Calibration Method for X-Ray Fluorescence Spectrometry
H. J. BEATTIE and R. M. BRISSEY
Thomson Laboratory, General Electric Co., Lynn, Mass.
i first-order linear correction for the absorption effects
of the various elements in an alloj on fluorescent x-ray
intensities is obtained bj essentially assuming a mono-
chromatic incident beam. The result is a set of simul-
taneous equations; the unknowns represent composi-
tion, the diagonal coefficients are determined by in-
tensity measurements on the unknown alloy and pure
elements, and the nondiagonal coefficients are calcu-
lated from intensitj measurements on analj aed binary
alloys. The simultaneous equations are orerdeter-
mined, and the calculating procedure, which minimizes
the error introduced by the assumption, has been
determined empirically. Analj ses based on this
niethod, where onlj pure elements are needed as stand-
ards, hate been performed on quaternary allojs con-
taining moll bdenum, nickel, iron, and chromium.
The calculated lalues agreed in nearlj all cases to
within xk5% of the ralues obtained b! accepted chemi-
cal procedures.
S HER31.4X has treated the problem of fluorescent x-ray
analysis st,atistically ( 3 ) . The following development con-
siders the physical mechanism of x-ray absorption and indicates
an empirical elemental calibration.
If it were not. for absorption effects, the fluorescent s-ray in-
tensity due to an element in an alloy would be proportional to
the atomic density of that element, and a quantit,ative analysis
could be made from the ratio of intensity of the element in the
alloy t o that of the pure element. The absorption effects arise be-
cause the incident x-rays penetrate a significant distance below
the specimen surface, although Koh and Caugherty ( 2 ) have
shown that this distance is not more than 0.004 inch. An ex-
pression for the absorption effects can be derived, but its precise
form is not usable owing t o the continuous spectrum of the inci-
dent, beam. The excitation probability, however, is very high
only for a limited region on the short-wave-length side, of the
absorption edge. By assuming the incident intensity and the
absorption coefficients of all the elements to be constant in this
region of high excitation probability, the first-order absorption
effects of all elements can be easily determined empirically from
binary alloys. The assumption of constant incident intensity
and absorption coefficients introduces an appreciable error when
the alloy contains elements \\-hose absorption edges or characteris-
tic spectra lie in the region of high excitation probability. These
conditions are abundant in the chromium-iron-nickel :illoys;
therefore, this ternary system would be a severe test for the as-
sumption. I n t,he investigation of the chromium-iron-nickel-
niolybdenum quaternary system, molybdenum was added t o
test for large differences in atomic number.
Gillam and Heal dealt with the mathematical problem of
fluorescent x-ray intensities ( 1 ) . Expressions of similar form
are derived here.
The fluorescent x-ray intensity due to a slab of thickness, dz,
a t a depth, x, below the surface is determined (Figure 1).
Let’ Zo(X)d X be the intensity of the incident radiation in the
range of wave lengths between X and +dX. At the slab, the
incident int,ensity is reduced by a factor of
exp [ - d h ) z a I
where p . ( X ) is the linear absorption coefficient of the specimen for
the incident wave length, A, and a is a geometric constant of the
980
b
apparatus. Roughly, 01 = chc e; more precisely, it is compli-
cated owing t o the incident beam divergence.
If ‘4 is the area of the mask in front of the specimen, and if
Sais the atomic density of the element, a, in the specimen, there
are .ITaA d x atoms of a in the slab. If Q&) is the excitation proba-
bility of element a a t wave length A, the intensity of fluorescent
x-radiation emitted by the slab in all directions is
l o ( X ) d X exp [ - ~ , ( h ) a z ] Q , ( X ) S , A d z
A fraction of this radiation within the infinitesimally small
solid angle, dw, which makes an angle 4 with the specimen surface
in the plane of incidence, ~ 1 be1 diffracted by the spectrometer
crystal to the Geiger counter. On emerging from the surface,
this radiation from the slab has been further reduced in intensity
by a factor of
exp ( - w z csc v)
where P J is the linear absorption coefficient of the specimen for
the monochromatic fluorescent x-radiation of element a. The
fluorescent intensity due to element a in the slab, excited by the
range of wave lengths between X and +
d2Z, = lo(X)Q,(X)SaAd~exp
dX is
+
[ - ( ~ ~ ( x ) 0 1 P/ csc ‘p)z]dzdX (1)
The intensity due to the entire volume of the specimen is ob-
tained by integrating Equation 1 from a depth of z = 0 to z = m,
giving
To obtain an exact expression for fluorescent x-ray intensity,
Equation 2 must be integrated with respect t o X; however, three
quantities in Equation 2 are unknown functions of A. At wave
lengths longer than the absorption edge of a , Q a ( X ) is zero; it
is maximum a t the absorption edge, and decreases rapidly as A
decreases from the edge. If we assume that Zo(X) and p e ( X ) are
essentially constant over the range of high &.(A), the approximate
expression for fluorescent x-ray intensity due to a is
(3)
-
- K,S,
Pea + Pf csc ‘p
where K , = Zo &. A d w is a constant of the apparatus and the
element, a, and is independent of the absorption in the specimen.
The intensity is seen to be proportional to the atomic density,
Sa, of a, modified by the absorption effect of the alloy in the
denominator of Equation 3. \l*e shall denote this absorption
effect by p J f a 8 , where p1 is the density of the alloy, subscript a
refers to the radiating element, and subscript s refers to the alloy
matrix of all the elements. The mass absorption coefficient,
p / p , of a mixture of elements is the sum of the mass absorption
coefficients of all elements in proportion t o their composiLion by
weight. If T Y b is the fractional weight of element 6 in the alloy,
then
the double subscript, ab, indicating the mass absorption coefficient
of element b for the excitation (first term of Equation 4) or char-
V O L U M E 2 6 , N O . 6, J U N E 1 9 5 4 981

acteristic (second term of Equation 1) radiation of element a . The unknowns are further governed ti>- the constraint:
From Equation 1,it follows that
TYa f wb f Tt'c + = 1 (11)
EXPERIMENTAL PROCEDURE

Fifteen specimen blocks were prepared: Four were the high-

If the fluorescent \-ray intensity I,,, fiom a pure element speci-
men of a is divided k)y that from the alloy, I,,, the ratio is (using
Equations 3 to 5 )
.I-.$ p r is the numbri of atoms of a per gram of alloy, and -VCa/pa
1-
the number of atoms of a per gram of n Therefore,
the number of grams of a per gram of alloy, or the fractional
weight of a. IIultiplying Equation 6 tiy It', and setting L i a b =
purity elements, chromium, iron, nickel, and molybdenum, 15 ith
respect to which the intensity ratios were taken; s i i were all of
the binary combinations of these elements for determining the
empirical absorption parameters of the system; the remaining
five were test alloys, three being chromium-iron-nickel ternaries,
and two being chromium-iron-nickel-molybdenum quaternarie8.
The binaries, ternaries, and quaternaries were chemicall!,
analyzed. These analyses were conducted R ith exceptional care
and the values from duplicate samples averaged. The pro;
cedures were checked by analyzing Sational Bureau of Standards
samples a t the same time.
,111 elemental, binary, and ternary samples Tere checked
qualitatively on the x-ray spectrometer for impurities, and none
xqas detected.
Fluorescent x-ray intensities were read using the General
Electric XRD-3S unit. The line intmsities were read in all
samples for a given element without changing the goniometer
setting or the stabilized x-ray tube current. The latter was
adjusted so that the intensity from the pure element did not
exceed 3500 counts per second. Appropriate background
intensities Jvere determined and subtracted from the observed
line intensities. Table I V lists intensity ratios for the alloys and
the actual intensities of the elements in counts per second. The
total count for each reading was 65,536. giving a statistical
probable error of 0.26%.

or
Table I. Chemical Analjses of the Binary Alloys
Analysis Binary Partner
b#a for Cr Fe Ni 310
The absorption parameters, .qO,,, can bcs easily obtained em- Cr 48 24 48.07 74.28
Fe 50'83 5 1 53 6 5 . 33
pirically by reading intensity ratios on chemically analyzed Xi 48.19 4R: 4 5 53 70
310 23.53 34.'44 l(i.27 ...
binary alloys. Thus, from Equation 8,
Table IT. Intensity Ratio for Binary Alloys
Radi-
atina - Binary Partner
(9) Element Cr Fe Xi 310
Cr 1 000 1 it30 1 81.5 1 841
Fe 3 3h0 1 000 1 013 1 835
N1 2 800 3 Ai0 1 000 1 954
310 3 770 2 787 2 461 1 000

Table 111. Absorption Parameters

Radi-
ating Absorbing Element
Element Cr Fe XI AI0
Cr 1.000 0 721 0 813 2,660
Fe 2.482 1.000 0,676 1.682
Xi 1,863 2.420 1.000 1.108
310 0.877 0.9435 1,260 1 ,000

The compositions of the binary alloys are given in Table I,

and the x-ray intensity ratios of the binaries are listed in Table
11. The data in these two tables are substituted into Equations
0 to determine the absorption parameters (Table 111). K i t h
these evaluated, the simultaneous equations for the chromium-
iron-nickel-molybdenum quaternary

Figure 1. X-Ray Fluorescence

+ 0.721Fe + 0.81:3Si + 2.6(iOlIo 0
- ( R c , - 1)Cr = (A)
2.482Cr - ( R F ~- 1)Fe + 0.676Si + 1.582110 0 = (B)
1.863Cr + 2.120Fe - ( R s ; - 1 ) s ; + 1.108lIo = 0 (C)
Once these absorption parameters are determined over an al-
loy system, any alloy in the system can be analyzed by reading 0.877Cr + 0.9435Fe + 1.200Si ; - 1)lIo 0
(R5c0 = (D)
fluorescent x-ray intensity ratios, R,, Ra, . . for each element .) h n alloy of this system is analyzed by reading the four s-ray
in the alloy and substituting them into simultaneous Equations intensity ratios from the alloy and the standard elements, sub-
a: stituting them into the nhovcl equations, and solving the equa-
-(Ra - 1 ) I v a + +
AabTT.'b + A o c Tf'c 0 tions.
~4baTZ'a
AcaWm
-(Rb -
+ B,bIl'a
+
1 ) w b f
- ( R , - l)W, +
0
~ 4 b e W e
= 0
(10) Since there are no constant terms in these equations, m-e can
merely obtain ratios among the unknowns. The equations are,
982 ANALYTICAL CHEMISTRY
therefore, overdetermined, since there are only n- 1 independent have the lowest intensity ratio; therefore, it is taken as the most
ratios among n quantities. A set of ratios can be obtained by abundant element. Equation B is omitted, the equations are
dividing through by any of the four unknowns. We can then divided through by Fe, the constant terms are transposed, and
omit any one of the four equations to solve for the three unknown the intensity ratios from Table IV are substituted.
ratios with the remaining three equations. Thus there are a t
least 16 independent ways to solve the above equations. -4.04Cr/Fe + 0.8131\'i/Fe + 2.660?*Io/Fe = -0.721 (A)
These 16- independent procedures yield 16 different sets of 1.863Cr/Fe - 5.640Si/Fe + 1.108?vfo/Fe = -2.420 (C)
solutions, because the equations are based on a rough assumption
and the coefficients are empirical. Thus the equations are mu- 0.877Cr/Fe + 1.260X/Fe - 14.34,11o/Fe -0.9435
= (D)
tually inconsistent to some extent. It is desirable to choose that If the denominator determinant of the coefficient4 times the
calculating procedure which tends to minimize the errors intro- adjustment factor is denoted by k , then the numerator dctermi-
duced by the assumption. Such a calculating procedure was nants are
developed empirically; however, the result indicates that it
can also be developed from the symmetry of the equations. 0.813 2 660
k Cr/Fe = -2 420 -5 640 1
1 108 = -108 61
1 ~ ~ .1 260
~ ~-14 ~ 34 5
CALCULATING PROCEDURE
1-4.04 -0,721 2.660 1
Initially, one element must be distinguished from the others. k N/Fe = 1 863 -2 420 1.108 = -163 42
1
~

Empirically it develops that the best solutions are obtained by 0 877 -0 9435 -14 34
distinguishing the most abundant element in the alloy. By 1-4 04 0 813 - 0 721 I
dividing throughout by the unknown representing the relative k RIo/Fe = 1 863 -5 640 - 2 420 = - 39 37
weight of the abundant element, the effective number of unknowns 1 0 877 1 260 - 0 94351
is reduced by one.
Multiplying Equation B by k/Fe and transposing the resulting
The result is four equations in three unknowns, each equation
constant term (with R F substituted
~ in from Table IV) yields
containing a constant term. The triple ratio among the three
unknowns, which is the triple ratio among the three elements 2.482 k Cr/Fe + 0.676 k S i / F e + 1.582 k Mo/Fe =
other than the abundant element, is best determined from the 1 . 3 4 6 k (B)
three equations which contain the intensity ratios of these ele-
The above three numerator determinants are substituted into
ments. Therefore, the equation which contains the intensity
this equation, and the result is solved for k, giving
ratio of the abundant element is omitted, and the triple ratio is
obtained by evaluating the three numerator determinants of k = - 328.62
the remaining three equations. These determinants are then
multiplied by an adjustment factor, whose value depends on Equation 11 can now be used to evaluate the four unknowns
the ratios of each of the elements to the abundant element, and,
as follows (disregarding the negative signs):
therefore, whose value is best obtained by substituting the three k Cr/Fe = 108 61
determinants into the fourth equation, which contains the inten- k Ni/Fe = 163 42
sity ratio of the abundant element. The final four individual k Mo/Fe = 39 37
k =k Fe/Fe = 328 62
solutions are determined by using Equation 11.
The use of the equations in combinations other than that just
k/Fe(Cr +Si + +
110 Fe) = 640 02 = k/Fe
described consistently yields solutions further from the accepted giving
values obtained by chemical analysis. The range of values cal-
culated among various combinations of the equations is particu- Cr = - = 16.97%
640.02
larly large for chromium-iron-nickel alloys, because the abundant
anomalies in the ranges of high excitation probability introduce Fe = 328'62
_ _ = 51.35%
large errors t o the assumption. Therefore, this calculating pro- 640 02
cedure apparently uses all equations in the best possible manner
to minimize these errors. The relation of this procedure to the
symmetry of the problem fortunately provides convenient rules-
of-thumb. 110 = 39'37 = 6.15%
~

640,02
SAMPLE CALCULATION
for the analysis.
If Alloy 4 from Table IV, which gives the intensity ratios
of the ternary and quaternary alloys, is chosen iron is seen to DISCUSSION

The results on all five test alloys, using the calculating pro-
cedure outlined above, are given in Table V, together with the
Table IV. Intensity Ratios for Ternary and Quaternary
Alloys chemical analyses. For Alloy 3, whose three elements are equally
Ra = Iooa/Ios
abundant, the best analysis is obtained by taking the mean of
Alloy the three sets of solutions, using each of the elements as the abun-
NO. Cr Fe Ni 110 dant one. This is not true for the remaining alloys, which, like
1 3.610 1.918 14.34 ... most practical alloys, have only one element in high abundance.
2 3,535 6.620 2.152 ...
3 2.477 3.665 5.373 The single calculating procedure, therefore, should be adequate in
4 5.04 2.346 6.640 1i:34 most cases.
5 4.01 1.942 14.72 41.35
I n the derivation of Equation 2, all of the excitation radiation
raa b 597 2796 3270C 1076
was assumed to come from the direction of the x-ray source.
a Intensity in counts per second on pure element corrected for back-
ground. When a certain element in the alloy emits K CY radiation in the
'Intensity from a 0.75 X 0.76 inch sample area. Bent mica crystal region of high K excitation probability for the element being
mas 0.0016 inch thick.
Nickel intensities were determined with a 1%'-targetx-ray tube operat- read, there will be a strong monochromatic component of the ex-
ing a t 24.1 ma. and 48.5 kvp. All other elements were determined under
4 9 . 6 ma. and 48.5 k v p . citation spectrum coming from all directions. This secondary
excitation radiation, however, will still be strongest from the
V O L U M E 26, NO. 6, J U N E 1 9 5 4 983

in Table V. I n Alloy 2, having a high nickel-low iron content,

Table V. Final Results the x-ray reading of iron is 8.6% too high; whereas, for Alloy
Composi- 5, in which the nickel-iron composition is reversed, the x-ray
Composi- tion by reading of iron is 2% too low. Thus it is well to have the binar-
Composi- tion by X-Ray b y
tion b y X-Rag Propor- ies, from which the parameters are determined, near the 50-50
Chemical Csing tional
Al- Ele- Analysis, Param- % Intensities, % composition.
loy ment 76 eters, % Error % Error
The last two columns in Table V give the compositions and
1 Cr 19.85 20.29 + 2 22 27.7 +39.5
Fe 65.94 05.95 + Nil 52.1 -21.0 relative errors, assuming the relative compositions to be inversely
Si 13.98 13 76 -1.57 6.98 -50.0 proportional to the intensity ratios of Table IV. The effects of
2 Cr 21.23 20 39 -3.96 28.7 f35.2 absorption and mutual fluorescence are readily observed, and
Fe 13.85 l5,04 +8.60
-0.42
15.1
46.5
+ 9.0
-28.3 the effectiveness of the absorption parameters in correcting for
Xi 64.84 64.57
these effects is in evidence.
3 Cr 32.13 32.44 +0.97 40.4 4-25 8
Fe 35.43 35.86 +1.21 27.3 -22.9 Analyses based on this procedure offer the significant advan-
Ni 31.86 31.66 -0.63 18.6 -41.5
tages of a physical method of analysis that is not restricted by
4 Cr 16 45 16.96 +3.10 19 8 +20.4 the need for reference or standard samples with a composition
re 50 20 51.35 $2.29 81 0 +61.4
S I 26 80 25.54 -4.70 15.1 -43 6 similar to the sample being analyzed.
110 6 55 6.15 -6.10 6.52 - 0.5
J Cr 18 43 19.63 +6.61 24.9 +35.1 CONCLUSIONS
Fe 66 26 64.93 -2.01 51 5 -22.3
12 99 13.23 +1.85 6.79 -48.4
N1
.\Io 2 32 2.22 -4.31 2.42 + 4.3 Absorption-excitation effects can be expressed in terms of em-
pirical parameters relating pairs of elements.
These parameters and flourescent intensity data can be com-
bined in theoretically simultaneous, linear equations.
direction of the surface. Gillam and Heal ( 1 ) gave, without der-
A solution of these equations ultimately gives the percentage
ivation, an expression for the effect, which they called mutual by weight of each element in an alloy.
fluorescence, and which was to be added to a term analogous to This definite system for solving these overdetermined equations
Equation 2. The inclusion of thls expression would prohibit greatly simplifies the mathematical manipulation.
the use of linear equations with empirical coefficients. The accuracy of results obtained on ternary and quaternary
The upe of empirically determined absorption parameters alloys compares favorably to that obtained by routine chemical
compensates for this omission. This can be seen from the values
procedures.
of the absorption parameters (Table 111). Thus A c ~ F A ~ c, ~ N ) , This method is not restricted by need for reference samples.
and A F ~ have
N ~ values less than unity-Le., less than the self-
absorption value. Since the absorbing element in each case is a
ACKNOWLEDGMENT
little heavier than the radiating element, the low absorption pa-
rameters can be due only to mutual fluorescencr. Hence, the The authors are indebted to Sol Gootman for the careful chemi-
empirically determined parameters give an automatic correc- cal analyses of the alloyP.
tion for mutual fluorescence. The three corresponding param-
eters with the subscripts reversed, on the other hand, are ap- LITERATURE CITED
preciably greatel than unity. Close examination shows that (1) Gillam, E., and Heal, H. T., Brit. J . A p p l . Phys.. 3, 363-8
there is a correspondence in the departure from unity between (1962).
the two sets of three parameters. This stems from the recipro- (2) Koh, P. K., and Caugherty, B., J . A p p l . P h y s . , 23, 427-33
cal relation: high excitation of A from B’s K a radiation (1 952).
(3) Sherman, J., Pittsburgh Conference on Analytical Chemistry
means high absorption of B’Rradiation by A . and Applied Spectroscopy, l l a r c h 1953.
That the absorption parameters with their automatic correc-
tion for mutual fluorescence are not strictly constant can be seen RECEIVEDfor review September 14, 1953 .\ccepted Rlarch 29, 1954.

Spectrophotometric Determination of Mumulone

Complex and Lupulone in Mops
G O R D O N A L D E R T O N , G. F. BAILEY, 1. C . LEWIS, and FRED STITT
Agricultural Research Service, Western Utilization Research Branch, U. S. Department o f , .Jriculture, Albany 6, Calif.

A rapid method for determination of lupulone and of
humulone complex (humulone plus cohumulone) in
petroleum ether extracts of hops is based on absorbance
measurements at wave lengths of 275, 325, and 355 mp.
The influence of solvent and of alkali in methanol on
the spectrophotometric stability of lupulone and
humulone complex has been studied. The background
absorption of hop extracts due to components other
than these two has been investigated by adsorption on
ion exchange resin and silica gel, by molecular distilla-
tion, and by alkaline extraction of hop extracts. Con-
ditions to w-hich a background absorption must con-
form in a ternary spectrophotometric analysis, so that
it makes no contribution to values for each of two real
components, are pointed out. These conditions are
shown to be approximated well by the variable back-
ground absorption of hop extracts due to components
other than lupulone and humulone complex.
T HERE is need for a rapid method for determination of the
resinous constituents of hops, which are important in brew-
ing. These include humulone, lupulone, and cohumulone.
Cohumulone was discovered recently by Rigby and Bethune
(11), who isolated it by countercurrent distribution. It is