Analyst, November, 1979, Vol. 104, pp. 1037-1049 1037 Determination of Microgram Amounts of Precious Metals Using X-ray Fluorescence Spectrometry Paul R. Oumo and Evert Nieboer Department of Chemistry, Laurentian University, Sudbury, Ontario, P3E 2C6, Canada Microgram amounts of noble metals were localised with ammonium sulphide on filter absorbent pads and in cellulose pellets for spectrometer counting. The Ka, lines of ruthenium, rhodium and palladium (tungsten tube) and the La, lines of osmium, iridium, platinum and gold (molybdenum tube) were employed in conjunction with a lithium fluoride (ZOO) analysing crystal. At the 95% confidence level, detection limits of 1.0 p g (ruthenium, rhodium and palladium) and 0.6 p g (osmium, iridium, platinumand gold) were observed for the pellet technique, with values of 0.6 and 0.2 p g , respectively, for the absorbent- pad method.The average coefficient of variation for the determination of 10 p g of the seven metals studied was 6.5% for both sample presentations. No inter-elemental matrix interferences were observed among the noble metals themselves. However, the presence of more than 200 p g of nickel or copper reduced the slopes of the calibration graph by a constant factor of 10% for the lighter metals, and amounts of more than 400 pg of the base metals reduced the slopes by 20% for the heavier members. Good agree- ment was found between the X-ray fluorescence procedures and standard atomic-absorption methods in analysis of ore concentrates. Keywords : X-ray fluorescence spectrometry ; precious metals ; absorbent-pad technique; cellulose-pellet technique The precious metals include silver and gold and metals of the platinum group (platinum, palladium, rhodium, ruthenium, iridium and osmium).Almost all of the platinum-group metals are associated with ultramafic rocks, and nickel and copper are the chief metals extracted from these rocks1 Consequently, substantial amounts of the noble metals are recovered as by-products of nickel and copper smelting procedures. The scarcity and ever growing industrial use of the precious metals have necessitated accurate and precise methods of analysis. The following analytical procedures are generally employed : spectrophoto- metry,2 atomic-absorption ~pectrophotometry,~~~ gra~imetry,~ spectrochemical methods273 and X-ray fluorescence ~pectrometry.~?~ Of these procedures, X-ray fluorescence (XRF) spectrometry appears to offer some special attractive features, such as simple sample preparation and presentation, as well as the capability of analysing a single sample for a multiplicity of elements.However, published reports3s5 on the application of XRF spectro- metry to the determination of microgram amounts of precious metals lack detail and often only consider individual members of the group. This study was launched in order to devise a simple spectrometer presentation method that permits the determination of all of the precious metals in a single sample aliquot, to evaluate the lowest limits of detection and to study matrix and inter-elemental interference effects.Experimental Procedures Materials and Chemicals All the ore concentrate samples analysed were of a nickel - copper sulphide polymetallic type and were supplied by INCO Metals Company, Sudbury, Ontario, Canada. All chemicals used were of analytical-reagent grade. Preparation of Standard Materials and Samples Gold, platinum and palladium standard stock solutions were prepared by dissolving the individual metal sponges in aqua regia. Iridium, rhodium, ruthenium and osmium standard stock solutions were prepared from ammonium hexachloroiridate( IV) , ammonium hexa- chlororhodate(II1) , ammonium hexachlororuthenate( IV) and ammonium hexachloro- osmate(IV), respectively.1038 OUMO AND NIEBOER: DETERMINATION OF MICROGRAM AMOUNTS AnabySt, vd. 104 Test samples for calibration graphs, matrix and other studies were prepared as described below.Exactly 0.50ml of the appropriate standard solution was added to a 2.2 cm diameter absorbent pad, previously impregnated with 0.30 ml of 23.7% ammonium sulphide solution. The absorbent-pad discs used were cut to 2.2 cm diameter from 4.7-cm filter absorbent pads (Cat. No. AP1003700) made by Millipore Corp., Redford, Mass., USA. During the sample preparation, the cut discs were positioned in grooves cut in a plastic sheet of 1.5-cm thickness. After the addition of the standard solution, the discs were dried for 30 min in the oven at 105 "C, after which they could be presented directly to the sample holder of the X-ray spectrometer for counting. The second procedure for presenting samples to the spectrometer involved the preparation of cellulose pellets.Samples (0.50 g) of microcrystalline cellulose were weighed into medium- sized porcelain crucibles, then 0.5 ml of 23.7% ammonium sulphide solution was added to each, followed by 1.00-ml aliquots of the appropriate sample solution. The sample mixtures were then dried in the oven at 105 "C for 45 min, allowed to cool, mixed thoroughly using a mortar and pestle, and subsequently pressed into pellets at an applied pressure of 15 ton in-2 for a duration of 15 s with a semi-automatic press. Spectrometer Details X-ray spectrometer. summarised in Table I. The X-ray fluorescence measurements were carried out with a Philips, Model PW 1220, Conditions under which the measurements were carried out are TABLE I SPECTROMETER DETAILS Parameter value for individual analvsis Parameter ' Pd Rh Ru All Pt Ir 0 s X-ray tube Voltage/kV .. CurrentlmA . . Crystal .. Collimator . . Counter* .. Analytical line Peak "20 .. Backkround,. "20 Background,, "20 Counting timels .. .. .. w W .. .. .. 95 95 .. .. .. 20 20 .. .. .. LiF (200) LiF (200) .. .. .. Fine Fine . . . . . . gfp + scint gfp + scint .. .. .. Kal KUl . . . . . . 16.76 17.51 .. .. .. 15.25 15.25 . . . . . . 18.76 18.76 .. .. .. 100 100 W 95 20 LiF (200) Fine gfp + scint Kal 18.41 15.26 18.76 100 Mo 95 20 LiF (200) Fine gfp + scint 36.97 36.23 La, - 100 Mo 96 20 LiF Fine (200) gfp + scint La1 38.01 36.23 - 100 Mo 95 20 LiF (200) Fine gfp + scint 39.20 36.23 Lal - 100 Mo 96 20 LiF (200) Fine gfp + scint 40.41 36.23 LEI - 100 * gfp, Gas-flow proportional counter; scint, scintillation counter.Spectrometer Calibration The standard samples described above were used to establish calibration graphs for each metal covering the concentration ranges 0-30 pg for the absorbent-pad technique and 0-80 pg for the cellulose-pellet method. A master standard pellet was prepared from cellulose containing 8Opg of each precious metal. In this approach to calibration, the ratios of the corrected intensities of standard samples relative to the corrected intensity of the master sample were plotted against the concentration of the precious metals in the standard samples. Unknown concentrations were evaluated by referring the corrected count ratios of the unknown pellets to this calibration graph. Matrix Effect Studies Studies of the matrix effect were carried out to investigate possible absorption and enhance- ment of the characteristic analytical lines of individual noble metals by other noble metals and the base metals copper and nickel.Matrix effects among the precious metals themselves were studied using the absorbent-pad sample preparation technique, whereas the effects of copper and nickel were studied by the pellet method. Precision Studies palladium or of iridium, platinum and gold, or 10 pg of all six noble metals. Replicate samples were prepared containing either 10 pg each of ruthenium, rhodium andNovember, 1979 OF PRECIOUS METALS USING X-RAY FLUORESCENCE SPECTROMETRY 1039 Lowest Limits of Detection The lowest limits of detection were calculated for the confidence limits of 99.7y0, 95.4% and 68.3y0, which correspond to 30b, 20b and 0b, respectively, with Ob representing the standard deviation associated with the evaluation of the background6 : where Nb is the number of counts, Rb the counting rate and Tb the counting time in seconds ( N b = &,Tb).The values of Ob used in the calculation of the lowest limits of detection were expressed in micrograms by dividing them by m, where m is the slope of the appropriate calibration graph with units of counts per second per microgram. Analysis of Ore Concentrates Ore concentrate samples were dissolved and pre-treated in order to isolate and concentrate the precious metals using one of the following procedures: (a) separation of the precious metals by leaching with 12 M hydrochloric acid followed by coprecipitation of the precious metals with tellurium'; (b) nickel sulphide collection of precious metals by fire assay8-10; (c) fusion with sodium peroxidell; and ( d ) digestion with aqua regia.The identification of the ore samples analysed and the corresponding pre-treatment employed in each instance is considered in the Discussion. In the hydrochloric acid leaching process, 30 g of finely ground ore sample were digested for 3 h on a hot-plate with 600 ml of 12 M hydrochloric acid. When the dissolution of the base metals was completed, 200 ml of hot distilled water were added, followed by 10 ml of tellurium(1V) chloride solution (2.5 g 1-l). The mixture was then boiled for 5 min, after which 30 ml of tin(I1) chloride solution (500 g 1-1 in 2.5% V/V hydrochloric acid) were added.After boiling for 10 min, the mixture was filtered and the residue was washed, on the filter, with 50% V/V hydrochloric acid. The filter and contents were digested in 250 ml of aqua regia for 1 h. This digestion mixture was then filtered and the residue was discarded after thorough washing with 50% V/V hydrochloric acid on the filter. Evaporation, nearly to dryness, of the filtrate was then carried out, and the residue was taken up in a suitable volume of 25% V/V hydrochloric acid - 1% V/V nitric acid and was diluted to volume. Additional dilutions of this stock solution were effected with 8% V/V hydrochloric acid. The sulphide collection procedure is especially suited to samples containing substantial amounts of elemental copper.Ore samples (10 g) were mixed with 60 g of disodium tetra- borate, 30 g of sodium carbonate, 15 g of silica, 16 g of nickel powder and 8 g of elemental sulphur. The mixture was fused in a furnace at 1000 "C for 2 h, and the melt was then quickly poured from the clay crucible into a steel mould. The nickel button obtained was weighed, ground, re-weighed and subjected to the 12 M hydrochloric acid leaching procedure described above. It involved fusing (in a flame) 250 mg of ore with 4 g of sodium peroxide in a zirconium crucible until the melt was clear (this usually required 10-15 min). The fusion mixture was leached with 25 ml of distilled water and was subsequently acidified with 40 ml of 12 M hydrochloric acid, and then diluted to a suitable volume.Additional dilutions of this stock solution were effected with 8% V/V hydrochloric acid. Aqua regia digestion was achieved by heating, for a suitable period of time, 3-g samples with 25 ml of aqua regia on a hot-plate. The supernatant was isolated by filtration using a fine sintered-glass filter, and was then evaporated nearly to dryness. The residue derived from this filtrate was dissolved in 5 ml of 25% V/V hydrochloric acid - 1% V/V nitric acid and diluted to volume (50 ml) with distilled water. Cellulose pellets and absorbent discs were prepared from stock sample solutions in the usual manner using 1 .OO- and 0.50-ml aliquots, respectively. The ore concentrate samples employed (see Discussion) were also analysed independently by INCO personnel using conventional flame atomic-absorption spectrophotometry after pre-treating the various samples in the manner described above, Peroxide fusion was employed for samples with low levels of nickel and copper.Ru Rh Pd Ir Pt Au Ru Rh 0 s TABLE I1 CALCULATION OF CORRECTED SPECTROMETER COUNTS For peak 28-values used see Table I ; C is the count rate, with the subscript denoting the angle relative to the peak 28-value; and CF is the correction factor.Met a1 Technique Corrected sample counts/s-l .. .. .. .. .. .. .. .. .. Absorbent pad Absorbent pad and pellet Absorbent pad and pellet Absorbent pad and pellet Absorbent pad and pellet Pellet C2e - G e - 1.51 + C2e + 2.tdCF12 C2e - (C2e - 2.97)CF C2e - (C2e - 1.dCF C2e - (C2e - 0.74lCF C2e - (C2e + o.&F Pellet Absorbent pad and pellet C2e - (C2e - 4.18)CF Definition and value of CF 2Cte / ( G e - 2.28 + G e + 1.26) 2 ~ 0 / ( G e - 1.51 + Cie + 2.00) G e I G e - 2.97 G e ICZe - 1.78 G e I G e - 0.74 G e ICte + 0.36 CF = 0.97 -j= 0.01 CF = 1.05 & 0.01 CF = 0.90 -+ 0.01 CF = 1.06 -j= 0.01 CF = 0.96 & 0.01 CF = 1.02 -J= 0.01 G e /C* 2e + 1.26 Cie K,", - 418 CF = 1.09 & 0.01 CF = 0.92 & 0.01 * Indicates that the count rates correspond to the blank sample identified in the last column.Blank Absorbent pad treated with 0.3 ml of (NH4),S solution and 0.5 ml of 8% V/V HCl As for Ru As for Ru As for Ru As for Ru As for Ru Pellet made with 0.53 g of cellulose treated with 0.5 ml of (NH4),S solution + 0.5 ml of 8% V/V HC1 As for Ru As for RuNovember, 1979 OF PRECIOUS METALS USING X-RAY FLUORESCENCE SPECTROMETRY 1041 Results Spectrometer Details and Evaluation of Backgrounds The appropriate spectrometer settings for each individual analysis are summarised in Table I.Similarly, the manner in which corrected counts were evaluated is provided in Table 11. The correction factor, CF, takes into account sloping backgrounds and tube- target contaminations and was evaluated with a metal-free blank (see the last column of Table 11). Table I11 gives a comparison of the backgrounds evaluated as directed in Table I1 with those obtained by a least-squares method applied to graphs of uncorrected counts vemas amount of precious metal. The agreement between the two sets of values is good, and in most instances is within the range covered by one standard deviation based on replicate determinations by the correction-factor method.TABLE I11 BACKGROUND EVALUATION FOR ABSORBENT-PAD TECHNIQUE See Tables I and I1 for appropriate peak 26 values and correction-factor definition ; samples used contained equal amounts of Ru, Rh, Pd, Os, Ir, Pt and Au. Background count rate at peak 2Cvalue Metal Pd . . .. .. Rh .. .. .. Ru . . .. .. Au . . .. .. Pt . . .. .. Ir . . .. .. 0 s . . .. .. w Correction-f actor method* 469 f 9 430 f 9 395 f 8 210 f 7 187 f 6 173 f 6 173 f 6 - Least-squares methodt 45 7 428 393 193 191 175 171 * Values & standard deviation; corresponds to replicate t Evaluated by extrapolating graphs of uncorrected samples. count rates versus amount of precious metal. Calibration Graphs In Table IV, a summary is provided of the regressional parameters pertaining to standard calibration graphs plotted as corrected count rates versus the amount of a single precious TABLE IV REGRESSIONAL PARAMETERS FOR CALIBRATION GRAPHS Graphs of corrected count rates versus amount of metal on absorbent pad.A least-squares method that minimised the sum of the squares only of the y-residuals was used. Samples used contained only the metal indicated. Metal Au . . Pt . . Ir . . Pd . . Rh .. Ru . . 0 s .. Slope (ratel . . 30.4 . . 12.6 . . 11.3 . . 11.4 . . 4.8 . . 8.6 . . 10.7 Correlation Intercept coefficient* at - 7.7 0.999 10.3 -8.7 1.000 2.2 1.8 1.000 3.5 2.4 0.997 9.8 -0.6 0.998 2.8 -3.1 0.998 6.3 0.9 0.999 3.9 * Correspond to a level of significance of < O .l % ( p < 0.001). t o = 1/C(y-residuals)Z/N, where N is the number of data, points.1042 OUMO AND NIEBOER DETERMINATION OF MICROGRAM AMOUNTS Analyst, vat?. 104 metal on the absorbent pad. The values of near unity of the correlation coefficients emphasise the high degree of linearity observed. It is worth noting that the highest sensitivity (largest slope) was observed for gold, and the lowest for rhodium. The standard calibration graphs for all seven metals studied for the cellulose-pellet sample presentation method gave straight line graphs through the origin, up to the maximum concentration tested of 80 pg per pellet. These consist of graphs of the ratio of corrected sample counts to the corrected master standard counts veysus the amount of metal in the pellet.The standard samples contained ruthenium, rhodium, palladium, osmium, iridium, platinum and gold at the same concentra- tion. Typical correlation coefficients (Y) were 1.000 for osmium, gold and platinum, 0.994 for iridium, 0.996 for rhodium, 0.999 for ruthenium and 0.998 for palladium; the levels of significance (9) were less than 0.001 in all instances. Interestingly, no drastic differences in sensitivities for the various metals were observed for the pellets. This uniform response is demonstrated by the nearly constant slopes for ruthenium, rhodium and palladium (approximately 1.2 x 10-2) and for iridium, platinum and gold (approximately 1 x low2) given in Table VI (samples with no copper or nickel added). Good linearity was again found over the experimental concentration range.Matrix Effects The data in Table V for the absorbent-pad technique show no strong trend in corrected count differences for sample mixtures containing 20 pg each of palladium, rhodium, ruthenium, platinum, gold, iridium and osmium compared with the data for samples with the same amount of a single metal. The observed slopes of the calibration graphs for rhodium and ruthenium were not altered significantly by 30 pg of palladium or platinum or by 30pg of both of these metals (absorbent-pad presentation, data not reported). Similarly, it is seen in the cellulose-pellet sample presentation method in Table VII (columns 5 and 8) that the same intensities were observed for samples containing either 10 pg of three or six noble metals, respectively.In Table VI, the effects of nickel and copper on the intensities of the analytical lines of the various noble metals are summarised. TABLE V COMPARISON OF CORRECTED COUNTING RATES OF EACH PRECIOUS METAL IN A MIXTURE AND WHEN PRESENT ALONE (ABSORBHNT-PAD TECHNIQUE) Amount of Corrected count rates each metal/ A > Sample te Pd Rh Ru Au Pt Ir 0 s Mixture of Pd, Rh, Ru, Pt, Au, Ir and 0 s . . . . .. . . 20.0 227 94 177 625 246 231 200 Individual precious metal . . . . . . 20.0 230 93 154 603 250 225 224 The presence of 50 pg of nickel generally increased the observed intensity, corresponding to 6-15% increases in the calibration graph slopes ; ruthenium, rhodium and palladium being most affected. In contrast, 100 pg of nickel had no measurable effect on the intensities of the six noble metals studied.However, line intensities were reduced by about 10% for ruthenium, rhodium and palladium by 200 pg of nickel, but those for iridium, platinum and gold were not. When substantial amounts of copper were present, namely 400 and 800 pg, slopes were also smaller by approximately 10% for the lighter members and up to 20% for the heavier members. Pellets containing up to 20 pg of each precious metal studied, effectively exhibited no changes in intensity when copper was added in amounts between 40 and 600 pg. This apparent contradiction must be qualified by the observation that the standard deviation for replicate measurements was of the same magnitude as the expected 10-20% modifications in intensity due to the copper (see the next section).In separate studies, copper in the amounts of 50400 pg had no measurable bearing on the magnitude of the correction factors for ruthenium, rhodium, palladium, iridium, platinum and gold (pellet method).November, 1979 OF PRECIOUS METALS USING X-RAY FLUORESCENCE SPECTROMETRY 1043 TABLE VI SUMMARY OF REGRESSIONAL PARAMETERS FOR CALIBRATION GRAPHS IN THE PRESENCE OF COPPER AND NICKEL (CELLULOSE-PELLET TECHNIQUE) In contrast to the information in Table 11, two angles were employed for Ru and Rh to evaluate the background. Samples contained a mixture of all six precious metals. Metal Ru Rh Pd Ir Pt Au Pellet content of matrix element No Ni or Cu added Ni content = 50 pg per pellet Ni content = 100 pg per pellet Ni content = 200 p g per pellet Cu content = 400 p g per pellet Cu content = 800 pg per pellet No Ni or Cu added Ni content = 50 pg per pellet Ni content = 100 pg per pellet Ni content = 200 pg per pellet Cu content = 400 pg per pellet Cu content = 800 pg per pellet No Ni or Cu added Ni content = 50 pg per pellet Ni content = 100 pg per pellet Ni content = 200 pg per pellet Cu content = 400 fig per pellet Slope 1.23 f 0.08 1.37 1.18 1.14 1.13 1.13 1.2 f 0.1 1.31 1.12 1.08 1.09 1.07 1.3 f 0.15 1.49 1.31 1.20 1.18 ( x 102) Cu content = 800 bg i)er pellet 1.15 No Ni or Cu added 1.01 f 0.03 Ni content = 25 pg per pellet 1.01 Ni content = 50 pg per pellet 1.02 Ni content = 100 pg per pellet 0.98 Ni content = 200 pg per pellet 1.00 Cu content = 400 pg per pellet 0.89 Cu content = 800 pg per pellet 0.90 No Ni or Cu added 1.02 f 0.02 Ni content = 25 pg per pellet 1.03 Ni content = 50 pg per pellet 1.13 Ni content = 100 pg per pellet 1.02 Ni content = 200 pg per pellet 1.04 Cu content = 400 pg per pellet 0.85 Cu content = 800 pg per pellet 0.88 No Ni or Cu added 1.05 f 0.01 Ni content = 25 pg per pellet 1.02 Ni content = 50 pg per pellet 1.11 Ni content = 100 pg per pellet 1.03 Ni content --- 200 p g per pellet 1.08 Cu content = 400 pg per pellet 0.85 Cu content = 800 pg per pellet 0.87 Intercept ( x 102) -1.2 f 0.4 - 5.66 0.13 1.73 0.20 - 0.88 -1.7 f 1.3 -2.38 3.00 4.72 0.62 1.81 -0.3 5 2 -5.15 -0.62 3.99 0.93 3.57 1.37 0.55 1.2 f 0.8 -0.33 - 1.44 1.08 1.22 1.4 f 0.5 -0.31 -1.12 -0.77 - 1.99 2.13 0.55 1.3 f 0.2 1.00 0.21 0.58 - 1.15 1.01 -0.26 Correlation coefficient * 0.991 0.998 0.997 0.997 1.000 0.988 0.999 0.997 0.997 0.997 0.988 0.998 0.997 0.998 0.995 1.000 1.000 0.999 0.993 0.997 1 .ooo 0.995 0.999 0.999 0.993 0.997 1.000 0.997 0.999 0.999 0.995 0.996 1 .ooo - - - - - - U t ( x 102) - 3.93 1.66 2.24 1.93 0.64 4.38 1.15 2.11 2.15 1.98 4.98 1.98 2.34 1.58 2.60 0.45 0.30 1.16 2.44 1.78 0.40 1.29 0.48 0.99 2.70 1.49 0.44 1.09 0.55 0.85 2.28 1.75 0.54 - - - - - Number of data points, N 6 6 6 6 6 5 6 6 6 6 6 5 6 6 6 6 6 5 6 5 6 6 6 6 5 6 5 6 6 6 6 5 6 5 6 6 6 6 5 * Level of significance is <O.l% ( p < 0.001).t a = 2/C(y-residuals)2/N, corresponding to graphs of count ratios (sample to master standard) veisus amount of the specific precious metal in the cellulose pellet. Precision Studies The reproducibility observed (Table VII) for replicate samples was comparable for the two sample presentation methods studied.For 10-pg amounts of the precious metals, the standard deviations were in the range 0.3-1.2 pg, with means of 0.7 and 0.6 pg for the absorbent-pad and cellulose-pellet presentation procedures, respectively. The standard deviation values in columns 4, 7 and 10 of Table VII when multiplied by the factor 10 are converted into the coefficient of variation: (a in micrograms per microgram of metal) x 100 yo. Finally, the reproducibility of calibration graph slopes for replicate sets of calibra- tion standards was generally good. For gold, platinum and iridium, average deviations of 1-3% were observed in the magnitudes of the slopes, while the reproducibility was not as good (average deviations in the slopes of 7-12%) for ruthenium, rhodium and palladium (see the “No Ni or Cu added” entries in Table VI).1044 OUMO AND NIEBOER: DETERMINATION OF MICROGRAM AMOUNTS Analyst, VoZ.104 TABLE VII PRECISION STUDIES Count ratios refer to the ratio of corrected counts of sample to master standard sample, multiplied by 100; multiplication of the quantities in columns 4, 7 and 10 by the factor 10 yields coefficients of variation; SD is the standard deviation. Absorbent-pad technique c Average count Metal ratio* Ru . . . . 20.3 Rh .. . . 17.4 Pd . . . . 29.2 Ir . . . . 19.8 Pt . . . . 23.9 Au . . . . 43.8 SD of ratio SD/pg 1.22 0.60 1.63 0.94 1.65 0.57 0.86 0.43 1.50 0.63 4.90 1.12 r Average count ratio 15.9t 17.3t 15.5t 11.4t l0.6$ 11 .O$ Cellulose-pellet technique SD of count SD of ratio SD/pg ratios ratio 0.86 0.54 16.2 1.52 1.61 0.93 17.9 0.62 1.00 0.64 15.6 1.88 0.52 0.46 11.7 0.57 0.36 0.33 10.9 0.42 0.56 0.53 10.3 0.69 A - Average SD/H 0.94 0.35 1.20 0.49 0.39 0.57 * Corresponds to 8 replicate samples; samples contained 10.0 pg of all six metals.t Corresponds to 4 replicate samples; samples contained 10.0 pg of Ru, Rh and Pd. $ Corresponds to 5 replicate samples; samples contained 10.0 pg of Ir, Pt and Au. Corresponds to 4 replicate samples; samples contained 10.0 pg of all six metals. Detection Limits The lowest limits of detection based on background count rates measured for a counting time of 100 s are summarised in Tables VIII and IX. It is evident that the detection limits calculated are better by a factor of about two for the absorbent-pad technique (Table IX) relative to the pellet method (Table VIII).The detection limits for the heavier elements osmium, iridium, platinum and gold are twice as good or better than those for ruthenium, rhodium and palladium. TABLE VIII LOWEST LIMITS OF DETECTION OF VARIOUS PRECIOUS METALS Results found using the cellulose-pellet technique with a counting time, TI,, of 100s. The lowest limit of detection was calculated as 3 4 m , 2a1,lm and ob/m for the confidence intervals of 99.7%, 95.4% and 68.3%, respectively. The slope factor m corresponds to the slope of calibration graphs consisting of plots of corrected counting rates zlevsus concentration ; U b is defined in equation (I), Atomic number 44 46 46 76 77 78 79 Metal Ru Rh Pd 0 s Ir Pt Au Slope factor, m/ counts s-l pg-l 6.4 5.6 4.7 5.0 5.3 5.5 5.6 Lowest limit of detection (pg g-l) at the specified confidence limit Qb/ I A counts s-l 99.7% 95.4% 68.3%’ 2.67 1.26 0.84 0.42 2.74 1.48 0.99 0.49 2.87 1.83 1.22 0.61 1.50 0.91 0.61 0.30 1.59 0.90 0.60 0.30 1.71 0.94 0.63 0.31 1.68 0.89 0.59 0.30 Analysis of Precious Metal Concentrates In Fig.1, the results of analysing the ore concentrates by XRF spectrometry are com- pared with the accepted values determined by flame at omic-absorpt ion spect ropho t omet ry . The calibration graphs used corresponded to those for mixtures of precious metals using the absorbent-pad technique and, except for one ore concentrate, to those for noble metal mixtures in the presence of copper for samples analysed in the pellet form (see Table VI). The exception was the sample designated MY (see Discussion), for which the calibration graphs in Table VI, without nickel or copper added, were used.Except for the platinum level in one sample, the agreement between the two methods was good, as indicated byNovember, 1979 OF PRECIOUS METALS USING X-RAY FLUORESCENCE SPECTROMETRY 1045 the regressional parameters reported in the legend of Fig. 1. Of the two sample presenta- tions used, the cellulose-pellet procedure showed the best accord with the atomic-absorption (AA) spectrophotometric values (slope of fitted line, rn = 1.0). The absorbent-pad results were generally 5% lower than the AA or pellet XRF values (1% = 0.95). 80 60 20 I / 0 10 20 30 40 50 60 70 80 0 10 20 30 40 50 60 70 80 Accepted value (AA) Accepted value (AA) Fig.1. Comparison of the analysis of ore concentrates by X-ray fluorescence spectro- metry and accepted value by flame atomic-absorption spectrophotometry. (a), Absorbent- pad presentation: slope, m = 0.95; intercept, b = 0.00; correlation coefficient, r = 0.996; and level of significance, p t0.001. (b), Cellulose-pellet presentation: m = 1.02; b = -0.10; r = 0.997; p <0.001. Perfect agreement is depicted by the dotted line, and the units of concentration are either % m/m or ounces per ton. The anomalous Pt values shown (indicated by question marks) were excluded from the regressional analyses. , Ruthenium; 0, palladium; A, gold; 0, indium; 0, rhodium; and A, platinum. Discussion Sample Presentation The ammonium sulphide impregnation technique for localising metal ions was devised after preliminary localisation studies with nickel(I1). Initially, spotted absorbent pads were exposed to anhydrous hydrogen sulphide in a closed atmosphere.It was found that the black nickel(I1) sulphide precipitate preferentially accumulated at the edges, and thus centrifugal migration of the spotted nickel(I1) occurred. The pre-treatment of the absorbent TABLE IX LOWEST LIMITS OF DETECTION OF VARIOUS PRECIOUS METALS Results found using the absorbent-pad technique with a counting time. Tb, of 100 s. The lowest limit of detection was calculated as 3ub/m, 2Ublm and ub/m for the con- fidence intervals of 99.7y0, 95.4% and 68.3%, respectively. The parameter m is the slope defined and given in Table IV.Atomic number 44 46 46 76 77 78 79 Lowest limit of detection (pg) at the specified confidence limit r I A Element 99.7% 95.4% 68.3% Ru 0.69 0.46 0.23 Rh 1.29 0.86 0.43 Pd 0.57 0.38 0.19 0 s 0.36 0.24 0.12 I r 0.36 0.24 0.12 Pt 0.33 0.22 0.11 Au 0.16 0.10 0.061046 OUMO AND NIEBOER: DETERMINATION OF MICROGRAM AMOUNTS Analyst, vd. 104 pads with ammonium sulphide circumvented this migration. As the precious metal samples were dissolved in 8% V/V hydrochloric acid, the formation of complex ammonium chloro derivatives and various sulphides presumably occurred on contact between the sample aliquot and the impregnated absorbent pad. To prevent losses due to liquid adsorption on to container walls, ammonium sulphide was also used as the precipitant in the cellulose-pellet sample presentation method.It appears that the two sample presentation techniques developed offer some definite advantages over other known micro-sampling methods. As implied in the previous para- graph, evaporation procedures are often accompanied by centrifugal migration of ions, resulting in a non-uniform localisation on the filter discs. MacNevin and Hakkila5 did not mention this problem although they may have overcome it by adding the liquid aliquot to the centre of an oblong of paper placed concave upwards and resting on its four corners. Another approach is to collect a precipitate on a membrane filter.3J2 For example, Pietzner and Werner3313 determined gold by XRF spectrometry after tellurium collection, and trapping the precipitate on a membrane.They, and others using filter membranes,12 point out that lack of adhesion was encountered with voluminous precipitates. Cracking and peeling of the precipitates occurred. In addition, the adsorption properties of some colloidal precipi- tates make handling difficult prior to and during the filtration step.12 In this work total amounts of 210 pg of noble metals could be handled in the absorbent-pad procedure, without difficulty, although matrix studies involving large amounts of nickel and copper (400- 800 pg) did result in poorly adhering precipitates. None of these problems were encountered using the pellet method. Finally, the anion-exchange collection used by Taylor and Beamish3314 to determine small amounts of ruthenium was reported to give a uniform distribution on the exchange paper, and appears to compare favourably with the absorbent- pad and pellet techniques.Background Evaluation and Resolution An examination of the X-ray fluorescence emission lines for other elements15J6 revealed that possible interferences by overlap with the chosen analytical lines (Table I) was minimal. Potentially interfering lines were either very weak, of high excitation energy, or belonged to elements normally absent in samples containing the precious metals (e.g., lanthanides and actinides). The only observed exception was the overlap of the copper K/3 line with the osmium Lal, preventing the determination of osmium in samples containing appreciable amounts of copper. In studies on concentrated solutions of copper - nickel mattes in hydrochloric acid, the K/i? line of copper also affected the determination of p l a t i n ~ m .~ ~ ~ 7 ~ 1 The interference by overlap reported by MacNevin and Hakkila5 of palladium Ka with rhodium Ka, and platinum La, with iridium La,, was not evident in our work. Tube-target contamination was not serious, as indicated by the values of the correction factor, CF, of 1.0 The near constancy of the background intensities for palladium, rhodium and ruthenium (tungsten tube) and for gold, platinum, iridium and osmium (molybdenum tube) reaffirms the good base-line linearity. Tube- target contamination precludes the use of the tungsten tube for gold, platinum, iridium and osmium. The considerably higher background count rates for the lighter members (palladium, rhodium and ruthenium) account for their poorer detection limits.0.1 reported in Table 11. Sensitivity, Detection Limits and Concentration Ranges A comparison of slopes expressed in count rate per niicrogram of metal on the absorbent pad (see Table IV) gave the relative sensitivity order gold > osmium, platinum, iridium, palladium > ruthenium > rhodium. Consequently, the lowest detection limits reported in Table IX should and do follow the same trend. In contrast, there was not much spread in the corresponding slopes for the pellet method (see Table VIII), or in the detection limits (Ta.ble VIII). The reduced sensitivity for the cellulose pellet is not surprising, as it corre- sponds to a thicker and more dense and compact matrix.(X-ray fluorescence intensity obeys Beer’s law, A = ppx, where A is the absorbance, p the mass absorption coefficient, p the density and x the thicknes6) The observed detection limits for the absorbent-pad technique between 0.10 and 0.50pg (excluding rhodium) a t the 95% confidence interval, asNovember, 1979 OF PRECIOUS METALS USING X-RAY FLUORESCENCE SPECTROMETRY 1047 well as the corresponding range of 0.60-1.0 pg (excluding palladium) for the cellulose pellets, are comparable or better when compared with the values reported for individual precious metals3 However, they are an improvement by several orders of magnitude over those reported in the only detailed XRF spectrometric analysis of microgram mixtures of noble met a k 5 The concentration range for which linear calibration graphs might be expected appears to be very wide indeed according to reports by other w0rke1-s.~~~ Linearity up to milligram levels has been reported for iridium, platinum, palladium and rhodium for an evaporation sample-presentation te~hnique,~ amounts up to 150 pg for ruthenium by the anion-exchange paper method and osmium and gold up to 100 pg for precipitate collection on filter-paper.3 Good linearity of calibration graphs for percentage-level concentration ranges are also known : 0.2-2y0 for ruthenium in silica pellets (and 0.08-0.20 g 1-1 in solution),19 0.08-0.16~0 for rhodium, l.2-l.8y0 for platinum, 1.5-3.0% for silver and 2-6% for palladium in pellets of industrial platinum concentrates, copper - nickel slimes and products of their conversion,20 and 04% for platinum in flattened silver beads obtained by cupellation in the classical fire- assay process.21 Similar ranges have been cited by Beamish et aL3 for less recent studies.Consequently, the limit of 80 pg of each precious metal corresponding to a total of 560 pg chosen as the arbitrary upper limit in the cellulose-pellet studies should not constitute the actual experimental limiting level. In contrast, and as mentioned above, the maximum amount that can be handled by the absorbent-pad procedure would be limited by the adhesive properties of the precipitate. Nevertheless, total amounts of noble metals up to 200 pg pose no problem. Precision The observed values of the coefficient of variation corresponding to 1Opg of precious metal between 3 and 12% with an average of 6.5% (see Table VII) are comparable to, or smaller in magnitude than, those reported for analogous3 and larger concentration^^^^^ of individual precious metals.Comparable estimates were associated with the reproducibility of calibration graphs. Even though the observed precision for some of the metals studied was poorer than that of existing analytical norms, it is felt that it is adequate for the simple, but relatively fast, sample preparation techniques employed. It is reasonable to assume that some streamlining of the procedure and the familiarity accompanying repeated use should improve the precision. Inter-elemental Effects AS not much information is available on inter-elemental effects, the results obtained in this work are discussed in some detail.The lack of any obvious inter-elemental effect in the data in Table V, in which a comparison of counts of 20 pg of individual noble metals is compared with the same amount present in mixtures, was reinforced by the count ratios summarised in Table VII. In this table, no differences in intensities are recorded for samples containing three or six noble metals. This result is not surprising as the absorption coefficients for the absorption of the La, lines of osmium, iridium, platinum and gold by these same metals and by ruthenium, rhodium and palladium are all in the range of 110-160 (for compilations of absorption coefficients, p, consult MullerZ2 and Jenkins and de Vriess), and thus mixtures of these metals provide a nearly constant matrix.The value of the mass absorption coefficients for the absorption by sulphur of these La, lines is considerably lower (about 60). Similarly, there is very little self-absorption by palladium, ruthenium and rhodium of their own and each others Kcc lines (p rn 15). Absorption of these lines by osmium, iridium, platinum and gold is also low (p w70), and by sulphur is negligible (p -5-7). Consequently, favourable absorption coefficients combine with the low relative atomic mass cellulose matrix in the pellet and the thin layer of precipitate spread over a relatively large area on the absorbent pad, to generate matrices relatively free of inter-noble metal matrix absorption and enhancement effects. A similar conclusion was reached by Coombes et aZ.2l as they found that small amounts of palladium, rhodium, gold and iridium did not affect the intensity corresponding to 1% of platinum (Lcc line) in silver cupellation beads. Platinum metals did not interfere appreciably with the determination of gold (LBl line) in hydrochloric acid solution^.^,^^1048 Analyst, “ol.104 The observed influences of nickel and copper may have been expected on the basis of their known emission and absorption properties. Firstly, the interference of the copper KP line in the determination of osmium has already been mentioned. Secondly, copper and nickel do not strongly absorb the Ka lines of ruthenium, rhodium and palladium (p WOO), but they do more significantly affect the La lines of iridium, platinum and gold (p = 220- 280; the values for osmium La line absorption are 40 for copper and 280 for nickel).As the matrix studies reported in Table VI correspond to mixtures of all seven noble metals, the presence of 50 pg of nickel in the pellet should have rendered the matrix lighter for the Kcc radiation of ruthenium, rhodium and palladium, but not for the La lines of osmium, iridium, platinum and gold (see values of p given earlier). The 6-15% increases in calibration graph slopes due to the 50 pg of nickel for ruthenium, rhodium and palladium plus the lack of any significant effect on the slopes for the heavier metals are therefore reasonable. The presence of larger amounts of copper or nickel might be expected to reduce the total amount of radiation reaching the surface, and thus to reduce the detection sensitivity for all the precious metals.This effect was observed, as reductions in calibration graph slopes of up to 20% were recorded (Table VI). Presumably, the levelling off of this absorption effect for additions of more than 200 pg of copper and nickel for the lighter members and more than 400 pg for the heavier members corresponds to the condition of adding a large amount of a weak to moderate absorber as a diluent to dominate the matrix absorption.6 For small amounts of precious metals (q., 10 pg) the reductions in intensity were not statistically significant. OUMO AND NIEBOER: DETERMINATION OF MICROGRAM AMOUNTS Analysis of Ore Concentrates The ore concentrates examined and the pre-treatment employed (given in parentheses) were as follows: MY, a copper - nickel concentrate (hydiochloric acid leaching) ; RR, concentrate MY treated to remove most of the nickel as the carbonyl derivative (nickel sulphide collection) ; ISRM, INCO standard reference material, which is a Bessemer con- verter matte (peroxide fusion); SC, sulphur cake, concentrate RR with most of the copper removed (aqua regia digestion) ; and a precious metal concentrate (peroxide fusion).Each sample was analysed for ruthenium, rhodium, palladium, iridium, platinum and gold. It was not possible to analyse for osmium because of overlap of its La emission line with the K/3 line of copper, as X-ray fluorescence spectrometer scans showed the presence of substantial amounts of residual copper in the XRF samples (except in the instance of the MY concentrate).The under-estimation of the metal levels by the absorbent-pad method relative to the cellulose-pellet procedure (as well as atomic absorption) may possibly be due to the fact that the calibration graphs used were determined in the absence of copper in the standard samples (see Results). Apart from the platinum level in a single sample, the agreement between the XRF methods developed and the AA procedure is that which one might expect from an analytical pro- cedure having average coefficients of variation of 6.5%. The anomalous results obtained for platinum corresponded to the SC ore concentrate. As no separation of base metals was carried out for this sample, it is conceivable that some unidentified element interfered. The remaining deviations from the accepted values were random.This observation indicates that the XRF spectrometric procedure is not dependent on the nature of the ore concentrates examined, or on the separation and dissolution pre-treatments employed. The implication of this lack of gross matrix dependence is that ore samples can be examined by XRF without any pre-treatment provided that the calibration standards are of approximately the same composition and that high enough concentrations of the precious metals occur in the ore samples. have indeed determined silver, palladium, rhodium, ruthenium, gold and platinum by direct measurements on pellets prepared from industrial platinum concentrates, copper - nickel slimes and products of their conversion. This is in contrast with those used for pellet samples.As already mentioned, Shestakov et The authors thank J. Bozic and S. Maggs, Central Analytical Service Laboratory, INCO Metals Co., Ontario Division, Copper Cliff, Ontario, for their assistance and advice. D. Guest of the Geology Department, Laurentian University, also provided technical help. Financial support €or P. R. Oumo by the Canadian International Development Agency is gratefully acknowledged, as is the research funding received from Laurentian University.November, 1979 OF PRECIOUS METALS USING X-RAY FLUORESCENCE SPECTROMETRY 1049 References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. Youngquist, W. L., “Investing in Natural Resources : Today’s Guide to Tomorrow’s Needs,’’ Dow The National Research Council (USA), “Platinum-Group Metals,” National Academy of Sciences, Beamish, F.E., Lewis, C. L., and Van Loon, J . C., Talanta, 1969, 16, 1. Gilchrist, R., and Wichers, E., J . Am. Chem. Soc., 1935, 57, 2565. MacNevin, W. M., and Hakkila, E. A., Analyt. Chem., 1957, 29, 1019. Jenkins, R., and de Vries, J. L., Palmer, I., and Streichert, G., “The Coprecipitation of Nobel Metals with Tellurium. Jones - Irwin, Homewood, Ill., 1975, p. 165. Washington, D.C., 1977, pp. 66-78. Practical X-ray Spectrometry,” Second Edition, Springer-Verlag, New York, 1969. I. Platinum, Palladium, Rhodium and Gold,” National Institute for Metallurgy, Johannesburg, Report No. 1273, 1971. RobCrt, R. V. D., van Wyk, E., and Palmer, R., “Concentration of the Noble Metals by a Fire- assay Technique Using Nickel Sulphide as the Collector,” National Institute for Metallurgy, Johannesburg, Report No. 1371, 1971. RobCrt, R. V. D., and van Wyk, E., “The Effects of Various Matrix Elements on the Efficiency of the Fire-assay Procedure Using Nickel Sulphide as the Collector,” National Institute for Metal- lurgy, Johannesburg, Report No. 1705, 1975. Dixon, K., Jones, E. A., Rasmussen, S., and Robkrt, R. V. D., “The Efficiency of the Fire-assay Procedure with Nickel Sulphide as the Collector in the Determination of Platinum, Silver, Gold and Iridium,” National Institute for Metallurgy, Johannesburg, Report No. 1714, 1975. Jeffery, P. G., “Chemical Methods of Rock Analysis,” Pergamon Press, Oxford, 1970, p. 29. Watanabe, H., Berman, S., and Russell, D. S . , Talanta, 1972, 19, 1363. Pietzner, H., and Werner, H., 2. Analyt. Chem., 1966, 221, 186. Taylor, H., and Beamish, F. E., Talanta, 1968, 15, 497. White, E. W., and Johnson, G. G., “X-ray Emission and Absorption Wavelengths and Two-Theta Tables,” ASTM Data Series DS 37A, Second Edition, American Society for Testing and Materials, Philadelphia, Pa., 1970. Vandorpe, B., and Durr, J.] Analusis, 1977, 5, 38. Strasheim, A., and Wybenga, F. T., Appl. Spectrosc., 1964, 18, 16. Wybenga, F. T., and Strasheim, A., Appl. Spectrosc., 1966, 20, 247. Leoni, L., Braca, G., Sbrana, G., and Giannetti, E., Analytica Chim. Acta, 1975, 80, 176. Shestakov, V. A., Arkhipov, N. A., Makarov, D. F., and Kukushkin, Yu. N., Zh. Prikl. Khim., Coombes, R. J., Chow, A., and Flint, R. W., Analytica Chim. Acta, 1977, 91, 273. Miiller, R. O., “Spectrochemical Analysis by X-ray Fluorescence,” Plenum, New York, 1972. Chow, A., and Beamish, F. E., Talanta, 1966, 13, 539. Leningr., 1974, 47, 1035. Received December 13th, 1978 Accepted April 12th, 1979