Improved Slurry Sampling Electrothermal Vaporization System Using a Tungsten Coil for Inductively Coupled Plasma Atomic Emission Spectrometry PETER BARTH SUSANNE HAUPTKORN AND VILIAM KRIVAN* Sektion Analytik und Ho� chstreinigung Universita�t Ulm D-89069 Ulm Germany vaporizers a much lower current is suYcient to reach very high heating rates and temperatures of up to 3000 °C. Therefore inexpensive and small power supplies are adequate. Sample volumes of up to 40 ml can be pipetted on to the coil and by using the slurry sampling technique the direct analysis of powdered solids is possible. The tungsten-coil vaporizer in combination with slurry sampling has already been applied to the determination of trace impurities in silicon carbide by ETV–ICP-AES18 and silicon dioxide by ETV–ICP-MS.19 Although the experimental set-up used for ETV–ICP-AES worked satisfactorily and provided accurate results,18 improvements concerning power supply data acquisition and processing spectrometer coupling automation and tungsten ablation were necessary.In this paper the improved ETV system is described. In addition the transport losses of several elements were determined by means of the radiotracer technique for diVerent matrices. The application of this system to the direct determination of trace impurities in silicon dioxide and silicon nitride by slurry sampling ETV–ICP-AES is described in a later paper.39 EXPERIMENTAL 2 An improved ETV system for the determination of trace elements in diverse samples by slurry sampling ETV–ICP-AES is presented.It consists of a tungsten coil vaporizer a simple computer controlled power supply with high reproducibility of the output voltage and a fast and eYcient data acquisition and processing system for the short transient signals. Coupling the ETV system with the spectrometer and control of additional functions were performed by means of an interface connected to the ETV computer allowing operation with a high degree of automation. The ablation of tungsten in the vaporization step could be significantly reduced by coating the coil with tungsten carbide and by a decrease in the concentration of traces of oxygen in the Ar–H carrier gas using a high-voltage discharge cell. The tungsten ablation was investigated for aqueous solutions and also for silicon carbide as an example of a refractory matrix.The transport losses of the analyte elements Au As Ca Cr Cu Co Fe La Mn Na Sb and Sc were determined for several matrices by means of the radiotracer technique. Transport losses were found to be in the range from 7% (La) to 54% (Cu). Keywords Inductively coupled plasma atomic emission spectrometry; electrothermal vaporization; tungsten coil; slurry sampling; silicon carbide; silicon dioxide; silicon nitride; transport losses; radiotracer technique Inductively coupled plasma atomic emission and mass spectrometry (ICP-AES and ICP-MS) are well established methods for the analysis of a wide variety of liquid and solid samples. However when these methods are applied to analysis of solid samples in their conventional form decomposition of the sample material is necessary.Particularly for refractory inorganic materials the decomposition is often diYcult and sometimes even impossible. In general this time consuming procedure represents a considerable limitation to the limits of detection and a source of systematic errors both caused mainly by blanks. For these reasons methods for the direct analysis of solids are of great interest. Among the solid sampling techniques that have been developed for ICP-AES and ICP-MS electrothermal vaporization (ETV) is increasingly gaining popularity for the direct determination of impurities in solids at the trace or ultratrace level.1–19 Graphite furnaces similar to those used in atomic absorption spectrometry (AAS) are the predominant type of ETV devices.1–11,20–25 Besides graphite refractory metals (tungsten or tantalum tubes and filaments) have been used as furnace materials.14,26–30 Double-layer tungsten coils normally manufactured in large numbers for low-voltage lamps also belong to this group and have been applied as vaporizers to AAS,31–37 ICP-AES18,38 and ICP-MS.19 The attraction of these coils is to be seen in their extremely low price and high reproducibility of geometric shape and physical properties.Mounted in a small quartz apparatus they form a simple inexpensive and eYcient ETV device. Compared with graphite or metal-tube Samples and Reagents Silicon carbide Type S933 (ESK Kempten Germany).The average particle diameter was at the sub-mm level and it did not exceed 5 mm. Silicon dioxide SiO2–1 99.9% pure -325 mesh lot no. X8653 (Cerac Milwaukee WI USA). SiO2–2 Aerosil 200 LOS 7638605 (Novartis Basle Switzerland) (high-purity sample). The particle size of both samples was estimated by electron microscopy to be less than 10 mm. Silicon nitride Type LC12 (H. C. Starck Goslar Germany). The average particle diameter was 0.48 mm with a Gaussian grain size distribution; 90% of the sample had a particle diameter of about 0.80 mm. element stock standard solutions (1 g l-1) (Merck Darmstadt Working standard solutions were prepared using single Germany). For dilution of the stock standard solutions and preparation of the sample slurries doubly distilled water was used.Hexane used for the coating of the coil and hydrofluoric acid (40%) and nitric acid (65%) used in the radiotracer experiments were of ‘reinst’ quality (Merck). An acid mixture containing HF and HNO3 (6 mol l-1 each) and dilute HNO3 (6 mol l-1) were prepared. In order to minimize the oxidation of the tungsten coil during heating an Ar–H mixture (6.5% v/v H2) (Linde Journal of Analytical Atomic Spectrometry December 1997 Vol. 12 (1351–1358) 2 1351 Munich Germany) normally used for welding purposes was used. Radiotracers Radiotracers were produced by irradiation of evaporated stock standard solutions of As Au Cu La Mn Na Sb and Sc in the FRM-1 reactor Garching Germany at a thermal neutron flux of 2.0×1013 cm-2 s-1.The irradiation and decay times for manganese and the other elements was 1 and 6 h and 1 and 2 d respectively. The irradiated compounds were dissolved in 1 ml of 6 mol l-1 HNO3. For copper a separate radiotracer solution was prepared. For the in situ experiments 100 mg portions of silicon carbide were irradiated. The radiotracers 47Ca 58Co 51Cr and 59Fe with specific activities of 5.4×103 4.6×108 1.03×107 and 1.02×105 Bq mg-1 respectively were supplied by Amersham-Buchler (Braunschweig Germany). The main nuclear data for the radionuclides are given in Table 1. Fig. 1 Schematic diagram of the set-up for the tungsten coil technique 1 carrier gas (Ar–H2; 6.5% v/v H2; 1 bar); 2 flow meter with built-in needle valve (0–1000 ml min-1); 3 flow meter with built-in needle valve (0–30 ml min-1); 4 high-voltage discharge cell (HVDC); 5 glass bottle with hexane (volume=10 ml); 6 coating valve (two-way valve electrically operated); 7 ETV device (quartz); 8 interface to the ICP (quartz tube with ball joint); 9 plasma torch; 10 computer (AT-286); 11 serial port; 12 parallel port; 13 interface; 14 power supply; 15 ultrasonic probe on/oV; 16 to spectrometer computer (triggering of the measurement); 17 ceramic stopper; 18 copper electrodes; 19 tungsten coil; 20 quartz stopper; 21 carrier gas inlet; A carrier gas stream (700 ml min-1); and B coating gas stream (9–16 ml min-1).The quartz apparatus for mounting of the tungsten coil and the quartz interface to the ICP were laboratory made.A 2 Type HD-70 ultrasonic probe (Bandelin Electronic Berlin Germany) was used for homogenization of the slurries. The laboratory made power supply was controlled by a portable Amount traced per vaporization/ ng Half-life Corresponding concentration in the matrix†/ mg g-1 Specific activity*/ Bq ng-1 c-Rays counted/ keV 559.1 411.8 1296.8 810.8 1.6 23 0.0007 52 750 5.4 455000 26.3 h 2.7 d 4.54 d 70.8 d 8 0.4 115 0.0035 0.14 0.8 1.5 0.8 10300 133 102 75 320.1 511.0 1099.2 1596.2 27.7 d 12.7 h 45.1 d 40.3 h 10 8 1 62 6.2 46 846.6 1368.6 564.0 2.58 h 15.0 h 2.72 d 0.7 4 7.5 4 50 40 5 4 0.8 57 889.3 83.6 d Instrumentation The instrumental set-up is shown in Fig.1. AES measurements were performed with a sequential JY-24 spectrometer extended with a JY-74 polychromator (Paschen–Runge mount 15 elements) both from Jobin-Yvon (Longjumeau France). The operating parameters are given in Table 2. Spectrometer control was eVected with the standard Jobin-Yvon software package running on a 386-IBM clone (33 MHz). For data acquisition a 16-channel 12-bit analogue-to-digital card (Type MC-PC20 BMC Puchheim Germany) plugged into an ISA slot of the 386 was used. The carrier and coating gas flow rates were adjusted by flow meters with built-in needle valves (Rota Wehr Germany; rotameter type). For the reduction of molecular oxygen in the carrier gas a high-voltage discharge cell (HVDC) was constructed.The Ar–H2 mixture flows through the space formed between two concentric quartz tubes. The electric discharge is generated between two copper electrodes located outside of this space (one inside the inner quartz tube and the other outside the outer quartz tube). The electrodes are connected to a highvoltage generator consisting of a pulse generator (rectangular wave shape frequency approximately 300 Hz) and a car ignition coil. The high voltage can be adjusted via the output voltage of the pulse generator. For coating the tungsten coil with tungsten carbide part of the carrier gas was led through a small glass bottle filled with hexane. The Ar–H –hexane mixture (‘coating gas’) could be switched into the carrier gas stream by means of an electrical valve (‘coating valve’).Table 1 Main data for the radiotracers used for the determination of transport eYciencies Radionuclide 76As 198Au 47Ca 58Co 51Cr 64Cu 59Fe 140La 56Mn 24Na 122Sb 46Sc * Specific activity corrected to the day when the experiments were performed. † Calculated for a slurry aliquot of 20 ml and a slurry concentration of 10 g l-1. 1352 Journal of Analytical Atomic Spectrometry December 1997 Vol. 12 Table 2 ICP-AES parameters used for slurry ETV–ICP-AES Plasma gas (Ar) flow rate Intermediate plasma gas (Ar) flow rate Aerosol carrier gas (Ar–H2; 6.5%v/v H2) flow rate Rf power Analyte emission line Element 396.152 208.959 313.042 393.366 226.502 228.616 267.716 324.754 259.940 279.553 257.610 231.604 220.353 334.941 213.858 193.001* 208.819* Al B Be Ca Cd Co Cr Cu Fe Mg Mn Ni Pb Ti Zn C W * Monochromator.286 computer (Type PP-1601 Charisma Taiwan). The coating valve the ultrasonic probe and the triggering of the data acquisition were controlled by a laboratory made interface connected to the serial port of the 286. Type 64655 HLX 24 V 250 W tungsten coils were supplied by Osram (Munich Germany). The diameter of the wire was 0.30 mm the area of the coil was 7.0×3.6 mm2 the distance between the double layers was 1.2 mm and the mass was 210.4±0.2 mg (n=20). The space formed between the double layers allows the introduction of up to 40 ml of sample solution or slurry.The tungsten coil was fixed in the quartz apparatus by means of a ceramic stopper with two copper electrodes for clamping and contacting the coil. Measurement of the tungsten coil temperature in the range 800–3000 °C was performed with an optical pyrometer (Type Cyclops 52 Land/Minolta Leverkusen Germany). For voltage measurements a true-r.m.s. digital multimeter was used (Type 87 Fluke Kassel Germany). Gamma-rays from the radiotracers were counted by a high resolution gamma-ray spec- Voltage/mV Step Drying Vaporization Clean-out Coating* 790 710 610 500 17100 19500 0 14500 11200 0 Cool down Additional functions— Action Baseline correction Measurement Coating valve Ultrasonic probe Table 3 ETV parameters used for slurry ETV–ICP-AES for a sample volume of 20 ml * Coating gas flow rate=9–16 ml min-1 depending on matrix and slurry concentration.From/s To/s Journal of Analytical Atomic Spectrometry December 1997 Vol. 12 Procedures For the determination of the power consumption of the tungsten coil vaporizer the actual electric current as a function of the output voltage was measured by the voltage drop at a high-precision resistor (0.001 V) switched in series with the tungsten coil. For measurement of the tungsten coil temperature the optics of the pyrometer were focused through the pipetting hole of the quartz apparatus on to the centre of the tungsten coil.A carrier gas flow rate of 700 ml min-1 was used throughout. For tungsten carbide coating executed after each sample vaporization the optimum coating gas flow rate was 9–16 ml min-1 depending on the type of matrix and slurry concentration. The coating was performed at 2400 °C and the coating gas was switched for 8 s into the carrier gas stream (see also Table 3). For coating a new tungsten coil this coating step was applied twice. For the optimization of the coating process the carbon and tungsten emission was monitored by setting the monochromator to 193.001 and 208.819 nm respectively. At a coating gas flow rate of 12 ml min-1 the maximum concenwas about 9.5 mg ml-1. tration of hexane in the ETV device during the coating step The slurries were prepared as described elsewhere.39 In the in situ radiotracer experiments the irradiated silicon carbide sample was used for slurry preparation.For all other radiotracer experiments inactive slurries or 10 ml of water were spiked with the radiotracer solution during ultrasonic agitation. As the activity of 64Cu can only be measured at the non-specific 511 keV annihilation line all experiments using 64Cu had to be performed separately from the other radionuclides. The amount traced in a 20 ml aliquot (sample volume used for vaporization) and the corresponding concentration in the solid for a 10 g l-1 slurry are given in Table 1. For the radiotracer experiments the ETV device was assembled in a hood. To retain the greatest part of the vaporized radionuclides a cellulose filter was placed at the end of the aerosol transport path.Aliqouts of the slurries (20 ml ) were pipetted on to the coil dried and vaporized using the voltage–time programme without a clean-out and coating step (see Table 3). After 20 Ramp/s 14 l min-1 0.3 l min-1 0.7 l min-1 900 W Wavelength/nm Interfering W emission line 228.629 267.728 259.964 257.617 Temperature/°C — — — — 2600 2700 Decreasing 2400 2200 Decreasing 99 110 94 100 125 175 117 160 trometer system consisting of an HPGe detector and a 16K multi-channel analyzer (EG&G Ortec Munich Germany). The gamma-rays from 64Cu were counted with a single channel analyzer equipped with an NaI(Tl ) detector and an automatic sample changer (Berthold Munich Germany).Elapsed time/s Hold/s 20 10 40 30 0 0 0 0 5 0 3 9 1 0 9 4 0 0 20 30 70 100 105 109 118 127 131 176 45 0 1353 3 3 (65%)+2 ml of HF 3 (65%)+0.5 ml of HF (40%) respect- observed so far indicating a remarkably high reproducibility sequential vaporization steps the ETV device was dismantled and the quartz parts were rinsed three times with 1 ml of the HF–HNO mixture. The tungsten coil and the filter were digested in a mixture of 1 ml of HNO (40%) and 2 ml of HNO ively. For the reference values the corresponding amounts of the radiotracers were pipetted directly into the vessels used for counting.Prior to the gamma-ray measurements all resulting solutions were made up to equal volumes. RESULTS AND DISCUSSION Fig. 2 Schematic diagram of set-up for the power supply used for the heating of the tungsten coil vaporizer. Power Supply and Data Acquisition Power supply When using a tungsten coil vaporizer as described above the requirements for a power supply are as follows. (a) For fast vaporization of the solvent or the suspension medium without any analyte losses during the drying step a precise and reproducible power setting in the low-power range must be possible. For example to remove a sample volume of 20 ml a power setting of 1.7W (790 mV 2.15 A) was used at the beginning of the drying step which to avoid analyte losses had to be reduced stepwise to a power setting of 0.91 W (500 mV 1.82 A).( b) For matrix vaporization and clean-out power settings in the range 150–200 W (16.2 V 9.3 A to 19.5 V 10.3 A) are necessary. (c) For comfortable operation and automation computer control of the power supply is desirable. We first used a 24 V ac power supply where the power setting was performed by means of a phase-shift control.18 However reproducible adjustment of low output power as needed during the drying step was not possible. To avoid analyte losses caused by too high temperatures of the tungsten coil the drying step had to be performed at lower output voltages resulting in a significantly prolonged drying step. The improved power supply presented in this work is based on a diVerent principle of power setting and combines simple construction low costs and very high reproducibility of the output voltage.The schematic set-up of the power supply is shown in Fig. 2. All functions are controlled by the parallel printer port of a computer. The output power can be digitally adjusted in 256 steps by pulse-width modulation (frequency approximately 122 Hz) of a stabilized dc voltage. To achieve a fine graduation of the output power in the low-power range the stabilized dc voltage can be switched from 22 to 5 V. As the pulse width is adjusted digitally the only part of the power supply which could influence the reproducibility of the power 1354 Journal of Analytical Atomic Spectrometry December 1997 Vol. 12 setting is the reference voltage needed for stabilizing the dc voltage.However only a few electronic components are necessary to achieve a stable and temperature compensated reference voltage. No drift of the dc voltage during operation has been of the power setting. As the temperature of the tungsten coil in the vaporization step was not controlled by a pyrosensor the temperature–time curve is essentially ballistic. Owing to the very high heating rates of the tungsten coil this heating characteristics do not have any disadvantageous eVect on the vaporization of the matrix. Simple and comfortable operation of the power supply is performed by a computer program written in PowerBasic (Kirschbaum Software Emmering Germany) which allows one to run voltage–time programmes similar to those in ETAAS (Table 3).The voltage values given in Table 3 (and at the beginning of this section) are the voltages of a dc source delivering the same electrical power to the tungsten coil as the pulsed dc output voltage of the power supply. The temperature and the power consumption of the tungsten coil as a function of the voltage of a dc source are shown in Fig. 3. The temperature values are the mean values of measurements performed with five diVerent new tungsten coils; the relative standard deviations were below 2.5%. The influence of the number of vaporizations coating and diVerent nature of the matrix on the temperature of the tungsten coil in the vaporization step was also investigated. The measured values were within the standard deviation of the results obtained for new tungsten coils.Data acquisition With computer programs normally used for ICP-AES instruments the registration of transient emission signals is often diYcult. Even with special software for transient signals as oVered by various manufacturers of ICP-AES systems the scanning rates could be too low for the very short signals produced by the tungsten coil vaporizer (about 1 s for most elements). In our previous work,18 only the transient signals of the polychromator could be integrated by the installed standard software. The integration had to be triggered manually before the vaporization step of the voltage–time programme. A long integration time was necessary to ensure complete integration of the emission signals of all elements which however caused an unfavourable signal-to-background ratio.Neither scanning of the emission signals nor coupling of the ETV system with the spectrometer was possible. For these reasons an improved data-acquisition and -processing procedure was developed. For the JY24/74 spectrometer the analog signals of the 16 photomultiplier–amplifiers 15 for the simultaneous channels and one for the monochromator are accessible with only minor modifications at the spectrometer. These signals are Fig. 3 Temperature (&) and power consumption ($) of the tungsten coil as a function of the output voltage of the power supply. recorded by means of a 16-channel analogue-to-digital card with a resolution of 12 bits.The data acquisition software written in PowerBasic has the following main features (a) coupling of the data acquisition system with the ETV system; (b) setting of individual integration times for each element; (c) automatic baseline correction; (d) scanning and storing of the intensity–time profiles of 16 elements with a scanning frequency of about 50 Hz each; (e) the data of the monochromator channel can be used for arithmetic correction of spectral interferences; (f ) the measured data can be exported to commercial computer progams (e.g. charting graphics and spreadsheet programs); and (g) automated measurement cycles are possible. The originally installed spectrometer software is only used for adjusting the mono- and polychromator before analysis.The data acquisition can be started by a trigger signal sent by the ETV computer (Fig. 1). With the above described improvements in instrumentation and software a high degree of flexibility and automation was achieved. Only the pipetting of the sample was done manually. However extension by an autosampler is possible. Reduction of Tungsten Ablation During the vaporization step of the voltage–time programme considerable tungsten ablation from the surface of the tungsten coil was observed.39 Owing to the extremely line-rich emission spectra of tungsten spectral interferences occurred for the analytes Mn Cr Co and Fe (see also Table 2) the extent of the interferences decreasing from Mn to Fe. As the polychromator used for simultaneous determinations (Paschen–Runge mount as utilized in many spectrometers) has a fixed wavelength for each element this problem could not be solved by simply choosing another analyte line.Furthermore the background has to be determined on the maximum of the emission lines making diYcult an appropriate background correction for the elements interfered with by tungsten. Hence the determination of these elements benefits from the reduction of tungsten ablation. As the vapour pressure of tungsten (mp 3410 °C bp 5660 °C)40 is very low (e.g. only 3 mPa at 2600 °C),40 the tungsten ablation during the vaporization step is mainly caused by volatile or easily decomposing compounds formed by the reaction of tungsten with the matrix or traces of oxygen and water in the carrier gas.Further investigations showed that the tungsten ablation caused by the carrier gas can mainly be attributed to traces of molecular oxygen. The concentration of molecular oxygen in the Ar–H2 carrier gas could be reduced by reaction with hydrogen induced by an electric discharge. For this purpose a simple HVDC was constructed and inserted into the carrier gas line after the flow meter (Fig. 1). A further possibility for reducing tungsten ablation is to coat the coil surface with a layer that is more resistant than tungsten to oxidation by oxygen water and many matrices. Tungsten carbide is a refractory compound (WC W2C mp approximately 2860 °C bp 6000 °C)40 of high hardness and chemical resistance. As a tungsten carbide layer on the surface of the coil can be easily formed by reaction with hydrocarbons at high temperatures the tungsten carbide coating (TCC) of the coil was investigated as a means of reduction of the tungsten ablation.Depending on the nature of the tungsten surface and on partial pressure of the hydrocarbon the reaction starts between 1000 and 1500 °C and the optimum reaction temperature is reached between 2000 and 2500 °C.41 Muzgin and co-workers42,43 used an argon–methane mixture for coating a tungsten coil atomizer for atomic absorption spectrometry with tungsten carbide. The aim was an improvement of the atomization and hence the sensitivity of several elements due to the reducing properties of carbon. However no data on tungsten ablation were given.In this work hexane was used as a coating reagent. For the Journal of Analytical Atomic Spectrometry December 1997 Vol. 12 formation of a tungsten carbide layer the coil was heated to 2400 °C in the presence of hexane. New tungsten coils were coated prior to analysis. With a coated coil tungsten ablation cannot be completely avoided obviously owing to partial removal of the carbide layer during the vaporization. To restore the carbide coating a coating step was inserted after each vaporization step (Table 3). In fact this successive carbide coating reduces the lifetime of the tungsten coil; the coil becomes increasingly brittle and the melting point decreases. For a maximum lifetime the carbide layer formed during the coating step should be as thin as possible but suYcient for a reliable decrease in tungsten ablation.Therefore the coating gas flow rate the duration of the coating and the timing of both the coating valve and voltage applied in the coating step (‘coating voltage’) had to be optimized for each matrix. If the coating valve and the coating voltage were switched oV simultaneously a significant decrease in the lifetime of the coil was observed. A possible explanation is the formation of a carbon layer on the surface of the coil during the cool down in presence of hexane the temperatures being suYcient for the pyrolysis of hexane but too low for the formation of tungsten carbide. In the subsequent vaporization step this carbon layer leads to a renewed formation of tungsten carbide which is responsible for the shortened lifetime.Therefore the coating valve has to be switched oV prior to the coating voltage (see also Table 3). In the following the eVect of the HVDC and TCC on tungsten ablation is discussed for the matrices water and silicon carbide. Water When processing water using an uncoated tungsten coil the tungsten emission signal during the vaporization step shows two diVerent sections. Section 1 consists of a strong and sharp signal at the beginning of the vaporization with a duration of about 0.4 s (see Fig. 4 trace A). This peak can be attributed to the vaporization of tungsten oxides previously formed during the drying step by reaction of tungsten with oxygen and water. The most likely reaction products are tungsten oxide WO (mp 1473 °C bp 1750 °C),40 and the tungsten 3 suboxides W2O5 and W4O11 (sublimation point 800–900 °C).40 In section 2 the tungsten signal increases continuously with increase in the tungsten coil temperature reaching its final value after about 1.1 s and then remaining constant until the end of the vaporization step (see Fig.4 trace A). This signal is caused by the vapour pressure of tungsten and by the corrosion of the hot tungsten coil by traces of oxygen and water in the carrier gas. As can be seen from Fig. 4 the HVDC mainly reduces the tungsten ablation in section 1 whereas with the TCC the tungsten ablation in section 2 can be almost Fig. 4 Signals measured using the monochromator at the tungsten emission line at 208.819 nm.Vaporization temperature 2600 °C (17 100 mV 165W). A Without HVDC and TCC; B with HVDC only; C with TCC only; and D with both HVDC and TCC. 1355 Table 4 Decrease in tungsten ablation using the HVDC and TCC for diVerent integration times measured at an interference free tungsten emission line (208.819 nm); n=4. Matrix water; vaporization temperature 2600 °C (17 100 mV 165W) Tungsten signal peak area/mV s 1.7 s 1.2 s TCC HVDC No No Yes Yes No Yes No Yes 2180±25 1305±35 850±25 175±2 1510±45 715±15 840±25 175±2 Fig. 5 Signals measured using the polychromator at the manganese emission line at 257.610 nm. Vaporization temperature 2600 °C (17 100 mV 165W). A–D Signals caused by tungsten interference; A without HVDC and TCC; B with HVDC only; C with TCC only; and D with both HVDC and TCC.Dotted line signal for 100 pg of manganese with HVDC and TCC corrected for tungsten interference by subtracting signal D. completely avoided. In Table 4 the eVect of the HVDC and TCC on the peak area is given for diVerent integration times at a vaporization temperature of 2600 °C. The influence of the HVDC and TCC on the spectral interference caused by tungsten is demonstrated in Fig. 5 for the manganese emission line at 257.610 nm as an example. With the HVDC and TCC the tungsten interference on a 100 pg manganese signal can be reduced from 218% to 24% (integration time 1.2 s see Table 5). Significant reductions were also achieved for the other emission lines interfered with by tungsten although not as pronounced as for the manganese line which suVered the strongest interference.Nevertheless the corrosion of the tungsten coil and thus the introduction of tungsten into the plasma cannot be completely avoided leading to a deterioration of the signalto-background ratio for the emission lines interfered with by tungsten. The optimum coating gas flow rate was found to be 11 ml min-1 resulting in a liftetime of the coil of about 80 vaporizations. Silicon carbide In the vaporization step silicon carbide reacts with tungsten at temperatures above 1400 °C44 to give tungsten carbides TCC HVDC Mn signal* (100 pg Mn) 124±2 No No No Yes Yes Yes No Yes * Mn signal corrected for tungsten interference.Table 5 Decrease in tungsten interference at the manganese emission line at 257.610 nm in processing water using the HVDC and TCC. Integration time=1.2 s; n=4. Vaporization temperature 2600 °C (17 100 mV 165W) Peak area/mV s Journal of Analytical Atomic Spectrometry December 1997 Vol. 12 Transport EYciency For the determination of transport losses i.e. the fraction of an analyte element vaporized in the ETV cell that does not reach the ICP for excitation the radiotracer technique is well suited. The vaporization of the radiotracers under conditions identical with those existing in sample analysis can only be achieved when using in situ labelling by irradiation of the sample with neutrons in a nuclear reactor.Owing to low analyte concentrations and/or low activation yields the activity of the radioisotopes in an irradiated sample was in general too low for the determination of transport losses by in situ labelling. Therefore the slurries were spiked with a radionuclide mixture leading to suYcient activity for the radiotracer experiments. If owing to a diVerent vaporization behaviour the analytes added by the spike are vaporized at a diVerent time to the analytes in the solid this selective volatilization of analytes without a concomitant matrix may lead to a change in transport eYciency.45–47 Therefore the influence of labelling in situ and by spiking the slurries with radiotracers was examined for silicon carbide as an example. The activity of the radioisotopes 76As 24Na and 122Sb deposited in the ETV device after 20 sequential vaporizations of 40 mg of irradiated silicon carbide was high enough to achieve good counting Signal caused by tungsten interference 270±7 124±4 158±2 128 30±1 24 1356 Decrease in tungsten ablation (-fold) 2.5 s 1.7 s 1.2 s 2.5 s 1.5 3.7 18.1 1.7 2.6 12.4 2.1 1.8 8.6 3180±40 — — — 2180±55 850±30 175±3 2 mp approximately 900 °C; W5Si3).41 At a (WC W2C mp approximately 2860 °C bp 6000 °C)40 and silicides (WSi vaporization temperature of 2600 °C which was normally used for analysis only the tungsten carbides are stable.When analysing a silicon carbide slurry (concentration 2 g l-1 sample volume 20 ml ) the carbide layer formed by reaction with the matrix has the same eVect on tungsten ablation as the carbide coating by hexane.With the HVDC the peak areas of the tungsten signal were found to be 165±14 166±15 and 167±14 mV s (mean±standard deviation of five consecutive measurements) for integration times of 1.2 1.7 and 2.5 s respectively. Thus the tungsten ablation caused by silicon carbide (only with the HVDC) is very similar to that given by water (with the HVDC and TCC). Using the HVDC and TCC the background correction for the elements interfered with by tungsten can be performed by measuring the intensity at the maximum of the emission line for the suspension medium water and subtracting this value from the signal for the silicon carbide slurry.Tungsten interference in relation to Mn signal (%) 218 100 Fig. 6 Transport losses for 12 elements determined by means of the radiotracer technique for the matrices water silicon carbide silicon nitride and silicon dioxide (20 sequential vaporization steps). Vaporization temperature 2600 °C; carrier gas flow rate 700 ml min-1. Slurry concentrations SiC 2 g l-1; and Si3N4 and SiO2–1 5 g l-1. Sample aliquot=20 ml. Values in parentheses are sample amounts per vaporization. statistics. The transport losses were found to be 41 44 and 40% for As Na and Sb respectively which are in good accordance with those obtained for the spiked silicon carbide (see Fig. 6). Obviously the very high heating rate of the tungsten coil vaporizer reduces the time diVerence between volatilization of diVerent species of the same element and consequently also the diVerences in transport eYciency between the analytes originating from the sample and from the spike.This is a presupposition for the applicability of a calibration by means of the standard addition method. In the presence of large amounts of a co-volatilizing matrix the absolute amount of analyte vaporized has only a minor influence on the transport eYciency.45–47 Nevertheless for most elements investigated the amount of each radiotracer present in a 20 ml slurry aliquot (sample volume used for vaporization) was kept at trace levels (see also Table 1). For this reason multiple sequential vaporization steps were necessary in order to obtain readily detectable activity of the radiotracers deposited in the ETV device for gamma-ray counting.When processing aqueous standard solutions or slurries of silicon carbide silicon nitride and silicon dioxide the RSDs of the peak areas of the analyte emisson signals were below 5% for 20 vaporizations indicating constant transport eYciencies during the enrichment procedure of the radiotracer experiments. Generally quantitative vaporization of the analytes was observed with the exception of Co Cr Fe and Sc for which residues on the tungsten coil of about 0.6% and 1.5% of the initial activity were found for SiC and Si3N4 respectively. As for a certain selected vaporization temperature the percentages of residues were the same for both coated and uncoated coils the TCC was not used in the radiotracer experiments.The influence of the slurry concentration on the transport losses was examined for silicon nitride as an example. Variations of the amount of sample applied in the range 40–160 mg caused only minor transport loss diVerences between 1.5% (Au) and 7.8% (Cu). Physical considerations of analyte transport losses suggest that the matrix acts as a physical carrier (transport modifier) and improves the transport eYciency with increasing amount of matrix.45–47 This could not be confirmed by the results of our radiotracer investigations under the experimental conditions used. Even with the lowest matrix concentration applied in our experiments suYcent transport modifier for the analytes was pro- Journal of Analytical Atomic Spectrometry December 1997 Vol.12 vided. Therefore a further increase in matrix concentration did not lead to an improvement in the transport eYciency. The transport losses of 12 elements obtained in processing aqueous solutions and slurries of silicon carbide silicon nitride and silicon dioxide are given in Fig. 6 (aqueous solution except for As Cu Mn Na; SiO2 except for Mn). Elements with very diVerent physical and chemical properties were selected permitting the estimation of the transport behaviour of other elements for which no suitable radiotracers were available. The volatilization characteristics of the matrix influence the release of the analytes and the particle size and concentration of the physical carrier formed by the evaporating matrix during the vaporization step.45 Therefore diVerent transport eYciencies would be expected for the matrices SiC Si3N4 and SiO2.However except for Ca La and Sb the transport losses were at about the same level for all matrices investigated; they were in the range 20–45% and the highest transport loss of 54% was obtained for Cu in SiO2. A possible explanation of the minor eVect of the matrix on transport eYciency could be seen again in the very high heating rate of the tungsten coil vaporizer which leads to similar vaporization rates even for very diVerent matrices. With aqueous solutions of Au Cr Fe La and Sc increased transport losses were observed obviously owing to the absence of a matrix (see Fig.6). However this was not found for Ca and Sb. For silicon carbide and silicon nitride the transport losses decreased in general with increasing boiling-point of the element this eVect being more pronounced for silicon nitride (see Figs. 7 and 8). For silicon dioxide the same trend was observed but with the boilingpoints of the respective oxides (see Fig. 9). The large diVerence between the boiling-points of calcium (1484 °C)40 and calcium oxide (2850 °C)40 could be a possible explanation for the significantly reduced transport loss of Ca obtained for silicon dioxide compared with the other matrices. With water as Fig. 7 Dependence of the transport losses for silicon carbide on the boiling-point of the elements. Slurry concentration 2 g l-1. Fig.8 Dependence of the transport losses for silicon nitride on the boiling-point of the elements. The given transport losses are mean values for three slurry concentrations (2 5 and 8 g l-1). 1357 Fig. 9 Dependence of the transport losses for silicon dioxide on the boiling-point of the oxides (As Sb2 2 O3). Slurry concentration 5 g l O-31 A . u2 O3 CaO Cr2O3 La2 O3 matrix the trends of the transport losses for the boiling-points of both the elements and the oxides were not as pronounced as with the other matrices. CONCLUSION Through improvements in instrumentation and software an ETV system for ICP-AES using a tungsten coil as vaporizer was optimized to give a high degree of reliability flexibility and automation. The ETV system was specially adapted for coupling with the ‘classical’ type of ICP spectrometers (separate mono- and polychromators photomultipliers polychromator in Paschen–Runge mount) which are still widely used.Compared with graphite furnace ETV systems the main advantages of the present tungsten coil ETV system include a simple small and inexpensive construction very high heating rates low dilution of the analyte vapours and the absence of memory eVects caused by the formation of stable carbides. On the other hand with graphite furnace ETV devices the introduction of larger sample amounts and direct solid sampling are possible and the furnace material carbon causes no spectral interferences in AES measurements. A substantial decrease in tungsten ablation during the vaporization step causing unspecific background and spectral interferences was achieved by means of a high-voltage discharge cell and tungsten carbide coating of the coil.The reduced entry of tungsten into the ICP has a positive eVect on both the unspecific background and spectral interferences. The radiotracer technique allowed an accurate determination of transport losses for 12 elements with very diVerent physical and chemical properties which were between about 10% and 50%. In general they decreased with increasing boiling-point of the element or the oxide. The described ETV system can be used for multi-element analyses of liquids but is especially advantageous for analysis of diYcult to digest powdered materials in the form of slurries.As only 10–40 ml sample volumes are required for an analysis cycle the method is well suited for the analysis of samples available in only small amounts. REFERENCES 1 Moens L. Verrept P. Boonen S. Vanhaecke F. and Dams R. Spectrochim. Acta Part B 1995 50 463. 2 Zaray D. and Kantor T. Spectrochim. Acta Part B 1995 50 489. 3 Gre�goire D. Miller-Ihli N. J. and Sturgeon R. E. J. Anal. At. Spectrom. 1994 9 605. 4 Wang J. Carey J. M. and Caruso J. A. Spectrochim. Acta Part B 1994 49 193. 5 Verrept P. Dams R. and Kurfu� rst U. Fresenius’ J. Anal. Chem. 1993 346 1035. 6 Nickel H. Zadgorska Z. and WolV G. Spectrochim. Acta Part B 1993 48 25. 1358 Journal of Analytical Atomic Spectrometry December 1997 Vol. 12 7 Vo� llkopf U.Paul M. and Denoyer E. R. Fresenius’ J. Anal. 8 Zucheng J. Bin H. Yongchao Q. and Yun’e Z. Microchem. J. 9 Fonseca R. W. and Miller-Ihli N. J. Appl. Spectrosc. 1995 Chem. 1992 342 917. 1996 53 326. 49 1403. 10 Alary J. F. Hernandez G. and Salin E. D. Appl. Spectrosc. 1995 49 1796. 11 Ren J. M. Rattray R. Salin E. D. and Gre�goire D. C. J. Anal. At. Spectrom. 1995 10 1027. Spectrochim. Acta Part B 1995 50 7. 12 Darke S. A. and Tyson J. F. Microchem. J. 1994 50 310. 13 Florian K. Hassler J. and Schroen W. Fresenius’ J. Anal. Chem. 1996 355 601. 14 Carey J. M. and Caruso J. A. Crit. Rev. Anal. Chem. 1992 23 397. 15 Broekaert J. A. C. Browner R. F. and Marcus R. K. 16 Nickel H. and Zadgorska Z. Spectrochim. Acta Part B 1995 50 527.17 Golloch A. Haveresch-Kock M. and Plantikow-Vossga�tter F. Spectrochim. Acta Part B 1995 50 501. 18 Barth P. and Krivan V. J. Anal. At. Spectrom. 1994 9 773. 19 Hauptkorn S. Krivan V. Gercken B. and Pavel J. J. Anal. At. Spectrom. 1997 12 421. 20 Escobar M. P. Smith B. W. and Winefordner S. D. Anal. Chim. Acta 1996 320 11. 21 Sparks C. M. Holcombe J. A. and Pinkston T. L. Appl. Spectrosc. 1996 50 86. 22 Tao S. and Kumamaru T. Anal. Chim. Acta 1995 310 369. 23 Argentine M. D. and Barnes R. M. J. Anal. At. Spectrom. 1994 9 1371. 24 Lamoureux M. M. Gre�goire D. C. Chakrabarti C. L. and Goltz D. M. Anal. Chem. 1994 66 3208. 25 Matousek J. P. and Mermet J. M. Spectrochim. Acta Part B 1993 48 835. 26 Erwen M. Zucheng J. and Zhenhuang L.Fresenius’ J. Anal. Chem. 1992 344 54. 27 Evans H. E. Caruso J. A. and Satzger D. R. Appl. Spectrosc. 1991 45 1478. 28 Shibata N. Fudagawa N. and Kubota M. Anal. Chem. 1991 63 636. 29 Tsukahara R. and Kubota M. Spectrochim. Acta Part B 1990 45 779. 30 Okamoto Y. Murata H. Yamamoto M. and Kumamuru T. Anal. Chim. Acta 1990 239 139. 31 Sanford C. L. Thomas S. E. and Jones B. T. Appl. Spectrosc. 1996 50 174. 32 Parsons P. J. Qiao H. Aldous K. M. Mills E. and Slavin W. Spectrochim. Acta Part B 1995 50 1475. 33 Bruhn C. G. Ambiado F. E. Cid H. J. Wo� rner R. Tapia J. and Garcia R. Anal. Chim. Acta 1995 306 183. 34 Krug F. J. Silva M. M. Oliveira P. V. and No� brega J. A. Spectrochim. Acta Part B 1995 50 1469. 35 Havesov I. Ivanova E. Berndt H. and Schaldach G. Fresenius J. Anal. Chem. 1990 336 484. 36 Berndt H. and Schaldach G. J. Anal. At. Spectrom. 1988 3 709. 37 Williams M. and Piepmeier E. H. Anal. Chem. 1972 44 1342. 38 Dittrich K. Berndt H. Schaldach G. and To� lg G. J. Anal. At. Spectrom. 1988 3 1105. 39 Barth P. Hauptkorn S. and Krivan V. J. Anal. Atom. Spectrom. 1997 12 1359. 40 Handbook of Chemistry and Physics ed. Weast R. C. CRC Press Boca Raton FL 62nd edn. 1982. 41 Gmelin Handbook of Inorganic Chemistry T ungsten System No. 54 suppl. vol. A5b Springer Berlin 8th edn. 1993. 42 Muzgin V. N. Atnashev V. B. Pupyshev A. A. and Atnashev Yu. B. Zh. Anal. Khim. 1986 41 1798. 43 Muzgin V. N. Atnashev Yu. B. Korepanov V. E. and Pupyshev A. A. T alanta 1987 34 197. 44 Gmelin Handbook of Inorganic Chemistry T ungsten System No. 54 suppl. vol. A7 Springer Berlin 8th edn. 1987. 45 Kantor T. Spectrochim. Acta Part B 1988 43 1299. 46 Ediger R. D. and Beres S. A. Spectrochim. Acta Part B 1992 47 907. 47 Kantor T. Fresenius’ J. Anal. Chem. 1996 355 606. Paper 7/05321E Received July 23 1997 Accepted