Utilization of Metallic Platforms in Electrothermal Vaporization Inductively Coupled Plasma Mass Spectrometry* ISAM MARAWI LISA K. OLSON JIANSHENG WANG AND JOSEPH A. CARUSOT Department of Chemistry University of Cincinnati Cincinnati OHI 45221-01 72 USA A method to improve the analytical performance of hydride trapping on Pd inside a graphite furnace and subsequent determination by ICP-MS was investigated. Strips of tungsten tantalum molybdenum and rhenium were coated (electroplated/sputtered) with palladium and used as platforms inside the graphite furnace. A typical pyrolytic graphite-coated graphite platform was also coated with Pd in the same manner for performance comparisons. Scanning electron microscopy (SEM ) and X-ray fluorescence spectrometry were used to characterize the surfaces.SEM showed a smoother Pd layer covering the metallic substrates when compared with the graphite. An analyte signal reproducibility study revealed that Ta Re and Mo were not ideal substrates for this purpose. The As peaks obtained from trapping on the graphite platform were broadened with a decrease in intensity after a few firings. The Pd-sputtered tungsten platform showed superior performance over the other platforms. The linear dynamic range obtained using this platform was improved by one order of magnitude in the upper limit (from 1 to 100 ng ml-') over that reported using a graphite platform in previous work. The limit of detection of arsenic was found to be 0.01 ng rnl-' within the same range obtained from the graphite platform. Keywords Hydride generation; electrothermal vaporization; inductively coupled plasma mass spectrometry; hydride trapping; metallic platforms Continuous hydride generation for sample introduction in analytical atomic spectrometric techniques is a popular method for the detection of arsenic and other volatile hydride-forming elements in environmental samples.'-12 The popularity of this method is due to several advantages that can lead to superior analytical performance over the more common methods of sample introduction such as solution neb~lization.'~-'~ Graphite furnace (GF) trapping of hydrides at relatively low temperatures in the presence of palladium is a potentially useful technique for analyte preconcentration. Lower concen- tration limits of detection for several volatile hydride-forming elements have been obtained by applying this method with electrothermal (graphite furnace) atomic absorption spec- trometry ( ETAAS),17*'8 microwave-induced plasma atomic emission spectrometry ( MIP-AES)19i20 and inductively coupled plasma mass spectrometry (ICP-MS).21 In the last method (ICP-MS detection) micrograms of Pd solution were deposited on a L'vov platform inside the graphite furnace.After drying the Pd solution the hydrides were vented through the furnace at a temperature between 200 and 600°C. At these tempera- tures the palladium trapped the analytes,22 or possibly they were adsorbed on active graphite sites.23924 When the atomiz- ation temperature was reached (2100-2600 "C) the analytes desorbed from the surface and were instantly purged into the *Presented in part at the 1994 Winter Conference on Plasma f To whom correspondence should be addressed.Spectrochemistry San Diego CA USA January 10-15 1994. Journal of Analytical Atomic Spectrometry ICP-MS using a suitable carrier gas (GF carrier gas) such as argon. Preliminary experiments demonstrated the ability to trap and subsequently determine several elements by ICP-MS in a single run,21 and a detailed performance comparison between the trapping method and the typical ETV and the direct HG was presented.21 The collection efficiency of arsine was found to be near 100% at ultra-trace levels (ngl-'). However collection at higher concentrations was limited. This decrease in efficiency limited the method's linear dynamic range and practical applications.In this work Pd-coated metallic platforms (electroplated/ sputtered) were investigated in an attempt to increase the analyte collection efficiency at higher analyte concentrations. The purpose was to examine the utility of a metallic substrate coated with Pd as a platform inside a graphite furnace for arsine trapping. This paper describes the preparation of these metal platforms and presents evidence of palladium coverage on their surfaces. It also discusses in detail the analyte signal reproducibility over an extended number of firings. EXPERIMENTAL Instrumentation Graphite furnace A modified graphite furnace unit (HGA 300; Perkin-Elmer Norwalk CT USA) was used with a graphite tube without the sample introduction hole (Sherba Analytical Lab Products New Port Richey FL USA) for the collection and atomization of the analyte.The Pd-coated platforms (metallic and graphite) were positioned inside the graphite tube in a manner similar to the L'vov platform. A new graphite tube was used with each different platform. Inductively coupled plasma mass spectrometry A VG Plasma Quad PQ I (VG Elemental Winsford Cheshire UK) was used. The detector was operated in the single-ion monitoring (SIM) mode in which the analyte signal can be collected versus time. The optimized operating conditions for this instrument were described in a previous paper.2' Method Hydride generation Reaction of As in acidic solution (4 moll-' HCl) with sodium tetrahydroborate ( 1 % m/m NaBH in 0.1 mol I-' NaOH) was used for the continuous generation of arsine.A two-channel peristaltic pump was used to feed both the analyte and the reagent solutions at the same flow rate (0.45 ml min-') into a polyethylene reaction tube (10 cm x 0.2 cm id.). The gaseous products (arsine and excess hydrogen) were carried by a purging gas (0.7501min-' argon) to the graphite furnace through a polypropylene porous membrane tube which was Journal of Analytical Atomic Spectrometry January 1995 Vol. 10 7utilized as a gas-liquid separator (GLS). The GLS and the instrumental set-up were described in detail previously.21 Preparation of the metallic platforms Electroplating. Tungsten and tantalum foils 1 mm thick (Johnson Matthey Ward Hill MA USA) were cut to similar dimensions as a typical L'vov platform by an electrodischarge machine (EDM) (Lambda Research Cincinnati OH USA).The platform-shaped metallic strips were etched in a hydroflu- oric acid-nitric acid solution for 4 h. This step was necessary to clean the surface to remove cutting process residues. Prior to plating they were rinsed in ethanol and air dried. The electroplating solution was 10%0 Pd metal in 10% nitric acid and a palladium wire 99.9% pure (Johnson Matthey) was used as an anode. The three platforms W Ta and graphite were each electroplated long enough to form a significant layer of Pd on their surfaces. After the plating they were placed inside a graphite furnace and the temperature was ramped to 2650 "C five times. Those platforms will be identified throughout this paper as graphite 1 tungsten 1 and tantalum 1.Sputtering deposition. Tungsten molybdenum tantalum and rhenium strips 0.1 mm thick (H. Cross Weehawken NJ USA) with the dimensions of a typical L'vov platform were used. As purchased the metallic strips had minimal surface roughness. Prior to sputtering the metallic strips were rinsed in ethanol blown dry with nitrogen and immediately placed in the sputter- ing system. The graphite and metallic strips were arranged in a circle on the substrate holder and sputtered in a single experiment at room temperature. A commercially available palladium plate (100 x 100 mm 0.1 mm thick 99.9% pure; Johnson Matthey) was used as a target for the sputtering source. A 1 pm thick film of palladium was deposited on one side of the platforms. The sputtering experiment was performed by the Material Science Department College of Engineering University of Cincinnati.The sputtered graphite tungsten and tantalum platforms are henceforth identified as graphite 2 tungsten 2 and tantalum 2. Directly after deposition (anealing at high temperature was not carried out) palladium surface roughness and coverage were determined by scanning electron microscopy (SEM) and X-ray fluorescence (XRF) spectrometry. Reagents All solutions were prepared with distilled de-ionized water (resistivity 18 mi2 cm) (Barnsted PCS Boston MA USA). Hydrochloric acid ( 12 moll-' Baker Instra-analyzed reagent; J.T. Baker Phillipsburg NJ USA) and nitric acid (70% certified ACS reagent; Fisher Scientific Fairlawn NJ USA) solutions were prepared by dilution with distilled de-ionized water to the desired concentrations.Sodium tetrahydroborate (NaBH,) (99% powder analytical-reagent grade; Johnson Matthey) solution was prepared by dissolving the powder in 0.1 moll-' NaOH (ACS reagent; Fisher Scientific). Arsenic working standard solution was prepared by serial dilution of a 1000 mg ml-' stock standard solution (atomic absorption standard solution; Aldrich Milwaukee WI USA). The pal- ladium electroplating solution (Baker Newark NJ USA) was used without any modification. Sample Drinking water HPS certified reference material lot No. 390225 (High-Purity Standards Charleston SC USA) and Freeze-Dried Urine (low level non-certified standard) Standard Reference Material SRM 2670 (NIST Gaithersburg MD USA) were analysed for arsenic. One volume of the drinking water was mixed with two volumes of 6 mol 1-1 HCl to give a final acid concentration of 4 moll-l.The urine sample was reconstituted in 4mollM1 HCl following NIST instructions. Procedure The arsine was generated continuously and was introduced into the furnace by the GLS carrier gas during the hydride trapping step of the GF temperature programme only; this step was 30 s in duration in all experiments. A four-way valve (Whitey Highland Heights OH USA) was used to control the carrier gas flowing into the graphite furnace. In one position of this valve only the GLS carrier gas is flowing into the graphite furnace whereas in the other position the GF carrier gas only flows in. The GF carrier gas was argon at a flow rate of 130mlmin-'. The GF carrier gas purged the furnace at all times with the exception of the hydride trapping step.The GF carrier gas transfers the analytes during the atomization step to the plasma. A representative GF tempera- ture programme used in this study is presented graphically in Fig. 1. RESULTS AND DISCUSSION Surface Coverage The XRF counts for Pd obtained from the electroplated tungsten 1 tantalum 1 and graphite 1 platforms were 2633 576 and 827 respectively. These counts represented Pd cover- ages of 48% on W and 29% on Ta. Elements with relative atomic mass less than 19 could not be determined using the XRF instrument utilized in this study so the coverage ratio on the graphite platform could not be determined. Fig. 2(a) (b) and (c) are scanning electron photomicrographs of the Pd film electroplated on the graphite tungsten and tantalum substrates respectively. The scanning electron photomicro- graph for the graphite platform presented in Fig.2(u) shows microscopic aggregates of palladium on the surface with vari- able sizes. Presumably the pyrolysed graphite surface had a limited interaction with Pd. The XRF count and the scanning electron photomicrograph obtained from the graphite platform sputtered with Pd (graph- ite 2 ) were similar to those obtained from the electroplated platform. On the other hand all sputtered metallic platforms had better than 95% Pd coverage. Fig. 3(a) (b) and (c) are scanning electron photomicrographs of the Pd film coating the tungsten 2 molybdenum and rhenium substrates respectively. Trapping Reproducibility The reproducibility of the arsine signal obtained by using each platform was evaluated by examining the relative standard 3000 Clean 2 2500 I;' 2000 2 5 1500 \ OI a E 1000 $ 500 20 40 60 80 100 120 140 160 0 Time/s Fig.1 Graphical representation of a prototype graphite furnace temperature programme 8 Journal of Analytical Atomic Spectrometry January 1995 Vol. 10Fig* Scanning photomicrograph of Pd On Fig. 3 Scanning electron photomicrographs of 1 pm thick fih of Pd (a) graphite (b) tungsten and (c) tantalum. For XRF counts and coverage see Surface Coverage sputtered on (a) tungsten (b) molybdenum and (c) rhenium. XRF counts indicated that Pd was covering more than 95% of these surfaces deviation (s,) for several trapping replicates. After seven blank firings the graphite 1 platform was used in eight trapping experiments with 1.35 ng of arsenic.The peak areas obtained for each of these experiments are presented graphically in Fig. 4(u) The s obtained for the eight trappings performed in this study was less than 26%. Following this study a sub- sequent XRF experiment showed no discernible Pd peak and presumably none of the palladium remained on the surface. The results from SEM supported this finding. Fig. 5(a) is a scanning electron photomicrograph of this surface showing only the dull graphite. After three blank firings the graphite 2 platform was used in 27 trapping experiments with 0.7ng of arsenic. The peak areas obtained for each of these experiments are presented in Fig. 6(u). The over-all s calculated for the 27 trappings per- formed in this study was less than 13%.After the 30 firings an XRF experiment was carried out on this platform the results showed that all of the Pd aggregates were evaporated Journal of Analytical Atomic Spectrometry January 1995 Vol. 10 911 t i 8 10 12 14 I cn v) 3 +I 8 150 0 m 5 100 L 2 (b) 50 m 0 10 20 30 40 50 60 70 150 Ic’ e No. of firings Fig.4 Signal reproducibility from trapping 1.35 ng of As on Pd-electroplated platforms (a) graphite 1 (b) tungsten 1 and (c) tanta- lum. Each bar represents the peak area obtained in a signal trapping experiment and the void areas represent blank firings and these results are similar to those obtained with the electroplated platform. The decrease in the arsenic signal [Fig. 4(a)] and the loss of the sharpness of the As peak realized in this study (see Fig.7) agreed with earlier reports about the effect of palladium on trapping experiments.21,22 When palladium was present inside the graphite tube the analyte collection efficiency was high and the analyte peak was sharp; however when the palladium evaporated (supported by the XRF findings) the trapping efficiency was lost and the analyte peak was much broader. The tungsten 1 platform was fired 75 times and 42 firings were trapping experiments with 1.35 ng of arsenic. The results are plotted in Fig.4(b) where the value obtained from each trapping experiment is presented as a bar and the void areas represent a blank firing. The overall (42 trapping runs) s value calculated is less than 52%. The XRF count showed that less than 2% of the surface was still covered with Pd and Fig.5(b) shows the shiny Pd in the grooves of the shattered tungsten surface after the total of 75 firings. The peak areas obtained with both tungsten 1 and tanta- lum 1 and plotted in Fig. 4(b) and (c) decreased within runs. After several blank firings the signal returned to near its original value and repeated the pattern. A possible explanation for the erratic behavior might be sub-surface involvement in the analyte trapping. This phenomenon will have an unpredict- able effect on the volatilization of the analytes. The analyte peak obtained during the cleaning step shown in Fig.7 (the second peak appearing in both tungsten and tantalum plots) may support this argument. Also a substrate thickness of 1 mm might have been a factor in the substrate-palladium annealing process.Consequently much thinner (0.1 mm) plat- Fig. 5 Scanning electron photomicrographs of the electroplated plat- forms after the reproducibility work; (a) graphite 1 (b) tungsten 1 and (c) tantalum 1 forms were utilized in subsequent experiments (sputtered platforms). The tungsten 2 platform was used for 47 trapping experi- ments with 0.7 ng of arsenic. The results obtained are plotted in Fig. 6(b). The over-all (47 runs) s value calculated is less than 12% and the value calculated for the last 35 runs is less than 5%. The XRF count showed that more than 2% Pd coverage remained on this surface and Fig. 8(a) shows one of the Pd aggregates remaining on the tungsten surface after 82 firings.10 Journal of Analytical Atomic Spectrometry January 1995 VoE. 105 9 13 17 21 25 29 1 5 9 13 17 21 25 29 33 37 41 45 49 . 16 18 20 22 24 26 28 30 No. of firings Fig. 6 Signal reproducibility of 0.7 ng of As trapped on Pd-sputtered platforms (a) graphite 2 (b) tungsten 2 and (c) molybdenum Tungsten 1 2oth firing 7+--4 120 140 160 ' l;O IkO li0 '- Ti me/s Tantalum 1 16th firing Graphite 1 13th firing Fig. 7 Time-resolved signals obtained from trapping of 1.35 ng of arsenic using three different platforms Fig. 8 Scanning electron photomicrographs of the sputtered plat- forms after the reproducibility work (a) tungsten 2 (b) molybdenum and (c) rhenium The tantalum 1 platform was used in 60 firings using the same temperature programme as described before and 44 of those firings were trapping experiments with 1.35 ng of arsenic.The results obtained are plotted in Fig. 4(c). The s value obtained for these experiments was less than 31%. Fig. 5(c) shows the post-trapping scanning electron photomicrograph of this platform. It appeared that the tantalum melted inside the graphite tube. Melting of this platform could have been due to tantalum and palladium forming an alloy with a melting-point lower than the atomization temperature. Fig. 9 is a phasephase diagram of the tantalum-palladium metals and indicates that these two metals can form an alloy with several compositions all of which have melting-points lower than 1950 "C. An attempt to trap using the tantalum 2 platform with a modified temperature programme having an atomiz- ation temperature at 1900 "C was not successful.Presumably at this temperature there was insufficient energy to desorb the arsenic. Similar work was carried out using a sputtered molybdenum platform. As molybdenum and palladium form a binary alloy that melts at temperatures higher than 1755°C,25 the atomiz- ation temperature in the temperature programme was adjusted to 1700°C. Thirty firings were performed on this platform using the modified temperature programme; the last 15 of Journal of Analytical Atomic Spectrometry January 1995 VoZ. 10 1 ITantalum (at.-%) 0 10 20 30 40 50 60 70 80 90 100 3200 I I I I I I I t I 3020°C I 70 2800 1 ( a ) - V 2400 1 = Y +- g 2000 Q E 8 1600 L __--- 1920 & 25°C / - t A- I 1 I I I I I I \ J I - 800 ' ' ' 1 0 10 20 30 40 50 60 70 80 90 100 Tantalum (% m/m) Fig.9 Pd-Ta phase diagram (reproduced from ref. 25 with permission) those firings were trapping experiments with 0.7 ng of As. The s value obtained for these experiments was less than 8%. However the arsenic signal observed was spread over three peaks (Fig. lo) and these peaks corresponded to changes in temperature during the temperature programme. Each peak represents some of the arsenic being volatilized off the surface. It appears as if the modification power of the Pd was lost in the presence of molybdenum and any addition of energy (heating) to the system caused some of the arsenic to be evaporated. A possible explanation for this behaviour is that molybdenum is a very reactive metal and is possibly involved in trapping the arsenic.'' Fig.8(b) shows the post-trapping scanning electron photomicrograph; the XRF count revealed approximately 1% of Pd remaining on this surface. The rhenium platform was also used in a similar manner to the other platforms. The temperature programme was also modified for the reasons mentioned above; the atomization temperature utilized was 1600°C. Fig. ll(a) shows the single- ion monitoring of the first trapping experiment performed on this platform. The increase in the background between 50 and 80s on the time scale represents the arsine that was not trapped inside the furnace; moreover the peak height and area realized during the atomization step between 90 and 120s were lower than those obtained using the other platforms. The signal obtained in the second trapping attempt is presented in Fig.11 (b) and clearly indicates very poor trapping efficiency. 400 u) 2 CI 2 : 2 300 C 2 200 0 a + 4- 100 50 6o i 30 40 i a - Lo 2 80 2 70 * 60 50 40 30 20 10 L 0 4- 90 120 150 0 30 60 Time/s Fig. 11 Time-resolved signal from 0.7 ng of As (ASH,) in an attempt to trap it on a fresh Pd-coated rhenium platform; (a) first run (b) second run. The broad peak appearing between 60 and 90 s represents the As which escaped the trapping during the hydride trapping step of the G F temperature programme (Fig. 1) The poor trapping on this surface could be due to the saturation of the surface with the analyte which possibly formed a refractory complex metallic compound with the palladium. An attempt to volatilize the analyte off the surface at a higher temperature resulted in melting of the platform.Fig. 8(c) shows the surface after the two trapping attempts and the XRF count indicated that about 5% Pd remained on this surface. Peak Shape and Intensity The arsenic peak shape obtained when using different platform substrates after several runs is demonstrated in Fig. 7. The peak obtained from graphite 1 after 13 firings is short and wide; this shape is similar to that observed on trapping the analyte on the wall of an old graphite tube. The peaks that resulted from the metallic substrates were sharper. The second peak due to the cleaning step suggested that subsurface collec- tion or a stable compound formation may be taking place with the substrate metal. A comparison of the peak intensities clearly shows the much higher analyte signal obtained from metallic substrates than that obtained using graphite.The low signal intensity obtained from the experiment employing the graphite 1 platform may be due to the loss of the palladium coating which was also suggested by the XRF count. 0 30 60 90 120 150 Tim e/s Fig. 10 Time-resolved signal from 0.7 ng of As (ASH,) trapped on Pd-sputtered Mo platform. The presence of multiple As peaks after the trapping step suggest the reactivity of Mo with As Memory Effect A new tungsten platform freshly electroplated with Pd was used to study the memory effect. The amount of arsenic (70 ng) utilized was sufficient to saturate the platform. The results 72 Journal of Analytical Atomic Spectrometry January 1995 Vol.10I 1,786,063 I Blank Blank Blank Blank Blank Sample 70 ng Arsenic Blank Blank Blank Fig. 12 Arsenic signals obtained during blank firings due to the memory effect obtained are presented in Fig. 12. The bar height represents the response of the detector to the analyte introduced into the plasma. The blank firings on the abscissa indicated that no arsenic was introduced for trapping in these experiments. The peaks obtained from these blank runs are due to a memory effect. Less than 10% of the original analyte signal was obtained in the first blank firing and Fig. 12 shows that the signal returned to the original background level within seven runs. This result was reasonable considering the high analyte concentration used and the sensitivity of the detector (ICP-MS).Analytical Figures of Merit The limit of detection (LOD) was calculated based on LOD = s/m where s is the standard deviation for a minimum of eight background measurements and m is the slope of the calibration graph. A detection limit for arsenic of 0.01 ngml-I was obtained. The precision of the method was determined by measuring the s for a minimum of six replicates of the standard used for the calibration graph. The s values obtained were less than 10 14 and 11% for concentrations of 1 I0 and 100 ngml-I respectively. Fig. 13 shows the log-log plot of the linear dynamic range over the concentrations mentioned above; the slope of this plot is 1.03. Sample Analysis Prelimenary work to test the accuracy of this method was carried out by the determination of the total arsenic concen- 0 1 2 Log [concentration ( p p b ) l Fig.13 Arsenic linear dynamic range in log space obtained from the trapping study tration in HPS drinking water. The result obtained was 60+ 3 ng ml- ' (average k 2s). This result is lower than that expected i.e. 8OkO.4 ng ml-I. A possible explanation for the low values obtained from the samples is a matrix effect on the hydride generation reaction.26-28 High concentrations of sodium calcium and magnesium in this matrix and the pres- ence of several transition metals may have a retarding effect on the efficiency of the hydride generation reaction. CONCLUSION A thin (0.1 mm thick) tungsten strip coated with Pd has the potential to be utilized as a platform inside a graphite furnace. The plasma discharge sputtered platforms showed superior results to the electroplated platforms.The sputtered tungsten platform retained the palladium for an extended number of firings more than 80 in this study. Utilization of this platform increased the palladium surface area available for analyte trapping which improved the linear dynamic range by an order of magnitude. Also in this study the Pd solution depos- ition step was eliminated which could improve the sample throughput and assist in the automation of this technique. However a number of blank firings (12-20) are necessary for conditioning the surface and could include cleaning and proper annealing of the palladium and tungsten. Pd-coated tantalum molybdenum and rhenium were found not to be suitable for this technique possibly owing to the reactivities of these metals at the temperature used (400 "C).I.M. acknowledges the University of Cincinnati's scholarship fund for their support and E. H. Evans for suggesting the problem. The authors thank professor A. T. Hubbard for helpful discussions. This project received partial support by grant No. ES 04908 and ES 03221 from NIEHS. The authors are also grateful to the Material Science Department for assisting with the SEM and XRF studies and particularly Mr. Ernest Clark whose advice and help were invaluable. REFERENCES 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Le X. C. Cullen W. R. and Reimer K. J. Appl. Organomet. Chem. 1992,6 161. Zhang B. Tao K. and Feng J. J. Anal. At. Spectrom. 1992,7 171. Le X. C. Cullen W.R. Reimer K. J. and Brindle I. D. Anal. Chim. Acta 1992 258 307. Ybanez N. Cervera M. L. and Montoro R. Anal. Chim. Acta 1992 258 61. Rohr U. and Meckel L. Fresenius' J. Anal. Chem. 1992,342,370. Cacho J. Ferreira V. and Nerin C. Analyst 1992 117 31. van Elteren J. T. Haselager N. G. Das H. A. de Ligny C. L. and Agterdenbos J. Anal. Chim. Acta 1991 252 89. Alverez G. H. and Stephen S . G. Anal. Lett. 1991 24 1695. Schramel P. and Xu L. Freesius' J. Anal. Chem. 1991 340 41. Tao H. and Miyazaki A. Anal. Sci. 1991 7 55. Ebdon L. Hill S. J. and Jones P. Talanta 1991 38 607. Hakala E. and Pyy L. J . Anal. At. Spectrom. 1992 7 191. Story W. C. and Caruso J. A. in Preconcentration Techniques for Trace Elements ed. Alfassi Z. B. and Wai C. M. CRC Press Boca Raton FL 1992 p.333. Heitkemper D. T. Wolnik K. A. Fricke F. L. and Caruso J. A. in Inductively Coupled Plasma in Analytical Atomic Spectrometry ed. Montaser A. and Golightly D. W. VCH New York 2nd edn. 1992 p. 781. Nakahara T. Prog. Anal. Spectrosc. 1983 6 163. Janghorbani M. and Bill T. G. Anal. Chem. 1989 61 701. Li Z. McIntosh S. Carnirck G. R. and Slavin W. Spectrochim. Acta Part B 1992 47 701. Doidge P. S. Sturman B. T. and Rettberg T. M. J. At. Anal. Spectrom. 1989 4 251. Matusiewicz H. Sturgeon R. E. and Berman S . S. Spectrochim. Acta Part B 1990 45 209. Bulska E. Tschopel P. Broekaert J. A. C. and Tolg G. Anal. Chim. Acta 1993 271 171. Journal of Analytical Atomic Spectrometry January 1995 Vol. 10 1321 Marawi I. Wang J. and Caruso J. A. Anal. Chim. Acta 1994 291 127. 22 Sturgeon R. E. Willie S. N. Sproule G. I. Robinson P. T. and Berman S. S. Spectrochim. Acta Part B 1989 44 667. 23 Lee D. S. Anal. Chem. 1982 54 1682. 24 Sturgeon R. E. Wille S. N. and Berman S. S. J. Anal. At. Spectrom. 1986 1 115. 25 Binary Alloy Phase Diagrams ASM International Materials Park OH 1986. 26 Smith A. E. Analyst 1975 100 300. 27 Yamamoto Y. and Kumamaru T. Fresenius’ 2. Anal. Chem. 1976,281 353. 28 Pierce F. D. and Brown H. R. Anal. Chem. 1976 48 693. Paper 4/01 51 8E Received March 14 1994 Accepted September 13 1994 14 Journal of Analytical Atomic Spectrometry January 1995 VoZ. 10