Simplified Analysis of Chlorinated Ethenes in Water Samples by Gas Chromatography with Flame Ionization Detection

Abstract

Chlorinated ethenes are among the most common environmental contaminants and are known or suspected carcinogens. This class of compounds includes perchloroethene (PCE), trichloroethene (TCE), and their breakdown products, including dichloroethene (DCE) isomers and vinyl chloride (VC). Engineers and scientists must be able to measure concentrations of these chemicals in water samples to assess site contamination, monitor clean-up progress, and test possible remediation technologies. Gas chromatography with flame ionization detection (GC/FID) is a common method for measuring these contaminants in environmental samples. In this study, we tested the hypothesis that FID response factors are equal for all chlorinated ethene compounds. The rationale for the investigation was that if the hypothesis is correct, a single calibration curve can be used for GC/FID analysis of all chlorinated ethene compounds, saving time and money during sample analysis. Based on our measurements, a single calibration curve (FID response versus mass of analyte injected) is applicable to analysis of PCE, TCE, and all three DCE isomers (r²=0.990, n=50 measurements), allowing for simplified quantification of those chemicals. However, the apparent FID response factor for VC was lower by ∼40%, indicating that a separate calibration curve would need to be used to accurately estimate the VC concentration in water samples. The difference in the apparent VC response factor is caused predominantly by losses of VC to volatilization during the analysis.

Introduction

Chlorinated ethenes are a class of compounds found at numerous contaminated sites in the United States (National Research Council, 1994). This class of compounds includes perchloroethene (PCE), trichloroethene (TCE), dichloroethene (DCE) isomers, and vinyl chloride (VC). Chlorinated ethenes are a risk to human health, because they are generally persistent in the environment and are known or suspected carcinogens (Wartenberg et al., 2000). Based on their frequent occurrence and health implications, these contaminants are the focus of significant remediation and environmental research.

To analyze environmental samples containing these contaminants, analytical methods for detecting the contaminants must be developed. Chlorinated ethenes are volatile or semivolatile, and thus good candidates for analysis by gas chromatography. Gas chromatography with flame ionization detection (GC/FID) is a common method for quantifying the concentration of these chemicals in environmental samples, including water samples (Otson and Williams, 1982; Ketola et al., 1997). When analyzing environmental samples, calibration curves for target chemicals must be created to quantify concentrations in samples of unknown concentration. Thus, in general, a scientist or engineer wishing to analyze chlorinated ethenes would need to develop six separate calibration curves, one for each chemical (PCE, TCE, VC, and three DCE isomers).

Numerous studies have shown that FID response factors for hydrocarbons are linearly proportional to the number of carbon atoms present (Ettre, 1962; Perkins et al., 1962; Sternberg et al., 1962; Leveque, 1967; Tong and Karasek, 1984; Slemr et al., 2004). Because all chlorinated ethenes have the same number of carbon atoms, it is therefore possible that all chlorinated ethenes would exhibit the same FID response factor. This is significant because, if validated, it may allow us to simplify the analysis of chlorinated ethenes (and/or other chlorinated hydrocarbons) in water samples. If the FID response factor for all chlorinated ethenes is the same, then one calibration curve could be used for all target analytes during analysis of samples with unknown concentration. Instead of creating six separate calibration curves, only one would be required, resulting in a savings of time and money during sample analysis.

However, the different chlorinated ethenes have different numbers of chlorine atoms, and the effect of the degree of chlorine substitution on the FID response factor is unknown. On one hand, some studies have considered the effect of heteroatoms and functional groups on the response factor and have shown that the effective carbon number for these substituted compounds may be different than the unsubstituted hydrocarbon (Perkins et al., 1962, 1963; Mockel, 1976; Jorgensen et al., 1990; Huang et al., 1990; Kallai et al., 2001). In other words, a hydrocarbon containing heteroatoms or functional groups may exhibit a different FID response factor than the unsubstituted hydrocarbon. On the other hand, it has also been observed that the presence of a single halogen atom has limited impact on FID response (Sternberg et al., 1962). Therefore, at the present time, it remains unclear if all chlorinated ethenes are likely to exhibit the same FID response factor. To the best of our knowledge, this has not yet been determined.

Therefore, the objective of this work was to test the hypothesis that all chlorinated ethenes have the same FID response factor. The implication of this hypothesis is that a single calibration curve could be used for GC/FID analysis of all chlorinated ethenes. The rationale for this study is that, if verified, a calibration curve could be generated for a single chemical and then used for all chemicals in the family, thereby reducing the need for separate calibration curves, and reducing time and cost of analysis.

Experimental Protocols

Chemicals and reagents

The chlorinated ethenes utilized in this research are listed in Table 1. VC (1000 μg/mL, from SPEX CertiPrep) and all DCE isomers (2000 μg/mL, from Restek) were obtained as analytical standards in methanol. Trichloroethene (99.5%, Fisher Scientific), tetrachloroethene (99%, Acros Organics), and methanol (99.9%, Fisher Scientific) were obtained as pure-phase liquids. Deionized water was produced from a Barnstead water purification system.

Table 1.

Properties of Chlorinated Ethenes Used in This Study

IUPAC name	Common name	Abbreviation	Molecular weight (g/mole)	Henry's constant at 21.7°C
Tetrachloroethene	Perchloroethylene	PCE	165.83	0.580
Trichloroethene	Trichloroethylene	TCE	131.39	0.338
cis-1,2-Dichloroethene	cis-Dichloroethylene	cis-DCE	96.94	0.150
trans-1,2-Dichlorothene	trans-Dichloroethylene	trans-DCE	96.94	0.386
1,1-Dichloroethene	1,1-Dichloroethylene	1,1-DCE	96.94	1.04
Chloroethene	Vinyl chloride	VC	62.50	0.936

Henry's constants are in dimensionless form and were estimated at 21.7°C using Equation (3) and Table 2 of Staudinger and Roberts (2001).

Table 2.

Concentrations of Chlorinated Ethenes in Aqueous Standard Solutions

Aqueous sample number	PCE conc. (mg/L)	TCE conc. (mg/L)	cis-DCE conc. (mg/L)	trans-DCE conc. (mg/L)	1,1-DCE conc. (mg/L)	VC conc. (mg/L)
1	24	21	7.0			3.5
2	7.9	7.1	2.3			1.2
3	2.6	2.4	0.77			0.39
4	0.87	0.79	0.26			0.13
5	0.29	0.26	0.086			0.043
6				9.9
7				3.3
8				1.1
9				0.37
10				0.12
11					9.9
12					3.3
13					1.1
14					0.37
15					0.12

PCE, perchloroethene; conc., concentration; TCE, trichloroethene; DCE, dichloroethene; VC, vinyl chloride.

Methanol stock solutions

We developed three stock solutions of chlorinated ethenes in methanol. The first stock solution was prepared by adding 1.2 mL of VC in methanol, 1.2 mL of cis-DCE in methanol, 5 μL of neat-phase TCE, 5 μL of neat-phase PCE, and 1 mL of methanol to a 5-mL glass screw-top vial. Separate stock solutions were made for trans-DCE and 1,1-DCE. Each of those was prepared by adding 1.0 mL of DCE in methanol and 1.0 mL of methanol to a 5-mL glass screw-top vial. The rationale for starting with three separate stock solutions was that by analyzing trans-DCE and 1,1-DCE as single solutes (rather than as components in a mixture), we could ensure that no DCE isomers would co-elute from the GC during analysis. All vials were closed and kept in a freezer to minimize volatilization when not in use.

Serial dilution of methanol stock solutions

Serial dilution of the original methanol stock solutions was performed to obtain solutions of target contaminants in methanol at desired concentrations. For each dilution, we added 0.5 mL of parent solution to a glass screw-top vial and then added 1 mL of methanol. The new solution was shaken for 10 min and allowed to sit undisturbed for another 10 min to ensure complete mixing. The resultant solution was then diluted further following the same procedure. This procedure was repeated to create a total of five solutions of known concentration starting from each stock solution. Each daughter solution was one-third the concentration of the parent solution. A total of fifteen methanol solutions were developed (five from each of the three stock solutions).

Preparation of aqueous standards

The methanol solutions described above were used to create aqueous standards of known concentrations. Aqueous standards were made by adding 2 mL of deionized water to a 5-mL (nominal size) glass screw-top vial, and then adding 20 μL of methanol solution containing the chlorinated ethenes. Vials were sealed with an open-hole cap with a polytetrafluoroethylene-lined septum, and then shaken on a shaker table for 30 min.

By using the fifteen different methanol solutions (described above), we produced aqueous standards of fifteen different concentrations. All aqueous standards were created in duplicate; thus, we had a total of thirty aqueous standards, that is, two samples each of fifteen different concentrations. Concentrations of the different aqueous standards are shown in Table 2. The concentrations of chlorinated ethenes ranged from 43 μg/L in the least concentrated to 24 mg/L in the most concentrated standard. This represents a realistic range of concentrations such as what might be found at a contaminated groundwater site.

Analysis of aqueous standards

Standard static headspace analysis was performed on each of the aqueous standards (cf. Dietz and Singley, 1979; Penton, 1992; Ketola et al., 1997). The aqueous standards in the 5-mL screw-top vials were allowed to equilibrate with the headspace in the vial for 12 h at room temperature (21.7°C). After equilibration between the aqueous phase and the headspace, 1 mL of headspace gas was withdrawn by piercing the septum with a gas-tight syringe. Headspace was analyzed by gas chromatography with flame ionization detection. The chromatography was performed on a Perkin-Elmer Clarus 500 GC. The chromatographic column used was RTX-1301 (Restek), 30-m length, 0.53-mm inner diameter, and 3-μm film thickness. The oven temperature in the GC was held at 35°C for 8 min, then increased to 200°C at 20°C/min, and then was held at 200°C for 1 min. The temperature of the FID was 240°C throughout the experiment. The flow rate of helium carrier gas was 3 mL/min. The temperature of the injector was 180°C, and the injector split ratio was 15:1.

Because aqueous standards were created by adding 20 μL of methanol solution to 2 mL of deionized water (as described above), the final samples are ∼0.99% methanol (by volume, neglecting volume change upon mixing). As noted by Schwarzenbach et al. (2003, p. 166), the activity of the chlorinated ethenes in the aqueous phase should not be affected by the presence of <1% methanol. Hence, we assume that the presence of the methanol does not affect the partitioning of the chlorinated ethenes between the aqueous phase and the headspace in the vial. This implies that, among other things, Henry's constants can be estimated using the relationship of Staudinger and Roberts (2001), as shown in Table 1.

Results and Analysis

GC/FID peak areas versus aqueous concentration

Results of the GC/FID analyses of the aqueous standards are shown in Fig. 1. The measured GC/FID peak area is plotted versus aqueous concentration for each of the six target contaminants. In other words, Fig. 1 presents the GC/FID calibration curves for each of the chlorinated ethenes in aqueous solution.

FIG. 1.

Gas chromatography with flame ionization detection (GC/FID) peak areas versus aqueous concentration for each of the six target analytes. Symbols represent measured data; solid lines represent best linear fits to the measured data. Equations are given for the best linear fits as determined with weighed least-square regression; linearity is indicated by the square of the correlation coefficient, r². Data are shown on a logarithmic scale to demonstrate linearity over two orders of magnitude in concentration.

Best-fit lines were determined assuming that the peak area is linearly proportional to concentration, A=m·C, with the intercept forced to be zero; that is, we require that a concentration of zero corresponds to a peak area of zero. The slopes m of the best-fit lines were determined using linear least-square regression with weighed residuals, using the inverse of the square of concentration as a weighting factor (Draper and Smith, 1998, p. 225). This puts approximately equal weight on data in all concentration ranges, rather than having the best-fit slope determined by only the high-concentration data. We verified that the weighted residuals are homoscedastic (results not shown), which implies that the selected weighting was appropriate.

The linearity of the data is quantified with r², the square of the correlation coefficient (Moore and McCabe, 1993, p. 162). It is worth noting that in this case, r² is not exactly equal to R², the coefficient of determination for the best-fit line, for two reasons: first, because we assume a zero-intercept model for the relationship between area and concentration, and second, because we used weighed residuals to estimate the best-fit slope m (Willett and Singer, 1988; Golberg and Cho, 2004, p. 293). However, Equation (1.6.10) of Draper and Smith (1998) and Equation (6.141) of Golberg and Cho (2004) both suggest that r² is a reasonable estimate of R² in this case, even though the two are not identically equal. Thus, we choose r² as an appropriate measure of the goodness-of-fit of the calibration lines.

The data in Fig. 1 are plotted with logarithmic axes to allow observation of the data points and the best-fit line across the entire measured concentration range. For comparison, Fig. 2 shows the calibration curve for 1,1-DCE with linear axes. The concentration range of the aqueous standards spans two orders of magnitude (as seen in Table 2), and thus low-concentration data are compressed in Fig. 2, but are visible in Fig. 1.

FIG. 2.

GC/FID peak areas versus aqueous concentration for 1,1-dichloroethene (DCE). Symbols represent measured data; the solid line represents the best-fit line from Fig. 1. Data are shown on linear scale for comparison to Fig. 1.

Measured peak areas are corrected using methanol as an internal standard, because all aqueous samples contain the same methanol concentration (0.99% by volume, neglecting change of volume upon mixing). The peak areas are corrected according to the following formula: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}A_ { analyte } ^ { corrected } = A_ { analyte } ^ { measured } \times \frac { A _ { methanol } ^ { average } } { A_ { methanol } ^ { measured } } \tag { 1 } \end{align*} \end{document}

where \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$A_{analyte}^{corrected}$$\end{document} is the corrected peak area for the target analyte in a given sample; \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$A_{analyte}^{measured}$$ \end{document} is the measured (uncorrected) peak area for that target analyte in that sample; \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$A_{methanol}^{measured}$$ \end{document} is the measured peak area for methanol in that sample, and \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$A _{methanol} ^{average}$$ \end{document} is the average methanol peak area measured for all 30 standards. When analyzing actual field samples, methanol would not likely be present, and an internal standard or a surrogate standard would need to be introduced into the samples (U.S. Army Corps of Engineers, 2005). In that case, it is likely that a standard would be chosen that is chemically similar to the analytes of concern. For instance, Herzfeld et al. (1989) used dibromomethane as an internal standard during headspace analysis of TCE and PCE in water samples. However, for the work performed here, methanol was already present in an appropriate concentration and could therefore function as the internal standard. We note that the measured methanol peak area was quite consistent in all samples analyzed (coefficient of variation=0.051), which suggests that both the creation of the aqueous standards and the headspace analysis thereof were performed consistently.

It can be seen from Fig. 1 that, as expected, the FID peak area is linear with respect to the aqueous concentration of the target analyte. This was observed for all six chlorinated ethenes tested. However, it is important to note that when peak area is plotted versus aqueous concentration, the slopes are different for each of the six calibration curves. The best-fit slopes (with 95% confidence intervals) of the six compounds are as follows: PCE 29,410±4970; TCE 26,950±4210; cis-DCE 20,980±4720; trans-DCE 35,740±3380; 1,1-DCE 55,440±6250; VC 51,450±26,790. These results show clearly that when calibration curves are presented as peak area versus aqueous concentration, a separate calibration curve is required for each target analyte.

GC/FID peak areas versus moles of target analyte injected into GC

The hypothesis of this project is that all chlorinated ethenes have the same FID response factor. To test this hypothesis, we compute the number of moles of each target analyte actually injected into the GC during the headspace analysis. This computation is performed as follows: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}M _i ^ { injected } = \frac { V^ { injected } } { V^ { air } } f_i^ { air } M_i^ { total } \tag { 2 } \end{align*} \end{document}

where H_i is the dimensionless Henry's constant for compound i at the equilibration temperature of 21.7°C (given in Table 1). Equation (4) assumes that the air and water reach full equilibrium within the 12-hr equilibration time. Equation (3) assumes that there are no mass losses (e.g., biodegradation or volatilization) of any of the contaminants once they are introduced into the vials. Note that, although the vials have a 5-mL nominal size, the actual volume of the vials is 6.21±0.10 mL (mean and standard deviation based on 10 measured values), of which 2.02 mL is aqueous solution. We therefore assumed a value of V^air=4.19 mL for all calculations. Even though the volume in an individual vial may differ slightly from this assumed value, deviations between vials are small enough that they may be ignored without introducing significant error.

Equations (2 –4) are used to calculate \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$M _i ^{injected}$$ \end{document} , the mass of contaminant injected, for each chemical in each vial. Figure 3 graphs the measured GC/FID peak area versus \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} $$M _i ^{injected}$$ \end{document} for each of the six target contaminants. Figure 3, like Figs. 1 and 2, can be considered a GC/FID calibration curve for chlorinated ethenes in an aqueous solution. Like Fig. 1, the data in Fig. 3 are plotted with logarithmic axes to allow observation of the data points across the entire measured concentration range. For comparison, Fig. 4 presents the same PCE, TCE, and DCE data as Fig. 3, but with linear axes instead of logarithmic axes.

FIG. 3.

GC/FID peak areas versus moles of contaminant injected into the GC. Symbols represent measured data; solid lines represent best linear fits to the measured data. Diamond symbols (⋄) are for perchloroethene (PCE); triangles (∇) are for trichloroethene (TCE); squares (□) are for cis-DCE; circles (○) are for trans-DCE; six-pointed stars () are for 1,1-DCE; and five-pointed stars () are for vinyl chloride (VC). Equations are given for the best linear fits as determined with weighed least-square regression; linearity is indicated by the square of the correlation coefficient, r².

FIG. 4.

GC/FID peak areas versus moles of contaminant injected into the GC for PCE, TCE, and three DCE isomers; linear axes are used for comparison with Fig. 3. Symbols represent measured data; solid lines represent best linear fits to the measured data. Diamond symbols (⋄) are for PCE; triangles (∇) are for TCE; squares (□) are for cis-DCE; circles (○) are for trans-DCE; and six-pointed stars () are for 1,1-DCE. Top panel shows all data (n=50). Bottom panel zooms in on the low-concentration range (area indicated by the dashed line in the top panel).

It can be seen clearly from Figs. 3 and 4 that PCE, TCE, and all three DCE isomers share the same calibration line when the GC/FID peak area is graphed versus the moles of compound injected. A single calibration line was determined from the combined PCE, TCE, and DCE data and was found to fit the measured data (n=50 measurements) with r²=0.990. The slope of the best-fit line (with 95% confidence interval) for the five compounds together is 17,308,000±3,058,000. If each of the five compounds is considered individually, then the best-fit slopes range from 16,164,000±1,527,000 (for trans-DCE) to 18,524,000±3,128,000 (for PCE). The 95% confidence intervals for all five compounds overlap with each other, and also overlap with the overall best-fit line for the five compounds together. Therefore, we conclude that the measured data are consistent with the hypothesis that these five chlorinated ethenes exhibit the same response factor on the FID. In other words, one micromole of PCE results in the same peak area as one micromole of TCE or one micromole of DCE.

However, the measured peak areas for VC do not lie on the same line (Fig. 3). The slope of the VC calibration curve is about 42% lower than the slope of the PCE/TCE/DCE curve, indicating that the apparent VC response factor is about 42% lower than the response factor for the other five analytes.

Discussion

Significance and implications of results

Because PCE, TCE, cis-DCE, trans-DCE, and 1,1-DCE all share the same GC/FID response factor, separate calibration curves do not need to be generated for all five compounds. Any one of these five compounds could be used to generate the calibration curve shown in Fig. 3, and the other four compounds would share the same calibration curve. The calibration curves in Fig. 1, which are presented as GC/FID peak area versus aqueous concentration, could be computed by starting with the best-fit calibration line from Fig. 3 and using Equations (2 –4). To the best of our knowledge, this is the first study to investigate if all chlorinated ethenes exhibit the same FID response factor.

We contend that this finding will allow scientists and engineers to save time and money in the future by obviating the need to generate separate calibration curves for each individual chlorinated ethene. The literature abounds with studies where researchers developed separate calibration curves for GC/FID headspace analysis of two or more of these chlorinated ethenes. For example, Herzfeld et al. (1989) analyzed surface water samples for both TCE and PCE (as well as for two chlorinated methanes) via headspace GC/FID. Because Herzfeld et al. (1989) generated calibration curves as a peak area ratio versus aqueous concentration (as shown in Figs. 1 and 2), TCE and PCE exhibited different slopes and were considered separately. Prommer et al. (2008) investigated the degradation of TCE in groundwater by use of zero-valent iron, and they analyzed for TCE and all three DCE isomers via headspace GC/FID; separate calibration curves were determined for all four compounds. Haest et al. (2010) and Mendoza-Sanchez et al. (2010) both studied biologically mediated reductive dechlorination of TCE in water; both groups analyzed for TCE, cis-DCE, and VC by headspace GC/FID, with separate calibration curves generated for each. These are just a few examples in which multiple chlorinated ethenes were analyzed by headspace GC/FID, with separate calibration curves generated for each. In all of these cases, the investigating team could have saved money on purchasing analytical standards, and/or could have saved time in creating and analyzing the FID response of those analytical standards. The studies cited above are not intended to be an exhaustive literature review, but merely to demonstrate the potential utility of the findings of this article.

Difference between response factor for VC and all other chlorinated ethenes

The observation that the apparent VC response factor is about 42% lower than the PCE/TCE/DCE response factor is surprising. We are able to imagine three possible reasons for this surprising observation:

1. the central hypothesis of this work is incorrect, and chlorinated ethenes do not all exhibit the same response factor when analyzed by FID;

2. the VC-in-methanol standard from which we constructed our aqueous standards had a concentration that was lower than the promised concentration of 1000 μg/mL, so that we have overestimated the VC concentration and mass in all samples; or

3. about 42% of the VC was lost out of the vials by volatilization during the 12-h equilibration period (and/or during the subsequent sampling and injection into the GC).

Of these three, the third hypothesis seems the most likely. VC has the lowest boiling point and the highest vapor pressure of the six chlorinated ethenes considered, and thus is the most likely to be lost by volatilization. We are unable to conceive of a plausible physical mechanism by which a flame ionization detector would respond differently to VC than it does to TCE, PCE, or any of the three DCE isomers (cf. Sternberg et al., 1962).

To test this hypothesis, we conducted a follow-up experiment to those described above. We obtained a gas standard (American Air Liquide) of VC in nitrogen gas, concentration 100 parts per million by volume (ppmv). The gas standard was used to fill a gas-tight bag at atmospheric pressure. Then, using a gas-tight syringe, we withdrew gas from the gas-tight bag and injected it directly into the GC/FID. The rationale for this experiment is that by directly injecting a VC gas standard, we eliminate the potential for volatilization of VC out of the vials during the 12-h air–water equilibration period. Hence, we expect the measured VC peak areas for the gas standards to coincide with the calibration curve for PCE, TCE, and DCE. We injected four different volumes of gas, namely 0.4, 0.6, 0.8, and 1.0 mL; duplicate experiments were performed for each volume. Using the ideal gas law with an air temperature of 21.7°C and a concentration of 100 ppmv, we calculate that these volumes correspond to 0.0017, 0.0025, 0.0033, and 0.0041 μmoles of VC injected. Peak areas were measured on the FID.

Figure 5 shows the results of the gas-standard-injection experiment. In Fig. 5, the upper solid line corresponds to the calibration line previously determined for PCE, TCE, and DCE isomers; the lower dashed line corresponds to the calibration line previously determined for VC. The symbols (stars) represent the measured data. It can be seen from Fig. 5 that, as expected, the measured VC data lie closer to the calibration line for PCE, TCE, and DCE, which is consistent with our hypothesis that previous low results for VC were caused by volatilization out of the vials. It can also be seen from Fig. 5 that, even when directly injecting gas standards, the measured peak areas for VC are still lower than expected by about 10% (on average). However, this discrepancy lies within the 95% confidence interval for the slope of the PCE/TCE/DCE best-fit line (slope=17,308,000±3,058,000). Therefore, we cannot state with 95% confidence that the VC response factor is different from the PCE/TCE/DCE response factor. It may be that, even when directly injecting VC gas standards into the GC, there was some volatilization out of the gas-tight syringe, and this may account for the remaining ∼10% difference between observed and expected values of the VC response factor.

FIG. 5.

GC/FID peak areas determined by direct injection of vinyl chloride gas standard. Symbols represent measured data. The upper (solid) line is the best-fit calibration line previously determined for PCE, TCE, and DCE. The lower (dashed) line is the best-fit calibration line for VC as determined by analysis of aqueous samples.

It may be possible to reduce some of the VC volatilization by using different equipment or different experimental protocols. We used a standard manual gas-tight syringe to sample headspace from the vials after equilibration (Prommer et al., 2008; Penton, 2010). Some syringes can be fitted with valves to reduce sample loss through the needle while the sample is transported to the GC. Other, more sophisticated sampling techniques are also available. For instance, Roberts et al. (1996) used a system where headspace equilibration was performed in the syringe itself, which may reduce losses during sampling and transfer to the GC.

If volatilization could be minimized, then perhaps the PCE/TCE/DCE calibration curve could be reliably applied to VC as well. However, based on the method we employed here, a separate calibration curve is required for VC during analysis of aqueous samples. Losses due to volatilization are significant during sample analysis, and must be taken into account by the corresponding calibration curve.

Summary and Conclusions

We tested the hypothesis that chlorinated ethenes all exhibit the same FID response factor. The hypothesis was tested by conducting headspace GC/FID analysis of aqueous standards of known concentration. We observed that PCE, TCE, and all three DCE isomers do exhibit the same FID response factor, as hypothesized. When the GC/FID peak area is graphed versus moles of contaminant injected into the GC, a single calibration line is able to fit the PCE, TCE, and DCE data with r²=0.990. However, the apparent VC response factor was ∼42% lower than that of the other three chemicals. Most of this difference is due to losses of VC by volatilization during a 12-h equilibration period (and perhaps during the subsequent sampling and injection into the GC). These results imply that it is not necessary to develop separate calibration curves for PCE, TCE, and DCE; this finding may save time and money during analysis of environmental samples containing chlorinated ethenes. However, it is necessary to develop a separate calibration curve for VC, so that differences in the response factor and/or losses from volatilization will be taken into account during sample analysis.

Footnotes

Acknowledgments

This material is based on work performed by the first two authors as part of a Research Experiences for Undergraduates (REU) program. The REU program was supported by the College of Engineering at the University of South Florida (USF), and by the National Science Foundation (NSF) through award number 0851910. Any opinions, findings, conclusions, or recommendations are those of the authors and do not necessarily represent the views of NSF or USF. The authors thank three anonymous reviewers for constructive reviews on two earlier versions of this manuscript. The authors thank Ms Margaret Cone of USF for assisting with some laboratory measurements.

Author Disclosure Statement

No competing financial interests exist.

References

Dietz

E.A.

Jr. , Singley

K.F.

1979. Determination of chlorinated hydrocarbons in water by headspace gas chromatography. Anal. Chem., 51:1809.

Draper

N.R.

, Smith

1998. Applied Regression Analysis, 3rd. New York: John Wiley & Sons, Inc.

Ettre

L.S.

1962. Relative molar response of hydrocarbons on the ionization detectors. Brenner

, Callen

J.E.

Weiss

M.D.

Gas Chromatography. New York: Academic Press, Inc., 307.

Golberg

M.A.

, Cho

H.A.

2004. Introduction to Regression Analysis. Southampton, United Kingdom: WIT Press.

Haest

P.J.

, Springael

, Smolders

2010. Dechlorination kinetics of TCE at toxic TCE concentrations: Assessment of different models. Water Res., 44:331.

Herzfeld

, van der Gun

K.-D.

, Louw

1989. Quantitative determination of volatile organochlorine compounds in water by GC-headspace analysis with dibromomethane as an internal standard. Chemosphere, 18:1425.

Huang

, Ou

, Yu

1990. Characteristics of flame ionization detection for the quantitative analysis of complex organic mixtures. Anal. Chem., 62:2063.

Jorgensen

A.D.

, Picel

K.C.

, Stamoudis

V.C.

1990. Prediction of gas chromatography flame ionization detector response factors from molecular structures. Anal. Chem., 62:683.

Kallai

, Veres

, Balla

2001. Response of flame ionization detectors to different homologous series. Chromatographia, 54:511.

10.

Ketola

R.A.

, Virkki

V.T.

, Ojala

, Komppa

, Kotiaho

1997. Comparison of different methods for the determination of volatile organic compounds in water samples. Talanta, 44:373.

11.

Leveque

R.E.

1967. Determination of C3-C8 hydrocarbons in naphthas and reformates utilizing capillary column gas chromatography. Anal. Chem., 39:1811.

12.

Mendoza-Sanchez

, Autenrieth

R.L.

, McDonald

T.J.

, Cunningham

J.A.

2010. Effect of pore velocity on biodegradation of cis-dichloroethene (DCE) in column experiments. Biodegradation, 21:365.

13.

Mockel

H.J.

1976. FID response factors for aliphatic sulphur compounds at higher concentration levels. Fresen. Z. Anal. Chem., 279:199.

14.

Moore

D.S.

, McCabe

G.P.

1993. Introduction to the Practice of Statistics, 2nd. New York: W.H. Freeman and Company.

15.

National Research Council (NRC). 1994. Alternatives for Ground Water Cleanup. Washington, DC: National Academies Press.

16.

Otson

, Williams

D.T.

1982. Headspace chromatographic determination of water pollutants. Anal. Chem., 54:942.

17.

Penton

1992. Optimization of conditions in static headspace GC. J. High Res. Chromatog., 15:834.

18.

Penton

Z.E.

2010. Headspace gas chromatography. Pawliszyn

, Lord

H.L.

Handbook of Sample Preparation. Hoboken, NJ: John Wiley & Sons, 25–38.

19.

Perkins

, Rouayheb

G.M.

, Lively

L.D.

, Hamilton

W.C.

1962. Response of the gas chromatographic flame ionization detector to different functional groups. Brenner

, Callen

J.E.

, Weiss

M.D.

Gas Chromatography. New York: Academic Press, Inc., 269–285.

20.

Perkins

, Laramy

R.E.

, Lively

L.D.

1963. Flame response in the quantitative determination of high molecular weight paraffins and alcohols by gas chromatography. Anal. Chem., 35:360.

21.

Prommer

, Aziz

L.H.

, Bolaño

, Taubald

, Schüth

2008. Modelling of geochemical and isotopic changes in a column experiment for degradation of TCE by zero-valent iron. J. Contam. Hydrol., 97:13.

22.

Roberts

A.L.

, Totten

L.A.

, Arnold

W.A.

, Burris

D.R.

, Campbell

T.J.

1996. Reductive elimination of chlorinated ethylenes by zero-valent metals. Environ. Sci. Technol., 30:2654.

23.

Schwarzenbach

R.P.

, Gschwend

P.M.

, Imboden

D.M.

2003. Environmental Organic Chemistry. Hoboken, NJ: Wiley-Interscience.

24.

Slemr

, Slemr

, D'Souza

, Partridge

2004. Study of the relative response factors of various gas chromatograph-flame ionisation detector systems for measurement of C₂-C₉ hydrocarbons in air. J. Chromatogr. A, 1061:75.

25.

Staudinger

, Roberts

P.V.

2001. A critical compilation of Henry's law constant temperature dependence relations for organic compounds in dilute aqueous solutions. Chemosphere, 44:561.

26.

Sternberg

J.C.

, Gallaway

W.S.

, Jones

T.L.

1962. The mechanism of response of flame ionization detectors. Brenner

, Callen

J.E.

, Weiss

M.D.

Gas Chromatography. New York: Academic Press, Inc., 231–267.

27.

Tong

H. Y.

, Karasek

F.W.

1984. Flame ionization detector response factors for compound classes in quantitative analysis of complex organic mixtures. Anal. Chem., 56:2124.

28.

United States Army Corps of Engineers (US ACE). 2005. Guidance for Evaluating Performance-Based Chemical DataEngineer Manual EM 200-1-10http://140.194.76.129/publications/eng-manuals/EM_200-1-10. 2012 November 2

29.

Wartenberg

, Reyner

, Scott

C.S.

2000. Trichloroethylene and cancer: epidemiologic evidence. Environ. Health Persp., 108:161.

30.

Willett

J.B.

, Singer

J.D.

1988. Another cautionary note about R²: Its use in weighed least-squares regression analysis. Am. Stat., 42:236.