T r Automated Sampling System for the Direct Determination of Trace Amounts of Heavy Metals in Gaseous Hydrogen Chloride by Atomic Absorption Spectrometry BERND BAASKE AND URSULA TELGHEDER* Department of Instrumental Analytical Chemistry University of Duisburg Lotharstr. 1 4705 7 Duisburg Germany An analytical procedure is described for the determination of iron nickel chromium and manganese in gaseous hydrogen chloride by means of a modified atomic absorption spectrometer. An automated sampling system allows a 'quasi' on-line monitoring of the investigated gas. The introduction of the gas was regulated by a magnetic valve which is connected to the autoprobe of the spectrometer and to a control unit. Typical parameters e.g. the temperature programme and the gas injection volume were optimized.Different types of calibration were carried out using element standard solutions standard additions of gaseous standards and standard solutions as well as gaseous standards. The detection limit for the described procedure was 39 pg for iron 8.7 pg for nickel 1.1 pg for chromium and 0.4 pg for manganese (temperature = 25 "C) using the standard calibration with element standard solutions. The reproducibility of the absorbance signals for the elements in gaseous hydrogen chloride varies between 5.3 and 13% (n= 10). Keywords Electrothermal atomic absorption spectrometry; analysis of reactive gases; automated gas sampling system; determination of heavy metals The analysis of reactive gases is of increasing importance particularly because of their use in the production of semiconductors.The types of metallic impurities in gases have also not been clarified. Reactive gases are generally contained under pressure in steel cylinders. The cylinder and the cylinder valve must be considered together with the sample as it is not possible to dispose of the packaging. The pressure of the gas must be reduced to atmospheric pressure for the purpose of analysis. The sources of contamination could be the valve system the corrosive gas itself or both. The analysis of the metallic contaminants of gases must not only take into account particle- bound impurities but also components in liquid and gaseous states. Flaherty et al.' have published results from the examin- ation of the dynamics of particles and metal contamination from gas cylinders regulators and valves.They found that for corrosive gases metallic impurities exist in the form of univer- sally distributed particles and that the cylinder valve is a significant source of metal contamination. Hence different methods exist for the investigation of the contaminants; for example indirect methods such as filtration and hydrolysis. It is possible to differentiate between particle-bound impurities and components in liquid and gaseous states by using these two methods. Faix et aL2 described two analytical procedures for the determination of particle-bound trace metals in high- purity hydrogen chloride. They used polycarbonate filters with * To whom correspondence should be addressed. Journal of Analytical Atomic Spectrometry a pore diameter of 0.05 pm for sampling.The subsequent determination of Br Mg Mn Na Sb Sn Te Ti and Zn was carried out by instrumental neutron activation analysis (NAA) and the determination of Cr Cu Fe Mn and Ni by electrother- mal atomic absorption spectrometry (ETAAS). Denyszyn et al. accomplished the sampling of trace metal impurities in gaseous hydrogen chloride in two steps ( 1 ) particulate material was collected on a membrane filter and (2) aerosol or gaseous material was collected in a Greenburg-Smith impinger. The determination of Fe Cr Ni and Cu was performed by ETAAS. The concentrations of the metal impurities measured in the gas-phase filters were in most instances 25-50% higher than the background measurements of the filters. The impinger data were higher than the filter results. Cui et aL4 have investigated trace amounts of Al Ca Cr Cu Fe U Mn Na Pb and Zn in high-purity gases such as N H2 HCl NH3 B2Hs SiH ASH and PH,.They discussed different sampling methods including the conditions for the absorption of the gas in solution and for matrix separation. Subsequently ETAAS methods were established. The determination of ASH in PH has been investigated by Scharf et aL5 The gas was absorbed in impinger gas-bottles filled with HNO and the solutions were analysed by hydride generation AAS with NaBH as the reductive reagent. Another possibility of sampling is the absorption of the trace metals on a solid material.6 Relatively few direct methods have been published for the determination of metallic contaminants particularly in reactive gases.Hutton et ~ 1 . ~ investigated the direct determination of As in SiH4 by inductively coupled plasma mass spectrometry (ICP-MS). They used an alloy sample cone which operated at a higher temperature than the commercially available nickel cones. The addition of hydrogen to the carrier gas further increased the analyte signal level. Schram' described a bypass-backflush balancing system for the direct introduction of gaseous hydro- gen chloride connected to an inductively coupled plasma atomic emission spectrometry (ICP-AES) instrument. A widely applied technique is the use of a sealed ICP (SICP) for the analysis of SiH4,11 HC112 and chlorine13 as described by Barnes and co-workers. The SICP provides complete sample containment. Only a small volume of the toxic gas needs to be used.The plasma is sustained with the discharge gas and sample enclosed together inside a quartz container. Initial investigations into the development of a microwave- induced plasma atomic emission spectrometry (MIP-AES) system to determine trace metals in gases have been de~cribed.'~ The atomic emission spectra were generated by using a medium power microwave-induced plasma with a rectangular-type cavity. A continuously running sampling system was developed for the direct determination of the iron concentration in gaseous hydrogen chloride. An analytical procedure was described for the determination of iron in gaseous hydrogen Journal of Analytical Atomic Spectrometry December 1995 Vol. 10 1077chloride by means of a modified atomic absorption spec- trometer.15 The gas inlet was realized by a special graphite capillary.Typical parameters were optimized and the cali- bration was carried out using iron standard solutions. The gas injection was carried out manually with a gas-tight syringe. Based on this fundamental knowledge the automation of this modified atomic absorption spectrometer and further improvements to the procedure are described here. EXPERIMENTAL Sampling The sampling procedure is important and difficult for the continuous analysis of reactive gases. The material that will be in contact with the sample gas must be inert and the transport without any variations. In principle a gas introduc- tion system using a peristaltic pump15 is suitable for this purpose because it is possible to regulate the gas flow after opening the hydrogen chloride cylinder during a measurement without moving any valves. On the other hand the pulsation of the pump does not allow a constant gas flow.Hence the calculation of the amount of gas during a monitoring process is very difficult if not impossible. For further investigations a system with a reducing valve made of stainless steel suitable for hydrogen chloride (Fig. l) was used. After passing the reducing valve (4) the gas flow is divided into two parts one leading to a mixing chamber ( 6 ) where the sample gas could be mixed with argon (2) or any calibration gas (3) the other to a scrubber (10). The sample gas flow is regulated by a flow meter ( 5 ) to ensure a constant analyte flow. The gas introduction is regulated by a magnetic valve (8).The flow meter and all tubes which come into contact with the reactive gases are made of non-metallic materials such as poly (propylene) poly (tetrafluoroethylene) (PTFE) or glass. Trace amounts of water vaporized from the scrubber that back-diffuse into the system are adsorbed by tubes filled with silica gel (7). 9 " I 7 Fig. 1 Gas introduction system 1 HCl; 2 Ar; 3 Fe(CO) in Ar; 4 reducing valve; 5 flow meter; 6 mixing chamber; 7 tube filled with silica gel; 8 magnetic valve; 9 washing bottle; 10 scrubber; and 11 Solaar 939 AA spectrometer with graphite furnace (Unicam) For the analysis of corrosive gases on atomic absorption spectrometer (Unicam) with a graphite furnace and with an autoprobe unit was used.The graphite tubes used for the autoprobe technique had two holes one for sample introduc- tion and a second wider hole drilled into the tube at 90" to the former hole. A suitable gas inlet was realized by a special quartz capillary connected to a tube made of poly(propy1ene) (Fig. 2). The outer diameter of the quartz capillary is 2 mm and the inner diameter is 1 mm. The length of the capillary is 80mm. The magnetic valve and the driving motor for the autoprobe are connected to a control unit. This control unit is a potentiometer which adjusts the rate of feeding for the capillary. The electric pulse to start the motor is given by the spectrometer programme. During the transport of the capillary to the graphite tube the motor has to operate against a spring which sustains the sledge in position during the injection step.The motor has been switched off by a light barrier before it starts. Subsequently the motor will be activated and the capillary will be led back by the spring. The dwell time of the capillary in the graphite tube and hence the sample introduc- tion can be fixed by the spectrometer programme. In order to ensure that the magnetic valve opens at the moment when the capillary is just at the hole of the graphite tube a period of retardation is set by a second potentiometer. This retardation has to be considered in the spectrometer programme. Reagents Germany.) Argon 4.6 99.996%. (Messer Griesheim Duisburg Hydrogen chloride 2.8 99.8%. (Messer Griesheim.) Hydrogen chloride 5.0 99.999%. (Messer Griesheim.) Iron calibration solution (1.000 g I-') iron(111) nitrate in 0.5 moll-' nitric acid. (Bernd Kraft Duisburg Germany.) horizontal position screw B sledge / / I vertical adjustment depthtadjustment screw A injection hole screw c slot for the probe q- Autoprobe graphite tube c I Fig.2 Autoprobe unit capillary 1078 Journal of Analytical Atomic Spectrometry December 1995 Vol. 10Nickel calibration solution (1.000 g 1-I) nickel(I1) nitrate in 0.5 moll-' nitric acid. (Bernd Kraft.) Copper calibration solution (1.000 g 1- l) copper(r1) nitrate in 0.5 mol 1-' nitric acid. (Bernd Kraft.) Manganese calibration solution ( 1.000 g 1-I) manganese(I1) nitrate in 0.5 moll-' nitric acid. (Bernd Kraft.) Dilutions were carried out in ultrapure water (Milli-Q-Plus Millipore Eschborn Germany) and the resulting solutions were stored in poly( propylene) bottles.Sub-boiled hydrochloric acid 37% pro analysi. (Merck Darmstadt Germany.) Calibration gas iron pentacarbonyl 6.3 ppm (v/v). (Messer Griesheim.) -$ 0.30 2 0.25 -e 0.20 -0 Q 0.15 0.10 0.05 - E - - Instrumentation - - - -. .- .- Flow meterlgas mixer. Suitable for hydrogen chloride (PTFE) range 0-5 ml min-' and 2-26 ml min-' (Novodirect Kehl Germany). Scrubber. Buchi 412 (Biichi Flawil Switzerland). Spectrometer. Solaar 939 AA spectrometer with Solaar GF90 Sampler. Solaar FS 90 (Unicam). electrothermal atomization unit (Unicam Kassel Germany). RESULTS AND DISCUSSION Experimental Parameters The elements iron nickel chromium and manganese in gaseous hydrogen chloride were determined with the automated ETAAS system described above.The parameters for the deter- mination of iron in gaseous hydrogen chloride will be discussed as an example. The measurements were carried out at the resonance wavelength of 248.3 nm with a slit-width of 0.2 nm. Deuterium background correction was used. Atomization could be achieved at a temperature of 2300°C. In contrast to earlier rneasurement~,'~ absorbance on the spectral line for iron was observed if pyrolytic graphite coated graphite tubes were used. It is assumed that traces of iron in the form of iron carbide are deposited on the walls of the graphite tube. Further investigations to explain this phenomenon are required. The injection volume can be controlled by opening the magnetic valve and depends on the opening time of the valve (see Fig.1); for all investigations a volume of 0.43 ml was used. Finally the temperature programme of the furnace which was used for the determination of trace amounts of iron in gaseous hydrogen chloride is shown in Table 1. Calibration of the Analytical System When performing a calibration it is important that the matrix of the sample and standards be as similar as possible. In this instance gaseous hydrogen chloride is not available as a calibration gas. However iron pentacarbonyl diluted in Ar can be used as a gaseous calibration standard. The decomposition of this gaseous standard starts at 50°C. Initially calibration was carried out with dilute hydrochloric acid standard solution^.'^ It was assumed that the iron atoms produced from the solutions and those produced from the gas have the same spatial distribution in the graphite tube and that the temperature is constant during both peaks.Furthermore the matrix of the gaseous hydrogen chloride has to be similar to the matrix of the dilute hydrochloric acid solutions after drying and ashing of the solutions. Additionally it would be sensible to investigate a gaseous standard in order to exclude the influence of the matrix. Gaseous iron penta- carbonyl diluted in Ar was used for calibration because no certified gaseous hydrogen chloride standard is available. In order to compare different calibration methods and to evaluate the best fitted type of calibration (closer to the sample matrix) the following calibration procedures were investigated. (i) Calibration with dilute hydrochloric acid standard solu- tions.For this calibration six standard solutions [sub-boiled hydrochloric acid (1 + 99) with concentrations of iron in the range 0.16-0.8 ng per 20 p13 were analysed. (ii) Standard additions of gaseous iron pentacarbonyl diluted in Ar [ciron = 6.3 ppm (v/v)] and dilute hydrochloric acid standard solutions. For this type of calibration a constant volume of standard solutions with concentrations in the range 0.02-0.8 ng of iron per 20 pl was injected into the graphite tube. After drying (T= 120°C) ashing (T= 1200°C) and cool- ing (T=30O0C) a constant volume of gaseous iron penta- carbonyl (mFe = 0.084 ng) was added. Subsequently the whole sample was atomized at 2300°C and the graphite tube was cleaned at 2500°C.(iii) Calibration with gaseous iron pentacarbonyl diluted in Ar [ciron = 6.3 ppm (v/v)] . The different calibration graphs are shown in Fig. 3. The intensities of the absorbance signals correspond to the peak areas. The average was calculated from six results. The iron concentrations were calculated in terms of absolute mass for the possibility of comparison. Supposing that iron pentacarbonyl reacts as an ideal gas the volume of 1 mol of iron is 22.414 1. An iron concentration of 6.3 ppm (v/v) is equivalent to 6.3 pl of iron per litre of Ar or 2.81 x lo-' mol of iron. Taking into consideration the gram- 0.35 A 1 x / / ,"' A:y = 0.4031~ + 0.0173 B:y = 0.3227~ +0.0258 0 ' C:y =0.3207~ +0.0034 0 d 0 - 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 Absolute mass of irordng Fig.3 Calibration graphs by A gaseous Fe(CO) in Ar; B standard additions of gaseous Fe(CO)5 in Ar and dilute HCl standard solutions; and C dilute HC1 standard solutions Table 1 Temperature programme for the determination of trace amounts of iron in gaseous hydrogen chloride Step Temperature Argon gas flow rate/ TemperaturerC Time/s increasePC s - ml min-' Sample introduction 1100 4 1100 Adsorption of trace amounts of iron on the graphite tube 1100 4 1100 Removal of the gaseous hydrogen chloride 1100 7 1100 Atomization 2300 8 TC* Cleaning 2500 5 TC* 200 0 300 0 300 * TC Temperature controlled by an optical sensor. Journal of Analytical Atomic Spectrometry December 1995 Vol. 10 1079molecular mass of iron (M = 55.85 g mol-l) the concentration of mass is 0.0157mg of iron per litre of Ar.The absolute masses in Fig. 3 depend on the injection time and the flow of the gas. Fig. 3 shows that the slopes of the calibration graphs obtained with dilute hydrochloric acid standard solutions and by the standard additions of gaseous iron pentacarbonyl-Ar and dilute hydrochloric acid standard solutions are almost identical. The small difference shows that the influence of the matrix is minimal. Thus the method of standard additions as described above is the most suitable type of calibration for the analysis of hydrogen chloride particularly since the result obtained by the extrapolation of the regression line to the abscissa (0.086 ng of iron) corresponds to the calculated mass (0.084ng of iron). Hence the calibration for other elements such as nickel chromium and manganese could be carried out with dilute hydrochloric acid standard solutions or by standard additions. The detection limit was found to be 39pg of iron [standard deviation s = 11.4% (n = 6 ) ] by calibration with standard solutions and 26 pg of iron [ s = 12.9% (n=4)] by calibration with gaseous iron pentacarbonyl-Ar.In both instances the detection limit (DL) was calculated by DL = 3sS-I (1) (s = standard deviation of the blank S = sensitivity of the calibration and 3 =statistical value). Calculation of Element Concentrations in Gaseous Hydrogen Chloride The determination of iron nickel chromium and manganese was carried out by the standard calibration procedure using the parameters described above. All the analytical signals were calculated by means of peak areas.The element concentrations in gaseous hydrogen chloride are shown in Table 2. Investigation of Gaseous Hydrogen Chloride of Different Quality Reactive gases are generally contained under pressure in steel cylinders. The cylinder and the cylinder valve must be con- sidered together with the sample as it is not possible to dispose of the packaging. The pressure of the gas must be reduced to atmospheric pressure for the purpose of analysis. The sources of contamination could be the valve system the corrosive gas itself or both. Hence the analysis of gaseous hydrogen chloride with purities of 99.999 and 99.8% respectively could show whether there is a difference between the content of contami- nants. Fig. 4 shows the dependence of the absorbance signal of iron on the timing of sampling.The sampling of hydrogen chloride was carried out every 70 s after opening the valve. In both instances the absorbance was highest just after opening the hydrogen chloride cylinder and then decreased slowly to a constant value. Thus a steady state was reached about 7 min after opening the gas cylinder. It was found that at this time no significant difference between the iron concentration in gaseous hydrogen chloride with purities of 99.999 and 99.8% respectively could be observed. Hence it seems that the valve system is one of the main sources for the contamination of gaseous hydrogen chloride with iron irrespective of the purity of the gas. CONCLUSION The described ETAAS system with an automated sampling system allows the determination of iron nickel chromium and manganese in gaseous hydrogen chloride.The problem of calibration has been discussed. A comparison of calibration Table 2 Element concentrations in gaseous hydrogen chloride (T= 25 "C) [values in parentheses are relative standard deviations (W) n= 101 ( E integrated absorbance) Element concentration Element Sensitivity1 DL*/pg in gas/pg 1-' E Pg-l Iron 0.0003 39 19.5 (5.3) Nickel 0.0004 8.7 3.4 (13) Chromium 0.0016 1.1 < DL-f Manganese 0.0043 0.4 < DLt * DL = Detection limit. t Below the detection limit. 0.6 1 4 8 12 16 20 24 Tirne/min Fig.4 Dependence of the integrated absorbance on the timing of sampling with standard aqueous solutions as well as standard additions and calibration with gaseous standards has shown that the influence of the matrix is minimal.This fact is interesting for the calibration for other elements for which no calibration gas exists. Nevertheless in those instances calibration using stan- dard solutions is possible. Further investigations with different gases used in the microelectronics industry are planned. This work is part of a JESSI-project and is supported by Messer Griesheim GmbH. The authors are grateful to Dr. Eschwey Messer Griesheim GmbH. REFERENCES 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Flaherty 1;. T. John L. and Amato A. F. Solid State Technol. 1992 35 S1. Faix W. G. Schramm W. Vix F. Weichbrodt G. and Henkelmann R. Fresenius' 2. Anal. Chem. 1988 329 847. Denyszyn R. B. Yin I. H. and Bandy T. paper presented at Microcontamination West Conference Anaheim 1989. Cui X. Xu X. Yan X. and Lang W. Bandaoti Xuebao 1989 10( 12) 94s. Scharf H. Hahn E. and Emrich G. 2. Chem. 1990,30(3) 107. Miyazaki K. and Nakagawa K. Koatsu Gasu 1992,29 281. Hutton R. C. Bridenne M. Coffre E. Marot Y. and Simondet F. J. Anal. At. Spectrom. 1990 5 463. Schram J. Fresenius' J. Anal. Chem. 1992 343 727. Jacksier T. and Barnes R. J. Anal. At. Spectrom. 1992 7 839. Jacksier T. and Barnes R. Spectrochim. Acta Part B 1993 48 (6/7) 941. Jahl M. J. and Barnes R. J. Anal. At. Spectrom. 1992 7 833. Jacksier T. and Barnes R. J. Anal. At. Spectrom. 1994 9 1299. Jacksier T. and Barnes R. Spectrochim. Acta Part B 1994 49 (8) 797. Kirschner S. Golloch A. and Telgheder U. J. Anal. At. Spectrom. 1994 9 971. Baaske B.. Golloch A. and Telgheder U. J. Anal. At. Spectrom. 1994 9 867. Paper 51039021 Received June 16 1995 Accepted August 18 1995 1080 Journal of Analytical Atomic Spectrometry December 1995 Vol. 10