Environ. Sci. Technol. 2004, 38, 234-239
Field Screening of Waterborne Petroleum Hydrocarbons by Thickness Shear-Mode Resonator Measurements MICHELLE S. APPLEBEE,† JOHN D. GEISSLER, ADAM P. SCHELLINGER, RICHARD J. JAEGER, AND DAVID T. PIERCE* Department of Chemistry, University of North Dakota, Grand Forks, North Dakota 58202
An inexpensive, field-portable sensor for direct, aggregate determination of aqueous petroleum hydrocarbons (PH) down to sub-ppm levels was developed. The basis of this sensor was an unusual, highly nongravimetric frequency response of 10 MHz (series fundamental) AT-cut quartz crystals when coated with rubbery silicone films. The response depended linearly and reliably on the total concentration of dissolved hydrocarbons over a range of 0.01-100 mg‚L-1 or up to aqueous solubility limits. Calibration sensitivities were measured individually for laboratory-prepared solutions of BTEX (benzene, toluene, ethylbenzene, and xylene isomers) and C6-C8 aliphatic components. Each component demonstrated a method detection limit (MDL) in the lowto sub-ppm range (benzene 10 mg‚L-1, n-hexane 0.54 mg‚L-1) for light coatings of a commercially available poly(dimethylsiloxane) gum (OV-1, >106 g‚mol-1) and lower MDLs for heavier coatings. Pairwise responses for the aliphatic and benzenoid standards were additive, indicating that aggregate determinations of mixtures (especially light fuels) were possible. Natural matrix interferences caused by sample turbidity and ionic strength were overcome by simple preparative methods. Fuelspiked natural waters were determined with respect to standards and verified by gas chromatography. A 0.19 mg‚L-1 MDL for gasoline was obtained for heavy OV-1 films. Field determinations of groundwater surrounding a leaking underground fuel tank demonstrated that the sensor and method were useful for on-site PH screening. Large differences between the equilibration times of aliphatic and benzenoid components also indicated one avenue for BTEX speciation with the device.
Introduction Most hydrocarbon contaminants found in natural water sources are related to petroleum products. Petroleum hydrocarbons (PHs) find their way into natural water sources by way of spills, improper disposal, and leaking storage tanks. Certain components of the gasoline and diesel cuts are known to be toxic (1), and legislation in the United States has targeted * Corresponding author phone: (701)777-2942; fax: (701)777-2331; e-mail:
[email protected]. † Current address: Department of Chemistry, Elmhurst College, Elmhurst, IL. 234
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 1, 2004
the remediation and control of major environmental sources, such as underground storage tanks (2). Methods for the determination of waterborne PHs typically require extraction from the aqueous matrix (liquidliquid, purge/trap, solid phase, and solid microphase) prior to determination by gas chromatography or spectroscopy (3, 4). For spectroscopic methods employing sorbent-coated waveguides (LIF (5, 6) and UV absorption (7), IR (8)), the extraction and determination steps can be carried out simultaneously with analysis, thus allowing direct probing of waterborne contaminants. Detection limits for these methods range from the low ppb (LIF) to low ppm range (IR). However, sizable instrument costs or limited field portability are often undesirable characteristics of these methods. Less specific or precise methods can also be valuable tools for initial spill investigations and subsequent remediation of petroleum release sites as long as they are rapid, simple to use, and have a low cost per sample. Screening methods of this nature include headspace analysis by flame ionization detection (e.g., EPA method 25A) (9), colorimetric analysis for total organic carbon (TOC) (10), immunoassay (e.g., EPA SW-846 method 4030) (11), and turbidimetric analysis (e.g., EPA SW-846 method 9074) (12, 13). When used in the field, such methods can improve sampling efficiency and decrease project costs associated more specific, precise, and expensive laboratory-based methods. Acoustic wave sensors based on the thickness shear-mode resonance (TSMR) of AT-cut quartz-crystal wafers are inexpensive and require little accompanying circuitry or equipment, thus making them attractive platforms for screening applications in the field. The resonant frequency response of these devices is often gravimetric, as with the quartz-crystal microbalance (QCM). This behavior was first described by Sauerbrey in 1959, wherein a positive change of mass that was rigidly attached to the crystal electrode resulted in a negative change in the resonance frequency from that of the bare electrode of the resonator (14). Although applied most often in gas or headspace sensors (15-17), TSMR devices have recently been used to directly determine waterborne organic compounds by allowing contact between liquid samples and permselective polymer films coated on one face of a resonant crystal. Josse et al. used such TSMR configurations for the determination of individual organic solvents in water (18, 19). Lucklum and others performed numerous studies of organic pollutants in water using TSMR sensors with various types of films including a variety of polymers (e.g., polyacrylates, polybutadienes, polystyrenes, and silicones) (20-22). Most of the films showed gravimetric response and gave detection limits in the low to sub-ppm range. However, in certain cases where thicker films were employed or samples had high ionic strength (e.g., seawater), some polymers showed significant nongravimetric contributions to the sensor response. Although sources of nongravimetric behavior have been aggressively explored for the past decade (23-25), no study to date has demonstrated that nongravimetric contributions can be reliably applied to analytical determinations. Herein, extensive characterization is provided for a reliable TSMR sensor that responds directly but in a nongravimetric fashion to concentrations of waterborne hydrocarbons. Work was conducted to evaluate the response for individual waterborne hydrocarbons as well as aggregates of petroleum hydrocarbons including gasoline. Work was also conducted to minimize matrix effects such as ionic strength, since these factors play important roles in the analysis of natural water 10.1021/es0344866 CCC: $27.50
2004 American Chemical Society Published on Web 11/25/2003
samples and adversely affect TSMR response (26). The method was evaluated in the laboratory with gasoline-spiked natural waters and also in the field at the remediation site of a leaking underground fuel storage tank.
Experimental Section Reagents. Methanol (MeOH), tetrahydrofuran (THF), chlorobenzene, BTEX (benzene, toluene, ethylbenzene, and xylene isomers), and aliphatic (n-hexane, n-heptane, n-octane, and 2-methylpentane) compounds were all of spectroscopic grade and used as received from Fisher (Fair Lawn, NJ), SigmaAldrich (St. Louis), or EM Science (Gibbstown, NJ). Poly(dimethylsiloxane) (OV-1; gum, 0.98 g‚mL-1, >106 g‚mol-1) was purchased from Alltech (Deerfield, IL). Commercial gasoline (87% octane rating) was obtained locally and stored with zero headspace at -18 °C in brown-glass bottles with Teflon-lined caps. Milli-Q reagent water (18 MΩ‚cm) was used to prepare all aqueous solutions other than natural water samples. Natural Samples. PH-free natural surface water was collected locally and, after an initial confirmation by purgeand-trap/gas chromatography, stored at room temperature until use. PH-contaminated water samples were collected locally at four groundwater monitoring wells surrounding the remediation site of a leaking underground gasoline storage tank. Contaminated wells were 25 m (MW1), 120 m (MW2), and 180 m (MW3). A fourth well 220 m from the source (MW4) did not show PH contamination by purgeand-trap/gas chromatography. Instrumentation. Polished 10 MHz AT-cut quartz crystals (1.5 cm diameter) with vapor-deposited gold electrodes and flag contacts (electrodes 0.51 cm diameter, 100 nm Au over 30 nm Cr) were purchased from International Crystal Manufacturing (Oklahoma City, OK). The piezoelectric sensing area of each crystal was ca. 0.205 cm2. After application of a polymer film, crystals were rigidly bonded to an HC-48/U crystal holder using conductive silver paint (SPI Supplies, West Chester, PA). A commercial acrylic flow-through cell (Universal Sensors Inc., Metairie, LA) was used to hold 70 µL of sample solution in contact with the film-coated side of a crystal during measurement. Because the opposing cell wall was slanted to aid the removal of air bubbles, the distance between the coated crystal surface and the opposing cell wall varied from 2 to 4 mm. Lengths of 1/16 in. PTFE tubing and Upchurch fittings at the cell inlet and outlet were used to convey solution through the sample cavity. TSMR measurement was performed by driving the fundamental, series-resonant frequency (fs) of the crystal using a home-built TTL(LS) oscillator circuit (27). The HC-48/U crystal holder plugged directly into a matching socket on the oscillator board to limit mechanical strain on the crystal holder or electrode flags. To insulate the system from minor changes in room temperature in the laboratory, the cell and oscillator were placed inside a foam block that maintained the crystal at a constant 30.5 ( 0.5 °C. The buffered, 50-Ω output of the oscillator was measured to (0.01 Hz by a HP 53131A frequency counter (HewlettPackard, Loveland, CO). Timed data from the HP 53131A were transferred to a PC via an internal IEEE card (National Instruments, Austin, TX) using HP Benchlink software. For measurements in the field, the cell, oscillator, and DCpowered frequency counter (B&K Precision 1803C, (1 Hz) were placed rigidly inside a thermoelectric ‘cooler’ (Powerchill 40 Qt, Coleman) that was modified with a precise DCpowered temperature controller (Oven Industries, TS67178). A 12 V automotive battery provided power, with the temperature controller drawing directly from the source and the oscillator/counter devices drawing through a regulator (LM317, 7.5 V) and filter (LC π). Tests showed that the crystal temperature could be adjusted from 4 to 65 ( 0.5 °C with this
apparatus. All field measurements were conducted with the cell at 38 ( 0.5 °C. At ambient temperatures in the range from 20 to 30 °C, the entire system drew an average power of ca. 48 W (4 A). Electrical impedance measurements of bare and filmcoated crystals, recorded as admittance (|Y|) and phase (φY) versus frequency, were performed with a HP E4916A Crystal Impedance/LCR Meter (Hewlett-Packard, Loveland, CO) controlled by a PC via an internal IEEE card using a program written in HP VEE, version 4.0. All crystals used for study were typically screened before use, and the impedance data were modeled using a Butterworth-Van Dyke (BVD) lumpedelement equivalent circuit (motional-branch RLC elements with parallel-branch static capacitance) (23). Motional resistances of bare crystals were typically 7 ( 3 Ω in air and 150 ( 10 Ω under water. Solid-phase microextraction/gas chromatography with flame ionization detection (SPME/GC/FID) was performed with a HP Series II 5890 GC/FID unit interfaced to a HP 3396 Series II integrator. A solid-phase microextraction holder with 100 µm OV-1 coated fibers (Supelco Inc., Bellefonte, PA) was used for sample introduction, and a 15 m DB-1 capillary column (J & W Scientific, Fulsom, CA) with an inner diameter of 0.32 mm and film thickness of 0.25 µm was used for separation. Purge-and-trap/gas chromatography with flame ionization detection (PT/GC/FID) was performed with an OI Analytical 4460A purge-and-trap concentrator equipped with a type E trap (Supelco Inc., Bellefonte, PA) and 10-mL sparger. The concentrator was coupled to a HP Series II 5890 GC with flame ionization detection and interfaced to a HP 3396 Series II integrator. The GC was equipped with a 30-m DB-5 capillary column (J & W Scientific, Fulsom, CA) with an inner diameter of 0.32 mm and film thickness of 0.25 µm. Volatile organics were purged for 11 min with helium at a flow rate of 40 mL/min and then desorbed from the trap at 180 °C for 4 min. The desorbed components were analyzed by GC/FID (injector port, 200 °C; temperature program, 5 min at 30 °C for, 10 °C‚min-1 until 200 °C, 3 min at 200 °C; detector, 250 °C). TSMR Preparation. Polymer solutions were prepared by dissolving 0.1 g of the OV-1 silicone in 25 mL of distilled THF. The same polymer solution was used to prepare two different types of films. Wipe-coated films were prepared by depositing 50 µL of the polymer solution onto one electrode of the quartz crystal and placing the crystal in an oven to cure at 105 °C for 1 h. After the crystal was cool, the polymer film was gently wiped with a soft, lint-free tissue to remove sufficient material to reach a desired frequency change of maximum conductance relative to the bare crystal (δf ). Spin-coated films were prepared by depositing a 100 µL of the polymer solution onto one electrode of the quartz crystal and spinning the crystal at a sufficient speed (1000-4000 rpm) to reach a desired δf. Spin-cast films were cured at 105 °C for 1 h. Freshly coated crystals were characterized by impedance analysis at room temperature prior to their use as TSMR sensors. Films were acceptable if the coated crystal showed a desired resonance frequency change from the uncoated crystal (δf of the conductance maximum, usually -1, -3, -6, or -10 kHz) and a phase change of more than 40° within 5 kHz of the resonance frequency. Reproduction of these parameters within (10% yielded consistent behavior of the sensor. Phase changes of less than 40° indicated an excessive acoustic energy loss by the film that would cause inconsistent sensor behavior or even a failure to resonate upon contact with water solutions. For acceptable films, fitting of impedance data to the BVD equivalent circuit typically yielded motional resistance values (air vs water) of 190 Ω vs 450 Ω (δf ) -3 kHz, wipe coated), 190 Ω vs 500 Ω (δf ) -6 kHz, spin coated), 200 Ω vs 550 Ω (δf ) -10 kHz, spin coated). VOL. 38, NO. 1, 2004 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
235
TABLE 1. Calibration Data and Figures of Merita for Analysis of Individual Hydrocarbons and Automotive Gasoline in Distilled Water with 0.4% MeOH compd
slope, m (Hz‚ppm-1)
intercept, b (Hz)
r2
LR
LOD (ppm)
MDL (ppm)
% RSD
benzene toluene ethylbenzene p-xylene o-xylene m-xylene n-hexane n-heptane n-octane 2-ethylpentane gasoline
3.34 10.6 26.1 22.1 14.5 11.7 134 181 138 75.0 99.6
4.03 9.42 -0.50 -9.80 -5.74 -2.62 21.1 49.8 -28.5 80.0 10.0
0.9993 0.9980 0.9978 0.9989 0.9978 0.9937 0.9944 0.9878 0.9799 0.9871 0.9975
225 714 727 615 250 320 2000 2000 2000 1333 667
0.89 0.28 0.11 0.13 0.32 0.25 0.02 0.02 0.02 0.04 0.03
6.77 1.63 0.33 0.37 0.74 0.87 0.03 b b 0.16 0.19
1.19 2.55 2.11 1.63 2.64 4.62 5.26 7.83 14.3 9.85 4.52
a
TSMR sensor with spin-coated OV-1 film (δf ) -10 kHz).
b
Not determined.
TSMR Calibration. Stock aqueous solutions of hydrocarbons or gasoline were prepared by dissolving microliter volumes of the analyte(s) in 2 mL of MeOH. This solution was added to a 500 mL volumetric flask containing 480 mL of distilled water and diluted to the mark. This 500 mL flask was capped with a dry ground glass stopper. Final solutions were place in 25 mL EPA vials and immediately capped with Teflon-coated silicone septa (23 mm × 85 mm, National Scientific), ensuring that no headspace remained in the vial. To limit losses by volatility, dilution and subsequent analysis was performed within 3 h. Before use, a TSMR resonator was equilibrated in the flowthrough cell by making successive 1 mL injections (Hamilton, 1000 series gastight syringe) of a 0.4% MeOH(aq) blank solution until a stable, baseline frequency was achieved (ca. six injections over 30 min). Calibration data were recorded by making repeated 1 mL injections of an analyte standard at 5 min intervals until a stable series resonance frequency was obtained (i.e., within (2 Hz). Once a stable frequency was obtained for a solution, another standard of higher analyte concentration was injected. Series resonance frequency changes (∆fs) were determined by subtracting the frequency obtained for the blank solution from the stable frequency obtained for each standard solution (28). Following calibration, the original baseline frequency was reestablished by flushing the cell at a rate of 2 mL‚min-1 with 50% aqueous methanol solution for a minimum of 300 mL. Limits of detection (LOD), method detection limits (MDL), and percent relative standard deviations (% RSD) were determined according to guidelines listed in Standard Methods (3). Linear ranges (LR) were determined as a ratio of the upper linear limit to the LOD. In all cases, the upper linear limit was imposed by the solubility limit of the component in the 0.4% MeOH(aq) blank solution. Values of % RSD were determined from three individual calibration curves. TSMR Determinations with Natural Waters. Solutions of natural surface or groundwaters were usually injected into the cell through an in-line filter to prevent suspended solids from settling on the sensor surface. Sensors treated in this manner displayed good reproducibility and lasted approximately 3 months without deterioration. Comparison Methods. TSMR results for gasoline-spiked samples were verified by comparison with the SPME/GC/ FID method. Split solutions were placed in a 25 mL EPA sample vial containing a small Teflon stir bar and capped with zero headspace using a Teflon-lined silicone septum. The stirred solution was sampled with a SPME extended fiber for 15 min, after which the fiber and holder were quickly removed from the sample and the extracted components were analyzed by GC/FID (injector port, 250 °C; temperature program, 5 min at 30 °C for, 10 °C‚min-1 until 200 °C, 3 min at 200 °C; detector, 300 °C). 236
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 38, NO. 1, 2004
Results and Discussion Individual Hydrocarbons. Calibration data (∆fs vs ppm) were measured individually for solutions of 10 hydrocarbons and gasoline using TSMR crystals with two different amounts of OV-1 film coating. The calibration data and corresponding figures of merit shown in Table 1 were measured with a heavily spin-coated crystal (δf ) -10 kHz). Related data for a lightly wipe-coated crystal are available as Supporting Information. The data illustrated a linear but nongravimetric (i.e., positive Hz‚ppm-1) response regardless of the hydrocarbon or film amount. Because the positive frequency change was reversible in all cases, film loss from the crystal surface was not a source of the unusual positive response. Responses were also found to correlate directly with octanol:water partition coefficients (Kow ) Coctanol/Cwater) for all of the hydrocarbons except octane (29). Because this correlation held for the majority of hydrocarbons tested (from benzenoids to branched alkanes), it suggested that the process causing TSMRs unusual response was influenced mostly by the concentration of hydrocarbon within the film and not by the type of hydrocarbon. Hydrocarbons Mixtures. To determine if the individual responses of individual hydrocarbons were additive, calibration data were also measured for solutions containing hydrocarbon pairs. Data obtained in triplicate for one- and two-component solutions of m-xylene and n-hexane are available as Supporting Information. In most cases, the measured frequency changes agreed at the 90% confidence level with the frequency changes calculated as a linear combination of calibration responses for the individual hydrocarbons (eq 1).
∆fs(calculated) )
∑ {[PH ,ppm]‚(m ,Hz‚ppm i
i
i
-1
)} (1)
These data confirmed that the OV-1-based sensors produced an additive response that correlated with the aggregate concentrations of hydrocarbons in solution. This additive response occurred regardless of the paired hydrocarbons or their relative proportions. Such behavior could only occur if sorption of one component did not significantly influence the sorption or nongravimetric response of another component. Overall, the highly nongravimetric response of this sensor appears to be unique among acoustic wave devices that have been previously developed for the determination of waterborne PHs. The results with laboratory solutions indicate that the unusual response can be used to perform reliable determinations of aggregate PHs in water. Equilibration Time and Mass-Transfer Effects. The time required for the sensor to reach a steady-state frequency after changing solution was measured for various hydro-
FIGURE 1. Sensor equilibration and recovery curves measured using a TSMR sensor with a wipe-coated OV-1 film (δf ) -1 kHz). Frequency data were recorded immediately prior to injection of fresh solutions at intervals of 2 (open symbols) or 10 min (solid symbols). Equilibrations were performed with solutions of 60 ppm m-xylene (circles) or 20 ppm n-hexane (squares). Recoveries were performed with a blank solution containing 0.4% methanol. carbon standards. Concentrations were selected to ensure the same absolute frequency change of ca. 40 Hz and thereby the same equilibrium concentration of hydrocarbon in the film. Using a lightly wipe-coated crystal (δf ) -1 kHz), each standard was injected in 2 min intervals until equilibrium was reached, followed by injections of blank solution (0.4% aqueous MeOH) in 2 or 10 min intervals until the frequency returned to its initial value. Data for m-xylene and n-hexane, as representative benzenoid and aliphatic hydrocarbons, are shown in Figure 1. BTEX compounds were found to equilibrate very rapidly with OV-1 (