1248
Anal. Chem. 1980, 5 2 , 1248-1251
is -0.12 cm2, t h e amount of material on the contacts is estimated as being equivalent to between 8 and 80 monolayers. I n summary, pyrolysis of dimethyl silicones reproducibly yields volatile fragments readily detected using GC/MS techniques. The capability of detecting subnanogram amounts suggests that in many situations silicone contamination can be identified before contamination related failures have occurred. These same techniques should facilitate the identification of silicones in environmental samples and perhaps promote t h e monitoring of these synthetic compounds as indicators of man's influence on his environment.
ACKNOWLEDGMENT We thank N. M. Kitchen, C. A. Russell, and B. T. Reagor for helpful discussions throughout the course of this work. LITERATURE CITED (1) W. J. Storck, Chem. Eng. News, 56 (32),8 (1978). (2) W. J. Storck, Chem. Eng. News, 57, (19), 22 (1979).
(3) H. W. Fox, P. W. Taylor, and W. A. Zisman, Ind. Eng. Chem., 39,1401 119471. (4) N.-M: Kitchen and C. A. Russell, "Silicone Oils on Electrical ContactsEffects. Sources and Countermeasures". in Proceedinas of the Twenty-First 'Annual Holm Seminar on Electrical Contacts, Ill%ois Institute of Technology, October 14-16, 1975. (5) R . Pellenberg, €nviron. Sci. Technol., 13, 565 (1979). (6) J. H. Beynon, Mass Spectrometry and Its Applications to Organic Chemistry", Elsevier, Amsterdam, 1960, Chapter 10. (7) M. G. Voronkov, V. P Mileshkevich, and Yu. A. Yuzhelevskii, "The Siloxane Bond", Plenum Publishing, New York, 1978, Chapter 3. (8) E. Stenhagen, S. Abrahamsson, and F. W. McLafferty, "Registry of Mass Spectral Data", John Wiley and Sons, New York. 1974. (9) M. F. Hunter, J. F. Hyde, E. L. Warrick, and H. J. Fletcher, J. Am. Chem. Soc., 66, 667 (1946). (IO) T. H. Thomas and T. C. Kendrick, J. Poiym. Sci., Part A-2, 7 , 537 (1969). (1 1) W. L. Budde and J. W. Eichelberger, "Organics Analysis Using Gas ChromatographylMass Spectrometry", Ann Arbor Science, Ann Arbor, Mich., 1979.
RECEIVED for review December 26, 1979. Accepted March 7 , 1980.
Determination of Silver in Silicate Rocks by Furnace Atomic Absorption Spectrometry John C. Eames" CSIRO Division of Mineralogy, North Ryde, N.S.W. 2 1 13, Australia
Jaroslav P. Matousek Department of Analytical Chemistiy, The University of New South Wales, Kensington, N.S. W. 2033, Australia
The silver content of two quartzite rocks was measured directly uslng a graphite cup furnace. The effects of particle size and calibration procedure on the precision and accuracy of the determination were investigated. The reduction of particle size to less than 10 pm significantly reduced the sampling error to yield an RSD of 8.1 %. The accuracy of the determination was also increased, probably because of the Improved release of the analyte during atomization. A large matrix interference was observed, but it was possible to overcome this partially by the use of a standard addition technique.
T h e direct atomization of powdered sample is one of the most challenging problems in atomic absorption spectrometry (AAS). There can be advantages in introducing the sample into the furnace as a powder rather than in solution. T h e laborious and time-consuming decomposition steps, dry ashing, wet digestions, and extractions can be omitted. Often no reagents are needed and consequently no contaminants are introduced. Early work by L'vov ( I ) described a device in which the sample was atomized from an independently heated graphite rod electrode combined with a graphite furnace which had t h e capability of directly atomizing solids. Several later approaches used flame as an atomization medium, and included premixing the powdered sample with solid propellant ( 2 , 3 ) , nebulization of suspensions ( 4 , 5 ) ,and delivery of the powder to the burner channel by a miniature spiral conveyor (6). The methods had limited success.
More recent furnace work has relied on various modifications of the furnace and furnace-in-flame-type atomizers, such as the graphite rod with a n air/acetylene flame, vertical cup, and horizontal tube furnaces with resistance heating or radiofrequency heating, the capsule-in-flame atomizer, circular-cavity furnace, and graphite microprobe (7). Most of t h e work reported on the analysis of solids using AAS has utilized specially constructured furnaces. In the past, workers have made only brief mention of possible factors affecting the accuracy and precision of the analysis of powdered samples. In this paper, factors affecting the accuracy and precision of the direct determination of silver in silicate rocks are investigated using conventional instrumentation with a view to applying this technique to routine types of analyses.
EXPERIMENTAL Apparatus. Experiments were performed on a Varian Techtron AA5 atomic absorption spectrometer equipped with a model 90 CRA furnace and a Mace FBQ-100 chart recorder. High purity nitrogen sheathed the pyrolytic graphite cup furnace. Both Ultra-Carbon and Ringsdorff graphite cups were employed, and Ringsdorff RW-C spectroscopic graphite was used for sample dilution. Furnace temperatures between 1000 and 3000 OC were measured using an Ircon Series 6000 radiation thermometer (Ircon Inc., Skokie, Ill. 60076). A chromel-alumel thermocouple was used for the lower temperature range. Electron micrographs were taken with a Cambridge S600 scanning electron microscope incorporating an EDAX X-ray analyzer. Sample Preparation. Reference quartzite rocks G246 and G248 were supplied by the NSW Department of Mineral Re-
0003-2700/80/0352-1248$01.00/0C 1980 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 52, NO. 8, JULY 1980
m
1249
. .
/
C3V
7 :
"
1 c
I 1 L
F
10 KEl.H1
'C
3
3 Y
'3
Figure 1. Plot of absorbance vs. sample weight after grinding sample-graphite mixture in a Grindex mixer/mill for 5 min sources and Development (Australia) and were ground to particle sizes of approximately 50 pm. Sample-graphite mixtures were ground and homogenized in a Grindex Series 2 mixer/mill and a Retsch Spectro mill. The Grindex Series 2 mixer/mill is a high-speed impact swing mill with agate container and pulverizing balls. During operation, the balls reciprocate against the sample. The agitation can be represented by components in three mutually perpendicular directions. The Retsch Spectro mill is an 8-mL capacity agate mortar and pestle. To reduce particle sizes to the required fineness, the mortar was rotated a t 400 rpm and the pestle a t 1500 rpm in a 12-min operation. General Procedure. For the determination of silver, 1-to 2-mg aliquots of sample mixed with graphite in a ratio of 1 : l O were weighed to kO.001mg into tared graphite cups, and analyses were performed in batches. For direct analysis of complex inorganic solids, mixing the sample with graphite is often necessary in order to (i) dilute the analyte to a level suitable for furnace AAS; (ii) prevent the ashing residue from melting or sintering into a bead, thus allowing more efficient removal of interfering components; (iii) help eliminate chemical interference problems by dispersing the sample in a reducing environment. Solutions were introduced into the cup with a Hamilton micropipet incorporating a Chaney adaptor and disposable Teflon tips (Diagnostic Division, Pfizer Inc., New York, N.Y.). The graphite mixture was dried for 15 s a t a selected voltage so that the graphite cup a t the end of drying reached a temperature of 110 "C. An ashing temperature of 800 "C for 15 s was sufficient to remove all volatile matrix components without loss of analyte. The final atomization temperature of 1800 "C with a hold time of 3.5 s ensured that all silver was completely atomized from the silicate samples; a ramp rate of 600 "C s-' gave maximum sensitivity. All absorbances were evaluated from peak height measurements at the most sensitive Ag 328.1-nm line. The level of background absorption was checked with a hydrogen continuum lamp and found to be negligible. The silver content of the two quartzite reference rocks was determined using a perchloriehydrofluoric acid digestion followed by both anodic stripping voltammetry and solution furnace AAS methods of analysis. The anodic stripping voltammetric procedure given by Jacobs (8)and Perone (9) was used. This involved plating silver onto a wax impregnated carbon electrode for 5 min with the potential of the carbon electrode held at -0.3 V vs. a standard calomel electrode (SCE). After 15 s, when the deposition current had decayed to a constant value, a linearly varying anodic scan from -0.3 to +0.7 V vs. SCE was applied to the electrode a t a rate of 5 mV s-'. During the scan, the potential change and corresponding anodic stripping current were recorded. This technique is applicable to the analysis of geochemical samples. Sample solutions were also analyzed for silver content in a graphite tube furnace using the operating conditions specified for solid sample analysis. A standard addition technique was used to overcome possible chemical interferences, and background interferences were eliminated during the ashing cycle. Lines of best fit shown in Figures 1,2, and 3 and line I in Figure 4 were obtained via a regression analysis which constrained the line to pass through the origin. Results obtained by both methods
00 10 15 20 SAMPLE WE>GHT I w g l
05
25
Figure 2. Plot of absorbance vs. sample weight after grinding sample-graphite mixture in a Grindex mixer/mill for 10 min
J'.
9a 0 3 m U
0.2
c 5
IO SAMPLE
15 WEIGtI
2 3
2 5
(mgi
Figure 3. Plot of absorbance vs. sample weight after grinding sample-graphite mixture in a Retsch Spectro mill
05
00
0 5
10
2c
15
S A M P L E WEIGHT
25
(mgi
Flgure 4. Experimental plot of absorbance vs. sample weight for sample-graphite mixture ground in a Retsch Spectro mill for 12 min without (line I) and with (line 11) the addition of the analyte element were in good agreement and the silver contents of samples G246 and G248 were established as 2.3 and 3.8 pg g-' respectively.
RESULTS AND DISCUSSION Effect of Particle Size. Several workers, notably Wilson ( I O ) , have discussed and calculated the size of a laboratory sample required for determining a n element in relation to:
1250
ANALYTICAL CHEMISTRY, VOL. 52, NO. 8, JULY 1980
Table I. Determination of Silver in Silicate Sample (2248 no.
mean
of
silver
yses
p g g-’
std. dev.
RSD,
type and grinding time
anal- concn., Grindex mixerimill, 5 min Grindex mixerimill, 10 min
38
1.25
0.18
14.4
39
1.23
0.14
11.2
Retsch Spectro mill
49
1.50
0.12
8.1
%
the accuracy of the analytical method, the abundance of the element, and the degree of fineness of the powdered rock sample. Trace elements in geological materials are either distributed evenly throughout the lattice of constituent minerals or concentrated in a minor mineral. T h e relative sampling error will be smaller in the former case. Using the Wilson modification of Benedetti-Pichler’s ( 1 1 ) application of the Bernoulli theorem of sampling a binary system, a 10-mg sample would require particles to be about 50 ,um, while smaller samples necessitate further reduction in particle size in order to obtain representative sampling. Graphite samples ground in the Grindex mixer/mill and the Retsch Spectro mill were examined microscopically in order to determine the size of constituent particles. The mixture ground in the Grindex mill contained particles predominantly in the size range 50 to 10 pm. Assuming the particle density to be 2.3 g ~ m - the ~ , number of particles for the two limiting cases for an arbitrary sample size can be calculated. For example, for a 1-mg aliquot and 50-pm particles, the number is 7 x lo3;for 10-pm particles it is 8 X lo5 particles. According to Wilson ( I O ) , in order to achieve desirable homogeneity, it is necessary for a sample to contain approximately lo5 particles. Therefore, unless the majority of particles in the aliquot is reduced in size to the lower limit, large sampling errors will occur if the analyte is segregated in the sample material. Grinding in the Retsch Spectro mill reduced particle sizes t o 5 pm, which, assuming the above density, would give 7 X lo6 particles in a 1-mg sample. This would reduce the sampling error to a very low level. Some workers, using aqueous solutions, have successfully analyzed solid samples directly to construct a calibration curve (7, 12-14). Brady e t al. (15) found t h a t both an aqueous standards calibration and a standard addition technique gave equivalent results for zinc in seabottom sediments. This indicates that, in this case, the matrix did not interfere during atomization. Likewise, Katskov et al. (13,14), using the two calibration techniques, obtained similar results for a number of trace metals in various solids including rocks, metals, and slags. Aqueous standards were also used by L’vov (12) to analyze for trace metals in boron nitride, graphite, and zirconium oxide. Aqueous standard was initially used to construct a calibration curve in the present work. Results of analysis using this calibration approach are presented in Table I and indicate t h a t the silver content for sample-graphite mixes ground in both the Grindex mixer/mill and the Retsch Spectro mill was less than half the established value of 3.8 pg g-l. The accuracy of the determination was not significantly improved by extended grinding of the sample in the Grindex mixer/mill although, as illustrated in Figures 1 and 2, there was an improvement in precision. The fine state of subdivision of the sample produced by the Retsch Spectro mill led to a significant improvement in precision of the silver determination (Figure 3) due to considerably reduced sampling error. At the same time, the accuracy also increased, probably as a consequence of improved release of analyte during atomization.
Wm
SAMPLE
WEIGHT,
W
Figure 5. Theoretical plot of absorbance vs. sample weight without (line I) and with (line 11) the addition of the analyte element for the standard addition method of callbration. W,,, = weight of sample equivalent to weight of element
Even though the sample mixes had been ground extremely fine, reducing the particle size to the order of 5 pm, and even though the relative standard deviation was acceptable, the silver content was always less than half the established value. The same difficulty was experienced by Langmyhr and Thomassen (16) who obtained low results for cesium and rubidium in silicate rocks using aqueous standards. Consequently, Langmyhr e t al. (17) later used solid standards for the determination of lead, silver, and zinc in silicate rocks. The results in Table I show that, despite the addition of graphite powder t o samples, chemical interferences were not completely eliminated. For this reason, the use of a standard additions method of calibration was investigated. Effect of Method of Standardization. The standard addition technique usually involves adding equal volumes of standards of various concentrations to equal aliquots of sample. However, in the present work, small sample weights could not be readily reproduced and the following alternative scheme was devised. The same volume of standard aqueous solution was added to various accurate weights of solid. This approach is illustrated in Figure 5 . Line I denotes the relationship between W , the weight of sample taken, and A , the measured absorbance produced by atomization of an element in the sample. With addition of the same amounts, m, of the element to varying weights of sample, the absorbance-weight relationship shifts to line 11. Line I1 extrapolated to zero absorbance meets the weight axis a t W , and the absorbance axis a t A,. The equation of line I1 through A,,, is
A=aW+A,
(1)
where a is the slope. W , gives a measure of the sample weight equivalent to the weight m of element added. Then the fraction of element in the sample, K , is
K = m/W, (2) and since m is known and W , can be measured, the analyte concentration can be calculated. Figure 4 illustrates a practical example of the standardization procedure. There is no significant difference in the relative standard deviation for deviations for the two lines. Results in Table I1 confirm that greatly improved accuracy with good precision was obtained as compared with the approach using the aqueous standards calibration. However, the standard addition technique was not entirely successful, since a discrepancy remained between the results shown in Table I1 and the corresponding established values. For the graphite cup furnace, where the absorbance is measured outside the highest temperature zone, the residual matrix effect is not totally unexpected. Further studies are needed to determine the nature of this effect. However a practical solution of the remaining problem could be the use of solid standards which have been employed successfully in the de-
ANALYTICAL CHEMISTRY, VOL.
52!,NO. 8,
JULY 1980
1251
Table 11. Determination of Silver4 method of no. of standardization analyses aqueousC standards standardC addition standardd addition
mean silver concn pg g-'
std. dev.
RSD, %
49
1.50 ( 3 . 8 )
0.12
8.1
17
3.30 ( 3 . 8 )
0.30
9.1
13
2.03 ( 2 . 3 )
0.18
8.9
Particles ground t o approximately 5 pm. Established value (ppm) in parentheses. Sample G248. Sample G246. Figure 7. Residual siliceous agglomerate formed during atomizing and cooling cycles (SEM)
Flgure 6. Breakdown of the pyrolytic coating of a graphite cup and exposure of underbing porous graphite after analysis of silicate samples (SEMI
termination of silver in silicate rocks with an induction-heated furnace ( I 7). Deterioration of Graphite Cups. When 3 X g Ag was atomized in cleaned graphite cups that had previously been used for analysis of powdered silicate samples, an intermittently nonreproducible absorbance signal and an overall reduction in absorbance of 3MO% were recorded. However, such effects were not noted for those cups used only for aqueous solutions. The variations in absorbance noted above occurred after approximately five silicate firings. Other workers have noted that pitting and exfoliation of the pyrolytic graphite surface inside the graphite furnace led to absorption of elements in the underlying porous substrate, resulting in memory effects and nonreproducible results (18). A scanning electron microscopy equipped with an EDAX X-ray analyzer was used to investigate possible deterioration of the pyrolytic graphite surface of the graphite furnace cups. Figure 6 shows exposed porous graphite underlying residual siliceous material. This area of discontinuity is probably due t o the formation of silicon carbide, resulting from chemical attack on the pyrolytic surface a t the temperature of atomization.
SiOz + C Si0
+ 2C
-
+ CO(,, a t 1460 "C Sic,,,+ CO,,, a t 1600 "C Si0
(3)
(4)
I t was thought t h a t the addition of graphite powder to silicate samples would minimize the formation of siliceous globules, thus preventing the pyrolytic coating from being attacked. However, successive layers of the coating buckled and separated from each other, a fact probably due, a t least in part, to the layers becoming brittle and shrinking during successive high-temperature atomizations. Repeated use led to almost complete exfoliation of the layers from the under-
lying porous graphite. In the graphite-quartzite mix in the standard addition procedure, the solution was not absorbed into the underlying porous graphite and hence was available for atomization. This contrasts with the absorption of aqueous standards in an empty cup as described previously. During the atomizing and cooling cycles, siliceous residues agglomerate, as shown in Figure 7 . Elements qualitatively detected by the EDAX and analyzer in the residual siliceous material were: Si, Fe, Ni, Cu, Zn, and Mn. If cups are not cleaned manually between determinations, entrainment and gradual enrichment of trace elements in siliceous residues can lead to increased chemical interferences. For badly deteriorated cups with a large amount of accumulated residue, repeated heating a t high temperature between samples had to be used to remove trapped silver but this approach was not effective in removing the siliceous residue. A wool-wound pipe cleaner, bent to fit the internal shape of a graphite cup, could only partly remove residue by abrasion (18). Because the cup furnace walls were too thin to be machined, the exterior pyrolytic coating (19) could not be removed in order t o ensure good electrical contact between the furnace and the supporting electrodes. This results in the contact zone becoming damaged by arcing and hot spots with a corresponding decrease in the reproducibility of measurements. All these factors contribute to fast deterioration of the pyrolytic graphite surface. It is therefore advisable to discard cups after approximately ten determinations.
LITERATURE CITED L'vov, B. V. Spectrochim. Acta 1961, 17, 761-770. Venghiattis, A. A. Spectrochim. Acta, Par7 B 1967, 23, 67-78. Venghiattis, A. A.; Whirlock, L. At. Absorpt. News/ 1967, 6 , 135-136. Oshirna, S.;Kashiki, M. Anal. Chim. Acta 1970, 5 7 , 387-392. Willis, J. B. Anal. Chem. 1975, 4 7 , 1752-1758. Vergnaut, J. M.; Coudent, M. Anal. Chem. 1970, 42, 1303-1304. L'vov, B. V. Talanta 1976, 23, 109- 118. Jacobs, E. S. Anal. Chem. 1963, 35,2112-2115. Perone, S. P. Anal. Chem. 1963, 35,2091-2094. Wilson, A. D. Analyst (London) 1964, 89, 18-30, Benedetti-Pichler, A. A. "Physical Methods in Chemical Analysis", Beri, W. M., Ed.; Academic Press: New York, 1956; Volume 3. L'vov, B. V. Russ. J. Anal. Chem. 1971, 26,590-680. Katskov. D. A.; Kruglikova, L. P.; L'vov, B. V.; Polzik, K. Zh. Prikl. Spectrosk. 1974, 20, 739-741. Katskov, D. A.; Kruglikova, L. P.; L'vuv. B. V. Russ. J . Anal. Chem. 1975, 30, 238-243. Brady, D.V.; Montalvo, J. G.; Glowacki, G.; Pisciotta, A. Anal. Chim. Acta 1974, 70, 446-452. Langrnyhr, F. J.; Thornassen, Y. Fresenius' 2. Anal. Chem. 1973, 264, 122-127. Langrnyhr. F. J.; Stubergh, J. R.; Thornassen, Y.; Hanssen. J. E.; Dolezal, J. Anal. Chim. Acta 1974, 7 1 , 35-42. Garnys, V. P.; Srnythe, L. E. Talanta 1975, 22,881-887. Matousek, J. P. Talanta 1977, 24. 3'15-319.
RECEIVED
1980.
for review October 1, 1979. Accepted March 27,
