values, the hydrogen ion also competes for resin positions but even a t low p H values the ionic strength of the solution outside the resin affects the distribution coefficient due to sorption of sodium chloride at the higher ionic strengths. Theoretically there is no justification for terminating Lhe series in Equation 6 with the second term but such a termination yields Equation 8 which is of the same form as Equation 2 which can be used to represent the data satisfactorily. This also indicates that some other factor such as sorption of sodium chloride which is related to the ionic strength also affects the distribution. Equation 2 thus satisfactorily represents the data since the ionic strength in the solution outside the resin controls the main factors both outside and inside the resin which affect, the distribution coefficient of the alkaline earth ions with the resin.
LITERATURE CITED
(1) Bates, Roger G., Bower, S-incent E.,
ANAL.CHEM.28, 1322 (1956).
( 2 ) Chaberek, S., Jr., Martell, A. E., J . Am. Chem. SOC.74, 5052 (1952). (3) Clark, W. AI., Lubs, H., J . B i d . Chem. 25, 479 (1916). (4) Collins, A. G., Pearson, C., Attaway, D. H., Ebrey, T. G., U . S . Bureau of Mines Rept. 6087 (1962).
(5) Dow Chemical Co., Midland, Michigan, “Dowex A-1 Chelating Resin” Form KO.164-80-64, February 1964. (6) Harned, H. S., Owen, B. B., “The Physical Chemistry of Electrolytic Solutions,” 3rd ed., p. 525, Reinhold, New York, 1958. ( 7 ) Helfferich, F., Sature 189, 1001 (1961). ( 8 ) Imoto, H., Bunseki Kagaku 10, 124 (19611. \ - - - - I
(9) Krasner, J. K., Marinsky, J. A., J . Phys. Chem. 67, 2559 (1963). (10) Kunin, Robert, “Ion Exchange Resins,’’ 2nd ed.,. 1-). 38,. Xilev, ” . New York, 1950. i l l ) Levden. Donald E.. Underwood. A. L., J.“‘Phyi.Chem. 68,2093 (1964).
(12) Loewnschuss, H., Schmuckler, G., Talanta 11, 1399 (1964). (13) Oelschlager, W., Z . Anal. Chem. 146, 339 (1955). (14) Olsen, R. L., Diehl, H., Collins, P. F., Ellestad, R. B., Talanta 7, 187 (1961). (15) Reyden, A. J. van der, van Linger, R. L. AI., 2. ilnal. Chem. 187, 241 (1962). (16) Ringbom, A., “Complexation in Analytical Chemistry,” p. 249, Interscience, New York, 1963. (17) Turse, R., Rieman, W.,111, Anal. China. Acta 24, 202 (1961). (18) Kelcher, Frank J., “The Analytical Uses of Ethylenediaminetetraacetic Acid,” p , 144, Tan Nostrand, New York, 1958. RECEIVED for review September 24, 1965. Accepted March 10, 1966. Ilivision of Analytical Chemistry, Winter RIeeting, ACS, Phoenix, AriA., January 1966. Taken in part from the thesis submitted by Jerry L. Side5 to the Graduate School of Southern Methodist University in partial fulfillment of the requirements for the degree of XIaster of Science.
Quantitative X-Ray Spectrographic Determination of Traces of Elements Using Direct Electron Excitation C. J. TOUSSAINT and GILBERT VOS Chemistry Deparfmenf, CCR, Euratom, Ispra, Italy
b An x-ray spectrographic method using direct electron excitation for the quantitative determination of traces of a number of elements in different matrixes is described. Limits of detection are reported for Zn, Cu, Si, and M g in aluminum, Co in low alloy steel, P in bronze, and N o in a graphite matrix. The results are compared wiih those obtained by other methods. Factors affecting further improvements in the sensitivity limits of low atomic number elements are discussed.
x
RAY
SPECTROGRAPHIC
ANALYSIS
using direct electron excitation has been known for nearly 40 years (6, 7 , 10). However little use was made of this technique for macro analysis owing to technical difficulties such as insulation and vacuum requirements. h revival of this method took place near 1960 and an excellent review of this recent period is given by Campbell and Brown (1). Electron excitation offers the advantages of reduced absorption and enhancement effects and smaller critical thickness. Disadvantages iiiclude practical application only to solids of sufficient conductivity (mixing the sample with a conducting material or
deposition of a conducting layer on the surface of the sample is of course possible) and inability to analyze quantitatively volatile samples or thoze with low melting points. With regard t o background, electron bombardment results in continuous radiation, the intensity of which increases with the acceleration voltage of the electrons and the atomic number of the target material. With fluorescence excitation, the background is caused principally by coherent or incoherent scattering of the primary radiation from the sample. Inasmuch as the detection of traces of light elements by fluorescence analysis is difficult because fluorescence yield decreases sharply with decreasing atomic number, detection limits are presented here for some low atomic number elements in various matrixes. EXPERIMENTAL
Equipment. T h e apparatus used was t h e Cristallobloc 31 x-ray spectrograph (Compagnie G6nBrale de Radiologie, Paris) equipped with a demountable tube (Figure 1) and a rhenium filament. The movable anode is a cylinder with vertical axis and six anticathode faces. When the x-ray tube is used for secondary emission the six anticathode faces consist of different target materials.
Targets available are gold, silver, platinum, nickel, rhenium, rhodium, cobalt, chromium, iron, molybdenum, tungsten, and copper. For electron excitation analysis, the metallic or solid samples are placed directly in slits machined in a solid copper anode. Since the anode is rotable under vacuum, six different samples ran be analyzed successively without breaking the vacuum. The electron sources consist of 7 turnings of tungjten or rhenium with a diameter of 0.22 mm. The sample surface irradiated by the electrons can vary between 5 X 1 and 1 X 1 mm.; in this work we used 5 X 1 mm. Changing from one series of 6 samples to another series (or using secondary emission from one series of six different targets to another) can be accomplished in a few minutes, owing to a bypasi and forevacuum system. The vacuum pumping group consists of a rotating pump with a 2 cubic-meters-per-hour output, a small mercury pump with an output of 8 liters per second and an oil diffusion pump having a 120 liters-per-second output. Instrumental Conditions. Instrumental conditions are given in Table I. T h e optimum excitation conditions for each element were chosen by considering a figure of merit ( I , $ ) , R d B where R is the peak to background ratio and B the intensity of the background in counts per second, Figures 2 and 3 show the figure of merit for the A1 Krv line from a pure VOL. 38,
NO. 6, MAY 1966
71 1
1
Figure 1. X-ray tube: ( 1 ) cooling system, (2) movable anode, (3) anodeholder, (4) position of the filament, (5) position of the sample, (6) filamentsupport, (7) cathode, (8) connection with the vacuum
aluminum sample, together with the peak t o background ratio and the total peak intensity. Figure 2 is obtained by changing stepwise the anode voltage from 6 to 20 kv. by a constant electron current of 0.5 ma., Figure 3 by changing the anode current from 0.1 ma. to 1.0 ma. and keeping a constant anode voltage of 10 kv. Examination of Table I shows that the optimum exciting voltages for Co, Cu, and Zn must be
about four times the ionization potential of the K-level of these elements. This is necessary to obtain a good peak-background ratio. For phosphorus however, the optimum value determined was about 6.5 and for sodium even 11 times the ionization potential of the excited level. This is much more than found for the medium and hard x-ray region, but agrees well with the results of Dolby (2) and Holliday (11). The latter, for instance, found that the optimum excitation voltage for C K a was about 15 times the ionization potential. T o determine the exact nature of this effect more experimental work will have to be done. Concerning the electron current, we found that applying more than 0.2 ma. gave no increase in the figure of merit. Sample Preparation. T h e metallic samples were cut in small sheets with dimensions of 10 X 16 mm. and a thickness of 1 mm. Thereafter they were polished using a 50-micron diamond paste t o avoid surface contamination and to reduce the relief t o a minimum. This surface polishing was absolutely necessary, because, as already stated, the critical thickness is very small (for esample, Stoddard (15) found the value for the penetration depth of 50 kv. electrons mm.), so in aluminum to be about t h a t the contribution of the surface layer to the intensity of the fluorescent radiation is dominant. The powder samples were, if necessary, mived with spectroscopically pure graphite to render the surface conductible to the electrons, and then pressed into pellets with the above mentioned dimensions. Because of particle size effects on intensity in powder samples, the same precautions should be taken as in x-ray fluorescence analysis. As a general rule large particle sizes cause an abnormal absorption of the x-radiation emitted, especially when it is in the longer wavelength range. Reference should be made to the evcellent work of Fonda (4) who made a detailed study of the effect of particle size on intensity in electron excitation analysis. Standards. Calibration curves for trace detection in aluminum were drawn using the aluminum standards of Pechiney (13). For phosphorus de-
Table 1.
termination in bronze, standards came from the Canadian Association for Applied Spectroscopy. No standard was available for cobalt detection in low alloy steel and various samples analyzed by spectrophotometry and activation analysis were used. Standard samples for the sodium determination were prepared by mixing small quantities of borax (Na2B407) with specpure graphite. RESULTS
Detection limits are listed in Table 11. Examples are given below where the trace is a light element in a light matrix, a light element in a heavy matrix, and a heavy element in a heavy matrix. Light Element in Heavy Matrix. The example presented is the analysis of Cu and Zn in aluminum. The sensitivity limits appear to be of the same order of magnitude as found by x-ray fluorescence analysis (19). Also the standard deviation of about 5% found for the detection of 0.098% of Cu in aluminum is about the same as by x-ray fluorescence (18). Table I11 shows good agreement with other analytical techniques. Light Element in Light Matrix. T h e detection limits shown by Table I1 for the detection of Si and M g in aluminum are distinctly better than those found by Fox (6) (who also used direct electron excitation) and those obtained by conventional x-ray fluorescence. The determination of 2.45% of llIg in aluminum was obtained with a standard deviation of 0.6% (seven determinations). Good agreement with other analytical techniques was also obtained for Si analysis as illustrated by Table 111. The detected limit of sensitivity for N a determination in graphite is of the same order of magnitude as given by Dunne (3) using the Henke tube with a KAP crystal and I-micron carbon detector window. A spectrometer recording of a sample containing 0.5% Na is shown by Figure 4. Notable is the sensitivity for Ka K a and the appearance of K a K w , ~satellite.
General Operating Conditions
Curved Johann-type crystal optics, 100-second counting time Element N a' hIga Sia Pa Cob Cub Znb
a
b c
Excitation conditions kv. ma. 12 0.2 10 10 14
30
30
30
Windowless x-ray tube used. With a 0.1 mm. Be window 6 micron mylar window.
712
ANALYTICAL CHEMISTRY
0.2
0.2 0 1
0.1 0.1 0.1
Crystal Mica Flat Gypsum R = 500 mm. Gypsum R = 500 mm. GvDsum R = 750 mm. LTF R = 750 mm. LiF R = 750 mm. LiF R = 750 mm.
Detector and operating voltage Flowcounter" Flowcounterc 1500 FlowcounterC 1500 Flowcounterc 1500 Flowcountere 1500 Scint. counter 1100 Scint. counter 1100
Discrimination, volts Window Base level width 12 24 12 18 16 20 20 24 .. .. 5 35 5 35
P* scc
R
RG
t t coo
1500
6000'
500 1400
7000-
*Oo0
70 01
IOO~WOO
i
5000
60 1200
200 .I 100
6oo\ 4 00
50 1100
100 1000
30001
300 1200
-
I
0
2
I
6
8
10
I2
14
16
KY
18 20
Q Y)
h 38
37
i
3000
2000 ,nA
--L
ql
that by using the scintillation counter instead of the flow counter, the figure of merit increased by a factor of 3. This remark should be also valid for the Co Ka which is of the same order of wave length as the Ta Lal and involves a detection limit of about 10 p.p.m. Source of Error. CARBONCONTAMINATION. It is well known t h a t contamination of t h e specimen surface b y carbonaceous material appears during electron bombardment caused by rubber O-rings, high vacuum grease, oil of the diffusion pump, etc. This deposit was no serious problem in quantitative analysis up till Ns as has been shown. Quantitative analysis of carbon, however, should only be possible by using ion pumps and metal sealings or an ultra high vacuum system. The often reported statement that carbon contamination should be reduced to a negligible value by only maintaining the anode at elevated temperatures seems to be incorrect (11). FILAMEKT CONTAMINATION. With a rhenium filament, quantitative
Element Si Zn
35
Figure 4. Sodium K x-ray emission spectra of a graphite sample containing 0.5% N a . Scanning speed O.So/min. Full scale: 50 C.P.S.
9000..
42 93 0,4 45
016
47
O 0,
too0
49 .$O
Figure 3. Variation for the AI Ka line from a pure aluminum sample, of the peak intensity PI peak to background ratio R and figure of merit R d B with the electron current at a constant anode voltage of 10 kv.
Table 111.
36
6060.
4000.
Figure 2. Variation for the AI K a line from a pure aluminum sample of the peak intensity, peak to background ratio, R and figure of merit R d B with the anode voltage at constant electron current of 0.5 mA
Light Element in Heavy Matrix. T h e detection limit for P in bronze is also much lower as detected by Fox (6) and of the same order of magnitude as by x-ray fluorescence. Heavy Element in a Heavy Matrix. No better results t h a n b y x-ray fluorescence (1.2) were obtained in the analysis of Co in low alloy steel. However, an advantage of this technique remains that, owing to the curved crystal optics, the Co Ka line is completely separated from F e Kp, in the first order, so that no rather complicated correction is needed due to the overlapping of Co K a by Fe KO. A similar situation occurs in the Ta determination in Niobium (17) where it was found
t
cu
a
analysis gave a good precision. An experiment was performed by bombarding an aluminum sample with electrons after preheating the filament. B u t even after a n irradiation of about one hour, we could not detect the pres-
Table 11. Limits of Sensitivity for Zn, Cu, Co, PI Si, Mg, and N a in Different Matrices
Limit Background
Element Zn Cu Co
of
detection" (p.p.m.) 2
Matrix (c.P.s.) Aluminum 4400 Aluminum 3600 3 Low alloy 1600 30 steel 600 20 P Bronze Si Aluminum 240 0.7 10 11 Aluminum Mg Na Graphite 4 250 Defined as the concentration that yields an intensity equal to three standard deviations of the background intensity for a counting time of 100 seconds.
Comparison of X-ray Results Obtained with Other Analytical Methods
X-ray electron excitation, % 0.30 0.029 0.0035
X-Ray fluorescence, % 0.25 0.026 0,0045
Spectrophotometry,
70
0.20 0.027 0.051
Activation analysis ND" 0.028 0.005%
ND: not determined.
VOL. 30,
NO. 6, MAY 1966
e
713
and Campbell (16). Considering the peak to background ratio, the best value found for magnesium, of about 700 was better than the figure of 100 obtained by Wyckhoff (21) who also used a windowless x-ray spectrometer.
10090
-
80 : 70
LITERATURE CITED
-
60-
Z
=2I 5 0 4 I-$
40-
(1) Campbell, W. J., Brown, J. D., ANAL.CHEM.36, 312 R (1964). (2) Dolby, R., Brit. J . Appl. Phys. 1 1 , 64 (1960). (3) Dunne, J. A., Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, 1964. ( 4 ) Fonda, G. R., J . Am. Chem. SOC. 55. 55, 123 (1933). __,_ (5) Fonda, E'onda, G. R., R., Collins, G. B., J . ~ Am.m . Chem. SOC.53, 113 (1931). (6) Fox, J. C. hl., J . Znst. M e t . 91, 239 ( 1962-1 963 ), ( 7 ) Glocker, It., Schreiber, H., Ann. Phys. 85, 1089 (1928). ( 8 ) Henke, B. L., Bdvan. X-Ray .4naE. 6 . 3- 6- -2 (1962). f1962) 6,362 (9) Henke, B. L., White, R., Lundberg, B., J . A p p l . Phys. 2 8 , 98 (1957). (10) Hevesy, Von, G., Coster, D., Nature 111, 79 (1923). (11) Holliday, J. E., J . Appl. Phys. 33, 3259 11963). (12) Michaelis, R. E., Alvarez, R., Kilday, B. A., J . Res. Nut. Bureau of Stds. 65'2-1, 71 (1961). (13) Pechiney. Centre de Recherches hfetallurgiques de Chambery, France, 1962. (14) Spielberg, N., Bradenstein, M., Appl. Spectr. 17, 6 (1963). (15) Stoddard, K. B., Phys. Rev. 46, 837 j -
ep 30 20 -
- 7
10
-
n -
~
I
H He L i
I
I
I
Be'B C
I
I
N 0
I
I
I
I
I
F Ne NaMg AI
I
I
SI
P S C1
I
I
I
Figure 5. Calculated x-ray transmission characteristics of different detector-window materials in the 4 to 70 A. region
ence of traces of rhenium on the surface of the aluminum sample by x-ray fluorescence analysis. As a precaution during the analysis, the samples were left the same time under the electron bombardment. Nevertheless a complete elimination of the eventual rhenium contamination is possible by deflecting the electron beam from a helical rhenium cathode situated below the anode, as used by Henke (8). DISCUSSION
Quantitatire determination of low atomic number elements by direct electron excitation has given good results. The limits of detection of light elements in various matrixes are better than those found by x-ray fluorescence and of the same order of magnitude using the Henke tube associated with very thin detector windows and recently developed analyzing crystals. However, no liquid or volatile samples can be analyzed by this method as comparing by the two latter techniques. Further improvements in lowering the limit of detection can be obtained in the future by various technics1 modifications. Recently developed crystals have stronger reflecting power and preliminary studies on the Na K a of an XaC1 sample, have shown (with a KAP instead of a mica crystal) a n increase of
714
ANALYTICAL CHEMISTRY
the peak to background ratio by a factor of 2 and of 4 for the figure of merit. dlso, new thinner window materials are very promising as for example nitrocellulose of 1.000 to 2.000 A. thickness. Figure 5 indicates what improvements can be expected from various window materials, their mass absorption coefficients being calculated from Victoreen (20) and Henke's (9) data. It is possible to eliminate the window by using windowless photomultipliers such as the Bendix h l 306. But some preliminary work with this photomultiplier har shown a peak to background ratio of only 3 for the A1 K a line. By the use of direct electron spectrograph as a windowless x-ray tube spectrometer some experiments were performed to compare the two techniques in the case of the K a ,Mg line from a pure magnesium sample. For the fluorescence excitation three different anode materials, Cr, hIo, and W, were employed. Results showed obviously the superiority of electron excitation with a figure of merit of 3.200. A t the other hand the hlo anode was a little more efficient than Cr and W for the excitation of magnesium, the figures of merit being for Mo, 970, for W, 920, and for Cr, 840. The same order of efficiency was found for aluminum excitation by these three anodes, in contrast to the results given by Thatcher
1 - - - - 1 .
i 1924) j - l _ -
(16) Thatcher, J. W., Campbell, W. J., Advan. X-Ray Anal. 7, 519 (1964). (17) Toussaint, C. J., Vos, G., Anal. Chim. Acta. 33, 279 (1965). (18) VOS, G., Toussaint, C. J., European $1. Energy Commun. EUR-21 f (1962). (19) Vos, G., Tonssaint, C. J., 26th Congress of GAMS, Paris, 1964. (20) Victoreen, J. A., J . Appl. Phys. 20, 1141 (1949). (21) Wyckhoff, W. G., Davidson, F. D., Rev. Sci. Znstr. 35, 381 (1964). RECEIVEDfor review April 15, 1965. Accepted February 15, 1966. Work performed in the field of the ORGEL Program, presented at the Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, hlarch 1-5 (1965).
Correction Improved Extraction Method for the Isolation of Trivalent Act inide-Lan t ha n ide Elements from Nitrate Solutions I n this article by Fletcher L. Moore [ANAL. CHEM.38, 510 (1966)) three errors appear on page 511 in column headings for Table 11. The headings for columns 3, 7 , and 10 should read z44Cm, 148Pm,and l7oTm, respectively.