JOURNAL O F ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 45 1 Determination of Silicon in Fine Gold by Solution and Solid Sample Graphite Furnace Atomic Absorption Spectrometry and Inductively Coupled Plasma Atomic Emission Spectrometry* Michael W. Hinds and Valentina V. Kogan Royal Canadian Mint 320 Sussex Drive Ottawa Ontario Canada K1A OG8 Four methods are described for determining silicon in fine gold. These include a solid sample electrothermal atomic absorption spectrometry (ETAAS) method with aqueous calibration standards an ETAAS solution based method with matrix matched standards an inductively coupled plasma atomic emission spectrometry (ICP-AES) method with matrix matched standards and a spark ablation ICP-AES method. The first three methods give comparable results although the solid sample ETAAS method is more error prone.Results from these techniques were used for the characterization of gold reference materials which were used as calibration standards for determining silicon in gold by spark ablation ICP-AES. In general limits of detection were 3 pg g-’ or better for the methods presented. Keywords Silicon; gold matrix; solid sample; electrothermal atomic absorption spectrometry; inductively coupled plasma atomic emission spectrometry The assay of high volumes of fine gold products can be accomplished by spark ablation inductively coupled atomic emission spectrometry (ICP-AES) and laser ablation induc- tively coupled plasma mass spectrometry (ICP-MS). In these methods the concentrations of common trace metal impurities are determined and the purity of gold is calculated by difference.Calibration is carried out using solid gold reference materials manufactured and characterized at the Royal Canadian Mint (RCM).l Silicon is a common low level impurity in gold. This is not surprising since most gold originates in siliceous ore. Consequently calibration standards manufactured to deter- mine impurities in gold by solid sample spectrometry must also contain silicon. Each trace metal in the gold reference materials was determined by at least two different analytical techniques. Two electrothermal atomic absorption spec- trometry (ETAAS) methods were developed for determining silicon in gold a solution method that used matrix matched standards and a solid sample method that used aqueous standards for calibration.Similarly two ICP-AES methods were also developed a solution based method and a spark ablation method for solid samples. The two ETAAS methods and the solution ICP-AES methods are described and com- pared. The spark ablation ICP-AES method is included to demonstrate that this solid sample ICP-AES method can produce acceptable silicon concentration values. Experimental Electrothermal atomic absorption spectrometers Two different atomic absorption spectrometers were used in the course of this work a PE 5000 and a PE 3100 (Perkin- Elmer Norwalk CT USA) both with an HGA 500 atomizer (Perkin-Elmer). Data collection was carried out using a per- sonal computer connected to the spectrometers via a DAS8 12 bit analogue-to-digital converter (Keithley Metrabyte Taunton MA USA).The data collection hardware and software were based on the work of Allen and Jackson2 Background corrected (continuum source) integrated absorbance values were calculated and used throughout this work. The outside wall temperature was measured by an optical pyrometer which was also computer interfaced.’ ~~ ~ * Presented at the XXVIII Colloquium Spectroscopicum Internationale (CSI) Post-Symposium on Graphite Atomizer Techniques in Analytical Spectroscopy Durham UK July 4-7 1993. A silicon hollow cathode lamp was operated at 40 mA. Two wavelengths were used the resonance line at 251.6 nm and the less sensitive non-resonance line at 221.1 nm. In both cases a slit-width of 0.2 nm was used. Temperature programmes for wall and platform atomization are listed in Table 1.Inductively coupled plasma atomic emission spectrometer The spectrometer used in this study was an ICAP 9000 inductively coupled plasma atomic emission spectrometer (Thermo Jarrell Ash Franklin MA USA). Samples could be introduced as liquids via a cross flow nebulizer or as solids (conducting solids only) via a spark ablation device (supplied by the manufacturer). An electronically controlled wave form spark source was used to sample solid metals and to generate a metal aerosol. A flow of argon gas swept the aerosol into the ICP torch where the material was excited. The resulting emission was measured by a 0.75 m focal length direct reading polychromator. A sapphire tipped torch was used to minimize background silicon emission.The optimum wavelength at 251.6nm was selected on the basis of sensitivity and freedom from interferences; these parameters were examined theoreti- cally (from a computer based wavelength library) and empiri- cally. The optimized instrumental parameters are listed in Table2. For solution analysis vanadium was used as an internal standard in 2% dissolved gold solutions and concen- trations were determined by the method of standard additions. Reagents Water used in these experiments was distilled and de-ionized by a Nanopure I1 system (Barnstead/Thermolyne Dubuque IW USA). Commercially prepared high-purity hydrochloric and nitric acids were used (Fisher Scientific Ottawa Ontario Canada) for the preparation of samples and standards. Calibration solutions were prepared from 1000 pg g-’ stock solutions (High Purity Standards Charleston SC USA) that are traceable to National Institute of Standards and Technology (NIST Gaithersburg MD USA) Standard Reference Materials.To ensure matrix matching each Cali- bration solution also contained an appropriate amount of dissolved Au 99.999% purity (Metalor USA Refining Corporation North Attleborough MA USA).452 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 Table 1 Temperature programmes for silicon in gold solutions L'vov platform Wall atomization Step Temperature/"C Ramp time/s Hold time/s Temperature/"C Ramp time/s Hold time/s 500 1 20 200 15 10 Dry Cool-down 20 1 10 20 1 10 Clean-out 2650 1 3 2650 1 3 Pyrolysis 1200 1 15 1400 1 15 Atomize* 2650 0 6 2650 0 6 * Read 1 s before atomization step and gas flow stopped during atomization.Table 2 Optimum instrumental parameters for trace element determi- nation in gold solutions and solid samples by ICP-AES Parameters Solution Spark ablation Forward r.f. power/W 1100 1100 Reflected r.f. power/W < 30 < 30 Outer gas flow rate/l min-' 19.0 19.0 Carrier flow ratefi min-' 0.4 0.4 Solution uptake rate/ml min- ' - 1.8 Sample Preparation So 1 u t io n For determinations by ETAAS gold samples (0.5g) were dissolved in 4 ml of aqua regia [HCl+HN03(3+1)] in a covered Teflon beaker under minimal heat for about 30min. The solution was either transferred into a 50ml plastic cali- brated flask and brought up to volume with distilled de-ionized water or 15 ml of water were added directly to the dissolved gold (evaporation losses were minimal).For determinations by ICP-AES 2% dissolved gold solu- tions were prepared by dissolving 2 g of a gold sample in 20 ml of aqua regia in a closed Teflon vessel. Samples were heated in a microwave oven (Model MDSSlD CEM Matthews NC USA) for 30min at 75% power. To minimize the risk of an explosion from hydrogen released in the dissolution process the vessels were purged with argon just prior to being sealed. A pressure controlling device maintained the pressure at 100 psi (1 psi "N 6.895 x lo3 Pa) but did not permit the pressure in the closed vessel to exceed this value. Upon dissolution samples were transferred into 100 ml plastic calibrated flasks as described above. Solid sample For ETAAS solid pieces of gold were cut from either shavings or chunks by using a fine sharp stainless-steel knife.Pieces between 0.2 and 0.5 mg were placed on a tared weighing pan and the exact mass was recorded. Two different balances were used in this work a Mettler AE163 analytical balance (smallest division 0.01 mg) and a Mettler UM3 electronic balance (small- est division 0.0001 mg). The weighed gold sample was trans- ferred into a 10 ml plastic cup where it could conveniently be picked up by a fine curved-nose steel forceps and placed in the graphite furnace. This was facilitated by the use of a funnel (a pipette tip cut off 7 mm from the tapered end) set in the dosing hole to assist dropping the sample into the atomizer. One could also insert the sample by dropping it directly through the dosing hole.Solid samples for spark ablation ICP-AES were easily prepared by machining a flat surface on a metal lathe and then mechanically polishing the surface with aluminium car- bide paper. Sample sizes ranged from 2 cm diameter and 0.1 cm thickness to 10 cm diameter and 5 cm thickness. Results and Discussion Dissolved Gold Solution Analysis Silicon is not typically soluble in aqua regia and it was thought that it would remain in particulate form once the gold matrix was dissolved. This was investigated by dissolving a gold sample (doped with silicon) in aqua regia diluting with water and then leaving the solution in a plastic centrifuge tube undisturbed. Silicon absorbance was measured by ETAAS when the solution was originally made up and then each day for five days.The absorbance values remained the same throughout the time of the experiment. This indicated that silicon remains in solution either as a soluble species or as a stable colloidal suspension in the dissolved gold solution (12.5 g 1-1 of Au and approximately 15% hydrochloric acid). However in more dilute solutions (4-10-fold dilution) inte- grated absorbance values for silicon decreased after 1 to 2 h. Upon re-agitation the original integrated signal values were obtained. It would appear then that silicon might not be fully dissolved in the solution and forms a stable suspension in concentrated solutions but is unstable in dilute solutions. Electrothermal Atomic Absorption Spectrometry Optimization of parameters Pyrolysis temperature does not greatly effect the silicon analyt- ical signal from the platform up to 1400"C as shown in Fig 1 (line A).The effect of pyrolysis temperature for 10 ng of Si atomized from the wall of the graphite furnace and observed at 221.1 nm was also studied (Fig. 1 line B). Silicon was stable up to 1600°C. This might be due to the larger analyte mass atomized and observation at the less sensitive wavelength which may mask the smaller changes observed with the platform experiment. The effect of the amount of gold deposited within the atomizer on the integrated absorbance measured for 2 ng of Si is shown in Fig. 2. The signal observed decreases as the 500 1000 1500 2000 TemperaturePC Fig. 1 Effect of pyrolysis temperature on the integrated absorbance for A 2ng of Si with 10 pg of Au from platform atomization at 251.6 nm; and B 10 ng of Si (aqueous solution) from wall atomization at 221.1 nmJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL.9 453 0.3 I I I 1 I I 0 5 10 15 20 Mass of gold/pg Fig.2 Effect of mass of gold in the atomizer on the integrated absorbance of 2 ng of Si from platform atomization at 251.6 nm 0.4 I 1 0 5 10 15 20 25 30 HC1 (%I Fig. 3 Effect of hydrochloric acid concentration on the integrated absorbance of 2 ng of Si from platform atomization at 251.6 nm amount of gold increases. This is consistent with experiments performed by Frech et a13. who proposed that this observed attenuation is due to the formation of fine condensed metal (or oxide) particles within the analyte volume. These particles can act as adsorptive surfaces for analyte atoms in the gas phase which leads to the reduction in the observed integrated signal.Also metal matrix condenses at the cooler ends of the atomizer which also acts as an adsorptive s ~ r f a c e . ~ However the former mechanism would probably be dominant because only very small amounts of gold re-condense at the ends of the atomizer (if at all) at the atomization temperature of 2650 "C and therefore would probably have little effect on the silicon atoms in the gas phase. The effect of hydrochloric acid was investigated since it is another major component of dissolved gold solutions. Chloride present as sodium chloride was reported to be an interferent5 The presence of hydrochloric acid did not appear to affect the integrated signal observed for silicon (Fig.3) however increases in acid concentration decreased the lifetime of the platform. Inductively coupled plasma atomic emission spectrometry There did not appear to be any interference from the gold matrix on the silicon wavelength as shown in Fig. 4. Nevertheless background correction was used because rela- tively high levels of emission were observed for both the acid blank and a high-purity gold blank. There was approximately a 30% decrease in the observed silicon emission due to the gold matrix as compared with the same concentration of silicon in water. This effect did reduce sensitivity but did not have a large detrimental effect on the determination. 150 I I 81 12 I I J Wavelengthhm 251.756 251.476 251.616 Fig. 4 Wavelength scan about the Si emission line at 251.616 nm for A a 2% high-purity gold blank; B 2.0ppm Si in a 2% high-purity gold solution; and C 5.0 ppm Si aqueous solution from ICP-AES atomization silicon atomic absorbance peak shapes from an aqueous solution and a solid gold sample were nearly coinci- dent (Fig.5). Temperature measurements indicated that atom- ization of the majority of silicon atoms from both sample types occurred as stabilized temperature conditions were established within the atomizer. It appears that silicon atoms originating from both the solid sample and the aqueous standards were exposed to similar temperature environments. Thus the residence time for each should be similar and the value of integrated absorbance should also be comparable.This means that aqueous silicon can be used as a calibration standard for the analysis of solid samples. Preliminary determi- nations using aqueous standards were in agreement with results for the determination of silicon by ICP-AES. This led to more detailed experiments being carried out as outlined below. Wall atomization Graphite tubes were modified by enlarging the dosing hole from 2 to 3.5mm in diameter in order to observe physical changes in 0.5mg gold samples. No change was observed when the temperature of the drying step was set to 200°C. However at the pyrolysis temperature ( 1400 "C) samples melted and formed small spheres after the pyrolysis step. Observations of the physical effects at atomization tempera- tures were carried out with a transversely heated graphite atomizer (THGA) (after the platform was taken out) because the THGA permitted convenient recovery of the remaining gold sample (compared with the Massmann type atomizer).A maximum temperature of 2600°C could only be used because of software limitations. After one complete atomizer firing (Table l) it was found that the sample was only reduced in mass by about 24%. The non-resonance line for gold at 274.8 nm was monitored during consecutive firings for one Solid Sample Analysis Electrothermal atomic absorption spectrometry Experimentation with direct solid sample introduction into a graphite atomizer was initiated after observing that with wall 0 1 2 3 4 Timels Fig.5 Absorbance peak profiles for long of Si from an aqueous solution (A) and from a 0.4 mg solid gold sample (B); C is the outside tube wall temperature454 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL.9 Table 3 Figures of merit for three different approaches to determining silicon in gold by solid sample ETAAS; value given +1 standard deviation Wall Wall Platform 221.1 nm 251.6 nm 221.1 nm gas stop gas flow gas stop Characteristic mass*/pg 226 44 750 + 110 190 i- 14 Detection limitt/pg g-' 4.6 16 - * Mass of analyte whose absorbance is equal to 0.0044 s. t k = 3 based on estimated 0.35 mg gold sample. gold sample (0.5 mg). High integrated absorbance signals were observed for three consecutive firings. No variation from the baseline was seen after the fourth firing. This was consistent with the observations of Irwin et aL6 for nickel samples who noted that the majority of the nickel sample remained in the atomizer after one heating cycle.As noted previously the peak shapes for silicon from solution and solid sample were nearly coincident (Fig. 5). Temperature measurements showed that heating at maximum power permit- ted the temperature to stabilize just as atomization occurred. This particular situation is mainly due to the refractory nature of silicon because more volatile elements would have atomized as the temperature of the atomizer was rapidly increasing. Two approaches to the determination of solid samples were used gas flow during atomization with detection at the resonance line for silicon (25 1.6 nm) and gas stop flow during atomization with detection at a non-resonance line for silicon (221.1 nm).In both cases peak shapes were quite similar. Sensitivity was fairly low for gas flow rates between 100 and 250mlmin-' and was similar to the value shown in Table 3 for a flow rate of lOOmlmin-'. Characteristic mass values for flow rates of 50 ml min-' approached those obtained from detection at 221.1 nm with gas stop conditions. Platform atomization Platform atomization was tried and the figures of merit are presented in Table 3. Improved sensitivity and reproducibility were observed compared with wall atomization. Unfortunately there was incomplete atomization of silicon from the solid gold samples. Subsequent re-firing of the atomizer after a determination in many cases produced another silicon signal that was about 20-30% of the original signal.This was not acceptable for quantitative determinations and was probably caused from the reduced heating rate of the platform which slowed the release of silicon from the solid sample and ulti- mately led to incomplete atomization. Method comparison Limits of detection (LODs) (Table 3) were based on the integrated absorbance measured for the empty atomizer firing immediately following the determination of silicon in a solid sample. As previously noted about 75% of the sample mass remains after the first atomization cycle. It was estimated that the average amount of gold remaining in the furnace used for the calculation of LODs would be 0.35 mg. Gold remaining in the atomizer did not appear to contain any measurable levels of silicon. High-purity gold was not used as a blank because it contains low levels of silicon (<1 pgg-') which may not be homogeneously distributed in the gold.The estimates of LOD indicate that the method utilizing the non-resonance line at 221.1 nm has a lower LOD than the method involving gas flow during atomization. No reasonable estimate could be obtained for atomization from a platform because of the frequent occurrence of memory effects Monitoring silicon absorbance at the non-resonance line (221.1 nm) was the more favoured technique because of the I I I 251.476 251.616 251.756 Wavel e ngt h/n m Fig. 6 Wavelength scan about the Si emission line at 251.616 nm for 90.pg g-' Si in A gold reference material FAUlO and B 27.8 pg g-' Si in gold reference material FAU8 from spark ablation ICP-AES better sensitivity and better LODs obtained compared with measuring the absorbance at the resonance line (251.6 nm) with a high gas flow during atomization (noted in the second column of Table 3).Spark ablution inductively coupled plasma atomic emission spectrometry Spark ablation ICP-AES is a well established technique for the determination of trace metals particularly for the iron and steel ind~stry.~ The accuracy of trace element determinations by this technique depends mainly on the concentration values assigned to solid sample calibration standards and the homo- geneity of the standards. Gold reference materials have been prepared by the RCM specifically for this purpose. Trace element concentrations have been determined by at least two independent methods.A detailed account of the manufacture and characterization of these reference materials has been outlined by Kogan et al.' The calibration graph was linear for the three standards covering silicon concentrations up to 28 pg g-l with a linear regression coefficient of 0.9964. A scan of the emission about the 251.6 nm silicon emission line confirmed that there was little evidence of interference from the gold matrix (Fig. 6). The peak areas for each of the profiles outlined in Fig. 6 are proportional to the silicon concentrations determined by the other methods described in this paper profile A 9.0k2.2 pg 8-l (k 1 standard deviation) and profile B 27.8k4.8 pg g-'. The average diameter of the ablated craters was 2.4k0.9 mm and the average depth was about 0.25 mm.The amount of gold ablated was about 1 mg determined by volume calculations (assuming a cylindrical crater shape) and by measuring the mass before and after ablation. Comparison of Analytical Results A comparison of analytical results for three of the methods is shown in Table 4. In general there is very good overlap between solution concentration values obtained by ETAAS and ICP- AES. Solid sample ETAAS as one would expect was less Table 4 Comparison of the determination of silicon in different gold reference materials by different methods; values given & 1 standard deviation CSil/Clg g-' Methods FAU8 FAUlO Solid sample ETAAS* (n = 4) 30$4 6.9 & 1.3 Solution ETAAS (n = 4) 9.0 & 1.4 Solution ICP-AES (TI = 5 ) 9.1 f 0.8 27.2 + 1.8 26.4 t_ 1.7 * 221.1 nm 0 gas flow during atomization.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL.9 455 precise and somewhat biased compared with the solution based methods. Nevertheless values obtained through solid sample ETAAS overlapped with the solution methods for FAU8 gold reference material (RCM). The solid sample value for FAUlO gold reference material (RCM) is significantly different from solution methods as determined by a t-test (n= 4 95% confidence level). This value can be considered to be somewhat comparable especially if one takes into account all the potential errors associated with solid sample methods.' Spark ablation ICP-AES was used to determine silicon in a gold reference material undergoing certification. Existing gold reference materials with different silicon concentration levels (determined by at least two of the three other aforementioned techniques) were used for calibration.A concentration value of 45 pg 8-l was obtained. This was in good agreement with the value of 47 pg g-' obtained from determination using solution ICP-AES. The experiment was repeated for another gold reference material. There was good agreement between the value of 29.1 f0.9 pg 8-l (+ 1 standard deviation; n=4) obtained from solution ICP-AES and the value of 29.10+ 1.63 pg 8-l (n = 3) from spark ablation ICP-AES. Figures of merit for the four techniques are compared in Table 5. Sample masses presented to the atom source were compared. Solution ICP-AES used the largest amount of sample whereas the smallest amount was consumed by the solution ETAAS method.The sample uptake rate for ICP- AES was 1.85 ml min-' for 30 s per determination. A factor of 0.02 was applied to account for the 2% nebulizer efficiency. Sample masses consumed by the solid sample methods (ETAAS and ICP-AES) were within a factor of 2 of each other. It is difficult to compare sensitivities between AAS and AES techniques. Comparisons were made between similar methods. As one would expect solution ETAAS is more sensitive than solid sample ETAAS yet both have about the same LOD. The AES methods are less sensitive than ETAAS techniques but appear to be more reproducible as denoted by the lower LODs. Conclusion Four different analytical methods for determining silicon in fine gold have been presented and the method involving the analysis of gold solutions by ETAAS appears to be the most sensitive technique.It was found that the determination of silicon in gold by solid sample ETAAS (wall atomization) with calibration by aqueous standards was reasonably accurate despite the physi- cal differences between the aqueous calibration standards and the solid gold sample. The main reason for this phenomenon appears to be that the atomization of silicon occurs as steady- state temperature conditions are established within the graphite furnace for aqueous solution and solid samples. However for this application solid sample ETAAS has limited utility Table 5 Comparison of figures of merit for solid sample and solution based techniques for the determination of silicon in gold Solid sample Solution methods methods ETAAS ICP-AES ETAAS ICP-AES 0.5 1 0.10 220 Sample mass/mg Sensitivity Characteristic mass*/pg 39 - 226 - Counts (ppm-') - Detection limit/pg g-' 3 1 3 1 30 - 22 * Mass of analyte whose integrated absorbance signal is equal to 0.0044 s.because it is more time consuming and less precise than solution ETAAS. The determination of silicon by solution ICP-AES is con- venient but there is a slight decrease in sensitivity owing to the dissolved gold matrix. Spark ablation ICP-AES is some- what more sensitive and is very convenient for the routine analysis of large numbers of solid samples. However this method requires calibration by solid gold reference materials which up till recently have not been available. The authors thank L. McKay G. Ocampo and G. Valente for their assistance in completing the experiments for this paper. M.H. also thanks K. W. Jackson (New York State Department of Health) for the graphite furnace temperature measurements and V. Luong (National Research Council of Canada) for assistance in computer interfacing the atomic absorption spectrometer used in these experiments. References Kogan V. Hinds M. W. Ocampo G. and Valente G. in Precious Metals 1993 ed. Mishra R. International Precious Metal Institute Allentown PA 1993. Allen E. and Jackson K. W. Anal. Chim. Acta 1987 192 355. Frech W. L'vov B. V. and Romanova N. P. Spectrochim. Acta Part B 1992,47 1461. Frech W. Li K. Berglund M. and Baxter D. C. J. Anal. At. Spectrom. 1992 7 141. Frech W. and Cedergren Anal. Chim. Acta 1980 113 227. Irwin R. Mikkelsen A. Michel R. G. Dougherty J. P. and Preli F. R. Spectrochim. Acta Part B 1990 45 903. Lemarchand A. Labarraque G. Masson P. Broekaert J. A. C. J. Anal. At. Spectrom. 1987 2 481. Bendicho C. and de Loos-Vollebregt M. T. C. J. Anal. At. Spectrom. 1991 6 353. Paper 310561 OD Received December 17 1993 Accepted October 12 1993