Development of an analytical scheme for the direct determination of antimony in geological materials by automated ultrasonic slurry sampling-ETAAS M. Jesu�s Cal-Prieto, Alatzne Carlosena,* Jose� M. Andrade, Soledad Muniategui, Purificacio�n Lo� pez-Mahý�a, Esther Ferna�ndez and Darý�o Prada Department of Analytical Chemistry, University of La Corun�a, Campus da Zapateira s/n, E-15071 La Corun�a, Spain. E-mail: alatzne@udc.es Received 21st September 1998, Accepted 7th January 1999 The ongoing application of ultrasonic probes for slurry homogenization gives new additional advantages with respect to the traditional slurry analyses by ETAAS, namely speed, automation, improved reproducibility, etc. Despite all these advantages, there are still diYculties in the optimization stages of the analytical procedures because of the large number of variables to take into account.Accordingly, several chemometric techniques were applied to help in the development of a systematic analytical scheme to determine Sb directly in soils and sediments by means of ultrasonic slurry sampling (USS)-ETAAS. They are intended not only to optimize the slurry preparation but also to diminish the amount of benchwork to be done.The chemometric techniques and their objectives were as follows: Plackett–Burman designs, to assess the influence of six analytical variables; optimization of the important analytical variables; a two-way ANOVA, to investigate the influence of sample cups and replicates; and control charts, to monitor the graphite tube performance.Additionally, two pipetting options and three quantification methods were evaluated. Analytical results for Sb determined in five certified reference materials confirmed the usefulness of the ultrasonic slurry sampler (USS-100) combined with ETAAS for a direct and almost fully automatic analysis of complex matrices. determination of this element in several environmental samples Introduction such as soils, sediments and air.Additionally, this technique Traditionally, very little attention has been paid to antimony in is increasingly being used in the direct analysis of solid samples soils, sediments and biomaterials, mainly because it is not either with solid sampling or slurry sampling. The main recognized as having nutritional significance and its contents in benefits of direct solid analysis by ETAAS are reduced time most materials are very low.Nevertheless, compounds of anti- and sample preparation requirements, decreased likelihood of mony are moderately toxic to most organisms, Sb3+ compounds analyte loss prior to analysis, minor use of acids and their being more toxic than Sb5+ compounds (similarly to the corresponding reduction in the production of wastes and a corresponding arsenic compounds).1 Exposure to antimony lower probability of sample contamination. During the 1990s, compounds causes cellular damage in the lungs, heart and the use of slurry sampling-based analytical procedures has kidneys, but the toxicity mechanisms are still not well known.2 been gaining acceptance to cope with diVerent types of samples, Many industrial processes such as the manufacture of alloys, such as soils and sediments,7–14 thanks to several practical paints, compact discs and bactericides involve the use of anti- advantages.Some of them are as follows: (i) conventional mony compounds which appear in most industrial eZuents.liquid autosamplers can be used; (ii) several replicates can be Furthermore, antimony may reach soils by wet and dry depos- made in just one aliquot; (iii) handling of higher masses of ition following emissions from incineration and fossil fuels sample; (iv) ability to change the slurry concentration; (v) combustion, vehicle traYc and by the addition of soil amend- better analytical performance can be achieved since the slurry ments such as chemical fertilizers, sewage sludge and fly-ash.3,4 technique combines the benefits of solid and liquid sampling; Antimony in air is considered to be associated with the smaller and (vi) improved repeatability and reproducibility.Moreover, particles, inhalation being the major cause of antimony intoxi- the development and application of an ultrasonic probe cation. The occurrence of this element in airbone dust and its (Perkin-Elmer USS-100)15–21 for slurry homogenization feacorresponding deposition and accumulation in vegetation were tured new additional advantages such as higher speed, autostudied by Dietl et al.,4 who confirmed antimony enrichment mation of slurry sample preparation and its introduction into near roadways by monitoring it in standardized grass cultures the furnace, improved reproducibility and analyte extraction and total depositions.Moreover, the antimony deposition might and improved representativeness of the aliquot analysed.also aVect the antimony cycling in the oceans.In that sense, Despite all these advantages, there are diYculties in the Takayanagi et al.5 investigated vertical profiles of dissolved optimization stages of the analytical procedures because of inorganic antimony in sea-water, reflecting higher concen- the large number of variables to take into account: mass of trations on the surface which decreased to a near constant level sample, volume of slurry, mixing time, power output of the with depth. Despite this, the natural low levels of antimony in ultrasonic probe, use of thixotropic agents, chemical characterthe earth’s crust (0.2–0.3 mg g-1)6 make it a potential marker istics of the diluent, number of replicates per sample, number of environmental metallic contamination.of particles present in the injected sample volume, etc. The good sensitivity of electrothermal atomic absorption Although the classical schemes of work (many of them ‘trial and error’ type) have given good results in previous method spectrometry (ETAAS) makes it a suitable technique for the J.Anal. At. Spectrom., 1999, 14, 703–710 703developments, few attempts have been made to incorporate a polyethylene autosampler cup and suspending it in a small chemometric techniques not only with optimization purposes volume of diluent (typically 0.5 or 1 ml ). Then, an aliquot is but also as a way to diminish the amount of work to be done. injected into the graphite furnace.The mass of the powder In this way, Koch et al.22 have shown an excellent approach will depend on the total amount of Sb that the sample is to deal with these kind of developments focused on the final expected to present. The higher the content, the lower is the measurement step. Here, the following topics are applied to sample mass. There are practical limits for both the minimum achieve a reliable and optimized procedure for the direct and maximum amounts of sample (see Results and discussion) determination of Sb in soils and sediments by ultrasonic slurry that can be considered (the latter situation is mostly caused sampling (USS)-ETAAS: by the relatively poor eYciency of conventional autosamplers (i) a Plackett–Burman saturated factorial design23 to deter- when used for the aspiration of a concentrated slurry aliquot). mine the influence of six analytical variables on the preparation The experimental conditions were carefully optimized for of the slurries; slurry preparation whereby sample aliquots (2–150 mg) were (ii) optimization of the critical variables; directly weighed in a polyethylene cup and suspended in 1 ml (iii) a two-way ANOVA to investigate the influence of both of 0.5% v/v HNO3 and the ultrasonic power setting and the the number of sample cups and number of replicates per agitation time were set at 40W and 10 s, respectively to obtain sample cup; levels for cups are 1, 2, 3 and 4 and levels for homogeneous slurries. The major steps are depicted in Fig. 1, replicates are 2, 3, 4 and 5; which illustrates the straightforward operations required by (iv) comparison of three quantification methods using the this analytical approach. Hotelling confidence intervals associated with the calibration Three roadside soil samples and three marine sediment line;24 samples were collected. Soil samples were from a metrol charts to monitor the behaviour and ageing of city, La Corun� a, NW Spain, at a 0–5 cm depth; two of them the pyrolytic graphite tubes (with a L’vov platform).were from city gardens and the other was from a crop soil. At every sampling site, several sub-samples were collected from a Experimental 0.5 m2 grid area and mixed to obtain one bulk sample. Regarding the sediments, they were collected from the La Apparatus Corun� a estuary in coastal areas. Sampling was made with a The determination of Sb was performed on a Perkin-Elmer ‘box-corer’ probe.The superficial sediment layer (0–10 cm) was (U� berlingen, Germany) Model 4100 atomic absorption spec- taken. Samples were lyophilized (-40 °C) for 36–48 h until trometer equipped with deuterium background correction, an constant mass, then sieved using a 2 mm mesh sieve and ground. HGA-700 graphite furnace, an AS-70 autosampler and a Accuracy and precision were assessed by using five certified USS-100 ultrasonic slurry sampler. Argon was used to provide reference materials, namely two NCR CRM (National the inert atmosphere within the furnace.Pyrolytic graphite- Research Council of Certified Reference Materials, China) coated graphite tubes (Z-tek, Amsterdam, The Netherlands) soils (GBW07401 and GBW07409), one NCR CNRC with preinserted pyrolytic L’vov platforms were employed. (National Research Council of Canada) marine sediment Atomization was accomplished by using the maximum power (BCSS-1) and two BCR CRM (Community Bureau of heating and gas stop flow.Peak heights were recorded and Reference, Brussels, Belgium) materials (a calcareous loam used to quantify the Sb analytical signal. Measurements were soil, CRM 141, and an estuarine sediment, CRM 277). made by using a hollow cathode lamp (Perkin-Elmer) at the 217.6 nm Sb wavelength, using a 0.7 nm spectral bandwidth. Analytical procedure For the statistical treatment the Statgraphics software package Sb was determined directly by using automated USS-ETAAS.was used. Slurry aliquots (5–20 ml ) were injected after ultrasonic agitation, then dried using two steps: 100 °C (5 s ramp, 10 s hold) Reagents and 130 °C (30 s ramp, 10 s hold). The optimum pyrolysis and All reagents were of analytical-reagent grade. High-purity atomization temperatures were found to be 900 °C (10 s ramp, water (Milli-Q Water System, Millipore, Madrid, Spain) was 30 s hold) and 1700 °C (0 s ramp, 2 s hold), respectively. Two employed throughout. The acids (HNO3 and HCl) were of slurries were prepared for each sample and they were analysed Suprapur grade (Merck, Darmstadt, Germany).An Sb stock in triplicate. Quantification was performed using peak height standard solution (1.000 g l-1) for atomic absorption was measurements and the method of addition calibration obtained from Panreac (Barcelona, Spain) and working Sb implemented in the spectrometer management software. The standard solutions were prepared daily by diluting appropriate slurry, Sb spike and diluent were introduced into the furnace aliquots of the commercial stock standard solution.Standard by the autosampler using the ‘pipette separate’ mode. The aqueous solutions of the chemical modifiers containing total volume injected was 20 ml. 10 000 g l-1 of both magnesium and ammonium dihydrogenphosphate were prepared from Mg(NO3)2 (99.995%) Results and discussion (Aldrich, Milwaukee, WI, USA) and NH4H2PO4 (99.999%) (Aldrich), respectively.A palladium solution was made by Results are presented for studies carried out to develop and dissolving Pd powder (99.999%) (Aldrich) in the minimum validate the analytical method intended to determine Sb in volume of aqua regia, followed by evaporation to near dryness geological materials by USS-ETAAS. The flow chart depicted and dilution to a final concentration of 1.000 g l-1 with in Fig. 2 shows each major analytical step and the associated 0.5% v/v HNO3. The same procedure was followed to prepare studies.This flow chart is not ‘strict’; this means that the order nickel solution. of each test or particular study might be changed to other All glassware, plasticware, pipette tips and storage bottles sequences. Such diYculty arises as a direct consequence of the were soaked in 10% v/v HNO3 for 24 h and rinsed with high- intrinsic complexity of the slurry-based analytical techniques. purity water at least three times prior to use. Sometimes, several tests have to be performed to establish some particular conditions, but which is first? Hence, this Slurry and sample preparation scheme represents an approach to the development of analytical methods to determine trace metals directly by slurry Grossly speaking, the preparation of the sample for the sampling-ETAAS and it resembles a general-purpose pro- ultrasonic slurry sampling using the USS-100 probe consists in weighing a certain amount of sample powder directly into cedure already published.17 704 J.Anal. At. Spectrom., 1999, 14, 703–710Fig. 1 System used for the direct and almost fully automatic analysis of soils and sediments by ultrasonic slurry sampling-ETAAS for the determination of Sb. media, mixed using the USS-100 probe and the Sb content determined in the supernatants (separated by centrifugation). In all cases, the Sb extraction into the liquid phase was negligible. This reveals that the use of high concentrations of nitric acid is not essential (as usually recommended for other metals12,16,17).Also, this will increase the lifetime of the graphite tube. Therefore, the representativeness of the analysed slurry aliquot becomes more critical. Selection of the critical variables. Regarding the slurry preparation itself, there are a large number of variables to deal with9,18 and, also, diVerent ways to study them simultaneously to evaluate their influence on the analytical response, e.g., simplex22 and evolutionary operation procedures (EVOP), Latin squares and experimental designs.22 Of these, the Plackett–Burman partial factorial designs are very appealing since they allow the maximum extraction of information with a minimum level of laboratory work.They have also been proposed by ASTM26 as a simple way to obtain ‘rugged’ analytical methods. Accordingly, one of the Plackett–Burman Fig. 2 General flow chart of the main steps needed to carry out designs was chosen to elucidate which variables are important analyses by slurry sampling-ETAAS showing the main statistical studies associated with each. (and to what extent) when preparing slurry solutions.Table 1 presents the variables considered in this study and also the experimental conditions for each trial, the correspond- The inherent diYculties with this analytical technique call ing experimental absorbance and the calculated eVect for each for an iterative process in order to optimize the method but variable (experimental t-value). All the calculations were made this is highly expensive and time- and labour-consuming, so it as usual.was only applied when strictly necessary. Taking the tabulated t-value (95% confidence) as 1.94, it can be observed that only the F factor (time of agitation) was Slurry preparation statistically significant. Nevertheless, the relatively high t- Stability of the slurry. Precision is highly dependent on the values for factors A (sample mass) and B (slurry volume) homogeneity of the analyte distribution in the sample and make it interesting to consider both of them in some more analyte partitioning between the solid and liquid phases.16 detail (although they are not statistically significant). When the analyte was extracted easily into the liquid phase Moreover, the confounding eVect most likely to happen is the (which is improved by ultrasonic agitation and acid content), interaction between variables A–B and variable F.This means slurry stability became less critical. that varying both factors, the mass of sample and the volume Accordingly, to study the slurry stability, various slurries of of slurry, may have the same eVect as varying the time of the same sample were prepared (50 mg ml-1 in 0.5% v/v agitationover, the negative sign means that the HNO3) using the USS-100 probe at a power setting of 40 W absorbance diminishes when the time of agitation increases for 10 s.Several aliquots were taken periodically and measured or, alternatively, when the mass of sample and volume of (homogenization was carried out before the first sampling). slurry increase. Of course, both variables point towards the The average absorbances showed that after 185 s the so-called mass/volume ratio so important in slurry studies.19 absorbance dropped from 0.158 to 0.003 owing to the high The explanation for the negative sign is that when the agitation sedimentation rate after stopping agitation and the small to time is too long (alternatively, the mass/volume ratio), a trend negligible amount of analyte extracted into solution.to form emulsions in the autosampler cup was observed. Unfortunately, this is the minimum time to make two consecutive injections using the AS-70 autosampler and the furnace Optimization of the selected variables. Following the experimental design, it seems that there is only one variable to be programme.Hence the need for ultrasonic agitation prior to each measurement was clearly established (see also Miller- optimized (time of agitation) and, therefore, a classical univariate scheme is adequate. All the variables not statistically Ilhi25 for analogous results for materials with high densities). To evaluate the analyte extraction eYciency, diVerent HNO3 significant were fixed at values considered either in the experimental design or within the experimental range, as follows: concentrations (from 0.5 to 5% v/v) were assayed as suspension J.Anal. At. Spectrom., 1999, 14, 703–710 705Table 1 Experimental conditions for each trial of the experimental design and statistical eVect calculated for each variable (experimental t-value, 95% confidence interval ) Variable Mass/ Volume/ [HNO3] [Triton X-100] Power/ Time/ Experimental Experiment mg ml (%) (%) W s ‘Dummy’ absorbance 1 100 1 5 0 30 10 — 0.00517 2 50 1 5 0.04 50 10 — 0.00428 3 50 0.5 5 0.04 30 30 — 0.00656 4 100 0.5 0.5 0.04 50 10 — 0.00447 5 50 1 0.5 0 50 30 — 0.00532 6 100 0.5 5 0 50 30 — 0.00523 7 100 1 0.5 0.04 30 30 — 0.05080 8 50 0.5 0.5 0 30 10 — 0.00642 Statistical eVect— Variable Mass Volume [HNO3] [Triton X-100] Power Time ‘Dummy’ EVect -0.00065 -0.00071 -0.000014 -0.00044 0.000126 -0.00097 0.000463 texp 1.38 1.52 0.03 0.94 0.27 2.07 1 mass=50 mg; volume=1 ml; Triton X-100=not used; L’vov platform.Therefore, the upper limit for the sample mass is well established (150 mg), and it seems reasonable to assume HNO3=0.5% v/v; and power=40 W.Fig. 3 shows that after a minimum agitation time of 5 s, no influence was observed that the lower limit will depend on the Sb content of the samples (ensuring a representative number of particles in the up to 30 s. This surprising result strongly suggests that the mass/volume ratio is the variable that needs more attention, furnace). The upper limit achieved here is much larger than those usually reported when slurries are prepared directly in so its optimization is required (agitation time fixed at 10 s).Hence the mass/volume ratio was studied as usual in the the sample cup, which allows one to improve the sensitivity and detection limits. Small masses of another certified reference related literature.27 Seven levels of mass from the NCR CRM GBW07401 soil material (BCR CRM 277 estuarine sediment; 3.9±1 mg g-1 Sb content, not certified) were assayed and the results obtained sample were suspended in 1 ml of the diluent and then analysed.For each level two samples were prepared, three were in good concordance with the certified value. There is another important factor to deal with, which is replicates per sample. No diVerences were observed in the Sb response employing masses from 20 to 150 mg (Table 2). In that the mass/volume ratio may also limit the volume of slurry to be introduced into the furnace. Table 3 summarizes the contrast, high and low responses were obtained when 10 and 200 mg were used, respectively.When 200 mg of sample were results obtained for several masses of the certified reference soil GBW07401, injecting diVerent volumes of each one. An weighed, the high concentration of the slurry led to ineYcient pipetting of the aliquot, whereas 10 mg of sample gave poor important diminution in absorbance can be observed when 200 mg are used. Both the mean value and associated RSD results owing to the small amount of Sb introduced onto the became negatively aVected to a large extent, as stated previously.Such a diminution becomes much greater as the volume of slurry injected in the furnace increases. From the foregoing, the following operational guidelines can be established: (i) samples with moderate Sb contents (2–<20 mg g-1) would be analysed by diluting the sample, either by using low masses (2–10 mg) or by injecting small volumes of slurry into the furnace (5–10 ml ); for higher concentrations, less sensitive non-resonance lines may be useful;9,17,18 Fig. 3 Peak height absorbances recorded after applying diVerent ultrasonic homogenization times for a soil slurry (0.5% v/v HNO3; Table 3 Relationship between the pipetted volume of slurry and the Tpyrolysis=700 °C; Tatomization=1700 °C; power=40 W). sample mass used to prepare 1 ml of soil slurry (total volume injected fixed at 20 ml using 0.5% v/v HNO3) Table 2 Analytical response and limits of detection (LOD) and quanti- Sample mass/mg fication (LOQ) as a function of the soil sample mass employed to prepare 1 ml of slurry (volume injected=20 ml ) 20 50 200 Sample mass/ Corrected LOD/ LOQ/ Slurry volume/ml Corrected absorbancea (n=3) mg absorbancea (n=3) mg g-1 mg g-1 5 7.1 6.9 5.3 2 — 0.620 2.06 10 6.2 7.0 4.4 10 7.0 0.124 0.41 15 6.8 6.6 3.6 20 6.3 0.062 0.21 20 6.3 6.7 2.6 40 6.4 0.031 0.10 x: 6.6 6.8 4.0 50 6.7 0.025 0.08 RSD (%) 6.4 2.7 28.9 92 5.7 0.013 0.04 150 5.8 0.008 0.03 aCorrection made for each volume injected and sample mass 200 2.6 — — Aabsorbance×103 mg × total volume/ml volume injected/mlB. aCorrection made for each sample mass (103 absorbance mg-1). 706 J. Anal. At. Spectrom., 1999, 14, 703–710(ii) samples with low Sb contents (<0.5 mg g-1) would be Therefore, it was decided to check the influence of changing processed either by weighing high masses of sample (<150 mg) the graphite tube (i.e., ‘tube 1’ and ‘tube 2’) on the absorbance.or injecting large volumes of slurry (20 ml ). The results obtained for the 2-way ANOVA are summarized in Table 4 considering absorbance values (they were also made Representativeness of the aliquot injected into the graphite considering concentrations and all the results agreed fairly furnace well ). The eVect of changing the number of replicates per sample cup is statistically significant, which can be verified An important concern when direct analyses of solid samples from Fig. 4, where all the confidence intervals (95%) for the are performed by ETAAS is the representativeness of the average values for each level overlap substantially except for aliquot introduced into the graphite furnace.8 The representathat corresponding to the two replicates level. Accordingly, tiveness of the aliquot will determine both the accuracy and three replicates should be performed to improve the precision precision of the analytical results, and it depends on the of the within-sample measurements and to avoid the slight homogeneity of the analyte distribution in the sample and the negative trend associated with performing only two replicates.analyte partitioning between the solid and liquid phases of the No special benefits are obtained with more than three repli- slurry.16,21 Taking into account that extraction of Sb into cates. These results reflected that the automatic ultrasonic the liquid phase was negligible, it seems mandatory to evaluate mixing provided a uniform distribution of particles in the the representativeness of the analysed slurry aliquot.autosampler cup and that the autosampler exhibited a low performance variability. Hence, the amount of sample ali- Number of particles. It is accepted that at least 50 individual quoted by the autosampler from the slurry is representative of particles of the sample should be present in the aliquot injected the slurry itself. into the graphite tube to avoid statistical limitations of the Regarding the number of sample cups, this factor was not sampling techniques.18 statistically significant (see Table 4).Therefore, it can be The sample powder density and the particle size, assuming deduced that Sb exhibits a homogeneous distribution in spherical particles, allow the determination of the number of the studied sample (certified reference soil GBW07401). particles present in an aliquot. The method previously proposed Nevertheless, it was decided to use two sample cups in all for the determination of density18 was found not to be well measurements combined with three replicates.One sample cup suited for the materials covered in this work because the could be adequate although the variance of the results might reproducibility was poor. Consequently, a standardized pycbe too high. These results agree with previous considerations nometer was used. The densities of the samples and soil and for slurry-sampling analyses of CRMs, although when real sediment standards were found to be around 2.6 g cm-3, which samples are slurry analysed this should be confirmed owing to agrees with densities found for other certified soils and their more questionable homogeneity.16,17 Attention should be sediments.8,18 drawn to the slightly negative trend observed in Fig. 4 when The particle size was determined by means of an optical four sample cups were considered. This could be attributed to microscope, placing 20 ml of a slurry (10 mg ml-1) on a slide the tube ageing, which decreases the analytical signal at the and drying it in an oven at 60 °C.A small area of the drop end of its lifetime. was examined to determine the number of particles and their Now, the eVects caused by the ageing of the graphite tube maximum and minimum diameters. The total area occupied can be studied by means of an additional one-way ANOVA. It by the drop was also determined to estimate the total number was considered that this factor should not be combined with of particles.For the GBW07401 CRM soil, the diameters were the previous two variables to develop a three-way ANOVA between 93.3 and 0.15 mm and the estimated number of because it is not current practice to analyse one sample using particles was 2912. This procedure was repeated for all the diVerent tubes. Nevertheless, the tube seemed an important certified materials and samples and similar results were factor when batches of samples are to be analysed.As expected, obtained. In all cases, the number of particles was much larger the factor ‘tube’ was significant and it caused large changes in than 50, indicating that after slurry homogenization using the the absorbances (Fig. 4). This confirms the need to maintain ultrasonic probe, the aliquot can be said to be representative strict control over the graphite tube performance and to recali- of the weighed sample.brate and validate carefully the results obtained after each change of tube. Moreover, it is important to take into account Number of replicates and sample cups. Despite the importance that faulty packs of tubes could be delivered and hence pur- of this topic, there is little agreement when dealing with slurry chased. Unfortunately, the influence of such tubes on the suspensions. Miller-Ihli17 recommended five replicates in each measurements is not always equal and their eVects depend upon of five sample cups for each sample. Here, it was thought that which element is being measured.One of the more straightfor- some kind of ‘optimization’ could be made to ascertain the ward and eYcient ways of checking tube performance is to adequacy of such a rule. Hence a two-way ANOVA was develop and maintain control charts (e.g., by using either an designed where ‘number sample cups’ ( levels=1, 2, 3, 4) and aqueous standard or a slurry sample).In this work, the control ‘replicates per sample cup’ ( levels=2, 3, 4, 5) were the factors. chart for an aqueous Sb standard (20 ng ml-1) revealed that The eVect of changing the graphite tube used for the analyses up to 350 atomization cycles can be made without a significant was also studied as a factor because one of the typical problems loss of sensitivity. Several control charts for soil slurries (10 ml associated with the slurry-based determination of metals by ETAAS is the accelerated degradation of the L’vov platform.of 50 mg ml-1 solution) allowed 90–100 atomization cycles to Table 4 Two-way ANOVA considering number of sample cups and number of replicates Source of Sum of Degrees of Mean Statistical variation squares freedom square F-ratio diVerence (95%) Sample cups 0.0006796 3 0.0002265 1.0 NO Replicates 0.0023607 3 0.0007869 3.6 YES Interaction 0.0038752 9 0.0004306 1.9 NO Residual 0.0250987 116 0.0002164 — — Total 0.0317421 131 — — — J.Anal. At. Spectrom., 1999, 14, 703–710 707Fig. 4 Average values and confidence intervals associated with each factor of the two diVerent ANOVA studies: (a) number of replicates; (b) number of sample cups; and (c) tubes. be established as the upper limit for the use of L’vov platform the reduced palladium provided the best results, as has been observed elsewhere.31 tubes when analysing this kind of slurry. In conclusion, 0.5% v/v HNO3 was found to be the most appropriate chemical modifier for the determination of Sb in Measurement procedure slurry soils by ETAAS, providing a good peak shape and a Chemical modifiers and temperature programme.The high satisfactory separation between the atomic and background volatility of Sb and the presence of several concomitants in peaks. The use of HNO3 as chemical modifier is also advantathe sample matrices can cause diYculties in its determination geous because its optimum concentration can be obtained by ETAAS.28 The use of the L’vov platform and chemical directly by preparing the slurries in this suspension medium.modification has been reported for the determination of this element in several matrices. Especially for geological samples, palladium, magnesium, lanthanum, nickel, nitric acid, copper Quantification and molybdenum have been reported.28,29 Also some suitable Pd mixtures such as Pd–Zr, Pd–Zr–citrate and Pd–W–citrate Calibration. There are two general calibration modes when working in ETAAS, namely direct calibration and the method have been used in the determination of volatile elements.30 Recently, a 0.1% m/v nickel nitrate hexahydrate solution has of the standard additions. The former assumes that the analytical signal is only caused by the analyte itself and that all been recommended as a chemical modifier for Sb determination in soil and sediment slurries.14 samples and standards are treated considering the same conditions.This method has the important advantage that one Systematic experiments were carried out to optimize the temperatures and times for the drying, pyrolysis and atomi- calibration line is useful for analysing many samples.The standard additions method is time consuming since a cali- zation steps considering nitric acid, palladium nitrate, a mixture of palladium and magnesium nitrate, palladium nitrate– bration line has to be obtained for each sample. This method accounts for most of the possible matrix interferences and it ascorbic acid, magnesium nitrate, ammonium dihydrogenphosphate and nickel as chemical modifiers.These conditions were is considered as the ‘reference’ method. More details can be found elsewhere.32 optimized for a slurry soil sample (90 mg of certified reference soil GBW07401 in 1 ml of 0.5% v/v HNO3) and aqueous Sb A third calibration method was implemented in the spectrometer software, called the method of addition calibration, standard solutions (20 ng ml-1).When a slurry sample is atomized, the high background and it takes characteristics from the two previous modes.33 It implies the development of one standard addition line for the causes the deuterium background correction system to overcorrect the atomic signal, sometimes giving negative integrated first sample and then the absorbance value of the sample is subtracted from the absorbance of each of the absorbance values. Therefore, only peak height absorbance was taken into account in this work.sample+standard solution. Subsequently, the absorbance of the second, third, etc., sample is added to each of the above In all cases, two drying steps (100 and 130 °C) were needed for eYcient, uniform drying of the soil slurries, being deter- calibration points and the sample is quantified. This method can be applied only if the matrix is more or less constant in mined through careful visual observation of the inner part of the tube.The use of 0.5% v/v HNO3 gave the most symmetric all the samples, which is the case here. To elucidate the most suitable calibration mode and to atomic peak and the best separation between the atomic and background signals, with pyrolysis and atomization tempera- assess if the calculated concentrations showed a dependence on sample mass, three diVerent masses of one reference soil tures of 700 and 1700 °C, respectively. NH4H2PO4, Mg(NO3)2 and Ni provided a slightly lower pyrolysis temperature of (GBW07401) were taken (number of cups and replicates as stated above) and analysed.Direct calibration was made with 800 °C (atomization temperature 1750 °C). These modifiers decreased the atomic peak height and altered the background aqueous standards in 0.5% v/v HNO3. Table 5 presents the quantification results and the recoveries. The slope of the signal appearance. Regarding the Pd-based modifiers, palladium nitrate and direct calibration regression did not diVer statistically (Student’s t-test, 95% confidence) from that of the standard reduced palladium (palladium nitrate–ascorbic acid) increased the pyrolysis temperature to 1000 °C without Sb losses (atomi- additions method, the main disadvantages of the direct calibration being the large confidence interval associated with zation temperature 1750 °C).In contrast, the atomic absorbance decreased considerably and the peaks obtained each prediction and the poor recoveries.The method of addition calibration and the method of standard additions were worse than those with HNO3. Pd(NO3 )2–Mg(NO3)2 (1.5+1) led to the highest pyrolysis temperature, 1200 °C gave the most satisfactory recoveries (83–99%) and quantifi- cations (accuracy and precision). The calibration confidence (atomization temperature 1800 °C), but the amount of modifier employed became critical. Thus, on increasing its amount, the intervals associated with each prediction were calculated considering the Hotelling confidence intervals of the regression atomic signal diminished and the overlap between the atomic and background signals increased.From all these Pd modifiers, lines.31 Accordingly, the method of addition calibration was 708 J. Anal. At. Spectrom., 1999, 14, 703–710Table 5 Comparison of three calibration methods for the quantification of Sb in one certified reference soil (GBW07401) using three diVerent levels of sample mass.Certified value: 0.75–0.87–0.99 mg g-1 Concentration/mg g-1 (average value±Hotelling confidence intervals) Direct calibration Standard additions method Method of addition calibration Sample mass/ mg Mean value Recovery (%) Mean value Recovery (%) Mean value Recovery (%) 40 0.09–0.66–1.23 76 0.52–0.86–1.19 99 0.39–0.72–1.05 83 92 0.46–0.69–0.94 76 0.60–0.86–1.11 99 0.77–0.84–1.19 97 150 0.94–1.04–1.15 119 0.94–1.09–1.26 125 0.77–0.82–1.04 94 selected because of its advantages (narrow confidence intervals, both the LOD and limits of quantitation (LOQ) (10 times the standard deviation of the blank).Note that for 200 mg ml-1, good recoveries, lower costs, high throughput, etc.). no limits were calculated owing to operational problems (the upper limit was set at 150 mg ml-1). The instrumental LOQ Pipetting options. Simultaneously with the calibration tests, we assessed how to introduce the slurries and metal standards was 4.13 ng ml-1. The characteristic mass (mo) was calculated for several into the furnace.Previously, the pipetting sequence was optimized in order to achieve good repeatability. The AS-70 sampler aqueous standards and slurry samples containing diVerent Sb concentrations. The mean values obtained were 17.3±1.7 and allows two pipetting options: (i) ‘pipette separate’: each of the slurry, standard aliquots 12.0±1.6 pg, respectively. and diluent are injected separately, the slurry being the first to be introduced, and a washing step is performed before Accuracy. From an operative point of view, accuracy can aspirating the standard aliquots; be studied in two ways: first, check the accuracy of the (ii) ‘pipette together’: slurry, standards and diluent are instrumental measurements; and second, test the overall pipetted and injected in one step.method accuracy. They allow one to investigate eventual bias A simple way to compare their performance is to develop in diVerent stages of the methodology.The use of the USS-100 either direct or standard addition regressions employing each probe did not cause any bias in the results (as shown below). pipetting mode and compare their slopes statistically. The lines The ‘instrumental accuracy’ has to be assessed using recovery for the pipette separate mode were as follows: calibration, assays. Accordingly, diVerent amounts of an aqueous Sb absorbance=0.0170+0.0045[Sb]; addition, absorbance= standard were added to several slurries obtained from the 0.0827+0.0053[Sb].The regression lines for the ‘pipette certified soil GBW07401, homogenized and then measured. together’ mode were as follows: calibration, absorbance= All recoveries were satisfactory, ranging from 93 to 110%. The 0.0155+0.0042[Sb]; addition, absorbance=0.1042+ method accuracy was evaluated using the reference materials 0.0027[Sb]. mentioned above. The ‘pipette separate’ mode gave much better results (the Despite all the method development carried out using certitwo lines are parallel at the 95% confidence level, texp=2.07 fied soils, several studies were replicated for sediments as a and ttab=2.23, n-1=10) than the ‘pipette together’ option.tentative extension of the developed methodology. Therefore, To our knowledge, this has not been reported previously and, if the two CRM sediments also performed well, this is strong therefore, the ‘pipette separate’ mode is recommended. It is evidence to consider that the methodology can be applied for thought that the diVerences between these two pipetting modes both soils and marine sediments.Table 6 collects the main increased when working with slurry samples because after results for each CRM. As for the instrumental accuracy, the aspirating the slurry the sampler tip remains dirty. This seems recoveries ranged from 82 to 109% and all the results were justified because whenever aqueous calibrations were carried very satisfactory.Not only were the experimental values out they showed the same slope using both pipetting modes. included in the confidence intervals of the certified or indicative In contrast, when the ‘pipette together’ mode was employed values, but they also exhibited excellent agreement with the to construct the standard additions line for slurry solutions, reference values. This confirms that the developed method can the slope diVered statistically from that of the aqueous be used successfully for the direct analysis of soils and sedicalibration. ments for Sb determination.Moreover, it was observed that the atomic peaks presented better characteristics when the ‘pipette separate’ mode was Precision. First, the precision of the instrumental measure- used and the quantification of one reference material ments was studied using two aqueous Sb standards (20 and (GBW07401) was much improved using such a pipetting mode 60 ng ml-1), either applying ultrasonic mixing prior to the (average value and confidence intervals for the ‘pipette separinjection into the furnace and without such homogenization.ate’ mode 0.60–0.86–1.11 mg g-1 and for the ‘pipette together’ Five replicates were performed for each situation and their mode 1.63–2.07–2.50 mg g-1; certified value, 0.75–0.87– relative standard deviations (RSD) were calculated. The values 0.99 mg g-1). were satisfactory and no statistical diVerences (Student’s ttest, 95% confidence) were found considering this variable: Analytical figures of merit with agitation, 20.14±0.59 ng ml-1 (RSD=2.9%); 61.60± 1.60 ng ml-1 (RSD=2.6%); without agitation, 20.55± Sensitivity. The instrumental limit of detection (LOD) is the signal which is statistically diVerent (99.9% confidence) from 0.47 ng ml-1 (RSD=2.3%); 60.63±1.80 ng ml-1 (RSD= 3.0%).Consequently, the ultrasonic homogenization achieved the blank.24 It was calculated from 10 independent blank measurements and using the average slope from six indepen- using the USS-100 probe did not increase the analytical response uncertainty.dent calibrations (which was 0.0048 ml absorbance ng-1). The instrumental LOD for the Sb determination was 1.24 ng ml-1. Second, two CRMs (BCR CRM 141 and BCR CRM 277) with diVerent contents of Sb were selected to evaluate the The method LOD, i.e., the LOD calculated following all the analytical methodology, has to be calculated for each mass to precision.One slurry of each of the two CRMs was prepared and then analysed six times. The RSDs obtained were 3.6 and volume ratio as is usual in slurry techniques.11 Table 2 gives J. Anal. At. Spectrom., 1999, 14, 703–710 709Table 6 Sb contents of five certified reference materials and six ‘real’ samples (average values±standard deviation; n=6) Concentration/mg g-1 RSD (%) Material Obtained value Certified value Recovery (%) (n=6) Soil GBW07401 0.86±0.06 0.87±0.12 99 7.0 Soil GBW07409 0.23±0.01 0.21±0.03 109 6.1 Calcareous loam soil BCR CRM 141 0.69±0.05 (0.7±0.3)a 99 7.2 Estuarine sediment BCR CRM 277 3.21±0.15 (3.9±1)a 82 4.7 Marine sediment BCSS-1 0.58±0.03 0.59±0.06 98 5.5 Soil 1 0.29±0.02 — — 7.0 Soil 2 0.46±0.02 — — 4.3 Soil 3 0.68±0.04 — — 5.9 Sediment 1 0.54±0.05 — — 9.2 Sediment 2 1.06±0.05 — — 4.7 Sediment 3 1.35±0.07 — — 5.2 aNot certified. 7 M. W. Hinds, K. W. Jackson and A. P. Newman, Analyst, 1985, 2.8%, respectively. Although the use of slurries increased the 110, 947.within-sample RSD (as expected), the precision values were 8 M. Hoening and P. Regnier, J. Anal. At. Spectrom., 1989, 4, 631. very satisfactory (<4%). It is thought that these good values 9 C. Bendicho and M. T. C. de Loos-Vollebregt, J. Anal. At. are derived from the excellent homogenization provided by Spectrom. 1991, 6, 353. the USS-100 probe and also the optimized preparation of the 10 P. Bermejo-Barrera, C. Barciela-Alonso, M. Aboal-Somoza and A.Bermejo-Barrera, J. Anal. At. Spectrom., 1994, 9, 469. slurry samples. 11 P. Bermejo-Barrera, J. Moreda-Pin� eiro, A. Moreda-Pin� eiro and Next, six diVerent portions of each CRM were weighed, A. Bermejo-Barrera, Anal. Chim. Acta, 1994, 296, 181. slurry prepared and analysed to calculate the overall method- 12 I. Karadjova, P. Mandujukov, S. Tsakovsky, V. Simeonov, J. A. ology precision. The precision values were good (7.2–4.7%). Stratis and G. A. Zachariadis, J. Anal.At. Spectrom., 1995, 10, Similar figures were achieved when six ‘real’ samples (three 1065. soils and three sediments) were analysed applying the above 13 P. Bermejo-Barrera, M. C. Barciela-Alonso, J. Moreda-Pin� eiro, C. Gonza�lez-Sixto and A. Bermejo-Barrera, Spectrochim. Acta, developed methodology (Table 6). This strongly suggests that Part B, 1996, 51, 1235. the numbers of cups and replicates optimized using CRMs are 14 I. Lo�pez-Garcý�a, M. Sa�nchez-Merlos and H. Herna�ndez-Co�rdoba, also valid when dealing with real samples.16 Therefore, these Spectrochim.Acta, Part B, 1997, 52, 437. results can be used to confirm that Sb is also homogeneously 15 N. J. Miller-Ilhi, J. Anal. At. Spectrom., 1989, 4, 295. distributed in the soil and sediment samples studied in this 16 M. S. Epstein, G. R. Carnrick and W. Slavin, Anal. Chem., 1989, work. The Sb contents for soils were comparable to natural 61, 1414. 17 N. J. Miller-Ihli, At. Spectrosc., 1992, 13, 1.reported values (0.2–1 mg g-1).34 Regarding sediments, they 18 N. J. Miller-Ihli, Fresenius’ J. Anal. Chem., 1993, 345, 482. exhibited slightly higher concentrations than soils. Never- 19 N. J. Miller-Ihli, J. Anal. At. Spectrom., 1994, 9, 1129. theless, these levels were similar to others already reported for 20 W. Klemm and G. Bombach, Fresenius’ J. Anal. Chem., 1995, estuarine sediments which are considered as normal values 353, 12. (0.04–1 mg g-1).35 For both types of samples, the Sb contents 21 R. Dobrowolski, Spectrochim. Acta, Part B, 1996, 51, 221. corresponded to unpolluted sites. 22 I. Koch, C. F. Harrington, K. J. Reimer and W. R. Cullen, Talanta, 1997, 44, 771. 23 R. L. Plackett and J. P. Burman, Biometrika, 1946, 33, 305. 24 J. C. Miller and J. N. Miller, Statistics for Analytical Chemistry, Acknowledgements Ellis Horwood, Chichester, 2nd edn., 1992. 25 N. J. Miller-Ilhi, Spectrochim. Acta, Part B, 1997, 52, 431. This research was supported by the Vicerrectorado de 26 Standard Guide for Conducting Ruggedness Tests, ASTM E 1169, Investigacio�n of the University of La Corun� a. The authors American Society for Testing and Materials, Philadelphia, PA, thank B. Ferna�ndez Souto (Servicios Xerais de Apoio a� 1997. Investigacio� n) for his collaboration in obtaining the optical 27 A. Carlosena, M. Gallego and M. Valca�rcel, J. Anal. At. microscopy photographs. Professor X. Toma�s is also acknowl- Spectrom., 1997, 12, 479. edged for his kind advice on the chemometric concepts. 28 H. Niskavaara, J. Virtasalo and L. H. J. Lajunen, Spectrochim. Acta, Part B, 1985, 40, 1219. 29 Y. Morishigue, K. Hirokawa and K. Yasuda, Fresenius’ J. Anal. Chem., 1994, 350, 410. References 30 I. Havezov, A. Detcheva and J. Rendl, Mikrochim. Acta, 1995, 1 S. Allen, Chemical Analysis of Ecological Materials, Blackwell, 119, 147. 31 E. Bulska and W. Jedral, J. Anal. At. Spectrom., 1995, 10, 49. Oxford, 2nd edn., 1989. 32 M. J. Gardner and A. M. Gunn, Fresenius’ J. Anal. Chem., 1988, 2 G. B. van der Voet and F. A. de WolV, in Toxicology of Metals, 330, 103. ed. L. W. Chang, CRC Press, Boca Raton, FL, 1996. 33 Atomic Absorption Laboratory Benchtop. User’s Guide, Perkin- 3 B. J. Alloway, Heavy Metals in Soils, Wiley, Chichester, 1st edn., Elmer, U� berlingen, 1991. 1990. 34 A. Kabata-Pendias and H. Pendias, Trace Elements in Soils and 4 C. Dietl, W. Reifenha�user and L. Peichl, Sci. Total Environ., 1997, Plants, CRC Press, Boca Raton, FL, 1992. 205, 235. 35 U. Fo� rstner and G. T. W. Wittman, Metal Pollution in the Aquatic 5 K. Takayanagi, D. Cossa and J.-M. Martin, Mar. Chem., 1996, Environment, Springer, Berlin, 1983. 54, 303. 6 N. N. Greenwood and A. Earnshaw, Chemistry of the Elements, Pergamon Press, Oxford, 1997. Paper 8/07359G 710 J. Anal.