Environ. Sci. Technol. 1996, 30, 1212-1219
Determination of Volatile Organic Compounds in Water by Solid Phase Microextraction and Infrared Spectroscopy DANIEL L. HEGLUND AND DAVID C. TILOTTA* The Department of Chemistry, Box 9024, University of North Dakota, Grand Forks, North Dakota 58202
A simple method is described combining solid phase microextraction (SPME) and infrared (IR) spectroscopy for determining volatile organic compounds (VOCs) in water. The solid phase consists of a small square of Parafilm (130 µm thick) that is used to selectively extract the VOCs from the water. Infrared transmission spectroscopy is used to detect the extracted VOCs directly in the Parafilm. Ten VOCs (e.g., benzene, chlorobenzene, toluene, chloroform, and p-chlorotoluene) are chosen to evaluate the SPME-IR procedure. Preliminary experiments show that detection limits are in the 66 ppb-1.3 ppm range for spiked solutions, and linear dynamic calibration ranges extend nearly to the water solubility limits for all VOCs examined. Although the formal equilibration times are determined to be 30-200 min for these VOCs, we demonstrate that reproducible extractions can also be performed at 30 min. The application of SPME-IR to a real water sample shows no significant matrix interferences. Finally, the potential for determining individual compounds in mixtures is investigated by extracting aromatic components from water samples contaminated with gasoline.
Introduction Volatile organic compounds (VOCs) found as contaminants in ground and surface waters include benzene, toluene, ethylbenzene, xylenes (collectively referred to as the BTEX compounds), and chlorinated hydrocarbons (such as chloroform and carbon tetrachloride). Because these organic compounds are toxic to humans and aquatic life, their sensitive and rapid determination is of critical importance. Current methods for determining VOCs in water include liquid/liquid extraction (1, 2), solid phase extraction (3), solid phase microextraction (SPME) (4, 5), and purge and trap gas chromatography (6). Although each of these methods has certain advantages and disadvantages, many of them are time-consuming, labor-intensive, and difficult to implement and require relatively expensive * Author to whom correspondence should be addressed: Fax: (701) 777-2331. E-mail:
[email protected].
1212
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 30, NO. 4, 1996
instrumentation. In addition, some are not readily adaptable to field determinations. Infrared (IR) spectroscopy is a sensitive analytical technique that provides compound-specific information via molecular structure. Infrared spectroscopy is also relatively inexpensive to implement, and analytes can be easily quantitated through the Beer-Lambert law (A ) bc, where A is the absorbance, is the molar absorptivity, b is the path length, and c is the concentration). Therefore, IR spectrometry can be a potentially useful method for the sensitive and selective determination of VOCs. However, it is well-known that the determination of most organic compounds directly in water matrices by IR transmission spectroscopy is difficult because of the spectral interference of the water, offering detection limits in the percent concentration range. As a means of circumventing this severe spectral interference, several groups have investigated using polymer coatings and evanescent wave spectroscopy to effectively move the organic from the water to the surface of an attenuated total reflectance (ATR) element. For example, Kellner and associates developed methodology for coating optical fibers (10-cm lengths) for determining chlorinated hydrocarbons in water and showed that sensitivities in the 1-50 ppm level can be obtained (7-11). In a similar experiment, Ertan-Lamontagne et al. coated the surface of an ATR cell with a chloroparaffin/poly(vinyl chloride) polymer and reported concentration enhancements for nitrobenzene (12). Although the direction of these experiments is toward the development of fiber-optic remote sensing systems, these workers showed that polymer membranes (coupled with evanescent wave technology in these cases) could be used to minimize the interfering effects of water through the selective removal of the organic analytes. SPME is a relatively new analytical technique that relies on the partitioning of organic compounds from a water matrix directly into a solid phase (4, 5). These solid phases are typically polymers [e.g., poly(dimethylsiloxane) or poly(acrylate)] that function as selective preconcentrators. The current technology for SPME involves coating the solid phase on a short length of glass fiber. The fiber, installed on the plunger of a modified syringe assembly, is then used to determine organics in water by extracting the analyte from the sample and thermally desorbing it in the injector of a gas chromatograph (GC). Standard GC methods are used to quantitate the extracted analytes. At equilibrium, the distribution of the organic compound between the solid phase and the aqueous phase has been shown to be described by classical extraction theory as
Cs ) K d Ca
(1)
where Cs is the concentration of the organic in the solid phase, Ca is the concentration of the organic in the aqueous phase, and Kd is the distribution constant for the solid phase/aqueous phase system (5). Since the Kd values have been shown to be quite large for many environmentally important compounds (e.g., factors of 100-100 000), the solid phase effectively preconcentrates the organic from the water. The selectivity of SPME arises from matching the properties of the analytes to those of the appropriate solid phase.
0013-936X/96/0930-1212$12.00/0
1996 American Chemical Society
FIGURE 1. Schematic diagram (exploded) of SPME-IR holder: BP, back plate (5.0 × 5.3 cm); N, nuts; W, wire spacer; P, Parafilm; FP, front plate (4.0 × 5.3 cm); B, bolts.
This paper describes the development of a simple and sensitive “dippable” probe for determining VOCs in water using SPME with IR detection. In this work, the analytes are removed from the water with a thin polymer film (Parafilm M) and then determined in the same polymer film by transmission IR spectroscopy. Solid phase microextraction of VOCs with IR detection offers the advantage of a simple method for the rapid testing of contaminated waters (both natural and anthropogenic). The only major piece of equipment necessary for SPME with IR detection is a Fourier transform infrared spectrometer (FT-IR). The accessory described in this work is a simple film holder that supports the SPME material so that it can be removed and reproducibly replaced in the sample compartment of the FT-IR. As opposed to the prior work using ATR methods, the major advantage of these transmission IR measurements is that they are inherently simpler to perform; that is, special polymer coatings and ATR elements are not required. As a result, this method is inexpensive to implement (8 h to vaporize the extracted compounds contained in the Parafilm. Infrared spectra of the VOCs dissolved in CS2 were obtained in a 0.2-mm ZnSe cell (Janos, Townshend, VT). The molar absorptivities of the various analytical VOC bands were determined from the slopes of the calibration curves prepared in this solvent. The heights of the infrared bands were used to provide quantitative data. Detection limits were obtained by extracting solutions that had concentrations producing absorbances ca. 3-8× the baseline noise. The detection limit is defined here as that concentration which produces an absorbance equivalent to twice the peak-to-peak baseline noise.
Results and Discussion Chromatographic Considerations. A simple, dippable probe constructed from a solid phase for determining VOCs in water by IR transmission spectroscopy must possess several features. Similarly to the application of SPME in gas chromatography, the solid phase must have an affinity for the analytes of interest, possess a reasonable equilibration time (or the ability to quantitate at nonequilibrium), and be insoluble in the water matrix. In addition, for IR transmission spectroscopy to be useful as a detector, the solid phase must be structurally rigid (or be bonded to a transparent solid support) to hold its shape during the analysis (i.e., a gum, rubber, or wax) and be optically transparent in the region of the analyte bands (freedom from spectral interferences). As an option that would make the analysis more attractive, the solid phase should be able to be reusable for several extractions. Parafilm was chosen as a nonbonded solid phase for the IR experiments because it possessed all of the characteristics described above. Other solid phases examined such as Apiezon-J and Apiezon-L (both mixed hydrocarbons), poly(dimethylsiloxane), and Teflon were rejected for use with IR spectroscopy since they either possessed absorption bands that obscured the bands of the analytes, were viscous fluids, or exhibited poor partitioning. Parafilm is a commercially available wax-impregnated polymer/rubber composite available in 2-20-in.-wide rolls up to 250 ft long. It is well suited for infrared transmission spectroscopy because the Parafilm is a moderately uniform thin film (i.e., a thickness of 130 ( 5 µm). In addition, it is inexpensive (3035, 27681500, 1335-1240, 1204-735, and 710-400 cm-1. The spectral region of 1200-460 cm-1 is the most useful for the determination of organic compounds partitioned into Parafilm. This spectral region encompasses the skeletal vibrations such as the out-of-plane C-H ring bends (900675 cm-1), the out-of-plane ring bends (710-675), methylene twisting and wagging (1350-1150 cm-1), and the C-Cl stretching vibrations (850-550 cm-1). Table 1 presents a listing of the useful analytical bands in the Parafilm optical window for several VOCs examined in this study. Since most of these VOCs have several vibrational bands, the relative absorbances of other bands observable in the Parafilm window are also listed (normalized to the major analytical band). For example, the major band of toluene occurs at 693 cm-1 with a second band at 464 cm-1 at 0.70× the absorbance of the major band and a third band at 1029 cm-1 at 0.14× the absorbance of the major band. Effect of Analyte Evaporation. Evaporative loss of the VOC from the Parafilm into the air occurs as soon as the Parafilm is removed from the water matrix. In addition, VOC evaporation is accelerated in this experiment because the Parafilm is heated by the IR beam of the spectrometer. Figure 3 shows an example of the real-time loss of benzene (one of the more volatile analytes examined in this study) from Parafilm as monitored by IR absorption spectroscopy. The graph in Figure 3 was obtained by extracting a 100 ppm aqueous benzene solution for 90 min with Parafilm
TABLE 1
Analytical Bands in the Parafilm Optical Window for Selected VOCs
a
chemical
anal. band,a cm-1
alternate bands, cm-1 (intens of anal. band, %)
benzene carbon tetrachloride chlorobenzene chloroform p-chlorotoluene ethylbenzene toluene m-xylene o-xylene p-xylene
673 786 739 761 804 695 693 767 741 795
1037 (6) nab 685 (32), 703 (21), 902 (6), 1023 (16), 1085 (21) na 483 (38), 634 (12), 1018 (14), 1092 (58), 1111 (4) 555 (10), 772 (8), 1032 (10) 464 (70), 1029 (14) 691 (47), 875 (6), 1059 (4), 1094 (5), 1170 (4) 985 (1), 1021 (4), 1053 (6), 1119 (3) 483 (53), 1042 (4), 1102 (3), 1119 (7)
Strongest absorption band.
b
na, not available.
FIGURE 4. Effect of scan co-addition on S/N for the 673-cm-1 benzene band. FIGURE 3. Evaporative loss of benzene from Parafilm following a 90-min extraction. The absorbance band at 673 cm-1 was monitored.
and plotting the absorbance of the strongest benzene band (673 cm-1) versus time. It should be noted that the Parafilm remained in the FT-IR spectrometer for the duration of the study. It can easily be seen from this figure that the IR signal decreases rapidly with time as a result of evaporative loss. In fact, the IR signal is reduced to half of the initial value after 15 min of exposure to the IR beam. The loss of the analyte as a function of time has an important ramification with respect to acquiring infrared spectra on an FT-IR spectrometer. Specifically, it is common practice in FT-IR spectrometry to co-add spectra in order to increase the signal-to-noise ratio (S/N) and, hence, improve the detection limits. It has been shown that as long as the IR signal is constant and the noise is random, S/Ns increase as the square root of the number of co-added scans (13). However, in this procedure, since the VOCs rapidly evaporate from the Parafilm on exposure to air and the IR beam, one might expect the S/N to actually decrease as the number of spectra are co-added (because the analyte signal is decreasing). Figure 4 shows a plot of the S/N of the 673-cm-1 IR band of benzene versus the number of co-added scans. It is evident upon inspection of this plot that the S/N is maximized for a specific number of co-added scans. Over the progression from one to eight scans, the noise in the FT-IR spectrometer decreases faster than the absorbance signal decreases until the S/N reaches a maximum at about eight scans (requiring 30 s). However at greater than eight scans, the absorbance signal decreases faster than the spectrometer noise and results in decreasing
TABLE 2
Equilibration Times for Selected VOCs in Parafilm compound
T90%, min
T100%, min
benzene carbon tetrachloride chlorobenzene chloroform p-chlorotoluene ethylbenzene toluene m-xylene o-xylene p-xylene
76 44 60 40 60 57 23 125 125 154
90 60 70 50 70 60 30 165 165 200
S/Ns. Thus, contrary to general practice, more co-added spectra in this experiment result in slightly poorer S/Ns. Similar results were determined for other VOCs and other concentrations examined in this study. Extraction Equilibria. The time required for each of the volatile compounds examined in this study to reach equilibrium between the aqueous phase and the Parafilm was determined by extracting 10 ppm solutions of each compound for increasing periods of time. The formal (100%) and 90% equilibration times for the various VOCs are shown in Table 2. The formal equilibration time (T100% in Table 2) is defined as the minimum time required for the absorbance signal to reach its maximum and become constant (within the RSDs of the measurements). Likewise, the 90% equilibration time (T90% in Table 2) is that time at which the absorbance is at 90% of its maximum. It was found that the extraction time for the various compounds
VOL. 30, NO. 4, 1996 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1215
did not vary with concentration as long as the solutions were stirred vigorously. As can be seen from Table 2, the formal equilibration times range from 30 to 200 min for the various VOCs. It is interesting to note that these times are significantly longer than those reported by Arthur et al. for poly(dimethylsiloxane) films of similar thickness (4). The increase in equilibrium time arises from the slower diffusion of the analytes through the Parafilm. It should be realized, however, that quantitative extractions do not have to be performed at equilibrium. As will be shown in a subsequent section, the extraction times can be substantially shortened with only minor negative analytical consequences. Assuming that the Kd values are greater than unity, the distribution constants in this system (as defined by eq 1) are potentially measurements of the improvement in sensitivities over those of the organics directly in the water matrix (although, as will be discussed in a subsequent section, none of these compounds could be detected directly in water at low concentrations by transmission IR spectroscopy). Prior work by Pawliszyn et al. has shown that the distribution constants for the poly(dimethylsiloxane)/ aqueous phase system are approximated by conventional octanol/water distribution constants (Kow’s) (4, 5). Thus, the existing Kow values could be used to predict the improvements in the detection limits as well as other analytical properties for the VOCs. Similarly, a correlation between the Kow values and the Parafilm/water distribution constants would be useful in predicting the detection limits and other analytical parameters from existing information. On the basis of eq 1 (and writing explicit expressions for the concentration terms and the mass balance of the system), it has been shown that Kd can be obtained experimentally by
Kd )
nsVa Vs(VaCa0 - ns)
(2)
where ns is the number of moles extracted at equilibrium, Va is the volume of the extracted aqueous phase, Vs is the volume of the solid phase, and Ca0 is the initial concentration of the analyte (14). For measurements performed directly in the solid phase with optical spectroscopy, ns can be obtained from Beer’s law using
ns ) AsVs/sbs
(3)
where As is the absorbance in the solid phase, s is the molar absorptivity in the solid phase, and bs is the thickness of the solid phase. Because the molar absorptivities for the various bands in Parafilm were not available (and cannot be easily measured unless the exact concentration of the analyte in the Parafilm can also be determined), we measured molar absorptivities in CS2 solvent and used those values to approximate molar absorptivities in Parafilm. This substitution is reasonable since the polarity of CS2 and the Parafilm would be expected to be similar (owing to similar interactions). Thus, their IR spectra should be approximately equivalent. The distribution constants of the various VOCs in the Parafilm/water system are listed in Table 3 and range from 2.09 (CHCl3) to 579 (p-chlorotoluene). Since all of these Kd values are greater than unity, the VOCs examined in this study will enjoy a preconcentration enhancement when Parafilm is used as the extraction medium. A comparison
1216
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 30, NO. 4, 1996
TABLE 3
Comparison of Kow and Kd Values for Selected VOCsa
a
compound
Kow
Kd
benzene carbon tetrachloride chlorobenzene chloroform p-chlorotoluene ethylbenzene toluene m-xylene o-xylene p-xylene
135 436 692 93 nab 1413 490 1584 589 1412
28.7 46.0 130 2.09 579 74.7 64.3 307 175 395
Kow values from refs 15 and 16. b na, not available.
of the available Kow values (also shown in Table 3) with the Kd values for the Parafilm/water system reveals two features. First, the distribution constants for this system are ca. 5-10× smaller than the distribution constants of the octanol/water system. This disparity may be because Parafilm may be slightly more polar than poly(dimethylsiloxane) or there may be inactive absorption sites in the Parafilm. Second, the Kd’s of these VOCs correlate remarkably well with their Kow’s. For example, the trends in the Kd values of the two classes of VOCs (i.e., the chlorinated hydrocarbons and the BTEX compounds) follow those of the Kow values (with the exception of ethylbenzene and o-xylene). It should be realized, however, that several assumptions were made in the determination of these Kd values that may lead to errors (e.g., the constancy of between the CS2 and the Parafilm matrices). Equilibrium Determinations. Analytical calibration information for the 10 VOCs is shown in Table 4. The detection limits for these compounds range from 66 ppb to 1.3 ppm and were obtained using their major analytical band at their equilibrium extraction times (shown in Table 2). A comparison of the detection limits with the Kd values shows that as the Kd increases in a given class (i.e., either the chlorinated hydrocarbons or the BTEX compounds), the detection limit decreases. The single exception to this trend is toluene, most likely because of a small molar absorptivity. Because the Parafilm is apparently moderately nonpolar (based on the Kd values determined in the prior section), the more polar organics have poorer detection limits. For example, the detection limit for CCl4 is ca. 6× lower than for CHCl3. The detection limits obtained in this work are governed by both the residual fringing remaining in the Parafilm spectra and the instrumental noise. Although the major bands are used for these analytical studies, analytical information for the other bands can be obtained via Table 1 by dividing the “major band” data by the reduction in absorbance. For example, the detection limit for p-chlorotoluene using the major analytical band at 804 cm-1 is 66 ppb and ca. 550 ppb (66 ppb/0.12) using the less intense band at 634 cm-1. Likewise, the linear dynamic range is also reduced to 12% of the LDR for the major band (because of the reduction in the detection limit). It should be borne in mind that none of these VOCs could be detected directly in water by IR transmission spectroscopy. The single-beam spectrum in Figure 5 shows the relative amount of light transmitted (the emittance) by a distilled water sample in a 0.2-mm ZnSe cell (this path length is similar to the thickness of the Parafilm). This
TABLE 4
Equilibrium Calibration Data for Selected VOCs Extracted into Parafilm
a
compound
slope, AU/ppm
benzene carbon tetrachloride chlorobenzene chloroform p-chlorotoluene ethylbenzene toluene m-xylene o-xylene p-xylene
2.11 × 10-3 1.65 × 10-3 3.07 × 10-3 4.76 × 10-4 7.46 × 10-3 2.14 × 10-3 7.04 × 10-4 5.07 × 10-3 4.84 × 10-3 6.78 × 10-3
Linear dynamic range.
b
calibration curve data intercept, AU -3.61 × 10-3 -7.14 × 10-4 -5.84 × 10-4 -3.71 × 10-3 -3.67 × 10-3 -8.43 × 10-4 -6.71 × 10-4 -7.70 × 10-5 -4.21 × 10-4 -2.60 × 10-4
r2
LDR,a ppb
detection limit, ppb (% RSD)b
0.9993 0.9997 0.9995 0.9992 0.9970 0.9985 0.9989 0.9999 0.9996 0.9999
750-100000 640-80000 600-50000 4000-800000 225-37500 480-30000 4000-25000 140-7000 140-7000 140-7000
182(9) 200(3) 187(8) 1290(11) 66(10) 182(9) 752(4) 80(4) 102(9) 66(3)
Percent relative standard deviation data obtained from triplicate extractions at concentrations ca. 6× detection limits.
Table 5. Extraction times of 30 min were chosen as an example, only. The data in Table 5 show that the only negative consequences of performing determinations at nonequilibrium are a slight degradation of both the detection limit and the LDR (by a factor of 2-3 for a 30-min extraction). Because the concentrations of most of these compounds reach equilibrium in the Parafilm rather slowly (as shown by comparing the T90% and T100% values in Table 2), the reproducibilities of these measurements are reasonably good (RSDs of 3-6%) and, in fact, are similar to those obtained at equilibrium.
FIGURE 5. Single-beam infrared spectrum of water obtained through a 0.2-mm ZnSe cell.
spectrum was acquired under the same scan conditions as those with the Parafilm. It is obvious from this spectrum that the strong absorption of the water matrix obscures the majority of the IR spectral region (e.g., no light transmission in the range of 1800-400 cm-1). In fact, none of the VOCs examined in this study could be detected in the water matrix (at their solubility limits) using transmission spectroscopy. The LDR of calibration for all VOCs by SPME-IR (shown in Table 4) extends from the detection limit to the highest concentration examined in this study. Calibration data were not obtained near the solubility limits of the VOCs in the water because of uncertainties in the actual solution concentration (i.e., from emulsion formation or phase separation of the solutions). The reproducibility of the concentration measurements for the VOCs using this technique is also shown in Table 4. The percent RSD for each determination was calculated from triplicate extractions of single-component solutions with concentrations at ca. 6× their detection limits. The RSDs range from 3 to 11% and are governed by volatility losses, solution preparation errors, and instrumental noise. Nonequilibrium Determinations. With the exception of toluene, the formal (100%) equilibration times for the various VOCs are greater than 50 min. In fact, the equilibration times for the xylenes are on the order of 3-4 h. There is actually no need, however, to perform determinations at equilibrium provided that the extractions are performed reproducibly. Quantitative information for determining these VOCs in water at 30 min is shown in
Selectivity. In general, the ability of SPME and optical detection to determine individual compounds in mixtures will depend upon the selectivities of both the solid phase and the spectroscopy. Since the Parafilm solid phase used in this work is moderately nonpolar, most nonpolar molecules will partition into it (as evidenced by the 10 VOCs examined in this study); therefore, in this case, the selectivity of SPME-IR with Parafilm as the solid phase will rely heavily on the IR spectroscopy. Of course, the ability of IR spectroscopy to speciate in this application will depend upon the complexity of the sample. For simple mixtures, speciation is relatively straightforward since most of these VOCs have several unique vibrational bands. For example, the spectrum shown in Figure 6 was obtained by extracting a simple threecomponent mixture of benzene, toluene, and p-chlorotoluene in distilled water. Each compound was at a concentration level of 10 ppm, and the entire mixture was extracted for the equilibration time of the slowest equilibrating compound (benzene in this case). An inspection of this spectrum clearly shows that all bands for all compounds are baseline resolved. To illustrate the potential of using this method to speciate more complex multicomponent mixtures, a 100 ppm solution of American Petroleum Institute (API) gasoline was extracted from distilled water. The spectra shown in Figure 7 were obtained over a portion of the Parafilm optical window (815-650 cm-1) and demonstrate the potential for speciation. A comparison of the API gasoline extract spectrum with each of the individual component spectra shows that nearly all six of the BTEX compounds can be discerned. The only significant spectral overlap occurs for toluene, m-xylene, and ethylbenzene in the region at ca. 700 cm-1. However, since m-xylene has two bands in the region shown, speciation is only adversely affected for toluene and ethylbenzene.
VOL. 30, NO. 4, 1996 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1217
TABLE 5
Calibration Data (30 min) for Selected VOCs Extracted into Parafilm
a
compound
slope, AU/ppm
benzene carbon tetrachloride chlorobenzene chloroform p-chlorotoluene ethylbenzene m-xylene o-xylene p-xylene
8.20 × 10-4 6.17 × 10-4 1.58 × 10-3 3.09 × 10-4 4.35 × 10-3 1.51 × 10-3 1.92 × 10-3 2.40 × 10-3 2.43 × 10-3
Linear dynamic range.
b
calibration curve data intercept, AU 3.89 × 10-5 -6.47 × 10-4 -1.19 × 10-3 -2.19 × 10-3 -1.93 × 10-3 -9.01 × 10-4 1.90 × 10-5 8.34 × 10-4 2.07 × 10-4
r2
LDR,a ppb
detection limit, ppb (% RSD)b
0.9998 0.9957 0.9997 0.9996 0.9987 0.9990 0.9978 0.9983 0.9997
1400-105000 960-80000 1000-50000 6400-800 000 300-37500 780-30000 315-10500 315-10500 315-10500
400(6) 297(6) 439(9) 1325(7) 99(6) 422(6) 220(7) 204(7) 155(2)
Percent relative standard deviation data obtained from triplicate extractions at concentrations ca. 6× detection limits.
TABLE 6
Recoveries (30 min) for Selected VOCs Spiked into Natural Creek Water av absorbance, AU
FIGURE 6. Infrared spectrum obtained in Parafilm following a 90min extraction of a water solution containing 10 ppm each of benzene (B), toluene (T), and p-chlorotoluene (C). The “blanked-out” region at ca. 725 cm-1 is due to the strong absorption of the Parafilm.
FIGURE 7. Infrared spectra obtained in Parafilm following extractions of water solutions containing 100 ppm gasoline and 10 ppm each of the BTEX compounds. The baselines for the spectra have been offset for clarity, and the “blanked-out” regions in the center of all spectra are due to the strong absorption of the Parafilm.
According to the certificate of analysis accompanying the API gasoline, this sample (containing over 190 identifiable compounds) contained, by weight, 1.22% benzene, 7.68% toluene and C8 hydrocarbons, 3.37% ethylbenzene, 2.13% p-xylene and C9 hydrocarbons, 2.64% o-xylene, and 5.31% m-xylene and C9 hydrocarbons. Simple univariate quantitation using band heights yields a benzene and o-xylene content of 2.99 (3.6% RSD) and 2.49 ppm (6.3% RSD), respectively, from quadruplicate extractions from the
1218
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 30, NO. 4, 1996
chemical
in DI water
in creek water
% recovery (% RSD)a
benzene carbon tetrachloride chlorobenzene chloroform p-chlorotoluene ethylbenzene toluene m-xylene o-xylene p-xylene
0.004 50 0.005 00 0.003 97 0.003 97 0.006 70 0.004 80 0.002 90 0.005 17 0.004 57 0.005 03
0.004 57 0.005 40 0.004 03 0.003 87 0.006 63 0.005 13 0.002 83 0.004 97 0.004 20 0.005 03
101.7(4.6) 103.2(3.7) 101.7(1.4) 97.5(6.0) 98.5 (5.7) 98.7(2.2) 97.7(8.9) 96.1(1.1) 92.0(9.5) 100.0(11)
a Percent RSD data obtained from triplicate extractions at concentrations ca. 10× detection limits.
100 ppm solution. Considering the simplicity of the calibration, these concentrations are in good agreement with the certified values. Because the exact quantities of toluene, p-xylene, and m-xylene are unknown (and because of the spectral overlap of toluene and ethylbenzene), the concentrations of the other components were not determined. However, using currently available multivariate methods (such as partial least squares or principle component regression), it should be possible to provide quantitative information for all components in such a complex extraction mixture. Analyses of Real Water Samples. As a preliminary test to determine whether real water matrices interfere with these extractions, each of the compounds was spiked into creek water samples at concentrations ca. 10× their detection limits. The creek was located in Grand Forks, ND, and the water was used as obtained (i.e., unfiltered and not chemically treated). This water had particulates in suspension, with 1240 mg/L of residual matter after air evaporation. The VOC spikes were introduced into the creek water matrix as methanol solutions (final methanol concentration did not exceed 0.5%). All solutions were extracted in triplicate for 30 min, and the absorbance values obtained from the river water extractions were compared to those obtained from distilled water extractions. The recovery data are shown in Table 6 along with the RSDs of the measurements. The percent RSDs were taken from the triplicate measurements in the creek water. As can be seen in Table 6, the recoveries are quite good, ranging from 92 to 103%. The solid matter in the water sample did
TABLE 7
Comparison of SPME-IR Determination of BTEX Components in a Wastewater Sample with Standard Purge and Trap GC/MS Determinationa chemical
SPME-IR result, ppm (% RSD)b
GC/MS result, ppm (% RSD)c
benzene toluene ethylbenzene o-xylene m-xylene p-xylene
9.63(8.9) ndd 4.74(5.5) 1.08(3.8) 2.72(7.9) 1.30(12.0)
10.7(5.0) 12.1(5.2) 1.17(9.9) 1.53(20) nde nde
a GC/MS determination based on U.S. EPA method 624. b Percent RSDs from quadruplicate determinations using main analytical bands except p-xylene (483 cm-1). c Percent RSDs from triplicate determinations. d nd, not determined because unique spectral bands could not be identified in this sample. e nd, not determined because of chromatographic limitations. However, m- + p-xylenes were determined to be 6.7 ppm (% RSD ) 23).
not significantly affect the extraction recoveries for any of the VOCs. Finally, the SPME-IR method was applied to the determination of the BTEX components in a petroleum industry wastewater sample. As discussed in the Experimental Section, this water sample was obtained from a local source and was known to be contaminated with gasoline. The wastewater had particulates in suspension with 1230 mg/L of residual matter after air evaporation. Quadruplicate extractions were performed for 30 min on the sample, and univariate calibration based on peak height was employed for quantitation. The results of the determinations, along with those results obtained from the standard EPA method, are shown in Table 7. With the exception of ethylbenzene, the deviation of the SPME-IR results from those of the standard method (purge and trap GC/MS) are on the order of 10-40% and are generally within the standard deviations of the measurements. The concentration of ethylbenzene, as determined by SPME-IR, is about 3× larger than that determined by GC/MS and may be because of the spectral overlap of other components known to be in gasoline but not examined in this study (e.g., ethyl toluene). Similarly, toluene could not be determined by SPME-IR in this sample because of the severe spectral overlap of other unknown gasoline components. It should also be noted that (because of chromatographic limitations) m- and p-xylenes could not be individually determined by the GC/MS method. However, the sum of the m- and p-xylene concentrations
determined by SPME-IR is approximately equal to the sum determined by GC/MS (see the footnote to Table 7). As found with the spiked solutions, the RSDs of the SPME-IR determinations are in the range of 4-12% and are governed by volatility losses and instrumental noise.
Acknowledgments The authors thank the North Dakota Water Resources Research Institute, Grant ND93-06, for providing financial support for this work. In addition, we also thank Sally Eckert-Tilotta of the UND Water Quality Laboratory for performing the GC/MS analysis of the wastewater sample. This paper was presented, in part, at the 1995 Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy as paper 1001.
Literature Cited (1) Liska, I.; Kurpick, J.; LeClerc, P. A. J. High Resolut. Chromatogr. 1989, 12, 577. (2) Glaze, W. H.; Lin, C. C.; Burleson, J. L.; Henderson, J. E.; Mapel, D.; Rawley, R.; Scott, D. R. Optimization of Liquid-Liquid Extraction Methods for the Analysis of Organics in Water; U.S. Dept. of Commerce, National Technical Information Service: Springfield, VA, 1983. (3) Junk, G. A.; Richard, J. J. Anal. Chem. 1988, 60, 451. (4) Arthur, C. L.; Killiam, L. M.; Motlagh, S; Lim, M.; Potter, D. W.; Pawliszyn, J. Environ. Sci. Technol. 1992, 26, 979-983. (5) Arthur, C. L.; Potter, D. W.; Buchholz, K. D.; Motlagh, S.; Pawliszyn, J. LC-GC 1994, 10, 656-661. (6) USEPA Method 502.2. Fed. Regist. 1984, 32. (7) Krska, R.; Taga, K.; Kellner, R.; Messica, A.; Katzir, A. Fresenius J. Anal. Chem. 1992, 343, 202. (8) Krska, R.; Taga, K.; Kellner, R. Appl. Spectrosc. 1993, 47, 14841487. (9) Rosenberg, E.; Krska, R.; Kellner, R. Fresenius J. Anal. Chem. 1994, 348, 560-562. (10) Go¨bel, R.; Krska, Neal, S.; Kellner, R. Fresenius J. Anal. Chem. 1994, 348, 514-519. (11) Go¨bel, R.; Krska, R.; Taga, K.; Kellner, R. Appl. Spectrosc. 1994, 48, 678-683. (12) Ertan-Lamontagne, M. C.; Parthum, K. A.; Seitz, W. R.; Tomellini, S. A. Appl. Spectrosc. 1994, 48, 1439-1544. (13) Griffiths, P. R.; deHaseth, J. A. Fourier Transform Infrared Spectrometry; John Wiley and Sons: New York, 1986; p 254. (14) Buchholz, K. D.; Pawliszyn, J. Environ. Sci. Technol. 1993, 27, 2844-2848. (15) Chiou, C. T.; Schmedding, D. W.; Manes, M. Environ. Sci. Technol. 1982, 16, 4-10. (16) Wang, L.; Govind, R.; Dobbs, R. A. Environ. Sci. Technol. 1993, 27, 152-158.
Received for review June 20, 1995. Revised manuscript received November 15, 1995. Accepted November 18, 1995.X ES9504303 X
Abstract published in Advance ACS Abstracts, February 15, 1996.
VOL. 30, NO. 4, 1996 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1219
