JOURNAL O F ANALYTICAL ATOMIC SPECTROMETRY MARCH 1993 VOL. 9 463 Determination of Silicon in Titanium Dioxide and Zirconium Dioxide by Electrothermal Atomic Absorption Spectrometry Using the Slurry Sampling Technique* Susanne Hauptkorn Germar Schneider and Viliam Krivant Sektion Analytik und Hochsfreinigung der Universitat Ulm AIbert-Einstein-Allee 7 7 0-89069 Ulm Germany A method for the determination of silicon in titanium dioxide and zirconium dioxide powders based on electrothermal atomic absorption spectrometry using the slurry sampling technique has been developed. The experimental conditions with regard to chemical modification the temperature programme the lifetime of the graphite furnace and the slurry concentration were optimized. The behaviour of zirconium and calcium in the graphite tube during the ashing atomization and cleaning steps was investigated using 47Ca and "Zr as radiotracers.Cali bration was performed by the standard additions method using aqueous standards. The results of this technique were compared with those for atomic emission spectrometry. The limits of detection were found to be 7 pg g-' in titanium dioxide and 2 pg g-' in zirconium dioxide respectively. Keywords Silicon determination; titanium dioxide; zirconium dioxide; slurry sampling; electrothermal atomic absorption spectrometry One of the main impurities in ceramic materials is silicon because of its prevalence in the lithosphere. In the production of zirconium dioxide in which zircon (ZrSiO,) is used as a basic material silicon represents a particularly important impurity.' Silicon has adverse effects on the mechanical chemi- cal and electrical properties of many materials because glassy grain boundary phases are formed during the sintering pro- c ~ s s .~ ~ ~ Apart from use in pigments titanium dioxide is also used in vitreous enamels electronic components welding rods synthetic sapphires and rubies. It is also used in capacitors positive temperature coefficient thermistors and piezoelectric materials in electrooptics.* Zirconium dioxide (25-0,) and partially stabilized zirconium dioxide (PSZ) ceramics find applications as cutting tools knife blades milling media pump components machinery wear parts insulation parts and solid electrolytes for fuel cells."" When zirconium dioxide is used as an oxygen sensor the electrical conductivity is influenced by the partial pressure of oxygen which depends on the impurity contents of aluminium and silicon.'2 For the determination of trace elements in ceramic materials mostly solution methods requiring decomposition of the sample and often also separation of the matrix and trace elements have been used.For the determination of silicon titanium dioxide has been decomposed by fusion with ammonium sulfate-sulfuric acid [( NH4)2S04-H2S0,] ,I3 whereas zirconium dioxide has been decomposed by using a mixture of sodium carbonate (Na,CO,) and sodium tetrabo- rate (Na2B207 borax)7 or a mixture of borax boric acid (HBOJ and lithium hydroxide (LiOH).I4 The most important digestion procedures for both matrices are based on utilization of media containing hydrofluoric acid ( HF).12*'5-'7 However using these media no mineralization of yttria stabilized zir- conium dioxide is possible.The limitation of the digestion methods is associated mainly with two problems i.e. with considerable blank values originating from the extraordinarily high over-all concentration of silicon and when using hydro- fluoric acid also with possible losses of silicon. Furthermore all the decomposition procedures are tedious and time consum- ing. Another problem associated with a digestion stage is the introduction of a matrix unsuitable for the subsequent determi- nation step. When hydrofluoric acid is used for the digestion of the samples the formation of silicon tetrafluoride in the * Presented at the XXVIII Colloquium Spectroscopicurn Inter- nationale (CSI) Post-Symposium on Graphite Atomizer Techniques in Analytical Spectroscopy Durham UK July 4-7 1993.t To whom correspondence should be addressed. graphite furnace makes the determination of silicon by electro- thermal atomic absorption spectrometry (ETAAS) impossible. For these reasons direct methods are often preferred for the analysis of ceramics. Titanium dioxide has been analysed for silicon by spark source mass spectrometry,'* atomic emission spectrometry (AES) with spark or d.c. arc excitation,'"*' inductively coupled plasma (ICP) AES with slurry sample introduction2* and X-ray fluorescence ~ p e c t r o m e t r y . ~ ~ ' ~ ~ For the analysis of zirconium dioxide slurry sampling ICP-AES has been ~ s e d . ~ ~ ~ ~ ~ ~ In recent years the slurry sampling technique has increas- ingly been used for the analysis of a wide variety of samples by ETAAS.25,26 However only a few applications to ceramics and related materials have been despite the fact that the samples to be analysed are often in the form of powders with sub-micrometre sized particles which are particu- larly suitable for slurry sampling.In the present work an ETAAS method for the determi- nation of silicon in titanium dioxide and zirconium dioxide based on slurry sampling which avoids the above mentioned disadvantages of the solution techniques has been developed. Experimental Samples Reagents and Radiotracers The particle size of the titanium dioxide sample (Type P25 Degussa Germany) was ~ 0 . 2 pm and that of the investigated zirconium dioxide samples Zr0,- 1 (Dynamit Nobel Troisdorf Germany) Zr02-2 and Zr02-3 (Magnesium Electron Twickenham London UK Type 9066/3 and 1011) was < 2 pm.The commercially available zirconium dioxide powder Z1-02-4 stabilized with 3 mol-% yttria (yttrium oxide) was supplied by Cerasiv (Plochingen Germany) and has a particle size of between 10 and 50 pm. The particle sizes of all samples were determined by scanning electron microscopy. Doubly distilled water was used for preparation of slurries standards and chemical modifier solutions. Calibration stan- dards were prepared by dilution of a stock standard solution (Merck Darmstadt Germany) with a concentration of 1 g 1-'. The hydrochloric acid of pro analysi quality (37% Merck) was purified by sub-boiling distillation.Magnesium nitrate and calcium nitrate were of Suprapur quality (Merck). All other reagents used were of pro analysi quality. The radiotracer experiments were carried out with a com- mercially available calcium radioisotope 47Ca in the chloride form (Amersham-Buchler Braunschweig Germany). The 47Ca tracer had a specific activity of 7.4 MBq pg-' of calcium and464 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1993 VOL. 9 the calcium concentration was 250 pg m1-I. The 97Zr radio- tracer was produced by irradiation of zirconium dioxide for 30 min in the FRM-1 reactor station Garching (Munich Germany) with a thermal neutron flux of 1.3 x loi3 n cm-2 s-'. A specific activity of 0.4 Bq pg-' was obtained. Instrumentation A Perkin-Elmer atomic absorption spectrometer Model 4100 ZL equipped with a THGA graphite furnace an AS-70 autosampler and a USS-100 slurry sampler was used.Background correction was performed using the longitudinal inverse Zeeman effect. For pre-treatment of the suspensions a Sonorex RK 255 H ultrasonic bath (Bandelin Electronic Berlin Germany) was used. Scanning electron micrographs of the titanium dioxide and zirconium dioxide powders were performed on a digital scanning electron microscope Model DSM 962 (Zeiss Oberkochen Germany). The ultrasonic probe Sonoplus HD70 (Bandelin Electronic) was used for producing slurry suspensions of titanium dioxide and zirconium dioxide for the radiotracer experiments. A high-resolution y-ray spectrometer system (EG & G Ortec Munich Germany) consisting of a germanium detector with an efficiency of 44% relative to a 3 x 3 in NaI(T1) detector an energy resolution of 1.72 keV at the 1.332 MeV y-ray of 6oCo was used for counting the 1.297 and 1.148 MeV y-rays of 47Ca and 97Zr respectively.Performance of the Radiotracer Experiments The labelled slurries were prepared by mixing 10mg of the inactive titanium dioxide and 50 mg of the labelled zirconium dioxide respectively with 10 ml of water 80 pg of inactive calcium nitrate modifier and 50 pl of the 47Ca radiotracer in 15 ml polystyrene vessels. The resulting suspension was homo- genized for 20s using an ultrasonic probe and then 20p1 of the slurry were pipetted into the graphite tube. The activity of the graphite tube was counted after the drying ashing atomiz- ation and cleaning steps using the y-ray spectrometer. The accumulation of calcium and zirconium in the tube was estimated by counting the y-rays of 47Ca and 97Zr after five runs were completed.Procedure Slurries of the samples were prepared by mixing 10mg of titanium dioxide or between 10 and 150mg of zirconium dioxide with a solution of 40mg of calcium nitrate in 10ml of doubly distilled water previously checked for the blank value in 15 ml polystyrene vessels by the selected pots pro- Table 1 Temperature programme and instrumental parameters used for slurry ETAAS Temperature programme - Ramp Hold Argon flow/ Step Temperature/"C time/s time/s ml min- Drying 110 1 20 250 130 5 30 250 Charring 1000 10 10 250 Atomization 2400 0 5 0 Cleaning 2600 1 3 250 Instrumental parameters - Wavelength 251.6 nm Sli t-wid t h Source Read 5 s Signal mode Peak area 0.2 nm Hollow cathode lamp 40 mA Sample volume 20 p1 ~ e d u r e .~ ~ ~ ' The suspensions were pre-treated for 10 min in an ultrasonic bath before analysis to disintegrate larger particle agglomerates. The beakers containing the slurries were used directly for autosampling. Before pipetting each aliquot of the slurry using the sampling capillary homogenization was per- formed by ultrasonic agitation with the USS-100 slurry sampler for 30 s at about 4 W. For standardization by the standard additions technique the titanium dioxide and zirconium dioxide slurries were spiked twice in sequence with 1 and 5 pg of silicon respectively. The standard solution containing 100 pg ml-1 of silicon was prepared by diluting the stock standard solution in a poly(propy1ene) calibrated flask and adjusting the pH to 5 with concentrated hydrochloric acid.Temperature programmes and instrumental parameters are summarized in Table 1. Results and Discussion Optimization of the Experimental Conditions When silicon is atomized from both the titanium dioxide and the zirconium dioxide matrix without addition of a chemical modifier the absorption signals have irregular shapes and have poor reproducibility with relative standard deviations up to 50% for integrated absorbance (QA) (n = 5) [see Fig. l(u) and (43. A similar problem was encountered when silicon was atomized from a boron nitride matrix.34 Magnesium nitrate has been reported to be a modifier of universal applicability also suitable for the determination of silicon.35 In the case of the boron nitride matrix the addition of magnesium nitrate as a chemical modifier lead to a significant improvement of both the signal shape and reproducibility while using calcium nitrate this could not be achieved although calcium nitrate has been proposed as an even more efficient modifier than magnesium nitrate for the determination of silicon by ETAAS.36937 In the case of zirconium dioxide the addition of magnesium nitrate leads to a considerable improvement of the signals with respect to the peak shape and reproducibility [see Fig.l(b)]. However when applied to titanium dioxide this modifier caused no improvement in the quality of the signal and it even lead to a reduction in the sensitivity as can be seen in Fig.l(e). For both the titanium dioxide and the zirconium dioxide matrices calcium nitrate proved to be the most suitable chemical modifier. When using this very smooth sharp peaks and an acceptable reproducibility with a relative standard deviation of between 3 and 7% (n=5) are obtained for both matrices [see Fig. l(c) and (f)]. The behaviour of the chemical modifier (calcium) and the matrix element (zirconium) in the graphite tube during the operation of the temperature pro- gramme was studied by using 47Ca and 97Zr as radiotracers. The radiotracer experiments showed that after execution of the whole temperature programme zirconium remained almost quantitatively in the tube (see Fig. 2). This can be explained by the formation of the highly refractory zirconium carbide which melts at 3540°C and volatilizes at 5100"C.38 The behaviour of titanium could not be examined in this way as no suitable radioisotope for this element is available.However a similar behaviour can be expected as titanium carbide with a melting-point of 3140 "C and a boiling-point of 4820 "C also shows refractory behaviour. It is shown in Fig. 3 that calcium also accumulates in the graphite tube with an increasing number of runs. However its behaviour differs from that of zirconium (compare Figs. 2 and 3) as it forms a less stable carbide at temperatures higher than 2000°C.39 While calcium is retained quantitatively in the tube after charring considerable losses occur during atomization and they are further slightly increased during the cleaning step (see Fig.2). For this reason accumulation of calcium deviates from linearity for a higher number of runs whereas zirconium shows an approximately linear behaviour throughout (see Fig. 3).465 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1993 VOL. 9 0.15 0.30 0.15 0 2.5 5.0 0 2.5 5.0 Time/s 0 2.5 5.0 Fig. 1 Influence of the chemical modifier on quality of the absorption signal (solid line) and background signal (broken line) of (a) zirconium dioxide slurry without modifier (integrated absorbance (QA =0.056) (b) zirconium dioxide slurry with 80 pg of magnesium nitrate (QA = 0.1 15) (c) zirconium dioxide slurry with 80 pg of calcium nitrate (QA =0.115) (d) titanium dioxide slurry without modifier (QA = 0.032) (e) titanium dioxide slurry with 80 pg of magnesium nitrate (QA=0.014); and (f) titanium dioxide slurry with 80 pg of calcium nitrate (QA=0.090).In all cases a sample of about 20 pg was applied Charring Atomization Cleaning ' t . B o o ti. Fig. 2 Retention of A 97Zr applied as labelled zirconium dioxide slurry; B 47Ca added as labelled calcium nitrate to zirconium dioxide slurry; and C 47Ca added as labelled calcium nitrate to titanium dioxide slurry I I I I I I I I 1 0 10 20 30 40 50 60 70 80 90 100 No. of heating cycles Although the addition of calcium nitrate as a chemical modifier to the slurry leads to a significant improvement in the quality of the signal for both matrices it does not prevent the build-up of titanium carbide and zirconium carbide resi- dues.To estimate the tube lifetime that is the maximum number of runs with still acceptable signal quality repeated determinations were performed in the same tube. Fig. 4 shows that for both a titanium dioxide and zirconium dioxide matrix peak heights cannot be used for evaluation because of extremely large variations and a rapid decrease in signal especially during the first 50 runs. The integrated signals on the other hand show satisfactory stability for about 300 runs. With further increase in the number of runs the signal quality rapidly decreases and sometimes even breaking of the tubes occurs. In the case of zirconium dioxide a slight decrease in signal with increasing number of runs can be observed during approximately the first 50 runs. However this has no significant Fig.3 Accumulation of A 97Zr applied labelled zirconium dioxide slurry; B 47Ca added as labelled calcium nitrate modifier to zirconium dioxide slurry; and C 47Ca added as labelled calcium nitrate to titanium dioxide slurry influence on the results when the standard additions method is used. Whereas with matrix-free solutions a longer tube lifetime is achievable than with slurries introduction of sample solutions usually containing the concentrated acid mixtures used for decomposition can lead to severe damage to the platform surface. Thus compared with methods requiring decomposition of the sample utilization of slurry samples does not cause shortening of the tube lifetimes. Standardization and Sample Analysis Generally the use of solid materials with certified concen- trations of the elements of interest and matrices corresponding466 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1993 VOL.9 0.671 0 t lu e g 0.335 2 0 - 0.20 0.15 0.10 $ 0.05 c - ( b ) - I I .y . ~ . I 1 C B A - ; 0' I I I I I J 1.0 b 0.8 pi, *. 0.6 0.4 0.2 I I I 1 I 0 50 100 150 200 250 300 No. of runs Fig. 4 Dependence of the absorbance on the number of runs applying (a) titanium dioxide slurry (0.1 YO m/v); and (b) zirconium dioxide slurry (0.1% m/v); A peak heights; and B integrated absorbance (s) 0 700 900 1100 1300 1500 1700 1900 Tern perat u rePC Fig.5 Charring curves of silicon obtained for A aqueous silicon standard (1 ng of Si); B titanium dioxide slurry (about 1.5 ng of Si); and C zirconium dioxide slurry (about 10 ng of Si).In all cases 80 pg of calcium nitrate were added to those of the samples is considered to be the most accurate standardization method for solid sample ETAAS including the slurry sampling technique. However these materials are costly and are not available for all matrices. The second best choice considering the accuracy achievable is calibration by standard additions using aqueous standard solutions. If accept- able accuracy can thus be achieved standardization by cali- bration curve with aqueous standards is because of greater simplicity and rapidity therefore the preferred method. In this work the last two calibration methods were tested for their accuracy and precision. For both calibration tech- niques accurate standardization requires similar behaviour of the silicon contained in the sample and of the silicon contained in the aqueous standard solution during the charring and the atomization stages. From the charring curves shown in Fig.5 obtained for an aqueous silicon standard solution and for the slurries of titanium dioxide and zirconium dioxide all contain- ing calcium nitrate as a chemical modifier it can be seen that this pre-condition is sufficiently fulfilled for the charring step for all three cases very similar charring curves are obtained. The absorption signals for unspiked and spiked slurries of both titanium dioxide and zirconium dioxide (see Fig. 6 ) show that the atomization behaviour of silicon originating from the sample and from the spiked standard solution is also very similar. Furthermore the characteristic masses of silicon for the aqueous solution titanium dioxide and zirconium dioxide using calcium nitrate as a chemical modifier were calculated in order to ascertain whether standardization by a calibration curve was possible.In this manner characteristic masses for silicon of 105 & 10 pg for the aqueous solution 128 & 11 pg for the titanium dioxide slurry and 420f 130 pg for the zirconium dioxide slurry were obtained. The relatively large standard deviation can be explained by a great increase in the character- istic mass during the first 50 runs. From these results it is evident that standardization by the calibration curve method 1.2 0.60 0 2:5 5.0 Time/s Fig. 6 Absorption signals of silicon for (a) aqueous solution; (b) titanium dioxide slurry A without addition of silicon standard B spiked with 2 ng of silicon and C spiked with 4 ng of silicon and (c) zirconium dioxide slurry A without addition of silicon standard B spiked with 10 ng of silicon and C spiked with 20 ng of silicon is not possible.Therefore the standard additions method was used for calibration. The silicon concentrations determined by this technique in four zirconium dioxide samples and one titanium dioxide samples are summarized in Table 2 along with results obtained by other independent methods. The result obtained for titanium dioxide by this method agrees very well with that obtained byTable 2 Silicon contents (pg g-') determined in titanium dioxide and zirconium dioxide by this method and by AES Slurry Slurry Solution Sample ETAAS * d.c.AESY ICP-AESf ICP-AESg Zr0,- 1 244 f 46 314+ 15 250 _+ 10 255+11 Zr0,-2 166 f 30 90,99 102_+7 105+7 - - Ti02 76f4 75+5 Zr0,-3 130 f 41 81,102 100 + 7 95+7 ZrO2-4 25$-4 - - - * n=7. 7 Ref. 40. f Ref. 15. § Solution ICP-AES involving fusion with NH4HS04.16 d.c. AES both methods producing low standard deviations. In the case of zirconium dioxide the agreement of the results with those obtained by d.c. AES ICP-AES with slurry nebuliz- ation and ICP-AES after decomposition of the sample is within the standard deviations except for the sample Zr0,-2. From a comparison of the results in Table 2 it can also be seen that the slurry ETAAS method provides considerably higher stan- dard deviations than the two ICP-AES techniques. This can be explained by the inhomogeneous distribution of silicon in - 20 pm - 50 pm Fig.7 Scanning electron micrographs of (a) an unstabilized zirconium dioxide sample (ZrO - 1) and (b) yttria stabilized zirconium dioxide JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1993 VOL. 9 467 sample ( ZrOz - 4) 0 0.4 0.8 1.2 1.6 2.0 Slurry concentration (%) Fig. 8 Dependence of the absorbance on the slurry concentration of A titanium dioxide and B zirconium dioxide the zirconium dioxide samples 1-3 and at the same time by the extremely low sample portions (20 pg) taken for the determination by the slurry ETAAS technique. Surprisingly the sample with the lowest silicon content and even with the highest particle size Zr0,-4 gives the best standard deviation and thus has obviously the most homogeneous silicon distri- bution.This sample is an yttria stabilized zirconium dioxide powder which in some respects has different characteristics from the unstabilized zirconium dioxide. For example the yttria stabilized zirconium dioxide powder consists of spherical particles whereas the particles of the unstabilized powders are irregular (see Fig. 7). To establish the linear working range for the proposed method different slurry concentrations were examined and for each slurry concentration the amount of the calcium nitrate modifier was also adjusted accordingly. As is evident from Fig. 8 the maximum slurry concentration applicable for titanium dioxide is about 0.6% m/v which is significantly lower than that for zirconium dioxide (1.25%). This is in accordance with the differences in the characteristic masses of silicon in these two matrices.Because with smaller sample portions the role of the sample homogeneity increases slurry concentrations lower than 0.1 YO were not used. The slurry sampling ETAAS method described enables a detection limit for silicon to be achieved estimated by using the criterion of three standard deviations of the blank of 7 pg g-l for titanium dioxide and 2 pg g-' for zirconium dioxide. For the zirconium dioxide matrix this is one order of magnitude better than the limit of detection achievable by atomic emission methods of 20-30 pg g-'.'5i16 Conclusion Slurry sampling ETAAS has proved to be an advantageous method for the determination of silicon in powdered titanium dioxide and zirconium dioxide samples when calcium nitrate is used as a chemical modifier to overcome otherwise strong matrix interferences. Compared with methods that involve sample decomposition and separation of the silicon this tech- nique leads to lower blanks it is less time consuming and no hazardous acids are needed.A possible drawback might be that contrary to many other applications of the slurry sampling technique the standard additions method must be used for calibration. The authors thank J. Pave1 for making available the d.c. AES results. This work was supported financially by the Bundesministerium fur Forschung und Technologie (G.S. Zr02) and the Deutsche Forschungsgemeinschaft (S.H. TiO,) Bonn. Germanv.468 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1993 VOL.9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 References Ayala J. M. Verdeja L. F. Garcia M. P. Llavona M. A. and Sancho J. P. J. Muter. Sci. 1992 27 458. Kobayashi S. J. Muter. Sci. Lett. 1985 4 266. Jaganathan J. Ewing K. J. Buckley E. A. Peitersen L. and Agganval I. D. Microchem. J. 1990 41 106. MacFarlane D. R. Mincely P. J. and Newman P. J. J. Non- Cryst. Solids 1992 140 335. Jun C. Guojan D. and Chengshan Z. J. Non-Cryst. Solids 1992 140 293. Broekaert J. A. C. Graule T. Jenett H. Tolg G. and Tschopel P. Fresenius 2. Anal. Chem. 1989 332 825. Martinez-Lebrusant C. and Barba F. Analyst 1990 115 1335. Hutton R. C. Anal. Proc. 1984 21 317. Clough D. J. Ceram. Eng. and Sci. Proc. 1985 6 1244. Hirano M. and Inada H. Ceram. Int. 1991 17 359. Stuhrhahn H.H. Dawihl W. and Thamerus G. Ber. Dtsch Keram. Ges. 1975 52 59. Carleer R. Van Poucke L. C. and Francois J.-P. Bull. SOC. Chim. Belg. 1986 95 385. Bastius H. CFI Ceram. Forum Int. 1984 61 140. Kruidhof H. Anal. Chim. Acta 1978 99 193. Lobinski R. Borm W. V. Broekart J. A. C. Tschopel P. and Tolg G. Fresenius’ J. Anal. Chem. 1992 342 563. Lobinski R. Broekart J. A. C. Tschopel P. and Tolg G. Fresenius’ J. Anal. Chem. 1992 342 569. Hiraide M. Namba K. and Kawaguchi H. Anal. Sci. 1990 6 609. Jackson P. F. S. and Whitehead J. Analyst 1966 91 418. Mitzuno T. and Matsumura T. Nippon Kinzoku Gakkaiski 1969,33 344. Rantschke R. and Rehfeld K. H. Spectrochim. Acta Part B 1972 27 211. Kato A. Osumi Y. Nakane M. and Miyake Y. Bunseki Kagaku 1974,23 1036. 22 23 24 25 26 27 ‘2 8 29 30 31 32 33 34 35 36 37 38 39 40 Broekaert J.A. C. Leis F. Raeymaekers B. and Zaray G. Spectrochim. Acta Part B 1988 43 339. Denton C. L. Himsworth G. and Whitehead J. Analyst 1972 97 461. Huang M. and Shen X.-e. Spectrochim. Acta Part B 1989 44 957. Miller-Ihli N. J. Anal. Chem. 1992 64 964A. Bendicho C. and de Loos-Vollebregt M. T. C. J. Anal. At. Spectrom. 1991 6 353. Slovak Z. and Docekal B. Anal. Chim. Acta 1981 129 263. Marecek J. and Synek V. J. Anal. At. Spectrom. 1990 5 385. Docekal B. and Krivan V. J. Anal. At. Spectrom. 1992,7 521. Docekal B. and Krivan V. J. Anal. At. Spectrom. 1993 8 637. Fuller C. W. Analyst 1976 101 961. Karwowska R. and Jackson K. W. Spectrochim. Acta Part B 1986 41 947. Karwowska R. and Jackson K. W. J. Anal. At. Spectrom. 1987 6 273. Hauptkorn S. and Krivan V. Spectrochim. Acta Part B 1994 49 221. Welz B. Schlemmer G. and Mudakavi J. R. J. Anal. At. Spectrom. 1992 7 1257. Gitelman H. J. and Alderman F. R. J. Anal. At. Spectrom. 1990 5 687. Holden A. J. Littlejohn D. and Fell G. S. Anal. Proc. 1992 29 260. Handbook of Chemistry and Physics ed. Weast R. C. CRC Press Boca Raton 74th edn. 1993. Gmelins Handbuch der anorganischen Chemie Calcium Syst. No. 28 Teil B Verlag Chemie Weinheim 8th edn. 1961 pp. 821 and 834. Pavel J. Ciba Geigy Basel personal communication. Paper 3 J04957D Received August 14 1993 Accepted October 20 1993