Determination of Trace Impurities in High-purity Quartz by Electrothermal Vaporization Inductively Coupled Plasma Mass Spectrometry Using the Slurry Sampling Technique SUSANNE HAUPTKORNa, VILIAM KRIVAN*a, BERTHOLD GERCKENb AND JIRI PAVELb aSektion Analytik und Ho� chstreinigung, Universita�t Ulm, D-89069 Ulm, Germany bNovartis Scientific Services, R-1055.4.02, Postfach, CH-4002 Basel, Switzerland A method has been developed for the determination of 14 ablation17–19 and electrothermal vaporization (ETV).20–22 ETV relevant trace impurities in high-purity quartz based on ETV- seems to be avery promising technique for routine applications, ICP-MS using slurry sampling.The ETV device consisted of a because of the inexpensive equipment, simple handling and the double layer tungsten coil. Experimental conditions were possibility to calibrate with aqueous standards. Furthermore, optimized with respect to the temperature program, the carrier owing to the spatial separation of volatilization (in the vaporgas flow, the i.d.of the aerosol tubing, ICP-MS measurement izer) on the one hand and atomization and ionization (plasma) parameters and internal standardization. Excluding U, on the other, an in situ analyte–matrix separation through calibration had to be carried out by the standard additions sequential volatilization of the sample components is method because of non-spectral matrix interferences. For U, feasible.23,24 simple quantification via calibration curves, recorded with However, when this is not possible, large amounts of matrix aqueous standards, was possible.The observed interferences can reach the plasma simultaneously with the analytes, possibly also aggravated the background evaluation, which seriously leading to matrix interferences and memory effects. The furnace limited the determination of Al and Fe. The method was material may cause additional interferences. Moreover, for applied to the determination of Al, Ba, Co, Cr, Cu, Fe, Li, quadrupole mass spectrometers, the multi-elemental capabili- Mg, Mn, Na, Pb, Sr, U and Zn in two quartz samples of ties for short transient signals (<5 s) are limited, because they different grades of purity. The accuracy of the results was work sequentially, although they are fairly fast.Another critical checked by their comparison with those obtained by aspect is the occurrence of losses during the aerosol transport, independent methods including instrumental neutron activation diminishing sensitivity and reproducibility.25–27 Some workers analysis.The achievable detection limits are between 2 ng g-1 have recommended sodium chloride or NASS-3 seawater ( Li, U) and 70 mg g-1 (Al ). reference material as modifiers for the enhancement of transport efficiency.28–31 However, owing to the possible introduc- Keywords: Inductively coupled plasma mass spectrometry; tion of contaminants, they are not really applicable to the electrothermal vaporization; tungsten coil furnace; slurry sampling; quartz analysis of high-purity microelectronic materials, the less so since the alkali and alkali earth elements, which are the main components of these modifiers, are amongst the most relevant analytes.Industry requirements on the purity of materials for microelec- Mainly on account of their widespread usage in ETAAS, tronic applications, such as quartz, call for analytical methods graphite tubes are the most commonly used vaporizers for allowing accurate determination of trace elemental impurities ETV, e.g., refs.28–33 Nevertheless, graphite vaporizers have a at the ng g-1 level and below.Only a few methods, i.e., neutron number of disadvantages, such as the occurrence of isobaric activation analysis (NAA),1,2 electrothermal atomic absorption interferences by carbon species (e.g., 52Cr and 40Ar12C),34 the spectrometry (ETAAS),3–5 total reflection X-ray fluorescence formation of refractory analyte carbides and the restriction to spectrometry (TXRF)6 and atomic mass spectrometry vaporization temperatures below #2600 °C.Therefore, in (MS),7–11 can provide the detection power necessary for these some cases refractory metal vaporizers are preferable.35–37 applications. In recent years, inductively coupled plasma mass Double layer tungsten coils, as manufactured for halogen spectrometry (ICP-MS)7–9 has developed into one of the most lamps, form simple and inexpensive but nonetheless efficient popular methods for ultratrace analysis.Major advantages of ETV devices, already successfully tested for ETV-ICP- ICP-MS are high instrumental sensitivity, the simplicity of the AES.38–40 They are easily obtained with highly reproducible mass spectra and the possibility of fast multi-element analysis. physical properties, enable high heating rates and temperatures However, these potentials cannot be fully exploited when of up to 3000 °C to be applied even with low cost power ICP-MS is used in combination with conventional nebulization supplies,and only a small piece of quartz apparatus is necessary of solutions, requiring sample digestion.As only sample solu- for mounting of the coil causing only low analyte vapour tions with low salt content can be analysed (usually <0.1%), dilution. Barth and Krivan have already demonstrated its in most cases analyte–matrix separation is also needed.7–9 suitability for the slurry ETV-ICP-AES analysis of silicon Sample dilution and contamination introduced in these carbide.40 additional steps can lead to a considerable increase in the Although the biggest potential of ETV-ICP-MS lies in the achievable LODs.12,13 Moreover, transport efficiencies of the combination with solid sampling or slurry sampling, up to conventional nebulization systems reach only 2–5%.now only a few papers dealing with the analysis of solids have Other sample introduction techniques for ICP-MS include direct insertion of solids,14,15 slurry nebulization,16 laser been published.19,41–49 In none of these works were high-purity Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12 (421–428) 421materials analysed, although this would be the ultimate field Temperature measurements of the tungsten coil were performed using IR thermometers: the Infrascope Mark 1 (Barnes of application for ETV-ICP-MS. In the present work, a slurry ETV-ICP-MS method has Engineering, Stamford, CT, USA), the Cyclops 52 (Land Infrarot, Leverkusen, Germany) and the Minolta Chromameter been developed for the determination of trace elemental impurities in high-purity quartz.Optimizations of the ICP-MS xy-1 (Minolta, Zu�rich, Switzerland) in the temperature ranges 100–400, 600–3000 and 1500–2900 °C, respectively. and ETV conditions were performed with regard to the applicability to routine analysis. Data Acquisition EXPERIMENTAL As the software employed, issue 2.03a, did not include an Samples and Reagents option for processing transient signals, the given possibilities had to be adapted to ETV measurements. Transient signals The analysed quartz samples were SiO2, -325 mesh, 99.9% were recorded using the scanning mode (qualitative scan pure, Lot No.X8653 from Cerac (Milwaukee, WI, USA), program, ‘mode 1’). This program is designed for recording further denoted as SiO2–1, and Aerosil 200, LOS 7638605 and displaying of mass spectra.However, instead of performing supplied by Ciba-Geigy (Basel, Switzerland), denoted as a scan through different m/z values, the quadrupole was set SiO2–2. Particle sizes were estimated by electron microscopy on a fixed m/z of interest. Therefore, the virtual ‘scan’ performed to be less than 10 and 1 mm for samples SiO2–1 and SiO2–2, by the computer was in reality a single ion monitoring during respectively. the length of time assigned for the scan, and the mass axis in For most elements, standard solutions were prepared by the spectrum has to be considered as a time axis.Scanning dilution of the multi-element stock standard solutions ICP- parameters chosen for these measurements are given in Table 1. Mehrelement-Standardlo�sung IV (1000 mg l-1NO3, Unfortunately, in this mode, only one m/z value can be Merck, Darmstadt, Germany) and Trace Elements I, Ground monitored during one measurement cycle. Consequently, quan- Water and Waste Water Pollution Standard (5–500 mg l-1 in titative multi-element measurements were performed in the 5% HNO3, Perkin-Elmer, Norwalk, CT, USA).Single element ‘peak hopping’ mode (‘mode 2’), which allows fast sequential standards were used for the analyte elements Fe, Al, Li measurement of several isotopes by setting the quadrupole (1000 mg l-1, Merck) and U [1000 mg l-1, prepared by directly on each selected signal maximum (-0.1–+0.1 m/z).dilution of aqueous UO2(NO3 )2 in 1% HNO3], and for the For peak hopping measurements, an exact calibration of the internal standards In (1000 mg l-1, Spex Plasma Standards, quadrupole with respect to the chosen isotopes was carried Spex Industries, Grasbrunn, Germany) and 233U (10g l-1 out, otherwise the signal maximum could not be hit correctly. NIST Certified Reference Material U-233, New Brunswick For this purpose, a so-called ‘marker’ has to be run, in which Laboratory, US Department of Energy, IL, USA).The HNO3 a scan is performed across the m/z ranges of interest. It was used was subboiled from concentrated HNO3 (pro analysi, found preferable to run the marker using continuous sample Merck). For the preparation of slurries and standards, introduction by the nebulizer. The peak hopping parameters de-ionized water was used. for marker and analytical measurement are listed in Table 1. Under optimized conditions, a maximum of four isotopes Instrumentation could be measured during a single run.With mode 2, only signal intensities (counts s-1) are obtained, the recording of A VG PlasmaQuad 1 (VG Elemental at Fisons Instruments, Winsford, UK) ICP-MS instrument was used. The ETV device signal profiles is not possible. For the synchronization of ICP-MS measurements and the was similar to the one described by Barth and Krivan.40 It consisted of a double layer tungsten coil, Type 64655 HLX, ETV temperature program, a connection between the ETV computer and the keyboard of the computer controlling the supplied by Osram (Munich, Germany), connected to a0–24 V, 250 W power supply.The power supply was controlled by a mass spectrometer was established, allowing the start of the data acquisition by simulating the operation of the correspond- Sharp PC-7000 computer (Sharp, Osaka, Japan) with software written in GW BASIC. ing key at a signal given by the ETV computer. However, this was only possible in mode 2, whereas in mode 1, the measure- The vaporizer was interfaced to the plasma via quartz tubing with a length of 78 cm and an i.d.of 5 mm. As the quartz ment had to be started manually. Therefore, comparison of the appearance times of signals obtained by ETV-ICP-MS connection is inflexible, the entire ETV set-up had to be fastened to the torch box, enabling it to move together with becomesrather difficult. In both modes, several seconds elapsed between the triggering and the actual commencing of data the torch.In order to prevent oxidation of the tungsten coil during acquisition. This was considered in the temperature program by a time buffer step after the start signal for data acquisition heating, an Ar–H2 mixture [6% H2 , Carbagas (Alphagaz), Basel, Switzerland] was used as carrier gas, the gas flow being and before the beginning of the vaporization step. 800 ml min-1. Lens settings of the mass spectrometer were optimized with the ETV using the 93Nb signal originating from Procedure the Nb impurity vaporized from the tungsten coil at temperatures above 2000 °C.The 93Nb was preferred over a tungsten Sample slurries were prepared by suspending between 5 and 700 mg of quartz in 10 ml of 0.5% HNO3, previously checked isotope, because being in the middle of the m/z range it should give better compromise conditions for multi-element determi- for blank values, in 15 ml polystyrene vessels (Greiner, Frickenhausen, Germany).Slurries with matrix concentrations nations. As optimum torch alignment and plasma conditions did not vary between wet and dry plasmas, optimization below #0.5 g l-1 were prepared by dilution of concentrated stock slurries.3 Homogenization during sampling was of these parameters was performed on 115In using the nebulization system. performed using an ultrasonic probe. The internal standards were added to the suspension media With the experimental set-up described, a fast exchange of nebulizer and ETV unit was possible, which is an important prior to the blank determinations.The concentrations varied between 5 and 20 ng ml-1 for In and between 5 and 25 ng ml-1 feature in routine laboratory work. The ultrasonic probe Labsonic 1510 (B. Braun Melsungen for 233U, depending on the momentary sensitivity of the ICP-MS instrument. For calibration by the standard additions AG, Melsungen, Germany) was used for homogenization of slurries.method, the slurries were spiked twice in sequence with appro- 422 Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12Table 1 Operating parameters for slurry ETV-ICP-MS ET V— Temperature program: Step Voltage/mV Temperature/°C Time/s Drying 120 Rising slowly 10 90 to 300 15 50 45 Thermal pre-treatment* 1000 1500 118 Time buffer 0 7 (mode 1)/3 (mode 2) Vaporization 12500 2700 5 Cool down 0 Decreasing 5 Clean out 12500 2700 3 Sample volume: 10 ml ICP-MS— Outer plasma gas flow rate 12.5 l min-1 Intermediate gas flow rate 1.0 l min-1 Aerosol carrier gas flow rate (Ar+6% H2) 0.8 l min-1 Forward power 1300 W Reflected power 40 W Data acquisition: Mode 1: recording of signal profiles in ‘qualitative scan’ program Range 100–200 m/z Sweeps 1 Dwell time 10 ms Number of channels 1024 Mode 2: quantitative measurements in ‘peak hopping mode’ Peak hopping sweeps 50 Dwell time 500 ms Dwell time for marker (nebulizer) 10 ms Points per peak 5 * Only for the determination of U.priate amounts of aqueous standard solutions; typically Volatile Cd was chosen as a test element to detect possible 50–250 ng of analyte were added to 10 ml of the slurry. losses of the analyte. Owing to the small mass of the vaporizer, Aqueous standards used for recording the calibration graphs the vaporization enthalpy of the water is not negligible and (typically 5–100 ng ml-1) were prepared by dilution of stock the temperature of the coil is dependent on the amount of standard solutions in 0.5% HNO3 .Standard and slurry ali- water present. Thus, with a fixed voltage setting, the temperaquots (10 ml) were pipetted manually onto the tungsten coil. ture increases near the end of the drying step when most of For each value at least four replicate measurements were the water has already been removed. By gradually reducing performed. In the determination of all elements excluding U, the voltage during drying, analyte losses could be avoided with the standard additions method was used for quantification.a temperature program of acceptable duration. As the vaporiz- The ICP-MS operating parameters and the ETV tempera- ation temperature, 2700 °C was chosen for all elements investiture program are given in Table 1, and the m/z values and gated, although some elements could be vaporized at maximum applicable slurry concentrations in Table 2. significantly lower temperatures (e.g., Pb at 1800 °C).However, at this temperature, complete vaporization of all analytes was ensured and, thus, it could be applied to multi-element determi- RESULTS AND DISCUSSION nations. As only in the determination of U did thermal pre- Optimization of ETV treatment prove to be beneficial, no pyrolysis step was included in the temperature program for all other elements (for a more For the optimization of the drying step, the evaporation of detailed discussion, see below). water was monitored by observing the ArO+ signal at m/z 56.The optimized carrier gas flow for ETV measurements of 800 ml min-1 was approximately higher by 100 ml min-1 than Table 2 m/z values and maximum slurry concentrations used for the the aerosol carrier gas flow rate for nebulization of solutions. determination of 14 elements in quartz This is in accordance with the observations of Becker and Hirner.46 They attributed this phenomenon to an earlier m/z value and relative Maximum slurry appearance of the maximum ion density in a dry plasma Element abundance (%) concentration/g l-1 comparedwith awet plasma.A higher gas flow rate counteracts Al 27 (100) 0.02 this effect by moving the maximum ion density forward to Ba 138 (71.9) 10 the interface. Co 59 (100) 10 Cr 52 (83.8) 10 Three different ids, i.e., 1.1, 2.2 and 5 mm, were tested for Cu 63 (69.1) 20 the quartz connection between the ETV unit and the plasma Fe 56/57 (91.7/2.14) 2.0 torch. They proved to have no significant influence on either Li 7 (92.5) 10 the signal shape or the sensitivity. However, the largest diam- Mg 24 (79) 10 eter (5 mm) was preferred, because the tubes with the smaller Mn 55 (100) 10 diameters tended to be blocked by matrix residues during the Na 23 (100) 10 Pb 208 (99.3) 70 analysis of a slurry.Sr 88 (82.6) 10 In the optimization of the peak hopping parameters, it was U 238 (99.3) 10 essential to adjust the measurement window to the duration Zn 64 (48.9) 20 of the signal to obtain an optimum signal-to-noise ratio.The Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12 423width of the window is determined mainly by the number of peak hopping sweeps, whereas the dwell time (measurement time at each point) is less significant. Although, in principle, three points per peak were sufficient, it was found that with five points per peak, better reproducibility was achieved when setting the quadrupole. With the optimized parameters listed in Table 1, a maximum of four isotopes could be determined in one run.One more isotope led to a loss of sensitivity, which could not be regained by a further adjustment of the peak hopping parameters. Several elements, including In, Y, V, Pd and Au, were considered as internal standards for correction of fluctuations of the transport and vaporization processes and for instrumental instability. Preferably, the element used as an internal standard should be close to the m/z value of the analyte elements, have an isotope with a large abundance, and thus give good sensitivity in the MS measurement, and impurities Fig. 2 ETV signal profiles obtained for 63Cu in (A) standard solution (100 pg),(B) quartz slurry (15.4 mg SiO2–1 per vaporization, containing or isobaric matrix interferences from both the samples and the about 26 pg Cu) and (C) slurry spiked with 10 pg Cu in standard standards should not contribute significantly to the signal solution.measured. The indium isotope 115In proved to be most suitable for this purpose. Its m/z value is well in the middle of the RSDs for uranium were improved by between a factor of two range, and thus, it is a good compromise for most elements and ten. determined in the simultaneous multi-element mode. It has a high abundance of 95.6%, gives satisfactory ETV signal profiles (see Fig. 1) and good reproducibility: the RSD of ten replicate Optimization of ETV-ICP-MS for Analysis of Quartz measurements was determined to be around 10% for In spiked to both the 0.5% HNO3 and quartz slurry.Moreover, contrary In Fig. 2, the signal profiles obtained for an aqueous solution, slurry and spiked slurry are given for Cu as a typical example to V and Y, no 115In signal was obtained for sample slurries even at high concentrations of the low purity SiO2–1 sample. for all analyte elements in question. The signals obtained for the slurry show a more pronounced tailing.More or less In the determination of heavy elements such as Pb and Ba, in particular, Au could also be used as an internal standard. pronounced double peaks were also obtained for the other elements. However, quantitative analysis using measurement However, there was no significant improvement compared with In. mode 2 should not be influenced by the signal shape since the signals are integrated. Also, the increase in signal intensity, Using In as the internal standard, the RSD of the ETVICP- MS measurements (n=4) could be improved in 50–60% obtained by adding a spike to the slurry, is evenly distributed over the signal profile (see Cu profiles in Fig. 2; similar of all cases, depending on the analyte, 10–20% of all determined RSD values remained unchanged, whereas 20–30% of the observations were made for the other elements), indicating that the spiked analyte and the analyte contained in the quartz RSDs even increased. Thus, with respect to reproducibility, the use of an internal standard is dispensable, especially when matrix behave similarly in the vaporization process; this is an important prerequisite for calibration via the standard considering that without correction by an internal standard RSDs are usually below 15%.Nevertheless, for the correction additions method. The differences in the appearance times of the ETV-ICP-MS of medium to long-term instrumental sensitivity drifts, an internal standard is still useful.signals, evident from Figs. 1, 2 and 5, were presumably caused by delays in starting the data acquisition by hand. Owing to Uranium constituted a special case inasmuch as it allowed the use of the artificial isotope 233U as a practically ‘ideal’ this uncertainty, it is not possible to obtain and interpret the real appearance times. internal standard, with exactly the same properties and consequently the same behaviour as the analyte. By this means, the As the concentrations of all elements except Na were below the limit of detection in sample SiO2–2, this sample was used to confirm the absence of isobaric interferences from the quartz matrix for the m/z values employed in the analysis (see Table 2).For Na (m/z=23), isobaric interferences by Si-containing species [Si, m/z=28 (92.2%), 29 (4.67%), 30 (3.1%)] can also virtually be excluded. Thus, regarding spectral interferences, it is possible to evaluate the background by measuring the de-ionized water. However, in the analysis of quartz, considerable non-spectral interferences were observed.For most elements determined, the mass response curve (i.e., signal intensity versus sample mass) was non-linear even for fairly low sample masses (see Fig. 3). Moreover, the slopes of the calibration curves obtained for aqueous standards were higher than those obtained by standard additions to the slurry. The signal suppression by the matrix was most pronounced for uranium: a standard spiked to a quartz slurry gave approximately a 20–40 times lower signal intensity than an aqueous standard solution.For this element, a considerable reduction, Fig. 1 ETV signal profiles obtained for 500 pg In in (A) 0.5% HNO3, though not a complete elimination, of the interferences was (B) a slurry of SiO2–1 in 0.5% HNO3 (5 mg SiO2 per vaporization) achieved by a thermal pre-treatment at 1500 °C as part of the and (C) a slurry of SiO2–2 in 0.5% HNO3 (100 mg SiO2 per vaporization).temperature program (see Table 1). By this means, the sensi- 424 Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12Up to the slurry concentrations given in Table 2, taking these as maxima, the sensitivities obtained for slurries and spiked slurries were the same, within the measurement uncertainties. For U, the use of the isotope 233U as an internal standard makes a correction for the matrix interferences possible (see Fig. 4). In this case, the calibration could be performed using aqueous standard solutions.Owing to their relatively high contents in the tungsten coil, Al and Fe showed high blank values (see the 56Fe signal trace of 0.5% HNO3 in Fig. 5A; the blank signal for 27Al has a similar shape). For this reason, the observed non-spectral matrix interferences aggravate the determination of the background, as the suppressing effect of the matrix cannot be accurately taken into account. An accurate background correction could only be performed with an analyte-free sample.Fig. 3 63Cu signal intensity as a function of the slurry concentration of sample SiO2–1 (A) without and (B) with correction by the internal standard 115In (200 pg). tivity obtained for the quartz slurry could be increased by a factor of about ten compared with the standard temperature program. An increase in the pre-treatment temperature above 1500 °C led to a further increase in sensitivity, but also caused analyte losses and, therefore, could not be applied.For all other elements, the maximum applicable pre-treatment temperatures were not high enough to achieve a similar reduction in the matrix interferences. The suppression of analyte signals in the presence of large amounts of matrix in the plasma is a phenomenon well known in ICP-MS,50 including ETV-ICP-MS analysis of solids, which cannot be completely removed during a pyrolysis step.47,48 The suppression can be caused by changes of the plasma Fig. 4 238U signal intensity as a function of the slurry concentration conditions during vaporization and decomposition of the (SiO2–1) (A) without and (B) with correction by the internal standard matrix aerosol, such as cooling of the plasma and dilution of 233U (250 pg). the analytes in the central channel after expansion of the gases formed. The ionization efficiency as well as the site of the maximum ion density might also be influenced. Interferences of this type have also been attributed to ion repulsion effects,51,52 whereby light analyte elements were usually influenced more strongly than heavy elements. However, in the analysis of quartz by ETV-ICP-MS, the extent of this interference proved to be independent of the mass of the analyte element; while U is most strongly influenced, Pb shows the least matrix interferences.Deposition of oxides originating from the matrix (in this case SiO2) on the interface has also been described as a possible source of signal suppression.51 However, as a standard solution measured immediately after a slurry does not show reduced sensitivity, in the present case, it does not seem to be of relevance.The thermal pre-treatment step employed in the determination of U obviously reduces the amount of matrix reaching the plasma. Presumably, the SiO2 is removed as SiO after reduction by either the tungsten of the coil53 or most probably by the hydrogen mixed with the aerosol carrier gas.54 At higher temperatures the reduction process and the volatilization of the reaction products is enhanced, leading to a further decrease in the matrix interference and thereby to an increased sensitivity for U.As the extent of suppression varies for the individual elements, it cannot be completely corrected for by the internal standard In, although it is subject to the same type of matrix interference as the analytes. Therefore, with the exception of U, calibration was performed by the standard additions method.Even then, accurate results were obtained only for slurry concentrations equal to or rather below the maximum values given in Table 2. They were determined by measuring Fig. 5 ETV signal profiles obtained for 56Fe in (A) 0.5% HNO3, (B) the intensities of the analytes for slurries of different concen- 0.5% HNO3 containing 200 pg Fe and (C) slurry of SiO2–1 (23 mg SiO2 containing about 8.5 ng Fe). trations and the same slurries spiked with aqueous standards.Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12 425However, even SiO2–2 is not pure enough with respect to Al In the quartz material SiO2–2, owing to its extremely high purity, the concentrations of all elements except for Na were and Fe for this purpose (see Table 3). Evidently, the relative error made in the background evaluation using de-ionized below the LODs of ETV-ICP-MS. The result obtained for Na is in excellent agreement with the results of the independent water is highest for samples with low analyte concentrations.For high element concentrations, such as in sample SiO2–1, methods, excluding ICP-MS. It is remarkable that the concentrations found by solution ICP-MS for the contamination risk the error resulting from the background evaluation can be neglected (see Table 3 and Fig. 5). elements Mg and Na are between two and three times higher than those obtained by the other methods. It is conceivable that the too high values are due to contaminations introduced Analysis of Samples and Detection Limits during the rather complex analyte–matrix separation procedure described for this method,7 which are difficult to control.In Table 3, the results obtained by slurry ETV-ICP-MS are For the ETV-ICP-MS measurements, on the other hand, the compared with those obtained by independent methods, includelimination of the sample preparation steps and the control of ing INAA, which due to some unique features, normally allows the blank value of the suspension medium prior to slurry a high degree of accuracy to be achieved.In addition, quartz preparation considerably reduce the risk of contamination.12 represents a very suitable matrix for INAA. The concentrations determined by TXRF for Al in both The concentrations determined by slurry ETV-ICP-MS in samples, Co in SiO2–1, and Cr, Cu, Fe and Pb in SiO2–2, sample SiO2–1 are on the whole in good, and in some instances differ considerably from the other results.In these cases, the even in excellent, accordance with those obtained by the other methods are obviously more reliable. independent methods. The mean Co content obtained by ETVDetection limits for 14 elements (Table 4) were calculated ICP-MS is higher by a factor of 1.5 and 2 than the mean as three times the standard deviation of the blank value. Since values determined by INAA and ICP-MS, respectively.the concentrations of all elements except Na were below their However, the uncertainty ranges of ETV-ICP-MS and the LODs in SiO2–2, it was possible to use slurries of this sample INAA results still overlap; for the ICP-MS measurements,7 no with maximum matrix concentrations as blank samples for standard deviations were available. The result determined by determination of the LODs in quartz. The sensitivity of each TXRF is clearly too low. A similar situation exists also with element was determined by standard additions to the same Sr, except that the TXRF result is in better accordance with slurry.This procedure ensures that matrix effects can be fully the results of all other methods excluding ICP-AES, the result taken into account. The concentration of Na was well above of which seems to be too low. The concentration of Pb the LOD in both quartz samples, hence, the standard deviation obtained by ETV-ICP-MS is approximately half those deterof the blank value of the suspension medium was used for mined by ICP-AES and ICP-MS.However, it is confirmed by evaluation of the LOD. The LODs of Al and Fe were estimated TXRF. Unfortunately, INAA is not a suitable method for the determination of this element. using the conditions applied to the analysis of sample SiO2–1 Table 3 Element concentrations determined by slurry ETV-ICP-MS in quartz and comparison with results of independent methods Concentration/mg g-1 This work Independent method Element Sample Slurry ETV-ICP-MS Slurry ETAAS* Solution ETAAS* INAA† ICP-AES† ICP-MS‡ TXRF‡ Al SiO2–1 3200±600 3300±400 2990±150 — 2500±3 — 1700 SiO2–2 <70 1.1±0.2 0.8±0.1 — 1.6±0.1 — 12 Ba SiO2–1 34±7 — — 56±13 26.6±0.3 31 26 SiO2–2 <0.2 — — <0.01 <0.01 <0.025 <0.2 Co SiO2–1 0.8±0.15 — — 0.55±0.1 — 0.4 <0.25 SiO2–2 <0.014 — — 0.0017±0.0001 — — <0.02 Cr SiO2–1 2.1±0.5 3.7±0.6 3.5±0.4 3.0±0.4 2.1±0.1 3 2.2 SiO2–2 <0.02 <0.02 0.007±0.001 0.015±0.004 <0.02 0.02 0.05–1.8 Cu SiO2–1 1.7±0.2 1.62±0.06 1.8±0.1 — 1.60±0.01 0.8 1.63 SiO2–2 <0.05 <0.07 <0.007 — <0.02 <0.03 0.07 Fe SiO2–1 369±18 360±30 390±24 348.0±0.2 233±1 400 306 SiO2–2 <2 0.4±0.1 0.7±0.1 0.5±0.2 0.80±0.01 0.3 0.5–16 Li SiO2–1 1.6±0.1 1.3±0.2 1.7±0.1 — 1.70±0.01 — — SiO2–2 <0.002 <0.012 <0.003 — <0.01 — — Mg SiO2–1 128±15 130±30 150±10 — 147±1 289 — SiO2–2 <0.7 0.21±0.03 0.19±0.01 — 0.30±0.01 0.35 — Mn SiO2–1 14±3 17±4 17.9±0.6 14.0±0.2 10.9±0.1 19 13.3 SiO2–2 <0.03 <0.04 0.012±0.002 <0.02 <0.05 0.029 0.03 Na SiO2–1 73±4 69±9 80±10 79±1 61.6±0.5 198 — SiO2–2 1.1±0.2 1.0±0.2 1.5±0.2 0.8±0.1 1.2±0.1 2.6 — Pb SiO2–1 3.1±0.6 — — — 6.1±0.5 5.5 3.82 SiO2–2 <0.006 — — — <0.01 <0.03 0.025 Sr SiO2–1 29±5 — — 36±5 19.9±0.1 37 27.5 SiO2–2 <0.05 — — <0.8 <0.01 0.003 <0.008 U SiO2–1 0.41±0.03 — — 0.60±0.01 — 0.67 0.50 SiO2–2 <0.002 — — <0.0017 — 0.0002 <0.02 Zn SiO2–1 1.0±0.1 — — <5 1.1±0.1 1 1.1 SiO2–2 <0.4 — — 0.21±0.09 0.10±0.01 <0.25 0.12 * From ref. 3.† From ref. 55. ‡ From ref. 7. 426 Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12Table 4 LODs obtained for ETV-ICP-MS using maximum slurry should be much less of a problem for matrices that are either concentrations and 10 ml aliquots, and comparison with those for more refractory or more volatile than quartz, and therefore ETAAS are better suited for a simple in situ analyte–matrix separation by sequential vaporization of analytes and matrix. Absolute LOD in quartz/ng g-1 Further improvement of this method can be expected from LOD*/pg Slurry Slurry Solution enhancement of the transport efficiency, which is currently Element ETV-ICP-MS ETV-ICP-MS ETAAS† ETAAS† under investigation. Al 20 70×103 200 500 Ba 5 200 — — Co 3 14 — — REFERENCES Cr 7 20 20 3 Cu 5 50 70 7 1 Franek, M., and Krivan, V., Anal.Chim. Acta, 1993, 274, 317. 2 De Corte, F., De Wisplaere, A., Van den Boer, M., Bossus, D., Fe 30 2000 500 30 Li 0.05 2 12 3 and van Slujs, R., Anal. Chim. Acta, 1991, 254, 127. 3 Hauptkorn, S., and Krivan, V., Spectrochim. Acta, Part B, 1996, Mg 20 700 2 0.4 Mn 2 30 40 4 51, 1197. 4 Phelan, V. J., and Powell, R. J. W., Analyst, 1984, 109, 1269. Na 20 100 7 13 Pb 0.1 6 — — 5 Nakamura, T., Sasagawa, R., and Sato, J., Bunseki Kagaku, 1992, 41, 89. Sr 0.1 50 — — U 0.03 2 — — 6 Reus, U., Spectrochim. Acta, Part B, 1989, 44, 533. 7 Baumann, H., and Pavel, J., Mikrochim. Acta, 1989, III, 423. Zn 5 400 — — 8 Herzner, P., and Heumann., K. G., Anal. Chem., 1992, 64, 2942. 9 Naka, H., and Kurayasu, H., ISIJ Int., 1993, 3, 1252. * In aqueous solution. † From ref. 3. 10 Milton, D. M. P., Hutton, R. C., and Ronan, G. A., Fresenius’ J. Anal. Chem., 1992, 343, 773. 11 Milton, D. M. P., Clark, J., Potter, D., and Hutton, R. C., Anal. Sci., 1991, 7, 1243. (Al, slurry concentration 0.01 g l-1; Fe, slurry concentration 12 Docekal, B., and Krivan, V., J.Anal. At. Spectrom., 1993, 8, 637. 2 g l-1; m/z 57) for which, due to the relatively high contents 13 Friese, K.-Ch., and Krivan, V., Anal. Chem., 1995, 67, 354. of these two elements, the error made in the 14 Hall, G. E. M., Pelchat, J.-C., Boomer, D. W., and Powell, M., J. Anal. At. Spectrom., 1988, 3, 791. background determination was negligible. 15 Karanassios, V., and Horlick, G., Spectrochim. Acta, Part B, 1989, On the whole, the LODs achievable in the analysis of quartz 44, 1361.by slurry ETV-ICP-MS are at the same level as those obtained 16 Mochizuki, T., Sakashita, A., Iwata, H., Ishibashi, Y., and Gunji, for slurry ETAAS.3 Detection limits are, with the exception of N., Fresenius’ J. Anal. Chem., 1991, 339, 889. Al and Fe, in the 1–100 ng g-1 range. The ETV-ICP-MS 17 Imbert, J. L., and Telouk, P., Mikrochim. Acta, 1993, 110, 151. technique developed was even more limited with respect to the 18 Voellkopf, U., Paul, M., and Denoyer, E.R., Fresenius’ J. Anal. Chem., 1992, 342, 917. maximum applicable slurry concentration than the ETAAS 19 Mochizuki, T., Sakashita, A., Iwata, H., Ishibashi, Y., and Gunji, method, because the matrix effects appeared more severe in N., Anal. Sci., 1991, 7, 151. ETV-ICP-MS. Moreover, the absolute LODs of this ETV- 20 Matusiewicz, H., Adv. At. Spectrosc., 1995, 2, 63. ICP-MS technique in aqueous solution, although still between 21 Carey, J.M., and Caruso, J. A., Crit. Rev. Anal. Chem., 1992, 0.03 and 30 pg, are in some instances significantly higher than 23, 397. those reported by other workers with different ETV-ICP-MS 22 Olson, L. K., Vela, N. P., and Caruso, J. A., Spectrochim. Acta, Part B, 1995, 50, 355. set-ups.20 In the present work, the detection capability was 23 Seubert, A., and Meinke, R., Fresenius’ J. Anal. Chem., 1994, found to be mainly limited by the background fluctuation, 348, 510.especially for Al and Fe, which were detectable impurities in 24 Argentine, M. D., and Barnes, R. M., J. Anal. At. Spectrom., 1994, the tungsten coil (see Fig. 5). The fluctuation was obviously 9, 1371. caused by the expansion of the aerosol carrier gas at the 25 Kantor, T., Spectrochim. Acta, Part B., 1988, 43, 1299. beginning of the vaporization step, leading to changes in the 26 Sparks, C. M., Holcombe, J. A., and Pinkston, T. L., Spectrochim. Acta, Part B, 1993, 48, 1607.plasma conditions, visible as a drop in the baseline. This effect 27 Ediger, R. D., and Beres, S. A., Spectrochim. Acta, Part B, 1992, might be less pronounced with graphite furnace ETV systems, 47, 907. as these furnaces cannot usually attain heating rates similar to 28 Hoffmann, E., Lu�dke, C., and Scholze, H., J. Anal. At. Spectrom., the comparatively small tungsten coil. On the other hand, the 1994, 9, 1237. small dimension of the coil made it possible to construct an 29 Hughs, D.M., Chakrabarti, C. L., Goltz, D. M., Gre�goire, D. C., ETV housing of very low dead volume, minimizing dilution of Sturgeon, R. E., and Byrne, J. P., Spectrochim. Acta, Part B, 1995, 50, 425. the analytes. Obviously, this advantage does not offset the 30 Lamoureux, M. M., Gre�goire, D. C., Chakrabarti, C. L., and increased background fluctuation. However, further improve- Goltz, D. M., Anal. Chem., 1994, 66, 3217. ment to the ETV unit might lead to a considerable 31 Sparks, C.M., Holcombe, J. A., and Pinkston, T. L., Appl. improvement in the LODs. Spectrosc., 1996, 50, 86. 32 Lamoureux, M. M., Gre�goire, D. C., Chakrabarti, C. L., and Goltz, D. M., Anal. Chem., 1994, 66, 3208. CONCLUSION 33 Hub, W., and Amphlett, H., Fresenius’ J. Anal. Chem., 1994, 350, 587. Compared with the conventional nebulization of solutions for 34 Gre�goire, D. C., and Sturgeon, R. E., Spectrochim. Acta, Part B, ICP-MS, the slurry technique combined with electrothermal 1993, 48, 1347.vaporization considerably reduces the risk of contamination, 35 Marawi, I., Olson, L. K., Wang, J., and Caruso, J. A., J. Anal. At. is less time consuming, easier to apply and avoids the use of Spectrom., 1995, 10, 7. 36 Tsukahara, R., and Kubota, M., Spectrochim. Acta, Part B, 1990, the highly toxic hydrofluoric acid as a digestion medium. 45, 779. However, the LODs of this technique are not superior to some 37 Shibata, N., Fudagawa, N., and Kubota, M., Anal. Chem., 1991, other methods, such as slurry ETAAS, and in some instances 63, 636. they are even inferior (see the LODs for solution ETAAS 38 Dittrich, K., Fuchs, H., Berndt, H., Broekaert, J. A. C., and in Table 4). Schaldach, G., Fresenius’ J. Anal. Chem., 1990, 336, 303. The strong matrix interference represents the main limitation 39 Dittrich, K., Mohamad, I., Nguyen, H. T., Niebergall, K., Pfeifer, M., and Wennrich, R., Fresenius&rsqu. Chem., 1990, 337, 546. of the technique as applied to the analysis of quartz. This Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12 42740 Barth, P., and Krivan, V., J. Anal. At. Spectrom., 1994, 9, 773. 49 Ren, J. M., Rattray, R., Salin, E. D., and Gre�goire, D. C., J. Anal. 41 Moens, L., Verrept, P., Boonen, S., Vanhaecke, F., and Dams, R., At. Spectrom., 1995, 10, 1027. Spectrochim. Acta, Part B, 1995, 50, 463. 50 Evans, E. H., and Giglio, J. J., J. Anal. At. Spectrom., 1993, 8, 1. 42 Gre�goire, D. C., Miller-Ihli, N. J., and Sturgeon, R. E., J. Anal. 51 Gre�goire, D. C., Spectrochim. Acta, Part B, 1987, 42, 895. At. Spectrom., 1994, 9, 605. 52 Beauchemin, D., McLaren, J. W., and Berman, S. S., Spectrochim. 43 Vanhaecke, F., Boonen, S., Moens, L., and Dams, R., J. Anal. At. Acta, Part B, 1987, 42, 467. Spectrom., 1995, 10, 81. 53 Gmelin Handbook of Inorganic Chemistry, T ungsten, System- 44 Wang, J., Carey, J. M., and Caruso, J. A., Spectrochim. Acta, Part No. 54, suppl. vol. A7, Springer-Verlag, Berlin, 8th edn., 1987. B, 1994, 49, 192. 54 Gmelins Handbuch der Anorganischen Chemie, Silicium, System- 45 Darke, S. A., Pickford, C. J., and Tyson, J. F., Anal. Proc., 1989, No. 15, Part B, Verlag Chemie, Weinheim/Bergstrasse, 1959. 26, 379. 55 Fritz, M., and Krivan, V., unpublished results. 46 Becker, S., and Hirner, A. V., Fresenius’ J. Anal. Chem., 1994, 350, 260. Paper 6/06027G 47 Vanhaecke, F., Galba�cs, G., Boonen, S., Moens, L., and Dams, Received September 2, 1996 R., J. Anal. At. Spectrom., 1995, 10, 1047. 48 Boonen, S., Vanhaecke, F., Moens, L., and Dams, R., Spectrochim. Accepted January 6, 1997 Acta, Part B, 1996, 51, 271. 428 Journal of Analytical Atomic Spectrometry, April 1997,