JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL. 9 1195 Determination of Ultra-trace Amounts of Cobalt in Ocean Water by Laser-excited Atomic Fluorescence Spectrometry in a Graphite Electrothermal Atomizer With Semi On-line Flow Injection Preconcentration" Alexander I. Yuzefovsky Robert F. Lonardo Mohui Wangt and Robert G. MichelS Department of Chemistry University of Connecticut Storrs CT 06269-3060 USA A method has been developed for the determination of trace and ultra-trace amounts of cobalt in sea-water. Samples of CASS-2 Nearshore Seawater and NASS-4 Open Ocean Water reference materials from the National Research Council of Canada were employed. Laser-excited atomic fluorescence spectrometry in an electrothermal atomizer (ET-LEAFS) was used and integrated with semi on-line flow injection microcolumn preconcentration.For cobalt the effects of pH on the preconcentration efficiency the concentration of the chelating agent and the distribution of cobalt in the ethanol eluate were studied. A bonded silica with octadecyl functional groups (C,8) in a 10 kl column was employed for preconcentration of cobalt in ocean water. Ocean water volumes of 0.40 and 1.00 ml were required for the determination of cobalt in CASS-2 and NASS-4 respectively. These volumes were almost two orders of magnitude smaller than those required by inductively coupled plasma mass spectrometry and some other competitive techniques. The preconcen- tration factors were 5- and 12.5-fold for CASS-2 and NASS-4 respectively. The detection limits (3s) based on 12.5-fold preconcentration were 0.08 and 1 .O ng I-' for cobalt in aqueous standard solutions and in Ocean Water Reference Materials respectively.Results for the determination of cobalt in CASS-2 and NASS-4 showed that there were no significant differences between the certified values and the measured values based on Student's t-test at the 95% confidence level. The relative standard deviations for the determinations of the concentrations of cobalt in CASS-2 and NASS-4 were 9 and 13% respectively. Keywords Laser-excited atomic fluorescence spectrometry; graphite furnace; flow injection; cobalt; ocean water; preconcentration The determination of trace amounts of cobalt in natural waters'-' is of great interest because cobalt is important for living species as complexed vitamin BI2.Vitamin BI2 is present in human and animal cells in the forms of adenosylcobal- amin(m) and methylcobalamin(1v). The deficiency of cobalt in ruminants usually results in different types of anaemia. Toxicological effects of large amounts of cobalt include vaso- dilation flushing and cardiomyopathy in humans and animals. The importance of cobalt in human and ruminant nutrition has led to work on the determination of cobalt in soils plants feedstuffs herbage natural waters and fertilizers. Investigations have extended to the biochemistry of cobalt in animals humans microorganisms and enzyme^.^-^ However there is still very little information available concerning the distribution and speciation of cobalt in the environment owing to analytical difficulties.There are two major problems in the determination of cobalt in ocean water. Firstly the high salt content of the ocean water matrix has often resulted in analytical inaccuracies. Secondly the concentration of cobalt in ocean water is below or very close to the detection limit of the most sensitive analytical techniques." For the direct determination of cobalt in aqueous standards the most recent detection limits reported were 0.1 pg I-' for electrothermal atomic absorption spec- trometry (ETAAS) by Slavin;" 7 ng 1-' for inductively coupled plasma mass spectrometry (ICP-MS) by Akatsuka et al.;' and 0.1 and 1.5 ng 1-' reported for laser-excited atomic fluorescence spectrometry in an electrothermal atomizer (ET-LEAFS) by Remy et and Irwin et ~ l .' ~ respectively. Detection limits for cobalt of between 1.0 and 3.0ng1-' have routinely been achieved with ET-LEAFS. * Presented at the XX Annual Meeting of the Federation of Analytical Chemistry and Spectroscopy Societies (FACSS) Detroit MI USA October 17-22 1993 as paper No. 0639. On leave from Chengdu College of Geology Chengdu People's Republic of China. $ To whom correspondence should be addressed. Recently flow injection microcolumn preconcentration has been employed to pre-treat samples prior to analyses in a variety of ways,14 and can be integrated with any of the techniques mentioned above. The technique addresses the two main problems of ocean water analyses as it allows for simultaneous matrix separation and analyte preconcentration during a fairly short analysis time.Flow injection preconcen- tration procedures reduce the sample consumption and are more rugged methods of analyses than other procedures. RdiiEka and Arndal'' demonstrated that extraction procedures for metals as their chelates from aqueous samples can be simplified miniaturized and automated by flow injection- sorbent extraction techniques. For example the chelate is formed in the flow stream sorbed onto c18 bonded silica eluted and then transferred into the detection system. One ancillary advantage of this flow injection procedure is that the apparatus is closed to the environment which reduces the likelihood of contamination. This is a critical consideration for the determination of ultra-trace amounts of metals in ocean water. For on-line flow injection determination of trace and ultra- trace amounts of cobalt in water Hirata et a1.I6 reported a detection limit of 2.9 mg 1-l by flame atomic absorption spec- trometry (FAAS) interfaced with ion-exchange column precon- centration.Backstrom and Danielsson17 reported 10 pg 1-' as the detection limit for cobalt by ETAAS interfaced with liquid- liquid extraction. A lower detection limit of 0.2 pg 1-l was reported by Fang et for ion-exchange column preconcen- tration interfaced with an ICP. Sperling et aLi9 employed c18 microcolumn separation and preconcentration with ETAAS for the determination of ultra-trace amounts of cobalt in natural waters. A detection limit of 1.7 ng I-' was reported for cobalt based on 700-fold preconcentration. Zhang et a1." obtained a detection limit of 0.47 ng I-' for cobalt in ocean water through the use of cathodic stripping square-wave voltammetry. Christian2' obtained a detection limit of 3 ng I-' for cobalt in sea-water through a reductive precipitation tech-1196 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL.9 nique using ICP-MS. The lowest detection limits for cobalt in ocean water reference materials have been obtained by flow injection microcolumn preconcentration interfaced with an ICP-MS instrument and were 50 ng 1-1 by Beauchemin and Berman;22 0.2 ng 1-' by McLaren et al.,23 with 50-fold precon- centration of the sample; and 0.1 ng I-' by Akatsuka et based on 90-fold preconcentration of the sample. It was anticipated that ET-LEAFS would be more sensitive than all or most of these approaches which was the stimulus for the work reported here.There were three aims addressed in the present work. The main goal was to realize the potential of the use of flow injection semi on-line microcolumn preconcentration in con- junction with ET-LEAFS. Femtogram detection limits can be routinely achieved by ET-LEAFS.11,24-26 Such detection limits are 1-4 orders of magnitude lower than those of ETAAS. In addition the linear dynamic range can be up to five orders of magnitude greater than that for ETAAS. Commercial electro- thermal atomic absorption equipment can be employed with- out modification. Thus a flow injection semi on-line microcolumn preconcentration system can be interfaced in the same way with ET-LEAFS as for ETAAS.27 All these features make flow injection ET-LEAFS a promising technique for the determination of ultra-trace amounts of elements in samples with detection limits comparable to or better than ICP-MS.A second aim was to decrease the sample consumption as much as possible and at the same time maintain good precision and accuracy. Sample volumes of only 400 and 100Op1 for CASS-2 and NASS-4 respectively were realized in the present work for the preconcentration of cobalt. This preconcentration volume was between ten and several hundred times smaller than the volumes required for other techniques for the same reference material^.^"^'^* A third aim of this work was to continue the investigation of sodium diethyldithiocarbamate (NaDDC) as a separating reagent with a c18 c o l ~ m n .~ * ~ ~ This approach has been ~ s e d ~ . ' ~ ~ ' successfully to preconcentrate and separate ultra- trace amounts of lead copper and cadmium by flow injection ETAAS for sea-water reference materials. Ethanol was used as the eluent to extract the heavy metal-DDC chelate complexes from the c18 column for direct introduction into a graphite furnace. The analytical conditions such as pH and flow rate were optimized. The preconcentration efficiencies for cobalt on a c18 10 pl column pre-loaded with NaDDC and an unloaded column were compared. Experimental Apparatus The instrumentation for ET-LEAFS (Table 1) has been dis- cussed in detail and is summarized briefly here. An excimer laser which was operated with xenon chloride (308 nm) at a repetition rate of 500 Hz was used to pump a tunable dye laser.Rhodamine 610,o-( 6-diethylamino-3-diethyl- imino-3H-xanthe-9-yl) benzoic acid (Exiton Dayton OH USA) was employed as the laser dye at concentrations of 0.91 and 0.30gl-' in absolute methanol for the oscillator and amplifier dye cells respectively. The frequency doubled output was passed through beam expansion to adjust the beam diameter to 2-3 mm before passage through the atomizer. The atomizer was a Perkin-Elmer HGA-500 graphite tube furnace equipped with an AS-40 autosampler and a L'vov platform. Both windows of the furnace were angled to reduce the stray laser background r a d i a t i ~ n . ~ ~ Cobalt atoms were excited at 304.4 nm. Fluorescence was detected at 340.5 nm in a scheme termed front-surface ill~mination.~~ An off-axis ellipsoidal mirror with an aluminium reflective surface and overcoated with magnesium fluoride was employed to collect the fl~orescence.~~ Table 1 ET-LEAFS instrumentation Component and Model No.Excimer laser EMG 104 MSC Dye laser FL 3002E Boxcar averager 162 165 Photomultiplier tube (PMT) Monochromator H-10 Graphite furnace HGA-500 9893QB-350 Triggering circuitry Data processing software Off-axis ellipsoidal mirror AlMgF coated f = 140 mm; f2 = 260 mm ~ ~~ Manufacturer Lambda Physik Acton MA Lambda Physik Acton MA PAR Princeton NJ USA Thorn-EMI Fairfield NJ ISA Metuchen NJ USA Perkin-Elmer Norwalk CT Laboratory made Asyst Software Rochester NY USA Aero Research Associates Port Washington NY USA USA USA USA USA The detection system consisted of a photomultiplier tube a preamplifier with a gain of ten and a boxcar integrator with a gate width of 5 ns a gate time constant of 0.5 ps and an output time constant of 10ms.The data processing was carried out with a personal computer Dell PC 200-80286. The integrated signal peak area was employed throughout the work. The graphite furnace temperature programme for the determination of cobalt in the ethanol eluate is given in Table 2. For preliminary work on the optimization of the analytical conditions for the determination of cobalt a Perkin-Elmer Model 5000 atomic absorption spectrometer with Zeeman- effect background correction was employed with an HGA-500 graphite furnace and an AS-40 autosampler. A cobalt hollow cathode lamp was operated at 20 mA. The absorption signals were processed with a Model 2108 personal computer WYSEpc 286 which was directly connected to the RS 232C/TTY port of the spectrometer.The graphite furnace temperature programme for the determination of cobalt in the ethanol eluate was similar to the one used for ET-LEAFS (Table 2). The flow injection semi on-line microcolumn preconcen- tration system (Fig. 1 ) was identical to that discussed pre- v i o ~ s l y . ~ ~ The c18 microcolumn was attached to a three-way valve (Rainin Instrument Co. Woburn MA USA) by means of 0.8 mm i.d. poly( tetrafluoroethylene) (PTFE) tubing (Rainin). The valve was used to direct solution through the column in either direction. Switching of the three-way valve was performed manually. Important details of the various steps in the procedures for flow injection labelled 1-4 in Fig.1 are discussed below. In the present work a 10 pl cylindrical microcolumn was employed for the sample analysis. For some preliminary studies an NaDDC loaded c18 column was used. To prepare the NaDDC loaded C column Table 2 Graphite furnace temperature programme for the determi- nation of cobalt by ETAAS and ET-LEAFS in the ethanol eluate Time/s Temperature/ Argon flow rate/ Step "C Ramp Hold ml min-' 1* 90 5 75 300 2 1000 5 40 300 3 20 1 5 300 5 2650 1 5 300 4 2200/2400t 0 7 O$ "Two sequential drying steps were used to dry the two 4Opl aliquots of the sample. t On the atomization step 2400 "C was used with ETAAS 2200 "C was used with ET-LEAFS (to reduce the black-body radiation gener- ated by a graphite furnace).2. Read.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL. 9 1197 Pe ri st a It ic Pump 3-Way valve -1 Microcolumn *- I J I I GraDhite furnace II IV v Fig. 1 Schematic diagram of the flow injection system (three-way valve in a sampling position) for ETAAS/ET-LEAFS. Arabic numbers correspond to procedural steps 1-4 described in the text. The arrows indicate the flow directions of the ethanol and aqueous solutions during the various steps; The containers I-V were used as follows I mixture of sample and NaDDC solutions; 11 sub-boiled distilled water; 111 ethanol eluate collector; IV waste solutions; and V ethanol the basic CI8 column was rinsed with ethanol for 3min and air dried for 1 min. Then 400 pl of a 0.05% NaDDC solution (in a buffer of pH 9) was passed through the column.This implies that DDC functional groups were pre-adsorbed onto the surface of the c18 particles in the column?7 Finally the c18 column was again air dried for 1 min. The flow rate to load the sample (0.15mlmin-') was comparable to the flow rate that was used by S l a ~ i n ~ ' with a cylindrical column but was slower than the flow rates that were used by Fang et aL2 (2.1 ml min-') and Sperling et aL3 (3.0 ml min-') with conical columns. The limitation in flow rate was due to the back pressure of the column which primarily depends on some physical characteristics of the system such as dimensions column capacity particle size and extent of compaction of the packing material. The same flow rate was employed for both preconcentration of the analyte onto the column and elution of the analyte from the column.Higher flow rates were attempted but could not be achieved. Reagents Sodium DDC (J. T. Baker Chemicals Phillipsburg NJ USA) which is soluble at pH values above about 7 was initially dissolved in a buffer solution (0.06 moll-' ammonia+ 0.03 mol 1-1 acetic acid pH 9). Ultimately during the cobalt preconcentration step the pH of the sample plus buffer solution was adjusted to the optimum range of 2.0-2.5 as discussed later. Other reagents included ultra-pure nitric acid (J. T. Baker Chemicals); absolute ethanol 200 proof (AAPER Alcohol and Chemicals Shelbyville KY USA). Reversed-phase silica bonded with an octadecyl functional group (CIS) 40 pm (J. T. Baker Chemicals) was used as the solid sorbent.The CASS-2 Nearshore Seawater and NASS-4 Open Ocean Water reference materials for trace metals which as delivered have a pH of 1.6 were obtained from the National Research Council of Canada Ottawa Ontario Canada while sub-boiled distilled water prepared immediately prior to all analyses was used throughout. Cobalt standard solutions in the concentration range of 0.05-5.00 pg 1-' were prepared daily by stepwise dilution of a 1000 mg 1-' stock solution (J. T. Baker Chemicals) with 0.2% v/v nitric acid. It was vitally important to maintain control of the contami- nation throughout all the experiments because the flow injec- tion system had some containers that were partially open to the environment during the analytical cycle. The following procedure was employed.All samples and standards were prepared in a class 100 (US Federal 209b) clean-air hood and on a class 100 clean bench. All containers and pipette tips used in this work were cleaned by soaking them first in 20% laboratory-reagent grade nitric acid for approximately 24 h followed by a rinse with de-ionized water then a rinse with a (1 + 1 v/v) mixture of 0.05% NaDDC in buffer solution and 0.2% nitric acid and a final rinse with sub-boiled distilled water.27 The 0.05% NaDDC solution for both the cleaning procedure and the determination of cobalt in CASS-2 was purified in advance by pumping the solution through a 500 pl CI8 conical column. For NASS-4 the purification of the NaDDC solution was performed twice because of a much lower concentration of the analyte in the sample.Recommended Procedure One of the important factors that affected the preconcentration efficiency for cobalt on a C column was the pH of the mixture of the sample and NaDDC. It was found that the optimum pH range for the preconcentration of cobalt was between a pH of 1.7 and 4.0 (Fig. 2). This was taken into account in step 1 of the following procedure for the preparation of sample-NaDDC mixtures. For the determination of cobalt in CASS-2 in step 1 400 pl of ocean water sample were mixed with 200 pl 0.05% NaDDC in the pH9 buffer. The pH of the resultant mixture was 2.5. For the determination of cobalt in NASS-4 in step 1 1000 p1 of ocean water sample were mixed with 400 ~10.05% NaDDC in the buffer at pH 9. The pH of the resultant mixture was 2.4.Hence the acidities of both CASS-2-NaDDC and NASS-4- NaDDC mixtures lay in the optimum pH range. The resultant mixture of each sample was held in an autosampler cup and pumped directly through the column at a flow rate of 0.15 mlmin-'. This allowed the CoDDC chelate to be adsorbed onto the column. In step 2 100 p1 of sub-boiled distilled water were passed through the column in the same direction and at the same flow rate in order to rinse out the residual ocean water matrix retained on the column. In step 3 ethanol was passed through the column at the same flow rate as the sample solution but in the reverse direction in order to extract the CoDDC chelate. An autosampler cup equipped with a PTFE plug on the top was used as the collector for the ethanol eluate.The plug was employed to prevent evaporation of the ethanol eluate from the container. A small hole in the plug allowed the eluate to be poured in and to be taken out of the collector by the autosampler (Fig. 1; 111). In step 4 the ethanol eluate which contained the analyte was delivered by the autosampler directly onto the L'vov platform. The first 80 pl of ethanol eluate were collected and then introduced into the electrother- mal atomizer in two separate deliveries with sequential drying of each 40 p1 aliquot. II " a 0.10 1 .o 3.0 5.0 7 .O 9.0 PH Fig. 2 Effect of pH on the preconcentration efficiency for cobalt (800 pg of cobalt in 400 pl standard with 200 pl of 0.05% NaDDC solution). A 10 p1 unloaded CI8 column was used1198 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL.9 Results and Discussion The experimental section of this project was carried out in two parts. In the first part all the experimental conditions such as flow rate volume of the column and pH of the sample- NaDDC mixture were optimized on the flow injection ETAAS system because the laser system was not available for the experimental work. In the second part of the project the determination of ultra-trace amounts of cobalt in CASS-2 and NASS-4 was carried out by flow injection ET-LEAFS by use of the conditions optimized by ETAAS. E ~ p e r i e n c e ~ ~ ’ ~ ~ with electrothermal atomization has indicated that there is no difference between the techniques of absorption and fluor- escence in terms of the optimization of analytical methods because all the physico-chemical processes are identical when the same furnace is used for both.Fluorescence detection merely improves the sensitivity by several orders of magnitude and the furnace transients always appear temporally identical for both spectrometric methods of detection. It is often useful to optimize the methodology on the simpler atomic absorption instrumentation because of the operational and maintenance difficulties of working with excimer pumped dye lasers. It is expected that this will change in the near future as dye lasers are substituted with solid-state lasers. Nature of the C Column In previous both unloaded and NaDDC loaded c18 columns with different shapes and volumes were compared with respect to the distribution of copper and cadmium in the eluate.It was found that a cylindrical c18 column had an advantage over a conical column because most of the analyte was distributed in the first portion (30 pl) of the eluate. However for a conical column the analyte was distributed over more than 100 pl of the eluate. Therefore a cylindrical c18 column was employed throughout. Diethyldithiocarbamic acid (HDDC) behaves as a bidentate univalent anionic ligand and forms very stable chelates with most of the heavy-metal ions.34-36 The effect of pH on the preconcentration efficiency of copper and cadmium with the proposed flow injection system was discussed in detail pre- v i o ~ s l y . ~ ~ It was pointed out that the pH had a significant effect on the preconcentration efficiency of the analyte when either NaDDC loaded or an unloaded column was used.Furthermore it was shown that the preconcentration efficiency for copper could change by more than one order of magnitude within the pH range between 4.0 and 8.0. This report contra- dicted earlier work by Fang et aL2 and Sperling et aL3 They stated that the formation of metal-DDC complexes and their subsequent preconcentration on a column did not depend on pH. Here the same effect was investigated for cobalt. A series of analyte solutions in the pH range from 1.7 to 8.5 was prepared. The total volume of the analyte solution was held constant at 600 p1. The analyte solution was a mixture of 400 p1 of a standard that contained 800 pg of cobalt and 200 p1 of 0.05% NaDDC solution. Different acidities in this series of analyte solutions were achieved by changing the concentration of the nitric acid in the 400 pl of cobalt standard.It was found that the optimum pH range for cobalt in terms of the precon- centration efficiency was between a pH of 1.7 and 4.0 (Fig. 2) and the preconcentration efficiency slightly degraded at pH values higher than 4.0. These results together with the data obtained earlier,27 show that the optimum pH range was different for different elements. The preconcentration efficiency of cobalt was not as strongly dependent on pH as it was for copper and cadmium. Nevertheless this effect could not be ignored. The pH values used in the present work for CASS-2 and NASS-4 were 2.5 and 2.2 respectively. The distributions of analyte in the ethanol eluate after preconcentration of cobalt on an unloaded column and on a NaDDC loaded c18 cylindrical column are shown in Fig. 3.A comparison of signal sizes obtained for cobalt indicated that there was no difference between these two types of cylindrical c18 columns in terms of the distribution and the preconcen- tration efficiency of the analyte. The unloaded c18 column was chosen for the analysis of ocean water samples. a comparison of various sizes of c18 cylindrical columns showed that 10 pl was the optimum volume for the column in terms of the preconcentration efficiency and the distribution of copper and cadmium in the ethanol eluate. Hence in the present work a 10 p.1 unloaded C18 cylindrical column was employed throughout. In previous Effect of the Concentration of NaDDC Solution A series of 400 pl standards which contained 800 pg of cobalt mixed with 200 pl of NaDDC solutions to give a concentration range of from 0.01 to 0.20% m/v was preconcentrated on the c18 column by use of the procedure described above.The concentration of analyte in the ethanol eluate was measured for each solution. The results indicated that with a 10 pl unloaded C18 column the preconcentration efficiency for cobalt was almost constant over a wide range of NaDDC concen- trations from 0.03% to 0.20% (Fig. 4). A concentration of 0.05% m/v NaDDC was chosen. These results were similar to those obtained for preconcentration of copper and cadmium.27 Effect of Standing Time on the Mixtures of Sample and NaDDC Solutions During a preliminary set of experiments it was observed that the integrated absorbance values of the analyte degraded when mixtures of NaDDC with each sample were allowed to stand in autosampler cups for more than 20-30min (Fig.5). This phenomenon was considered to be a threat to good precision because of the relatively long period of time that was required for each sample to be loaded onto the C column. To avoid possible degradation in the precision of the analyses the effect 0.20 (a) 1 0.15 0.10 S!? 0.05 0 m f! $ 0 a 0.20 E Q 0 c 0.15 - 0.10 0.05 0 I 1 Volume of ethanol eluate/ml Fig. 3 Distribution of cobalt (800 pg of cobalt in 400 pl of standard) in the ethanol eluate after preconcentration on an (a) unloaded and (b) NaDDC loaded 10 p1 CIS columnsJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL.9 1199 0.20 1 0 0.05 0.10 0.15 0.20 INaDDCl(%) Fig. 4 Effect of the concentration of NaDDC on the preconcentration efficiency for cobalt (800 pg of cobalt in 400 pl of standard) 0.20 1 C D 0 100 200 S t a n d i n g ti me/m i n Fig. 5 Effect of standing time of the mixture of cobalt (800 pg of cobalt in 400 p1 of standard) and NaDDC (400 p1 of 0.05% NaDDC) on the preconcentration efficiency for cobalt A mixture was open to the air; B air C oxygen and D hydrogen gases were bubbled through the mixture tration efficiency of the analyte the solutions were passed through the preconcentration column within 20 min of prepar- ing the sample mixture. Effect of the Matrix in Ocean Water The presence of large amounts of matrix materials in ocean water requires preliminary separation from the sample prior to analysis. An investigation was carried out to determine whether or not dilution could be useful to remove interferences given sufficient analytical sensitivity. The results (Fig.6) showed that the laser scatter signal from a CASS-2 solution injected directly into graphite furnace was at least one order of magnitude larger than the fluorescence signal from 10 pg of standard cobalt which was equivalent to the amount of cobalt in 400 pl of CASS-2 reference material. Further dilution experi- ments indicated that the scatter signal for the matrix was so large that it was impossible to isolate the fluorescence signal of the analyte from the laser scatter without prior separation on the c18 column. Preconcentration allows separation of most of the matrix elements such as alkali and alkaline earth elements from cobalt and other heavy-metal ions but residual matrix elements can still be retained on the column.The retained matrix is rinsed into the detection system along with the analyte by the of standing time on the integrated absorbance values of cobalt was investigated in more detail. A series of mixtures of 200 p1 of 0.05°/~ NaDDC with 400 p1 of cobalt standard (800 pg of cobalt) was allowed to stand in autosampler cups open to air for the different periods of time. Thereafter cobalt from each mixture was preconcentrated on the CI8 column and the preconcentration efficiency was meas- ured. For mixtures that stood in containers for longer than 20-30 min before they were pumped through the c18 column the integrated absorbance values for cobalt decreased as a function of time.It was postulated that the effect was due to the decomposition of the chelating reagent in the sample- NaDDC mixtures. In order to identify the nature of this effect the experiment was repeated with different gases bubbled through the sample-NaDDC mixture. It was expected that if the driving force of the effect was oxidation of DDC by oxygen from the air or from the sample solutions an oxidative environment (oxygen gas) should accelerate the process and a reductive environment (hydrogen gas) should inhibit it. In order to keep the experimental conditions all the same air was also bubbled through the sample-NaDDC mixture as a reference point. It was found that the integrated absorbance values for cobalt did not depend on the nature of the gases that were passed through the mixtures.Moreover the signals decreased more rapidly as a function of time compared with the case when the mixtures were just open to the air. A clear explanation for these observations could not be found. It is possible that the stirring caused by the gases accelerated the kinetics of the process of decomposition of DDC molecules. It is also possible that some photo- or thermo-degradation process(es) took place in the sample solutions. This was not investigated further. To avoid possible losses of the preconcen- 304.360 304.400 304.440 Excitation wavelength/nm Fig. 6 Excitation spectra for cobalt A 20 pl of 0.025 pg 1-' solution of CASS-2; and B cobalt 10 pg in aqueous standard.Monochromator slit-width 0.5 mm (4 nm bandpass) and An = 340.5 nm e 0 0 5 0 Time/s 5 Fig. 7 Absorbance profiles for 500 pg of cobalt (a) without and (b) (c) ( d ) in the presence of residual ocean water matrix materials (a) cobalt in 400 pl of standard solution A = 0.073 s; (b) (c) ( d ) cobalt spiked in 400 pl of CASS-2; (b) without rinsing step A=0.085 s; (c) rinsed with 50 p1 of 0.02% nitric acid A =0.043 s; and ( d ) rinsed with 50 pl of sub-boiled distilled water A =0.074 s1200 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL. 9 eluate and can affect the precision and accuracy of the determination of cobalt in the ocean water. In previous work a washing step was a necessary part of the separation procedure prior to elution with ethanol.27 In order to find the best washing reagent for the removal of residual matrix retained on the CI8 column 500pg of cobalt were spiked into 400 pl aliquots of ocean water and either sub- boiled distilled water or 0.02% nitric acid was used during the washing step.The recoveries of the cobalt in the ethanol eluates were compared with the absorbance value obtained Table 3 Comparison of the efficiencies of the different volumes of the sub-boiled distilled water rinse of the residual matrix materials ~~ ~ Volume of sub-boiled distilled water/ Pl 20 0.072 50 0.074 100 0.072 150 0.057 Cobalt integrated abs o r bance/s * * Results were obtained by use of 500 pg of cobalt spiked in 400 pl of CASS-2. directly from 500pg of cobalt in an aqueous solution [Fig.7(a)]. For the first mixture the washing step was omitted completely and as expected the presence of the residual matrix in the eluate solution manifested itself as a large spike in the absorbance profile [Fig. 7(b)]. For the second and the third mixtures equal amounts of 0.02% nitric acid [Fig. 7(c)] and sub-boiled distilled water [Fig. 7 ( d ) ] were employed respectively to rinse out the residual matrix. From the results obtained it was clear that the use of 0.02% nitric acid gave worse recovery of the analyte than sub-boiled distilled water. Sub-boiled distilled water which gave essentially 100% recov- ery was used as the washing solution during all subsequent analyses. Different amounts of sub-boiled distilled water were used to find the optimum volume that was required for the separation of the residual matrix from the analyte.The results (Table 3) indicated that any amount of water rinse between 20 and 100 pl could be employed for the removal of the residual matrix. It was also observed that a further increase in the amount of water rinse to more than 100 pl resulted in a decreased integrated absorbance signal for cobalt owing to the leaching of some portion of the analyte from the C column. It could have been useful to buffer the water rinse Table 4 Comparison of results (in ng 1-l) for the determination of cobalt in ocean water reference materials; data are f95% confidence limits Reference material Certified value Present work* CASS-2 NASS- 1 NASS-2 4+1 - 4+1 - NASS-3 NASS-4 -tt 4.4 f 0.6 (13 n = 5 ~ Zhang et a1.t Sperling et a1.S Akatsuka et a1.6 23 t 0.4 27+4 -7 (2 n=6)11 (15 n=5)11 - - 3.9 f 0.4 - 4.8 f 0.4 - - (10 n=4)1i (8 n=8)11 - - 3.1 f0.5 (16 n=6-12**)11 * Analytical volume of ocean water 0.4 ml (CASS-2) 1.0 ml (NASS-4).t Ref. 20; Analytical volume of ocean water 10 ml. $ Ref. 19; Analytical volume of ocean water 5.6 ml (CASS-2) 28 ml (NASS-2). 6 Ref. 5; Analytical volume of ocean water 900 ml. 7 Not available. 11 RSD% n =number of determinations. ** The number of samples was not specified exactly for cobalt. t'f NASS-4 had not been certified at the time of this analysis but NASS-4 was nominally the same as NASS-1 2 3 ref. 37. Table 5 Detection limits for cobalt and performance of different techniques Work Method Present work Flow injection Zhang et al.11 Stripping square- Sperling et al.tt Flow injection Akatsuka et al.fl Flow injection ET-LEAFS wave voltammetry ETAAS ICP-MS [nstrumental Method Ocean water Column packing- detection limit/ detection limit/ detection limit*/ Concentration separating reagent ng 1-' ng I-' ng 1-' factor CI8-NaDDC 17 0.081 1 .w 12.5 C1,-NaDDC low 1.7 7 m - Silica-immobilized 7 8-h ydroxyquinoline * Estimated for the NASS Open Ocean Water reference materials.Based on the preconcentration factors for each method or as indicated. t Based on the variability (3s) of an aqueous blank. $ Estimated from the instrumental detection limit based on 12.5-fold preconcentration of the aqueous standards. 6 Calculated from measurements of the variability (3s) of the procedural blank which was 1.8 ng 1-'.7 Detection limit probably controlled by contamination of ethanol with cobalt. /I Ref. 20. No preconcentration was done. ** Not available. tt Ref. 19. $$ Ref. 11. &j Based on the calculation that 40 p1 of eluate were collected from 28 ml of sample. fl Ref. 5. 11 11 Estimated minimum concentration of cobalt in ocean water that can be determined ref. 5.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL. 9 1201 but this was not tried owing to the risk of contamination and because the procedures described here already provided accu- rate results. When air was allowed through the column before the water rinse it did not degrade the effectiveness of the removal of the retained matrix materials and did not affect the preconcen- tration efficiency for cobalt.These data contradicted previous observations where the presence of air in the column signifi- cantly degraded the effectiveness of the removal of the residual matrix from the column during the determination of copper in CASS-2.” An explanation for this difference could not be found. Results for the Ocean Water Sample Analysis Ultra-trace amounts of cobalt were determined in CASS-2 and NASS-4 Ocean Water reference materials (Table 4). Aqueous calibration proved possible. There was no difference found between the integrated absorption signal for cobalt from the ethanol solution with and without chemical modification with magnesium nitrate. Hence no chemical modifier was employed not only because it was not necessary but also to minimize the risk of contamination.The results indicated that both the precision and the accuracy of the method were satisfactory compared with the certified and values from other works. The relative standard deviations ( RSD) for the determinations of cobalt in CASS-2 and NASS-4 were 9 and 13% respectively (Table 4). Based on Student’s t-test at the 95% confidence level there were no significant differences between the meas- ured values and the values for the determination of cobalt in certified CASS-2 and non-certified NASS-4. There is a possibil- ity that some of the manual procedures used could have had some degradation effect on the precision of the technique. Detection limits for cobalt by a variety of techniques are presented in Table 5. The ET-LEAFS instrumental detection limit was determined after subtraction of an aqueous blank signal by extrapolation of the analytical curve to a signal level equal to three times the standard deviation of 16 measurements of the blank.Calculations for the detection limit of the present method were based on the instrumental detection limit and the 12.5-fold preconcentration factor. The ocean water detec- tion limits were measured from the standard deviation of the procedural blanks (3s) except where noted in Table 5. Despite the order of magnitude better instrumental detection limit of ET-LEAFS the ocean water detection limit for cobalt in the present work was essentially the same as that achieved by other techniques in Table 5. Although ET-LEAFS was nominally sensitive enough to perform the determination of such small amounts of cobalt without the preliminary precon- centration of the sample on the C column the preconcen- tration procedure was required to separate the matrix from the sample.The procedure introduced contamination via the reagents which negated the improvement in detection limit caused by preconcentration. This is evidenced by the pro- cedural blank level of 1.8 ng 1-’ (Table 5 ) . The same precision and accuracy of a n a l y s e ~ ~ ’ ~ ~ ~ ~ was achieved as those obtained by the alternative techniques dis- cussed in the literature but the present method required less ocean water with volumes at least 1-2 orders of magnitude smaller than those used by other workers (Table4). If the volumes of the ocean water were to be increased the detection limit and the overall precision and accuracy of the method would possibly improve further with caveats about reagent borne contamination.The authors are very grateful to James W. McLaren of the National Research Council of Canada Marine Analytical Chemistry Standards Program for provision of several ocean water standards. Also we thank John T. McCaffrey and Susan McIntosh of Perkin-Elmer Walter Slavin of Bonaire Technologies Zhang Li of AMSPEC and our colleague Evelyn G. Su all of whom helped with parts of this work. This work was supported by an American Chemical Society Division of Analytical Chemistry Fellowship sponsored by Perkin-Elmer (awarded to A.I.Y.). This research employed some equipment that was purchased under grant number GM32002 from the National Institutes of Health.M.W. was supported by grants from the Government of the People’s Republic of China and the University of Connecticut Research Foundation. 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 References Porta V. Abollino O. Mentasti E. and Sarzamini C. J. Anal. At. Spectrom. 1991 6 119. Fang Z.-L. Sperling M. and Welz B. J. Anal. At. Spectrom. 1990 5 639. Sperling M. Yin X. and Welz B. J . Anal. At. Spectrom. 1991 6 295. Azeredo L. C. Sturgeon R. E. and Curtius A. J. Spectrochim. Acta Part B 1993 48 91. Akatsuka K. McLaren J. W. Lam J. W. and Berman S. S. J. Anal. At. Spectrom. 1992 7 889. Frieden E. Biochemistry of the Essential Ultratrace Elements Plenum New York 1984. Mertz W. Trace Elements in Human and Animal Nutrition Fifth Edition Academic Press New York 1987.Lavi N. and Altassi Z. B. Analyst 1990 115 817. Blanchflower W. J. Cannavan A. and Kennedy D. G. Analyst 1990 115 1323. Boniforti R. Ferraroli R. Frigieri P. Heltai D. and Queirazza G. Anal. Chim. Acta 1984 33 162. Slavin W. Graphite Furnace AAS. A Source Book Perkin-Elmer Ridgefield CT 1984. Remy B. Verhaeghe I. and Mauchien P. Appl. Spectrosc. 1990 44 1633. Irwin R. L. Wei G.-T. Butcher D. J. Liang Z. Su E. G. Takahashi J. Walton A. P. and Michel R. G. Spectrochim. Acta Part B 1992 41 1497. Carbonell V. Salvador A. and de la Guardia M. Fresenius’ J. Anal. Chem. 1992 342 529. RfiLiEka J. and Arndal A. Anal. Chim. Acta 1989 216 243. Hirata S. Honda K. and Kumamaru T. Bunzeki Kagaku 1987 36 678.Backstrom K. and Danielsson L. G. Anal. Chim. Acta 1990 232 301. Fang Z.-L. Xu S. and Zhang S. Anal. Chim. Acta 1987 200 35. Sperling M. Yin X. and Welz B. J. Anal. At. Spectrom. 1991 6 615. Zang H. Wollast R. Vire J.-C. and Patriarche G. J. Analyst 1989 114 1597. Christian J. D. Environ. Lab. 1993 5 10. Beauchemin D. and Berman S. S. Anal. Chem. 1989 61 1857. McLaren J. W. Mykytiuk A. P. Willie S. N. and Berman S. S. Anal. Chem. 1985 57 2907. Butcher D. J. Dougherty J. P. Preli F. R. Walton A. P. Wei G.-T. Irwin R. L. and Michel R. G. J. Anal. At. Spectrom. 1988 3 1059. Wei G.-T. Dougherty J. P. Preli F. R. Jr. and Michel R. G. J . Anal. At. Spectrom. 1990 5 249. Sjostrom S. and Mauchien P. Spectrochim. Acta Rev. 1993 15 153. Wang M. Yuzefovsky A. I. and Michel R. G. Microchem. J. 1993 48 326. Dougherty J. P. Preli F. R. McCaffrey J. T. Seltzer M. D. and Michel R. G. Anal. Chem. 1987 59 1112. Goforth D. and Winefordner J. Anal. Chem. 1986 58 2598. Yuzefovsky A. I. Lonardo R. F. and Michel R. G. Anal. Chem. in the press. Slavin W. Perkin-Elmer Norwalk CT personal communi- cation 1993. Dougherty J. P. Preli F. R. and Michel R. G. Talanta 1989 36 151. Butcher D. J. Irwin R. L. Takahashi J. Su E. G. Wei G.-T. and Michel R. G. Appl. Spectrosc. 1990 44 1521.1202 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY NOVEMBER 1994 VOL. 9 34 Cheng K. L. and Ueno K. Handbook of Organic Analytical Reagents CRC Press Boca Raton 1982. 35 Stary J. Solvent Extraction of Metal Chelates Pergamon 37 McLaren J. W. National Research Council Canada Institute for Environmental Science personal communication 1993. Morrison G. H. and Freiser H. Solvent Extraction in Analytical Chemistry Wiley New York 1957. Oxford 1964. 36 Paper 31073274 Received December 13 1993 Accepted June 30 1994