LIMIT CYCLES OF CUBIC POLYNOMIAL VECTOR

FIELDS VIA THE AVERAGING THEORY

JAUME GIN

E
2
AND JAUME LLIBRE
3
Abstract. In this paper we study the maximum number of limit
cycles that can bifurcate from the period annulus surrounding the
origin of a class of cubic polynomial dierential systems using the
averaging method. More precisely , we prove that the perturbation
of the period annulus of the center located at the origin of the
cubic polynomial dierential system x = yf(x, y), y = xf(x, y),
where f(x, y) = 0 is a conic such that f(0, 0) = 0, by arbitrary
cubic polynomial dierential systems provide at least 6 limit cycles
bifurcating from the periodic orbits of the period annulus using
only the rst order averaging method.
1. Introduction and statement of the main results
Probably the main open problem in the qualitative theory of real pla-
nar dierential systems is the determination and distribution of their
limit cycles. A classical way to produce limit cycles is by perturbing
a system which has a center, in such a way that limit cycles bifurcate
in the perturbed system from some of the periodic orbits of the pe-
riod annulus of the center of the unperturbed system, see for instance
Pontrjagin [17].
It is well known that perturbing the period annulus of the linear
center by arbitrary polynomials p and q of degree n, i.e., x = y +
p(x, y), x = x + q(x, y), we can obtain at most [(n 1)/2] limit
cycles using only the rst order Melnikov method, where [ ] denotes
the integer part function; see, for instance, [7]. Also it is known that
1991 Mathematics Subject Classication. Primary 34C23. Secondary 34C35,
34C25, 34C29.
Key words and phrases. limit cycle, averaging method, bifurcation from a center,
abelian integral.
The rst author is partially supported by a DGICYT grant number MTM2005-
06098-C02-02 and by a CICYT grant number 2005SGR 00550, and by DURSI
of Government of Catalonia Distincio de la Generalitat de Catalunya per a la
promocio de la recerca universit`aria. The second author is partially supported by
a DGICYT grant number MTM2005-06098-C02-01 and by a CICYT grant number
2005SGR 00550.
1
2 J. GIN

E AND J. LLIBRE
perturbing the quadratic polynomial dierential center x = y(1 +x),
x = x(1+x) (note that essentially it is the linear center with a straight
line of singular points) inside the polynomial dierential systems of
degree n we can obtain at most n limit cycles using only the rst order
averaging method, see [15].
There are essentially four usual methods for studying the number of
limit cycles which bifurcate from the periodic orbits of a period annulus
of a center. The rst one is based on the Poincare return map, see for
instance [2] and [6]. The second is based on the PoincarePontrjagin
Melnikov integrals or abelian integrals that are equivalent in the plane,
see Section 6 of Chapter 4 of [10] and Section 5 of Chapter 6 of [1].
The third is based on the inverse integrating factor, see Section 6 of [7]
or [8, 9, 21]. Finally, the last one in based on the averaging theory, see
for instance [3, 18, 20]. The rst two methods only give information on
the number of periodic orbits of the unperturbed system that turn into
limit cycles with the perturbation. The last two methods additionally
can give the shape of the bifurcated limit cycle up to any order in the
perturbation parameter, see [8, 9] and [14].
In this paper we are interested in the study of the number of limit
cycles that can bifurcate from the period annulus surrounding the origin
of a cubic polynomial dierential system of the form
(1) x = yf(x, y), y = xf(x, y),
where f(x, y) = 0 is a conic such that f(0, 0) = 0, when we perturb it
inside the class of all cubic polynomial dierential systems having the
origin as a singular point, that is, for suciently small we want to
study the number of limit cycles of the systems
(2)
x = yf(x, y) + P (x, y) ,
y = xf(x, y) + Q(x, y) ,
which bifurcate from the period annulus surrounding the origin of sys-
tem (1), where P(x, y) =
3

k=1
P
k
(x, y), and Q(x, y) =
3

k=1
Q
k
(x, y),
with P
k
and Q
k
homogeneous polynomials of degree k. Thus we write
P
k
=

i+j=k
p
ij
x
i
y
j
and Q
k
=

i+j=k
q
ij
x
i
y
j
.
The goal of this paper is to study for which coecients p
ij
and q
ij
system (2) has the maximum number of hyperbolic limit cycles bi-
furcating from the period annulus of system (1) up to rst order in .
Here, the exact meaning of up to rst order in is that the bifurcated
limit cycles will be computed using the rst order averaging method
LIMIT CYCLES VIA THE AVERAGING THEORY 3
described in Theorem 3. The main result of this paper is the following
one.
Theorem 1. There exist values of the coecients p
ij
and q
ij
such that
system (2) has at least 6 hyperbolic limit cycles up to rst order in
bifurcating from the period annulus surrounding the origin of system
(1).
Theorem 1 will be proved in Section 4. In this section we provide a
stronger result than Theorem 1. Thus, we give information about the
number of limit cycles of system (2) that can be obtained according
the dierent kind of conics f(x, y) = 0.
We want to say that up to now the maximum number of limit cycles
found in a cubic polynomial dierential system is 12 (see, for instance,
[25]). But this 12 limit cycles do not surround the same singular point.
As a precursor work where it was obtained 11 limit cycles see [12]. In
[24],

Zol adek provide the existence of a cubic polynomial dierential
system with 11 innitesimal limit cycles surrounding a unique singular
point. Many authors give examples of cubic polynomial dierential
systems having large limit cycles surrounding a unique singular point;
see, for instance, [11, 13, 16, 19, 22, 23]. All these last papers provide
examples of cubic polynomial dierential systems with at most 4 large
limit cycles.
In the study of the limit cycles which bifurcate from a period annulus,
the averaging method coincides with the abelian integral method in
certain cases, see for instance Theorems 5.1 and 5.2 of [3]. The following
proposition provides other examples.
Proposition 2. Let f, P and Q real polynomials in the variables x and
y. Consider a polynomial dierential system (2) for which f(0, 0) =
0. Then, the rst order averaging method coincides with the abelian
integral method for studying the number of limit cycles which bifurcate
from the period annulus of the center located at the origin when = 0.
Proposition 2 is proved in Section 3. In particular it shows that the
averaging method and the abelian integral method coincide when we
apply them to the cubic polynomial dierential systems (2).
2. The first order averaging method
In few words we can say that the averaging method [18, 20] gives a
quantitative relation between the solutions of some nonautonomous
periodic dierential system and the solutions of its averaged dieren-
tial system, which is an autonomous one. The next theorem provides
a rst order approximation in for the periodic solutions of a periodic
4 J. GIN

E AND J. LLIBRE
dierential system, for a proof see Theorem 2.6.1 of Sanders and Ver-
hulst [18] and Theorem 11.5 of Verhulst [20]. The original theorem is
given for a system of dierential equations, but since we will use it only
for one dierential equation we state them in this case.
Theorem 3. We consider the following two initial value problems
(3) x = F(t, x) +
2
G(t, x, ), x(0) = x
0
,
and
(4) y = f
0
(y), y(0) = x
0
.
where x, y, x
0
D an open interval of R, t [0, ), (0,
0
], F and
G are periodic of period T in the variable t, and f
0
(y) is the averaged
function of F(t, x) with respect to t, i.e.,
f
0
(y) =
1
T
_
T
0
F(t, y)dt.
Suppose: (i) F, F/x,
2
F/x
2
, G and G/x are dened, contin-
uous and bounded by a constant independent on in [0, ) D and
(0,
0
]; (ii) T is a constant independent of ; (iii) y(t) belongs to
D on the timescale 1/.
Then the following statements hold. (a) On the time scale 1/ we
have that x(t) y(t) = O(), as 0. (b) If p is an equilibrium point
of the averaged system (4) such that
(5)
f
0
y

y=p
= 0,
then there exists a Tperiodic solution (t, ) of equation (3) which is
close to p such that (t, ) p as 0. (c) If (5) is negative, then
the corresponding periodic solution (t, ) of equation (4) in the space
(t, x) is asymptotically stable for suciently small. If (5) is positive,
then it is unstable.
In [3] the authors have succeeded to weaken the hypotheses of the
rst order and second order averaging method using the Brouwer degree
theory, and also they present a third order averaging method.
3. Proof of Proposition 2
In order to prove Proposition 2 we need to recall the basic result on
abelian integrals. Consider a small C
1
perturbation of a C
1
Hamilton-
ian system, i.e. consider
x =
H
y
+ A(x, y, ), y =
H
x
+ B(x, y, ).
LIMIT CYCLES VIA THE AVERAGING THEORY 5
A periodic orbit (c) lying on a level curve H = c is called a generating
periodic orbit if the perturbed system has a limit cycle which depends
continuously on for small || and which tends to (c) as 0.
The abelian integral method for Hamiltonian systems is stated in the
following theorem, a proof of it can be found in [17].
Theorem4. If (c) is a generating periodic orbit, then I(c) =
_
(Bdx
Ady) = 0 where the integral is computed along (c). Conversely, if c
is simple zero of I(c), then (c) is a generating periodic orbit, and the
limit cycle of the perturbed system that it generates is hyperbolic for
small || and depends continuously on .
Proof of Proposition 2: As we have seen, the application of the rst
order averaging method to system (2), reduces to compute the number
of simple zeros of the function f
0
(r) dened by
(6)
f
0
(r) =
1
2
_
2
0
cos P(r cos , r sin ) + sin Q(r cos , r sin )
f(r cos , r sin )
d.
On the other hand, in order to apply the method of the abelian inte-
gral to systems (2) we must write it as a perturbation of a Hamiltonian
system, i.e. we consider the equivalent system
x = y +
P(x, y)
f(x, y)
, y = x +
Q(x, y)
f(x, y)
.
Taking into account that we integrate on the circles with constant ra-
dius the method of the abelian integral (see Theorem 4) applied to
these systems gives the same integral (6) multiplied by 2r. This
completes the proof of Proposition 2.
In general the averaging method and the method of abelian inte-
gral do not coincide because, in general, it is not possible to write an
arbitrary system of the form
x = f(x, y) + P(x, y), y = g(x, y) + Q(x, y),
into the canonical form which needs the averaging method described
in Theorem 3. Another dierence between both methods is that the
averaging method restricted to onedimensional dierential equation
(as it is stated in Theorem 3) only says that if we have a simple zero of
the averaged function, then this zero provides a hyperbolic limit cycle.
But in general the converse is not true. While this converse always is
true for the abelian integral method, see Theorem 4.
In short, the results obtained in this paper are equivalent to study
systems (2) with the method of the abelian integral. Other interesting
6 J. GIN

E AND J. LLIBRE
work for the future will be the study of the number of limit cycles which
bifurcate from the period annuli studied here but using the averaging
methods of second and third order described in [3].
4. Proof of Theorem 1
By means of the polar change of variables x = r cos , y = r sin
system (2) takes the form
r =
3

k=1
f
k
()r
k
,

= f(r cos , r sin ) +

3

i=k
g
k
()r
k1
,
where f
k
() = cos P
k
(cos , sin ) + sin Q
k
(cos , sin ) and g
k
() =
cos Q
k
(cos , sin ) sin P
k
(cos , sin ). Therefore, we have
(7)
dr
d
=
3

k=1
f
k
()r
k
f(r cos , r sin )
+
2
G(r, , ) .
Equation (7) veries the assumptions for applying the averaging method;
i.e., it is a system of the form (3). Therefore, in order to prove Theorem
1, according with Theorem 3, we must study the number of hyperbolic
singular points of the averaged dierential system
dr
d
= f
0
(r),
where
f
0
(r) =
1
2
3

k=1
r
k
_
2
0
f
k
()
f(r cos , r sin )
d.
Therefore, we obtain that f
0
(r) =
3

k=1
C
k
(r)r
k
, where
C
k
(r) =
1
2

i+j=k
(p
ij
I
i+1,j
(r) + q
ij
I
i,j+1
(r)) ,
and
I
p,q
=
_
2
0
cos
p
sin
q

f(r cos , r sin )

d .
The rest of this section is devoted, rst to compute the integral f
0
(r),
and second to study the number of its simple zeros.
The integrals I
p,q
are well studied (see, for instance, [5]). They are
integrals of the form
I =
_
2
0
R(sin , cos ) d,
LIMIT CYCLES VIA THE AVERAGING THEORY 7
where R(x, y) is a rational function which has no poles in the circle
x
2
+y
2
= 1. Taking z = e
i
, when increases from 0 to 2, z describes
the unit circle. So, by the Residue Theorem the integral I is equal to
the product of 2i with the sum of the residues of the function
1
iz
R
_
1
2i
_
z
1
z
_
,
1
2
_
z +
1
z
__
,
in the poles contained in the unit disc.
The conics in R
2
are classied as ellipses, complex ellipses, hyper-
bolas, two complex straight lines intersecting in a real point, two in-
tersecting real straight lines, parabolas, two real parallel straight lines,
two complex parallel straight lines and one double real straight line. In
this paper we study system (2) having a conic f = f(x, y) = 0 given
by one of the following nine cases:
(E) Ellipse f = (x + a)
2
+ (y + b)
2
1 = 0 with a
2
+ b
2
= 1.
(CE) Complex ellipse f = (x + a)
2
+ (y + b)
2
+ 1 = 0.
(H) Hyperbola f = (x + a)
2
y
2
1 = 0 with a
2
= 1.
(CL) Two complex straight lines intersecting in a real point f =
(x + a)
2
+ (y + b)
2
= 0. In this case we allow that f(0, 0) = 0.
(RL) Two real straight lines intersecting in a point (LotkaVolterra
systems) f = (x + a)(y + b) = 0 with ab = 0.
(P) Parabola f = x a y
2
with a = 0.
(RPL) Two real parallel straight lines (x + a)
2
1 = 0 with a
2
= 1.
(CPL) Two complex parallel straight lines (x + a)
2
+ 1 = 0.
(DL) One double invariant real straight line (x +a)
2
= 0 with a = 0.
4.1. Ellipse. For the case of the ellipse (E) system (1) has two period
annuli A
1
= {(x, y) : 0 < x
2
+y
2
< r
2
1
}, and A
2
= {(x, y) : r
2
2
< x
2
+y
2
}
where r
2
1
= min{x
2
+y
2
: (x +a)
2
+ (y +b)
2
= 1}, and r
2
2
= max{x
2
+
y
2
: (x + a)
2
+ (y + b)
2
= 1}. In this case every integral I
p,q
has as
denominator the expression z
k
((a+bi)r+(a
2
+b
2
+r
2
1)z+(abi)rz
2
)
for some k {2, 3, 4}. So the poles of these integrals are z
1
= 0 with
multiplicity k, and
z
2,3
=
1 a
2
b
2
r
2

_
(a
2
+ b
2
+ r
2
1)
2
4r
2
(a
2
+ b
2
)
2(a bi)r
,
both with multiplicity 1.
We must study the zeros of the function f
0
(r) in two intervals, as-
sociated to the two period annuli. That is, in the interval (0, r
1
) with
r
1
=
_
1 + a
2
+ b
2
2

a
2
+ b
2
, and in the interval (r
2
, ) with r
2
=
_
1 + a
2
+ b
2
+ 2

a
2
+ b
2
. We note that the three poles z
i
, i = 1, 2, 3
8 J. GIN

E AND J. LLIBRE
are contained in the interior of the unit disc, independently if r is in
(0, r
1
), or in (r
2
, ).
We compute the function f
0
(r) using the Residue Theorem and we
get that
f
0
(r) =
(a
0
+ a
2
r
2
+ a
4
r
4
+ a
6
r
6
+ a
8
r
8
) + (b
0
+ b
2
r
2
+ b
4
r
4
+ b
6
r
6
)g
r(a
2
+ b
2
)
4
g
,
where g =
_
(a
2
+ b
2
1 2r r
2
)(a
2
+ b
2
1 + 2r r
2
) and the co-
ecients a
i
and b
i
are polynomials in the coecients of P, Q, a and b.
In fact, there are only 6 independent parameters a
i
and b
i
with respect
to the parameters p
ij
and q
ij
.
In order to bound the number of zeros of the numerator of f
0
(r), it
is sucient to bound the number of zeros of h(r) = (a
0
+a
2
r
2
+a
4
r
4
+
a
6
r
6
+ a
8
r
8
)
2
(b
0
+ b
2
r
2
+ b
4
r
4
+ b
6
r
6
)
2
g
2
. Therefore, h is at most of
degree 8 in r
2
, but substituting the values of a
i
and b
i
we obtain
h(r) = 4r
2
(a
2
+ b
2
)
4
(d
0
+ d
2
r
2
+ d
4
r
4
+ + d
10
r
10
),
where d
i
are polynomials in a, b, p
ij
and q
ij
. Hence, we have that, f
0
(r)
at most can have 5 simple zeros in (0, r
1
) (r
2
, ).
The cubic system (2), with a = 2, b = 1 and coecients p
10
= 2,
p
01
= 2, p
20
= 32.429..., p
11
= 4, p
02
= 142.287..., p
30
= 6, p
21
= 7,
p
12
= 8, p
03
= 9, q
10
= 1, q
01
= 2, q
20
= 3, q
11
= 4, q
02
= 5, q
30
=
9.959..., q
21
= 7, q
12
= 15.005... and q
03
= 9, has a function f
0
(r)
with the following 5 positive simple real zeros s
i
= (
_
6 2

5)/(i +1)
for i = 1, 2, 3, 4, 5 in the interval (0,
_
6 2

5). Consequently, this

system realize the upper bound of 5 limit cycles for the ellipse case.
In short, we have proved the following result.
Proposition 5. System (2) with f = (x+a)
2
+(y+b)
2
1 has at most
5 limit cycles up to rst order in bifurcating from the period annuli
surrounding the origin of system (1). Moreover, there exist values of the
coecients p
ij
and q
ij
such that system (2) has 5 hyperbolic limit cycles
bifurcating from the period annuli surrounding the origin of system (1).
4.2. Complex ellipse. For the case of the complex ellipse (CE) sys-
tem (1) has a unique period annulus, namely R
2
\ {(0, 0)}. In this case
every integral I
p,q
has as denominator the expression z
k
((a + bi)r +
(a
2
+b
2
+r
2
+1)z +(a bi)rz
2
) for some k {2, 3, 4}. So the poles of
these integrals are z
1
= 0 with multiplicity k, and
z
2,3
=
(1 + a
2
+ b
2
+ r
2
)
_
(a
2
+ b
2
+ r
2
+ 1)
2
4r
2
(a
2
+ b
2
)
2(a bi)r
,
LIMIT CYCLES VIA THE AVERAGING THEORY 9
both with multiplicity 1. It is easy to check that the three poles are
contained in the interior of the unit disc.
We must study the positive zeros of the function f
0
(r). Using the
Residue Theorem we get that
f
0
(r) =
(a
0
+ a
2
r
2
+ a
4
r
4
+ a
6
r
6
+ a
8
r
8
) + (b
0
+ b
2
r
2
+ b
4
r
4
+ b
6
r
6
)g
r(a
2
+ b
2
)
4
g
,
where g =
_
(1 + a
2
+ b
2
)
2
2(1 + a
2
+ b
2
)r
2
+ r
4
and the coe-
cients a
i
and b
i
are polynomials in the coecients of P, Q, a and
b. In fact, there are only 6 independent parameters between the a
i
and
b
i
with respect to the parameters p
ij
, q
ij
, a and b. In order to bound
the number of zeros of the numerator of f
0
(r), it is sucient to bound
the number of zeros of h(r) = (a
0
+a
2
r
2
+a
4
r
4
+a
6
r
6
+a
8
r
8
)
2
(b
0
+
b
2
r
2
+ b
4
r
4
+ b
6
r
6
)
2
g
2
. Therefore, h is at most of degree 8 in r
2
, but
substituting the values of a
i
and b
i
we obtain
h(r) = 4r
2
(a
2
+ b
2
)
4
(d
0
+ d
2
r
2
+ d
4
r
4
+ + d
10
r
10
),
where d
i
are polynomials in a, b, p
ij
and q
ij
. Hence f
0
(r) can have at
most 5 simple zeros.
The cubic system (2) with a = 2, b = 1 and coecients p
10
= 2,
p
01
= 2, p
20
= 121.469..., p
11
= 4, p
02
= 425.906..., p
30
= 6, p
21
= 7,
p
12
= 8, p
03
= 9, q
10
= 1, q
01
= 2, q
20
= 3, q
11
= 4, q
02
= 5,
q
30
= 27.509..., q
21
= 7, q
12
= 30.896... and q
03
= 9 has a function
f
0
(r) with the following 5 positive simple real zeros s
i
= 1/(i + 1) for
i = 1, 2, 3, 4, 5. Consequently, this system realize the upper bound of 5
limit cycles for the complex ellipse case.
In short, for the complex ellipse case we get a similar proposition to
Proposition 5.
4.3. Hyperbola. For the case of the hyperbola (H) system (1) has
a unique period annulus A = {(x, y) : 0 < x
2
+ y
2
< r
2
1
}, where
r
1
= min{x
2
+y
2
: (x+a)
2
y
2
= 1}. In this case every integral I
p,q
has
as denominator the expression z
k
(r
2
+2arz +2(a
2
1)z
2
+2arz
3
+r
2
z
4
)
for some k {1, 2, 3}. So the poles of these integrals are z
1
= 0 with
multiplicity k, and
z
2,3
=
a

2 a
2
+ 2r
2

_
2 2r
2
+ 2a

2 a
2
+ 2r
2
2r
,
and
z
4,5
=
a +

2 a
2
+ 2r
2

_
2 2r
2
2a

2 a
2
+ 2r
2
2r
,
10 J. GIN

E AND J. LLIBRE
with multiplicity 1. We must study the zeros of the function f
0
(r)
in the interval (0, r
1
). In this case r
2
1
= (a
2
2)/2 if |a| >

2, and
r
1
= min{|1 a|, | 1 a|} if |a|

2. When we want to study the

zeros of f
0
(r) in the rst case we have that z
1
, z
2
and z
4
are the poles
contained in the interior of the unit disc. But when we study the zeros
of f
0
(r) in the second case the poles contained in the interior of the
unit disc are z
1
, z
2
and z
5
. If we compute the function f
0
(r) using the
Residue Theorem we obtain for both cases
f
0
(r) =
1
rg
1
g
2
hk
_
r(a
0
+ a
1
r + a
2
r
2
+ a
3
r
3
+ a
4
r
4
+ a
5
r
5
)g
1
+
(b
0
+ b
1
r + b
2
r
2
+ b
3
r
3
+ b
4
r
4
)hg
1
+ r(c
0
+ c
1
r + c
2
r
2
+
c
3
r
3
+ c
4
r
4
+ c
5
r
5
)g
2
+ (d
0
+ d
1
r + d
2
r
2
+ d
3
r
3
+ d
4
r
4
)hg
2
+
(e
0
+ e
2
r
2
)g
1
g
2
hk
_
,
where g
1
=

1 a
2
+ 3r
2
+ 2h, g
2
=

1 a
2
+ 3r
2
2h with h =
_
r
2
(2 a
2
+ 2r
2
), k =

1 a
2
+ 2ar r
2
and the coecients a
i
and
b
i
are polynomials in the coecients of P, Q and a.
Using this function f
0
(r) we have found that the cubic system (2),
with a = 2 and coecients p
10
= 24.277..., p
01
= 2, p
20
= 39.236...,
p
11
= 4, p
02
= 5, p
30
= 12.662..., p
21
= 7, p
12
= 43.007..., p
03
= 9,
q
10
= 1, q
01
= 24.277..., q
20
= 3, q
11
= 4, q
02
= 5, q
30
= 6, q
21
= 7,
q
12
= 8 and q
03
= 9, has a function f
0
(r) with the following 5 positive
simple real zeros s
i
= 1/(i + 1) for i = 1, 2, 3, 4, 5 in the interval (0, 1).
Consequently, this system has 5 limit cycles. Therefore the hyperbola
case gives at least 5 limit cycles.
In short, we have proved the following proposition.
Proposition 6. System (2) with f = (x + a)
2
y
2
1 has at least 5
limit cycles up to rst order in bifurcating from the period annulus
surrounding the origin of system (1).
4.4. Two complex straight lines intersecting in a real point.
For the case of two complex straight lines intersecting in a real point
(CL) system (1) has two period annuli A = {(x, y) : 0 < x
2
+ y
2
<
a
2
+ b
2
}, and A
2
= {(x, y) : a
2
+ b
2
< x
2
+ y
2
} if a
2
+ b
2
= 0, and a
unique period annulus R
2
\ {(0, 0)} if a = b = 0.
Firstly, we consider the case a
2
+ b
2
= 0. In this case every integral
I
p,q
has as denominator the expression z
k
(r + az bzi)(a + rz + bi)
for some k {2, 3, 4}. So the poles of these integrals are z
1
= 0 with
multiplicity k, z
2
= (a + bi)/r, and z
3
= (a + bi)r/(a
2
+ b
2
) both
with multiplicity 1.
LIMIT CYCLES VIA THE AVERAGING THEORY 11
We must study the zeros of the function f
0
(r) in (0,

a
2
+ b
2
)
(

a
2
+ b
2
, ). Since the modulus of the poles are |z
1
| = 0, |z
2
| =

a
2
+ b
2
/r and |z
3
| = r/

a
2
+ b
2
, when we want to study the zeros of
f
0
(r) in (0,

a
2
+ b
2
) we have that z
1
and z
3
are the two poles contained
in the interior of the unit disc. But when we study the zeros of f
0
(r)
in (

a
2
+ b
2
, ) the poles contained in the interior of the unit disc are
z
1
and z
2
. Using the Residue Theorem we get that
f
0
(r) =
_

_
r(a
0
+ a
2
r
2
+ a
4
r
4
+ a
6
r
6
)
8(a
4
+ b
4
)(a
2
+ b
2
r
2
)
if r (0,

a
2
+ b
2
),
b
0
+ b
2
r
2
+ b
4
r
4
8r(a
2
+ b
2
r
2
)
if r (

a
2
+ b
2
, ),
the coecients a
i
and b
i
are polynomials in the coecients of P, Q,
a and b. Hence, f
0
(r) at most can have 5 simple zeros in (0, ) \
{

a
2
+ b
2
}.
The cubic system (2), with a = b = 1 and coecients p
10
= 1172/518,
p
01
= 18929/581, p
20
= 3155/581, p
11
= 32518/581, p
02
= 5, p
30
=
8034/581, p
21
= 7, p
12
= 8, p
03
= 9, q
10
= 1, q
01
= 2, q
20
= 3,
q
11
= 4, q
02
= 5, q
30
= 6, q
21
= 7, q
12
= 8 and q
03
= 9, has a function
f
0
(r) with the following 5 positive simple real zeros: two of them are
s
i
= (i + 1)

2 for i = 1, 2 in the interval (

2, ) and the others are

s
i
=

2/(i 1) for i = 3, 4, 5 in the interval (0,

2). Consequently,
this system realize the upper bound of 5 limit cycles for the case of two
complex straight lines intersecting in a real point.
In the case a
2
+b
2
= 0, the unique pole of the integrals I
p,q
is z
1
= 0
with multiplicity k for some k {3, 4, 5}. If we compute the function
f
0
(r) using the Residue Theorem we obtain
f
0
(r) =
1
r
(4p
10
+ 4q
01
+ (3p
30
+ p
12
+ q
21
+ 3q
03
)r
2
).
Hence, f
0
(r) at most can have 1 simple positive zero. The cubic system
(2), with a = b = 0 and coecients p
10
= 23/4, p
01
= 2, p
20
= 3,
p
11
= 4, p
02
= 5, p
30
= 6, p
21
= 7, p
12
= 8, p
03
= 9, q
10
= 1,
q
01
= 2, q
20
= 3, q
11
= 4, q
02
= 5, q
30
= 6, q
21
= 7, q
12
= 8 and
q
03
= 9, has a function f
0
(r) with the positive simple real zero s
1
= 1/2.
Consequently, this system realize the upper bound of 1 limit cycles for
the case of two complex straight lines intersecting in a real point with
a
2
+ b
2
= 0.
In short, for the case of two complex straight lines intersecting in a
real we get a similar proposition to Proposition 5.
12 J. GIN

E AND J. LLIBRE
4.5. Two real straight lines intersecting in a real point. For the
case of two real straight lines intersecting in a real point (RL) system
(1) has a unique period annulus A = {(x, y) : 0 < x
2
+ y
2
< r
2
1
} where
r
1
= min{|a|, |b|}. In this case every integral I
p,q
has as denominator
the expression z
k
(r +2az +rz
2
)(r +2bzi +rz
2
) for some k {1, 2, 3}.
So the poles of these integrals are z
1
= 0 with multiplicity k, z
2
=
(a

a
2
r
2
)/r, z
3
= (a +

a
2
r
2
)/r, z
4
= (

r
2
b
2
bi)/r
and z
5
= (

r
2
b
2
bi)/r with multiplicity 1.
Since the modulus of the poles are |z
1
| = 0, and |z
2
|, |z
3
|, |z
4
| and
|z
5
| depend on the values of a, b and r and we want to study the zeros
of f
0
(r) in (0, r
1
), we have that z
1
, z
3
z
5
are the three poles contained
in the interior of the unit disc.
Using the Residue Theorem we get that
f
0
(r) =
1
gh(a
2
+ b
2
r
2
)
_
(a
0
+ a
2
r
2
+ a
4
r
4
)g
+(b
0
+ b
2
r
2
+ b
4
r
4
)h + (a
2
+ b
2
r
2
)(c
0
+ c
2
r
2
)gh
_
,
where g =

a
2
r
2
and h =

b
2
r
2
and the coecients a
i
, b
i
and c
i
are polynomials in the coecients of P, Q, a and b. In fact, there are
only 6 independent parameters between the a
i
, b
i
and c
i
with respect
to p
ij
, q
ij
, a and b. In order to bound the number of zeros of the
numerator of f
0
(r), it is sucient to bound the number of zeros of
g
2
[(a
0
+a
2
r
2
+a
4
r
4
)+(a
2
+b
2
r
2
)(c
0
+c
2
r
2
)h]
2
(b
0
+b
2
r
2
+b
4
r
4
)
2
h
2
= 0
or equivalent of
(a
2
r
2
)(a
0
+ a
2
r
2
+ a
4
r
4
)
2
+[(a
2
r
2
)(a
2
+ b
2
r
2
)
2
(c
0
+ c
2
r
2
)
2
(b
0
+ b
2
r
2
+ b
4
r
4
)
2
](b
2
r
2
)
+ 2(a
2
r
2
)(a
0
+ a
2
r
2
+ a
4
r
4
)(a
2
+ b
2
r
2
)(c
0
+ c
2
r
2
)h = 0.
Finally, to bound the number of zeros of the previous expression we
must bound the zeros of the polynomial.
k = 4(a
2
r
2
)
2
(a
0
+ a
2
r
2
+ a
4
r
4
)
2
(a
2
+ b
2
r
2
)
2
(c
0
+ c
2
r
2
)
2
(b
2
r
2
)
{(a
2
r
2
)[(a
0
+ a
2
r
2
+ a
4
r
4
)
2
+ (a
2
+ b
2
r
2
)
2
(c
0
+ c
2
r
2
)
2
(b
2
r
2
)
(b
0
+ b
2
r
2
+ b
4
r
4
)
2
(b
2
r
2
)}
2
.
Therefore, k is at most of degree 12 in r
2
, but substituting the values
of a
i
, b
i
and c
i
we obtain
k = r
2
(a
2
+ b
2
r
2
)
2
(d
0
+ d
2
r
2
+ d
4
r
4
+ + d
18
r
18
),
where the d
i
s are polynomials in a
i
, b
i
and c
i
. Hence, we have that,
f
0
(r) at most can have 9 simple zeros in (0, r
1
). In [4] it is found a better
upper bound for the number of limit cycles of system (2) bifurcating
LIMIT CYCLES VIA THE AVERAGING THEORY 13
from the period annulus surrounding the origin of system (1) in this
case and this number is 7.
The cubic system (2), with a = 2, b = 1 and coecients p
10
=
3.998..., p
01
= 13.935..., p
20
= 14.023..., p
11
= 4, p
02
= 5, p
30
= 6,
p
21
= 3.934..., p
12
= 8, p
03
= 10.105..., q
10
= 1, q
01
= 3.998...,
q
20
= 3, q
11
= 4, q
02
= 5, q
30
= 6, q
21
= 7, q
12
= 8 and q
03
= 9,
has a function f
0
(r) with the following 6 positive simple real zeros
s
i
= 1/(i + 1) for i = 1, 2, 3, 4, 5, 6 in the interval (0, 1). Consequently,
this system has 6 limit cycles. Therefore the case of two real straight
lines intersecting in a real point gives at least 6 hyperbolic limit cycles
and has an upper bound of 7 limit cycles.
In short, we have proved the following result.
Proposition 7. System (2) with f = (x + a)(y + b) has at most 7
limit cycles up to rst order in bifurcating from the period annuli
surrounding the origin of system (1). Moreover, there exist values of the
coecients p
ij
and q
ij
such that system (2) has 6 hyperbolic limit cycles
bifurcating from the period annuli surrounding the origin of system (1).
4.6. Parabola. For the case of the parabola (P) system (1) has a
unique period annulus A = {(x, y) : 0 < x
2
+ y
2
< r
2
1
}, where r
2
1
=
min{x
2
+ y
2
: x a y
2
= 0}. In this case every integral I
p,q
has as
denominator the expression z
k
(r
2
+2(2ai)rz +2(2a
2
+r
2
)z
2
+2(2a+
i)rz
3
+ r
2
z
4
) for some k {1, 2, 3}. So the poles of these integrals are
z
1
= 0 with multiplicity k, and
z
2,3
=
1 +

1 + 4a + 4r
2

_
2 + 4a 2

1 + 4a + 4r
2
2r
,
and
z
4,5
=
1

1 + 4a + 4r
2

_
2 + 4a + 2

1 + 4a + 4r
2
2r
,
with multiplicity 1. We must study the zeros of the function f
0
(r) in
the interval (0, r
1
). In this case r
1
= |a| if a > 1/2 and dierent
from zero, and r
1
=

1 4a/2 if a 1/2. When we want to study

the zeros of f
0
(r) in (0, |a|) we have that z
1
, z
3
and z
4
are the poles
contained in the interior of the unit disc. But when we study the zeros
of f
0
(r) in (0,

1 4a/2) the poles contained in the interior of the

unit disc are z
1
, z
2
and z
4
. If we compute the function f
0
(r) using the
14 J. GIN

E AND J. LLIBRE
Residue Theorem we obtain for both cases
f
0
(r) =
1
rg
1
g
2
hk
_
r(a
0
+ a
1
r + a
2
r
2
+ a
3
r
3
+ a
4
r
4
+ a
5
r
5
)g
1
+
(b
0
+ b
1
r + b
2
r
2
+ b
3
r
3
+ b
4
r
4
)hg
1
+
r(c
0
+ c
1
r + c
2
r
2
+ c
3
r
3
+ c
4
r
4
+ c
5
r
5
)g
2
+
(d
0
+ d
1
r + d
2
r
2
+ d
3
r
3
+ d
4
r
4
)hg
2
+
(e
0
+ e
2
r
2
)g
1
g
2
hk
_
,
where g
1
=

a + 2r
2
rh, g
2
=

a + 2r
2
+ rh, h =

1 + 4a + 4r
2
,
k =

a + r and the coecients a

i
, b
i
, c
i
, d
i
and e
i
are polynomials in
the coecients of P, Q and a.
The cubic system (2), with a = 1 and coecients p
10
= 2, p
01
= 2,
p
20
= 3.066..., p
11
= 4, p
02
= 26.755..., p
30
= 5.601..., p
21
= 7,
p
12
= 10.875..., p
03
= 9, q
10
= 1, q
01
= 2, q
20
= 3, q
11
= 4, q
02
= 5,
q
30
= 6, q
21
= 7, q
12
= 8 and q
03
= 9, has a function f
0
(r) with the
following 5 positive simple real zeros s
i
= 1/(i + 1) for i = 1, 2, 3, 4, 5
in the interval (0, 1). Consequently, this system has 5 limit cycles.
Therefore the parabola case provides at least 5 hyperbolic limit cycles.
In short, for the parabola case we get a similar proposition to Propo-
sition 6.
4.7. Two real parallel straight lines. For the case of two real par-
allel straight lines (RPL) system (1) has a unique period annulus A =
{(x, y) : 0 < x
2
+y
2
< r
2
1
}, where r
1
= min{|a1|, |a+1|}. In this case
every integral I
p,q
has as denominator the expression z
k
(r 2z +2az +
rz
2
)(r + 2z + 2az + rz
2
) for some k {1, 2, 3}. So the poles of these
integrals are z
1
= 0 with multiplicity k, z
2
= (1a
_
(a 1)
2
r
2
)/r,
z
3
= (1 a +
_
(a 1)
2
r
2
)/r, z
4
= (1 a
_
(a + 1)
2
r
2
)/r and
z
5
= (1 a +
_
(a + 1)
2
r
2
)/r with multiplicity 1.
Since the modulus of the poles are |z
1
| = 0, and |z
2
|, |z
3
|, |z
4
| and
|z
5
| depend on the values of a and r and we want to study the zeros of
f
0
(r) in (0, r
1
), we have that z
1
, z
3
z
5
are the three poles contained in
the interior of the unit disc. Using the Residue Theorem we get that
f
0
(r) =
(a
0
+ a
2
r
2
+ a
4
r
4
)g + (b
0
+ b
2
r
2
+ b
4
r
4
)h + (c
0
+ c
2
r
2
)gh
gh
,
where g =
_
(a 1)
2
r
2
and h =
_
(a + 1)
2
r
2
and the coecients
a
i
, b
i
and c
i
are polynomials in the coecients of P, Q and a. In fact,
there are only 6 independent coecients between the a
i
, b
i
and c
i
with
respect to the coecients p
ij
, q
ij
and a. In order to bound the number
LIMIT CYCLES VIA THE AVERAGING THEORY 15
of zeros of the numerator of f
0
(r), it is sucient to bound the number
of zeros of
g
2
[(a
0
+ a
2
r
2
+ a
4
r
4
) + (c
0
+ c
2
r
2
)h]
2
(b
0
+ b
2
r
2
+ b
4
r
4
)
2
h
2
= 0,
or equivalent of
((a 1)
2
r
2
)(a
0
+ a
2
r
2
+ a
4
r
4
)
2
+((a + 1)
2
r
2
)[((a 1)
2
r
2
)(c
0
+ c
2
r
2
)
2
(b
0
+ b
2
r
2
+ b
4
r
4
)
2
]
+ 2((a 1)
2
r
2
)(a
0
+ a
2
r
2
+ a
4
r
4
)(c
0
+ c
2
r
2
)h = 0 .
Finally, to bound the number of zeros of the previous expression we
must bound the zeros of the polynomial.
k = 4((a 1)
2
r
2
)
2
(a
0
+ a
2
r
2
+ a
4
r
4
)
2
(c
0
+ c
2
r
2
)
2
((a + 1)
2
r
2
)
{((a 1)
2
r
2
)[(a
0
+ a
2
r
2
+ a
4
r
4
)
2
+ (c
0
+ c
2
r
2
)
2
((a + 1)
2
r
2
)]
(b
0
+ b
2
r
2
+ b
4
r
4
)
2
((a + 1)
2
r
2
)}
2
.
Therefore, k is at most of degree 20 in r
2
, but substituting the values
of a
i
, b
i
and c
i
we obtain
k = r
2
(d
0
+ d
2
r
2
+ d
4
r
4
+ + d
16
r
16
),
where the d
i
s are polynomials in a
i
, b
i
and c
i
. Hence f
0
(r) has at most
8 simple zeros in (0, r
1
).
The cubic system (2), with a = 2 and coecients p
10
= 29.106...,
p
01
= 2, p
20
= 33.346..., p
11
= 4, p
02
= 35.758..., p
30
= 6, p
21
= 7,
p
12
= 18.791..., p
03
= 9, q
10
= 1, q
01
= 2, q
20
= 3, q
11
= 4, q
02
= 5,
q
30
= 6, q
21
= 7, q
12
= 8 and q
03
= 9, has a function f
0
(r) with the
following 5 positive simple real zeros s
i
= 1/(i + 1) for i = 1, 2, 3, 4, 5
in the interval (0, 1). Consequently, this system has 5 hyperbolic limit
cycles. Therefore the case of two real parallel straight lines provides at
least 5 limit cycles and has an upper bound of 8 limit cycles.
In short, we have proved the following result.
Proposition 8. System (2) with f = (x + a)
2
1 has at most 8
limit cycles up to rst order in bifurcating from the period annuli
surrounding the origin of system (1). Moreover, there exist values of the
coecients p
ij
and q
ij
such that system (2) has 5 hyperbolic limit cycles
bifurcating from the period annuli surrounding the origin of system (1).
4.8. Two complex parallel straight lines. For the case of two
complex parallel straight lines (CPL) system (1) has a unique pe-
riod annulus, namely R
2
\ {(0, 0)}. In this case every integral I
p,q
has as denominator the expression z
k
(r
2
+ 4arz + 2(2 + a
2
+ r
2
)z
2
+
4arz
3
+ r
2
z
4
) for some k {1, 2, 3}. So the poles of these integrals
are z
1
= 0 with multiplicity k, z
2
= (a + i
_
(a i)
2
r
2
)/r,
16 J. GIN

E AND J. LLIBRE
z
3
= (a+i +
_
(a i)
2
r
2
)/r, z
4
= (ai
_
(a + i)
2
r
2
)/r and
z
5
= (a i +
_
(a + i)
2
r
2
)/r with multiplicity 1.
We must study the zeros of the function f
0
(r). Since the modulus of
the poles are |z
1
| = 0, and |z
2
|, |z
3
|, |z
4
| and |z
5
| depend on the values
of a and r, we have that z
1
, z
3
and z
5
are the three poles contained in
the interior of the unit disc. Using the Residue Theorem we get that
f
0
(r) =
(a
0
+ a
2
r
2
+ a
4
r
4
)g + (b
0
+ b
2
r
2
+ b
4
r
4
)h + (c
0
+ c
2
r
2
)gh
gh
,
where g =
_
(a i)
2
r
2
and h =
_
(a + i)
2
r
2
and the coecients
a
i
, b
i
and c
i
are polynomials in the coecients of P, Q and a. In fact,
there are only 6 independent coecients between the a
i
, b
i
and c
i
with
respect to the coecients p
ij
, q
ij
and a. In order to bound the number
of zeros of the numerator of f
0
(r), it is sucient to bound the number
of zeros of
g
2
[(a
0
+ a
2
r
2
+ a
4
r
4
) + (c
0
+ c
2
r
2
)h]
2
(b
0
+ b
2
r
2
+ b
4
r
4
)
2
h
2
= 0
or equivalent of
((a i)
2
r
2
)(a
0
+ a
2
r
2
+ a
4
r
4
)
2
+((a + i)
2
r
2
)[((a i)
2
r
2
)(c
0
+ c
2
r
2
)
2
(b
0
+ b
2
r
2
+ b
4
r
4
)
2
]
+ 2((a i)
2
r
2
)(a
0
+ a
2
r
2
+ a
4
r
4
)(c
0
+ c
2
r
2
)h = 0 .
Finally, to bound the number of zeros of the previous expression we
must bound the zeros of the polynomial.
k = 4((a i)
2
r
2
)
2
(a
0
+ a
2
r
2
+ a
4
r
4
)
2
(c
0
+ c
2
r
2
)
2
((a + i)
2
r
2
)
{((a i)
2
r
2
)[(a
0
+ a
2
r
2
+ a
4
r
4
)
2
+ (c
0
+ c
2
r
2
)
2
((a + i)
2
r
2
)]
(b
0
+ b
2
r
2
+ b
4
r
4
)
2
((a + i)
2
r
2
)}
2
.
Therefore, k is at most of degree 20 in r
2
, but substituting the values
of a
i
, b
i
and c
i
we obtain
k = r
2
(d
0
+ d
2
r
2
+ d
4
r
4
+ + d
16
r
16
),
where the d
i
s are polynomials in a
i
, b
i
and c
i
. Hence f
0
(r) at most
has 8 simple positive zeros.
The cubic system (2), with a = 1 and coecients p
10
= 3338.616...,
p
01
= 2, p
20
= 4924.060..., p
11
= 4, p
02
= 395.544..., p
30
= 1983.121...,
p
21
= 7, p
12
= 1745.939..., p
03
= 9, q
10
= 1, q
01
= 2, q
20
= 3, q
11
= 4,
q
02
= 5, q
30
= 6, q
21
= 7, q
12
= 8 and q
03
= 9, has a function f
0
(r)
with the following 5 positive simple real zeros s
i
= i for i = 1, 2, 3, 4, 5.
Consequently, this system has 5 hyperbolic limit cycles. Therefore the
case of two complex parallel straight lines provides at least 5 limit
cycles and has an upper bound of 8 limit cycles.
LIMIT CYCLES VIA THE AVERAGING THEORY 17
In short, for the case of two complex parallel straight lines we get a
similar proposition to Proposition 8.
4.9. One double invariant real straight line. For the case of one
double invariant real straight line (DL) system (1) has a unique period
annulus A = {(x, y) : 0 < x
2
+ y
2
< a
2
}. In this case every integral
I
p,q
has as denominator the expression z
k
(r + 2az + rz
2
)
2
for some
k {1, 2, 3}. So the poles of these integrals are z
1
= 0 with multiplicity
k, z
2
= (a

a
2
r
2
)/r and z
3
= (a+

a
2
r
2
)/r with multiplicity
2.
Since the modulus of the poles are |z
1
| = 0, and |z
2
| and |z
3
| depend
on the values of a and r and we want to study the zeros of f
0
(r) in
(0, |a|), we have that z
1
and z
3
are the two poles contained in the
interior of the unit disc. Using the Residue Theorem we get that
f
0
(r) =
(a
2
r
2
)(a
0
+ a
2
r
2
)f + (b
0
+ b
2
r
2
+ b
4
r
4
)
2r(a
2
r
2
)f
,
where f =

a
2
r
2
and the coecients a
i
and b
i
are polynomials in
the coecients of P, Q and a. In fact, there are only 4 independent
coecients between the a
i
, b
i
and c
i
with respect to the coecients p
ij
,
q
ij
and a. From the function g = (a
2
r
2
)
2
(a
0
+a
2
r
2
)
2
f
2
(b
0
+b
2
r
2
+
b
4
r
4
)
2
we have that, f
0
(r) at most can have 4 simple zeros in (0, |a|).
The cubic system (2), with a = 2 and coecients p
10
= 2.003...,
p
01
= 2, p
20
= 9.881..., p
11
= 4, p
02
= 23.250..., p
30
= 6, p
21
= 7,
p
12
= 8, p
03
= 9, q
10
= 1, q
01
= 2, q
20
= 3, q
11
= 4, q
02
= 5, q
30
= 6,
q
21
= 7, q
12
= 8 and q
03
= 9, has a function f
0
(r) with the following
3 simple real zeros s
i
= 2/(i + 1) for i = 1, 2, 3 in the interval (0, 2).
Consequently, this system has 3 hyperbolic limit cycles. Therefore the
case of one double invariant real straight line gives at least 3 limit cycles
and has an upper bound of 4 limit cycles.
Proposition 9. System (2) with f = (x+a)
2
has at most 4 limit cycles
up to rst order in bifurcating from the period annuli surrounding the
origin of system (1). Moreover, there exist values of the coecients p
ij
and q
ij
such that system (2) has 3 hyperbolic limit cycles bifurcating
from the period annuli surrounding the origin of system (1).
Remark. We think that one of the main contributions of the present
paper is to show that 6 of the other cases provide less number of limit
cycles than case 4.5, and that the remaining two cases (hyperbola and
parabola) can provide at least 5, but we do not know for these cases
an upper bound.
18 J. GIN

E AND J. LLIBRE
References
[1] V.I. Arnold, V. I. and Y.S. Ilyashenko, Dynamical systems I, Ordinary
dierential equations, Encyclopaedia Math. Sci., Vol. 1, Springer, Berlin, 1988.
[2] T.R. Blows and L.M. Perko, Bifurcation of limit cycles from centers and
separatrix cycles of planar analytic systems, SIAM Rev. 36 (1994), 341376.
[3] A. Buic a and J. Llibre, Averaging methods for nding periodic orbits via
Brouwer degree, Bull. Sci. Math. 128 (2004), 722.
[4] A. Buic a and J. Llibre, On the number of limit cycles of a perturbed cubic
polynomial dierential center, to appear in Chaos Solitons Fractals.
[5] H. Cartan, Theorie elementaire des fonctions analytiques dune ou plusieurs
variables complexes, Hermann, Paris, 1961.
[6] C. Chicone and M. Jacobs, Bifurcation of limit cycles from quadratic
isochrones, J. Dierential Equations 91 (1991), 268326
[7] H. Giacomini, J. Llibre and M. Viano, On the nonexistence, existence
and uniqueness of limit cyles, Nonlinearity 9 (1996), 501516.
[8] H. Giacomini, J. Llibre and M. Viano, On the shape of limit cycles that
bifurcate from Hamiltonian centers, Nonlinear Anal. 41 (2000), Ser. A: Theory
Methods, 523537.
[9] H. Giacomini, J. Llibre and M. Viano, On the shape of limit cycles that
bifurcate from non-Hamiltonian centers, Nonlinear Anal. 43 (2001), Ser. A:
Theory Methods, 837859.
[10] J. Guckenheimer and P. Holmes, Nonlinear oscillations, dynamical sys-
tems, and bifurcations of vector elds. Applied Mathematical Sciences, 42.
Springer-Verlag, New York, 1986.
[11] M. Han, The number and distributions of limit cycles in a cubic system, Chi-
nese Ann. Math. Ser. A 23 (2002), 143152.
[12] Li, Jibin and Huang, Qi Ming, Bifurcations of limit cycles forming com-
pound eyes in the cubic system, Chinese Ann. Math. Ser. B 8 (1987), 391403.
[13] C. Li, W. Li, J. Llibre and Z. Zhang, Linear estimation of the number
of zeros of abelian integrals for some cubic isochronous centers, J. Dierential
Equations 180 (2002), 307333.
[14] J. Llibre, Averaging theory and limit cycles for quadratic systems, Rad. Mat.
11 (2002/3), 215228.
[15] J. Llibre, J.S. P erez del Ro, J. A. Rodrguez, Averaging analysis of
a perturbated quadratic center, Nonlinear Anal. Ser. A: Theory Methods 46
(2001), 4551.
[16] Z. Meng, M. Han and S. Gu, Global bifurcations of limit cycles for a cubic
system, Ann. Dierential Equations 16 (2000), 263269.
[17] L.S. Pontrjagin,

Uber Autoschwingungssysteme, die den hamiltonshen nahe
liegen, Physikalische Zeitschrift der Sowjetunion, Band 6, Heft(12) (1934),
2528.
[18] J.A. Sanders and F. Verhulst, Averaging methods in nonlinear dynamical
systems, Applied Mathematical Sci., Vol. 59, Springer-Verlag, New York, 1985.
[19] Y. Song, The Poincare bifurcation of cubic Hamiltonian systems with a global
center, Acta Math. Sinica 47 (2004), 291298.
[20] F. Verhulst, Nonlinear dierential equations and dynamical systems, Uni-
versitext. Springer-Verlag, Berlin, 1996.
LIMIT CYCLES VIA THE AVERAGING THEORY 19
[21] M. Viano, J. Llibre and H. Giacomini, Arbitrary order bifurcations for
perturbed Hamiltonian planar systems via the reciprocal of an integrating fac-
tor, Nonlinear Anal. 48 (2002), Ser. A: Theory Methods, 117136.
[22] Y. Zhao and Z. Zhang, Abelian integrals for cubic vector elds, Ann. Mat.
Pura Appl. (4) 176 (1999), 251272.
[23] Y. Zhao and Z. Zhang, Bifurcations of limit cycles from cubic Hamiltonian
systems with a center and a homoclinic saddleloop, Publ. Mat. 44 (2000),
205235.
[24] H.

Zoladek, Eleven small limit cycles in a cubic vector eld, Nonlinearity 8
(1995), 843860.
[25] P. Yu and M. Han, Small limit cycles bifurcating from ne focus points in
cubic order Z
2
equivariant vector elds, Chaos Solitons Fractals 24 (2005),
329348.
1
Departament de Matem` atica, Universitat de Lleida, Av. Jaume II,
69. 25001 Lleida, Spain
E-mail address: gine@eps.udl.es
2
Departament de Matem` atiques, Universitat Aut` onoma de Barcelona,
08193 Bellaterra, Barcelona, Spain
E-mail address: jllibre@mat.uab.es