Analyst, January 1996, Vol. 121 (31-35) 31 Determination of Ultra-trace Amounts of Selenium(iv) by Flow Injection Hydride Generation Atomic Absorption Spectrometry with On-line Preconcentration by Co-precipitation with Lanthanum Hydroxide Part II. On-line Addition of Co-precipitating Agent Steffen Nielsen, Jens J. Sloth and Elo H. Hansen Chemistry Department A, Technical University of Denmark, Building 207, DK-2800 Lyngby, Denmark A flow injection procedure for the determination of ultra-trace amounts of selenium(rv) is described, which combines hydride generation atomic absorption spectrometry (HGAAS) with on-line preconcentration of the analyte by co-precipitation-dissolution in a filterless knotted Microline reactor. Based on a previously published procedure that requires the off-line premixing of sample and co-precipitating agent, the present approach facilitates on-line addition of the co-precipitant to the time-based aspirated sample.The sample and the coprecipitating agent (lanthanum nitrate) are mixed on-line and merged with an ammonium buffer solution of pH 9.1, which promotes precipitation and quantitative collection on the inner walls of an incorporated knotted Microline reactor. The SeIV preconcentrated by coprecipitation with the generated lanthanum hydroxide precipitate is subsequently eluted with hydrochloric acid, allowing an ensuing determination via hydride generation. At different sample flow rates, i.e., 4.8, 6.4 and 8.8 ml min-1, enrichment factors of 30, 40 and 46, respectively, were obtained at a sampling frequency of 33 samples h-l.The detection limit (3s) was 0.005 pg 1-1 at a sample flow rate of 6.4 ml min-1 and the precision (relative standard deviation) was 0.5% (n = 11) at the 0.1 pg 1-1 level. Keywords: On-line co-precipitation-preconcentration; co-precipitation with lanthanum hydroxide; on-line addition of the co-precipitant; flow injection hydride generation atomic absorption spectrometry; selenium(rv) assay Introduction Flow injection (FI) on-line pre-concentration procedures by collection and dissolution of precipitates were first described by Jimenez et al. in 1987 for the indirect determination of anions.',* In these experiments a stainless-steel filter was used for the collection of the precipitates. Based on this conceptual idea, the authors proceeded with the pre-concentration of organic constituents,36 and subsequently with the preconcen- tration of trace elements,7-10 primarily employing flame atomic absorption spectrometry (AAS) as the means of detection.However, the use of an incorporated filtering sytem may not only limit the efficiency of the on-line collection of the precipitate formed, but it might also, to some extent, influence the ensuing on-line dissolution process, because the presence of the filtering device can give rise to the generation of considerable back-pressure. As a consequence, the sample loading is reduced which, in turn, affects the enrichment factor (EF)" of the system. Besides, the filtering device often detracts from the reproducibility of the determination. To achieve an effective on-line preconcentration an approach is therefore needed that promotes effective collection of the precipitate formed, allows its ensuing instantaneous dissolution and obviates the problems associated with generation of unwanted back-pressure.In 1991 Fang et a1.12 introduced a simple and most elegant solution to this problem, demonstrating that the collection of precipitate could effectively be achieved on the inner walls of a tube provided that it was tightly knotted. This device, the so-called knotted reactor (KR), which thus eliminates the need for the use of a filter, has proved to be very effective, providing optimum conditions for on-line precipita- tion-dissolution manipulations. 1 1 The applicability of the approach has been demonstrated in a series of papers by Fang and co-workers,12-14 Welz et al.15 and Min and Hansen16 for the determination of low levels of various metal ionic species by means of co-precipitation in an FI-AAS system.In all instances, the on-line collection was effected by means of an organic co-precipitating agent, which in turn required the use of an organic solvent for the ensuing dissolution procedure. For obvious reasons, it would be preferable if an inorganic solvent, such as a mineral acid, could be employed, because many organic solvents exhibit toxic or unpleasant properties. Although one can take advantage of the fact that the FI system is a closed one, the necessity of using organic solvents entails additional problems. Thus, these chemicals cannot be handled by most pump tubes and therefore require the use of displacement facilities, which makes the system more complicated.A system based on employing co-precipitation and ensuing dissolution in an inorganic solvent was recently described by Tao and Hansen.17 Used for the determination of SeIV, the authors employed an FI manifold combining hydride generation and atomic absorption spectrometry (HGAAS) with time based sampling, where on-line preconcentration of Se"' was effected by co-precipitation with La1'' at a pH around 9.1. The dissolution procedure was made by hydrochloric acid (Fig. I), which is the medium for the hydride generation process. In the procedure, the co-precipitant was premixed off-line with the sample, which, of course, is a severe practical limitation of the approach, and it would therefore be of interest to explore the possibilities of accomplishing this procedure on-line.This is not only because the sample manipulations would be significantly reduced in the actual assay, but also, and most importantly,32 Analyst, January 1996, Vol. 121 because such an investigation would reveal if this approach had general applicability. The present methodological study is thus devoted to these matters, using as its basis the approach of Tao and Hansenl7 (Fig. 1) but extending it to the determination of trace-level amounts of Sew via the on-line addition of co-precipitating agent. Because Tao and Hansen demonstrated that the La co- precipitation procedure as such was successful for practical assays of a number of 'real samples', the present procedure was for the same reason confined to aqueous standards, and hence it was found redundant to apply it to other types of sample.However, as it turned out, and as one intuitively might expect by considering the dynamic conditions under which the precipitat- ing reaction takes place, this investigation proved to be anything but a trivial exercise. Experimental Apparatus A Perkin-Elmer (Norwalk, CT, USA) Model 2100 atomic absorption spectrometer was used in combination with a Perkin- Elmer Model FIAS-200 flow injection unit (equipped with two individually controlled peristaltic pumps and a five-port FI- valve), with a hydride generation accessory (the gas-liquid separator used in the chemifold was a Perkin-Elmer W- configuration unit). A selenium hollow cathode lamp (S.& J. Juniper, Harlow, Essex, UK) was used at a wavelength of 196.1 nm with a spectral bandpass of 2.0 nm, and was operated at 7 mA. The temperature of the quartz atomizer cell was set at 9OO0C, and the argon carrier flow rate was fixed at 100 ml min-l. The output signals were processed with a time constant of 0.5 s in the peak-height mode and recordings from the graphics screen were printed out by an Epson Model FX-850 printer. The acutation times of the injector valve and the two pumps were programmed with the use of the FI software of the Model 2 100 atomic absorption spectrometer. The filterless knotted reactor precipitate collectors were made from 0.5 mm id, 1.8 mm od Microline tubing (cross-linked ethyl vinyl acetate) by tying interlaced knots (the optimum length of the knotted reactor, LKR, was 100 cm).The knots were made with approximately 5 mm diameter loops. All the other reaction coils, connections and conduits in the FI-manifold (Fig. 2) Fig. 1 Schematic diagram of the on-line co-precipitation-dissolution HGAAS system with off-line addition of precipitating agent (La'", added to individual samples), shown in (a) the loading (precipitating) sequence and (b) the elution stage. QTA, quartz cell; Ar, argon; RC, reaction coil; SP, gas- liquid separator; P1 and P2, peristaltic pumps; V, valve; W, waste; and KR, filterless knotted reactor (Redrawn from ref. 17, with permission from the Royal Society of Chemistry). Fig. 2 Schematic diagram of the FI-HGAAS system for the on-line co- precipitation-dissolution HGAAS system with on-line addition of La"'.MC, mixing coil (6 cm); other symbols as in Fig. 1 . The pumping rate of the sample solution (X) was investigated at three different levels (i.e., 4.8, 6.4 and 8.8 ml min-1, respectively; for details, see text).Analyst, January 1996, Vol. 121 33 consisted of 0.5 mm id PTFE (polytetrafluoroethylene) tub- ing. Reagents and Standard Solutions All the reagents were of analytical-reagent grade, and distilled water was used throughout. Sodium tetrahydroborate solution [0.3% (m/v) in 0.05 mol 1-1 sodium hydroxide solution] was prepared freshly daily. Lanthanum nitrate hexahydrate solution, 0.5% m/v, was made by dissolving 0.6662 g of lanthanum nitrate hexahydrate in 100 ml of distilled water. The buffer solutions, freshly prepared every day, were in all instances 0.2 moll-' ammonium chloride adjusted to the appropriate pH buffer (9.1-9.2) by addition of 0.2 mol 1-1 ammonia solution (the optimum pH value depended on the sample flow rate; see Results and Discussion).Standard solutions of selenium(1v) for calibration purposes were prepared by three-stage aqueous dilutions of a 1000 mg 1-1 stock solution, which was made by dissolving 1.4053 g of selenium dioxide in 1000 ml of 1.0 mol 1-1 hydrochloric acid. All glassware was soaked for at least 24 h in 1 moll-' nitric acid, and finally rinsed in distilled water before use. Operational Procedure The time-based FI-HGAAS system with on-line addition of the co-precipitant, Lar1*, is shown in Fig. 2, depicting the optimized experimental parameters.In the precipitation procedure (Fig. 2a), the sample, hydrochloric acid and sodium tetrahydroborate solutions were introduced by pump 1, and the buffer and lanthanum nitrate solutions were delivered via pump 2. During the precipitation sequence both pumps were activated for a period of 99 s. The sample and La'r1 were pre-mixed in the mixing coil (MC) of length (LMC) 6 cm, and subsequently merged with the buffer solution at the entrance to the knotted reactor (KR) which had a length (LKR) of 100 cm. The precipitate, which was formed instantaneously after the merging point of KR, was collected on the inner walls of the knotted reactor. The effluent emerging from the reactor was discarded. Simultaneously, the acid was pumped through the by-pass of the rotor of the valve and directed into the hydride generation system.During this stage, the baseline for the final readout was established. At the end of the precipitation period pump 2 was stopped, and the valve was actuated automatically from the fill-mode to the inject-mode for a period of 10 s (dissolution-hydride- generation procedure; Fig. 2b), by which means the acid was introduced to the knotted reactor where the precipitate adhering to the inner walls of the knotted reactor was dissolved. This concentrated zone was directed from the knotted reactor to the hydride generating system, where the analyte was merged with a reducing solution of sodium tetrahydroborate. After passing through a reaction coil (RC) (LRC = 35 cm), the gas-liquid mixture was guided into the gas-liquid separator (SP) in which the hydrogen selenide and the evolved hydrogen were separated and swept into the atomizer cell by a steady argon carrier flow.The absorption signal was then recorded. The waste from the gas-liquid separator, which included unreacted sodium tetrahydroborate, acid and some argon, was removed by aspiration. After 10 s in the inject mode, the valve was returned to the fill mode, allowing the sample solution to be interchanged so that the remainder of the previous sample in the sample pump tube could be effectively washed out and the next sample kept ready for precipitation. Results and Discussion Preliminary Investigations In their experiments with off-line addition of co-precipitant, Tao and Hansen17 anived at the optimized operational par- ameters shown in Fig.1, pointing out that the pH of the precipitation reaction was very critical. When the present authors tried to reproduce the assay, difficulties were encoun- tered in achieving comparable characteristics, i.e., the peak heights obtained were somewhat lower (about 25%) than those previously reported. While this might possibly be due to the 'knotting efficiency' of the incorporated reactor in the two systems, it was nevertheless decided to make a closer scrutiny of the flow rate of the ammonia buffer used. It was thus found that with all other parameters fixed as indicated in Fig. 1 (the concentration of La added to each sample being 20 mg 1-l, and the pH of all sample solutions adjusted to 3), progressively lower flow rates of the buffer (pH 9.1) yielded increasingly higher signals, the optimum flow rate (QBuffer) being 0.5 ml min-1, that is, one third of the flow rate used by Tao and Hansen. With these conditions, comparable results were obtained, and therefore the low buffer pumping rate was used in the following on-line approach.Optimization of the On-line Co-precipitation-Dissolution Preconcentration System with On-line Addition of the Co-precipitant The FI-manifold used for the on-line addition of precipitation agent is shown in Fig. 2. As a first approximation, the optimization procedure might be effected by reproducing the favourable kinetic conditions prevailing in the off-line system of Tao and Hansen, that is, by ensuring that the presentation of the samples to the co-precipitation reactor and that the reaction parameters within it (ix., concentrations of constituents and the pH of the buffer) are identical in the two instances.To reproduce the sample presentation of the off-line system, the aqueous sample solutions were therefore pre-mixed with an Larr1 stream, the pH of which was adjusted to 3 by means of the addition of hydrochloric acid. Thus, with the individual pumping rates used and with due consideration to the dilutions of the merging streams, the concentration of the La stream yielding the optimum concentration, as found for the off-line system (20 mg 1-I), might readily be calculated. However, with different lengths of the added mixing coil MC (varied between 50 and 200 cm), it was only possible to reach about 35% of the results that Tao and Hansen achieved with off-line addition of Lar1'.This proved that the conversion from the off-line to the on- line addition of co-precipitating agent was no trivial task, very likely because the precipitation reaction takes place under dynamic conditions, where it is very critical how the precipitate is formed in order to be entrapped in the knotted reactor. Thus, it is essential that the hydrophilic precipitate consist of small, curdy particles which willingly adhere to the hydrophilic Microline tube and that the formation of larger particles, which might be flushed through the reactor, is prevented. A feasible avenue to improve the performance of the system, and possibly reach higher enrichment factors (EF, the ratio between calibration curves with and without preconcentration), would therefore be to increase the delivery of sample solution during each cycle.This could be carried out either by prolonging the time for the aspiration of sample, which would have a negative effect on the sampling frequency, or by increasing the flow rate of the sample solution. It should be noted that in the latter instance, which would be operationally preferable, it is a condition that quantitative entrapment of precipitate is still achieved, and therefore a conscientious optimization procedure is called for. However, with increasing accumulated precipitate in the knotted reactor, the back pressure of the system (as expected) tended to increase, and in order to avoid troubles due to potential disruptions of tube connections Tao and Hansen therefore limited the sample flow rate (Qs) to 4.0 ml min-1.In the present investigation, where higher pumping rates were also34 Analyst, January 1996, Vol. 121 attempted, care was taken to fasten the acid and sample tubes securely with metal wire tighteners, which therefore permitted us to overcome possibly higher back-pressures and hence to increase the sample flow rates, which in turn allowed this parameter to be altered in order to achieve higher sensitiv- ities. The sample flow rate was tested at levels of Qs = 4.8,6.4 and 8.8 ml min-l, respectively, with the length of the knotted reactor ( L K ~ ) affixed at 100 cm. For each sample flow rate used, the FI system was optimized accordingly, that is, the parameters which have any influence on the co-precipitation ([LaTrr], pHBuffer, QL~III and QBuffer) were investigated, while the experimental parameters for the hydride generation procedure were identical with those described previously17 (Fig.1 b). As it turned out, the experimental parameters for the co-precipitation procedure were rather similar to those reported by Tao and Hansen, except that it was necessary to increase the pH of the buffer slightly as the sample flow rate was increased. The optimized parameters are depicted in Fig. 2 and the results obtained are shown in Table 1. While it should be expected that the limit of determination is improved with increasing pumping rate of the sample, it is significant that the sensitivity of measurement increased with increasing sample consumption.Thus, at Qs = 8.8 ml min-l, the sensitivity was 0.776 specific absorbance (pg l-l)-l, or almost twice that achieved with the off-line addition. During the optimization procedure of the co-precipitation reaction it was found that it was unnecessary to add acid to the lanthanum nitrate solution. Therefore, for a given sample flow rate, optimum signals could be achieved simply by optimizing the pH of the buffer solution (see Table 1). Thus, by ensuring that it had sufficient buffering capacity (0.2 mol 1-l), identical 0.4 0.39 0.38 0.37 0.36 0.35 0.34 0.32 0.33 ~ 170 180 190 200 210 Po 230 240 2% La"' (pprn) Fig. 3 Effect of the lanthanum nitrate concentration on the peak absorbance of a 0.5 pg 1-1 Se'" standard. The pumping rate (X) of the sample solution was 8.8 ml min-1 (pH 9.17); all other experimental parameters as in Fig.2. signals were achieved for the aspirated sample solutions even if the pH of these solutions was varied within a pH range of 3-7. Therefore, hydrochloric acid was not added to the lanthanum nitrate solutions in further work. For a sample flow rate of Qs = 8.8 ml min-l the optimization of the LarrT concentration is shown in Fig. 3. For all the sample flow rates used the optimum concentration of La(NO& was found to be 210 ppm. This is not surprising because the premixing of the sample and the Larr1 solution in the very short premixing coil, MC (LMc = 6 cm), is minimal. Rather, the mixing of the sample and the lanthanum nitrate solution was optimized by optimizing the length of the knotted reactor, L m , which was established to be 100 cm.In the preconcentration stage, the main objective is to ensure that the precipitation is quantitative or that as much as possible of the precipitate formed is entrapped in the knotted reactor. Therefore, it should be expected that it would be preferable to utilize relatively small flow rates of the buffer and the lanthanum nitrate solutions to minimize the total flow rate through KR and thereby avoid unnecessary disturbance of the co-precipitation process. This was, in fact, confirmed in this investigation. Furthermore, it was found, which in retrospect is not surprising, that the optimized flow rates of the buffer and lanthanum nitrate solutions were constant at all sample flow rates used, and that the optimum pH of the buffer increased with increasing sample flow rates.However, as was shown experi- mentally, the optimized pH value of the buffer is a very critical parameter for each individual sample flow rate; this is noted in Table 1. Indeed, if the pH was varied between 9.1 and 9.2 at the various flow rates tested, variations up to 2630% in the recorded signals were actually obtained. Yet, at fixed values, which are readily maintained by ensuring a sufficient buffering capacity of the buffer, the system is very robust. 0.8 0.7 S 0 0.6 0.5 8 0.4 0.3 h .- c - v 0 5 0.2 0.1 0 1 2 3 4 5 6 7 8 9 Flow rate/rnl rnin-' Fig. 4 Graphical representation of the slopes of the individual calibration curves as a function of the flow rates of the sample solution.Experimental parameters as in Fig. 2 and as shown in Table 1. Table 1 Characteristics for the FI on-line co-precipitation-preconcentration HGAAS system with on-line addition of Lallr at different sample flow rates Sample flow-rate/ml min-1 pH of buffer Concentration of La(N03)3/mg 1-1 Calibration range/pg 1-1 Regression equation in calibration range (6 standards, n = 3, CS, in yg 1-I) Sample volume per assaylml (loading time, 99 s) Sample frequency cf) (samples h-1) Relative standard deviation Limit of detection (3 s)/pg I-' Enrichment factor (EF) Concentration efficiency (n = 11; 0.1 pg 1-1) (%) (CE = EF fl60) 4.8 9.13 0.01-0.30 210 0.510 Cs, + 0.005 (Y = 1.000) 7.9 33 0.5 0.006 30 16.5 6.4 9.15 0.01-0.30 (r = 0.999) 210 0.678 Cs, + 0.003 10.6 33 0.5 0.005 40 22.0 8.8 9.17 0.005-0.30 ( r = 0.999) 210 0.776 Cs, + 0.009 14.5 33 0.8 0.004 46 25.3Analyst, January 1996, Vol.121 35 By depicting the slope of each calibration curve versus the sample flow rate (Fig. 4), it is feasible to obtain useful information as to the capacity of the knotted reactor of LKR = 100 cm. Thus, according to this figure, there is a tendency to a linear relationship until a sample flow rate of 6.4 ml min-1 is reached, which strongly indicates quantitative co-precipitation. Above this level, the curve tends to bend off, indicating a lack of sufficient capacity of the knotted reactor at these sample flow rates (and the co-precipitation period of 99 s). Very likely this ‘breakthrough’ of the capacity of the reactor can be explained by an insufficient adherence of the precipitate on the inner walls of the knotted reactor because of a too high flow through the knotted reactor at high sample flow rates.This explanation was, in fact, verified by halving the sample flow rate and doubling the time for the co-precipitation while maintaining all other experimental parameters as in Fig. 2. Conclusion The conceptual idea of Tao and Hansenl7 and the work presented in this paper demonstrate that it is possible to arrange an effective on-line FI procedure for preconcentration of trace levels of elements by co-precipitation-dissolution in inorganic media. The performance of the previously described system has been improved from off-line addition to on-line addition of the co- precipitant to the sample. The optimization by on-line addition yielded very satisfactory results, notably allowing higher sensitivity of measurement and yielding better enrichment factors (EF) and improved concentration efficiencies (CF = EF f/60, see Table l), the former being twice as high as previously obtained.17 These advantages primarily resulted from the on- line system allowing the consumption of higher sample volumes at higher sample flow rates, which, in turn, was enabled by mechanical improvements to the FI system, notably the use of metal wire tighteners to fasten the sample and acid pump tubes.With the principles of this FI-HGAAS system for precon- centration by on-line co-precipitationdissolution with on-line addition of the co-precipitant, it could be of interest to include other hydride forming elements.Further, it would be interesting to test the possibilities for developing the FI system by incorporating on-line speciation of the hydride forming ele- ments. Work in this direction is currently being conducted at this laboratory. The authors wish to extend their appreciation to Julie Damm’s Foundation (Denmark) for partial financial assistance of this research programme. References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 Martinez-JimCnez, P., Gallego, M., and Valcircel, M., J . Anal. At. Spectrom., 1987, 2, 21 1. Martinez-JimCnez, P., Gallego, M., and Valcircel, M., Anal. Chem., 1987,59,69. Montero, R., Gallego, M., and Valcircel, M., J . Anal. At. Spectrom., 1988,3,725. Montero, R., Gallego, M., and Valcircel, M., Anal. Chim. Acta, 1988, 215, 241. Martinez Calatayud, J., and Garcia Mateo, J. V., J . Pharm. Biochem. Anal., 1989, 7 , 1441. Du, K. P., Wang, Y. Z., and Fang, Z. L., Shenyang Yaoxueyuan Xuebao, 1992,9, 130. Martinez-JimCnez, P., Gallego, M., and Valcircel, M., Analyst, 1987, 112, 1233. Santelli, R. E., Gallego, M., and Valcarcel, M., Anal. Chem., 1989, 61, 1427. Santelli, R. E., Gallego, M., and Valcircel, M., J . Anal. At. Spectrom., 1989, 4, 547. Esmadi, F., Kharoaf, M., and Attiyat, A. S . , Microchem. J., 1989,39, 71. Fang, Z. L., Flow Injection Separation and Preconcentration, VCH, Weinheim, Germany, 1993. Fang, Z. L., Sperling, M., and Welz, B., J . Anal. At. Spectrom., 1991, 6, 301. Fang, Z. L., and Dong, L., J . Anal. At. Spectrom., 1992, 7 , 439. Chen, H., Xu, S., and Fang, Z. L., Anal. Chim. Acta, 1994, 298, 167. Welz, B., Xu, S., and Sperling, M., Appl. Spectrosc., 1991, 45, 1433. Min, R. W., and Hansen, E. H., Chem. Anal. (Warsaw), 1995, 40, 243. Tao, G. H., and Hansen, E. H., Analyst, 1994, 119, 333. Paper 51031 99K Received May 19,1995 Accepted September 7,1995