Environ. Sci. Technol. 2001, 35, 1202-1208
A Water Extraction, Static Headspace Sampling, Gas Chromatographic Method to Determine MTBE in Heating Oil and Diesel Fuel TIMOTHY M. CUMMINS,† G A R Y A . R O B B I N S , * ,‡ BRENT J. HENEBRY,‡ C. RYAN GOAD,‡ EDWARD J. GILBERT,‡ MICHAEL E. MILLER,† AND JAMES D. STUART† Department of Chemistry, U-3060, 55 North Eagleville Road, University of Connecticut, Storrs, Connecticut 06269-3060, and Department of Geology and Geophysics, U-2045, 345 Mansfield Road, University of Connecticut, Storrs, Connecticut 06269-2045
A method was developed to determine the fuel/water partition coefficient (KMTBE) of methyl tert-butyl ether (MTBE) and then used to determine low parts per million concentrations of MTBE in samples of heating oil and diesel fuel. A special capillary column designed for the separation of MTBE and to prevent coelution and a gas chromatograph equipped with a photoionization detector (PID) were used. MTBE was partitioned from fuel samples into water during an equilibration step. The water samples were then analyzed for MTBE using static headspace sampling followed by GC/PID. A mathematical relationship was derived that allowed a KMTBE value to be calculated by utilizing the fuel/water volume ratios and the corresponding PID signal. KMTBE values were found to range linearly from 3.8 to 10.9 over a temperature range of 5-40 °C. This analysis method gave a MDL of 0.7 ppm MTBE in the fuel and a relative average accuracy of (15% by comparison with an independent laboratory using purge and trap GC/ MS analysis. MTBE was found in home heating oil in residential tanks and in diesel fuel at service stations throughout the state of Connecticut. The levels of MTBE were found to vary significantly with time. Heating oil and diesel fuel from terminals were also found to contain MTBE. This research suggests that the reported widespread contamination of groundwater with MTBE may also be due to heating oil and diesel fuel releases to the environment.
Introduction MTBE has recently received increased interest from national and local news organizations and in several scientific publications (1, 2). The U.S. Environmental Protection Agency (EPA) has announced its plan to ban MTBE as an additive in gasoline within the next 3 years due to its unreasonable risk to public health and to the environment. MTBE has been * To whom correspondence should be addressed. Phone: (860) 486-1392; fax: (860) 486-1383; e-mail:
[email protected]. † Department of Chemistry. ‡ Department of Geology and Geophysics. 1202
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 6, 2001
used extensively for the past 20 years as a gasoline additive (up to 15 wt %) to reduce automobile carbon monoxide and hydrocarbon emissions. The fact that MTBE is highly soluble in water (approximately 5 wt %) (3) and chemically inert when compared to other fuel constituents causes it to be often detected at high concentrations in groundwater in the vicinity of gasoline spills. The EPA has reported that low levels of MTBE in drinking water (above 40 µg/L) may cause unpleasant taste and odors and has designated MTBE as a possible human carcinogen (4). Past studies have concentrated on the reporting of MTBE levels in groundwater near gasoline spills. Happel et al. reported an MTBE occurrence rate of approximately 78% at locations where hydrocarbons have impacted groundwater (5). Johnson et al. estimate that 9000 leaking underground fuel tanks have caused MTBE contamination at community water supplies in the 31 states surveyed (excluding California and Texas) (6). Robbins et al. reported finding a significant number of MTBE detections in groundwater samples taken at sites in Connecticut known to be contaminated by heating oil spills (7). Later, this same research group reported finding MTBE contamination to range from 9.7 to 906 mg/L in heating oil and from 74 to 120 mg/L in diesel fuel in samples collected from storage tanks in Connecticut (8). The method used to analyze these samples was based on fuel-water partitioning and GC analysis. This present study provides the detailed basis for that analytical method. MTBE fuel-water partition coefficients as a function of temperature, which are critical to the method, are also presented. This study also reports on variations in MTBE levels as a function of time observed at several residences and a service station. Analytical results are reported for samples taken from terminals as part of an effort to assess the sources of MTBE in heating oil and diesel fuel.
Method Theory. The following derivation is adapted from a paper by Robbins et al. (9) Known volumes of clean water and heating oil are placed in a sample vial and agitated to equilibrium. When equilibrium is reached, the initial number of moles of MTBE in the heating oil is now divided between the fuel and water phases as is expressed in eq 1
no ) nf + nw
(1)
where no is the initial moles of MTBE in the heating oil, nf is the moles MTBE in heating oil after equilibration, and nw is the moles MTBE in water after equilibration. Both sides of eq 1 are divided by Vf, which is the volume of heating oil utilized in the equilibration.
no nf nw ) + Vf Vf Vf
(2)
Equation 2 can be rewritten as eq 3 through simple substitutions. Co ) (no/Vf) is defined as the initial MTBE concentration in the heating oil and Cf ) (nf/Vf) is defined as the MTBE concentration in heating oil after equilibration. Additionally, nw/Vf is multiplied by Vw/Vw (i.e., 1) where Vw is the volume of water used in the equilibration.
Co ) Cf +
( )( ) nw V w Vf V
(3)
Cw is defined as the MTBE concentration in water after 10.1021/es001355l CCC: $20.00
2001 American Chemical Society Published on Web 02/08/2001
equilibration. Using the fact that Cw ) nw/Vw, an additional substitution is made, and eq 3 is then rearranged as shown
Co ) Cf + Cw
() Vw Vf
(4)
The Cf term is eliminated by substituting Cf ) CwK, as K is defined as the fuel/water partition coefficient.
()
Co ) CwK + (Cw)
Vw Vf
(5)
Both sides of eq 5 are divided first by Cw.
()
Co Vw )K+ Cw Vf
(6)
and then both sides of eq 6 are divided by Co to obtain
( ) ( )( )
1 1 Vw 1 ) K+ Cw Co Co V f
(7)
Equation 7 may be used to determine KMTBE values. This process entails determining 1/Cw as a function of different fuel/water volume ratios (Vw/Vf). Because Co is held constant, a plot of 1/Cw versus Vw/Vf should result in a straight line with a y-intercept value of (1/Co)K and a slope of 1/Co. The value of K may now be determined by dividing the intercept by the slope. Equipment Utilized. All experiments were carried out using a Hewlett-Packard model 5890A gas chromatograph (Hewlett-Packard, Wilmington DE) equipped with an OI model 4430 photoionization detector (PID, OI Analytical, College Station, TX) which was serially connected to an OI model 4410 flame ionization detector (FID). The capillary GC column used in this study was a J&W DB-MTBE, 0.53 mm i.d, 30 m length, 3 µm film thickness (J&W Scientific, Folsom, CA). The serial arrangement of the detectors allowed the entire column eluent to be subjected to each detector, producing two independent detector responses. Two separate HP 3396A integrators (Hewlett-Packard, Wilmington, DE) allowed for the data collection. A Tekmar model 1020 Constant Temperature Water Bath-Shaker (Tekmar-Dohrmann, Cincinnati, OH), with a temperature range between 0 and 55 ((0.5) °C, was utilized for static headspace and vial temperature-sensitive sample equilibrations. A water/antifreeze solution was circulated through a cooling coil inserted into the shaker for uses when temperature equilibrations below 25 °C were required. Equilibration of heating oil and diesel fuel samples were completed using a Burrell Wrist Action Shaker 75 BB (Burrell Scientific, Pittsburgh, PA). For all runs, the gas chromatographic splitless injector was set at 210 °C and the detector temperatures were set at 230 °C. The GC oven was programmed to run isothermally at 35 °C for 3.5 min, then to be increased in temperature at 5 °C/min to 60 °C and finally increased in temperature at 70 °C/min and held at 250 °C for 6.5 min. The rapid temperature increase after 60 °C was utilized to quickly clear the column of the numerous higher boiling components found in the diesel/heating fuels that elute after MTBE. Collection of Fuel Oil Samples. An oil siphoning method was developed to collect samples from both aboveground storage tanks (AST) and underground storage tanks (UST). The method involved snaking disposable, 0.25 in. (0.64 cm) o.d. polyethylene tubing into a tank through the fill pipe. A 0.75 in. (1.9 cm) diameter ceramic ball was attached to the end of the tubing with a 0.13 in. (0.32 cm) diameter, threaded brass rod. The ball helped to snake the tubing around pipe elbows that were common in basement located ASTs. Once
the tubing was positioned in the tank, a disposable 60 mL plastic syringe was connected to it with 0.25 in. (0.64 cm) o.d. gasoline-resistant Tygon tubing. The fuel was siphoned from each tank and placed in two 40 mL VOA vials. Approximately 1 mL of headspace was left in the vials to allow for thermal expansion of the product. For secondary containment the VOA vials were placed in 250 mL widemouth Nalgene bottles containing absorbent pads and placed in a cooler with ice packs for transportation to the laboratory. For samples collected from commercial stations, diesel fuel was dispensed into cleaned, dedicated, one-gallon, plastic fuel containers. Subsequently, the diesel fuel was transferred to 40 mL VOA vials and transported to the laboratory. Samples collected at terminals were obtained using on-site fuel sampling devices or from collection ports. Several samples had been previously collected and stored on-site by terminal operators. On-site sampling equipment consisted of clean bottles or sampling devices dedicated to a particular fuel. Analytical Method. A known volume of fuel contaminated with MTBE is placed into a 40 mL VOA vial with a known quantity of water (care is taken to ensure that no headspace exists between the liquid phases). After equilibration, 10 mL of water is extracted and transferred into a 20 mL VOA vial where an additional equilibration of MTBE occurs with the headspace. This headspace is then sampled for MTBE concentration. The accurate volume of each 40 mL VOA vial was determined by weighing the vial empty and then completely filled with temperature-equilibrated water. Knowing the accurate weight of the water at a given temperature and using the water’s known density, an accurate volume of each vial was obtained. The partitioning of the MTBE between the fuel oil and water was accomplished as follows. A known quantity of deionized/distilled water was added to the vial. The vial was then overfilled with heating oil. The actual amount of heating oil added to the vial was determined by subtracting the accurately known water volume added from the total volume of the vial. For KMTBE determinations, the capped vials were placed into the water bath shaker and allowed to reach temperature equilibration for thirty minutes without agitation. The vials were then agitated by hand for 1 min and placed back into the water-bath where they were agitated for different periods of time. Heating oil and diesel fuel sample testing was completed as above except that 25 mL of water was utilized for the equilibration and all equilibrations were completed using the Burrell Wrist Action Shaker at ambient temperature 23 ((0.5) °C. After both chemical and thermal equilibrium were reached, the fuel/water layers were allowed to settle for 10 min with the vial in an inverted position (cap/septa down). While still inverted, a long needle (22 guage/3.5 in. long) was inserted through the heavy septa and into the fuel layer to vent the mixture and prevent turbulence during extraction. A second needle attached to a plastic disposable syringe was inserted into the water layer and greater than 10 mL of the equilibrated water was carefully removed. This water sample was placed into a 10 mL vial, sealed with a Teflon coated screw cap (without headspace) and stored at 4 ((0.5) °C until time for the GC analysis. For the GC analysis, 10 mL of the equilibrated water sample was transferred into a 20 mL sample vial. In addition, 4.4 g of NaCl was added to the 20 mL vial to increase the analyte partitioning into the headspace and improve method sensitivity. This vial was once again agitated by hand for one minute, placed into a water bath at 25 ((0.5) °C and gently agitated/equilibrated for 30 min. A 200 µL headspace sample was taken from the equilibrated water/headspace sample vial using a 250 µL gastight, locking syringe (SGE 250R-GSG with push/pull valve, 031906) and injected into the GC. The VOL. 35, NO. 6, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1203
FIGURE 1. Determination of the equilibration time for MTBE extraction from heating oil to water at 5 °C with Vw/Vf ) 4.4. Where Vw is the volume of water utilized in the equilibration and Vf is the volume of fuel. analysis of each headspace sample was repeated three times with the average of the resulting PID detector peak area counts employed.
Results and Discussion Equilibration Time Determination. In order for the relationship between the MTBE concentration in the extracted water phase and the water/fuel volume ratio (eq 7) to be linear, the system must be in a state of equilibrium. A study was performed to determine the agitation time required to reach equilibrium. Vials with the same fuel/water volume ratio were tested over a range of temperatures. In addition, vials with both high and low fuel/water volume ratios were compared. Figure 1 depicts a graph of MTBE peak intensities as a function of the equilibration time at 5 °C (the lowest temperature tested) with a water/fuel volume ratio of 4.4. Equilibration was reached in about 60 min at all temperatures and fuel/water variations tested. Hence, in all later tests a conservative time of 120 min was employed so that equilibration between the fuel/water would be assured. This equilibration time is in agreement with contact times of 60-90 min previously published (10, 11). KMTBE Variation With Temperature. To determine the value of the partition coefficient, KMTBE, from eq 7, the value of Cw must be known. For the purposes of this study, the analytical signal (Sigw) produced by MTBE from the headspace above the equilibrated water was utilized in lieu of Cw. Cw is related to Sigw, through a direct relationship utilizing a constant Ks. By substituting this Cw/Sigw relationship back into eq 7 the following is determined.
( ) ( )( )
Ks Ks Vw 1 ) K+ Sigw Co Co V f
(8)
When the value of K is determined by dividing the y-intercept by the slope, the constant Ks is eliminated as it appears in both the numerator and the denominator. Figure 2 depicts a graph of 1/Sigw versus Vw/Vf for the three different temperatures tested. The values of KMTBE were as follows: at 5 °C, K ) 3.8 ( 0.7; at 25 °C, K ) 8.0 ( 1.9; and at 40 °C, K ) 10.9 ( 1.4. KMTBE uncertainties were calculated by determining and combining the errors in the regression line slope and y-intercept line as described by Miller et al. (12). Additionally, a plot of the variation of KMTBE with 1204
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 6, 2001
temperature was obtained and is shown in Figure 3. The relationship between KMTBE and temperature was determined to be linear over the temperature range of 5-40 °C. The values of KMTBE determined by this study are generally about half those previously published for gasoline (11, 13, 14). Cline et al. indicated that their empirical results for gasoline constituents compared very well with predicted ideal behavior based on their equation for an ideal mixture of liquids. The equation predicts the partition coefficient as a function of organic liquid density, average molecular weight for the organic phase, and the aqueous solubility of the constituent of interest. Substituting a fuel oil density of 0.83 g/mL, a solubility of 0.54 mol/L (48 000 mg/L), and an average molecular weight for diesel fuel of 200 (15) into their ideal equation predicts a KMTBE value of 7.7 at 25 °C. This is in close agreement with the KMTBE value of 8 obtained here. MTBE Concentration Determination in Fuel Oil Samples. To obtain the concentration of MTBE in a fuel sample, analytical results were compared to those from a calibration curve. All calibration curves had a linear coefficient of determination (R2) of more than 0.998. After determining the MTBE concentration in the water, this value was related to the MTBE concentration in the fuel using eq 9.
Cf ) CwKMTBE
(9)
In all runs the signal average for three replicate injections per sample had a typical relative standard deviation (RSD) value of less than 3%. The ambient temperature recorded during the sample extraction was utilized to correct for temperature variations in KMTBE using Figure 3. Accuracy Verification. This method was found to have several advantages over the conventional GC/MS purge and trap method (EPA method 524.2). Though purge and trap has a lower method detection limit (MDL) than this static headspace method (0.5 ppb compared to 700 ppb), it is not possible to directly analyze petroleum products for MTBE by the GC/MS purge and trap method. This is due to the long run time (numerous late eluting/high boiling point compounds) and the possibility of fouling the purge-and-trap glassware, trap and transfer system, and the GC column. In addition, more dilution steps are required in the EPA method 524.2 for samples with higher MTBE concentrations. It was found that these dilution steps contribute large errors to the
FIGURE 2. Graphical determination of KMTBE at 5, 25, and 40 °C.
FIGURE 3. KMTBE as a function of temperature from 5 to 40 °C. Error bars are developed by combining slope and y-intercept regression errors from Figure 2. values for MTBE reported by the purge-and-trap, GC/MSD. The photoionization detector (PID) utilized in this method has a large linear dynamic range when compared to the purge and trap method (MDL to about 100 mg/L for the PID, in comparison from the MDL to about 80 ppb for the purge and trap GC/MS system). This large linear range eliminates the need to develop numerous calibration curves when using the PID detector. To relate this method to a known analysis method, a comparison study was completed. One unspiked sample and three samples of the same heating oil to which MTBE had been added were equilibrated utilizing the above-described method. Duplicate samples were analyzed by this static headspace method and by an independent analytical testing laboratory using EPA method 524.2. The comparative results are presented in Table 1. The “multiple t” test (with N ) 4) was utilized to compare the results obtained from the water
TABLE 1. Comparisons of the Results between the Developed Static Headspace GC Method Compared with Purge and Trap GC/MS (EPA Method 524.2)
heating oil sample
extraction method (mg/L MTBE in water)
GC/MS purge and trap (mg/L MTBE in water)
absolute difference
relative % error
unspiked B C D
0.15 2.3 7.2 10.4
0.16 2.8 7.8 14.3
0.01 0.47 0.63 3.9
7.2 16.8 8.1 27.4
equilibration method to the standard purge-and-trap GC/ MS method. A t value of 1.106 was determined which indicated that at a 95% confidence level, there is no statistical difference between the two methods. VOL. 35, NO. 6, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
1205
that both methods had the same precision because each was from the same population and was identically analyzed. A calculated value of t ) 0.94 determined that at the 95% confidence level there is no difference between the averages of the duplicate runs. In addition, the duplicate results had an overall average % RSD of less than 2%.
FIGURE 4. Coelution studysstandards with equal concentrations of MTBE and 2-methylpentane. Figure 4a is the FID chromatogram while Figure 4b is the PID chromatogram. GC conditions as stated in Experimental Section. In addition, a coelution study was completed. Lacy et al. determined 2-methylpentane to coelute with MTBE (16). The sample chromatogram supplied by J&W Scientific with the DB-MTBE column indicated 2-methylpentane to have the closest elution time to MTBE (17). The chromatographic results of a standard containing equal concentrations of MTBE and 2-methylpentane as run under our experimental conditions are depicted in Figure 4. Both panels, 4a for the FID and 4b for the PID, are from the same sample injection. Results indicated retention time resolution between the two peaks with 5.7 min. for MTBE and 6.0 min. for 2-methylpentane. The retention time resolution is most evident in the FID chromatogram due to its increased sensitivity toward 2-methylpentane when compared to the PID. Figure 4b shows that there is about a 300 times enhanced sensitivity for MTBE to 2-methylpentane as indicated by the indiscernible 2-methylpentane peak in the PID chromatogram. These two factors significantly eliminate the possibility of 2-methylpentane coelution with MTBE when utilizing the PID detector, the specified GC column, and the described analysis procedure. Reproducibility. Duplicate sample runs were completed on approximately 20% of the samples tested. The duplicate runs consisted of dividing the original fuel sample between two different extraction vials and completing an analysis of each sample. Results of the duplicate samples from these two sets of data were compared using the “paired t” test (with N1 and N2 ) 3). For this calculation, it was assumed
Determination of MDL. The MDL was determined using the EPA suggested method (18). Heating oil with a known concentration of MTBE (1.1 mg/L) was equilibrated with water as described above. The resulting equilibrated water (which had a MTBE concentration of about 0.15 mg/L) was diluted by a ratio of 1:5 (10 mL into 50 mL) for a final MTBE concentration of 0.03 mg/L. This MTBE concentration gave an analytical signal of approximately 5 times the signal/noise ratio. Three 10 mL VOA vials were overfilled and stored for GC testing. Each sample was prepared for GC analyses as indicated above and a total of eleven replicate runs were used to obtain an MDL as per ref 18. An MDL of 0.096 mg/L MTBE was determined for MTBE in water. This value was converted to a MTBE concentration in the fuel by utilizing eq 9 such that an MDL of 0.71 mg/L MTBE in the fuel was determined. Temporal Monitoring Results. As noted above, the results of a statewide residential and service station sampling effort were detailed in Robbins et al. (8) That effort entailed a onetime sampling event. To evaluate how MTBE levels in heating oil and diesel fuel vary temporally, fuel samples were collected at four heating oil tank locations and one diesel tank location over a 9 month period. Temporal data are presented in Figure 5. The Manchester and Norwalk heating oil tanks showed low MTBE levels with little variance. The Colchester and Ashford tanks had higher MTBE concentration and variance (over a factor of 10 times). The diesel tank showed considerable variance with time (up to a factor of about 30 times). The temporal variations observed likely reflect a combination of factors including the MTBE concentration of the fuel in the tanks prior to delivery, the MTBE concentration in the delivered fuel, and the amount of fuel delivered. Three tanks were sampled at the beginning of the study in August 1999 and then again in November 1999. Over the 3-month period these tanks were not used or refilled. These results are listed in Table 2. In two of the tanks the MTBE concentration in the fuel decreased significantly over this
FIGURE 5. Temporal variations in the MTBE concentration in heating fuel from samples taken November 1999 to April 2000. 1206
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 6, 2001
TABLE 2. Variances in MTBE Concentration Stored in Unused Fuel Tanks sample location Colchester Ashford Manchester
fill date
sample date
MTBE in fuel (mg/L)
dispersion period
% difference
11-Jun-99 11-Jun-99 25-May-99 25-May-99 22-Apr-99 22-Apr-99
19-Aug-99 18-Nov-99 13-Aug-99 11-Nov-99 10-Aug-99 11-Nov-99
107.6 71.1 67.1 39.4 17.7 17.9
3 months
-51
3 months
-70
3 months
1
TABLE 3. Results of the Terminal and Barge Study by Terminal Locationa sample
date delivered
New Haven, tank 21 New Haven, northeast dock East Hartford, diesel tank 2 East Hartford, heating tank 4 East Hartford, diesel mid East Hartford, heating start East Hartford, heating mid Bridgeport, diesel Bridgeport, heating New London, tank 7
20-Sep-99 20-Sep-99 unknown unknown 10-Dec-99 10-Dec-99 10-Dec-99 unknown 05Jan-00 13-Jan-00
New Haven, barge D New Haven, barge R New Haven, barge O New London, barge
11-Aug-99 27-Aug-99 20-Sep-99 13-Jan-00
date sampled
MTBE in water average (mg/L)
std. dev
RSD in %
MTBE in oil (mg/L)
Terminals 29-Oct-99 29-Oct-99 10-Dec-99 10-Dec-99 10-Dec-99 10-Dec-99 10-Dec-99 05-Jan-00 05-Jan-00 13-Jan-00
0.2 1.0 0.8 4.6 0.0 1.9 2.9 1.1 4.3 4.9
0.00 0.01 0.05 0.04 0.01 0.04 0.04 0.01 0.08 0.14
1.47 1.15 5.99 0.91 18.97 1.92 1.22 0.84 1.90 2.88
1.8 7.5 5.8 34.2
