EVectiveness of fullerene as a sorbent for the determination of trace amounts of cobalt in wheat flour by electrothermal atomic absorption spectrometry Ma Mar Gonza�lez, Mercedes Gallego and Miguel Valca�rcel* Department of Analytical Chemistry, Faculty of Sciences, University of Co�rdoba, E-14004 Co�rdoba, Spain Received 19th October 1998, Accepted 12th February 1999 A flow injection system for the on-line preconcentration of ultratrace levels of cobalt by sorption with ammonium pyrrolidinedithiocarbamate on a C60 fullerene column was developed. The eluate volume is reduced and dispersion minimized by using a nitrogen (or air) stream to carry the eluent (IBMK) for elution of the adsorbed chelate, which is collected in a 500 ml PTFE autosampler cup.The analytical figures of merit for the determination of cobalt are as follows: limit of detection (3s), 8 ng l-1; precision (RSD), 4% for 0.2 ng ml-1; preconcentration factor, 40 (using 25 ml of sample and 500 ml of IBMK).Similar experiments involving C18-bonded silica as sorbent revealed ethanol to be the most suitable eluent under these conditions. The results obtained in the determination of cobalt in wheat flour testify to the usefulness of the proposed method. The high hydrodynamic impedance of the column to using Introduction high sample loading flow rates, and its short lifetime, can both The significance of cobalt to human and ruminant nutrition be overcome by adsorbing metal complexes on the inner walls has aroused great interest in its determination in a variety of of a knotted PTFE reactor.These novel preconcentration sample types. Cobalt, as vitamin B12, is essential to living systems have been extended to ETAAS for the automatic species; however, it is toxic in large amounts. Plants are determination of metals in water samples.14–16 By using this generally low in cobalt; however, they are the main source of preconcentration system, the adsorbed Co–PDC complex is this metal for the human diet.1 quantitatively eluted with 45 ml of ethanol and transferred The very low concentration of cobalt in foods and the directly to an ETAAS instrument.16 An enhancement factor presence of matrix interferences call for the use of a separation of 47 (6.7 ml of sample) was thus obtained at a sampling technique for its determination.2 Recently developed automatic frequency of 30 h-1.techniques provide a variety of eVective approaches to the Ever since its existence was confirmed,17 fullerene has problem, particularly those including a continuous preconcen- received substantial attention in various scientific fields.18 The tration unit.3–5 The advantages and pitfalls of liquid–liquid main problem with fullerenes continues to be the separation extraction in relation to other alternatives such as column of isomers and homologues; Jinno et al.have published about extraction procedures based on chelating resins and sorbents 25 papers on the separation and isolation of fullerenes from have been discussed.6 The use of on-line flow injection (FI ) carbon soot on common and specialized stationary phases by column preconcentration for the atomic spectrometric determi- HPLC.19,20 Numerous reviews and a recent monograph on nation of cobalt was reviewed by Fang et al.7 Based on this topic have also been published.21 The purity of fullerene solid-phase extraction with C18 silica gel, Sperling and precursors and products has also been examined; thus, data co-workers8–11 developed FI systems for determining this metal on trace element impurities in these products have been by electrothermal atomic absorption spectrometry (ETAAS).reported.22 The analytical potential of C60 fullerene as a For example, ultratrace amounts of cobalt in sea-water were sorbent material for the preconcentration of metals was first determined by using NaDDC as chelating reagent and demonstrated by Gallego et al.;23 subsequent experiments with reversed-phase C18-bonded silica as sorbent;10 an enhancement C60 and C70 fullerenes in continuous systems (and IBMK as factor of 42–210 was obtained with a 2–10 min sample loading eluent) were successfully applied to FAAS for the determitime (5–30 ml of sample).The principle of eluate zone sam- nation of copper24 and cadmium25 in biological samples. Both pling, whereby a 40 ml portion of the eluate containing the fullerenes were found to exhibit better sorbent properties in most concentrated fraction of eluted analyte (in ethanol ) is metal preconcentration than did conventional solid materials introduced directly into the graphite furnace through the (e.g.RP-C18, activated carbon and resins). Better sensitivity capillary in the autosampler arm, enables processing of high and selectivity were obtained with neutral chelates than by sample volumes with minimal contamination and analyte loss.ion-pair formation.24,25 Ma and Adams12 have investigated the applicability of alkyldi- Cobalt concentrations in cereals can be as low as 10 ng g-1 thiophosphates as chelating agents for cobalt and other metals; dry matter; the determination of such small amounts obviously C18-bonded silica gel as sorbent, methanol or ethanol as eluent requires a very sensitive analytical method. The purpose of and FAAS/ ETAAS detection were used. More recently, a this work was to extend the high analytical usefulness of preconcentration column packed with Muromac A-1 chelating fullerenes for metal preconcentration to ETAAS in order to resin was inserted at the tip of an autosampler arm to determine ultratrace amounts of cobalt in wheat flour.Following decomposition of organic matter by a simple, rapid preconcentrate several metals (cobalt excluded) in sea-water.13 J. Anal. At. Spectrom., 1999, 14, 711–716 711wet ashing procedure, cobalt, in 0.1 mol l-1 HNO3, is intro- columns packed with 80 mg of C60 fullerene or silica RP-C18.The mini-columns were made from PTFE capillaries of 3 mm duced into an FI system. The adsorbed Co–PDC complex is quantitatively eluted with IBMK and the whole eluate is id and sealed at both ends with small cotton-wool plugs to prevent material losses. They were initially flushed with directly transferred to a stoppered autosampler cup. 0.1 mol l-1 HNO3 (subsequent use of IBMK as eluent in each operating cycle was suYcient to make them ready for re-use). Experimental C60 fullerene columns were usable for at least 9 months.Reagents and standard solutions Sample preparation A 1000 mg l-1 cobalt stock solution was prepared by dissolving 1.000 g of the metal in a small volume of concentrated nitric Flour samples were prepared at a pilot plant; the raw material was wheat produce from diVerent Spanish locations. An acid and diluting to 1 l with 1% v/v nitric acid. A 0.1% m/v aqueous solution of APDC (Aldrich, Madrid, Spain) was amount of 0.5–1 kg of wheat containing ca. 15% m/m water was ground in a metal rolling mill to obtain 60% of white prepared; the solution remained stable for at least 3 d. A 1% m/v solution of neocuproine (Merck, Darmstadt, flour and 40% of by-products. The flour was screened through a 132 mm sieve. Flour samples were dried to constant mass in Germany) in ethanol was also prepared. Hydroxylammonium chloride and IBMK (Merck) were also used.C60 fullerene an oven at 103 °C. The wheat flours were prepared as follows: an accurately (>99.4%, Hoechst, Frankfurt-am-Main, Germany) and polygosyl- bonded silica reversed-phase sorbent with octadecyl weighed amount of 0.25 g was digested with 2 ml of 65% HNO3 and a few drops of 30% H2O2 in a glass beaker. The functional groups (RP-C18), 60–100 mm particle size (Millipore, Madrid, Spain), were employed as sorbent mate- mixture was heated on a hot-plate at about 200 °C to neardryness (about 8 min).Once cool, the residue was diluted with rials. Standard solutions (25 ml ) containing 0.02–1 ng ml-1 cobalt were all freshly prepared by appropriate dilution of a 0.1 mol l-1 HNO3 and transferred quantitatively into a 25 ml calibrated flask to which 1 g of hydroxylammonium chloride stock standard solution (1000 mg l-1) in 0.1 mol l-1 HNO3. An organic stock solution ofh;PDC was made by placing and 0.5 ml of 1% neocuproine (to avoid the interference of iron and copper, respectively) were added before making up 1 ml of the 1000 mg l-1 cobalt solution, 10 ml of 1 mol l-1 acetic acid–acetate buVer (pH 4.7) and 10 ml of the 0.1% to volume with the same nitric acid solution. A reagent blank was prepared in parallel.The diluted sample (25 ml ) was APDC solution in a separating funnel and adding 50 ml of IBMK. The mixture was then shaken gently for 5 min; after analysed immediately after preparation by inserting it into the manifold of Fig. 1. the two phases had been decanted, the IBMK phase was transferred to a PTFE stoppered bottle containing anhydrous sodium sulfate. Working-strength solutions were prepared Preconcentration procedure daily by appropriate dilution of the above 20 mg ml-1 stock The continuous preconcentration and elution system is shown solution with IBMK. in Fig. 1. In the preconcentration step, a volume of 25 ml of Solutions of potentially interfering ions were prepared by standard or treated sample, containing 0.02–1 ng ml-1 Co in dissolving the amount of each metal/salt needed to obtain a 0.1 mol l-1 HNO3, was continuously introduced into the 100 mg ml-1 concentration of each ion.system at 3.0 ml min-1 and mixed thoroughly with the chelating reagent solution (0.1% APDC) at 0.3 ml min-1. The cobalt Apparatus chelate was adsorbed on the C60 mini-column, located inside the loop of the injection valve (IV1), the sample matrix being A Perkin-Elmer Model 1100-B atomic absorption spectrometer (U� berlingen, Germany) equipped with deuterium arc back- sent to waste.Residual aqueous solution inside the column and FI connectors was flushed by passing a nitrogen stream ground correction, an HGA-700 graphite furnace atomizer and an AS-70 furnace autosampler were used throughout. at 2.0 ml min-1 through the carrier line of the second valve (IV2) for 2 min; simultaneously, the loop of IV2 was filled Measurements were made at 240.7 nm, using a single-element hollow cathode lamp for cobalt that was operated at 30 mA with eluent (IBMK).As IV2 was switched, 500 ml of IBMK were injected into the nitrogen stream and passed through the and a bandwidth of 0.2 nm. Argon at a flow rate of 300 ml min-1 was used as the inert gas except during the column to elute the chelate (position in bold lines in Fig. 1). The extract was collected in the stoppered PTFE cup (500 ml atomization step, where the flow was stopped; the injected volume was 20 ml.Pyrolytic graphite L’vov platforms inserted into pyrolytic graphite coated tubes were obtained from Perkin-Elmer. Background-corrected integrated absorbance was used as the analytical signal. The furnace programme is shown in Table 1. The flow manifold consisted of a Gilson Minipuls-2 peristaltic pump ( Villiers-le-Bel, France) furnished with poly(vinyl chloride) tubes, two Rheodyne 5041 injection valves (Cotati, CA, USA) and laboratory-made sorption mini- Table 1 Graphite furnace temperature–time programme for the determination of cobalt following sorbent extraction in the FI system Time/s Step Temperature/ °C Ramp Hold 1 90 5 15 2 300 10 20 Fig. 1 FI manifold for the on-line preconcentration of cobalt and its 3 1500 6 20 oV-line determination by ETAAS. Bold lines denote lines relevant to 4 2600 0 6 the elution step. IV, Injection valve; W, waste; IBMK, isobutyl methyl 5 2650 1 3 ketone (eluent); GT, graphite tube. 712 J. Anal. At. Spectrom., 1999, 14, 711–716capacity) of the instrument’s autosampler. A blank consisting solution extracted with IBMK prepared in parallel with the organic standards but in the absence of cobalt; the other was of 20 ml of IBMK (0.005 A s) was used. an IBMK solution. Identical results were obtained with both blanks (ca. 0.005 A s), so the IBMK blank (20 ml ) was used Results and discussion in subsequent experiments. There are two methods for the determination of trace amounts Sorption preconcentration system of cobalt in feed by ETAAS following complexation with 2-nitroso-1-naphthol and solvent extraction with xylene26 or The eVect of chemical and flow variables was studied by heptan-2-one.27 In this work, APDC was selected to form a introducing an aqueous standard solution of 0.2 ng ml-1 neutral chelate with cobalt because this is the most widely cobalt into the flow system for ca. 8 min (sample flow rate used chelating reagent for metal enrichment in AAS techniques 3 ml min-1, sample volume 25 ml ) and merging it with a 0.1% and in combination with C60 fullerene it exhibits improved APDC stream.The mini-column (1.0 cm×3 mm id) was consensitivity and selectivity.23–25 Isobutyl methyl ketone is the structed from a PTFE capillary and packed with 80 mg of C60 preferred eluent because the PDC chelate is the easiest to fullerene. For comparison, a similar mini-column packed with dissolve and hence to desorb in this medium.A previously 80 mg of RP-C18 (1.5 cm×3 mm id) was also tested. The reported on-line column preconcentration system based on eluent used was IBMK in all instances. A volume of 500 ml FAAS, with some changes (e.g., the eluent was carried by a (the capacity of the PTFE autosampler cup) was initially stream of nitrogen instead of water), was used.25 Using air as selected. The cup was fitted with a pierced anti-evaporation eluent carrier was initially discarded as cobalt(II) complexes stopper allowing insertion of the flow line in order to minimize are known to be easily oxidized by oxygen to trivalent contamination and losses.complexes that are too inert for elution.28 The eVect of pH on chelate adsorption was studied over the range 0–7 by adjusting the cobalt sample with dilute HNO3 Graphite furnace heating programme or NH3 as required. The optimum pH range (1–5) was wider for C60 fullerene than for RP-C18 (1–1.5), probably because The low surface tension of IBMK makes sample delivery the adsorption constant of the former is greater, consistent diYcult as the ketone tends to creep along the length of the with experimental results obtained for Pb–PDC (adsorption furnace tube, thus severely limiting sample volumes; in constants were 575 and 155 for C60 and RP-C18, respectaddition, the analytical response from organometallic com- ively),23 and Cd–PDC chelates.25 Although the maximum pounds is often diVerent from that of inorganic salts.29 chelate adsorption for RP-C18 was achieved at a low sample Additional problems often posed by IBMK include the fact pH, a further plateau was observed between pH 3 and 5, in that the analytical response varies with the injected solvent addition to a signal decrease by about 25% relative to the first volume and potential pre-atomization losses of volatile metal plateau.It is interesting that the acid zone (lower than 2) is chelates or their decomposition products.In this work, pyro- unusual in liquid–liquid extraction procedures based on this lytic graphite coated graphite tubes with platforms were used ligand; this suggests that the acid medium favours retention in order to restrict penetration of the solvent into the graphite of the chelate on both sorbents. In order to simplify the and spreading of the sample.30 A simultaneous study was operating procedure and increase the selectivity of the method, performed with the Co–PDC chelate in IBMK and aqueous 0.1 mol l-1 HNO3 was used to prepare the samples for both standards of cobalt in 0.2% HNO3.No problems were encoun- mini-columns. The eVect of the APDC concentration was tered in pipetting 20 ml volumes of IBMK into the graphite studied over the range 0.001–0.5%; both mini-columns profurnace. Aqueous and organic standards of 25 ng ml-1 cobalt vided similar results (the preconcentration eYciency gradually were employed in all instances.No significant influence on increased as the APDC concentration was raised to 0.01% m/v sensitivity or precision was observed at drying temperatures and then levelled oV ). For subsequent work, a 0.1% m/v of ca. 110 °C for aqueous solutions; on the other hand, two APDC solution was chosen. Replacing the APDC streawith drying steps at 90 and 300 °C were necessary to avoid splatter- a water stream (the sample was also circulated) resulted in an ing of the organic solutions. The integrated absorbance of area in the elution step similar to that obtained by replacing cobalt remained constant on varying the pyrolysis temperature the sample stream with 0.1 mol l-1 HNO3 (blank), ca.from 500 to 1500 °C in both types of medium. This indicates 0.005 A s, which corresponds to the eluent signal. IBMK was that the thermal stability of Co–PDC in IBMK is similar to thus used as the blank. that of cobalt in the aqueous solution. The high pyrolysis The influence of the sample flow rate on the signal was temperature used (1500 °C) favoured separation of the cobalt examined over the range 0.5–4.0 ml min-1 by using an overall from more volatile concomitants present in the matrix sample, sample volume of 25 ml; the signal remained constant through- thus reducing the background signal.At a pyrolysis temperaout the range studied, so 3.0 ml min-1 was selected. A reagent ture of 1500 °C, peak areas for cobalt increased with increasing flow rate of 0.3 ml min-1 was chosen because higher flow rates atomization temperature up to 2600 °C (with both aqueous resulted in concomitant/sample dilution and hence in decreased and organic standards).On the other hand, the use of mag- atomic signals. The optimum length of the preconcentration nesium nitrate as modifier (15 ml volume of standard and 5 ml coil ( located before the sorbent column) ranged between 100 of 10 g l-1 chemical modifier) provided no advantages with and 300 cm for both mini-columns; a length of 200 cm was aqueous or organic standards.At a pyrolysis temperature of thus used throughout. 1500 °C and an atomization temperature of 2600 °C, the integrated absorbance was 10% higher with Co–PDC in IBMK Elution process standards than with identical amounts of cobalt in 0.2% HNO3. Background absorption was found not to depend on As shown elsewhere,10,12,16 ethanol is the most widely used eluent by virtue of its eVective elution of adsorbed chelates the nature of the cobalt solution medium; thus, the signals were similar for aqueous and IBMK standards and ranged from classical sorbents; it has thus been frequently used with RP-C18 sorbents.As in this paper a parallel study was per- from 0.004 to 0.012 A s. The blank proved the most important single sample to be formed with C60 and RP-C18, we compared ethanol and IBMK as eluents with both sorbents. A volume of 25 ml of a standard carried through the analytical procedure; running two blanks simultaneously with the samples for the organic medium solution containing 0.2 ng ml-1 Co in 0.1 mol l-1 HNO3 was introduced into the FI system of Fig. 1 for this purpose. proved advantageous. One blank consisted of a PDC buVer J. Anal. At. Spectrom., 1999, 14, 711–716 713Following retention, the column was dried with nitrogen and chelate were completely eluted by ethanol, even if the volume needed exceeded that of IBMK as eluent, the analytical signal the retained chelate was eluted by using a nitrogen stream at 2.0 ml min-1 as eluent carrier.The eZuent from the sorbent of cobalt should be similar under the optimum conditions for ethanol with equal volumes of both eluents. Because the column was collected in a PTFE autosampler cup of 500 ml capacity. favourable eVect might lie in boosted sorption (i.e. sorption of the Co–PDC chelate might be increased by conditioning The eVect of the eluent volume was studied between 100 and 500 ml by using loops of variable length in the injection with IBMK), a series of quintuplicate experiments was conducted that provided the following results: (i) when the fullerene valve (IV2 in Fig. 1).In order to study only the desorption process and to correct the opposing eVect of dilution, the column was conditioned with 500 ml of IBMK before the eluate was always diluted to 500 ml with IBMK or ethanol. preconcentration step and further elution with 500 ml of etha- With C60, desorption eYciency increased with increasing nol, the signal was 90% of that obtained when eluting with injected IBMK volume up to 200 ml, above which the analytical 500 ml of IBMK; (ii) when, following conditioning with 500 ml signal remained constant.A second injection of the same of IBMK, the column was flushed with 1 ml of 0.1 mol l-1 eluent volume (200 ml ) revealed the absence of carry-over. The HNO3, elution with 500 ml of ethanol or IBMK provided elution experiments with ethanol revealed that at least 400 ml signals which bore the same relation as in (i), but were both were necessary for complete desorption of the Co–PDC che- 30% lower; and (iii) conditioning with 500 ml of ethanol and late; in addition, as can be seen in Fig. 2, the peak area eluting with IBMK provided results similar to those obtained decreased by ca. 40% relative to IBMK eluent. Fig. 2 also by eluting with ethanol. One can therefore conclude that shows the atomization signal for cobalt following preconcen- adsorption is favoured by conditioning the C60 fullerene tration on C60 and elution with 500 ml of 2mol l-1 HNO3. column with IBMK, probably because its keto group (an Nitric acid resulted in no signal diVerence between the sample electron donor) boosts the sorption capacity of fullerenes and blank.IBMK provided the better results (diVerence (electron acceptors), which seemingly act via p-electron interbetween sample and blank); also, the background signal was actions.21 Alternatively, the water immiscibility of IBMK negligible and similar for the three eluents tested.A second solutions may give rise to the formation of a film within the study was performed by using RP-C18 as sorbent under the column that might subsequently facilitate the adsorption of above-described conditions. The better eluent in this case was the Co–PDC chelate; however, this eVect is not observed with ethanol, which resulted in an atomic signal 15% higher than the RP-C18 sorbent, so the previous explanation seems the with IBMK.In addition, the signal for 0.2 ng ml-1 cobalt more plausible. On the other hand, if the potential bonding obtained with RP-C18 (and ethanol as eluent) was similar to mechanism for the adsorption of Co–PDC on RP-C18 is based that obtained with C60 (with IBMK as eluent). Therefore, on a hydrophobic eVect, then the increased eluting capacity calibration graphs with RP-C18 were constructed by using of ethanol can be ascribed to the chelate desorption being ethanol and IBMK as eluents (500 ml ).The sensitivity (slope governed by a reduced solvent polarity. of the calibration graphs) for ethanol and IBMK was 0.40 Finally, in order to optimize the performance of the and 0.35 A s per ng ml-1, respectively. proposed method, an injected volume of 500 ml of IBMK was The above experiments with C60 fullerene provided better used in order to ensure complete elution of the chelate and results with IBMK than with ethanol as eluent (see Fig. 2). conditioning of the column before the aqueous sample was However, this is diYcult to explain because, if the Co–PDC preconcentrated. Flow rates of the nitrogen stream (the carrier of the eluent volume) between 1 and 3 ml min-1 had no eVect on the cobalt signal. No blank was required; the sample signal was calculated by subtracting the eluent signal (20 ml of IBMK) obtained before the sample extract was injected into the graphite furnace tube.Air, introduced via the peristaltic pump, can be used to dry the sorbent column and FI connections, and also as the eluent carrier as it provides results similar to those of nitrogen, probably because the Co–PDC chelate is not oxidized. Analytical performance Several calibration graphs were run by using mini-columns packed with 80 mg of C60 or RP-C18 and IBMK as eluent. Aqueous standard solutions were processed along the preconcentration flow system depicted in Fig. 1. The sensitivities (expressed as the slopes of the calibration graphs) were 0.40±0.02 and 0.35±0.03 A s per ng ml-1 and the linear ranges for cobalt were 0.02–1 and 0.05–1 ng ml-1 for C60 and RP-C18, respectively, equivalent to a sample volume of 25 ml at a sampling time of ca. 8 min. The detection limit was calculated as three times the standard deviation of the peak area for 15 injections of 20 ml of IBMK (blank) and was found to be 8 and 20 ng l-1 for C60 and RP-C18, respectively.The precision (as relative standard deviation) was checked on 11 standard solutions containing 0.2 ng ml-1 cobalt with both Fig. 2 Atomization signals for 0.2 ng ml-1 Co after on-line mini-columns. It was higher for C60 (4%) than for RP-C18 preconcentration of Co–PDC chelate on C60 fullerene sorbent, using (7%). The preconcentration factor (viz. the ratio between the 500 ml of diVerent eluents: (a) IBMK, peak area (AA-BG)= slope of the calibration graphs provided by the proposed 0.078 A s, peak area (BG)=0.011 A s; (b) ethanol, peak area method and by direct insertion of Co–PDC standards in (AA-BG)=0.048 A s, peak area (BG)=0.010 A s; and (c) 2 mol l-1 HNO3, peak area (AA-BG)=0.002 A s, peak area (BG)=0.009 A s.IBMK) was 40 and 35 for C60 and RP-C18, respectively. 714 J. Anal. At. Spectrom., 1999, 14, 711–716Table 3 Cobalt content (ng g-1±s) in various wheat flour samples as Table 2 Tolerated concentrations of foreign cations in the determination of 0.2 ng ml-1 cobalt with APDC and C60 sorbent determined by the proposed FI method (n=9) and directly in mineralized samples (n=6) Tolerated Ion [metal ]5[Co] ratio Sample FI-ETAAS Conventional ETAAS 1 11.0±0.9 9.5±0.7 Al3+ 10 000 Mn2+ 10 000 2 19.9±0.7 20.1±0.5 3 18.3±0.8 19.1±0.6 Sn2+ 10 000 Hg2+ 8000 4 16.6±0.8 15.5±0.5 5 16.3±0.7 15.5±0.5 Zn2+ 7000 Bi3+ 2000 6 18.2±0.7 19.0±0.6 7 13.3±0.6 12.8±0.4 Ni2+ 700 Cu2+ 400–1000a 8 13.8±0.6 13.5±0.4 Cd2+ 300 Fe3+ 100–8000b Pb2+ 100 a,bTolerated ratio with 0.02% m/v neocuproine and 4% m/v replicate are given in Table 3.For comparison, identical hydroxylammonium chloride, respectively. samples (No. 1, 2 and 5 in Table 3) were pre-treated by ashing and analysed by direct ETAAS using aqueous standards of cobalt. For this purpose, about 5 g of flour were charred in a Interference testing burner for ca. 2 h and then ashed at 650 °C for 3 h. The results were consistent with those provided by the proposed FI The influence of metals that might react with APDC and method; however, the procedure was more time-consuming replace cobalt in the original chelate was investigated in order and used a greater amount of sample.The proposed method to identify potential interferences. Table 2 lists the cations is, therefore, suitable for the determination of cobalt in this examined and their tolerated ratios in the determination of type of sample; the sample preparation time is minimal (ca. 0.2 ng ml-1 cobalt with C60 at pH 1; the maximum concen- 10 min) as the preconcentration method aVords the processing tration tested was 10 000 times that of the analyte. Interferents of small amounts of sample (ca. 0.25 g) that can be readily decreased the cobalt signal in all instances by competing for mineralized within a short time. and consuming the reagent; as a result, uncomplexed cobalt was not retained on the column. As can be seen in Table 2, no interference was caused by Al3+, Mn2 + or Sn2+ at the Conclusions highest concentration tested (2 mg ml-1).Only Pb2+ and Fe3+ The ETAAS determination of trace amounts of cobalt in interfered at concentration ratios up to 100. Unless the samples organic solvents warrants several interesting comments. Thus, are highly contaminated, only Cu2+ and Fe3+ are encountered the complexing agent (APDC) and the solvent (IBMK) have at concentrations of ca. 6 and 70 mg g-1, respectively, in flour negligible eVects relative to aqueous standard solutions (peak samples; such contents interfere with the proposed method.areas are 10% higher in the organic medium and background Masking agents may thus be required to avoid the interference signals are similar and negligible). The analytical potential of of high concentrations of Fe3+ and Cu2+ in analysing wheat fullerene as a sorbent in ETAAS was studied here for the first flour. We examined the eVectiveness of several commonly used time.Fullerenes perform better in metal preconcentration than masking agents. Sodium citrate (8% m/v), tartaric acid (4% do conventional C18-bonded silica sorbents when using FAAS m/v) and Tiron (10% m/v), which forms a highly stable and IBMK as eluent;23–25 however, water-miscible solvents complex with Fe3+ in an acidic medium, were tested as (e.g., ethanol ) should be avoided in order to minimize dilution masking agents for iron, with unfavourable results at high while water is being driven to the nebulizer.In this work, iron concentrations. Since Fe2+ has a lower aYnity for APDC eluents were carried by a nitrogen (or air) stream, and IBMK than Fe3+, reductants such as hydroxylammonium chloride and ethanol were tested in parallel with C60 and RP-C18 and ascorbic acid were tested. Hydroxylammonium chloride sorbents. The results aVord some interesting conclusions. First, and ascorbic acid at levels up to 4% m/v increased the amount the sensitivity is higher with C60 than with RP-C18 when using of tolerated iron to 1.6 mg ml-1 (Fe5Co ratio 800051).IBMK as eluent; however, RP-C18 is similar to C60 when using Bathocuproine and neocuproine at variable concentrations ethanol instead of IBMK. Second, adsorption is increased by were tested to overcome the interference of copper. Only conditioning C60 fullerene with IBMK before the preconcen- neocuproine was found to complex Cu2+ eVectively (in the tration step, probably as a result of the keto group in IBMK presence of a reductant such as hydroxylammonium chloride) favouring the adsorption mechanism via p-electron inter- with no disturbing eVect on cobalt. As can be seen in Table 2, actions.Future experiments will be conducted to identify, in 0.02% m/v neocuproine suppressed the interference of Cu2+ more rigorous terms, the best conditions (e.g., chemical vari- at levels at least up to 1000 times higher than that of cobalt, ables, selectivity) for RP-C18 and Co–PDC, using ethanol as which suYced since copper normally occurs at concentrations eluent.If similar figures of merit are obtained under the only ca. 600 times higher than that of cobalt in flour. optimum conditions for each sorbent (C60 and RP-C18), then the sorbent of choice will be the more aVordable and readily Analysis of wheat flour available. Because no certified values for the cobalt content in wheat flour reference materials were available, the applicability of Acknowledgements the proposed method was checked by analysing wheat flour samples from diVerent Spanish locations.These materials were This work was supported by grant PB95–0977 from Spain’s dissolved as described under Experimental and the cobalt DIGICyT. content in the resulting solutions was determined by using the recommended procedure. The signals for the blanks, which References contained the reagents in the digested sample, corresponded to a cobalt concentration below the detection limit.The results 1 World Health Organization, Trace Elements in Human Nutrition and Health, WHO, Geneva, 1996. obtained for three sample replicates and three injections per J. Anal. At. Spectrom., 1999, 14, 711–716 7152 A. Mizuike, Enrichment Techniques for Inorganic Trace Analysis, 17 H. W. Kroto, J. R. Heath, S. C. O’Brien, R. F. Curl and Springer-Verlag, Berlin, 1983. R. E. Smalley, Nature, (London), 1985, 318, 162. 3 Flow Injection Atomic Spectroscopy, ed. J. L. Burguera, Marcel 18 J. F. Anacleto and M. A. Quilliam, Anal. Chem., 1993, 65, 2236. Dekker, New York, 1989. 19 K. Jinno and H. Nakamura, Chromatographia, 1994, 39, 285. 4 Z. Fang, Flow Injection Atomic Absorption Spectrometry, Wiley, 20 K. Jinno, C. Okumura, M. Harada, Y. Saito and M. Okamoto, New York, 1995. J. Liq. Chromatogr., 1996, 19, 2883. 5 Z. L. Fang and G. H. Tao, Fresenius’ J. Anal. Chem., 1996, 355, 21 M. S. Dresselhaus, G. Dresselhaus and P. C. Eklund, Science of 576. Fullerenes and Carbon Nanotubes, Academic Press, San Diego, 6 B. Welz, Microchem. J., 1992, 45, 163. CA, 1996. 7 Z. L. Fang, S. Xu and G. H. Tao, J. Anal. At. Spectrom., 1996, 22 H. Rausch and T. Braun, Anal. Chem., 1997, 69, 2312. 11, 1. 23 M. Gallego, Y. Petit de Pen�a and M. Valca�rcel, Anal. Chem., 1994, 8 Z. L. Fang, M. Sperling and B.Welz, J. Anal. At. Spectrom., 1990, 66, 4074. 5, 639. 24 Y. Petit de Pen� a, M. Gallego and M. Valca�rcel, Anal. Chem., 1995, 9 M. Sperling, X. Yin and B. Welz, J. Anal. At. Spectrom., 1991, 67, 2524. 6, 295. 25 Y. Petit de Pen� a, M. Gallego and M. Valca�rcel, J. Anal. At. 10 M. Sperling, X. Yin and B. Welz, J. Anal. At. Spectrom., 1991, Spectrom., 1997, 12, 453. 6, 615. 26 O. K. Borggaard, H. E. M. Christensen and S. P. Lund, Analyst, 11 M. Sperling, X. Yin and B. Welz, Spectrochim. Acta, Part B, 1991, 1984, 109, 1179. 46, 1789. 27 W. J. Blanchflower, A. Cannavan and D. G. Kennedy, Analyst, 12 R. Ma and F. Adams, Anal. Chim. Acta, 1995, 317, 215. 1990, 115, 1323. 13 Y. H. Sung, Z. S. Liu and S. D. Huang, J. Anal. At. Spectrom., 28 K. Isshiki and E. Nakayama, Anal. Chem., 1987, 59, 291. 1997, 12, 841. 29 R. E. Sturgeon, S. S. Berman, A. Desaulniers and D. S. Russel, 14 M. Sperling, X. P. Yan and B. Welz, Spectrochim. Acta, Part B, Talanta, 1980, 27, 85. 1996, 51, 1891. 30 E. Tserovsky and S. Arpadjan, J. Anal. At. Spectrom., 1991, 6, 15 E. Ivanova, K. Benkhedda and F. Adams, J. Anal. At. Spectrom., 487. 1998, 13, 527. 16 X. P. Yan, W. Van Mol and F. Adams, Lab. Rob. Autom., 1997, 9, 191. Paper 8/08095J 716 J. Anal. At. Spectrom., 1999, 14, 7