JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 257 Direct Solid Sample Analysis of Silicon Carbide Powders by Direct Current Glow Discharge and Direct Current Arc Emission Spectrometry* K. Florian TU Kosice Department of Chemistry Letna 9 SK-0400 I Kosice Slovakia W. Fischer and H. Nickel Research Centre Jiilich GmbH Institute of Materials for Energy Systems P. 0. Box 19 13 0-52425 Julich Germany Glow discharge atomic emission spectrometry (GD-AES) and the classical spectrometric method with a d.c. arc source (D.c.-arc-AES) were applied to the direct solid sample analysis of SIC powders. The homogenity of pellets prepared from mixtures of Sic with copper powder and used for the GD experiments was investigated in detail. Various methods were tested for the analytical calibration and applied to the direct analysis of technical Sic powders. The experimental data were evaluated by using chemometric procedures.The results of the two methods were compared. Keywords Glow discharge atomic emission spectrometry; direct current arc emission spectrometry; direct solid sample analysis; silicon carbide powder; calibration with model standards Ceramic materials based on silicon carbide are used for such applications as slidings seals soot filters heat exchangers and for other high temperature applications. The mechanical behav- iour of the components sintered from Sic powders are influ- enced by the purity of the powder itself. Impurities of technical importance in Sic powders are Al Fe V Ti Ni Ca Cr and B. They are introduced by the starting materials as well as in the production process.There is a substantial need for the Sic manufacturers to check the evolution of contamination during the production and to certify the products by a simple and quick multi-elemental routine analysis with good precision accuracy and limit of detection (LOD). Spectrometric analytical methods have been used for the quantitative determination of the impurities mentioned in ceramic materials.’-3 The application of a conventional dissolu- tion method for sample preparation is limited for refractory compounds like Sic owing to their resistance to chemical attack. This resistance increases the potential danger of intro- ducing additional impurities during sample preparation. Therefore techniques for the direct analysis of the solid sample are preferred.Direct current glow discharge atomic emission spectrometry (GD-AES) utilizing the sputtering process for the destruction of the sample and d.c. arc emission spectrometry (d.c.-arc- AES) were applied for the direct analysis of Sic powders. Both methods need a certified reference material (CRM) for analyt- ical calibration. In general such CRMs are not available. Therefore two types of calibration samples were prepared (i) the matrix was modelled as an equimolar mixture of silicon and graphite powders (Si+C); and (ii) superpure silicon car- bide powder was used as a matrix. The analytes were added to both matrices as oxides of spectral purity. Furthermore a set of technical Sic powders and the National Institute of Standards and Technology (NIST) Standard Reference Material (SRM) 112b Silicon Carbide all well characterized by various methods including wet chemical analysis were used for checking the accuracy of both spectro- metric methods.A test of the homogeneity of the calibration samples com- parison of the results of the calibration of GD and d.c.-arc- AES and calculation of the precision accuracy and LOD values are the subject of this study. * Presented at the XXVIII Colloquium Spectroscopicum Inter- nationale (CSI) York UK June 29-July 4 1993. Experimental Samples and Sample Preparation Two types of matrices were prepared for analytical calibration (i) an equimolar mixture of Si powder (Merck No. 12497 > 99% Si < 120 pm) and graphite (Ringsdorff type RW-A/T >99.9999% C <60 pm); and (ii) superpure Sic powder a special product of the Elektroschmelzwerk Kempten Germany (made as a reference material for a round robin test of the purity of Sic used for technical purposes; the concentrations of the impurities of interest were kept some orders of magnitude lower than usual).The analytes were added to these matrices as oxides of spectral purity in the concentrations given in Table 1. The mixtures were homogenized by grinding in an agate mortar for 5 min. For the GD-AES measurements the calibration mixtures were diluted with copper powder (Merck No.2704 3 99.9% 60 pm) in a ratio of 3:7. In the next step they were compacted into pellets in a hydraulic press at 800 MPa for 120 s together with pure copper powder.’ The arrangement of the calibration mixture and the surrounding copper powder in the pressing tool was such that the calibration mixture formed a centered inlet (10 mm diameter 1 mm in thickness) on one side of the pellet surrounded and mechanically stabilized by a copper ‘holder’. For the d.c.-arc-AES investigations graphite powder and the spectrochemical additive CoF2 + Ba(N0,)26 were added to the calibration mixture in the ratio 1 1 1.A 15 mg portion of the mixture was used for each analysis after intense homogenization. Table 2 lists the impurity concentrations in the set of techni- Table 1 Analyte concentrations (YO) in calibration samples si+c Sic Sample No. Al Fe V Ca B Cr Al Fe V B Cr Ca 1 1.0 0.3 16 1 .o 2 0.7 0.2 0.7 3 0.316 0.1 0.316 4 0.1 0.0316 0.1 5 0.03 16 0.01 0.05 6 0.01 0.03 16 -258 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL.9 Table 2 Set of technical Sic powders used for check of the calibration accuracy (RSDs not available) Concentralion (%) Analyte A1 Fe V B Cr Ca Grain size/pm F-400 0.12 0.31 0.1 1 0.01 0.03 1 0.16 17 F-320 0.025 0.030 0.060 0.008 Unknown 0.003 1 27-30 ‘EXTRA’ 0.128 0.0364 0.0346 Unknown 0.0005 0.003 7 z 60 NMP-1 0.010 0.0271 0.012 Unknown 0.0005 0.0005 < 15 NMP-2 0.034 0.007 0.0185 Unknown Unknown 0.0006 < 35 NIST 0.34 0.1 1 0.019 Unknown Unknown 0.17 Unknown cal SIC powders and NIST SRM 112b. These data have been used to check the accuracy. GD-AES Measurements The experiments were performed with the commercial equip- ment SPECTRUMAT 1000s (Leco Germany); some technical details are summarized in Table 3.For data on the analytical lines used see Table 5. To avoid fluctuations of the emission intensity caused by air leakages at the sample gasket or by the porosity of the pellet itself an evacuated cup was used.2 Fig. 1 shows a schematic diagram of the slice-by-slice removal of the centered inlet of a pellet during each GD measurement. During the compacting the hard constituents of a calibration mixture are pressed into the softer components Table 3 GD-AES equipment Source Discharge tube Discharge parameters Discharge atmosphere Spectrometer Polychromator Entrance slit Exit slits Grating Dispersion range Detectors Grimm-type 8 mm diameter Constant voltage mode; U = Ar 99.99%; mass flow 300 water-cooled cathode plate lo00 V; Current z 60 mA ml min-’ (pAr ~ 0 .9 hPa) Paschen-Runge configuration; focal length 1000 mm evacuated chamber 50 pm 10 pm Holographic 2 160 grooves mm - 130-430 nm (first order) D= Hamamatsu photomultiplier 0.34 nm mm-’ special entrance windows for vacuum UV region Pellet weighed Fig. 1 Schematic diagram of the slice-by-slice removal of a GD pellet of the pellet so the first slices near to the surface of a new pellet are not representative of the whole inlet. Therefore these upper slices were removed during the first three unevaluated measurements from M(-2) to M(0). The sputtering time for each slice M(i) was fixed to 30 s. Nine slices i.e. nine measure- ments were made on each pellet. Integrated intensities were used for all further analytical evaluations. The nine measurements are arranged into three groups (see Fig.1) the upper group (U) representing the mean of the first three slices M(l)-M(3) the middle (M) for the next three and the lower group (L) for the last three slices. These mean values U M and L were used for the statistical test of the homogeneity of the pellets. D.c.-arc-AES Measurements The experimental conditions for the d.c.-arc experiments are given in Table 4. The analytical lines used for the calibration (Table 5 ) are not identical with those of the GD-AES measure- Table 4 D.c.-arc-AES spectrograph Source Arc discharge Excitation Electrodes Spectrograph Grating spectrograph Entrance slit Grating Dispersion range Detection/evaluation Photographic plate Exposition time Spectra evaluation UBI-1 (Zeiss Jena) Free burning d.c.arc; 4 mm electrode gap; carrier electrode as anode; 1=8.5 A High resistance carbon electrodes in water-cooled holders carrier electrode type SW 380 counter electrode type SW 202 (Elektrokarbon Topolcany Slovakia) PGS-2 (Zeiss Jena) focal length 20 pm 651 grooves mm-’ 250-330 nm (second order) 2000 mm D=0.36 nm mm-’ ORWO WU-3 80 s (total evaporation) Modernized 4D densitometer 0 < S < 4 1 transformation Table 5 Analytical lines (wavelengths in nm) Analyte A1 Fe V B Cr Ca ~ ~~~ GD-AES D.c.-arc- AES A1 1396.15 A1 1308.22 Fe I1 249.32 Fe I 302.11 V 1411.18 V 1318.34 B 1208.94 B 1249.68 Cr I 302.16 No sensitive UV line Cr T 425.43 Ca I1 393.37JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 259 ments. The two devices are equipped with spectrometers with different dispersion ranges.Calibration was provided on the basis of a 5-fold measure- ment of each calibration sample. The precision was calculated from ten replicate measurements for two selected calibration samples (see Table 6). Results and Discussion Homogeneity Tests The first test concerns the homogeneity of the elemental distribution within a GD pellet by comparison of groups of slices. The statistical check was performed according to Lord‘s ~riterion.~ This test utilizes the largest difference AI, of the emission intensities within each group of slices instead of the standard deviations. This type of testing is useful if only a small number of repeated measurements is available (in the present case n = 3). The test criterion is the value u calculated from eqn.(1 ) where iA and IB are the averaged emission intensities for any analyte in the groups tested. The value of u has to be compared with its counterpart UTab (ref. 7) UTab = 0.636 for a statistical precision of 95% and n = 3; if u < UTab then the groups compared can be considered homogeneous. The test results of one arbitrary GD calibration pellet are given in Table 7. The test results confirm the homogeneity of the pellet in the first two groups U and M for most analytes. It reveals a general inhomogeneity of Fe and B. Therefore correct analyt- ical results can be expected only if the evaluation is limited to the first three slices i.e. to the group U. The homogeneity of different pellets prepared from the same calibration mixture was tested in the same manner.The mean value of the first groups U of the different pellets have been used for the comparison. The results are presented in Table 8. The numbers 1 2 and 3 stand for arbitrary identifiers of different pellets. Table 6 RSD (YO) values for two arbitrarily selected calibration samples (n = 10) sample 1 sample 2 Analyte c (%) GD D.c.-arc c (%) GD D.c.-arc A1 0.316 1.2 20 0.1 5.2 5.7 Fe 0.316 9.7 24 0.1 30 6.6 V 0.316 2.3 32 0.1 3.4 7.3 B 0.1 10 15 0.0316 16 7.6 Cr 0.1 1.8 27 0.0316 5.8 20 Ca 0.316 16 - 0.1 19 - Table 9 RSDs (YO) of repeated d.c.-arc measurements Calibration sample Analyte 1 2 3 4 5 6 A1 24 20 5.7 10 30 1 1 Fe 25 24 6.6 17 10 25 31 32 7.3 29 19 - V B 19 15 26 7.5 6.4 12 Cr 31 27 20 20 8.8 20 The results of this second test do not show a clear tendency.Therefore it is assumed that the pellets are equivalent to each other. Owing to the total evaporation of a calibration sample in a d.c.-arc measurement a similar test to that used for GD is excluded. Consequently the relative standard deviations (RSDs) of repeated measurements must be compared. These RSD data are in good agreement with the data given in Table 6 for the precision of the d.c.-arc method. It can be concluded therefore that the d.c.-arc calibration mixtures are homogeneous. Analytical Calibration Owing to the limitation of the dynamic range of the detection system of the SPECTRUMAT 1000s to two orders of magni- tude the intensity ratios y = IJIsi have been plotted directly against the concentration (c) of the analytes.A straight line [eqn. (2)] was fitted to these data by linear regression Y=A+Bxc (2) For the d.c.-arc measurements with photographic registration and the evaluation of the plate blackening over four orders of magnitude a double logarithmic plot of log(y) uersus log(c) is more convenient [eqn. (3)] (3) The results of the analytical calibration are listed in Table 10 together with the so-called ‘expected precision’ RSD (c)(%) of the concentration calculated according to eqn. (4) log(y) = A’ + B’ x log(c) RSD(c) = (sreS/B) x J( 1/N + l/n) x 100 (4) where s, is the residual standard deviation after linear regression N the number of calibration samples and n the number of repeated measurements. Table 11 gives the LOD cL calculated using the 3s criterion from blank measurement data (15 repetitions).Figs. 2-12 show the calibration plots. Calcium could not be determined with the d.c.-arc owing to the absence of a sensitive emission line within the dispersion region of the PGS-2 spectrometer. Table 7 Results of the homogeneity test within a GD pellet (U M and L indicate the groups of slices compared + identical; - different; and x irrelevant comparison) ~~ A1 Fe V B Cr Ca U M L U M L U M L U M L U M L U M L U x + - x - - x + + x - - x + - x + - x - + x - + x - M + x - X X X X + - L X - x - + x - - - - - - - - - - - - Table 8 Pellet A1 Fe V B Cr Ca Results of the homogeneity test of different GD pellets (1 2 and 3 are arbitrary pellet identifiers); symbols as in Table 7 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 2 3 1 x - + x + + x - + x + - x - + x + + x - + x - 2 - x + + x + 3 + + x + + x + + x - - X + - X + - x - x + + x - -260 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL.9 Table 10 Numerical results of the analytical calibration GD-AES Si+C Sic Analyte A B RSD (c) (%) A B RSD (c) (Yo) A1 0.060 1.870 2.3 0.023 2.763 2.9 Fe 0.052 0.323 5.4 0.042 0.366 1.5 v 0.023 0.480 2.1 0.038 0.492 0.7 B 0.008 2.158 0.7 Cr 0.117 5.852 1.4 0.145 6.098 0.7 Ca 0.104 1.828 2.4 0.061 2.823 0.6 - - - D.c.-arc-AES ~ ~~~ s i + c Sic Analyte A' B' RSD (c) (YO) A' B' RSD (c) (YO) A1 0.84 0.65 8.8 0.60 0.48 27 Fe 0.35 0.53 18 0.38 0.76 29 v 0.74 0.89 20 0.63 1.08 18 - 0.38 0.43 25 B Cr 0.416 0.66 18 0.57 1.07 7.8 - - Table 11 Limits of detection (cL) in pg g-' Analyte A1 Fe V B Cr Ca GD-AES - Si+C Sic 50 98 557 765 646 325 44 8.5 10 27 106 - 0.4 1 + 0.3 -* 3 0.2 0.1 LOD -* I I I I I I 0 0.2 0.4 0.6 0.8 1 .o LFel (YO) Fig.3 GD calibration graph of Fe in B Si+ C; A Sic; 0 Sic F320; and + SIC F400 matrices 0.6 I 1 0.5 D.c.-arc- AES 0.4 ~~ Si+C Sic 0.3 0.3 1 .o 15 4.6 23 5.1 56 - - - - < 0.3 -2' 0.2 0.1 LOD + ~~ 0 0.2 0.4 0.6 0.8 7 .O [VI (%I Fig. 4 GD calibration graph of V in Si + C; A Sic; 0 Sic F320; and + SIC F400 matrices A 1.0 1 0 0.2 0.4 0.6 0.8 1 .o [All (%I Fig.2 GD calibration graph of A1 in B Si+C; A SIC; 0 Sic F320; and + Sic F400 matrices The differences in the parameters A and B of the GD calibration curves of the two matrices are small (Table 10). Further conclusions from the values of A should not be made since they correspond to an unallowed extrapolation below the lower calibration limit.They include the influence of the impurities in the diluent and the substances used for matrix modelling as well as the error of the linear regression. A more detailed discussion of this low concentration range requires further experiments to be carried out. The calibration results confirm the general experience that in GD-AES the matrix effect is either small or completely absent. Boron is the only exception its calibration curve being non-linear. Neither additional contamination of the pellets introduced during the preparation nor a line overlap can be the reason for this deviation. It may be that chemical reactions in the discharge 0 0.1 0.2 0.3 0.4 [Bl (Yo) Fig. 5 GD calibration graph of B in .Si +C; A Sic; 0 Sic F320 matrices plasma e.? the formation of B-C compounds are responsible. In contrast the d.c.-arc calibration results show a pro- nounced influence of the sample matrix (see results for Fe V or Cr). The slope of the calibration curves becomes stronger when the Si+C matrix is substituted by Sic. The changes in the A' value are negligibly small. The stronger slopes of the SIC calibration curves are important especially in the case of higher analyte concentrations. This matrix effect calls for26 1 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 0 0.2 [Crl (%) 0.4 Fig. 6 GD calibration graph of Cr in . Si+C; A Sic; 0 Sic F320; and + Sic F400 matrices 2 2 ,u . 1 LOD -+ 0 0.2 0.4 0.6 0.8 1 .o [Cal (%I Fig.7 GD calibration graph of Ca in W Si+C; A Sic; 0 Sic F320; and + Sic F400 matrices 1 -1 1 I I 1 -3 -2 -1 0 Log [Cn (%)I Fig.8 D.c.-arc calibration graph of A1 in m Si +C; A Sic; x Sic NMP-1; 0 Sic F320; A SIC NMP-2; + Sic F400; 0 Sic NIST; and 0 Sic 'extra' matrices further investigation of the evaporation behaviour of the calibration samples. To date the spectrochemical additives have been optimized only with respect to a quick and complete destruction of the Sic crystal lattice3 and total sample evapor- ation necessary for quantitative analysis. The additives were not optimized from the point of view of the matrix effect. The fit of a straight line to the GD calibration data is better than that to the d.c.-arc data. In d.c.-arc the correlation I I J -3 - 2 -1 0 Log [CFe (%)I Fig. 9 D.c.-arc calibration graph of Fe in 1 Si + C; A Sic; A Sic NMP-2; x Sic NMP-1; 0 SIC F320; + Sic F400; 0 SIC NIST; and 0.Sic 'extra' matrices -3 - 2 - 1 Log [C" (%)I 0 Fig. 10 D.c.-aRC calibration graph of V in m Si +C; A Sic; x SIC NMP-1; A Sic NMP-2; 0 Sic NIST; 0 SIC 'extra'; + Sic F400; and 0 Sic F320 matrices 0.5 I - 0.5 -1 I I I -3 - 2 -1 0 Log Ice (%)I Fig. 11 matrices; the arrow indicates the B contamination level Attempt at d.c.-arc calibration of B in . Si+C and A Si coefficients of the linear regression ranged from 0.95 (B) to 0.99 (Al Fe) compared with >0.99 for all analytes in GD. The 'expected precision' of the determination of the concen- trations in GD is significantly better than in d.c.-arc. Possible reasons for this finding are the more even sample atomization in the sputtering process combined with the photoelectric detection of the analytical signal.The LODs for the two262 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 -2.0 L I I I - 3 -2 - 1 0 Log [cc (%)I Fig. 12 D.c.-arc calibration graph of Cr in W Si+C and A Sic matrices 0.05 - s - -0 0.04 .- E 0.03 -0 z 0.02 0 .- .I- E E 0.01 a u 0 u 0 Fig. 13 0.75 (a) N M P-2 I I I I 0.5 0.25 0.01 0.02 0.03 0.04 0.05 0 0.25 0.5 0.75 Given concentration of Al (YO) Comparison of the accuracy of the analytical d.c.-arc results ofvarious Sic powders using calibration data of both the Si+C (0) and the Sic (+ 1 matrices methods (see Table 11) show opposite tendencies of the detect- ibility of an analyte and the precision of its determination. Obviously the precision of the d.c.-arc method is worse than that obtained with GD but its advantage is the low LOD.First experiments with improved equipment where the photo- graphic plate is substituted by a photoelectric detection system indicate that a higher precision might be achievable with the d.c.-arc also.g The accuracy of the GD analyses could be tested only for the Sic powder F-400 because of the relatively high consump- tion of material in the preparation of pellets. The RSDs between the determined and the given concentrations ranged from -44% for Fe to +50% for Al. The sample consumption of d.c.-arc measurements (15 mg per measurement) is considerably lower than in GD. Therefore six Sic powders were tested. The Yuden plot7 of Al shown in Fig. 13 is representative of the other analytes too.In the worst case the deviation between the determined and the given value increases to 80%. Conclusions The present study demonstrates that both GD-AES and d.c.- arc-AES can be used for the direct solid sample analysis of impurities in Sic powders. The limitation of the investigation to the impurity elements discussed does not reflect a limitation of the method but rather of the spectroscopic equipment available. Samples for calibration can be prepared from powder mixtures consisting of Si+C or Sic with additions of the analytes. The homogeneity of the calibration samples is ensured if the preparation procedure described is applied. Both methods differ in their precision and their LODs; the precision of the d.c.-arc method is not as good as that achieved with GD-AES but the LODs are 1-2 orders of magnitude lower.No matrix effect was observed in the GD measurements. The pronounced matrix effect in the d.c.-arc measurements requires further studies especially in relation to the influence of chemical modifiers. The accuracy of the analyses is dependent on the analyte and its concentration level. It is believed that the accuracy achieved is acceptable for the routine certification of technical Sic powders. The present investigation was carried out as a sub-task of the contract on scientific-technical cooperation between Germany and Czechoslovakia. The authors thank the ministries of both countries for financial support. Furthermore the authors thank Mr. J. Hassler (Elektroschmelzwerk Kempten Germany) and Mr. Dr. G. Wolff (Research Centre Julich GmbH Germany) for making available the Sic materials investigated. References Broekaert J. A. C. and Tolg G. Mikrochim. Acta (Wien) 1990 11 173. Ehrlich G. Stahlberg U. and Hoffmann V. Spectrochim. Acta Part B 1992 46 115. Florian K. Nickel H. and Zadgorska Z. Fresenius J. Anal. Chem. 1993,345,445. Hassler J. personal communication. Guntur D. S. Fischer W. Mazurkiewicz M. Naoumidis A. and Nickel H. Reports of the Research Centre Julich Jiilich Germany Jul-Report No. 2592 1992. Nickel H. Zadgorska Z. and Wolff G. Spectrochim. Acta Part B 1993 48 25. Eckschlager K. Horsack I. and Kodejs Z. Evaluation of Analytical Results and Methods SNTL Publishing Company Prague 1980 p. 43 (in Czech). Fischer W. Naoumidis A. and Nickel H. paper presented at the XXVIII Colloquium Spectroscopicum Internationale (CSI) Post- Symposium on Analytical Applications of Glow Discharge in Optical and Mass Spectrometry York UK July 4-7 1993. Florian K. Fischer W. Hassler J. and Nickel H. paper presented at the XXXVI Hungarian Annual Conference on Spectral Analysis Lilafiired Hungary August 24-27 1993. Paper 3103941 B Received July 7 1993 Accepted November 15 1993