JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1991 VOL. 6 615 Flow Injection On-line Separation and Preconcentration for Electrothermal Atomic Absorption Spectrometry Part 2.* Determination of Ultra-trace Amounts of Cobalt in Water Michael Sperling Xuefeng Yint and Bernhard Welzt Department of Applied Research Bodenseewerk Perkin-Elmer GmbH 0-7770 Uberlingen Germany Ultra-trace amounts of cobalt in sea-water were determined by flow injection on-line sorbent extraction-precon- centration coupled with electrothermal atomic absorption spectrometry. With sodium diethyldithiocarbamate as the complexing agent and C,,-bonded silica reversed-phase sorbent as the column material a 42-fold enhancement of peak area compared with that for direct 40 PI sample introduction was obtained with a 120 s preconcentration and 5.6 ml of sample.The atomization signal was directly proportional to the preconcentration time. An enhancement factor of 210 and a detection limit of 1.7 ng I-’ could be achieved with a 10 min sample loading time (28 ml). Hence the extremely low cobalt concentration of 4.0 ng I-’ in National Research Council of Canada certified reference material NASS-1 Open Ocean Seawater could be determined within 12 min. Results obtained for sea-water and estuarine water certified reference materials agreed with the certified values with a relative standard deviation of 5.0%. Keywords Sea-water; ultra-trace cobalt determination; on-line solid-sorbent extraction; electrothermal atomic absorption spectrometry The determination of cobalt in sea-water is complicated by various factors.The most important of these are the very low concentrations of analyte element and the high total salt content of the matrix. Many methods such as coprecipita- tion co-crystallization solvent extraction column extrac- tion and electrolysis have been developed for the precon- centration of trace metals in sea-water. Liquid-liquid extraction of chelated metal ions is one of the most widely used techniques for the preconcentration of trace metals from water. A variety of chelating agents have been used such as ammonium pyrrolidin-1-yldithio- formate [pyrrolidine dithiocarbamate (APDC) ],2-5 diethyl- ammonium diethyldithiocarbamate (DDC)6 or different mixtures of these reagent~.~~’-l Because large volume extractions are impractical both from a theoretical and a physical point of view an upper limit is placed on the sample preconcentration factor that can be achieved with single-stage separations.A 30- to SO-fold preconcentration of trace metals from a 1 1 sample of sea-water could not be attained by simple extraction into an organic solvent such as 4-methylpentan- 2-one (isobutyl methyl ketone IBMK) not only because of the unfavourable magnitude of the distribution coefficients of the chelates,12 but primarily because the solubility of IBMK in sea-water necessitated the use of at least 2 ml of organic phase for 100 ml of aqueous phase.4 ‘Clean’ separations were very difficult because a thin band of emulsion formed during phase separation between the two liquid phases. This phenomenon resulted in a variation in the volume of the IBMK extract through the carryover of water present in the em~lsion.~ Another disadvantage of liquid-liquid extraction was the instability of many metal carbamates in organic solu- tion,2*10J1J3 which limited the time available for analysis after extraction.Also the analytical response of organo- metallic compounds was often different from that of in- organic salts making the determination difficult unless organometallic standards and the same solvent were ~sed.41415 *For Part 1 see ref. 43. ?On leave from Shandong Provincial Institute of Environmental Protection Shandong Provincial Environmental Monitoring Centre Jinan China. $To whom correspondence should be addressed. An additional problem with electrothermal atomic ab- sorption spectrometry (ETAAS) was the low surface tension of IBMK making sample delivery difficult. The IBMK tended to creep along the length of the furnace tube limiting the sample volume sever el^.^ Storage of the extracts in open cups on the autosampler was problematic because the analyte concentration increased during analysis owing to solvent evaporation.These and many other problems with the analysis of organic extracts by ETAAS16 were avoided when the metals were back-extracted into an aqueous solution. Unfortu- nately back-extraction was often slow and inefficient for metals such as cobalt copper and iron.” C~precipitation~*l~-~~ has the advantage that most ele- ments except for the alkali and alkaline-earth metals are efficiently concentrated and the procedures are generally simple.However it has disadvantages in that the coprecipi- tation reagent introduces a new matrix and that impurities from the large excess of coprecipitation reagent and from the alkali used for pH adjustment are a serious source of contamination. Very often the precipitate is difficult to filter calling for lengthy ageing procedures20 or centrifuga- tion which is cumbersome when large volumes are in- volved. Some of the problems can be overcome by flotation techniques. 19921+23 Preconcentration by coprecipitation for determination by ETAAS is hampered additionally by the large excess of precipitant which can cause severe matrix effects and background absorption.1° Because of these problems column extraction procedures have attracted increasing attention.These techniques have two potential advantages over liquid-liquid extraction procedures (i) a relatively high enrichment factor and (ii) the ability to treat large sample volumes in a closed system hence reducing the risk of contamination. The column techniques applied to the preconcentration of cobalt from sea-water can be classified according to the types of solid- phase materials used (i) ligand-immobilized ma- teria1;3-5*7-9*24-36 (ii) ligand-impregnated ~orbent;~’J~ and (iii) reversed-phase s ~ r b e n t . ~ ~ . ~ ~ In the first two examples metals are usually eluted from the column by dissociating them from the complexing agent by using mineral acids. In this process most metals are easily released except for a few such as cobalt and chromium.It is known that cobalt(I1) complexes are easily616 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL. 6 oxidized by dissolved oxygen or by their own ligands in the aqueous and organic phases and that the resulting trivalent complexes are too inert to be easily dissociated even by strong acids. Low recoveries of cobalt reported by several worker^^^^^ were caused by incomplete elution of such inert cobalt species and by incomplete complexation.30~40 Chelex- 100 has been widely used as sorbent material for preconcentration of trace metals from sea- water,4*5*7~9~2a*27*30*31-35~36 because it is a strong chelator and removes metal ions from most naturally occurring chelates in sea-water. However the resin does not remove metals held in organic and inorganic colloids which could be present even after ultrafiltration leading to low recoveries from natural water^.^ While a high column capacity results in an improved tolerance to chemical interferences it can also lead to incomplete separation from the matrix.Alkali and alkaline-earth ions occupy the resin sites not occupied by transition metals and are co-eluted by acids. The salts formed frequently impair instrumental techniques such as ETAAS,34*36-41 inductively coupled plasma mass spectro- metry (ICP-MS)24+2* and ICP atomic emission spectrometry (AES).26 Column capacity must be optimized in order to balance the tolerance capacity for interferences in the sample-loading and elution and determination stages. For sea-water calcium and magnesium had to be removed from the resin before elution of the trace metals by well- contrived washing p r ~ c e d u r e s ~ ~ ~ ~ ~ J ~ which were essential for the success of the method.The process of preconcentra- tion was slow because flow rates of less than 2 ml min-' had to be used in order to avoid incomplete retention.30 Elution from the resin was also slow with severe tailing such that a minimum of 30 ml of acid were required for complete e l ~ t i o n . ~ Consequently only low concentration factors could be achieved unless very large sample volumes were used. In addition to this problem swelling and contraction of the resin created difficulties in maintaining column flow rates necessitating frequent operator intera~tion.~Jl Another widely used column material is immobilized 8- An efficient preconcentra- tion and separation method for a number of metals utilizing silica-immobilized 8-hydroxyquinoline and an HN03-HCl mixture to elute the trace metals from the column prior to their determination by ETAAS was described by Sturgeon et ~ 1 1 .~ ~ and Willie et aZ.33 However this procedure required a relatively large acid volume of 10 ml in order to elute the sequestered metals from the column. Consequently a large sample volume (500-900 ml) was required for the analysis of uncontaminated sea- water in order to obtain the necessary sensitivity and it was difficult to attain enrichment factors greater than 100 with less than 1 1 of sample. For cobalt quantitative recovery could not be obtained. When metal-ligand complexes are formed in the aqueous phase and collected with a sorbent-packed ~ o 1 ~ m n ~ ~ ~ ~ ~ organic solvents can be used for elution overcoming the elution problems mentioned above and allowing for the selection of the functional group from a wider range of reagents.A serious drawback of these flow-through methods based on solid-sorbent extraction was the loss of analyte element by precipitation and adsorption of the complexes on to the container walls.40 All such methods invariably increase sample manipula- tion and the relatively large amounts of reagents and the container surfaces coming into contact with the sample often give rise to unacceptably high and/or random proce- dural blanks. This requires extensive purification of re- agents rigorous cleaning of laboratory ware and sample preparation under clean-room condition^.^^ The sample volume available for analysis is often limited e.g.owing to the high costs of water sampling. However the need for large enrichment factors persists requiring the develop- ment of preconcentration techniques that demand low sample volumes and yet provide high enrichment factors with minimal contamination. In only four papers7~25+28~40 are detection limits reported that are sufficiently low for the determination of cobalt in unpolluted open ocean sea-water by atomic spectrometry. In one of these papers,28 ICP-MS is used as the instrumen- tal system for the determination whereas the others used ETAAS.7*25*40 Most of the procedures start with 500-1000 ml volumes of sea-water in order to achieve the large enrichment factors necessary for the determination of this element.In addition all the methods used are off-line and require a number of procedural steps which makes them lengthy and requiring a class 10 (or at least class 100) clean- air working environment. This means that none of the procedures is straightforward and routinely applicable in a normal laboratory environment. Interest in flow injection (FI) AAS has increased not only for sample introduction but also as a technique for sample pre-treatment such as analyte preconcentration and separa- tion from the bulk of the matrix. On-line coupling of a FI-preconcentration system with ETAAS42,43 offers solu- tions to most of the problems discussed above. The closed system allows for processing of high sample volumes with a minimum of wetted surface hence reducing the risk of contamination and analyte loss by sorption on container walls.The integrated system permits fully automated operation avoiding time-consuming manual work which also enhances reproducibility and precision. The FI system can work in parallel with the graphite furnace such that the graphite furnace cycle time can be used for sample processing consequently the over-all processing time is not increased. The combination of liquid-liquid extraction principles with sorption on a solid-phase column permits an extensive selection of functional groups to be used. The possibility of choosing a functional group for chelating and a suitable solid-phase material for sorption provides the high selectivity required in order to separate ultra-trace amounts of transition metal ions from the bulk of the alkali and alkaline-earth elements in the sea-water matrix.The very high sample-to-eluate volume ratio provided by on-line sorbent extraction together with the high efficiency of sample introduction offers the potential of sufficiently high preconcentration factors to enable the determination of elements in the low ng 1-1 range. This work describes investigations into whether the combination of NaDDC (sodium diethyldithiocarbamate) as the chelating agent and c18 reversed-phase material as the sorbent which has been applied successfully to the determination of a number of elements in sea-water and related samples,43 could also be used for the precise determination of ultra-trace amounts of cobalt in sea-water.The aim was to reduce sample consumption increase the speed of analysis and improve the reliability of the determination. Experimental A Perkin-Elmer Zeemad3030 atomic absorption spectro- meter with an HGA-BOO graphite furnace and AS-60 autosampler equipped with a cobalt hollow cathode lamp operated at 35 mA was used throughout this work. The wavelength was set to 240.7 nm with a spectral slit-width of 0.2 nm. Pyrolytic graphite coated electrographite tubes with pyrolytic graphite platforms were used exclusively; the graphite furnace temperature programme is shown in Table 1. The atomization signals were recorded by high-resolution graphics and printed out with a Perkin-Elmer PR-100 printer. Integrated absorbance was used exclusively for determinations.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1991 VOL.6 617 Table 1 Graphite furnace temperature programme for the deter- mination of cobalt in sorbent-extraction eluates with use of pyrolytic graphite coated electrographite tubes with a pyrolytic graphite platform Time/s Temperature/ Argon flow Step "C Ramp Hold rate/ml min-' 1 90 5 30 300 2 1000 5 20 300 3 2550 0 5 0 (Read) 4 2650 1 5 300 A Perkin-Elmer Model FIAS-200 FI accessory for AAS was used for the preconcentration of cobalt from standards and samples. The rotation speeds of the pumps and the (a) Ethanol HNO DDC Sample W ( C) v C Ethanol -1 W n P2 HNO Sample PC DDC - W ( e) P1 sequence of operation of the pumps and valve were programmed by an Epson PC+ computer working inde- pendently from the spectrometer.Tygon pump tubes were used for all aqueous solutions and solvent-resistant Verdo- prene pump tubes for ethanol. The manufacturing proce- dure for the conical sorbent extraction-preconcentration microcolumns with 15 pl of sorbent material has been described elsewhere.42 Small-bore (0.35 mm id.) poly(tet- rafluoroethylene) tubing was used for all connections. The FI manifold and the sequence of its operation are shown in Fig. l(a-e). The duration and function of each sequence are summarized in Table 2. A complete cycle of preconcentration and eluate introduction into the graphite furnace consisting of seven stages typically took 207 s with a sample loading period of 120 s (corresponding to 5.6 ml of sample).( b ) Ethanol DDC Sample HN03 & U P1 Ethanol .t W W DDC Sample P1 . W GF v 1 Fig. 1 Flow injection manifold and sequence of operation for sorbent extraction preconcentration ETAAS. For details see text. PI and P2 peristaltic pumps; C conical column ( 15 pl packed with RP-C18 sorbent); PC precolumn (500 pl packed with RP-C18 sorbent); W waste; and GF graphite furnace. (a) Sample loading sequence. (b) Column rinsing sequence. (c) Pre-elution to waste. (dj Analyte elution into graphite furnace. (e) Column cleaning. The two additional sequences (sequences 4 and 6 in Table 2) in which the autosampler arm is moved into and out of the graphite tube respectively are not shown here618 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1991 VOL. 6 Table 2 Sequence of operation of the on-line sorbent extraction-preconcentration system for the detemination of cobalt by ETAAS Time/ Pump I/ Pump 2/ Pumped Stage of Sequence S ml min-' ml min-' Injector medium operation 1 [Fig. I(@] 120 2.8 0 Fill Sample 1.5 NaDDC 2 [Fig.I(@] 25 2.8 0 Inject 0.5% HNO 3 [Fig. l(c)] 8* 0 1.5* Inject Ethanol 4 10 0 0 Inject - 6 10 0 0 Inject - 5 [Fig. l ( 4 ] 4 0 0.7 Inject Ethanol 7 [Fig. l(e)] 30 0 0.8 Inject Ethanol Total 207 s *To be optimized for maximum sensitivity. Load sample Wash column Elute into capillary Capillary into furnace Inject sample Tubing into waste port Elute to waste A sample loading period of 10 min corresponding to 28 ml of sample was used for the determination of the lowest cobalt concentration in the National Research Council of Canada (NRCC) certified reference material (CRM) NASS- 1 Open Ocean Seawater.The principle of eluate zone sampling whereby a 40 pl portion of the eluate containing the most concentrated fraction of the eluted analyte is introduced directly into the graphite furnace through the capillary of the autosampler arm has been explained in detail el~ewhere.~*q~~ The over-all efficiency of this sampling strategy for cobalt was found to be approximately 35%. All reagents were of analytical-reagent grade (Merck) and doubly de-ionized water was used throughout. A 0.05% m/v NaDDC solution was prepared in a buffer solution (0.06 mol dm-3 ammonia + 0.03 mol dm-3 acetic acid; pH 9) and purified on-line by passing the solution through a high- capacity (500 pl) precolumn filled with the same material as the preconcentration column (see Fig. 1).A working stock standard solution (0.100 mg 1-l) was prepared by step-wise dilution of a 1000 mg 1-l stock solution with 0.2Oh m/v nitric acid. Cobalt reference solutions (0.025-0.20 pg 1-l) were prepared by adding a suitable volume of working stock standard solution to 100 ml of 0.2% nitric acid with use of a diluter (TAM Dispenser) and balance (Mettler PM 2000) instead of calibrated flasks in order to reduce the risk of contamination. The NRCC CRMs used in this work were CASS-2 Nearshore Seawater NASS- 1 Open Ocean Seawater and SLEW-1 Estuarine Water. Results and Discussion Optimization of Analytical Parameters The dependence of the extraction efficiency on the pH of the solution is one of the important parameters that can have a significant influence on the over-all performance of the solid-sorbent extraction method.If a narrower pH range must be maintained slight changes in the acidity from sample to sample can result in a deterioration of the extraction efficiency and hence in an increased variability of the results. The efficiency of preconcentration expressed as the integrated absorbance of a cobalt reference solution was studied as a function of the pH by measuring the signal after 1 min preconcentration and the pH of the column eMuent in sequence 1. For this purpose cobalt reference solutions (0.2 pg 1-I) were prepared in nitric acid of different concentrations before on-line preconcentration. No change was observed in the pH range 2.5-9.The effect of the concentration of the chelating agent (NaDDC) on the efficiency of the on-line sorbent extraction of cobalt was investigated by using the same reference solution (0.2 pg 1-l Co) and a 1 min preconcentration time. For the investigated concentration range of 0.02-0.1% m/v NaDDC there was no influence on the integrated absorbance signal for cobalt. While the lowest concentra- tion of the chelating agent would be sufficient for complexa- tion of the traces of cobalt typically found in sea-water 0.05% NaDDC was used for all further experiments. Other heavy metals which can also form complexes and hence consume NaDDC,40*44 can be present in sea-water samples. The higher concentration of the chelating agent could therefore avoid possible interference from other heavy metal ions.Loss of complexed metals on container walls etc. as reported by other was avoided by the on- line formation of the complex and use of the carefully designed manifold with a short connection between the confluence point of sample with the NaDDC solution and the column. There is a direct relationship between the loading flow rate and the sampling frequency so that high flow rates are more desirable to achieve a high sample throughput. However high sampling flow rates have been reported to impair the efficiency of some solid sorbents. In the system discussed in this paper with C18 as the sorbent material no deterioration of the column efficiency was found with flow rates increasing up to about 5 ml min-l. The loading flow rate was only limited by the back-pressure produced by the column which eventually impaired the precision of the liquid volume transported per unit time.A total flow rate of sample and complexing agent of 4.3 ml min-l was therefore used throughout this work which resulted in relatively short loading times and good precision. The importance of rinsing all tubes and the column carefully before elution has been discussed in detail previ~usly.~~ It has also been shown that the direction of rinsing has a decisive influence. Rinsing in a direction counter to the direction of loading was more efficient than rinsing in the same direction even when much longer times were used.43 The reason for this is that with a reverse flow in the rinsing step not only are the non-adsorbed constituents of the matrix removed but possibly even those concomi- tant elements that are retained on the column. Retention of concomitant elements occurs either because they also form DDC complexes or because of the ion-exchange capacity of the sorbent material.Depending on the pH of the rinsing solution and the stability of the complexes formed more weakly bound elements could be removed while more strongly bound elements would be retained on the column. When the column was rinsed in the direction of sample loading less strongly bound elements migrated along theJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL. 6 619 < 0.10 I + 2 0.08 P 2 c m -! Fig. 2 Effect of pyrolysis temperature (see Table 1 step 2) on the integrated absorbance of 0.2 pg 1-l cobalt reference solution after 1 min preconcentration column as in chromatography while they were removed from the column very effectively when the direction of flow was reversed.The most important parameter in this context is the kinetic stability of the analyte-DDC complex and the strength of sorptive bonding to the column in acid solution. Consequently careful optimization of the pH is essential for efficient rinsing. For this purpose the solid-sorbent column was rinsed with nitric acid in concentrations ranging from 0.02 to 2% d v . The CoDDC complex was found to be stable and was retained quantitatively on the column even when 2% nitric acid was used for rinsing. This result is essentially in agreement with those of Isshiki and Nakayama,"o who reported that cobalt was retained on solid-sorbent columns with a wide range of ligands even when 1 mol dm-3 hydro- chloric acid was used for rinsing.Nitric acid (0.5%) was used in all further experiments to remove the sea-water matrix. This concentration was found to be very effective and gave low background signals in the ETAAS detection step. When ETAAS is used for determinations the volatility and thermal stability of the analyte-DDC complexes in ethanolic solution are important. Many chelate complexes and other organometallic compounds are known to sublime undecomposed. l6 In ETAAS this could result in analyte loss during pyrolysis. In order to investigate the volatility of the cobalt complex the effect of pyrolysis temperature on the absorbance of ethanolic CoDDC extracts was ascertained.For this purpose the column eluate was dispensed by the FI system into the cavity of a pyrolytic graphite platform mounted in the graphite tube. Use of the platform results in improved atomization conditions because spreading of the sample in the tube is restricted. The results in Fig. 2 show that the integrated absorbance of cobalt remains constant up to a pyrolysis temperature of 1500 "C indicating that the thermal stability of cobalt from the chelate-ethanol solu- tion is almost the same as that of cobalt in aqueous solution in a graphite furnace (this is in good agreement with results obtained by Komhek and co-worker~).~~~~ The high stabil- ity offers the possibility of separating the analyte element from more volatile concomitants in the pyrolysis stage and hence of increasing the specificity and reducing background absorbance.A pyrolysis temperature of 1000 "C was found to be ideal and was used throughout this work for all analytical applications. Performance of the Sorbent Extraction-Preconcentration System The performance of on-line sorbent extraction-preconcen- tration ETAAS for cobalt is summarized in Table 3. Because of the low cobalt concentration typically found in sea-water a preconcentration time of 120 s was used and a 42-fold enhancement in peak area compared with that for Table 3 Performance of the on-line sorbent extraction-precon- centration system for cobalt in sea-water Preconcentration time/min Enrichment factor Characteristic concentratiodng 1-I Detection limit (30)/ng 1-l Precision (% RSD) Sample throughputh-l Sample consumptiodml Reagent consumption per sample/ml Ethanol 0.05% NaDDC 2 42 5.8 6.4 5* 5.6 0.7 3 17 *Using SLEW- 1 Estuarine Water (n = 9).tUsing NASS-1 Open Ocean Seawater (n=4). 10 210 1.2 1.7 1 ot 5 28 0.7 15 0.2 I 1 0 1 .o 2.0 3.0 Time/s Fig. 3 Superimposed atomization signals for cobalt in different aqueous reference solutions after sorbent extraction preconcentra- tion with a 2 min loading time A 25 ng 1-l; B 50 ng 1-I; and C 75 ng 1-l 40 pl direct introduction was obtained (corresponding to an over-all efficiency of 35%). An analytical curve was established by using matrix-free cobalt reference solutions in the range 0.005-0.1 pg l-l and the linear regression equation was y= -0.00006 +0.758x with a correlation coefficient of 0.99997 (three superimposed atomization signals for cobalt reference solutions are shown in Fig.3). Because the cobalt concentration in NRCC CRM NASS-1 was lower than the detection limit obtained with a 2 min sample preconcentration further experiments were carried out in order to evaluate the column capacity. The integrated absorbance signal was found to be directly proportional to the preconcentration time at least up to 10 min and the standard deviation for the blank did not increase much with preconcentration time. The mean integrated absorbance values (n = 10) of the blank for 2 and 10 min preconcentrations were 0.00 1 6 and 0.002 1 respec- tively. Typical blank signals for 2 and 10 rnin preconcentra- tion as displayed by the instrument are shown in Fig.4(a and b). With a sample loading time of 10 min correspond- ing to 28 ml of sample an enrichment factor of 2 10 and a detection limit of 1.7 ng 1-l (30) could be achieved. 1 1 0 1 .o 2.0 3.0 Time/s Fig. 4 Atomization signals for 0.2% nitric acid blank solution after on-line sorbent extraction using different loading times (a) 2; and (b) 10 min620 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL. 6 Table 4 Determination of cobalt in sea-water and estuarine water standard reference materials by on-line sorbent extraction-precon- centration ETAAS Certified reference Certified Found?/ CASS-2 5 0.025 & 0.006 0.027 & 0.004 SLEW-1 9 0.046 & 0.007 0.053 f 0.003 NASS- 1 $ 4 0.004 & 0.00 1 0.0039 k 0.0004 material Replicates value*/pg 1-I Pg I-' *The uncertainties represent 95% confidence limits.tMean values and standard deviations are given. 10 min preconcentration. 0 1 .o 2.0 3.0 Time/s Fig. 5 Atomization signals for cobalt in certified reference materials after sorbent extraction preconcentration with 2 min loading time (a) CASS-2; and (6) SLEW-1 The accuracy of the proposed method for the determina- tion of low concentrations of cobalt in samples with a high total dissolved solids content was tested by the analysis of NRCC CRMs CASS-2 NASS-1 and SLEW-1. The samples which were acidified for conservation purposes to pH 1.6 by the supplier were used without further treatment. The results shown in Table 4 agree well with the certified values with precisions that are more than adequate for such low cobalt concentrations.Typical cobalt atomization signals are shown in Fig. 5(a and b). The background absorbance signals for sea-water samples are low and are not higher than the background signals obtained from aqueous standard solutions. The background absorbance detected by the instrument is in essence due to incomplete resolution of the Zeeman line splitting which produces a slight overlap of the cr components of the absorption profile with the emission line. Zeeman-effect background correction was not necessary for this applica- tion although the Zeeman 3030 was the instrument used for this study. Conclusion Combining a micro-scale FI preconcentration system on- line with ETAAS results in a powerful integrated hybrid system. The most outstanding advantage is the greatly improved detection limit achieved by preconcentration and matrix separation. The totally closed system together with on-line purification results in very low and reproducible blanks allowing the determination of ultra-trace concentra- tions of elements even in laboratories not equipped with clean-room facilities.The automated sample processing avoids time-consuming manual work and operator interac- tion thereby ehancing reproducibility and precision. Solid- sorbent extraction offers high preconcentration and selec- tivity for the separation of ultra-trace amounts of metals from sea-water matrices. Direct introduction of the eluate helps to overcome the problems with the analysis of organic extracts in ETAAS as discussed in the introduction.Processing samples and standards by the same system avoids calibration problems caused by a different response from organic and aqueous solutions and results in excellent accuracy. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 References Leyden D. E. and Wegscheider W. Anal. Chem. 1981 53 1059A. Tao H. Miyazaki A. Bansho K. and Umezaki Y. Anal. Chim. Acta 1984 156 159. Allen E. A. Bartlett P. K. N. and Ingram G. Analyst 1984 109 1075. Sturgeon R. E. Berman S. S. Desaulniers A. and Russell D. S. Talanta 1980 27 85 Smits J. Nelissen J. and van Grieken R. Anal. Chim. Acta 1979 111 215. Chakraborti D. Adams F. Van Mol W. and Irgolic K. J. Anal. Chim. Acta 1987 196 23. Boniforti R. Ferraroli R. Frigieri P. Heltai D.and Queirazza Q. Anal. Chim. Acta 1984 162 33. McLeod C. W. Otsuki A. Okamoto K. Haraguchi H. and Fuwa K. Analyst 1981 106 419. Sturgeon R. E. Berman S. S. Desaulniers J. A. H. Mykytiuk A. P. McLaren J. W. and Russell D. S. Anal. Chem. 1980 52 1585. Danielsson L.-G. Magnusson B. and Westerlund S. Anal. Chim. Acta 1978 98 47. Kinrade J. D. and van Loon J. C. Anal. Chem. 1974 46 1894. Brooks R. R. Presley B. J. and Kaplan I. R. Talanta 1967 14 809. Hulanicki A. Talanta 1967 14 137 1. Betz M. Gucer S. and Fuchs F. Fresenius 2. Anal. Chem. 1980 303 4. Karwowska R. Bulska E. Barakat K. A. and Hulanicki A. Chem. Anal. (Warsaw) 1980 25 1043. Volynsky A. B. Spivakov B. Ya. and Zolotov Yu. A. Talanta 1984 31 449. Magnusson B. and Westerlund S. Anal. Chim. Acta 1981 131 63.Shan X.-q. Tie J. and Xie G.-g. J. Anal. At. Spectrom. 1988 3 259. Caballero M. Lopez R. Cela R. and Perez-Bustamante J. A. Anal. Chim. Acta 1987 196 287. Akagi T. Fuwa K. and Haraguchi H. Anal. Chim. Acta 1985 177 139. Hiraide M. Ito T. Baba M. Kawaguchi H. and Mizuike A. Anal. Chem. 1980 52 804. Hudnik V. Gomiscek S. and Gorenc B. Anal. Chim. Acta 1978 98 39. Hiraide M. Yoshida Y. and Mizuike A. Anal. Chim. Acta 1976 81 185. Beauchemin D. and Berman S. S. Anal. Chem. 1989 61 1857. Nakashima S. Sturgeon R. E. Willie S. N. and Berman S. S. Fresenius Z. Anal. Chem. 1988 330 592 Van Berkel W. W. Overbosch A. W. Feenstra G. and Maessen F. J. M. J. J. Anal. At. Spectrom. 1988 3 249. Fang. Z. Xu S. and Zhang S. Anal. Chim. Acta 1987 200 35. McLaren J. W. Mykytiuk A.P. Willie S. N. and Berman S. S. Anal. Chem. 1985 57 2907. Marshall M. A. and Mottola H. A. Anal. Chem. 1985 57 729. Hartenstein S. D. RbiiEka J. and Christian G. D. Anal. Chem. 1985,57,21. Fang Z. RbiiEka J. and Hansen E. H. Anal. Chim. Acta 1984 164 23. Malamas F. Bengstsson M. and Johansson G. Anal. Chim. Acta 1984 160 1. Willie S. N. Sturgeon R. E. and Berman S. S. Anal. Chim. Acta 1983 149 59. Sturgeon R. E. Berman S. S. Willie S. N. and Desaulniers. J. A. H. Anal. Chem. 1981 53 2337.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL. 6 62 1 35 36 37 38 39 40 41 42 Mykytiuk A. P. Russell D. S. and Sturgeon R. E. Anal. Chern. 1980,52 128 1. Kingston H. M. Barnes I. L. Brady T. J. Rains T. C. and Champs M. A. Anal. Chem. 1978,50 2064. Braun T. and Abbas M. N. Anal. Chim. Acta 1980,119 1 13. Stella R. Ganzerli Valentini M. T. and Maggi L. Anal. Chern. 1985,57 1941. Plantz M. R. Fritz J. S. Smith F. G. and Houk R. S. Anal. Chern. 1989,61 149. Isshiki K. and Nakayama E. Anal. Chern. 1987 59 291. Hydes D. J. Anal. Chem. 1980 52 959. Fang Z. Sperling M. and Welz B. J. Anal. At. Spectrom. 1990 5 639. 43 Sperling M. Yin X. and Welz B. J. Anal. At. Spectrom. 1991 6 295. 44 RGiiEka J. and Arndal A. Anal. Chim. Acta 1989 216 243. 45 Komdrek J. and Sommer L. Talanta 1982 29 159. 46 Komdrek J. Kolcava D. and Sommer L. Collect. Czech. Chem. Commun. 1980,45 3313. Paper 1 /01839F Received April I9th 1991 Accepted July 2nd 1991