Optimisation of supercritical fluid extraction of polycyclic aromatic hydrocarbons and their nitrated derivatives adsorbed on highly sorptive diesel particulate matter

Analytica Chimica Acta 651 (2009) 48–56

Contents lists available at ScienceDirect

Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca

Optimisation of supercritical fluid extraction of polycyclic aromatic hydrocarbons and their nitrated derivatives adsorbed on highly sorptive diesel particulate matter F. Portet-Koltalo a,∗ , K. Oukebdane a , F. Dionnet b , P.L. Desbène a a b

Laboratoire d’Analyse des Systèmes Organiques Complexes, IRCOF et IFRMP, Université de Rouen, 55 rue Saint Germain, 27000 Evreux, France CERTAM, Technopole du Madrillet, 1 rue Joseph Fourier, 76800 Saint Etienne du Rouvray, France

a r t i c l e

i n f o

Article history: Received 3 April 2009 Received in revised form 16 July 2009 Accepted 17 July 2009 Available online 22 July 2009 Keywords: Polycyclic aromatic hydrocarbons Nitrated polycyclic aromatic hydrocarbons Highly sorptive diesel particulates Supercritical fluid extraction Factorial design

a b s t r a c t Supercritical fluid extraction (SFE) was performed to extract complex mixtures of polycyclic aromatic hydrocarbons (PAHs), nitrated derivatives (nitroPAHs) and heavy n-alkanes from spiked soot particulates that resulted from the incomplete combustion of diesel oils. This polluted material, resulting from combustion in a light diesel engine and collected at high temperature inside the particulate filter placed just after the engine, was particularly resistant to conventional extraction techniques, such as soxhlet extraction, and had an extraction behaviour that differed markedly from certified reference materials (SRM 1650). A factorial experimental design was performed, simultaneously modelling the influence of four SFE experimental factors on the recovery yields, i.e.: the temperature and the pressure of the supercritical fluid, the nature and the percentage of the organic modifier added to CO2 (chloroform, tetrahydrofuran, methylene chloride), as a means to reach the optimal extraction yields for all the studied target pollutants. The results of modelling showed that the supercritical fluid pressure had to be kept at its maximum level (30 MPa) and the temperature had to be kept relatively low (75 ◦ C). Under these operating conditions, adding 15% of methylene chloride to the CO2 permitted quantitative extraction of not only light PAHs and their nitrated derivatives, but also heavy n-alkanes from the spiked soots. However, heavy polyaromatics were not quantitatively extracted from the refractory carbonaceous solid surface. As such, original organic modifiers were tested, including pyridine, which, as a strong electron donor cosolvent (15% into CO2 ), was the most successful. The addition of diethylamine to pyridine, which enhanced the electron donor character of the cosolvent, even increased the extraction yields of the heaviest PAHs, leading to a quantitative extraction of all PAHs (more than 79%) from the diesel particulate matter, with detection limits ranging from 0.5 to 7.8 ng for 100 mg of spiked material. Concerning the nitrated PAHs, a small addition of acetic acid into pyridine, as cosolvents, gave the best results, leading to fair extraction yields (approximately 60%), with detection limits ranging from 18 to 420 ng. © 2009 Elsevier B.V. All rights reserved.

1. Introduction The emission of particulate matter from diesel engines represents a serious environmental concern in urban areas, and affects the quality of life and human health [1]. Notably, diesel particulate matter, of which the carbonaceous part is called diesel soot, can have a very complex chemical composition. Diesel soot is typically composed of highly agglomerated carbonaceous particles and of a variety of adsorbed hydrocarbon materials (aliphatic, aromatic and oxygenated compounds) that represent the soluble organic fraction (SOF), whereas the insoluble fraction contains sulphates, nitrates, nitrites and metals [2]. Consequently, diesel vehicles can

∗ Corresponding author. Tel.: +33 02 32 29 15 35; fax: +33 02 32 29 15 35. E-mail address: fl[email protected] (F. Portet-Koltalo). 0003-2670/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.aca.2009.07.038

be regarded as one of the more serious anthropogenic sources of polycyclic aromatic hydrocarbons (PAHs), which are produced as a result of the incomplete combustion of fossil fuels [3,4]. Many PAHs have been shown to be mutagenic or carcinogenic. There are also indications that oxygenated PAHs, as well as nitrated PAHs (also known as nitroPAHs), contribute significantly to the mutagenecity of the diesel particulate matter [5–7]. For example, benzo[a]pyrene, which is considered to be one of the most toxic PAHs, has a toxic equivalent factor of 1, which indicates that it has a smaller contribution to the toxicity of a sample of diesel particulate matter than 6-nitrochrysene, which has a toxic equivalent factor of 10. In other words, the dose needed to obtain the same effect is ten times smaller for 6-nitrochrysene. For that reason, it is not only important to identify the PAHs, but also the nitrated derivatives, adsorbed on diesel particulates. As such, because the individual products from incomplete combustion induce different

F. Portet-Koltalo et al. / Analytica Chimica Acta 651 (2009) 48–56

health risks, informations about total PAH emissions are less important than informations about the composition of such emissions. Chromatographic methods are very well suited for the separation, identification and quantification of each PAH. Analyses of complex mixtures of PAHs and nitroPAHs have been performed using a gas chromatograph fitted with flame ionisation detection [8] or electron capture detection (for nitroPAHs) [9], but more typically hyphenated with mass spectrometry (MS) [10,11]. Moreover some authors have used liquid chromatography with fluorescence detection [12,13], MS [14], or supercritical fluid chromatography [15]. However, the key step in quantitative analysis is often the process of sample extraction from the solid surface. Analysts often use conventional liquid solvent extraction methods that have changed very little over the decades, as is the case for soxhlet or sonication extractions [1,16]. These techniques require several hours for quantitative extractions (8 h in the case of a soxhlet extraction of PAHs and longer in the case of nitroPAHs), and they use large amounts of organic solvents. After extraction, the dilute extracts obtained have to be concentrated via solvent evaporation (which can lead to important losses of volatile analytes). Often a clean-up step is necessary before analysis to remove more polar matrix organics. Other extraction techniques have been developed more recently and used for the extraction of PAHs from environmental solids. These techniques include: microwave-assisted extraction (MAE) [17,18], assisted solvent extraction (ASE) [19], dynamic sonication [20] and supercritical fluid extraction (SFE) [21,22]. These techniques allow for a considerable reduction in extraction time and solvent consumption. In fact, MAE seemed particularly efficient at extracting PAHs that were strongly adsorbed on diesel particulates [18]. However, MAE needed to be compared to SFE, which also leads to rapid extractions. Indeed, SFE generates extracts that are ready for analysis and does not require additional concentration by means of solvent evaporation, which is not the case for MAE extracts. Moreover, SFE provides clean extracts because of its higher selectivity when compared to liquid solvent extraction techniques [23]. Then time-consuming clean-up steps that are also a source of analytical error are not necessary after SFE extractions [24]. Additionally, supercritical fluids offer the opportunity to control the solvating power of the extraction fluid by varying the temperature, pressure or modifier content [25]. In fact, carbon dioxide (the most used supercritical fluid) sometimes requires the addition of small amounts of cosolvent into the CO2 to increase PAH solubility [26]. The combined advantages of SFE led us to use this technique to carry out PAH and nitroPAH extractions, followed by a GC/MS analysis. Because spiked target pollutants were strongly adsorbed on the solid particulate material obtained from the combustion of diesel oils, SFE factors needed to be well understood and optimised to obtain good extractions. This was done herein using a factorial experimental design. The simultaneously studied parameters included the pressure and the temperature of the supercritical fluid, and also the nature and the percentage of the cosolvent added into the CO2 . Moreover, the effects of individual or combined original modifiers on PAH recovery yields were also investigated to elucidate the several mechanisms of desorption from the highly sorptive carbonaceous surface and to achieve quantitative extractions, not only for the PAHs but also for their nitrated derivatives.

2. Experimental 2.1. Samples and reagents Tetrahydrofuran, pyridine, diethylamine, chloroform (all of HPLC grade) and acetic acid (purity 99.5%) were purchased from Acros Organics (Noisy le Grand, France). Methylene chloride and toluene (HPLC grade) were purchased from Sigma–Aldrich (St. Quentin-

49

Fallavier, France). Carbon dioxide (quality for SFC) was provided by Air Liquide (Grand Quevilly, France). Magnesium sulphate (purity >99%) and sand, washed with sulphuric acid, were obtained from VWR International France (Fontenay sous Bois, France). Native diesel soot was provided by the CERTAM (Saint Etienne du Rouvray, France). Particulate exhaust samples were directly collected during engine tests using a particulate ceramic filter (a cordierite monolithic honeycomb) installed at the exhaust pipe just after the diesel engine (Europe’s Euro 4 standard). After a few hours of sampling, the accumulated soot particles were blown into glass flasks at 100 ◦ C. During combustion tests, the sampling temperature (>300 ◦ C) was continuously measured at the cordierite filter inlet. The four cylinder direct injection engine was operated under steady state conditions (power: 80 kW, torque: 220 N m, regime: 1500 rpm). Soot particles were stored at −20 ◦ C after collection. For spiking experiments, a test mixture of PAHs, nitrated PAHs and n-alkanes was diluted into toluene (concentration equal to 100 ␮g mL−1 ) and was composed of: naphthalene (NAPH, purity 99%), biphenyle (BIPH, purity 99%), acenaphtene (ACE, 99%), fluorene (FLUO, 99%), phenanthrene (PHEN, 96%), anthracene (ANT, 99%), fluoranthene (FLT, 98%), pyrene (PYR, 98%), benzo[k]fluoranthene (B(k)FLT, 99%), benzo[e]pyrene (B(e)PYR, 99%), benzo[a]pyrene (B(a)PYR, 97%), perylene (PER, 99%) and dibenz[a,h]anthracene (DB(ah)ANT, 98%), all of which were provided by Sigma–Aldrich–Fluka, whereas acenaphtylene (ACTY, 98%), chrysene (CHRY, 98%) and indeno[1,2,3-c,d]pyrene (I(1,2,3-c,d)PYR, 98%) were purchased from Interchim (Monluc¸on, France). Benz[a]anthracene (B(a)ANT, 99%), benzo[b]fluoranthene (B(b)FLT, 99%) and benzo[g,h,i]perylene (B(ghi)PER, 99%) were purchased from Supelco (Bellefonte, PA, USA). 1-nitronaphthalene (1N-NAPH, 99%), 1,5-dinitronaphthalene (1,5N-NAPH, 98%), 2nitrofluorene (2N-FLUO, 98%), 9-nitroanthracene (9N-ANT, 97%), 3-nitrofluoranthene (3N-FLT, 90%), 2,7-dinitrofluorene (2,7N-FLUO, 97%), heneicosane (HENEI, 99.5%), tetracosane (TETRA, 99.5%) and triacontane (TRIA, 99.5%) were provided by Sigma–Aldrich–Fluka. Two internal standards were used for quantisation after chromatographic analysis: 2-methylanthracene (named IS1 , purity 97%) and 7-methylbenzo[a]pyrene (named IS2 , purity 98%), and both were purchased from Sigma–Aldrich. 2.2. Extractions 2.2.1. Soot clean-up procedure A Soxhlet apparatus was used to clean 10 g of diesel particulate matter using methylene chloride. The solvent was refluxed for 8 h with approximately four cycles per hour. At the end, the diesel soot was dried and ground for homogenisation, and two successive extractions were performed under the same conditions, with fresh methylene chloride. The soots were considered to be clean after their supercritical fluid extractions produced blank samples, showing no aliphatic or aromatic impurities when analysed under the same conditions as those used for sample extracts. Cleaned soots were then stored at 4 ◦ C. 2.2.2. Supercritical fluid extractions Supercritical fluid extractions were performed using a Carlo Erba instrument (Thermo Fisher Scientific, Waltham, MA, USA) fitted with SFC 300 syringe pumps for carbon dioxide delivery. To generate modified CO2 , the syringe pump was connected to a sample loop (volume: 8.3 mL) to introduce the solvent directly inside the pump cylinder (total volume: 150 mL). Particulate matter (0.1 g) was introduced into an empty stainless steel column used as an extraction cell (4.6 mm i.d. and 50 mm length) purchased from Grace Davison France (Templemars, France). Gas tightness was ensured using special stainless steel frits (0.5 ␮m) with carbon reinforced peek rings, also purchased from Grace Davison France. Knowing that both

50

F. Portet-Koltalo et al. / Analytica Chimica Acta 651 (2009) 48–56

high flow-rates and long extraction times generally result in lower recovery yields of volatile compounds, the supercritical fluid flowrate was maintained at relatively low level (about 0.3–0.4 mL min−1 , measured as liquid fluid at the pump), using a capillary-fused silica restrictor (25 ␮m i.d. and 12 cm length) purchased from Thermo Fisher Scientific. To avoid restrictor plugging, the capillary tube was placed into a thermostatically controlled, heated block kept at 100 ◦ C. However, plugging of the restrictor often occurred when pure CO2 was used or when organic modifiers were added directly into the extraction cell. Indeed organic solvents used as polar modifiers could contain water, so frozen water could plug the outlet of the restrictor [27–29]. The addition of magnesium sulphate to the extraction cell was then absolutely necessary to absorb humidity. Moreover, sample matrices containing high concentrations of extractable components are highly prone to precipitating phenomena. In our case, this problem was markedly enhanced when n-alkanes were added to the PAHs and nitroPAHs mixtures as spiked pollutants. Adding an organic modifier inside the extraction cell prevented effectively the precipitating and plugging phenomena by increasing the solubility of the analytes. Unfortunately, when the modifier was entirely transferred to the bulk collection solvent after only several minutes of a dynamic extraction, precipitation occurred again into pure CO2 at the end of the restrictor. Then, it was more appropriate to continuously and dynamically introduce the organic modifier by mixing it with CO2 into the pump cylinder. Extracting cells, prepared 1 h before their extraction, were filled consecutively with: clean sand (to avoid dead volumes); 0.05 g of cleaned diesel soot spiked with 50 ␮L of the test mixture; 0.05 g of cleaned diesel soot and, finally, a mixture of sand, magnesium sulphate and copper granulates (to remove elemental sulphur). The extracts were collected by inserting the restrictor outlet into a colourless glass vial containing 2 mL of toluene. Toluene was preferred over methylene chloride as a collecting solvent [29] because it was not only a good solvent for gas chromatography (GC), but also because methylene chloride vaporised noticeably after 30 min of a dynamic extraction, even though the collection vial was capped and possessed only a little hole. After the extraction, vials were shaken and sonicated to avoid any problem of solute condensation on the vial walls, and the volume of collection solvent was measured. No phase separation was observed even when modifiers containing acetic acid or diethylamine were used. The final volume, which never exceeded 3 mL, was quite low, so we decided to avoid the evaporation step generally used to reduce the solvent volume. Then, extracts were analysed without any further clean-up process. A 980 ␮L volume of collecting solvent was transferred into a vial and 10 ␮L of each internal standard (IS1 and IS2 initially at 100 ␮g mL−1 ) was added before GC analysis. 2.3. GC/MS analysis Analysis of n-alkanes, PAHs and nitroPAHs was performed using a Hewlett–Packard model 5890 gas chromatograph (Agilent Technologies, Santa Clara, CA, USA) fitted with a 50 m DB5-MS capillary column (250 ␮m i.d., 0.25 ␮m film thickness, J&W Scientific, Folsom, CA, USA). The splitless injector was kept at 250 ◦ C and the injected volume equalled 1 ␮L. Helium (0.9 mL min−1 ) was used as the carrier gas. The analyses were performed using the following temperature program: 55 ◦ C isothermal for 2.05 min; then, the temperature was increased (first heating rate: 40 ◦ C min−1 up to 167 ◦ C, second heating rate: 3.9 ◦ C min−1 up to 300 ◦ C); finally, the final temperature was kept at 300 ◦ C for 8 min. The gas chromatograph was hyphenated with a HP5972 mass detector (electron impact ionization: 70 eV; electron multiplier voltage: 2200 V). The temperature of the MS transfer line was kept at 290 ◦ C. Peak identification was based on the retention times and the full mass spectra of standard PAHs (total ion current mode). Quantisation

was based on selected ion monitoring (SIM mode) for the molecular ion of each PAH, with some principal fragmentation ions being added for nitroPAHs and n-alkanes. The internal standard, previously termed IS1 , was used for the quantisation of the PAHs and nitrated derivatives containing up to four aromatic rings as well as for the n-alkanes having fewer than 25 carbons. IS2 was used for the quantisation of the heaviest target pollutants. Eight calibration plots (standard mixtures ranging from 0.1 to 8 ␮g mL−1 ) were used to determine the response factors of the 28 compounds. The detection thresholds in SIM mode, calculated as three times the standard deviation of blank sample noise, ranged from 0.15 to 2.5 ␮g L−1 for PAHs, from 40 to 60 ␮g L−1 for alkanes and from 5 to 100 ␮g L−1 for nitro and dinitroPAHs. Recovery yields obtained from quantisation after GC–MS analyses were analysed using JMP 5.1 software (SAS Institute, Cary, NC, USA) for statistical calculations and modelling after chemometric optimisation of the supercritical fluid extraction.

3. Results and discussion 3.1. Specifics of the studied diesel particulate matter Before discussing the optimisation of the extractions of PAHs and nitroPAHs from the spiked diesel particulate matter, we would like to mention some experimental difficulties that were inherent in the solid matrix that was studied. Indeed the diesel soot matrix was a real contaminated environmental material and was obtained from engine tests by direct collection from a particulate filter placed downstream of the diesel engine. It is well known that more and more diesel engines are equipped with particulate filters and there was concern over whether the generated soots had exactly the same surface properties and retention behaviours as generally studied soots (these latter ones being collected at the exhaust pipe without a particulate filter, under cold conditions). In fact, we used a spiking procedure to mimic the extraction of n-alkanes, PAHs and nitroPAHs from the soot material generated from the particulate filter, although several authors have demonstrated that commonly used spiking methods are not really valid for determining quantitative extraction conditions from environmental matrices [30]. However, it was experimentally impossible to reproduce the conditions that occurred during the deposition of these pollutants on real diesel soots. Moreover, it appeared that our diesel soot was effectively quite different from the standard reference material SRM 1650 diesel exhaust particulate matter purchased from the national institute of standard and technology (NIST). Indeed, this reference material, obtained from heavy duty diesel engines (produced in 1985), was collected at the exhaust pipe using a dilution tunnel at 52 ◦ C, whereas our samples, obtained from a recent light duty vehicle, were collected at temperatures greater than 300 ◦ C without any air dilution. In those collecting conditions, the SOF retained on our diesel particulates was markedly poorer than the soluble organic fraction that could be extracted from the SRM 1650 material (less than 10% for our particulates versus 20.2% for SRM 1650b). A similar diesel soot material, generated at 350 ◦ C, was described by Turrio-Baldassari et al. [31], who concluded that diesel soot with low content of soluble organic substances was likely to strongly retain all PAHs, particularly heavy ones. Effectively, thermal treatment studies of soot have demonstrated the important release of the most volatile part of the SOF at high temperatures, which constitutes approximately half of the SOF. Moreover, these studies demonstrated an increase in the specific surface area of the particulate matter at high temperatures [2]. Now, it is well known that accompanying pollutants in the soot sample may hamper the extraction of PAHs. And effectively, if SRM 1650 soots were extracted relatively easily under SFE conditions using

F. Portet-Koltalo et al. / Analytica Chimica Acta 651 (2009) 48–56

51

Fig. 2. Soxhlet extraction of 0.1 g of spiked Diesel soot, using 150 mL of solvent, during approximately 8 h (60 cycles). Fig. 1. Influence of the nature of the solid matrix (m = 0.1 g) on the SFE extraction of spiked PAHs, nitroPAHs and n-alkanes; Experimental conditions: CO2 containing 10% toluene as organic modifier, temperature: 100 ◦ C and pressure: 30 MPa; Static time: 15 min, dynamic time: 30 min.

CO2 and 10% toluene as an organic modifier [32], it was not the case for our diesel particulates, even when they were simply spiked samples. Indeed, Fig. 1 shows the poor extraction of our spiked diesel material using 10% toluene in CO2 , at 30 MPa and 100 ◦ C. To demonstrate that the nature and strength of the interactions between the solid surface and the PAHs limited the extraction (rather than the solubility in the extracting fluid), spiked sand was extracted under the same conditions. As expected, and although these extraction conditions were not optimised, PAHs and n-alkanes were easily and quantitatively extracted from the polar inorganic sand, with recoveries exceeding 80%, even for the heaviest PAHs, while extractions from the carbonaceous diesel soot led to poorer recoveries, especially for the heavy PAHs, for which recovery yields did not exceed 15%. To confirm the extremely difficult extraction of the heavy PAHs from our diesel soot, experiments were also performed with a conventional soxhlet extractor using methylene chloride or toluene as extracting solvents over 8 h. Fig. 2 shows that the recovery yields of the heaviest PAHs did not exceed 10–20% with methylene chloride and 15–25% with toluene. Consequently, it was clear that our diesel particulate matter was particularly refractory to conventional liquid extractions or supercritical methods of extraction, even more than reference materials such as SRM 1650. Those results showed us that even though spikes do not necessarily interact with the same active sites as the native analytes, spiking may be a method to elucidate the influence of different factors throughout the extraction process. It may also be a way to find many solutions to the problem of difficult extractions from very specific environmental matrices. 3.2. Chemometric optimisation of SFE extractions Multivariate optimisation of SFE extraction permits the study of several variables simultaneously, taking into account the interaction between these variables [33]. Moreover, the number of experiments involved is considerably smaller than the classical univariate approach. This is a great benefit when four operating variables are considered: the pressure and temperature of CO2 as well as the identity and percentage of the organic modifier added to CO2 . Several other factors concerning the collection efficiency were previously studied, as recommended by S. B. Hawthorne et al. [34],

by spiking the relatively inert sand matrix. These factors include: the nature and volume of the collecting solvent, the collection vial shape, the temperature of restrictor heater and the size of restrictor (see experimental section). The amount of diesel particulates (0.1 g) and the quantity of spiked compounds (50 ␮L of a 100 ␮g mL−1 solution) were also kept constant. Finally, some experiments were performed with spiked soot, which let us to conclude that 15 min for static extractions and 30 min for dynamic ones were sufficient for obtaining recoveries which did not significantly change with longer times. At this point, we could start our chemometric optimisation using a full central composite design comprising star points which allowed for estimation of the response curvature. The number of centrepoint runs (called zero points) was set to six, which were dispersed throughout the design matrix (while other runs were randomised). They ensured that the variance of prediction was the same over all the experimental space (uniform precision design). Table 1 presents the experimental space: supercritical fluid pressure (P) and temperature (T) levels were chosen with regard to instrumental constraints. The percentage of organic modifier (%) added to CO2 was investigated over a wide range (5–25%). In contrast, the nature of cosolvent (Solv) studied was a nominal factor and it was not easy to choose among all the organic modifiers that could be used with carbon dioxide [35,36]. Thus, only three cosolvents were selected, according to their classification by L. R. Snyder et al. [37]. Although some authors have said that, for the case of non-polar solutes containing no functional groups (like PAHs), the cosolvent-induced solubility is quite similar for all cosolvents [38], we wondered if any specific interactions between PAHs and modifiers would occur. To this end, we chose cosolvents that might have pronounced selective effects on many solutes, due to their extreme position within Snyder’s solvent scheme. Indeed, knowTable 1 Description of the experimental domain according to the four factors studied and the coded levels of the factorial design. Factors P (MPa) T (◦ C) % Solvent

Levels −␣

−1

0

+1

+␣

10 50 5 CHCl3

15 75 10 CHCl3

20 100 15 THF

25 125 20 CH2 Cl2

30 150 25 CH2 Cl2

52

F. Portet-Koltalo et al. / Analytica Chimica Acta 651 (2009) 48–56

ing that three types of interactions could define the selectivity of a solvent, the selected modifiers were classified into three different groups that could principally participate in one kind of interaction: proton donor with chloroform (group VIII), dipolar interaction with methylene chloride (group V) and proton acceptor with tetrahydrofuran (group III). A mathematical model using multiple linear regressions permitted the adjustment of the experimental response (recovery yields) obtained from the 30 experiments. Parametric second order equations were calculated using the JMP software and allowed relating the observed values to the factors and to their combinations. Then, analysis of variance could permit one to test for the statistical significance of the estimates (main effects for factors and second or third order interactions). Table 2 summarises the main estimates, interactions or quadratic effects really influencing the extracted amounts. As can be seen in Table 2, only representative compounds were chosen to illustrate the results of modelling, with PAHs subdivided according to their molecular weights: light volatile PAHs (from NAPH to ANT), medium semi-volatile PAHs (from FLT to CHRY), heavy non-volatile PAHs (from B(b)FLT to B(ghi)PER), nitrated PAHs and n-alkanes. It is clear from Table 2 that all main factors significantly influenced the recovery yields, but to different extents. In general, all extracted pollutants were mainly influenced by the nature of the modifier (Solv) and, to a lesser extent, by the pressure (P) of the extracting mixture. The fact that these two factors had positive estimates showed that methylene chloride (+1 level) was always markedly better than THF (0 level) as an organic modifier, which itself was better than chloroform (−1 level), and that an increase in the fluid pressure was favourable for the extraction of all compounds. On one hand, it is well known that an increase in the pressure causes an increase in the supercritical fluid viscosity, which results in an increase in the solvating power of CO2 , thus permitting a better extraction of solutes. However, as mentioned earlier, solubility does not seem to be the limiting factor for supercritical extractions from diesel particulates, since all PAHs could be extracted from spiked sand and not from our spiked diesel soot. In fact, quantitative extractions require extraction fluids that can also overcome matrix-analyte interactions. The efficiency enhancement resulting from the addition of modifiers to CO2 has been effectively explained by the displacement of analytes from their sorption sites on the solid, breaking the analyte-matrix interactions. Modifiers appear to act by competitively sorbing to high-energy matter sites, thus decreasing the heterogeneity of available sites. This is why the modifier identity generally has a larger effect on the extraction efficiency than the pressure. Differences in chemical or physical properties of modifiers also have a greater effect than their percentage into CO2 [39]; this tendency was also well observed in Table 2. Next, it should be noted that of all the factors, fluid temperature had the weakest significant effect, though negative values for estimates showed that an increase in this factor was rather unfavourable to the extraction. It is generally recognised that an increase in temperature causes a general decrease of the supercritical fluid density

and, consequently, a decrease of the supercritical fluid’s solvating power [10]. Finally, the effect of modifier percentage (%) into CO2 was more difficult to interpret directly from Table 2 because this factor occurred not only as a main factor but interacted also with other factors, a quadratic term appearing furthermore for some PAHs and n-alkanes. Some main and quadratic effects were even contradictory, showing that an increase in the modifier percentage could be favourable at relatively low percentages (homogeneous supercritical state) but could be disastrous at the highest levels, where the extracting fluid may be more probably at the subcritical state. The shape of the fitted overall responses could be best summarised in three-dimensional graphs, generating response surface plots. Fig. 3 shows different ways to illustrate the PAH extraction yields according to the four factors, with one factor kept constant in each. For example, Fig. 3a effectively shows that a maximum pressure is recommended for the PAH extractions and that methylene chloride is the better modifier. Once the pressure has been kept at its maximum level, Fig. 3b illustrates that the temperature must be quite low for better recoveries and that very high percentages of modifier are not really favourable to PAH extractions, especially when methylene chloride is used as the organic modifier. In fact, this tendency was more pronounced for PAHs heavier than anthracene. For high molecular weight PAHs, which interact greatly with the active sites of the soot, a concentration of 15% modifier added to the CO2 was found to be the best value. Concerning the extraction of nitrated PAHs and n-alkanes, Fig. 4 shows again the general tendencies already described. It must be noted here that nitroPAHs were quite insensitive to the percentage of modifier (see Fig. 4(a)) and that n-alkanes were less sensitive than the PAHs or their nitrated derivatives to the nature of the cosolvent (see Fig. 4(b)). Effectively, these latter compounds were always quantitatively extracted when the pressure was kept at 30 MPa (which is above their threshold pressure), regardless of the modifier added, or if a modifier was not added at all. To summarise all these observations, it can be concluded that the best conditions for simultaneously extracting PAHs, nitroPAHs and n-alkanes are: pressure at the highest level (30 MPa), temperature at a low level (75 ◦ C) and use of methylene chloride as a modifier at the mid level (15%). Table 3 shows the extraction recoveries obtained under such conditions, with the accuracy of all the procedure being estimated to average 15%. The robustness of this method was evaluated by the experimental design, using tests at ±10% of the optimal values (with methylene chloride as a modifier). The least robust factor was the percentage of modifier. However, for all the compounds, these minor changes in the variables led to changes in the recoveries that were always smaller than the variance. Therefore, the extraction method could be considered robust. Unfortunately, while PAHs from NAPH to CHRY, light nitroPAHs and n-alkanes were extracted quantitatively, the heaviest PAHs and their nitrated derivatives were not. Consequently, the main dipolar interaction developed by methylene chloride with the active sites and the analytes, which is more efficient than proton donor or proton acceptor forces, is not sufficient to disrupt the strong inter-

Table 2 Estimated effects of the studied factors and/or the statistically significant interactions (t-test approach) for some typical compounds. Volatile PAHs

P T % Solv % × Solv P × Solv P×T P × % × Solv (%)2

Semi-volatile PAHs

Heavy PAHs

Nitrated PAHs

PYR

B(ghi)PER

1N-NAPH

NAPH

PHEN

2.52 −2.14 8.31 13.03 −5.46

8.32 −1.9

9.52

17.55 −5.74 9.45

22.03 −6.59

−7.17

B(a)ANT 5.13

1.42

4.9 17.24

1.41 2.37

6.11

n-alkanes 3N-FLT 2.62 −2.04

14.51 −5.96

15

−5.84

−4.35

TETRA 4.48 10.4 13.29 −10.6

2.96 5.28 −7.25 −8.41

2.06

−6.8

TRIA 10.5 11.3 17.8 −9.88

F. Portet-Koltalo et al. / Analytica Chimica Acta 651 (2009) 48–56

53

Table 3 Recovery yields (%) and Relative Standard Deviations (%) calculated with optimal SFE extraction conditions: 100 mg of spiked Diesel soot were extracted using 15% methylene chloride into CO2 , at 75 ◦ C and 30 MPa.

NAPH BIPH ACE ACTY FLUO 1N-NAPH PHEN ANT 1,5N-NAPH FLT HENEI 2N-FLUO PYR 9N-ANT a b

Recovery yieldsa (%)

R.S.D.b (%)

96 92 98 91 95 84 91 83 65 63 97 30 79 27

9.3 15.7 10.9 11.2 13.8 14.7 14.4 17.3 24.0 24.5 15 49 16.9 46

TETRA B(a)ANT CHRYS 3N-FLT 2,7N-FLUO B(b)FLT B(k)FLT B(e)PYR B(a)PYR PER TRIA I(1,2,3-c,d)PYR DB(ah)ANT B(ghi)PER

Recovery yieldsa (%)

R.S.D.b (%)

84 74 72 35 20 41 37 36 29 25 82 12 12 10

19.7 18 19.7 25.3 10.5 12.3 14.3 11.4 14.8 23.6 23.0 18.6 21.0 23.2

Average recovery yields calculated on five replicates. Relative Standard Deviations estimated using the variance analysis of the factorial design.

actions between five and six ring PAHs (and heavy nitroPAHs) and the diesel soot matrix.

3.3. Use of aromatics and mixtures of organic modifiers The carbon present in diesel particulate matter has an important aromatic structure, which is certainly mainly responsible for the strong adsorption of PAHs. Indeed, some authors have calculated the percentage of sp2 hybridization for a standard diesel exhaust soot to be 56% [40]. Thus, many aromatic solvents, which might better compete for the matrix adsorption sites, were tested as organic modifiers. Fig. 5 presents the results obtained when 15% toluene, nitrobenzene or pyridine was added to the CO2 . Recoveries were slightly better when 15% toluene was added to CO2 at 75 ◦ C and 30 MPa than the results presented in Fig. 1, where extraction conditions were not optimised. Nevertheless, toluene did not markedly enhance the whole recovery yields when compared to methylene chloride, except those of the three heaviest PAHs, as can be seen in Table 3. In contrast, pyridine and nitrobenzene were much better cosolvents for extracting PAHs and nitroPAHs, respectively, even the heaviest ones. Considering the native PAHs, pyridine was the best modifier and could enhance ten to forty times the extraction of heavy PAHs when compared to pure CO2 and two to four times when compared to the CO2 /methylene chloride mixture (see Fig. 5). It was obvious that its aromatic character was not sufficient to explain this phenomenon, as the simply aromatic toluene was unable to disrupt all matrix/solutes interactions; its basic character also had to be considered. As such, we enhanced the basic character of the modifier by adding a proportion of diethylamine into the pyridine. Fig. 6 shows that, in particular, the 6/1 mixture of pyridine/diethylamine (v/v) was the best modifier, permitting the quantitative extraction of all PAHs, even the three heaviest ones, with recovery yields exceeding 80%. In fact, the success of such a modifier mixture was based on an important interaction force that had not been taken into account with our first choice of cosolvents. Specifically, it had become clear that the ␲-electron donor/acceptor interactions were mainly governing the adsorption of the PAHs. Effectively, the diesel soot surface was able to strongly adsorb the aromatic species by forming charge transfer interactions with the PAHs, with these interactions involving the partial shift of ␲-electron density from one molecule to another. An attack of a strong electron donor such as pyridine resulted in a disruption of the ␲–␲* interactions between the soot surface ␲-acceptor sites and the PAHs (␲-electron donors). The PAHs were next rapidly solvated by the supercritical fluid and then transported away from their adsorption sites [41]. Moreover, the addition of a base could

also decrease the ␲-acceptor ability of the surface aromatic moieties [42], thus increasing the modifier capacity to extract the PAHs. While the combined effect of the pyridine and the diethylamine base led to a quantitative desorption of native PAHs from the soot surface, it was noticed that the addition of diethylamine was inefficient at desorbing the nitrated PAHs (see Fig. 6). Moreover, it must be recalled that, contrary to PAHs, nitrated derivatives were incompletely extracted from the sand matrix (see Fig. 1) owing to the fact that many strong polar active sites could interact with the nitrated moieties. If we suspected that H-bonding interactions predominated in the case of the sand matrix, it was not true for the adsorption of nitroaromatics on the soot surface; otherwise, THF (as a proton acceptor) would have been the best modifier to replace and then desorb the nitroPAHs (weak proton acceptors), which was not the case. In fact, nitrobenzene, which is a good electron acceptor solvent, was a better organic modifier than THF or pyridine (see Fig. 5). Again, this result was used to support the existence of electron donor–acceptor interactions between the nitroPAHs and the polyaromatic soot surface, though this time on different active sites (electron donor sites) [43]. Finally, we tried to add acid to pyridine, thinking that it could help displace the nitrated PAHs from the soot surface more easily than could a base. Fig. 7 shows that the ternary mixture composed of CO2 , 1% of acetic acid and 14% of pyridine was favourable for the extraction of nitroPAHs. This mixture was less efficient than nitrobenzene for extracting light nitrated PAHs but more efficient for heavy nitroPAHs. In contrast, it was obvious that the addition of an acid into pyridine was not favourable for the extraction of native PAHs or even n-alkanes (see Fig. 7).

3.4. Validation of the method The optimised SFE extraction procedure, using original mixtures of modifiers to enhance the extraction of heavy PAHs and nitroPAHs from the highly sorptive diesel particulate matter, was evaluated in terms of recoveries, precision (as relative standard deviation), sensitivity and linear range. Although accuracy is generally assessed by analysing a certified reference material, we noticed, as shown previously, that our diesel particulate matter exhibited extraction behaviour that was markedly different from the reference material SRM 1650. Thus, we worked on blank soots spiked with known concentrations of the analytes to validate our method. Table 4 shows the mean recoveries, the accuracy and the sensitivity of the method. The precision of the method was established from five replicates, with 5 ␮g of each pollutant spiked on 0.1 g of clean soot. Values of RSD were lower than those presented in Table 3 because these values were estimated from the variance analysis of the facto-

54

F. Portet-Koltalo et al. / Analytica Chimica Acta 651 (2009) 48–56

Fig. 3. Response surfaces (recovery yields in %) after SFE extractions of anthracene, using methylene chloride, tetrahydrofuran or chloroform as organic modifiers, according to: (a) supercritical fluid temperature and pressure (modifier percentage: 15%); (b) fluid temperature and modifier percentage (pressure: 30 MPa).

Fig. 4. Response surfaces (recovery yields in %) after SFE extractions using methylene chloride, tetrahydrofuran or chloroform as organic modifiers (pressure: 30 MPa): (a) 3-nitrofluoranthene; (b) tetracosane.

Table 4 Figures of merit of the SFE extraction method for the determination of PAHs, nitroPAHs and n-alkanes from highly sorptive spiked diesel soot (m = 0.1 g), using modifiers containing pyridine/diethylamine (PAHs and n-alkanes) or pyridine/acetic acid (nitroPAHs) into CO2 .

NAPH BIPH ACE ACTY FLUO 1N-NAPH PHEN ANT 1,5N-NAPH FLT HENEI 2N-FLUO PYR 9N-ANT a

Mean recoverya (%)

RSD (%)

89 91 90 97 96 63 84 83 60 92 98 61 85 66

6.1 5.6 5.8 6.5 6.4 10 6.8 7.1 12.7 7.0 7.0 11.8 7.5 14.6

LOD (ng) 0.9 0.5 0.6 0.5 1.9 18 1 1 76 2 100 37 2.1 22

Average recovery yields calculated on five replicates.

LOQ (ng g−1 ) 0.30 0.17 0.20 0.17 0.63 6.0 0.33 0.33 25 0.67 33 12 0.70 7.3

TETRA B(a)ANT CHRYS 3N-FLT 2,7N-FLUO B(b)FLT B(k)FLT B(e)PYR B(a)PYR PER TRIA I(1,2,3-c,d)PYR DB(ah)ANT B(ghi)PER

Mean recoverya (%)

RSD (%)

LOD (ng)

LOQ (ng g−1 )

91 81 82 58 59 86 84 81 79 79 90 81 81 79

12.5 8.9 10.0 13.9 14.7 11.1 11.2 9.2 9.1 12.0 12.5 14.9 13.9 15.1

150 2 1.8 25 420 1.6 1.7 1.6 2.7 2.2 170 6.5 7.8 6.5

50 0.67 0.60 8.3 140 0.53 0.57 0.53 0.90 0.73 57 2.2 2.6 2.2

F. Portet-Koltalo et al. / Analytica Chimica Acta 651 (2009) 48–56

Fig. 5. Influence of the nature of aromatic modifiers (added to CO2 ) on the SFE extractions of many PAHs, nitroPAHs and n-alkanes; Experimental conditions: CO2 containing 15% organic modifier, temperature: 75 ◦ C and pressure: 30 MPa.

rial design and bias effects introduced on the entire experimental domain studied were higher in this case. Effectively, experiments were randomised and extracting conditions differed considerably between two successive extractions (temperature, pressure, nature and percentage of modifier varying). In contrast, values of accuracy presented in Table 4 came from samples analysed on different days by two different operators, but with the same extracting conditions. These values are in agreement with those presented in different works for PAHs extracted by SFE from soils (1–22% RSD) [44,45], from spiked urban sewage sludge (8–26% RSD) [12] and from marine sediments (2–7.6% RSD) [10]. Precisions presented by the certificate of analysis of the certified diesel particulate matter SRM 1650b, combining different methods of extraction and analysis, were also in this range (1.7–19.5% RSD). The precision was lower for the extraction of nitroPAHs, but it was in agreement with results of T. Paschke et al (7.1–25% RSD) [32]. Quantification limits (calcu-

Fig. 6. Influence of the addition of diethylamine into pyridine (total percentage of solvent mixture into CO2 : 15%) on the SFE extractions of many PAHs and nitroPAHs (pressure: 30 MPa, temperature: 75 ◦ C).

55

Fig. 7. Influence of the addition of acetic acid into pyridine (total percentage of solvent mixture into CO2 : 15%) on the SFE extraction of many PAHs and nitroPAHs (pressure: 30 MPa, temperature: 75 ◦ C).

lated as ten times the standard deviation of blank sample noise) were ranging from 0.2 to 3 ng g−1 for PAHs, from 6 to 140 ng g−1 for nitroPAHs and from 33 to 57 ng g−1 for n-alkanes. The sensitivity was estimated by calculating the detection limits of each compound (defined for signals equivalent to the blank signal plus three times its standard deviation) (see Table 4). Then, the linear range was evaluated and was satisfactory for spikes containing 0.2 ␮g of each pollutant to spikes containing 40 ␮g of each pollutant. Finally, knowing that the procedure was optimised using cleaned diesel soots, no interference appeared in chromatographic analysis. However, it must be underlined that, compared to SFE extracts, soxhlet extracts were darker. The absence of interference could then be also explained by a greater selectivity of the SFE process over the conventional soxhlet extraction. 4. Conclusions Extraction recoveries of PAHs adsorbed on two different diesel particulates were first compared. SFE extractions performed on a certified diesel particulate matter (SRM 1650) by T. Paschke et al. [32] led to quantitative recoveries, but this was not the case for our diesel soot, which was not collected in the same way as the certified material (collecting temperatures, in real operating conditions of the diesel engine, being markedly higher and modifying the SOF rate and probably the specific surface area). A comparison with the conventional soxhlet extraction permitted us to conclude that this carbonaceous surface was highly sorptive for heavy PAHs and nitroPAHs, and that conventional extracting conditions would not permit quantitative extraction of these target pollutants. Then, a second order central composite design, intended to optimise the extraction of the complex mixture of PAHs, nitroPAHs and nalkanes from the very refractory diesel soot surface allowed us to understand the influence of four factors on their supercritical fluid extraction: the nature of the organic modifier had the highest influence, followed by the supercritical fluid pressure, then by the percentage of organic modifier added into CO2 and finally, to a lesser extent, by the temperature of the fluid. Three tested organic modifiers, each of which could develop one principal kind of interaction with the surface active sites, led us to conclude that neither dipolar forces nor proton donor–acceptor forces were sufficient for quantitatively displacing and extracting heavy PAHs and their

56

F. Portet-Koltalo et al. / Analytica Chimica Acta 651 (2009) 48–56

nitrated derivatives from the mainly sp2 hybridised soot surface. ␲electron donor–acceptor interactions were consequently taken into account by using pyridine as an organic modifier added into CO2 , which turned out to be the most successful cosolvent. The addition of a basic solvent into pyridine further enhanced the capacity of the extracting fluid to desorb the heaviest native PAHs, while the addition of a small proportion of an acid to the pyridine further enhanced the extraction of nitrated compounds. Unfortunately, while the extraction of native PAHs and n-alkanes was quantitative with the addition of 15% of the 6:1 (v:v) pyridine/diethylamine mixture into CO2 , the extraction of heavy nitroPAHs was only fair (recoveries exceeding 60%) when using nitrobenzene or a mixture composed of pyridine and 1% of acetic acid into CO2 . Moreover, heavy PAHs and nitrated PAHs could not be simultaneously and quantitatively desorbed by the same extracting supercritical fluid: a choice of the modifier identity must be made before SFE extractions of target pollutants can be analysed. At last, we can conclude that the next step of this study will be to use this method not only for extracting naturally polluted diesel materials (collected inside wall-flow particulate filters), but also certified standard materials, in order to be sure that the optimised extracting mixtures are also capable of desorbing aged matrices, where links are known to be stronger. References [1] R. de Abrantes, J.V. de Assunc¸ao, C.R. Pesquero, Atmos. Environ. 38 (2004) 1631–1640. [2] S. Collura, N. Chaoui, B. Azambre, G. Finqueneisel, O. Heintz, A. Krzton, A. Koch, J.V. Weber, Carbon 43 (2005) 605–613. [3] A. Duran, M. Carmona, J.M. Monteagudo, Chemosphere 56 (2004) 209–225. [4] H.H. Mi, W.J. Lee, C.B. Chen, H.H. Yang, S.J. Wu, Chemosphere 41 (2000) 1783–1790. [5] J.P.A. Neeft, M. Makkee, J.A. Moulijn, Fuel Process. Technol. 47 (1996) 1–69. [6] N.Y. Kado, R.A. Okamoto, P.A. Kuzmicki, C.J. Rathbun, D.P.H. Hsieh, Chemosphere 33 (1996) 495–516. [7] D. Villemin, D. Cherquaoui, A. Mesbah, J. Chem. Inf. Comput. Sci. 34 (1994) 1288–1293. [8] M. Barzegar, A. Daneshfar, M. Ashrafkhorassani, Anal. Chim. Acta 349 (1997) 245–252. [9] R. Deuster, N. Lubahn, C. Friedrich, W. Kleibömer, J. Chromatogr. A 785 (1997) 227–238. [10] V. Librando, O. Hutzinger, G. Tringali, M. Aresta, Chemosphere 54 (2004) 1189–1197.

[11] N. Yassaa, B.Y. Meklati, A. Cecinato, F. Marino, Atmos. Environ. 35 (2001) 1843–1851. [12] C. Miege, J. Dugay, M.C. Hennion, J. Chromatogr. A 823 (1998) 219–230. [13] N. Hüsers, W. Kleibömer, J. Chromatogr. A 697 (1995) 107–114. [14] L. Bonfanti, M. Careri, A. Mangia, P. Manini, M. Maspero, J. Chromatogr. A 728 (1996) 359–369. [15] F. Gritti, G. Felix, M.F. Achard, F. Hardouin, Chromatographia 53 (2001) 201–206. [16] K. Karar, A.K. Gupta, Atmos. Res. 84 (2007) 30–41. [17] M. Letellier, H. Budzinski, J. Bellocq, J. Connan, Org. Geochem. 30 (1999) 1353–1365. [18] F. Portet-Koltalo, K. Oukebdane, F. Dionnet, P.L. Desbene, Anal. Bioanal. Chem. 390 (2008) 389– 398. [19] K. Li, M. Landriault, M. Fingas, M. Llompart, J. Hazard. Mater. 102 (2003) 93– 104. [20] C. Domeno, M. Blasco, C. Sanchez, C. Nerin, Anal. Chim. Acta 569 (2006) 103–112. [21] T.L. Chester, J.D. Pinkston, D.E. Raynie, Anal. Chem. 70 (1998) 301R–319R. [22] G. Anitescu, L.L. Tavlarides, J. Supercrit. Fluids 38 (2006) 167–180. [23] J.W. King, J. Chromatogr. Sci. 27 (1989) 355–364. [24] S. Bowadt, S.B. Hawthorne, J. Chromatogr. A 703 (1995) 549–571. [25] T.M. Young, W.J. Weber, Anal. Chem. 69 (1997) 1612–1619. [26] V. Camel, A. Tambuté, M. Caude, J. Chromatogr. A 642 (1993) 263–281. [27] M.D. Burford, S.B. Hawthorne, D.J. Miller, T. Braggins, J. Chromatogr. A 609 (1992) 321–332. [28] C. Nerin, R. Batlle, M. Sartaguda, C. Pedrocchi, Anal. Chim. Acta 464 (2002) 303–312. [29] J.J. Langenfeld, M.D. Burford, S.B. Hawthorne, D.J. Miller, J. Chromatogr. 594 (1992) 297. [30] M.B. Burford, S.B. Hawthorne, D.J. Miller, Anal. Chem. 65 (1993) 1497–1505. [31] L. Turrio-Baldassari, C.L. Battistelli, Anal. Bioanal. Chem. 375 (2003) 589–595. [32] T. Paschke, S.B. Hawthorne, D.J. Miller, J. Chromatogr. 609 (1992) 333–340. [33] C. Nerin, E. Asensio, C. Jimenez, Anal. Chem. 74 (2002) 5831–5836. [34] S.B. Hawthorne, D.J. Miller, M.D. Burford, J.J. Langenfeld, S. Eckert-Tilotta, P.K. Louie, J. Chromatogr. 642 (1993) 301–317. [35] V. Janda, K.D. Bartle, A.A. Clifford, J. Chromatogr. 642 (1993) 283–299. [36] R. Batlle, C. Nerin, C. Crescenzi, H. Carlsson, Anal. Chem. 77 (2005) 4241– 4247. [37] L.R. Snyder, J.J. Kirkland, Introduction to Modern Chromatography, Wiley, New York, 1999. [38] G. Anitescu, L.L. Tavlarides, J. Supercrit. Fluids 11 (1997) 35–51. [39] C. Lutermann, E. Willems, W. Dott, J. Hollender, J. Chromatogr. A 816 (1998) 201–211. [40] R.J. Hopkins, A.V. Tivanski, B.D. Marten, M.K. Gilles, J. Aerosol Sci. 38 (2007) 573–591. [41] J. Hollender, J. Scheine, W. Dott, M. Heinzel, H.W. Hagemann, G.K.E. Götz, J. Chromatogr. A 776 (1997) 233–243. [42] D. Zhu, S. Hyun, J.J. Pignatello, L.S. Lee, Environ. Sci. Technol. 38 (2004) 4361–4368. [43] M. Sanders, J.J. Pignatello, Environ. Sci. Technol. 39 (2005) 1606–1615. [44] B.W. Wenclawiak, T. Paschke, M. Krappe, Fresenius J. Anal. Chem. 357 (1997) 1128–1132. [45] N. Saim, J.R. Dean, P. Md. Abdullah, Z. Zakaria, J. Chromatogr. A 791 (1997) 361–366.