摘要:

Critical Review Determination of Selenium and Tellurium in Environmental Samples Alessandro D’Ulivo CNR, Istituto di Chimica Analitica Strumentale, Via Risorgimento 35, 56126 Pisa, Italy Summary of Contents 1 Sample Collection and Storage 1.1 Sample Collection Vessels and Procedures 1.1.1 Water and sediments 1.1.2 Air and gaseous sampling 1.2 Storage Conditions: Physical 1.3 Storage Conditions: Chemical 1.4 Reagents 2 Sample Treatment 2.1 Digestion 2.1.1 Mineralization 2.1.2 Protein hydrolysis 2.2 Extraction 2.2.1 Selective phase extraction 2.2.2 Organic matter fractionation 2.3 Preconcentration 2.3.1 Cold trapping 2.3.2 In situ trapping of hydrides 2.3.3 Coprecipitation and precipitation 2.3.4 Chelation–extraction 2.4 Control of Interferences 2.4.1 Hydride derivatization 2.4.2 Piazselenol derivatization 3 Speciation 3.1 Selective Extraction 3.1.1 Natural waters 3.1.2 Sediments 3.1.3 Biological samples 3.2 Chromatographic Methods 3.2.1 Gas chromatography 3.2.2 Liquid chromatography 3.2.3 Liquid chromatography with on-line detection 4 Detection 4.1 Non-specific Detectors 4.1.1 Gas Chromatographic detectors 4.1.2 Fluorimetric detectors 4.2 Element-specific Detectors 4.2.1 Atomic absorption detectors 4.2.1.1 Total element determination 4.2.1.2 Chromatographic detection 4.2.2 Atomic emission detectors 4.2.3 Atomic fluorescence detectors 4.2.4 Mass spectrometric detectors 4.3 Species-specific Detectors 4.3.1 Electrochemical detectors 4.3.2 Spectrometric detectors 5 Conclusions 6 References Keywords: Review; selenium; tellurium; environmental samples Both selenium and tellurium are extremely rare elements in the Earth’s crust, the relevant estimated abundance being around 0.05–0.09 mg g21 for selenium1,2 and even lower for tellurium,3 but their use in many technological processes results in local enrichment and release.Industrial applications of selenium are in the manufacture of semiconductor materials, photoelectric cells, glass, pigments and pharmaceuticals and in metallurgy.Tellurium has similar applications in the semiconductor industry, electronics and metallurgy. 4 The average environmental levels for selenium are typically higher by one or more orders of magnitude than that of tellurium.3–8 The emission of inorganic selenium and tellurium compounds in the environment may create serious problems owing to the toxicity of these elements.In addition, following their microbial activities, both elements can be transformed into volatile organometalloid compounds with consequent modification in their transport pattern and toxicological behavior.4, 9,10 Although both are regarded as toxic elements, selenium has received much more attention than tellurium, since the discovery that it is an essential trace element11and also plays a biochemical role as a component of glutathione peroxidase.12 Many other organoselenium compounds have been identified in the environment and in living systems ranging from simple inorganic forms, simple methylated species, selenoamino acids Alessandro D’Ulivo graduated in Chemistry at the University of Pisa, Italy in 1978.He has been a researcher of the Italian National Research Council since 1981, developing research at the Institute of Instrumental Analytical Chemistry in Pisa in the field of trace element determination by atomic spectrometry. He is also a lecturer for graduate courses at the Department of Chemistry and Industrial Chemistry of the University of Pisa.He is author and co-author of numerous papers on fundamental aspects and applications of hydride and vapour generation techniques coupled with atomic fluorescence spectrometry (AFS) detection, AFS element-specific detectors for gas chromatography, sample preparation techniques, and chemical modifiers in ETAAS. His current main interest is in the development of improved AF spectrometers and atomisers to be used in the element-specific detection for vapour generation techniques and chromatography.Analyst, December 1997, Vol. 122 (117R–144R) 117Rand related compounds4,13–16 up to the more complex selenoenzymes and selenium nucleic acids17 (see Table 1). Many harmful and beneficial properties of selenium compounds have been demonstrated for humans, plants and animals.Among them, the relationships with immunity and cancer perhaps attract the greatest current interest (for a review, see Masukawa18). The interaction of selenium with heavy metals and toxic elements is also well known. Selenium compounds are thought to be detoxifying agents, playing an antagonistic role towards mercury,19,20.21 methylmercury,22,23 cadmium,24 silver, 25 lead26 and many others.18 Interactions of selenium with other toxic elements in living organisms are continuing to be reported.27–32 The natural occurrence of organotellurium compounds in forms more complex than dimethyl telluride or dimethyl ditelluride, has not yet been demonstrated, and also no essential biological function for tellurium is recognized.Dimethyl telluride, dimethyl ditelluride and telluroamino acids have been identified during laboratory experiments on fungi, bacteria and yeasts incubated in the presence of inorganic tellurite33–36 (see Table 1).Tellurium is still regarded only as a rare, toxic, nonessential element.37,38 The concentrations of some selenium and tellurium species in samples of different origin are reported in Tables 2–5.Only a few papers on the determination of tellurium in the environment have been published, in contrast to the environmental selenium literature. The instability of its organo compounds toward air and light and the rarity of tellurium in nature, with the consequent necessity for high sensitivity and sophistication in the analytical procedures for its determination, may have contributed to such a lack of information.In this review, attention is mainly focused on the determination of organic derivatives of selenium and tellurium, but speciation and total element determinations are also considered. Analytical techniques and instrumentation that seem promising in environmental analysis are also discussed. 1 Sample Collection and Storage Almost all considerations made for other trace elements39 are valid also for the determination of selenium and tellurium in environmental samples.Sample manipulation prior to the analytical determination should be kept to the minimum to avoid contaminations, losses or speciation changes. Considering the low concentration levels of selenium and, in particular, of tellurium in environmental samples, great care must be taken, after the sampling step, to avoid gross sample contamination. 1.1 Sample Collection Vessels and Procedures 1.1.1 Water and sediments Yoon et al.3 collected sub-surface sea-water samples to be analyzed for tellurium using an acid-leached Niskin bottle and filtering the samples through a 0.45 mm membrane to remove metallic sulfide precipitates. For the same purpose, Lee and Edmond8 employed either Niskin bottles with a Teflon-coated internal spring hung on steel wire or rosette mounted Niskin bottles.Modified GO-FLO bottles and filtration through 0.3 mm Nuclepore filters were employed by Cutter and Bruland40 for sub-surface sampling of sea-water to be analyzed for selenium. Measures et al.41 collected surface sea-water samples while the ship was cruising at about 10 knots using an acid-leached, Teflon-lined, polypropylene tube attached to and protruding toward a towed fish: the analyses for selenium were made onboard ship.Tanzer and Heumann42 employed a simple 10 l bucket to collect surface water, in the Atlantic Ocean, to be analyzed for dimethyl selenide and trimethylselenonium (TMSe) compounds. Selenium and tellurium levels in sediments are several orders of magnitude higher than in water samples.This greatly reduces the risks of sample contamination. Velinsky and Cutter43 sampled sediment cores in a coastal salt marsh by carefully driving a butyrate core liner (6 cm in diameter) into the sediment. The cores were immediately sealed from the atmosphere and returned to the laboratory for further processing (separation of pore water from sediment) and selenium determination. 1.1.2 Air and gaseous sampling Atmospheric alkyl selenides44–46 were collected by passing air at 3 l min21 in a cryogenic U-trap, after filtration through a 0.4 mm membrane, and subsequently delivered to a gas chromatographic column by thermal stripping. Detailed studies were made in order to optimize the collection efficiency.Both the temperature and the type of adsorbents filling the U-trap were found to be critical.46,47 The best adsorbents were found to be acid-washed poly(vinyl) chloride and Pyrex Glass-wool (silanized or untreated). The collection temperature had to be maintained @2130 °C in order to obtain a quantitative collection efficiency for dimethyl diselenide, diethyl selenide and, in particular, the most volatile dimethyl selenide.The main disadvantage of the trap was its clogging due to ice formation after 4 h of collection. Muangnoicharoen et al.48 collected selenium and tellurium in the atmosphere by adsorption on either gold-coated quartz beads or charcoal. Near 100% collection efficiency was obtained at sampling flow rates of 1 l min21 for both adsorbents, but only charcoal maintained quantitative efficiency with sampling flow rate up to 5 l min21.The trapping systems are not affected by ice formation and their collection efficiency was tested with volatile inorganic SeIV and SeVI and TeIV and TeVI species. The collection efficiency of the gold-coated quartz beads was improved by increasing the gold coating from 10 to 15%.In this way, quantitative collection efficiency was obtained for sampling flow rates up to 7 l min21. The method Table 1 Some selenium and tellurium compounds encountered in environmental and biological systems Type Selenium Tellurium Inorganic species Selenite, selenate, Se0, Se2II Tellurite, tellurate, Te0 Simple organic and MeSeH, Me2Se, Me2Se2, Me3Se+, Me2Te, Me2Te2 methylated species Me2SeO2, MeSeO(OH), Me2SeO, MeSSeMe, SeNC(NH2)2 Amino acids and low Selenomethionine, selenocysteine, selenocystine, Telluromethionine, tellurocysteine, molecular mass species Se-methylselenocysteine, selenocysteic acid, tellurocystine Se-methylselenomethionine, selenomethionine selenoxide, selenoniocholine, selenobetaine Other compounds Selenoproteins, selenoenzymes, Se–metal metallothioneins (e.g., Se–Hg, Se–Zn) 118R Analyst, December 1997, Vol. 122was applied to collect elemental Se and Te and SeIV, SeVI, TeIV and TeVI volatile compounds in air by using four sampling serial adsorption tubes. The thermal desorption was replaced by a sequential leaching (see Section 3.2.2) in order to preserve selenium and tellurium speciation.49 A method for the direct determination of selenium associated with atmospheric particulate matter has been proposed.It consists of an electrographite probe filter for filtering the particulates following the determination using a graphite probe furnace with atomic absorption spectrometric detection (detection limit 36 pg of Se).50 Charcoal possesses a good collection efficiency for the volatile alkyl selenides and was employed for the collection of methyl selenides evolved from incubation experiments on plants and soils.51–53 A glass U-tube (26 cm long, 6 mm diameter) packed with 3% OV-1 on Chromosorb W was employed to collect alkyl selenides evolved from lake sediments. 54 The trap, maintained at about 280 °C , gave 95–98% recoveries for dimethyl selenide and dimethyl diselenide. The trapped compounds were delivered to a gas chromatograph following thermal desorption.Volatile alkyl selenides evolved from incubated soil were collected by sweeping the incubation vessel headspace with nitrogen at 200 ml min21 and then trapped in a microtube containing 0.5 ml of hexane cooled in a dry-ice bath.55 1.2 Storage Conditions: Physical Storage in a cold place is the simplest and widely accepted method for the preservation of solid samples to be analyzed for selenium.Solid samples, biological material and sediments can be stored frozen during the period between sampling and analysis.43,56,57 For water samples, the presence of suspended matter may lead to composition changes through both sorption processes and bacterial action, but filtration, for storage purposes, is not always recommended except for waters containing large amounts of suspended matter.39 Bacterial activity can be reduced by storage at 4 °C or, better, by freezing at 220 °C.58 However, rapid freezing of spiked water samples in dry-ice followed by storage in a deep-freezer led to 12% losses of selenite and 48% losses of selenate after 3 d.Quick freezing in liquid nitrogen following by storage in a freezer gave only minimal losses of selenite and selenate.59 The major drawback of freezing procedures is the contamination of Table 2 Concentrations of selenium in samples of different origin Sample Source/type Total selenium/mg g21 Other selenium species/mg g21 Ref.Sediments Nagoya Harbour 0.1–11 319 Laurentian Trough 0.37–1.28 320 Pacific 0.05–0.4 321 Venice Lagoon 0.15–0.71 322 Central Venice 0.4–1.8 322 Lakes, mining area (Canada) 4.4–161 86 Lake (Antartica) 0.082–0.083 322 Arno river mouth 0.03–0.3 322 Rivers (Finland) 0.029–3.94 323 River (Germany) Me2Se (0.0010) Me2Se2 (0.0021) 119 Soil Non-seleniferous < 0.1–2.0 1 Seleniferous 1–2000 2 Brazil < 0.01–0.5 324 Canada 0.08–2.09 324 USA < 0.1–4.0 324 New Zealand 0.12–2.65 324 Finland 0.005–1.24 324 Norway 0.08–1.7 324 Sweden 0.16–0.98 324 Denmark 0.2–1.44 324 England/Wales 0.2–1.8 324 England (peat) 92–320 324 Ireland 0.02–0.07; 3–360 (peat) 324 Scotland 0.02–0.36 324 Coal Missouri 3.81–4.30 325 Illinois 2.04–2.75 325 Plants Grass 0.005–0.023 (Canada), 0.019–0.134 (France), 0.01–0.04 (USA) 324 Barley grain 0.027–0.042 (France), 0.002–0.11 (Denmark) 324 Oat grain 0.004–0.023 (Canada), 0.15–1 (USA), 0.003–0.054 (Denmark) 324 Wheat grain 0.001–0.117 (Australia) 324 Vegetables (USA) 0.150 (cabbage leaves), 0.064 (carrot), 0.130 (kale), 0.042 (onion), 0.036 (tomato) 324 Algae Queensland 0.063 326 South Australia (Phaeophyceae) 0.014–0.135ß Distribution of Se species investigated: 109 South Australia (Rhodophyceae) 0.153–0.434¶ 56–73% of the total Se associated with proteins 109 South Australia (Chlorophyceae) 0.053–0.264ƒ and 7–31% with amino acids 109 Fish Australia: Fish 0.35 ± 0.16ß Distribution of Se species investigated: 27 Molluscs 1.03 ± 1.3 ¶ 78–83% of the total Se associated with proteins 27 Crustaceans 0.5 ± 0.3 ƒ and 11–13% with amino acids 27 North Thyrrenian Sea: Nephrops norvegicus 1.6–3.7 327 Merluccius merluccius 1.2–3.0 327 Solea vulgaris 1.1–4.9 327 Eledone cirrhosa 2.0–11.2 327 Analyst, December 1997, Vol. 122 119Rthe dissolved aqueous phase due to rupture of phytoplankton cell membranes.39,58 For this reason, the samples must be filtered before the freezing step if the phase speciation also has to be determined. Even if freezing procedures seem to represent the best choice for the preservation of chemical speciation, none of the above methods appeared adequate for the preservation of samples to be analyzed for volatile methyl selenium species.Cutter59 was not able to detect any free methyl selenides in water samples and this was attributed to the inadequacy of the storage conditions. Air-tight containers were only good for 1 d; after that period a complete loss of volatile methyl selenides occurred.Cooke and Bruland60 performed in-field purging and cryogenic trapping into a glass tube packed with silanized glass beads. The volatile alkyl selenide species could be stored for up to 3 d in liquid nitrogen. Storage of water samples in flame-sealed glass (silylated) containers followed by deep freezing for the determination of dimethyl selenide in natural sea-water samples was recently reported by Tanzer and Heumann.42,61 Amoroux et al.62 performed analysis and calibration onboard during a cruise for the determination of volatile methyl selenides (MeSeH, Me2Se, Me2Se2) by immediately purging surface sea-water samples with helium and trapping the volatile compounds in a packed chromatographic trap kept cold in liquid Table 3 Concentration of selenium species in environmental water samples Concentration Sample Source units* Se total SeIV SeVI Se0,2II Me2Se Me2Se2 Me3Se+ Organo-Se Ref. Sea-water ng l21 < 5.0 80.2 59 ng l21 < 5.0 58.4 59 nmol l21 0.63 0.85 107 SeIV+VI Atlantic ng l21 1–5 42 Ocean ng l21 2–31 18–51 2–6 75 Mediterra- ng l21 0.01–0.075 62,259 nean sea (MeSeH, Me2Se, Me2Se2) nmol 0.1–0.7 nd-0.5† nd-2.7 140 kg21 pmol l21 4–20‡ 328 Pacific nmol l21 0.489– 0.038– 0.226– 0.086– 112 Ocean 1.930 0.715 0.765 0.669 Pacific nmol l21 0.638– 0.000– 0.122– 0.232– 112 Ocean 2.176 0.488 0.640 1.436 Pacific nmol l21 0.784– 0.266– 0.135– 0.088– 112 Ocean 1.655 0.716 0.737 0.702 Osaka nmol l21 0.05–0.31 0.09–0.50 0.14–0.77 329 Bay Estuary nmol l21 0.53 SeIV+VI 0.84 107 Freshwater River ng l21 2–16 3–302 5–12 75 River ng l21 14.5 26.5 nd 203 River ng l21 12.9 24.5 nd 203 River nmol l21 0.12–0.37 0.02–0.32 0.32–0.18 329 River ng l21 35.4 23.0 nd 203 River ng g21 0.20–0.35 0.051– 0.118– 183 0.056 0.28 River nmol l21 1.25 SeIV + VI 0.82 107 River ng l21 32–180 Lake ng l21 6.2 10.5 18.8 203 Lake ng l21 17.70 < 5.0 59 Lake ng g21 0.83 0.046 0.805 183 Lake ng g21 0.19 0.030 0.171 183 Lake nmol l21 0.04–0.36 0.00–0.36 0.00–1.97 329 Marsh ng l21 50.85 7.10 59 Pond nmol l21 943 38.3 271§ 60 Moorland pg g21 213 27 35 13 49¶ 183 lake 94· Moorland pg g21 240 141 47 < 10 18¶ 183 lake 28· Groundwater — nmol l21 130 13.2 60 — ng l21 nd-37.0 10.0–99.0 nd-8.9 203 — ng l21 nd-34 38.0–39.0 nd 203 — ng l21 nd 7.5–32.3 nd-6.7 203 — ng g21 13.66–15.66 0.59–1.29 12.26–15.09 183 — ng l21 25.3 1.9 6.9 330 — ng l21 216 2.0 194 330 — ng l21 504 7.9 418 330 — ng g21 1.80 0.13 1.63 183 * Mass concentration units are referred to selenium. † nd = not detected.‡ Selenoamino acids. § Dimethylselenonium species RMe2Se+. ¶ Neutral/basic organoselenium compounds. · Acid organoselenium compounds. 120R Analyst, December 1997, Vol. 122nitrogen. The trap was directly connected to the elementspecific detection system (see Section 4.2.3). Sunlight was found to have no significant effect on either SeIV or SeVI species at the 10 and 50 ng ml21 levels, during storage for up to 1 year and under different experimental conditions.63 1.3 Storage Conditions: Chemical The chemical composition of storage containers, their chemical pre-treatment and the additions of chemicals are the most important chemical parameters for achieving proper storage. Polyethylene and borosilicate glass are the most widely employed materials but they require to be pre-cleaned with strong mineral acids (HCl, HNO3) followed by pH conditioning to prevent sample contamination.6,39 Glass containers are preferred for the storage of the organometallic forms59 and, in some cases, silylation was performed for surface deactivation. 42,60 Acidification of natural water samples is a widely accepted method for the preservation of the total element but phase speciation can be altered by releasing processes from the particulate matter. However, excessive acid concentrations are reported to change the chemical speciation.A sea-water reference material certified for SeIV and other trace elements is stable for over 5 years if stored in a cool place. The samples are unfiltered, acidified with nitric acid to pH 1.6 and stored in polyethylene bottles (NASS-2, National Research Council of Canada). Selenate showed a 60% conversion59 into selenite after 7 d of storage in 4 mol l21 HCl whereas both selenite and selenate were preserved in 1 mol l21 HCl.56,59 Sea-water samples stored at pH 2 with HCl in either polyethylene or glass containers showed no changes in SeIV and total selenium concentration over a 4.5 month period.64 Under similar conditions the concentrations of TeIV, TeVI and total tellurium in sea-water samples were preserved for over 1 year of storage in polyethylene bottles at pH < 2 in HCl.8 Selenium losses in river, ground, snow-melt and tap water samples containing selenium at concentrations in the range 44.5–138 ng l21 were tested at 4 °C in a 1 l polyethylene container.65 No selenium losses are detected for up to 15 d of storage.For ground water the stability was up to 13 months. The stability of selenium compounds in purified water samples followed the order selenate > selenomethionine > selenite.65 Cobo et al.63 studied the stability of SeIV and SeVI species in pure aqueous solution under different conditions of acidity (H2SO4 at pH 2, or no acidification at pH 6), temperature (220, 20 and 40 °C) and container material (polyethylene or PTFE).Their conclusions were that the best conditions were 220 °C without the need for acidification.They did not find any significant variation in the concentration of selenium species over 12 months of storage, but at room temperature the maximum time of sample storage was found to be 2 and 9 months (pH 6) for polyethylene and PTFE containers, respectively. Selenite and selenate are more stable at 40 °C than at room temperature (pH 6 and polyethylene container) and the presence of chloride tends to stabilize both species. Sample acidification, although it does not resolve the storage problem due to the volatility of alkyl selenides, is thought to have an adverse effect on the organic chemical speciation of non-volatile selenium compounds.However, it was reported that dimethylselenonium compounds, (CH3)2Se+R (e.g., Semethylselenomethionine), are stable at acidic pH, but decompose rapidly at pH 8 with formation of dimethyl selenide.60 Selenomethionine slowly decomposes in hydrochloric or hydrobromic acid solutions to form volatile dimethyl diselenide.The reaction take place even in aqueous solutions containing selenomethionine acidified at pH 2 with HCl and stored at 4 °C.66 1.4 Reagents The collection and storage of samples require, in any case, the use of high purity acids and high purity water to perform acid leaching and pH conditioning of both collection and storage bottles.The use of high purity acids is even more recommended for the chemical preservation of water samples. With regard to selenium, most of the commercially available analytical-reagent grade mineral acids possess adequate purity.A higher degree of purity can be obtained by sub-boiling distilled acids, but at an increased cost. The selenium concentration in both analytical-reagent grade and sub-boiled HCl and in sub-boiled HNO3 was < 0.01 ng ml21 whereas higher selenium levels were reported for ultrapure HClO4 (50.1 ng ml21) and sub-boiled H2SO4 (50.06 ng ml21).67 The extreme low concentration of tellurium species in open ocean sea-water required a HCl with a tellurium content below 1 ng l21 in order to avoid sample contamination.It was prepared by passing analytical-reagent grade, 12 mol l21 HCl through an anion-exchange column (40 mm 3 12 mm id; AG1-X2, 100–200 mesh).3 A single passage reduced the tellurium level from 7 to 0.33 ng l21.3 Sodium tetrahydroborate reagent, employed in the hydride generation (HG) technique, can be purified from selenium by coprecipitation with metal hydroxides.67 The best results were obtained by dissolving 0.5 g of NaOH, 1 g of NaBH4 and about 50 mg of BaCl2 in 50 ml of water.Thereafter, the solution was filtered with a 0.01 mm membrane and diluted to 100 ml.The selenium concentration was lowered from 0.17 to @0.01 ng ml21.67 2. Sample Treatment 2.1 Digestion 2.1.1 Mineralization The mineralization of samples for the total determination of trace amounts of selenium and tellurium has been performed typically after wet acid digestion. For selenium the main problems arise from the peculiar chemical resistance of some organoselenium compounds (selenonium salts, selenoamino Table 4 Concentration of selenium species in air samples (concentrations in ng m23 as Se) Sample/site Me2Se Me2Se2 Me2SeO2 Se0 SeIV SeVI Ref.Near lake 0.9–1.0 0.5–0.7 44 (Belgium) Near lake 0.8–0.9 0.5–0.7 44 (Belgium) Missouri 0.92–3.05 Se Tot. 48 Missouri 0.97– 0.72– 0.09 48 1.13 0.81 Laboratory 2.21 2.64 0.75 49 air (Missouri) Laboratory 1.55 2.09– 0.55 48 air 2.30 (Missouri) Near lake 0.47 ± 0.35 ± < 0.2 46 (Belgium) 0.03 0.05 Smelta Belgium 1.41 ± 0.07 0.63 ± 0.1 0.3 ± 0.01 46 Fishing pond (Belgium): I < 0.15 < 0.30 < 0.20 46 II 2.40 ± < 0.30 18.8 ± 46 0.04 0.70 Analyst, December 1997, Vol. 122 121Racids) to conversion into inorganic forms. On the other hand, the use of elevated temperatures in open systems may result in serious losses by volatilization of selenium compounds formed during the digestion.5,68–70 The use of hydrochloric acid during digestion must be avoided owing the volatility of some selenium chlorides and oxychlorides.68 The availability of microwave digestion in pressurized PTFE bombs greatly reduced the digestion time and the risk of analyte losses.Nevertheless, the use of some chemicals, particularly perchloric acid, is not recommended owing to explosion hazards, and this may entail incomplete decomposition owing to the chemical resistance of some organoselenium compounds.Nitric acid digestion gave unsatisfactory recoveries of selenonium compounds and some selenoamino acids such as selenomethionine even in pressurized digestion systems.68–70 Nitric acid–sulfuric acid digestion requires the use of refluxing apparatus to avoid losses by volatilization.71 Nitric acid–perchloric acid digestion and nitric acid–perchloric acid– sulfuric acid digestion with a gradual increase of temperature up to 210 and 310 °C, respectively, have proved to be effective both in the complete mineralization of organoselenium compounds and in the prevention of volatilization losses.69–72 Nitric acid–perchloric acid–sulfuric acid wet digestion was adapted to an open microwave digestion system by Hocquellet and Candillier.73 The presence of perchloric acid at elevated temperature may create some risks due to explosion hazards.Furthermore, owing to the high oxidation potential of acid mixtures containing perchloric acid, most or part of the inorganic selenium is converted into selenate, which is then not suitable for most analytical techniques based on HG, derivatization to piazselenol or electrochemical detection.Therefore, a pre-reduction step for the conversion of selenium into selenite is necessary. Several chemical methods to convert selenate into selenite have been reported,6 the simplest being the reduction with HCl solutions at different concentrations and temperatures.Table 5 Concentration of tellurium species in environmental samples Source Concentrations units* Total Te TeIV TeVI Te0 Me2Te Ref. Sea-water Florida pmol l21 1.9–6.8 —† 7 Gulf of Mexico pmol l21 3.0 —† 7 Angola Basin pmol kg21 0.6–1.3 0.12–0.36 0.43–0.9 8 Panama Basin pmol kg21 0.5–1.4 0.1–0.55 0.5–1.14 8 N.Atlantic pmol kg21 0.8–1.6 8 Pacific pmol l21 0.46–1.39 0.12–0.37 0.37–1.04 3 Atlantic pmol l21 0.56–1.03 0.17–0.20 0.39–0.84 3 Canada pmol l21 2.5–10.8 0.72–2.01 0.57–3.77 3 Submarine hydrothermal water: I pmol l21 < 40 ± 20 3 II pmol l21 1310 ± 90 3 III pmol l21 < 20 ± 10 3 River-water Orinoco pmol l21 2.22 3 Mississipi pmol l21 4.17 3 Amazon pmol l21 10.3 3 Rainwater Bermuda Island pmol l21 0.58–7.68 3 Cambridge, MA pmol l21 2.6–6.1 3 Pacific pmol l21 2.64 3 Korea pmol l21 189–205 3 Florida pmol l21 4.0–2.6 7 Snow (Korea) pmol l21 15.0 3 Snow (Missouri) ng l21 4.2 331 Marine sediments N.Pacific ng g21 140–150 3 S. Pacific ng g21 25–880 3 Antarctic lake ng g21 < 5 322 Arno River, mouth ng g21 < 5–60 322 Venice Lagoon ng g21 5–93 322 Venice lagoon, ng g21 180–1000 322 old town centre Air Missouri ng m23 0.10–0.34 48 Laboratory air (Missouri) ng m23 0.93 0.20 0.26 0.47 48 Laboratory air (Missouri) ng m23 0.67–0.75 0.18–0.19 0.15–0.20 0.34–0.36 48 Laboratory air (Missouri) ng m23 0.21–0.26 0.05–0.06 0.11–0.13 0.05–0.07 48 Aerosol (Bermuda) ng m23 0.64–15.1 3 Aerosol (Missouri) ng m23 0.45 332 Other samples River sediments mg kg21 0.0002 119 Soil (Japan) mg kg21 0.010–0.16 333 Coal (Missouri) mg kg21 0.91–0.92 325 Coal (Illinois) mg kg21 0.58–0.84 325 Coal (Japan) mg kg21 0.023–0.093 334 Volcanic ash (Japan) mg kg21 0.27 333 Vehicle exhaust particles mg kg21 0.15 335 Petroleum ash mg kg21 0.55 333 Coal fly ash (Japan) mg kg21 0.25–0.37 334 Garlic ng g21 0.02 ± 0.005 3 Plants ng g21 0.7–6 324 Vegetables ng g21 0.013–0.35 324 * Mass concentration units are referred to tellurium.† No evidence of TeVI was found. 122R Analyst, December 1997, Vol. 122The redox–buffer properties of HBr in the presence of bromine were employed to convert all oxidation states of selenium into the tetravalent form.74,75 Hydrobromic acid–bromine digestion was able to recover quantitatively inorganic selenite and selenate, selenomethionine, 6-selenopurine, selenocystine and trimethylselenonium compounds.76 Bromine was generated in situ by the addition of KBrO3 solution to hydrobromic acid.The same digestion was successfully adapted for total selenium determination in aqueous and biological samples. In all cases, the final chemical form was inorganic tetravalent selenium.66,77 Hydrobromic acid digestion (1.8–9 mol l21 HBr) was able to recover traces of organoselenium species present at the parts per billion level in solution,78 but this was attributed to the presence of parts per million free bromine impurities in hydrobromic acid solutions. 66,76 Indeed, in a further investigation, the role of bromine was identified as its oxidative addition to divalent selenium to form bromoselenonium intermediates, which are dealkyklated by the bromide ion.The chain of the two reactions continues until all organic selenium has been dealkylated and hydrolyzed to inorganic selenite.66 An on-line HBr–Br2 microwave digestion method has been developed for total selenium determination by flow injection coupled with HG and electrothermal atomization atomic absorption spectrometry (FI–HG– ETAAS)79 and post column pre-treatment after high-performance liquid chromatographic (HPLC) separation followed by HG derivatization and ETAAS detection.80 Hydrobromic acid proved a good pre-reduction reagent and a suitable reaction medium for the determination of total selenium by fluorimetry following piazselenol derivatization with 2,3-diamino-1,4-dibromonaphthalene. 81 Yamada and Hattori82 digested soil extracts, prior to total selenium determination, in an autoclave at 120 °C for 30 min in the presence of potassium peroxodisulfate using an alkaline solution.After cooling, the reduction of SeVI to SeIV was performed in the presence of HCl–KBr at 100 °C for 15 min in the same autoclave.Lan et al.83 succesfully determined selenium in fish tissue after HNO3–H2SO4–H2O2 microwave digestion in pressurized PTFE vessels followed by an HCl reduction step. For the determination of total selenium in sea-water, Measures and Burton64 used a different approach: the sample was placed in a quartz tube and irradiated for 5 h with a 1250 W UV lamp in the presence of a borax solution containing dilute hydrogen peroxide. The use of hydrofluoric acid can be necessary in order to ensure complete dissolution of sediments, soils and particulate matter.In that case, the formation of volatile selenium fluorides SeF4 and SeF6, may be a problem in open digestion systems.5,84 Itoh et al.,85 using a multi-step digestion with nitric acid– sulfuric acid–perchloric acid followed by a second step with nitric acid–hydrofluoric acid and a final step with perchloric acid, obtained quantitative recoveries for selenium in different certified reference materials when the digestion was performed in a Teflon beaker heated on a hot-plate.The use of HCl– HNO3–HF digestion in microwave-heated PTFE bombs was successfully reported for the determination of selenium in lake sediments and mine tailings.86 For tellurium, fewer problems due to volatilization losses are found, since, in particular, the chlorides are not as volatile as for selenium and pre-reduction of TeVI to TeIV with hot hydrochloric acid is relatively fast and reliable,87 whereas prereduction with TiCl3 at room temperature is fast but must be controlled to avoid reduction to the elemental state.88 Digestion or leaching with aqua regia89,90 and nitric acid–sulfuric acid91 has been reported.Yoon et al.3 performed the digestion for total tellurium determination in aerosol particles, biological and geological materials by means of complete dissolution with HNO3, HClO4 and HCl–HF. After the dissolution, they performed a prereduction step by gently boiling the digest with 4 mol l21 HCl. 2.1.2 Protein hydrolysis In a recent study, Lei and Marshall92 emphasized that the major challenge is devising a procedure that liberates selenium compounds efficiently from biological samples where most of selenium is protein-bound. Identification of selenoamino acids present in proteins requires special conditions in order to perform the protein hydrolysis while preventing any chemical modification in the structure of selenoamino acids.Acid hydrolysis in 6 mol l21 HCl, at 110 °C and in an anaerobic environment, followed by ion chromatographic determination are the conditions normally used for protein analysis using amino acid analyzers. Under these conditions, Shepherd and Huber93 reported that selenomethionine was broken down after 7 h whereas Broderick et al.94 reported that both selenomethionine and selenocysteine are relatively stable up to 22 h of acid hydrolysis.However, Broderick et al.,94 in addition to acid hydrolysis, also investigated proteolytic digestion with pronase and prolidase and digestion with thioethanesulfonic acid. They concluded that a universal hydrolysis method did not exist for proteins containing both selenocysteine and selenomethionine.The protection of selenocystine with reducing agents, such as reduced glutathione or other thiol groups of biological origin, created an artifact on the selenomethionine elution peak. The proteolytic digestion produced inconsistent results for selenomethionine, depending on the kind of sample, whereas the thioethanesulfonic acid digestion completely destroyed selenocystine after 3 h and selenomethionine after 20 h.Protection of selenocystine before protein hydrolysis is better realized by derivatization reactions able to produce stable products. Forstrom et al.95 used iodoacetic acid and ethylenimine to form stable carboxymethyl and aminoethyl derivatives of the selenol group of selenocysteine, respectively. This reaction leaves unchanged the carboxylic and the amino groups, and further derivatization to N-acetyl and O-methyl derivatives had to be performed after hydrolysis for mass spectrometric (MS) identification.The derivatization by carboxymethylation prior proteins acid hydrolysis has been widely employed in the identification of Se-cysteine in plant proteins,96 in proteins from rat and monkey plasma,97 in glutathione peroxidase and other selenoproteins of rat tissues98 and in metallothioneins.99,100 For the identification of selenocysteine in glutathione peroxidase by MS, acid hydrolysis was performed only after the protection of the selenocysteine group.101,102 The enzyme was reduced with sodium tetrahydroborate(iii) followed by reaction with 1-fluoro-2,4-dinitrobenzene to form the Se-(2,4-dinitrophenyl) derivative.The presence of guanidine was found to be an important factor to obtain a near quantitative yield of the dinitrophenyl derivative. The inactivated enzyme was then hydrolyzed in 6 mol l21 HCl at 110 °C for 20 h under anaerobic conditions. Under these conditions the Se-cysteine derivative is stable but it was not found suitable for MS identification.For this reason, after hydrolysis, Se-(2,4-dinitrophenyl)- selenocysteine was converted into the more stable Se-methyl-N- (2,4-dinitrophenyl)selenocysteine with sodium barbital and methyl iodide (Smiles rearrangement). Identification by MS was performed on the methyl ester derivative.101,102 For the identification of selenomethionine in soybean proteins,57, low-temperature defatted soybean flakes were subfractionated.The 7S globulin fraction was subjected to a proteolytic enzyme digestion in a dialysis tube, first with pepsin in 0.1 mol l21 HCl at 37 °C for 15 h. After neutralization and lyophilization, the digestion was continued in the presence of pancreatin at pH 7.5 for 24 h at 37 °C and then for 5 min in a boiling water-bath.The digestion was completed by the Analyst, December 1997, Vol. 122 123Raddition of actinase E and incubation for 24 h at pH 7.5. After a further sub-fractionation with gel permeation chromatography and thin-layer chromatography, the dried fraction containing selenium was trimethylsilylated with N,O-bis(trimethylsilyl)- acetamide and analyzed by gas chromatography–mass spectrometry (GC–MS).The method was not suitable for the identification of selenocysteine and selenocystine since no precautions were taken to prevent their oxidation.57 For the direct determination of seleno amino acids in fish tissues Cavalli and Cardellicchio30 employed microwave heating- gas-phase hydrolysis of proteins. The sample (100 mg of purified dolphin liver proteins) was placed in a PTFE reaction vessel with 10 ml of 6 mol l21 HCl and 0.5% phenol solution, the latter added in order to protect easily oxidizable amino acids.The reaction vessels were then evacuated and purged with nitrogen, the pressure set at 55 lb in22 (temperature 150 °C). Hydrolysis was performed with microwave irradiation at 645 W for 25 min. Recoveries of added selenomethionine and selenocysteine were > 90%. 2.2 Extraction 2.2.1 Selective phase extraction Several procedures have been investigated for extracting selenium selectively from the different components of sediments, soils and particulate matter, in order to perform a phase speciation for the element. The extraction is typically performed with a sequential, selective dissolution of the different phases.The general, sequential extraction scheme proposed by Tessier et al.103 to investigate the particulate trace metal distribution in the ‘exchangeable’, ‘carbonate’, ‘iron and manganese oxide’, ‘organic’ and ‘mineral’ phases was discussed by Cutter56 for the determination of sedimentary selenium. The first observation was that the five sequential leaches (MgCl2 at pH 7.0; NaOAc–HOAc at pH 5.0; NH2OH·HCl in 25% HOAc at pH Å 2; H2O2–HNO3 followed by HOAc; HF and HClO4) allowed the determination only of the total selenium in each phase, since they destroyed the chemical speciation. Further, Cutter stressed the fact that only the above phase definitions were operative.The organic fraction, for example, represented all the selenium solubilized by oxidation and that could include organic and inorganic selenide in addition to the elemental selenium.The peroxide– nitric acid leach gave an incomplete recovery of the organic selenium and was replaced with the nitric aicd–dilute perchloric acid procedure used for total selenium. No selenium was detected in the mineral phase and the hydrofluoric acid– perchloric leach was found not to be necessary for the total Se determination.In relation to the total determination of selenium in geological material, Kuldvere104 found that selenium was quantitatively extracted by an overnight treatment at 95–100 °C with either aqua regia or reversed aqua regia. The procedure yielded selenium in the tetravalent form, which made unnecessary the pre-reduction step before its determination.Quantitative extraction of Se and Te in sediments was obtained by D’Ulivo et al.105 by using either aqua regia or reversed aqua regia in PTFE bombs heated in a microwave oven. The sequential extraction scheme of Tessier et al.103 was reexamined by Gruebel et al.106 in particular for the reductive dissolution of the iron phase and the oxidation of organic material.They found that some minerals adsorbed SeIV, also at low pH values, more strongly than SeVI. The release of the latter from minerals depended on pH, ionic strength and the concentration of competing ions such as sulfate. During the reductive dissolution of amorphous iron oxide with NH2OH·HCl at pH Å 2, SeIV adsorbed on montmorillonite could be released, causing over-estimation of selenium in the iron oxide phase.Under the same conditions, SeIV could be adsorbed on goethite causing under-estimation of selenium. The successive oxidative treatment for the determination of the organic-bound selenium oxidized part of the adsorbed SeIV to SeVI, then released it in solution, and an over-estimation of the organic selenium resulted.Selenium in the colloidal phase of natural water samples was determined by Takayanagi and Wong107 after fractionation with ultrafiltration. Water samples were filtered through a glass-fiber filter immediately after sample collection. Sub-samples of this filtrate were individually further filtered through UM2, YM5 and YM10 membranes. The cut-off points of these ultrafilters are at nominal molecular masses of 1000, 5000 and 10 000, respectively. Only the material passing through the UM2 ultrafilter was considered to be in the truly dissolved form.Total inorganic selenium and organic selenium were determined in each of the fractions including the sample without ultrafiltration. 2.2.2 Organic matter fractionation A sequential extraction method was proposed by Maher108,109 in order to investigate the biochemical association and the chemical forms of selenium in marine organisms and macroalgae.The samples were extracted at 25 °C and under mechanical shaking and sonification. The first step [2 3 50 ml of CHCl3–CH3OH (2 + 1) for 36 h followed by KCl addition] extracted the lipid–lipoprotein fraction. The second step [2 350 ml of ethanol–water (9 + 1) for 12 h] extracted the fraction containing the amino acids, organic acids and sugars.Third and final step (3 3 30 ml of 0.1 mol l21 TRIS–HCl solution containing 0.1% m/v NaCl + 1% m/v sodium dodecyl sulfate + 0.05 mol l21 dithiothreitol, pH 7.5, for 12 h) extracted the fraction containing proteins. In each fraction the total selenium was determined fluorimetrically after digestion with nitric acid– perchloric acid.In a subsequent study, Maher110 used a simplified procedure consisting of extraction with chloroform– ethanol (2 + 1) to remove lipids and lipoproteins followed by extraction with ethanol–water (9 + 1) to remove soluble amino acids, organic acids and sugars. The residue consisted of protein-bound selenium. Maher’s results showed that in marine algae most of selenium is present in the selenoamino acid and protein fractions and in marine animals most of selenium is associated with the protein fractions (see also Table 2).27,108–110 Abrams and Burau111 extracted organic matter from soil with the aim of quantifying and identifying organic selenium compounds associated with different fractions.The extraction was performed under a nitrogen flow using a mixture of NaOH and Na2P2O7 at a concentration of 0.1 mol l21 in each component.An extraction of 33% of total soil selenium was obtained. In the extracts, humic acids were precipitated at pH 1.8 whereas the soluble fulvic acids were fractionated into hydrophobic (acidic, neutral and basic) and hydrophilic fulvates using Amberlite XAD-8 resin with a pH gradient. Aono et al.112 performed a fractional separation, using Starikova’s method,113 of typical organic compounds present in natural waters: humic acids, organic acids, amino acids and carbohydrates.The separation from the inorganic selenium, prior to fractionation, was performed by ultrafiltration and/or gel filtration on Sephadex G-15. The organic compoundcontaining fractions were concentrated under vacuum and treated with hot acidic ethanol [HCl–ethanol (2 + 8)].The residues contained the humic acids. The ethanolic solutions were vacuum distilled and 1 ml of 2 mol l21 H2SO4 was added to the residue that was completely wiped with a membrane filter. Organic acids were obtained after 18 h by extraction with diethyl ether in a Soxhlet extractor.The residue was treated with hot acidic ethanol and evaporated in a vacuum evaporator. 124R Analyst, December 1997, Vol. 122Acidic acetone extraction [HCl–acetone (2 + 8)] of the residue recovered selenoamino acids whereas carbohydrates remained in the residue. Total selenium determination on each fraction was employed to quantify organic selenium compounds.Radiotracer experiments performed by the addition of [75Se]selenomethionine demonstrated that 93% of selenomethionine was recovered in the selenoamino acids fraction. 2.3 Preconcentration Preconcentration procedures are often necessary for the determination of selenium and tellurium because most analytical techniques do not possess an adequate sensitivity for their direct determination.In extreme cases, as in determinations of tellurium in natural uncontaminated waters, the use of two different preconcentration steps was applied to achieve reliable determination and speciation far below the ng l21 level.3,7 Preconcentration techniques, in some cases, also allow the separation of the analyte from the sample matrix, thus preventing interference effects in the final detection step.The gaseous sampling techniques applied to the volatile chemical species of selenium and tellurium also represent a useful preconcentration step prior the analytical determination of species which are present in air at or below the ng m23 level and have been described in Section 1.1.2. In addition, it must be emphasized that most separation techniques based on column liquid chromatography (LC) (Section 3.2.2) have the advantage of also achieving attractive preconcentration factors. 2.3.1 Cold trapping Selenium compounds that are already volatile or derivatizable to volatile chemical species can be easily preconcentrated from aqueous solutions by purge and cold trap procedures.59,60, 114 Cutter59 performed the preconcentration of hydrogen selenide after HG from inorganic SeIV.The hydride was trapped in a Pyrex U-tube (18 cm 3 6 mm od) kept at liquid nitrogen temperature. Another U-tube, cooled with dry-ice–propan-2-ol, was interposed between the hydride generator and the sample trap to remove water. Delivery of the trapped hydride to the AAS detector was simply accomplished by allowing the sample trap to warm to room temperature.With minor modifications the system was adapted to the determination of dimethyl selenide and dimethyl diselenide in water samples; a gas chromatograph was interposed between the above stripping and trapping apparatus and the AAS detector according to the scheme proposed by Chau et al.54 A quartz tube packed with silanized Chromosorb W (30–60 mesh), cooled at 2140 °C was used to trap hydrogen selenide after generation with sodium tetrahydroborate(iii) reduction.Water and aerosol were removed from the purging gas stream with a U-trap maintained at 270 °C. The trapped hydrogen selenide was swept to the AAS detector by thermal stripping at 100 °C under a stream of helium. The sensitivity (detection limit 6 pg) was increased by about 2–3 orders of magnitude with respect to the direct transfer method.67,115 A specially designed U-trap, partially packed with Teflon strands and immersed in liquid nitrogen, was employed by Apte and Howard172 to trap hydrogen selenide generated from 50 ml water samples.The gas stream was dried by allowing the water to freeze in an unpacked section of the Utube. The ice so formed was removed by immersing the trap in hot water between successive determinations.The trapped hydrogen selenide was swept to the AAS detection system with nitrogen while the trap was allowed to warm to room temperature. � Ornemark et al.116 determined total selenium in water samples by cold trapping the hydrogen selenide, evolved by borohydride reduction, in a U-trap filled with a small silanized glass-wool plug, after the conversion of selenium to SeIV.The water was removed by means of a U-trap maintained at 215 °C. For the measurement, the hydrogen selenide trap was immersed in an ethanol bath at 215 °C and swept to the atomizer with a stream of nitrogen. A special purge and trap apparatus has been described by Tanzer and Heumann61 for the preconcentration of volatile selenium, sulfur and iodine species from water samples.The system was built using silanized glass and Teflon and provided filtration on a 0.45 mm membrane prior the delivery of the sample to the purging unit. There, the volatile compounds were extracted under a stream of helium, passed through a potassium carbonate or Nafion drying tube and trapped in a capillary trap (250 3 0.32 mm id) at liquid nitrogen temperature. Thermal stripping with a hot air gun was employed to sweep the trapped compound to a gas chromatograph.Cold trapping systems have also been described for the purpose of multi-element trapping and analysis of volatile compounds such as hydride, alkyl and mixed hydride–alkyl derivatives of Se, Te, Sn, Ge, Pb, Bi, As, Sb and Hg.117,118,119 A simple PTFE tube (2.83 m31.6 mm id) cooled at liquid nitrogen temperature was employed for preconcentration of selenium hydride, stibine, arsine and stannane (not tellurium) before their GC separation with on-line photoionization detection. 117 A more sophisticated system has been adopted for the qualitative or semiquantitative determination of a number of volatile compounds, including methylated selenium and tellurium species, by GC–ICP-MS.118 The system consists of a drying tube (9.2 3 2.2 cm od) filled with anhydrous CaCl2 followed by a cryogenic glass trap (22 cm 3 6 mm id) filled with 10% SP-2100 (60–80 mesh).The system has been used for the determination of volatile species stripped from samples of domestic waste deposits118 and sediments119 before and after HG with NaBH4. 2.3.2 In situ trapping of hydrides The cold trapping methods cannot be applied to hydrogen telluride because of its low thermal stability causing decomposition during the stripping from the cold trap. The in situ trapping of hydride into a graphite atomizer is an useful alternative for the atomic spectrometric determination of the hydride-forming elements and some volatile organometallic species.120,121 The advantages of this techniques are the increased sensitivity (@ng l21) and the virtual elimination of the atomization interferences which are a serious problem in the electrothermal quartz tube atomizer.121 Concerning liquid-phase interferences, better control can be achieved by appropriate sample dilution.This technique has been applied exstensively for the determination of extremely low concentrations of tellurium in water and environmental samples after a further preconcentration step such as ion exchange122 and coprecipitation with magnesium hydroxide.3,7 Andreae7 trapped the hydrogen telluride directly in a graphite furnace atomizer, after reduction of the inorganic TeIV with sodium tetrahydroborate (iii), from a 25 ml volume of sample.Trapping was performed by thermolysis at 300 °C and it was followed by a determination step consisting in atomization at 2750 °C and detection by ETAAS. Applying a method similar to one employed in bismuthine trapping,123 Lee and Edmond8 trapped hydrogen telluride in a carbon rod atomizer. A further improvement was achieved by Yoon et al.3 through a slight modification of the carbon rod atomizer.This improved the collection efficiency and made it almost constant in the temperature range 20–800 °C. For routine work, the trapping was performed at 110 °C followed, for the Analyst, December 1997, Vol. 122 125Rdetermination, by an ashing step at 700 °C and atomization at 2000 °C (absolute detection limit 2–4 pg of Te). Trapping at 600 °C in a graphite furnace was also applied to hydrogen selenide, after the usual sodium tetrahydroborate(iii) reduction, from 20–50 ml of samples, followed by atomization at 2600 °C and AAS detection (detection limit 70 pg of Se).124 The efficiency of the in situ concentration of hydrides in a graphite furnace can be enhanced by the presence of palladium coatings.In this case the trapping temperature may range between 100 and 1000 °C and between 100 and 900 °C for selenium and tellurium hydrides, respectively.The sensitivities were improved by factors of 2.6 and 4.9-fold for tellurium and selenium, respectively, compared with those obtained in absence of the palladium coating.125,126 Palladium-coated graphite furnaces have also been used for the adsorption of alkyl selenides such as dimethyl selenide, diethyl selenide and dimethyl diselenide.55,127 The optimum adsorption temperatures were different for each of the alkyl selenides, ranging between 500 and 700 °C.55 The major disadvantage of palladium coating is the volatility of palladium itself, which is lost during the atomization and the clean-out step of the of the furnace.Long-term stability of the trapping coating coupled with a good trapping efficiency of hydrogen selenide were obtained by using a mixed Pd–Ir coating, which allows up to 300 complete trapping and atomization cycles.128 The use of iridium deposited on tungsten- or zirconium-treated graphite tubes has recently been investigated by Tsalev et al.129 and proposed as a permanent modifier for the in situ trapping of hydride-forming elements. 130,131 2.3.3 Coprecipitation and precipitation Coprecipitation of selenium and tellurium is mainly performed with the aid of insoluble metal hydroxides acting as collectors. The most popular are La(OH)3 132,133 Fe(OH)3 134,135,136 for both selenium and tellurium. In addition, coprecipitation with ZrIV has been employed for selenium137 and coprecipitation with MgII for tellurium.3,7 Coprecipitation with magnesium showed a 90–100% collection efficiency for both TeIV and TeVI.Andreae7 coprecipitated tellurium from sea-water with Mg(OH)2, simply by addition of sodium hydroxide to 4 l of sample. Magnesium chloride was added, before the addition of sodium hydroxide, to water samples containing low levels of magnesium.A similar preconcentration method was used also by Yoon et al.3 for the determination of tellurium in natural water samples. In spite of the preconcentration by coprecipitation, the extremely low concentrations of tellurium required a further preconcentration step. For this purpose, magnesium hydroxide, after separation, was dissolved in hydrochloric acid and tellurium was determined by HGAAS with in situ trapping in a graphite7 or carbon rod atomizer.3 Lanthanum hydroxide coprecipitation has also been employed for the preconcentration of selenium in environmental waters.132, 138 The difference between this method, in comparison with Fe(OH)3 coprecipitation, is the ability of the former to recover both SeIV and SeVI whereas the latter recovers only SeIV.However, some workers reported that sulfate seriously interferes with the recovery of selenate.138 Some workers recommended pre-reduction of selenate to selenite before coprecipitation.132 Coprecipitation with Fe(OH)3 is reported to allow a better separation than La(OH)3 from potential interfering elements such as Au, Cu, Ir, Pb, Pd, Pt and Rh present in complex sample matrices.136 Both SeIV and SeVI present in natural waters have been reduced to Se0 in the presence of hydrazine sulfate and coprecipitated with tellurium (reductive precipitation). 107,139, 140 Reductive precipitation with sodium tetrahydroborate(iii) in the presence of In–Pd141 and chelation with ammonium tetramethylenedithiocarbamate followed by coprecipitation wi its oxidation products in the presence of hydrogen peroxide142 have also been proposed as preconcentration methods for selenium. 2.3.4. Chelation–extraction These methods are typically employed for the preconcentration of selenium and tellurium present in the inorganic tetravalent state. Inorganic selenite and tellurite react with dithiocarbamates to give the corresponding bis(dithiocarbamates), which are easily extracted with organic solvents or solvent mixtures. Sodium diethyldithiocarbamate and ammonium tetramethylenedithiocarbamate are commonly used reagents.139,140,143–145 and selective extraction of tetravalent selenium and tellurium has been reported.EDTA is often added to control interferences from foreign ions. Dithiocarbamates have also been investigated in a resin form by Mentasti et al.141 In a comparison of different preconcentration methods, they concluded that poly- (dithiocarbamate) resins are suitable for selenium speciation studies, furthermore providing preconcentration factors as high as 100-fold.141 A sulfonic acid derivative of bismuthiol-II, loaded on an anion-exchange column, adsorbed SeIV selectively and quantitatively, through the formation of a selenotrisulfide, and it was applied to sea-water analysis.146 The formation of piazselenols is specific for selenite, and subsequent extraction into a suitable organic phase can result in a preconcentration factor of 100–500-fold when applied to natural water samples.64,75 Extraction of selenium and tellurium from sulfuric acid– bromide media with a mixture of hex-1-ene and toluene was applied by Torgov et al.147 in the analysis of minerals, rocks and ores.The system is based on the formation of Se and Te organometallic compounds which are irreversibly extracted into the organic phase. Tellurium(iv) can be extracted from natural waters as a halide-complex by tributyl phosphate.148 The solvent extraction of TeIV has been reviewed by Havezov and Jordanov.149 2.4 Control of Interferences 2.4.1 Hydride derivatization Hydride generation coupled with element-specific detection is one of the most powerful analytical methods for the determination of selenium and tellurium when present in an aqueous solution and as the tetravalent forms selenite and tellurite.121 The main problem with this technique is the interference effects that may arise in the different steps of the procedure.The most serious interferences are those generated from the presence of the ions of several transition elements such as Cu, Ni, Co, precious metals and the group of hydride-forming elements.150–156 This kind of interference can manifest itself at a wide range of concentrations: from parts per million levels as with the iron to a few parts per billion as with palladium.Other interference effects may arise from nitrites and nitric acid.154,155,157 The magnitude of a given interference effect depends on many experimental parameters: (a) the kind of hydride generation apparatus (batch, continuous, flow injection) and its design, (b) the chemical nature of the reagents employed in the hydride reaction, (c) their concentrations and their mixing order, 126R Analyst, December 1997, Vol. 122(d) the manipulation of evolved hydrides (direct transfer, collection) and (e) the atomization mode (flame, plasma, silica furnace, graphite furnace). As several workers used laboratory-made HG apparatus, notable discrepancies can be found among the interference figures reported. Consequently, wide discrepancies can be found among the methods proposed to remove interference effects.Further details on HG techniques can be found in reviews by Nakahara,158,159 and D�edina160 and in the comprehensive monograph by D�edina and Tsalev.121 Interference effects, occurring in the liquid phase and due to transition metals are more pronounced in batch generation apparatus and decrease in the order batch > continuous > flow injection.161 The simplest method allowing better control of liquid phase interferences is sample dilution, whenever possible.The tolerance of selenium to the interference effects of Cu, Ni, Co and Fe was considerably augmented by increasing the HCl concentration from 0.5 to 5 mol l21 in the batch reaction apparatus.162 Similar beneficial effects have also been reported for the determination of selenium in the presence of several other transition metal ions.151 A further improvement in the tolerance limit of selenium towards NiII and CoII interferences was obtained by decreasing the concentration of the borohydride reducing solution from 3 to 0.5%.163 In the case of more serious interference problems, the use of masking agents has required.The use of effective masking agents against transition metal interferences has not been reported for selenium and tellurium determination by HG. The use of l-cysteine and thiourea seems to produce poor results and in some cases the interference effects become worse164,165 The use of cyanoborohydride157 is able to mask both copper and nickel interference on selenium.The presence of iron(iii) was found to be effective in masking the interference of nickel and copper ion in the determination of tellurium166 and in masking the interference of copper and bismuth in the determination of selenium.85 For both selenium and tellurium, reduction with borohydride in alkaline media followed by acidification was found to be effective.167–169 In this case, the use of complexing agents such as diethylenetriaminepentaacetic acid or EDTA is effective in alkaline media for some transition metals such as Cu, Co and Ni.The tolerance limits, better than 250 mg dm23, appears more than adequate for the analysis of environmental samples for selenium and tellurium. The addition of potassium iodide (0.1–1 mol l21 ) to the borohydride reducing solution has also been proposed in order to remove several kinds of interferences.71 The role of halide ions (iodide, bromide and chloride) was explained by their catalytic action, which produces an acceleration in the reduction rate of selenite to selenide.An important consequence of this behavior is that in presence of KI, the amount of NaBH4 necessary to obtain complete reduction of selenite is lowered by more than one order of magnitude.The pre-treatment of 20 ml of sample with a mixture of 3 ml of 20% KI solution, 8 ml of concentrated HCl and 1 ml of 5% thiourea solution was found to prevent the interference of Cu, Ag, Au, Fe, Ni, Co, Pt, Pd and Os, present at a 106-fold mass ratio with respect to the analyte, in the simultaneous determination of Se, Te and the other hydride-forming elements.170 Interferences from nitric acid, nitrate and nitrite ions may be a serious problem in systems employing cold trapping of hydrogen selenide,157,171 owing to the formation of nitrogen oxides, in particular after digestion procedures using nitric acid.71 These types of interferences are removed by the addition of sulfamic acid157 or sulphanilamide.171,172 Nitric acid and nitrite gave lesser problems in batch generation systems with the direct transfer of evolved hydrogen selenide.78 Hydrofluoric acid does not generate interference problems for either selenium or tellurium at levels up to 0.3 and 0.6 mol l21, respectively, by using continuous flow generation and miniature diffusion flame atomization.105 The use of separation techniques, e.g., coprecipitation, solvent extraction 7,87 or separation by ion-exchange chromatography (IC),173,174 can be also employed to prevent interference effects.Natural organic substances and proteins dissolved in the sample interfere in the HG of selenium. Organic matter can be removed by LC on Amberlite XAD-8 resin at pH 1.6–1.8, while preserving the speciation of inorganic selenium,175 but the separation of inorganic selenium species from the organic matter was not always satisfactory, still producing some residual interference.56,111 � Ornemark and Olin176 improved the separation of SeVI from the organic matter present in freshwater samples by first passing the sample through Amberlite XAD-8 resin.After this step the amount of organic matter present in the eluate was decreased to 20&nd%.Then SeVI was preconcentrated on a strong anion-exchange column, Dowex 1-X8 (chloride form),177 from weakly acid solution and eluted with 5 mol l21 HCl. The residual amount of organic matter was estimated to be in the range 2–10%, with SeVI being preconcentrated by a factor of 10. 2.4.2 Piazselenol derivatization Reaction of aromatic o-diamines with SeIV to form the corresponding selenodiazoles (piazselenols) is one of the most commonly used derivatization procedures prior to the determination of selenium by fluorimetry or GC.Interferences may occur both during the derivatization–extraction process and in the detection step. In the fluorimetric determination of selenium in sediments, 178 control of interference was achieved with a buffer composed of acetate–hydrochloric acid, EDTA, sodium fluoride and ammonium sulfamate, used before the reaction of SeIV with 2,3-diaminonaphthalene (DAN).Under the above conditions the presence of 104-fold excess of Fe, Mn, Cu, Zn, Pb, Ni, Sn and Al, and a 105-fold excess of nitrate, phosphate and sulfate did not interfere. Interferences, however, were reported for 2 3 104-, 104-, 103- and 103-fold excesses of iron, nitrite, thiosulfate and cysteine, respectively.The interference of FeIII was complex because it reacts with DAN, thus enhancing the fluorescence intensity, but at the same time decreases the efficiency of Se–DAN piazselenol formation. It was also shown that coprecipitation with tellurium in a strongly acidic medium was effective in removing all the interference problems, except for thiosulfate, the interference of which was greatly reduced, however.178 Gas chromatographic detection of volatile piazselenol derivatives is typically performed using electron-capture detection. This can be prone to artifacts if reagent impurities or matrix compounds are co-extracted with piazselenol into the organic phase.Uchida et al.,75 in the determination of selenium in water samples, washed the acidified samples with toluene to remove organic phase-soluble matter before the formation of the 4,6-dibromopiazselenol. An alternative method based on the coprecipitation of selenium with lanthanum hydroxide was employed for biological samples after digestion.179 The presence of concomitant elements, also at concentrations between 10 and 100 times their sea-water maximum concentration levels, did not interfere with the GC determination of selenium using 4-nitro-1,2-diaminobenzene derivatization in presence of EDTA; the latter can be omitted in the analysis of oceanic water samples.64 Analyst, December 1997, Vol. 122 127RThe concentration of perchloric acid was found to affect the extraction efficiency of 5-nitropiazselenol180 and it is commonly used for the clean-up of the organic extracts containing the piazselenol derivatives to be analyzed.179,181,182 3 Speciation It has been recognized that environmental selenium can be present in different inorganic (SeIV, SeVI, Se0 and Se2II) and organic chemical forms.40,46,49 The latter may include, in addition to the simple methyl selenides, also oxidized methylated selenium compounds, selenonium compounds, selenoamino acids and their derivatives.40,46,60,183 Much less information is available on environmental tellurium.The natural occurrence of inorganic TeIV, TeVI (refs. 3, 7 and 49) and Te0 (ref. 49) has been reported, while the presence of organotellurium was hypothesized in sea-water on the basis of the difference between the concentrations of total tellurium and [TeIV + TeVI] in the same natural waters samples,3 while only recently dimethyl telluride has been identified in river sediments (see Table 5).119 Several speciation schemes are available for selenium.The most complete appear to be that one proposed by Cooke and Bruland60 for dissolved selenium in waters and that proposed by Velinsky and Cutter for sediments.43,184 The scheme proposed by Tanzer and Heumann183 for the speciation of dissolved selenium in natural water is limited to the non-volatile dissolved selenium fraction.It is based on the independent determination of single species after their chromatographic separation. The sum of the single species is then compared with the total selenium content determined using an independent procedure.Several chemical reactions are employed for the derivatization of selenium and tellurium species in order to realize speciation schemes involving either selective extraction or chromatographic techniques. Some of them have already been cited in previous sections. A summary of derivatization reactions is given in Table 6.Several recent reviews on selenium speciation can be found in the literature.185–190 3.1 Selective Extraction 3.1.1 Natural waters Inorganic SeIV can be selectively extracted into organic phases after reaction with several complexing or chelating agents, as already described in Section 2.3.4 . Among them the derivatization- extraction of piazselenols appears to be the most reliable.The formation of gaseous derivatives using borohydride reduction techniques directly on the water sample without any pre-treatment cannot be considered selective for inorganic selenite, because several simple organoselenium compounds are simultaneously derivatized to volatile forms (see Section 3.2.1 and Table 6) Inorganic SeVI and TeVI are usually determined after conversion into their tetravalent forms, mainly because they cannot be directly detected by most analytical procedures.The reduction is performed mostly with HCl using various concentrations, temperatures and reaction times,3,59 but HBr has also been used.75 This approach has been critically revised and discussed by different workers. The main problem with these reduction procedures is the non-selectivity of the reduction process itself, so that an undetermined fraction of the other inorganic(0, 2II) and organic species may be converted into the tetravalent form during the HCl or HBr reduction step, thus causing an over-estimation of the hexavalent inorganic form.66,78,107,176,140,183 The separation of SeVI and TeVI from the inorganic tetravalent species and from organic species is typically performed by ion exchange.3,176,177,183 The fraction containing SeVI or TeVI is then treated with HCl or HBr for their conversion into the tetravalent state.Considering that the organic selenium content is not directly determined but is estimated by difference between the total selenium and total inorganic selenium (SeIV + SeVI) the accuracy of the estimation is thus strictly related to the selectivity associated with the separation of inorganic species from the organic species.� Ornemark and Olin176,177 developed a two-step method based on filtration through an XAD-8 column followed by selective preconcentration of SeVI on the strong anion exchanger Dowex 1-X8. After elution only 2–10% of organic matter was present in the eluate containing SeVI, which resulted in a preconcentration factor of 10.Nevertheless, they could not exclude the over-estimation of SeVI due to the presence of residual organic matter in the eluate. Reddy et al.191 proposed a method for the selective field extraction of both selenite and selenate ionic species present in groundwater samples in the range 22–151 mg l21 and in the presence of sulfate at levels as high as 11.3 g l21.Selenite and selenate were adsorbed on CuO particles at pH 5.5 and desorbed at pH 12.5. Ion pairs such as MgSeO4 were not adsorbed. Considering that selenate was determined by HG after reduction to SeIV with hot 7 mol l21 HCl, the previous consideration on extraction selectivity applies also to this method.Concerning organoselenium compounds, the only selective derivatization–extraction procedure appears to be that reported for trimethylselenonium compounds. Natural water samples are treated in highly alkaline solution at 90 °C in order to produce thermal decomposition of trimethylselenonium. The evolved gaseous dimethyl selenide was trapped in a concentrated nitric acid solution with a total average recovery of 80%.183 The selectivity of this reaction, however, was not tested with other dimethylselenonium compounds such as Me2Se+2R (e.g., Se-methylselenomethionine). 3.1.2 Sediments Velinsky and Cutter184 developed an analytical scheme for selenium in sediments based on selective extractions to provide information about the concentration of SeIV + VI, Se0, inorganic Se2II or ‘pyrite-Se’, organic Se and total selenium. An alkaline leach with 1 mol l21 NaOH for 4 h in an ultrasonic bath was an effective procedure for the recovery of inorganic SeIV and SeVI from biological particles and mineral sediments, irrespective of their phase distribution and without changes in chemical speciation.After the leaching step, the organic matter was removed from the solution by extraction with an XAD-8 resin column.56 The reaction between Se0 and sulfite to form soluble selenosulfate was employed for the selective determination of elemental selenium in sediments.184 Elemental selenium was then selectively extracted from sediments by means of a sulfite leach in a ultrasonic bath at pH 7 and determined by HGAAS after conversion into SeIV.Inorganic sedimentary forms of Se2II, defined as ‘pyrite-Se’, may be present in sediments in the form of FeSe2, FeSe and FeSSe. They were selectively converted into hydrogen selenide by reaction with CrII in acidic solution. The hydrogen sulfide evolved was separated in a Porapak-PS chromatographic column to avoid interference in the AAS determination of selenium.184 The determination of pyrite-Se was performed only after the extraction of elemental selenium to avoid partial mobilization of Se0 from the CrII treatment.Total selenium was determined after a nitric acid– perchloric acid digestion procedure. The organic selenium was then determined by difference from total selenium and the sum of determined inorganic species. All the procedures were verified to possess an acceptable degree of extraction efficiency and selectivity and were applied to the study of the geochemical cycle of selenium in a coastal marsh.43 Some of the results are reported in Table 7 and, for the sake of simplicity, are limited for each sampled site to the fraction of the sediment exibiting 128R Analyst, December 1997, Vol. 122the maximum and minimum concentration of total selenium along the the sampled core. 3.1.3 Biological samples Selenomethionine was reacted with CNBr to form homoserine and methyl selenocyanate, CH3SeCN.93 The latter can be extracted into chloroform, digested and converted into a piazselenol for GC determination.192,193 The method was employed for the indirect determination of selenomethionine in selenium yeast and for the determination of the bound selenomethionine in plants and biological materials. The most serious interference problems derived from the presence of other compounds containing the CH3SeR group (e.g., selenoethers, Se-methylselenocysteine and Se-phenylselenomethionine), which also reacted with the cyanogen bromide to form CH3SeCN.193 The derivatization with CNBr was extended to the identification of telluroamino acids in tellurium yeast.Also in this case all the telluroamino acids containing the CH3Te group (telluromethionine, Te-methyltellurocysteine) react with CNBr reagent to give CH3TeCN.36 3.2 Chromatographic Methods 3.2.1 Gas chromatography Volatile alkyl selenides and alkyl tellurides can be separated by means of various GC methods without any derivatization step and can then be detected using element-specific or selective detectors. In early work, separations were performed using packed glass columns interfaced to the element-specific detector through an external transfer line made of stainless steel54 or nickel.44 In more recent work, the trends has been Table 6 Summary of derivatization reactions for selenium and tellurium species Species Reagent Reaction products and comments Section SeIV NaBH4 Reduction to H2Se. SeVI not reduced.Not specific for SeIV; see the reaction of different organoselenium compounds with NaBH4 2.3.1; 2.3.2; 2.4.1 SeIV Aromatic o-diamines Formation of 2,1,3-arenoselenodiazoles (piazselenols). Specific for SeIV. No reaction for TeIV 2.4.2; 2.3.4; 3.2.1; 4.1.1 SeIV Sodium tetraethylborate, NaB(C2H5)4 Formation of diethyl selenide.Not specific for SeIV; dialkydiselenides, R2Se2, form ethylalkylselenides Et-Se-R 3.2.1 SeIV Thiols (RSH) Formation of bis(alkythio)selenides, RSSeSR (also termed selenotrisulfides). Not very stable, except for some particular thiols (e.g., penicilamine) 2.4.4; 4.3.1 SeIV Dithiocarbamates Formation bis(dithiocarbamates) 3.2.3; 2.4.4 SeVI HCl or HBr (various concentrations and temperatures) Reduction to SeIV.Not selective for SeVI. Organoselenium compounds can be degraded or converted into inorganic SeIV 1.3; 2.1.1; 3.1.1 TeIV NaBH4 Reduction to H2Te. TeVI is not reduced 2.3.2; 2.4.1 TeIV Dithiocarbamates Formation bis(dithiocarbamates) 2.4.4 TeIV (4-Fluorophenyl)MgBr Formation of Te(C6H4F)2 after chelation–extraction of TeIV as a bis(dithiocarbamate) into the organic phase 3.2.1 TeVI HCI (various concentrations and temperatures) Reduction to TeIV.Not selective for TeVI. The same considerations made for the reduction of SeVI to SeIV with HCl can be applied also in this case 1.3; 2.1.1; 3.1.1 Se2II ‘pyrite’ Se CrII Formation of H2Se in acidic solution 3.1.2 Se0 Sulfite Formation of soluble senenosulfate 3.1.2 RASeSeR, RSeO(OH) NaBH4 Reduction to selenol, RSeH 3.2.1 R2SeO, R2SeO2 NaBH4 Reduction to selenides, R2Se 3.2.1 R(CH3)2Se+ NaBH4 Formation of (CH3)2Se 3.2.1 (CH3)3Se+ NaOH (hot) Formation of (CH3)2Se.Probably not specific for (CH3)3Se+ 3.1.1 Selenols (RSeH, H2Se) 1-Fluoro-2,4-dinitrobenzene, F- (DNP) Formation of RSeDNP or Se(DNP)2 for H2Se 4.3.2 Se-cysteine F-(DNP) Se-cysteine reacts like alkaneselenols to form CH(COOH)(NH2)CH2SeDNP, which can be rearranged to more stable CH3SeCH2CH(NHDNP)(COOH).The other Se-amino acids react to form RCH(COOH)NHDNP 2.1.2; 3.2.3 Se-cysteine Iodoacetic acid, ICH2COOH Carboxymethylation. Formation of –SeCH2COOH. Free carboxylic and amino functional groups are left unchanged 2.1.2; 4.3.2 Se-cysteine Ethylenimine, (CH2CH2)NH Aminoethylation.Formation of –SeCH2CH2NH2. Free carboxylic and amino functional groups are left unchanged 2.1.2 Se-cysteine N-(Iodoacetylaminoethyl)- 5-naphthylamine-1-sulfonic acid Formation of fluorophore for HPLC with fluorimetric detection 3.2.3; 4.2.1 Se trisulfides 7-Fluoro-4-nitrobenz-2,1,3-oxadiazole Formation of fluorophore for HPLC with fluorimetric detection 3.2.3; 4.2.1 Se-amino acids Various silylating agents Formation of various volatile O- and N-trimethylsilyl derivatives. Not specific for Se-amino acids 3.2.1; 4.3.2 Se-amino acids Ethyl chloroformate, ClCO2Et Simultaneous, fast derivatization of several functional groups: carboxylic group is derivatized to –COOEt; amino group is derivatised to –NHCO2Et and selenol group is derivatized to –SeCO2Et 3.2.1 Se-methionine CNBr Formation of CH3SeCN.Not specific; organoselenium compounds containing the MeSe–group give the same reaction product of Se-methionine 3.1.3; 4.1.1 Te-methionine CNBr Formation of CH3TeCN. Not specific; organotellurium compounds containing the MeTe–group give the same reaction product as Te-methionine 3.1.3; 4.1.1 Total selenium HBr + Br2 (hot) Conversion of selenoamino acids, selenonium ions and inorganic selenium species into SeIV 2.1.1 Analyst, December 1997, Vol. 122 129Rtowards the use of capillary columns35,42,61 and wide-bore capillary columns194 directly connected to the detection system.Chasteen et al.35 achieved the separation of sulfur, selenium and tellurium compounds, for a total of 10 species, evolved from microbial experiments (see Table 8), by use of a 30 m 3 0.25 mm id DB-5 capillary column with a 1 mm film thickness of methylphenyl(5%)silicone. The compounds could be detected simultaneously and selectively with a fluorine-induced chemiluminescence detector (see Section 4.1.1). Derivatization techniques for selenium determination are widely employed in GC.Piazselenol derivatives, specific for the determination of inorganic SeIV, were separated by means of either packed glass columns181,195,134 or capillary columns.182 The use of capillary columns resulted in the separation of the piazselenol peak from the background also in the absence of a clean-up step. Derivatization at 100 °C reduced the reaction time for quantitative piazselenol formation to 5 min.182 Sensitive detection was achieved with electron-capture detection but element-specific detectors are also employed for the detection of volatile piazselenols.196 The GC separation of selenoamino acids could be performed after their conversion into less polar, more volatile derivatives.Conversion to the corresponding trimethylsilyl derivatives was adopted by Caldwell and Tappel.197 Recently, for the identification of selenomethionine in soyabean proteins and soil extracts, the reagents N,O,-bis(trimethylsilyl)acetamide57 and N-methyl- N-(tert-butyltrimethylsilyl)trifluoroacetamide,111 respectively, were used as silylating agents.Janak et al.198 performed the separation of selenomethionine, selenoethionine and selenocysteine by capillary GC after their rapid derivatization with ethyl chloroformate.This reagent has the advantage of performing simultaneous derivatization of the carboxyl group to the ethyl ester while the amino group and the selenol group are esterified to –NH(CNO)OEt and –Se(CNO)OEt, respectively.199 The reduction of SeIV to hydrogen selenide, with the classical NaBH4 reduction, cannot be considered selective for inorganic selenite.It has been reported that other oxidized methylated species such as dimethyl selenoxide, dimethylselenone and dimethylselenonium species can be reduced to volatile dimethyl selenide.59,60,114 Methaneseleninic acid60 and diselenides66 can be converted into volatile methaneselenol and RSH, respectively, and alkyldimethylselenonium compounds give dimethyl selenide following reaction with sodium tetrahydroborate( iii).60 For this reason, the reaction with sodium tetrahydroborate(iii) was employed in pre-column derivatization for GC.The gaseous species evolved from natural water samples, following borohydride reduction, were purged with nitrogen and collected in a cold trap packed with 5% OV-3 on Chromosorb W HP (80–100 mesh).The controlled warming of the trap from 2196 to 100 °C in 4 min allowed the separation of the trapped selenium compounds, which were then detected on-line using an AAS Se-specific detector.60 A similar approach was adopted by Masscheleyn et al.114 who used an empty Utrap. The conversion of several inorganic selenium compounds, Na2Se, SeO2, Na2SeO3 and Na2SeO4, into diethyl selenide was first reported by Clark and Craig200 by using sodium tetraethylborate as derivatizing reagent.The reaction was carried out in sealed vials in HCl at pH 1. The gas-phase products were identified by GC–AAS and GC–MS. Diethyl selenide was the only reaction product, irrespective of the chemical form of selenium. Volatilization of SeIV into diethyl selenide with sodium tetraethylborate was employed for the determination of selenium using capillary GC coupled with microwave-induced plasma AES.201 The sample acidity was 4 mol l21 in HCl, and under this condition diethyl selenide was generated from SeIV with an SeIV : SeVI selectivity ratio of better than 1 : 10.Reaction of sodium tetraethylborate with dimethyl selenide and dimethyl Table 8 Methylated sulfur, selenium and tellurium compounds detected in microbial headspace samples by capillary GC coupled with fluorineinduced chemiluminescence detection.Data taken from ref. 35 Compound* Retention time/min Detected in bacteria and/or fungi† CH3SH 2.73 Bacteria CH3SeH 3.55 Bacteria CH3SCH3 4.35 Bacteria CH3SeCH3 5.25 Bacteria, fungi CH3TeCH3 6.55 Bacteria, fungi CH3SSCH3 7.55 Bacteria CH3SeSCH3 8.45 Bacteria CH3SeSeCH3 9.25 Bacteria, fungi CH3SSSCH3 10.25 Bacteria CH3TeTeCH3 11.25 Fungi * All biologically methylated compounds were verified by GC–MS analysis, except for dimethyl trisulfide, which was identified by its GC retention time.Methylated Se and Te compounds were only seen when cultures were supplemented with inorganic Se or Te salts. † Bacteria: Pseudomonas fluorescens.Fungi: Acremonium falciforme and Penicillium citrinum). Table 7 Speciation of selenium in sediments Concentration Sample Source Depth/cm units Total Se SeIV+VI Se0 Pyrite-Se Se(org) Ref. Marine sediment (coastal Great Marsh 6–9 mg g21 0.95 ± 0.09 0.068 ± 0.009 0.32 ± 0.02 < 0.025 0.55 ± 0.10 184 salt marsh)* 33–36 0.45 ± 0.03 0.026 ± 0.002 0.14 ± 0.03 0.044 ± 0.005 0.24 ± 0.04 Great Marsh, 4 April 1985 0–2.9 nmol g21 8.86 ± 0.25 0.481 ± 0.089 4.69 ± 0.38 < 0.32 3.80 ± 0.51 43 23–25.9 3.80 ± 0.13 < 0.01 1.65 ± 0.13 < 0.32 1.90 Great Marsh, 19 June 1985 19.1–21.8 nmol g21 6.71 ± 0.63 0.659 3.54 ± 0.38 < 0.32 2.53 ± 0.76 43 38.8–42.0 2.66 ± 0.38 0.144 ± 0.013 1.01 ± 0.13 < 0.32 1.65 ± 0.38 Great Marsh, 26 March 1986 9–12 nmol g21 12.4 ± 0.4 1.90 ± 0.13 6.84 ± 0.51 < 0.32 3.55 ± 0.63 43 36–39 4.43 ± 0.13 0.342 ± 0.038 1.39 ± 0.38 < 0.32 2.15 ± 0.38 Great Marsh, 26 June 1986 2.9–5.9 nmol g21 8.87 1.39 ± 0.13 4.69 ± 0.25 < 0.32 2.79 ± 0.63 43 31.6–34 5.19 ± 0.13 0.380 ± 0.051 1.27 ± 0.25 < 0.32 3.55 ± 0.25 Freshwater sediments Hyco Reservoir 0–2 mg g21 21.1 ± 0.03 0.9 ± 0.04 17.1 ± 1.1 < 0.025 3.1 ± 1.1 184 8–10 34.0 ± 0.5 5.20 ± 0.20 29.1 ± 1.2 < 0.025 nd Philpott Lake 0–2 mg g1 1.02 ± 0.03 0.04 ± 0.006 0.27 ± 0.01 < 0.025 0.71 ± 0.04 184 8–10 0.76 ± 0.05 0.01 ± 0.004 0.50 ± 0.03 < 0.025 0.25 ± 0.05 * Determination performed after separation from pore water. 130R Analyst, December 1997, Vol. 122diselenide was investigated.Whereas dimethylselenide was left unchanged, the latter was converted into methylethyl selenide. The indirect determination of selenomethionine using the CNBr reaction (see Section 3.1.3) can be performed also by directly analyzing the CH3SeCN reaction products by GC. In this way, the risks of interference from other selenoamino acids (e.g., Se-phenylselenomethionine) are reduced but the method cannot be considered selective because all species containing the –SeCH3 group are able to produce methyl selenocyanate. 193 Aggarwal et al.202 developed a method for the derivatization of TeIV into volatile compounds based on reaction with a Grignard reagent. After digestion, drying, redissolution and pH adjustment the sample containing TeIV was chelated with lithium bis(trifluoroethyl)dithiocarbamate.The chelate, TeL2, was extracted with toluene, the extract was evaporated to dryness at 60 °C under a stream of argon and the residue was reacted with the Grignard reagent, (4-fluorophenyl)magnesium bromide, in dry diethyl ether to give the final product, Te(FC6H4)2 . After the excess of Grignard reagent had been destroyed, the tellurium derivative was extracted with toluene, the extract was evaporated to dryness and the residue was reconstituted in methylene chloride.The method was found to be suitable for the GC–MS isotope dilution determination of tellurium in urine using 120Te as internal standard and was found to be free from appreciable memory and carry-over effects. 3.2.2 Liquid chromatography Several workers have performed a preliminary fractionation of selenium and tellurium species before their analytical determination, in order to avoid errors during the speciation studies.Non-polar Sep-Pak C18 resin was employed to isolate the organic selenides dissolved in waters that could be associated with soluble peptides, proteins and higher molecular mass organic compounds. The polar free amino acids, such as selenomethionine, were not quantitatively recovered.60 The trimethylselenonium ion was isolated and preconcentrated from natural water samples using a Dowex 50W-X8 cation-exchange resin in the H+ form and in presence of thiosulfate.183,203 In the selenium speciation of saturation soil extracts, Amberlite XAD-8 resin was found to isolate hydrophobic organic selenium.204 Tanzer and Heumann,183 in speciation studies of selenium in natural waters by LC–isotope dilution MS, employed Amberlite XAD-2 resin for the isolation of neutral and basic organoselenium species at pH 8, the acidic organoselenium compounds being retained at pH 3.Anion-exchange resins are employed in the separation of inorganic species of selenium and tellurium from water samples. Both SeIV and SeVI were isolated on Dowex 1-X4 resin and simultaneously eluted with 1 mol l21 HCl.203 AG1-X2 resin was employed for the separation of TeIV from TeVI after their preconcentration from sea-water samples by magnesium hydroxide coprecipitation.3 AG1-X8 anion-exchange resin in the chloride form isolated both SeIV and SeVI from water samples.The selenite was then eluted with 1 mol l21 HCOOH, whereas the selenate was eluted with 3 mol l21 HCl after washing the column with water.183 Muagnoicharoen et al.49 extensively applied ion exchange to the speciation of selenium and tellurium in the atmosphere. Selenium and tellurium inorganic species, adsorbed from air on gold-coated quartz beads (see Section 1.1.2), were leached successively with (i) warm distilled water at 60 °C for SeIV, SeVI and TeVI, (ii) 1 mol l21 HCl for TeIV and (iii) 3 mol l21 HNO3 for Se0 and Te0.Fractions (i) and (ii) were mixed and diluted to 0.05 mol l21 HCl to avoid speciation changes. The solution obtained was loaded on to an AG-1X8 (100–200 mesh, Cl2 form) anion exchange resin; the SeVI was retained whereas SeIV, TeIV and TeVI were eluted with 0.05 mol l21 HCl.The SeVI was then eluted with 0.3 mol l21 HCl. The eluate containing SeIV, TeIV and TeVI was passed through an Amberlite IR-120 (60–80 mesh, H+ form) column, where TeIV was retained and SeIV and TeVI were eluted with 0.05 mol l21 HCl. The adsorbed TeIV were eluted with 0.3 mol l21 HCl. The eluate containing TeVI and SeIV was evaporated nearly to dryness and the TeVI was then reduced to TeIV using HCl.The solution obtained and containing SeIV and TeIV was processed again on an Amberlite IR-120 column to separate the two ions. Fraction (iii) was evaporated nearly to dryness, treated with HCl and SeIV and TeIV were separated as described above. In each of the six fractions obtained selenium and tellurium were determined by ETAAS, after evaporation nearly to dryness. 3.2.3 Liquid chromatography with on-line detection HPLC and IC, which are widely used in biomedical studies for the determination and speciation of Se compounds,205 so far seem to be of only limited application in environmental analysis for Se and Te species. Some recent studies, however, appear to be interesting for their possible application also in the environmental field, thanks to interfacing with sensitive element-specific detection methods such as ICP-MS.Selenonium compounds, like trimethylselenonium compounds, selenoniocholine and acethylselenocholine, were separated on a cyanopropyl-bonded phase (5 mm silica support, LCCN) using a methanol mobile phase containing 20% v/v diethyl ether, 1% v/v acetic acid, 0.1% v/v triethylamine and 0.2 mg l21 trimethylsulfonium iodide.206,207 The species were detected online with AAS detection after thermochemical HG (see Section 4.2.1).Using an improved interface for AAS with thermochemical HG as described by Momplaisir et al.208 (see Section 4.2.1) capable of operating with either organic or aqueous HPLC mobile phases, Lei and Marshall92 achieved the separation of a number of selenium compounds.Using a cyanopropyl-bonded phase (5 mm silica support, LC-CN), selenate, selenite, selenomethionine, selenocystine and selenoethionine were separated with an aqueous acetic acid (0.015% v/v) mobile phase containing 0.1% m/v ammonium acetate. In the same study, N-2,4-dinitrophenyl derivatives of selenocystine, selenomethionine and phenylmercury cysteineselenoate were separated on a Nucleosil 5-NO2 column using a methanolic mobile phase containing acetic acid and triethylamine (0.05 and 0.8 mg ml21, respectively), while the selenium oxoanions (methaneseleninate, methaneselenonate and selenite) were separated by IC on a strong anion-exchange phase (PL-SAX, 8 mm) eluted with 0.1% m/v aqueous ammonium carbonate.Trimethylselenonium, selenite and selenate were separated on a Waters IC-PAK anion-exchange column using 0.08 mol l21 ammonium citrate as the mobile phase and detected online by ICP-AES.209 Several organoselenium and organosulfur compounds of biological importance such as selenols (RSeH = Se-cysteamine), thiols (GSH = reduced glutathione; PSH = penicillamine) and some of their derivatives, diselenides (RSeSeR), disulfides, selenenyl sulfides (RSeSG) and bis(alkylthio)selenides (PSSeSP, PSSeSG, GSSeSG) were separated by reversedphase HPLC on a Biophase ODS column (100 32 mm id, 3 mm particle size).210 In the elution of RSeH, GSH, GSSG, GSSeR and RSeSeR a mobile phase consisting of 0.05 mol l21 Na2H2PO4, 5% v/v CH3CN, 40 mg l21 sodium octylsulfate and sufficient H3PO4 to give a pH of 2.9 was employed.In order to achieve a satisfactory separation in a reasonable time of GSSG, GSSP, PSSP, GSSeSG, PSSeSG and PSSeSP, gradient elution using a mixture of two mobile phases was necessary.211 In all Analyst, December 1997, Vol. 122 131Rcases an on-line, specially designed, electrochemical detector was employed210 (See section 4.3.1). The selective derivatization of selenocysteine with the fluorescent reagent N-(iodoacetylaminoethyl)-5-naphthylamine- 1-sulfonic acid has been proposed for the determination of this selenoamino acid by HPLC with fluorimetric detection.212 The derivative is separated on an Ultrasphere ODS column (250 34.6 mm id, 5 mm particle size) and eluted at room temperature with an 8 + 2 v/v mixture of A (50 mmol l21 formic acid, 60 mmol l21 acetic acid, adjusted to pH 1.5 with sodium hydroxide) and B [solution A–propan-2-ol (97 + 3)].Determination of inorganic species is also performed by LC with on-line detection and also using derivatization techniques. The simultaneous determination of selenite and selenate in an aqueous extract of seleniferous soil was performed, without any derivatization, by using single column IC (also called nonsuppressed IC) with conductimetric detection.213 Selenite has also been separated by HPLC after derivatization to piazselenol,214,215 diethyldithiocarbamate complexes214,216 or benzylproponitriledithiocarbamate.217 On-line detection was used, with a UV detector for the carbamates and by fluorimetry for the piazselenol derivatives.Nakagawa et al.218 determined SeIV after its reaction with penicillamine (PSH) to form the selenotrisulfide (PSSeSP), which was subsequently derivatized to a fluorophore by reaction with 7-fluoro-4-nitrobenz-2,1,3-oxadiazole.The fluorophore was separated by reversed-phase HPLC on a C18 column at 40 °C and using acetonitrile–water–phosphoric acid (400 + 600: + 1,v/v) containing 10 mmol l21 of lithium sulfate (pH 2.6) as the eluent. 4. Detection 4.1 Non-specific Detectors 4.1.1 Gas chromatographic detectors This type of detector is typically operated on-line in most commercially available GC apparatus. The lack of specificity and, in some cases, sensitivity may represent a serious limitation for identification work. These disadvantage are often counterbalanced by their wide commercial availability, relatively low cost, well experienced analytical features and, in several cases, excellent analytical performance.Flame ionization detection (FID) was proposed and applied in early studies on the biological production of volatile alkyl selenides.219,220 FID was also employed for the determination of dimethyl selenide in breath from mice orally administered with selenium221 and for the quantification of methylselenocyanate for the indirect determination of selenomethionine192 (Sections 3.1.3 and 3.2.1).A detection limit of about 5 ng of Se was obtained in both cases. Similary, FID was employed for the detection of CH3TeCN in the indirect identification of telluroamino acids in tellurium yeast.36 Janak et al.198 used capillary GC–FID for the determination of selenomethionine, selenoethionine and selenocysteine after their derivatization with ethyl chloroformate.199 Today FID is rarely used, not only because of its limited sensitivity towards some alkylselenium and alkyltellurium compounds, but mainly owing to its lack of selectivity.Electron-capture detection (ECD) is an excellent method for the determination of volatile piazselenols formed by reaction of SeIV with substituted 1,2-diaminobenzene.The nature of the substituents on the aromatic ring has a considerable effect on the volatility, the extractability (typically into toluene) and on the chromatographic behavior, but also, and this is of particular relevance here, on the response factor of the ECD.181,195,222 One of the most widely used derivatizing agents is 3,5-dibromo- 1,2-diaminobenzene, but 3-bromo-5-trifluoromethyl-1,2-diaminobenzene has proved to be superior in terms of sensitivity and analysis time.195 In most cases the reported detection limits are a few picograms, but a sub-picogram detection limit has also been achieved.182 Flame photometric detection (FPD), widely employed in the selective detection of sulfur and phosphorus organic compounds, can also be employed for the selective detection of organoselenium and organotellurium compounds, but with one and two orders of magnitude lower sensitivity respectively, than for sulfur.223,224 The response of FPD to selenium and tellurium (in addition to sulfur) is exponential but can be forced to linearity by doping the make-up gas with sulfur.223 FPD is versatile element-selective method, possessing, at trace levels, a selectivity of Se versus C of 102–103. The broad bands of the emitting species S2 and Se2 centered at about 400 and 480 nm, respectively, are overlapped to a considerable extent, which limits the possibility of obtaining a good Se/S selectivity ratio.The selectivity of Se versus S can be enhanced, to a limited extent, by using optical filter discrimination and/ or by methane doping of the carrier gas according to the procedures proposed by Hanckock et al.,225 but in all cases the selectivity increase was small and, at the same time, the sensitivity was lowered.For these reasons, the simultaneous presence in the environment of several volatile sulfur compounds may represent a serious limitation to the application of FPD for the determination of volatile alkylselenium compounds.42,61 Fluorine-induced chemiluminescence detection (FCD) was employed by Chasteen et al.35 in the determination of volatile methylated compounds of selenium, tellurium and sulfur evolved from fungal or bacterial cultures.The working principle is similar to that proposed for the sulfur chemiluminescence detector, where the gas-phase reaction of molecular fluorine with reduced sulfur-containing compounds (RSH, RSR and RSSR) produces vibrationally excited HF and electronically excited CH2S.The radiational de-excitation of these species produces an emission in the visible (red) portion of the spectra.226 FCD is relatively non-responsive to H2S, H2Se and H2Te and its detection limits range from 6 pg of Te for dimethylditelluride up to 8 and 22 pg of Se for dimethyl diselenide and dimethyl selenide, respectively. The linear dynamic range covers at least three decades of concentration. The selectivity over normal hydrocarbons is !107.The sensitivity of FCD allowed the direct, simultaneous determination of selenium, tellurium and sulfur gas in the headspace of fungal and bacterial cultures without any preconcentration after a simple GC separation on a capillary column.35 An interesting application of photoionization detection (PID) has been reported for the simultaneous determination of selenium, arsenic, tin and antimony after HG with NaBH4 and separation of the generated hydrides with a Tenax GC column.117 The column plays an important role, possessing substantial inertness towards H2Se.At the same time, the column retains unwanted by-products (HCl, CO2, H2O, boranes) that would otherwise give a large interference signal. A detection limit of 25 pg for selenium was obtained.117 4.1.2 Fluorimetric detectors Fluorimetric detection (FD) is widely employed for the determination of selenium compounds after derivatization.The possibility of changing both the excitation and emission wavelengths and with the specificity of some of the derivatization procedures lead to determinations with good selectivity. Derivatization of SeIV with 2,3-diaminonaphthalene (DAN) and some of its halogenated derivatives, such as 2,3-diamino- 1,4-dichloronaphthalene (Cl2-DAN) and 2,3-diamino-1,4-dibromonaphthalene (Br2-DAN), to give the corresponding 132R Analyst, December 1997, Vol. 1224,5-benzopiazselenol has been used by several workers for the determination of SeIV at levels down to 0.01 ng ml21.81,227–229 One of the limiting factors using DAN derivatization is the adverse effect of atmospheric oxygen, which contributes both to the formation of fluorescing species that enhance the analytical blank and to the quenching of fluorescence of the piazselenol, thus decreasing the sensitivity.Purging of oxygen from the solvents and reagents results in a detection limit of < 0.05 ng ml21 of selenium in a semi-automated fluorimetric procedure. In addition, DAN derivatization is time consuming and requires pH adjustment to around 1–2.Recent investigations of piazselenol derivatization with Cl2- DAN and Br2-DAN led to substantial improvements in terms of reaction time and sensitivity; pH adjustment was not necessary for acid-digested samples.81,228,229 Further, Cl2-DAN and Br2- DAN allow the simultaneous pre-reduction of SeVI–piazselenol derivatization in HCl and HBr media, respectively, at temperatures of 85-100 °C in less than 15 min.81,228,229 Pre-column fluorescence derivatization of SeIV with DAN coupled with HPLC and FD resulted in a separation of the analyte from the background contribuiting species (detection limit 0.15 ng).215 Fluorophore different other than Se–DAN have been used.Nakagawa et al.218 proposed the derivatization of SeIV with penicillamine (RSH) to form a stable trisulfide followed by fluorophore formation by reaction with 7-fluoro- 4-nitrobenzene-2,1,3-oxadiazole (detection limit 0.015 ng of Se).Selective derivatization of selenocysteine with N-(iodoacetylaminoethyl)- 5-naphthylamine-1-sulfonic acid followed bt HPLC–FD resulted in a detection limit of 67 ng ml21 for selenocysteine.212 FD has been also applied to the determination of seleno methionine, selenocystine and TMSe using thin-layer chromatography (TLC) on a silica gel sintered plate.After the separation and digestion of the spots directly on the plate with nitric and perchloric acid, a derivatization step with DAN was applied, again directly on the plate. The fluorescence at wavelengths above 540 nm was measured by means of a TLC scanner after excitation at 388 nm.A detection limit of 0.4 ng of Se and linearity from 3 up to 250 ng of Se were obtained.230 4.2 Element-specific Detectors 4.2.1 Atomic absorption detectors 4.2.1.1 Total element determination. AAS is perhaps the most widely employed technique for the determination of total selenium and tellurium in environmental samples. The techniques employed are essentially three: electrothermal atomization in a graphite atomizer (ETAAS) with direct sample injection, HG with quartz tube atomization and HG with in situ trapping in a graphite furnace.Atomization in graphite furnaces with direct sample injection offers good sensitivities for both Se and Te with LODs in the range 10–30 pg.231 The major problem with the technique is represented by interferences due to the volatility of both of these elements, which hinders the use of high pyrolysis temperatures. 231,232 Moreover, some matrix components (e.g., chlorides, sulfates, phosphates, iron) may give rise to signal depression and background effects. Some background effects are not corrected by using a deuterium lamp and require Zeeman-effect correction.231,232 Chemical modification is thus an indispensable component of ETAAS determinations of Se and Te, and a vast literature exists on the subject.Pd– Mg(NO3)2 is one of the most useful mixed chemical modifiers allowing thermal stabilization of the analyte and better control of interefences from chlorides and sulfates. The interference of sulfate in the determination of selenium in water samples can be minimized by the use of a Pd–Mg(NO3)2–Ba(NO3)2 chemical modifier.233 The performances of mixed Pd–Mg modifiers were found superior to those of Cu–Mg and Ni–Mg mixed modifiers in the stabilization of inorganic and organic selenium species for the determination of total selenium in marine sample digests.234 Even though thermal stabilization with palladiumbased chemical modifiers was found to be effective for both selenium and tellurium compounds,233,235,236 many other metals have been tested, e.g., Pt, Ir, Cu and Ni, alone or mixed.231,232 The use of iridium deposited on carbide-coated platforms, pre-treated with either tungsten or zirconium, was investigated for a number of volatile analytes including Se and Te, and it was proposed as a permanent modifier for the determination of Se and Te and the other hydride forming elements by ETAAS.129,130 HG with on-line detection by AAS is realized by using a flame or electrothermally heated quartz tube atomizer.Sensitivity and interference figures are mostly related to the HG methods. In general, both absolute sensitivity and interference control are improved in the sequence batch > continuous > flow injection.161 Detection limits as low as 20 pg of Se237 and 30 pg of Te238 have been reported using flow injection– HGAAS.Quartz tube atomizers are prone to atomization interferences, mainly due to the deterioration of the inner quartz surfaces and mutual interference effects generated by the elements that form volatile hydrides.156 These interferences during the atomization step are related to the atomization mechanism of the hydrides and to the fate of free atoms.150,121,239 They are lowered by several orders of magnitude by using a hydrogen flame-in-tube atomizer,150 a miniature argon–hydrogen diffusion flame78,240 or graphite atomizer.241 In situ trapping of hydrides in a graphite atomizer represents an elegant and reliable technique for Se and Te determination, allowing preconcentration and better control of interferences both in the liquid phase and in the atomization step.(see also Section 2.3.2). When coupled with flow injection technique, the methods allow detection limits at the ng l21 level, automation and better control of liquid-phase interferences.120 4.2.1.2 Chromatographic detection.Sensitive element-specific detectors for chromatography have also been made employing quartz or graphite furnaces as atomizers, properly interfaced with the chromatographic column. The GC determination of volatile dimethyl selenide and dimethyl diselenide using AAS as a selenium-specific detector was firstly reported by Chau et al.54 and was used in their studies on the methylation of selenium in lake water sediments. 242 The interface between the silica furnace, electrothermally heated at 1000 °C, and the GC column, was simply a piece of stainless-steel tubing (2 mm od). The GC effluents were pre-combusted by means of a small flame generated by admitting a small flow of hydrogen (120 ml min21) and air (120 ml min21) in the side-arm of the silica furnace.Pre-combustion prevented interferences due to absorption at the 196 nm selenium resonance line given caused by chlorinated solvents and hydrocarbon compounds. This precaution was indispensable since the deuterium background correction system failed to correct effectively for this kind of interference. With the pre-combustion zone in operation no interference was observed in presence of 5 ml of acetone, alcohol, toluene, diethyl ether, benzene, chloroform, methylene chloride and carbon tetrachloride.Another important experimental parameter was the presence of an auxiliary hydrogen flow inside the furnace (120 ml min21), which resulted in an improved sensitivity. The detection limit was 0.1 ng of Se, with a linear range up to 50 ng and an RSD of 8% for 10 ng of Se.242 Cutter59 also used a silica furnace atomization system for the determination of volatile alkyl selenides in water samples.It was very similar to the system described by Chau et al.54 with appropriate modifications for the different sampling technique, and, in fact, exhibited very similar analytical performances. Analyst, December 1997, Vol. 122 133RAlthough quartz furnace atomizers continue to be used in recent studies on selenium speciation,60,114 their analytical performance, for the GC–AAS determination of alkyl-selenides, does not seem that have been improved since the first report by Chau et al.54 Radziuk and Van Loon243 built an integral quartz device incorporating a silica quartz atomizer and a GC column.The custom-built system had the advantage of avoiding the use of an external transfer line.Graphite furnaces have also been employed in as atomizers GC–AAS. Jiang et al.44 interfaced the GC column with a furnace with a 1 m length of nickel tubing maintained at 20 °C. The atomizer was operated at 1800 °C during the elution of the alkyl selenides, which required 4 min. Also in this case the addition of hydrogen (10%) to the argon carrier gas improved both the sensitivity and the selectivity of determinations.However, its analytical performance was almost the same as that obtained with silica furnace atomizers. An improvement was achieved by using a palladium-coated graphite furnace.54 The GC column was interfaced to the furnace by means of length of Teflon tubing (200 3 1 mm id) tipped with a short piece of quartz tube.The quartz tip was inserted into the pyrolytic graphite furnace through the sample introduction port and kept in contact with the opposite interior wall of the furnace. The compounds, dimethyl selenide, diethyl selenide and dimethyl diselenide, were determined sequentially during the GC separation by adsorption on the palladium-coated graphite furnace.In fact, the time interval between the elution of each peak was long enough to allow the operation of the atomization cycle of the furnace. Detection limits of 5 and 20 pg of Se were obtained for dimethyl selenide and diethyl selenide, and for dimethyl diselenide, respectively, with an RSD of 1.7–3.7% at nanogram levels. The incompatibility of electrothermal atomizers with a continuous flow of liquid sample represents a severe limitation in obtaining a simple and sensitive on-line metal-specific AAS detector for LC.An off-line Se-specific detector was realized by Chakraborti et al.244 by interfacing a ion chromatograph with a Zeeman-effect graphite furnace AA spectrometer. The IC apparatus also had a conductivity detector whose response was simultaneously recorded with that obtained from the AAS detector. Once every 80 s, a slider injection valve took away a small volume of eluate already passed through the conductivity detector, to inject it into the graphite atomizer.The resulting AAS trace had a discontinuous shape and was time-shifted with respect to the real-time trace obtained with a conductivity detector. For selenite and selenate, the detection limit was around 5 ng of Se.The determination of these ions could be achieved also in the presence of several parts per million of chloride, sulfate and phosphate, which was impossible with a conductivity detector alone. In a more recent paper, the fraction collection of HPLC eluates followed by hot injection into a graphite furnace was optimized for selenite, selenate and trimethylselenonium with detection limits in the range 0.76–1.67 ng of Se for the considered species.245 Post-column derivatization was adopted by Blais et al.206 for the on-line detection and determination of trimethylselenonium compounds and selenoniocholine by HPLC–AAS using quartz furnace atomization.A specially designed interface was built for the derivatization, in which the eluate was thermosprayed and pyrolyzed in presence of oxygen and the resulting selenium species (e.g., SeO2) were thermochemically derivatized to the hydride in a hydrogen flame.The hydride formed was atomized in the quartz furnace in presence of an oxygen–hydrogen flame. The limits of detection were 5 and 7 ng of Se for selenoniocholine and trimethylselenonium, respectively.The principal limitation of this interface was the requirement for a predominantly methanolic mobile phase to support the combustion process, which poses severe limitations in the choice of conditions for the chromatographic stage. In addition, the analytical response of the selenate was only 20% of the response of selenite and trimetilselenonium ion.An improved interface for thermochemical HG was developed by Momplaisir et al.208 This improved interface was capable of operating with either an organic or aqueous mobile phase and gave the same analytical response for seleno oxoanions and selenoamino acids. Further, the detection limits were improved and were in the the range 1–2 ng for the investigated selenium compounds. 4.2.2 Atomic emission detectors Inductively coupled plasma atomic emission spectrometry (ICP-AES) is one of the most widely employed techniques for multi-element determinations.Its sensitivity, depending on the elements, can reach the parts per billion level and the calibration graphs are linear over a range spanning three to five decades of concentration. Combination with HG techniques for total element determination results in improved detection limits with respect to sample nebulization.Nevertheless, HG–ICP-AES is significantly less sensitive than other techniques combining HG with AAS, AFS and ICP-MS detection. Detection limits in the range 0.055–1.2 ng ml21 for Se and 1 ng ml21 for Te have been reported for ICP-AES combined with various HG techniques (batch, continuous, flow injection).121 Its use as an elementspecific detector is mainly devoted to LC owing to the compatibility of the nebulization system with the continuous flow of the eluate from the LC column.Selenite and selenate were determined with detection limits of 140 and 91 ng, respectively, by simply connecting an HPLC column to the nebulizer of the ICP system with Teflon tubing.246 The nebulizer design was important in determining the sensitivity and the compatibility of ICP towards HPLC solvents and in maintaining the chromatographic peak shape.209,247 Laborda et al.209 made a comparison between cross-flow and thermochemical nebulization.It was found that the latter gave a threefold improvement in sensitivity and higher tolerance towards the use of organic solvents.The fraction of methanol in the methanol–water eluent introduced into the plasma could be as high as 75%, compared with the 25% tolerated by the crossflow nebulizer. The system was tested on the separation of trimethylselenonium, selenite and selenate, obtaining detection limits ranging from 14 to 54 ng of Se. Post-column hydride derivatization has been also tested for the determination of SeIV and SeVI by means of IC–ICP-AES, and it gave detection limits of 1.6 and 2.5 ng ml21 of Se respectively, in the original sample.248 Colon and Barry249 evaluated the performance of post-column hydride derivatization for the determination of SeIV by using ac arc plasma-AES.A detection limit of 60 pg s21 of selenium was obtained, corresponding to 20 ng ml21 in the original sample.The microwave-induced plasma (MIP) is one of the major plasma sources for multi-element-specific detection in GC determinations. It possesses sensitivities in the low picogram range, a wide dynamic range and the possibility of the determination of elemental ratios.250 Talmi and Andren180 reported a detection limit of 40 pg of Se for the determination of piazselenols by using GC–MIP-AES, which was not very different from the 62 pg of Se obtained by Estes et al.251 for the determination of diethylselenium using a capillary column.In the latter case the Se versus C selectivity was 1.09 3 104.251 Reamer and Zoller252 applied GC–MIP-AES to studies of selenium biomethylation in soil and sewage sludge, obtaining a detection limit of 20 pg of dimethyl selenide.More recently, a commercial GC–MIP-AES unit has been marketed. Based on a photodiode-array detection system, the 134R Analyst, December 1997, Vol. 122apparatus allows the automated, simultaneous determination of many elements during a single GC run. Such an instrument was employed by Tanzer and Heumann42,61 for the simultaneous determination of volatile sulfur and selenium compounds in Atlantic Ocean sea-water.The detection limit for dimethyl selenide was 50 pg l21 Se as using a 50 ml sample volume (corresponding to an absolute amount of 2.5 pg of Se). A good example of the potential of elementspecific detectors and of the limitations of non-specific, even if selective, detector systems is the following. Tanzer and Heumann,42,61 in their investigation, also employed nonspecific GC detectors, e.g., the FPD, in the determination of sulfur compounds.A peak due to an unidentified sulfur compound was present in the FPD trace. The same determination performed with MIP-AES allowed the simultaneous acquisition of two element-specific traces: selenium at 196 nm and sulfur at 181 nm. MIP-AES detection revealed that the FPD peak is due to the contribution of two species, dimethyl selenide and an unidentified sulfur compound, unresolved by the GC column. 4.2.3 Atomic fluorescence detectors This kind of detector possesses good sensitivity, a wide dynamic range and multi-element capabilities. At the moment, the major drawback of atomic fluorescence spectrometry (AFS) appears to be its relatively limited commercial exploitation, if a comparison is made with other techniques such as AAS and AES.Simple and sensitive non- dispersive (ND) apparatus can easily be assembled using commercially available components. 253 AFS is well suited for coupling with HG using small hydrogen–argon diffusion flames as atomizers. Detection limits of 100 pg254 and 27 pg240 for Se and 80 pg255 for Te with calibration graphs linear over 3–4 decades of concentration can be obtained by use of simple in-house assembled apparatus, based on electrodeless discharge lamps as the excitation light source and a batch hydride generator.Detection limits of 18 pg for Se256 and 17 pg for Te148 were obtained by batch HG from non-aqueous media. Guo et al.257. reported detection limits of 35 and 20 pg for Se and Te, respectively, using flow injection– HG–ND-AFS.257 Commercially available HG–ND-AFS detection systems consist of a boosted-output hollow-cathode lamp as the light source and a miniature argon–hydrogen flame as an atomizer. 258 The isolation of the main AF lines for selenium (196.09, 203.99 and 206.28 nm) is achieved by means of an interference filter.Using continuous-flow HG, a detection limit of 0.05 ng ml21 of Se is obtained with a calibration graph linear up to 100 ng ml21. Similar performances have been also obtained for tellurium. The same ND-AFS detection system was used as an element-specific detector for both GC and HPLC speciation of selenium species. The determination of methyl selenides onboard during a cruise was achieved by interfacing a packed GC column with the ND-AFS detector through a PTFE line; detection limits of around 5 pg or 0.5 pg l21 of Se were achieved.62,259 For speciation of SeIV and SeVI, the inorganic species were separated on an anion-exchange column followed by on-line microwave reduction of SeVI to SeIV in hydrochloric acid.The subsequent on-line reduction of SeIV to the volatile hydride and detection by ND-AFS allowed limits of detection of 0.2 and 0.3 ng ml21 for SeIV and SeVI, respectively.260 The use of continuous HG coupled with ND-AFS has been investigated for several hydride forming elements, including Se and Te.105 One of the major problems was recognized as light scattering caused by small droplets generated during the sodium tetrahydroborate(iii) decomposition.Lowering the NaBH4 concentration from 1 to 0.1% suppressed the scattering signals with a consequent improvement in both signal-to-noise ratio and accuracy. For selenium, the addition of 0.1 mol l21 KI, to the borohydride solution was effective in mantaining the same generation efficency, according to the catalytic role played by halide ions in the reduction of selenite ion.71 For tellurium no such effect was found to operate but the borohydride concentration can be lowered at the expense of a smaller linear dynamic range.Using a sample flow rate of 7 ml min21, detection limits of 2–5 pg ml21 can be obtained with calibration curves linear over three or four decades of concentration.105 ND-AFS has also been employed for element-specific GC detection in the simultaneous determination of alkyl compounds of Se, Pb and Sn.194 By directly coupling a wide-bore capillary column to a simple borosilicate glass burner sustaining a small hydrogen–argon diffusion flame, a detection limit of 10 pg of Se was obtained for both dimethyl selenide and dimethyl diselenide.The selectivity was 107 for CS2, 9 3 107 for benzene and better than 108 for CCl4, 1,1,1-trichloroethane, ethanol, methyl acetate and butan-2-one.The most serious interference problem by the overlap of the solvent peak (0.5–2 ml) with the analyte peak. In this case the fluorescence signal of the analyte peak could be greatly reduced, probably owing to quenching effects, and the replacement of the solvent with another one, possessing a different retention time, was advised.194 ICP-AFS coupled with HG and cold trapping has been widely investigated by Rigin261 for several elements, including selenium, for which a detection limit of 3 pg was obtained.An argon shielded, highly fuel rich, hydrogen–oxygen diffusion micro flame has recently neen developed for hydride atomization.262 For selenium the signal-to-noise ratio was improved of about one order of magnitude with respect to a miniature argon–hydrogen diffusion flame.Impressive analytical performances can be obtained by using laser-excited atomic fluorescence spectrometry (LE-AFS). Liang et al.263 using of a graphite furnace atomizer combined with LEAFS, reported for tellurium a detection limit of 0.02 pg with calibration curves linear over seven decades of concentration.Comparable results were obtained for selenium by Heitmann et al.264 with a detection limit of 0.015 pg and calibration curves linear over six decades of concentration. The disadvantages of LEAFS instruments are mainly their very high cost and the necessity for expert operators, but their use could lead to significant contributions to the knowledge of the environmental behavior of rare elements as tellurium. 4.2.4 Mass spectrometric detectors. The use of MS in elemental analysis has grown in the last decade thanks to the commercial exploitation of inductively coupled plasma mass spectrometry (ICP-MS) as a simultaneous multi-element-specific detection method. ICP-MS possesses excellent selectivity and sensitivity and to this it must be added the possibility of performing also isotopic determinations.However, this technique suffers from many interferences, which are discussed at the end of the section. In addition, ICPMS instrumentation is expensive and not readily accessible for most laboratories. In relation to total element determination, ICP-MS has been coupled with several sample introduction techniques.A detection limit of 50 pg for Te has been reported with ultrasonic nebulization and aerosol desolvation,265 and 0.3 ng ml21 for Se266 was obtained after removal of interferences. HG gave the most impressive results allowing detection limits of 0.5 pg for Te267 and 1–6.4 pg for Se.267–269 Detection limits at the picogram level were also reported for selenium by using electrothermal atomization ICP-MS.270 Gallus and Heumann, 271 in the development of an isotope dilution method based on piazselenol derivatization of selenite and GC–ICP-MS detection, reported a 0.02 ng ml21 detection limit by injection of 1–6 ml of sample extract.Analyst, December 1997, Vol. 122 135RCoupling various LC techniques with ICP-MS for the speciation of inorganic and organic selenium compounds gave detection limits of 100 pg for both selenite and selenate,265 @100–200 pg for selenomethionine, selenocystine and trimethylselenonium272 –274 and down to 30 pg using a low-flow interface with a capillary nebulizer.275 It must be emphasized that at present ICP-MS appears to be the only technique able to be directly interfaced with LC, thus resulting in a simultaneous multi-element-specific on-line detector with sensitivity at the sub-ng ml21 level.A comparison among different atomic element-specific detection systems for LC was reported by K�olbl.188 He concluded that ICP-MS offers unmatched sensitivity (0.1 ng absolute detection limit) and wide linear dynamic ranges compared with graphite furnace AAS, which, can be used for off-line detection in LC.Interesting applications of LC interfaced with ICP-MS have been reported. 100,276,277 In a waste water treatment plant the problem existed of removing the tellurium present in the waste water in the form of organic and inorganic compounds at 0.5–1.5 mg l21 levels and in the presence of about 5000 mg l21 of organic compounds. The waste water treatment consisted of an ‘aging step’ to degrade the organotellurium to inorganic tellurite and tellurate, followed by a ‘retaining column step’ to remove the inorganic tellurium species.They used IC–ICP-MS and HPLC– ICP-MS, also sometimes coupled with on-line isotope dilution with an enriched 125Te isotope, to verify the effectiveness of the treatment. The degradation of organotellurium into inorganic tellurium species, the only species retained by the ‘columns’, was successively improved by using NaOH in the presence of EDTA.276,277 Selenium incorporation into cyanobacterial metallothionein induced by zinc was investigated by Takatera et al.100 They coupled ICP-MS with different HPLC techniques, namely sizeexclusion HPLC, reversed-phase HPLC and cation-exchange HPLC, in order to achieve the separation of metallothioneins, the separation of metallothionein isoforms and the speciation of selenium species, respectively. The simultaneous acquisition of Zn-specific and Se-specific chromatograms allowed them to verify the correlation between the two elements in metallothionein and metallothionein isoforms.The speciation of selenium was performed on the carboxymethylated derivatives.Also in this case, acid hydrolysis of metallothionein was applied only after carboxymethylation with iodoacetic acid in order to protect selenocysteine (see Section 2.1.2). Their results showed that about 70% of selenium was incorporated into the cyanobacterial metallothionein as selenocysteine, 4% as selenomethionine and 19% as an unidentified selenium compound. Negative thermal ionization mass spectrometry (NTI-MS) was employed by Tanzer and Heumann183 for the determination of selenium by use of the isotope dilution technique.Sample spiking was performed using 82Se-enriched selenium compounds and the selenium was determined by measuring the 80Se/82Se ratio. Using a 0.25 l water sample, a detection limit of 12 pg g21 was obtained.The restricting factor for the detection limit was the precision in the determination of isotope ratio. However, ICP-MS is not free from interference problems. Evans and Giglio, in a review on interferences in ICP-MS, classified the interferences into two categories, spectroscopic and non-spectroscopic.278 Spectroscopic interferences cause an analytical signal enhancement which is originated by isotopes of different elements or from molecular ions formed in the plasma torch from plasma gases themselves, entrained atmospheric gases, and species arising from sample matrix, including reagents employed for sample preparation and dissolution.Among molecular ion interferences the following have been reported for selenium: 40Ar40Ar+ on 80Se (the most abundant selenium isotope, 49.82% isotopic abundance),279,280 40Ar36Ar+ on 76Se,281 40Ar38Ar+ on 78Se,281 40Ar37Cl+ on 77Se,281–285 37Cl37Cl+ on 74Se284 and 81Br1H on 82Se.286 Tellurium determination by ICP-MS appears to be fe from these problems and the most abundant natural isotopes are used in most cases.267,276,277 Non-spectroscopic interferences may cause either an enhancement or a depression of the analytical signal and originate from the sample matrix.They affect the signal magnitude through the influence on sample transport, ionization in the plasma, ion extraction or ion throughput in the resultant ion beam.278 Most spectroscopic and non-spectroscopic interferences are removed by sample pre-treatment involving extraction, precipitation and coprecipitation of the analyte from the sample matrix interferences.HG for Se and Te allows to the analyte to be separated from the sample matrix, increases the sensitivity and greatly reduces the spectroscopic interferences of 40Ar37Cl+ on 77Se and 37Cl37Cl+ on 74Se.282,287,288 Similar beneficial effects are to be expected on coupling GC with ICP-MS to those demonstrated by LC techniques allowing the separation of the analyte from the sample matrix265,276,277 and on-line removal of interferences by anion-exchange column separation has recently been proposed.266 Spectroscopic interferences due to polyatomic species of the type ArCl+, ClCl+ and ArAr+ can be attenuated by up to three orders of magnitude by the addition of less than 5% N2 to the central channel or spiking the sample with 10% propan- 2-ol.289,290 The interference of 40Ar37Cl+, 40Ar38Ar+ and 40Ar2H2 + with 77Se, 78Se and 82Se determinations, can be alleviated by addition of methanol to the sample (510%).269 Methanol additions also induced a sensitivity enhancement by a factor of 4–5 for selenium and tellurium, using either pneumatic nebulization or HG.269 The use of an alternative plasma source such as He MIP and N2 MIP has been widely investigated278 and should results in an interference-free determination of 77Se, 78Se and 80Se. 4.3. Species-specific Detectors 4.3.1 Electrochemical detectors These detectors offer excellent sensitivities for selenium coupled with simple and inexpensive instrumentation. Electrochemical techniques employed for the determination of total selenium are based on the reduction of the electroactive inorganic SeIV,205 and for this reason all selenium must be converted into selenite form before determination, or it can be used in the selective detection of inorganic SeIV.Differentialpulse voltammetric techniques have been applied to the determination of selenium in contaminated waters (detection limit 2 ng ml21)291 and fish samples.83 In the latter method a sensitive catalytic polarographic wave was obtained with the complex Se(O)SO3 22, in KIO3–NH3–NH4Cl medium, which is formed from reaction of selenite and sulfite.The detection limit was 60 pg ml21 and the calibration curves were linear up to 8 ng ml21.83 A 1.5th-differential polarographic determination of selenium in natural waters was reported by Wu et al.292 with a linear dynamic range from 0.01 to 1 ng ml21.Stripping voltammetric technique offers detection limits below 0.5 ng ml21 at a dropping mercury electrode293 and 0.1 ng ml21 at a gold-film electrode.294 The addition of CuII or RhIII (electrodeposition and stripping of Cu2Se or Rh2Se3 at the dropping mercury electrode) resulted in detection limits of better than 2295,296 and 0.5 pg ml21,297 respectively.Differential- pulse cathodic stripping voltammetry at a dropping electrode offers detection limits around or below the 100 pg ml21 level,298–300 but in the presence of added CuII the detection limits can be lowered to a few pg ml21 of Se.301 The use of piazselenol derivatization has also been reported in electrochemistry in combination with differential-pulse polarography, 302 differential pulse voltammetry,303 cathodic and adsorptive stripping voltammetry,304 and differential-pulse 136R Analyst, December 1997, Vol. 122cathodic stripping voltammetry,305 with detection limits in the range 10–400 pg ml21. The electrochemical behavior of tellurium is similar to that of selenium and only TeIV is electroactive.The addition of CuII substantially enhances the sensitivity in stripping techniques but also in this case the conditions must be strictly controlled in terms of pH, Cu concentration and deposition time.306,307 The simultaneous determination of selenium and tellurium at sub-parts per billion level has been reported.306–308 Detection limits for tellurium varied from 20 ng ml21 for differential-pulse polarography309 down to 0.1 ng ml21 for differential-pulse cathodic stripping voltammetry.307 The use of a preconcentration technique (sulphydryl cotton fiber) allowed the determination of @1 ng l21 of tellurium in uncontaminated natural water samples.310 The application of polarography to tellurium in process water has been evaluted and compared with other analytical techniques in the chemical industry that use tellurium catalysts.309 The simultaneous determination of selenium and tellurium in biotic materials and coal by differential-pulse cathodic stripping voltammetry has been reported.307 Interesting applications of electrochemical detectors coupled on-line with LC can be found in the literature.Killa and Rabenstein210,211 coupled electrochemical detection with HPLC (see also Section 3.2.3) for the determination of selenols, thiols, diselenides, disulfides, selenenyl sulfides and bis(alkylthio) selenides.The detector cell consisted of two 3.2 mm diameter amalgamated gold electrodes. The dual electrodes were positioned in series in the flow channel and an Ag/AgCl reference electrode was used. The upstream Au/Hg electrode, kept at a suitable potential, provided the reduction of disulfides, diselenides and selenenyl sulfides to thiols, selenols and thiols and selenols, respectively, according to a reaction of the type RSSR + 2H+ + 2e2 ? 2 RSH (1) The reduction of bis(alkylthio)selenides was thought to follow the reaction scheme: RSSeSR + 2H+ + 2e2 ? 2 RSH + Se (2) RSSeSR + 4H+ + 4e2 ? 2 RSH + H2Se (3) The downstream electrode was typically set at +0.15 V versus Ag/AgCl and performed the detection of thiols and selenols originally present in the sample and the upstream electrode those produced by the reduction of RSSR, RSeSeR and RSSeR.Among the reduction products of reactions (2) and (3), presumably only the species RSH and H2Se were electroactive at the downstream electrode.By appropriate choice of the potential for the upstream electrode, the diselenides and the selenenyl sulfides could be selectively detected in the presence of disulfides.210 Also, the bis(alkylthio)selenides could be selectively detected in the presence of disulfides.211 For bis(penicillamine)selenides and using a 20 ml sample volume, the detection limit was < 1027 mol l21.Also, the range of linearity of the detector was very attractive, extending from 1.1 3 1027 up to 1.2 3 1024 mol l21.211 The application of commercially available integrated pulsed amperometric detector (IPAD)311,312 for high-performance anion-exchange chromatography has been reported for the determination of selenocysteine and selenomethionine in dolphin liver protein hydrolyzates.30 The detection is based on the oxidation at +0.35 V of the amino group in an alkanine medium (pH > 11) at an Au working electrode (Ag/AgCl reference electrode).The removal of reaction products and the cleaning of electrode surface were performed at +0.9 and 20.9 V, respectively. Absolute detection limits are in the range 20–30 pg for selenomethionine and selenocysteine with a dynamic range covering up to two decades of concentration.30 Interferences are a serious problem when using electrochemical techniques. They arise from concomitant elements, organic substances and surface-active substances, and are the main limitation to electrochemical techniques. Hence a sample pre-treatment step is often required before determination.Dissolved organic substances present in natural waters seriously interfere with the electrodeposition of selenium in several stripping techniques.In many cases the destruction of organic substances is required. Mattsson et al.296 found that for concentrations of dissolved organic carbon around the mg l21 level the determination of selenium is not possible by cathodic stripping voltammetry or by differential-pulse cathodic stripping voltammetry.UV photolysis was employed for the destruction of organic substances but produced the simultaneous reduction of selenium in the tetravalent form, and for these reasons it is only useful for total dissolved selenium determination. 291,295,296,313 Interferences from foreign ions and other matrix components are best removed by using ion-exchange columns (e.g., Amberlite IRA-400,293 Chelex-100,291 Chelex- 100 in the FeIII form298P), which do not produce alterations in selenium speciation and also allow useful preconcentration factors to be achieved.The use of piazselenol derivatization is not free from interference problems. Interferences from CrVI, CuII, MoVI, NiII, SnII, TeIV and VV produce signals close to the reduction signal of piazselenol, and they are removed by sample pre-treatment with a Chelex-100 resin column, with a 99.5% selenium recovery.302 Surface-active organic substances present in natural waters interfere strongly by decreasing the amount of deposited selenium and also oxidation products of 3,3A-diaminobenzidine employed for piazselenol formation hinder the accurate measurement of the selenium stripping peak.305 The extraction of piazselenol into toluene followed by its back-extraction into acidic solution renders the analysis free from both interferences.305 4.3.2 Spectrometric detectors The identification of organic Se and Te compounds using chromatographic techniques coupled with element-specific detectors can be effected solely by comparison with chromatograms of authentic standards of the analytes that had to be determined.In the case of chromatographic peaks originating from unknown compounds, the structural recognition may then be made only on a trial-and-error base. The only other alternative is the use of detectors able to give also structural information. Among the spectrometric techniques providing this kind of information and possessing adequate sensitivity for environmental applications, only MS has been applied to the identification of organoselenium and organotellurium compounds.For instance, organoselenium and organotellurium compounds evolved during incubation experiments were identified by GC coupled with element-specific detectors, but in many cases their structures could be confirmed only by the use of GC–MS.35,252 Chau et al.,54,242 using GC–AAS, detected an unidentified volatile organoselenium compounds during incubation experiments on lake sediments spiked with selenium compounds.Later, Reamer and Zoller252 identified the compound as (CH3)2SeO2 by using GC–MS, but the mass spectrum did not allowed them to discriminate between the two possible molecular structures, dimethylselenone or methyl methylselenite.Chasteen,314 after experiments performed by GC–FCD (see Section 4.1.1), CG–MS, and also considering the boiling point data, concluded that the above selenium compound is not dimethylselenone or methyl methylselenite, but dimethyl selenenyl sulfide, CH3SSeCH3. The identification of selenoamino acids, in particular selenocysteine, required derivatization reactions in order to produce more stable and less polar compounds with respect to the Analyst, December 1997, Vol. 122 137RTable 9 Summary of analytical techniques employed for Se and Te determinations in environmental samples Technique Comments Section Atomic absorption spectrometry— ETAAS Direct injection of liquid sample into graphite furnace for total element determination.Required the use of appropriate background correction systems (e.g., Zeeman). Use of chemical modifiers is mandatory to reduce volatilization losses of Se and Te and for a better control of matrix interferences (e.g., sulfates, chlorides, phosphates, iron). LOD 0.5–2 ng ml21 or 10–30 pg 4.2.1 HGAAS (direct transfer) Hydride generation with on-line atomization, mostly with quartz tube atomizer.For total Se and Te determination. LODs (0.05–0.1 ng ml21) and magnitude of liquid-phase interferences depend on the HG generation technique (batch, continuous, flow injection). The other hydride-forming elements may generate serious interferences in the quartz tube atomizer. Modification of quartz tube surfaces deteriorates analytical performance 4.2.1; 2.4.1 HGAAS (cold trapping) Trapping volatile compounds after HG step followed by thermal stripping with on-line atomization in a quartz tube.Not usable for Te owing to its thermal instability. Possibility of speciation of simple methylated organoselenium and inorganic selenium converted into the corresponding volatile derivatives (H2Se, MeSeH, Me2Se) by GC trap.LOD @ng l21 level. Interferences are the major problems as in the case of direct transfer HGAAS. Additional problem due to the accumulation of volatile interfering species into the cold trap 2.3.1; 3.2.1 FI–HG–ETAAS Hydride generated by flow injection technique followed by in situ trapping in a graphite furnace or in a carbon rod atomizer. For total element determination. LODs in the range of 2–40 pg or in the low ng l21 range.Remarkable reduction of liquid-phase interferences and virtual elimination of atomization interferences. Fully automated 2.3.2; 2.3.3; 4.2.1 GC–ETAAS GC coupled with on-line quartz tube atomization. Volatile alkyl selenides or simple methylated selenium species (e.g., MeSeH, Me2Se, Me2Se2, Me2SeO2) are collected and preconcentrated in a GC column followed by thermal stripping and on-line atomization in a quartz tube.Interferences due to organic molecules reduced by hydrogen addition to the atomizer. LOD 0.1 ng for Se. Applicable to both air and water samples 1.1.2; 4.2.1 HPLC–ETAAS (graphite atomizer) Off-line detector for speciation of SeIV, SeVI and Me3Se+. Best reported LOD 0.76–1.76 ng 4.2.1 HPLC–ETAAS (thermochemical HG) On-line detection by thermochemical HG realized with a specially designed quartz furnace.LODs 1–2 ng of Se. Used for inorganic species, Se-amino acids, selenoniocholine, Me3Se+ 3.2.3; 4.2.1 Atomic emission spectrometry— HG–ICP-AES ICP-AES with various HG techniques. LOD 0.055–1.2 ng ml21 for Se and around 1 ng ml21 for Te 4.2.2 HPLC–ICP-AES Tested mainly for SeIV, SeVI and Me3Se+ detemination.LOD 14–141 ng depending on design of HPLC–ICP interface 4.2.2 IC–HG–AES Ion chromatography with on-line post-column HG. LOD 1.6–2.5 ng ml21 of SeIV and SeVI in the original sample by ICP-AES detected and 20 ng ml21 for SeIV by alternating current arc plasma AES 4.2.2 GC–MIP–AES GC coupled on-line with microwave-induced plasma AES.Simultaneous acquisition of elementspecific chromatograms for difference elements. LOD 2.5–62 pg of Se depending on Se compound (alkylselenides, piazselenols) 4.2.2 CV–GC–MIP–AES Derivatization of SeIV with sodium tetraethylborate(iii) to diethyl selenide followed by cold trapping and capillary GC with on-line MIP–AES detection. LOD 8 pg ml21 for 5 ml sample volume 3.2.1 Atomic fluorescence spectrometry— HG–ND-AFS Hydride generation coupled with non-dispersive apparatus.Miniature flame mostly used as the atomizer. For total element determination of Se and Te. Excellent control of atomization interferences. LOD in the range 17–100 pg and a few ng l21, according to the HG technique employed (batch, continuous, flow injection). Dynamic ranges span over 3–5 decades 4.2.3 HG–ICP-AFS Hydride generation–cold trapping followed by ICP atomization and AFS detection.LOD 3 pg of Se 4.2.3 GC–ND-AFS Employed for detection and determination of methyl selenides (MeSeH, Me2Se, Me2Se2) in air and water samples. LOD 5–10 pg of Se and 0.5 pg l21 3.2.1; 4.2.3 HPLC–ND-AFS Separation and determination of selenite and selenate. On-line reduction followed by on-line HG and ND-AFS detection.LOD 0.2–0.3 ng ml21 4.2.3 LEAFS Graphite furnaced atomization combined with laser-excited AFS. LODs 0.015 pg for Se and 0.02 pg for Te. Dynamic ranges span over six–seven decades 4.3.2 Inductively coupled plasma mass spectrometry— ICP-MS Solution nebulization leads to sub-ng ml21 detection limits for both Se and Te. Matrix interferences for Se and Te.Isobaric interferences for selenium. Interference control and sensitivity enhancement obtained by sample spiking with propan-2-ol or methanol 4.2.4 HG–ICP-MS Various HG technques coupled on-line with ICP-MS for total Se or Te determination. Isobaric interferences reduce sensitivity for Se. LOD 1–6.4 pg or Se and 0.5 pg of Te 4.2.4 ETV–ICP-MS Electrothermal vaporization coupled on-line with ICP-MS.Simultaneous multi-element determination. LOD Å 1 pg for Se 4.2.4 GC–ICP-MS GC of piazselenol derivative of SeIV. LOD at the femtogram level for Se: 0.02 ng ml21 with 1–6 ml sample injection 4.2.4 HG–GC–ICP-MS Simultaneous qualitative and semiquantitative determination of several hydride-forming elements and their volatile compounds (Me2Se, M2Se2, Me2Te). Not for H2Te owing to cold trapping step 2.3.1 LC–ICP-MS Various LC techniques coupled on-line with ICP-MS.Application to determination of SeIV, SeVI, selenoamino-acids, Me3Se+, organotellurium in waste water, Se–Zn metallothioneins. LOD 30–200 pg for Se 4.2.4 138R Analyst, December 1997, Vol. 122original selenoamino acid. The idendification of selenocysteine in glutathione peroxidase enzyme was performed by direct probe MS after a series of derivatization reactions, the first of which was performed in order to protect the selenol group before protein hydrolysis (see also Table 6).95,101 Then the aminoacid is converted to the less polar, more volatile N-acetyl and/or O-methyl derivatives before MS identification (see also Section 2.1.2.).The derivatisation with 1-fluoro-2,4-dinitrobenzene was proposed by Ghanter and Kraus315,316 for the MS identification of H2Se and CH3SeH (see also Table 6).GC–MS of the trimethylsilyl derivative was employed by Yasumoto et al.57 for the identification of selenomethionine in soybean proteins. Abrams and Burau111 identified and made a semi-quantitative determination of selenomethionine in soil after derivatization with a silylating reagent.The determination was performed by GC–MS and led to the results that about 20% of selenium (corresponding to about 4 mg g21 of Se) in the hydrophilic fulvate fraction (see Section 2.2.2) was present as selenomethionine. LC–MS and Fourier transform infrared (FTIR) spectrometry (the latter could be coupled with several separation techniques) do not appear to have been applied to the identification of selenium and tellurium compounds in the environment.The simultaneous use of MS and FTIR in identification work could be of great interest considering that they provide complementary structural information. The major limitation of FTIR detection, compared with MS, is its relatively low sensitivity. In spite of the most recent improvements in coupled techniques, FTIR detection limits are in the range from 50 pg (for GC) to 500 pg (for LC).317 Compact instrumentation for GC–MS and LC–MS, suitable for routine application at the femtogram level, is nowadays commercially available. Recently, some attempts at the identification of selenium compounds in enzyme extracts of cooked cod have been Table 9 continued— Gas chromatography— GC–FID Flame ionization detection.Used for determination of volatile alkyl selenides; MeSeCN and MeTeCN in the indirect determination of Se- and Te-methionine. LOD 5 pg. Poor selectivity 4.1.1 GC–FPD Flame photometric detection. Poor Se/S selectivity ratio. LOD 2 mg s21 for Se under optimized conditions. LODs for Te are worse by about one order of magnitude 4.1.1; 4.2.2 GC–ECD Electron-capture detection.Selective determination of SeIV after piazselenol derivatization with various o-diaminobenzene reagents followed by extraction into a suitable solvent. LOD in the range 0.2–5 pg of Se or a few ng l21 in the original samples. Disadvantages: manipulation of toxic reagents such as aromatic o-diamines 4.1.1 GC–FCD Fluorine-induced chemiluminescence detection.Simultaneous detection of S, Se and Te alkyl compounds. Good selectivity ratio versus hydrocarbons, not responsive for hydrides. LOD 6–22 pg, dynamic ranges span at least over three orders of magnitude 4.1.1 HG–GC–PID Photoionization detection. Hydride generation and trapping followed by GC separation and on-line PID. Simultaneous determination of Se, As, Sn and Sb.LOD 25 pg for Se 4.1.1 Fluorimetry— DAN–FD Selective derivatization of SeIV with varioius 2,3-diaminonaphthalene reagents followed by extraction and spectrofluorimetric determination. LOD 0.01–0.05 ng ml21. Interferences due to oxidation product of DAN and from sample matrix species. Disadvantage: manipulation of toxic reagents such as aromatic o-diamines 4.1.2 HPLC–FD Formation of various fluorophores for SeIV and Se-cysteine, followed by HPLC separation on-line fluorimetric detection. LOD 0.15–0.015 pg for SeIV and 67 ng ml21 for Se-cysteine 3.2.3; 4.1.2 Electrochemical methods— DPP, DPV Various differential pulse polarographic and voltammetric techniques. Determination of the electroactive species SeIV and TeIV. Serious interferences due to organic substances, surface-active substances and concomitant elements. LOD 0.01–2 ng ml21 for SeIV and 20 ng ml21 for TeIV. Dynamic ranges span over two order of magnitude 4.3.1 CSV, DPCSV Various stripping voltammetric techniques for determination of electroactive species SeIV and TeIV. Possibility of performing simultaneous determination of SeIV and TeIV. Interferences as for DPV and DPP techniques. LOD 0.5–100 pg ml21 for SeIV and 100 pg ml21 for TeIV (below 1 pg ml21 after preconcentration) 4.3.1 Piazelenol–DPP, –DPV, –CSV, –DPCSV Piazselenol derivatization of SeIV combined with various polarographic and voltammetric techniques. Interference from some concomitant elements and organic substances still a problem. LOD 10–400 pg ml21 4.3.1 HPLC–ED On-line selective electrochemical detection for HPLC determination of RSSR, RSSeR, RSSeSR, RSH and RSeH. LOD below 1027 mol l21 for bis(penicillamine)selenides (20 ml sample injection). Dynamic ranges span over three decades 3.2.3; 4.3.1 HPLC–IAPD On-line integrated pulsed amperometric detector for HPLC determination of Se-amino acids. Not specific for Se compounds. LOD 20–30 pg for Se-methionine and Se-cysteine. Dynamic ranges span over two decades 2.1.2; 4.3.1 Mass spectrometry— GC–MS GC coupled on-line with MS. Several applications to identification of unknown volatile selenium and tellurium compounds: methyl selenides, methyltellurides, dimethylselenone, dimethyl selenenyl sulfide 4.3.2 FNB–GC–MS Derivatization of hydrogen selenide and RSeH with FNB (see Table 6) followed by GC–MS. Used for identification of Se-cysteine in proteins and enzymes 4.3.2; 2.1.2 Silylation–GC–MS Various silylating agents used for derivatization of Se-amino acids to volatile compounds. Used for identification of Se-methionine in soil extracts and proteins 4.3.2; 2.1.1 LC–ES-MS LC interfaced on-line with electrospray MS detection. Identification of organoselenium compounds. Sensitivity impaired with respect standard performance of ES-MS 4.2.4 LC–ID–NTIMS Various LC techniques coupled off-line with isotope dilution negative thermal ionization MS. Determination of SeIV, SeVI, neutral/basic and acidic organoselenium species. LOD 12 pg g21 of Se for 250 ml water sample 4.2.4; 3.2.2 Analyst, December 1997, Vol. 122 139Rreported. These compounds were detected by HPLC–ICP-MS but their identification was not possible. HPLC–electrospray MS was then attempted in order to gain structural information, but the sensitivity was impaired with respect to the standard performance of the system. The authors concluded that a careful optimization of both the chromatographic and instrumental conditions is required.318 It seems very probable that the application of separation techniques coupled with MS or FTIR detection will give a decisive advance in environmental studies on organoselenium and organotellurium compounds. 5 Conclusions A summary of the analytical techniques employed for the determination of selenium and tellurium in environmental samples is reported in Table 9. General conclusions about the state of the art for the determination of selenium and tellurium species in samples of environmental origin cannot be drawn in a simple way. First, much more effort has been made for selenium than for tellurium about the development of appropriate analytical methods, owing to the recognized biochemical role played by selenium in living systems. However, many analytical problems still have to be solved, even for selenium determination, in fields such as the identification of unknown species and the speciation/determination of already known species. In both cases a common problem is represented by sample preparation and manipulation and by the availability of suitable procedures able to liberate selenium and tellurium compounds (mainly from the protein-bound fraction) from the samples without alteration of their original chemical structure. The second problem is related to the lack of suitable instrumentation able to give structural information on chemical species that can be present in the original samples at concentrations as low as a few parts per billion, sometimes lower. At present only GC–MS techniques appear to possess adequate sensitivity, but they require complex sample manipulations. The development and optimization of sensitive LC– MS techniques will have a great impact on identificative work for both selenium and tellurium compounds. In speciation analysis of already known compounds, the use of selective reactions and selective reagents, with a few exceptions, is of doubtful accuracy and is better achieved with the additional aid of separation techniques. Future trends should be devoted to the development of separation techniques coupled with highly sensitive element-specific/selective detectors that could be useful also for in-field determinations of those compounds which undergo degradation during sample storage. Currently, much effort is being devoted to the optimization of LC–ICP-MS techniques, and further improvements are to be expected soon. AFS is interesting for its sensitivity towards both selenium and tellurium when combined with gaseous sampling techniques (HG, GC).The impressive performance achievable by LEAFS could be of great impact in achieving highly sensitive element specific detection for chromatography. Finally, concerning the determination of total selenium and tellurium, for total selenium determination many adequate analytical methods have been developed and investigated for most environmental samples, involving the use of chromatography, electrochemistry, spectrochemistry and various coupled techniques. Also, the dissemination of knowledge about the advantages and disadvantages of the different procedures, including sample preparation and manipulation, is valuable. For total tellurium determination, the choice is mainly restricted to the most sensitive atomic spectrometric techniques such as ETAAS, HG–ETAAS and HG–ICP-MS. Further developments in vapor generation techniques coupled with AFS and application of LEAFS techniques could be of great interest in obtaining highly sensitive analytical tools for total element determination. The author thanks Professor S. Rapsomanikis and Professor D. L. Tsalev for their interest in and helpful discussions of this work. 6. (References here)

ISSN:0003-2654    

DOI:10.1039/a704759b

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Aqueous Sample Direct Extraction and Analysis by Membrane Extraction With a Sorbent Interface† Yu Z. Luo, Marc Adams and J. Pawliszyn* Guelph-Waterloo Center for Ground Water Research, Department of Chemistry, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1 A mathematical model was developed for aqueous sample analysis by membrane extraction with a sorbent interface (MESI). The model used in this paper includes the consideration of boundary layers which are located inside and outside of the membrane.In this study, benzene, toluene, ethylbenzene, trichloroethylene and hexane were the standard test analytes. Distribution constants of these analytes between the membrane and water were measured. Some significant parameters were investigated, such as agitation, temperature and headspace effect. Good agreement was found between the model and the experimental results. Keywords: Membrane extraction; gas chromatography; sample preparation; water analysis The field of membrane separation technology is presently in a state of rapid growth and innovation.Many different membrane separation processes have been developed. Membrane introduction mass spectrometry (MIMS) has been widely applied in air, water and biological analysis.1–5 The supported liquid membrane (SLM) technique has been coupled with GC and LC for the analysis of ionizable and charged species.6–8 Membrane extraction with a sorbent interface (MESI) coupled to a gas chromatograph was conceived as an exceptionally simple method for the sampling and analysis of trace compounds in the environment.9–12 The first application of this approach was published in 1992 by Pawliszyn et al.11,12 In that method, an aqueous sample was pumped through a single hollow fiber membrane while an inert gas flowed countercurrently around the exterior of the fiber.The volatile organic compounds (VOCs) permeated from the liquid phase across the membrane and into the gas phase where they were collected by cryofocusing and then thermally desorbed for GC analysis. Recently, more applications related to the MESI method have been reported.13–16 MESI is adaptable to continuous monitoring or field analysis.Understanding the mechanisms of MESI in terms of well established basic scientific theory will enable MESIAs rapid development for reliable, effective analysis in various analytical applications. A theory is put forward to explain the processes occurring in MESI extraction directly from stirred water.The theory includes the fluid dynamics around the membrane since diffusion through water is a significant part of the extraction process. This paper continues the study of membrane extraction with a sorbent interface, in which MESI was evaluated for air and headspace analysis.17,18 In those two papers, resistance to mass transfer in air and headspace was considered negligible. In order to validate the theory, some non-polar and polar compounds were selected as test analytes.The response times of the membrane to these compounds were measured and discussed. Extraction under different stirring speeds was studied and compared. Formation of bubbles on the membrane surface occurred when water samples had headspace, and their effect on extraction rate was studied. The agreement between model and experiment is sufficient to conclude that the model describes the dominant natural processes in MESI.Some implications for design and operation of the MESI system have been derived from the theory. Theory Fig. 1 shows the geometry of the membrane probe used in this study. The theory analyses MESI as diffusion according to Fick’s law through a hollow cylinder with mass transfer resistance at the boundaries and a mobile phase flowing inside. Fig. 2 illustrates the influence of the boundary layers in membrane extraction and the concentration gradients. The basic equations are from established theory and will apply to flowing gas or liquid samples, any membrane with Fickian diffusion, and gas or liquid mobile phase.Analyte transport in the MESI system is divided into five steps: 1 Convection and Diffusion Through Sample to the Membrane Outer Surface This complex phenomenon is understood by fluid dynamics theory. At the membrane outer surface, i.e., at r = b where r represents the radius from the membrane axis and b represents † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997.Fig. 1 Geometry of the hollow fiber membrane. Fig. 2 Schematic of boundary layers. Analyst, December 1997, Vol. 122 (1461–1469) 1461the radius from the membrane axis to the membrane outer surface, the mass transfer can be described by:19 C C K h D C r r b r b s ms o m - æ è ç ö ø ÷ = = = ¶ ¶ (1) where Cs is the analyte concentration in the bulk sample and is assumed to be constant (mg l21); ho is the coefficient of mass transfer from the sample to the membrane outer surface (kg s21 m22); Kms is the distribution constant of the analyte between the sample and the membrane, concentration in membrane divided by concentration in sample at the interface (dimensionless, mg l21: mg l21); Cýr=b is the analyte concentration in the membrane at its outer surface (mg l21); and Dm is the analyte’s diffusion coefficient in the membrane (cm2 s21).The mass transfer coefficient is given by ho = Nuo·Ds/2b where Nuo is the Nusselt number, and Ds is the analyte’s diffusion coefficient in the fluid.The Nusselt number for a cylinder in a fluid cross flow is:19 Nu Sc Sc Sc o d d d for and = + + ( ) é ë êê ù û úú > < 0 3 0 62 1 0 4 0 2 10000 0 5 1 4 1 3 2 3 . . . . , . Re Re Re (2) where Red is the Reynolds number and Sc the Schmidt number of the fluid, defined Red = u2b/n, Sc = n/Ds where u is fluid velocity (cm s21) and n its kinematic viscosity (cm2 s21).Let k1A = KmsDm/bho = 2KmsDm/NuoDs. The parameter k1A is a measure of the resistance to mass transfer at the membrane outer surface. When k1A = 0 concentration in the membrane at r = b is equal to sample bulk concentration times the distribution constant, i.e., concentration in the membrane is at its maximum possible. As k1A increases the concentration at r = b decreases from this maximum possible, i.e., a concentration drop occurs due to mass transfer resistance of the boundary layer. 2 Partitioning Between Sample and Membrane at its Outer Surface This process follows Henry’s law for an air sample, and Nernst’ law for a water sample. Henry’s law states that the ratio of the analyte partial pressure to Cýr=b at the interface is constant over low concentrations, and varies with exp(DH/RgT) where DH is the heat of sorption from sample to membrane and Rg is the gas constant. Since Kms is a ratio of mass-per-volume concentrations and pressure is assumed to be constant at 1 atm, Kms will vary with exp(DH/RT)/T.For non-polar solutes partitioning between air and polymer, DH is approximately equal to the heat of vaporization, available from published tables.20 Nernst’ law states that Kms is constant over low concentrations and varies with exp(DH/RT). 3 Diffusion Through Membrane Initial analyte concentration is constant throughout the sample, and zero in the membrane and mobile phase. End effects of the membrane are assumed to be negligible so diffusion is symmetric about and along the membrane axis, this is described by Fick’s law in one dimension, radius r: D r r r r C r t t C r t m ¶ ¶ ¶ ¶ ¶ ¶ , , ( ) æè ç öø ÷ = ( ) (3) where t is the time from the start of extraction. The analyte diffusion coefficient in a polymer varies with exp(2Ed/RgT) where Ed is the apparent activation energy for diffusion and Rg is the gas constant. 4 Partitioning Between Membrane and Mobile Phase at the Membrane Inner Surface This mass transfer step follows Henry’s or Nernst’ law, as in step 2. 5 Diffusion and Convection of Analyte Into Mobile Phase Which Flows out the Membrane This mass transfer step is understood by fluid dynamics theory, as in step 1, but concentration is not constant in the mobile phase.The boundary condition at r = a is modeled: D C r h C K C r a r a m i mg g ¶ ¶ = = = - æ è ç ö ø ÷ (4) where hi is the coefficient of mass transfer from the membrane to the mobile phase, Kmg is the distribution constant between the membrane and the mobile phase, concentration in membrane divided by concentration in mobile phase (dimensionless, mg l21 : mg l21), and Cg is the average bulk concentration in the mobile phase (mg l21).The bulk concentration in the mobile phase changes along the length of the membrane as it picks up analyte. The average concentration in the mobile phase is assumed to be a constant fraction, f, of the exit concentration.In other words the concentration profile along the length of the membrane is assumed to have the same shape from the start of extraction until steady-state. Flux through the membrane inner surface (mass in) must equal mobile phase exit concentration times flow rate (mass out) which gives: A D r C Q C f r a i m g ¶ ¶ = = (5) where Ai is the membrane inner surface area (cm2), Ai = 2paL, L is the length of the membrane, and Q is the mobile phase volumetric flow rate (ml min21) and is assumed to be constant.(Extraction rate change is assumed to be slow compared with contact time between membrane and an average element of mobile phase.) The mass transfer coefficient hi is calculated from the Nusselt number, the same way as ho, which for flow inside a cylinder is:19 Nu a Sc a Sc if i d d d = + + [ ] < 3 65 0 13 1 0 04 2 2300 2 3 . . . Re Re Re (6) where Red and Sc are as defined for eqn. (2) above except b is replaced by a. Combining eqns.(4), (5) and the relation between hi and Nui gives: a D K Nu D Lf Q C r C r a 2 1 0 m mg i g + æ è ç ö ø ÷ - = = p ¶ ¶ (7) k D K Nu D Lf Q 1 2 1 = + æ è ç ö ø ÷ m mg i g p . Let The parameter k1 is a measure of resistance to mass transfer at the membrane inner surface, similar to the significance of k1A at the outer surface. To estimate f it is determined for the steady state case. Let x represent distance along the membrane. At the membrane outer surface r = b the boundary condition is as described by eqn. (1).At the membrane inner surface r = a, bulk concentration in the mobile phase changes with x according to: 1462 Analyst, December 1997, Vol. 122C x a Q D r C x x r a g m d ( ) = ò = 0 2p ¶ ¶ (8) so the boundary condition can be written as: D C r h C K aD Q C r x r a r a x m i g m d ¶ ¶ p ¶ ¶ = = = - æ è ç ö ø ÷ ò 2 0 (9) (Diffusion along the axis in the membrane and in the mobile phase is assumed to be insignificant.) The steady-state boundary value problem specified by eqns.(1), (3) and (9) is solved by applying the Laplace transform in x to obtain a differential equation in one variable, r. This one dimensional problem is solved in ref. 21. From that result the inverse Laplace solution has the form C(r,x) = Kms[1 2 B(r)e2P0x] with two poles at p = 0 and p = p0: p Q K K Nu D Nu D K D 0 1 1 2 = + + - p f ms mg o s i g mg m ln (10) where f = b/a. Substituting this formula for C(r,x) in eqn. (8) C L C x L g g = ( ) ò 1 0 gives a formula for Cg(x) which is used in to give finally: f C C L e p L p L = ( ) = - - - g g 1 1 1 0 0 (11) independent of the exact expression for B(r).Note 0.5 < f < 1. Model The boundary value problem specified by eqns. (1), (3) and (7) is solved in ref. 21, giving an expression for C(r, t). From this solution the expression for extraction rate can be written down in the dimensionless parameters k1A, k1, f and W = Dmt/a2 as: G t A D r C r t A a D K C k k e F k J J k J J r a n n n n n n n n n ( ) = ( ) = ¢ + + + ( ) ì í ï î ï ¢ ( ) - ( ) ( ) + ( ) ü ý ï � ï = - � å i m i m s s ¶ ¶ f a fa fa fa a a a a , ln [ ][ ] 1 2 1 1 1 1 1 0 1 1 0 2 W (12) where ±a1, ±a2, ±a3 ...are the roots of: [k1aJ1(a) + J0(a)][k1AfaY1(fa) 2 Y0(fa)] 2 [k1aY1(a) + Y0(a)][k1AfaJ1(fa) 2 J0(fa)] = 0 and F(a) = (k1A2f2a2 + 1)[k1aJ1(a) + J0(a)]2 2 (k1 2a2 + 1)[k1AfaJ1(fa) 2 J0(fa)]2 where Ji and Yi are the Bessel functions of the first and second kind, of order i.A formula for response time, the time when extraction rate reaches 90% of its steady state, accurate to ± 15% for 1.3 < f < 5 was derived as explained in ref. 21 using a symbolic algebra program:22 t a D k k k k k k 90 2 2 2 2 1 2 2 1 2 1 1 1 1 2 1 1 1 2 1 2 2 1 % ln ln ln ln = + ( ) - + + - - ( ) - - - ( ) ¢ + - ( ) ¢ ¢ + + m f f f f f f f f f f (13) If k1A and k1 are not significant response time: t a D b a D 90 2 2 2 2 2 1 1 0 32 % ln . = - + - - æ è ç ö ø ÷ » - ( ) m m f f f (14) The formula for steady state extraction rate is seen from eqn.(12) to be: G LD K C k k LC Nu D K K D f L Q D K e m ms s s o s mg ms g m ms = ¢ + + = + + æ è ç ö ø ÷ + 2 1 1 3 65 2 1 1 p f p p f ln . ln (15) Concentration in the carrier gas approaches equilibrium if: L Q K K Nu D K D > + æ è ç ö ø ÷ 1 6 1 2 . ln p f ms mg o s ms m (16) from eqn. (8), and under this condition: Gc = 0.9 CsQKms/Kmg ± 10% (17) Note Kms/Kmg is 1 for extraction from air, and is the analyte Henry’s constant for extraction from water.Experimental Apparatus and Reagents Samples were prepared with purified deionized water from a Barnstead water filtration system (Barnstead, Dubuque, IA, USA) and with 99% pure standards of benzene, toluene, ethylbenzene, hexane, trichloroethylene (TCE), ethanol and butanol (Supelco Canada, Mississauga, Canada). Aqueous standards were prepared at ppm m/m level: benzene 7 ml, toluene 9 ml, ethylbenzene 9 ml, TCE 3 ml and hexane 5 ml. Samples were prepared by diluting the concentrated aqueous standard one hundred fold, by pipetting 5 ml into a 500 ml volumetric flask of water, or by syringe addition of 0.4 ml to a 40 ml vial of water.In all aqueous samples the exact concentration for benzene, toluene, ethylbenzene, trichloroethylene, hexane, ethanol and butanol was 123, 156, 156, 88, 66, 98 and 95 ppb m/m. Samples were held in 40 ml amber vials with hole caps, Teflon lined septa and a 2.0 cm magnetic stir bar, filled completely unless otherwise stated.Samples were always agitated by magnetic stir bar at 1500 rpm, using a VWR stirrer/ hot plate (VWR Scientific, Mississauga, Canada). The average speed of water flowing past the membrane was estimated as 55 cm s21 using formula 3.41 in ref. 23. The temperature for all experiments was 25 ± 0.5 °C, and was monitored by thermocouple. Standards for GC calibration by syringe injection were prepared at 100 ppm m/m in reagent grade methanol, each analyte being stored in a separate 5 ml volumetric flask.A one point, one replicate calibration check was carried out each day using benzene. Whenever the response factor changed by more than 15% from the previous calibration curve, a new five point calibration curve was determined, with three replicates determined for each point. A five point calibration was always carried out for each new series of experiments, before or after, and at least once every two weeks in any case.The MESI system was described in detail previously.18 The gas line from the tank regulator was connected to the GC needle valve for carrier gas, and from there it was connected to the membrane module. Ultra-high-purity (UHP) grade nitrogen (Praxair, Waterloo, Canada) was used as the carrier gas at a flow rate of 2.2 ml min21, unless otherwise stated, as measured at the Analyst, December 1997, Vol. 122 1463membrane exit. UHP grade nitrogen, zero air and hydrogen (Praxair) were used as make-up, combustion and fuel gases, respectively for the FID with corresponding flow rates of 50, 300 and 30 ml min21, for all experiments.Flow rates were measured with a bubbeter at ambient temperature and pressure. The Varian 3500 GC instrument (Varian Canada, Mississauga, Canada) used had a split/splitless injector. For syringe or solid-phase micro-extraction (SPME) injections the column and gas line were disconnected from the membrane module and connected to the split/splitless injector.No split was used. Fig. 3 is a schematic diagram of the apparatus used for measuring the distribution constant of an analyte between the membrane and water. An SPME fibre assembly (Supelco Canada) was modified by replacing the outer needle with 5 cm long 8 gauge stainless steel tubing glued to the septum, and the fibre support rod was replaced with 7 cm long 31 gauge stainless steel tubing. The plastic hub from the original fibre support rod was glued to the top end of the 31 gauge tubing as a plastic plunger cap.At the bottom end of the 31 gauge tubing a 1 cm piece of membrane was forced on. The method development and procedure followed the standard methods of SPME as described in ref. 23. In this research, a 40 ml sample volume was chosen for the K value measurements, the aqueous samples were kept at 25 °C ± 0.5. The desorption conditions were 60 s at 200 °C. The piece of membrane was exposed to the stirred sample at the center of the vial for progressively longer times, 10 s up to 60 s, then 20 s up to 120 s, and finally 2.5, 3, 5 and 10 min. The GC oven program was 40 °C (hold for 0.2 min) to 200 °C (hold 5 min) at 20 °C min21.Calibration was by syringe injection, using the same oven program. For experiments to determine response time the membrane probe was connected directly to the GC detector via a 20 cm 3 0.32 mm id deactivated silica tubing (Supelco Canada).The membrane probe was initially exposed to air in a clean vial for 1 min to get a flat base line. The membrane probe was then placed in a vial full of solution and stirred. In a time period of 7 min, a rising signal and then a flat signal line were observed. Finally, the membrane probe was taken out of the solution and placed in a vial of pure stirred water. During this time a dropping signal line was observed. For the extraction rate measurements the MESI sorbent trapping time was 1 min and at steady state the peak areas were recorded.The peak area counts were calibrated by the syringe injection calibration explained before. Results and Discussion Tables 1 and 2 list the parameter values determined for the experiments in this paper. The distribution constants were determined by the SPME method, using the equation: K = nmvs/[vm(c0vs 2 nm)] (18) where nm is the amount extracted by the membrane piece used as an SPME fibre, vs is sample volume, vm is the membrane volume, c0 is the initial concentration of sample.The results show that nonpolar compounds have better affinities for the hydrophobic membrane. The distribution constants must also be measured for mobile phase. The distribution constants of analytes between membrane and nitrogen are the same as between the membrane and air.27 The distribution constants between membrane and air were measured by the SPME method and reported previously.18 Diffusion coefficient measurements have been performed and reported previously.18 Siloxane absorbs some water, but water is assumed to have a negligible effect on the diffusion coefficient in the polymer. A few studies suggest that water does not affect polymer diffusion coefficients of hydrophobic compounds,28 although this observation is not an established principle.An experiment using air extraction under varying humidity showed no significant effect of humidity on extraction rates.To determine the response times the membrane was connected directly to the FID and the signal versus time profile was recorded. Fig. 4 shows a typical FID signal versus time profile, Fig. 3 Schematic of assembly of SPME membrane. Table 2 Physico-chemical parameter values at 25 °C Kms water Dm/1026 cm2 s21 (determined Dwater †/ (determined Analyte experimentally) Kms * air 1025 cm2 s21 Dair ‡/cm2 s21 experimentally) DN2§ Benzene 136 485 1.09 0.093 2.12 0.38 Toluene 346 1872 0.95 0.085 1.59 — Ethylbenzene 847 3380 0.9 0.0755 1.09 — TCE 182 443 0.96 0.0875 1.81 — Hexane 126 224 0.9 0.0732 2.27 — Methanol 29 — 1.66 0.152 — — Isobutyl alcohol 52 — 0.93 0.088 — — * See ref. 18. † See ref. 24. ‡ See ref. 25. § See ref. 26. Table 1 Specifications of the experimental apparatus Parameter Value a, Membrane inner radius 0.0015 cm b, Membrane outer radius 0.00315 cm L, Length of membrane 4 cm Q, Flow rate of carrier gas 2.2 ml min21 Water sample temperature 25 °C Water sample speed 55 cm s21 1464 Analyst, December 1997, Vol. 122produced by benzene. Table 3 summarizes the response times determined for the different compounds with this MESI system. Good agreement is found between the theoretical predictions and the experimental results. Theory and experiment were also compared for extraction rates. The comparisons of extraction rates for the test compounds are listed in Table 3. An extraction versus time profile involves both response time and extraction rate.In Fig. 5, curve A is the model prediction of the benzene extraction versus time profile and curve B is the experimental result, showing good agreement. Figs. 6(a) and 6(b) show the chromatograms obtained from a benzene sample with and without stirring. The figures show a decrease in response time and an increase in extraction rate with stirring. In Fig. 6(b), the peaks declined after 12 min of extraction, attributed to depletion of analyte in the sample.Figs. 7(a) and 7(b) show the extraction versus time profiles for a benzene sample at 400 and 1200 rpm, respectively. As expected sample stirring enhanced extraction rate and reduced response time. The agreement between model and experiment is sufficient to conclude that the model of membrane extraction describes accurately the dominant natural processes in MESI. Figs. 8(a) and 8(b) illustrate the theoretical relationship between response time and mass transfer resistance at the outside membrane surface, i.e., t90% vs k1A, and the relationship between response time and mass transfer resistance at the inside membrane surface, i.e., t90% vs k1.Calculations with eqns. (13) and (15) show that response time and steady state extraction rate will be influenced by boundary effects at the membrane outside surface if: 1 0 015 Nu D K D o s ms m > . (19) i.e., if k1A > 0.03, for the aspect ratio of the membrane used in this study, f = 2.08. Likewise response time and steady state extraction rate will be influenced by boundary effects at the membrane inside surface if: 1 3 65 0 03 .. D f L Q K D g g m + > p (20) i.e., if k1 > 0.06, for the aspect ratio f = 2.08. Table 4 summarizes the range of parameter values possible for VOCs in air and water. Theory predicts that boundary effects can be significant for extraction from air or water, especially when compounds have high distribution constants between membrane and sample, based on the criteria of eqns.(19) and (20) and the parameter values in Table 4. The outside boundary effect, i.e., k1A, will likely be significant for extraction from air if Kms > 2000, and for extraction from water in most cases. One consequence of this characteristic is if the outside boundary effect is significant the system must have a constant sample flow speed for reproducibility. The membrane can be surrounded by a chamber with a stirrer to control sample flow speed.Such a mechanism to increase sample flow speed will also lower response time and minimize trapping time. General conclusions on the properties of the MESI system are not simple because diffusion coefficients of VOCs in polymers can vary by orders of magnitude for one compound in different polymers, or for different compounds in one polymer: for example, at room temperature benzene has D = 0.48 3 10212 cm2 s21 in polyvinyl acetate, and D = 1.3 3 1025 cm2 s21 in natural rubber; in polyacrylate, D = 6.2 3 10210 cm2 s21 for benzene and D = 1.0 31027 cm2 s21 for methanol.28 Diffusion in the polymer can also be sensitive to compounds absorbed in the polymer so extraction rate could vary irreproducibly for samples with high concentrations of organics, such as effluents or biological fluids.The diffusion coefficient is assumed constant for the analyte concentration ranges in this study. The boundary condition at the inside surface is measured by the value of k1.This parameter can be controlled by the experimental settings, and it varies little with experimental conditions such as temperature. A short membrane and high mobile phase flow rate will reduce the value of k1. In some cases experimental settings can be chosen to make boundary effects negligible on extraction rate and response time, i.e., to make k1A and k1 negligible. An experimental setup with negligible boundary effects was used to measure Dm using eqn.(14), by extraction from air with a high mobile phase flow rate.18 No simple relationship with respect to temperature is derived. Therefore, as an example, the relationships of response time and extraction rate with temperature were calculated for benzene, shown in Figs. 9(a) and (b). These calculations used the parameter values determined for this experimental system, and all of their individual relationships with T as described in Table 4. In this case the variation in response time with T is dominated by the variation in the polymer diffusion coefficient, Dm.The extraction rate variation is dominated by variation in the value of Kms Dm . The overall change of extraction rate depends on the Fig. 4 FID signal time profiles of benzene in response time measurement. Table 3 Comparison of theoretical and experimental results. t0–90% represents the time of signal climbing and from 0% to 90% of steady-state intensity, and t100–10% dropping from 100% to 10%.RSD are based on 3 replicates % Difference % Difference between between Response time†/s theoretical and theoretical and experimental Ge § (RSD)/ experimental Analyte t90% */s t0–90% (% RSD) t100–10% (% RSD) t0–90% Ge ‡/ng s21 ng s21 Ge Benzene 136 113 (3.9) 106 (6.2) 16 0.55 0.55 (3.3) 0 Toluene 347 269 (5.8) 308 (7.8) 22 0.73 0.72 (4.2) 21.3 Ethylbenzene 701 641 (5.9) 856 (6.5) 8.5 0.91 0.95 (3.8) 24.4 TCE 163 131 (3.7) 157 (2.4) 19 0.45 0.41 (3.4) 8.9 Hexane 113 120 (2.1) 296 (6.7) 26.9 0.32 0.28 (3.7) 12 * Determined by eqn. 13. † Determined experimentally. ‡ Determined by eqn. 15. § Determined experimentally. Analyst, December 1997, Vol. 122 1465change of KmsDm. Fig. 10 gives the experimental results showing variation of the extraction amount of benzene from aqueous solution under different temperatures. It can be seen that with the increase of temperature the amount of benzene extracted increased, hence the extraction rate increased.Fig. 5 Extraction tie profiles of benzene without trapping. A, Model prediction and B, experimental result. Fig. 6 Extraction time profile of benzene, (a) from a relatively static sample and (b) with stirring. Fig. 7 Extraction time profile of benzene at different stirring speeds (rpm): (a) 400 and (b) 1200. Fig. 8 Relationship between response time and parameter of mass transfer at the membrane surfaces, (a) outside surface k1A, and (b) inside surface k1. The solid line is calculated with the exact model [eqn. (12)], the dashed line with the approximation formula [eqn.(13)]. Fig. 9 Relationship of (a) response time and (b) extraction rate with temperature, as calculated by the relationships described in Table 4. 1466 Analyst, December 1997, Vol. 122Experiments also revealed that a small headspace can increase the extraction rate from aqueous samples. Comparing the extraction amounts without headspace and with 0.5 ml headspace shows that the total extraction amount for the first 20 min increased by 10% to 20% depending on the compound.With headspace, a large number of small gas bubbles adhere to the membrane surface. A compressed headspace allows carrier gas to penetrate the membrane wall and form bubbles on the membrane surface. Although the analyte still needs to diffuse through water to the gas bubbles, mass transfer was enhanced because the small gas bubbles had a large surface area, the molecules of analyte could easily diffuse through those gas bubbles and, in general, compounds have larger distribution constants between membrane–air than between membrane– water.With a headspace concentration in the aqueous solution drops somewhat because analytes distribute into the headspace, but overall the extraction rate is increased if the headspace is small. Some implications for design and operation of the MESI system can be derived from the theory. Quantitative analysis by the MESI method depends on calibration, and on control of significant factors to ensure reproducibility.The simplest situation in which to calibrate the steady state extraction rate is when the carrier gas reaches equilibrium before exiting the membrane. This situation will occur under the conditions specified by eqn. (16). If constraints do not allow the carrier gas to reach equilibrium, then extraction rate will be sensitive to the several factors listed in eqn. (15), and controlling the significant factors or compensating for their variations will be more complicated.The diffusion coefficient in the polymer will be especially difficult to evaluate, since this parameter is sensitive to so many factors as explained above. Therefore, when designing a MESI system for a new application a membrane with a low value of lnf/KmgDm is desirable since it can make practical calibration by eqn. (17), for the simplest and most reliable calibration. Membrane types could be compared according to their value of lnf/KmgDm. Likewise when choosing membrane length and carrier gas flow rate, a value of L/Q that allows calibration by eqn.(17) would simplify. Increasing L/Q does increase the response time, but by at most a factor of 2 for aspect ratio f = 2.08 by eqn. (13). Increasing carrier gas flow rate can also increase extraction rate, so a shorter trapping time, and thus a shorter effective response time, would obtain the same response. However, a high carrier gas flow rate results in analyte breakthrough in the sorbent trap.Membrane choice also depends on characteristics such as selectivity, and mechanical or chemical durability. Besides the factors described above, if MESI is applied directly to an effluent stream or biological system rich in organic compounds, significant interference could cause high background noise or affect the parameters that govern response time and extraction rate in a non-reproducible manner.Humic or other materials in a sample could foul a membrane. In that situation, a headspace approach is more suitable.17 Table 4 Range of parameter values at 300 K and how each varies with T over ambient temperatures in different phases of the MESI system. All symbols are defined in the Glossary Phase n/cm2 s21 D, for VOCs/cm2 s21 Kms, for VOCs (dimensionless) Other Air 0.15, Å T1.8 0.05–0.16, Å T1.8 (ref. 19) 0.1–115 000, Å e2DH/RT NuoDs = 0.015 + 0.099(ub)1/2 (ref. 19) (refs. 29, 30) to 0.045 + 0.39(ub)1/2 in g cm s Example values of DH/Rg:20 units. An analyte at constant 5160 for benzene mole fraction will vary 4700 for toluene concentration [ML23] with 1/T. Water n(T) Å 0.1083– 0.5–1.5 3 1025, Å T/n(T) 0.1–7000, refs. 29, 31 NuoDs = 0.00057(ub)1/2 to 0.000332T (ref. 19) Å e2DH/RT or 0.0012(ub)1/2 in g cm s (ref. 19) Å 1/solubility (T). units, Å n(T)2/3/T5/6. Example value of DH/Rg, estimated from water solubility tables32: 525 for benzene Polymer 10213–1025, Å e2EdRgT D is sensitive to compounds membrane (refs. 26, 28) absorbed in the polymer.Example values of Ed/Rg: (PDMS will absorb 720 mg cm23 7700 benzene in natural rubber, water33 with D of water = 2900 toluene in polystyrene, 7 3 1025 cm2 s21 28) e.g., in 300 benzene in the membrane polyvinylacetate: compared to used in this study, estimated 1 ppm, 1000 ppm of allyl from measurements reported chloride changes its D 60%; before.10 water in polymer greatly accelerates the sorption rate of acetone.28 Gas mobile phase 0.1–1.0, Å T1.8 0.022–0.38, Å T1.8 Partition coefficients between At a constant mass flow rate the (ref. 19) (refs. 19, 26) polymer and any gas are volumetric flow rate Q will practically the same as between vary with T. (If tank and polymer and air.27 needle valve are at ambient temp T and the valve is not adjusted mass flow rate will not vary significantly with T.34) Fig. 10 Temperature effect on extraction rate of benzene solution.Analyst, December 1997, Vol. 122 1467This project has been financially supported by the Dow Chemical Company and the Natural Sciences and Engineering Research Council of Canada. Appendix Glossary a Radius from the membrane axis to the membrane inner surface (cm; see Fig. 1). b Radius from the membrane axis to the membrane outer surface (cm; see Fig. 1). f Ratio of average concentration to exit concentration in mobile phase (dimensionless). ho,hi Mass transfer coefficients at the membrane outer and inner surfaces, respectively (kg s21 m22).k1A KmsDm/bho = 2KmsDm/NuoDs (dimensionless). This parameter is a measure of the resistance to mass transfer at the membrane outer surface. 2 1 D K Nu D Lf Q m g i g + æ è ç ö ø ÷ p k1 (dimensionless). This parameter is a measure of the resistance to mass transfer at the membrane inner surface. r Radius since the membrane axis (cm). t Time from the start of extraction (s). t90% Response time, t, at which extraction rate reaches 90% of its steady state value (s).This is also the time from any change in Cs until 90% of the resulting change in steady state extraction rate. u Sample fluid velocity (cm s21). x Distance along the length of the membrane (cm). x = 0 at the point where mobile phase enters the membrane, and x = L at the exit. Ai Membrane inner surface area (cm2), Ai = 2paL. B(r) A function of r in the expression for C(r,x), not given explicitly because it cancels out.C or C(r,t) Analyte concentration in the membrane, a function of r and t (mg l21). Cs Analyte concentration in the bulk sample (mg l21). Cg(x) Analyte bulk concentration in the mobile phase as a function of x (mg l21). C Average bulk concentration in the mobile phase (mg l21). D Diffusion coefficient (cm2 s21). Ds Analyte’s diffusion coefficient in the sample (cm2 s21). Dm Analyte’s diffusion coefficient in the membrane (cm2 s21). Dg Analyte’s diffusion coefficient in the mobile phase (cm2 s21).Ed Analyte’s apparent activation energy for diffusion in a polymer. G(t) Overall extraction rate of the MESI membrane and mobile phase (ng s21). Ge Overall steady-state extraction rate (ng s21). DH Analyte’s heat of sorption from sample to membrane. Ji,Yi Bessel functions of the first and second kind, of order i. Kms Analyte distribution constant between membrane and sample, concentration in membrane divided by concentration in sample at the interface (dimensionless; mg l21: mg l21).Kmg Analyte distribution constant between membrane and mobile phase, concentration in membrane divided by concentration in mobile phase (dimensionless; mg l21: mg l21). L Length of the membrane (cm). Nuo,Nui Nusselt numbers at the outer and inner membrane surfaces, respectively (dimensionless). P0 Pole in Laplace transform. Rg Gas constant. Q Mobile phase volumetric flow rate (ml min21). Red Reynold’s number of a fluid, defined as Red = ud/ n where d is diameter of the membrane outer or inner surface depending on context (dimensionless).Sc Schmidt number of a fluid, defined as Sc = n/Ds or n/Dg (dimensionless). T Absolute temperature in degrees Kelvin. n Fluid kinematic viscosity (cm2 s21). f Ratio of membrane’s outer to inner radius, = b/a (dimensionless). W = Dmt/a2, dimensionless time parameter (dimensionless). References 1 Lotiaho, T., Lauritsen, F. R., Choudhury, T. K., Cooks, R. G., and Tsao, G. T., Anal.Chem., 1991, 63, 875A. 2 Lauritsen, F. R., Int. J. Mass Spectrom. Ion Proc., 1990, 95, 259. 3 Carlsen, H. N., Jørgensen, L., and Degn, H., Appl. Microbiol. Biotechnol., 1991, 35, 124. 4 LapPack, M. A., Tou, J. C., and Enke, C. G., Anal. Chem., 1990, 62, 1265. 5 Virkki, V. T., Ketola, R. A., Ojala, M., Kotiaho, T., Komppa, V., Grove, A., and Faccchetti, S., Anal. Chem., 1995, 67, 1421. 6 J�onsson, J. Å., and Mathiasson, L., Trends Anal. Chem., 1992, 11, 106. 7 Lindegråd, B., J�onsson, J.Å., and Mathiasson, L., J. Chromatogr., 1992, 573, 191. 8 Thordarson, E., P�almarsd�ottir, S., Mathiasson, L., and J�onsson, J. Å., Anal. Chem., 1996, 68, 2559. 9 Yang, M. J., Harms, S., Luo, Y. Z., and Pawliszyn, J., Anal. Chem., 1994, 66, 1339. 10 Yang, M. J., Luo, Y. Z., and Pawliszyn, J., Chemtech., 1994, 24, 31. 11 Pratt, K. F., and Pawliszyn, J., Anal. Chem., 1992, 64, 2101. 12 Pratt, K. F., and Pawliszyn, J., Anal. Chem., 1992, 64, 2107. 13 Mitra, S., Zhang, L., Zhu, N., and Guo, X., J.Chromatogr., 1996, 736, 165. 14 Burger, B. V., Burger, W. J. G., and Burger, I., J. High Resolut. Chromatogr., 1996, 19, 571. 15 Yang, M. J., and Pawliszyn, J., Anal. Chem., 1993, 65, 2538. 16 Ortner, E. K., and Rohwer, E. R., J. High Resolut. Chromatogr., 1996, 19, 336. 17 Yang, M. J., Adams, M., and Pawliszyn, J., Anal. Chem., 1996, 68, 2782. 18 Luo, Y. Z., Adams, M., and Pawliszyn, J., Anal. Chem., in the press. 19 Transfer Processes, ed. Edwards, K., Dennry, V.E., and Mills, A. F., Hemisphere Publishing, New York, 2nd edn., 1978. 20 Martos, P. A., and Pawliszyn, J., Anal. Chem., 1996, 69, 206. 21 Carslaw, H. S., and Jaeger, J. C., Conduction of Heat in Solids, Clarendon Press, Oxford, 2nd edn., 1986, section 13.4. 22 Maple V Release 4, Waterloo Maple, Waterloo, Canada, 1996. 23 Pawliszyn, J., SPME, John Wiley and Sons, New York, 1997. 24 Hayduk, W., and Laudie, H., AIChE J., 1974, 20, 611. 25 Lugg, G. A., Anal.Chem., 1968, 40, 1072. 26 Jost, W., Diffusion in Solids, Liquids, Gases, Academic Press, New York, 1960, p. 412. 27 Berezkin, V. G., Korolev, A. A., and Maluykova, I. V., Proc. Int. Symp. Capillary Chromatogr., 18th, 1996, pp. 396–404. 28 Diffusion in Polymers, ed. Crank, J., and Park, G. S., Academic Press, London, 1968. 29 Pan, L., Adams, M., and Pawliszyn, J., Anal. Chem., 1995, 67, 4396. 1468 Analyst, December 1997, Vol. 12230 Zhang, Z., and Pawliszyn, J., J. High Resolut.Chromatogr., 1996, 2, 155. 31 Langenfeld, J., Hawthorne, S., and Miller, D., Anal. Chem., 1996, 68, 144. 32 Freier, R. K., Aqueous Solutions, Walter de Gruyter, Berlin, 1976. 33 Klunder, G. L., and Russo, R. E., Appl. Spectrosc., 1995, 49, 379. 34 McGraw-Hill Encyclopedia of Science and Technology, McGraw- Hill, 1997, vol. 12, pp. 124–126. Paper 7/06441A Received September 3, 1997 Accepted November 4, 1997 Analyst, December 1997, Vol. 122 1469 Aqueous Sample Direct Extraction and Analysis by Membrane Extraction With a Sorbent Interface† Yu Z.Luo, Marc Adams and J. Pawliszyn* Guelph-Waterloo Center for Ground Water Research, Department of Chemistry, University of Waterloo, Waterloo, Ontario, Canada N2L 3G1 A mathematical model was developed for aqueous sample analysis by membrane extraction with a sorbent interface (MESI). The model used in this paper includes the consideration of boundary layers which are located inside and outside of the membrane. In this study, benzene, toluene, ethylbenzene, trichloroethylene and hexane were the standard test analytes.Distribution constants of these analytes between the membrane and water were measured. Some significant parameters were investigated, such as agitation, temperature and headspace effect. Good agreement was found between the model and the experimental results. Keywords: Membrane extraction; gas chromatography; sample preparation; water analysis The field of membrane separation technology is presently in a state of rapid growth and innovation. Many different membrane separation processes have been developed.Membrane introduction mass spectrometry (MIMS) has been widely applied in air, water and biological analysis.1–5 The supported liquid membrane (SLM) technique has been coupled with GC athe analysis of ionizable and charged species.6–8 Membrane extraction with a sorbent interface (MESI) coupled to a gas chromatograph was conceived as an exceptionally simple method for the sampling and analysis of trace compounds in the environment.9–12 The first application of this approach was published in 1992 by Pawliszyn et al.11,12 In that method, an aqueous sample was pumped through a single hollow fiber membrane while an inert gas flowed countercurrently around the exterior of the fiber.The volatile organic compounds (VOCs) permeated from the liquid phase across the membrane and into the gas phase where they were collected by cryofocusing and then thermally desorbed for GC analysis.Recently, more applications related to the MESI method have been reported.13–16 MESI is adaptable to continuous monitoring or field analysis. Understanding the mechanisms of MESI in terms of well established basic scientific theory will enable MESIAs rapid development for reliable, effective analysis in various analytical applications. A theory is put forward to explain the processes occurring in MESI extraction directly from stirred water.The theory includes the fluid dynamics around the membrane since diffusion through water is a significant part of the extraction process. This paper continues the study of membrane extraction with a sorbent interface, in which MESI was evaluated for air and headspace analysis.17,18 In those two papers, resistance to mass transfer in air and headspace was considered negligible. In order to validate the theory, some non-polar and polar compounds were selected as test analytes.The response times of the membrane to these compounds were measured and discussed. Extraction under different stirring speeds was studied and compared. Formation of bubbles on the membrane surface occurred when water samples had headspace, and their effect on extraction rate was studied. The agreement between model and experiment is sufficient to conclude that the model describes the dominant natural processes in MESI.Some implications for design and operation of the MESI system have been derived from the theory. Theory Fig. 1 shows the geometry of the membrane probe used in this study. The theory analyses MESI as diffusion according to Fick’s law through a hollow cylinder with mass transfer resistance at the boundaries and a mobile phase flowing inside. Fig. 2 illustrates the influence of the boundary layers in membrane extraction and the concentration gradients. The basic equations are from established theory and will apply to flowing gas or liquid samples, any membrane with Fickian diffusion, and gas or liquid mobile phase.Analyte transport in the MESI system is divided into five steps: 1 Convection and Diffusion Through Sample to the Membrane Outer Surface This complex phenomenon is understood by fluid dynamics theory. At the membrane outer surface, i.e., at r = b where r represents the radius from the membrane axis and b represents † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997.Fig. 1 Geometry of the hollow fiber membrane. Fig. 2 Schematic of boundary layers. Analyst, December 1997, Vol. 122 (1461–1469) 1461the radius from the membrane axis to the membrane outer surface, the mass transfer can be described by:19 C C K h D C r r b r b s ms o m - æ è ç ö ø ÷ = = = ¶ ¶ (1) where Cs is the analyte concentration in the bulk sample and is assumed to be constant (mg l21); ho is the coefficient of mass transfer from the sample to the membrane outer surface (kg s21 m22); Kms is the distribution constant of the analyte between the sample and the membrane, concentration in membrane divided by concentration in sample at the interface (dimensionless, mg l21: mg l21); Cýr=b is the analyte concentration in the membrane at its outer surface (mg l21); and Dm is the analyte’s diffusion coefficient in the membrane (cm2 s21).The mass transfer coefficient is given by ho = Nuo·Ds/2b where Nuo is the Nusselt number, and Ds is the analyte’s diffusion coefficient in the fluid. The Nusselt number for a cylinder in a fluid cross flow is:19 Nu Sc Sc Sc o d d d for and = + + ( ) é ë êê ù û úú > < 0 3 0 62 1 0 4 0 2 10000 0 5 1 4 1 3 2 3 .. . . , . Re Re Re (2) where Red is the Reynolds number and Sc the Schmidt number of the fluid, defined Red = u2b/n, Sc = n/Ds where u is fluid velocity (cm s21) and n its kinematic viscosity (cm2 s21).Let k1A = KmsDm/bho = 2KmsDm/NuoDs. The parameter k1A is a measure of the resistance to mass transfer at the membrane outer surface. When k1A = 0 concentration in the membrane at r = b is equal to sample bulk concentration times the distribution constant, i.e., concentration in the membrane is at its maximum possible. As k1A increases the concentration at r = b decreases from this maximum possible, i.e., a concentration drop occurs due to mass transfer resistance of the boundary layer. 2 Partitioning Between Sample and Membrane at its Outer Surface This process follows Henry’s law for an air sample, and Nernst’ law for a water sample.Henry’s law states that the ratio of the analyte partial pressure to Cýr=b at the interface is constant over low concentrations, and varies with exp(DH/RgT) where DH is the heat of sorption from sample to membrane and Rg is the gas constant. Since Kms is a ratio of mass-per-volume concentrations and pressure is assumed to be constant at 1 atm, Kms will vary with exp(DH/RT)/T.For non-polar solutes partitioning between air and polymer, DH is approximately equal to the heat of vaporization, available from published tables.20 Nernst’ law states that Kms is constant over low concentrations and varies with exp(DH/RT). 3 Diffusion Through Membrane Initial analyte concentration is constant throughout the sample, and zero in the membrane and mobile phase. End effects of the membrane are assumed to be negligible so diffusion is symmetric about and along the membrane axis, this is described by Fick’s law in one dimension, radius r: D r r r r C r t t C r t m ¶ ¶ ¶ ¶ ¶ ¶ , , ( ) æè ç öø ÷ = ( ) (3) where t is the time from the start of extraction.The analyte diffusion coefficient in a polymer varies with exp(2Ed/RgT) where Ed is the apparent activation energy for diffusion and Rg is the gas constant. 4 Partitioning Between Membrane and Mobile Phase at the Membrane Inner Surface This mass transfer step follows Henry’s or Nernst’ law, as in step 2. 5 Diffusion and Convection of Analyte Into Mobile Phase Which Flows out the Membrane This mass transfer step is understood by fluid dynamics theory, as in step 1, but concentration is not constant in the mobile phase. The boundary condition at r = a is modeled: D C r h C K C r a r a m i mg g ¶ ¶ = = = - æ è ç ö ø ÷ (4) where hi is the coefficient of mass transfer from the membrane to the mobile phase, Kmg is the distribution constant between the membrane and the mobile phase, concentration in membrane divided by concentration in mobile phase (dimensionless, mg l21 : mg l21), and Cg is the average bulk concentration in the mobile phase (mg l21).The bulk concentration in the mobile phase changes along the length of the membrane as it picks up analyte. The average concentration in the mobile phase is assumed to be a constant fraction, f, of the exit concentration. In other words the concentration profile along the length of the membrane is assumed to have the same shape from the start of extraction until steady-state.Flux through the membrane inner surface (mass in) must equal mobile phase exit concentration times flow rate (mass out) which gives: A D r C Q C f r a i m g ¶ ¶ = = (5) where Ai is the membrane inner surface area (cm2), Ai = 2paL, L is the length of the membrane, and Q is the mobile phase volumetric flow rate (ml min21) and is assumed to be constant.(Extraction rate change is assumed to be slow compared with contact time between membrane and an average element of mobile phase.) The mass transfer coefficient hi is calculated from the Nusselt number, the same way as ho, which for flow inside a cylinder is:19 Nu a Sc a Sc if i d d d = + + [ ] < 3 65 0 13 1 0 04 2 2300 2 3 . . . Re Re Re (6) where Red and Sc are as defined for eqn. (2) above except b is replaced by a. Combining eqns. (4), (5) and the relation between hi and Nui gives: a D K Nu D Lf Q C r C r a 2 1 0 m mg i g + æ è ç ö ø ÷ - = = p ¶ ¶ (7) k D K Nu D Lf Q 1 2 1 = + æ è ç ö ø ÷ m mg i g p .Let The parameter k1 is a measure of resistance to mass transfer at the membrane inner surface, similar to the significance of k1A at the outer surface. To estimate f it is determined for the steady state case. Let x represent distance along the membrane. At the membrane outer surface r = b the boundary condition is as described by eqn.(1). At the membrane inner surface r = a, bulk concentration in the mobile phase changes with x according to: 1462 Analyst, December 1997, Vol. 122C x a Q D r C x x r a g m d ( ) = ò = 0 2p ¶ ¶ (8) so the boundary condition can be written as: D C r h C K aD Q C r x r a r a x m i g m d ¶ ¶ p ¶ ¶ = = = - æ è ç ö ø ÷ ò 2 0 (9) (Diffusion along the axis in the membrane and in the mobile phase is assumed to be insignificant.) The steady-state boundary value problem specified by eqns.(1), (3) and (9) is solved by applying the Laplace transform in x to obtain a differential equation in one variable, r. This one dimensional problem is solved in ref. 21. From that result the inverse Laplace solution has the form C(r,x) = Kms[1 2 B(r)e2P0x] with two poles at p = 0 and p = p0: p Q K K Nu D Nu D K D 0 1 1 2 = + + - p f ms mg o s i g mg m ln (10) where f = b/a. Substituting this formula for C(r,x) in eqn. (8) C L C x L g g = ( ) ò 1 0 gives a formula for Cg(x) which is used in to give finally: f C C L e p L p L = ( ) = - - - g g 1 1 1 0 0 (11) independent of the exact expression for B(r).Note 0.5 < f < 1. Model The boundary value problem specified by eqns. (1), (3) and (7) is solved in ref. 21, giving an expression for C(r, t). From this solution the expression for extraction rate can be written down in the dimensionless parameters k1A, k1, f and W = Dmt/a2 as: G t A D r C r t A a D K C k k e F k J J k J J r a n n n n n n n n n ( ) = ( ) = ¢ + + + ( ) ì í ï î ï ¢ ( ) - ( ) ( ) + ( ) ü ý ï � ï = - � å i m i m s s ¶ ¶ f a fa fa fa a a a a , ln [ ][ ] 1 2 1 1 1 1 1 0 1 1 0 2 W (12) where ±a1, ±a2, ±a3 ...are the roots of: [k1aJ1(a) + J0(a)][k1AfaY1(fa) 2 Y0(fa)] 2 [k1aY1(a) + Y0(a)][k1AfaJ1(fa) 2 J0(fa)] = 0 and F(a) = (k1A2f2a2 + 1)[k1aJ1(a) + J0(a)]2 2 (k1 2a2 + 1)[k1AfaJ1(fa) 2 J0(fa)]2 where Ji and Yi are the Bessel functions of the first and second kind, of order i.A formula for response time, the time when extraction rate reaches 90% of its steady state, accurate to ± 15% for 1.3 < f < 5 was derived as explained in ref. 21 using a symbolic algebra program:22 t a D k k k k k k 90 2 2 2 2 1 2 2 1 2 1 1 1 1 2 1 1 1 2 1 2 2 1 % ln ln ln ln = + ( ) - + + - - ( ) - - - ( ) ¢ + - ( ) ¢ ¢ + + m f f f f f f f f f f (13) If k1A and k1 are not significant response time: t a D b a D 90 2 2 2 2 2 1 1 0 32 % ln . = - + - - æ è ç ö ø ÷ » - ( ) m m f f f (14) The formula for steady state extraction rate is seen from eqn.(12) to be: G LD K C k k LC Nu D K K D f L Q D K e m ms s s o s mg ms g m ms = ¢ + + = + + æ è ç ö ø ÷ + 2 1 1 3 65 2 1 1 p f p p f ln . ln (15) Concentration in the carrier gas approaches equilibrium if: L Q K K Nu D K D > + æ è ç ö ø ÷ 1 6 1 2 . ln p f ms mg o s ms m (16) from eqn. (8), and under this condition: Gc = 0.9 CsQKms/Kmg ± 10% (17) Note Kms/Kmg is 1 for extraction from air, and is the analyte Henry’s constant for extraction from water.Experimental Apparatus and Reagents Samples were prepared with purified deionized water from a Barnstead water filtration system (Barnstead, Dubuque, IA, USA) and with 99% pure standards of benzene, toluene, ethylbenzene, hexane, trichloroethylene (TCE), ethanol and butanol (Supelco Canada, Mississauga, Canada). Aqueous standards were prepared at ppm m/m level: benzene 7 ml, toluene 9 ml, ethylbenzene 9 ml, TCE 3 ml and hexane 5 ml.Samples were prepared by diluting the concentrated aqueous standard one hundred fold, by pipetting 5 ml into a 500 ml volumetric flask of water, or by syringe addition of 0.4 ml to a 40 ml vial of water. In all aqueous samples the exact concentration for benzene, toluene, ethylbenzene, trichloroethylene, hexane, ethanol and butanol was 123, 156, 156, 88, 66, 98 and 95 ppb m/m. Samples were held in 40 ml amber vials with hole caps, Teflon lined septa and a 2.0 cm magnetic stir bar, filled completely unless otherwise stated.Samples were always agitated by magnetic stir bar at 1500 rpm, using a VWR stirrer/ hot plate (VWR Scientific, Mississauga, Canada). The average speed of water flowing past the membrane was estimated as 55 cm s21 using formula 3.41 in ref. 23. The temperature for all experiments was 25 ± 0.5 °C, and was monitored by thermocouple. Standards for GC calibration by syringe injection were prepared at 100 ppm m/m in reagent grade methanol, each analyte being stored in a separate 5 ml volumetric flask.A one point, one replicate calibration check was carried out each day using benzene. Whenever the response factor changed by more than 15% from the previous calibration curve, a new five point calibration curve was determined, with three replicates determined for each point. A five point calibration was always carried out for each new series of experiments, before or after, and at least once every two weeks in any case. The MESI system was described in detail previously.18 The gas line from the tank regulator was connected to the GC needle valve for carrier gas, and from there it was connected to the membrane module.Ultra-high-purity (UHP) grade nitrogen (Praxair, Waterloo, Canada) was used as the carrier gas at a flow rate of 2.2 ml min21, unless otherwise stated, as measured at the Analyst, December 1997, Vol. 122 1463membrane exit. UHP grade nitrogen, zero air and hydrogen (Praxair) were used as make-up, combustion and fuel gases, respectively for the FID with corresponding flow rates of 50, 300 and 30 ml min21, for all experiments.Flow rates were measured with a bubble meter at ambient temperature and pressure. The Varian 3500 GC instrument (Varian Canada, Mississauga, Canada) used had a split/splitless injector. For syringe or solid-phase micro-extraction (SPME) injections the column and gas line were disconnected from the membrane module and connected to the split/splitless injector.No split was used. Fig. 3 is a schematic diagram of the apparatus used for measuring the distribution constant of an analyte between the membrane and water. An SPME fibre assembly (Supelco Canada) was modified by replacing the outer needle with 5 cm long 8 gauge stainless steel tubing glued to the septum, and the fibre support rod was replaced with 7 cm long 31 gauge stainless steel tubing. The plastic hub from the original fibre support rod was glued to the top end of the 31 gauge tubing as a plastic plunger cap.At the bottom end of the 31 gauge tubing a 1 cm piece of membrane was forced on. The method development and procedure followed the standard methods of SPME as described in ref. 23. In this research, a 40 ml sample volume was chosen for the K value measurements, the aqueous samples were kept at 25 °C ± 0.5. The desorption conditions were 60 s at 200 °C. The piece of membrane was exposed to the stirred sample at the center of the vial for progressively longer times, 10 s up to 60 s,n 20 s up to 120 s, and finally 2.5, 3, 5 and 10 min.The GC oven program was 40 °C (hold for 0.2 min) to 200 °C (hold 5 min) at 20 °C min21. Calibration was by syringe injection, using the same oven program. For experiments to determine response time the membrane probe was connected directly to the GC detector via a 20 cm 3 0.32 mm id deactivated silica tubing (Supelco Canada).The membrane probe was initially exposed to air in a clean vial for 1 min to get a flat base line. The membrane probe was then placed in a vial full of solution and stirred. In a time period of 7 min, a rising signal and then a flat signal line were observed. Finally, the membrane probe was taken out of the solution and placed in a vial of pure stirred water. During this time a dropping signal line was observed. For the extraction rate measurements the MESI sorbent trapping time was 1 min and at steady state the peak areas were recorded.The peak area counts were calibrated by the syringe injection calibration explained before. Results and Discussion Tables 1 and 2 list the parameter values determined for the experiments in this paper. The distribution constants were determined by the SPME method, using the equation: K = nmvs/[vm(c0vs 2 nm)] (18) where nm is the amount extracted by the membrane piece used as an SPME fibre, vs is sample volume, vm is the membrane volume, c0 is the initial concentration of sample.The results show that nonpolar compounds have better affinities for the hydrophobic membrane. The distribution constants must also be measured for mobile phase. The distribution constants of analytes between membrane and nitrogen are the same as between the membrane and air.27 The distribution constants between membrane and air were measured by the SPME method and reported previously.18 Diffusion coefficient measurements have been performed and reported previously.18 Siloxane absorbs some water, but water is assumed to have a negligible effect on the diffusion coefficient in the polymer.A few studies suggest that water does not affect polymer diffusion coefficients of hydrophobic compounds,28 although this observation is not an established principle. An experiment using air extraction under varying humidity showed no significant effect of humidity on extraction rates. To determine the response times the membrane was connected directly to the FID and the signal versus time profile was recorded.Fig. 4 shows a typical FID signal versus time profile, Fig. 3 Schematic of assembly of SPME membrane. Table 2 Physico-chemical parameter values at 25 °C Kms water Dm/1026 cm2 s21 (determined Dwater †/ (determined Analyte experimentally) Kms * air 1025 cm2 s21 Dair ‡/cm2 s21 experimentally) DN2§ Benzene 136 485 1.09 0.093 2.12 0.38 Toluene 346 1872 0.95 0.085 1.59 — Ethylbenzene 847 3380 0.9 0.0755 1.09 — TCE 182 443 0.96 0.0875 1.81 — Hexane 126 224 0.9 0.0732 2.27 — Methanol 29 — 1.66 0.152 — — Isobutyl alcohol 52 — 0.93 0.088 — — * See ref. 18. † See ref. 24. ‡ See ref. 25. § See ref. 26. Table 1 Specifications of the experimental apparatus Parameter Value a, Membrane inner radius 0.0015 cm b, Membrane outer radius 0.00315 cm L, Length of membrane 4 cm Q, Flow rate of carrier gas 2.2 ml min21 Water sample temperature 25 °C Water sample speed 55 cm s21 1464 Analyst, December 1997, Vol. 122produced by benzene. Table 3 summarizes the response times determined for the different compounds with this MESI system. Good agreement is found between the theoretical predictions and the experimental results. Theory and experiment were also compared for extraction rates. The comparisons of extraction rates for the test compounds are listed in Table 3. An extraction versus time profile involves both response time and extraction rate. In Fig. 5, curve A is the model prediction of the benzene extraction versus time profile and curve B is the experimental result, showing good agreement. Figs. 6(a) and 6(b) show the chromatograms obtained from a benzene sample with and without stirring. The figures show a decrease in response time and an increase in extraction rate with stirring. In Fig. 6(b), the peaks declined after 12 min of extraction, attributed to depletion of analyte in the sample. Figs. 7(a) and 7(b) show the extraction versus time profiles for a benzene sample at 400 and 1200 rpm, respectively.As expected sample stirring enhanced extraction rate and reduced response time. The agreement between model and experiment is sufficient to conclude that the model of membrane extraction describes accurately the dominant natural processes in MESI. Figs. 8(a) and 8(b) illustrate the theoretical relationship between response time and mass transfer resistance at the outside membrane surface, i.e., t90% vs k1A, and the relationship between response time and mass transfer resistance at the inside membrane surface, i.e., t90% vs k1.Calculations with eqns. (13) and (15) show that response time and steady state extraction rate will be influenced by boundary effects at the membrane outside surface if: 1 0 015 Nu D K D o s ms m > . (19) i.e., if k1A > 0.03, for the aspect ratio of the membrane used in this study, f = 2.08. Likewise response time and steady state extraction rate will be influenced by boundary effects at the membrane inside surface if: 1 3 65 0 03 .. D f L Q K D g g m + > p (20) i.e., if k1 > 0.06, for the aspect ratio f = 2.08. Table 4 summarizes the range of parameter values possible for VOCs in air and water. Theory predicts that boundary effects can be significant for extraction from air or water, especially when compounds have high distribution constants between membrane and sample, based on the criteria of eqns.(19) and (20) and the parameter values in Table 4. The outside boundary effect, i.e., k1A, will likely be significant for extraction from air if Kms > 2000, and for extraction from water in most cases. One consequence of this characteristic is if the outside boundary effect is significant the system must have a constant sample flow speed for reproducibility. The membrane can be surrounded by a chamber with a stirrer to control sample flow speed. Such a mechanism to increase sample flow speed will also lower response time and minimize trapping time. General conclusions on the properties of the MESI system are not simple because diffusion coefficients of VOCs in polymers can vary by orders of magnitude for one compound in different polymers, or for different compounds in one polymer: for example, at room temperature benzene has D = 0.48 3 10212 cm2 s21 in polyvinyl acetate, and D = 1.3 3 1025 cm2 s21 in natural rubber; in polyacrylate, D = 6.2 3 10210 cm2 s21 for benzene and D = 1.0 31027 cm2 s21 for methanol.28 Diffusion in the polymer can also be sensitive to compounds absorbed in the polymer so extraction rate could vary irreproducibly for samples with high concentrations of organics, such as effluents or biological fluids.The diffusion coefficient is assumed constant for the analyte concentration ranges in this study. The boundary condition at the inside surface is measured by the value of k1.This parameter can be controlled by the experimental settings, and it varies little with experimental conditions such as temperature. A short membrane and high mobile phase flow rate will reduce the value of k1. In some cases experimental settings can be chosen to make boundary effects negligible on extraction rate and response time, i.e., to make k1A and k1 negligible. An experimental setup with negligible boundary effects was used to measure Dm using eqn. (14), by extraction from air with a high mobile phase flow rate.18 No simple relationship with respect to temperature is derived.Therefore, as an example, the relationships of response time and extraction rate with temperature were calculated for benzene, shown in Figs. 9(a) and (b). These calculations used the parameter values determined for this experimental system, and all of their individual relationships with T as described in Table 4. In this case the variation in response time with T is dominated by the variation in the polymer diffusion coefficient, Dm.The extraction rate variation is dominated by variation in the value of Kms Dm . The overall change of extraction rate depends on the Fig. 4 FID signal time profiles of benzene in response time measurement. Table 3 Comparison of theoretical and experimental results. t0–90% represents the time of signal climbing and from 0% to 90% of steady-state intensity, and t100–10% dropping from 100% to 10%. RSD are based on 3 replicates % Difference % Difference between between Response time†/s theoretical and theoretical and experimental Ge § (RSD)/ experimental Analyte t90% */s t0–90% (% RSD) t100–10% (% RSD) t0–90% Ge ‡/ng s21 ng s21 Ge Benzene 136 113 (3.9) 106 (6.2) 16 0.55 0.55 (3.3) 0 Toluene 347 269 (5.8) 308 (7.8) 22 0.73 0.72 (4.2) 21.3 Ethylbenzene 701 641 (5.9) 856 (6.5) 8.5 0.91 0.95 (3.8) 24.4 TCE 163 131 (3.7) 157 (2.4) 19 0.45 0.41 (3.4) 8.9 Hexane 113 120 (2.1) 296 (6.7) 26.9 0.32 0.28 (3.7) 12 * Determined by eqn. 13. † Determined experimentally. ‡ Determined by eqn. 15. § Determined experimentally. Analyst, December 1997, Vol. 122 1465change of KmsDm. Fig. 10 gives the experimental results showing variation of the extraction amount of benzene from aqueous solution under different temperatures. It can be seen that with the increase of temperature the amount of benzene extracted increased, hence the extraction rate increased. Fig. 5 Extraction tie profiles of benzene without trapping.A, Model prediction and B, experimental result. Fig. 6 Extraction time profile of benzene, (a) from a relatively static sample and (b) with stirring. Fig. 7 Extraction time profile of benzene at different stirring speeds (rpm): (a) 400 and (b) 1200. Fig. 8 Relationship between response time and parameter of mass transfer at the membrane surfaces, (a) outside surface k1A, and (b) inside surface k1. The solid line is calculated with the exact model [eqn.(12)], the dashed line with the approximation formula [eqn. (13)]. Fig. 9 Relationship of (a) response time and (b) extraction rate with temperature, as calculated by the relationships described in Table 4. 1466 Analyst, December 1997, Vol. 122Experiments also revealed that a small headspace can increase the extraction rate from aqueous samples. Comparing the extraction amounts without headspace and with 0.5 ml headspace shows that the total extraction amount for the first 20 min increased by 10% to 20% depending on the compound.With headspace, a large number of small gas bubbles adhere to the membrane surface. A compressed headspace allows carrier gas to penetrate the membrane wall and form bubbles on the membrane surface. Although the analyte still needs to diffuse through water to the gas bubbles, mass transfer was enhanced because the small gas bubbles had a large surface area, the molecules of analyte could easily diffuse through those gas bubbles and, in general, compounds have larger distribution constants between membrane–air than between membrane– water.With a headspace concentration in the aqueous solution drops somewhat because analytes distribute into the headspace, but overall the extraction rate is increased if the headspace is small. Some implications for design and operation of the MESI system can be derived from the theory. Quantitative analysis by the MESI method depends on calibration, and on control of significant factors to ensure reproducibility.The simplest situation in which to calibrate the steady state extraction rate is when the carrier gas reaches equilibrium before exiting the membrane. This situation will occur under the conditions specified by eqn. (16). If constraints do not allow the carrier gas to reach equilibrium, then extraction rate will be sensitive to the several factors listed in eqn. (15), and controlling the significant factors or compensating for their variations will be more complicated.The diffusion coefficient in the polymer will be especially difficult to evaluate, since this parameter is sensitive to so many factors as explained above. Therefore, when designing a MESI system for a new application a membrane with a low value of lnf/KmgDm is desirable since it can make practical calibration by eqn. (17), for the simplest and most reliable calibration. Membrane types could be compared according to their value of lnf/KmgDm.Likewise when choosing membrane length and carrier gas flow rate, a value of L/Q that allows calibration by eqn. (17) would simplify. Increasing L/Q does increase the response time, but by at most a factor of 2 for aspect ratio f = 2.08 by eqn. (13). Increasing carrier gas flow rate can also increase extraction rate, so a shorter trapping time, and thus a shorter effective response time, would obtain the same response. However, a high carrier gas flow rate results in analyte breakthrough in the sorbent trap. Membrane choice also depends on characteristics such as selectivity, and mechanical or chemical durability.Besides the factors described above, if MESI is applied directly to an effluent stream or biological system rich in organic compounds, significant interference could cause high background noise or affect the parameters that govern response time and extraction rate in a non-reproducible manner. Humic or other materials in a sample could foul a membrane.In that situation, a headspace approach is more suitable.17 Table 4 Range of parameter values at 300 K and how each varies with T over ambient temperatures in different phases of the MESI system. All symbols are defined in the Glossary Phase n/cm2 s21 D, for VOCs/cm2 s21 Kms, for VOCs (dimensionless) Other Air 0.15, Å T1.8 0.05–0.16, Å T1.8 (ref. 19) 0.1–115 000, Å e2DH/RT NuoDs = 0.015 + 0.099(ub)1/2 (ref. 19) (refs. 29, 30) to 0.045 + 0.39(ub)1/2 in g cm s Example values of DH/Rg:20 units.An analyte at constant 5160 for benzene mole fraction will vary 4700 for toluene concentration [ML23] with 1/T. Water n(T) Å 0.1083– 0.5–1.5 3 1025, Å T/n(T) 0.1–7000, refs. 29, 31 NuoDs = 0.00057(ub)1/2 to 0.000332T (ref. 19) Å e2DH/RT or 0.0012(ub)1/2 in g cm s (ref. 19) Å 1/solubility (T). units, Å n(T)2/3/T5/6. Example value of DH/Rg, estimated from water solubility tables32: 525 for benzene Polymer 10213–1025, Å e2EdRgT D is sensitive to compounds membrane (refs. 26, 28) absorbed in the polymer. Example values of Ed/Rg: (PDMS will absorb 720 mg cm23 7700 benzene in natural rubber, water33 with D of water = 2900 toluene in polystyrene, 7 3 1025 cm2 s21 28) e.g., in 300 benzene in the membrane polyvinylacetate: compared to used in this study, estimated 1 ppm, 1000 ppm of allyl from measurements reported chloride changes its D 60%; before.10 water in polymer greatly accelerates the sorption rate of acetone.28 Gas mobile phase 0.1–1.0, Å T1.8 0.022–0.38, Å T1.8 Partition coefficients between At a constant mass flow rate the (ref. 19) (refs. 19, 26) polymer and any gas are volumetric flow rate Q will practically the same as between vary with T. (If tank and polymer and air.27 needle valve are at ambient temp T and the valve is not adjusted mass flow rate will not vary significantly with T.34) Fig. 10 Temperature effect on extraction rate of benzene solution. Analyst, December 1997, Vol. 122 1467This project has been financially supported by the Dow Chemical Company and the Natural Sciences and Engineering Research Council of Canada. Appendix Glossary a Radius from the membrane axis to the membrane inner surface (cm; see Fig. 1). b Radius from the membrane axis to the membrane outer surface (cm; see Fig. 1). f Ratio of average concentration to exit concentration in mobile phase (dimensionless). ho,hi Mass transfer coefficients at the membrane outer and inner surfaces, respectively (kg s21 m22).k1A KmsDm/bho = 2KmsDm/NuoDs (dimensionless). This parameter is a measure of the resistance to mass transfer at the membrane outer surface. 2 1 D K Nu D Lf Q m g i g + æ è ç ö ø ÷ p k1 (dimensionless). This parameter is a measure of the resistance to mass transfer at the membrane inner surface. r Radius since the membrane axis (cm). t Time from the start of extraction (s). t90% Response time, t, at which extraction rate reaches 90% of its steady state value (s).This is also the time from any change in Cs until 90% of the resulting change in steady state extraction rate. u Sample fluid velocity (cm s21). x Distance along the length of the membrane (cm). x = 0 at the point where mobile phase enters the membrane, and x = L at the exit. Ai Membrane inner surface area (cm2), Ai = 2paL. B(r) A function of r in the expression for C(r,x), not given explicitly because it cancels out. C or C(r,t) Analyte concentration in the membrane, a function of r and t (mg l21).Cs Analyte concentration in the bulk sample (mg l21). Cg(x) Analyte bulk concentration in the mobile phase as a function of x (mg l21). C Average bulk concentration in the mobile phase (mg l21). D Diffusion coefficient (cm2 s21). Ds Analyte’s diffusion coefficient in the sample (cm2 s21). Dm Analyte’s diffusion coefficient in the membrane (cm2 s21). Dg Analyte’s diffusion coefficient in the mobile phase (cm2 s21). Ed Analyte’s apparent activation energy for diffusion in a polymer.G(t) Overall extraction rate of the MESI membrane and mobile phase (ng s21). Ge Overall steady-state extraction rate (ng s21). DH Analyte’s heat of sorption from sample to membrane. Ji,Yi Bessel functions of the first and second kind, of order i. Kms Analyte distribution constant between membrane and sample, concentration in membrane divided by concentration in sample at the interface (dimensionless; mg l21: mg l21).Kmg Analyte distribution constant between membrane and mobile phase, concentration in membrane divided by concentration in mobile phase (dimensionless; mg l21: mg l21). L Length of the membrane (cm). Nuo,Nui Nusselt numbers at the outer and inner membrane surfaces, respectively (dimensionless). P0 Pole in Laplace transform. Rg Gas constant. Q Mobile phase volumetric flow rate (ml min21). Red Reynold’s number of a fluid, defined as Red = ud/ n where d is diameter of the membrane outer or inner surface depending on context (dimensionless). Sc Schmidt number of a fluid, defined as Sc = n/Ds or n/Dg (dimensionless). T Absolute temperature in degrees Kelvin. n Fluid kinematic viscosity (cm2 s21). f Ratio of membrane’s outer to inner radius, = b/a (dimensionless). W = Dmt/a2, dimensionless time parameter (dimensionless). References 1 Lotiaho, T., Lauritsen, F. R., Choudhury, T. K., Cooks, R. G., and Tsao, G. T., Anal. Chem., 1991, 63, 875A. 2 Lauritsen, F. R., Int. J. Mass Spectrom. Ion Proc., 1990, 95, 259. 3 Carlsen, H. N., Jørgensen, L., and Degn, H., Appl. Microbiol. Biotechnol., 1991, 35, 124. 4 LapPack, M. A., Tou, J. C., and Enke, C. G., Anal. Chem., 1990, 62, 1265. 5 Virkki, V. T., Ketola, R. A., Ojala, M., Kotiaho, T., Komppa, V., Grove, A., and Faccchetti, S., Anal. Chem., 1995, 67, 1421. 6 J�onsson, J. Å., and Mathiasson, L., Trends Anal. Chem., 1992, 11, 106. 7 Lindegråd, B., J�onsson, J. Å., and Mathiasson, L., J. Chromatogr., 1992, 573, 191. 8 Thordarson, E., P�almarsd�ottir, S., Mathiasson, L., and J�onsson, J. Å., Anal. Chem., 1996, 68, 2559. 9 Yang, M. J., Harms, S., Luo, Y. Z., and Pawliszyn, J., Anal. Chem., 1994, 66, 1339. 10 Yang, M. J., Luo, Y. Z., and Pawliszyn, J., Chemtech., 1994, 24, 31. 11 Pratt, K. F., and Pawliszyn, J., Anal. Chem., 1992, 64, 2101. 12 Pratt, K. F., and Pawliszyn, J., Anal. Chem., 1992, 64, 2107. 13 Mitra, S., Zhang, L., Zhu, N., and Guo, X., J. Chromatogr., 1996, 736, 165. 14 Burger, B. V., Burger, W. J. G., and Burger, I., J. High Resolut. Chromatogr., 1996, 19, 571. 15 Yang, M. J., and Pawliszyn, J., Anal. Chem., 1993, 65, 2538. 16 Ortner, E. K., and Rohwer, E. R., J. High Resolut. Chromatogr., 1996, 19, 336. 17 Yang, M. J., Adams, M., and Pawliszyn, J., Anal. Chem., 1996, 68, 2782. 18 Luo, Y. Z., Adams, M., and Pawliszyn, J., Anal. Chem., in the press. 19 Transfer Processes, ed. Edwards, K., Dennry, V. E., and Mills, A. F., Hemisphere Publishing, New York, 2nd edn., 1978. 20 Martos, P. A., and Pawliszyn, J., Anal. Chem., 1996, 69, 206. 21 Carslaw, H. S., and Jaeger, J. C., Conduction of Heat in Solids, Clarendon Press, Oxford, 2nd edn., 1986, section 13.4. 22 Maple V Release 4, Waterloo Maple, Waterloo, Canada, 1996. 23 Pawliszyn, J., SPME, John Wiley and Sons, New York, 1997. 24 Hayduk, W., and Laudie, H., AIChE J., 1974, 20, 611. 25 Lugg, G. A., Anal. Chem., 1968, 40, 1072. 26 Jost, W., Diffusion in Solids, Liquids, Gases, Academic Press, New York, 1960, p. 412. 27 Berezkin, V. G., Korolev, A. A., and Maluykova, I. V., Proc. Int. Symp. Capillary Chromatogr., 18th, 1996, pp. 396–404. 28 Diffusion in Polymers, ed. Crank, J., and Park, G. S., Academic Press, London, 1968. 29 Pan, L., Adams, M., and Pawliszyn, J., Anal. Chem., 1995, 67, 4396. 1468 Analyst, December 1997, Vol. 12230 Zhang, Z., and Pawliszyn, J., J. High Resolut. Chromatogr., 1996, 2, 155. 31 Langenfeld, J., Hawthorne, S., and Miller, D., Anal. Chem., 1996, 68, 144. 32 Freier, R. K., Aqueous Solutions, Walter de Gruyter, Berlin, 1976. 33 Klunder, G. L., and Russo, R. E., Appl. Spectrosc., 1995, 49, 379. 34 McGraw-Hill Encyclopedia of Science and Technology, McGraw- Hill, 1997, vol. 12, pp. 124–126. Paper 7/06441A Received September 3, 1997 Accepted November 4, 1997 Analyst, December 1997, Vol

ISSN:0003-2654    

DOI:10.1039/a706441a

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Determination of Formaldehyde by Conversion to Hexahydrooxazolo[3,4- a]pyridine in a Denuder Tube With Recovery by Thermal Desorption, and Analysis by Gas Chromatography–Mass Spectrometry† C. L. Paul Thomasa, Charlotte D. McGillb and Robert Towillc a Department of Instrumentation and Analytical Science, UMIST, PO Box 88, Manchester, UK M60 1QD b Department of Chemistry, University of Reading, Whiteknights, PO Box 224, Reading, UK RG6 6AD c Jones Chromatography, Hengoed, Mid Glamorgan, UK CF8 8AU The implications for formaldehyde analysis arising from a recent study where the inter-laboratory relative standard deviation for formaldehyde [CAS Registry number: 50-00-0] monitoring was reported to fall in the range 20 to 60% are discussed.A thermal desorption approach is proposed as a possible way to address the requirement for less intensive sample processing in formaldehyde analysis; as well as meeting other important criteria such as reduced sample handling and increased overall sensitivity.A denuder tube coated with 2-hydroxymethylpiperidine [CAS Registry number: 3433-37-2] and connected to an adsorbent sampler packed with Tenax TA was used to derivatise formaldehyde vapour. The volatile derivative, hexahydrooxazolo[3,4-a]pyridine [CAS Registry number: 274-50-0], was retained on the adsorbent trap. Analysis of the derivative was accomplished using a thermal desorption technique coupled to gas chromatography–mass spectrometry.To test the approach a test atmosphere generator based on a permeation tube containing a-polyoxymethylene was constructed. A dilution chamber heated to 80 °C was used to prevent repolymerisation of the released formaldehyde. Gravimetric measurements indicated that formaldehyde was generated at a rate of 1.9 3 1024 g h21 (95% confidence limits = ± 1 3 1026 h21) and atmospheres of formaldehyde at concentrations of 5.92, 3.35 and 1.79 mg m23, with a water concentration less than 17 mg m23 at 20 °C and 1.013 kPa were validated using NIOSH Method 3500 for formaldehyde determination. Formaldehyde masses in the range 0.1 to 16 mg were sampled using the system described.The contents of the adsorbent trap were then analysed. The data obtained supported the hypothesis that formaldehyde derivatives could be recovered in an analytically useful way by thermal desorption. Mass spectrometric data confirmed previous assignments for hexahydrooxazolo[3,4-a]pyridine and that thermal decomposition did not occur to a detectable extent during desorption. The relationships between mass sampled and instrument response were linear up to the point of breakthrough, with limits of detection in the range 0.03 to 0.51 mg m23 formaldehyde in air.Keywords: Formaldehyde; denuder tubes; adsorbent trap; Tenax TA; 2-hydroxymethylpiperidine; hexahydrooxazolo[3,4-a]pyridine; test atmosphere generation; thermal desorption gas chromatography–mass spectrometry; sampling Aldehydes in general, and formaldehyde in particular, are acknowledged to be important volatile organic compounds which are present in many industrial operations.Exposure to formaldehyde at low concentrations results in acute and chronic health effects. The odour detection threshold has been estimated to fall in the range 0.06 to 1.2 mg m23 and the onset of irritation to eyes and upper respiratory tract takes place in the range 0.01 to 3.1 mg m23.1 Formaldehyde causes cancer in rats and is a suspected human carcinogen. The exposure limits in the United Kingdom are set at 2.5 mg m23 for short-term exposure over 15 min, as well as long-term time-weighted average exposure over 8 h.2 Formaldehyde is also a precursor of photochemical smog, and in the presence of water produces formic acid and methanol.The potential damage to artefacts in museums, art galleries and other archives arising from formaldehyde pollution requires a reappraisal of the materials used currently for the construction of display and storage facilities.Exposure to formaldehyde also occurs in domestic dwellings, where maximum indoor concentrations have been estimated to reach levels of 0.2 to 4 mg m23,3,4 while mean concentrations lie in the range 0.02 to 0.06 mg m23,5 an unfavourable comparison with the maximum exposure limits mentioned above. It is not surprising, therefore, that much effort has been devoted to developing and improving methods for the determination of formaldehyde.A recent paper provides a timely and authoritative comparison of the techniques currently available,6 although the use of direct reading fuel cell monitors was not included in this work. Extensive reviews of formaldehyde determination have also been published.7,8 Detection limits for formaldehyde determination are generally quoted as falling in the range 0.24 mg m23–0.25 mg m23, but perhaps as significant are the values of measurement uncertainty associated with formaldehyde determination.Under controlled laboratory conditions inter-laboratory values for relative standard deviation (RSD) have been reported to fall in the range 20–60%.6 Such variance in the sample collection procedures means that the resources required to run monitoring programmes, where sample heterogeneity would only serve to increase the uncertainty in the measurement still further, are likely to be prohibitively large. Table 1 illustrates this point with the numbers of samples needed to meet a range of data quality objectives (DQO) when the overall RSD, including measurement and sampling contributions to the uncertainty, is 50%.Even modest DQO require large numbers of independent samples obtained from a random sampling plan. Given the recognised need to monitor formaldehyde, sampling methods are required that either: (i) reduce the complexity and cost of the sampling systems involved and thus enable higher specifications of DQO to be set; (ii) reduce the experimental uncertainties involved and so reduce the numbers † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997.Analyst, December 1997, Vol. 122 (1471–1476) 1471of independent samples needed to yield high specification DQO; (iii) lower the limits of detection for formaldehyde determination to enable more complete monitoring of formaldehyde in the outdoor environment; (iv) or a combination of all three requirements.As well as these considerations there is the need to acknowledge that the materials used in the construction of laboratories may themselves be a significant source of interference in formaldehyde analysis; concentrations of 0.61 mg m23 of formaldehyde in air have been reported in laboratory atmospheres.10 Consequently, methods that reduce the exposure of the sample to laboratories and other potential sources of interference may also be seen to offer benefits.This work seeks to address the issues outlined above by testing the hypothesis that volatile derivatives of formaldehyde generated during sampling by a denuder tube may be trapped on an adsorbent bed and that the derivatives may be recovered using a thermal desorption technique for analysis by gas chromatography. Derivatisation approaches are used often in gas chromatography, to either stabilise analytes, enhance the chromatographic behaviour of an analyte or impart characteristics that enable more selective and sensitive detection to be used.11 However the direct recovery of such a derivative by thermal desorption has not been previously reported.Such an approach appears attractive because: (i) it removes the need for solvent elution of the derivatised formaldehyde thereby reducing the cost and complexity of the sampling procedures; (ii) it is amenable to automation; (iii) it isolates the sampled formaldehyde from the laboratory; (iv) it offers enhanced sampling efficiency in that all the derivatised material will be transferred to the gas chromatograph instead of an aliquot of an extract; (v) and offers the opportunity to use cryogenic focusing to attain lower limits of detection.Denuder tubes in the analysis of trace gases were originally conceived as a method for removing interfering species from a gas sample12 and later as a means for isolating analytes from the sample,13,14 including formaldehyde.15,16 Other approaches involved heating denuder tubes to produce thermochemical transformations in some of the sampled species, reported first in 1978.Thus a denuder tube used to derivatise the analyte to produce a volatile product, subsequently retained in an adsorbent trap, may be seen to be a development of this idea. The theory and uses of denuder tubes have been reviewed previously.13 Thus this work used a denuder tube coated with 2-hydroxymethylpiperidine (2-HMP) connected to a Tenax TA adsorbent trap.Formaldehyde diffused to the walls of the denuder tube and reacted yielding hexahydrooxazolo[3,4-a]pyridine which desorbed and was swept into the adsorbent trap. 2-HMP was chosen as it is the basis of established and accepted methodologies for the determination of formaldehyde and produces a volatile derivative.17,18 An alternative approach would have been to have impregnated the Tenax TA with the 2-HMP directly; which would have been analogous to the approach of commercially available formaldehyde sampling devices from Supelco (Orbo tubes, Fancy, Poole, Dorset, UK) which are based on an XAD Resin and use solvent elution.The advantages of a denuder tube approach over an impregnated adsorbent were envisaged to be: (i) the prevention of experimental artefacts arising from the release of formaldehyde through thermal degradation of the adsorbent as it is thermally desorbed; and (ii) the consequences of competitive adsorption between the 2-HMP coating and hexahydrooxazolo[3,4-a]pyridine were not known.Experimental Generation of Standard Formaldehyde Atmospheres Formaldehyde vapour is stable as a monomer between 80 and 150 °C. Below 80 °C polymerisation occurs, and above 150 °C decomposition takes place.19 Further work has noted discrepancies between gravimetric calibration of formaldehyde permeation tube sources based upon a-polyoxymethylene and observed concentrations of monomeric formaldehyde.20 Suggested reasons for this were: (i) other degradation products, as well as formaldehyde, being released from the permeation tube; (ii) thermal decomposition of formaldehyde; and (iii), impurities in the a-polyoxymethylene.The vapour generator was designed therefore to maintain the released vapours from a permeation tube at 80 °C until dilution had been achieved; to quench any possible repolymerisation due to reaction with coreleased impurities.Further, all the internal surfaces of the test atmosphere generator were made from either PTFE or silanised borosilicate glass. Once diluted the test atmosphere was allowed to cool to room temperature before it was passed into a sampling manifold. The permeation tubes were made from 6.25 mm od PTFE tubing filled with a-polyoxymethylene (Merck, Poole, Dorset, UK) and capped with silanised glass plugs. It was calibrated by a series of mass-loss measurements made over a period of approximately 2500 h.The data showed that the release of material from the permeation tube was constant throughout this time at a rate of 1.9 3 1024 g h21 ± 1 3 1026 g h21 (95% confidence limits for 27 mass loss measurements). As mass-loss measurements involved disruption of the vapour source a stabilisation period of 1 week was invoked after each mass loss determination before experimentation could recommence. The diluent gas used was bottled air passed through purification media (Phase Separations, London, UK) before use.All flow controls were calibrated and the estimated concentrations from the gravimetric data were compared with observed concentrations obtained from NIOSH Method 3500,21 the chromotropic acid method. Three standard atmospheres were used in this work, 5.92, 3.35 and 1.79 mg m23, with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa (0.1% relative humidity, RH). At each of these concentrations seven different sample masses were taken across the range 10–140 mg. The concentrations of the test atmosphere and sampled masses determined by NIOSH Method 3500 were correlated against the gravimetric data.The results of this validation are presented in Table 2. The inference from this experiment was that under the operating conditions described above gravimetric data could be used to estimate test atmosphere concentrations to within the limits of uncertainty of NIOSH Method 3500.Studies undertaken for extended sampling periods and volumes showed an absence of sampling artefacts in the data; demonstrating a lack of surface activity in the test atmosphere generator and sampling train. Fig. 1 is a schematic diagram of the test atmosphere generator. Preparation of the Denuder Tube A glass tube, 15.8 cm long, 0.6 cm od and 0.3 cm id was coated with 3.2 mg of 2-HMP (Aldrich, Poole, Dorset, UK). The inlet zone was 1 cm long. Two 10 ml aliquots of 2-HMP in methanol (Merck) at a concentration of 160 mg cm23 were applied to the Table 1 Number of samples required to achieve a range of data quality objectives for a method with an overall RSD of 50%.Calculations based on methodology described in ref. 9. % Maximum acceptable % Confidence error in result (same units as RSD) 5 10 20 80 166 43 12 90 273 70 19 95 387 99 27 99 668 170 46 1472 Analyst, December 1997, Vol. 122internal surfaces of the denuder tube and the tube was rotated and tilted while the methanol evaporated leaving a coating of 2-HMP on the interior surfaces of the tube.The denuder tube was then immediately sealed with PTFE plugs. Preparation of the Adsorbent Bed for Use in a Programmable Temperature Vaporiser (PTV) Injector The denuder tubes described above were connected to an injector liner (Optic, Atas, Cambridge, UK) packed with 0.089 g of Tenax TA (Phase Separations). The Tenax TA trap was conditioned for 3 h before use by placing it in the PTV injector (an Optic unit) and then heating it to 350 °C while passing a flow of helium through it at approximately 20 cm3 min21.The sampling units were sealed with PTFE end caps immediately after they were prepared. Determination of Sampled Mass of Formaldehyde by Recovery of Oxazolidine The sampler was connected to the sampling manifold using a silanised glass connector and a sampling pump, set to sample at 15 cm3 min21. Three concentrations of 1.79, 3.35 and 5.92 mg m23 formaldehyde in air at ambient pressure and temperature, with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa were sampled to give a range of sampled masses of formaldehyde at different concentrations. The Tenax packed injection liner was then placed in the PTV injector and the trapped materials were thermally eluted onto a gas chromatography column.Detection was by electron-ionisation quadrupole mass spectrometry. The analysis parameters are summarised in Table 3, and the results of these runs are in Fig. 2.Discussion Generation of Formaldehyde Test Atmospheres The results in Table 2 show that the test atmosphere generator produced stable and reproducible standards of formaldehyde vapour, and that gravimetric data obtained from the permeation tubes could be used to calibrate formaldehyde monitoring methodologies. It was found that a stabilisation time of 1 week was needed after any parameter affecting the concentration of formaldehyde in the unit had been changed. The stabilisation period ensured that residual surface activity within the system attained equilibrium with the test atmospheres and did not interfere with subsequent experiments.The stabilisation period was especially important for low concentration studies. At- Table 2 Summary of the correlation of the formaldehyde concentrations, obtained from the gravimetric data, against the observed formaldehyde concentrations using the NIOSH Method 3500* Formaldehyde concentration/ mg m23 Gradient Intercept/mg r2 1.79 1.223 ± 0.11 23.5 ± 13.8 0.96 3.35 0.974 ± 0.05 28.7 ± 5.9 0.99 5.92 1.011 ± 0.04 22.0 ± 5.4 0.99 * 27 gravimetric measurements were made over a period of 2500 h.Each gravimetric determination was based on five separate measurements. These data were linearly correlated to time to give a release rate of 1.49 3 1024 g h21 ± 1.1026 g h21 at 95% confidence limits. Test atmospheres at the above concentrations were generated and determined by the NIOSH 3500 Method for seven sampled masses.Each test atmosphere concentration determination was based on five independent measurements using the NIOSH 3500 Method. For the two methods to agree the gradient of the regression should be 1 with an intercept of zero.22 95% confidence limits are given, and it can be seen that within the limits of experimental error the concentrations of the test atmosphere determined from the gravimetric approach and the NIOSH 3500 Method agree.Fig. 1 Schematic diagram of the test atmosphere used to produce the formaldehyde standards for this study: 1, purified air supply; 2, flow controllers; 3, flow meters; 4, permeation tube holder made from 15 cm of 1.3 cm od glass tubing; 5, permeation tube; 6, dilution chamber made from 20 cm long by 2.5 cm od glass tubing fused to the permeation tube holder; 7, sampling manifold; 8, charcoal trap; 9, denuder tube coated with 2-HMP connected to the inlet of a Tenax TA filled adsorbent trap; 10, sampling pump; 11, exhaust.Table 3 Summary of the analysis parameters for the determination of hexahydrooxazolo[3,4-a]pyridine recovered from the Tenax TA adsorption traps by thermal desorption Instrument variable Level Sampler type Adsorbent trap fitted with a denuder tube. Adsorbent holder is an Optic compatible thermal desorption tube. Adsorbent 8.9 mg of Tenax TA Denuder tube dimensions 15 cm long, 0.6 cm od and 0.3 cm id Denuder tube coating 3.2 mg of 2-hydroxymethylpiperidine Instrument type Fisons 8035 gas chromatograph fitted with a programmable temperature vaporising injection unit (Optic) in conjunction with a Fisons Trio 1000 Quadrupole mass spectrometer.Desorption temperature Start temperature: 50 °C Initial time: 1 min Temperature ramp: 960 °C min21 Final temperature: 200 °C Carrier gas Helium Desorption gas flow 102.3 cm3 min21 Split flow 100 cm3 min21 Gas chromatography column and J&W DB-Wax PEG 30 m, dimensions 0.32 mm id, and film thickness 0.25 mm.Carrier gas flow 2.3 ± 0.3 cm3 min21 Gas chromatography temperature Start temperature: 35 °C programme Initial time: 0 min Ramp rate 1: 50 °C min21 Temp 2: 110 °C Time at Temperature 2: 1 min Ramp rate 2: 15 °C min21 Final temperature: 190 °C Final time: 1 min Detector Mass spectrometer settings were set by the instrument under ‘auto-tune’ Analyst, December 1997, Vol. 122 1473tempts to reduce the stabilisation time resulted in increases in the variances of the data obtained.The importance of a heated permeation tube holder and dilution chamber was demonstrated by alternatively reducing the heating and the dilution flow. In both cases repolymerisation of formaldehyde was observed as a white coating on the interior surfaces of the test atmosphere generator. Finally, if the temperature of the permeation tube was taken too high, greater than 130 °C, the concentration of released formaldehyde vapour increased beyond the effective capacity of the dilution chamber, resulting in repolymerisation of formaldehyde throughout the test atmosphere generator.Mass Spectra Table 4 presents the mass spectrometric data obtained for hexahydrooxazolo[3,4-a]pyridine along with tentative assignments. The data confirm those reported previously and the assignment of the hexahydrooxazolo[3,4-a]pyridine peak.16 No evidence was found of thermal decomposition of hexahydrooxazolo[ 3,4-a]pyridine caused by thermal desorption.Analysis of Recovered Hexahydrooxazolo[3,4-a]pyridine The data in Fig. 2 are consistent with the behaviour of a volatile compound in an adsorbent sampling bed, used in the active mode. Initially the recovery of derivative increases linearly with sampled mass, see Fig. 3. As breakthrough starts the plot of peak area versus sampled mass becomes non-linear until the whole adsorbent bed reaches equilibrium at complete breakthrough.The equilibrium capacity of the adsorbent was observed to depend upon the concentration of the formaldehyde vapour; which is what would be expected for Type I physical adsorption.23 In which case the mass trapped, Mt, is related to the concentration of the sampled vapour, [i], by the expression, M K i K i K j j j n t = + + = å [ ] [ ] [ ] 1 1 (1) where K is a constant analogous to the adsorption constant, and Kj and [j] refer to the constants and concentrations for the other adsorbates present in the sampled air. In pure diluent air, as found in the test atmosphere generator, and at low concentrations of formaldehyde, eqn. 1 may be approximated to, Mt = K[i] (2) which indicates that the capacity of an adsorption trap will decrease in direct proportion to the concentration of the analyte. Analysis of the observed trap capacities showed a linear relationship between the formaldehyde concentration and the mass of material recovered by thermal desorption; mass Fig. 2 Graph showing the relationship of the peak areas obtained from the thermally desorbed hexahydrooxazolo[3,4-a]pyridine with increasing mass of sampled formaldehyde at concentrations of 1.79, 3.35 and 5.92 mg m23 formaldehyde in air at ambient temperature and pressure with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa. Table 4 Summary of the mass spectrometric data, confirming the identity of the hexahydrooxazolo[3,4-a]pyridine peak Reactive ion Relative mass/u abundance (%) Assignment 127 30 Molecular ion 126 52 Loss of H from molecular ion 98 21 97 100 Loss of H2CO from molecular ion 82 6 70 15 69 70 Loss of NCH2OCH2 from molecular ion 68 18 56 18 55 21 Loss of CH2 from reactive ion mass 69 54 9 41 58 Loss of CH2 from reactive ion mass 55 Fig. 3 Graphs showing the linear relationships, before breakthrough in Fig. 2, 95% confidence limits shown. 1474 Analyst, December 1997, Vol. 122ratio = 0.097 3[H2CO] + 9 31026.The mass ratio is given by the expression Mt Ma , where Ma is the mass of the adsorbent used, in this case 8.9 mg, while the units for the constant K are m3 g21. The correlation coefficient was 0.998. It is likely that the sampling volumes associated with this study could be increased significantly by increasing the mass of adsorbent used in the trap; the small adsorbent masses were used to ensure rapid heating of the trap during desorption. Doing this would increase the sensitivity of the technique as well as lower the detection limits, perhaps enough to enable background concentrations of formaldehyde in unpolluted air to be monitored.To do this, cryogenic focusing of the thermally desorbed products would need to be incorporated and Fig. 4 is a comparison of the chromatography obtained with and without cryogenic focusing. However, as the capacity of the trap diminishes with decreasing analyte concentration a limit will be reached beyond which increasing the trap size will not yield significant benefits in terms of sensitivity and lowered limits of detection.The cryogenic focusing was used to investigate the hexahydrooxazolo[ 3,4-a]pyridine residues still within the denuder tube after sampling and to establish if the denuder tube might work as a stand alone sampling unit. In this case the denuder tube was desorbed directly into a gas chromatograph system. Table 5 summarises the experimental parameters used.The data obtained from this study for concentrations of formaldehyde below 5.92 g m23 formaldehyde in air at ambient pressure and temperature, with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa indicated trace residues only remaining within the denuder tube. At these concentrations the data were not reproducible and there was no clear correlation between the mass of formaldehyde sampled and the amount of recovered material. At 5.92 g m23 a linear relationship between peak area and sampled mass was observed and linear regression analysis gave the relationship, Peak area = 43.3 3 mass of formaldehyde sampled + 0.58.The units for peak area were mV min21 and the units for mass were mg. The correlation coefficient was 0.992 and the RSD values were in the range 10–30%, see Fig. 5. These findings supported the assumption that the volatile derivative was efficiently transferred from the denuder tube to the adsorbent trap and confirmed that the equilibrium capacity of the denuder tube was not large enough at concentrations below 5.92 g m23 for it to be used as the basis of an effective stand alone device.Conclusions These data show that it is possible to recover volatile derivatives for analysis by gas chromatography by thermal desorption. This is a previously untried approach. In this study hexahydrooxazolo[ 3,4-a]pyridine has been produced from sampled formaldehyde with a denuder tube coated with 2-HMP and trapped on a Tenax adsorbent bed used in conjunction with a thermal desorption technique.The system investigated determined formaldehyde at concentrations below current exposure limits, and the estimated limits of detection were respectively 0.51, 0.03 and 0.05 mg per sample, for the formaldehyde concentrations of 5.92, 3.35 and 1.79 mg m23 formaldehyde in air at ambient pressure and temperature, with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa.Which may be compared to the 1 mg per sample stated by NIOSH Method 2541,17 based on 2-HMP. The variations in these values may be attributed to variations in the ion source of the mass spectrometer as these runs were conducted over a period of many weeks. In this study the 2-HMP was separated from the adsorbent so that all the observed hexahydrooxazolo[3,4-a]pyridine could be attributable to the formaldehyde vapour studied and the possibility of experimental artefacts arising from thermal degradation of the adsorbent excluded. The technique may be simplified still further by undertaking the derivatisation within the adsorbent trap; perhaps used in the passive mode.Such a study would be a logical continuation of this work and would improve the overall efficiency of the process. It is also important to note that the effects of residence time, temperature and Fig. 4 Examples of the chromatography obtained using (a) the PTV injector with no cryogenic focusing, analyte at 5.29, and (b) a thermal desorption unit with cryogenic focusing, analyte is peak number 5 and the derivative is peak number 11.Table 5 Summary of the analysis parameters for the determination of hexahydrooxazolo[3,4-a]pyridine recovered from the denuder tubes by thermal desorption with cold focusing Instrument variable Level Denuder tube dimensions 15 cm long, 0.6 cm od and 0.3 cm id Denuder tube coating 3.2 mg of 2-hydroxymethylpiperidine Instrument type Chrompack CP 9000 fitted with an integrated TCT/PTI system (thermal desorption unit) Desorption temperature 150 °C Desorption time 2 min Cold trap temperature 2100 °C Cold trap desorption temperature 150 °C Carrier gas oxygen-free nitrogen Desorption gas flow 32.3 ± 1.6 cm3 min21 Split flow 30.0 ± 1.2 cm3 min21 Gas chromatography column and J&W DB-Wax PEG 30 m, dimensions 0.32 mm id, and film thickness 0.25 mm Carrier gas flow 2.3 ± 0.3 cm3 min21 Gas chromatography temperature Start temperature: 70 °C programme Initial time: 1 min after cold trap reaches 150 °C Ramp: 15 °C min21 Final temperature: 205 °C Final time: 7.5 min Detector Flame ionisation detector held at 300 °C Analyst, December 1997, Vol. 122 1475environmental factors such as relative humidity need to be characterised, and these are the focus of ongoing work. Finally, investigating the extension of this approach to other aldehydes, and other derivatisation techniques for gas chromatographic analysis suggest themselves as obvious next stages of development for this technique.The authors acknowledge the support provided for C.D.McG. by EPSRC and Jones Chromatography Ltd. as part of the CASE Award programme. References 1 World Health Organisation, Air Quality Guidelines for Europe, European Series No. 23, WHO Regional Publications, Copenhagen, 1987. 2 Health and Safety Executive, HSE EH40/91 Occupational Exposure Limits 1991, Health and Safety Executive, London, 1991. 3 Mathews, T. G., Hawthorne, A. R., Howell, T. C., and Metcalfe, C. E., Environ. Int., 1982, 8, 143. 4 Shirtliffe, C. J., Rousseau, M. Z., Young, I. C., Silwinski, J. F., and Sim, P. G., Formaldehyde, Analytical Chemistry and Toxicology, Advances in Chemistry Series 210, American Chemical Society, Washington, DC, USA, 1985. 5 Molhave, L., Indoor Air Pollution by Formaldehyde in European Countries, European Concerted Action Report 7, EUR 13216 EN, Commission of the European Communities, Luxembourg, 1990. 6 Goelen, E., Lambrechts, M., and Geyskens, F., Analyst, 1997, 122, 411. 7 Otson, R., and Fellin, P., Sci. Total Environ., 1988, 77, 95. 8 Vairavamurthy, A., Roberts, J. M., and Newman, L., Atmos. Environ., 1992, 26A, 1965. 9 Keith, L. H., Patton, G. L., Lewis, D. L., and Edwards, P. G., Principles of Environmental Sampling, ed. Keith, L. H., 2nd edn., American Chemical Society, Washington, DC, USA, 1996, ch. 1. 10 Yasuhara, A., and Shibamoto, T., J.Assoc. Off. Anal. Chem., 1989, 72, 899. 11 Handbook of Derivatives for Chromatography, ed. Blau, K., and Halket, J. M., 2nd edn., Wiley, Chichester, UK, 1993. 12 Crider, W. L., Barkley, N. P., Knott, M. J., and Slater, R. W., Anal. Chim. Acta, 1969, 47, 237. 13 Ferm, M., Atmos. Environ., 1979, 13, 1385. 14 Ali, Z., Thomas, C. L. P., and Alder, J. F., Analyst, 1989, 114, 759. 15 Possanzini, M., Ciccioli, P., Di Palo, V., and Draisel, R., Chromatographia, 1987, 23, 829. 16 Cecchini, F., Febo, A., and Possanzini, M., Anal. Lett., 1985, 18, 681. 17 Kennedy, E. R., O’Connor, P. F., and Gagnon, Y. T., Anal. Chem., 1984, 56, 2120. 18 National Institute of Occupational Safety and Health, NIOSH Manual of Analytical Methods, Method 2541, National Institute of Occupational Safety and Health, Washington, DC, USA, 1989. 19 Walker, F. R., Formaldehyde, 2nd edn., Reinhold Publishing, New York, USA, 1964. 20 Ho, M. H., Formaldehyde, Analytical Chemistry and Toxicology, ed.Turoski, V., Advances in Chemistry, American Chemical Society, Washington, DC, USA, 1985. 21 NIOSH, Manual of Analytical Methods, Method 3500, National Institute of Occupational Safety and Health, Washington, DC, USA, 1989. 22 Miller, J. C., and Miller, J. N., Statistics for Analytical Chemistry, 2nd edn., Ellis Horwood, Chichester, UK, 1988. Paper 7/04731B Received July 4, 1997 Accepted September 29, 1997 Fig. 5 Graph showing relationship between mass of formaldehyde sampled and instrument response for the analysis of the residues within the denuder tube used to sample a concentration of 5.92 mg m23 formaldehyde in air at ambient pressure and temperature, with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa.The error bars are the 95% confidence intervals. Data points without error bars were obtained with a single analysis only. 1476 Analyst, December 1997, Vol. 122 Determination of Formaldehyde by Conversion to Hexahydrooxazolo[3,4- a]pyridine in a Denuder Tube With Recovery by Thermal Desorption, and Analysis by Gas Chromatography–Mass Spectrometry† C.L. Paul Thomasa, Charlotte D. McGillb and Robert Towillc a Department of Instrumentation and Analytical Science, UMIST, PO Box 88, Manchester, UK M60 1QD b Department of Chemistry, University of Reading, Whiteknights, PO Box 224, Reading, UK RG6 6AD c Jones Chromatography, Hengoed, Mid Glamorgan, UK CF8 8AU The implications for formaldehyde analysis arising from a recent study where the inter-laboratory relative standard deviation for formaldehyde [CAS Registry number: 50-00-0] monitoring was reported to fall in the range 20 to 60% are discussed.A thermal desorption approach is proposed as a possible way to address the requirement for less intensive sample processing in formaldehyde analysis; as well as meeting other important criteria such as reduced sample handling and increased overall sensitivity.A denuder tube coated with 2-hydroxymethylpiperidine [CAS Registry number: 3433-37-2] and connected to an adsorbent sampler packed with Tenax TA was used to derivatise formaldehyde vapour. The volatile derivative, hexahydrooxazolo[3,4-a]pyridine [CAS Registry number: 274-50-0], was retained on the adsorbent trap. Analysis of the derivative was accomplished using a thermal desorption technique coupled to gas chromatography–mass spectrometry. To test the approach a test atmosphere generator based on a permeation tube containing a-polyoxymethylene was constructed. A dilution chamber heated to 80 °C was used to prevent repolymerisation of the released formaldehyde.Gravimetric measurements indicated that formaldehyde was generated at a rate of 1.9 3 1024 g h21 (95% confidence limits = ± 1 3 1026 h21) and atmospheres of formaldehyde at concentrations of 5.92, 3.35 and 1.79 mg m23, with a water concentration less than 17 mg m23 at 20 °C and 1.013 kPa were validated using NIOSH Method 3500 for formaldehyde determination. Formaldehyde masses in the range 0.1 to 16 mg were sampled using the system described. The contents of the adsorbent trap were then analysed.The data obtained supported the hypothesis that formaldehyde derivatives could be recovered in an analytically useful way by thermal desorption. Mass spectrometric data confirmed previous assignments for hexahydrooxazolo[3,4-a]pyridine and that thermal decomposition did not occur to a detectable extent during desorption.The relationships between mass sampled and instrument response were linear up to the point of breakthrough, with limits of detection in the range 0.03 to 0.51 mg m23 formaldehyde in air. Keywords: Formaldehyde; denuder tubes; adsorbent trap; Tenax TA; 2-hydroxymethylpiperidine; hexahydrooxazolo[3,4-a]pyridine; test atmosphere generation; thermal desorption gas chromatography–mass spectrometry; sampling Aldehydes in general, and formaldehyde in particular, are acknowledged to be important volatile organic compounds which are present in many industrial operations.Exposure to formaldehyde at low concentrations results in acute and chronic health effects. The odour detection threshold has been estimated to fall in the range 0.06 to 1.2 mg m23 and the onset of irritation to eyes and upper respiratory tract takes place in the range 0.01 to 3.1 mg m23.1 Formaldehyde causes cancer in rats and is a suspected human carcinogen.The exposure limits in the United Kingdom are set at 2.5 mg m23 for short-term exposure over 15 min, as well as long-term time-weighted average exposure over 8 h.2 Formaldehyde is also a precursor of photochemical smog, and in the presence of water produces formic acid and methanol. The potential damage to artefacts in museums, art galleries and other archives arising from formaldehyde pollution requires a reappraisal of the materials used currently for the construction of display and storage facilities.Exposure to formaldehyde also occurs in domestic dwellings, where maximum indoor concentrations have been estimated to reach levels of 0.2 to 4 mg m23,3,4 while mean concentrations lie in the range 0.02 to 0.06 mg m23,5 an unfavourable comparison with the maximum exposure limits mentioned above. It is not surprising, therefore, that much effort has been devoted to developing and improving methods for the determination of formaldehyde. A recent paper provides a timely and authoritative comparison of the techniques currently available,6 although the use of direct reading fuel cell monitors was not included in this work.Extensive reviews of formaldehyde determination have also been published.7,8 Detection limits for formaldehyde determination are generally quoted as falling in the range 0.24 mg m23–0.25 mg m23, but perhaps as significant are the values of measurement uncertainty associated with formaldehyde determination.Under controlled laboratory conditions inter-laboratory values for relative standard deviation (RSD) have been reported to fall in the range 20–60%.6 Such variance in the sample collection procedures means that the resources required to run monitoring programmes, where sample heterogeneity would only serve to increase the uncertainty in the measurement still further, are likely to be prohibitively large. Table 1 illustrates this point with the numbers of samples needed to meet a range of data quality objectives (DQO) when the overall RSD, including measurement and sampling contributions to the uncertainty, is 50%.Even modest DQO require large numbers of independent samples obtained from a random sampling plan. Given the recognised need to monitor formaldehyde, sampling methods are required that either: (i) reduce the complexity and cost of the sampling systems involved and thus enable higher specifications of DQO to be set; (ii) reduce the experimental uncertainties involved and so reduce the numbers † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997.Analyst, December 1997, Vol. 122 (1471–1476) 1471of independent samples needed to yield high specification DQO; (iii) lower the limits of detection for formaldehyde determination to enable more complete monitoring of formaldehyde in the outdoor environment; (iv) or a combination of all three requirements.As well as these considerations there is the need to acknowledge that the materials used in the construction of laboratories may themselves be a significant source of interference in formaldehyde analysis; concentrations of 0.61 mg m23 of formaldehyde in air have been reported in laboratory atmospheres.10 Consequently, methods that reduce the exposure of the sample to laboratories and other potential sources of interference may also be seen to offer benefits.This work seeks to address the issues outlined above by testing the hypothesis that volatile derivatives of formaldehyde generated during sampling by a denuder tube may be trapped on an adsorbent bed and that the derivatives may be recovered using a thermal desorption technique for analysis by gas chromatography. Derivatisation approaches are used often in gas chromatography, to either stabilise analytes, enhance the chromatographic behaviour of an analyte or impart characteristics that enable more selective and sensitive detection to be used.11 However the direct recovery of such a derivative by thermal desorption has not been previously reported.Such an approach appears attractive because: (i) it removes the need for solvent elution of the derivatised formaldehyde thereby reducing the cost and complexity of the sampling procedures; (ii) it is amenable to automation; (iii) it isolates the sampled formaldehyde from the laboratory; (iv) it offers enhanced sampling efficiency in that all the derivatised material will be transferred to the gas chromatograph instead of an aliquot of an extract; (v) and offers the opportunity to use cryogenic focusing to attain lower limits of detection.Denuder tubes in the analysis of trace gases were originally conceived as a method for removing interfering species from a gas sample12 and later as a means for isolating analytes from the sample,13,14 including formaldehyde.15,16 Other approaches involved heating denuder tubes to produce thermochemical transformations in some of the sampled species, reported first in 1978.Thus a denuder tube used to derivatise the analyte to produce a volatile product, subsequently retained in an adsorbent trap, may be seen to be a development of this idea. The theory and uses of denuder tubes have been reviewed previously.13 Thus this work used a denuder tube coated with 2-hydroxymethylpiperidine (2-HMP) connected to a Tenax TA adsorbent trap.Formaldehyde diffused to the walls of the denuder tube and reacted yielding hexahydrooxazolo[3,4-a]pyridine which desorbed and was swept into the adsorbent trap. 2-HMP was chosen as it is the basis of established and accepted methodologies for the determination of formaldehyde and produces a volatile derivative.17,18 An alternative approach would have been to have impregnated the Tenax TA with the 2-HMP directly; which would have been analogous to the approach of commercially available formaldehyde sampling devices from Supelco (Orbo tubes, Fancy, Poole, Dorset, UK) which are based on an XAD Resin and use solvent elution.The advantages of a denuder tube approach over an impregnated adsorbent were envisaged to be: (i) the prevention of experimental artefacts arising from the release of formaldehyde through thermal degradation of the adsorbent as it is thermally desorbed; and (ii) the consequences of competitive adsorption between the 2-HMP coating and hexahydrooxazolo[3,4-a]pyridine were not known.Experimental Generation of Standard Formaldehyde Atmospheres Formaldehyde vapour is stable as a monomer between 80 and 150 °C. Below 80 °C polymerisation occurs, and above 150 °C decomposition takes place.19 Further work has noted discrepancies between gravimetric calibration of formaldehyde permeation tube sources based upon a-polyoxymethylene and observed concentrations of monomeric formaldehyde.20 Suggested reasons for this were: (i) other degradation products, as well as formaldehyde, being released from the permeation tube; (ii) thermal decomposition of formaldehyde; and (iii), impurities in the a-polyoxymethylene.The vapour generator was designed therefore to maintain the released vapours from a permeation tube at 80 °C until dilution had been achieved; to quench any possible repolymerisation due to reaction with coreleased impurities. Further, all the internal surfaces of the test atmosphere generator were made from either PTFE or silanised borosilicate glass.Once diluted the test atmosphere was allowed to cool to room temperature before it was passed into a sampling manifold. The permeation tubes were made from 6.25 mm od PTFE tubing filled with a-polyoxymethylene (Merck, Poole, Dorset, UK) and capped with silanised glass plugs. It was calibrated by a series of mass-loss measurements made over a period of approximately 2500 h.The data showed that the release of material from the permeation tube was constant throughout this time at a rate of 1.9 3 1024 g h21 ± 1 3 1026 g h21 (95% confidence limits for 27 mass loss measurements). As mass-loss measurements involved disruption of the vapour source a stabilisation period of 1 week was invoked after each mass loss determination before experimentation could recommence. The diluent gas used was bottled air passed through purification media (Phase Separations, London, UK) before use.All flow controls were calibrated and the estimated concentrations from the gravimetric data were compared with observed concentrations obtained from NIOSH Method 3500,21 the chromotropic acid method. Three standard atmospheres were used in this work, 5.92, 3.35 and 1.79 mg m23, with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa (0.1% relative humidity, RH). At each of these concentrations seven different sample masses were taken across the range 10–140 mg.The concentrations of the test atmosphere and sampled masses determined by NIOSH Method 3500 were correlated against the gravimetric data. The results of this validation are presented in Table 2. The inference from this experiment was that under the operating conditions described above gravimetric data could be used to estimate test atmosphere concentrations to within the limits of uncertainty of NIOSH Method 3500. Studies undertaken for extended sampling periods and volumes showed an absence of sampling artefacts in the data; demonstrating a lack of surface activity in the test atmosphere generator and sampling train.Fig. 1 is a schematic diagram of the test atmosphere generator. Preparation of the Denuder Tube A glass tube, 15.8 cm long, 0.6 cm od and 0.3 cm id was coated with 3.2 mg of 2-HMP (Aldrich, Poole, Dorset, UK). The inlet zone was 1 cm long. Two 10 ml aliquots of 2-HMP in methanol (Merck) at a concentration of 160 mg cm23 were applied to the Table 1 Number of samples required to achieve a range of data quality objectives for a method with an overall RSD of 50%.Calculations based on methodology described in ref. 9. % Maximum acceptable % Confidence error in result (same units as RSD) 5 10 20 80 166 43 12 90 273 70 19 95 387 99 27 99 668 170 46 1472 Analyst, December 1997, Vol. 122internal surfaces of the denuder tube and the tube was rotated and tilted while the methanol evaporated leaving a coating of 2-HMP on the interior surfaces of the tube.The denuder tube was then immediately sealed with PTFE plugs. Preparation of the Adsorbent Bed for Use in a Programmable Temperature Vaporiser (PTV) Injector The denuder tubes described above were connected to an injector liner (Optic, Atas, Cambridge, UK) packed with 0.089 g of Tenax TA (Phase Separations). The Tenax TA trap was conditioned for 3 h before use by placing it in the PTV injector (an Optic unit) and then heating it to 350 °C while passing a flow of helium through it at approximately 20 cm3 min21.The sampling units were sealed with PTFE end caps immediately after they were prepared. Determination of Sampled Mass of Formaldehyde by Recovery of Oxazolidine The sampler was connected to the sampling manifold using a silanised glass connector and a sampling pump, set to sample at 15 cm3 min21. Three concentrations of 1.79, 3.35 and 5.92 mg m23 formaldehyde in air at ambient pressure and temperature, with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa were sampled to give a range of sampled masses of formaldehyde at different concentrations.The Tenax packed injection liner was then placed in the PTV injector and the trapped materials were thermally eluted onto a gas chromatography column. Detection was by electron-ionisation quadrupole mass spectrometry. The analysis parameters are summarised in Table 3, and the results of these runs are in Fig. 2. Discussion Generation of Formaldehyde Test Atmospheres The results in Table 2 show that the test atmosphere generator produced stable and reproducible standards of formaldehyde vapour, and that gravimetric data obtained from the permeation tubes could be used to calibrate formaldehyde monitoring methodologies. It was found that a stabilisation time of 1 week was needed after any parameter affecting the concentration of formaldehyde in the unit had been changed.The stabilisation period ensured that residual surface activity within the system attained equilibrium with the test atmospheres and did not interfere with subsequent experiments. The stabilisation period was especially important for low concentration studies. At- Table 2 Summary of the correlation of the formaldehyde concentrations, obtained from the gravimetric data, against the observed formaldehyde concentrations using the NIOSH Method 3500* Formaldehyde concentration/ mg m23 Gradient Intercept/mg r2 1.79 1.223 ± 0.11 23.5 ± 13.8 0.96 3.35 0.974 ± 0.05 28.7 ± 5.9 0.99 5.92 1.011 ± 0.04 22.0 ± 5.4 0.99 * 27 gravimetric measurements were made over a period of 2500 h.Each gravimetric determination was based on five separate measurements. These data were linearly correlated to time to give a release rate of 1.49 3 1024 g h21 ± 1.1026 g h21 at 95% confidence limits. Test atmospheres at the above concentrations were generated and determined by the NIOSH 3500 Method for seven sampled masses.Each test atmosphere concentration determination was based on five independent measurements using the NIOSH 3500 Method. For the two methods to agree the gradient of the regression should be 1 with an intercept of zero.22 95% confidence limits are given, and it can be seen that within the limits of experimental error the concentrations of the test atmosphere determined from the gravimetric approach and the NIOSH 3500 Method agree.Fig. 1 Schematic diagram of the test atmosphere used to produce the formaldehyde standards for this study: 1, purified air supply; 2, flow controllers; 3, flow meters; 4, permeation tube holder made from 15 cm of 1.3 cm od glass tubing; 5, permeation tube; 6, dilution chamber made from 20 cm long by 2.5 cm od glass tubing fused to the permeation tube holder; 7, sampling manifold; 8, charcoal trap; 9, denuder tube coated with 2-HMP connected to the inlet of a Tenax TA filled adsorbent trap; 10, sampling pump; 11, exhaust.Table 3 Summary of the analysis parameters for the determination of hexahydrooxazolo[3,4-a]pyridine recovered from the Tenax TA adsorption traps by thermal desorption Instrument variable Level Sampler type Adsorbent trap fitted with a denuder tube. Adsorbent holder is an Optic compatible thermal desorption tube. Adsorbent 8.9 mg of Tenax TA Denuder tube dimensions 15 cm long, 0.6 cm od and 0.3 cm id Denuder tube coating 3.2 mg of 2-hydroxymethylpiperidine Instrument type Fisons 8035 gas chromatograph fitted with a programmable temperature vaporising injection unit (Optic) in conjunction with a Fisons Trio 1000 Quadrupole mass spectrometer.Desorption temperature Start temperature: 50 °C Initial time: 1 min Temperature ramp: 960 °C min21 Final temperature: 200 °C Carrier gas Helium Desorption gas flow 102.3 cm3 min21 Split flow 100 cm3 min21 Gas chromatography column and J&W DB-Wax PEG 30 m, dimensions 0.32 mm id, and film thickness 0.25 mm.Carrier gas flow 2.3 ± 0.3 cm3 min21 Gas chromatography temperature Start temperature: 35 °C programme Initial time: 0 min Ramp rate 1: 50 °C min21 Temp 2: 110 °C Time at Temperature 2: 1 min Ramp rate 2: 15 °C min21 Final temperature: 190 °C Final time: 1 min Detector Mass spectrometer settings were set by the instrument under ‘auto-tune’ Analyst, December 1997, Vol. 122 1473tempts to reduce the stabilisation time resulted in increases in the variances of the data obtained.The importance of a heated permeation tube holder and dilution chamber was demonstrated by alternatively reducing the heating and the dilution flow. In both cases repolymerisation of formaldehyde was observed as a white coating on the interior surfaces of the test atmosphere generator. Finally, if the temperature of the permeation tube was taken too high, greater than 130 °C, the concentration of released formaldehyde vapour increased beyond the effective capacity of the dilution chamber, resulting in repolymerisation of formaldehyde throughout the test atmosphere generator. Mass Spectra Table 4 presents the mass spectrometric data obtained for hexahydrooxazolo[3,4-a]pyridine along with tentative assignments.The data confirm those reported previously and the assignment of the hexahydrooxazolo[3,4-a]pyridine peak.16 No evidence was found of thermal decomposition of hexahydrooxazolo[ 3,4-a]pyridine caused by thermal desorption.Analysis of Recovered Hexahydrooxazolo[3,4-a]pyridine The data in Fig. 2 are consistent with the behaviour of a volatile compound in an adsorbent sampling bed, used in the active mode. Initially the recovery of derivative increases linearly with sampled mass, see Fig. 3. As breakthrough starts the plot of peak area versus sampled mass becomes non-linear until the whole adsorbent bed reaches equilibrium at complete breakthrough.The equilibrium capacity of the adsorbent was observed to depend upon the concentration of the formaldehyde vapour; which is what would be expected for Type I physical adsorption.23 In which case the mass trapped, Mt, is related to the concentration of the sampled vapour, [i], by the expression, M K i K i K j j j n t = + + = å [ ] [ ] [ ] 1 1 (1) where K is a constant analogous to the adsorption constant, and Kj and [j] refer to the constants and concentrations for the other adsorbates present in the sampled air.In pure diluent air, as found in the test atmosphere generator, and at low concentrations of formaldehyde, eqn. 1 may be approximated to, Mt = K[i] (2) which indicates that the capacity of an adsorption trap will decrease in direct proportion to the concentration of the analyte. Analysis of the observed trap capacities showed a linear relationship between the formaldehyde concentration and the mass of material recovered by thermal desorption; mass Fig. 2 Graph showing the relationship of the peak areas obtained from the thermally desorbed hexahydrooxazolo[3,4-a]pyridine with increasing mass of sampled formaldehyde at concentrations of 1.79, 3.35 and 5.92 mg m23 formaldehyde in air at ambient temperature and pressure with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa. Table 4 Summary of the mass spectrometric data, confirming the identity of the hexahydrooxazolo[3,4-a]pyridine peak Reactive ion Relative mass/u abundance (%) Assignment 127 30 Molecular ion 126 52 Loss of H from molecular ion 98 21 97 100 Loss of H2CO from molecular ion 82 6 70 15 69 70 Loss of NCH2OCH2 from molecular ion 68 18 56 18 55 21 Loss of CH2 from reactive ion mass 69 54 9 41 58 Loss of CH2 from reactive ion mass 55 Fig. 3 Graphs showing the linear relationships, before breakthrough in Fig. 2, 95% confidence limits shown. 1474 Analyst, December 1997, Vol. 122ratio = 0.097 3[H2CO] + 9 31026. The mass ratio is given by the expression Mt Ma , where Ma is the mass of the adsorbent used, in this case 8.9 mg, while the units for the constant K are m3 g21.The correlation coefficient was 0.998. It is likely that the sampling volumes associated with this study could be increased significantly by increasing the mass of adsorbent used in the trap; the small adsorbent masses were used to ensure rapid heating of the trap during desorption. Doing this would increase the sensitivity of the technique as well as lower the detection limits, perhaps enough to enable background concentrations of formaldehyde in unpolluted air to be monitored.To do this, cryogenic focusing of the thermally desorbed products would need to be incorporated and Fig. 4 is a comparison of the chromatography obtained with and without cryogenic focusing. However, as the capacity of the trap diminishes with decreasing analyte concentration a limit will be reached beyond which increasing the trap size will not yield significant benefits in terms of sensitivity and lowered limits of detection.The cryogenic focusing was used to investigate the hexahydrooxazolo[ 3,4-a]pyridine residues still within the denuder tube after sampling and to establish if the denuder tube might work as a stand alone sampling unit. In this case the denuder tube was desorbed directly into a gas chromatograph system. Table 5 summarises the experimental parameters used. The data obtained from this study for concentrations of formaldehyde below 5.92 g m23 formaldehyde in air at ambient pressure and temperature, with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa indicated trace residues only remaining within the denuder tube. At these concentrations the data were not reproducible and there was no clear correlation between the mass of formaldehyde sampled and the amount of recovered material.At 5.92 g m23 a linear relationship between peak area and sampled mass was observed and linear regression analysis gave the relationship, Peak area = 43.3 3 mass of formaldehyde sampled + 0.58.The units for peak area were mV min21 and the units for mass were mg. The correlation coefficient was 0.992 and the RSD values were in the range 10–30%, see Fig. 5. These findings supported the assumption that the volatile derivative was efficiently transferred from the denuder tube to the adsorbent trap and confirmed that the equilibrium capacity of the denuder tube was not large enough at concentrations below 5.92 g m23 for it to be used as the basis of an effective stand alone device. Conclusions These data show that it is possible to recover volatile derivatives for analysis by gas chromatography by thermal desorption.This is a previously untried approach. In this study hexahydrooxazolo[ 3,4-a]pyridine has been produced from sampled formaldehyde with a denuder tube coated with 2-HMP and trapped on a Tenax adsorbent bed used in conjunction with a thermal desorption technique.The system investigated determined formaldehyde at concentrations below current exposure limits, and the estimated limits of detection were respectively 0.51, 0.03 and 0.05 mg per sample, for the formaldehyde concentrations of 5.92, 3.35 and 1.79 mg m23 formaldehyde in air at ambient pressure and temperature, with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa. Which may be compared to the 1 mg per sample stated by NIOSH Method 2541,17 based on 2-HMP.The variations in these values may be attributed to variations in the ion source of the mass spectrometer as these runs were conducted over a period of many weeks. In this study the 2-HMP was separated from the adsorbent so that all the observed hexahydrooxazolo[3,4-a]pyridine could be attributable to the formaldehyde vapour studied and the possibility of experimental artefacts arising from thermal degradation of the adsorbent excluded.The technique may be simplified still further by undertaking the derivatisation within the adsorbent trap; perhaps used in the passive mode. Such a study would be a logical continuation of this work and would improve the overall efficiency of the process. It is also important to note that the effects of residence time, temperature and Fig. 4 Examples of the chromatography obtained using (a) the PTV injector with no cryogenic focusing, analyte at 5.29, and (b) a thermal desorption unit with cryogenic focusing, analyte is peak number 5 and the derivative is peak number 11.Table 5 Summary of the analysis parameters for the determination of hexahydrooxazolo[3,4-a]pyridine recovered from the denuder tubes by thermal desorption with cold focusing Instrument variable Level Denuder tube dimensions 15 cm long, 0.6 cm od and 0.3 cm id Denuder tube coating 3.2 mg of 2-hydroxymethylpiperidine Instrument type Chrompack CP 9000 fitted with an integrated TCT/PTI system (thermal desorption unit) Desorption temperature 150 °C Desorption time 2 min Cold trap temperature 2100 °C Cold trap desorption temperature 150 °C Carrier gas oxygen-free nitrogen Desorption gas flow 32.3 ± 1.6 cm3 min21 Split flow 30.0 ± 1.2 cm3 min21 Gas chromatography column and J&W DB-Wax PEG 30 m, dimensions 0.32 mm id, and film thickness 0.25 mm Carrier gas flow 2.3 ± 0.3 cm3 min21 Gas chromatography temperature Start temperature: 70 °C programme Initial time: 1 min after cold trap reaches 150 °C Ramp: 15 °C min21 Final temperature: 205 °C Final time: 7.5 min Detector Flame ionisation detector held at 300 °C Analyst, December 1997, Vol. 122 1475environmental factors such as relative humidity need to be characterised, and these are the focus of ongoing work. Finally, investigating the extension of this approach to other aldehydes, and other derivatisation techniques for gas chromatographic analysis suggest themselves as obvious next stages of development for this technique.The authors acknowledge the support provided for C.D.McG. by EPSRC and Jones Chromatography Ltd. as part of the CASE Award programme. References 1 World Health Organisation, Air Quality Guidelines for Europe, European Series No. 23, WHO Regional Publications, Copenhagen, 1987. 2 Health and Safety Executive, HSE EH40/91 Occupational Exposure Limits 1991, Health and Safety Executive, London, 1991. 3 Mathews, T. G., Hawthorne, A. R., Howell, T. C., and Metcalfe, C. E., Environ. Int., 1982, 8, 143. 4 Shirtliffe, C. J., Rousseau, M. Z., Young, I. C., Silwinski, J. F., and Sim, P. G., Formaldehyde, Analytical Chemistry and Toxicology, Advances in Chemistry Series 210, American Chemical Society, Washington, DC, USA, 1985. 5 Molhave, L., Indoor Air Pollution by Formaldehyde in European Countries, European Concerted Action Report 7, EUR 13216 EN, Commission of the European Communities, Luxembourg, 1990. 6 Goelen, E., Lambrechts, M., and Geyskens, F., Analyst, 1997, 122, 411. 7 Otson, R., and Fellin, P., Sci. Total Environ., 1988, 77, 95. 8 Vairavamurthy, A., Roberts, J. M., and Newman, L., Atmos. Environ., 1992, 26A, 1965. 9 Keith, L. H., Patton, G. L., Lewis, D. L., and Edwards, P. G., Principles of Environmental Sampling, ed. Keith, L. H., 2nd edn., American Chemical Society, Washington, DC, USA, 1996, ch. 1. 10 Yasuhara, A., and Shibamoto, T., J. Assoc. Off. Anal. Chem., 1989, 72, 899. 11 Handbook of Derivatives for Chromatography, ed. Blau, K., and Halket, J. M., 2nd edn., Wiley, Chichester, UK, 1993. 12 Crider, W. L., Barkley, N. P., Knott, M. J., and Slater, R. W., Anal. Chim. Acta, 1969, 47, 237. 13 Ferm, M., Atmos. Environ., 1979, 13, 1385. 14 Ali, Z., Thomas, C. L. P., and Alder, J. F., Analyst, 1989, 114, 759. 15 Possanzini, M., Ciccioli, P., Di Palo, V., and Draisel, R., Chromatographia, 1987, 23, 829. 16 Cecchini, F., Febo, A., and Possanzini, M., Anal. Lett., 1985, 18, 681. 17 Kennedy, E. R., O’Connor, P. F., and Gagnon, Y. T., Anal. Chem., 1984, 56, 2120. 18 National Institute of Occupational Safety and Health, NIOSH Manual of Analytical Methods, Method 2541, National Institute of Occupational Safety and Health, Washington, DC, USA, 1989. 19 Walker, F. R., Formaldehyde, 2nd edn., Reinhold Publishing, New York, USA, 1964. 20 Ho, M. H., Formaldehyde, Analytical Chemistry and Toxicology, ed. Turoski, V., Advances in Chemistry, American Chemical Society, Washington, DC, USA, 1985. 21 NIOSH, Manual of Analytical Methods, Method 3500, National Institute of Occupational Safety and Health, Washington, DC, USA, 1989. 22 Miller, J. C., and Miller, J. N., Statistics for Analytical Chemistry, 2nd edn., Ellis Horwood, Chichester, UK, 1988. Paper 7/04731B Received July 4, 1997 Accepted September 29, 1997 Fig. 5 Graph showing relationship between mass of formaldehyde sampled and instrument response for the analysis of the residues within the denuder tube used to sample a concentration of 5.92 mg m23 formaldehyde in air at ambient pressure and temperature, with a water concentration of less than 17 mg m23 at 20 °C and 1.013 kPa. The error bars are the 95% confidence intervals. Data points without error bars were obtained with a single analysis only. 1476 Analyst, December 1997, Vol. 122

ISSN:0003-2654    

DOI:10.1039/a704731b

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Determination of Dissolved Reactive Phosphorus in Estuarine Waters Using a Reversed Flow Injection Manifold† Stefan Auflitschab, Darren M. W. Peata, Ian D. McKelviec and Paul J. Worsfold*a a Department of Environmental Sciences, Plymouth Environmental Research Centre, University of Plymouth, Drake Circus, Plymouth, Devon UK PL4 8AA b Institute of Analytical Chemistry, Vienna University of Technology, Getreidemarkt 9/151, A-1060 Vienna, Austria c Water Studies Centre, Chemistry Department, Monash University, Caulfield East, Victoria 3145, Australia The Schlieren or refractive index (RI) effect is a major problem in the determination of dissolved reactive phosphorus in estuarine waters using conventional flow injection (FI) manifolds with sample injection.This is because differences in RI between the injected sample zone and the carrier stream give rise to a lensing effect which is superimposed on the blank response and causes significant error in quantitation.A simple reversed flow injection (rFI) manifold using spectrophotometric detection which removes these quantitation errors is reported. Acidic molybdate is injected into a sulfuric acid carrier stream of the same refractive index and sequentially merged with sample and reductant (ascorbic acid). Reduction of phosphomolybdate to phosphomolybdenum blue is carried out in a coil thermostated at 60 °C. Dissolved reactive phosphorus has been successfully determined in estuarine waters with salinities ranging from 0 to 30‰ using calibration standards prepared in deionized water, with a detection limit of 2 mg l21 PO4–P and a linear range of 2–100 mg l21 PO4–P (r2 = 0.9998).Keywords: Dissolved reactive phosphorus; phosphate; salinity; refractive index; Schlieren effect; flow injection; estuarine water Phosphorus is an essential and often limiting nutrient for algal and bacterial growth1 and the total phosphorus content of natural waters comprises both particulate and dissolved forms, the latter being operationally defined as the fraction which passes through a 0.45 mm membrane.This total dissolved phosphorus (TDP) fraction can be further sub-divided into dissolved inorganic phosphorus (DIP) and dissolved organic phosphorus (DOP). The DIP fraction comprises orthophosphate and condensed phosphates with orthophosphate being the most frequently determined form of DIP and also referred to as dissolved reactive phosphorus (DRP).Flow injection (FI) with spectrophotometric detection has been widely reported for monitoring nutrients in natural waters.2,3 However, the application of conventional FI to the determination of DRP in estuarine waters has been very limited due to the so-called refractive index (RI) or Schlieren effect.4 This occurs when samples with differing ionic strengths are injected into a carrier stream of lower ionic strength.5 Under laminar flow conditions the geometry of the sample zone is parabolic and differences in refractive index between the sample zone and carrier stream lead to a lensing effect, in which the sample zone acts as a dispersing and collecting lens.As a result the Schlieren effect causes a negative frontal peak followed by a positive peak. Large errors in quantitation are caused by this effect, especially at low analyte concentrations. 4 When measuring DRP in seawater the RI problem can be overcome using a matched carrier stream (typically 39 g l21 NaCl).In estuarine waters, however, the salinity varies in the range 0–35‰ and a matrix matching approach would require the preparation of carrier solutions or calibration standards covering the whole range of salinities encountered, which is not a practical solution for routine laboratory work or field deployment. In a recent paper4 a reversed flow injection (rFI) manifold was reported for the determination of DRP in estuarine waters. This method overcame a number of difficulties encountered when conventional FI manifolds are used.Schlieren effects were eliminated by using a rFI manifold in which molybdate was injected into a NaCl RI-matched carrier stream which was continuously merged with the sample stream. The effect of varying sample chloride concentration on the efficiency of the tin(ii) chloride reduction step was masked by a swamping chloride concentration (60‰ as NaCl) in the carrier stream. This allowed one set of aqueous DRP standards to be universally applied over the salinity range 0–30‰.However, the analytical signal obtained was a doublet peak due to the formation of a pronounced acid concentration gradient in the injected molybdate zone. Kinetic inhibition of molybdenum blue complex formation was most severe in the centre of the sample zone where the acid concentration was highest. It was not possible to match both the RI and acid concentrations because the chloride concentration of the carrier stream was critical in order to avoid interference in the tin(ii) chloride reduction step.6 This effect also restricted the carrier stream to sample stream flow rate ratio that could be used (2.5 : 1) which limited the sensitivity due to on-line sample dilution.This paper describes the successful use of a new salinity compensation manifold that allows the direct determination of DRP in estuarine waters (0–35‰) using a single set of deionized water standards. Reduction was performed using ascorbic acid because this reaction is not influenced by the matrix chloride concentration.7 This allowed the acid concentration in the carrier stream and the injected molybdate reagent stream to be matched without susceptibility to chloride interference and resulted in a singlet analytical peak. In addition the freedom to choose a more favourable carrier stream to sample stream split ratio resulted in a much improved detection limit and sensitivity. These improvements make the method ideally suited to the rapid determination of DRP in estuarine waters, particularly at the lower end of the typical concentration range (5–50 mg l21 PO4–P) found in these waters.The † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997. Analyst, December 1997, Vol. 122 (1477–1480) 1477simplicity of the FI manifold and the calibration procedure facilitates the rapid determination of phosphate in estuarine waters and makes it ideal for shipboard deployment.Experimental Reagents Phosphate stock solution, 1000 mg l21 P as KH2PO4. Oven dried potassium dihydrogen orthophosphate (4.3937 g) was diluted to 1000 ml with deionized water, and stored at 4 °C. Working standards in the range 5–200 mg l21 PO4–P were prepared each day. Ammonium molybdate reagent. Ammonium molybdate [20 g; (NH4)6Mo7O24.4H2O] was dissolved in 700 ml of deionized water, 40 ml of concentrated sulfuric acid was added and the solution made up to 1000 ml.Ascorbic acid–antimony reagent. Ascorbic acid (20 g; C6H8O6) and 1 g of antimony potassium oxide (+)-tartrate 0.5 hydrate ( KSbO.C4H4O6.1/2H2O ) were dissolved in 700 ml of deionized water, 50 ml of concentrated sulfuric acid was added and the solution made up to 1000 ml. A fresh solution was prepared each day. Carrier solution. Concentrated sulfuric acid (56 ml) was added to 700 ml of deionized water and made up to 1000 ml.AsV and SiIV standards. Merck (Poole, Dorset, UK) standards containing 1000 mg l21 of SiIV and AsV were used to spike the phosphate standards. All solutions were prepared from BDH (Merck) AnalaR grade reagents using high purity Milli-Q (Millipore, Milford, MA, USA) water, and were sonicated for 15 min before use. The refractive indices of the ammonium molybdate and the carrier solutions were matched to within ±0.0005 with an Abb�e refractometer (Bellingham and Stanley, Tunbridge Wells, UK) by adding concentrated sulfuric acid to the carrier solution.Instrumentation and Procedures The manifold shown in Fig. 1 was used to determine DRP in standards and estuarine samples collected from the Tamar Estuary, Devon, UK. These samples were filtered (0.45 mm) in the laboratory within 6 h of collection and stored in a refrigerator for aaximum of 18 h prior to analysis. A Tecator 5020 FI analyser (Perstorp Analytical, Maidenhead, UK) and a Philips PU8620 Series UV/VIS/NIR (Philips, Cambridge, UK) single beam spectrophotometer were used to carry out the experiments.PTFE tubing of 0.5 mm id was used throughout. The flow cell used for the detection of phosphomolybdenum blue at 882 nm had an optical path length of 10 mm and an analytical volume of 18 ml. A Windows 95/NT Visual Basic application (Science Engineering Systems, Plymouth, UK) was used to control the instrument via a serial communication protocol and to store data from the instrument in a relational database.In this rFI manifold, ammonium molybdate solution was injected into a RI-matched sulfuric acid carrier and merged with the sample stream. The sample stream was changed manually. The formation of phosphomolybdic acid took place in a reaction coil, RC1, of 20 cm length. The phosphomolybdic acid was then reduced to phosphomolybdenum blue by ascorbic acid, with antimony acting as a catalyst, in a reaction coil, RC2, of 150 cm length.The main reaction coil (RC2) was immersed in a thermostated water-bath at 60 °C. Results and Discussion Optimisation of Molybdate Concentration To investigate the effect of molybdate concentration all other parameters were kept constant (as described in the Experimental section) and the molybdate concentration was varied over the range 1.7–40 g l21. Sample (200 mg l21 PO4–P) absorbance increased with molybdate concentration and was highest at 40 g l21, as shown in Fig. 2. However, at concentrations > 20 g l21 the gain in sensitivity was offset by the appearance of a trailing shoulder as shown in Fig. 3. The shape of this peak was variable and seriously degraded precision at lower phosphate concentrations. This is probably caused by direct reduction of the molybdate reagent8 at the high molybdate to acid concentration ratio which exists after merging with the sample stream and therefore a concentration of 20 g l21 was used for all subsequent studies.Interference From SiIV and AsV Sulfuric acid concentration strongly affects the formation kinetics of molybdenum heteropolyacid complexes with PO4 32, SiIV and AsV, with increasing acid concentration decreasing the rate of formation of all of these complexes.6 In the range 0.14–0.29 m H2SO4 the rate of formation of the PO4 32 complex is relatively stable9,10 but the formation rate of the SiIV and AsV complexes decreases dramatically at the higher acid concentra- Fig. 1 FI salinity compensation manifold.Fig. 2 Effect of ammonium molybdate concentration on the sample and blank responses. RC2 was 150 cm and the PO4–P concentration was 200 mg l21. Error bars represent ±3s (n = 3). A, sample; B, blank. Fig. 3 FI traces showing the effect of ammonium molybdate concentration on the peak profile. 1478 Analyst, December 1997, Vol. 122tion. In many estuarine waters SiIV is present at relatively high concentrations11 and hence the choice of acid concentration is a compromise between sensitivity and susceptibility to SiIV interference. This can be achieved using an acid concentration at the lower pH end of the acid stability plateau for orthophosphate.At this acidity (0.29 m) the rate of formation of phosphomolybdenum blue decreases only slightly compared with that at 0.14 m, whereas the rate of formation of silicomolybdenum blue and arsenomolybdenum blue decreases dramatically. In the rFI manifold reported here the optimum pH will be different from that for the batch method due to the dynamic nature of the system.In addition, the pH for the reduction step must be controlled by adjusting the pH of the ascorbic acid stream because the acidity of the molybdate and carrier streams has to be fixed in order to eliminate RI and acid gradient effects. The optimum acid concentration (0.9 m; 50 ml l21) gave a linear response over the range 0–100 mg l21 PO4–P [response (arbitrary units, AU) = 0.0004 concentration (mg l21 PO4–P) + 0.0026] with a r2 of 0.9994. Using this manifold, silicate interference does not occur below 1 mg l21 SiIV and over-estimation at 2 mg l21 SiIV is only equivalent to 3.5 mg l21 PO4–P, which is similar to the results reported for batch methods.6 Similarly, AsV showed no interference up to 5 mg l21, which is much higher than the level normally encountered in estuarine waters.However, 10 mg l21 AsV resulted in a phosphate overestimation of 5 mg l21 P.Optimisation of Phosphomolybdate Reduction The strong temperature dependence of the overall reaction is shown in Fig. 4, with the detector response rising from 0.14 AU at room temperature to 0.18 AU at 60 °C for a 500 mg l21 PO4–P standard. The blank signal concomitantly increased from 0.002 to only 0.004 AU and there was no degradation in precision (n = 3) at the higher temperature. Use of a heated reaction coil was necessary in order to increase the rate of reduction induced by ascorbic acid, because of the non-equilibrium conditions that exist in the FI manifold.12 Bubble formation in the reaction coil occurred at 70 °C, and although the sensitivity was improved at this temperature this was negated by a large increase in the blank response and a decrease in reproducibility.Operation of the manifold was therefore optimal when the reactor coil was thermostated at 60 °C. The length of RC1 had little effect on the overall performance of the manifold and therefore was set at 20 cm to facilitate a high sample throughput.Increasing the length of RC2 increased the dispersion but the absorbance still increased because of additional reaction time for the ascorbic acid reduction step. Fig. 5 shows that the calibration graph was sigmoidal at a coil length of 12.0 m whereas a coil length of 1.5 m gave a linear calibration (r2 = 0.9994) over the range 0–100 mg l21 PO4–P and was therefore used for all subsequent studies.Intermediate coil lengths showed increasingly sigmoidal behaviour. PTFE tubing (0.5 mm id) was also used throughout to minimise dispersion. Analytical Figures of Merit The effectiveness of this new salinity compensation manifold was demonstrated by comparing the calibration data for a series of PO4–P standards (5–100 mg l21) prepared at differing salinities (0–30‰). Regression equations (Table 1) for calibration standards prepared in salinities of 5, 10, 20 and 30‰ were co-linear with calibration standards prepared in deionized water (0‰), with gradients of 0.0006 AU per mg l21 PO4–P and blanks of 0.005 AU, showing that a single set of calibration Fig. 4 Graph showing the temperature dependence of the absorbance signal. The length of RC2 was 250 cm and the PO4–P concentration was 500 mg l21. Table 1 The effect of varying salinity (0–30‰) on the FI responses for orthophosphate standards over the range 0–100 mg l21 PO4–P. Results expressed as absorbance (arbitrary units) Salinity PO4–P/mg l21 0‰ 5‰ 10‰ 20‰ 30‰ 0 0.0050 0.0050 0.0050 0.0050 0.0050 20 0.0183 0.0187 0.0193 0.0193 0.0183 40 0.0310 0.0317 0.0313 0.0317 0.0320 60 0.0430 0.0430 0.0443 0.0430 0.0437 80 0.0547 0.0557 0.0560 0.0557 0.0555 100 0.0667 0.0673 0.0687 0.0677 0.0690 Calibration y = 0.0006x y = 0.0006x y = 0.0006x y = 0.0006x y = 0.0006x +0.0058 +0.0056 +0.0059 +0.0061 +0.0056 r2 0.9994 0.9992 0.9992 0.9990 0.9993 Fig. 5 Calibration graphs obtained with two different reaction coil lengths.A, 12 m; and B, 1.5 m. Analyst, December 1997, Vol. 122 1479standards can be used for the determination of orthophosphate in estuarine waters. A detection limit of 2 mg l21 PO4–P was calculated using the method described by Miller and Miller13 (n = 7). The calibration graphs were linear over the range 2–100 mg l21 PO4–P and r2 values were in the range 0.9990–0.9994. Method Validation Samples were collected from the Tamar Estuary, Devon, UK, filtered through a 0.45 mm filter and stored in a refrigerator until analysed.Comparative results were obtained using a reference ascorbic acid batch spectrophotometric method.14 The results for a transect of the estuary (Fig. 6) show that there was excellent agreement between the batch method and the salinity compensation FI manifold over the full range of salinities. Samples were also collected at a fixed location (Halton Quay, grid reference 41.2/65.4) over a 12 h period and the results for the complete tidal cycle are shown in Fig. 7. Conclusions The proposed rFI salinity compensation manifold using a refractive index-matched sulfuric acid carrier and ascorbic acid as the reducing agent is effective in suppressing the Schlieren effect. It gives a single analytical peak in the determination of DRP in estuarine waters with a detection limit of 2 mg l21 PO4– P. This manifold design has generic potential for FI determinations in samples of widely varying ionic strength.S.A. would like to thank the EU for providing a grant under the ERASMUS scheme as part of the ‘Analytical Chemistry’ network. The authors also thank D. Whitworth for his assistance with the field work. P.J.W. would like to thank NERC for research grant GST/02/669 to support initial studies in this area. References 1 Stumm, W., and Morgan, J. J., Aquatic Chemistry, Wiley, New York, 3rd edn., 1996, p. 891. 2 Andrew, K. N., Blundell, N. J., Price, D., and Worsfold, P.J., Anal. Chem., 1994, 66, 916A. 3 Robards, K., McKelvie, I. D., Benson, R. L., Worsfold, P. J., Blundell, N., and Casey, H., Anal. Chim. Acta, 1994, 287, 143. 4 McKelvie, I. D., Peat, D. M. W., Matthews, G. P., and Worsfold, P. J., Anal. Chim. Acta, 1997, 351, 265. 5 Ham, G., Anal. Proc., 1981, 18, 69. 6 Broberg, O., and Petterson, K., Hydrobiologia, 1988, 170, 45. 7 Murphy, J., and Riley, J. P., Anal. Chim. Acta, 1962, 27, 31. 8 Crouch, S. R., and Malmstadt, H. V., Anal.Chem., 1967, 39, 1084. 9 Ciavatta, C., Antisari, L. V., and Sequi, P., J. Environ. Qual., 1990, 19, 761. 10 Rodriguez, J. B., Self, J. R., and Soltanpour, P. N., Soil Sci. Soc. Am. J., 1994, 58, 866. 11 House, W. A., Leach, D., Warwick, M. S., Whitton, B. A., Pattinson, S. N., Ryland, G., Pinder, A., Ingram, J., Lishman, J. P., Smith, S. M., Rigg, E., and Denison, F. H., Sci. Total Environ., 1997, 194–195, 303. 12 Johnson, K. S., and Petty, R. L., Anal. Chem., 1982, 54, 1185. 13 Miller, J. C., and Miller, J. N., Statistics for Analytical Chemistry, Ellis Horwood, Chichester, 1988, p. 227. 14 American Public Health Association, Standard Methods for the Examination of Water and Wastewater, method 4500-P.E, 18th edn., APHA–AWWA–WEF, 1992, p. 108. Paper 7/05363K Received July 24, 1997 Accepted August 27, 1997 Fig. 6 Validation of the salinity compensation FI method. Comparison of DRP data obtained for transect samples from the Tamar Estuary, Devon, UK using the proposed FI salinity compensation manifold (error bars show ±3s, n = 3) and a standard batch reference method (RM) with a mean error of ±5 mg l21 PO4–P.Fig. 7 DRP data obtained for tidal cycle samples from the Tamar Estuary, Devon, UK using the proposed FI salinity compensation manifold (error bars show ±3s, n = 3). 1480 Analyst, December 1997, Vol. 122 Determination of Dissolved Reactive Phosphorus in Estuarine Waters Using a Reversed Flow Injection Manifold† Stefan Auflitschab, Darren M.W. Peata, Ian D. McKelviec and Paul J. Worsfold*a a Department of Environmental Sciences, Plymouth Environmental Research Centre, University of Plymouth, Drake Circus, Plymouth, Devon UK PL4 8AA b Institute of Analytical Chemistry, Vienna University of Technology, Getreidemarkt 9/151, A-1060 Vienna, Austria c Water Studies Centre, Chemistry Department, Monash University, Caulfield East, Victoria 3145, Australia The Schlieren or refractive index (RI) effect is a major problem in the determination of dissolved reactive phosphorus in estuarine waters using conventional flow injection (FI) manifolds with sample injection.This is because differences in RI between the injected sample zone and the carrier stream give rise to a lensing effect which is superimposed on the blank response and causes significant error in quantitation. A simple reversed flow injection (rFI) manifold using spectrophotometric detection which removes these quantitation errors is reported.Acidic molybdate is injected into a sulfuric acid carrier stream of the same refractive index and sequentially merged with sample and reductant (ascorbic acid). Reduction of phosphomolybdate to phosphomolybdenum blue is carried out in a coil thermostated at 60 °C. Dissolved reactive phosphorus has been successfully determined in estuarine waters with salinities ranging from 0 to 30‰ using calibration standards prepared in deionized water, with a detection limit of 2 mg l21 PO4–P and a linear range of 2–100 mg l21 PO4–P (r2 = 0.9998).Keywords: Dissolved reactive phosphorus; phosphate; salinity; refractive index; Schlieren effect; flow injection; estuarine water Phosphorus is an essential and often limiting nutrient for algal and bacterial growth1 and the total phosphorus content of natural waters comprises both particulate and dissolved forms, the latter being operationally defined as the fraction which passes through a 0.45 mm membrane. This total dissolved phosphorus (TDP) fraction can be further sub-divided into dissolved inorganic phosphorus (DIP) and dissolved organic phosphorus (DOP). The DIP fraction comprises orthophosphate and condensed phosphates with orthophosphate being the most frequently determined form of DIP and also referred to as dissolved reactive phosphorus (DRP).Flow injection (FI) with spectrophotometric detection has been widely reported for monitoring nutrients in natural waters.2,3 However, the application of conventional FI to the determination of DRP in estuarine waters has been very limited due to the so-called refractive index (RI) or Schlieren effect.4 This occurs when samples with differing ionic strengths are injected into a carrier stream of lower ionic strength.5 Under laminar flow conditions the geometry of the sample zone is parabolic and differences in refractive index between the sample zone and carrier stream lead to a lensing effect, in which the sample zone acts as a dispersing and collecting lens.As a result the Schlieren effect causes a negative frontal peak followed by a positive peak. Large errors in quantitation are caused by this effect, especially at low analyte concentrations. 4 When measuring DRP in seawater the RI problem can be overcome using a matched carrier stream (typically 39 g l21 NaCl). In estuarine waters, however, the salinity varies in the range 0–35‰ and a matrix matching approach would require the preparation of carrier solutions or calibration standards covering the whole range of salinities encountered, which is not a practical solution for routine laboratory work or field deployment.In a recent paper4 a reversed flow injection (rFI) manifold was reported for the determination of DRP in estuarine waters. This method overcame a number of difficulties encountered when conventional FI manifolds are used. Schlieren effects were eliminated by using a rFI manifold in which molybdate was injected into a NaCl RI-matched carrier stream which was continuously merged with the sample stream.The effect of varying sample chloride concentration on the efficiency of the tin(ii) chloride reduction step was masked by a swamping chloride concentration (60‰ as NaCl) in the carrier stream. This allowed one set of aqueous DRP standards to be universally applied over the salinity range 0–30‰. However, the analytical signal obtained was a doublet peak due to the formation of a pronounced acid concentration gradient in the injected molybdate zone.Kinetic inhibition of molybdenum blue complex formation was most severe in the centre of the sample zone where the acid concentration was highest. It was not possible to match both the RI and acid concentrations because the chloride concentration of the carrier stream was critical in order to avoid interference in the tin(ii) chloride reduction step.6 This effect also restricted the carrier stream to sample stream flow rate ratio that could be used (2.5 : 1) which limited the sensitivity due to on-line sample dilution.This paper describes the successful use of a new salinity compensation manifold that allows the direct determination of DRP in estuarine waters (0–35‰) using a single set of deionized water standards. Reduction was performed using ascorbic acid because this reaction is not influenced by the matrix chloride concentration.7 This allowed the acid concentration in the carrier stream and the injected molybdate reagent stream to be matched without susceptibility to chloride interference and resulted in a singlet analytical peak.In addition the freedom to choose a more favourable carrier stream to sample stream split ratio resulted in a much improved detection limit and sensitivity. These improvements make the method ideally suited to the rapid determination of DRP in estuarine waters, particularly at the lower end of the typical concentration range (5–50 mg l21 PO4–P) found in these waters.The † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997. Analyst, December 1997, Vol. 122 (1477–1480) 1477simplicity of the FI manifold and the calibration procedure facilitates the rapid determination of phosphate in estuarine waters and makes it ideal for shipboard deployment. Experimental Reagents Phosphate stock solution, 1000 mg l21 P as KH2PO4.Oven dried potassium dihydrogen orthophosphate (4.3937 g) was diluted to 1000 ml with deionized water, and stored at 4 °C. Working standards in the range 5–200 mg l21 PO4–P were prepared each day. Ammonium molybdate reagent. Ammonium molybdate [20 g; (NH4)6Mo7O24.4H2O] was dissolved in 700 ml of deionized water, 40 ml of concentrated sulfuric acid was added and the solution made up to 1000 ml. Ascorbic acid–antimony reagent. Ascorbic acid (20 g; C6H8O6) and 1 g of antimony potassium oxide (+)-tartrate 0.5 hydrate ( KSbO.C4H4O6.1/2H2O ) were dissolved in 700 ml of deionized water, 50 ml of concentrated sulfuric acid was added and the solution made up to 1000 ml.A fresh solution was prepared each day. Carrier solution. Concentrated sulfuric acid (56 ml) was added to 700 ml of deionized water and made up to 1000 ml. AsV and SiIV standards. Merck (Poole, Dorset, UK) standards containing 1000 mg l21 of SiIV and AsV were used to spike the phosphate standards.All solutions were prepared from BDH (Merck) AnalaR grade reagents using high purity Milli-Q (Millipore, Milford, MA, USA) water, and were sonicated for 15 min before use. The refractive indices of the ammonium molybdate and the carrier solutions were matched to within ±0.0005 with an Abb�e refractometer (Bellingham and Stanley, Tunbridge Wells, UK) by adding concentrated sulfuric acid to the carrier solution. Instrumentation and Procedures The manifold shown in Fig. 1 was used to determine DRP in standards and estuarine samples collected from the Tamar Estuary, Devon, UK. These samples were filtered (0.45 mm) in the laboratory within 6 h of collection and stored in a refrigerator for a maximum of 18 h prior to analysis. A Tecator 5020 FI analyser (Perstorp Analytical, Maidenhead, UK) and a Philips PU8620 Series UV/VIS/NIR (Philips, Cambridge, UK) single beam spectrophotometer were used to carry out the experiments. PTFE tubing of 0.5 mm id was used throughout.The flow cell used for the detection of phosphomolybdenum blue at 882 nm had an optical path length of 10 mm and an analytical volume of 18 ml. A Windows 95/NT Visual Basic application (Science Engineering Systems, Plymouth, UK) was used to control the instrument via a serial communication protocol and to store data from the instrument in a relational database. In this rFI manifold, ammonium molybdate solution was injected into a RI-matched sulfuric acid carrier and merged with the sample stream. The sample stream was changed manually.The formation of phosphomolybdic acid took place in a reaction coil, RC1, of 20 cm length. The phosphomolybdic acid was then reduced to phosphomolybdenum blue by ascorbic acid, with antimony acting as a catalyst, in a reaction coil, RC2, of 150 cm length. The main reaction coil (RC2) was immersed in a thermostated water-bath at 60 °C. Results and Discussion Optimisation of Molybdate Concentration To investigate the effect of molybdate concentration all other parameters were kept constant (as described in the Experimental section) and the molybdate concentration was varied over the range 1.7–40 g l21.Sample (200 mg l21 PO4–P) absorbance increased with molybdate concentration and was highest at 40 g l21, as shown in Fig. 2. However, at concentrations > 20 g l21 the gain in sensitivity was offset by the appearance of a trailing shoulder as shown in Fig. 3. The shape of this peak was variable and seriously degraded precision at lower phosphate concentrations. This is probably caused by direct reduction of the molybdate reagent8 at the high molybdate to acid concentration ratio which exists after merging with the sample stream and therefore a concentration of 20 g l21 was used for all subsequent studies. Interference From SiIV and AsV Sulfuric acid concentration strongly affects the formation kinetics of molybdenum heteropolyacid complexes with PO4 32, SiIV and AsV, with increasing acid concentration decreasing the rate of formation of all of these complexes.6 In the range 0.14–0.29 m H2SO4 the rate of formation of the PO4 32 complex is relatively stable9,10 but the formation rate of the SiIV and AsV complexes decreases dramatically at the higher acid concentra- Fig. 1 FI salinity compensation manifold. Fig. 2 Effect of ammonium molybdate concentration on the sample and blank responses.RC2 was 150 cm and the PO4–P concentration was 200 mg l21. Error bars represent ±3s (n = 3). A, sample; B, blank. Fig. 3 FI traces showing the effect of ammonium molybdate concentration on the peak profile. 1478 Analyst, December 1997, Vol. 122tion. In many estuarine waters SiIV is present at relatively high concentrations11 and hence the choice of acid concentration is a compromise between sensitivity and susceptibility to SiIV interference. This can be achieved using an acid concentration at the lower pH end of the acid stability plateau for orthophosphate.At this acidity (0.29 m) the rate of formation of phosphomolybdenum blue decreases only slightly compared with that at 0.14 m, whereas the rate of formation of silicomolybdenum blue and arsenomolybdenum blue decreases dramatically. In the rFI manifold reported here the optimum pH will be different from that for the batch method due to the dynamic nature of the system. In addition, the pH for the reduction step must be controlled by adjusting the pH of the ascorbic acid stream because the acidity of the molybdate and carrier streams has to be fixed in order to eliminate RI and acid gradient effects. The optimum acid concentration (0.9 m; 50 ml l21) gave a linear response over the range 0–100 mg l21 PO4–P [response (arbitrary units, AU) = 0.0004 concentration (mg l21 PO4–P) + 0.0026] with a r2 of 0.9994.Using this manifold, silicate interference does not occur below 1 mg l21 SiIV and over-estimation at 2 mg l21 SiIV is only equivalent to 3.5 mg l21 PO4–P, which is similar to the results reported for batch methods.6 Similarly, AsV showed no interference up to 5 mg l21, which is much higher than the level normally encountered in estuarine waters.However, 10 mg l21 AsV resulted in a phosphate overestimation of 5 mg l21 P. Optimisation of Phosphomolybdate Reduction The strong temperature dependence of the overall reaction is shown in Fig. 4, with the detector response rising from 0.14 AU at room temperature to 0.18 AU at 60 °C for a 500 mg l21 PO4–P standard. The blank signal concomitantly increased from 0.002 to only 0.004 AU and there was no degradation in precision (n = 3) at the higher temperature. Use of a heated reaction coil was necessary in order to increase the rate of reduction induced by ascorbic acid, because of the non-equilibrium conditns that exist in the FI manifold.12 Bubble formation in the reaction coil occurred at 70 °C, and although the sensitivity was improved at this temperature this was negated by a large increase in the blank response and a decrease in reproducibility.Operation of the manifold was therefore optimal when the reactor coil was thermostated at 60 °C. The length of RC1 had little effect on the overall performance of the manifold and therefore was set at 20 cm to facilitate a high sample throughput. Increasing the length of RC2 increased the dispersion but the absorbance still increased because of additional reaction time for the ascorbic acid reduction step.Fig. 5 shows that the calibration graph was sigmoidal at a coil length of 12.0 m whereas a coil length of 1.5 m gave a linear calibration (r2 = 0.9994) over the range 0–100 mg l21 PO4–P and was therefore used for all subsequent studies. Intermediate coil lengths showed increasingly sigmoidal behaviour. PTFE tubing (0.5 mm id) was also used throughout to minimise dispersion.Analytical Figures of Merit The effectiveness of this new salinity compensation manifold was demonstrated by comparing the calibration data for a series of PO4–P standards (5–100 mg l21) prepared at differing salinities (0–30‰). Regression equations (Table 1) for calibration standards prepared in salinities of 5, 10, 20 and 30‰ were co-linear with calibration standards prepared in deionized water (0‰), with gradients of 0.0006 AU per mg l21 PO4–P and blanks of 0.005 AU, showing that a single set of calibration Fig. 4 Graph showing the temperature dependence of the absorbance signal. The length of RC2 was 250 cm and the PO4–P concentration was 500 mg l21. Table 1 The effect of varying salinity (0–30‰) on the FI responses for orthophosphate standards over the range 0–100 mg l21 PO4–P. Results expressed as absorbance (arbitrary units) Salinity PO4–P/mg l21 0‰ 5‰ 10‰ 20‰ 30‰ 0 0.0050 0.0050 0.0050 0.0050 0.0050 20 0.0183 0.0187 0.0193 0.0193 0.0183 40 0.0310 0.0317 0.0313 0.0317 0.0320 60 0.0430 0.0430 0.0443 0.0430 0.0437 80 0.0547 0.0557 0.0560 0.0557 0.0555 100 0.0667 0.0673 0.0687 0.0677 0.0690 Calibration y = 0.0006x y = 0.0006x y = 0.0006x y = 0.0006x y = 0.0006x +0.0058 +0.0056 +0.0059 +0.0061 +0.0056 r2 0.9994 0.9992 0.9992 0.9990 0.9993 Fig. 5 Calibration graphs obtained with two different reaction coil lengths. A, 12 m; and B, 1.5 m. Analyst, December 1997, Vol. 122 1479standards can be used for the determination of orthophosphate in estuarine waters.A detection limit of 2 mg l21 PO4–P was calculated using the method described by Miller and Miller13 (n = 7). The calibration graphs were linear over the range 2–100 mg l21 PO4–P and r2 values were in the range 0.9990–0.9994. Method Validation Samples were collected from the Tamar Estuary, Devon, UK, filtered through a 0.45 mm filter and stored in a refrigerator until analysed. Comparative results were obtained using a reference ascorbic acid batch spectrophotometric method.14 The results for a transect of the estuary (Fig. 6) show that there was excellent agreement between the batch method and the salinity compensation FI manifold over the full range of salinities. Samples were also collected at a fixed location (Halton Quay, grid reference 41.2/65.4) over a 12 h period and the results for the complete tidal cycle are shown in Fig. 7. Conclusions The proposed rFI salinity compensation manifold using a refractive index-matched sulfuric acid carrier and ascorbic acid as the reducing agent is effective in suppressing the Schlieren effect.It gives a single analytical peak in the determination of DRP in estuarine waters with a detection limit of 2 mg l21 PO4– P. This manifold design has generic potential for FI determinations in samples of widely varying ionic strength. S.A. would like to thank the EU for providing a grant under the ERASMUS scheme as part of the ‘Analytical Chemistry’ network.The authors also thank D. Whitworth for his assistance with the field work. P.J.W. would like to thank NERC for research grant GST/02/669 to support initial studies in this area. References 1 Stumm, W., and Morgan, J. J., Aquatic Chemistry, Wiley, New York, 3rd edn., 1996, p. 891. 2 Andrew, K. N., Blundell, N. J., Price, D., and Worsfold, P. J., Anal. Chem., 1994, 66, 916A. 3 Robards, K., McKelvie, I. D., Benson, R. L., Worsfold, P. J., Blundell, N., and Casey, H., Anal. Chim. Acta, 1994, 287, 143. 4 McKelvie, I. D., Peat, D. M. W., Matthews, G. P., and Worsfold, P. J., Anal. Chim. Acta, 1997, 351, 265. 5 Ham, G., Anal. Proc., 1981, 18, 69. 6 Broberg, O., and Petterson, K., Hydrobiologia, 1988, 170, 45. 7 Murphy, J., and Riley, J. P., Anal. Chim. Acta, 1962, 27, 31. 8 Crouch, S. R., and Malmstadt, H. V., Anal. Chem., 1967, 39, 1084. 9 Ciavatta, C., Antisari, L. V., and Sequi, P., J. Environ. Qual., 1990, 19, 761. 10 Rodriguez, J. B., Self, J. R., and Soltanpour, P. N., Soil Sci. Soc. Am. J., 1994, 58, 866. 11 House, W. A., Leach, D., Warwick, M. S., Whitton, B. A., Pattinson, S. N., Ryland, G., Pinder, A., Ingram, J., Lishman, J. P., Smith, S. M., Rigg, E., and Denison, F. H., Sci. Total Environ., 1997, 194–195, 303. 12 Johnson, K. S., and Petty, R. L., Anal. Chem., 1982, 54, 1185. 13 Miller, J. C., and Miller, J. N., Statistics for Analytical Chemistry, Ellis Horwood, Chichester, 1988, p. 227. 14 American Public Health Association, Standard Methods for the Examination of Water and Wastewater, method 4500-P.E, 18th edn., APHA–AWWA–WEF, 1992, p. 108. Paper 7/05363K Received July 24, 1997 Accepted August 27, 1997 Fig. 6 Validation of the salinity compensation FI method. Comparison of DRP data obtained for transect samples from the Tamar Estuary, Devon, UK using the proposed FI salinity compensation manifold (error bars show ±3s, n = 3) and a standard batch reference method (RM) with a mean error of ±5 mg l21 PO4–P. Fig. 7 DRP data obtained for tidal cycle samples from the Tamar Estuary, Devon, UK using the proposed FI salinity compensation manifold (error bars show ±3s, n = 3). 1480 Analyst, December 1997, Vol. 122

ISSN:0003-2654    

DOI:10.1039/a705363k

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Ambient Volatile Organic Compound Monitoring by Diffusive Sampling. Compatibility of High Uptake Rate Samplers With Thermal Desorption† Matthew Batesa,b, Norbert Gonzalez-Flesca*a, Vincenzo Cocheoc, and Ranjeet Sokhib a INERIS, B.P.2, Parc Technologique Alata, 60550 Verneuil en Halatte, France b University of Hertfordshire, Hatfield, Herts. UK c Fondazione Salvatore Maugeri, Padova, Italy Field and laboratory validation studies were carried out on a novel, high uptake rate, radial diffusive sampler loaded with Carbotrap, a thermally desorbable adsorbent.For field experiments parallel techniques were employed for comparison. The goal was to assess the compatibility of the device for subsequent thermal desorption and preconcentration of the sample for analysis. Some unexpected results were obtained especially for benzene, where it was observed that, under certain conditions the longer the device was exposed to an atmosphere, the less sample was collected.A possible explanation for these results is put forward by applying knowledge previously acquired on saturation and competitivity effects. An alteration to the sampler geometry is proposed which, through further laboratory and field work, is shown to overcome this problem. Keywords: Diffusive sampling; ambient air; volatile organic compounds; high uptake rate; thermal desorption Certain volatile organic compounds (VOC) have long been recognised as direct hazards to human health1–4 and, within the scope of industrial hygiene, diffusive samplers are often used to measure personal exposure at the work-place.5–7 More recently however, since VOC have been recognised as tropospheric ozone precursors and as indoor air pollutants (coming from modern furniture, paints, cigarette smoking, etc.), it has become necessary to monitor their levels inside buildings8,9 and in ambient air.10–12 Diffusive sampling is a well established and common technique for indoor air monitoring.Many different techniques are available for ambient VOC monitoring (such as passivated canisters, on-line chromatography and spectrometric techniques) however for expensive multi-site campaigns where detailed concentration mapping is required the comparitively low cost of diffusive samplers makes them an indispensible tool. For example to efficiently map a medium sized town of Å 250 000 inhabitants, 100 sampling sites would be necessary, this can only be achieved with low cost equipment.A passive sampler may be defined13 as a device which is capable of taking samples of gas or vapour pollutants from the atmosphere at a rate controlled by a physical process such as diffusion through a static air layer or permeation through a membrane, but which does not involve the active movement of air through the sampler. Passive samplers were originally designed to measure concentrations at ppm levels over an eight hour period in a work-place atmosphere; however for indoor or ambient air applications their detection limits need to be below 1 ppb for target compounds.Conventional passive samplers fall into two main geometrical categories: tube-type and badge-type devices. Tube-type samplers are characterized by their relatively low sampling rates due to a long axial diffusion path compared to a low crosssectional area, deep adsorbent bed. Badge-type samplers typically have higher uptake rates due to the combination of a shorter diffusion path-length with a thin, greater cross sectional area adsorbent bed.Two different desorption techniques exist for these devices; solvent and thermal, but due to physical constraints badge type, and hence all conventional high uptake rate samplers, have until recently been incompatible with the thermal option. A relatively new high uptake rate device (commercially available as Radiello) can be used in conjunction with both options.14 However as we will go on to show, care must be taken when using these devices with subsequent thermal desorption.The aim of this work is to demonstrate and attempt to explain the difficulties encountered when trying to quantitatively recover certain VOC from ambient air using high uptake rate diffusive samplers. Having identified the nature of these problems an alternative arrangement is suggested and its performance assessed. Theory In simple terms a diffusive sampler consists of an adsorbent bed separated from a sampler opening by a zone of still air across which matter is transferred down an induced concentration gradient.Axial Diffusive Sampler A schematic representation of an axial (conventional badge or tube-type) diffusive sampler is given in Fig. 1. Fick’s first law may be used to describe the rate of mass transfer down the concentration gradient induced in the diffusion air gap: d d d d m t DA C x = (1) where: dm/dt = rate of mass transfer; D = diffusion coefficient; A = cross-sectional surface area, and dC/dx = concentration gradient along the axis. Integration of eqn. 1 with the parameters defined in Fig. 1 gives: † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997. Fig. 1 Schematic of an axial diffusion path sampler. Analyst, December 1997, Vol. 122 (1481–1484) 1481m t DA L C C t = - ( ) o a (2) where: mt = mass transferred in time t; t = sampling duration; L = diffusion path-length; Co = ambient concentration; and Ca = concentration just above adsorbent surface.The collection of parameters DA/L for a given device and compound at a given temperature is often referred to as the uptake rate (cm3 s21) and is useful when describing sampler performance. Radial Diffusive Samplers A schematic diagram of a radial sampler is given in Fig. 2. Again Fick’s first Law may be used to describe the mass transfer across the still air zone. In cylindrical coordinates this gives eqn. 3: d d d d m t D rh C r = 2p (3) where: r = radius; h = length of the cylinder, and dC/dr = concentration gradient along the radius. which when integrated using the parameters defined in Fig. 2 gives: m t D h C C r r t = - 2p o a o a ln / (4) where: ro = radius of outer cylinder, and ra = radius of inner cylinder (adsorbent cartridge). and the uptake rate is now represented by: D h r r 2p ln / o n It is important to note that the theoretical uptake rate for benzene on Radiello is more than two orders of magnitude greater than that for commercially available axial tubes and, as a consequence, limits of detection are much improved.Nevertheless, as we will go on to show, this benefit can be overshadowed by the incompatibility of thermally desorbable adsorbents and elevated uptake rates. Experimental. Part 1 The preliminary experimental work involved firstly the assessment of recovery coefficients for spiked samples. This was performed on 3.9 mm od (Supelco compatible) Radiello cartridges loaded with Carbotrap-B adsorbent.Then four field experiments were conducted to assess sampler performance using parallel techniques for comparison. Material and Equipment Adsorbent For all diffusive and active sampling, the adsorbent used was Carbotrap-B 20/40 mesh (Supelco, Bellefonte, PA, USA) a high purity graphitized carbon black particularly suitable for sampling C4–C8 organic compounds.15 It is compatible with thermal desorption exhibiting good thermal stability up to 400 °C and effective desorption of adsorbed compounds between 300—340 °C.16 Adsorbent cartridges Stainless steel mesh cylinders [100-mesh hole size; 0.1 mm wire diameter; 60 mm long; od 3.9 mm (Supelco glass tube compatible, filled with 125 mg of Carbotrap B) or 4.8 mm (Perkin Elmer tube compatible, filled with 150 mg of Carbotrap B) were loaded with Carbotrap and sealed at both ends by stainless steel mesh caps (from Radiello, Fondazione Salvatore Maugeri, Centro di Ricerche Ambientali, Padova, Italy). Analytical and sample preparation apparatus Dynatherm Model 890, thermal desorption unit (TDU) and Dynatherm tube conditioner Model 60 (Supelco).An automatic thermal desorption apparatus (ATD400; Perkin-Elmer, Beaconsfield, Buckinghamshire, UK). Chrompack CP 9002 GC– FID instrument and Chrompack CP-Sil-5CB column (25 m 3 0.25 mm od Film thickness, 0.4 mm; Chrompack, Middelburg, The Netherlands). Determination of Recovery Coefficients For conditioning and analysis an adsorbent cartridge must be placed inside an auxiliary tube if existing desorption equipment is to be used.To determine recovery coefficients, a standard solution of benzene, toluene and xylenes (BTX) was prepared in methanol. Different volumes of this solution were then evaporated using a TDU and adsorbed from the vapour phase on Radiello cartridges. Tubes were analysed with the equipment described above (TDU; 300 °C/7 min/14 ml min21; split flow: 12.1 ml min21; cryofocus: 290 °C; GC: analytical program 35–105 °C (2.5 °C min21) then 105–235 °C (10 °C min21); column flow 1.9 ml min21.Results are shown for benzene, toluene and m/p xylene in Fig. 3. From Fig. 3 it can be seen that a linear and 100% desorption efficiency was obtained for these compounds at these concentrations. It must be noted that this experiment in dynamic spiking of Radiello cartridges was carried out solely to verify that the combination of a cartridge with an auxiliary tube could be easily handled and that matter could be quantitively recovered by thermal desorption.Field Experiments A series of four field experiments of varying duration and under various conditions were carried out to assess the Radiello sampler performance with respect to parallel equivalent techniques. Fig. 2 Schematic of a radial diffusive sampler. Fig. 3 Recovery of spiked benzene and m/p-xylene on Radiello cartridges. 1482 Analyst, December 1997, Vol. 122Conditions (1) Seven day duration. Suburban site. Average temperature, 7 °C. Average relative humidity, 70%. Three Radiello samplers (3.9 mm od cartridges). Parallel technique: two INERIS tubes (Supelco compatible axial diffusive sampling tubes designed, manufactured and assessed at INERIS, Verneuil en Halatte). (2) Six day duration. Suburban site. Average temperature, 9 °C. Average relative humidity, 65%. Three Radiello samplers (3.9 mm od cartridges).Parallel technique: two INERIS tubes. (3) Eleven hour duration. Urban site. Average temperature, 16 °C. Average relative humidity, 31%. Three Radiello samplers (3.9 mm od cartridges). Parallel technique: one passivated canister (Restek, Bellefonte, PA, USA). (4) Six hour duration. Suburban site. Average temperature, 10 °C. Average relative humidity, 75%. Three Radiello samplers (3.9 mm od cartridges). Parallel technique: six RPE (Radiello Perkin Elmer cf.Experimental Part 2) tubes used consecutively in active sampling mode (using a low flow personal air sampler, Gilian Instrument, NJ, USA. Flow rate 40 ml min21). All results are shown in Table 1. Discussion From Table 1 the most striking observation to be made is that after seven days exposure no benzene was found on the Radiello cartridges (even though it was detected by the parallel technique), after six days a fifth of the expected amount was observed, after eleven hours exposure to an urban atmosphere, approximately half of the expected amount was recovered and finally after six hours exposure (with an overall uptake of approximately one tenth that of the previous three experiments) to a suburban atmosphere, the expected and observed amounts showed good agreement.Hence, as far as benzene is concerned, it can be observed that, the longer one exposes this device to an atmosphere, the less sample one collects. This somewhat unexpected result for benzene recovery could be explained as follows.From previous work17,18 it is accepted that many VOC exhibit Langmuirian behaviour with respect to their adsorption on solid adsorbents. For an individual VOC at equilibrium, the mass of adsorbate per unit mass of the adsorbent mads, is given by: m bcm l bc ads = + max (5) where: b = distribution coefficient (adsorption to desorption equilibrium coefficient ratio). c = gaseous phase VOC concentration, and mmax = maximum adsorbable mass.In the case of a VOC mixture, it has been demonstrated that the mass of a given adsorbate is dependent on the nature and amount of the other constituents in the mixture. Mass adsorbed mj for a compound j in a mixture of i compounds can be written as follows: m b c m b c j j j i i j = +å max 1 (6) where: ·bici, indicates that the sum of the product must be performed for all compounds. Eqn. 6 quantifies to some extent the competitive adsorption effects between different VOC.The adsorption parameters and hence the affinity for different BTX have been determined on Tenax–GC (Supelco). It was shown that affinity increases with the degree of alkylation of the aromatic ring. So, at equilibrium the relative masses adsorbed follow the trend: mbenzene <mtoluene <mxylene From as yet unpublished work, it can be shown that Carbotrap- B exhibits a similar behaviour. In the case of these field experiments, one can expect some competitive effects between benzene and other aromatics, especially if one considers that in areas polluted by automotive traffic the atmospheric toluene concentration, for example, is often three times that of benzene.Having identified this problem of competitivity at equilibrium on Carbotrap two possible options were open. Either to use the most common configuration of Radiello with a stronger adsorbent, such as activated charcoal, and revert to solvent desorption. Or, to stay with thermal desorption and reduce the effective uptake rate of the Radiello cartridge by altering its host sampler geometry.The latter choice was taken and the performance assessed by a series of laboratory and field experiments. Experimental. Part 2 An empty Perkin Elmer tube was fitted with a 4.8 mm od Radiello cartridge (see Fig. 4) previously loaded with Å 150 mg of Carbotrap-B held in place by a retaining spring. Diffusion path length can be precisely controlled by the positioning of the Radiello cartridge front-end cap.For the following experiments the front-end cap was inserted 4 mm bringing the total diffusion path length to 21 mm. As earlier, the preliminary laboratory work involved the assessment of the recovery of spiked samples from the new sampler arrangement. Experiments were conducted in the same manner and, as before linear, 100% desorption efficiencies were observed for all BTX. Next, in order to validate theoretical uptake rates, four RPE were exposed for six hours to high concentrations of BTX in the exposure chamber of a dynamic atmosphere generator.19 The atmosphere was controlled using classic dynamic sampling tubes and the results are shown in Fig. 5. Even though the concentrations used in this experiment are much higher than those encountered in ambient air, no saturation or competitive effects were observed. It must be Table 1 Comparison of measured values by Radiello cartridges and parallel techniques* Measured atmospheric concentrations/mg m23 Duration Device Benzene Toluene m/p-Xylene o-Xylene 7 d Axial tubes 1.9 8.1 4.4 1.4 Radiello 0.0 6.0 4.3 1.3 6 d Axial tubes 1.7 5.3 2.7 0.9 Radiello 0.4 4.8 3.0 1.0 11 h Canister 16.3 59.2 24.3 8.6 Radiello 7.2 55.5 27.0 9.7 6 h Active tubes 2.8 8.3 2.6 1.0 Radiello 2.8 8.5 3.1 1.2 * Diffusion coefficients taken from the literature.20 Fig. 4 Schematic of the Radiello Perkin Elmer configuration. Analyst, December 1997, Vol. 122 1483noted that the total uptake in this experiment represents only a fraction (1%) of the uptake on the radial mode Radiellos during the eleven hour field experiment.The maximum RSD observed was 1.5% and there was very little deviation from theoretial uptake rates. Field Experiments As a final performance assessment, six RPE tubes were exposed alongside six INERIS tubes in the field (under the following conditions: seven days exposure. Suburban site. Average temperature, 14 °C.Average relative humidity, ND). The results are shown in Table 2. The results show little dispersion even at low atmospheric concentrations. Benzene is well recovered as predicted by laboratory experiments. It is also interesting to note, that in both cases, measured concentrations are closely matched for the devices, even though theoretical uptake rates are not the same (due to differing geometries). This suggests that no competition problems were encountered and supports the assumption made earlier of a linear uptake rate throughout the sampling period for these lower uptake rate devices.This conformity with theoretical expectations of sampler behaviour is due to the control that is now possessed over the diffusion path length, and hence the uptake rate of the Radiello cartridge in this axial configuration. Conclusions Diffusive samplers are undoubtedly very practical tools for air quality monitoring expecially if extensive campaigns or personal monitoring are planned.If correctly used, high uptake rate devices allow the measurement of low concentrations of VOC even for short duration exposure. The relatively new radial sampler, Radiello, offers the possibility of solvent or thermal desorption. However under certain conditions it has been shown that the high uptake rate of this device can cause recovery problems for some compounds when using thermally desorbable adsorbents, such as Carbotrap. The unexpected results may be explained with the help of previously published work.The solution proposed here combines the versatility of Radiello cartridges with a popular auxiliary tube. This arrangement including the adjustable diffusion path length gives new possibilities to commercial products already on the market and allows one to cover almost any sampling scenario with just one analytical chain. Acknowledgements A. Frezier for her technical assistance and Sigma Aldrich, France, for their kind donation of equipment.References 1 Guicherit, R., and Schulting, F. L., Sci. Total Environ., 1985, 43, 193. 2 Tancrede, M., Wilson, R., Zeise, L., and Crouch, E. A. C., Atmos. Environ., 1987, 21, 2187. 3 Agency for Toxic Substances and Disease Registry, US Public Health ATSDR PB93-182-384, Atlanta GA (1993). 4 Snyder, R., Witz, G., and Goldstein, B. D., Environ. Health Perspect., 1993, 100, 293. 5 Brown, R. H., Charlton, J., and Saunders, K. J., Am. Ind. Hyg. Assoc.J., 1981, 42, 865. 6 Tompkins, F. C., Jr., and Goldsmith, R. L., Am. Ind. Hyg. Assoc. J., 1977, 38, 371. 7 Palmes, E. D., and Gunnison, A. F., Am. Ind. Hyg. Assoc. J., 1973, 34, 78. 8 Brown, V. M., Crump, D. R., and Gardiner, D., Environ. Technol., 1992, 13, 367. 9 Brown, V. M., Crump, D. R., Gardiner, D., and Yu, C. W. F., Environ. Technol., 1993, 14, 771. 10 M�egie, G., Bonte, J., Carlier, P., Chavaudra, J., Dizengremel, P., Feugier, A., Granier, C., Hauglustaine, D., Kanakidou, M., Le Bras, G., Marenco, A., Mouvier, G., Tissot, B., Toupance, G., and Truhau, R., Rapp.l’Acad. Sci., 1993, 30, 10. 11 Derwent, R. G., Middleton, D. R., Field, R. A., Goldstone, M. E., Lester, J. N., and Perry, R., Atmos. Environ. 1995, 29, 923. 12 Carler, R. E., Thomas, M. J., Marotz, G. A., Lane, D. D., and Huson, J. I., Environ. Sci. Technol., 1992, 26, 2175. 13 Brown, R. H., Harvey, R. P., Purnall, C. J., and Saunders, K. J., Am. Ind. Hyg. Assoc. J., 1984, 45, 67. 14 Cocheo, V., Boaretto, C., and Sacco, P., Am. Ind. Hyg. Assoc. J., 1996, 57, 897. 15 Supelco GC bulletin 846B, Supelco Inc, Bellefonte, USA, 1986. 16 Cao, X. L., and Hewitt, C. N., Atmos. Environ. 1993, 27A, 1865. 17 Comes, P., Gonzalez-Flesca, N., M�enard, T., and Grialt, J. O., Anal. Chem., 1993, 65, 1048. 18 Comes, P., Gonzalez-Flesca, N., Bader, F., and Grimalt, J. O., J. Chromatogr. A, 1996, 723, 293. 19 Jaouen, P., Gonzalez-Flesca, N., and Carlier, P., Environ. Sci. Technol., 1995, 29, 2718. 20 Lugg, G. A., Anal. Chem., 1968, 40, 1072. Paper 7/05610I Received August 4, 1997 Accepted October 31, 1997 Fig. 5 Concentration comparison for dynamic and passive measurements. Table 2 Results from parallel exposure of INERIS and RPE tubes Measured atmospheric concentrations/mg m23 (% RSD of six results) Device Benzene Toluene m/p-Xylene o-Xylene INERIS 2.7 8.2 3.9 1.4 tubes (4) (3) (5) (5) RPE 2.8 8.4 3.8 1.4 tubes (3) (2) (3) (4) 1484 Analyst, December 1997, Vol. 122 Ambient Volatile Organic Compound Monitoring by Diffusive Sampling. Compatibility of High Uptake Rate Samplers With Thermal Desorption† Matthew Batesa,b, Norbert Gonzalez-Flesca*a, Vincenzo Cocheoc, and Ranjeet Sokhib a INERIS, B.P.2, Parc Technologique Alata, 60550 Verneuil en Halatte, France b University of Hertfordshire, Hatfield, Herts. UK c Fondazione Salvatore Maugeri, Padova, Italy Field and laboratory validation studies were carried out on a novel, high uptake rate, radial diffusive sampler loaded with Carbotrap, a thermally desorbable adsorbent.For field experiments parallel techniques were employed for comparison. The goal was to assess the compatibility of the device for subsequent thermal desorption and preconcentration of the sample for analysis. Some unexpected results were obtained especially for benzene, where it was observed that, under certain conditions the longer the device was exposed to an atmosphere, the less sample was collected.A possible explanation for these results is put forward by applying knowledge previously acquired on saturation and competitivity effects. An alteration to the sampler geometry is proposed which, through further laboratory and field work, is shown to overcome this problem. Keywords: Diffusive sampling; ambient air; volatile organic compounds; high uptake rate; thermal desorption Certain volatile organic compounds (VOC) have long been recognised as direct hazards to human health1–4 and, within the scope of industrial hygiene, diffusive samplers are often used to measure personal exposure at the work-place.5–7 More recently however, since VOC have been recognised as tropospheric ozone precursors and as indoor air pollutants (coming from modern furniture, paints, cigarette smoking, etc.), it has become necessary to monitor their levels inside buildings8,9 and in ambient air.10–12 Diffusive sampling is a well established and common technique for indoor air monitoring.Many different techniques are available for ambient VOC monitoring (such as passivated canisters, on-line chromatography and spectrometric techniques) however for expensive multi-site campaigns where detailed concentration mapping is required the comparitively low cost of diffusive samplers makes them an indispensible tool. For example to efficiently map a medium sized town of Å 250 000 inhabitants, 100 sampling sites would be necessary, this can only be achieved with low cost equipment. A passive sampler may be defined13 as a device which is capable of taking samples of gas or vapour pollutants from the atmosphere at a rate controlled by a physical process such as diffusion through a static air layer or permeation through a membrane, but which does not involve the active movement of air through the sampler.Passive samplers were originally designed to measure concentrations at ppm levels over an eight hour period in a work-place atmosphere; however for indoor or ambient air applications their detection limits need to be below 1 ppb for target compounds.Conventional passive samplers fall into two main geometrical categories: tube-type and badge-type devices. Tube-type samplers are characterized by their relatively low sampling rates due to a long axial diffusion path compared to a low crosssectional area, deep adsorbent bed. Badge-type samplers typically have higher uptake rates due to the combination of a shorter diffusion path-length with a thin, greater cross sectional area adsorbent bed.Two different desorption techniques exist for these devices; solvent and thermal, but due to physical constraints badge type, and hence all conventional high uptake rate samplers, have until recently been incompatible with the thermal option. A relatively new high uptake rate device (commercially available as Radiello) can be used in conjunction with both options.14 However as we will go on to show, care must be taken when using these devices with subsequent thermal desorption.The aim of this work is to demonstrate and attempt to explain the difficulties encountered when trying to quantitatively recover certain VOC from ambient air using high uptake rate diffusive samplers. Having identified the re of these problems an alternative arrangement is suggested and its performance assessed. Theory In simple terms a diffusive sampler consists of an adsorbent bed separated from a sampler opening by a zone of still air across which matter is transferred down an induced concentration gradient. Axial Diffusive Sampler A schematic representation of an axial (conventional badge or tube-type) diffusive sampler is given in Fig. 1. Fick’s first law may be used to describe the rate of mass transfer down the concentration gradient induced in the diffusion air gap: d d d d m t DA C x = (1) where: dm/dt = rate of mass transfer; D = diffusion coefficient; A = cross-sectional surface area, and dC/dx = concentration gradient along the axis.Integration of eqn. 1 with the parameters defined in Fig. 1 gives: † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997. Fig. 1 Schematic of an axial diffusion path sampler. Analyst, December 1997, Vol. 122 (1481–1484) 1481m t DA L C C t = - ( ) o a (2) where: mt = mass transferred in time t; t = sampling duration; L = diffusion path-length; Co = ambient concentration; and Ca = concentration just above adsorbent surface.The collection of parameters DA/L for a given device and compound at a given temperature is often referred to as the uptake rate (cm3 s21) and is useful when describing sampler performance. Radial Diffusive Samplers A schematic diagram of a radial sampler is given in Fig. 2. Again Fick’s first Law may be used to describe the mass transfer across the still air zone.In cylindrical coordinates this gives eqn. 3: d d d d m t D rh C r = 2p (3) where: r = radius; h = length of the cylinder, and dC/dr = concentration gradient along the radius. which when integrated using the parameters defined in Fig. 2 gives: m t D h C C r r t = - 2p o a o a ln / (4) where: ro = radius of outer cylinder, and ra = radius of inner cylinder (adsorbent cartridge). and the uptake rate is now represented by: D h r r 2p ln / o n It is important to note that the theoretical uptake rate for benzene on Radiello is more than two orders of magnitude greater than that for commercially available axial tubes and, as a consequence, limits of detection are much improved.Nevertheless, as we will go on to show, this benefit can be overshadowed by the incompatibility of thermally desorbable adsorbents and elevated uptake rates. Experimental. Part 1 The preliminary experimental work involved firstly the assessment of recovery coefficients for spiked samples.This was performed on 3.9 mm od (Supelco compatible) Radiello cartridges loaded with Carbotrap-B adsorbent. Then four field experiments were conducted to assess sampler performance using parallel techniques for comparison. Material and Equipment Adsorbent For all diffusive and active sampling, the adsorbent used was Carbotrap-B 20/40 mesh (Supelco, Bellefonte, PA, USA) a high purity graphitized carbon black particularly suitable for sampling C4–C8 organic compounds.15 It is compatible with thermal desorption exhibiting good thermal stability up to 400 °C and effective desorption of adsorbed compounds between 300—340 °C.16 Adsorbent cartridges Stainless steel mesh cylinders [100-mesh hole size; 0.1 mm wire diameter; 60 mm long; od 3.9 mm (Supelco glass tube compatible, filled with 125 mg of Carbotrap B) or 4.8 mm (Perkin Elmer tube compatible, filled with 150 mg of Carbotrap B) were loaded with Carbotrap and sealed at both ends by stainless steel mesh caps (from Radiello, Fondazione Salvatore Maugeri, Centro di Ricerche Ambientali, Padova, Italy).Analytical and sample preparation apparatus Dynatherm Model 890, thermal desorption unit (TDU) and Dynatherm tube conditioner Model 60 (Supelco). An automatic thermal desorption apparatus (ATD400; Perkin-Elmer, Beaconsfield, Buckinghamshire, UK). Chrompack CP 9002 GC– FID instrument and Chrompack CP-Sil-5CB column (25 m 3 0.25 mm od Film thickness, 0.4 mm; Chrompack, Middelburg, The Netherlands).Determination of Recovery Coefficients For conditioning and analysis an adsorbent cartridge must be placed inside an auxiliary tube if existing desorption equipment is to be used. To determine recovery coefficients, a standard solution of benzene, toluene and xylenes (BTX) was prepared in methanol. Different volumes of this solution were then evaporated using a TDU and adsorbed from the vapour phase on Radiello cartridges. Tubes were analysed with the equipment described above (TDU; 300 °C/7 min/14 ml min21; split flow: 12.1 ml min21; cryofocus: 290 °C; GC: analytical program 35–105 °C (2.5 °C min21) then 105–235 °C (10 °C min21); column flow 1.9 ml min21.Results are shown for benzene, toluene and m/p xylene in Fig. 3. From Fig. 3 it can be seen that a linear and 100% desorption efficiency was obtained for these compounds at these concentrations. It must be noted that this experiment in dynamic spiking of Radiello cartridges was carried out solely to verify that the combination of a cartridge with an auxiliary tube could be easily handled and that matter could be quantitively recovered by thermal desorption.Field Experiments A series of four field experiments of varying duration and under various conditions were carried out to assess the Radiello sampler performance with respect to parallel equivalent techniques. Fig. 2 Schematic of a radial diffusive sampler. Fig. 3 Recovery of spiked benzene and m/p-xylene on Radiello cartridges. 1482 Analyst, December 1997, Vol. 122Conditions (1) Seven day duration. Suburban site. Average temperature, 7 °C. Average relative humidity, 70%. Three Radiello samplers (3.9 mm od cartridges). Parallel technique: two INERIS tubes (Supelco compatible axial diffusive sampling tubes designed, manufactured and assessed at INERIS, Verneuil en Halatte). (2) Six day duration. Suburban site. Average temperature, 9 °C. Average relative humidity, 65%.Three Radiello samplers (3.9 mm od cartridges). Parallel technique: two INERIS tubes. (3) Eleven hour duration. Urban site. Average temperature, 16 °C. Average relative humidity, 31%. Three Radiello samplers (3.9 mm od cartridges). Parallel technique: one passivated canister (Restek, Bellefonte, PA, USA). (4) Six hour duration. Suburban site. Average temperature, 10 °C. Average relative humidity, 75%. Three Radiello samplers (3.9 mm od cartridges). Parallel technique: six RPE (Radiello Perkin Elmer cf.Experimental Part 2) tubes used consecutively in active sampling mode (using a low flow personal air sampler, Gilian Instrument, NJ, USA. Flow rate 40 ml min21). All results are shown in Table 1. Discussion From Table 1 the most striking observation to be made is that after seven days exposure no benzene was found on the Radiello cartridges (even though it was detected by the parallel technique), after six days a fifth of the expected amount was observed, after eleven hours exposure to an urban atmosphere, approximately half of the expected amount was recovered and finally after six hours exposure (with an overall uptake of approximately one tenth that of the previous three experiments) to a suburban atmosphere, the expected and observed amounts showed good agreement.Hence, as far as benzene is concerned, it can be observed that, the longer one exposes this device to an atmosphere, the less sample one collects.This somewhat unexpected result for benzene recovery could be explained as follows. From previous work17,18 it is accepted that many VOC exhibit Langmuirian behaviour with respect to their adsorption on solid adsorbents. For an individual VOC at equilibrium, the mass of adsorbate per unit mass of the adsorbent mads, is given by: m bcm l bc ads = + max (5) where: b = distribution coefficient (adsorption to desorption equilibrium coefficient ratio). c = gaseous phase VOC concentration, and mmax = maximum adsorbable mass.In the case of a VOC mixture, it has been demonstrated that the mass of a given adsorbate is dependent on the nature and amount of the other constituents in the mixture. Mass adsorbed mj for a compound j in a mixture of i compounds can be written as follows: m b c m b c j j j i i j = +å max 1 (6) where: ·bici, indicates that the sum of the product must be performed for all compounds. Eqn. 6 quantifies to some extent the competitive adsorption effects between different VOC.The adsorption parameters and hence the affinity for different BTX have been determined on Tenax–GC (Supelco). It was shown that affinity increases with the degree of alkylation of the aromatic ring. So, at equilibrium the relative masses adsorbed follow the trend: mbenzene <mtoluene <mxylene From as yet unpublished work, it can be shown that Carbotrap- B exhibits a similar behaviour. In the case of these field experiments, one can expect some competitive effects between benzene and other aromatics, especially if one considers that in areas polluted by automotive traffic the atmospheric toluene concentration, for example, is often three times that of benzene.Having identified this problem of competitivity at equilibrium on Carbotrap two possible options were open. Either to use the most common configuration of Radiello with a stronger adsorbent, such as activated charcoal, and revert to solvent desorption.Or, to stay with thermal desorption and reduce the effective uptake rate of the Radiello cartridge by altering its host sampler geometry. The latter choice was taken and the performance assessed by a series of laboratory and field experiments. Experimental. Part 2 An empty Perkin Elmer tube was fitted with a 4.8 mm od Radiello cartridge (see Fig. 4) previously loaded with Å 150 mg of Carbotrap-B held in place by a retaining spring. Diffusion path length can be precisely controlled by the positioning of the Radiello cartridge front-end cap.For the following experiments the front-end cap was inserted 4 mm bringing the total diffusion path length to 21 mm. As earlier, the preliminary laboratory work involved the assessment of the recovery of spiked samples from the new sampler arrangement. Experiments were conducted in the same manner and, as before linear, 100% desorption efficiencies were observed for all BTX. Next, in order to validate theoretical uptake rates, four RPE were exposed for six hours to high concentrations of BTX in the exposure chamber of a dynamic atmosphere generator.19 The atmosphere was controlled using classic dynamic sampling tubes and the results are shown in Fig. 5. Even though the concentrations used in this experiment are much higher than those encountered in ambient air, no saturation or competitive effects were observed. It must be Table 1 Comparison of measured values by Radiello cartridges and parallel techniques* Measured atmospheric concentrations/mg m23 Duration Device Benzene Toluene m/p-Xylene o-Xylene 7 d Axial tubes 1.9 8.1 4.4 1.4 Radiello 0.0 6.0 4.3 1.3 6 d Axial tubes 1.7 5.3 2.7 0.9 Radiello 0.4 4.8 3.0 1.0 11 h Canister 16.3 59.2 24.3 8.6 Radiello 7.2 55.5 27.0 9.7 6 h Active tubes 2.8 8.3 2.6 1.0 Radiello 2.8 8.5 3.1 1.2 * Diffusion coefficients taken from the literature.20 Fig. 4 Schematic of the Radiello Perkin Elmer configuration. Analyst, December 1997, Vol. 122 1483noted that the total uptake in this experiment represents only a fraction (1%) of the uptake on the radial mode Radiellos during the eleven hour field experiment. The maximum RSD observed was 1.5% and there was very little deviation from theoretial uptake rates. Field Experiments As a final performance assessment, six RPE tubes were exposed alongside six INERIS tubes in the field (under the following conditions: seven days exposure. Suburban site. Average temperature, 14 °C.Average relative humidity, ND). The results are shown in Table 2. The results show little dispersion even at low atmospheric concentrations. Benzene is well recovered as predicted by laboratory experiments. It is also interesting to note, that in both cases, measured concentrations are closely matched for the devices, even though theoretical uptake rates are not the same (due to differing geometries). This suggests that no competition problems were encountered and supports the assumption made earlier of a linear uptake rate throughout the sampling period for these lower uptake rate devices.This conformity with theoretical expectations of sampler behaviour is due to the control that is now possessed over the diffusion path length, and hence the uptake rate of the Radiello cartridge in this axial configuration. Conclusions Diffusive samplers are undoubtedly very practical tools for air quality monitoring expecially if extensive campaigns or personal monitoring are planned.If correctly used, high uptake rate devices allow the measurement of low concentrations of VOC even for short duration exposure. The relatively new radial sampler, Radiello, offers the possibility of solvent or thermal desorption. However under certain conditions it has been shown that the high uptake rate of this device can cause recovery problems for some compounds when using thermally desorbable adsorbents, such as Carbotrap. The unexpected results may be explained with the help of previously published work.The solution proposed here combines the versatility of Radiello cartridges with a popular auxiliary tube. This arrangement including the adjustable diffusion path length gives new possibilities to commercial products already on the market and allows one to cover almost any sampling scenario with just one analytical chain. Acknowledgements A. Frezier for her technical assistance and Sigma Aldrich, France, for their kind donation of equipment.References 1 Guicherit, R., and Schulting, F. L., Sci. Total Environ., 1985, 43, 193. 2 Tancrede, M., Wilson, R., Zeise, L., and Crouch, E. A. C., Atmos. Environ., 1987, 21, 2187. 3 Agency for Toxic Substances and Disease Registry, US Public Health ATSDR PB93-182-384, Atlanta GA (1993). 4 Snyder, R., Witz, G., and Goldstein, B. D., Environ. Health Perspect., 1993, 100, 293. 5 Brown, R. H., Charlton, J., and Saunders, K. J., Am. Ind. Hyg. Assoc. J., 1981, 42, 865. 6 Tompkins, F. C., Jr., and Goldsmith, R. L., Am. Ind. Hyg. Assoc. J., 1977, 38, 371. 7 Palmes, E. D., and Gunnison, A. F., Am. Ind. Hyg. Assoc. J., 1973, 34, 78. 8 Brown, V. M., Crump, D. R., and Gardiner, D., Environ. Technol., 1992, 13, 367. 9 Brown, V. M., Crump, D. R., Gardiner, D., and Yu, C. W. F., Environ. Technol., 1993, 14, 771. 10 M�egie, G., Bonte, J., Carlier, P., Chavaudra, J., Dizengremel, P., Feugier, A., Granier, C., Hauglustaine, D., Kanakidou, M., Le Bras, G., Marenco, A., Mouvier, G., Tissot, B., Toupance, G., and Truhau, R., Rapp. l’Acad. Sci., 1993, 30, 10. 11 Derwent, R. G., Middleton, D. R., Field, R. A., Goldstone, M. E., Lester, J. N., and Perry, R., Atmos. Environ. 1995, 29, 923. 12 Carler, R. E., Thomas, M. J., Marotz, G. A., Lane, D. D., and Huson, J. I., Environ. Sci. Technol., 1992, 26, 2175. 13 Brown, R. H., Harvey, R. P., Purnall, C. J., and Saunders, K. J., Am. Ind. Hyg. Assoc. J., 1984, 45, 67. 14 Cocheo, V., Boaretto, C., and Sacco, P., Am. Ind. Hyg. Assoc. J., 1996, 57, 897. 15 Supelco GC bulletin 846B, Supelco Inc, Bellefonte, USA, 1986. 16 Cao, X. L., and Hewitt, C. N., Atmos. Environ. 1993, 27A, 1865. 17 Comes, P., Gonzalez-Flesca, N., M�enard, T., and Grialt, J. O., Anal. Chem., 1993, 65, 1048. 18 Comes, P., Gonzalez-Flesca, N., Bader, F., and Grimalt, J. O., J. Chromatogr. A, 1996, 723, 293. 19 Jaouen, P., Gonzalez-Flesca, N., and Carlier, P., Environ. Sci. Technol., 1995, 29, 2718. 20 Lugg, G. A., Anal. Chem., 1968, 40, 1072. Paper 7/05610I Received August 4, 1997 Accepted October 31, 1997 Fig. 5 Concentration comparison for dynamic and passive measurements. Table 2 Results from parallel exposure of INERIS and RPE tubes Measured atmospheric concentrations/mg m23 (% RSD of six results) Device Benzene Toluene m/p-Xylene o-Xylene INERIS 2.7 8.2 3.9 1.4 tubes (4) (3) (5) (5) RPE 2.8 8.4 3.8 1.4 tubes (3) (2) (3) (4) 1484 Analyst, December 1997, Vol.

ISSN:0003-2654    

DOI:10.1039/a705610i

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Development of Analytical Methods for the Determination of Synthetic Mud Base Fluids in Marine Sediments† Lynda Webstera, Peter R. Mackieb, Simon J. Hirdc, Pauline D. Munroa, Nigel A. Browna and Colin F. Moffat*a a FRS Marine Laboratory, P.O. Box 101, Victoria Road, Aberdeen, UK AB11 9DB b Department of Medicine and Therapeutics, University of Aberdeen, Foresterhill, Aberdeen, UK AB25 7AI c Central Science Laboratory, Sand Hutton, York, UK YO4 1LZ A method for the extraction and quantification of synthetic drilling fluids in sediment, which could be adapted for different concentrations and types of synthetic fluid, was developed.Six different base fluids were investigated at four different concentrations (@5, 100, 500 and 5000 mg g21). The method used varied depending on the type and concentration of fluid in the sediment. Extraction of the base fluids was achieved by sonication in dichloromethane and methanol. The extract was purified using normal-phase high-performance liquid chromatography.Quantitative analysis was carried out by gas chromatography with flame ionisation detection. Recoveries and limits of detection were calculated for each of the six synthetic fluids at each of the four concentration levels by spiking clean sediment. This method was used to monitor synthetic fluid concentrations in samples from sea-bed surveys and from a solid-phase test designed to assess the biodegradability of these compounds. Keywords: Offshore drilling; synthetic mud base fluids; high-performance liquid chromatography; gas chromatography; sonication The deposition of drill cuttings on the sea-bed is probably the single most important factor in the environmental impact of offshore drilling operations in the northern North Sea.The cuttings are discharged into the sea at the side of the platform, resulting in cutting piles on the sea-bed which may be 20–30 m high. The contaminated area may extend several kilometres from the source.1 Since the late 1970s, the use of oil-based drilling muds in the North Sea has increased owing to their improved drilling efficiency over water-based muds in deep and directionally drilled wells.The first oil-based muds contained diesel oil, which contains a high proportion of toxic aromatic hydrocarbons. Concern over the possible long-term effects of diesel oil resulted in its replacement with low-toxicity mineral oils, which contain lower concentrations of aromatic compounds.In 1992, the Paris Commission for the prevention of oil pollution set down maximum discharge levels of 10% mineraloil- on-cuttings with a further stipulation of a 1% threshold for all North Sea drilling operations by January 1997. This led to the development of a diverse range of synthetic drilling mud base fluids (SMs), which were thought to be more biodegradable and less toxic to the environment than mineral oils. Consequently, the number of wells drilled using SM fluids had increased to more than 100 in 1996.Furthermore, SM-oncuttings discharges are not currently regulated to a specific target, although industry is requested to meet the 10% threshold previously operative on mineral oil-based muds. There is an obvious requirement to monitor these chemicals in sea-bed sediment samples and also in experimental systems designed to study the biodegradation of these compounds. Therefore, a method was developed which could be readily adapted to detect a variety of SM fluids at concentrations ranging from <1 mg g21 to 5 mg g21 dry sediment.The method had to be capable of extracting the complete range of generic types of SM base fluid currently in use in the off-shore industry, including an ester, an acetal derivative and the hydrocarbons, linear a-olefin (LAO), poly-a-olefin (PAO), internal olefin (IO) and an n-alkane (n-paraffin). One of the primary degradation products of the ester, namely 2-ethylhexan- 1-ol, was also examined.The final extraction method was varied in relation to SM concentration with respect to the mass of sediment extracted and the extract volumes prepared prior to high-performance liquid chromatography (HPLC) and gas chromatography (GC). Recoveries and limits of detection were determined for each method (expected concentrations @5, 100, 500 and 5000 mg g21 dry mass sediment) for all six SM fluids. Experimental Synthetic Mud Base Fluids The SM fluids were supplied to the Marine Laboratory by the various mud companies under the auspices of the UK Offshore Operators Association (UKOOA) and the European Oilfield Speciality Chemicals Association (EOSCA).These fluids consisted of an n-paraffin (a mixture of n-alkanes with chain lengths ranging from C11 to C16); a PAO, which mainly consisted of a hydrogenated decene dimer, C20H42 (95%), with lesser amounts of C30H62 (4.8%) and C10H22 (0.2%); an IO, which consisted of a series of isomeric forms of C16 and C18 alkenes; an LAO, consisting of a series of isomeric forms of C14 and C16 monoenes; a monoester, namely 2-ethylhexan-1-ol esterified to even chain length, aliphatic carboxylic acids in the C8–C16 range; an acetal derivative made from isobutyl aldehyde; and 2-ethylhexan-1-ol.The composition of the SMs was confirmed by gas chromatography–mass spectrometry using an HP 5890 gas chromatograph (Hewlett-Packard, Stockport, UK) interfaced with a high-resolution Autospec mass spectrometer (Micromass, Manchester, UK).The GC conditions and injection were as detailed below except that the carrier gas was helium (10 lb in22). Standard positive electron ionisation (EI+) mass spectra were collected using an electron energy of 70 eV. Perfluorokerosene, obtained from Micromass, was used as the mass spectrometer calibrant. Apparatus The extraction was performed in 250 ml glass centrifuge tubes, the solvent–sediment mixture being sonicated in an F5 Minor † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997.Analyst, December 1997, Vol. 122 (1485–1490) 1485sonic bath (Decon Ultrasonics, Hove, Sussex, UK) to aid extraction. Concentration of organic solutions was performed either under reduced pressure using a rotary evaporator (Bibby Sterilin, Stone, Staffordshire, UK) or under a stream of oxygenfree nitrogen. A GS 6R refrigerated centrifuge [Beckman Instruments (UK), High Wycombe, UK] was used to separate the solvent extract from the sediment.Isolation of the analytes from more polar material was achieved by HPLC using a Spectra-Physics Series P100 isocratic pump (Thermo Separation Products, San Jose, CA, USA) and a Genesis metal-free column (SIL 4 mm, 25 cm 34.6 mm id; Jones Chromatography, Hengoed, UK). Standards were weighed to an accuracy of 0.01 mg using a self-calibrating analytical balance (Sartorius RC210S; European Instruments, Oxford, UK).Reagents Analytical-reagent grade isohexane, dichloromethane (DCM), methanol, acetone and water were supplied by Rathburn Chemicals (Walkerburn, UK). Anhydrous sodium sulfate was purchased from BDH (Poole, Dorset, UK). Squalane (Kodak Eastman Fine Chemicals, Rochester, NY, USA), heptamethylnonane (HMN), methyl tricosanoate and octan-1-ol (all from Aldrich, Gillingham, Dorset, UK) were used as internal standards. Sediment The sediment used for the recovery experiments and the solidphase biodegradation test was obtained at low water from Loch Thurnaig, by Loch Ewe, Wester Ross, Scotland.It was wet sieved with sea-water through a 2 mm, then a 1 mm mesh sieve to remove large stones, molluscs, crustaceans, polychaetes and organic debris. The total organic carbon and nitrogen contents of the sediment, determined using a Model 2400 CHN elemental analyser (Perkin-Elmer, Beaconsfield, Buckinghamshire, UK), were 1.2% and 0.1% respectively. The moisture content was 20%.Preparation of Standard Solutions A concentrated stock solution of each SM base fluid in isohexane (approximately 25 mg ml21) was prepared for spiking clean sediment. A dilute stock solution (approximately 500 mg ml21) of each SM was also prepared in isohexane. Working solutions of each SM fluid were prepared by making dilutions of the appropriate stock SM solution to cover the range of spiking levels. Five standard solutions were prepared (2.5, 50, 100, 150 and 600 mg ml21) and each solution was analysed by gas chromatography with flame ionisation detection (GC–FID).Several of the SMs, specifically the nparaffin (Fig. 1) and ester [Fig. 6(a)], contained several discrete components. The composition for the individual components covered the range 4–49% for the n-paraffin and 2–63% for the ester. The total SM concentration of the most dilute solution was selected such that each individual component could be readily identified and quantified.The total peak area was plotted against concentration for each SM to give a calibration curve and the gradient of the curve and the correlation coefficient (r) were determined. Internal standard solutions (5000 and 50 mg ml21) were prepared in isohexane to check that these compounds were also recovered satisfactorily using the SM extraction method. A mixture of heptamethylnonane and squalane was used for the hydrocarbon SM fluids as these compounds have been shown to have response factors that are similar to those of hydrocarbons, give a linear response when analysed by GC–FID over the required concentration range and are the recognised standards for aliphatic hydrocarbon environmental analysis.Their retention times were distinct from all the determinants of interest. Methyl tricosanoate was used for the oxygenated SM fluids as it is chemically similar to the ester SM, was found to give a similar GC–FID response to the oxygen-containing determinants and gave a linear response over the required concentration range.Extraction of SM Fluid from Sediment: Recovery Experiment Many of the SMs are primarily hydrocarbons and the method was based on that for the isolation of mineral oils from sediment, which is well established2 and utilises a monophasic, binary solvent system.3 Sediment (5–20 g) was weighed into a 250 ml glass centrifuge tube and the appropriate amount of internal standard was added (Table 1). The sediment was then spiked to give SM fluid concentrations of 5, 100, 500 or 5000 mg g21.Clean sediment (20 g), to which only the internal standards were added, was also analysed with each spiking experiment and extracted in accordance with the method for a sediment estimated to contain @5 mg g21 SM. The SMs were extracted with a mixture of methanol (20 ml) and DCM (20 ml) by sonication for 5 min followed by centrifugation for 10 min at 2000 rpm and 2 °C.The supernatant was transferred into a separating funnel containing HPLC-grade water (18 ml) and shaken thoroughly for 2 min. The layers were allowed to separate and the lower, organic phase was transferred into a conical flask. The sediment was re-extracted with DCM (20 ml), repeating the ultrasonication and centrifugation. The solvent was decanted into the separating funnel and shaken. The lower, DCM layer was combined with the first extract and dried over anhydrous sodium sulfate (10 g) before concentration to approximately 0.5 ml by rotary evaporation.Solvent exchange was achieved by addition of isohexane (25 ml) followed by rotary evaporation to about 0.5 ml. This process was repeated a further twice before transferring the solution, with washings, into a vial or calibrated flask. The isohexane was concentrated by rotary evaporation and then under a stream of oxygen-free nitrogen, if necessary, to the appropriate volume (Table 1).External standard methodology was used to assess the recovery of the SMs; hence, for volumes !1 ml the samples were prepared in calibrated flasks. For smaller volumes, the actual volume of solvent was calculated by obtaining the mass of the final solution and converting to volume units using a value of 0.660 g ml21 for the density of isohexane.4 The SM fluids were separated from polar contaminants by normal-phase HPLC. An aliquot of the isohexane extract (150 ml) was injected on to a metal-free Genesis SIL 4 mm HPLC column (25 cm 34.6 mm id) with elution with isohexane at a flow rate of 2 ml min21.The hydrocarbons (IO, PAO, LAO and n-paraffin), which eluted with isohexane, were collected between 0 and 3 min following injection. For the acetal and ester the solvent was switched to isohexane–acetone (99 + 1 Table 1 Concentration specific variations used in the extraction of both hydrocarbon- and oxygen-containing SMs from sediments Expected Amount of Volume of Final extract concentra- Mass of standard/ extract prior volume for tion/mg g21 sediment/g mg g21 to HPLC/ml GC/ml @5 20 5 0.5 0.05* 100 10 250 1 0.25* 500 10 1000 5 1 5000 5 5000 20 2 * An increased final volume was used for the acetal; 0.2 ml (@5 mg g21) and 1.0 ml (100 mg g21). 1486 Analyst, December 1997, Vol. 122v/v) after 3.75 min and the fraction eluting between 3.75 and 20 min was collected in a round-bottomed flask. The eluate in both cases was concentrated to the appropriate volume (Table 1) before analysis by GC–FID.Extraction of 2-Ethylhexan-1-ol from Sediment One of the primary degradation products of the ester SM, 2-ethylhexan-1-ol, was also extracted and quantified. Extraction of 2-ethylhexan-1-ol from sediment samples was achieved as for the SM fluids, but using octan-1-ol as the internal standard since the response factor and linear range for octan- 1-ol were similar to those of 2-ethylhexan-1-ol. The concentrated extract was purified by HPLC using DCM as the mobile phase at a flow rate of 2 ml min21.An aliquot (150 ml) of the concentrated extract (final volume 1000 ml) was injected on to a metal-free column as before and 2-ethylhexan-1-ol was collected in the fraction eluting between 8 and 20 min. This fraction was collected in a round-bottomed flask and concentrated to an appropriate volume (300 ml) prior to analysis by GC–FID. GC–FID Analysis Samples (1 ml) were chromatographed on an HP 5890 Series II gas chromatograph equipped with an HP 7673 automated, cool on-column injector and fitted with a non-polar Ultra 1 column (25 m 3 0.2 mm id, film thickness 0.33 mm; Hewlett-Packard).The carrier gas was ECD-grade nitrogen (16 lb in22) and the oven temperature programmes were as detailed in Table 2. The detector was maintained at 300 °C throughout. Data were collected using a PE Nelson 600 series link box and processed using a Turbochrom 3 data station (Perkin-Elmer).Peaks were assigned as SM fluids on the basis of retention time relative to the original SM base fluid, summed as necessary and quantified on the basis of the relevant calibration curve covering the concentration range 2.5–600 mg ml21, taking into account the volumes prepared at each stage of the extraction. When analysing sea-bed samples or sediments from the biodegradation tests, quantification was on the basis of the added internal standard. The Turbochrom 3 data station takes into account the small differences between the analyte and the internal standard response during the calculations of unknowns.Furthermore, in the internal standard calculation, the sample volumes at each stage were not so critical and variations in recoveries are accounted for. 2-Ethylhexan-1-ol was determined by injecting an aliquot (1 ml) on to an HP 5880 gas chromatograph using an HP 7673A cool, on-column automated injector (Hewlett-Packard). A polar DB-23 fused-silica column (30 m 3 0.25 mm id, film thickness 0.25 mm; J & W Scientific, Folsom, CA, USA) was required for the analysis.The column oven was maintained at 40 °C for 5 min after injection and then programmed at 6 °C min21 to 180 °C. This temperature was held for 20 min then increased at 20 °C min21 to 220 °C and held there for 5 min. ECD-grade nitrogen was used as the carrier gas (16 lb in22). The temperature of the detector was 300 °C. Data processing was carried out using an HP 5880A Level 4 integrator, which computes the area of peaks, determines the chromatographic baseline and corrects the area accordingly.Determination of Limits of Detection Limits of detection were determined empirically by spiking clean sediment with concentrations of SM equivalent to 4–5% of the specified concentration for each method. The sediments were extracted according to the appropriate protocol. Each component of the SM, at its limit of detection, had to be clearly identifiable on the gas chromatogram and quantifiable.Determination of the Dry Mass of Sediments To enable the data to be quoted as mg g21 dry mass of sediment, sediment (10 g) was accurately weighed into an aluminium dish and heated at 80 °C in an oven for 22 ± 2 h. Previous unpublished data from this laboratory had shown this method to be applicable to a range of marine sediments containing contaminant hydrocarbons. Analysis of Sea-bed Survey Samples and Samples from a Biodegradation Test The above methodology was applied to samples from sea-bed surveys and from sediment samples from a solid-phase test designed to assess the biodegradability of the SM fluids.5–7 The sediments were extracted in a manner identical with that in the recovery experiments, having added the appropriate amount of internal standard, this being dependent on the expected concentration of the SM in the sediment (Table 1).Solvent blanks, prepared according to the method for sediments containing an estimated @5 mg g21 SM, were analysed with every batch of six samples.Peaks were assigned as SM fluids on the basis of retention time relative to the original SM base fluid and quantified using the internal standard. The solid-phase biodegradation test5–7 involved mixing clean sediment with six different SM fluids to give nominal concentrations of 100, 500 and 5000 mg g21 dry mass of sediment. The jars containing the spiked sediments were placed in a flowing sea-water system.At set times (0, 14, 28, 56 and 120 d) three equivalent jars were removed and the SM fluid was extracted using the above methodology, appropriate for each spiking level, and analysed by GC–FID. This allowed the determination of the primary degradation rates of the different fluid types. A number of sea-bed surveys have been carried out to monitor concentrations of SM fluids found in the environment. Samples from the environment can contain SM concentrations from in excess of 0.5% m/m down to trace amounts.An estimation of the amount of SM fluid that a sea-bed sample might contain can be made by relating it to the distance from where the sample was obtained to the drilling platform and this, together with the information on the expected type of SM, was used to select the appropriate method. Results and Discussion The overall objective was to produce a method, the basis of which was similar for all the SMs, which could be used to determine concentrations in excess of 0.5% m/m and as low as @5 mg g21.To this end, the basic extraction was the same in all cases and the variables were the starting mass of sediment, the volume of the extract prior to HPLC and the final volume of the test solution prior to GC–FID. Furthermore, the eluents for Table 2 Gas chromatograph injection temperatures and oven temperature profiles for the analysis of SMs Initial Final Total Injector hold Final hold run tempera- time/ Ramp/ tempera- time/ time/ SM ture/°C min °C min21 ture/°C min min Ester 60 0 8 300 15 45 Acetal 60 0 10 300 10 34 LAO 60 0 8 300 10 40 PAO 60 0 10 300 10 34 IO 60 0 10 300 15 39 n-Paraffin 60 3 4 300 30 93 Analyst, December 1997, Vol. 122 1487HPLC were varied depending on whether the test compound was a hydrocarbon SM or an oxygen-containing SM. At the higher concentrations of SM, interference from biogenic hydrocarbons can be ignored. However, at sites remote from platforms, or in a sediment where substantial biodegradation has occurred, the concentration of the SM is such that interference from the natural hydrocarbons can be a problem.Although the aliphatic hydrocarbon profile of a sediment is generally dominated by long-chain, odd-carbon alkanes including heptacosane, nonacosane and hentriacontane, the shorter chain alkanes (undecane to octadecane) are present at trace levels and cannot be readily distinguished from the equivalent compounds in the n-paraffin SM (Fig. 1). Under these circumstances it is necessary to review the individual aliphatic hydrocarbon profile and compare it with that for the SM so as to assess the relative abundance of the natural components. The problem is far less acute with the LAO, IO and PAO since they consist of discrete groups of molecules and give specific profiles when analysed by GC–FID (Figs. 2, 3 and 4, respectively), these being distinct from biogenic compounds.The methodology for the oxygen-containing compounds is such that aliphatic hydrocarbons are eluted from the HPLC column prior to the determinant and are therefore less of a problem when dealing with the very low SM concentrations. Although adventitious hydrocarbon contamination is always a complication with this type of analysis, it can be minimised by using the procedures described by Webster et al.8 All the SMs, with the exception of the acetal (Fig. 5), are mixtures and this obviously influences the limit of detection on the basis of total SM.This was most acute for the ester [Fig. 6(a)] and n-paraffin (Fig. 1). At the two higher concentrations, quantification of the SM at 4–5% of the specified concentration, including individual components in the case of the ester and n-paraffin, was possible. There were no problems with interfering compounds. The signal-to-noise ratio for the smallest components in both the ester and n-paraffin were approximately 30 : 1 and consequently the limit of determination was set at 25 and 200 mg g21 for the 500 and 5000 mg g21 methods, respectively (Table 3).At the two lower concentrations, the methodology was such that a larger amount of total SM was injected on to the GC column following isolation of the SM from sediment spiked with 5% of the specified concentration. The various components for each SM Fig. 1 Gas chromatogram of an n-paraffin SM extracted from a sediment obtained during a survey of the sea-bed close to an oil platform operating in the northern North Sea.The internal standard was heptamethylnonane (HMN). Fig. 2 Gas chromatogram of an LAO SM extracted from a sediment as part of a study into the biodegradation of SM fluids. The sediment, mean starting SM concentration 100.2 mg g21 dry sediment (SE = 2.9 mg g21, n = 3), was removed from the flowing sea-water test rig after 28 d. The concentration of the LAO (49.1 mg g21 dry sediment) was calculated from the combined peak areas for LAO1 and LAO2, using HMN as the internal standard.The internal standard for the hydrocarbon SMs included squalane (SQ). Fig. 3 Gas chromatogram of an extract from a sediment sample spiked with 5000 mg g21 of IO as part of a recovery experiment. The area of the two multiple peaks (IO1 and IO2) was summed prior to calculation of the SM concentration. The internal standard was HMN. Fig. 4 Chromatogram of an extract from a sediment sample spiked with 500 mg g21 of PAO as part of a recovery experiment. The total area of the partially resolved complex (PAO) was used to calculate the concentration of the SM fluid using squalane (SQ) as the internal standard. 1488 Analyst, December 1997, Vol. 122were again readily identified but there was minor interference from the biogenic hydrocarbons, specifically pentadecane, when determining the n-paraffin at both 250 ng g21 and 5 mg g21 using the @5 and 100 mg g21 methods, respectively.The smallest component in the ester, the C16 ester, was quantifiable at both test concentrations but the size of the peak was such that the limit of determination was set at 250 ng g21 and 5 mg g21 for the @5 and 100 mg g21 methods, respectively (Table 3). For the LAO, PAO and IO it was possible to detect 2.5 mg g21 of SM using the 100 mg g21 method. The consequence of the acetal being a single component (Fig. 5) was that the volumes of the final extract were increased for the @5 and 100 mg g21 methodologies (Table 1).The calibration curves for the internal standards and the various SMs were linear over the concentration range 2.5–600 mg ml21, equivalent to 2.5–600 ng on-column. The correlation coefficients of the calibration curves for all the SMs and internal standards were > 0.999. Differences in detector response between the analytes and the internal standard were taken into account by the data system when performing the internal standard calculation. There is a close similarity between the n-paraffin SM (Fig. 1) and the well established low-toxicity drilling fluids with respect to both carbon number range and distribution.9 A marked difference between the two types of drilling fluid is, however, the absence of branched isomers in the SM. As the analytical method was based on the technique for the analysis of mineral oil-type base fluids, it would be expected to be suitable for the n-alkane-type SM. The combination of number of base fluids studied and the range of concentrations investigated meant that only single recovery experiments were performed for each SM at each concentration.Although this limits the available data, it does show that recoveries in excess of 70% were achieved at the four concentrations for all the SMs (Table 3). The mean recoveries for all the SMs at 5, 100, 500 and 5000 mg g21 were 84, 92, 95 and 92%, respectively. An analysis of variance showed that there was no significant difference between the means at each of the four concentrations or between the recoveries of the individual SM fluids. The mean recovery of the n-paraffin was 86% [standard error of the mean (SE) = 6.0%, n = 4].The only SM which gave a lower mean recovery was the ester at 84% (SE = 5.6%, n = 4). All other SMs gave a mean recovery of > 90%, the highest being the acetal at 98% (SE = 3.0%, n = 4). During the analysis of samples from the solid-phase test, duplicate analyses were performed on three equivalent test samples which had been dosed with the n-paraffin at 100 mg g21 SM and maintained in the flowing sea-water system for 120 d.The mean concentration was 17.8 mg g21 (SE = 1.2 mg g21, n = 6; Table 4). Quadruplicate analyses of sea-bed survey samples containing the n-paraffin gave a mean of 320 mg g21 Fig. 5 Chromatogram showing the single peak of the acetal SM fluid extracted from a sea-bed survey sample collected close to an installation operating in the northern North Sea.The internal standard was methyl tricosanoate (23 : 0). Fig. 6 Gas chromatograms of (a) the five components of the ester SM (E1- E5) and (b) 2-ethylhexan-1-ol. The ester was extracted from a sediment as part of a study into the biodegradation of SM fluids. The fifth component (E5) is the smaller of the two peaks eluting around 24.1 min. The internal standard was methyl tricosanoate (23:0). The 2-ethylhexan-1-ol was extracted from a sediment after 28 d in a flowing sea-water solid-phase test.The concentration was calculated using octan-1-ol as the internal standard. Table 3 Summary of recovery data for each of the SMs at the four different concentrations together with the concentration specific limit of determination (LD). The mean recovery across the four concentrations together with the associated standard error of the mean (SE) is also presented Recovery (%) Spiked synthetic fluid concentration/ mg g21 dry sediment SM @5 100 500 5000 Mean SE Ester 76 92 73 95 84 5.6 Acetal 91 95 99 105 98 3.0 LAO 84 101 100 91 94 4.0 PAO 97 87 92 82 90 3.2 IO 81 104 99 83 92 5.7 n-Paraffin 78 74 91 100 86 6.0 LD/mg g21 0.25 5 25 200 Analyst, December 1997, Vol. 122 1489(SE = 4.9 mg g21; Table 4), indicating good reproducibility of the method for both the biodegradation test system and sea-bed survey samples. Triplicate jars of sediment, originally spiked with a nominal concentration of 100 mg g21 dry sediment of LAO (Fig. 2) in the solid-phase test system, were found to contain 49.1, 45.1 and 46.6 mg g21 dry sediment after 28 days, giving a mean of 46.9 mg g21 dry sediment (SE = 1.2 mg g21, n = 3). Equivalent replication was obtained at the higher concentrations (Table 4). As with the n-paraffin, the data were calculated against HMN. The chromatogram of the IO (Fig. 3), taken from the recovery experiment, shows a similar profile to that of the LAO.Both of these SMs are alkenes but the IO is a combination of C16 and C18 alkenes whereas the LAO consists of C14 and C16 alkenes. Furthermore, the mixture of components of the IO appears to be more complex, with both the peaks having the appearance of being composites rather than individual components, as is the case with the LAO. Sediment from a solid-phase test, spiked with the IO and maintained in flowing sea-water for 56 d, was found to contain 5050 mg g21 dry sediment (SE = 12 mg g21, n = 3).The PAO was only partially resolved under the GC conditions employed. Nevertheless, this did not affect the quantification of the total SM as there were distinct start and end points for the unresolved complex mixture (UCM), as is evident for the sample spiked with 500 mg g21 of PAO as part of the recovery experiments (Fig. 4). The acetal was the simplest of all the synthetic mud base fluids containing only a single compound (Fig. 5). It was slightly more polar than the hydrocarbons, requiring 1% v/v of acetone in the HPLC mobile phase and more time to elute from the column.The same extraction procedure could be applied, although increased volumes were required before GC, owing to the simple composition of the SM. The acetal gave the highest mean recovery across the four concentrations at 98% (SE = 3.0%, n = 4). Good reproducibility was evident in replicate analyses from the solid-phase test with standard errors of 2–12 mg g21 (n = 3) for sediment taken from three individual jars at four time intervals with concentrations of approximately 500 mg g21 SM.Similar replication was obtained at other concentrations (Table 4). Methyl tricosanoate was well resolved from the determinant (Fig. 5). This standard was also resolved from the five different components of the ester SM fluid [Fig. 6(a)]. This particular sample, which originally contained 500 mg g21 of SM, was found to have an ester concentration of 130 mg g21 after 56 d in the flowing sea-water system.The polarity of the ester was only slightly greater than that of the acetal, hence the same extraction and clean-up procedure could be employed, yielding satisfactory results, with recoveries ranging from 73 to 95% (Table 3). Marine sediments contain bacteria which produce enzymes which are readily able to cleave ester bonds. One of the primary degradation products of the ester SM is 2-ethylhexan-1-ol and, because of the very rapid degradation of the ester in marine sediments, it was deemed appropriate to utilise a degradation product as part of the environmental assessment of this SM.Extraction of the alcohol was achieved using the same method as for the SMs. It was necessary, however, to use an alternative HPLC procedure which required a more polar mobile phase. Similarly, a more polar capillary column was required for the GC analysis but good peak shape and resolution of the determinant and internal standard were achieved using a DB-23 column [Fig. 6(b)]. This particular example originated from a sediment which had been incubated for 28 d in a flowing seawater system. The sediment was initially spiked with 500 mg g21 of the ester SM. After 28 d, 75.6 mg g21 of 2-ethylhexan- 1-ol were found in the sediment extract. The method developed for the analysis of SM fluids in sediment was readily applied to all the generic types of SM base fluids currently in use in the off-shore industry.It was also possible to extract 2-ethylhexan-1-ol, a primary degradation product of the ester, from sediment using this method. The same extraction procedure was used for all test compounds and only the HPLC conditions were varied for the different fluids. By altering the volumes before HPLC and GC–FID, the method could be used to quantify a wide range of concentrations (from < 1 mg g21 to > 5000 mg g21) of fluids in sediment samples. Recoveries of > 80 % were achieved in 82% of the recovery experiments.This method has been used successfully to monitor SM fluid concentrations in samples from sea-bed surveys, despite the different sediment types and range of concentrations encountered, and in a solid-phase test developed to monitor the biodegrability of these compounds. Investigations into adapting the method for the determination of SM fluids in biota are under way together with studies into the application of the extraction methodology for the isolation of olive oil from sediment, this triglyceride oil being used as a positive control in the solid phase test.References 1 Davies, J. M., Addy, J. M., Blackman, R. A., Blanchards, J. R., Ferbrache, J. E., Moore, D. C., Sommerville, H. J., Whitehead, A., and Wilkinson, T., Mar. Pollut. Bull., 1984, 15, 363. 2 Stagg, R. M., McIntosh A., and Mackie, P. R., Aquat. Toxicol., 1995, 33, 245. 3 Bligh, E. G., and Dyer, W. J. A., Can. J. Biochem. Physiol., 1959, 37, 911. 4 Handbook of Chemistry and Physics, ed. Weast, R. C., CRC Press, Cleveland, OH, USA, 69th edn., 1988–89. 5 Croce, B., McIntosh, A. D., Moffat, C. F., Hird, S. J., and Stagg, R. M., Biodegradation of Base-Oils Used in Synthetic Drilling Muds in an Experimental Solid-Phase System, Fisheries Research Services Report No. 3/96, Marine Laboratory, Aberdeen, 1996. 6 Munro, P. D., Moffat, C. F., Couper, L., Brown, N. A., Croce, B. and Stagg R. M., Degradation of Synthetic Mud Base Fluids in a Solid- Phase Test System, Fisheries Research Services Report No. 1/97, Marine Laboratory, Aberdeen, 1997. 7 Munro, P. D., Moffat, C. F., and Stagg, R. M., in Contributing to Environmental Progress—Getting the Message Across, Proceedings of SPE/UKOOA European Environmental Conference, Aberdeen, Scotland, 1997, Society of Petroleum Engineers, Dallas, TX, USA, p. 205. 8 Webster, L., Angus, L., Topping, G., Dalgarno, E. J., and Moffat, C. F., Analyst, 1997, 122, 1491. 9 Parker, J. G., Howgate, P., Mackie, P. R., and McGill, A. S., Oil Chem. Pollut., 1990, 6, 263. Paper 7/05975B Received August 14, 1997 Accepted October 13, 1997 Table 4 Examples of analyses of sediments from a solid-phase biodegradation test and from sea-bed surveys. The sediments from the solid-phase test were incubated in a flowing sea-water system for 120 d SM concentration (mean ± SE/mg g21 dry sediment) Solid phase biodegradation test (n = 3) Starting concentration/mg g21 dry Sea-bed sediment surveys SM 100 500 5000 (n = 4) n-Paraffin 17.8 ± 1.2* 446 ± 38 6101 ± 58 320 ± 4.9 LAO 12.1 ± 0.6 229 ± 17 4391 ± 309 1946 ± 46 Acetal 88.4 ± 4.4 480 ± 2 4553 ± 169 — * Result from three jars each analysed in duplicate. 1490 Analyst, December 1997, Vol. 122 Development of Analytical Methods for the Determination of Synthetic Mud Base Fluids in Marine Sediments† Lynda Webstera, Peter R. Mackieb, Simon J. Hirdc, Pauline D. Munroa, Nigel A. Browna and Colin F.Moffat*a a FRS Marine Laboratory, P.O. Box 101, Victoria Road, Aberdeen, UK AB11 9DB b Department of Medicine and Therapeutics, University of Aberdeen, Foresterhill, Aberdeen, UK AB25 7AI c Central Science Laboratory, Sand Hutton, York, UK YO4 1LZ A method for the extraction and quantification of synthetic drilling fluids in sediment, which could be adapted for different concentrations and types of synthetic fluid, was developed. Six different base fluids were investigated at four different concentrations (@5, 100, 500 and 5000 mg g21). The method used varied depending on the type and concentration of fluid in the sediment.Extraction of the base fluids was achieved by sonication in dichloromethane and methanol. The extract was purified using normal-phase high-performance liquid chromatography. Quantitative analysis was carried out by gas chromatography with flame ionisation detection. Recoveries and limits of detection were calculated for each of the six synthetic fluids at each of the four concentration levels by spiking clean sediment.This method was used to monitor synthetic fluid concentrations in samples from sea-bed surveys and from a solid-phase test designed to assess the biodegradability of these compounds. Keywords: Offshore drilling; synthetic mud base fluids; high-performance liquid chromatography; gas chromatography; sonication The deposition of drill cuttings on the sea-bed is probably the single most important factor in the environmental impact of offshore drilling operations in the northern North Sea.The cuttings are discharged into the sea at the side of the platform, resulting in cutting piles on the sea-bed which may be 20–30 m high. The contaminated area may extend several kilometres from the source.1 Since the late 1970s, the use of oil-based drilling muds in the North Sea has increased owing to their improved drilling efficiency over water-based muds in deep and directionally drilled wells.The first oil-based muds contained diesel oil, which contains a high proportion of toxic aromatic hydrocarbons. Concern over the possible long-term effects of diesel oil resulted in its replacement with low-toxicity mineral oils, which contain lower concentrations of aromatic compounds. In 1992, the Paris Commission for the prevention of oil pollution set down maximum discharge levels of 10% mineraloil- on-cuttings with a further stipulation of a 1% threshold for all North Sea drilling operations by January 1997.This led to the development of a diverse range of synthetic drilling mud base fluids (SMs), which were thought to be more biodegradable and less toxic to the environment than mineral oils. Consequently, the number of wells drilled using SM fluids had increased to more than 100 in 1996. Furthermore, SM-oncuttings discharges are not currently regulated to a specific target, although industry is requested to meet the 10% threshold previously operative on mineral oil-based muds.There is an obvious requirement to monitor these chemicals in sea-bed sediment samples and also in experimental systems designed to study the biodegradation of these compounds. Therefore, a method was developed which could be readily adapted to detect a variety of SM fluids at concentrations ranging from <1 mg g21 to 5 mg g21 dry sediment. The method had to be capable of extracting the complete range of generic types of SM base fluid currently in use in the off-shore industry, including an ester, an acetal derivative and the hydrocarbons, linear a-olefin (LAO), poly-a-olefin (PAO), internal olefin (IO) and an n-alkane (n-paraffin). One of the primary degradation products of the ester, namely 2-ethylhexan- 1-ol, was also examined.The final extraction method was varied in relation to SM concentration with respect to the mass of sediment extracted and the extract volumes prepared prior to high-performance liquid chromatography (HPLC) and gas chromatography (GC).Recoveries and limits of detection were determined for each method (expected concentrations @5, 100, 500 and 5000 mg g21 dry mass sediment) for all six SM fluids. Experimental Synthetic Mud Base Fluids The SM fluids were supplied to the Marine Laboratory by the various mud companies under the auspices of the UK Offshore Operators Association (UKOOA) and the European Oilfield Speciality Chemicals Association (EOSCA).These fluids consisted of an n-paraffin (a mixture of n-alkanes with chain lengths ranging from C11 to C16); a PAO, which mainly consisted of a hydrogenated decene dimer, C20H42 (95%), with lesser amounts of C30H62 (4.8%) and C10H22 (0.2%); an IO, which consisted of a series of isomeric forms of C16 and C18 alkenes; an LAO, consisting of a series of isomeric forms of C14 and C16 monoenes; a monoester, namely 2-ethylhexan-1-ol esterified to even chain length, aliphatic carboxylic acids in the C8–C16 range; an acetal derivative made from isobutyl aldehyde; and 2-ethylhexan-1-ol.The composition of the SMs was confirmed by gas chromatography–mass spectrometry using an HP 5890 gas chromatograph (Hewlett-Packard, Stockport, UK) interfaced with a high-resolution Autospec mass spectrometer (Micromass, Manchester, UK). The GC conditions and injection were as detailed below except that the carrier gas was helium (10 lb in22). Standard positive electron ionisation (EI+) mass spectra were collected using an electron energy of 70 eV.Perfluorokerosene, obtained from Micromass, was used as the mass spectrometer calibrant. Apparatus The extraction was performed in 250 ml glass centrifuge tubes, the solvent–sediment mixture being sonicated in an F5 Minor † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997. Analyst, December 1997, Vol. 122 (1485–1490) 1485sonic bath (Decon Ultrasonics, Hove, Sussex, UK) to aid extraction.Concentration of organic solutions was performed either under reduced pressure using a rotary evaporator (Bibby Sterilin, Stone, Staffordshire, UK) or under a stream of oxygenfree nitrogen. A GS 6R refrigerated centrifuge [Beckman Instruments (UK), High Wycombe, UK] was used to separate the solvent extract from the sediment. Isolation of the analytes from more polar material was achieved by HPLC using a Spectra-Physics Series P100 isocratic pump (Thermo Separation Products, San Jose, CA, USA) and a Genesis metal-free column (SIL 4 mm, 25 cm 34.6 mm id; Jones Chromatography, Hengoed, UK).Standards were weighed to an accuracy of 0.01 mg using a self-calibrating analytical balance (Sartorius RC210S; European Instruments, Oxford, UK). Reagents Analytical-reagent grade isohexane, dichloromethane (DCM), methanol, acetone and water were supplied by Rathburn Chemicals (Walkerburn, UK).Anhydrous sodium sulfate was purchased from BDH (Poole, Dorset, UK). Squalane (Kodak Eastman Fine Chemicals, Rochester, NY, USA), heptamethylnonane (HMN), methyl tricosanoate and octan-1-ol (all from Aldrich, Gillingham, Dorset, UK) were used as internal standards. Sediment The sediment used for the recovery experiments and the solidphase biodegradation test was obtained at low water from Loch Thurnaig, by Loch Ewe, Wester Ross, Scotland. It was wet sieved with sea-water through a 2 mm, then a 1 mm mesh sieve to remove large stones, molluscs, crustaceans, polychaetes and organic debris. The total organic carbon and nitrogen contents of the sediment, determined using a Model 2400 CHN elemental analyser (Perkin-Elmer, Beaconsfield, Buckinghamshire, UK), were 1.2% and 0.1% respectively.The moisture content was 20%. Preparation of Standard Solutions A concentrated stock solution of each SM base fluid in isohexane (approximately 25 mg ml21) was prepared for spiking clean sediment. A dilute stock solution (approximately 500 mg ml21) of each SM was also prepared in isohexane.Working solutions of each SM fluid were prepared by making dilutions of the appropriate stock SM solution to cover the range of spiking levels. Five standard solutions were prepared (2.5, 50, 100, 150 and 600 mg ml21) and each solution was analysed by gas chromatography with flame ionisation detection (GC–FID). Several of the SMs, specifically the nparaffin (Fig. 1) and ester [Fig. 6(a)], contained several discrete components. The composition for the individual components covered the range 4–49% for the n-paraffin and 2–63% for the ester. The total SM concentration of the most dilute solution was selected such that each individual component could be readily identified and quantified. The total peak area was plotted against concentration for each SM to give a calibration curve and the gradient of the curve and the correlation coefficient (r) were determined.Internal standard solutions (5000 and 50 mg ml21) were prepared in isohexane to check that these compounds were also recovered satisfactorily using the SM extraction method. A mixture of heptamethylnonane and squalane was used for the hydrocarbon SM fluids as these compounds have been shown to have response factors that are similar to those of hydrocarbons, give a linear response when analysed by GC–FID over the required concentration range and are the recognised standards for aliphatic hydrocarbon environmental analysis.Their retention times were distinct from all the determinants of interest. Methyl tricosanoate was used for the oxygenated SM fluids as it is chemically similar to the ester SM, was found to give a similar GC–FID response to the oxygen-containing determinants and gave a linear response over the required concentration range. Extraction of SM Fluid from Sediment: Recovery Experiment Many of the SMs are primarily hydrocarbons and the method was based on that for the isolation of mineral oils from sediment, which is well established2 and utilises a monophasic, binary solvent system.3 Sediment (5–20 g) was weighed into a 250 ml glass centrifuge tube and the appropriate amount of internal standard was added (Table 1).The sediment was then spiked to give SM fluid concentrations of 5, 100, 500 or 5000 mg g21. Clean sediment (20 g), to which only the internal standards were added, was also analysed with each spiking experiment and extracted in accordance with the method for a sediment estimated to contain @5 mg g21 SM.The SMs were extracted with a mixture of methanol (20 ml) and DCM (20 ml) by sonication for 5 min followed by centrifugation for 10 min at 2000 rpm and 2 °C. The supernatant was transferred into a separating funnel containing HPLC-grade water (18 ml) and shaken thoroughly for 2 min. The layers were allowed to separate and the lower, organic phase was transferred into a conical flask.The sediment was re-extracted with DCM (20 ml), repeating the ultrasonication and centrifugation. The solvent was decanted into the separating funnel and shaken. The lower, DCM layer was combined with the first extract and dried over anhydrous sodium sulfate (10 g) before concentration to approximately 0.5 ml by rotary evaporation. Solvent exchange was achieved by addition of isohexane (25 ml) followed by rotary evaporation to about 0.5 ml.This process was repeated a further twice before transferring the solution, with washings, into a vial or calibrated flask. The isohexane was concentrated by rotary evaporation and then under a stream of oxygen-free nitrogen, if necessary, to the appropriate volume (Table 1). External standard methodology was used to assess the recovery of the SMs; hence, for volumes !1 ml the samples were prepared in calibrated flasks. For smaller volumes, the actual volume of solvent was calculated by obtaining the mass of the final solution and converting to volume units using a value of 0.660 g ml21 for the density of isohexane.4 The SM fluids were separated from polar contaminants by normal-phase HPLC. An aliquot of the isohexane extract (150 ml) was injected on to a metal-free Genesis SIL 4 mm HPLC column (25 cm 34.6 mm id) with elution with isohexane at a flow rate of 2 ml min21.The hydrocarbons (IO, PAO, LAO and n-paraffin), which eluted with isohexane, were collected between 0 and 3 min following injection. For the acetal and ester the solvent was switched to isohexane–acetone (99 + 1 Table 1 Concentration specific variations used in the extraction of both hydrocarbon- and oxygen-containing SMs from sediments Expected Amount of Volume of Final extract concentra- Mass of standard/ extract prior volume for tion/mg g21 sediment/g mg g21 to HPLC/ml GC/ml @5 20 5 0.5 0.05* 100 10 250 1 0.25* 500 10 1000 5 1 5000 5 5000 20 2 * An increased final volume was used for the acetal; 0.2 ml (@5 mg g21) and 1.0 ml (100 mg g21). 1486 Analyst, December 1997, Vol. 122v/v) after 3.75 min and the fraction eluting between 3.75 and 20 min was collected in a round-bottomed flask. The eluate in both cases was concentrated to the appropriate volume (Table 1) before analysis by GC–FID. Extraction of 2-Ethylhexan-1-ol from Sediment One of the primary degradation products of the ester SM, 2-ethylhexan-1-ol, was also extracted and quantified. Extraction of 2-ethylhexan-1-ol from sediment samples was achieved as for the SM fluids, but using octan-1-ol as the internal standard since the response factor and linear range for octan- 1-ol were similar to those of 2-ethylhexan-1-ol. The concentrated extract was purified by HPLC using DCM as the mobile phase at a flow rate of 2 ml min21.An aliquot (150 ml) of the concentrated extract (final volume 1000 ml) was injected on to a metal-free column as before and 2-ethylhexan-1-ol was collected in the fraction eluting between 8 and 20 min. This fraction was collected in a round-bottomed flask and concentrated to an appropriate volume (300 ml) prior to analysis by GC–FID.GC–FID Analysis Samples (1 ml) were chromatographed on an HP 5890 Series II gas chromatograph equipped with an HP 7673 automated, cool on-column injector and fitted with a non-polar Ultra 1 column (25 m 3 0.2 mm id, film thickness 0.33 mm; Hewlett-Packard). The carrier gas was ECD-grade nitrogen (16 lb in22) and the oven temperature programmes were as detailed in Table 2.The detector was maintained at 300 °C throughout. Data were collected using a PE Nelson 600 series link box and processed using a Turbochrom 3 data station (Perkin-Elmer). Peaks were assigned as SM fluids on the basis of retention time relative to the original SM base fluid, summed as necessary and quantified on the basis of the relevant calibration curve covering the concentration range 2.5–600 mg ml21, taking into account the volumes prepared at each stage of the extraction.When analysing sea-bed samples or sediments from the biodegradation tests, quantification was on the basis of the added internal standard. The Turbochrom 3 data station takes into account the small differences between the analyte and the internal standard response during the calculations of unknowns. Furthermore, in the internal standard calculation, the sample volumes at each stage were not so critical and variations in recoveries are accounted for. 2-Ethylhexan-1-ol was determined by injecting an aliquot (1 ml) on to an HP 5880 gas chromatograph using an HP 7673A cool, on-column automated injector (Hewlett-Packard). A polar DB-23 fused-silica column (30 m 3 0.25 mm id, film thickness 0.25 mm; J & W Scientific, Folsom, CA, USA) was required for the analysis. The column oven was maintained at 40 °C for 5 min after injection and then programmed at 6 °C min21 to 180 °C.This temperature was held for 20 min then increased at 20 °C min21 to 220 °C and held there for 5 min. ECD-grade nitrogen was used as the carrier gas (16 lb in22). The temperature of the detector was 300 °C. Data processing was carried out using an HP 5880A Level 4 integrator, which computes the area of peaks, determines the chromatographic baseline and corrects the area accordingly. Determination of Limits of Detection Limits of detection were determined empirically by spiking clean sediment with concentrations of SM equivalent to 4–5% of the specified concentration for each method.The sediments were extracted according to the appropriate protocol. Each component of the SM, at its limit of detection, had to be clearly identifiable on the gas chromatogram and quantifiable. Determination of the Dry Mass of Sediments To enable the data to be quoted as mg g21 dry mass of sediment, sediment (10 g) was accurately weighed into an aluminium dish and heated at 80 °C in an oven for 22 ± 2 h.Previous unpublished data from this laboratory had shown this method to be applicable to a range of marine sediments containing contaminant hydrocarbons. Analysis of Sea-bed Survey Samples and Samples from a Biodegradation Test The above methodology was applied to samples from sea-bed surveys and from sediment samples from a solid-phase test designed to assess the biodegradability of the SM fluids.5–7 The sediments were extracted in a manner identical with that in the recovery experiments, having added the appropriate amount of internal standard, this being dependent on the expected concentration of the SM in the sediment (Table 1).Solvent blanks, prepared according to the method for sediments containing an estimated @5 mg g21 SM, were analysed with every batch of six samples. Peaks were assigned as SM fluids on the basis of retention time relative to the original SM base fluid and quantified using the internal standard.The solid-phase biodegradation test5–7 involved mixing clean sediment with six different SM fluids to give nominal concentrations of 100, 500 and 5000 mg g21 dry mass of sediment. The jars containing the spiked sediments were placed in a flowing sea-water system. At set times (0, 14, 28, 56 and 120 d) three equivalent jars were removed and the SM fluid was extracted using the above methodology, appropriate for each spiking level, and analysed by GC–FID.This allowed the determination of the primary degradation rates of the different fluid types. A number of sea-bed surveys have been carried out to monitor concentrations of SM fluids found in the environment. Samples from the environment can contain SM concentrations from in excess of 0.5% m/m down to trace amounts. An estimation of the amount of SM fluid that a sea-bed sample might contain can be made by relating it to the distance from where the sample was obtained to the drilling platform and this, together with the information on the expected type of SM, was used to select the appropriate method.Results and Discussion The overall objective was to produce a method, the basis of which was similar for all the SMs, which could be used to determine concentrations in excess of 0.5% m/m and as low as @5 mg g21. To this end, the basic extraction was the same in all cases and the variables were the starting mass of sediment, the volume of the extract prior to HPLC and the final volume of the test solution prior to GC–FID. Furthermore, the eluents for Table 2 Gas chromatograph injection temperatures and oven temperature profiles for the analysis of SMs Initial Final Total Injector hold Final hold run tempera- time/ Ramp/ tempera- time/ time/ SM ture/°C min °C min21 ture/°C min min Ester 60 0 8 300 15 45 Acetal 60 0 10 300 10 34 LAO 60 0 8 300 10 40 PAO 60 0 10 300 10 34 IO 60 0 10 300 15 39 n-Paraffin 60 3 4 300 30 93 Analyst, December 1997, Vol. 122 1487HPLC were varied depending on whether the test compound was a hydrocarbon SM or an oxygen-containing SM. At the higher concentrations of SM, interference from biogenic hydrocarbons can be ignored. However, at sites remote from platforms, or in a sediment where substantial biodegradation has occurred, the concentration of the SM is such that interference from the natural hydrocarbons can be a problem. Although the aliphatic hydrocarbon profile of a sediment is generally dominated by long-chain, odd-carbon alkanes including heptacosane, nonacosane and hentriacontane, the shorter chain alkanes (undecane to octadecane) are present at trace levels and cannot be readily distinguished from the equivalent compounds in the n-paraffin SM (Fig. 1). Under these circumstances it is necessary to review the individual aliphatic hydrocarbon profile and compare it with that for the SM so as to assess the relative abundance of the natural components.The problem is far less acute with the LAO, IO and PAO since they consist of discrete groups of molecules and give specific profiles when analysed by GC–FID (Figs. 2, 3 and 4, respectively), these being distinct from biogenic compounds. The methodology for the oxygen-containing compounds is such that aliphatic hydrocarbons are eluted from the HPLC column prior to the determinant and are therefore less of a problem when dealing with the very low SM concentrations.Although adventitious hydrocarbon contamination is always a complication with this type of analysis, it can be minimised by using the procedures described by Webster et al.8 All the SMs, with the exception of the acetal (Fig. 5), are mixtures and this obviously influences the limit of detection on the basis of total SM. This was most acute for the ester [Fig. 6(a)] and n-paraffin (Fig. 1). At the two higher concentrations, quantification of the SM at 4–5% of the specified concentration, including individual components in the case of the ester and n-paraffin, was possible.There were no problems with interfering compounds. The signal-to-noise ratio for the smallest components in both the ester and n-paraffin were approximately 30 : 1 and consequently the limit of determination was set at 25 and 200 mg g21 for the 500 and 5000 mg g21 methods, respectively (Table 3). At the two lower concentrations, the methodology was such that a larger amount of total SM was injected on to the GC column following isolation of the SM from sediment spiked with 5% of the specified concentration.The various components for each SM Fig. 1 Gas chromatogram of an n-paraffin SM extracted from a sediment obtained during a survey of the sea-bed close to an oil platform operating in the northern North Sea. The internal standard was heptamethylnonane (HMN). Fig. 2 Gas chromatogram of an LAO SM extracted from a sediment as part of a study into the biodegradation of SM fluids.The sediment, mean starting SM concentration 100.2 mg g21 dry sediment (SE = 2.9 mg g21, n = 3), was removed from the flowing sea-water test rig after 28 d. The concentration of the LAO (49.1 mg g21 dry sediment) was calculated from the combined peak areas for LAO1 and LAO2, using HMN as the internal standard. The internal standard for the hydrocarbon SMs included squalane (SQ). Fig. 3 Gas chromatogram of an extract from a sediment sample spiked with 5000 mg g21 of IO as part of a recovery experiment.The area of the two multiple peaks (IO1 and IO2) was summed prior to calculation of the SM concentration. The internal standard was HMN. Fig. 4 Chromatogram of an extract from a sediment sample spiked with 500 mg g21 of PAO as part of a recovery experiment. The total area of the partially resolved complex (PAO) was used to calculate the concentration of the SM fluid using squalane (SQ) as the internal standard. 1488 Analyst, December 1997, Vol. 122were again readily identified but there was minor interference from the biogenic hydrocarbons, specifically pentadecane, when determining the n-paraffin at both 250 ng g21 and 5 mg g21 using the @5 and 100 mg g21 methods, respectively. The smallest component in the ester, the C16 ester, was quantifiable at both test concentrations but the size of the peak was such that the limit of determination was set at 250 ng g21 and 5 mg g21 for the @5 and 100 mg g21 methods, respectively (Table 3).For the LAO, PAO and IO it was possible to detect 2.5 mg g21 of SM using the 100 mg g21 method. The consequence of the acetal being a single component (Fig. 5) was that the volumes of the final extract were increased for the @5 and 100 mg g21 methodologies (Table 1). The calibration curves for the internal standards and the various SMs were linear over the concentration range 2.5–600 mg ml21, equivalent to 2.5–600 ng on-column.The correlation coefficients of the calibration curves for all the SMs and internal standards were > 0.999. Differences in detector response between the analytes and the internal standard were taken into account by the data system when performing the internal standard calculation. There is a close similarity between the n-paraffin SM (Fig. 1) and the well established low-toxicity drilling fluids with respect to both carbon number range and distribution.9 A marked difference between the two types of drilling fluid is, however, the absence of branched isomers in the SM.As the analytical method was based on the technique for the analysis of mineral oil-type base fluids, it would be expected to be suitable for the n-alkane-type SM. The combination of number of base fluids studied and the range of concentrations investigated meant that only single recovery experiments were performed for each SM at each concentration. Although this limits the available data, it does show that recoveries in excess of 70% were achieved at the four concentrations for all the SMs (Table 3).The mean recoveries for all the SMs at 5, 100, 500 and 5000 mg g21 were 84, 92, 95 and 92%, respectively. An analysis of variance showed that there was no significant difference between the means at each of the four concentrations or between the recoveries of the individual SM fluids. The mean recovery of the n-paraffin was 86% [standard error of the mean (SE) = 6.0%, n = 4].The only SM which gave a lower mean recovery was the ester at 84% (SE = 5.6%, n = 4). All other SMs gave a mean recovery of > 90%, the highest being the acetal at 98% (SE = 3.0%, n = 4). During the analysis of samples from the solid-phase test, duplicate analyses were performed on three equivalent test samples which had been dosed with the n-paraffin at 100 mg g21 SM and maintained in the flowing sea-water system for 120 d. The mean concentration was 17.8 mg g21 (SE = 1.2 mg g21, n = 6; Table 4).Quadruplicate analyses of sea-bed survey samples containing the n-paraffin gave a mean of 320 mg g21 Fig. 5 Chromatogram showing the single peak of the acetal SM fluid extracted from a sea-bed survey sample collected close to an installation operating in the northern North Sea. The internal standard was methyl tricosanoate (23 : 0). Fig. 6 Gas chromatograms of (a) the five components of the ester SM (E1- E5) and (b) 2-ethylhexan-1-ol.The ester was extracted from a sediment as part of a study into the biodegradation of SM fluids. The fifth component (E5) is the smaller of the two peaks eluting around 24.1 min. The internal standard was methyl tricosanoate (23:0). The 2-ethylhexan-1-ol was extracted from a sediment after 28 d in a flowing sea-water solid-phase test. The concentration was calculated using octan-1-ol as the internal standard. Table 3 Summary of recovery data for each of the SMs at the four different concentrations together with the concentration specific limit of determination (LD).The mean recovery across the four concentrations together with the associated standard error of the mean (SE) is also presented Recovery (%) Spiked synthetic fluid concentration/ mg g21 dry sediment SM @5 100 500 5000 Mean SE Ester 76 92 73 95 84 5.6 Acetal 91 95 99 105 98 3.0 LAO 84 101 100 91 94 4.0 PAO 97 87 92 82 90 3.2 IO 81 104 99 83 92 5.7 n-Paraffin 78 74 91 100 86 6.0 LD/mg g21 0.25 5 25 200 Analyst, December 1997, Vol. 122 1489(SE = 4.9 mg g21; Table 4), indicating good reproducibility of the method for both the biodegradation test system and sea-bed survey samples. Triplicate jars of sediment, originally spiked with a nominal concentration of 100 mg g21 dry sediment of LAO (Fig. 2) in the solid-phase test system, were found to contain 49.1, 45.1 and 46.6 mg g21 dry sediment after 28 days, giving a mean of 46.9 mg g21 dry sediment (SE = 1.2 mg g21, n = 3).Equivalent replication was obtained at the higher concentrations (Table 4). As with the n-paraffin, the data were calculated against HMN. The chromatogram of the IO (Fig. 3), taken from the recovery experiment, shows a similar profile to that of the LAO. Both of these SMs are alkenes but the IO is a combination of C16 and C18 alkenes whereas the LAO consists of C14 and C16 alkenes. Furthermore, the mixture of components of the IO appears to be more complex, with both the peaks having the appearance of being composites rather than individual components, as is the case with the LAO.Sediment from a solid-phase test, spiked with the IO and maintained in flowing sea-water for 56 d, was found to contain 5050 mg g21 dry sediment (SE = 12 mg g21, n = 3). The PAO was only partially resolved under the GC conditions employed. Nevertheless, this did not affect the quantification of the total SM as there were distinct start and end points for the unresolved complex mixture (UCM), as is evident for the sample spiked with 500 mg g21 of PAO as part of the recovery experiments (Fig. 4). The acetal was the simplest of all the synthetic mud base fluids containing only a single compound (Fig. 5). It was slightly more polar than the hydrocarbons, requiring 1% v/v of acetone in the HPLC mobile phase and more time to elute from the column. The same extraction procedure could be applied, although increased volumes were required before GC, owing to the simple composition of the SM.The acetal gave the highest mean recovery across the four concentrations at 98% (SE = 3.0%, n = 4). Good reproducibility was evident in replicate analyses from the solid-phase test with standard errors of 2–12 mg g21 (n = 3) for sediment taken from three individual jars at four time intervals with concentrations of approximately 500 mg g21 SM. Similar replication was obtained at other concentrations (Table 4). Methyl tricosanoate was well resolved from the determinant (Fig. 5). This standard was also resolved from the five different components of the ester SM fluid [Fig. 6(a)]. This particular sample, which originally contained 500 mg g21 of SM, was found to have an ester concentration of 130 mg g21 after 56 d in the flowing sea-water system. The polarity of the ester was only slightly greater than that of the acetal, hence the same extraction and clean-up procedure could be employed, yielding satisfactory results, with recoveries ranging from 73 to 95% (Table 3).Marine sediments contain bacteria which produce enzymes which are readily able to cleave ester bonds. One of the primary degradation products of the ester SM is 2-ethylhexan-1-ol and, because of the very rapid degradation of the ester in marine sediments, it was deemed appropriate to utilise a degradation product as part of the environmental assessment of this SM. Extraction of the alcohol was achieved using the same method as for the SMs.It was necessary, however, to use an alternative HPLC procedure which required a more polar mobile phase. Similarly, a more polar capillary column was required for the GC analysis but good peak shape and resolution of the determinant and internal standard were achieved using a DB-23 column [Fig. 6(b)]. This particular example originated from a sediment which had been incubated for 28 d in a flowing seawater system. The sediment was initially spiked with 500 mg g21 of the ester SM. After 28 d, 75.6 mg g21 of 2-ethylhexan- 1-ol were found in the sediment extract. The method developed for the analysis of SM fluids in sediment was readily applied to all the generic types of SM base fluids currently in use in the off-shore industry.It was also possible to extract 2-ethylhexan-1-ol, a primary degradation product of the ester, from sediment using this method. The same extraction procedure was used for all test compounds and only the HPLC conditions were varied for the different fluids. By altering the volumes before HPLC and GC–FID, the method could be used to quantify a wide range of concentrations (from < 1 mg g21 to > 5000 mg g21) of fluids in sediment samples. Recoveries of > 80 % were achieved in 82% of the recovery experiments. This method has been used successfully to monitor SM fluid concentrations in samples from sea-bed surveys, despite the different sediment types and range of concentrations encountered, and in a solid-phase test developed to monitor the biodegrability of these compounds. Investigations into adapting the method for the determination of SM fluids in biota are under way together with studies into the application of the extraction methodology for the isolation of olive oil from sediment, this triglyceride oil being used as a positive control in the solid phase test. References 1 Davies, J. M., Addy, J. M., Blackman, R. A., Blanchards, J. R., Ferbrache, J. E., Moore, D. C., Sommerville, H. J., Whitehead, A., and Wilkinson, T., Mar. Pollut. Bull., 1984, 15, 363. 2 Stagg, R. M., McIntosh A., and Mackie, P. R., Aquat. Toxicol., 1995, 33, 245. 3 Bligh, E. G., and Dyer, W. J. A., Can. J. Biochem. Physiol., 1959, 37, 911. 4 Handbook of Chemistry and Physics, ed. Weast, R. C., CRC Press, Cleveland, OH, USA, 69th edn., 1988–89. 5 Croce, B., McIntosh, A. D., Moffat, C. F., Hird, S. J., and Stagg, R. M., Biodegradation of Base-Oils Used in Synthetic Drilling Muds in an Experimental Solid-Phase System, Fisheries Research Services Report No. 3/96, Marine Laboratory, Aberdeen, 1996. 6 Munro, P. D., Moffat, C. F., Couper, L., Brown, N. A., Croce, B. and Stagg R. M., Degradation of Synthetic Mud Base Fluids in a Solid- Phase Test System, Fisheries Research Services Report No. 1/97, Marine Laboratory, Aberdeen, 1997. 7 Munro, P. D., Moffat, C. F., and Stagg, R. M., in Contributing to Environmental Progress—Getting the Message Across, Proceedings of SPE/UKOOA European Environmental Conference, Aberdeen, Scotland, 1997, Society of Petroleum Engineers, Dallas, TX, USA, p. 205. 8 Webster, L., Angus, L., Topping, G., Dalgarno, E. J., and Moffat, C. F., Analyst, 1997, 122, 1491. 9 Parker, J. G., Howgate, P., Mackie, P. R., and McGill, A. S., Oil Chem. Pollut., 1990, 6, 263. Paper 7/05975B Received August 14, 1997 Accepted October 13, 1997 Table 4 Examples of analyses of sediments from a solid-phase biodegradation test and from sea-bed surveys. The sediments from the solid-phase test were incubated in a flowing sea-water system for 120 d SM concentration (mean ± SE/mg g21 dry sediment) Solid phase biodegradation test (n = 3) Starting concentration/mg g21 dry Sea-bed sediment surveys SM 100 500 5000 (n = 4) n-Paraffin 17.8 ± 1.2* 446 ± 38 6101 ± 58 320 ± 4.9 LAO 12.1 ± 0.6 229 ± 17 4391 ± 309 1946 ± 46 Acetal 88.4 ± 4.4 480 ± 2 4553 ± 169 — * Result from three jars each analysed in duplicate. 1490 Analyst, December 1997, Vol. 122

ISSN:0003-2654    

DOI:10.1039/a705975b

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Long-term Monitoring of Polycyclic Aromatic Hydrocarbons in Mussels ( Mytilus edulis) Following the BraerOil Spill† Lynda Webstera, Lindsay Angusb, Graham Toppinga, Eric J. Dalgarnoa and Colin F. Moffat*a a FRS Marine Laboratory Aberdeen, P.O. Box 101, Victoria Road, Aberdeen, UK AB11 9DB b North Atlantic Fisheries College, Port Arthur, Scalloway, Shetland, UK ZE1 0UN On January 5, 1993, 84 700 t of Norwegian Gullfaks crude oil was released into the coastal region of south Shetland when the tanker MV Braer grounded at Garths Ness.A Fisheries Exclusion Zone was designated under the Food and Environment Protection Act 1985 (FEPA), prohibiting the taking or harvesting of fish or shellfish within the Zone so as to prevent contaminated products reaching the marketplace. The criteria set for lifting of the Order were that the particular species must be free from any petrogenic taint and the concentration of polycyclic aromatic hydrocarbons (PAHs) must be within the range for reference samples.Between April 1993 and February 1995 the Order was progressively lifted for wild fish, salmon, crustacea, excluding Nephrops norvegicus (Norway lobster), and molluscs, with the exception of mussels. As part of the monitoring exercise, mussels from a reference site were transplanted in June 1995 to three sites within the Zone, where they were suspended in plastic mesh boxes from rafts to a depth of 5 m. Samples were collected at regular intervals over the following 12 months and the concentration and composition of PAHs were determined by gas chromatography with mass spectrometric detection.The total measured PAH concentration at the control site increased from 13.7 to 66.1 ng g21 wet mass of tissue between June 1995 and February 1996. This trend was reversed by July 1996 when the PAH concentration was 12.8 ng g21. The mean across the year for the control site was 24.0 ng g21 (SE = 8.9 ng g21, n = 6). A similar seasonal trend in PAH concentration over the year was observed at all sites within the Zone, but the PAH concentration was consistently greater at these sites, reaching a maximum concentration of 316 ng g21 in February 1996.Although no taint was detected in any of the mussels, these results meant that it was not possible to lift the Prohibition Order for mussels. Further monitoring at three sites outwith the Zone and three sites within the Zone is under way together with investigations into the specific source of the PAHs.Keywords: Polycyclic aromatic hydrocarbons; mussels; gas chromatography–mass spectroscopy; oil spill The tanker MV Braer grounded on Garths Ness, Shetland, on January 5, 1993. Over the following 7 d the entire cargo of approximately 85 000 t of Norwegian Gullfaks light crude oil, a naturally biodegraded oil resulting in a more naphthenic and aromatic crude, was released from the ship, together with some bunker fuel oil. A Fisheries Exclusion Zone was designated on January 8, 1993, by Order under the Food and Environment Protection Act 1985 (FEPA).The Order prohibited the harvesting of farmed or wild fish or shellfish within the Zone to prevent contaminated products reaching the marketplace. The Zone was extended 5 miles westward on January 27, 1993 (see Fig. 1).1 The Scottish Office Agriculture, Environment and Fisheries Department (SOAEFD) instituted a programme of † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997.Fig. 1 Map of Shetland showing the FEPA Exclusion Zone together with the location of the reference site at Olna Firth (R1) and the sites within the Zone (Z1–Z3). Z1, Sandsound Voe; Z2, Stromness Voe, which contained a northerly (N) and southerly (S) site; and Z3, Merry Holm/Trondra. The Braer grounded at Garths Ness on the southerly tip of the Shetland Islands close to the Bay of Quendale. Analyst, December 1997, Vol. 122 (1491–1495) 1491marine monitoring with the aims of ensuring that the original decision to institute the FEPA Zone was a sensible one, that its limits had been correctly drawn and to provide support for any decision-making process with regard to the possible extension or future lifting of the Zone. The criteria for revoking the FEPA Order for fish and shellfish were first, that the fish and shellfish from within the Zone should not contain taint that is associated with crude oils and petroleum fractions, and second, that the concentration of aliphatic hydrocarbons and PAHs in fish and shellfish from within the Zone should fall within the background range of values for fish and shellfish outside the Exclusion Zone.2 It was apparent from the first analytical measurements of hydrocarbons in fish and shellfish that the concentration of aliphatic hydrocarbons in samples from within the Zone was generally low and that the distribution of these hydrocarbons in Gullfaks crude was not distinctive.Therefore, in terms of the aims of the programme, attention was focused on the PAHs. Naphthalenes (75.6%), phenanthrenes (12.5%) and dibenzothiophenes (5.4%) were the major PAHs in Gullfaks crude oil with lesser amounts of the fluoranthenes and pyrenes (4.2%), benzophenanthrenes, benzanthracenes and chrysenes (1.5%) and five- and six-ring PAHs (0.7%).2 As a result of the monitoring programme, the Order was lifted for wild fish on April 23, 1993, farmed salmon on December 8, 1993, crustaceans, with the exception of Nephrops norvegicus (Norway lobster), on September 30, 1994, and molluscs, with the exception of mussels, on February 9, 1995.The blue mussel (Mytilus edulis), a circumboreal species,3 is farmed from rafts in the sea lochs around the Shetland Islands. Some of these mussel farms were within the Exclusion Zone. Mussels from the western area of the Exclusion Zone were sampled during 1993 and 1994 and the PAH concentration, defined as the combined concentration of the two- to six-ring parent and branched PAHs, was determined in the edible tissue.A total measured PAH concentration of 1450 ng g21 wet mass of tissue was determined in a sample taken from Stromness Voe in March 1993, this being the greatest concentration found in mussels.2 The mean PAH concentration in the samples collected from various locations within the Zone during October 1994 was 221 ng g21 (SE = 60 ng g21, n = 5).This contrasts with a mean for reference samples, collected from a mussel farm in Olna Firth (March 1993) and from Aith Voe (February, June and October 1994), of 56.8 ng g21 (SE = 17.2 ng g21, n = 4).2 Mussels are known to accumulate trace contaminants, such as heavy metals and hydrocarbons, present in the water column.4 As such, it was concluded that mussels could be utilized as an organism for monitoring long-term hydrocarbon pollution in marine waters.5 For these reasons, the decision was taken to transplant animals from a site well outside the Zone to various sites within the Zone and to monitor the change in PAH concentration with time.Experimental Reagents Methanol, isohexane, dichloromethane and acetone were glassdistilled reagents specifically prepared for hydrocarbon analysis by Rathburn Chemicals (Walkerburn, UK). HPLC-grade water was also supplied by Rathburn Chemicals. Individual batches of all solvents were checked for contaminants as described previously.6 Analytical-reagent grade nitric acid was purchased from BDH (Poole, Dorset, UK).Sodium chloride, sodium hydroxide and anhydrous sodium sulfate were analyticalreagent grade reagents from Fisons Scientific Equipment (Loughborough, UK). Deuteriated naphthalene, biphenyl, dibenzothiophene, anthracene, pyrene and benzo[a]pyrene were obtained from C/D/N/Isotopes through K&K-Greeff (Croydon, UK). Preventative Measures for Reducing Casual PAH Contamination PAHs are ubiquitous in the environment and great care must be taken to avoid adventitious contamination of samples.To this end, all glassware was washed and dried in a GW 4000 glassware washer (Camlab, Cambridge, UK). Prior to use, the glassware was rinsed twice with dichloromethane and then twice with isohexane, the latter being allowed to evaporate before proceeding. The columns used for the sodium sulfate filtration were soaked at regular intervals in concentrated nitric acid to clean the frits.The columns were then flushed with copious volumes of water before being washed as described above. The sodium sulfate was cleaned by washing with isohexane in an ultrasonic bath for 10 min. The solvent was decanted to waste and the sodium sulfate placed in an oven at 110 °C overnight. The use of Socorex PTFE-lined pipettes (Camlab) with disposable glass Pasteur pipettes, the minimum presence of any plastics, a strict regime for storage of samples, environmental control of the laboratory and assignment of all equipment to specific areas of analysis are all further precautions taken to avoid such contamination.Test Sites The test sites included a reference site at Olna Firth (Site R1, Fig. 1) and three sites within the Exclusion Zone: Sandsound Voe (Site Z1), Stromness Voe (Site Z2, two farms) and Merry Holm (Site Z3). Three of the sites, R1, Z1 and Z2, were located within discrete voes which were remote from any urban or industrial areas, were associated with only one minor road and would have only small-boat traffic. In contrast, Merry Holm (Z3) was a more open, in-shore site approximately 2.5 km south west of Scalloway harbour.Samples of sediment were collected at each site, from the Fisheries Research Vessel Clupea, using a Day grab. The sediments at Z3 were characterised as medium sand/shell sand whereas those at Sandsound (Z1) were a mixture of fine sand and mud.The reference site and Stromness Voe (Z2) contained muddier sediments. Total organic carbon concentration (mean ± s, n = 6) for the sediments, determined using a Model 2400 CHN Elemental Analyser (Perkin-Elmer, Beaconsfield, Bucks., UK), was 5.080 ± 0.807%, 5.155 ± 1.082% and 6.160 ± 0.678% for Sandsound Voe, Stromness Voe and Olna Firth, respectively. The total organic carbon content of the sediment around Merry Holm was lower at 1.384 ± 0.124%. Mussels Mussels (Mytilus edulis), of uniform age, were transplanted from the reference mussel farm in Olna Firth (Site R1) in mid- June 1995 to the three sites within the Zone (Fig. 1). Samples of approximately 70 mussels were placed in numbered plastic mesh boxes and suspended from rafts to a depth of 5 m. Sampling, which comprised collection of one of the mesh boxes, took place during August 1995, October 1995, February 1996 and June/July 1996. The mussels were thoroughly iced and dispatched to Aberdeen by overnight ferry.On arrival in Aberdeen, approximately 20 mussels were removed for sensory assessment. The tissue was removed from the shell of the bulk of the remaining animals, combined and homogenised. A portion was taken for chemical analysis. All residual material was packaged and stored at 230 °C in case repeat analysis was required. Isolation of Hydrocarbons The method was based on that of Grimmer and B�ohnke.7 To a homogenised sample of mussel (10 g) were added the aliphatic hydrocarbon internal standards heptamethylnonane and squalane (approximately 3.2 mg of each).A mixture of deuteriated 1492 Analyst, December 1997, Vol. 122naphthalene, biphenyl, dibenzothiophene, anthracene, pyrene and benzo[a]pyrene (100 ml; approximately 1 mg ml21 each) was then added. This was mixed with sodium hydroxide (10% m/v) in methanol–water (9 + 1 v/v; 40 ml) and 3–5 pre-washed anti-bumping granules. The mixture was refluxed for 3 h 45 min before the addition of water (10 ml) and then refluxing was continued for a further 15 min.The resulting hot solution was extracted with isohexane (80 ml) following the addition of methanol–water (4 + 1 v/v; 40 ml). A second extraction of the aqueous solution with isohexane (80 ml) was performed. The first organic extract was washed with methanol–water (1 + 1 v/v; 40 ml) and this aqueous solution was then used to wash the second organic extract. The two isohexane extracts were combined and washed with water (3 3 40 ml).The resulting organic solution was dried by passage through a column (11 3 1.5 cm id) containing sodium sulfate (approximately 60 g). The column was washed with isohexane (50 ml) and the combined solvent concentrated to approximately 300 ml by rotary evaporation (water bath, < 30 °C). The concentrate was transferred into a vial and concentrated back to approximately 300 ml under a stream of scrubbed nitrogen. The PAHs were isolated from the aliphatic hydrocarbons by isocratic normal-phase HPLC.An aliquot (150 ml) of the concentrated isohexane solution was injected on to a Genesis SIL 4 mm HPLC column (25 3 0.46 cm id) (Jones Chromatography, Hengoed, UK) and eluted with isohexane at 2 ml min21. The aliphatic fraction was collected between 0 and 2.75 min and the aromatic fraction between 2.75 and 20 min. The resulting eluates were separately concentrated under reduced pressure prior to transfer to a chromatographic vial insert (Hewlett- Packard, Stockport, UK), where they were further concentrated to approximately 15–20 ml under a stream of scrubbed nitrogen.The sides of the vial insert were carefully washed down with the concentrate before being capped. A procedural blank was analysed with each batch of samples. Determination of Polycyclic Aromatic Hydrocarbons The concentration and composition of the PAHs were determined by gas chromatography with mass spectrometric detection (GC–MS). Samples (1 ml) were chromatographed on an HP 5890 Series gas chromatograph equipped with an HP 7673A oncolumn injector and fitted with a fused silica capillary column (25 m 3 0.2 mm id) coated with a 0.33 mm film of Ultra 1, a cross-linked methylsilicone gum (Hewlett-Packard).Injections were made at 50 °C and the oven temperature was held constant for 3 min, after which it was increased at 20 °C min21 to 100 °C. This was followed by a slower ramp of 4 °C min21 up to a final temperature of 270 °C.The oven temperature remained constant until the end of the analysis. Helium (10 lb in22) was used as the carrier gas. The gas chromatograph was interfaced with an HP 5970 Series mass selective detector (Hewlett-Packard), which was set for selective ion monitoring (SIM) with a dwell time of 50 ms. A total of 25 ions plus the six internal standard ions were measured over the period of the analysis, as detailed previously. 2 Thus, the analysis incorporated two- to six-ring, parent and branched PAHs.This does not cover all of the many PAH compounds that exist. Thus, all references to PAH concentrations and distributions relate to the measured PAHs, details of which are presented in Table 1. Perfluorotributylamine was used as the mass spectrometric calibrant. Standards for all the parent and branched PAHs cannot be obtained, but the limit of detection, calculated as three times the standard deviation of the mean value from six procedural blanks, was found to be < 0.2 ng g21 for benzo[k]fluoranthene and benzo[a]pyrene and < 0.3 ng g21 for chrysene.Good reproducibility was generally obtained for individual PAHs (Table 1). Further quality control was assured through participation in the PAH programme of QUASIMEME (Quality Assurance of Information for Marine Environmental Monitoring in Europe). Assessment of Taint The mussels were steamed for approximately 4 min or until the shells were completely open.At this point the meat was transferred to a lidded casserole and kept warm on an electric hot-plate for the duration of the tasting session (15–20 min). The mussels were assessed ‘blind’ by each member of a panel Table 1 Duplicate determination of the PAHs (ng g21 wet mass of tissue) in a sample of mussel tissue. The numbers following the name refer to the molecular mass. The sum of the molecular mass groups is also presented. The total PAH concentration was 316.3 and 328.5 ng g21 for samples 1A and 1B respectively Sample Sample PAH 1A 1B PAH 1A 1B Naphthalene 0.9 0.6 C3 202 11.5 11.4 C1 Naphthalenes 1.0 0.9 Sum of 202s 55.0 69.1 C2 Naphthalenes 6.1 5.3 Benzo[c]phenanthrene (228) 0.8 0.9 C3 Naphthalenes 16.4 17.3 Benz[a]anthracene (228) 1.7 4.4 C4 Naphthalenes 0.8 0.5 Chrysene + triphenylene (228) 4.7 1.3 Sum of naphthalenes 25.24.6 Benz[a]anthracene (228) nd nd Phenanthrene 5.5 5.6 C1 228 7.2 6.3 Anthracene nd* nd C2 228 3.1 3.4 C1 178 31.8 32.6 Sum of 228s 17.5 16.3 C2 178 63.9 60.4 Benzofluoranthene (252) 8.0 6.7 C3 178 50.3 49.7 Benzo[e]pyrene (252) 3.7 4.0 Sum of 178s 151.5 148.3 Benzo[a]pyrene (252) 1.0 0.9 DBT† 0.5 0.6 Perylene (252) 1.4 1.3 C1 DBT 8.3 8.1 C1 252 3.1 3.1 C2 DBT 22.2 21.7 C2 252 nd 2.0 C3 DBT 16.2 18.3 Sum of 252s 17.2 18.0 Sum of DBTs 47.2 48.7 Indenopyrene (276) 1.2 1.2 Fluoranthene (202) 4.2 4.8 Benzoperylene (276) 1.5 1.8 Pyrene (202) 4.0 4.6 C1 276 nd 0.5 C1 202 17.3 14.8 C2 276 nd nd C2 202 18.0 33.5 Sum of 276s 2.7 3.5 * nd, Not detected.† DBT, dibenzothiophene. Analyst, December 1997, Vol. 122 14930 59 136 247 361 (Jun '95) (Aug '95) (Oct '95) (Feb '96) (Jun '96) Z2(N) Z2(S) Z1 R1 350 300 250 200 150 100 50 0 Time/d [PAH]/ng g–1 of 8–10 staff trained to recognise petroleum-derived taints by odour and taste and scored as described by Whittle et al.1 Results and Discussion It is essential in any environmental monitoring programme to have a benchmark against which the sites of potential contamination can be assessed.Consideration must also be given to the fact that the test matrix is a living, growing animal. The objective of transferring samples from a single reference site to several locations within the Zone was to ensure that there was a commonality for the test matrix. The total measured PAH concentration of the mussels at time of transfer was 13.7 ng g21. All the individual parent PAHs and associated branched groups were present at < 2 ng g21 and many were not detected.The largest individual grouping was the three-ring compounds (phenanthrene, molecular mass 178, and the C1–C3 substituted compounds) which comprised 38% of the PAHs. The naphthalenes comprised 24% of the PAHs and the five-ring compounds 18%. The concentration of PAHs at the control site increased progressively over 8 months to a maximum of 66.1 ng g21 in the sample collected in February 1996 (Fig. 2). This was a result of an increase in the concentration of the four- to six-ring compounds, the concentration of the naphthalenes and three-ring PAHs remaining relatively consistent over the year. The greatest PAH group concentration was 19.7 ng g21, determined for the five-ring PAHs isolated from the February sample. The observed increase in PAH concentration over autumn and winter was reversed by July 1996, when the PAH concentration was found to be 12.8 ng g21. This resulted from a decrease in the concentration of the larger ring PAHs.Thus, the proportion of two- and three-ring compounds again dominated the PAH profile. The mean PAH concentration across the year for the control site was 24.0 ng g21 (SE = 8.9 ng g21, n = 6). No taint was detected in any of the mussels from the reference site. After 2 months, an increase in the PAH concentration was detected in the mussels from both sites in Stromness Voe (Site Z2) but not the more northerly site at Sandsound Voe (Site Z1, Fig. 2). This change was a result of an increase in the concentration of all PAH groupings but there was a relative decrease in the proportion of naphthalenes present. The PAH distributions at the two sites in Stromness Voe were very similar. The three-ring compounds still dominated at 38% but the naphthalenes comprised only 18% and 19% of the PAHs at the northerly and southerly sites, respectively. In contrast, the proportion of four-ring compounds had increased from 15% at the time of transplanting to 26% at the northerly site and 25% at the southerly site.All sites within the Zone displayed a progressive increase in PAH concentration between August 1995 and February 1996 with a maximum concentration of 316 ng g21 in Stromness Voe (Fig. 2). Neither this, nor any other sample from within the Zone, was found to be tainted. The dominant group was the C16 four-ring PAHs (pyrene, fluoranthene and C1–C3 substituted compounds), which comprised approximately 30% of the PAHs (Table 2).Although there was no August sample for site Z3, samples were obtained from this location during October 1995 and February 1996 when the PAH concentrations were determined to be 133 and 279 ng g21, respectively. Thus, the trend was maintained at this site. Although the concentration in the October sample from Z3 was similar to that of the northerly site in Stromness Voe, the PAH distribution was distinct, having 35% naphthalenes.In contrast, the PAH distribution of the February sample from Z3 was similar to the other sites within the Zone (Table 2). Indeed, the percentage compositions of the PAHs across the sites within the Zone for the samples collected in February were very similar. The mean percentage (with standard error) for the various PAH groupings at the four sites were: naphthalenes 4% (0.6%); 178, 18% (0.5%); DBTs 2% (0.3%); 202, 30% (0.6%); 228, 18% (0.4%); 252, 24% (0.8%); 276, 5% (0.4%).Severe storms resulted in the loss of some of the mesh boxes from the rafts. This meant that there was only one Zone sample left, at the northerly site in Stromness Voe, in the summer of 1996. As with the equivalent reference sample, a decrease in PAH concentration was observed, relative to the February sample, but the value of 108 ng g21 was greater than that at the reference site. As with the mussels from Olna Firth, the dominant PAH grouping was the three-ring compounds at 40% with the naphthalenes comprising 20% of the measured PAHs.There was an apparent seasonal trend for the PAH concentration in mussels, the concentration increasing over winter and declining in the spring. From autumn until spring lipids may be saved for gametogenesis8 and this increase in lipid content would permit the retention of increased amounts of lipophilic compounds such as PAHs. The maturation of the gametes is Fig. 2 Variation in PAH concentration (ng g21 wet mass of tissue) with time for mussels collected from a reference site (R1) and from sites within the Zone.In all cases there is a progressive increase in PAH concentration with time over the autumn and winter months. The final samples from both the reference site and northerly site in Stromness Voe [Z2(N)], collected in early summer, show a decrease relative to the February sample. The PAH concentrations are consistently greater at sites within the Zone relative to the reference site.Table 2 Percentage distribution of the various PAH groupings for the PAHs isolated from mussels collected at sites within the Exclusion Zone during February 1996 Site within the Exclusion Zone PAH group* Z1 Z2 (North) Z2 (South) Z3 Naphthalenes 4 6 3 4 178 17 19 18 19 DBT 1 1 2 2 202 31 29 30 28 228 19 17 18 18 252 25 23 22 25 276 4 5 6 5 * Naphthalenes, naphthalene and C1–C4 branched compounds; 178, phenanthrene/anthracene and C1–C3 branched compounds; DBT, dibenzothiophene and C1–C3 branched compounds; 202, C16 four-ring PAHs and C1–C3 branched compounds; 228, C18 four-ring PAHs and C1–C2 branched compounds; 252, C20 five-ring PAHs and C1–C2 branched compounds; 276, C22 six-ring PAHs and C1–C2 branched compounds. 1494 Analyst, December 1997, Vol. 122under several exogenous (e.g., water temperature, lunar periodicity, depth, food abundance and availability, light intensity) and endogenous (e.g., genetic, hormonal) controls. Of all the external controls, temperature is generally regarded as the most important and the act of spawning is initiated when the water temperature generally exceeds 10–12 °C for Mytilus edulis,9 although in Shetland the relevant temperature range is regarded as 8–12 °C.The principal spawning time is, thus, late spring/ early summer, but it is obviously affected by the prevailing weather conditions in any year. During 1996, some mussels in Shetland were observed to spawn in early June and this coincides with the decrease in PAH concentration.The seasonal trend can, therefore, be related to physical and chemical changes in the animals. In order for bioaccumulation of PAHs to occur, these compounds must be bioavailable. PAHs may be present dissolved in the water or in the sediment where they are incorporated in the biota, associated with particulate organic and inorganic matter, associated with dissolved organic matter and truly dissolved in the interstitial water.Filter-feeding molluscs are exposed to both water and suspended particulate material (SPM), which would include resuspended sediments. The relative importance of dissolved and particulate PAHs as a source for PAH accumulation in mussels is uncertain. Pruell et al.10 concluded that the dissolved phase was the direct source of contaminants accumulated by mussels. In contrast, Naes et al.11 suggested that particulate PAH is the main source for mussels. This being so, winter storms would result in a degree of resuspension of the sediments, which, if they contained PAHs, would mean an increase in the exposure of the animals to these organic xenobiotics over the winter.The higher concentrations observed for sites within the Zone could result from either a greater exposure to SPM or a greater concentration of PAHs within the SPM. The primary objective of the monitoring programme was to provide data that enabled pertinent decisions to be made with regard to a change in status of the Exclusion Zone.It was evident that the PAH concentration for mussels within the Zone was consistently greater than for the reference site, hence it was not possible to lift the Zone for mussels. Further monitoring is necessary and a second, 3 year, mussel experiment was initiated in September 1996. The seasonal variation, evident for the PAHs, illustrates the clear need for pertinent reference sites which must be sampled at the same time as the test sites and to this end the new experiment incorporates a further two reference sites, at Mangaster Voe and Vaila (Fig. 1). Studies into the source of the PAHs have been initiated and additional investigations into the association of fat content with PAH concentration are under way. The authors acknowledge Mrs. Nicky Shepherd for her assistance with the preparation of the mussels for determination of taint and all colleagues on the sensory panels. References 1 Whittle, K. J., Anderson, D.A., Mackie, P. R., Moffat, C. F., Shepherd, N. J., and McVicar, A. H., in The Impact of an Oil Spill in Turbulent Waters: The Braer, ed. Davies, J. M., and Topping, G., The Stationery Office, Edinburgh, 1997, ch. 11. 2 Topping, G., Davies, J. M., Mackie, P. R., and Moffat, C. F., in The Impact of an Oil Spill in Turbulent Waters: The Braer, ed. Davies, J. M., and Topping, G., The Stationery Office, Edinburgh, 1997, ch. 10. 3 Jamieson, G. S., World Aquacult., 1989, 20, 94. 4 Murray, A. P., Richardson, B. J., and Gibbs, C. F., Mar. Pollut. Bull., 1991, 22, 595. 5 Fossato, V. U., and Canzonier, W. J., Mar. Biol., 1976, 36, 243. 6 McGill, A. S., Moffat, C. F., Mackie, P. R., and Cruickshank, P., J. Sci. Food Agric., 1993, 61, 357. 7 Grimmer, G., and B�ohnke, H., J. Assoc. Offic. Anal. Chem., 1975, 58, 725. 8 Voogt, P. A., in The Mollusca, Metabolic Biochemistry and Molecular Biomechanics, ed. Hochachka, P. W., Academic Press, New York, 1983, ch. 7. 9 Mackie, G. L., in The Mollusca, Reproduction, ed. Tompa, A. S., Verdonk, N. H., and van den Biggelaar, J. A. M., Academic Press, New York, 1984, ch. 5. 10 Pruell, R. J., Lake, J. L., Davis, W. R., and Quinn, J. G., Mar. Biol., 1986, 91, 497. 11 Naes, K., Bakke, T., and Konieczny, R., Mar. Freshwater Res., 1995, 46, 275. Paper 7/05977I Received August 14, 1997 Accepted October 10, 1997 Analyst, December 1997, Vol. 122 1495 Long-term Monitoring of Polycyclic Aromatic Hydrocarbons in Mussels ( Mytilus edulis) Following the BraerOil Spill† Lynda Webstera, Lindsay Angusb, Graham Toppinga, Eric J.Dalgarnoa and Colin F. Moffat*a a FRS Marine Laboratory Aberdeen, P.O. Box 101, Victoria Road, Aberdeen, UK AB11 9DB b North Atlantic Fisheries College, Port Arthur, Scalloway, Shetland, UK ZE1 0UN On January 5, 1993, 84 700 t of Norwegian Gullfaks crude oil was released into the coastal region of south Shetland when the tanker MV Braer grounded at Garths Ness.A Fisheries Exclusion Zone was designated under the Food and Environment Protection Act 1985 (FEPA), prohibiting the taking or harvesting of fish or shellfish within the Zone so as to prevent contaminated products reaching the marketplace. The criteria set for lifting of the Order were that the particular species must be free from any petrogenic taint and the concentration of polycyclic aromatic hydrocarbons (PAHs) must be within the range for reference samples. Between April 1993 and February 1995 the Order was progressively lifted for wild fish, salmon, crustacea, excluding Nephrops norvegicus (Norway lobster), and molluscs, with the exception of mussels. As part of the monitoring exercise, mussels from a reference site were transplanted in June 1995 to three sites within the Zone, where they were suspended in plastic mesh boxes from rafts to a depth of 5 m.Samples were collected at regular intervals over the following 12 months and the concentration and composition of PAHs were determined by gas chromatography with mass spectrometric detection.The total measured PAH concentration at the control site increased from 13.7 to 66.1 ng g21 wet mass of tissue between June 1995 and February 1996. This trend was reversed by July 1996 when the PAH concentration was 12.8 ng g21. The mean across the year for the control site was 24.0 ng g21 (SE = 8.9 ng g21, n = 6). A similar seasonal trend in PAH concentration over the year was observed at all sites within the Zone, but the PAH concentration was consistently greater at these sites, reaching a maximum concentration of 316 ng g21 in February 1996.Although no taint was detected in any of the mussels, these results meant that it was not possible to lift the Prohibition Order for mussels. Further monitoring at three sites outwith the Zone and three sites within the Zone is under way together with investigations into the specific source of the PAHs.Keywords: Polycyclic aromatic hydrocarbons; mussels; gas chromatography–mass spectroscopy; oil spill The tanker MV Braer grounded on Garths Ness, Shetland, on January 5, 1993. Over the following 7 d the entire cargo of approximately 85 000 t of Norwegian Gullfaks light crude oil, a naturally biodegraded oil resulting in a more naphthenic and aromatic crude, was released from the ship, together with some bunker fuel oil. A Fisheries Exclusion Zone was designated on January 8, 1993, by Order under the Food and Environment Protection Act 1985 (FEPA).The Order prohibited the harvesting of farmed or wild fish or shellfish within the Zone to prevent contaminated products reaching the marketplace. The Zone was extended 5 miles westward on January 27, 1993 (see Fig. 1).1 The Scottish Office Agriculture, Environment and Fisheries Department (SOAEFD) instituted a programme of † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997.Fig. 1 Map of Shetland showing the FEPA Exclusion Zone together with the location of the reference site at Olna Firth (R1) and the sites within the Zone (Z1–Z3). Z1, Sandsound Voe; Z2, Stromness Voe, which contained a northerly (N) and southerly (S) site; and Z3, Merry Holm/Trondra. The Braer grounded at Garths Ness on the southerly tip of the Shetland Islands close to the Bay of Quendale. Analyst, December 1997, Vol. 122 (1491–1495) 1491marine monitoring with the aims of ensuring that the original decision to institute the FEPA Zone was a sensible one, that its limits had been correctly drawn and to provide support for any decision-making process with regard to the possible extension or future lifting of the Zone.The criteria for revoking the FEPA Order for fish and shellfish were first, that the fish and shellfish from within the Zone should not contain taint that is associated with crude oils and petroleum fractions, and second, that the concentration of aliphatic hydrocarbons and PAHs in fish and shellfish from within the Zone should fall within the background range of values for fish and shellfish outside the Exclusion Zone.2 It was apparent from the first analytil measurements of hydrocarbons in fish and shellfish that the concentration of aliphatic hydrocarbons in samples from within the Zone was generally low and that the distribution of these hydrocarbons in Gullfaks crude was not distinctive.Therefore, in terms of the aims of the programme, attention was focused on the PAHs. Naphthalenes (75.6%), phenanthrenes (12.5%) and dibenzothiophenes (5.4%) were the major PAHs in Gullfaks crude oil with lesser amounts of the fluoranthenes and pyrenes (4.2%), benzophenanthrenes, benzanthracenes and chrysenes (1.5%) and five- and six-ring PAHs (0.7%).2 As a result of the monitoring programme, the Order was lifted for wild fish on April 23, 1993, farmed salmon on December 8, 1993, crustaceans, with the exception of Nephrops norvegicus (Norway lobster), on September 30, 1994, and molluscs, with the exception of mussels, on February 9, 1995.The blue mussel (Mytilus edulis), a circumboreal species,3 is farmed from rafts in the sea lochs around the Shetland Islands. Some of these mussel farms were within the Exclusion Zone. Mussels from the western area of the Exclusion Zone were sampled during 1993 and 1994 and the PAH concentration, defined as the combined concentration of the two- to six-ring parent and branched PAHs, was determined in the edible tissue.A total measured PAH concentration of 1450 ng g21 wet mass of tissue was determined in a sample taken from Stromness Voe in March 1993, this being the greatest concentration found in mussels.2 The mean PAH concentration in the samples collected from various locations within the Zone during October 1994 was 221 ng g21 (SE = 60 ng g21, n = 5).This contrasts with a mean for reference samples, collected from a mussel farm in Olna Firth (March 1993) and from Aith Voe (February, June and October 1994), of 56.8 ng g21 (SE = 17.2 ng g21, n = 4).2 Mussels are known to accumulate trace contaminants, such as heavy metals and hydrocarbons, present in the water column.4 As such, it was concluded that mussels could be utilized as an organism for monitoring long-term hydrocarbon pollution in marine waters.5 For these reasons, the decision was taken to transplant animals from a site well outside the Zone to various sites within the Zone and to monitor the change in PAH concentration with time.Experimental Reagents Methanol, isohexane, dichloromethane and acetone were glassdistilled reagents specifically prepared for hydrocarbon analysis by Rathburn Chemicals (Walkerburn, UK). HPLC-grade water was also supplied by Rathburn Chemicals. Individual batches of all solvents were checked for contaminants as described previously.6 Analytical-reagent grade nitric acid was purchased from BDH (Poole, Dorset, UK). Sodium chloride, sodium hydroxide and anhydrous sodium sulfate were analyticalreagent grade reagents from Fisons Scientific Equipment (Loughborough, UK). Deuteriated naphthalene, biphenyl, dibenzothiophene, anthracene, pyrene and benzo[a]pyrene were obtained from C/D/N/Isotopes through K&K-Greeff (Croydon, UK).Preventative Measures for Reducing Casual PAH Contamination PAHs are ubiquitous in the environment and great care must be taken to avoid adventitious contamination of samples.To this end, all glassware was washed and dried in a GW 4000 glassware washer (Camlab, Cambridge, UK). Prior to use, the glassware was rinsed twice with dichloromethane and then twice with isohexane, the latter being allowed to evaporate before proceeding. The columns used for the sodium sulfate filtration were soaked at regular intervals in concentrated nitric acid to clean the frits.The columns were then flushed with copious volumes of water before being washed as described above. The sodium sulfate was cleaned by washing with isohexane in an ultrasonic bath for 10 min. The solvent was decanted to waste and the sodium sulfate placed in an oven at 110 °C overnight. The use of Socorex PTFE-lined pipettes (Camlab) with disposable glass Pasteur pipettes, the minimum presence of any plastics, a strict regime for storage of samples, environmental control of the laboratory and assignment of all equipment to specific areas of analysis are all further precautions taken to avoid such contamination.Test Sites The test sites included a reference site at Olna Firth (Site R1, Fig. 1) and three sites within the Exclusion Zone: Sandsound Voe (Site Z1), Stromness Voe (Site Z2, two farms) and Merry Holm (Site Z3). Three of the sites, R1, Z1 and Z2, were located within discrete voes which were remote from any urban or industrial areas, were associated with only one minor road and would have only small-boat traffic.In contrast, Merry Holm (Z3) was a more open, in-shore site approximately 2.5 km south west of Scalloway harbour. Samples of sediment were collected at each site, from the Fisheries Research Vessel Clupea, using a Day grab. The sediments at Z3 were characterised as medium sand/shell sand whereas those at Sandsound (Z1) were a mixture of fine sand and mud. The reference site and Stromness Voe (Z2) contained muddier sediments. Total organic carbon concentration (mean ± s, n = 6) for the sediments, determined using a Model 2400 CHN Elemental Analyser (Perkin-Elmer, Beaconsfield, Bucks., UK), was 5.080 ± 0.807%, 5.155 ± 1.082% and 6.160 ± 0.678% for Sandsound Voe, Stromness Voe and Olna Firth, respectively.The total organic carbon content of the sediment around Merry Holm was lower at 1.384 ± 0.124%. Mussels Mussels (Mytilus edulis), of uniform age, were transplanted from the reference mussel farm in Olna Firth (Site R1) in mid- June 1995 to the three sites within the Zone (Fig. 1). Samples of approximately 70 mussels were placed in numbered plastic mesh boxes and suspended from rafts to a depth of 5 m. Sampling, which comprised collection of one of the mesh boxes, took place during August 1995, October 1995, February 1996 and June/July 1996. The mussels were thoroughly iced and dispatched to Aberdeen by overnight ferry. On arrival in Aberdeen, approximately 20 mussels were removed for sensory assessment.The tissue was removed from the shell of the bulk of the remaining animals, combined and homogenised. A portion was taken for chemical analysis. All residual material was packaged and stored at 230 °C in case repeat analysis was required. Isolation of Hydrocarbons The method was based on that of Grimmer and B�ohnke.7 To a homogenised sample of mussel (10 g) were added the aliphatic hydrocarbon internal standards heptamethylnonane and squalane (approximately 3.2 mg of each). A mixture of deuteriated 1492 Analyst, December 1997, Vol. 122naphthalene, biphenyl, dibenzothiophene, anthracene, pyrene and benzo[a]pyrene (100 ml; approximately 1 mg ml21 each) was then added. This was mixed with sodium hydroxide (10% m/v) in methanol–water (9 + 1 v/v; 40 ml) and 3–5 pre-washed anti-bumping granules. The mixture was refluxed for 3 h 45 min before the addition of water (10 ml) and then refluxing was continued for a further 15 min.The resulting hot solution was extracted with isohexane (80 ml) following the addition of methanol–water (4 + 1 v/v; 40 ml). A second extraction of the aqueous solution with isohexane (80 ml) was performed. The first organic extract was washed with methanol–water (1 + 1 v/v; 40 ml) and this aqueous solution was then used to wash the second organic extract. The two isohexane extracts were combined and washed with water (3 3 40 ml).The resulting organic solution was dried by passage through a column (11 3 1.5 cm id) containing sodium sulfate (approximately 60 g). The column was washed with isohexane (50 ml) and the combined solvent concentrated to approximately 300 ml by rotary evaporation (water bath, < 30 °C). The concentrate was transferred into a vial and concentrated back to approximately 300 ml under a stream of scrubbed nitrogen. The PAHs were isolated from the aliphatic hydrocarbons by isocratic normal-phase HPLC.An aliquot (150 ml) of the concentrated isohexane solution was injected on to a Genesis SIL 4 mm HPLC column (25 3 0.46 cm id) (Jones Chromatography, Hengoed, UK) and eluted with isohexane at 2 ml min. The aliphatic fraction was collected between 0 and 2.75 min and the aromatic fraction between 2.75 and 20 min. The resulting eluates were separately concentrated under reduced pressure prior to transfer to a chromatographic vial insert (Hewlett- Packard, Stockport, UK), where they were further concentrated to approximately 15–20 ml under a stream of scrubbed nitrogen.The sides of the vial insert were carefully washed down with the concentrate before being capped. A procedural blank was analysed with each batch of samples. Determination of Polycyclic Aromatic Hydrocarbons The concentration and composition of the PAHs were determined by gas chromatography with mass spectrometric detection (GC–MS). Samples (1 ml) were chromatographed on an HP 5890 Series gas chromatograph equipped with an HP 7673A oncolumn injector and fitted with a fused silica capillary column (25 m 3 0.2 mm id) coated with a 0.33 mm film of Ultra 1, a cross-linked methylsilicone gum (Hewlett-Packard). Injections were made at 50 °C and the oven temperature was held constant for 3 min, after which it was increased at 20 °C min21 to 100 °C.This was followed by a slower ramp of 4 °C min21 up to a final temperature of 270 °C.The oven temperature remained constant until the end of the analysis. Helium (10 lb in22) was used as the carrier gas. The gas chromatograph was interfaced with an HP 5970 Series mass selective detector (Hewlett-Packard), which was set for selective ion monitoring (SIM) with a dwell time of 50 ms. A total of 25 ions plus the six internal standard ions were measured over the period of the analysis, as detailed previously. 2 Thus, the analysis incorporated two- to six-ring, parent and branched PAHs.This does not cover all of the many PAH compounds that exist. Thus, all references to PAH concentrations and distributions relate to the measured PAHs, details of which are presented in Table 1. Perfluorotributylamine was used as the mass spectrometric calibrant. Standards for all the parent and branched PAHs cannot be obtained, but the limit of detection, calculated as three times the standard deviation of the mean value from six procedural blanks, was found to be < 0.2 ng g21 for benzo[k]fluoranthene and benzo[a]pyrene and < 0.3 ng g21 for chrysene.Good reproducibility was generally obtained for individual PAHs (Table 1). Further quality control was assured through participation in the PAH programme of QUASIMEME (Quality Assurance of Information for Marine Environmental Monitoring in Europe). Assessment of Taint The mussels were steamed for approximately 4 min or until the shells were completely open. At this point the meat was transferred to a lidded casserole and kept warm on an electric hot-plate for the duration of the tasting session (15–20 min).The mussels were assessed ‘blind’ by each member of a panel Table 1 Duplicate determination of the PAHs (ng g21 wet mass of tissue) in a sample of mussel tissue. The numbers following the name refer to the molecular mass. The sum of the molecular mass groups is also presented. The total PAH concentration was 316.3 and 328.5 ng g21 for samples 1A and 1B respectively Sample Sample PAH 1A 1B PAH 1A 1B Naphthalene 0.9 0.6 C3 202 11.5 11.4 C1 Naphthalenes 1.0 0.9 Sum of 202s 55.0 69.1 C2 Naphthalenes 6.1 5.3 Benzo[c]phenanthrene (228) 0.8 0.9 C3 Naphthalenes 16.4 17.3 Benz[a]anthracene (228) 1.7 4.4 C4 Naphthalenes 0.8 0.5 Chrysene + triphenylene (228) 4.7 1.3 Sum of naphthalenes 25.2 24.6 Benz[a]anthracene (228) nd nd Phenanthrene 5.5 5.6 C1 228 7.2 6.3 Anthracene nd* nd C2 228 3.1 3.4 C1 178 31.8 32.6 Sum of 228s 17.5 16.3 C2 178 63.9 60.4 Benzofluoranthene (252) 8.0 6.7 C3 178 50.3 49.7 Benzo[e]pyrene (252) 3.7 4.0 Sum of 178s 151.5 148.3 Benzo[a]pyrene (252) 1.0 0.9 DBT† 0.5 0.6 Perylene (252) 1.4 1.3 C1 DBT 8.3 8.1 C1 252 3.1 3.1 C2 DBT 22.2 21.7 C2 252 nd 2.0 C3 DBT 16.2 18.3 Sum of 252s 17.2 18.0 Sum of DBTs 47.2 48.7 Indenopyrene (276) 1.2 1.2 Fluoranthene (202) 4.2 4.8 Benzoperylene (276) 1.5 1.8 Pyrene (202) 4.0 4.6 C1 276 nd 0.5 C1 202 17.3 14.8 C2 276 nd nd C2 202 18.0 33.5 Sum of 276s 2.7 3.5 * nd, Not detected.† DBT, dibenzothiophene. Analyst, December 1997, Vol. 122 14930 59 136 247 361 (Jun '95) (Aug '95) (Oct '95) (Feb '96) (Jun '96) Z2(N) Z2(S) Z1 R1 350 300 250 200 150 100 50 0 Time/d [PAH]/ng g–1 of 8–10 staff trained to recognise petroleum-derived taints by odour and taste and scored as described by Whittle et al.1 Results and Discussion It is essential in any environmental monitoring programme to have a benchmark against which the sites of potential contamination can be assessed.Consideration must also be given to the fact that the test matrix is a living, growing animal. The objective of transferring samples from a single reference site to several locations within the Zone was to ensure that there was a commonality for the test matrix. The total measured PAH concentration of the mussels at time of transfer was 13.7 ng g21. All the individual parent PAHs and associated branched groups were present at < 2 ng g21 and many were not detected.The largest individual grouping was the three-ring compounds (phenanthrene, molecular mass 178, and the C1–C3 substituted compounds) which comprised 38% of the PAHs. The naphthalenes comprised 24% of the PAHs and the five-ring compounds 18%. The concentration of PAHs at the control site increased progressively over 8 months to a maximum of 66.1 ng g21 in the sample collected in February 1996 (Fig. 2). This was a result of an increase in the concentration of the four- to six-ring compounds, the concentration of the naphthalenes and three-ring PAHs remaining relatively consistent over the year.The greatest PAH group concentration was 19.7 ng g21, determined for the five-ring PAHs isolated from the February sample. The observed increase in PAH concentration over autumn and winter was reversed by July 1996, when the PAH concentration was found to be 12.8 ng g21. This resulted from a decrease in the concentration of the larger ring PAHs.Thus, the proportion of two- and three-ring compounds again dominated the PAH profile. The mean PAH concentration across the year for the control site was 24.0 ng g21 (SE = 8.9 ng g21, n = 6). No taint was detected in any of the mussels from the reference site. After 2 months, an increase in the PAH concentration was detected in the mussels from both sites in Stromness Voe (Site Z2) but not the more northerly site at Sandsound Voe (Site Z1, Fig. 2).This change was a result of an increase in the concentration of all PAH groupings but there was a relative decrease in the proportion of naphthalenes present. The PAH distributions at the two sites in Stromness Voe were very similar. The three-ring compounds still dominated at 38% but the naphthalenes comprised only 18% and 19% of the PAHs at the northerly and southerly sites, respectively. In contrast, the proportion of four-ring compounds had increased from 15% at the time of transplanting to 26% at the northerly site and 25% at the southerly site. All sites within the Zone displayed a progressive increase in PAH concentration between August 1995 and February 1996 with a maximum concentration of 316 ng g21 in Stromness Voe (Fig. 2). Neither this, nor any other sample from within the Zone, was found to be tainted. The dominant group was the C16 four-ring PAHs (pyrene, fluoranthene and C1–C3 substituted compounds), which comprised approximately 30% of the PAHs (Table 2).Although there was no August sample for site Z3, samples were obtained from this location during October 1995 and February 1996 when the PAH concentrations were determined to be 133 and 279 ng g21, respectively. Thus, the trend was maintained at this site. Although the concentration in the October sample from Z3 was similar to that of the northerly site in Stromness Voe, the PAH distribution was distinct, having 35% naphthalenes. In contrast, the PAH distribution of the February sample from Z3 was similar to the other sites within the Zone (Table 2).Indeed, the percentage compositions of the PAHs across the sites within the Zone for the samples collected in February were very similar. The mean percentage (with standard error) for the various PAH groupings at the four sites were: naphthalenes 4% (0.6%); 178, 18% (0.5%); DBTs 2% (0.3%); 202, 30% (0.6%); 228, 18% (0.4%); 252, 24% (0.8%); 276, 5% (0.4%). Severe storms resulted in the loss of some of the mesh boxes from the rafts.This meant that there was only one Zone sample left, at the northerly site in Stromness Voe, in the summer of 1996. As with the equivalent reference sample, a decrease in PAH concentration was observed, relative to the February sample, but the value of 108 ng g21 was greater than that at the reference site. As with the mussels from Olna Firth, the dominant PAH grouping was the three-ring compounds at 40% with the naphthalenes comprising 20% of the measured PAHs.There was an apparent seasonal trend for the PAH concentration in mussels, the concentration increasing over winter and declining in the spring. From autumn until spring lipids may be saved for gametogenesis8 and this increase in lipid content would permit the retention of increased amounts of lipophilic compounds such as PAHs. The maturation of the gametes is Fig. 2 Variation in PAH concentration (ng g21 wet mass of tissue) with time for mussels collected from a reference site (R1) and from sites within the Zone.In all cases there is a progressive increase in PAH concentration with time over the autumn and winter months. The final samples from both the reference site and northerly site in Stromness Voe [Z2(N)], collected in early summer, show a decrease relative to the February sample. The PAH concentrations are consistently greater at sites within the Zone relative to the reference site. Table 2 Percentage distribution of the various PAH groupings for the PAHs isolated from mussels collected at sites within the Exclusion Zone during February 1996 Site within the Exclusion Zone PAH group* Z1 Z2 (North) Z2 (South) Z3 Naphthalenes 4 6 3 4 178 17 19 18 19 DBT 1 1 2 2 202 31 29 30 28 228 19 17 18 18 252 25 23 22 25 276 4 5 6 5 * Naphthalenes, naphthalene and C1–C4 branched compounds; 178, phenanthrene/anthracene and C1–C3 branched compounds; DBT, dibenzothiophene and C1–C3 branched compounds; 202, C16 four-ring PAHs and C1–C3 branched compounds; 228, C18 four-ring PAHs and C1–C2 branched compounds; 252, C20 five-ring PAHs and C1–C2 branched compounds; 276, C22 six-ring PAHs and C1–C2 branched compounds. 1494 Analyst, December 1997, Vol. 122under several exogenous (e.g., water temperature, lunar periodicity, depth, food abundance and availability, light intensity) and endogenous (e.g., genetic, hormonal) controls. Of all the external controls, temperature is generally regarded as the most important and the act of spawning is initiated when the water temperature generally exceeds 10–12 °C for Mytilus edulis,9 although in Shetland the relevant temperature range is regarded as 8–12 °C.The principal spawning time is, thus, late spring/ early summer, but it is obviously affected by the prevailing weather conditions in any year. During 1996, some mussels in Shetland were observed to spawn in early June and this coincides with the decrease in PAH concentration. The seasonal trend can, therefore, be related to physical and chemical changes in the animals.In order for bioaccumulation of PAHs to occur, these compounds must be bioavailable. PAHs may be present dissolved in the water or in the sediment where they are incorporated in the biota, associated with particulate organic and inorganic matter, associated with dissolved organic matter and truly dissolved in the interstitial water. Filter-feeding molluscs are exposed to both water and suspended particulate material (SPM), which would include resuspended sediments.The relative importance of dissolved and particulate PAHs as a source for PAH accumulation in mussels is uncertain. Pruell et al.10 concluded that the dissolved phase was the direct source of contaminants accumulated by mussels. In contrast, Naes et al.11 suggested that particulate PAH is the main source for mussels. This being so, winter storms would result in a degree of resuspension of the sediments, which, if they contained PAHs, would mean an increase in the exposure of the animals to these organic xenobiotics over the winter. The higher concentrations observed for sites within the Zone could result from either a greater exposure to SPM or a greater concentration of PAHs within the SPM. The primary objective of the monitoring programme was to provide data that enabled pertinent decisions to be made with regard to a change in status of the Exclusion Zone. It was evident that the PAH concentration for mussels within the Zone was consistently greater than for the reference site, hence it was not possible to lift the Zone for mussels. Further monitoring is necessary and a second, 3 year, mussel experiment was initiated in September 1996. The seasonal variation, evident for the PAHs, illustrates the clear need for pertinent reference sites which must be sampled at the same time as the test sites and to this end the new experiment incorporates a further two reference sites, at Mangaster Voe and Vaila (Fig. 1). Studies into the source of the PAHs have been initiated and additional investigations into the association of fat content with PAH concentration are under way. The authors acknowledge Mrs. Nicky Shepherd for her assistance with the preparation of the mussels for determination of taint and all colleagues on the sensory panels. References 1 Whittle, K. J., Anderson, D. A., Mackie, P. R., Moffat, C. F., Shepherd, N. J., and McVicar, A. H., in The Impact of an Oil Spill in Turbulent Waters: The Braer, ed. Davies, J. M., and Topping, G., The Stationery Office, Edinburgh, 1997, ch. 11. 2 Topping, G., Davies, J. M., Mackie, P. R., and Moffat, C. F., in The Impact of an Oil Spill in Turbulent Waters: The Braer, ed. Davies, J. M., and Topping, G., The Stationery Office, Edinburgh, 1997, ch. 10. 3 Jamieson, G. S., World Aquacult., 1989, 20, 94. 4 Murray, A. P., Richardson, B. J., and Gibbs, C. F., Mar. Pollut. Bull., 1991, 22, 595. 5 Fossato, V. U., and Canzonier, W. J., Mar. Biol., 1976, 36, 243. 6 McGill, A. S., Moffat, C. F., Mackie, P. R., and Cruickshank, P., J. Sci. Food Agric., 1993, 61, 357. 7 Grimmer, G., and B�ohnke, H., J. Assoc. Offic. Anal. Chem., 1975, 58, 725. 8 Voogt, P. A., in The Mollusca, Metabolic Biochemistry and Molecular Biomechanics, ed. Hochachka, P. W., Academic Press, New York, 1983, ch. 7. 9 Mackie, G. L., in The Mollusca, Reproduction, ed. Tompa, A. S., Verdonk, N. H., and van den Biggelaar, J. A. M., Academic Press, New York, 1984, ch. 5. 10 Pruell, R. J., Lake, J. L., Davis, W. R., and Quinn, J. G., Mar. Biol., 1986, 91, 497. 11 Naes, K., Bakke, T., and Konieczny, R., Mar. Freshwater Res., 1995, 46, 275. Paper 7/05977I Received August 14, 1997 Accepted October 10, 1997 Analyst, December 1997, Vol. 122 14

ISSN:0003-2654    

DOI:10.1039/a705977i

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Use of Chemical Ionization in Multianalysis Gas and Liquid Chromatography Combined With a Single Mass Spectrometer for the Ultra-trace Level Determination of Microcontaminants in Aqueous Samples† Arjan J. H. Louter*‡ , Ariadne C. Hogenboom, Jaroslav Slobodn�ýk, Ren�e J. J. Vreuls and Udo A. Th. Brinkman Department of Analytical Chemistry, Free University, De Boelelaan 1083, Amsterdam, 1081 HV, The Netherlands. E-mail: arjan.louter@unilever.com An automated system that comprises a single mass spectrometer (MS) in combination with gas (GC) and liquid chromatography [LC; particle beam (PB) interface] has been used for the trace-level detection and identification of contaminants in water samples. The analytes were enriched by on-line solid-phase extraction (SPE).In the present study, the potential of negative chemical ionization (NCI) detection in this so-called multianalysis system was explored. Attention was devoted to the enhancement of selectivity and sensitivity as well as to the additional spectral information obtained for the identification of unknown compounds.Nine chlorinated pesticides representing three major groups, i.e., triazines, anilides and organophosphorus pesticides, were used as test compounds. Among the three reagent gases used for NCI, isobutane, methane and ammonia, methane gave the best results. For six of the nine pesticides, a 3- to 30-fold increase in sensitivity was observed in the NCI mode as compared with the electron impact (EI) mode.As expected, the NCI mass spectra showed little fragmentation. Electron capture appears to be the dominant ionization mechanism. In order to study the potential of the total on-line SPE–LC/GC–MS set-up, the pesticides were spiked to tap and surface water samples. The detection limits obtained in the NCI (full-scan) mode ranged from 0.1–3 ng l21 for GC–MS and 50 to 200 ng l21 for LC–PB–MS for 100-ml tap water samples. The potential of NCI–MS was demonstrated by the identification of several unknown microcontaminants in a river water sample. Keywords: Gas chromatography–mass spectrometry; liquid chromatography–mass spectrometry; chemical ionization; solid-phase extraction; water samples; pesticides; microcontaminants Mass spectrometry (MS) is the preferred technique for the tracelevel detection of microcontaminants in environmental samples because it can be optimized for target compounds (identification and quantification) as well as unknowns (provisional identification).The complexity of most environmental samples requires a separation step prior to MS detection. Capillary gas chromatography –mass spectrometry (GC–MS) combines a high separation power and selective as well as sensitive detection. For the on-line coupling of column liquid chromatography and mass spectrometry (LC–MS), several interfaces have become available in the past decade. These include thermospray (TSP), particle beam (PB) and atmospheric pressure ionisation (API) interfaces, which offer the use of various ionization modes.1 Although API interfaces are considered to have a promising future because of their excellent sensitivity and the possibility of regulated fragmentation in the ion source,2 the PB interface will certainly remain attractive as the only interface which can generate electron impact (EI) and solvent-independent chemical ionization (CI) spectra.In GC–MS and LC–PB–MS detection, ionization is generally carried out in the positive EI mode because EI–MS provides very reproducible mass spectra and a wide variety of mass spectral libraries containing up to 130000 entries are available for fast identification.3,4 CI with positive (PCI) or negative (NCI) ion detection can provide additional information for provisional identification of unknowns when no relevant information is contained in such MS libraries.5 NCI is especially recognized for the improved selectivity and sensitivity that can be obtained in the detection of chlorinated pesticides.6,7 Usually, only a few ions of high abundance are observed in the NCI mass spectra; this fact enhances analyte detectability if the selected ion monitoring (SIM) mode is applied.8–10 Since the analytes of interest are generally present at very low concentration levels, a concentration step prior to actual analysis is required.Using on-line solid-phase extraction (SPE) for introducing large sample volumes into GC–MS or LC–MS has several advantages.On-line SPE has a good automation potential, requires the use of small amounts of organic solvents and offers a high sensitivity due to the quantitative transfer of the analytes to the chromatograph. Several automated and online SPE–GC–MS and SPE–LC–MS systems have been developed in the recent past.11–16 SPE of microcontaminants from aqueous samples can be carried out on a pre-column with typical dimensions of 10 mm 3 2 mm id, which is packed with a hydrophobic phase such as a polystyrene–divinylbenzene copolymer or C18 bonded silica.In recent studies, SPE–LC has been combined with TSP–MS for target analysis11 and PB–MS for the detection of unknown compounds.12 In SPE–GC, desorption is carried out with a volume of 50–100 ml of ethyl acetate, which can be directly introduced into a gas chromatograph after a proper drying step13 using retention gap techniques involving partially concurrent solvent evaporation.14 The set-up of an automated, software-controlled SPE–GC–MS system has also been reported previously.15,16 In a previous study, a socalled multianalysis system has been described.17 This system, which uses a single MS detector combined with on-line SPE– GC and SPE–LC, was found to be a powerful tool for the detection of microcontaminants in aqueous samples.The present study describes the potential of combined SPE– LC–PB–MS and SPE–GC–MS using NCI detection for the target analysis of chlorinated pesticides and for the identification of unknown compounds in real-life samples.† Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997. ‡ Present address: Unilever Research Laboratorium, P.O. Box 114, 3130 AC Vlaardingen, The Netherlands. Analyst, December 1997, Vol. 122 (1497–1503) 1497Experimental Chemicals and Samples Ethyl acetate, hexane, methanol and HPLC-grade water were from J.T. Baker (Deventer, The Netherlands). The ethyl acetate, hexane and methanol were glass-distilled prior to use. The chlorinated pesticides, alachlor, atrazine, bromophos, chlorpyriphos, coumaphos, cyanazine, fenchlorphos, metolachlor and tetrachlorvinphos (see Table 1 for details), were all of 95%, or better, purity and were purchased from Riedel de Haen (Seelze, Germany). Stock solutions of 200 mg l21 were prepared in ethyl acetate and diluted to a final concentration of 10 mg l21 of each compound in a mixture.This was used for direct on-column injections and for the preparation of spiked tap and surface water samples. River water samples were collected from the Nitra River in Slovakia, transported to The Netherlands in a portable refrigerator and stored at 4 °C. The river water samples were filtered through a 0.45 mm acetylcellulose filter (Schleicher and Schuell, Dassel, Germany) prior to preconcentration. Prior to analysis, surface water samples were spiked with metoxuron and propazine (1 mg l21) which served as internal standards. Trace Enrichment Trace enrichment was performed on a Prospekt sample preparation system of Spark Holland (Emmen, The Netherlands).This system consists of three six-port HPLC valves, a cartridge exchange system and a solvent delivery system (SDU). The SDU is provided with a six-port solvent selection valve, a pulse dampener and a single-piston analytical LC pump.Water samples were preconcentrated on 10 3 2.0 mm id stainless-steel pre-columns containing 15–25 mm PLRP-S copolymer (Polymer Laboratories, Church Stretton, Shropshire, UK). All experiments were performed with the setshown in Fig. 1. Prior to analyte enrichment, the pre-column was conditioned with 4 ml of methanol and 2 ml of HPLC-grade water. Next, 10 or 100 ml of sample were preconcentrated on the pre-column in order to trap the analytes of interest. Gas chromatography For the combination of the SPE module with GC–MS, a 50 cm 3 50 mm id fused silica capillary was used as a connection between the Prospekt module and the GC on-column injector.After trace enrichment, 1 ml of HPLC-grade water was pumped through the pre-column in order to prevent the introduction of salts into the GC system. Next, the pre-column was dried for 30 min with a stream of nitrogen (40 ml min21). The capillaries were prepressurized with ethyl acetate and, after switching valve V1 to the elute position, the analytes were eluted from the pre-column into the retention gap (details given below).A microMetric metering pump from Milton Roy (Riviera Beach, FL, USA) was used to deliver the desorption solvent, ethyl acetate. Liquid chromatography For the combination of the SPE module with LC–PB–MS the trapped analytes were eluted from the precolumn by the LC gradient onto the analytical column. Gas Chromatography A Hewlett-Packard (Palo Alto, CA, USA) Model 5890 Series II gas chromatograph equipped with a pressure-programmable oncolumn injector was used for GC analysis.The split/splitless injector was replaced by a home-made solvent vapour exit (SVE) which was controlled by a 24 V triggering signal of the GC. The injector was connected to a 5 m 30.32 mm id retention gap, which is essentially an uncoated fused silica capillary deactivated with diphenyltetramethyldisilazane (DPTMDS) (BGB Analytik, Z�urich, Switzerland), a 2 m 3 0.20 mm id, retaining pre-column and a 10 m 3 0.20 mm id capillary GC column, both containing HP-1 [100% dimethyl siloxane (Hewlett-Packard)] with a film thickness of 0.33 mm.Helium (99.99999% purity, Hoekloos, Schiedam, The Netherlands) was the carrier gas at an inlet pressure of 6.0 psi. Connections were made with conventional glass press-frits and a glass pressfrit Y-piece (BGB Analytik). The analytes trapped on the pre-column were eluted with 100 ml of ethyl acetate (75 ml min21) into the retention gap; the excess of solvent was removed through the SVE (see ref. 17 for a detailed description). The observed evaporation rate was 58 ml min21. After closing of the SVE (total open time, 2.2 min), the analytes were transferred to the analytical column for separation and MS detection. The initial temperature was 85 °C (5 min hold time), which was increased to 270 °C at a rate of 10 °C min21. The initial carrier gas pressure was 6.0 psi, which was programmed to 46.0 psi (at 99 psi min21) during the SVE open time in order to Table 1 Relevant information on pesticides used in the present study Relative Molecular Electronegative Analyte Type* mass formula groups (number) Atrazine TRI 215.1 C8H14ClN5 –Cl Alachlor ANA 269.8 C14H12ClNO2 –Cl, –NO2 Fenchlorphos OPP 321.6 C8H8Cl3O3PS –Cl (3), –P ester Cyanazine TRI 240.1 C9H13ClN6 –Cl, C·N Metolachlor ANA 283.8 C15H22ClNO2 –Cl, –NO2 Chlorpyriphos OPP 350.6 C9H11Cl3NO3PS –Cl (3), –P ester Bromophois OPP 366.0 C8H8BrCl2O3PS –Cl (2), –Br, –P ester Tetrachlorvinphos OPP 365.9 C10H9Cl4O4P –Cl (4), –P ester Coumaphos OPP 362.0 C14H16ClO5PS –Cl, P ester * TRI = triazine, ANA = anilide, OPP = organophosphorus pesticide. Fig. 1 Set-up for multianalysis system which combines SPE–GC–MS and SPE–LC–DAD–PB–MS in one integrated set-up. HP 1090, liquid chromatograph; HP 5890 II, gas chromatograph; MS, mass spectrometer; PB, particle beam interface; DAD, UV diode array detector; PROSPEKT, automated valve-switching, solvent selection and cartridge exchange unit; SDU, solvent delivery unit; M, MUST, automated six-port switching valve; S, syringe pump; PR, SPE precolumn; AC, LC analytical column; C, GC analytical column; RP, retaining precolumn; RG, retention gap; SVE, solvent vapour exit; INJ, on-column injector; N2, nitrogen; W, waste; V1–V5 = six-port switching valves. 1498 Analyst, December 1997, Vol. 122enhance the solvent evaporation rate and reduce the transfer time.Next, the pressure was reduced to 6 psi (1.95 min hold time) and programmed to 12.4 psi (at 0.3 psi min21). Liquid Chromatography Analyses were performed on a HP 1090 liquid chromatograph (Hewlett-Packard) equipped with an automatic six-port switching valve (Rheodyne, Berkeley, CA, USA). A 250 3 4.6 mm id stainless-steel column packed with 5 mm C18-bonded silica of 100 Å pore size (Supelco LC-18-DB; Supelchem, Leusden, The Netherlands) was used for separation.A HP 1050 UV diodearray detector (Hewlett-Packard) was operated at 210 nm wavelength with a 10 nm bandwidth. The eluent consisted of a mixture of acetonitrile and HPLCgrade water (Riedel de Ha�en). The analytes were eluted with a linear gradient (flow 0.4 ml min21) during which the acetonitrile content was increased from 10 to 95% (0–45 min) with a hold time of 10 min (45–55 min). Mass Spectrometric Detection A HP 5889A MS Engine (Hewlett-Packard), equipped with a high energy dynode, was used as detector.The MS was operated in the full-scan mode, using the scan range 45–400 amu for positive EI and 65–400 amu for both NCI and PCI detection; the scan rate was 1 scan s21. The operating temperature of the MS ion source was 250 °C and that of the quadrupole, 100 °C. The PB nebulizing gas, high-purity helium (Hoekloos) was kept at 248 kPa (36 psi, about 2.5 ml min21). The temperature of the PB desolvation chamber was 70 °C. Methane, isobutane and ammonia (all three from Hoekloos; purity > 99.999 95%) were used for CI experiments; their pressure in the ion source was maintained at 0.9, 0.7 and 0.9 Torr, respectively.Multianalysis System All parts of the set-up were controlled by the SPE/WIN software (version A.03.14B) of the DOS-based HP Vectra 486/66XM computer (Hewlett-Packard, Waldbronn, Germany). For a detailed description of the operation of the Multianalysis system (Fig. 1), one should consult ref. 21. Results and Discussion With the ion source of the mass spectrometer included in our Multianalysis system, next to the EI mode, both the NCI and PCI mode of operation can be studied for LC–PB–MS as well as GC–MS.This provides the unique situation that spectra obtained after two completely different types of chromatographic separation can be directly compared. As regards CI operation, the influence of the reagent gases on MS response and fragmentation was studied. Methane and ammonia were used for PCI operation; for NCI operation, these two gases and isobutane were studied.Nine chlorinated pesticides, which could be expected to give good responses in the NCI mode, were used as test analytes. Comparison of GC–MS and LC–PB–MS Spectra EI–MS The mass spectra generated in the Multianalysis system using the same detector for GC–MS and LC–PB–MS in the EI mode are presented in Table 2. Eight out of the nine test analytes exhibited the same base peak for both GC–MS and LC–PB–MS.The exception was chlorpyriphos which exhibited m/z 97 {[(HO)2PS]+} as the base peak in the GC–MS spectrum and m/z 314 ([M–Cl]+) in the LC–PB–MS spectrum. The molecular ion was the base peak in the EI spectrum for one analyte (coumaphos) only. This was not unexpected in view of the extensive fragmentation generally observed in the EI mode. In most instances the GC–MS and LC–PB–MS spectra were very similar, but some differences in relative abundances and order of the secondary mass fragments was observed.CI–MS. In the NCI, and also in the PCI mode, the GC–MS and LC–PB– MS spectra were almost identical. They always exhibited the same base peak, and the differences in relative abundance and order of the secondary mass fragments were negligible. No doubt this is partly due to the fact that CI spectra showtle fragmentation. Relevant data on NCI and PCI data are presented in Table 3 and are discussed in some detail below. NCI–MS The responses and fragmentation patterns of all analytes were studied by GC–MS with methane, ammonia and isobutane.The mass spectra obtained with the three gases were highly similar. Relevant data on the NCI spectra are presented in Table 3. Despite their electronegative -Cl and -NO2 groups, alachlor and metolachlor were not detected even when 10 ng were injected, while the response of atrazine was just above the detection limit. The rather poor detectability of atrazine in NCI (methane)–MS has also been observed by other workers.6,18,19 To our knowledge, there are no NCI–MS data on alachlor and metolachlor in the literature.The analyte responses invariably were highest with methane as reagent gas; the only exception was atrazine, which showed a 3-fold higher response with ammonia. The results obtained with isobutane were rather poor. To quote an example, 1-mg amounts had to be injected in LC– Table 2 Major ions and relative abundances obtained by GC–MS and LC–PB–MS under EI conditions Ions: m/z (per cent.relative abundance) GC–MS LC–PB–MS Analyte Atrazine 200 (100) 58 (75) 215 (60) 68 (40) 173 (27) 200 (100) 215 (53) 138 (35) 173 (34) 202 (25) Alachlor 45 (100) 160 (51) 188 (42) 146 (20) 122 (14) —* Fenchlorphos 285 (100) 287 (67) 125 (45) 79 (25) 109 (23) 285 (100) 287 (61) 86 (35) 125 (33) Cyanazine 212 (100) 68 (85) 225 (46) 198 (41) 96 (40) 212 (100) 68 (83) 225 (47) 96 (42) 172 (40) Metolachlor 162 (100) 238 (21) 163 (11) 146 (11) 150 (9) 162 (100) 238 (14) 163 (12) 146 (10) 150 (8) Chlorpyriphyos 97 (100) 199 (75) 197 (68) 314 (46) 316 (34) 314 (100) 197 (93) 97 (75) 316 (62) 131 (53) Bromophos 331 (100) 329 (72) 125 (79) 79 (45) 109 (36) 331 (100) 329 (82) 125 (65) 75 (64) 335 (32) Tetrachlorvinphos 109 (100) 329 (89) 331 (83) 333 (30) 207 (21) 109 (100) 329 (81) 331 (74) 333 (30) 79 (19) Coumaphos 362 (100) 226 (49) 109 (41) 364 (41) 97 (36) 362 (100) 226 (77) 109 (75) 97 (67) 210 (48) * Coelutes with metolachlor.Analyst, December 1997, Vol. 122 1499PB–NCI–MS to detect a mere four out of the nine test analytes, viz. tetrachlorvinphos, coumaphos, fenchlorphos and bromophos. For obvious reasons, methane was selected as reagent gas for further work. The NCI spectra of the organophosphorus pesticides exhibited typical group-specific fragments, viz., m/z 125 for the dimethylphosphates and tetrachlorvinphos, m/z 141 for the dimethylphosphorothionates, bromophos (low abundance) and fenchlorphos, and m/z 169 for the diethylphosphorothionates, coumaphos and chlorpyrifos, which were also observed by other workers.20 As for the triazines, the NCI spectrum of atrazine exhibited [M–H]2 as the base peak whereas cyanazine showed [M]2 due to the presence of a C·N group under electron-capture conditions.21 PCI–MS Methane and ammonia were used as reagent gases for LC–PB– PCI–MS, and methane for GC–PCI–MS.As was to be expected, the majority of the PCI mass spectra (Table 3) exhibited [M + H]+ as the base peak.As regards the two exceptions, tetrachlorvinphos and alachlor, the former analyte exhibited m/z 127 as the base peak, which can be assigned to [C2PH8O4]+ (counterpart of the m/z 125 fragment due to [C2PH6O4]2 observed in NCI–MS). Alachlor exhibited m/z 238, due to [M2CH3O]+, as its base peak. For six out of the nine compounds the MS responses in the PCI mode were substantially lower than in the NCI mode. However, the detectability of atrazine, metolachlor and alachlor was better in the PCI (methane) mode.When ammonia was used as a reagent gas for LC–PB–PCI– MS, atrazine gave a 2–3-fold better response than with methane. Prominent [M + C2H5]+ and [M + C3H5]+ adduct ions (with methane, m/z 244 and 256, respectively) were detected for atrazine, which have also been observed by other workers.14,22 For the other eight analytes [M + C2H5]+ adduct ions were also detected, but their intensity was typically about 10% of that of the [M + H]+ peak and, invariably, the responses were lower than with atrazine.As an example, Fig. 2 shows the EI, NCI and PCI mass spectra of coumaphos obtained by LC–PB–MS. In the EI mode, the base peak can be assigned to the molecular ion (m/z 362), while in the NCI spectrum the base peak observed at m/z 225 is due to the thiophenolate [M2C4H12O3P]2 ion, and m/z 227 to its 37Cl isotope. The peaks observed at m/z 191 and 169 are due to [M2C4H12O3PS]2 and [C4H10O3PS]2, respectively. The latter is the group-specific fragment already mentioned above.The base peak in the PCI mode (m/z 363) can be assigned to [M + H]+, while m/z 329 is due to [M - Cl + H]+. The peak observed at m/z 391 is due to a [M + C2H5]+ adduct. Analytical Data of Chlorinated Pesticides by SPE–LC–PB–MS/GC–MS (EI and NCI Detection) The analytical performance of the GC and LC–PB parts of the multianalysis system were tested in the EI and NCI mode, using Table 3 Relative molecular masses (Mr), major ions and relative abundances obtained under NCI and PCI conditions (GC; 10 ng injection) Ions: m/z (per cent.relative abundance) NCI PCI Analyte Mr Atrazine 215 214 (100) 179 (30) 216 (100) 180 (95) 218 (32) 244 (17) 256 (10) Alachlor 269 238 (100) 240 (34) 270 (14) Fenchlorphos 320 213 (100) 211 (87) 215 (36) 141 (20) 270 (10) 323 (100) 321 (97) 125 (86) 111 (39) 285 (33) Cyanazine 240 240 (100) 242 (34) 212 (18) 120 (17) 204 (16) 214 (100) 241 (55) 205 (41) 178 (38) Metolachlor 283 284 (100) 252 (78) 286 (33) 254 (27) 238 (26) Chlorpyriphos 349 313 (100) 315 (73) 214 (64) 212 (61) 169 (60) 350 (100) 352 (96) 354 (34) 153 (28) 322 (23) Bromophos 364 257 (100) 270 (65) 255 (59) 259 (52) 272 (47) 367 (100) 365 (60) 369 (30) 125 (30) 331 (22) Tetrachlorvinphos 364 125 (100) 224 (14) 222 (13) 127 (100) 367 (18) 365 (16) Coumaphos 362 225 (100) 227 (49) 191 (26) 169 (23) 362 (21) 363 (100) 329 (42) 365 (33) 391 (11) 139 (10) Fig. 2 LB–PB–MS spectra of coumaphos obtained in the EI mode, methane–NCI and PCI mode. 1500 Analyst, December 1997, Vol. 122on-line SPE to improve analyte detectability. Table 4 summarizes the detection limits (S/N = 3) that were obtained for 100 ml tap water samples. For six out of the nine test compounds, detectability in the NCI mode is distinctly, i.e., 10–30-fold, better than in the EI mode. However, there are notable exceptions as well, although these are not really unexpected after the earlier discussion of the NCI–MS data.Another striking observation is, of course, that for the same mass spectrometer, the detectability in LC–PB–MS is two–three orders of magnitude less good than in GC–MS, i.e. in the low to sub ng l21 versus the mg l21 range for the (identical) samples analysed. Fig. 3 shows SPE–GC–MS chromatograms of the analysis of 10 ml spiked (1 mg l21) tap water run in the NCI and EI mode. Analyte detectability can be improved by one order of magnitude with time-scheduled selected ion monitoring (SIM), as is demonstrated in Fig. 4, where extracted ion and SIM chromatograms are compared.Obviously, for the six test compounds showing high analyte response in NCI–MS detection, both LC- and GC-based separation can conveniently be used for ultra-trace level confirmation purposes at levels required by legislative bodies (typically, 0.1 mg l21 for tap water). Actually, it can be inferred from our data that with timescheduled operation, 1 ml of water sample will be sufficient to obtain the required sensitivity.The linearity was tested in both the GC and LC operation of the system. For 10-ml samples values of the squares of the regression coefficients (R2) for SPE–GC–NCI–MS ranged from 0.9906 to 0.9929 (4 data points; range 0.01–1 mg l21); in the EI mode, the values ranged from 0.9981 to 0.9999 (4 data points; range 0.05–5 mg l21). As was to be expected on the basis of published experience,23 for SPE–LC–PB–NCI–MS curved rather than linear calibration plots were obtained.Still, as was also observed before, for the rather limited concentration range of interest in studies of environmental samples (typically, 0.1–10 mg l21), a three- or four-point calibration will enable reliable quantification.21 Application: Identification of Unknown Microcontaminants As an application the combined EI and NCI operation was used for the analysis of a river water sample by SPE–GC–MS and SPE–LC–PB–MS.In the SPE–GC–MS chromatograms of the surface water sample (Fig. 5), the EI and NCI traces were found to be distinctly different, with several compounds (e.g., No. 1 and 2) responding better in the latter mode. The EI and NCI spectra of the unknown compound eluting at 12.4 min (peak No. 1) indicate it to be (2-methylthio)-benzothiazole, presumably a transformation product of benzothiazole, which was detected in the same sample with SPE–LC–DAD–UV.The base peak in the NCI mass spectrum is most likely to be the thiolate anion (m/z 166; [C7NS2H4]2), which is known to be very stable. The relative abundance of the peak at m/z 168 was 8.7%, which is supporting evidence for the presence of two sulfur atoms. The base peak in the EI mode was m/z 181, which is the molecular ion; here the typical sulfur isotope peak (m/z 183; relative abundance, 8.9%) was also observed. The spectrum gave a good match (94 out of 100) with the Wiley library spectrum.The unknown compound, designated peak No. 2, was more difficult to identify. The EI spectrum [Fig. 6(a)] is obviously a mixture of two or three coeluting compounds. The corresponding NCI spectrum [Fig. 6(b)] shows much better selectivity or, in other words, a clearer mass spectrum. Again, the base peak in the NCI mass spectrum was m/z 166, with a sulfur isotope peak at m/z 168. On the basis of the m/z 166 and m/z 168 peaks, the peak at m/z 182 can provisionally be assigned to [C7NS2H4O]2 (see Fig. 6, below), also because of the peak ratio typical for two sulfur atoms (relative abundance of m/z 184, 8.7% of m/z 182). The EI spectrum contained characteristic ions at m/z 108, 134, 148, 166 and 197, which support the presence of an oxidation product of (2-methylthio)-benzothiazole with a relative molecular mass of 197. The same sample was also analysed with SPE–LC–PB–MS. Despite the known low sensitivity of PB–MS for benzothiazole- Table 4 Detection limits of on-line SPE–GC–MS (ng l21) and SPE–LC– PB–MS (mg l21) for 100 ml tap water samples SPE–GC–MS SPE–LC–PB–MS Analyte EI NCI EI NCI Atrazine 1 100 0.5 > 10 Alachlor 7 — 7 — Fenchlorphos 5 0.2 5 0.2 Cyanazine 10 3 0.5 0.15 Metolachlor 0.5 — 0.5 — Chlorpyriphos 30 1 5 0.2 Bromophoa 5 0.3 3 0.2 Tetrachlorvinphos 1 0.1 0.5 0.15 Coumaphos 30 2 0.7 0.05 Fig. 3 Full-scan SPE–GC–MS chromatograms (EI and methane–NCI) of 10 ml tap water spiked at the 1 mg l21 level.Peak designation; 1, atrazine; 2, alachlor; 3, fenchlorphos; 4, cyanazine; 5, metolachlor; 6, chlorpyriphos; 7, bromophos; 8, tetrachlorvinphos; 9, coumaphos. Fig. 4 SPE–GC–NCI–MS chromatograms of 10 ml tap water spiked at the 10 ng l21 level with chlorinated pesticides. Left, extracted ion chromatograms of m/z of base peak of investigated compounds. Right, TS SIM chromatogram. Peak designation; 1, fenchlorphos (m/z 213); 2, chlorpyriphos (m/z 313); 3, bromophos (m/z 257); 4, tetrachlorvinphos m/z 125); 5, coumaphos (m/z 225).Analyst, December 1997, Vol. 122 1501type compounds, the oxidation product of benzothiazole proposed above was detected under NCI conditions, as is evident from the extracted ion chromatogram of m/z 166 in Fig. 7. Here, 200 ml of sample were preconcentrated to obtain sufficient sensitivity. In the EI mode the most abundant LC peak had a mass spectrum with m/z 138 as base peak [Fig. 7, insert (a)].Insert (b) shows the NCI mass spectrum recorded for the same peak, which is dominated by m/z 222. The relative molecular mass of the unknown compound probably is 254, as can be derived from the peaks at m/z 254 and 253, designated as [M]2 and [M-H]2 in the NCI spectrum and as [M]+ in the EI spectrum, respectively, with m/z 222 then representing [MSH]+. In both the EI and NCI modes the peak at m/z 256 had a relative abundance of 20% of the m/z 254 (molecular ion) peak, which indicates the presence of four or five sulfur atoms. Unfortunately, further identification was not possible on the basis of these data.This will probably require the use of tandem MS techniques. Conclusions An automated, so-called multianalysis, system which combines SPE–LC–PB–MS and SPE–GC–MS using only one sample preparation module and one mass spectrometric detector, has been used to study CI–MS detection. Nine pesticides containing different electronegative groups were used as test analytes.As expected, NCI detection provided enhanced sensitivity and selectivity compared with EI detection for most of the analytes. For five chlorinated pesticides, a 10–30-fold increase in sensitivity was observed. However, the presence of a chlorine or nitro-substituent did not always give an increased response; atrazine, alachlor and metolachlor were such exceptions. The LC–PB–MS and GC–MS spectra recorded on the same MS detector were found to be very similar in the EI mode and almost identical in the NCI mode.For six of the nine test analytes, the detection limits in on-line SPE–LC/GC–NCI–MS (full-scan) were 0.1–3 ng l21 (GC) and 50–200 ng l21 (LC–PB) for 100-ml tap water samples. Obviously, for on-line SPE–GC samples sizes of 1–10 ml are more appropriate in view of the directives for detection limits in drinking water. For SPE–LC–PB–MS the NCI mode is an interesting option for sub mg l21 detection and identification of a number of microcontaminants in water.SPE–LC approaches using TSP–MS or API–MS may be more sensitive, but they are not ‘universal’ either, and PB–NCI–MS has the advantage of providing more structural information. These results again Fig. 5 Chromatograms of 10 ml river Nitra water samples obtained by SPE–GC–MS in the full-scan EI and methane–NCI mode. The sample was spiked with 1 mg l21 of the internal standard propazine (P). (a) EI and (b) NCI spectra of the compound eluted at 12.4 min are shown as inserts.Fig. 6 EI and methane–NCI spectra of unknown peak 2 from Fig. 5. 1502 Analyst, December 1997, Vol. 122indicate that more attention should be devoted to improve the efficiency of the particle beam interface. The practicality of NCI detection in the combined SPE–LC– PB–MS and SPE–GC–MS approach was demonstrated by the (partial) identification of unknown microcontaminants in rather polluted sample where EI detection alone was not sufficient. We thank the European Union (EV5V–CT92–0105) and the River Basin Program (Amsterdam) for their financial support.Dr. Ben van Baar is acknowledged for valuable discussions. References 1 Niessen, W. M. A., and van der Greef, J., Liquid Chromatography– Mass Spectrometry, Principles and Applications, Marcel Dekker, New York, USA, 1992. 2 Voyksner, R. D., and Peck, T. D., Rapid Commun. Mass Spectrom., 5, (1991) 263. 3 Levsen, K., Org. Mass. Spectrom., 1988, 23 406. 4 Schmidt, E.R., Chromatographia, 1990, 30, 573. 5 Mattina, M. J. I., J. Chromatogr., 1991, 542, 385. 6 Ong, V. S., and Hites, R. A., Mass Spectrom. Rev., 1994, 13, 259. 7 Mattina, M. J. I., Trends Anal. Chem. 1993, 12, 328. 8 Busch, K. L., Bursey, M. M., Hass, J. R., and Sovocool, G. W., Appl. Spectrosc., 1978, 32, 388. 9 Stan, H.-J., and Kellner, G., Biomed. Environ. Mass Spectrom., 1982, 9, 483. 10 Stan, H.-J., and Kellner, G., Biomed. Environ. Mass Spectrom., 1989, 18, 645. 11 Bagheri, H., Brouwer, E.R., Ghijsen, R. T., and Brinkman, U. A. Th., J. Chromatogr., 1993, 647, 121. 12 Bagheri, H., Slobodnik, J., Marc�e Recasens, R. M., Ghijsen, R. T., and Brinkman, U. A. Th., Chromatographia, 1993, 37, 159. 13 Vreuls, J. J., Ghijsen, R. T., de Jong, G. J., and Brinkman, U. A. Th., J. Chromatogr., 1992, 625, 237. 14 Grob, K., Fr�olich, D., Schilling, B., Neukom, H.-P., and N�ageli, P., J. Chromatogr., 1984, 295, 55. 15 Louter, A. J. H., Brinkman, U. A. Th., and Ghijsen, R.T., J. Microcol. Sep., 1993, 5, 303. 16 Louter, A. J. H., van Beekvelt, C. A., Cid Montanes, J. Slobodn�ýk, J., Vreuls, J. J., and Brinkman, U. A. Th., J. Chromatogr., 1996, 625, 67. 17 Slobodnik, J., Hogenboom, A. C., Louter, A. J. H., and Brinkman, U. A. Th., J. Chromatogr., 1996, 730, 353. 18 Huang, L. Q., and Mattina, M. J. I., Biomed. Environ. Mass Spectrom., 1989, 18, 828. 19 Bagheri, H., Vreuls, J. J., Ghijsen, R. T., and Brinkman, U. A. Th., Chromatographia, 1992, 34, 5. 20 Durand, G., and Barcel�o, D., Anal. Chim. Acta, 1991, 243, 259. 21 Slobodnik, J., Ph.D. Thesis, Free University, Amsterdam, The Netherlands, 1996. 22 Crespo, C., Marc�e, R., and Borull, F., J. Chromatogr. A, 1994, 670, 135. 23 Bellar, T. A., Behymer, T. D., and Budde, W. L., J. Am. Soc. Mass Spectrom., 1990, 1, 92. Paper 7/05209J Received July 21, 1997 Accepted September 25, 1997 Fig. 7 Chromatograms of 200 ml river Nitra water samples obtained by SPE–LC–PB–MS in the full-scan EI and methane–NCI mode.Samples were spiked with 1 mg l21 of internal standards propazine (P) and metoxuron (M). MS chromatograms represent extracted ions obtained at the m/z values of the base peak of each investigated analyte. The inserts show the (a) EI and (b) methane NCI spectra of the compound eluted at 28.9 min. Analyst, December 1997, Vol. 122 1503 Use of Chemical Ionization in Multianalysis Gas and Liquid Chromatography Combined With a Single Mass Spectrometer for the Ultra-trace Level Determination of Microcontaminants in Aqueous Samples† Arjan J.H. Louter*‡ , Ariadne C. Hogenboom, Jaroslav Slobodn�ýk, Ren�e J. J. Vreuls and Udo A. Th. Brinkman Department of Analytical Chemistry, Free University, De Boelelaan 1083, Amsterdam, 1081 HV, The Netherlands. E-mail: arjan.louter@unilever.com An automated system that comprises a single mass spectrometer (MS) in combination with gas (GC) and liquid chromatography [LC; particle beam (PB) interface] has been used for the trace-level detection and identification of contaminants in water samples.The analytes were enriched by on-line solid-phase extraction (SPE). In the present study, the potential of negative chemical ionization (NCI) detection in this so-called multianalysis system was explored. Attention was devoted to the enhancement of selectivity and sensitivity as well as to the additional spectral information obtained for the identification of unknown compounds.Nine chlorinated pesticides representing three major groups, i.e., triazines, anilides and organophosphorus pesticides, were used as test compounds. Among the three reagent gases used for NCI, isobutane, methane and ammonia, methane gave the best results. For six of the nine pesticides, a 3- to 30-fold increase in sensitivity was observed in the NCI mode as compared with the electron impact (EI) mode. As expected, the NCI mass spectra showed little fragmentation.Electron capture appears to be the dominant ionization mechanism. In order to study the potential of the total on-line SPE–LC/GC–MS set-up, the pesticides were spiked to tap and surface water samples. The detection limits obtained in the NCI (full-scan) mode ranged from 0.1–3 ng l21 for GC–MS and 50 to 200 ng l21 for LC–PB–MS for 100-ml tap water samples. The potential of NCI–MS was demonstrated by the identification of several unknown microcontaminants in a river water sample.Keywords: Gas chromatography–mass spectrometry; liquid chromatography–mass spectrometry; chemical ionization; solid-phase extraction; water samples; pesticides; microcontaminants Mass spectrometry (MS) is the preferred technique for the tracelevel detection of microcontaminants in environmental samples because it can be optimized for target compounds (identification and quantification) as well as unknowns (provisional identification). The complexity of most environmental samples requires a separation step prior to MS detection.Capillary gas chromatography –mass spectrometry (GC–MS) combines a high separation power and selective as well as sensitive detection. For the on-line coupling of column liquid chromatography and mass spectrometry (LC–MS), several interfaces have become available in the past decade. These include thermospray (TSP), particle beam (PB) and atmospheric pressure ionisation (API) interfaces, which offer the use of various ionization modes.1 Although API interfaces are considered to have a promising future because of their excellent sensitivity and the possibility of regulated fragmentation in the ion source,2 the PB interface will certainly remain attractive as the only interface which can generate electron impact (EI) and solvent-independent chemical ionization (CI) spectra.In GC–MS and LC–PB–MS detection, ionization is generally carried out in the positive EI mode because EI–MS provides very reproducible mass spectra and a wide variety of mass spectral libraries containing up to 130000 entries are available for fast identification.3,4 CI with positive (PCI) or negative (NCI) ion detection can provide additional information for provisional identification of unknowns when no relevant information is contained in such MS libraries.5 NCI is especially recognized for the improved selectivity and sensitivity that can be obtained in the detection of chlorinated pesticides.6,7 Usually, only a few ions of high abundance are observed in the NCI mass spectra; this fact enhances analyte detectability if the selected ion monitoring (SIM) mode is applied.8–10 Since the analytes of interest are generally present at very low concentration levels, a concentration step prior to actual analysis is required.Using on-line solid-phase extraction (SPE) for introducing large sample volumes into GC–MS or LC–MS has several advantages. On-line SPE has a good automation potential, requires the use of small amounts of organic solvents and offers a high sensitivity due to the quantitative transfer of the analytes to the chromatograph.Several automated and online SPE–GC–MS and SPE–LC–MS systems have been developed in the recent past.11–16 SPE of microcontaminants from aqueous samples can be carried out on a pre-column with typical dimensions of 10 mm 3 2 mm id, which is packed with a hydrophobic phase such as a polystyrene–divinylbenzene copolymer or C18 bonded silica.In recent studies, SPE–LC has been combined with TSP–MS for target analysis11 and PB–MS for the detection of unknown compounds.12 In SPE–GC, desorption is carried out with a volume of 50–100 ml of ethyl acetate, which can be directly introduced into a gas chromatograph after a proper drying step13 using retention gap techniques involving partially concurrent solvent evaporation.14 The set-up of an automated, software-controlled SPE–GC–MS system has also been reported previously.15,16 In a previous study, a socalled multianalysis system has been described.17 This system, which uses a single MS detector combined with on-line SPE– GC and SPE–LC, was found to be a powerful tool for the detection of microcontaminants in aqueous samples.The present study describes the potential of combined SPE– LC–PB–MS and SPE–GC–MS using NCI detection for the target analysis of chlorinated pesticides and for the identification of unknown compounds in real-life samples.† Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997. ‡ Present address: Unilever Research Laboratorium, P.O. Box 114, 3130 AC Vlaardingen, The Netherlands. Analyst, December 1997, Vol. 122 (1497–1503) 149emicals and Samples Ethyl acetate, hexane, methanol and HPLC-grade water were from J. T. Baker (Deventer, The Netherlands). The ethyl acetate, hexane and methanol were glass-distilled prior to use.The chlorinated pesticides, alachlor, atrazine, bromophos, chlorpyriphos, coumaphos, cyanazine, fenchlorphos, metolachlor and tetrachlorvinphos (see Table 1 for details), were all of 95%, or better, purity and were purchased from Riedel de Haen (Seelze, Germany). Stock solutions of 200 mg l21 were prepared in ethyl acetate and diluted to a final concentration of 10 mg l21 of each compound in a mixture. This was used for direct on-column injections and for the preparation of spiked tap and surface water samples.River water samples were collected from the Nitra River in Slovakia, transported to The Netherlands in a portable refrigerator and stored at 4 °C. The river water samples were filtered through a 0.45 mm acetylcellulose filter (Schleicher and Schuell, Dassel, Germany) prior to preconcentration. Prior to analysis, surface water samples were spiked with metoxuron and propazine (1 mg l21) which served as internal standards.Trace Enrichment Trace enrichment was performed on a Prospekt sample preparation system of Spark Holland (Emmen, The Netherlands). This system consists of three six-port HPLC valves, a cartridge exchange system and a solvent delivery system (SDU). The SDU is provided with a six-port solvent selection valve, a pulse dampener and a single-piston analytical LC pump. Water samples were preconcentrated on 10 3 2.0 mm id stainless-steel pre-columns containing 15–25 mm PLRP-S copolymer (Polymer Laboratories, Church Stretton, Shropshire, UK).All experiments were performed with the set-up shown in Fig. 1. Prior to analyte enrichment, the pre-column was conditioned with 4 ml of methanol and 2 ml of HPLC-grade water. Next, 10 or 100 ml of sample were preconcentrated on the pre-column in order to trap the analytes of interest. Gas chromatography For the combination of the SPE module with GC–MS, a 50 cm 3 50 mm id fused silica capillary was used as a connection between the Prospekt module and the GC on-column injector. After trace enrichment, 1 ml of HPLC-grade water was pumped through the pre-column in order to prevent the introduction of salts into the GC system.Next, the pre-column was dried for 30 min with a stream of nitrogen (40 ml min21). The capillaries were prepressurized with ethyl acetate and, after switching valve V1 to the elute position, the analytes were eluted from the pre-column into the retention gap (details given below).A microMetric metering pump from Milton Roy (Riviera Beach, FL, USA) was used to deliver the desorption solvent, ethyl acetate. Liquid chromatography For the combination of the SPE module with LC–PB–MS the trapped analytes were eluted from the precolumn by the LC gradient onto the analytical column. Gas Chromatography A Hewlett-Packard (Palo Alto, CA, USA) Model 5890 Series II gas chromatograph equipped with a pressure-programmable oncolumn injector was used for GC analysis.The split/splitless injector was replaced by a home-made solvent vapour exit (SVE) which was controlled by a 24 V triggering signal of the GC. The injector was connected to a 5 m 30.32 mm id retention gap, which is essentially an uncoated fused silica capillary deactivated with diphenyltetramethyldisilazane (DPTMDS) (BGB Analytik, Z�urich, Switzerland), a 2 m 3 0.20 mm id, retaining pre-column and a 10 m 3 0.20 mm id capillary GC column, both containing HP-1 [100% dimethyl siloxane (Hewlett-Packard)] with a film thickness of 0.33 mm.Helium (99.99999% purity, Hoekloos, Schiedam, The Netherlands) was the carrier gas at an inlet pressure of 6.0 psi. Connections were made with conventional glass press-frits and a glass pressfrit Y-piece (BGB Analytik). The analytes trapped on the pre-column were eluted with 100 ml of ethyl acetate (75 ml min21) into the retention gap; the excess of solvent was removed through the SVE (see ref. 17 for a detailed description). The observed evaporation rate was 58 ml min21. After closing of the SVE (total open time, 2.2 min), the analytes were transferred to the analytical column for separation and MS detection. The initial temperature was 85 °C (5 min hold time), which was increased to 270 °C at a rate of 10 °C min21. The initial carrier gas pressure was 6.0 psi, which was programmed to 46.0 psi (at 99 psi min21) during the SVE open time in order to Table 1 Relevant information on pesticides used in the present study Relative Molecular Electronegative Analyte Type* mass formula groups (number) Atrazine TRI 215.1 C8H14ClN5 –Cl Alachlor ANA 269.8 C14H12ClNO2 –Cl, –NO2 Fenchlorphos OPP 321.6 C8H8Cl3O3PS –Cl (3), –P ester Cyanazine TRI 240.1 C9H13ClN6 –Cl, C·N Metolachlor ANA 283.8 C15H22ClNO2 –Cl, –NO2 Chlorpyriphos OPP 350.6 C9H11Cl3NO3PS –Cl (3), –P ester Bromophois OPP 366.0 C8H8BrCl2O3PS –Cl (2), –Br, –P ester Tetrachlorvinphos OPP 365.9 C10H9Cl4O4P –Cl (4), –P ester Coumaphos OPP 362.0 C14H16ClO5PS –Cl, P ester * TRI = triazine, ANA = anilide, OPP = organophosphorus pesticide.Fig. 1 Set-up for multianalysis system which combines SPE–GC–MS and SPE–LC–DAD–PB–MS in one integrated set-up. HP 1090, liquid chromatograph; HP 5890 II, gas chromatograph; MS, mass spectrometer; PB, particle beam interface; DAD, UV diode array detector; PROSPEKT, automated valve-switching, solvent selection and cartridge exchange unit; SDU, solvent delivery unit; M, MUST, automated six-port switching valve; S, syringe pump; PR, SPE precolumn; AC, LC analytical column; C, GC analytical column; RP, retaining precolumn; RG, retention gap; SVE, solvent vapour exit; INJ, on-column injector; N2, nitrogen; W, waste; V1–V5 = six-port switching valves. 1498 Analyst, December 1997, Vol. 122enhance the solvent evaporation rate and reduce the transfer time. Next, the pressure was reduced to 6 psi (1.95 min hold time) and programmed to 12.4 psi (at 0.3 psi min21). Liquid Chromatography Analyses were performed on a HP 1090 liquid chromatograph (Hewlett-Packard) equipped with an automatic six-port switching valve (Rheodyne, Berkeley, CA, USA).A 250 3 4.6 mm id stainless-steel column packed with 5 mm C18-bonded silica of 100 Å pore size (Supelco LC-18-DB; Supelchem, Leusden, The Netherlands) was used for separation. A HP 1050 UV diodearray detector (Hewlett-Packard) was operated at 210 nm wavelength with a 10 nm bandwidth. The eluent consisted of a mixture of acetonitrile and HPLCgrade water (Riedel de Ha�en).The analytes were eluted with a linear gradient (flow 0.4 ml min21) during which the acetonitrile content was increased from 10 to 95% (0–45 min) with a hold time of 10 min (45–55 min). Mass Spectrometric Detection A HP 5889A MS Engine (Hewlett-Packard), equipped with a high energy dynode, was used as detector. The MS was operated in the full-scan mode, using the scan range 45–400 amu for positive EI and 65–400 amu for both NCI and PCI detection; the scan rate was 1 scan s21.The operating temperature of the MS ion source was 250 °C and that of the quadrupole, 100 °C. The PB nebulizing gas, high-purity helium (Hoekloos) was kept at 248 kPa (36 psi, about 2.5 ml min21). The temperature of the PB desolvation chamber was 70 °C. Methane, isobutane and ammonia (all three from Hoekloos; purity > 99.999 95%) were used for CI experiments; their pressure in the ion source was maintained at 0.9, 0.7 and 0.9 Torr, respectively. Multianalysis System All parts of the set-up were controlled by the SPE/WIN software (version A.03.14B) of the DOS-based HP Vectra 486/66XM computer (Hewlett-Packard, Waldbronn, Germany).For a detailed description of the operation of the Multianalysis system (Fig. 1), one should consult ref. 21. Results and Discussion With the ion source of the mass spectrometer included in our Multianalysis system, next to the EI mode, both the NCI and PCI mode of operation can be studied for LC–PB–MS ell as GC–MS.This provides the unique situation that spectra obtained after two completely different types of chromatographic separation can be directly compared. As regards CI operation, the influence of the reagent gases on MS response and fragmentation was studied. Methane and ammonia were used for PCI operation; for NCI operation, these two gases and isobutane were studied.Nine chlorinated pesticides, which could be expected to give good responses in the NCI mode, were used as test analytes. Comparison of GC–MS and LC–PB–MS Spectra EI–MS The mass spectra generated in the Multianalysis system using the same detector for GC–MS and LC–PB–MS in the EI mode are presented in Table 2. Eight out of the nine test analytes exhibited the same base peak for both GC–MS and LC–PB–MS. The exception was chlorpyriphos which exhibited m/z 97 {[(HO)2PS]+} as the base peak in the GC–MS spectrum and m/z 314 ([M–Cl]+) in the LC–PB–MS spectrum.The molecular ion was the base peak in the EI spectrum for one analyte (coumaphos) only. This was not unexpected in view of the extensive fragmentation generally observed in the EI mode. In most instances the GC–MS and LC–PB–MS spectra were very similar, but some differences in relative abundances and order of the secondary mass fragments was observed. CI–MS. In the NCI, and also in the PCI mode, the GC–MS and LC–PB– MS spectra were almost identical.They always exhibited the same base peak, and the differences in relative abundance and order of the secondary mass fragments were negligible. No doubt this is partly due to the fact that CI spectra show little fragmentation. Relevant data on NCI and PCI data are presented in Table 3 and are discussed in some detail below. NCI–MS The responses and fragmentation patterns of all analytes were studied by GC–MS with methane, ammonia and isobutane.The mass spectra obtained with the three gases were highly similar. Relevant data on the NCI spectra are presented in Table 3. Despite their electronegative -Cl and -NO2 groups, alachlor and metolachlor were not detected even when 10 ng were injected, while the response of atrazine was just above the detection limit. The rather poor detectability of atrazine in NCI (methane)–MS has also been observed by other workers.6,18,19 To our knowledge, there are no NCI–MS data on alachlor and metolachlor in the literature.The analyte responses invariably were highest with methane as reagent gas; the only exception was atrazine, which showed a 3-fold higher response with ammonia. The results obtained with isobutane were rather poor. To quote an example, 1-mg amounts had to be injected in LC– Table 2 Major ions and relative abundances obtained by GC–MS and LC–PB–MS under EI conditions Ions: m/z (per cent. relative abundance) GC–MS LC–PB–MS Analyte Atrazine 200 (100) 58 (75) 215 (60) 68 (40) 173 (27) 200 (100) 215 (53) 138 (35) 173 (34) 202 (25) Alachlor 45 (100) 160 (51) 188 (42) 146 (20) 122 (14) —* Fenchlorphos 285 (100) 287 (67) 125 (45) 79 (25) 109 (23) 285 (100) 287 (61) 86 (35) 125 (33) Cyanazine 212 (100) 68 (85) 225 (46) 198 (41) 96 (40) 212 (100) 68 (83) 225 (47) 96 (42) 172 (40) Metolachlor 162 (100) 238 (21) 163 (11) 146 (11) 150 (9) 162 (100) 238 (14) 163 (12) 146 (10) 150 (8) Chlorpyriphyos 97 (100) 199 (75) 197 (68) 314 (46) 316 (34) 314 (100) 197 (93) 97 (75) 316 (62) 131 (53) Bromophos 331 (100) 329 (72) 125 (79) 79 (45) 109 (36) 331 (100) 329 (82) 125 (65) 75 (64) 335 (32) Tetrachlorvinphos 109 (100) 329 (89) 331 (83) 333 (30) 207 (21) 109 (100) 329 (81) 331 (74) 333 (30) 79 (19) Coumaphos 362 (100) 226 (49) 109 (41) 364 (41) 97 (36) 362 (100) 226 (77) 109 (75) 97 (67) 210 (48) * Coelutes with metolachlor.Analyst, December 1997, Vol. 122 1499PB–NCI–MS to detect a mere four out of the nine test analytes, viz.tetrachlorvinphos, coumaphos, fenchlorphos and bromophos. For obvious reasons, methane was selected as reagent gas for further work. The NCI spectra of the organophosphorus pesticides exhibited typical group-specific fragments, viz., m/z 125 for the dimethylphosphates and tetrachlorvinphos, m/z 141 for the dimethylphosphorothionates, bromophos (low abundance) and fenchlorphos, and m/z 169 for the diethylphosphorothionates, coumaphos and chlorpyrifos, which were also observed by other workers.20 As for the triazines, the NCI spectrum of atrazine exhibited [M–H]2 as the base peak whereas cyanazine showed [M]2 due to the presence of a C·N group under electron-capture conditions.21 PCI–MS Methane and ammonia were used as reagent gases for LC–PB– PCI–MS, and methane for GC–PCI–MS.As was to be expected, the majority of the PCI mass spectra (Table 3) exhibited [M + H]+ as the base peak.As regards the two exceptions, tetrachlorvinphos and alachlor, the former analyte exhibited m/z 127 as the base peak, which can be assigned to [C2PH8O4]+ (counterpart of the m/z 125 fragment due to [C2PH6O4]2 observed in NCI–MS). Alachlor exhibited m/z 238, due to [M2CH3O]+, as its base peak. For six out of the nine compounds the MS responses in the PCI mode were substantially lower than in the NCI mode. However, the detectability of atrazine, metolachlor and alachlor was better in the PCI (methane) mode.When ammonia was used as a reagent gas for LC–PB–PCI– MS, atrazine gave a 2–3-fold better response than with methane. Prominent [M + C2H5]+ and [M + C3H5]+ adduct ions (with methane, m/z 244 and 256, respectively) were detected for atrazine, which have also been observed by other workers.14,22 For the other eight analytes [M + C2H5]+ adduct ions were also detected, but their intensity was typically about 10% of that of the [M + H]+ peak and, invariably, the responses were lower than with atrazine.As an example, Fig. 2 shows the EI, NCI and PCI mass spectra of coumaphos obtained by LC–PB–MS. In the EI mode, the base peak can be assigned to the molecular ion (m/z 362), while in the NCI spectrum the base peak observed at m/z 225 is due to the thiophenolate [M2C4H12O3P]2 ion, and m/z 227 to its 37Cl isotope. The peaks observed at m/z 191 and 169 are due to [M2C4H12O3PS]2 and [C4H10O3PS]2, respectively. The latter is the group-specific fragment already mentioned above.The base peak in the PCI mode (m/z 363) can be assigned to [M + H]+, while m/z 329 is due to [M - Cl + H]+. The peak observed at m/z 391 is due to a [M + C2H5]+ adduct. Analytical Data of Chlorinated Pesticides by SPE–LC–PB–MS/GC–MS (EI and NCI Detection) The analytical performance of the GC and LC–PB parts of the multianalysis system were tested in the EI and NCI mode, using Table 3 Relative molecular masses (Mr), major ions and relative abundances obtained under NCI and PCI conditions (GC; 10 ng injection) Ions: m/z (per cent.relative abundance) NCI PCI Analyte Mr Atrazine 215 214 (100) 179 (30) 216 (100) 180 (95) 218 (32) 244 (17) 256 (10) Alachlor 269 238 (100) 240 (34) 270 (14) Fenchlorphos 320 213 (100) 211 (87) 215 (36) 141 (20) 270 (10) 323 (100) 321 (97) 125 (86) 111 (39) 285 (33) Cyanazine 240 240 (100) 242 (34) 212 (18) 120 (17) 204 (16) 214 (100) 241 (55) 205 (41) 178 (38) Metolachlor 283 284 (100) 252 (78) 286 (33) 254 (27) 238 (26) Chlorpyriphos 349 313 (100) 315 (73) 214 (64) 212 (61) 169 (60) 350 (100) 352 (96) 354 (34) 153 (28) 322 (23) Bromophos 364 257 (100) 270 (65) 255 (59) 259 (52) 272 (47) 367 (100) 365 (60) 369 (30) 125 (30) 331 (22) Tetrachlorvinphos 364 125 (100) 224 (14) 222 (13) 127 (100) 367 (18) 365 (16) Coumaphos 362 225 (100) 227 (49) 191 (26) 169 (23) 362 (21) 363 (100) 329 (42) 365 (33) 391 (11) 139 (10) Fig. 2 LB–PB–MS spectra of coumaphos obtained in the EI mode, methane–NCI and PCI mode. 1500 Analyst, December 1997, Vol. 122on-line SPE to improve analyte detectability. Table 4 summarizes the detection limits (S/N = 3) that were obtained for 100 ml tap water samples. For six out of the nine test compounds, detectability in the NCI mode is distinctly, i.e., 10–30-fold, better than in the EI mode. However, there are notable exceptions as well, although these are not really unexpected after the earlier discussion of the NCI–MS data.Another striking observation is, of course, that for the same mass spectrometer, the detectability in LC–PB–MS is two–three orders of magnitude less good than in GC–MS, i.e. in the low to sub ng l21 versus the mg l21 range for the (identical) samples analysed. Fig. 3 shows SPE–GC–MS chromatograms of the analysis of 10 ml spiked (1 mg l21) tap water run in the NCI and EI mode. Analyte detectability can be improved by one order of magnitude with time-scheduled selected ion monitoring (SIM), as is demonstrated in Fig. 4, where extracted ion and SIM chromatograms are compared. Obviously, for the six test compounds showing high analyte response in NCI–MS detection, both LC- and GC-based separation can conveniently be used for ultra-trace level confirmation purposes at levels required by legislative bodies (typically, 0.1 mg l21 for tap water). Actually, it can be inferred from our data that with timescheduled operation, 1 ml of water sample will be sufficient to obtain the required sensitivity. The linearity was tested in both the GC and LC operation of the system. For 10-ml samples values of the squares of the regression coefficients (R2) for SPE–GC–NCI–MS ranged from 0.9906 to 0.9929 (4 data points; range 0.01–1 mg l21); in the EI mode, the values ranged from 0.9981 to 0.9999 (4 data points; range 0.05–5 mg l21).As was to be expected on the basis of published experience,23 for SPE–LC–PB–NCI–MS curved rather than linear calibration plots were obtained. Still, as was also observed before, for the rather limited concentration range of interest in studies of environmental samples (typically, 0.1–10 mg l21), a three- or four-point calibration will enable reliable quantification.21 Application: Identification of Unknown Microcontaminants As an application the combined EI and NCI operation was used for the analysis of a river water sample by SPE–GC–MS and SPE–LC–PB–MS.In the SPE–GC–MS chromatograms of the surface water sample (Fig. 5), the EI and NCI traces were found to be distinctly different, with several compounds (e.g., No. 1 and 2) responding better in the latter mode. The EI and NCI spectra of the unknown compound eluting at 12.4 min (peak No. 1) indicate it to be (2-methylthio)-benzothiazole, presumably a transformation product of benzothiazole, which was detected in the same sample with SPE–LC–DAD–UV. The base peak in the NCI mass spectrum is most likely to be the thiolate anion (m/z 166; [C7NS2H4]2), which is known to be very stable. The relative abundance of the peak at m/z 168 was 8.7%, which is supporting evidence for the presence of two sulfur atoms.The base peak in the EI mode was m/z 181, which is the molecular ion; here the typical sulfur isotope peak (m/z 183; relative abundance, 8.9%) was also observed. The spectrum gave a good match (94 out of 100) with the Wiley library spectrum. The unknown compound, designated peak No. 2, was more difficult to identify. The EI spectrum [Fig. 6(a)] is obviously a mixture of two or three coeluting compounds. The corresponding NCI spectrum [Fig. 6(b)] shows much better selectivity or, in other words, a clearer mass spectrum. Again, the base peak in the NCI mass spectrum was m/z 166, with a sulfur isotope peak at m/z 168. On the basis of the m/z 166 and m/z 168 peaks, the peak at m/z 182 can provisionally be assigned to [C7NS2H4O]2 (see Fig. 6, below), also because of the peak ratio typical for two sulfur atoms (relative abundance of m/z 184, 8.7% of m/z 182).The EI spectrum contained characteristic ions at m/z 108, 134, 148, 166 and 197, which support the presence of an oxidation product of (2-methylthio)-benzothiazole with a relative molecular mass of 197. The same sample was also analysed with SPE–LC–PB–MS. Despite the known low sensitivity of PB–MS for benzothiazole- Table 4 Detection limits of on-line SPE–GC–MS (ng l21) and SPE–LC– PB–MS (mg l21) for 100 ml tap water samples SPE–GC–MS SPE–LC–PB–MS Analyte EI NCI EI NCI Atrazine 1 100 0.5 > 10 Alachlor 7 — 7 — Fenchlorphos 5 0.2 5 0.2 Cyanazine 10 3 0.5 0.15 Metolachlor 0.5 — 0.5 — Chlorpyriphos 30 1 5 0.2 Bromophoa 5 0.3 3 0.2 Tetrachlorvinphos 1 0.1 0.5 0.15 Coumaphos 30 2 0.7 0.05 Fig. 3 Full-scan SPE–GC–MS chromatograms (EI and methane–NCI) of 10 ml tap water spiked at the 1 mg l21 level. Peak designation; 1, atrazine; 2, alachlor; 3, fenchlorphos; 4, cyanazine; 5, metolachlor; 6, chlorpyriphos; 7, bromophos; 8, tetrachlorvinphos; 9, coumaphos. Fig. 4 SPE–GC–NCI–MS chromatograms of 10 ml tap water spiked at the 10 ng l21 level with chlorinated pesticides. Left, extracted ion chromatograms of m/z of base peak of investigated compounds. Right, TS SIM chromatogram. Peak designation; 1, fenchlorphos (m/z 213); 2, chlorpyriphos (m/z 313); 3, bromophos (m/z 257); 4, tetrachlorvinphos m/z 125); 5, coumaphos (m/z 225). Analyst, December 1997, Vol. 122 1501type compounds, the oxidation product of benzothiazole proposed above was detected under NCI conditions, as is evident from the extracted ion chromatogram of m/z 166 in Fig. 7. Here, 200 ml of sample were preconcentrated to obtain sufficient sensitivity. In the EI mode the most abundant LC peak had a mass spectrum with m/z 138 as base peak [Fig. 7, insert (a)]. Insert (b) shows the NCI mass spectrum recorded for the same peak, which is dominated by m/z 222.The relative molecular mass of the unknown compound probably is 254, as can be derived from the peaks at m/z 254 and 253, designated as [M]2 and [M-H]2 in the NCI spectrum and as [M]+ in the EI spectrum, respectively, with m/z 222 then representing [MSH]+. In both the EI and NCI modes the peak at m/z 256 had a relative abundance of 20% of the m/z 254 (molecular ion) peak, which indicates the presence of four or five sulfur atoms. Unfortunately, further identification was not possible on the basis of these data.This will probably require the use of tandem MS techniques. Conclusions An automated, so-called multianalysis, system which combines SPE–LC–PB–MS and SPE–GC–MS using only one sample preparation module and one mass spectrometric detector, has been used to study CI–MS detection. Nine pesticides containing different electronegative groups were used as test analytes. As expected, NCI detection provided enhanced sensitivity and selectivity compared with EI detection for most of the analytes.For five chlorinated pesticides, a 10–30-fold increase in sensitivity was observed. However, the presence of a chlorine or nitro-substituent did not always give an increased response; atrazine, alachlor and metolachlor were such exceptions. The LC–PB–MS and GC–MS spectra recorded on the same MS detector were found to be very similar in the EI mode and almost identical in the NCI mode. For six of the nine test analytes, the detection limits in on-line SPE–LC/GC–NCI–MS (full-scan) were 0.1–3 ng l21 (GC) and 50–200 ng l21 (LC–PB) for 100-ml tap water samples.Obviously, for on-line SPE–GC samples sizes of 1–10 ml are more appropriate in view of the directives for detection limits in drinking water. For SPE–LC–PB–MS the NCI mode is an interesting option for sub mg l21 detection and identification of a number of microcontaminants in water. SPE–LC approaches using TSP–MS or API–MS may be more sensitive, but they are not ‘universal’ either, and PB–NCI–MS has the advantage of providing more structural information.These results again Fig. 5 Chromatograms of 10 ml river Nitra water samples obtained by SPE–GC–MS in the full-scan EI and methane–NCI mode. The sample was spiked with 1 mg l21 of the internal standard propazine (P). (a) EI and (b) NCI spectra of the compound eluted at 12.4 min are shown as inserts. Fig. 6 EI and methane–NCI spectra of unknown peak 2 from Fig. 5. 1502 Analyst, December 1997, Vol. 122indicate that more attention should be devoted to improve the efficiency of the particle beam interface. The practicality of NCI detection in the combined SPE–LC– PB–MS and SPE–GC–MS approach was demonstrated by the (partial) identification of unknown microcontaminants in rather polluted sample where EI detection alone was not sufficient. We thank the European Union (EV5V–CT92–0105) and the River Basin Program (Amsterdam) for their financial support. Dr. Ben van Baar is acknowledged for valuable discussions. References 1 Niessen, W. M. A., and van der Greef, J., Liquid Chromatography– Mass Spectrometry, Principles and Applications, Marcel Dekker, New York, USA, 1992. 2 Voyksner, R. D., and Peck, T. D., Rapid Commun. Mass Spectrom., 5, (1991) 263. 3 Levsen, K., Org. Mass. Spectrom., 1988, 23 406. 4 Schmidt, E. R., Chromatographia, 1990, 30, 573. 5 Mattina, M. J. I., J. Chromatogr., 1991, 542, 385. 6 Ong, V. S., and Hites, R. A., Mass Spectrom. Rev., 1994, 13, 259. 7 Mattina, M. J. I., Trends Anal. Chem. 1993, 12, 328. 8 Busch, K. L., Bursey, M. M., Hass, J. R., and Sovocool, G. W., Appl. Spectrosc., 1978, 32, 388. 9 Stan, H.-J., and Kellner, G., Biomed. Environ. Mass Spectrom., 1982, 9, 483. 10 Stan, H.-J., and Kellner, G., Biomed. Environ. Mass Spectrom., 1989, 18, 645. 11 Bagheri, H., Brouwer, E. R., Ghijsen, R. T., and Brinkman, U. A. Th., J. Chromatogr., 1993, 647, 121. 12 Bagheri, H., Slobodnik, J., Marc�e Recasens, R. M., Ghijsen, R. T., and Brinkman, U. A. Th., Chromatographia, 1993, 37, 159. 13 Vreuls, J. J., Ghijsen, R. T., de Jong, G. J., and Brinkman, U. A. Th., J. Chromatogr., 1992, 625, 237. 14 Grob, K., Fr�olich, D., Schilling, B., Neukom, H.-P., and N�ageli, P., J. Chromatogr., 1984, 295, 55. 15 Louter, A. J. H., Brinkman, U. A. Th., and Ghijsen, R. T., J. Microcol. Sep., 1993, 5, 303. 16 Louter, A. J. H., van Beekvelt, C. A., Cid Montanes, J. Slobodn�ýk, J., Vreuls, J. J., and Brinkman, U. A. Th., J. Chromatogr., 1996, 625, 67. 17 Slobodnik, J., Hogenboom, A. C., Louter, A. J. H., and Brinkman, U. A. Th., J. Chromatogr., 1996, 730, 353. 18 Huang, L. Q., and Mattina, M. J. I., Biomed. Environ. Mass Spectrom., 1989, 18, 828. 19 Bagheri, H., Vreuls, J. J., Ghijsen, R. T., and Brinkman, U. A. Th., Chromatographia, 1992, 34, 5. 20 Durand, G., and Barcel�o, D., Anal. Chim. Acta, 1991, 243, 259. 21 Slobodnik, J., Ph.D. Thesis, Free University, Amsterdam, The Netherlands, 1996. 22 Crespo, C., Marc�e, R., and Borull, F., J. Chromatogr. A, 1994, 670, 135. 23 Bellar, T. A., Behymer, T. D., and Budde, W. L., J. Am. Soc. Mass Spectrom., 1990, 1, 92. Paper 7/05209J Received July 21, 1997 Accepted September 25, 1997 Fig. 7 Chromatograms of 200 ml river Nitra water samples obtained by SPE–LC–PB–MS in the full-scan EI and methane–NCI mode. Samples were spiked with 1 mg l21 of internal standards propazine (P) and metoxuron (M). MS chromatograms represent extracted ions obtained at the m/z values of the base peak of each investigated analyte. The inserts show the (a) EI and (b) methane NCI spectra of the compound eluted at 28.9 min. Analyst, December 1997, V

ISSN:0003-2654    

DOI:10.1039/a705209j

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Direct On-filter Immunoassay of Some b-Lactam Antibiotics for Rapid Analysis of Drug Captured From the Workplace Atmosphere† Frederick J. Rowell*a, Zhi-Feng Miaoa, Roger N. Reevesa and Robert H. Cummingb a North East Biotechnology Centre, School of Health Sciences, University of Sunderland, Sunderland, UK SR2 7EE b North East Biotechnology Centre, School of Science and Technology, University of Teesside, Middlesbrough, UK TS1 3BA A simple competitive enzyme-linked immunoassay for the antibiotic ceftazidime and structurally similar b-lactam antibiotics has been developed which can be performed directly on the surface of a cellulose nitrate filter used to capture the airborne drug during workplace monitoring for health and safety purposes.Post sampling analysis is performed on the filter retained within the air sampler. It involves two steps; the first a 10 min incubation in which the captured drug is dissolved and competes with drug immobilised within a protein conjugate on the surface of the filter for an enzyme-labelled antibody reagent, and the second, following washing under vacuum in situ, a 5 min incubation of substrate solution when colour develops on the spot at the location of the immobilised drug–protein conjugate.The intensity of the spot can be assessed visually within the sampler to ascertain the presence or absence of captured drug, or quantitative results can be obtained using an optical scanner. The intensity of the spots is linear from 10 ng to 1 mg (r2 = 0.9996, n = 3) and the limit of detection is 1.9 ng of captured drug (10 ng for visual discrimination between this standard and the zero). The assay is precise with between-assay RSD values of < 4% over the linear range of the assay.Keywords: Personal monitoring; workplace atmosphere sampling; enzyme-linked immunoassay; b-lactam antibiotics; health and safety Beta-lactam antibiotics are potential skin and respiratory sensitising agents and occupational exposure limits (OELs) have been adopted within industry for their airborne concentrations.The current method for monitoring individual worker exposure is to capture the aerosol from the workplace air onto a filter which is located on a sampler attached to the clothing of the worker. Air is sucked through the filter at about 2 l min21 with a portable battery driven pump which is also carried by the worker. Since the current OEL for these drugs is 100 mg m23,1 sampling for 15 min periods to obtain time weighted average values will deposit a mass of 3 mg of the drug on the filter for air containing the drug at the OEL.We have developed rapid ELISAs for determining the cephalosporin, ceftazidime, and have used them on site to determine levels of captured drug eluted from filters following sampling.2,3 Although such ELISAs have advantages over the current HPLC method for post sampling analysis,4 in that they can be performed on site immediately following sampling due to their simplicity, both HPLC and ELISA methods require removal of filters from the filter holders and extraction of captured drug prior to analysis.A better approach would be to perform such ELISA determinations directly on the filter still retained within the sampler, to generate a coloured signal which would indicate whether the OEL had been exceeded or not on sampling. We now report on such an assay. Experimental Chemicals All reagents were purchased from Sigma (Poole, Dorset, UK) unless otherwise stated.The buffer salts were of analyticalreagent grade (BDH, Poole, Dorset, UK). Ceftazidime pentahydrate was a gift from GlaxoWellcome (Barnard Castle, Co. Durham, UK). The antiserum used had been raised in rabbits to a ceftazidime–porcine thyroglobulin conjugate5 and was affinity purified as previously described3 to isolate the ceftazidimespecific antibodies. Phosphate buffered saline with Tween, pH 7.4 (PBST); phosphate buffered saline, pH 7.4 (PBS); methanol –sodium dihydrogen phosphate (1 + 3), pH 5.5; carbonate buffer, pH 9.6; and diethanolamine buffer, pH 9.8 were prepared as previously described.5 5-Bromo-4-chloro-3-indolyl phosphate (BCIP) stock solution was 50 mg ml21 BCIP in 100% dimethyl formamide (DMF).Nitroblue tetrazolium (NBT) stock solution was 75 mg ml21 NBT in 70% DMF. Substrate solution for alkaline phosphatase (AP) was prepared from a mixture of BCIP stock, NBT stock and carbonate buffer with a volume ratio of 1 : 1 : 100.Equipment Cellulose nitrate (CN) membrane filters, diameter 25 mm, pore size 0.1 mm, were from Whatman, Maidstone, Kent, UK. Cellulose nitrate membrane filter images were captured by an EPSON GT8000 scanner, EPSON UK, Hemel Hempstead, Hertfordshire, UK, and Micrografix Picture Publisher V3.1 picture processing software, Micrografix, Woking, Surrey, UK. The intensities of colour on CN membrane filters were measured by Band Leader Application V2.01 shareware software, produced by Maayan, Techknowledge, Israel.Institute of Occupational Medicine (IOM) personal inhalable dust samplers for the membrane filters were purchased from SKC, Blandford, Dorset, UK. Preparation of the Single Antibody Reagent The single antibody reagent was prepared from a mixture of equal volumes of affinity purified 1:1000-diluted rabbit anticeftazidime serum5 in PBST and 1 : 175-diluted goat anti-rabbit IgG conjugated with AP in PBST.The mixture was preincubated for over 1 h at room temperature prior to use.2 This reagent was then stable for over 7 d at 4 °C. Preparation of Coated CN Membrane Filters Cellulose nitrate membrane filters were placed in IOM samplers and treated as follows. The filters were coated with 1.5 ml of 1, † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997. Analyst, December 1997, Vol. 122 (1505–1508) 150510 or 60 mg ml21 ceftazidime–BSA conjugates in coating buffer as single spots at its centre.After incubation for 10 min at room temperature the now dry spots on the filters were washed 3 times with 3 ml of PBST under vacuum which was supplied by a pump to suck PBST through filters from top to bottom. The filters were allowed to air dry for about 10 min when aliquots of 25 ml of PBST containing different amounts of ceftazidime (0.1 ng to 100 mg) were added as single drops on the above treated filters adjacent to the central spots.The filters stood at room temperature until air dried (about 10 min). They were then stored desiccated in plastic envelopes and sealed in aluminium foil until used in the assay. Assay Optimisation Aliquots (300 ml) of single antibody–enzyme reagent were added dropwise to cover the whole surface of each filter. The filters were then incubated for 2–10 min at room temperature with 200 motions min21 shaking (Janke and Kunkel HS 250 shaker from S.H.Scientific, Blyth, Northumberland, UK). After incubation with this single reagent, the filters were washed with 3 ml of distilled water 3 times under vacuum. To the now dry filters were added 10 ml of substrate solution as a drop to the centre of each filter. After incubation for 5–13 min at room temperature, the vacuum was re-applied for 1 min to remove excess substrate and to terminate colour development. The colour intensity of the central spots on the filters was then ascertained visually or by use of an EPSON scanner using Band Leader software. Effect of Air Flow on the Coated Filters The CN filters coated with drug–protein conjugate and ceftazidime standards (0 and 100 ng) and filters not coated with drug–protein conjugate were placed in IOM samplers which were then subjected to a vacuum to induce an air flow through the filters of clean laboratory air for 20 min at a rate of 1.5 l min21 (equivalent to exposure of the filters to a total volume of 30 l of air).The filters were then used in the standard ELISA together with identical coated and uncoated filters which had not been subjected to this treatment and the intensities of the resulting spots determined. Precision of the Assay and its Limit of Detection (LOD) Replicates of coated filters (n = 6) spiked with 0, 10 ng, 100 ng and 1 mg of ceftazidime were processed in the ELISA to determine the RSD at each spiking concentration. The standard deviation of the zero filter was used to calculate the assay’s LOD at 95% confidence from the resulting standard curve.Specificity of the Assay Pre-coated filters were spiked with different antibiotics of the penicillin, cephalosporin and non-b-lactam types, with no drug and 1 mg of drug per filter (n = 3 per standard). Each batch of filters was subjected to the assay and the intensities of the resulting spots determined. The cross-reaction of the drugs was calculated as a percentage of the signal reduction seen with the ceftazidime spiked filters.Stability of the Coated Filters on Storage Thirty-six filters pre-coated with the drug–protein conjugate were stored as three batches: at 4 °C, at room temperature in a desiccator, or at 37 °C. In each case the filters were stored as described above. The filters were removed at 7 and 28 d and spiked with drug standards (zero and 100 ng per filter). They were then subjected to the ELISA and the intensities of the resulting spots compared with those of corresponding freshly prepared and spiked filters.Results Optimisation of the Assay The optimal conditions and protocol chosen for the assay were as follows: (i) coating concentration of 1.5 ml of a 60 mg ml21 solution of drug–BSA conjugate; (ii) single antibody reagent, 300 ml; (iii) initial incubation time of 10 min; (iv) wash conditions under vacuum, 1 min with 3 3 3 ml of distilled water; (v) substrate addition, 10 ml applied to the centre of the filter; (vi) substrate incubation time, 5 min followed by application of vacuum for 1 min; and (vii) assess colour intensity of central spots on the sample filters against those of the zero and 1 mg standards (and additional standards if required) either by eye with the filters within the samplers or by use of an optical scanner following removal of the filters from their sample holders.A typical standard curve obtained following scanning is shown in Fig. 1. The plot is linear between 10 ng and 1 mg (r2 = 0.9996, n = 3, p < 0.001). Effect of Air Flow on the Coated Filters and Filter Stability Significant differences in intensities of the spots were observed for drug-spiked and unspiked filters exposed to an air flow of 1.5 l min21 for 20 min and identically treated filters which were not exposed to this treatment (p < 0001, Student’s t-test). The percentage fall in intensity of the unspiked (zero) filter was 25% of the signal of the corresponding zero filter not exposed to passage of 30 l of air through the filter over 20 min but a satisfactory standard curve was obtained under these conditions.Reduction of the sampling time to 5 min now gave no significant differences in these intensities. Limit of Detection and Precision of the Assay The assay’s LOD was 1.9 ng per filter at 95% confidence limits following scanning. Visual observation of the intensities of spots for the standards enabled discrimination of the zero filter and the filter spiked with the 10 ng standard.The RSD values at Fig. 1 Standard curve for the on-filter ELISA for ceftazidime following scanning of the filters using an optical scanner. Signals are shown as means ± 1 s. 1506 Analyst, December 1997, Vol. 12210 ng, 100 ng and 1 mg per filter (n = 3 per concentration) were 3.7, 2.2 and 1.9%, respectively. Specificity of the Assay The relative signal differences observed for the drugs examined compared with those seen between the 0 and 1 mg standards for ceftazidime are shown in Table 1.Cross reaction was only observed with antibiotics possessing an intact b-lactam substructure. Thus, certain penicillins and cephalosporins having similar side chains to ceftazidime can be detected in the assay although with less sensitivity than for ceftazidime, while structurally dissimilar antibiotics and non-antibiotic drugs such as chloramphenicol, ondansetron and dexamethasone do not interfere in the assay.Stability of the Coated Filters on Storage The signal differences following ELISA development between filters stored at 4 °C, room temperature and at 37 °C for 1 and 4 weeks are shown in Fig. 2. Signal differences between the zero and 100 ng standards after storage are compared as a percentage of the signal drop observed for these standards with filters which were freshly coated with the drug–protein conjugate. The coated filters are stable at ambient temperature and at 4 °C for 4 weeks since the signal differences observed for these conditions are 89% and 66%, respectively, of the signal drop for corresponding standards on the freshly coated filters.This compares with a corresponding value of only 37% for the filters stored at 37 °C for this period. Discussion The conditions for a direct filter-based ELISA for determining the drug ceftazidime captured during air sampling have been determined with respect to the choice of filter (CN was found superior to commercial Teflon and PTFE membrane filters, data not shown; glass wool filters were not considered due to the known instability of the drug on this type of filter3), concentration of coating conjugate on the filter, constitution of the antibody reagents, and incubation times.This has resulted in an assay that can be performed within 17 min without the need to remove the filters from the personal monitors used in the air sampling stage. The end point is a stable blue spot, the colour of which indicates the concentration of drug captured on the filter.This intensity can be observed by eye; the presence of 10 ng or more of drug can be discriminated from the intensity of the spot corresponding to the absence of the drug on a control filter. Alternatively the intensities can be determined using an optical scanner when the limit of detection is 1.9 ng per filter at 95% confidence limits.In this format the standard curve is linear over the range 10 ng to 1 mg. The assay is precise with RSD below 4% over the linear part of the standard curve. The performance of the assay was also unaffected by the passage of 10 l of air through the filters at a rate corresponding to that normally used for personal air monitoring in industry, namely about 2 l min21. This sampling takes 5 min. Although this is less than the 30 l normally sampled over 15 min, the sensitivity of the assay is sufficient to accommodate this smaller sample volume.If the air contained the drug at its OEL of 100 mg m23, a total mass of 3 mg of drug would be captured on the filter at the end of a 15 min sampling period. This mass is outside the linear range of the assay but the reduced sampling time of 5 min would capture 1 mg of drug on the filter which is within this range. The on-filter ELISA produces a spot which could be easily distinguished by eye from the corresponding spots on a control filter without drug and a standard filter spiked with 1 mg of drug.Alternatively the assay’s sensitivity can be reduced by increasing the amount of drug–protein conjugate immobilised on the filter (details not shown) thereby enabling a 15 min sampling period to be followed. Another format is also under investigation in which a number of spots each containing a different amount of drug–protein conjugate are immobilised on each filter. This should produce a simple test to determine the mass of captured drug over a wider concentration range than is observed in the assay described in this paper and thus enabling a range of differing sampling times to be used according to the needs of the user.The spots thus obtained on the filters are stable for many months. The hapten-coated filters are also reasonably stable with 11% signal loss following storage within a desiccator for 4 weeks at ambient temperature. Storage at 37 °C for 4 weeks resulted in signal loss by 73%. The filters can be processed within their sampling devices and the whole process of reagent addition, sample incubation and shaking and filtration could easily be automated leading to a free-standing assay for batch analysis of many samples.We currently process a batch of 12 inter-linked samplers attached to a vacuum system within a shaker for one assay run but use manual addition of reagents. Table 1 Specificity of the assay tested at a concentration of 1 mg (free base) per filter for each drug Difference Cross in signal reactivity intensity Structural similarity relative to (1 mg– ceftazidime Drug zero drug) R1 R2 b-Lactam† (%) Ceftazidime 44 *** *** Yes 100 Cefotaxime 15 * X‡ Yes 36 Cephaloridine 0§ X *** Yes 0 Cephalexin 22 X X Yes 49 Cefuroxime 0§ X X Yes 0 Cefixime 25 ** X Yes 56 Penicillin G 21 X X Yes 47 Ampicillin 8 X X Yes 19 Chloramphenicol 0§ na¶ na No 0 Ondansetron 0§ na na No 0 Dexamethasone 0§ na na No 0 † See Fig. 3 for the structures of the cephalosporins used in this study. * Similar substituent to ceftazidime. ** Very similar substituent to ceftazidime. *** Identical substituent to ceftazidime. ‡ X; Structurally dissimilar substituent to ceftazidime. § No significant difference between the signal at 1 mg of drug and the signal in the absence of drug. ¶ na; Not applicable since not a b-lactam antibiotic. Fig. 2 Stability profiles of filters precoated with drug–protein conjugates and stored under various conditions for 1 and 4 weeks.The results are expressed as the percentage of the ELISA signal difference between coated filters spiked with 100 ng and no drug (zero) for freshly coated compared with stored filters. Analyst, December 1997, Vol. 122 1507N S CONH O COOH R2 H H R1 Cephalosporins N S NH2 N O C(CH3)2COOH R1 N+ CH2 R2 Ceftazidime S CH2 N+ CH2 Cephaloridine N S NH2 N OCH2COOH CH CH2 Cefixime N S NH2 N OCH3 CH2OCOCH3 Cefotaxime Cefuroxime Sodium O O NOCH3 CH2OOCNH2 CH NH2 Cephalexin CH3 From the results of the specificity study, the system appears to be applicable to the detection of examples of both penicillins and cephalosporins.The antiserum thus appears to recognise drugs possessing intact b-lactam rings carrying side chains similar to ceftazidime through the amide bridge (R1 in Fig. 3) (Table 1) since cefixime exhibits the greatest cross reaction (56%). The antiserum appears to accommodate relatively simple R1 phenyl-methylene side chains associated with cephalexin, ampicillin and penicillin G whilst being more selective to cephalosporins which have dissimilar R2 substituents to ceftazidime. Thus cephaloridine which has an identical R2 substituent to ceftazidime but a dissimilar R1 side chain exhibited no significant cross reaction in the assay. The greater cross reaction for cephalexin (49%) compared with ampicillin (19%) represents discrimination of the two ring structures since the former is a cephalosporin and the latter a penicillin although both have identical R1 side chains.This partial generic specificity renders the assay of some value for these classes of compounds for which workplace atmospheric monitoring is required. Further studies are required to assess the sensitivity, precision and robustness of the system with additional examples of these compounds. The system will now be further validated under actual workplace conditions and the results obtained correlated with those obtained for samples by HPLC.The on-filter ELISA described herein should be applicable to other airborne biochemicals for which such workplace monitoring is required and for which antibodies can be produced. The method could also find application for general environmental air monitoring outside the workplace and for screening for concealed illicit substances and explosives. References 1 Ceftazidime Product Information. GlaxoWellcome Operations, Barnard Castle, Co.Durham, UK, 1992. 2 Cumming, R., Farrell, C., Nitescu, I., Rowell, F. J., and Tang, L. X., Anal. Chim. Acta, 1995, 311, 377. 3 Rowell, F. J., Farrell, C., Nitescu, I., Cumming, R. H., and Stewart, I. W., J. Aerosol Sci., 1997, 28, 493. 4 Leeder, J. S., Spino, M., Tesoro, A. M., and MacLeod, S. M., Antimicrob. Agents Chemother., 1983, 24, 720. 5 Farrell, C., Rowell, F. J., D’Silva, C., and Cumming, R. H., Analyst, 1994, 119, 2411. Paper 7/04728B Received July 4, 1997 Accepted August 20, 1997 Fig. 3 Structures of the cephalosporins used in the study. 1508 Analyst, December 1997, Vol. 122 Direct On-filter Immunoassay of Some b-Lactam Antibiotics for Rapid Analysis of Drug Captured From the Workplace Atmosphere† Frederick J. Rowell*a, Zhi-Feng Miaoa, Roger N. Reevesa and Robert H. Cummingb a North East Biotechnology Centre, School of Health Sciences, University of Sunderland, Sunderland, UK SR2 7EE b North East Biotechnology Centre, School of Science and Technology, University of Teesside, Middlesbrough, UK TS1 3BA A simple competitive enzyme-linked immunoassay for the antibiotic ceftazidime and structurally similar b-lactam antibiotics has been developed which can be performed directly on the surface of a cellulose nitrate filter used to capture the airborne drug during workplace monitoring for health and safety purposes.Post sampling analysis is performed on the filter retained within the air sampler.It involves two steps; the first a 10 min incubation in which the captured drug is dissolved and competes with drug immobilised within a protein conjugate on the surface of the filter for an enzyme-labelled antibody reagent, and the second, following washing under vacuum in situ, a 5 min incubation of substrate solution when colour develops on the spot at the location of the immobilised drug–protein conjugate. The intensity of the spot can be assessed visually within the sampler to ascertain the presence or absence of captured drug, or quantitative results can be obtained using an optical scanner.The intensity of the spots is linear from 10 ng to 1 mg (r2 = 0.9996, n = 3) and the limit of detection is 1.9 ng of captured drug (10 ng for visual discrimination between this standard and the zero). The assay is precise with between-assay RSD values of < 4% over the linear range of the assay. Keywords: Personal monitoring; workplace atmosphere sampling; enzyme-linked immunoassay; b-lactam antibiotics; health and safety Beta-lactam antibiotics are potential skin and respiratory sensitising agents and occupational exposure limits (OELs) have been adopted within industry for their airborne concentrations.The current method for monitoring individual worker exposure is to capture the aerosol from the workplace air onto a filter which is located on a sampler attached to the clothing of the worker. Air is sucked through the filter at about 2 l min21 with a portable battery driven pump which is also carried by the worker.Since the current OEL for these drugs is 100 mg m23,1 sampling for 15 min periods to obtain time weighted average values will deposit a mass of 3 mg of the drug on the filter for air containing the drug at the OEL. We have developed rapid ELISAs for determining the cephalosporin, ceftazidime, and have used them on site to determine levels of captured drug eluted from filters following sampling.2,3 Although such ELISAs have advantages over the current HPLC method for post sampling analysis,4 in that they can be performed on site immediately following sampling due to their simplicity, both HPLC and ELISA methods require removal of filters from the filter holders and extraction of captured drug prior to analysis.A better approach would be to perform such ELISA determinations directly on the filter still retained within the sampler, to generate a coloured signal which would indicate whether the OEL had been exceeded or not on sampling.We now report on such an assay. Experimental Chemicals All reagents were purchased from Sigma (Poole, Dorset, UK) unless otherwise stated. The buffer salts were of analyticalreagent grade (BDH, Poole, Dorset, UK). Ceftazidime pentahydrate was a gift from GlaxoWellcome (Barnard Castle, Co. Durham, UK). The antiserum used had been raised in rabbits to a ceftazidime–porcine thyroglobulin conjugate5 and was affinity purified as previously described3 to isolate the ceftazidimespecific antibodies. Phosphate buffered saline with Tween, pH 7.4 (PBST); phosphate buffered saline, pH 7.4 (PBS); methanol –sodium dihydrogen phosphate (1 + 3), pH 5.5; carbonate buffer, pH 9.6; and diethanolamine buffer, pH 9.8 were prepared as previously described.5 5-Bromo-4-chloro-3-indolyl phosphate (BCIP) stock solution was 50 mg ml21 BCIP in 100% dimethyl formamide (DMF).Nitroblue tetrazolium (NBT) stock solution was 75 mg ml21 NBT in 70% DMF.Substrate solution for alkaline phosphatase (AP) was prepared from a mixture of BCIP stock, NBT stock and carbonate buffer with a volume ratio of 1 : 1 : 100. Equipment Cellulose nitrate (CN) membrane filters, diameter 25 mm, pore size 0.1 mm, were from Whatman, Maidstone, Kent, UK. Cellulose nitrate membrane filter images were captured by an EPSON GT8000 scanner, EPSON UK, Hemel Hempstead, Hertfordshire, UK, and Micrografix Picture Publisher V3.1 picture processing software, Micrografix, Woking, Surrey, UK.The intensities of colour on CN membrane filters were measured by Band Leader Application V2.01 shareware software, produced by Maayan, Techknowledge, Israel. Institute of Occupational Medicine (IOM) personal inhalable dust samplers for the membrane filters were purchased from SKC, Blandford, Dorset, UK. Preparation of the Single Antibody Reagent The single antibody reagent was prepared from a mixture of equal volumes of affinity purified 1:1000-diluted rabbit anticeftazidime serum5 in PBST and 1 : 175-diluted goat anti-rabbit IgG conjugated with AP in PBST.The mixture was preincubated for over 1 h at room temperature prior to use.2 This reagent was then stable for over 7 d at 4 °C. Preparation of Coated CN Membrane Filters Cellulose nitrate membrane filters were placed in IOM samplers and treated as follows. The filters were coated with 1.5 ml of 1, † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997.Analyst, December 1997, Vol. 122 (1505–1508) 150510 or 60 mg ml21 ceftazidime–BSA conjugates in coating buffer as single spots at its centre. After incubation for 10 min at room temperature the now dry spots on the filters were washed 3 times with 3 ml of PBST under vacuum which was supplied by a pump to suck PBST through filters from top to bottom. The filters were allowed to air dry for about 10 min when aliquots of 25 ml of PBST containing different amounts of ceftazidime (0.1 ng to 100 mg) were added as single drops on the above treated filters adjacent to the central spots.The filters stood at room temperature until air dried (about 10 min). They were then stored desiccated in plastic envelopes and sealed in aluminium foil until used in the assay. Assay Optimisation Aliquots (300 ml) of single antibody–enzyme reagent were added dropwise to cover the whole surface of each filter.The filters were then incubated for 2–10 min at room temperature with 200 motions min21 shaking (Janke and Kunkel HS 250 shaker from S.H. Scientific, Blyth, Northumberland, UK). After incubation with this single reagent, the filters were washed with 3 ml of distilled water 3 times under vacuum. To the now dry filters were added 10 ml of substrate solution as a drop to the centre of each filter. After incubation for 5–13 min at room temperature, the vacuum was re-applied for 1 min to remove excess substrate and to terminate colour development.The colour intensity of the central spots on the filters was then ascertained visually or by use of an EPSON scanner using Band Leader software. Effect of Air Flow on the Coated Filters The CN filters coated with drug–protein conjugate and ceftazidime standards (0 and 100 ng) and filters not coated with drug–protein conjugate were placed in IOM samplers which were then subjected to a vacuum to induce an air flow through the filters of clean laboratory air for 20 min at a rate of 1.5 l min21 (equivalent to exposure of the filters to a total volume of 30 l of air).The filters were then used in the standard ELISA together with identical coated and uncoated filters which had not been subjected to this treatment and the intensities of the resulting spots determined. Precision of the Assay and its Limit of Detection (LOD) Replicates of coated filters (n = 6) spiked with 0, 10 ng, 100 ng and 1 mg of ceftazidime were processed in the ELISA to determine the RSD at each spiking concentration. The standard deviation of the zero filter was used to calculate the assay’s LOD at 95% confidence from the resulting standard curve.Specificity of the Assay Pre-coated filters were spiked with different antibiotics of the penicillin, cephalosporin and non-b-lactam types, with no drug and 1 mg of drug per filter (n = 3 per standard).Each batch of filters was subjected to the assay and the intensities of the resulting spots determined. The cross-reaction of the drugs was calculated as a percentage of the signal reduction seen with the ceftazidime spiked filters. Stability of the Coated Filters on Storage Thirty-six filters pre-coated with the drug–protein conjugate were stored as three batches: at 4 °C, at room temperature in a desiccator, or at 37 °C. In each case the filters were stored as described above.The filters were removed at 7 and 28 d and spiked with drug standards (zero and 100 ng per filter). They were then subjected to the ELISA and the intensities of the resulting spots compared with those of corresponding freshly prepared and spiked filters. Results Optimisation of the Assay The optimal conditions and protocol chosen for the assay were as follows: (i) coating concentration of 1.5 ml of a 60 mg ml21 solution of drug–BSA conjugate; (ii) single antibody reagent, 300 ml; (iii) initial incubation time of 10 min; (iv) wash conditions under vacuum, 1 min with 3 3 3 ml of distilled water; (v) substrate addition, 10 ml applied to the centre of the filter; (vi) substrate incubation time, 5 min followed by application of vacuum for 1 min; and (vii) assess colour intensity of central spots on the sample filters against those of the zero and 1 mg standards (and additional standards if required) either by eye with the filters within the samplers or by use of an optical scanner following removal of the filters from their sample holders.A typical standard curve obtained following scanning is shown in Fig. 1. The plot is linear between 10 ng and 1 mg (r2 = 0.9996, n = 3, p < 0.001). Effect of Air Flow on the Coated Filters and Filter Stability Significant differences in intensities of the spots were observed for drug-spiked and unspiked filters exposed to an air flow of 1.5 l min21 for 20 min and identically treated filters which were not exposed to this treatment (p < 0001, Student’s t-test).The percentage fall in intensity of the unspiked (zero) filter was 25% of the signal of the corresponding zero filter not exposed to passage of 30 l of air through the filter over 20 min but a satisfactory standard curve was obtained under these conditions. Reduction of the sampling time to 5 min now gave no significant differences in these intensities. Limit of Detection and Precision of the Assay The assay’s LOD was 1.9 ng per filter at 95% confidence limits following scanning.Visual observation of the intensities of spots for the standards enabled discrimination of the zero filter and the filter spiked with the 10 ng standard. The RSD values at Fig. 1 Standard curve for the on-filter ELISA for ceftazidime following scanning of the filters using an optical scanner. Signals are shown as means ± 1 s. 1506 Analyst, December 1997, Vol. 12210 ng, 100 ng and 1 mg per filter (n = 3 per concentration) were 3.7, 2.2 and 1.9%, respectively.Specificity of the Assay The relative signal differences observed for the drugs examined compared with those seen between the 0 and 1 mg standards for ceftazidime are shown in Table 1. Cross reaction was only observed with antibiotics possessing an intact b-lactam substructure. Thus, certain penicillins and cephalosporins having similar side chains to ceftazidime can be detected in the assay although with less sensitivity than for ceftazidime, while structurally dissimilar antibiotics and non-antibiotic drugs such as chloramphenicol, ondansetron and dexamethasone do not interfere in the assay.Stability of the Coated Filters on Storage The signal differences following ELISA development between filters stored at 4 °C, room temperature and at 37 °C for 1 and 4 weeks are shown in Fig. 2. Signal differences between the zero and 100 ng standards after storage are compared as a percentage of the signal drop observed for these standards with filters which were freshly coated with the drug–protein conjugate.The coated filters are stable at ambient temperature and at 4 °C for 4 weeks since the signal differences observed for these conditions are 89% and 66%, respectively, of the signal drop for corresponding standards on the freshly coated filters. This compares with a corresponding value of only 37% for the filters stored at 37 °C for this period.Discussion The conditions for a direct filter-based ELISA for determining the drug ceftazidime captured during air sampling have been determined with respect to the choice of filter (CN was found superior to commercial Teflon and PTFE membrane filters, data not shown; glass wool filters were not considered due to the known instability of the drug on this type of filter3), concentration of coating conjugate on the filter, constitution of the antibody reagents, and incubation times.This has resulted in an assay that can be performed within 17 min without the need to remove the filters from the personal monitors used in the air sampling stage. The end point is a stable blue spot, the colour of which indicates the concentration of drug captured on the filter. This intensity can be observed by eye; the presence of 10 ng or more of drug can be discriminated from the intensity of the spot corresponding to the absence of the drug on a control filter.Alternatively the intensities can be determined using an optical scanner when the limit of detection is 1.9 ng per filter at 95% confidence limits. In this format the standard curve is linear over the range 10 ng to 1 mg. The assay is precise with RSD below 4% over the linear part of the standard curve. The performance of the assay was also unaffected by the passage of 10 l of air through the filters at a rate corresponding to that normally used for personal air monitoring in industry, namely about 2 l min21.This sampling takes 5 min. Although this is less than the 30 l normally sampled over 15 min, the sensitivity of the assay is sufficient to accommodate this smaller sample volume. If the air contained the drug at its OEL of 100 mg m23, a total mass of 3 mg of drug would be captured on the filter at the end of a 15 min sampling period. This mass is outside the linear range of the assay but the reduced sampling time of 5 min would capture 1 mg of drug on the filter which is within this range.The on-filter ELISA produces a spot which could be easily distinguished by eye from the corresponding spots on a control filter without drug and a standard filter spiked with 1 mg of drug. Alternatively the assay’s sensitivity can be reduced by increasing the amount of drug–protein conjugate immobilised on the filter (details not shown) thereby enabling a 15 min sampling period to be followed. Another format is also under investigation in which a number of spots each containing a different amount of drug–protein conjugate are immobilised on each filter.This should produce a simple test to determine the mass of captured drug over a wider concentration range than is observed in the assay described in this paper and thus enabling a range of differing sampling times to be used according to the needs of the user. The spots thus obtained on the filters are stable for many months. The hapten-coated filters are also reasonably stable with 11% signal loss following storage within a desiccator for 4 weeks at ambient temperature.Storage at 37 °C for 4 weeks resulted in signal loss by 73%. The filters can be processed within their sampling devices and the whole process of reagent addition, sample incubation and shaking and filtration could easily be automated leading to a free-standing assay for batch analysis of many samples. We currently process a batch of 12 inter-linked samplers attached to a vacuum system within a shaker for one assay run but use manual addition of reagents.Table 1 Specificity of the assay tested at a concentration of 1 mg (free base) per filter for each drug Difference Cross in signal reactivity intensity Structural similarity relative to (1 mg– ceftazidime Drug zero drug) R1 R2 b-Lactam† (%) Ceftazidime 44 *** *** Yes 100 Cefotaxime 15 * X‡ Yes 36 Cephaloridine 0§ X *** Yes 0 Cephalexin 22 X X Yes 49 Cefuroxime 0§ X X Yes 0 Cefixime 25 ** X Yes 56 Penicillin G 21 X X Yes 47 Ampicillin 8 X X Yes 19 Chloramphenicol 0§ na¶ na No 0 Ondansetron 0§ na na No 0 Dexamethasone 0§ na na No 0 † See Fig. 3 for the structures of the cephalosporins used in this study. * Similar substituent to ceftazidime. ** Very similar substituent to ceftazidime. *** Identical substituent to ceftazidime. ‡ X; Structurally dissimilar substituent to ceftazidime. § No significant difference between the signal at 1 mg of drug and the signal in the absence of drug.¶ na; Not applicable since not a b-lactam antibiotic. Fig. 2 Stability profiles of filters precoated with drug–protein conjugates and stored under various conditions for 1 and 4 weeks. The results are expressed as the percentage of the ELISA signal difference between coated filters spiked with 100 ng and no drug (zero) for freshly coated compared with stored filters. Analyst, December 1997, Vol. 122 1507N S CONH O COOH R2 H H R1 Cephalosporins N S NH2 N O C(CH3)2COOH R1 N+ CH2 R2 Ceftazidime S CH2 N+ CH2 Cephaloridine N S NH2 N OCH2COOH CH CH2 Cefixime N S NH2 N OCH3 CH2OCOCH3 Cefotaxime Cefuroxime Sodium O O NOCH3 CH2OOCNH2 CH NH2 Cephalexin CH3 From the results of the specificity study, the system appears to be applicable to the detection of examples of both penicillins and cephalosporins. The antiserum thus appears to recognise drugs possessing intact b-lactam rings carrying side chains similar to ceftazidime through the amide bridge (R1 in Fig. 3) (Table 1) since cefixime exhibits the greatest cross reaction (56%). The antiserum appears to accommodate relatively simple R1 phenyl-methylene side chains associated with cephalexin, ampicillin and penicillin G whilst being more selective to cephalosporins which have dissimilar R2 substituents to ceftazidime. Thus cephaloridine which has an identical R2 substituent to ceftazidime but a dissimilar R1 side chain exhibited no significant cross reaction in the assay. The greater cross reaction for cephalexin (49%) compared with ampicillin (19%) represents discrimination of the two ring structures since the former is a cephalosporin and the latter a penicillin although both have identical R1 side chains. This partial generic specificity renders the assay of some value for these classes of compounds for which workplace atmospheric monitoring is required. Further studies are required to assess the sensitivity, precision and robustness of the system with additional examples of these compounds. The system will now be further validated under actual workplace conditions and the results obtained correlated with those obtained for samples by HPLC. The on-filter ELISA described herein should be applicable to other airborne biochemicals for which such workplace monitoring is required and for which antibodies can be produced. The method could also find application for general environmental air monitoring outside the workplace and for screening for concealed illicit substances and explosives. References 1 Ceftazidime Product Information. GlaxoWellcome Operations, Barnard Castle, Co. Durham, UK, 1992. 2 Cumming, R., Farrell, C., Nitescu, I., Rowell, F. J., and Tang, L. X., Anal. Chim. Acta, 1995, 311, 377. 3 Rowell, F. J., Farrell, C., Nitescu, I., Cumming, R. H., and Stewart, I. W., J. Aerosol Sci., 1997, 28, 493. 4 Leeder, J. S., Spino, M., Tesoro, A. M., and MacLeod, S. M., Antimicrob. Agents Chemother., 1983, 24, 720. 5 Farrell, C., Rowell, F. J., D’Silva, C., and Cumming, R. H., Analyst, 1994, 119, 2411. Paper 7/04728B Received July 4, 1997 Accepted August 20, 1997 Fig. 3 Structures of the cephalosporins used in the study. 1508 Analyst, December 1997, Vol. 122

ISSN:0003-2654    

DOI:10.1039/a704728b

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Filter holder Air pump 5 m tube Dry-gas meter OUT IN Application of Chemometrics to the Identification of Trends in Polynuclear Aromatic Hydrocarbons and Alkanes in Air Samples From Oporto† Teresa A. P. Rocha and Armando C. Duarte Department of Chemistry, University of Aveiro, 3810 Aveiro, Portugal A sampling system for collecting airborne particulate matter was installed at Oporto Town Hall. Sampling was carried out every seven days and took place from June 1995 to June 1996. A total of 56 samples were collected and analysed for 16 PAHs and 23 alkanes by GC–MS, and organic and elemental carbon by thermal oxidation and detection of CO2 by infra-red.In addition meteorological data (precipitation, wind speed and direction, maximum and minimum temperature) were also obtained. Both cluster and principal component analysis were used for detecting patterns and trends in the data set. Cluster analysis detected a seasonal effect from the data on polynuclear aromatic hydrocarbons (PAHs).Two clusters were identified: one from June to September 1995 and January to June 1996 and another from September 1995 to January 1996. These clusters could be related to changes in both maximum and minimum temperatures as well as precipitation during these periods of time. Principal component analysis identified two factors: the first factor (36.4% variance) contained nine PAHs characteristic of urban areas while the second factor (22.7% variance) contained four alkanes also characteristic of biological contribution in urban areas.Keywords: Polynuclear aromatic hydrocarbons; alkanes; principal component analysis; cluster analysis Both natural (combustion and biosynthesis) and anthropogenic (mobile and stationary categories) sources give rise to polynuclear aromatic hydrocarbon (PAH) formation by means of incomplete combustion of organic materials containing carbon and hydrogen. The incomplete combustion of these organic materials in urban areas, besides causing an environment prone to mutagenic and carcinogenic activity, promotes the soiling of building surfaces through deposition of airborne particulate matter.In the urban environment, anthropogenic activities have a fundamental role in stone soiling and carbonaceous fuels may be the principal cause for the increase in soiling rates. Among a variety of factors, the composition of airborne particulate matter must play an important role in the soiling process.The identification of the origin of air samples through the analysis of their PAH components has been studied by several authors (Khalili et al.,1 Dyremark et al.,2 Freeman and Cattel,3 Simoneit,4 Smith and Harrison5). Khalili et al.1 found the following six PAHs with highest concentrations in traffic samples: naphthalene, acenaphthylene, fluorene, phenanthrene, pyrene and acenaphthene while the predominant PAHs in diesel emissions were fluoranthene, naphthalene, acenaphthylene, phenanthrene and anthracene.Smith and Harrison5 reported the presence of phenanthrene, fluoranthene and pyrene as diesel markers. In gasoline engine samples Khalili et al.1 found naphthalene, fluorene, benzo[e]pyrene, acenaphthylene, pyrene and acenaphthene. Smith and Harrison5 reported the presence of fluorene and indeno[1,2,3-cd]pyrene as a result of combustion of lubricant oil and benzo[ghi]perylene as a petrol marker. Khalili et al.1 found naphthalene, acenaphthylene, phenanthrene, fluorene, anthracene and fluorene in domestic oven emissions and found acenaphthylene, naphthalene, anthracene, phenanthrene, benzo[a]pyrene and benzo[e]pyrene in wood smoke emission samples.Freeman and Cattel3 reported the presence of anthracene, phenanthrene, dibenzo[a,h]anthracene, fluoranthene, benzo[ghi]fluoranthene, benzo[a]pyrene, benzo[ e]pyrene and cyclopenta[cd]pyrene in wood combustion emissions. Simoneit4 found fluoranthene, pyrene, benzanthracene and benzofluoranthene and explained that the data were comparable to similar data in environments with forest fire smoke.Five PAHs (phenanthrene, anthracene, fluoranthene, pyrene and chrysene/triphenylene) have been considered by Dyremark et al.2 as good tracers in emissions from charcoal smoke without meat smoke. For the alkanes Hildemann et al.6 suggested six possible sources: (a) green vegetative detritus, where odd numbered carbon predominate with a maximum for C31; (b) fried hamburger emissions where a predominance of odd number carbon with a maximum for C21 also occurs; (c) catalystequipped automobile emissions, with a maximum for C25 but no predominance for either odd or even numbered carbons; (d) cigarette smoke, where odd numbered carbons predominate in the range C27–C34 and reaching a maximum for C31; (e) fire place emissions of pine wood with a homogeneous distribution and a maximum for C29; (f) urban road dust where an odd number of carbons predominate in the range C25–C34 with a maximum for C29.Simoneit4 reported that petroleum residues present in aerosols are comprised of resolved normal alkanes, with no carbon number predominance and usually range from C15 to C26. Broddin et al.7 report that hydrocarbons of biological origin show a pronounced predominance of odd carbon numbers in the range C27–C31. The more data obtained, the more complex the data structure becomes and therefore an immediate and efficient interpretation of the data is often very difficult.In order to overcome some of † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997. Fig. 1 Sampling equipment for hydrocarbon monitoring. Analyst, December 1997, Vol. 122 (1509–1515) 1509the difficulties, chemometrics, mainly multivariate analysis, has been widely used in analytical chemistry to extract maximum relevant chemical information from chemical data.8,9 The aim of this study, and considering that Oporto is the second largest town in Portugal with a population of about 1.2 million inhabitants, was to speciate some of the organic compounds present in an urban aerosol and to check the origin of such compounds through the use of multivariate analysis based on techniques such as cluster and principal component analysis (PCA).Experimental Sampling Airborne particulate matter was collected with low volume pumps operating at 0.2 m3 h21 and running in parallel with either a Charles Magnetron air pump, Model Dynax 2, or a large pump fitted with a flow limiting orifice.Whatman (Clifton, NJ, USA) QM-A Quartz, 47 mm filters (Catalogue No. 1851037) were used in the low volume sampler to take into account not only the speciation of organic compounds but also the Fig. 2 Variation of total PAH concentration at Oporto from June 1995 to June 1996. Fig. 3 Variation of total alkane concentration at Oporto from June 1995 to June 1996. 1510 Analyst, December 1997, Vol. 122determination of total organic carbon in the samples. The filters were previously conditioned at 500 °C for 8 h. Filters were weighed before and after sampling in order to assess the amount of aerosol gravimetrically. Two systems of filter holders were tested: one that was commercially available and another that was home made from Teflon. The home made system, which is shown as part of the sampling arrangement in Fig. 1, has a major advantage over the commercially available system: after sampling the filter could be removed without losing part of the filter rim in the process of unscrewing the support. The sampling system was installed at Oporto Town Hall at a height of about 25 m. Sampling was carried out every seven days, and filters were removed from the holders, stored in Petri dishes for transport and kept in a refrigerator at 4 °C before analysis. Each sample thus represents an integration of weekly variations.Sampling took place for one year from June 1995 to June 1996. Extraction An ultrasonic bath was preferred for the extraction procedure for removing hydrocarbons from the particulate matter because it allows better extraction yields and is both less time and solvent consuming than Soxhlet extraction. The extraction was performed with a mixture of toluene (Riedel-de Haen, No. 32249 Hannover, Germany) and methanol (Riedel-de Haen, No. 32213). The filter was placed in a 100 3 1023 dm3 round bottomed flask and 25 31023 dm3 of toluene were added to the flask and left in a ultrasonic bath for 1 h at 50 °C.The extract was removed and another extraction was performed with 25 3 1023 dm3 of methanol under the same conditions. After Fig. 4 Observed values for maximum and minimum temperatures, precipitation and total PAH concentration at Oporto. Analyst, December 1997, Vol. 122 1511removing the methanol the process was repeated once again.The toluene and methanol extracts were mixed, filtered through QM-A quartz filters (previously conditioned at 500 °C for 8 h), evaporated to dryness in a rotary evaporator at 35 °C and redissolved in dichloromethane (Riedel-de Haen, No. 32222). The resulting solution was transferred into a small glass sample tube. The sample tube was then allowed to stand in the fume cupboard to allow the air flow to produce a final dry extract, and stored in darkness for further analysis.Analysis Both the PAHs and alkanes were measured using a GC–MS Shimadzu (Kyoto, Japan) Model QP 1100Ex operated in selected ion monitoring (SIM) mode. Separation was attained in both cases using a 0.25 mm phenylmethylsilicone OV5-column, 30 m 30.32 mm (Catalogue No. 530-3202). The concentrations of PAHs were determined by GC–MS and the concentration of individual compounds determined by comparison with an Environmental Protection Agency (EPA) standard solution (EPA 610 Polynuclear aromatic hydrocarbons mix in methanol –methylenechloride (1 + 1), supplied by Supelco, Bellefonte, PA, USA, Catalogue No. 4-8743). The samples for PAH analysis were dissolved in 250 3 1026 dm3 of dichloromethane (DCM) and a 60 3 1024 dm3 aliquot was injected into the splitless injector at 280 °C. The oven temperature was 60 °C and increased 6 °C min21 to 282 °C. The concentrations of alkanes were determined by comparison with the following standard solutions: saturated straight chain hydrocarbons (Sigma, St.Louis, MO, USA, Lot 55H 9005) containing the low to intermediate molecular mass hydrocarbons (n-decane to n-docosane) and saturated straight chain hydrocarbons (Sigma, Lot 016H 9058) containing the high molecular mass hydrocarbons (n-tricosane to n-tetratriacontane with exception of n-hentriacontane and n-tritriacontane). The samples for alkane determination were dissolved in 250 3 1026 dm3 DCM and a 60 3 1024 dm3 aliquot was injected into the splitless injector at 270 °C.The oven temperature was held at 70 °C for 1 min, then increased at 15 °C min21 to 295 °C, and held again for 9 min. The carrier gas used for both PAH and alkane determination was Helium (N55 Arliquido). The organic carbon and elemental carbon were measured with a thermal-optical carbon analyser. The analyser comprises a quartz tube with two heating zones, a laser, a detector for the laser beam, a non-dispersive infra-red spectrophotometer for CO2 detection, a temperature controller and a computer for continuous data acquisition.Results and Discussion Figs. 2 and 3 show the total concentration of PAHs and alkanes, respectively. Total PAH concentration (expressed as the sum of the 16 measured compounds) varied between 1 and 75 ng m23. This range is lower than the range of concentrations found in four UK urban areas reported by Halsall et al.10 throughout 1991 where the range varied from 27–377 ng m23 for Stevenage, 11–555 ng m23 for Cardiff, 70–288 ng m23 for Manchester and 35–735 ng m23 for London.Our data range is more similar to the value found in Ghent (residential) of 11–91 ng m23 and reported by Broddin et al.7 Total alkane concentrations (expressed as the sum of the 23 measured compounds) varied between 113 and 595 ng m23. This range is three times higher than the range of concentrations (41–201 ng m23) reported by Broddin et al.7 also for Ghent (residential).Compared to Ghent data, as an example of a residential area, in Oporto there are higher levels of alkanes but, on the other hand, unexpected lower levels of PAHs were observed for the sampling period. Cluster analysis by k-means splitting method based on nearest centroid sorting, using SPSS,11 detected a seasonal effect from the data on PAHs. Two clusters were identified: one from June to September 1995 and January to June 1996 and another cluster from September 1995 to January 1996 as shown in Fig. 2. These clusters could be somewhat related to changes in both maximum and minimum temperatures and also in precipitation during these periods of time (Fig. 4). Both maximum and minimum temperatures started decreasing in September 1995, until about January 1996 when the temperature increased to a maximum in June 1996. The most intense episodes of Fig. 5 Total carbon and elemental carbon at Oporto. 1512 Analyst, December 1997, Vol. 122precipitation occur also from September 1995 to January 1996, coinciding with the trend of decreasing temperatures and with maximum PAH concentrations.This finding agrees with the reports from Kamens et al.12 who observed that higher humidity can be associated with higher concentration of PAHs. The total PAH concentrations in September 1995 to January 1996 were in general much higher than the total PAH concentration corresponding to the second cluster, that is from June to September 1995 and January to June 1996.This coincides with a temperature decrease, since in the first cluster both maximum and minimum temperatures were much lower than in the second cluster. Such a finding also agrees well with Smith and Harrison5 who explained that the PAH concentrations may vary due to meteorological conditions: in winter they are less favourable for dispersion while greater rates of loss occur in summer due to physico-chemical and meteorological factors, which may significantly affect the atmospheric degradation of some of the reactive PAHs.Baek et al.13 also demonstrated that losses of PAHs may occur by photochemical and/or chemical reactions on the filter-deposited particles during the sampling period with gaseous air pollutants under varying meteorological conditions. The data on PAH concentrations of the second cluster from June to September 1995 show episodes of high concentrations which could be attributed to the forest fires that occur every summer in the northern region of Portugal.For data on total and elemental carbon (Fig. 5) as well as on the alkanes (Fig. 3) no significant cluster division was identified. Fig. 6 Distribution of mean concentration for individual PAHs. Fig. 7 Distribution of mean concentration for individual alkanes. Analyst, December 1997, Vol. 122 1513Fig. 6 shows the distribution of average concentration for each PAH and it compares the values obtained in winter with those obtained in summer.As expected an obvious difference is observed in the average concentration between summer and winter: in winter the concentration of individual PAHs are usually higher than in summer. Fig. 7 shows the distribution of average concentration for each alkane, and an alternating pattern is observed with higher concentrations of odd carbon number for alkanes from C27 to C30. This observation may be attributed to natural emissions of biological origin (e.g., vegetation7).The total concentration of the lower alkanes (C10 to C26) do not exhibit an alternating pattern but shows a monotonous increasing trend which attains a maximum at C26. Factor extraction using PCA was applied to all data [alkanes, PAHs, elemental carbon, total carbon, minimum temperature (tmin), maximum temperature (tmax) and precipitation]. Taking into consideration the ingredients of a good factor analysis solution11 such as the utilisation of factors of appropriateness such as the Kayer–Meyer–Olkin measure of sampling adequacy, the Bartlett test of esphericity and communality reduced the data matrix from 46 to 22 variables.Two factors were identified and as shown in Table 1 the first factor (36.4% explained variance) has a strong contribution, that is large Table 1 Unrotated and rotated factor matrices obtained using PCA Unrotated factors Rotated factors Variable Factor 1 Factor 2 Factor 1 Factor 2 Acenaphthene 0.65361 0.37663 0.63385 0.40902 Anthracene 0.82024 0.28252 0.80499* 0.32341 Benzo[a]pyrene 0.57691 20.15792 0.58412 20.12871 Benzo[a]anthracene 0.94446 20.08371 0.9474* 20.03611 Benzo[b]fluoranthene 0.78029 20.18199 0.78845* 20.14252 Benzo[ghi]perylene 0.66468 20.09728 0.66873 20.06374 Benzo[k]fluoranthene 0.73897 20.26726 0.75147* 20.22976 Chrysene 0.93383 20.08800 0.93708* 20.40930 Dibenzo[a,h]anthracene 0.46090 20.23414 0.47209 20.21067 Dotriacontane 20.14419 0.73370 20.18091 0.72552* Fluoranthene 0.87642 0.19101 0.86571* 0.23484 Fluorene 0.85737 0.18358 0.84705* 0.22646 Heneicosane 0.31268 0.45023 0.28964 0.46538 Heptacosane 0.14455 0.66670 0.11084 0.67313 Hexacosane 20.08614 0.85438 20.12899 0.84897* Indeno[1,2,3-cd]pyrene 0.75491 20.16799 0.76240* 20.12981 Nonacosane 20.0710 0.87467 20.15094 0.86818* Nonadecane 0.18408 0.50770 0.15832 0.51631 Octacosane 20.17921 0.90403 20.22444 0.89387* Phenanthrene 0.76922 0.20916 0.75772* 0.24758 Tetratriacontane 0.28065 0.49340 0.25549 0.50689 Triacontane 20.16476 0.64565 0.19702 0.63655 * Considered a strong contribution (!0.70).Fig. 8 Air samples ploted in a plane defined by the first two PCs (+ for cluster 1 and D for cluster 2). 1514 Analyst, December 1997, Vol. 122loadings, from anthracene, benzo[a]anthracene, benzo[b]fluoranthene, benzo[k]fluoranthene, chrysene, fluoranthene, fluorene, indeno [1,2,3-cd]pyrene and phenanthrene while the second factor (22.7% variance explained) has a strong contribution from dotriacontane, hexacosane, octacosane, and nonacosane.Table 1 contains both the unrotated as well as the rotated factors obtained by Equamax rotation (Kaiser normalisation). 11 The rotation did not bring any substantial modification to the contribution of the variables identified as important to each factor. It is important to observe that the first factor is only constituted by PAHs and the second factor is only constituted by alkanes. This can be explained by the fact that these PAHs and alkanes may be from different origins and comparing our results with those mentioned earlier we suggest that the alkanes may be from biological origin (including forest fires) and/or urban road dust (presence of pulverised vegetation as a component of the resuspended fine aerosol) while the PAHs may be from anthropogenic origin (fossil fuel combustion, diesel, petrol).Fig. 8 shows the air samples plotted in a plane defined by the first two PCs (59.1% total variance explained) and classified into two clusters as defined in Fig. 4 (+ for cluster 1 and D for cluster 2). PC1 on the x-axis can be mainly attributed to anthropogenic sources (combustion of fossil fuel and lubricant oil and diesel emissions) and PC2 can be related to biological derived sources (green urban vegetation derived detritus, cigarette smoke, pulverised vegetation urban in road dust). This claim can also be supported by the different pattern of wind directions associated with the two clusters of data (see inset in Fig. 8): cluster 1 is dominated by SE winds (urban and industrial areas) while cluster 2 is associated with NW and SE winds (strong contribution from clean winds from the Atlantic Ocean although containing also a component from urban and industrial areas from SE direction). In general terms the observation of the composition of each principal component allows us to conclude that the predominant alkanes and PAHs in Oporto may be characteristic of urban dust and urban area sources (diesel and fossil fuel combustion).This work was partially financed through EC funded project Particulate Pollution and Stone Damage (research contract EV5V/ CT94/0519). References 1 Khalili, N. R., Scheff, P. A., and Holsen, T. M., Atmos. Environ., 1995, 29, 533. 2 Dyremark, A., Westerholm, R., Overvik, E., and Gustavsson, J.-A., Atmos. Environ., 1995, 29, 1553. 3 Freeman, D. J., and Cattel, C. R., Environ. Sci. Technol., 1990, 24, 1581. 4 Simoneit, B. R. T., Atmos. Environ., 1984, 18, 51. 5 Smith, D. J. T., and Harrison, R. M., Atmos. Environ., 1996, 30, 2513. 6 Hildemann, L. M., Rogge, W. F., Cass, G. R., Mazurek, M. A., and Simoneit, B. R., J. Geophys. Res., 1996, 101, 19 541. 7 Broddin, G., Cautreels, W., and Cauwenberghe, K. V., Atmos. Environ., 1980, 14, 895. 8 Thielemans, A., and Massart, D. L., Separatdruck Chim., 1985, 39, 236. 9 Hopke, P. K., and Natusch, D. F. S., Analytical Aspects of Environmental Chemistry, ed.Wiley, New York, 1983, ch. 6. 10 Halsall, C., Burnett, V., Davis, B., Jones, P., Pettit, C., and Jones, K. C., Chemosphere, 1993, 26, 2185. 11 Norusis, M. J., SPSS Professional Statistics 6.1., ed SPSS Inc., Chicago, 1994. 12 Kamens, R. M., Fulcher, J. N., and Zhishi, G., Atmos. Environ., 1986, 20, 1579. 13 Baek, S. O., Goldstone, M. E., Kirk, P. W., Lester, J. N., and Perry, R., Environ. Technol., 1991, 12, 107. Paper 7/04732K Received July 4, 1997 Accepted September 12, 1997 Analyst, December 1997, Vol. 122 1515 Filter holder Air pump 5 m tube Dry-gas meter OUT IN Application of Chemometrics to the Identification of Trends in Polynuclear Aromatic Hydrocarbons and Alkanes in Air Samples From Oporto† Teresa A. P. Rocha and Armando C. Duarte Department of Chemistry, University of Aveiro, 3810 Aveiro, Portugal A sampling system for collecting airborne particulate matter was installed at Oporto Town Hall. Sampling was carried out every seven days and took place from June 1995 to June 1996. A total of 56 samples were collected and analysed for 16 PAHs and 23 alkanes by GC–MS, and organic and elemental carbon by thermal oxidation and detection of CO2 by infra-red.In addition meteorological data (precipitation, wind speed and direction, maximum and minimum temperature) were also obtained. Both cluster and principal component analysis were used for detecting patterns and trends in the data set.Cluster analysis detected a seasonal effect from the data on polynuclear aromatic hydrocarbons (PAHs). Two clusters were identified: one from June to September 1995 and January to June 1996 and another from September 1995 to January 1996. These clusters could be related to changes in both maximum and minimum temperatures as well as precipitation during these periods of time. Principal component analysis identified two factors: the first factor (36.4% variance) contained nine PAHs characteristic of urban areas while the second factor (22.7% variance) contained four alkanes also characteristic of biological contribution in urban areas.Keywords: Polynuclear aromatic hydrocarbons; alkanes; principal component analysis; cluster analysis Both natural (combustion and biosynthesis) and anthropogenic (mobile and stationary categories) sources give rise to polynuclear aromatic hydrocarbon (PAH) formation by means of incomplete combustion of organic materials containing carbon and hydrogen. The incomplete combustion of these organic materials in urban areas, besides causing an environment prone to mutagenic and carcinogenic activity, promotes the soiling of building surfaces through deposition of airborne particulate matter.In the urban environment, anthropogenic activities have a fundamental role in stone soiling and carbonaceous fuels may be the principal cause for the increase in soiling rates. Among a variety of factors, the composition of airborne particulate matter must play an important role in the soiling process. The identification of the origin of air samples through the analysis of their PAH components has been studied by several authors (Khalili et al.,1 Dyremark et al.,2 Freeman and Cattel,3 Simoneit,4 Smith and Harrison5).Khalili et al.1 found the following six PAHs with highest concentrations in traffic samples: naphthalene, acenaphthylene, fluorene, phenanthrene, pyrene and acenaphthene while the predominant PAHs in diesel emissions were fluoranthene, naphthalene, acenaphthylene, phenanthrene and anthracene.Smith and Harrison5 reported the presence of phenanthrene, fluoranthene and pyrene as diesel markers. In gasoline engine samples Khalili et al.1 found naphthalene, fluorene, benzo[e]pyrene, acenaphthylene, pyrene and acenaphthene. Smith and Harrison5 reported the presence of fluorene and indeno[1,2,3-cd]pyrene as a result of combustion of lubricant oil and benzo[ghi]perylene as a petrol marker.Khalili et al.1 found naphthalene, acenaphthylene, phenanthrene, fluorene, anthracene and fluorene in domestic oven emissions and found acenaphthylene, naphthalene, anthracene, phenanthrene, benzo[a]pyrene and benzo[e]pyrene in wood smoke emission samples. Freeman and Cattel3 reported the presence of anthracene, phenanthrene, dibenzo[a,h]anthracene, fluoranthene, benzo[ghi]fluoranthene, benzo[a]pyrene, benzo[ e]pyrene and cyclopenta[cd]pyrene in wood combustion emissions.Simoneit4 found fluoranthene, pyrene, benzanthracene and benzofluoranthene and explained that the data were comparable to similar data in environments with forest fire smoke. Five PAHs (phenanthrene, anthracene, fluoranthene, pyrene and chrysene/triphenylene) have been considered by Dyremark et al.2 as good tracers in emissions from charcoal smoke without meat smoke. For the alkanes Hildemann et al.6 suggested six possible sources: (a) green vegetative detritus, where odd numbered carbon predominate with a maximum for C31; (b) fried hamburger emissions where a predominance of odd number carbon with a maximum for C21 also occurs; (c) catalystequipped automobile emissions, with a maximum for C25 but no predominance for either odd or even numbered carbons; (d) cigarette smoke, where odd numbered carbons predominate in the range C27–C34 and reaching a maximum for C31; (e) fire place emissions of pine wood with a homogeneous distribution and a maximum for C29; (f) urban road dust where an odd number of carbons predominate in the range C25–C34 with a maximum for C29.Simoneit4 reported that petroleum residues present in aerosols are comprised of resolved normal alkanes, with no carbon number predominance and usually range from C15 to C26. Broddin et al.7 report that hydrocarbons of biological origin show a pronounced predominance of odd carbon numbers in the range C27–C31. The more data obtained, the more complex the data structure becomes and therefore an immediate and efficient interpretation of the data is often very difficult.In order to overcome some of † Presented at the Symposium on Analytical Science and the Environment, Newcastle, UK, June 30–July 3, 1997. Fig. 1 Sampling equipment for hydrocarbon monitoring. Analyst, December 1997, Vol. 122 (1509–1515) 1509the difficulties, chemometrics, mainly multivariate analysis, has been widely used in analytical chemistry to extract maximum relevant chemical information from chemical data.8,9 The aim of this study, and considering that Oporto is the second largest town in Portugal with a population of about 1.2 million inhabitants, was to speciate some of the organic compounds present in an urban aerosol and to check the origin of such compounds through the use of multivariate analysis based on techniques such as cluster and principal component analysis (PCA). Experimental Sampling Airborne particulate matter was collected with low volume pumps operating at 0.2 m3 h21 and running in parallel with either a Charles Magnetron air pump, Model Dynax 2, or a large pump fitted with a flow limiting orifice.Whatman (Clifton, NJ, USA) QM-A Quartz, 47 mm filters (Catalogue No. 1851037) were used in the low volume sampler to take into account not only the speciation of organic compounds but also the Fig. 2 Variation of total PAH concentration at Oporto from June 1995 to June 1996.Fig. 3 Variation of total alkane concentration at Oporto from June 1995 to June 1996. 1510 Analyst, December 1997, Vol. 122determination of total organic carbon in the samples. The filters were previously conditioned at 500 °C for 8 h. Filters were weighed before and after sampling in order to assess the amount of aerosol gravimetrically. Two systems of filter holders were tested: one that was commercially available and another that was home made from Teflon.The home made system, which is shown as part of the sampling arrangement in Fig. 1, has a major advantage over the commercially available system: after sampling the filter could be removed without losing part of the filter rim in the process of unscrewing the support. The sampling system was installed at Oporto Town Hall at a height of about 25 m. Sampling was carried out every seven days, and filters were removed from the holders, stored in Petri dishes for transport and kept in a refrigerator at 4 °C before analysis.Each sample thus represents an integration of weekly variations. Sampling took place for one year from June 1995 to June 1996. Extraction An ultrasonic bath was preferred for the extraction procedure for removing hydrocarbons from the particulate matter because it allows better extraction yields and is both less time and solvent consuming than Soxhlet extraction. The extraction was performed with a mixture of toluene (Riedel-de Haen, No. 32249 Hannover, Germany) and methanol (Riedel-de Haen, No. 32213). The filter was placed in a 100 3 1023 dm3 round bottomed flask and 25 31023 dm3 of toluene were added to the flask and left in a ultrasonic bath for 1 h at 50 °C. The extract was removed and another extraction was performed with 25 3 1023 dm3 of methanol under the same conditions. After Fig. 4 Observed values for maximum and minimum temperatures, precipitation and total PAH concentration at Oporto.Analyst, December 1997, Vol. 122 1511removing the methanol the process was repeated once again. The toluene and methanol extracts were mixed, filtered through QM-A quartz filters (previously conditioned at 500 °C for 8 h), evaporated to dryness in a rotary evaporator at 35 °C and redissolved in dichloromethane (Riedel-de Haen, No. 32222). The resulting solution was transferred into a small glass sample tube. The sample tube was then allowed to stand in the fume cupboard to allow the air flow to produce a final dry extract, and stored in darkness for further analysis.Analysis Both the PAHs and alkanes were measured using a GC–MS Shimadzu (Kyoto, Japan) Model QP 1100Ex operated in selected ion monitoring (SIM) mode. Separation was attained in both cases using a 0.25 mm phenylmethylsilicone OV5-column, 30 m 30.32 mm (Catalogue No. 530-3202). The concentrations of PAHs were determined by GC–MS and the concentration of individual compounds determined by comparison with an Environmental Protection Agency (EPA) standard solution (EPA 610 Polynuclear aromatic hydrocarbons mix in methanol –methylenechloride (1 + 1), supplied by Supelco, Bellefonte, PA, USA, Catalogue No. 4-8743). The samples for PAH analysis were dissolved in 250 3 1026 dm3 of dichloromethane (DCM) and a 60 3 1024 dm3 aliquot was injected into the splitless injector at 280 °C. The oven temperature was 60 °C and increased 6 °C min21 to 282 °C. The concentrations of alkanes were determined by comparison with the following standard solutions: saturated straight chain hydrocarbons (Sigma, St.Louis, MO, USA, Lot 55H 9005) containing the low to intermediate molecular mass hydrocarbons (n-decane to n-docosane) and saturated straight chain hydrocarbons (Sigma, Lot 016H 9058) containing the high molecular mass hydrocarbons (n-tricosane to n-tetratriacontane with exception of n-hentriacontane and n-tritriacontane). The samples for alkane determination were dissolved in 250 3 1026 dm3 DCM and a 60 3 1024 dm3 aliquot was injected into the splitless injector at 270 °C.The oven temperature was held at 70 °C for 1 min, then increased at 15 °C min21 to 295 °C, and held again for 9 min. The carrier gas used for both PAH and alkane determination was Helium (N55 Arliquido). The organic carbon and elemental carbon were measured with a thermal-optical carbon analyser. The analyser comprises a quartz tube with two heating zones, a laser, a detector for the laser beam, a non-dispersive infra-red spectrophotometer for CO2 detection, a temperature controller and a computer for continuous data acquisition.Results and Discussion Figs. 2 and 3 show the total concentration of PAHs and alkanes, respectively. Total PAH concentration (expressed as the sum of the 16 measured compounds) varied between 1 and 75 ng m23. This range is lower than the range of concentrations found in four UK urban areas reported by Halsall et al.10 throughout 1991 where the range varied from 27–377 ng m23 for Stevenage, 11–555 ng m23 for Cardiff, 70–288 ng m23 for Manchester and 35–735 ng m23 for London.Our data range is more similar to the value found in Ghent (residential) of 11–91 ng m23 and reported by Broddin et al.7 Total alkane concentrations (expressed as the sum of the 23 measured compounds) varied between 113 and 595 ng m23. This range is three times higher than the range of concentrations (41–201 ng m23) reported by Broddin et al.7 also for Ghent (residential).Compared to Ghent data, as an example of a residential area, in Oporto there are higher levels of alkanes but, on the other hand, unexpected lower levels of PAHs were observed for the sampling period. Cluster analysis by k-means splitting method based on nearest centroid sorting, using SPSS,11 detected a seasonal effect from the data on PAHs. Two clusters were identified: one from June to September 1995 and January to June 1996 and another cluster from September 1995 to January 1996 as shown in Fig. 2. These clusters could be somewhat related to changes in both maximum and minimum temperatures and also in precipitation during these periods of time (Fig. 4). Both maximum and minimum temperatures started decreasing in September 1995, until about January 1996 when the temperature increased to a maximum in June 1996. The most intense episodes of Fig. 5 Total carbon and elemental carbon at Oporto. 1512 Analyst, December 1997, Vol. 122precipitation occur also from September 1995 to January 1996, coinciding with the trend of decreasing temperatures and with maximum PAH concentrations. This finding agrees with the reports from Kamens et al.12 who observed that higher humidity can be associated with higher concentration of PAHs. The total PAH concentrations in September 1995 to January 1996 were in general much higher than the total PAH concentration corresponding to the second cluster, that is from June to September 1995 and January to June 1996.This coincides with a temperature decrease, since in the first cluster both maximum and minimum temperatures were much lower than in the second cluster. Such a finding also agrees well with Smith and Harrison5 who explained that the PAH concentrations may vary due to meteorological conditions: in winter they are less favourable for dispersion while greater rates of loss occur in summer due to physico-chemical and meteorological factors, which may significantly affect the atmospheric degradation of some of the reactive PAHs.Baek et al.13 also demonstrated that losses of PAHs may occur by photochemical and/or chemical reactions on the filter-deposited particles during the sampling period with gaseous air pollutants under varying meteorological conditions. The data on PAH concentrations of the second cluster from June to September 1995 show episodes of high concentrations which could be attributed to the forest fires that occur every summer in the northern region of Portugal.For data on total and elemental carbon (Fig. 5) as well as on the alkanes (Fig. 3) no significant cluster division was identified. Fig. 6 Distribution of mean concentration for individual PAHs. Fig. 7 Distribution of mean concentration for individual alkanes. Analyst, December 1997, Vol. 122 1513Fig. 6 shows the distribution of average concentration for each PAH and it compares the values obtained in winter with those obtained in summer.As expected an obvious difference is observed in the average concentration between summer and winter: in winter the concentration of individual PAHs are usually higher than in summer. Fig. 7 shows the distribution of average concentration for each alkane, and an alternating pattern is observed with higher concentrations of odd carbon number for alkanes from C27 to C30. This observation may be attributed to natural emissions of biological origin (e.g., vegetation7).The total concentration of the lower alkanes (C10 to C26) do not exhibit an alternating pattern but shows a monotonous increasing trend which attains a maximum at C26. Factor extraction using PCA was applied to all data [alkanes, PAHs, elemental carbon, total carbon, minimum temperature (tmin), maximum temperature (tmax) and precipitation]. Taking into consideration the ingredients of a good factor analysis solution11 such as the utilisation of factors of appropriateness such as the Kayer–Meyer–Olkin measure of sampling adequacy, the Bartlett test of esphericity and communality reduced the data matrix from 46 to 22 variables.Two factors were identified and as shown in Table 1 the first factor (36.4% explained variance) has a strong contribution, that is large Table 1 Unrotated and rotated factor matrices obtained using PCA Unrotated factors Rotated factors Variable Factor 1 Factor 2 Factor 1 Factor 2 Acenaphthene 0.65361 0.37663 0.63385 0.40902 Anthracene 0.82024 0.28252 0.80499* 0.32341 Benzo[a]pyrene 0.57691 20.15792 0.58412 20.12871 Benzo[a]anthracene 0.94446 20.08371 0.9474* 20.03611 Benzo[b]fluoranthene 0.78029 20.18199 0.78845* 20.14252 Benzo[ghi]perylene 0.66468 20.09728 0.66873 20.06374 Benzo[k]fluoranthene 0.73897 20.26726 0.75147* 20.22976 Chrysene 0.93383 20.08800 0.93708* 20.40930 Dibenzo[a,h]anthracene 0.46090 20.23414 0.47209 20.21067 Dotriacontane 20.14419 0.73370 20.18091 0.72552* Fluoranthene 0.87642 0.19101 0.86571* 0.23484 Fluorene 0.85737 0.18358 0.84705* 0.22646 Heneicosane 0.31268 0.45023 0.28964 0.46538 Heptacosane 0.14455 0.66670 0.11084 0.67313 Hexacosane 20.08614 0.85438 20.12899 0.84897* Indeno[1,2,3-cd]pyrene 0.75491 20.16799 0.76240* 20.12981 Nonacosane 20.0710 0.87467 20.15094 0.86818* Nonadecane 0.18408 0.50770 0.15832 0.51631 Octacosane 20.17921 0.90403 20.22444 0.89387* Phenanthrene 0.76922 0.20916 0.75772* 0.24758 Tetratriacontane 0.28065 0.49340 0.25549 0.50689 Triacontane 20.16476 0.64565 0.19702 0.63655 * Considered a strong contribution (!0.70).Fig. 8 Air samples ploted in a plane defined by the first two PCs (+ for cluster 1 and D for cluster 2). 1514 Analyst, December 1997, Vol. 122loadings, from anthracene, benzo[a]anthracene, benzo[b]fluoranthene, benzo[k]fluoranthene, chrysene, fluoranthene, fluorene, indeno [1,2,3-cd]pyrene and phenanthrene while the second factor (22.7% variance explained) has a strong contribution from dotriacontane, hexacosane, octacosane, and nonacosane.Table 1 contains both the unrotated as well as the rotated factors obtained by Equamax rotation (Kaiser normalisation). 11 The rotation did not bring any substantial modification to the contribution of the variables identified as important to each factor. It is important to observe that the first factor is only constituted by PAHs and the second factor is only constituted by alkanes.This can be explained by the fact that these PAHs and alkanes may be from different origins and comparing our results with those mentioned earlier we suggest that the alkanes may be from biological origin (including forest fires) and/or urban road dust (presence of pulverised vegetation as a component of the resuspended fine aerosol) while the PAHs may be from anthropogenic origin (fossil fuel combustion, diesel, petrol).Fig. 8 shows the air samples plotted in a plane defined by the first two PCs (59.1% total variance explained) and classified into two clusters as defined in Fig. 4 (+ for cluster 1 and D for cluster 2). PC1 on the x-axis can be mainly attributed to anthropogenic sources (combustion of fossil fuel and lubricant oil and diesel emissions) and PC2 can be related to biological derived sources (green urban vegetation derived detritus, cigarette smoke, pulverised vegetation urban in road dust). This claim can also be supported by the different pattern of wind directions associated with the two clusters of data (see inset in Fig. 8): cluster 1 is dominated by SE winds (urban and industrial areas) while cluster 2 is associated with NW and SE winds (strong contribution from clean winds from the Atlantic Ocean although containing also a component from urban and industrial areas from SE direction). In general terms the observation of the composition of each principal component allows us to conclude that the predominant alkanes and PAHs in Oporto may be characteristic of urban dust and urban area sources (diesel and fossil fuel combustion). This work was partially financed through EC funded project Particulate Pollution and Stone Damage (research contract EV5V/ CT94/0519). References 1 Khalili, N. R., Scheff, P. A., and Holsen, T. M., Atmos. Environ., 1995, 29, 533. 2 Dyremark, A., Westerholm, R., Overvik, E., and Gustavsson, J.-A., Atmos. Environ., 1995, 29, 1553. 3 Freeman, D. J., and Cattel, C. R., Environ. Sci. Technol., 1990, 24, 1581. 4 Simoneit, B. R. T., Atmos. Environ., 1984, 18, 51. 5 Smith, D. J. T., and Harrison, R. M., Atmos. Environ., 1996, 30, 2513. 6 Hildemann, L. M., Rogge, W. F., Cass, G. R., Mazurek, M. A., and Simoneit, B. R., J. Geophys. Res., 1996, 101, 19 541. 7 Broddin, G., Cautreels, W., and Cauwenberghe, K. V., Atmos. Environ., 1980, 14, 895. 8 Thielemans, A., and Massart, D. L., Separatdruck Chim., 1985, 39, 236. 9 Hopke, P. K., and Natusch, D. F. S., Analytical Aspects of Environmental Chemistry, ed. Wiley, New York, 1983, ch. 6. 10 Halsall, C., Burnett, V., Davis, B., Jones, P., Pettit, C., and Jones, K. C., Chemosphere, 1993, 26, 2185. 11 Norusis, M. J., SPSS Professional Statistics 6.1., ed SPSS Inc., Chicago, 1994. 12 Kamens, R. M., Fulcher, J. N., and Zhishi, G., Atmos. Environ., 1986, 20, 1579. 13 Baek, S. O., Goldstone, M. E., Kirk, P. W., Lester, J. N., and Perry, R., Environ. Technol., 1991, 12, 107. Paper 7/04732K Received July 4, 1997 Accepted September 12, 1997 Analyst, December 1997, Vol. 122 1515

ISSN:0003-2654    

DOI:10.1039/a704732k

出版商:RSC    

年代:1997

数据来源: RSC