DEVELOPMENT OF A SENSITIVE THERMAL DESORPTION GC-MS METHOD USING SELECTIVE ION MONITORING FOR A WIDE RANGE OF VOCS C Jia *, S Batterman and S Chernyak Department of Environmental Health Sciences, University of Michigan 1420 Washington Heights, Ann Arbor, MI 48109-2029 ABSTRACT Thermal desorption (TD) of volatile organic compounds (VOCs) collected on adsorbent tubes is increasingly used in IAQ applications. This paper develops and evaluates the use of a TD-GC-MS method using selective ion monitoring () that can significantly improve method detection limits (MDLs). Two protocols were developed for 99 target compounds: a full scan method with ions ranging from 26 to 250 amu, and a method using 16 windows over the 37.15 min GC run time. In the laboratory, calibration curves and MDLs were determined for the target compounds using both modes. In the field, duplicate samples were taken inside and outside of 38 residences, and analyzed using both scan and modes. Calibration curves for scan and modes were very similar, allowing a single calibration for both modes. MDLs in mode were 1.5 to 6 times lower than those in the scan mode and often well below 0.01 μg m -3. In the field, many more VOCs were detected and quantified using mode. The new method offers greater sensitivity and maintains the high selectivity of the traditional GC-MS method. INDEX TERMS Adsorbents, Mass spectrometry, Sampling, Volatile organic compounds, VOCs. INTRODUCTION Adsorbent sampling followed by thermal desorption, gas chromatography and mass spectrometry (TD-GC-MS) has been extensively used for VOC analysis (Peng and Batterman 2000, Woolfenden 1997, Heavner et al. 1992). The MS detector is considered to provide definitive identification of compounds since the ion spectrum obtained by the MS can often resolve co-eluting pollutants. Co-elution is a significant problem in environmental sampling using less specific detectors, e.g., flame ionization detectors (FID) and electron capture detectors (ECD). Compounds with similar structures often co-elute, e.g., isomers such as p-xylene and m-xylene. Sometimes compounds with very different structures co-elute, e.g., benzene and carbon tetrachloride. Either case can result in misidentification and incorrect quantification using non-specific detectors. Non-specific detectors have the advantage that they can often provide much higher sensitivity than MS detectors, e.g., FID have high sensitivity to many hydrocarbons and a wide linear range. Thus, GC-MS is sometimes coupled to FID, with the FID used for quantification (Na et al. 2004, Mohamed et al. 2002, Jurvelin et al. 2001, Brickus et al. 1998, Cheng et al. 1997). Other configurations include dual-column tandem GC/ECD-FID (Begerow et al. 1996), or GC/FID-RIM (retention index monitoring), again with the purpose of confirming the identification of compounds and simultaneously obtaining high sensitivity. These methods have been used in a variety of settings. For example, Fernandez-Martinez et al. (1999) developed a method for 15 VOCs using GC-FID and GC-MS, and obtained MDLs for 4 L samples from 0.0004 to 0.030 μg m -3 (GC-FID), 0.001 to 0.008 μg m -3 (GC-MS, mode), and 0.004 to 0.08 μg m -3 (GC-MS, SCAN mode). However, multi-detector approaches require specialized equipment, are relatively costly, and calibrations and analyses can be time consuming. A mass spectrometer has two modes: scan and selective ion monitoring (). In scan mode, the MS scans every ion over a specified (and typically wide) range. In mode, only specific ions are measured. mode is routinely applied to identify compounds in environmental samples since many compounds may be present and the ion spectrum can be used to confirm the identification of compounds. However, scan mode may not provide sufficient sensitivity for many important environmental contaminants. MS sensitivity can be significantly increased using the selective ion monitoring () mode, potentially yielding * Corresponding author email: chunrong@umich.edu 2194
sensitivities that are comparable to non-specific detectors like FIDs (Message 1984, Chen 1979). Within a specified ion interval, the sensitivity can be increased significantly, for example, the Agilent 5973 inert GC/MS system has electronic impact (EI) scan sensitivity of 60:1 s/n (signal to noise) for 1 pg OFN (octofluoronaphthalene) scanning from 50 300 amu at nominal m/z 272 ion, and EI sensitivity of 10:1 s/n for 0.020 pg OFN at nominal m/z 272 ion (Agilent). Adjusted to the same s/n, the mode sensitivity increase is about 8 times. MS in mode has been occasionally used for VOC analyses. Almasi et al. (1993) detected 41 VOCs with concentration range from 0.8 to 80 μg m -3 using small sampling volumes, i.e., 0.06 L. Heavner et al. (1992) used a multisorbent TD-GC-MS method for 28 target VOCs and obtained LODs from 0.03 to 0.8 μg m -3 for 4 L samples. Otson et al. (1994) collected passive samples in 757 Canadian residences and analyzed 52 target VOCs using a method, but samples were desorbed using solvent extraction, which is less sensitive than thermal desorption. The goal of this study is to develop and evaluate a method for VOCs measured in indoor and ambient air. The method is evaluated using laboratory and field tests. EXPERIMENTAL Thermal desorption tubes (TDTs) The TDTs have been previously described by Peng and Batterman (2000). Each stainless steel tube is packed with two sorbents (Tenax GR followed by Carbosieve SIII) for active sampling, or with a single sorbent (Tenax GR) for passive sampling since VOCs cannot reach the second sorbent bed by diffusion. Target compounds Ninety-nine VOCs were selected based on their toxicity and occurrence in indoor and outdoor air. Many of these VOCs are listed in the US EPA TO-14 (US EPA 1999A) and TO-15 (US EPA 1997) methods, and include aromatics, halogenated compounds, terpenes, alkanes, etc. This paper describes results for six of these VOCs in two groups, selected on the basis of their occurrence and concentration in indoor and outdoor environments in a previous study (Jia et al. 2004). VOCs in the first group are frequently detected at relatively high concentrations: toluene, limonene and 1,4-dichlorobenzene. VOCs in the second group are found at low concentrations and are detected less frequently: chloroform, styrene and heptadecane. The six VOCs were also selected to represent aromatic, terpene, chlorinated and aliphatic compounds. Laboratory performance evaluation Seven point calibration curves were obtained in scan and modes using an Agilent 6890/5973 running Chemstation (G1701BA, Version B.01, Hewlett-Packard, Palo Alto, CA, USA) at concentrations from 0.15 to 50.0 ng μl -1, spanning the range of interest to determine instrument sensitivity and linearity. Calibrations used 2 μl aliquots of liquid standards and duplicate TDTs at each concentration. Method detection limits (MDLs) were determined using low concentration standard solutions that contained 0.015 and 0.005 ng μl -1 of each compound for scan and modes, respectively. Each tube was loaded with 2 μl of these standards, thus obtaining injection masses of 0.03 and 0.01 ng tube -1 for scan and modes, respectively. The MDL was calculated as 3.14 S where S = the standard deviation of 7 replicate analyses, and 3.14 = Student s t-value for 99% confidence level (US EPA 1999B). Field study The field study was conducted in Ann Arbor, MI, USA. Duplicate passive samplers were placed inside and outside of houses for 3-4 weekdays during August and September, 2004. The estimated sampling volume was ~2 L. Collected samples were analyzed using scan and modes. Overall, 38 pairs of indoor measurements and 38 pairs of outdoor measurements were collected. Of these, one indoor and one outdoor sample failed quality assurance/quality control requirements, so comparisons omit these two sample pairs. Sample analysis and quality control The GC-MS scan method and quality control procedure is described elsewhere (Peng and Batterman 2000). In mode, we applied the same GC conditions, but divided the target compounds into 16 groups, based on their retention times, and assigned from 4 to 19 ions to each time window in the 37.15 min GC run. Several compounds were included in two or even three adjacent windows due to the possibility of shifting peaks during the run. A target ion (for both quantification and confirmation) and one or two qualifier ions (for confirmation) were selected for each compound. Totally 79 different ions (from 29 to 260 m/z) were monitored over the whole run. 2195
RESULTS Laboratory performance For each mode, the precisions for duplicate tubes across the concentrations used in the calibration were mostly below 10%, and were below 20% at the lowest concentrations (below 0.5 ng μl -1 ). Calibration curves were based on linear regressions with intercepts set to 0 in order to increase the sensitivity. R 2 values were high, showing good linearity over the concentration range tested. and mode calibrations showed good agreement. Based on a sample volume of 4 L used in routine analysis (US EPA 1999B), MDLs for target compounds were below 0.025 μg m -3 in the scan mode and below 0.015 μg m -3 in mode (Table 1). -mode sensitivity was 1.5 to 5.7 times better than that for scan mode. This increase was lower than expected, in part because we analyzed 100 compounds in each run, and most window included 10 compounds. With fewer target compounds, sensitivity will increase significantly. However, the increased sensitivity obtained here was still valuable as the mode detected ions that were not found by scan mode in the low concentration samples, as seen in the field study described next. Table 1. Calibration curves and MDLs measured in scan and modes. Compound MDL Slope R 2 MDL (μg ( 礸 m -3 ) Slope R 2 MDL (μg ( 礸 m -3 ) / Toluene 1.2164 0.9999 0.020 1.2013 00 0.013 1.52 1,4-Dichlorobenzene 2.0475 0.9999 0.010 2.0098 0.9996 0.002 4.97 d-limonene 0.9265 0.9985 0.021 0.7646 0.9948 0.007 2.97 Chloroform 0.3572 0.9993 0.023 0.3615 0.9994 0.006 3.74 Styrene 2.4096 0.9998 0.017 1.9848 0.9997 0.003 5.66 n-hexadecane 1.3596 0.9936 0.013 0.9421 0.9816 0.006 2.32 Field study The six VOCs were detected in over 90% of the homes sampled. In ambient air, toluene, d-limonene and chloroform were frequently detected. 1,4-dichlorobenzene and styrene were less frequently detected. n-hexadecane was not detected in ambient air. In the first group of VOCs, indoor concentrations of toluene and d-limonene frequently exceeded 10 μg m -3 (Table 2). The large standard deviations reflected the differences between homes. Outdoor concentrations were much lower. Concentrations of VOCs in the second group were low indoors (<0.5 μg m -3 ) and even lower outdoors (<0.1 μg m -3 ). As expected, standard deviations of outdoor concentrations for all VOCs were small, reflecting the limited range of concentrations found. Detection frequencies of VOCs in scan and modes were computed after pooling the indoor and outdoor data (Table 2). Four VOCs were detected more frequently using mode, e.g., chloroform was detected in 19% of the samples in the scan mode, but 97% in the mode. Detection frequencies for toluene and n-hexadecane were unchanged. Table 2. Concentrations and detection probabilities for 6 compounds. Indoor Outdoor Sim Compound Mean Median St. Dev. Detected Mean St. Dev. Median Detected Detected Detected (µg m -3 ) (µg m -3 ) (µg m -3 ) (%) (µg m -3 ) (µg m -3 ) (µg m -3 ) (%) (%) (%) Toluene 15.61 4.80 30.49 100 1.14 1.51 0.74 100 100 100 d-limonene 10.63 4.96 16.09 100 0.15 0.12 0.11 97 57 100 1,4-Dichlorobenzene 1.29 0.16 3.65 95 0.03 0.04 0.03 65 32 82 Chloroform 0.22 0.13 0.33 92 0.06 0.02 0.06 95 19 97 Styrene 0.28 0.16 0.36 100 0.02 0.04 0.01 78 50 90 n-hexadecane 0.28 0.21 0.16 100 0.00 0.00 0.00 0 46 50 Figure 1 shows typical scan and chromatograms for ion 68 representing d-limonene. The mode peak is higher and smoother than that in the scan mode. The increased sensitivity is derived from the abundance of the 2196
quantification ions. The abundances of ion 68 are 5063 and 7342 in scan and modes, respectively, an increase of ~1.5 times. Since abundances of internal standards increase proportionally, one calibration still works. : Abundance 21.50 22.00 22.50 23.00 Retention time : Abundance 21.50 22.00 22.50 23.00 Retention time Figure 1. Target ions of d-limonene in scan and modes. Concentrations in both modes were calculated using the same calibration curves, plotted and fitted using linear regression (Figure 2). Concentrations were within 10% for the two modes for toluene, d-limonene, 1,4-dichlorobenzene and styrene. The small differences might result from sampling, analysis and calibration errors. For chloroform, mode measurements were 19% lower than scan mode determinations, though the correlation was high (R 2 =0.97). Most chloroform concentrations were <1 μg m -3. and scan mode measurements for n-hexadecane showed good agreement, though lower correlation (R 2 =0.51). Again, this might result from the low concentrations (<0.5 μg m -3 ) encountered. 1000.00 100.00 10.00 n = 72 y = 0.9784x R 2 = 0.9593 Toluene 100.00 10.00 n = 41 y = 0.9244x R 2 = 0.9570 d-limonene 0.10 0.10 0.10 10.00 100.00 1000.00 0.01 0.01 0.10 10.00 100.00 100.00 10.00 n = 23 y = 0.9124x R 2 = 0.9972 1,4-Dichlorobenzene 2.00 n = 14 1.50 y = 0.8072x R 2 = 0.9743 Chloroform 0.10 0.50 0.01 0.01 0.10 10.00 100.00 0.00 0.00 0.50 1.50 2.00 2.50 Styrene 2.50 2.00 1.50 n = 36 y = 0.9674x R 2 = 0.9525 0.50 0.00 0.00 0.50 1.50 2.00 2.50 1.40 1.20 0.80 0.60 0.40 0.20 0.00 n = 33 y = 1.0276x R 2 = 0.5145 n-hexadecane 0.00 0.20 0.40 0.60 0.80 Figure 2. Concentrations of 6 compounds measured by scan and modes. 2197
DISCUSSION While the laboratory-based determination of -mode MDLs showed modest (1.5 to 5.7-fold) improvements compared to scan mode, -mode analyses significantly increased the detection frequency of VOCs in the field study. A scan-mode chromatogram reflects the many substances found in complex environmental samples. mode reduces noise and achieves better signal-to-noise ratios, resulting in clearer peaks that are easier to identify and quantify. Thus, in practice, mode detects about 50% more compounds than the traditional scan mode, including compounds that are rarely detected in scan mode analyses, e.g., 1,1,1-trichloroethane, trichloroethylene and phenol. and scan mode calibration curves did not change significantly, e.g., calibration curves were within 15%, thus allowing a single and convenient calibration for both modes. It is important to measure even low concentration VOCs. Some VOCs may affect human health even at low levels. Also, more complete detection provides information that can be used to identify sources. The mode may also facilitate the detection of compounds that are masked by the solvent from the internal standard solution, e.g., 1,3-butadiene, a very reactive and toxic compound, as well as methanol and other VOCs. The method offers greater sensitivity while maintaining the high selectivity of traditional MS detector. The new method appears to provide a viable option for analyzing environmental samples. ACKNOWLEDGEMENTS Portions of the study were financially supported by the American Chemistry Council (Grant 2401) and the Michigan Education and Research Center which is funded by the National Institute of Occupational Safety and Health (Grant T42 CC5410428). REFERENCES Agilent 5973 Inert GC/MS Specifications, http://www.lqa.com/data/pdf_specifications_products/5973inert_spec.pdf Almasi E., Kirshen N. and Kern H. 1993. The determination of sub part-per-billion levels of volatile organic-compounds in air by preconcentration from small sample volumes, International Journal of Environmental Analytical Chemistry. 52 (1-4): 39-48. Begerow J., Jermann E., Keles T., Koch T. and Dunemann L. 1996. Screening method for the determination of 28 volatile organic compounds in indoor and outdoor air at environmental concentrations using dual-column capillary gas chromatography with tandem electron-capture-flame ionization detection, Journal of Chromatography A. 749 (1): 181-191. Chen ECM. 1979. Selective ion monitoring, Practical Mass Spectrometry. Edited by Middleditch B.S. Plenum Press, New York. Cheng L., Fu L., Angle RP. and Sandhu HS. 1997. Seasonal variations of volatile organic compounds in Edmonton, Alberta, Atmospheric Environment. 31 (2): 239-246. Fernandez-Martinez G.; Lopez-Mahia P.; Muniategi-Lorenzo S. et al., 1999. Development of a method for determination of volatile organic compounds (C6-C9) by thermal desorption-gas chromatography: application to urban and rural atmospheres, Analytical Letters. 32 (14): 2851-2870. Heavner DL., Ogden MW. and Nelson PR. 1992. Multisorbent thermal desorption/gas chromatography/mass selective detection method for the determination of target volatile organic compounds in indoor air, Environ Science & Techn. 26 (9):1737-46. Jia C., Godwin C., Batterman S. and Alfred F. 2004. Sources and significance of VOC exposures in three microenvironments: schools, residences and vehicles, Air & Waste Management Association Conference, Indianapolis, IN, USA, June 2004. Jurvelin J., Edwards R., Saarela K., Laine-Ylijoki J., et al. 2001. Evaluation of VOC measurements in the EXPOLIS study, Journal of Environmental Monitoring. 3(1):159-65. Jurvelin J., Edwards R., Saarela K., Laine-Ylijoki J., et al. 1998. Distributions of indoor and outdoor air pollutants in Rio de Janeiro, Brazil: implications to indoor air quality in Bayside offices, Environmental Science and Technology. 32 (22): 3485-3490. Message GM. 1984. Practical Aspects of Gas Chromatography/Mass Spectrometry. John Wiley & Sons, Inc. pp. 218. Mohamed M., Kang D. and Aneja V. 2002. Volatile organic compounds in some urban locations in United States, Chemosphere. 47 (8): 863-882. Na K., Kim YP., Moon I. and Moon KC. 2004. Chemical composition of major VOC emission sources in the Seoul atmosphere, Chemosphere. 55 (4): 585 594. Otson R., Fellin P. and Tran Q. 1994. VOCs in representative Canadian residences, Atmospheric Environment, 28 (22): 3563-3569. 2198
Peng C. and Batterman S. 2000. Performance evaluation of a sorbent tube sampling method using short path thermal desorption for volatile organic compounds, Journal of Environmental Monitoring. 2 (4):313-324. US EPA. 1997. Compendium of Methods for the Determination of Toxic Organic Compounds in Ambient Air, TO-15, U. S. Environmental Protection Agency, Washington, DC, 2nd ed., Report No. EPA/625/R-96/010b. US EPA. 1999A. Compendium Method TO-14A. Determination of volatile organic compounds in ambient air using specially prepared canisters with subsequent analysis by gas chromatograpy. January 1999. US EPA. 1999B. Compendium Method TO-17. Determination of volatile organic compounds in ambient air using active sampling onto sorbent tubes. January 1999. Woolfenden E. 1997. Monitoring VOCs in air using sorbent tubes followed by thermal desorption capillary GC analysis: Summary of data and practical guidelines, Journal of the Air & Waste Management Association. 47 (1):20-36. 2199