Dc arc spectrometry of solids: some new aspects of an old method† Karol Flo�ria�n,*ab Ju�rgen Haßlerc and Eva Surova�a aDepartment of Chemistry, Faculty of Metallurgy, Technical University of Kos¡ice, Letna� 9, SK-042 00 Kos¡ice, Slovakia bDepartment of Physical and Analytical Chemistry, Faculty of Sciences, P.J. S¡ afa�rik University, Moyzesova 11, SK-041 54 Kos¡ice, Slovakia cElektroschmelzwerk GmbH, P.O. Box 1526, D-87405 Kempten, Germany Received 10th September 1998, Accepted 24 November 1998 The critical evaluation of old-fashioned dc arc excitation under modern conditions and in connection with a multichannel spectrometer is reported.As typical solid samples in all evaluation steps, various kinds of SiC powders were used; as analyte elements, the most important impurities (Al, Fe, Ni, Ti and V) were chosen. The broad use of the method is also demonstrated in the ecologically relevant analysis of sediments, where the main toxic and essential elements (Cr, Cu, Ni, Pb, V and Zn) were analysed. Contrary to our earlier investigations, in this case, neither a special working atmosphere nor the dilution of samples with spectrochemical additives was applied. The main figures of merit of the method were determined (precision of the method, calibration characteristics, limit of determination, etc.).Their values confirmed the applicability of the modernized dc-arc-OES method as a rapid alternative to existing solid sampling (SS)-OES and SS-AAS methods.For a long period of spectroscopic history, direct current (dc) enables the destruction of such complicated ceramic matrices as SiC or B4C and, at the same time, the total evaporation of arc excitation was the leading technique for the direct analysis of powdered samples, especially in the specific area of geo- impurities from the powder sample. These aspects of the forgotten dc arc technique were reported in the last Solid chemical analyses. This initially semi-quantitative method, after much optimization and quantification1 (internal Sampling Colloquium.9 If the disadvantages of old-fashioned dc arc spectrography, the non-controlled free burning of the standard method, Lomakin–Scheibe equation, blackeningtransformation procedures, etc.), became a rapid, widely used arc and the photographic registration of spectra, are avoided, a new modernized dc arc connected with quartz-fibre optics multi-element quantitative method.The latter optimization, in the evaluation of photographically registered spectra1 (modern to a multichannel spectrometer will give a good alternative spectrochemical method to the existing ICP-ETV and/or densitometers), led to an average precision of about 15% with relatively low detection limits.Some modifications of the dc SS-ETV-AAS techniques in the routine analysis of complicated powder samples. These statements are underlined with the arc itself were also developed, e.g.the use of stationary magnetic fields,2 inhomogeneous magnetic fields3 and the availability of commercial dc arc sources from ARL Fisons (Crawley, W. Sussex, UK) or Thermo Jarrel Ash (Franklin, double-plasma method.4 Also, continuous sample introduction techniques were successfully evaluated and used in the practical MA, USA). The research reported in this paper confirms the statements analysis of powdered samples. The intensive development of new excitation sources in given above.As a model for all evaluation steps, SiC powder was chosen—a typical sample in solid sampling. The use of atomic spectroscopy and the commercial use of modern spectrometric techniques halted the evaluation of dc arc excitation, the proposed dc arc-OES method is evaluated on the basis of the analysis of sediment samples of environmental importance. and research in the field of the direct analysis of solids was consequently directed in both atomic emission and atomic Special attention was paid to the calibration procedure: calibration with diVerent weights of a single reference material absorption spectrometry towards the use of various models of electrothermal vaporization (ETV).5 In connection with the (RM) was preferred and evaluated in detail. most successful excitation source of the last three decades, the ICP, other sample introduction techniques were evaluated Experimental (slurry nebulization, direct sample insertion, laser ablation, etc.), but ICP-ETV and ETV-AAS seem to be the most Instrumentation widespread.The experiments were carried out using an LECO-750 (St The use of the ETV technique has certain disadvantages, in Joseph, MI, USA) simultaneous spectrometer with quartz- particular the upper temperature limit of about 2900 °C. fibre optics and a computer-controlled dc arc excitation source Therefore, in special cases of solid sampling (SS) analysis with optimized ramping (see Table 1) of the arc current.The (ceramic powders, sediments), the use of either working atmosconnection between the dc arc light source and the LECO-750 pheres6 of special gas mixtures or of spectrochemical additives7 spectrometer was realized by two vertically oriented quartz is necessary. fibres with collecting lenses (double optics, DO arrangement); The dc arc excitation source in connection with highresistance carbon auxiliary electrodes achieves on the top of the carrier electrode, a temperature8 of about 4000 °C, which Table 1 The arc current–time programme of the dc arc source Burning time/s 0–5 5–8 8–16 16–29 29–65 65 †Presented at the 8th Solid Sampling Spectrometry Colloquium, Arc current/A 2.7 6.9 3.9 8.4 13.5 0 Budapest, Hungary, September 1–4, 1998.J. Anal. At. Spectrom., 1999, 14, 559–564 559Table 2 The characterization of the reference samples of SiC used Reference sample SiC-2495 SiC-3757 SiC-628 SiC NMP-6 d50/mm 54–59 10.8–13.3 1.5 80–90 Al/mg g-1 825 350 265 458 B/mg g-1 — — 40 19.5 Ca/mg g-1 60 19 23 74.4 Cu/mg g-1 — — 3 5.1 Fe/mg g-1 520 140 415 702 Ni/mg g-1 135 — 100 134 Ti/mg g-1 290 140 90 183 V/mg g-1 780 380 410 217 Usually, the calculation of the straight line of the calibration is based on the least-squares method, but the fulfilment Fig. 1 The arrangement of the quartz-fibre optics.of predicted conditions is rarely reported. First of all, the normality of repeated measurements (signals) should be checked with the David test, in which the value in the special case of arc stability testing, a single (SO) quartz fibre was used (see Fig. 1). D� =Ri/si for each data set of repeated signal measurements for given Evaluation procedure calibration concentrations is calculated (Ri is the range and si the standard deviation of the given data set) and compared Testing of instrumental parameters with the table data. If the calculated value lies inside the range In the first step, the optimum observation position of the (2.21<D<2.71 for K=5 repetitions, and statistical signifi- collecting lens of the quartz fibre was determined (see Fig. 2). cance level a=0.10), the normality of the data set is confirmed In the following steps, the influence of the sample homogeneity, (denoted as + in Tables 9–13). sample amount weighed into the cavity of the carrier electrode In the second evaluation step, the homoscedasticity (homoand slit width of the spectrometer was investigated.In all of geneity of variances si2) should be checked;10 when K1=K2= these experiments, the influence on both the integrated intensity .....KN=K=const., the value (Cochran test) means and the RSD values was followed, using SiC powders of variable grain size as model samples. G� =smax2/ S N i=1 si2 Calibration and its characteristics is calculated and compared with the table data. If The analytical calibration using SiC standards (see Table 2) G� <Gtab and eight sediment RMs (see Table 3) was based on the linear the homoscedasticity is confirmed (denoted as + in Tables calibration function: 9–13).integrated intensity=f [concentration] The goodness of fit (test of the linearity, test of the adequacy of the linear model ) can be evaluated on the basis of two The integration coen for each element. The variances:10 the residual variance experimental conditions varied as a function of the number of RMs used, caused by the non-linearity of the calibration line in some cases.Besides the classical calibration with a set of sres.2= 1 N-2 S N i=1 (yi-y� i)2 RMs, calibration with diVerent weights of a single RM was used. and the variance of repeated measurements sw2= S N i=1 S K j=1 ( yi,j-yi)2 N(K-1) with the linear calibration function y=A+Bc and N calibration samples (concentrations c), each with K repetitions. If, for the calculated value FLIN= sres.2 sw2 FLIN<FTAB is true (FTAB of the Fisher distribution; degrees of freedom n1=N-2; n2=N(K-1); and statistical level of significance a=0.05), then the goodness of fit is confirmed (denoted as + in Tables 9–13).The figures of merit Besides the mentioned characteristics of calibration (the residual variance is the most important characteristic), other figures of merit were calculated: Fig. 2 The experimental arrangement of various observation positions of the quartz-fibre optics. (i) the average relative standard deviation RSDa6 [the RSDi 560 J.Anal. At. Spectrom., 1999, 14, 559–564Table 3 The characterization of reference materials (RMs) of sediments used Element/mg g-1 RM Cu Zn Cd Pb Cr Co Ni V B BCR-277 101.7 547 11.8 146 192 43.4 BCR-280 70.5 291 1.6 80.2 114 73.6 BCR-320 44.1 142 0.53 42.3 138 75.2 GWB-07305 137 243 0.82 112 70 18.9 34 109 51 GWB-07309 32 78 0.26 23 85 14.4 32 97 54 GWB-07312 1230 498 4.0 285 35 8.8 12.8 46.6 NIST-2704 98.6 438 3.45 161 135 14 44.1 95 SL-1 30 223 0.26 37.7 104 19.8 44.9 170 (39) Table 4 The dependence of the integrated intensity means and RSDs obtained using the position in the middle of the electrode gap, on the observation position (six repeated measurements, each of them but with lower signal values.As the precision of the measuretwice, see Fig. 2) ments is most important, the observation position 4 was chosen for future experiments. Position Homogenization of the SiC sample is of greater importance Element Parameter 1 2 3 4 5 6 when powders of larger grain size are examined (Table 5), but a homogenization time above 25 min had a negative influence Al SI/a.u. 67.7 23.3 22.6 26.7 40.1 37.5 on the RSD values. The distribution of grain size changed RSD (%) 11.9 7.8 10.3 7.2 10.0 11.6 from a typical normal distribution to a skewed one in the Fe SI/a.u. 39.3 22.5 20.6 21.3 35.0 48.5 direction of smaller particles. This fact can explain the worsen- RSD (%) 9.1 5.6 7.0 7.1 13.6 10.0 ing of the RSD values, especially after the second homogeniz- Ni SI/a.u. 24.8 15.5 15.1 18.5 34.2 65 RSD (%) 13.7 4.2 5.7 6.3 16.9 12.1 ation, due to the inhomogeneity of the sample. These Ti SI/a.u. 72.7 34.6 37 41.7 63.4 79.8 conclusions are in good agreement with the calculated data of RSD (%) 17 4.0 9.4 7.2 9.9 9.9 the so-called span:13 V SI/a.u. 41.4 20.2 21.6 27.2 46.4 47.2 RSD (%) 19.9 10.2 10.9 5.6 10.8 8.5 span= d90-d10 d50 (i=1-N) data of K=5 repeated measurements were averaged which are 1.30–1.46–1.92 for the sample with larger grain size for the series of measurements of calibration samples]; (SiC-2495) and 0.38–0.73 for the sample with smaller grain (ii) the relative precision of the method (RSDMETHOD) size (SiC-3757).The signal values did not show any dependence calculated according to the German norm;11 on the homogenization, with the exception of Ti, where the (iii) the detection limit (LOD), calculated12 as the upper doubling of the signal can be explained only by the release of limit of the expected range of the intercept (A) of the calibration Ti from the large grains of the SiC skeleton (in the case of line (y=A+Bc).fine SiC powder, this increase was not observed). The dependence of the signal values on the amount of Results and discussion sample weighed into the electrode cavity is, as expected almost linear; the diVerent evaporation rates of the diVerent amounts Optimization of instrumental parameters of SiC samples is the basis of the diVerent RSD values.These are better when samples with smaller grain size or samples The optimum observation position is element dependent (Table 4) and the two chosen parameters (signal/RSD) contra- after homogenization are used (Table 6). The slit width of the spectrometer can also have an influence dict each other. The maximum signal values were obtained at the observation position near to the counter-electrode (cath- on the tested parameters, e.g. the signal value and its reproducibility. A decrease in the slit width led, for some elements, to ode), due to the well known cathode-layer eVect, but they are associated with the highest RSDs. The optimal RSDs were typical minimum RSD values at a slit width of 1.5 mm Table 5 The dependence of the integrated intensity means and RSDs on the homogeneity of the SiC samples with various grain sizes (sample weight=6 mg) Homogenization of the sample 0 min 25 min 2×25 min SiC sample Element SI/a.u.RSD (%) SI/a.u. RSD (%) SI/a.u. RSD (%) 2495 Al 42.6 8.3 47.0 5.3 48.4 4.3 Fe 48.3 10.0 52.1 6.6 51.0 12.0 Ni 36.1 19.0 41.5 5.9 39.7 15.0 Ti 37.4 16.0 71.7 5.7 64.3 14.0 V 40.9 22.0 47.0 7.8 45.7 23.0 d50/mm 54–59 36.5 25.1 3757 Al 20.4 4.3 22.2 5.4 Fe 18.5 5.3 19.7 3.5 Ni 12.6 4.7 13.4 4.7 Ti 36.5 8.5 32.2 4.3 V 25.6 7.5 22.7 5.9 d50/mm 10.8–13.3 10.6 J. Anal. At. Spectrom., 1999, 14, 559–564 561Table 6 The dependence of the integrated intensity means and RSDs on the amount of SiC sample weighed into the cavity of the carrier electrode Amount of sample 9 mg 6 mg 3 mg 1.5 mg SiC sample Element SI/a.u.RSD SI/a.u. RSD SI/a.u. RSD SI/a.u. RSD (%) (%) (%) (%) 3757 Al 37.7 8.1 20.4 4.3 11.5 14.0 6.2 13.0 (10.8–13.3 mm) Fe 26.3 7.1 18.5 5.3 14.0 13.0 11.2 8.9 Ni 17.5 5.4 12.6 4.7 9.7 12.0 7.2 10.0 Ti 47.7 3.6 36.5 8.5 20.9 9.7 14.0 9.8 V 39.2 7.5 25.6 7.5 12.6 8.7 6.6 15.0 2495 Al 73.0 11.2 42.6 8.3 20.7 3.6 (54–59 mm) Fe 76.9 13.1 48.3 10.0 26.4 13.3 Ni 56.5 19.5 36.1 18.5 21.2 12.0 Ti 95.9 30.3 37.4 15.5 37.3 27.5 V 72.4 21.0 40.9 21.9 18.6 18.9 2495 Al 80.8 2.3 47.0 5.3 22.8 8.3 (25 min homog.) Fe 79.1 6.0 52.1 6.6 28.2 13.0 (36.5 mm) Ni 62.2 8.0 41.5 5.9 22.1 17.0 Ti 95.4 10.0 71.7 5.7 35.9 19.0 V 80.6 6.8 47.0 7.8 15.6 46.0 Table 7 The dependence of the integrated intensity means and RSDs on the slit width of the spectrometer (SiC-3757, 6 mg, n=5) Slit width/mm 2.5 2.0 1.5 1.0 Element SI/a.u.RSD SI/a.u.RSD SI/a.u. RSD SI/a.u. RSD (%) (%) (%) (%) Al 46.6 4.7 27.3 5.7 38.4 5.7 12.5 15.1 Fe 40.7 5.6 26.3 3.5 28.3 3.9 8.2 8.3 Ni 33.6 5.4 22.1 1.7 20.2 5.3 5.3 42.0 Ti 78.5 5.7 52.1 3.1 52.0 2.9 16.9 11.0 V 50.8 9.0 33.9 10.0 34.9 4.3 12.1 20.0 Table 8 The dependence of the integrated intensity means and RSDs on the observation mode used (10 repeated measurements; SO, single optics=only one glass fibre optic; DO, double optics=two, 90° oriented, fibre optics; see also Fig. 1) SiC Sediment SO (m=8 mg) DO (m=8 mg) SO (m=6 mg) DO (m=3 mg) Element SI/a.u.RSD SI/a.u. RSD SI/a.u. RSD SI/a.u. RSD (%) (%) (%) (%) Al 12.4 12.0 18.2 12.0 B 23.0 7.1 45.9 7.8 Ca 9.5 11.0 14.9 9.5 Cr 72.0 3.2 89.7 4.7 Cu 5.1 9.5 8.6 8.5 72.0 4.5 79.0 6.1 Fe 23.2 7.8 37.1 7.3 Mg 3.7 8.2 6.6 7.0 Ni 16.9 7.0 31.9 7.6 53.8 7.9 49.9 6.8 Pb 14.9 8.5 19.5 9.9 Ti 17.7 11.0 30.2 11.0 V 30.2 15.0 48.6 14.0 68.6 8.7 80.0 8.1 Zn 21.5 11.0 21.1 6.7 (Table 7), but without a common trend. The highest signals about a factor of 1.5–2.0, but there were no significant diVerences in the RSDs.This allows the conclusion to be were obtained at a slit width of 2.5 mm with, on average, acceptable values of RSD. Therefore this value was used in drawn that the stability of arc is high and the so-called ‘dancing of arc’ is totally eliminated in the electronically all subsequent experiments. The arrangement used for optical signal processing allows controlled dc arc. In the second part of the experiment, in order to obtain similar signal values in the single optic the signal dependence and arc stability to be studied.When the same amount of SiC sample was excited and the signals arrangement, two fold higher amounts of the sediment sample werepplied. Similar signal values were characterized in this were detected first using only one (SO) quartz fibre and then using both (DO) quartz fibres, the following results were case with no significant diVerences in RSDs.These experiments confirm the perfect function of the quartz fibre coupling. obtained (Table 8): the signal values increased, on average, by 562 J. Anal. At. Spectrom., 1999, 14, 559–564Table 9 The figures of merit for calibration with a single calibration sample (SiC-628: concentration of elements, see Table 2) at diVerent weights (12, 8, 4, 2 and 1 mg) in the carrier electrode cavity Element Parameter Al B Ca Cu Fe Ni Ti V RSDa (%) 12.7 8.1 14.6 12.9 9.9 11.3 14.8 14.1 Norm./homosc.(+/-) +/- +/+ +/- +/+ +/+ +/+ +/+ +/- r 0.989 0.998 0.976 0.994 0.998 0.995 0.991 0.991 A -0.44 4.80 4.30 1.99 6.90 10.2 8.60 -3.90 B 0.118 1.48 0.780 3.10 0.112 0.327 0.394 0.220 RSDMETHOD (%) 9.4 5.8 19 9.6 5.1 8.1 12 11 FLIN (a=0.01) + + + + + + + + LOD/mg g-1 39 3.6 6.7 0.5 33 13 16 72 Table 10 The figures of merit for calibration with a single calibration sample (SiC NMP-6: concentration of elements, see Table 2) at diVerent weights (12, 8, 4, 2 and 1 mg) in the carrier electrode cavity Element Parameter Al B Ca Cu Fe Ni Ti V RSDa(%) 11 12 13 13 13 14 16 15 Norm./homosc.(+/-) +/+ +/+ +/+ +/- +/+ +/+ +/- +/- r 0.997 0.999 0.980 0.998 0.999 0.999 0.995 0.995 A -0.11 2.15 10.7 1.77 4.53 10.3 8.21 0.52 B 0.138 1.52 0.420 2.91 0.119 0.309 0.391 0.179 RSDMETHOD (%) 6.9 3.0 17 5.2 4.5 2.3 8.5 8.6 FLIN (a=0.01) + + + + + + + + LOD/mg g-1 62 0.9 20 0.4 49 4.7 24 29 Table 13 The figures of merit for calibration with a single sediment Table 11 The figures of merit for calibration with four sediment RMs (GBW-07305, -07309, -07312, BCR-277: concentration of elements, RM (SL-1: concentration of elements, see Table 3) at diVerent weights of the sample (12, 8, 6, 4, 2 and 1 mg) in the carrier electrode cavity see Table 3) at two diVerent weights (4 and 2 mg) in the carrier electrode cavity Element Element Parameter Cu Cr Ni Pb V Zn Parameter Cu Cr Ni V Zn RSDa (%) 7.0 10 11 10 10 11 Norm./homosc.+/+ +/+ +/+ +/+ +/+ +/+ RSDa (%) 7.8 7.3 5.9 6.8 6.0 Norm./homosc.(+/-) +/+ +/- +/- +/+ +/+ (+/-) r 0.962 0.986 0.976 0.969 0.983 0.996 r 0.975 0.997 0.978 0.966 0.959 A 19.9 0.946 12.7 1.20 4.82 A 10.6 6.2 30.9 9.1 7.5 9.3 B 0.90 0.67 0.70 0.22 0.58 0.07 B 0.722 0.976 0.484 0.535 0.026 RSDMETHOD (%) 14 4.0 13 14 15 RSDMETHOD (%) 20 11 15 17 14 6.8 FLIN (a=0.01) + + + + + + FLIN (a=0.01) + + + + + LOD/mg g-1 23 5.3 7.4 28 261 LOD/mg g-1 8.2 15 13 10 30 19 Table 12 The figures of merit for calibration with eight sediment RMs preparation of calibration sets.With the exception of two (concentration of elements, see Table 3) and a weight of 6 mg in the analytes (Ca and V), no significant diVerences in the main carrier electrode cavity figures of merit were observed (Tables 9 and 10) in some cases, the homogeneity of variance was not confirmed and a Element revision of the data sets or the use of the weighted least- Parameter Cu Cr Ni V Zn squares method in future optimization is needed.Both the RSDa and RSDMETHOD values, defining the precision, are RSDa (%) 13 16 12 10 8.5 acceptable for the direct solid sampling method, but the values Norm./homosc. +/- +/+ +/+ +/+ +/+ obtained for analytes Ca, Ti and V need some optimization. (+/-) The adequacy of the linear model used was confirmed in all r 0.996 0.959 0.962 0.992 0.993 cases, with the exception of Ca and V, there are no significant A 16.7 10.3 27.3 6.5 23.1 B 0.340 0.470 0.460 0.130 0.220 diVerences in the parameters of the calculated straight lines.RSDMETHOD (%) 5.0 12 13 11 5.5 The LOD values are, with the exception of Al, also acceptable, FLIN (a=0.01) + + + + + because the concentrations of the elements lie above these LOD/mg g-1 10 33 15 30 18 limits in commercial SiC products which should be controlled. For the calibration of sediments, three diVerent kinds of calibration process were used. Firstly (Table 11), a calibration Calibration and method validation with four diVerent RMs at two diVerent weights in the carrier electrode; secondly, a calibration with eight diVerent RMs (see As mentioned, the analytical calibration was carried out using two typical samples: SiC and sediments.For SiC calibration, Table 12); thirdly, only a single RM (SL-1) at six diVerent weights (Table 13). In most of these cases, all three basic tests only the method using diVerent weights of a single calibration sample was employed because of a lack of suitable RMs.Two (normality of data sets, homoscedasticity and linearity) were confirmed, but there are significant diVerences in the param- SiC samples with diVerent grain sizes were used for the J. Anal. At. Spectrom., 1999, 14, 559–564 563eters (A and B) of the straight lines. These diVerences could Acknowledgements be caused by classical matrix eVects, due to the diVerent origin This work was supported by the Slovak Grant Agency VEGA, of the RMs having SiO2 as the main component, but also the Grant.No 1/4365/97, and by common Slovak–German coop- type of species in which the elements are present can have eration project SLAX 262.8. some eVect. Future experiments are needed to answer these questions. The RSD values above 10% should also be revised, but values of about 5% (Cr, Cu, V) are fully acceptable. The References LOD values depend partially on the ‘goodness of fit’ of the 1 K. Zimmer, K. Flo� ria�n and Gy. Heltai, Prog.Anal. Atom. calibration lines; on average they are acceptable, because the Spectrosc., 1982, 5, 341 (and references cited therein). lowest concentrations of the elements in sediments from 2 D. Lummerzheim and H. Nickel, Z. Anal. Chem., 1969, 245, 267. Slovakia are: 26 mg g-1 for Zn, 2.2 mg g-1 for Cr, 5.8 mg g-1 3 M. Todorovic�, V. Vukanovic� and V. Georgijevic�, Spectrochim. for Pb and 5.2 mg g-1 for Cu. Acta, 1969, 24B, 571. 4 D. Vukanovic�, V. Simic�, V. Vukanovic�, H. Nickel and M. Mazurkiewicz, Spectrochim. Acta, 1977, 32B, 305. 5 D. C. Baxter and W. Frech, Spectrochim. Acta, 1995, 50B, 655. 6 Gy.Za� ray and T. Ka�ntor, Spectrochim. Acta, 1995, 50B, 489. Conclusions 7 H. Nickel and Z. Zadgorska, Spectrochim. Acta, 1995, 50B, 527. Using typical samples for solid sampling spectrometry, a 8 H. Muller, M. Mazurkiewicz and H. Nickel, Ber. Kernforschungsanlage Ju�lich, Nr. 1449, 1977. complex methodical evaluation of dc arc spectrometry was 9 K. D. Ohls, Spectrochim. Acta, 1996, 51B, 245. made. Dc arc OES is a possible alternative to the modern, but 10 K. Danzer and L. A. Currie, Pure Appl. Chem., 1998, 70, 993. more complicated, methods of solid sampling analysis, because 11 DIN 32645, 1994. of its simple handling and lack of need for sample preparation 12 V. Damman, G. Donnevert and W. Funk, Quality Assurance in or the use of special working atmospheres to achieve complete Analytical Chemistry, Verlag Chemie,Weinheim, 1995. evaporation of the analytes. The figures of merit obtained 13 W. Mutter, Int. Lab. News, 1993, 12, 26. show the need for further optimization steps, directed in particular to the special problems of classical matrix eVects. Paper 8/07088A 564 J. Anal. At. Spectrom., 1