Talanta 48 (1999) 451 459

Headspace solid phase microextraction (HSSPME) for the determination of volatile and semivolatile pollutants in soils
Maria Llompart a,*, Ken Li b, Merv Fingas b
Departamento de Quimica Analitica Nutricion y Bromatologia, Facultad de Quimica, Uni6ersidad de Santiago de Compostela, E -15706 Santiago de Compostela, Spain b Emergencies Science Di6ision, En6ironment Canada, En6ironmental Technology Centre, 3439 Ri6er Road, Ottawa, ON, Canada
a

Received 3 April 1998; received in revised form 10 July 1998; accepted 3 August 1998

Abstract We have investigated the use of headspace solid phase microextraction (HSSPME) as a sample concentration and preparation technique for the analysis of volatile and semivolatile pollutants in soil samples. Soil samples were suspended in solvent and the SPME bre suspended in the headspace above the slurry. Finally, the bre was desorbed in the Gas Chromatograph (GC) injection port and the analysis of the samples was carried out. Since the transfer of contaminants from the soil to the SPME bre involves four separate phases (soil-solvent-headspace and bre coating), parameters affecting the distribution of the analytes were investigated. Using a well-aged articially spiked garden soil, different solvents (both organic and aqueous) were used to enhance the release of the contaminants from the solid matrix to the headspace. It was found that simple addition of water is adequate for the purpose of analysing the target volatile organic chemicals (VOCs) in soil. The addition of 1 ml of water to 1 g of soil yielded maximum response. Without water addition, the target VOCs were almost not released from the matrix and a poor response was observed. The effect of headspace volume on response as well as the addition of salt were also investigated. Comparison studies between conventional static headspace (HS) at high temperature (95C) and the new technology HSSPME at room temperature ( 20C) were performed. The results obtained with both techniques were in good agreement. HSSPME precision and linearity were found to be better than automated headspace method and HSSPME also produced a signicant enhancement in response. The detection and quantication limits for the target VOCs in soils were in the sub-ng g 1 level. Finally, we tried to extend the applicability of the method to the analysis of semivolatiles. For these studies, two natural soils contaminated with diesel fuel and wood preservative, as well as a standard urban dust contaminated with polyaromatic hydrocarbons (PAHs) were tested. Discrimination in the response for the heaviest compounds studied was clearly observed, due to the poor partition in the headspace and to the slow kinetics of all the processes involved in HSSPME. 1999 Elsevier Science B.V. All rights reserved. Keywords: Solid phase microextraction; Headspace; VOC; Soil analysis

* Corresponding author. Tel./fax: + 34 981 547141; e-mail: qblvrlgb@usc.es 0039-9140/99/$ - see front matter 1999 Elsevier Science B.V. All rights reserved. PII S0039-9140(98)00263-X

452

M. Llompart et al. / Talanta 48 (1999) 451459

1. Introduction A recent advance in sample preparation for trace analysis is solid phase microextraction (SPME) technology. In this solvent-free extraction technique, developed in 1989 by Pawliszyn [1 5], the analytes are adsorbed directly from an aqueous [3] or gaseous phase [6] onto a fused-silica bre coated with a polymeric phase. Hence sampling, extraction and concentration are accomplished in a single step. The entire assembly is mounted in a modied syringe needle which, after exposure to the sampling media (water or air), is inserted into the heated injection port of a GC. The analytes are desorbed in the GC injector and analysis of the samples is carried out. Separation and detection then proceed in the usual manner. The SPME bre can also be suspended in the headspace above the aqueous or solid sample (headspace solid phase microextraction (HSSPME)), which eliminates interference problems because the bre is not in contact with the sample [7,8]. SPME has become very popular in the last 2 or 3 years, especially in environmental analysis [914]. Environmental contamination with volatile organic chemicals (VOCs) associated with fuel/ petroleum usage is widespread in the world. Spills during transport and leaking storage tanks are the main sources of contamination to soils, ground water and air. Other sources of petroleum product contamination are heating oil tanks, reneries, aboveground tanks, terminals, pipelines, or accidental spills from other sources. These compounds have gained prominence in environmental pollution control over the past decade as a result of increased environmental and health concerns and the introduction of new regulations. The most commonly used protocols for the analysis of VOCs in water are static headspace (HS) or purge-and-trap techniques (PAT). These techniques have also been applied to the analysis of soils and sediments [15]. In this paper we present the development of a HSSPME method to extract volatile organic compounds (VOCs) from soil matrices. Model VOCs include toluene, ortho -, meta - and para -xylene, chlorobenzene, 1,2-, 1,3- and 1,4-dichlorobenzene.

Soil samples were suspended in solvent and the SPME bre suspended in the headspace above the slurry. Finally the bre was desorbed in the GC injection port and the analysis of the samples was carried out. Different parameters affecting the distribution of the analytes between the different phases were investigated. The HSSPME method at room temperature ( 20C) was compared with conventional automated HS sampling at high temperature (95C). Due to the lack of adequate reference soil material of VOC, a typical urban soil was spiked with a mixture of target compounds and aged for several months to simulate real soil samples. Conventional HS results were compared to the HSSPME results and found to be in good agreement. The proposed method offered signicant advantage in terms of linearity, precision and detection limits. Although this method shown to be well suited for VOC analysis, compounds of low volatility partition poorly into HS and are generally not amenable to headspace type of analysis. The limitation of the HSSPME method developed in this work for the analysis of semivolatiles was studied through three naturally contaminated samples: one with hydrocarbons and another two with PAHs.

2. Experimental

2.1. Instrumentation
Static headspace analysis was performed using a Hewlett Packard HP19395A headspace sampler and a HP5890 Series II gas chromatograph equipped with a 5970 mass selective detector (MSD). Experimental parameters of the HS sampler were as follows: equilibration time, 30 min (nominal); bath temperature, 95C; sample loop, 3 ml; valve/loop temperature, 110C; valve operation sequence of pressurisation, 10 s; venting and lling of loop, 5 s; and injection 15 s. The carrier gas was helium at 80 ml min 1; and the auxiliary pressure of 1.5 bar. Normal HS was run using a constant heating time accessory on the headspace sampler, consisted of a sample magazine and the controlling pneumatics which dropped one vial at a time into the heated carousel as the previous

M. Llompart et al. / Talanta 48 (1999) 451459

453

Fig. 1. Effect on the response after the addition of different polar solvents to the soil.

sample was being analysed, such that each sample vial was equilibrated for the time equivalent to one GC run (heating up and cooling down; nominally 30 min). A manual SPME holder was used with a 100mm polydimethylsiloxane bre assembly (Supelco, Mississauga, ON). The analysis was performed on the above system with the headspace transfer line detached from the injection port. GC conditions were the same in normal HS and in HSSPME analysis, and were as follows: inlet temperature, 225C; inlet mode, split operation with split ratio 1:10 (splitless operation in SPME); split vent ow, 60 ml min 1; oven temperature, 40C hold 5 min, rate 7.5C min 1 to nal temperature 200C; column, 30-m SPB-1, 0.53 mm i.d., 1.5-mm lm, column ow, 7.5 ml min 1 nominal; linear veloc-

ity, 40 cm s 1 at 100C. An open-split interface was used to limit the ow to the MSD to 0.7 ml min 1. The MSD was operated in SIM mode. For PAHs analysis another GC/MSD system similar to the one above was employed with a 30-m DB-5 GC column (0.25 mm i.d., 0.25 mm lm). GC temperature program was: oven temperature, 80C hold 1 min, rate 15C min 1 to nal temperature 300C. The injector temperature was 260C and the capillary interface temperature was 300C. The MSD was operated in selected ion monitoring mode with two or three monitored ions per compound. Both systems were controlled by a HPChem station (DOS series). For total hydrocarbon analysis, a 5890 GC equipped with ame ionisation detector (FID) was used. A 30-m DB-5 GC column (0.25 mm

454

M. Llompart et al. / Talanta 48 (1999) 451459

Fig. 2. Effect on the response after the addition of different amounts of water to the soil.

i.d., 0.25 mm lm) was used. Oven temperature was 50C for 1 min and heated to 310C at the rate of 15C min 1. Injector and detector temperature were 280 and 310C, respectively. The GC column ow was nominally 1 ml min 1.

2.2. Reagents and chemicals

A ten-components VOC standard of 2000 mg ml 1 of each of the target analytes was obtained from HP (part no. 8500-6080). The target analytes were benzene, toluene, chlorobenzene, ethylbenzene, p -, m -, o -xylene, 1,2-, 1,3- and 1,4dichlorobenzene. The standard was diluted ten times in methanol to give a spiking solution of 200 mg ml 1. All the solvents (analytical grade) were purchased from Caledon (Belleville, ON, Canada). For the study of volatiles, experiments were performed in a sandy-loam sub-surface soil col-

lected from a garden in an urban area in Ottawa. This soil consists of 48.4% sand, 32.3% silt and 13.3% clay with 2.0% organic carbon. After removing twigs and extraneous material it was dried in an oven at 90C for 4 h and then homogenised by crushing in a mortar and screened to a particle size of 200 mm. This soil did not contain any of the VOCs studied in this work. About 100 g was weighed out in a jar, to which the VOC standard mixture dissolved in 100 ml of methanol was added to give a soil concentration of 2 mg g 1 of each of the VOC target compounds. The slurry was allowed to stand, loosely covered to protect it from dust, and stirred occasionally until the methanol completely evaporated (approximately 2 days). The soil was then capped and kept in a desiccator. The nal VOC content of this soil was not known due to evaporation loss during preparation and storage (approximately 18 months). Benzene has not been included in this study be-

M. Llompart et al. / Talanta 48 (1999) 451459

455

Fig. 3. Effect on the response with the headspace volume.

cause due to its high volatility, it was not detected in the aged soil. Soils spiked in this manner, and aged for a long period of time, resembled a real sample more than the common technique of spikTable 1 HSSPME detection and quantication limits for the target VOCs Detection limit (ng g1) Toluene Chlorobenzene Ethylbenzene p+m -Xylene o -Xylene 1,3-Dichlorobenzene 1,4-Dichlorobenzene 1,2-Dichlorobenzene 0.23 0.15 0.08 0.05 0.07 0.14 0.13 0.15 Quantication limit (ng g1) 0.78 0.49 0.27 0.16 0.23 0.47 0.42 0.49

ing one spot in the soil matrix just before analysis, because the target analytes were in more intimate contact with the soil particles, and thus maximising analyte/matrix interaction. For calibration studies a standard addition protocol was used. Different levels of VOCs were added to the spiked soil. A 1-g soil aliquot was added into a 20-ml vial and spiked with a 100-ml of MeOH containing the VOCs. After capping, the vials were tumbled for 0.5 h and allowed to equilibrate at room temperature for 24 h before sampling and analysis. For the analysis of semivolatiles three real samples were used: (i) a soil contaminated with diesel fuel from a gas process plant, in which we analysed the total hydrocarbon content; (ii) a wood preservative contaminated soil (SRS103100) purchased from Fisher Scientic (Fair Lawn, NJ) for the analysis of PAHs; and (iii) an urban dust (SRM 1649) purchased from NIST (Gaithersburg, MD) for the analysis of PAHs.

456

M. Llompart et al. / Talanta 48 (1999) 451459

Table 2 HSSPME versus HS: linear correlation coefcients and ratio of responses (HSSPME/HS) Correlation coefcients HSSPME Toluene Chlorobenzene Ethylbenzene p+m -Xylene o -Xylene 1,3-Dichlorobenzene 1,4-Dichlorobenzene 1,2-Dichlorobenzene 0.9993 0.9999 0.9992 0.9989 0.9999 0.9997 0.9993 0.9991 HS 0.9971 0.9981 0.9987 0.9996 0.9998 0.9991 0.9975 0.9973 2 2 3 10 7 8 7 7 Ratio of responses (HSSPME/HS)

2.3. Con6entional HS sampling

HS sampling and analysis was performed following the method proposed by Voice and Kolb [15]. Two grams of soil were introduced into a 22-ml HS-vial. Three millilitres of water were added to each vial and then the vial was placed in the HS bath at 95C for 0.5 h before analysis.

3.1. Optimisation of HSSPME conditions

The degree of partitioning of VOCs between the soil and the HS is generally very low. To promote the release of volatiles, water can be added to the soil [7,16]. The effect of adding water as well as organic polar solvents to the sample matrix was studied. Fig. 1 shows the responses obtained after adding 100 ml of different solvents to 1 g of sample. As can be seen, there was an increase in the response in all cases. The highest results, however, were obtained using water, which produced a 100-fold increment to the signal obtained without the addition of solvent. Different mechanisms are involved in the desorption of volatiles from the soil into the headspace when a polar solvent is used. One of them is the displacement of the analytes from the active sites in the soil. The active sites are usually polar functional groups that have more afnity for polar compounds than for less polar ones. The polar compounds added to the soil will substitute, at least in part, the VOCs on these active sites and, in consequence, the VOCs will be released into the headspace. Another mechanism involved is the competition between the matrix and the solvent for the analytes. Part of the analytes can be desorbed from the soil into the solvent by solvation, and then from the solvent into the headspace. The quantity of water added to 1 g of soil was investigated next. The experiments were carried out with the addition of water between 0 and 15 ml. The results are shown in Fig. 2. This plot

2.4. HSSPME sampling

Different amounts of water or organic solvents were added into 1 g soil in a 22-ml HS-vial, to promoted the release of volatiles. A 1-cm stir bar was placed in each vial, which was sealed with an aluminium cap with a Teon-lined septum. The SPME bre was suspended in the HS for 30 min. The watersoil mixture was stirred during the sampling time to promote faster equilibrium. Once sampling was complete, the bre was immediately inserted into the GC injector port for desorption. A desorption time of 3 min at 260C was enough for quantitative desorption of all the analytes studied and no carry over was observed.

3. Results and discussion The bre exposure time required for equilibrium between phases was not studied since it is a matrix dependant parameter. However, we found equilibrium was reached after 30 min. Further exposure up to 1 h did not show any increase in the response.

M. Llompart et al. / Talanta 48 (1999) 451459 Table 3 Concentration of VOCs (ng g1) found in the well-aged spiked soil by HSSPME and by HS HSSPME Mean (ng g1)a Toluene Chlorobenzene Ethylbenzene p+m -Xylene o -Xylene 1,3-Dichlorobenzene 1,4-Dichlorobenzene 1,2-Dichlorobenzene
a

457

HS RSD (%) 10.0 2.2 4.0 4.2 5.6 2.1 3.9 4.2 Mean (ng g1)a 22.7 52.8 61.3 95.0 75.7 628.5 747.2 881.1 RSD (%) 12.8 8.9 7.0 10.0 8.9 6.2 11.5 8.2

33.1 79.0 64.7 104.6 75.9 583.9 651.8 807.2

n = 3.

exhibits a maximum response for 1 ml of water added to the system. However, the responses obtained between 100 ml and 5 ml were not signicantly different. When the amount of water added was 5 ml or more, the responses decreased. The results clearly demonstrate that the addition of water to the samples is necessary to release the VOCs into the gas phase. The fact that the responses were practically not dependent on the amount of water between 100 ml and 5 ml for a 1-g sample size has a signicant practical advantage, because in real soil samples, the moisture content can be very different, and is often not known. The results of this study indicate that the HSSPME method proposed for the target VOCs is not sensitive to the moisture level of the soil. For subsequent studies, the amount of water added to 1 g of soil was set at 1 ml. The effect of the addition of salt to the samples was also studied. One gram of KCl was added to each vial and HSSPME was performed in the same manner. In contrast to other authors [7], no increase in the response was observed after the addition of salt. The addition of salt to aqueous samples is frequently used to enhance the sensitivity of headspace analysis of polar compounds; for non-polar compounds this effect is expected to be insignicant [17,18]. The analytes studied in this paper have very low polarity and, consequently no signicant enhancement in sensitivity should be expected after the addition of salt.

The effect of the headspace volume on extraction efciency was investigated next. A set of experiments was performed using 1 g of soil and 1 ml of water in different size vials, with a headspace volume of 4, 9, 21 and 39 ml. The results obtained for toluene, ethylbenzene, p xylene and 1,2-dichlorobenzene are shown in Fig. 3. The responses obtained were similar for headspace volumes between 4 and 21 ml. The ability to work with bigger HS volume without decreasing the response of the method has an important practical advantage; if the bre is situated further from the stirred soil/water mixture there is a lesser tendency for soil particles to stick on the bre, which results in a cleaner operation. It also removes the constraint of using a particular size container. For a headspace volume of approximately 40 ml the response was lower, between 54 and 82% of the response obtained with the other three headspace volumes.

3.2. Repeatability, reproducibility, detection limit and quantication limit

A series of ve consecutive extractions were performed on different aliquots of soil in order to evaluate the repeatability of the HSSPME method. The precision of the HSSPME method was very good and the relative standard deviations (%RSD) were between 1 and 5% for all the compounds.

458

M. Llompart et al. / Talanta 48 (1999) 451459

Table 4 MSD Response obtained for PAHs in urban dust (SRM1649) after HSSPME at 20 and 100C (The reference concentration values are also given) Reference value (mg g1) Response (area cts) 20C Phenanthrene Fluoranthene Pyrene Benzo(a)anthracene Chrysene Benzo(b,k)uoranthene Benzo(a)pyrene Benzo(g,h,i)perylene Indeno(1,2,3-cd)pyrene nd, not detected. 4.5 7.1 6.6 2.6 3.6 8.2 2.9 4.5 3.3 2170 1641 nd nd nd nd nd nd nd 100C 1714626 1146076 62831 150455 53840 12027 7623 801 674

Another series of ve extractions of the same soil were performed in different days to evaluate the reproducibility of the method. The precision was also very good and the %RSD were between 3 and 6% for all the compounds, with the exception of toluene (RSD = 10%). The detection and quantication limits (signalto-noise ratio = 3 and 10, respectively) were also determinated and are reported in Table 1. Detection and quantication limits for all the target VOCs were in the sub-ng g 1 level.

3.3. Linearity study and 6alidation of the method: con6entional HS 6ersus HSSPME
Calibration studies for both techniques, HS and HSSPME, were performed. To accommodate for possible matrix effects and the incomplete transfer of analytes to the HS, a standard addition protocol was performed on the well-aged spiked soil. A ve-level calibration study was carried out with the addition of each target analyte between 0 and 2 mg g 1. One gram of soil was spiked with a 100-ml of MeOH containing the analytes. After capping, the vials were tumbled for 0.5 h and allowed to equilibrate at room temperature for 24 h before sampling and analysis. Both techniques gave linear calibration curves for all the target analytes, and the regression coefcients obtained for each compound are shown in Table 2. HSSPME was found to have better linearity than

conventional HS, with correlation coefcients in 0.999 or greater for all the compounds. The concentration of analytes found in the soil with HSSPME and conventional HS are shown in Table 3. Statistical tests showed that the two techniques gave equivalent results for all the analytes with the exception of chlorobenzene and toluene. For these compounds HSSPME gave higher results than HS. The precision obtained with HSSPME was also better than conventional HS (Table 3). The sensitivity of both techniques was also compared. The responses obtained with HSSPME method were between two and 11 times higher than those with the HS method (Table 2).

3.4. Application of the method to less 6olatile compounds

For soil samples, the application of SPME is limited to sampling the headspace over the soil or soil water mixture to avoid the contact between the bre and the soil. We tried to extend the application of the HSSPME method proposed in this paper to the analysis of semivolatiles. The possibility of analysing semivolatile compounds in solid samples at room temperature is largely limited by the poor release of these compounds into the vapour phase. For these studies three naturally contaminated samples, one with diesel and the other two with PAHs, were tested. Discrimination in the responses obtained was observed in

M. Llompart et al. / Talanta 48 (1999) 451459

459

both cases for the high boiling compounds, due to their low partition into the headspace and to the slow kinetics that these compounds have at room temperature. After 30 min sampling, the system is still far away from equilibrium. In the diesel soil sample, the characteristic diesel hump around C14C20 was only barely detectable in the chromatogram when compared to that from conventional solvent extraction. In the wood preservative soil and in the urban dust, both contaminated with PAHs, it was only possible to identify PAHs smaller than four rings. For quantitative analysis of semivolatiles in soils it is necessary to promote the partition of the compounds into the headspace and speed up the mass transfer process. One of the possibilities is to heat the soil-water slurry. In fact, increasing the temperature from 20 to 100C and sampling for the same time (30 min), all the PAHs in the soil could be identied. In the urban dust, which concentration in PAHs is much more lower than in the wood preservative soil, all the PAHs contained could be detected, but the responses for benzo(g,h,i)perylene and indeno(1,2,3-cd)pyrene were very low. Table 4 shows the MSD responses obtained for this certied reference material after HSSPME at 20 and 100C. The certied concentration values are also given. Even at 100C the discrimination in the responses with the size of the PAH is very signicant. Further research in the application of HSSPME for the analysis of semivolatiles in soils is being performed.

or spill samples. For semivolatile compounds in the mid-boiling range, the HSSPME proposed method did not perform well, due to the poor partition of these compounds into the headspace and to the slow kinetics of all the process involved in HSSPME that these compounds have at room temperature. At higher temperature, an adequate screening analysis of a soil sample contaminated with PAHs could be done, but still very low response for the six-ring analytes was obtained. Further studies are being performed to extend the application of HSSPME to the quantitative analysis of semivolatiles.

References
[1] J. Pawliszyn, Solid Phase Microextraction. Theory and Practice, Wiley-VHC, New York, 1997. [2] C.L. Arthur, J. Pawliszyn, Anal. Chem. 62 (1990) 2145. [3] C.L. Arthur, L.M. Killan, S. Motlagh, M. Lim, D.W. Potter, J. Pawliszyn, Environ. Sci. Technol. 26 (1992) 979. [4] A.A. Boyd-Boland, M. Chai, Y.Z. Luo, Z. Zhang, M.J. Yang, J. Pawliszyn, T. Gorecki, Environ. Sci. Technol. 28 (1994) 569A. [5] Z. Zhang, M.J. Yang, J. Pawliszyn, Anal. Chem. 66 (1994) 844A. [6] M. Chai, J. Pawliszyn, J. Environ. Sci. Technol. 29 (1995) 693. [7] Z. Zhang, J. Pawliszyn, J. High Resolut. Chromatog. 16 (1993) 689. [8] J. Czerwinsky, B. Zygmunt, J. Namiesnik, Fresenius J. Anal. Chem. 356 (1996) 80. [9] K.D. Buchholz, J. Pawliszyn, J. Anal. Chem. 66 (1994) 160. [10] A.A. Boyd-Boland, S. Magdic, J. Pawliszyn, Analyst 121 (1996) 929. [11] H. Lakso, W.F. Ng, Anal. Chem. 69 (1997) 1866. [12] T.K. Choudhury, K.O. Gerhardt, T.P. Mawhinney, Environ. Sci. Technol. 30 (1996) 3259. [13] B.L. Wittkamp, S.B. Hawthorne, D.C. Tilotta, Anal. Chem. 69 (1997) 1204. [14] J. Poerschmann, Z. Zhang, F.-D. Kopinke, J. Pawliszyn, J. Anal. Chem. 69 (1997) 597. [15] T.C. Voice, B. Kolb, Environ. Sci. Technol. 27 (1993) 709. [16] Z. Zhang, J. Pawliszyn, Anal. Chem. 67 (1995) 34. [17] B. Kolb, L.S. Ettre, Static Headspace-Gas Chromatography. Theory and Practice, Wiley-VHC, New York, 1997. [18] Z.E. Penton, in: P.R. Brown, E. Grushka (Eds.), Advances in Chromatography, vol. 37, Chapter 5, Marcel Dekker, New York, 1997.

4. Conclusions HSSPME at room temperature ( 20C) has been shown to be a very suitable methodology for the quantitative determination of VOCs in soil samples. The addition of water to the soil is required to release the volatiles into the gas phase. The enhancement in the sensitivity over conventional HS at 95C is signicant. The technique offers good performance in terms of precision (RSD 5%) and linearity, and because of simplicity and low equipment cost, is well suited for the rapid quantication of residual volatiles in landll
.