LITERATURE CITED
Finally, a vote among the three and five most similar spectra can be taken and the unknown predicted to belong to the class with the majority of near neighbors. The results of the 3 and 5 most similar spectra predictions are also tabulated in Table IV for each approach discussed in this paper. As has been found in other investigations (7), the best results are obtained from predictions using only the one most similar spectrum.
(1) E. Fix and J. L. Hodges, Jr., "Discriminatory Analysis, Non-Parametric Discrimination, Consistency Properties," USAF School of Aviation Medicine, Randolph Field, Texas, Project 21-49-004, Rept. 4, Feb. 1951. (2) E. Fix and J. L. Hodaes. Jr.. "Discriminaton, Analvsis: , ~Small - Samole Performance," USAF- School of Aviation Medicine, Randolph-Fieid, Texas, Project 21-49-004, Rept. 11, Aug. 1952. D. 0. Loftsgaarden and C. P. Quesenberry, Ann. Math. Stat., 36, 1049 (1965). T. M. Cover and P. E. Hart, lfff Trans. Inf. Theory, IT-13, 21 (1967). T. M. Cover, I f f € Trans. lnf. Theory, IT-14, 50 (1968). B. R. Kowalski and C. F. Bender, Anal. Chem., 44, 1405 (1972). J. B. Justice and T. L. Isenhour, Anal. Chem., 46, 223 (1974). M. A. Pichler and S.P. Perone, Anal. Chem., 46, 1790 (1974). G. L. Ritter, H. 6 . Woodruff, S. R. Lowry, and T. L. Isenhour, "An Algorithm for a Selective Nearest Neighbor Decision Rule," to be published in / € E Trans. lnf. Theory, November 1975. H. B. Woodruff, S.R . Lowry, and T. L. Isenhour, "A Comparison,,of Two Discriminant Functions for Classifying Binary Infrared Data, Appl. Spectrosc., 29 (3),226 (1975). D.J. Rogers and T. T. Tanimoto,,Science, 132, 1115 (1960). S.R. Lowry and T. L. Isenhour, A Feature Selection Technique for Binary infrared Spectra." submitted to J. Chem. lnf. Comput. Sci. ~
CONCLUSIONS Two similarity measures have been discussed concerning the classification of binary spectra. For each case, the Tanimot0 similarity measure gave improved overall classification over the conventional technique of measuring distances. For the overall prediction results, the improvement ranged from 0.7 to 3.8%.While this does not mean that one should exclusively use the Tanimoto measure instead of a distance measure, the fact that the former technique gave superior results in all cases is noteworthy. In addition, the fact that eliminating the less important background peaks resulted in improved classification is very satisfying. Not only do reduced features give better results, they do it more rapidly since fewer machine operations are involved.
RECEIVEDfor review December 16, 1974. Accepted July 7, 1975. T. L. Isenhour is an Alfred P. Sloan Fellow, 1971-75. The financial support of the National Science Foundation is gratefully acknowledged. Presented in part at the 169th ACS National Meeting, Philadelphia, Pa., April 7, 1975.
X-Ray Fluorescence Spectrometric Determination of Gold, Bromine, and Iodine in Biological Fluids J. T. Purdham and 0. P. Strausz Department of Chemistry, University of Alberta, Edmonton, Alberta, Canada
K. 1. Strausz Toxicology-ProvincialAnalyst Laboratory, Alberta Department of Agriculture, Edmonton, Alberta, Canada
X-Ray fluorescence spectrometry is a powerful tool for the determination of elements in a wide variety of sample matrices. The method can be adapted to trace element analysis if matrix absorption and enhancement effects are minimized or held essentially constant. This can be achieved by the use of dilute matrices (1) or alternatively by precipitating the element of interest in the form of a thin film on a suitable substrate (2). Deutsch ( 3 ) recently described the use of X-ray fluorescence for the direct determination of bromine in sea water, and Rose and Cuttita in saline water ( 4 ) . Lund and Mathies ( 5 ) have determined serum bromine by evaporating serum on a filter paper, leaving a thin film of solid behind. Wahlberg and Myers also reported a similar concentration method (6). Natelson and Sheid determined protein-bound iodine in serum using a concentration technique (7). In the present study, we found that for biological fluids such as serum and urine, solvent evaporation is unnecessary at the concentration levels commonly encountered with drug metabolites, and the bromine and iodine concentrations can be measured directly in the sample, thereby significantly reducing analysis time. We also wish to report here a rapid and relatively simple method for the determination of gold in serum. The need for this has arisen from the use of gold salts in the treatment of arthritis which requires careful monitoring of gold levels in the serum. These levels are normally too low for 2030
direct determination, and we employed sulfide precipitation, after digestion, as a means of concentration, as originally suggested by Luke (8). EXPERIMENTAL Apparatus. A Philips PW 1540 spectrometer with a silver target tube was used, employing excitation conditions of 50 kV and 20 mA. Isolation of the Br K a , I K a and Au La lines was achieved by LiF (200) crystal dispersion together with appropriate pulse height selection. A sodium iodide fluorescent crystal scintillation counter served as detector. Standards. Stock solutions of KBr containing 0-24 mg/100 ml Br-, NaI containing 0-65 mg/100 ml I-, and AuC13 containing 1 ppm Au3+ were used for standardization. Procedure. The Br K a line was measured a t 28 = 29.92' with background intensity taken a t 29.0' and 30.84', all measurements being made for 100 sec. The iodine line was measured a t 28 = 12.375' and the background a t 13.1' for 100 sec. The Au concentrations were determined by measuring the Au La line intensities a t 28 = 36.95', for four minutes. The background was taken for the same time a t 20 = 36.0'. All peak intensities were corrected for background radiation.
RESULTS AND DISCUSSION Bromine a n d Iodine. Standard calibration curves were obtained in distilled water solution for these two elements. The calibration curve of net Br K a counts/100 sec vs. mg/ 100 ml Br, is linear, passes through the origin, extends to 400,000 counts for 24 mg/100 ml Br, and is defined by the
ANALYTICAL CHEMISTRY, VOL. 47, NO. 12, OCTOBER 1975
~
~
_
_
_
~~
_
~
Table IV. Determination of Known Amounts of Iodine i n Serum
Table I. Determination of Known Amounts of Bromine in Urine Concentration of bromine mg/100 m i
Veasued concentration, mg/100 m l
3.0 6.0 9.0 13.23 18.52 23.81 a Average 98.9 f 2.3%.
3.10 5.84 8.91 12.95 18.03 23.37
Recober),
Concenixation of iodine mg/ 100
a
lieasured concentration, mg!100 m l
4.15 4.15 6.98 6.98 14.24 14.24 20.70 Average 98.8 f 1 4%.
4.11 4.03 6.90 7.10 14.04 14.00 20.32
mgI100 m l
8.0 8.0 16.0 16.0 24.0 24.0 33.17 33.17 46.98 46.98 a Average 99.6
103.3 97.4 99.0 97.9 97.4 98.2
Table 11. Determination of Known Amounts of Bromine in Serum Concentration of b r o m i n e rng:100 m i
Measured Concentration,
d1
loa
Recovery, 5 5 a
Recovery,
7.70 7.90 15.70 16.06 24.82 24.50 33.36 32.68 46.53 46.51
* 2.1%.
96.3 98.8 98.1 100.4 103.4 102.1 100.6 98.5 99.0 99.0
Table V. Analysis of Bromine-Iodine Mixtures in Aqueous Solution
99.0 97.1 98.9 101.7 98.6 98.3 98.2
Concentration, mgI100 m l
Br
I
Measured, mgi100 m l
I
Br
I*
Br a
8.66 6.15 102.5 6.0 8.44 54.82 5.95 53.48 99.2' 6.0 11.95 7.88 99.6 8.44 12.0 53.54 11.95 99.6 12.0 54.82 a Average 100.2 f 1.57~.Average 97.8 f 3.870.
-
._
Recovery,
102.6 97.6 93.4 97.7 ~
Table 111. Determination of Known Amounts of Iodine in Urine Concentration u! iodine mgi 100 41
5.06 8.44 16.88 27.41 54.82 a Average 100.2 f 6 . 8 7 ~ .
\leasued concentration, mqI100 m l
4.53 8.90 17.47 27.18 56.67
Reco\er),
Table VI. Recovery of Added Gold from Serum Amount of Au added, u g
a
89.5 105.5 103.5 99.2 103.4
equation, y = 1 6 2 4 1 ~- -178. All nine calibration points, ranging from 0 to 24 mg/100 ml Br lie directly on the line. 'l'he calibration curve of net I Kcu counts/100 sec is also linear, passing through the origin, extends to -260,000 counts for 64 mg/100 ml I and is defined by the equation, y = 234 4092x. All nine calibration points ranging from 0 to 64 mg/100 ml I lie directly on the line. The linearity of both curves indicates that, over these concentration ranges, the dilution effect is sufficient to overcome matrix effects. Since both serum and urine are light element matrices, we assumed that the calibration lines produced from aqueous solution could he used for bromine and iodine measurements in these media. The assumption was tested and found to be valid by adding known amounts of Br- and Ito samples of serum and urine and comparing the results of the analysis using the calibration curves with the known prepared concentrations. The results of these experiments are summarized in Tables I-IV. Recoveries were generally within f 3 % of the amount added. Table V, which gives the results on solutions containing both iodine and bromine, points to another advantage of the method, namely, that these two elements can be measured in the presence of each other without interference. Bromine and iodine measurements are simply carried out by placing 8 ml of sample (6 ml produces the critical depth), into a sample holder, measuring the appropriate line, correcting for background, and obtaining the concentration from the calibration lines.
+
Amount of A u measured by X.R.F., u ¶
2.0 2.0 2.0 2.0 5.0 5.0 5.0 5.0 7.0 7.0 7.0 7.0 10.0 10.0 10.0 10.0 a Average 97.9 f 6.5% from
Recovery, -6"
94.5 93.5 101.5 106.0 92.8 103.0 104.8 105.0 102.8 99.1 93.7 101.9 79.2 96.3 95.2 96.4
1.89 1.87 2.03 2.12 4.64 5.15 5.24 5.25 7.20 6.94 6.56 7.13 7.92 9.63 9.52 9.64 16 measurements ~
~
~~
Gold. Two hundred bg of Te are added to 3-5 ml of the serum sample, which is digested using 10 ml concentrated nitric acid to oxidize the organic material. The digest is taken just to dryness and 10 ml of dilute aqua regia (4:1:5 HCl:HNO$:H*O) are added and again the solution is evaporated just to dryness. Finally, the residue is dissolved in concentrated hydrochloric acid and heated a little longer to ensure that all the nitric acid and oxides of nitrogen have been driven off. The solution obtained is diluted to 30-40% in hydrochloric acid and heated to approximately 7 0 OC. Hydrogen sulfide gas is bubbled through for ten minutes and the precipitate produced is filtered through a 1.2-bmpore filter membrane using a millipore filter apparatus. The precipitate is dried for 1-2 hours in air and placed in a sample holder. A 1.5-mm thick Teflon disc is placed over the membrane to keep it flat and the Au La line intensity is measured.
ANALYTICAL CHEMISTRY, VOL. 47, NO. 12, OCTOBER 1975
2031
Standards were produced by precipitating known amounts of Au in the 0- to 12-pg range as the sulfide. Measurement of the Au La line was made as described above. The calibration curve produced was linear, passes through the origin, extends to -28,000 counts for 10 pg Au and is defined by the equation, y = 2838x - 33. All eight calibration points lie close to the line. The linearity suggests that the precipitation of Au is complete. T o ascertain whether gold is lost during the digestionprecipitation-filtration processes, an experiment was carried out to test the recovery of gold. Various amounts of gold were added to serum samples and the samples were then put through the complete procedure. The results of these recovery experiments are presented in Table VI. In general, the recovery was reasonably good with the majority of the results clustering within f 6 % of the correct value. The results show that X-ray fluorescence analysis can be used as a rapid and accurate method for bromine and iodine analysis. The method for gold analysis is somewhat slow due to the necessity for digestion of the sample prior to analysis, but is a relatively simple procedure. The theoretical minimum concentration of bromine that can be analyzed by this method, taken as the concentration giving a peak which has a net intensity of three times its standard deviation as calculated from counting statistics, is 0.1 mg/100 ml. For iodine the minimum is 0.8 mg/100 ml. Both can be lowered by using longer counting times. However, in the case of bromine, this is not really necessary as the limit is already below the range of bromide levels normally found in serum, estimated as 0.25-2.0 mg/100 ml (as NaBr) (9). Levels above this occur from drug administration. Intoxication occurs a t levels above 150 mg/100 ml and more serious toxicosis above 250 mg/100 ml (10). Samples containing bromide a t levels above 24 mg/100 ml require dilution with water before measurement, and then the sample matrix even more closely approximates that of the standards. The normal levels for iodine in serum are in the range 4.8-8.6 pg/lOO ml ( 1 1 ) which is well below the detectable limit of this system. Ours is not the best fluorescence sys-
tem for iodine measurement. The choice of an Ag target means that the iodine K a line, occurs in the region of highest background intensity. Also, only the higher energy continuous radiation can be used to excite the iodine K a line since the intense Ag K a and KP lines are too low in energy. However, we still find the method useful for measuring levels of iodine-containing compounds in spinal fluids, where they have been used as X-ray contrast media. The gold analysis suffers the disadvantage of requiring digestion before measurement, unlike an earlier atomic absorption method (12). But it is still a relatively simple procedure and, with a detection limit as previously defined of 0.2 pg, is reasonably sensitive and, as can be seen from Table VI, quite accurate. The lower limit of detection is equivalent to 0.05 ppm on a 4-ml sample or 5 pg/lOO ml, which compares very favorably with the atomic absorption study and is well within the clinical range of interest, -100 pg/lOO ml (12). As in the previous study, the accuracy of measurement a t or just above the limit of detection can be improved by addition of known amounts of gold. The standard deviation in a sample containing 1 pg Au is f 0 . 2 pg while for 10 pg, it is f0.25 pg.
LITERATURE CITED (1)F. Claisse, Que. Dep. Mines Preliminary Rep., 402 (1960). (2)E. L. Gunn, Anal. Chem., 33, 921 (1961). (3)Y . Deutsch. Anal. Chem., 46, 437 (1974). (4)H. J. Rose, Jr., and F. Cuttita, Adv. X-Ray Anal., 11, 23 (1968). (5) R. K. Lund and J. C. Mathies, Norelco Rep., 7 , 134 (1960). (6)J. S.Wahlberg and A. T. Myers, U.S. Geol. Surv., Prof. Pap. No. BOO-D, 214 (1968). (7)S.Natelson and B. Shied, Clin. Chem., 8, 17 (1962). (8)C. L. Luke, Anal. Chim. Acta, 41, 237 (1968). (9)G. A. Harrison, "Chemical Methods in Clinical Medicine", J. and A. Churchill Ltd., London, 1957,p 358. (IO)E. Rentoul and H. Smith, "Glaisters Medical Jurisprudence and Toxicology", 13th ed., Churchiii Livingston, Edinburgh and London, 1973,p 558. (11) "Handbook of Biological Data", National Acad. of Sciences, Nat. Res. Council, W. B. Saunders and Co., Philadelphia and London, 1956,p 52. (12)A. Lorber, R. L. Cohen, C. C. Chung, and H. E. Anderson, Arthritis Rheum., 11, 170 (1968).
RECEIVEDfor review March 31, 1975. Accepted June 9, 1975. The authors are grateful to the National Research Council of Canada for partial financial support.
X-Ray Fluorescence Spectrometric Determination of Selenium in Biological Materials K. 1. Strausz Toxicology-Provincial Analyst Laboratory, Alberta Department of Agriculture, Edmonton, Alberta, Canada
J. T. Purdham and 0. P. Strausz Department of Chemistry, University of Alberta, Edmonton, Alberta, Canada
Selenium has been shown to be an essential nutrient (1) and deficiencies in animal diets lead to serious diseases. At the same time, selenium levels slightly above the nutritional requirement are highly toxic and the permissible intake level has been estimated to lie in the range 0.5 to 3.5 pg/g of feed (2). The recognition of the biological role of selenium and the critically narrow concentration range between deficiency and toxic levels has given stimulus to the development of new sensitive analytical methods for the determination of this element. 2032
The techniques most commonly used for selenium determination are neutron activation analysis ( 3 ) , fluorometry (4, 5 ) , atomic absorption (6, 7) and spectrophotometry (8). The first method has the obvious disadvantages of being expensive and time-consuming, while the last three often require elaborate, multi-step procedures and are subject to interferences. We wish to report here a rapid and simple method for the analysis of selenium in biological matrices using X-ray fluorescence spectrometry. The technique is not sensitive
ANALYTICAL CHEMISTRY, VOL. 47, NO. 12, OCTOBER 1975