Quantitative Analysis of Zirconium Oxide by Direct Current Glow Discharge Mass Spectrometry Using a Secondary Cathode WIM SCHELLES AND RENE� VAN GRIEKEN* Department of Chemistry, University of Antwerp (UIA), Universiteitsplein 1, B-2610 Antwerpen, Belgium A dc GD mass spectrometer has been used for the analysis of of the research work in this area has been restricted to qualitative analysis of only a few matrices. In the present work, ZrO2 samples with known amounts of artificial impurities.For this purpose the secondary cathode technique has been used. three samples of compacted ZrO2 with added amounts of artificial impurities are measured, mainly aiming at a method- The research work focused both on the methodological aspects of the atomization of the ZrO2 and on some quantitative ological investigation. This results in sputter–atomization data, specifically for this type of matrix, and an evaluation of some aspects, with direct analytical interest.The reproducibility was found to be better than 10% RSD in most cases; the accuracy quantitative aspects. Moreover, from the application point of view, dc GDMS in of the ‘standardless’, raw results was within a factor of 2–3 of the known concentration. combination with the secondary cathode technique can be considered as a novel approach for trace analysis of refractory Key words: Glow discharge mass spectrometry; nonconductor insulators such as ZrO2, which have always been a challenge analysis; secondary cathode; quantitative results to the analytical chemist.22–24 These sample types are commonly measured with typical solid analysis techniques (X-ray In the last decade GDMS has become a powerful and accepted methods,25 laser-based methods)26 or wet-chemical techniques tool for trace analysis of solid samples.Its main advantages such as ICP-OES and ICP-MS (using fusion and subsequent are the ability to measure all elements, even with isotopic dissolution for sample preparation, slurry nebulization, matrix information, with minimum sample preparation, over a wide removal)22–24,27 or (ET) AAS.28 Because all these techniques dynamic range (sub-ppb to 100%) and with a relatively uniform have intrinsic advantages and disadvantages, the present study response over the mass range.1–4 The concept of the GD, in can add dc GDMS to this list, without necessarily being which the sample acts as a cathode, seems to restrict the superior in all aspects to the alternative techniques.applications to conducting materials. However, considerable efforts have been undertaken to extend the inherent advantages of GDMS to insulating materials as well. A well known method to overcome the conductivity requirement is to substitute the EXPERIMENTAL classical dc GD source by an rf powered source.5–8 Here it GDMS becomes possible to sputter–atomize directly nonconducting samples such as glasses and ceramic materials. The develop- The work reported in this study was performed with a VG9000 ments over the last years concerning this type of instrumen- double-focusing dc GD mass spectrometer (VG Elemental, tation are promising, and rf sources have even been coupled Fisons Instruments, Winsford, UK), described in detail else- with commercial mass spectrometers.9–11 It is, on the other where.29 A working resolution of 3500–4000 was routinely hand, also possible to apply dc GD sources to the analysis of used for these studies.The detection system consists of a nonconducting materials. A common approach is to mix the combination of a Faraday cup and a Daly detector, providing powdered sample with a conducting binder material.12–15 a dynamic range of more than nine orders of magnitude (i.e. However, when the sample under investigation is a solid <1×10-18–1×10-9 A). The flat cell of the second generation, material, grinding can be troublesome and can cause severe often referred to as ‘the new flat cell’, was used. It has already contamination.To handle solid sample types with a dc GDMS been described elsewhere.30 The opening of the anode body instrument (commercially available, unlike rf GDMS), a con- was 7.5 mm, as determined by the opening in the front plate ducting diaphragm should be placed in front of the sample, of the sample holder. The cell was cryogenically cooled to the so-called secondary cathode technique.16–21 This method, reduce the background due to residual gases.The glow dis- introduced by Milton and Hutton in 1993,16 has not yet been charge was supported with high-purity argon (Air Liquide, widely applied. The concept is based on continuously coating Lie`ge, Belgium, 99.9997%) that was not further purified. the nonconducting sample with a thin conducting film, sputterdeposited from the secondary cathode. The major advantages are the ease and low cost of the technique; the major limitations are the restricted operating conditions and, for some elements, Materials the possible blank contribution due to the sputtering of the secondary cathode material.In general, one can state that Three samples of 5 g each, based on Tosoh-Zirconia TZ-3Y (93.96% ZrO2, and an uncertified amount of Y2O3), produced the secondary cathode technique can, in certain cases, be a viable alternative for rf GDMS, as previously proven in the by Ceraten (Madrid, Spain) and supplied by VG Elemental, were measured.These samples contained small amounts of restricted number of publications concerning this method.16–21 The analysis of a new type of matrix with the secondary Mg, Al, Si, Ca and Fe, resulting in samples with 10, 100 and 1000 ppm of these elements. The diaphragms used as secondary cathode method is not yet straightforward, mainly because of limited operating conditions to create stable atomization of cathodes were made of 0.25 mm thick tantalum (Goodfellow, Cambridge, UK, 99.9%).the nonconductor under investigation.17–19 Therefore most Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12 (49–52) 49RESULTS AND DISCUSSION larities between the sputtering of SiO2 and ZrO2 are striking. For Si and Zr, the MO+5M ratio is 3.0 and 4.1%, respectively; Atomization of ZrO2 for the MO2+5M ratio (the molecular species) these values are 0.011 and 0.015%, respectively. This resemblance can be When using a secondary cathode, it is important to know whether the matrix under investigation is ‘poorly conducting’ understood as both sample types were measured under the same operating conditions and as the M–O bond strength is or ‘nonconducting’. It has previously been shown that there is, for GDMS measurements, a certain ‘edge of conductivity’ comparable for both (for Si and Zr, it is 798.7±8.4 and 759.8±8.4 kJ mol-1, respectively).31 The fact that cluster (about 1010–1011 V cm).17 Below this edge, samples can be directly used as a cathode resulting in a stable discharge.The species are only present in the discharge at relatively low concentrations is favourable from the analytical point of view. sputter yield can, however, be extremely low, for example for samples with an electrical resistivity close to the edge (e.g. Clusters can interfere with analytical peaks; moreover, the absence of clusters facilitates quantitative procedures, because about 109 V cm). This makes it useless to perform the analysis of these ‘poorly conducting’ samples directly, i.e., without a the matrix peak(s) can reliablybe used, as afirst approximation, as internal standard (see below). The other values in Table 1 secondary cathode.If samples with an electrical resistivity greater than about 1010 V cm are used directly, no discharge have more analytical significance, and will therefore be discussed in the next section. can be obtained. In these cases, the secondary cathode is absolutely necessary to create stable atomization of the nonconducting sample.Poorly conducting and nonconducting Quantitative Results and Analytical Figures of Merit samples require other optimum operating conditions and Three ZrO2 based samples with known amounts of artificial poorly conducting terials reveal better analytical character- impurities were measured under the operating conditions istics when measured with the secondary cathode technique. described in the previous section.The matrix element Zr Therefore, information concerning the electrical resistivity is (74% m/m of ZrO2) was used as an internal standard. This is significant. The most useful way to distinguish between ‘non- the well known ion beam ratio approach, a predecessor of the conducting’ and ‘poorly conducting’ materials is to run the relative sensitivity factors (RSFs) approach commonly used sample directly as a cathode in the dc GD, without exposing for the analysis of high purity conducting samples; in that case, any auxiliary conductor to the plasma.17 If no discharge at all the matrix is considered to have a 100% concentration.The can be obtained, this means that the sample can be classified raw, uncorrected concentrations were calculated according to: as ‘nonconducting’. With this method, the compacted ZrO2 samples were found to be ‘nonconducting’, i.e., the use of the Concentration X (ppm)= Signal intensity X Signal intensity Zr ×0.74×106 secondary cathode was absolutely necessary to create a stable discharge.When a nonconductor is measured with a secondary cath- For each sample (containing 10, 100 or 1000 ppm of each artificial impurity), at least five data were acquired over a ode, the discharge conditions (pressure, voltage, current) cannot be chosen arbitrarily. Therefore, initially, two different range of more than half an hour of sputtering. The mean uncorrected concentrations and their RSDs are listed in sets of discharge conditions, which had previouslybeen successfully applied for ‘nonconductor’ analysis (i.e. for the analysis Table 2.The rather poor reproducibility for Ca in comparison with the other elements is caused by the low abundance of the of glass and of Macor19) were used to sputter–atomize these samples in a stable way. A 3 mA/0.7 kV discharge, used for isotope measured (44Ca, 2.13% abundance). Generally, it is clear that the atomization and ionization of ZrO2 in a dc GD the atomization of glass, always caused instabilities and/or a tantalum coated sample, thereby preventing the underlying using a tantalum secondary cathode can be considered as reproducible.ZrO2 from being sputtered. On the other hand, the (slightly modified) conditions used for Macor, allowed reproducible Table 2 also reveals that acceptable semi-quantitative results can be obtained (within a factor of 2–3), based only on an and successful analysis of ZrO2 .The conditions were a 0.8 mA/1.15 kV discharge, and, as for the glass analysis, the internal and no external standards. The exception is the 10 ppm level, where deviations obtained were rather high. This is either use of a tantalum secondary cathode with a 4 mm hole, an anode body opening of 7.5 mm diameter (as determined by the due to blank values in the ZrO2 sample or to blank values in the secondary cathode. To evaluate this and to have a general opening in the sample holder front plate), and a 0.5 mm thick Teflon spacer between the cathode and anode. The reason for overview of the influence of the blank contribution of the tantalum mask, three panoramic analyses of the tantalum were the different optimum discharge conditions is not yet completely understood; the sample surface itself and the adhesion performed.The data obtained for the impurities are listed in Table 3 in decreasing order of average measured concentration. of the redeposited tantalum atoms seem to play a major role.19 In the first part of this study, the signal intensities of the These ‘real’ blank values should be multiplied with a weight factor of the secondary cathode sputtering to obtain apparent matrix species and the tantalum secondary cathode were evaluated.To make comparisons possible, the raw, isotopic blank values, i.e. the blank contribution of the mask in the spectrum of the sample. The weight of the blank values in the intensities were converted into elemental signal intensities by means of the isotopic abundances.These intensities and their mask is given by the Ta5Zr signal intensity ratio (5.0, see Table 1) and the concentration of Zr in the sample (74%). ratio to the elemental Zr signal intensity are listed in Table 1. When one compares the data with previous results obtained This results in a weight factor of less than 4 (namely 5.0×0.74). This factor is significantly lower than the weight factor 10, with the secondary cathode technique on Macor,19 the simireported for the analysis of Macor ceramic.19 In practice, this means that, for the selected elements, the impurities in the Table 1 Mean maximum elemental signal intensities (A) for the matrix constituents and their ratio to the Zr signal ZrO2 rather than the impurities in the tantalum secondary cathode can be considered as the cause of the significant Mean value Ratio to Zr deviations at the 10 ppm level.O+ 4.9×10-13 1.0×10-2 Based on the known concentrations and the measured, Zr+ 5.2×10-11 1 uncorrected concentrations calibration graphs were drawn.ZrO+ 2.2×10-12 4.1×10-2 The correlation coefficients for the different elements are listed ZrO2+ 8.2×10-16 1.5×10-5 in Table 2. The concept of a calibration curve is useful, Ta+ 2.5×10-10 5.0 especially if a blank contribution is involved. In GDMS, 50 Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12Table 2 Measured, uncorrected concentrations (ppm), based on Zr (74%) as internal standard.Values between brackets are relative standard deviations over at least five measurements on the same sample. Correlation coefficients and calculated RSFsZrOx are based on the calibration lines Measure concentration (ppm) Correlation 10 100 1000 coefficient RSFZrOx RSFstand RSFoxid RSFMacor Mg 6 46 556 0.99985 1.9 1.5 1.0 0.61 (5.3%) (2.4%) (2.9%) Al 28 89 797 0.99994 1.4 1.4 0.57 0.30 (1.4%) (5.5%) (3.3%) Si 18 109 562 0.99723 2.0 1.8 1.3 (5.2%) (7.3%) (5.6%) Ca 18 128 1200 0.99999 0.91 0.55 (25%) (16%) (2.6%) Fe 37 152 1114 0.99989 1.0 1.0 1.0 1.0 (3.8%) (3.7%) (1.6%) Table 3 Mean measured concentrations (ppm) of impurities in the tantalum secondary cathode Concentration Concentration Concentration Impurity (ppm) Impurity (ppm) Impurity (ppm) Nb 185 Ni 0.039 Ti 0.0056 O 39 Ga 0.037 Sm 0.0047 C 20 Li 0.030 U 0.0047 W 14 Cr 0.025 Dy 0.0047 Sn 1.1 Hf 0.025 Pd 0.0037 Mo 0.73 Th 0.023 Er 0.0037 K 0.54 Cd 0.022 Tl 0.0033 Fe 0.45 P 0.022 Mn 0.0026 Na 0.43 Sb 0.016 La 0.0021 In 0.27 Pb 0.015 Eu 0.0020 Cu 0.23 B 0.015 Rb 0.0019 S 0.19 Ba 0.013 Be 0.0019 Ca 0.18 Sr 0.012 Ce 0.0018 Zn 0.092 Zr 0.0097 Y 0.0017 Mg 0.072 Ag 0.0087 Co 0.0016 Pt 0.055 Nd 0.0067 Rh 0.0013 Au 0.052 Yb 0.0063 Pr 0.0010 Al 0.049 Bi 0.0057 V 0.0006 Si 0.045 Gd 0.0057 Sc 0.0006 quantification is, however, commonly performed by means of view (i.e.if one wants to generalize towards other matrices), only blank values due to the sputtering of the tantalum should RSFs.This is defined as the reciprocal value of the slope of the calibration curve, and is calculated relative to Fe (RSF= be taken into account (and not those due to the ZrO2 impurities). As mentioned earlier, these values are rather low 1). Actually, the RSF is a measure of the ‘in-sensitivity’ of an element. The obtained RSFs are also listed in Table 2. For for most elements (see Table 3).In practice, blank contributions restrict the LODs for several elements to a level of about 100 reasons of comparison, ‘standard RSFs’ (obtained for metallic samples),32 RSFs obtained for powdered oxide-based samples14 ppb or even higher. On the other hand, even sub-ppb LODs have been reported for U and Th, using the secondary cathode and limited data obtained for Macor19 (also an oxide-based ceramic, but measured as a solid, unlike the compacted ZrO2 technique.21 In this case, the low resolution mode (and the inherent signal enhancement) and extremely long integration powder samples used in this study) are also listed in Table 2.Although there is no absolute match, the same trend can be times were applied. Generally, one can state that about 85% of the elements can routinely be measured at sub-ppm levels. seen in the different sets: RSFCa<RSFFe<RSFAl etc. Nevertheless, it is clear that RSFs acquired for varying materials, even for ‘matrix-matched’ materials, are not exchangeable, W.S. acknowledges financial support by the Vlaams Instituut if an accurate result is required. voor de bevordering van het Wetenschappelijk-technologisch For the LODs, two factors have to be taken into account Onderzoek in de Industrie (IWT). The authors thank VG for each element: the sensitivity and the blank contribution. Elemental for the supply of samples and K. De Cauwsemaecker The sensitivity for the selected elements, expressed as the and R. Saelens for technical support.maximum signal intensity/ppm is between 2×10-18 A/ppm (44Ca, 2.13% abundant) and 1×10-16 A/ppm (27Al, 100% REFERENCES abundant). Depending on the integration time used, it is possible, for elements not subject to interference, to distinguish 1 King, F. L., Teng, J., and Steiner, R. E., J. Mass Spectrom., 1995, peak heights of 2×10-17 A (20 ms per channel) to 1×10-18 A 30, 1061. (8 s per channel) from the background. In this context, it may 2 King, F.L., and Harrison, W. W., in Glow Discharge Spectroscopies, ed. Marcus, R. K., Plenum Press, New York, 1993, be useful to know that, commonly, 60–100 channels are pp. 175–214. measured for each isotope. Thus, only taking into account the 3 Gijbels, R., van Straaten, M., and Bogaerts, A., in Advances in sensitivity, LODs for most elements are in the range between Mass Spectrometry, ed. Cornides, I., Horva�th, G., and Ve� key, K., 10 ppb (high elemental sensitivity, long integration time) and Wiley, New York, 1995, vol. 13, pp. 241–256. 10 ppm (poor elemental sensitivity, short integration time). 4 Harrison, W. W., J. Anal. At. Spectrom., 1992, 7, 75. However, in addition, the restrictions due to blank values have 5 Coburn, J. W., and Kay, E., Appl. Phys. L ett., 1971, 19, 350. 6 Duckworth, D. C., and Marcus, R. K., Anal. Chem., 1989, 61, 1879. to be taken into account. From the methodological point of Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12 517 Marcus, R. K., Harville, T. R., Mei, Y., and Shick, C. R., Jr., Anal 22 Broekaert, J. A. C., Graule, T., Jenett, H., To� lg, G., and Tscho�pel, Chem., 1994, 66, 902A. P., Fresenius’ Z. Anal Chem., 1989, 332, 825. 8 Harville, T. R., and Marcus, R. K., Anal. Chem., 1995, 67, 1271. 23 Broekaert, J. A. C., and To� lg, G., Mikrochim. Acta, 1990, II, 9 Duckworth, D. C., Donohue, D. L., Smith, D. H., Lewis, T. A., 173. and Marcus, R. K., Anal. Chem., 1993, 65, 2478. 24 Broekaert, J. A. C., Lathen, C., Brandt, R., Pilger, C., Pollmann, 10 Shick, C. R., Jr., Raith, A., and Marcus, R. K., J. Anal. At. D., Tscho�pel, P., and To� lg, G., Fresenius’ J. Anal. Chem., 1994, Spectrom., 1993, 8, 1043. 349, 20. 11 Saprykin, A. I., Becker, J. S., and Dietze, H.-J., J. Anal. At. 25 Radha Krishna, G., Ravindra, H. R., Gopalan, B., and Spectrom., 1995, 10, 897. Syamsunder, S., Anal. Chim. Acta, 1995, 309, 333. 12 Tong, S. L., and Harrison, W.W., Spectrochim. Acta, Part B, 26 Becker, J. S., and Dietze, H.-J., Fresenius’ J. Anal. Chem., 1993, 1993, 48, 1237. 346, 134. 13 Woo, J. C., Jakubowski, N., and Stuewer, D., J. Anal. At. 27 Lobinsky, R., Van Borm, W., Broekaert, J. A. C., Tscho�pel, P., Spectrom., 1993, 8, 881. and To� lg, G., Fresenius’ J. Anal. Chem., 1992, 342, 563. 14 De Gendt, S., Schelles, W., Van Grieken, R. E., and Mu�ller, V., 28 Hauptkorn, S., Schneider, G., and Krivan, V., J. Anal. At. J. Anal. At. Spectrom., 1995, 10, 681. Spectrom., 1994, 9, 463. 15 Duckworth, D. C., Barshick, C. M., and Smith, D. H., J. Anal. 29 Robinson, K., and Nayler, R., Eur. Spectrosc. News, 1986, 68, 18. At. Spectrom., 1993, 8, 875. 30 van Straaten, M., Gijbels, R., and Vertes, A., Anal. Chem., 1992, 16 Milton, D. M. P., and Hutton, R. C., Spectrochim. Acta, Part B, 64, 1855. 1993, 48, 39. 31 Handbook of Chemistry and Physics, West, R. C., Ed., 55th edn., 17 Schelles, W., De Gendt, S., Muller, V., and Van Grieken, R. E., Appl. Spectrosc., 1995, 49, 939. CRC Press, Cleveland, 1982–1983, pF185. 18 Schelles, W., De Gendt, S., Maes, K., and Van Grieken, R. E., 32 VG9000, software version 5.3, Fisons Instruments, VG Elemental, Fresenius’ J. Anal. Chem., 1996, 355, 858. Winsford, Cheshire, UK, 1990. 19 Schelles, W., and Van Grieken, R. E., Anal. Chem., 1996, 68, 3570. 20 Schelles, W., De Gendt, S., and Van Grieken, R. E., J. Anal. At. Paper 6/05397A Spectrom., 1996, 11, 937. Received August 1, 1996 21 Betti, M., Giannarelli, S., Hiernaut, T., Rasmussen, G., and Koch, L., Fresenius’ J. Anal. Chem., 1996, 355, 642. Accepted October 10, 1996 52 Journal of Analytical Atomic Spectrometry, J