Continuous Hydride Generation System for the Determination of Trace Amounts of Bismuth in Metallurgical Materials by Atomic Absorption Spectrometry Using an On-line Stripping-type Generator/ Gas–Liquid Separator SOLANGE CADORE* AND NIVALDO BACCAN Universidade Estadual de Campinas, Instituto de Quý�mica, Caixa Postal 6154, 13.083–970 Campinas, SP, Brazil A method for the determination of trace amounts of bismuth chemist with an excellent tool to detect and quantify hydrideusing flow injection and atomic absorption spectrometry with forming elements, owing to its high sensitivity and hydride generation was developed. The introduction of 50 ml of simplicity.13–16 sample and tetrahydroborate solution into the HCl and The concept of the technique of hydride generation was aqueous carriers in a merging zones manifold allows the developed by Holak17 and is based on the reaction between formation of bismuthine, which is separated from the liquid the acidified sample and a reducing agent, which forms the phase in an on-line stripping-type generator/gas–liquid volatile hydride.Since then, eorts have continually been made separator. The calibration graph is linear from 0.1 to to improve the reproducibility and sensitivity, introducing 100 ng ml-1 Bi with a detection limit (3s) of 320 pg ml-1 Bi tetrahydroborate as reductor18 and quartz atomization tubes.19 (corresponding to 16 pg Bi). The relative standard deviation In order to increase the analytical signal it is recommended for 20 replicates varies from 10% for 0.1 ng ml-1 Bi to 1.9% that the hydride should be generated and transferred to the for 100 ng ml-1 Bi, with an injection frequency of up to 150 atomizer as quickly as possible, diminishing its dilution by the samples h-1.NiII, CoII, AgI, HgII, SeIV and SbIII interfere, but carrier gas. The concept of flow injection analysis (FIA)20 is they can be masked with a thiourea–KI solution. The prescribed as the method of choice to overcome this problem applicability of the proposed method to metallurgical samples of dilution and increase the sample throughput.Low sample was demonstrated by the analysis of certified reference volumes are involved and miniaturization, which reduces the materials. contact time between the reagents, minimizes interference eects.21 The combination of bismuthine (BiH3) generation Keywords: Bismuth determination; metallurgical materials ; with atomic absorption spectrometry has been the subject of flow injection hydride generation; atomic absorption a number of investigations with automated methods,21–25 in spectrometry dierent materials.A° stro�m21 used an FIA system in which a bismuth sample of Quality control of industrial products demands a continuous 700 ml was injected into a continuous flowing stream of HCl. development and improvement of new analytical chemical Under the best conditions for the system, a detection limit of methods.The presence of elements at trace levels requires 0.08 ng ml-1 Bi was obtained. Chan et al.22 described an instrumental methods which in turn might be automated. In automated method in which the sample reacted with a masking particular, the presence of hydride-forming elements in steel reagent and the resulting mixture was introduced into an HCl may impart deleterious changes to the physical properties of stream. The hydride, once generated, was separated and trans- this material, as desirable or undesirable eects. Among these ported to the atomizer by an argon flow, after passing through elements, bismuth is important because its addition to metallur- an H2SO4 impinger in order to eliminate excess of water gical materials can aect their quality, positively or negatively, vapour, resulting in a detection limit of 20 ng g-1 Bi.Chan depending on its concentration and the composition of the and Hon23 utilized a flow injection (FI) system with the material.1,2 It has been reported that the addition of bismuth introduction of 200 ml of sample solution, obtaining a detection to Al–Mg alloys improves their corrosion resistance due to limit of 0.17 ng of bismuth using calibration by standard the protective action of Bi2Mg3 compounds in its structure.3 additions.The determination of bismuth and other elements As well as avoiding the formation of graphite nodules, bismuth forming volatile covalent hydrides was described by Schmidt also promotes iron carbide stabilization during the solidifi- et al.;24 according to these workers, the sensitivity of the cation process, when added to iron or steels.4 On the other automated system is comparable to, and, in most cases, better hand, even small amounts of this element may produce a than the manual technique and is considered superior with decrease in hot ductility, workability and cause the rupture of respect to reproducibility and ease of operation. A detection alloys5 and steels.6 limit of 0.24 ng ml-1 Bi was obtained with a high sample flow Among the methods that have been used for bismuth rate (50–100 ml min-1).Yamamoto et al.25 found that gas determination are those based on spectrophotometry7,8 and segmentation with N2 in FIA is an eective method to prevent polarography,9 but they lack sensitivity or selectivity. Flame sample zone broadening during mixing before gas generation. atomic absorption spectrometry10 shows selectivity but the When 0.5 ml of sample was used, Bi, As, Se, Sb and Te were sensitivity is not sucient for metallurgical analysis.Graphite determined with detection limits of 0.04–0.3 ng. furnace atomization has good sensitivity,5,11,12 but occasionally The eciency of the separation of the generated hydride matrix elements other than the analyte may be volatilized from the liquid phase is important in order to ensure that the together, causing interferences.The hydride generation method, coupled with atomic absorption, has provided the analytical analyte is transferred to the atomizer. This is provided by a Journal of Analytical Atomic Spectrometry, June 1997, Vol. 12 (637–642) 637Fig. 1 Schematic diagram of the hydride generation system. L= Injection volume; RC=reaction coil; PP=peristaltic pump; AAS= atomic absorption spectrometer. gas–liquid separator and the most common type is a U-tube;21,26 however, sometimes it is dicult to obtain a constant drain of the liquid phase and problems such as dilution of the generated hydride, which causes a low signal and low reproducibility, might occur.Many other designs have been described in the literature, making modifications to improve the hydride separation. A packed U-tube,27 a cooled U-tube,28 porous membranes29 and a porous tube30 separator have all been proposed, having smaller dead volumes than Utube separators. Despite the higher sensitivities claimed, most of these separators show the best performance with a high acid concentration, producing a large amount of H2 when in contact with the tetrahydroborate solution.A very convenient method to remove the generated hydride from the aqueous Fig. 2 Stripping-type gas–liquid reactor–separator used in the con- medium was described by Schmidt et al.24 as a typical forced tinuous-flow hydride generation system. outlet separator made with a medium porosity Bu�chner fritted glass funnel with a stopper inserted in the upper end.Three (Micronal, Model B332 II). Samples and reagents were aspir- tubes were inserted through the stopper to accommodate the ated through Tygon tubes (Technicon) with appropriate flow sample input and gaseous hydride, argon and waste removal. rates for each solution. Sample loops, reaction coils and The argon used to purge the hydride was introduced through transmission lines were prepared using Teflon tubing (CPL, the lower end of the funnel.This system works with large 0.8 mm id). For the introduction of solutions into the system, volumes on-line. Brindle and co-workers31,32 described another a proportional injector made from acrylic was used.33 For the forced outlet gas–liquid separator based on a glass frit for the separation of the formed hydride from the liquid phase a determination of arsenic and antimony using DC In stripping-type gas–liquid reactor–separator was utilized, both designs, sample and reducing agent solutions are continuinstead of a U-shaped25 separator.The separator used consists ously introduced into the generator/separator chamber. An of three parts made of acrylic and is shown in Fig. 2. The hole argon flow of 0.4–0.5 l min-1 was introduced into the generator through which the mixture (gaseous hydride+liquid fraction) through the glass frit to remove the hydrides from solution. enters the separator and those through which the liquid Another flow of argon (2 l min-1) was introduced into the fraction leaves were at the same height, allowing the presence vessel from the top. The high flow of argon sweeps the hydrides of a small liquid residue (the residence volume is about 200 ml) into the plasma through the outlet part of the device via a U-tube buer tank, partly filled with water.This buer tank is inside the separator. N2 introduction at the bottom strips the necessary in this design in order to moderate any pressure hydride through a sintered glass frit to the atomizer, using a fluctuation in the hydride transportation line to the plasma.Teflon tube (10 cm×3 mm), in a uniform and constant way. This paper describes a simple stripping-type gas–liquid This enhances the separation eciency which is related to the separator that eciently enhances the separation and transpor- sensitivity. tation of BiH3 to the quartz atomization cell. This eciency After separation, the hydride was transported with a conis reflected in a greater sensitivity and excellent reproducibility, trolled N2 flow rate to an electrothermal cell, constructed from due to the low residence volume (about 200 ml) inside the a 170×8 mm id quartz tube, fused at the centre with a chamber.Furthermore, small sample and reagent volumes 100×2 mm id quartz tube to form a T-shaped atomizer. The (50 ml) are used in a merging-zones configuration of the FI cell was wound with 2 m of Ni–Cr wire (0.45 mm diameter; manifold.The N2 flow, used as purge gas, was 120 ml min-1, 8.3 V m-1) and isolated with an asbestos strip, except for the causing an ecient mixing and removal of the bismuth hydride 2 cm extremities of the tube. This allowed more stable signals with subsequent transference to the atomization cell. The because of the elimination of ignition of the H2 at the end of relevant parameters were optimized taking advantage of this the tube.34 In order to obtain higher temperatures and to stripping-type generator/gas–liquid separator, increasing the minimize heat changes, the cell was surrounded with a two eciency of the separation of BiH3 from the liquid phase.part aluminium tube, with a 2 mm thick tube-wall, fixed The interference eect of a number of ions was evaluated as together under pressure, as shown in Fig. 3. Before use, the well as the use of masking agents. The method developed was cell was washed with a 1+9 HF solution and then treated successfully applied to the determination of bismuth in metal- with 5% m/v dichlorodimethylsilane or trimethylchlorosilane lurgical materials. in toluene for 2 h and dried for 1 h at 90–100 °C.The silane solution reacts with hydroxyl groups, eliminating active sites on the glass surface, allowing the best sensitivity and repeat- EXPERIMENTAL ability of the desired reaction.35 This treatment was repeated Apparatus after about 500 injections, and could be repeated as often as necessary.The BiH3 atomization temperature was obtained The FI system is shown in Fig. 1 and consists of a mergingzones manifold33 with a six-channel peristaltic pump with a Varivolt regulator, which is connected to the Ni–Cr 638 Journal of Analytical Atomic Spectrometry, June 1997, Vol. 12pared with those obtained with graphite furnace atomization, according to the procedure described by Gladney.36 RESULTS AND DISCUSSION Using the merging zones manifold, equal volumes of sample and reducing agent are introduced into separate carrier streams, which have the same flow rate in order to ensure perfect homogenization of the solutions.The generation of BiH3 occurs in a reaction coil which is located after the confluence Fig. 3 Heated quartz tube used as the electrothermal atomizer for on-line atomization of BiH3. point of the system. As the hydride is generated in an acidic medium the carrier for the sample is an HCl solution while de-ionized water was chosen for the tetrahydroborate carrier, because of the formation of H2 in the presence of acid.When the tetrahydroborate reacts with the acidified sample solution the following reactions take place: BH4-+3H2O+H+�H3BO3+8H Bi3++6H �BiH3+3H+ The hydride is separated from the liquid phase in a gas– liquid separator and then swept to the atomizer where, according to Welz and Melcher,37 the atomization in the quartz tube Fig. 4 Dependence of the absorbance on the reaction coil length for is due to collisions between hydrogen radicals: dierent analyte concentrations. CHCl=1.0 mol l-1 ; CNaBH4 =1.0%; Tat=900 °C; N2 flow rate=110 ml min-1.BiH3+H � BiH2+H2 BiH2+H � BiH+H2 coil of the cell. The electrothermal cell was aligned in the optical path of an atomic absorption spectrometer (Varian, BiH+H�Bi+H2 Model Gemini AA 12/1475) equipped with a deuterium background corrector. During all the optimization steps, the background corrector was kept on in order to check for the existence of significant background.We did not observe any Optimization of FI Hydride Generation System relevant background that needed to be corrected. The bismuth Reaction coil (reaction time) and injection volume hollow cathode lamp was operated at 8 mA with a wavelength of 223.1 nm and a slit-width of 0.2 nm. The signals obtained The BiH3 is formed when the sample reacts with a tetrahydrowere registered by an Epson LX-800 printer. borate solution; hence, the time for which the solutions are in contact was evaluated.Dierent lengths (10, 15, 30, 50 and 75 cm) of reaction coil were prepared and tested while also Reagents varying the injection volume (30, 50, 75, 100 and 150 ml) for All reagents used were of analytical-reagent grade. De-ionized three dierent concentrations of bismuth (40, 80 and water was used throughout. 120 ng ml-1). For each reaction coil the absorbance signal The bismuth standard solution was prepared from metallic increases with the increase in injected volumes.Fig. 4 shows bismuth (J. T. Baker, 99.99%) treated with concentrated HCl the eect of the reaction coil in the determination of bismuth and HNO3 and finally diluted with 1.0 mol l-1 HCl. Appropriate dilution was made from this solution, whenever necessary, with 1.0 mol l-1 HCl. Solutions of sodium tetrahydroborate were prepared by dissolving NaBH4 powder (Merck) in 0.05 mol l-1 KOH and were stored in plastic bottles, under refrigeration.The solution is stable for about 2 weeks, with no loss of the observed absorption signal. Solutions of interferent ions and masking agents were prepared by the dissolution of appropriate salts in acid or de-ionized water. Fig. 5 Eect of the amount of NaBH4 used on the hydride generation with dierent analyte concentrations. Vinj=50 ml; RC=15 cm; N2 flow Samples rate=120 ml min-1; Tat=900°C; carrier flow rate (HCl and H2O)= 1.9 ml min-1.T in alloy. A 0.1 g amount of certified alloy was dissolved in 3 ml of concentrated HNO3 and heated. The filtered solution Table 1 Peak area for the injection of 4 ng Bi. Reactor–separator was treated with concentrated HCl and heated. The sample height: 6 cm; 1.0 mol l-1 HCl; 1.0% NaBH4 in 0.05 mol l-1 KOH; N2 was diluted to 50.0 ml with 1.0 mol l-1 HCl. at 120 ml min-1 Steels. A 0.1 g amount of material was treated with 3 ml of Injection volume/ Bismuth concentration/ Peak area/ concentrated HNO3. After evaporation, concentrated HCl was ml ngml-1 As added and the solution was heated.The sample was diluted 30 133.3 3.849 to 10.0 ml with 1.0 mol l-1 HCl. 50 80.0 3.901 Bronze/Brass. Appropriate amounts of material were treated 75 53.3 4.058 with concentrated HNO3 and HCl and heated. After filtration, 100 40.0 3.846 the solutions werwith 1.0 mol l-1 HCl. The results 150 26.7 3.665 obtained with hydride generation measurements were com- Journal of Analytical Atomic Spectrometry, June 1997, Vol. 12 639for a sample and reducing agent volume of 50 ml. The absorbance is slightly better for a 15 cm coil, decreasing when the reaction time increases, due to sample dilution. In addition, the instability of the BiH3 should be considered. As has previously been reported, bismuth hydride is unstable and thermal decomposition38 might occur, even at ambient temperature. 21 Fujita and Tanaka38 concluded that bismuth hydride is very unstable, decomposing slowly, even at a temperature of 25°C. They included a kinetic factor to be considered, inducing the decomposition of the hydride.Fig. 7 Influence of the height of the separator on the absorbance of Furthermore, they suggested that hydride generation was dierent analyte concentrations. Analytical conditions as in Fig. 5. relatively rapid, but the product hydride was thermally unstable CHCl=1.0 mol l-1. and should be removed from the solution as quickly as possible to produce the maximum absorbance value.This is a condition 0.5 to 5.0 mol l-1; hence, control of the acidity during sample that was used in our approach with the stripping-type gas– preparation is not a critical step for bismuth determination. liquid separator. Thus, considering the sensitivity and repro- However, this is valid only for HCl medium. If there is a ducibility, a reaction coil of 15 cm is recommended for the mixture of HCl with another acid, this becomes a serious determination of bismuth.With respect to injection volume, problem, considering that the absorbance signal decreases with the reproducibility and the frequency decrease when volumes an increase in the concentration of the second acid. Replacing greater than 100 ml are injected. To choose this parameter, HCl with HNO3, H2SO4, H3PO4 or HClO4 leads to a lower however, it is important to consider the eciency of the sensitivity owing to interactions between the analyte and hydride separation from the liquid phase.Injecting dierent NO3-, SO42- or ClO4- or even the formation of less reactive volumes of sample with variable concentration but with the species between H3PO4 and BH4-. same final mass of bismuth, the peak area was calculated. Table 1 shows that volumes between 30 and 100 ml can be injected without significant change in the area. In this instance, Eect of the reactor–separator height the stripping-type gas–liquid separator is ecient for this The proposed stripping-type reactor–separator consists of three system and a volume of 50 ml was used.separate parts. The intermediate part can have dierent heights, which might aect the final result because it increases the Eect of sodium tetrahydroborate concentration distance to the atomizer. Three dierent heights were tested: 4, 6 and 10 cm. An increase in the length leads to a decrease The purpose of the tetrahydroborate is to provide hydrogen in the absorbance signal (Fig. 7), reflecting the dilution of the radicals to react with bismuth in order to generate the hydride. hydride. The best eciency was obtained with the intermediate As can be seen in Fig. 5, the absorbance increases with the part at 4 cm; however, it was also noticed that some water concentration of BH4- up to 2.0% m/v, for all the bismuth vapour condensed in the Teflon tube wall used to connect the concentrations tested. A concentration of 1.0% m/v was separator to the atomizer, causing less reproducibility of the chosen, considering that higher concentrations of NaBH4 signals.This cannot be controlled by the use of drying agents generate an excess of H2 , which leads to low reproducibility such as H2SO4 or CaCl2. Thus, the 6 cm intermediate part due to the dilution of BiH3. was found to be the most appropriate and was used during the remainder of this work. Eect of acid concentration Using the same concentration of HCl for dilution of the Eect of carrier gas flow rate bismuth solution and as its carrier, it was observed (Fig. 6) Flow rates of N2 less than 90 ml min-1 showed lack of that the generation of BiH3 is virtually independent of the reproducibility and large, poorly shaped signals. Above acidity above 0.5 mol l-1, as has previously been reported 130 ml min-1, a higher injection frequency associated with by A° stro�m21 and Yamamoto et al.25 It is important to emphasdilutionof the BiH3 was observed and, consequently, a decrease ize that the presence of the acidic medium is essential to in the signals.For this work, flow rates between 110 and generate the hydride. When the sample is prepared with 120 ml min-1 were chosen. de-ionized water instead of HCl, the absorbance signal can decrease up to five times, showing that the presence of H+ is needed to liberate the H2 that will react with the Bi3+ ion. On Eect of reagent flow rate the other hand, no change in absorbance is observed when the When the merging zones manifold is used the flow rate of both sample carrier is kept constant (1.0 mol l-1) and the concencarrier reagents must be the same.It was observed that up to tration of the acid used to prepare the sample changes from a flow rate of 2.8 ml min-1 the absorbance signal increases Fig. 6 Eect of the HCl concentration on the eciency of hydride generation with dierent analyte concentrations. CNaBH4 =1.0% in Fig. 8 Bismuth absorbance as a function of the atomization tempera- 0.05 mol l-1 KOH; Vinj=50 ml; carrier flow rate (HCl and H2O)= 1.9 ml min-1. ture with dierent analyte concentrations. Conditions as in Fig. 5. 640 Journal of Analytical Atomic Spectrometry, June 1997, Vol. 12Fig. 11 Interferences of some cations on the formation of BiH3 in the Fig. 9 Calibration graph for bismuth using the on-line hydride gener- aqueous phase. CBi=40 ng ml-1. ation system. MnII, MoVI, TiIV, VV and ZnII did not aect the absorbance and above this value the sample and reducing agent do not signal even when they were present in a 5000- or 10000-fold have sucient time to react before reaching the reactor– excess and were not considered as liquid phase interferents.separator. Considering the injection frequency and the sensi- AsIII, AsV and PbII did not interfere in the gaseous phase in a tivity, a flow rate of 1.9 ml min-1 was considered suitable for 10000-fold excess, nor did SbV and SnIV in a 5000-fold excess.bismuth determination. However, a significant reduction in the bismuth signal was caused by small amounts of CoII, CuII, NiII, WVI, HgII, SbIII Eect of the temperature of atomization and AgI. The eect of the interferents in the liquid and gaseous phases is shown in Figs. 11 and 12, respectively. The compe- Once separated, the hydride is quickly transported to the tition for the reducing agent can explain the interference in atomization cell and no critical eects of the changes in the the liquid phase as suggested by Smith.40 In the gaseous phase, atomization temperature were observed.It was verified that the competition for both the tetrahydroborate and the hydro- the atomization temperature is not critical above 850 °C, as gen radicals inside the atomizer39 seems to be the cause of the can be seen in Fig. 8, and 900 °C is suggested as a working absorbance signal reduction. temperature. This value is consistent with those reported In order to overcome these interferences, several reagents by A° stro�m,21 Chan et al.22 and Crock,28 with a similar profile were tested as masking agents.Considering that there is a to the plot of absorbance versus quartz cell temperature. competition for the reducing agent it is advisable to introduce a second reductor species into the system. To maintain the Analytical Performance of the Method simplicity of the FI manifold, the studied reagents were dissolved in 1.0 mol l-1 HCl and used as the sample carrier.The calibration graph with typical analytical signals (Figs. 9 When the sample and the reductor are kept in contact, the and 10) is linear from 0.1 to 100 ng ml-1 of bismuth. The masking agent is already acting over the interferent species, detection limit, calculated as three times the standard devon thus reducing the competition for the tetrahydroborate. As of the blank signal, was 320 pg ml-1, which corresponds to thiourea, L-cysteine and thiosemicarbazide have a disulfide 16 pg of bismuth. The relative standard deviation, for 20 group that can easily be reduced,41 they were investigated as replicate determinations, varies from 10% for 0.1 ng ml-1 Bi masking agents.The results showed that a 0.2% thiourea to 1.9% for 100 ng ml-1 Bi. The injection frequency, under the solution and a mixture of thiourea (0.2%)–L-cysteine (1%) optimized conditions, was 120–150 samples h-1. were ecient in preventing interferences if only one interferent species was present in the sample.Ascorbic acid, citric acid, Interferences hydroxylamine and potassium iodide were also studied but were not suciently ecient to overcome all the interferent According to Dedina,39 the interferences shown by the hydride species. The use of a thiourea (0.2%)–KI (10%) solution generation technique can be classified into two groups: the allowed the recovery of the bismuth signal in the presence of liquid phase interferences, where hydride formation occurs, one or more foreign ions.Table 2 shows the recovery of the and the gaseous phase interferences, occurring either during analytical signal for bismuth spiked with dierent types of hydride transport or in the atomizer. In order to identify interferents in the presence of the thiourea–KI solution. For species that could interfere with the determination of bismuth samples with a high content of tin, the Sn2+ was separated by in the liquid phase, 18 foreign ions were studied.Five species precipitation as metastannic acid, which is formed in an HNO3 (Sn, Sb, As, Se, Pb) which generate hydrides, and are classified medium, and filtered prior to injection. as potential gaseous interferents, were also studied. Arsenic, tin and antimony were investigated in dierent valency states. In addition, mercury was also investigated because it can be Applications reduced to Hg0 when it is in contact with tetrahydroborate.A The accuracy of the method was examined by analysing several species was considered as a potential interferent if the dierence NIST SRMs. The determination of bismuth was also performed between the absorbance for bismuth and that for bismuth in the presence of a particular interferent was higher than 10%. Under these conditions, AlIII, CaII, CdII, CrIII, FeII, FeIII, MgII, Fig. 10 Typical analytical signals showing the reproducibility of the Fig. 12 Influence of some elements occurring in the gaseous phase injections. (The dotted line represents the monitoring of the background.) on the absorbance of the analyte. CBi=40 ng ml-1. Journal of Analytical Atomic Spectrometry, June 1997, Vol. 12 641Table 2 Recovery of the analytical signal of Bi in the presence of interferent species under the masking eect of thiourea (0.2%).KI (10%) Sample Bi5interferent ratio Analytical signal recovery for Bi (%) Bi5NiII5CoII5CuII5AgI 15500055000510005200 94 Bi5SnIV 155000 100 Bi5SbIII 15500 100 151000 90 Bi5SeIV 1510 100 15100 91 Bi5NiII5CoII5CuII5AgI5SeIV5SbIII5SnIV 15100051000510005100525550051000 93 Table 3 Bismuth determination in samples of metallurgical interest 4 Aborn, R.H., Bull. Bismuth Inst., 1975, 7, 1. 5 Headridge, J. B., and Thompson, R., Anal. Chim. Acta, 1978, 102, 33. Amount of Bi (%) 6 Zhou, N., Frech, W., and Lundberg, E., Anal. Chim. Acta, 1983, 153, 23. Sample* This work ETAAS Certified value 7 Marczenko, Z., Spectrophotometric Determination of Elements, SRM 54D 0.044¡¾0.001 . 0.044¡¾0.005 Ellis Horwood, Chichester, 1976. SRM 361 0.00036¡¾0.00001 . 0.0004¢Ó 8 Beinrohr, E., and Hofbauerova¢¥, H.,Mikrochim. Acta, 1989, II, 119. SRM 362 0.0018¡¾0.0001 . 0.002¢Ó 9 Rooney, R. C., Analyst, 1976, 101, 749. SRM 364 0.00096¡¾0.00002 . 0.0009¢Ó 10 Barnett, W. B., and McLaughlim, E. A., Jr., Anal. Chim. Acta, Brass 0.0011¡¾0.0002 0.0011¢Ô 1975, 80, 285. Bronze de Can.o¢¥ n 0.0059¡¾0.0001 0.0067¢Ô 11 Andrews, D.G., and Headridge, J. B., Analyst, 1977, 102, 436. Fluorescent bronze 0.00075¡¾0.00001 ND¡× 12 Headridge, J. B., and Smith, D. R., T alanta, 1972, 19, 833. 13 Drinkwater, J. E., Analyst, 1976, 101, 672. * SRM 54D: Sn (88.5%); Sb (7.04%); Cu (3.62%). SRM 361: 14 Hon, P. K., Lau, O. N., Chung, W. C., and Wong, M. C., Anal. Ni (2.0%); Cr (0.69%); Mn (0.66%); Mo (0.19%). SRM 362: Chim. Acta, 1980, 115, 355. Mn (1.04%); Cu (0.50%); Ni (0.59%); Co (0.30%); Cr (0.30%). 15 Vanloo, B., Dams, J., and Hoste, J., Anal. Chim. Acta, 1983, SRM 364: Mn (0.25%); Cu (0.24%); Ni (0.14%); Mo (0.45%); 151, 391. Co (0.15%); Ti (0.24%). Brass: Cu (50.90%); Zn (20.40%); 16 Welz, B., and Melcher, M., Spectrochim. Acta, PartB, 1981, 36, 439. Pb (0.2%); Sn (0.6%); P (0.0.1%); As (0.0.1%); Sb (0.0.1%); 17 Holak, W., Anal. Chem., 1969, 41, 1712. Ni (<0.25%); Fe (0.1%). Bronze de Can.o¢¥ n: Cu (60.97%); 18 Braman, R. S., Justen, L.L., and Foreback, C. C., Anal. Chem., Sn (10.35%);Zn (<2%). Fluorescent bronze: Cu (#90%);Sn (#10%); 1972, 44, 2195. Zn (<2%); P (0.2.1.5%). 19 Chu, R. C., Barron, G. P., and Baumgarner, P. A. W., Anal. ¢Ó Value for reference only, not certified. Chem., 1972, 44, 1476. ¢Ô Standard deviation within 10%. 20 Ru¡Æz¢§ ic¢§ka, J., and Hansen, E. H., Flow Injection Analysis, Wiley, ¡× ND: Not detected. New York, 2nd edn., 1988. 21 A¡Æ stro¡§m, O., Anal. Chem., 1982, 54, 190. 22 Chan, C. Y., Baig, M. W. A., and Pitts, A. E., Anal. Chim. Acta, in dierent metallurgical samples. The results obtained 1979, 111, 169. (Table 3) were compared with certified values or graphite 23 Chan, W.-F., and Hon, K.-P., Analyst, 1990, 115, 567. furnace measurements. The close agreement between the results 24 Schmidt, F. J., Royer, J. L., and Muir, S. M., Anal. L ett., 1975, obtained by the proposed method and the certified values 8, 123. shows the good accuracy of the method.This was also con- 25 Yamamoto, M., Makoto, Y., and Yamamoto, Y., Anal. Chem., firmed by the analysis of steels, bronze and brass, the results 1985, 57, 1382. 26 Vijan, P. N., and Wood, G. R., At. Absorpt. Newsl., 1974, 13, 33. of which are included in Table 3. 27 Pierce, F. D., Lamoreaux, T. C., and Fraser, K. S., Appl. Spectrosc., 1976, 30, 38. CONCLUSIONS 28 Crock, J. G., Anal. L ett., 1986, 19, 1367. 29 Pacey, G. E., Strata, M. R., and Gord, J. R., Anal. Chem., 1986, Bismuth determination using hydride generation atomic 58, 502. absorption spectrometry coupled with an FI system was found 30 Yamamoto, M., Takada, K., Kumamaru, T., Yasuda, T., and Yokoyama, S., Anal. Chem., 1987, 59, 2446. to be simple, with low reagent consumption, relatively inter- 31 Brindle, I. D., Alarabi, H., Karshman, S., Le, X., Zheng, S., and ference-free and sensitive. This sensitivity combined with signal Chen, H., Analyst, 1992, 117, 407. reproducibility is mainly due to the stripping-type gas.liquid 32 Chen, H., Brindle, I. D., and Zheng, S., Analyst, 1992, 117, 1603. separator because of the low residence volume of the liquid 33 Bergamin Fo, H., Zagatto, E. A. G., Krug, F. J., and Reis, B. F., phase therein, which leads to better separation eciency. Anal. Chim. Acta, 1978, 101, 17. The results obtained for metallurgical samples showed good 34 Thompson, K. C., and Thomerson, D. R., Analyst, 1974, 99, 595. 35 Parisis, N. E., and Heyndrickx, A., Analyst, 1986, 111, 281. accuracy and precision. 36 Gladney, E. S., At. Absorpt. Newsl., 1977, 16, 114. 37 Welz, B., and Melcher, M., Analyst, 1983, 108, 213. The authors thank Carol H. Collins for assistance with the 38 Fujita, K., and Tanaka, T., T alanta, 1986, 33, 203. English in this manuscript. 39 Dedina, J., Anal. Chem., 1982, 54, 2097. 40 Smith, A. E., Analyst, 1975, 100, 300. 41 Boampong, C., Brindle, I. D., Le, X., Pidwerbesky, L., and REFERENCES Ponzoni, C. M. C., Anal. Chem., 1988, 60, 1185. 1 Fleming, H. D., and Ide, R. G., Anal. Chim. Acta, 1976, 83, 67. 2 Yamamoto, M., Yamamoto, Y., and Yamashige, T., Analyst, 1984, Paper 6/06553H 109, 1461. Received September 24, 1996 3 Baba, Y., Hagiwara, M., and Hamada, J., Bull. Bismuth Inst., 1974, 4, 1. Accepted February 18, 1997 642 Journal of Analytical Atomic Spectrometry, June 1997, Vol. 12