USAMTS 1998 - 2017 en With Solutions

U S A Mathematical Talent Search
PROBLEMS
Round 1 - Year 10 - Academic Year 1998-99
1/1/10. Several pairs of positive integers (m,

n) satisfy the condition 19m + 90 + 8n = 1998 .
Of these, (100,
1) is the pair with the smallest value for n . Find the pair with the smallest
value for m .
r 1 1 1 r
2/1/10. Determine the smallest rational number - such that --- + ---- + --- - whenever k, m, and
s k m n s
1 1 1
n are positive integers that satisfy the inequality --- + ---- + --- < 1 .
k m n
3/1/10. It is possible to arrange eight of the nine numbers

2, 3, 4, 7, 10, 11, 12, 13, 15 1
in the vacant squares of the 3 by 4 array shown on the right so that the 9 5
arithmetic average of the numbers in each row and in each column is 14
the same integer. Exhibit such an arrangement, and specify which one
of the nine numbers must be left out when completing the array.
y
4/1/10. Show that it is possible to arrange seven distinct points in the
plane so that among any three of these seven points, two of the points
are a unit distance apart. (Your solution should include a carefully pre
pared sketch of the seven points, along with all segments that are of unit
length.)
R2 R1
5/1/10. The gure on the right shows the ellipse
2 2
( x 19 ) ( y 98 )
---------------------- + ---------------------- = 1998 .
19 98
Let R 1 , R 2 , R 3 , and R 4
denote those areas within the ellipse that are
R3 R4 x
in the rst, second, third, and fourth quadrants, respectively. Determine
the value of R 1 R 2 + R 3 R
4
********************
Complete, well-written solutions to at least two of the problems above, accompanied by a com-
pleted Entry Form and Cover Sheet, should be sent to the following address and postmarked no
later than September 12, 1998.
USA Mathematical Talent Search
COMAP Inc., Suite 210
57 Bedford Street
Lexington, MA 02173
PROBLEMS
1/2/10. Determine the unique pair of real numbers ( x, y) that satisfy the equation
2 2
(4 x + 6 x + 4)(4 y 12 y + 25 ) = 28 .
2/2/10. Prove that there are innitely many ordered triples of positive integers (a, b, c) such
2 2 2 2 2 2
that the greatest common divisor of a, b, and c is 1, and the sum a b + b c + c a is the
square of an integer.
3/2/10. Nine cards can be numbered using positive half-integers (1/2, 1, 3/2, 2, 5/2, ... ) so that
the sum of the numbers on a randomly chosen pair of cards gives an integer from 2 to 12 with
the same frequency of occurrence as rolling that sum on two standard dice. What are the num
bers on the nine cards and how often does each number appear on the cards?
4/2/10. As shown on the gure, square PQRS is
inscribed in right triangle ABC, whose right C
angle is at C, so that S and P are on sides BC and
CA, respectively, while Q and R are on side AB. P S
Prove that AB 3QR and determine when
equality occurs.
A Q R B
5/2/10. In the gure on the right, ABCD is a con
vex quadrilateral, K, L, M, and N are the mid
points of its sides, and PQRS is the quadrilateral K C
D
formed by the intersections of AK, BL, CM, and
DN. Determine the area of quadrilateral PQRS
S
if the area of quadrilateral ABCD is 3000, and
the areas of quadrilaterals AMQP and CKSR are R
513 and 388, respectively. L N
P
Q
********************
A B
M
Complete, well-written solutions to at least two of the problems above, accompanied by a com
pleted Cover Sheet, should be sent to the following address and postmarked no later than
November 14, 1998.
57 Bedford Street
Lexington, MA 02173
PROBLEMS
1/3/10. Determine the leftmost three digits of the number
1 2 3 999 1000
1 + 2 + 3 + + 999 + 1000 .
2/3/10. There are innitely many ordered pairs ( m, n ) of positive integers for which the sum
m + (m + 1) + (m + 2) + + (n 1) + n
is equal to the product mn . The four pairs with the smallest values of m are (1, 1), (3, 6),
(15, 35), and (85, 204). Find three more (m, n) pairs.
3/3/10. The integers from 1 to 9 can be arranged into a 3 3 array (as shown
A B C
on the right) so that the sum of the numbers in every row, column, and diago
nal is a multiple of 9.
D E F
(a.) Prove that the number in the center of the array must be a multiple of 3.
G H I
(b.) Give an example of such an array with 6 in the center.
6 6 6 6
4/3/10. Prove that if 0 < x < 2 , then sec x + csc x + ( sec x)( csc x) 80 .
A
5/3/10. In the gure on the right, O is the cen-
ter of the circle, OK and OA are perpendicu
lar to one another, M is the midpoint of OK,
BN is parallel to OK, and
AMN = NMO . Determine the measure
B N
of ABN in degrees.
K O
M
********************
Complete, well-written solutions to at least two of the problems above, accompanied by a com-
pleted Cover Sheet, should be sent to the following address and postmarked no later than Janu
ary 9, 1999. Each participant is expected to develop solutions without help from others.
57 Bedford Street
Lexington, MA 02173
PROBLEMS
1/4/10. Exhibit a 13-digit integer N that is an integer multiple of 213 and whose digits consist of
only 8s and 9s.
2/4/10. For a nonzero integer i, the exponent of 2 in the prime factorization of i is called ord2(i).
For example, ord2(9) = 0 since 9 is odd, and ord2(28) = 2 since 28 = 22 x 7. The numbers
3n - 1 for n = 1, 2, 3,... are all even, so ord2(3n - 1) > 0 for n > 0.
a) For which positive integers n is ord2(3n - 1) = 1?
b) For which positive integers n is ord2(3n - 1) = 2?
c) For which positive integers n is ord2(3n - 1) = 3?
Prove your answers.
1
3/4/10. Let f be a polynomial of degree 98, such that f ( k ) = --- for k = 1, 2, 3, ..., 99. Deter-
k
mine f(100).
4/4/10. Let A consist of 16 elements of the set {1, 2, 3., 106} , so that no two elements of A
differ by 6, 9, 12, 15, 18, or 21. Prove that two elements of A must differ by 3.
5/4/10. In ABC , let D, E, and F be the midpoints of C

the sides of the triangle, and let P, Q, and R be the
midpoints of the corresponding medians, AD , BE ,
and CF , respectively, as shown in the gure at the
E R D
right. Prove that the value of
2 2 2 2 2 2
AQ + AR + BP + BR + CP + CQ
P
-----------------------------------------------------------------------------------------------
2 2 2
- Q
AB + BC + C A
does not depend on the shape of ABC and nd
that value. A F B
*****************
pleted Cover Sheet, should be sent to the following address and postmarked no later than
March 13, 1999. Each participant is expected to develop solutions without help from others.
57 Bedford Street
Lexington, MA 02173
PROBLEMS
1/1/11. The digits of the three-digit integers a, b, and c are the nine nonzero digits 1, 2, 3, ..., 9,
each of them appearing exactly once. Given that the ratio a:b:c is 1:3:5, determine a, b, and c.
2/1/11. Let N = 111...1222...2, where there are 1999 digits of 1 followed by 1999 digits of 2.
Express N as the product of four integers, each of them greater than 1.
3/1/11. Triangle ABC has angle A measuring 30 , angle B measuring 60 , and angle C measur
ing 90 . Show four different ways to divide triangle ABC into four triangles, each similar to
triangle ABC, but with one quarter of the area. Prove that the angles and sizes of the smaller
triangles are correct.
4/1/11. There are 8436 steel balls, each with radius 1 centimeter, stacked in a tetrahedral pile,
with one ball on top, 3 balls in the second layer, 6 in the third layer, 10 in the fourth, and so on.
Determine the height of the pile in centimeters.
5/1/11. In a convex pentagon ABCDE the sides have lengths 1, 2, 3, 4, and 5, though not neces
sarily in that order. Let F, G, H, and I be the midpoints of the sides AB, BC, CD, and DE,
respectively. Let X be the midpoint of segment FH, and Y be the midpoint of segment GI.
The length of segment XY is an integer. Find all possible values for the length of side AE.
*****************
pleted Cover Sheet and a completed Entry Form, should be sent to the following address and post
marked no later than September 13, 1999. Each participant is expected to develop solutions
without help from others.
57 Bedford Street
Lexington, MA 02173
PROBLEMS
1/2/11. The number N consists of 1999 digits such that if each pair of consecutive digits in N
were viewed as a two-digit number, then that number would either be a multiple of 17 or a
multiple of 23. The sum of the digits of N is 9599. Determine the rightmost ten digits of N.
2/2/11. Let C be the set of non-negative integers which can be B

expressed as 1999s + 2000t, where s and t are also non-negative
integers.
(a) Show that 3,994,001 is not in C.
(b) Show that if 0 n

3,994,001 and n is an integer not in C,
then 3,994,001 - n is in C.
3/2/11. The gure on the right shows the map of Squareville, where
each city block is of the same length. Two friends, Alexandra and A R
Brianna, live at the corners marked by A and B, respectively. They
start walking toward each others house, leaving at the same time,
walking with the same speed, and independently choosing a path to
the others house with uniform distribution out of all possible mini-
mum-distance paths [that is, all minimum-distance paths are S
equally likely]. What is the probability they will meet?
P Q
4/2/11. In PQR , PQ = 8, QR = 13, and RP = 15. Prove that there is
a point S on line segment PR , but not at its endpoints, such that PS and QS are also integers.
5/2/11. In ABC , AC > BC, CM is the median, C
and CH is the altitude emanating from C, as
shown in the gure on the right. Determine the
measure of MCH if ACM and BCH
each have measure 17 .

A M H B
*****************
pleted Cover Sheet should be sent to the following address and postmarked no later than
November 15, 1999. Each participant is expected to develop solutions without help from others.
57 Bedford Street
Lexington, MA 02173
PROBLEMS
1/3/11. We dene the repetition number of a positive integer n to be the number of distinct digits
of n when written in base 10. Prove that each positive integer has a multiple which has a rep
etition number less than or equal to 2.
2/3/11. Let a be a positive real number, n a positive integer, and dene the power tower an
recursively with a1 =a, and a(i+1) = a(a) for i = 1, 2, 3, . For example, we have
4
(4 ) 256
43 = 4 = 4 , a number which has 155 digits. For each positive integer k, let xk
denote the unique positive real number solution of the equation xk = 10(k+1). Which is
larger: x42 or x43?
3/3/11. Suppose that the 32 computers in a certain network are numbered with the 5-bit integers
00000, 00001, 00010, ..., 11111 (bit is short for binary digit). Suppose that there is a one-way
connection from computer A to computer B if and only if A and B share four of their bits with
the remaining bit being 0 at A and 1 at B. (For example, 10101 can send messages to 11101
and to 10111.) We say that a computer is at level k in the network if it has exactly k 1s in its
label (k = 0, 1, 2, ..., 5). Suppose further that we know that 12 computers, three at each of the
levels 1, 2, 3, and 4, are malfunctioning, but we do not know which ones. Can we still be sure
that we can send a message from 00000 to 11111?
4/3/11. We say a triangle in the coordinate plane is integral if its three vertices have integer
coordinates and if its three sides have integer lengths.
(a) Find an integral triangle with perimeter of 42.
(b) Is there an integral triangle with perimeter of 43?
5/3/11. We say that a nite set of points is well scattered on the surface of a sphere if every open
hemisphere (half the surface of the sphere without its boundary) contains at least one of the
points. The set { (1,0,0), (0,1,0), (0,0,1) } is not well scattered on the unit sphere (the sphere
of radius 1 centered at the origin), but if you add the correct point P it becomes well scattered.
Find, with proof, all possible points P that would make the set well scattered.
*****************
pleted Cover Sheet should be sent to the following address and postmarked no later than Janu
ary 10, 2000. Each participant is expected to develop solutions without help from others.

57 Bedford Street
Lexington, MA 02173
PROBLEMS
1/4/11. Determine the unique 9-digit integer M that has the following properties: (1) its digits
are all distinct and nonzero; and (2) for every positive integer m = 2, 3, 4, ..., 9, the integer
formed by the leftmost m digits of M is divisible by m.
2/4/11. The Fibonacci numbers are dened by F1 = F2 = 1 and Fn = Fn-1 + Fn-2 for n > 2. It is
well-known that the sum of any 10 consecutive Fibonacci numbers is divisible by 11. Deter
mine the smallest integer N so that the sum of any N consecutive Fibonacci numbers is divisi
ble by 12.
3/4/11. Determine the value of

1 1 1 1 1 1
S = 1 + ----2- + ----2- + 1 + ----2- + ----2- + + 1 + -------------2- + -------------2
1 2 2 3 1999 2000
4/4/11. We will say that an octagon is integral if it is y

equiangular, its vertices are lattice points (i.e., points
with integer coordinates), and its area is an integer.
For example, the gure on the right shows an integral
octagon of area 21. Determine, with proof, the
smallest positive integer K so that for every positive
integer k K , there is an integral octagon of area k.
2
5/4/11. (Revised 2-4-2000) Let P be a point interior to
1
square ABCD so that PA = a, PB = b, PC = c, and c2
= a2 + 2b2. Given only the lengths a, b, and c, and 1 2
x
using only a compass and straightedge, construct a
square congruent to square ABCD.
*****************
pleted Cover Sheet should be sent to the following address and postmarked no later than March
13, 2000. Each participant is expected to develop solutions without help from others.

57 Bedford Street
Lexington, MA 02173
PROBLEMS
1/1/12. Determine the smallest ve-digit positive integer N such that 2N is also a ve-digit inte
ger and all ten digits from 0 to 9 are found in N and 2N.
24
2
2/1/12. It was recently shown that 2 + 1 is not a prime number. Find the four rightmost dig
its of this number.
3/1/12. Determine the integers a, b, c, d, and e for which

2 3 2 5
( x + ax + b)( x + cx + dx + e) = x 9 x 27 .
4/1/12. A sequence of real numbers s 0, s 1, s 2, has the property that

s i s j = s i + j + s i j for all nonnegative integers i and j with i j ,
s i = s i + 12 for all nonnegative integers i, and
s0 > s1 > s2 > 0 . 5
Find the three numbers s0, s1, and s2.
5 5
6 6
5/1/12. In the octahedron shown on the right, the base
and top faces are equilateral triangles with sides
6 6
measuring 9 and 5 units, and the lateral edges are all
of length 6 units. Determine the height of the octa 6 6
hedron; i.e., the distance between the base and the
top face.
9 9
9
*****************
pleted Cover Sheet and a completed Entry Form (both available on the web site http://
www.nsa.gov/programs/mepp/usamts.html), should be sent to the following address and post
marked no later than September 11, 2000. Each participant is expected to develop solutions
without help from others.

57 Bedford Street
Lexington, MA 02173
PROBLEMS
1/2/12. A well-known test for divisibility by 19 is
as follows: Remove the last digit of the number,
6 7 9 4 4 4 4 9 7 6
add twice that digit to the truncated number, and
keep repeating this procedure until a number less 8 1 2
than 20 is obtained. Then, the original number is 6 8 0 2 4 5 0 9
divisible by 19 if and only if the nal number is 4 1 8
19. The method is exemplied on the right; it is 6 8 4
4 6 8
easy to check that indeed 67944 is divisible by
8 1 6
19, while 44976 is not. 7 6 6 2
1 2 4
Find and prove a similar test for divisibility by
29. 1 9 1 0
1492! 1492!
2/2/12. Compute 1776 (mod 2000) ; i.e., the remainder when 1776 is divided by
2000. (As usual, the exclamation point denotes factorial.)
3/2/12. Given the arithmetic progression of integers

308, 973, 1638, 2303, 2968, 3633, 4298 ,
determine the unique geometric progression of integers,
b 1, b 2, b 3, b 4, b 5
, b 6
so that
308 < b 1 < 973 < b 2 < 1638 < b 3 < 2303 < b 4 < 2968 < b 5 < 3633 < b 6 < 4298 .
4/2/12. Prove that every polyhedron has two vertices at which the same number of edges meet.
[Editors comment: Proofs are difcult to grade. Please try to keep your solution short and
well written.]
5/2/12. In ABC , segments PQ, RS, and TU are parallel to

C
sides AB, BC, and CA, respectively, and intersect at the
points X, Y, and Z, as shown in the gure on the right.

S T
Determine the area of ABC if each of the segments PQ,

Z
RS, and TU bisects (halves) the area of ABC , and if the
P Y Q
X
area of XYZ is one unit. Your answer should be in the
form a + b 2 , where a and b are positive integers. A U R B
*****************
pleted Cover Sheet (available on the web site http://www.nsa.gov/programs/mepp/usamts.html),
should be sent to the following address and postmarked no later than November 13, 2000.
Also include an Entry Form if you did not participate in Round 1. Each participant is expected
to develop solutions without help from others.

57 Bedford Street
Lexington, MA 02420-4350
PROBLEMS
1/3/12. Find the smallest positive integer with the property that it has divisors ending with every
decimal digit; i.e., divisors ending in 0, 1, 2, ..., 9.
2/3/12. Assume that the irreducible fractions between 0 and 1, with denominators at most 99,
17
are listed in ascending order. Determine which two fractions are adjacent to ------ in this listing.
76
5 2 2
3/3/12. Let p ( x ) = x + x + 1 have roots r 1, r 2, r 3, r 4, r 5 . Let q ( x ) = x 2 . Determine
the product q ( r 1 )q ( r 2 )q ( r 3 )q ( r 4 )q ( r 5 ) .

4/3/12. Assume that each member of the sequence i i = 1 is either a + or a sign.
Determine the appropriate sequence of + and signs so that
2 = 6 1 6 2 6 3 .
Also determine what sequence of signs is necessary if the sixes in the nested roots are
replaced by sevens. List all integers that work in the place of sixes and the sequences of signs
that are needed with them.
5/3/12. Three isosceles right triangles are M

erected from the larger side of a rectan
gle into the interior of the rectangle, as
shown on the right, where M is the mid
point of that side. Five circles are
inscribed tangent to some of the sides
and to one another as shown. One of the
circles touches the vertex of the largest
triangle.
Find the ratios among the radii of the
ve circles.
****************
pleted Cover Sheet (available on the web site http://www.nsa.gov/programs/mepp/usamts.html),
should be sent to the following address and postmarked no later than January 8, 2001. Also
include an Entry Form if this is your rst submission this academic year. Each participant is
expected to develop solutions without help from others.

57 Bedford Street
PROBLEMS
1/4/12. Determine all positive integers with the property that they are one more than the sum of
the squares of their digits in base 10.
n n n n
2/4/12. Prove that if n is an odd positive integer, then N = 2269 + 1779 + 1730 1776 is
an integer multiple of 2001.
3/4/12. The gure on the right. can be divided into two congruent halves that are
related to each other by a glide reection, as shown below it. A glide reection
reects a gure about a line, but also moves the reected gure in a direction
parallel to that line. For a square-grid gure, the only lines of reection that

keep its reection on the grid are horizontal, vertical, 45 diagonal, and 135
diagonal. Of the two gures below, divide one gure into two congruent halves
related by a glide reection, and tell why the other gure cannot be divided like
that.
4/4/12. Let A and B be points on a circle which are not diametrically opposite, and let C be the
midpoint of the smaller arc between A and B. Let D, E, and F be the points determined by the
intersections of the tangent lines to the circle at A, B, and C. Prove that the area of DEF is
greater than half of the area of ABC .
5/4/12. Hexagon RSTUVW is constructed by starting with

U
a right triangle of legs measuring p and q, constructing
squares outwardly on the sides of this triangle, and then
connecting the outer vertices of the squares, as shown in
the gure on the right. T

V
Given that p and q are integers with p > q , and that the q
area of RSTUVW is 1922, determine p and q. W p
****************
Complete, well-written solutions to at least two of the prob
lems above, accompanied by a completed Cover Sheet R S
(available on the web site http://www.nsa.gov/programs/mepp/usamts.html), should be sent to the
following address and postmarked no later than March 12, 2001. Also include an Entry Form
if you have not submitted one for this academic year. Each participant is expected to develop
solutions without help from others.
57 Bedford Street
PROBLEMS
1/1/13. Determine the unique positive two-digit integers m and n for which the approximation
m m
---- 0.2328767 is accurate to seven decimals, i.e., 0.2328767 ---- < 0.2328768 .
n n
2/1/13. It is well known that there are innitely many triples of integers (a, b, c) whose greatest
2 2 2
common divisor is 1 and which satisfy the equation a + b = c . Prove that there are also
innitely many triples of integers (r, s, t) whose greatest common divisor is 1 and which sat
2 2 2
isfy the equation ( rs ) + ( st ) = ( rt ) .
cos3 x 1 sin3x
3/1/13. Suppose --------------- = --- for some angle x, 0 x --- . Determine -------------- for the same x.
cos x 3 2 sin x
4/1/13. The projective plane of order three consists of 13 points and 13 lines. These lines
are not Euclidean straight lines; instead they are sets of four points with the properties that
each pair of lines has exactly one point in common, and each pair of points has exactly one
line that contains both points. Suppose the points are labeled 1 through 13, and six of the lines
are A = {1, 2, 4, 8} , B = {1, 3, 5, 9} , C = {2, 3, 6, 10} , D = {4, 5, 10, 11} ,
E = {4, 6, 9, 12} , and F = {5, 6, 8, 13} . What is the line that contains 7 and 8?
5/1/13. In PQR , QR < PR < PQ so that the

Q T
exterior angle bisector through P intersects ray
QR at point S, and the exterior angle bisector P
R
at R intersects ray PQ at point T, as shown on
the right. Given that PR = PS = RT , deter
mine, with proof, the measure of PRQ .
****************
Complete, well-written solutions to at least two of the problems above, accompanied by a Cover
Sheet and Entry Form should be sent to the address listed on the USAMTS web site
http://www.nsa.gov/programs/mepp/usamts.html
and postmarked no later than 7 October 2001. Each participant is expected to develop solu
tions without help from others.
PROBLEMS
1/2/13. How many positive ve-digit integers are there consisting of the digits 1, 2, 3, 4, 5, 6, 7,
8, 9, in which one digit appears once and two digits appear twice? For example, 41174 is one
such number, while 75355 is not.
2/2/13. Determine, with proof, the positive integer whose square is exactly equal to the number
2001
(4i 2 )
3
1+ .
i=1
[A computer solution will be worth at most 1 point.]
3/2/13. Factor the expression

2 2 2 2
30(a + b + c + d ) + 68ab 75ac 156ad 61bc 100bd + 87cd .
4/2/13. Let X = ( x 1, x 2, x 3, x 4, x 5, x 6, x 7, x 8, x 9 ) be a 9-long vector of integers. Determine X

if the following seven vectors were all obtained from X by deleting three of its components:
Y 1 = (0, 0, 0, 1, 0, 1) , Y 2 = (0, 0, 1, 1, 1, 0) , Y 3 = (0, 1, 0, 1, 0, 1) , Y 4 = (1, 0, 0, 0, 1, 1),
Y 5 = (1, 0, 1, 1, 1, 1) , Y 6 = (1, 1, 1, 1, 0, 1) , Y 7 = (1, 1, 0, 1, 1, 0).
5/2/13. Let R and S be points on the sides BC and AC , respectively, of ABC , and let P be the
intersection of AR and BS . Determine the area of ABC if the areas of APS , APB , and
BPR are 5, 6, and 7, respectively.
****************
Sheet, should be sent to the address listed on the USAMTS web site
and postmarked no later than 25 November 2001. Each participant is expected to develop solu
PROBLEMS
1/3/13. We will say that a rearrangement of the letters of a word has no xed letters if, when the
rearrangement is placed directly below the word, no column has the same letter repeated. For
instance, the blocks of letters below show that E S A R E T is a rearrangement with no xed
letters of T E R E S A , but R E A S T E is not.
T E R E S A T E R E S A
E S A R E T R E A S T E
How many distinguishable rearrangements with no xed letters does T E R E S A have? (The
two Es are considered identical.)
2/3/13. Without computer assistance, nd ve different sets of three positive integers {k, m, n}
1 1 1 19
such that k < m < n and --- + ---- + --- = ------ .
k m n 84
[To think about, but not a part of the problem: How many solutions are there?]
n n1
3/3/13. Suppose p ( x ) = x + a n 1 x + + a 1 x + a 0 is a monic polynomial with integer
n 2
coefcients. [Here monic polynomial just means the coefcient of x is one.] If ( p ( x )) is a
polynomial all of whose coefcients are non-negative, is it necessarily true that all the coef
cients of p ( x ) must be non-negative? Justify your answer.
4/3/13. As shown in the gure on

C = (30s, 30t)
the right, in ACF , B is the
midpoint of AC , D and E divide D
side CF into three equal parts, E
while G, H, and I divide side
F = (0, 0)
A = (60r, 0)
FA into four equal parts. G H I
Seventeen segments are
drawn to connect these six
points to one another and to the opposite vertices of the triangle. Determine the points interior
to ACF at which three or more of these line segments intersect one another.
To make grading easier, we have embedded the triangle into the rst quadrant with point F
at the origin, point C at (30s, 30t) , and point A at (60r, 0) , where r, s, and t are arbitrary
positive real numbers. Please use this notation in your solutions.
5/3/13. Two perpendicular planes intersect a sphere in two circles. These circles intersect in
two points, spaced 14 units apart, measured along the straight line connecting them. If the
radii of the circles are 18 and 25 units, what is the radius of the sphere?
****************
********************************
Sheet, should be mailed to
DDM Co.
279 East Central Street, Suite 246
Franklin, MA 02038-1317
and postmarked no later than 6 January 2002. Each participant is expected to develop solu
PROBLEMS
1/4/13. In a strange language there are only two letters, a and b, and it is postulated that the let
ter a is a word. Furthermore, all additional words are formed according to the following rules:
A. Given any word, a new word can be formed from it by adding a b at the righthand end.
B. If in any word a sequence aaa appears, a new word can be formed by replacing the aaa
by the letter b.
C. If in any word the sequence bbb appears, a new word can be formed by omitting bbb.
D. Given any word, a new word can be formed by writing down the sequence that consti
tutes the given word twice.
For example, by (D), aa is a word, and by (D) again, aaaa is a word. Hence by (B) ba is a
word, and by (A) bab is also a word. Again, by (A), babb is a word, and so by (D), babbbabb
is also a word. Finally, by (C) we nd that baabb is a word.
Prove that in this language baabaabaa is not a word.
2/4/13. Let f ( x ) = x x x x for all positive real numbers x, where y denotes the
greatest integer less than or equal to y.
(1) Determine x so that f ( x ) = 2001 .
(2) Prove that f ( x ) = 2002 has no solution.
3/4/13. Let f be a function dened on the set of all integers, and assume that it satises the fol
lowing properties:
A. f ( 0 ) 0 ;
B. f ( 1 ) = 3 ; and
C. f ( x ) f ( y ) = f ( x + y) + f ( x y) for all integers x and y.
Determine f ( 7 ) .
4/4/13. A certain company has a faulty telephone system that sometimes transposes a pair of
adjacent digits when someone dials a three-digit extension. Hence a call to x318 would ring
at either x318 , x138 , or x381 , while a call received at x044 would be intended for either
x404 or x044 . Rather than replace the system, the company is adding a computer to deduce
which dialed extensions are in error and revert those numbers to their correct form. They have
to leave out several possible extensions for this to work. What is the greatest number of three-
digit extensions the company can assign under this plan?
5/4/13. Determine the smallest number of squares into which one can dissect a 11 13 rectan
gle, and exhibit such a dissection. The squares need not be of different sizes, their bases
should be integers, and they should not overlap.
********************************
DDM Co.
and postmarked no later than 17 March 2002. Each participant is expected to develop solu
PROBLEMS
1/1/14. Some unit cubes are stacked atop a at 4 by 4 square. The South View East View
gures show views of the stacks from two different sides. Find
the maximum and minimum number of cubes that could be in the
stacks. Also give top views of a maximum arrangement and a
minimum arrangement with each stack marked with its height.
2/1/14. Find four distinct positive integers, a , b , c , and d , such that each of the four sums
a + b + c , a + b + d , a + c + d , and b + c + d
is the square of an integer. Show that
innitely many quadruples (a, b, c, d) with this property can be created.
3/1/14. For a set of points in a plane, we construct the perpendicular bisectors of the line
segments connecting every pair of those points and we count the number of points in which
these perpendicular bisectors intersect each other. If we start with twelve points, the maximum
possible number of intersection points is 1705. What is the maximum possible number of
intersection points if we start with thirteen points?
4/1/14. A transposition of a vector is created by switching exactly two entries of the vector. For
example, (1, 5, 3, 4, 2, 6, 7) is a transposition of (1, 2, 3, 4, 5, 6, 7) . Find the vector X if
S = (0, 0, 1, 1, 0, 1, 1) , T = (0, 0, 1, 1, 1, 1, 0) , U = (1, 0, 1, 0, 1, 1, 0) , and
V = (1, 1, 0, 1, 0, 1, 0) are all transpositions of X. Describe your method for nding X.
5/1/14. As illustrated below, we can dissect every triangle ABC into four pieces so that piece 1
is a triangle similar to the original triangle, while the other three pieces can be assembled into
a triangle also similar to the original triangle. Determine the ratios of the sizes of the three
triangles and verify that the construction works.
B
D
3
G 2 F Piece 2
H rotated
4 1 4
1 180 3
2
A C
E
Complete, well-written solutions, accompanied by a Cover Sheet and an Entry Form, should be
mailed to the address on the USAMTS web site
and postmarked no later than 6 October 2002. Each participant is expected to develop solutions
without help from other people.
PROBLEMS
1/2/14. Each member of the sequence 112002, 11210, 1121, 117, 46, 34, is obtained by add
ing ve times the rightmost digit to the number formed by omitting that digit. Determine the
billionth (109th) member of this sequence.
2/2/14. The integer 72 is the rst of three consecutive integers 72, 73, and 74, that can each be
expressed as the sum of the squares of two positive integers. The integers 72, 288, and 800 are
the rst three members of an innite increasing sequence of integers with the above property.
Find a function that generates the sequence and give the next three members.
3/2/14. An integer lattice point in the Cartesian plane is a point ( x, y) where x and y are both
integers. Suppose nine integer lattice points are chosen such that no three of them lie on the
same line. Out of all 36 possible line segments between pairs of those nine points, some line
segments may contain integer lattice points besides the original nine points. What is the mini
mum number of line segments that must contain an integer lattice point besides the original
nine points? Prove your answer.
4/2/14. Let f ( n ) be the number of ones that occur in the decimal representations of all the num
bers from 1 to n . For example, this gives f ( 8 ) = 1 , f ( 9 ) = 1 , f ( 10 ) = 2 , f ( 11 ) = 4 ,
100
and f ( 12 ) = 5 . Determine the value of f (10 ).
5/2/14. For an isosceles triangle ABC where AB = AC , it is possible A

to construct, using only compass and straightedge (see next page),
an isosceles triangle PQR where PQ = PR such that triangle
PQR is similar to triangle ABC , point P is in the interior of line
segment AC , point Q is in the interior of line segment AB , and P
Q
point R is in the interior of line segment BC . Describe one method
of performing such a construction. Your method should work on
every isosceles triangle ABC , except that you may choose an upper B R C
limit or lower limit on the size of angle BAC .
National Conference Services, Inc.
6440-C Dobbin Road
Columbia, MD 21045-4770
and postmarked no later than November 24, 2002. Each participant is expected to develop
solutions without help from others. For the cover sheet and other details, see the USAMTS web
site http://www.nsa.gov/programs/mepp/usamts.html.
Compass and Straightedge Constructions
Constructing geometric gures by compass and straightedge actually uses several idealized tools:
a pen to draw with, paper to draw on, a compass to draw circles, a straightedge to draw lines, and
a divider to copy distances. Classic geometers preferred these tools, as opposed to other tools such
as the pins and string for drawing ellipses, because working with denite shapes and denite dis
tances made proofs easier. Since some schools might not cover compass and straightedge con
structions in their geometry curriculum, some USAMTS participants might benet from a brief
summary of some basic compass-and-straightedge constructions.
The following list covers some basic constructions. Combine them for more elaborate construc
tions. Points, lines, and circles can be used only if they are already drawn. A known distance is
any distance between two points. A known angle is any angle between intersecting lines.
1. Random Point. Draw a point at a random location in the plane, on a line, on a line segment, on
the circumference of a circle, or in the interior of any geometric gure.
2. Line. Draw a line through any two points. Draw a line through any point parallel to another
line. Draw a line through any point perpendicular to another line. Extend a line segment into a
line.
3. Circle. Draw a circle whose center is a point and whose radius is a known distance. Draw the
circumscribed and inscribed circles of any triangle. Find the center point of a circle. Extend an
arc into a circle.
4. Intersection. Find all points where two lines, two circles, or a line and a circle intersect.
5. Distance. Given a point on a line, draw another point on that line at a known distance from the
rst point in either direction. Can do the same with a line segment instead of a line, though it
might be extend the line segment. Create any distance that is a rational or square root multiple
of another known distance.
6. Line Segment. Draw a line segment between two points. Draw a copy of a line segment with
one endpoint at any point and the copy rotated by any known angle.
7. Angle. Draw a line through a point on another line such that the angle between those lines
copies any known angle. Draw an angle whose measure is equal to the sum or difference of
the measures of any two known angles.
8. Arc. Draw an arc on a circle, starting at any point on the circumference, whose angle measure
is any known angle.
9. Partitioning. Divide a line segment into any number of segments whose lengths are equal or
are proportional to any collection of known distances.
10. Angle Bisection. Draw a line through the intersection of two other lines that bisects the angle
between them. Trisection is not possible with compass and straightedge, though it is possible
to construct 30, 60, and 90 angles.
11. Tangent. Draw two lines through a point exterior to a circle that are tangent to the circle.
Draw a line through a point on the circumference of a circle that is tangent to the circle. Draw
the common tangent lines to two circles, provided that a circle is not contained inside the
other.
12. Projection. Copy any geometric gure build out of line segments and arcs, with its size
rescaled by any ratio between known distances, its orientation rotated by any known angle,
and one point of the copy at any desired point.
PROBLEMS
Round 3 - Year 14 - Academic Year 20022003
1/3/14. The integer n, between 10000 and 99999, is abcde when written in decimal notation. The
digit a is the remainder when n is divided by 2, the digit b is the remainder when n is divided
by 3, the digit c is the remainder when n is divided by 4, the digit d is the remainder when n is
divided by 5, and the digit e is the remainder when n is divided by 6. Find n.
2/3/14. Given positive integers p, u, and v such that u 2 + 2v 2 = p , determine, in terms of u and v,
integers m and n such that 3m 2 2mn + 3n 2 = 24 p . (It is known that if p is any prime num-
ber congruent to 1 or 3 modulo 8, then we can nd integers u and v such that u 2 + 2v 2 = p .)
m
3/3/14. Determine, with proof, the rational number ---- that equals
n
1 1 1 1
--------------------------- + --------------------------- + --------------------------- + + --------------------------------------------------------------------------------------------------- .
1 2+2 1 2 3+3 2 3 4+4 3 4012008 4012009 + 4012009 4012008
4/3/14. The vertices of a cube have coordinates ( 0, 0, 0 ) , ( 0, 0, 4 ) , ( 0, 4, 0 ) , ( 0, 4, 4 ) ,

( 4, 0, 0 ) , ( 4, 0, 4 ) , ( 4, 4, 0 ) , and ( 4, 4, 4 ) . A plane cuts the edges of this cube at the points
( 0, 2, 0 ) , ( 1, 0, 0 ) , ( 1, 4, 4 ) , and two other points. Find the coordinates of the other two
points.
5/3/14. A fudgeake is a planar fractal gure with 120 rota-
tional symmetry such that three identical fudgeakes in the
same orientation t together without gaps to form a larger
fudgeake with its orientation 30 clockwise of the smaller 1
fudgeakes orientation, as shown on the right. If the distance
between the centers of the original three fudgeakes is 1,
what is the area of one of those three fudgeakes? Justify
your answer.
Complete, well-written solutions to at least two of the problems above,

accompanied by a Cover Sheet, should be mailed to
6440-C Dobbin Road
and postmarked no later than January 5, 2003. Each participant is expected to develop solu-
tions without help from others. For the cover sheet and other details, see the USAMTS web site
http://www.nsa.gov/programs/mepp/usamts.html.
PROBLEMS
Round 4 - Year 14 - Academic Year 20022003
1/4/14. The sequence of letters TAGC is written in succession 55 times on a strip, as shown
below. The strip is to be cut into segments between letters, leaving strings of letters on each
segment, which we will call words. For example, a cut after the rst G, after the second T, and
after the second C would yield the words TAG, CT, and AGC. At most how many distinct
words could be found if the entire strip were cut? Justify your answer.
TAGCTAGCTAG ... CTAGC
2/4/14. We dene the number s as

10
1 1 1 1 1
s = ---------------
i1
- = --- + ------ + --------- + ------------ + = 0.12232424 .
9 99 999 9999
i=1
We can determine the nth digit right of the decimal point of s without summing the entire in
nite series because after summing the rst n terms of the series, the rest of the series sums to
less than 2 10 n + 1 . Determine the smallest prime number p for which the pth digit right of
the decimal point of s is greater than 2. Justify your answer.
3/4/14. Find the real-numbered solution to the equation below and demonstrate that it is unique.
36 9
------- + ------- = 42 9 x y
x y
4/4/14. Two overlapping triangles could divide the plane into up to eight regions, and three over
lapping triangles could divide the plane into up to twenty regions. Find, with proof, the maxi
mum number of regions into which six overlapping triangles could divide the plane. Describe
or draw an arrangement of six triangles that divides the plane into that many regions.
5/4/14. Prove that if the cross-section of a cube cut by a plane is a
pentagon, as shown in the gure on the right, then there are two E
adjacent sides of the pentagon such that the sum of the lengths A
of those two sides is greater than the sum of the lengths of the
other three sides. For ease of grading, please use the names of D
the points from the gure on the right in your solution.
Complete, well-written solutions to at least two of the problems
above, accompanied by a Cover Sheet, should be mailed to: C
USA Mathematical Talent Search B
6440-C Dobbin Road
and postmarked no later than March 16, 2003. Each participant is expected to develop solu
tions without help from others. For the cover sheet and other details, see the USAMTS web site:
PROBLEMS
Round 1 Year 15 Academic Year 20032004
1/1/15. Let f (x) = x + 10x + 100x + 1000x for all real numbers x, where for any
real number r, r means the greatest integer less than or equal to r. For example,
f () = 3 + 31 + 314 + 3141 = 3489. Find, with proof, a positive integer n less than
2003 such that f (x) = n has no solution for x, but f (y) = n + 11 and f (z) = n + 111
have solutions for y and z.
2/1/15. Find all primes p for which 6p + 1 is the fth power of an integer. Prove that you
found all of them.
3/1/15. We attempted to arrange the integers 1, 2, 3, ..., 12 around a circle so that all
sums of pairs of adjacent integers are either
(a) all perfect squares, or

(b) all triangular numbers, which are numbers of the form n(n + 1)/2.
One case out of (a) and (b) succeeded. Which one? For the successful case, show a
valid arrangement. For the unsuccessful case, explain why it is impossible.
4/1/15. Two players play a game. Each player, in turn, has to name a positive integer that
is less than the previous number but at least half the previous number. The player
who names the number 1 loses. If the rst player starts by naming 2003 and after that
both players play with the the best strategy, who wins? Describe the strategy and
prove it works.
5/1/15. A rectangular piece of paper, ABCD, is folded A B A C B

along line segment EF , where point E is on side AD
S

S
and point F is on side BC, so that C ends up at the F D
S

SF

midpoint of side AB. Determine the length of EF if E
E
S
the length of AB is 240 and the length of BC is 288.
D
C
Complete, wellwritten solutions to at least two of the problems above, accompanied by

an entry form and cover sheet, should be mailed to the administrative address listed on
the USAMTS web site and postmarked no later than 5 October 2003. Each participant
is expected to develop solutions without help from others. For the cover sheet and other
details, see the USAMTS web site: http://www.nsa.gov/programs/mepp/usamts.html.
PROBLEMS
1/2/15. The faces of 27 unit cubes are painted red, white, and blue in such a manner that
we can assemble them into three dierent congurations: a red 3 3 3 cube, a white
3 3 3 cube, and a blue 3 3 3 cube. Determine, with proof, the number of unit
cubes on whose faces all three colors appear.
2/2/15. For any positive integer n, let s(n) denote the sum of the digits of n in base 10.
Find, with proof, the largest n for which n = 7s(n).
3/2/15. How many circles in the plane contain at least three of the nine points q q q
(0, 0), (0, 1), (0, 2), (1, 0), (1, 1), (1, 2), (2, 0), (2, 1), (2, 2)? Rigorously q q q
verify that no circle was skipped or counted more than once in the result.
q q q
4/2/15. In how many ways can one choose three angle sizes, , , and ,
with from the set of integral degrees, 1 , 2 , 3 , . . ., 178 , SS

such that those angle sizes correspond to the angles of a nondegenerate S
S
triangle? How many of the resulting triangles are acute, right, and

obtuse, respectively?
5/2/15. Clearly draw or describe a convex polyhedron that has exactly three pentagons
among its faces and the fewest edges possible. Prove that the number of edges is a
minimum.
Complete, wellwritten solutions to at least two of the problems above, accompanied by a

Cover Sheet, should be mailed to:
DDM Co
279 East Central St Suite 246
Franklin, MA 020381317
and postmarked by Sunday, 23 November 2003. Each participant is expected to develop

solutions without help from others. For the cover sheet and other details, see the USAMTS
web site: http://www.nsa.gov/programs/mepp/usamts.html.
PROBLEMS
1/3/15. Find, with proof, all pairs of twodigit positive integers ab and cd such that all the
digits a, b, c, and d are dierent from one another and (ab)(cd) = (dc)(ba).
2/3/15. Find the smallest positive integer n such that the product (2004n + 1)(2008n + 1)
is a perfect square. Prove that n is as small as possible.
3/3/15. Pebbles are put on the vertices of a combinatorial graph. For a vertex with two or
more pebbles, a pebbling step at that vertex removes one pebble at the vertex from the
graph entirely and moves another pebble at that vertex to a chosen adjacent vertex.
The pebbling number of a graph is the smallest number t t t
such that no matter how t pebbles are distributed on the J
graph, the distribution would have the property that for

t
J
J
every empty vertex a series of pebbling steps could move
J
a pebble to that one vertex. For example, the pebbling t t
Jt t
number of the graph formed from the vertices and edges of J
J
J
J
a hexagon is eight. Find, with proof, the pebbling number Jt

Jt

of the graph illustrated on the right.
4/3/15. An innite sequence of quadruples begins with the ve quadruples (1, 3, 8, 120),
(2, 4, 12, 420), (3, 5, 16, 1008), (4, 6, 20, 1980), (5, 7, 24, 3432). Each quadruple (a, b, c, d)
in this sequence has the property that the six numbers ab + 1, ac + 1, bc + 1, ad + 1,
bd + 1, and cd + 1 are all perfect squares. Derive a formula for the nth quadruple in
the sequence and demonstrate that the property holds for every quadruple generated
by the formula.
5/3/15. In triangle ABC the lengths of the sides of the triangle A
opposite to the vertices A, B, and C are known as a, b, and
B
c, respectively. Prove there exists a constant k such that if N
B
B
the medians emanating from A and B are perpendicular to HH B
one another, then a2 + b2 = kc2 . Also nd the value of k.
PHH B
HHB
C M B
Complete, wellwritten solutions to at least two of the problems above, accompanied by a

DDM Co.
Franklin, MA 020381317
and postmarked by Sunday, 4 January 2004. Each participant is expected to develop solu
tions without help from others. For the cover sheet and other details, see the USAMTS web
site: http://www.nsa.gov/programs/mepp/usamts.html.
PROBLEMS
1/4/15. Find, with proof, the smallest positive integer n for which the sum of the digits of
29n is as small as possible.
2/4/15. For four integer values of n greater than six, there exist right triangles whose side
lengths are integers equivalent to 4, 5, and 6 modulo n, in some order. Find those
values. Prove that at most four such values exist. Also, for at least one of those values
of n, provide an example of such a triangle.
3/4/15. Find a nonzero polynomial f (w, x, y, z) in the four indeterminates w, x, y, and z of
minimum degree such that switching any two indeterminates in the polynomial gives
the same polynomial except that its sign is reversed. For example, f (z, x, y, w) =
f (w, x, y, z). Prove that the degree of the polynomial is as small as possible.
4/4/15. For each nonnegative integer n define the function fn (x) by
2 4
fn (x) = sinn (x) + sinn (x + ) + sinn (x + )
3 3
for all real numbers x, where the sine functions use radians. The functions fn (x) can
be also expressed as polynomials in sin(3x) with rational coefficients. For example,
3 3
f0 (x) = 3, f1 (x) = 0, f2 (x) = , f3 (x) = sin(3x),
2 4
9 15 27 3
f4 (x) = , f5 (x) = sin(3x), f6 (x) = + sin2 (3x),
8 16 32 16
for all real numbers x. Find an expression for f7 (x) as a polynomial in sin(3x) with
rational coefficients, and prove that it holds for all real numbers x.
5/4/15. Triangle ABC is an obtuse isosceles trian- B
gle with the property that three squares of equal ES ZF
Z
D S ZG
size can be inscribed in it as shown on the right. SS Z
HZ I ZZ
SS
The ratio AC/AB is an irrational number that is Z
S Z
Z
the root of a cubic polynomial. Determine that SZ Z
polynomial. A J K LM C
Complete, well-written solutions to at least two of the problems above, accompanied by a

DDM Co.
and postmarked by Sunday, 14 March 2004. Each participant is expected to develop solutions
without help from others. For the cover sheet and other details, see the USAMTS web site:
Round 1 Problems
Year 16 Academic Year 20042005
www.usamts.org
1/1/16. The numbers 1 through 10 can be arranged along 1

the vertices and sides of a pentagon so that the sum of the 5 Z Z8
three numbers along each side is the same. The diagram 10 Z7
B
on the right shows an arrangement with sum 16. Find, 2B

6

with proof, the smallest possible value for a sum and give B
B
an example of an arrangement with that sum. 4 9 3
2/1/16. For the equation
(3x2 + y 2 4y 17)3 (2x2 + 2y 2 4y 6)3 = (x2 y 2 11)3 ,

Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer
determine its solutions (x, y) where both x and y are integers. Prove that your answer
lists all the integer solutions.
3/1/16. Given that 5r + 4s + 3t + 6u = 100, where r s t u 0 are real numbers,

find, with proof, the maximum and minimum possible values of r + s + t + u.
4/1/16. The interior angles of a convex polygon form an arithmetic progression with a
common difference of 4 . Determine the number of sides of the polygon if its largest
interior angle is 172 .
A
5/1/16. Point G is where the medians of the triangle ABC in- b
BBb
@
tersect and point D is the midpoint of side BC. The triangle @b
F BP@bb E
BDG is equilateral with side length l. Determine the lengths, PPG b
AB, BC, and CA, of the sides of triangle ABC. B 1 @P1 Pb
B @ PP b
1
B D C
Round 1 Solutions must be submitted by October 4, 2004.

Please visit http://www.usamts.org for details about solution submission.

c 2004 Art of Problem Solving Foundation
Round 2 Problems
www.usamts.org
Please follow the rules below to ensure that your paper is graded properly.
1. Put your name, username, and USAMTS ID# on every page you submit.
2. Once you send in your solutions, that submission is final. You cannot resubmit solu-
tions.
3. Confirm that your email address in your Art of Problem Solving Forum Profile is
correct. You can do so by logging into the Forum, then clicking the Profile button
near the top of the Forum. If you are registered for the USAMTS and havent received
any email from us about the USAMTS, your email address is probably wrong in your
Forum Profile.
4. Do not fax solutions written in pencil.
5. No single page should contain solutions to more than one problem.
6. By the end of October, Round 1 results will be posted at www.usamts.org. To see your
results, log in to the Art of Problem Solving Forum, then go to My USAMTS in the
USAMTS pages.
7. Submit your solutions by November 22, 2004 (postmark deadline), via one of the
methods below.
(a) Email: solutions@usamts.org. Please see usamts.org for a list of acceptable file
types. Do not send .doc Microsoft Word files.
(b) Fax: (619) 445-2379
(c) Snail mail: USAMTS, P.O. Box 2090, Alpine, CA 919032090.
8. Reread item 1.
Round 2 Problems
www.usamts.org
1/2/16. The numbers 1 through 9 can be arranged in the triangles

labeled a through i illustrated on the right so that the numbers in aTT
each of the 2 2 triangles sum to the same value n; that is T
T c T
b T dT
a + b + c + d = b + e + f + g = d + g + h + i = n. T
T T
e T f g T h i T
T T T
For each possible sum n, show such an arrangement, labeled with
n
the sum as shown at right. Prove that there are no possible ar-
rangements for any other values of n.
to buy Virtual PDF Printer
Create PDF with GO2PDF for free, if you wish to remove this line, click here
2/2/16. Call a number a b 2 with a and b both positive integers tiny if it is closer to zero
than any number c d 2 such that c and d are positive
integers
with c < a and d < b.
Three numbers which are tiny are 1 2, 3 2 2, and 7 5 2. Without using a
calculator or computer, prove whether or not each of the following is tiny:

(a) 58 41 2 , (b) 99 70 2 .
3/2/16. A set is reciprocally whole if its elements are distinct integers greater than 1 and
the sum of the reciprocals of all those elements is exactly 1. Find a set S, as small
as possible, that contains two reciprocally whole subsets, I and J, which are distinct
but not necessarily disjoint (meaning they may share elements, but they may not be
the same subset). Prove that no set with fewer elements than S can contain two
reciprocally whole subsets.
4/2/16. How many quadrilaterals in the plane have four of the nine points q q q
(0, 0), (0, 1), (0, 2), (1, 0), (1, 1), (1, 2), (2, 0), (2, 1), (2, 2) as vertices? Do q q q
count both concave and convex quadrilaterals, but do not count figures
q q q
where two sides cross each other or where a vertex angle is 180 . Rigorously
verify that no quadrilateral was skipped or counted more than once.
5/2/16. Two circles of equal radius can tightly fit inside right triangle ABC, which has
AB = 13, BC = 12, and CA = 5, in the three positions illustrated below. Determine
the radii of the circles in each case.
A A A
! ! !

!!
! !! !
! !

!!
!! !! !!
! !
! !! !!
!

B C B C B C
Case (i) Case (ii) Case (iii)
Round 2 Solutions must be submitted by November 22, 2004.

Round 3 Problems
www.usamts.org
tions.
3. If you have already sent in an Entry Form and a Permission Form, you do not need to
resend them.
4. Confirm that your email address in your USAMTS Profile is correct. You can do so by
logging into the site, then clicking on My USAMTS on the sidebar, then click Profile.
If you are registered for the USAMTS and havent received any email from us about
the USAMTS, your email address is probably wrong in your Profile.
7. By the end of December, Round 2 results will be posted at www.usamts.org. To see

your results, log in to the USAMTS page, then go to My USAMTS.
8. Submit your solutions by January 3, 2005 (postmark deadline), via one of the methods
below.
(b) Fax: (619) 445-2379
9. Reread item 1.
Round 3 Problems
www.usamts.org
1/3/16. Given two integers x and y, let (xky) denote the concatenation of x by y, which
is obtained by appending the digits of y onto the end of x. For example, if x = 218 and
y = 392, then (xky) = 218392.
(a) Find 3-digit integers x and y such that 6(xky) = (ykx).
(b) Find 9-digit integers x and y such that 6(xky) = (ykx).
2/3/16. Find three isosceles triangles, no two of which are congruent, with integer sides, such
that each triangles area is numerically equal to 6 times its perimeter.
3/3/16. Define the recursive sequence 1, 4, 13, . . . by s1 = 1 and sn+1 = 3sn + 1 for all
positive integers n. The element s18 = 193710244 ends in two identical digits. Prove that
all the elements in the sequence that end in two or more identical digits come in groups of
three consecutive elements that have the same number of identical digits at the end.
4/3/16. Region ABCDEF GHIJ consists of 13 equal squares and is Q BX R

inscribed in rectangle P QRS with A on P Q, B on QR, E on RS, and XXX C
XX X
H on SP , as shown in the figure on the right. Given that P Q = 28 and
X
XX X
D
QR = 26, determine, with proof, the area of region ABCDEF GHIJ. AX
X
XX XXE
JX
X XX
XG F
IXX
P H S
5/3/16. Consideran isosceles triangle ABC with side lengths F
AB = AC = 10 2 and BC = 10 3. Construct semicircles
P , Q, and R with diameters AB, AC, BC respectively, such
that the plane of each semicircle is perpendicular to the plane D E
B C
of ABC, and all semicircles are on the same side of plane ABC
as shown. There exists a plane above triangle ABC that is
tangent to all three semicircles P , Q, R at the points D, E, and
F respectively, as shown in the diagram. Calculate, with proof, A
the area of triangle DEF .
Round 3 Solutions must be submitted by January 3, 2005.


Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Pri
Round 4 Problems
www.usamts.org
tions.
resend them.
7. Submit your solutions by March 14, 2005 (postmark deadline), via one of the methods
below.
(b) Fax: (619) 445-2379
8. Reread item 1.
Round 4 Problems
www.usamts.org
1/4/16. Determine with proof the number of positive integers n such that a convex regular
polygon with n sides has interior angles whose measures, in degrees, are integers.
2/4/16. Find positive integers a, b, and c such that

q

a + b + c = 219 + 10080 + 12600 + 35280.
Prove that your solution is correct. (Warning: numerical approximations of the values do
not constitute a proof.)
3/4/16. Find, with proof, a polynomial f (x, y, z) in three variables, with integer coefficients,
such that for all integers a, b,c, the
sign of f (a, b, c) (that is, positive, negative, or zero) is
3 3
the same as the sign of a + b 2 + c 4.
4/4/16. Find, with proof, all integers n such that there is a solution in nonnegative real
numbers (x, y, z) to the system of equations
2x2 + 3y 2 + 6z 2 = n and 3x + 4y + 5z = 23.
5/4/16. Medians AD, BE, and CF of triangle ABC meet

at G as shown. Six small triangles, each with a vertex at G, A
are formed. We draw the circles inscribed in triangles AF G,
BDG, and CDG as shown. Prove that if these three circles
are all congruent, then ABC is equilateral.
E F
C D B
Round 1 Problems
www.usamts.org
1. If you have not already sent an Entry Form, download an Entry Form from the Forms
page at
http://www.usamts.org/MyUSAMTS/U MyForms.php
and submit the completed form with your solutions.
resend them.
3. Put your name and USAMTS ID# on every page you submit.
tions.
7. No single page should contain solutions to more than one problem. Every solution you
submit should begin on a new page.
results, log in to the USAMTS page, then go to My USAMTS. Check that your email
address in your USAMTS Profile is correct; you will receive an email when the scores
are available.
9. Submit your solutions by October 3, 2005 (postmark deadline), via one (and only one!)
of the methods below.
(b) Fax: (619) 445-2379 (Please include a cover sheet indicating the number of pages
you are faxing, your name, and your User ID.)
10. Reread Items 19.

Round 1 Problems
www.usamts.org
1/1/17. An increasing arithmetic sequence with infinitely many terms is determined as follows.
A single die is thrown and the number that appears is taken as the first term. The die is
thrown again and the second number that appears is taken as the common difference between
each pair of consecutive terms. Determine with proof how many of the 36 possible sequences
formed in this way contain at least one perfect square.
2/1/17. George has six ropes. He chooses two of the twelve loose ends at random (possibly
from the same rope), and ties them together, leaving ten loose ends. He again chooses two
loose ends at random and joins them, and so on, until there are no loose ends. Find, with
proof, the expected value of the number of loops George ends up with.
3/1/17. Let r be a nonzero real number. The values of z which satisfy the equation
r4 z 4 + (10r6 2r2 )z 2 16r5 z + (9r8 + 10r4 + 1) = 0
are plotted on the complex plane (i.e. using the real part of each root as the x-coordinate
and the imaginary part as the y-coordinate). Show that the area of the convex quadrilateral
with these points as vertices is independent of r, and find this area.
4/1/17. Homer gives mathematicians Patty and Selma each a different integer, not known to
the other or to you. Homer tells them, within each others hearing, that the number given to
Patty is the product ab of the positive integers a and b, and that the number given to Selma
is the sum a + b of the same numbers a and b, where b > a > 1. He doesnt, however, tell
Patty or Selma the numbers a and b. The following (honest) conversation then takes place:
Patty: I cant tell what numbers a and b are.
Selma: I knew before that you couldnt tell.
Patty: In that case, I now know what a and b are.
Selma: Now I also know what a and b are.
Supposing that Homer tells you (but neither Patty nor Selma) that neither a nor b is greater
than 20, find a and b, and prove your answer can result in the conversation above.
5/1/17. Given triangle ABC, let M be the midpoint of side AB and N be the midpoint of
side AC. A circle is inscribed inside quadrilateral N M BC, tangent to all four sides, and
that circle touches M N at point X. The circle inscribed in triangle AM N touches M N at
point Y , with Y between X and N . If XY = 1 and BC = 12, find, with proof, the lengths
of the sides AB and AC.


Round 2 Problems
www.usamts.org
page at
resend them.
tions.

your results, log in to the USAMTS page, then go to My USAMTS. Check that your
email address in your USAMTS Profile is correct; you will receive an email when the
scores are available.
9. Submit your solutions by November 21, 2005 (postmark deadline), via one (and only
one!) of the methods below.

Round 2 Problems
www.usamts.org
1/2/17.
At the right is shown a 4 4 grid. We wish to fill in the grid such that each 1 2 3 4
row, each column, and each 2 2 square outlined by the thick lines contains
the digits 1 through 4. The first row has already been filled in. Find, with
proof, the number of ways we can complete the rest of the grid.
2/2/17. Write the number

1

2 32
as the sum of terms of the form 2q , where q is rational.
(For
example, 21 + 21/3 + 28/5 is a
sum of this form.) Prove that your sum equals 1/( 2 3 2).
3/2/17.
An equilateral triangle is tiled with n2 smaller congruent
equilateral triangles such that there are n smaller triangles
along each of the sides of the original triangle. The case
n = 11 is shown at right. For each of the small equilateral
triangles, we randomly choose a vertex V of the triangle
and draw an arc with that vertex as center connecting the
midpoints of the two sides of the small triangle with V as
an endpoint. Find, with proof, the expected value of the
number of full circles formed, in terms of n.
4/2/17. A teacher plays the game Duck-Goose-Goose with his class. The game is played as
follows: All the students stand in a circle and the teacher walks around the circle. As he
passes each student, he taps the student on the head and declares her a duck or a goose.
Any student named a goose leaves the circle immediately. Starting with the first student,
the teacher tags students in the pattern: duck, goose, goose, duck, goose, goose, etc., and
continues around the circle (re-tagging some former ducks as geese) until only one student
remains. This remaining student is the winner.
For instance, if there are 8 students, the game proceeds as follows: student 1 (duck),
student 2 (goose), student 3 (goose), student 4 (duck), student 5 (goose), student 6 (goose),
student 7 (duck), student 8 (goose), student 1 (goose), student 4 (duck), student 7 (goose)
and student 4 is the winner. Find, with proof, all values of n with n > 2 such that if the
circle starts with n students, then the nth student is the winner.
Round 2 Problems
www.usamts.org
5/2/17. Given acute triangle 4ABC in plane P, a point Q in space is defined such that
AQB = BQC = CQA = 90 . Point X is the point in plane P such that QX is
perpendicular to plane P. Given ABC = 40 and ACB = 75 , find AXC.


Round 3 Problems
www.usamts.org
page at
resend them.
tions.
8. In early March, Round 3 results will be posted at www.usamts.org. To see your results,
log in to the USAMTS page, then go to My USAMTS. Check that your email address
in your USAMTS Profile is correct; you will receive an email when the scores are
available.
9. Submit your solutions by January 9, 2006 (postmark deadline), via one (and only one!)

Round 3 Problems
www.usamts.org
1/3/17.
For a given positive integer n, we wish to construct a circle of 662 626
six numbers as shown at right so that the circle has the following
properties: 866 264
(a) The six numbers are different three-digit numbers, none of
whose digits is a 0. 486 648
(b) Going around the circle clockwise, the first two digits of each
number are the last two digits, in the same order, of the previous number.
(c) All six numbers are divisible by n.

The example above shows a successful circle for n = 2. For each of n = 3, 4, 5, 6, 7, 8, 9,
either construct a circle that satisfies these properties, or prove that it is impossible to do
so.
2/3/17. Anna writes a sequence of integers starting with the number 12. Each subsequent
integer she writes is chosen randomly with equal chance from among the positive divisors of
the previous integer (including the possibility of the integer itself). She keeps writing integers
until she writes the integer 1 for the first time, and then she stops. One such sequence is
12, 6, 6, 3, 3, 3, 1.
What is the expected value of the number of terms in Annas sequence?
3/3/17.
Points A, B, and C are on a circle such that 4ABC is an acute A
triangle. X, Y , and Z are on the circle such that AX is perpen- Y
dicular to BC at D, BY is perpendicular to AC at E, and CZ is E
perpendicular to AB at F . Find the value of Z F
AX BY CZ
+ + , B C
AD BE CF D
and prove that this value is the same for all possible A, B, C on the X
circle such that 4ABC is acute.
4/3/17. Find, with proof, all triples of real numbers (a, b, c) such that all four roots of the
polynomial f (x) = x4 + ax3 + bx2 + cx + b are positive integers. (The four roots need not be
distinct.)
Problem 5 on next page.
Round 3 Problems
www.usamts.org
5/3/17. Lisa and Bart are playing a game. A round table has n lights evenly spaced around
its circumference. Some of the lights are on and some of them off; the initial configuration
is random. Lisa wins if she can get all of the lights turned on; Bart wins if he can prevent
this from happening.
On each turn, Lisa chooses the positions at which to flip the lights, but before the lights
are flipped, Bart, knowing Lisas choices, can rotate the table to any position that he chooses
(or he can leave the table as is). Then the lights in the positions that Lisa chose are flipped:
those that are off are turned on and those that are on are turned off.
Here is an example turn for n = 5 (a white circle indicates a light that is on, and a black
circle indicates a light that is off):
1
5 2
Initial Position.
4 3
1
Lisa says 1, 3, 4.
5 2
Bart rotates the table one
position counterclockwise.
4 3
1
Lights in positions 1, 3, 4 are 5 2
flipped.
4 3
Lisa can take as many turns as she needs to win, or she can give up if it becomes clear
to her that Bart can prevent her from winning.
(a) Show that if n = 7 and initially at least one light is on and at least one light is off,
then Bart can always prevent Lisa from winning.
(b) Show that if n = 8, then Lisa can always win in at most 8 turns.


Round 4 Problems
www.usamts.org
page at
resend them.
tions.
8. In April, Round 4 results will be posted at www.usamts.org. To see your results, log in
to the USAMTS page, then go to My USAMTS. Check that your email address in your
USAMTS Profile is correct; you will receive an email when the scores are available.
9. Submit your solutions by March 13, 2006 (postmark deadline), via one (and only one!)

Round 4 Problems
www.usamts.org
1/4/17. AB is a diameter of circle C1 . Point P is on C1 such that AP > BP . Circle C2 is

centered at P with radius P B. The extension of AP past P meets C2 at Q. Circle C3 is
centered at A and is externally tangent to C2 . R is on C3 such that AR AQ. Circle C4
passes through A, Q, and R. Find, with proof, the ratio between the area of C4 and the area
of C1 , and show that this ratio is the same for all points P on C1 such that AP > BP .
2/4/17. Centered hexagonal numbers are the numbers of dots used to create hexagonal arrays
of dots. The first four centered hexagonal numbers are 1, 7, 19, and 37, as shown below.
qqq qqqqqqqqq
q qqqqqqq qqqqqqqqqqqqq qqqqqqqqqqqqqqqqqqq
qqq qqqqqqqqq
1 7 19 37
Centered Hexagonal Numbers
Consider an arithmetic sequence 1, a, b and a geometric sequence 1, c, d, where a, b, c, and
d are all positive integers and a + b = c + d. Prove that each centered hexagonal number is a
possible value of a, and prove that each possible value of a is a centered hexagonal number.
3/4/17. We play a game. The pot starts at $0. On every turn, you flip a fair coin. If you flip
heads, I add $100 to the pot. If you flip tails, I take all of the money out of the pot, and
you are assessed a strike. You can stop the game before any flip and collect the contents
of the pot, but if you get 3 strikes, the game is over and you win nothing. Find, with proof,
the expected value of your winnings if you follow an optimal strategy.
4/4/17. Find, with proof, all irrational numbers x such that both x3 6x and x4 8x2 are
rational.
5/4/17. Sphere
S is inscribed in cone C. The height of C equals its radius, and both equal
12 + 12 2. Let the vertex of the cone be A and the center of the sphere be B. Plane P
is tangent to S and intersects segment AB. X is the point on the intersection of P and C
closest to A. Given that AX = 6, find the area of the region of P enclosed by the intersection
of C and P.
Round 4 Solutions must be submitted by March 13, 2006.


Round 1 Problems
www.usamts.org
1. You must show your work and prove your answers on all problems. If you just send a
numerical answer for a problem with no proof, you will get no more than 1 point.
page at
resend them.
tions.
are available.
10. Submit your solutions by October 10, 2006 (postmark deadline), via one (and only

Round 1 Problems
www.usamts.org
1/1/18. When we perform a digit slide on a number, we move its units digit to the front of
the number. For example, the result of a digit slide on 6471 is 1647. What is the smallest
positive integer with 4 as its units digit such that the result of a digit slide on the number
equals 4 times the number?
2/1/18.
(a) In how many different ways can the six empty circles in the diagram 1
at right be filled in with the numbers 2 through 7 such that each number is
usedforonce,
Create PDF with GO2PDF free, if youand
wish to each number
remove this line, click hereis either
to buy greater
Virtual PDF Printer than both its neighbors, or less
than both its neighbors?
(b) In how many different ways can the seven empty circles in the diagram
1
used once, and each number is either greater than both its neighbors, or less
3/1/18.
(a) An equilateral triangle is divided into 25 congruent smaller equi- 1
lateral triangles, as shown. Each of the 21 vertices is labeled with a
number such that for any three consecutive vertices on a line segment,
their labels form an arithmetic sequence. The vertices of the original
equilateral triangle are labeled 1, 4, and 9. Find the sum of the 21 labels.
4 9
(b) Generalize part (a) by finding the sum of the labels when there are
2
n smaller congruent equilateral triangles, and the labels of the original equilateral triangle
are a, b, and c.
4/1/18. Every point in the plane is colored either red, green, or blue. Prove that there exists
a rectangle in the plane such that all four of its vertices are the same color.
5/1/18. ABCD is a tetrahedron such that AB = 6, BC = 8, AC = AD = 10, and BD =

CD = 12. Plane P is parallel to face ABC and divides the tetrahedron into two pieces of
equal volume. Plane Q is parallel to face DBC and also divides ABCD into two pieces of
equal volume. Line ` is the intersection of planes P and Q. Find the length of the portion
of ` that is inside ABCD.


Round 2 Problems
www.usamts.org
page at
resend them.
tions.
are available.

Round 2 Problems
www.usamts.org
1/2/18. Find all positive integers n such that the sum of the squares of the digits of n is 2006
less than n.
2/2/18. ABCD is a trapezoid with BC k AD, ADC = 57 , B M C

DAB = 33 , BC = 6, and AD = 10. M and N are the
midpoints of BC and AD, respectively.
A D
N
(a) Find M N A.
(b) Find M N .
3/2/18. The expression bxc means the greatest integer that is smaller than or equal to x, and
dxe means the smallest integer that is greater than or equal to x. These functions are called
the floor function and ceiling function, respectively. Find, with proof, a polynomial f (n)
equivalent to
n2 j
X k l m
k + k
k=1
for all positive integers n.
4/2/18. For every integer k 2, find a formula (in terms of k) for the smallest positive integer
n that has the following property:
No matter how the elements of {1, 2, . . . , n} are colored red and blue, we can find
k elements a1 , a2 , . . . , ak , where the ai are not necessarily distinct elements, and
an element b such that:
(a) a1 + a2 + + ak = b, and
(b) all of the ai s and b are the same color.
5/2/18. In triangle ABC, AB = 8, BC = 7, and AC = 5. We extend AC D

past A and mark point D on the extension, as shown. The bisector of E A
DAB meets the circumcircle of 4ABC again at E, as shown. We draw F
a line through E perpendicular to AB. This line meets AB at point F .
Find the length of AF . C


Round 3 Problems
www.usamts.org
page at
resend them.
tions.

your results, log in to the USAMTS page, then go to My USAMTS. Check that your
email address in your USAMTS Profile is correct; you will receive an email when the
scores are available.

Round 3 Problems
www.usamts.org
1/3/18. In how many distinguishable ways can the edges of a cube be colored such that each
edge is yellow, red, or blue, and such that no two edges of the same color share a vertex?
(Two cubes are indistinguishable if they can be rotated into positions such that the two
cubes are colored exactly the same.)
2/3/18. Find, with proof, all real numbers x between 0 and 2 such that
tan 7x sin 6x = cos 4x cot 7x.

3/3/18. Three circles with radius 2 are drawn in a plane such that each circle is tangent to
the other two. Let the centers of the circles be points A, B, and C. Point X is on the circle
with center C such that AX + XB = AC + CB. Find the area of 4AXB.
4/3/18. Alice plays in a tournament in which every player plays a game against every other
player exactly once. In each game, either one player wins and earns 2 points while the other
gets 0 points, or the two players tie and both players earn 1 point. After the tournament,
Alice tells Bob how many points she earned. Bob was not in the tournament, and does not
know what happened in any individual game of the tournament.
(a) Suppose there are 12 players in the tournament, including Alice. What is the smallest
number of points Alice could have earned such that Bob can deduce that Alice scored
more points than at least 8 other players?
(b) Suppose there are n players in the tournament, including Alice, and that Alice scored
m points. Find, in terms of n and k, the smallest value of m such that Bob can deduce
that Alice scored more points than at least k other players.
5/3/18. Let f (x) be a strictly increasing function defined for all x > 0 such that f (x) > x1
and f (x)f (f (x) + x1 ) = 1 for all x > 0. Find f (1).


Round 4 Problems
www.usamts.org
page at
resend them.
tions.
are available.

Round 4 Problems
www.usamts.org
n
X
1/4/18. Let S(n) = (1)i+1 i. For example, S(4) = 1 2 + 3 4 = 2.
i=1
(a) Find, with proof, all positive integers a, b such that S(a) + S(b) + S(a + b) = 2007.
(b) Find, with proof, all positive integers c, d such that S(c) + S(d) + S(c + d) = 2008.
2/4/18. For how many integers n between 1 and 102007 , inclusive, are the last 2007 digits of n
and n3 the same? (If n or n3 has fewer than 2007 digits, treat it as if it had zeros on the left
to compare the last 2007 digits.)
3/4/18. Let ABCD be a convex quadrilateral. Let M be the midpoint C

D
of diagonal AC and N be the midpoint of diagonal BD. Let O be the M
intersection of the line through N parallel to AC and the line through M O
parallel to BD. Prove that the line segments joining O to the midpoints N
of each side of ABCD divide ABCD into four pieces of equal area. A
4/4/18. We are given a 2 n array of nodes, where n is a positive integer. A valid connection
of the array is the addition of 1-unit-long horizontal and vertical edges between nodes, such
that each node is connected to every other node via the edges, and there are no loops of any
size. We give some examples for n = 3:
Valid Valid Invalid: loop Invalid: not connected
Let Tn denote the number of valid connections of the 2 n array. Find T10 .
5/4/18. A sequence of positive integers (x1 , x2 , . . . , x2007 ) satisfies the following two conditions:
(1) xn 6= xn+1 for 1 n 2006, and

x1 + x2 + + xn
(2) An = is an integer for 1 n 2007.
n
Find the minimum possible value of A2007 .


Round 1 Problems
www.usamts.org
Please follow the rules below:
page at
http://www.usamts.org/MyUSAMTS/U_MyForms.php
resend them.
tions.
6. Confirm that your email address in your USAMTS Profile is correct. You can do so
by logging into www.usamts.org, then clicking on My USAMTS on the sidebar, then
clicking Profile. If you are registered for the USAMTS and havent received any email
from us about the USAMTS, your email address is probably wrong in your Profile.
submit should begin on a new page, and you should only submit work on one side of
each piece of paper.
9. Round 1 results will be posted at www.usamts.org by mid-November. To see your

are available.
10. Submit your solutions by October 9, 2007 (postmark deadline), via one (and only one!)
types. Do not send Microsoft Word files.
(b) Fax: (619) 445-2379 (You must include a cover sheet indicating the number of
pages you are faxing, your name, and your USAMTS ID#.)

Round 1 Problems
www.usamts.org
1/1/19. Gene has 2n pieces of paper numbered 1 through 2n. He removes n pieces of paper
that are numbered consecutively. The sum of the numbers on the remaining pieces of paper
is 1615. Find all possible values of n.
2/1/19. A regular 18-gon is dissected into 18 pentagons, each of which is congruent to pentagon
ABCDE, as shown. All sides of the pentagon have the same length.
C Z
B Y
X
D
A E
(a) Determine angles A, B, C, D, and E.
(b) Show that points X, Y , and Z are collinear.
3/1/19. Find all positive integers a b c such that

1 1 1
arctan + arctan + arctan = .
a b c 4
4/1/19. In convex quadrilateral ABCD, AB = CD, ABC = 77 , and BCD = 150 . Let
P be the intersection of the perpendicular bisectors of BC and AD. Find BP C.
5/1/19. Let c be a real number. The sequence a1 , a2 , a3 , . . . is defined by a1 = c and

an = 2a2n1 1 for all n 2. Find all values of c such that an < 0 for all n 1.


Round 2 Problems
www.usamts.org
page at
resend them.
tions.
9. Round 2 results will be posted at www.usamts.org by mid-January. To see your results,

log in to the USAMTS page, then go to My USAMTS. Check that your email address
in your USAMTS Profile is correct; you will receive an email when the scores are
available.

Round 2 Problems
www.usamts.org
1/2/19. Find the smallest positive integer n such that every possible coloring of the integers
from 1 to n with each integer either red or blue has at least one arithmetic progression of
three different integers of the same color.
2/2/19. Let x, y, and z be complex numbers such that x + y + z = x5 + y 5 + z 5 = 0 and

x3 + y 3 + z 3 = 3. Find all possible values of x2007 + y 2007 + z 2007 .
3/2/19. A triangular array of positive integers is called remarkable if all of its entries are
distinct,
Create PDF with GO2PDF for free, ifand each
you wish entry,
to remove other
this line, click here tothan those
buy Virtual in the top row, is the quotient of the two numbers
PDF Printer
immediately above it. For example, the following triangular array is remarkable:
7 42 14
6 3
2
Find the smallest positive integer that can occur as the greatest element in a remarkable
array with four numbers in the top row.
4/2/19. Two nonoverlapping arcs of a circle are chosen. Eight distinct points are then chosen
on each arc. All 64 segments connecting a chosen point on one arc to a chosen point on the
other arc are drawn. How many triangles are formed that have at least one of the 16 points
as a vertex?
A sample figure is shown below:
5/2/19. Faces ABC and XY Z of a regular icosahedron are parallel, with the vertices labeled
such that AX, BY , and CZ are concurrent. Let S be the solid with faces ABC, AY Z,
BXZ, CXY , XBC, Y AC, ZAB, and XY Z. If AB = 6, what is the volume of S?


Round 3 Problems
www.usamts.org
page at
resend them.
tions.
9. Round 3 results will be posted at www.usamts.org by mid-February. To see your

are available.

Round 3 Problems
www.usamts.org
1/3/19. We construct a sculpture consisting of infinitely many cubes, as follows. Start with
a cube with side length 1. Then, at the center of each face, attach a cube with side length
1
3
(so that the center of a face of each attached cube is the center of a face of the original
cube). Continue this procedure indefinitely: at the center of each exposed face of a cube in
the structure, attach (in the same fashion) a smaller cube with side length one-third that of
the exposed face. What is the volume of the entire sculpture?
2/3/19. Gene starts with the 3 3 grid of 0s shown at left below. He then repeatedly chooses
a 2 2 square within the grid and increases all four numbers in the chosen 2 2 square by
1. One possibility for Genes first three steps is shown below.
0 0 0 0 0 0 0 1 1 0 1 1
0 0 0 1 1 0 1 2 1 2 3 1
0 0 0 1 1 0 1 1 0 2 2 0
How many different grids can be produced with this method such that each box contains
an integer from 1 to 12, inclusive? (The numbers in the boxes need not be distinct.)
3/3/19. Consider all polynomials f (x) with integer coefficients such that f (200) = f (7) = 2007
and 0 < f (0) < 2007. Show that the value of f (0) does not depend on the choice of
polynomial, and find f (0).
4/3/19. Prove that 101 divides infinitely many of the numbers in the set
{2007, 20072007, 200720072007, 2007200720072007, . . .}.
p
5/3/19. For every rational number 0 < < 1, where p and q are relatively prime, construct a
q 1
p 1 1
circle with center , and diameter 2 . Also construct circles centered at 0, and
1 q 2q 2 q 2
1, with diameter 1.
2
(a) Prove that any two such circles intersect in at most
1 point.
X

1
i=1 i3
(b) Prove that the total area of all of the circles is 1+ X1
.
4
4
i=1 i


Round 4 Problems
www.usamts.org
page at
resend them.
tions.
9. Results will be posted at www.usamts.org by mid-April. To see your results, log in to

the USAMTS page, then go to My USAMTS. Check that your email address in your
USAMTS Profile is correct; you will receive an email when the scores are available.

Round 4 Problems
www.usamts.org
1/4/19. In the diagram at right, each vertex is labeled with a different

positive factor of 2008, such that if two vertices are connected by an
edge, then the label of one vertex divides the label of the other vertex.
In how many different ways can the vertices be labeled? Two labelings
are considered the same if one labeling can be obtained by rotating
and/or reflecting the other labeling.
n + n + n + n < n,
2/4/19. Determine, with proof, the greatest integer n such that
2 3 11 13
where bxc is the greatest integer less than or equal to x.
3/4/19. Let 0 < < 1. Define a sequence {an } of real numbers by a1 = 1 and for all integers
k 1,
a2k = ak ,
a2k+1 = (1 )ak .
Find the value of the sum

Xa

2k a2k+1 in terms of .
k=1
4/4/19. Suppose that w, x, y, z are positive real numbers such that w + x < y + z. Prove that
it is impossible to simultaneously satisfy both
(w + x)yz < wx(y + z) and (w + x)(y + z) < wx + yz.
(Continued on next page.)

Round 4 Problems
www.usamts.org
5/4/19. Let P1 P2 P3 P13 be a regular 13-gon. For 1 i 6, let di = P1 Pi+1 . The 13

diagonals of length d6 enclose a smaller regular 13-gon, whose side length we denote by s.
Express s in the form
s = c1 d1 + c2 d2 + c3 d3 + c4 d4 + c5 d5 + c6 d6 ,
where c1 , c2 , c3 , c4 , c5 , and c6 are integers.
P4
P5
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer P3
P6
P2
d4 d3 d2
P7 d5 d1
d6
P1
P8
P13
P9
P12
P10
P11


Round 1 Problems
www.usamts.org
Important information:
1. You must show your work and prove your answers on all problems. If you
just send a numerical answer for a problem with no proof, you will get no more than
1 point.
each
Create PDF with GO2PDF for free, if youpiece of paper.
wish to remove this line, click here to buy Virtual PDF Printer
4. Submit your solutions by October 14, 2008, via one (and only one!) of the methods
below:
(a) Web: Log on to www.usamts.org to upload a PDF file containing your solutions.
(No other file type will be accepted.)
(b) Mail: USAMTS, P.O. Box 2090, Alpine, CA 919032090.
(Solutions must be postmarked on or before the deadline.)
tions.
6. Confirm that your email address in your USAMTS Profile is correct. You can do
so by logging onto www.usamts.org and visiting the My USAMTS pages. (If you
are registered for the USAMTS and havent received any email from us about the
USAMTS, then your email address is probably wrong in your Profile.)
7. Round 1 results will be posted at www.usamts.org when available. To see your results,
log on to the USAMTS website, then go to My USAMTS. You will also receive an
email when your scores and comments are available (provided that you did item #6
above).
These are only part of the complete rules.

Please read the entire rules on www.usamts.org.
Round 1 Problems
www.usamts.org
Each problem is worth 5 points.
1/1/20. 27 unit cubes25 of which are colored black and 2 of

which are colored whiteare assembled to form a 3 3 3
cube. How many distinguishable cubes can be formed? (Two
cubes are indistinguishable if one of them can be rotated to
appear identical to the other. An example of two indistin-
guishable cubes is shown at right.)
2/1/20. Find all positive integers n for which it is possible to find three positive factors x, y,
and z of n 1, with x > y > z, such that x + y + z = n.
3/1/20. Let a, b, c be real numbers. Suppose that for all real numbers x such that |x| 1, we
have |ax2 + bx + c| 100. Determine the maximum possible value of |a| + |b| + |c|.

4/1/20. A pointP inside a regular tetrahedron ABCD is such that P A = P B = 11 and
P C = P D = 17. What is the side length of ABCD?

5/1/20. Call a positive real number groovy if it can be written in the form n + n + 1 for
some positive integer n. Show that if x is groovy, then for any positive integer r, the number
xr is groovy as well.


Round 2 Problems
www.usamts.org
1 point.
each
4. Submit your solutions by November 24, 2008, via one (and only one!) of the methods
below:
(Solutions must be postmarked on or before the deadline.)
tions.
above).

Round 2 Problems
www.usamts.org
1/2/20. Sarah and Joe play a standard 3-by-3 game of tic-tac-toe.

Sarah goes first and plays X, and Joe goes second and plays O. They
alternate turns placing their letter in an empty space, and the first
to get 3 of their letters in a straight line (across, down, or diagonal)
wins. How many possible final positions are there, given that Sarah
wins on her 4th move? (Dont assume that the players play with any
sort of strategy; one example of a possible final position is shown at
right.)
2/2/20. Let C1 , C2 , C3 , and C4 be four circles, with radii 1, 1, 3, and 3, respectively, such
that circles C1 and C2 , C2 and C3 , C3 and C4 , and C4 and C1 are externally tangent. A
fifth circle C is smaller than the four other circles and is externally tangent to each of them.
Find the radius of C.
3/2/20. Find, with proof, all polynomials p(x) with the following property:
There exists a sequence a0 , a1 , a2 , . . . of positive integers such that p(a0 ) = 1 and

for all positive integers n:
(a) p(an ) is a positive integer, and

n1
X 1 1
(b) + = 1.
a
j=0 j
p(an )
4/2/20. Find, with proof, the largest positive integer k with the following property:
There exists a positive number N such that N is divisible by all but three of the
integers 1, 2, 3, . . . , k, and furthermore those three integers (that dont divide N )
are consecutive.
5/2/20. The set S consists of 2008 points evenly spaced on a circle of radius 1 (so that S forms
the vertices of a regular 2008-sided polygon). 3 distinct points X, Y, Z in S are chosen at

random. The expected value of the area of 4XY Z can be written in the form r cot 2008 ,
where r is a rational number. Find r.


Round 3 Problems
www.usamts.org
Important! New deadline for web submission: 3 PM Eastern / Noon Pacific
1 point.
4. Submit your solutions by January 12, 2009, via one (and only one!) of the methods
below:
Deadline: 3 PM Eastern / Noon Pacific on January 12
(Solutions must be postmarked on or before the deadline day.)
tions.
above).

Round 3 Problems
www.usamts.org
1/3/20. Let S be the set of all 10-digit numbers (which by definition may not begin with 0) in
which each digit 0 through 9 appears exactly once. For example, the number 3,820,956,714
is in S. A number n is picked from S at random. What is the probability that n is divisible
by 11?
2/3/20. Two players are playing a game that starts with 2009 stones. The players take turns
removing stones. A player may remove exactly 3, 4, or 7 stones on his or her turn, except
that if only 1 or 2 stones are remaining then the player may remove them all. The player who
removes the last stone wins. Determine, with proof, which player has a winning strategy,
the first or the second player.
3/3/20. Let a, b, c be three positive integers such that
(lcm(a, b))(lcm(b, c))(lcm(c, a)) = (abc) gcd(a, b, c),
where lcm means least common multiple and gcd means greatest common divisor.
Given that no quotient of any two of a, b, c is an integer (that is, none of a, b, c is an integer
multiple of any other of a, b, c), find the minimum possible value of a + b + c.
4/3/20. Given a segment BC in plane P, find the locus of all points A in P with the following
property:
There exists exactly one point D in P such that ABDC is
a cyclic quadrilateral and BC bisects AD, as shown at right.
A
5/3/20. Let b be an integer such that b 2, and let a > 0 be a
1 1
real number such that + > 1. Prove that the sequence
a b B C
bac, b2ac, b3ac, . . .

D
contains infinitely many integral powers of b. (Note that bxc is defined to be the greatest
integer less than or equal to x.)

Web deadline: 3 PM Eastern / Noon Pacific on the due date
Mail deadline: Postmarked on or before due date

Round 4 Problems
www.usamts.org
Important! New deadline for web submission: 3 PM Eastern / Noon Pacific
1 point.
4. Submit your solutions by March 9, 2009, via one (and only one!) of the methods below:
Deadline: 3 PM Eastern / Noon Pacific on March 9
(Solutions must be postmarked on or before the deadline day.)
tions.
above).

Round 4 Problems
www.usamts.org
a2n1
1/4/20. Consider a sequence {an } with a1 = 2 and an = for all n 3. If we know that
an2
a2 and a5 are positive integers and a5 2009, then what are the possible values of a5 ?
2/4/20. There are k mathematicians at a conference. For each integer n from 0 to 10, inclusive,
there is a group of 5 mathematicians such that exactly n pairs of those 5 mathematicians
are for
Create PDF with GO2PDF friends. Find
free, if you wish (with
to remove this line,proof) theVirtual
click here to buy smallest
PDF Printer possible value of k.
3/4/20. A particle is currently at the point (0, 3.5) on the plane and is moving towards the
origin. When the particle hits a lattice point (a point with integer coordinates), it turns with
equal probability 45 to the left or to the right from its current course. Find the probability
that the particle reaches the x-axis before hitting the line y = 6.
4/4/20. Find, with proof, all functions f defined on nonnegative integers taking nonnegative
integer values such that
f (f (m) + f (n)) = m + n
for all nonnegative integers m, n.
5/4/20. A circle C1 with radius 17 intersects a circle C2 with C2

radius 25 at points A and B. The distance between the C1 A
centers of the circles is 28. Let N be a point on circle C2
such that the midpoint M of chord AN lies on circle C1 .
Find the length of AN .
M

Web deadline: 3 PM Eastern / Noon Pacific on the due date
Mail deadline: Postmarked on or before due date

Round 1 Problems
www.usamts.org
1 point.
each
below:
Deadline: 3 PM Eastern / Noon Pacific on October 13
(Solutions must be postmarked on or before October 13.)
tions.
above).

Round 1 Problems
www.usamts.org
1/1/21. Fill in the circles in the picture at right with the digits 1-8, one
digit in each circle with no digit repeated, so that no two circles that
are connected by a line segment contain consecutive digits. In how
many ways can this be done?
2/1/21. The ordered pair of four-digit numbers (2025, 3136) has the
property that each number in the pair is a perfect square and each digit
of the second number is 1 more than the corresponding digit of the
first number. Find, with proof, all ordered pairs of five-digit numbers
and ordered pairs of six-digit numbers with the same property: each
number in the pair is a perfect square and each digit of the second
number is 1 more than the corresponding digit of the first number.
3/1/21. A square of side length 5 is inscribed in a square of side length 7.

If we construct a grid of 1 1 squares for both squares, as shown to the
right, then we find that the two grids have 8 lattice points in common.
If we do the same construction by inscribing a square of side length 1489
in a square of side length 2009, and construct a grid of 1 1 squares in
each large square, then how many lattice points will the two grids of 1 1
squares have in common?
4/1/21. Let ABCDEF be a convex hexagon, such that F A = AB, BC = CD, DE = EF ,

and F AB = 2EAC. Suppose that the area of ABC is 25, the area of CDE is 10, the
area of EF A is 25, and the area of ACE is x. Find, with proof, all possible values of x.
5/1/21. The cubic equation x3 +2x1 = 0 has exactly one real root r. Note that 0.4 < r < 0.5.
(a) Find, with proof, an increasing sequence of positive integers a1 < a2 < a3 < such
that
1
= ra1 + ra2 + ra3 + .
2
(b) Prove that the sequence that you found in part (a) is the unique increasing sequence
with the above property.


Round 2 Problems
www.usamts.org
1 point.
each
below:
Deadline: 3 PM Eastern / Noon Pacific on November 23
(Solutions must be postmarked on or before November 23.)
tions.
above).

Round 2 Problems
www.usamts.org
1/2/21. Jeremy has a magic scale, each side of which holds a positive integer. He plays the
following game: each turn, he chooses a positive integer n. He then adds n to the number
on the left side of the scale, and multiplies by n the number on the right side of the scale.
(For example, if the turn starts with 4 on the left and 6 on the right, and Jeremy chooses
n = 3, then the turn ends with 7 on the left and 18 on the right.) Jeremy wins if he can
make both sides of the scale equal.
(a) Show that if the game starts with the left scale holding 17 and the right scale holding 5,
then Jeremy can win the game in 4 or fewer turns.
(b) Prove that if the game starts with the right scale holding b, where b 2, then Jeremy
can win the game in b 1 or fewer turns.
2/2/21. Alice has three daughters, each of whom has two daughters; each of Alices six grand-
daughters has one daughter. How many sets of women from the family of 16 can be chosen
such that no woman and her daughter are both in the set? (Include the empty set as a
possible set.)
3/2/21. Prove that if a and b are positive integers such that a2 + b2 is a multiple of 72009 , then
ab is a multiple of 72010 .
4/2/21. The Rational Unit Jumping Frog starts at (0, 0) on the Cartesian plane, and each
minute jumps a distance of exactly 1 unit to a point with rational coordinates.
1
(a) Show that it is possible for the frog to reach the point 51 , 17

in a finite amount of time.
(b) Show that the frog can never reach the point 0, 41 .

5/2/21. Let ABC be a triangle with AB = 3, AC = 4, and BC = 5, let P be a point on

BC, and let Q be the point (other than A) where the line through A and P intersects the
circumcircle of ABC. Prove that
25
PQ < .
4 6


Round 3 Problems
www.usamts.org
1 point.
each
below:
(Solutions must be postmarked on or before January 11.)
tions.
above).

Round 3 Problems
www.usamts.org
1/3/21. Let ABCD be a convex quadrilateral with AC BD, and let P be the intersection
of AC and BD. Suppose that the distance from P to AB is 99, the distance from P to BC
is 63, and the distance from P to CD is 77. What is the distance from P to AD?
2/3/21. Find, with proof, a positive integer n such that
(n + 1)(n + 2) (n + 500)
500!
is an integer with no prime factors less than 500.
3/3/21. We are given a rectangular piece of white paper with length 25 and width 20. On the
paper we color blue the interiors of 120 disjoint squares of side length 1 (the sides of the
squares do not necessarily have to be parallel to the sides of the paper). Prove that we can
draw a circle of diameter 1 on the remaining paper such that the entire interior of the circle
is white.
4/3/21. Let a and b be positive integers such that all but 2009 positive integers are expressible
in the form ma + nb, where m and n are nonnegative integers. If 1776 is one of the numbers
that is not expressible, find a + b.
5/3/21. The sequences (an ), (bn ), and (cn ) are defined by a0 = 1, b0 = 0, c0 = 0, and
cn1 an1 bn1

an = an1 + , bn = bn1 + , cn = cn1 +
n n n
for all n 1. Prove that
an n + 1 < 2

3 3n
for all n 1.


Round 4 Problems
www.usamts.org
1 point.
each
4. Submit your solutions by March 8, 2010, via one (and only one!) of the methods below:
Deadline: 3 PM Eastern / Noon Pacific on March 8
(Solutions must be postmarked on or before March 8.)
tions.
7. Round 4 and full-year results will be posted at www.usamts.org when available. To

see your results, log on to the USAMTS website, then go to My USAMTS. You will
also receive an email when your scores and comments are available (provided that you
did item #6 above).

Round 4 Problems
www.usamts.org
1/4/21. Archimedes planned to count all of the prime numbers between 2 and 1000 using the
Sieve of Eratosthenes as follows:
(a) List the integers from 2 to 1000.
(b) Circle the smallest number in the list and call this p.
(c) Cross out all multiples of p in the list except for p itself.
(d) Let p be the smallest number remaining that is neither circled nor crossed out. Circle p.
(e) Repeat steps (c) and (d) until each number is either circled or crossed out.
At the end of this process, the circled numbers are prime and the crossed out numbers
are composite.
Unfortunately, while crossing off the multiples of 2, Archimedes accidentally crossed out
two odd primes in addition to crossing out all the even numbers (besides 2). Otherwise,
he executed the algorithm correctly. If the number of circled numbers remaining when
Archimedes finished equals the number of primes from 2 to 1000 (including 2), then what is
the largest possible prime that Archimedes accidentally crossed out?
2/4/21. Let a, b, c, d be four real numbers such that
a + b + c + d = 8,
ab + ac + ad + bc + bd + cd = 12.
Find the greatest possible value of d.

3/4/21. I give you a deck of n cards numbered 1 through n. On each turn, you take the top
card of the deck and place it anywhere you choose in the deck. You must arrange the cards
in numerical order, with card 1 on top and card n on the bottom. If I place the deck in a
random order before giving it to you, and you know the initial order of the cards, what is the
expected value of the minimum number of turns you need to arrange the deck in order?
4/4/21. Let S be a set of 10 distinct positive real numbers. Show that there exist x, y S
such that
(1 + x)(1 + y)
0<xy < .
9
5/4/21. Tina and Paul are playing a game on a square S. First, Tina selects a point T inside S.
Next, Paul selects a point P inside S. Paul then colors blue all the points inside S that are
closer to P than T . Tina wins if the blue region thus produced is the interior of a triangle.
Assuming that Paul is lazy and simply selects his point at random (and that Tina knows
this), find, with proof, a point Tina can select to maximize her probability of winning, and
compute this probability.


Round 1 Problems
www.usamts.org
1 point.
below:
tions.
by logging onto www.usamts.org and visiting the My USAMTS pages.
above).

Round 1 Problems
www.usamts.org
1/1/22.
Given a set S of points in the plane, a line is called happy if it contains at
least 3 points in S. For example, if S is the 3 3 grid of points shown at right,
then there are 8 happy lines as shown.
(a) If S is the 3 9 grid shown below, how many happy lines are there?
(b) Find, with proof, a set S (in the plane) with 27 points that has exactly 49 happy lines.
2/1/22. Youre at vertex A of triangle ABC, where B = C = 65 . The sides of the triangle
are perfectly reflective; if you shoot a laser from A to the midpoint of BC, it will reflect
once and return to A. Suppose you fire at a point on BC other than its midpoint, and the
beam still returns to A after reflecting some number of times. What is the smallest number
of reflections the beam can make before returning to A? What is the smallest angle between
AB and the initial beam that produces this number of reflections?
3/1/22. Find c > 0 such that if r, s, and t are the roots of the cubic
f (x) = x3 4x2 + 6x + c,
then
1 1 1
1= + 2 + 2 .
r2 +s 2 s +t 2 t + r2
4/1/22. Sasha has a compass with fixed radius s and Rebecca has a compass with fixed radius r.
Sasha draws a circle (with his compass) and Rebecca then draws a circle (with her compass)
that intersects Sashas circle twice. We call these intersection points C and D.
Charlie draws a common tangent to both circles, meeting Sashas circle at point A and
Rebeccas circle at point B, and then draws the circle passing through A, B, and C. Prove
that the radius of Charlies circle does not depend on where Sasha and Rebecca choose to
draw their circles, or which of the two common tangents Charlie draws.
Round 1 Problems
www.usamts.org
5/1/22. A convex polygon P is called peculiar if: (a) for some n 3, the vertices of P are a
subset of the vertices of a regular n-gon with sides of length 1; (b) the center O of the n-gon
lies outside of P ; and (c) for every integer k with 0 < k n2 , the quantity 2k
n
is the measure
of exactly one AOB, where A and B are vertices of P . Find the number of non-congruent
peculiar polygons.
6/1/22. There are 50 people (numbered 1 to 50) and 50 identically wrapped presents around
a table at a party. Each present contains an integer dollar amount from $1 to $50, and no
two forpresents
Create PDF with GO2PDF free, if you wishcontain
to remove this the same
line, click amount.
here to buy Each person is randomly given one of the presents.
Virtual PDF Printer
Beginning with player #1, each player in turn does one of the following:
1. Opens his present and shows everyone the contents; or
2. If another player at the table has an open present, the player whose turn it is may
swap presents with that player, and leave the table with the open present. The other
player then immediately opens his new present and shows everyone the contents.
For example, the game could begin as follows:
Player #1 opens his present. (The game must always begin this way, as there are no
open presents with which to swap.)
Player #2 decides to swap her present with Player #1. Player #2 takes the money
from her newly acquired present and leaves the table. Player #1 opens his new present
(which used to belong to Player #2).
Player #3 opens her present. (Now Players #1 and #3 have open presents, and Player
#2 is still away from the table.)
Player #4 decides to swap his present with Player #1. Player #4 takes the money
from his newly acquired present and leaves the table. Player #1 opens his new present
The game ends after all the presents are opened, and all players keep the money in their
currently held presents.
Suppose each player follows a strategy that maximizes the expected value that the player
keeps at the end of the game.
(a) Find, with proof, the strategy each player follows. That is, describe when each player
will choose to swap presents with someone, or keep her original present.
(b) What is the expected number of swaps?


Round 2 Problems
www.usamts.org
1 point.
below:
tions.
above).

Round 2 Problems
www.usamts.org
1/2/22. Show that there is a unique way to place positive integers in the grid
to the right following these three rules:
1. Each entry in the top row is one digit.
2. Each entry in any row below the top row is the sum of the two
entries immediately above it.
3. Each pair of same-color squares contain the same integer. These five distinct integers
are used exactly twice and no other integer is used more than once.
(You can find a larger labeled version of this diagram on the back page.)
2/2/22. A sequence is called tworrific if its first term is 1 and the sum of every pair of
consecutive terms is a positive power of 2. One example of a tworrific sequence is 1, 7, 5,
7, 57.
(a) Find the shortest possible length of a tworrific sequence that contains the term 2011.
(b) Find the number of tworrific sequences that contain the term 2011 and have this
shortest possible length.
3/2/22. Richard, six of his friends, and a Gortha beast are standing at different vertices of a
cube-shaped planet. Richard has a potato and is a neighbor to the Gortha. On each turn,
whoever has the potato throws it at random to one of his three neighbors. If the Gortha
gets the potato he eats it. What is the probability that Richard is the one who feeds the
Gortha?
4/2/22. Let A, B, C, and D be points in the plane such that AD k BC. Let I be the incenter of
4ABC and assume that I is also the orthocenter of 4DBC. Show that AB +AC = 2BC.
5/2/22. Zara and Ada are playing a game. Ada begins by picking an integer from 1 to 2011
(inclusive). On each turn Zara tries to guess Adas number. Ada then tells Zara whether her
guess is too high, too low, or correct. If Zaras guess is not correct, Ada adds or subtracts 1
from her number (always constructing a new number from 1 to 2011). Assuming Zara plays
optimally, what is the minimum number of turns she needs to guarantee that she will guess
Adas number?
6/2/22. The roving rational robot rolls along the rational number line. On each turn, if the
robot is at pq , he selects a positive integer n and rolls to p+nq
q+np
. The robot begins at the
rational number 2011. Can the roving rational robot ever reach the rational number 2?
Round 2 Problems
www.usamts.org



Round 1 Problems
www.usamts.org
1 point.
4. Submit your solutions by Tuesday, October 11, 2011, via one (and only one!) of the
methods below:
tions.
above).

Round 1 Problems
www.usamts.org
1/1/23. The grid on the right has 12 boxes and 15 edges connect-
ing boxes. In each box, place one of the six integers from 1 to 6
such that the following conditions hold:
For each possible pair of distinct numbers from 1 to
6, there is exactly one edge connecting two boxes
with that pair of numbers.
If an edge has an arrow, then it points from a box
with a smaller number to a box with a larger number.
You do not need to prove that your configuration is the only one possible; you merely need
to find a configuration that satisfies the constraints above. (Note: In any other USAMTS
problem, you need to provide a full proof. Only in this problem is an answer without
justification acceptable.)
2/1/23. Find all integers a, b, c, d, and e, such that
a2 = a + b 2c + 2d + e 8,
b2 = a 2b c + 2d + 2e 6,
c2 = 3a + 2b + c + 2d + 2e 31,
d2 = 2a + b + c + 2d + 2e 2,
e2 = a + 2b + 3c + 2d + e 8.
3/1/23. (Corrected from an earlier release.) You have 14 coins, dated 1901 through 1914.
Seven of these coins are real and weigh 1.000 ounce each. The other seven are counterfeit
and weigh 0.999 ounces each. You do not know which coins are real or counterfeit. You also
cannot tell which coins are real by look or feel.
Fortunately for you, Zoltar the Fortune-Weighing Robot is capable of making very precise
measurements. You may place any number of coins in each of Zoltars two hands and Zoltar
will do the following:
If the weights in each hand are equal, Zoltar tells you so and returns all of the coins.
If the weight in one hand is heavier than the weight in the other, then Zoltar takes one
coin, at random, from the heavier hand as tribute. Then Zoltar tells you which hand
was heavier, and returns the remaining coins to you.1
1
In the earlier version, this sentence read, Then Zoltar tells you the result of the measurement, and
returns the remaining coins to you. The correction clarifies that the result of the measurement was
meant to refer only to which hand was heavier, not to the actual weight in either hand.
Round 1 Problems
www.usamts.org
Your objective is to identify a single real coin that Zoltar has not taken as tribute. Is there
a strategy that guarantees this? If so, then describe the strategy and why it works. If not,
then prove that no such strategy exists.
4/1/23. Let ABCDEF and ABC 0 D0 E 0 F 0 be regular planar hexagons in three-dimensional

space with side length 1, such that EAE 0 = 60 . Let P be the convex polyhedron whose
vertices are A, B, C, C 0 , D, D0 , E, E 0 , F , and F 0 .
(a) Find the radius r of the largest sphere that can be enclosed in polyhedron P.
(b) Let S be a sphere enclosed in polyhedron P with radius r (as derived in part (a)). The
set of possible centers of S is a line segment XY . Find the length XY .
5/1/23. In the game of Tristack Solitaire, you start with three stacks of cards, each with a
different positive integer number of cards. At any time, you can double the number of cards
in any one stack of cards by moving cards from exactly one other, larger, stack of cards to
the stack you double. You win the game when any two of the three stacks have the same
number of cards.
For example, if you start with stacks of 3, 5, and 7 cards, then you have three possible
legal moves:
You may move 3 cards from the 5-card stack to the 3-card stack, leaving stacks of 6,
2, and 7 cards.
5, and 4 cards.
10, and 2 cards.
Can you win Tristack Solitaire from any starting position? If so, then give a strategy for
winning. If not, then explain why.


Round 2 Problems
www.usamts.org
1 point.
4. Submit your solutions by Monday, November 14, 2011, via one (and only one!) of the
methods below:
Deadline: 3 PM Eastern / Noon Pacific on November 14, 2011.
tions.
above).

Round 2 Problems
www.usamts.org
1/2/23. Find all the ways of placing the integers 1, 2, 3, . . . , 16 in the boxes below, such that
each integer appears in exactly one box, and the sum of every pair of neighboring integers
is a perfect square.
2/2/23. Four siblings are sitting down to eat some mashed potatoes for lunch: Ethan
has 1 ounce of mashed potatoes, Macey has 2 ounces, Liana has 4 ounces, and Samuel
has 8 ounces. This is not fair. A blend consists of choosing any two children at random,
combining their plates of mashed potatoes, and then giving each of those two children half
of the combination. After the childrens father performs four blends consecutively, what is
the probability that the four children will all have the same amount of mashed potatoes?
3/2/23. Find all integers b such that there exists a positive real number x with
1 1 1
= + .
b b2xc b5xc
Here byc denotes the greatest integer that is less than or equal to y.
4/2/23. A luns with vertices X and Y is a region bounded by two circular arcs meetingat the

endpoints
X and Y . Let A, B, and V be points such that AV B = 75 , AV = 2 and
BV = 3. Let L be the largest area luns with vertices A and B that does not intersect the

lines V A or V B in any points other than A and B. Define k as the area of L. Find the value
k
.
(1 + 3)2
5/2/23. Miss Levans has 169 students in her history class and wants to seat them all in a
13 13 grid of desks. Each desk is placed at a different vertex of a 12 meter by 12 meter
square grid of points she has marked on the floor. The distance between neighboring vertices
is exactly 1 meter.
Each student has at most three best friends in the class. Best-friendship is mutual: if
Lisa is one of Shannons best friends, then Shannon is also one of Lisas best friends. Miss
Levans knows that if any two best friends sit at points that are 3 meters or less from each
other then they will be disruptive and nobody will learn any history. And that is bad.
Prove that Miss Levans can indeed place all 169 students in her class without any such
disruptive pairs.


Round 3 Problems
www.usamts.org
just send a numerical answer for a problem other than Problem 1 with no proof, you
will get no more than 1 point.
4. Submit your solutions by Tuesday, January 17, 2012, via one (and only one!) of the
methods below:
Deadline: 3 PM Eastern / Noon Pacific on January 17, 2012.
tions.
above).

Round 3 Problems
www.usamts.org
1/3/23. Fill in the circles to the right with the numbers 1

1 through 16 so that each number is used once (the
8 2
number 1 has been filled in already). The number in 11
any non-circular region is equal to the greatest differ- 7
ence between any two numbers in the circles on that

regions vertices. 2
You do not need to prove that your configuration is 10
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer 2
the only one possible; you merely need to find a valid
configuration. (Note: In any other USAMTS problem,
you need to provide a full proof. Only in this problem 4
4
is an answer without justification acceptable.) 8 5
2/3/23. Let x be a complex number such that x2011 = 1 and x 6= 1. Compute the sum
x2 x4 x6 x4020
+ + + + .
x1 x2 1 x3 1 x2010 1
3/3/23. A long, 1-inch wide strip of cloth can be folded into the figure below.
When the cloth is pulled tight and flattened, the result is a knot with two trailing strands.
The knot has outer boundary equal to a regular pentagon as shown below.
Instead, a long 1-inch wide strip of cloth is folded into the next figure, following the given
turns and crossings.
Round 3 Problems
www.usamts.org
The knot has outer boundary equal to a regular heptagon. The trailing strands of the
heptagonal knot are both cut at the outer (heptagonal) boundary of the knot. Then the
knot is untied. What is the area of one side of the resulting quadrilateral of cloth? (Your
answer may contain trigonometric expressions.)
4/3/23. Renata the robot packs boxes in a warehouse. Each box is a cube of side length 1
foot. The warehouse floor is a square, 12 feet on each side, and is divided into a 12-by-12
grid of square tiles 1 foot on a side. Each tile can either support one box or be empty. The
warehouse has exactly one door, which opens onto one of the corner tiles.
Renata fits on a tile and can roll between tiles that share a side. To access a box, Renata
must be able to roll along a path of empty tiles starting at the door and ending at a tile
sharing a side with that box.
(a) Show how Renata can pack 91 boxes into the warehouse and still be able to access any
box.
(b) Show that Renata cannot pack 95 boxes into the warehouse and still be able to access
any box.
5/3/23. Let k > 2 be a positive integer. Elise and Xavier play a game that has four steps, in
this order.
1. Elise picks 2 nonzero digits (1-9), called e and f .
2. Xavier then picks k nonzero digits (1-9), called x1 , . . . , xk .
3. Elise picks any positive integer d.
4. Xavier picks an integer b > 10.
Each players choices are known to the other player when the choices are made.
The winner is determined as follows. Elise writes down the two-digit base b number efb .
Next, Xavier writes the k-digit base b number that is constructed by concatenating his digits,
(x1 . . . xk )b .
They then compute the greatest common divisor (gcd) of these two numbers. If this gcd is
greater than or equal to the integer d then Xavier wins. Otherwise Elise wins.
(As an example game for k = 3, Elise chooses the digits (e, f ) = (2, 4), Xavier chooses
(4, 4, 8), and then Elise picks d = 100. Xavier picks base b = 25. The base-25 numbers 2425
and 44825 are, respectively, equal to 54 and 2608. The greatest common divisor of these two
is 2, which is much less than 100, so Elise wins handily.)
Find all k for which Xavier can force a win, no matter how Elise plays.


Round 1 Problems
www.usamts.org
just send a numerical answer with no proof for a problem other than Problem 3, you
4. Submit your solutions by Monday, October 22, 2012, via one (and only one!) of the
methods below:
tions.
above).

Round 1 Problems
www.usamts.org

1/1/24. Several children were playing in the ugly tree when suddenly they all fell.
Roger hit branches A, B, and C in that order on the way down.
Sue hit branches D, E, and F in that order on the way down.
Gillian hit branches G, A, and C in that order on the way down.
Marcellus hit branches B, D, and H in that order on the way down.

Juan-Phillipe hit branches I, C, and E in that order on the way down.
Poor Mikey hit every branch A through I on the way down. Given only this information, in
how many different orders could he have hit these 9 branches on the way down?
2/1/24. Three wooden equilateral triangles of side length 18 inches

are placed on axles as shown in the diagram to the right. Each axle
is 30 inches from the other two axles. A 144-inch leather band is
wrapped around the wooden triangles, and a dot at the top corner is
painted as shown. The three triangles are then rotated at the same
speed and the band rotates without slipping or stretching. Compute
the length of the path that the dot travels before it returns to its
initial position at the top corner. 30
3/1/24. The symmetric difference, 4, of a pair of sets is the set of

elements in exactly one set. For example,
{1, 2, 3}4{2, 3, 4} = {1, 4}.
There are fifteen nonempty subsets of {1, 2, 3, 4}. Assign each subset
to exactly one of the squares in the grid to the right so that the following
conditions are satisfied.
(i) If A and B are in squares connected by a solid line then A4B has exactly one element.
(ii) If A and B are in squares connected by a dashed line then the largest element of A is
equal to the largest element of B.
Round 1 Problems
www.usamts.org
4/1/24. Let bxc denote the greatest integer less than or equal to x. Let m be a positive integer,
m 3. For every integer i with 1 i m, let
m
2 1 mi
Sm,i = n2 + 1 : n = 1, 2, 3, . . . .
2i1
For example, for m = 3,
S3,1 = {b7n 3c : n = 1, 2, 3, . . . }
= {4, 11, 18, . . . },

7
S3,2 = n 1 : n = 1, 2, 3, . . .
2
= {2, 6, 9, . . . },

7
S3,3 = n : n = 1, 2, 3, . . .
4
= {1, 3, 5, . . . }.
Prove that for all m 3, each positive integer occurs in exactly one of the sets Sm,i .
5/1/24. An ordered quadruple (y1 , y2 , y3 , y4 ) is quadratic if there exist real

numbers a, b, and c such that
yn = an2 + bn + c
for n = 1, 2, 3, 4.
Prove that if 16 numbers are placed in a 4 4 grid such that all four rows are quadratic
and the first three columns are also quadratic then the fourth column must also be quadratic.
(We say that a row is quadratic if its entries, in order, are quadratic. We say the same
for a column.)
Round 1 Problems
www.usamts.org
Larger diagram for Problem 3/1/24.


Round 2 Problems
www.usamts.org
4. Submit your solutions by Monday, November 26, 2012, via one (and only one!) of the
methods below:
tions.
above).

Round 2 Problems
www.usamts.org
1/2/24. Fill in each of the ten boxes with a 3-digit num-

0 1 2
ber so that the following conditions are satisfied.
1. Every number has three distinct digits that sum
to 15. 0 may not be a leading digit. One digit
3 4 5 6
of each number has been given to you.
2. No two numbers in any pair of boxes use the
same
Create PDF with GO2PDF for free, three
if you wish digits.
to remove Forhereexample,
this line, click itPrinter
to buy Virtual PDF is not allowed
for two different boxes to have the numbers 456 7 8 9
and 645.
3. Two boxes joined by an arrow must have two numbers that share an equal hundreds
digit, tens digit, or ones digit. Also, the smaller number must point to the larger.
2/2/24. Find all triples (a, b, c) of positive integers with a b c such that

1 1 1
1+ 1+ 1+ = 3.
a b c
3/2/24. Let f (x) = x x1 , and define f 1 (x) = f (x) and f n (x) = f (f n1 (x)) for n 2. For each
n, there is a minimal degree dn such that there exist polynomials p and q with f n (x) = p(x) q(x)
and the degree of q is equal to dn . Find dn .
4/2/24. Let n be a positive integer. Consider an n n grid of unit squares. How

many ways are there to partition the horizontal and vertical unit segments of
the grid into n(n + 1) pairs so that the following properties are satisfied?
(i) Each pair consists of a horizontal segment and a vertical segment that
share a common endpoint, and no segment is in more than one pair.
(ii) No two pairs of the partition contain four segments that all share the same endpoint.
(Pictured above is an example of a valid partition for n = 2.)
5/2/24. (Corrected from an earlier release.) A unit square ABCD is given in the plane,
with O being the intersection of its diagonals. A ray ` is drawn from O. Let X be the unique
point on ` such that AX + CX = 2, and let Y be the point on ` such that BY + DY = 2.
Let Z be the midpoint of XY , with Z = X if X and Y coincide. Find, with proof, the
minimum value of the length of OZ. 1
1
In the earlier version, the final sentence read Find, with proof, the minimum value of |OZ|, which was
inconsistent with how the rest of the problem statement indicated lengths of segments.
Round 2 Problems
www.usamts.org
0 1 2
3 4 5 6
7 8 9


Round 3 Problems
www.usamts.org
below:
Deadline: 3 PM Eastern / Noon Pacific on January 7, 2013
tions.
above).

Round 3 Problems
www.usamts.org

1/3/24. (Corrected from an earlier release.) In the 8 8 grid 24 1 3 20 13 11
shown, fill in 12 of the grid cells with the numbers 112 so that
3
the following conditions are satisfied:
18
1. Each cell contains at most one number, and each
number from 112 is used exactly once. 21
2. Two cells that both contain numbers may not touch,
even at a point. 20
3. A clue outside the grid pointing at a row or column gives 13
the sum of all of the numbers in that row or column.
1
Rows and columns without clues have an unknown sum.
2/3/24. Palmer and James work at a dice factory, placing dots on dice. Palmer builds his
dice correctly, placing the dots so that 1, 2, 3, 4, 5, and 6 dots are on separate faces. In a
fit of mischief, James places his 21 dots on a die in a peculiar order, putting some nonneg-
ative integer number of dots on each face, but not necessarily in the correct configuration.
Regardless of the configuration of dots, both dice are unweighted and have equal probability
of showing each face after being rolled.
Then Palmer and James play a game. Palmer rolls one of his normal dice and James
rolls his peculiar die. If they tie, they roll again. Otherwise the person with the larger roll
is the winner. What is the maximum probability that James wins? Give one example of a
peculiar die that attains this maximum probability.
3/3/24. In quadrilateral ABCD, DAB = ABC = 110 , BCD = 35 , CDA = 105 ,
and AC bisects DAB. Find ABD.
4/3/24. (Corrected from an earlier release.) Denote by bxc the greatest integer2
less than or equal to x. Let m 2 be an integer, and let s be a real number between
0 and 1. Define an infinite sequence of real numbers a1 , a2 , a3 , . . . by setting a1 = s and
ak = mak1 (m 1)bak1 c for all k 2. For example, if m = 3 and s = 85 , then we get
a1 = 85 , a2 = 15
8
, a3 = 29
8
, a4 = 39
8
, and so on.
Call the sequence a1 , a2 , a3 , . . . orderly if we can find rational numbers b, c such that
5
ban c = bbn + cc for all n 1. With the example above where m = 3 and s = 8 , we get an
3n 3
orderly sequence since ban c = 2 2 for all n. Show that if s is an irrational number and
m 2 is any integer, then the sequence a1 , a2 , a3 , . . . is not an orderly sequence.
5/3/24. Let P and Q be two polynomials with real coefficients such that P has degree greater
than 1 and
P (Q(x)) = P (P (x)) + P (x).
Show that P (x) = P (x) + x.
Round 3 Problems
www.usamts.org
24 1 3 20 13 11
3
18
21
20
13
3
Notes
1
The second sentence about rows and columns without clues was not in the original
version. It was added to make the problem clearer.
2
The original version erroneously said greatest positive integer less than or equal to x.


Round 1 Problems
www.usamts.org
just send a numerical answer with no proof for a problem other than Problem 2 , you
below:
tions.
above).

Round 1 Problems
www.usamts.org
1/1/25. Alex is trying to open a lock whose code is a sequence that is three letters long, with
each of the letters being one of A, B or C, possibly repeated. The lock has three buttons,
labeled A, B and C. When the most recent 3 button-presses form the code, the lock opens.
What is the minimum number of total button presses Alex needs to guarantee opening the
lock?
2/1/25. In the 5 6 grid shown, fill in all of the grid cells with
the digits 09 so that the following conditions are satisfied:
7
8 6
1. Each digit gets used exactly 3 times.
2. No digit is greater than the digit directly above it. 2 4
3. In any four cells that form a 2 2 subgrid, 5 1
the sum of the four digits must be a multiple of 3.
3
You do not need to prove that your configuration is the only
one possible; you merely need to find a configuration that works. (Note: In any other
USAMTS problem, you need to provide a full proof. Only in this problem is an answer
without justification acceptable.)
3/1/25. An infinite sequence of positive real numbers a1 , a2 , a3 , . . . is called territorial if for

all positive integers i, j with i < j, we have |ai aj | 1j . Can we find a territorial sequence
a1 , a2 , a3 , . . . for which there exists a real number c with ai < c for all i?
4/1/25. Bunbury the bunny is hopping on the positive integers. First, he is told
a positive integer n. Then Bunbury chooses positive integers a, d and hops on
all of the spaces a, a + d, a + 2d, . . . , a + 2013d. However, Bunbury must make
these choices so that the number of every space that he hops on is less than n and relatively
prime to n.
A positive integer n is called bunny-unfriendly if, when given that n, Bunbury is unable
to find positive integers a, d that allow him to perform the hops he wants. Find the maximum
bunny-unfriendly integer, or prove that no such maximum exists.
5/1/25. Niki and Kyle play a triangle game. Niki first draws 4ABC with area 1, and Kyle
picks a point X inside 4ABC. Niki then draws segments DG, EH, and F I, all through
X, such that D and E are on BC, F and G are on AC, and H and I are on AB. The ten
points must all be distinct. Finally, let S be the sum of the areas of triangles DEX, F GX,
and HIX. Kyle earns S points, and Niki earns 1 S points. If both players play optimally
to maximize the amount of points they get, who will win and by how much?


Round 2 Problems
www.usamts.org
below:
tions.
above).

Round 2 Problems
www.usamts.org
1/2/25. In the 3 5 grid shown, fill in each empty box with 9

a two-digit positive integer so that:
(a) no number appears in more than one box, and
(b) for each of the 9 lines in the grid consisting of three
boxes connected by line segments, the box in the mid-
dle of the line contains the least common multiple of the 4
numbers in the other two boxes on the line.
You do not need to prove that your answer is the only one possible; you merely need to
find an answer that satisfies the constraints above. (Note: In any other USAMTS problem,
you need to provide a full proof. Only in this problem is an answer without justification
acceptable.)
2/2/25. Let ABCD be a quadrilateral with AB k CD, AB = 16, CD = 12, and BC < AD.
A circle with diameter 12 is inside of ABCD and tangent to all four sides. Find BC.
3/2/25. For each positive integer n 2, find a polynomial Pn (x) with rational coefficients
1
such that Pn n 2 = . (Note that n
2 denotes the positive nth root of 2.)
1+ n2
4/2/25. An infinite sequence of real numbers a1 , a2 , a3 , . . . is called spooky if a1 = 1 and for

all integers n > 1,
na1 + (n 1)a2 + (n 2)a3 + + 2an1 + an < 0,
n2 a1 + (n 1)2 a2 + (n 2)2 a3 + + 22 an1 + an > 0.
Given any spooky sequence a1 , a2 , a3 , . . ., prove that
20133 a1 + 20123 a2 + 20113 a3 + + 23 a2012 + a2013 < 12345.
5/2/25. Let S be a planar region. A domino-tiling of S is a partition of S into

1 2 rectangles. (For example, a 2 3 rectangle has exactly 3 domino-tilings, as
shown to the right.) The rectangles in the partition of S are called dominoes.
(a) For any given positive integer n, find a region Sn with area at most 2n
that has exactly n domino-tilings.
(b) Find a region T with area less than 50000 that has exactly 100002013
domino-tilings.


Round 3 Problems
www.usamts.org
just send a numerical answer with no proof for a problem other than Problem 1 , you
below:
tions.
above).

Round 3 Problems
www.usamts.org
1/3/25. In the grid shown, fill in each empty space with a

number, such that after the grid is completely filled in, 4 1
the number in each space is equal to the smallest posi- 1 1 3
tive integer that does not appear in any of the touching
spaces. (A pair of spaces is considered to touch if they 2 6
both share a vertex.) 7
You do not need to prove that your configuration is
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer 3 5
the only one possible; you merely need to find a con-
figuration that satisfies the constraints above. (Note:
2 4 3
In any other USAMTS problem, you need to provide a 3 2
full proof. Only in this problem is an answer without
2/3/25. Let a1 , a2 , a3 , . . . be a sequence of positive real numbers such that ak ak+2 = ak+1 + 1
for all positive integers k. If a1 and a2 are positive integers, find the maximum possible value
of a2014 .
3/3/25. Let A1 A2 A3 . . . A20 be a 20-sided polygon A7 A5

P in the plane, where all of the side lengths of P
A6
are equal, the interior angle at Ai measures 108 A9 A3
A8 A4
degrees for all odd i, and the interior angle at Ai
measures 216 degrees for all even i. Prove that
the lines A2 A8 , A4 A10 , A5 A13 , A6 A16 , and A7 A19 A10 A2
all intersect at the same point. A11 A1
A12 A20
A13 A14 A18 A19

A16
A15 A17
Round 3 Problems
www.usamts.org
4/3/25. An infinite sequence (a0 , a1 , a2 , . . .) of positive integers is called a ribbon if the sum
of any eight consecutive terms is at most 16; that is, for all i 0,
ai + ai+1 + + ai+7 16.
A positive integer m is called a cut size if every ribbon contains a set of consecutive elements
that sum to m; that is, given any ribbon (a0 , a1 , a2 , . . .), there exist nonnegative integers
k ` such that
ak + ak+1 + + a` = m.
Find, with proof, all cut sizes, or prove that none exist.
5/3/25. For any positive integer b 2, we write the base-b numbers as follows:
(dk dk1 . . . d0 )b = dk bk + dk1 bk1 + + d1 b1 + d0 b0 ,
where each digit di is a member of the set S = {0, 1, 2, . . . , b 1} and either dk 6= 0 or k = 0.

There is a unique way to write any nonnegative integer in the above form.
If we select the digits from a different set S instead, we may obtain new representations
of all positive integers or, in some cases, all integers. For example, if b = 3 and the digits are
selected from S = {1, 0, 1}, we obtain a way to uniquely represent all integers, known as the
balanced ternary representation. As further examples, the balanced ternary representation
of the numbers 5, 3, and 25 are:
5 = (1 1 1)3 , 3 = (1 0)3 , 25 = (1 0 1 1)3 .
However, not all digit sets can represent all integers. If b = 3 and S = {2, 0, 2}, then no
odd number can be represented. Also, if b = 3 and S = {0, 1, 2} as in the usual base-3
representation, then no negative number can be represented.
Given a set S of four integers, one of which is 0, call S a 4-basis if every integer n has at
least one representation in the form
n = (dk dk1 . . . d0 )4 = dk 4k + dk1 4k1 + + d1 41 + d0 40 ,
where dk , dk1 , . . . , d0 are all elements of S and either dk 6= 0 or k = 0.
(a) Show that there are infinitely many integers a such that {1, 0, 1, 4a+2} is not a 4-basis.
(b) Show that there are infinitely many integers a such that {1, 0, 1, 4a + 2} is a 4-basis.


Round 1 Problems
www.usamts.org
below:
Deadline: 3 PM Eastern / Noon Pacific on November 5, 2014.
(b) Mail: USAMTS
PO Box 4499
New York, NY 10163
(Solutions must be postmarked on or before November 5, 2014.)
tions.
above).

Round 1 Problems
www.usamts.org

1/1/26. Divide the grid shown to the right into more than one 6 5 6
region so that the following rules are satisfied.
4 2 4
1. Each unit square lies entirely within exactly 1 region.
3 3 4
2. Each region is a single piece connected by the edges of 4
its unit squares. 4 3
3. Each region contains the same number of whole unit 4 4 4
squares.
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer 1 1 1
4. Each region contains the same sum of numbers.
to find a configuration that works. (Note: In any other USAMTS problem, you need to
provide a full proof. Only in this problem is an answer without justification acceptable.)
2/1/26. Find all triples (x, y, z) such that x, y, z, x y, y z, x z are all prime positive
integers.
3/1/26. A group of people is lined up in almost-order if, whenever person A is to the left
of person B in the line, A is not more than 8 centimeters taller than B. For example, five
people with heights 160, 165, 170, 175, and 180 centimeters could line up in almost-order
with heights (from left-to-right) of 160, 170, 165, 180, 175 centimeters.
(a) How many different ways are there to line up 10 people in almost-order if their heights
are 140, 145, 150, 155, 160, 165, 170, 175, 180, and 185 centimeters?
(b) How many different ways are there to line up 20 people in almost-order if their heights
are 120, 125, 130, 135, 140, 145, 150, 155, 160, 164, 165, 170, 175, 180, 185, 190, 195,
200, 205, and 210 centimeters? (Note that there is someone of height 164 centimeters.)
4/1/26. Let P and Q be two circles of radius 1, intersecting in points A and B. Let P
and Q be two regular n-gons (for some positive integer n 4) inscribed in P and Q ,
respectively, such that A and B are vertices of both P and Q. Suppose a third circle of
radius 1 intersects P at two of its vertices C, D and intersects Q at two of its vertices E, F .
Further assume that A, B, C, D, E, F are all distinct points, that A lies outside of , and
that B lies inside . Show that there exists a regular 2n-gon that contains C, D, E, F as
four of its vertices.
5/1/26. Let a0 , a1 , a2 , . . . be a sequence of nonnegative integers such that a2 = 5, a2014 = 2015,
and an = aan1 for all positive integers n. Find all possible values of a2015 .

Please visit www.usamts.org for details about solution submission.

Round 2 Problems
www.usamts.org
4. Submit your solutions by December 8, 2014, via one (and only one!) of the methods
below:
Deadline: 3 PM Eastern / Noon Pacific on December 8, 2014.
(b) Mail: USAMTS
PO Box 4499
New York, NY 10163
(Solutions must be postmarked on or before December 8, 2014.)
tions.
above).

Round 2 Problems
www.usamts.org
1/2/26. The net of 20 triangles shown to the right can be folded to

form a regular icosahedron. Inside each of the triangular faces,
write a number from 1 to 20 with each number used exactly once. 2
Any pair of numbers that are consecutive must be written on 6
faces sharing an edge in the folded icosahedron, and additionally, 15
14
1 and 20 must also be on faces sharing an edge. Some numbers
have been given to you.
Youfordo
Create PDF with GO2PDF free, ifnot
you wishneed
to removeto prove
this line, click herethat your
to buy Virtual answer is the only one
PDF Printer
possible; you merely need to find an answer that satisfies the

constraints above. (Note: In any other USAMTS problem, you
need to provide a full proof. Only in this problem is an answer
2/2/26. Let a, b, c, x, y be positive real numbers such that
ax + by bx + cy cx + ay.
Prove that b c.
3/2/26. Let P be a square pyramid whose base consists of the four vertices (0, 0, 0), (3, 0, 0),
(3, 3, 0), and (0, 3, 0), and whose apex is the point (1, 1, 3). Let Q be a square pyramid whose
base is the same as the base of P, and whose apex is the point (2, 2, 3). Find the volume of
the intersection of the interiors of P and Q.
4/2/26. A point P in the interior of a convex polyhedron in Euclidean space is called a

pivot point of the polyhedron if every line through P contains exactly 0 or 2 vertices of the
polyhedron. Determine, with proof, the maximum number of pivot points that a polyhedron
can contain.
5/2/26. Find the smallest positive integer n that satisfies the following: We can color each
positive integer with one of n colors such that the equation
w + 6x = 2y + 3z
has no solutions in positive integers with all of w, x, y, z the same color. (Note that w, x, y, z
need not be distinct: for example, 5 and 7 must be different colors because (w, x, y, z) =
(5, 5, 7, 7) is a solution to the equation.)
Round 2 Solutions must be submitted by December 8, 2014.


Round 3 Problems
www.usamts.org
below:
(b) Mail: USAMTS
PO Box 4499
New York, NY 10163
tions.
above).

Round 3 Problems
www.usamts.org
1/3/26. Fill in each blank unshaded cell

3 59
with a positive integer less than 100, such
that every consecutive group of unshaded
cells within a row or column is an arith-
10
metic sequence.
You do not need to prove that your an-
swer is the only one possible; you merely
needfor free,
Create PDF with GO2PDF to iffind
you wishan answer
to remove that
this line, click here tosatisfies
buy Virtual PDFthe
Printer
31 26 59
constraints above. (Note: In any other USAMTS problem, you need to provide a full proof.
Only in this problem is an answer without justification acceptable.)
2/3/26. Let A1 A2 A3 A4 A5 be a regular pentagon with side length 1. The

sides of the pentagon are extended to form the 10-sided polygon shown
in bold at right. Find the ratio of the area of quadrilateral A2 A5 B2 B5
(shaded in the picture to the right) to the area of the entire 10-sided
polygon.
3/3/26. Let a1 , a2 , a3 , . . . be a sequence of positive real numbers such that:
(i) For all positive integers m and n, we have amn = am an , and
(ii) There exists a positive real number B such that for all positive integers m and n with
m < n, we have am < Ban .
Find all possible values of log2015 (a2015 ) log2014 (a2014 ).
4/3/26. Nine distinct positive integers are arranged in a circle such that the product of any
two non-adjacent numbers in the circle is a multiple of n and the product of any two adjacent
numbers in the circle is not a multiple of n, where n is a fixed positive integer. Find the
smallest possible value for n.
5/3/26. A finite set S of unit squares is chosen out of a large grid of unit squares. The
squares of S are tiled with isosceles right triangles of hypotenuse 2 so that the triangles do
not overlap each other, do not extend past S, and all of S is fully covered by the triangles.
Additionally, the hypotenuse of each triangle lies along a grid line, and the vertices of the
triangles lie at the corners of the squares. Show that the number of triangles must be a
multiple of 4.


Round 1 Problems
www.usamts.org
below:
(b) Mail: USAMTS
P.O. Box 4499
New York, NY 10163
(Solutions must be postmarked on or before October 19. We strongly recom-
mend that you keep a copy of your solutions and that you pay for tracking on
the mailing. With large envelopes, there have been significant delays in the past.)
tions.
above).

Round 1 Problems
www.usamts.org

1/1/27. Fill in the spaces of the grid to the right with positive
integers so that in each 2 2 square with top left number a, top 3 9
right number b, bottom left number c, and bottom right number
d, either a + d = b + c or ad = bc.
11 7 2
10 16
You do not need to prove that your answer is the only one pos-
sible; you merely need to find an answer that satisfies the con- 15
straints above. (Note: In any other USAMTS problem, you need
to provide
Create PDF with GO2PDF for free, if youawish
full proof.
to remove Only
this line, into buy
click here this problem
Virtual PDF Printer is an answer without 20 36 32
2/1/27. Suppose a, b, and c are distinct positive real numbers such that
abc = 1000,
bc(1 a) + a(b + c) = 110.
If a < 1, show that 10 < c < 100.
3/1/27. Let P be a convex n-gon in the plane with vertices labeled V1 , . . . , Vn in counterclock-
wise order. A point Q not outside P is called a balancing point of P if, when the triangles
QV1 V2 , QV2 V3 , . . . , QVn1 Vn , QVn V1 are alternately colored blue and green, the total areas of
the blue and green regions are the same. Suppose P has exactly one balancing point. Show
that the balancing point must be a vertex of P .
4/1/27. Several players try out for the USAMTS basketball team, and they all have integer
heights and weights when measured in centimeters and pounds, respectively. In addition,
they all weigh less in pounds than they are tall in centimeters. All of the players weigh at
least 190 pounds and are at most 197 centimeters tall, and there is exactly one player with
every possible height-weight combination.
The USAMTS wants to field a competitive team, so there are some strict requirements.
(i) If person P is on the team, then anyone who is at least as tall and at most as heavy
as P must also be on the team.
(ii) If person P is on the team, then no one whose weight is the same as P s height can
also be on the team.
Assuming the USAMTS team can have any number of members (including zero), how many
different basketball teams can be constructed?
Round 1 Problems
www.usamts.org
5/1/27. Find all positive integers n that have distinct positive divisors d1 , d2 , . . . , dk , where
k > 1, that are in arithmetic progression and
n = d1 + d2 + + dk .
Note that d1 , d2 , . . . , dk do not have to be all the divisors of n.


Round 2 Problems
www.usamts.org
1. You must prove your answers on all problems (other than Problem 1). This
means showing all necessary steps to demonstrate that your result is true; however,
please leave out any work that is unnecessary to the proof. If you just send a numerical
answer with no proof for a problem other than Problem 1, you will get no more than
1 point.
below:
(b) Mail: USAMTS
P.O. Box 4499
New York, NY 10163
(Solutions must be postmarked on or before November 30. We strongly recom-
mend that you keep a copy of your solutions and that you pay for tracking on
the mailing. With large envelopes, there have been significant delays in the past.)
tions.
above).

Round 2 Problems
www.usamts.org

1/2/27. In the grid to the right, the shortest path through unit squares between the pair of
2s has length 2. Fill in some of the unit squares in the grid so that
(i) exactly half of the squares in each row and column 6

contain a number,
10 1
(ii) each of the numbers 1 through 12 appears exactly
twice, and
2 8
(iii) for n = 1, 2, . . . , 12, the shortest path between
5 2
the pair of ns has length exactly n. 7 9
3
acceptable.)
2/2/27. A net for a polyhedron is cut along an edge to give two pieces. For example, we may
cut a cube net along the red edge to form two pieces as shown.
Are there two distinct polyhedra for which this process may result in the same two pairs of
pieces? If you think the answer is no, prove that no pair of polyhedra can result in the same
two pairs of pieces. If you think the answer is yes, prove an example; a clear example will
suffice as proof.
3/2/27. For all positive integers n, show that

n n

1 X k k! k
= 1.
n k=1 nk
4/2/27. Find all polynomials P (x) with integer coefficients such that, for all integers a and b,
P (a + b) P (b) is a multiple of P (a).
Round 2 Problems
www.usamts.org
5/2/27. Let n > 1 be an even positive integer. A 2n 2n grid of unit squares is given, and it
is partitioned into n2 contiguous 2 2 blocks of unit squares. A subset S of the unit squares
satisfies the following properties:
(i) For any pair of squares A, B in S, there is a sequence of squares in S that starts with
A, ends with B, and has any two consecutive elements sharing a side; and
(ii) In each of the 2 2 blocks of squares, at least one of the four squares is in S.
An example for n = 2 is shown below, with the squares of S shaded and the four 2 2 blocks
of squares outlined in bold.
In terms of n, what is the minimum possible number of elements in S?


Round 3 Problems
www.usamts.org
1. You must prove your answers on all problems (other than Problem 1). This
means showing all necessary steps to demonstrate that your result is true; however,
please leave out any work that is unnecessary to the proof. If you just send a numerical
answer with no proof for a problem other than Problem 1, you will get no more than
1 point.
below:
Deadline: 3 PM Eastern / Noon Pacific on January 4, 2016
(b) Mail: USAMTS
P.O. Box 4499
New York, NY 10163
(Solutions must be postmarked on or before January 4, 2016. We strongly
recommend that you keep a copy of your solutions and that you pay for tracking
on the mailing. With large envelopes, there have been significant delays in the
past.)
tions.
above).

Round 3 Problems
www.usamts.org
1/3/27. Fill in each space of the grid with either a 0 or a 1 so that all 16 strings of four
consecutive numbers across and down are distinct.
0
1 0
0 1
1 0
0 Printer1
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF
0
0
0
1
acceptable.)
2/3/27. Fames is playing a computer game with falling two-dimensional blocks. The playing
field is 7 units wide and infinitely tall with a bottom border. Initially the entire field is
empty. Each turn, the computer gives Fames a 1 3 solid rectangular piece of three unit
squares. Fames must decide whether to orient the piece horizontally or vertically and which
column(s) the piece should occupy (3 consecutive columns for horizontal pieces, 1 column
for vertical pieces). Once he confirms his choice, the piece is dropped straight down into the
playing field in the selected columns, stopping all three of the pieces squares as soon as the
piece hits either the bottom of the playing field or any square from another piece. All of
the pieces must be contained completely inside the playing field after dropping and cannot
partially occupy columns.
If at any time a row of 7 spaces is all filled with squares, Fames scores a point.
Unfortunately, Fames is playing in invisible mode, which prevents him from seeing the state
of the playing field or how many points he has, and he has already arbitrarily dropped some
number of pieces without remembering what he did with them or how many there were.
For partial credit, find a strategy that will allow Fames to eventually earn at least one more
point. For full credit, find a strategy for which Fames can correctly announce I have earned
at least one more point and know that he is correct.
3/3/27. For n > 1, let an be the number of zeroes that n! ends with when written in base n.
Find the maximum value of ann .
Round 3 Problems
www.usamts.org
4/3/27. Let 4ABC be a triangle with AB < AC. Let the angle bisector of BAC meet BC
at D, and let M be the midpoint of BC. Let P be the foot of the perpendicular from B to
AD. Extend BP to meet AM at Q. Show that DQ is parallel to AB.
5/3/27. Let a1 , a2 , . . . , a100 be a sequence of integers. Initially, a1 = 1, a2 = 1 and the

remaining numbers are 0. After every second, we perform the following process on the
sequence: for i = 1, 2, . . . , 99, replace ai with ai + ai+1 , and replace a100 with a100 + a1 . (All
of this is done simultaneously, so each new term is the sum of two terms of the sequence
from before any replacements.) Show that for any integer M , there is some index i and some
time t for which |ai | > M at time t.


Round 1 Problems
www.usamts.org
below:
Deadline: 8 PM Eastern / 5 PM Pacific on October 17, 2016
(b) Mail: USAMTS
PO Box 4499
New York, NY 10163
Deadline: Solutions must be postmarked on or before October 17, 2016.
We strongly recommend that you keep a copy of your solutions and that you
pay for tracking on the mailing. With large envelopes, there have been significant
delays in the past.
tions.
by logging on to www.usamts.org and visiting the My USAMTS pages.
above).

Please read the entire rules at www.usamts.org.
Round 1 Problems
www.usamts.org

1/1/28. Fill in each cell of the grid with one of the numbers 1, 2 1
2, or 3. After all numbers are filled in, if a row, column, or any 3 2
diagonal has a number of cells equal to a multiple of 3, then it 2 3 2
must have the same amount of 1s, 2s, and 3s. (There are 10 2 1 1 3
such diagonals, and they are all marked in the grid by a gray 3 3 3
dashed line.) Some numbers have been given to you. 2 2 3
3 2 3 2 2 3
3 1
1 3
acceptable.)
2/1/28. A tower of height h is a stack of contiguous rows of squares

of height h such that
(i) the bottom row of the tower has h squares,
(ii) each row above the bottom row has one fewer square
than the row below it, and within each row the squares
are contiguous,
(iii) the squares in any given row all lie directly above
a square in the row below.
A tower is called balanced if when the squares of the tower are colored black and white in a
checkerboard fashion, the number of black squares is equal to the number of white squares.
For example, the figure above shows a tower of height 5 that is not balanced, since there are
7 white squares and 8 black squares.
How many balanced towers are there of height 2016?
3/1/28. Find all positive integers n for which (xn + y n + z n )/2 is a perfect square whenever
x, y, and z are integers such that x + y + z = 0.
4/1/28. Find all functions f (x) from nonnegative reals to nonnegative reals such that
f (f (x)) = x4 and f (x) Cx2 for some constant C.
5
5/1/28. Let ABCD be a convex quadrilateral with perimeter 2
and AC = BD = 1. Determine
the maximum possible area of ABCD.
Round 1 Problems
www.usamts.org
Problems by Evan Chen, Aaron Doman, Billy Swartworth, and USAMTS Staff.

Round 2 Problems
www.usamts.org
1. You must show your work and prove your answers for all problems. If you
below:
Deadline: 8 PM Eastern / 5 PM Pacific on November 28, 2016
(b) Mail: USAMTS
PO Box 4499
New York, NY 10163
(Solutions must be postmarked on or before November 28, 2016.)
tions.
above).

Round 2 Problems
www.usamts.org

1/2/28. Shade in some of the regions in the grid to the
right so that the shaded area is equal for each of the 11
rows and columns. Regions must be fully shaded or fully
unshaded, at least one region must be shaded, and the
area of shaded regions must be at most half of the whole
grid.
acceptable.)
2/2/28. Find all triples of three-digit positive integers x < y < z with x, y, z in arithmetic
progression and x, y, z + 1000 in geometric progression.
For this problem, you may use calculators or computers to gain an intuition about how to
solve the problem. However, your final submission should include mathematical derivations
or proofs and should not be a solution by exhaustive search.
3/2/28. Suppose m and n are relatively prime positive integers. A regular m-gon and a regular
n-gon are inscribed in a circle. Let d be the minimum distance in degrees (of the arc along
the circle) between a vertex of the m-gon and a vertex of the n-gon. What is the maximum
possible value of d?
Round 2 Problems
www.usamts.org
4/2/28. On Binary Island, residents communicate using special paper. Each sheet of paper is
a 1 n row of initially uncolored squares. To send a message, each square on the paper must
be colored either red or green. Unfortunately the paper on the island has become damaged,
and each sheet of paper has 10 random consecutive squares each of which is randomly colored
red or green.
Malmer and Weven would like to develop a scheme that allows them to send messages of
length 2016 between one another. They would like to be able to send any message of length
2016, and they want their scheme to work with perfect accuracy. What is the smallest value
of n for which they can develop such a strategy?
Note that when creating a message, one can see which 10 squares are randomly colored
and what colors they are. One also knows on which square the message begins, and on which
square the message ends.
5/2/28. Let n 4 and y1 , . . . , yn real with

n
X n
X n
X
yk = kyk = k 2 yk = 0
k=1 k=1 k=1
and
yk+3 3yk+2 + 3yk+1 yk = 0
for 1 k n 3. Prove that
n
X
k 3 yk = 0.
k=1
Problems by Aaron Doman, Mehtaab Sawhney, Billy Swartworth, and USAMTS Staff.

Round 3 Problems
www.usamts.org
below:
Deadline: 8 PM Eastern / 5 PM Pacific on January 3, 2017
(b) Mail: USAMTS
PO Box 4499
New York, NY 10163
(Solutions must be postmarked on or before January 3, 2017.)
tions.
above).

Round 3 Problems
www.usamts.org

1/3/28. Fill in each square of the grid with a number from 1 to 16, + 13 28 25 23
using each number exactly once. Numbers at the left or top give
13 7
the largest sum of two numbers in that row or column. Numbers
at the right or bottom give the largest difference of two numbers in 28
that row or column. 8
23 8

7 12 12
acceptable.)
2/3/28. Malmer Pebane, Fames Jung, and Weven Dare are perfect logicians that always tell
the truth. Malmer decides to pose a puzzle to his friends: he tells them that the day of his
birthday is at most the number of the month of his birthday. Then Malmer announces that
he will whisper the day of his birthday to Fames and the month of his birthday to Weven,
and he does exactly that.
After Malmer whispers to both of them, Fames thinks a bit, then says Weven cannot know
what Malmers birthday is.
After that, Weven thinks a bit, then says Fames also cannot know what Malmers birthday
is.
This exchange repeats, with Fames and Weven speaking alternately and each saying the other
cant know Malmers birthday. However, at one point, Weven instead announces Fames and
I can now know what Malmers birthday is. Interestingly, that was the longest conversation
like that we could have possibly had before both figuring out Malmers birthday.
Find Malmers birthday.

Round 3 Problems
www.usamts.org
3/3/28. An n-city is an n n grid of positive integers such that every entry greater than 1 is
the sum of an entry in the same row and an entry in the same column. Shown below is an
example 3-city.
1 1 2
2 3 1
6 4 1
(a) Construct a 5-city that includes some entry that is at least 150. (It is acceptable simply
to write the 5-city. You do not need to explain how you found it.)
(b) Show that for all n 1, the largest entry in an n-city is at most 3( 2 ) .
n
4/3/28. Let A1 , . . . , An and B1 , . . . , Bn be sets of points in the plane. Suppose that for all
points x,
D(x, A1 ) + D(x, A2 ) + + D(x, An ) D(x, B1 ) + D(x, B2 ) + + D(x, Bn ),
where D(x, y) denotes the distance between x and y. Show that the Ai s and the Bi s share
the same center of mass.
5/3/28. Consider the set S = {q + 1q , where q ranges over all positive rational numbers}.
(a) Let N be a positive integer. Show that N is the sum of two elements of S if and only
if N is the product of two elements of S.
(b) Show that there exist infinitely many positive integers N that cannot be written as the
sum of two elements of S.
(c) Show that there exist infinitely many positive integers N that can be written as the
Problems by Billy Swartworth, Remus Nicoara, and USAMTS Staff.


Round 3 Solutions
www.usamts.org
13 7
23 8

7 12 12
acceptable.)
Solution
+ 13 28 25 23
13 3 9 2 4 7
28 1 12 11 16
8 15 14 7 8
23 5 13 10 6 8

7 12 12
Round 3 Solutions
www.usamts.org
is.

Solution
We imagine that Fames and Weven are each given a number between 1 and 12 (representing
either the day or the month of Malmers birthday respectively).
We analyze what happens at the first step, when Fames claims Weven cannot know what
Malmers birthday is. Since Weven is given the month, and the day can be any number less
than or equal to the month, he could only know Malmers birthday if he were given a 1 (as
this would imply Malmers birthday was 1/1). So, Weven must not have a 1. In order for
Fames to know that Weven doesnt have a 1, Fames also must not have a 1. So, we conclude
that neither Fames nor Weven was given a 1.
Next, Weven claims that Fames also cannot know what Malmers birthday is. Since Fames
is given the day, and the month can be any number greater than or equal to the day, he could
only know Malmers birthday if he were given a 12 (as this would imply Malmers birthday
was 12/12). So, Fames must not have a 12. In order for Weven to know that Fames doesnt
have a 12, he also must not have a 12. So, we conclude that neither Fames nor Weven was
given a 12.
Similarly, if this exchange were to happen again, Famess statement would allow us to con-
clude that neither Fames nor Weven was given a 2, and Wevens statement would allow us
to conclude that neither Fames nor Weven was given an 11.
Round 3 Solutions
www.usamts.org
In order for this to continue for as long as possible, both players must continue to eliminate
numbers until Fames eliminates 6. Once this happens, both players know that only 7 is
possible, and Malmers birthday is (7, 7) (July 7th).
Round 3 Solutions
www.usamts.org
example 3-city.
1 1 2
2 3 1
6 4 1
n
Solution
(a) An example of a 5-city with an entry that is at least 150 is shown below. You may
have found a different example.

1 2 5 3 4
26 1 6 25 10

67 69 1 15 12

326 259 138 1 11
52 121 127 26 1
(b) An n-city must have at least one 1 in every row and column. So an n-city must have
at least n 1s, and hence can have at most n2 n = 2 n2 entries greater than 1. Let

us list these entries as a1 , a2 , . . . , a2(n) from least to greatest.

2
Notice that a1 can be at most 1 + 1 = 2. Then a2 can be at most a1 + 1 = 3, and a3

can be at most a2 + a1 = 3 + 2 = 5. In general, we see that ak is at most ak1 + ak2 .
Hence, we conclude that ak Fk+2 for all k, where Fm is the m-th Fibonacci number.
Therefore, we have a2(n) F2(n)+2 . So, it suffices to show that F2(n)+2 3( 2 ) .

n
2 2 2
To that end, we claim that F2m+2 3m for all m 0. For m = 0 both sides of the
inequality are 1. Then,
F2m+2 = F2m+1 + F2m
= 2F2m + F2m1
3F2m .
Hence, by induction F2m+2 3(3m1 ) = 3m . Thus,
a2(n) F2(n)+2 3( 2 ) .
n
2 2
Round 3 Solutions
www.usamts.org
So, every entry in an n-city is at most 3( 2 ) as desired.

n
Round 3 Solutions
www.usamts.org
points x,
Solution
b2
Lemma: If a > b, then a2 + b 2 a 2a
.
Proof of Lemma: We see that
b4
a2 + b 2 + a2 + b 2 ,
4a2
and both sides are positive. Taking the square root of both sides, we get
b2 + 2a2
a2 + b 2 .
2a
Subtracting a from both sides gives the desired claim.
Suppose that the centers of mass of the Ai and Bi are MA and MB respectively, and that
d is a real number with D(Ai , MA ) d and D(Bi , MB ) d for all i. Let D(MA , MB ) = `,
and suppose for the sake of contradiction that ` > 0. Choose a real number q such that
d2
q >d+ .
2`
Let P be the point on line MA MB with D(P, MA ) = q and D(P, MB ) = q + `. We claim
that
D(P, A1 ) + D(P, A2 ) + + D(P, An ) < D(P, B1 ) + D(P, B2 ) + + D(P, Bn ).
For each i, let A0i be the projection of Ai onto line MA MB , and similarly define Bi0 . We see
that P A0i Ai is a right triangle with hypotenuse P Ai , so applying the Pythagorean Theorem
gives us D(P, Ai )2 = D(P, A0i )2 + D(A0i , Ai )2 . Since A0i is the closest point to Ai on the line
MA MB , we have D(Ai , A0i ) D(Ai , MA ) d. So, D(P, Ai )2 D(P, A0i )2 + d2 . For brevity,
0
let ui = D(P,p Ai ) and vi = D(P, Ai )p for each i. We rewrite our inequality as u2i d2 + vi2 .
Thus, ui d2 + vi , and ui vi d2 + vi2 vi . Applying the Lemma gives us
2
d2
ui vi .
2vi
Round 3 Solutions
www.usamts.org
Applying the Triangle Inequality to triangle P A0i MA , we get

D(P, A0i ) + D(A0i , MA ) D(P, MA ). Since D(A0i , MA ) D(Ai , MA ) d, we have vi + d q.
2
Combining this inequality with the inequality q > d + d2` and subtracting d, we get
2
vi q d > d2` .
2 2
d
Applying the inequality vi > d2` to the right side of the inequality ui vi 2vi
, we get
Pn
ui vi < `. So, ui < vi + `. P
And since i=1 vi = nq by D(P, MA ) = q and the definition of
the center of mass, we have ni=1 ui < nq + n`.
We also have D(P, Bi ) D(P, Bi0 ), so
n
X n
X
D(P, Bi ) D(P, Bi0 ) = nD(P, MB ) = nq + n`.
i=1 i=1
Therefore,
n
X n
X
D(P, Ai ) < nq + n` D(P, Bi ).
i=1 i=1
This is a contradiction of the condition, so ` = 0 as desired.

Round 3 Solutions
www.usamts.org
Solution
(a) Note that the right side implies left side implication is true even without the require-
ment that N is an integer. Indeed, assume that N = (q + 1q ) (r + 1r ), with q, r rational.
By expanding the product, we obtain:

1 q r 1 1 q
N = qr + + + = x+ + y+ , where x = qr and y = are rational.
qr r q x y r
We now show that the converse is true. Let N be an integer such that
1 1
N = (x + ) + (y + ),
x y
where x, y are positive rational numbers. We can write x, y as irreducible fractions
x = ab and y = dc , with a, b, c, d positive integers satisfying gcd(a, b) = gcd(c, d) = 1.
After expanding the formula for N we obtain:
(a2 + b2 )cd + (c2 + d2 )ab
N= .
abcd
Since N is an integer, it follows that ab must divide the numerator. Thus, ab must
divide (a2 + b2 )cd. Since gcd(a, b) = 1, it follows that gcd(ab, a2 + b2 ) = 1. So, we
must have that ab divides cd. Similarly, by using the fact that cd must also divide the
numerator of the fraction, we obtain that cd divides ab. It follows ab = cd. Note now
that if we let p = cb and q = ac , then we have:

1 1 c b a c
p+ q+ = + +
p q b c c a
a b c2 ab
= + + + 2
b a ab c
a b c d
= + + +
b a d c
1 1
= x+ + y+
x y
= N.
Round 3 Solutions
www.usamts.org
c2
We used above the equality ab
= dc , which is equivalent to ab = cd.
Remark. Another way to finish the proof from ab = cd is to argue that there must
exist positive integers s, t, u, v such that a = st, b = uv, c = su, and d = tv. It follows
that N = ( vs + vs )( ut + ut ).
(b) We show that if N is an integer divisible by 8, then it is not possible to write N as the
sum of two elements of S. Equivalently, it suffices to show that N is not the product
of two elements of S. (It is somewhat faster to work with the product than the sum.)
Assume N = ( ab + ab ) ( dc + dc ), where a, b, c, d are positive integers such that ab and dc
are irreducible, positive fractions. After expanding the product, we obtain
(a2 + b2 ) (c2 + d2 )
= N.
abcd
Since N is divisible by 8, it follows that 8 must divide the numerator (a2 +b2 )(c2 +d2 ).
Thus, at least one of the two factors of the product must be divisible by 4. Assume for
instance that 4 divides a2 + b2 . By considering all the possibilities for a, b (mod 4), it
is easy to see that a, b must both be even. However, this contradicts the fact that the
fraction ab is irreducible.
Remark. Another option is to show that if N has a prime factor p with p 3
(mod 4), then N cannot be written as the product of two elements of S.
(c) We show that there exist infinitely many integers N of the form N = (a + a1 )(b + 1b ),
with a, b positive integers. Note that this statement is stronger than the statement of
(c). We have:
a2 + 1 b 2 + 1

1 1
a+ b+ = .
a b b a
We construct recursively infinitely many pairs of integers (a, b), with a < b and such
that a divides b2 + 1, and b divides a2 + 1. First, observe that the pair (1, 2) works.
2
Now, assuming that (a, b) satisfies the requirements, we show that (b, b a+1 ) also works.
2
Indeed, if we let c = b a+1 , then we know that c is an integer. The fact that b < c is
equivalent to ab < b2 + 1, which is true since a < b. The fact that c divides b2 + 1
follows from b2 + 1 = ac. Finally, the fact that b divides c2 + 1 is equivalent to b divides
(b2 +1)2 +a2
a2
. This is true because (b2 + 1)2 + a2 1 + a2 0 (mod b), and gcd(a, b) = 1.
Remark. The first few pairs obtained through our construction are (1, 2), (2, 5), (5, 13), ....
The corresponding values of N are 5, 13, 68, . . ..


PROBLEMS / SOLUTIONS / COMMENTS

Gene A. Berg, Editor
Students names and solutions are published with permission.
1/1/10. Several pairs of positive integers (m, n) satisfy the condition 19m + 90 + 8n = 1998 .
Of these, (100, 1 ) is the pair with the smallest value for n . Find the pair with the smallest
value for m .
Solution 1 by Michael Castleman (12/MA): First we solve the given equation for n in terms of
m, which gives n = 238.5 2.375m . Trying 1, 2, or 3 for m gives non-integer values for n, but
trying 4 for m gives n = 229. Thus the answer is (m, n) = (4, 229).
Solution 2 by Amanda Felder (12/TX): First, subtract 90 from each side of the equation, so it
reads 19m + 8n = 1908 . Since there is a remainder of 4 when 1908 is divided by 8, 19m must
also have a remainder of 4 when divided by 8, if 8n, and thus n, is to be an integer. Thus m must
be divisible by four since 19m leaves a remainder of four when divided by 8. The smallest posi
tive integer m divisible by 4 is 4. The pair with the smallest value for m is (4, 229).
Solution 3 by Brian Cruz (11/TN): Rewrite the equation as 19m = 1908 8n . Since n is a
positive integer, this statement is equivalent to 19m 1908 (mod 8) given the restriction
0 < 19m < 1908 . It can be further reduced to 3m 4 (mod 8), or m 4 (mod 8). The smallest
value of m with this characteristic is, of course, 4. Therefore the pair of positive integers (m, n)
satisfying 19m + 90 + 8n = 1998 with the smallest value of m is (4, 229).
Solution 4 by Blair Dowling (GA): Rewrite the equation as 19m + 8n = 1908 . First solve
19M + 8N = 1. Apply the extended Euclids algorithm to inputs 19 and 8:
19 = 2x8 + 3 Magic Box to generate a solution to 19M + 8N = 1.
8 = 2x3 + 2
3 = 1x2 + 1
2 2 1 2
2 = 2x1 + 0
0 1 2 5 7 19
Thus 1 0 1 2 3 8
19(3) + 8(-7) = 1
so M = 3 and N = -7.
First solution to 19m + 18n = 1908:

m = (M)(1908) = 3(1908) = 5724
n = (N)(1908) = (-7)(1908) = -13356.
We also have a solution to 19M + 8N = 0, namely M = -8 and N = 19.

So all solutions to 19m + 8n = 1908 are given by (m, n) = (5724, -13356) + a(-8, 19) for positive
integer a. We want the solution with the smallest positive value for m.
Since 5724/8 = 715.5, we set a = 715 to get

(m, n) = (5724 - 8x715, -13356 + 19x715) = (4, 229), our solution.
Solution 5 by Yen-Chieh Tseng (11/GA): This is a Diophantine equation of the type

Ax By = D , solvable using continued fractions. Converted into this form, the equation
becomes 19m 8( n ) = 1908 . Convert A/B into a continued fraction. The result is
<2, 2, 1, 2> (see the editors comment which follows for notation). Calculate the convergents, and
then solve for the smallest possible value of m. The work is shown below:
First, the partial quotients, ai, are given in the following table
i -2 -1 0 1 2 3
ai 2 2 1 2
pi 0 1 2 5 7 19
qi 1 0 1 2 3 8
i
Here, p i q i 1 q i p i 1 = ( 1 ) for each integer i.
Set M = Dq 2 + Bt = 1908 3 + 8t . With 0 < M we have 0 < 5724 + 8t and 715.5 < t . Sim
ilarly set N = D p 2 + At so N = 1908 7 19t . With 0 < N we have 0 < 13356 19t and
t < 702.9 .
Set t = 715 , the smallest value possible, and obtain M=4 and N=229. So the integer pair (m, n)
with the smallest value of m is (4, 229). Checking these values in the original equation conrms
the solution.
Editors comments: Mr. Tsengs solution gives an opportunity to expand on two interesting
mathematical topics, Diophantine equations and continued fractions.
The theory of Diophantine equations is concerned with integer solutions to polynomial equations.
Consider, for example, the equation
2 2
x 2y = 1 .
Solutions (x, y) to this equation include (1, 1), (3, 2), and (7, 5).
Continued fractions yield solutions for certain types of Diophantine equations, including a type
known as Pells equations, which includes this equation. Let D be a positive real number greater
than 1. Consider the following sequence of equations. In each equation ai is a positive integer
and 0 R i < 1 . If Ri = 0 we stop; but if R i 0 , then we can continue with the next equation. If
1
0 < R i < 1 , then we can write R i = ------------------------------ , where ai+1 is a positive integer and 0 R i < 1 .
ai + 1 + Ri + 1
D = a0 + R0
1
= a 0 + -----------------
a1 + R1
1
= a 0 + -----------------------------
1
a 1 + -----------------
a2 + R2
1
= a 0 + ------------------------------------------ = a 0, a 1, a 2, a 3,
1
a 1 + -----------------------------
1
a 2 + ----------------
a3 +
This is called the simple continued fraction expansion of D. Let
p
----0- = a 0 = a 0 ,
q0
p 1
----1- = a 0 + ----- = a 0, a 1 .
q1 a1
In general let
p 1
----i = a 0 + ----------------------------- = a 0, a 1, , a i
qi 1
a 1 + -----------------
1
+ ----
ai
These rational numbers are called the convergents, and give increasingly accurate approximations
to D.
2 2
The continued fraction expansion of 2 yields solutions to our equation, x 2y = 1 . The
continued fraction expansion of 2 is
1 1
2 = 1 + ( 2 1 ) = 1 + ------------------------------ = 1 + ---------------------------------- = 1, 2, 2, 2,
2 + ( 2 1) 1
2 + ------------------------
1
2 + --------------
2+
A few of the convergents are given in the following table:
i 0 1 2 3 4 5 ...
ai 1 2 2 2 2 2 ...
pi 1 3 7 17 41 99 ...
qi 1 2 5 12 29 70 ...
p 1 1.5 1.4 1.41666... 1.41379...1.41428... ...
----i
qi
First, observe that the convegents are increasingly good estimates of 2 , alternating between too
2 2
large and too small. Second, observe that pi 2q i = 1 for each i. Thus there are an innite
2 2
number of solutions to x 2y = 1 , and we are able to generate solutions very efciently.
Third, the similarities of the tables in solutions 4 and 5 reveal a close connection between the
extended Euclidean algorithm and continued fractions.
r 1 1 1 r
2/1/10. Determine the smallest rational number - such that --- + ---- + --- - whenever k, m, and
s k m n s
1 1 1
n are positive integers that satisfy the inequality --- + ---- + --- < 1 .
k m n
Solution by Irena Foygel (10/IL): Without loss of generality, let k m n .
If k = 2: Then m > 2.
1 1 1 1 1
If m = 3: --- + --- + --- < 1 , --- < --- , n > 6. The minimum integer is n = 7.
2 3 n n 6
Maximum --- + --- + --- = --- + --- + --- = ------ .

1 1 1 1 1 1 41
2 3 n 2 3 7 42
1 1 1 1 1
2 4 n n 4
Maximum --- + --- + --- = --- + --- + --- = ------ .

1 1 1 1 1 1 19
2 4 n 2 4 5 20
1 1 1 1 1
If m > 4: --- + ---- + --- < 1 , --- ---- , n > 4. The minimum integer is n = m.
2 m n n m
Maximum --- + ---- + --- = --- + --- + --- = ------ .

1 1 1 1 1 1 9
2 m n 2 5 5 10
If k = 3: Then m > 2.
1 1 1 1 1
3 3 n n 3
Maximum --- + --- + --- = --- + --- + --- = ------ .

1 1 1 1 1 1 11
3 3 n 3 3 4 12
1 1 1 1 1
If m > 3: --- + ---- + --- < 1, --- ---- , n m . The minimum integer is n = m.
3 m n n m
Maximum --- + ---- + --- = --- + ---- + ---- = --- .

1 1 1 1 1 1 5
3 m n 3 m m 6
1 1 1
If k > 3: --- + ---- + --- < 1 , m k , n k . For minimum m, m = k. For minimum n, n = k.
k m n
Maximum --- + ---- + --- = --- + --- + --- = ---

1 1 1 1 1 1 3
.
k m n k k k 4
In summary, for --- + ---- + --- < 1 , k, m, and n positive integers, the maximum --- + ---- + --- = ------ .
1 1 1 1 1 1 41
k m n k m n 42
r 41
So - = ------ .
s 42
3/1/10. It is possible to arrange eight of the nine numbers

2, 3, 4, 7, 10, 11, 12, 13, 15 1
in the vacant squares of the 3 by 4 array shown on the right
so that the arithmetic average of the numbers in each row and in 9 5
each column is the same integer. Exhibit such an arrangement, 14
and specify which one of the nine numbers must be left out when
completing the array.
Solution by Raymond Chen (10/CA): We are given 13 integers (1 through 15 except 6 and 8), of
which we must use twelve to ll the array. Since the average of the numbers in each row and col
umn is an integer [given], and there are 3 rows and 4 columns, the sum of the 12 numbers must be
divisible by 3 and 4; therefore the sum must be divisible by 12. Since the sum of the thirteen
numbers given is 106 10 (mod 12), we must leave out 10 from our
12 numbers. The sum of the remaining numbers is 96, so the sum for
each row is 32 = 96/3 and the sum for each column is 24 = 96/4. 1 A B C
Thus, D 9 E 5
(a) D + E = 18, so D and E must be 3 and 15 or 7 and 11
in some order.
F G 14 H
(b) D + F = 23, so D and F must be 11 and 12, in some order.
From (a) and (b), D = 11, so F = 12 and E = 7, from which B = 3.
Now,
(c) G + H = 6, so G and H must be 2 and 4 in some order.
(d) A + G = 15, so A and G must be 2 and 13 in some order.
1 13 3 15
From (c) and (d), G = 2, so H = 4 and A = 13, from which C = 15. 11 9 7 5
12 2 14 4
Therefore, the desired array is shown on the right,
with 10 left out of the given nine numbers.
4/1/10. Show that it is possible to arrange seven distinct points in the plane so that among any
three of these seven points, two of the points are a unit distance apart. (Your solution should
include a carefully prepared sketch of the seven points, along with all segments that are of unit
length.)
Solution by Michael Seeman (11/CA): See the diagram at A

the right. All marked segments have a length of one unit. E
Note that it is not a regular pentagon since not all angles
are congruent.
F
Requirement #1: Given any three points chosen
B G
from the gure, two are a unit apart.
D
(a) A pentagon with unit sides meets
requirement #1, because it is impossible to divide 5
into 3 groups of 2 (or more). (Note the angles are not congruent.) C
(b) Placing point F one unit away from points A, B, and C, still meets requirement #1, because the
three points including F and either A, B, or C, have a distance one unit between F and either A, B,
or C. The only three points remaining are F, E, and D, in which E and D are one unit apart.
(c) Use similar reasoning for the placement of point G. The three remaining possibilities, with a
unit length segment, are:
Three Points Unit Length Segment
G, B C BC
G, B, F BF
G, C, F CF.
(d) All possible sets of three points have two that are a unit apart, so this gure of seven points
meets requirement #1.
5/1/10. The gure on the right shows the ellipse y
2 2
( x 19 ) ( y 98 )
---------------------- + ---------------------- = 1998 .
19 98
Let R1, R2, R3, and R4 denote those areas within the ellipse
that are in the rst, second, third, and fourth quadrants,
respectively. Determine the value of R1 - R2 + R3 - R4.
R2 R1
Solution by Alex Wissner-Gross (12/NY): The ellipse
may be expressed in the form
2 2
( x 19 ) ( y 98 )
----------------------- + ----------------------- = 1 .
19 1998 98 1998
R3 R4 x
In this form it is clear the ellipse is centered at (19, 98) and
has lines of both horizontal and vertical symmetry that run
through its center (shown with dashed lines on the following
diagram).
If the regions of the ellipse between these lines of symmetry and

the x and y axes are labeled A, B, C, D, and E as shown in the diagram,
and if the total area of the ellipse is denoted by S, then R1, R2, R3, and R4 may be expressed
R1 = S/4 + A + C + D
y
R2 = S/4 - A + B
R3 = S/4 - B - C - E
A
R4 = S/4 - D + E.
R1 - R2 + R3 - R4 may thus be expressed:
(S/4 + A + C + D) - (S/4 - A + B) + (S/4 - B - C - E) - (S/4 - D + E) R2 R1

= 2A - 2B + 2D - 2E.
B C D
Because of vertical and horizontal symmetry,
A=C+E
and D = B + C. R3 R4 x
Therefore, 2A - 2B + 2D - 2E = E
=2(C + E) - 2B + 2(B + C) - 2E
=4C.
Since C is bounded by the axes ( x = 0 and y = 0 ) and the

lines x = 19 and y = 98, C has an area of 19 98 = 1862 .
Therefore the area of R1 - R2 + R3 - R4 is 4C = 7448 units2.
Editors Comment: During the summer of 1998, Mr. Wissner-Gross represented the United
States at the International Olympiad in Informatics (IOI 98), after having been ranked number 2
in America at the USA Computer Olympics training camp. I have said before that many
USAMTS alumni go on to represent the United States in the various science and mathematics
Olympiads. Although Alex is not yet an alumnus, he has proven me correct again.

1/2/10. Determine the unique pair of real numbers ( x, y) that satisfy the equation
2 2
(4 x + 6 x + 4)(4 y 12 y + 25) = 28 .
2
Solution 1 by Robert Kotredes (11/ME): The polynomial 4 x + 6 x + 4 can be written as
3 2 7
4 x + --- + --- , and therefore has a range of --- . The polynomial 4 y 12y + 25 can be writ
7 2
4 4 4
3 2
ten as 4 x --- + 16 , and therefore has a range of 16 . Because --- 16 = 28 , the only possi
7
2 4
3 3
ble values for each polynomial are their minimums, which occur at x = --- and y = --- ,
4 2
respectively. So the unique pair of real numbers ( x, y) is ---, --- .

3 3
4 2
2
Solution 2 by Kim Won Jong (12/CA): Let A = (4 x + 6 x + 4) , then
2
A (4y 12y + 25) = 28
2 28
(4y 12y + 25) = -----
A
2 28
4 y 12 y + 25 ------ = 0
A
2
Since there is a unique pair ( x, y) , the discriminant b 4ac of the quadratic formula must equal
zero, or
12 4 4 25 ------ = 0
2 28
A
448
144 400 + --------- = 0
A
7
A = --
4
So,
2 7
A = 4 x + 6 x + 4 = --
4
2 9
4 x + 6 x + --- = 0
4
2
b b 4ac b 3
x = -------------------------------------- = ------ = --- , since again, the discriminant is zero.
2a 2a 4
Substituting A into the original equation,
7 2
--- (4 y 12y + 25) = 28
4
2
4 y 12 y + 9 = 0
b 12 3
y = ------ = ---------- = --
2a 24 2
Therefore, the unique pair of real numbers ( x, y) that satisfy the equation
2 2
(4x + 6x + 4)(4y 12y + 25) = 28
is ---, ---
.
3 3
4 2
Solution 3 by Robert Klein (12/PA): Completing the square within each term we get:
2
4 x 2 + 6 x + 9--- + 7--- ((4y 2 12y + 9 ) + 16 ) = 2x + 3--- + 7--- ((2 y 3 ) 2 + 16 ) = 28
4 4 2 4
3 2
Substituting a = 2x + --- and b = (2 y 3 ) , the equation becomes a + --- (b + 16 ) = 28 .
2 7
2 4
7
Observing a 0 and b 0 , we must have a = 0 and b = 0 since --- 16 = 28 .
4
3 2
Solving a = 2x + --- = 0 and b = (2 y 3 ) = 0 , the unique pair of real numbers satisfying
2
2
the equation is (-3/4, 3/2).
2
Solution 4 by Adam Salem (12/NY): This solution uses calculus. Let f ( x ) = 4 x + 6 x + 4 and
2
g ( y ) = 4 y 12 y + 25 . Thus f ( x ) g ( y ) = 28 . To nd the extreme points of both f and g, set
their derivatives equal to 0. f( x ) = 8 x + 6 and f( x ) = 8 > 0 , so f(x) has a minimum at
x = 0.75 . Similarly, g( y ) = 8y 12 and g( y ) = 8 > 0 , so g(y) has a minimum at y = 1.5 .
This is equivalent to f ( x ) f ( 0.75 ) = 1.75 and g ( y ) g(1.5) = 16 . Multiplying the inequal-
ities yields f ( x )g ( y ) 28 . This is notable because 28 is the number from the original equation,
which means that we are simply trying to nd numbers x and y that minimize f ( x ) g ( y ) . Since
both f(x) and g(y) are always positive, it follows that the smallest value of f ( x ) g ( y ) is the prod
uct of their minimums. Thus the smallest value of f ( x ) g ( y ) is produced when
( x, y) = (0.75, 1.5)
Editors Comment: This problem is based on a similar problem (E: 11418) proposed by Petre
Btrnetu of Galati, Romania, in Issue 7-8/1997 of the Gazeta Mathematic.
2/2/10. Prove that there are innitely many ordered triples of positive integers (a, b, c) such
2 2 2 2 2 2
that the greatest common divisor of a, b, and c is 1, and the sum a b + b c + c a is the
square of an integer.
Solution 1 by Irena Foygel (10/IL): Let x and y be relatively prime positive integers such that
2 2
x 1 (mod 2). Let a = x , b = 2y , and c = xy . Because x and y are relatively prime and x is
not divisible by 2, a and b are relatively prime; therefore gcd(a, b, c) = 1.
Now,
2 2 2 2 2 2 4 4 4 2 2 2 2 4
a b + b c + c a = ( x )(4 y ) + (4y )( x y ) + ( x y )( x )
2 2 2 2 4 4 2 2 2
= x y (4x y + 4y + x ) = ( xy( x + 2 y ))
2 2 2 2 2 2 2 2 2
Set n = xy ( x + 2y ) and observe a b + b c + c a = n .
Because there are an innite number of pairs (x, y) meeting the above requirements, there are an
innite number of triples (a, b, c) meeting the requirements.
Solution 2 by Michael Castleman (12/MA): For relatively prime integers a, b, and c, the sum
2 2 2 2 2 2
a b + b c + c a is a square of an integer if one of the numbers equals the sum of the other
two. We shall now prove this.
Without loosing the generality of the proof, assume that a + b = c. Replacing a + b for c and sim
plifying, we get:
2 2 2 2 2 2
a b + b (a + b) + (a + b) a
2 2 2 2 2
= a b + (a + b )(a + b)
4 3 2 2 3 4
= a + 2a b + 3a b + 2ab + b
2 2 2
= (a + ab + b )
2 2
Since a + ab + b is an integer, the result is the square of an integer. Since there are an innite
number of ordered triples (a, b, c) such that a, b, and c are relatively prime and a + b = c, and, for
2 2 2 2 2 2
all of those pairs, a b + b c + c a is an integer, there exist an innite number of ordered tri
ples which meet the given criteria.
2 2 2 2 2 2
Solution 3 by Andy Large (11/TN): In order for a b + b c + c a to be the square of an inte
2 2 2
ger, it must be possible to write it in the form x + 2xy + y or ( x + y) . Conveniently, we have a
2 2
trinomial with squared rst and third terms: ( ab ) and ( ca ) . This means we only need the mid
dle term to be equivalent to 2 times the product of the said squared terms. That is
2 2 2
b c = 2 ( ab ) ( ac ) = 2a bc
bc 2
------ = a
2
bc
a = -----
2
a will be an integer under either of the conditions:
2n + 1 2 2
(i) b = 2 m and c = s for m, n, s integers.
2n + 1 2 2
(2 m )s n
Here a = ------------------------------------ = 2 ms
2
(ii) c and b are switched in (i) above.
n 2n + 1 2 2
Consider the triple (a, b, c) = (2 ms, 2 m , s ) , where m, n, and s are positive integers.
When m and s are relatively prime and when s and 2 are relatively prime, then GCD(a, b, c) = 1,
and this triple meets the requirements above. There are an innite number of such triples.
Solution 4 by Jeffrey Palmer (12/NY): Integers a, b, and c have GCD = 1 if two of the members
are distinct primes.
2 2 2 2 2 2
a b + b c + c a is the square of an integer when a
square can be completed. ac
2 2 2 2
The a c and a b terms are accounted by
the shaded areas (see diagram at right).
ab
If
2 2 2
b c = 2cba
2
2a
c = -------
b
If b = 2 and a is an odd prime, then innitely many tri
2
ples can be created which meet the requirements: (a, b, c) = (a, 2, a ) .
2 2 2 2 2 2 2
Solution 5 by George Lee (11/CA): Suppose a b + b c + c a = d , and rewrite the equa
2 2 2 2 2 2 2 2 2 2
tion as d (a + b )c = a b . With a = 1, this reduces to d (b + 1 )c = b .
2 2
Now let b = 1 to obtain d 2c = 1 . This is a Pells equation, which has innitely many solu
tions - each corresponding to an ordered triple (1, 1, c). [For a discussion of Pells equation see the
Editors comment following Solution 6.] For example, (c, d) = (c 1, d 1 ) = (2, 3) satises the
equation. Some ways that successive solutions can be generated are by:
(i) using the recursive relations c n + 1 = 3c n + 2d n and d n + 1 = 4c n + 3d n ;
(ii) using the single recursive relationc n + 2 = 6c n + 1 c n where c1 = 2 and c2 = 12; or
n n
(iii) using the equation c n = ((3 + 2 2) (3 2 2) ) ( 2) .
For this solution, we will show that the rst method works by induction:
2 2 2 2
d n + 1 2c n + 1 = (4c n + 3d n ) 2 ( 3c n + 2d n )
2 2 2 2
= 16c n + 24c n d n + 9d n 18c n 24c n d n 8d n
2 2
= d n 2c n
= 1.
Since cn+1 and dn+1 are larger than cn and dn, we can generate innitely many solutions (c, d) =
(cn, dn). Also, 1, 1, and cn are relatively prime. Thus we can generate innitely many correspond
ing solutions (1, 1, c).
Solution 6 by David Fithian (11/OR): We are to show that innitely many positive integer sets
2 2 2 2 2 2 2
(a, b, c, n) satisfy the equation a b + b c + c a = n with gcd(a, b, c) = 1. Without loss of
generalization, let c = 1. This automatically makes gcd(a, b, c) = 1, regardless of the values of a
2 2 2 2 2
and b. We are left with a b + a + b = n with a, b, n N . Now, set b = 1, so that
2 2 2 2
2a + 1 = n . Rearranging, we see that n 2a = 1 ; since this is a Fermat-Pell equation of the
2 2
form x d y = 1 , and since d = 2 is prime, there are innitely many solution pairs (a, n), and
thus the equation has innitely many positive integral solutions with gcd(a, b, c) equal to 1.
These solutions would be (a, b, c, n) = (r, 1, 1, s), where s/r is a certain fractional convergent of
2 . In particular, the rst solution pairs (a, n) are (2, 3), (12, 17), (70, 99), (408, 577),... . It can be
veried that n/a indeed converges to 2.
Editors Comments: This problem is due to Suresh T. Thaker of Bombay, India. We are grateful
for his contribution. A brief discussion of Pells equations is available in the solutions to Round
1 of Year 10. Mr. Lee (Solution 5 above) presents two more methods to generate solutions using
Pells equations which we have not included. Pells equations are used in Solution 6 as well. In the
following note, Erin Schram continues the discussion.
Solution by Pells Equation, summary by Erin J. Schram, longtime grader.

Several test-takers solved problem 2 by Pells equation. Gene Berg described Pells equation in
the solutions to round 1, but he did not expect Pells equation to be of any use in this round. The
creativity of the students who take the USAMTS is surprising and refreshing.
Pells equation is the quadratic Diophantine equation of the form

2 2
x Dy = N
where D and N are integer constants and x and y are integer unknowns. We start with the equation
2 2 2 2 2 2 2
a b +a c +b c = x
and we pick arbitrary integers for a and b. This changes the equation to a Pells equation in the
variables x and c. Although setting a and b to arbitrary values wont give every solution to
2 2 2 2 2 2 2
a b + a c + b c = x , the problem asked for only an innite family of solutions, and the
solutions to the Pells equation are an innite family. Besides, with a and b held constant, we can
pick them so that their greatest common divisor is 1, forcing the greatest common divisor of a, b,
and c to be 1.
For example, the most common choice was a = 1 and b = 1. Then the equation simplies to
2 2
x 2c = 1,
which is a Pells equation. This equation is almost identical to the equation from Gene Bergs
2 2
discussion, x 2y = 1 . Gene Bergs use of continued fractions to solve that Pells equation
2 2
can be adapted to solve problem 2, since every other term from the solutions to x 2y = 1
2 2
satises x 2y = 1 .
For variety, I will discuss the recursive solution to the Pells equation that results from setting a =
2 2
1 and b = 2, the second most common choice. This gives x 5c = 4 .
2 2
Suppose we have the Pells equation x Dy = N and one solution to it, ( x 0, y 0 ) . Further
2 2
more, suppose (u, v) is a solution to u Dv = 1 : note that I replaced the integer N with the
2 2
integer 1 in that equation. Then (ux 0 + vDy 0, vx 0 + uy 0 ) will be a solution to x Dy = N .
2 2
We have x 5c = 4 and some trial and error gives us ( x, c) = (3, 1) as one solution. For a
2
solution to u 2 Dv = 1 , lets be lazy and cut our solution to the previous equation in half,
which gives (u, v) = (1.5, 0.5 ) . Even though the recursion will not be built from integers, it is
possible that its results will still be integers. The result is a solution ( x, c) = (3, 1) and a recur
sive relation:
( x i + 1, c i + 1 ) = (1.5 x i + 2.5c i, 0.5 x i + 1.5c i )
Hence, we can rearrange the recursive relation to the following:

( x 0, c 0 ) = (3, 1)
( x 1, c 1 ) = (1.5 3 + 2.5 1, 0.5 3 + 1.5 1) = (7, 3)
( x 2, c 2 ) = (1.5 7 + 2.5 3, 0.5 7 + 1.5 3) = (18, 8 )
( x 3, c 3 ) = (1.5 18 + 2.5 8, 0.5 18 + 1.5 8) = (47, 21 )
( x 4, c 4 ) = (1.5 47 + 2.5 21, 0.5 47 + 1.5 21) = (123, 55 )
We also have the following relations:.
c i = 0.5 x i 1 + 1.5c i 1 , so x i 1 = 2c i 3c i 1
x i = 1.5 x i 1 + 2.5c i 1 , so x i = 1.5 ( 2c i 3c i 1 ) + 2.5c i 1 = 3c i 2c i 1
c i + 1 = 0.5 x i + 1.5c i , so c i + 1 = 0.5 ( 3c i 2c i 1 ) + 1.5c i = 3c i c i 1
The recursive relation c i + 1 = 3c i c i 1 tells us the solutions will all be integers. Some test-tak-
ers noticed that the values for c, which are 1, 3, 8, 21, 55, 144, 377,..., are every other term from
the Fibonacci sequence. The Fibonacci sequence, being the simplest nontrivial recursive
sequence, appears a lot in recursive relations.
3/2/10. Nine cards can be numbered using positive half-integers (1/2, 1, 3/2, 2, 5/2,...) so that the
sum of the numbers on a randomly chosen pair of cards gives an integer from 2 to 12 with the
same frequency of occurrence as rolling that sum on two standard dice. What are the numbers
on the nine cards and how often does each number appear on the cards?
Solution 1 by Megan Guichard (11/WA): With nine cards, there are 9C2 = 36 possible ways to
choose two cards. As it happens, there are also 6 6 = 36 possible outcomes when two dice are
rolled, with the following frequencies of occurrence:
Sum 2 3 4 5 6 7 8 9 10 11 12
Frequency 1 2 3 4 5 6 5 4 3 2 1
Therefore, there must be exactly one pair of cards with a sum of 2, exactly two pairs of cards with
a sum of 3, and so on.
Since all possible sums must be integers, the cards must be numbered either with all integers or
with all odd integer multiples of 1/2. Since half-integers are specically mentioned, it seems a
good assumption that all cards are numbered with odd integer multiples of 1/2 (at least until this
assumption is proven either true or false).
Operating under this assumption, there is only one way to get a sum of 2: 1/2 + 3/2. Thus there
must be exactly one card numbered 1/2 and exactly one numbered 3/2. Then there must be two
pairs of cards with sum 3; this can be accomplished by either 1/2 + 5/2 or 3/2 + 3/2. However, we
already know there is only one card numbered 3/2, so there must be two ways to draw 1/2 + 5/2.
Only one card is numbered 1/2 so there must be two cards numbered 5/2.
With the four cards we now have (1/2, 3/2, 5/2, 5/2) there are only two ways to draw a pair of
cards with sum 4: 3/2 + 5/2 and 3/2 + 5/2 (there are two cards numbered 5/2). Therefore we need
to add one more card that will create a sum of 4 when paired with an existing card. However the
new card must be greater than 5/2, because there are exactly the right number of cards less than or
equal to 5/2 already; thus the new card must be 7/2, which, when paired with 1/2, yields the third
way to get a sum of 4.
Continuing this line of reasoning shows that the remaining four cards must be numbered 9/2, 9/2,
11/2, and 13/2, meaning that the complete set of none cards is numbered as follows
(1/2, 3/2, 5/2, 5/2, 7/2, 9/2, 9/2, 11/2, 13/2)
With these cards, exactly one pair results in a sum of 2, exactly two pairings result in a sum of 3,
three pairs yield a sum of 4, and so on, with each sum having the same frequency as on a pair of 6
sided dice.
Editors comments: Erin Schram of the National Security Agency contributed this clever prob
lem.
4/2/10. As shown on the gure, square PQRS is

C
inscribed in right triangle ABC, whose right angle is
at C, so that S and P are on sides BC and CA, respec- P S
tively, while Q and R are on side AB. Prove that
AB 3QR and determine when equality occurs.
Solution 1 by Suzanne Armstrong (11/MO): A Q R B

(i) Since A CPS RSB
B CSP QPA
and AQP C SRB = 90
then ABC is similar to SRB .
(ii) Let AQ = 1 and let x = PQ = PS = QR = RS .
(iii) So
AQ PQ 1 x 2
-------- = -------- and hence --- = ------- and therefore BR = x .
RS BR x BR
2
(iv) AB = BR + QR + AQ = x + x + 1
2
(v) Now to prove that AB 3QR , we must show x + x + 1 3x .
2 2
This is equivalent to x 2 x + 1 0 or ( x 1 ) 0 which is clear
and is further emphasized by its graph (at right).
(vi) As our last step, we determine when equality occurs. When
QR = 1 = AQ , then AB = 3QR = 3 ; also,
when AB = 3QR = 3 , all of the triangles are isosceles
45, 45, 90 .
Thus, AB 3QR with equality when AQ = QR .
Solution 2 by Xuejing Chen (12/OK): Observe that

C
APQ = SBR and PAQ = BSC , so APQ is
similar to SRB . P S
a b 2
Thus --- = --- and a = bc . a
c a A b Q R c B
2
Since ( b c) 0 , it follows that b + c 2 bc .
So b + c 2a , and b + c + a 3a .
Thus AB 3QR .
Now, if b = c, then b + c = 2a so we have equality. This occurs if CAB is isosceles.
Solution 3 by Daniel Moraseski (11/FL):

AQ AQ AC
RB RB BC
By similarity we have and -------- = ------- = -------
-------- = -------- = --------
QR PQ BC
QR RS AC
s+t
The Arithmetic Mean - Geometric Mean inequality states ---------- st for nonnegative numbers s
2
and t. [See in Solution 2 above for proof.]
AC BC
So AB = QR + AQ + RB = QR 1 + -------- + -------- QR 1 + 2 -------- -------- = 3QR
AC BC
BC AC BC AC
AC BC
Equality is satised when -------- = -------- because this is the equality condition for the AM-GM. This
BC AC
means AC = BC and it is an isosceles right triangle.
Editors Comments: We are most grateful to Professor Hiroshi Okumura of the Maebashi Insti
tute of Technology for this excellent problem. Professor Okumura is in charge of the Japanese
counterpart of the USAMTS.
1
Many solutions used the inequality n + --- 2 (for positive number n) in one form or another.
n
2 2 2 1
Since (n 1 ) 0 we can write n 2n + 1 0 , so n + 1 2n and n + --- 2 . If this fact was
n
used in a solution, full credit for the solution required proof of this fact.
5/2/10. In the gure on the right, ABCD is a

convex quadrilateral, K, L, M, and N are the K C
D
midpoints of its sides, and PQRS is the
quadrilateral formed by the intersections of S
AK, BL, CM, and DN. Determine the area
R
of quadrilateral PQRS if the area of quadri
L N
lateral ABCD is 3000, and the areas of P
quadrilaterals AMQP and CKSR are 513 Q
and 388, respectively.
Solution 1 by Mary Tian (10/TX): A B

M
Notation: [P1P2...Pv] denotes the area of poly
gon P1P2...Pv.
Connect AC. Because M, K are midpoints of
AB, CD respectively, then [ACK] = [ACD]/2,
and [CAM] = [CAB]/2.
Hence
[AMCK] = [ACK] + [CAM] = ( [ACD]/2 ) + ( [CAB]/2 )
= [ABCD]/2
Thus
[PQRS] = [AMCK] - [AMQP] - [CKSR]
= ( [ABCD]/2 ) - [AMQP] - [CKSR]
= (3000/2) - 513 - 388
= 599.
So area of quadrilateral PQRS is 599.
Editors comment: We are thankful to Professor Gregory Galperin of Eastern Illinois University
for proposing an earlier version of this problem. His many contributions are most appreciated.

Gene A. Berg <gaberg@ieee.org>, Editor
1/3/10. Determine the leftmost three digits of the number

1 2 3 999 1000
1 + 2 + 3 + + 999 + 1000 .
Solution 1 by Jason Oh (11/MD): The leftmost three numbers of this sum are 100. To prove this
I will show that the only term that contributes to the leftmost three digits is the 10001000 term.
The way to see this is by considering the series
1 2 3 999 1000
1000 + 1000 + 1000 + + 1000 + 1000
Without much consideration, this series can be seen to add up to
1001001001001000
In this series the terms from 10001 to 1000999 have no effect on the three leftmost digits. In the
original series, the sum of the terms from 11 to 999999 will be smaller than the sum of the terms
10001 to 1000999 in the second series since each term is smaller and the original series is domi
nated by the second series. Thus in the original series, the rst 999 terms will also not contribute
to the leftmost three digits of the sum. The leftmost three terms of the sum come from the
10001000 term, and are therefore 100.
Solution 2 by Marcus Aaron (10/TX): Setting

1 2 3 999 1000
x = 1 + 2 + 3 + + 999 + 1000
it is clear that
1000 1 2 3 999 1000
1000 < x < 1000 + 1000 + 1000 + + 1000 + 1000
or that x is between 1000000... and 100100100... . Since both these numbers have exactly 3001
digits, the rst three digits of x must be 100.
Editors comment: Thanks are due to Professor Gregory Galperin of Eastern Illinois University
for this nice problem.
2/3/10. There are innitely many ordered pairs (m, n) of positive integers for which the sum
m + (m + 1 ) + (m + 2 ) + + (n 1 ) + n
is equal to the product mn . The four pairs with the smallest values of m are (1, 1), (3, 6), (15,
35), and (85, 204). Find three more (m, n) pairs.
Solution 1 by Lucy Jiang (12/MD): The equation can be rewritten as:
nm+1
(m + n) ---------------------- = mn ,
2
which simplies to
2 2
m m(1 2n) n n = 0 .
Although 0 is not a positive integer, it can be used in determining a pattern for m and n because
0 + 0 = 0 0.
Suppose (m 0, n 0 ) = (0, 0) , (m 1, n 1 ) = (1, 1) , (m 2, n 2 ) = (3, 6) , and so on. We observe that

n i + 2 = 6n i + 1 n i , as shown below:
(0,
0)
(1, 1) 1
6 0 = 6
(3, 6) 6
6 1 = 35
(15,
35 ) 35 6 6 = 204
(85, 204 )
Using this pattern, the next value for n is 204 6 35 = 1189 . Substituting n into the equation
2 2
m m(1 2n) n n = 0 , we obtain m = 493 . Thus the ordered pair is (493, 1189), and
this works.
By the same method, two other ordered pairs are: (2871, 6930) and (16731, 40391).
i(i + 1)
Solution 2 by Trevor Bass (12/NY): Using the fact that the sum 1 + 2 + 3 + + i = ----------------- ,
2
where i is an integer, we can rewrite the equation
m + (m + 1 ) + (m + 2 ) + + (n 1 ) + n = mn
as
n(n + 1) m(m + 1)
-------------------- ----------------------- = mn ,
2 2
so
2 2
m + (2n 1 )m + ( n n) = 0 .
In order for m to be an integer, the discriminant of this quadratic equation must be the square of an
integer, so
2 2 2
(2n 1 ) 4( n n) = i ,
where i is an arbitrary integer. This equation can be rewritten as
2 2
8n + 1 = i .
The rst seven values of n for which there exist is are: 1, 6, 35, 204, 1189, 6930, and 40391. For
each n, the corresponding m can be computed by solving the quadratic equation
2 2
m + (2n 1 )m + (n n) = 0 . The rst four ns are given in the problem statement, so the
next three (n, m) pairs are (493, 1189), (2871, 6930), and (16731, 40391).
Solution 3 by David Walker (11/NE): The expression m + (m + 1 ) + + (n 1 ) + n can be
nm+1
written (m + n) ---------------------- . Set this equal to mn and obtain (m + n)(n m + 1 ) = 2mn . If you
2
put 1, 3, 15, or 85 in for m in this equation, you get two solutions for each m. The solutions for 1
are 1 and 0. The solutions for 3 are 6 and -1. The solutions for 15 are 35 and -6, and for 85 are
204 and -35. I noticed the pattern that the solution to an m that is negative is the positive solution
to the previous m value. So I put -204 into the equation and got 493. Then I put -493 back in for
m and got 1189. I continued doing this and got the three (m, n) pairs (493, 1189), (2871, 6930),
and (16731, 4039).
Solution 4 by Oaz Nir (10/CA): Three pairs are: (m, n) = (493, 1189), (2871, 6930), and (16731,
40391).
Claim: If (m, n) is a solution, then (2n + m, 5n + 2m - 1) also is.
Using this claim, we easily recursively compute the three pairs above. Starting with the given
solution (85, 204), we use the transformation (m, n) (2n + m, 5n + 2m 1 ) three times in suc
cession.
To prove the claim, we rst use the arithmetic series sum formula to transform the given equation
into the following equivalent form:
(sum of the rst and last terms)(number of terms)/2 = mn
(m + n)(n m + 1 ) = 2mn .
Now we see that (2n + m, 5n + 2m 1 ) is a solution if
([2n + m] + [5n + 2m 1 ])([5n + 2m 1 ] [2n + m] + 1 ) = 2 [ 2n + m][5n + 2m 1 ]
(7n + 3m 1 )(3n + m) = 2 [ 2n + m][5n + 2m 1 ]
2 2 2 2
21n + 16mn + 3m 3n m = 20n + 18mn 4n + 4m 2m
2 2
n + n m + m = 2mn
2 2
(n + n mn ) + ( m + mn + m) = 2mn
n ( n m + 1 ) + m ( n m + 1 ) = 2mn
(n + m)(n m + 1 ) = 2mn .
That is, if (m, n) is a solution then (2n + m, 5n + 2m - 1) is also a solution.
Solution 5 by Reid Barton (10/MA): Answer: (493, 1189), (2871, 6930), and (16731, 40391).
Since m + (m + 1 ) + + (n 1 ) + n = (n m + 1 )(m + n) 2 , we want to nd m and n such

that
(n m + 1 )(m + n)
mn = ---------------------------------------------- .
2
We have
(n m + 1 )(m + n)
mn = ---------------------------------------------- .
2
2 2
2mn = nm + n m mn + m + n
2 2
m + 2mn n m n = 0
2 2
(m + n) (m + n) 2n = 0
2 2
4(m + n) 4(m + n) + 1 8n = 1
2 2
(2m + 2n 1 ) 8n = 1
2 2
t 8n = 1
This is Pells equation with primitive solution (3,1), so its solutions are given by
k
t n 8 = (3 8) for k 1 . For k = 1, 2, 3, and 4 we obtain (t, n) = (3, 1), (17, 6), (99, 35),
(577, 204), which give (m, n) = (1, 1), (3, 6), (15, 35), (85, 204) respectively, as m = (t-2n +1)/2.
Letting k = 5, 6, and 7 gives (t, n) = (3363, 1189), (19601, 6930), and (114243, 40391), which
give (m, n) = (493, 1189), (2871, 6930), (16731, 40391), three more solutions to the original equa
tion.
Solution 6 by Mike Fliss (12/NJ): The sum on the left side of the equation (the sum of the inte
gers from m to n) can be rewritten as (n m + 1 )(n + m) 2 since this is the number of terms
times the average term. We have
(n m + 1 )(n + m) 2 = nm
2 2
n m + n + m = 2nm
2
Now, in Ax + Bx + C = 0 form, this is
2 2
n + (1 2m)n + (m m ) = 0
We use the quadratic formula to obtain solutions for n. We want positive solutions for n. For all
2
m 1 , observe that 2m 1 8m 8m + 1 , so we do not consider the , just the + in the
quadratic formula to get positive values for n. So
2
2m 1 + 8m 8m + 1
n = -----------------------------------------------------------
2
2
8m 8m + 1 1
n = m + ---------------------------------------------
2
Set
2
8m 8m + 1 1
k = ---------------------------------------------
2
Since m and n are integers, by the additive closure of the integers, k is also an integer.
2
8m 8m + 1 = 2k + 1
2 2
8m 8m + 1 = (2k + 1 )
2 2
8m 8m + 2 = (2k + 1 ) + 1
2 2
(2k + 1 ) 2 (m 1 ) = 1
2 2
s 2r = 1
where s = 2k + 1 and r = m 1 . This is Pells equation, solvable by a continued fraction
expansion of 2 , because
2
s +1
2 = -------------
2
-.
r
Continued fraction expansion yields fractions (s/r). If (s, r) is a solution to the equation
2 2
s 2r = 1 , then the following argument shows that (s+2r, s+r) is also a solution:
2 2 2 2 2 2
(s + 2r ) 2(s + r) = s + 4sr + 4r 2s 4sr 2r
2 2
= s + 2r
= + 1
Therefore, the sequence of fractional expansions follows this form:
s s i + 2r i
---i , ----------------
- , ...
ri si + r i
Thus, the answers for the (s/r) expansion are
1 3 7 17 41 99 239 577 1393 3363 8119 19601 47321 114243 275807

1 2 5 12 29 70 169 408 985 2378 5741 13860 33461 80782 195025
BUT, these are the answers for +1 and -1; we only want solutions for -1. Therefore, we take every
other term.
1 7 41 239 1393 8119 47321 275807

1 5 29 169 985 5741 33461 195025
Since the bottom number corresponds to r, and 2m-1 = r, we get the m values (rst 7) in the table
2
8m 8m + 1 1
below. Plugging these ms back into the original n = m + ---------------------------------------------- , we get the cor
2
responding ns.
m 1 3 15 85 493 2871 16731

n 1 6 35 204 1189 6930 40391
Three new pairs (m, n) are listed in bold in the table.
Editors comment: We are indebted to Professor Suresh T. Thaker of India for this excellent
problem.
3/3/10. The integers from 1 to 9 can be arranged into a 3 3 array (as shown
on the right) so that the sum of the numbers in every row, column, and diag
A B C
onal is a multiple of 9.
D E F
(a.) Prove that the number in the center of the array must be a multiple of 3.
(b.) Give an example of such an array with 6 in the center.

G H I
Solution 1 by Rachel Johnson (9/MN): (a) There are nine possible combina
tions of three distinct numbers from 1 to 9 that have a sum of 9 or 18 (multiples of 9). They are:
{1, 2, 6} and {2, 3, 4} have a sum of 9, while {1, 8, 9}, {2, 7, 9}, {3, 6, 9}, {3, 7, 8}, {4, 5, 9}, and
{4, 6, 8} have a sum of 18. These combinations have either one or three multiples of 3 in each of
them.
Since all possible combinations have a multiple of 3 and there are only three
multiples of 3, the multiples of 3 must be shared. This can be done by placing 9 1 8
two (of 3, 6, and 9) in opposite corners. This provides a multiple of 3 for the
four outside combinations. To provide a multiple of three for the other combi 5 6 7
nations, the remaining multiple of 3 would have to be placed in the center. 4 2 3
(b) At the right is a possible arrangement of the array with 6 in the middle.
Solution 2 by Michelle Rengarajan (8/CA): (a) According to the given information:

1) A + B + C = 9k1
2) D + E + F = 9k2
3) G + H + I = 9k3
4) A + D + G = 9k4
5) B + E + H = 9k5
6) C + F + I = 9k6 1 8 9
7) A + E + I = 9k7 5 6 7
8) C + E + G = 9k8
3 4 2
Solve for E in equations 2, 7, and 8.
E = 9k2 -(D + F)
E = 9k7 -(A + I)
E = 9k8 -(C + G)
Add these three equations together to nd the value of 3E:
3E = 9k2 + 9k7 + 9k8 -(A + D + G + C + F + I).
So, according to equations 4 and 6 7 8 3
3E = 9k2 + 9k7 + 9k8 - 9k4 - 9k6.
2 6 1
Therefore, 3E is a multiple of 9, so E must be a multiple of 3.
9 4 5
(b) An array with 6 in the middle is shown on the right.
Editors comment: This clever problem was posed by Dr. Erin Schram of the National Security
Agency. He claims there are 24 possible arrays for part (b), all of which are rotations or reec
tions of one of the three arrays shown above.
4/3/10. Prove that if 0 < x < 2 , then
6 6 6 6
sec x + csc x + ( sec x)( csc x) 80 .
Solution 1 by Lucy Jin (11/MI): (Proof by contradiction.)

6 6 6 6
sec x + sec x + ( sec x)( csc x)
1 1 1
= -------------6
- + ------------
6
- + ------------------------------------
6 6

cos x sin x ( cos x)( sin x)
1 1 1
= --------------------3- + -------------------3- + ------------------------------------
6 6

( cos x)
2
( sin x)
2
( cos x)( sin x)
2 3 2 3
( sin x) + ( cos x) + 1
= --------------------------------------------------------
6 6

( sin x)( cos x)
2 2 4 2 2 4
( sin x + cos x)( sin x sin x cos x + cos x) + 1
= ----------------------------------------------------------------------------------------------------------------------
6 6
( sin x)( cos x)
4 2 2 4
sin x sin x cos x + cos x + 1
= ---------------------------------------------------------------------------
6 6

( sin x)( cos x)
2 2 2 2 2
( sin x + cos x) 3 sin x cos x + 1
= --------------------------------------------------------------------------------------
6 6

( sin x)( cos x)
2 2
2 3 sin x cos x
= --------------------------------------
6 6
( sin x)( cos x)
The last expression is equivalent to the original trigonometric expression.
Now we begin the proof by contradiction. Assume

2 2
2 3 sin x cos x
---------------------------------------
6 6
< 80 .
( sin x)( cos x)
We show that this assumption leads to a contradiction, and therefore must be false.
Let u = sin x cos x . The inequality above can be written
2
2 3u
-----------------
6
< 80
u
6 2
80 u + 3 u 2 > 0
u + 1--- u 1--- (80u 4 + 20u 2 + 8) > 0
2 2
Observe 80u + 20u + 8 is always positive. Therefore u + --- u --- > 0 , so u > --- and either
4 2 1 1 2 1
2 2 4
1 1
u < --- or u > --- . Recall u = sin x cos x , so
2 2
1 1
sin x cos x < --- sin x cos x > --
2 2
sin2 x < 1 sin2 x > 1 .
Both these cases are contradictory, because 1 sin 1 .
Therefore, our assumption was false and we must have

2 2
2 3( sin x cos x)
6 6
- 80 .
-------------------------------------------
( sin x)( cos x)
6 6 6 6
and sec x + csc x + ( sec x)( csc x) 80 .
Solution 2 by Mark Tong (12/VA): This proof makes extensive use of the Arithmetic Mean-
Geometric Mean (AM-GM) Inequality, which states that for positive numbers a and b,
12 2
(a +
b) 2 (ab) . For the rst step of the proof, substitute into AM-GM with a = sin x and
2 2 2 12 2 2
b = cos x . So ( sin x cos x) ( sin x + cos x) 2 sin x cos x 1 2

6 6 6 6
sin x cos x 1 64 sec x csc x 64 .
6 6
Using this new equation and substituting into AM-GM again with a = sec x and b = csc x :
6 6 6 6 12 6 6 12
( sec x + csc x) 2 ( sec x csc x) sec x + csc x 2 ( 64 ) = 16 .
6 6 6 6
Since sec x csc x 64 and sec x + csc x 16 ,
6 6 6 6
sec x + csc x + sec x csc x 80 .
5/3/10. In the gure on the right, O is the center A
of the circle, OK and OA are perpendicular to
one another, M is the midpoint of OK, BN is
parallel to OK, and AMN = NMO .
Determine the measure of ABN in degrees.
B N
Solution by Chi-Bong Chan (12/NJ): For conve
nience and without loss of generality let the radius
of the circle be 2. Then OM = 1, OA = 2, and the K O
M
Pythagorean Theorem gives MA = 5 . The
angle bisector theorem gives MO/MA = NO/NA.
Rewrite as
MO NO
------------------------- = ----------------------- .
MA + MO NA + NO
Since AN + NO = OA = OB, we have
NO MO 1 51
-------- = ------------------------- = ---------------- = ---------------- .
OB MA + MO 5+1 4
OO 2NO 51
If we reect O over BN to get O, then ---------- = ------------ = ---------------- .
OB OB 2
This is the golden mean, and one interest

ing fact involving the golden mean is that A
the ratio of the side to the diagonal of a
regular pentagon is the golden mean. O
Proof of the fact: Keep in mind that B N
1) because it is a regular pentagon, the ve

arcs are equal, and K O
M
2) an angle subtending an arc has one-half

the measure of the arc.
We shall rst show that each diagonal is

parallel to the opposite side, for example
in the diagram of the regular pentagon
below, AD || BC . We have
ADC = (arcAC ) 2 = ((2 5)(360)) 2 = 72 ,
and BCD = (arcBAD) 2 = ((3 5)(360)) 2 = 108 .
Now AD || BC follows directly from the fact that ADC and BCD are complementary.
Now, without loss of generality, let x be the length of a
A
side and 1 be the length of a diagonal, so that their ratio is
simply x. Since each diagonal is parallel to the opposite
side, AFDE is a parallelogram, and therefore DF = AB =
E B x. Moreover, since AD || BC , CBF = FDA and
1 1-x BCF = FAD . Therefore, ADF = BCF , so
F (BC ) (DA) = (BF ) (DF ) . That means
2
x 1 = (1 x) x or x = 1 x , which has the golden
x mean as the positive root, as desired. This completes the
proof of the fact.
D C
Back to our problem, we see that OBO

A
would be similar to DAC in the pentagon
because their sides are proportional
(OO/BO and DC/DA are both equal to the O
golden mean, and both triangles are isosce
les). Therefore BOO = ADC of the
B N B
pentagon, and so BOO = 72 from the
earlier calculation. 72
K O
M
Finally, to nd ABN , reect B across
AO to get B. Then
B = (arcAB) 2 = (72) 2 = 36
because an angle subtending an arc has
half the measure of the arc. Since B is
the image of ABN under the reection,
we have at last ABN = B = 36 .
Editors comment: Again we thank Professor Gregory Galperin of Eastern Illinois University
for this, his second beautiful problem included in this round.

Gene A. Berg, Editor (phone 301-688-3175)
1/4/10. Exhibit a 13-digit integer N that is an integer multiple of 213 and whose digits consist of
only 8s and 9s.
Solution 1 by Ann Marie Cody (12/MA): Let N = ABCDEFGHIJKLM, where A, B, ..., M are
digits.
N is divisible by 2, so M must be even. Therefore M = 8.
N = ABCDEFGHIJKx100 + LM and is divisible by 213. N and ABCDEFGHIJKx100 are
each divisible by 4, so LM must be divisible by 4. LM = L8, 88 is divisible by 4, and 98 is not, so
LM = 88.
N = ABCDEFGHIJx1000 + KLM is divisible by 8 and 1000 is divisible by 8, so KLM
must be divisible by 8. 8 does not divide 988, so KLM = 888.
N = ABCDEFGHIx10,000 + JKLM. N and 10,000 are divisible by 24, so JKLM is also.
JKLM = J888. 8888 is not divisible by 16 but 9888 is, so JKLM = 9888.
N = ABCDEFGHx100,000 + IJKLM. N and 100,000 are divisible by 25, so IJKLM is
also. IJKLM = I9888. 99888 is not divisible by 32 but 89888 is, so IJKLM = 89888.
N = ABCDEFGx1,000,000 + HIJKLM. N and 1,000,000 are divisible by 26, so HIJKLM
is also. HIJKLM = H89888. 889888 is not divisible by 64 but 989888 is, so HIJKLM = 989888.
N = ABCDEFx10,000,000 + GHIJKLM. N and 10,000,000 are divisible by 27, so
GHIJKLM is also. GHIJKLM = G989888. 8989888 is not divisible by 128 but 9989888 is, so
GHIJKLM = 9989888.
N = ABCDEx100,000,000 + FGHIJKLM. N and 100,000,000 are divisible by 28, so
FGHIJKLM is also. FGHIJKLM = F9989888. 99989888 is not divisible by 256 but 89989888 is,
so FGHIJKLM = 89989888.
N = ABCDx109 + EFGHIJKLM. N and 109 are divisible by 29, so EFGHIJKLM is also.
EFGHIJKLM = E89989888. 889989888 is not divisible by 512 but 989989888 is, so
EFGHIJKLM = 989989888.
N = ABCx1010 + DEFGHIJKLM. N and 1010 are divisible by 210, so DEFGHIJKLM is
also. DEFGHIJKLM = D989989888. 9989989888 is not divisible by 210 but 8989989888 is, so
DEFGHIJKLM = 8989989888.
N = ABx1011 + CDEFGHIJKLM. N and 1011 are divisible by 211, so CDEFGHIJKLM is
also. CDEFGHIJKLM = C8989989888. 88989989888 is not divisible by 211 but 98989989888
is, so CDEFGHIJKLM = 98989989888.
N = Ax1012 + BCDEFGHIJKLM. N and 1012 are divisible by 212, so BCDEFGHIJKLM
is also. BCDEFGHIJKLM = B98989989888. 998989989888 is not divisible by 212 but
898989989888 is, so BCDEFGHIJKLM = 898989989888.
N and 1013 are divisible by 213. N=ABCDEFGHIJKLM = A98989989888.
9898989989888 is not divisible by 213 but 8898989989888 is, so
N = 8898989989888
Solution 2 by Oaz Nir (10/CA): Answer: 8898989989888

Proof: More generally, we prove that for all positive integers n and non-zero digits a and
b, with a even and b odd, there exists an integer xn with exactly n digits that is an integer multiple
of 2n, whose digits consist of only as and bs. The proof is by mathematical induction. For the
base case we take x1 = a. Clearly this number is divisible by 21 = 2, as required. For the induction
step, we assume that an integer xn exists with the desired properties, and we show how to con
struct xn+1. We need two cases:
Case 1: xn is divisible by 2n+1.
n
Take xn+1 equal to xn with the digit a appended to the left. That is, x n + 1 = a 10 + x n .
n
It is easy to see that xn+1 is divisible by 2n+1 (because both terms a 10 and xn are).
Case 2: xn is not divisible by 2n+1.
n
Take xn+1 equal to xn with the digit b appended to the left. That is, x n + 1 = b 10 + x n .
n
It is easy to see that xn+1 is divisible by 2n+1 (because both terms b 10 and xn are odd multiples
of 2n, so their sum is an even multiple of 2n, and is therefore a multiple of 2n+1 as required).
This completes the proof of the general result. We can use our method of proof to actually
calculate the numbers xn. In particular, when a = 8, b = 9, and n = 13, we proceed as follows: x1
= 8, x2 = 88, x3 = 888, x4 = 9888, x5 = 89888, x6 = 989888, x7 = 9989888, x8 = 89989888,
x9 = 989989888, x10 = 8989989888, x11 = 98989989888, x12 = 898989989888, and x13 =
8898989989888, where the numbers x2, x3, x5, x8, x10, x12, and x13 were formed as in Case 1,
and the numbers x4, x6, x7, x9, and x11, were formed as in Case 2.
Editors comments: We congratulate Mr. Nir for winning the Brilliancy Award from the Bay
Area Mathematical Olympiad for a very elegant solution. The BAMO is a regional mathematical
competition in the San Francisco area.
Solution 3 by Zhihao Liu (9/IL): Answer: 8898989989888

N is divisible by 213. Therefore for every integral value of k from 1 to 13, the last k digits
of N are divisible by 2k. Working from right to left, the unit, tens, and hundreds digits must all be
8, since 888 is divisible by 23.
Also, if a k-digit number is divisible by 2k, then when it is divided by 2k+1, the remainder
is either 0 or 2k. Using this fact, we can deduce the other digits of N. Since
888 8 9000mod16 , thus 9000 + 888 8 + 8 , and 9888 is a multiple of 16. Similarly,
9888 0mod32 , hence 32 divides 89888. Performing this algorithm eight more times:
89888 32mod64
989888 64mod128
9989888 0mod256
89989888 256mod512
989989888 0mod1024
8989989888 1024mod2048
98989989888 0mod2048
898989989888 0mod4096
8898989989888
Indeed, 8898989989888 is divisible by 213 as it can be expressed as 8192x1086302489.
Therefore, N = 8898989989888.
Editors comments on modular arithmetic: This may be a good opportunity to briey discuss
modular arithmetic as used in Mr. Lius solution above. Instead of the standard clock with twelve
digits, 1, 2, 3, ..., 12, consider a clock that has p digits, 0, 1, 2, ..., p-1. As a specic example, con
sider a clock that has p = seven digits, 0, 1, 2, 3, 4, 5, 6. With this clock we add as follows: 4 + 5
corresponds to the time starting at four oclock and advancing ve hours to nine oclock, which
on this clock is 2 oclock. Thus, on this clock 4 + 5 = 2. We write this 4 + 5 2 , or to make sure
everyone knows we are doing clock arithmetic on a clock with 7 digits, we might write
4 + 5 2mod7 . We read this expression as four plus ve is congruent to 2 modulo seven. Simi
larly, 6 + 3 2 , 4 + 3 0 , etc. Notice in each case, we only care what the remainder is after
division by seven; in the example 4 + 5, notice 4 + 5 = 9, and 9 divided by 7 is 1 with a remain
der of 2, so 4 + 5 2mod7 . We dont seem to care what the quotient is, we only care about the
remainder after division by seven. We can also subtract with clock arithmetic: 4 - 5 corresponds
to starting at four oclock and backing up ve hours, which on this clock is 6 oclock. Thus,
4 5 6 . Observe that 4 4 4 + 3 0 , so we may think of 3 as -4; that is the additive inverse
of 4 is 3 [ i.e. three is the number we must add to four to get zero]. Knowing how to add allows
us to multiply: notice that 3
5 corresponds to adding 5 three times, and we already know how
to add, so 3 5 = 5 + 5 + 5
1 . With this idea we can now multiply in this clock arithmetic
system. Other examples of multiplication are 6 3 4 , 4 3 5 , etc. So now we can add, sub
tract, and multiply on this clock. Division is more difcult. If our clock has a prime number of
digits, then we can always divide by nonzero numbers. But if our clock has p digits where p is not
prime, we may not always be able to divide by every nonzero number. In our example with p = 7
observe that 3 5 = 5 + 5 + 5 1 , so ve is the multiplicative inverse of three [i.e. ve is the
1 1
number we must multiply by three to get one]. We write --- 5mod7 . Similarly, --- 4mod7 ,
3 2
1 1
--- 2mod7 , etc. Just to make sure you understand, before reading ahead, nd --- ; that is, nd the
4 6
number you must multiply by 6 to get 1 modulo 7.
When p is a prime, Euclids algorithm helps us nd multiplicative inverses. When p is 7 it
is easy to do this just by guessing and trying. But if p is large, say p = 163 or an even larger prime
number, it is more difcult to guess and check. With p = 163 we use the Euclidean algorithm as
in the following example. Suppose we have the number 25 and we wish to nd the multiplicative
inverse of 25 modulo 163. That is, we wish to nd an integer k such that 25 k 1mod163 . The
Euclidean algorithm gives us the greatest common divisor of 25 and 163. Since 163 is prime and
0<25<163, we know this will be one. The process of the Euclidean algorithm gives us more infor
mation, which solves our problem. This process is sometimes called the extended Euclidean
algorithm:
Divide 25 into 163 to get quotient 6 and remainder 13. 13 = 163 6 25 .
Divide 13 into 25 to get quotient 1 and remainder 12. 12 = 25 1 13 .
Divide 13 by 12 to get quotient 1 and remainder 1. 1 = 13 1 12 .
Now back up through this process.
1 = 13 1 12
1 = (1 13) 1 (25 1 13) = (2 13) (1 25)
1 = 2 (163 6 25) (1 25)
1 = (2 163) (13 25)
Thus, the greatest common divisor of 25 and 163 is one, and can be written as in the last
equation above. Since I only care about the remainder after division by 163, and the term 2 163
will contribute zero to this remainder, I can write
1 13 25mod163
Since 13 150mod163 , we rewrite this as
1 150 25mo 163 .
So the multiplicative inverse of 25 modulo 163 is 150. With clock arithmetic on a clock with p =
163, division by 25 is equivalent to multiplication by 150.
Again, when p is prime we are guaranteed that each nonzero digit will have a multiplica
tive inverse, so we can divide by all nonzero digits. When p is not a prime we can still add, sub
tract, and multiply modulo p, but we cannot always divide by nonzero numbers because on these
clocks there are nonzero divisors of zero.
1
--- 6mod7 is the answer to the exercise posed above. Demonstrate it using Euclids
6
Algorithm.
Editors comments: Thanks go to Professor George Berzsenyi, the creator of the USAMTS and
our continuing supporter, for posing this interesting problem. Round 4 completes the tenth year
of the USAMTS.
2/4/10. For a nonzero integer i, the exponent of 2 in the prime factorization of i is called ord2(i).
For example, ord2(9) = 0 since 9 is odd, and ord2(28) = 2 since 28 = 22 x 7. The numbers
3n - 1 for n = 1, 2, 3,... are all even, so ord2(3n - 1) > 0 for n > 0.
a) For which positive integers n is ord2(3n - 1) = 1?
b) For which positive integers n is ord2(3n - 1) = 2?
c) For which positive integers n is ord2(3n - 1) = 3?
Prove your answers.
n
Solution 1 by Aaron Marcus (10/TX): (a.) 3 1 can be factored into
2 n1
(3 1)(1 + 3 + 3 + + 3 ) . Since the second factor consists of only odd elements, the sum
will be odd if and only if there are an odd number of elements. Since there are (n-1)+1=n ele
ments (exponents 0 through n-1), ord2(3n - 1) = 1 when n is odd, or when n = 2k + 1.
(b.) Consider when n is even. The terms in the second factor can be paired. Thus,
n 2 n1
3 1 = (3 1 )(1 + 3 + 3 + + 3 )
2 2 n2 n2
= 2 (( 1 + 3 1) + (3 + 3 3 ) + ( 3 +33 ))
2 n2
= 2 (( 4 1) + (4 3 ) + ( 4 3 ))
2 n2
= 8 ( 1 + 3 + 3 )
Since n can only be even or odd, there is no case where ord2(3n - 1) = 2 (it must be either 1 or
greater than or equal to 3).
2 n2
(c.) From above, for ord2(3n - 1) = 3, (1 + 3 + + 3 ) must be odd.
n
--- 1
2 n2 2 2
1+3 ++3 = 1+9+9 ++9 . Since all terms are odd, this can only be odd if
n n
there are an odd number of terms. There are --- terms in the factor, so ord2(3n - 1) = 3 when --- is
2 2
odd, or when n = 4k + 2.
Solution 2 by Reid W. Barton (10/MA): Use to represent if and only if. Note that
k k+1 k k+1
(ord 2 ( n
) = k ) 2 divides n and 2 does not divide n n 2 (mod2 ) .
(a.) Answer: n odd.
n n n 1
We have ord 2 (3 1 ) = 1 3 1 2 ( mod 4) 3 3 ( mod 4) . Now 3 3 ( mod 4) and
2 n
3 1 ( mod 4) , so 3 3 ( mod 4) if and only if n is odd.
(b.) Answer: no n.
n n n 1
Since ord 2 (3 1 ) = 2 3 1 4 ( mod 8) 3 5 ( mod 8) , and 3 3 ( mod 8) ,
2 n
3 1 ( mod 8) , there is no n for which 3 5 ( mod 8) , and therefore no n for which ord2(3n - 1)
= 2.
(c.) Answer: n 2 ( mod 4) .
n n n
in this case, ord 2 (3 1 ) = 3 3 1 8 ( mod 16) 3 9 ( mod 16) , and
1 2 3 4
3 3 ( mod 16), 3 9 ( mod 16) , 3 11 ( mod 16) , and 3 1 ( mod 16) , so
n
3 9 ( mod 8) if and only if n 2 ( mod 4) .
Editors comment: This problem was posed by Dr. Erin Schram of the National Security
Agency. Dr. Schram contributed one problem for each round of Year 10. His involvement is
greatly appreciated.
1
3/4/10. Let f be a polynomial of degree 98, such that f ( k ) = --- for k = 1, 2, 3, ..., 99. Deter-
k
mine f(100).
1
Solution 1 by Wei-Han Liu (8/TN): Consider the expression f ( x ) --- . This expression has
x
roots at 1, 2, 3, 4, ..., 99. However, this is not a polynomial. Next consider the function
g ( x ) = xf ( x ) 1 . (a)
Because f(x) is a polynomial of degree 98, g(x) is a polynomial of degree 99. It has the roots 1, 2,
3, 4, ..., 99. Therefore it must be of the form
g ( x ) = c ( x 1 )(x 2 )(x 3 )(x 99 ) (b)
where c is a constant. To determine c we nd that g(0) = -1 from equation (a). So we put 0 in for
x in equation (b) and get g(0) = -c(99!) = -1, so c = 1/(99!) and
(x 1 )(x 2 )(x 3 )(x 99 )
g ( x ) = ----------------------------------------------------------------------------- . (c)
99!
We want to nd f(100), so we plug 100 into both equation (a) and (c) and get
(100 1 )(100 2 )(100 3 )(100 99 )
100( f (100)) 1 = ----------------------------------------------------------------------------------------------------
99!
100( f (100)) 1 = 1
100( f (100)) = 2
1
f (100) = -----
50
1
Therefore, f (100) = ------ .
50
Solution 2 by Daniel Moraseski (11/FL): Generalize by replacing 99 with n. Since f(k)k = 1 for
all integers k between 1 and n,
f ( x )x 1 = A ( x 1 )(x 2 )(x n) ,
where A is a constant, is also true for these values of k. Now plug in x = 0 and get
n
1 = An! ( 1 )
n
( 1 )
A = ------------- .
n!
If we set A equal to this, the above equation is also true at x = 0. Since it is of degree n and is
valid at n+1 values, it is true for all x. Now we nd f(n+1).
n1
( 1 ) n1
f ( n + 1 )(n + 1 ) 1 = -------------------- n! = ( 1 )
n!
n1
( 1 ) +1
f ( n + 1 ) = ----------------------------
n+1
2
f ( n + 1 ) = 0 for n even. f ( n + 1 ) = ------------ for n odd.
n+1
Applied to our specic problem, n = 99, so f(100) = 1/50.
4/4/10. Let A consist of 16 elements of the set {1, 2, 3., 106} , so that no two elements of A
differ by 6, 9, 12, 15, 18, or 21. Prove that two elements of A must differ by 3.
Solution 1 by Kasia Kobeszko (12/LA): (Proof by contradiction) Since all the differences for
bidden within A (including the difference in question, 3) are multiples of three, the set A can be
separated into numbers coming from three disjoint sources: numbers in the form 3n, in the form
3n+1, and in the form 3n+2, for integer ns. The largest of these sources has 36 elements.
Let us assume that no two members of A differ by 3.
This, combined with the stated difference restrictions, means that two members of A from
any one source must differ by at least 24 to both be eligible for membership in A, or when count
ing by threes, they must be at least eight apart. That means that the most members that can come
from one source is (size of source) 8 36 8 = 5 , where w is the ceiling function indi
cating the smallest integer w .
With at most 5 elements from each of three sources, A can have at most 15 elements. This
contradicts the denition of A as a 16-element set, and disproves our assumption.
Solution 2 by Emily Kendall (10/IN): Assume temporarily that no two members of A differ by
three. Consider all members of A congruent to c (mod 3), where c = 0, 1, or 2. Because these
members all differ by multiples of three, they must differ by at least 24.
106 can be divided into at most 4 intervals of 24. Since any two consecutive members of
A congruent to c (mod 3) must be separated by one of these intervals, there can be at most 5 such
members.
Therefore, assuming that no two members differ by three, there are at most ve members
of A congruent to 1 (mod 3), ve congruent to 2 (mod 3), and ve congruent to 0 (mod 3), for a
total of 15. So in order to have 16 elements in set A, two of them must differ by three.
Solution 3 by Vikki Kowalski (11/AR): What follows is an attempt to construct a set of 16 ele
ments that meet all the required conditions of set A and have no two elements which differ by 3.
That would mean that for each new element added to A, 14 other potential elements are
made invalid for consideration to become part of the set A. The maximum number of times that a
single element could be counted (overlap between the set that one member of A makes invalid
and the set that another member of A makes invalid) is twice.
This is because in order for an element to be removed from consideration twice, it must be
3n (Let us dene n and all integers ni as integers such that -8 < n < 8) greater than a member of A
and 3n units less than a member of A. If an element was removed from consideration three times,
this would mean that there must have been an element chosen for set A that was 3n1 from it and
one 3n2 from it in the same direction (since there are only 2 directions and 3 elements, pigeonhole
principle). However, this must mean that they are 3(n1 - n2) units apart, and -8 < n1 - n2 < 8, so
they could not actually both be members of the set A.
The only problem that remains is the elements near the upper and lower limits of the set.
If items for set A are selected sufciently close to the rst element in the set, they will each elimi
nate only 7 elements from consideration for set A. Without loss of generality, we may choose the
elements near the lower limit of the set (1, 2, and 3). Selecting these three will cause a total of
only 8 elements to be eliminated from consideration (for each one selected). Following these, 3
elements may be selected at 24-unit intervals (because these three, and each three following, elim
inate a total of 24 elements). The maximum number which can be chosen for set A using this pro
cess before all possible elements have been either selected or eliminated from consideration is 15
(this may be repeated ve times, terminating with 97, 98, and 99).
The only elements which are left to be selected that satisfy the initial constraints (but not
the extra one applied originally) will be the ones which have a difference of 3 with one of the ele
ments selected for A.
Therefore, two elements of A must differ by 3.
5/4/10. In ABC , let D, E, and F be the mid C

points of the sides of the triangle, and let P,
Q, and R be the midpoints of the correspond
ing medians, AD , BE , and CF , respec
tively, as shown in the gure at the right.

E R D
Prove that the value of
2 2 2 2 2 2
AQ + AR + BP + BR + CP + CQ
P
-----------------------------------------------------------------------------------------------
2 2 2
- Q
AB + BC + C A
does not depend on the shape of ABC and
nd that value. A F B
Solution 1 by Luke Gustafson (10/MN):

ABC can be placed on a coordinate sys
C (4y, 4z)
tem as shown on the right, with A on the
origin, B at (4x, 0), and C at (4y, 4z). By the
midpoint theorem, D = (2x +2y, 2z),
E = (2y, 2z), F = (2x, 0), P = (x+y, z), Q = D
E R
(2x + y, z), and R = (x+2y, 2z).
Using the distance formula, P

Q
AQ2 = (2x+y)2+z2 = 4x2+4xy+ y2+ z2
AR2 = (x+2y)2+4z2= x2+4xy+4y2+4z2
BP2=(3x-y)2+z2= 9x2-6xy+ y2 + z2 A F B (4x, 0)
BR2=(3x-2y)2+4z2= 9x2-12xy+4y2+4z2
CP2=(x-3y)2+9z2= x2-6xy+9y2+9z2
CQ2=(2x-3y)2+9z2= 4x2-12xy+9y2+9z2.
Adding these up gives 28(x2-xy+y2+z2).
Furthermore,
AB2=16x2
BC2=16x2-32xy+16y2+16z2
CA2=16y2+16z2
Adding these up gives 32(x2-xy+y2+z2).
2 2 2 2 2 2 2 2 2
AQ + AR + BP + BR + CP + CQ 28 ( x xy + y + z ) 7
That makes -----------------------------------------------------------------------------------------------
2 2 2
- = --------------------------------------------------
2 2 2
- = --
AB + BC + C A 32 ( x xy + y + z ) 8
Since this construction used an arbitrary triangle, this value is independent of the shape of
ABC .

1/1/11. The digits of the three-digit integers a, b, and c are the nine nonzero digits 1, 2, 3, ..., 9,
each of them appearing exactly once. Given that the ratio a:b:c is 1:3:5, determine a, b, and c.
Solution 1 by Antoinette Realica (10/GU):

a: b: c
1: 3: 5
--- --- ---
a, b, c have 9 nonzero digits with each of them appearing exactly once.
The smallest possible number: 123
The last digit of c should be a multiple of 5 ending in 5, so the rst digit of a should end in an odd
digit (3, 7, or 9). The rst digit of a should be 1 because if it were larger, then c would be four
digits.
Process of trial and error:

a: b: c
1: 3: 5
123: 369 615 3 and 6 repeated
127: 381: 635 1 and 3 repeated
129: 387 645 This works!
a=129, b = 387, c = 645.
Solution 2 by Vladimir Novakovski (10/VA): Let a = a2a1a0, b = b2b1b0, c = c2c1c0.
Since c = 5a < 1000, a < 200, so a2 = 1. Let us examine a0. If

a0 is even, then c0 = 0, which is impossible. If a0 = 5, then b0 = c0 = 5, which is also impossible.
If a0 = 7, then b0 = 1, which is impossible as well (since a2 = 1). So a 0 {3, 9} .
Case 1: a0 = 3
Consider a1, b1, and c1. From our hypothesis, it follows that b 1 3a 1 (mod10) , and that
c 1 5a 1 + 1 ( mod10 ) ( the latter since 5 3 = 10 + 5 , and the 10 contributes to the tenths dig
its). Therefore a1 cannot be even, since then c1 = 1. Also, a 1 5 since otherwise a1 = b1 = 5;
and a 1 7 since otherwise b1 = 1. Since a0 = 3, a 1 3 , and since b 0 3 a 0 mod 10 = 9,
a 1 9 . So there are no solutions.
Case 2: a0 = 9
Again, consider a1, b1, and c1. Now b 1 3a 1 + 2 ( mod10 ) , and c 1 5a 1 + 4 ( mod10 ) . Now
a 1 3 since then b1 = 1; a 1 4 since then b1 = 4; a 1 5 since then c1 = 9; a 1 6 since then
b1 = 0; a 1 7 since then c1 = 9; a 1 1 and a 1 9 . So the only possibilities are 2 and 8.
Now 189 5 > 900 , but c 0 9 . It can be checked by arithmetic that 129 does work, so the num
bers a, b, and c are 129, 387, and 645.
Editors Comments: This problem was proposed by Professor Sndor Rka, the founding editor
of Hungarys Abacus, a problem solving journal for students in grades 4 through 8.
2/1/11. Let N = 111...1222...2, where there are 1999 digits of 1 followed by 1999 digits of 2.
Express N as the product of four integers, each of them greater than 1.
Solution 1 by Kirsten Rutschman (12/WA):

To solve, look for patterns:
1122 11 = 102
111222 111 = 1002
11112222 1111 = 10002
N A = B where A is 1999 digits of 1 and B = 1000...02 has 1998 digits of 0.
102 2 = 51
1002 2 = 501
10002 2 = 5001
B2 = C where C = 5000...01 with 1997 digits of 0.
51 3 = 17
501 3 = 167
5001 3 = 1667
C3 = D where D = 1666...67 with 1997 digits of 6.
N = A B C D = 1111 2 3 166667 .
Solution 2 by Luke Gustafson (11/MN):

n n
10 1 is a string of n 9s, hence (10 1 ) 9 is a string of n ones. Then we have
1999 1999
1999 10
------------------------ + 2 ------------------------
1 10 1
N = 10
9 9
1999
N = ------------------------ (10
10 1 1999
+ 2)
9
1999
N = 2 ------------------------ ( 5 10
10 1 1998
+ 1)
9
1999 1998
1 5 10
N = 2 3 ------------------------ -------------------------------
10 +1
9 3
1998
The last factor is an integer since the sum of the digits of 5 10 + 1 (a ve followed by a
string of zeros and then a one) is 6, implying it is divisible by three. Hence
N = 2 3 (1111) (166667)
1998
It might be interesting to note that N is also divisible by 23. I will show that 5 10 + 1 is
divisible by 23. This proof makes use of Fermats Theorem: if x 0 ( mod23 ) , then
22
x 1 ( mod23 ) .
1998 22 90 18
5 10 + 1 5 (10 ) (10 ) + 1 ( mod23 )
18
5 10 +1
9
5 (100 ) + 1
9
58 +1
22 5
5 (2 ) (2 ) + 1
5
52 +1
161
0 ( mod23 )
This yields a more complete factorization of N:
1999 1998
5 10
N = 2 3 23 ------------------------ --------------------------------
10 1 +1
9 69
Solution 3 by Alex Utter (11/CA): N is even so it is a multiple of 2, but the last two digits are not
divisible by 4, so N is not a multiple of 4. The sum of its digits is 1999 1 + 1999 2 = 1999 3 ,
so it is a multiple of 3 but not 9. 5 is not a factor because N does not end with 0 or 5.
Potential third factors can be tested by a technique that I devised based on remainder repetition
intervals. I will use the division of 7 into a short series of ones and twos as an example:
1 5 8 7 3 0 3 1 7 4 6
7 ) 11 14 16 15 12 10 22 21 25 23 24 20
As you can see, it divides evenly into 6 ones, and divides evenly into 6 twos. Because the remain
der at the end of each section is zero, it can loop around as many times as desired. In addition, at
equal distances from the change from 1 to 2 are two remainders of 2, so the cycle can be inter
rupted at that point. Therefore, 7 is a factor of any number like N with 6x ones and 6x twos, or a
number with 6x+5 ones and 6x+5 twos, with x as an integer greater than or equal to 0. I write this
cycle structure as 6x and 6x+5. Since 1999 does not equal 6x or 6x+5 for any integer x, 1999 does
not t sevens cycle, and 7 is not a factor.
Now, this method of testing primes until a factor is found can be used. The results starting after 7
are:
Prime Cycle structure
11 2x
13 6x
17 16x, 16x+2
19 18x, 18x+8
23 22x, 22x+19
1999 = 22 90 + 19 = 22 x + 19 , so the third factor of N is 23. The results of division cycle

with the remainder cycles, so N/23 can be expressed as follows:
(Notation: X indicates that the previous series of numbers should be repeated the indicated num
ber of times, ... indicates that the next number should be concatenated to the previous one)
N/23 = 0048309178743961352657X90...
004830917874396135266183574879227...
00966178357487922705314X90.
Each section alone is not a multiple of 3, but the three sections together are. In addition, the mid
dle section (put together) is also a multiple of 3. So
N
------------- =
3 23
001610305958132045088566827697262479871175523349436392914653784219X30...
0016130595813204508872785829307568438...
003220611916264090177133655394524959742351046698872785829307568438X30
Only the rst section is odd, but the remainder can be contained in the rst digit of the next repe
tition. Therefore, the nal solution is
N = 2 * 3 * 23 *
280008051529790660225544283413848631239935587761672718196457326892109...
500805152979066022544283413848631239935587761674718196457326892109X29...
50080515297906602254436392914653784219...
001610305958132045488566827697262479871175523349436392914653784219X30
Editors Comments: We thank Dr. George Berzsenyi, the creator of the USAMTS and now
retired from Rose-Hulman Institute of Technology to the mountains of Colorado, for this beautiful
problem.
The solutions to these problems give us an opportunity to briey introduce congruences and mod
ular arithmetic, and state an important theorem from Number Theory. Our goal is to introduce
some of the notation and terminology, and briey introduce this subject to young mathematicians
seeing this for the rst time, and possibly help them understand the proofs.
Two integers a and b are said to be congruent modulo m if m divides a b . We write

a b(modm) , and say that a is congruent to b modulo m. The positive integer m is called the
modulus, and the relation is called a congruence. For example,
86 6 ( mod10) 1 16 ( mod5) 3 4 ( mod7) 88 0 ( mod11)
Observe that if you divide a by m and get remainder r, then a r (modm) . If a b(modm) and
c d (modm) , then a + b c + d (modm) and a b c d (modm) .
This gives us an interesting new arithmetic system where we can add, subtract, and multiply. We
refer to this arithmetic system as arithmetic modulo m. [Do you see a connection between
clock arithmetic and arithmetic modulo 12?]
If m is a prime integer, then we can also divide by nonzero elements. For example, observe that
1
2 4 1 ( mod 7) , so we think of 2 as the multiplicative inverse of 4; that is, 2 --- in this arith
4
metic system. Thus, dividing by 4 is the same as multiplying by 2 when you are working modulo
7. Whenever m is a prime number, and a is not congruent to 0 modulo m, then we can nd a sec
ond number b such that a b 1 ( modm) ; dividing by a is the same as multiplying by b, so we
now can divide by any nonzero element a. Now we have an arithmetic system where we can
add, subtract, multiply, and divide by nonzero numbers.
This sounds good, you say, but given a, how do you nd the corresponding b? The process is
straightforward and efcient, and uses the extended Euclidean algorithm for nding Greatest
Common Divisors, an algorithm already familiar to you, perhaps. If p is prime and p does not
divide a, then GCD(p,a) = 1. Therefore there are integers r and s such that 1 = ar + ps. So
1 ar + ps(mod p) ar (mod p) , and r is the multiplicative inverse of a, as desired.
1
As an example, we will nd 19 modulo 61. That is, we nd an integer r such that
19 r 1 ( mod61 ) . First we use the division algorithm, divide 61 by 19, and get remainder 4:
4 = 61 (3 19 )
Next, divide 19 by 4 and obtain remainder 3:
3 = 19 (4 4)
Finally
1=4-3
So 1 is the Greatest Common Divisor of 61 and 19. Substituting back into the above equations we
obtain:
1=4-3
= 4 (19 4 4) = 5 4 1 19
= 5 ( 61 3 19 ) 1 19
= 16 19 + 5 61
45 19 + 5 61 ( mod61 )
45 19 ( mod61 ) .
1
Thus, 19 modulo 61 is 45.
Now we state a beautiful theorem. If n is a positive integer, then an corresponds to

a a a a with n factors of a. We dene a0 = 1. So we also know how to take positive
exponents.
p1
Fermats Little Theorem: If p is a prime integer, and p does not divide a, then a 1 ( mod p) .
For example,
2 6 22 10
5 1 ( mod 3 ) 2 1 ( mod7 ) 123456 1 ( mod23 ) ( 3 ) 1 ( mod11 )
This theorem is a key part of Luke Gustafsons proof above.
3/1/11. Triangle ABC has angle A measuring 30 , angle B measuring 60 , and angle C measur
ing 90 . Show four different ways to divide triangle ABC into four triangles, each similar to
triangle ABC, but with one quarter of the area. Prove that the angles and sizes of the smaller
triangles are correct.
Solution 1 by Neeraj Kumar (12/NC): Let us start with triangle ABC having sides AB, BC, and
3
AC of lengths 2, 1, and 3 , respectively. The area of this triangle is ------- . Therefore, each of the
2
3 3
smaller triangles must have area ------- and sides of lengths 1, 1/2, and ------- .
8 2
Here is the rst way to divide ABC:
B
We choose E to be the midpoint of BC, and connect

it to AB (parallel to AC). Since angle B is common
to both triangles ABC and DBE, and E is also a right D
E
angle, the two triangles are similar. Also, since BE is

1/2, DE must equal ( 3) 2 , and the area of DBE is
A F C
3
indeed ------- . If we now choose F to be the midpoint
8
of AC, then again we create a similar situation with
shared angle A of the two triangles ADF and ABC, and so ADF is also a valid smaller triangle.
Now DC = DB = DA, by Side-Angle-Side congruence in the triangles BED and CED (BE = EC,
angle BED = angle DEC, DE = DE), and in triangles ADF and CDF. Therefore, each of DEC and
DFC is also a valid smaller triangle.
For the second way to divide the triangle, we can simply choose the other diagonal of rectangle
DECF, i.e. FE rather than DC.
For the nal two ways, we can see that triangle DBC is equilateral (because each side has length
1). Therefore, the altitude DE can be moved to any one of the three sides. Thus the four different
ways are given here:
Editors Comments: This problem, along with many earlier USAMTS problems, was proposed
by Dr. Erin Schram of the National Security Agency. We are also thankful to him for his thought
ful evaluation of the problems considered for the program.
4/1/11. There are 8436 steel balls, each with radius 1 centimeter, stacked in a tetrahedral pile,
with one ball on top, 3 balls in the second layer, 6 in the third layer, 10 in the fourth, and so on.
Determine the height of the pile in centimeters.
Solution 1 by Shaili Jain (12/MI): The number of balls in each level of the pyramid are: 1, 3, 6,
10, 15, 21, ..., n(n+1)/2.
n
i(i + 1)
-----------------
2
= 8436
i=1
n = number of layers in the pyramid. We want to solve for n.
n 2 n n
1 1 n ( n + 1 )(2n + 1 ) n ( n + 1 )
i i
i +i 2
8436 = ------------ = --- + = --- ----------------------------------------- + -------------------
2 2 2 6 2
i=1 i=1 i = 1
8436 = --- n ( n + 1 )(n + 2 ) = n + 2 , a binomial coefcient.

1
6 3
n = 36.
The sides of the tetrahedron are
35 2 = 70 cm each (they go through the 30
centers of the outer balls). 120
sin 120 sin30

------------------- = ---------------- by Law of Sines
70 x
70
3 Base of
x = ------------
3 tetrahedron
70 3 2
70 ------------- = ------------
2 70 6
Tetrahedron height =
3 3
70 6
The height of the pile of balls is ------------- + 2 cm. because it is +1 cm. from the center of the top
3
ball to the top, and +1 cm. from the center of a bottom ball to the bottom.
Solution 2 by Wendy Pang (11/CA): 8436 is the sum of 36 trian

gle numbers, meaning that there are 36 layers of balls. The centers
of four adjacent balls form a tetrahedron (equilateral) with slant
2 6 3 x 2
height of 2, which means that the height of the tetrahedron is ---------- .
3
Height of pile = (layers - 1)(height of tetrahedron) + 2 radius .
= (36 1 ) ---------- + 2 ( 1 )
2 6
3 3
70 6
= ------------- + 2 cm.
3
Editors Comments: This problem was suggested 20 years ago by Professor Endre Hdi, long
time leader of Hungarys successful teams to the IMOs.
5/1/11. In a convex pentagon ABCDE the sides have lengths 1, 2, 3, 4, and 5, though not neces
sarily in that order. Let F, G, H, and I be the midpoints of the sides AB, BC, CD, and DE,
respectively. Let X be the midpoint of segment FH, and Y be the midpoint of segment GI.
The length of segment XY is an integer. Find all possible values for the length of side AE.
Solution 1 by Melody Chan (12/NY): Use the notation A = (4a1, 4a2) to mean the coordinates
of the point A are (4a1, 4a2) .
Let A = (4a1, 4a2), B = (4b1, 4b2), C = (4c1, 4c2), D = (4d1, 4d2), and E = (4e1, 4e2) .
Then F = (2a1+2b1, 2a2+2b2)
G = (2b1+2c1, 2b2+2c2)
H = (2c1+2d1, 2c2+2d2)
I = (2d1+2e1, 2d2+2e2)
and X = (a1+b1+c1+d1, a2+b2+c2+d2)

Y = (b1+c1+d1+e1, b2+c2+d2+e2)
Thus
2 2
XY = [(b 1 + c 1 + d 1 + e 1 ) (a 1 + b 1 + c 1 + d 1 )] + [(b 2 + c 2 + d 2 + e 2 ) (a 2 + b 2 + c 2 + d 2 )]
2 2
= (e 1 a 1 ) + (e 2 a 2 ) which is given to be an integer.
Also
2 2
AE = (4e 1 4a 1 ) + (4e 2 4a 2 )
2 2
= 4 ( e 1 a 1 ) + ( e 2 a 2 ) = 4 XY
So AE is a multiple of 4. Thus AE = 4 is the only possible solution.
Solution 2 by Kartik Lamba (12/IL): The

solution becomes simple when one uses vec
tors. Consider the pentagon shown at right with
any of the given following dimensions:
B G C
A+B B+C
F = -------------- G = --------------
2 2
F X
H
C+D D+E Y
H = -------------- I = --------------
2 2 0
A
D
A+B C+D B+C D+E
-------------- + ------------- -------------- + -------------- I
2 2 2 2
X = ----------------------------------- Y = ---------------------------------- E
2 2
A+B+C+D B+C+D+E AE
X Y = ----------------------------------- ----------------------------------- = ------------
4 4 4
4 XY = AE
In order for both lengths XY and AE to have integer values between 1 and 5, inclusive, XY must
be 1 and AE must be 4. Thus
AE = 4 .
Solution 3 by Dmitry Portnyagin (12/NY): We draw

line segment BE, and mark its midpoint Z. Now, G and B G C
H are midpoints of the sides of BCD , Z and I are
midpoints of the sides of BED . Thus X
F H
GC H BCD , ZEI BED , and in both cases Y
the sides of the smaller triangle are half the size of the Z
A
sides of the larger triangles. Thus, ZI = GH = BD/2, D
and ZI || GH || BD . Thus, GHIZ is a parallelogram.
I
In a parallelogram the diagonals bisect each other.
Thus Y is the midpoint of HZ. E
F and Z are midpoints of AB and BE, respectively.
Thus, once again, FZ = AE/2. X and Y are midpoints
of FH and HZ, and once again XY = FZ/2. Thus, XY =
AE/4. But XY has to be an integer, and out of the ve possible lengths, only 4 is divisible by 4.
AE = 4
Editors Comments: This clever problem was created for USAMTS by Professor Gregory Galp
erin, who is the author of a soon-to-be published compendium of problems for the famous Mos
cow Mathematical Olympiads.

1/2/11. The number N consists of 1999 digits such that if each pair of consecutive digits in N
were viewed as a two-digit number, then that number would either be a multiple of 17 or a
multiple of 23. The sum of the digits of N is 9599. Determine the rightmost ten digits of N.
Solution 1 by Ann Chi (8/IN):

Multiples of 17: 17, 34, 51, 68, 85. Multiples of 23: 23, 46, 69, 92.
68,85,51,17 can be used only at the very end because no other numbers start in 8, 5, 1, or 7.
Another possible sequence of numbers is 23,34,46,69,92,23,... starting with any of these numbers.
2 + 3 + 4 + 6 + 9 = 24
1999 5 = 399 R4
399 24 = 9576
9599 9576 = 23
So there must be an end sequence of 4 numbers adding up to 23. Since picking any four of the
original repeating sequence will not add up to 23, we have to use part of the sequence 923468517
(there must be a 46 before the 68).
4+6+8+5 = 23. This means the sequence of numbers must be
469234692346...4685.
The last ten digits of this sequence are 3469234685.
Solution 2 by Nick Masiewicki (11/MD): Consider all two-digit multiples of 17 and 23:
17, 23, 34, 46, 51, 68, 69, 85, 92.
Each digit 1-9 is represented exactly once as a 1s
digit of a number on that list. Thus, one and only 46 68 85 51 17
one of those numbers may precede a number on
that list in N in the sense that ab precedes bc in 34
abc. Thus we know 51 precedes 17, 85 precedes 69
51, and so on (see diagram at right). Thus, no
matter what the rightmost digit of N is, well 23
92
always get to the sequence 69234 (read left to
right) repeated for the rest of N. Since that
1999 4
sequence is at most four steps away from any digit, that sequence must repeat --------------------- = 399
5
times. The sum of its digits is 24. Thus we have accounted for 399 24 = 9576 of the sum.
That leaves 23 left. We must nd a four-member subset of consecutive digits in 346923468517
whose sum is 23. 4685 is the only such subset. Thus the rightmost ten digits of N are
3469234685.
Editors Comment: Once again, we are indebted to Professor Sndor Rka, the founding editor
of Hungarys Abacus, for the idea of this problem. For an English translation of the problems and
for their solution, the reader is referred to: http://www.gcschool.org/abacus.html. This competi
tion is open to all students in grades 3 to 8.
2/2/11. Let C be the set of non-negative integers which can be expressed as 1999s + 2000t,
where s and t are also non-negative integers.
(a) Show that 3,994,001 is not in C.
(b) Show that if 0 n 3,994,001 and n is an integer not in C, then 3,994,001 - n is in C.
Solution by David Gaebler (12/IA):

(a) Proof by contradiction: Suppose instead that 3,994,001 is in C, so
1999s + 2000t = 3994001 . Taking both sides mod 1999, t 1998 ( mod1999 ) . Since t 0 , this
implies t 1998 , so 2000t 3996000 . But then s must be negative, contrary to the hypothesis
that 3994001 is in C. Thus, 3,994,001 is not in C.
(b) Let n = 1999a + b where 0 b 1998 . Since n = 1999(a b) + 2000b , if n C , then

b >
a .
3994001 n = (2000 1998 1999 ) (1999a + b)
2000 ( 1998 b) + 1999(b a 1 )
Since b > a , b
a 1 0 . Since b 1998 , 1998 b 0.
Therefore, 3,994,001 - n is in C.
Editors Comment: Appreciation is hereby expressed to Dr. Peter Anspach of the National
Security Agency for proposing this problem.
3/2/11. The gure on the right shows the map of Squareville, where B
each city block is of the same length. Two friends, Alexandra and
Brianna, live at the corners marked by A and B, respectively. They
start walking toward each others house, leaving at the same time,
walking with the same speed, and independently choosing a path to
the others house with uniform distribution out of all possible mini-
mum-distance paths [that is, all minimum-distance paths are equally
likely]. What is the probability they will meet?
A
Solution 1 by Seth Kleinerman (12/NY): 35 125 290 B
The number of paths from A to B is equal to
the number of paths from B to A. To determine
this number, we note that the number of paths 35 90 165
from A to a vertex is equal to the sum of the
number of paths from A to the vertex below it 1 5 15 35 55 75
and the number of paths from A to the vertex
to the left of it. So starting at A, we can induc
tively nd the number of paths from A to B as 1 4 10 20 20 20
shown on the left.
1 3 6 10
There are 290 possible paths from A to B. If
the two girls are to meet somewhere in the
middle, they must have each traveled 5 1/2 1 2 3 4
blocks, since the total trip is 11 blocks and
they start at the same time and walk at the
same speed. They can only meet in the middle 1 1 1 1
of one of the four marked streets in the above A
diagram (not at a street corner), and if they
cross the same marked street they must meet. We shall call the streets 1, 2, 3, and 4 from top-left
to bottom-right.
Let us repeat the above inductive process to 1 1 1

determine the number of paths from both A B
and B to the corners lying on either side of
these streets (see diagram at right). There are 3 2 1
5 6 = 30 paths going through street 1,
6 10 = 60 paths going through street 2, 1 5 6 6 3 1
10 10 = 100 paths going through street 3,
and 10 10 = 100 paths going through street 1 4 10 10 4 1
4. So the probability of crossing each of these
streets is 30/290, 60/290, 100/290, and 100/
290 for streets 1, 2, 3, and 4 respectively. 1 3 6 10
The probability that both girls will choose to 1 2 3 4

2
cross at street 1 is (30 290) , that both will
2
choose to cross at street 2 is (60 290) , and 1 1 1 1
so on. So the probability that they meet A
(equivalently stated, the probability of both choosing the same numbered street to cross) is
2 2 2 2
30 + 60 + 100 + 100 245
----------------------------------------------------------
2
- = --------
290 841
Solution 2 by Andrew Dudzik (11/CA):
Lemma: The number of equal-distance paths from one corner of an a by b rectangle is
a + b
.
a
Proof of lemma: The total distance to be traveled is a + b. At any of these a + b steps, we can
choose to go down instead of left, but we must do so exactly a times (or b times, depending on the
orientation). So the total number is
a + b
.
a
Problem solution: See the diagram at the right. Because Alexandra and B
Brianna are moving at the same speed and start at the same time, they can
only meet at one of the streets labeled W, X, Y, and Z. The number of W
paths that go through any given street is equal to the number that go from X
A to one side of it times the number of paths that go from B to the other Y
Z
side of it. So the number of paths through W is = 30 . The
5 4
1 2
number through X, Y, and Z are = 60 , = 100 , and A

5 4 5 5
2 2 2 2
5 5
= 100 , respectively. Since these choices are mutually exclusive, the probability that A
2 2
will select the same street as B is
30 2 60 2 100 2 100 2
-------- 245
- + --------- + --------- + --------- = ---------
290 290 290 290 841
Solution 3 by Megan Guichard (12/WA):

Label the coordinates of the corners on the map as if they were points in the plane, using one city
block as a unit, with the origin at A. Then A has coordinates (0,0) and B has coordinates (5,6).
Since all minimum-length paths are 11 units long, Alexandra and Brianna will meet only if their
paths contain the same block as the sixth block each passes over; that is, they must both choose
the segment from (1,4) to (2,4), the segment from (2,3) to (2,4), the segment from (2,3) to (3,3), or
the segment from (3,2) to (3,3). Since there are the same number of paths going from A to B as
going from B to A, it is sufcient to nd the probability of passing over a given one of these four
crucial segments while traveling from A to B. If we start from (1,4) and go through (2,4), there is
one way to get to the point (3,4), one way to get from there to point (3,5), and one way to get from
(3,4) to point (4,4). After this there is one way to get to (3,6), two ways to get to (4,5), and one
way to get to (5,4). Then there are three ways to get to each of (4,6) and (5,5), and nally, six
ways to get to B from (1,4). Carrying out similar calculations for the other three crucial segments,
there are 6 ways to get from B to (2,4), 10 ways to get from B to (3,3), 5 ways to get from A to
(1,4), 10 ways to get from A to (2,3), and 10 ways to get from A to (3,2).
Therefore, there are a total of 5 6 = 30 ways to go from A to B passing through the segment
from (1,4) to (2,4); there are a total of 10 6 = 60 ways to go from A to B passing through the
segment from (2,3) to (2,4); there are 10 10 = 100 ways to go from A to B passing through the
segment from (2,3) to (3,3); and 10 10 = 100 ways to go from A to B passing through the seg
ment from (3,2) to (3,3). There are 30 + 60 + 100 + 100 = 290 ways to get from A to B.
2
30
We then calculate the relevant probabilities. There is a ----------2- chance that both Alexandra and
290
2
60
Brianna will pass through the rst crucial segment, a ----------2- chance that both will pass through the
290
2 2
100 100
second segment, a ----------2- chance that both will pass through the third segment, and a ----------2- chance
290 290
that both will pass through the fourth segment.
245
Adding these probabilities together results in --------- , the chance that Alexandra and Brianna will
841
meet.
Editors Comment: We thank our problem editor, Dr. George Berzsenyi, for this problem.
4/2/11. In PQR , PQ = 8, QR = 13, and RP = 15. Prove that there is a point R

S on line segment PR , but not at its endpoints, such that PS and QS are also
integers.
Solution 1 by Frank Chemotti (11/WI): By the law of cosines: S

2 2 2
13 = 15 + 8 2 15 8 cos RPQ P Q
cos RPQ = 0.5
R Let x be the length of segment PS and y be the length of

segment QS. By the law of cosines:
2 2 2
y = x + 8 2 x 8 cos RPQ
2 2 2
y = x + 8 2 x 8 (0.5)
2 2
15-x y = x 8 x + 64
13 2
y = x 8 x + 64
(0 < x < 15)
Integer solutions exist for this equation: (3, 7), (5, 7),
S and (8, 8).
x y
Therefore, PS and SQ are both integers when point S is
P 8 Q 3, 5, or 8 units away from point P.
Solution 2 by Daren Zou (12/NC): Draw R
circle Q with radius 8. Since RPQ is not
opposite the longest side of the triangle, it
must be acute. Intersect PR at S. 5
15
By Power of the Point Theorem
RA RB = RS RP A
5 21 = RS 15 S
RS = 7 8
So PS = 8 = QS are both integers.
8
Editors Comment: Although this solution
does not give all solutions, it is unique and
quite interesting. To nd the remaining two
solutions, draw circle Q with radius 7 (rather P 8 Q
than 8) and show RP intersects the circle in
two points S and S. Then, to show the dis
tances from P to the two points are 3 and 5,
apply the Power of the Point Theorem twice,
using lines PQB and RSS.
Using this theorem, can you show these are

all the points?
B
This problem was also created by Dr. George
Berzsenyi.
5/2/11. In ABC , AC > BC, CM is the median, C
and CH is the altitude emanating from C, as
shown in the gure on the right. Determine the
measure of MCH if ACM and BCH
each have measure 17 .

A M H B
Solution 1 by Yuran Lu (11/ME): Let us prove the
general case:

Lemma. If ACM = BCH = and MCH = , then 2 + = --- .
2
Proof of lemma. Let CH = 1. AM = AH - MH, BM = BH + MH.

AH = 1 tan ( + ) , BH = 1 tan , and MH = 1 tan .
Since AM = BM we have:
1 tan ( + ) 1 tan = 1 tan + 1 tan
tan + tan
-------------------------------- tan = tan + tan
1 tan tan
tan + tan = ( tan + 2 tan )(1 tan tan )
2 2
tan + tan = tan + 2 tan tan tan 2 tan tan
2 2
tan tan tan 2 tan tan = 0
2
1 tan 2 tan tan = 0
2
1 tan
tan = -------------------------- = --------------- = cot2 = tan --- 2
1
2 tan tan2 2

= --- 2
2
This completes the proof of the lemma.
Problem solution: If ACM = BCH = 17, then by the lemma

MCH = 90 2 ( 17) = 56 .
Solution 2 by Jacqueline Ou (11/MA): C

Let MCH = x . Then HBC = 73 and
CAH = (73 x) . We also know that AM = BM
since M is the midpoint of AB.
A M H B
Applying the law of sines to ACM and to BCM ,
we obtain
sin 17 sin (73 x) sin (17 + x) sin ( 73 )
------------- = --------------------------- and ---------------------------- = -----------------
AM CM AM CM
So
sin 17 sin (73 x)
---------------------------- = --------------------------
sin (17 + x) sin 73
Expanding sin (17 + x) and sin (73 x) , and cross multiplying, we obtain
( sin73 cos 17 sin 17 cos 73 ) ( sin x cos x) + sin 17 sin73 cos 2 x = sin 17 sin 73
2 2
sin56 sin x( 1 sin x) + sin17 sin 73 2 sin17 cos 17 sin x = sin 17 sin 73
2
sin56( 1 sin x) sin 34 sin x = 0
2 2 2 2 2
sin 56 sin 56 sin x = sin 34 sin x
2 2
sin 56 = sin x
x = 56 .
Editors Comment: We are thankful to Professor Gregory Galperin for this problem and for his
many other contributions to the USAMTS.

1/3/11. We dene the repetition number of a positive integer n to be the number of distinct digits
of n when written in base 10. Prove that each positive integer has a multiple that has a repeti
tion number less than or equal to 2.
Solution 1 by Peter Ruse (12/NY): Since there are nitely many possible remainders on division
by n, two of the numbers 1, 11, 111, ..., must have the same remainder. Their difference, a num
ber of the form 111...100...0, is a multiple of n and has repetition number of 2.
1
Solution 2 by Michael Catlin (10/IN): We can express --- as the repeating decimal
n
0.b1b2b3...bka1a2...am
m
1 (10 1 )(b 1 b 2 b k ) + (a 1 a 2 a m )
k m
-.
Then, --- = --------------------------------------------------------------------------------------
n 10 (10 1 )
Thus, n divides 10k(10m - 1), a number with only the digits 9 and 0.
2/3/11. Let a be a positive real number, n a positive integer, and dene the power tower an
recursively with a1 =a, and a(i+1) = a(a) for i = 1, 2, 3, . For example, we have
4
(4 ) 256
43 = 4 = 4 , a number which has 155 digits. For each positive integer k, let xk
denote the unique positive real number solution of the equation xk = 10(k+1). Which is
larger: x42 or x43?
Solution 1 by Lisa Leung (9/MD): By denition, xkk = 10(k+1), where both xk and 10 are
positive real numbers. Thus, xk > 10 for all k.
x
x4243 = x42( 4242) = x421043 > 101043 = 1044 = x4343.
Thus, x42 is greater than x43.
Editors comments: This interesting problem was suggested by Roger Pinkham of Stevens Insti
tute of Technology.
3/3/11. Suppose that the 32 computers in a certain network are numbered with the 5-bit integers
00000, 00001, 00010, ..., 11111 (bit is short for binary digit). Suppose that there is a one-way
connection from computer A to computer B if and only if A and B share four of their bits with
the remaining bit being 0 at A and 1 at B. (For example, 10101 can send messages to 11101
and to 10111.) We say that a computer is at level k in the network if it has exactly k 1s in its
label (k = 0, 1, 2, ..., 5). Suppose further that we know that 12 computers, three at each of the
levels 1, 2, 3, and 4, are malfunctioning, but we do not know which ones. Can we still be sure
that we can send a message from 00000 to 11111?
Solution 1 by Jennifer Balakrishnan (10/GU):

Computer at level 0: 00000
Computers at level 1: 00001, 00010, 00100, 01000, 10000
Computers at level 2: 11000, 10100, 10010, 10001, 01100, 01010, 01001, 00110, 00101, 00011
Computers at level 3: 00111, 01011, 01101, 01110, 10011, 10101, 10110, 11001, 11010, 11100
Computers at level 4: 11110, 11101, 11011, 10111, 01111
Computer at level 5: 11111
Assuming that malfunctioning means will not accept information, then we cannot be sure that
we can send a message from 00000 to 11111.
At level 0, the computer can send the information I to any of 5 computers. At level 1, each of
these ve computers can send I to four computers on level 2. At level 2 each computer can send I
to three computers on level 3. On level 3 each computer can send I to two computers on level 4.
On level 4 each computer can send I to 1 computer on level 5.
Suppose at level 1, three computers are affected. I still can be sent to the other two computers at
level 1. These two computers, P and Q, can normally send to seven level 2 computers; assume
among these seven that the three that P does not send to are malfunctioning, so Q sends I to one
good level 3 computer, and P sends to this one good level 3 computer (call it R) and three other
good level 3 computers (call them S, T, and U). R sends to three level 3 computers and together,
R, S, T, and U send to six level 3 computers. Assume that among these six, the computers R does
not send to are malfunctioning. Now on level four, only the three computers which are grand
children of R will be sent I. Assume these three malfunction. So the information stops here on
level four and cannot be transmitted to level 5.
Solution 2 by Asher Walkover (12/NY): Rather than limit my answer to the case with 25 com
puters in the network, I will prove a more general result. Consider any system of 2n computers
(where 4 < n), each computer given a binary name, networked as described. I shall prove that it
will always be possible to select n-2 computers on levels 1, 2, n 2 , and n 1 , such that, if they
are malfunctioning, a message sent from the level 0 computer (000...) will not get through to the
level n computer (111...).
On level 1, each computer has a single 1 in its name. Its name is n digits long so there are n pos
sible ways a level 1 computer can be named, which means there are n computers on level 1.
Since there are only n level 1 computers, of which n 2 are malfunctioning, only two level 1
computers will be working. Let those two computers be the computer with 1 as rst digit (call it
computer A) and the computer with 1 as the second digit (call it computer B).
Since only A and B are functioning at level 1, all level 2 computers who receive the message
receive it from A or B or both. From computer B we can access all n 1 level 2 computers with
a 1 as their second digit and one more 1 somewhere else in the name. Consider, however, that one
of these computers (the one whose rst two digits are 1s, the rest 0s) can be reached by both
computers A and B. Thus there are only n 2 level 2 computers which are reachable only by
computer B. Let those n 2 level 2 computers be malfunctioning. It follows, therefore, that all
functioning level 2 computers which can receive a message are the ones that are linked to com
puter A. A has a 1 as the rst digit, and so must all the level 2 computers linked to A. Thus, the
only computers on level 2 that receive the message are the computers with 1 as the rst digit.
After the message has traveled through all the levels (if any exist) between level 2 and level
,n 2 , all level n 2 , computers with a 1 as rst digit may have the message. If the computer
does not have a 1 as the rst digit, it cannot have the message, since the only level 2 computers to
receive and pass on the message had a 1 as rst digit. Thus, all level n 2 computers who have
the message have the rst two digits either 11 or 10. Consider the later case. In the remaining
,n 2 digits of those computers names, there is only one 0. There are n 2 such computers
(computers on level n 2 that begin with 10). If these n 2 computers are all malfunctioning,
then the only level n 2 computers to receive the message will be those whose rst digits are 11.
Since all the level n 2 computers with the message begin with 11, they can only send a message
to a level n 1 computer that begins with a 11.
On level n 1 , each computer has n 1 , 1s and one 0. There are n such computers. n 2
computers are malfunctioning at that level, so only two are working. However, if we let the two
working computers be the ones whose rst digits are 10 and 01 (with all the rest of the digits 1s)
then neither will be able to receive the message froma level n 2 computer that begins with 11.
But the level n 2 computers that begin with 11 are the only functioning level n 2 computers
with the message, so neither one of the working level n 1 computers will receive the message,
and it will not be passed on to the computer on level n.
If we let n = 5, then we have 3 computers malfunctioning on levels 1, 2, 3, and 4, as stipulated, so

we see it is not necessarily possible that a message sent from the level 0 computer will reach the
level n computer.
Editors comments: This problem is a special case of a theorem of our Problem Editor, Bela
Bajnok of Gettysburg College, and Shahriar Shahriari of Pomona College. The general result,
which appeared in the Journal of Combinatorial Theory, Series A, in 1996, says that it is possible
to remove n 2 nodes at each of the levels 1, 2, ..., n 1 , of the Boolean lattice of order n (our
network) in such a way that the source gets disconnected from the target, and that this is not pos
sible if we remove n 3 at each level.
4/3/11. We say a triangle in the coordinate plane is integral if its three vertices have integer
coordinates and if its three sides have integer lengths.
(a) Find an integral triangle with perimeter of 42.
(b) Is there an integral triangle with perimeter of 43?
Solution 1 by Clayton Myers (11/NJ): (a) The triangle with vertices (0,0), (14,0), and (9,12) is
an integral triangle with perimeter 42 and sides of lengths 13, 14, and 15.
(b) Since an integral triangle can be translated horizontally and vertically in integral steps without
effect, let us assume that a hypothetical integral triangle with perimeter 43 has vertices (0,0),
(x1,y1), and (x2,y2). Now, since all side lengths are integers and their sum is 43, we know that
either all three sides are odd, or two are even and one is odd. Therefore, since their squares must
have the same parity, the sum of their squares must be odd. This number can be expressed as
2 2 2 2 2 2 2 2 2 2
x 1 + y 1 + x 2 + y 2 + ( x 1 x 2 ) + ( y 1 y 2 ) = 2( x 1 + y 1 + x 2 + y 2 x 1 x 2 y 1 y 2 )
which is even. This contradiction means that our assumption that an integral triangle with perim
eter 43 exists is false.
Solution 2 by Robert Kotredes (12/ME): (a) The triangle with vertices (0,0), (12,9), and
(12,16) has sides of length 20, 15, and 7, so is an integral triangle with perimeter 42.
(b) By Picks Theorem, the area of any convex polygon with integer coordinates is given by
A = I + --- 1 , where I is the number of integer points interior to the polygon, and B is the
B
2
number of integer points along its border. This is a rational number.
By Herons Formula, the area of a triangle with sides of lengths a, b, and c, is given by
A = s ( s a)(s b)(s c) , where s = (a + b + c)/2. For any triangle with perimeter 43 this
43 ( 43 2a)(43 2b)(43 2c) 1
becomes A = ---------------------------------------------------------------------------- = --- 43 ( 43 2a)(43 2b)(43 2c) .
16 4
Observe 43 is prime. Also (43 -2a), (43 - 2b), and (43 - 2c) are positive integers smaller that 43,
so dont have 43 as a factor. The area of the triangle is therefore N 43 where N is a number.
Hence the area is an irrational number. So if there were an integral triangle with perimeter 43 it
would have both an irrational and a rational area, a contradiction. Thus, there is no integral trian
gle with perimeter 43.
Editors comments: This problem was derived from a problem proposal by Robert Ward of
NSA. He came upon the problem as a result of his work with the Ask Dr. Math program where
he serves as a consultant to deal with questions submitted by high school students. For more Ask
Dr. Math details see http://forum.swarthmore.edu/dr.math.
5/3/11. We say that a nite set of points is well scattered on the surface of a sphere if every open
hemisphere (half the surface of the sphere without its boundary) contains at least one of the
points. The set { (1,0,0), (0,1,0), (0,0,1) } is not well scattered on the unit sphere (the sphere
of radius 1 centered at the origin), but if you add the correct point P it becomes well scattered.
Find, with proof, all possible points P that would make the set well scattered.
Solution 1 by Alexey Gorshkov (12/MA):

z
Answer: All points on the sphere that have all
three of their coordinates negative. That is, (0,0,1)
one eighth of the unit sphere without its
boundary. (See Figure at right.)
Proof (it consists of two sub-proofs):
Proof that all points not belonging to the (0,1,0)

region do not make the set well-scattered: y
Since the point is not in the region described
above, at least one of its coordinates is positive
or zero.Then the open hemisphere that has all (1,0,0)
the points with this coordinate negative will
not contain any of the four points, which
proves that all points not belonging to the
region do not make the set well scattered. x
Proof that all points belonging to the region do make the set well scattered: Lets suppose we
found an open hemisphere H that does not contain the three given points and some point P in the
region described above in the answer. Since all three coordinates of P are negative, the point P
diametrically opposite to P lies in the triangular region determined by the three points given ini
tially. Since any open hemisphere that contains a point in this triangular region must contain at
least one of its vertices(1), the open hemisphere H does not contain the point P. However two
diametrically opposite points (P and P) can be both outside an open hemisphere only if they lie
on its boundary, which cannot be the case(2). So H does not exist. So all the points belonging to
the region described in the answer do make the case well scattered.
(1)Ifthis is not obvious, then here is a proof. Lets suppose there is an open hemisphere G con
taining a point X inside the triangular region and not containing any of the three vertices.
Since G does not contain any of the vertices, it does not contain any points of the sides (this is
because two circles of radii equal to that of the sphere either coincide or intersect in two diametri
cally opposite points). Since G does not contain any points on the boundary of the triangular
region and does not contain a point inside, it must be completely inside, which is impossible.
(2)If
this is not obvious, then the proof is identical to the one in the previous paragraph with the
only change that G is a closed hemisphere (so that it contains P), which does not make any differ
ence in the proof.
Solution 2 by Daniel Kane (9/WI): The set of points that would make it well scattered are the
points inside the spherical triangle with vertices (-1,0,0), (0,-1,0), and (0,0,-1); call this triangle A.
You can see that any point outside of A would not make the set well scattered by looking at the
hemispheres determined by the great circles through (1,0,0) and (0,1,0), (1,0,0) and (0,0,1), and
(0,1,0) and (0,0,1).
Any point within the spherical triangle A is diametrically opposite a point inside the spherical tri
angle (1,0,0), (0,1,0),(0,0,1). Therefore, the tetrahedron determined by the four points in the set
would contain the center of the sphere. Given any four points on the same hemisphere (of the
ball) the tetrahedron they determine would also be contained in that hemisphere. Therefore if the
tetrahedron contains the center of the sphere, the points are well scattered.
Therefore, given P is not in the spherical triangle A, the points are not well scattered, and given
that point P is in triangle A, {(1,0,0), (0,1,0), (0,0,1), P} is well scattered.
Editors comments: This attractive problem is based on the work of Andrew Lenard, a retired
Hungarian mathematics professor at Indiana University.

1/4/11. Determine the unique 9-digit integer M that has the following properties: (1) its digits
are all distinct and nonzero; and (2) for every positive integer m = 2, 3, 4, ..., 9, the integer
formed by the leftmost m digits of M is divisible by m.
Solution 1 by Lisa Leung (9/MD): Let the 9-digit integer M be represented by

abcdefghi where each letter represents a unique digit from the set of {1, 2, ...9}. c d
Since abcde must be divisible by 5, e can only be 5 or 0. However the digits are
non-zero, so e = 5. b, d, f, and h are even numbers since they are the last digits of 1 2
numbers that are divisible by 2, 4, 6, and 8. When examining combinations of cd
1 6
and fgh, where cd will be divisible by 4 and fgh will be divisible by 8, one can con
clude from Table 1 (at right) that d is 2 or 6, and h is 2 or 6, since c and g cannot be 2 4
even. The possible values for cd are 12, 16, 32, 36, 52, 56, 72, 76, 92, and 96.
2 8
The possible values for fgh are 416, 432, 472, 496, 816, 832, 872, 896. Table 2 3 2
(below) shows the possible digits of each position after the initial analysis. a+b+c
must be divisible by 3 and d+e+f must be divisible by 3. The only possibilities of 3 6
a, b, c, d, e, and f left are given in Table 3. 4 8
a b c d e f g h i a b c d e f ghi 5 2
1 1 2 5 1 2 1 1 4 7 2 5 8 963 5 6
3 4 3 4 3 3 7 4 1 2 5 8 963 6 4
7 7 6 7 6 7 1 8 3 6 5 4 729 6 8
9 8 9 8 9 9 3 8 1 6 5 4 729 7 2
Table 2: 9 8 1 6 5 4 729 7 6
Table 3: 8 4
Among the ve possibilities, only 3816547 is divisible by 7. 9 2
Thus, M = 381654729. 9 6
Table 1:
Solution 2 by Zhihao Liu (10/IL): Answer: 381654729.
Let abcdefghi =M, where letters a through i represent digits 1 through 9, let abcde
represent a ve digit number, and so on. Since 5 | abcde (read as, 5 divides
abcde), we conclude e = 5. Note that b, d, f, and h are all even, thus a, c, g, and i are odd. Since
abcd and abcdefgh are divisible by 4 and their tens digits are odd, we conclude that d and h are 2
and 6 (in some order), and b and f are 4 and 8 (in some order). Also, 3 | a+b+c, 3 |
a+b+c+d+e+f, and 3 | a+b+c+d+e+f+g+h+i = 45, so 3 | d+e+f, and 3 | g+h+i. From this we
see that def is either 258 or 654. Since 8 | fgh and f is even, we conclude that 8 | gh. With these
restrictions, we eliminated all 9-digit numbers except: 147258963, 183654729, 189654327,
381654729, 741258963, 789654321, and 987654321. However, only 381654729 possesses the
property that the number formed by its rst seven digits is divisible by 7.
Therefore, M = 381654729.
Editors Comment: We express thanks for this problem to Sndor Rka, the founding editor of
Abacus, Hungarys mathematical journal for students age 10 - 14.
2/4/11. The Fibonacci numbers are dened by F1 = F2 = 1 and Fn = Fn-1 + Fn-2 for n > 2. It is
well-known that the sum of any 10 consecutive Fibonacci numbers is divisible by 11. Deter
mine the smallest integer N so that the sum of any N consecutive Fibonacci numbers is divisi
ble by 12.
Solution 1 by Chi Cao Minh (12/TX): Let x and y represent two consecutive Fibonacci num
bers. Continuing the sequence:
Fibonacci Sum of Fibonacci Sum of Sequence

Number Sequence Number
1 x x 13 89x + 144y 233x +376 y
2 y x+y 14 144x + 233y 377x + 609y
3 x+y 2x + 2y 15 233x +377 y 610x + 986y
4 x + 2y 3x + 4y 16 377x + 610y 987x + 1596y
5 2x + 3y 5x + 7y 17 610x + 987y 1597x + 2583y
6 3x + 5y 8x + 12y 18 987x + 1597y 2584x +4180 y
7 5x + 8y 13x + 20y 19 1597x + 2584y 4181x + 6764y
8 8x + 13y 21x + 33y 20 2584x +4181 y 6765x + 10945y
9 13x + 21y 34x + 54y 21 4181x + 6765y 10946x + 17710y
10 21x + 34y 55x + 88y 22 6765x + 10946y 17711x + 28656y
11 34x + 55y 89x + 143y 23 10946x + 17711y 28657x + 46367y
12 55x + 89y 144x + 232y 24 17711x + 28657y 46368x + 75024y
Dividing the Sum of the Sequence by 12 shows that the smallest number N is 24.
Solution 2 by Ricky Liu (10/MA):
Suppose we have found N such that F k + F k + 1 + + F k + N 1 is always divisible by 12. Then
( F k + 1 + F k + 2 + + F k + N ) (F k + F k + 1 + + F k + N 1 ) = F k + N F k must be divisible by
12. In other words, given any two Fibonacci numbers with indices greater than one and separated
by N, their difference must be divisible by 12. To simplify things, notice that if
F 2 F N + 2 (mod12) and if F 3 F N + 3 (mod12) , then F k F N + k (mod12) for all k > 1. This
is easily shown by induction: F k + 2 F k + F k + 1 F N + k + F N + k + 1 F N + k + 2 (mod12) . So,
we need only nd the second pair of consecutive Fibonacci numbers equivalent to 1 and 2 (mod
12) respectively. These occur at F26 and F27. Thus, N = 26 -2 = 24. N = 24.
-
Editors Comment: This problem was created by the editor.
3/4/11. Determine the value of

1 1 1 1 1 1
S = 1 + ----2- + ----2- + 1 + ----2- + ----2- + + 1 + -------------2- + -------------2
1 2 2 3 1999 2000
Solution by Emily Kendall (11/IN):

1999

1 1
S = 1 + ----2- + ------------------2
a=1 a (a + 1 )
1999 4 3 2

a + 2a + 3a + 2a + 1
= ---------------------------------------------------------
2 2

a=1 a (a + 1 )
1999 2
-----------------------
a +a+1
= 2

a=1 a +a
1999
1 + --------------
1
= 2
a=1 a +a
1999
--a- a-----------
- , a telescoping sum.
1 1
= 1999 +
+ 1
a=1
1
= 1999 + 1 -----------
2000
1999
= 1999 + -----------
2000
Editors Comment: This timely problem was adapted by our Problem Editor, Dr. George Berzse
nyi, from MatLap, Transylvanias Hungarian language mathematics journal for students at the
middle and high school levels.
4/4/11. We will say that an octagon is integral if it is y
equiangular, its vertices are lattice points (i.e., points
with integer coordinates), and its area is an integer.
For example, the gure on the right shows an integral
octagon of area 21. Determine, with proof, the
smallest positive integer K so that for every positive
integer k K , there is an integral octagon of area k.
2
Solution 1 by Sawyer Tabony (11/VA): The integer K
1
that I came up with is K = 13, and I will show you how I
came up with this. First, I will show you that there is an 1 2 x
integral octagon of area k for all k greater than or equal
13. Here are the diagrams of integral octagons of areas
13 through 21:
13 14 15 16
17
18 19 20
21
Notice that for the last ve (17-21) there is a column shaded within each octagon exactly ve units
tall and one wide. To nd an octagon of area 22, one would only need to widen the column in the
octagon of area 17 by one unit. This would create the extra ve square units. Since we have ve
consecutive octagons with this feature, it is clear that by expanding the width of these columns we
can eventually reach any area.
So now I must show that there does not exist an integral octagon of
area 12. The smallest octagon whose height is ve is the one of
area 13 on the previous page, so clearly both height and width of an
octagon of area 12 must be less than 5 (height and width are the
dimensions of the smallest rectangle with vertical and horizontral
11
sides into which the integral octagon ts). The largest four by four
integral octagon is shown above and has an area of 14. The only
other one is to the right, and it has area 11.
Next we move down to four by three octagons. There is only one

octagon with dimensions four by three; it is shown at right (with an
area of 10). 10
The nal octagon is one of height and width three that has area of
seven. This is shown at the right.
So there is no integral octagon of area 12.

7
This makes K = 13.
Comment by Abhijit Mehta (10/OH): The smallest integral octagon, none of whose sides are
parallel to the coordinate axes, is shown below. Its area is 35, so it does not alter the solution.
35
Editors Comment: Dr. Berzsenyi based this problem on a similar, but more difcult problem,
proposed by A. C. Heath in the March 1976 issue of The Mathematical Gazette.
5/4/11. (Revised 2-4-2000) Let P be a point interior to square ABCD so that PA = a, PB = b, PC

= c, and c2 = a2 + 2b2. Given only the lengths a, b, and c, and using only a compass and
straightedge, construct a square congruent to square ABCD.
Solution 1 by Nyssa Thompson (10/HI): I am assuming basic construction knowledge: con
structing a right angle and a square with a given side length.
Connecting point P to points A, B, and C creates two trian-
gles, PAB and PBC , that share the same side PB.
Since ABCD is a square, AB = BC = CD = DA. We can A

B
refer to any of the sides as s. Furthermore, we can see that a b
ABP + CBP = 90 because they sum to ABC P
which is a corner of the square ABCD. Also, since it was
given that AP = a , BP = b , and CP = c , we can refer c
to these lengths as such. Thus PAB has sides of lengths a,
s, and b, and PBC has sides of lengths b, s, and c.
D C
We can rotate these triangles so that instead of sharing side
b, they share the s side. Then instead of the sides of length s
or
forming a right angle, the sides of length b form a right
angle (see below). A
B
c b a
b
C
s P
B
A a
b c
D
C
or
b
c
C s
B
A
a
b
Thus, we have a polygon with two adjacent sides of length b forming a right angle at B, and
remaining sides a and c. The segment from the vertex at B to the vertex at A and C is of length s.
With this knowledge we can construct the square ABCD if we are given a, b, and c using only
compass and straight edge.
1. Construct a right angle.
2. Set the compass to length b, place its point on the vertex of the right angle, and mark off
length b on each side of the angle.
3. Next set the compass to length a or c, place the compass point on the end of one of the lengths
b (not the vertex), and make a circle with the radius a or c.
4. Repeat step 3 with the remaining segment length a or c accordingly. Construct the circle on
the opposite bs endpoint.
5. Now, these two circles should either intersect at two points or be tangent.
6. Choose the point which lies within the right angle and, using the straightedge, connect it to the
vertex of the right angle. This length is s.
7. Construct a square with side length s. It will be congruent to square ABCD!
Solution 2 by Megan Guichard (12/WA): Let s be the length of the sides of ABCD and let
2 2 2
ABP = . Then, by the law of cosines, we have a = s + b 2sb cos ( ) ,
2 2 2 2 2 2
cos ( ) = ---------------------------- , and = cos ---------------------------- . We also know, again by the law of
s +b a 1 s + b a
2sb 2sb
cosines, that
2 2 2 2 2 2 2
c = s + b 2sb cos (CBP) = s + b 2sb cos (90 ) = s + b 2sb sin ( ) . Since
2 2 2
we know that = cos ---------------------------- ,
+b a 1 s
2sb
2
sin ( ) = 1 cos ( )
2 2 4 2 2 4 2 2 2 2 4
4s b s + 2a s a 2s b + 2a b b
= ------------------------------------------------------------------------------------------------------------
2 2

4s b
2 2 2 2 2 2 4 4 4
2a s + 2s b + 2a b s a b
= --------------------------------------------------------------------------------------------
2sb
2 2 2 2 2 2 4 4 4 2 2 2
Then, 2sb sin ( ) = 2a s + 2s b + 2a b s a b . Since c = s + b 2sb sin ( ) ,
2 2 2 2 2 2 2 2 2 2
2sb sin ( ) = s + b c = s + b a 2b = s a b
Squaring both sides of this equation gives

2 2 2 2 2 2 4 4 4 4 2 2 4 2 2 2 2 4
2a s + 2s b + 2a b s a b = s 2a s + a 2b s + 2a b + b
4 2 2 4 2 2 4
s 4a s + 2a 4s b + 2b = 0
4 2 2 2 4 4
s 2(a + b )s + (a + b ) = 0
4 2 2 2 2 2 2 2 2
s 2(a + b )s + (a + b ) = 2a b
2 2 2 2 2 2
(s (a + b )) = 2a b
2 2 2
s (a + b ) = ab 2
2 2 2
s = a + b + ab 2
s = a + b 2 ab -------
2 2 2 2
2
2 2 2
s = a + b 2 ab cos (135) .
Notice that this is simply the law of cosines for a triangle with sides a, b, and s, with the angle
opposite s measuring 135 . Since 2 sides and their included angle determine a triangle, we can
construct s by constructing a triangle with two sides a and b such that the angle between a and b is
135 . (We can construct an angle of 135 by constructing a 90 angle, bisecting it to get a
45 angle, and then adding another 90 angle.) Once we have s, we can construct a square
congruent to ABCD fairly easily by constructing 90 angles and new sides of length s until we
have a square with four sides of length s.
2 2 2
Editors Comment: Notice that the rst solution does not use c = a + 2b . Given that the
square ABCD exists, does this construction work for any point P (interior or exterior) and set of
distances PA = a, PB = b, PC = c? We are indebted to Professor Gregory Galperin of Eastern Illi
nois University for suggesting this wonderful problem.

1/1/12. Determine the smallest ve-digit positive integer N such that 2N is also a ve-digit inte
ger and all ten digits from 0 to 9 are found in N and 2N.
Solution 1 by Alex Lang (8/WI):

First step: Assign variables to the digits. N is assigned the variables ABCDE and 2N is assigned
the variables FGHIJ.
Second step: Solve for N and 2N. A cannot equal 0 since N would then be considered a 4 digit
number and to satisfy the conditions N needs to be a 5 digit number. The next smallest number
that A could equal is 1. Therefore, A will be assumed equal to 1 until all possible solutions with A
= 1 have been proved false. F = 2 since 2A = 2(1) = 2. B will be assumed equal to 3 since that is
the smallest number not already used. C will be assumed equal to 4. Then G = 6 since 2B = 2(3)
= 6. D will be assumed to equal 5. H = 9 since 2C + carry over from 2D = 2(4) + 1 = 9. But, if E
= 8, 2(D) + carry over = 2(5) + 1 = 11, so I = 1. That is a contradiction because A also equals 1
and each number is supposed to be used once. Also, if E = 0, 2E would also equal 0, and another
contradiction would occur. Therefore, D 5 . D will then be assumed equal to 7. That is because
G already equals 6, therefore 7 is the next smallest number. Then H = 9 since 2C + carryover =
2(4) + 1 = 9. Then E would have to equal 8 because if E = 0 then 2E would also equal 0 and a
contradiction would arise. But, if E = 8 then 2E = 2(8) = 16 and therefore J = 6, another contra
diction occurs since G already equals 6. Therefore, D 7 . D will then be assumed to equal 8.
Then H = 9 since 2C + carry over = 2(4) + 1 = 9. E will then be assumed to equal 5. Then 2E =
2(5) = 10, but J would only equal the last digit which is 0, I would then equal 2D + carry over =
2(8) + 1 = 17, so I = 7. That solution satises the specied conditions.
Therefore, N = 13,485 and 2N = 26,970.
Solution 2 by Valerie Lee (10/NY):

N is comprised of digits ABCDE, while 2N is FGHIJ.
1. Let A = 1.
2. If A is 1, then B 2 because F would equal 2, and B 2 because if C < 5, then G = 0 and
C = 5-9, then G = 1. Let B = 3.
3. C 0-3, so let C = 4.
Thus far we have 134DE 2 = 26 [ 8, 9 ]IJ .
4. a) D 0-4, so let D = 5, so we would have 1345E 2 = 269 [ 0,

1 ]J and there would be
no way to conclude J = 7 or 8, so D 5 .
b) Let D = 6, so 1346E 2 = 269 [ 2,

3 ]J , but 6 is used twice, so D 6 .
c) Let D = 7, so 1347E 2 = 269 [ 4, 5 ]J , but there is no answer for E so that J = 0 or 8, so
D 7.
d) Let D = 8, so 1348E 2 = 269 [ 6, 7 ]J . Now E = 0 or 5. E = 0 gives contradiction. So
let E = 5. 13485 2 = 26970 . This uses all ten digits.
Answer: N = 13485.
Solution 3 by Paul Wrayno (11/NC): Let N = ABCDE, 2N = FGHIJ.

To be smallest, ideally A = 1, and F = 2, B = 3, G = 6, if it is possible to create N and 2N with these
values, because they are the absolute least values for the rst two digits. The lowest remaining
digit is 4, so ideally C = 4, causing H = 9 because the remaining digits guarantee a carry from the
2 D. This leaves 0, 5, 7, and 8. Since 0 can only be achieved by 2 5 without a carry, E = 5
and J = 0. 8 2 + 1 = 17 , which ts the other four digits, so D = 8 and J = 7. This is the only
(N, 2N) pair that has the ideal rst three digits, so N = 13485 is the smallest.
Answer: N = 13485.
Editors Comment: We are indebted to Dr. Bla Bajnok of Gettysburg College for posing this
problem and for his continuing assistance with USAMTS. Dr. Bajnok claims the next smallest
solutions (after 13485) are 13548, 13845, and 14538. He notes they all use the same ve digits.
24
2
2/1/12. It was recently shown that 2 + 1 is not a prime number. Find the four rightmost dig
its of this number.
Solution 1 by Christopher Lyons (12/CA): This problem is equivalent to asking, What is the
24
2
remainder when 2 + 1 is divided by 10,000? Since we are only concerned with the remain
der, I chose to use the mathematical tool that is all about remainders: the congruence. My strategy
in attacking this problem is to use the rules of congruences to build up from
1 24
2 2
2 = 4 4 ( mod 10,000 ) to 2 = 7536 ( mod 10,000 ) , and to add 1 to get
24
2
2 + 1 = 7537 ( mod 10,000 ) .
1
2
We start by evaluating the congruence of 2 modulo 10,000. In fact, throughout the description
of this problem, all congruences will be modulo 10,000.
1
2
2 = 4 4 ( mod 10,000 )
Since we can multiply congruences just like equations, we will multiply this congruence by itself.
But rst we must show an identity that will be useful for the rest of the problem:
n n n n n n+1
2 2 (2 + 2 ) 2( 2 ) 2
2 2 = 2 = 2 = 2
n n n+1 6 6 7
2 2 2 2 2 2
Thus, when we multiply 2 by 2 , we get 2 . For example 2 2 = 2 . We can
1 24
2 2
apply this identity repeatedly with congruences to get all the way from 2 to 2 . Let us write
out the rst few steps:
2 1 1
2 2 2
2 = 2 2 = 4
4 = 16
2
2
2 = 16
3
2 2
2 2 2
2 = 2 2 = 16 16 = 256
2
2 = 256
4 3 3
2 2 2
2 = 2 2 = 256 256 = 65536 5536 ( mod 10,000 )
4
2
2 5536
5 4 4
2 2 2
2 = 2 2 = 65536 65536 = 30647296 7296
5
2
2 7296
6 7 24
2 2 2
We continue this process, obtaining the congruences of 2 , 2 , ..., 2 . The following is a
table of all these congruences. (I used an eight-digit hand held calculator to nd these values,
since this was the easiest way for me.)
1 9 17
2
2 4 2
2
4096 2
2
3696
2 10 18
2 2 2
2 16 2 7216 2 416
3 11 19
2 2 2
2 256 2 656 2 3056
4 12 20
2 2 2
2 5536 2 336 2 9136
5 13 21
2 2 2
2 7296 2 2896 2 6496
6 14 22
2 2 2
2 1616 2 6816 2 8016
7 15 23
2 2 2
2 1456 2 7856 2 6256
8 16 24
2 2 2
2 9936 2 6736 2 7536
Now that we have obtained this last congruence, we may add it to another congruence to produce
the desired result:
24
2
2 + 1 7536 + 1 7537 ( mod 10,000 ) .
This nal congruence is equivalent to writing

24
2
2 + 1 = 10000 K + 7537
where K is some positive integer (a large one, probably!). Looking at the right side of this equa
tion, we see that the multiple of 10,000 will have (at least) four trailing zeros. When we add the
multiple and its four trailing zeros to 7537, we get a number whose four rightmost digits are 7537.
24
2
The four rightmost digits of 2 + 1 are 7537.
Solution 2 by Aleksandr Kivenson (10/NY): To begin, it is necessary to declare that to obtain

the four rightmost digits of a product, only the four rightmost digits of the factors need to be mul
tiplied (for example, to get the four rightmost digits of the product of 12,345 and 67,890, all you
have to do is multiply 2,345 by 7,890 and take the four rightmost terms of the result). Since my
approach to this problem is to use known large numbers and multiply their four rightmost digits to
get the answer, I will use this method.
Since when you multiply numbers which have the same base but different powers you add the
2 3 2+3
powers (for example 2 2 = 2 ), it would be easy to nd numbers whose rightmost four
digits I should multiply by looking for large numbers expressed as powers of 2, nding the ones
24
whose powers add up to 2 , and then multiplying the rightmost 4 digits of these numbers to get
24
2
the rightmost four digits of 2 .
Conveniently, such a list of known powers of two exists in the form of Mersenne prime numbers.
x
These are a special type of prime numbers that are expressed as 2 1 . Many such prime num
bers are known and I used a web page
http://www.isthe.com/chongo/tech/math/prime/mersenne.html
to nd numerical values for these primes. I chose the powers of 2 whose powers added up to the
24
exponent 2 = 16777216 . I ignored the -1 because when I chose the powers I would use I
obtained the last four digits of their numerical value, added one to each, and then multiplied them
24
2
out to get the last four digits of 2 . The powers which I used and their numerical value are as
follows:
:
Power of 2 Last four digits of that power

12 4096
2
607 8128
2
607 8128
2
19937 1472
2
44497 8672
2
216091 8448
2
216091 8448
2
1398269 5712
2
2976221 1152
2
2976221 1152
2
2976221 1152
2
2976221 1152
2
2976221 1152
2
Sums of powers: 16,777,216 Last four digits of above values: 7,536
From multiplying the last four digits of the numerical value of each power, I obtained a number
24
2
whose last four digits were 7,536. This means the last four digits of 2 are 7,536.
24
2
Therefore, the last four digits of 2 + 1 are 7,537.
Solution 3 by Zhihao Liu (11/IL):
Answer: 7537
Proof: [See the Editors Comment below for a quick review of some of these terms and concepts.]
24 9 2
Note that 2 ( 2 ) ( 64 ) 144 64 = 9216 216 ( mod 500 ) . By Eulers Theorem,
24
2 x 4 24 4
2 2 (mod 5 ) , if 2 x(mod ( 5 )) . Since (625) = 500 , Eulers Theorem says

24 24
2 x 24 2 216
2 2 (mod 625), if 2 x(mod 500) . It follows that 2 2 (mod 625) . By doubling
27
for a while we nd 2 = 134217728 228 ( mod 625 ) , and so
216 8 2 4 4 2 2 2
2 228 (228 ) 109 (109 ) 6 36 ( mod 625 ) . Since 24 and 54 are relatively
4 5
prime, by the Chinese Remainder Theorem there is a unique n, 0 n < 2 2 = 10000 that sat
ises both n 36 ( mod 625 ) , and n 0 ( mod 16 ) . Note that n = 7536 satises both of these
24 24
2 2
congruences, and 2 n(mod 10000) . Therefore the last four digits of 2 + 1 are 7537.
Solution 4 by Jacob Licht (11/CT):

(m)
Eulers Theorem states: If (a,m) = 1 , then a 1 ( modm) . If a = 2 and m = 625,
4 3
(625) = ( 5 ) = 5 (5 1) = 500 . By the Division Algorithm q, r Z
24
2 = 500q + r and 0 r < 500 . [ Read: By the Division Algorithm there exist q and r, ele
24
ments of the integers Z, such that 2 = 500q + r and 0 r < 500 . ] Since
10 24 10 2 4 2
2 = 1024 24 ( mod 500 ) , 2 = (2 ) 2 (24 )16 216 ( mod 500 ) . So r = 216, and
24
2 500q + 216 500 q 216 q 216 216
2 = 2 = (2 ) 2 (1) 2 2 (mod 625) . Note that 28 = 256, so
16 2
2 = 256 = 65536 536 ( mod 625 )
32 2
2 = 536 421 (
mod 625 )
64 2
2 = 421 366 (
mod 625 )
128 2
2 = 366 206 (
mod 625 ) .
216 128 + 64 + 16 + 8
So, 2 = 2 (206)(366)(536)(256) 36 ( mod 625 ) .
24 24 24
2 4 2 2
Thus 2 36 ( mod 635 ) , and since 2 = 16 divides2 so 2 0 ( mod 16 ) . We also have
that 625 1 ( mod 16 ) and 36 4 ( mod 16 ) , so 12 ( 625 ) + 36 0 ( mod 16 ) . Now by the Chinese
24
2
Remainder Theorem 2 12 ( 625 ) + 36 7536 ( mod 10000 ) .
24
2
So the four rightmost digits of 2 + 1 are 7537.
Editors Comment: In what he described as the deepest computation in history whose result
was a simple yes/no answer, Richard Crandall of the Center for Advanced Computation at Reed
College, together with Ernest Myer, formerly of Case Western Reserve University, and Jason Pap
24
2
adopoulos of the University of Maryland, have veried that the 24th Fermat number, 2 + 1 , is
not a prime number. For more information about the 24th Fermat number, visit the web site
www.perfsci.com/. This problem was created by Gene Berg of NSA.
The solutions to these problems give us an opportunity to briey introduce Eulers function
( m ) , Eulers generalization of Fermats Theorem, and the Chinese Remainder Theroem. Our
goal is to introduce some of the notation and terminology of this subject to young mathematicians
seeing this for the rst time, and possibly help them understand the proofs. For more details see
(a) An Introduction to the Theory of Numbers by G. H. Hardy and E. M. Wright, published by
Clarendon Press, or (b) The Art of Problem Solving, Volumes 1 and 2, by Richard Rusczyk and
Sandor Lehoczky, published by Greater Testing Concepts, P.O. Box 5014, New York, NY 10185
5014.
For a brief discussion of Congruences, Fermats Theorem, and the Extended Euclidean Algo
rithm (EEA) for nding Greatest Common Divisors see the Solutions to Problem 1/2/11 from Year
11 of the USAMTS.
Denition (Eulers function ( m ) ): For an integer m, let ( m ) denote the number of positive
integers less than m and relatively prime to m. For example, consider m = 20: there are eight pos
itive integers less than 20 which are relatively prime to 20, namely 1, 3, 7, 9, 11, 13, 17, and 19, so
( 20 ) = 8 . Since m = 17 is prime, all sixteen positive integers less than 17 are prime to 17,
and ( 17 ) = 16 . If p is prime, then ( p ) = p 1 --- . If m and n are relatively prime inte

k k 1
p
gers, then (mn) = ( m ) ( n ) . For example:
( 20 ) = (4 5) = ( 4 ) ( 5 ) = 4 1 --- [ 4 ] = 8
1
2
Theorem (Eulers generalization of Fermats Theorem): If a and m are integers with Greatest
Common Divisor GCD(a, m) = 1, then
(m)
a 1 ( mod m).
For example, if a is any integer relatively prime to 20 [i.e. a {1, 3, 7, 9, 11, 13, 17, 19} ], then
8
a 1 ( mod 20) .
Chinese Remainder Theroem: If m 1, m 2, , m k are positive integers that are pairwise relatively
prime [ that is, GCD( mi, mj) = 1 for 1 i < j k ], then for any integers a 1, a 2, , a k the system
of congruences y a i (mod m i ), i = 1, 2, , k , has a simultaneous solution y that is uniquely
determined modulo m = m 1 m 2 m k . [A similar theorem applies to polynomials.]
As an example, nd an integer c with 0 c < 3 7 11 13 = 3003 such that

c 2 ( mod 3) ,
c 4 ( mod 7) ,
c 6 ( mod 11 ) , and
c 8 ( mod 13 ) .
Solution: We do this in three steps, solving for the rst two equations, then for the rst three equa
tions, and nally for all four equations. In each step we use the Extended Euclidean Algorithm
(EEA) (for examples of this algorithm in use, see the Solution to Problem 2/1/11 of Year 11).
Step 1. Find x satisfying x 2 ( mod 3 ) a 1 (modm 1 ) and x 4 ( mod 7 ) a 2 (modm 2 ) .
GCD(m1, m2) = 1 so by the EEA there are integers f1 and f2 so that
1 f 1 m 1 + f 2 m 2 f 1 3 + f 2 7 . The EEA nds f 1 = 5 and f 2 = 2 . Now choose
x = a 1 + (a 2 a 1 ) f 1 m 1 = 2 + (4 2 ) 5 3 = 32 .
Observe x satises the requirements of step 1.
Step 2. Find y satisfying y 2 ( mod 3 ) a 1 (modm 1 ) , y 4 ( mod 7 ) a 2 (modm 2 ) , and

y 6 ( mod 11 ) a 3 (modm 3 ) . GCD(m1m2, m3) = 1 so by the EEA there are integers g1 and g2
so that 1 g 1 m 1 m 2 + g 2 m 3 g 1 21 + g 2 11 ( 1 ) 21 + 2 11 . Now choose
y = a 1 + (a 2 a 1 ) f 1 m 1 + (a 3 x) g 1 m 1 m 2 = 32 + (6 32)( 1 ) ( 3 ) ( 7 ) = 578 .
where
x = a 1 + (a 2 a 1 ) f 1 m 1 .
Observe y = 578 satises step 2.
Step 3. Find c satisfying c 2 ( mod 3 ) a 1 (modm 1 ) , c 4 ( mod 7 ) a 2 (modm 2 ) ,

c 6 ( mod 11 ) a 3 (modm 3 ) and c 8 ( mod 13 ) . GCD(m1m2m3, m4) = 1 so by the EEA there
are integers h1 and h2 so that
1 h 1 m 1 m 2 m 3 + h 2 m 4 h 1 3 7 11 + h 2 13 4 231 + ( 71 ) 13 . Now choose
z = a 1 + (a 2 a 1 ) f 1 m 1 + (a 3 x) g 1 m 1 m 2 + (a 4 y) h 1 m 1 m 2 m 3
= y + (a 4 y) h 1 m 1 m 2 m 3 .
where
x = a 1 + (a 2 a 1 ) f 1 m 1 and y = a 1 + (a 2 a 1 ) f 1 m 1 + (a 3 x) g 1 m 1 m 2 .
Observe z = 578 + (8 578 ) 4 3 7 11
= 526680 , and reduced to
c = 2426 z(mod 3003) satises step 3 and the original requirement.
3/1/12. Determine the integers a, b, c, d, and e for which

2 3 2 5
( x + ax + b)( x + cx + dx + e) = x 9 x 27 .
Solution 1 by Christopher Church (10/KY): This problem essentially asks us to factor

5
x 9 x 27 into a quadratic and a cubic polynomial. Theoretically, a TI-89 graphing calculator
could factor this polynomial. I, however, do not have access to the TI-89. However, I did use the
TI-85 to calculate the roots of the polynomial. My calculator returned the ve values shown here:
2.1541, 42291.9998i, 1.5 0.8660i

3 3
The last one, based on my knowledge of the quadratic formula, was --- --- i . This I could
2 2
transfer into a polynomial of degree two. The result, using the fact that x equals the values above,
2
is x + 3 x + 3 . Using long division of polynomials, I found that
2 3 2 5
( x + 3 x + 3)( x 3 x + 6 x 9 ) = x 9x 27 .
Thus the integers that the problem asked for are found here. They are a = 3 , b = 3 ,
c = 3 , d = 6 , and e = 9 .
Solution 2 by Laura Pruitt (11/MA): Multiply out the left side of the equation to get
5 4 3 2 5
x + (a + c) x + (ac + b + d) x + (ad + bc + e) x + (ae + bd)x + be = x 9 x 27
Comparing coefcients in this new equation, it is easy to nd equations relating the variables a, b,
c, d, and e:
1. a + c = 0
2. ac + b + d = 0
3. ad + bc + e = 0
4. ae + bd = 9
5. be = 27
Initial observations on these equations:

(i) Simple substitutions: a = c , or c = a , b = 27 e or e = 27 b .
(ii) Since all variables are integers, there are only eight possibilities for
(b,e) : (1, 2 ), ( 1,27), (27, 1), (2 ,1), (3, 9), ( 3,9), (9, 3), ( 9,3) .
This leaves us with three equations (2, 3, and 4) and one variable (d) that we have not used yet.
Substitute c for a and solve 2, 3, and 4 for d in terms of a, b, c, and e:
2 2
2. c + b + d = 0 -> d = c b
3. cd + bc + e = 0 -> d = (bc + e) c
4. ce + bd = 9 -> d = (ce 9 ) b
2
Therefore c b = (bc + e) c = (ce 9 ) b .
In pairs, solve for c in terms of b and e:
Comb. 1 (use 2 and 3) Comb. 2 (use 2 and 4) Comb. 3 (use 3 and 4)
2 2
c b = ( bc + e ) c c b = ( ce 9 ) b ( bc + e ) c = ( ce 9 ) b
3 2 2 2 2
c 2bc e = 0 bc ec + ( 9 b ) = 0 ec ( b + 9 )c be = 0
2 2 2 2
2
[ e e 4b ( 9 b ) ]
2 b + 9 ( b + 9 ) + 4be
discard: cubic c = ------------------------------------------------------- c = -----------------------------------------------------------------------
2b 2e
In order for c to be an integer, which it must be, the discriminant must be a perfect square. Test
the eight possible solutions for (b,e) :
For Comb. 2: (b,e) {(9, 3), (3, 9)}
For Comb. 3: (b,e) = (3

, 9)
Check (b,e) = (3, 9) in the full quadratics from Comb. 2 and Comb. 3. It checks.
If (b,e) = (3, 9) , then c {0, 3} (the answers to the quadratics), but looking at equation 3
when solved for d (d = (bc + e) c ), we see that c 0 since division by 0 is undened. There
fore c = 3 and a = c = 3 .
It is now simple to solve for d; simply substitute the values of a, b, c, and e into any of the original
equations containing d to yield d = 6 .
Solution: (a, b, c, d, e ) = (3, 3, 3, 6, 9 )

The desired factorization is
2 3 2 5
( x + 3 x + 3)( x 3 x + 6 x 9 ) = x 9x 27 .
Solution 3 by Sarah Emerson (12/WA): Expand the equation to obtain:

5 4 3 2 5
x + (a + c) x + (ac + b + d) x + (ad + bc + e) x + (ae + bd)x + be = x 9 x 27
Therefore
(1) a + c = 0 c = a
(2a) d + ac + b = 0 2
da +b = 0
(3a) e + ad + bc = 0 e + ad ba = 0
(4) ae + bd = 9
(5) be = 27
To solve, try all possible values of b and e, plugging the values into the other equations to deter
mine if they work.
b = 1 (4) 27a + d = 9 d = 9 27a a does not exist.
e = 27 (3) 27 + ad a = 0 27 + a ( 9 27a ) a = 0
2
27 9a 27a a = 0
2
27a 10a 27 = 0
10 100 4 ( 27 ) ( 27 )
a = ---------------------------------------------------------------
2 ( 27 )
b = 3 (4) 9a + 3d = 9 d = 3 + 3a a = 3
e = 9 (3) 9 + ad 3a = 0 2 b = 3
9 3a + 3a 3a = 0
2 c = 3
a 2a 3 = 0
d = 6
(a 3)(a + 1) = 0
e = 9
a = 3 or a = 1
it works!
Solution.
Editors Comment: We thank our Problem Editor, Dr. George Berzsenyi, for this problem. It
5 2
stems from a recent article, The Factorization of x p x k and Fibonacci Numbers, pub
lished in the November 1999 issue of the Fibonacci Quarterly.
4/1/12. A sequence of real numbers s 0, s 1, s 2, has the property that s i s j = s i + j + s i j for

all nonnegative integers i and j with i j , s i = s i + 12 for all nonnegative integers i, and
s0 > s1 > s2 > 0 . Find the three numbers s0, s1, and s2.
Solution 1 by Jennifer Dawson (11/AK):

Answer: s 0 = 2 , s 1 = 3 , and s 2 = 1 .
Solution:
s 1 s 0 = s 1 + s 1 = 2s 1
s0 = 2
s 1 s 1 = s 2 + s 0
2 2
s2 = s1 s0 = s1 2
s2 s1 = s3 + s1
2
[s 1 2 ] s 1 = s 3 + s 1
s 3 = s 1 3s 1
s 3 s 3 = s 6 + s 0
3 2
[s 1 3s 1 ] = s 6 +
2
6 4
2
s 1 6s 1 + 9s 1 = s 6 +
2
6 4 2
s 6 = s 1 6s 1 + 9s 1 2
s 6 s 6 = s 12
+ s 0 = s 0 + s 0 = 4
6 4
2 2
[s 1 6s 1 + 9s 1 2 ] = 4
s 1 { 3, 3, 2, 2, 1, 1, 0}
Since s1 must be strictly between 0 and 2, all but 1 and 3 are eliminated.
First, try s 1 = 1 . s 1 s 1 = s 2 + s 0 yields s 2 = 1 , a contradiction. So s1 1 .
Second, try s 1 = 3. s 1 s 1 = s 2 + s 0 yields 3 = s 2 + 2 or s 2 = 1 . It works!
So, s 0 = 2 , s 1 = 3 , and s 2 = 1 .
Editors Comment: Once again, we are most grateful to Dr. Erin Schram of NSA for this intrigu
ing problem. An indirect, but interesting solution begins with the observation that
1 1
cos cos = --- cos ( + ) + --- cos ( ) and modeling the sequence as s i = 2 cos ( i ) for
2 2
some .
5/1/12. In the octahedron shown on the right, the base
and top faces are equilateral triangles with sides
measuring 9 and 5 units, and the lateral edges are all 5
of length 6 units. Determine the height of the octa 5 5
hedron; i.e., the distance between the base and the
6 6
top face.
6 6
Solution 1 by Anna Maltseva (12/MI): 6 6
By symmetry, the top face is parallel to the base, and the
line connecting the centers of the triangles of the top
face and the base is perpendicular to both the top face
9 9
and the base. 9
The top view of the octahedron looks like the gure at left. Let
O1 denote the center of the top face, O2 denote the center of the
base, A denote a vertex of the top face, and B denote the mid
point of the corresponding side of the base. Imagine dropping a
A line from A perpendicular to the base and let K denote the point
B where it intersects the plane of the base. Then triangle ABK will
O1 be a right triangle and AK will be the height of the octahedron.
2 2 2 2 2
So AB = AK + BK = AK + (O 2 B O 1 A)
2 2 2 2
6 = AB + (4.5) so AB = 15.75 .
A
A 5
-------
4.5 3
-------
3 2.5
O2
4.5 B
30
Therefore,
15.75 = AK + ------- ------- .
2 4.5 5 2
3 3
63 1 2
------ ------ = AK
4 12
188 47 141
AK = --------- = ------ = ------------- .

12 3 3
Solution 2 by Alexander Power (11/IA):
Let the equilateral triangle with side 5 have vertices A, B, and C, and let the equilateral triangle
with side 9 have vertices D, E, and F, with sides of the octahedron AD, BD, BE, CE, CF, AF.
Then, the height of the octahedron is the same as the height of tetrahedron ABCD and the same as
the height of tetrahedron CDEF. We know the length of all the sides except CD. A formula for
T
the volume of a tetrahedron ABCD is --------- , where T is the determinant of the 5 5 matrix
288
0 1 1 1 1
2 2 2
1 0 AB AC AD
2 2 2
1 AB 0 BC BD
2 2
2
1 AC BC 0 CD
2 2 2
1 AD BD CD 0
Plugging in values of the two tetrahedra, we get
0 1 1 1 1
1 0 25 25 36
4 2
T 1 = det 1 25 0 25 36 = 50 x + 4850 x 6050

2
1 25 25 0 x
2
1 36 36 x 0
and
0 1 1 1 1
1 0 81 81 36
4 2
T 2 = det 1 81 0 81 36 = 162 x + 24786 x 328050

2
1 81 81 0 x
2
1 36 36 x 0
as the determinants of the two tetrahedra, with T1 as the determinant of tetrahedron ABCD and T2
the determinant of CDEF. Since the tetrahedra have the same height, their volumes are propor
tional to the areas of their bases (by Cavalieris Principle). The areas of their bases are propor
tional to a square of a side. Thus
T1 T2
81 --------
- = 25 --------
-.
288 288
25
2
This means that T 2 ------ = T 1 . Thus
81
( 50 x + 4850 x 6050 ) = ( 162x + 24786x 328050 ) ------

4 2 4 2 25
81
or
2800 4 22400 2
--------------- x + --------------- x + 25200 = 0
81 9
and x2 is 81 or -9, but -9 is extraneous. This means that CD = 9 . Thus, T2 is
0 1 1 1 1
1 0 81 81 36
T 2 = 1 81 0 81 36 = 616734
1 81 81 0 81
1 36 36 81 0
27 81 3
and the volume of tetrahedron CDEF is ------ 47 . Since the base has area ------------- , the height is
4 4
47
------ 3.958 .
3
Editors Comment: Dr. Berzsenyi based this problem on a recent article in Mathematics Maga
zine (vol. 72, no. 4, pp.277-286).

1/2/12. A well-known test for divisibility by 19 is

as follows: Remove the last digit of the number,
6 7 9 4 4 4 4 9 7 6
add twice that digit to the truncated number, and
keep repeating this procedure until a number less 8 1 2
than 20 is obtained. Then, the original number is 6 8 0 2 4 5 0 9
divisible by 19 if and only if the nal number is 4 1 8
19. The method is exemplied on the right; it is 6 8 4
4 6 8
easy to check that indeed 67944 is divisible by
8 1 6
19, while 44976 is not. 7 6 6 2
1 2 4
Find and prove a similar test for divisibility by
29. 1 9 1 0
Solution 1 by Nina Boyarchenko (10/PA):

Method: Remove the last digit of the number and add three times that digit to the truncated num
ber. Keep repeating this procedure until a number less than 30 is obtained. Then the original num
ber is divisible by 29 if and only if the nal number is 29.
Proof: The given number can be represented as 10a + b , where b is the last digit of the number
and a is the number when b is removed. When b is removed and when 3b is added to the trun
cated number, it becomes a + 3b .
10a + b is divisible by 29 if and only if 10a + b + 29b is divisible by 29.

10a + b + 29b = 10a + 30b = 10 (a + 3b) is divisible by 29 if and only if a + 3b is divisible by
29 as 10 is not divisible by 29 and 29 is prime. Also, since 29 is the only number between 0 and
30 that is divisible by 29, the original number 10a + b is divisible by 29 if and only if the nal
number is divisible by 29.
Solution 2 by Benjamin Armbruster (12/AZ): The test for divisibility by 29 is as follows:

Remove the last digit of the number (x), add three times that digit to the truncated number (obtain
ing x), and keep repeating this procedure until a nal number less than 30 is obtained. Then, the
original number is divisible by 29 if and only if the nal number is 29.
Proof. Let the original number be x. The last digit of x is x(mod 10) . After removing the last
digit one has
x ( x(mod 10))
---------------------------------------
10
If one then adds three times that last digit one obtains
x ( x(mod 10)) x + 29 ( x(mod 10))
x = --------------------------------------- + 3( x(mod 10)) = ---------------------------------------------
10 10
Hence,
x ([( x + 29 )( x(mod 10))] 10)(mod 29)
10 x ( x + 29 )( x(mod 10))(mod 29)
10 x x(mod 29) .
Since 29 is prime, this means that x 0 if and only if x 0 . This means that the nal number is
( x + 29 )( x(mod 10))
divisible by 29 if and only if the original number was. Because x = --------------------------------------------------- , x
10
will always be less than x for positive x. This means that if you repeat the procedure long enough,
you will get a number less than 30. Then, any number less than 30 is divisible by 29 if and only if
it is 29.
Solution 3 by Sean Markan (11/MA):

A similar test for divisibility by 29 is to remove the last digit of the number, add three times that
digit to the truncated number, and repeat the process until the resulting number is less than 30.
The original number is divisible by 29 iff the nal number is 29.
To prove this test, we rst show that 10a + b is a multiple of 29 if and only if a + 3b is a multiple
of 29:
a + 3b 0 ( mod 29 )
10a + 30b 0 ( mod 29 )
10a + b 0 ( mod 29 )
Now, starting with a number ab, where a is an integer, b is a single digit, and ab represents
10a + b , tripling the last digit and adding it to the truncated number yields a + 3b , which is a
multiple of 29 if and only if the previous number is. If 10a + b > 29 , then a + 3b is a smaller
number also. So, by repeating this process we arrive at a number less than 30 which is a multiple
of 29 iff the original number was. Therefore, if the nal number is 29, then the original number
was divisible by 29; otherwise the original number was not divisible by 29.
Extension by Laura Pruitt (11/MA): An essentially identical rule will work for any number one
less than a multiple of ten: for 39, multiply by 4; for 49, by 5; etc. This format is also the basis of
the more common divisibility rule for 9, which simply adds up the digits to get a multiple of 9.
This is the essential difference: by removing all available multiples of the number at each step (by
bypassing all tens), our rule gives the number itself as the nal answer rather than any multiple
of that number.
Editors Comment: We thank our Problem Editor, Dr. George Berzsenyi, for this problem.
1492! 1492!
2/2/12. Compute 1776 (mod 2000) ; i.e., the remainder when 1776 is divided by
2000. (As usual, the exclamation point denotes factorial.)
Solution 1 by Jason Chiu (12/NY): Answer: 1376.

Powers of 1776 (mod 2000)
1
1776 has remainder 1776,
2
3
4
5
6
7
and so on.
6 1 n n5
Since 1776 1776 1776 ( mod 2000 ) , 1776 1776 (mod 2000) for all n > 5 and we
may consider the exponent (mod 5). It is plain that 1492! is divisible by 5, so that
1492! 5
1776 1776 1376 ( mod 2000 )
1492!
Therefore, 1776 2000 has remainder 1376.
Solution 2 by Eugene Fridman (12/IL): Answer: 1376.

We begin by introducing a lemma.
n
Lemma: For all positive integers n, 1376 1376 ( mod 2000 ) .
Proof of Lemma: We prove the lemma by induction. We rst notice that
1 2
1376 1376 ( mod 2000 ) and 1376 = 1893376 1376 ( mod 2000 ) . For our induction
k
hypothesis, we assume the statement is true for n = k , i.e. that 1376 1376 ( mod 2000 ) .
We then prove that the statement is true for n = k + 1 . That is, we prove that
k+1
1376 1376 ( mod 2000 ) . To do so, we start with the inductive hypothesis and multiply
k 2
both sides of the congruence equation by 1376 to obtain 1376 1376 1376 (mod 2000) ,
k+1 2 2 1
or 1376 1376 (mod 2000) . Since 1376 1376 (mod 2000) , the congruence above
k+1
can be written as 1376 1376 ( mod 2000 ) , and our proof is complete.
5
We now compute the required quantity by noticing that 1776 1376 ( mod 2000 ) . Since 5
divides 1492! we can say that
1492! 1492!
-------------- --------------
1492! 5 5 5
1776 (1776 ) 1376 (mod 2000)
According to the lemma above,
1492!
--------------
5
1376 1376 ( mod 2000 )
so
1492!
1776 1376 ( mod 2000 ) .
Hence our answer is 1376.
Solution 3 by Charles Wang (12/IL):

Since 2000 = 16 125 , we can look at this number mod 125 and mod 16 and use the Chinese
1492!
Remainder Theorem to nd this number mod 2000. First, 1776 0 ( mod 16 ) since 1776 is a
multiple of 16 so any power of 1776 is a multiple of 16 also. It is also well known that
(n)
x 1 ( modn) where (n, x) = 1 and ( n ) is the totient function, a function that counts the
number of positive integers less than and relatively prime to n. Since (125) is less than 125,
which in turn is less than1492, we know that (125) 1492! 1492! = (125) k where
kZ.
(125) (125) k k
Since (125, 1776 ) = 1 , 1776 1 ( mod 125 ) 1776 1 1 ( mod 125 ) .
1492! (125) k
Therefore 1776 1776 1 ( mod 125 ) .
1492!
Using the Chinese Remainder Theorem1776 0 125 a + 1 16 b(mod 16 125) , where
125 a 1 ( mod 8 ) and 16 b 1 ( mod 125 ) . Since 125 a is multiplied by 0, we need only
solve for b in this equation. Running quickly through the equations we nd that
16 86 1 (
mod 125 ) . Plugging back in, we nd that
1492! 1492!
1776 16 86 ( mod 2000 ) 1776 1376 (
mod 2000 ) .
Solution 4 by Anatoly Preygel (10/MD):

(m)
Eulers Extension to Fermats Little Theorem states: If (a,m) = 1 , then a 1 ( modm) .
100
So we have a 1 ( mod 125 ) for all a with gcd(a, 125) = 1.
1492!
We see that 16 125 = 2000 , and 16 divides 1776 so 1776 0 ( mod 16 ) .
1492!
Let us consider 1776 (mod 125) . By Eulers theorem above, since gcd(125, 1776) = 1, and
since 5|1492!:
1492!
1776 1 ( mod 125 ) .
We now wish to solve for the system of congruences n 1 ( mod 125 ) , n 0 ( mod 16 ) , which we
can do by simple trial. Of all the numbers between 0 and 2000 congruent to 1 mod 125, namely
251, 376, 501, 626, 751, 876, 1001, 1251, 1376, 1501, 1626, 1751, 1876,
only 1376 is divisible by 16.
1492!
Thus we see that 1776 1376 ( mod 2000 )
Editors Comment: We are thankful to Dr. Peter Anspach of NSA for this nice problem.
1776!
Remarkably, 1492 (mod 2000) = 1376 as well, but seems harder to prove. For a brief intro
duction to Eulers totient function ( m ) , Eulers generalization of Fermats Theorem, and the
Chinese Remainder Theorem see the Solutions to Round 1 of Year 12 on these web pages.
3/2/12. Given the arithmetic progression of integers

308, 973, 1638, 2303, 2968, 3633, 4298 ,
determine the unique geometric progression of integers,
b 1, b 2, b 3, b 4, b 5
, b 6
so that
308 < b 1 < 973 < b 2 < 1638 < b 3 < 2303 < b 4 < 2968 < b 5 < 3633 < b 6 < 4298 .
Solution by Rishi Gupta (8/CA): First I tried to nd a number x such that b 1 x = b 2 ,

b 2 x = b 3 , b 3 x = b 4 , etc.
I started by nding the smallest value x could equal. Since b6 and b1 are 5 sequence numbers
5
apart, b 6 = b 1 x . I took the largest value for b1, which is 972, and the smallest value for b6,
5
which is 3634. Therefore, 3634 972 x and 1.301 x .
Then I found the maximum value for x, using the smallest b4 and the largest b6, so
2
4297 2304 x , and x 1.37 .
Therefore 1.301 x 1.37 .
5
Now since b 1 x = b 6 and the sequence is integral, when x is written as a fraction b1 must be
divisible by the denominator to the fth power. Any fraction with a denominator greater than 3
5
will not work, because 4 = 1024 and b1 must be less than 973. Neither 1/1 nor 1/2 have multi
ples between 1.30 and 1.37, so that leaves us with a denominator of 3. Since 4/3 is the only mul
tiple of 1/3 that is between 1.30 and 1.37, x = 4 3 .
5
Since 3 = 243 , b1 is a multiple of 243. Because b1 is between 309 and 972, it can only be 486,
729, or 972. Also, since 729 (4 3) is not bigger than 973 as required for b2, 486 and 729 are
ruled out. That leaves b1 equal to 972, b 2 = 972 (4 3) = 1296 ,
b 3 = 1296 (4 3) = 1728 , b 4 = 1728 (4 3) = 2304 , b 5 = 2304 (4 3) = 3072 , and
b 6 = 3072 (4 3) = 4096 .
Therefore the solution is

b 1 = 972
b 2 = 1296
b 3 = 1728
b 4 = 2304
b 5 = 3072
b 6 = 4096
Editors Comment: This problem was inspired by Problem 7 of the Second Selection Examina
tion held in Bucharest, on April 25, 1999. We are indebted to Kroly Dn of Romania for calling
this problem to our attention.
4/2/12. Prove that every polyhedron has two vertices at which the same number of edges meet.
Solution 1 by Lisa Fukui (12): At least three edges must meet at every vertex of a polyhedron.
If a polyhedron had n vertices and every vertex had different numbers of edges meeting, then the
number of edges meeting at the vertex with the most edges would be at least n + 2 .
A polyhedron with n vertices cannot have a vertex with more than n 1 edges meeting, since
each edge is a segment between two vertices.
Therefore, there are no polyhedra that have a different number of edges meeting at every vertex.
So every polyhedron has two vertices at which the same number of edges meet.
Solution 2 by Agustya Mehta (9/OH): Let the polyhedron have n vertices.
The minimum number of edges that can meet at a vertex is 3. The maximum number of edges
that can meet at a vertex is n 1 . Let us assume that we have a pigeon sitting on each vertex of
our polyhedron (we assume that these pigeons can count and read), and we have pigeonholes
marked 3, 4, 5, ..., n 1 . The pigeons count the number of edges that meet at their vertex, and y
to the pigeonhole with the same number. Since we have n pigeons and only n 3 pigeonholes,
there must be at least one pigeonhole with more than one occupant. The vertices from where
these pigeon roommates originally ew, have the same number of edges meeting. Thus there are
at least two vertices that have the same number of edges meeting. (The argument is still valid
even if we do not have pigeons that can count and read <smile>!)
Solution 3 by Soa Leibman (8/OH): In a polyhedron with n vertices, the number of edges that
can meet at one vertex is an integer between 3 and n 1 . But it is impossible to choose n differ
ent numbers from n 3 numbers (3, 4, 5, ..., n 1 ). So there must be at least two vertices where
the same number of edges meet.
Editorscomment: This problem parallels Problem M15, which appeared in the September/
October 1990 issue of Quantum. It demonstrates the importance of the extreme case in problem
solving.
5/2/12. In ABC , segments PQ, RS, and TU are parallel to

C
sides AB, BC, and CA, respectively, and intersect at the
points X, Y, and Z, as shown in the gure on the right. S T
Determine the area of ABC if each of the segments PQ, Z

RS, and TU bisects (halves) the area of ABC , and if the P Y Q
X
area of XYZ is one unit. Your answer should be in the
form a + b 2 , where a and b are positive integers. A U R B
Solution by Rachel Johnson (11/MN): The area of triangle ABC can be determined using ratios.
Let AB be x. When a triangles area is cut in half by a line parallel to the base, the ratio of the base
of the original to the new base is 2 :1. So PQ is ( x 2) 2 . Since triangles PCQ, UTB, and ASR
have the same areas (half the total), and have the same angles, they are congruent. It follows that
PX and YQ equal (2 x x 2) 2 . Subtracting PS and YQ from PQ shows that XY equals
(3x 2 4x ) 2 . From this, the ratio of the base of ABC to that of XYZ is 2 ( 3 2 4 ) :1. To get
the ratio of the area of ABC to that of XYZ, the ratio of bases is squared: 2 ( 17 12 2) :1. Since
the area of XYZ is 1, the area of ABC is 2 ( 17 12 2) , or in a + b 2 form, 34 + 24 2 .
Editors comments: For more information on the area bisectors of a triangle, the reader is
referred to the article Halving the Triangle by J. A. Dunn and J. E. Pretty in Number 396 (May
1972) of The Mathematical Gazette. We thank our problem editor, George Berzsenyi for posing
this problem.

1/3/12. Find the smallest positive integer with the property that it has divisors ending with every
decimal digit; i.e., divisors ending in 0, 1, 2, ..., 9
Solution 1 by Rishi Gupta (8/CA): Let us rst look at the prime factors of the answer. There
must be a 5, because any number ending in 5 is divisible by 5. There must also be a 2, for the
same reason. Therefore, so far, we have factors: 2 5 = 10 , and the numbers 0, 1, 2, and 5 are
covered.
We have the numbers 3, 4, 6, 7, 8, and 9 left.
If we can nd multiples of the three odd numbers (3, 7, and 9), then the even numbers are covered
because of the multiple of 2 ( 2 3 = 6 , 2 7 = 14 , 2 9 = 18 ). Therefore all we need to
worry about is 3, 7, and 9.
One solution (which may not be smallest) would be to use the LCM of 3, 7, and 9, which is 63.
So the answer would have to be less than or equal to 63 10 .
Lets see if any of the numbers ending in 7 have factors ending in 3 and 9. Seven and 17 are
primes, but 27 has 3 and 9 as factors. So now 27 10 = 270 is a smaller solution. Any other
possible solution below 27 would have to use a multiple of 7 or 17. The only candidates are 14
and 21, neither of which has a factor of 9 or 19.
Verication: 270 is divisible by 10, 1, 2, 3, 54, 5, 6, 27, 18, and 9.
Therefore the solution is 27 10 = 270.
Solution 2 by Matthew Pelc (12/PA): Solution 270.
Call the desired integer N. All positive integers have 1 as divisor, so it is immediately dealt with.
The only way a divisor of N can end in zero is for it to be a multiple of 10, thus 2 and 5 must be
prime factors of N, and N is a multiple of 10. Divisors ending in 5 would have to be multiples of
5, but we already have taken care of 5. So 2 5 covers 0, 1, 2, and 5.
Any multiplication with 5 just produces another number ending in 0 or 5, so 5 is not really a con
sideration in determining the remaining needed factors. Any divisor ending with 7 automatically
gives us another ending with 4 (because 2 is already a factor of N), and similarly any ending with
9 gives us another factor ending with 8. The smallest and most economical solution for 7 and 9
3
would be 27 = 3 = 3 9 . The 4 (27 2 = 54 ) and 8 (9 2 = 18 ) are taken care of simulta
neously. Notice 3 and 6 are covered since 3 is a factor and 6 = 2 3 . So 2 5 27 = 270 is
the smallest positive integer with divisors ending in each decimal digit. If we are not sure, we can
check each multiple of 10 up to 270; but notice only 70, 140, 170 and 210 have divisors ending
with 7. Of these only 210 has a divisor ending in 3, but 210 has no divisor ending in 8. 27 is the
best because it contains 3 and 9, and with 2 it takes care of 3, 4, 6, 7, and 9 all at once.
Solution N = 270.
Solution 3 by Mike Church (12/PA): The smallest such integer is 270.
Proof: First, since the desired integer has a divisor ending in 0, it is clear that this integer must be
divisible by 10. Hence we can eliminate all integers not divisible by 10.
That our integer is divisible by an integer ending in 7 implies it is divisible by an integer among
the set {7, 17, ...}. Hence, we can remove all other integers from our consideration. Thus the ve
smallest candidates for our desired integer are:
70, 140, 170, 210, 270
Next, it is desired that our integer have a factor ending with 3, so it must be divisible by one
among 3, 13, 23, and so on. Thus we can eliminate 70, 140, and 170, and hence there remain
210 and 270
The divisors of 210 are 1, 2, 3, 5, 6, 7, 10, 14, 15, 21, 30, 35, 42, 70, 105, and 210. As this set
lacks elements whose last digits are 8 or 9, 210 can also be disqualied.
Hence, all integers less than 270 have been disqualied. Now evaluate the divisors of 270:
1, 2, 3, 5, 6, 9, 10, 15, 18, 27, 30, 45, 54, 90, 135, and 270.
On inspection, 270 indeed does have the desired property.
Editors Comment: We are grateful to Professor Bruce Reznick of the University of Illinois for
this nice problem, which he rst published (via Martin Gardner) in Science Fiction Puzzle Tales
in the early 1980s.
2/3/12. Assume that the irreducible fractions between 0 and 1, with denominators at most 99,
17
are listed in ascending order. Determine which two fractions are adjacent to ------ in this listing.
76
Editors comment: For a discussion of continued fractions and Farey series see the Editors
Comments following the solutions.
Solution 1 by Christopher Lyons (12/CA):
17
We consider the continued fraction expansion of ------ :
76
17 1
------ = 0 + ---------------------- = [0 ;4, 2, 8]
76 1
4 + -----------
1
2 + ---
8
To nd the fraction that is directly below 17 76 , we must realize that changing the 8 in the
expansion to a higher number would, in fact, make the overall number smaller. So we must gure
out how much to add to the 8 in order to turn the overall fraction into one directly below it on the
list. Let us write the number below 17 76 as having the continued fraction expansion
1
x = 0 ;4, 2, 8 + --- ,
a
where a is a positive real number. When we condense x into its common fractional form, we have
17a + 2
x = ------------------ .
76a + 9
90
We know the denominator is less than or equal to 99, so a ------ . We also know that both 17a and
76
76a must be positive integers. But since 17 and 76 are coprime, a must be an integer itself. The
only positive integer less than 90 76 is 1, so a = 1 , and
19
x = ------ .
85
To nd the fraction directly above 17 76 , we must subtract some amount from the 8. We call
this larger fraction y, and write its continued fraction expansion as
1
y = 0 ;4, 2, 8 --- ,
b
where b is a positive real number. When we condense y into its traditional rational form, we nd
17b 2
y = ------------------ .
76b 9
108
Again, we know the denominator cannot exceed 99, so b --------- . Once more, b must be an integer
76
due to the lack of common factors between 17 and 76. Therefore, b = 1 , and
15
y = ------ .
67
So
19 17 15
------ < ------ < ------ .
85 76 67
Solution 2 by Lisa Leung (10/MD): The irreducible fractions between 0 and 1 with denomina
tors at most 99, listed in ascending order, describes a Farey series of order 99.
Two basic theorems that describe characteristics of successive terms in a Farey series Fn of order
n are:
h h
(1) If --- and ---- are two successive terms of the Farey series, then kh hk = 1 .
k k
h h
(2) If --- and ---- are two successive terms of the Farey series, then k + k > n .
k k
By using theorem (1) with h k set to 17 76 where h is 17 and k is 76,

76h 17k = 1 .
This is similar to solving 76h 1 ( mod 17) . This is solved when h 15 ( mod 17) . Substituting
15 into the rst equation yields k = 67 , and h k = 15 67 .
Thus
17 15
------ < ------ .
76 67
Next, by using theorem (1), but with h k as 17 76 ,
17k 76h = 1 .
This is similar to solving 17k 1 ( mod 76) . This is satised when k 9 ( mod 76) . However,
when k = 9, it contradicts theorem (2). Thus k 9 . When k = 85, it satises both theorems.
h 19
When substituted into the equation, --- = ------ .
k 85
19 17 15
Thus, ------ < ------ < ------ .

85 76 67
19 17 15
Solution 3 by Jason Chiu (12/WY): Answer: ------ < ------ < ------ .
85 76 67
h h
Theorem. If --- and ---- are two successive terms of the Farey series Fn, then kh hk = 1 .
k k
For several proofs of this well-known theorem, see G.H. Hardy and E.M. Wright, An Introduction
to the Theory of Numbers, Fifth Edition, Oxford University Press, London (1979).
a a
Let --- and ---- respectively denote the fractions left-adjacent and right-adjacent to 17 76 in F99.
b b
By the contrapositive of the Theorem, 17b 76a = 1 and 17b 76a = 1 . To solve the rst
Diophantine equation, apply Euclids Algorithm to obtain
76 = 4 17 + 8 ,
17 = 2 8 + 1 ,
8 = 1 8 + 0.
Backing up through these equations gives
1 = 17 2 8 = 17 2 ( 76 4 17) = 9 17 2 76 ,
yielding the solution (a, b) = (2, 9) to the equation 17b 76a = 1 . An alternate way to obtain
17
this solution is to compute the convergents to the continued fraction expansion of ------ . The con
76
vergents in this expansion are
P P 1 1 P 1 2 P 1 17
------0 = 0 , ------1 = 0 + --- = --- , ------2 = 0 + ------------ = --- , ------3 = 0 + ---------------------- = -----
Q0 Q1 4 4 Q2 1 9 Q3 1 76
4 + --- 4 + -----------
2 1
2 + ---
8
P 2
------2 = --- also gives the solution (a, b) = (2, 9) to the equation 17b 76a = 1 .
Q2 9
Since gcd(17,76) = 1, the general solution of the equation 17b 76a = 1 is a = 2 + 17t ,
b = 9 + 76t . For any solution (a,
b) to 17b 76a = 1 ,
17 a 1
k = ------ --- = --------

76 b 76b
so that choosing the largest value of b minimizes k. Hence the ordered pair (a, b) = (19, 85)
gives the smallest value of k and 19 85 is left adjacent to 17 76 .
Similarly, the general solution to 17b 76a = 1 is a = 2 + 17t , b = 9 + 76t so the

ordered pair (a, b) = (15, 67) gives the smallest value
a 17 1
k = ---- ------ = ----------
b 76 76b
and 15 67 is right-adjacent to 17 76 .
Editors Comment: A similar problem appeared in the January 1999 issue of Mathematical
Digest, an excellent South African mathematical journal for high school students. This problem
does allow for computer solutions; we welcome them, but commend such solutions only if they
are based on a clever algorithm and deal with accuracy. We thank our problem editor, Dr. George
Berzsenyi, for this problem.
This interesting problem gives us an opportunity to expand on two interesting mathematical top
ics, continued fractions and Farey series. Continued fractions were discussed in the Solutions to
Round 1 of Year 10, available on the USAMTS web site.
h
By a Farey series Fn of order n, we mean the set of all fractions --- with 0 h k , gcd(h,k) = 1,
k
1 k n , and arranged in ascending order of magnitude. For example, F5 is
0 1 1 1 2 1 3 2 3 4 1
---, ---, ---, ---, ---, ---, ---, ---, ---, ---, --
1 5 4 3 5 2 5 3 4 5 1
These series have remarkable properties, some of which are mentioned in the solutions above. I
restate two of the theorems here so you might easily conrm them for this example:
h h
(1) If --- and ---- are two successive terms of the Farey series, then kh hk = 1 .
k k
h h
(2) If --- and ---- are two successive terms of the Farey series, then k + k > n .
k k
Proofs of these and other properties are given in G.H. Hardy and E.M. Wright, An Introduction to
the Theory of Numbers, Fifth Edition, Oxford University Press, London (1979).
5 2 2
3/3/12. Let p ( x ) = x + x + 1 have roots r 1, r 2, r 3, r 4, r 5 . Let q ( x )
= x 2 . Determine
the product q ( r 1 )q ( r 2 )q ( r 3 )q ( r 4 )q ( r 5
) .
5 2
Solution 1 by Soa Leibman (8/OH): We can write the polynomial x + x + 1 in the form
(x r 1 )(x r 2 )(x r 5 ) .
2 2 2
The product q ( r 1 )q ( r 2 )q ( r 5 ) = (r 1 2 )(r 2 2 )(r 5 2 ) .
2
r i 2 = ( 2 r i )( 2 r i ) .
So the product can be written as
( 2 r 1 )( 2 r 2 )( 2 r 5 )( 2 r 1 )( 2 r 2 )( 2 r 5 )
= p( 2) p( 2)
5 2 5 2
= [( 2) + ( 2) + 1] [( 2) + ( 2) + 1]
= 23
The solution is -23.
Solution 2 by Sarah Emerson (12/WA):

Any polynomial of degree n with leading coefcient 1 can be factored as
(x z 1 )(x z 2 )(x z n 1 )(x z n )
where z 1, z 2, z n are the roots of the polynomial, are complex numbers of the form a + bi , (a
and b are real numbers and either a or b, or both, can be zero).
5 2
Therefore, p ( x ) = x + x + 1 can be expressed as
p ( x ) = (x r 1 )(x r 2 )(x r 3 )(x r 4 )(x r 5 ) .
2
Also, q ( x ) = x 2 has two roots, s1 and s2, so q ( x ) = (x s 1 )(x s 2 ) .
Then q ( r 1 ) = (r 1 s 1 )(r 1 s 2 ) and
q ( r 1 )q ( r 2 )q ( r 3 )q ( r 4 )q ( r 5 ) =
= (r 1 s 1 )(r 1 s 2 )(r 2 s 1 )(r 2 s 2 )(r 3 s 1 )(r 3 s 2 )(r 4 s 1 )(r 4 s 2 )(r 5 s 1 )(r 5 s 2 )
= (s 1 r 1 )(s 1 r 2 )(s 1 r 3 )(s 1 r 4 )(s 1 r 5 )(s 2 r 1 )(s 2 r 2 )(s 2 r 3 )(s 2 r 4 )(s 2 r 5 )
since we changed an even number of signs and reordered terms. This is equivalent to
p ( s1 ) p ( s2 ) .
2
Factoring q ( x ) = x 2 = (x 2)(x + 2) gives the two roots of q(x): s 1 = 2 and
s 2 = 2 .
Plug these roots of q(x) into the original equation for p(x).
q ( r 1 )q ( r 2 )q ( r 3 )q ( r 4 )q ( r 5 ) = p( 2) p( 2)
5 2 5 2
= [( 2) + ( 2) + 1] [( 2) + ( 2) + 1]
= 23 .
This method will work with any two polynomials; however, if both polynomials are of orders that
are odd numbers, the product of f (roots of g ( x )) will be the negative of g(roots of f ( x ))
because the odd number of negative signs will not all cancel out.
Editors Comment: We are thankful to Professor Rob Hochberg of the University of Connecti
cut for this interesting problem.

4/3/12. Assume that each member of the sequence i i = 1
is either a + or a sign. Deter-
mine the appropriate sequence of + and signs so that
2 = 6 1 6 2 6 3 .
Also determine what sequence of signs is necessary if the sixes in the nested roots are
replaced by sevens. List all integers that work in the place of sixes and the sequences of signs
that are needed with them.
Solution by Ho Seung (Paul) Ryu (9/KS):

Firstly, we notice that 2 = 6 2 . So, if we replace the 2 on the right side with the identical
expression 6
2 in innite number of times, we will have obtained
2 = 6 6 6 6 . So the sequence of signs for 6 is simply -,-,-,-,... .
For number 7, 2 = 7 3 , which is of little help, but 3 = 7 + 2 , so 2 = 7 7 + 2.
So, by replacing 2 on the right side, we get 2 = 7 7 + 7 7 + , just alternating

signs -,+,-,+, ... .
For the number 8, we will prove that 8 cannot be used. That is for a simple reason, that
2 = 8 4 , but no sequence of radicals and 8s can be larger than 8 + 8 + . This quan
2 2
tity is given by x = 8+ 8+ = 8 + x . x = 8 + x and thus x x 8 = 0.
1 1 4 1 ( 8 ) 1 33
x = ------------------------------------------------ = ------------------- . This is less than 4, so 8 does not work in the problem.
2 2
No numbers greater than 8 satisfy the conditions set forth in the previous statement, so now we
can investigate numbers less than 6.
Now for 5, 2 = 5 1 , 1 = 5 4 , 4 = 5 + 11 , 11 = 5 + 116 . The numbers just
never cease to increase, so we can never have a working sequence of signs.
For 4, 2 = 4 0 , and we are stuck right away.
For 3, 2 = 3 + 1 , 1 = 3 2 , and so 2 = 3 + 3 3 + 3 . Thus the sequence of

signs is +,-,+,-, ... . .
As for 2, 2 = 2 + 2 , so 2 = 2 + 2 + and we have sequence +,+,+,+, ... .
Lastly, for 1, 2 = 1 + 3, 3 = 1 + 8, 8 = 1 + 63 , and we just keep increasing. Therefor,
1 does not work.
Thus for the number six, the required sequence of signs is just a string of minus signs -,-,-,-, ... .
The numbers that work in place of six, and the sequence of signs needed are:
2: All plus signs +,+,+,+, ... .
3: Alternating plus/minus +,-,+,-, ... .
7: Alternating minus/plus -,+,-,+, ... .
Editors comment: This wonderful problem was proposed by Dr. Rodrigo Gomez of NSA.
5/3/12. Three isosceles right triangles are M

erected from the larger side of a rectan
gle into the interior of the rectangle, as
shown on the right, where M is the mid
point of that side. Five circles are
inscribed tangent to some of the sides
and to one another as shown. One of the
circles touches the vertex of the largest
triangle.
Find the ratios among the radii of the
ve circles.
Solution by Lisa Leung (10/MD): Since M

the triangles are isosceles right triangles
that are erected from the larger side of a
a a
rectangle, the two circles marked as a are
congruent and the two circles marked as c
are congruent.
Without loss of generality, let the length of

the rectangle be 1 unit, and ra, rb, and rc be
the radius of circle a, b and c respectively. c c
b
As shown in Figure 2,
Figure 1.
1
r a +
r a 2 = ---
4
1 21
r a = ------------------------ = ---------------- (1)
4(1 + 2) 4 1/2
From the base of gure 3,
ra
1
2 2
--- =
( r c + r b ) ( r c r b ) = 2 r b r c + r c (2)
2
From the width of the rectangle in Figure 3,
(1 +
2)r c = ------- --- + 2r b
1 1
22 Figure 2.
r b = 1 + ------- r c ---
2 1
(3)
2 4
Substitute into (2),
2 2 r
--- = 2 1 + ------- r c ----c + r c
1 1/2
2 2 4
2 2 r
--- r c = 2 1 + ------- r c ---c
1 1/2
2 2 4
2 2 r
(1 2r c ) = 16 1 + ------- r c ---c
2
2 4 rc
2 2
4r c 4r c + 1 = (16 + 8 2)r c 4r c
2 1 1 rc
r c = ---------------------------- = --------------------------2 rb r
4 ( 3 + 2 2) 4 ( 1 + 2) b rc-rb
Take the positive square root, rb
1 21
r c = ------------------------ = ---------------- = 2r a (4)
2 ( 1 + 2) 2 Figure 3.
Substitute into (3),
r b = 1 + ------- ---------------- --- = ---------------- = r a
2 21 1 21
2 2 4 4
Thus, the three smallest circles have the same radii. The ratios among the radii of the ve circles
are:
a:a:b:c:c = 1:1:1:2:2.
Editors comments: We are indebted to Professor Hiroshi Okumura of Japan for this wonderful
shungaku problem. Professor Okumura is a longtime enthusiastic promoter of the Japanese
equivalent of the USAMTS.

1/4/12. Determine all positive integers with the property that they are one more than the sum of
the squares of their digits in base 10.
Solution 1 by Steve Byrnes (10/MA): Let the number be, in base ten, ...a5a4a3a2a1a0, where
any of the as may be zero. Then the condition is equivalent to
2 2 2 2 2 1 2 3 4
1 + a 0 + a 1 + a 2 + a 3 + a 4 + = a 0 + a 1 10 + a 2 10 + a 3 10 + a 4 10 + ,
or
0 = 1 + a 0 (a 0 1 ) + a 1 (a 1 10 ) + a 2 (a 2 100 ) + a 3 (a 3 1000 ) + a 4 (a 4 10000 ) +
a 1 (a 1 10 ) + a 2 (a 2 100 ) + a 3 (a 3 1000 ) + = 1 a 0 (a 0 1 ) 1 9 8 = 73 .
All terms on the left side are negative, since the ai are integers 0 to 9. Also, after the leftmost
term, the terms are all zero or smaller that 73 . Hence, we must conclude that
0 = a2 = a3 = a4 = a5 = .
Therefore a 0 (a 0 1 ) + 1 = a 1 (10 a 1 ) .
As a0 goes from 0 to 9, the left side can be 1, 3, 7, 13, 21, 31, 43, 57, or 73. As a1 goes from 0 to
9, the right side can be 0, 9, 16, 21, 24, 25, 21, 16, 9, or 0. The only number these lists have in
common is 21, so a 0 = 5 and a 1 = 3 or 7. Hence, the two possible numbers are 35 and 75.
Solution 2 by Yuen-Joyce Liu (9/MA): 35 and 75 are the only two positive integers with the
property that they are one more than the sum of the squares of their digits.
Table 1:
Number of digits in Maximal integer repre maximal sum of the squares of k digits
integer (n) sented by k digits (10n-1) plus 1 (92n+1)
1 1 82
2 10 163
3 100 244
4 1000 325
5 10000 406
Table 1 suggests that the largest possible number of digits in an integer which satises the prob
lem requirement is 3. We use mathematical induction to prove
Page 1
n1 2
Claim: 10 > 9 n + 1 = 81n + 1 for integer n > 3 .
41
Proof of Claim: For n = 4, 10 = 1000 > 81 4 + 1 = 325 .
k1
Assume for n = k > 3 , that 10 > 81k + 1 . Then for n = k + 1 we have
(k + 1 ) 1 k1
10 = 10 10
> 10 ( 81k + 1 )
= 810k + 10
= 81(k + 1 ) + (729k 71 )
> 81(k + 1 ) + 1
Therefore the Claim is proved, and an integer which satises the problem requirement can have at
most 3 digits.
The maximal sum of squares of three digits plus one is 244. So there can be no integer n meeting
the requirements with n > 244 . The largest possible integer of three digits with 1 or 2 in the hun
dreds digit is 299, whose sum of the square of its digits plus one is 166, so there can be no integer
n meeting the requirements with n > 166 . For 100 n 166 , the value of n which has largest
sum of its squares plus one is n = 159, whose sum of squares plus one equals 107. So there can be
no integer n meeting the requirements with n > 107 . For 100 n 107 , the value of n which
has largest sum of its squares plus one is n = 107, whose sum of squares plus one equals 50. So
there can be no integer n meeting the requirements with n > 99 . So the largest possible number
of digits in an integer which satises the requirement is reduced to 2.
In order for a positive integer with the tens digit a and the units digit b to satisfy the problem
requirement, we have
2 2
10a + b = 1 + a + b
2 2
a 10a + ( b b + 1 ) = 0
2
10 100 4 ( b b + 1 )
a = ---------------------------------------------------------------
2
Table 2:
b 0 1 2 3 4 5 6 7 8 9
2 9.80 9.80 9.38 8.49 6.93 4 Unreal Unreal Unreal Unreal

100 4 ( b b + 1 )
a ... ... ... ... ... 7 or 3 ... ... ... ...
As calculated and shown in Table 2, we nd that 35 and 75 are the only two positive integers with
the property that they are one more than the sum of the squares of their digits in base 10.
1/4/12. Editors Comment: We are grateful to Professor Bruce Reznick of the University of
Illinois for communicating this problem to us. He created this problem for the 1990 Friendly
Competition (The Indiana College Mathematics Competition).
Page 2
n n n n
2/4/12. Prove that if n is an odd positive integer, then N = 2269 + 1779 + 1730 1776 is
an integer multiple of 2001.
Solution 1 by Anatoly Preygel (10/MD): Note that 2001 factors into primes as 3 23 29 .
Thus it is sufcient that 3 N , 23 N , and 29 N .
n n n1 n2 n2 n1
Note that for all integers n, a b = (a b)(a +a b++ab +b ) and for
n n n1 n2 n2 n1
odd integer n, a + b = (a + b)(a a b+ab +b ) . Thus, for odd n,
n n n n
(a + b) (a + b ) and (a b) (a b ) .
n n n n
In this case, note that N = 2269 + 1779 + 1730 1776 and that
n n
(2269 + 1779 ) (2269 + 1779 )
2269 + 1779 = 4048 = 23 176
n n
(1730 1776 ) (1730 + 1776 )
1730 1776 = 46 = 23 ( 2 )
Thus, 23 N .
n n n n
Now, note that N = 2269 1776 + 1779 + 1730 (just reordering terms), and that
n n
(2269 1776 ) (2269 + 1776 )
2269 1776 = 493 = 29 17
n n
(1779 + 1730 ) (1779 + 1730 )
1779 + 1730 = 3509 = 29 121
Thus, 29 N .
Last, note that 1779 = 3 593 , 1776 = 3 592 , and that
n n
(2269 + 1730 ) (2269 + 1730 )
2269 + 1730 = 3999 = 3 1333
Thus 3 N .
n n n n
Thus we have 2001 N = 2269 + 1779 + 1730 1776 for odd integer n.
Solution 2 by Valerie Lee (10/NY):

n n n n
N = 2269 + 1779 + 1730 1776
n n n n
N is multiple of 2001?: 2001 x = 2269 + 1779 + 1730 1776 .
n n n n
Factor 2001: 3 23 29 x = 2269 + 1779 + 1730 1776 .
If N is divisible by the factors of 2001, then it has to be divisible by 2001. Writing N modulo
these factors, the expressions should be equal to zero. If this is true, then N is a multiple of 2001.
Page 3
n n n n
Consider mod 29: 2269 + 1779 + 1730 1776 (mod 29)
n n n n
Replace terms by remainders mod 29. 7 + 10 + 19 7 (mod 29)
n n n n n n
19 ( 10 ) (mod 29) 7 + 10 + ( 10 ) 7 (mod 29)
Everything cancels if n is odd 0
n n n n
Continue mod 23: 2269 + 1779 + 1730 1776 (mod 23)
n n n n
n n n n n n
15 ( 8 ) (mod 23) ( 8 ) + 8 + 5 5 (mod 23)
n n n n
Last mod 3: 2269 + 1779 + 1730 1776 (mod 3)
n n n n
n n n n n n
2 ( 1 ) (mod 3) 1 + 0 + ( 1 ) 0 (mod 3)
Thus, N is divisible by 2001 when n is odd.
Editors Comment: We thank our problem editor, Dr. George Berzsenyi, for this timely prob
lem.
3/4/12. The gure on the right can be divided into two congruent halves that are
related to each other by a glide reection, as shown below it. A glide reection
reects a gure about a line, but also moves the reected gure in a direction
parallel to that line. For a square-grid gure, the only lines of reection that

keep its reection on the grid are horizontal, vertical, 45 diagonal, and

135 diagonal. Of the two gures below, divide one gure into two congruent
halves related by a glide reection, and tell why the other gure cannot be
divided like that.
Page 4
Solution 1 by Katherine Herbig (11/WA): The gure must slide along a line
parallel to the line it is reected across. Therefore, the gure and its reection
must have the same number of squares that either cut the line or share an edge
with it. Therefore, the line of reection must cut through an even number of
squares, or touch the same number of edges on each side of the line.
In addition, there must be the same number of squares on each side of the line
because all squares on one side of the line at the beginning of a glide reection
will be on the other side after the glide reection.
There is only one line in either of the squares that ts both requirements. It is
shown in the top diagram on the right. The division is shown in the lower dia
gram.
The other gure cannot be divided by a glide reection because it has only one
line that is even near the middle of the gure. This line crosses through an even number of
squares, but does not divide the gure evenly.
Solution 2 by Katheryn Green (11/WV):
The gure can be divided using the 45 glide reection:
For a glide reection to be possible, the reecting line must divide the
gure so that there is an equal number of squares on each side. Notice that there is a total of six
squares on each side of the line in the example, and a total of nine on each side in the solution
above. The gure at left below cannot be divided into two congruent halves related by a glide
reection because, as shown, it is impossible to divide the gure in half with a horizontal, vertical,
45 , or 135 line that lies along the grid lines.
5
9.5 9.5
7 11
8.5 13 8.5
11.5 8.5 8.5 10
6.5 9.5 9.5 8
Page 5
Solution 3 by Igor Zhitnitskiy (10/NY): It is given C
that the line of reection can only be horizontal, ver
tical, or diagonal at 45 degrees in either direction. 18 17 16
Thus for a given gure, there is a maximum of four B
lines of reection. However, a line of reection must 15 14 13 12
also have an equal area on either side (since a glide
reection only moves the reected portion parallel to
the line and does not change its area). It must also go 11 10 9
through corners or midpoints of lines on the square
grid; it cannot go through any other point because a 8 7 6 5
reection would not be possible. Keeping this in
mind, it is simple to verify, by trial and error, that
only one line of reection exists for the gure at right 4 3 2 1
(expressed as a broken line from the corner of block 4 A
to the corner of block 13). The other gure has no
such lines of reection that maintain equal areas
while intersecting only corners and midpoints; thus
any kind of glide reection is impossible.
Once the line of reection is identied the procedure for determining the congruent halves is as
follows. Since blocks 1 and 18 are the only two blocks with a distance of 2 diagonals from their
distant corners to the line of reection, they must be corresponding parts of the two congruent
halves. Let us denote blocks associated with block 1 as white and the blocks associated with
block 18 as gray, and the side of reection on which block 18 is located as the gray side,
with the other side of the reection line as the white side. From this relationship it is clear that
blocks on the white side are related to their corresponding blocks on the gray side by a reection
and 1 block movement up along the line of reection (as in the diagram A B C ). If block
12 were white, then its glide reection would be 1 block along the axis away from block 16.
However there is no block there. Thus block 12 must be gray. There must then be a white block
reected across and 1 block down from 12. This is block 14; thus block 14 is white. These two
objects must have their blocks connected to form one gure, so blocks 18 and 12 must be
bridged by either blocks 17, 16, and 13, or blocks 15, 11, 10, 9,and 13. However if the later is
used, block 14 is isolated from block 1. Thus blocks 17, 16,and 13 are gray. The corresponding
blocks for these are white, so blocks 5, 9, and 10 are white. If block 8 is gray, then there must be
a white block one block down from block 3 along the axis. There is no block there, so block 8
must be white, and its counterpart, block 6, is gray. 6 must not be isolated from the gray gure;
thus 15, 11, and 7 must be gray. Their respective counterparts 2, 3, and 4 must be white. As
stated earlier the other given gure has no applicable lines of reection, so this is the one and only
glide reection possible for the two gures.
Editors Comment: We are indebted to Dr. Erin Schram of the National Security Agency for
formulating this problem, and to Professor Kimmo Eriksson of Sweden, whose article Splitting a
Polygon into Two Congruent Pieces in the May 1996 issue of The American Mathematical
Monthly served as an inspiration for this problem.
Page 6
4/4/12. Let A and B be points on a circle which are not diametrically opposite, and let C be the
midpoint of the smaller arc between A and B. Let D, E, and F be the points determined by the
intersections of the tangent lines to the circle at A, B, and C. Prove that the area of DEF is
greater than half of the area of ABC .
Solution 1 by Rishi Gupta (8/CA): Let O be the

center of the circle, with r as the radius. FO splits F
the diagram in half, so one side is symmetric to the
other.
First, I saw that AF > w because of right GAF . m

y
AF = m + n , as shown in the diagram, so
m + n > w . Now DA DC (because both are tan
gents), so we can replace our equation with D n C E
m + DC > w . Because of right CDF , m > DC . x
Therefore, 2m > w . A H G B
w
r
x y O
Then, I wanted to prove that --- = ---- . I drew a per-
n m
pendicular line segment down from D to point H on
AB, so that DH = x . Now HAD CDF
because both are right triangles and
DAG FDC ( DE AB ), Therefore,
HD CF x y
--------- = -------- , or --- = ---- .
DA FD n m
Going back to my previous inequality, 2m > w , we have:

2m > w
2m ---- > w ---

y x
m n
wx
ny > ------- .
2
The area of DEF = ny and the area of ABC = wx . Substituting, we have the area of
DEF is greater than half the area of ABC .
Page 7
Solution 2 by Nathaniel Jones (12/IA): In the g
ure at the right, let CG = a , BG = b , EG = x , E
and CF = y . Since m AC = mBC ,

mABF = 2m CBF . Let
mABC = mCBF = . Using trigonometry,
a x-a
tan = --- , and
b
F 2 y C D
2 ---
a
tan b
tan 2 = ---------------------
- = -------------------
2ab
- = ----------------
x
= --- . a
2 2 2 b B A
1 tan a 2 b a b G
1 ----2-
b
2
2ab
Thus, x = ----------------
2 2
. Since both AB and DF are per
b a
pendicular to EG, mABE = mDFE . This
2ab 2
---------------- a
2ab xa b
2
a
2

means that tan2 = ----------------
2 2
= ----------- = -.
------------------------------
b a y y
2 2
a +b
Combining this with the expression above for tan2 gives y = ----------------- .
2b
2ab
The area of ABC is --------- = ab , and
2
1 a + b 2ab
2 2 2 5 3 2 4
the area of DEF is 2 --- ----------------- ----------------
a + 2a b + ab
a = - . To show that the area of
----------------------------------------
2 2b b 2 a 2 3
2(b a b)
2
5 3 2 4
1 a + 2a b + ab
DEF is greater than half the area of ABC , it must be proven that --- ab < ----------------------------------------
3 2
- , or
2 2(b a b)
5 3 2 4
a + 2a b + ab 1 4 2 2
- --- ab . This statement simplies to 0 < a + 3a b , and since both a
that 0 < ----------------------------------------
3
2(b a b)
2 2
and b are always positive, the statement is always true. So the area of DEF is greater than half
the area of ABC .
Page 8
Solution 3 by Kevin Yang (11/IL): Let W be the
midpoint of line segment AB. Line segment AD is F
congruent to DC because of the two-tangent rule.
Now, we know that AC is longer than AW, because
the hypotenuse of any right triangle is longer than
either leg. Also, the lengths of AD plus that of DC
will be greater than the length of AC because of the
triangle inequality. So, the length of AD plus that of
DC will be greater than the length of AW. And thus, D C E
DC will be longer than half the length of AW.
From this we know that DE is more than half the A B

W
length of AB. Since the triangles DFE and ABF are
similar (because AB is parallel to DE), we also know
that FC is more than half the length of FW. In other
words, FC is longer than CW. The area of a trian
gle is (1/2)(base)(height), so since the base of tri
angle DEF is more than half that of triangle ABC,
and its height greater than that of triangle ABC,
then the area of triangle DEF must be more than
half the area of triangle ABC.
Editors comment: This historically interesting problem was proposed by Dr. Peter Anspach of
NSA. He found it in a paper by mathematician/astronomer Christiaan Huyghens, who used it as a
stepping stone for calculating the digits of pi.
5/4/12. Hexagon RSTUVW is constructed by starting with

U
a right triangle of legs measuring p and q, constructing
squares outwardly on the sides of this triangle, and then
connecting the outer vertices of the squares, as shown in
the gure on the right. T
V
Given that p and q are integers with p > q , and that the q
area of RSTUVW is 1922, determine p and q. W p
R S
Page 9
Solution 1 by Laura Pruitt (11/MA):
From the given information we can deter- U
mine the following segment lengths,
areas, and angles:
2 2
2 2 p +q
( ABUT ) = p +q
2
( ACRS) = p p +q
2 2
2
(BCWV ) = q T
pq
(ABC) = ----- B 2 2
2 V q p +q
2 2
pq p +q
(CRW ) = -----
2 q q
We now need only the areas of BUV q

C
W p A
and AST .
m(VBU ) + m(CBA) = 180 so p p

p 2 2
sin (VBU ) = sin (CBA) = --------------------- p +q
2 2
p +q
Similarly p
q R S
sin (TAS) = sin (BAC) = ---------------------
2 2
p +q
So
2
(BUV ) = --- q p + q sin (VBU ) = --- q p + q ---------------------- = ------
1 2 2 1 2 p pq
2 2 p 2 + q 2 2
2
(AST ) = --- p p + q sin (TAS) = --- q p + q ---------------------- = ------
1 2 2 1 2 q pq
2 2 p +q 2 2 2
We now know the areas of all the pieces of the hexagon in terms of p and q. So
1922 = ( ABUT ) + ( ACRS) + (BCWV ) + (ABC) + (CRW ) + (BUV ) + (AST )
2 2 2 2 pq pq pq pq 2 2
= p + q + p + q + ------ + ------ + ------ + ------ = 2 p + 2q + 2 pq
2 2 2 2
2 2
2 2 q q 4(q 961 )
So 961 = p + pq + q . The quadratic equation gives p = ---------------------------------------------------------- . Looking
2
2 2 2
at the determinant, q 4(q 961 ) = 3844 3q , we see that, since the discriminant must be
greater than or equal to zero, since q < p , and since q is an integer, it follows that 1 q 17 .
Furthermore, from this range we can see that the square root of the discriminant must be between
55 and 61 inclusive. Call this value s.
Page 10
2
3844 s 2
---------------------- = q , which must be a perfect square. Of the range 55 to 61, only one value of s
3
2
3844 3481
works: s = 59. -------------------------------- = 121 giving q = 11 .
3
2 2
11 11 4 ( 11 961 )
This gives p = ------------------------------------------------------------------- = 24 .
2
So p = 24, q = 11.
Solution 2 by Ho Seung (Paul) Ryu (09/KS): To make

U
referring to the regions easier, label the vertices of the trian-
gle A, B, and C. [ABC], the notation meaning the area of
p
pq
ABC, is obviously ------ . We also have q q T
2
A
V p
[ABWV] = q2,
q
[BCSR] = p2
W B p C
[ACTU] = p2 + q
2 from the Pythagorean Theorem.
Additionally, by looking at the diagram, we also see that
pq
[VAU] = [CST] = ----- , so [RSTUVW] is the sum of all the
2 R S
2 2
individual areas, which sum to 2 p + 2 pq + 2q . So we
2 2 2 2 2
now have that 1922 = 2 p + 2 pq + 2q , or simply 961 = p + pq + q = ( p + q) pq , or
2 2
961 + pq = ( p + q) . Since 961 = 31 , p + q is greater than 31.
If p + q is 32, pq must equal 63. No such combination exists. The same occurs for p + q equal
to 33 and 34. But if p + q = 35 , then ( p, q) = (24, 11 ) satises the inequality.
So if (p, q) = (24, 11), then the area of RSTUVW is 1922.
Editors comments: This problem was inspired by Problem 54 in Quantum Quandaries, pub
lished in 1996 by the National Science Teachers Association. We thank Dr. Berzsenyi for posing
this problem.
Page 11

1/1/13. Determine the unique positive two-digit integers m and n for which the approximation
m m
---- 0.2328767 is accurate to seven decimals, i.e., 0.2328767 ---- < 0.2328768 .
n n
Solution 1 by Grigori Avramidi (11/NM): We can use Euclids Algorithm to obtain a contin
ued fraction approximation for 0.2328767 and then check whether it ts our bounds. A continued
fraction is of the form
1
a 1 + ------------------------------------------ .
1
a 2 + -----------------------------
1
a 3 + -----------------
1
a 4 + -----

It can also be written as [a 1, a 2, a 3, a 4, ] . This version of Euclids Algorithm works as fol
lows. One takes a number and expresses it as the greatest integer of it (this is the number in
parenthesis) plus the remainder. This greatest integer corresponds to a 1 in the continued fraction.
Next one takes the inverse of the remainder and does the same thing. This can be repeated an
unlimited number of times, each reiteration increasing the accuracy of the calculation. A simple
continued fraction may be nite or innite. It may also be repeating.
0.2328767 = ( 0 ) 1 + 0.2328767 (1)
------------------------- ( 4 ) 1 + 0.2941179 (2)
0.2328767
------------------------- ( 3 ) 1 + 0.3999974 (3)
0.2941179
------------------------- ( 2 ) 1 + 0.5000164 (4)
0.3999974
------------------------- ( 1 ) 1 + 0.9999344 (5)
0.5000164
1
------------------------- ( 1 ) 1 + 0.0000656 (6)
0.9999344
------------------------- (15243) 1 + 0.9024390 (7)

0.0000656
This tells us that 0.2328767 is approximated by the nite continued fraction
Page 1
[0, 4, 3, 2, 1, 1, 15234] .
Now we set up a table to obtain the convergents to 0.2328767 .

m x = a x (m x 1 + m x
2 )
n x = a x (n x 1 + n x 2 )
x - - 1 2 3 4 5 6 7
ax - - 0 4 3 2 1 1 15243
mx 0 1 0 1 3 7 10 17 259141
nx 1 0 1 4 13 30 43 73 1112782
m 17
We see that the closest approximation ---- where m and n are two-digit numbers is ------ . We check
n 73
17 17
and nd that ------ = 0.23287671 so 0.2328767 ------ < 0.2328768 .
73 73
Thus
m = 17 and n = 73 .
Another way to evaluate a continued fraction is by writing out the continued fraction.
1 1 1 1 17
0 + ----------------------------------------- = 0 + ---------------------------------- = 0 + ---------------------- = --------------- = ------ .
1 1 1 5 73
4 + ------------------------------- 4 + ------------------------ 4 + ------------ 4 + -----
1 1 2 17
3 + ---------------------- 3 + -------------- 3 + ---
1 (5 2) 5
2 + -----------
1
1 + ---
1
We see that we get the same result

m = 17 and n = 73 .
Solution 2 by Mike Hansen (12/IA): We will nd m and n by the continued fractions method.
1 1 1
0.2328767 ------------------------------------- -------------------------------------------- ------------------------------
4 + 0.29411787 1 1
4 + ---------------------------------- 4 + ---------------------
3 + 0.3999974 1
3 + -----------
1
2 + ---
2
Page 2
1
----------------------------------------------
1
4 + ------------------------------------
3 + 0.39999737
--------------------------------------------------------
1
4 + ----------------------------------------------
1
3 + ------------------------------------
2 + 0.50001643
------------------------------
1
4 + ---------------------
1
3 + -----------
1
2 + ---
2
17 m
= ------ = ---- 0.2328767
73 n
Notice that we stopped at --- because if we kept going, m and n would not both be two digit
2
numbers.
So m = 17 and n = 73 .
Solution 3 by Shyam Amin (10/NY): Use Farey sequences with = 0.2328767 .

0 1 0 1 1
0 = --- < < --- = 1 . The next fraction between --- and --- with least denominator is --- = 0.5 .
1 1 1 1 2
0 1 0 1 1
--- < < --- . The next fraction between --- and --- with least denominator is --- = 0.3 .
1 2 1 2 3
0 1 0 1 1
1 3 1 3 4
0 1 0 1 1
1 4 1 4 5
1 1 1 1 2
5 4 5 4 9
2 1 2 1 3
--- < < --- . The next fraction between --- and --- with least denominator is ------ = 0.2307692 .
9 4 9 4 13
3 1 3 1 4
------ < < --- . The next fraction between ------ and --- with least denominator is ------ = 0.2352941 .
13 4 13 4 17
3 4 3 4 7
------ < < ------ . The next fraction between ------ and ------ with least denominator is ------ = 0.23 .
13 17 13 17 30
Page 3
3 7 3 7 10
------ < < ------ . The next fraction between ------ and ------ with l. d. is ------ = 0.232558 .
13 30 13 30 43
10 7 10 7 17
------ < < ------ . The next fraction between ------ and ------ with l. d. is ------ = 0.2328767 .
43 30 43 30 73
m
When m = 17 and n = 73 , then ---- 0.2328767 .
n
Editors Comment: This problem, created by our Problem Editor, George Berzsenyi, was
inspired by a similar problem in the January 1999 issue of Mathematical Digest. It is also related
to problem 2/3/12; continued fractions and Farey sequences are discussed further in the solutions
to that problem available on our web site.
2/1/13. It is well known that there are innitely many triples of integers (a, b, c) whose greatest
2 2 2
common divisor is 1 and which satisfy the equation a + b = c . Prove that there are also
innitely many triples of integers (r, s, t ) whose greatest common divisor is 1 and which sat
2 2 2
isfy the equation ( rs ) + ( st ) = ( rt ) .
Solution by Anna Pierrehumbert (11/IL):

2 2 2
Claim: If a, b, and c are integers such that a + b = c and gcd(a, b, c) = 1 , then r = ac ,
2 2 2
s = ab , and t = bc satisfy ( rs ) + ( st ) = ( rt ) and gcd(r, s, t ) = 1 .
2 2 2 2 2 2 2 2 2 2
Proof: Multiplying a + b = c by (abc) gives (a bc) + (ab c) = (abc ) , which can be
2 2 2
written as (( ac ) ( ab )) + (( ab ) ( bc )) = (( ac ) ( bc )) . Letting r = ac , s = ab , and t = bc .
2 2 2
yields ( rs ) + ( st ) = ( rt ) . It is given that gcd(a, b, c) = 1 . Furthermore, no two of a, b,
and c have a common factor, because if they did, the common factor would be factored from two
2 2 2
terms of a + b = c and not the third. Therefore, because r, s, and t share no common factor
a, b, or c, gcd(r, s, t ) = 1 .
2 2 2
There are innitely many integers a, b, c such that a + b = c and gcd(a, b, c) = 1 . As
shown above, each triple (a, b, c) corresponds to a triple (r, s, t ) satisfying
2 2 2
( rs ) + ( st ) = ( rt ) and gcd(r, s, t ) = 1 . Consequently, there are innitely many triples
2 2 2
(r, s, t ) satisfying ( rs ) + ( st ) = ( rt ) and gcd(r, s, t ) = 1 . Q.E.D.
Editors Comment: The author of this problem is Mr. Surat Intasang, a former USAMTS partic
ipant, who recently initiated the equivalent of the USAMTS in his native Thailand. We thank him
for his contribution and wish him well in his endeavors.
Page 4
cos3 x 1 sin3x
3/1/13. Suppose --------------- = --- for some angle x, 0 x --- . Determine -------------- for the same x.
cos x 3 2 sin x
Solution 1 by Julia Mundy (12/NY): There are two methods of solving this problem. The use of
technology will produce an explicit value for x. This value can be substituted into the second
expression and the answer found. However, the solution can also be found more elegantly with
out the use of technology or an explicit value for x.
cos ( 3 x ) = cos (2x + x) . This expression can be evaluated using the trigonometric double angle
and summation rules.
cos ( 3 x ) = cos ( 2x ) cos ( x ) sin ( 2x ) sin ( x )

2
(1 2 sin x) cos ( x ) 2 sin ( x ) cos ( x )
= ----------------------------------------------------------------------------------------
sin ( x )
Factoring out cos ( x

) and simplifying, we receive:
2
cos ( 3 x ) = cos ( x
)(4 sin x + 1 ) .
cos ( 3 x ) 2 1
Therefore, -------------------- = 4 sin x + 1 = --- .
cos ( x ) 3
sin ( 3 x )
We can nd a similar expression for ------------------- using the trigonometric identities:
sin ( x )
2 sin ( 3x ) 2
sin ( 3x ) = sin ( x )(4 cos x 1 ) and ------------------- = 4 cos x 1 .
sin ( x )
cos ( 3 x ) 2 1
-------------------- = 4 sin x + 1 = --- .
cos ( x ) 3
2 2
As sin x = 1 cos x , we can substitute to receive
1 2
--- = 4 ( 1 cos x) + 1
3
1
2
--- = 4 cos x 3
3
7 2 sin ( 3 x )
or --- = 4 cos x 1 , which is our expression for ------------------- .

3 sin ( x )
sin ( 3x ) 7
So, ------------------- = --- .
sin ( x ) 3
Page 5
Solution 2 by Victor Li (11/CA):
sin ( 3 x ) cos ( 3x ) sin ( 3x ) cos ( x ) cos ( 3 x ) sin ( x )
------------------- -------------------- = -----------------------------------------------------------------------------
sin ( x ) cos ( x ) sin ( x ) cos ( x
)
sin ( 2 x
)
= -------------------------------
sin ( x ) cos ( x )
sin ( 2x )
= 2 ------------------
sin ( 2x )
= 2.
So,
sin ( 3 x ) 1
------------------- --- = 2
sin ( x ) 3
or
sin ( 3 x ) 1 7
------------------- = 2 + --- = ---
sin ( x ) 3 3
Editors Comment: We are indebted to Dr. Bruce Reznik of the University of Illinois for the
idea leading to this problem. By the way, the value of x is near, but not equal to 24 .
4/1/13. The projective plane of order three consists of 13 points and 13 lines. These lines
are not Euclidean straight lines; instead they are sets of four points with the properties that
each pair of lines has exactly one point in common, and each pair of points has exactly one
line that contains both points. Suppose the points are labeled 1 through 13, and six of the lines
are A = {1, 2, 4, 8} , B = {1, 3, 5, 9} , C = {2, 3, 6, 10} , D = {4, 5, 10, 11} ,
E = {4, 6, 9, 12} , and F = {5, 6, 8, 13} . What is the line that contains 7 and 8?
Solution 1 by Bob Hough (11/MI): Observe that if there are 13 lines, 13 points, 4 points per line
where no point can occur in a given line more than once, then each point must appear in 4 distinct
lines. Since no point may appear with another point in more than one line, any pairing of two
points on a given line is unique to that line. Therefore, if three of the lines containing a given
point are known, then the points that constitute the fourth line containing that point are also
known. There are three points that occur three times in lines A - F . They are 4, 5, and 6. There
fore, it is possible to decide that lines {3,4,7,13}, {2,5,7,12}, and {1,6,7,11} are in the projective
plane. Now given that three lines are known that contain 7, it is possible to determine that the
fourth line containing both 7 and 8 is {7,8,9,10}.
Page 6
Solution 2 by Scott Wilbur (11/MA): Here are the lines we know:
A: 1 2 4 8
B: 1 3 5 9
C: 2 3 6 10
D: 4 5 10 11
E: 4 6 9 12
F: 5 6 8 13
Because each set of two points denes a line, and:
4 has not yet been used with 3, 7, or 13;
5 has not yet been used with 2, 7, or 12; and
6 has not yet been used with 1, 7, or 11;
we know that these lines are also a part of the plane:
G: 4 3 7 13
H: 5 2 7 12
I: 6 1 7 11
Because 7 has not yet been used with 8, 9, or 10, we can add this line:
J: 7 8 9 10
This is the line we are looking for:

{7, 8, 9, 10} .
Editors comment: This problem was proposed by Dr. Erin Schram of the National Security
Agency. We are most thankful for his continuing support of the USAMTS.
5/1/13. In PQR , QR < PR < PQ so that the

Q T
exterior angle bisector through P intersects ray
QR at point S, and the exterior angle bisector P
R
at R intersects ray PQ at point T, as shown on
the right. Given that PR = PS = RT , deter
mine, with proof, the measure of PRQ .
Solution 1 by Elena Ruse (10/NY): S

(Note: All angles are measured in degrees.)
Let X be a point on PR with R between P and X. Let m(QRX ) = . Notice that we need to

nd 180 . Since TRX is exterior to triangle TPR, --- = m(TRX ) = 2m(TPR) ,
2
180 ---
4
because TPR is isosceles. So m(TPR) = --- . Then m(RPS) = ----------------------- = 90 --- . Now
4 2 8
Page 7

we know all three angles of isosceles triangle PSR: 90 --- , , and (because
8
15
PSR = PRS = QRS ). So 90 --- + + = 180 ; --------- = 90 ; = 48 . So
8 8
180 = 132 .
So the measure of PRQ is 132 degrees.
Solution 2 by Kristin Berger (11/PA): As shown
on the right, each essential angle measure is

Q T
c
assigned a variable.
e
P c d e
R
a b
a
b
The following system of ve linear equations in
ve unknowns is obtained:
S
2a + c = 180 addition property for s
2b +
a = 180 Isosceles Thm., addition property
b +
d = 180 Supplementary s
b 2e = 0 Vertical s are equal
2c + d + e = 180 Isosceles Thm., addition property for s
This system of equations forms the following matrix equation:

2 0 1 0 0 a 180
1 2 0 0 0 b 180
0 1 0 1 0 c = 180
0 1 0 0 2 d 0
0 0 2 1 1 e 180
Solving this matrix equation yields:
Page 8
8 3 4 2
4
------ ------ ------ ------ ------
13 13 13 13 13
1 4 8 2 1 2
a 20 1 0 0 180 ------ ------ ------ ------ ------ 180 84
13 13 13 13 13
b 12 0 0 0 180 180 48
3 6 8 4 8
c = 0 1 0 1 0 180 = ------ ------ ------ ------ ------ 180 = 12
13 13 13 13 13
d 01 0 0 2 0 0 132
4 8 15 1 2
e 00 2 1 1 180 ------ ------ ------ ------ ------ 180
24
13 13 13 13 13
2 4 1 7
1
------ ------ ------ ------ ------
13 13 13 13 13
Each of the end values corresponds to an angle value. Notice there has been no loss of accuracy
since each expression is exact.
d corresponds to PRQ , so m(PRQ) = 132 .
Editors comments: This triangle is known as the Bottema triangle. To learn more about it, see
I. F. Sharygins article on The Steiner-Lehmus Theorem in the November-December 1998 issue
of Quantum magazine. Unfortunately, Quantum is no longer being published. This problem was
proposed by Dr. Berzsenyi.
Page 9

1/2/13. How many positive ve-digit integers are there consisting of the digits 1, 2, 3, 4, 5, 6, 7,
8, 9, in which one digit appears once and two digits appear twice? For example, 41174 is one
such number, while 75355 is not.
Solution 1 by Diana Davis (11/NH):

The rst step is to choose one digit to appear once: 9C1 = 9 .
Then choose the two digits that appear twice: 8C2 = 28 .
This yields the same result as choosing the two digits rst: 9C2 =
36 ,
and then choosing the digit to appear once: 7C1 =
7 ,
because 9 28 = 36 7 = 252 .
The second step is to nd the number of arrangements of ve numbers (a, b, b, c, c) .

5! (number of digits)
This is --------------- = ------------------------------------------------------------- = 30 .
2!2!1! (redundant arrangements)
Then we multiply these numbers to nd the total possibilities:
252 30 = 7560 positive ve digit integers with this property.
9
Solution 2 by Elana Ruse (10/NY): There are C 2 ways to choose the two digits that appear
9 5!
twice, and then 7 ways to choose the remaining digit. Then there are C 2 7 ---------- = 7560 such
2!2!
5!
numbers, because there are ---------- distinct sequences of the 5 objects A, A, B, B, C, as in each
2!2!
sequence, the 2 As or the 2 Bs can be permuted among their positions, though the sequence really
remains the same, in 2! ways.
Solution 3 by Simon Rubenstein-Salzedo (11/CA): First, we should count numbers of the form
aabbc , where a, b, and c are distinct and from 1 to 9. There are 7 of these, since a and b
9
2
5!
are interchangeable, and we do not want to over count by a factor of two. Next, there are ----------
2!2!
ways of arranging aabbc into different orders. Multiplying these, we get the number of numbers
Page 1
in the form of the problem to be 7 ---------- = 7560 distinct numbers.
9 5!
2 2!2!
Editors Comment: This problem was created by our Problem Editor, Dr. George Berzsenyi.
2/2/13. Determine, with proof, the positive integer whose square is exactly equal to the number
2001
(4i 2 )
3
1+ .
i=1
[A computer solution will be worth at most 1 point.].
Solution by Jim Castelaz (12/TN): Recall that for summations from i = 1 to n:

1 = n
i = ( n )(n + 1 ) 2
i = ( n
)(n + 1 )(2n 1 ) 6
i = [(n ( n + 1 )) 2]
3 2
Let all the following summations be from i = 1 to 2001

1 + (4i 2 ) =
3
= 1 + (64i 96i + 48i 8 )
3 2
= 1 + 64 i 96 i + 48 i 8 1
3 2
2001 2002 2 2001 2002 4001 2001 2002

= 1 + 64 --------------------------- 96 --------------------------------------------- + 48 ---------------------------- (8 2001)
2 6 2
= 256512352096009
12
(256512352096009) = 16016003 .
Editors Comment: We are thankful to Dr. Jeremy Dover of NSA for proposing this nice prob
lem.
Page 2
3/2/13. Factor the expression
2 2 2 2
30(a + b + c + d ) + 68ab 75ac 156ad 61bc 100bd + 87cd .
Solution 1 by Devin Averett (12/UT):

The factorization holds for all values of a, b, c, and d, so we can eliminate two variables at a time
by setting them equal to zero, and then factor the resulting expression.
2 2
a & b: 30a + 68ab + 30b
= (5a + 3b)(6a + 10b) OR * (3a + 5b)(10a + 6b) (3a + 5b) (10a + 6b)

2 2
a & c: 30a 75ac + 30c
= (6a 3c)(5a 10c) OR * (3a 6c)(10a 5c) -6c -5c

2 2
a & d: 30a 156ad + 30d
= (10a 2d)(3a 15d) * OR (2a 10d)(15a + 3 d) -15d -2d
In each pair of alternate factorizations above, one of the alternates uses 3a and 10a. These are
marked with *.
The resulting factorization is

(3a + 5b 6c 15d)(10a + 6b 5c 2d)
Solution 2 by Nataliya Ostrovskaya (11/NY): Assume the given expression factors as

( Aa + Bb + Cc + Dd)(Ea + Fb + Gc + Hd) .
30 30 30 30
Since the coefcients of a2, b2, c2 and d2 are all 30, E = ------ , F = ------ , G = ------ , and H = ------ .
A B C D
The expression becomes
( Aa + Bb + Cc + Dd) ------ a + ------ b + ------ c + ------ d .

30 30 30 30
A B C D
Since we get the ab term by multiplying the a term from the rst factor with the b term of the sec
ond, or the b term of the rst by the a term of the second, the coefcient of ab in the expansion is
A B A 2 A 5 3
30--- + 30--- = 68 . Let r = --- . Then 30r 68r + 30 = 0 , so --- = --- or --- . Similarly, we get
B A B B 3 5
A 1 A 1 B 6 5 B 1 C 2 5
---- = 2 or ------ ; ---- = ------ or 5 ; ---- = ------ or ------ ; ---- = 3 or ------ ; ---- = --- or --- .
C 2 D 5 C 5 6 D 3 D 5 2
Now,
2 or -----
1
- = ---- = --- ---- = --- or ------- ------ or ------
A A B 5 3 6 5
2 C B C 3 5 5 6
Page 3
implies, by inspection, that ---, ---- = ---,
------ or ---, ------ .
A B 3 5 5 6
B C 5 6

3
5
Similarly, since
--5- or ------
3
- = --- = ---- ---- = ------ or ------ 2 or ------
B B C 6 5 1
3 5 A C A 5 6 2
we must have
----, ---- = ------, ------ or ---, 2 .
B C 6 1 5

C A 5 2
6
Finally,
--2- or ------
5
- = ---- = ---- ---- = 2 or ------ ------ or -5
C C A 1 1
5 2 D A D 2 5
implies that ----, ---- =

2, ------ or ------, 5 .
C A 1 1
A D 5
2
If --- = --- , then ---, ----, ----, ---- = ---, ------,

2, ------ , and so ( A, B, C, D ) = A
, -------, 2 A, 5 A . It
A 3 A B C A 3 5 1 5A
B 5 B
C
A
D
5 6 5 3
is easy to check that all the pair-wise ratios of these numbers are indeed the ratios we found
above.
Letting A
= 3 , we get that the expression factors as
(3a + 5b 6c 15d)(10a + 6b 5c 2d) .
Editors Comment: We are indebted to Professor Harold Reiter of the University of North
Carolina in Charlotte, NC, for this interesting problem. Dr. Reiter is a staunch friend and sup
porter of the USAMTS program.
4/2/13. Let X
=
(
x 1, x 2, x 3, x 4, x 5, x 6, x 7, x 8, x 9 ) be a 9-long vector of integers. Determine X
if the following seven vectors were all obtained from X by deleting three of its components:
Y
1 = ( 0, 0, 0, 1, 0, 1 ) ,
Y
2 = ( 0, 0, 1, 1, 1, 0 ) ,
Y
3 = ( 0, 1, 0, 1, 0, 1 ) ,
Y
4 = ( 1, 0, 0, 0, 1, 1 ),
Y
5 = ( 1, 0, 1, 1, 1, 1 ) , Y
6 =
( 1, 1, 1, 1, 0, 1 )
, Y
7 = ( 1, 1, 0, 1, 1, 0 ).
Solution 1 by Jeffrey Kuan (8/IL): From Y1, we know that at least four components of X are 0.
From Y5 and Y6 we know that at least ve components of X are 1. So exactly ve components are
1 and exactly four components are 0. Looking at Y1, one can deduce that x 9 = 1 . Why? Well, at
least one 1 occurs after the fourth (and last) 0. Observing Y6, at least one 0 occurs between the
last and second-to-last 1. So, x 8 = 0 . Again, looking at Y1, at least one 1 occurs between the
Page 4
third and fourth 0. Since x8 is the fourth 0, x 7 0 . x7 must equal 1. Continuing in a similar
manner you nd that
the answer is X = (1, 0, 0, 1, 0, 1, 1, 0, 1) .
Solution 2 by Tamara Broderick (11/OH): X = ( x 1, x 2, x 3, x 4, x 5, x 6, x 7, x 8, x 9 ) .
In Y5, ve 1 terms occur, and in Y1, four 0 terms occur. Since these terms must be in the original
vector X, vector X must have four 0 terms and ve 1 terms.
Then, since Y1 already contains all the zero terms, no 0 terms have been deleted from it. Because
its last term is 1, no 0 terms could have occurred in X after this 1, and x 9 = 1 . Y6 already con
tains all of the 1 terms (i.e. none have been omitted), and the last 1 shown is x9, so x 8 = 0 .
Again Y1 is not missing a 0 so x8 and x9 are represented by the nal 0 and 1, and x7 must be a 1.
Next, looking at Y4, we know that the two nal digits of X are not both 1; and 0 must therefore
come between or after the last two 1s in Y4. With this 0, all the 0 terms of X are represented and
we know x 1 = 1 . Likewise, all of the 1 terms are shown in Y5 so x 2 = 0 . In Y2 the 1s on both
ends are missing. When supplied, the total number of 1s is represented and x3 must equal 0.
Finally, Y7 is obviously missing the nal 1 and the 0s of the x2 and x3 terms. Substituting these
in, we have the whole 9-term vector
X = (1, 0, 0, 1, 0, 1, 1, 0, 1) .
Editors comment: This problem was proposed by Dr. Gene Berg of NSA on the basis of an
article by Vladimer I. Levenshtein, published in the January 2001 issue of the IEEE Transactions
of Information Theory. The article answers several questions such as, What is the minimum
number m of distinct binary vectors Y of length n k from which one can reconstruct any n
dimensional vector?
Page 5
5/2/13. Let R and S be points on the sides BC and AC , respectively, of ABC , and let P be the
intersection of AR and BS . Determine the area of ABC if the areas of APS , APB , and
BPR are 5, 6, and 7, respectively.
Solution by Connie Yee (10/NY): When two Not to scale.

adjacent triangles share a height, the ratio of A
their areas is equal to the ratio of their bases.
For ASP and ABP , whose common height 5
is AM S
SP 5 P M
------- = --- . 6 N
PB 6 x y
For SPR and BPR , whose common height 7
is RN B R C
SP x
------- = --- .
PB 7
Therefore x = 35 6 .
BR 7+x
Similarly, for BSR and RSC , -------- = -----------
RC y
BR 7+x
and for ABR and ACR , -------- = ------------ .
RC y
7+x 13 (7 + x)(5 + x) 5005
Therefore ------------ = --------------------- , and y = ---------------------------------- = ------------ .

y 5+x+y 6x 6
35 5005
So Area(ABC) = 5 + 6 + 7 + x + y = 5 + 6 + 7 + ------ + -----------
6 6
= 858 .
Editors comments: We thank former USAMTS participant Surat Intasang for proposing this
problem, inspired by Problem 6 of the 1985 American Invitational Mathematics Examination.
We hereby wish Surat success with his efforts to transplant the USAMTS to his native Thailand.
Page 6

Erin Schram, Grader
1/3/13. We will say that a rearrangement of the letters of a word has no xed letters if, when the
rearrangement is placed directly below the word, no column has the same letter repeated. For
instance, the blocks of letters below show that E S A R E T is a rearrangement with no xed
letters of T E R E S A, but R E A S T E is not.
T E R E S A T E R E S A
E S A R E T R E A S T E
How many distinguishable rearrangements with no xed letters does T E R E S A have? (The
two Es are considered identical.)
Comment: Our Problem Editor, Professor George Berzsenyi, suggested this problem.
Solution 1 for 1/3/13 by Dmitriy Yegoshin (12/CA):

First lets position our Es since they will present the biggest problem. There are only 6 valid
arrangements of the 2 Es where both of them are in different positions than in TERESA. They are:
1 E * E * * *
2 E * * *E *
3 E * * * * E
4 * * E * E *
5 * * E * * E
6 * * * * E E
Now that we have the Es positioned, we can go ahead and position the other 4 letters. But
since the other 4 letters are not repeated in TERESA, the number of different placements for those
4 letters is the same for each of the 6 combinations of Es.
Lets count all the possible combinations for one of the Es combinations:
T E R E S A T E R E S A T E R E S A
E T E R A S E R E A T S E A E T R S
E T E S A R E S E T A R E A E R T S
E T E A R S E S E R A T E A E S T R
E R E T A S E S E A T R E A E S R T
E R E S A T E S E A R T
We get 14 different combinations.

So now since each of the 6 Es combinations has 14 sub-combinations, we multiply
6 14 = 84 .
Therefore, there are a total of 84 distinguishable arrangements.
Solution 2 for 1/3/13 by Elena Ruse (10/NY):
To get a suitable rearrangement, we must move each of the Es to one of the positions 1, 3, 5,
2
or 6. There are C 4 = 6 ways to do that. Say we move the two Es to where the letters X, Y, which
are two of the letters in {T, R, S, A}, were originally. Our problem is now to place the other two
letters, which well call Z and W, in 2 of the remaining 4 positions so that neither is in its original
position. (the Es are now where X and Y were originally, so, of course, neither X nor Y will
remain xed no matter where we put them). Now, we can either move Z to where W was, in which
case we can move W to any of the 3 remaining positions, or move Z to the original positions of
one of the Es (which we can do 2 ways), in which case we must move W either to where Z was or
where the other E was. So, once weve placed the Es (which we can do in 6 ways), there are
3 + 2 2 = 7 ways to place Z and W, and then 2 ways to place X and Y in the remaining 2 posi
tions. The total number of suitable rearrangements is, therefore, 6 7 2 = 84 .
Solution 3 for 1/3/13 by Dheera Venkatraman (12/NJ):

There are 6! = 720 ways of arranging six letters. Since two are identical (there are two Es),
there are 360 distinct arrangements. Of these:
One-sixth (360/6 = 60) will have T as the rst letter and, therefore, have a xed letter. These
are immediately excluded, leaving 300 arrangements.
One-sixth of the original 360 arrangements will have R as the third letter. Of this one-sixth,
one-fth will also have T as the rst letter, which has already been counted. This creates
60 60 5 = 48 rearrangements which can be eliminated. This leaves 300 48 = 252
arrangements.
One-sixth of the original 360 will have S as the fth letter. Following similar logic: one-fth of
these will have R as the third letter, which has been counted; another one-fth will have T as
the rst letter; but one-fourth of each of those will have R as the third letter and T as the rst
letter, and this intersection of the two sets (the set with R as the third letter and the set with T
as the rst letter) must not be subtracted twice. This creates 60 (2 (60 5) (60 5) 4) =
39 arrangements which can be eliminated, leaving a total of 213 arrangements.
Following this pattern for those with the sixth letter being A,
60 (3 (60 5) (3 ((60 5) 4) ((60 5) 4) 3)) = 32 arrangements which are
eliminated, leaving 181 arrangements.
To eliminate the occurrences of the Es, a slightly different pattern must be used. Here,
because there are two Es, one-third (360 3 = 120 ) of the original sequence bear E as the
second letter. Of those, one-fth will have A as the sixth letter, one-fth will have S as the fth
letter, one-fth will have R as the third letter, and one-fth will have T as the rst letter. Also,
of the one-fth that have A as the sixth letter, one-fourth of those will also have S as the fth
letter, and so on. This gives
120 (4 (120 5) (6 ((120 5) 4) (4 (((120 5) 4) 3) (((120 5) 4) 3) 2)))
= 53 arrangements with E as the second letter that have not been eliminated yet. Likewise,
there are 53 arrangements with E as the fourth letter that have not been eliminated yet. This
makes 53 + 53 = 106 ; but that counts arrangements with Es as the second letter and fourth
letter twice. Of the original 360, one-fth of one-third, 24 arrangements, have Es in both
places. Of those, one-fourth will have A as the sixth letter (since two letters are in place
already, there are only four places of the A), one-fourth will have S as the fth letter, and so
on. This gives
24 (4 (24 4) (6 ((24 4) 3) (4 (((24 4) 3) 2) (((24 4) 3) 2) 1))) = 9
arrangements not to subtract twice. 53 + 53 9 = 97 , leaving 84 arrangements.
There are, therefore, 84 permutations of the letters TERESA that contain no xed letters.
An interesting note: Note the coefcients of each term of the expressions of the number of
permutations eliminated in each step of the sequence for letters of a single occurrence:
[ 1 ] 60 eliminations for T
[ 1 ] 60 [ 1 ] (60 5) for R
[ 1 ] 60 ([ 2 ] (60 5) [ 1 ] (60 5 4)) for S
[ 1 ] 60 ([ 3 ] (60 5) ([ 3 ] (60 5 4) [ 1 ] (60 5 4 3))) for A
Note the coefcients which follow Pascals triangle. To illustrate why this occurs, not the deriva
tion of the coefcients of the last step shown here for A: One must rst count the occurrences of A
itself in its xed position; then the number of times each of the other letters appears xed by itself
(i.e., T, R, or S), followed by the number of times each group of two letters appears xed (i.e., after
subtracting the appearances of T, R, and S, you need to make sure you do not subtract the occur
rences of T and R, R and S, or S and T twice), followed by each group of three letters, and nally
the single group of all four letters.
2/3/13. Without computer assistance, nd ve different sets of three positive integers {k, m, n}
1 1 1 19
such that k < m < n and --- + ---- + --- = ------ .
k m n 84
[To think about, but not a part of the problem: How many solutions are there?]
Comment: Our Problem Editor, Professor George Berzsenyi, suggested this problem too. There
are 22 different sets of {k, m, n}. The second solution below lists them all.
Solution 1 for 2/3/13 by Boris Alexeev (10/GA):

Answer: (5, 60, 105), (6, 21, 84), (6, 28, 42), (7, 14, 84), (7, 21, 28)
We will attempt to nd only ve sets, not all of them. Our rst step will be to look for solu
tions of the form {k, m, n} where k, m, and n are divisors of 84, the denominator of the righthand
side. So we let a = 84 k , b = 84 m , and c = 84 n , and multiply both sides of the equation
1 k + 1 m + 1 n = 19 84 by 84 to get a + b + c = 19 and a > b > c . Since k is a factor of 84,
so is a. Likewise for b and c. {a, b, c} {1, 2, 3, 4, 6, 7, 12, 14, 21, 28, 42, 84} .
Its not hard to see that a is either 7, 12, or 14. If a is 7, we must nd b > c with b + c = 12 ,
but this is impossible. If a is 12, we must nd b > c with b + c = 7 . There are two such pairs:
b = 6 , c = 1 and b = 4 , c = 3 . These yield k = 7 , m = 14 , n = 84 and k = 7 , m = 21 ,
n = 28 , respectively. If a is 14, we must nd b > c with b + c = 5 . There are two such pairs,
b = 4 , c = 1 and b = 3 , c = 2 . These yield k = 6 , m = 21 , n = 84 and k = 6 , m = 28 ,
n = 42 , respectively.
Now, that is only four sets, to nd a fth, we notice that we already have solutions with k = 6
and k = 7 , so we arbitrarily try k = 5 . We then get the equation 1 m + 1 n = 11 420 . We
use the same trick as before, multiplying by 420 this time, to get d + e = 11 , where d and e are
factors of 420. It doesnt take much work at all to nd one possible solution: d = 7 , e = 4 ,
yielding a fth solution of k = 5 , m = 60 , n = 105 .
Note that there are, in fact, many more than ve solutions: there are 22.
Solution 2 for 2/3/13 by George Lin (10/NJ):

1 1 1
Since 0 < k < m < n , --- < ---- < --- , and therefore,
n m k
1 19 1 1 1 3
--- < ------ = --- + ---- + --- < --
k 84 k m n k
From this we get,
84 3 84
------ < k < ------------- .
19 19
Since k is an integer, this simplies to 5 k 13 .
19 1 1
Notice that ------ = --- + ------ . Let us consider k = 7 . Here, we need only solve m and n such that
84 7 12
1 1 1
---- + --- = ------ . (1)

m n 12
From
1 1 1 1 2
---- < ------ = ---- + --- < ---
m 12 m n m
we have 12 < m < 24 . Since m is an integer, this implies that 13 m 23 . Solving equation (1)
for n, we get
12m
n = ---------------- .
m 12
Now, let us nd n for m = 13 to 23, and see which values produce integers.
m 13 14 15 16 17 18 18 20 21 22 23
n 156 84 69 48 204
---------
36 228
---------
30 28 132
---------
276
---------
5 7 5 11
Integer? Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Using the same method for other values of k, we can nd all the solutions. They are:
k m n k m n k m n k m n
5 39 1820 6 17 1428 7 13 156 8 10 840
5 40 840 6 18 252 7 14 84 8 12 56
5 42 420 6 20 105 7 15 60
5 45 252 6 21 84 7 16 48
5 56 120 6 24 56 7 18 36
5 60 105 6 28 42 7 20 30
5 70 84 7 21 28
Therefore, there are 22 integer solutions total.
n n1
3/3/13. Suppose p ( x ) = x + a n 1 x + + a 1 x + a 0 is a monic polynomial with integer
n 2
coefcients. [Here monic polynomial just means the coefcient of x is one.] If ( p ( x )) is a
polynomial all of whose coefcients are non-negative, is it necessarily true that all the coef
cients of p ( x ) must be non-negative? Justify your answer.
Comment: This problem was created by professor Bruce Reznick of the University of Illinois and
was rst featured in the 1990 Indiana College Mathematics Competition.
Solution 1 for 3/3/13 by Andrew Newman (11/AL):

No. A counterexample sufces:
4 3 2 2
( x + 2x x + 3 x + 4) = x 8 + 4 x 7 + 2x 6 + 2x 5 + 21x 4 + 10 x 3 + x 2 + 24 x + 16 .
The conjecture that non-negative integer coefcients in ( p ( x )) 2 implies non-negative coefcients
in p ( x ) is true for all rst, second, and third degree p ( x ) , but it is not true in general for any
higher degree. We will consider the cases of degrees 1, 2, 3, and 4, and then give a construction
for higher degrees.
Case 1: p( x) = x + a
( p ( x )) 2 = x 2 + 2ax + a 2
2a 0 implies a 0 , so the conjecture is true for rst degree p ( x ) .
Case 2: p ( x ) = x 2 + ax + b
( p ( x )) 2 = x 4 + 2ax 3 + (a 2 + 2b) x 2 + 2abx + b 2
2a 0 implies a 0 . If a > 0 , then 2ab 0 implies b 0 . If a = 0 , then a 2 + 2b 0 implies
b 0 . So the conjecture is true for second degree p ( x ) ) .
Case 3: p ( x ) = x 3 + ax 2 + bx + c
( p ( x ))
2 = x 6 + 2ax 5 + (a 2 + 2b) x 4 + (2ab + 2c) x 3 + (b 2 + 2ac)x 2 + 2bcx + c 2
2a 0 implies a 0 . Since 2bc 0 , either b and c are both positive, they are both negative, or at
least one is zero. If they are both negative, then 2ab + 2c would be negative, a contradiction. If
a > 0 and b or c is zero, then 2ab + 2c 0 makes the other one non-negative. If a = 0 , we use
a 2 + 2b 0 and 2ab + 2c 0 instead. Thus, both b and c are non-negative, and the conjecture is
true for third degree p ( x ) .
Case 4: p ( x ) = x 4 + ax 3 + bx 2 + cx + d
( p ( x ))
2 = x 8 + 2ax 7 + (a 2 + 2b) x 6 + (2ab + 2c) x 5 + (b 2 + 2ac + 2d)x 4
+ (2ad + 2bc)x 3 + (c 2 + 2bd) x 2 + 2cdx + d 2
As before, a 0 . Since 2cd 0 , if neither of c and d is zero, they have the same sign. Suppose c
and d are positive and let b = 1 . Since a 2 + 2b 0 , we have a 2 2 . Let a = 2 . From
2ab + 2c 0 we have c 2 . Let c = 3 . There are four conditions on d. The rst is
b 2 + 2ad + 2d 0 , which is satised since we assumed d is positive. The second is
2ad + 2bc 0 , which is equivalent to d 1.5 . The third is c 2 + 2bd 0 , or d 4.5 . The fourth
is 2cd 0 , which we made sure was satised when we supposed that c and d are positive. Thus,
any integer value for d in [1.5, 4.5], namely 2, 3, or 4, provides a counterexample where all the
coefcients of ( p ( x )) 2 will be non-negative but p ( x ) will have a negative coefcient, namely
b = 1 . The counterexample with d = 4 appears at the beginning of this solution.
For any degree n greater than 4, a counterexample to the conjecture can be obtained by multi
plying the above counterexample by x n 4 .
It should also be noted that the conjecture is true for a polynomial of any degree that has only
real roots, no imaginary or non-real complex roots. Squaring p ( x ) introduces no new unique
roots. Since ( p ( x )) 2 has all non-negative coefcients, it follows by Descartes Rule of Signs that
it has no positive roots., and since all its roots are real, it must have only negative and zero roots.
p ( x ) has the same roots (which each root occurring with half the multiplicity) and so may be
written as p ( x ) = ( x + r 1 )( x + r 2 )( x + r n ) with each r i being positive or zero, so that when
expanded p ( x ) must have all non-negative coefcients.
4/3/13. As shown in the gure on
C = (30s, 30t)
the right, in ACF , B is the
midpoint of AC , D and E divide D
side CF into three equal parts, E
while G, H, and I divide side
F = (0, 0)
A = (60r, 0)
FA into four equal parts. G H I
Seventeen segments are
drawn to connect these six
points to one another and to the opposite vertices of the triangle. Determine the points interior
to ACF at which three or more of these line segments intersect one another.
To make grading easier, we have embedded the triangle into the rst quadrant with point F
at the origin, point C at (30s, 30t) , and point A at (60r, 0) , where r, s, and t are arbitrary
positive real numbers. Please use this notation in your solutions.
Comment: An article about M.C. Eschers work in the April 1998 issue of Pythagoras, an excel
lent journal for high school students in the Netherlands, inspired Professor Berzsenyi to propose
this problem. Escher divided the sides of the triangle into 3, 4, and 5 parts and discovered that
there are 17 such triple points of intersection within that triangle.
A visual solution by drawing all the segments and looking for triple intersections is a valid
proof that there are at most two triple intersections, but the two triple intersections have to be ver
ied, since each seeming triple intersection might instead be three separate intersections very
close to each other.
Solution 1 for 4/3/13 by William Groboski (11/WI):

The rst thing is to draw the triangle and see where any three points nearly cross.
The points are circled.
B
E
F A
G H I
We can see there are two possible points, where GC , EH , and FB cross and where HB , IC , and
AD cross.
The next step is to nd where the points are:
A = (60r, 0) D = (20s, 20t) G = (15r, 0)
B = (15s + 30r, 15t) E = (10s, 10t) H = (30r, 0)
C = (30s, 30t) F = (0, 0) I = (45r, 0)
The next step is to dene the lines, this is easy. First you nd the slope, and then use the point-
slope form of a line. The equations are:
GC y = -------------- x --------------
2t 30rt
2s r 2s r
EH y = -------------- x --------------
t 30rt
s 3r s 3r
FB y = -------------- x
t
s + 2r
We will set two right sides equal to each other and solve for x and then input that x to nd y. We
should do this for each pair to nd their crossing.
It is found that they all cross at (12r + 6s, 6t) .
We do the same for the other three lines:
HB y = - x ----------
t 30rt
s s
IC y = ----------------- x -----------------
2t 90rt
2s 3r 2s 3r
AD y = -------------- x --------------
t 60rt
s 3r s 3r
We nd where they all intersect and they also all intersect at one point. They all meet at
(30r + 10s, 10t) .
The points in the interior of the triangle were three or more segments intersect one another are
(12r + 6s, 6t) and (30r + 10s, 10t) .
Solution 2 for 4/3/13 by Dimitri Pavlichin (11/PA)

There are 17 segments that divide triangle ABC. Of these 17, segments EG , DB , and IB only
cross segments that all intersect at a vertex, thus eliminating them from any possible intersection
of at least 3 segments in the interior. Now there are only 14 segments to examine.
A list of possible two-segment intersections was made. Parallel segments, segments that can
not intersect inside the triangle (such as DI and EH ) and segments that share a common starting
point (such as EH and EI ) were not included.
AD BE AE BG BE CG BF CI BG DH CG DI
AD BF AE BH BE CH BF DG BG DI CG EH
AD BG AE CG BE CI BF DH BG EH CG EI
AD BH AE CH BE DG BF DI BG EI CH DI
AD CG AE CI BE DH BF EH BH CI CH EI
AD CH AE DG BE DI BF EI BH DI DG EH
AD CI AE DH BF CG BG CH BH EI DG EI
AE BF AE DI BF CH BG CI CG DH DH EI
The coordinates of each of the points on the sides of the triangle were derived. For example,
the coordinate of point B was derived by the midpoint formula:
X mid = ( X 1 + X 2 ) 2 Y mid = (Y 1 + Y 2 ) 2.
The x-values of the coordinates of point G, H, and I were derived by multiplying the x-value
of point A, which was 60r, by 1/4, 1/2, and 3/4, respectively, since those points divided segment
AF into four segments of equal length.
The x- and y-values of the coordinates of point D and E were derived by multiplying the x- and
y-values of point C, which were (30s, 30t ) , by 1/3 and 2/3. respectively, since these points
divided segment CF into segments of equal length, and if perpendicular lines were drawn from
points C, D, and E onto the x- and y-axes, each axis would also have been divided into segments
of equal length.
Thus, the coordinates are as listed: B = (15s + 30r, 15t ) , D = (20s, 20t ) , E = (10s, 10t ) ,
G = (15r, 0) , H = (30r, 0) , I = (45r, 0) .
The equations of the lines that contain these line segments were derived. Only 14 equations
were written, since 3 segments were excluded as described above.
AD tx + (3r s)y = 60rt CH tx + (r s)y = 30rt
AE tx + (6r s)y = 60rt CI 2tx + (3r 2s) y = 90rt
BE tx (6r + s) y = 60rt DG 4tx + (3r 4s)y = 60rt
BF tx (2r + s)y = 0 DH 2tx (3r 2s)y = 60rt
BG tx (r + s) y = 30rt DI 4tx + (9r 4s)y = 180rt
BH tx sy = 30rt EH tx + (3r s)y = 30rt
CG 2tx (r 2s)y = 30rt EI 2tx + (9r 2s)y = 90rt
These equations were used in systems of 3 equations to see whether the 3 segments contained in a
system of 3 equations intersect in a common point. (Note from Erin Schram: though Mr. Pavlichin
does not mention it, checking his list of 48 line segment intersections, I nd 39 cases were 3 line
segments each intersect the other two, though not necessarily at a common point. So he had to
check 39 systems of 3 equations. It would be about as easy to simply nd the 48 points of intersec
tion for all the pairs of line segments and check for matching points.)
The intersection of AD , BH , and CI is a common point: (30r + 10s, 10t) .
The intersection of BF , CG , and EH is a common point: (12r + 6s, 6t) .
No other points of triple intersection were found.
5/3/13. Two perpendicular planes intersect a sphere in two circles. These circles intersect in two
points, spaced 14 units apart, measured along the straight line connecting them. If the radii of
the circles are 18 and 25 units, what is the radius of the sphere?
Comment: This interesting geometry problem was originally proposed for the American Invita
tional Mathematics Examination (AIME) by the late Professor Joseph Konhauser, a superb prob
lemist and a great friend of Professor Berzsenyi.
Solution 1 for 5/3/13 by Konstantin Getmanchuk (11/NY)

We have two perpendicular planes intersection a sphere. The intercepted section of a plane
and a sphere is a circle.
18
18
r 18 x
18 x x 7 7
25 25
25
7
25
Viewing the circle of radius 25 as a base, the height, x, of the center of the sphere above that base
is the same as the height of the center of the circle of radius 18 above that base. x = 10 2 7 2 =
2 2
275 . As seen in the diagram, the radius, r, of the sphere is r = x + 25 = 275 + 625 = 30.
Solution 2 for 5/3/13 by Daniel Halpern-Leistner (11/PA)
18
a
n r
b
25
r
This diagram is a cross-section perpendicular to both planes that intersect the sphere and
coplanar with the centers of the two circles of the two circles formed by intersection and with the
center of the sphere itself. First the values of a and b must be determined as functions of r via the
2 2 2 2 2
2
Pythagorean theorem: b = r 25 and a = r 18 . Again by the Pythagorean theorem,
2 2 2
n = a + b . n is also the base of a triangle whose height is half the between the two intersec
tions of the circles and whose hypotenuse is the radius of the sphere, because it connects the cen
2 2
2
ter with a point on the surface. Therefore, the Pythagorean relationship, r = n + 7 , holds.
2 2 2 2 2 2
Substituting one equation into the next: n = r 25 + r 18 = 2r 949 and
2 2
2 2
r = n + 49 = 2r 900 , so r = 900 and r = 30 .
Solution 3 for 5/3/13 by Yan Zhang (11/VA)

We rotate the our coordinate system so that the xy-plane is parallel to the rst plane intersect
ing the sphere, the yz-plane is parallel to the second plane, and the center of the sphere is at
(0, 0, 0)
. Clearly, the intersections are symmetric about the xy-plane, and their z-coordinates must
be opposites of each other. So let their coordinates be ( x, y, 7) and ( x, y, 7 ) .
2 2
2
Now all points on the rst circle satisfy x + z = 18 , and all the points on the second circle
2 2 2 2 2 2 2
2 2
satisfy y + z = 25 . Since ( x, y, 7) is on both circles, x + 7 = 18 and y + 7 = 25 . The
point ( x, y, 7) is on the sphere, whose radius is r, so
2 2 2
2
r = x +y +7
2 2 2 2 2
= (18 7 ) + (25 7 ) + 7
= 900
and the radius is 30.

solutions edited by Erin Schram
1/4/13. In a strange language there are only two letters, a and b, and it is postulated that the let
ter a is a word. Furthermore, all additional words are formed according to the following rules:
A. Given any word, a new word can be formed from it by adding a b at the righthand end.
B. If in any word a sequence aaa appears, a new word can be formed by replacing the aaa
by the letter b.
C. If in any word the sequence bbb appears, a new word can be formed by omitting bbb.
D. Given any word, a new word can be formed by writing down the sequence that consti
tutes the given word twice.
For example, by (D), aa is a word, and by (D) again, aaaa is a word. Hence by (B) ba is a
word, and by (A) bab is also a word. Again, by (A), babb is a word, and so by (D), babbbabb
is also a word. Finally, by (C) we nd that baabb is a word.
Prove that in this language baabaabaa is not a word.
Comment: Our Problem Editor, Professor George Berzsenyi proposed this problem. It is similar
to a problem mentioned in Mathematical Challenges II, published by the Scottish Mathematical
Council in 1995.
Solution 1 for 1/4/13 by Jenna Le (12/MN):

(B) and (D) are the only rules that affect the number of times the letter a appears in your word.
Suppose you have a word in which the latter a appears n times. If you apply rule (B), you will
get a new word in which the letter a appears n 3 times. If you apply rule (D) instead, you will
get a new word in which the letter a appears 2n times.
Unless n is a multiple of three, neither n 3 nor 2n can be a multiple of 3, by Lemmas 1 and
2 below. Therefore, unless you start out with a word in which the number of times the letter a
appears is a multiple of three, you will never be able to form a word in which the letter a appears
a multiple of three times. You start out with the word a, which contains one letter a, and one is not
a multiple of three. Therefore, you will never be able to form the word baabaabaa, since baabaa
baa contains the letter a six times, and six is a multiple of three.
QED
Lemma 1. If 3 divides n 3 , then 3 divides n .
Suppose 3 (n 3 ) . Then there exists an integer x such that n 3 = 3x . If follows that
n = 3x + 3 = 3( x + 1 ) , so 3 n .
Lemma 2. If 3 divides 2n , then 3 divides n .
Suppose 3 ( 2n) . Then the Fundamental Theorem of Arithmetic implies that, since 3 does not
divide 2 and 3 is prime, 3 n .
Solution 2 for 1/4/13 by Mauro Braunstein (12/FL)
We have here an axiomatic system with axiom a and rules of inference A, B, C, and D:
AXIOM: a is a word.
RULES: A. If x is a word, so is xb.
B. If xaaay is a word, so is xby.
C. If xbbby is a word, so is xy.
D. If x is a word, so is xx.
In the above rules, x and y stand for strings of letters, which could be the empty string . Notice
that is not postulated as a word. We will show you that it isnt a word at all, and even its addi
tion to the set of words would still keep baabaabaa from being a word.
We will focus on the number of as for our problem, so let a ( x ) be the number of times a
occurs in x. For example, a ( baabaabaa ) = 6 . Rule A adds a b to a word, so the number of as
clearly remains unchanged. Rule B replaces three as by a b, so the number of as decreases by
three. Rule C simply removes three bs, so it also leaves the number of as unchanged. Rule D
duplicates the word, therefore clearly doubling the number of as. If we let n = a ( x ) ,
a ( A( x )) = n
a ( B( x )) = n3
a ( C( x )) = n
a ( D( x )) = 2n
a ( a ) = 1 , and from there we want to create a case where a ( x ) = 6 . What order of multiplying
by two and subtracting three will take 1 to 6? None. Consider n (mod 3). Subtracting 3 in rule B
will leave n (mod 3) unchanged. Look at doubling. We are given 1; 2 1 2 (mod 3), and
2 2 1 (mod 3), so doubling will yield a repetitive cycle: 1, 2, 1, 2, . However, 6 0 (mod 3),
and no amount of doubling will yield 0 from 1. Since a ( baabaabaa ) = 6 , it is impossible to
construct it from a string m such that a ( m ) = 1 , particularly m = a . Likewise, a ( ) = 0 , so
cannot be constructed either.
If were accepted as an axiom, no rule would allow an increase in its as. A and C keep the
number of as xed at 0, D doubles the number of as from 0 to 0, and B cannot be used because it
needs at least three as, an impossible requirement since we started with zero as. Therefore, the
word baabaabaa would not be possible even under this new system.
2/4/13. Let f ( x ) = x x x x for all positive real numbers x, where y denotes the
greatest integer less than or equal to y.
(1) Determine x so that f ( x ) = 2001 .
(2) Prove that f ( x ) = 2002 has no solution.
Comment: This problem was suggested by Professor Bela Bajnok of Cornell University. It is
adapted from a Chech-Slovakian problem given a few years ago.
Solution 1 to 2/4/13 by David Barmore (10/IL)

First, we will determine x such that f ( x ) = 2001 . For x 1 , f ( x ) is strictly increasing.
Since f (6.9965) = 2000.999 < 2001 and f (6.9966) = 2001.0276 > 2001 , if f ( x ) = 2001
f ( x)
then 6.9965 < x < 6.9966 . Also, since x = ---------------------------- by denition of f ( x ) , x is rational for
x x x
2001
all integral values of f ( x ) . For our problem, x = ---------------------------- . For 6.9965 < x < 6.9966 ,
x x x
x x x
= 286 , so x could only be 2001/286, which checks.
Now, we will prove that no x exists such that f ( x ) = 2002 . First, if

there were some number
2002 ------------ ) = 2001 < 2002 and
x such that f ( x ) = 2002 , then x = ---------------------------- . Clearly, since f ( 2001
286
x x x
f ( 7 ) = 2401 > 2002 and f ( x ) is strictly increasing on this interval, 2001
------------
286
< x < 7. For x on this
2001 41
interval, x = 6, x x = 6x = 41 since ------------
286
> ------ ,
6
x x x = 41x = 286 since
2001
------------
286
> 286
--------- .
41
So x could only be 2002/286 = 7. However, f ( 7 ) 2002 . Since we have arrived at a
contradiction, our supposition that there exists x such that f ( x ) = 2002 must be false. There
fore, there does not exist x such that f ( x ) = 2002 .
Solution 2 to 2/4/13 by Agustya Mehta (10/OH)

Part 1.
It is easy to see that f ( x ) increases as x increases, because x x x either remains the
same or increases when x increases.
4 4
We see that f ( 6 ) = 6 = 1296 and f ( 7 ) = 7 = 2401 , so if f ( x) = x x x x =
2001, then 6 < x < 7 .
Let m = x x x . m is an integer by the denition of . So, f ( x ) = x m = 2001 or
x = 2001 m . Since x < 7 , we have 286 m (Inequality 1). Also since 6 < x < 7 ,
x = 6
x x < 42
x
x 41 (Inequality 2)
x x x < 287
x
x x 286
So by both the inequalities, 286
m 286 ; thus, m = 286 . Thus x = 2001
286 .
Part 2.
4 4
As we have shown in part 1, f ( 6 ) = 6 = 1296 , f ( 7 ) = 7 = 2401 , and 1296 < 2002 <
2401. So if f ( x ) = 2002 has a solution x
, it must satisfy 6 < x < 7 . This is the same condition
that gave us Inequality 2 in part 1: x
x
286 . So x
x
< 2002 . Thus,
f ( x ) = 2002 has no solution.
Solution 3 by Daniel McLaury (11/OK)

x
must be a rational number, since we are proposing to multiply it by an integer and get an
integer result, so let x
=
n + a-
- , where n , a , and b are integers and 0 < a < b . Then we know
b
4
that n =
6 , since if we plug in an
x
-value less than 6, well get a number less than 6 = 1296,
and if x
is 7 or more, we get 7 = 2401 or more.
Start from inside the nested oor brackets. Obviously, x = 6 + a-- = 6 . At the next oor
b
a a a
bracket, we get 36 + ( 6 ) ( --- ) . Since 0 < --- < 1 , we know that 0 < ( 6 ) ( --- ) < 6 , so that it becomes
b b b
an integer between 0 and 5.
a a 5
a
Suppose ( 6 ) ( --- ) is 4, which means --- < -- . We would get 40 for 36 + ( 6 ) ( --- ) and then
b
b 6 b
a
multiplying by x
would give 240 +
( 40 ) ( --- ) . The oor of that has a maximum value of 273, since
b
a 5
--- < --- . Now 2001 is more than 7 times 273, so obviously this is too small. Therefore, by contradic
b 6
a
tion, we know that ( 6 ) ( --- )
= 5
.
b
a
So we have that x x = 41 , which means that x x x = 246 + ( 41 ) ( --- ) . Now 2001
b
a
divided by 7 is just less than 286, so anything smaller than that wont work. Since -- < 1 ,
b
a a
( 41 ) ( --- )
< 41 , so ( 41 ) ( --- ) 40 . But we need 286, so it must be at least 40. So it is 40.
b b
a
So nally, x x x x is simply ( 6 + a-- )(286) =
1716 +
(286) ( --- ) . We can set b = 286 ,
b b
a
and reduce ---
b
later if necessary. We get 1716 + a = 2001 , giving a = 285 . So our nal answer
is 6 + 285
--------- .
286
Letting x be anything between 6 +

285
--------- and 7 would still give f ( x ) = 286x , which could not

286
4
be an integer result. And f ( 7 ) = 7 is much larger than 2002, so it is impossible to get 2002.
I dont like this problem for some reason. There just seems to be way too much brute force
arithmetic. Any problem with an answer like 6 + 285
--------- just seems to be severely contrived and lacks
286
any real beauty.
Comment by Erin Schram on Mr. McLaurys last observation: We select our problems for
solutions of appropriate difculty that require mathematical understanding. We would love to
have beauty in every problem, but given the limited pool of problems that we can beg, borrow, or
invent, sometimes we have to make do without beauty. However, our problems are not contrived
in the sense of deliberately selecting ugly answers to add more difculty. We want the mathemat
ics of the problems to be as bold and clear as possible.
3/4/13. Let f be a function dened on the set of all integers, and assume that it satises the fol
lowing properties:
A. f (0) 0 ;
B. f ( 1 ) = 3 ; and
C. f ( x ) f ( y ) = f ( x + y ) + f ( x y ) for all integers x and y.
Determine f ( 7 ) .
Comment: This problem is based on a similar problem used in 1987 for the selection of Hun-
garys IMO team, and was proposed by Prof. Berzsenyi.
Solution 1 for 3/4/13 by George Khachatryan (11/TX)

f (1) f (0) = f (1) + f (1) 3 f (0) = 3 + 3 f (0) = 2
f (1) f (1) = f (2) + f (0) 3 3 = f (2) + 2 f (2) = 7
f ( 2 ) f ( 1 ) = f ( 3 ) + f ( 1 ) 7 3 = f ( 3 ) + 3 f ( 3 ) = 18
f ( 3 ) f ( 1 ) = f ( 4 ) + f ( 2 ) 18 3 = f ( 4 ) + 7 f ( 4 ) = 47
f ( 4 ) f ( 3 ) = f ( 7 ) + f ( 1 ) 47 18 = f ( 7 ) + 3 f ( 7 ) = 843
so f ( 7 ) = 843 .
However, what is interesting about f ( n )

is that it has a closed form. The sequence we saw
actually is special. Over the positive integers, f ( n 1 ) f ( 1 ) = f ( n ) + f ( n 2 ) , which implies
f ( n 1 ) 3 = f ( n ) + f ( n 2 ) and
f ( n ) = 3 f ( n 1 ) f ( n 2 ).
This is the recursive denition for the even bisection of the Lucas numbers, L ( n ) . The Lucas
number L ( n ) is dened as L ( 0 ) = 2 , L ( 1 ) = 1 , and L ( n ) = L ( n 1 ) + L ( n 2 ) . The
sequence is: 2, 1, 3, 4, 7, 11,18, 29, 47, 76, 123, . (c.f. the Fibonacci numbers).
The even bisection of the Lucas numbers, which are our f ( n ) ,
has a closed form:
1 + 5 2n
1 5 2n
f ( n ) = ---------------- + ---------------- . Setting n = 7 indeed yields 843.
2 2
4/4/13. A certain company has a faulty telephone system that sometimes transposes a pair of
adjacent digits when someone dials a three-digit extension. Hence a call to x318 would ring
at either x318 , x138 , or x381 , while a call received at x044 would be intended for either
x404 or x044 . Rather than replace the system, the company is adding a computer to deduce
which dialed extensions are in error and revert those numbers to their correct form. They have
to leave out several possible extensions for this to work. What is the greatest number of three-
digit extensions the company can assign under this plan?
Comment: This problem was proposed by Dr. Peter Anspach of the National Security Agency.
Solution 1 for 4/4/13 by Tamara Broderick (11/OH)

How many three-digit extensions can be assigned so that the computer can deduce the correct
extension every time?
For extensions of the form aaa:
All possible outputs of the faulty system: aaa
Therefore, every extension of the form aaa (10 total) may continue to be used as it will not be
confused with any other extension.
For extensions of the form aab:

Possible extensions of form aab Possible outputs of the system
aab aab, aba
aba aba, baa, aab
baa baa, aba
Since there are no two extensions that have no shared possible outputs, all of these extensions
could be confused with the others, and only one can, thus, be used by the computer system. There
fore, 1/3 of all extensions of this form can be assigned.
For extensions of the form abc:

Possible extensions of form abc Possible outputs of the system
abc abc, bac, acb
acb acb, cab, abc
bac bac, abc, bca
bca bca, cba, bac
cab cab, acb, cba
cba cba, bca, cab
Each extension can be confused with two other extensions. Therefore, the maximum number of
extensions that may be veried as unique by the new computer out of 6 such extensions is 2. For
example, abc and cba share no possible outputs in common. Thus, 2/6 or 1/3 of these extensions
can be used by the computer.
Since 1/3 of all forms besides form aaa are usable, 1/3 of their remaining 990 total three-digit
extensions (after subtracting all 10 of form aaa) are usable. Therefore, the greatest number of
assignable extensions is 10 + 990 3 = 340 .
Solution 2 for 4/4/13 by David Galkowski (10/NY)

1. The possible extension numbers can be divided into 3 groups: extension numbers with three
different digits, extension numbers with two different digits with one repeated, and extension
numbers with one digit used three times.
2. Extension numbers with three different digits are the largest group. There are 10 choose 3
ways to pick three different digits. Each set of three digits has two possible extensions that
will always produce different numbers when garbled. For example, if the digits 3, 2, and 1
were chosen, we could chose x123 and x321. When x123 was dialed, the computer would
receive 123, 213, or 132. When x321 was dialed, the computer would receive 321, 312, or
231. Since neither share numbers, yet all possible combinations of 3, 2, and 1 are used, there
are two possible extensions for each extension number with three different digits. the number
of extension numbers with three different digits is 2 10 = 240 .
3
3. The number of extension numbers with two different digits, one repeated once, is 10 9 .
There are 10 choices for the number that repeats and 9 choices for a different number. If two
3s and a 1 were chosen for an extension, only one number could be used, since all three pos
sible extensions could share a possible outcome of 313.
x133 could change to 313 or 133.
x313 could change to 133, 313, or 331.
x331 could change to 313 or 331.
The number of extension numbers with two different digits is 10 9 = 90 .
4. The number of extension numbers with 1 digit repeated twice more is 10 since there are only
10 digits and there is only one possible extension for each digit.
5. The greatest number of three-digit extensions the company can assign is 240 + 90 + 10 = 340
extensions.
5/4/13. Determine the smallest number of squares into which one can dissect a 11 13 rectan
gle, and exhibit such a dissection. The squares need not be of different sizes, their bases
should be integers, and they should not overlap.
Comment: We thank Prof. Berzsenyi for proposing this interesting problem. For an excellent
introduction to the topic of dissecting rectangles into squares, the reader is referred to Chapter 17
of Scientic Americans Second Book of Mathematical Puzzles and Diversions by Martin Gard
ner.
Comment by Erin Schram: To a mathematician, determine means nd and verify. Proving
that six is the smallest number of squares in a dissection is half the work in the problem, and a
solution would be incomplete without such a verication.
Solution 1 for 5/4/13 by Paul H. Ryu (10/CA)
This gures shows a dissection with six squares. Of
the six squares, two are of size 4 4 , and the other four
squares are of sizes 1 1 , 5 5 , 6 6 , and 7 7 .
Since the dissection is valid, we only have to verify that
there are no possible dissections with fewer than 6
squares.
To do that, we have to rst gure out the set of
whole numbers whose squares sum up to the area of the
rectangle, which is 143.
With the aid of a computer program, one can nd
sets of whole numbers whose squares sum up to the area
of the rectangle. There are no sets of two or three whole
numbers whose squares sum to 143. There are ve sets of four whole numbers whose squares sum
to 143. They are {1, 5, 6, 9}, {2, 3, 3, 11}, {2, 3, 7, 9}, {3, 3, 5, 10}, and {3, 6, 7, 7}. However,
none of them actually work, because the squares do not t together. That is because the combined
length of the two largest squares exceeds 13; thus, two of them cannot be laid on the 11 13 rect
angle without exceeding the boundary. The two squares have to be laid down somehow, but if two
of them have sides summing greater than 13, their sides opposite each other will be more than 13
apart.
Now we nd sets of ve whole numbers whose squares sum to 143. There are eight of them:
{1, 1, 2, 4, 11}, {1, 1, 4, 5, 10}, {1, 2, 5, 7, 8}, {1,3, 4, 6, 9}, {2, 4, 5, 7, 7}, {2, 5, 5, 5, 8}, {3, 3,
3, 4, 10}, and {3, 3, 5, 6, 8}. None can actually be arranged into an 11 13 rectangle. The
squares simply do not t. As before, the sum of the largest two squares exceeds 13, except with
the {2, 5, 5, 5, 8} case, in which the 5 5 squares cannot ll the gaps left by the 8 8 square.
Thus, we see that our dissection, which uses the set {1, 4, 4, 5, 6, 7}, is optimal, and the small
est number of squares into which one can dissect an 11 13 rectangle is six squares.
Solution 2 for 5/4/13 by Xuan Wu (12/CA)

Testing for smallest number of squares:
1? No. Biggest square that will t in 11 13 is 11 11 , which does not cover the remaining
2 11 .
2? No. Only conguration of two squares that make a rectangle is the gure to the right. The
length has to be 2 times longer than the width, L = 2W . However, 13 is not 2 11 , so an
11 13 rectangle cannot be formed by 2 squares.
n
n n
3? No. Two congurations:
n L = 3W 2L = 3W
n
n n n 2n
n
2n n
4? No. Five congurations:
n n
L = 4W L = W
n n n n n
n n
3n n n
n n
3n n 3L = 4W 3L = 5W
2n
n
3n n 3n 2n
n Considered to be the same

2n 2L = 5W as the case to the left; any
n
time the same squares can
2n 2n n be rearranged to form the
same dimensions, it will be
considered the same cong
uration.
5? No. Eleven congurations:
n n
L = 5W 2L = 7W
n n n n n n
2n 2n 2n n
n n
L = 2W 3L = 7W
n n
2n n n n
3n 3n n
2n 2n 2n 3n 3n n n
n
2n 5L = 6W 3L = 8W
2n
3n 3n 3n 2n
3n 3n
n n
n 4L = 5W 4L = 7W
n 3n
n
4n n 4n 3n
2n 2n 3n 5n n n
6L = 7W
n 5L = 7W
3n
2n
2n
3n
5n 2n
4n 3n
5n 2n n
5L = 8W
n
3n
5n 3n
6? Yes. One conguration, shown below, works. The smallest number of squares into which one
can dissect a 11 13 rectangle is 6.
6n 7n
11L = 13W
6n
7n
n
n n
n
5n
4n
5n 4n 4n
Solution 3 for 5/4/13 by Robert Cordwell (9/NM)
This is the solution with six squares.
6 5
1
4
7
4
It is relatively easy to prove that at least six squares are needed. We

know that there must be at least one square at each corner for a mini
mum solution, since we know that a solution with 6 squares exists and
that a solution with one square of size 11 covering two corners takes 8 11
squares. We prove that there must be 6 squares by showing that a solu
tion with ve or fewer squares is not possible.
1
2 2 2 2 2 1
Notice that all the sides of the rectangles must be covered completely by the sides of the
squares, and each corner square of length x covers x units on one of the rectangles sides of length
13 and x units on one of the rectangles sides of length 11. It should be fairly clear that with ve or
fewer squares at least three sides of the rectangle must be covered solely by the corner squares,
including at least one 13-long side. Let those two squares on the long side have sides of lengths a
and b. In order to have at most ve squares, the other two corner squares must have lengths of
11 a and 11 b , which sum up to 9, leaving a gap of 4 on the other side. Thus, there must be a
square of side length 4 (it could be done with more squares, but with ves squares we can use only
one non-corner square) on a 13-long side.
We must now try to place the four corner squares so that they cover
the rest of the rectangle. Notice that a square opposite the square of
length 4 must have side length 7 to avoid placing a sixth square between b 11 b
them.
4
a = 7
11 a
This in turn means that we have squares on the two adjacent corners
of side lengths 4 and 6.
6 11 b
4
7
4
Finally, this means that there must be a square with side length 5 in
the last corner.
6 5
4
7
4
However, when one adds up the areas of all ve large squares: 16 + 16 + 25 + 36 + 49 , one
nds that it is 142, one short of the area of the rectangle, 11 13 = 143 . Thus, there is no solu
tion with ve squares. By adding the one missing square, we have instead found our solution with
six squares.
1/1/14. Some unit cubes are stacked atop a at 4 by 4 square. The South View East View
gures show views of the stacks from two different sides. Find
the maximum and minimum number of cubes that could be in the
stacks. Also give top views of a maximum arrangement and a
minimum arrangement with each stack marked with its height.
Comment: This problem was inspired by Problem 1097 in Sndor Rkas excellent collection of
2000 problems in elementary mathematics.
Solution 1 for 1/1/14 by Ameya Velingker (9/PA):

First, assign a coordinate system to the unit squares inside the 4 by 4
(1,1)(1,2)(1,3)(1,4)
square as shown in gure 1.1. Let (i, j) denote the square that is in the i th
(2,1)(2,2)(2,3)(2,4)
row and j th column. Also let S ( i, j) denote the height of the stack on square (3,1)(3,2)(3,3)(3,4)
(i, j) .
(4,1)(4,2)(4,3)(4,4)
Let us rst nd the maximal arrangement of cubes that will give the
appropriate views. From looking at the south view, we can deduce that Figure 1.1
S ( k, 1) 4 for 1 k
4
S ( k, 2) 1 for 1 k
4
S ( k, 3) 4 for 1 k
4
S ( k, 4) 2 for 1 k 4
By looking at the east view, we can also deduce that

S(1, k) 3 for 1 k 4
S(2, k) 4 for 1 k 4
S(3, k) 2 for 1 k 4
S(4, k) 4 for 1 k 4
For each square (i, j) , we can choose the largest value for S ( i, j) that satis
4 1 4 2
es the above inequalities. This gives us the maximal arrangement with top
view shown in gure 1.2. Each square in the gure contains the height of the 2 1 2 2
stack on top of it. The total number of cubes used in this maximal congura 4 1 4 2
tion is 38. 3 1 3 2
Now let us nd a minimal arrangement. Looking at the east view, we have Figure 1.2
max(S(1, 1), S(1, 2), S(1, 3), S(1, 4)) = 3
max(S(2, 1), S(2, 2), S(2, 3), S(2, 4)) = 4
max(S(3, 1), S(3, 2), S(3, 3), S(3, 4)) = 2
max(S(4, 1), S(4, 2), S(4, 3), S(4, 4)) = 4
So, the number of cubes needed is at least 13. However, looking at the stack of height 1 in the
south view, there is at least one value of k such that S ( k, 2) = 1 . Therefore, at least 14 cubes are
needed. Figure 1.3 shows the top view of a working conguration using 14 cubes. Hence, this is a
minimal conguration.
4
2
4
1 3
Figure 1.3
Solution 2 for 1/1/14 by James Albrecht (11/IL):

To prove the mximum and minimum for the number of blocks, rst I will establish bounds and
then I will nd congurations that exist at the bounds.
Minimum: 14
According to the stacks seen from the east and the south, for the minimum
4
we need at least:
2 one stack of height 1,
4 one stack of height 2,
one stack of height 3, and

3
two stacks of height 4.
4 1 4 2
So min 1 + 2 + 3 + 4 + 4 = 14 .
There is a conguration with 14 cubes, so that is the minimum.
4 1 0 0
0 0 0 2
0 0 4 0
0 0 3 0
Maximum: 38
For the maximum, the south view lets us have 4s down the rst column, 1s
4
down the second column, 4s down the third column, and 2s down the fourth
2 column.
4 But the east view limits the rows. In row 2 each entry must be no higher than
2. In row 4 each entry is no higher than 3. This leaves a conguration with 38
3
cubes.
4 1 4 2
4 1 4 2
2 1 2 2
4 1 4 2
3 1 3 2
2/1/13. Find four distinct positive integers, a , b , c , and d , such that each of the four sums
a + b + c , a + b + d , a + c + d , and b + c + d
is the square of an integer. Show that
innitely many quadruples (a, b, c, d) with this property can be created.
Comment: This problem is based on Problem 5 of the 1993-94 Scottish Mathematical Challenge.
The interested reader is directed to Mathematical Challenges II, published by the Scottish Mathe
matical Council in 1995.
Solution 1 for 2/1/14 by Matthew Walsh (12/NY):

One quadruple satisfying the conditions is (9, 198, 269, 522).
If we let a = 3x 2 18 x 39 , b = 3x 2 + 6 , c = 3 x 2 + 18 x + 33 , and d = 3 x 2 + 36x + 42
where x is some integer greater than 7 (so that a, b, c, and d are all positive), all of the four sums
given in the problem will be perfect squares.
a + b + c = (3 x 2 18 x 39 ) + (3 x 2 + 6 ) + (3 x 2 + 18 x + 33) = 9 x 2 = ( 3 x ) 2
a + b + d = (3 x 2 18 x 39 ) + (3 x 2 + 6 ) + (3 x 2 + 36 x + 42) = 9 x 2 + 18 x + 9 = (3x + 3 ) 2
a + c + d = (3x 2 18 x 39 ) + (3 x 2 + 18 x + 33) + (3 x 2 + 36 x + 42) = 9 x 2 + 36 x + 36 = (3 x + 6 ) 2
b + c + d = (3x 2 + 6 ) + (3 x 2 + 18 x + 33) + (3 x 2 + 36 x + 42) = 9 x 2 + 54 x + 81 = (3 x + 9 ) 2
Because an innite number of values for x exists, there also exists an innite number of quadru
ples (a, b, c, d) such that each of the four sums a + b + c , a + b + d , a + c + d , and b + c + d
Solution 2 for 2/1/14 by Rohit Dewan (9/MD):

Let w, x, y, and z be the squares a + b + c , a + b + d , a + c + d , and b + c + d . So
1 1 1 0 a w
1 1 0 1 b = x
1 0 1 1 c y
0 1 1 1 d z
1 1 1 0 13 13 13 2 3
Using a calculator, I found the inverse of 1 1 0 1 to be 1 3 13 2 3 1 3 . Therefore,
1 0 1 1 13 2 3 13 13
0 1 1 1 2 3 13 13 13
a 13 13 13 2 3 w
b = 13 13 2 3 13 x
c 13 2 3 13 13 y
d 2 3 13 13 13 z
which means
a = (1 3)w + (1 3) x + (1 3)y (2 3)z
b = (1 3)w + (1 3) x (2 3)y + (1 3)z
c = (1 3)w (2 3) x + (1 3)y + (1 3)z
d = (2 3)w + (1 3)x + (1 3)y + (1 3)z
If w, x, y, and z are divisible by 3, then a, b, c, and d would be integers. I used 482 as w, 512 as x,
542 as y, and 572 as z. 48 2 = 2304 , 51 2 = 2601 , 54 2 = 2916 , and 57 2 = 3249 .
441 = (1 3)(2304) + (1 3)(2601) + (1 3)(2916) (2 3)(3249)
774 = (1 3)(2304) + (1 3)(2601) (2 3)(2916) + (1 3)(3249)
1089 = (1 3)(2304) (2 3)(2601) + (1 3)(2916) + (1 3)(3249)
1386 = (2 3)(2304) + (1 3)(2601) + (1 3)(2916) + (1 3)(3249)
So (441, 774, 1089, 1386) is a working example.

If I were to multiply the 2304, 2601, 2916, and 3249 by any square, the new a, b, c, and d
would be four different distinct positive integers. Therefore, innitely many quadruples
(a, b, c, d) can be created.
Comment from the solutions editor: Rohit Dewans method for creating an innite number of
quadruples is simple, but the equations that are deduced are very powerful for creating quadru
ples. With a little rearrangement, we get
w+x+y+z
a = ------------------------------- z
3
w+x+y+z
b = ------------------------------- y
3
w+x+y+z
c = ------------------------------- x
3
w+x+y+z
d = ------------------------------- w
3
So in order for a, b, c, and d to be integers, we need only have the sum of the squares,
w + x + y + z divisible by 3 rather than having each individual square divisible by 3. With two
more requirements that the squares w, x, y, and z be distinct, so that a, b, c, and d will be distinct,
and that the largest of the squares be smaller than half the sum of the other three squares, so that
the smallest of a, b, c, and d will be positive, these equations can generate all possible quadruples
(a, b, c, d) .
By using 82 as w, 92 as x, 102 as y, and 112 as z (note that 64 + 81 + 100 + 121 = 3 122 ),
we get the smallest quadruple (1, 22, 41, 58 ) .
3/1/13. For a set of points in a plane, we construct the perpendicular bisectors of the line
segments connecting every pair of those points and we count the number of points in which
these perpendicular bisectors intersect each other. If we start with twelve points, the maximum
possible number of intersection points is 1705. What is the maximum possible number of
intersection points if we start with thirteen points?
Comment: This problem was found in an unpublished manuscript by Jnos Pataki, a noted
Hungarian mathematician.
We wish to apologize for a aw in the statement of the problem. As Agustya
Mehtas solution mentions below, it is possible to have an innite number of points
of intersection between the perpendicular bisectors, starting with as few as four
points, if two line segments sharing the same perpendicular bisector are viewed as
generating two superimposed perpendicular bisectors. We said that starting with
twelve points gave 1705 intersections, so we felt that we had implicitly eliminated
innite intersections as a solution. However, we should have explicitly eliminated the
innite case by asking for the intersections of distinct perpendicular bisectors, as the editted prob
lem below does. We simply had not noticed that the innite case was possible in our translation
from the problems original Hungarian wording.
3/1/14. (editted) For a set of points in a plane, we construct the perpendicular bisectors of the
line segments connecting every pair of those points. We count the number of points in which
distinct perpendicular bisectors constructed this way intersect each other. If we start with
twelve points, the maximum possible number of intersection points is 1705. What is the max
imum possible number of intersection points if we start with thirteen points?
Solution 1 for 3/1/14 by Agustya Mehta (11/OH):

We assume that no two bisectors are common to each other, as this would result in innite
points of intersection.
x( x 1)
If we start with x points, we will have x = ------------------- connecting lines, and thus also
2 2
x( x 1)
------------------- perpendicular bisectors.
2
If we have n oblique lines with no restrictions on their placement, the mximum number of
n(n 1)
points of intersection that they create is n = -------------------- .
2 2
In our case, any three of the starting points form a triangle, and thus their perpendicular bisec
tors will all intersect at a single point. This means that, for every set of three starting points, we
will have a single point of intersection instead of the three given by .
3
2
Thus, we must subtract 2 from the total possible number of points of intersection given by
x
3
n(n 1) x (
x 1 )
the formula -------------------- where n = ------------------- to correct the count.
2 2
Therefore, the maximum number of points of intersection for the perpendicular bisectors of
the ( x ( x 1 )) 2 lines is
x ( x 1 ) x ( x 1 )
------------------- ------------------- 1
2 2 x ( x 1 )( x 2 x 2 ) x ( x 1 )( x 2 )
------------------------------------------------------------ 2 = ------------------------------------------------ 2 -----------------------------------
x
2 3 222 321
x ( x 1 )( x 2 )( x + 1 ) x ( x 1 )( x 2 )
= ------------------------------------------------------ -----------------------------------
8 3
= x ( x 1 )( x 2 ) ------------ ---
x+1 1
8 3
x ( x 1 )( x 2 )(3 x 5 )
= --------------------------------------------------------
24
x ( x 1 )( x 2 )(3 x 5 ) 12 ( 11 ) ( 10 ) ( 31 )
So, when x = 12 , --------------------------------------------------------- = --------------------------------------- = 1705
24 24
x ( x 1 )( x 2 )(3 x 5 ) 13 ( 12 ) ( 11 ) ( 34 )
and when x = 13 , --------------------------------------------------------- = --------------------------------------- = 2431 .
24 24
Solution 2 for 3/1/14 by Joel Lewis (12/NY):

We will approach the problem generally, for n points in the plane.
Any three points determine three segments, and thus three perpendicular bisectors. The three
perpendicular bisectors are concurrent at the circumcenter of the triangle determined by the three
points. So for every three points, we have one intersection. Thus, for n points in the plane, there
are C ( n, 3) points of intersection of this type.
Also, if you take two separate pairs of points that are not in some special position, the perpen
dicular bisectors of their segments must intersect in a point. There are C ( n, 2) ways to pick the
rst pair of points and C ( n 2, 2) ways to pick the second pair. Order doesnt matter, so we have
C ( n, 2) C (
n 2, 2) 2 total points of intersection of this type.
So in general, we have a maximum of C ( n, 3) + C ( n, 2) C (
n 2, 2) 2 points of intersec
tion. For n = 12 this gives us 220 + 66 45 2 = 220 + 1485 = 1705 , and for n = 13 , this
gives 286 + 78 55 2 = 286 + 2145 = 2431 .
Solution 3 for 3/1/14 by Mitka (Dmitry) Vaintrob (8/OR):

The answer is C ( C(13, 2), 2) 2C(13,
3) = 2431 . In general, if there are n points, there are
at most C ( C ( n, 2), 2) 2C (
n, 3) intersection points. This is because for every pair of points, P
and Q , there is one bisector B P, Q and each two bisectors intersect in at most one point. So the
total number of pairs of bisectors C ( k, 2) , where k is the number of bisectors C ( n, 2) , is an
upper bound on the number of intersection points.
For every three points, P , Q , and R , the three bisectors B P, Q , B P, R , and B Q, R are either
parallel (this happens when P , Q , and R are collinear) or concurrent, intersecting at the circum-
center of triangle PQR . So for the three pairs of bisectors {B P, Q, B P, R } , {B P, Q, B Q, R } ,
{B P, R, B Q, R } , we get at most one intersection point, So we have to subtract at least two pairs of
bisectors for each triple of points. This gives an upper bound of C ( C ( n, 2), 2) 2C ( n, 3) for the
maximum number of intersection points.
Let us now show that this number of intersections can indeed be achieved.
We will show by induction on n that there exist n points in the plane such that any two distinct
bisectors of pairs of those points intersect and the only time three bisectors are concurrent is when
they are of the form B P, Q , B P, R , B Q, R for any three points P , Q , R , and no four or more bisec
tors are concurrent.
Let us suppose the hypothesis is true for n 1 points and P 1 , P 2 , , P n 1 are a congura
tion of points for which this is true. We will show that it is possible to add another point P n so that
the formula for number of intersections still hold for n points.
Notice that B P, Q is parallel to B R, S only if PQ is parallel to RS , so if P i P n is not parallel to
P j P k for any 1 i, j, k < n with j < k then the bisectors B Pi, Pn and B P j, Pk will intersect. So if
P n does not lie on a certain collection of lines through the points P i , then all the bisectors will
intersect.
Also notice that for a point O , the bisector B P, Q passes through O only if the distance OP
equals the distance OQ . So given any point O where two or more existing bisectors intersect, the
bisector B Pi, Pn will not pass throught O if OP n does not equal OP i for 1 i < n . So if P n does
not lie on a certain collection of circles whose centers are the points of intersection for the cong
uration of n 1 points, then no new perpendicular bisector will be concurrent with two or more
existing perpendicular bisectors.
There is just one more type of concurrence that we didnt consider: a concurrence with two or
more new bisectors involving P n , such as a concurrence between B Pi, Pn , B P j, Pn , and B Pk, Pl , for
1 i, j, k, l < n with i < j and k < l . In the example here, we must notice that the point of inter
section of B Pi, Pn and B P j, Pn is also on the old bisector B Pi, P j since P i , P j , and P n form a trian
gle. So if { i, j } {k, l} , both B Pi, P j and B Pk, Pl would pass through that point, so we would have
the concurrence at the intersection of two old bisectors, which brings this example into a previous
case. The only other example possible would be when B Pi, Pn , B P j, Pn , and B Pk, Pn concur, but that
intersection point is also on B Pi, P j , B Pi, Pk , and B P j, Pk , which is again that previous case.
So, as long as P n is not is not on a certain nite set of circles nor on a certain nite set of
lines, the points P 1 , P 2 , , P n 1 , P n satisfy all the conditions.
Our induction hypothesis is true when n = 3 , for when we have three points P 1 , P 2 , and P 3
not all on one line, the three bisectors intersect at the circumcenter of the triangle P 1 P 2 P 3 and
C ( C(3, 2), 2) 2C(3, 3) = C(3, 2) 2C(3, 3) = 3 2 1 = 1 .
Therefore, the number of intersections of bisectors for the collection of points we constructed
is equal to C ( C ( n, 2), 2) 2C ( n, 3) . Which shows that our upper bound is indeed the maximum
number. When n = 13 , this gives the answer C ( C(13, 2 ), 2) 2C(13, 3 ) = 2431 .
4/1/13. A transposition of a vector is created by switching exactly two entries of the vector. For
example, (1, 5, 3, 4, 2, 6, 7) is a transposition of (1, 2, 3, 4, 5, 6, 7) . Find the vector X if
S = (0, 0, 1, 1, 0, 1, 1) , T = (0, 0, 1, 1, 1, 1, 0) , U = (1, 0, 1, 0, 1, 1, 0) , and
V = (1, 1, 0, 1, 0, 1, 0) are all transpositions of X. Describe your method for nding X.
Comment: This problem was created by Drs. Gene Berg and George Berzsenyi on the basis of an
article by Vladimer I. Levenshtein, published in the January 2001 issue of the IEEE Transactions
on Information Theory.
Solution 1 for 4/1/13 by Brian Ho (12/IL):

To nd the vector X, we want to look for a vector as similar to S, T, U, and V as possible, since
vector transposition does not change the vector very much.
S = (0, 0, 1, 1, 0, 1, 1)
T = (0, 0, 1, 1, 1, 1, 0)
U = (1, 0, 1, 0, 1, 1, 0)
V = (1, 1, 0, 1, 0, 1, 0)
We can notice that positions 2 and 7 have three of the four vectors as 0. Positions 3, 4, and 6
have three or four vectors as 1. Based on the above reasoning, we can say that
X = (?, 0, 1, 1, ?, 1, 0) .
If we compare this X with vector V, we notice that the entries at positions 2 and 3 are switched.
The unknown entries of X must thus be the same as the corresponding entries in V, because only
two entries are switched in a transposition. Thus X = (1, 0, 1, 1, 0, 1, 0) .
We can verify that this is a viable vector X by testing that it is a transposition of S, T, and U as
well as of V. Switching positions 1 and 7 in X results in S, switching positions 1 and 5 in X results
in T, switching positions 4 and 5 results in U, and switching positions 2 and 3 results in V. We
have, therefore, found X.
Comment from the solutions editor: Brian Hos nice solution uses a method mathematicians
call majority vote. It could be improved by determining when majority vote would yield correct
information about X. As it is, Mr. Ho had to verify his answer at the end to be rigorous.
If the writer of the problem had added a fth transposition W = (0, 1, 1, 1, 0, 1, 0) to the
other four transpositions, majority vote would have given (0, 0, 1, 1, 0, 1, 0) as the result, which
is not X and has too many zeros. Common sense tells us that given too many zeros, the 3-to-2
votes for 0 are the suspect entries, so the result converts to (?, 0, 1, 1, ?, 1, 0) , which we can han
dle as Mr. Ho did. But it is wise to support common sense with rigorous mathematics.
As Mr. Ho argued, a transposition is not changed from X by much: it differs in two entries (we
will ignore switching two identical entries, since that would yield X itself and testing shows that
neither S, T, U, nor V is X). So the four transpositions combined differ from X at exactly eight
entries. The two columns at positions 1 and 5 must use up four of those different entries, two in
each column. That leaves four other different entries. We have four columns with a 3-to-1 vote, so
the minority values in those column must be the four other entries that dont match X. Thus, we
have proven Mr. Hos claim that X = (?, 0, 1, 1, ?, 1, 0) .
Solution 2 for 4/1/14 by Yuyin Chen (11/MI):
I will list every transposition of S, T, U, and V.
S T U V
(0, 0, 0, 1, 1, 1, 1) (0, 0, 0, 1, 1, 1, 1) (0, 0, 1, 0, 1, 1, 1) (0, 1, 0, 1, 0, 1, 1)

(0, 0, 1, 0, 1, 1, 1) (0, 0, 1, 0, 1, 1, 1) T = (0, 0, 1, 1, 1, 1, 0) (0, 1, 0, 1, 1, 1, 0)
S = (0, 0, 1, 1, 0, 1, 1) S = (0, 0, 1, 1, 0, 1, 1) (0, 1, 1, 0, 1, 1, 0) (0, 1, 1, 1, 0, 1, 0)
(0, 0, 1, 1, 1, 0, 1) (0, 0, 1, 1, 1, 0, 1) (1, 0, 0, 0, 1, 1, 1) (1, 0, 0, 1, 0, 1, 1)
T = (0, 0, 1, 1, 1, 1, 0) T = ( 0, 0, 1, 1, 1, 1, 0) (1, 0, 0, 1, 1, 1, 0) (1, 0, 0, 1, 1, 1, 0)
(0, 1, 0, 1, 0, 1, 1) (0, 1, 0, 1, 1, 1, 0) (1, 0, 1, 0, 0, 1, 1) (1, 0, 1, 1, 0, 1, 0)
(0, 1, 1, 0, 0, 1, 1) (0, 1, 1, 0, 1, 1, 0) (1, 0, 1, 0, 1, 0, 1) (1, 1, 0, 0, 0, 1, 1)
(0, 1, 1, 1, 0, 0, 1) (0, 1, 1, 1, 0, 1, 0) U = (1, 0, 1, 0, 1, 1, 0) (1, 1, 0, 0, 1, 1, 0)
(0, 1, 1, 1, 0, 1, 0) (0, 1, 1, 1, 1, 0, 0) (1, 0, 1, 1, 0, 1, 0) (1, 1, 0, 1, 0, 0, 1)
(1, 0, 0, 1, 0, 1, 1) (1, 0, 0, 1, 1, 1, 0) (1, 0, 1, 1, 1, 0, 0) V = (1, 1, 0, 1, 0, 1, 0)
(1, 0, 1, 0, 0, 1, 1) U = (1, 0, 1, 0, 1, 1, 0) (1, 1, 0, 0, 1, 1, 0) (1, 1, 0, 1, 1, 0, 0)
(1, 0, 1, 1, 0, 0, 1) (1, 0, 1, 1, 0, 1, 0) (1, 1, 1, 0, 0, 1, 0) (1, 1, 1, 0, 0, 1, 0)
(1, 0, 1, 1, 0, 1, 0) (1, 0, 1, 1, 1, 0, 0) (1, 1, 1, 0, 1, 0, 0) (1, 1, 1, 1, 0, 0, 0)
Now I will systematically search for any vector that is in all four lists. If a vector is in the S list
and not in the T list, I do not have to search for it in the U list, and if a vector is in the S and T lists
and not in the U list, I do not have to search for it in the V list. So 13 vectors in the S list reduce to
7 vectors in the T list, then to 3 vectors in the U list, and then to 1 vector in the V list, which is
(1, 0, 1, 1, 0, 1, 0)
. It is a transposition of all four vectors; therefore, all four vectors are transpo
sitions of it. So our solution is X = (1, 0, 1, 1, 0, 1, 0)
.
Comment from the solutions editor: I sorted Yuyin Chens lists into lexicographical order,
because such an ordering makes comparisons between lists easier.
Solution 3 to 4/1/14 by Michael Chmutov (12/OM):

Suppose the rst entry of X is 0. Then in the vectors U and V it was changed by the transposi
tion. In the vector U that 0 would have been switched with either the second, fourth, or seventh
entry, which would make X either (0, 1, 1, 0, 1, 1, 0) , (0, 0, 1, 1, 1, 1, 0) , or (0, 0, 1, 0, 1, 1, 1) .
But V is not a transposition of any of these. Consequently, the rst entry of X is 1.
In the vector S that rst entry of X was switched with either the third, fourth, sixth, or seventh
entry, which makes X either (1, 0, 0, 1, 0, 1, 1) , (1, 0, 1, 0, 0, 1, 1) , (1, 0, 1, 1, 0, 0, 1) , or
(1, 0, 1, 1, 0, 1, 0) . The only one of these for which T is a transposition is (1, 0, 1, 1, 0, 1, 0) .
Checking it against U and V, they also turn out to be transpositions of it.
Answer: X = (1, 0, 1, 1, 0, 1, 0) .
Solution 4 for 4/1/14 by Michael Lieberman (12/PA):
First I will note that if two transpositions of X are different in four places, then two of those
places must have been transposed in one and the other two in the other. Therefore, the remaining
three places must all be the same as the original vector X. For example, U and V are different in
positions 2, 3, 4, and 5, so their entries at positions 1, 6, and 7 must be the entries in X.
U = (1, 0, 1, 0, 1, 1, 0)
V = (1, 1, 0, 1, 0, 1, 0)
X = (1, ?, ?, ?, ?, 1, 0 )
Similarly, by comparing S with U and S with V the rest of X can be determined.
S = ( 0, 0, 1, 1, 0, 1, 1 ) S = ( 0, 0, 1, 1, 0, 1, 1 )
U = ( 1, 0, 1, 0, 1, 1, 0 ) V = ( 1, 1, 0, 1, 0, 1, 0 )
X = ( ?, 0, 1, ?, ?, 1, ? ) X = ( ?, ?, ?, 1, 0, 1, ? )
(1, ?, ?, ?, ?, 1, 0)
(?, 0, 1, ?, ?, 1, ?)
(?, ?, ?, 1, 0, 1, ?)
X = (1, 0, 1, 1, 0, 1, 0)
Note that vector T is not even needed, but it can be used to check the correct answer. The fol
lowing diagram shows which entries were switched.
X = (1, 0, 1, 1, 0, 1, 0)
S = (0, 0, 1, 1, 0, 1, 1)
T = (0, 0, 1, 1, 1, 1, 0)
U = (1, 0, 1, 0, 1, 1, 0)
V = (1, 1, 0, 1, 0, 1, 0)
Solution 5 for 4/1/14 by Brandon Levin (12/OH):

Let X = ( x 1, x 2, x 3, x 4, x 5, x 6, x 7 ) . Observe that X must contain four 1s and three 0s since
S, T, U, and V do. Also note that none of S, T, U, and V can be X.
Now the dot product X X is 4, since there are four 1s. (A dot product of two vectors is the
sum of all the pairwise products of entries in the same positions.) However, when one of the 1s is
switched with a 0 to make a transposition, the dot product of X and the transposition will be less
by 1, in other words, the dot product will be 3.
Therefore, the dot product of X with any of the transpositions S, T, U, and V will be 3. Also the
dot product of known vector, such as S, with X will be the sum of the entries in X : the variables
x 1 , x 2 , x 3 , x 4 , x 5 , x 6 , and x 7 . So, each dot product of X with the transpositions S, T, U, and V
will create an equation in the variables x 1 , x 2 , x 3 , x 4 , x 5 , x 6 , and x 7 .
Therefore,
X S = x3 + x4 + x6 + x7 = 3
X T = x3 + x4 + x5 + x6 = 3
X U = x1 + x3 + x5 + x6 = 3
X V = x1 + x2 + x4 + x6 = 3
For solving the equations, we can use the extra information that the seven variables consist of
four 1s and three 0s. So three of the variables in each equation above are 1 and the fourth vari
able is 0. Since x 6 is in every equation, if it were 0, then all the other variables would have to
be 1, which cannot be true. So x 6 = 1 .
Also, subtracting the rst two equations from each other gives that x 5 = x 7 and subtracting
the middle two equations from each other gives that x 1 = x 4 . Substituting for x 4 , x 6 , and x 7
leaves two distinct equations:
x1 + x3 + x5 = 2
2 x1 + x2 = 2
That last equation can be solved with 0s and 1s only by x 1 = 1 and x 2 = 0 .
We know three of the four 1s in X already: x 1 , x 4
, and x 6 . There is only one 1 left and since
x 5 = x 7 , they cannot be 1. So x 2 , x 5 , and x 7 are 0, which leaves that x 3 = 1 .
Finally, X = (1, 0, 1, 1, 0, 1, 0) .
Comment from the solutions editor: Brandon Levin could have used the all-ones vector
1 = (1, 1, 1, 1, 1, 1, 1) to create a fth equation:
X 1 = x1 + x2 + x3 + x4 + x5 + x6 + x7 = 4
which represents the information that X contains exactly four 1s.
5/1/14. As illustrated below, we can dissect every triangle ABC into four pieces so that piece 1
is a triangle similar to the original triangle, while the other three pieces can be assembled into
a triangle also similar to the original triangle. Determine the ratios of the sizes of the three
triangles and verify that the construction works.
B
D
3
G 2 F Piece 2
H rotated
4 1 4
1 180 3
2
A C
E
Comment: This problem was inspired by Problem 1365 in Sndor Rkas excellent collection of
2000 problems in elementary mathematics, which is also Problem 1113 in his earlier collection of
only 1500 problems.
Some solutions showed that several people were unsure what the instruction verify that the
construction works meant. It meant to show that the pieces t together properlyby showing
that sides that had to t together had the same length and angles that had to t together into a
straight line were supplementaryand that the three triangles were similar.
We counted both the ratios of the lengths, 5 : 4 : 3, and the ratios of the areas, 25 : 16 : 9, as
correct answers. Size is an ambiguous word, but the size of a polygon usually means the lengths
of its sides.
Solution 1 for 5/1/14 by Kristen Kozak (11/WA):

The construction works if we make the dissection lines of triangle ABC parallel to its sides as
shown below.
B
G F
A C
E
Making the lines parallel also allows us to make the following equalities
DB = HF Opposite sides of parallelogram DBFH.

DH = BF Opposite sides of parallelogram DBFH.
GF = EC Opposite sides of parallelogram GFCE.
GE = FC Opposite sides of parallelogram GFCE.
By looking at the assembled triangle, we obtain these equalities.
DG = GH Aligned where pieces 2 and 3 touch on assembled triangle.

HE = BF Aligned where pieces 3 and 4 touch on assembled triangle.
We nd the ratio of the size of triangle ADE to the size of triangle ABC by nding the ratio of the
length of segment DE to the length of segment BC , because DE and BC are corresponding
edges. First we dene x to be the length of segment DG . Then we use the equalities above to nd
the lengths of the the other line segments on segments DE and BC in terms of x. These lengths
are shown below.
D 2x
x
G F
x
H 3x
2x
A C
E
We nd the length of DE by summing the lengths of DG , GH , and HE . We nd the length of
BC by summing the lengths of BF and FC . DE = 4x . BC = 5x . The ratio of the length of
side DE to the length of side BC is 4 x : 5 x , which is 4 : 5. We square the ration of the lengths to
nd the ratio of the area of triangle ADE to the area of triangle ABC. We get 16 : 25.
We nd the ratio of the size of the assembled triangle to the size of triangle ABC by nding
the ratio of the length of segment FC to the length of segment BC , because FC and BC are cor
responding edges. We know that the length of FC is 3 x and the length of BC is 5 x , so the ratio
of their lengths is 3 : 5. We square the ratio of the lengths to nd the ratio of the areas of the two
triangles. We get 9 : 25.
So, the ratio of the area of triangle ABC to the area of triangle ADE to the area of the assem
bled triangle is 25 : 16 : 9.
Solution 2 for 5/1/14 by Sam McVeety (11/MN):

In verifying that this construction works, I found a Euclidean method of creating the gure
that clearly shows both the area ratios and veries the construction. First, one must divide base
AC into ve equal parts.1 We now construct parallels to side BC at the four points dividing those
ve parts of AC . Those parallel lines divide side AB into ve equal parts. Next, construct paral
1. Using Euclids method, we draw a line AM where M is a point not on line AC and
create ve adjacent equal segments on AM beginning at A, labeling the endpoint of the
last segment as N. We next draw line segment CN and construct a parallel to CN at
every endpoint of the equal segments on AM . These parallel lines will divide line seg
ment AC into ve equal parts.
lels to side AC at the four points on AB where the parallels to side BC intersected. Those paral
lel lines divide side BC into ve equal parts. Finally, constuct parallels to side AB from the four
points on AC where the parallels to side AC intersected. Those parallel lines pass through the
original four points of penta-section on side AC . We have now created 25 congruent triangles,
which are similar to the original triangle ABC because the parallel lines cut equal angles at their
intersections with parallel lines, and are congruent to each other because adjacent triangles share a
matching side and triangles along each side of the large triangle ABC have sides that are from ve
equal parts of that side.
B
D D

G F

H

A C
A 1 E 2 3 4
E
The divided triangle and its similar triangles, with important angles labeled
In the construction, line segment DE lies over one of the parallels to BC , line segment FG
lies over one of the parallels to AC , and line segment FH lies over one of the parallels to AB , as
shown in the diagram above. Piece 1 of triangle ABC consists of 16 of the 25 tiny triangles, piece
2 is one of the 25 tiny triangles, piece 3 consists of three of the 25 tiny triangles, and piece 4 con
sists of ve of the 25 tiny triangles. Since the tiny triangles t together, pieces 2, 3, and 4 t
together to form a triangle made of nine of the tiny triangles.
Clearly, the ratio of the areas of triangle ABC to triangle ADE to the assembled triangle is
25 : 16 : 9.
Extending this construction to dividing triangles into n parts in similar fashion,1 where n is
even and greater than 2, yields the following ratio of areas:
2 2n + 2 2 n 2 2n 2
n--------------------------
- : ----------------- : (n 1 ) 2
2 2
1. The triangle is divided so that one large part is similar to the original triangle, and the
other n 1 pieces are a small triangle and several trapezoids of increasing size that are
assembled into a triangle similar to the original. For example, n = 6 would be:
Rotate 180
1/2/14. Each member of the sequence 112002, 11210, 1121, 117, 46, 34, is obtained by add
ing ve times the rightmost digit to the number formed by omitting that digit. Determine the
billionth (109th) member of this sequence.
Comments: This problem was based on a problem from issue 3/2001 of Matlap, Transylvanias
excellent Hungarian language mathematics journal for students at the middle school and high
school levels.
Solution for 1/2/14 by Aleksandr Arkhipov (9/NJ):

This sequence generates a new member by taking a function, f ( x ) , of the preceding
member x , so that f ( nth member) is the (n + 1 ) st member. The function f ( x ) can be simplied
if we express the integer x as 10k + y , where y = x mod10 ; i.e., y is the ones digit of x . Then
we can say that f (10k + y) = k + 5 y , since the function removes the last digit y , leaving behind
the integer part of x /10, which is k , and adds ve times that last digit y .
Next, the value of any member of the sequence depends only on the value of the preceding
member. Therefore, as soon as any member appears that is equal to any previous member in the
sequence, the rest of the series will repeat again and again generating the same value every N th
time, where N is the distance between the original two identical members. This leads to three
possibilities:
1) The sequence settles on one number x for which f ( x ) = x .
2) The squence enters a cycle in which a block of N numbers repeats over and over again.
3) No member ever repeats.
In option 3 the sequence increases boundlessly toward innity or bounces around without
repeating in an innite range, since its range has to hold innitely many values. However, when
ever any member of the sequence is under 49, all members thereafter will be under 49. Therefore,
the sequence will stay under 49 after the fth member 46. So option 3 is ruled out.
Forty-nine is the maximum value of f ( x ) for x 49 , because if k 4 in the 10k + y
form of x , then f (10k + y) = k + 5 y 4 + 5 ( 9 ) = 49 .
Next, I wrote a simple computer program to output the sequence. I found the sequence
started with the nonrepeating members 112002, 11210, 1121, and 117, and after that a block of 42
numbers (46, 34, 23, 17, 36, 33, 18, 41, 9, 45, 29, 47, 39, 48, 44, 24, 22, 12, 11, 6, 30, 3, 15, 26,
32, 13, 16, 31, 8, 40, 4, 20, 2, 10, 1, 5, 25, 27, 37, 38, 43, 19) repeats. This means that the 5th,
47th, and 89th members were all 46, continuing to all positions that equal 5 modulo 42 (that
means the position n can be expressed as 42i + 5 for some integer i). By the same logic, since
10 9 reduced modulo 42 equals 34, (10 9 = 238095238 42 + 34 ), the billionth member is the
same as as the 34th member, which is 40.
2/2/14. The integer 72 is the rst of three consecutive integers 72, 73, and 74, that can each be
expressed as the sum of the squares of two positive integers. The integers 72, 288, and 800 are
the rst three members of an innite increasing sequence of integers with the above property.
Find a function that generates the sequence and give the next three members.
Comment: This problem is based on a similar problem that appeared in the 1991 Indiana College
Mathematics Competition, initiated by Professor Paul Mielke.
Solution 1 for 2/2/14 by George Lin (11/NJ):

We want a function f ( n ) such that f ( 1 ) = 72 , f ( 2 ) = 288 , f ( 3 ) = 800 , and f ( n ) ,
f ( n ) + 1 , and f ( n )
+ 2 can all be expressed as the sum of the squares of two positive integers for
all positive integers n . First, let us write out the known three sets of consecutive integers as the
sum of the squares of two positive integers:
f (1) f (2) f ( 3
)
2 2
2 2 2 2
72 = 6 + 6 288 = 12 + 12 800 = 20 + 20
2 2
2 2 2 2
73 = 3 + 8 289 = 8 + 15 801 = 15 + 24
2 2 2 2 2 2
74 = 5 + 7 290 = 11 + 13 802 = 19 + 21
Notice that all of the rst of the three consecutive integers, 72, 288, and 800, are of the form
a 2 + a 2 . Thus, assume f ( n ) = a n2 + a n2 = 2a n2 .
Notice the all the last of the three consecutive integers, 74, 290, and 802, are of the form
2 2
(a 1 ) + (a + 1 ) . If we expand this out, we can see this clearly works for any f ( n ) = 2a n2 :
2 2
f ( n ) + 2 = (a n 1 ) + (a n + 1 ) = (a n2 2a n + 1 ) + (a n2 + 2a n + 1) = 2a n2 + 2.
Now, we need only deal with the second of the three consecutive integers. We have a 1
= 6 ,
a 2 = 12 , and a 3 = 20 , so
2 2
f ( 1 ) + 1 = (a 1 3 ) + (a 1 + 2 )
2 2
f ( 2 ) + 1 = (a 2 4 ) + (a 2 + 3 )
2 2
f ( 3 ) + 1 = (a 3 5 ) + (a 3 + 4 )
Based on the rst three members of the sequence f ( n ) + 1 , we extrapolate the trend and assume
2 2
that f ( n ) + 1 = (a n (n + 2 )) + (a n
+ (n + 1 )) . This gives
2 2
f ( n )
+ 1 = (a n (n + 2 )) + (a n + (n + 1 ))
2 2
= a n2 2a n (n + 2 ) + (n + 2 ) + a n2 + 2a n (n + 1 ) + (n + 1 )
= 2a n2 2a n + 2n 2 + 6n + 5
2
Since f ( n ) + 1 = 2a n2 + 1 , we get the equation
2a n2 + 1 = 2a n2 2a n + 2n 2 + 6n + 5
2a n = 2n 2 + 6n + 4
2
a n = n + 3n + 2
2
Therefore, we can let a n = n + 3n + 2 and f ( n ) = 2a n2 , which gives
f ( n ) = a n2 + a n2
2 2
f ( n ) + 1 = (a n (n + 2 )) + (a n + (n + 1 ))
2 2
f ( n ) + 2 = (a n 1 ) + (a n + 1 ) ,
and f ( 1 ) = 72 , f ( 2 ) = 288 , and f ( 3 ) = 800 . Letting n = 4, 5, 6 , we nd the next three
terms in the sequence: 1800, 3528, 6272.
Solution 2 for 2/2/14 by Sherry Gong (8/PR):

2 2
Answer: The function is generated by 2(n + 1 ) (n + 2 ) . The next three terms are 1800,
3528, and 6272.
Proof:
2 2 2 2
2(n + 1 ) (n + 2 ) = ((n + 1 )(n + 2 )) + ((n + 1 )(n + 2 ))
= 2n 4 + 12n 3 + 26n 2 + 24n + 8
2 2 2 2
2(n + 1 ) (n + 2 ) + 1 = (n ( n + 2 )) + ((n + 1 )(n + 3 ))
= 2n 4 + 12n 3 + 26n 2 + 24n + 9
2 2 2 2
2(n + 1 ) (n + 2 ) + 2 = ((n + 1 )(n + 2 ) 1 ) + ((n + 1 )(n + 2 ) + 1 )
= 2n 4 + 12n 3 + 26n 2 + 24n + 10
3/2/14. An integer lattice point in the Cartesian plane is a point ( x, y) where x and y are both
integers. Suppose nine integer lattice points are chosen such that no three of them lie on the
same line. Out of all 36 possible line segments between pairs of those nine points, some line
segments may contain integer lattice points besides the original nine points. What is the mini
mum number of line segments that must contain an integer lattice point besides the original
nine points? Prove your answer.
Comment: This problem was modied by Gene Berg and George Berzsenyi from a problem sug
gested by Professor Tom Hoeholdt of the Technical University of Denmark.
Solution 1 for 3/2/14 by Irena Wang (11/IL):

Answer: 6
If two points (a, b) and (c, d) have lattice points on the segment between them, then a c
and b d must both be divisible by some common factor q with q > 1 . So a c (mod q ) and
b d (modq ) , which we can write as (a, b) (c, d) (modq ) . For example,
(10, 2 ) (3, 9) (mod 7) .
For modulo n where n is 3 or higher, it is possible for nine points to have no congruent pairs
modulo n, since there are n 2 different possible types of points modulo n.
2
For modulo 2, however, there are only 2 = 4 possiblities: (even, even), (even, odd),
(odd, even), and (odd, odd), so there will be congruent pairs of points. For the least number of
such pairs, we can make 3 points of one type, such as (even, even), and 2 points in each of the
other types. Since the segments determined by congruent pairs of points contain lattice points, this
will give us at least 32 + 22 + 22 + 22 = 6 segments containing lattice points.
Now that a lower bound has been established, we need a real example of nine points that
determine only six segments containing other lattice points to complete the proof. One such
example is (0, 0) , (1, 6) , (2, 2) , (2, 4) , (3, 7) , (4, 5) , (5, 6) , (6, 5) , and (7, 1) . Note that
there are no two points congruent in any modulus beyond 2.
8
7
6
5
4
3
2
1
1 2 3 4 5 6 7 8
Solution 2 for 3/2/14 by Jonathan Sasmor (12/NY):

Two integers have the same parity if they are both even or both odd. Consider two lattice
points to have the same parity if their x-coordinates have the same parity and their y-coordinates
have the same parity. If two lattice points are of the same parity, the differences in their x- and
y-coordinates will be even, and thus, by the midpoint formula, the midpoint of the line segment
connecting them will be a lattice point. We seek to minimize the number of line segments con
necting lattice points of the same parity.
Regardless of how we choose our nine lattice points, there will be at least six line segments
connecting lattice points of the same parity.
If there is any group of four or more lattice points of the same parity, there will be at least
4 = 6 line segments connecting these points.
2
If there are two groups of three lattice points of the same parity, there must be at least
2 32 = 6 line segments connecting the points in each group.
Since there are only four lattice point parities: (even, even), (odd, odd), (even, odd),
(odd, even)and nine lattice points, then from the pigeonhole principle, there is at least
one group of three points of the same parity. If there is only one such group, the other
three parities must have two points each. The number of line segments with midpoint
lattice points would be 3
2 + 3 22 = 6 .
Thus, in all possible cases, there must be at least six line segments with lattice point midpoints.
For the set of lattice points { (0, 0) , (0, 1) , (1, 1) , (1, 2) , (2, 1) , (2, 5) , (3, 2) , (4, 6) ,
(11, 3 ) }, no three points are collinear and exactly 6 of the 36 connecting line segments contain a
lattice point besides the original nine points. Therefore the minimum possible number of such line
segments is 6.
7
6
5
4
3
2
1
1 2 3 4 5 6 7 8 9 10 11
4/2/14. Let f ( n ) be the number of ones that occur in the decimal representations of all the num
bers from 1 to n . For example, this gives f ( 8 ) = 1 , f ( 9 ) = 1 , f ( 10 ) = 2 , f ( 11 ) = 4 ,
100
and f ( 12 ) = 5 . Determine the value of f (10 ).
Comment: Professor Roger Pinkham of Stevens Institute of Technology suggested the original
problem that used this function, and Dr. Michael Zieve of IDA/CCR simplied the suggestion into
a problem appropriate for this competition.
Solution 1 for 4/2/14 by Daniel McLaury (12/OK):

100
It is helpful to think of numbers between 0 and 10 1 as strings of exactly 100 digits. For
example, 1417 would be represented as 000000001417. Then simple combinatorics show a
pretty simple solution.
100
Note that 10 1 is 99999999, all its digits are 9. Observe that exactly one tenth of the
100
numbers between 0 and 10 1 will have a 1 in the ones place; similarly, one tenth will have a
99
1 in the tens place, etc., and the pattern continues all the way up to the 10 s place. (We are free
to add 0 to the list of numbers that we count 1s in, because it has no 1s in it and therefore does
not affect the total.)
0 1 2
Since there are 100 places in each number (the 10 s place, the 10 s place, the 10 s place,
99 10 99
up to the 10 s place) and in each place 100 10 = 10 many 1s will show up, there are
99 101 100
100 10 = 10 many 1s in the numbers between 0 and 10 1 . Then all that is left to
100 101
count is the single 1 in the googols place in 10 itself. Thus, there are 10 + 1 many 1s in
100 101
all. So f (10 ) = 10 + 1.
a1 a
In general, a10 + 1 many 1s are used in writing all the numbers from 1 to 10 .
Solution 2 for 4/2/14 by Daniel Walton (12/WA):

1 10
Using combinatorics, I found f (10 ) through f (10 ) . Here is an example of my method for
3
f (10 ) , i.e., f (1000) :
A number between 1 and 1000 contains zero ones, one one, two ones, or three
ones. Consider all numbers between 1 and 1000 that contain exactly n ones. The
number of ones found in all such numbers is
( 3 n)
n (The number of ways to position n ones) 9 ,
except when n is one, in which case 1000 itself gives an additional one.
For numbers with only one one (besides 1000), that is 1 31 9 = 243.
2
For numbers with exactly two ones, that is 2 32 9 = 54.

1
For numbers with exactly three ones, that is 3 33 9 = 3.

0
6
There is another one in the number 1000, that is 1.
243 + 54 + 3 + 1 = 301 .
n
I observed a pattern in the function f (10 ) .
1
f (10 ) = 2 = 1 +
1
f (10 2 ) = 21 = 20 +
1
3
f (10 ) = 301 = 300 +
1
4
f (10 ) = 4001 = 4000 + 1
5
f (10 ) = 50001 = 50000 + 1
n n1
The pattern is f (10 ) = n10 + 1.
100 99 101
Therefore, f (10 ) = 100 ( 10 ) + 1 = 10 + 1.
Comment from the solutions editor: It would have been more elegant for Mr. Walton to calcu
100
late f (10 ) directly with his combinatorial method, as
) = 1
100 99
9 + 2
100 98
9 + 3
100 97
9 + + 99
100 1
9 + 100
100 100 0
f (10 9 + 1,
1 2 3 99 100
but since that would require calculating and summing 100 numbers, most of them very large, we
understand that he had to resort to a shortcut.
We can derive Mr. Waltons pattern from his combinatorial method by using more advanced
combinatorics. Mr. Waltons method gives that
n
i i 9
n n ni
f (10 ) = 1 + .
i=1
Consider the product i ni .

(n 1 )!
i = i --------------------- = ----------------------------------- = n ----------------------------------- = n
n n! n! n 1
i i!(n i)! (i 1 )!(n i)! (i 1 )!(n i)! i 1
So
n n
1
(n 1 ) j
n
n 1 n i

n 1

n j n1
f (10 ) = 1 + 9 = 1+n (1) (9) = 1 + n(1 + 9)
i 1 j
i=1 j=0
n n1
where j = i 1 and the last step is by the binomial theorem. Thus, f (10 ) = 1 + n10 .
7
5/2/14. For an isosceles triangle ABC where AB = AC , it is possi- A
ble to construct, using only compass and straightedge, an isosceles
triangle PQR where PQ = PR such that triangle PQR is similar
to triangle ABC , point P is in the interior of line segment AC ,
P
point Q is in the interior of line segment AB , and point R is in the
Q
interior of line segment BC . Describe one method of performing
such a construction. Your method should work on every isosceles
triangle ABC , except that you may choose an upper limit or lower B R C
limit on the size of angle BAC .
Comments: This problem was devised by George Berzsenyi, the founder and writer of the
USAMTS.
The graders of this problem noticed that few of the submitted constructions bothered being
precise about the upper limit on the size of BAC (no construction needed a lower limit). Rather
than calculating the largest angle for which the construction would work, most participants
merely declared that the angle was acute, which did t the letter of the problem. Thus, the graders
decided that the precise upper limit was worth zero points.
Solution 1 to 5/2/14 by Eline Boghaert (11/NY):

1. In order for this method of construction to work, the size of BAC should be less than 90.
2. Find the midpoints of AC , BC , and AB , and label them P , R , A
and Z , respectively. (To draw the midpoint of AC , set your com
line of
pass so that it is wider than half of AC s length, such as setting it reection
to AC s length. Draw a circle centered at point A . Using the
Z P
same compass radius, draw a circle centered at point C . These
circles intersect at two points: connect these points of intersection
Q
with a line segment. This line segment is the perpendicular bisec
tor of AC ; the midpoint of AC is where the perpendicular bisec-

tor intersects AC . Construct midpoints for BC and AB by the B R C
same method.) Connect the midpoints P , R , and Z with line segments. A triangle RPZ simi
lar to the larger triangle ABC is created, with PRZ congruent to BAC .
3. Draw the perpendicular bisector to PR (same process as before). Since the line segment con
necting any two midpoints of sides of a triangle is parallel to the side it does not intersect, PR
is parallel to AB . Thus, the perpendicular bisector of PR is also perpendicular to AB .
4. Reect RPZ through the perpendicular bisector of PR . PR reects onto itself. To reect
point Z , copy the distance between Z and line of reection to the other side of the line of
reection. Since AB is perpendicular to the line of reection, point Z s reection, call it Q ,
will also be on AB . Draw the line segments PQ and RQ to complete the triangle PQR .
Since PQR is a reection of RPZ , it is similar to ABC , with QPR congruent to BAC .
Comment from the solutions editor: Given points R and Z in Ms. Boghaerts construction,
there are many ways to nd point Q . For example, RQ = RB , so we could draw a circle of
radius RB centered at R . Many participants found different ways to construct this triangle.
If the size of BAC were more than 90, then QPR and Q
RPC together would measure more than 180, so point Q , A More than
P 180
though still on line AB , would be on the wrong side of A .
B C
R
Solution 2 for 5/2/14 by Anton Kriksunov (11/NY):
To construct the triangle, I rst made a bisection of angle ABC A
and another bisection of angle BAC . Point P is the intersection of
the bisector of angle ABC and the side AC . Next at point P I copied 1 1
half of angle BAC onto both sides of BP , creating angles QPB and
RPB , where Q is where the upper ray intersects AB and R is where P
Q 1
the lower ray intersects BC . Then I drew line segment QR to nish 1
2
triangle PQR . It is similar to triangle ABC because both are isosce
2
les and have congruent vertex angles, which means all their angles are
congruent. B R C
This method works well for all acute and right isosceles triangles, but there is a problem with
obtuse ones. After their vertex angles reach a certain amount, the constructed triangle is no longer
contained within the original triangle, so the construction does not fulll the conditions of the
problem.
This similar triangle is just on the border of A=Q
the original triangle: if its vertex angle
x
increases any more, point Q will move outside y P
x/2
the original triangle. This then is the limit for
x/2
the size of angle BAC . To nd the angle, I set y
y/2
up two equations: y/2 y
x + y + y = 180 from triangle ABC , and C
B R
x + y 2 + x 2 = 180 from triangle ABP .
Solving gives x = 108 and y = 36 . So the restrictions on angle BAC are that
0 < m BAC < 108 .
9
Solution 3 for 5/2/14:
From the solutions editor: Another method for constructing triangle PQR is by constructing a
triangle that is the right shape and right orientation and rescaling it to the proper size by projec
tion. Alas, the example of this method selected by the graders did not have the solvers signature
on the cover sheet giving permission to use his name. It would be improper to use his solution
without giving credit. So I will write my own version of a projection construction.
Draw a point Q anywhere on the interior of side AB . Draw a A
point P anywhere on the interior of side AC except that BQP 54
must be larger than ABC . Locate point R in the interior of P
BQP as follows. We draw a ray from point Q so that the angle Q 63 54
between the ray and QP copies ABC and draw another ray from
point P so that the angle between that ray and QP copies BAC . R
63 63
Point R is the intersection of those two rays.
B C
Because BQP is larger than RQP (which is congruent to A
ABC ), ray QR is in the interior of BQP . We can also show
that CPR is congruent to angle AQP , so ray PR is in the P
interior of CPQ . Therefore, point R is in the interior of BAC . Q P
If point R happens to be on side BC , we are done, but more Q

R
likely it is above or below that side. In that case, draw ray AR . Since
the ray is in the interior of BAC , it intersects side BC , in the interior B R C
of that side. Call the point of intersection R . Draw a line through point
R that is parallel to QR . Let Q be the point where that line and line AB intersect. Draw a line
through point R that is parallel to PR . Let P be the point where that line and line AC intersect.
Draw line segment QP .
Triangle PQR is similar to triangle PQR because it is a projection about point A of trian
gle PQR. Thus, it is similar to triangle ABC . For those unfamilar with the properties of projec
tion, note that triangle AQR is similar to triangle AQR and triangle APR is similar to triangle
APR , both in the same ratio of dimensions. Since QRP QRP and
PR QR = PR QR , the triangles PQR and PQR are similar.
Because AQR is larger than ABC , point R is lower than point Q when viewing side BC
as the base of triangle ABC . So Q is above side BC and below point A , so it is in the interior of
side AB. Tracing angles around the triangle gives that APR is congruent to BQP , which is
larger than ACB , so P is likewise in the interior of side AC .
This construction works for all triangles, regardless of the size of BAC .
10
solutions edited by Dr. Erin Schram
1/3/14. The integer n, between 10000 and 99999, is abcde when written in decimal notation.
The digit a is the remainder when n is divided by 2, the digit b is the remainder when n is
divided by 3, the digit c is the remainder when n is divided by 4, the digit d is the remainder
when n is divided by 5, and the digit e is the remainder when n is divided by 6. Find n.
Comment: This problem is from a 1966 issue of Abacus, a Hungarian mathematics journal for
middle school students.
Solution for 1/3/14 by Tamara Broderick (12/OH):

n = abcde .
a = n mod 2 ; therefore, a is equal to 0 or 1, but since n 10000 , a is 1, and n is odd.
Given this restriction, since e = n mod 6 , it may be only {1, 3, 5} . When we divide e by 3,
we get the same remainder as when we divide n by 3, so b = e mod 3 . Also, because d is the
remainder when n is divided by 5, d is determined entirely by e, the last digit of n.
e 1 3 5
b 1 0 2
d 1 3 0
The remainder, c, when n is divided by 4 is determined by the last two digits of n (since
100 mod 4 = 0 ). So c = de mod 4 .
de 11 33 05
c 3 1 1
Therefore, we have three possibilities for n:
a 1 1 1
b 1 0 2
c 3 1 1
d 1 3 0
e 1 3 5
The sum of the digits of n is congruent to n mod 3. Only in the rst column of the above table is
the sum of the digits, reduced mod 3, equal to b: (1 + 1 + 3 + 1 + 1) mod 3 = 1 . So n = 11311 .
1
2/3/14. Given positive integers p, u, and v such that u 2 + 2v 2 = p , determine, in terms of u and
v, integers m and n such that 3m 2 2mn + 3n 2 = 24 p . (It is known that if p is any prime
number congruent to 1 or 3 modulo 8, then we can nd integers u and v such that
u 2 + 2v 2 = p .)
Comments: This problem was developed by Dr. Robert Ward, a retired mathematician living in
Maryland and active in the Ask Dr. Math program. It was inspired by Problem 2 for Grade 8 in the
1992 Spring Mathematical Competition of Bulgaria, which was called to our attention by the late
Professor Ljubomir Davidov of Soa.
Solution 1 for 2/3/14 by Subrahmanya Krishnamoorthy (10/NY):

We have u 2 + 2v 2 = p and 3m 2 2mn + 3n 2 = 24 p . Dividing the latter equation through
out by three and multiplying the former equation throughout by eight, we have
2
m 2 --- mn + n 2 = 8 p = 8u 2 + 16v 2 .
3
Splitting the n 2 and rearranging gives
2 1 8
m 2 --- mn + --- n 2 + --- n 2 = 16v 2 + 8u 2 .
3 9 9
This groups as
2 2
m n--- + 8 n--- = ( 4v ) 2 + 8u 2 .
3 3
Now we can see that the partitioning was wise, because setting the insides of the leftmost squares
on each side equal and the insides of the rightmost squares on each side equal gives an easy way
to make both sides equal:
n n
m --- = 4v and --- = u.
3 3
Thus, n = 3u and m = 4v + u . Because both u and v are integers, m and n must also be integers
by additive closure of integers.
Solution 2 for 2/3/14 by William Carlson (10/KS):

First, recognize that p, u, and v are integers that can be positive, and m and n are integers. It is
also given that u 2 + 2v 2 = p and 3m 2 2mn + 3n 2 = 24 p . So
2
m 2 --- mn + n 2 = 8 p = 8u 2 + 16v 2 .
3
To go about nding m and n in terms of u and v, I set up a table. For my p values I selected
several primes congruent to 1 or 3 mod 8. I found u, v, m, and n values to go with those p values.
At rst, I didnt know which of those paired m and n values were m or n, because they are inter
changeable in m 2 (2 3)mn + n 2 = 8 p , but they are not interchangeable when I was trying to
nd fuchtions f ( u, v) and g ( u, v) such that m = f ( u, v) and n = g ( u, v) . The table given here
has the m and n values in the proper order for my argument, though I had to switch the order of
one pair in my original table.
p 3 11 17 19 41 43
u 1 3 3 1 3 5
v 1 1 2 3 4 3
m 3 9 9 3 9 15
n 5 7 11 13 19 17
From this table, it seemed that m = 3u .

So I put 3u in for m in the equation m 2 (2 3)mn + n 2 = 8u 2 + 16v 2 to nd out what n
would be. This gave
( 3u ) 2 (2 3)(3u)n + n 2 = 8u 2 + 16v 2
9u 2 2un + n 2 = 8u 2 + 16v 2
u 2 2un + n 2 = 16v 2
(n u) 2 = ( 4v ) 2
n u = 4v
n = u 4v
Setting n = u + 4v made all the ms and ns in the table work. Besides, if m = 3u and
n = u + 4v are plugged back into the equation m 2 (2 3)mn + n 2 = 8u 2 + 16v 2 , the result is
8u 2 + 16v 2 = 8u 2 + 16v 2 .
One thing left: m and n have to be integers. If u and v are integers, and m = 3u and
n = u + 4v , then m and n must also be integers.
So m = 3u and n = u + 4v .
Solution 3 for 2/3/14 by Alexander Yee (9/CA):

Given 3m 2 2mn + 3n 2 = 24 p and u 2 + 2v 2 = p .
First, use the quadratic formula to solve 3m 2 2mn + 3n 2 = 24 (u 2 + 2v 2 ) for m in terms of
n, u, and v.
2n (2n) 2 4 ( 3 )(3n 2 24u 2 48v 2 )
m = ---------------------------------------------------------------------------------------------------
2(3)
n 2 2n 2 + 18u 2 + 36v 2
= -----------------------------------------------------------------
3
In order to get an integer, we must rst eliminate all radicals. We need to select a value for n that
when subsituted into the discriminant, 2n 2 + 18u 2 + 36v 2 , will give a square of an integer. We
can do that by setting n to 3u .
( 3u ) 2 2 ( 3u) 2 + 18u 2 + 36v 2

m = ---------------------------------------------------------------------------------
3
3u 2 36v 2
= ------------------------------
3
3u 12v
= --------------------
3
= u 4v
We have found two solutions. There are eight in all. Setting n to 3u also eliminates the radical
and gives m = u 4v . Notice that the equation 3m 2 2mn + 3n 2 = 24 p is symmetrical in m
and n. Therefore, we can switch the values of m and n and still have a solution. That gives eight
solutions for (m, n) : (u + 4v, 3u) , (u 4v, 3u) , ( u + 4v, 3u) , ( u 4v, 3u) ,
(3u, u + 4v) , (3u, u 4v) , (3u, u + 4v) , and (3u, u 4v) .
Solution 4 for 2/3/14 by Mircea (Bobby) Georgescu (9/CA):

Notation: x y means that x is a factor of y.
We are given 3m 2 2mn + 3n 2 = 24 p (equation 1). So 2mn = 3m 2 + 3n 2 24 p =
3(m 2 + n 2 8 p) . Therefore, 3 2mn . Since 3 is prime and 3 is not a factor of 2, 3 m and/or 3 n .
Let us consider the case where 3 m (the 3 n case has a similar solution). We set m = 3k for
some integer k. Substituting m = 3k in equation 1 gives 27k 2 6kn + 3n 2 = 24 p , which sim
plies to 9k 2 2kn + n 2 = 8 p (equation 2).
Equation 2 is equivalent to k 2 2kn + n 2 = 8( p k 2 ) , so 8 (k 2 2kn + n 2 ) . This is
8 (k n) 2 , which means that 4 (k n) . So k n = 4 j for some integer j.
We substitute n = k 4 j and equation 2 becomes 8k 2 + 16 j 2 = 8 p , which is equivalent to
k 2 + 2 j 2 = p (equation 3). We know that u 2 + 2v 2 = p , so setting k = u and j = v gives a
solution to equation 3. This gives m = 3u and n = u 4v as a solution to equation 1.
m
3/3/14. Determine, with proof, the rational number ---- that equals
n
1 1 1 1
--------------------------- + --------------------------- + --------------------------- + + --------------------------------------------------------------------------------------------------- .
1 2+2 1 2 3+3 2 3 4+4 3 4012008 4012009 + 4012009 4012008
Comment: This problem was inspired by a similar problem in Matlap, Transylvanias outstand
ing Hungarian-language mathematics journal for middle school and high school students.
Solution 1 for 3/3/14 by Eric Stansifer (10/FL):

All of the terms of the expression given were of the form
1
---------------------------------------------------.
a a + 1 + (a + 1) a
I decided to simplify this into an easier-to-manipulate form, so I multiplied the numerator and
denominator by the conjugate of the denominator:
1 1 a a + 1 (a + 1) a
--------------------------------------------------- = --------------------------------------------------- -------------------------------------------------
a a + 1 + (a + 1) a a a + 1 + (a + 1) a a a + 1 (a + 1) a
a a + 1 (a + 1) a
= --------------------------------------------------------
-
a ( a + 1 ) + ( a + 1 )2( a )
2
a a+1 (a + 1) a
= --------------------------------------------------------
- --------------------------------------------------------
-
a ( a + 1 ) + ( a + 1 ) ( a ) a ( a + 1 ) + ( a + 1 )2( a )
2 2 2
a+1 a
= ---------------------------------------------2- -------------------------------
-
a(a + 1) (a + 1) a a(a + 1) 2
a+1 a
= -------------------- -------
( a + 1 ) a
1 1
= ------- ----------------
a a+1
When each term of the given expression is modied in the above fashion, it yields a telescoping
series which simplies to a rational form. The modied series is
1
- ------- + ------- ------- + ------- ------- + + ------------------------- ------------------------- .
------ 1 1 1 1 1 1 1
1 2 2 3 3 4 4012008 4012009
Canceling out all but the rst and last terms yields
1 1 1 1 2002
------- ------------------------- = --- ------------ = ------------ .
1 4012009 1 2003 2003
The answer is 2002/2003.
5
Solution 2 for 3/3/14 by Hyun-Soo Kim (10/NJ):
The sum given in the problem can be written in summation notation as
k
--------------------------------------------------
1
-
n=1
n n + 1 + (n + 1) n
where in this case k = 4012008 .

Upon calculating the sum for smaller values of k, such as 2 and 3, a pattern was established.
The pattern is that
k
n--------------------------------------------------
-
1 1
= 1 ---------------- .
n=1
n + 1 + (n + 1 ) n k+1
We will prove by induction that the above conjecture is true.
Base case:
k = 1.
1
--------------------------------------------------
-
1 1 1 2 2 1 1
= --------------------------- = ---------------- = ---------------- = 1 ------- = 1 ---------------- .
n=1
n n + 1 + (n + 1 ) n 1 2+2 1 2+ 2 2 2 k +
1
Assuming that the induction hypothesis holds for a given value k, prove that it holds for the value
k + 1:
k+1 k
--------------------------------------------------
- = --------------------------------------------------
- + -------------------------------------------------------------------------
1 1 1
n=1
n n + 1 + (n + 1) n n=1
n n + 1 + (n + 1) n (k + 1) k + 2 + (k + 2) k + 1
1 1
= 1 ---------------- + -------------------------------------------------------------------------
k + 1 (k + 1) k + 2 + (k + 2) k + 1
( 1 ) ( ( k + 1 ) k + 2 + ( k + 2 ) k + 1 ) + k + 1
= 1 + ----------------------------------------------------------------------------------------------------------------
( k + 1)((k + 1) k + 2 + (k + 2) k + 1)
( k + 1 ) k + 1 ( k + 1 ) k + 2
= 1 + -------------------------------------------------------------------------------------------
(k + 1) k + 1 k + 2 + (k + 1)(k + 2)
( 1 ) ( k + 1 ) ( k + 1 + k + 2 )
= 1 + ----------------------------------------------------------------------------------
( k + 2)(k + 1)( k + 1 + k + 2)
1
= 1 ----------------
k+2
1
= 1 ------------------------------
(k + 1) + 1
So by induction, our conjecture is true. The rational number that equals the given sum is:
4012008
----------------------------------------------------
1
- = 1 ---------------------------------- = 1 ------------ = ------------.
2002

1 1
n n + 1 + (n + 1) n 4012008 + 1 2003 2003
n=1
6
4/3/14. The vertices of a cube have coordinates (0, 0, 0) , (0, 0, 4) , (0, 4, 0) , (0, 4, 4) ,
(4, 0, 0) , (4, 0, 4) , (4, 4, 0) , and (4, 4, 4) . A plane cuts the edges of this cube at the points
(0, 2, 0) , (1, 0, 0) , (1, 4, 4) , and two other points. Find the coordinates of the other two
points.
Comments: This problem was contributed by Prof. George Berzsenyi, the founder and problem
editor of the USAMTS. We are thankful for all the ways in which Prof. Berzsenyi supports the
USAMTS.
Solution 1 for 4/3/14 by Kevin Lin (10/OH):

The general equation of the plane is aX + bY + cZ = d .
Put in the three points on the plane: (0, 2, 0) , (1, 0, 0) , and (1, 4, 4) :
(a)(0) + (b)(2) + (c)(0) = d
(a)(1) + (b)(0) + (c)(0) = d
(a)(1) + (b)(4) + (c)(4) = d
Solve for a, b, and c in terms of d. We have
d d
a = d, b = ---, c = --- .
2 2
Setting d = 2 , the equation of the plane is
2X + Y Z = 2.
The cube has 12 edges, each dened by a pair of constraints:
Edge 1 X = 0 & Y = 0
Edge 2 X = 0 & Y = 4
Edge 3 X = 0 & Z = 0 The plane cuts this edge at known point (0, 2, 0)
Edge 4 X = 0 & Z = 4
Edge 5 X = 4 & Y = 0
Edge 6 X = 4 & Y = 4
Edge 7 X = 4 & Z = 0
Edge 8 X = 4 & Z = 4
Edge 9 Y = 0 & Z = 0 The plane cuts this edge at known point (1, 0, 0)
Edge 10 Y = 0 & Z = 4
Edge 11 Y = 4 & Z = 0 The plane cuts this edge at known point (1, 4, 4)
Edge 12 Y = 4 & Z = 4
If an edge of the cube is cut by the plane, every coordinate of the intersection point has to be
between 0 and 4, inclusive, or that point is not on the cube. Checking the intersection points of the
nine edges without known points, we nd that the plane cuts edge 2 at (0, 4, 2) and cuts edge 10
at (3, 0, 4) .
7
Solution 2 for 4/3/14 by Karl Jiang (12/FL):
To the left is a gure of the cube suggested. The
three known intersection points of the plane and the
G 1 C (1,4,4) edges are A, B, and C, shown labeled with their
y 2 coordinates. If two intersection points share a face
E of the cube, the line segment drawn between them
is the intersection of the plane and that face. We
4 can see that for faces opposite each other on the
cube, the slopes of those plane-face intersection
z segments are going to be the same. In this way, we
can use the proportions of those slopes to get the
F 2 D
remaining two plane-edge intersection segments.
B (0,2,0)
4 Line segment AB is the intersection of the
2 plane and a face. Naming the vertex with coordi
x nates (0, 0, 0) as H, we see that BH = 2 and
1
A (1,0,0) AH = 1 . The opposite face must contain a right
H
triangle CDF similar to triangle BAH, with point F
directly below point C at coordinates (1, 0, 0) and
with line segment CD being the intersection of the plane and the face. CF = 4 and DF / CF =
AH / BH = 1/2, so DF = 2 and we can conclude that D = (3, 0, 4) .
Line segment AD is the intersection of the plane and a face. We see that AF = 4 and
DF = 2 . The opposite face must contain a right triangle CEG similar to triangle DAF, with
point G at the vertex (0, 4, 4) and with line segment CE being the intersection of the plane and
the face. CG = 1 and EG / CG = AF / DF = 2, so EG = 2 and we can conclude that
E = (0, 4, 2)
.
Answer: The two other intersection points are (3, 0, 4) and (0, 4, 2)
.
8
5/3/14. A fudgeake is a planar fractal gure with 120 rota-
tional symmetry such that three identical fudgeakes in the
same orientation t together without gaps to form a larger
fudgeake with its orientation 30 clockwise of the smaller 1
fudgeakes orientation, as shown on the right. If the distance
between the centers of the original three fudgeakes is 1,
what is the area of one of those three fudgeakes? Justify
your answer.
Comments: This problem was written by Dr. Erin Schram, the

solutions editor of the USAMTS. The fudgeake came from The
Fractal Geometry of Nature by Benoit Mandelbrot, who credits
a 1970 paper by Davis and Knuth in the Journal of Recreational Mathematics.
Solution 1 for 5/3/14 by Guy David (9/FL):

We can rst connect the centers of the three fudgeakes,
creating an equilateral triangle with side length 1. We note the
region that the triangle cuts out from one fudgeake, which is
shaded in red in the diagram to the left.
As the fudgeakes have 120

rotational symmetry, three copies of
this region occur in the fudgeake.
We can put a fourth fudgeake above the others. It ts the one to its lower left just like the
other two fudgeakes t together, so their centers are 1 apart. Likewise the centers of that fudge-
ake and the fudgeake to its lower right are 1 apart. This gives us a different equilateral triangle
that cuts out a different region, shaded in blue.
This blue region can be found in three places within the
fudgeake. Three copies of the red region and three copies of the
blue region combine to form a full fudgeake. The sum of the
areas of all three red pieces is equal to the area of the equilateral
triangle with side 1, and the sum of the areas of all three blue
pieces is the same. Therefore, the area of one fudgeake is twice
3 3
the area of an equilateral triangle with side 1: 2 ------- = ------- .
4 2
= +
9
Solution 2 for 5/3/14 by Boris Hanin (10/ID):
Observe that we can completely tesselate any part of the plane using the fractal. Also, by con-
necting the centers of adjacent fractals, we get a grid made of equilateral triangles with side 1.
Consider the sequence of nested rhombuses shown below.
An n n rhombus is contained in ( n + 1 ) 2 fractals and contains ( n 1 ) 2 fractals. This means

that n 2 fractals arranged in a rhombus pattern contain an ( n 1 ) ( n 1 ) rhombus and are con-
tained in an ( n + 1 ) ( n + 1 ) rhombus. If we take those two rhombuses, their areas give us an
upper bound and a lower bound for the area of the n 2 fractals.
The area of a rhombus with side a is a 2 sin where is the angle between two adjacent
sides. Observing that the grid of equilateral triangles between the centers of the fractals gives
a2 3
= 60 , the area of a rhombus with side a is ------------- . Denoting the area of a fractal as A, we get
2
( n 1 )2 3 ( n + 1 )2 3
-------------------------- < n 2 A < --------------------------- .
2 2
Equivalently,
( n 1 )2 3 ( n + 1 )2 3
-------------------------- < A < --------------------------
-.
2n 2 2n 2
( n 1 )2 3 ( n + 1 )2 3 3
Observe that lim -------------------------- = lim --------------------------
- = - , so letting n and applying the
------
n 2n 2 n 2n 2 2
3
squeeze theorem, we get A = ------- .
2
10
Comment from the solutions editor: One rare type of solution, such as the one below, requires
nding a reference that describes the construction of a fudgeake, such as Mandelbrots The
Fractal Geometry of Nature. Outside references like this are perfectly acceptable, though a solu
tion should always cite any special references used.
Solution 3 for 5/3/14 by Jason Allen (12/MN):

The fudgeake is constructed as follows. You begin with a regular hexa
gon like the one pictured at right. Then for each of the six sides, you replace
the side with the motif at right, alternating directions so that the motif bends
in at one side, out at the adjacent side, in again for the third side, out again for
the fourth side, and so on. The motif has a 120 angle and its missing base is
as long as a side of the hexagon. Repeat this again and again with the result- Regular Hexagon
ing gure, scaling the motif smaller at each iteration, and always starting
with the same vertex and with the same direction at each iteration. Three iter-
Motif
ations are shown below.
Start
point
Iteration 1 Iteration 2 Iteration 3

In every iteration, we are adding a certain number of motif triangles to the gure and we are
subtracting the same number of triangles with the same area from the gure, so the area in each
successive iteration remains constant. Note that the center does not move between iterations.
If we transform two identical hexagons that share a side into fudgeakes, if we coordinate
their motif patterns, the resulting fudgeakes t together the same way that the fudgeakes in this
problem t together. Since the fudgeakes have the same area as the initial hexagons, we just need
to nd the area of those hexagons.
Label the centers of the fudgeakes as A and B. We are given
that AB = 1 and we deduced that A and B are the centers of
A 1 B congruent regular hexagons that share a side and have the same
area as the fudgeakes. Congruent hexagons have congruent
apothems, so AD = BD = 1 2 and line segment AD is per
perpendicular to side CE . Triangle ACE is equilateral, since the
C hexagon is regular, so mACD = 60 . Triangle ACD is a
A D B 30-60-90 triangle, so CD = 3 6 . The area of triangle ACD is
1/2 1/2 (1 2) ( 3 6) (1 2) = 3 24 . The hexagon centered at
E point A is made up of 12 triangles congruent to triangle ACD, so
the area of the hexagon is 3 2 . Therefore, the area of one of
3
the fudgeakes is ------- .
2
11
solutions edited by Dr. Erin Schram
1/4/14. The sequence of letters TAGC is written in succession 55 times on a strip, as shown
below. The strip is to be cut into segments between letters, leaving strings of letters on each
segment, which we will call words. For example, a cut after the rst G, after the second T, and
after the second C would yield the words TAG, CT, and AGC. At most how many distinct
words could be found if the entire strip were cut? Justify your answer.
TAGCTAGCTAG ... CTAGC
Comment: This problem was inspired by Problem 1759 of Sndor Rkas excellent collection of
2000 problems in elementary mathematics (2000 Feladat az Elemi Matematika Krbl, Typotex,
Budapest, 2000).
Solution 1 to 1/4/14 by Becky Wright (9/UT):

I will assume that a word must have at least one letter. If that is not the case, please add one
word to my total to account for the one word with no letters in it.
I claim that the maximum number of distinct words is 40.
Forty distinct words can be constructed as follows. Make three cuts to cut the original strip of
220 letters into four consecutive segments containing 55 letters each. Next cut each of these seg
ments in the same way: make the rst cut after the rst letter, make the second cut after two more
letters, make the third cut after three more letters, and so on until the ninth cut after nine more let
ters. The segment after the last cut has ten letters, since 1 + 2 + 3 + + 9 + 10 = 55 . So each
segment of 55 letters has been cut into ten words, and this gives a total of 40 words from the orig
inal strip of 220 letters. Next, note that these 40 words contain no repetitions. These 40 words
contain four words of length one, four words of length two, and so on up to four words of length
ten. For any given length, the four words of that length are offset from the previous word of that
length by 55 letters. Since 55 3 (mod 4 ) , then for a given length, the second word of that length
will start three letters later in the sequence than the rst word, the third word will be six (which is
the same as two) letters later in the sequence, and the fourth word will be nine (which is the same
as one) letters later in the sequence. Therefore, the four words must start with different letters and
are distinct.
To see that no more than 40 words is possible, consider the following. First of all, there are at
most four possible distinct words of a given length. The four letters repeat in the same order
everywhere on the strip, so the starting letter and the word length completely determine the word.
Since there are only four possible starting letters, this forces no more than four possible words of
a given length. The way to construct the largest collection of words is to use words of shortest
length as much as possible, which is four times for each length. So the 40 shortest distinct words
use four words each of lengths one through ten. This uses exactly 220 letters. Hence, any collec
tion of more than 40 distinct words will have to use more than 220 letters, so 40 is the maximum.
Solution 2 for 1/4/14 by Alan Gostin (11/TX):

There are four distinct words for each length of word, one starting with each of the four let
ters. Obviously, since the most words are wanted, it is best to use the shortest words possible and
not to waste any words on repeating words. The 40 smallest words, four of each length from one
to ten, add to 220 letters. Thus, the largest number of words that might be possible is 40.
However, it is as of yet uncertain whether these words can all be t into the 220 letters. The
solution is simple, though. By cutting one string of each of the ten lengths off the beginning, ten
distinct strings are created using 55 letters. If this is repeated, taking special care to cut in the
same order as the rst cutting, ten more strings are obtained. These strings must be different from
the rst ten, because instead of starting with the rst letter T, this set of words starts with the 56th
letter C. The rst letter of each new set must be different because the greatest common factor of
55 and 4 is 1. By following this method of cutting, 40 distinct strings may be created, as illustrated
by the example below.
Example cutting (First set of ten words on the rst line)

T AG CTA GCTA GCTAG CTAGCT AGCTAGC TAGCTAGC TAGCTAGCT AGCTAGCTAG
C TA GCT AGCT AGCTA GCTAGC TAGCTAG CTAGCTAG CTAGCTAGC TAGCTAGCTA
G CT AGC TAGC TAGCT AGCTAG CTAGCTA GCTAGCTA GCTAGCTAG CTAGCTAGCT
A GC TAG CTAG CTAGC TAGCTA GCTAGCT AGCTAGCT AGCTAGCTA GCTAGCTAGC
Solution 3 for 1/4/14 by Gleb Kuznetsov (9/UT):

By analyzing the given series of letters, it is evident that since the order of letters doesnt
change, there are only four distinct words for each size of word, each starting with a different let
ter.
By listing out the letters, and marking where the strip would be cut, I found that alternating
between the four words of an odd size and the four words of an even size creates distinct words of
each length. Once all combinations of those sizes are used, a new pair of sizes is used; i.e., after
all 1 and 2 lengths are cut, move on to 3 and 4, and so on.
Since there are 4 55 = 220 letters on the strip, the pattern can be continued until that
many letters are used. And since there are four possibilities for each word size, the number of let
ters used to make all words up to that size can be calculated by 4 ( 1 + 2 + 3 + ) , increasing the
number of terms in the sum until all 220 letters are used. Since 4 ( 1 + 2 + 3 + + 10) = 220 ,
ten word sizes are used, and 40 distinct words are made.
1&2: T|AG|C|TA|G|CT|A|GC|
3&4: TAG|CTAG|CTA|GCTA|GCT|AGCT|AGC|TAGC|
5&6: TAGCT|AGCTAG|CTAGC|TAGCTA|GCTAG|CTAGCT|AGCTA|GCTAGC|
7&8: TAGCTAG|CTAGCTAG|CTAGCTA|GCTAGCTA|GCTAGCT|AGCTAGCT|
AGCTAGC|TAGCTAGC|
9&10: TAGCTAGCT|AGCTAGCTAG|CTAGCTAGC|TAGCTAGCTA|GCTAGCTAG|
CTAGCTAGCT|AGCTAGCTA|GCTAGCTAGC
If any other word sizes are tried, the new-size word would have to replace two or more smaller
words to t into the 220 letters. Since that would make fewer words, 40 words is the maximum.
2/4/14. We dene the number s as

---------------
1 1 1 1 1
s = - = --- + ------ + --------- + ------------ + = 0.12232424 .
10 i 1 9 99 999 9999
i=1
We can determine the nth digit right of the decimal point of s without summing the entire in
nite series because after summing the rst n terms of the series, the rest of the series sums to
less than 2 10 n + 1 . Determine the smallest prime number p for which the pth digit right of
the decimal point of s is greater than 2. Justify your answer.
Comments: Igor Zhinitsky of New York, one of the contestants in the USAMTS, submitted the
original version of this problem. We are thankful for his contribution and for his enthusiasm to
participate in the USAMTS in more than one way. However, in the future, we are going to delay
using any student-submitted problems until after the student has graduated from high school, in
order to avoid a contestant submitting a solution to a problem that he or she wrote, as happened
with Mr. Zhinitsky. However, this does give me an opportunity to present the problem writers
own solution as an ofcial solution.
Solution 1 to 2/4/14 by Igor Zhinitsky (12/NY):

The answer to this problem is 47.
All the terms of the series have only 1s and 0s in all the digits after the decimal place. The
nth digit of s can, thus, be found by calculating the number of terms that have a 1 in the nth place.
1/9 has a 1 in every place; thus, every place number divisible by 1 has a 1 from that term. 1/99 has
a 1 in every other place; thus, every place number divisible by 2 has a 1 from that term. 1/999 has
a 1 in every third place; thus, every place number divisible by 3 has a 1 from that term. And so
forth.
Summing these values, we see that the nth digit is the number of factors of n, disregarding for
the moment the possibility of 10 or more factors. More precisely, the nth digit D ( n ) in general is
exactly equal to
F (n + 1) + F (n + 2) + F (n + 3) +
--------------------------------------------------------------------
D ( n ) = F ( n ) + -----------------------------------------------------------------------------------------------------
10 mod 10
10
-

where F ( n ) denotes the number of factors of n and x denotes the greatest integer less than or
equal to x.
Clearly, the number of factors of any prime number is two; thus, the pth digit will be 2 unless
the greatest integer part of the formula for D ( n ) is non-zero, which will happen for the rst time
when p + 1 has 10 or more factors. 48 is the rst number that has 10 or more factors; thus, the
47th digit will be F ( 47 ) + 10 +0
--------------- , which comes out to 3, because 47 is prime. Thus, 47 is the
10
solution.
Solution 2 to 2/4/14 by Hilary Palevsky (12/PA):

When written in decimal form,
1
--- = 0.111111 = 0.1
9
1
------ = 0.010101 = 0.01
99
1
--------- = 0.001001 = 0.001
999
1
and so on. This can be generalized that ---------------
i
- yields a number such that every nth digit right of
10 1
the decimal point is 1 for all positive integers n divisible by i and all other digits are 0s.

From this, it can be seen that in the number s such that s =
1
---------------
i
- , the value of the
i = 1 10
1
nth digit to the right of the decimal point is primarily determined by the number of factors of n.
The sixth digit to the right of the decimal point, for example, is 4, because 6 has four factors: 1, 2,
3, and 6.
Thus, it would seem that when n is prime, the nth digit could not be greater than 2, since by
the denition of a prime number, there cannot be more than two factors (1 and n). The exception,
however, occurs when n + 1 has 10 or more factors, which would carry the digit from the tens
place and add it to the digit 2 from the number ns own factors.
The smallest integer with 10 or more factors is 48, which has exactly ten factors. Since the
number before 48, 47, is a prime number, the 47th digit to the right of the decimal point will be
2 + 1 = 3.
Hence, 47 is the smallest prime number p for which the pth digit right of the decimal point of
s is greater than 2.
3/4/14. Find the real-numbered solution to the equation below and demonstrate that it is unique.
36 9
------- + ------- = 42 9 x y
x y
Comment: This problem was dug up from deep in Prof. George Berzsenyis archives, which
means we have no-one to blame but ourselves for the aw in it. There are two legitimate ways to
interpret the problem, and one of those interpretations has ve solutions instead of one. So we
graded the problem as if it were the participants choice out of the following two problems:
3/4/14 version (a). Working solely in the real numbers, nd the solution to the equation below
and demonstrate that it is unique.
36 9
------- + ------- = 42 9 x y
x y
(In this version, the square root of a negative number is undened.)
3/4/14 version (b). Working in the complex numbers, nd all solutions to the equation below in
which both x and y are real numbers.
36 9
------- + ------- = 42 9 x y
x y
(In this version, the square root of a negative number n is ( n )i where i = 1 .)
We apologize for the confusion that the potential four additional solutions caused. However, we
are elated that some of the USAMTS participants were insightful enough to nd those four addi
tional solutions despite the contradicting statement about uniqueness.
Solution to 3/4/14 version (b) by Andrew Altheimer (10/NC):

First, get all the xs on one side and the ys on the other:
36 9
------- + 9 x = 42 ------- y (1)
x y
Pull out common factors on each side, including the 1 ( x) and 1 ( y) , and complete the
squares:
9 1
------- ( ( x ) 2 + 4 ) = ------- ( ( y ) 2 42 y + 9 )
x y
9 9 1 1
------- ( ( x ) 2 4 x + 4 ) + ------- ( 4 x ) = ------- ( ( y ) 2 6 y + 9 ) + ------- ( 36 y )
x x y y
9 1
------- ( ( x ) 2 4 x + 4 ) + 36 = ------- ( ( y ) 2 6 y + 9 ) + 36
x y
9 1
------- ( ( x ) 2 4 x + 4 ) = ------- ( ( y )2 6 y + 9 )
x y
9
1
------- ( x 2 ) 2 = ------- ( y 3 ) 2
x y
If x and y are positive so that their square roots are real numbers, then the left side would be
greater than or equal to zero while the right hand side would be less than or equal to zero. Since
the only overlap is zero, the equation can be solved by setting both sides equal to zero. This gives
x = 4 and y = 9 .
Although this nds a solution to the problem, it does not prove that no other solutions exist,
because we have not examined the case where x or y could be negative. (Neither is zero due to the
division by x and y .)
Both x and y cannot both be negative, because this would leave only one real term in the orig
inal equation, the 42, with no other nonzero real term to match it on the other side.
If x is negative while y is positive, then we have two imaginary terms, both on the left side in
equation (1) above. The only situation where this left side can match the real number on the right
36
side is when both imaginary terms cancel each other out: ------- + 9 x = 0 . That gives x = 4 .
x
9
Plugging in x = 4 into equation (1) gives 42 ------- y = 0 . Simplify it as ( y) 2 42 y + 9
y
= 0 and use the quadratic formula to nd y = 21 12 3 . Squaring that gives
y = 873 504 3 .
If y is negative while x is positive, then we again have two imaginary terms and again need
9
them to cancel out: ------- y = 0 . That gives y = 9 . Plugging this back into equation (1)
y
36
gives ------- + 9 x = 42 . Simplify it as 9( x) 2 42 x + 36 = 0 and use the quadratic formula to
x
7 13 62 14 13
nd x = ------------------- . Squaring that gives x = ---------------------------- .
3 9
62 14 13
This gives ve solutions to the equation: (4, 9) , ( 4, 873 504 3 ) , and ----------------------------, 9 .
9
Plugging these numbers back into the original equation then proves that all these are valid real-
number solutions to the equation. Although four of these solutions will form imaginary numbers,
these numbers cancel out and since the solutions themselves are real, they should be considered
real-numbered solutions.
Solution 1 to 3/4/14 version (a) by Kristina Simmons (12/WI):
The given equation is rearranged to be 9 x + ------- + y + ------- = 42.

4 9
x y
4 9
Substituting m = x + ------- and n = y + ------- gives 9m + n = 42 . Because x 0 and
x y
y 0 by denition, and x 0 and y 0 to avoid division by 0, m and n must both be posi
tive. Now the task is to nd the solution to 9m + n = 42 for which x and y are real numbers.
4
The equation m = x + ------- , or equivalently x m x + 4 = 0 , is quadratic in x with solu
x
m m 2
16
tions x = --------------------------------- . For there to be at least one real solution, we need m 2 16 0 , so
2
9 n n 2 36
m 4 , because m must be positive. Similarly, n = y + ------- gives y = ------------------------------- , so
y 2
n 6 . In other words, there is a real solution to the original equation only if m 4 and n 6 .
However, if m 4 and n 6 , then 9m + n 42 . Since 9m + n = 42 , this forces m = 4 and
4 4 2 16
n = 6 . Now we can nd x and y. Substituting for m, we get x = ------------------------------- = 2 , and substi
2
6 6 2 36
tuting for n, we get y = ------------------------------- = 3 . Therefore, the unique solution to the equation is
2
( x, y) = (4, 9) .
7
Solution 2 to 3/4/14 version (a) by Joanne Kong (11/NY):
This problem can be simplied by comparing arithmetic means with geometric means. For
two positive numbers a and b , the arithmetic mean of the numbers is greater the geometric mean,
a+b
or equal if the numbers are equal. ------------ ab , so a + b 2 ab .
2
36 9
The equation to be solved can be arranged into ------- + 9 x + ------- + y = 42 , and the sums
x y
36 9
------- + 9 x and ------- + y can be both viewed as twice an arithmetic mean. Every term is positive,
x y
since square roots are nonnegative and these square roots are not zero. So
------- + 9 x 2 ------- (9 x) = 2 324 = 36

36 36
x x
and
------- + y 2 ------- ( y) = 2 9 = 6.
9 9
y y
The sum of the minimum values, 36 and 6, is 42, so equality is necessary. This means that the two
36 9 36 36
terms inside each sum ------- + 9 x and ------- + y are equal. ------- = 9 x gives ------ = x x = x ,
x y x 9
9
so x = 4 . ------- = y gives 9 = y y = y , so y = 9 .
y
The solution is x = 4 and y = 9 .
8
Solution 3 to 3/4/14 version (a) by Daniel Walton (12/WA):
36 9
Start by substituting a and b for x and y , respectively. ------- + ------- = 42 9 x y
x y
36 9
becomes ------ + --- = 42 9 a b . Now multiply both sides by ab and rearrange to solve for b.
a b
36 b + 9 a = 42 ab 9 a 2 b ab 2
ab 2 + 9 a 2 b 42 ab + 36 b + 9 a = 0
( a )b 2 + (9 a 2 42 a + 36 )b + (9 a) = 0
Use the quadratic formula and simplify.
(9 a 2 42 a + 36 ) (9 a 2 42 a + 36 ) 2 4 ( a ) ( 9 a )
b = --------------------------------------------------------------------------------------------------------------------------------
2a
9 a 2 + 42 a 36 (81 a 4 756 a 3 + 2412 a 2 3024 a + 1296 ) (36 a 2 )
b = -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
2a
9 a 2 + 42 a 36 81a 4 756 a 3 + 2376a 2 3024 a + 1296
b = ---------------------------------------------------------------------------------------------------------------------------------------------------
2a
The discriminant 81 a 4 756 a 3 + 2376 a 2 3024 a + 1296 factors as 3 (a 2 ) 2 (3 a 2 16 a + 12 ) ,

so it is nonnegative whenever a = 2 or a 0.903 or a 4.43 . Remember that both a and b must
be positive, since each is a real square root and is used as a denominator. So the bounds on a are
a = 2 or 0 < a 0.903 or a 4.43 .
Now examine the value of b for those values of a. For 0 < a 0.903 , b < 0 . For a = 2 ,
b = 3 . For a 4.43 , b < 0 . By elimination, a must be 2.
Does it work? x = a 2 = 2 2 = 4 , y = b 2 = 3 2 = 9 .
36 9 ?
------- + ------- = 42 9 4 9
4 9
?
18 + 3 = 42 18 3

21 = 21 Yes!
Solution 4 to 3/4/14 version (a) by Ryan Hendrickson (12/MD):

Answer: x = 4 , y = 9 .
Proof of uniqueness:
36 9
------- + ------- = 42 9 x y
x y
36 9
------- + 9 x + ------- + y = 42
x y
f ( x ) + g ( y ) = 42
36 9
where f ( z ) = ------ + 9z and g ( z ) = --- + z for all real numbers z with z > 0 . (Thats all the
z z
domain we need, and it will make things easier.)
df 36 dg 9
-----( z ) = ------ + 9 and ------ ( z ) = ---- + 1 . Fortunately, we restricted the domain of the func
dz z 2 dz z2
tions to positive values of z to avoid the singularity at z = 0 . We nd the minimum values of
df dg
f ( z ) and g ( z ) by setting their derivatives to zero. -----( z ) = 0 gives z = 2 , and ------ ( z ) = 0
dz dz
gives z = 3 . We take the second derivatives to verify that both solutions are local minima:
d2 f 72 d2 f 18 2
------------2- ( 2 ) = ------ = 9 > 0 and ------------2- ( 3 ) = -----3- = --- > 0 so the extrema points are minima.
( dz ) 23 ( dz ) 3 3
At the boundaries of our domain, lim f (z) = + , lim f ( z ) = + , lim g ( z ) = + ,
z 0+ z + z 0+
and lim g ( z ) = + , so the local minima are absolute minima.
z +
36 9
f min = ------ + 9 2 = 36 and g min = --- + 3 = 6 , so the absolute minimum value of

2 3
f ( x) + g ( y) is 36 + 6 = 42 . This is the required value, and it only happens when x = 2
and y = 3 , which implies that x = 4 , y = 9 is the only solution.
10
4/4/14. Two overlapping triangles could divide the plane into up to eight regions, and three over
lapping triangles could divide the plane into up to twenty regions. Find, with proof, the maxi
mum number of regions into which six overlapping triangles could divide the plane. Describe
or draw an arrangement of six triangles that divides the plane into that many regions.
Comments: This problem was developed by Prof. George Berzsenyi, the founder of the
USAMTS.
The graders of this problem requested that I explain the weaknesses of the greedy algorithm,
because many submitted solutions relied on it. The greedy algorithm is a rule of thumb for making
efcient constructions that are too complicated to construct in one fell swoop: rst construct the
best solution you can nd for size 1; next, extend the size 1 result as efciently as possible to
make a size 2 result; continue this process, extending the size i result as efciently as possible to
construct size i + 1 , until you reach the desired size.
The greedy algorithm does indeed nd the correct maximum number of regions for problem
4/4/14. However, nding a result by the greedy algorithm does not count as a proof that that result
is the best, because for some problems, the greedy algorithm fails to nd the best solution, even
when each step uses the provably best extension of the previous result. In those counterexamples,
the best solution for size i does not contain the best solution for size i 1 inside it, so extending
the ( i 1 ) st solution does not yield the ith solution.
For example, consider the problem of nding the smallest positive integer with exactly i fac
tors. The answer for one factor is that 1 has one factor. By the greedy algorithm, to nd a potential
solution for two factors, we multiply 1 (the previous result) by 2 (the smallest prime) to nd that 2
is the smallest integer with two factors, { 1, 2} . By the greedy algorithm, to nd a potential solu
tion for three factors, we multiply 2 by 2 to nd that 4 is the smallest integer with three factors,
{ 1, 2, 4} . By the greedy algorithm, to nd a potential solution for four factors, we multiply 4 by
2 to nd that 8 is a small integer with four factors, { 1, 2, 4, 8} . But it is not the smallest integer
with four factors, because 6 is. The greedy algorithm failed.
The following solution by Piotr Wojciechowski uses the greedy algorithm, but he is careful to
never demand that the solution for n triangles must contain the solution for n 1 triangles, so his
argument is rigorous. The second solution, by Connie Yee, also uses the greedy algorithm; in
additon, she proves that the greedy algorithm does give the best solution. The third solution, by
Greg Evans, does not use the greedy algorithm.
Solution 1 to 4/4/14 by Piotr Wojciechowski (9/WV):

Let R n be the maximum number of regions into which n triangles can divide the plane. We
will show by induction that R n 3n ( n 1 ) + 2 .
Obviously, R 1 = 2 . Suppose T 1, T 2, , T n are n triangles in the plane that divide the plane
into R n regions. Since a line can intersect a triangle at at most 2 points, a triangle (ignoring its
interior) can intersect another triangle at at most 6 points. So a triangle can intersect n 1 trian
gles at at most 6( n 1 ) points. Thus, triangle T n intersects the other n 1 triangles T 1 , T 2 , ,
T n 1 in at most 6( n 1 ) . If T n intersects the other triangles at k points, then the perimeter of
11
T n is divided into k intervals each of which divides a region created by the n 1 triangles into
two regions. Thus, the number of regions into which the n triangles T 1 , T 2 , , T n divide the
plane into is k larger than the number of regions into which the n 1 triangles T 1 , T 2 , , T n 1
divide the plane. The number of regions created by the n 1 triangles is at most R n 1 and the
value of k is at most 6(n 1 ) . It follows thatR n R n 1 + 6(n 1 ). By the induction hypothe
sis,R n 1 3(n 1 )(n 2 ) + 2. So R n 3(n 1 )(n 2 ) + 2 + 6(n 1 ) = 3n ( n 1 ) + 2.
In particular, R 6 92 .
We will show that R 6 = 92 , by showing that there is an arrangement of six triangles T 1 , T 2 ,
, T 6 on that plane that divide the plane into 92 regions. By the argument above, it is enough to
arrange them so that any two triangles intersect at six points and no three triangles intersect at the
same point.
The required arrangement can be obtained as follows. Take C3
C2
C4
C1
18 points A 1 , A 2 , , A 6 , B 1 , , B 6 , C 1 , , C 5 , C 6 on a C5
circle, lying in that speciced order along the circle. Form the C6
triangles T 1 = A 1 B 1 C 1 and T 2 = A 2 B 2 C 2 . Clearly, T 1

A1
and T 2 intersect at six points. If the segment A 3 B 3 goes B6

A2
through any of those intersection points, move point B 3 to A3

A4 B5
anywhere else in the arc between B 2 and B 4 so that the seg B4
ment A 3 B 3 does not go through any intersection points. This A5

B3
A6 B2
B1
is possible since there are innitely many possible positions
for the point B 3 and only nitely many of those positions are bad. If any of the segments A 3 C 3
and B 3 C 3 go through an intersection point, then we can move point C 3 to a position so that the
segments do not go through intersection points, as we did with point B 3 . Once points B 3 and C 3
are in good positions, we form triangle T 3 = A 3 B 3 C 3 . Repeat this process for B 4 and C 4 so
that none of A 4 B 4 , A 4 C 4 , and B 4 C 4 go through any of the intersection points between T 1 , T 2 ,
and T 3 , and form triangle T 4 = A 4 B 4 C 4 . Repeat this process for B 5 , C 5 , B 6 , and C 6 to form
triangles T 5 = A 5 B 5 C 5 and T 6 = A 6 B 6 C 6 . Because of the order of the points along the cir
cle, any two triangles intersect at 6 points. Because we made sure to eliminate any triple intersec
tions, no three triangles intersect at the same point.
Triangle T 2 is split into six parts by its points of intersection with triangle T 1 , triangle T 3 is
split into twelve parts by its points of intersection with triangles T 1 and T 2 , and so on up to trian
gle T 6 split into 30 parts by its intersection with triangles T 1 , T 2 , T 3 , T 4 , and T 5 . Since trian
gle T 1
alone divided the plane into two regions, the total number of regions is
2 + 6
+ 12 + 18 + 24 + 30 = 92.
12
Solution 2 to 4/4/14 by Connie Yee (11/NY):
One triangle divides the plane into two regions.
For the best case with two triangles, each side of the second triangle cuts through each angle
of the rst triangle at two points. Each point of intersection corresponds to one new region. The
three sides with two intersections each generate six more regions.
For the best case with three triangles, each side of the third triangle cuts through an angle of
the rst triangle and an angle of the second triangle for a total of four intersection points per side.
Thus, the third triangle generates 3 4 = 12 more regions.
The fourth triangle can be set up so that each side of the fourth tri- I
angle cuts through an angle of the rst triangle, an angle of the second II
triangle, and an angle of the third triangle, as shown in the diagram to III
the right. This gives a total of six intersection points per side, generat- IV
ing 3 6 = 18 more regions.
So long as we can send each side of the new triangle through an
angle of each of the previous triangles, the pattern goes on. For each
new intersection point, we get a new region. Each side can intersect a
previous triangle at at most two points, by going through an angle. So
the pattern give the maximum possible number of regions. The growth proceeds as follows:
Number of general
triangles 1 2 3 4 5 6 n
Maximum 2 8 20 38 62 92 Rn
number of =2+321 =8+322 =20+323 =38+324 =62+325 = Rn-1 +
regions 6(n - 1)
The recursive formula R n = R n 1 + 6(n 1 ) gives a direct formula R n = 3n ( n 1 ) + 2 .
Can we always continue the pattern? Look at the four-triangle A I
diagram again. Any side of triangle IV cuts through an angle of II
A III
each of triangles I, II, and III, and that side and the next side of tri
angle IV make an angle of triangle IV. Label the rst point where
IV
that side of triangle IV cuts through triangle I as A , the rst point B
B
where the next side of triangle IV cuts through triangle I as B , and
the rst point of where the third side of triangle IV cuts through tri- C
C
angle I as C . Line AB cuts through an angle of every triangle, but
line segment AB is barely too short to nish cutting through triangles I and IV, since its endpoints
are on their perimeters. That happens to lines AC and BC , too. Points A , B , and C are next to
the exterior region; therefore, we can nd points A , B , and C near them in the exterior region
so that each side of triangle ABC cuts through an angle of every previous triangle. However,
drawing a new side though an intersection point of previous triangles would not make a new inter
section point and a new region. So we avoid drawing any new side through an old intersection.
We can make the sixth triangle by the same method.
Therefore, the maximum number of regions into which 6 overlapping triangles could divide a
plane is 92.
13
Solution 3 to 4/4/14 by Greg Eden (10/MD):
To create the most regions in a plane with two overlapping triangles, the triangles should be
placed so that they share an interior region but their vertices are rotated relative to each other so
that each vertex of each triangle extends outside the other triangle. This creates 3+3 triangular
regions from the corners of the triangles in addition to the region around the triangles and the
polygonal region inside both triangles. This adds up to eight regions for two triangles, like the
problem indicates.
If there are x overlapping triangles, then there are
x( x 1)
i = 1 i = -------------------
x1
distinct pairs of triangles. Each distinct pair of triangle splits off six distinct
2
triangular regions besides the common interior and exterior of each triangle. The interior and
exterior regions are not created by the pairing of the triangles, but the six triangular regions are.
Actually, with more than three overlapping triangles, these triangular regions lose their triangular
shape, because pieces of them are claimed by the triangular regions split off to form the triangular
regions of other pairs of triangles. But the number of formerly triangular regions is unaffected by
x( x 1)
losing pieces. This means that there are 6 ------------------- = 3x ( x 1 ) total formerly triangular
2
regions formed by x overlapping triangles. When the original interior region and exterior region
are added, there are a total of r ( x ) = 3 x ( x 1 ) + 2 regions in the gure.
If a pair of overlapping triangles are positioned differently relative to each other, they cannot
create more triangular regions than six, so r ( x ) is the maximum.
For six overlapping triangles, this means we have r ( 6 ) = 3 6 5

+ 2 = 92 regions.
Colorful illustrations of how the triangular regions are split off are below.
2 triangles, 8 regions 3 triangles, 20 regions

Blue-green pair makes 3 solid blue Red-blue pair makes 3 pale red
regions and 3 striped green regions. regions and 3 striped blue regions.
Red-green pair splits 3 solid blue
regions into red and blue and
3 striped green regions into striped
and pale.
14
4 triangles, 38 regions 5 triangles, 62 regions

Magenta-red pair makes 3 hatched magenta Cyan-blue pair makes 3 barred cyan regions and
regions and 3 striped red regions. 3 pale blue regions.
Magenta-blue pair splits 3 pale red regions Cyan-green pair splits 3 solid blue regions into
into magenta and red and 3 striped blue cyan and blue and 3 pale green regions into
regions into striped and hatched. barred and pale.
Magenta-green pair splits 3 solid red Cyan-magenta pair splits 3 striped blue regions
regions into magenta and red and 3 striped into cyan and blue and 3 pale magenta regions
green regions into striped and hatched. into barred and pale.
Cyan-red pair splits 3 hatched blue regions into
cyan and blue and 3 pale red regions into barred
and pale.
6 triangles, 92 regions
Yellow-green pair makes 3 speck-
led yellow regions and 3 solid
green regions.
Yellow-magenta pair splits 3
striped green regions into yellow
and green and 3 solid magenta
regions into speckled and solid.
Yellow-red pair splits 3 hatched
green regions into yellow and green
and 3 solid red regions into speck-
led and solid.
Yellow-cyan pair splits 3 pale green
regions into yellow and green and 3
solid cyan regions into speckled
and solid.
Yellow-blue pair splits 3 barred
green regions into yellow and green
and 3 solid blue regions into speck-
led and solid.
15
5/4/14. Prove that if the cross-section of a cube cut by a plane is a
pentagon, as shown in the gure on the right, then there are two E
adjacent sides of the pentagon such that the sum of the lengths A
of those two sides is greater than the sum of the lengths of the
other three sides. For ease of grading, please use the names of D
the points from the gure on the right in your solution.
Comments: This problem was suggested by Professor Gregory Gal

perin of Eastern Illinois University. We are thankful for the many C
ne problems Prof. Galperin has sent the USAMTS over the years. B
Solution 1 to 5/4/14 by Ari Ofcer (10/IL):

Extend the lines ED and BC . Since both lines are on the plane cutting through the cube, they
meet at a point F . The gure ABFE is a parallelogram, since if two parallel planes, such as
opposite sides of a cube, are intersected by a third plane, then the lines of intersection are parallel.
Line segment AE is the same length as line segment BF , and line segment AB is the same
length as line segment EF , because opposite sides of a parallelogram are of equal length. Thus,
the sum of lengths AB and AE equals the sum of lengths BF and EF .
However, in the original pentagon, side CD cuts the corner of the parallelogram, removing F ,
which results in a decreased perimeter. The perimeter decreases in going from parallelogram to
pentagon because the some of the lengths of two sides of a triangle, CF and DF , is greater than
the lenght of the third side, CD . Since the sum of DE , CD , and BC is less than the sum of EF
and BF , it is also less than the sum of AB and AE . Hence, the sum of the lengths of the two
adjacent sides AB and AE is greather than the sum of the lengths of the other three sides BC ,
CD , and DE .
F
C
B
16
Solution 2 for 5/4/14 by Judith Stanton (12/IN):

I start by creating a coordinate system. The vertex of the cube nearest points C and D is the
origin, the edge containing C is on the positive x-axis, the edge containing D is on the positive
z-axis, and the third edge out from the origin is on the positive y-axis. Let s be the length of an
edge of the cube. Then I can write the coordinates of the vertices of the pentagon as:
A = (s, y A, s), B = (s, y B, 0), C = (x C, 0, 0), D = (0, 0, z D ), E = (0, y E, s).
Since the plane of the pentagon does not contain the origin, I can assume its equation is
mx ny + lz = s. (1)
and the reason of the subraction of the middle term will become obvious. Then substituting the
coordinates for points A , B , C , D , and E into equation (1) gives
s(1 m l) s(1 m) s s s(1 l)
y A = ----------------------------, y B = --------------------, x C = ----, zD = - , y E = ------------------ .
n n m l n
Since x C and z D are between 0 and s, I see that m, l > 1 . That means 1 m l < 1 , and since
y A is between 0 and s, it must be that n < 1 m l < 1 , which gives n > 1 .
Let AB denote the distance between A and B , etc. Then
( sl ) 2 s
AB = ( s s) 2 + (y A y B ) 2 + (s 0 ) 2 = - + s 2 = --- l 2 + n 2
----------
n 2 n
( s ( m 1 )) 2 (s ( m 1 )) 2
- = --- ------- m 2 + n 2
s s
BC = ( s x C ) 2 + (y B 0 ) 2 + (0 0) 2 = ---------------------------
- + --------------------------
m 2 n 2 n mn
s2 s2 s
CD = ( x C 0 ) 2 + (0 0) 2 + (0 z D ) 2 = -----2- + ----2 = ------ l 2 + m 2
m l lm
( s ( l 1 )) 2 (s ( l 1 )) 2
- = --- ----- l 2 + n 2
s s
DE = ( 0 0) 2 + (0 y E ) 2 + (z D s) 2 = - + -----------------------
------------------------
n 2 l 2 n ln
(sm) 2 s
AE = ( s 0 ) 2 + (y A y E ) 2 + (s s) 2 = s 2 + -------------
- = --- m 2 + n 2 .
n 2 n
So
s s
AB + AE BC DE = ----- l 2 + n 2 + ------- m 2 + n 2 > 0
ln mn
and
s 2 s 2 2s 2 2s 2 2
( AB + AE BC DE) 2 = ----2 + ------2 + -------
2
- + -----------2- l + n 2 m 2 + n 2
l m n lmn
s2 s2 s 2
> ----2 + ------2 = ------ (l 2 + m 2 ) = (CD) 2
l m lm
Therefore, AB + AE BC DE > CD , so AB + AE > BC + CD + DE . Thus, the sum of the
lengths of the adjacent sides AB and AE is greater than the sum of the lengths of the other three
sides.
17

CREDITS and QUICK ANSWERS
1/1/15. Let f (x) = x + 10x + 100x + 1000x for all real numbers x, where for any real number r, r
means the greatest integer less than or equal to r. For example, f () = 3 + 31 + 314 + 3141 = 3489.
Find, with proof, a positive integer n less than 2003 such that f (x) = n has no solution for x, but
f (y) = n + 11 and f (z) = n + 111 have solutions for y and z.
The function f (x) = x+10x+100x+1000x was invented by George Berzsenyi,
the founder of the USAMTS.
Since f is constant over all real numbers in the halfopen interval [i/1000, (i + 1)/1000)
for all integers i, to nd all possible values of f we need look only at the values of
f (i/1000) for integers i. The values of f (i/1000) and f ((i + 1)/1000) are consecutive
integers unless i + 1 is divisible by 10. n = 1108 is in the gap between f (999/1000) =
1107 and f (1000/1000) = 1111, n + 11 = 1119 = f (1008/1000), and n + 111 = 1219 =
f (1099/1000).
2/1/15. Find all primes p for which 6p + 1 is the fth power of an integer. Prove that you found all of them.
This problem was also devised by George Berzsenyi.
Suppose 6p+ 1 is a fth power n5 . This gives 6p = n5 1 = (n 1)(n4 +n3

+n2 + n +1).
The second factor, n4 + n3 + n2 + n + 1, is greater than 6, since n is at least 2. So p
must be a factor of n4 + n3 + n2 + n + 1. This leaves n 1 as a factor of 6, so n could
be only 2, 3, 4, or 7. Of those, only n = 7 works, giving 75 = 6 2801 + 1. Thus, the
only answer for p is the prime 2801.
3/1/15. We attempted to arrange the integers 1, 2, 3, ..., 12 around a circle so that all sums of pairs of
adjacent integers are either
(a) all perfect squares, or
(b) all triangular numbers, which are numbers of the form n(n + 1)/2.
One case out of (a) and (b) succeeded. Which one? For the successful case, show a valid arrange
ment. For the unsuccessful case, explain why it is impossible.
This problem was also devised by George Berzsenyi.
Case (b) succeeded.
(a) The only summation involving 12 that sums to a perfect square is 4 + 12 = 16.
Since we would need two sums involving 12 in order to arrange the integers 1, 2, 3, ...,
12 around a circle, this case is impossible.
(b) The triangular numbers are 1, 3, 6, 10, 15, 21, 28, etc. 6 9
The circle to the right has the numbers 1 through 12 ar 4 12
11 3
ranged so that all sums of pairs of adjacent numbers are
triangular. 10 7
5 8
1 2
4/1/15. Two players play a game. Each player, in turn, has to name a positive integer that is less than the
previous number but at least half the previous number. The player who names the number 1 loses.
If the rst player starts by naming 2003 and after that both players play with the the best strategy,
who wins? Describe the strategy and prove it works.
This problem was invented by David Grabiner, a mathematician at NSA.
The second player will win.
The best strategy is to name only numbers of the form 3(2n ) 1, which are 1535, 767,
383, 191, 95, 47, 23, 11, 5, and 2. If a player names a number of the form 3(2n ) 1,
then the other player has to name a number between 3(2n1 ) and 3(2n ) 2, inclusive.
No matter which of those numbers the other player names, the strategic player can
legally respond by naming 3(2n1 ) 1. The last number the strategic player names
will be 3(20 ) 1 = 2. The other player can respond only by naming 1, so that player
will lose. In the game from the problem, the rst player started with the nonstrategic
number 2003, so the second player gets to claim the best strategy by naming 1535.
5/1/15. A rectangular piece of paper, ABCD, is folded along line seg A B A C B

ment EF , where point E is on side AD and point F is on side BC, S

so that C ends up at the midpoint of side AB. Determine the length F D SF

of EF if the length of AB is 240 and the length of BC is 288. E S
E
S
D C
This problem is based on Problem 2.3.3 of Traditional Japanese Problems of the 18th
and 19th Century by Fukagawa and Rigby.
C
A B A fold reects points; therefore, the line EF is the perpen
B dicular
bisector of line segment CC . The length of CC is
(120 + 2882 ) = 312. Angle CC B is congruent to angle
B 2

B
B F
EF C, so there exists a point G on ray F C such that triangle
B
EF G is a right triangle similar to triangle CC B. Side EG is
E G
B
B
B
parallel to side AB, so EG = 240. By ratios of corresponding
B lengths on similar triangles, EF/EG = CC /BC. Therefore,
D C EF = (240)(312)/288 = 260.

1/2/15. The faces of 27 unit cubes are painted red, white, and blue in such a manner that we can assemble
them into three dierent congurations: a red 3 3 3 cube, a white 3 3 3 cube, and a blue
3 3 3 cube. Determine, with proof, the number of unit cubes on whose faces all three colors
appear.
This problem was devised by George Berzsenyi, Professor Emeritus of RoseHulman
Institute for Technology and founder of the USAMTS. We are thankful for all the work
Prof. Berzsenyi puts into this contest.
Painting the surface of three 3 3 3 cubes requires 162 square units, which equals
the total surface area of the 27 unit cubes. So no face is left unpainted. In the white
3 3 3 cube the eight cubes on the corner each have three white faces, the twelve
cubes on the edges each have two white faces, the six cubes on the face centers each
have one white face, and the one cube in the center has zero white faces. The same
happens for red and blue. The only way for a unit cube to miss a color is if it serves
as the center cube of that colors 3 3 3 cube. A cube missing a color must contain
three faces of each of the other two colors, so each center of a monotone 3 3 3 cube
is distinct. Thus, exactly three cubes have two colors and exactly twentyfour cubes
have three colors.
2/2/15. For any positive integer n, let s(n) denote the sum of the digits of n in base 10. Find, with proof,
the largest n for which n = 7s(n).
This problem was inspired by a Problem of the Month proposal of Professor Bela
Bajnok of Gettysburg College.
Consider a positive integer n with at most d digits. We write it as ad1 ad2 . . . a2 a1 a0
with each ai being adigit from 0 to 9, starting
d1with leading zeros if n has fewer than
d1 i
d digits. Then n = i=0 10 ai and s(n) = i=0 ai . The equation n = 7s(n) becomes
10d1 ad1 + . . . + 100a2 + 10a1 + a0 = 7ad1 + . . . 7a2 + 7a1 + 7a0 . (1)
Cancelling like terms on opposite sides leaves
(10d1 7)ad1 + . . . + 93a2 + 3a1 = 6a0 . (2)
Since the maximum value of the righthand side of equation (2) is 6 9 = 54, we have
that ai = 0 for all i > 1. This leaves 3a1 = 6a0 , that is a1 = 2a0 . The largest digits
that work for a1 and a0 are a1 = 8 and a0 = 4, giving n = 84.
q q q
3/2/15. How many circles in the plane contain at least three of the nine points (0, 0), (0, 1), (0, 2), q q q
(1, 0), (1, 1), (1, 2), (2, 0), (2, 1), (2, 2)? Rigorously verify that no circle was skipped or q q q
counted more than once in the result.
We are thankful to Professor Harold B. Reiter, the President of Mu Alpha Theta, for
this problem.
1
There are 34 circles: one through (0, 0), (0, 2), (2, 0), and (2, 2); one through (0, 1),
(1, 0), (1, 2), and (2, 1); four like the one through (0, 0), (0, 1), (1, 0), and (1, 1); four
like the one through (0, 0), (0, 1), (2, 0), and (2, 1); four like the one through (0, 0),
(0, 1), (1, 2), and (2, 2); eight like the one through (0, 0), (1, 1), and (1, 2); four like the
one through (0, 0), (0, 2), and (1, 1); four like the one through (0, 0), (0, 2), and (1, 2);
and four like the one through (0, 0), (1, 2), and (2, 1).
To verify we did not skip a circle, we count that every set of three points has been
covered. There are ( 93 ) = 84 such sets. The 14 circles with four points each account for
4 14 = 56 sets. The 20 circles with three points each account for 20 sets. There are
eight lines that cover three points each, which do not count as circles. 56 + 20 + 8 = 84,
accounting for every set of three points.
4/2/15. In how many ways can one chose three angle sizes, , , and , with
from the set of integral degrees, 1 , 2 , 3 , . . ., 178 , such that those angle sizes SS

correspond to the angles of a nondegenerate triangle? How many of the resulting S
triangles are acute, right, and obtuse, respectively?
S
The rst question came from Hungarys famous high school mathematics journal
KoMaL more than a century ago. The second question was added by George Berzsenyi.
For acute triangles, angle ranges from 60 to 89 , angle ranges from (180 )/2
to , and angle is 180 . That adds up as 1+2+4+5+7+8+ +43+44 = 675
acute triangles. For right triangles, angle is 90 , angle ranges from 45 to 89 , and
angle is 90 . That gives 45 right triangles. For obtuse triangles, angle ranges
from 91 to 178 , angle ranges from (180 )/2 to 179 , and angle is
180 . That adds up as 44 + 44 + 43 + 43 + . . . + 1 + 1 = 1980 obtuse triangles.
That gives 2700 triangles total.
5/2/15. Clearly draw or describe a convex polyhedron that has exactly three pentagons among its faces and
the fewest edges possible. Prove that the number of edges is a minimum.
This problem was inspired by a problem from the keynote address given by Dr. Robert
Geretschlager at the Congress of the World Federation of National Mathematics Com
petitions in Melbourne, Australia, in August 2002. Dr. Geretschlanger is the Leader of
Austrias IMO team. His permission to use the original problem is most appreciated.
Each pentagon has ve edges on it, but any pair of pentagons can share one edge. That
gives at least 12 edges on the three pentagons. At least one vertex on each pentagon
does not touch another pentagon. Such a vertex has at least one more edge from it,
which we will count as half an edge, because we might also count the other end of that
edge elsewhere. Thirteen and a half edges round up to a minimun of 14 edges.
There are two topologicallydistinct convex polyhe @ H
@H @X
dra with three pentagonal faces and 14 edges. The @
A
@ H
A X
faces of both polyhedra consist of three pentagons, A A
A

one quadrilateral, and three triangles. We can con A A
struct either out of a triangular prism, one by making
@ @
two triangular cuts on adjacent corners of a triangle @ @
and the other by making two triangular cuts on a @ @
single corner, as shown on the right.

1/3/15. Find, with proof, all pairs of twodigit positive integers ab and cd such that all the digits a, b, c, and
d are dierent from one another and (ab)(cd) = (dc)(ba).
This is an old problem that Professor Emeritus George Berzenyi, the founder of the US
AMTS, remembered for this contest without recalling its original source.
The numerals in (ab)(cd) = (dc)(ba) expand out as
(10a + b)(10c + d) = (10d + c)(10b + a)

100ac + 10ad + 10bc + bd = 100bd + 10ad + 10bc + ac
99ac = 99bd
ac = bd.
So we need only look for all pairs of distinct digits whose products are the same, (1)(6) =
(2)(3), (1)(8) = (2)(4), (2)(6) = (3)(4), (2)(9) = (3)(6), and (3)(8) = (4)(6), and arrange
them in all possible orders:
(12)(63) = (36)(21)
(13)(62) = (26)(31)
(12)(84) = (48)(21)
(14)(82) = (28)(41)
(23)(64) = (46)(32)
(24)(63) = (36)(42)
(23)(96) = (69)(32)
(26)(93) = (39)(62)
(34)(86) = (68)(43)
(36)(84) = (48)(63).
2/3/15. Find the smallest positive integer n such that the product (2004n + 1)(2008n+1) is a perfect square.
Prove that n is as small as possible.
This problem was devised by Professor Emeritus George Berzsenyi and modied to be more
challenging by Dr. David Grabiner, a mathematician at the National Security Agency.
Since 501(2004n + 1) 500(2008n + 1) = 1, the two factors 2004n + 1 and 2008n + 1
are relatively prime. So both factors must be perfect squares. Let x2 = 2004n + 1 and
y 2 = 2008n + 1.
Since both x and y are odd integers, and y is larger than x, we have y x + 2. So y 2 x2
(x + 2)2 x2 = 4x + 4. We also have that y 2 x2 = (2008n + 1) (2004n + 1) = 4n. Thus,
x+1 n. Subtracting 2 from both sides of the inequality
gives x1 n2. The values x+1,
x 1, n, and n 2 are positive, since x is at least 2004(1) + 1 = 44.78, so we can multiply
the inequalities to get (x + 1)(x 1) n(n 2). Because (x + 1)(x 1) = x2 1 = 2004n,
this gives 2004n n(n 2), so 2004 n 2 and 2006 n.
Trying n = 2006 gives (2004n + 1)(2008n + 1) = (20052 )(20072 ) = 40240352 .
1
3/3/15. Pebbles are put on the vertices of a combinatorial graph. For a vertex with two or more pebbles, a
pebbling step at that vertex removes one pebble at the vertex from the graph entirely and moves
another pebble at that vertex to a chosen adjacent vertex. The pebbling number of a graph is
the smallest number t such that no matter how t pebbles are distributed r r
on the graph, the distribution would have the property that for every empty JJ
r
vertex a series of pebbling steps could move a pebble to that one vertex.
J
For example, the pebbling number of the graph formed from the vertices and r r
J r r
edges of a hexagon is eight. Find, with proof, the pebbling number of the J
J
Jr
Jr
graph illustrated on the right.
Pebbling was invented in 1987 to simplify a number theoretic proof. Check the work of Glenn
Hurlbert at the Southeastern Combinatorics/Graph Theory Conference at Florida Atlantic
University for details. Dr. Michelle Wagner of the NSA brought pebbling to the attention
of the USAMTS. Dr. Erin Schram of the NSA chose the graph, which, incidentally, is the
graph of the states of a hexahexaexagon.
The pebbling number is more than 13, since the 13pebble distri 7r 1r
bution illustrated on the right cannot move a pebble to the target JJ0r

vertex in a series of pebbling steps. Thus, we want to see whether
J
3r 0r
J0r 1r
14 pebbles will always be enough to put a pebble on a target vertex. J1
r
J
J
Jr target
If the target vertex is one of the corner vertices, assign a point value 1r 1r
to each vertex of the graph, as illustrated on the right. Each pebble JJ2
r
on the graph is worth the point value of its vertex. The points are
J
2r
J4r
designed so that whenever a pebbling step moves a pebble toward 1r 1r
J1
J
the target vertex, the total point value of all pebbles on the graph Jr
Jr target
does not change unless we move a pebble onto the target vertex.
Also, the total point value of one pebble on each of the eight vertices beside the target vertex
is 13. Therefore, whenever the total point value of all the pebbles is more than 13, some
vertex besides the target vertex has two or more pebbles and we can make a pebbling step
to move a pebble toward the target vertex. Suppose we start with 14 pebbles. We would
either start with a pebble on the target vertex or start with 14 or more total points. With a
total of 14 or more points, we can repeatedly make pebbling steps toward the target vertex
until we nally move a pebble onto the target vertex.
If the target vertex is one of the central vertices, use the point 1r 1r
system illustrated on the right instead. We can always make a JJ2
r
pebbling step toward the target vertex if we have 11 or more points. J

1r 2r
J0r 1r
Since 14 pebbles give us 14 or more points unless a pebble is on the J 1
J 1
target vertex, that means that 14 pebbles give us enough to move Jr

target Jr

a pebble onto that target vertex.
Therefore, the pebbling number of the graph is 14.
4/3/15. A innite sequence of quadruples begins with the ve quadruples (1, 3, 8, 120), (2, 4, 12, 420),
(3, 5, 16, 1008), (4, 6, 20, 1980), (5, 7, 24, 3432). Each quadruple (a, b, c, d) in this sequence has
the property that the six numbers ab + 1, ac + 1, bc + 1, ad + 1, bd + 1, and cd + 1 are all perfect
squares. Derive a formula for the nth quadruple in the sequence and demonstrate that the property
holds for every quadruple generated by the formula.
This problem was inspired by some of the work of Neven Juric of Croatia.
The quadruples are given by the formula (n, n + 2, 4n + 4, 16n3 + 48n2 + 44n + 12).
The six perfect squares are
(n)(n + 2) + 1 n2 + 2n + 1 = (n + 1)2 ,
=
(n)(4n + 4) + 1 4n2 + 4n + 1 = (2n + 1)2 ,
=
(n + 2)(4n + 4) + 1 4n2 + 12n + 9 = (2n + 3)2 ,
=
3
(n)(16n + 48n2 + 44n + 12) + 1 16n4 + 48n3 + 44n2 + 12n + 1
=
= (4n2 + 6n + 1)2 ,
(n + 2)(16n3 + 48n2 + 44n + 12) + 1 = 16n4 + 80n3 + 140n2 + 100n + 25
= (4n2 + 10n + 5)2 ,
(4n + 4)(16n3 + 48n2 + 44n + 12) + 1 = 64n4 + 256n3 + 368n2 + 224n + 49
= (8n2 + 16n + 7)2 .
5/3/15. In triangle ABC the lengths of the sides of the triangle opposite to A
the vertices A, B, and C are known as a, b, and c, respectively. Prove
B

there exists a constant k such that if the medians emanating from A N B
and B are perpendicular to one another, then a2 + b2 = kc2 . Also nd H
PH HHB
B
the value of k. H
C M B
Prof. George Berzsenyi invented this problem, but he doubts that he was the rst to come
up with it.
Let m denote the length M P and n denote the length N P . Since the medians of a triangle
intersect one third of their length from their base end, this gives that AP = 2m and BP = 2n.
The perpendicular medians create three right triangles with the sides of triangle ABC and
we can use the Pythagorean theorem to nd relations between the medians and the sides:
A
b/2B BM P gives m2 + (2n)2 = (a/2)2 , so a2 = 4m2 + 16n2 (1)
2 2 2 2 2 2
N n 2m B c AN P gives (2m) + n = (b/2) , so b = 16m + 4n (2)
B
ABP gives (2m)2 + (2n)2 = c2 , so c2 = 4m2 + 4n2

H
(3)
H
P H 2n B

m HHB
H B
B
C M a/2 B
Summing equations (1) and (2) gives a2 + b2 = 20m2 + 20n2 . Since c2 = 4m2 + 4n2 , this
means a2 + b2 = 5c2 . The value of k is 5.
3
1/4/15. Find, with proof, the smallest positive integer n for which the sum of the digits of 29n is as small as
possible.
This problem was fashioned after Problem 3 of the 2nd Lithuanian Olympiad administered
at Vilnius University on September 30, 2000.
We search for 29n rather than directly for n. Numbers with only a single nonzero digit are
that digit times a power of ten, which would not be divisible by 29. Of the numbers with two
nonzero digits, the smallest digit sum occurs with both nonzero digits being 1. The smallest
such number must begin and end in 1, because if it ended in zero, we could divide by 10. So
we are looking for the smallest integer of the form 10e + i that is a multiple of 29.
The powers of 10 mod 29 are 1, 10, 13, 14, 24, 8, 22, 17, 25, 18, 6, 2, 20, 26, 28, 19, 16, 15,
5, 21, 7, 12, 4, 11, 23, 27, 9, 3, 1, . . .. so 1014 28 (mod 29). Therefore 1014 + 1 is divisible
by 29, and n = (1014 + 1)/29 = 3448275862069.
2/4/15. For four integer values of n greater than six, there exist right triangles whose side lengths are integers
equivalent to 4, 5, and 6 modulo n, in some order. Find those values. Prove that at most four such
values exist. Also, for at least one of those values of n, provide an example of such a triangle.
This problem was devised by Dr. Peter Anspach of the NSA, after a lunchtime conversation
among mathematicians about trying to create a (4, 5, 6) right triangle.
Let a, b, and c be the sides of such a right triangle, with a2 + b2 = c2 . If c 4 (mod n),
then 52 + 62 42 (mod n) so n is a factor of 52 + 62 42 = 45. If c 5 (mod n), then n is
a factor of 42 + 62 52 = 27. If c 6 (mod n), then n is a factor of 42 + 52 62 = 5, but
since n > 6, we can reject this case. Thus, n is a factor of 27 or 45, so n is 9, 15, 27, or 45.
Given an n, a triple (a, b, c) that forms a right triangle can be found by trial and error, by
applying number theory to an equation such as (9x + 5)2 + (9y + 6)2 = (9z + 4)2 , or by the
(r(s2 t2 ), 2rst, r(s2 + t2 )) method of finding Pythagorean triples after solving for r, s, and
t that give the right values mod n. The first three solutions for each case are:
for n = 9 and c 4 (mod 9): (24, 32, 40), (51, 68, 85), (78, 104, 130);
for n = 9 and c 5 (mod 9): (40, 96, 104), (15, 112, 113), (60, 175, 185);
for n = 15 and c 4 (mod 15): (65, 156, 169), (140, 336, 364), (215, 516, 559);
for n = 27 and c 5 (mod 27): (168, 490, 518), (33, 544, 545), (814, 1248, 1490);
for n = 45 and c 4 (mod 45): (1176, 2975, 3199), (3920, 16926, 17374), (1760, 17556, 17644).
1
3/4/15. Find a nonzero polynomial f (w, x, y, z) in the four indeterminates w, x, y, and z of minimum degree
such that switching any two indeterminates in the polynomial gives the same polynomial except
that its sign is reversed. For example, f (z, x, y, w) = f (w, x, y, z). Prove that the degree of the
polynomial is as small as possible.
This problem was invented by Dr. David Grabiner, an NSA mathematician who helps with
the evaluation and grading of problems for the USAMTS.
Consider any two of the indeterminants, say w and x. If the standard form of f (w, x, y, z) has
a term kwa xb y c z d , then it must have a matching term kwb xa y c z d . Grouping them together
gives either k(wab xab )wb xb y c z d or k(xba wba )wa xa y c z d , which both are divisible by
w x. Since the entire polynomial can be grouped this way, it is divisible by w x.
Thus, for every pair of indeterminants, the polynomial is divisible by their difference. This
means f (w, x, y, z) is divisible by (w x)(w y)(w z)(x y)(x z)(y z). Letting
f (w, x, y, z) = (wx)(wy)(wz)(xy)(xz)(yz) keeps the degree to a minimum, and it
works because switching any two indeterminants in (w x)(w y)(w z)(xy)(xz)(y z)
gives the same polynomial except that its sign is reversed.
4/4/15. For each nonnegative integer n define the function fn (x) by

2 4
fn (x) = sinn (x) + sinn (x + ) + sinn (x + )
3 3
for all real numbers x, where the sine functions use radians. The functions fn (x) can be also
expressed as polynomials in sin(3x) with rational coefficients. For example,
3 3
f0 (x) = 3, f1 (x) = 0, f2 (x) = , f3 (x) = sin(3x),
2 4
9 15 27 3
f4 (x) = , f5 (x) = sin(3x), f6 (x) = + sin2 (3x),
8 16 32 16
for all real numbers x. Find an expression for f7 (x) as a polynomial in sin(3x) with rational coeffi-
cients, and prove that it holds for all real numbers x.
This problem was devised by the distinguished Hungarian mathematician and poet Mihaly
Bencze of Brasso, Transylvania.
The triple angle formula for sine is sin(3x) = 3 sin(x) 4 sin3 (x). But since sin(3x) also
equals sin(3(x + 2
3
)) and sin(3(x + 4
3
)), it also expands to 3 sin(x + 2
3
) 4 sin3 (x + 2
3
)
4 3 4
and 3 sin(x + 3 ) 4 sin (x + 3 ). Thus, for all real numbers x,
2 4
fn (x) sin(3x) = sinn (x) sin(3x) + sinn (x + ) sin(3x) + sinn (x + ) sin(3x)
3 3
= sinn (x) 3 sin(x) 4 sin3 (x)

2 2 2
+ sinn (x + ) 3 sin(x + ) 4 sin3 (x + )
3 3 3
4 4 4
+ sinn (x + ) 3 sin(x + ) 4 sin3 (x + )
3 3 3
2 4
= 3 sinn+l (x) + sinn+l (x + ) + sinn+l (x + )
3 3
2 4
4 sinn+3 (x) + sinn+3 (x + ) + sinn+3 (x + )
3 3
= 3fn+l (x) 4fn+3 (x).
2
Setting n to 4 and rearranging gives f7 (x) = 43 f5 (x) 14 f4 (x) sin(3x) = ( 43 )( 15
16
) sin(3x)
( 14 )( 89 ) sin(3x) = ( 63
64
) sin(3x).
5/4/15. Triangle ABC is an obtuse isosceles triangle B

with the property that three squares of equal E
ZS ZF
size can be inscribed in it as shown on the right. D S ZG
SS Z
The ratio AC/AB is an irrational number that
HSZ I ZZ
S
is the root of a cubic polynomial. Determine
Z
S Z Z
that polynomial. SZ Z
A J K KM C
This triangle was apparently discovered by the Italian mathematician Eugenio Calabi. It is
the only non-equilateral triangle into which we can fit three equal squares in this manner.
The triangle is displayed on page 266 of The Book of Numbers, by John H. Conway and
Richard K. Guy, published by Springer-Verlag in 1996.
Let s be the length of a side of the squares. By symmetry, AJ = M C, so 2(AJ) =
AC JM = AC s. AD = AB BD = AB s. Triangle AEJ is congruent to triangle
AKD, so AD = AJ. So AC s = 2(AB s), giving s = 2(AB) AC and AJ = AC AB.
Let P be the midpoint of side AC. Triangle AEJ is similar to triangle ABP . So AE/AJ =
AB/AP = 2(AB/AC). AE = 2(AC AB)(AB/AC).
By the Pythagorean Theorem (AE)2 = (AJ)2 + s2 , which converts to
2
AB
2(AC AB) = (AC AB)2 + (2(AB) AC)2 (1)
AC
(2(AC AB)(AB))2 = (AC)2 (AC AB)2 + (AC)2 (2(AB) AC)2 . (2)
Let x = AC/AB. Equation (2) converts to
4(x 1)2 = x2 (x 1)2 + x2 (2 x)2

(4 x2 )(x 1)2 x2 (2 x)2 = 0
(2 x)(x + 2)(x 1)2 x2 (2 x)2 = 0
(2 x) (x + 2)(x 1)2 x2 (2 x)

= 0
(2 x)(2x3 2x2 3x + 2) = 0
By the triangle inequality, AC < AB + BC = 2(AB), so x < 2 and 2 x 6= 0. Therefore,

AC/AB is the root of the cubic polynomial 2x3 2x2 3x + 2.
3
Solutions to Problem 1/1/16
www.usamts.org
1/1/16. The numbers 1 through 10 can be arranged along 1

the vertices and sides of a pentagon so that the sum of the 5
Z 8
Z
three numbers along each side is the same. The diagram on 10 Z7
B
the right shows an arrangement with sum 16. Find, with proof, 2

6
B
the smallest possible value for a sum and give an example of an B
B
arrangement with that sum. 4 9 3
Credit This problem was invented by George Berzsenyi, inspired by Team Question 1 of
the 2001 South East Asian Mathematics Olympiad.
Comments Some students solved this problem by pointing at that 10 must be on at least
one side, so 10 + 1 + 2 = 13 is the smallest possible side sum. Such proofs go on to show
that such a side sum is impossible, though this approach has the drawback of requiring many
steps that must be discussed. Skipping any step led to a lack of rigor in many solutions.
One student found a novel way to examine this approach by grouping the possible sums of
13 into those where the lowest number was 1, 2, or 3.
Solution 1 by: Aaron Pribadi (9/AZ)
The variables a through j are the values arranged around the pentagon.
Since the sums of the sides are the same:
a + b + c = c + d + e = e + f + g = g + h + i = i + j + a = the sum of one side
Thus,
(a + b + c) + (c + d + e) + (e + f + g) + (g + h + i) + (i + j + a)
= the sum of one side
5
or it can be rearranged to
(a + b + c + d + e + f + g + h + i + j) + (a + c + e + g + i)
5
Since a + b + c + d + e + f + g + h + i + j consists of the sum of the numbers one through
10, it equals 1 + 2 + 3 + 4 + 5 + 6 + 7 + 8 + 9 + 10, or 55. Thus
(55) + (a + c + g + e + i) a+c+g+e+i
= 11 + = the sum of one side
5 5
www.usamts.org
The numbers are positive, so to minimize the expression above, a + c + g + e + i must be

minimized.
It is obvious that the values to minimize the sum a + c + g + e + i are the numbers 15 out
of the possible numbers of 1 through 10, since those are the 5 smallest numbers. So,
1+2+3+4+5
11 + = 11 + 3 = 14 the minimum sum of one side
5
14 is the smallest possible value for a sum.
Now that an optimal minimum value has been found, an example will show that the value
can be attained. One such example is in the following diagram:
Solution 2 by: Nathan Pflueger (12/WA)

Define S for any arrangement of the numbers 1 through 10 on the edges and vertices of a
pentagon to be the sum of the sums of the three numbers along each side of the pentagon.
Notice that, in this summation, every number found on an edge will be added once (when
the numbers on that edge are added), but each number on a vertex will be added twice (once
for each edge connected to that vertex). Thus S = (edge numbers)+2(vertex numbers).
However, observe that (edge numbers)+(vertex numbers)= (all numbers), and the sum
of all numbers is the sum 1 + 2 + + 10 = 55, thus S = 55 + (vertex numbers). This can
be minimized by making the vertex numbers 1 through 5, thus S 55 + 15 = 70. If the sum
of the numbers on each edge are equal, then each such sum is S/5, thus the least possible
such sum is 70/5 = 14. Indeed, this sum is possible by using the arrangement shown below.
www.usamts.org
www.usamts.org
2/1/16. For the equation
(3x2 + y 2 4y 17)3 (2x2 + 2y 2 4y 6)3 = (x2 y 2 11)3 ,
determine its solutions (x, y) where both x and y are integers. Prove that your answer lists
all the integer solutions.
Credit The original version of the problem was invented by Dr. George Berzsenyi, the
creator of the USAMTS, and was modified into its current form by Dr. Erin Schram of
NSA.
Comments Nearly all correct solutions followed one of the three strategies outlined in
the published solutions below. Solutions 1 and 2 by Tony Liu and Zachary Abel exhibit
two methods using substitution and algebraic manipulation. Solution 3 by Meir Lakhovsky
employs Fermats Last Theorem.
Solution 1 by: Tony Liu (10/IL)
To simplify the algebra, let us denote
a = 3x2 + y 2 4y 17
b = 2x2 + 2y 2 4y 6
so that
a b = x2 y 2 11
It follows that the original equation is equivalent to
a3 b 3 = (a b)3
a3 b 3 = a3 3a2 b + 3ab2 b3
0 = 3a2 b + 3ab2
0 = 3ab(b a)
Now, we will have solutions if and only if a = 0, b = 0, or a = b.

Case 1: If a = 0, then
3x2 + y 2 4y 17 = 0
3x2 + (y 2)2 = 21
Because the right hand side is divisible by 3, we conclude that 3|(y 2)2 . Since squares
are nonnegative, it follows that (y 2)2 = 0 or 9. If (y 2)2 = 0, then we have 3x2 = 21,
which has no integral solutions. If (y 2)2 = 9, then y = 1 or 5, and 3x2 = 12, so x = 2.
Our solutions (x, y) are thus
www.usamts.org
(2, 1) (2, 5) (2, 1) (2, 5)
Case 2: If b = 0, then
2x2 + 2y 2 4y 6 = 0
x2 + y 2 2y 3 = 0
x2 + (y 1)2 = 4
Since the only squares less than or equal to 4 are 0, 1, and 4, we must have one term
equal to 0 and the other equal to 4. If x2 = 0, then (y 1)2 = 4, and y = 1 or 3. If x2 = 4,
then x = 2 and (y 1)2 = 0, so y = 1. Thus our solutions for this case are
(0, 1) (0, 3) (2, 1) (2, 1)
Case 3: If a = b, then
3x2 + y 2 4y 17 = 2x2 + 2y 2 4y 6
x2 y 2 = 11
(x + y)(x y) = 11
From this equation, we note that 11 is prime to conclude that x + y = 1 or 11, and
x y = 11 or 1. Solving these four systems of equations, we obtain the solutions
(6, 5) (6, 5) (6, 5) (6, 5)
Our final solution list is thus
(2, 1) (2, 5) (2, 1) (2, 5)

(0, 1) (0, 3) (2, 1) (2, 1)
(6, 5) (6, 5) (6, 5) (6, 5)
Since we have considered all of the cases, these are indeed all the solutions.
www.usamts.org
Solution 2 by: Zachary Abel (11/TX)

If we let a = x2 y 2 11 and b = 2x2 + 2y 2 4y 6, then we (conveniently) have
a + b = 3x2 + y 2 4y 17. So the given equation becomes
0 = (a + b)3 a3 b3 = 3ab(a + b)
Thus, either a = 0, b = 0, or a + b = 0. We do these cases separately.
Case 1: a = 0.
We have (x y)(x + y) = 11. But since 11 decomposes into the product of 2 integers in only
four ways, we obtain the following systems of equations:

xy = 1 x y = 11 x y = 1 x y = 11
; ; ;
x + y = 11 x+y = 1 x + y = 11 x + y = 1
These systems give four solutions: (x, y) = (6, 5), (6, 5), (6, 5), and (6, 5).
Case 2: b = 0.
The equation 2x2 + 2y 2 4y 6 = 0 is equivalent to x2 + (y 1)2 = 4. There are only a few
cases.
If |x| = 0, then (y 1)2 = 4, giving the solutions (x, y) = (0, 3) and (0, 1).
If |x| = 1, then (y 1)2 = 3, which is not solvable in integers.
If |x| = 2, then (y 1)2 = 0, giving the solutions (x, y) = (2, 1) and (2, 1).
If |x| > 2, then 0 (y 1)2 = 4 x2 < 0, which is impossible.
Case 3: a + b = 0.
We have 3x2 + y 2 4y 17 = 0, i.e. 3x2 + (y 2)2 = 21. There are again 4 cases:
If |x| = 0, then (y 2)2 = 21, which is impossible in integers.
If |x| = 1, then (y 2)2 = 18, impossible in integers.
If |x| = 2, then (y 2)2 = 9, leading to (x, y) = (2, 5), (2, 1), (2, 5), and (2, 1).
If |x| 3, then (y 2)2 = 21 3x2 21 3 32 < 0, which is impossible since squares

are non-negative.
So there are 12 solutions: (x, y) = (6, 5), (6, 5), (6, 5), (6, 5), (0, 3), (0, 1), (2, 1),
(2, 1), (2, 5), (2, 1), (2, 5), and (2, 1).
www.usamts.org
Solution 3 by: Meir Lakhovsky (9/WA)

Fermats Last Theorem states that an + bn = cn has integer solutions for n 3 if and only
if a, b, and/or c = 0. We let (3x2 + y 2 4y 17) = c; (2x2 + 2y 2 4y 6) = b; and
(x2 y 2 11) = a, and note that our equation has the form a3 + b3 = c3 . Thus, either
[a = 0, b = c 6= 0] or [b = 0, a = c 6= 0] or [ c = 0, a = b 6= 0 ] or [ a = b = c = 0 ].
Case A: a = 0, b = c 6= 0
x2 y 2 11 = 0 (x y)(x + y) = 11. Since 11 is prime, its only factorizations are 11 1
and (11) (1); therefore (x y) and (x + y) equal either: (1, 11) or (11, 1) or (1, 11) or
(11, 1) respectively. Solving each case individually, we get, (6, 5), (6, 5), (6, 5), and
(6, 5) respectively. Furthermore, in all the cases b = c 6= 0. But, we notice, that whenever
a = 0, we have b 6= 0, thus the case a = b = c = 0 is impossible.
Case B: b = 0, a = c 6= 0
2x2 + 2y 2 4y 6 = 0 x2 + y 2 2y 3 = 0 x2 + (y 1)2 = 4. Since no two non-zero
squares add up to 4, either x2 or (y 1)2 equal 0, while the other equals 4. Thus, the integral
solutions in this case are (2, 1), (2, 1), (0, 3), and (0, 1). All of these solutions give us
a = c 6= 0, thus, all of them are valid.
Case C: c = 0, a = b 6= 0
3x2 + y 2 4y 17 3(x2 ) + (y 2)2 = 21. Through a little trial and error (plugging 0,1,
4, and for x2 ), we see that the only possible value of x2 which leaves y integral is 4. This
yields the solutions (2, 5), (2, 1), (2, 5), and (2, 1). In all of these cases a = b 6= 0,
thus, they are all valid.
There are no other integral solutions because we covered every case which follows from
Fermats Last Theorem. In conclusion, there are 12 solutions: (6, 5), (6, 5), (6, 5),
(6, 5), (2, 1), (2, 1), (0, 3), (0, 1), (2, 5), (2, 1), (2, 5), and (2, 1).
www.usamts.org
3/1/16. Given that 5r + 4s + 3t + 6u = 100, where r s t u 0 are real numbers,

find, with proof, the maximum and minimum possible values of r + s + t + u.
Credit This problem was inspired by a similar problem posed 35 years ago in the first round
of Hungarys Daniel Arany Mathematical Competition for students of advanced standing.
Comments Three elegant algebraic solutions are presented below. Some students also
solved this problem by considering the region of 4dimensional space described by the in-
equalities r s t u 0. The minimum and maximum of r+s+t+u must located at the
corners of this space. Thus, we must test (x, 0, 0, 0); (x, x, 0, 0); (x, x, x, 0); and (x, x, x, x)
by finding the value of x in each case which satisfies the given 5r + 4s + 3t + 6u = 100 and
evaluating r + s + t + u at the resulting points.
Solution 1 by: Yakov Berchenko-Kogan (10/NC)
Let:
u + a = t
u + a + b = s
u + a + b + c = r
Since r s t u 0, we know a, b, c R+
0 . Note that:
r + s + t + u = 4u + 3a + 2b + c
Substituting:
5r + 4s + 3t + 6u = 100
5(u + a + b + c) + 4(u + a + b) + 3(u + a) + 6u = 100
18u + 12a + 9b + 5c = 100
(2u + b + c) + 4(r + s + t + u) = 100
Clearly, in order to maximize r + s + t + u we must minimize 2u + b + c. Since all values are
positive, this can easily be done by setting u = b = c = 0. Now, we can find what exactly
the maximum value is:
4(r + s + t + u) = 100
r + s + t + u = 25
25
Thus 25 is the maximum value of r + s + t + u, achieved when r = s = t = 3
and u = 0.
Now we must find the minimum value:
18u + 12a + 9b + 5c = 100
5(r + s + t + u) (2u + 3a + b) = 100

www.usamts.org
Similarly to before, in order to minimize r + s + t + u we must minimize 2u + 3a + b, and

this is easily done by setting u = a = b = 0. Again, we can easily find what exactly the
minimum value is:
5(r + s + t + u) = 100
r + s + t + u = 20
Thus the minimum value of r + s + t + u is 20, achieved when r = 20 and s = t = u = 0.
So, in summary, 20 r + s + t + u 25.

Define S = r + s + t + u. Since r s t u 0, the numbers r s, s t, t u, and u

are non-negative. To find the lower bound, we calculate as follows:
S = r+s+t+u
= (r s) + 2(s t) + 3(t u) + 4u
(r s) + 59 (s t) + 12
5
(t u) + 18
5
u
1

= 5
5r + 4s + 3t + 6u
1

= 5
100
= 20.
The minimum of 20 can be achieved when (r, s, t, u) = (20, 0, 0, 0). We similarly find the
upper bound:
S = r+s+t+u
= (r s) + 2(s t) + 3(t u) + 4u
54 (r s) + 94 (s t) + 3(t u) + 92 u
= 14 5r + 4s + 3t + 6u

= 41 100

= 25.
This maximum is attained when (r, s, t, u) = ( 25 , 25 , 25 , 0). Thus, the minimum and maxi-
3 3 3
mum values of S are 20 and 25 respectively.
www.usamts.org
Solution 3 by: Feiqi Jiang (9/MA)

Since r t, we have r t 0. Also, u 0 implies 2u 0. Adding this to r t 0 gives
r t + 2u 0
Note that
4(r + s + t + u) + (r t + 2u) = 5r + 4s + 3t + 6u = 100.
Therefore,
100 4(r + s + t + u) = (r t + 2u) 0
100 4(r + s + t + u) 0
100 4(r + s + t + u)
25 r+s+t+u
Hence the maximum value of r + s + t + u is 25.

We take a similar approach for the minimum: s u implies s u 0. Adding this to
2t 0 gives s u + 2t 0.
Note that
5(r + s + t + u) (s u + 2t) = 5r + 4s + 3t + 6u = 100.

Therefore
5(r + s + t + u) 100 = s u + 2t 0
5(r + s + t + u) 100 0
5(r + s + t + u) 100
r+s+t+u 20
Thus the minimum value of r + s + t + u is 20.

www.usamts.org
4/1/16. The interior angles of a convex polygon form an arithmetic progression with a
common difference of 4 . Determine the number of sides of the polygon if its largest interior
angle is 172 .
Credit This problem was modeled after Problem 106 on page 75 of Volume 3, Num-
ber 1 (Spring 1981) of AGATE, which was edited in Alberta, Canada, by Professor Andy
Liu, a long-time friend of the USAMTS program. Dr. Liu referenced the Second Book of
Mathematical Bafflers, edited by A. F. Dunn and published by Dover Publications in 1983.
Comments The most straightforward solution is as in Solution 1. A slightly more elegant
solution, using exterior angles instead of interior angles, is shown in Solution 2. Note that
both method leads to the same quadratic equation. It is also possible to use a guess and
check method on this problem.
Solution 1 by: Yakov Berchenko-Kogan (10/NC)
Let n be the number of sides of the polygon. The first step to solving this problem is to
determine the sum of the interior angles determined by the arithmetic progression in terms
of n. It is fairly clear that the progression is 172, 168, 164, ..., 172 4(n 1). Thus we can
find its sum to be:
n(n 1)
172n 4
2
= 172n 2n(n 1)
= 174n 2n2 .
We also know the formula for the sum of the interior angles of an n-sided polygon: 180(n
2) = 180n 360. Thus we can equate these two and solve for n:
180n 360 = 174n 2n2

2n2 + 6n 360 = 0
n2 + 3n 180 = 0
(n + 15)(n 12) = 0
Obviously n = 15 is an extraneous solution, and so we know that n = 12, and thus the
polygon is a dodecagon.
Solution 2 by: Benjamin Lee (9/MD)
The sum of the exterior angles of the polygon is always 360 . So,
Sumexterior angles = 8 + (8 + 4(1)) + (8 + 4(2)) + + (8 + 4(n 2)) + (8 + 4(n 1)) = 360,
where n is the number of sides of the polygon.

The equation can be simplified to
www.usamts.org
8n + 4(1 + 2 + + (n 2) + (n 1)) = 360

(n 1)(n)
2n + = 90
2
4n + (n 1)(n) = 180
n2 + 3n 180 = 0
(n 12)(n + 15) = 0
Therefore n = 12 or n = 15. Since n must be a positive integer, n = 12 and we are done.

www.usamts.org
A
5/1/16. Point G is where the medians of the triangle ABC intersect bBBb
@
and point D is the midpoint of side BC. The triangle BDG is @bb
@ bE
equilateral with side length l. Determine the lengths, AB, BC, and F BPP
CA, of the sides of triangle ABC. B P@GPb 1 1 Pb
B @ PPb
1
B D C
Credit The problem was proposed by Professor Gregory Galperin, a member of the Com-
mittee in charge of the USA Mathematical Olympiad. He has suggested many excellent
problems to the USAMTS over the years.
Comments Our participants found numerous solutions to this problem. Below are 4 of the
most common approaches. Other solutions involved complex numbers, clever rotations or line
extensions, and complicated analytic geometry. Our first solution below, from Timothy Zhu,
exhibits the basic geometric approach. The second solution, from David Benjamin, employs
the law of cosines, while Lawrence Chan gives us a third solution using analytic geometry.
Solution 4 from Ameya Velingker uses Stewarts Theorem to prove a formula relating median
lengths to the side lengths of a triangle, and Solution 5 from Joshua Horowitz uses vectors
and analytic geometry.
www.usamts.org
Solution 1 by: Timothy Zhu (12/NH)

Lemma 1: BE CF
Proof: Since D is the midpoint of BC, DC has side length 1. Since 4BGD is equilateral,
its angles are all 60 . Therefore, GDC = 180 60 = 120 . Since DG = DC, 4GDC

is isosceles. Thus, DGC and DCG are both 180 GDC2
= 30 , so BGC = BGD +
DGC = 90 . Therefore, BE CF .
end Lemma 1
A
Since AD, BE, and CF are medians, D, E, and F are midpoints.
b
Therefore,
BBb
@
7 13
2 @b 2
AB = 2F B F BP@bb E
@GP1b
P

BC = 2 2B 901 P
7
P3 b
13
2
CA = 2EC B 1@ P 1
b
P
B D C
Since BE CF (Lemma 1), F B, EC, and GC can be calculated using the Pythagorean
Theorem. q
2 2
AB = 2F B = 2 BG + GF
BC = 2
q
2 2
CA = 2EC = 2 GC + GE

q
2 2
GC = BC BG = 22 12 = 3
It is well known that the centroid, G, divides the medians in 2:1 ratios. Thus,
GC BG
GF = and GE =
2 2
So,
!2
s v
2 u

q
2 2 2 GC u 3
AB = 2F B = 2 BG + GF = 2 BG + = 2t12 + = 7
2 2
BC = 2
s s
2 2
2
q
2 2 2 BG 1
CA = 2EC = 2 GC + GE = 2 GC + =2 3 + = 13
2 2
www.usamts.org
Solution 2 by: David Benjamin (8/IN)
Side BC
D is the midpoint of BC, so DC = 1, and BC = 2.
Side AB
AG
G is the centroid of the triangle, so GD = 2, so AG = 2, so AD = 3.
4BDG is an equilateral triangle, so GDB = 60 .
So, by the Law of Cosines,
(AB)2 = (BD)2 + (AD)2 2(AD)(BD) cos GDB

= 12 + 32 2 3 1 cos 60
1
= 1+96
2
= 10 3
= 7

AB = 7
Side AC
GDC and GDB are supplementary, so GDC = 120 .
So, by the Law of Cosines,
(AC)2 = (DC)2 + (AD)2 2(DC)(AD) cos GDC

= 12 + 32 2
1 3 cos 120
1
= 1+96
2
= 10 + 3
= 13

AC = 13
www.usamts.org
Solution 3 by: Lawrence Chan (11/IL)

We will solve this problem using a coordinate plane with B as the origin and an x-axis
running parallel to BC. BD = DC because AD is a median, so the coordinates of C are
twice those of D. Since D = (1, 0), C = (2, 0). GBD = 60 and BG = 1 because 4BGD
is equilateral with side length 1, so using trigonometry we get
!
1 3
G = (cos 60 , sin 60 ) = ,
2 2
G is the centroid, since it is the intersection of the medians. Centroids have the property
that their coordinates are the averages of the coordinates of the vertices. Let A = (a! , a ),
1 2
1 3
B = (0, 0), and C = (2, 0) be our three vertices. Since the centroid G = , has
2 2
coordinates that are the averages of the respective coordinates in the vertices, we can set up
two equations as follows:

1 a1 + 0 + 2
=
2 3

a2 + 0 + 0 3
=
3 2
!
1 3 3
From these we can deduce that A = ,
2 2
Applying the distance formula to the three sides of the triangle gives us

AB = 7, BC = 2, CA = 13

www.usamts.org
Solution 4 by: Ameya Velingker (11/PA)

Let mA = AD, mB = BE, mC = CF be the lengths of the medians from A, B, and C,
respectively. Also, for convenience, we let a = BC, b = CA, and c = AB be the sides of the
triangle. By Stewarts Theorem,
a mA 2 + BD CD = b2 BD + c2 CD

a2

2 a a
a mA + = b2 + c 2
4 2 2
1
m =
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF PrinterA
2b2 + 2c2 a2 (1)
2
Clearly, a = 2BD = 2 and mA = 3DG = 3 since the centroid divides each median in a 2 : 1
ratio. Substituting these expressions in to (1) and simplyfying, we obtain b2 + c2 = 20. Now,
we note the equation
1 2
mB = 2c + 2a2 b2 (2)
2
which can be derived in the same way we derived (1). Observe that mB = 32 BG = 32 .
2 2
2 2
that 2c b=
Substituting this expression along with a = 2 into (2) and rearranging, we find
1. Solving this equation simultaneously with b +c = 20, we get b = 13 and c = 7.
Thus, the side lengths of the triangle are 2, 13, and 7.
www.usamts.org
Solution 5 by: Joshua Horowitz (10/CT)

We apply coordinates to our diagram, so that B = (0, 0), D = (1, 0), and C = (2, 0). BDG
~ = (A
is equilateral, so G = ( 2 , 23 ). But also, since G is the centroid of ABC, G
1 ~ +B~ + C)/3
~
(using vector notation). We can use this to solve for A:
~+B
A ~ +C ~
~ =
G
3
!
1 3 ~ + (0, 0) + (2, 0)
A
, =
2 2 3
!Printer
3 3 3 ~ + (2, 0)
, = A
2 2
!
1 3 3 ~
, = A.
2 2
Determining side lengths is now just an application of the distance formula:

v !2
u 2
u 1 3 3
AB = t 0 + 0 = 7
2 2
q
BC = (0 2)2 + (0 0)2 = 2
!2
v
u 2
u 1 3 3
CA = t 2 + 0 = 13.
2 2
www.usamts.org
1/2/16. The numbers 1 through 9 can be arranged in the triangles

labeled a through i illustrated on the right so that the numbers in aTT
each of the 2 2 triangles sum to the same value n; that is T
T c T
b T dT
a + b + c + d = b + e + f + g = d + g + h + i = n. T
T T
e T f g T h i T
T T T
For each possible sum n, show such an arrangement, labeled with
n
the sum as shown at right. Prove that there are no possible arrange-
ments for any other values of n.
Credit This is a take-off on a Hungarian problem that appeared in the book Brainteasers
for Upperclassmen by Imrecze, Reiman, and Urban in Hungarian in 1986.

Comments There are basically two steps to the solution: finding bounds on the possible
values of n, and then finding which values within those bounds have valid arrangements.
Many students simplified the argument by noting the symmetry between arrangements sum-
ming to n and arrangements summing to 40 n. (Solution 2 below uses this fact.) The
major variation amongst different solutions is the method by which the cases n = 18 and
n = 22 were shown to be impossible. Solution 1 shows a nice casework approach. Solution
2 uses a clever observation to eliminate all of the cases at once.
Solution 1 by: Eric Paniagua (12/NY)
9TT 6TT 1TT 2TT 4TT

T T T T T
T 4 T T 8 T T 9 T T 7 T T 3 T
3T 1T 1T 4T 7T 3T 3T 9T 9T 7T
T T
T T
T
T T
T T
T T
5T 7 2T 6 8T 9 T 2 7 T 3 5 T 6 T 2 5 T 4 8 T 8T 4 6T 5 1T 1T 5 8T 6 2T
T T T T
T T T T T T T T T T T T T T T
17 19 20 21 23
From the problem, we have
3n =a + b + c + d + e + f + g + h + i + (b + d + g)
3n =45 + b + d + g
where {a, b, c, d, e, f, g, h, i} = {1, 2, 3, 4, 5, 6, 7, 8, 9}. Clearly, the maximum and minimum
values of this sum are
max =45 + 7 + 8 + 9 = 69
min =45 + 1 + 2 + 3 = 51
so we have the bounds on n:
51 3n 69
17 n 23
www.usamts.org
Example arrangements for n = 17, 19, 20, 21, 23 are given above.
Proof that no arrangement exists for n = 18.
If n = 18 then b+d+g = 3(18)45 = 9 and {b, d, g} equals exactly one of {4, 2, 3}, {1, 5, 3}, {1, 2, 6}.
By the symmetry of the positions of b, d, g in the triangle all assignments of these to the
numbers in one of these sets are equivalent.
First, assume b = 4, d = 2, g = 3. Then we have
n = 18 = a + b + c + d = a + c + 6
and a + c = 12, so {a, c} = {7, 5} because this is the only decomposition of 12 not using 2,
3, or 4. Similarly, e + f = 18 b g = 11 and {e, f } = {6, 5} contradicting the fact that

a, c, e, f are distinct.
Now assume b = 1, d = 5, g = 3. We have a + c = 18 b d = 12, so {a, c} = {8, 4}.
Similarly, h + i = 18 d g = 10 and {h, i} equals {6, 4} or {8, 2} contradicting the fact
that a, c, h, i are distinct.
Finally, assume b = 1, d = 2, g = 6. Then a + c = 18 b d = 15, so {a, c} = {8, 7}.
Similarly, e + f = 18 b g = 11 and {e, f } equals {7, 4} or {8, 3} contradicting the fact
that a, c, e, f are distinct.
n 6= 18.
Proof that no arrangement exists for n = 22.
If n = 22 then b+d+g = 3(22)45 = 21 and {b, d, g} equals exactly one of {6, 7, 8}, {9, 7, 5}, {9, 4, 8}.
Assume b = 6, d = 7, g = 8. Then a + c = 22 b d = 9, so {a, c} = {5, 4}. Similarly,
e + f = 22 b g = 8 and {e, f } = {5, 3} contradicting the fact that a, c, e, f are distinct.
Assume b = 9, d = 7, g = 5. Then a + c = 22 b d = 6, so {a, c} = {4, 2}. Similarly,
e + f = 22 b g = 8 and {e, f } = {6, 2} contradicting the fact that a, c, e, f are distinct.
Assume b = 9, d = 4, g = 8. Then a + c = 22 b d = 9, so {a, c} equals {7, 2} or
{6, 3}. Similarly, e + f = 22 b g and {e, f } = {3, 2} contradicting the fact that a, c, e, f
are distinct.
n 6= 22.
Solution 2 by: Adam Hesterberg (10/WA)
Answer: The possible values of n are 17, 19, 20, 21, and 23, as shown below:
9TT 9TT 7TT 1TT 1TT

T T T T T
1 T5 2 T 1 T2 7 T 4 T3 6 T 9 T8 3 T 9 T5 8 T
T T T T T T T T T T
T
T
T T
T T
T T
T T
7TT 6 3T 4 8T 6TT 8 4T 3 5T 9TT 2 5T 1 8T 4TT 2 6T 7 5T 3TT 4 7T 6 2T
T T T T T T T T T T T T T T T
17 19 20 21 23
www.usamts.org
First, we prove that 17 n 23. Note that
3n = (a + b + c + d) + (b + e + f + g) + (d + g + h + i)
X
= (all the numbers) + b + d + g
= 45 + b + d + g
45+6 45+24
Since b+d+g is between 1+2+3 = 6 and 7+8+9 = 24, n is between 3
= 17 and 3
= 23.
Thus, the only values for n left to consider are 18 and 22. Note that if 18 could be con-
structed, so could 22, by replacing each entry x by 10 x. Therefore, we need only consider
18.
n = 18 implies b + d + g = 3 18 45 = 9. Without loss of generality, let 9 be a (it will not
be b, d, or g since the sum of the other two would then have to be 0). Then,
18 = a + b + c + d
= 9 + c + (b + d + g) g
=9+9+cg
0=cg
c=g
However, c and g were to be distinct, so this is impossible. Therefore, neither 18 nor 22 can
be constructed, so the only possible values for n are the ones constructed above.
www.usamts.org

2/2/16. Call a number a b 2 with a and b both positive integers tiny if it is closer to
zero than any number c d 2 such that
c and d are positiveintegers with c < a and d < b.
Three numbers which are tiny are 1 2, 3 2 2, and 7 5 2. Without using a calculator
or computer, prove whether or not each of the following is tiny:

(a) 58 41 2 , (b) 99 70 2 .
Credit We are indebted to Dr. David Grabiner of the NSA for this problem. David is a
former multiple winner of the USAMO, whose continued support of the USAMTS is most
appreciated.
Comments Solution 1 shows the most straightforward solution. Solution 2 usesthe shape
of the graph of y = x. Solution 3 uses the continued fraction representation of 2. Other
solutions
are possible, including listing (by hand!) all of the smallest numbers of the form
|a b 2| for each positive integer a up through 100.

(a) We claim that 58 41 2 is not tiny. Indeed, from 1 < 2, we have

|58 41 2|
|58 41 2| >
2

= |29 2 41|

= |41 29 2|

Thus 41 29 2 is closer to zero than 58 41 2. Since 41 < 58, and 29 < 41, we conclude
that 58 41 2 is not tiny.

(b) We claimthat 9970 2 is tiny. Assume, for the sake of contradiction, that there exists a
number cd 2 closer to zero, with c < 99, and d < 70. Since 992 2702 = 98019800 = 1,
we have

1 = |(99 70 2)(99 + 70 2)|

> |(c d 2)(99 + 70 2)|

> |(c d 2)(c + d 2)|
= |c2 2d2 |
www.usamts.org
2 2

Because c and d are positive integers,
this implies that c 2d = 0, or c = d 2, which is
impossible. It follows that 99 70 2 is indeed tiny.
Solution 2 by: Johnny Hu (10/AL)

The three examples for tiny numbers are all in the form, x + 1 x or x x + 1,
where x is an integer. Since the graph for x is half a parabola that opensto the positive
side that rises more and more slowly as x increases, the difference between x + 1 and x
becomes smaller and smaller click hereas
Create PDF with GO2PDF for free, if you wish to remove this line,
x increases.
PDF Printer Since x + 1 and x are consecutive integers and
to buy Virtual
thedifferencebetween x +1 and x becomes smaller as x increases, numbers in the form
of x + 1 x and x x + 1 must be tiny because all values smaller than x will not
produce a number closer to zero.

Since 58 41 2 can
be written as 3364 3362, it
is in the form of x + 2 x. The
graph of x + 2 x is above the graph of x + 1 x so 58 41 2 is not a tiny number
as there exists a number in the form of c d 2, where c < 58 and d < 41, which is closer
to zero.

To verify this, we must find a number in the form of y + 1 y or y y + 1, since
these will most likely to be smaller than x + 2 x (This will be proven later in the page).
a b 2 = 58 41 2

58 41 2 = a b 2

= a2 2b2
Also:
a2 = 2b2 + 2
= 2(b2 + 1)

To make this equation into the form of y y + 1:
Let:
(d)( 2) = a
Then:
2d2 = 2(b2 + 1)
d2 = b2 + 1
b2 = y
d2 = y + 1
www.usamts.org

Therefore, y y + 1 = b2 d2

Substituting our original value, we have 1681 1682
= 41 29 2.
To verify that 41 29 2 is closer to zero than 58 41 2:

|41 29 2| < |58 41 2|

|41 29 2|2 < |58 41 2|2

3363 2378 2 < 6726 4756 2

Since 2(3363 2378 2) = 6726 4756 2, |41 29 2| < |58 41 2| and 58 41 2 is not
a tiny number.

Since 9970 2 can be written as 9801 9800,it is in the form of x + 1 x. Numbers
in this form are always tiny numbers, so 99 70 2 is a tiny number.
This problem follows from a (well known?) theorem concerning the approximation ability
of continued fractions.
Theorem. For a given irrational number , the number p q is tiny if and only if p/q is
a convergent of .
The proof is in two parts.
Lemma 1. If pn /qn is a convergent for the irrational number and p/q 6= pn /qn is an
arbitrary fraction with 0 < q < qn+1 , then
|pn qn | < |p q|.
Proof. The key to this proof is to try to write
(pn qn )x + (pn+1 qn+1 )y = p q
by solving the system

qn x + qn+1 y = q
(1)
pn x + pn+1 y = p
for x and y. Using the fact that pn+1 qn pn qn+1 = (1)n , we find from the above system
that
x = (1)n qpn+1 pqn+1 and y = (1)n pqn qpn

This tells us a lot! First of all, both x and y are integers. Next, neither x nor y is 0. Indeed,
if x = 0 then p/q = pn+1 /qn+1 , which is impossible for q < qn+1 since gcd(pn+1 , qn+1 ) = 1,
and if y = 0 then p/q = pn /qn , which was assumed to be false.
www.usamts.org
We can obtain even more information from the sytem in (1): x and y must have opposite
sign. If both were positive, then q = qn x + qn+1 y > qn+1 , and if both were negative, then q
would be negative.
Now were ready for the final step. Since lies between pn /qn and pn+1 /qn+1 , the numbers
pn qn and pn+1 qn+1 have opposite signs. Since x and y also have opposite signs, the
two numbers (pn qn )x and (pn+1 qn+1 )y have the same sign. Thus,
(p qn )x + (pn+1 qn+1 )y = p q
n
(pn qn )x + (pn+1 qn+1 )y = p q

Create PDF with GO2PDF for free, if you wish to remove this
(pn qn )x +PDF
line, click here to buy Virtual (p n+1 qn+1 )y
Printer = p q

p n qn x < p q

p n qn < p q
Notice that this lemma shows that the number pn qn is tiny. This next lemma shows
that there are no other tiny numbers.
Lemma 2. If p/q is not a convergent of , then p q is not tiny.
Proof. Since p/q isnt a convergent to , we can find two successive convergents pn /qn and
pn+1 /qn+1 with qn < q < qn+1 . Then by the first lemma, |pn qn | < |p q|, and so p q
is not tiny.
These two lemmas show that pn qn is tiny for each n and that there are no other tiny
numbers. So the main theorem has been proven.

Because of this
theorem with = 2, the tiny numbers can be found by calculating the
convergents of 2. The continued fraction representation of 2 is
1
2=1+
1
2+
1
2+
2 +
Using the recurrence relations
pn = an pn1 + pn2
qn = an qn1 + qn2
www.usamts.org
and the fact that an = 2 for n 1, we can easily calculate the convergents. We get
p0 p2 1 7 p3 p1
17 3
= = = =
q0 q21 5 q3 q1
12 2
p4 41
p6 239 p7 p5577 99
= = = =
q4 29
q6 169 q7 q5408 70

Since 99/70 is one of the convergents,
99 70 2 is a tiny number, whereas 58/41 is not a
convergent and so 58 41 2 is not tiny.
www.usamts.org
3/2/16. A set is reciprocally whole if its elements are distinct integers greater than 1 and
the sum of the reciprocals of all those elements is exactly 1. Find a set S, as small as possible,
that contains two reciprocally whole subsets, I and J, which are distinct but not necessarily
disjoint (meaning they may share elements, but they may not be the same subset). Prove
that no set with fewer elements than S can contain two reciprocally whole subsets.
Credit We are thankful to Dr. Kent D. Boklan of the National Security Agency for devising
this nice problem.
Comments There are many possible 5-element sets which satisfy the conditions of the
problem; probably the one most commonly cited in students solutions was {2, 3, 4, 6, 12}.
The key to this problem was rigorously proving that a set with 4 or fewer elements is
impossible. Solution 1 is an especially concise example. Some solutions, such as Solution 2,
prove along the way that {2, 3, 6} is the unique reciprocally whole set with 3 elements.
Solution 1 by: Zhou Fan (11/NJ)
A reciprocally whole set must have at least three elements, since the reciprocals of only
two distinct integers greater than 1 can sum to at most 1/2 + 1/3 = 5/6. Thus a set S
with two reciprocally whole subsets must contain at least four elements. Suppose such a set
exists: S = {a, b, c, d}. The entire four element set cannot be reciprocally whole if there is
a smaller reciprocally whole subset, so S must contain two three-element reciprocally whole
subsets. WLOG, assume that they are {a, b, c} and {a, b, d}. Then a1 + 1b + 1c = 1 and
1
a
+ 1b + d1 = 1, which implies that 1c = d1 , or c = d: a contradiction. Thus S has at least 5
elements. Such a five-element set exists: S = {2, 3, 6, 7, 42} satisfies the problem conditions
since 1/2 + 1/3 + 1/6 = 1 and 1/2 + 1/3 + 1/7 + 1/42 = 1.
Solution 2 by: Jason Ferguson (12/TX)
Consider the set S = {2, 3, 6, 9, 18}. S is a 5-element set that contains two distinct
reciprocally whole subsets, {2, 3, 6} and {2, 3, 9, 18} (since 21 + 31 + 61 = 12 + 31 + 91 + 18
1
= 1).
We will now show that there are no sets with two distinct reciprocally whole subsets with
cardinality 0, 1, 2, 3, or 4.
To this end, we first show that there cannot be a reciprocally whole set of cardinality
zero, one, or two and the only reciprocally whole set of cardinality three is {2, 3, 6}. Clearly
the null set cannot be reciprocally whole, and if the reciprocal of a number is 1, then that
number must be one. Thus, there is only 1 one-element set S which has the property that the
sum of the reciprocals of its elements, and that set is {1}, but this is not a reciprocally whole
set (all elements must be greater than 1). Suppose T is a reciprocally whole set of cardinality
two, and let x be the smaller element and y the larger (the two elements are distinct). Then
{x, y} is a reciprocally whole set, so x1 + y1 = 1. However, because x < y, x1 > y1 , so x1 > 12 .
Then x < 2. But all elements in a reciprocally whole set must be integers greater than one.
www.usamts.org
From this contradiction we conclude that there are no 2-element reciprocally whole sets.
Suppose now that U is a three-element reciprocally whole set. Then, let a be the smallest
element, b the middle element, and c the largest. Then a1 + 1b + 1c = 1, and a < b < c.
Therefore, a1 > 1b > 1c , so a1 > 13 . Then a < 3, so a = 2. Because a1 + 1b + 1c = 1 and a = 2, it
follows that 1b + 1c = 12 . Because 1b > 1c , it follows that 1b > 14 . Then b < 4. Since 2 = a < b,
it follows that b = 3. Because 1b + 1c = 12 and b = 3, it follows that 1c = 21 31 = 16 . Then,
c = 6, so (a, b, c) = (2, 3, 6), and we conclude that the only reciprocally whole three-element
set is {2, 3, 6}. Therefore, there is only one reciprocally whole set with cardinality less than
or equal to 3: {2, 3, 6}.
Create PDF with GO2PDFA setif you
for free, with
wish tocardinality
remove this line, click of
here 0 orVirtual
to buy 1 cannot
PDF Printer have two distinct, nonempty subsets. Since a
reciprocally whole set has to be nonempty, it follows that a set of cardinality 0 or 1 cannot
have two distinct reciprocally whole subsets. Now, if a set V has two distinct reciprocally
whole subsets, then neither of those two subsets can be V itself, for if V was reciprocally
whole, then the other reciprocally whole subset of V must be a proper subset of V . Then
the sum of the reciprocals of the elements of the proper subset would be less than the sum
of the reciprocals of the elements of V , which is 1. From this contradiction we conclude that
if a set V has two reciprocally whole subsets, then both of them must be proper subsets
of V . Thus, if a set with cardinality 2 had two distinct, reciprocally whole subsets, then
both of them would have to have cardinality less than or equal to 1. This is impossible, as
there are no reciprocally whole sets of cardinality 0 or 1. Similarly, because there are also
no reciprocally whole sets of cardinality 2, a set with cardinality 3 cannot have two distinct
reciprocally whole subsets. Finally, if a set with cardinality 4 had two distinct, reciprocally
whole subsets, then both of them would have to have cardinality less than or equal to 3.
This is impossible, as there is only one reciprocally whole set of cardinality less than or equal
to 1.
So we conclude that there are no sets with two distinct, reciprocally whole subsets whose
cardinality is less than or equal to 4, but the set S = {2, 3, 6, 9, 18} is a five-element set
with this property. We also conclude that S is a minimal set with two distinct, reciprocally
whole subsets. QED
www.usamts.org
5/2/16.
Two circles of equal radius can tightly fit inside right triangle ABC, which
has AB = 13, BC = 12, and CA = 5, in the three positions illustrated
below. Determine the radii of the circles in each case.
A A
!! !

!! !!

! ! !
!! !!
! !
!!
B C B C
Case (i) A Case (ii)
!

!
! !!
!!
! !
B C
Case (iii)
Credit This problem was inspired by Problem 5.3.2 in Traditional Japanese Mathematics
Problems of the 18th and 19th Centuries published in 2002 by SCT Publishing of Singapore.
We are thankful to Mr. Willie Yong, the Publisher, for sending us a copy of this wonderful
book.
Comments There are many ways to solve this problem though most of them used right
angles drawn from the centers of the circles to points of tangency and the use of similar
triangles (whether or not they were implied by the use of trigonometry).
Solution 1 by: Jeffrey Manning (9/CA)
The method for solving case (iii) is different than the method for solving cases (i) and
(ii) so we will look at these seperatly.
Cases (i) and (ii)
Draw the line through the point of intersection of the two circles that is tangent to both
circles. Call the point where this line intersects the hypotenuse, E, and call the point where
it intersects the leg that is tangent to both circles (BC in case (i) and CA in case (ii)), D.
A A
E
5-2r
2r
E D
2r 2r
B 12-2r D 2r C B C
www.usamts.org
In each case ED is perpendicular to one leg and parallel to the other. This means that in
case (i) 4ABC 4EBD and in case (ii) 4ABC 4AED. In each of these cases one of the
circles is the incircle of the new triangle. Also because in each case the distance from ED to
the leg it is parallel to is 2r (where r is the length of the radii of the circles) we have DC = 2r
ab
The length of the radius, R, of the incircle of a right triangle is given by R = .
a+b+c
Since a = 5, b = 12 and c = 13, we have R = 2. R will be proportional to r.
We will now solve each case seperatly.

Case (i)
Since DC = 2r we have BD = 12 2r so, since 4ABC 4EBD,

BD r
=
BC R
12 2r r
=
12 2
24 4r = 12r
16r = 24
r = 3/2
Case (ii)
Since DC = 2r we have AD = 5 2r so, since 4ABC 4EBD,

AD r
=
AC R
5 2r r
=
5 2
10 4r = 5r
9r = 10
r = 10/9
Case (iii)
Draw the line tangent to both circles similarly to cases (i) and (ii). Call the points
where it intersects AB, BC and the line formed by AC, D, E and F respectively.
www.usamts.org
A
x
D
13-x x
B E C
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer F
4EBD and 4AF D are both right triangles because ABDF . Also ABC = EBD and
BAC = F AD, so 4EBD 4ABC 4AF D. Since 4EBD 4AF D and the radii
of their incircles are equal we have 4EBD
= 4AF D, so AD = DE.
Let x = AD = DE
Since AD = x, DB = 13 x which gives:

x 13 x
=
CA BC
12x = 65 5x
17x = 65
x = 65/17
r x
=
R CA
r x
=
2 5
2
r = x
5

2 65
r =
5 17
r = 26/17
www.usamts.org
This can be generalized for AB = c, BC = a and CA = b:
Case (i)
We have BD = a 2r so
aR
r=
a + 2r
ab
Substituting for R gives:
a +thisb line,
Create PDF with GO2PDF for free, if you wish to remove + cclick here to buy Virtual PDF Printer
ab
r=
a + 3b + c
ab
Similarly in case (ii) r =
3a + b + c
Case (iii)
Since AD = x we have DB = c x so
x cx
=
b a
bc
x =
a+b
So
R Rc
r= x=
b a+b
abc
r=
(a + b + c)(a + b)
www.usamts.org
Solution 2 by: Chenyu Feng (12/IL)

Case (i)
A
!!

! !
!
!! O P
!!

B C
Call the centers of the two circles O and P , with P being closer to point C. Because the
area of the triangle is 5 12/2 = 30, we will sum the partial areas to find r.
K(trapezoid BCP O) + K(4AOB) + K(4AOP ) + K(4AP C) = K(4ABC)

r 13r 2r(5 r) 5r
(2r + 12) + + + = 30
2 2 2 2
r = 3/2
Thus the radii of both circles in case (i) is 3/2 .
Case (ii)
A
!
!
!! O

!!
!!
! P

B C
Call the centers of the two circles O and P , with P being closer to point C. Because the
area of the triangle is 5 12/2 = 30, we will again sum the partial areas to find r.
K(trapezoid AOP C) + K(4BOA) + K(4BOP ) + K(4CP B) = K(4ABC)

r 13r 2r(12 r) 12r
(2r + 5) + + + = 30
2 2 2 2
r = 10/9
Thus the radii of both circles in case (ii) is 10/9 .

www.usamts.org
Case (iii)
We will use similar triangles to solve this one. Firstly, draw the two circles common
internal tangent, intersecting line AB at D and line BC at E. Extend lines DE and AC
so that they meet at point F . Because BDE = ACE (theyre both right angles),
DEB = A and the two incircles are the same size, 4ADF = 4BDE. It immediately
follows that AD = DE. Because 4BDE and 4ABC each have a right angle and they share
B, it follows that 4BDE 4ABC. Then, we set up a ratio:
DE BD
=
AC BC
DE 13 DE
=
5 12
65
DE =
17
156
BD = 13 DA = 13 DE =
s 17
2 2
65 156 169
BE = DE 2 + BD2 = + =
17 17 17
And since in a right triangle, 2r = a + b c, where a and b are the lengths of the legs
and c is the length of the hypotenuse:
2r = a + b c
a+bc
r=
2
156 65 169
+
r = 17 17 17
2
26
r=
17
And thus the radius of the circles, r, is 26/17 .

www.usamts.org
Solution 3 by: Benjamin Lee (9/MD)
Case (i) Case (ii) Case (iii)
In all cases, Area3triangles = Area4ABC = 12 (BC)(AC) = 21 (12)(5) = 30

In all cases, we use the formula for a triangle area for the 3 triangles (4EBA, 4EAC,
4ECB) and 4ABC, and r, b, h, for the radius, base, and height of the small triangle,
respectively, so we have
1 1 1
BC hBC + AC hAC + AB hAB = 30
2 2 2
5 13
6hBC + hAC + hAB = 30.
2 2
Case (i)
Drawing auxillary lines (orange) from the center of the leftmost circle to the vertices, we
have three small triangles whose area add up to the area of 4ABC. Thus,
Using the values hBC = r, hAC = 3r, hAB = r, we get
5 13
6r + (3r) + r = 30
2 2
20r = 30
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer 3 Create PDF with GO2PDF for free, if you wish to remove this line, click here
r=
2
Case (ii)
Drawing auxillary lines (orange) from the center of the topmost circle to the vertices, we
have three small triangles whose area add up to the area of 4ABC. Thus,
Using the values hBC = 3r, hAC = r, hAB = r, we get

6(3r) + 25 r + 13
2
r = 30
27r = 30
r = 10 9
www.usamts.org
Case (iii)
Drawing auxillary lines (orange) from the center of the leftmost circle to the vertices, we
have three small triangles whose area add up to the area of 4ABC.
To find the distance from the center of the leftmost circle to AC, because 4ABC
4DEF we can use the proportion
hAC r DE 2r
= = .
BC AB AB
From this proportion,
hAC = r + r.
13
24
Using the values hBC = r, hAC = 13
r + r, hAB = r, we get
13 5 24
6r + r + (r + r) = 30
2 2 13
26
r=
17
Thus the radii of the circles of Cases (i),(ii), and (iii), are 32 , 10
9
, and 26
17
, respectively.
www.usamts.org
4/2/16. How many quadrilaterals in the plane have four of the nine points q q q
(0, 0), (0, 1), (0, 2), (1, 0), (1, 1), (1, 2), (2, 0), (2, 1), (2, 2) as vertices? Do q q q
count both concave and convex quadrilaterals, but do not count figures
q q q
where two sides cross each other or where a vertex angle is 180 . Rigorously
verify that no quadrilateral was skipped or counted more than once.
Credit This problem was proposed by Professor Harold Reiter, the president of Mu Alpha
Theta Mathematics Honor Society and a long-time supporter of the USAMTS program.
Comments Though a combinatorial argument is the easiest place to start with this problem,
many students found interesting ways to show that they covered all possible cases. Many of
those methods failed due to the difficulty of making sure every case was covered. Most such
methods would scale poorly to problems with larger grids.
Solution 1 by: Derrick Sund (11/NC)

There are 94 = 126 different ways to choose four of the points to be the vertices of a quadri-
lateral. However, from this we must subtract the number of ways there are to choose those
four points such that three of them are colinear. There are 8 ways to choose three points
such that all of them are colinear, and for all of them, there are 6 ways to choose the fourth
point, so the number of ways to choose four points such that they can form a quadrilateral
9
that actually has four distinct sides is 4 6 8 = 78.
We are not done yet. Some of those sets of four points will give us exactly one quadri-
lateral (such as (0, 0), (0, 1), (1, 1), (1, 0)), while others (such as (0, 0), (1, 2), (1, 1), (2, 1)) give
us three. A set of four points with no three of the points colinear will give us three quadri-
laterals if one of the points is inside the triangle formed by the other three; otherwise, the
four points give us one quadrilateral. There are exactly eight sets of points such that one is
inside the triangle formed by the other three:
(1, 1), (0, 2), (1, 0), (2, 1)
(1, 1), (2, 2), (1, 0), (0, 1)

(1, 1), (2, 0), (0, 1), (1, 2)
(1, 1), (0, 0), (2, 1), (1, 2)
(1, 1), (1, 2), (0, 0), (2, 0)
(1, 1), (2, 1), (0, 0), (0, 2)
(1, 1), (1, 0), (0, 2), (2, 2)
(1, 1), (0, 1), (2, 0), (2, 2)
The other 70 possible sets of points all give us one quadrilateral. Therefore, the answer to
the problem is 1(70) + 3(8) = 94.
www.usamts.org
Solution 2 by: Keone Hon (10/HI)

There are 94 such quadrilaterals. We will separate the quadrilaterals into a number of cases
as a primary means of enumerating them:

Rectangles: There are 10 rectangles in all: one 2 2, one 2 2, four 1 1s, and
four 1 2s. An example of each is shown below:
q q q q q q q q q q q q
@
q q q q q @q q q q q q q
@
q q q q @q q q q q q q q
1 1 4 4
Parallelograms (other than rectangles): there are two types of parallelograms with
non-perpendicular sides:
q q q q q q

q q q q q q

q q q q q q
4 8
There are four parallelograms of the first shape, since the pair of currently horizontal sides
may be any of the four pairs of opposite sides on an edge of the figure. There are eight
parallelograms of the second shape, since for each of the four 1 2 rectangles, there are two
such parallelograms inside. In all, there are then 12 non-rectangle parallelograms.
Trapezoids: There are three types of trapezoids:

q q q q q q q q q
HH @
q q q q qHHq q @q q
@ @ @
q q @q q q q q @q @q
16 8 4
There are 16 trapezoids of the first type, since each of the four 1 2 rectangles contains
four such trapezoids (each one is formed by cutting off a corner, and there are four corners).
There are 8 trapezoids of the second type, since any of the four edges of the array can be
the side that is currently on the bottom, and for each of those four choices, the longer of the
two bases can be chosen from among two bases. There are four trapezoids of the third type,
since the two non-parallel edges can (when extended) meet at any of the four corners of the
array. In all, there are 28 trapezoids.
Kites: There is only one type of kite. There are four of this type, since the right-angle
vertex can be chosen from any of the four corners of the array.
www.usamts.org
q q q

q q q

q q q
8
Other convex figures: There are two other types of convex figures with no parallel sides.
For the first shape, any of the four edges of the array can be chosen as the side with length
2, and from there, any of the two adjacent edges can be chosen as the side with length 1.
Thus there are 8 of this shape. For the second shape, any of the 8 segments on the edge of
the array with length 1 can be chosen as the side with length 1; from there, the rest of the
shape is determined. Thus there are 8 of this shape. In all, then, there are 16 such figures.
q q q q q q
A @
q qA q q q @q
A
q q Aq q q q
8 8
Concave figures: There are four types of convex figures. There are four of the first
shape, since any of the four corners of the array can be chosen as the vertex currently in
the bottom left corner. There are eight of the second shape, since in addition to the same
four choices for the corner vertex, there are also two choices for which side the obtuse angle
will open towards. There are eight of the third shape, since there are four choices for the
side of length 2 (it can lie on any of the four edges of the array) and two choices for which
side the obtuse angle will open towards. There are four of the last shape, since the obtuse
angle can open towards any of the four edges of the array. In all, there are 24 concave figures.
q q q q q q q q q q q q
@ A
q q q q q @q q q q q qA q
@ @A
q
q q q
q q q q @q q q @
Aq
4 8 8 4
These are all the shapes. In all, there are 10 + 12 + 28 + 4 + 16 + 24 = 94 quadrilaterals.
To verify that all the shapes have been counted exactly once, consider the following.
There are 94 = 126 ways to choose 4 points out of the 9. Of all combinations of four points,
the only ones that do not form quadrilaterals are where three points are collinear. Three
points can be collinear if they are on any of the eight lines passing through three points, as
shown below. The fourth point can be chosen from any of the other 6 points not on that
line. Thus, there are 48 combinations of four points that cannot form quadrilaterals. So
there are 126 48 = 78 combinations of four points that do form quadrilaterals.
www.usamts.org
q q q
@
q @q q
@
q q @q
Given any collection of four noncollinear points, exactly one convex figure can be drawn
through them. Thus, 70 combinations of four points correspond to 70 convex quadrilaterals.
However, this is not true for concave figures; more than one concave figure can be drawn
through four points. Each of the remaining eight combinations can be formed into three
distinct quadrilaterals, as demonstrated below. Thus there are 24 concave quadrilaterals. In
all, there
Create PDF with GO2PDF arewish70
for free, if you + 24this=line,94
to remove quadrilaterals,
click here to buy Virtual PDF Printer which agrees with our previous count.
q q q q q q q q q
@ @
q q q q q @q q q @q

q
q q q
q q q
q q
www.usamts.org
1/3/16. Given two integers x and y, let (xky) denote the concatenation of x by y, which
is obtained by appending the digits of y onto the end of x. For example, if x = 218 and
y = 392, then (xky) = 218392.
(a) Find 3-digit integers x and y such that 6(xky) = (ykx).
(b) Find 9-digit integers x and y such that 6(xky) = (ykx).
Credit The 3-digit variety of the problem was inspired by Problem 28 in the Singapore
Mathematical Olympiad (Junior Section) in 2001. The 9-digit extension is due to USAMTS
founder Dr. George Berzsenyi.
Comments Many students took a trial-and-error appraoch. The most common algebraic
approach to part (a) is reflected in Jason Blands solution. Many students used this approach
for part (b), but a few students used the slick approach of using (a) to get (b) as shown in
Nathan Pfluegers solution below. Still others used the number 1,000,001,000,001 as Jason
Bland illustrates below. Solutions edited by Richard Rusczyk.
(a)
Let (x, y) = (142, 857). Multiplication yields 6(x||y) = 6 142857 = 857142 = (y||x).
(b)
Let (x, y) = (142, 857) as above. Let (u, v) = (x||y||x, y||x||y). It was shown above that
6(x||y) = (y||x) thus 6(u||v) = 6(x||y||x||y||x||y) = (y||x||y||x||y||x) = (v||u), thus u and v
are the 9-digit integers we seek: 142857142 and 857142857, respectively. Alternating con-
catenations such as this can also be used to select two such integers for any number of digits
of the form 3 + 6n.
www.usamts.org
Solution 2 by: Jason Bland (10/PA)
(a) Because x and y each have 3 digits, we can write (x||y) = 1000x + y. Therefore, we have
6(1000x + y) = 1000y + x
6000x + 6y = 1000y + x
5999x = 994y
857x = 142y
x = 142
y = 857
(b) (x||y) has 6 digits when x and y have 3 digits each and 18 digits when x and y have
9 digits each, so multiplying the equation involving (x||y) and (y||x) for 3-digit x and y by
1,000,001,000,001 gives the equation involving (x||y) and (y||x) for 9-digit x and y.
6 142,857 = 857,142
6 142,857,142,857,142,857 = 857,142,857,142,857,142
x = 142,857,142 y = 857,142,857
www.usamts.org
2/3/16. Find three isosceles triangles, no two of which are congruent, with integer sides,
such that each triangles area is numerically equal to 6 times its perimeter.
Credit This is a slight modification of a problem provided by Suresh T. Thakar of India.
The original problem asked for five isosceles triangles with integer sides such that the area
is numerically 12 times the perimeter.
Comments Many students simply set up an equation using Herons formula and then
turned to a calculator or a computer for a solution. Below are presented more elegant
solutions. Zachary Abel shows how to reduce this problem to finding Pythagorean triples
which have 12 among the side lengths. Adam Hesterberg gives us a solution using Herons
formula. Finally, Kristin Cordwell shows how to take an intelligent trial-and-error approach
to construct the solutions. Solutions edited by Richard Rusczyk.
In 4ABC with AB = AC, we use the common notations r = inradius, s = semiperimeter,

p = perimeter, K = area, a = BC, and b = AC. The diagram shows triangle ABC with its
incircle centered at I and tangent to BC and AC at M and N respectively.
A
x
h
12
I a/2
12
B M a/2 C
The area of the triangle is given by rs = K = 6p = 12s, which implies r = 12. The area is
also
a a a 2
K = AM = AC 2 M C 2 = 4b a2 .
2 2 4
Since K is an integer (since it is six times the perimeter), 4b2 a2 must be a perfect square.
If a were odd, then 4b2 a2 3 mod 4, which is not possible for a perfect square. Thus, a
is even. So x = AN = b a2 is an integer, and so is h = AM 12 = 12 4b2 a2 12. So,
since AN I is a right triangle, the integers x, 12, and h form a Pythagorean triple.
It is easy to check that 12 can be the leg in only four Pythagorean triangles: (5, 12, 13),
(12, 16, 20), (9, 12, 15), and (12, 35, 37). So these give all the possibilities for x and h.
www.usamts.org
Similar triangles AN I and AM C show that 12 x

= h+12
(a/2)
, i.e. a = 24(h+12)
x
, which gives
168
a = 120, 48, 72, 5 for the four respective cases. Since a is an integer, the fourth case
doesnt work. Now, since b = x + a2 , we find that b = 65, 40, 45 respectively for the three
remaining cases. Therefore, the three triangles have dimensions (120, 65, 65), (48, 40, 40),
and (72, 45, 45), and there are no others that satisfy the conditions of the problem.
Answers: Triangles with sides (72,45,45), (48,40,40), (120,65,65).

Let the sides of the triangle be (a, b, b). Then the six times the
q perimeter of the triangle
b + 2 2 2 b a2 .
a
a a
is 6a + 12b, and the area of the triangle, by Herons formula, is
r
a a a a
6a + 12b = b+ b
2 2 2 2
a 2 a 2 a a
144 b + = b b+
2 2 2 2
a 2 a
144b + 72a = b
2 2
2b + a a 2
=
2b a 24
Trying multiples of 24 for a leads to (a = 48, b = 40), (a = 72, b = 45), and (a = 120,
b = 65). These have areas of 972, 768, and 1500, respectively, all of which are 6 times their
perimeters.
Solution 3 by: Kristin Cordwell (8/NM)
Three isosceles triangles whose area equals six times their perimeter are 45 by 45 by 72,
40 by 40 by 48, and 65 by 65 by 120.
To begin with, we notice that two congruent right triangles stuck together at a common
leg form an isosceles triangle. We then consider common right triangles: 3,4,5 and 5,12,13.
The other feature that we need to note is that, if we scale the perimeter by a factor , then
the area scales by 2 .
If we take two 3,4,5 right triangles and join them along the short side, we get an isosce-
les triangle of P = 18, and A = 12. Since P = 2 32 and A = 22 3, we see that, if we scale P
by 3, A will scale by 32 , and they will be in the proportion A :: P = 22 33 :: 2 33 . This isnt
quite what we want, but if we scale P by 3 one more time, we will end up with 22 35 :: 2 34 ,
or A = 6P .
www.usamts.org
If we take two 3,4,5 right triangles and join them along the long side, we get an isosce-
les triangle of P = 16, and A = 12. Note that we already have a factor of 3 in the A to P
ratio, but that we need another net factor of 23 to get the overall ratio of 6. Since we gain
a net factor of 2 for every doubling of the perimeter, we scale P by 8, which gives P = 128
and A = 768 = 6P .
Finally, if we consider two 5,12,13 triangles and glue them together at the short leg, we
have a triangle of sides 13, 13, and 24, with P = 50 = 2 52 , and A = 60 = 22 3 5. The
powers
Create PDF with GO2PDF ofif you
for free, 2 wish
and 3 are
to remove what
this line, weto buy
click here wish, we need to scale P by 5 (and A by 52 , in order to
butPrinter
Virtual PDF
have the powers of 5 balance. We then obtain P = 250 and A = 1500 = 6P .
In some cases, we can scale by a fraction. For example, if we look at two 7,24,25 right
triangles joined at the short edge, we have P = 98 = 2 72 and A = 168 = 23 3 7. If we
scale P by 72 , we get a new perimeter of P = 73 and an area A = 2 3 73 . This works because
we have a starting power of 2 in the perimeter that can be canceled.
www.usamts.org
3/3/16. Define the recursive sequence 1, 4, 13, . . . by s1 = 1 and sn+1 = 3sn + 1 for all
positive integers n. The element s18 = 193710244 ends in two identical digits. Prove that
all the elements in the sequence that end in two or more identical digits come in groups of
three consecutive elements that have the same number of identical digits at the end.
Credit This problem was devised by Erin Schram of the NSA. It is based on an Olympiad
Problem of a 2003 issue of the Gazeta Matematica magazine that was posted on the Art of
Problem Solving forum.
Comments Many students proved that the last two digits repeat in a cycle of 20, and
used this cycle to prove that the elements in the sequence that end in two or more identical
digits come in groups of three consecutive elements. Fewer students proved the second half
that within each of these groups of three, the three numbers have the identical number of
repeating digits at the end. Jeffrey Manning gives a clear, concise explanation, and Cary
Malkiewich gives us a more formal solution. Solutions edited by Richard Rusczyk.
If a number ends in two or more identical digits its last two digits must be identical.
Working out the sequence modulo 100 gives:
1, 4, 13, 40, 21, 64, 93, 80, 41, 24, 73, 20, 61, 84, 53, 60, 81, 44, 33, 00, 1, . . .
Since the sequence is recursive and s21 s1 1 mod 100 the sequence modulo 100
must repeat every 20 elements which means that all elements that end in two or more iden-
tical digits come in groups of three consecutive elements, where the digits are 4s in the first
element, 3s in the second and 0s in the third. Now we must prove that they end in the same
number of identical digits.
Let n be the number of 4s a the end of some element of the sequence. Since,
3(444
| {z. . . 4}) + 1 = 1 |333{z
. . . 3} and 3(333
| {z. . . 3}) + 1 = 1 |000{z
. . . 0}
n digits n digits n digits n digits
each element must end in at least as many identical digits as the previous element.
For the second element to end in more than n identical digits the last n + 1 digits of the
first element must be X 444
| {z. . . 4}, where X is a digit other than 4 such that 3X + 1 3 mod
n digits
10, but the only single digit that would satisfy this is 4 which is a contradiction. This means
that the second element must end in exactly n identical digits.
www.usamts.org
Similarly, for the third element to end in more than n digits the last n + 1 digits of the
second element must be Y 333| {z. . . 3} where Y 6= 3 and 3Y + 1 0 mod 10, but similarly the
n digits
only single digit that would satisfy this is 3 which is a contradiction. So all three elements
must end in exactly n elements. The proof is complete.
Solution 2 by: Cary Malkiewich (12/MA)
Define 0 : Z10 Z10 and 1 : Z10 Z10 as follows:

0 (a) = 3a mod 10
1 (a) = 3a + 1 mod 10
Lemma: The functions 0 and 1 are bijective.

Proof: This is proven simply by listing out elements.
0 (0) = 0 0 (5) = 5 1 (0) = 1 1 (5) = 6
0 (1) = 3 0 (6) = 8 1 (1) = 4 1 (6) = 9

0 (2) = 6 0 (7) = 1 1 (2) = 7 1 (7) = 2
0 (3) = 9 0 (8) = 4 1 (3) = 0 1 (8) = 5
0 (4) = 2 0 (9) = 7 1 (4) = 3 1 (9) = 8
Since every element of Z10 appears exactly once in the range of each function, each function
is bijective.
As a result of this lemma, we can define 1 1

0 and 1 to be the inverses of the above
functions.
Since 1 (1) = 4, 1 (4) = 3, 1 (3) = 0, and 1 (0) = 1, the units digit in the given
sequence cycles through 1,4,3,0. These are only 4 numbers that could form the repeating
digits at the end of sn .
In order to rigorously prove the assertion, we must prove all three of these statements
(k > 1):
1. Iff sn ends in exactly k 4s, sn+1 ends in exactly k 3s.
2. Iff sn+1 ends in exactly k 3s, sn+2 ends in exactly k 0s.
3. sn can never end in two or more 1s.
www.usamts.org
1. Iff sn ends in exactly k 4s, sn+1 ends in exactly k 3s.

Assume a term sn of the sequence ends in a string of 4s that is exactly k digits long: (k > 1)
. . . x444 . . . 444
When sn is multiplied by 3, every 4 becomes a 12. A 1 is carried in every column, resulting
in a string of (k 1) 3s followed by a 2:
. . . [1 (x)]333 . . . 332
Then 1 is added, and the new number 3sn + 1 = sn+1 ends in a string of 3s that is k digits
long:
. . . [1 (x)]333 . . . 333
The (k + 1) digit from the right, 1 (x), is not a 3. Since 1 is bijective, this would imply
that the digit x in sn is 1
1 (3) = 4, and we have assumed that only the last k digits were 4.
For the converse, suppose that sn+1 ends in exactly k 3s. sn ends in a 4, so suppose
sn ends with exactly m 4s. By the argument above, sn+1 ends in exactly m 3s. Therefore
m = k, and sn ends with exactly k 4s.
2. Iff sn+1 ends in exactly k 3s, sn+2 ends in exactly k 0s.

Suppose sn+1 ends in a string of 3s that is exactly k digits long: (k > 1)
. . . y333 . . . 333
Then 3sn+1 will end in a string of 9s that is exactly k digits long:
. . . [0 (y)]999 . . . 999
3sn+1 + 1 = sn+2 will then end in a string of 0s that is k digits long:
. . . [1 (y)]000 . . . 000
The (k + 1) digit from the right, 1 (y), is not a 0. This would imply that the digit y in sn+1
is 1
1 (0) = 3, and we have assumed that only the last k digits were 3.
For the converse, suppose that sn+2 ends in exactly k 0s. sn+1 ends in a 3, so suppose
sn+1 ends with exactly m 3s. By the argument above, sn+2 ends in exactly m 0s. Therefore
m = k, and sn+1 ends with exactly k 3s.
3. sn can never end in two or more 1s.

The last two digits form this repeating sequence of 20 terms:
01, 04, 13, 40, 21, 64, 93, 80, 41, 24, 73, 20, 61, 84, 53, 60, 81, 44, 33, 00
www.usamts.org
Since 11 is not a member of this sequence, no member of the original sequence sn can end
in two or more 1s.
Now that all three statements have been proven, it follows that all elements in the
sequence that end in two or more identical digits (4, 3, 0) come in groups of three consecutive
elements (44 . . . 44, 33 . . . 33, 00 . . . 00) that have the same number (k) of identical digits at
the end.
www.usamts.org
4/3/16. Region ABCDEF GHIJ consists of 13 equal squares and is Q BX R

inscribed in rectangle P QRS with A on P Q, B on QR, E on RS, and XXX C
XX X
H on SP , as shown in the figure on the right. Given that P Q = 28 and
X
XX X
D
QR = 26, determine, with proof, the area of region ABCDEF GHIJ. AX
X
XX XXE
JX X XX
XG F
IXX
P H S
Credit This problem was inspired by Problem 2 of the First Round of the 2001 Japanese
Mathematical Olympiad.
Comments There were many different successful approaches to solving this problem. The
most aesthetically pleasing comes from Zachary Abel (11/TX). Others took a similar ap-
proach, using projection without Zacharys clever dissection. Noah Cohen gives an example
of this approach. Nicholas Zehender shows that a subset of the figure formed by the little
squares could itself be inscribed in a square and used this fact to solve the problem. Finally,
Michael John Griffin gives us another dissection to solve the problem.

Inside each square in the diagram, draw two horizontal segments and
two vertical segments as shown to the right. Let the two indicated lengths
be a and b. The whole diagram looks like this: a
b
a
b
The total horizontal length of the red segments is 5a + b, which is equal to the width of
the rectangle, i.e. 5a + b = 26. Likewise, the total vertical length of the blue segments is
www.usamts.org
5a + 3b, which is equal to the height of the rectangle, 28. So we have the system

5a + b = 26
,
5a + 3b = 28

whose solution is a = 5 and b = 1. So the side length of each square is a2 + b2 = 26, the
area of each square is 26, and the total area of ABCDEF GHIJ is 26 13 = 338.
Solution 2 by: Noah Cohen (11/ME)

Via angle chasing, it can be seen that BCL GHN OAJ EF M , call this angle
, and call the length of one square x
www.usamts.org
We can now represent the height and and width of the rectangle with the following equations
5x sin + x cos = 26
5x sin + 3x cos = 28
x cos = 1
x sin = 5
(x sin ) + (x cos )2
2
= x2

x = 26

The area of the polygon ABCDEF GHIJ can be represented by 13x2 , or (13)( 26)2 , which
is equal to 338.
Solution 3 by: Nicholas Zehender (11/VA)
Draw a line parallel to QR through point V . CDEF GHIJAT U V is the same if you
rotate it 90 degrees, so Y S = W Y = QR = 26. RY = RS Y S = 28 26 = 2, so XV = 2.
AF K V BX because AF k V B, F K k BX, and KA k XV . EF Z = 18090AF K =
90 AF K = F AK, and EZF = F KA = 90, so F EZ is also similar to AF K and
V BX.
ZF/XV = EF/BV
ZF/2 = 1/2
ZF = 1
F K = ZK ZF = 26 1 = 25
KA/XV = AF/BV
KA/2 = 5/2
KA = 5
www.usamts.org

AF = F K 2 + KA2
AF = 625 + 25
AF = 650

The side of one of the squares is AF/5 = 26, so the area of ABCDEF GHIJ is 13( 26)2 =
338.
Solution 4 by: Michael John Griffin (12/UT)

Let Point K be the point one third of the way between points Q B``` U R
```
A and B so that it is also in line with point E. Also, let points C

`
```
T , U , V , and W be points along lines P Q, QR, RS, and SP ``
respectively, such that line T V is perpendicular to line P Q and T ```

` V
K ``D ``
passes through point K, and line U W is perpendicular to line A````` `E
`
QR and passes through point C, as shown in the figure at right. J G
```
``
`` F

I ``
P H W S
Triangles ABQ, BCU , AKT , KEV , and CHW are all similar, with ABQ =
BCU and KEV = CHW . All of these triangles have one right angle and the other
corresponding angles equal. If two given angles (such as ABQ and CBU ) and a right
angle are all collinear, the two angles are complimentary. In the case of AKT and ABQ,
Euclids Coresponding Angles postulate works well.
KV = CW , U C = QB, and T K = 1/3QB (given the ratio of the hypotenuses similar
triangles). Notice that U C + CW = 28 and T K + KV = 26.
U C + CW = T K + KV + 2
QB
QB + CW = + CW + 2
3
2 QB
= 2
3
QB = 3
2 2
Since U C = QB and U C + CW = 28, CW = 25. CW = 5AQ 3
, so AQ = 15. AQ + QB =
2 2 2 2 2
AB , so AB = 15 + 3 = 225 + 9 = 234. AB just happens to be the area of 9 of the 13
squares, so the total area of all the squares is 13/9 234 = 338 units2
www.usamts.org
5/3/16. Consider an isosceles triangle

ABC with side lengths F
AB = AC = 10 2 and BC = 10 3. Construct semicircles
P , Q, and R with diameters AB, AC, BC respectively, such
that the plane of each semicircle is perpendicular to the plane D E
B C
of ABC, and all semicircles are on the same side of plane ABC
as shown. There exists a plane above triangle ABC that is
tangent to all three semicircles P , Q, R at the points D, E, and
F respectively, as shown in the diagram. Calculate, with proof, A
the area of triangle DEF .
Credit This problem was contributed by Professor Vladimir Fainzilberg of the Department
of Chemistry at the C. W. Post Campus of Long Island University.
Comments The most common mistake in this problem was asserting that points D and E
are at the midpoints of the their respective semi-circles. Some students successfully slogged
through this problem with calculus or coordinates. Lawrence Chan shows us a geometric
solution and Tony Liu mixes in a little trigonometry. Solutions edited by Richard Rusczyk
Solution 1 by: Lawrence Chan (11/IL) Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Pri
We begin this problem by first drawing lines through A, B, and C perpendicular to

the plane of triangle ABC. Then, we draw the lines where the plane containing triangle
DEF (that is, the plane tangent to all three circles) intersects each of the three circles
plane (labeling intersection points as shown above). Because BG and CH were drawn
perpendicular to the plane of triangle ABC, and because of symmetry due to the two
closest
circles being congruent, we know that BGHC is a rectangle. Thus, BG = CH = 5 3.
www.usamts.org
We now turn our focus to the plane of the circle with diameter AB.
SinceBG, GD, DI, and IA are all tangent to the semicircle, we know that BG =
GD = 5 3 and DI = IA. Let us call the length of IA = x. If we draw a line parallel
to BA and passing through I, we form a right triangle GJI and a rectangle BJIA. Thus,
JB = IA = DI = x and JI = BA. Using the Pythagorean Theorem on GJI gives us the
following result:
GJ 2 + JI 2 = GI 2
(GB JB)2 + BA2 = (GD + DI)2

(5 3 x)2 + (10 2)2 = (5 3 + x)2

75 10x 3 + x2 + 200 = 75 + 10x 3 + x2

20x 3 = 200
10 3
x =
3
We now our attention to the plane containing triangles DEF and GHI.
Since the two trapezoids below GI and HI are congruent and the two circles below the
same lines are also congruent, we know that the triangle GHI posses symmetry about F I.
www.usamts.org
Thus, we know that DE and GH are parallel, and we can consequently form similar triangles
GHI and DEI. We can then set up the following relations.
DE ID
=
GH IG
DE x
=
GH x + DG
10 3
DE 3
=
10 PDF3Printer 103 3 + 5 3
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual

DE = 4 3
All that is left is to find F M . We first find the length of F I. Since we have symmetry,
we know that GF I is a right triangle, and thus we can use the Pythagorean Theorem.

FI = GI 2 GF 2
u 103
v !2
u
2
FI = t +5 3 5 3
3

20 3
FI =
3
Now we can use the similar triangles GF I and DM I.
GD FM
=
GI FI
GD FM
=
GD + x FI
5 3 FM
=
20 3 10 3
3
5 3 + 3

FM = 4 3

Finally, the area of DEF = 12 (DE)(F M ) = 12 (4 3)(4 3) = 24
AreaDEF = 24

www.usamts.org
Construct a line through F , parallel to BC, and let it intersect plane BDA at point F 0 .
Since BC||DE, F 0 lies on the plane DEF . Now, let us focus on plane BDA and employ
methods of two-dimensional geometry.
Let D0 be the foot of the perpendicular from D to AB, and denote the midpoint of AB by X.
Additionally, let the foot of the perpendicular from D to BF 0 be Y . Note that F 0 D = F 0 B
are tangents to the circle, since F 0 lies in the planes BF C and DEF . This, along with
BX 0 0
that 4F XB = 4F XD. We note that F X = BX 0 2 + F 0B2 =
= DX implies 0
50 + 75 = 5 5, since F B is a radius of the semicircle with diameter BC = 10 3. Letting
= F 0 XB = F 0 XD, we have,

2 1 1
cos = = cos 2 = 2 cos2 1 = = cos (180 2) =
5 5 5

and since DXD0 = 180 2, it follows that XD0 = 2, so DD0 = 50 2 = 4 3. In
particular, we observe that BXD is obtuse, so D0 lies on segment DA. Next, we note that
BY DD0 is a rectangle (by construction) so BY = DD0 and BD0 = Y D. We will use these
results later on.

Note that 4DEF is isosceles (by symmetry of 4ABC), and since AD0 = 4 2, by symmetry
and similar isosceles triangles (projecting DE onto 4ABC), we deduce that DE = 4 3.
Now, let h denote the altitude of 4DEF . By using the perpendicular bisector of DE (par-
allel to plane ABC) and a perpendicular from F to BC, we can calculate h by using the
www.usamts.org
1 DE
Pythagorean Theorem. The right triangle has one leg of length 200 75 = 3 5
BC
and another of length F 0 Y = 75 72 = 3. Thus, weget h = 45 + 3 = 4 3, and
consequently the area of 4DEF is 21 h DE = 12 4 3 4 3 = 24.
www.usamts.org
1/4/16. Determine with proof the number of positive integers n such that a convex regular
polygon with n sides has interior angles whose measures, in degrees, are integers.
Credit We are grateful to Professor Gregory Galperin, one of the worlds most powerful
problem posers, for suggesting this problem for the USAMTS program.
Comments Most students took the straightforward approach illustrated by Jake Snell and
Kim Scott below. A few looked at the exterior angles instead of the interior angles, as shown
by Zachary Abel. Solutions edited by Richard Rusczyk.
Solution 1 by: Jake Snell (11/NJ)
We know that the sum of interior angles in any n-gon is 180 (n 2). Therefore, since we
are considering only regular polygons, each interior angle is congruent and now
180 (n 2) 180 n 360 360

m(each interior angle) = = = 180
n n n
360
Each interior angle will have an integer measure in degrees only if is an integer. Thus, n
n
must be a factor of 360. We can construct these factors since we know that 360 = 23 32 5.
We seek all nonnegative integers a, b and c such that 2a 3b 5c divides 23 32 5. Since 2, 3,
23
and 5 are all primes, 2a |23 , 3b |32 , and 5c |5. 2a |23 implies a = 2(3a) is an integer. Therefore,
2
a {0, 1, 2, 3}. In a similar manner, we find that b {0, 1, 2} and c {0, 1}. Since 2a 3b 5c
is the prime factorization of a nonnegative integer, and since no two nonnegative numbers
have the same prime factorization, given a unique combination of a, b and c, 2a 3b 5c is a
unique nonnegative integer. The number of unique combinations of a, b and c is simply the
product of the number of different values they could be, or 4 3 2 = 24. However, we must
subtract 2 from this result since 1 = 20 30 50 and 2 = 21 30 50 do not yield polygons.
Our answer is n = 24 2 = 22. //
www.usamts.org
Solution 2 by: Kim Scott (10/MA)

Since any polygon with n 3 sides can be divided into n 2 triangles (each with total angle
measure 180 ), the total degree measure of its interior angles is 180(n 2) . In a regular
polygon, every interior angle has the same degree measure, which must thus be
180(n 2) 360
= 180
n n
This is an integer exactly when 360 n
is an integer, which is true iff n is a factor of 360.
Since 360 = 23 32 5, its factors are of the form 2a 3b 5c , with 0 a 3, 0 b 2,
and 0 c 1. Since there are 4 values for a, 3 for b, and 2 for c, 360 has 4 3 2 = 24
integral factors, and 360 n
is an integer for 24 values of n. However, this count includes n = 1
and n = 2, which do not correspond to valid values for the number of sides of a regular
polygon.
Therefore, there are 24 2 = 22 positive integers n such that a convex regular polygon
with n sides has interior angles whose measures, in degrees, are integers.
The interior angle is an integer if and only if the exterior angle is an integer because these
two angles add to 180 . Since the exterior angle measures 360 /n, the condition holds if and
only if n is a divisor of 360. Since 360 = 23 32 5, this number has 4 3 2 = 24 factors. But
n must be at least 3, so we reject the possibilities that n = 1 or n = 2 and conclude that n
may equal any of the other 22 divisors of 360.
www.usamts.org
2/4/16. Find positive integers a, b, and c such that

q

a + b + c = 219 + 10080 + 12600 + 35280.
Prove that your solution is correct. (Warning: numerical approximations of the values do
not constitute a proof.)
Credit This is based on Problem 80 on page 42 of Problems from the History of Mathematics,
by Levardi and Sain, a book published in Hungarian in Budapest, 1982. Problem 80 was
attributed to the Indian mathematician Bhaskara (1114 - ca. 1185).
Comments Most students squared both sides of the given equation to get the solution.
Examples are given below by Jason Ferguson and Tony Liu. Solutions edited by Richard
Rusczyk.
Solution 1 by: Jason Ferguson (12/TX)
If the ordered triple of real numbers (a, b, c) satisfy the problem condition, then so will
any permutation (a0 , b0 , c0 ) of (a, b, c). Thus, we may assume without loss of generality that
a b c.
Upon squaring both sides of the equation

q

a + b + c = 219 + 10080 + 12600 + 35280,
we obtain

a + b + c + 2 ab + 2 ac + 2 bc = 219 + 10080 + 12600 + 35280

= 219 + 12 70 + 30 14 + 84 5.
As
a, b, and
c are with a b c, it can be the case that a + b + c = 219,
integers
2 ab = 12 70, 2 ac = 30 14, and 2 bc = 84 5. Then

ab = 6 70, (1)

ac = 15 14, (2)

bc = 42 5. (3)
Multiplying (1), (2), and (3) gives
abc = 264600, (4)
and squaring (1), (2), and (3) gives
ab = 2520, (5)
ac = 3150, (6)
bc = 8820, (7)
www.usamts.org
respectively. Dividing (4) by (7), (6), and (5), respectively give a = 30, b = 84, and c = 105.
Then indeed a + b + c = 219, so (a, b, c) = (30, 84, 105) satisfy

q

a+ b+ c = 219 + 10080 + 12600 + 35280,
as desired. QED
Squaring the given equation, we obtain

a + b + c + 2 ab + 2 bc + 2 ca = 219 + 10080 + 12600 + 35280.
Since there are three radical terms on the right side (which are not integers), the three
radicals on the left side can match up correspondingly. Also note that because a, b, c are
positive integers, this implies a + b + c = 219. Without loss of generality, we may assume
b a c ab bc ca to obtain the following system of equations:
1
ab = 10080 = 6 70
2
1
bc = 12600 = 15 14
2
1
ca = 35280 = 42 5
2

Multiplying the three gives abc = 264600 abc = 210 6. Thus, we can solve for a, b, c:

abc 210 6
a = = = 2 21 a = 84.
bc 15 14

abc 210 6
b= = = 30 b = 30.
ca 42 5

abc 210 6
c= = = 105 c = 105.
ab 6 70
Checking, we note that a + b + c = 84 + 30 + 105 = 219 still holds. Finally, we conclude that
a = 84, b = 30, c = 105 satisfy the given equation.
www.usamts.org
3/4/16. Find, with proof, a polynomial f (x, y, z) in three variables, with integer coeffi-
cients, such that for all integers a, b,
c, the
sign of f (a, b, c) (that is, positive, negative, or
zero) is the same as the sign of a + b 3 2 + c 3 4.
Credit This problem was devised by Dr. Erin Schram of the National Security Agency.
Comments Most students used algebraic manipulation to arrive at a solution; Tony Liu
and Laura Starkston give example solutions.
Solutions edited by Richard Rusczyk.

We claim that the polynomial f (a, b, c) = a3 + 2b3 + 4c3 6abc has the desired properties.
Our proof begins with the following lemma:
Lemma: The expressions s = p3 + q 3 + r3 3pqr and t = p + q + r have the same sign for
real numbers p, q, r that are not all equal.
Proof: We note the following identity:
p3 + q 3 + r3 3pqr = (p + q + r)(p2 + q 2 + r2 pq qr rp)

1
= (p + q + r)((p q)2 + (q r)2 + (r p)2 ),
2
or equivalently,
t
s = ((p q)2 + (q r)2 + (r p)2 ).
2
Note that (p q)2 + (q r)2 + (r p)2 0, with equality if and only if p = q = r. By hy-
pothesis, this cannot hold, so (pq)2 +(q r)2 +(r p)2 > 0. Thus, t = 0 if and only if s = 0.
Moreover, when s, t 6= 0, we may divide by t to get st = 12 ((p q)2 + (q r)2 + (r p)2 ) > 0,
and the result follows.

Now, we set p = a, q = b 3 2, and r = c 3 4, so by our lemma,
p3 + q 3 + r3 3pqr = a3 + 2b3 + 4c3 6abc

has the same sign as p + q + r =a + b 3
2 + c 3 4, provided that p = q = r does not hold. If
p = q = r does hold, then a = b 3 2 = c 3 4, which implies a = b = c = 0 because a, b, c are
www.usamts.org
integers. Thus, f (a, b, c) = a3 + 2b3 + 4c3 6abc = a + b + c = 0, and this case is covered as
well. This concludes our proof.
Solution 2 by: Laura Starkston (11/AZ)
If the signs must be the same, the zeros must be the same so...
3

3
a+b 2+c 4 = 0
3
3
a 2 + b 4 + 2c = 0
3

3
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual a 2 b 4 = 2c
PDF Printer
Keep this in mind. Rewrite the original equation:

3

3
a+b 2+c 4 = 0
3
3
3
3
a+b 2 = c 4

a3 + 3 2a2 b + 3 4ab2 + 2b3 = 4c3
3 3

a3 + 2b3 + 4c3 = 3 2a2 b 3 4ab2
3 3

a3 + 2b3 + 4c3 =
3 3
(a 2 b 4)(3ab)
Combine the equations:
a3 + 2b3 + 4c3 = 6abc

a3 + 2b3 + 4c3 6abc = 0
Since the only operations performed were multiplication by a constant (which does not
change the sign or the zeros, only the magnitude of the values) and cubing (which does
not change the zeros; it makes each zero occur 3 times, but each zero is still the same; it
preserves the sign because it is an odd number power), f (a, b, c) where thefunction
is defined
3 3 3 3 3
as f (x, y, z) = x + 2y + 4z 6xyz should have the same sign as a + b 2 + c 2.
www.usamts.org
4/4/16. Find, with proof, all integers n such that there is a solution in nonnegative real
numbers (x, y, z) to the system of equations
2x2 + 3y 2 + 6z 2 = n and 3x + 4y + 5z = 23.
Credit This problem was created by USAMTS Director Richard Rusczyk.

Comments There are a variety of solutions to this problem. Zhou Fan and Nathan Pflueger
present algebraic approaches. Lawrence Chan considers a geometric interpretation of the
problem,
Create PDF with GO2PDF for free, if and
you wishRichard McCutchen
to remove this line, uses
click here to buy Virtual PDF vectors.
Printer
Let 3x = x1 , 4y = y1 , and 5z = z1 . Then x1 + y1 + z1 = 23 and 92 x21 + 3 2

y
16 1
+ 6 2
z
25 1
= n. We
can find bounds for n as follows:
For the upper bound, we note:

2 2 3 6
n = x1 + y12 + z12
9 16 25
6 2
(x + y12 + z12 )
25 1
6
(x1 + y1 + z1 )2
25
3174
=
25
We can obtain equality when (x1 , y1 , z1 ) = (0, 0, 23).
For the lower bound, we note that 207 28

+ 184
21
+ 575
84
= 23, so we can let (x1 , y1 , z1 ) =
207 184 575
( 28 + a, 21 + b, 84 + c) where a + b + c = 0. Then
2 2 3 6
n = x1 + y12 + z12
9 16 25
2 2 2
2 207 3 184 6 575
= +a + +b + +c
9 28 16 21 25 84
2 2 23 13 23 6 23 529
= a + a + b2 + b + c 2 + c +
9 7 16 7 25 7 14
2 2 13 2 6 2 529
= a + b + c +
9 16 25 14
529

14
www.usamts.org
We can obtain equality when a, b, c = 0 and thus (x1 , y1 , z1 ) = ( 207 , 184 , 575 ).
28 21 84
We note that the ranges of values of x1 , y1 , and z1 are all continuous intervals since the
original x, y, z can be any nonnegative real numbers. Thus, we can have a continuous change
between (x1 , y1 , z1 ) = ( 207 , 184 , 575 ) and (x1 , y1 , z1 ) = (0, 0, 0), so n can take any real value
28 21 84
between 529
14
and 317425
. Thus, if n is restricted to the integers, then n can be any integer
between 38 and 126 inclusive.

We shall show that all integers n, 38 n 126, there exist nonnegative reals (x, y, z)
such that 2x2 + 3y 2 + 6z 2 = n and 3x + 4y + 5z = 23, and these are the only integers n with
this characteristic.
Define f (x, y, z) = 2x2 + 3y 2 + 6z 2 , for any nonnegative reals x, y, z, satisfying 3x + 4y +

5z = 23. Note that f (x, y, z) = a carries both the implication that 2x2 + 3y 2 + 6z 2 = a as
well as 3x + 4y + 5z = 23, f only being defined on its given domain.
LEMMA: If f (x1 , y1 , z1 ) = a1 , and f (x2 , y2 , z2 ) = a2 , then for all a between a1 and a2 ,

there exist x, y, z such that f (x, y, z) = a.
PROOF: Define the function g(t) = f (x1 + (x2 x1 )t, y1 + (y2 y1 )t, z1 + (z2 z1 )t). The
arguments given for f lie between the (x1 , y1 , z1 ) and (x2 , y2 , z2 ), so they are nonnegative,
and they satisfy 3x + 4y + 5z = 23, so g is defined for all t [0, 1]. g expands out to
a polynomial function of t, so it is continuous. Since g(0) = a1 and g(1) = a2 , for all a
between a1 and a2 there is some t [0, 1] such that g(t) = a, and thus some (x, y, z) such
that f (x, y, z) = a.

By the Cauchy-Schwarz inequality, (x 2 3 + y 3 4 + z 6 5 )2 (2x2 + 3y 2 +
2 3 6
2
6z 2 )( 92 + 16
3
+ 25
6
) = 14 f (x, y, z), so f (x, y, z) 141
(3x + 4y + 5z)2 = 23
14
= 529
14
. By applying
69 46 115 529
the equality condition of Cauchy, we find that f ( 28 , 21 , 84 ) = 14 .
3
By the trivial ineqality and the fact that 16 < 29 < 256
, f (x, y, z) = 92 (3x)2 + 16
3
(4y)2 +
6 6
25
(5z)2 15 ((3x) + (4y) + (5z) ). Since x, y, z > 0 and due to the expansion of (a + b + c)2 ,
2 2 2
6 6
this is in turn less than or equal to 15 (3x + 4y + 5z)2 = 15 232 = 3174
25
. Thus we have
3174 23 3174
f (x, y, z) 25 . The equality case is achievable; f (0, 0, 5 ) = 25 .
Thus we have determined that 529 14

f (x, y, z) 3174
25
, and there exist x, y, z satisfying
both equality cases. Therefore, by the lemma, for any integer n [ 529 14
, 3174
25
], there exist
nonnegative reals (x, y, z) such that f (x, y, z) = n. 529
14
= 37 11
14
, and 3174
25
= 126 24
25
, so all
possible integer values of n are the integers from 38 to 126.
www.usamts.org
Solution 3 by: Lawrence Chan (11/IL)
There is an interesting geometric solution to this problem. We begin by noting that

2
2x + 3y 2 + 6z 2 = n is an ellipsoid and 3x + 4y + 5z = 23 is a plane. If the ellipsoid becomes
too small, it will not be able to touch the plane. If it becomes too big, its intersection with
the plane will be outside the set of triples of positive reals. We will begin by finding the
lower bound.
Since we want to find the point where the ellipsoid just touches the plane, we want to
find n such that the ellipsoid is tangent to the plane. To simplify things, we will rearrange
variables
you wishdefine new
to remove this variables.
line, click here to buy Virtual PDF Printer
2x2 + 3y 2 + 6z 2 = n
2 2
( 2x) + ( 3y) + ( 6z)2 = n
We will define

u := 2x

v := 3y

w := 6z
Our equation then becomes

u2 + v 2 + w 2 = n

which is a sphere centered at (0,0,0) with radius n.
We will also change the variables in the equation of the plane to get
3 4 5
u + v + w = 23
2 3 6
If a plane is tangent to a sphere centered at the origin, the distance from the plane to
the origin is equal to the radius of the sphere. Thus,
| 32 (0) + 43 (0) + 56 (0) 23|

n = q
( 32 )2 + ( 43 )2 + ( 56 )2
232
n =
14
n 37.7857
We want the the first n that works, so our lower bound is n = 38. To find the upper
bound, we check the axes since those are the farthest reaching (in other words, we let x and
www.usamts.org
y equal 0 to find the upper bound by z, and we do this for each of the three variables). We
want to find the largest n such that
r
n 23
= checking z axis
6 5
r n 126.96
n 23
= checking y axis
3 4
r n 99.19
n PDF Printer
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual 23
= checking x axis
2 3
n 117.56
The highest n that still works is 126, so our answer is as follows:

n= {x : x Z and 38 x 126}

Solution 4 by: Richard McCutchen (10/MD)

A change of variables will make this problem a bit easier. Let x = u/ 2, y = v/ 3,
and z = w/ 6; the new variables u, v, and w are nonnegative iff x, y, and z are. Our two
equations become
u2 + v 2 + w2 = n, (1)
3 4 5
u + v + w = 23. (2)
2 3 6
Introduce a rectangular uvw coordinate system; we consider only the region where
u, v, w 0 (henceforth
the first octant). (1) asserts that the distance from (u, v, w) to
the origin is n, while (2) asserts that (u, v, w) is on a plane (call it P) that does not de-
pend on n. Thus, if we can determine at what distances the first-octant points on P lie from
the origin, well know for which values of n the system is solvable.
For convenience, let

3 4 5 ~b = u, v, w .

~a = , , and
2 3 6
We can then rewrite (1) and (2) as

|~b| = n, (10 )
~a ~b = 23. (20 )
www.usamts.org

It is also useful to know that |~a| = 14:
s 2 2 2 r
9 16 25

3 4 5 3 4 5
, , = + + = + + = 14.
2 3 6 2 3 6 2 3 6
The first-octant point of P nearest the origin

The Cauchy-Schwartz Inequality on vectors states that
(~a ~b) |~a| |~b|,

(3)
where ~a and ~b are as above. Id like to use this to obtain a lower bound on the distance
from points ~b on P from the origin; this distance is |~b|. Since ~b is on P, we can substitute
for the left side using (20 ), and we already know what |~a| is:

23 14|~b| |~b| 23/ 14. (4)
This is interesting.
There is, in fact, a point of P in the first octant that lies at a distance of 23/ 14 from
the origin. It is ~b = (23/14)~ 2
a. This point lies on P because (23/14)~a ~a = (23/14)|~a| = 23,
and |~b| = (23/14)|~a| = 23/ 14.
The first-octant point of P furthest from the origin

What first-octant point of P is furthest from the origin? To find this out, I would like to
prove a lemma.
Lemma. Let A = A1 A2 . . . Ak be a polygon lying in 3-space, and let P be a point in 3-

space. Of all the points on or inside A, the point furthest from P is one of the vertices Ai .
(If multiple points on or inside A tie for furthest, they are all vertices of A.)
Proof. Let Q be any point of A other than a vertex; I will prove that there is a point of A
further from P than Q. There must be a line segment U V lying on or inside A and having
Q as its midpoint. (If Q is on a side of A, take a short segment along this side. Otherwise,
any sufficiently short segment with midpoint Q and in the plane of A will do.)
~ = QV
Let w ~ = U~Q. Suppose w ~ P~Q 0. Then V is further from P than Q is, because
P V 2 = |P~Q + w
~ |2 = (P~Q + w
~ ) (P~Q + w
~ ) = P Q2 + |~
w|2 + 2P Q|~
w| > P Q2 + 2P Q|~
w| P Q2 .
~ P~Q 0.) If w
(The last inequality follows because w ~ P~Q < 0, the proof is similar; U is
further from P than Q is. Either way, we have proved the lemma. 2
www.usamts.org
It should be clear that the intersection of the fixed plane P with the first octant is a
triangle (call it T ) with one vertex on each axis. Using (2), these vertices are
! ! !
23 2 23 3 23 6
, 0, 0 , 0, ,0 , 0, 0, .
3 4 5

A quick calculator approximation shows that (0, 0, 23 6/5)
is the furthest from the origin
among these three. The lemma then says that (0,0, 23 6/5) is furthest from the origin
among all points of T . Of course, the distance is 23 6/5.
Putting it together
So, we know that, of the points of the triangle T , (23/14)~a
(~a was the messyvector from
a while ago) is the closest to the origin ata distance of 23 14, and (0, 0, 23 6/5) is the
furthest from the origin at a distance of 23 6/5. Now, back to the system
u2 + v 2 + w2 = n, (1)
3 4 5
u + v + w = 23. (2)
2 3 6

(2) says that (u, v, w) is on or inside T . (1) says that (u, v, w) is ata distance n from the
origin. Thus, the system is solvable for a given n iff T has a point n units from the origin.
Thus, if n could be a real number, it could range from

(23/ 14)2 = 529/14 37.78 to (23 6/5)2 = 3174/25 = 126.96.
Since n must be an integer, the system is solvable for 38 n 126.

www.usamts.org
5/4/16. Medians AD, BE, and CF of triangle ABC meet at

G as shown. Six small triangles, each with a vertex at G, are A
formed. We draw the circles inscribed in triangles AF G, BDG,
and CDG as shown. Prove that if these three circles are all
congruent, then ABC is equilateral.
E F
C D B
Credit This problem was contributed by Professor Gregory Galperin, a long-time contrib-
utor of problems to the USAMTS.
Comments Most solutions involved first showing that 4CGD = 4BGD by first showing
that the perimeters of these triangles are equal. Students took a variety of approaches to
showing AF = CD, some using a purely geometric approach, as Benjamin Dozier illustrates,
some using a more trigonometric approach like that of Shotaro Makisumi, and some using a
little mix of the two, like Dan Li does. Solutions edited by Richard Rusczyk
Solution 1 by: Benjamin Dozier (9/NM)
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Vir
M
N
E K F
G L
O I P
C J D B
The area of 4CDG equals the area of 4BDG as they share the altitude from G to
BC and they have bases of equal length. 4CDG can be dissected into 4OCD, 4ODG
and 4COG. Likewise, 4BDG can be dissected into 4P BD, 4P DG and 4BP G. The
www.usamts.org
area of 4OCD equals the area of 4P BD because they share a base and the respective
altitudes to that base are of the same length. Likewise, the area of 4ODG equals the area
of 4P DG. Thus the area of 4COG equals the area of 4BP G. Since these two triangles have
altitudes of the same length, they must have bases of the same length. Therefore CG = BG.
We know CD = DB, so 4CDG = 4BDG by Side-Side-Side congruence. Furthermore,
mGDC + mGDB = 180 and so mGDC = mGDB = 90 . Since median AD is also

the altitude, we know that 4ABC is isosceles with AC = AB.

Now, since O and N lie on the angle bisectors of CGD and AGF respectively, and
CGD = AGD, we know that OGI = N GL. Also, N L = OI and both OIG and
GLN
Create PDF with GO2PDF for free,are right,
if you wish sothis4OGI
to remove
to buy
line, click here= 4N GL
Virtual

= 4OGH
PDF Printer = 4N GK. Now 4OCH = 4OCJ by
ASA congruence. Likewise 4ODI = 4ODJ, 4N LF = 4N M F , and 4N AM = 4N AK.
All of these triangles have an altitude of common length, the inradius, which we will call r.
The area of 4CDG is the same as the area of 4GF A as the medians dissect a triangle into
six smaller triangles all of the same area. Thus:
1 1 1 1 1 1
(2)( r)GH + (2)( r)JD + (2)( r)CJ = (2)( r)GK + (2)( r)AM + (2)( r)M F
2 2 2 2 2 2
Since GH = GK:
(r)JD + (r)CJ = (r)AM + r(M F )

JD + CJ = AM + M F
AF = CD
which implies that AB = AC = CB and thus the triangle is equilateral.
Solution 2 by: Shotaro Makisumi (9/CA)

Since the centroid divides each median into segments of proportion 1 : 2, each of the six
small triangles has a base that is half of and a height a third of 4ABC, and so they all have
the same area. We know that the radii of the incircles of 4AF G, 4BDG, and 4CDG are
equal. Since 2A = rp, where A is the area of the triangle, p is the perimeter, and r is the
radius of the incircle, are all equal, these triangles all have equal perimeter. That is,
CD + CG + DG = BD + BG + DG = AF + AG + F G (1)
But CD = BD, so CG = BG. By SSS congruence, 4CDG = 4BDG. This implies

CDG = 90 .
We let x = F G and = mCGD = mAGF . Then we have CG = 2x. Since 4CDG is a
right triangle, CD = 2x sin and DG = 2x cos , and so AG = 2DG = 4x cos . We apply
www.usamts.org
the Law of Cosines on 4AF G:

(AF )2 = (F G)2 + (AG)2 2(F G)(AG) cos(mAGF )
(AF )2 = x2 + (4x cos )2 2x(4x cos ) cos
(AF )2 = x2 + 8x2 cos2

AF = x 1 + 8 cos2
Now we can rewrite the second equality of (1) as follows:

2x + 2x cos + 2x sin = x 1 + 8 cos2 + 4x cos + x
Since x 6= 0, we can divide through by x and simplify:

1 + 2 sin 2 cos = 1 + 8 cos2
2 2
4 sin + 4 sin + 1 + 4 cos 4 cos 8 sin cos = 1 + 8 cos2
2 cos2 cos + 1 + sin 2 sin cos = 0
(sin + cos + 1)(1 2 cos ) = 0
This is satisfied when sin + cos + 1 = 0 or 1 2 cos = 0. For (0 , 90 ), the former
has no solution, since sin > 0 and cos > 0. We solve the second equation to obtain
1
cos =
2
Finally, AG = 4x cos = 2x = CG = BG. Since the longer portions of the medians are
congruent, the shorter portions are also congruent, and all six smaller triangles are congruent.
This occurs only if 4ABC is equilateral.
Q.E.D.
Solution 3 by: Dan Li (10/CA)
Lemma 5.1. The inradius, r, of

a triangle with sides m, n, o and angle opposite the side
n+om
of length m is r = tan .
2 2
Proof. Let the triangle be 4M N O, with N O = m, M N = n, M O = o, and N M O = .
Let the incenter of 4M N O be I. Let the points of tangency on M O, M N , N O be P , Q,
R, respectively.
O'
'
''
P ? '''
? ?
??I ggg'g'' R
lll gg'''
lllll '
ll
M Q N
www.usamts.org
Because I is equidistant from M O and M N (IP = IQ = r), it lies on the bisector of

N M O. Therefore,
N M O
IM Q = = (2)
2 2
Because the segments from one point to two points of tangency have equal length (e.g.
M P = M Q),
M Q = n QN = n RN = n (m RO)
= n m + P O = n m + (o M P ) = n m + o M Q (3)
2M Q = n + o m
(4)
n+om
MQ = (5)
2
Thus,
n+om
r = IQ = M Q(tan IM Q) = tan (6)
2 2
Let i = GD, j = GE, k = GF , a = F A, b = EA, c = DB. Because the distance

from the centroid (G) to a vertex is twice the distance from the centroid to the midpoint of
the side opposite the vertex, GA = 2i, GB = 2j, GC = 2k. By the definition of median,
F B = a, EC = b, DC = c.
Let = CGD = AGF . Because the inradii of 4CGD and 4AGF are equal and by
Lemma 5.1,

CG + GD DC CGD AG + GF F A AGF
tan = tan (7)
2 2 2 2

2k + i c 2i + k a
tan = tan (8)
2 2 2 2
2k + i c = 2i + k a (9)
k+a = i+c (10)
It is well-known that all the medians together divide [a triangle] into six equal parts
(http://mathworld.wolfram.com/TriangleCentroid.html). Therefore, the areas of 4AGF ,
4CGD, and 4BGD are equal. It is similarly well-known that the product of the semi-
perimeter and the inradius equals the area of a triangle (see (7) at
http://mathworld.wolfram.com/TriangleArea.html; a proof is given at
http://mathworld.wolfram.com/Inradius.html). Since the inradii of the three triangles
www.usamts.org
are equal (since their incircles are congruent) and the areas are equal, their semiperimeters
must be equal. Therefore,
2i + k + a 2k + i + c 2j + i + c
= = (11)
2 2 2
The second and third parts of (11) result in
2k + i + c 2j + i + c
= (12)
2 2
k = j
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer (13)
Therefore, 4EGC
= 4F GB by SAS (EG = j = k = F G, CG = 2k = 2j = BG,

EGC = F GB). Hence, EC = F B and
b=a (14)
The first and second parts of (11) yield
2i + k + a 2k + i + c
= (15)
2 2
k+c = i+a (16)
Subtracting (10) from (16) yields
ca = ac (17)
c = a (18)
Combining (14) and (18),

a = b = c
2a = 2b = 2c
AB = AC = BC
Hence, 4ABC is equilateral.
www.usamts.org
1/1/17. An increasing arithmetic sequence with infinitely many terms is determined as

follows. A single die is thrown and the number that appears is taken as the first term. The
die is thrown again and the second number that appears is taken as the common difference
between each pair of consecutive terms. Determine with proof how many of the 36 possible
sequences formed in this way contain at least one perfect square.
Credit This problem was taken from the book St. Marys College Mathematics Contest
Problems.
Comments This is a straight-forward problem using modular arithmetic, requiring only

some basic casework. We would like to point out that technically, the term quadratic
residue only applies when the modulus is prime, and 0 is not included. For example, the
quadratic residues modulo 7 are 1, 2, and 4. Otherwise, the term square modulo m should
be used. Solutions edited by Naoki Sato.
Solution 1 by: Derrick Sund (12/WA)
Note: throughout this problem, I will use (a, b) to denote the infinite arithmetic sequence
obtained from first rolling the number a, and then rolling the number b.
It is a well-known fact that if i is a quadratic residue (mod j), there are infinitely many
perfect squares congruent to i (mod j), and that if k is not a quadratic residue (mod j),
then there are no perfect squares congruent to k (mod j). Thus, if a is a quadratic residue
(mod b), then the sequence (a, b) (which consists of all numbers greater than or equal to a
which are congruent to a (mod b)) must contain a perfect square, and likewise, if a is not a
quadratic residue (mod b), the sequence (a, b) cannot contain a perfect square.
Therefore, the sequence (a, b) will contain a perfect square if and only if a is a quadratic
residue (mod b). Since it is also well-known that you can determine all quadratic residues
(mod n) simply by squaring all numbers from 1 to n, inclusive, and finding their residues
(mod n), we can finish the problem by finding the quadratic residues for mods 2, 3, 4, 5, and
6 ((mod 1) need not be considered, since (a, 1) trivially contains all perfect squares greater
than or equal to a).
The quadratic residues (mod 2) are 0 and 1. Therefore, (1,2), (2,2), (3,2), (4,2), (5,2),
(6,2) all contain perfect squares.
The quadratic residues (mod 3) are 0 and 1. Therefore, (1,3), (3,3), (4,3), (6,3) all con-
tain perfect squares, while (2,3), (5,3) do not.
www.usamts.org
The quadratic residues (mod 4) are 0 and 1. Therefore, (1,4), (4,4), (5,4) all contain
perfect squares, while (2,4), (3,4), (6,4) do not.
The quadratic residues (mod 5) are 0, 1, and 4. Therefore, (1,5), (4,5), (5,5), (6,5) all
contain perfect squares, while (2,5), (3,5) do not.
The quadratic residues (mod 6) are 0, 1, 3, and 4. Therefore, (1,6), (3,6), (4,6), (6,6) all
contain perfect squares, while (2,6), (5,6) do not.
Thus, since 6 is the highest number that a die can roll, we have 27 sequences with perfect
squares: (1,1), (1,2), (1,3), (1,4), (1,5), (1,6), (1,2), (2,2), (3,2), (4,2), (5,2), (6,2), (1,3),
(3,3), (4,3), (6,3), (1,4), (4,4), (5,4), (1,5), (4,5), (5,5), (6,5), (1,6), (3,6), (4,6), (6,6).
Solution 2 by: Jeff Nanney (12/TX)
Denote the result of the first die toss d. Denote the result of the second die toss a.
Naturally, a, d N such that 1 a, d 6. We now seek to determine which ordered pairs
(a, d) will yield at least one perfect square of the form a(n 1) + d, where n N. Though a
variety of approaches are available, the most natural is to examine the 6 cases according to
the values of a. In particular, we will use the basic property that squaring all the members
of a residue system yields each possible residue for a perfect square in that modulus. In
general, we are seeking to find a solution in positive integers to the equation an + d = x2 ,
which is equivalent to finding for which d there exists some x such that x2 d (mod a).
1. Let a = 1. Thus, for 1 d 6, we must find some integer x such that x2 d (mod 1).
Since all positive integers are congruent modulus 1, we know that all d are candidates
to produce perfect squares in the sequence. To verify, we implement a simple check, im-
mediately noting that 9 is a perfect square attainable by all the sequences, regardless of
the value of d. Thus, we have 6 sequences so far for which a perfect square is produced.
2. Let a = 2. For 1 d 6, we must find some integer x such that x2 d (mod 2).
Because 02 = 0 and 12 = 1, and all members of the residue system are perfect squares,
we know that all d are candidates to produce perfect squares in the sequence. To verify,
we implement a simple check, immediately noting that 9 is a perfect square attainable
when d is odd, and 16 is a perfect square attainable when d is even. Thus, we have 6
more sequences for which a perfect square is produced.
Because 02 = 0, 12 = 1, and 22 1 (mod 3), we know that d 0, 1 (mod 3), or
www.usamts.org
d = 1, 3, 4, 6, are candidates to produce perfect squares in the sequence. To verify, we

implement a simple check, noting that 9 is a perfect square attainable when d 0
(mod 3), and 16 is a perfect square attainable when d 1 (mod 3). Thus, we have 4
Because 02 = 0, 12 = 1, 22 0 (mod 4), and 32 1 (mod 4), we know that d 0, 1
(mod 4), or d = 1, 4, 5, are candidates to produce perfect squares in the sequence. To
verify, we implement a simple check, noting that 16 is a perfect square attainable when
d 0 (mod 4), and 9 is a perfect square attainable when d 1 (mod 4). Thus, we
have 3 more sequences for which a perfect square is produced.
5. Let a = 5. For 1 d 6, we must find some integer x such that x2 d (mod 5). Be-
cause 02 = 0, 12 = 1, 22 4 (mod 5), 32 4 (mod 5), and 42 1 (mod 5), we know
that d 0, 1, 4 (mod 5), or d = 1, 4, 5, 6, are candidates to produce perfect squares
in the sequence. To verify, we implement a simple check, noting that 25 is a perfect
square attainable when d 0 (mod 5), 16 is a perfect square attainable when d 1
(mod 5), and 4 is a perfect square attainable when d 4 (mod 5). Thus, we have 4
Because 02 = 0, 12 = 1, 22 4 (mod 6), 32 3 (mod 6), 42 4 (mod 6), and
52 1 (mod 6), we know that d 0, 1, 3, 4 (mod 5), or d = 1, 3, 4, 6, are candidates
to produce perfect squares in the sequence. To verify, we implement a simple check,
noting that 25 is a perfect square attainable when d 1 (mod 6), 9 is a perfect square
attainable when d 3 (mod 6), 16 is a perfect square attainable when d 4 (mod 6),
and 36 is a perfect square attainable when d 0 (mod 6). Thus, we have 4 more
sequences for which a perfect square is produced.
Combining the conclusions from each of the above 6 cases, we find that of the 36 possible
sequences, exactly 6 + 6 + 4 + 3 + 4 + 4 = 27 sequences contain at least one perfect square.
www.usamts.org
2/1/17. George has six ropes. He chooses two of the twelve loose ends at random (possibly
from the same rope), and ties them together, leaving ten loose ends. He again chooses two
loose ends at random and joins them, and so on, until there are no loose ends. Find, with
proof, the expected value of the number of loops George ends up with.
Credit This problem, or some variation, is often used in interviews for quantitative positions
on Wall Street.
Comments The easiest approach, as the solution by James Sundstrom illustrates, is to

develop a recursive formula for the number of expected loops formed when starting with n
ropes. However, it is also possible to find the expected value by counting all possible loops.
This was done in a clever way by Scott Kovach. Solutions edited by Naoki Sato.
Solution 1 by: James Sundstrom (11/NJ)
Let En denote the expected value of the number of loops from this process starting with
n ropes. Then we have the following lemma:
1
Lemma. For all natural numbers n, En = En1 + 2n1
.
Proof. If the process starts with n ropes, after one loose end is selected there are 2n 1
1
loose ends remaining, giving a probability of 2n1 that the other end of the same rope will
be selected as the second choice. If this occurs, there is one loop already formed and n 1
loose ropes left, so the expected value for the number of loops formed is 1 + En1 . There is
a probability of 2n2
2n1
that an end of a different rope will be chosen, leaving n 1 ropes (1
longer one and n 2 short ones). Then the expected value of the number of loops is En1 .
Combining the two possibilities gives:
1 2n 2
En = (1 + En1 ) + En1
2n 1 2n 1
1 1 2n 2
= + + En1
2n 1 2n 1 2n 1
1
= En1 + .
2n 1
2
www.usamts.org
It is obvious that E1 = 1, so we have:

1 4
E2 = 1 + = ,
3 3
4 1 23
E3 = + = ,
3 5 15
23 1 176
E4 = + = ,
15 7 105
176 1 563
E5 = + = ,
105 9 315
563 1 6508
E6 = + = .
315 11 3465
Therefore, if George has six ropes which he ties together by randomly selecting two loose
ends at a time to tie together, the expected value of the number of loops he will end with is
6508
3465
.
Solution 2 by: Scott Kovach (10/TN)
First consider the number of different ways to tie n ropes together. The first tie can be
done in 2n leaving 2n 2 loose ends. The next can be done in 2n2

2
ways, 2
ways, the next
2n4
in 2 , and so on. The order that the ties are made doesnt matter, however, so we must
divide the product of these binomial coefficients by n!. The number of ways therefore is
(2n)! (2n2)! 2!
2n 2n2 2 2(0)!

2 2
2 2(2n2)! 2(2n4)! (2n)!
f (n) = = = .
n! n! 2n n!
Now consider the number of ways to tie n ropes together to make a loop. The ropes can
be sequentially tied to each other in any order, so there are (n 1)! ways to order them.
The first end of the first rope can be tied to either end of the second, the remaining end of
the second to either end of the third, and so on. There are 2n1 ways to do this, so the total
number of ways to tie the n ropes together into a single loop is g(n) = 2n1 (n 1)!.
Finally, we count the total number of loops among all the possible tyings. There are
6
1
ways to choose one rope and g(1) ways to tie it into one loop, and f (5) ways to tie the
remaining ropes together. Similarly, there are 62 ways to choose two ropes, g(2) ways to
tie them into one loop, and f (4) ways to tie the other four ropes together. Extending this
process counts every possible loop of any size. The total number of loops is
6
X 6
g(k)f (6 k) = 19524.
k=1
k
www.usamts.org
There are f (6) = 10395 tyings, so the expected value is

19524 6508
= .
10395 3465
www.usamts.org
3/1/17. Let r be a nonzero real number. The values of z which satisfy the equation
r4 z 4 + (10r6 2r2 )z 2 16r5 z + (9r8 + 10r4 + 1) = 0
are plotted on the complex plane (i.e. using the real part of each root as the x-coordinate
and the imaginary part as the y-coordinate). Show that the area of the convex quadrilateral
with these points as vertices is independent of r, and find this area.
Credit This problem was proposed by Dave Patrick of AoPS and Erin Schram of the NSA.
Comments Many students seem to have been intimidated by the complicated looking
quartic equation and the setting of the complex plane. The first step is to find the factors
of the quartic. This is really the bulk of the problem, and was accomplished with a variety
of approaches, as the following solutions illustrate. The next step is to plot the roots in
the complex plane, which are found to form a trapezoid. Some students merely plugged the
quartic into software such as Mathematica, but you still need to show justification that the
roots so produced are in fact correct. Solutions edited by Naoki Sato.
Solution 1 by: Daniel Jiang (11/IN)
The constant coefficient of the equation can be factored so that the equation becomes:
r4 z 4 + (10r6 2r2 )z 2 16r5 z + (9r4 + 1)(r4 + 1) = 0.
Factoring would lead us to the roots of the equation as a function of r. From what we have
so far, we can see that the factors of the equation may look like
[r2 z 2 + + (r4 + 1)][r2 z 2 + + (9r4 + 1)].
The given equation has powers of z 4 , z 2 , and z, so at this stage, we let the factors take the
form
[r2 z 2 + az + (r4 + 1)][r2 z 2 + bz + (9r4 + 1)].
Expanding, we get:
r4 z 4 + (ar2 + br2 )z 3 + (10r6 + 2r2 + ab)z 2 + (a + b + 9ar4 + br4 )z + (9r8 + 10r4 + 1) = 0.
Comparing this to the given equation, we know that there is no z 3 term, so a + b = 0, and
from the z 2 term, we see that ab = 4r2 . Using these, it is easy to see that (a, b) = (2r, 2r)
or (2r, 2r), but then testing in the z term, we see that a must be 2r, so (a, b) = (2r, 2r).
The factored form of the equation is then
(r2 z 2 2rz + r4 + 1)(r2 z 2 + 2rz + 9r4 + 1) = 0.

www.usamts.org
Now the quadratic formula can be applied to each of the quadratic factors, which gives us
the four roots as: 1r ri and 1r 3ri.
Plotting the points 1r , r , 1r , r , 1r , 3r , 1r , 3r , we have a trapezoid with two

2
bases of length |6r| and |2r| and a height of |r| . The area of a trapezoid is b1 +b
2
2
h, so the area
is |8r|
2
2
|r| = 8. The r cancels out; thus, the area of this convex quadrilateral is 8 independent
of r.
Solution 2 by: Joshua Horowitz (11/CT)

The given expression can be written and factored as the difference of two squares:
r4 z 4 + (10r6 2r2 )z 2 16r5 z + (9r8 + 10r4 + 1) = 0

[r2 z 2 + (5r4 + 1)]2 (2rz + 4r4 )2 = 0
[r2 z 2 + 2rz + (9r4 + 1)][r2 z 2 2rz + (r4 + 1)] = 0.
So the roots of the original equation are the roots of P (z) = r2 z 2 + 2rz + (9r4 + 1) combined
with the roots of Q(z) = r2 z 2 2rz + (r4 + 1). Each of these is a quadratic with real
coefficients. The quadratic P has discriminant 36r6 and the quadratic Q has discriminant
4r6 . Both of these are negative (since r is a nonzero real) so the roots of P and the roots
of Q form conjugate pairs.
Using the quadratic formula we can compute a root p = 1r + 3ri of P and a root
q = 1r + ri of Q. These two roots and their conjugates will form an isosceles trapezoid with
the real axis as an axis of symmetry. The area of this trapezoid (the desired answer to this
problem) will be the height times the sum of half the bases:

2 2
|Re p Re q| (|Im p| + |Im q|) = (|3r| + |r|) = |4r| = 8,
r r
where Re z and Im z denote the real and imaginary parts of the complex number z, respec-
tively.
Solution 3 by: Linda Liu (11/GA)
We have that
r4 z 4 + (10r6 2r2 )z 2 16r5 z + (9r8 + 10r4 + 1) = 0

r4 z 4 + (6r6 2r2 )z 2 + 9r8 6r4 + 1 + 4r6 z 2 16r5 z + 16r4 = 0
r4 z 4 + 2(3r4 1)r2 z 2 + (3r4 1)2 + (2r3 z 4r2 )2 = 0
(r2 z 2 + 3r4 1)2 + (2r3 z 4r2 )2 = 0,
www.usamts.org
so
(r2 z 2 + 3r4 1)2 = (2r3 z 4r2 )2 .
Taking the square root of both sides gives the two equations
r2 z 2 + 3r4 1 = (2r3 z 4r2 )i r2 z 2 2r3 zi + 3r4 + 4r2 i 1 = 0,
and
r2 z 2 + 3r4 1 = (2r3 z 4r2 )i r2 z 2 + 2r3 zi + 3r4 4r2 i 1 = 0.
Applying the quadratic formula to the first quadratic equation produces the roots
1 1
+ 3ri and ri.
r r
Applying the quadratic formula to the second quadratic equation produces the roots
1 1
3ri and + ri.
r r
These four complex numbers then form a trapezoid with height 2/|r| and bases |2r| and
|6r|, so the area of the trapezoid is
|2r| + |6r| 2
= 8.
2 r
www.usamts.org
4/1/17. Homer gives mathematicians Patty and Selma each a different integer, not known
to the other or to you. Homer tells them, within each others hearing, that the number given
to Patty is the product ab of the positive integers a and b, and that the number given to
Selma is the sum a + b of the same numbers a and b, where b > a > 1. He doesnt, however,
tell Patty or Selma the numbers a and b. The following (honest) conversation then takes
place:
Patty: I cant tell what numbers a and b are.
Selma: I knew before that you couldnt tell.
Patty: In that case, I now know what a and b are.
Create PDF with GO2PDFSelma:
for free, if youNow I also
wish to remove know
this line, click herewhat a PDF
to buy Virtual and b are.
Printer
Supposing that Homer tells you (but neither Patty nor Selma) that neither a nor b is greater
than 20, find a and b, and prove your answer can result in the conversation above.
Credit This problem comes from the Carnegie Mellon Math Studies Problem Seminar.
Comments What makes this problem tricky is that its not just a problem: Its a problem
inside a problem. It requires you to place yourselves in the shoes of Patty and Selma, and
not only make the same deductions they make, but deduce which conditions could have
led to those deductions. Some careful casework leads to the answer. Note that in the
following solutions, Meir Lakhovsky shows that (a, b) = (4, 13) is a viable solution, and
Jeffrey Manning shows that it is the only solution. Solutions edited by Naoki Sato.
Solution 1 by: Meir Lakhovsky (10/WA)
Some trial and error leads us to a = 4, b = 13. Let us show that the conversation
mentioned could take place. Patty was given the number 52 and Selma was given the
number 17.
Patty said, I cant tell what numbers a and b are, which is true since the product 52
could be achieved through either the numbers (2,26) or (4,13).
Selma answers I knew before that you couldnt tell, which is also true, since for each
possible pair (a, b) that adds up to 17, namely (2,15), (3,14), (4,13), (5,12), (6,11), (7,10),
and (8,9), there are at least two possible solutions that give rise to the product ab.
Patty then says, In that case, I now know what a and b are, which is true because
if (a, b) was equal to (2,26), then Selma would have had the number 28, which means she
could not have made her former statement because (a, b) could have been (5,23) in which
case Patty would have been able to figure out what a and b are before any statements were
made. Since Patty knows the product is 52 and (a, b) is not (2,26), she knows that (a, b) is
(4,13), the only other option.
www.usamts.org
Selma now says, Now I also know what a and b are, which is true since the only pair
from (2,15), (3,14), (4,13), (5,12), (6,11), (7,10), and (8,9) in which Patty could have made
her later statement is (4,13) (reasoning shown above). Let us show why (2,15), (3,14), (5,12),
(6,11), (7,10), and (8,9) dont work. If (2,15) were the numbers, then Patty would have had
the number 30 and would not have been able to make her later statement since both (2,15)
and (5,6) yield sums for which Selma would have been able to make her former statement.
Likewise, for (3,14), Patty would not have been able to make her statement since both (3,14)
and (2,21) yield sums for which Selma would have been able to make her former statement.
By the same logic, the pair for (5,12) is (3,20); the pair for (6,11) is (2,33); the pair for (7,10)
is (2,35);
you wishthe pair
to remove forclick(8,9)
this line, is Virtual
here to buy (3,24).PDF Printer
Thus, (a, b) = (4, 13) could have resulted in the described conversation.
Notice that Patty could tell what a and b were if and only if there is exactly one way to
factor ab into the product of two distinct integers, both greater than one (for the rest of the
solution we will use the word factorization to mean factorization into two distinct factors
both greater than 1), in that case a and b are the two factors. This would only happen when
ab = pq where p and q are primes, in which case a and b would be p and q or when ab = p3
where p is prime, in which case a = p2 and b = p.
For Selma to already know that Patty couldnt tell what a and b were, it must be
impossible to write a + b as the sum of two distinct primes or as the sum of a prime and
its square. Since a + b 40, the possible values for a + b are 11, 17, 23, 27, 29, 35, and 37
(notice that we do need to consider primes and squares of primes greater than 20 because
Selma doesnt know that a, b 20).
For the third line of the conversation to be true, there must be exactly one factorization
of ab such that the sum of the factors cannot be written as the sum of two distinct primes or
as the sum of a prime and its square. If a, b > 1, then (a 1)(b 1) = ab (a + b) + 1 > 0,
so a + b < ab + 1, so Patty knows that a + b < 401. This means Patty knows that a + b is
either odd or twice a prime (Goldbachs conjecture states that any even integer 4 is the
sum of two primes, not necessarily distinct. Although this has not been proven, it has been
verified for all values we are concerned with). So we only need to consider factorizations of
ab such that the sum of the factors is not divisible by 4 (we cant have a + b = 2(2) = 4
because that would mean ab is 3 or 4). Notice that if ab = 4p where p is an odd prime, then
the only possible values of a and b are 4 and p.
For the fourth line to be true there must be only one way to write a + b as the sum of
two numbers (> 1) whose product satisfies these conditions.
We have:
www.usamts.org
If a + b = 11 = 4 + 7 = 2 + 9, then we could have ab = 28 or 18.

We will show that each of these possible values of ab satisfy the conditions for the third
line to be true. The numbers 28, 76, 92 and 124 are all in the form 4p so they all work.
If ab = 18 = 2 3 3 then the only factorization, other than 2 9, is 3 6 which gives
a + b = 9 = 2 + 7.
If ab = 50 = 2 5 5 then the only factorization, other than 2 25, is 5 10 which gives
a + b = 15 = 2 + 13.
If ab = 54 = 2 3 3 3 then the only factorizations, other than 2 27, are 6 9 which gives
a + b = 15 = 2 + 13, and 3 18 which gives a + b = 21 = 2 + 19.
If ab = 112 = 2 2 2 2 7 then the only factorizations other than 7 16 such that 4 - a + b,
are 8 14 which gives a + b = 22 = 5 + 17, and 2 56 which gives a + b = 58 = 5 + 53.
If ab = 208 = 222213 then the only factorizations other than 1316 such that 4 - a+b,
If ab = 96 = 2 2 2 2 2 3 then the only factorizations other than 3 32 such that 4 - a + b,
If ab = 160 = 2 2 2 2 2 5 then the only factorizations other than 5 32 such that
4 - a+b, are 1016 which gives a+b = 26 = 3+23, and 280 which gives a+b = 82 = 29+53.
If ab = 232 = 2 2 2 29 then the only factorizations, other than 8 29, are 4 58 which
gives a + b = 62 = 3 + 59, and 2 116 which gives a + b = 118 = 5 + 113.
Therefore a + b = 11, 23, 27, 29, 35, and 37 dont satisfy the fourth line, so a + b = 17.
Thus the possible ordered pairs (a, b) are:
(2, 15) ab = 30, but this can be factored as 5 6 and 5 + 6 = 11.

(4, 13) ab = 52, which can be written in the form 4p so it satisfies the third line.
www.usamts.org
Thus the only value of ab that satisfies the third line is 52, so a + b = 17 satisfies the
fourth line. Thus the only possible value of (a, b) is (4,13).
www.usamts.org
5/1/17. Given triangle ABC, let M be the midpoint of side AB and N be the midpoint
of side AC. A circle is inscribed inside quadrilateral N M BC, tangent to all four sides, and
that circle touches M N at point X. The circle inscribed in triangle AM N touches M N at
point Y , with Y between X and N . If XY = 1 and BC = 12, find, with proof, the lengths
of the sides AB and AC.
Credit This problem was proposed by Richard Rusczyk.
Comments This geometry problem is best solved using a side chase (as opposed to an
angle chase), in which the relations between lengths of line segments are written down until
there are a sufficient number of them that they can be solved algebraically. Any approach of
this kind will almost inevitably lead to the answer. But this is not the only possible approach,
as the following solutions will show. We recommend that when submitting a solution to a
geometry problem to also provide a diagram, so that the grader does not have to draw one.
Solutions edited by Naoki Sato.
Let AB and AC be tangent to the incircle of 4AM N at D and E. Similarly, let AB

and AC be tangent to the circle inscribed in M N CB at P and Q. Let AC = 2b, AB = 2c.
Note that M N = 12 BC = 6, and let s = 12 (b + c + 6) be the semiperimeter of 4AM N .
Because M N CB has an inscribed circle, M N +BC = M B +N C, or b+c = 18. By equal

tangents around the incircle of 4AM N , it is easy to see that M Y = s b and N Y = s c.
www.usamts.org
For example,
2 MY = MY + MD
= (AM AD) + (M N Y N )
= AM AE + M N EN
= AM + M N AN
= AM + M N + AN 2AN
= 2s 2b
Create PDF with GO2PDF for free, if you wish to remove this line, clickhereM = sPDFPrinter
Y Virtual
to buy b.
Moreover, the circle inscribed in M N CB is the excircle of 4AM N , opposite of A. This

implies that N X = M Y . Indeed, we have
2 AP = AP + AQ
= (AM + M X) + (AN + N X)
= 2s
AP = AQ = s.
Thus, M Y = s b = AQ AN = N Q = N X. From this, we have 1 = XY = N X N Y =

(sb)(sc) = cb. Combined with b+c = 18, we have AB = 2c = 19 and AC = 2b = 17.
Solution 2 by: Alan Deng (12/NY)
Let M X = a and N Y = b. By properties of tangents to a circle and the fact that triangle
AM N is similar to triangle ABC with a ratio of similitude of 1:2 (since M is the midpoint
of AB and N is the midpoint of AC), we can label the following diagram on the left, except
for the length of T H, where H is the foot of the altitude from A to BC.
www.usamts.org
A A
2a + 1
2b + 1
M0 Y0
N0
X0
a+1
O0 O0
b
M a Y b M Y
N N
a X1
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer X1
b+1
2a + 2 O O
2b
B 2a + 2 T 2 H 2b 2 C B T 2 H C
The ratio of the radius of circle O to that of O0 is 2:1, since they are inscribed in similar
triangles of that ratio. The distance XY is 1. Draw a line M 0 N 0 parallel to M N that is also
tangent to circle O0 , as shown in the diagram on the right. We also have points X 0 and Y 0
on M 0 N 0 that correspond to points X and Y on M N . In fact, X 0 and Y 0 are the images of
X and Y , respectively, of the similitude through A by ratio 1:2, so X 0 Y 0 = XY /2 = 1/2.
Furthermore, X 0 Y is parallel to AH since M 0 N 0 and M N are parallel tangents to circle O0 .
If we continue this process of drawing tangents and circles towards point A, we obtain line
segments of length 1, 1/2, 1/4, 1/8, and so on, and the union of their orthogonal projections
onto BC is T H. Hence,
1 1 1
T H = 1 + + + + = 2.
2 4 8
Now, we have two right triangles ABH and ACH that share the same height. Then
AB 2 BH 2 = AH 2 = AC 2 CH 2
(6a + 4)2 (2a + 4)2 = (6b + 2)2 (2b 2)2
32a2 + 32a = 32b2 + 32b
4a2 + 4a = 4b2 + 4b
4a2 + 4a + 1 = 4b2 + 4b + 1
(2a + 1)2 = (2b + 1)2 ,
www.usamts.org
so a = b. Since BC = 2a + 2b + 2 = 12, a = b = 5/2. Hence, AB = 6a + 4 = 19 and

AC = 6b + 2 = 17.
Solution 3: Based on the solution by Logan Daum (11/AK)
By the side chasing argument in Solution 1, we get that M X = N Y . (For example,

this follows immediately from M Y = N X.) Since XY = 1 and M N = 6, we have that
M X = N Y = 5/2, and we can argue that T H = 2 as in Solution 2. (Actually, Logan argues
this in a more direct way than Alan. Let Z be the intersection of AH and M N . What do
you for
Create PDF with GO2PDF notice
free, if youabout triangles
wish to remove this line, clickAZY
here to buyand T XY
Virtual PDF Printer?)
Now, T is the image of Y under the homothety through A by a factor of 2, so CT =

2N Y = 5, which means BT = BC CT = 12 5 = 7. Also, CH = CT T H = 3 and
BH = BC CH = 9.
A
M Y
N
X1
B 7 T 2 H 3 C
Let r be the inradius of triangle ABC, so the height of trapezoid BM N C is 2r, so then
the height of triangle ABC is AH = 4r. Then
B OT r
tan = = ,
2 BT 7
and
AH 4r
tan B = = .
BH 9
www.usamts.org
By the double angle formula,
2 tan(B/2)
tan B = .
1 tan2 (B/2)
Substituting, we obtain
4r 2r/7 14r
= 2
=
9 1 (r/7) 49 r2
196 4r2 = 126
70 35
r2 = =
r 4 2
35
r=
2
r
4r 4 35 2 70
tan B = = = .
9 9 2 9
Since tan B is positive, B is acute. Also,
cos2 B
cos2 B =
cos2 B + sin2 B
1
= sin2 B
1+ cos2 B
1
=
1 + tan2 B
1
=
1 + 470
81
81
= ,
361
so r
81 9
cos B = = .
361 19
Hence,
BH 9
AB = = = 19.
cos B 9/19
Similarly, r
AH 4r 4 35 2 70
tan C = = = = ,
CH 3 3 2 3
www.usamts.org
so C is also acute, and

1
cos2 C =
1 + tan2 C
1
=
1 + 470
9
9
= ,
289
so r
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer 9 3
cos C = = .
289 17
Hence,
AH 3
AC = = = 17.
cos C 3/17
www.usamts.org
1/2/17. At the right is shown a 4 4 grid. We wish to fill in the grid such 1 2 3 4
that each row, each column, and each 2 2 square outlined by the thick lines
contains the digits 1 through 4. The first row has already been filled in. Find,
with proof, the number of ways we can complete the rest of the grid.
Credit This problem was proposed by George Berzsenyi, founder of the USAMTS.
Comments This is a relatively simple counting problem (inspired by the latest Sudoku
puzzles and simplified to a 4 4 grid), where you need a little care to make sure that you
have covered all the cases and that you havent counted any grids twice. Lynnelle Ye shows
a particularly nice approach. Solutions edited by Naoki Sato.
Solution 1 by: Lynnelle Ye (8/CA)
The numbers in the second row must be one of the following: 3 4 1 2, 4 3 2 1, 3 4 2 1, or

4 3 1 2.
In the first two cases, the first and third columns are missing the same numbers, and
the second and fourth columns are missing the same numbers, so the number chosen for the
third row, first column determines the number chosen for the third row, third column, and
the number chosen for the third row, second column determines the number chosen for the
third row, fourth column. Once the third row is chosen, the fourth row is determined. For
example, with the first case for the second row,
1 2 3 4
3 4 1 2
2 or 4 1 or 3 4 or 2 3 or 1
So these two cases contribute a total of 23 = 8 possibilities.
In the second two cases, the number chosen for the third row, first column determines
the number chosen for the third row, either third or fourth column. This then determines
the number in the third row, second column, which then determines the third row, whatever
column is left. Again, the fourth row is determined. For example, with the third case for
the second row and the third row, first column equal to 2,
1 2 3 4
3 4 2 1
2 1 4 3
The third row, fourth column must be 3, which means that the third row, second column
must be 1, which means that the third row, third column must be 4. So these two cases
contribute a total of 22 = 4 possibilities.
Therefore, there are a total of 8 + 4 = 12 ways to complete the rest of the grid.
www.usamts.org
2/2/17. Write the number

1

2 32
as the sum of terms of the form 2q , where q is rational.
(For
example, 21 + 21/3 + 28/5 is a
3
sum of this form.) Prove that your sum equals 1/( 2 2).
Credit This problem was proposed by Sidney Kravitz.
Comments This problem is effectively an exercise in rationalizing the denominator, but

the twist is the presence of both the square root and the cube root. A knowledge of basic
algebraic identities can take care of both. James Sundstrom arrives at the answer using
two steps, Justin Hsu shows how to the same calculation in one step, and Jeffrey Manning
provides a clever solution using geometric series. Solutions edited by Naoki Sato.
The identities (a b)(a + b) = a2 b2 and (a b)(a2 + ab + b2 ) = a3 b3 suggest

the
following
approach for rationalizing the denominator. For example, setting a = 2 and
3
b = 2 gives
3 3 3
( 2 2)( 2 + 2) = 2 4,
so
1 2+ 32 2+ 32
= = .
2 32 ( 2 3 2)( 2 + 3 2) 2 34

Then setting a = 2 and b = 3 4 gives
3

3

3
(2 4)(4 + 2 4 + 2 2) = 8 4 = 4,
so

2+ 32 ( 2 + 3 2)(4 + 2 3 4 + 2 3 2)
=
2 34 (2 3 4)(4 + 2 3 4 + 2 3 2)

4 2+432+234 2+4+232 2+234
=
4

3
3
4 2 3
2 2 3
4
= 2+ 2+ +1+ +
1 1 2
2 1 2
1
2 1 2
1
+ 2 + 2 3 2 2 21 + 2 3 21
0

= 22 + 23 + 23 22 2
1 1 1 1 1
= 2 2 + 2 3 + 2 6 + 20 + 2 6 + 2 3 .
www.usamts.org
Solution 2 by: Justin Hsu (11/CA)
We have the identity
(a b)(a5 + a4 b + a3 b2 + a2 b3 + ab4 + b5 ) = a6 b6 .

Now, we let a = 2 and b = 3 2, and multiply the numerator and denominator by a5 + a4 b +
a3 b2 + a2 b3 + ab4 + b5 . This gives us

1 4 2+432+234 2+4+232 2+234
this line, click
Create PDF with GO2PDF for free, if you wish to remove here to Printer
buy Virtual PDF
2 32 4 2+432+234 2+4+232 2+234

4 2+432+234 2+4+232 2+234
=
4
1 1 1 1 1
= 2 2 + 2 3 + 2 6 + 2 + 2 6 + 2 3 ,
0
giving us the required sum of rational powers of 2.
Notice that if x 6= 0, then the sum x3 + x2 + x + 1 + x1 + x2 is a geometric series with

common ratio x1 and initial term x3 . If x 6= 1, then we can use the formula for a geometric
series to get
x3 (1 x6 ) x3 x3 (x3 x3 )x3 x6 1
x3 + x2 + x + 1 + x1 + x2 = = = = .
1 x1 1 x1 (1 x1 )x3 x3 x2
Applying this formula with x = 21/6 gives
1 21 1
=
2 32 21/2 21/3
(21/6 )6 1
= 1/6 3
(2 ) (21/6 )2
= (21/6 )3 + (21/6 )2 + 21/6 + 1 + (21/6 )1 + (21/6 )2
= 21/2 + 21/3 + 21/6 + 20 + 21/6 + 21/3 .
www.usamts.org
3/2/17. An equilateral triangle is tiled with n2 smaller

congruent equilateral triangles such that there are n smaller
triangles along each of the sides of the original triangle. The
case n = 11 is shown at right. For each of the small equilat-
eral triangles, we randomly choose a vertex V of the triangle
and draw an arc with that vertex as center connecting the
midpoints of the two sides of the small triangle with V as
an endpoint. Find, with proof, the expected value of the
number of full circles formed, in terms of n.
Comments Trying to count in how many cases the circles appear gets very complicated,
as these cases are not independent. (In other words, whether a circle appears at one vertex
affects whether a circle can appear at adjacent vertices.) However, a fundamental property
of expected value is that the expected value of a sum is simply the sum of the expected
values, a property mentioned in our Expected Value article. Once you see this, the problem
actually becomes quite easy. Solutions edited by Naoki Sato.
Solution 1 by: Derrick Sund (12/WA)
Consider a vertex in such a triangle that has six small triangles around it. Each of these
triangles has a 31 probability of its arc being the arc of a circle centered on that vertex.
1
Therefore, the probability that that vertex has a full circle around it is 316 = 729 . To get
the expected value of the number of full circles, we simply need to multiply this probability
by the number of vertices that can have full circles around them. If the original triangle is
divided into n2 smaller triangles, then the number of such vertices will be
(n 1)(n 2)
(n 2) + (n 3) + (n 4) + + 2 + 1 = ,
2
so the desired expected value is
(n 1)(n 2) 1 (n 1)(n 2)
= .
2 729 1458
www.usamts.org
4/2/17. A teacher plays the game Duck-Goose-Goose with his class. The game is played
as follows: All the students stand in a circle and the teacher walks around the circle. As he
passes each student, he taps the student on the head and declares her a duck or a goose.
Any student named a goose leaves the circle immediately. Starting with the first student,
the teacher tags students in the pattern: duck, goose, goose, duck, goose, goose, etc., and
continues around the circle (re-tagging some former ducks as geese) until only one student
remains. This remaining student is the winner.
For instance, if there are 8 students, the game proceeds as follows: student 1 (duck),
student 2 (goose), student 3 (goose), student 4 (duck), student 5 (goose), student 6 (goose),
student
Create PDF with GO2PDF for free,7 (duck),
if you student
wish to remove this line, click8here
(goose),
to buy Virtual student
PDF Printer 1 (goose), student 4 (duck), student 7 (goose)
and student 4 is the winner. Find, with proof, all values of n with n > 2 such that if the
circle starts with n students, then the nth student is the winner.
Credit This problem was proposed by Mathew Crawford.
Comments Adam Hesterberg nicely solves the problem by modifying the game slightly to
one that has a recursive nature. Solutions edited by Naoki Sato.
We claim that the nth student is the winner if and only if n of the form 3k 2 or 2 3k 2,
where k is a positive integer.
Suppose that the game was actually Goose-Goose-Duck, and there were two more
people at the start of the circle. This game is equivalent to Duck-Goose-Goose, since the
two new people will be tagged immediately, and the pattern is equivalent to Duck-Goose-
Goose from there. In the first round, people numbered with multiples of 3 survive, so either
the winner is a multiple of 3, or there are only two people left and he is one of them.
In the latter case, the game ends with that round. In the former case, if the winner is a
multiple of 3, then the first two people in the next round will be tagged, and the third will
live. In general, if the last person survived the (k 1)st round, then the survivors of the
k th round will be the multiples of 3k . Therefore, for the last person to be the final survivor,
he must have the greatest power of 3 as a factor. In case there are two numbers with the
same greatest power of 3 as a factor, the lower number gets tagged earlier, so the last person
would still win. Therefore, the possible numbers for the last person are 3k and 2 3k greater
multiples of 3k would produce someone with a greater power of 3. Subtract 2 to revert to
the original game, getting n = 3k 2 or n = 2 3k 2.
www.usamts.org
5/2/17. Given acute triangle 4ABC in plane P, a point Q in space is defined such
that AQB = BQC = CQA = 90 . Point X is the point in plane P such that QX is
perpendicular to plane P. Given ABC = 40 and ACB = 75 , find AXC.
Credit This problem was proposed by Sandor Lehoczky.
Comments The key insight in this problem is to realize that X is the orthocenter of triangle
ABC. This, in turn, can be proven by showing that Q must lie on certain spheres. Once
you identify that the point X is the orthocenter, the rest of the problem becomes an easy
anglefor chase.
Create PDF with GO2PDF Solutions
free, if you wish edited
to remove this line, click hereby Naoki
to buy Virtual PDFSato.
Printer
Solution 1 by: Philip Shirey (12/PA)
It is well known that in two dimensions, given points A and B, the locus of points P
such that AP B = 90 is the circle with diameter AB. (See diagram below.) The three-
dimensional locus, such as in this problem, is a sphere. Thus, point Q is the intersection of
three spheres.
We can better visualize the intersection of the three spheres by taking their intersection
with the plane P.
The triangle shown above is acute, so each circle interesects with the adjancent segments.
As mentioned in the above theorem, any point on a circle will form a right triangle with its
diameter. Because of this, these intersections form the altitudes of the triangle.
www.usamts.org
The intersection of two spheres is a circle. In the 2D representation of our 3D figure,

these circles would be perpendicular to the plane of the page, which means they would be
shown as segments in the 2D image. The endpoints of these segments are the interesections
of the projected spheres (i.e., the circles). So, the altitudes represent the intersections of the
spheres. As point X lies on plane P, X is the intersection of the altitudes, otherwise known
as the orthocenter.
Lastly, AXC = 140 because the congruent angle on the opposite side of X forms a
quadrilateral with two 90 angles with ABC, which is 40 . So AXC = 180 40 = 140 .
Solution 2 by: Tan Zou (10/IN)
Let PACQ be the plane of 4ACQ. Draw a line through B and X such that it intersects
side AC at point E. Draw EQ and let PBEQ be the plane of 4BEQ. Then BQ AQ and
www.usamts.org
BQ CQ, so PBEQ PACQ . Hence, BE AC, because BE and AC are in PBEQ and
PACQ , respectively.
Similarly, if we draw AD and CF through point X, we can prove that they are perpen-
dicular to BC and AB, respectively. Therefore, X is the orthocenter of 4ABC.
We know that if H is the orthocenter of 4ABC, then A + BHC = 180 . To see this,
let A = . Then ABH = 90 , so BHF = , and BHC = 180 .
Since X is the orthocenter and B = 40 , AXC = 180 40 = 140 .

Let X be the origin of coordinate space, and let us use the notation ~a for the vector from
point X to point A, etc.
Since QX is perpendicular to the plane containing ABC,
~q ~a = ~q ~b = ~q ~c = 0.
Since AQB = BQC = CQA = 90 ,
(~q ~a) (~q ~b) = (~q ~b) (~q ~c) = (~q ~c) (~q ~a) = 0.
Expanding (by the distributive property of dot products), and substituting 0 for ~q ~a, ~q ~b,
and ~q ~c gives ~a ~b = ~b ~c = ~c ~a = |~q|2 . But if ~a ~b = ~a ~c, then ~a (~b ~c) = 0, so AX is
perpendicular to BC. Similarly, BX and CX are perpendicular to AC and AB, respectively,
so X is the orthocenter of ABC.
Let CX intersect AB at C1 and AX intersect BC at A1 ; then 4AC1 X = 4AA1 B because
AC1 X = AA1 B = 90 . Therefore, AXC1 = ABA1 = 40 , and AXC = 140 .

www.usamts.org
1/3/17. For a given positive integer n, we wish to construct a 662 626

circle of six numbers as shown at right so that the circle has the
following properties: 866 264
(a) The six numbers are different three-digit numbers, none of
whose digits is a 0. 486 648
(b) Going around the circle clockwise, the first two digits of each
number are the last two digits, in the same order, of the previous number.
(c) All six numbers are divisible by n.

The example above shows a successful circle for n = 2. For each of n = 3, 4, 5, 6, 7, 8,

9, either construct a circle that satisfies these properties, or prove that it is impossible to do
so.
Credit This problem was based on a proposal by George Berzsenyi, founder of the
USAMTS.
Comments First, you must determine for each given n whether such a circle of numbers
exists or not. When it exists, such a circle is not hard to find. When it does not exist, well-
known divisibility rules of numbers can be used to give a rigorous proof. Solutions edited by
Naoki Sato.
Circles can be constructed for n = 3, 4, 6, and 7. An example of each is shown below.
We will show that such a construction is impossible for n = 5, 8, and 9.

n = 5: Each number must end in 0 or 5 for divisibility by 5, but 0 cannot be used, so all
numbers must end in 5. Then, going around the circle, the ten digits must also all be 5, as
the hundred digits. Thus, all numbers must be 555, which violates rule (a).
n = 8: All units digits must be even for divisibility by 8. Then the ten digits and also
the hundred digits must all be even. Since 8 divides 200, 8 must also divide the last two
www.usamts.org
digits. The only possibilities are x24, x48, x64, and x88, where x is an even digit. Going
clockwise around the circle, x24 and x64 both force x48 to be the next number, which then
forces x88 as the next number, and then 888. Thus, 888 will necessarily be repeated before
the circle is complete, violating rule (a).
n = 9: Assume such a construction is possible, and pick a number abc (or 100a + 10b + c)
in the cycle. A number is divisible by 9 if and only if the sum of its digits equals a multiple
of 9, so 9|(a + b + c). If we let the next number clockwise be bcd, then 9|(b + c + d), so
9|[(a + b + c) (b + c + d)] or 9|(a d). Since 1 a, d 9, we must have a = d, so bcd = bca.
Continuing clockwise, we see by the same argument that the next numbers are cab and abc.
The number abc must then appear twice, which again violates rule (a).
www.usamts.org
2/3/17. Anna writes a sequence of integers starting with the number 12. Each subsequent
integer she writes is chosen randomly with equal chance from among the positive divisors of
the previous integer (including the possibility of the integer itself). She keeps writing integers
until she writes the integer 1 for the first time, and then she stops. One such sequence is
12, 6, 6, 3, 3, 3, 1.
What is the expected value of the number of terms in Annas sequence?
Credit This problem was proposed by Mathew Crawford.

Comments This problem is similar to problem 2/1/17, in which we also calculated an

expected value using a recursive formula. Solutions edited by Naoki Sato.
Solution 1 by: Garrett Marcotte (12/CA)
Let (an ) be a sequence such as described in the problem, and let E(a1 ) be the expected
number of terms of (an ). To calculate E(a1 ), suppose that d1 , d2 , . . . , dk are the positive
divisors of a1 , with dk = a1 . Then there is a k1 probability that any given divisor di is chosen
as a2 . Thus, based on the method of generating the sequence, we can calculate E(a1 ) as
follows:
1 1 1
[E(d1 ) + 1] + [E(d2 ) + 1] + + [E(dk ) + 1]
E(a1 ) =
k k k
1 1
= [k + E(d1 ) + E(d2 ) + + E(dk1 )] + E(dk )
k k
k1 1
E(a1 ) = [k + E(d1 ) + E(d2 ) + + E(dk1 )]
k k
1
E(a1 ) = [k + E(d1 ) + E(d2 ) + + E(dk1 )].
k1
Now we apply this result to find E(12). By the definition of the sequence, E(1) = 1. The
numbers 2 and 3 have the same number of divisors, namely 2, so
1
E(2) = E(3) = [2 + E(1)] = 3.
1
The number 4 has three divisors, namely 1, 2, and 4, so
1 7
E(4) = [3 + E(1) + E(2)] = .
2 2
The number 6 has four divisors, namely 1, 2, 3, and 6, so
1 11
E(6) = [4 + E(1) + E(2) + E(3)] = .
3 3
www.usamts.org
Finall, the number 12 has six divisors, namely 1, 2, 3, 4, 6, and 12, so

1
E(12) = [6 + E(1) + E(2) + E(3) + E(4) + E(6)]
5
1 7 11
= 6+1+3+3+ +
5 2 3
121
= .
30
Solution 2 by: Gaku Liu (10/FL)

We will count the expected value of the number of each of the integers 12, 6, 4, 3, 2, and
1 in the sequence separately. (Note: We will use the term decomposition to denote a term
changing from one integer to a different one.)
The integer 12 always appears as the first term of the sequence. The next integer has an
equal chance of being any one of 12s six divisors, so a second 12 will appear an expected 16
times. Then, a third 12 will appear an expected ( 61 )2 times, etc., so the expected value of
the number of 12s is 2
1 1 6
1+ + + = .
6 6 5
The integer 6 can only decompose from the integer 12. 12 has an equal chance of decom-
posing into any of its five proper divisors, so 6 has a 51 chance of appearing in the sequence.
6 has four divisors, so the expected value of the number of 6s is
" 2 #
1 1 1 1 4 4
1+ + + = = .
5 4 4 5 3 15
The integer 4 can only decompose from the integer 12, so it has a 15 chance of appearing
in the sequence. 4 has three divisors, so the expected value of the number of 4s is
" 2 #
1 1 1 1 3 3
1+ + + = = .
5 3 3 5 2 10
The integer 3 can decompose from either 12 or 6. It has a 15 chance of decomposing from
12. 6 has a 15 chance of appearing in the sequence, and has three proper divisors it can
decompose into, so there is a 15 31 = 15
1
chance the integer 3 will decompose from 6. Hence,
1 1 4
there is a 5 + 15 = 15 chance 3 will appear in the sequence. The integer 3 has two divisors,
so the expected value of the number of 3s is
" 2 #
4 1 1 4 8
1+ + + = 2= .
15 2 2 15 15
www.usamts.org
The integer 2 can decompose from either 12, 6, or 4. It has a 51 chance of decomposing
from 12, a 51 13 = 15
1
chance of decomposing from 6, and a 15 12 = 10
1
chance of decomposing
from 4. Hence, 2 has a 15 + 15 1 1
+ 10 11
= 30 chance of appearing in the sequence. The integer 2
has two divisors, so the expected value of the number of 2s is
" 2 #
11 1 1 11 11
1+ + + = 2= .
30 2 2 30 15
The integer 1 will always appear exactly 1 time. Hence, the expected value of the total
number of terms is
6 4 3 8 11 121
+ + + + +1= .
5 15 10 15 15 30
www.usamts.org
3/3/17. Points A, B, and C are on a circle such that 4ABC is A

an acute triangle. X, Y , and Z are on the circle such that AX is Y
perpendicular to BC at D, BY is perpendicular to AC at E, and CZ E
is perpendicular to AB at F . Find the value of Z F
AX BY CZ
+ + ,
AD BE CF B D C
and prove that this value is the same for all possible A, B, C on the X
circle such that 4ABC is acute.
Credit This problem was proposed by Naoki Sato.
Comments This geometry problem can be solved by recognizing that the given ratios can
be expressed as ratios of certain areas, and using the fundamental result that HD = DX,
where H is the orthocenter of triangle ABC. A solution using power of a point is also
possible. Solutions edited by Naoki Sato.
Solution 1 by: Justin Hsu (11/CA)
Let H be the orthocenter of 4ABC. First, 4BHD is similar to 4BCE, since they are
both right triangles and they share CBE, so BCE = BHD. Also, BXA = BCA =
BHD, since they both are inscribed angles that intercept the same arc BA. Now, 4BXH
is isosceles, which means that BD is the perpendicular bisector of segment HX. Therefore,
4BDH 4BDX, and HD = DX. Similarily, this can be extended to the other sides of
the triangle to show that HE = EY and HF = F Z.
Now,
AX BY CZ AD + DX BE + EY CF + F Z
+ + = + +
AD BE CF AD BE CF
DX EY FZ
=1+ +1+ +1+
AX BY CZ
HD HE HF
=3+ + + .
AD BE CF
But each fraction is a ratio between the altitudes of two triangles with the same base, so
we can rewrite this sum in terms of area, where [ABC] denotes the area of 4ABC:
HD HE HF [HBC] [HCA] [HAB]
3+ + + =3+ + +
AD BE CF [ABC] [ABC] [ABC]
[ABC]
=3+
[ABC]
= 3 + 1 = 4.
www.usamts.org
By the Power of a Point Theorem,

AD DX = BD CD.
Therefore,
DX BD CD
= = cot B cot C.
AD AD AD
Similarly,
EY
= cot C cot A,
BE
FZ
= cot A cot B.
CF
We can calculate
AX BY CZ AD DX BE EY CF FZ
+ + = + + + + +
AD BE CF AD AD BE BE CF CF
DX EY FZ
=3+ + +
AD BE CF
= 3 + cot B cot C + cot C cot A + cot A cot B
tan A + tan B + tan C
=3+ .
tan A tan B tan C
We claim that
tan A + tan B + tan C = tan A tan B tan C
for all acute triangles 4ABC (acuteness of 4ABC means that tan A, tan B, and tan C
exist). [Ed: As the following argument shows, this identity holds for all triangles ABC
where both sides are defined.]
We have that
tan C = tan( A B) = tan(A B) = tan(A + B),
and
tan A + tan B
tan(A + B) = ,
1 tan A tan B
so
tan A + tan B
. tan C =
1 tan A tan B
This can be re-arranged to become tan A + tan B + tan C = tan A tan B tan C.
Hence,
AX BY CZ
+ + = 3 + 1 = 4.
AD BE CF
www.usamts.org
4/3/17. Find, with proof, all triples of real numbers (a, b, c) such that all four roots of the
polynomial f (x) = x4 + ax3 + bx2 + cx + b are positive integers. (The four roots need not be
distinct.)
Credit This problem was based on a proposal by Brian Rice.
Comments To find all possible sets of roots (which is what the problem is effectively asking
for), you must use both the fact they are positive and integers. The first condition can be
used to find bounds on the roots and narrow down to a finite number of cases, and the
second
Create PDF with GO2PDF condition
for free, canthisbe
if you wish to remove used
line, click tobuyfind
here to them
Virtual PDF Printerspecifically. Solutions edited by Naoki Sato.
Let p, q, r, and s be the positive integer roots of f (x) = x3 + ax3 + bx2 + cx + b. We have
f (x) = x4 + ax3 + bx2 + cx + b

= (x p)(x q)(x r)(x s)
= x4 (p + q + r + s)x3 + (pq + pr + ps + qr + qs + rs)x2
(pqr + qrs + rsp + spq)x + pqrs.
Comparing coefficients, we note that it suffices to find all quadruples (p, q, r, s) of positive
integers such that
b = pqrs = pq + pr + ps + qr + qs + rs,
for then we obtain reals (which in fact are integers) a and c as
a = (p + q + r + s) and c = (pqr + qrs + rsp + spq),
and hence obtain all triples (a, b, c). First, let us rewrite our equation containing b by dividng
through by pqrs. We have
1 1 1 1 1 1
+ + + + + = 1.
pq pr ps qr qs rs
1
Without loss of generality, we have rs 61 so rs 6 for some pair of the roots (for
instance, the two smallest). Assume r s and p q. We now proceed with some casework.
Case 1: rs = 6. We either have (r, s) = (6, 1) or (3, 2). If (r, s) = (6, 1), we have
6pq = pq + 7(p + q) + 6 (5p 7)(5q 7) = 49 + 30 = 79,
which admittedly does not have any integer solutions (p, q) since 79 is prime and we must
have 5p 7 = 79 and 5q 7 = 1, but this is clearly impossible. If (r, s) = (3, 2) we have
6pq = pq + 5(p + q) + 6 5(pq p q) = 6,

www.usamts.org
which again does not have any integer solutions (p, q) since 6 is not divisible by 5.
Case 2: rs = 5. We have (r, s) = (5, 1) so
5pq = pq + 6(p + q) + 5 2(pq 3p 3q) = 5,
which does not have any integer solutions (p, q) since 5 is odd.
Case 3: rs = 4. We either have (r, s) = (4, 1) or (2, 2). If (r, s) = (4, 1), then
4pq = pq + 5(p + q) + 4 (3p 5)(3q 5) = 25 + 12 = 37,

so 3p 5 = 37 p = 14 and 3q 5 = 1 q = 2, since 37 is prime. Thus, we obtain
(p, q, r, s) = (14, 2, 4, 1), whence a = 21, b = 112, and c = 204. If (r, s) = (2, 2), we have
4pq = pq + 4(p + q) + 4 (3p 4)(3q 4) = 16 + 12 = 28,
which decomposes as a product of two positive integers as 28 1 = 14 2 = 7 4. It is easily

verified that only the cases 3p 4 = 14 and 3q 4 = 2 yields a valid solution (p, q) = (6, 2).
We obtain (p, q, r, s) = (6, 2, 2, 2), whence a = 12, b = 48, and c = 80.
3pq = pq + 4(p + q) + 3 2(pq 2p 2q) = 3,
which does not have any integer solutions (p, q) since 3 is odd.
2pq = pq + 3(p + q) + 2 (p 3)(q 3) = 9 + 2 = 11,
and we have p 3 = 11 p = 14 and q 3 = 1 q = 4. Thus, we get (p, q, r, s) =

(14, 4, 2, 1), which we obtained earlier in a different order with (r, s) = (4, 1).
pq = pq + 2(p + q) + 1 2(p + q) + 1 = 0,
which is clearly absurd, so there are no positive integer solutions (p, q).
Thus, we have determined all desired triples (a, b, c), namely (21, 112, 204) and (12, 48, 80).
Note: The number of cases can be reduced by the following argument. First, not all of p,
q, r, and s can be odd. If they were, then pqrs would be odd, but then pq+pr+ps+qr+qs+rs,
as the sum of six odd integers, would be even. Hence, at least one of them must be even.
WOLOG, let p be even. Then
pqrs pq pr ps = p(qrs q r s) = qr + qs + rs,

www.usamts.org
so qr+qs+rs is even. If q, r, and s were all odd, then qr+qs+rs would be odd, contradiction,
so at least one of them must be even. WOLOG, let q be even.
Then
rs = pqrs pq pr ps qr qs.
Each term in the RHS contains a factor of p or q, so the RHS is even. Then rs is even, so
one of r and s must be even. Hence, of the four positive integers p, q, r, and s, at least three
must be even. This argument, among other things, allows us to eliminate cases 2, 4, and 6
above.
www.usamts.org
5/3/17. Lisa and Bart are playing a game. A round table has n lights evenly spaced around
its circumference. Some of the lights are on and some of them off; the initial configuration
is random. Lisa wins if she can get all of the lights turned on; Bart wins if he can prevent
this from happening.
On each turn, Lisa chooses the positions at which to flip the lights, but before the lights
are flipped, Bart, knowing Lisas choices, can rotate the table to any position that he chooses
(or he can leave the table as is). Then the lights in the positions that Lisa chose are flipped:
those that are off are turned on and those that are on are turned off.
Here is an example turn for n = 5 (a white circle indicates a light that is on, and a black
circlefor indicates
Create PDF with GO2PDF free, if you wish to a light
remove that
this line, is tooff):
click here buy Virtual PDF Printer
1
5 2
Initial Position.
4 3
1
Lisa says 1, 3, 4.
5 2
Bart rotates the table one
position counterclockwise.
4 3
1
Lights in positions 1, 3, 4 are 5 2
flipped.
4 3
Lisa can take as many turns as she needs to win, or she can give up if it becomes clear
to her that Bart can prevent her from winning.
(a) Show that if n = 7 and initially at least one light is on and at least one light is off,
then Bart can always prevent Lisa from winning.
(b) Show that if n = 8, then Lisa can always win in at most 8 turns.
Credit This problem was based on a problem from the Puzzle TOAD page at
http://www.cs.cmu.edu/puzzle.
They credit the problem to Ron Holzman of the TechnionIsrael Institute of Technology.
www.usamts.org
Comments The case n = 7 can be solved by considering what moves must make Bart lose
to Lisa. The case n = 8 can be solved by an induction argument. Solutions edited by Naoki
Sato.
Solution 1 by: Hannah Alpert (11/CO)
For part (a), Lisa sees at least one light on and at least one light off, and she wants to
turn all the lights the same color (we will say that two lights are the same color if they are
both on or both off). For Lisa to win, the game must end with Bart being stuck: He must
either
Create PDF with GO2PDF turn
for free, if you all the
wish to off
remove this lights onto buy
line, click here orVirtual
all PDF
thePrinter
on lights off. Then as a last move, if all the lights
are off, Lisa will turn all of them on.
However, since 7 is odd, the number of off lights cannot be the same as the number of on
lights. Therefore, Bart cannot end up with such a decision; for example, if there are 3 lights
on and 4 lights off, and Lisa picks three positions, Bart cannot possibly be forced to turn all
the lights the same color, because it is not even possible to turn all the off lights on! Thus,
for any odd n and not all lights the same color, Bart can prevent Lisa from winning.
For part (b), we will prove by induction on k that if there are n = 2k lights, Lisa can
always win in at most 2k turns. Then obviously, we can apply this fact to the case where
n = 8.
In the base case, k = 0, there is one light. If it is on, Lisa wins; if it is off, Lisa requests
to turn it on. That takes only one turn.
Now assume that if there are 2k lights, then Lisa can win in at most 2k turns. We want
to show that if there are 2 2k lights, then Lisa can win in at most 2 2k turns. Observe
that since the number of lights, 2k+1 , is even, each light has a partner light across from it
on the table. Also notice that when Lisa chooses positions, the partners are preserved; for
example, if there are 8 lights and Lisa chooses positions 1 and 5, then now matter how Bart
rotates the table, she knows that some light and its partner will both change color.
Lisas aim will be first to get each light the same color as its partner and then to turn
all the pairs on. To get each light the same color as its partner, she restricts her choices
such that she never chooses both a light and its partner. Since there are 2 2k lights, there
are 2k sets of partners. Lisa imagines an equivalent table with 2k lights, where each light
corresponds to one set of partners on the big table. For each pair on the big table that has
two different colors, their light is off on the small table; for each pair on the big table with
the same color, their light is on at the small table. Then she makes each light the same color
as its partner in 2k moves, just as she would solve the small table in 2k moves.
[Ed: To clarify, here is an example. Suppose the big table has 8 lights, and positions 1,
2, and 6 are on. Then the small table has 4 lights, and positions 2, 3, and 4 are on. Looking
at the small table, Lisa chooses position 1. That means on the big table, she can choose
position 1 or 5. After she does so and Bart rotates the table, exactly one light (among two
www.usamts.org
partners) is flipped, which means that one light gets flipped on the small table. Choosing
exactly one light among partners ensures that this will always be the case. Lisa continues
making moves towards getting all the lights on the small table, and when that happens, for
any pair of partner lights on the big table, both are on, or both are off.]
Once every light is the same color as its partner, Lisa restricts her choices such that for
every light she chooses, she also chooses its partner. This time she imagines another table
with 2k lights. On this small table, each light still corresponds to one set of partners on the
big table, but this time the light at the small table is the same color as the pair (since the
partners now match). She finishes solving the big table in 2k moves, just as she would solve
Create PDF with GO2PDF for free, if you wish to removekthis line, click here to buy Virtual PDF Printer
the small table in 2 moves. Lisa has then solved the big table with 2 2k lights in 2 2k
moves. Our induction is complete.
(b) We identify the lights and the switching patterns with polynomials by calling the first
of the consecutive eight lights (points) 1, the second x, the third x2 , and so on, all the way
up to x7 . A light will be on if the coefficient of its respective power of x is 1; otherwise,
the light will be off. We will also think of the switching pattern being rotated, rather
than the lights. Similarly, a coefficient of 1 in a switching pattern polynomial will indicate
that that position is to be switched. To rotate a switching pattern n positions, we multiply
the corresponding polynomial by xn , and we take this new polynomial mod x8 + 1, because
x8 = 1, having gone all the way around the circle. When we add and multiply polynomials,
we take the coefficients mod 2, since switching (adding a power of x, with a coefficient of 1)
changes an off (0) to an on (1) and vice-versa.
[Ed: For example, suppose the table has 8 lights, and positions 1, 2, and 6 are on. Then
the polynomial corresponding to this set of lights is
x11 + x21 + x61 = 1 + x + x5 .
Lisa chooses positions 1, 2, and 6. Suppose Bart rotates the table by five positions. Then
the lights in positions 6, 7, and 3 get flipped, so the lights in positions 1, 2, 3, and 7 are on.
In terms of the polynomials, we multiply 1 + x + x7 by 1 + x5 to get
(1 + x + x5 )(1 + x5 ) = 1 + x + 2x5 + x6 + x10 .
We reduce all coefficients mod 2, because flipping a light twice does nothing. Also, going
around the circle, position 10 is the same as position 2, which is the same as reducing the
polynomial mod x8 + 1. Hence,
1 + x + 2x5 + x6 + x10 1 + x + x2 + x6 = x11 + x21 + x31 + x71 ,
which confirms that the lights at positions 1, 2, 3, and 7 are on.]

www.usamts.org
We now define A0 (x) to be the initial polynomial for the lights that are on and off, and
An (x) the polynomial for the lights at the end of the nth round. We claim that if Lisa gives
Bart a switching pattern equal to the on lights at the beginning of that round, then Lisa
will win in eight or fewer rounds.
When Lisa gives Bart A0 (x) for the switching pattern, we end up with
A1 (x) = A0 (x) + xa1 A0 (x) = (1 + xa1 )A0 (x)
as the new light pattern, where Bart has rotated the switching pattern by a1 positions.
Similarly,
A2 (x) = (1 + xa2 )A1 (x),

where Bart rotates the second switching pattern by a2 positions. Continuing, we get
A3 (x) = (1 + xa3 )A2 (x),

A4 (x) = (1 + xa4 )A3 (x),
...,
A8 (x) = (1 + xa8 )A7 (x).
Note that if x = 1, then for any i,
1 + xai = 1 + 1ai = 2 0,
so 1 + x is a factor of 1 + xai . Indeed,
(1 + x)(1 + x + x2 + + xai 1 )
= 1 + x + x2 + + xai 1 + x + x2 + + xai
= 1 + 2x + 2x2 + + 2xai 1 + xai
1 + x ai .
Thus, setting Pi (x) = 1 + x + x2 + + xai 1 , we can write
1 + xai (1 + x)Pi (x).
So,
A8 (x) = (1 + xa8 )A7 (x)

= (1 + xa8 )(1 + xa7 )A6 (x)
=
= (1 + xa8 )(1 + xa7 ) (1 + xa1 )A0 (x)
(1 + x)P8 (x)(1 + x)P7 (x) (1 + x)P1 (x)A0 (x)
= (1 + x)8 P8 (x)P7 (x) P1 (x)A0 (x).
www.usamts.org
Now,
(1 + x)8 = 1 + 8x + 28x2 + 56x3 + 70x4 + 56x5 + 28x6 + 8x7 + x8

1 + x8 (reducing the coefficients mod 2)
0, (reducing the polynomial mod 1 + x8 )
so A8 (x) 0. Therefore, all lights will always be off after eight rounds. Looking back at the
equation
A8 (x) = (1 + xa8 )A7 (x),
we see that the polynomial A7 (x) must then have the property that for any a8 , multiplying
A7 (x) by 1 + xa8 gives the zero polynomial. The only polynomials A7 (x) that have this
property are A7 (x) = 0 and A7 (x) = 1 + x + x2 + + x7 . Thus, after seven turns, the lights
are either all on or all off. If the lights are all on, Lisa has won at the end of the seventh
round (or sooner, if they were all on sooner). If they are all off, Lisa gives Bart the switching
pattern of change everything, and Lisa wins at the end of the eighth round (or sooner, if
they were all off sooner).
This generalizes to all n that are powers of two, so that Lisa can always win in at most
n rounds.
www.usamts.org
1/4/17. AB is a diameter of circle C1 . Point P is on C1 such that AP > BP . Circle C2

is centered at P with radius P B. The extension of AP past P meets C2 at Q. Circle C3 is
centered at A and is externally tangent to C2 . R is on C3 such that AR AQ. Circle C4
passes through A, Q, and R. Find, with proof, the ratio between the area of C4 and the area
of C1 , and show that this ratio is the same for all points P on C1 such that AP > BP .
Credit This problem was proposed by Ismor Fischer, University of Wisconsin.
Comments This problem can be easily solved by labeling lengths, and using right angles
to apply Pythagorass theorem. Solutions edited by Naoki Sato.
Solution 1 by: Matt Superdock (9/PA)
Let D be the center of C1 , and let E be the center of C4 . Let T be the point of tangency
of C2 and C3 . Let a be the radius of C3 , and let p be the radius of C2 . Then we have
AT = AR = a and P B = P Q = P T = p.
C4
Q
R E
P
p
p
a C2
a T p
A B
D
C3
C1
Since AR AQ, RAQ is a right angle. Since right angles inscribe the diameter of the
circle, RQ is the diameter of C4 . Since AP B inscribes AB, the diameter of C1 , AP B is
also a right angle.
The ratio we are seeking is
[C4 ] (ER)2 (QR)2 (AR)2 + (AQ)2 (AR)2 + (AT + T P + P Q)2

= = = =
[C1 ] (AD)2 (AB)2 (AP )2 + (BP )2 (AT + T P )2 + (BP )2
2 2 2 2
a + (a + p + p) 2a + 4ap + 4p
= 2 2
= 2 = 2.
(a + p) + p a + 2ap + 2p2
Since we made no assumptions, the ratio is 2 for all points P on C1 such that AP > BP .
www.usamts.org
2/4/17. Centered hexagonal numbers are the numbers of dots used to create hexagonal
arrays of dots. The first four centered hexagonal numbers are 1, 7, 19, and 37, as shown
below.
qqq qqqqqqqqq
q qqqqqqq qqqqqqqqqqqqq qqqqqqqqqqqqqqqqqqq
qqq qqqqqqqqq
1 7 19 37
Centered Hexagonal Numbers
Consider an arithmetic sequence 1, a, b and a geometric sequence 1, c, d, where a, b, c,
and d are all positive integers and a + b = c + d. Prove that each centered hexagonal number
is a possible value of a, and prove that each possible value of a is a centered hexagonal
number.
Credit This problem was proposed by Richard Rusczyk and Erin Schram.
Comments This problem requires some algebra to find the nth centered hexagonal number,
and then a little number theory to show the equivalence. Solutions edited by Naoki Sato.
Solution 1 by: Mike Nasti (11/IL)
We want to find an explicit formula for the nth centered hexagonal number. Partitioning
the dots as above, we see immediately that the nth centered hexagonal number is 1 more
than 6 times the (n 1)th triangular number. Thus, the nth centered hexagonal number is
1 + 6 (n1)(n)
2
= 3n2 3n + 1.
The arithmetic sequence 1, a, b has common difference a 1, so it can be written in one
variable as 1, a, 2a 1, so b = 2a 1. The geometric sequence 1, c, d has common ratio c, so
it too can be written in one variable as 1, c, c2 , so d = c2 . Then a + b = c + d a + 2a 1 =
c + c2 3a 1 = c(c + 1).
Let c = 3n2 for some integer n. Then c is an integer, so we know 3a1 = (3n2)(3n
2+1) = 9n2 9n+2 = 3(3n2 3n+1)1. Thus, whenever c = 3n2, a = 3n2 3n+1 which
is the nth centered hexagonal number. So each centered hexagonal number is a possible value
of a.
www.usamts.org
Now, if c(c + 1) = 3a 1 for an integer a, we can say c(c + 1) 1 2 (mod 3). Then we
know that c 1 (mod 3) because if c 0 (mod 3), then c(c + 1) 0 (mod 3), and if c 2
(mod 3), then c(c + 1) 0 (mod 3). Since c 1 2 (mod 3), every possible value of c can
be written in the form 3n 2 for some integer n. Therefore the set of possible values of a
is equal to the set of centered hexagonal numbers, so every possible value of a is a centered
hexagonal number.
www.usamts.org
3/4/17. We play a game. The pot starts at $0. On every turn, you flip a fair coin. If you
flip heads, I add $100 to the pot. If you flip tails, I take all of the money out of the pot, and
you are assessed a strike. You can stop the game before any flip and collect the contents
of the pot, but if you get 3 strikes, the game is over and you win nothing. Find, with proof,
the expected value of your winnings if you follow an optimal strategy.
Credit This problem was proposed by Dave Patrick.
Comments This problem is similar to problems 2/1/17 and 2/3/17, in that both involved
expected value, and both used the technique of reducing the problem to a simpler problem.
This particular problem is best solved by considering in sequence what happens when you
have one strike left, then two strikes, then the full three strikes. Solutions edited by Naoki
Sato.
Solution 1 by: Wei Hao (11/VA)
Consider first the case when I already have two strikes. In order to find the optimal time
to stop, let us assume that there are x dollars in the pot. I will toss again only when I will,
on average, make more than x dollars, or
1 1
(x + 100) + 0 > x,
2 2
which implies
x < 100. (1)
So if x < 100, it is advantageous to risk the last strike by tossing again. But the only possible
value for x < 100 is x = 0. Therefore, with two strikes, I will toss once and stop the game
no matter what the outcome is. The expected return for this toss is then $50.
Consider next when I have only one strike. I can use the same logic to decide when to
stop the game, except that if I get a tail, I have one more strike to give. As a result, I should
use the $50 expected value from the above discussion as the expected winnings if I get a tail.
So, assuming again that there are x dollars in the pot before a toss, it is worthwhile to risk
another strike only when
1 1
(x + 100) + 50 > x,
2 2
which gives
x < 150. (2)
To find the expected winnings in this case, I will follow the following strategy: Since imme-
diately after a strike, there is $0 dollar in the pot, so I will flip the coin again. If I get a
head, there will be $100 dollars in the pot, and I still have only one strike. But $100 is less
than $150, therefore, I can flip again. On the second flip, I will either get a head and stop
www.usamts.org
the game because there will be $200 in the pot, or I will get a tail and face the two strike
problem discussed before. If I get a tail on the first flip, I will also face the same two strike
problem. So the expected winnings with one strike is

1 1 1 1
200 + 50 + 50 = 87.5. (3)
2 2 2 2
Finally, I can use the same reasoning to decide when to stop the game when I have no
strikes. The only difference is that in the event of getting a tail, I must use the result of
equation
Create PDF with GO2PDF for free, if(3) asto the
you wish removeexpected winnings
this line, click here since
to buy Virtual PDF Printer I will have one strike then. Assuming that there
are x dollars in the pot to begin with, it is advantageous to try another flip when
1 1
(x + 100) + 87.5 > x,
2 2
or
x < 187.5. (4)
The expected winnings for following this strategy can be calculated in the same way as the
one strike case. I start by flipping the coin. If I get a head, there will be $100 in the pot.
But that is less than the $187.5 of equation (4). So I will flip again. If I get a head again,
there will be $200 in the pot and I will stop the game. But if I get a tail on the second flip,
the problem is reduced to the one strike problem with an expected winnings of $87.5. The
same thing happens if I get a tail on the first flip. Therefore, the expected winnings for this
game is
1 1 1 1
200 + 87.5 + 87.5 = 115.625. (5)
2 2 2 2
So the expected winnings for my strategy is $115.625.
www.usamts.org
4/4/17. Find, with proof, all irrational numbers x such that both x3 6x and x4 8x2
are rational.
Credit This problem was proposed by Erin Schram.

Comments Many students were able to find the solutions 6, but there are four additional
solutions. One must be careful about making assumptions about irrational numbers, such
as what they must look like. The best strategy here is to play the two given polynomials off
of each other. Solutions edited by Naoki Sato.

We claim that x = 1 3 (taking all four combinations of signs) and x = 6 are the
only six irrational x such that both x3 6x and x4 8x2 are rational. Now, we prove that
these are the only such values.
Assume we have some irrational x such that both x3 6x and x4 8x2 are rational. Let
a = x2 4, so a2 = x4 8x2 + 16 is rational. Let b = x3 6x = x(x2 6) = x(a 2) which
is also rational by hypothesis. We have
b2 = x2 (a 2)2 = (a + 4)(a 2)2 = a3 12a + 16 = a(a2 12) + 16.
In particular, because b2 is rational, a(a2 12) must be rational. 2 2

If a 12 6= 0, then a 12
is rational so a must be rational as well. Otherwise, a = 2 3.
If a2 12 6= 0 and a is rational, note that b = x(a 2) isrational. Because x is irrational
(and x 6= 0) we must have a = 2. Thus, x2 = 6 and x = 6.

If a = 2 3 then x2 = 42 3 = (1 3)2 so x = 1 3 (taking all four combinations
of signs). It is easily verified that all six solutions make x3 6x and x4 8x2 rational and
this concludes our proof.
www.usamts.org
5/4/17. Sphere S is inscribed in cone C. The height of C equals its radius, and both equal
12 + 12 2. Let the vertex of the cone be A and the center of the sphere be B. Plane P
is tangent to S and intersects segment AB. X is the point on the intersection of P and C
closest to A. Given that AX = 6, find the area of the region of P enclosed by the intersection
of C and P.
Comments As with most problems in three-dimensional geometry, a solution can be found

by considering relevant cross-sections of the figure. Many students made incorrect assump-
tions about the figure, such as assuming that the altitude of the cone from A intersects the
ellipse at its center. An accurately drawn figure helps prevent such errors. Solutions edited
by Naoki Sato.
Solution 1 by: Eric Chang (11/CA)
A
X Y
N P
C B
M
Let Y be the furthest point on the intersection of P and S from A. Let O be the center
of sphere S. Let BC be a diameter of the base of cone C. Let N and P be the points where
the sphere S is tangent to AB and AC, respectively. Finally, let M be the center of the base
of cone C.
First of all, we see that the intersection of plane P and cone C will be an ellipse, which
we call E. If we take a cross section of the cone and the sphere along plane ABC, we get
the above picture. Since the area of an ellipse is ab, with a and b as the semi-major and
semi-minor axis, respectively, we can solve for the area if we can find the length of the major
and minor axis. It is obvious that XY is the major axis of ellipse E since it is the longest
line in the ellipse. Also, by symmetry M is the midpoint of side BC, and A, O amd M are
collinear.
Because OP A, ON A, and N AP are all right angles, N OP must be a right angle
also. Since all segments tangent to a circle from the same point have the same length,
AN = AP and we see that AN OP is a square. Let r be the radius of S, then we see from
the picture AM = r 2 + r = 12 + 12 2, since it is the height of the cone. Solving we get
r = 12. Now we will consider square ON AP , reproduced below:
www.usamts.org
N X A
O P
Let R be the point of tangency of XY to the circle centered at O. Since all sides of
a square are congruent, they are all equal to 12, therefore, N X = 12 6 = 6. Since all
segments tangent to the circle from the same point are congruent, XR = 6 and RY = P Y .
Labeling AY = x, we see that P Y = 12 x and XY = 18 x. Also, since AXY is a right
triangle, XY 2 = x2 + 62 . Equating the two expressions for XY and solving for x, we get
x = 8, and therefore, AY = 8 and XY = 10.
We will construct the diagram below in the following paragraph. Draw a line through X
parallel to BC, and call D its point of intersection with AB, and then draw a line through Y
parallel to BC also, and the point E will be its intersection with AC. Now, since these lines
are parallel to the base, by similar triangles we have AX = AD = 6 and AE = AY = 8,
which implies XE = DY = 2. If we draw line F G parallel to XD and EY and go through
the midpoint of XE, it will also go through the midpoint of XY and DY because parallel
lines cut all transversals in the same ratio. Call H its intersection with XY , the the minor
axis will pass through this point perpendicular to XY .
X D
F G
E IH Y
Now draw the segment AI, where I is the midpoint of F G, then by a property of isosceles
triangles, AI is perpendicular to F G. As a result, we can find HI by using the Pythagorean
theorem on triangle AHI. Since XAY is a right triangle, by a well-known theorem, the
segment AH will be congruent to XH and HY , therefore AH = 5. Also, using the45-45-
90 triangle AF I, we find that AF = AX + XF = 6 + 1 = 7. Therefore, AI = 7 2 2 and
2
HI = 5 7 2 2 = 21 , so HI = 22 .
2 2
www.usamts.org
I G
H
Now we take a top view of the cross section of the cone at the
circle centered at I. Let JK
7 2
be the minor axis. We can see that the radius of the circle is 2 from previous calculations,
and since HI is perependicular to JK, which is JH, using the Pythagorean theorem again:
!2 !2
7 2 2
JH 2 = = 24,
2 2

so JH = 24 = 2 6. Now we can plug in 2 6 and 10/2 = 5 into out formula ab. As a
result, the area of ellipse E is 10 6.
Note: There is a beautiful solution using Dandelin spheres. Not only does this approach
solve the problem in a nice, synthetic way, it also explains why the cross-section is an ellipse.
See
http://www.artofproblemsolving.com/Community/AoPS_Y_MJ_Transcripts.php?mj_id=128
for a transcript of this solution.

www.usamts.org
1/1/18. When we perform a digit slide on a number, we move its units digit to the front of
the number. For example, the result of a digit slide on 6471 is 1647. What is the smallest
positive integer with 4 as its units digit such that the result of a digit slide on the number
equals 4 times the number?
Comments Let n be the number in the problem. Since the last digit of n is 4, the last
digit of 4n is the same as the last digit of 4 4 = 16. But 4n is also the number obtained by
performing
Create PDF with GO2PDF for free, if you a digit
wish slide
to remove this line,on
click n,
here so the
to buy Virtuallast two digits of n are 64. One may repeat this process
PDF Printer
to find all the digits of n. Solutions edited by Naoki Sato.
Solution 1 by: Caroline Suen (11/CA)
Letting X be the positive integer in question, 4X is the result of the digit slide on X.
The units digit of X is 4, and 4 4 = 16, so the units digit of 4X is 6, and the last two digits
of X are 64.
We can continue the argument as follows:
64 4 = 256, so the last two digits of 4X are 56, and the last three digits of X are 564,
564 4 = 2256, so the last three digits of 4X are 256, and the last four digits of X are
2564,
2564 4 = 10256, so the last four digits of 4X are 0256, and the last five digits of X are
02564,
02564 4 = 10256, so the last five digits of 4X are 10256, and the last six digits of X are
102564, and finally
102564 4 = 410256, which just happens to be the result of a digit slide on 102564.
Hence, 102564 is the smallest positive integer with 4 as its units digit such that the result
of a digit slide on the number equals 4 times the number.
Solution 2 by: Howard Tong (11/GA)
Let x be the number formed by the digits other than the digit 4, and let x have k digits.
Then the original number is 10x+4, and the number obtained from the digit slide is 410k +x.
Therefore,
4 (10x + 4) = 4 10k + x,
which implies that
39x = 4 10k 16.
www.usamts.org
The RHS is not divisible by 39 for k = 1, 2, 3, or 4, but when k = 5, 39x = 4 105 16 =

399984 x = 10256. Therefore, the smallest possible number is 102564.
Solution 3 by: Shobhit Vishnoi (12/SC)
Let the number we are looking for be S. We have that S = dn dn1 dn2 . . . d2 4, where each
dk represents a digit of the decimal expansion of S. Let us construct a rational repeating
decimal number N , where
N = 0.d d n n1
d n2
PDF Printer
. . . d2 4dn dn1 dn2 . . . .
By the conditions given in the problem, 4N must equal 0.4dn dn1 dn2 . . . d2 4dn dn1 dn2 . . . .
Thus, we have the following equations:
N = 0.dn dn1 dn2 . . . d2 4dn dn1 dn2 . . . , (1)

4N = 0.4dn dn1 dn2 . . . d2 4dn dn1 dn2 . . . . (2)
Multiplying equation (2) by 10, we get
40N = 4.dn dn1 dn2 . . . d2 4dn dn1 dn2 . . . . (3)
Subtracting equation (1) from equation (3) gives us 39N = 4, so

4
N= = 0.102564102564102564 . . . = 0.102564.
39
The repeating part of N is the desired number. Therefore, S = 102564. Checking, we see
that 4 102564 = 410256, and indeed satisfies the conditions.
Additional Comments. This problem resembles problem 1 from the 1962 IMO:
Find the smallest natural number n which has the following properties:
(a) Its decimal representation has 6 as the last digit.

(b) If the last digit 6 is erased and placed in front of the remaining digits, the
resulting number is four times as large as the original number n.
www.usamts.org
2/1/18.
(a) In how many different ways can the six empty circles in the diagram 1
(b) In how many different ways can the seven empty circles in the diagram
1
thanforboth
Create PDF with GO2PDF its
free, if you wishneighbors?
to remove this line, click here to buy Virtual PDF Printer
Comments This problem may be solved by dividing into cases, by considering which
numbers must be greater than or less than their neighbors. Solutions edited by Naoki Sato.
Solution 1 by: Aaron Pribadi (11/AZ)
(a) Starting from a position in the loop, if the next value is greater, then the third value
is less than the second, because the second value must be greater than both the first and
third. If the next value is less, then the third value is greater than the second, because the
second value must be less than both the first and third. This creates a higher, lower, higher,
lower pattern. This pattern cannot exist with an odd number of positions; therefore, it is
impossible to complete a seven-position diagram.
(b) Designate four high positions as those greater than their neighbors, and four low
that are less than their neighbors, in an alternating pattern.
L
H H
L L
H H
L
Neither 1 nor 2 may be in a high position because it would require two numbers lower
4

than it. Similarly, neither 7 nor 8 may be in a low position. So, there are 2 = 6 ways of
dividing the numbers 1 to 8 into the two groups high and low.
Case 1. Low: 1, 2, 3, 4 High: 5, 6, 7, 8
Because all of the lows are less than all of the highs and all of the highs are greater than
all of the lows, any low may be in any of the four low positions, and any high may be in
www.usamts.org
any high position. Because the 1 is in four possible positions rather than one fixed position,
there is a uniform overcounting by a factor of four. Therefore, for this case there are
4! 4!
= 144
4
possible arrangements.
Case 2. Low: 1, 2, 3, 5 High: 4, 6, 7, 8
For a given arrangement of the lows (4! arrangements), the 4 cannot be next to the 5,
so it has 2 possible positions. The other three highs may be in any of the three remaining
positions and still produce a valid arrangement (3! arrangements). Therefore, there are
4! 2 3!
= 72
4
Case 3. Low: 1, 2, 4, 5 High: 3, 6, 7, 8
For a given arrangement of the highs, neither the low-4 nor the low-5 can be next to
the high-3, so there are two positions and 2! arrangements. The other two lows may be in
any of the two remaining positions and still produce a valid arrangement (2! arrangements).
Therefore, there are
4! 2! 2!
= 24
4
Case 4. Low: 1, 2, 3, 6 High: 4, 5, 7, 8
For a given arrangement of the lows, neither the high-4 nor the high-5 can be next to
the low-6, so there are two positions and 2! arrangements. The other two highs may be in
any of the two remaining positions and still produce a valid arrangement (2! arrangements).
Therefore, there are
4! 2! 2!
= 24
4
Case 5. Low: 1, 2, 4, 6 High: 3, 5, 7, 8
For a given position of the 6 (4 possible locations), the 3 must be in a non-adjacent
location (2 possibilities). The 7 and 8 both must be next to the low-6, and the 1 and 2 both
must be next to the high-3, but the 7 and 8 can switch with each other, and the 1 and 2
can also switch (2 2 possibilities). That leaves the low-4 and high-5, each with only one
possible low or high location. Therefore, there are
4222
=8
4
www.usamts.org
Case 6. Low: 1, 2, 5, 6 High: 3, 4, 7, 8
The 7 and 8 are the only highs greater than the low 5 and 6, but the 7 and 8 cannot
surround both the 5 and 6, so no arrangement is possible.
Therefore, there are a total of 144 + 72 + 24 + 24 + 8 = 272 valid arrangements.
(a) Assume such a numbering exists. Let the numbers in the circles, starting from 1 going
counterclockwise, be 1, a1 , a2 , . . . , a6 . Since a1 > 1, we must also have a1 > a2 . It follows

that a2 < a3 , a3 > a4 , and so on. In general, if i is odd, ai is greater than its neighbors,
and if i is even, ai is less than its neighbors. Then a6 < 1, a contradiction. So there are no
possible numberings. This generalizes to any even number of circles.
(b) Let An be the answer to the general problem for n circles. We will find a recursive
formula for odd n. Let the numbers in the circles, starting from 1 going counterclockwise,
be 1, a1 , a2 , . . . , a2m+1 , where m is a nonnegative integer. As in part (a), ai is less than
its neighbors if and only if i is even. Since there are no two numbers the number 2 can be
greater than, ai = 2 only if i is even.
Suppose a2k = 2. Then the numbers 1 and 2 divide the larger circle into two arcs with
2k 1 and 2m 2k + 1 circles each. The numbers 3 through 7 can be distributed between
2m
the two arcs in 2k1 ways. Consider the numbers in the arc with 2k 1 circles. Since
every number is greater than 1 and 2, they can be arranged in the arc in the same number
of ways as they can be arranged in the original diagram with 2k 1 circles, which is A2k1 .
Similarly, the numbers in the other arc can be arranged in A2m2k+1 ways. Therefore, the
2m
total number of ways the empty circles can be filled in given a2k = 2 is 2k1 A2k1 A2m2k+1 .
Summing up the values for k = 1, 2, . . . , m, we have
m
X 2m
A2m+1 = A2k1 A2m2k+1 .
k=1
2k 1
We have A1 = 1, so A3 = 21 A21 = 2, A5 = 41 A1 A3 + 43 A3 A1 = 16, and

6 6 2 6
A7 = A1 A 5 + A3 + A5 A1 = 6 16 + 20 22 + 6 16 = 272.
1 3 5
Hence, the answer is 272.

www.usamts.org
Additional Comments. Using exponential generating functions, James Sundstrom

derived that the the number of arrangements for 2n numbers is
22n1 (22n 1)|B2n |

,
n
where Bn denotes the nth Bernoulli number.
www.usamts.org
3/1/18.
(a) An equilateral triangle is divided into 25 congruent smaller equi- 1
lateral triangles, as shown. Each of the 21 vertices is labeled with a
number such that for any three consecutive vertices on a line segment,
their labels form an arithmetic sequence. The vertices of the original
equilateral triangle are labeled 1, 4, and 9. Find the sum of the 21 labels.
4 9
(b) Generalize part (a) by finding the sum of the labels when there are
n2 smaller congruent equilateral triangles, and the labels of the original equilateral triangle
are a,
Create PDF with GO2PDF b, ifand
for free, c.to remove this line, click here to buy Virtual PDF Printer
you wish
Comments The desired sums can be systematically computed by recognizing that for any
consecutive vertices lying on a line, the labels form an arithmetic sequence. A symmetry
argument can also be used to give a quick solution. Solutions edited by Naoki Sato.
Solution 1 by: Erik Madsen (12/CA)
This solution develops a general equation for part (b) and then applies it to solve part
(a).
Without loss of generality, label the top vertex of the triangle a, the lower-left vertex b,
and the lower-right vertex c. Now note that in a triangle composed of n2 smaller triangles,
we have n + 1 horizontal rows of vertices the first row has 1 vertex, the second has 2, and so
on, with the bottom row having n + 1 vertices. The conditions of the problem indicate that
the vertices of each of these horizontal rows form an arithmetic sequence (since if any three
consecutive vertices form an arithmetic sequence, then any number of consecutive vertices
form an arithmetic sequence), as do the vertices of each of the diagonal edges of the triangle.
Therefore, the value of the left vertex in the mth row is
ba
lm = a + (m 1),
n
with m ranging from 1 to n + 1. Similarly, the value of the right vertex in the mth row is
ca
rm = a + (m 1),
n
with m ranging from 1 to n + 1. Using the formula for the sum of an arithmetic series, the
sum of the values of the vertices in row m is
m
sm = (lm + rm ) ,
2
www.usamts.org
where m ranges from 1 to n + 1.

To find the sum S of all vertices of the triangle, we must sum sm over all m:
n+1 n+1
X X m
S= sm = (lm + rm )
m=1 m=1
2
n+1
X ba ca m
= a+ (m 1) + a + (m 1)
m=1
n n 2
n+1
Create PDF with GO2PDF for free, if you wish to remove this b
X line, 2a
+clickc here m
to buy Virtual PDF Printer
= 2a + (m 1)
m=1
n 2
n+1 n+1
X b + c 2a X b + c 2a
= a m+ m2
m=1
2n m=1
2n

b + c 2a (n + 1)(n + 2) b + c 2a (n + 1)(n + 2)(2n + 3)
= a +
2n 2 2n 6

b + c 2a b + c 2a 2n + 3 (n + 1)(n + 2)
= a +
2n 2n 3 2

b + c 2a 2n + 3 (n + 1)(n + 2)
= a+ 1
2n 3 2

b + c 2a 2n (n + 1)(n + 2)
= a+
2n 3 2

b + c 2a (n + 1)(n + 2)
= a+
3 3
(a + b + c)(n + 1)(n + 2)
= .
6
Therefore,
(a + b + c)(n + 1)(n + 2)
S(n, a, b, c) = .
6
Applying this formula to part (a), we have S(5, 1, 4, 9) = 98.
(a) Consider two triangles, both divided into 25 smaller vertices and labeled such that the
labels of any three consecutive vertices on a line segment form an arithmetic sequence. Obtain
a third labeled triangle by adding the corresponding labels of the original two triangles (call
this operation addition of labeled triangles). The new triangle will also have the property that
www.usamts.org
the labels of any three consecutive vertices on a line segment form an arithmetic sequence,
because, if a1 , b1 , c1 and a2 , b2 , c2 are both arithmetic sequences, then
(c1 + c2 ) (b1 + b2 ) = (c1 b1 ) + (c2 b2 )

= (b1 a1 ) + (b2 a2 )
= (b1 + b2 ) (a1 + a2 ),
and so a1 + a2 , b1 + b2 , c1 + c2 is also an arithmetic sequence. It is easy to see that the sum

of the labels of the new triangle is equal to the sum of the sums of the labels of the original
triangles.
Say that a triangle is labeled with (a, b, c) if a is at the top vertex, b is at the bottom left
vertex, and c is at bottom right vertex, and all other labels follow the arithmetic sequence
rule. Add the three triangles labeled with (1, 4, 9), (4, 9, 1), and (9, 1, 4). The result is the
triangle labeled with (14, 14, 14). Since, by symmetry, the original three triangles must have
the same label sum, the label sum of this triangle is equal to three times the label sum of
the triangle labeled with (1, 4, 9). All the labels of the triangle labeled with (14, 14, 14) are
the same, and since the triangle in question has 21 vertices, its label sum is 21 14 = 294.
Therefore the label sum of the triangle labeled with (1, 4, 9) is 294/3 = 98.
(b) Let fn (a, b, c) denote the sum of the labels of a triangle labeled with (a, b, c) when
there are n2 smaller triangles. The function fn is additive, i.e.
fn (a1 , b1 , c1 ) + fn (a2 , b2 , c2 ) = fn (a1 + a2 , b1 + b2 , c1 + c2 ),
as noted in part (a). By symmetry, fn (a, b, c) = fn (b, c, a) = fn (c, a, b), so
3fn (a, b, c) = fn (a, b, c) + fn (b, c, a) + fn (c, a, b) = fn (a + b + c, a + b + c, a + b + c).
Let Tn denote the nth triangular number. Then all Tn+1 labels of the triangle labeled with
(a+b+c, a+b+c, a+b+c) are the same, namely a+b+c, so fn (a+b+c, a+b+c, a+b+c) =
Tn+1 (a + b + c). Therefore,
1
fn (a, b, c) = fn (a + b + c, a + b + c, a + b + c)
3
1
= Tn+1 (a + b + c)
3
(n + 1)(n + 2)(a + b + c)
= .
6
www.usamts.org
4/1/18. Every point in the plane is colored either red, green, or blue. Prove that there
exists a rectangle in the plane such that all four of its vertices are the same color.
Credit This problem was proposed by Dave Patrick, and comes from the Carnegie Mellon
Mathematical Studies Problem Seminar.
Comments Once you have the idea of using a sufficiently large grid of points, the problem
quickly reduces to an application of the Pigeonhole principle. Note to USAMTS students:
It is more important to get the stated problem correct, before moving onto proving a gener-
alization.
Create PDF with GO2PDF Solutions
for free, if you edited
wish to remove this line, clickby
hereNaoki
to buy VirtualSato.
PDF Printer
Consider a 4 82 rectangle of points in the plane, such as {(x, y) Z2 | 0 x 3, 0

y 81}. For each column, there are 4 points and 3 possible colors per point, for a total
of 34 = 81 possible colorings. With 82 columns, by the Pigeonhole Principle, there are two
columns with the same coloring. Also, there are 4 points per column and 3 possible colors,
so by the Pigeonhole Principle, some color appears twice. From each of the two columns,
take some corresponding two points of a color that appears twice. These form a rectangle
all of whose vertices are the same color.
Solution 2 by: Sam Elder (11/CO)
Consider a 4 19 grid of points in this plane. For each row of 4 points, by the Pigeonhole
Principle, two must be the same color, for instance green. Denote such a row green (a
row can be two colors simultaneously) and consider the colors of all 19 rows. Again by the
Pigeonhole Principle, 7 must be the same color. Without loss of generality, assume this color
is green.
4
Now consider the placement of the two green points out of four in each row. There are
2
= 6 ways to place two green points in four spots, so again by the Pigeonhole Principle,
two of the seven rows must have the same placement. By choosing the four green points in
those two rows, we form a monochromatic rectangle, as desired.
www.usamts.org
5/1/18. ABCD is a tetrahedron such that AB = 6, BC = 8, AC = AD = 10, and

BD = CD = 12. Plane P is parallel to face ABC and divides the tetrahedron into two
pieces of equal volume. Plane Q is parallel to face DBC and also divides ABCD into two
pieces of equal volume. Line ` is the intersection of planes P and Q. Find the length of the
portion of ` that is inside ABCD.
Comments This problem is best solved by using similar tetrahedra, and drawing a nice
diagram. To solve three-dimensional geometry problems, one technique that may help is to
consider the two-dimenaional analogue. Solutions edited by Naoki Sato.
Let A0 denote the intersection of plane P and AD, and define points B 0 and C 0 similarly.
Let B 00 denote the intersection of plane Q and AB, and define points C 00 and D00 similarly.
Let B 000 denote the intersection of P, Q, and face ABD, and let C 000 denote the intersection
of P, Q, and face ACD. Then the problem asks for the length of B 000 C 000 .
www.usamts.org
Tetrahedrons ABCD and A0 B 0 C 0 D are similar because plane P is parallelto face ABC.
0 0 0 0 3 3
The volume of ABCD is twice the volume of A B C D, so A D = AD/ 2 = 10/ 2.
00 3 0 00 0 00 0 00 3
Similarly,
AD = 10/ 2. Since A D+AD = AD+A D , we find that A D = 20/ 210 =
3
10( 4 1).
Since plane P and face ABC are parallel, and plane Q and face DBC are parallel,
tetrahedrons ABCD and A0 B 000 C 000 D00 are similar. Therefore,

000 000 BC A 0 00
D 8 10( 3
4 1)
3
B C = = = 8( 4 1).
AD 10
www.usamts.org
1/2/18. Find all positive integers n such that the sum of the squares of the digits of n is
2006 less than n.
Credit This problem was proposed by Dave Patrick.
Comments The simplest approach in this problem begins with establishing good bounds,
which helps to reduce the casework. For example, with the proper bounds, you can show
that n must have exactly four digits. Solutions edited by Naoki Sato.
Solution 1 by: Carlos Dominguez (10/OH)

We first show that n must have exactly 4 digits. Since the sum of the squares of the
digits is positive, n > 2006. Now suppose n has d digits with d 5. The smallest integer
with d digits is 10d1 , and the largest possible sum of the squares of the digits is 92 d = 81d.
We claim that
81d < 10d1 2006.
When d = 5, we get 405 < 10000 2006 = 7994, which is true. Now suppose that it
holds for d = k, for some positive integer k 5. Then add 81 to both sides to get
81(k + 1) < (10k1 + 81) 2006 < 10k 2006,
which completes the induction. This implies if d 5, then the sum of squares of the digits
of n must be less than n 2006. Therefore, n must have exactly 4 digits.
We want to find digits (a, b, c, d), where n = 1000a + 100b + 10c + d, such that a2 + b2 +
c2 +d2 = 1000a+100b+10c+d2006. The maximum value of a2 +b2 +c2 +d2 is 92 4 = 324,
which means that n is at most 324 + 2006 = 2330. Since we already know that n > 2006,
the first digit must be a = 2.
Substituting into our equation, we get
b2 + c2 + d2 = 100b + 10c + d 10.
Case 1: b = 0. We have
c2 10c + d2 d = 10
2
2 1 1
(c 5) + d = 25 + 10
2 4
2 2
(2c 10) + (2d 1) = 61.
We see that 61 can be written as the sum of two squares only as 62 + 52 . This gives the
solutions (c, d) = (2, 3), (8,3). So two possible values of n are 2023 and 2083. Indeed,
22 + 02 + 22 + 32 + 2006 = 2023 and 22 + 02 + 82 + 32 + 2006 = 2083.
www.usamts.org
Case 2: b > 0. If b = 1, then
c2 10c + d2 d = 89
2
2 1 1
(c 5) + d = 89 + 25 +
2 4
(2c 10)2 + (2d 1)2 = 457.
However, since the left hand side is at most (2 0 10)2 + (2 9 1)2 = 389, equality is
impossible. ((2c 10)2 is maximized when c is the the smallest digit 0, whereas (2d 1)2 is
maximized when d is the largest digit 9.) For the same reason, the cases b = 2 and b = 3
are impossible as well.
Therefore, the only answers are 2023 and 2083.
www.usamts.org
2/2/18. ABCD is a trapezoid with BC k AD, ADC = B M C

57 , DAB = 33 , BC = 6, and AD = 10. M and N are the
midpoints of BC and AD, respectively.
A D
N
(a) Find M N A.
(b) Find M N .
Credit This problem was proposed by Gregory Galperin.

Comments The fact that the angles 57 and 33 are complementary is a strong indication to
complete the right triangle, as in the solution below. Once this right triangle is constructed,
the required data is easy to calculate. The fact that lines AB, CD and M N are concurrent
must be justified. Solutions edited by Naoki Sato.
Solution 1 by: David Corwin (10/MA)
Extend AB and DC to meet at point E. Because BC k AD, segments BC and AD are

homothetic with respect to point E, with ratio AD
BC
= 10
6
= 53 . Because M is the midpoint of
BC, its corresponding homothetic point on AD is the midpoint of AD, which is N , so M
and N are homothetic with respect to point E, and therefore E, M , and N are collinear.
B C
M
A D
N
(a) By triangle AED, AED = 180 EAD EDA = 180 33 57 = 90 ,

so triangle AED is right. Because N is the midpoint of hypotenuse AD, N E = AN , so
triangle AN E is isosceles, and N EA = N AE = 33 . Then by triangle AN E, AN E =
AN M = 180 N EA N AE = 180 33 33 = 114 .
(b) Because N is the midpoint of AD, N E = AN = AD
2
= 10
2
= 5. By the homothety,
NE 5
M E = 5 = 5 = 3. Then M N = N E M E = 5 3 = 2.
3 3
www.usamts.org
3/2/18. The expression bxc means the greatest integer that is smaller than or equal to x,
and dxe means the smallest integer that is greater than or equal to x. These functions are
called the floor function and ceiling function, respectively. Find, with proof, a polynomial
f (n) equivalent to
n2 j
X k l m
k + k
k=1
for all positive integers n.
Credit This problem was proposed by Scott Kominers, a past USAMTS participant.
Comments The first thing we want to do in this sum is remove the floor and ceiling
notation. Since k is an integer when k is a perfect square, we can consider what happens
when k lies between consecutive perfect squares. Once the floor and ceiling brackets have
been removed, the rest of the problem is an exercise in algebra using standard summation
formula. Solutions edited by Naoki Sato.
Let m be a positive
integer. For
1)2 + 1 k m2 1, we have (m 1)2 <
(m
k< m2 m 1 < k < m b kc + d ke = (m 1) + m = 2m 1. For k = m ,
2
b kc + d ke = m + m = 2m. Hence,
m 2
X
(b kc + d ke) = [(m2 1) (m 1)2 ](2m 1) + 2m
k=(m1)2 +1
= (m2 1 m2 + 2m 1)(2m 1) + 2m
= (2m 2)(2m 1) + 2m
= 4m2 4m + 2,
www.usamts.org
which implies

n2 n 2
m
X X X
(b kc + d ke) = (b kc + d ke)
k=1 m=1 k=(m1)2 +1
n
X
= (4m2 4m + 2)
m=1
Xn n
X n
X
2
=4 m 4 m+2 1
m=1 m=1 m=1
n(n + 1)(2n + 1) n(n + 1)
=4 4 + 2n
6 2
4(2n3 + 3n2 + n) 12(n2 + n) 12n
= +
6 6 6
3
8n + 4n
=
6
4n3 + 2n
= .
3
Therefore,
4n3 + 2n
f (n) = .
3
www.usamts.org
4/2/18. For every integer k 2, find a formula (in terms of k) for the smallest positive
integer n that has the following property:
No matter how the elements of {1, 2, . . . , n} are colored red and blue, we can find
k elements a1 , a2 , . . . , ak , where the ai are not necessarily distinct elements, and
an element b such that:
(a) a1 + a2 + + ak = b, and
(b) all of the ai s and b are the same color.
Credit This problem was proposed by Dave Patrick, and is a generalization of a problem
that appeared on the 2004 British Mathematical Olympiad.
Comments There are two parts to this problem: You must show that for n = k 2 + k 2,
there is a coloring that does not satisfy the given property, and you must show that for
n = k 2 + k 1, any coloring satisfies the given property.
The first part can be accomplished by explicitly constructing a counter-example, and the
second part can be shown by considering the colors of only a few key numbers. Solutions
edited by Naoki Sato.
The answer is n = k 2 + k 1.
First, we show that for n = k 2 + k 2, we can produce a coloring that does not satisfy
these criteria. Let the numbers 1 to k 1 be red, k to k 2 1 be blue, and k 2 to k 2 + k 2
be red. Any k blue numbers sum to at least k 2 , and all numbers at least k 2 are red. Also,
if we choose k red numbers less than k, we get a total sum of at most k(k 1) < k 2 but at
least k, and all of these numbers are blue. Moreover, if we choose at least one red number
that is at least k 2 , our sum is at least k 2 + k 1, which is not in our set. So no matter which
k identically-colored numbers we choose, their sum is not the same color.
Now, we show that n = k 2 + k 1 does work. Assume for the sake of contradiction that
we cannot find k + 1 such integers as described in the problem. Without loss of generality,
let 1 be red. Then k must be blue and k 2 must be red. Summing k 2 + |1 + {z
+ 1}, k 2 + k 1
k1
must also be blue. Now this means k + 1 must be red, because otherwise we would have
k + (k + 1) + + (k + 1) = k 2 + k 1, with k, k + 1 and k 2 + k 1 blue, contradiction.
| {z }
k1
But then we get 1 + (k + 1) + + (k + 1) = k 2 , and 1, k + 1 and k 2 are red, contradiction.
| {z }
k1
Therefore, for n = k 2 + k 1, any coloring satisfies the given property.
www.usamts.org
5/2/18. In triangle ABC, AB = 8, BC = 7, and AC = 5. We extend D

AC past A and mark point D on the extension, as shown. The bisector of E A
DAB meets the circumcircle of 4ABC again at E, as shown. We draw F
a line through E perpendicular to AB. This line meets AB at point F .
Find the length of AF . C
Credit This problem was proposed by Richard Rusczyk. B
Comments An angle chase shows that triangle BEC is equilateral. Then

the length of AF can be found with an application of Ptolemys theorem. Solutions edited
by Naoki Sato.
Solution 1 by: Scott Kovach (11/TN)
D
E A
Applying the law of cosines to triangle ABC, we see that

82 + 52 72 1
cos BAC = = ,
285 2
so BAC = 60 . Then EAF = DAF/2 = (180 BAC)/2 = 60 as well.
Now, BEC subtends the same arc as BAC, and EBC subtends the arc comple-
mentary to EAC, so EBC = BEC = BAC = 60 , which makes triangle BEC
equilateral.
Quadrilateral EACB is cyclic, so by Ptolemys theorem,
EA BC + EB AC = AB EC
EA 7 + 7 5 = 8 7
EA = 3.
Finally, triangle EAF is a 30 -60 -90 triangle, so AF = EA/2 = 3/2.

www.usamts.org
1/3/18. In how many distinguishable ways can the edges of a cube be colored such that
each edge is yellow, red, or blue, and such that no two edges of the same color share a
vertex? (Two cubes are indistinguishable if they can be rotated into positions such that the
two cubes are colored exactly the same.)
Credit This problem was proposed by Richard Rusczyk and Sam Vandervelde.
Comments This problem is best solved by systematic casework and well-drawn diagrams.
Solutions edited by Naoki Sato.
Solution 1 by: Igor Tolkov (10/WA)
The four possible colorings are as shown.
Without loss of generality, color three mutually adjacent edges blue, red, and yellow as
shown in the top layer. Then, consider the edge marked by 1. Because an adjacent edge
is red, this edge can be either yellow or blue.
First, assume that it is yellow. Then consider the edge marked by 2. This edge can be
either blue or red. If it is red, then all blue edges are uniquely determined and there exist
two options for the remaining four edges. This produces two colorings: One with parallel
www.usamts.org
edges of the same color and one with red and yellow edges alternating. If edge 2 is blue,
then the other edges are uniquely determined and we obtain a coloring with blue and red
edges alternating.
Now, if edge 1 is blue, then the other edges are again uniquely determined and we obtain
a coloring with blue and yellow edges alternating. This covers all cases.
Note that all four cubes have a plane of symmetry so orientation does not matter. In
other words, each cube can be rotated to account for orientation.
www.usamts.org
2/3/18. Find, with proof, all real numbers x between 0 and 2 such that
tan 7x sin 6x = cos 4x cot 7x.
Credit This problem was proposed by Marcin E. Kuczma from the University of Warsaw.
Comments By expressing everything in terms of sine and cosine, it is seen that the given
equation is actually an inequality in disguise, where equality must occur. The equation can
then be solved by finding where equality occurs. Solutions edited by Naoki Sato.
Solution 1 by: Bohua Zhan (12/NJ)
Writing everything in terms of sine and cosine and rearranging, we have:

sin 7x cos 7x
sin 6x = cos 4x
cos 7x sin 7x
sin 7x cos 7x
+ = cos 4x + sin 6x
cos 7x sin 7x
sin2 7x + cos2 7x
= cos 4x + sin 6x
sin 7x cos 7x
1
= cos 4x + sin 6x
sin 7x cos 7x
2
= cos 4x + sin 6x
sin 14x
2 = sin 14x(cos 4x + sin 6x).
Since the range of sine and cosine are [1, 1], | sin 14x| 1 and | cos 4x + sin 6x| 2 for
all x. Since the product of these two expressions is 2, they must all attain the maximum
value. That is, | sin 14x| = 1, | sin 6x| = 1, and cos 4x = sin 6x. There are two cases:
Case 1: If sin 14x = 1, then cos 4x = sin 6x = 1. So 4x = k, where k is an odd
integer. Then for x between 0 and 2, we have x = 4 , 3
4
, 5
4
, 7
4
. It is not difficult to verify
that only x = 4 and x = 5
4
satisfy the other two equations.
Case 2: If sin 14x = 1, then cos 4x = sin 6x = 1. So 4x = k, where k is an even integer.
For x between 0 and 2, we have x = 0, 2 , , 32
, 2. Note that for all four possible values
of x, 6x is a multiple of , and sin 6x = 0. Therefore, there are no solutions in this case.
5
In conclusion, the solutions of x between 0 and 2 are 4
and 4
.
www.usamts.org
3/3/18. Three circles with radius 2 are drawn in a plane such that each circle is tangent
to the other two. Let the centers of the circles be points A, B, and C. Point X is on the
circle with center C such that AX + XB = AC + CB. Find the area of 4AXB.
Comments The condition AX + XB = AC + CB leads to X lying on an ellipse, whose

equation can be determined with analytic geometry. Solutions edited by Naoki Sato.

Position the three circles in the coordinate plane

so that A = (2,
0) and B = (2, 0),and
2 2
let O be the origin. We can easily calculate CO = CA OA = 2 3, so let C = (0, 2 3).
Now, note that the locus of all points X such that AX + XB = AC + CB is an ellipse E.
More specifically, E is centered at O and has foci at A and B. If we let A0 = (4, 0) and
B 0 = (4, 0), then AA0 + A0 B = AB 0 + B 0 B = 8, so A0 and B 0 lie on the ellipse as well. The
ellipse passes through C as well, so the equation of E is given by
x2 y 2 4y 2
+ =1 x2 + = 16.
16 12 3
To locate X, we want to find the point at which the ellipse E intersects the circle centered
at C, which has the equation
x2 + (y 2 3)2 = 4.
Substituting, we obtain
4y 2
y2
4 (y 2 3)2 + = 16
+ 4 3y = 24.
3
3

Solving this quadratic and taking the positive root yields y = 6 5 6 3. This is equal to
the altitude of
triangle
AXB, with respect to base AB, so the area of this triangle is simply
1
2
4 y = 12 5 12 3.
www.usamts.org
4/3/18. Alice plays in a tournament in which every player plays a game against every
other player exactly once. In each game, either one player wins and earns 2 points while
the other gets 0 points, or the two players tie and both players earn 1 point. After the
tournament, Alice tells Bob how many points she earned. Bob was not in the tournament,
and does not know what happened in any individual game of the tournament.
(a) Suppose there are 12 players in the tournament, including Alice. What is the smallest
number of points Alice could have earned such that Bob can deduce that Alice scored
more points than at least 8 other players?
(b) Suppose there are n players in the tournament, including Alice, and that Alice scored
m points. Find, in terms of n and k, the smallest value of m such that Bob can deduce
that Alice scored more points than at least k other players.
Credit This problem is based on a problem that appeared in Problem Solving Journal for
secondary students, a British publication.
Comments To show that that the answer is m = n + k 1, we must prove two things:
First, we must prove that it is possible for Alice to score n + k 2 points while beating at
most k 1 other players. Second, we must prove that if Alice scores n + k 1 points, then
she must have scored more points than at least k other players. Solutions edited by Naoki
Sato.
Solution 1 by: Matt Superdock (10/PA)
We will find a generalization in part (b) and use it to find the answer to part (a).
We claim that the smallest value of m such that Bob can deduce that Alice scored more
points than at least k other players is m = n + k 1.
If Alice scores more points than at least k other players, then there are at most n k
players, including Alice, that scored m or more points. It is sufficient to prove that it is
impossible for n k + 1 players to each score n + k 1 or more points, and that it is possible
for n k + 1 players to each score n + k 2 or more points.
Suppose for the sake of contradiction that n k + 1 players each score n + k 1 or more
points. Call these n k + 1 players good players, and call the other k 1 players bad players.
The good players must collectively score a total of at least (n k + 1)(n + k 1) points.
There are nk+1

2
games between two good players, and there are two points available
in each game. In these games, good players can accumulate a total of 2 nk+1

2
= (n k +
1)(n k) points. There are (n k + 1)(k 1) games between one good player and one bad
player, and there are again two points available in each game. In these games, good players
www.usamts.org
can accumulate a total of (n k + 1)(2k 2) points. Good players cannot accumulate points
in games between two bad players. Therefore, good players can accumulate a total of at
most (n k + 1)(n k) + (n k + 1)(2k 2) = (n k + 1)(n + k 2) points, which is less
than (n k + 1)(n + k 1). We have reached a contradiction, so it is impossible for n k + 1
players to each score n + k 1 of more points.
It remains to be shown that it is possible for n k + 1 players to each score n + k 2
or more points. Again, call these n k + 1 players good players, and call the other k 1
players bad players. Every good player plays n k games against good players and k 1
games against bad players. Suppose that every game between two good players results in a
tie, and every game between a good player and a bad player results in a win for the good
player. In this scenario, each good player gets exactly n k + 2(k 1) = n + k 2 points,
as desired.
Therefore the answer is indeed m = n + k 1.
Applying our generalization to part (a), we find that the answer is 12 + 8 1 = 19.
Further Comments. Some students attempted to prove that if Alice scores n + k 1

points, then she must have scored more points than at least k other players by setting up a
worst-case scenario, consisting of a number of top players that tie each other, and beat all
worse players.
Proofs that simply state this worst-case scenario are not sufficiently rigorous. Proofs that
began with this scenario and used the idea that for every point one player gained, another
player lost, were also generally unconvincing. You must prove that Alice scores more points
than at least k other players in all possible scenarios, and the easiest way is as done above.
www.usamts.org
5/3/18. Let f (x) be a strictly increasing function defined for all x > 0 such that f (x) > x1
and f (x)f (f (x) + x1 ) = 1 for all x > 0. Find f (1).
Credit This problem was proposed by Joe Jia.
Comments One important step in solving this functional equation is to substitute f (x)+1/x
for x into the functional equation itself, a step which is suggested by the form of the functional
equation. Then the strictly increasing condition can be used to solve for f (x). Solutions
Solution 1 by: Vlad Firoiu (9/MA)
From the given equation,

1 1
f f (x) + = .
x f (x)
1
Since y = f (x) + > 0 is in the domain of f , we have that
x

1 1 1
f f (x) + f f f (x) + + = 1.
x x f (x) + x1
1 1

Substituting f f (x) + x
= into the above equation yields
f (x)

1 1 1
f + = 1,
f (x) f (x) f (x) + x1
so that
1 1
f + 1 = f (x).
f (x) f (x) + x
Since f is strictly increasing, it must be 1 to 1. In other words, if f (a) = f (b), then

a = b. Applying this to the above equation gives
1 1
+ 1 = x.
f (x) f (x) + x
Solving yields that

1 5
f (x) = .
2x
Now, if for some x in the domain of f ,

1+ 5
f (x) = ,
2x
www.usamts.org
then
1 5 1+ 5
f (x + 1) = < = f (x).
2x + 2 2x
This contradicts the strictly increasing nature of f , since x < x + 1. Therefore,

1 5
f (x) =
2x
for all x > 0. Plugging in x = 1 yields

1 5
f (1) = .
2
www.usamts.org
n
X
1/4/18. Let S(n) = (1)i+1 i. For example, S(4) = 1 2 + 3 4 = 2.
i=1
(a) Find, with proof, all positive integers a, b such that S(a) + S(b) + S(a + b) = 2007.
(b) Find, with proof, all positive integers c, d such that S(c) + S(d) + S(c + d) = 2008.
Credit This problem was proposed by Dave Patrick, and was based on a discussion at the
2006 World Federation of National Mathematics Competitions conference.
Comments Since both parts have the form S(m) + S(n) + S(m + n), it is easiest to analyze
this form first to solve for a, b, c and d. Solutions edited by Naoki Sato.
If n is even, then
n
S(n) = (1 2) + (3 4) + + [(n 1) n] = 1 1 1 = .
| {z } 2
n/2 1s
If n is odd, then S(n) = S(n 1) + n = n1

2
+n = n+1
2
. We now consider the expression
T (m, n) = S(m) + S(n) + S(m + n).
Case 1. Both m and n are odd. Then m + n is even, so
m+1 n+1 m+n
T (m, n) = + = 1.
2 2 2
Case 2. Both m and n are even. Then m + n is even, so

m n m+n
T (m, n) = = m n < 0.
2 2 2
Case 3. m is odd and n is even. Then m + n is odd, so

m+1 n m+n+1
T (m, n) = + = m + 1,
2 2 2
which is even.
Case 4. n is odd and m is even. Analogously with the previous case, T (m, n) = n + 1,
which is again even.
None of these cases yield T (m, n) = 2007, so there are no solutions to part (a). For part
(b), we can use either case 3 or 4, with the only difference being the ordering in the pairs.
In Case 3, m = 2007 and n is even, and in Case 4, n = 2007 and m is even. Hence, the
solutions are (c, d) = (2007, n) and (c, d) = (n, 2007), where n is any even positive integer.
www.usamts.org
2/4/18. For how many integers n between 1 and 102007 , inclusive, are the last 2007 digits
of n and n3 the same? (If n or n3 has fewer than 2007 digits, treat it as if it had zeros on
the left to compare the last 2007 digits.)
Credit This problem was proposed by Paul Bateman, Professor Emeritus at the University
of Illinois at Urbana-Champaign.
Comments When solving a congruence with modulus m, we can look at the congruence
with respect to each prime factor of m. Then, the solutions can be sewn together using the
Chinese
Create PDF with GO2PDF for free, Remainder Theorem.
if you wish to remove this line, click here toSolutions edited by Naoki Sato.
buy Virtual PDF Printer
Saying that the last 2007 digits of n and n3 are the same is equivalent to saying that
n n3 (mod 102007 ), or
n(n 1)(n + 1) 0 (mod 102007 ).
Therefore, n(n 1)(n + 1) is divisible by both 52007 and 22007

Since only one of n, n 1, and n + 1 can be divisible by 5, whichever one is divisible by
5 must also be divisible by 52007 , so
n 0, 1, or 1 (mod 52007 ).
Similarly, if n is even, then both n 1 and n + 1 are odd, so n 0 (mod 22007 ). On

the other hand, if n is odd, then (n 1)(n + 1) 0 (mod 22007 ). However, the difference
between n 1 and n + 1 is 2, so only one of them can be divisible by 4. Call this one n 1.
Hence, n 1 is divisible by 2 but not 4. Therefore, n 1 must be divisible by 22006 in order
that (n 1)(n + 1) 0 (mod 22007 ), so n 1 (mod 22006 ). Hence, if n is odd,
n 1, 22006 1, 22006 + 1, or 22007 1 (mod 22007 ).
Recall that if n is even, then n 0 (mod 22007 ).

There are three possible values of n modulo 52007 and five possible values of n modulo
2 . By the Chinese Remainder Theorem, there are 15 possible values of n modulo 102007 ,
2007
which means there are 15 solutions n for 1 n 102007 .

www.usamts.org
3/4/18. Let ABCD be a convex quadrilateral. Let M be the midpoint C

D
of diagonal AC and N be the midpoint of diagonal BD. Let O be the M
intersection of the line through N parallel to AC and the line through M O
parallel to BD. Prove that the line segments joining O to the midpoints N
of each side of ABCD divide ABCD into four pieces of equal area. A
B
Credit This problem is based on a problem from the Canadian IMO
training program.
Comments This problem can be succinctly solved by using formulas for the areas of
triangles involving their bases and heights. Solutions edited by Naoki Sato.
Solution 1 by: Kenan Diab (12/OH)
C
R
D
M
S O
Q
N
Let P , Q, R, and S be the midpoints of AB, BC, CD, and DA, respectively. Consider
quadrilateral AP M S. By definition of midpoint, we have AB = 2AP , AC = 2AM , and
AD = 2AS. Thus, a homothecy centered at A maps quadrilateral AP M S to quadrilateral
ABCD with a factor of 2. Hence, [AP M S] = [ABCD]/4 and SP k BD.
But, we are given OM k BD, so OM k SP . Thus, O and M are the same distance from
SP . Since 4OSP and 4M SP share side SP , it follows that [OSP ] = [M SP ]. Thus,
[ABCD]
[OP AS] = [OSP ] + [ASP ] = [M SP ] + [ASP ] = [AP M S] = .
4
Analogous homothecies centered at B, C, and D give [OP AS] = [OQBP ] = [ORCQ] =
[OSDR] = [ABCD]/4, as desired.
www.usamts.org
4/4/18. We are given a 2 n array of nodes, where n is a positive integer. A valid

connection of the array is the addition of 1-unit-long horizontal and vertical edges between
nodes, such that each node is connected to every other node via the edges, and there are no
loops of any size. We give some examples for n = 3:
Valid Valid Invalid: loop Invalid: not connected
Let Tn denote the number of valid connections of the 2 n array. Find T10 .
Comments By constructing valid connections on a 2 n array from smaller arrays, we can

obtain a recursive formula for Tn . Solutions edited by Naoki Sato.
Solution 1 by: Drew Haven (11/CA)
It is trivial to note that T1 = 1 because there is only one way to connect two nodes. Let
us compute Tn+1 from Tn . Given any valid connection of 2 n nodes, adding two nodes to
the right gives an array of size 2 (n + 1). The additional two nodes may be connected in
one of three ways:
... , ... , ...
None of these result in any loops. This gives a total of 3Tn valid connections.
However, it is possible that a valid connection on a 2 (n + 1) array does not contain
a valid connection in its leftmost 2 n nodes. Let us consider the case where the leftmost
2 (n 1) nodes form a valid connection, but the leftmost 2 n nodes do not. There are two
different ways to connect the nodes on the right to make a valid connection on a 2 (n + 1)
array:
... , ...
This adds 2Tn1 ways to the total count. Similarly, if only the leftmost 2 (n 2) nodes
make a valid connection, there are 2Tn2 ways:
... , ...
www.usamts.org
Likewise, there are 2Tk ways for each k < n that come from the leftmost 2 k nodes
forming a valid connection. The last case to consider is the case when the leftmost two nodes
are not connected, and there are no valid connections of any 2 k leftmost subarray up to
k = n. There is only one such valid connection:
...
Summing these ways gives a formula for Tn+1 :

Tn+1 = 3Tn + 2Tn1 + 2Tn2 + + 2T1 + 1. (1)
To find a simpler recurrence, we subtract

Tn = 3Tn1 + 2Tn2 + + 2T1 + 1
from this to give
Tn+1 Tn = (3Tn + 2Tn1 + 2Tn2 + + 2T1 + 1)
(3Tn1 + 2Tn2 + + 2T1 + 1)
= 3Tn Tn1
Tn+1 = 4Tn Tn1 ,
which can be rewritten as
Tn+2 = 4Tn+1 Tn . (2)
From (1), T2 = 3T1 + 1 = 4. Then from (2),

T3 = 4 4 1 = 15,
T4 = 4 15 4 = 56,
T5 = 4 56 15 = 209,
T6 = 4 209 56 = 780,
T7 = 4 780 209 = 2911,
T8 = 4 2911 780 = 10864,
T9 = 4 10864 2911 = 40545,
T10 = 4 40545 10864 = 151316.
As a side note, an explicit formula for Tn can be found using generating functions:

(2 + 3)n (2 3)n
Tn = . (3)
2 3
Substituting 10 for n here gives the same answer, T10 = 151316.
www.usamts.org
5/4/18. A sequence of positive integers (x1 , x2 , . . . , x2007 ) satisfies the following two con-
ditions:
(1) xn 6= xn+1 for 1 n 2006, and

x1 + x2 + + xn
(2) An = is an integer for 1 n 2007.
n
Find the minimum possible value of A2007 .
Credit This problem was a former proposal for the Canadian Mathematical Olympiad.
Comments Finding the optimal sequence is not difficut, but a high degree of rigor and
careful reasoning must be employed to show conclusively that you have the minimum value.
In particular, using a greedy algorithm is not sufficient. Both conditions (1) and (2) must
be used effectively. Solutions edited by Naoki Sato.
We claim that the minimum value of An is n+1

2
. This value is achieved for the sequence

n+1
for odd n,

2

xn =

3n

for even n.
2
Indeed, if n 2 is even, then xn1 = n/2 and xn+1 = (n + 2)/2, both of which are less
than xn = 3n/2. Hence, no two consecutive terms are equal, so condition (1) is satisfied.
For even n,
(x1 + x3 + + xn1 ) + (x2 + x4 + + xn )

An =
n
(1 + 2 + + n/2) + (3 + 6 + + n/2)
=
n
4(1 + 2 + + n/2)
=
n
4 n/2 (n + 2)/2
=
2n
n+2 n+1
= = ,
2 2
www.usamts.org
and for odd n,

(x1 + x3 + + xn2 ) + (x2 + x4 + + xn1 ) + xn
An =
n
[1 + 2 + + (n 1)/2] + [3 + 6 + + 3(n 1)/2] + (n + 1)/2
=
n
4[1 + 2 + + (n 1)/2] + (n + 1)/2
=
n
4 1/2 (n 1)/2 (n + 1)/2 + (n + 1)/2
=
here to buyVirtualnPDF Printer
Create PDF with GO2PDF for free, if you wish to remove this line, click
n+1 n+1
= = .
2 2
We now prove this is the minimum through induction. It is true for n = 1, because the
minimum of A1 is 1 = 1+1 2
. For n = 2, if A2 = 1, then x1 + x2 = 2 x1 = x2 = 1, which
2+1
contradicts (1). Hence, the minimum of A2 is 2 = 2 .
Now, assume that A2m 2m+1

2
= m + 1 for some positive integer m. Let Sn =
x1 + x2 + + xn . In particular, Sn must be a multiple of n. We have S2m = 2mA2m
2m(m + 1) = 2m2 + 2m. Also,
2m2 + m < 2m2 + 2m < 2m2 + 3m + 1

m(2m + 1) < 2m2 + 2m < (m + 1)(2m + 1),
so the least multiple of 2m + 1 greater than 2m2 + 2m is (m + 1)(2m + 1). Since S2m+1 >
S2m 2m2 + 2m, we have S2m+1 (m + 1)(2m + 1), so

(2m + 1) + 1
A2m+1 m + 1 = .
2
Note that 2m2 + 2m = m(2m + 2) is a multiple of 2m + 2. The next greatest multiple of

2m + 2 is (m + 1)(2m + 2). Suppose that S2m+2 = (m + 1)(2m + 2) = 2m2 + 4m + 2. Then
2m2 + 3m + 1 < 2m2 + 4m + 2 < 2m2 + 5m + 2

(m + 1)(2m + 1) < 2m2 + 4m + 2 < (m + 2)(2m + 1),
so the greatest multiple of 2m + 1 less than 2m2 + 4m + 2 is (m + 1)(2m + 1). Since

S2m+1 < S2m+2 = 2m2 + 4m + 2, we have S2m+1 (m + 1)(2m + 1). But we have already
shown that S2m+1 (m + 1)(2m + 1), so S2m+1 = (m + 1)(2m + 1).
Also,
2m2 + 2m < 2m2 + 3m + 1 < 2m2 + 4m

(m + 1)2m < 2m2 + 3m + 1 < (m + 2)2m,
www.usamts.org
so 2m2 + 2m is the greatest multiple of 2m less than S2m+1 . Since S2m < S2m+1 = (m +
1)(2m + 1), we have S2m (m + 1)2m. But S2m (m + 1)2m, so S2m = (m + 1)2m. Then
x2m+1 = S2m+1 S2m = (m+1)(2m+1)(m+1)2m = m+1, and x2m+2 = S2m+2 S2m+1 =
(m + 1)(2m + 2) (m + 1)(2m + 1) = m + 1, which contradicts (1).
Hence, S2m+2 (m + 2)(2m + 2), so

(2m + 2) + 1
A2m+1 m+2= ,
2
2007+1
completing the induction. Therefore, the minimum value of A2007 is
2
= 1004.
www.usamts.org
1/1/19. Gene has 2n pieces of paper numbered 1 through 2n. He removes n pieces of paper
that are numbered consecutively. The sum of the numbers on the remaining pieces of paper
is 1615. Find all possible values of n
Comments The first step in the problem is to use algebra to find suitable bounds on n.
We can then use divisibility properties of integers to find the solutions. Solutions edited by
Naoki Sato.
Solution 1 by: Carlos Dominguez (11/OH)
The minimum sum of the numbers on the remaining pieces is 1 + 2 + + n = n(n+1)

2
,
n(n+1) 2
so 2
1615. Clearing the denominator and expanding gives n + n 3230. Since
56 + 56 = 3192 < 3230 and 572 + 57 = 3306 > 3230, we must have n 56.
2
The maximum sum of the numbers on the remaining pieces is (n+1)+(n+2)+ +2n =
n(3n+1)
2
,so n(3n+1)
2
1615, which implies 3n2 + n 3230. Since 3 322 + 32 = 3104 < 3230
and 3 332 + 33 = 3300 > 3230, we must have n 33.
Of the n numbers removed, let k be the first number. Then the sum of the n remaining
numbers is
(1 + 2 + + 2n) [k + (k + 1) + + (k + n 1)]
2n(2n + 1) (2k + n 1)n
=
2 2
= 1615.
Multiplying both sides by 2/n and expanding, we get

3230
4n + 2 (2k + n 1) = 3n 2k + 3 = .
n
Since 3n 2k + 3 is an integer, 3230/n is also an integer. In other words, n is a factor
of 3230. The factors of 3230 are 1, 2, 5, 10, 17, 19, 34, 38, 85, 95, 170, 190, 323, 646, 1615,
and 3230. The only factors between 33 and 56 (inclusive) are n = 34 and n = 38. The
corresponding values of k are 5 and 16, respectively, which are both viable, so the possible
values of n are 34 and 38.
www.usamts.org
2/1/19. A regular 18-gon is dissected into 18 pentagons, each of which is congruent to

pentagon ABCDE, as shown. All sides of the pentagon have the same length.
C Z
B Y
X
D
A E
(a) Determine angles A, B, C, D, and E.
(b) Show that points X, Y , and Z are collinear.
Comments Part (a) can be done by considering appropriate combinations of angles in the
regular 18-gon. Part (b) can be done by showing that XY Z = 180 . Solutions edited by
Naoki Sato.
Solution 1 by: Luyi Zhang (9/CT)
(a) At the center of the 18-gon, six pentagons join together by their angle that corresponds
to A. Therefore, A = 360 /6 = 60 . Since all sides of the pentagon are equal, triangle
ABE is equilateral and quadrilateral BCDE is a rhombus.
ABC is an interior angle of the 18-gon, so B = ABC = 160 . Then
EBC = ABC ABE = 160 60 = 100 ,
so D = CDE = EBC = 100 and
C = BED = 180 EBC = 180 100 = 80 .
Finally, E = AED = AEB + BED = 60 + 80 = 140 .

To summarize, A = 60 , B = 160 , C = 80 , D = 100 , and E = 140 .
www.usamts.org
B C
A
E D
C
B A
D A A
A A
A
A E
(b) To show that points X, Y , and Z are collinear we will show that XY Z = 180 .
Label points M , N , O, and P , as shown below.
O N
M
Z 80
60 100 60
P Y
X
Since all the sides are of equal length, we can easily create isosceles triangles to assist
in our angle search. In triangle M XY , M X = M Y and XM Y = 80 , so M XY =
M Y X = (180 80 )/2 = 50 .
In triangle P Y Z, P Y = P Z and ZP Y = ZP O + OP Y = 60 + 100 = 160 , so
P ZY = P Y Z = (180 160 )/2 = 10 .
Then in triangle OP Y , P O = P Y and OP Y = 100 , so P Y O = P OY = N Y O =
N OY = (180 100 )/2 = 40 , so ZY O = P Y O P Y Z = 40 10 = 30 . Then
XY Z = M Y X + M Y N + N Y O + ZY O = 50 + 60 + 40 + 30 = 180 ,
and we are done.

www.usamts.org
3/1/19. Find all positive integers a b c such that

1 1 1
arctan + arctan + arctan = .
a b c 4
Comments First, we can use the properties of the arctan function to establish bounds on
a. Then we can transform the given equation into an algebraic equation, from which we can
deduce
Create PDF with GO2PDF the
for free, if you solutions.
wish to remove this Solutions edited
line, click here to buy byPrinter
Virtual PDF Naoki Sato.
Solution 1 by: Damien Jiang (10/NC)
We first establish bounds on a. Since arctan x is increasing on (0, 1],

1 1 1
arctan arctan arctan .
a b c
Hence,
1 1 1 1
= arctan + arctan + arctan 3 arctan ,
4 a b c a
so
1 1 1
arctan tan =2 3 a = 2 + 3 < 4.
a 12 a 12 2 3
Additionally,
1 1 1 1
= arctan + arctan + arctan > arctan ,
4 a b c a
so
1
< tan = 1 a > 1.
a 4
Therefore, the only possible values of a are a = 2 and a = 3.
From the original equation, we subtract arctan 1c , and take the tangent of both sides to
get
1
a
+ 1b 1 1c
1 = .
1 ab 1 + 1c
Note that this equation is equivalent with the original because tan x is injective on (0, 1].
Multiplying, clearing denominators, and rearranging, we get
abc + 1 = ab + ac + bc + a + b + c.
www.usamts.org
If a = 2, then
2bc + 1 = 2(b + c) + bc + 2 + b + c
bc 3(b + c) = 1
(b 3)(c 3) = 10.
Because c > b, we have b = 4, c = 13 or b = 5, c = 8.

If a = 3, then
3bc + 1 = 3(b + c) + bc + 3 + b + c
2bc 4(b + c) = 2
(b 2)(c 2) = 5.
Because c > b, we have b = 3, c = 7.

Therefore, the only solutions are (a, b, c) = (2, 4, 13), (2, 5, 8), and (3, 3, 7).
www.usamts.org
4/1/19. In convex quadrilateral ABCD, AB = CD, ABC = 77 , and BCD = 150 .

Let P be the intersection of the perpendicular bisectors of BC and AD. Find BP C.
Comments Since P lies on the perpendicular bisector of BC, P B = P C. This and similar
observations lead to the construction of congruent triangles which determine BP C. In
addition, the solution below rigorously estalishes the location of point P . Solutions edited
by Naoki Sato.
Solution 1 by: Carl Lian (9/MA)
A
N
D
B
M C
Note that there are three distinct cases for the position of P : Either outside quadrilateral
ABCD on the side of BC, that is, P M < P N ; outside quadrilateral ABCD on the side of
AD, that is, P N < P M ; or inside quadrilateral ABCD. We first deal with the first case,
and then prove that the second and third cases are impossible.
Let M be the midpoint of BC and N the midpoint of AD. We have BM = M C and
AN = N D, and BM P = CM P = AN P = DN P = 90 , so 4BM P = 4CM P and
4AN P = 4DN P . From these congruences, BP = CP and AP = DP , and we are given
that AB = CD. Therefore, 4ABP = 4DCP , and ABP = DCP .
Let = CBP . Then DCP = ABP = 77 + , and BCP = . Now, by the angles
around C, we have DCB + BCP + P CD = 150 + + 77 + = 360 , so 2 = 133 .
Hence, BP C = 2BP M = 2(90 ) = 180 2 = 47 .
www.usamts.org
For the second case, assume by way of contradiction that P lies outside quadrilateral
ABCD, on the side of AD. Again, 4ABP = 4DCP . We have P BA = P CD, and
also P BM = P CM from 4P BM = 4P CM . Adding these gives P BA + P M B =
P CD + P CM , and thus ABC = BCD, but this is a contradiction because ABC =
77 , and BCD = 150 , so P cannot lie outside quadrilateral ABCD on the side of AD.
For the third case, assume by way of contradiction that P lies inside quadrilateral ABCD.
Again, 4ABP = 4DCP . We have P BA = P CD, and also P BM = P CM from
4P BM = 4P CM . Adding these gives P BA + P M B = P CD + P CM , and thus
ABC = BCD, but this is a contradiction because ABC = 77 and DBC = 150 , so
P cannot lie inside quadrilateral ABCD.
Therefore, the first case is the only possible case, and our assertion that BP C = 47
still holds.
www.usamts.org
5/1/19. Let c be a real number. The sequence a1 , a2 , a3 , . . . is defined by a1 = c and

an = 2a2n1 1 for all n 2. Find all values of c such that an < 0 for all n 1.
Comments It is not difficult to show that the value c = 12 works. If c 6= 12 , then the
terms of the sequence must diverge from 12 , to the point where they become positive. The
following solution uses a rigorous bounding argument. Solutions edited by Naoki Sato.

The only value is c = 21 .

If an = 12 , then an+1 = 2a2n 1 = 2( 21 )2 1 = 12 , so if c = 12 , then an = 12 < 0 for
all n 1 and the result is achieved.
Assume c 6= 21 , and define the sequence bn = 2an + 1. Assume that an < 0 for all n, so
bn < 1 for all n. A recursion for the bn is derived from that for the an :
2
bn 1 bn1 1
=2 1
2 2
bn 1 = (bn1 1)2 2
bn = bn1 (bn1 2)
bn = bn2 (2 bn2 )(2 bn1 )
for all n > 2. If bn = 0, then bn1 = 0 or bn1 = 2. However, by assumption, bn < 1 for all
n, and b1 = 2a1 + 1 = 2c + 1 6= 0, so bn 6= 0 for all n.
If bn < 0, then bn+1 = bn (bn 2) > 0. Likewise, if bn > 0, then bn+1 < 0 since bn < 1 and
so bn 2 < 1 < 0. Hence, the terms bn alternate in sign, so for all n, one of bn1 and bn2
is negative. The other is less than 1, so
bn
= (2 bn2 )(2 bn1 ) > (2 0)(2 1) = 2.
bn2
Let m = 1 if b1 is positive, and m = 2 if b2 is positive, so bm is positive. Take l sufficiedntly

large so that bm > 2l . Then
bm+2l > 2bm+2(l1) > 22 bm+2(l2) > > 2l1 bm+2 > 2l bm > 1,
a contradiction. So c = 12 is the only solution.

www.usamts.org
1/2/19. Find the smallest positive integer n such that every possible coloring of the integers
from 1 to n with each integer either red or blue has at least one arithmetic progression of
three different integers of the same color.
Comments Any solution to this problem will inevitably require some casework. However,
by choosing them carefully, the number of cases can be considerably reduced. Solutions
Solution by: Adrian Chan (12/CA)

We first prove that n 8 does not suffice. To do so, it is sufficient to give a counterex-
ample for n = 8: Let 1, 4, 5, and 8 be red, and let 2, 3, 6, and 7 be blue. We see that there
are no arithmetic sequences among the red numbers, or the blue numbers.
Now we show that for every coloring of the integers from 1 to 9, there is always an arith-
metic sequence of three different integers of the same color. For the sake of contradiction,
suppose that there is a coloring that does not produce any such arithmetic sequences. With-
out loss of generality, let 5 be blue. Then at least one of 1 and 9 must be red, otherwise 1,
5, and 9 will form a blue arithmetic sequence.
Case 1: 1 is blue and 9 is red, or 1 is red and 9 is blue.
First, assume that 1 is blue and 9 is red. Then 3 must be red, otherwise 1, 3, and 5 will
form a blue arithmetic sequence. Next, 6 must be blue, otherwise 3, 6, and 9 will form a red
arithmetic sequence. Next, both 4 and 7 must be red, otherwise 4, 5, and 6, or 5, 6, and 7
will form a blue arithmetic sequence. Finally, both 2 and 8 must be blue, otherwise 2, 3, and
4, or 7, 8, and 9 will form a red arithmetic sequence. However, we end up with 2, 5, and 8
forming a blue arithemtic sequence, contradiction. The case that 1 is red and 9 is blue can
be similarly proven, by reversing the order of the colors.
Case 2: Both 1 and 9 are red.
First, assume that 7 is red. Then both 4 and 8 must be blue, otherwise 1, 4, and 7, or 7,
8, and 9 will form a red arithmetic sequence. Next, both 3 and 6 must be red, otherwise 3,
4, and 5, or 4, 5, 6 will form a blue arithmetic sequence. However, we end up with 3, 6, and
9 forming a red arithmetic sequence, contradiction.
Now, asssume that 7 is blue. Then 3 must be red, otherwise 3, 5, and 7 will form a
blue arithmetic sequence. Next, 6 must be blue, otherwise 3, 6, and 9 will form a red
arithmetic sequence. However, we end up with 5, 6, and 7 forming a blue arithmetic sequence,
contradiction.
We conclude that n = 9 is the smallest possible integer that satisfies the desired condi-
tions.
www.usamts.org
2/2/19. Let x, y, and z be complex numbers such that x + y + z = x5 + y 5 + z 5 = 0 and

x3 + y 3 + z 3 = 3. Find all possible values of x2007 + y 2007 + z 2007 .
Comments There are different ways to appoach this problem. The solution below uses a
substitution to eliminate one of the variables, and determines the values of x3 , y 3 , and z 3
directly. Solutions edited by Naoki Sato.
Solution by: Kristin Cordwell (11/NM)

First, x+y +z = 0, so x+y = z. We then cube both sides to get x3 +3x2 y +3y 2 x+y 3 =
z . We rearrange the equation to get x + y 3 + z 3 = 3x2 y 3y 2 x. We know that
3
3
x3 + y 3 + z 3 = 3, so we get 3xy(x + y) = 3, or xy(x + y) = 1. This also tells us that

neither xy nor x + y can be equal to 0.
Now we take the fifth power of x + y = z to get
x5 + 5x4 y + 10x3 y 2 + 10x2 y 3 + 5xy 4 + y 5 = z 5 .
Rearranging the equation gives us
x5 + y 5 + z 5 = 5xy(x3 + 2x2 y + 2xy 2 + y 3 ).
We know that x5 + y 5 + z 5 = 0, so we get 5xy(x3 + 2x2 y + 2xy 2 + y 3 ) = 0, or
xy(x3 + 2x2 y + 2xy 2 + y 3 ) = 0.
Also, we know that xy 6= 0, so we get
x3 + 2x2 y + 2xy 2 + y 3 = x3 + y 3 + 2xy(x + y) = 0.
We know that xy(x + y) = 1, so this simplifies as x3 + y 3 = 2. Finally, we know that

x3 + y 3 + z 3 = 3, so z 3 must equal 1. By symmetry, x3 = y 3 = 1. Since 2007 is divisible by
3, x2007 + y 2007 + z 2007 = 3.

To show that this value is possible, let x = 1, y = 21 + 23 i and z = 21 23 i, which
are the cube roots of unity. Then x + y + z = x5 + y 5 + z 5 = 0, and x3 + y 3 + z 3 =
x2007 + y 2007 + z 2007 = 3.
www.usamts.org
3/2/19. A triangular array of positive integers is called remarkable if all of its entries are
distinct, and each entry, other than those in the top row, is the quotient of the two numbers
immediately above it. For example, the following triangular array is remarkable:
7 42 14
6 3
2
Find the smallest positive integer that can occur as the greatest element in a remarkable
array with four numbers in the top row.
Comments It is fairly easy to find an example where the greatest number in the array is
120. Then, one can prove that 120 is the mininum by showing that some number in the top
row is the product of four different integers, all at least 2. Solutions edited by Naoki Sato.
Solution by: Dmitri Gekhtman (11/IN)

A remarkable array cannot contain a 1, because this would mean that it contains at least
two equal numbers. Denote the integer in the bottom row by a1 . Then the second row from
the bottom contains integers a2 and a1 a2 , with a2 6= a1 . In the third row from the bottom,
and above a1 a2 , there are integers a3 and a1 a2 a3 (in either order), with a3 different from a1
and a2 . Finally, in the top row, and above a1 a2 a3 , there are integers a4 and a1 a2 a3 a4 (in
either order), with a4 different from a1 , a2 , and a3 .
This means that the greatest number in the top row is at least a1 a2 a3 a4 . Since a1 a2 a3 a4
is a product of four different integers, all at least 2, it is greater or equal than 2 3 4 5 = 120.
Therefore, the answer to the problem is at least 120. An example with the greatest element
in the array equal to 120 is as follows:
40 5 120 30
8 24 4
3 6
2
Hence, the answer is 120.

www.usamts.org
4/2/19. Two nonoverlapping arcs of a circle are chosen. Eight distinct points are then
chosen on each arc. All 64 segments connecting a chosen point on one arc to a chosen point
on the other arc are drawn. How many triangles are formed that have at least one of the 16
points as a vertex?
A sample figure is shown below:
Comments We can count the number of triangles by counting the number of ways to choose
points on the circle that give rise to these triangles. However, for each case, we must be
careful to factor in just how many triangles are created by the points we choose. Solutions
Solution by: Santhosh Karnik (11/GA)

To generalize, suppose m points are chosen on one arc and n points are chosen on the
other arc. Let the arc with m points be arc A and the arc with n points be arc B. Since all
triangles formed must have at least one vertex on the circle, a triangle must have either 1,
2, or 3 vertices on the circle.
If all 3 vertices of a triangle are on the circle, then at least 2 of the 3 vertices must be on
the same arc. But none of the segments connect 2 points on the same arc. Thus there are
no triangles formed with all 3 vertices on the circle.
If 2 vertices of a triangle are on the circle, then one must be on the arc A, and the
other must be on arc B. The third vertex can be anywhere inside the circle as long as it is
connected to the other two vertices. Since all segments connect a point on arc A to a point
on arc B, the two segments of the triangle that are connected to the third vertex can each be
extended to a point on the circle. This gives a total of 4 points, 2 on each arc. Connecting
all possible segments on these 4 points yields 2 triangles with 2 vertices on the circle. Thus,
selecting 2 points on arc A and 2 points on arc B always yields 2 triangles with exactly two
vertices on the circle, and all such triangles can be formed by choosing 2 points on each arc
and connecting all possible segments.
www.usamts.org
A A A
A
B
B B B
Therefore, there are 2 m2 n2 triangles formed such that 2 vertices of a triangle are on the

circle.
If a triangle has only 1 vertex on the circle, then the other 2 vertices must be inside the
circle and connected to each other and the first vertex. By extending all the segments of
the triangle to points on the circle, 4 additional points are obtained, 1 on the same arc as
the first vertex, and 3 on the other arc. Connecting all possible segments on these 5 points
yields 2 triangles with exactly one vertex on the circle. Thus, selecting 2 points on arc A
and 3 points on arc B or vice-versa always yields 2 triangles with 1 vertex on the circle, and
all such triangles can be formed by choosing 2 points on one arc and 3 points on the other
arc and connecting all possible segments.
A A A A
A B
OR
B A
B B
B B
Therefore there are 2 m2 n3 + 2 m3 n2 triangles formed such that 1 vertex of the triangle

is on the circle. Therefore, the total number of triangles formed with at least one vertex on
the circle is
m n m n m n
T (m, n) = 2 + + .
2 2 2 3 3 2
In the problem, there are m = n = 8 points on each arc. So the total number of triangles
formed with at least one vertex on the circle is

8 8 8 8 8 8
T (8, 8) = 2 + + = 7840.
2 2 2 3 3 2
www.usamts.org
5/2/19. Faces ABC and XY Z of a regular icosahedron are parallel, with the vertices
labeled such that AX, BY , and CZ are concurrent. Let S be the solid with faces ABC,
AY Z, BXZ, CXY , XBC, Y AC, ZAB, and XY Z. If AB = 6, what is the volume of S?
Comments The volume of S can be found by relating it to other polyhedra whose volumes
are known. In the solution below, the polyhedron S is related to a hexagonal prism. Solutions
Solution by: Luyi Zhang (9/CT)

First, we claim that in a regular pentagon, the ratio of the lengths of a diagonal to a side

is 1+2 5 : 1. Let ABCDE be a regular pentagon. Let F be the intersection of AC and BE,
and let G be the intersection of AC and BD.
B
36
x+y 36
y y
36 108 72 72
A y x C
F G
E D
Let x = F G and y = BG. Triangle BF G is isosceles, so BF = BG = y. Triangle ABF

is isosceles, so AF = BF = y. Then AG = x + y, and since triangle ABG is isosceles,
AB = x + y.
By AAA, triangles ABG and BF G are similar, so
AB BF x+y y
= = ,
BG FG y x
which simplifies as y 2 xy x2 = 0. By the quadratic formula,

x 5x2 1 5
y= = x.
2 2
www.usamts.org

1 5
Since is negative,
2
1+ 5
y= x.
2
By AAA, triangles ABC and AF B are similar, so the ratio of the lengths of a diagonal
to a side is
AC AB x+y y 1+ 5
= = = = ,
AB AF y x 2
as desired.
Now, the polyhedron S has eight faces. Two of them are equilateral triangles with the
same side length as the icosahedron, which is 6. The other six are isosceles triangles. In
these six faces, the base is the same as the side length of the icosahedron, which
is 6, and
the legs are the diagonals of a regular pentagon with side length 6, which is 3(1 + 5).
A
C
Z
X
Let A0 , B 0 , and C 0 be the projections of A, B, and C onto the plane of triangle XY Z, and
let X 0 , Y 0 , and Z 0 be the projections of X, Y , and Z onto the plane of triangle ABC. Then
AY 0 CX 0 BZ 0 A0 Y C 0 XB 0 Z is a regular hexagonal prism. (This follows from the symmetry of
the icosahedron.)
www.usamts.org
B
Z0
X0
A
C
Y0
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer B0
Z
X
A0
C0
Y
We see that this hexagonal prism is the union of polyhedron S (whose edges are in
red) and the six tetrahedra AA0 Y Z, BB 0 XZ, CC 0 XY , XX 0 BC, Y Y 0 AC, and ZZ 0 AB, all
of which are congruent. Thus, we can find the volume of S by finding the volume of the
hexagonal prism, and then subtracting the volume of the six tetrahedra.
0 0 0
Let b denote the area of
regular hexagon AY CX BZ . Since AB = 6, the side length of
0 0 0
hexagon AY CX BZ is 2 3. Hence, the hexagon is composed of six equilateral triangles of
side length 2 3, so
3 2
b=6 (2 3) = 18 3.
4
0 2
We can then use Pythagoras to findthe height h of the
prism. Since AA Y = 90 , AY =
0 2 0 2 0
(AA ) + (A Y ) . But AY = 3(1 + 5) and A Y = 2 3, so
(AA0 )2 = AY 2 (A0 Y )2

= [3(1 + 5)]2 (2 3)2

= 42 + 18 5.
p
Therefore, h = AA0 = 42 + 18 5 = 3(3 + 5).
The volume of the hexagonal prism is bh. Since triangle A0 Y Z has 1
6
th the area of
hexagon AY 0 CX 0 BZ 0 , the volume of tetrahedron AA0 Y Z is
1 1 1
b h = bh.
3 6 18
www.usamts.org
All six tetrahedra AA0 Y Z, BB 0 XZ, CC 0 XY , XX 0 BC, Y Y 0 AC, ZZ 0 AB have the same
volume, so the volume of S is
6 1 2 2
bh bh = bh bh = bh = 18 3 3(3 + 5) = 108 + 36 5.
18 3 3 3
www.usamts.org
1/3/19. We construct a sculpture consisting of infinitely many cubes, as follows. Start

with a cube with side length 1. Then, at the center of each face, attach a cube with side
length 13 (so that the center of a face of each attached cube is the center of a face of the
original cube). Continue this procedure indefinitely: at the center of each exposed face of a
cube in the structure, attach (in the same fashion) a smaller cube with side length one-third
that of the exposed face. What is the volume of the entire sculpture?
Comments Once the geometry of the sculpture has been determined, the volume can be
found by summing an infinite geometric sequence. Solutions edited by Naoki Sato.
Since the cube with side length 1 has 6 faces, the sculpture has 6 cubes of side length 13
and volume ( 31 )3 = 27
1
. Since each of the 6 cubes of side length 13 has 5 exposed faces, there
1 2
are 6 5 cubes of side length ( 31 )2 and volume ( 27 ) . By the same reasoning, there are 6 52
1 n 1 n
cubes of volume ( 27 ) . In general, the sculpture contains 6 5n1 cubes of volume ( 27 ) ,
where n is a positive integer. The sculpture contains one cube of volume 1. Therefore, the
total volume of the sculpture is
n n
X
n1 1 X 6 5
1+ 65 =1+ .
n=1
27 n=0
27 27
6 5
Since the infinite sum is an infinite geometric series with first term 27
and common ratio 27
,
the volume is
6
14
1 + 27 5 = .
1 27 11
www.usamts.org
2/3/19. Gene starts with the 3 3 grid of 0s shown at left below. He then repeatedly
chooses a 2 2 square within the grid and increases all four numbers in the chosen 2 2
square by 1. One possibility for Genes first three steps is shown below.
0 0 0 0 0 0 0 1 1 0 1 1
0 0 0 1 1 0 1 2 1 2 3 1
0 0 0 1 1 0 1 1 0 2 2 0
How many different grids can be produced with this method such that each box contains
an integer from 1 to 12, inclusive? (The numbers in the boxes need not be distinct.)
Comments The number of different grids can be counted by identifying each grid with a
quintuple of positive integers. Then the number of such quintuples can be found using a
standard partition argument. Solutions edited by Naoki Sato.
Solution by: Matt Superdock (11/PA)

Any 2 2 square inside the 3 3 grid includes the center box, so the integer in the center
box is equal to the number of steps Gene makes. Therefore, Gene can make at most 12 steps,
or else the middle box will contain an integer greater than 12. Additionally, each corner box
of the 3 3 grid is included in only one of the four 2 2 squares. Therefore, Gene must
choose each 2 2 square at least once, or else one of the corner boxes will contain a 0. The
final grid depends only on how many times Gene chooses each 2 2 square, not which order
he chooses them.
Let a, b, c, and d be the numbers of times Gene chooses each 2 2 square. Then a, b,
c, and d are positive integers, and a + b + c + d 12. Let e = 13 a b c d, so that
e is also a positive integer, and a + b + c + d + e = 13. To find the number of solutions to
this equation, we consider partitioning 13 objects into 5 non-empty groups. We arrange the
13 objects in a row, and we partition them by placing dividers in 4 of the 12 spaces between
12

adjacent objects. There are 4 = 495 ways to do this, so there are 495 grids that can be
produced.
www.usamts.org
3/3/19. Consider all polynomials f (x) with integer coefficients such that f (200) = f (7) =
2007 and 0 < f (0) < 2007. Show that the value of f (0) does not depend on the choice of
polynomial, and find f (0).
Comments This problem may be solved by using the following crucial result: If p(x) is a
polynomial with integer coefficients, then for any integers a and b, p(a) p(b) is a multiple
of a b. In particular, for the case b = 0, p(a) p(0) is a multiple of a. Solutions edited by
Naoki Sato.
Solution by: Andy Zhu (11/NJ)

Since all polynomials P (x) with integer coefficients can be expressed in the form P (x) =
x Q(x) + P (0), where Q(x) is a polynomial with integer coefficients, P (n) P (0) (mod n)
for all positive integers n. Thus f (0) f (7) 2007 (mod 7) and f (0) f (200) 2007
(mod 200).
Since gcd{7, 200} = 1, we can apply the Chinese Remainder Theorem to get f (0)
2007 607 (mod 1400). The unique value which f (0) takes in the range 0 < f (0) < 2007
is 607.
www.usamts.org
4/3/19. Prove that 101 divides infinitely many of the numbers in the set
{2007, 20072007, 200720072007, 2007200720072007, . . .}.
Comments There are several possible approaches, but the following solution elegantly
reduces the elements of the set modulo 101 to a simple formula. Solutions edited by Naoki
Sato.
Solution by: Daniel Tsai (12/NJ)

Modulo 101, we have 100 1, 101 10, 102 100, 103 91, and 104 1. Therefore,
modulo 101, for each integer k 0, 104k 1, 104k+1 10, 104k+2 100, and 104k+3 91.
Consequently, for each positive integer n,
20072007 2007 (2 91 + 0 100 + 0 10 + 7 1)n 189n 88n (mod 101),
where the left-hand-side consists of n 2007s. For each positive multiple n of 101, 88n 0
(mod 101), thus 101 divides infinitely many numbers in the set
{2007, 20072007, 200720072007, 2007200720072007, . . . }.

www.usamts.org
p
5/3/19. For every rational number 0 < < 1, where p and q are relatively prime,
q
p 1 1
construct a circle with center , 2 and diameter 2 . Also construct circles centered at
q 2q q
1 1
0, and 1, with diameter 1.
2 2
(a) Prove that any two such circles intersect in at most 1 point.

X 1

i=1
i3

(b) Prove that the total area of all of the circles is 1 +
.
X 1

i=1
i4
Comments Part (a) may be solved by comparing the distance between the centers of two
circles to the sum of the radii. Part (b) may be solved by expressing the same sum two
different
P ways, oneP
involving the circles in the problem, and the other involving the infinite
sums 1
i=1 i3 and 1
i=1 i4 . Solutions edited by Naoki Sato.

(a) Let r and s be two nonnegative integers such that 0 r s and either r and s are
relatively prime or s = 1. For all possible values of r and s, construct a circle with center
( rs , 2s12 ) and diameter s12 . The resulting set of circles is the same as the one described in the
problem. (The parameters r = 0, s = 1 and r = 1, s = 1 specify the circles centered at (0, 12 )
and (1, 12 ) with diameter 1, respectively.)
Choose two distinct circles from the set of constructed circles. Let one have center ( sr11 , 2s12 )
1
and diameter s12 and the other have center ( sr22 , 2s12 ) and diameter s12 . Since, for each circle,
1 2 2
the radius is equal to the y-coordinate of the center, both circles are tangent to the x-axis
and lie above it. Therefore, one circle cannot lie in the interior of the other circle. The
x-coordinate of the center of each circle is the same as the x-coordinate of the intersection of
the circle and the x-axis. Since the circles are tangent to the the x-axis at different points,
they obviously cannot be internally tangent. Therefore, either (1) the two circles are external
to each other and do not intersect at any point, (2) the two circles are externally tangent to
each other and intersect at exactly one point, or (3) the two circles intersect at exactly two
points.
Let d be the distance between the centers of the two circles, and let R be the sum of the
their radii. In case (1), d > R. In case (2), d = R. In case (3), d < R. The distance between
the centers of the two circles is
s 2 2
r1 r2 1 1
d= + .
s1 s2 2s21 2s22
www.usamts.org
The sum of the radii of the two circles is

1 1
R= 2
+ 2.
2s1 2s2
So
2 " 2 2 #
r1 r2 1 1 1 1 1
d2 R2 = + 2
2 2
+ 2
s1 s2 4 s1 s2 s1 s2
r12 2r1 r2 r22 1
= 2
+ 2 2 2
s1to buy Virtual
Create PDF with GO2PDF for free, if you wish to remove this line, click here s2 s1 s2
s1 s2PDF Printer
r s 2r1 s2 r2 s1 + r22 s21 1
2 2
= 1 2
s21 s22
(r1 s2 r2 s1 )2 1
= .
s21 s22
Since r1 , r2 , s1 , and s2 are integers, r1 s2 r2 s1 is an integer. Suppose that r1 s2 r2 s1 = 0.

Then
r1 r2
= .
s1 s2
r1 r2
But s1 and s2 are distinct rational numbers, so r1 s2 r2 s1 6= 0. Since r1 s2 r2 s1 is a nonzero
integer, (r1 s2 r2 s1 )2 1, so
(r1 s2 r2 s1 )2 1
d2 R2 = 0,
s21 s22
which means d2 R2 . Since d and R are positive, d R. Therefore, the circles cannot
intersect at two points. Furthermore, d can equal R (for example, when r1 = 0, s1 = 1,
r2 = 1, and s2 = 1, d = R = 1). So any two circles can intersect in at most 1 point. This
completes the proof.
(b) Consider all pairs of nonnegative integers a and b such that 0 a b and b 6= 0.
The sum over all such pairs (a, b) of b14 is
X b
X 1 X b+1
4
= .
b=1 a=0
b b=1
b4
Note that each pair (a, b) can be uniquely written in the form (nr, ns), where n is a positive
integer and (r, s) is a pair of integers of the type described in part (a). Thus, we may write
1
the sum over all such pairs (a, b) of b14 as the sum over all triples (r, s, n) of (ns) 4 . In other
words (note that all series below are absolutely convergent),

X b+1 XX 1 X 1 X 1
= = .
b4 (ns)4 s4 n=1 n4
b=1 (r,s) n=1 (r,s)
www.usamts.org
Therefore,

X b+1 X 1 X1 X 1
+
X 1 b=1
b4 i=1
i4 i=1 i3 i=1
i3
= = =1+ .
s4 X 1 X 1 X 1
(r,s)
n=1
n4 i=1
i4 i=1
i4
Multiplying both sides of this equation by 4 , we get

X 1
X i=1
i3

= 1 +
.
4s4 4 X 1

(r,s)
i=1
i4
The left side of the equation is the sum over all such pairs (r, s) of the area of a circle of
diameter s12 . Hence, the total area of all of the circles is

X 1

i=1
i3
1 + .

4
X1
i 4
i=1
This completes the proof.

www.usamts.org
1/4/19. In the diagram at right, each vertex is labeled with a different

positive factor of 2008, such that if two vertices are connected by an
edge, then the label of one vertex divides the label of the other vertex.
In how many different ways can the vertices be labeled? Two labelings
are considered the same if one labeling can be obtained by rotating
and/or reflecting the other labeling.
Comments We can solve the problem by using casework. The number

of cases can be reduced by using the symmetry in the diagram. Starting with the factors
8 and
Create PDF with GO2PDF 251
for free, if youis also
wish a this
to remove good idea,
line, click because
here to buy they are the least connectable factors. Solutions
Virtual PDF Printer
Solution by: Sam Elder (12/CO)

The positive factors of 2008 are 1, 2, 4, 8, 251, 502, 1004, and 2008. Six pairs of these
numbers do not satisfy the given property (one divides the other): (2, 251), (4, 251), (4, 502),
(8, 251), (8, 502), and (8, 1004). These pairs cannot be connected by an edge.
Denote the vertices as N, NW, W, SW, S, SE, E, and NE as directions on a compass.
N
NW NE
W E
SW SE
S
Rotate any working grid so 251 is at S. Each vertex in this diagram is connected by edges
to four others, and not connected to the other three. Since 251 cannot be connected to 2, 4,
or 8, these three numbers must be in the northernmost three places in the diagram, N, NE
and NW. Consider the different positions 8 can be in.
Case 1: 8 is at N. The only two factors that can lie at W and E are 1 and 2008. Reflect
any diagrams across the north-south axis so 1 is at W and 2008 is at E.
8
2008
1 2008
251
2 and 4 occupy NE and NW, so 502 and 1004 occupy SE and SW. The only other
forbidden pair is (4, 502), so if 4 is at NE then 502 is at SW, and if 4 is at NW then 502 is
at SE. These are the two solutions for this case:
www.usamts.org
8 8
4 2 2 4
1 2008 1 2008
1004 502 502 1004

251 251
Case 2: 8 is not at N. Reflect across the north-south axis so 8 is at NE. Then 2 and 4
are at N and NW, while 502 and 1004 are at W and SW to avoid edges with 8. The only
way 4 and 502 cannot share an edge is if 4 is at N and 502 at SW. Then 2 is at NW and
1004 at W, and the only factors left to arrange are 1 and 2008, which can go either at E or
SE:
4 4
2 8 2 8
1004 1 1004 2008
502 2008 502 1

251 251
With two possible labelings in each case, there are 4 possible labelings in all.
www.usamts.org
2/4/19. Determine, with proof, the greatest integer n such that

jnk jnk j n k j n k
+ + + < n,
2 3 11 13
where bxc is the greatest integer less than or equal to x.
Credit This problem was proposed by Andy Niedermaier.
Comments In an intuitive (but not rigorous) sense, n should leave the maximum remainder
when divided by 2, 3, 11, and 13, i.e. n should leave a remainder of 12 when divided by
13, and so on. The following proof rigorously establishes the answer by finding a bound in
terms of these remainders. Solutions edited by Naoki Sato.
Solution by: Wenyu Cao (9/NJ)

We claim that the greatest integer that satisfies the given inequality is 1715. First, we
check that n = 1715 satisfies the given inequality:

1715 1715 1715 1715
+ + + = 857 + 571 + 155 + 131 = 1714 < 1715.
2 3 11 13
Next, we claim that

xk+1 jxk

k k
for all positive integers x and k. By the Division Algorithm, there exist integers q and r
such that x = qk + r and 0 r k 1. Then
j x k qk + r j rk
= = q+ = q,
k k k
since 0 r/k < 1, and
xk+1 qk + r k + 1 r (k 1) jxk
= =q+ q= ,
k k k k
as desired.
Now, let n be a positive integer that satisfies the given inequality:
jnk jnk j n k j n k
+ + + < n.
2 3 11 13
Since both sides of the inequality are integers,
jnk jnk j n k j n k
+ + + n 1.
2 3 11 13
www.usamts.org
Therefore, from the result above,

jnk jnk j n k j n k
n1 + + +
2 3 11 13
n 1 n 2 n 10 n 12
+ + +
2 3 11 13
859 2573
= n .
858 858
Then
n 1715
,
858 858
so n 1715. Since we have shown that n = 1715 works, we are done.
www.usamts.org
3/4/19. Let 0 < < 1. Define a sequence {an } of real numbers by a1 = 1 and for all
integers k 1,
a2k = ak ,
a2k+1 = (1 )ak .

X
Find the value of the sum a2k a2k+1 in terms of .
k=1
Credit This problem was proposed by Sandor Lehoczky, and modified by Dave Patrick.
Comments This following solution deftly finds the required sum by directly using the given
recursion relations. Solutions edited by Naoki Sato.
Solution by: Tony Jin (10/CA)

By the definition of {an },

X
X
X
a2k a2k+1 = [ak (1 )ak ] = (1 ) a2k .
k=1 k=1 k=1
P
We can split up the sum k=1 a2k as follows:

X
X
X
a2k = a21 + a22k + a22k+1
k=1 k=1 k=1

X X
= a21 + 2 a2k + (1 )2 a2k
k=1 k=1

X
X
= a21 + 2
a2k + (1 ) 2
a2k
k=1 k=1

X
= a21 + [2 + (1 )2 ] a2k .
k=1
Therefore,

X
[1 2 (1 )2 ] a2k = a21 ,
k=1
so
X a21 1
a2k = 2 2
= .
k=1
1 (1 ) 2(1 )
www.usamts.org
Finally, the sum we seek is

X X 1
a2k a2k+1 = (1 ) a2k = .
k=1 k=1
2
www.usamts.org
4/4/19. Suppose that w, x, y, z are positive real numbers such that w + x < y + z. Prove
that it is impossible to simultaneously satisfy both
(w + x)yz < wx(y + z) and (w + x)(y + z) < wx + yz.
Comments Since we want to show that not all three inequalities can hold simultaneously,
we can approach the problem by using contradiction. Solutions edited by Naoki Sato.
Solution 1 by: Andy Zhu (11/NJ)

For the sake of contradiction, suppose that all the given inequalities hold. Multiplying
the inequalities wx(y + z) > (w + x)yz and wx + yz > (w + x)(y + z), we get
wx(y + z)(wx + yz) > yz(w + x)2 (y + z).
By the AM-GM inequality, (w + x)2 4wx, so
wx(y + z)(wx + yz) > yz(w + x)2 (y + z) 4wxyz(y + z).
Dividing by wx(y + z) (which is positive), we get
wx + yz > 4yz,
so wx > 3yz.
Also, since y + z > w + x and wx + yz > (w + x)(y + z),
wx + yz > (w + x)(y + z) > (w + x)2 4wx,
so yz > 3wx. Multiplying the inequalities wx > 3yz and yz > 3wx, we get wxyz > 9wxyz,
contradiction. Thus, not all the given inequalities can hold simultaneously.
www.usamts.org

We argue by contradiction. Suppose that the positive real numbers w, x, y, z satisfy all
the given inequalities, so w + x < y + z,
(w + x)yz < wx(y + z) wxy + wxz wyz xyz > 0,
and
(w + x)(y + z) < wx + yz wx + yz wy xy wz xz > 0.
Now consider the polynomial p(s) = (s w)(s x)(s + y)(s + z). Expanding this, we
have
p(s) = s4 + (y + z w x)s3 + (wx + yz wy xy wz xz)s2
+ (wxy + wxz wyz xyz)s + wxyz.
The coefficients of s3 , s2 , and s are all positive, and wxyz > 0 because w, x, y, z > 0.
Therefore, p(s) > 0 for all s > 0. However, p(w) = 0 and w > 0, contradiction. Therefore,
all three inequalities cannot simultaneously hold.
www.usamts.org
5/4/19. Let P1 P2 P3 P13 be a regular 13-gon. For 1 i 6, let di = P1 Pi+1 . The 13

diagonals of length d6 enclose a smaller regular 13-gon, whose side length we denote by s.
Express s in the form
s = c1 d1 + c2 d2 + c3 d3 + c4 d4 + c5 d5 + c6 d6 ,
where c1 , c2 , c3 , c4 , c5 , and c6 are integers.
Comments In geometry, when trying to find relationships between lengths, it is often

useful to find line segments that add up to the lengths in question. The following solution
constructs these line segments by using the symmetry of the regular 13-gon. Solutions edited
by Naoki Sato.
Solution by: Rui Jin (11/CA)

P4
P5
P3
P6
P2
P7
AB P1
P8
C
P13
P9 D
P12
P10
P11
Let A be the intersection of P2 P8 and P3 P9 , B the intersection of P1 P7 and P2 P8 , C the

intersection of P1 P9 and P2 P10 , and D the intersection of P1 P11 and P10 P13 . Since P1 P7 ,
P2 P8 , and P3 P9 are all diagonals of length d6 , AB is a side of the smaller regular 13-gon, so
s = AB.
By symmetry, BP2 = AP8 . Let x = BP2 = AP8 . Since P2 P8 = P1 P7 = d6 , AP2 = d6 x.
Diagonals P3 P9 and P2 P10 are parallel, and diagonals P2 P8 and P1 P9 are parallel, so
quadrilateral P2 AP9 C is a parallelogram. Hence, AP2 = d6 x = P9 C. Since P1 P9 =
P1 P6 = d5 , P9 C = d6 x = d5 P1 C.
Since P2 P10 and P1 P11 are parallel and P1 P9 and P13 P10 are parallel, quadrilateral
P1 CP10 D is a parallelogram. Hence, P1 C = DP10 and we can rewrite the equation above as
d6 x = d5 DP10 . Since P10 P13 = P1 P4 = d3 , d6 x = d5 (d3 DP13 ).
www.usamts.org
Finally, since P1 P11 and P13 P12 are parallel and P13 P10 and P12 P11 are parallel, quadrilat-
eral P13 DP11 P12 is a parallelogram. Hence, DP13 = P11 P12 and we can rewrite the equation
above as d6 x = d5 (d3 P11 P12 ). Since P11 P12 = P1 P2 = d1 , d6 x = d5 (d3 d1 ).
Rearranging the last equation, we find that x = d6 d5 + d3 d1 . Since s = d6 2x, we
have
s = d6 2(d6 d5 + d3 d1 ) = 2d1 2d3 + 2d5 d6 .
Round 1 Solutions
www.usamts.org
1/1/20. 27 unit cubes25 of which are colored black and 2 of

which are colored whiteare assembled to form a 3 3 3
cube. How many distinguishable cubes can be formed? (Two
cubes are indistinguishable if one of them can be rotated to
appear identical to the other. An example of two indistin-
guishable cubes is shown at right.)
If one of the two white cubes is the central cube (that is, the cube that is not visible from
the outside), then there are only 3 distinguishable places for the other white cube to be: a
corner, an edge, or a face-center, as shown left-to-right below:
Otherwise, we have the following possibilities:
Two corners: 3 waysthe corners lie on opposite ends of an edge, a face diagonal, or
an interior diagonal, as shown below:
One corner, one edge: 4 waysthe white cubes are adjacent, or give 2 non-adjacent
white squares on the same face (in two different distinguishable ways), or do not give
white squares on the same face:
One corner, one face-center: 2 wayseither the face-center shares a face with the
corner white cube or it does not:
Round 1 Solutions
www.usamts.org
Two edges: 5 waysthere can be white squares on the same face in two ways (either
diagonally adjacent or opposite, as in the first two pictures below), or they can be on
different faces in three distinguishable ways:
One edge, one face-center: 3 waysthe face-center may be on the same face as one of
the white edge cubes white squares, or the face-center is on a face adjacent to both
of the white edge cubes white squares, or the face-center is on a face adjacent to only
one of the white edge cubes white squares (shown left-to-right below):
Two face-centers: 2 waysthe face-centers are either on adjacent or opposite faces:
(The last diagram has another white face-center on the face directly opposite the face
with the visible white face-center.)
This gives a total of 3 + 3 + 4 + 2 + 5 + 3 + 2 = 22 distinguishable cubes.

Round 1 Solutions
www.usamts.org
2/1/20. Find all positive integers n for which it is possible to find three positive factors x, y,
and z of n 1, with x > y > z, such that x + y + z = n.
Let a = (n 1)/x, b = (n 1)/y, and c = (n 1)/z, so

n1 n1 n1
n=x+y+z = + + ,
a b c
where a, b, and c are all positive integers, and a < b < c.
We start by finding bounds on a. If a = 1, then x = n 1, so y + z = 1. But y and z are
positive integers, so y + z 2, giving a contradiction. Therefore, a 2.
n1 n1
On the other hand, if a 3, then x 3
, and similarly z < y < x 3
. So
n1 n1 n1
n=x+y+z < + + = n 1,
3 3 3
giving a contradiction.
Therefore, a = 2. This gives x = (n 1)/2 and thus
n1 n+1
y+z =nx=n = .
2 2
But recall y = (n 1)/b and z = (n 1)/c, so we have
n+1 n1 n1
= + . ()
2 b c
Now we look for bounds on b. We already know b > a = 2. If b 4, then c > b 4, and
by () we have
n+1 n1 n1 n1 n1 n1
= + < + = ,
2 b c 4 4 2
giving a contradiction. So b < 4, and combining this with b > 2, we must have b = 3.
This means that y = (n 1)/3, and by () we get
n1 n+1 n1 n+5
= = .
c 2 3 6
If c 6, then
n+5 n1 n1
= ,
6 c 6
giving a contradiction. Therefore, c = 4 or c = 5.
If c = 4, then
n+5 n1
= ,
6 4
Round 1 Solutions
www.usamts.org
and we get n = 13. We see that 13 is breakable, because 13 = 6 + 4 + 3.

If c = 5, then
n+5 n1
= ,
6 5
and we get n = 31. We see that 31 is breakable, because 31 = 15 + 10 + 6.
So the possible values of n are 13 and 31 .
Round 1 Solutions
www.usamts.org
3/1/20. Let a, b, c be real numbers. Suppose that for all real numbers x such that |x| 1, we
have |ax2 + bx + c| 100. Determine the maximum possible value of |a| + |b| + |c|.
Without loss of generality, we may assume that a 0. (If a < 0, then multiplying the
quadratic by 1 doesnt affect the bound nor change the value of |a| + |b| + |c|.) Also,
replacing x by x gives the function ax2 bx + c and does not change the bound nor the
value of |a| + |b| + |c|. So we may also, without loss of generality, assume that b 0.
Let f (x) = ax2 + bx + c. Since f (0) = c and f (1) = a + b + c, we must have |c| 100
and a + b + c 100. If c 0, then |a| + |b| + |c| = a + b + c 100. However, if c < 0, then
|a| + |b| + |c| = a + b c = (a + b + c) 2c 100 + 200 = 300,
so 300 is an upper bound for the value of |a| + |b| + |c|.

This bound can be achieved by the function f (x) = 200x2 100; note that the graph of
this is a parabola passing through the points (1, 100), (0, 100), and (1, 100).
The answer is 300 .
Round 1 Solutions
www.usamts.org

4/1/20. A pointP inside a regular tetrahedron ABCD is such that P A = P B = 11 and
P C = P D = 17. What is the side length of ABCD?
Solution 1 : Let X be the midpoint of side AB and let Y B

be the midpoint of side CD. Note that the locus of points
P inside of ABCD such that P A = P B and P C = P D is C
the line segment XY (minus its endpoints).
P Y
Let 2s be the side length of ABCD, so that AX = CY =
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer X x y
s. Note that XY isa leg of the right triangle AXY with
hypotenuse D
AY = s 3 and other leg AX = s, so we have
XY = s 2. If we let x = P X and y = P Y , then we have
the following system of equations (using the Pythagorean A
Theorem on right triangles P XA and P Y C):
x2 + s2 = 11,
y 2 + s2 = 17,

x + y = s 2.

Astute
solvers might see by inspection that the solution to
this system is s = 3, x = 2,
y = 2 2, but we can also solve it algebraically. Using y = s 2 x from the third equation,
substitute into the second equation:

17 = (s 2 x)2 + s2

= 2s2 2 2sx + x2 + s2

= 2s2 2 2sx + 11,
where the last step above uses x2 + s2 = 11 from our earlier system of equations. Solving
for x in terms of s, we get
s2 3
x= .
s 2
Now we plug this into the first equation:
2 2
s 3
+ s2 = 11,
s 2
and multiply through by 2s2 (to clear the denominator) and expand and simplify to get
3s4 28s2 + 9 = 0.
This factors as (3s2 1)(s2 9) = 0, so s2 = 9 or s2 = 13 . Only s2 = 9 works with the

original system of equations, so s = 3.
Round 1 Solutions
www.usamts.org
Thus the side length of ABCD is 2s = 6 .

Solution 2 : Place the tetrahedron in Cartesian 3-space using the coordinates
A = (a, a, a), B = (a, a, a), C = (a, a, a), D = (a, a, a).

Note that the side length is 2 2a.
Let P = (x, y, z). Then P A = P B implies that y = z, and P C = P D implies that
y = z, so we must have y = z = 0, and hence P = (x, 0, 0).
click here to buy Virtual PDFPrinter
Computing P A = 11 and P C = 17 gives us the system
(x a)2 + 2a2 = 11,

(x + a)2 + 2a2 = 17.
3
Simplifying gives 4ax = 6, or x = 2a
. Substituting this back in gives
2
3
a + 2a2 = 11,
2a
and after multiplying both sides by 4a2 , we have (3 2a2 )2 + 8a4 = 44a2 . This expands to
a quadratic in a2 :
12a4 56a2 + 9 = 0,
9
which factors as (2a2 9)(6a2 1) = 0, so a2 = 2
or a2 = 16 . But a2 = 1
6
does not give a
valid solution, so a2 = 29 , and hence a = 32 .

Therefore, the side length of ABCD is 2 2a = 2 2 32 = 6 .
Round 1 Solutions
www.usamts.org

5/1/20. Call a positive real number groovy if it can be written in the form n + n + 1 for
some positive integer n. Show that if x is groovy, then for any positive integer r, the number
xr is groovy as well.
We prove two lemmas, which we will combine to prove the overall result.
Lemma 1: If x is groovy and r is odd, then xr is groovy.

Proof: Let x = n + 1 + n and let y = n + 1 n; note that xy = 1. Also note the
expansions
r
r
X r rk
k
x = n+1 n ,
k=0
k
r
rk

k r k
X
r
y = (1) n+1 n .
k=0
k
Since r is odd, adding xr and y r will cancel all of the terms in the above expansions in
which k is odd, and subtracting y r from xr will cancel all of the terms where k is even. Thus,
we see that
p
xr + y r = 2a n + 1 = 4a2 (n + 1), (1)

r r
x y = 2b n = 4b2 n, (2)
where a and b are integers. Squaring each equation in the system gives us
x2r + 2 + y 2r = 4a2 (n + 1), (3)
x2r 2 + y 2r = 4b2 n. (4)
Subtracting (4) from (3) gives 4 = 4a2 (n + 1) 4b2 n, so 4a2 (n + 1) = 4b2 n + 4.
Then, adding (1) and (2) gives us
p
2xr = 4a2 (n + 1) + 4b2 n = 4b2 n + 4 + 4b2 n,

and hence xr = b2 n + 1 + b2 n. Thus xr is groovy. 2
Lemma 2: If x is groovy, then x2 is groovy.

Proof : Let x = n + 1 + n. We now compute:
2
x2 = n+1+ n
p
= n + 1 + 2 n(n + 1) + n
p
= (2n + 1) + 2 n(n + 1)

= 4n2 + 4n + 1 + 4n2 + 4n.
Round 1 Solutions
www.usamts.org
So x2 is groovy. 2
a
To complete the proof, we prove that x(2 r) is groovy for any nonnegative integer a and
any positive odd integer r, by induction on a. The base case of the induction (where a = 0)
is Lemma 1, and the inductive step is Lemma 2.
Credits: All problems and solutions were written by USAMTS staff. Dr. Erin Schram sug-
gested the wording for problem 2/1/20.

Round 2 Solutions
www.usamts.org
1/2/20. Sarah and Joe play a standard 3-by-3 game of tic-tac-toe.

Sarah goes first and plays X, and Joe goes second and plays O. They
alternate turns placing their letter in an empty space, and the first
to get 3 of their letters in a straight line (across, down, or diagonal)
wins. How many possible final positions are there, given that Sarah
wins on her 4th move? (Dont assume that the players play with any
sort of strategy; one example of a possible final position is shown at
right.)
There are two cases.
Case 1: Sarah wins in a horizontal or vertical row. There are 6 such winning rows.
The three Os can be placed in any of the remaining 6 squares, except that they cannot be
placed all in either of the two lines that are parallel to the winning line of Xs. So there are
6

3
2 = 18 ways to place the Os. This leaves 3 squares in which to place the fourth X.
Thus there are 6 18 3 = 324 positions in this case.
Case 2: Sarah wins in a diagonal. There are 2 such winning lines. The three Os can be
placed in any of the remaining 6 squares, so there are 63 = 20 ways to place the Os. This

leaves 3 squares in which to place the fourth X. Thus there are 2 20 3 = 120 positions in
this case.
We have a total of 324 + 120 = 444 possible final positions.
Round 2 Solutions
www.usamts.org
2/2/20. Let C1 , C2 , C3 , and C4 be four circles, with radii 1, 1, 3, and 3, respectively, such
that circles C1 and C2 , C2 and C3 , C3 and C4 , and C4 and C1 are externally tangent. A
fifth circle C is smaller than the four other circles and is externally tangent to each of them.
Find the radius of C.
Denote the centers of C1 , C2 , C3 , C4 as Z1 , Z2 , Z3 , Z4 , Z2 Z1

respectively, and the center of C as Y . Then,
because 4Y Z2 Z3 = 4Y Z1 Z4 and 4Y Z1 Z2 is
isosceles, we have Z3 Z2 Z1
= Z4 Z1 Z2 . Thus, Y
Z1 Z2 Z3 Z4 is an isosceles trapezoid with bases
Z1 Z2 = 2 and Z3 Z4 = 6 and legs Z2 Z3 =
Z3 Z4
Z4 Z1 = 4, as shown at right. Dropping a per-
pendicular from Z1 to Z3 Z4 gives a right trian-
gle with hypotenuse 4 and base 2, so the height
of the trapezoid is 2 3.
Z2 1 Z1 The point Y lies on the segment con-

necting the two midpoints of the bases

2 3h (because Y lies on the perpendicular bi-
sector of each of the parallel segments Z1 Z2
Y and Z3 Z4 ). Suppose that the distance
from Y to Z3 Z4 is h. Then the distance
h
from Y to Z1 Z2 is 2 3 h. Applying the
Pythagorean Theorem, we get
q
Z3 3 Z4 Y Z1 = 1 + (2 3 h)2 ,

Y Z4 = 9 + h2 .
But we also know that

Y Z1 1 = Y Z4 3 = radius of C.
Thus, Y Z1 + 2 = Y Z4 and hence
q
1 + (2 3 h)2 + 2 = 9 + h2 .
Square both sides to get

q
1 + h 4 3h + 12 + 4 1 + (2 3 h)2 + 4 = 9 + h2 .
2
Round 2 Solutions
www.usamts.org
Simplifying gives q
1 + (2 3 h)2 = 3h 2.
Squaring again gives
1 + h2 4 3h + 12 = 3h2 4 3h + 4.

This simplifies to 2h2 = 9, so h = 3/ 2.
Finally, this gives
p p p
YZ =
Create PDF with GO2PDF for free, if you wish to remove this line, 9 + h2 = 9 + (9/2) =
4 click here to buy Virtual PDF Printer (27/2) = 3 3/2,
and we conclude that

r ! !
3 6
radius of C = Y Z4 3 = 3 1 =3 1 .
2 2
Round 2 Solutions
www.usamts.org
3/2/20. Find, with proof, all polynomials p(x) with the following property:
There exists a sequence a0 , a1 , a2 , . . . of positive integers such that p(a0 ) = 1 and

for all positive integers n:
(a) p(an ) is a positive integer, and

n1
X 1 1
(b) + = 1.
a
j=0 j
p(an )
For any positive integer k, comparing equation (b) for n = k and n = k + 1 gives
1 1 1
= + .
p(ak ) ak p(ak+1 )
In particular,
1 1
> 0.
p(ak ) ak
p(ak )
Thus, we see that p(ak ) < ak for all k, and therefore < 1. Since the an grow arbitrarily
ak
n1
X 1 p(x)
large (otherwise, we would have > 1 for large n), we have 0 < < 1 for arbitrarily
j=0
a j x
large x. This means that p(x) must be linear or constant: if p(x) had a term of degree 2 or
higher, then p(x)/x would have a term of degree 1 or higher, and thus would grow arbitrarily
large.
1 1
Plugging in n = 1 to (b) gives + = 1, so since both a0 and p(a1 ) are positive
a0 p(a1 )
integers, we must have a0 = p(a1 ) = 2. Then, since p(a0 ) = 1, we have p(2) = 1. Since p(x)
is linear or constant, we need only determine one additional value of p in order to write the
formula for p(x). Since we know p(a1 ) = 2, it suffices to find a1 to determine p(x).
Plugging in n = 2 gives
1 1 1
+ + = 1,
2 a1 p(a2 )
so, since a1 and p(a2 ) must be positive integers, we must have one of the following cases.
Case 1: a1 = 3 and p(a2 ) = 6. Then, using p(a1 ) = 2, we have p(3) = 2, and hence
n1
Y
p(x) = x 1. We can verify that this works with the sequence a0 = 2 and an = 1 + ai .
i=0
Round 2 Solutions
www.usamts.org
p(x) = x/2. We can verify that this works with the sequence an = 2n+1 .
p(x) = (x/4) + 1/2. However, by p(a2 ) = 3, we have a2 = 10, and plugging n = 3 into (b)
gives
1 1 1 1
+ + + = 1,
2 6 10 p(a3 )
giving p(a3 ) = 30/7, which is not an integer.
So the only valid polynomials are x 1 and x/2 .
Round 2 Solutions
www.usamts.org
4/2/20. Find, with proof, the largest positive integer k with the following property:
There exists a positive number N such that N is divisible by all but three of the
integers 1, 2, 3, . . . , k, and furthermore those three integers (that dont divide N )
are consecutive.
Lemma 1: The three consecutive non-divisors must each be a positive power of a prime
number.
Proof: On the contrary, suppose that ab is one of the non-divisors of N , where a and b are
each greater than 1 and have no prime factors in common. Then either a or b does not
divide N ; suppose without loss of generality that a does not divide N . We then must have
ab 2 = a (since a and ab must be two of three consecutive numbers), so a(b 1) = 2, but
this means that a = b = 2, contradicting the assumption that a and b had no prime factors
in common. 2
Lemma 2: 7, 8, 9 are the three greatest consecutive positive integers that are each a positive
power of a prime.
Proof: Other than 2, 3, 4, there is no sequence of three consecutive such integers in which
the middle one is odd, because the only powers of 2 that are 2 apart are 2 and 4. So any
greater sequence of three consecutive integers must be of the form
2n 1, 2n , 2n + 1
for some positive integer n 2. Setting n = 2 gives the sequence 3, 4, 5. If n > 2, then we
cannot have n even, since otherwise
2n 1 = (2(n/2) 1)(2(n/2) + 1)
cannot be the power of an odd prime. Therefore we must have n odd. Also, any three
consecutive integers must contain a multiple of 3, so one of our numbers must be a power of
3. And, since 2n 2 (mod 3) for odd n, we must have 2n + 1 = 3m for some positive integer
m. Also, since 2n 0 (mod 4) for n 2, we must have m even (so that 3m 1 (mod 4)).
But then we can factor
2n = 3m 1 = (3(m/2) 1)(3(m/2) + 1),
which can only be a power of 2 if the two factors on the right side above (which differ by 2)
are 2 and 4, giving n = 3 and m = 2. This gives the three consecutive integers 7, 8, 9. 2
Round 2 Solutions
www.usamts.org
So if N is not divisible by 7, 8, or 9, then it is possible that N might be divisible by

10, 11, 12, 13, but N cannot be divisible by 14 (since it is not divisible by 7). Thus the
upper bound on k is 13. We can easily verify that N = 4 3 5 11 13 is divisible by all of
{1, 2, 3, 4, 5, 6, 10, 11, 12, 13} but none of {7, 8, 9}. Therefore the maximum value of k is 13 .
Round 2 Solutions
www.usamts.org
5/2/20. The set S consists of 2008 points evenly spaced on a circle of radius 1 (so that S forms
the vertices of a regular 2008-sided polygon). 3 distinct points X, Y, Z in S are chosen at

random. The expected value of the area of 4XY Z can be written in the form r cot 2008 ,
where r is a rational number. Find r.
Let n = 2008. Fix one of the points A in S. We can assume without loss of generality
that A is one of the vertices of our randomly chosen triangle. Let B and C be the other
two vertices so that A, B, C are in clockwise order, and let O be the center of the n-sided
polygon.
Then
2i 2j 2k
AOB = , BOC = , COA = ,
n n n
where i, j, k are positive integers such that
1 i n 2, 1 j n 1 i, k = n i j.
Note that the angles sum to 2 and the bounds ensure that i, j, k are all positive.
The angles of triangle ABC are in
, j
n
, and k
n
. We use the result that a triangle with
angles , , and circumradius R has area
2R2 (sin )(sin )(sin ).
(This formula is an immediate consequence of the Law of Sines: it can be proved by using the
general area formula 21 ab sin , where a and b are the lengths of the sides adjacent to angle
, and then using (a/ sin ) = (b/ sin ) = 2R.) We have R = 1 and = i/n, = j/n,
and = k/n, so our area is:
Area of ABC = 2(sin )(sin )(sin ).
A double use of the product-to-sum formulas, and the fact that + + = , gives
Area of ABC = 2(sin )(sin )(sin )
1
= (sin( + + ) + sin( + ) + sin( + ) sin( + + ))
2
1
= (sin( 2) + sin( 2) + sin( 2) + sin())
2
1
= (sin 2 + sin 2 + sin 2),
2
Therefore,
1 2i 2j 2k
Area of ABC = sin + sin + sin .
2 n n n
Round 2 Solutions
www.usamts.org
Summing over all triangles with vertex A, we get

n2 n1i
1X X 2i 2j 2(n i j)
Total area = sin + sin + sin .
2 i=1 j=1 n n n
We note that for any positive integer 1 a n 2, the term sin 2a

n
will appear exactly
n 1 a times in each position in the above sum. Therefore, we have
n2
3X 2a
Create PDF with GO2PDF for free, if you wish to remove this line, click Total
here to buy area =Printer (n 1 a) sin
Virtual PDF .
2 a=1 n
2i
Let S be the above summation (so that the total area is 32 S). Define = e n . Then S
is the imaginary part of
z = (n 2) + (n 3) 2 + (n 4) 3 + + n2 .
Then
z = (n 2) 2 + (n 3) 3 + (n 4) 4 + + n1 .
and thus
(1 )z = (n 2) 2 3 n2 n1
= (n 1) ( + 2 + + n1 )
= (n 1) + 1.
Thus
(n 1) + 1 n
z= = + 1.
1 1
Let = ei/n . Then
n
z= 1
1
n 2
= 1
1 2
n
= 1
(1/)
n(cos n + i sin n )
= 1
2i sin n
n cos n n
= i 1.
2 sin n 2
Round 2 Solutions
www.usamts.org
Thus
n
S = Im(z) = cot ,
2 n
and the total area is 23 S = 34 n cot n .
Therefore, since there are n1

2
choices for the points B and C, the expected value of
the area is
3n
4
cot n 3n
n1
= cot .
2
2(n 1)(n 2) n
3(2008) 502 502

Plugging in n = 2008, we see that the answer is = = .
2(2007)(2006) (669)(1003) 671007
Note: Unbeknownst to the USAMTS staff and our problem reviewers at the time that
the problems were being prepared, the solution to Problem 5/2/20 appeared in:
Shova KC and Anna Madra, Randomly generated triangles whose vertices are vertices
of a regular polygon, Rose-Hulman Undergraduate Mathematics Journal, 7, No. 2, 2006.
A copy of this paper (which also has the solution to the continuous version of this problem)
is available at http://www.rose-hulman.edu/mathjournal/v7n2.php
Credits: All problems and solutions were written by USAMTS staff; see special credit for
Problem 5/2/20 above. An earlier version of Problem 4/2/20 was suggested by Chris Jeuell.

Round 3 Solutions
www.usamts.org
1/3/20. Let S be the set of all 10-digit numbers (which by definition may not begin with 0) in
which each digit 0 through 9 appears exactly once. For example, the number 3,820,956,714
is in S. A number n is picked from S at random. What is the probability that n is divisible
by 11?
Let the number n be represented as abcdef ghij, where each letter represents a digit.
Then n is divisible by 11 if and only if the odd-digit sum a + c + e + g + i and the
even-digit sum b + d + f + h + j differ by a multiple of 11. However, since we know that
a + b + c + d + e + f + g + h + i + j = 0 + 1 + 2 + 3 + 4 + 5 + 6 + 7 + 8 + 9 = 45,
the only possibility is that one sum (the odds or the evens) is 17 and the other is 28. We
can now list the sets of 5 distinct digits that sum to 28 (the other 5 digits will necessarily
have to sum to 17):
9+8+7+4+0 9+8+7+3+1 9+8+6+5+0 9+8+6+4+1

9+8+6+3+2 9+8+5+4+2 9+7+6+5+1 9+7+6+4+2
9+7+5+4+3 8+7+6+5+2 8+7+6+4+3
Thus, there are 11 ways to split the digits into one group of five that sum to 28 and one
group of five that sum to 17. For each of these 11 ways, we have 2 choices which digits get
put into the odd positions and which get put into the even positions, and then 5! ways to
arrange the five digits in each group. However, exactly 1/10 of these will end up with 0 as
the leading digit, which is not allowed. Therefore, there are
9 99
11 2 (5!)2 = (5!)2
10 5
numbers in S that are divisible by 11. There are 9 9! numbers in S, so the probability is
99 (5!)2 11 4! 11
= = .
5 9 9! 9876 126
Round 3 Solutions
www.usamts.org
2/3/20. Two players are playing a game that starts with 2009 stones. The players take turns
removing stones. A player may remove exactly 3, 4, or 7 stones on his or her turn, except
that if only 1 or 2 stones are remaining then the player may remove them all. The player who
removes the last stone wins. Determine, with proof, which player has a winning strategy,
the first or the second player.
In a winning position, there exists a move that either wins the game immediately or
leaves the opponent in a losing position. In a losing position, every move results in a winning
position (for the other player). We can start with a small number of stones and work
backwards to classify the winning and losing positions. We will list all positions with 20 or
fewer stones:
# of stones Type of position Winning move (if any)
1 Winning take 1
2 Winning take 2
3 Winning take 3
4 Winning take 4
5 Losing must leave 1 or 2
6 Losing must leave 2 or 3
7 Winning take 7
8 Winning take 3
9 Winning take 3
10 Winning take 4
11 Losing must leave 4, 7, or 8
12 Winning take 7
13 Winning take 7
14 Winning take 3
15 Winning take 4
18 Winning take 7
19 Winning take 3
20 Winning take 3
Clearly we do not want to continue doing this up to 2009, but we notice that the pattern
repeats every 11 stones. Specifically, we note that the position type (winning or losing) only
depends on the number of stones modulo 11:
Round 3 Solutions
www.usamts.org
# of stones (mod 11) Type of position Winning move (if any)

1 Winning take 7 to leave 5 (or take 1 to win)
2 Winning take 7 to leave 6 (or take 2 to win)
3 Winning take 3 to leave 0
(Note that that stones that are left in the above table indicates the number of stones
modulo 11.)
Again, we note that from each winning position, there exists a move to a losing position,
and from each losing position, all moves go to winning positions.
Since 2009 = (182)(11) + 7 7 (mod 11), the first player wins by taking 7 stones to
leave a multiple of 11.
Round 3 Solutions
www.usamts.org
3/3/20. Let a, b, c be three positive integers such that
(lcm(a, b))(lcm(b, c))(lcm(c, a)) = (abc) gcd(a, b, c),
where lcm means least common multiple and gcd means greatest common divisor.
Given that no quotient of any two of a, b, c is an integer (that is, none of a, b, c is an integer
multiple of any other of a, b, c), find the minimum possible value of a + b + c.
Let p be any prime, and suppose that pr , ps , pt are the maximum prime powers that are
factors of a, b, c, respectively. Without loss of generality, suppose that r s t. Then,

isolating the powers of p in our given equation gives us
(ps )(pt )(pt ) = pr+s+t pr ,
so s + 2t = 2r + s + t, giving t = 2r.
Therefore, for any prime that appears as a factor in a, b, c, the greatest exponent of that
prime (in a, b, or c) must be exactly twice the smallest exponent of that same prime (in a, b,
or c). In particular, if any prime divides one of a, b, and c, then it must divide all of them.
We cannot have a, b, c all be a power of the same prime, since then clearly the smallest
one would divide the largest one.
Suppose that each of a, b, c are divisible by exactly two primes p and q. We cannot
have one of a, b, c have the largest power of both primes simultaneously, since otherwise the
other two numbers would divide it. So without loss of generality, suppose that a has the
largest power of p and b has the largest power of q. Since c does not divide a or b, we must
have that c has a higher power of q than a and a higher power of p than b. So we cannot
have a = p2 q and b = pq 2 , since this would imply c = p2 q 2 and now a and b each divide c.
The next smallest possibility is a = p4 q 2 , b = p2 q 4 , and c = p3 q 3 . Plugging in the smallest
possible primes p = 2 and q = 3 gives
a = 24 32 = 144, b = 22 34 = 324, c = 23 33 = 216,
so the sum is a + b + c = 684. This is the smallest sum with two primes.
Now suppose that each of a, b, c are divisible by exactly three primes p,q,r. The smallest
possibility is if, for each prime, two of the numbers are divisible by the prime and the third
is divisible by the prime squared; that is, a = p2 qr, b = pq 2 r, and c = pqr2 . Plugging in the
smallest primes p = 2, q = 3, and r = 5 gives
a = 22 3 5 = 60, b = 2 32 5 = 90, c = 2 3 52 = 150,

Round 3 Solutions
www.usamts.org
giving a sum of a + b + c = 60 + 90 + 150 = 300. This is the smallest sum with three primes,
and is the smallest sum weve found so far.
If there are 4 or more primes, then each number must be at least 2 3 5 7 = 210, so the
sum must be at least 630, which is larger than our previous example.
So the smallest possible sum is 300 .
Round 3 Solutions
www.usamts.org
4/3/20. Given a segment BC in plane P, find the locus of all points A in P with the following
property:
There exists exactly one point D in P such that ABDC is

a cyclic quadrilateral and BC bisects AD, as shown at right.
A
We set up coordinates: let B = (1, 0) and C = (1, 0). Let
A = (x, y) be a point in our locus. Let E = (s, 0) be the point
B C
of intersection
Create PDF with GO2PDF for free, if you wish toof AD
remove this and
line, clickBC, with
here to buy Virtual1 < s < 1. Then, since E
PDF Printer
is the midpoint of AD, we get D = (2s x, y). Since ABDC

is cyclic, we use Power of a Point at the point E to get the D
equation (EB)(EC) = (EA)(ED); after plugging in the coordinates, this gives
(1 s)(1 + s) = (s x)2 + y 2 .
Simplifying gives
1 s2 = s2 2xs + x2 + y 2 ,
which gives the quadratic in s:
2s2 2xs + x2 + y 2 1 = 0.
Because the point D is uniquely determined, this equation must have exactly one solution
in s, so its discriminant (as a quadratic in s) must be 0. That is,
4x2 8(x2 + y 2 1) = 0.
Simplifying this expression gives
x2
+ y 2 = 1.
2
So the locus of A is the set of solutions (x, y) to the above equation such that ABC is a
nondegenerate triangle. Thus, we get an ellipse, centered at (0, 0) (the midpoint of BC),
with major axis along BC with length 2(BC), and minor axis with length BC; however,
the two points on the ellipse at the ends of the major axis are not part of the locus (since at
these points ABC is degenerate). Note that B and C are the foci of the ellipse.
B C
D
Round 3 Solutions
www.usamts.org
1 1
5/3/20. Let b be an integer such that b 2, and let a > 0 be a real number such that + > 1.
a b
Prove that the sequence
bac, b2ac, b3ac, . . .
contains infinitely many integral powers of b. (Note that bxc is defined to be the greatest
integer less than or equal to x.)
We proceed by contradiction. Suppose that, for some positive integer M and all integers
m click here to buy Virtual PDF Printer
m M , the number b is not equal to bkac for any positive integer k. This means that for
any given m M , there exists a number n such that
na < bm < bm + 1 (n + 1)a. (1)
We first prove that (1) implies
bna < bm+1 < bm+1 + 1 (bn + 1)a. (2)
Since na < bm , we get bna < bm+1 after multiplication by b, which gives the left inequality
of (2).
We now prove the right half of (2). Since bm + 1 (n + 1)a from (1), we multiply by b
to get
bm+1 + b (bn + b)a. (3)
1 1
Since + > 1, we get that a(b 1) < b, and by adding bm+1 we get
a b
bm+1 + a(b 1) < bm+1 + b. (4)
Combining (4) with (3), we get that
bm+1 < (bn + b)a a(b 1) = (bn + 1)a. (5)
However, we cannot have b(bn+1)ac = bm+1 , since this would violate our original assumption
that bm+1 is not of the form bkac for any positive k. Therefore, in light of (5), we must have
bm+1 + 1 (bn + 1)a,
and the right half of (2) is proved.

We now apply (2) repeatedly to get the inequalities
br na < bm+r < bm+r + 1 (br n + 1)a,

Round 3 Solutions
www.usamts.org
for all integers r 1. But, dividing by br , this means that

m 1
na < b n + r a (6)
b
for all integers r 1. Subtracting na from (6) gives

a
0 < bm na ,
br
a
which cannot be true for all positive integers r, since
br
becomes arbitrarily small as r gets
sufficiently large. This is our contradiction.
Therefore, our original assumption was false, and there are infinitely many integral powers
of b in the sequence {bnac}.
Credits: Problem 3/3/20 and its solution are due to Palmer Mebane, and Dr. Erin Schram
suggested the version of Problem 2/3/20. All other problems and solutions are by USAMTS
staff.

Round 4 Solutions
www.usamts.org
a2n1
1/4/20. Consider a sequence {an } with a1 = 2 and an = for all n 3. If we know that
an2
a2 and a5 are positive integers and a5 2009, then what are the possible values of a5 ?
Since a1 and a2 are positive integers, all of the subsequent terms must be positive. Divide
both sides of the recursion by an1 to get
an an1
= .
an1 an2
Thus, the ratio of consecutive terms is constant, and the sequence is a geometric sequence.
x 4 x4
If a2 = x, then the ratio between consecutive terms is x/2. Hence a5 = 2 = .
2 8
For this to be an integer, given that x is an integer, it is necessary and sufficient that x be
a multiple of 2.
The inequality a5 2009 gives us
x4
2009 x4 16072.
8
Note that 114 < 16072 < 124 , so we must have x 11. But since x must be even, we must
have x {2, 4, 6, 8, 10}. Plugging these values of x into a5 = x4 /8 gives:
a5 {2, 32, 162, 512, 1250}

Round 4 Solutions
www.usamts.org
2/4/20. There are k mathematicians at a conference. For each integer n from 0 to 10, inclusive,
there is a group of 5 mathematicians such that exactly n pairs of those 5 mathematicians
are friends. Find (with proof) the smallest possible value of k.
There must be 5 mathematicians that are all friends (giving 10 pairs of friends for that
group), and 5 mathematicians that all are not friends (giving 0 pairs of friends for that
group). If k 8, then these conditions cannot both be simultaneously satisfied: if there are
5 mathematicians that are all friends, then any group of 5 mathematicians will contain at
least 2 from the group of 5 that are all friends, so we cannot find a group of 5 with no pairs
of friends.
Thus we must have k 9. We will show that k = 9 is achievable.
Let A, B, C, D, E be group of 5 mathematicians that are all friends, and let W, X, Y, Z
be a group that are all not friends. Further, suppose:
A is friends with W , X, Y , and Z
B is friends with W , X and Y (and not friends with Z)
C is friends with W and X (and not friends with Y and Z)
D is friends with W (and not friends with X, Y , and Z)
E is not friends with any of W , X, Y , and Z
Then we have the following groups with the required exact number of friends:
Subset Number Pairs of friends

{E, W, X, Y, Z} 0 none
{D, W, X, Y, Z} 1 {D, W }
{C, W, X, Y, Z} 2 {C, W }, {C, X}
{B, W, X, Y, Z} 3 {B, W }, {B, X}, {B, Y }
{A, W, X, Y, Z} 4 {A, W }, {A, X}, {A, Y }, {A, Z}
{B, C, D, X, Z} 5 {B, C}, {B, D}, {C, D}, {B, X}, {C, X}
{B, C, D, E, Z} 6 all 6 pairs in {B, C, D, E}
{A, B, C, D, Z} 7 {A, Z}, all 6 pairs in {A, B, C, D}
{A, B, C, D, Y } 8 {A, Y }, {B, Y }, all 6 pairs in {A, B, C, D}
{A, B, C, D, X} 9 {A, X}, {B, X}, {C, X}, all 6 pairs in {A, B, C, D}
{A, B, C, D, E} 10 all 10 pairs in {A, B, C, D, E}
Thus the smallest possible value of k is k = 9 .

Round 4 Solutions
www.usamts.org
3/4/20. A particle is currently at the point (0, 3.5) on the plane and is moving towards the
origin. When the particle hits a lattice point (a point with integer coordinates), it turns with
equal probability 45 to the left or to the right from its current course. Find the probability
that the particle reaches the x-axis before hitting the line y = 6.
Note that the direction of the first move is irrelevant because of the symmetry. After
that, we can sketch the possibilities:
y
Start
x
The green arrows are guaranteed wins. If the particle follows the blue arrow ending at
(4, 3), then the probability of winning from there is 21 , by symmetry.
Let:
p be the probability of winning from the start circle at (0, 3)
q be the probability of winning from the square at (2, 2)
r be the probability of winning from the diamond at (3, 1)
We then note, by symmetry, that:
the probability of winning from the circle at (5, 3) is 1 p
the probability of winning from the square at (6, 2) is q
the probability of winning from the diamond at (4, 5) is 1 r
Round 4 Solutions
www.usamts.org
Therefore, we can write the following system of equations:

1 1
p= + q,
2 2
1 1 1
q= + r + (1 r),
8 2 4
3 1 1
r= + q + (1 p).
4 8 8
We can clear the denominators and collect terms:
2p = 1 + q,
8q = 3 + 2r,
8r = 7 p + q.
Substituting the 3rd equation into the 2nd equation gives:
2p = 1 + q,
31q = 19 p.
So the first equation becomes
62p = 31 + 31q = 50 p,
50
hence 63p = 50 and p = .
63
Round 4 Solutions
www.usamts.org
4/4/20. Find, with proof, all functions f defined on nonnegative integers taking nonnegative
integer values such that
f (f (m) + f (n)) = m + n
for all nonnegative integers m, n.
Let a = f (0). Plugging in m = n = 0 to the equation gives
0 = m + n = f (f (m) + f (n)) = f (2f (0)) = f (2a).

So f (2a) = 0. Then, plugging in m = n = 2a gives
4a = m + n = f (f (m) + f (n)) = f (f (2a) + f (2a)) = f (0 + 0) = f (0) = a.
So 4a = a, hence a = 0. Thus f (0) = 0.

Now, plugging in n = 0 for an arbitrary m gives
m = m + 0 = f (f (m) + f (0)) = f (f (m) + 0) = f (f (m)),
so f (f (m)) = m for all m. In particular, apply f to both sides of the original equation to
get
f (m) + f (n) = f (f (f (m) + f (n))) = f (m + n).
In particular, letting n = 1 gives f (m + 1) = f (m) + f (1).
Let f (1) = b, so that (by a trivial induction) we have f (m) = mb for all nonnegative
integers m. But m = f (f (m)) = f (mb) = mb2 , so we must have b2 = 1, hence b = 1.
Therefore, the only function that satisfies the functional equation is f (m) = m for all m .
Round 4 Solutions
www.usamts.org
5/4/20. A circle C1 with radius 17 intersects a circle C2 with C2

radius 25 at points A and B. The distance between the C1 A
centers of the circles is 28. Let N be a point on circle C2
such that the midpoint M of chord AN lies on circle C1 .
Find the length of AN .
M
Let C3 be the image of C2 under a dilation through A B

by a factor of 1/2. Let O1 , O2 , O3 be the centers of C1 , N
C2 , C3 , respectively, so O3 is the midpoint of AO2 .
C2
C3
C1 A
O3
O1
O2
M
Then M is the image of N under this dilation. However, M also lies on C1 , so M is the
intersection of C1 and C3 , other than A.
Let P be the intersection of O1 O3 and AM . Since AM is a common chord of circles C1
and C3 , AM O1 O3 , so AP is the height from vertex A to base O1 O3 in triangle AO1 O3 .
Let = O1 AO3 . Note that AO1 = 17, AO2 = 25, and O1 O2 = 28, so by the Law of
Cosines,
172 + 252 282 13
cos = = .
2 17 25 85
Then
132 7056 842
sin2 = 1 2 = = ,
85 852 852
so
84
sin = .
85
(Since 0 < < , we take the positive root.)
Then
1 1 25 84
[O1 AO3 ] = AO1 AO3 sin = 17 = 105,
2 2 2 85
Round 4 Solutions
www.usamts.org
and again by the Law of Cosines,
(O1 O3 )2 = (AO1 )2 + (AO3 )2 2AO1 AO3 cos

252 25 13
= 172 + 2 17
4 2 85
625
= 289 + 65
4
1521
=
4
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual2PDF Printer
39
= 2,
2
39
hence O1 O3 = .
2
Therefore,
2[O1 AO3 ] 2 105 140
AP = = = .
O1 O3 39/2 13
Finally, P is the midpoint of AM , and M is the midpoint of AN , so
560
AN = 4AP = .
13
Credits: All problems and solutions are by USAMTS staff.


Round 1 Solutions
www.usamts.org
1/1/21. Fill in the circles in the picture at right with the digits 1-8, one
digit in each circle with no digit repeated, so that no two circles that
are connected by a line segment contain consecutive digits. In how
many ways can this be done?
We observe first that the lower center circle (shaded in the diagram
below to the right) is connected to all but one of the other circles. If
we place any digit other than 1 or 8 in this circle we will not be able
to complete the grid, since the digits 27 each have two other digits
that are not allowed to be adjacent to it. Furthermore, if the lower center circle contains 1,
then the top circle must contain 2, because 2 cannot be placed adjacent to 1.
We notice that if we are given a solution to this problem, then 2

replacing every number x in the graph by 9 x gives us a different so-
lution. Furthermore this operation gives a 1-1 correspondence between
the solutions with 1 in the lower center circle and the solutions with
8 in the lower center circle. Thus, we count the number of solutions
where 1 is in this circle, and the total will be twice this count.
1
The remaining six empty circles form a hexagon. We must place
the digit 3 in one of the bottom 3 circles (since it cannot be adjacent
to 2). Once we have chosen where to place 3, we are left filling the
hexagon with the digits 4 through 8. At this point the locations of 1
and 2 are irrelevant, so we can count each of these three possibilities for the location of 3 by
counting three times the number of possibilites when 3 is placed at the bottom of a hexagon
and the remaining digits are placed elsewhere. We now see that the answer to the original
problem is 2 3 = 6 times the number of ways to fill a six-circle loop with the digits 3 to 8
such that 3 is on the bottom and no two consecutive digits are adjacent.
Next we place the digit 4. There are two cases:
Case 1: 4 is opposite from 3. From here we have two choices 4

for the location of 5 (either of the circles adjacent to 3), and that
determines where 7 is placed uniquely, as in the figure to the
7
right. At this point we can place 6 and 8 in either of other two
nodes, giving a total of 2 2 = 4 possibilities for this case.
3
Round 1 Solutions
www.usamts.org
Case 2: 4 is two positions away from 3. There are two choices for the location of 4,
and these are symmetric, so we assume 4 is on the right-hand side of the diagram (and we
remember that we will get a factor of two from this symmetry). Now we have two non-
symmetric choices for placing 5. If we place 5 adjacent to 3 as in the left image below, there
is only one way to complete the diagram, as shown. If we place 5 a distance two from 3 (as
in the picture to the right below), then the location of 6 is now determined (as shown) and 7
and 8 may be placed freely. This gives two more options. So there are a total of 2(1+2) = 6
possibilities in this case.
6
8 4 5 4
5 7 6
3 3
Summing the two cases, this gives a total of 4 + 6 = 10 possibilities once 3 is placed.
Recalling the factor of 6 from earlier, we conclude that the total number of possibilities is
6 10 = 60 .
Round 1 Solutions
www.usamts.org
2/1/21. The ordered pair of four-digit numbers (2025, 3136) has the property that each number
in the pair is a perfect square and each digit of the second number is 1 more than the
corresponding digit of the first number. Find, with proof, all ordered pairs of five-digit
numbers and ordered pairs of six-digit numbers with the same property: each number in the
pair is a perfect square and each digit of the second number is 1 more than the corresponding
digit of the first number.
We observe first that if the digits of n2 are one less than the digits of m2 , then m2 n2 =
Create PDF with GO2PDF for free, if you wish
2 to remove
2 this line, click here to buy Virtual PDF Printer
11111 or m n = 111111 (depending on the lengths of m2 and n2 ). Thus we start by
looking for solutions to these two equations and then test all of our solutions.
If (n2 , m2 ) is an ordered pair of 5-digit numbers satisfying the desired property, then we
must have
11111 = m2 n2 = (m n)(m + n)
The number 11111 has only two factorizations into a product of two factors: 11111 = 41 271
and 11111 = 1 11111. Checking the two factorizations we have
(n, m) = (115, 156)

(n, m) = (5550, 5551)
The second pair will fail since n > 1000 implies that n2 must have at least 7 digits. We
check the first:
(n2 , m2 ) = (13225, 24336).
This pair works. So the only 5-digit pair is (13225, 24336).
Next we look for ordered pairs (n2 , m2 ) of six-digit numbers with the desired property.
We must have
111111 = m2 n2 = (m n)(m + n).
Again we factor: 111111 = 3 7 11 13 37. Thus, 111111 has 25 = 32 divisors, so
there are 16 ways to factor 111111 into the product of a pair of positive integers. But we
must have 100000 n2 m2 999999, so this severely restricts the possibilities: taking
the square root, we must have
317 n < m 999.
This means that the larger factor m + n must be at least 635 but at most 1997. This restricts
the choice of (m n, m + n) to the following pairs:
{(143, 777), (111, 1001), (91, 1221), (77, 1443)}.
These correspond to (n, m) equaling one of
{(317, 460), (445, 556), (565, 656), (683, 760)}.

Round 1 Solutions
www.usamts.org
We check these by computing (n2 , m2 ):
{(100489, 211600), (198025, 309136), (319225, 430336), (466489, 577600)}.
None of these pairs satisfies the desired property.

Our answer is:
There is one pair of 5-digit numbers which satisfies the property, (13225, 24336) .
There are no pairs of 6-digit numbers which satisfy the property.
Round 1 Solutions
www.usamts.org
3/1/21. A square of side length 5 is inscribed in a square of side length 7.

If we construct a grid of 1 1 squares for both squares, as shown to the
right, then we find that the two grids have 8 lattice points in common.
If we do the same construction by inscribing a square of side length 1489
in a square of side length 2009, and construct a grid of 1 1 squares in
each large square, then how many lattice points will the two grids of 1 1
squares have in common?
Let the two squares be ABCD and P QRS, and let A x P B
x = AP = BQ = CR = DS. Then P B = QC = RD =
SA = 2009 x. Then by the Pythagorean Theorem on
triangle AP S, 2009 x
1489 Q
x2 + (2009 x)2 = 14892 ,
S
which has solutions x = 689 and x = 1320. The lengths
689 and 1320 produce equivalent diagrams, so assume
x = 689.
D R C
For brevity, let a = 689, b = 1320, and c = 1489; note that c is prime. Consider
the diagram on the coordinate plane with D as the origin, so that C = (2009, 0) and
A = (0, 2009). Then, by thinking of starting at point S and moving in units in the directions

of SR and SP , we see that points on the lattice of P QRS are points of the form

b a a b
(0, a) + u , +v ,
c c c c
where 0 u c and 0 v c. This is a point on both lattices if and only if the coordinates
are integers; that is, if
ub + va ua + vb
and
c c
are both integers. So our goal is to count all (u, v) with 0 u c and 0 v c such that
ub + va and ua + vb are integer multiples of c; that is,
ub + va 0 (mod c), (1)

ua + vb 0 (mod c). (2)
Clearly if (u, v) {(0, 0), (0, c), (c, 0), (c, c)} then this condition is satisfiedthese corre-
spond to the four corners of the smaller square. We claim that if 0 < u < c, then there exists
a unique v with 0 < v < c satisfying (1) and (2). In particular, since c is prime, equation
(1) can be rewritten as
ub
v (mod c),
a
Round 1 Solutions
www.usamts.org
where division by a is well-defined modulo c. (Indeed, with a = 689, c = 1489 we can

check that a1 389 (mod c), since (389)(689) = 268021 = (180)(1489) + 1.) So there is a
unique v with 0 < v < c satisfying (1). To verify that this also satisfies (2), note that since
a2 + b2 = c2 , we have b2 a2 (mod c), and thus
b2 a2

ub
ua + vb ua + b a+ u a+ u 0 (mod c).
a a a
Thus, each 0 < u < c gives exactly one ordered pair (u, v) that corresponds to a lattice
point of both grids. This gives c 1 such points, so together with the four corners of P QRS,
we get a total of c + 3 points. Therefore, since c = 1489, the answer is 1492 .
Round 1 Solutions
www.usamts.org
4/1/21. Let ABCDEF be a convex hexagon, such that F A = AB, BC = CD, DE = EF ,

and F AB = 2EAC. Suppose that the area of ABC is 25, the area of CDE is 10, the
area of EF A is 25, and the area of ACE is x. Find, with proof, all possible values of x.
Since F AB = 2EAC, and the hexagon is convex, we have
EAC = F AB EAC = F AE + BAC.
Let X be the reflection of B across AC. Then CX = CB = CD. Also, AX = AB = AF ,

and
XAE = EAC XAC = (F AE + BAC) BAC = F AE.
Since AX = AF and XAE = F AE, we conclude that X is also the reflection of F across
AE. Hence EX = EF = ED.
Since CX = CD and EX = ED, we have two possibilities: either X = D, or X is the
reflection of D across CE. Both cases are shown below.
A F
A F
B
B
X
C E C E
D D
Let [P QR] denote the area of triangle P QR. In the first case,
[ACE] = [ADC] + [ADE] [CDE] = [ABC] + [AF E] [CDE] = 25 + 25 10 = 40.
In the second case,
[ACE] = [AXC] + [AXE] + [CXE] = [ABC] + [AF E] + [CDE] = 25 + 25 + 10 = 60.
Hence, the area of triangle ACE must be either 40 or 60. However we have not yet shown
that we can achieve these values. It remains to construct convex hexagons with the initial
conditions such that triangle ACE has these areas.
Round 1 Solutions
www.usamts.org
A
For the first case, let ACE be an equilateral tri-
angle with area 40, and let M be the midpoint of
CE. Let CM = a, so AM = a 3. Let D be the F
point on the perpendicular bisector of CE, on the B

opposite side of CE from A, such that DM = a 4 3 . M
Since triangle CDE shares a base with triangle ACE C E
but its altitude to that base is 1/4 the altitude of
D
ACE, its area is 1/4 that of ACE; thus, [CDE] =
1
(40)
Create PDF with GO2PDF
4
=if you
for free, 10.wishReflect
to remove this D across
line, click AC
here to buy Virtualto
PDFget
PrinterB, and reflect D across AE to get F . We have
[ABC] = [ADC] = [AM

C] + [DM C] = 12 (40) + 12 (10) = 25, and similarly [AF E] = 25.

Also, tan M CD = a( a3/4) = 43 < 33 = tan 30 , so M CD < 30 . Thus, BCD =
2ACD = 2(ACM + M CD) < 2(60 + 30 ) = 180 . Similarly, F ED < 180 , and
F AB = 2EAC = 2(60 ) < 180 . The remaining angles are obviously less than 180 , so
hexagon ABCDEF is convex. We have thus constructed the first case.
For the second case, let ACE be an equilateral A

triangle with area 60, and let M be the midpoint of
CE. Let X be the point on segment AM such that
XM = 16 AM . Then [CXE] = 16 [ACE] = 61 (60) = 10. F
B
Thus, [AXC] = 12 ([ACE][CXE]) = 12 (6010) = 25,
and similarly [AXE] = 25. Let B, D, and F be the
reflections of X across AC, CE, and EA, respectively.
Then [ABC] = 25, [CDE] = 10, and [EF A] = 25. X
Also, F AB = 2EAC = 2(60 ) < 180 , and simi- C E
larly BCD < 180 and DEF < 180 . The remain-
ing angles are obviously less than 180 , so hexagon D
ABCDEF is convex. Thus, we have constructed the second case.
Therefore, we conclude that 40 and 60 are the two possible areas of triangle ACE.
Round 1 Solutions
www.usamts.org
5/1/21. The cubic equation x3 +2x1 = 0 has exactly one real root r. Note that 0.4 < r < 0.5.
(a) Find, with proof, an increasing sequence of positive integers a1 < a2 < a3 < such
that
1
= ra1 + ra2 + ra3 + .
2
(b) Prove that the sequence that you found in part (a) is the unique increasing sequence
with the above property.
(a) Since r is a root of x3 + 2x 1, we know that 1 r3 = 2r, so

1 r
= = r(1 + r3 + r6 + r9 + ) = r + r4 + r7 + r10 +
2 1 r3
Furthermore, since |r| < 1, this sequence converges. Therefore an = 3n 2 solves the
equation above.
(b) Suppose that a1 < a2 < a3 < and b1 < b2 < b3 < are distinct sequences of
positive integers such that
1
ra1 + ra2 + ra3 + = rb1 + rb2 + rb3 + = .
2
Eliminating duplicate terms, we have
rs1 + rs2 + = rt1 + rt2 + = a > 0,
where s1 < s2 < . . . and t1 < t2 < . . . are positive integers, and all the ss and ts are
distinct. (Note that either series above might in fact be finite, but this does not affect
the validity of the argument to follow.) Assume with loss of generality that s1 < t1 .
Then
rs1 rs1 + rs2 + = rt1 + rt2 + . . . ,
so that, after dividing by rs1 , we have
1 ru1 + ru2 + . . . ,
where ui = ti s1 > 0. Thus, since 0 < r < 21 , we have 1

1r
< 2, and hence
1
1 ru1 + ru2 + r + r2 + r3 + = 1 < 1,
1r
a contradiction.
Round 1 Solutions
www.usamts.org
Credits: Problem 2/1/21 is based on a problem appearing in the February, 1997, issue of
Crux Mathematicorum with Mathematical Mayhem.
Problem 4/1/21 was proposed by Gaku Liu.
All other problems and solutions by USAMTS staff.

Round 2 Solutions
www.usamts.org
1/2/21. Jeremy has a magic scale, each side of which holds a positive integer. He plays the
following game: each turn, he chooses a positive integer n. He then adds n to the number
on the left side of the scale, and multiplies by n the number on the right side of the scale.
(For example, if the turn starts with 4 on the left and 6 on the right, and Jeremy chooses
n = 3, then the turn ends with 7 on the left and 18 on the right.) Jeremy wins if he can
make both sides of the scale equal.
(a) Show that if the game starts with the left scale holding 17 and the right scale holding 5,
then Jeremy can win the game in 4 or fewer turns.
(b) Prove that if the game starts with the right scale holding b, where b 2, then Jeremy
can win the game in b 1 or fewer turns.
(a) Jeremy wins as follows:
Turn n Left Right

Start 17 5
1 1 18 5
2 1 19 5
3 1 20 5
4 5 25 25
(b) Let a be the starting number on the left scale. First, notice that if on any turn Jeremy
chooses n = 1, then the left side increases by 1 and the right side remains unchanged. Thus,
our strategy is to have Jeremy choose 1 each turn until the left side is a multiple of b 1.
This will take at most b 2 turns: if a is already a multiple of b 1, no turns are needed;
otherwise, if a = d(b 1) + r, where 1 r b 2, the left side will be a multiple of b 1
after b 1 r b 2 turns in which Jeremy chooses 1.
After this, the left side equals k(b 1), where k is some positive integer, and the right
side is still b. If Jeremy now chooses k, the left side becomes k(b 1) + k = kb and the right
side also becomes kb, so the game ends. This occurs in at most (b 2) + 1 = b 1 turns, as
desired.
Round 2 Solutions
www.usamts.org
2/2/21. Alice has three daughters, each of whom has two daughters; each of Alices six grand-
daughters has one daughter. How many sets of women from the family of 16 can be chosen
such that no woman and her daughter are both in the set? (Include the empty set as a
possible set.)
We proceed by cases. If Alice is in the set, then we cannot choose any of her daughters,
and for each of the 6 granddaughters, we can choose either that granddaughter, that grand-
daughters daughter, or neither, for a total of 3 choices per granddaughter. Thus, there are
36 =for 729
free, if you wish to remove this line, click here to buy Virtual PDF Printer
possible sets that include Alice.
If Alice is not in the set, then we view Alices three daughters as matriarchs of three dis-
joint identical family trees. For each of her three daughters, we can choose (a) 22 = 4 subsets
that include the daughter and any number of that daughters grandchildren, or (b) 32 = 9
subsets that do not include the daughter, but include her children and/or grandchildren (for
each daughter, we can take her, her daughter, or neither). This gives 4 + 9 = 13 subsets
associated to each of Alices three daughters, and thus there are (13)3 = 2197 subsets that
do not include Alice.
This gives a total of 729 + 2197 = 2926 allowed subsets.
Round 2 Solutions
www.usamts.org
3/2/21. Prove that if a and b are positive integers such that a2 + b2 is a multiple of 72009 , then
ab is a multiple of 72010 .
We first construct the following chart of squares modulo 7:
k 0 1 2 3 4 5 6
k2 0 1 4 2 2 4 1
The only pair of squares modulo 7 that can add to 0 (mod 7) is 0 + 0. Thus, a2 b2 0
(mod 7). This means that a = 7a1 and b = 7b1 for some positive integers a1 and b1 . But
then (7a1 )2 + (7a2 )2 is a multiple of 72009 , so a21 + a22 is a multiple of 72007 . By the same
reasoning, a1 = 7a2 and b1 = 7b2 for some positive integers a2 and b2 ; hence a = 72 a2 and
b = 72 b2 . But then a22 + b22 is a multiple of 72005 , so we can repeat the process.
Inductively, we get that a = 71005 a1005 and b = 71005 b1005 for some positive integers a1005
and b1005 , so that ab = 72010 (a1005 b1005 ), as desired.
Round 2 Solutions
www.usamts.org
4/2/21. The Rational Unit Jumping Frog starts at (0, 0) on the Cartesian plane, and each
minute jumps a distance of exactly 1 unit to a point with rational coordinates.
(a) Show that it is possible for the frog to reach the point 51 , 17
1

in a finite amount of time.
(b) Show that the frog can never reach the point 0, 41 .

(a) We use the Pythagorean triples {3, 4, 5} and {8, 15, 17} to construct the path from (0, 0)
1 1
to 5 , 17 shown below:

3 4 6 1 1 15 8 1 16 1 1
(0, 0) , ,0 ,0 , , , .
5 5 5 5 5 17 17 5 17 5 17
(b) Suppose the frog jumps such that its position changes by (r, s), where r and s are rational.
Let c be the least common denominator of r and s, so that r = ac , s = cb for some integers a
2 2
and b with gcd(a, b) = 1. The length of each jump is 1, so we have ac + cb = 1, which
gives a2 + b2 = c2 .
If a and b are both odd, then c2 = a2 + b2 2 (mod 4). Since no square is congruent to
2 mod 4, we conclude that a and b cannot both be odd. We cannot have a and b both even
because gcd(a, b) = 1. Therefore, one of a and b is even and the other is odd, so c2 = a2 + b2
is odd, which means c is odd.

th ai bi
Letting the i jump change the frogs location by ci , ci , the frogs location after k
jumps is !
k k
X ai X bi
, .
c
i=1 i i=1 i
c
Since all of the ci are odd, the denominator of each of these sums is odd. Therefore, the frog
cannot reach a point with an even denominator. Specifically, the frog cannot reach 0, 41 .

Round 2 Solutions
www.usamts.org
5/2/21. Let ABC be a triangle with AB = 3, AC = 4, and BC = 5, let P be a point on

BC, and let Q be the point (other than A) where the line through A and P intersects the
circumcircle of ABC. Prove that
25
PQ < .
4 6
Consider the triangle ABC. Points A, C, Q, and B are A

concyclic, so triangles AP C and BP Q are similar, which means
AC/BQ
Create PDF with GO2PDF = wish
for free, if you P C/P Q.
to remove Likewise,
this line, click here to buytriangles ABP and CQP are
Virtual PDF Printer 4
similar, so AB/CQ = P B/P Q. Then P QAC = BQCP and 3
P Q AB = BP CQ. Taking the product of these equations, B P
C
we get
P Q2 AB AC = BP CP BQ CQ
P Q2 3 4 = BP CP BQ CQ Q
By the AM-GM Inequality,

2
BP + CP 25
BP CP = ,
2 4
which tells us
25
P Q2 BQ CQ.
48
R Now we must maximize BQ CQ. This value is the twice

B C
area of the triangle BCQ, since Q is also a right angle. Let R
be the base of the altitude of BCQ from Q. Then
Q BQ CQ = 2[BCQ] = BC RQ = 5RQ.
Since Q lies on the unit circle and R lies on a diameter, the length of RQ is maximized
when R is the center of the circle, making RQ the radius, so
5 25
BQ CQ = 5RQ 5 =
2 2
This tells us that
25 25 25
P Q2 BQ CQ ,
48 48 2
Round 2 Solutions
www.usamts.org
or
25
PQ .
4 6
Finally notice that we used two inequalities in this argument. We maximized BP CP

via the AM-GM Inequality, which is uniquely maximized at BP = CP , so when P is the
center of the circle. We also maximized the area [BQC], which is uniquely maximized when
AQ is the angle bisector of A. However the angle bisector can only pass through the center
of the circle if AB = AC. Therefore we do not achieve both maxima simultaneously and the
inequality
Create PDF with GO2PDF for free, if youis strict:
25
PQ < .
4 6
Credits: Problem statements and solutions were written by USAMTS staff.

Problem 1/2/21 was submitted by Gaku Liu.
Problem 2/2/21 was submitted by Ben Dozier.
Problem 3/2/21 is based on a problem that appeared in the March 1997 issue of Crux Math-
ematicorum with Mathematical Mayhem.
Problem 4/2/21 is based on a problem from the 1994 Old Mutual Mathematics Olympiad and
appeared in the January 1995 issue of Mathematical Digest.
Problem 5/2/21 is a simplified version of a problem that appeared in issue 4/96 of Mathe-
matics and Informatics Quarterly.

Round 3 Solutions
www.usamts.org
1/3/21. Let ABCD be a convex quadrilateral with AC BD, and let P be the intersection
of AC and BD. Suppose that the distance from P to AB is 99, the distance from P to BC
is 63, and the distance from P to CD is 77. What is the distance from P to AD?
Denote by a the distance from A to P . Likewise, let b, P

c, and d be the distances from P to B, C, and D. Since
AC BD, the triangles AP B, BP C, CP D, and DP A all a b
99
have right angles at P , and the distance from P to a side is
the length of the altitude of the respective triangle. Consider A
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer B
triangle P AB at right. Then 99 (the distance from P to AB) a2 + b 2
is the length of the altitude, and computing the area of P AB in two ways gives

2(Area) = ab = 99 a2 + b2 .
Squaring and rearranging gives

a2 b 2 2
99 = 2 .
a + b2
The reciprocal of this equation will be more useful to us:
1 1 1
2
= 2 + 2.
99 a b
Denoting the distance from P to AD (the value we want to find) by x, we likewise have
1 1 1
2
+ 2 = 2
b c 63
1 1 1
2
+ 2 = 2
c d 77
1 1 1
2
+ 2 = 2.
d a x
Combining these in pairs we get
1 1 1 1 1 1 1 1
2
+ 2 = 2 + 2 + 2 + 2 = 2 + 2,
99 77 a b c d 63 x
or
1 1 1 1 1
2
= 2+ 2 2 = .
x 99 77 63 2312
Therefore x = 231 .
Round 3 Solutions
www.usamts.org
2/3/21. Find, with proof, a positive integer n such that
(n + 1)(n + 2) (n + 500)
500!
is an integer with no prime factors less than 500.
We can rewrite our expression as

(n + 1)(n + 2) (n + 500)

n+1 n+2 n+3 n + 500
=
500! 1 2 3 500
n n n n
= +1 +1 + 1 +1
1 2 3 500
500
Y n
= +1 .
j=1
j
We claim that n = (500!)2 satisfies the condition of the problem. To prove this, notice that
(500!)2
for every j with 1 j 500, the quantity is an integer and is furthermore divisible
j
2
by 500!. Therefore (500!)
j
is divisible by every prime less than 500. So for any prime p < 500,
500 500
(500!)2
Y Y
+1 11 (mod p),
j=1
j j=1
proving the result.

Round 3 Solutions
www.usamts.org
3/3/21. We are given a rectangular piece of white paper with length 25 and width 20. On the
paper we color blue the interiors of 120 disjoint squares of side length 1 (the sides of the
squares do not necessarily have to be parallel to the sides of the paper). Prove that we can
draw a circle of diameter 1 on the remaining paper such that the entire interior of the circle
is white.
If we are able to place our circle entirely on the paper such that it does not overlap any
squares, the center of the circle must have two properties:
I. The center of the circle must be a distance at least 1/2 from the boundary of the paper.
II. The center of the circle must be outside the neighborhood of distance 1/2 of every blue
square.
The area of the region that satisfies condition (I) is 24 19.
Consider the diagram to the right. The neighborhood of

distance 1/2 of a square is a region consisting of the square
(of area 1), 4 rectangles each with dimensions 1 12 , and four
quarter-circles each of radius 12 . Thus, the total area of each
neighborhood is

3+ .
4
The circle will not intersect the square if and only if the center
of the circle is outside this region.
Therefore, the total restricted region under condition (II) is
the union of the 120 neighborhoods of the blue squares. This area is at most 120 3 + 4 =

360 + 30 (it may be less if the neighborhoods overlap). Subtracting this area from the
allowed area under condition (I) gives an overall allowed area of at least
(24)(19) (360 + 30) = 96 30 = 6(16 5).
Since 16 > 5, the area of the allowed region is positive, and a circle of diameter 1 centered
at any point in this region will lie entirely on the page and will not overlap any of the squares.
Round 3 Solutions
www.usamts.org
4/3/21. Let a and b be positive integers such that all but 2009 positive integers are expressible
in the form ma + nb, where m and n are nonnegative integers. If 1776 is one of the numbers
that is not expressible, find a + b.
We begin by considering the question: Given general a and b, what is the set of numbers
which are not expressible as ma + nb where m and n are nonnegative integers? First notice
that if g = gcd(a, b) > 1, then g is a divisor of any ma + nb, so this set is infinite. We may
assume, then, that a and b are relatively prime. Further, since we are trying to find the sum
a + b, we may assume a < b.
The following is a known theorem, related to the Frobenius Coin Problem and Chicken
McNugget Problem.
Theorem 1: Let a and b be relatively prime positive integers, N = ab a b, and let S be
the set of nonnegative integers which are not expressible in the form an + bm for nonnegative
m, n. Then
1. Every element of S is less than or equal to N , and

2. If 0 x, y N and x + y = N , then exactly one of x or y is in S.
(For completeness, we include a proof of this theorem at the end of the solution.)
From this theorem we see that S contains exactly half of the integers from 0 to N , so
has size N2+1 = (a1)(b1)
2
(notice that N is odd). By assumption, there are 2009 elements of
S, so N = 2 2009 1 = 4017 and
(a 1)(b 1) = 4018 = 2 72 41.
The possible values of a < b are
a 2 3 8 15 42 50
b 4019 2010 575 288 99 83
Of these, only (2, 4019), (8, 575), and (50, 83) are relatively prime.
We know that, for any of these pairs, exactly one of 1776 and N 1776 = 4017 1776 =
2241 is expressible as a sum of the form ma + nb. If we let (a, b) = (50, 83), then we notice
2241 = 0 50 + 27 83,
proving that (a, b) = (50, 83) satisfies the criteria. Checking the other two pairs, we see
1776 = 888 2 + 0 4019
1776 = 222 8 + 0 575,
Round 3 Solutions
www.usamts.org
showing that these are not solutions. Therefore
a + b = 50 + 83 = 133 .
Proof of Theorem 1: Let
T = {ma + nb | 0 m < b and 0 n < a}.
By the Chinese Remainder Theorem, this set spans the residue classes modulo ab: each
residue class contains exactly one element of T . Every element in T is less than 2ab. Notice
that if c S, then c is not in T (although the converse is false).
For any integer c > ab, either c T or c = c0 + kab for c0 T and some k 1, so c
/ S.
Therefore S is a subset of the positive integers less than ab, so is finite.
Notice that the largest element of T is a(b 1) + b(a 1) = 2ab a b. If 2ab a b <
c < 2ab, then the element of T in the residue class modulo ab of c is c ab. Therefore T
contains all integers d with ab a b < d < ab, so the largest element of S is at most
ab a b. This proves the first part of the theorem.
Next, assume that 0 < c < N can be expressed as c = am + bn for nonnegative m and n
(so c T ). Clearly m < b and n < a. Then letting c0 = N c, we have
c0 = N c = (ab a b) (am + bn) = a(b m 1) + b(n 1).
This tells us that

c0 + ab = a(b m 1) + b(a n 1) T,
so the residue class of c0 mod (ab) is represented in T by c0 + ab. This tells us c0
/ T.
Furthermore, we can reverse this argument: If c0
/ T and 0 c0 N , then c0 + ab T ,
so
c0 = am + bn ab,
for nonnegative m < b and n < a. Therefore
c = ababc0 = abab(am+bnab) = 2ababambn = a(bm1)+b(an1).
This is an expression of c = N c0 as a nonnegative combination of a and b, proving
c T N c 6 T,
for all 0 c N . From this we can conclude that if 0 c N , then
c S N c 6 S.
Round 3 Solutions
www.usamts.org
5/3/21. The sequences (an ), (bn ), and (cn ) are defined by a0 = 1, b0 = 0, c0 = 0, and
cn1 an1 bn1
an = an1 + , bn = bn1 + , cn = cn1 +
n n n
for all n 1. Prove that
an n + 1 < 2

3 3n
for all n 1.
Solution 1. Computing the first few values of an , bn , and cn , we find a1 = 1, b1 = 1,

c1 = 0, a2 = 1, b2 = 1 + 12 , c2 = (12)
1
, and
1 1 1 1 1 1
a3 = 1 + , b3 = 1 + + , c3 = + + .
123 2 3 12 13 23
By a straightforward induction argument, we can prove the following: For a set X of real
numbers, let (X) denote the product of the elements of X. Set () = 1. Then
X 1 X 1 X 1
an = , bn = , cn = .
(S) (S) (S)
S{1,2,...,n} S{1,2,...,n} S{1,2,...,n}
|S|0 (mod 3) |S|1 (mod 3) |S|2 (mod 3)
The inequality in the problem is easily checked for n = 1 and n = 2, so henceforth,

assume that n 3. For 0 k n, let
X 1
sk = .
(S)
S{1,2,...,n}
|S|=k
Then an = s0 + s3 + s6 + , and the generating function for the sk is

x x x
f (x) = s0 + s1 x + s2 x2 + + sn xn = 1 + 1+ 1 + .
1 2 n
Let = e2i/3 . Then setting x = 1, , 2 , respectively, we obtain the equations
s0 + s1 + s2 + + sn = f (1),
s0 + s1 + s2 2 + + sn n = f (),
s0 + s1 2 + s2 4 + + sn 2n = f ( 2 ).
Adding, we get
3an = f (1) + f () + f ( 2 ),
Round 3 Solutions
www.usamts.org
so
f (1) + f () + f ( 2 )
an = .
3
Now,
1 1 1 2 3 n+1
f (1) = 1 + 1+ 1 + = = n + 1.
1 2 n 1 2 n
To get an estimate for f (), for 1 k n, let

1+ = rk eik ,
k
where rk > 0 and 0 < k < . Then
2
rk2 = 1 +

k
2
1 + 3i
= 1 +

2k

2
1 3
= 1 + 2
2k 4k
1 1 3
=1 + 2 + 2
k 4k 4k
1 1
=1 + 2
k k
k2 k + 1
= ,
k2
so
2
|f ()|2 = 1 + 1+ 1 +

1 2 n

= r12 r22 rn2
n
Y k2 k + 1
=
k=1
k2
1 3 7 n2 n + 1
=
12 22 32 n2
2
3 7 13 n n+1 1
= 2 2 2
1 2 3 (n 1)2 n2
n1
1 Y k2 + k + 1
= 2 .
n k=1 k2
Round 3 Solutions
www.usamts.org
Note that
k2 + k + 1 k
2
<
k k1
for all positive integers k 2, since this is equivalent to k 3 1 < k 3 . Hence,
n1
1 Y k2 + k + 1
|f ()|2 =
n2 k=1 k2
n1
3 Y k2 + k + 1
= 2
n k=2 k2
n1
3 Y k
< 2
n k=2 k 1
3(n 1)
=
n2
3n 3
< 2 = ,
n n
so r
3
|f ()| < .
n
Since f (x) is a polynomial in x with real coefficients,
f ( 2 ) = f () = f (),
so
|f ( 2 )| = |f ()| = |f ()|.
Hence, r
f () + f ( 2 ) |f ()| + |f ( 2 )|

n + 1 2 3 2
an = < = .
3 3 3 3 n 3n
Round 3 Solutions
www.usamts.org
Solution 2. By a straightforward induction argument, an + bn + cn = n + 1 for all n 0.

Let
dn = (an bn )2 + (an cn )2 + (bn cn )2
for all n 0. Then for all n 1,
dn = (an bn )2 + (an cn )2 + (bn cn )2
cn1 an1 2
= an1 + bn1
n n
2
c here to buy Virtual PDF Printer
Create PDF with GO2PDF for free, if you wish to remove this line, click n1
bn1
+ an1 + cn1
n n
2
an1 bn1
+ bn1 + cn1
n n
2
2(n n + 1) 2
= (an1 + b2n1 + c2n1 an1 bn1 an1 cn1 bn1 cn1 )
n2
n2 n + 1
= [(an1 bn1 )2 + (an1 bn1 )2 + (bn1 cn1 )2 ]
n2
n2 n + 1
= dn1 .
n2
Hence, !
n
Y k2 k + 1
dn = d0 .
k=1
k2
Since d0 = 2, by the same argument as in Solution 1, dn < 6/n for all n 1.
Now we introduce a lemma.
Lemma. Let x, y, z, and be real numbers such that x + y + z = 0 and
(x y)2 + (x z)2 + (y z)2 < .
Then
2
max{|x|, |y|, |z|} < .
3
Proof. Expanding, we get
(x y)2 + (x z)2 + (y z)2 = 2x2 + 2y 2 + 2z 2 2xy 2xz 2yz
= 2x2 + 2(y + z)2 4yz 2x(y + z) 2yz
= 6x2 6yz.
But
(y + z)2 (y z)2 x2 (y z)2
yz = = ,
4 4
Round 3 Solutions
www.usamts.org
so
3[x2 (y z)2 ]
(x y)2 + (x z)2 + (y z)2 = 6x2
2
9x2 + 3(y z)2
= .
2

2 2 2
Then 9x2 9x2 + 3(y z)2 < 2, so |x| < 3
. Similarly, |y| < 3
and |z| < 3
.
Take x = an n+1
3
, y = bn n+1
3
, and z = cn n+1
3
. Then x+y+z = an +bn +cn (n+1) = 0,
and
6
(x y)2 + (x z)2 + (y z)2 = (an bn )2 + (an cn )2 + (bn cn )2 = dn < ,
n
so by the lemma, p
n + 1 = |x| < 2 6/n = 2 .

an
3 3 3n

Problem 1/3/21 was submitted by George Berzsenyi.
Problem 2/3/21 is based on a problem from the January 1995 issue of Mathematical Digest.
Problem 4/3/21 is loosely based on Problem A5 from the 1971 Putnam Competition.

Round 4 Solutions
www.usamts.org
1/4/21. Archimedes planned to count all of the prime numbers between 2 and 1000 using the
Sieve of Eratosthenes as follows:
(a) List the integers from 2 to 1000.
(b) Circle the smallest number in the list and call this p.
(c) Cross out all multiples of p in the list except for p itself.
(d) Let p be the smallest number remaining that is neither circled nor crossed out. Circle p.
(e) Repeat steps (c) and (d) until each number is either circled or crossed out.
At the end of this process, the circled numbers are prime and the crossed out numbers
are composite.
Unfortunately, while crossing off the multiples of 2, Archimedes accidentally crossed out
two odd primes in addition to crossing out all the even numbers (besides 2). Otherwise,
he executed the algorithm correctly. If the number of circled numbers remaining when
Archimedes finished equals the number of primes from 2 to 1000 (including 2), then what is
the largest possible prime that Archimedes accidentally crossed out?
Let q and r be the two primes that Archimedes accidentally crossed out, with q < r. Our
goal is to maximize r subject to the conditions of the problem.
The composite numbers that will not get crossed out are q 2 , qr, and r2 . Since we must
have two uncrossed composite numbers between 2 and 1000 (to make up for the 2 primes
that accidentally got crossed out), we must have
q 2 < qr < 1000 < r2 .
Since q 3 we must have r 1000 3

. The largest prime number less than 10003
is 331, and
indeed, q = 3 and r = 331 works: if they get crossed out on the first step, then 9 and 993
will not get crossed out and will mistakenly be considered primes. Thus, the answer is 331 .
Round 4 Solutions
www.usamts.org
2/4/21. Let a, b, c, d be four real numbers such that
a + b + c + d = 8,
ab + ac + ad + bc + bd + cd = 12.
Find the greatest possible value of d.
We factor d out of the second equation, and use the fact (from the first equation) that
a + b + c = 8 d:
12 = ab + ac + bc + d(a + b + c) = ab + ac + bc + d(8 d). (?)
However, we know that a2 + b2 + c2 ab + ac + bc: this is an application of what is

known as the Rearrangement Inequality, and it also follows by expanding and simplifying
(a b)2 + (b c)2 + (c a)2 0. Thus, we have
(a + b + c)2 = a2 + b2 + c2 + 2(ab + ac + bc) 3(ab + ac + bc), (??)
giving
1 1
ab + ac + bc (a + b + c)2 = (8 d)2 .
3 3
Therefore, (?) becomes
1
12 (8 d)2 + d(8 d),
3
2
which simplifies to d 4d 14 0. The largest value
of d satisfying this inequality is the
2
larger root of d 4d 14 = 0, which is d = 2 + 3 2. We verify that equality holds in (??)
when a = b = c, giving
1 1
a = b = c = (8 d) = (6 3 2) = 2 2,
3 3
and it is easily checked that

ab+ac+ad+bc+bd+cd = 3(2 2)2 +3(2 2)(2+3 2) = 3((64 2)+(2+4 2)) = 12,

as necessary. Therefore, the answer is 2 + 3 2 .
Round 4 Solutions
www.usamts.org
3/4/21. I give you a deck of n cards numbered 1 through n. On each turn, you take the top
card of the deck and place it anywhere you choose in the deck. You must arrange the cards
in numerical order, with card 1 on top and card n on the bottom. If I place the deck in a
random order before giving it to you, and you know the initial order of the cards, what is the
expected value of the minimum number of turns you need to arrange the deck in order?
In any arrangement, there will be k cards at the bottom (where 1 k n) that 3

are in ascending numerical order, such that the (k + 1) cards at the bottom (if k < n) 6
are not in ascending numerical order. (If k = n then the entire deck is already in 5
correct order, and 0 turns are required.) For example, if the deck is arranged as at 1
right (for n = 6), then k = 3 since the 3 bottom cards are in ascending order1, 2, 4 2
but the 4 bottom cards5, 1, 2, 4are not. Since the (k + 1)st card must be moved 4
to its correct position, all of the cards above the bottom k cards must be moved at
least once. We claim that this is sufficient: on the ith turn, move the top card to its correct
position among the bottom k 1 + i cards (which by assumption are already in the proper
order). Thus n k turns are required. For example, below we show the sequence of moves
for the deck at right: note that after i turns the bottom 3 + i cards are in order, but the card
immediately above this group is not (the cards at the bottom in correct order are boxed):
3 6 5 1
6 5 1 2
5 1 2 3

1 2 3 4
2 3 4 5
4 4 6 6
Let En be the expected number of turns required for a deck of size n. We complete the
solution via two different methods:
Method 1 : Unless the deck is already in correct order, we must move the top card. What
remains, disregarding the card that we just moved (which is now in its correct position in
the group at the bottom of the deck) is a deck of size n 1, which will take an expected En1
turns to arrange correctly. So we require an expected 1 + En1 number of turns. However,
there is a n!1 probability that the deck will start in its correct order, in which case we will
not even require the first turn. Thus, we have
1
En = 1 + En1 .
n!
Starting with E1 = 0, we have
n n
X 1 X 1
En = (n 1) = n .
k=2
k! k=1
k!
Round 4 Solutions
www.usamts.org
Method 2 : Suppose the k cards at the bottom are in ascending numerical order, but the
(k + 1) cards at the bottom are not, so that n k turns are required. To count the number
n

of such initial arrangements, we note that for each of the k+1 choices of k + 1 cards, there
are k ways to order the cards such that the final k are in ascending order, but the final
k + 1 are not (all such arrangements can be formed by starting with the k + 1 cards in order,
then moving one of the final k cards to the top of this group of k + 1 cards). There are
then (n (k + 1))! ways to order the remaining cards atop these k + 1 cards. So, there are
n
k(n (k + 1))! k+1 such arrangements, and we require n k turns for each.
Summing over all possible k, we have
n1 n1
X n X n!
(n k)k(n (k + 1))! (n k)k n1
k+1 (k + 1)! nk k 2
En = k=1 = k=1
X
= .
n! n! k=1
(k + 1)!
We can write this in a more appealing form with some manipulation that leads to telescoping
series:
n1
X nk k 2
En =
k=1
(k + 1)!
n1
X n(k + 1) n k(k + 1) + k
=
k=1
(k + 1)!
n1
Xn(k + 1) n k(k + 1) + (k + 1) 1
=
k=1
(k + 1)!
n1 X n1 Xn1
X n n 1 1 1
= + + .
k=1
k! (k + 1)! k=1
(k 1)! k! k=1
(k + 1)!
The first two summations telescope, and we are left with
n1
n n 1 1 X 1
En = +
1! n! 0! (n 1)! k=1 (k + 1)!
n1
1 1 X 1
=n 1+
(n 1)! (n 1)! k=1 (k + 1)!
n
X 1
= n .
k=1
k!
Note: there is no closed form for the above summation, but, using calculus, we can show
that En is very close to n + 1 e as n gets large, where e 2.71828 . . . is the base of the
natural logarithm.
Round 4 Solutions
www.usamts.org
4/4/21. Let S be a set of 10 distinct positive real numbers. Show that there exist x, y S
such that
(1 + x)(1 + y)
0<xy < .
9
Let the set be {x1 , x2 , . . . , x10 }, where 0 < x1 < x2 < < x10 . Let
9
ai = ,
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer xi + 1
9 ai
so 0 < a10 < a9 < < a1 < 9 and xi = .
ai
Let i < j. Then xi < xj , and
9 9
(1 + xi )(1 + xj ) 9 aj 9 ai ai
aj
xj x i
9 aj ai 9
9ai 9aj 9

ai aj ai aj
ai aj 1.
Suppose that no such x and y exist that satisfy the given condition. Then a1 a2 > 1,
a2 a3 > 1, . . . , a9 a10 > 1. Adding these inequalities, we get a1 a10 > 9. But a1 < 9
and a10 > 0, so a1 a10 < 9, a contradiction. Therefore, such x and y must exist.
Round 4 Solutions
www.usamts.org
5/4/21. Tina and Paul are playing a game on a square S. First, Tina selects a point T inside S.
Next, Paul selects a point P inside S. Paul then colors blue all the points inside S that are
closer to P than T . Tina wins if the blue region thus produced is the interior of a triangle.
Assuming that Paul is lazy and simply selects his point at random (and that Tina knows
this), find, with proof, a point Tina can select to maximize her probability of winning, and
compute this probability.
Let the square be ABCD, and for simplicity assume it has D C

side length 1. Let Tinas point be T and Pauls be P , and
let the perpendicular bisector of P T be `. All points on the P
same side of ` as T are closer to T than to P , and conversely
all points on the same side of ` as P are closer to P than to T
T . Thus, the blue region is the intersection of the interior of
the square with the half-plane containing P determined by `. `
This is a triangle if and only if exactly 1 of {A, B, C, D} is
in this half-plane, meaning that Tina wins if and only if T is
closer than P to three of the four vertices of the square. This A B
occurs if and only if T is closer than P to either pair of opposite vertices (either A and C,
or B and D).
Consider the circle centered at A with radius AT and

the circle centered at C with radius CT ; we call these D WD (T ) C
A (T ) and C (T ). The point T is closer than P to both
A and C if and only if P is outside both circles, since
T0
such a location of P gives P A > T A and P C > T C.
We call the two regions inside the square, but outside T
both circles, winning regions, because Tina wins if
Paul chooses a point in either of these regions. There
WB (T )
is a winning region WB (T ) of points P for which P is
closer to B than T but where T is closer to each of A,
C, and D than P , as shown in the diagram. Similarly, A B
there is a winning region WD (T ). Because the diagram
is symmetric about AC, we have WB (T ) = WD (T ). There is a similar pair of winning regions
WA (T ) and WC (T ) in which T is closer than P to both B and D.
We claim that Tina maximizes the total area of the winning regions by choosing T to be
the center of the square. We show this separately for the pair A, C and the pair B, D, in
two steps.
For any choice of T not on AC, let T 0 be the intersection of AC and A (T ). We claim
that WB (T 0 ) WB (T ) and WD (T 0 ) WD (T ). This is clear because C (T 0 ) lies inside
Round 4 Solutions
www.usamts.org
C (T ), and hence the winning regions of T 0 are larger than those of T , since they contain an
additional region that is inside C (T ) but outside C (T 0 ). Thus, we conclude that for any
point T not on AC, there is a point T 0 AC such that the total area of WB (T 0 ) and WD (T 0 )
is greater than the total area of WB (T ) and WD (T ). Therefore, the point that maximizes
the winning regions adjacent to B and D must lie on the diagonal AC.
Next, we show that of all points on AC, Tina maxi-

mizes the area of the winning regions WB (T ) and WD (T ) D WD (T ) C
by choosing the center of the square. Note that if T
AC,for then
Create PDF with GO2PDF free, if youthe
wish toarea ofline,the
remove this winning
click here to buy Virtualregions
PDF Printer is the entire
area of the square minus the areas of the portions of

T
A (T ) and C (T ) lying inside the square (the white ar-
eas in the diagram). So our goal is to minimize the white
areas. If AT and CT are both less than or equal to 1,
WB (T )
this white region is two quarter-circles whose radii sum
to the diagonal of the square, which is 2. Thus, if
the
radius of A (T ) is r, then the radius of C (T ) is A B
2 r, and the combined areaof the white regions is
2
(r + ( 2 r)2 ) = 4 (2r2 2 2r + 2). This quadratic is minimized at r = 2
, which is
4 2
when each radius is half the diagonal, or 2/2, giving a white area of
!2
1 2
2 = .
4 2 4
If AT > 1 or CT > 1, then the corresponding white region is not a quarter circle; however,
in this case the white region is larger than a quarter-circle of radius 1, which has area 4 , so
the white region is still larger than the minimum area.
Since the center point of the square maximizes the sum of the areas of the winning regions
WB (T ) and WD (T ), and by symmetry simultaneously maximizes the sum of the areas of the
winning regions WA (T ) and WC (T ), and these regions are all disjoint, we conclude that the
center point is the optimal point for Tina to select. As computed above, the regions WB (T )
and WD (T ) have a total area of 1 4 , as do the regions WA (T ) and WC (T ), so the total
winning area for Tina is 2(1 4 ) = 2 2 . Thus, since the total area of the square is 1, the

probability of Tina winning is just the total area of the winning regions, or 2 .
2

Problem 2/4/21 was submitted by Phan Van Thuan, Vietnam National University.
Problem 3/4/21 was originally suggested by Matt Superdock.

Round 1 Solutions
www.usamts.org
1/1/22.
Given a set S of points in the plane, a line is called happy if it contains at

least 3 points in S. For example, if S is the 3 3 grid of points shown at right,
then there are 8 happy lines as shown.
(a) If S is the 3 9 grid shown below, how many happy lines are there?
(b) Find, with proof, a set S (in the plane) with 27 points that has exactly 49 happy lines.
(a) We count the happy lines in sets. First notice that there are three horizontal happy
lines.
Any other happy line must contain a point in the second row. We count the remaining
happy lines by counting how many pass through a given middle point. There is exactly
one happy line through the leftmost point.
There are three happy lines through the second point.
There are five happy lines through the third point.
The other points have 7, 9, 7, 5, 3, and 1 happy lines respectively.

Round 1 Solutions
www.usamts.org
The total number of happy lines in the diagram is
3 + (1 + 3 + 5 + 7 + 9 + 7 + 5 + 3 + 1) = 44 .
(b) There are many solutions to this problem. For instance, we can take a 3 3 3 cube
of dots in space and project its 33 = 27 vertices onto the plane in such a way that all of
the lines in space which contain three vertices are taken to distinct lines in the plane
and no new lines are created 1 . We let S be the projection and count the happy lines.
First we may count the lines which all lie on horizontal planes. This gives 24 lines.
The remaining lines must intersect each of the three horizontal planes one time, so
must contain one vertex on each plane. We enumerate these by considering which of
the 9 vertices are on the line. By the center square we will mean the nine vertices
on the central horizontal plane. Each of the four corner vertices of the central square
meet only one non-horizontal line (in red, below). Each of the four edge vertices of the
center square are found on three non-horizontal lines (in blue below).
Finally, there are nine non-horizontal lines through the center vertex.
This gives a total of 24 + 4 1 + 4 3 + 9 = 49 happy lines.
1
We can always do this: the noncollinear subsets of 3 points from our 27 points generate a set of at most
27
3 planes. If we project to a plane that is not perpendicular to any plane in this set, we will not create
any new lines. Further, if we project to a plane that is not perpendicular to any of our existing lines, we will
not lose any lines upon projection.
Round 1 Solutions
www.usamts.org
2/1/22. Youre at vertex A of triangle ABC, where B = C = 65 . The sides of the triangle
are perfectly reflective; if you shoot a laser from A to the midpoint of BC, it will reflect
once and return to A. Suppose you fire at a point on BC other than its midpoint, and the
beam still returns to A after reflecting some number of times. What is the smallest number
of reflections the beam can make before returning to A? What is the smallest angle between
AB and the initial beam that produces this number of reflections?
We label the edge opposite A as a, the edge opposite B as b and the edge opposite C as
c. We will measure the initial angle from AB, where 0 < < 50 .
Since the room is perfectly reflective, standing at point A we see an infinite region in
which many copies of A, B, and C appear. We are looking for the minimum number of walls
that the beam may cross before returning to A. Furthermore, from the point of view of the
beam, it is traveling in a straight line and passing through a number of consecutive isosceles
chambers.
Notice that the solution set is symmetric, so any closed path with
angle > 25 is the horizontal reflection of a closed path with angle
50 . Therefore we may assume that the path which passes the
fewest faces meets a and then c. The sum of the three acute angles
at B is 3 65 = 195 > 180 , so the beam must subsequently pass
through edge b. The shaded region shows where the beam may pass.
The continuation of the beams possible paths at left includes

the set of all chambers that the beam can visit in five or fewer
reflections with an initial angle less that 25 . Since there is one
reflection of A other than A0 , the minimum number of reflections
must be five. Since B and C 0 are both reflex angles, the beam
cannot reach A00 .
Round 1 Solutions
www.usamts.org
We now know that the minimum number of reflections is 5 and is

achieved by the path at left. Let M denote the midpoint of the beam
AA00 . By symmetry M lies on CA0 , and the beam intersects this wall
with angle 90 . Now consider the quadrilateral AM C 0 B. This quadri-
lateral has total angle
360 = A + 90 + 65 + (3 65 ) = A + 350 .
Therefore we must fire the beam at an angle 10 in order to return in

the
Create PDF with GO2PDF for free, if you wish to remove this fewest
line, click number
here to buy of reflections (other than firing at the midpoint of
Virtual PDF Printer
BC).
Round 1 Solutions
www.usamts.org
3/1/22. Find c > 0 such that if r, s, and t are the roots of the cubic
f (x) = x3 4x2 + 6x + c,
then
1 1 1
1= + 2 + 2 .
r2 +s 2 s +t 2 t + r2
We know that x3 4x2 + 6x + c = (x r)(x s)(x t), therefore

r + s + t = 4,
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer (1)
rs + rt + st = 6, (2)
rst = c. (3)
Next we compute the sum of the squares of the roots using (1) and (2) above:
r2 + s2 + t2 = (r + s + t)2 2(rs + rt + st) = 42 2(6) = 4. (4)
Thus, using (4), our identity becomes
1 1 1
1= 2 + +
r + s2 s2 + t2 t2 + r2
1 1 1
= 2
+ 2
+
4t 4r 4 s2
(4 r )(4 s ) + (4 s2 )(4 t2 ) + (4 t2 )(4 r2 )
2 2
=
(4 r2 )(4 s2 )(4 t2 )
48 8(r2 + s2 + t2 ) + r2 s2 + r2 t2 + s2 t2
= . (5)
64 16(r2 + s2 + t2 ) + 4(r2 s2 + r2 t2 + s2 t2 ) r2 s2 t2
Substituting using (3) and (4) gives
48 8(4) + r2 s2 + r2 t2 + s2 t2 16 + r2 s2 + r2 t2 + s2 t2
1= = . (6)
64 16(4) + 4(r2 s2 + r2 t2 + s2 t2 ) c2 4(r2 s2 + r2 t2 + s2 t2 ) c2
Next, observe that by (2)
36 = (rs + rt + st)2 = r2 s2 + r2 t2 + s2 t2 + 2rst(r + s + t) = r2 s2 + r2 t2 + s2 t2 8c,
so that
r2 s2 + r2 t2 + s2 t2 = 36 + 8c. (7)
Combining (6) and (7) gives
16 + (36 + 8c) 52 + 8c
1= 2
= ,
4(36 + 8c) c 144 + 32c c2
and hence c2 24c 92 = 0. The solutions to this quadratic are

242 + 4(92) 24
c=
2

and the only positive solution is c = 12 + 2 59 .
Round 1 Solutions
www.usamts.org
4/1/22. Sasha has a compass with fixed radius s and Rebecca has a compass with fixed radius r.
Sasha draws a circle (with his compass) and Rebecca then draws a circle (with her compass)
that intersects Sashas circle twice. We call these intersection points C and D.
Charlie draws a common tangent to both circles, meeting Sashas circle at point A and
Rebeccas circle at point B, and then draws the circle passing through A, B, and C. Prove
that the radius of Charlies circle does not depend on where Sasha and Rebecca choose to
draw their circles, or which of the two common tangents Charlie draws.
Let Sashas circle be S and Rebeccas be R. Once S and R

R are chosen, Charlie has two choices for the common tangent.
Suppose he chooses the tangent such that the tangent is closer D
S
to C than to D, as shown in the diagram at the right. Had he
chosen the other tangent, the resulting circumcircle of 4ABC
would be congruent to the shown circumcircle of 4ABD, by
symmetry. So, we will show that Charlies two options for the C B
common tangent produce the same circumradius by showing A
that the circumradii of 4ABC and 4ABD are equal. We do
so by showing that the circumcircle of 4ABC is the reflection

of the circumcircle of 4ABD over AB.
D0
Clearly the reflections of A and B over AB are on the cir-

cumcircle of 4ABD. We need only show that the reflection of D over AB, which we call
D0 , is on the circumcircle of 4ABC. Since angles ADC and CAB are both inscribed in
of S, we have ADC = CAB. Similarly, we have CDB = CBA because these
arc AC
of R. Therefore, we have
two angles are inscribed in arc CB
ADB + ACB = ADC + CDB + ACB = CAB + CBA + ACB = 180 ,
so ADB is supplementary to ACB, which means AD0 B and ACB are supplementary.

Since D and C are on the same side of AB, points D0 and C are on opposite sides of AB.
Combining this with the fact that AD0 B +ACB = 180 , we know that AD0 BC is a cyclic
quadrilateral, so D0 is on the circumcircle of 4ACB.
Round 1 Solutions
www.usamts.org
Let Charlies circle be E. Let X, Y , Z be the centers of R

S, R, E, respectively. We will find the radius of E in terms
of r and s. Since CAB is inscribed in AC of S, we have D
S Y
CXA = AC = 2CAB. Similarly, since CAB is inscribed
of E, we have CZB = BC
in BC = 2CAB. Therefore, X
we have CZB = CXA, so CZY = CXZ (because ZY
bisects CZB and XZ bisects CXA). Similarly, we have C B
BY C = 2CBA = CZA, so CY Z = CZX. Combining A
CZY = CXZ and CY Z = CZX, we have 4CXZ
4CZY . Therefore, we have CX CZ
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy VirtualCY
= CZ , so
PDF Printer
CZ 2 = CX CY = Z
rs, which means that the radius of E is rs. We conclude
that Charlies circle has radius rs no matter where Sasha and E
Rebecca locate their circles (as long as the circles intersect).
Round 1 Solutions
www.usamts.org
5/1/22. A convex polygon P is called peculiar if: (a) for some n 3, the vertices of P are a
subset of the vertices of a regular n-gon with sides of length 1; (b) the center O of the n-gon
lies outside of P ; and (c) for every integer k with 0 < k n2 , the quantity 2k
n
is the measure
of exactly one AOB, where A and B are vertices of P . Find the number of non-congruent
peculiar polygons.
We notice first that if n is even, then letting k = n2 tells us we need

one central angle of 2n = . However if this is the case then the diameter
Create PDF with GO2PDF for free, if you wish to remove thiskline, click here to buy Virtual PDF Printer
between the two corresponding points is a subset of the peculiar polygon,
so the center of the n-gon lies in P , which violates condition (b). Therefore
n must be odd. First we give two examples of peculiar polygons. To the
right we have a peculiar triangle for n = 7, and below we have a peculiar
quadrilateral for n = 13.
Let v be the number of vertices of P , and let r = n1 2

(note r is an
integer because n is odd). Each pair of vertices describes a unique central
angle less than . Since this set of angles is

2 1 2 2 2 r
, ,..., ,
n n n

v
there are r such angles. Therefore 2
= r, or
n = v 2 v + 1.
We intend to show that there are no peculiar polygons beyond the two given above.
Assume we have a peculiar polygon. This polygon must contain

a pair of vertices, A and B, of maximal central angle AOB =
(n1)
n
. The chord AB separates the remaining vertices of the n-gon B
A
into two sets of size r 1 and r. If any points from the larger set are
also vertices of P , then the center will be interior to P . Therefore,
all vertices of P must lie on the short side of AB. In particular, AB
is an edge of P .
We now relabel the vertices from A to B as A = A0 , A1 , . . . , Ar = B. Next we look for a
pair of vertices forming an angle of (n3)
n
. Such a pair is only possible if one member of the
pair is either A0 or Ar . We may assume this is A0 by congruence. Then, the other vertex
must be Ar1 . Therefore P must have vertices A0 , Ar1 , and Ar .
In the case v = 3, these three vertices above describe the peculiar triangle. We now
assume that v > 3.
Round 1 Solutions
www.usamts.org
Notice that, by the uniqueness condition of property (c), the points A1 and Ar2 cannot

be vertices of P since A
0 A1 and Ar2 Ar1 are both congruent to Ar1 Ar .
(n5)
The only arcs of angle n
that lie above AB are

A
0 Ar2 , A1 Ar1 , and A2 Ar .
Since A1 and Ar2 are not vertices of P , we have that A2 must be a vertex. Therefore, P
must have the vertices A0 , A2 , Ar1 , and Ar . If v = 4 then n = 42 4 + 1 = 13 gives the
peculiar quadrilateral above.
Now we assume that v > 4. We have the four vertices from

above, and these four vertices span the six angles Ar1 A2
2 4 (n 7) (n 5) (n 3) (n 1) Ar
, , , , , and . A0
n n n n n n
We claim that there is no way to add a vertex (or more) in
order to get the angle (n9)
n
without creating another angle
we have already constructed.
(n9)
The only arcs of angle n
which lie above A0 Ar are

A
0 Ar4 , A1 Ar3 , A2 Ar2 , A3 Ar1 , and A4 Ar .
The only new vertex on this list not at an angle 2

n
or 4
n
from an existing vertex is Ar4 .
Therefore any peculiar polygon must contain this point as well.
If v = 5 then r = 10, so the vertices are A0 , A2 , A6 , A9 , and A10 . However this is not
a possible peculiar polygon as the angle from A2 to A6 is the same as that from A6 to A10 .
There is no peculiar pentagon.
Assume that v > 5 and that we have the vertices A0 , A2 , Ar4 , Ar1 , and Ar . We need
to place a point to give an angle of (n9)
n
. The only possiblilites are

A
0 Ar5 , A1 Ar4 , A2 Ar3 , A3 Ar2 , A4 Ar1 , A5 Ar .
None of these pairs is a subset of the current vertices since

v2 v 52 5
r5= 5> 5 = 5.
2 2
Furthermore, each of the new potential vertices is an angle of 2
n
, 4
n
, or 6
n
from an existing
vertex, so is not permissible. Therefore there are no peculiar polygons with 5 or more
vertices.
By our construction, there is exactly one peculiar triangle, one peculiar quadrilateral,
and zero other peculiar polygons, up to congruence. Therefore, the answer is 2 .
Round 1 Solutions
www.usamts.org
6/1/22. There are 50 people (numbered 1 to 50) and 50 identically wrapped presents around
a table at a party. Each present contains an integer dollar amount from $1 to $50, and no
two presents contain the same amount. Each person is randomly given one of the presents.
Beginning with player #1, each player in turn does one of the following:
1. Opens his present and shows everyone the contents; or
2. If another player at the table has an open present, the player whose turn it is may
swap presents with that player, and leave the table with the open present. The other
player then immediately opens his new present and shows everyone the contents.
For example, the game could begin as follows:
Player #1 opens his present. (The game must always begin this way, as there are no
open presents with which to swap.)
Player #2 decides to swap her present with Player #1. Player #2 takes the money
from her newly acquired present and leaves the table. Player #1 opens his new present
Player #3 opens her present. (Now Players #1 and #3 have open presents, and Player
#2 is still away from the table.)
Player #4 decides to swap his present with Player #1. Player #4 takes the money
from his newly acquired present and leaves the table. Player #1 opens his new present
The game ends after all the presents are opened, and all players keep the money in their
currently held presents.
Suppose each player follows a strategy that maximizes the expected value that the player
keeps at the end of the game.
(a) Find, with proof, the strategy each player follows. That is, describe when each player
will choose to swap presents with someone, or keep her original present.
(b) What is the expected number of swaps?
(a) At all times, each player knows the set of amounts in the unopened presents. We
prove the following:
Lemma: The optimal is strategy is: if the largest opened present in play (that
is, available to be swapped) is greater than the amount of at least one unopened
Round 1 Solutions
www.usamts.org
present, the player whose turn it is should swap his present for the largest opened
present in play. Otherwise, he should open his present. If all players follow this
strategy, then any player who opens his own present will end the game with the
lowest amount of those unopened at the time of his turn.
Proof of Lemma: We prove by induction, in reverse order of the players numbers.

Base case: Player #50 will clearly benefit by following the strategy: if there is a present
in play that is larger than the final unopened present, then she should swap for the larger
amount.
Create PDF with GO2PDF Otherwise,
for free, if you theclickunopened
wish to remove this line, present
here to buy Virtual PDF Printer is the largest amount available to her, so she
should open it: the game will immediately end and she will get to keep the amount that she
just opened.
Inductive step: Let 1 k < 50 be an integer, and assume that it is Player #ks turn,
and that (by inductive hypothesis) Players #(k + 1) through #50 will follow the strategy as
outlined in the Lemma.
If all the available amounts are less than all the unopened amounts, then Player #k
will do strictly worse by swapping than by opening: swapping is guaranteed to result in a
lower amount than opening. So Player #k should open his present; after this, all subsequent
players will swap with Player #k as long as there is a smaller-valued present remaining
unopened, hence Player #k is guaranteed to be stuck with the smallest remaining unopened
amount at the end of the game.
On the other hand, suppose that Player #m, with m < k, holds an opened present with
an amount M that is larger than some amount(s) that are currently unopened. If Player
#k chooses the opposite strategy to the Lemmas, and opens his present, there are two
possibilities:
Case 1: Player #k opens an amount less than M . Players will only swap with Player
#k if he has the largest opened present and there is an unopened present with smaller value.
Once they start swapping with Player #k, they will keep swapping until Player #k either
receives that unopened present or is no longer the largest opened present; in either case
Player #ks value will decrease, and will still be less than M .
Case 2: Player #k opens an amount greater than M . Then since there is still an unopened
present of value less than M , all subsequent players will follow the strategy and swap with
Player #k, until Player #k is stuck with a present with amount less than M , at which point
we revert to Case 1.
In any event, Player #k ends up with a final amount less than M if he does not swap,
whereas he ends up with M if he swaps. So he should swap.
This completes the proof. 2
(b) Based on the Lemma, there is no swap if and only if all of the opened presents are
less than all of the unopened presents. (Note this includes Player #1s non-swap.) So we
Round 1 Solutions
www.usamts.org
count the expected number of non-swaps and subtract from n.

Player #1s turn is always a non-swap, and the next non-swap occurs after the player
who holds (and swaps away) the present with $1. The next non-swap occurs after the player
holding the minimum value of all those remaining, and so on. This allows us to set up a
recurrence.
Let Ek denote the expected number of non-swaps in the k-player version of the game.
Note that E0 = 0 and E1 = 1. If the $1 present is held by person #j, then there are Ekj
swaps expected after the initial swap. Therefore, since the $1 present is equally likely to be
held by any player, the recurrence is
1 1 k1
X
Ek = 1 + (Ek1 + Ek2 + + E1 + E0 ) = 1 + Ej .
k k j=0
We prove by induction that
k
1 1 1 X 1
Ek = 1 + + + + = .
2 3 k j=1 j
1
It suffices to show that Ek Ek1 = k
(since E1 = 1) but this follows from:

1 k1
X 1 k2
X
Ek Ek1 = 1+ Ej 1+ Ej
k j=0 k 1 j=0
k2
1 1 1 X
= Ek1 + Ej
k k k1 j=0
k2
X
1 1
= Ek1 Ej
k k(k 1) j=0

1 1 1 k2
X
= Ek1 1+ Ej 1
k k k 1 j=0
1 1
= Ek1 (Ek1 1)
k k
1
= ,
k
proving the formula.
Thus, the expected number of swaps in the 50-person game is 50 minus the expected
number of non-swaps, which is

1 1 141008987635075780359241
50 E50 = 50 1 + + + = 45.501.
2 50 3099044504245996706400

Round 1 Solutions
www.usamts.org
1/2/22. Show that there is a unique way to place positive integers in the grid
to the right following these three rules:
1. Each entry in the top row is one digit.

2. Each entry in any row below the top row is the sum of the two
entries immediately above it.
3. Each pair of same-color squares contain the same integer. These five distinct integers
are used exactly twice and no other integer is used more than once.
Label additional squares as shown at right. Comparing the two x y z

pink boxes, we have
= + 3 = y + z 17,

since y and z are distinct and each a single digit. Thus 5. Next,

comparing the two orange boxes, we see that
= x + 3 = + ,
so x = 2. But 5, so in order to have x > 0 we must have equal to 1 or 2. The
only possibilities are thus
(, ) {(1, 3), (1, 4), (1, 5), (2, 5)}.
The first of these gives x = 1, which is disallowed since then x and are not distinct. The
second gives x = 2, but this is disallowed because then the box directly below the two blue
boxes would also be 2. So we conclude that = 5 and that (, x) {(1, 3), (2, 1)}.
If we attempt to use = 2 and x = 1, we can sum downwards to get the grid on the left
below. However, we then also have 46 in the green box to the right of the box with 28, and
subtracting upwards gives the grid on the right below; this makes y = 14 in the top row,
which violates condition 1. So this case cannot occur.
1 2 2 5 5 y z 1 2 2 5 5 14 z
3 4 7 10 3 4 7 10 19
7 11 17 7 11 17 29
18 28 18 28 46
46 46
Round 1 Solutions
www.usamts.org
The only remaining possibility is the case = 1, x = 3, which will fill in at left below.
Summing downwards gives = 38. We can place 38 in the other green box and subtract
upwards, and then complete the picture as shown at right below. This completed picture
satisfies all the requirements, and it is the only possibility.
3 1 1 5 5 y z 3 1 1 5 5 7 9
4 2 6 10 4 2 6 10 12 16
6 8 16 6 8 16 22 28
14 24 14 24 38 50
38 38 62 88
100 150
250
Round 1 Solutions
www.usamts.org
2/2/22. A sequence is called tworrific if its first term is 1 and the sum of every pair of
consecutive terms is a positive power of 2. One example of a tworrific sequence is 1, 7, 5,
7, 57.
(a) Find the shortest possible length of a tworrific sequence that contains the term 2011.
(b) Find the number of tworrific sequences that contain the term 2011 and have this
shortest possible length.
(a) We claim that the shortest possible length of a tworrific sequence that contains the
term 2011 is 5. First, the sequence
1, 7, 5, 37, 2011
contains the term 2011 and has length 5. The sums of pairs of consecutive terms are 1+7 = 8,
7 + (5) = 2, (5) + 37 = 32, and 37 + 2011 = 2048, so this sequence is tworrific.
Let
a1 , a2 , a3 , . . . , ak
be a tworrific sequence of length k ending with the term 2011, so a1 = 1 and ak = 2011.
Then
a1 + a2 = 2e1 ,
a2 + a3 = 2e2 ,
...,
ak1 + ak = 2ek1
for some nonnegative integers e1 , e2 , . . . , ek , so
ak = 2ek1 ak1
= 2ek1 2ek2 + ak2
= 2ek1 2ek2 + 2ek3 ak3
=
= 2ek1 2ek2 + 2ek3 + (1)k 2e1 + (1)k+1 a1
= 2ek1 2ek2 + 2ek3 + (1)k 2e1 + (1)k+1 .
We claim that that there are no tworrific sequences of length 1, 2, 3, or 4 that contain the
term 2011.
Case 1: The tworrific sequence has length 1.

Round 1 Solutions
www.usamts.org
This case is trivial, because the first term is always 1.
In this case, we seek a nonnegative integer e1 such that
2e1 1 = 2011.
Then 2e1 = 2012. But 2012 = 22 503 is not a power of 2, so there is no such nonnegative
integer e .
Create PDF with GO2PDF for free, if1you wish to remove this line, click here to buy Virtual PDF Printer
In this case, we seek nonnegative integers e1 and e2 such that
2e2 2e1 + 1 = 2011.
Then
2e2 2e1 = 2010.
Clearly e2 > e1 , so we can write
2e1 (2e2 e1 1) = 2010 = 2 1005.
The factor 2e1 is a power of 2, and the factor 2e2 e1 1 is odd, so we must have 2e2 e1 1 =
1005, which means 2e2 e1 = 1006. But 1006 = 2 1003 is not a power of 2, so there are no
such nonnegative integers e1 and e2 .
In this case, we seek nonnegative integers e1 , e2 , and e3 such that
2e3 2e2 + 2e1 1 = 2011.
Then
2e3 + 2e1 = 2e2 + 2012.
Note that a nonnegative integer (that is at least 2) can be written in the form 2e3 + 2e1 if
and only if it has exactly one or two 1s in its binary representation. (We get one 1 if e3 = e1 ,
which makes 2e3 + 2e1 = 2e1 +1 .) Thus, we seek a nonnegative integer e2 such that the binary
representation of 2e2 + 2012 has one or two 1s.
But the binary representation of 2012 is
2012 = 111110111002 .
Round 1 Solutions
www.usamts.org
It is easy to check that no value 0 e2 10 will work, and for e2 11, the binary
representation of 2e2 + 2012 is
1 |00 {z
. . . 0} 111110111002 .
e2 11 0s
Hence, there are no such nonnegative integers e1 , e2 , and e3 .
Therefore, the shortest possible length of a tworrific sequence that contains the term 2011
is 5.
(b) Finding all tworrific sequences of length 5 that contain the term 2011 is equivalent
to finding all nonnegative integers e4 , e3 , e2 , and e1 such that
2e4 2e3 + 2e2 2e1 + 1 = 2011,
or
2e4 + 2e2 = 2e3 + 2e1 + 2010.
Hence, the binary representation of 2e3 + 2e1 + 2010 must have exactly one or two 1s.
The binary representation of 2010 is

2010 = 111110110102 .
Checking all values where 0 e1 , e3 10, we find that the binary representation of 2e3 +
2e1 + 2010 has exactly one or two 1s only for (e1 , e3 ) = (3, 5) and (5, 3), for which
2e3 + 2e1 + 2010 = 1000000000102 .
Hence, (e2 , e4 ) = (1, 11) or (11, 1).
For e1 , e3 11, the binary representation of 2e3 + 2e1 + 2010 is

. . . 0} 111110110102
1 |00 {z
e1 10 0s
if e1 = e3 , and
1 |00 {z
. . . 0} 1 |00 {z
. . . 0} 111110110102
e1 e3 1 0s e3 11 0s
if e1 > e3 (and similarly for e1 < e3 ).
Hence, there are four quadruples of nonnegative integers (e1 , e2 , e3 , e4 ), namely (3, 1, 5, 11),
(3, 11, 5, 1), (5, 1, 3, 11), and (5, 11, 3, 1). This gives us the four tworiffic sequences
1, 7, 5, 37, 2011,
1, 7, 2041, 2009, 2011,
1, 31, 29, 37, 2011,
1, 31, 2017, 2009, 2011.
Round 1 Solutions
www.usamts.org
3/2/22. Richard, six of his friends, and a Gortha beast are standing at different vertices of a
cube-shaped planet. Richard has a potato and is a neighbor to the Gortha. On each turn,
whoever has the potato throws it at random to one of his three neighbors. If the Gortha
gets the potato he eats it. What is the probability that Richard is the one who feeds the
Gortha?
The planet is shown at right, where R is Richard and G is the q r

Gortha. Let p be the probability that Richard feeds the Gortha
given that he is holding the potato. For each non-Gortha neighbor
of Richard, let q be the probability that Richard feeds the Gortha R G
given that the neighbor is holding the potato. (By symmetry, it p
is the same probability for each neighbor.) For each non-Richard s
neighbor of the Gortha, let r be the probability that Richard feeds 1
3
the Gortha given that the neighbor is holding the potato. (Again,
by symmetry, it is the same probability for each of the Gorthas
other neighbors.) Finally, let s be the probability that Richard q r
feeds the Gortha given that the person opposite from Richard is holding the potato. Note
also that if the person who is opposite the Gortha has the potato, then the probability that
Richard feeds the Gortha is 31 , by symmetry, since each of the Gorthas three neighbors is
then equally likely to be the feeder. These probabilities are labeled on the diagram above.
When Richard has the potato, he has probability 13 of immediately feeding the Gortha,
and probability 23 of passing it to a neighbor. Therefore,
1 2
p= + q.
3 3
When one of Richards non-Gortha neighbors has the potato, she has probabilty 13 of throwing
the potato back to Richard, probability 31 of throwing the potato to one of the Gorthas other
neighbors, and probability 13 of throwing the potato to the person opposite the Gortha.
Therefore,
1 1 1
q = p+ r+ .
3 3 9
When one the Gorthas other neighbors has the potato, he has probability 13 of feeding the
Gortha (in which case Richard does not feed the Gortha), probability 13 of throwing the
potato to one of Richards neighbors, and probability 31 of throwing the potato to the person
opposite Richard. Therefore,
1 1
r = q + s.
3 3
Finally, when the person opposite Richard has the potato, she has probability 23 of throwing
it to one of the Gorthas non-Richard neighbors, and probability 13 of throwing it to the
Round 1 Solutions
www.usamts.org
person opposite the Gortha. Therefore,

2 1
s= r+ .
3 9
Thus we have the system of equations:

1 2
p= + q, (1)
3 3
1 1 1
q=
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer p + r+ , (2)
3 3 9
1 1
r = q + s, (3)
3 3
2 1
s= r+ . (4)
3 9
Substituting equation (4) into equation (3) gives:
1 2
p= + q, (5)
3 3
1 1 1
q = p+ r+ , (6)
3 3 9
1 2 1
r= q+ r+ . (7)
3 9 27
Equation (7) rearranges to give r = 37 q + 1
21
, so substituting this into (6) gives:
1 2
p= + q, (8)
3 3
1 1 8
q = p+ q+ . (9)
3 7 63
Clearing denominators gives
3p = 1 + 2q, (10)
54q = 21p + 8. (11)
7
Solving this system gives p = 12
and q = 38 , so the probability that Richard feeds the Gortha
7
is .
12
Round 1 Solutions
www.usamts.org
4/2/22. Let A, B, C, and D be points in the plane such that AD k BC. Let I be the incenter of
4ABC and assume that I is also the orthocenter of 4DBC. Show that AB +AC = 2BC.
Let a = BC, b = AC, and c = AB. Let s, r, and K denote the semiperimeter, inradius,
and area of triangle ABC, respectively. Let h be the common heights of triangles ABC and
DBC, with respect to base BC. Let T be the foot of the perpendicular from I to BC.
A D
I I
B T C B T C
Since I is the orthocenter of triangle DBC, CDT = 90 DCB = IBT , so triangles

IBT and CDT are similar. Hence,
CT DT
= .
IT BT
Since BT is the tangent from B to the incircle of triangle ABC, BT = s b. Similarly,
CT = s c, so
sc h
= .
r sb
Cross-multiplying, we get
(s b)(s c) = rh.
But r = K/s and h = 2K/a, so

2K 2
(s b)(s c) = ,
as
or
as(s b)(s c) = 2K 2 .
By Herons formula, K 2 = s(s a)(s b)(s c), so
as(s b)(s c) = 2s(s a)(s b)(s c).
Dividing both sides by s(s b)(s c), we get
a = 2(s a) = 2s 2a = b + c a,
so b + c = 2a, as desired.
Round 1 Solutions
www.usamts.org
5/2/22. Zara and Ada are playing a game. Ada begins by picking an integer from 1 to 2011
(inclusive). On each turn Zara tries to guess Adas number. Ada then tells Zara whether her
guess is too high, too low, or correct. If Zaras guess is not correct, Ada adds or subtracts 1
from her number (always constructing a new number from 1 to 2011). Assuming Zara plays
optimally, what is the minimum number of turns she needs to guarantee that she will guess
Adas number?
Zara needs a minimum of 1008 moves to guarantee a win.

Before demonstrating the formal proof, let us sketch the algorithm that Zara should
follow. At each guess, we will assume that Ada reacts in a way so as to maximize the
number of guesses Zara will need.
1. Zara guesses 1006 (the middle number). Without loss of generality assume Ada says
lower (if she says higher the remaining strategy is essentially the same).
2. Before Ada changes her number, Adas number must be in the set {1, 2, . . . , n}, where
n = 1005 initially.
3. After Ada changes her number, Adas number must be in the set {1, 2, . . . , n + 1}.
4. Zara guesses n.
5. If Ada says lower, then Adas number must be in the set {1, 2, . . . , n 1} (before
she changes her number), so we go back to Step 2 with n reduced by 1.
6. If Ada says higher, then continue to step 7 below with m = n + 1.
7. Before Ada changes her number, Adas number must be m.
8. After Ada changes her number, Adas number must be either m 1 or m + 1.
9. Zara guesses m + 1. We assume this is the wrong guess (otherwise Zara wins immedi-
ately).
10. Adas number must be m 1. If m 1 > 1, go back to Step 7 (with m reduced by 1).
11. Adas number must be 1, so when she changes it, her new number must be 2.
12. Zara guesses 2 and wins.
Notice that the above algorithm has four basic parts:
The initial guess of 1006.

Round 1 Solutions
www.usamts.org
A guess for each time we loop through steps 2-5. Each guess (except the last guess in
the loop) reduces n by 1, where n starts at 1005.
A guess for each time we loop through steps 7-10. Each guess (except the last guess
in the loop) reduces m by 1, where m starts at n + 1 for whatever value of n causes us
to end the first loop
The final guess of 2, which occurs after we have reduced to m = 2.
We run the two loops for a total of 1006 guesses, as we reduce from n = 1005 to m = 2:
each guess decreases the current variable (n or m by 1), except for the guess that shifts to
m = n + 1 somewhere in the middle, and the final guess of the second loop at m = 2. Adding
these 1006 guesses to the start and end guess gives us our total of 1008 guesses necessary.
Now we prove that this algorithm is the best we can do.
Let X = {1, 2, 3, . . . , 2011}. Call a subset S X an interval if it is of the form
S = {n | a n b}
for some a, b X with a b. We will denote this set as S = [a, b] (using the same notation
as we would use for an interval on the real line). For any n X, let
e(n) = min{x, 2012 x}
denote the distance from n from the boundary of X; that is, e(n) is the distance from n to
the closer of 0 or 2012. Note that e(1) = e(2011) = 1 and e(1006) = 1006.
For any nonempty subset S X, define
(
e(n) if S = {n} (that is, if S is the single element n),
E(S) = 2 + max(e(n)) if S has more than 1 element.
nS
(This value E will measure the least number of turns it takes for Zara to guess Adas number
from any of the sets At which interest us.)
Let t > 0 denote the turn (i.e. the number of guesses Zara has made). On each turn:
Zara guesses an integer zt X. If zt is Adas number, the game is over; otherwise Ada
replies higher or lower.
Zara gains information about Adas numbers. Let At denote the set of Adas possible
numbers before she adds or subtracts 1 to her number.
Ada adds or subtract 1 to her number. Let Bt denote the set of Adas possible numbers
after she adds or subtracts 1 to her number.
Round 1 Solutions
www.usamts.org
Note that if At is an interval with more than one element, then Bt is also an interval; in
particular if At = [a, b] then Bt = [a 1, b + 1] X. Also note that if Bt is an interval, then
either
At+1 = Bt {x X | x < zt }
or
At+1 = Bt {x X | x > zt }.
In either case At+1 is an interval. So, since B0 = X is an interval (that is, Adas number
could be anything at the beginning of the game), we have that At and Bt will always be
intervals until At becomes a set with a single element.
Lemma 1: If At is an interval with more than one element, then for any guess
zt Ada can reply so that E(At+1 ) E(At ) 1.
Proof of Lemma 1:
If 1006 At then 1005 At or 1007 At , and E(At ) = 1008. Also note that
[1005, 1007] Bt and at least one of 1004 and 1008 is also in Bt . Without loss of gen-
erality (by symmetry) we may assume Bt contains [1004, 1007]. Guessing anything other
zt = 1006 may result in [1005, 1006] or [1006, 1007] being a subset of At+1 . In this case
E(At+1 ) = 1008 = E(At ). Guessing zt = 1006 may result in [1004, 1005] At+1 . In this
case, E(At+1 ) = 1007 = E(At ) 1.
Otherwise we have 1006 6 At . Without loss of generality assume that At = [a, b] with
b < 1006, so that E(At ) = b + 2. (The case where a > 1006 is symmetric.) Then Bt =
[a 1, b + 1] (or [a, b + 1] if we have a = 1, but this does not affect our argument). If
zt is anything other than b or b + 1, then we may have [b, b + 1] At+1 , in which case
E(At+1 ) (b + 1) + 2 = b + 3 > E(At ).
If zt = b + 1 then (assuming this is not the correct guess) we have At+1 = [a 1, b], and
then E(At+1 ) = b + 2 = E(At ).
If zt = b then there are two possibilities. If Ada says lower then At+1 = [a 1, b 1],
and E(At+1 ) = 2 + (b 1) = b + 1 = E(At ) 1. If Ada says higher then At+1 = {b + 1}
and E(At+1 ) = b + 1 = E(At ) 1.
In all cases E(At+1 ) E(At ) 1, so this proves the Lemma.
2
Lemma 2: If At = {a} for some a, then Zara needs at least e(a) more guesses
to guarantee a win.
Proof of Lemma 2:
Round 1 Solutions
www.usamts.org
Zara can only win immediately if Bt consists of a single element. But the only possible
At that produce a single element are At = {1} and At = {2011}; if any 1 < n < 2011 are in
At , then both n 1 and n + 1 are in Bt .
Suppose without loss of generality that 1 < a 1006 (the proof for a > 1006 is sym-
metric). Then Bt = {a 1, a + 1}, and in particular no element less than a 1 is in At+1 .
Similarly, no element less than a 2 is in At+2 , so we cannot have 1 At+k until k a 1.
Thus we require at least a 1 guesses to have a possible winning move, plus 1 move (for the
winning guess itself). Thus Zara needs at least a = e(a) more guesses.
2
Now we are ready to complete the proof. The game starts in an interval phase for which
At is always an interval with at least 2 elements until some guess causes At to be a single
element. Specifically, by Lemma 1, At is the single {k} after at least E(A1 ) k guesses
(since E(At ) can decrease by at most 1 each guess), and then by Lemma 2 another k guesses
are required to guarantee Zara a win. So we require at least 1 + E(A1 ) guesses to win (the
extra 1 is the first guess). If z1 is anything other than 1006, then 1006 A1 and then
E(A1 ) = 1008. If z1 = 1006, then either A1 = [1, 1005] or A1 = [1007, 2011]. In either case
E(A1 ) = 1007, so 1007 in the minimum possible value of E(A1 ). Therefore we need at least
1 + 1007 = 1008 guesses.
The algorithm at the beginning describes how 1008 guesses will guarantee a win, so we
are done.
Round 1 Solutions
www.usamts.org
6/2/22. The roving rational robot rolls along the rational number line. On each turn, if the
robot is at pq , he selects a positive integer n and rolls to p+nq
q+np
. The robot begins at the
rational number 2011. Can the roving rational robot ever reach the rational number 2?
Solution 1. Let nj be the integer the robot chooses for step j. Let aj and bj the
sequences defined recursively by a0 = 2011, b0 = 1, and
aj = aj1 + nj bj1 ,
bj = bj1 + nj aj1 .
aj
Then at step j, the robots location is bj
. Notice now that
aj + bj = (1 + nj )(aj1 + bj1 ),
aj bj = (1 nj )(aj1 bj1 ).
Qk Qk
Let + = j=1 (1 + nj ) and = j=1 (1 nj ), so
ak + bk = + (a0 + b0 ),
ak bk = (a0 b0 ).
Also notice that a0 + b0 = 2012 and a0 b0 = 2010. Therefore
2ak = 2012+ + 2010 ,

2bk = 2012+ 2010 .
At step k the robots location is

2ak 2012+ + 2010
= .
2bk 2012+ 2010
If this is equal to 2, then
2012+ + 2010 = 2(2012+ 2010 ),
or
3 2010 = 2012+ .
We need to determine whether there is some set of positive integers n1 , . . . , nk for some k
such that
k
Y k
Y
3 (2 3 5 67) (1 nj ) = (2 2 503) (1 + nj ).
j=1 j=1
Round 1 Solutions
www.usamts.org
Clearly k must be even so that the left side of the above is positive. First we eliminate the
powers of two by choosing n1 = 5. Then we want to solve
n
Y n
Y
3 (2 3 5 67)(1 5) (1 nj ) = (2 2 503)(1 + 5) (1 + nj ).
j=2 j=2
or n n
Y Y
3 2 3 5 67 4 (1 nj ) = 2 2 503 6 (1 + nj ).
j=2 j=2
Cancellation gives
n
Y n
Y
3 5 67 (1 nj ) = 503 (1 + nj )
j=2 j=2
Now we choose our nj to reduce the size of the largest primes. For example if n2 = 504, then
we reduce to solving
n
Y n
Y
3 5 67(1 504) (1 nj ) = 503(1 + 504) (1 + nj )
j=3 j=3
which simplifies to
n
Y n
Y
3 5 67 (1 nj ) = 5 101 (1 + nj )
j=3 j=3
Choosing the sequence for the nj
5, 504, 102, 104, 66, 12, 10, 8
gives the products
3 (2 3 5 67) Q(1 cj ) = (1)8 3 (2 3 5 67) (4 503 101 103 65 11 9 7)

Q
(2 2 503) (1 + cj ) = (2 2 503) (6 505 103 105 67 13 11 9).
Since these are equal, this sequence gives a valid solution. Explicitly the rationals the roving
rational robot sees are
2011 5 / 84 504 / 84 102 / 18 104 / 86 66 / 4 12 / 16 10 / 2 8 / 2
2 419 17 85 19 17 5 5 1
Round 1 Solutions
www.usamts.org
Solution 2. Suppose the robot is at the rational number p, where p is a positive integer,
p 2. Taking n = p2 p 1, the next number is
p+n p2 1 p2 1 1
= 3 2
= 2
= .
1 + np p p p+1 (p 1)(p 1) p1
Next, suppose the robot is at the rational number 1/q, where q is a positive integer,
q 2. Taking n = q 2 q 1, the next number is
3 Virtual2PDF Printer
1 + nq q q q+1 (q 2 1)(q 1)
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy
= = = q 1.
q+n q2 1 q2 1
Hence, the robot can take the path
1 1 1
2011 2009 3.
2010 2008 4
When the robot is at 3, taking n = 9, the next number is

3+91 12 3
= = ,
1+93 28 7
and taking n = 11, the next number is
3 + 11 7 80
= = 2.
7 + 11 3 40
Credits: Problem 1/2/22 was proposed by Palmer Mebane.


Round 1 Solutions
www.usamts.org
1/1/23. The grid on the right has 12 boxes and 15 edges connect-
ing boxes. In each box, place one of the six integers from 1 to 6
such that the following conditions hold:
For each possible pair of distinct numbers from 1 to

6, there is exactly one edge connecting two boxes
with that pair of numbers.
If an edge has an arrow, then it points from a box

with
Create PDF with GO2PDF for free, if you aremove
wish to smaller
this line,number toVirtual
click here to buy a box with a larger number.
PDF Printer
First, we notice that each number 1 through 6 must be placed

in exactly two boxes. For each number, one of the two boxes must
have 2 neighbors and the other must have 3. Of the boxes with 3
1 ? neighbors, there is exactly one that does not have any incoming
arrows. This box must contain the value 1. Likewise, only two
? ? ? of the boxes with three neighbors have no outgoing arrows, and
two of the boxes with two neighbors have no outgoing arrows.
Therefore the two sixes must lie in these four boxes. We label these as ? in our first diagram.
We note that wherever the two sixes land, they must have 5
distinct neighbors. In particular, they cannot land in a pair of
boxes that share a neighbor. Among these four, the only 2- and
3-valent boxes (boxes with 2 or 3 neighbors respectively) that do 1 6
not share a neighbor are the two to the right.
Now that weve taken care of the simplest cases, we will need 6
to be much more careful with the remaining boxes. Lets label them as in the first figure on
the next page, using letters early in the alphabet to represent the 2-valent boxes (containing
1-5) and letters late in the alphabet for the 3-valent boxes (containing 2-5).
Round 1 Solutions
www.usamts.org
A B X C The values of D and E must be equal to two of the values from

X, Y , Z, and W . However, we know that two boxes with the
same value cannot be neighbors and also cannot share neighbors,
Y 1 6 Z so the values D and E must equal the values X and Y in some
order. By the symmetric argument, the values A and B must be
6 W D E equal to the values Z and W in some order. The specific upshot
of this is that we now know that we must have C = 1.
The remaining entries are all from 2 to 5. We know that A, A 2 D 1
X, and
Create PDF with GO2PDF for free, Y
if youare
wish todistinct and
remove this line, allto greater
click here Printer B, so B = 2. Recall
buy Virtual PDFthan
also that B must be equal to one of Z or W . Since W > D, W

cannot be equal to 2, and so Z must be 2. Since {A, B} = {W, Z} E 1 6 2
and B = Z, we must have W = A. Along the same lines, since
W = A and A neighbors Y , we cannot have Y = D, so instead 6 A D E
we must have Y = E and X = D. We compile this to the right.
Finishing the grid is straightforward from here. We know that A, D, and E must be
equal to 3, 4, and 5 in some order. The bottom of the grid tells us that D is the smallest of
the three, so D = 3. The top of the grid tells us that A < E, so we have A = 4 and E = 5.
The solution is given below, and a direct check shows that all conditions hold. Our
argument above shows that this solution is unique.
4 2 3 1
5 1 6 2
6 4 3 5
Round 1 Solutions
www.usamts.org
2/1/23. Find all integers a, b, c, d, and e, such that

a2 = a + b 2c + 2d + e 8,
b2 = a 2b c + 2d + 2e 6,
c2 = 3a + 2b + c + 2d + 2e 31,
d2 = 2a + b + c + 2d + 2e 2,
e2 = a + 2b + 3c + 2d + e 8.
We begin by bringing all terms to the left sides,

a2 a b + 2c 2d e + 8 = 0,
b2 + a + 2b + c 2d 2e + 6 = 0,
c2 3a 2b c 2d 2e + 31 = 0,
d2 2a b c 2d 2e + 2 = 0,
e2 a 2b 3c 2d e + 8 = 0.
When we add these five equations together we get
a2 6a + b2 4b + c2 2c + d2 10d + e2 8e + 55 = 0.
Now we can complete the squares,
(a2 6a + 9) + (b2 4b + 4) + (c2 2c + 1) + (d2 10d + 25) + (e2 8e + 16) = 0,
or
(a 3)2 + (b 2)2 + (c 1)2 + (d 5)2 + (e 4)2 = 0.
From here we see that any quintuple (a, b, c, d, e) satisfying the original equations must
necessarily satisfy this final equation. However, since squares of integers are always nonneg-
ative, this can only be solved when each of the squares is equal to zero. Therefore the only
possible solution is
(a, b, c, d, e) = (3, 2, 1, 5, 4).
We now need to check that these values satisfy all of the original equations.
To check that these values give a solution we must substitute them all back into the
original equations. Indeed, all 5 equations are satisfied by these values,
32 = 3 + 2 2 1 + 2 5 + 4 8,
22 = 3 2 2 1 + 2 5 + 2 4 6,
12 = 3 3 + 2 2 + 1 + 2 5 + 2 4 31,
52 = 2 3 + 2 + 1 + 2 5 + 2 4 2,
42 = 3 + 2 2 + 3 1 + 2 5 + 4 8.
Round 1 Solutions
www.usamts.org
Therefore there is a single, unique answer: (a, b, c, d, e) = (3, 2, 1, 5, 4).
Round 1 Solutions
www.usamts.org
3/1/23. (Corrected from an earlier release.) You have 14 coins, dated 1901 through 1914.
Seven of these coins are real and weigh 1.000 ounce each. The other seven are counterfeit
and weigh 0.999 ounces each. You do not know which coins are real or counterfeit. You also
cannot tell which coins are real by look or feel.
Fortunately for you, Zoltar the Fortune-Weighing Robot is capable of making very precise
measurements. You may place any number of coins in each of Zoltars two hands and Zoltar
will do the following:
If the weights in each hand are equal, Zoltar tells you so and returns all of the coins.
If the weight in one hand is heavier than the weight in the other, then Zoltar takes one
coin, at random, from the heavier hand as tribute. Then Zoltar tells you which hand
was heavier, and returns the remaining coins to you.1
Your objective is to identify a single real coin that Zoltar has not taken as tribute. Is there
a strategy that guarantees this? If so, then describe the strategy and why it works. If not,
then prove that no such strategy exists.
We begin the process by placing one coin in each of Zoltars hands. If the coins are the
same weight, we replace the coin in Zoltars left hand with a new coin and eventually the
hands will be imbalanced. Zoltar then identifies a real and a fake coin and takes the real coin
as tribute. We now have 13 coins left: 6 of these are real and we have 1 known counterfeit
coin.
If we set the known counterfeit coin aside, we reduce the problem to the situation where
we now have 6 real coins and 6 counterfeit coins. We again weigh some chosen coin against the
other 11 coins until Zoltar takes a real coin and leaves a known counterfeit coin. Discarding
the counterfeit again lets us reduces the problem and we now have 5 real and 5 counterfeit
coins.
If we repeat this reduction process three more times we will reduce our set of coins to
exactly 4 unknown coins, 2 of which are real and 2 of which are counterfeit. Now we place
two coins in each of Zoltars hands. One of the three possible pairings that we can try will
lead to an imbalance. This imbalance happens when both real coins are in one of his hands
and both counterfeit coins are in the other. When this happens, Zoltar will take one of the
real coins and we are assued that the remaining coin from that side of the scale is also real.
1
In the earlier version, this sentence read, Then Zoltar tells you the result of the measurement, and
returns the remaining coins to you. The correction clarifies that the result of the measurement was
meant to refer only to which hand was heavier, not to the actual weight in either hand.
Round 1 Solutions
www.usamts.org
This process requires exactly 6 unbalanced weighings and leaves us with 8 coins. Seven
of these coins are known counterfeits and the eighth is the desired real coin.
Round 1 Solutions
www.usamts.org
4/1/23. Let ABCDEF and ABC 0 D0 E 0 F 0 be regular planar hexagons in three-dimensional

space with side length 1, such that EAE 0 = 60 . Let P be the convex polyhedron whose
vertices are A, B, C, C 0 , D, D0 , E, E 0 , F , and F 0 .
(a) Find the radius r of the largest sphere that can be enclosed in polyhedron P.
(b) Let S be a sphere enclosed in polyhedron P with radius r (as derived in part (a)). The
set of possible centers of S is a line segment XY . Find the length XY .
(a) This polyhedron consists of 2 hexagonal faces meeting along an edge, a pair of opposite
triangles, 4AF F 0 and 4BCC 0 , and three quadrilaterals connecting the remaining pairs of
corresponding sides. We are looking for the set of all spheres of maximal radius that fit
inside the region bounded by the 6 planes containing these polygons.
We begin by studying only the three planes containing the two hexagons and the quadri-
lateral DEE 0 D0 . We will find the maximal radius for a sphere bounded by these planes,
giving an upper bound for the solution of the problem, and then show that such a sphere
can be placed inside our polyhedron.
Consider the triangular prism T bounded by the three segments AB, ED, D
and E 0 D0 , and let T be the infinite triangular prism bounded by the lines E
B
containing these segments. These three lines are parallel and we take a cross- A
section of T orthogonal to these lines. Since EAE 0 = 60 , this cross-section is D
E
an equilateral triangle, and the radius of the largest sphere that can be enclosed
in T is the inradius of this triangle.
Since triangle ADE isa 30 -60 -90 triangle and DE = 1, the side length of the equi-
lateral triangle is AE = 3. The inradius of an equilateral triangle with side length s is
3
, so the inradius of this triangle is = 1 . Therefore the maximal radius of any sphere
s
2 3 2 3 2
bounded by T is 12 . This tells us we have an upper bound of 21 on the radius of any sphere
living inside P.
On the other hand, notice that since AB = 1, there does exist a sphere of radius 12 (so
diameter 1) tangent to all of the faces of T . Since T lies inside P this shows that 12 is
achievable as the maximal radius of sphere enclosed by P.
(b) We learn slightly more from the previous part. We know that all spheres enclosed
by P of radius 21 must be tangent to the faces of T , so the centers of these spheres must lie
on its line of symmetry: the line ` containing the incenters of 4AEE 0 and 4BDD0 . The
boundary of the desired segment XY is determined by the centers of the spheres that are
tangent to the remaining four faces of P. We find these boundary points now.
Round 1 Solutions
www.usamts.org
Let C 00 , D00 , E 00 , and F 00 be the midpoints of CC 0 , DD0 , EE 0 ,

and F F 0 , respectively. By symmetry, the hexagon ABC 00 D00 E 00 F 00
C
is planar and lies on the plane containing AB and `. B

D
0 0 0 0
Since the lines CC , DD , EE , and F F are all orthogonal to
the plane of ABC 00 D00 E 00 F 00 , the five faces of P that we have not A E
yet discussed are also all orthogonal to this plane. Therefore any F

sphere with center in the plane of ABC 00 D00 E 00 F 00 that is tangent
to any of the other faces will be tangent at a point on the hexagon
00 00 00 00 1
ABC
Create PDF with GO2PDF Dif you
for free, E wish
F to .remove
This reduces
this line, the
click here to buy problem
Virtual PDF Printer to classifying all circles with radius 2
that are
enclosed by this hexagon. We now solve this reduced problem.
Let M and N be the midpoints of AE 00 and BD00 , respectively. Then M N is parallel
to AB, and M N = AB = 1. By symmetry, M and N lie on C 00 F 00 , and C 00 N = F 00 M , so
C 00 N = F 00 M = (C 00 F 00 M N )/2 = (2 1)/2 = 1/2.
C 00 Since BDD0 is an equilateral triangle, triangle BDD00 is a 30 -

60 -90 triangle. Then

B D00
N 3 3 3
BD00 = BD = 3= .
2 2 2
Hence, BN = N D00 = BD00 /2 = 3/4. Then by Pythagoras on
M
rigtht triangle C 00 N D00 ,
A E 00
p
C 00 D00 = (C 00 N )2 + (N D00 )2
F 00 s
2 2
1 3
= +
2 4
r
1 9
= +
4 16
r
13
=
16

13
= .
4
The line ` is located a distance 21 from the D00 and E 00 , so specifically is closer to D00 E 00

than to AB. Since C 00 F 00 is a line of symmetry of this hexagon the extremal enclosed circles

must be tangent to C 00 D00 or E 00 F 00 . One endpoint of line segment XY , say X, is at a distance
of 1/2 from the segment C 00 D00 , and the other endpoint Y is at a distance of 1/2 from the
segment E 00 F 00 .
Round 1 Solutions
www.usamts.org
Let the sphere centered at X with radius 1/2 be tangent to D00 E 00 C 00 V P

and C 00 D00 at T and U , respectively. Let P be the projection of C 00
onto D00 E 00 , and let the angle bisector of P D00 C 00 intersect P C 00 U
at V . Then D00 V and D00 X are the internal and external angle D00
bisectors of P D00 C 00 , respectively, so they are perpendicular. This
implies that right triangles D00 P V and XT D00 are similar. Since X
T
XT = P D00 = 1/2, the two triangles are in fact congruent.
Y
By the angle bisector theorem, S
00 this line, click here to buy Virtual PDF Printer

Create PDF with GO2PDF for free, if you wish to remove
PD 1/2 3 3 E 00
PV = P C 00 = = .
P D00 00
+C D 00
1/2 + 13/4 4 2(2 + 13)
Rationalizing the denominator, we get

3( 13 2) 3( 13 2) 13 2
PV = = = .
2( 13 + 2)( 13 2) 29 6
Hence,
00 13 2
TD = PV = .
6
Let the sphere centered at Y with radius 1/2 be tangent to D00 E 00 at S. Then by sym-
metry, SE 00 = T D00 , so

00 00 00 00 13 2 5 13
XY = ST = D E T D SE = 1 2 = .
6 3
Round 1 Solutions
www.usamts.org
5/1/23. In the game of Tristack Solitaire, you start with three stacks of cards, each with a
different positive integer number of cards. At any time, you can double the number of cards
in any one stack of cards by moving cards from exactly one other, larger, stack of cards to
the stack you double. You win the game when any two of the three stacks have the same
number of cards.
For example, if you start with stacks of 3, 5, and 7 cards, then you have three possible
legal moves:
2, and 7 cards.
5, and 4 cards.
10, and 2 cards.
Can you win Tristack Solitaire from any starting position? If so, then give a strategy for
winning. If not, then explain why.
We present two solutions to this problem.

Solution 1:
First, we show the winning strategy for a particular starting configuration:
Lemma 1: Suppose the three stacks have a, b, and c cards, where 0 < a b < c, and
suppose that b is a multiple of a; that is, b = na for some positive integer n. Then there is
a winning strategy in which the first two stacks (the stacks that start with a and b cards)
result in the same number of cards.
Proof of Lemma 1 : For ease of referral, call the stack that starts with a cards the first
stack, the stack that starts with b cards the middle stack, and the stack that starts with
c cards the last stack. Note that if n = 1 then weve already won, so assume that n > 1.
Write n = (1dk1 . . . d1 d0 )2 as a base-2 integer, where k = blog2 nc and each di is a binary
digit (0 or 1). Then we can make the first and middle stacks equal in k steps, as follows:
On step i (for 0 i < k), if di = 1, then double the first stack by moving cards
from the middle stack, and if di = 0, then double the first stack by moving cards
from the last stack.
Round 1 Solutions
www.usamts.org
At the end of the process, we will have doubled the first stack k times, so there will be 2k a
cards in the first stack. Since we are removing 2i cards from the middle stack if and only if
di = 1 for all 0 i < k, we will remove exactly (dk1 dk2 . . . d1 d0 )2 a cards from the middle
stack, leaving (1000 . . . 0)2 a = 2k a cards in the middle stack. Thus the first and middle stacks
will be equal. Since we are moving a total of (2k 1)a cards, and (2k 1)a < 2k a b < c,
we will always have sufficient cards in the last stack for all the steps in which di = 0. 2
Lemma 1 establishes that we can win the game if the middle stack is an integer multiple
of the smallest stack. Next, we show that if condition does not hold, then we can improve
our forposition
Create PDF with GO2PDF into remove
free, if you wish a well-defined
this line, click here toway.
buy Virtual PDF Printer
Lemma 2: Suppose the three stacks have a, b, and c cards, where 0 < a < b < c, and
suppose that b is not a multiple of a. Then there is a series of moves that produces of stack
of size r, where 1 r < a.
Proof of Lemma 2: Write b = na + r, where n is a positive integer and 0 < r < a.
(Note that r is simply the remainder upon division of b by a). We perform exactly the same
algorithm as in Lemma 1. After this algorithm concludes, we have stacks of size 2k a, 2k a + r,
and a third stack whose size is irrelevant to this argument. One additional move of doubling
the 2k a stack from the 2k a + r stack leaves us with our stack of size r. 2
Proof of the Original Claim: Therefore, we can either win right away (using Lemma 1),
or we can use Lemma 2 to produce a position with a stack that is smaller than any of the
stacks we started with. We repeat this use of Lemma 2 until we arrive at a position in which
we can use Lemma 1 to win the gamewe are guaranteed that this must happen within a
finite number of uses of Lemma 2 (the worst case scenario is that the smallest stack only
decreases by 1 upon each application of Lemma 2, but eventually it will decrease down to
1, at which point Lemma 1 will apply since the size of the medium stack will always be a
multiple of 1).
Solution 2:
For an integer n 6= 0, we define ord2 (n) as the largest integer k such that 2k divides n.
For example, ord2 (24) = 3.
Lemma 1: Given stacks initially containing a and b cards, where ord2 (a) > ord2 (b), we
can make a series of Tristack Solitaire moves on those two stacks that changes their respective
sizes to a0 and b0 , where a0 = a/2 and min{ord2 (a0 ), ord2 (b0 )} = min{ord2 (a), ord2 (b)}.
Proof of Lemma 1: The sizes a and b are respectively even and odd multiples of 2ord2 (b) ,
so a + b is an odd multiple of 2ord2 (b) .
Since a 6= b, we can make a unique Tristack Solitaire move on the two stacks. The number
of cards moved is a multiple of 2ord2 (b) , so both stack sizes remain multiples of 2ord2 (b) after
the move. Moreover, one of the stack sizes must still be an even multiple of 2ord2 (b) and one
must be an odd multiple of 2ord2 (b) , since the sum of their sizes is still a + b. Therefore, the
Round 1 Solutions
www.usamts.org
sizes of the two stacks remain unequal, and we can continue making moves on these two
stacks perpetually. We also see that the minimum value of ord2 over the sizes of the two
stacks is invariant throughout this process.
Each move doubles the sizes of both stacks modulo a + b. Consider the set of multiples
of 2ord2 (b) modulo a + b. Doubling (mod a + b) is a one-to-one operation on this set, because
a + b is an odd multiple of 2ord2 (b) . In particular, doubling has an inverse operation on this
set, so as we perform Tristack Solitaire moves on the two stacks, their sizes are periodic
(mod a + b). This implies that their sizes are in fact periodic, since their sizes are always
between
Create PDF with GO2PDF for free, if 0 wish to a
youand + b.this Therefore,
remove line, click here to buythe
Virtual stacks
PDF Printer eventually return to their original sizes. Suppose
this happens after k Tristack Solitaire moves (k > 0). In that case, one stack must have
move to half its original size after k 1 moves, since every Tristack Solitaire move doubles
one stack. Moreover, this stack must be the first stack, because the minimum value of ord2
over the two stack sizes cannot decrease. This proves the lemma.
Given two stacks with distinct ord2 , we refer to a series of Tristack Solitaire moves on
those two stacks that halves the larger order stack as the inverse move.
Here is an algorithm for playing Tristack Solitaire:
1. If two stacks are equal, stop and declare victory. Otherwise, go to step 2.
2. If possible, choose two stacks of sizes a and b such that ord2 (a) = ord2 (b). Perform the
unique Tristack Solitaire move on those two stacks, then go to step 1. If there are no
such stacks, go to step 3.
3. Choose two stacks of sizes a and b such that ord2 (a) ord2 (b) + 2. Perform an inverse
move on those two stacks, then go to step 1.
Lemma 2: The algorithm can be performed.

Proof of Lemma 2: The only thing to check is that, when we get to step 3, we can
always choose two stacks of sizes a and b such that ord2 (a) ord2 (b) + 2. This is true
because we only get to step 3 from step 2, and then only if no two stacks have sizes of
the same ord2 . In this case, we may denote the sizes of the stacks by a, b, and c, where
ord2 (a) > ord2 (c) > ord2 (b). It follows that ord2 (a) ord2 (b) + 2.
Lemma 3: Let ` and m respectively denote the largest and second largest of the values
ord2 (x), ord2 (y), ord2 (z), where x, y, and z are the sizes of three stacks at any given time.
Each time we perform step 2 of our algorithm, m increases. Each time we perform step 3,
m does not change, and ` m decreases.
Proof of Lemma 3: When we perform step 2, we choose two stacks of sizes a and b such
that ord2 (a) = ord2 (b). Thus, a and b are both odd multiples of 2ord2 (b) . The number of
cards we move is one of these odd multiples, so after the move, both stack sizes are even
multiples of 2ord2 (b) . This means the ord2 of both stack sizes increases, while the third stack
Round 1 Solutions
www.usamts.org
is unchanged. The other stack had either the largest or smallest order (including the case
that all orders are equal), so the value of m increases.
Now let us examine step 3. Suppose the stack sizes are a, b, and c, where ord2 (a) >
ord2 (c) > ord2 (b). We perform the inverse move on the stacks of size a and b, changing
their sizes to a0 = a/2 and b0 . Thus ord2 (a0 ) = ord2 (a) 1 ord2 (c). We know that
ord2 (b) < ord2 (a) 1. It follows by Lemma 1 that ord2 (b0 ) = ord2 (b) < ord2 (c). Therefore,
m does not change, and ` m decreases by 1.
Lemma 4: The algorithm terminates, and we win.
Proof of Lemma 4: Each time we perform step 3 of the algorithm, ` m (as defined
in Lemma 3) decreases, and we check the condition on step 2 before returning to step 3.
Therefore, after finitely many steps, we have ` = m, triggering a move in step 2. Each time
this happens, m increases. There is an upper bound on m, since the sum of the stack sizes
is constant. Therefore, the algorithm must terminate, and the only way this can happen is
in step 1, where we win.
Credits: Problem 4/1/23 is based on a proposal by Luyi Zhang.


Round 2 Solutions
www.usamts.org
1/2/23.
Find all the ways of placing the integers 1, 2, 3, . . . , 16 in the boxes below, such that
each integer appears in exactly one box, and the sum of every pair of neighboring integers
First, we construct a graph with 16 vertices, where each 16

8
1
vertex corresponds to one of the integers from 1 through
9 15
16, and two vertices are joined by an edge if the sum of the
integers corresponding to the vertices is a perfect square. 7 10
This gives us the graph to the right.

2 6
A solution to this problem corresponds to a path in
this graph that passes through every vertex exactly once
14 3
(also known as a Hamiltonian path). The vertices corre-
sponding to the integers 8 and 16 have only one edge, so 11 13
the path must start at the vertex labelled 8 and end at 5 12

4
the vertex labelled 16, or vice versa.
If the first number is 8, then the next number must be 1. We then have two choices for
the number after 1, namely 3 and 15. If the number after 1 is 3, then the path cannot reach
the vertex labelled 15 (because the last number must be 16), so the number after 1 must be
15, and then all numbers after that are uniquely determined. This gives us the first solution
shown below.
If the first number is 16, then the next numbers must be 9, 7, and so on, until 3. We
then have two choices for the number after 3, namely 1 and 6. If the number after 3 is 1,
then the path cannot reach the vertex labelled 6 (because the last number must be 8), so the
number after 3 must be 6, and then all numbers after that are uniquely determined. This
gives us the second solution shown below.
8 1 15 10 6 3 13 12 4 5 11 14 2 7 9 16
16 9 7 2 14 11 5 4 12 13 3 6 10 15 1 8
Round 2 Solutions
www.usamts.org
2/2/23.
Four siblings are sitting down to eat some mashed potatoes for lunch: Ethan has 1 ounce
of mashed potatoes, Macey has 2 ounces, Liana has 4 ounces, and Samuel has 8 ounces. This
is not fair. A blend consists of choosing any two children at random, combining their plates
of mashed potatoes, and then giving each of those two children half of the combination.
After the childrens father performs four blends consecutively, what is the probability that
the four children will all have the same amount of mashed potatoes?
To see how all four children can end up with the same amount of mashed potatoes after
four blends, we start with the final amounts and work backwards. The total amount of
mashed potatoes at the start, in ounces, is 1 + 2 + 4 + 8 = 15. This total never changes, so
at the end, each child must have 154
ounces of mashed potatoes.
We now consider the blends in reverse order. It is possible that the father kept blending
after everyone reached equal amounts of 15 4
, but at some point, the father must have blended
two different amounts to produce two equal amounts of 15 4
. Let these two different amounts
15 15
be 4 + x and 4 x, where x is a positive real number. So at this point, the amounts among
the children are 15
4
, 15
4
, 15
4
+ x, and 15
4
x.
We continue to consider the blends in reverse order. A blend always produces two equal
amounts, and the only two equal amounts that we have are 15 4
and 15
4
. Therefore, at some
point, the father must have blended another two different amounts to produce two equal
amounts of 154
. Let these two different amounts be 15
4
+ y and 15
4
y, where y is a positive
real number. At this point, the amounts among the children are 15 4
+ x, 15
4
x, 15
4
+ y, and
15
4
y.
Not all these amounts can be integers (for example, it is impossible for both 15
4
+ x and
15
4
x to be integers), so in particular, these amounts cannot be the original amounts of 1, 2,
4, and 8. This means that further blends are required to get to these amounts. However, a
blend always produces two equal amounts. The only way that there are two equal amounts
among 15 4
+ x, 15
4
x, 15
4
+ y, and 15
4
y is if x = y.
Hence, our amounts are now 154
+ x, 15
4
+ x, 15
4
x, and 15
4
x. Since a blend always
produces two equal amounts, we see that at some point, the father performed a blend to
produce two equal amounts of 15
4
+ x, and another blend to produce two equal amounts of
15
4
x. We have shown that the following four blends are required:
15
(A) A blend to produce two equal amounts of 4
+ x.
15
(B) A blend to produce two equal amounts of 4
x.
15 15
(C) Blending an amount of 4
+ x and an amount of 4
x to produce two equal amounts
Round 2 Solutions
www.usamts.org
15
of 4
.
15 15
(D) Blending another amount of 4
+ x and another amount of 4
x to produce two equal
amounts of 15
4
.
The father performs a total of four blends, so these must be the only blends.
Blends (A) and (B) must come before blends (C) and (D). After blends (A) and (B),
the amounts are 15 4
+ x, 15 4
+ x, 15 4
x, and 15 4
x. After any first blend, the amounts are
always of the form a, a, b, and c, where a,
b, and c are distinct. (We know that a, b, and
c are distinct because among the original amounts of 1, 2, 4, and 8, no amount is equal to
the average of any other two amounts.) The only way to get two pairs of equal amounts
after the next blend is to blend the amounts of b and c. The probability that this occurs is
1
= 16 .
(42)
We then have the amounts 15
4
+ x, 15
4
+ x, 15
4
x, and 15
4
x. The next blend must
15 15
blend an amount of 4 + x and an amount of 4 x. The probability that this occurs is
22
= 46 = 23 .
(42)
Finally, we have the amounts 15
4
, 15
4
, 15
4
+ x, and 15
4
x. The next blend must blend the
amounts of 4 + x and 4 x. The probability that this occurs is 14 = 16 .
15 15
(2)
Therefore, the probability that after four blends, all four children have the same amount
of mashed potatoes is
1 2 1 2 1
= = .
6 3 6 108 54
Round 2 Solutions
www.usamts.org
3/2/23.
Find all integers b such that there exists a positive real number x with
1 1 1
= + .
b b2xc b5xc
Here byc denotes the greatest integer that is less than or equal to y.
We claim that the only such positive integers b are 3 and all positive multiples of 10.
Let n = bxc. Then there exists a unique integer r, 0 r 9, such that
r r+1
n+ x<n+ .
10 10
For each such value of r, we can express b2xc and b5xc in terms of n.
r b2xc b5xc
0 2n 5n
1 2n 5n
2 2n 5n + 1
3 2n 5n + 1
4 2n 5n + 2
5 2n + 1 5n + 2
6 2n + 1 5n + 3
7 2n + 1 5n + 3
8 2n + 1 5n + 4
9 2n + 1 5n + 4
Solving for b in the given equation, we find
b2xcb5xc
b= .
b2xc + b5xc
We divide into cases.

Case 1: r = 0 or 1. In this case,
2n 5n 10n
b= = .
2n + 5n 7
If b is an integer, then n must be a multiple of 7. Let n = 7k. Then b = 10k, so b can take
on all positive multiples of 10.
Round 2 Solutions
www.usamts.org

2n(5n + 1) 10n2 + 2n
b= = .
2n + (5n + 1) 7n + 1
If b is an integer, then
70n2 + 14n 4n
7b 10n = 10n =
7n + 1 7n + 1
4n
must be an integer. But 0 < < 1, so b cannot be an integer.
7n+1
Case 3: r = 4. In this case,

2n(5n + 2) 10n2 + 4n
b= = .
2n + (5n + 2) 7n + 2
70n2 + 28n n2
7b (10n + 1) = (10n + 1) = .
7n + 2 7n + 2
must be an integer. If n = 2, then b = 3.
n2
If n = 1, then b = 14/9, which is not an integer, and if n 3, then 0 < 7n+2
< 1, so b is
not an integer.
Case 4: r = 5. In this case,
(2n + 1)(5n + 2) 10n2 + 9n + 2
b= = .
(2n + 1) + (5n + 2) 7n + 3
70n2 + 63n + 14 5n + 2
7b (10n + 4) = (10n + 4) = .
7n + 3 7n + 2
5n+2
must be an integer. But 0 < 7n+2
< 1, so b cannot be an integer.
(2n + 1)(5n + 3) 10n2 + 11n + 3
b= = .
(2n + 1) + (5n + 3) 7n + 4
70n2 + 77n + 21 2n + 1
7b (10n + 5) = (10n + 5) = .
7n + 4 7n + 2
2n+1
Round 2 Solutions
www.usamts.org
(2n + 1)(5n + 4) 10n2 + 13n + 4

b= = .
(2n + 1) + (5n + 4) 7n + 5
70n2 + 91n + 28 6n + 3
7b (10n + 5) = (10n + 5) = .
7n + 5 7n + 5
6n+3
Therefore, the only possible values of b are 3 and all positive multiples of 10.
Round 2 Solutions
www.usamts.org
4/2/23.
A luns with vertices X and Y is a region bounded by two circular arcs meetingat the

endpoints
X and Y . Let A, B, and V be points such that AV B = 75 , AV = 2 and
BV = 3. Let L be the largest area luns with vertices A and B that does not intersect the

lines V A or V B in any points other than A and B. Define k as the area of L. Find the value
k
.
(1 + 3)2
We first consider the question of which circular arcs from A to B lie

entirely inside the region bounded by the 75 angle to the left. Any such
A
arc is the arc of a circle with center equidistant from A and B. There-
fore the locus of possible centers for these arcs lie on the perpendicular
bisector, `, of the segment AB.
75
V B
If we choose a center for our circle, the circle defines two `

different arcs, one to the left of AB and one to the right
or AB. For the sake of this problem, we will define left as
A
pertaining to the half-plane bounded by AB containing V

and right as pertaining to the half-plane bounded by AB O
not containing V .
For any given center, we must explore whether either V B

or both of these arcs intersects either the line V A or the

line V B. Let O be the center of an arbitrary circle that

intersects V B at B. (Note that O and B uniquely determine

the circle.) If this circle is tangent to V B, then V BO = 90 . If the circle intersects V B

to the left of B then V BO < 90 . If the circle intersects V B to the right of B then
V BO > 90 .
Round 2 Solutions
www.usamts.org

`
In the figure to the left, weve drawn the lines V B and `.
The point XB is the intersection of ` and the perpendicular

to V B at B. The point XB is the center of the black circle

A
XB and the black circle intersects V B only at B. In particular,

both arcs from A to B in this circle lie above the line V B,
(excepting the point B).
V B The green region of ` is the set of points on ` to the right

of XB B. These are the centers of circles that intersect V B
toVirtual
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy the PDFright
Printer of B. For such a circle, only the left arc from A

to B lies above V B. The blue region of ` is the set of points

on ` to the left of XB B. These are the centers of the circles that intersect V B to the left of

B. For such a circle, only the right arc from A to B lies above V B. Notice that every valid
arc lies inside the circle centered at XB containing the point B.

We consider an identical construction for the line V A in `
the figure to the right. The point XA is the center of the
unique circle for which both the left and right arcs from A to
A
B do not intersect V A. The green points to the right of XA
are the centers of the circles for which the left arc from A to A X

B does not intersect V A. The blue points to the left of XA
are those points for which the right arc does not intersect
V B
V A. Notice that all of these arcs are in the interior of the
circle centered at XA .
`
Next we combine these figures. We drop both perpendic-
ulars through A and B. Since V B > V A, the point XB is to
the right of XA . If the center of a circle lies on XA XB , then
A
XB the left arc of the circle intersects V B twice and the right

XA arc of the circle intersects V A twice. Therefore neither arc
is an arc of a luns.
V B If the center of a circle lies to the right of XB on ` (colored
green here), then the left arc of this circle does not intersect
either line except at A and B. If the center of a circle lies
to the left of XA (colored blue), then the right arc of the
circle does not intersect either line, except at A and B. Therefore the green region of `
parameterizes the set of all valid left arcs and the blue region of ` parameterizes all of the
valid right arcs.
Consider the black luns in this figure. It has left arc with center XB and right arc with
center XA . This figure is a luns, and every valid luns is bounded by a pair of arcs that
Round 2 Solutions
www.usamts.org
lie inside this figure. Therefore every valid luns is a subset of this luns and this luns has
the maximal area of any luns satisfying the assumptions. Now we compute this area, by
computing the sum of the areas of the green region and the blue region.

Define the lengths a = V B = 3, b = V A = 2, and c = AB. The law of cosines gives
c2 = (V A)2 + (V B)2 2(V A)(V B) cos 75

6 2
=2+32 2 3
4
PDF Printer6 2
=52 6
4
62 3
=5
2
= 2 + 3.
Notice also that, since

(1 + 3)2 = 4 + 2 3 = 2c2 ,

and 1 + 3 is positive,
1+ 3
c= .
2
Finally, we apply the law of sines to find the other angles in V AB.
Since

1+
3 A
c 2 4(1 + 3)
= = = 2, 60
sin V 6+ 2 2 3+2
1+

3
2
4 2
we know that 75 45
b 2 V

3 B
2= = ,
sin B sin B
so sin B = 1 and V BA = 45 . Subtracting gives V AB = 60 .
2
First we compute the area of the green region. Since ABV = 45

and AXB = BXB , the triangle AXB B is right isosceles. The sector
containing the green region is a quarter of a circle of radius c2 , so the A
XB
2 2
entire sector has area 41 c2 = c8 . To find the green region we
subtract the area of the triangle to get
V B
c2 c2 2 2
= c .
8 4 8
Round 2 Solutions
www.usamts.org
Next we compute the area of the blue region. Since triangle AXA B
is isosceles and V AB = 60 , we get that XA = 120 . Therefore the
blue region is the union of 23 of the circle with center XA plus the area A
XA
of triangle AXA B. The triangle has altitude 2c 3 with respect to base
AB and the radius of this circle is c3 . Therefore the total area of the
blue region is 2 V B

2 c 1 c 3 3 + 8 2
+ c = c .
3 3 2 2 3 36
This makes the total area of the luns

2 2 3 3 + 8 2 6 3 + 25 18 2
k= c + c = c .
8 36 72

(1+ 3)2
Since c2 = ,
2
k k 6 3 + 25 18
= 2 = .
(1 + 3)2 2c 144
Round 2 Solutions
www.usamts.org
5/2/23.
Miss Levans has 169 students in her history class and wants to seat them all in a 13 13
grid of desks. Each desk is placed at a different vertex of a 12 meter by 12 meter square grid
of points she has marked on the floor. The distance between neighboring vertices is exactly
1 meter.
Each student has at most three best friends in the class. Best-friendship is mutual: if
Lisa is one of Shannons best friends, then Shannon is also one of Lisas best friends. Miss
Levans knows that if any two best friends sit at points that are 3 meters or less from each
other then they will be disruptive and nobody will learn any history. And that is bad.
Prove that Miss Levans can indeed place all 169 students in her class without any such
disruptive pairs.
There are only a finite number of ways of arranging the 169 students among the 169
desks. Consider the arrangement that has the least number of disruptive pairs. We claim
that this minimal arrangement has no disruptive pairs.
For the sake of contradiction, suppose that there is a disruptive pair in this arrangement.
We say that two students are neighbors if they sit within 3 meters of each other. As shown
below, every student has at most 28 neighbors.
Let students S and T be a disruptive pair, which means they are best friends and neigh-
bors. Let A be the set of all best friends of neighbors of S, so |A| 3 28 = 84. Let B be
the set of all neighbors of best friends of S, so |B| 28 3 = 84. Note that S is in both A
and B, so there are at most 83 + 83 = 166 members in A B other than S. But there are
a total of 169 students, so there must be at least 169 1 166 = 2 students, other than S,
who are not in A or B.
Round 2 Solutions
www.usamts.org
Let S 0 be one of these students, other than T . We swap the locations of students S and
S 0 . This breaks up the disruptive pair S and T . Since S 0 is not in A, S 0 is not best friends
with any of his neighbors in his new location (where S originally was), and since S 0 is not
in B, S is not best friends with any of his neighbors in his new location (where S 0 originally
was) either. So by swapping S and S 0 , we have reduced the number of disruptive pairs by
at least one. We now have an arrangement with less disruptive pairs, which contradicts
the minimality assumption. Therefore, in the minimal arrangement, there are no disruptive
pairs, so such a seating is possible.
Credits: Problem 4/2/23 is based on a proposal by Sandor Lehoczyk.

Problem 5/2/23 is based on a problem from the Bulgarian journal Matematika.

Round 3 Solutions
www.usamts.org
1/3/23. Fill in the circles to the right with the numbers 1

1 through 16 so that each number is used once (the
8 2
number 1 has been filled in already). The number in 11
any non-circular region is equal to the greatest differ- 7
ence between any two numbers in the circles on that

regions vertices. 2
10
You do not need to prove that your configuration is 2
the only one possible; you merely need to find a valid

configuration. (Note: In any other USAMTS problem,
you need to provide a full proof. Only in this problem 4
8 5
is an answer without justification acceptable.)
First, we point out that if a triangular region contains the 1 A B
number 2, then the three vertices for the region must contain 8 2
11
three consecutive numbers. In particular, if we look at the top- 7
C D
left region of the diagram, as in the figure to the right, then we see
2
that both {A, B, D} and {C, D, E} must be sets of consecutive 10
2
numbers, in some order. Since the greatest difference between E
two elements in the set {1, A, D} is 8, and the smallest element 4
4
is 1, A or D must be equal to 9. 8 5
Also, the greatest difference between two elements in the set

{1, C, D} is 7, so D cannot be 9, which means A = 9 and D 8. Also, C or D must be
equal to 8. If C = 8, then D 7, but then {A, B, D} cannot contain three consecutive
numbers, so we must have D = 8. We know the set {C, D, E} = {C, 8, E} must contain
three consecutive numbers, and none of the elements are 9, so {C, E} = {6, 7} in some order.
The set {A, B, D} = {9, B, 8} contains three consecutive numbers, and B cannot be 7, so
B = 10.
1 9 10 Now we have the figure to the left, where C and E are 6 and
8 2
11
7 in some order. We will call a number small if it is in the set
7
{2, 3, 4, 5}, and we will call a number large if it is in the set
C 8
2
10
{11, 12, 13, 14, 15, 16}.
2
E The unlabeled circles contain exactly those numbers that are

4 either small or large. There are 4 small numbers and 6 large
4
8 5 numbers.
Consider the numbers in the four vertices of the bottom-left square containing the number
4, including E. Since E is at most 7, and all these numbers are within 4 of each other, all
Round 3 Solutions
www.usamts.org
three of the remaining numbers must be small. There is one more small number elsewhere
in the grid. Now consider the numbers in the four vertices of the top-right square containing
the number 11. Two of these numbers differ by 11, so they cannot be only 10 and large
numbers. Hence, one of these numbers must be small. Also, the difference between two of
these numbers is 11, but neither of these numbers can be the 10. Thus, we can replace the
square containing the number 11 with a triangle containing the number 11, as shown below.
We can do the same for the center square containing the 1 9 10
number 10. Among the four numbers attached to this square, 8 2
two foroffree,them
Create PDF with GO2PDF if you wishdiffer
to remove by 10.
this line, Neither
click here ofPDFthese
to buy Virtual Printer numbers can be the 7 11
C 8
8, so we can replace the square containing the number 10 with
2
a triangle containing the number 10, as shown to the right. We 10 2
let F , G, and H be the three small numbers in the lower-left. E F
4
Next, consider the triangles labeled 2, 4, and 5 in the right 4
8 5
side of the figure. Since the smallest difference between a small G H
number and a large number is 11 5 = 6, the five numbers
attached to these triangles are either all large or all small. Since there are only four small
numbers, all five of these numbers must be large. Let these five numbers be V , W , X, Y ,
and Z, as in the figure below. The final two vertices contain the one remaining small number
and one remaining large number. We label these two vertices S and T .
1 9 10 T We have the following:

8 2
7 11 {C, E} = {6, 7}
C 8 S V
{F, G, H} {2, 3, 4, 5}
2
10 2 {V, W, X, Y, Z} {11, 12, 13, 14, 15, 16}
E F W X
4
4 Either S is small and T is big, or T is small and S is big.
8 5
G H Y Z
Consider the numbers W , Y , and Z. We know that the dif- 1 9 10 T
ference between two of these numbers is 5. But these numbers 8 2
are all large, and the only way that the difference between two 7 11
C 8 S V
large numbers is 5 is if the numbers are 11 and 16. Suppose W
2
or Y is equal to 16. We know that the greatest difference among 10 2
the numbers H, W , and Y is 8, which means H must be equal E F W X
to 8. However, we have already used 8, so neither W nor Y can 4
4
be 16, which means Z = 16. Then W or Y is equal to 11. We 8 5
know that the greatest difference among the numbers X, W , and G H 11 16
Z (which is 16) is 4, so W cannot be 11, which means Y = 11.

Round 3 Solutions
www.usamts.org
It also follows that W or X is equal to 12. We know that the greatest difference among
the numbers V , W , and X is 2, so these three numbers are consecutive. Also, all of these
numbers are large, and we have already used 11, so these numbers are 12, 13, and 14, in some
order. Therefore, the remaining large number must be 15. We know that one of M and N
is the remaining large number, and the other is the remaining small number. Furthermore,
these numbers differ by 11, so the remaining small number is 15 11 = 4.
Hence, the small numbers F , G, and H are 2, 3, and 5, in some order. We know that
the greatest difference among the numbers E, F , G, and H is 4. But the largest number is
E, and
Create PDF with GO2PDF for free,the smallest
if you wish number
to remove this line, click hereis 2,Virtual
to buy so E PDF 2 = 4, which means E = 6. Then C = 7.
Printer
Now consider the numbers W , H, and Y = 11. We know 1 9 10 T
that the greatest difference among these numbers is 8. Since W 8 2
is either 12, 13, or 14, and H is 2, 3, or 5, the greatest difference 7 11
7 8 S V
among these numbers is W H. Furthermore, the only values
2
that satisfy W H = 8 are W = 13 and H = 5. 10 2
Then X is 12 or 14. But the greatest difference among the 6 F W X
4
numbers W = 13, X, and Z = 16 is 4, so X = 12, which makes 4
8 5
V = 14. G H 11 16
1 9 10 T Finally, F and G are 2 and 3 in some order, and M and N

8 2 are 4 and 15 in some order. The greatest difference among F , N ,
7 11
and W = 13 is 10. By a quick check, the only values for which
7 8 S 14
this occurs are F = 3 and N = 4. Then G = 2 and M = 15,
2
10 2
which completes the grid.
6 F 13 12
4
4
8 5 1 9 10 15
G 5 11 16
8 2
11
7
7 8 4 14
2
10
2
6 3 13 12
4
4
8 5
2 5 11 16
Round 3 Solutions
www.usamts.org
2/3/23. Let x be a complex number such that x2011 = 1 and x 6= 1. Compute the sum
x2 x4 x6 x4020
+ + + + .
x1 x2 1 x3 1 x2010 1
Let S denote the given sum, so

2010
x2 x4 x4020 X x2k
S= + 2 + + 2010 = . (1)
x1 x 1 x 1 k=1 xk 1
We can reverse the order of the terms, to get

2010
x4020 x4018 x2 X x40222k
S = 2010 + 2009 + + = .
x 1 x 1 x 1 k=1 x2011k 1
Since x2011 = 1,
x40222k x2k 1 1
2011k
= k
= k 2k
= k ,
x 1 x 1 x x x (1 xk )
so
2010
X 1
S= . (2)
k=1
xk (1 xk )
Adding equations (1) and (2), we get

2010 2010
X x2k X 1
2S = +
k=1
xk 1 k=1 xk (1 xk )
2010
x2k

X 1
= +
k=1
xk 1 xk (1 xk )
2010
x3k

X 1
=
k=1
xk (xk 1) xk (xk 1)
2010
X x3k 1
= .
k=1
xk (xk 1)
Round 3 Solutions
www.usamts.org
We can factor x3k 1 as (xk 1)(x2k + xk + 1), so

2010
X (xk 1)(x2k + xk + 1)
2S =
k=1
xk (xk 1)
2010
X x2k + xk + 1
=
k=1
xk
2010
X
k 1
= x +1+ k
x PDF Printer
k=1

1 2 1 2010 1
= x+1+ + x + 1 + 2 + + x + 1 + 2010
x x x
1 1 1
= (x + x2 + + x2010 ) + 2010 + + 2 + + 2010 .
x x x
Since x2011 = 1, we have that x2011 1 = 0, which factors as
(x 1)(x2010 + x2009 + + x + 1) = 0.
We know that x 6= 1, so we can divide both sides by x 1, to get
x2010 + x2009 + + x + 1 = 0.
Then
1 1 1
2S = (x + x2 + + x2010 ) + 2010 + + 2 + + 2010
x x x
2010 2009
x + x + + x
= (x + x2 + + x2010 ) + 2010 + 2011
x
1
= (1) + 2010 +
1
= 2008,
so S = 1004.
Round 3 Solutions
www.usamts.org
3/3/23. A long, 1-inch wide strip of cloth can be folded into the figure below.
The knot has outer boundary equal to a regular pentagon as shown below.
Instead, a long 1-inch wide strip of cloth is folded into the next figure, following the given
turns and crossings.
The knot has outer boundary equal to a regular heptagon. The trailing strands of the
heptagonal knot are both cut at the outer (heptagonal) boundary of the knot. Then the
knot is untied. What is the area of one side of the resulting quadrilateral of cloth? (Your
answer may contain trigonometric expressions.)
We are given that the knot, when pulled tight, forms a regular heptagon
(with two trailing strands). This tells us that the angles at each crease will all
be the interior angles of a heptagon, 5
7
. The piece of cloth between two creases
(when the strip is unfolded) is then forced to be an isosceles trapezoid with
acute angles 2
7
, and whose legs are equal to the short base. Furthermore, the height of each
trapezoid is equal to the width of the cloth, namely 1 inch.
When the trailing strands are cut and the cloth is untied, the cloth be-
5
7
2
7
5
7
comes the union of the seven trapezoids. Our goal is now to compute the
area of one of these trapezoids.
2
7
Round 3 Solutions
www.usamts.org
Each leg of the trapezoid is the hypotenuse of a right triangle with angle 2 7
and opposite
side 1, so the length of each leg is sin12 . This means that the short base has length sin12 as
7 7
well.
1
sin 2
7
1
sin 2
7 2
7
1
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual PDF Printer cos 2
7
sin 2
7
1
1
1 sin 2
7
sin 2
7
2
7
cos 2
7
sin 2
7
Furthermore, the length of the long base is then
cos 2
7 1 cos 2
7
2 cos 2
7
+1
+ + = .
sin 2
7
sin 2
7
sin 2
7
sin 27
The height of the trapezoid is 1, so the area of the trapezoid is equal to the average of
the lengths of the bases, which is
2 cos 2 +1 cos 2 +1

1 1 7 7
2 + 2 = 2 .
2 sin 7 sin 7 sin 7
Using the identity tan 2 = sin

1+cos
, we can simplify this to
cos 2
7
+1
2 = cot .
sin 7 7
Finally, the entire strip consists of seven of these trapezoids, so the area of the strip, in
square inches, is

7 cot .
7
Round 3 Solutions
www.usamts.org
4/3/23. Renata the robot packs boxes in a warehouse. Each box is a cube of side length 1
foot. The warehouse floor is a square, 12 feet on each side, and is divided into a 12-by-12
grid of square tiles 1 foot on a side. Each tile can either support one box or be empty. The
warehouse has exactly one door, which opens onto one of the corner tiles.
Renata fits on a tile and can roll between tiles that share a side. To access a box, Renata
must be able to roll along a path of empty tiles starting at the door and ending at a tile
sharing a side with that box.
(a) Show how Renata can pack 91 boxes into the warehouse and still be able to access any
box.
(b) Show that Renata cannot pack 95 boxes into the warehouse and still be able to access
any box.
(a) The following diagram shows that Renata can have access to 91 boxes.
(b) Solution 1: We say that a platform is occupied if it has a box on it, and that a
platform is reachable if it is not occupied and Renata can roll to it from the door. Two
platforms are neighboring if they share a side. We refer to platforms that have 4, 3, or 2
neighboring platforms as interior, edge, and corner platforms, respectively.
Suppose Renata has packed boxes in such a way that she can access any box. Let rI , rE ,
and rC be respectively the number of interior, edge, and corner platforms that are reachable,
and let r = rI + rE + rC be the total number of reachable platforms. Let s be the number of
unoccupied platforms that are unreachable. Thus, the number of boxes is 144 r s.
Round 3 Solutions
www.usamts.org
Let N be the number of ordered pairs (x, y) of neighboring platforms such that y is
reachable. We count N in two different ways. On one hand,
X
N= (# neighbors of y),
reachable
platforms y
which gives us N = 4rI + 3rE + 2rC .

On the other hand, we can write
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buyX
Virtual PDF Printer
N= (# reachable neighbors of x),

all plat-
forms x
then split the terms of this sum into two groups:

X X
N= (# reachable neighbors of x) + (# reachable neighbors of x).
occupied unoccupied
platforms x platforms x
Each term in the first sum is at least 1, since every box must neighbor a reachable platform.
Thus, the first sum is at least 144 r s. To bound the second term, we note that if we
think of the platforms as vertices of a graph with edges connecting neighboring platforms,
then the reachable platforms form a connected subgraph. This subgraph has r vertices, so
it has at least r 1 edges, making the second sum greater than or equal to 2r 2 (since this
sum counts pairs of neighboring reachable platforms with order).
Therefore, N (144 r s) + (2r 2) 142 + r s. Comparing this to our first way
of counting N , we get
4rI + 3rE + 2rC 142 + r s.
Subtracting r from both sides gives
3rI + 2rE + rC 142 s,
and so
3r = 3rI + 3rE + 3rC 142 s + rE + 2rC .
In particular, we have
3r + 3s 142 + 2s + rE + 2rC
r+s= .
3 3
Note that rC 1, since the platform with the door is reachable, and rE + 2s 4, since every
corner must either have a reachable neighbor or be unoccupied and unreachable. Thus,
148
r+s > 49.
3
Round 3 Solutions
www.usamts.org
Since r + s is the number of unoccupied platforms, we know there are at least 50 platforms
without boxes, and thus no more than 94 with boxes.
Solution 2: Let f (n) be the maximum number of boxes that Renata can reach, where n
is the number of tiles that she can roll onto. Note that any configuration of the first 3 squares
of the path allows Renata to reach 3 boxes, so f (3) = 4. Each additional square added to
the path allows an increase of at most 2 to the number of reachable boxes: it removes the
box on the paths added square, but that added square potentially reaches 3 new boxes. So,
f (n) 2n 2 for all n 3. Noting that there are 144 tiles total, we have

2n 2 if 3 n 48,
f (n) min {144 n, 2n 2} =
144 n if n 49.
The only positive integer n for which f (n) 95 is n = 49, so to reach 95 boxes, we must have
a 49-tile path. By the initial construction, such a path could optimally reach 2(49) 2 = 96
boxes. If we can lower this bound for a 49-tile path by 2, we will have shown that no path
of length 49 can reach 95 or more boxes. Therefore, there exists no path that reaches 95 or
more boxes.
In order for Renata to reach 95 boxes on a 49-tile path, since 95 + 49 = 144, Renatas
path must contain at least one neighbor of each of the other three corner tiles (either to
access a box on that tile, or to reach that tile in her path). However, whenever Renata adds
one of these three tiles that neighbor a corner to her path, she only increases the number
of reachable squares by at most 1 (the wall of the warehouse taking the place of the second
new reachable tile). Therefore, Renata can reach at most 96 3 = 93 tiles from a path of
length 49.
An alternative argument goes as follows. Every time the path branches (that is, a new
square is added to the path anywhere other than at the end of the path), only at most
1 additional box is reached since at least 1 of the three new boxes reached by the new
square was already reachable by the path). Every time the path turns a corner, only at most
1 additional box is reached since at least 1 of the three new boxes reached by the new
square was again already reachable by the path. Because the warehouse is a 12 12 grid,
a 49-square path must turn or branch at least twice after the first three tiles. Such a path
can reach at most 96 2 = 94 boxes, so a path that reaches 95 boxes is not possible.
Round 3 Solutions
www.usamts.org
5/3/23. Let k > 2 be a positive integer. Elise and Xavier play a game that has four steps, in
this order.
1. Elise picks 2 nonzero digits (1-9), called e and f .

2. Xavier then picks k nonzero digits (1-9), called x1 , . . . , xk .
3. Elise picks any positive integer d.
4. Xavier picks an integer b > 10.
Each players choices are known to the other player when the choices are made.
The winner is determined as follows. Elise writes down the two-digit base b number efb .
Next, Xavier writes the k-digit base b number that is constructed by concatenating his digits,
(x1 . . . xk )b .
They then compute the greatest common divisor (gcd) of these two numbers. If this gcd is
greater than or equal to the integer d then Xavier wins. Otherwise Elise wins.
(As an example game for k = 3, Elise chooses the digits (e, f ) = (2, 4), Xavier chooses
(4, 4, 8), and then Elise picks d = 100. Xavier picks base b = 25. The base-25 numbers 2425
and 44825 are, respectively, equal to 54 and 2608. The greatest common divisor of these two
is 2, which is much less than 100, so Elise wins handily.)
Find all k for which Xavier can force a win, no matter how Elise plays.
We claim that Xavier has a winning strategy if and only if k is even.

Let k be an even integer. After Elise chooses her digits e and f , Xavier can set his digits
to be x1 = e, x2 = f , x3 = e, x4 = f , . . . , xk1 = e, and xk = f . Then Elises number
efb divides Xaviers number ef ef . . . efb , so their greatest common divisor is simply Elises
number efb = be + f .
Then no matter what positive integer d Elise chooses, Xavier can choose an integer b
sufficiently large so that be + f d, and Xavier wins.
Now, let k be an odd integer. Elise can start by choosing e = 1 and f = 9. Let Xavier
choose his digits x1 , x2 , . . . , xk . Let F (t) = t + 9 and
G(t) = x1 tk1 + x2 tk2 + + xk1 t + xk .
Note that F (b) and G(b) are equal to Elises number and Xaviers number in base b, re-
spectively. We will show there exists a constant M such that gcd(F (b), G(b)) < M for all
integers b > 10.
Round 3 Solutions
www.usamts.org
When the polynomial F (t) is divided into the polynomial G(t), we obtain a quotient Q(t)
and remainder R(t), so
G(t) = F (t)Q(t) + R(t).
Both G(t) and F (t) have integer coefficients, and the leading coefficient of F (t) is 1, so both
Q(t) and R(t) have integer coefficients as well. Furthermore, F (t) is a linear polynomial, so
the remainder R(t) is a constant, say c.
We will show that this constant c is positive. Since F (t) = t + 9, we can set t = 9 in
the equation above, to get
c = G(9) = x1 (9)k1 + x2 (9)k2 + + xk2 (9)2 + xk1 (9) + xk .
Since k is odd, and each xi is a digit from 1 to 9, this expression is minimized when x1 =
x3 = = xk = 1 and x2 = x4 = = xk1 = 9. Hence,
c = G(9)
= x1 (9)k1 + x2 (9)k2 + + xk2 (9)2 + xk1 (9) + xk
(9)k1 + 9 (9)k2 + (9)k3 + 9 (9)k4 + + (9)2 + 9 (9) + 1
= 1.
Therefore, c is positive.
Now, setting t = b in the equation above, we get
G(b) = F (b)Q(b) + c.
Since gcd(F (b), G(b)) divides both F (b) and G(b), gcd(F (b), G(b)) also divides G(b)F (b)Q(b) =
c. In particular, gcd(F (b), G(b)) must be less than or equal to c, for any integer b > 10.
Elise can then choose any integer d greater than c (say d = c + 1). Then no matter what
base b Xavier chooses, gcd(F (b), G(b)) will be less than or equal to c, so it will be less than
d. Thus, Elise has a winning strategy when k is odd.
Credits: Problem 1/3/23 and 5/3/23 proposed by Palmer Mebane.


Round 1 Solutions
www.usamts.org
1/1/24. Several children were playing in the ugly tree when suddenly they all fell.
Roger hit branches A, B, and C in that order on the way down.
Sue hit branches D, E, and F in that order on the way down.
Gillian hit branches G, A, and C in that order on the way down.
Marcellus hit branches B, D, and H in that order on the way down.
Juan-Phillipe hit branches I, C, and E in that order on the way down.

Poor Mikey hit every branch A through I on the way down. Given only this information, in
how many different orders could he have hit these 9 branches on the way down?
Let X < Y mean that Mikey must hit branch X before branch Y . Then the given
data is:
A<B<C (1),
D<E<F (2),
G<A<C (3),
B<D<H (4),
I<C<E (5).
Combining (1), (3) and the second half of (5) gives
G<A<B<C<E (6),
and combining (2) and the first half of (4) gives
B<D<E<F (7).
Combining (6) and (7), we find that C and D must both be between B and E, but can
appear in either order. So we must have either (8) or (9) below:
G<A<B<C<D<E<F (8),
G<A<B<D<C<E<F (9).
We only have to insert H and I into these chains. H must go after D, and I must go before
C. We now break into cases based on whether (8) or (9) is true.
In case (8): We have 3 choices for where to insert H and 4 choices for where to insert I.
These choices are independent, for a total of 3 4 = 12 possibilities.
Round 1 Solutions
www.usamts.org
In case (9): We have 4 choices for where to insert H and 5 choices for where to insert I.
These choices are independent, but if we place both H and I between D and C, they can go
in either order. Thus we have a total of 4 5 + 1 = 21 possibilities.
Thus, there are a total of 12 + 21 = 33 possibilities.
Round 1 Solutions
www.usamts.org
2/1/24. Three wooden equilateral triangles of side length 18 inches

are placed on axles as shown in the diagram to the right. Each axle
is 30 inches from the other two axles. A 144-inch leather band is
wrapped around the wooden triangles, and a dot at the top corner is
painted as shown. The three triangles are then rotated at the same
speed and the band rotates without slipping or stretching. Compute
the length of the path that the dot travels before it returns to its
initial position at the top corner. 30
Let a third of a turn of the triangles be called a thirn, and

assume the triangles are turning clockwise. Label the triangles A
A, B, C as in Figure 1. After each thirn, the dots position will
be exactly 18 inches farther along the perimeter of the band.
Because the band is 144 inches long, it will take 144/18 = 8
thirns for the dot to return to its original position. As this is
an integer, the dot will return to its original position after only C B
one complete rotation of the band.
During the first thirn, the dot follows the path of the vertex 30
of triangle A, tracing a third of the circumcircle of triangle A. Figure 1
The radius ofthis circle is the distance from center to the vertex of thetriangular
wheels,
which is 18/ 3. Thus, the distance traveled by the point is 2/3 18/ 3 = 4 3 inches.
(See Figure 1.)
During the second thirn, notice that the distance between

the top corners of triangles A and B will remain 30 inches A
throughout the duration of the thirn. The 30 inch part of the
band between these two corners will thus remain parallel to the
sides of the original triangle, and each point along the band will
be translated the same amount as the corners of the triangles
A and B. Figure 2 shows this visually, with the paths of the C B
triangle corners shown in red, a snapshot of the bands position
drawn as a black segment between the red arcs, and the blue 30
arc showing the path of the dot. Since the blue arc traces the Figure 2
same path shape as the corners of the triangles A and B, the
dot moves the same distance as it did in the first thirn, which is 4 3 inches.
Round 1 Solutions
www.usamts.org
The dot ends the second thirn 6 inches away from the top
corner of triangle B, which means it is also 6 inches from the A
center of triangle B. During the third thirn, the dot stays on
triangle B, tracing a third of a circle with radius whose radius
can be computed to be 6, traveling 2/3 6 = 4 inches. (See
Figure 3.)
C B
For reasoning identical to the second thirn, for the fourth
thirn the dot also moves 4 3 inches, ending up in center of the
bottom section of the band. This means that thirns 5 through 30
8 are a reflection of thirns 1 through 4, so the total distance Figure 3

traveled by the dot is 6 4 3 + 2 4 = (24 3 + 8) . Figure 4 shows the full path the
dot travels.
C B
30
Figure 4
Round 1 Solutions
www.usamts.org
3/1/24. The symmetric difference, 4, of a pair of sets is the set of

elements in exactly one set. For example,
{1, 2, 3}4{2, 3, 4} = {1, 4}.
There are fifteen nonempty subsets of {1, 2, 3, 4}. Assign each subset
to exactly one of the squares in the grid to the right so that the following
conditions are satisfied.
(i) If A and B are in squares connected by a solid line then A4B has exactly one element.
(ii) If A and B are in squares connected by a dashed line then the largest element of A is
equal to the largest element of B.
The image to the left shows a numbering of the 15 1 2 3 4

boxes that we will refer to in this solution.
Before we begin, we note a certain logical configu-
ration, which we will refer to as an isosceles triangle, 5 6 7 8
in which we have three squares A, B, C where A, B are
connected by a dashed line and C is connected by a
solid line to each of A and B. An example of an isosce-
9 10 11 12
les triangle is (A, B, C) = (8, 11, 12). In an isosceles
triangle, not only do A and B have the same max-
imum element x, but C has that maximum element
also. This is because if C had a different maximum 13 14 15
y, then either CA = CB = {y} (if y > x) or
CA = CB = {x} (if y < x). Either case implies
A = B, contradicting the condition that all squares have a distinct subset.
By enumerating all of the 15 subsets, we can determine that there are:
8 subsets with largest element 4,
1 subset with largest element 1.
Round 1 Solutions
www.usamts.org
Note that squares 8, 11, and 14 must have the same largest element, and since they are
part of isosceles triangles, 12 and 15 must have the same largest element as these three. So
squares 8, 11, 12, 14, and 15 must all have the same largest element. There are not five
subsets with largest element 1, 2, or 3, so largest element of squares 8, 11, 12, 14, and 15
must be 4.
Similarly, squares 1, 5, 6, and 10 all have the same largest element since (1, 6, 5) and
(5, 10, 6) are isosceles triangles. If this largest element were 4, then nine subsets would have
largest element 4, which is too many. On the other hand, there are not four subsets with
largest
Create PDF with GO2PDF for free,element 1 orthis2.line,So
if you wish to remove theto buy
click here largest
Virtual PDFelement
Printer of squares 1, 5, 6, and 10 is 3. Squares 2, 3,
and 7 also form an isosceles triangle and have the same largest element. We know where all
sets with largest element 3 are, and this is too many subsets have to have largest element 1
or 2. So squares 2, 3, and 7 must have largest element 4. Out of the three remaining squares,
9 and 13 have a dashed line, so they have largest element 2, and 4 has largest element 1.
Finally, note that a solid line always joins a set 1 2 3 4

with an odd number of elements to a set with an even 3 E 4 O 4 E 1 O
number of elements, and that seven subsets have an

even number of elements. Therefore, squares 1, 3, 6,
5 6 7 8
8, 9, 11, 14 have an even number of elements, and the 3 O 3 E 4 O 4 E
other squares have an odd number of elements.
The diagram shown collects all the information we
9 10 11 12
have so far. A number in a box is the largest element, 2 E 3 O 4 E 4 O
an E means the boxs subset has an even number of
elements, and an O means it has an odd number of
elements. 13 14 15
2 O 4 E 4 O
The information in the diagram will drive the rest
of our progress. We immediately get square 4 is {1}. Since we know square 9 has an even
number of elements, it has {1, 2}, while square 13 has {2}. These three squares are all
connected to ones with a higher maximum element, so that maximum element must be the
one added, and squares 3, 5, 14 have sets {1, 4}, {1, 2, 3}, and {2, 4} respectively.
Square 10 is the only other square with maximum element 3 and an odd number of
elements, so it has {3}. Then square 11, which has max element 4, is {3, 4}. The only
square with maximum element 4 and an even number of elements we have yet to determine
is square 8; it must have {1, 2, 3, 4}.
Now consider the possible locations of {4}. It cant touch {1, 2, 3, 4} eliminating squares
7 and 12. If square 2 has {4}, then there is no way for square 1 to have maximum element
3. Therefore, square 15 has {4}. Likewise, {2, 3, 4} must go in one of 2, 7, and 12 but cant
touch {1, 4}, so it goes in square 12. This tells us that squares 2 and 7 must contain {1, 2, 4}
and {1, 3, 4} in some order.
Round 1 Solutions
www.usamts.org
Since square 1 has largest element 3 and not 4, the edge from square 2 to square 1 must
remove a 4, so square 2 has {1, 3, 4} and square 1 has {1, 3}. Then square 6, the last square
with maximum element 3 has {2, 3}. From our determining of square 2, we also now know
that square 7 is {1, 2, 4}. This is the last box we needed to determine. The unique solution
we have found is shown below.
1 2 3 4
1 3 1 34 1 4 1
5 6 7 8
123 23 12 4 1234
9 10 11 12
12 3 34 234
13 14 15
2 2 4 4
Round 1 Solutions
www.usamts.org
4/1/24. Let bxc denote the greatest integer less than or equal to x. Let m be a positive integer,
m 3. For every integer i with 1 i m, let
m
2 1 mi
Sm,i = n2 + 1 : n = 1, 2, 3, . . . .
2i1
For example, for m = 3,
S3,1 = {b7n 3c : n = 1, 2, 3, . . . }
= {4, 11, 18, . . . },

7
S3,2 = n 1 : n = 1, 2, 3, . . .
2
= {2, 6, 9, . . . },

7
S3,3 = n : n = 1, 2, 3, . . .
4
= {1, 3, 5, . . . }.
Prove that for all m 3, each positive integer occurs in exactly one of the sets Sm,i .
Define
2m 1
n 2mi + 1.
f (m, i, n) =
2i1
so that Sm,i = {bf (m, i, n)c : n = 1, 2, 3, . . .}. Let
Tm,i = {bf (m, i, n)c : n = 1, 2, 3, . . . , 2i1 }.
We claim that Tm,i is the set of numbers less than 2m that end with exactly m i zeroes
when written in binary. Rewrite f (m, i, n) as
2m 1 mi mi 2i1 n
f (m, i, n) = n 2 + 1 = 2 (2n 1) + .
2i1 2i1
On the right side the first term is an integer, and since 1 n 2i1 the second term is
at least 0 and less than 1. Therefore the floor of the expression is equal to the first term,
meaning that bf (m, i, n)c = 2mi (2n 1).
A number ends in exactly m i zeroes if and only if it is an odd multiple of 2mi , and
we see above that bf (m, i, n)c ranges over the first 2i1 odd multiples of 2mi as n ranges
from 1 to 2i1 . The 2i1 th odd multiple is 2m 2mi , the last one less than 2m . This proves
our claim that Tm,i is the set of numbers less than 2m which end with exactly m i 0s when
written in binary.
Round 1 Solutions
www.usamts.org
Now notice that

2m 1
f (m, i, n + 2i1 ) = (n + 2i1 2mi + 1
2i1
2m 1
= i1 n 2mi + 2m
2
= f (m, i, n) + (2m 1).
Therefore,
S
Create PDF with GO2PDF for free, if you wish to remove this line,m,i
= {t + k(2m 1) : t Tm,i , k = 1, 2, 3, . . .}.
click here to buy Virtual PDF Printer
This means a number is in Sm,i if and only if its remainder when divided by 2m 1 is in Tm,i
(where we treat a remainder of 0 as a remainder of 2m 1). Since each number from 1 to
2m 1 occurs in exactly one of the Tm,i , this means every positive integer occurs in exactly
one of the Sm,i , as desired.
Round 1 Solutions
www.usamts.org
5/1/24. An ordered quadruple (y1 , y2 , y3 , y4 ) is quadratic if there exist real

numbers a, b, and c such that
yn = an2 + bn + c
for n = 1, 2, 3, 4.
Prove that if 16 numbers are placed in a 4 4 grid such that all four rows are quadratic
and the first three columns are also quadratic then the fourth column must also be quadratic.
(We say that a row is quadratic if its entries, in order, are quadratic. We say the same
for a column.)
We start with an important lemma.

Lemma: (y1 , y2 , y3 , y4 ) is quadratic if and only if y4 3y3 + 3y2 y1 = 0.
Proof: Suppose (y1 , y2 , y3 , y4 ) is quadratic via the quadratic polynomial
f (x) = ax2 + bx + c
so that yn = f (n) for n = 1, 2, 3, 4. Note that for all n,
f (n + 1) f (n) = a(n + 1)2 + b(n + 1) + c (an2 + bn + c) = 2an + a + b.
Let g(n) = 2an + a + b. Then
g(n + 1) g(n) = 2a(n + 1) + a + b (2an + a + b) = 2a.
Plugging in n = 3 gives
2a = g(3) g(2) = (f (3) f (2)) (f (2) f (1)) = y3 2y2 + y1 , (1)
and plugging in n = 4 gives
2a = g(4) g(3) = (f (4) f (3)) (f (3) f (2)) = y4 2y3 + y2 . (2)
Thus y3 2y2 + y1 = y4 2y3 + y2 , hence y4 3y3 + 3y2 y1 = 0.

Conversely, it is well-known given any triple (y1 , y2 , y3 ) we can find a quadratic f (x) =
2
ax + bx + c such that f (n) = yn for n = 1, 2, 3. Then equations (1) and (2) tell us that
f (4) 3y3 + 3y2 y1 = 0. So we must have f (4) = y4 , meaning (y1 , y2 , y3 , y4 ) is quadratic.
2
Let ai,j be the number in the ith row and jth column for 1 i, j 4. Define for each
i = 1, 2, 3, 4:
ri = ai,1 3ai,2 + 3ai,3 ai,4 , ci = a1,i 3a2,i + 3a3,i a4,i .

Round 1 Solutions
www.usamts.org
By the lemma, row i is quadratic if and only if ri = 0, and column i is quadratic if and only
if ci = 0. We are given that ri = 0 for all i and ci = 0 for i {1, 2, 3}, and we wish to show
c4 = 0.
Define the two quantities
R = r1 3r2 + 3r3 r4 , C = c1 3c2 + 3c3 c4 .
Suppose we expand out each of these sums in terms of the ai,j . We find that
R= a1,1 3a1,2 + 3a1,3 a1,4
3a2,1 + 9a2,2 9a2,3 + 3a2,4
+ 3a3,1 9a3,2 + 9a3,3 3a3,4
a1,1 + 3a1,2 3a1,3 + a1,4 = C
3
3
(The coefficient in front of ai,j can be written explicitly as i1 j1
(1)i+j . Notice that
this is symmetric in i and j; this is the reason behind R and C giving the same expression.)
Thus, we have learned that R = C. Since r1 = r2 = r3 = r4 = 0 and c1 = c2 = c3 = 0,
we have from the original definitions of R and C that R = 0 and C = c4 . This gives us
c4 = 0, as desired.
Remark : The ideas in this solution can be used to solve a generalized version of this
problem, where the grid is n n and the values of n rows and n 1 columns are the first n
values of a polynomial of degree at most n 2.

Round 2 Solutions
www.usamts.org
1/2/24. Fill in each of the ten boxes with a number so

0 1 2
that the following conditions are satisfied.
1. Every number has three distinct digits that sum

to 15. 0 may not be a leading digit. One digit 3 4 5 6
of each number has been given to you.
2. No two numbers in any pair of boxes have the

same unordered set of three digits. For example, 7 8 9
it is not allowed for two different boxes to have
the numbers 456 and 645.
3. Two boxes joined by an arrow must have two numbers which share an equal hundreds
digit, tens digit, or ones digit. Also, the smaller number must point to the larger.
Solution
To begin, notice that there are exactly 10 sets of three distinct unordered digits that sum
to 15:
{0,6,9} {0,7,8} {1,5,9} {1,6,8} {2,4,9}

{2,5,8} {2,6,7} {3,4,8} {3,5,7} {4,5,6}
Since we have ten boxes to fill and each box must use a different set of digits, each of the
digit sets above must be used for exactly one box.
Call the box with the digit d given the d-box. We know the 0-box either contains the
digits {0, 6, 9} or {0, 7, 8}. However, the 0-box has an arrow to the 3-box, so one of the digits
in the 0-box is also contained in a number using the digit 3. None of 0, 6, and 9 can be
contained as a digit in a box using the number 3. Therefore, the 0-box cannot use {0, 6, 9}
as its digit set, and must use {0, 7, 8} instead.
Now consider the digit shared between the 0-box and 1-box. The only possible sets for
the 1-box are {1, 5, 9} and {1, 6, 8}. Only {1, 6, 8} shares a digit with the 0-boxs set {0, 7, 8},
so the 1-boxs digit set is {1, 6, 8}. Additionally, we also know that the 1-box has the greater
number. Since the 0-boxs number cant begin with 0, it is at least 700. The only digit the
1-box has that is at least 7 is 8, so 8 must be the hundreds digit. As 8 is the shared digit
between the 0-box and 1-box, the 0-box also has 8 as the hundreds digit. So the 0-box is
either 807 or 870. The latter would be greater than anything the 1-box can be, so the 0-box
is 807.
Round 2 Solutions
www.usamts.org
We also have that the 3-box must have a number greater than 807. It cant have the
digit 9, so its hundreds digit must be 8, and that is also the shared digit with the 0-box.
The 3-box is either 834 or 843.
We now find the digit sets of the remaining boxes:
Consider the 8-box. The only remaining digit set containing 8 is {2, 5, 8}, so this is
the digit set of the 8-box.
Consider the 7-box, which shares a digit with the sets {3, 4, 8} and {2, 5, 8}. The
only remaining set that contains 7 and one digit from each of these sets is {3, 5, 7}.
Therefore box 7 has {3, 5, 7} as its set.
Consider the 4-box. It is larger than the 3-boxs number that is at least 800. However
all of the digit sets with 8 have been used up, so the 4-box must have 9 as its hundreds
digit and thus the set {2, 4, 9}.
Consider the 9-box. The only unused digit sets with 9 are {1, 5, 9} and {0, 6, 9}. Only
{1, 5, 9} shares a digit with the 8-boxs set {2, 5, 8}, so the 9-boxs set is {1, 5, 9}.
Consider the 5-box. The only unused digit set with 5 is {4, 5, 6}, so this the 5-boxs
set.
Only the 2-box and 6-box remain, containing digit sets {2, 6, 7} and {0, 6, 9} between
them. As only one has the digit 2, the 2-box must have {2, 6, 7} and then the 6-box
has {0, 6, 9}.
At this point we have determined the information shown in the image below.
Next, we will determine the hun- 816
dreds digit of each number. We al- 807 861
{2, 6, 7}
ready know the 0-box, 1-box, and 3-
box all have hundreds digit 8, and the
4-box has hundreds digit 9. Consider
the 5-box. It shares the digits 4 and 6 834 924
with the 4-box and 1-box respectively, {4, 5, 6} {0, 6, 9}
843 942
neither of which are a hundreds digit.
Therefore 5 is the hundreds digit. The
2-box, being smaller but having only
one digit less than 5, must have 2 as
its hundreds digit. In the 6-box, 0 is {3, 5, 7} {2, 5, 8} {1, 5, 9}
not allowed to be a hundreds digit and
the 6 is shared with the 5-box, leaving 9 as the hundreds digit there. As it shares the 9 with
the 9-box, the 9-box also has 9 as the hundreds digit. Finally, the 8-box shares the digits
2 and 5 with the 4-box and 9-box respectively, neither of which are hundreds digits, so the
8-box has 8 as its hundreds digit. The same logic applied to the 7-box shows it has 7 as the
hundreds digit.
Round 2 Solutions
www.usamts.org
There are now only two choices left for each undetermined box. The image below sum-
marizes these possibilities.
We are given that the 6-box, which 816 267
is 906 or 960, is less than the 9-box, 807 861 276
which is 915 or 951. Thus the 6-box is
906. From this the 5-box, 1-box, and
2-box, which all share the digit 6, must
have 6 as their units digit and be 546, 834 924 546 906
816,forand
Create PDF with GO2PDF 276
free, if you wishrespectively. Then
to remove this line, click here to buy the 4- Printer
Virtual PDF
843 942 564 960
box and 3-box, which both share the
digit 4 with the 5-box, must have 4 as
their tens digit and be 942 and 843 re-
spectively. The 7-box is 753 since it
735 825 915
shares the 3 with the 3-box. Finally,
753 852 951
the 8-box and 9-box both share the digit
5 with the 7-box, so they are 852 and 951 respectively.
This gives the unique solution shown below. It can be verified that all of the constraints
in the problem are satisfied by this configuration.
807 816 276
843 942 546 906
753 852 951

Round 2 Solutions
www.usamts.org
2/2/24. Find all triples (a, b, c) of positive integers with a b c such that

1 1 1
1+ 1+ 1+ = 3.
a b c
Solution
Notice that since a b c,

3
1 Virtual PDF Printer
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy 1 1 1
1+ 1+ 1+ 1+ ,
a b c a
so 3
1
1+ 3.
a
1 3
is decreasing. Also, (1 + 13 )3 = 64

As a function in a, 1 + a 27
< 3, so a 2.
Case 1. a = 1. In this case, the given equation becomes

1 1
2 1+ 1+ = 3.
b c
This equation simplifies to bc 2b 2c = 2. By adding 4 to both sides and factoring, this

equation becomes
(b 2)(c 2) = 6.
Since 1 b c, we get b 2 = 1 and c 2 = 6, or b 2 = 2 and c 2 = 3. This leads to
the solutions (a, b, c) = (1, 3, 8) or (1, 4, 5).
Case 2. a = 2. In this case, the given equation becomes

3 1 1
1+ 1+ = 3.
2 b c
This equation simplifies to bc b c = 1. By adding 1 to both sides and factoring, this

equation becomes
(b 1)(c 1) = 2.
Since 2 b c, we get b 1 = 1 and c 1 = 2. This leads to the solution (a, b, c) = (2, 2, 3).
Hence, the three solutions are (a, b, c) = (1, 3, 8), (1, 4, 5), or (2, 2, 3).
Round 2 Solutions
www.usamts.org
3/2/24. Let f (x) = x x1 , and define f 1 (x) = f (x), f n (x) = f (f n1 (x)) for n 2. For each
n, there is a minimal degree dn such that there exist polynomials p and q with f n (x) = p(x) q(x)
and the degree of q is equal to dn . Find dn .
Solution
pn (x)
Let f n (x) = qn (x)
, where pn and qn do not share a common factor. We claim that for all
n 1,
(pn+1 (x), qn+1 (x)) = (p2n (x) qn2 (x), pn (x)qn (x)). (1)
First, notice that
p2 (x) qn2 (x)

pn+1 (x) 1 pn (x) qn (x)
n+1 n n
= f (x) = f (f (x)) = f x = = n .
qn+1 (x) x qn (x) pn (x) pn (x)qn (x)
It remains to check that p2n (x) qn2 (x) and pn (x)qn (x) do not share a common factor. If
they did, then p2n (x) qn2 (x) must share a common factor d(x) with either of pn (x) or qn (x).
However, if d(x) divides both p2n (x) qn2 (x) and pn (x), this would imply that d(x) also
divides qn2 (x). This implies the existence of a common factor dividing both pn (x) and
qn (x), contradicting our definition that pn (x) and qn (x) have no common factor.
We can obtain a similar contradiction in the case that d(x) divided both p2n (x) qn2 (x)
and qn (x). We conclude that p2n (x) qn2 (x) and pn (x)qn (x) do not have any common factors.
This completes the proof of equation (1).
Next, we prove by induction on n that pn has degree 2n and qn has degree 2n 1 for all
2
n 1. For the base case n = 1, notice that f 1 (x) = x x1 , so p1 (x) = x2 1 and q1 (x) = x.
We see that pn has degree 21 = 1 and qn has degree 21 1 = 1, so the base case has been
verified.
Now assume that for some particular value of n, pn has degree 2n and qn has degree 2n 1.
By (1), we have pn+1 (x) = p2n (x) qn2 (x). The first term has degree 2 2n = 2n+1 and the
second term is of smaller degree, so the degree of the difference must be 2n+1 . Additionally,
(1) gives that qn+1 = pn qn , so by the inductive hypothesis it has degree 2n +(2n 1) = 2n+1 1.
This completes the induction.
Since dn is equal to the degree of qn , the answer to the original problem is 2n 1.
Round 2 Solutions
www.usamts.org
4/2/24. Let n be a positive integer. Consider an n n grid of unit squares. How

many ways are there to partition the horizontal and vertical unit segments of
the grid into n(n + 1) pairs so that the following properties are satisfied?
(i) Each pair consists of a horizontal segment and a vertical segment that
share a common endpoint, and no segment is in more than one pair.
(ii) No two pairs of the partition contain four segments that all share the same endpoint.
(Pictured above is an example of a valid partition for n = 2.)

Solution
The grid contains (n + 1)2 vertices which are intersections of gridlines. Call such a vertex
a joint if one of the pairs in the partition has this vertex as a common endpoint of its two
segments. By condition (ii) no two pairs of the partition have the same joint. There are
n(n + 1) pairs and (n + 1)2 vertices, so n + 1 of the vertices are not joints. We will refer to
these n + 1 vertices as isolated vertices.
Suppose there is a row or column of the grid in which all n + 1 vertices are joints. If it is
a row, then each of the joints must use a horizontal segment of that row, but there are n + 1
joints and only n segments, a contradiction. We get a similar contradiction for a column
since there will be n + 1 joints but only n vertical segments. Therefore each row and column
must have at least one isolated vertex. Since there are n + 1 rows and columns of vertices
and n + 1 total isolated vertices, each row and column has exactly one isolated vertex.
Call a set S of n + 1 vertices in an n n grid allowable if
each row and column of vertices contains exactly one element
of S. We claim that for every allowable set S, there exists a
unique valid partition that has the vertices of S as its isolated
vertices. Take an arbitrary endpoint P that is a joint. Its row
and column both contain exactly one unused endpoint. Have
P s horizontal segment point toward the unused endpoint of
the row, meaning that if the unused endpoint is to the left of
P , then P s horizontal segment should be to the left of P also.
Similarly have P s vertical segment point toward the unused
endpoint of the column. The image to the right shows a valid partition for n = 5 and has
arrows drawn on the segments to clarify where they are pointing.
This construction will not have any segment of the grid be used by two endpoints. If
a segment were reused, then it pointed in one direction for one of its endpoints and the
other direction for the other endpoint, implying that there is an unused endpoint in both
directions of the segment. This contradicts our earlier claim that no row or column contains
two unused endpoints. Additionally, every segment of the grid will be used by this partition.
This is because there are 2n(n + 1) total horizontal and vertical segments of the grid, and
we assigned 2 of them to each of n(n + 1) endpoints without any collisions.
Round 2 Solutions
www.usamts.org
Additionally, this construction is unique. A different construction would have required a

joint P to have a segment which points away from the unused endpoint of the row or column.
Call such a segment bad. Without loss of generality, suppose a joint P has a bad horizontal
segment ` in a row. This bad segment points at another vertex of the row, which we will
call Q. Q is a joint, so it is the shared endpoint of a horizontal segment. This segment
must also point away from the isolated vertex of the row and thus be bad. We can continue
this reasoning to find a sequence of bad segments until we eventually point at a vertex R on
the edge of the grid. R is a joint, but there is no available horizontal segment that it is an
endpoint of, a contradiction. Thus, all horizontal segments must point toward the isolated
vertex of their row and all vertical segments must point toward the isolated vertex of their
column, meaning the construction we proposed is the only one possible.
Since every allowable set S gives a unique valid partition with the vertices of S as isolated
vertices, there is a 1-1 correspondence between allowable sets and valid partitions. It remains
to count the number of allowable sets. Each row and column has exactly one vertex in S.
There are n + 1 choices for the vertex in the first row. In the second row, we can then include
in S any of the n vertices that are not in the same column as the vertex chosen in the first
row. Then the third row has n 1 choices for the vertex in S, since we cannot use the
two vertices in the same column as those chosen from the previous two rows. This pattern
continues, with n + 1 k ways to choose the vertex in the kth row. From this we obtain a
total of (n + 1) n (n 1) 1 = (n + 1)! choices for an allowable set S. This is the number
of valid partitions.
Round 2 Solutions
www.usamts.org
5/2/24. A unit square ABCD is given in the plane, with O being the intersection of its
diagonals. A ray ` is drawn from O. Let X be the unique point on ` such that AX +CX = 2,
and let Y be the point on ` such that BY + DY = 2. Let Z be the midpoint of XY , with
Z = X if X and Y coincide. Find, with proof, the minimum value of the length of OZ.
Solution

Place the points on the coordinate plane so that O is the origin, A = 22 , 0 , B =

0, 22 , C = 2 2 , 0 , and D = 0, 22 . Let be the real number in [0, 2) such
that the ray ` makes an angle of with the x-axis, meaning that ` is the set of points
{(r cos , r sin ) | r 0}.
The point X can be at any point on the plane whose sum of distances from the points
A and C is 2. The set of such points is an ellipse with foci A and C. Let EX be this ellipse.
The center of the ellipse is O, since this is the midpoint of its foci A and C. The major axis
of EX is along the x-axis since both of its foci are on the x-axis, and the minor axis is along
the y-axis since it is perpendicular to the major axis. Notice that since AB +CB = 1+1 =2
and AD + CD = 1 + 1 = 2, B and D both lie on E X . Since the length of BD is 2, the
length of the minor axis of EX is 2. Since the sum of the distances to the foci is 2, the
length of the major axis is 2. From the center of the ellipse and the length of its major and
minor axes, we get that the equation for EX is
x2 + 2y 2 = 1.
The coordinates of the point X must satisfy this equation. We can do similar reasoning for
set of possible positions for the point Y , finding that it forms an ellipse with foci B and D
and center O, with the same lengths for the major and minor axes as EY . Then the equation
for EY , which the coordinates for the point Y must satisfy, is
2x2 + y 2 = 1.
Let rX and rY be the positive real numbers such that X = (rX cos , rX sin ) and Y =
(rY cos , rY sin ). Using our equations for EX and EY , we have
2
rX (cos2 + 2 sin2 ) = 1,
rY2 (2 cos2 + sin2 ) = 1.

Simplifying, we get that rX = 1
and rY = 1
1+cos2
.
1+sin2
rX +rY
We are looking for the minimum length of OZ, which is equal to 2
. Therefore, we
Round 2 Solutions
www.usamts.org
need to minimize rX + rY . Note that

1 1 2
(rX + rY )2 = 2 + 2
+p
1 + sin 1 + cos 2
(1 + sin )(1 + cos2 )
3 2
= 2 2
+p
(1 + sin )(1 + cos ) 2
(1 + sin )(1 + cos2 )
3 2
= 2 +
2 + sin cos2 2 + sin2 cos2
3 2
= +
2 q
2 + sin 4(2) 2
2 + sin 4(2)
12 4
= 2 +p .
8 + sin (2) 8 + sin2 (2)
Thus it suffices to maximize sin2 (2). Since sin2 (2) 1, we set sin2 (2) = 1. This occurs
whenever makes an angle of 45 with one of the axes, meaning q ` is parallel
q to a side of the
rX +rY
square. For these values of , (rX + rY )2 = 38 , so OZ = 2
= 2
3
, and 2
3
is our answer.

Round 3 Solutions
www.usamts.org
1/3/24. In the 8 8 grid shown, fill in 12 of the grid cells with the 24 1 3 20 13 11
numbers 112 so that the following conditions are satisfied:
3
1. Each cell contains at most one number, and each 18
number from 112 is used exactly once.
21
2. Two cells that both contain numbers may not touch,
even at a point. 20
3. A clue outside the grid pointing at a row or column gives 13
the sum of all of the numbers in that row or column. 3
Rows and columns without clues have an unknown sum.
Solution
Number the rows and columns 18 starting from the top and left, and denote the space
in row m and column n by RmCn. For example, the top left square is R1C1. This solution
will solve the problem without using the 18 clue in row 2, which is actually unnecessary to
have a unique solution.
The only ways to have a row or column with numbers summing to 3 is for it to have only
3 or only 1 and 2. The only way to have a row or column with numbers summing to 1 is to
have just 1. From this, we conclude that column 2 has just 1, and that column 3 has just 3,
since it cannot have 1.
Now consider the first and last row. One of them has only 24 1 3 20 13 11
3, and the other one has only 1 or 2. The one with 3 must have
3
it in column 3. The one with 1 and 2 has the 1 in column 2.
As a result, the 1 in column 2 and the 3 in column 3 must be in 18
the first and last rows. The other spaces of these two columns
cannot have numbers. To the right is an image summarizing 21
the progress so far.
20
Notice that by the second rule, a 2 2 subblock of the
13
grid can contain at most one number. Consider rows 6 and 7,
3
which have 20 and 13 as clues. Since we are not allowed to
have numbers 13 or greater, each row must have at least two numbers. But since the two
rows can be partitioned into four 2 2 subblocks each with at most number, we can have
at most four numbers in the two rows, so each row has exactly two numbers. The same
logic can be applied to columns 5 and 6 with 20 and 13 as clues to show both columns have
exactly two numbers.
Round 3 Solutions
www.usamts.org
Suppose now that R6C5, in the intersection of the row and column with 20, had a number
a. Then the second number in row 6 must be 20 a, and the same is true of column 5.
Since we cant have the same number twice, this is a contradiction, so R6C5 cannot have a
number. Similarly, R7C6 cannot have a number either.
Looking at rows 6 and 7 again, and accounting for the fact 24 1 3 20 13 11
that columns 2 and 3 cant have numbers in these two rows,
3 3
we notice that column 1 can have at most one number, and the
only way to fit 3 numbers in the last five columns without any 18
touching
Create PDF with GO2PDF is wish
for free, if you to toput a this
remove number into each
line, click here ofPDFcolumn
buy Virtual Printer 4, column 6, and
column 8. Above, we determined R7C6 cannot have a number, 21
so the only place for a number in column 6 is R6C6; call this
number A. Now consider how we might fit four numbers in 20 A
columns 5 and 6. The first four rows can have at most two 13
numbers. Row 6 has a number, so the only place for the fourth 3 1 2
number is row 8. Above, we determined that the only number in rows 1 and 8 that can
occur outside of columns 2 and 3 is a 2. Therefore, a 2 appears in either R8C5 or R8C6.
This means row 8 has the 1 and 2 and row 1 has the 3. The shown image summarizes our
deductions so far.
We know row 7 has two numbers, but there are only two spaces that remain. Therefore
these two spaces, R7C4 and R7C8, have numbers. This means the two cells in row 6 with
numbers are R6C1 and R6C6. Also, to avoid touching R7C4, the 2 must go in R8C6. Since
column 6 must sum to 13, its second number must be 11, which is the value of A in R6C6.
Then in row 6, with sum 20, the second number in R6C1 must be a 9. Now notice that
the only way to fit two numbers into column 5 is to put them both in R2C5 and R4C5.
Furthermore, since two distinct numbers from 1 to 12 can sum to at most 23, column 1 must
have at least three numbers, and the only way to fit them is to use rows 2, 4, 6.
At this point, we have determined the positions of eleven 24 1 3 20 13 11
numbers, and there is only one more. Currently column 8 has
3 3
only one number, but it needs two since the 11 has been placed
already. Similarly, row 4 has only two numbers, and the only 18 F D
two number combinations summing to 21 are 9+12 and 10+11.
Since the 11 and 9 have been used, neither works, so it needs a 21 G E H
third number. Combining these, we see the 12th number must
be in R4C8. The image shows the grid with all numbers placed 20 9 11
and gives names to all of the numbers which we have yet to 13 B C
determine the value of. 3 1 2
Round 3 Solutions
www.usamts.org
Notice that B is the only number not clued by any of the column sums. The column
sums total to 72, which is 6 less than 78, the sum of the numbers 1 through 12. Therefore
B = 6. Using the 13 clue in row 7 and the 11 clue in column 8, we then get C = 7 and
H = 4.
The four numbers that remain are 5, 8, 10, 12. We have D +E = 20, E +G = 214 = 17,
and F + G = 24 9 = 15. By considering the possible sums of pairs of these four numbers,
all of which are distinct, we find that D, E can only be 8, 12 in some order, E, G are 5, 12
in some order, and F, G are 5, 10 in some order. Together, this information gives D = 8,
E = 12, G = 5, and F = 10. This reaches the solved grid shown below, which is the only
answer to the problem.
24 1 3 20 13 11
3 3
18 10 8
21 5 12 4
20 9 11
13 6 7
3 1 2
Remark: As noted, this solution did not make use of the 18 clue in row 2. If one takes
it into account, then the row clues sum to 78, so there can be no numbers in rows 3 and 5.
This provides some alternate, easier methods to find the positions of the twelve numbers in
the grid.
Round 3 Solutions
www.usamts.org
2/3/24. Palmer and James work at a dice factory, placing dots on dice. Palmer builds his
dice correctly, placing the dots so that 1, 2, 3, 4, 5, and 6 dots are on separate faces. In a
fit of mischief, James places his 21 dots on a die in a peculiar order, putting some nonneg-
ative integer number of dots on each face, but not necessarily in the correct configuration.
Regardless of the configuration of dots, both dice are unweighted and have equal probability
of showing each face after being rolled.
Then Palmer and James play a game. Palmer rolls one of his normal dice and James
rolls his peculiar die. If they tie, they roll again. Otherwise the person with the larger roll
is the winner. What is the maximum probability that James wins? Give one example of a
peculiar die that attains this maximum probability.
Solution
Let p be the probability James wins the game after it ends. Note that if Palmer and James
tie on a roll, then by definition James has probability p of winning in the subsequent rolls
that follow. Let f (k) be the probability that if James rolls a k on his die, he will eventually
be the winner. Note that f (0) = 0 and f (k) = 1 for k > 6. For k {1, 2, 3, 4, 5, 6}, we have
f (k) = (k1)+p
6
. This is because Palmers die has k 1 numbers less than k, and there is also
1
a 6 chance of tying. Then
f (a1 ) + f (a2 ) + + f (a6 )
p= .
6
Suppose James has exactly m faces with a positive number of dots on his die. Notice that
if m 3, then at least half of the faces on Jamess die lose automatically, and his probability
of winning is at most 21 . For m {4, 5, 6}, assume a1 , a2 , . . . , am are the nonzero faces of the
die and note that f (k) (k1)+p
6
for all k > 0. Then we have
m
f (a1 ) + f (a2 ) + + f (am ) X (ai 1) + p 21 m + mp
p= = .
6 i=1
36 36
21m
This simplifies to p 36m . For m {4, 5, 6}, the maximum value this expression takes is
17
32
. As this is higher than 2 , we have that p 17
1
32
in all cases.
We now construct a die obtaining this probability. Note that our bound f (k) (k1)+p 6
has equality if k {1, 2, 3, 4, 5, 6} and is a strict inequality if k > 6. Let a5 = a6 = 0 and
a1 , a2 , a3 , a4 be numbers in {1, 2, 3, 4, 5, 6} summing to 21, such as 6, 6, 6, 3. Then p = 17+4p
36
,
17
which simplifies to p = 32 . Therefore the upper bound we found is achievable, and this is
the maximum probability.
Remark: Using the result of the last paragraph, we can verify that a die obtains the
maximum probability 17
32
only when (a1 , a2 , a3 , a4 , a5 , a6 ) is a permutation of one of
(6, 6, 6, 3, 0, 0), (6, 6, 5, 4, 0, 0), (6, 5, 5, 5, 0, 0).

Round 3 Solutions
www.usamts.org
3/3/24. In quadrilateral ABCD, DAB = ABC = 110 , BCD = 35 , CDA = 105 ,

and AC bisects DAB. Find ABD.
Solution
Take E on BC so that DE is parallel to AB. Since BAD = ABE and AB is

parallel to DE, quadrilateral ABED is an isosceles trapezoid, which means it is cyclic, so
ABD = AED.
Also, DEC = ABE = 110 , and CDE = 180 DEC ECD = 180 110
for free, ifyou wish to remove this line, click here to buy Virtual PDF Printer
35 = 35 , so triangle CDE is isosceles with CE = DE.
B
110 E
110

A 110 35 C
35
Since CAD = 12 BAD = 12 CED, A lies on the circle centered at E with radius
CE = DE, so AED = 2ACD. From triangle ACD, ACD = 180 CAD ADC =
180 55 105 = 20 , so AED = 2 20 = 40 , which means ABD = 40 .
Round 3 Solutions
www.usamts.org
4/3/24. Denote by bxc the greatest integer less than or equal to x. Let m 2 be an integer,
and let s be a real number between 0 and 1. Define an infinite sequence of real numbers
a1 , a2 , a3 , . . . by setting a1 = s and ak = mak1 (m 1)bak1 c for all k 2. For example,
if m = 3 and s = 85 , then we get a1 = 85 , a2 = 15 8
, a3 = 29
8
, a4 = 39
8
, and so on.
Call the sequence a1 , a2 , a3 , . . . orderly if we can find rational numbers b, c such that
5
ban c = bbn + cc for all n 1. With the example above where m = 3 and s = 8 , we get an
3n 3
orderly sequence since ban c = 2 2 for all n. Show that if s is an irrational number and
m 2 is any integer, then the sequence a1 , a2 , a3 , . . . is not an orderly sequence.
Solution
Let {z} denote the fractional part of a real number z, and note that {z} = z bzc. We
rewrite the given recurrence as
ak = bak1 c + m{ak1 }.
Let the decimal expansion of {s} in base m be 0.d1 d2 d3 . . ., written as (0.d1 d2 d3 . . .)m , where
0 di m 1 for all i and the sequence is not eventually constant at m 1. First we show
by induction that
{ak } = (0.dk dk+1 dk+2 . . .)m .
The base case k = 1 is true by how weve defined the dk . Suppose it is true for k. No-
tice that if we take the fractional part of both sides of the recurrence, we get {ak+1 } =
{mak }. Furthermore, m (0.dk dk+1 dk+2 . . .)m = (dk .dk+1 dk+2 . . .)m , which has fractional part
(0.dk+1 dk+2 dk+3 . . .)m as desired.
From the above claim, we obtain that bm{ak }c = dk . Taking the floor of both sides of
the recurrence, we have
bak c = bak1 c + bm{ak1 }c.
The last term is equal to dk1 . Therefore,
bak c bak1 c = dk1 . (1)
Suppose the sequence a1 , a2 , a3 , . . . is orderly. We will show this implies that s is rational,
which will finish the problem. We know there exists rational numbers b and c such that
ban c = bbn + cc. Let q be a positive integer such that qb is an integer, which exists since b
is rational. Then we have that bb(n + q) + cc = bbn + cc + bq. Therefore,
ban+q+1 c ban+q c = ban+1 c ban c
for all n, and hence by (1) we get dn+q = dn for all n. This means the sequence d1 , d2 , d3 , . . .
is periodic, which means that s is rational. This completes the proof.
Round 3 Solutions
www.usamts.org
5/3/24. Let P and Q be two polynomials with real coefficients such that P has degree greater
than 1 and
P (Q(x)) = P (P (x)) + P (x).
Show that P (x) = P (x) + x.
Solution
Let n, m be the respective degrees of P, Q, and note that n > 1. From the given equation
we have that the degree of P (Q(x)) is mn and the degree of P (P (x)) + P (x) is n2 . Therefore
m = n. Let
P (x) = pn xn + pn1 xn1 + + p0 , Q(x) = qn xn + qn1 xn1 + + q0 .

2
Here pn , qn are nonzero. Consider the coefficient of xn for each side of the given equation.
Since n > 1, the P (x) term of the given equation will not affect this coefficient, so we have
pn pnn = pn qnn . Since these are real numbers, pn = qn . We break into two cases:
Case 1: pn = qn . We show by strong induction on k that pnk = qnk for 0 k n 1.
The base case k = 0 was assumed for this case. Given it is true for all values less than a
2
certain k, consider the coefficient of xn k in the given equation. Note that since k n 1
and n > 1, we have n2 k > n, so the P (x) term will not affect this coefficient. We compute
that
P (Q(x)) = pn (Q(x))n + pn1 (Q(x))n1 + + p0 ,

= pn (qn xn + qn1 xn1 + + q0 )n + pn1 (qn xn + qn1 xn1 + + q0 )n1 + + p0 .
2
Because n2 k > n(n 1), the only term that will contribute to the xn k coefficient is the
term pn (Q(x))n . We can expand this using the multinomial theorem, obtaining
X
pn (Q(x))n = pn qi1 qi2 qin xi1 +i2 ++in .
0i1 ,i2 ,...,in n
2 k
Let Ak be the coefficient of xn in this expansion. Then we have
X
Ak = p n qi1 q i2 qin .
i1 +i2 ++in =n2 k
Notice that the only time any terms on the right side are not from the set qn , qn1 , . . . , qnk+1
are when n 1 of the ij are equal to n and the last one is equal to n k. In all other cases,
all terms in the sum of the ij must be at least n k + 1.
2
Let Bk be the xn k coefficient of P (P (x)). By applying the same method of reasoning
to P (P (x)) that we used for P (Q(x)), we find that
X
Bk = pn p i1 pi2 pin .
i1 +i2 ++in =n2 k
Round 3 Solutions
www.usamts.org
But Ak = Bk . Furthermore, since pnj = qnj for all 0 j < k, when we set Ak and Bk
equal, all terms of the sum will cancel except those terms where n 1 of the ij equal n and
the remaining ij is equal to n k. Accounting for only these terms of the sum, we get
npn qnn1 qnk = npnn pnk ,
Cancelling the npn qnn1 from both sides using pn = qn , we are left with qnk = pnk , com-
pleting the inductive step.
Since the only coefficient that is not necessarily equal in P (x) and Q(x) is the constant
term, we have Q(x) P (x) = a for some real number a. We can verify P (x) = Q(x) cannot
satisfy the given equation for nonzero polynomials, so a 6= 0. The given equation becomes
P (P (x) + a) = P (x) + P (P (x)). Then letting y = P (x), we have P (y + a) P (y) = y.
Since P is nonconstant, y can take values in an infinite set, so since this is an equality of
polynomials, P (y + a) P (y) = y identically. The left side is a finite difference of P , so it has
degree n1 exactly. Since the right side has degree 1, n = 2, and thus P (x) = p2 x2 +p1 x+p0 .
Taking the linear coefficient of both sides of P (x + a) P (x) = x, we get 2p2 a = 1. Setting
the constant term of P (x + a) P (x) equal to 0 and using 2p2 a = 1, we get that p1 = 12 .
Notice that P (x) P (x) = 2p1 x = x, so the required identity is satisfied in this case.
Case 2: pn = qn . Observe that Q(x) P (x) divides (Q(x))k (P (x))k for any k.
Therefore by taking an appropriate linear combination of terms of the form (Q(x))k (P (x))k ,
we find that Q(x) P (x) divides P (Q(x)) P (P (x)). By substituting the given identity in,
we get that Q(x) P (x) divides P (x), so Q(x) P (x) also divides Q(x) by the Euclidean
Algorithm. However, notice that Q(x) P (x) has degree n since the leading coefficient of
the two polynomials is different. This means the same degree n polynomial divides P (x)
and Q(x), both also degree n polynomials. As a result, P (x) and Q(x) are scalar multiples
of each other. Since pn = qn , we conclude that P (x) = Q(x).
Plugging this into the equation in the original problem gets P (P (x)) = P (P (x))+P (x).
Let y = P (x) to obtain that P (y) = P (y) + y. This identity is true for an infinite number
of y since P (x) takes values on an infinite set, so therefore it is true for all y, finishing this
case.
Remark: Both of these cases admit solutions for P and Q. In case 1, we can have
P (x) = x2 x2 and Q(x) = x2 x2 + 12 . In case 2, we can have P (x) = x4 x2 and
Q(x) = x4 + x2 . Both of these pairs will satisfy P (Q(x)) = P (P (x)) + P (x).

Round 1 Solutions
www.usamts.org
1/1/25. Alex is trying to open a lock whose code is a sequence that is three letters long, with
each of the letters being one of A, B or C, possibly repeated. The lock has three buttons,
labeled A, B and C. When the most recent 3 button-presses form the string, the lock opens.
What is the minimum number of total button presses Alex needs to be sure to open the
lock?
Solution
The answer is 29. This can be achieved with the following sequence of presses:
AAACCCBCCACBBCBACABCAABBBABAA
There are 3 3 3 = 27 different strings of three letters with each letter being one of A, B,
or C. All 27 of these strings appear consecutively in the above sequences of presses.
Each press of the button corresponds to attempting at most one more string, namely the one
formed by the previous three presses. No string can be attempted after the first two presses.
Therefore, the first time that all 27 strings can be tried on the lock is after 27 + 2 = 29
presses of the button, so 29 is indeed the minimum.
Round 1 Solutions
www.usamts.org
2/1/25. In the 5 6 grid shown, fill in all of the grid cells with
the digits 09 so that the following conditions are satisfied:
7
8 6
1. Each digit gets used exactly 3 times.
2. No digit is greater than the digit directly above it.
2 4
3. In any four cells that form a 2 2 subgrid, 5 1
the sum of the four digits must be a multiple of 3. 3
You do not need to prove that your configuration is the only
one possible; you merely need to find a configuration that satisfies the constraints above.
(Note: In any other USAMTS problem, you need to provide a full proof. Only in this
problem is an answer without justification acceptable.)
Solution
Label the unknown cells with the variables a through v as shown to the a b c d 7 e
right. Using rule 3 repeatedly in the top two rows, we have f 8 g h i 6
j k 2 4 l m
b+8+c+g c+g+d+h d+h+7+i 7+i+e+6 0 (mod 3). 5 n o p 1 q
r 3 s t u v
The first equivalence above gives us b + 8 d + h (mod 3), and the third
equivalence gives us d + h e + 6 (mod 3). Thus b + 8 e + 6 (mod 3). Since 8 b 9
and 6 e 9, we must have that (b, e) is either (8, 7) or (9, 8).
Applying the technique above repeatedly gives us the following fourth rule:
4. If two columns are an even distance apart, then any adjacent pair of cells in the first
column has the same sum modulo 3 as the corresponding adjacent pair in the second
column. Similarly, if two rows are an even distance apart, then any adjacent pair of
cells in the first row has the same sum modulo 3 as the corresponding adjacent pair in
the second row.
Using rule 4, we have 5 + r 1 + u (mod 3). In particular, since u must be 0 or 1, we must

have r be either 0 or 2 modulo 3.
Now consider the sum of all numbers of the board. We have three copies a b c d 7 e
of every digit, so this sum is a multiple of 3. We also know that the six f 8 g h i 6
2 2 squares shown in the diagram to the right all sum to a multiple of j k 2 4 l m
3. Subtracting these six squares out from the sum of the whole board, 5 n o p 1 q
we get that r 3 s t u v
7 + e + 2 + 4 + r + 3 = 16 + e + r 1 + e + r 0 (mod 3).
But e must be 7 or 8 from above, and r must be 0 or 2 modulo 3, so the only possibility is
that e = 8 and r 0 (mod 3). Hence (b, e) = (9, 8) and also u = 1.
Round 1 Solutions
www.usamts.org
Now the grid looks as shown. Next, note that i 0 (mod 3) using the a 9 c d 7 8
upper-right 2 2 corner of the grid. So by rule 4 applied to the top two f 8 g h i 6
rows, we have j k 2 4 l m
5 n o p 1 q
a+f c+g 7+i1 (mod 3), r
0 or 3 3 s t 1 v
9+8d+h8+62 (mod 3).
In particular, no column can contain two 9s, so all three 9s must be in the top row.
Additionally, by rule 4, c + d 2 + 4 0 (mod 0), so since one of c and d must be a 9, they
bothfor free,
Create PDF with GO2PDF must be tomultiples
if you wish of here
remove this line, click 3. to Also
buy VirtualaPDF
isPrinter
a multiple of 3 since by rule 4, a + 9 r + 3 0
(mod 3). So a, c, d, i, and r are all multiples of 3.
At this point, we have shaded the boxes that we know must be mul- a 9 c d 7 8
tiples of 3. We know enough of the cells modulo 3 for simple appli- f 8 g h i 6
cations of rules 3 and 4 to give us the value of all of the cells mod- j k 2 4 l m
ulo 3. For example, d + 7 + h + i 0 + 1 + h + 0 0 (mod 3), so 5 n o p 1 q
h 2 (mod 3). Repeating this around the grid gives us the following r 3 s t 1 v
chart:
0
mod 3 9 0 0
mod 3 mod 3 7 8
1
mod 3 8 1 2 0
mod 3 mod 3 mod 36
2 1
mod 3 mod 3 2 4 0 0
mod 3 mod 3
5 1 2 1
mod 3 mod 3 mod 3 1 2
mod 3
0
mod 3 3 0 0
mod 3 mod 3 1 2
mod 3
We now just need to apply rules 1 and 2 repeatedly to determine the exact values of each
cell. This is mostly a case of filling in from the top and/or the bottom using rule 2, and
keeping track of how many times each digit is used to apply rule 1. To start, the only place
for the three 0s is the bottom row, and the only place for the remaining 1 is the 4th column.
The grid now looks as at right. The only numbers that fit in the left a 9 c d 7 8
column with the correct residues modulo 3 are (a, f, j) = (9, 7, 5), and f 8 g h i 6
the only remaining place for the last 8 is h. This makes d = 9 and c = 6. j k 2 4 l m
Then o = 2 and (q, v) = (5, 2) are the only ways to place the remaining 5 n o 1 1 q
numbers that are 2 modulo 3, and the rest of the grid easily fills to give 0 3 0 0 1 v
us the final answer below.
9 9 6 9 7 8
7 8 4 8 3 6
5 7 2 4 3 6
5 4 2 1 1 5
0 3 0 0 1 2
Round 1 Solutions
www.usamts.org
3/1/25. An infinite sequence of positive real numbers a1 , a2 , a3 , . . . is called territorial if for

all distinct positive integers i, j with i < j, we have |ai aj | 1j . Can we find a territorial
sequence a1 , a2 , a3 , . . . for which there exists a real number c with ai < c for all i?
Solution
We will construct a territorial sequence exists whose maximum value is 2. This allows us to
choose c to be any number greater than 2.
Consider the sequence
1 3 1 3 5 7 1 3 5 7 9 11 13 15 1 3
(ai ) = 2, 1, , , , , , , , , , , , , , , , , . . . .
2 2 4 4 4 4 8 8 8 8 8 8 8 8 16 16
Specifically, a1 = 2, and for any positive integer n 2, let k be the unique nonnegative
k )1
integer such that 2k < n 2k+1 , and then an = 2(n2 2k
. Note that all the numbers in
the sequence are distinct.
Suppose we are given two positive integers i, j with i < j. Let k be the unique nonnegative
integer such that 2k < j 2k+1 . Then aj = 2tk for some odd integer t. Since i 2k+1 , we also
have ai = 2uk for some integer u with t 6= u and u not necessarily odd. Then |ai aj | 21k > 1j ,
as desired.
Round 1 Solutions
www.usamts.org
4/1/25. Bunbury the bunny is hopping on the positive integers. First, he is told
a positive integer n. Then Bunbury chooses positive integers a, d and hops on
all of the spaces a, a + d, a + 2d, . . . , a + 2013d. However, Bunbury must make
these choices so that the number of every space that he lands on is less than n and relatively
prime to n.
A positive integer n is called bunny-unfriendly if, when given that n, Bunbury is unable to
find positive integers a, d that allow him to perform the hops he wants. Find the maximum
bunny-unfriendly integer, or prove that no such maximum exists.
Solution
Let M be the product of all prime numbers less than 2014. We claim that the maximum
bunny-unfriendly integer is 2013M . Note that 2013 = 3 11 63 is not prime, so all of the
prime divisors of 2013M are less than 2013.
First, we verify that 2013M is bunny-unfriendly. Suppose, for sake of contradiction, that pos-
itive integers a and d could be chosen so that a+2013d < 2013M and gcd(a+kd, 2013M ) = 1
for all 0 k 2013. If there is a prime p less than 2014 that does not divide d, then d1
exists modulo p. Choose 0 k < p such that k ad1 (mod p). Then a + kd is a multiple
of p, and gcd(a + kd, 2013M ) p, a contradiction. Therefore no such p exists, and thus
d must be a multiple of M . But then a + 2013d > 2013d > 2013M , also a contradiction.
Thus, Bunbury will not be able to find an a and d to use for his hopping, and hence 2013M
is bunny-unfriendly.
Next, we verify that all numbers greater than 2013M are bunny-friendly. Let n be an integer
greater than 2013M , and let p be the largest prime divisor of n. We break into cases based
on the properties of p.
Case 1: p < 2014, so that all the prime divisors of n are less than 2014.
Let x be the product of ns distinct prime divisors. Let a = 1 and d = x, and note that we
have that gcd(n, 1 + kx) = 1 for all integers k. Furthermore, 1 + 2013x < 2014x, which is at
most n because nx > M n
> 2013. Thus, this choice of a and d proves that n is bunny-friendly.
Case 2: p > 2014.
We first consider a = 1 and d = np . If d is still a multiple of p, then gcd(n, 1 + kd) = 1 for all
integers k. Furthermore, 1 + 2013d = 1 + 2013n p
< n. So this choice of a and d proves that n
is bunny-friendly.
But if d is not a multiple of p, then there is a unique b with 0 b < p such that 1 + bd
is a multiple of p, and hence gcd(n, 1 + bd) = p. For all other 0 k < p with k 6= b, we
have gcd(n, 1 + kd) = 1. Thus, if b > 2013, choosing a = 1 and d = np still proves that n is
bunny-friendly, by the argument above.
If b < 2014, then we can try to start our hopping after the bad multiple of d. That is, we
let a = 1 + (b + 1)d. This works provided that
a + 2013d = 1 + (b + 2014)d < n.

Round 1 Solutions
www.usamts.org
So it is sufficient to have 1 + 4027d < n. If p > 4028, then this will always be the case,
because then 1 + 4027d < 1 + 4027
4028
n < n.
So our remaining case is when d is not a multiple of p and p 4027. In this case we have to
try something different. Let a = p+d. Note that gcd(n, p+(k+1)d) = 1 for all 0 k 2013.
Furthermore,
2014n p 2014
p + 2014d = p + = + n,
p n p
but np is very small (recall that n > 2013M is very large relative to p < 4028); in particular
p
2
, so
n >forpfree, < p1 , and
if you wish
n to remove
hence
this line, click here to buy Virtual PDF Printer

1 2014 2015
p + 2014d < + n= n < n.
p p p
Thus this choice of a and d proves that n is bunny-friendly.

In all cases, any n > 2013M is bunny-friendly. Thus, 2013M is the largest bunny-unfriendly
integer.
Round 1 Solutions
www.usamts.org
5/1/25. Niki and Kyle play a triangle game. Niki first draws 4ABC with area 1, and Kyle
picks a point X inside 4ABC. Niki then draws segments DG, EH, and F I, all through
X, such that D and E are on BC, F and G are on AC, and H and I are on AB. The ten
points must all be distinct. Finally, let S be the sum of the areas of triangles DEX, F GX,
and HIX. Kyle earns S points, and Niki earns 1 S points. If both players play optimally
to maximize the amount of points they get, who will win and by how much?
Solution
(We use the common notation that [P QR] denotes the area of triangle P QR.)
We claim that the game will end with S = 31 . In particular, for any X that Kyle chooses, we
will show that Niki can choose her points so that S 13 , and if X is the centroid of ABC,
then the best that Niki can do is S = 13 .
Note that none of the argument below depends on how triangle ABC is drawn, so how Niki
draws ABC is irrelevant.
A
In the diagram at right, S is the sum of the shaded
areas. Consider the sum of the areas [AF I] + [BEH] + H
[CDG]. Notice that this sum counts each white region

G
once but each shaded region twice: for example, [HIX]
F X
is counted in both [AF I] and [BEH]. Therefore, I
[AF I] + [BEH] + [CDG] = 1 + S,

C E D B
and hence
S = [AF I] + [BEH] + [CDG] 1.
We will show that for any X Kyle chooses, Niki can choose her points such that
4
[AF I] + [BEH] + [CDG] ,
3
which implies that S 13 .
Let X be Kyles chosen point. Let a, b, and c be the lengths of BC, AC, and AB, respectively,
and let hA , hB , and hC be the distances from X to BC, AC, and AB, respectively.
We will focus on Nikis choices of F and I to minimize A
[AF I]. Let r = AI and s = AF . On the one hand,

s
r
1 hc
[AF I] = [AF X] + [AIX] = (rhB + shC ). F
hb
2
X I
On the other hand, since [ACB] = 1, we have
[AF I] rs sin F AI rs C B
[AF I] = = = .
[ACB] bc sin CAB bc
Round 1 Solutions
www.usamts.org
Combining the above two equations, we have

1 rs
[AF I] = (rhB + shC ) = . ()
2 bc
Set x = [AF I]. By the AM-GM inequality,
1 p
x = (rhB + shC ) rhb shC .
2
But rs = 1 bc(rh + sh ) = bcx by (), therefore
Create PDF with GO2PDF for free, if you B this line,Cclick here to buy Virtual PDF Printer
2 wish to remove
p
x bchB hC x.
Squaring and solving for x gives

x bchB hC ,
with equality if and only if rhB = shC . In the equality case, () gives rhB = rs
bc
, so s = bchB ,
and similarly r = bchC . Thus, Niki minimizes [AF I] by choosing F and I so that r = bchC
and s = bchB . She will able to make these choices provided that bhB 1 and chC 1.
So, if Kyle picks X such that ahA , bhB , and chC are all less than 1, then Niki can choose
her points so that
[AF I] + [BEH] + [CDG] = bchB hC + cahC hA + abhA hB .
Lets dispose of the contrary case first: suppose Kyle GA H

picks X such that (without loss of generality) ahA 1, F I
or equivalently ha a1 . Niki can choose F and I so X
that F I k CB. Then by similarity, and using the fact

that the height from A to BC is a2 , we have
2 2 1 2
a
ha a 1
[AF I] = 2 2 = .
a a
4 C E D B
Furthermore, Niki can pick G and H very close to A so that [BEH] + [CDG] is as close to
1 as she wishes (in the extreme case, G = H = A and [BEH] + [CDG] = 1), in particular,
Niki can get [BEH] + [CDG] < 13 12
. Thus, Niki can achieve
1 13 4
[AF I] + [BEH] + [CDG] < + = .
4 12 3
Now for the remaining cases: let p = ahA , q = bhB , and r = chC , and assume that Kyle
picks X such that p, q, r are all less than 1. Then we have
[AF I] + [BEH] + [CDG] = qr + rp + pq.

Round 1 Solutions
www.usamts.org
But we also have the general inequality

1
(qr + rp + pq) (p + q + r)2 ,
3
with equality if and only if p = q = r. (This is a corollary of the rearrangement inequality
pq+qr+rp p2 +q 2 +r2 , which itself follows from the inequality (pq)2 +(qr)2 +(rp)2
0.) On the other hand,
p + q + r = 2[BCX] + 2[CAX] + 2[ABX] = 2[ABC] = 2,

so Nikis choices result in

1 4
[AF I] + [BEH] + [CDG] = qr + rp + pq (p + q + r)2 = ,
3 3
and this inequality is strict unless p = q = r. Since Kyle wants to choose X to maximize the
amount of area that Niki must choose, he wants to force the inequality to be an inequality
by choosing X such that p = q = r. This means that ahA = bhB = chC , which means that
X is the centroid of ABC.
In all cases, Niki can pick points such that
1
S = [AF I] + [BEH] + [CDG] 1 ,
3
and Kyle can force her into S = 13 by picking X to be the centroid of ABC. Thus, with
optimal play, Niki will score 32 and Kyle will score 31 .

Round 2 Solutions
www.usamts.org
1/2/25. In the 3 5 grid shown, fill in each empty 9

box with a two-digit positive integer so that:
(a) no number appears in more than one box, and
(b) for each of the 9 lines in the grid consisting of
three boxes connected by line segments, the box in
the middle of the line contains the least common
multiple of the numbers in the other two boxes on 4
the line.
acceptable.)
Solution
Let x, y, z be three numbers on a line in that order, so that y = lcm(x, z). Notice that if
x is a divisor of z, then y = lcm(x, z) = z. But we must have y 6= z since all the numbers
must be distinct, so we conclude that x cannot be a divisor of z. By the same reasoning, z
cannot be a divisor of x. That is, neither number at the end of a line can be a divisor of the
number at the other end. Call this the anti-divisor property.
Label the unknown numbers in the grid with the a b c d 9
variables a through m as shown at right. By the
anti-divisor property, m is not a multiple of 4 or 9.
But if ms only prime divisors are 2 and/or 3, then e f g h i
m 6, which contradicts the requirement that m
is a 2-digit number. Therefore, m must have a
prime divisor p with p 5. Then, h, i, and l are
all multiples of m, so they each are multiples of p.
j k 4 l m
g is also a multiple of h, so it is a multiple of p. Since lcm(4, c) = g and p divides 4 but not

g, p must also divide c. Then, d is a multiple of c, so d is a multiple of p. In summary, all
of c, d, g, h, i, l, m are multiples of p.
In particular, d and i are each multiples of 9p, so we must have p = 5, because if p 7 then
there is at most 1 two-digit multiple of 9p. Hence d and i are 45 or 90 in either order. This
means that c and m are each multiples of 5 but divisors of 90, so they must be 10, 15, or
30. However, if one of them were 30, then the line chm would violate the anti-divisor
property, so c and m must be 10 and 15 in either order. But if c = 10 and m = 15, this
makes l = 60 and g = 20, a contradiction since g l. Thus, c = 15 and m = 10.
Round 2 Solutions
www.usamts.org
We can now fill in all of the rightmost three columns a b 15 45 9

of the grid, as shown to the right.
Next, we see that b must be a two-digit multiple of
3 but a divisor of 60. The only possibility not yet e f 60 30 90
used in the grid is b = 12. Thus f is a multiple of
12, but it cannot be a multiple of 8 because then
a would be a multiple of 8, contradicting the anti- j k 4 20 10
divisor property on the line af 4. This leaves
the only possibilities for f , after also excluding the multiples of 12 already in the grid, as
36 or 84. But a and k must both be two-digit divisors of f , and the only two-digit proper
divisors of 36 are 12 and 18, and 12 is already in the grid (at b). Thus we must have f = 84.
The grid is now as shown at right. Next, we see a 12 15 45 9
that a must be a multiple of 21 and a proper divisor
of 84, so it must be 21 or 42. But a = 42 leaves no
value for e, so we must have a = 21. Then e must e 84 60 30 90
be 42 or 63, but e = 63 leaves no two-digit value
for j, so we must have e = 42, which forces j = 14.
Finally, k is a two-digit proper divisor of 84 not yet
used in the grid, and the only value remaining is
j k 4 20 10
k = 28.
The completed grid is shown below, and our argument above proves that this solution is
unique.
21 12 15 45 9
42 84 60 30 90
14 28 4 20 10
Round 2 Solutions
www.usamts.org
2/2/25. Let ABCD be a quadrilateral with AB k CD, AB = 16, CD = 12, and BC < AD.
A circle with diameter 12 is inside of ABCD and tangent to all four sides. Find BC.
Solution
0
C y F y D
6
y
6 O
G PDF Printer
6
x
B x E x0 A
Let O be the center of the circle, let E and F be the points of tangency to AB and CD,
respectively, and let G be the point of tangency to BC, as in the picture above.
Let x = EB, and note that since 4OEB = 4OGB, we also have BG = x. Let y = F C,

and note that since 4OF C = 4OGC, we also have CG = y. Further, notice that
1 1 1
EOB = EOG = (180 F OG) = 90 F OG = 90 F OC.
2 2 2
Thus, EOB and F OC are complementary angles, and hence 4EOB and 4F CO are
similar right triangles. Therefore, EB
EO
= FF O
C
, giving x6 = y6 , or xy = 36.
Let x0 = EA and y 0 = F D. Using the same reasoning as above in trapezoid AEF D, we have
x0 y 0 = 36.
We also have the equations
x + x0 = 16, (1)
y + y 0 = 12. (2)
36 36
Using y = x
and y 0 = x0
, (2) becomes
1 1 1
+ 0 = . (3)
x x 3
x + x0 1
But (3) is equivalent 0
= , which simplifies to 3(x + x0 ) = xx0 , so using (1) we have
xx 3
xx0 = 48. (4)
Now, equations (1) and (4) together tell us that x and x0 are the roots of the quadratic
equation t2 16t + 48 = 0, which factors as (t 12)(t 4) = 0, so x and x0 are 4 and 12 (in
some order). Using xy = x0 y 0 = 36, we find that the two possible solutions are
(x, y, x0 , y 0 ) = (4, 9, 12, 3) or (x, y, x0 , y 0 ) = (12, 3, 4, 9).
However, recall that BC = x + y and AD = x0 + y 0 , and we must have BC < AD. This
forces the first solution for (x, y, x0 , y 0 ), giving a final answer of BC = 13.
Round 2 Solutions
www.usamts.org
3/2/25. For each positive integer n 2, find a polynomial Pn (x) with rational coefficients
1
such that Pn n 2 = n
. (Note that n 2 denotes the positive nth root of 2.)
1+ 2
Solution
We use the factorization

1 rk
= 1 + r + r2 + + rk1 .
1r
n
Let k = 2n and r = 2. Then the above equation becomes
14
= ( 2)2n1 + ( 2)2n2 + ( 2)2n3 + ( 2)2n4 + + ( 2) + 1.
n n n n n

n
1+ 2
Therefore,

1 ( n 2)2n1 ( n 2)2n2 + ( n 2)2n3 ( n 2)2n4 + + n 2 1
= .
1+ n2 3
Hence, we may choose

2n1
x2n1 x2n2 + x2n3 x2n4 + + x 1 X (1)i+1
Pn (x) = = xi .
3 i=0
3
Round 2 Solutions
www.usamts.org
4/2/25. An infinite sequence of real numbers a1 , a2 , a3 , . . . is called spooky if a1 = 1 and for

all integers n > 1,
na1 + (n 1)a2 + (n 2)a3 + + 2an1 + an < 0,
n2 a1 + (n 1)2 a2 + (n 2)2 a3 + + 22 an1 + an > 0.
Given any spooky sequence a1 , a2 , a3 , . . ., prove that
20133 a1 + 20123 a2 + 20113 a3 + + 23 a2012 + a2013 < 12345.
Solution
Define the following quantities for each positive integer k:

tk = ka1 + (k 1)a2 + + 2ak1 + ak ,
uk = k 2 a1 + (k 1)2 a2 + + 22 ak1 + ak ,
vk = k 3 a1 + (k 1)3 a2 + + 23 ak1 + ak .
Note that from the given conditions, t1 = u1 = 1, and tk < 0 < uk for all k > 1.
We compute:
tk + tk+1 + + t1 = (k + (k 1) + + 1)a1
+ ((k 1) + (k 2) + + 1)a2
+ ((k 2) + (k 3) + + 1)a3
.
+ ..
+ ak .
n2 +n
By using the fact that 1 + 2 + + n = 2
, we have:
k2 + k (k 1)2 + (k 1) (k 2)2 + (k 2)
tk + tk1 + + t1 = a1 + a2 + a3 + + ak
2 2 2
1
= (uk + tk ).
2
Thus,
uk = 2(t1 + t2 + t3 + + tk1 ) + tk .
In particular, note that uk < 2t1 = 2 for all k > 1.
Similarly, we compute:
uk + uk+1 + + u1 = (k 2 + (k 1)2 + + 12 )a1
+ ((k 1)2 + (k 2)2 + + 12 )a2
+ ((k 2)2 + (k 3)2 + + 12 )a3
.
+ ..
+ ak .
Round 2 Solutions
www.usamts.org
2n3 +3n2 +n
By using the fact that 12 + 22 + + n2 = 6
, we have:
2k 3 + 3k 2 + k 2(k 1)3 + 3(k 1)2 + (k 1)

uk + uk1 + + u1 = a1 + a2 + + ak
6 6
1
= (2vk + 3uk + tk ).
6
Thus,
3uk + tk 3uk tk
v = 3(u + u
Create PDF with GO2PDF for free, if youkwish to remove + + u )
click here to buy Virtual1PDF Printer
k this line,k1 = 3(uk1 + + u1 ) + .
2 2
Since uk < 2 and tk < 0 for all k > 1, we have
3(2)
vk < 3(2(k 2) + 1) + = 6k 6.
2
In particular, v2013 < 6(2013) 6 = 12072 < 12345, as desired.
Round 2 Solutions
www.usamts.org
5/2/25. Let S be a planar region. A domino-tiling of S is a partition of S into

1 2 rectangles. (For example, a 2 3 rectangle has exactly 3 domino-tilings,
as shown to the right.) The rectangles in the partition of S are called dominoes.
(a) For any given positive integer n, find a region Sn with area at most 2n
that has exactly n domino-tilings.
(b) Find a region T with area less than 50000 that has exactly 100002013
domino-tilings.
Solution
(a) Consider the infinitely repeating pattern of squares

shown in the diagram to the right, where we have col-
ored some squares gray and the rest white. For each 11 10 10
positive integer n, define Sn to be the region contain-
ing the squares in this figure numbered 1 to n inclusive.
11 9 8 8
We claim that Sn has exactly n domino-tilings, which 9 7 6 6
we will prove below. Since Sn has area 2n, these are
our desired regions. 7 5 4 4
First, for n 2, define Bn to be the same region as Sn 5 3 2 2
except omitting the two white squares labeled n and
n 1. We prove that Bn has a unique domino-tiling,
3 1 1
by induction on n. The base case n = 2 is clear: B2 consists only of the gray squares
labeled 1 and 2. For the inductive step, let k 2 be given and assume that Bk has a unique
domino-tiling, and consider Bk+1 . The gray square labeled k + 1 touches only the white
square labeled k 1, so in any domino-tiling of Bk+1 those two squares must be part of the
same piece of the partition (i.e. must be covered by the same domino). But what remains
is Bk , which by inductive hypothesis has a unique domino-tiling. Thus, Bk+1 has a unique
domino-tiling.
Now we prove our claim that Sn has exactly n domino-tilings, by induction on n. The base
cases of n = 1 and n = 2 are easily checked: S1 has a unique domino-tiling since it has
only two squares, and S2 is a 2 2 square, so the only possible domino-tilings are either
two horizontal or two vertical dominos, yielding exactly 2 domino-tilings. Now let k 2 be
a given positive integer, and assume that Sk has exactly k domino-tilings. Consider Sk+1 ,
and in particular the domino including the white square labeled k + 1. If this domino also
includes the white square labeled k, then what remains is Bk+1 , and thus there is only one
way to complete the domino-tiling of Sk+1 . Otherwise, our initial domino covers both squares
labeled k + 1. Then, the region that remains is Sk , so by the inductive hypothesis, there are
k ways to complete the domino-tiling of Sk+1 . Combining the two cases, we have k + 1 total
ways of tiling Sk+1 , completing the inductive step.
Round 2 Solutions
www.usamts.org
(b) Take two copies of S10000 and one copy of S2013 . Z4

Z3
Use them to construct the shape T shown in the di- Z1 Z2
Y4
agram at right. (Some squares have been labeled Y1 Y2 Y3
Xi , Yi , or Zi , for use in our proof below.) X4
S10000
X3
The area of T is X1 X 2
S2013
30 + 2(2 10000) + 2 2013 = 44056,

S10000
which is less than 50000. We now show that T has

exactly 100002013 domino-tilings, by showing that all
domino-tilings fall into one of the following two cases.

Case 1: The same domino covers X1 and X2 . Z4
Z3
There are now an odd number of remaining squares in Z1 Z2
Y4
the bottom left copy of S10000 , so X3 and X4 are also Y1 Y2 Y3
covered by the same domino. This forces the dominos X4
S10000
X3
as shown to the right. Both copies of S10000 have X1 X 2
had their two top left squares (X2 , X3 and Z2 , Z3 ) S2013
already used, meaning that what remains (for each)

S10000
is B10000 and each has only one way to be domino-
tiled. In contrast, the copy of S2013 is intact, so it has
2013 ways to be domino-tiled. Thus there are 2013
domino-tilings in this case.
Case 2: Different dominoes cover X1 and X2 . Z4

Z3
Then X3 and X4 cannot be covered by the same Z1 Z2
Y4
domino, since this would leave an odd number of Y1 Y2 Y3
squares in S10000 . This forces the dominos as shown X4
S10000
X3
to the right. Both copies of S10000 are intact, meaning X1 X 2
they each have 10000 ways of being domino-tiled, and S2013
the tilings of each copy of S10000 are independent of

S10000
each other. On the other hand, the copy of S2013 has
both squares Y2 , Y3 already used, so what remains is
B2013 , and this has only one domino-tiling. So there
are 10000 1 10000 = 100002 domino-tilings in this
case.
Adding up the two cases, we get 2013 + 100002 = 100002013 total domino-tilings of T , as
required.

Round 3 Solutions
www.usamts.org
1/3/25. In the hexagonal grid shown, fill in each space

with a number. After the grid is completely filled in, 4 1
the number in each space must be equal to the small- 1 1 3
est positive integer that does not appear in any of the
touching spaces. 2 6
7
You do not need to prove that your configuration is
the only one possible; you merely need to find a con- 3 5
figuration that satisfies the constraints above. (Note:
2 4 3
In any other USAMTS problem, you need to provide a
full proof. Only in this problem is an answer without 3 2
Solution
We label the cells a through v lexicographically as in

4 a b 1 the diagram to the left.
1 c 1 d 3 The 4 in row 1 must touch a 2, but the 2 cant be c
e 2 f g 6 h because c already touches a 2. Therefore, a = 2. The
4 must also touch a 3, so c = 3. Finally, b must be
i j k 7 ` m n at least 3, but cant be greater than 3 because there is
no way to place a 3 in a neighboring space. Therefore,
o 3 p q 5 r b = 3.
2 s 4 t 3 The 6 in the third row must touch a 5. That 5 cannot
be ` or m because those each neighbor the 5 in the fifth
3 u v 2 row. The 5 also cannot be d or h because there are not
enough empty spaces for d or h to neighbor all of 1, 2, 3, and 4. Therefore, g = 5. Now the
three vacant spaces touching g must have the numbers 2, 3, and 4. Since f touches a 2 and
a 3, it must be the 4. Since d touches a 3, it must be the 2. This leaves ` = 3. The progress
so far is summarized in the diagram to the right.
Consider the 4 in the sixth row. It must touch a 3,
but the only space surrounding it that does not touch 4 2 3 1
a 3 is v, so v = 3. The 2 in row 7 must touch a 1, so 1 3 1 2 3
t = 1. Next, consider ` = 3. It must touch a 1 and
a 2, so m and q must have a 1 and 2 between them. e 2 4 5 6 h
But q touches a 1, so q = 2 and m = 1. To finish the
right edge, consider the 5 in the fifth row. It must
i j k 7 3 m n
touch a 4, so r = 4. We can quickly determine that o 3 p q 5 r
n = 2 and h = 4 from here.
2 s 4 t 3
3 u v 2
Round 3 Solutions
www.usamts.org
Our progress so far is summarized in the diagram to

the right. Next, the 7 in the center must touch the 4 2 3 1
six numbers 1 to 6 once each. We already found the 1 3 1 2 3
numbers 2 through 5, so k and p are 1 and 6 between
them. Notice that regardless of which of the two is the e 2 4 5 6 4
6, the remaining vacant space it touches (either j or s)
must be a 5. In particular, this means each of j and s
i j k 7 3 1 2
is either 5 or neighbors a 1, so neither can be a 1. Since o 3 p 2 5 4
the 3 in row 7 must touch a 1, that 1 must be u. Then
s = for5,free,since
Create PDF with GO2PDF ittonow
if you wish remove touches
this line, click here
2 s 4 1 3
1 tothrough 4,Printer
buy Virtual PDF but p 6= 5.
The 2 in row 6 must touch a 1, so o = 1. All of the 3 u 3 2
vacant spaces around i touch a 2, so i is at most 2. It
is also at least 2 since it touches a 1, so i = 2. By similar reasoning, e is a 3 and j is a 4.
Since j is not a 5, the 6 must be p, and the 1 goes in k. This completes the grid. The only
solution is the one shown below.
4 2 3 1
1 3 1 2 3
3 2 4 5 6 4
2 4 1 7 3 1 2
1 3 6 2 5 4
2 5 4 1 3
3 1 3 2
Round 3 Solutions
www.usamts.org
2/3/25. Let a1 , a2 , a3 , . . . be a sequence of positive real numbers such that ak ak+2 = ak+1 + 1
for all positive integers k. If a1 and a2 are both positive integers, find the maximum possible
value of a2014 .
Solution
First we use the recurrence relation ak ak+2 = ak+1 + 1 to compute a2014 in terms of a1 and
a2 , then well choose values of a1 and a2 that maximize a2014 .
The recurrence relation can be rewritten as
ak+1 + 1
ak+2 = .
ak
Using this, we compute
a2 + 1
a3 =
a1
a2 +1
a3 + 1 a1
+1 a2 + a1 + 1
a4 = = = ,
a2 a2 a1 a2
a2 +a1 +1
a4 + 1 a1 a2
+1 (a1 a2 + a2 + a1 + 1)(a1 ) (a1 + 1)(a2 + 1)(a1 ) a1 + 1
a5 = = a2 +1 = = = ,
a3 a1
(a 1 a 2 )(a 2 + 1) (a 1 a 2 )(a 2 + 1) a 2
a1 +1 a1 +a2 +1
a5 + 1 a2
+1 a2
a6 = = a2 +a1 +1 = a1 +a 2 +1
= a1 ,
a4 a1 a2 a1 a2
a6 + 1 a1 + 1
a7 = = a1 +1 = a2 .
a5 a2
Since each term of the sequence depends only on the previous two terms and we have shown
a1 = a6 and a2 = a7 , we conclude that the sequence is periodic with period 5. Therefore, for
any k, we have ak = ar , where r is the remainder when k is divided by 5. So
a2 + a1 + 1 1 1 1
a2014 = a4 = = + + .
a1 a2 a1 a2 a1 a2
We are given that a1 and a2 are positive integers, so each term in this expression is at most
1 and the sum is at most 3. For a1 = a2 = 1, the sum is equal to 3. The maximum possible
value of a2014 is 3 from the sequence
1, 1, 2, 3, 2, 1, 1, 2, 3, 2, . . . .
Round 3 Solutions
www.usamts.org
3/3/25. Let A1 A2 A3 . . . A20 be a 20-sided polygon A7 A5

P in the plane. Suppose all of the side lengths
A6
of P are 1, the interior angle at Ai measures 108 A9 A3
A8 A4
degrees for all odd i, and the interior angle at
Ai measures 216 degrees for all even i. Prove
that that the lines A1 A9 , A2 A12 , A3 A15 , A4 A18 , A10 A2
and A6 A20 all intersect at the same point. A11 A1
Solution A12 A20
Let O be the center of P . Let B2i be the point A14 A18
A13 A19
on the line segment A2i O such that A2i B2i has A16
length 1. For each i, draw segment A2i B2i and
B2i B2i+2 . The resulting figure is shown below A15 A17
left, with the line segments A2 A8 , A4 A10 , A5 A13 ,
A6 A16 , and A7 A19 in gray.
A7 A5
We claim that the 10 small pentagons are regular
A6
unit pentagons. By symmetry we may check only
A8 A4
A9 A3 A2 A3 A4 B4 B2 , so consider only this pentagon for
now. First note that A2 B2 bisects a 216 angle so
B6
A10 B8 B4 A2 A2 = 108 . Since A2 = A3 = A4 = 108
B10 B2 and B4 = B2 , the pentagon is equiangular.
O
A11 A1 By construction, 4 edges have length 1, so all 5
B12 B20 edges have length 1 and the pentagon is regular.
A12 B14 B18 A 20 In particular, since the pentagon is regular, the
B16 line segments B2i B2i2 have length 1. By sym-
A13 A19 metry, the inner decagon is also equiangular so
A14 A18 it must be regular. Specifically, each angle of the
A16 decagon has measure 144 .
A15 A17
We now claim that A2 A8 , A4 A10 , A5 A13 , A6 A16 , and A7 A19 all intersect at B6 .
Since B6 lies on A6 O by definition, A6 A16 necessarily passes through B6 . If A2 A8 passes
through B6 , then A4 A10 passes through B6 , because A4 A10 is the reflection of A2 A8 through
A6 A16 . Similarly, if A5 A13 passes through B6 , then so does A7 A19 . So to solve the problem,
it suffices to show that A2 A8 and A5 A13 both contain B6 .
Lines B4 B6 and A2 A8 are both perpendicular to A5 A15 so are parallel (and hopefully equal).
The point B6 lies on A2 A8 if and only if these lines are equal. This occurs if and only if
B4 B6 A8 = 180 . To prove this, we first write B4 B6 A8 = B4 B6 B8 + B8 B6 A8 . Since
B4 B6 B8 is the interior angle of a regular decagon, its measure is 144 . Further, 4B6 B8 A8
is isosceles with vertex angle 108 , so the base angle is B8 B6 A8 = 36 . This forces
B4 B6 A8 = 144 + 36 = 180 ,
Round 3 Solutions
www.usamts.org
and A2 A8 passes through B6 .

Similarly, B6 B12 is parallel to A5 A13 , so A5 A13 A7 A5
A6
contains B6 if and only if B6 B12 A13 = 180 . To
A8 A4
prove this, first write A9 A3
B6 B12 A13 = B6 B12 B10 + B10 B12 A12 B6
A10 B8 B4 A2
+ A12 B12 A13 . B10 B2
O
A11 A1
Since A B A is the smaller angle formed by
12 wish
Create PDF with GO2PDF for free, if you 12to remove
13 this line, click here to buy Virtual PDF Printer B12 B20
a diagonal and a side in a regular pentagon, its A12 B14 B18 A20
measure is 36 . Since B10 B12 A12 is an interior B16
angle of a regular pentagon, its measure is 108 . A13 A19
Therefore A14 A18
A16
A15 A17
B6 B12 A13 = B6 B12 B10 + 108 + 36
= B6 B12 B10 + 144 .
To find the measure of B6 B12 B10 , we view it as an interior angle of the quadrilateral
B6 B8 B10 B12 . The sum of the interior angles of a quadrilateral is 360 . Two of the interior
angles of the quadrilateral are interior angles of a regular decagon, and therefore have measure
144 . The other two angles are also equal to one another, since they are reflections of each
other about the line A9 A19 . So we have
144 + 144 + 2B6 B12 B10 = 360 .
Dividing by 2 gives
B6 B12 A13 = B6 B12 B10 + 144 = 180 .
therefore A5 A13 passes through B6 . All 5 lines meet at B6 .
Round 3 Solutions
www.usamts.org
4/3/25. An infinite sequence (a0 , a1 , a2 , . . .) of positive integers is called a ribbon if the sum
of any eight consecutive terms is at most 16; that is, for all i 0,
ai + ai+1 + + ai+7 16.
A positive integer m is called a cut size if every ribbon contains a set of consecutive elements
that sum to m; that is, given any ribbon (a0 , a1 , a2 , . . .), there exist nonnegative integers
k ` such that
X `
ai = m.
i=k
Find, with proof, all cut sizes, or prove that none exist.
Solution
We claim that m is a cut size if and only if m is a positive multiple of 16. First we show
that any positive integer m that is not a multiple of 16 is not a cut size. We do this by
constructing counter-examples based on the highest power of 2 that divides m.
If m is odd, then m is not a sum of terms from the ribbon
2, 2, 2, 2, 2, 2, 2, 2, . . . .
Thus any cut size is even. If m is an odd multiple of 2 (meaning m 2 (mod 4)) then m is
not a sum of terms from the ribbon
3, 1, 3, 1, 3, 1, 3, 1, . . . .
Specifically, any sum of an even number of consecutive terms from this sequence is a multiple
of 4 and any odd number of terms gives an odd sum. Therefore any cut size is a multiple of
4. Likewise, the ribbon
5, 1, 1, 1, 5, 1, 1, 1, . . .
shows that any cut size must be a multiple of 8 and the period 8 ribbon
9, 1, 1, 1, 1, 1, 1, 1, . . .
shows that any cut size must be a multiple of 16.

Now we must show that for any m that is a multiple of 16, every ribbon contains a consecutive
set of terms that sums to m. Let some m = 16n be given.
Consider first all ribbons a0 , a1 , . . . such that each set of 8 consecutive terms gives a sum of
16:
ai + ai+1 + + ai+8 = 16.
We call such a ribbon maximal. Given a maximal ribbon, the sum of the first 8n terms is
a0 + a1 + + a8n = 16n = m
Round 3 Solutions
www.usamts.org
showing that m can be obtained as a sum of consecutive terms from this ribbon. Now we
are left to show that m can be achieved as a sum of consecutive terms for any ribbon that
is not maximal.
For any i let si = a0 + a1 + + ai be the sum of the first i + 1 terms. Since the ai are all
positive integers, si is a strictly increasing sequence of positive integers. We wish to find i, j
such that sj si = m, which means that ai+1 + ai+2 + + aj = m.
Since our ribbon is not maximal, there exists some k such that sk+8 sk < 16. Consider the
set T = {sk , 1 + sk , . . . 15 + sk }. Since sk+8 < 16 + sk , the nine partial sums si , si+1 , . . . , si+8
are for
Create PDF with GO2PDF members
free, if you wishof T . this
to remove Next consider
line, click U PDF
here to buy Virtual = {m
Printer + sk , m + 1 + sk , . . . m + 15 + sk }. We will show
that at least 8 of these values are also partial sums. Let sb be the largest partial sum smaller
than sk + m. This means sb+1 is an element of U . Since
sb+8 sb = ab+1 + ab+2 + + ab+8 16
we know
sb+8 16 + sb < m + 16 + sk
so sb+1 through sb+8 are all elements of U .
Now we know that T contains at least 9 partial sums from our sequence and U contains at
least 8 partial sums. By the Pigeonhole Principle, one of the sixteen sets
{sk , m + sk }, {1 + sk , m + 1 + sk }, . . . {15 + sk , m + 15 + sk }
contains two partial sums. These two give us the desired si , sj with difference m, so the
proof is complete.
Round 3 Solutions
www.usamts.org
5/3/25. For any positive integer b 2, we write the base-b numbers as follows:
(dk dk1 . . . d0 )b = dk bk + dk1 bk1 + + d1 b1 + d0 b0 ,
where each digit di is a member of the set S = {0, 1, 2, . . . , b 1} and either dk 6= 0 or k = 0.

There is a unique way to write any nonnegative integer in the above form.
If we select the digits from a different set S instead, we may obtain new representations of
all positive integers or, in some cases, all integers. For example, if b = 3 and the digits are
selected from S = {1, 0, 1}, we obtain a way to uniquely represent all integers, known as the
balanced ternary representation. As further examples, the balanced ternary representation
of the numbers 5, 3, and 25 are:
5 = (1 1 1)3 , 3 = (1 0)3 , 25 = (1 0 1 1)3 .
However, not all digit sets can represent all integers. If b = 3 and S = {2, 0, 2}, then no
odd number can be represented. Also, if b = 3 and S = {0, 1, 2} as in the usual base-3
representation, then no negative number can be represented.
Given a set S of four integers, one of which is 0, call S a 4-basis if every integer n has at
least one representation in the form
n = (dk dk1 . . . d0 )4 = dk 4k + dk1 4k1 + + d1 41 + d0 40 ,
where dk , dk1 , . . . , d0 are all elements of S and either dk 6= 0 or k = 0.
(a) Show that there are infinitely many integers a such that {1, 0, 1, 4a + 2} is not a
4-basis.
(b) Show that there are infinitely many integers a such that {1, 0, 1, 4a + 2} is a 4-basis.
Solution
Given a set S of 4 integers with one of them 0, we will say a nonzero integer n is repre-
sentable using S if we can write n in the form
(dk dk1 . . . d0 )4 = dk 4k + dk1 4k1 + + d1 41 + d0 40 ,
where dk , dk1 , . . . , d0 are all elements of S and dk 6= 0.

Part a:
We will show that if 4a + 2 is a multiple of 3, we do not get a 4-basis. Any value of a 1
4a + 2
(mod 3) will force 3 | 4a + 2 so there are infinitely many such S. Let b = and suppose
3
that S = {1, 0, 1, 3b} is a 4-basis. Since b is representable using X,
b = (dk dk1 . . . d0 )4 = dk 4k + dk1 4k1 + + d1 41 + d0 40 ,

Round 3 Solutions
www.usamts.org
where dk , dk1 , . . . , d0 are all elements of X and dk 6= 0. Choose some representation with k
minimal. Notice that
b 3b 4a + 2 2 (mod 4)
but our expression for b also gives
b (dk dk1 . . . d0 )4 d0 (mod 4).
The only choice for d0 in S that is congruent to 2 modulo 4 is d0 = 3b. Then removing the
final digit we get
bclick=here(dtokbuy
Create PDF with GO2PDF for free, if you wish to remove this line, dk1
Virtual.PDF
. . dPrinter
0 )4 = 4 (dk dk1 . . . d1 )4 + 3b.
Isolating b gives b = (dk dk1 . . . d1 )4 . However, this gives us a representation for b with
fewer digits, contradicting the minimality of k. So there could not have been a representation
of b, and {1, 0, 1, 4a + 2} is not a 4-basis.
Part b:
We claim that Sb = {1, 0, 1, 4b 2} is a 4-basis for every positive integer b. Since 4b 2 is
of the form 4a + 2, this will suffice.
For a given n, let d0 be the element of Sb such that n d0 (mod 4). Note that if
n d0
= dk 4k + dk1 4k1 + + d1 41 + d0
4
then
n = dk 4k+1 + dk1 4k + + d1 42 + d0 41 + d0 .
Therefore,
nd0
Lemma 1: If 4
is representable by Sb , then n is representable by Sb .
nd0
The number | 4 |will often be smaller than |n|, suggesting that we may be able to use the
fact that n is representable when |n| is small to prove that every n is representable. We first
0
explore when | nd4
| < |n|.
Assume that n 6= 0. If d0 = 1, 0, or 1, then
n d0 |n| + 1

4 |n|.

4
However, if d0 = 4c 2 then
n d0 |n| + (4c 2)

4

4
c
which could be larger than n. In the case that |n| 43 , we have 4c 3|n| so
n d0 |n| + (4c 2) |n| + 4c

|n| + 3|n|

4 < |n|.
4 4 4
In summary, we see that:

Round 3 Solutions
www.usamts.org
4b 0
Lemma 2: If |n| 3
then | nd
4
| < |n|.
b
Suppose that we know that every n with 0 < |n| < 43 is representable by Sb . Then we can
prove that every n is representable by strong induction on |n| as follows. Suppose that n is
0
representable when |n| N , and suppose that |n| = N + 1. If nd 4
= 0 then n is in Sb so is
0
representable. Assume not. By the first lemma, n is representable if nd 4
is representable.
nd0 4b nd0 0 b
Either 0 < | 4 | < 3 , in which case 4 is representable by assumption, or | nd 4
| 43 , in
0 0
which case | nd 4
| < |n|, by our second lemma, so that nd 4
is representable by our inductive
hypothesis. So to prove that every n is representable, it suffices to prove that every n with
4bwish to remove this line, click here to buy Virtual PDF Printer
0 < for|n|
Create PDF with GO2PDF free,<
if you
3
is representable.
As an example, the b = 1 case is follows since values of n for which 0 < |n| < 34 are 1 and
1, and these values are in S1 .
For values of n such that n is representable by S1 with k b digits, we can use such a
representation to construct a representation of n by Sb as follows. Suppose that
n = dk1 4k1 + dk2 4k2 + + d1 41 + d0 40
where d0 , d1 , . . . , dk1 S1 and k b. Then
n = (dk1 )4k1 + (dk2 )4k2 + + (d1 )41 + (d0 )40
We want to use this to represent n by Sb . Consider a new sum,
ek+b1 4k+b1 + ek+b2 4k+b2 + + e0 40
where ei is defined as follows. If i < k then

di if di {1, 0, 1}
ei = b
4 2 if di = 2
If k i < b then we set ei = 0. When i b, let

0 if dib {1, 0, 1}
ei =
1 if dib = 2
Then if di 6= 2,
ei+b 4i+b + ei 4i = 0 + (di )4i = (di )4i .
Furthermore, if di = 2, we get
ei+b 4i+b + ei 4i = (1)4i+b + (4b 2)4i = 2 4i = (di )4i

Round 3 Solutions
www.usamts.org
as well. Therefore
k1
X
n= (di )4i
i=0
k1
X
= ei+b 4i+b + ei 4i
i=0
k+b1
X
= ei 4i .
i=0
Each ei is in Sb so we are done.

b
We now use mathematical induction to prove that every n with |n| < 43 can be represented
by S1 using b or fewer digits.
The b = 1 case is that 1 may be represented with a single digit. This is true since these
are indeed in S1 . Now we suppose that the claim is true for b and prove it is true for b + 1.
0
That is, if nd
4
is representable using b digits, then n is representable using b + 1 digits. So
b+1 0 b
to prove the claim, it suffices to show that if |n| < 4 3 then | nd 4
| < 43 .
If d0 {1, 0, 1} then n d0 is the nearest multiple of 4 to n. If d0 = 2 then n d0 is the next
lower multiple of 4. So replacing n by n d0 rounds n to the nearest multiple of 4, rounding
down to break ties.
b+1
Suppose that |n| < 4 3 . Since n is an integer and 4b+1 is one more than a multiple of 3,
1 4b+1 4b+1 1
n .
3 3
When we round all elements in this interval to the nearest multiple of 4 we find
4 4b+1 4b+1 4
n d0 .
3 3
This means
1 4b n d0 4b 1
.
3 4 3
n d0

making < 4b . So if |n| < 4b+1 , then | nd0 | < 4b . This proves that every n with
3 3 4 3
4
b
|n| < 43 is representable by S1 using no more than b digits.
b
Since every n with |n| < 43 is representable by S1 using at most b digits, every such n is
representable in Sb and this gives us the base cases for our induction.

Round 1 Solutions
www.usamts.org
1/1/26. Divide the grid shown to the right into more than one 6 5 6
region so that the following rules are satisfied.
4 2 4
1. Each unit square lies entirely within exactly 1 region. 3 3 4
2. Each region is a single piece connected by the edges of 4
its unit squares. 4 3
3. Each region contains the same number of whole unit 4 4 4
squares. 1 1 1
4. Each region contains the same sum of numbers.
Solution
Let there be n regions, and let a be the area of each region and s be the sum of the numbers
in each region. Observe that the area of the entire grid is 56 and that the sum of all of the
numbers in the grid is 63. Therefore, we have na = 56 and ns = 63. But we also know that
n, a, s are integers and n > 1. Therefore, n must be a nontrivial common divisor of 56 and
63. The only such number is 7, so therefore n = 7, which gives a = 8 and s = 9. Thus, there
will be 7 regions, each region will contain 8 squares, and each regions numbers will sum to
9.
Next, observe that no region can contain any two of the 6s 6 5 6
and 5 in the top right, since its sum would be too high.
Additionally, no region can contain both a number at least 4 2 4
5 and a 1, since the regions have area 8 and these numbers 3 3 4
are too far away from each other. Therefore, the only way 4
to form a region containing a 6 is to have it contain 6 and
4 3
3. Similarly, the only way to make a region containing a 5
is to have it contain a 5 and a 4. So two regions contain 4 4 4
the numbers 3,6 and one other contains 4,5; the remaining 1 1 1
numbers in the grid are 1,1,1,2,3, and seven 4s. Among these
numbers, the only way to form a sum of 9 containing a 1 is 1,4,4. Since there are three 1s,
three regions contain the numbers 1,4,4. The final region contains the numbers 2,3,4. So,
we have that the seven regions contain these sets of numbers:
1, 4, 4; 1, 4, 4; 1, 4, 4; 2, 3, 4; 3, 6; 3, 6; 4, 5.
From here, it is fairly straightforward by trial-and-error to construct the following partition

of the grid into the 7 desired regions, as shown at right above.
Round 1 Solutions
www.usamts.org
2/1/26. Find all triples (x, y, z) such that x, y, z, x y, y z, x z are all prime positive
integers.
Solution
Note that since x y, y z, x z are positive, we must have x > y > z.

We cannot have more than one of x, y, z be even, since there is only one even prime. But if
x, y, z are all odd, then x y and x z are distinct even primes, a contradiction. Therefore
exactly one of x, y, z is an even prime, and since 2 is the smallest prime, we must have z = 2.
Thus x and y are both odd. But then x y is even and prime, so x y = 2 and hence
x = y + 2. Therefore our triple is (y + 2, y, 2). This means that all of y + 2, y, and y 2 are
prime. But at least one of these is a multiple of 3, and the only multiple of 3 that is prime
is 3. The only possibility is y 2 = 3, giving y = 5 and y + 2 = 7.
Therefore, the only such triple is (x, y, z) = (7, 5, 2) .
Round 1 Solutions
www.usamts.org
3/1/26. A group of people is lined up in almost-order if, whenever person A is to the left
of person B in the line, A is not more than 8 centimeters taller than B. For example, five
people with heights 160, 165, 170, 175, and 180 centimeters could line up in almost-order
with heights (from left-to-right) of 160, 170, 165, 180, 175 centimeters.
(a) How many different ways are there to line up 10 people in almost-order if their heights
are 140, 145, 150, 155, 160, 165, 170, 175, 180, and 185 centimeters?
(b) How many different ways are there to line up 20 people in almost-order if their heights
are 120, 125, 130, 135, 140, 145, 150, 155, 160, 164, 165, 170, 175, 180, 185, 190, 195,
200, 205, and 210 centimeters? (Note that there is someone of height 164 centimeters.)
Solution
(a) We prove the following Lemma:

Lemma: If we have n people whose heights are a, a + 5, . . . , a + (n 1)5 centimeters
for some a, then there are Fn+1 ways to line them up in almost-order, where Fk is the
k th Fibonacci number, defined by F0 = 0, F1 = 1, and Fk = Fk1 + Fk2 for all k 2.
Proof of Lemma: We prove by induction on n. The base cases n = 1 and n = 2 have
F2 = 1 and F3 = 2 almost-orderings, respectively.
For the inductive step, let n > 2 be given and assume that the result holds for fewer
than n people. Let L be a valid almost-ordered line up of the n people.
If the tallest person is at the rightmost end of L, then the preceding n 1 people are
almost-ordered, so by the inductive hypothesis there are Fn such almost-orderings.
If the tallest person is not at the rightmost end of L, then he must be no more than
5 centimeters taller than any person to his right. But there is only 1 such person:
the next-tallest person who is 5 centimeters shorter than the tallest person. So we
must have the next-tallest person at the far right end of L, and the tallest person to
his immediate left. This means that the preceding n 2 people are almost-ordered
(and they are all shorter than the two people at the far-right end), so by the inductive
hypothesis there are Fn1 such almost-orderings.
Combining both cases, we get Fn + Fn1 = Fn+1 total ways to almost-order a group of
n people, completing the induction. 2
Applying this Lemma to the given set of 10 people gives F11 = 89 possible almost-
orderings.
Round 1 Solutions
www.usamts.org
(b) First, we note that only the people of heights 160 or 170 can be lined up between
the people of heights 164 and 165. There are two cases. If the person of height 164
is to the left of the person with height 165, then by the almost-ordering condition, if
someone of height h is between them, we have 164 < h + 9 and h < 165 + 9. Therefore
155 < h < 174, so h is either 160 or 170. In the case that the person of height 164 is
to the right of the person with height 165, we obtain 165 < h + 9 and h < 164 + 9, so
again h must be either 160 or 170.
We now consider four cases based on the number of people who are between the people
of heights 164 and 165.
Case 1: No person appears between the people of heights 164 and 165. By the Lemma,
there are F20 = 6765 almost-orderings of the 19 people of heights 140, 145, ..., 205,
210 (excluding the person of height 164). For any almost-order of these 19 people, the
person of height 164 can be added directly to the left or the directly to the right of
the person of height 165. Thus, the number of almost-orderings including the person
of height 164 is twice the number of almost-orderings of the 19 people of heights 140,
145, ..., 205, 210. There are 2 6765 = 13, 530 total almost-orderings in this case.
Case 2: Only the person of height 160 is between the people of heights 164 and 165.
Suppose that the person of height 164 is to the left of the person of height 165. Then
all the people of height 155 or less must be to the left of the person of height 164, and
all the people of height 170 or more must be to the right of the person of height 165.
By the Lemma, there are F9 = 34 ways to almost-order the 8 people on the left, and
F10 = 55 ways to almost-order the 9 people on the right. This yields 34 55 = 1870
almost-orderings. Since an identical analysis applies if the person of height 165 is to
the left of the person of height 164, there are 2 1870 = 3740 almost-orderings in this
case.
Case 3: Only the person of height 170 is between the people of heights 164 and 165.
all the people of height 160 or less must be to the left of the person of height 164, and
all the people of height 175 or more must be to the right of the person of height 165.
By the Lemma, there are F10 = 55 ways to almost-order the 9 people on the left, and
the left of the person of height 164, there are 2 1870 = 3740 almost-orderings in this
case.
Case 4: Both the person of height 160 and the person of height 170 are between the
people of heights 164 and 165. Because of the almost-ordering condition, the person
of height 160 must be to the left of the person of height 170. Thus, the only two
almost-orderings of these four people are 164, 160, 170, 165 and 165, 160, 170, 164.
all people of height 155 or less must be to the left of the person of height 164, and
all people of height 175 or more must be to the right of the person of height 165. By
Round 1 Solutions
www.usamts.org
the Lemma, there are F9 = 34 ways to almost-order the eight people on the left, and
the left of the person of height 164, there are 2 1156 = 2312 total almost-orderings
in this case.
Adding up all the cases gives us 13, 530 + 3740 + 3740 + 2312 = 23, 322 total almost-
orderings of the 20 people.
Round 1 Solutions
www.usamts.org
4/1/26. Let P and Q be two circles of radius 1, intersecting in points A and B. Let P
and Q be two regular n-gons (for some positive integer n 4) inscribed in P and Q ,
respectively, such that A and B are vertices of both P and Q. Suppose a third circle of
radius 1 intersects P at two of its vertices C, D and intersects Q at two of its vertices E, F .
Further assume that A, B, C, D, E, F are all distinct points, that A lies outside of , and
that B lies inside . Show that there exists a regular 2n-gon that contains C, D, E, F as
four of its vertices.
Solution
Note that the statement of the problem asks for the ex-
istence of a 2n-gon whenever the construction is com-
plete. Although the construction is not possible for all C
n (for example, it cannot be done for n = 4 or n = 5), B E
the result does hold whenever the construction can be

completed as follows. P F Q
D
Since B is inside , the three circles have nontrivial
intersection, so one of C, D is inside Q and one of A
E, F is inside P . Without loss of generality, suppose
that D is inside Q and F is inside P , as shown in the picture to the right. This means
that the order of points on is C, F, D, E.
Let s = 2n
, the length of the arc between any two adjacent points in the regular n-gon. Note
that the fact that A, B, C, D are vertices of a regular n-gon means that arcs AD, DB, and
BC
all have lengths that are positive integer multiples of s, and similarly for arcs AF, F B,
and BE.

Let x, y, z be the lengths of arcs CF , F D, and DE,

respectively. To prove our result, it
suffices to show that x, y, and z are each half-integer multiples of s, because in a 2n-gon the
length of the arc between any two adjacent points is s/2.
Note that there is an arc CD in both and , and since the circles have the same radius,
P
this arc has the same length in both circles. In circle P , we know that this arc length is a
multiple of s, so it is also a multiple of s in circle , and thus x + y is an integer multiple
of s. Similarly, by looking at arc EF in circles and , we see that y + z is an integer
Q
multiple of s.
Next, since the angles of triangle BDF add up to , we have BDF +DF B +F BD = .
But BDF is inscribed in arcs F C and BC,
so in terms of arc lengths, F C + CB
= 2BDF .
Making similar computations with the other two angles in BDF gives us
F
C + CB
+ BE
+ ED
+ DA
+ AF
= 2.
But F
C + DE
= x + z, and the other four arcs are all integer multiples of s. Therefore since
2 = ns, we have that x + z is an integer multiple of s as well.
Round 1 Solutions
www.usamts.org
Finally, observe that

(x + y) + (x + z) (y + z)
x= ,
2
so since the numerator is an integer multiple of s, we conclude that x is a half-integer multiple
of s. Similar computations show that y and z are each half-integer multiples of s. This proves
our result.
Round 1 Solutions
www.usamts.org
5/1/26. Let a0 , a1 , a2 , . . . be a sequence of nonnegative integers such that a2 = 5, a2014 = 2015,

and an = aan1 for all positive integers n. Find all possible values of a2015 .
Solution
Let a3 = x and a4 = y for some nonnegative integers x and y. Note that
x = a3 = aa2 = a5 = aa4 = ay , (1)
and that
y = a4 = aa3 = ax . (2)

x if n is odd,
Claim: For all n 3, an =
y if n is even.
Proof of Claim: We prove by induction. The base cases n = 3 and n = 4 are true by
definition.
Let n > 4 be given, and assume the claim is true for all ak with 3 k < n. If n is odd, then
an = aan1 = ay = x (by (1)). If n is even, then an = aan1 = ax = y (by (2)). Therefore the
claim is true for an , and hence for all n 3 by induction. 2
Since a2014 = 2015, we must have y = 2015. Thus the sequence is
a0 , a1 , 5, x, 2015, x, 2015, x, . . .
with ax = 2015 and a2015 = x.

Next, we must have 5 = a2 = aa1 , which leads to four possibilities:
Case 1: a1 is an odd integer greater than 1 and x = 5. But then ax = a5 = x = 5,
contradicting the fact that ax = 2015. So this case cannot happen.
Case 2: a1 is an even integer greater than 2. But then aa1 = 2015, a contradiction. So this
case cannot happen.
Case 3: a1 = 2. Now 2 = a1 = aa0 , which gives two subcases:
Subcase a: a0 = 1, giving the sequence
1, 2, 5, x, 2015, x, 2015, x, . . . .
The condition ax = 2015 forces x to be either 2015 or an even integer greater than 2.
Subcase b: a0 is an odd integer greater than 1, and x = 2. But then we must have
a2 = 2015, which cannot occur since a2 = 5. So this subcase cannot happen.
Case 4: a1 = 0 and a0 = 5. This satisfies a2 = aa1 = a0 = 5. However, a1 = aa0 = a5 = x
forces x = 0, which contradicts ax = 2015 since a0 = 5. Thus, this case cannot happen.
Thus, a2015 = x can be 2015 or any even integer greater than 2 .

Round 2 Solutions
www.usamts.org
1/2/26. The net of 20 triangles shown to the right can be folded to

form a regular icosahedron. Inside each of the triangular faces,
write a number from 1 to 20 with each number used exactly once. 2
Any pair of numbers that are consecutive must be written on 6
faces sharing an edge in the folded icosahedron, and additionally, 15
14
1 and 20 must also be on faces sharing an edge. Some numbers
have been given to you.
You do not need to prove that your answer is the only one
possible; you merely need to find an answer that satisfies the
constraints above. (Note: In any other USAMTS problem, you
need to provide a full proof. Only in this problem is an answer
Solution
3 19
2 4 18
1 6 5 17
7 15
14 16
12
13 11
9
8 10
20
Round 2 Solutions
www.usamts.org
2/2/26. Let a, b, c, x, y be positive real numbers such that
ax + by bx + cy cx + ay.
Prove that b c.
Solution
The first inequality is equivalent to
(b c)y (b a)x,
and the second inequality is equivalent to
(b c)x (a c)y.
If b > c, then the left sides of the two inequalities above are positive, so the right sides are
positive as well. In particular, this means that b a > 0 and a c > 0. But then b > a > c,
giving ax > cx and by > ay. Adding these last two inequalities gives ax+by > cx+ay, which
contradicts the relation between the first and third expressions in the original inequality
chain.
Thus b c.
Round 2 Solutions
www.usamts.org
3/2/26. Let P be a square pyramid whose base consists of the four vertices (0, 0, 0), (3, 0, 0),
(3, 3, 0), and (0, 3, 0), and whose apex is the point (1, 1, 3). Let Q be a square pyramid whose
base is the same as the base of P, and whose apex is the point (2, 2, 3). Find the volume of
the intersection of the interiors of P and Q.
Solution
Consider the cross-sections of P and Q given by the plane z = c for some 0 c 3.

The cross-section of P with z = c is the square with vertices:
Create PDF with GO2PDF for free, if you wish to remove this line, click
c c here to 2c buy Virtual PDF
c
Printer

2c 2c

c 2c

, ,c , 3 , ,c , 3 ,3 ,c , ,3 ,c .
3 3 3 3 3 3 3 3
The cross-section of Q with z = c is the square with vertices:

2c 2c c 2c c c 2c c
, ,c , 3 , ,c , 3 ,3 ,c , ,3 ,c .
3 3 3 3 3 3 3 3
Thus the point (x, y, z) is in the interior of both P and Q if and only if:
z 2z 2z z
<x<3 and <x<3 ,
3 3 3 3
z 2z 2z z
<y <3 and <y <3 .
3 3 3 3
Combining these conditions gives
2z 2z 2z 2z
<x<3 and <y <3 .
3 3 3 3
This is equivalent to
(x, y, z)
lying in the interior of a pyramid with the same base as P and
3 3 9
Q, but with apex , , . (We note that the two inequalities above only have solutions
2 2 4
9
for z < , and are symmetric in x and y.) This pyramid has a base of area 9 and a height
4
9
of , so its volume is
4
1 1 9 27
bh = (9) = .
3 3 4 4
Round 2 Solutions
www.usamts.org
4/2/26. A point P in the interior of a convex polyhedron in Euclidean space is called a

pivot point of the polyhedron if every line through P contains exactly 0 or 2 vertices of the
polyhedron. Determine, with proof, the maximum number of pivot points that a polyhedron
can contain.
Solution
A polyhedron can contain 1 pivot point: for example, the center of a cube is a pivot point.
However, we claim that a convex polyhedron cannot contain more than one pivot point.
First, we prove the following:
Lemma: Let X be a pivot point of a convex polyhedron P, and let be any plane containing
X. Then there are an equal number of vertices in the two half-spaces determined by .
Proof: Let f be a mapping from the set of vertices of P to itself, defined by setting f (A) to

be the unique vertex of P, other than A, on the line AX. (This vertex must exist, and be
uniquely determined, by the definition of a pivot point.) Note that f is its own inverse, so
it is a bijection of the vertices of P.
If A is a vertex of P, then the points between X and A are all interior to P by convexity.
Hence, f (A) cannot lie between X and A. Symmetrically, A cannot lie between X and f (A).
Therefore X lies between A and f (A). Thus the segment Af (A) must either lie on or cross
. Therefore, f establishes a 1-1 correspondence between the vertices in one half-space and
the vertices in the other half-space, and thus there are an equal number of them. 2
Now we can prove our claim. Suppose, for the sake of contradiction, that there exists a
convex polyhedron P that contains two distinct pivot points X and Y . Choose distinct
parallel planes X through X and Y through Y such that X contains at least one vertex
of P. Let x and y be the number of vertices of P on X and Y , respectively, and note that
by construction x > 0. Also let z be the number of vertices of P that lie between X and
Y , and let x0 and y 0 (respectively) be the number of vertices in the half-space lying on the
side of X (respectively Y ) that does not contain Y (respectively X ). The diagram below
shows a side view of the planes X and Y , along with the number of points on, between,
and to either side of the planes.
x0 points
x points (x > 0) x
z points
y points y
0
y points
Round 2 Solutions
www.usamts.org
Then by the Lemma, we have
x0 = z + y + y 0 ,
y 0 = z + x + x0 .
Adding these equations and canceling x0 + y 0 from both sides gives
0 = 2z + x + y.
However, x is positive and z and y are nonnegative, giving the contradiction.

Thus, a convex polyhedron can contain at most 1 pivot point.

Round 2 Solutions
www.usamts.org
5/2/26. Find the smallest positive integer n that satisfies the following: We can color each
positive integer with one of n colors such that the equation
w + 6x = 2y + 3z
has no solutions in positive integers with all of w, x, y, z the same color. (Note that w, x, y, z
need not be distinct: for example, 5 and 7 must be different colors because (w, x, y, z) =
(5, 5, 7, 7) is a solution to the equation.)
Solution
The minimum number of colors is 4.

First, we show that 4 colors is achievable. We color every positive integer with one of four
colors according to its base-3 representation, as follows:
Number of terminating 0s Right-most nonzero digit

in base-3 representation in base-3 representation Color
even 1 red
even 2 blue
odd 1 green
odd 2 yellow
We show that, using the above coloring, there are no solutions of our equation in which all
four variables are of the same color. Assume this coloring does admit a solution. If there is a
blue solution (w, x, y, z), then the quadruple (2w, 2x, 2y, 2z) is red and is also a solution. If
there exists a green solution (w, x, y, z), then (3w, 3x, 3y, 3z) is a red solution. If there exists
a yellow solution (w, x, y, z) then (6w, 6x, 6y, 6z) is a red solution. Therefore if there is any
solution in this coloring then there is a red solution. We need only prove that this coloring
admits no red solutions.
If we are given a red solution (w, x, y, z), then each variable ends in one of 00, 01, 11, or 21
in base 3, and so each variable is congruent to 0, 1, 4, or 7 (mod 9). The values that w + 6x
can take on are summarized in the following table (with values of w along the top and values
of x along the side):
0 1 4 7
0 0 1 4 7
1 6 7 1 4
4 6 7 1 4
7 6 7 1 4
The values that 2y + 3z can take on are summarized in the following table (with values of y
along the top and values of z along the side):
Round 2 Solutions
www.usamts.org
0 1 4 7
0 0 2 8 5
1 3 5 2 8
4 3 5 2 8
7 3 5 2 8
Therefore w + 6x is congruent to 0, 1, 4, 6, or 7 (mod 9). Likewise, 2y + 3z is congruent
to 0, 2, 3, 5, or 8 (mod 9). The only way to satisfy w + 6x = 2y + 3z is for both to be
congruent to 0, meaning all four of (w, x, y, z) are multiples of 9.
The contradiction is quick from here. Let (w, x, y, z) be some red solution that minimizes w.
Since each variable is a multiple of 9, dividing by 9 gives us a new solution which is still red
(because we have removed an even number of terminating 0s), contradicting the minimality
of w.
Now we show that coloring the positive integers using 3 (or fewer) colors is insufficient. Note
that for any positive integer k, the 4-tuple (2k, k, k, 2k) is a solution to the equation, as are
the 4-tuples (3k, k, 3k, k) and (3k, 2k, 3k, 3k). Thus, k, 2k and 3k must all be different colors,
which shows that at least 3 colors are necessary.
Suppose we have only 3 colors (red, blue, green). Without loss of generality, since {1, 2, 3}
must all be different colors, we can color 1 red, 2 blue, and 3 green. Then since {2, 4, 6} and
{3, 6, 9} must each be all different colors, 6 must be a different color than both 2 and 3, so
6 must be red, which makes 4 green and 9 blue.
Below is a chart showing the colors so far:
1 2 3 4 5 6 7 8 9
R B G G R B
Next, note that {4, 8, 12} and {6, 12, 18} must each be all different colors. In particular, 12
must be a different color than both 4 and 6, so 12 must be blue. Then, 8 must be different
than both 4 and 12, so 8 must be red.
1 2 3 4 5 6 7 8 9 10 11 12
R B G G R R B B
But now, we are not able to color 5, because:
If 5 is red, then (1, 5, 8, 5) is an all-red solution to the equation.
If 5 is green, then (5, 3, 4, 5) is an all-green solution to the equation.
If 5 is blue, then (9, 5, 12, 5) is an all-blue solution to the equation.
Therefore, we cannot color the integers using just 3 (or fewer) colors.

Round 3 Solutions
www.usamts.org
1/3/26. Fill in each blank unshaded cell

3 59
with a positive integer less than 100, such
that every consecutive group of unshaded
cells within a row or column is an arith-
10
metic sequence.
You do not need to prove that your an-
swer is the only one possible; you merely 31 26 59
need to find an answer that satisfies the
constraints above. (Note: In any other USAMTS problem, you need to provide a full proof.
Only in this problem is an answer without justification acceptable.)
Solution
3 6 9 12 15 18 59 71 83 95
13 18 20 35 50 65
3 10 17 24 22 41 59 77 95
27 30 24 32 77
51 46 41 36 31 26 23 35 47 59
Round 3 Solutions
www.usamts.org
2/3/26. Let A1 A2 A3 A4 A5 be a regular pentagon with side length 1. The B1

sides of the pentagon are extended to form the 10-sided polygon shown
A3 A4
in bold at right. Find the ratio of the area of quadrilateral A2 A5 B2 B5 B5 B2
(shaded in the picture to the right) to the area of the entire 10-sided A2 A5
polygon. A1
B4 B3
Solution
Decompose the pentagon into 8 triangles as shown below with areas x, y, and z (by symmetry,
A3 A4 A5 = A5 A1 A2 , so these two triangles have the same area, y). Note that 4 triangles are
shaded and 4 are white.
B1
x
B5 A3 A4 B2
x y x
z
A2 y A5
x A1 x
B4 B3
The interior angles of the pentagon are 108 , so

1
A4 A3 A5 = (180 108 ) = 36 ,
2
and thus
A2 A3 A5 = 108 36 = 72 .
From this, we see that the triangle with area z is isosceles with base angles 72 . Each of the
outer triangles has base angle equal to the exterior angle of a pentagon, so these are also
isosceles with base angle 72 . Since z and x are areas of similar triangles that share a base,
these triangles are congruent and x = z.
The white area is then 3x + y and the shaded area is also 2x + y + z = 3x + y. Therefore
1
the shaded area is exactly of the entire area.
2
Round 3 Solutions
www.usamts.org
3/3/26. Let a1 , a2 , a3 , . . . be a sequence of positive real numbers such that:

(i) For all positive integers m and n, we have amn = am an , and
(ii) There exists a positive real number B such that for all positive integers m and n with
m < n, we have am < Ban .
Find all possible values of log2015 (a2015 ) log2014 (a2014 ).
Solution
k Printer
Condition (i) implies that a = (a ) for all positive integers n and k, by induction on k.
nk n
Hence,
(an )k = ank < Ba(n+1)k = B(an+1 )k ,
k
an an
and thus an+1 < B. As k can be arbitrarily large, we must have an+1 1, or an an+1 .
That is, a1 , a2 , a3 , . . . is a nondecreasing sequence. Also note that condition (i) with m = 1
gives an = a1 an , so a1 = 1.
If a2 = 1, then a2k = 1k = 1 for all positive integers k, and hence (because the as are
nondecreasing) we must have an = 1 for all n. In this case, log2015 (1)log2014 (1) = 00 = 0.
If a2 = > 1, then we claim that logn (an ) = log2 () for all n 2. To prove the claim, let
p be any positive integer such that 2p > n, and then set q to be the unique positive integer
such that
nq < 2p nq+1 .
Since the as are nondecreasing, anq a2p anq+1 . Thus, (an )q (a2 )p (an )q+1 . Because
a2 = ,
(an )q p (an )q+1 .
Taking the logarithm base 2 of both inequality chains gives
q log2 (n) < p (q + 1) log2 (n),
q log2 (an ) p log2 () (q + 1) log2 (an ).
Since all of these logarithms are positive, we can divide the bottom inequality chain by the
top one (in reverse order) to get
q log2 (an ) (q + 1) log2 (an )
log2 () .
(q + 1) log2 (n) q log2 (n)
This simplifies to

q q+1
logn (an ) log2 () logn (an ).
q+1 q
q
But as p grows, q will grow as well, and both q+1 and q+1q
can be made arbitrarily close to
1. Therefore, logn (an ) = log2 (), proving the claim.
Hence, in this case, log2015 (a2015 ) log2014 (a2014 ) = log2 () log2 () = 0.
Therefore, the only possible value of log2015 (a2015 ) log2014 (a2014 ) is 0.
Round 3 Solutions
www.usamts.org
4/3/26. Nine distinct positive integers are arranged in a circle such that the product of any
two non-adjacent numbers in the circle is a multiple of n and the product of any two adjacent
numbers in the circle is not a multiple of n, where n is a fixed positive integer. Find the
smallest possible value for n.
Solution
For any positive integer m and prime p, let p (m) be the exponent of the largest power of p
that divides m. In other words, p (m) is the exponent of p in the prime factorization of m.
Notefor free,
Create PDF with GO2PDF also that
if you wish to the
removefunction pto buy
this line, click here satisfies p (ab) = p (a) + p (b) for any positive integers a, b.
Virtual PDF Printer
Assume that n satisfies the condition above with the numbers on the circle being x1 , x2 , . . . , x9 .
Set x10 = x1 and x11 = x2 (to simplify the notation). For each 1 i 9, since n does not
divide xi xi+1 , there must be a prime p such that p (xi xi+1 ) < p (n). Let pi be one such
prime, so pi (xi ) + pi (xi+1 ) < pi (n).
Suppose that pi = pj for two nonconsecutive indices i, j (where 1 and 9 are treated as
consecutive indices). Note that xi , xi+1 , xj , xj+1 are all distinct. Let p = pi = pj (to make
the notation simpler), and set p (n) = k. Then we have p (xi ) + p (xi+1 ) < k and p (xj ) +
p (xj+1 ) < k. Summing, this gives
p (xi ) + p (xi+1 ) + p (xj ) + p (xj+1 ) < 2k. (1)
On the other hand, xi and xj are nonadjacent, so xi xj is a multiple of n, and hence p (xi ) +
p (xj ) k. Similarly, xi+1 and xj+1 are nonadjacent, so p (xi+1 ) + p (xj+1 ) k. Summing,
this gives
p (xi ) + p (xj ) + p (xi+1 ) + p (xj+1 ) 2k. (2)
(2) contradicts (1), so we cannot have pi = pj for two nonconsecutive indices i, j.
Now suppose that pi = pi+1 . Again, set p = pi = pi+1 to simplify the notation. If p (n) = 1,
then neither xi xi+1 nor xi+1 xi+2 is a multiple of p, so none of xi , xi+1 , xi+2 is a multiple of p.
But xi xi+2 must be a multiple of n, and hence a multiple of p, a contradiction. Therefore,
p (n) 2, and hence p2 divides n.
To summarize: a prime can appear at most twice in the sequence p1 , p2 , . . . , p9 . All primes
p1 , . . . , p9 must divide n, since pi (n) > 0 by definition. If any prime appears twice, we
know that the square of that prime divides n. So, we may conclude that p1 p2 p9 divides
n. We minimize this product by choosing the four smallest primes twice each and the fifth
smallest prime once. This gives n p1 p2 p9 22 32 52 72 111 , so the minimum possible n is
n = 22 32 52 72 111 = 485100. This is achievable by using these numbers in order:
n n n n n n n n n
, 2, , 2, , 2, , 2, .
2 11 2 2 3 3 3 5 5 5 7 7 7 11
Round 3 Solutions
www.usamts.org
5/3/26. A finite set S of unit squares is chosen out of a large grid of unit squares. The
squares of S are tiled with isosceles right triangles of hypotenuse 2 so that the triangles do
not overlap each other, do not extend past S, and all of S is fully covered by the triangles.
Additionally, the hypotenuse of each triangle lies along a grid line, and the vertices of the
triangles lie at the corners of the squares. Show that the number of triangles must be a
multiple of 4.
Solution
Think of the hypotenuse of each triangular tile as a perfectly reflecting mirror and the legs
as transparent glass. From the midpoint of the hypotenuse of one of the tiles, shine a laser
toward the midpoint of either leg of the tile. The unit square containing the leg must be
completely covered by tiles, so a second tile lies on the other side of this leg, and the laser
will strike the midpoint of that tiles hypotenuse and be reflected. Therefore, the laser will
be reflected continually, and it will first retrace its path when it returns to the tile it started
in. We will show the number of unit squares the laser passed through on this loop must be
a multiple of 4.
Let a be the number of squares in which the laser travelled northwest, b be the number of
squares in which the laser travelled northeast, c be the number of squares in which the laser
travelled southwest, and d be the number of squares in which the laser travelled southeast.
Since the laser returned to where it started, it travelled north as many times as it did
south. Thus, a + b = c + d. Likewise, it travelled west as many times as it did east, so
a + c = b + d. Finally, the laser alternated between travelling northwest or southeast and
travelling northeast or southwest each time it was reflected. So there is a 1-1 correspondence
between the two pairs of move types, and a + d = b + c. That is, we have
a + b = c + d, a + c = b + d, a + d = b + c.
Adding the first two equations gives 2a = 2d so a = d. Likewise adding the first and third
equation gives a = c and the final two equations gives a = b so a = b = c = d. The total
number of squares the laser passed through is a + b + c + d, which is 4a and thus a multiple
of 4.
To complete the proof, place a laser in one of the tiles as specified above, and remove all
the tiles it passes through from the board. This removes whole unit squares only, and the
number of unit squares removed is a multiple of 4. Repeat this process: placing a laser in
a tile and removing all the tiles it passes through. Eventually we must remove all the tiles,
and since the number of tiles removed for each laser was a multiple of 4, the total number
of tiles was also a multiple of 4, as desired.

Round 1 Solutions
www.usamts.org
1/1/27. Fill in the spaces of the grid to the right with positive
integers so that in each 2 2 square with top left number a, top 3 9
right number b, bottom left number c, and bottom right number
d, either a + d = b + c or ad = bc.
11 7 2
You do not need to prove that your answer is the only one pos- 10 16
sible; you merely need to find an answer that satisfies the con-
straints above. (Note: In any other USAMTS problem, you need 15
to provide a full proof. Only in this problem is an answer without
justification acceptable.) 20 36 32
Solution
Consider the following 2 2 square
a b
c d
and suppose a < c. By hypothesis, we must have either a + d = b + c or ad = bc. Either way,
it follows that b < d. Since 9 < 11, repeatedly applying this observation tells us that every
number in the top row is less than the number directly below it. Similarly, every other pair
of consecutive rows must satisfy the same constraint, and we conclude that every column is
strictly increasing from top to bottom.
Given the same 2 2 configuration, if a positive integer n divides c but not ad, then we must
have a + d = b + c.
Finally, consider the following 3 2 configuration:
a e c
b f d
In this case, we claim that if gcd(a, b) = 1 and b a > d c, then we must have a + f = b + e.
To see this, let b = a + k and suppose af = be. This can be rearranged to a(f e) = ke, or
a e
= .
ba f e
Since a and b a are relatively prime, we conclude that f e b a and e a. This implies
that we must have f c = ed. Rewriting e as afb yields ad = bc, and by the same argument we
have d c b a, a contradiction. So we must have a + f = b + e.
Combining these observations with a little trial-and-error, one can construct the following
unique solution.
Round 1 Solutions
www.usamts.org
3 9 12 6 1
5 11 14 7 2
10 22 28 21 16
15 27 33 26 21
20 36 44 37 32
Round 1 Solutions
www.usamts.org
2/1/27. Suppose a, b, and c are distinct positive real numbers such that
abc = 1000,
bc(1 a) + a(b + c) = 110.
If a < 1, show that 10 < c < 100.
Solution
By adding the two equations we have ab + ac + bc = 1110. So a, b, and c are the roots of the
polynomial
f (x) = (x a)(x b)(x c) = x3 dx2 + 1110x 1000,
where d > 0. We compare this to the polynomial
g(x) = (x 1)(x 10)(x 100) = x3 111x2 + 1110x 1000.
Since a < 1, we know that g(a) < 0. Notice that g(x) f (x) = (d 111)x2 . Since
g(a) = g(a) f (a) < 0, we conclude that g(x) f (x) = (d 111)x2 is negative for all x 6= 0.
Therefore,
g(b) = g(b) f (b) < 0
and
g(c) = g(c) f (c) < 0.
This means that b and c are in (0, 1) (10, 100). Since abc = 1000, we see that bc > 1000,
which implies that b and c are both greater than 10. Thus, 10 < b, c < 100, as desired.
Note: This proof can be generalized. Given any two cubic polynomials differing in exactly
one coefficient and having six distinct positive roots, one can show that there are only two
possible orderings of the six roots.
Round 1 Solutions
www.usamts.org
3/1/27. Let P be a convex n-gon in the plane with vertices labeled V1 , . . . , Vn in counterclock-
wise order. A point Q not outside P is called a balancing point of P if, when the triangles
QV1 V2 , QV2 V3 , . . . , QVn1 Vn , QVn V1 are alternately colored blue and green, the total areas of
the blue and green regions are the same. Suppose P has exactly one balancing point. Show
that the balancing point must be a vertex of P .
Solution
Define a function f : P R by setting

f (Q) = [QV1 V2 ] [QV2 V3 ] + + (1)n1 [QVn V1 ].
That is, f (Q) is the area of the blue regions minus the area of the green regions given a fixed
point Q. We examine what happens to f as we move Q along a fixed line. To do this, we
look at the area of the triangle with base V1 V2 .
V1
V2
The base stays fixed and the height changes linearly as Q moves along this line. The same
is true for all other sides of P , which means that f (Q) changes linearly as Q moves along a
fixed line inside P .
Now suppose A is not a vertex and f (A) = 0. Well show that P contains at least one more
balancing point. In order to show this, we will consider two cases: (1) A is in the interior of
P and (2) A is on an edge of P .
Suppose first that A is in the interior of P . Take two segments `1 and `2 containing A in
their interiors inside P . If either segment contains another balancing point, were done. So
suppose neither does. Since f changes linearly as we move along either segment, we know
that we can find two points B and C on `1 and `2 , respectively, such that f (B) and f (C)
have opposite signs.
V1 B C
A
V2 `1
`2
By convexity, we know that the segment connecting B and C is completely contained in P .

Since f changes linearly along this segment, there must be some D between B and C for
which f (D) = 0. Therefore, we have found another balancing point.
Round 1 Solutions
www.usamts.org
Now suppose that A is on an edge, Vi Vi+1 . Pick some point R inside P and consider the
segment QR. Again, if either Vi Vi+1 or QR contains another balancing point were done, so
suppose neither does. Since f changes linearly as we move along Vi Vi+1 and QR, we know
that we can find two points B 0 and C 0 on Vi Vi+1 and QR, respectively, such that f (B 0 ) and
f (C 0 ) have opposite signs. By convexity, we know that the segment connecting B 0 and C 0 is
completely contained in P . Since f changes linearly on this segment, there must be some D0
between B 0 and C 0 for which f (D0 ) = 0, and once again we have found another balancing
point.
Combining both of these cases, we see that if P has exactly one balancing point, it must be
a vertex.
Note: A hexagon with vertices at (0, 0), (0, 2), (1, 10), (3, 10), (4, 14/3), and (4, 8/3) has
exactly one balancing point at x = 0.
One can calculate the areas of the four triangles (from left to right) to be 1, 10, 13, and 4,
and we see that (0, 0) is indeed a balancing point.
Round 1 Solutions
www.usamts.org
4/1/27. Several players try out for the USAMTS basketball team, and they all have integer
heights and weights when measured in centimeters and pounds, respectively. In addition,
they all weigh less in pounds than they are tall in centimeters. All of the players weigh at
least 190 pounds and are at most 197 centimeters tall, and there is exactly one player with
every possible height-weight combination.
The USAMTS wants to field a competitive team, so there are some strict requirements.
(i) If person P is on the team, then anyone who is at least as tall and at most as heavy
as P must also be on the team.
(ii) If person P is on the team, then no one whose weight is the same as P s height can
also be on the team.
Assuming the USAMTS team can have any number of members (including zero), how many
distinct basketball teams can be constructed?
Solution
We interpret players as points in the plane with coordinates (x, y) = (weight, height) and
interpret the requirements graphically. Each player is a lattice point on or within the right
triangle defined by the lines y = x + 1, x = 190, and y = 197.
(196, 197)
height
(190, 191)
weight
Requirement (i) states that if a player (a, b) is on the team, then so is everyone on or within
the rectangle above and to the left of (a, b). An example for player (192, 194) is shown below.
(196, 197)
height
(190, 191)
weight
Requirement (ii) states that if a player (a, b) is on the team, then no one on the line x = b
can be on the team. Combined with requirement (i), this additionally implies that no one to
the right of the line x = b can be on the team if a player (a, b) is on the team. An example
for player (190, 193) is shown below.
Round 1 Solutions
www.usamts.org
height
(190, 193)
x = 193
weight
Notice that the red line is determined by the shortest player selected for the team on the line
x = 190. Call this player p0 , and let his height be 190 + m. Then requirement (ii) tells us
that all players selected for the team are on the lines x = 190, x = 191, . . . , x = 190+(m1).
Requirement (i) implies that the players selected for the team on the line x = 190 + k are
determined by the shortest player selected for the team on this line. Let pk be the shortest
player on the line x = 190 + k. Then it suffices to count the number of ways to select
p0 , . . . , pm1 . (If there are no players selected for the team on the line x = 190 + k, we simply
ignore pk , . . . , pm1 .)
To do this, draw a path containing only steps up and to the right starting at p0 and ending
at one of the m dots in the top row to the left of the line x = 190 + m, such that the lowest
point on the path with x-coordinate 190 + k is pk (if there are no players with x-coordinate
190 + k, our path will stay completely to the left of the line x = 190 + k). An example path
is shown below with p0 = (190, 193), p1 = (191, 194), and p2 = (192, 195).
height
(190, 193)
x = 193
weight
These paths correspond to teams because they determine p0 , . . . , pm1 . Notice that these
paths are in one to one correspondence with paths starting at p0 that end at (189 + m, 198):
we simply remove the portion of any given path above the line y = 197. Each path from our
initial player to (189 + m, 198) must have m 1 steps to the right and 7 (m 1) steps up.
7
Thus the total number of paths from our initial player to (189 + m, 198) is m1 . Summing
over all m 7, we get
7
X 7
= 27 1
m=1
m 1
paths. This corresponds to the number of teams with at least one player, so in total the
number of possible teams is 27 = 128.
Challenge: Can you give a bijective proof mapping possible teams to binary strings of 0s
and 1s of length 7?
Round 1 Solutions
www.usamts.org
5/1/27. Find all positive integers n that have distinct positive divisors d1 , d2 , . . . , dk , where
k > 1, that are in arithmetic progression and
n = d1 + d2 + + dk .
Note that d1 , . . . , dk do not need to be all the divisors of n.
Solution
n = 6 works by taking d1 = 1, d2 = 2, and d3 = 3. Similarly, n = 6m works by taking

d1 = m, d2 = 2m, and d3 = 3m. We will show that these are the only possible values of n.
Well assume without loss of generality that d1 < d2 < < dk . Then, by dividing out by
any common factor, well assume that d1 , d2 , . . . , dk share no (non-trivial) common divisor,
making our goal to show that n = 6.
Suppose two consecutive terms di and di+1 have a common divisor r. Then r| (di+1 di ),
which is the common difference of the arithmetic sequence. This implies that r divides all
di , but we assumed that d1 , d2 , . . . , dk shared no common divisor, so r = 1.
Since dk and dk1 are relatively prime and are both factors of n, we have n dk dk1 . Then,
dk k, so n kdk1 . For k 3, dk1 is greater than or equal to the average of the di ,
which means that n is at least k times the average of di . That is,
n d1 + d2 + + dk .
In order for equality to hold, we need n = kdk1 = dk dk1 . The first equality implies that
dk1 is the average of the k-term arithmetic sequence, so k = 3. The second equality then
tells us that d3 = 3. That is, d1 = 1, d2 = 2, d3 = 3, and n = 6, as desired.
To conclude it now suffices to eliminate k = 2. In this case, we have n = d1 + d2 and d1 6= d2 ,
which means that d2 > n2 , contradicting the fact that d2 is a divisor of n. Thus k = 2 is
impossible.

Round 2 Solutions
www.usamts.org
1/2/27. In the grid to the right, the shortest path through unit squares between the pair of
2s has length 2. Fill in some of the unit squares in the grid so that
(i) exactly half of the squares in each row and column 6

contain a number,
10 1
(ii) each of the numbers 1 through 12 appears exactly
twice, and
2 8
(iii) for n = 1, 2, . . . , 12, the shortest path between
5 2
the pair of ns has length exactly n. 7 9
3
acceptable.)
Solution
In this solution, we use the notation Ri Cj to denote the square in row i and column j. For
example, the given 5 is in R4 C2 .
The larger numbers give us the strongest restrictions, so we examine them first.
The only pairs of squares that are a distance of 12 apart are 11 6 12
the two pairs of diagonally opposite corners (R1 C1 , R6 C8 )
10 1
and (R1 C8 , R6 C1 ). However, the only square a distance of
10 from the given 10 is R6 C8 . So the 10 must be placed there. 2 8
Thus, we can place the 12s in the corners (R1 C8 , R6 C1 ). 5 2
For the pair of 11s to be a distance of 11 apart, one of them 7 9
must be in a corner. Since three corners are already taken, we
12 3 11 10
place an 11 in the remaining corner, R1 C1 . The two squares
that are a distance of 11 from this are R5 C8 and R6 C7 . Since column 8 already contains 12,
8, and 10, we cannot place any more numbers in that column and we must place the 11 in
R6 C7 .
Round 2 Solutions
www.usamts.org
Next we resolve the 9, 8, and 7. There are two possible posi- 11 7 6 12

tions for the 7: it can be either in R3 C1 or R1 C3 . Similarly,
9 10 1
the remaining 8 can be placed in R4 C1 , R6 C3 , or R5 C2 and
the remaining 9 can be placed in R2 C1 or R1 C2 . The 8 can- 2 8
not be placed in R6 C3 because row 6 already contains 12, 3, 5 2
11, and 10. So all possible positions for the non-given 8 and 8 7 9
9 are in columns 1 and 2. This means that columns 1 and 2
will each contain 3 numbers after the 8 and 9 are placed. So 12 3 11 10
we cannot place the 7 in column 1, as this would force column 1 to contain 4 numbers. So
we place the 7 in R1 C3 . This means that row 1 contains four numbers and we must place
the 9 in R2 C1 . Once the 9 is placed, column 1 contains three numbers and we must place
the 8 in R5 C2 .
Now we are forced to place the 6 in R5 C3 . This forces us to place the 1 in R2 C4 . The only
way to place two more numbers in row 3 is to place the 5 in R3 C6 and the 4 in R3 C7 . We
conclude by placing the remaining 4 in R4 C4 and the remaining 3 in R4 C6 .
11 7 6 12
9 10 1 1
2 5 4 8
5 4 2 3
8 6 7 9
12 3 11 10
Round 2 Solutions
www.usamts.org
2/2/27. A net for a polyhedron is cut along an edge to give two pieces. For example, we may
cut a cube net along the red edge to form two pieces as shown.
Are there two distinct polyhedra for which this process may result in the same two pairs of
pieces?
Solution
The answer is yes and there are many possible solutions. We present one here.
We start with a cube and glue square pyramids on two of its faces. There are two distinct
ways to do this: the two pyramids can replace either adjacent faces or opposite faces. We
show that we can find nets for each of these polyhedra that produce the same pair of polygons.
In the diagram below, we replace two opposite faces in the cube net with square pyramids,
then cut the resulting net along the red edge to produce two polygons.
In the diagram below, we replace two adjacent faces in the cube net with square pyramids,
then cut the resulting net along the red edge to produce two polygons.
Round 2 Solutions
www.usamts.org
3/2/27. For all positive integers n, show that

n n

1 X k k! k
= 1.
n k=1 nk
Solution 1
Fix n. Let n n

X k k! k
Sr =
k=nr
nk
be the sum of the last r + 1 terms of the given sum. We claim that
n!
. Sr =
r!nnr1
For r = 0, notice that we can write the last term in the sum as
n n! nn

n n! n!
n
= n
= .
n n 0!nn1
So the claim holds for r = 0. Now suppose the claim holds for r = m 1 and we will show
that its true for r = m (assuming m < n).
The sum of the last m + 1 terms is the sum of the last m terms, plus the (m + 1)st term
from the end. That is,
(n m)n!/m!
Sm = Sm1 + .
nnm
By the inductive hypothesis, the sum of the last m terms is
n!
Sm1 = .
(m 1)!nnm
So the sum of the last m + 1 terms is
n! (n m)n!/m! n!(n m) + n! m
Sm = nm
+ nm
=
(m 1)!n n (m)!nnm
n! n
=
m!nnm
n!
= .
m!nnm1
n!
Therefore, for all r < n, the sum of the last r + 1 terms is r!nnr1 . Substituting r = n 1,
we have
n
k k! nk

X n!
k
=
k=1
n (n 1)!nn(n1)1
n
= 0 = n.
n
Dividing through by n gives the result.
Round 2 Solutions
www.usamts.org
Solution 2
Suppose we roll an n-sided die repeatedly until we roll any number for the second time. We
calculate the probability that the first repeat is on the (k + 1)st roll.
In order for the (k + 1)st roll to be the first repeat, the previous k rolls must have all been
n
distinct. The number of sequences of k distinct rolls is k! k , because we can choose any
k numbers from our die to show up and place them in any order. Since there are nk total
k!(n)
sequences of k rolls, the probability that the first k rolls are distinct is nkk . Since the
(k + 1)st roll is a repeat, it must be one of the k numbers already seen. The probability of
thisforhappening
Create PDF with GO2PDF is nk .thisIn
free, if you wish to remove line, total,
click here tothe probability
buy Virtual PDF Printer that our (k + 1)st roll is the first repeat is
n
k k! nk

k k! k
= .
n nk nk+1
Summing over all k n, we find that the probability that we have at least one repeat after
n + 1 rolls is
n
k k! nk

X
.
k=1
nk+1
Since were rolling an n-sided die, were guaranteed to have a repeat by the (n + 1)st roll.
In particular,
n
k k! nk

X
k+1
= 1.
k=1
n
Factoring an n out of the denominator, we have
n n

1 X k k! k
= 1.
n k=1 nk
Round 2 Solutions
www.usamts.org
4/2/27. Find all polynomials P (x) with integer coefficients such that, for all integers a and b,
P (a + b) P (b) is a multiple of P (a).
Solution
Fix an integer b and write H(x) = P (x + b) P (b). If H is identically 0, then P (x) is a

constant function. Substituting P (x) = c, the condition says that c divides 0, which is true.
So constant functions work. For the remainder, we will assume that P (x) is non-constant.
In that case, H(x) and P (x) have the same leading term, so we can write
H(x) = P (x) + r(x),
where r(x) is either identically 0 or deg r(x) < deg P (x). For any integer a, we have that
H(a) = P (a) + r(a) is a multiple P (a). In particular, for any integer a, r(a) is a multiple of
P (a).
The degree of r(x) is less than the degree of P (x) (or r(x) is identically 0), so we can choose
some M such that for all integers a M , |P (a)| > |r(a)|. Since r(a) is a multiple of P (a)
for each such a, this implies that r(a) = 0 for all integers a M . Therefore, r(x) is a
polynomial with infinitely many zeros, and must be identically 0. Hence H(x) = P (x) and
P (x + b) P (b) = P (x).
Since b was an arbitrary integer, this equation holds for all integers b and real numbers x.
Plugging in x = b = 0 yields P (0) = 0. Plugging in x = 1, b = 1 gives us P (2) = 2P (1).
Then plugging in x = 2, b = 1 gives us P (3) = 2P (1) + P (1) = 3P (1). Continuing this way
by induction, plugging in x = k, b = 1 gives us
P (k + 1) = kP (1) + P (1) = (k + 1)P (1).
Therefore, P (x) xP (1) is a polynomial with infinitely many zeros, and must be identically
0. So if P (x) is not constant, P (x) = xP (1) is a linear function with no constant term.
Therefore, the only non-constant solution is P (x) = cx for some constant c. Substituting
P (x) = cx, the given condition says that ca + cb cb = ca is a multiple of ca, which is true.
So both of these classes of functions work and our solution is P (x) = cx or P (x) = c for
some integer c.
Note: We ended up deriving an equation of the form P (x + b) = P (x) + P (b). This is a

version of Cauchys functional equation. It turns out that if a function f : Q Q satisfies
f (x + y) = f (x) + f (y) for all x and y, then it must be of the form f (x) = cx. The proof
is similar to what we did above, except we can no longer take advantage of the fact that we
know f is a polynomial.
Round 2 Solutions
www.usamts.org
5/2/27. Let n > 1 be an even positive integer. A 2n 2n grid of unit squares is given, and it
is partitioned into n2 contiguous 2 2 blocks of unit squares. A subset S of the unit squares
satisfies the following properties:
(i) For any pair of squares A, B in S, there is a sequence of squares in S that starts with
A, ends with B, and has any two consecutive elements sharing a side; and
(ii) In each of the 2 2 blocks of squares, at least one of the four squares is in S.
An example for n = 2 is shown below, with the squares of S shaded and the four 2 2 blocks
of squares outlined in bold.
In terms of n, what is the minimum possible number of elements in S?
Solution
The answer is 32 n2 2. Let n = 2k and we will show that the minimum number of elements
in S is 6k 2 2. We begin with a Lemma.
Lemma: Let k be a positive integer. If we have a (2k + 1) (2k + 1) grid of squares, with
some squares shaded so that all shaded squares are connected by side, as in condition (i) in
the problem, and no 2 2 subgrid is entirely unshaded, then the number of shaded squares
is at least 2k 2 1.
Proof: Place the (2k + 1) (2k + 1) grid on the coordinate plane with each square having
area 1. The shaded squares together form a polygon whose area is equal to the total number
of shaded squares. Since no 2 2 subgrid is unshaded, all of the interior vertices of the
grid are inside or on the boundary of the polygon; there are 4k 2 such vertices. By Picks
Theorem, we have that the area is B2 + I 1 where B is the number of boundary vertices
and I is the number of interior vertices. As B + I 4k 2 , we have
B B+I 4k 2
+I 1 1 1 = 2k 2 1,
2 2 2
as desired.
Let S be a valid subset of the original grid, and we will show that it contains at least 6k 2 2
squares. Consider the set of vertices of the n2 blocks of 2 2 squares.
Round 2 Solutions
www.usamts.org
There are (2k + 1)2 such vertices. Construct a (2k + 1) (2k + 1) grid G of squares so
thatforeach
Create PDF with GO2PDF square
free, if you corresponds
wish to remove this line, click here toto one PDF
buy Virtual these (2k + 1)2 vertices. Shade a square in G if the
of Printer
corresponding vertex touches a square of S. An example of such a grid and shading is shown
below.
By (i), the shaded squares in G are connected by side, and by (ii) no 2 2 block of squares in
G can be fully unshaded. Therefore, by the lemma we have at least 2k 2 1 shaded squares
in G.
Next, construct a graph G0 as follows: the vertices correspond to the shaded squares in G.
Whenever we have two adjacent squares in the same 2 2 block both in S, we draw an edge
in G0 connecting the two vertices that the two squares in S touch. An example graph is
shown below.
By (i), G0 is a connected graph. Let S(G0 ) be a spanning tree of G0 . Since G0 has at least
2k 2 1 vertices, we know that S(G0 ) has at least 2k 2 2 edges.
Notice that each edge in S(G0 ) corresponds to two unit squares in the same 2 2 block both
being shaded. Since S(G0 ) contains no cycles, each edge corresponds to a unique such pair
of unit squares. We use this to place a lower bound on the total number of shaded squares
in the original grid.
In the original grid, we require at least one shaded square in each of the 4k 2 blocks. Also,
an additional shaded square in the original grid is required for each edge of S(G0 ). There
are at least 2k 2 2 edges in S(G0 ), so in total we have at least 6k 2 2 shaded squares.
Round 2 Solutions
www.usamts.org
We now exhibit a construction of 6k 2 2 shaded squares. In the accompanying diagrams,

we demonstrate the construction for n = 4. Number the rows and columns from 1 to 4k,
and denote the square in row i column j as (i, j). Shade in the 4k 2 squares (i, 2) for
2 i 4k 1.
Next, for the k values of i with 1 i 4k and i 2 (mod 4), shade in 4k 3 squares (i, j)
for 3 j 4k 1.
Finally, shade in k(2k 1) squares of the form (2i 1, 4j 1), for 2 i 2k and 1 j k.
The total number of shaded squares is

3
4k 2 + k(4k 3) + k(2k 1) = 4k 2 3k + 4k 2 + 2k 2 k = 6k 2 2 = n2 2,
2
as desired.

Round 3 Solutions
www.usamts.org
1/3/27. Fill in each space of the grid with either a 0 or a 1 so that all sixteen strings of four
consecutive numbers across and down are distinct.
0
1 0
0 1
1 0
0 1
0
0
0
1
acceptable.)
Solution
Let row k be the k-th row with a horizontal word and let column j be the j-th column with
a vertical word. (In particular, the lone 0 in the top left corner is not considered its own
row.) We use the notation Ri Cj to denote the square in row i and column j. For example,
we are given that R1 C2 is a 1, while R1 C1 is not given to us.
There are only two squares that intersect exactly one word, 0
the top left and bottom right. Every other square is part of
1 0
exactly two words. Therefore, if the bottom right square is
a 1, the total number of 0s in all 16 strings is odd. But we 0 1
know that there are 32 total 0s, so the bottom right square 1 0
must be a 0. 0 1
This leaves column 7 as the only possible position for the 0 1
string 1111. There are two possible locations for the string 0 1 0
0000: column 5 and column 6. However, if column 5 were 0 0 1
0000, then there would be 5 strings in the grid of the form
0 1 1 0
0x1x (rows 2, 4, 5, 6, and 7). But we know there are only 4
such strings, so the string 0000 must be placed in column 6.
Round 3 Solutions
www.usamts.org
Looking at the grid, we see that all strings of the form 0x1x 0
are accounted for (rows 2, 4, 7, and 8). Thus, looking at
1 0
column 3, we see that R3 C3 is 0, looking at row 5, we see
that R5 C5 is 1, and looking at row 6, we see that R6 C5 is 1. 0 1
After placing these, looking at column 5 tells us that R4 C5 1 0 0
is 1. Hence, column 5 is 1110. 0 1 1
Since row 6 is 1010, row 5 must be 1011. Since row 5 is 1 0 1
1011, column 8 must be 1001. Then column 5 and row 6 1 0 1 0
are both of the form 1x10, which means R3 C2 must be 0. 0 0 1
Hence row 3 is 1000, and column 2 must be 1100. Since row
0 1 1 0
8 is 0110, we see that row 2 must be 0111, which means that
row 4 must be 0011. Columns 2 and 3 are both of the form x100, so looking at row 1 tells
us that R1 C4 must be a 1. Finally, there are only 7 strings that start with a 0 in our grid so
far, so R1 C1 is a 0. This completes the grid.
0
0 1 0 1
0 1 1 1
1 0 0 0
0 0 1 1
1 0 1 1
1 0 1 0
0 0 1 0
0 1 1 0
Round 3 Solutions
www.usamts.org
2/3/27. Fames is playing a computer game with falling two-dimensional blocks. The playing
field is 7 units wide and infinitely tall with a bottom border. Initially the entire field is
empty. Each turn, the computer gives Fames a 1 3 solid rectangular piece of three unit
squares. Fames must decide whether to orient the piece horizontally or vertically and which
column(s) the piece should occupy (3 consecutive columns for horizontal pieces, 1 column
for vertical pieces). Once he confirms his choice, the piece is dropped straight down into the
playing field in the selected columns, stopping all three of the pieces squares as soon as the
piece hits either the bottom of the playing field or any square from another piece. All of
the pieces must be contained completely inside the playing field after dropping and cannot
partially occupy columns.
If at any time a row of 7 spaces is all filled with squares, Fames scores a point.
Unfortunately, Fames is playing in invisible mode, which prevents him from seeing the state
of the playing field or how many points he has, and he already arbitrarily dropped some
number of pieces without remembering what he did with them or how many there were.
For partial credit, find a strategy that will allow Fames to eventually earn at least one more
point. For full credit, find a strategy for which Fames can correctly announce I have earned
at least one more point and know that he is correct.
Solution
Drop a horizontal piece occupying columns 3, 4, and 5. Then drop two horizontal pieces
occupying columns 1, 2, and 3 and columns 5, 6, and 7. We illustrate this in the diagram
below (the new pieces dropped at this step are in gray).
Since the three pieces we just dropped span every column, one of latter two pieces is guar-
anteed to be our highest piece on the board at this point. Without loss of generality, we
assume the piece on the right is now our highest piece (as in the diagram above). We repeat
this pattern. When we drop the horizontal piece occupying columns 3, 4, and 5, it will be
Round 3 Solutions
www.usamts.org
stopped by the horizontal piece occupying columns 5, 6, and 7 that we dropped previously.
Each of our successive pieces will then be stopped by this piece. Thus, our board looks like
this:
At this point, Fames can simply drop one vertical piece in the middle and exclaim I have
earned at least one more point, as desired.
Note: If Fames repeatedly dropped vertical pieces in each column, he would eventually earn
a point, but he cannot know when that has happened.
Round 3 Solutions
www.usamts.org
3/3/27. For n > 1, let an be the number of zeroes that n! ends with when written in base n.
Find the maximum value of ann .
Solution
We claim that the maximum value of ann is 21 . In binary, 2! is written as 10, which ends with
1 zero. So a2 = 1, and a22 = 21 . So to conclude, it suffices to show that ann is at most 21 for all
n.
Notice that the number of zeroes that n! ends with when written in base n is the same as
the maximum power of n that divides n!.
Fix n and write n = pk m,
where p is some prime dividing n and m is relatively prime to p.
click here to buy Virtual PDF Printer
Then an is at most the largest power of pk dividing n!. By Legendres formula, the maximum
power of p dividing n! is

X n
p (n!) = .
j=1
pj
So we have
1X n 1X n nX 1
an < = .
k j=1 pj k j=1 pj k j=1 pj
1
The sum on the far right is geometric, and sums to p1
. Thus,
n
an < .
k(p 1)
Dividing through by n gives

an 1
< .
n k(p 1)
Therefore, if n has any prime divisor greater than 2, or is divisible by the square of any
1
prime, ann < 12 . So we see that the maximum value of ann is .
2
Round 3 Solutions
www.usamts.org
4/3/27. Let 4ABC be a triangle with AB < AC. Let the angle bisector of BAC meet BC
at D, and let M be the midpoint of BC. Let P be the foot of the perpendicular from B to
AD. Extend BP to meet AM at Q. Show that DQ is parallel to AB.
Solution
D
P M
A L C
Let BP meet AC at L. By construction, AP B = AP L = 90 and P AL = P AB

since AP is the angle bisector from A. So AP B and AP L are similar. Since they share side
AP , we see that AP B is congruent to AP L. So P is the midpoint of BL. Since M is the
midpoint of BC, this implies that BP M is similar to BLC with ratio 2, and P M is parallel
to AC.
Since P M is parallel to AC, triangles DP M and DAC are similar. Hence, M P
AC
= DM
DC
. To
finish, it suffices to show M DQ is similar to M BA, which means that it suffices to show
that MDB
D
=M QA
Q
.
Since AQL = P QM and P M is parallel to AL, we see that QP M is similar to QLA.
Thus M QA
Q
=M AL
P
and we just need to show that M
DB
D
=M
AL
P
.
a c
Notice that if b
= d
6= 12 , then a
b2a
= c
d2c
. Applying this to MP
AC
= DM
DC
gives us
MP MD
= .
AC 2P M CD 2M D
But 2P M = LC because BP M BLC with ratio 2. So
MP MP MP
= = .
AC 2P M AC LC AL
Similarly, we have CM = BD + DM , so CD = BD + DM + DM = BD + 2DM . Thus,
MD MD
= .
CD 2M D BD
Round 3 Solutions
www.usamts.org
Putting these together, we have

MP MD
= ,
AL BD
as desired.
Note: This problem and solution were proposed by Xinke Guo-Xue.
Round 3 Solutions
www.usamts.org
5/3/27. Let a1 , a2 , . . . , a100 be a sequence of integers. Initially, a1 = 1, a2 = 1 and the

remaining numbers are 0. After every second, we perform the following process on the
sequence: for i = 1, 2, . . . , 99, replace ai with ai + ai+1 , and replace a100 with a100 + a1 . (All
of this is done simultaneously, so each new term is the sum of two terms of the sequence
from before any replacements.) Show that for any integer M , there is some index i and some
time t for which |ai | > M at time t.
Solution
Fix a positive integer M . Throughout this solution we consider indices modulo 100, so a102
is really just a2 . Let an,k denote an after k seconds, with an being the original sequence
(an = an,0 ). We want to show that the for some n and k we have |an,k | > M . We have
a1,1 = a1 + a2
a1,2 = a1,1 + a2,1 = (a1 + a2 ) + (a2 + a3 ) = a1 + 2a2 + a3
a1,3 = a1,2 + a2,2 = (a1 + 2a2 + a3 ) + (a2 + 2a3 + a4 ) = a1 + 3a2 + 3a3 + a4 .
Continuing this pattern, we claim that

k k
an,k = an + an+1 + an+2 +
1 2
for all n. We show this by induction on k. Its clearly true for k = 0. Now suppose its true
for k and well show its true for k + 1. By induction, we have
an,k = an,k1 + an+1,k1

k1 k1 k1 k1
= an + an+1 + an+2 + + an+1 + an+2 + an+3 +
1 2 1 2

k1 k1 k1
= an + + 1 an+1 + + an+2 + .
1 2 1
But by Pascals identity, k1 + k1 = kr . Hence, we have the result.

r r1
Of course, most of these terms are actually 0. Crossing out the 0s we have

n n n n
ai,n = ai + a101 + a201 + + a102 + a202 +
101 i 201 i 102 i 202 i

X n X n
= .
k k
k1i (mod 100) k2i (mod 100)
Where the last equality follows because a1 = 1 and a2 = 1.

Let be a 100th root of unity. Suppose toward a contradiction that for all i and n, |ai,n | < M .
Then by the triangle inequality

X100
i
ai,n < 100M.

i=1
Round 3 Solutions
www.usamts.org
However, we can rewrite this sum as

100 100
X X X n i X n i
ai,n i =
i=1 i=1
k k
k1i (mod 100) k2i (mod 100)
n n
!
X n k X n k+1
= .
k=1
k k=1
k
By the binomial theorem, we can rewrite this as

(1 + 1 )n (1 + 1 )n = ( 2 )(1 + 1 )n .

Taking magnitudes, we have

X100
1 n
i

2

a i,n = 1 +
.

i=1
But |1 + 1 | > 1, so we can make this absolute value arbitrarily large by considering large
enough n. This contradicts
X100
i
ai,n < 100M,

i=1
and we see that we must have some time t and index i for which |ai | > M at time t.
Note: This problem can be generalized. Its false if 100 is replaced by 2 or 3, but true for
any other replacement, by the same argument as above.

Round 1 Solutions
www.usamts.org
1/1/28. Fill in each cell of the grid with one of the numbers 1, 2 1
2, or 3. After all numbers are filled in, if a row, column, or any 3 2
diagonal has a number of cells equal to a multiple of 3, then it 2 3 2
must have the same amount of 1s, 2s, and 3s. (There are 10 2 1 1 3
such diagonals, and they are all marked in the grid by a gray 3 3 3
dashed line.) Some numbers have been given to you. 2 2 3
3 2 3 2 2 3
3 1
1 3
acceptable.)
Solution
Round 1 Solutions
www.usamts.org
2/1/28. A tower of height h is a stack of contiguous rows of squares

of height h such that
(i) the bottom row of the tower has h squares,
(ii) each row above the bottom row has one fewer square
than the row below it, and within each row the squares
are contiguous,
(iii) the squares in any given row all lie directly above
a square in the row below.
A tower is called balanced if when the squares of the tower are colored black and white in a
checkerboard fashion, the number of black squares is equal to the number of white squares.
For example, the figure above shows a tower of height 5 that is not balanced, since there are
7 white squares and 8 black squares.
How many balanced towers are there of height 2016?
Solution
Each row with an even number of squares has the same number of black and white squares.
Each row with an odd number of squares has a number of black and white squares differing
by 1. So, a balanced tower must have the same number of odd rows with one more black
square as odd rows with one more white square.
In particular, a balanced tower of height 2016 must have 504 odd rows with one more white
square and 504 odd rows with one more black square. Given the previous rows, there are
two options for each row of the tower, and for odd rows exactly one choice results in one
more white square and exactly one choice results in one more black square.
We now obtain the answer by constructive counting. Each even row past the first row can
be chosen in two ways, for 21007 choices total. Among the 1008 odd rows, we select 504 row
numbers to have one more black square, and the rest to have one more white square. This
uniquely determines each row, and there are 1008 ways to select the row numbers. The
504
1008
total number of balanced towers is 21007 .
504
Round 1 Solutions
www.usamts.org
3/1/28. Find all positive integers n for which (xn + y n + z n )/2 is a perfect square whenever
x, y, and z are integers such that x + y + z = 0.
Solution
Letting x = 2, y = 1, and z = 1, we get

xn + y n + z n
= 2n1 + (1)n ,
2
which must be a square. We deal with the cases where n is even and odd separately.
n even: We have
2n1 + 1 = m2
for some positive integer m. Subtracting 1 from both sides and factoring gives us
2n1 = (m + 1)(m 1).
This means that m + 1 and m 1 must both be powers of 2, so m = 3, and hence, n = 4.

n odd : We have
2n1 1 = m2
for some nonnegative integer m. Since n is odd, the left-hand side is a difference of squares,
so
2(n1)/2 + 1 2(n1)/2 1 = m2 .

We can see that n = 1 is a solution. If n > 1, 2(n1)/2 + 1 and 2(n1)/2 1 are relatively
prime and must both be squares. However, no two squares differ by 2, so this is impossible
and n = 1 is the only solution.
Therefore, the only possibilities are n = 1 and n = 4. We show that these actually work. If
x + y + z = 0, then
x+y+z
= 02 ,
2
so n = 1 works. For n = 4, we have
x4 + y 4 + z 4 x4 + y 4 + (x + y)4
=
2 2
= x4 + 2x3 y + 3x2 y 2 + 2xy 3 + y 4
= (x2 + xy + y 2 )2 ,
which is a perfect square. Therefore, n = 1 and 4 are both solutions.

Round 1 Solutions
www.usamts.org
4/1/28. Find all functions f (x) from nonnegative reals to nonnegative reals such that
f (f (x)) = x4 and f (x) Cx2 for some constant C.
Solution
One solution is f (x) = x2 . We will prove that this is the only solution.
f (t)
First, since f (x) Cx2 , we know that f (0) = 0. Suppose that for some t, we have t2
=
a 6= 1. Then, using the functional equation for f , we have
2 PDF Printer
Create PDF with GO2PDF for free, if you wish to remove this line, click here to buy Virtual (at2 )2
4
f (at ) = f (f (t)) = t = .
a2
Using the functional equation again gives
f (t4 ) = f (f (at2 )) = (at2 )4 = a4 t8 .
Continuing in this manner, we get a sequence htn i with t0 = t such that

n
f (tn ) = a(2) t2n .
n
However, a(2) is unbounded for any a 6= 0 or 1. So, for all x, we have either f (x) = x2 or
f (x) = 0. However, if f (r) = 0, then applying f to each side of the equation r4 = f (0) = 0,
so r = 0. Thus f (x) = 0 if and only if x = 0.
So, the only solution to the functional equation is f (x) = x2 .
Round 1 Solutions
www.usamts.org
5
5/1/28. Let ABCD be a convex quadrilateral with perimeter 2
and AC = BD = 1. Determine
the maximum possible area of ABCD.
Solution
D
S
A
x O R
P
y
B Q C
Let P , Q, R, S denote the midpoints of AB, BC, CD, DA, respectively. Varignons theorem
tells us P QRS is a parallelogram, and we also see that P Q = AC
2
, etc., so
1
P Q = QR = RS = SP = .
2
Hence P QRS is in fact a rhombus and its diagonals are perpendicular. Suppose its diagonals
meet at O.
Using the inequalities P R AD+BC
2
and QS AB+CD2
, we have
AB + BC + CD + DA 5
P R + QS =
2 4
1 1
Set x = P O = 2 P R and y = QO = 2 QS, noting that the area of P QRS is 2xy, and that
this is half the area of ABCD. We know that
5 1
x + y and x2 + y 2 = P O2 + QO2 = P Q2 = .
8 4
Therefore, 2
5 1 9
2xy =
8 4 64
9 9
so the area of P QRS is at most 64
. Thus, the area of ABCD is at most .
32

5+ 7 5 7
Note that equality occurs when ABCD is a rectangle with sides 8
and 8
.
Problems by Evan Chen, Aaron Doman, Billy Swartworth, and USAMTS Staff.

Round 2 Solutions
www.usamts.org
1/2/28. Shade in some of the regions in the grid

to the right so that the shaded area is equal for
each of the 11 rows and columns. Regions must be
fully shaded or fully unshaded, at least one region
must be shaded, and the area of shaded regions
must be at most half of the whole grid.
acceptable.)
Solution
Round 2 Solutions
www.usamts.org
2/2/28. Find all triples of three-digit positive integers x < y < z with x, y, z in arithmetic
progression and x, y, z + 1000 in geometric progression.
For this problem, you may use calculators or computers to gain an intuition about how to
solve the problem. However, your final submission should include mathematical derivations
or proofs and should not be a solution by exhaustive search.
Solution
Since x, y, and z are in arithmetic progression, we have
x+z
y= ,
2
and since x, y, and z + 1000 are in geometric progression, we have
p
y = x(z + 1000).
Combining these equations gives us

x+z p
= x(z + 1000).
2
Rearranging and squaring gives us
x2 + 2xz + z 2 = 4xz + 4000x,
or
(z x)2 = 4000x.
Since 4000 = 10 202 , we know that 10x must be a square, and hence x = 10n2 for some
positive integer n. Therefore,
(z 10n2 )2 = (200n)2 ,
or z = 10n2 + 200n = n(10n + 200).
Since z is a three-digit number, n must be at most 4. But n 4 since x is also a three-

digit number, so we must have n = 4. It follows that the only solution is
(x, y, z) = (160, 560, 960) .
Round 2 Solutions
www.usamts.org
3/2/28. Suppose m and n are relatively prime positive integers. A regular m-gon and a regular
n-gon are inscribed in a circle. Let d be the minimum distance in degrees (of the arc along
the circle) between a vertex of the m-gon and a vertex of the n-gon. What is the maximum
possible value of d?
Solution
180
The answer is .
mn
It is equivalent to consider each point on the circle as a number from 0 to 1, with each
number treated mod 1. The regular m-gon becomes an infinite arithmetic progression with
common difference 1/m, and the regular n-gon becomes an infinite arithmetic progression
1
with common difference 1/n. Then our claimed answer is equivalent to 2mn being the largest
value of the smallest difference of two terms of these progressions. The construction is sim-
1
ple: let the m-progression start at 0 and the n-progression start at 2mn . Then every term
1
in the m-progression is an integer multiple of while each term in the n-progression is a
mn
1
half-integer times . Therefore, the difference between terms in different progressions is
mn
1 1
at least .
2 mn
To show this is best,
define
d to be the smallest distance between terms of the progression.
1
We will show that d mn is a difference between their terms as well. Suppose x y = d,

WLOG with x in the m-progression and y in the n-progression (since if x y = d we can
use the terms x and y instead). Since m and n are relatively prime, we can use Bezouts
lemma to see that there are integers p, q with pm qn = 1. Consider the terms x + q/m and
y + p/n. We have
q p
pm qn 1
y+ =xy
x+ =d .
m n mn mn

1
So the difference between the terms is d mn . Since d was chosen to be the smallest
possible distance between terms of the progression, it must be the case that

1
d d ,
mn
1
which implies that d 2mn
as desired.
Round 2 Solutions
www.usamts.org
4/2/28. On Binary Island, residents communicate using special paper. Each piece of paper
is a 1 n row of initially uncolored squares. To send a message, each square on the paper
must either be colored either red or green. Unfortunately the paper on the island has be-
come damaged, and each sheet of paper has 10 random consecutive squares each of which is
randomly colored red or green.
Malmer and Weven would like to develop a scheme that allows them to send messages of
length 2016 between one another. They would like to be able to send any message of length
2016, and they want their scheme to work with perfect accuracy. What is the smallest value
of n for which they can develop such a strategy?
Note that when sending a message, one can see which 10 squares are colored and what
colors they are. One also knows on which square the message begins, and on which square
the message ends.
Solution
The answer is 2016 + 10 = 2026 .
Its clear that no smaller n can work. Once 10 out of n squares have been filled in, there
are only 2n10 ways to fill in the remaining squares. Since Weven and Malmer want to be
able to send 22016 different messages, we must have n 10 2016.
Now we present a strategy that allows Weven and Malmer to send a message of length
2016 when n = 2026.
Number the squares from 1 to 2026. Divide the squares into 10 sets of squares, where
set i consists of the squares whose number is congruent to i mod 10. Observe that each
set contains exactly one initially colored square. So it suffices to find a strategy to trans-
mit a message of length k 1 given k squares with 1 square initially covered. Malmer and
Weven can then use this strategy to send 10 interleaved submessages with a total length of
2026 10 = 2016.
Malmer and Weven will use the following decoding scheme:
(i) The message is least k squares.
(ii) If the first square is green, flip all the other squares and read that message.
(iii) If the first square is red, leave all the other squares as is and read that message.
Round 2 Solutions
www.usamts.org
Then we can encode our message as follows:
(i) If the bad square is first:
(a) If the first square is red: then fill the remaining squares with what you want your
message to be.
(b) If the first square is green: then flip the colors of your message and fill the re-
maining squares.
(ii) If the bad square is in the middle:

(a) If the bad square is in the color you desire it to be already: color the first square
red and fill the remaining squares with what you want your message to be.
(b) If the bad square is not the color you desire it to be already: color the first square
green, flip all the colors of your message and fill the remaining squares.
Round 2 Solutions
www.usamts.org
5/2/28. Let n 4 and y1 , . . . , yn real with

n
X n
X n
X
yk = kyk = k 2 yk = 0
k=1 k=1 k=1
and
yk+3 3yk+2 + 3yk+1 yk = 0
for 1 k n 3. Prove that
n
X
k 3 yk = 0.
k=1
Solution
The second condition implies that there exists a quadratic f (x) = ax2 + bx + c such
that yk = f (k) for each k. To see why, note that all such f (x) satisfy the given recurrence
relation, and we can solve the system of equations
a + b + c = y1 ,
4a + 2b + c = y2 ,
9a + 3b + c = y3
to find a particular f (x) that also matches the first three terms of our sequence of yk . Since
the entire sequence of yk is determined by those initial three terms, the entire sequence of
yk matches f (k).
Pn m
Pn
Let pm = k=1 k . The condition that k=1 yk = 0 can be rewritten as
n
X
(ak 2 + bk + c) = 0,
k=1
Pn Pn Pn
or a k=1 k2 + b k=1 k1 + c k=1 k 0 = 0. In terms of pm , this equation is
ap2 + bp1 + cp0 = 0.
Rewriting the entire first condition in the same way gives the following system of equations:
ap2 + bp1 + cp0 = 0,

ap3 + bp2 + cp1 = 0,
ap4 + bp3 + cp2 = 0.
Round 2 Solutions
www.usamts.org

p2 p1 p0 a
Let M = p3 p2 p1 and v = b . Our system of equations states that
p4 p3 p2 c

0
M v = 0 .
0
To solve this equation for v, we will try to invert M , which is possible as long as det(M ) 6= 0.
We have
det(M ) = p4 (p21 p0 p2 ) p3 (p2 p1 p0 p3 ) + p2 (p22 p1 p3 ).

There are well-known formulas for p0 , p1 , p2 , and p3 and a less-well-known formula for p4 :
p0 = n,
1
p1 = n(n + 1),
2
1
p2 = n(n + 1)(2n + 1),
6
1 2
p3 = n (n + 1)2 ,
4
1
p4 = n(n + 1)(2n + 1)(3n2 + 3n 1).
30
Using these formulas, det(M ) simplifies to
1 3
n (n + 1)2 (n 1)2 (4 n2 ),
2160
which is negative since n 4.
Since the det(M ) < 0, M is invertible, so we have

0 0
1
v = M 0 = 0 .
0 0
Therefore, we have a = b = c = 0, so all the yk are 0, which implies the desired conclusion.
Problems by Aaron Doman, Mehtaab Sawhney, Billy Swartworth, and USAMTS Staff.

Round 3 Solutions
www.usamts.org
13 7
23 8

7 12 12
acceptable.)
Solution
+ 13 28 25 23
13 3 9 2 4 7
28 1 12 11 16
8 15 14 7 8
23 5 13 10 6 8

7 12 12
Round 3 Solutions
www.usamts.org
is.

Solution
We imagine that Fames and Weven are each given a number between 1 and 12 (representing
either the day or the month of Malmers birthday respectively).
We analyze what happens at the first step, when Fames claims Weven cannot know what
Malmers birthday is. Since Weven is given the month, and the day can be any number less
than or equal to the month, he could only know Malmers birthday if he were given a 1 (as
this would imply Malmers birthday was 1/1). So, Weven must not have a 1. In order for
Fames to know that Weven doesnt have a 1, Fames also must not have a 1. So, we conclude
that neither Fames nor Weven was given a 1.
Next, Weven claims that Fames also cannot know what Malmers birthday is. Since Fames
is given the day, and the month can be any number greater than or equal to the day, he could
only know Malmers birthday if he were given a 12 (as this would imply Malmers birthday
was 12/12). So, Fames must not have a 12. In order for Weven to know that Fames doesnt
have a 12, he also must not have a 12. So, we conclude that neither Fames nor Weven was
given a 12.
Similarly, if this exchange were to happen again, Famess statement would allow us to con-
clude that neither Fames nor Weven was given a 2, and Wevens statement would allow us
to conclude that neither Fames nor Weven was given an 11.
Round 3 Solutions
www.usamts.org
In order for this to continue for as long as possible, both players must continue to eliminate
numbers until Fames eliminates 6. Once this happens, both players know that only 7 is
possible, and Malmers birthday is (7, 7) (July 7th).
Round 3 Solutions
www.usamts.org
example 3-city.
1 1 2
2 3 1
6 4 1
n
Solution
(a) An example of a 5-city with an entry that is at least 150 is shown below. You may
have found a different example.

1 2 5 3 4
26 1 6 25 10

67 69 1 15 12

326 259 138 1 11
52 121 127 26 1
(b) An n-city must have at least one 1 in every row and column. So an n-city must have
at least n 1s, and hence can have at most n2 n = 2 n2 entries greater than 1. Let

us list these entries as a1 , a2 , . . . , a2(n) from least to greatest.

2
Notice that a1 can be at most 1 + 1 = 2. Then a2 can be at most a1 + 1 = 3, and a3

can be at most a2 + a1 = 3 + 2 = 5. In general, we see that ak is at most ak1 + ak2 .
Hence, we conclude that ak Fk+2 for all k, where Fm is the m-th Fibonacci number.
Therefore, we have a2(n) F2(n)+2 . So, it suffices to show that F2(n)+2 3( 2 ) .

n
2 2 2
To that end, we claim that F2m+2 3m for all m 0. For m = 0 both sides of the
inequality are 1. Then,
F2m+2 = F2m+1 + F2m
= 2F2m + F2m1
3F2m .
Hence, by induction F2m+2 3(3m1 ) = 3m . Thus,
a2(n) F2(n)+2 3( 2 ) .
n
2 2
Round 3 Solutions
www.usamts.org
So, every entry in an n-city is at most 3( 2 ) as desired.

n
Round 3 Solutions
www.usamts.org
points x,
Solution
b2
Lemma: If a > b, then a2 + b 2 a 2a
.
Proof of Lemma: We see that
b4
a2 + b 2 + a2 + b 2 ,
4a2
and both sides are positive. Taking the square root of both sides, we get
b2 + 2a2
a2 + b 2 .
2a
Subtracting a from both sides gives the desired claim.
Suppose that the centers of mass of the Ai and Bi are MA and MB respectively, and that
d is a real number with D(Ai , MA ) d and D(Bi , MB ) d for all i. Let D(MA , MB ) = `,
and suppose for the sake of contradiction that ` > 0. Choose a real number q such that
d2
q >d+ .
2`
Let P be the point on line MA MB with D(P, MA ) = q and D(P, MB ) = q + `. We claim
that
D(P, A1 ) + D(P, A2 ) + + D(P, An ) < D(P, B1 ) + D(P, B2 ) + + D(P, Bn ).
For each i, let A0i be the projection of Ai onto line MA MB , and similarly define Bi0 . We see
that P A0i Ai is a right triangle with hypotenuse P Ai , so applying the Pythagorean Theorem
gives us D(P, Ai )2 = D(P, A0i )2 + D(A0i , Ai )2 . Since A0i is the closest point to Ai on the line
MA MB , we have D(Ai , A0i ) D(Ai , MA ) d. So, D(P, Ai )2 D(P, A0i )2 + d2 . For brevity,
0
let ui = D(P,p Ai ) and vi = D(P, Ai )p for each i. We rewrite our inequality as u2i d2 + vi2 .
Thus, ui d2 + vi , and ui vi d2 + vi2 vi . Applying the Lemma gives us
2
d2
ui vi .
2vi
Round 3 Solutions
www.usamts.org
Applying the Triangle Inequality to triangle P A0i MA , we get

D(P, A0i ) + D(A0i , MA ) D(P, MA ). Since D(A0i , MA ) D(Ai , MA ) d, we have vi + d q.
2
Combining this inequality with the inequality q > d + d2` and subtracting d, we get
2
vi q d > d2` .
2 2
d
Applying the inequality vi > d2` to the right side of the inequality ui vi 2vi
, we get
Pn
ui vi < `. So, ui < vi + `. P
And since i=1 vi = nq by D(P, MA ) = q and the definition of
the center of mass, we have ni=1 ui < nq + n`.
We also have D(P, Bi ) D(P, Bi0 ), so
n
X n
X
D(P, Bi ) D(P, Bi0 ) = nD(P, MB ) = nq + n`.
i=1 i=1
Therefore,
n
X n
X
D(P, Ai ) < nq + n` D(P, Bi ).
i=1 i=1
This is a contradiction of the condition, so ` = 0 as desired.

Round 3 Solutions
www.usamts.org
Solution
(a) Note that the right side implies left side implication is true even without the require-
ment that N is an integer. Indeed, assume that N = (q + 1q ) (r + 1r ), with q, r rational.
By expanding the product, we obtain:

1 q r 1 1 q
N = qr + + + = x+ + y+ , where x = qr and y = are rational.
qr r q x y r
We now show that the converse is true. Let N be an integer such that
1 1
N = (x + ) + (y + ),
x y
where x, y are positive rational numbers. We can write x, y as irreducible fractions
x = ab and y = dc , with a, b, c, d positive integers satisfying gcd(a, b) = gcd(c, d) = 1.
After expanding the formula for N we obtain:
(a2 + b2 )cd + (c2 + d2 )ab
N= .
abcd
Since N is an integer, it follows that ab must divide the numerator. Thus, ab must
divide (a2 + b2 )cd. Since gcd(a, b) = 1, it follows that gcd(ab, a2 + b2 ) = 1. So, we
must have that ab divides cd. Similarly, by using the fact that cd must also divide the
numerator of the fraction, we obtain that cd divides ab. It follows ab = cd. Note now
that if we let p = cb and q = ac , then we have:

1 1 c b a c
p+ q+ = + +
p q b c c a
a b c2 ab
= + + + 2
b a ab c
a b c d
= + + +
b a d c
1 1
= x+ + y+
x y
= N.
Round 3 Solutions
www.usamts.org
c2
We used above the equality ab
= dc , which is equivalent to ab = cd.
Remark. Another way to finish the proof from ab = cd is to argue that there must
exist positive integers s, t, u, v such that a = st, b = uv, c = su, and d = tv. It follows
that N = ( vs + vs )( ut + ut ).
(b) We show that if N is an integer divisible by 8, then it is not possible to write N as the
sum of two elements of S. Equivalently, it suffices to show that N is not the product
of two elements of S. (It is somewhat faster to work with the product than the sum.)
Assume N = ( ab + ab ) ( dc + dc ), where a, b, c, d are positive integers such that ab and dc
are irreducible, positive fractions. After expanding the product, we obtain
(a2 + b2 ) (c2 + d2 )
= N.
abcd
Since N is divisible by 8, it follows that 8 must divide the numerator (a2 +b2 )(c2 +d2 ).
Thus, at least one of the two factors of the product must be divisible by 4. Assume for
instance that 4 divides a2 + b2 . By considering all the possibilities for a, b (mod 4), it
is easy to see that a, b must both be even. However, this contradicts the fact that the
fraction ab is irreducible.
Remark. Another option is to show that if N has a prime factor p with p 3
(mod 4), then N cannot be written as the product of two elements of S.
(c) We show that there exist infinitely many integers N of the form N = (a + a1 )(b + 1b ),
with a, b positive integers. Note that this statement is stronger than the statement of
(c). We have:
a2 + 1 b 2 + 1

1 1
a+ b+ = .
a b b a
We construct recursively infinitely many pairs of integers (a, b), with a < b and such
that a divides b2 + 1, and b divides a2 + 1. First, observe that the pair (1, 2) works.
2
Now, assuming that (a, b) satisfies the requirements, we show that (b, b a+1 ) also works.
2
Indeed, if we let c = b a+1 , then we know that c is an integer. The fact that b < c is
equivalent to ab < b2 + 1, which is true since a < b. The fact that c divides b2 + 1
follows from b2 + 1 = ac. Finally, the fact that b divides c2 + 1 is equivalent to b divides
(b2 +1)2 +a2
a2
. This is true because (b2 + 1)2 + a2 1 + a2 0 (mod b), and gcd(a, b) = 1.
Remark. The first few pairs obtained through our construction are (1, 2), (2, 5), (5, 13), ....
The corresponding values of N are 5, 13, 68, . . ..


USAMTS 1998 - 2017 en With Solutions

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

USAMTS 1998 - 2017 en With Solutions

Uploaded by

Copyright:

Available Formats

U S A Mathematical Talent Search

1/1/10. Several pairs of positive integers (m,

3/1/10. It is possible to arrange eight of the nine numbers

5/4/10. In ABC , let D, E, and F be the midpoints of C

2/2/11. Let C be the set of non-negative integers which can be B

(b) Show that if 0 n

5/2/11. In ABC , AC > BC, CM is the median, C

and CH is the altitude emanating from C, as

shown in the gure on the right. Determine the

measure of MCH if ACM and BCH

each have measure 17 .

USA Mathematical Talent Search

3/4/11. Determine the value of

4/4/11. We will say that an octagon is integral if it is y

square congruent to square ABCD.

USA Mathematical Talent Search

3/1/12. Determine the integers a, b, c, d, and e for which

4/1/12. A sequence of real numbers s 0, s 1, s 2, has the property that

USA Mathematical Talent Search

19, while 44976 is not. 7 6 6 2

3/2/12. Given the arithmetic progression of integers

5/2/12. In ABC , segments PQ, RS, and TU are parallel to

points X, Y, and Z, as shown in the gure on the right.

Determine the area of ABC if each of the segments PQ,

form a + b 2 , where a and b are positive integers. A U R B

USA Mathematical Talent Search

5/3/12. Three isosceles right triangles are M

Find the ratios among the radii of the

COMAP Inc., Suite 210

5/4/12. Hexagon RSTUVW is constructed by starting with

squares outwardly on the sides of this triangle, and then

connecting the outer vertices of the squares, as shown in

the gure on the right. T

5/1/13. In PQR , QR < PR < PQ so that the

the right. Given that PR = PS = RT , deter

mine, with proof, the measure of PRQ .

3/2/13. Factor the expression

4/2/13. Let X = ( x 1, x 2, x 3, x 4, x 5, x 6, x 7, x 8, x 9 ) be a 9-long vector of integers. Determine X

4/3/13. As shown in the gure on

side CF into three equal parts, E

while G, H, and I divide side

Prove that in this language baabaabaa is not a word.

5/2/14. For an isosceles triangle ABC where AB = AC , it is possible A

4/3/14. The vertices of a cube have coordinates ( 0, 0, 0 ) , ( 0, 0, 4 ) , ( 0, 4, 0 ) , ( 0, 4, 4 ) ,

Complete, well-written solutions to at least two of the problems above,

2/4/14. We dene the number s as

6440-C Dobbin Road

(a) all perfect squares, or

5/1/15. A rectangular piece of paper, ABCD, is folded A B A C B

Complete, wellwritten solutions to at least two of the problems above, accompanied by

Round 2 Year 15 Academic Year 20032004

Complete, wellwritten solutions to at least two of the problems above, accompanied by a

USA Mathematical Talent Search

279 East Central St Suite 246

and postmarked by Sunday, 23 November 2003. Each participant is expected to develop

graph, the distribution would have the property that for

a hexagon is eight. Find, with proof, the pebbling number Jt

of the graph illustrated on the right.

Complete, wellwritten solutions to at least two of the problems above, accompanied by a

279 East Central Street, Suite 246

Complete, well-written solutions to at least two of the problems above, accompanied by a

1/1/16. The numbers 1 through 10 can be arranged along 1

2/1/16. For the equation

(3x2 + y 2 4y 17)3 (2x2 + 2y 2 4y 6)3 = (x2 y 2 11)3 ,