Continuous extraction with acidiÆed subcritical water of arsenic, selenium and mercury from coal prior to on-line derivatisation-atomic Øuorescence detection V. Ferna�ndez-Pe�rez, M. M. Jime�nez-Carmona and M. D. Luque de Castro Analytical Chemistry Division, Faculty of Sciences, University of Co�rdoba, E-14004 Co�rdoba, Spain. E-mail: qa1lucam@uco.es Received 27th July 1999, Accepted 9th September 1999 AcidiÆed subcritical water is proposed for the continuous extraction of minor pollutants (namely, selenium, arsenic and mercury) from coal prior to continuous derivatisation (by hydride formation for Se and As, and cold vapor formation for Hg) and determination by atomic Øuorescence.Coal samples (3 g) were subjected to a 15 min static extraction followed by a 90 min dynamic extraction with water modiÆed with 4% (v/v) HNO3 for both steps. An in-depth study of the variables affecting the continuous leaching step, as well as those referring both to preconcentration (for mercury), derivatisation and detection (all of them) was performed.The linear ranges of the calibration curves for all analytes were at the ng ml21 level, with correlation coefÆcients, r2, better than 0.999 for Se and As and better than 0.99 for Hg. The method was validated by using a bituminous coal reference material (NIST SRM 1635). The good precision of the method [RSDs (n~6) of 12.0, 4.7 and 6.5% for As, Se and Hg, respectively], together with its safety and rapidity make it a good alternative for the determination of these analytes in coal.Introduction Coal combustion has been identiÆed as a potential source of pollution from volatile trace metals. Studies1 on the fate of such elements during combustion have shown that up to 85, 60 and 55% of the mercury, arsenic and selenium, respectively, originally present in the coal could not be accounted for in the waste streams examined. Other research2 has suggested that almost all of the arsenic and selenium present in the coal appears in the Øy ash in the exhaust gases.Arsenic is a metalloid used in many industrial processes and applications, thus making possible the contamination of water, soil and food.3 The toxic effect of mercury compounds has for long been recognised. Selenium is a metalloid present in the environment, mainly as a result of human activity, whose content in environmental materials has had to be established because of its ambivalent character (both toxic and essential). 4,5 Thus, due to the high toxicity and proved inØuence in the environment of these analytes, it is mandatory to search for a reliable quantitative determination for them in coal. Several methods have been applied for arsenic, selenium and mercury determination in coal. Spectrophotometry,6,7 X-ray Øuorescence spectrometry,8 colorimetry9 and atomic absorption spectrometry10 have been applied to the determination of arsenic in coals. Selenium is present in coal as both organic and inorganic compounds and it has been determined by Øuorimetry, 11,12 activation analysis13 and atomic absorption spectrometry. 14,15 In the case of mercury, the methods described have been based on colorimetry±Øuorimetry,16 gas chromatography, 17 neutron activation18 and atomic absorption spectrometry. 19 Other methods for the analysis of coals have been developed, such as slurry sampling with graphite furnace AAS.20,21 The methods thus developed for the analysis of trace elements in coal involve pretreatment of the matrix prior to the measurement step, consisting mainly of an ashing procedure, used especially by standard methods.22 In recent years, the digestion step has been carried out under very drastic conditions, this being thought the best way to achieve quantitation.23 The environmentally aggressive character of these methods, as well as their slowness, makes the search for analytical alternatives urgent.Subcritical water extraction, the technique based on the use of water as an extractant, at temperatures between 100 and 374 �C and pressures high enough to maintain the liquid state, is emerging as a powerful alternative for solid sample extraction.Thus, subcritical water extraction has been used for the extraction of organic pollutants within a wide range of polarities from environmental samples.24±28 The aim of this research has been to develop a rapid, clean and efÆcient method for extraction of trace metals from coal, based on continuous subcritical water29±31 extraction modiÆed if required, and the joint use of static and dynamic extraction.Experimental Apparatus The acidiÆed subcritical water extractions were performed using a prototype extractor (designed by Salvador and Mercha�n32) consisting of a stainless steel cylindrical extraction chamber (150611 mm id), closed with screws at either end, that permit the circulation of the leaching Øuid through them. Both screw caps contain stainless steel Ælter plates (2 mm in thickness and 1/4 in id) to ensure that the sample remains in the extraction chamber.The chamber, together with a stainless steel preheater, is located in an oven designed to work up to 300 �C and controlled by a Toho TC-22 temperature controller. A cooler system (consisting of a loop made from 1 m length stainless steel tubing and cooled with water at room temperature) was used to cool the extract from the oven to a temperature close to 25 �C.A Shimadzu (Tokyo, Japan) LC10AD pump with digital Øow-rate and pressure readouts was used to impel the extractant through the system. An Excalibur-Merlin atomic Øuorescence detector (PS Analytical, Orpington, Kent, UK) Ætted with boosted discharge hollow cathode lamps for Se, As and Hg, two Gilson Minipuls-3 peristaltic pumps (Gilson, Middleton, WI, USA) Ætted with rate selectors, two gas±liquid separators (for hydride J. Anal. At. Spectrom., 1999, 14, 1761±1765 1761 This Journal is # The Royal Society of Chemistry 1999generation and mercury separation), two Rheodyne (Cotati, CA, USA) 504 injection valves and PTFE tubing of 0.5 mm inner diameter were used to build the Øow injection (FI) manifold for the detection step. A Knauer recorder (Bad Harzburg, Germany) was used in order to record the signals. Reagents and solutions Ultrapure water from a Milli-Q system (Millipore, Bedford, MA, USA) was used throughout.Water modiÆed with 4% v/v nitric acid was used as the extractant. A 2% w/v NaBH4 solution (Sigma-Aldrich, Deisenhufen, Germany) in 0.1 mol l21 NaOH (Merck, Darmstadt, Germany) was used for selenium and arsenic derivatisation and 5% SnCl2 was used as the reagent in the mercury derivatisation step; 6.0, 3.0 and 0.3 mol l21 HCl solutions were used for Se and As reduction, hydride generation and mercury reduction, respectively. Solutions 561022% of 1-pyrrolidinecarbodithioic acid ammonium salt (APDC, Aldrich, Milwaukee, WI, USA) in 30 mmol l21 ammonium acetate (Merck)±acetic acid (Panreac, Barcelona, Spain) buffer, pH 6.5, and ethanol were used as complexing agent and eluent, respectively, for the mercury preconcentration step.The preconcentration column was packed with C18 bonded phase, obtained from Bond Elut cartridges (Varian, Harbor City, CA, USA). Argon (Carburos Meta� licos, Barcelona, Spain) was used to Øush the analytes to the detector.A stock solution of 1 g l21 AsIII was prepared by dissolving 1.320 g of As2O3 (Merck) in 25 ml of 20% m/v KOH solution, neutralizing with 20% H2SO4 and diluting to 1 l with 1% v/v H2SO4. One g l21 solution of SeIV (Merck) and HgII (Aldrich) were prepared in 3 mol l21 HCl and 1% HNO3, respectively. All reagents were of analytical grade and prepared daily. Procedure Extraction step. Coal (3 g) was placed in the extraction cell. After assembling the cell and locating it in the oven, this was brought up to the work temperature (180 �C) and pressurised with y50 bar by opening the inlet valve from the pump.The valve was then closed and static extraction was developed for 15 min. Both the inlet and outlet valves were then opened. AcidiÆed water oven at a Øow-rate of 2.5 ml min21 and the extract was collected in a vial after being cooled in the refrigerant at 25 �C. For kinetic experiments, volumes of 40 ml of extracts were collected at intervals of 16 min.Derivatisation and detection steps. Two aliquots from each extraction were subjected to different treatments, one for selenium and arsenic and the other for mercury. Reduction step (AsV and SeVI). These analytes are in their highest oxidation states as a consequence of the presence of nitric acid in the extractant. They were reduced to AsIII and SeIV from which the corresponding hydride was generated. The reduction step was performed by adding 12.5 ml of 6.0 mol l21 HCl to 12.5 ml of extract and heating the mixture at 75 �C for 45 min.33 Determination of AsIII and SeIV.Once the sample had been reduced, it was injected into a 3.0 mol l21 HCl carrier stream which then merged with a 2% NaBH4 stream [see Fig. 1(a)]. The volatile hydride was formed and swept out of the gas± liquid separator by an argon stream into the chemically generated hydrogen diffusion Øame, which was maintained by the excess of hydrogen produced in the reaction between NaBH4 and HCl.The hydride was atomised in the Øame and detected by atomic Øuorescence spectrometry. Preconcentration±detection step (Hg). The low concentration of mercury made necessary the development of a preconcentration34,35 step in an on-line Øow injection (FI) system. The manifold used in this step is shown in Fig. 1(b). A solution of 561022% chelating agent (APDC) was prepared in 30 mmol l21 ammonium acetate adjusted to pH 6.5 with 2 mmol l21 acetic acid.APDC (2.5 ml) was added to 100 ml of extract. The sample (or the standard solution) was aspirated and allowed to pass through the preconcentration column for 15 min. Then, the two injection valves were simultaneously switched to the injection position and the retained complex was eluted with an ethanol stream. The HgII was reduced to HgO by an SnCl2 stream, swept out of the gas±liquid separator by an argon stream into the atomic Øuorescence detector, and the signal recorded.The volume of sample injected was in all cases the inner volume of the injection valve (50 ml). Results and discussion Optimisation The experimental variables were optimised with the following aims: (a) to achieve quantitative extraction of the analytes; (b) to shorten the extraction time as much as possible; (c) to reduce both extractant consumption, in order to save reagent (nitric acid), and waste; and (d) to decrease analyte dilution. With these aims a hybrid discontinuous/continuous extraction method was intended.The univariate method was used in all instances. Ranges over which the effect of the variables was studied and the optimum values found are shown in Table 1. Extraction variables. The variables concerning leaching (namely temperature, concentration and composition of the leacher) were studied in order to Ænd the optimum conditions for this step. Three g of coal, a static extraction time of 10 min and a dynamic extraction time of 30 min were used for optimisation experiments.Since the extractor was not Ætted with a pressure controller, a pressure of 50 bar (created in the system by the working conditions) was used in all instances. The temperature of the extraction chamber was studied by performing a 30 min dynamic extraction with water modiÆed with nitric acid to 4% (v/v). The stainless steel extractor allowed the extraction to be performed up to 180 �C. An increase in the extraction efÆciency was observed when the temperature increased.Temperatures higher than 180 �C did not permit Fig. 1 Flow injection manifolds for the derivatisation±determination of arsenic and selenium (a), and for preconcentration±derivatisation of mercury (b). PP, peristaltic pump; IV, injection valve; W, waste; PC, preconcentration column; MC, mixing coil; RC, reaction coil; GLS, gas±liquid separator; AFD, atomic Øuorescence detector; R, recorder. 1762 J. Anal. At. Spectrom., 1999, 14, 1761±1765us to maintain the acidiÆed water in liquid state because the pressure was not high enough.A temperature of 180 �C was selected as optimum for further experiments. The leacher composition was studied in previous work29 and HNO3 was selected as extractant. The leacher concentration, ranging from 1 to 5% (v/v), was studied too. The efÆciency of the extraction increased with more acidic media, but nitric acid concentrations higher than 4% gave rise to overpressure in the system which hindered extraction, so 4% nitric acid concentration was Æxed for subsequent experiments.The hydrodynamic variables, namely Øow-rates and static and dynamic extraction times, were also evaluated. The Øowrate of the leaching agent was studied, with an extraction time of 16 min and values of the variable from 0.5 to 5.0 ml min21 were investigated. The efÆciency increased with the Øow-rate up to 2.5 ml min21, and a drop in yield occurred for higher values. This trend could be due to a higher compactness with increased Øow-rates that inhibited sample±extractant contact. Thus, a Øow-rate of 2.5 ml min21 was selected as the optimum.Quantitative leaching of the analytes was attained by a static extraction time of 15 min followed by a dynamic extraction time of 90 min. Derivatisation±detection variables. The efÆciency of hydride generation was studied for different Øow-rates of HCl and NaBH4. The HCl stream was Æxed at 4.5 ml min21 to ensure a sufÆciently acidiÆed medium.A NaBH4 Øow-rate lower than 2.5 ml min21 gave rise to Øame extinction owing to the low hydrogen concentration produced. Therefore, a HCl Øow-rate of 4.5 ml min21 and a NaBH4 Øow-rate of 2.5 ml min21 were selected for further experiments. The NaBH4 concentration is an important parameter for hydride generation. This concentration was studied in an interval between 0.5 and 4% w/v in 0.1 mol l21 NaOH solution. A concentration higher than 2% w/v was required in order to obtain both an appropriate hydride formation and hydrogen generation; for a lower concentration than this the Øame was extinguished, while for NaBH4 concentration higher than 3.5% the instability of the Øame caused by an excess of hydrogen gave a poor reproducibility and the signal to noise ratio was smaller.When the FI system was optimised for mercury determination, the Øow-rates of SnCl2 and HCl were Æxed at 1 and 2.5 ml min21, respectively. The best results for the mercury determination were found for 5% SnCl2 and 0.3 mol l21 HCl.Different manifolds were tested for the mercury preconcentration step, and the maximum recovery was obtained with the values of the variables showed in Table 1. Discontinuous±continuous extraction versus continuous extraction. A comparison between hybrid discontinuous± continuous extraction (consisting of a prior static extraction step, followed by a dynamic extraction step in which the extractant was continuously passed through the extraction chamber) and continuous extraction (in which only the latter step was applied) was performed.Different static extraction times were tested with the aim both of increasing the efÆciency and minimising dilution of the sample, thus reaching quantitative extraction in a substantially lower extract volume. As an example, the inØuence of this step on the extraction kinetics of the analytes is shown in Fig. 2, where the kinetic curves for As and Se with and without application of a static extraction step can be observed.A range from 0 to 30 min static extraction time was studied. The Table 1 Optimisation of the method Type of variable Variable Range studied Selected value Extraction variables Temperature/�C 80±250 180 Pressure/bar – 50 Nitric acid (%) 1±5 4 Flow-rate/ml min21 0.5±5 2.5 Static extraction time 0±30 15 Dynamic extraction time 15±180 90 Derivatisation±deteles for As and Se.HCl Øow-rate/ml min21 0±5.0 4.5 NaBH4 Øow-rate/ml min21 0.6±4.5 2.5 [NaBH4] (%) 0±5 2 [HCl]/mol l21 0.5±4 3 Derivatisation±detection variables for Hg HCl Øow-rate/ml min21 1.8±5 3.5 SnCl2 Øow-rate/ml min21 0.5±3.5 1.0 [HCl]/mol l21 0±3.5 0.3 [SnCl2] (%) 0±10 5 Type of eluent H2O Acetonitrile±H2O Acetonitrile Ethanol Ethanol Type of complexing agent DDC APDC APDC [APDC] (%) 561024±0.5 561022 Preconcentration time/min 0±30 15 Eluent volume/ml 0.2±2 1.5 Fig. 2 Kinetic extraction curves obtained by acidiÆed subcritical water with and without static extraction for As (–+– 0 min and –r– 15 min of static extraction), and Se (–6– 0 min and –&– 15 min static extraction).J. Anal. At. Spectrom., 1999, 14, 1761±1765 1763efÆciency increased with time up to 15 min, observing no signiÆcant improvement in the kinetics for longer times. Thus, a static extraction time of 15 min was selected as it provided the best analytical performance. The application of a 90 min dynamic extraction time after 15 min static extraction time yielded quantitative extraction of the analytes.The absence of static extraction time hindered quantitative extraction of the analytes even after 2 h, which demonstrates the necessity of applying this hybrid extraction mode in order to achieve completeness of extraction in a relatively short time. Features of the method Calibration curves were obtained using eight individual standard solutions of the analytes.Table 2 shows the equation for the calibration curves, the linear range for each analyte at the ng ml21 level. The sensitivity of the detector was changed in order to cover a wide concentration range. For this reason, two different slopes of the calibration curve (one for each value of the sensitivity) were obtained for each analyte. The standards and samples were injected in triplicate into the FI system in all instances. The method shows a good linearity with correlation coefÆcients, r2, better than 0.999 for Se and As, and better than 0.99 for Hg.The precision of the derivatisation±detection step for each analyte, expressed as RSD%, is also shown in Table 2. As can be seen in the table, the repeatibility study (for n~11 and concentrations of 40 ng ml21 for selenium and arsenic and 0.1 ng ml21 and 100 ng ml21 for mercury with and without preconcentration, respectively), yielded values lower than 3.50 in all instances, but for HgII without preconcentration the value was 5.9 owing to the difÆculty in measuring small signals.Validation of the method The accuracy of the method was evaluated by analysing a bituminous coal reference material with a sulfur content of approximately 0.3% (NIST-SRM 1635). As can be seen in Table 3, the results obtained are in good agreement with the certiÆed values. Finally, the method was applied to three coals from different locations. The study of the precision of the whole process (including the extraction and detection steps), expressed as RSD, was carried out performing extraction of 3 g of coal under the optimum working conditions.The RSDs (n~6) for arsenic, selenium and mercury were 12.0, 4.7 and 6.5, respectively. All these results are shown in Table 3. Conclusions A clean and rapid method for the determination of trace pollutants (namely, selenium, arsenic and mercury) in coal based on continuous subcritical extraction with acidiÆed water (in a static±dynamic mode) as a step prior to preconcentration± derivatisation (for mercury), on-line derivatisation based on hydride formation for As and Se and atomic Øuorescence detection (for all) is proposed.The establishment of a static extraction step prior to the continuous dynamic extraction is the key to both avoiding the dilution effect caused by continuous passage of the extractant through the chamber and shortening the time required for complete extraction. The proposed method is thus quicker, cleaner and safer than other methods established with this aim.Acknowledgements The Spanish Comisio�n Interministerial de Ciencia y Tecnologý�a (CICyT) is thanked for Ænancial support (Project No. PB96- 1265). References 1 N. E. Bolton, J. A. Carter, J. F. Emery, C. Felman, W. Fulkerson, L. D. Hulett and W. S. Lyon, in Trace Elements in Fuel, Advances in Chemistry Series, ed. S. P. Babu, American Chemical Society, Washington, DC, USA, 1975. 2 G. T.Moore and V. J. Elia, 71st Annual Meeting of Pollution Control Association, Houston, Texas, 1978, 78-34.2. 3 R. M. Harrison and S. Rapsomanakis, Environmental Analysis Using Chromatography Interfaced with Atomic Spectrometry, Wiley, Chichester, UK, 1989. 4 H. A. Schroder, D. V. Frost and J. Balassa, J. Chronic Dis., 1970, 23, 227. 5 W. O. Robinson, J. Assoc. Off. Anal. Chem., 1993, 16, 423. 6 N. A. Kevin and V. E. Marincheva, Zavod. Lab., 1970, 36, 1061. Table 2 Features of the method Analyte Equation Linear range/ng ml21 Regression coefÆcient/r2 Repeatability RSD (%) (n~11) Detection limit/ng ml21 AsIII Range 1 y~45.9z1.40x 1±20 0.997 3.3 0.22 Range 2 y~62.9z0.43x 20±100 0.9993 SeIV Range 1 y~4.12z4.80x 1.5±30 0.9990 3.44 0.46 Range 2 y~87.1z0.82x 30±150 0.998 HgII ay~22.8z0.15x 0.01±1 0.992 5.9 0.015 by~7.20z0.25x 5±250 0.98 2.7 4.5 aWith preconcentration.bWithout preconcentration. Table 3 Application of the method to natural samples and CRM Sample Se/mg kg21 As/mg kg21 Hg/mg kg21 Coal A 2.12°0.09 3.15°0.34 0.128°0.006 Coal B 0.88°0.09 3.54°0.28 0.096°0.002 Coal C 2.26°0.07 6.27°0.31 0.059°0.003 RSD (%) 4.7 12 6.5 NIST-RSM 1635 CertiÆed value/mg kg21 Found value/mg kg21 Selenium 0.9°0.3 0.755°0.005 Arsenic 0.42° 0.15 0.350°0.056 Mercury 0.02 0.022°0.002 1764 J.Anal. At. Spectrom., 1999, 14, 1761±17657 G. I. Spielholtz and R. B. Diehl, Talanta, 1996, 13, 991. 8 E. Hafter, D. Schmidt, P. Freimann and W. Gerwinski, Spectrochim.Acta, Part B, 1997, 52, 935. 9 J. Messerschmidt, A. Von-Bohlen, F. Alt and R. Klockenka»mper, J. Anal. At. Spectrom., 1997, 12, 1251. 10 T. Kubota, T. Yamaguchi and T. Okutani, Talanta, 1998, 46, 1311. 11 W. H. Gutenmann, G. J. Doss and D. J. Lisk, Chemosphere, 1998, 37, 389. 12 I. S. Palmer and N. Thiex, J. AOAC Int., 1997, 80, 469. 13 A. M. Yusof, Z. ManaÆah and A. K. H. Wood, Sci. Total Environ., 1998, 214, 247. 14 L. C. Robles and A. J. Aller, Anal. Sci., 1996, 12, 783. 15 G. J. Kumar and N. N. Meeravali, At. Spectrom., 1996, 17, 27. 16 I. Vedrina-Dragojevic, D. Dragojevic and S. Cadez, Anal. Chim. Acta, 1997, 355, 151. 17 K. C. Bowles and S. C. Apte, Anal. Chem., 1998, 70, 395. 18 L. J. Blanchard and J. D. Robertson, Radioanal. Nucl. Chem., 1998, 235, 255. 19 M. W. Hinds, Spectrochim. Acta, Part B, 1998, 53, 1063. 20 L. Ebdon and W. C. Pearce, Analyst, 1982, 107, 942. 21 D. Bradshaw and W. Slavin, Spectrochim. Acta, Part B, 1989, 44, 1245. 22 Australian Standard 1038, Part 14.2, Analysis of Higher Rank Coal Ash and Coke, Standard Association of Australia, Sydney, Australia, 1985. 23 E. Ikavalko, T. Laitinen and H. Revitzer, Fresenius J. Anal. Chem., 1999, 363, 314. 24 R. W. Shaw, T. B. Brill, A. A. Clifford, C. A. Ackert and E. U. Franck, Chem. Eng. News, 1991, 69, 26. 25 V. K. Jam, Environ. Sci. Technol., 1993, 27, 806. 26 S. B. Hawthorne, Y. Yang and D. J. Miller, Anal. Chem., 1994, 66, 2912. 27 Y. Yang, S. B. Hawthorne, S. Bowardt and D. J. Miller, Anal. Chem., 1995, 67, 4571. 28 Y. Yang, S. B. Hawthorne and D. J. Miller, J. Environ. Sci. Technol., 1997, 31, 430. 29 M. M. Jimenez-Carmona, J. J. Manclus, A. Montoya and M. D. Luque de Castro, J. Chromatogr., 1997, 785, 329. 30 M. M. Jimenez-Carmona, V. Ferna�ndez-Pe�rez and M. M. Luque de Castro, Anal. Chim. Acta, in the press. 31 V. Ferna�ndez-Pe�rez, M. M. Jime�nez-Carmona and M. D. Luque de Castro, J. Chromatogr., submitted for publication. 32 US Pat. 5,400,642, 1995. 33 L. Pitts, P. J. Worsfold and S. J. Hill, Analyst, 5. 34 R. Falter and H. F. Scho» ler, Fresenius J. Anal. Chem., 1995, 353, 34. 35 M. Ferna�ndez-Garcý�a, R. Pereiro-Garcý�a, N. Berdel-Garcý�a and A. Sanz-Mendel, Talanta, 1994, 41, 1833. Paper 9/906094D J. Anal. At. Spectrom., 1999, 14, 1761±1