Development of an Analytical Procedure for ... - ACS Publications

Aug 14, 2008 - TES Bretby, P.O. Box 100, Bretby Business Park, Burton-on-Trent, DE15 0XD, U.K., Centre for Resource. Management and Efficiency, Sustai...
1 downloads 11 Views 111KB Size
Anal. Chem. 2008, 80, 7090–7096

Development of an Analytical Procedure for Weathered Hydrocarbon Contaminated Soils within a UK Risk-Based Framework Graeme C. Risdon,† Simon J. T. Pollard,‡ Kirsty J. Brassington,‡ Jamie N. McEwan,† Graeme I. Paton,§ Kirk T. Semple,| and Fre´de´ric Coulon*,‡ TES Bretby, P.O. Box 100, Bretby Business Park, Burton-on-Trent, DE15 0XD, U.K., Centre for Resource Management and Efficiency, Sustainable Systems Department, School of Applied Sciences, Cranfield University, Cranfield, MK43 0AL, U.K., Department of Plant and Soil Sciences, School of Biological Sciences, Cruickshank Building, University of Aberdeen, Aberdeen, Scotland, AB24 3UU, U.K., and Department of Environmental Science, Lancaster University, Lancaster, LA1 4YQ, U.K. A sequential ultrasonic extraction method for contaminated soils with weathered hydrocarbons is presented. The method covers the determination of total petroleum hydrocarbons between nC8 and nC40, and subranges of hydrocarbons including diesel range organic compounds, kerosene range organic compounds, and mineral oil range organic compounds in soils. Further modifications to the carbon banding may be made as requested for risk assessment. These include a series of ranges known as Texas banding (from the Texas Risk Reduction Program) as well as separation of the aliphatic and aromatic fractions. The method can be routinely used for measuring hydrocarbons down to 10 mg kg-1 in soil. Lower limits can be achieved by employing a suitable solvent concentration step following extraction; however, this would result in increased cycle time. Detection limits may vary for individual carbon ranges calculated on the percentage of the full range they cover. With an extraction efficiency and recovery between g95 and 99%, this method can be easily positioned as a good alternative to Soxhlet extraction and shows a good potential for implementation as a standard method potentially providing further insight to the contaminated land sector. Risk assessment is a well-established and important tool for environmental management decisions, which is increasingly being used in petroleum hydrocarbon-contaminated land management.1 However, these frameworks are not always supported by suitable and robust analytical protocols, especially in the case of matrixes contaminated with weathered hydrocarbons.1 The authors have had a long interest in improved analytical protocols and methods for the analysis of risk- critical compounds in hydrocarbon * To whom correspondence should be addressed. E-mail: f.coulon@ cranfield.ac.uk. Tel: +44 (0)1234 750 111. Fax: +44 (0)1234 751 671. † TES Bretby. ‡ Cranfield University. § University of Aberdeen. | Lancaster University. (1) Brassington, K. J.; Hough, R. L.; Paton, G. I.; Semple, K. T.; Risdon, G. C.; Crossley, J.; Hay, I.; Askari, K.; Pollard, S. J. T. Crit. Rev. Environ. Sci. Technol. 2007, 37, 199–232.

7090

Analytical Chemistry, Vol. 80, No. 18, September 15, 2008

matrixes.2-5 Many hydrocarbon-contaminated sites (former refineries, coal carbonization plants, and integrated steelworks) contain (i) oils that are weathered because the source term has aged since release,3,6 (ii) heavy fuel oil residues,7 or (iii) viscous tars and solid bituminous process residues that are difficult to treat biologically.8 The comparison of reference analytical methods used for petroleum risk assessment protocols1 highlights the need for practical and simple extraction procedures that allow a better characterization of both aliphatic and aromatic hydrocarbon fractions within oil-contaminated samples, including soil and sediment samples with high moisture levels.1,9 Within a UK context, the development of novel methods should also allow the identification of risk indicator compounds within each hydrocarbon fraction and the monitoring of recalcitrant biomarkers to enable verification of treatment success. Concerns exist over the performance of the current reference methods used, specifically in terms of poor extraction efficiencies and analytical losses imparted by sample handling. The alteration of chemical composition with time may also affect the accuracy of final measurements and lead to misrepresentations of human health risk. In this respect, the impact of calibration on final measurement needs to be evaluated for a range of weathered products. To date, while the UK approach sets out guidelines for evaluating human health risks from petroleum hydrocarboncontaminated soils, as yet there has been no specification or adoption of recommended analytical procedures for these con(2) Dale, M. J.; Jones, A. C.; Pollard, S. J. T.; Langridge-Smith, P. R. R. Analyst 1994, 119, 571–578. (3) Pollard, S. J. T.; Hrudey, S. E.; Fedorak, P. M. Waste Manage. Res. 1994, 12, 173–194. (4) Whittaker, M.; Pollard, S. J. T.; Fallick, T. E. Environ. Technol. 1995, 16, 1009–1033. (5) Pollard, S. J. T.; Hrudey, S. E.; Rawluk, M.; Fuhr, B. J. J. Environ. Monit. 2004, 6, 713–718. (6) Westlake, D. W. S.; Jobson, A.; Phillippe, R.; Cook, F. D. Can. J. Microbiol. 1974, 20, 915–928. (7) Uhler, A. D.; Stout, S. A.; McCarthy K. J.; Emsbo-Mattingly, S.; Douglas, G. S.; Beall, P. W. Soil Sediment and Water, 2002. (8) Gray, M. R.; Banerjee, D. K.; Dudas, M. J.; Pickard, M. A. Biorem. J. 2000, 4, 249–257. (9) Environment Agency. Science Report P5-080/TR3, ISBN 1 84432 342 0, Almondsbury, Bristol, 2005. 10.1021/ac800698g CCC: $40.75  2008 American Chemical Society Published on Web 08/14/2008

taminants. Furthermore, the framework itself notes that currently adopted methods for petroleum hydrocarbon analysis may not be suitable for the heavier compounds and questions whether it is practical or relevant for analyzing weathered hydrocarbons.9 These observations further highlight the need to develop a suitable and robust analytical procedure to inform risk assessment. The development of a novel analytical procedure must (i) be capable of analyzing petroleum hydrocarbon fractions and risk indicators as required, (ii) incorporate an extraction method that is suitable for the weathered hydrocarbon range, (iii) be practical in application, and (iv) not entail excessive cost. Soxhlet extraction is a widely used, benchmarked, exhaustive, and easily standardized technique for the extraction of petroleum hydrocarbon-contaminated soils.10 Disadvantages, including lengthy extraction times, degradation of thermally liable compounds, use of large volumes of organic solvents, and the need to concentrate samples, have resulted in the investigation into alternative exhaustive and robust methods.4,11,12 Ultrasonic extraction has been investigated elsewhere and has potential for wider use in this area of analysis.13,14 Ultrasonication is a quick, easy, and cost-effective method that is now widely used in environmental analysis. However, analytical procedures using ultrasonication vary not only in the method used (e.g., type and volume of solvents, cycle duration, etc.) but also in the type of ultrasonic apparatus used (sonic probe or ultrasonic bath). Some of the more detailed investigations have shown that ultrasonic methods have the potential to produce equivalent or better efficiencies than currently used methods such as Soxhlet.13,15-17 Conversely, other investigations have shown the opposite to this with worse efficiencies compared to alternative methods.12,18 We conclude that if sonication is to be used in place of traditional methods, it needs to be clearly defined and optimized. In this paper, we present a novel solvent ultrasonic extraction procedure for soils contaminated with weathered hydrocarbons, allowing petroleum hydrocarbon class fractionation and identification of risk-indicator compounds. This research is part of an investigation conducted by a research consortium (PROMISE) on optimizing biopile processes for weathered hydrocarbons within a risk management framework, with the principal objective of improving end-user confidence in biopile technology. EXPERIMENTAL SECTION A sequential ultrasonic solvent extraction method has been evaluated using four different soil matrixes: (i) silty soil, (ii) clay soil, (iii) sandy soil, and (iv) a granular matrix comprising ash, (10) Shu, Y. Y.; Lai, T. L.; Lin, H-S.; Yang, T. C.; Chang, C-P. Chemosphere 2003, 52, 1667–1676. (11) Hawthorne, S. B.; Grabanski, C. B.; Martin, E.; Miller, D. J. J. Chromatogr., A 2000, 892, 421–433. (12) Hollender, J.; Koch, B.; Lutermann, C.; Dott, W. Int. J. Environ. Anal. Chem. 2003, 83, 21–32. (13) Banjoo, D. R.; Nelson, P. K. J. Chromatogr., A 2005, 1066, 9–18. (14) Sanz-Landaluze, J.; Bartolome, L.; Zuloaga, O.; Gonza´lez, L.; Dietz, C.; Ca´mara, C. Anal. Bioanal. Chem. 2006, 384, 1331–1340. (15) Heemken, O. P.; Theobald, N.; Wenclawiak, B. W. Anal. Chem. 1997, 69, 2171–2180. (16) Sun, F.; Littlejohn, D.; Gibson, M. D. Anal. Chim. Acta 1998, 364, 1–11. (17) Sporring, S.; Bøwadt, S.; Svensmark, B.; Bjo¨rklund, E. J. Chromatogr., A 2005, 1090, 1–9. (18) Song, Y. F.; Jing, X.; Fleischmann, S.; Wilke, B. M. Chemosphere 2002, 48, 993–1001.

brick, and concrete fragments selected to represent the surface soils found at many industrial sites (“made”ground). Each soil matrix was spiked with a mixture of diesel and lubricating oil at levels corresponding to 20 (10 000 mg kg-1) and 80% (40 000 mg kg-1) of the concentration range typical to environmental soil samples.19,20 Standards, Solvents, and Reagents. All solvents used were HPLC grade and purchased from Lab3 (Northampton, UK). Silica gel grade 923, o-terphenyl (CAS 92-94-4), squalane (CAS 111-01-3), 2,2,4,4,6,8,8-heptamethylnonane (CAS 4390-04-9), and 2-fluorobiphenyl (CAS 321-60-8) were purchased from Sigma-Aldrich (Dorset, UK). Anhydrous sodium sulfate was purchased from Fisher Scientific (Loughborough, UK). Silica gel and anhydrous sodium sulfate were heated at 110 °C for 12 h and at 400 °C for 4 h, respectively, before use. The removal of any total petroleum hydrocarbons (TPH) was confirmed by the analysis of the blank control method. Diesel fuel and Mineral Oil Standards (neat motor oil, 15w-50) were used as quality control standards and purchased commercially from fuel filling stations. Proof of purchase was retained to provide traceability for calibration solutions created from primary fuels. Florida Total Recoverable Petroleum Hydrocarbon standard (C8-C40 Florida TRPH) and semivolatile Calibration Mix 5 (16 priority EPA polycyclic aromatic hydrocarbon (PAH) mix EPA 8310) were supplied by Thames Restek Ltd. (Saunderton, UK). Soil certified reference material was purchased from RTC Corp. (Catalogue Reference Number CRMPR9583) containing TPH at 9510 mg kg-1. Sequential Ultrasonic Solvent Extraction. Soil samples (5 ± 0.05 g) were chemically dried with 5 g of anhydrous sodium sulfate. In order to evaluate the recovery from the extraction method, dried samples were spiked with 1 mL of a surrogate solution containing o-terphenyl, squalane, heptamethylnonane, and 2-fluororbiphenyl at a concentration of 200 µg mL-1each in acetone. Spiked samples were manually shaken for 1 min and allowed to equilibrate for at least 12 h at 4 °C, minimizing the potential loss of volatile compounds. Spiked soil samples were then extracted with 4 mL of acetone and sonicated for 2 min at 20 °C. Hexane and acetone were added to the samples in a 1:1 ratio. Samples were sonicated for a further 10 min, followed by manual shaking to break up and mix the sample matrix. This step was repeated twice and followed by centrifugation for 5 min at 750 rpm. After passing the supernatant through a filter column fitted with glass receiver tube, a sequential step series, including resuspension of the samples in 10 mL of acetone/hexane (1:1), sonication for 15 min at 20 °C, centrifugation for 5 min at 750 rpm, and then decantation into a filter column, was repeated twice. The final extract volume was adjusted to 40 mL with a mixture of acetone/hexane (1:1) and homogenized by manual shaking before gas chromatography analysis. Microscale Silica Gel Column Chromatography: Cleanup and Class Fractioning. The extract from the ultrasonic process can be directly used to analyze TPH content without further need for concentration or dilution. It should be mentioned here that TPH analysis typically used for risk assessment protocols does (19) Risk-based methodologies for evaluating petroleum hydrocarbon impacts at oil and natural gas E&P Sites, API Publication 4709; API Publishing Services: Washington DC, 2001. (20) Sarkar, P.; Ferguson, M.; Datta, R. Birnbaum Sci. Environ. Pollut. 2005, 136, 187–195.

Analytical Chemistry, Vol. 80, No. 18, September 15, 2008

7091

Figure 1. Analytical schematic recommended for analyzing soil contaminated with weathered hydrocarbons.

not cover polar compounds but only nonweakly polars as they are usually not considered for evaluating human health risk from petroleum hydrocarbons in soils.1,9 A silica gel column chromatography procedure was used to split the extracted hydrocarbons into aliphatic and aromatic class fractions (Figure 1). Approximately 80 mL of polished water and a spatula of sodium chloride (baked at 400 °C for 4 h) were added to the sample extracts, partitioning out any acetone into the water and ensuring the removal of the nonpolar content. TPH silica cleanup was then carried out by passing 1 mL of the upper phase from the partitioned samples through a silica gel column, eluting with 3 mL of dichloromethane (DCM). In contrast, the split of the aliphatic/aromatic fractions was realized by eluting with 3 mL of hexane followed by 3 mL of DCM, respectively. Instrumental Analysis. Quantification of TPH content was carried out using a gas chromatograph fitted with a flame ionization detector (GC-FID Agilent model 6890). Chromatography was performed on a fused-silica capillary column (15 m × 0.32 mm internal diameter) coated with HP-5 (0.25-µm film thickness; Agilent). Helium was used as the carrier gas at 2.5 mL min-1, and the FID detector temperature kept at 350 °C. Pulsed splitless injection with a sample volume of 3 µL was applied. The oven temperature was increased from 45 to 250 °C at a gradient of 120 °C min-1, then increased to 300 at 100 °C min-1, and finally increased to 340 at 90 °C min-1 and held at this temperature for 4 min. The total run time was 6.2 min. Quantification of TPH and subranges of hydrocarbons including Texas banding compounds 7092

Analytical Chemistry, Vol. 80, No. 18, September 15, 2008

(Texas 1-5),12 diesel range organic (DRO), kerosene range organic (KRO), and mineral oil range organic (MRO) compounds in soils was made by integrating peak areas using Agilent Chemstation Software Revision B.01.01 (164) SR1 (2001-2005) and by comparison of these peaks with the response of a known concentration of diesel and mineral oil standards. External multilevel calibrations were carried out for both diesel/mineral oil fractions and surrogates, quantification ranging from 0.5 to 2500 µg mL-1 and from 1 to 5 µg mL-1, respectively. Surrogate standards were used to determine the extraction efficiency of the method. For quality control, a 500 µg mL-1 diesel standard and mineral oil were run every 10 samples. In addition, duplicate blank controls and reference material were systematically used. The blank control was treated in exactly the same manner as the samples but contained no soil. When analyzed, if the blank concentration exceeded the method reporting limits either of 2 or 10 mg kg-1 for Texas banding and the TPH, DRO, MRO, and KRO hydrocarbon ranges, respectively, the samples series was re-extracted. The reference material was an uncontaminated soil of known characteristics and was spiked with a diesel and mineral oil standard at a concentration equivalent to 16 000 mg kg-1. Polycyclic aromatic hydrocarbons were identified and quantified by GC/MS using an Agilent 6890 gas chromatograph coupled to an Agilent 5975 mass spectrometer operated at 70 eV in positive ion mode. Splitless injection with a sample volume of 1 µL was applied. The oven temperature was increased from 80 to 150 °C

Table 1. Mean Concentration, Precision, and Bias for Duplicate Sample Extraction for Each Soil Matrix matrix silty soil

clay soil

sandy soil

madegrounda

Matrix RTC CRMPR 9583

spike concentration, mg kg-1

mean concentration, mg kg-1

precision, mg kg-1

precision, % RSD

bias, mg kg-1

0 1000 5000 10000 30000 0 1000 5000 10000 30000 0 1000 5000 10000 30000 0 1000 5000 10000 30000 9510

65 942 5876 9988 29280 81 487 5479 10142 30104 9 510 5387 9727 28759 286 616 5588 10802 31166 11124

7 38 235 300 878 23 15 219 710 2107 6 15 215 389 575 63 18 224 324 1558 334

11 4 4 3 3 28 3 4 7 7 62 3 4 4 2 22 3 4 3 5 3

n/a -58 876 -12 -720 81 -513 479 142 104 n/a -490 387 -273 -1241 n/a -384 588 802 1166 1614

bias, %

total error, %

n/a -6 18 0 -2 -51 10 1 0 n/a -49 8 -3 -4 n/a -38 12 8 4 17

n/a 14 26 6 8 56 57 18 15 14 n/a 55 16 11 8 n/a 44 20 14 14 23

a Madeground, a granular matrix comprising ash, brick, and concrete fragments, which was selected to represent surface soil found on many industrial sites (n/a not applicable).

at 30 °C min-1 and then increased to 320 at 15 °C min-1 and held at this temperature for 4 min. The mass spectrometer was operated using the full scan mode (range m/z 50-500) for quantitative analysis of target PAHs. For each compound, quantification was performed by integrating the peak at specific m/z. Internal multilevel calibration was carried out ranging from 0.01 to 5 ng µL-1. For quality control, a 1.0 ng µL-1 PAH Calibration Mix 5 Standard solution (Restek) was analyzed every 15 samples. Validation Procedure. Validation provides confidence that the established performance characteristics are based on robust experimental determinations and are statistically sound. Each spiked soil matrix was allowed to stand for 24 h at room temperature before commencing extraction, to allow the spike to interact with the soil matrix. Performance characteristics were determined with a minimum of 10 degrees of freedom by analyzing each certified reference material or spiked samples in duplicate in different analytical batches. Eleven batches of duplicates were analyzed for each matrix at each spiking level thus providing 10 degrees of freedom in each validation experiment. RESULTS AND DISCUSSION The sequential solvent ultrasonic method presented here has been evaluated using several different soils and concentration levels in addition to the use of a certified reference material. The mean TPH concentrations achieved for each tested soil, concentration levels, precision, and bias of the samples are shown in Table 1. The results showed that the method had good extraction efficiency and recovery independently of soil type with relative standard deviation (RSD) values for all the soils of below 10% for all of the spiked soils. However, recoveries below 5000 mg kg-1 would not be quite as high but at least g50% recovery may be expected by using this technique. The relatively high bias obtained may be due to the lack of evaporation step within the method. The highest degree of variability was obtained in the clay soil

(Table 1), a trend also reported in a previous study,21 where it was suggested that the organic compounds were strongly binding to the clay matrix bounds, reducing extractability and increasing variability. It is important to validate extraction efficiencies using certified reference materials as the recoveries obtained with spiked compounds may not be representative of those obtained with native compounds. This is because spiked analytes are usually lightly coated on the surface of the matrix, whereas native compounds can be strongly absorbed to the soil matrix. Although important, to the best of our knowledge, very few certified commercial data sets are available within the open literature. Extraction of certified reference matrix RTC CRMPR 9583 gave good precision (Table 1). Here a RSD of 3% (Table 1) was achieved, which is a better degree of precision than that achieved by Sanz-Landaluze et al.14 of 14% when validating their method using a reference material. These results are in good agreement with those of Banjoo and Nelson,13 where extraction efficiency greater than 90% for optimized sonication method was obtained. Differences between the method used in our study and those of Banjoo and Nelson13 occur in extraction duration, solvent volume, and addition sequence. Additionally, samples were not agitated in the Banjoo and Nelson method. In the present study, agitation was used to ensure full contact between sample and solvent increasing extraction efficiency. Banjoo and Nelson13 used evaporative techniques in their postextraction sample preparation. Here, no such step is required, which may first minimize loss of volatiles and second account for some of the differences in extraction efficiencies of these two methods, as losses may have occurred during extract concentration steps. While chemical analysis used in this work did not quantify directly the potential loss of volatiles compounds of the front end bands ranging between C8 and C12, some of our current work undertaken on kerosene extraction from soils suggests that retention of the lighter end compounds is very good (i.e., recovery (21) Shin, H-S.; Kwon, O.-S. Bull. Korean Chem. Soc. 2000, 21, 1101–1105.

Analytical Chemistry, Vol. 80, No. 18, September 15, 2008

7093

Table 2. DRO, KRO. and MRO Hydrocarbon Banding Mean Concentration Extracted, Precision, and Bias for Soils Spiked with 20 and 80% of Typical Hydrocarbon Concentration Found in Environmental Samples silty soil carbon band DRO C10-C24 KRO C8-C14 MRO C22-C34 a

clay soil

sandy soil

madeground

% spike levela

mean, mg kg-1

precision, mg kg-1 (% RSD)

mean, mg kg-1

precision, mg kg-1

mean, mg kg-1

precision, mg kg-1

mean, mg kg-1

precision, mg kg-1

20 80 20 80 20 80

8401 18957 1663 4963 3933 11415

1092 (13) 568 (3) 66 (4) 198 (4) 158 (4) 342 (3)

6574 19801 1641 4986 3708 10862

526 (8) 1584 (8) 148 (9) 499 (10) 334 (9) 760 (7)

6468 19135 1639 5005 3527 10451

323 (5) 382 (2) 82 (5) 150 (3) 141 (4) 523 (5)

7052 20677 1724 5201 4123 11513

212 (3) 1034 (5) 69 (4) 364 (7) 124 (3) 461 (4)

Percentage of expected concentration range encountered in environmental samples.

Table 3. Texas Risk Bandings Mean Concentration Extracted, Precision, and Bias for Soils Spiked with 20 and 80% of Typical Hydrocarbon Concentration Found in Environmental Samples silty soil

clay soil

sandy soil

madeground

carbon band

% spike level*

mean, mg kg-1

precision, mg kg-1 (% RSD)

mean, mg kg-1

precision, mg kg-1

mean, mg kg-1

precision, mg kg-1

mean, mg kg-1

precision, mg kg-1

TPH C8-C10

20 80 20 80 20 80 20 80 20 80

224 669 531 1593 2317 7173 2363 6783 4304 12143

11 (5) 20 (3) 21 (4) 48 (3) 116 (5) 287 (4) 47 (2) 271 (4) 215 (5) 364 (3)

234 673 550 1697 2383 7615 2523 7352 4173 12100

30 (13) 47 (7) 33 (6) 102 (6) 167 (7) 533 (7) 177 (7) 515 (7) 334 (8) 847 (7)

231 655 536 1633 2328 7340 2445 6919 3880 11461

28 (12) 33 (5) 27 (5) 33 (2) 93 (4) 147 (2) 73 (3) 138 (2) 155 (4) 458 (4)

228 642 556 1690 2502 7832 2653 7531 4471 12678

23 (10) 39 (6) 22 (4) 152 (6) 125 (5) 548 (7) 80 (3) 301 (4) 134 (3) 380 (3)

TPH C10-C12 TPH C12-C16 TPH C16-C21 TPH C21-C35

>70%; data not shown). Further research will need to be performed with light hydrocarbons such as isooctane, nonane, or decane to test this out. As also shown in Table 1, certified and some spiked compounds gave recovery of >100% and higher recovery efficiency was observed with clay soil rather than sandy soil. The reasons for the observed over-recoveries are quite common with reference compound recoveries and may be associated with inherent integration error with unresolved complex mixtures at the baseline or possible coelution of analytes.19,27 In addition, calibration variance observed in analytical systems particularly where bias is small may contribute to this observation. Ensuring reliability and validity of the results is an important consideration. Quality control is built into this method to allow for continued evaluation and validation of the method, through the analysis of a reagent blank and a spiked reference material with every e20 samples. Additionally, sample recovery is monitored using surrogate spikes. Analyses of KRO, DRO, MRO, and Texas risk carbon banding fractions show RSD values ranging between 2 and 13% with the highest overall degree of variability occurring when analyzing the lightest fractions including DRO C10-C24 ranges and TEXAS1-TPH C8-C10 ranges (Tables 2 and 3). With the exception of the madeground soil, a higher degree of precision was typically observed when extracting higher concentrations of hydrocarbons. It is only when analyzing the aromatic fraction (Table 4) where RSD values rise above 20%. The greatest degree of variability was observed when extracting low concentrations of the C8-C10 range for all of the soils tested, (22) Coulon, F.; Pelletier, E.; St Louis, R.; Gourhant, L.; Delille, D. Environ. Toxicol. Chem. 2004, 23, 1893–1901.

7094

Analytical Chemistry, Vol. 80, No. 18, September 15, 2008

possibly due to volatile losses or thermal decomposition of compounds. Here, the diminution curve showed that precision was lower due to the detection limits of the GC used (data not shown). This is consistent with the precision obtained for ultrasonic methods elsewhere. Shin and Kwon21 and SanzLandaluze et al.14 demonstrated RSD values of C8fC10b >C10fC12 >C12fC16 >C16fC21 >C21-C35

mean, mg kg-1 160 83 31 77 10 10 21 22 77

precision, mg kg-1 (% RSD) 29 (18) 12 (15) 5 (15) 13 (17) 2(16) 2 (15) 3 (16) 4 (18) 14 (18)

bias mg kg-1 100

aromatic hydrocarbons bias % 0.1

mean mg kg-1 166 84 12 89

precision, mg kg-1 (% RSD) 27 (16) 14 (16) 2 (14) 15 (17)

8 29 33 94

2 (22) 6 (20) 5 (16) 16 (17)

bias, mg kg-1 104

bias, % 4

a Certified reference materials for aliphatic and aromatic fractions are not available consequently the spiked matrix data were used to determine the accuracy and precision of the method. The level of spike used throughout was 160 mg kg-1. b PAH mixed used to spike the soil contained no aromatic PAH >C8-C10 fraction

extraction.10,23 Here, the extraction of field moist samples can suffer interference from water when hydrophobic solvents (hexane and DCM) are used, reducing extraction efficiencies. Within the method described in this paper, acetone, a hydrophilic solvent, allows penetration and extraction of contaminants from field moist samples, simultaneously disrupting the soil matrix and enhancing extraction rates. This, followed by the addition of hexane, enables the extraction of nonpolar compounds. This initial use of a polar solvent negates the need for oven or freeze-drying, which has been shown to reduce extraction efficiencies,13,24,25 particularly for the lower molecular weight equivalent carbon fractions. The effectiveness of this step in overcoming this interference from water has been reported elsewhere.13,24 The effectiveness of polar solvents such as acetone was also shown by Schwab et al.,22 when investigating the effect of solvent, soil type, extraction cycles, soil quantity, and aging on the efficiency of mechanical shaking. The authors found that soil moisture played a key role and of the solvents studied acetone was the least effected by soil moisture and type. Banjoo and Nelson13 optimized an ultrasonic extraction procedure for the determination of PAHs and aliphatic hydrocarbons ranging from C12 to C24, in sediments. This was compared against a reflux with the methanolic KOH method. The investigation showed that ultrasonic extraction of dried sediment with acetone/ hexane mix (1:1) gave a concentration of the PAHs studied comparable to the reflux method, with lower variation in the reproducibility. The advantage of using acetone prior to the addition of other solvents was highlighted. The authors found that extraction efficiency increased when samples were initially sonicated with acetone only prior to addition of hexane. Additionally, no increase in efficiency was seen upon increasing extraction duration from 30 to 60 min, and the optimum number of extraction cycles was four. (23) Schwab, A. P.; Su, J.; Wetzel, S.; Pekarek, S.; Banks, M. K. Environ. Sci. Technol. 1999, 33, 1940–1945. (24) Bergknut, M.; Kitti, A.; Lundstedt, S.; Tysklind, M.; Haglund, P. Environ. Toxicol. Chem. 2004, 23, 1861–1866. (25) CCME. Reference methods for the Canada-wide standard for petroleum hydrocarbons in soil-Tier 1 Method. Publication No. 1310, CCME, Winnipeg, Manitoba, Canada, 2001. (26) MADEP. Method for the determination of extractable petroleum hydrocarbons, Revision 1.1, Massachusetts Department of Environmental Protection, Executive Office of Environmental Affairs, Commonwealth of Massachusetts, Boston, MA, 2004.

The effect of solvent choice has also been demonstrated by Shin and Kwon21 and Sanz-Landaluze et al.,14 where acetone/DCM (1:1, v/v) and hexane, respectively, were shown to be the solvents of choice during sonication. Bergknut et al.24 compared pressurized liquid extraction (PLE) to Soxhlet extraction and also assessed the effects of other parameters of the PLE process. The effects of different organic solvents, pretreatment, and extraction time on the availability of PAHs extracted by PLE was assessed by sequentially extracting soil in water, methanol, 1-butanol, acetone, n-hexane, and toluene by PLE. Here the sample was extracted using the solvents one after the other. PLE extraction using methanol demonstrated equivalent extraction efficiency in 14 min as a 2-h toluene Soxhlet extraction. The studies highlighted the different solvents’ affinity for different molecular weight PAHs. Toluene and acetone extracted roughly even concentrations of all PAHs tested in the study comprising 2-, 3-, 4-, and 5-fused rings, whereas 1-butanol extracted higher concentrations of high molecular weight PAHs (>4- and 5-fused rings). Many of the methods currently in use for the analysis of weathered petroleum hydrocarbon-contaminated soils incorporate, where necessary, a sample cleanup method.1,25,26 In some cases, methods will, in the same step, fractionate a sample into aliphatic and aromatic fractions.1,26 Risk assessments are increasingly evaluating aliphatic and aromatic compounds separately; therefore, it is important that this is incorporated into methods not only to satisfy the risk assessment requirements but to ensure the results are not skewed by the presence of biogenic TPH. The method described here uses a microscale silica gel column chromatography method to fractionate extracts into aromatic and aliphatic fractions and where required remove interfering polar compounds. Comparison against other available methods is helpful in evaluating the applicability of a method to the extraction of weathered hydrocarbons in soil on a broader analytical scale, enabling its position within the group of weathered hydrocarbon soil extraction methods to be defined. Alternative methods have been demonstrated to give a range of efficiencies from 75 and 77% for supercritical fluid extraction15 and Soxhlet,27 respectively, up to g97 and 99% for accelerated solvent extraction15 and microwave-assisted extraction,27 respectively. In this study, a (27) Saifuddin, N.; Chua, K. H. Malaysian J. Chem. 2003, 5, 30–33.

Analytical Chemistry, Vol. 80, No. 18, September 15, 2008

7095

minimum recovery of g95% has been demonstrated with maximum recoveries in the range >99%, easily positioning this methods within the best of the alternatives to Soxhlet methods. The method is compatible with the current risk assessment frameworks, allowing for carbon banding and class fractionation. The method is faster than a typical Soxhlet method, allowing 24 (or more depending upon size of centrifuge and sonic bath) samples to be extracted within ∼70 min. If no cleanup/ fractionation step is required, extracts can be directly analyzed by GC/MS without further evaporation. Soxhlet methods can take between 2 and 24 h (depending on the protocol), often produce relatively dirty samples requiring sample cleanup, and often require sample evaporation. Soxhlet also requires more glassware that is fragile and comparatively expensive to that used in sonication. Soxhlet extraction protocols can require up to 250 mL of extracting solvent, whereas the sonication method described here uses 40 mL. Lower solvent consumption not only reduces costs but reduces the environmental impact of their subsequent disposal and health risks to the operator. Other authors have had success with adapted sonication techniques. In comparing USEPA Methods 3540 (Soxhlet) and 3550 (Sonication), for example, Ilias28 presented a rapid fuel isolation, identification, and quantitation method for TPH in soils, reportedly one-third of the cost of conventional Soxhlet techniques. Our work reinforces this observation as the cycle time of the sequential ultrasonic method presented here was good as most the apparatus was used once and disposable, keeping capital outlay low and minimizing crosscontamination. CONCLUSIONS This sequential ultrasonic extraction method for contaminated soils allows high efficiency for several levels of analysis, TPH (28) Ilias, A. M. Fuel isolation, identification and quantitation in soils, In Analysis of soils contaminated with petroleum constituents; O’Shay, T. A., Hoddinott, K. B., Eds.; ASTM STP 1221; American Society of Testing for Materials: Philadelphia, 1994; pp 12-26.

7096

Analytical Chemistry, Vol. 80, No. 18, September 15, 2008

concentration ranges, various hydrocarbon banding, aliphatic and aromatic fractions (with and without hydrocarbon banding), and specific target compound identification. The extraction method is the same for all the different levels of analytical detail achievable; here the use of column chromatography and GC-FID and GC/ MS integration techniques allows for the additional data to be easily extrapolated. Thus, this method meets the needs of the current risk assessments while also enabling the provision of data important for bioremediation studies. We believe that the ultrasonic solvent extraction method described here is an improvement on conventional methods, as it saves time and lowers solvent consumption. Further, there are no evaporation steps preventing the potential loss of front end hydrocarbon bands, and the use of water partitioning facilitates an effective solvent exchange prior to fractionation. The method involves fewer handling steps, and disposable apparatus eliminates potential cross-contamination. Importantly, it is easy and simple to reproduce. The method is also environmentally friendly, due to its reduced chlorinated solvent usage. The method has shown a good potential for implementation as a standard method, potentially capable of providing (through use) further insight and knowledge to the contaminated land sector.

ACKNOWLEDGMENT This research and F.C. were funded by a Department for Business Enterprise and Regulatory Reform/BBSRC/Environment Agency Grant (BIOREM_35). K.J.B. was funded by an EPSRC CASE Award supported by the FIRSTFARADAY partnership.

Received for review April 8, 2008. Accepted July 11, 2008. AC800698G