in precision. A further gain in detecTable II.
Matrix Na2C03
Effect of Matrix on l e a d Determination
Pb added, wt. % 27.9
Av. ZrOz
27.9
Av .
2.7y0 for lead concentrations ranging from 15 to 1 0 0 ~ o . The error due to sample preparation a n d the error due to the spectrographic process can be assessed by using the d a t a obtained from the samples containing sodium carbonate and zirconium oxide (Table 11). For each sample, four determinations were made on each of four fusions. The standard deviation within sets of fusions was 0.84 wt. yofor the sodium carbonate sample and 0.82 wt. yo for the zirconium oxide sample. Similarly, the standard deviation
1 29.1 27.6 25.6 27.8 27.5 27.9 27.4 27.1 26.2 27.2
Pb found, wt. % Fusion 2 28.9 28.3 28.7 29.2 28.8 27.5 27.5 27.2 27.1 27.3
3 27.8 29.3 28.3 28.2 28.4 26.4 26.9 25.1 26.4 26.2
4 27.4 27.6 27.6 27.7 27.6 25.2 25.2 25.7 27.1 25.8
between sets was 1.20 and 1.17 wt. %, respectively. This indicates that about one half of the error is due to sample preparation. I n practice, it is convenient to average results of four determinations from one fusion mixture. This average should be within 5% of the true value at the 95% confidence level. By substituting carbon for graphite in the sample preparation and by using an applied voltage of 250 volts, the method can be extended to lead concentrations as low as o.o.!iyOwith some loss
tion can be made by fusing the sample with a smaller amount of potassium pyrosulfate. Some work done in determining light elements such as silicon and calcium, transition elements such as iron and nickel, and heavy elements such as copper and bismuth indicate that the method can be applied as a general spectrographic procedure. LITERATURE CITED
(1) Am. SOC. Testing Mater., Philadelphia, “Methods for Emission Spectrochemical Analyses,’” 1957. (2) Danielson, A., Lundgren, F., Sundkvist, G., Spectrochim. Acta 15, 122 (1959). (3) Eeckhout, J., Zbid., 3,575 (1949). (4) Feldman, C., Ellenburg, J., ANAL. CHEM.27, 1714 (1959). (5) Harmon, D. D., Russel, R. G., Zbid., 23, 125 (1951). (6) Jaycox, E. K., Applied Spectroscopy 12, 87 (1958). (7) Milbourn, M., Hartley, H. E. R., Spectrochim. Acta 3, 320 (1948). (8) Steinberg, R. H., Belic, H. J., ANAL. CHEW20, 72 (1948).
RECEIVED for review November 18, 1963. Accepted January 9, 1964. 11th Detroit Anachem Conference, Detroit, Mich., October 21, 1963.
Intensity Ratio Technique for X-Ray Spectrometric Analysis of Binary Samples Particular Application to Determination of Niobium and Tin on Nb3Sn-Coated Metal Ribbon EUGENE P. BERTIN Commercial Receiving Tube and Semiconductor Division, Radio Corporation of America, Harrison,
b X-ray fluorescence spectrometric determinations are usually calibrated with curves of net intensity vs. concentration derived from measurements on standards. This technique usually fails when the samples differ from one another or from the standards in surface condition or in the area, shape, or effective thickness exposed in the x-ray spectrometer, and when it is impractical to reproducibly position or orient samples unsuited to masking. For binary systems, this difficulty can sometimes be resolved by measuring the intensity of a line of each component and making a log-log plot of net intensity ratio vs. concentration ratio. The curves are linear, and the scatter in the points is reduced. The technique was applied to W-Re and Bi-Sn alloys, to mixtures of Cr203and Ag20 and of FesOl and CuO powders, to briquetted Nb-Sn alloy pow826
ANALYTICAL CHEMISTRY
ders, to Nb-Sn coatings on metal ribbons and ceramic plates, and to G a and In in Ga-In arsenide alloys having constant molar As concentration.
X
-RAY FLUORESCENCE spectrometric
determinations are usually calibrated with curves of net intensity us. concentration derived from measurements on standards. This technique usually fails under the following conditions: when the samples differ from one another or from the standards in the form (dimensions, area, shape, effective thickness, etc.) presented to the x-ray spectrometer, or in surface relief and texture (unevenness, graininess, porosity, chipping, cracking, machine or grind marking, etc.); and when it is impractical to reproducibly position or orient samples which, be-
N. 1.
cause of small size, irregular shape, etc., cannot be made to present a uniform surface to the x-ray spectrometer by the usual technique of placement behind a mask. I n such cases, all samples and standards may be converted to a common form such as powder, briquet, fusion mixture, glass disk, or solution. However, if the analysis involves a binary system, it may be possible to deal with the samples without preparation by measuring the net intensity of a line of each constituent and making a log-log plot of net intensity ratio us. concentration ratio. The principle derives from the basic relationship between emitted intensity I and concentration C: ZA
=
(KC“)A
ZB = (KC”)B
-
where K and n are constants different for the elements A and B. If 12.4 nB,
I
0 50 COMPOSITION, mole X
Figure 1.
0 Re
W/Re CONCENTRATION RATIO I m d e Xl
""""I
1
COMWIITION. .I. Y
Figure 2.
Calibration curves for Bi-Sn alloys
Clslibration curves for W-Re alloys
the following derives from these relationships :
Thus a log-log plot 01'Ia/lB as a function of C A / C Bshould be linear. The calibration curve is established from measurements on a series of standards having known A-B compositions. The advantages of the technique are speed and convenience, linearity and improved reproducibility of the calibration curves, reduced scatter in the data, and insensitivity t o reasonable variations in sample dimensions, texture, position, and orientaticn. However, the technique is inapplicable very near the composition extrem2s where a small increment in concentration corresponds to a large change in intensity ratio and vice versa. I n general, the intensity measurements must be made on both elements without dismrbing the sample. It follows that both lines must be measurable with thtb same x-ray tube target, crystal, collimator, and detector, unless the spectrometer is provided with means for rapidly changing these components during operation. Finally, the technique doe:, not necessarily compensate for all the differences that may occur among samples and standards. The technique was evaluated for six binary alloy systems in a variety of representative samp e forms: W-Re as polished disks; 15-Sn as cast bar, wire, ribbon, drops, and small disks; Crz03 and Ag20 af, mixed powders; Fe304 and CuO as mixed powders; Kb-Sn as briquetted mixtures of alloy powder and starch and as superconductive Nb-Sn coatings on metal and ceramic substrates; and Ga-In arsenide as thin slices of ingot. The six binary systems also represent a variety of excitation conditions. I n the W-Re, h'b-Sn, Ga-In, and Cr-Ag systems, the two elements are separated by progressively larger intervals in atomic number-1, !3, 18, and 23, respectively; also, the two lines used are in the same series ( K or, for W-Re, L )
and are subject to minimum specific absorption and enhancement effects. The Bi-Sn system involves an L and a K line. I n the Fe-Cu system, CuKn is strongly absorbed by Fe with consequent enhancement of FeKa. The GaAs-InAs samples consisted of various proportions of the pure end phases with no free As. Thus Ga and I n concentrations varied in an As matrix of fixed molar concentration, enabling application of the ratio technique to a certain type of ternary system. The Crs03-Ag,0 and Fe304-Cu0 systems may be regarded as binary for the purpose of this study because of the virtually negligible effect of the oxygen. These two systems were prepared synthetically and were studied solely to evaluate the applicability of the ratio technique to cases involving large differences in excitation potential, wavelengths of the two lines, absorption of one element for the line of the other (Cr-Ag), and serious absorption and enhancement effects (Fe-Cu). By far the most work was done on the Nb-Sn coated samples. The requirements of these determinations were particularly demanding. In addition to being rapid, convenient, nondestructive, reasonably precise and accurate, and insensitive to sample position and orientation, the method had to be applicable to various forms of substrate, including various sizes of metal ribbon and wire, disks, plates, and cylinders; various substrate compositions, including Pt, Pt-plated Xi-base alloy, and ceramic; a range of coating weight per unit area; a wide range of Nb-Sn composition; and small and/or irregularly shaped samples. The more conventional x-ray spectrometric techniques were unsuitable for the Xb-Sn coating determinations because they require relatively large samples and/or relatively elaborate and time-consuming sample preparation, and because they are destructive. X-ray spectrometric methods developed (3) for determination of plate metals on plated wires involve either the plate metal line intensity from uniform re-
producible areas of sample, or the ratio of the intensities of a plate and substrate line. These methods are not applicable to systems where there are two plate metals in various proportions, Other x-ray spectrometric methods for multiple platings are not applicable because they presuppose a more or less distinct layer structure so that the outermost plating can be determined from its x-ray emission without absorption correction. Then a lower layer can be determined in either of two ways. Its emission may be corrected for attenuation by the overlying layer ( 7 ) ,or a set of curves may be established relating the intensity of a substrate line with thickness of the lower layer, each curve for a different outer layer thickness (4). A technique reported recently by some Japanese workers (6) appears to be applicable to the problem, but the work came to the writers' attention after the ratio method reported here had been developed. EXPERIMENTAL
Equipment. The work was done on a General Electric model X R D - 3 x-ray secondary emission (fluorescence) spectrometer t h a t had flatcrystal optics and conventional-that is, not inverted-geometry, and was fitted with the following components: iMachlett type ,4EG-50S x-ray tube with tungsten target, LiF crystal, 15/8 X 0.070-inch soller source collimator, 31/2X 0.005-inch soller detector collimator, and G. E. type SPG-2 sealed Kr-filled proportional counter. The original No. 1 electronic circuitry was retained, including the 14-stage binary scaler, the capacity of which was increased 100-fold by addition of a mechanical count register. The conventional sample drawer was fitted with sample masks having circular or rectangular windows of various dimensions. The Kb-Sn briquets were measured in a special rotating sample drawer (1) and the powders in Lucite cells ( I ) . Calibration Standards. IT-Re disks 0.50 inch in diameter and inch thick were cut from chemically analyzed ingots and polished on one face. Buttons of Bi-Sn alloys were prepared by VOL. 36, NO. 4, APRIL 1964
a
827
yu%04
Figure 3. mixtures
Calibration curves for CrzOs-AgzO powder
casting the pure metals in various proportions. The Crz03-Agz0 and FesOaCuO mixtures were prepared from Reagent Grade powders by weighing in suitable proportions and mixing in a Spex Industries Mixer Mill. Nb and Sn powders were mixed in various proportions, pressed into bars at 10,000 pounds per square inch, sintered in a sealed evacuated quartz tube at 1050°C. for 16 hours, and reduced to powder having particle size