Appendix C: XRF Theory

Appendix C:
XRF Theory
An introduction
to XRF
Analysis
X-rays, which form a part of the electromagnetic spectrum as shown in
Figure C-3 are generated when atoms of elements are struck either by
particles, such as electrons, or other higher energy radiation. The
process of primary X-ray emission which takes place inside an X-ray
tube can be considered simply as the release of characteristic X-ray
wavelengths of the tube anode element when it is irradiated by a beam
of high energy electrons.
s the tomic !umber of the element increases, the wavelength of the
most sensitive characteristic emission spectral line, known as the " line,
decreases. #ow atomic number elements are often called $light
elements% and their radiations are absorbed by air. Conversely, high
atomic numbers are called heavy elements.
Page C.&
Figure C-3: Part of the electromagnetic radiation spectrum
Although the X-ray region could be defned by frequency in MHZ, it is more convenient to use
avelengths specifed in either angstrom units !"# or A$%"&
-'
cm( or the )* unit nanometers
!"nm % "&
- +
m( or energy ,e-.
Appendix C: XRF Theory
Characteristic X-rays
are often also
e'pressed in terms of
energy in ke(
)inversely
proportional to
wavelength*. This
means that the
energy of the X-ray
emission lines in ke(
increase with
increasing tomic
!umber. The
relationship between
energy and
wavelength is as
follows+
E keV
wavelength
( )
.
(
=
12 4
)
,t is necessary to understand that each element has a number of
characteristic lines and that these occur as a result of internal electronic
transitions within the atom. These lines occur as the result of the e'cess
energy liberated when electrons from higher energy levels $fall down% to
take the place of ", # and - electrons which have been e.ected by the
e'citing radiation.
These discrete energy di/erences give rise to characteristic element
lines and are shown in Figure C-0. 1ere the classical 2iegbahn
description of the line is used, i.e.", ", etc.
3hen X-rays are used as a source of radiation to e'cite other elements
the process is known as X-Ray Fluorescence.
Selective
excitation
and priary
!ea
"ltration in
#$XRF
The energy dispersive spectrometer has a single detector which sees X-
rays of all energies simultaneously. This advantage has to be used
carefully to ensure a ma'imum countrate of typically, 34,444 counts per
second, which the detector will accept.
,n the 56X7F spectrometer, therefore, we must lower the percentage of
the spectrum taken up by background counts and this is done by
selective e'citation and primary beam 8ltration. ,n practice, this means
that the e'citation voltage used for a particular range of elements is
only that which e/ectively e'cites the highest energy line and no more.
This avoids producing large amounts of continuum background above
the highest element to be analysed under these conditions. ,n addition,
the use of a 8ter may be placed between the X-ray tube and the sample
to modify the position of the peak energy available for e'citation.
,n this way, the X-ray beam reaching the sample has a spectral
distribution which favours good e'citation and low backgrounds for a
particular region of the spectrum. To accomplish this, the #ab-X 3944
energy dispersive spectrometer is 8tted with programmable multiple
8lter changers which provide automatic changing between di/erent
ranges of elements in multi-element applications. The choice of 8lter is
matched with an appropriate
Page C.:
Appendix C: XRF Theory
Incident X-ray
Characteristic
X-ray
e-
K
L
M
e-
ejected
electron
16
s
Fluorescence of sulphur atom
Figure C-%: /luorescence of sulphur atom
combination of tube voltage and beam current. 2uch analytical
procedures may be stored for automatic repeat use, where the stored
sets )methods* of analytical conditions are re-used.
Page C.3
Appendix C: XRF Theory
A!sorption o&
X-rays
"nowledge of the nature of absorption of X-rays is fundamental to the
understanding of nearly everything that happens in X-ray analysis. Two
basic facts need to be grasped as follows+
)i* X-rays are absorbed e'ponentially according to ;eer%s law.
I I e
x o
x
=

3here
,
'
< the intensity passing through a thickness '
*
o
< the initial incident intensity
= < the mass absorbtion coe>cient
< the density
' < the thickness of the absorber
)ii* The value of the mass absorption coe>cient changes with
wavelength and a plot of the variation between absorption and
wavelength shows ma'ima followed by a sharp drop - the
absorption edge. Figure C-9 shows this e/ect.
bsorption edges
occur when
energy is close to
the critical value
needed to e'cite a
particular spectral
line of an element
and is absorbed
very strongly. n
abrupt fall in
absorption takes
place when the
energy is .ust
insu>cient to e.ect the appropriate electron.
Page C.0
Figure C-'
Appendix C: XRF Theory
From this it follows
that an element is
e/ectively
transparent to its own
radiation and that
elements a few
atomic numbers lower
will be strong
absorbers. For
e'ample, l will
absorb 2i" radiation
as strong as Fe, even
though Fe is much
more dense and has a
much higher atomic
number.
5lements of higher
atomic number are
normally called heavy
elements and
because they have
many e'tra nuclear
electrons they have a
series of absorption
edges ", #, -, etc.
The higher the atomic
number the higher
the energy )shorter
the wavelength* of
the " spectral lines.
Conversely, elements
of low atomic number
are called light
elements and have "
spectral lines of low
energy )or long
wavelength*.
From a study of a set of mass absorption tables it will be seen that the
magnitude of the mass absorption coe>cient at the energy being
studied plays a principle role. This is the whole basis of inter-element
)matri'* correction procedures, which are needed in the analysis of
multi-element materials. The make-up of the material or the matri'
decides how much of the e'cited characteristic radiation escapes from
the sample which is measured and related by calibration to the
concentration of the element.
Single Filter
(ethod
single 8lter may be employed where the line of the element to be
measured )analyte* is lower in energy than the element)s* to be
discriminated against and where the absorption edge of the 8lter lies to
the high energy side of the analyte line. This is shown diagrammatically
in Figure C-?.
The use of a single 8lter mounted in front of the detector, known as a
secondary 8lter, plays a very important role. Filters may be
programmed to selectively absorb elements of higher energy )normally
higher tomic !umber*.
These 8lters work particularly well in the region from l to Ti, where it is
now possible to separate l from 2i, @ from 2, 2 from Cl and so on.
2pecial
Page C.9
Figure C-)
Appendix C: XRF Theory
techniAues have been
used for these
elements to fabricate
8lters of high
e>ciency.
nother powerful tool in limiting interference from higher energy
elements is the use of speci8c programmed kilo voltages on the X-ray
tube. For e'ample, it is possible to nearly eliminate the e'citation of Ca,
while still retaining a useful signal from an element such as 2 in, for
e'ample, a 8nished cement sample.
*eneral
Principles o&
Cali!ration
X-ray analysis, as with other forms of spectroscopy, is a comparative
techniAue. This means it is necessary to have a series of reference
standards which cover the elements to be measured in the ranges of
interest. These are measured and the intensities for the elements to be
analysed are recorded so that calibrations of concentration versus
intensity can be derived and stored in the data processing computer for
future analysis of unknown samples of the same type.
;efore the standard samples can be measured it is necessary to decide
the sample preparation methods that will be employed )see ppendi'
;*.
,t must always be remembered that even though X7F analysis is a
volume analysis, the depth of sample involved in the analysis depends
greatly on the matri' material )i.e. absorption* and the energy of the X-
rays being measured. For e'ample, @b being measured as the @b#
spectral line in a steel comes from only some &4 microns deep, whereas
Page C.?
Figure C-+
Appendix C: XRF Theory
@bl measured in a
gasoline comes from
more than &4mm
depth in the sample.
2olid samples
therefore need a Bat
surface for analysis
and because the
surface of dense
solids represent the
mechanical reference
with the
instrument, it is important that these are Bat to &4-:4 microns, if the
ultimate of accuracy is to be achieved. This point will be touched on
again in the sections describing precision and accuracy.
The simplest form of calibration is a straight line as shown in Figure C-C
and this basic relationship between concentration and intensity is the
sensitivity of the instrument. The eAuation of the straight line is given
by+
C a a I
o m
= +
1
*
such that the concentration for an $unknown% sample can be calculated
by measuring the intensity from the instrument and using the a
0
and a
1

coe>cients. This eAuation is derived from the following+
y mx c = +
From Figure C-C, y is the intensity measured ),
m
*, m is the sensitivity )D*
and c is ,
4
. The eAuation becomes+
I Q C I
m o
= + ( )
7earrangement of this gives+
C
I
Q
I
Q
m o
=
The magnitude of the term a
0
, which is normally negative, is called the
;ackground 5Auivalent Concentration commonly known as the ;5C. ,t is
possible to use an intensity intercept I
0
, but this is less common.
The inverse of the slope term &Ea
&
is eAual to the number of
countsEsecondEunit of concentration used. This is referred to as 0 in our
documentation and will be used in formulae to e'press the precision of
the instruments.
s it is reAuired to use a calibration over a very long time, it is
necessary to compensate for instrument drift which, even though small,
will occur. This is done by the use of high and low reference points for
each element. 2pecial samples known as 2F2s )2etting Fp 2amples*,
Page C.C
Appendix C: XRF Theory
which are long term
stable, are
made to contain enough of the elements being checked to give these
high and low intensity values for each element.
Correction o&
inter&erence
s
There are two main types of interference to be considered+
)i* ,nstrument imperfections )Gverlaps due to insu>cient resolution*
)ii* -atri' e/ects )due to large changes of matri' absorption*
,n addition to these problems occurring between elements in the
sample being measured, there can also be changes of spectral
background a/ecting the measurement of elements in low
concentration. s a general rule, if it is necessary to measure below
&444 ppm it will be necessary to consider some form of background
correction unless the overall matri' is Auite constant.
The spectral background varies inversely with the average atomic
number of the samples. That is to say the higher the average atomic
number )and mass absorption coe>cient* of the sample, the lower the
scattered background. This e/ect is also shown in the intensities
measured from the scattered X-7ay tube anode lines. These
characteristic lines are scattered from the sample and the lower the
average atomic number of the sample the higher their intensities will
be.
2cattered intensities are directly related to the spectral background and,
if measured, can be used to make measured background corrections of
other element gross peak intensities.
Corrections ,t can be seen that the simplest form of background correction is the
subtraction of a 8'ed amount eAual to ,
4
from each measured intensity,
commonly referred to as $;lank 2ubtraction%. 2uch a procedure produces
a series of net intensities related to concentration by+
C a I =
1
*
The principal use of this form of background correction is for the
determination of silicone on paper.
Page C.H
Appendix C: XRF Theory
n overlap correction
can be applied to
either a tube line or
an overlapping
element line. The
corrected, or net,
intensity )I
N
* is
calculated as follows+
I I K I
N measure B
= *
The value of K for a
tube line is
determined by
measuring a $blank%
base material. For
e'ample, if
measuring sulphur in
a diesel oil, a suitable
blank would be a low
viscosity mineral oil
which does not
contain any sulphur.
2pectral overlaps are
corrected in a similar
way but here the
special samples
needed to calculate
the correction factor
must contain both a
signi8cant amount of
the overlapping
element and none of
the analyte element
present. G'ford
,nstruments supply 2etting Fp 2amples )2F2s* which are suitable for
this purpose.
Gnce all the corrections to intensity have been made and a set of $pure%
net intensities e'ist for all elements, correction for change of sensitivity
)a&* can be made. These corrections rely on the basic premise that the
change of sensitivity, either positive or negative, is proportional to both
the concentration of the analyte being measured and that of the
interfering element causing the change+
C a a I k I
i o i
n
n n
1
1
1
1
1
1 = + +

( )( * )
s concentration is directly proportional to the net intensity of each
element, the matri' correction eAuation is e'pressed as an intensity,
where a
0
and a
1
have the same meaning, as e'plained earlier, and ,
&
n
is
the intensity of the nth element.
,f there are ! values to be calculated empirically then a minimum of !
:

standards are normally reAuired for a good statistical 8t.
The intensity model is the easiest to apply because not all the element
percentages need to be known. The sign and magnitude of the
calculated factors will depend on the number of elements used in the
eAuation, because elements omitted contribute to the $balance of the
matri'.% ,t must be remembered that the e/ect of changing an element
for another depends on
Page C.I
Appendix C: XRF Theory
the resultant
change of
absorption for the analyte radiation by what is considered the balance
of the matri'.
Precision The term $precision% is used to describe the repeatability of the
instrument and, to give it some numerical signi8cance, it is normally
described as the standard deviation.
Figure C-H shows a typical Jaussian distribution which can be used for
any measured value where random events are involved. From this it is
seen that the deviation from the mean value for appro'imately ??K of
all the measured values is a number which is called sigma )*. Twice
this value takes in I9K of all the measured values and three times
II.ICK of all. These break-points are used freAuently to describe
various levels of con8dence..
That is to say, there is a & in :4 chance in getting a result with a
deviation larger than twice the one sigma )* value.
3ith X-ray analysis the measurements are made up of a series of
numbers accumulated by counting photons of a particular energy
entering a detector. This counting process obeys a Jaussian distribution
and the standard deviation & sigma )* is eAual to+
1 = total number o !hotons counte"
but
Page C.&4
??K
I9K
: - -:
=

( ) x x
n 1
Figure C-,
Appendix C: XRF Theory
total number o
!hotons counte" # $C
% BEC) & Q)
2ee Figure C-C, where+
;5C < ;ackground 5Auivalent Concentration
D < 2ensitivity
T < CountingEanalysis time
C < Concentration
1 = + ( ) * C BEC Q' counts
however+
Q' counts # ()
so that+
1
1
=
+
=
( ) *
%
( ) C BEC Q'
Q'
C * *
'
o
1 =
+ ( )
%
C BEC
Q'
N+'E, * is negative
o
From this it is seen that+
at -ero = =
BEC
Q'
* *
'
o
%
1
Three times this value, ie+
BEC
Q'
is known as the limit of detection
#G6.
Accuracy The terms $accuracy and $precision% are often interchanged and wrongly
used. The de8nition is, however, Auite clear+ Precision is the ability of
the instrument to repeat a result from the same sample.
,t should be remembered that precision and time are intimately related.
For e'ample, if it is reAuired to improve the sigma at a particular
concentration level by a factor of :, then the counting time must be
increased by a factor of 0. That is to say, the time must increase by the
sAuare of the improvement reAuired.
Page C.&&
Appendix C: XRF Theory
Accuracy is the
correctness of this
result compared with
other reference
methods of analysis.
useful way of
e'pressing the
accuracy is in terms
of the standard error,
)standard deviation of
di/erences between
X-ray and given
values* of the
calibration curve.
From this, it follows
that an instrument
can have e'cellent
precision or
reproducibility but
that the value being
produced could be
Auite wrong.
classical cause for
this would be a poor
calibration, possibly
caused by insu>cient
calibration standards,
inaccurate
concentration values,
errors in sample
preparation, etc.
Page C.&: