Article pubs.acs.org/ac
Identification and Quantification of Aqueous Aromatic Hydrocarbons Using SH-Surface Acoustic Wave Sensors Florian Bender,† Rachel E. Mohler,‡ Antonio J. Ricco,§ and Fabien Josse*,† †
Department of Electrical and Computer Engineering, Marquette University, Milwaukee, WI 53201-1881, United States Chevron Energy Technology Co., 100 Chevron Way, Richmond, CA 94801, United States § Department of Electrical Engineering, Center for Integrated Systems, Stanford University, Stanford, CA 94305-4075, United States ‡
ABSTRACT: A need exists for compact sensor systems capable of in situ monitoring of groundwater for accidental releases of fuel and oil. The work reported here addresses this need, using shear horizontal surface acoustic wave (SH-SAW) sensors, which function effectively in liquid environments. To achieve enhanced sensitivity and partial selectivity for hydrocarbons, the devices are coated with thin chemically sensitive polymer films. Various polymer materials are investigated with the goal of identifying a set of coatings suitable for a sensor array. The system is tested with compounds indicative of fuel and oil releases, in particular, the BTEX compounds (benzene, toluene, ethylbenzene, and xylenes), in the low milligrams/liters to high micrograms/liters concentration range. Particular emphasis is placed on detection of benzene, a known carcinogen. It was observed that within the above concentration range, responses to multiple analytes in a mixture are additive, and there is a characteristic response time for each coating/analyte pair, which is largely independent of concentration. With the use of both the steady-state and transient-response information of SHSAW sensor devices coated with three different polymer materials, poly(ethyl acrylate), poly(epichlorohydrin), and poly(isobutylene), a response pattern was obtained for benzene that is easily distinguishable from those of the other BTEX compounds. The time courses of the responses to binary analyte mixtures were modeled accurately using dual-exponential fits, yielding a characteristic concentration-independent time constant for each analyte/coating pair. Benzene concentration was quantified in the aqueous phase in the presence of the other BTEX compounds.
A
application. For these reasons, polymers are by far the most common materials for hydrocarbon sensing.10 The performance of the sensors and coating materials investigated in this work has recently been studied under varying environmental conditions, including changes in temperature, acidity, and salinity.11 The BTEX compounds (benzene, toluene, ethylbenzene, and xylenes) are present in crude oil and its refined products12 and are regulated by government agencies.13 Therefore, these aromatic compounds have been selected as indicators of fuel and oil releases for the present work. The BTEX compounds can be detected by polymer-coated SHSAW sensors in liquid phase at low concentrations. Among the BTEX compounds, benzene is of particular concern due to its carcinogenicity;13 thus, it is of paramount importance, and in some cases a legal requirement, to measure the concentration of benzene in water. This task is challenging, not only because relevant concentrations are in the low ppm (mg/L) to low ppb (μg/L) range5,13 but also because of the frequent presence of similar aromatic compounds, including toluene, ethylbenzene, and xylenes. Note that among the BTEX compounds, benzene shows the highest solubility in water,14
ccidental releases of fuel and oil from storage tanks, pipelines, and other sources into groundwater, lakes, rivers, and oceans pose potential threats to the environment and public health.1−4 Timely detection of such releases can minimize the impact on public health and reduce environmental damage and cleanup costs. Consequently, periodic monitoring is in many cases required by law.1 The current practice (the analysis in the laboratory of field-collected samples),2 however, is too time-consuming for continuous monitoring. Therefore, compact sensor systems are under development that will allow frequent or continuous in situ monitoring of groundwater and surface water at critical sites.5 Sensor systems based on SH-SAWs (shear-horizontal surface acoustic waves) represent a promising approach for this application. SH-SAWs combine the advantages of high surface confinement of the acoustic energy (which facilitates high sensitivity to the analyte) and the ability to propagate along a solid−liquid interface without excessive acoustic wave attenuation.6,7 For this work, a 103-MHz 36° YX-LiTaO3 sensor platform was selected.8,9 In order to enhance sensitivity and provide partial selectivity for the analytes of interest, the sensors were coated with thin (≤1 μm) organic polymer films. Many polymers show favorable physical and chemical characteristics for hydrocarbon sensing, are easy to prepare, and offer the possibility of tailoring their chemical structure for a particular © 2014 American Chemical Society
Received: November 15, 2013 Accepted: January 7, 2014 Published: January 7, 2014 1794
dx.doi.org/10.1021/ac403724f | Anal. Chem. 2014, 86, 1794−1799
Analytical Chemistry
Article
resulting in lower polymer/water partition coefficients11,15 and lower sensitivities for polymer-coated sensor devices. Thus, a sensor system for quantification of benzene in the presence of other BTEX compounds must provide a high degree of selectivity. A common approach to providing selectivity toward an analyte lacking a unique functional group is the use of an array of sensors with different coatings, each showing a degree of partial selectivity for the analyte based on weak physicochemical interactions.16 After evaluation of the steady-state sensor responses using suitable pattern recognition algorithms, identification of the analyte will in many cases be possible.17 Previous investigations have shown that this approach is feasible for the detection of BTEX in water,11 but the degree of selectivity achieved was limited. On the other hand, the various BTEX compounds show very different solubilities in water14 and, consequently, very different partition coefficients for common polymers in contact with the aqueous phase.11,15 Since the partition coefficient is related to the dynamics of analyte sorption/desorption into and out of the polymer, this feature can be exploited by evaluating both the steady-state frequency shift and the time-dependent behavior of the sensor response. This approach will be incorporated into the final system design, together with the use of an array of sensors coated with different polymers, to achieve identification and quantification of benzene in water in the presence of other interferents. As long as the interaction between a polymer and an analyte is nonspecific (physisorptive) and analyte concentrations are sufficiently low, the sorption of the analyte into the polymer from the aqueous phase can be described by a single partition coefficient,15 and Henry’s law can be applied.18 As a result, at low concentrations the response time for sorption of an analyte into a given coating is independent of analyte concentration in the aqueous phase, and responses to multiple analytes are additive. Because one of the main objectives here is to quantify benzene in a mixture, it was critical to first investigate the validity of Henry’s law in the concentration range of interest for the various polymers (i.e., low parts per million to parts per billion range). It is demonstrated that the response time is specific to each coating/analyte pair, and unique response patterns can be obtained for benzene, toluene, and ethylbenzene. (Note that ethylbenzene and o-, m-, and p-xylene are all chemical isomers, and to distinguish between them has proven to be beyond the capabilities of the approach described in this paper.) Binary mixtures of aromatic analytes were investigated and the resulting response curves were evaluated using dual-exponential fits in order to determine benzene concentration in the presence of a single chemical interferent. Finally, the accuracy of this approach was determined for the current sensor system design.
Δfa − m Δfc − m
=
ma ma = mp + m w Vc(ρp + Cw )
(1)
In eq 1, Δfa−m describes the frequency shift due to the sorbed mass of the analyte; Δfc−m is the frequency shift due to the mass of the coating (including sorbed water); ma, mp, and mw denote the mass of the sorbed analyte, the polymer (i.e., the coating material), and the water sorbed in the coating, respectively (mw is assumed not to change during analyte sorption); Vc is the volume of the coating; ρp is the polymer density; and Cw is the concentration of water (in units of mass/volume) sorbed in the coating. The distribution of an analyte between a sorbent polymer phase and an adjacent aqueous phase at equilibrium can be described by the partition coefficient, Kp−w: K p−w =
Ca m /V = a c Camb Camb
(2)
In eq 2, Ca and Camb are the concentrations of the analyte in the coating and in the aqueous phase (ambient), respectively. For two of the polymer coatings investigated in this work, polymer/water partition coefficients for BTEX compounds have been calculated previously15,21 and are listed in Table 1. Table 1. Partition Coefficients, Kp−w, for Two Polymers, Poly(epichlorohydrin) (PECH) and Poly(isobutylene) (PIB), Calculated for the BTEX Compounds15,21 analyte
Kp−w (PECH)
Kp−w (PIB)
benzene toluene ethylbenzene xylenes
67 249 673 700
30 140 464 458
For the third polymer coating investigated in this work, poly(ethyl acrylate) (PEA), no values were calculated in the above references. However, from the experimental results presented below, taking into account the different thicknesses of the polymer coatings investigated, it can be estimated that the partition coefficients for PEA are similar to those of PECH. Note that an exact calculation cannot be made based on the experimental results since the sensitivity of the sensor will depend on the acoustic waveguiding effect,7 which in turn depends on the thickness and the mechanical properties of the polymer coating. Substituting eq 2 into eq 1 gives: Δfa − m = Δfc − m
CambK p − w ρp + Cw
(3)
It should be noted that surface waves (including the SHSAW) are also sensitive to the viscoelastic properties of a surface coating.22 In addition, the polymer thicknesses investigated in this work are between 0.6 and 1.0 μm, corresponding to between 1.5 and 2.5% of the acoustic wavelength (40 μm). For this range, a small phase lag will occur between the top and bottom surfaces of the polymer. Moreover, analyte sorption tends to soften or plasticize the polymer, leading to a change in this phase lag.19 As a result, an additional contribution to the frequency shift due to viscoelastic effects, Δfa−v, is observed. However, in this study it was experimentally determined that for small analyte concentrations, this contribution is also proportional to the analyte concentration in the coating and, thus, to Δfa−m:
■
THEORY As stated above, both the steady-state and transient information associated with a sensor response can be useful for analyte identification. Both components will be investigated for the case of small analyte concentrations in the following. Steady-State Response. Initially, the sensor coating will be assumed acoustically thin (i.e., the entire coating vibrates in phase with the substrate surface).19 Ignoring effects other than mass loading, the following equation describes the observed frequency shift due to the presence of the analyte:20 1795
dx.doi.org/10.1021/ac403724f | Anal. Chem. 2014, 86, 1794−1799
Analytical Chemistry
Article
Δfa − v = κ Δfa − m
Note that the baking step was found to be crucial in order to ensure reproducible results when the polymer-coated sensors were used over extended periods of time (up to 3 months). The sensor platform utilizes a dual-delay-line design, with the second delay line coated with poly(methyl methacrylate) (Scientific Polymer Products) and baked for 120 min at 180 °C, resulting in a glassy, effectively nonsorbent coating. The purpose of this reference line is to compensate for the influence of temperature drifts and other secondary effects. All BTEX analytes had purities of ≥98.5%, were purchased from SigmaAldrich, and used without further purification. The experimental setup consisted of a network analyzer (Agilent 8753ES) and a switch/control system (Agilent 3499A) to switch between the two SH-SAW delay lines. For some measurements, an Agilent E5061B network analyzer was used (switch/control system: Agilent 34980A); in this case, the averaging function of the instrument was turned on to reduce random data noise. The sensor was placed inside a liquid flow cell made in-house which has a volume of 130 μL. A peristaltic pump (Eppendorf EVA/Ismatec Reglo Digital MS) was used and set to a sample flow rate of 0.4 mL/min, resulting in a sample exchange time of 20 s. The temperature was held constant at 22.0 ± 0.1 °C. The data sampling time was 30 s, which is currently limiting the accuracy in analyzing the fastest responding component (benzene). All data shown are corrected for linear baseline drift. BTEX samples were prepared by pipetting the required amount of the BTEX compound into a vial filled with deionized water, followed by magnetic stirring for at least 120 min. In order to minimize loss of the volatile analytes, the amount of headspace in the vial was kept as small as possible while avoiding any spill, and Teflon lined caps and Teflon tubing were used throughout. Samples were used within 24 h after preparation.
(4)
where κ is a constant.20,22 This yields a total frequency shift, Δf 0, given by Δf0 = Δfa − m + Δfa − v = Δfc − m
CambK p − w ρp + Cw
(1 + κ ) (5)
In practice, if the mechanical properties of the coating are unknown, it is difficult to separate the two contributions, Δfa−m and Δfa−v, in the experimentally observed frequency shift. It should also be noted that if the polymer coating becomes thicker than a few percent of the acoustic wavelength, κ will no longer be constant.22 Coating thicknesses larger than 1 μm were therefore avoided in the present study. Transient Response Information. To obtain the transient response, it is first necessary to model the time response of the sensor. It is first assumed that the rate of analyte absorption at time, t, is proportional to the difference between the concentration of analyte in the coating at equilibrium and the concentration of analyte in the coating at time, t. This leads to a first-order absorption model given by23 Cȧ (t ) =
1 [K p − wCamb(t ) − Ca(t )] τ
(6)
where Ċ a(t) = dCa(t)/dt and τ is a response time that depends on the type of analyte, coating material, and coating thickness but not on analyte concentration. For a step change in ambient concentration of a single analyte, using eqs 5 and 6, it is found that the resulting frequency shift as a function of time has an exponential form: Δf (t ) = Δf0 [1 − exp( −t /τ )]
(7)
where Δf 0 is given by eq 5. If a simultaneous step change in the ambient concentrations of two analytes is applied, assuming no interaction between the analytes (for small analyte concentrations), a dual exponential response curve is obtained as
■
RESULTS AND DISCUSSION The main focus of the present work is on analyzing binary analyte mixtures. However, this analysis requires first investigating responses to single analytes. Previous experiments with single analytes in the concentration range of 0−10 ppm demonstrated that transient responses can be modeled using a single exponential fit, in agreement with eq 7. A characteristic response time constant was found for each coating/analyte combination which, within experimental error, was independent of analyte concentration in the range of interest. As examples, Figure 1 shows (a) the response of a sensor coated with 1.0 μm PEA to various concentrations of benzene and (b) the response of a sensor coated with 0.8 μm PIB to various concentrations of ethylbenzene. Low analyte concentrations were tested in order to experimentally approach the detection limits, LD. LD values, defined as the analyte concentrations resulting in a frequency shift of three times the RMS noise level, were also calculated from the experimental data. For these calculations, the RMS noise levels were measured using data averaging, as discussed in the Experimental Section. For a 1.0 μm thick PEA coating, an RMS noise level of 26 Hz was measured. On the basis of this value, the detection limits listed in Table 2 were calculated. In the above measurements, no significant deviation from Henry’s law was observed. Therefore, response curves to binary mixtures were modeled according to eq 8. Response times, τ1 and τ2, were obtained from single-analyte measurements. The
Δf (t ) = Δf1 [1 − exp( −t /τ1)] + Δf2 [1 − exp( −t /τ2)] (8)
where the subscripts (1, 2) refer to the two analytes. Equation 8 contains four parameters (Δf1, Δf 2, τ1, and τ2) that can be extracted from experimental data, and which can all convey useful information for pattern recognition for the purpose of analyte identification and quantification. It is noted that at steady-state, from eq 8, the total sensor response to a binary analyte mixture is Δf = Δf1 + Δf 2.
■
EXPERIMENTAL SECTION The SH-SAW sensor platform selected for this work has been previously described.8,9 It uses a delay line design where each delay line consists of two identical interdigital transducers (IDTs) on a 36° YX-LiTaO3 substrate, with a metallized surface region between the IDTs to eliminate acoustoelectric interactions with the load (i.e., the contacting aqueous phase and its contents). The IDTs have a periodicity of 40 μm, resulting in a frequency of operation of 103 MHz for the polymer-coated device. The sorbent polymers selected for this study are poly(ethyl acrylate) (PEA), poly(epichlorohydrin) (PECH), and poly(isobutylene) (PIB) (supplier: SigmaAldrich). Sorbent polymers were deposited from solution by spin coating9 and baked for 15 min at 60 °C, resulting in thicknesses ranging from 0.6 to 1.0 μm, as indicated below. 1796
dx.doi.org/10.1021/ac403724f | Anal. Chem. 2014, 86, 1794−1799
Analytical Chemistry
Article
Table 4. Measured Response Times, τ (in s), for Three Different Polymer Coatings to Various BTEX Analytes, Together with Their Standard Errors polymer
τbenzene
τtoluene
τethylbenzene
1.0 μm PEA 0.6 μm PECH 0.8 μm PIB
36.1 (±10.0) 26.5 (±8.4) 29.3 (±7.8)
76.7 (±6.0) 77.6 (±2.8) 84.2 (±6.5)
204 (±4.5) 175 (±13) 245 (±14)
indicates a scatter of about 13%. Other contributing factors, such as variations in coating thickness among devices (which was measured using profilometry and ellipsometry), are assumed to be smaller. The relative error on the response times (Table 4) is always below 8% for toluene and ethylbenzene but significantly larger for benzene (27−32%). Since benzene shows by far the fastest response, the measured response times strongly depend on the first one or two data points of the response curve. It is assumed that the error margins for benzene can be improved by increasing the data sampling rate and/or by minimizing the mechanical disturbance of the system when switching from water (baseline) to the aqueous analyte sample. Changes in the system design to this end are currently under way. In order to assess the ability of the sensor system to distinguish between different BTEX analytes, response patterns were created using six input parameters, specifically, the response times and the ratios between the steady-state frequency shifts, measured for three different coatings each. Results are shown in the form of a radial plot in Figure 2.
Figure 1. (a) Response of a SH-SAW sensor coated with 1.0 μm PEA, successively exposed to various samples of benzene in water (concentrations are indicated in the graph in parts per billion). (b) Response of a SH-SAW sensor coated with 0.8 μm PIB to various samples of ethylbenzene in water (concentrations in parts per billion).
Table 2. Calculated Detection Limits, LD, Literature Values for the Aqueous Solubilities, CS,14 and Molar Masses, MW,14 and Normalized Detection Limits, LD × MW/CS, of Three Different BTEX Analytes for a SH-SAW Device Coated with 1.0 μm PEA analyte benzene toluene ethylbenzene
LD (ppb) CS (ppm) 320 110 35
1780 531 161
MW (g/mol)
LD × MW/CS (g/mol)
78.1 92.1 106.2
0.014 0.019 0.023
steady-state frequency shifts, Δf1 and Δf 2, are proportional to the concentrations of the two analytes in the binary mixture. If the sensitivity to each analyte is known from single-analyte measurements, the analyte concentrations can be readily extracted. Both the values for the measured frequency sensitivity, S = Δf 0/Camb, and τ are obtained from multiple measurements and are listed as average values for various coating/analyte combinations in Tables 3 and 4, respectively. It is noted that the accuracies of the sensitivities (Table 3) are adversely affected by the inaccuracy in the manual sample mixing procedure. Verification of the actual sample concentrations by GC/MS (gas chromatography/mass spectroscopy)
Figure 2. Radial plot showing the response times, τ (in units of 100 s), and ratios of the steady-state frequency shifts, Δf, for various BTEX analytes detected using three different sensor coatings.
For the response patterns shown in Figure 2, the response times of the different BTEX compounds appear better separated than the ratios of frequency shifts. Indeed, no overlap occurred between the response times of different analytes for all the concentrations and polymer coatings investigated. Therefore, the response time measured using a single sensor device would be sufficient to reliably identify a given single analyte. However, the use of multiple input parameters will still be important in the data analysis of analyte mixtures. In the ratios of the steady-state frequency shifts, some overlap between different analytes is observed; see, for example, the Δf(PEA)/Δf(PECH) axis in Figure 2. Therefore, the frequency shifts are less useful than the response times for the purpose of analyte identification. The observed overlaps are in part caused by the error in analyte concentration due to the
Table 3. Measured Sensitivities, S (in Hz/ppm), for Three Different Polymer Coatings to Various BTEX Analytesa polymer
Sbenzene
Stoluene
Sethylbenzene
1.0 μm PEA 0.6 μm PECH 0.8 μm PIB
244 (±27) 109 (±9) 63 (±5)
690 (±160) 435 (±25) 344 (±43)
2240 (±460) 1450 (±240) 1670 (±110)
a
Standard errors (68% confidence interval), obtained from multiple measurements, are given in parentheses. 1797
dx.doi.org/10.1021/ac403724f | Anal. Chem. 2014, 86, 1794−1799
Analytical Chemistry
Article
manual sample mixing procedure, which will affect the steadystate frequency shifts but not the response times. However, note that the frequency shifts are still needed for analyte quantification since the response times are independent of concentration. In Figure 2, only two parameters (Δf and τ) were used for each sensor coating because the sensor response was modeled using a single-exponential fit. However, the use of a dualexponential fit of the form given in eq 8 offers the potential to extract more information from the sensor response to binary mixtures because deviations from a single-exponential response curve can be used to quantify the two analytes in the mixture. Figure 3 shows the frequency responses, Δf(t), of a sensor device coated with 1.0 μm PEA to two BTEX analytes and their
After normalization, it is seen that PEA shows a higher affinity, and thus, a lower detection limit, for benzene than for the other BTEX compounds (see Table 2, last column). Figure 4 also illustrates the response times for the analytes, the value for benzene being the shortest. For binary mixtures, the transient response is the sum of a faster response (to benzene) and a slower response (second component). In good approximation (for the low concentrations), the steady-state frequency shifts for single analytes add to give the steady-state frequency shift observed for their binary mixture. In order to extract the concentrations of the individual analytes in the mixture, response curves to binary mixtures were fit using eq 8 and the single-analyte parameters listed in Tables 3 and 4. Results are presented in Table 5. Table 5. Actual Ambient Concentrations, Camb, and Concentrations Extracted from Dual-Exponential Fits, Cfit (in ppm), for Three Different BTEX Analytes and Three Polymer Coatingsa benzene polymer 1.0 μm PEA
0.6 μm PECH
Figure 3. Response of a SH-SAW sensor coated with 1.0 μm PEA to toluene, ethylbenzene, and their binary mixture. All concentrations are 10 ppm (binary mixture: 10 ppm toluene + 10 ppm ethylbenzene). Experimental data are modeled with single- and dual-exponential fits for single- and two-component analytes, respectively.
0.8 μm PIB
ethylbenzene
Cfit
Camb
Cfit
Camb
Cfit
2 3 5 3 5 4 5
1.4 3.5 3.4 1.5 5.4 1.9 5.4
2 5
1.9 5.2
5
4.5
5
2.8
5
4.0
2
1.3
2
1.9
a
Values for Cfit are averages of multiple experiments (n = 5) on binary mixtures of benzene and toluene, and binary mixtures of benzene and ethylbenzene, respectively.
binary mixture. The figure demonstrates that the response to the binary mixture can be modeled using a dual-exponential fit with response times, τ1 and τ2, taken from the single analyte responses, as indicated by eq 8. Figure 4 shows the frequency response, Δf(t), of a sensor device coated with 0.6 μm PECH to several aromatic analytes
The errors in Table 5 are again related to the same sources as discussed for Tables 3 and 4. It is noted that analyte concentrations in binary mixtures are more often underestimated than overestimated, possibly indicating that a small amount of the first analyte in a binary mixture was lost when the second analyte was added, though care was taken to minimize such losses. The fact that responses to multiple analytes are additive was independently verified by a different set of experiments. For this purpose, a sensor device was first exposed to 10 ppm of a single analyte. As soon as equilibrium was attained, the device was immediately (i.e., without intermediate flushing with water) exposed to a binary mixture, again containing 10 ppm of the above analyte together with a second analyte. The resulting response to the binary mixture is expected then to show a transient that reflects only the addition of the second analyte, since the concentration of the first analyte remains constant. As indicated by the results presented in Figure 5, this is indeed the case within experimental error.
Figure 4. Response of a SH-SAW sensor coated with 0.6 μm PECH to various single analytes and binary mixtures. Concentrations used are 3 ppm (benzene) and 5 ppm (toluene and ethylbenzene), respectively.
■
CONCLUSION It was demonstrated that a single BTEX compound can readily be identified in water samples using response patterns based on both the steady-state signals and response times of an array of polymer-coated SH-SAW sensor devices. The concentration of the analyte can be estimated based on the steady-state frequency shifts, with detection limits as listed in Table 2. Data for the sensitivity and response time are listed in Tables 3 and 4, respectively. Within the concentration range relevant for
and binary mixtures. It can be seen that the sensitivities to the analytes fall in the order Sethylbenzene. > Stoluene > Sbenzene, in agreement with Table 3. Note that in order to determine the true affinity of a polymer for an analyte, sensitivity has to be normalized to the solubility in the aqueous phase, CS, and the molar mass, MW, of the analyte: Snormalized = (Δf /MW )/(ΔCamb/CS)
toluene
Camb
(9) 1798
dx.doi.org/10.1021/ac403724f | Anal. Chem. 2014, 86, 1794−1799
Analytical Chemistry
Article
Washington, D.C., 2012 (http://www.epa.gov/oust/pubs/fy11_ annual_ust_report_3-12.pdf). (2) Wang, Z.; Stout, S. A.; Fingas, M. Environ. Forensics 2006, 7, 105−146. (3) Liu, L.; Hao, R. X.; Cheng, S. Y. Journal of Environmental Informatics 2003, 2, 31−37. (4) Einarson, M. D.; Mackay, D. M. Environ. Sci. Technol. 2001, 35, 66A−73A. (5) Ho, C. K.; Robinson, A.; Miller, D. R.; Davis, M. J. Sensors 2005, 5, 4−37. (6) Länge, K.; Rapp, B. E.; Rapp, M. Anal. Bioanal. Chem. 2008, 391, 1509−1519. (7) Gizeli, E. Smart Mater. Struct. 1997, 6, 700−706. (8) Josse, F.; Bender, F.; Cernosek, R. W. Anal. Chem. 2001, 73, 5937−5944. (9) Li, Z.; Jones, Y.; Hossenlopp, J.; Cernosek, R.; Josse, F. Anal. Chem. 2005, 77, 4595−4603. (10) Pejcic, B.; Eadington, P.; Ross, A. Environ. Sci. Technol. 2007, 41, 6333−6342. (11) Bender, F.; Josse, F.; Ricco, A. J. Joint Conference of the IEEE IFCS and the EFTF 2011, 422−427. (12) Arey, J. S.; Gschwend, P. M. Environ. Sci. Technol. 2005, 39, 2702−2710. (13) U.S. Environmental Protection Agency. National Primary Drinking Water Regulations. EPA 816-F-09−004, 2009. http://www. epa.gov/safewater/consumer/pdf/mcl.pdf. (14) Lide, D. L. Handbook of Chemistry and Physics, 82nd ed.; CRC Press: Boca Raton, FL, 2001−2002; p 8−95. (15) Jones, Y. K.; Li, Z.; Johnson, M. M.; Josse, F.; Hossenlopp, J. M. IEEE Sens. J. 2005, 5, 1175−1184. (16) Zellers, E. T.; Batterman, S. A.; Han, M.; Patrash, S. J. Anal. Chem. 1995, 67, 1092−1106. (17) Ricco, A. J.; Crooks, R. M.; Osbourn, G. C. Acc. Chem. Res. 1998, 31, 289−296. (18) Hierlemann, A.; Ricco, A. J.; Bodenhöfer, K.; Göpel, W. Anal. Chem. 1999, 71, 3022−3035. (19) Martin, S. J.; Frye, G. C.; Senturia, S. D. Anal. Chem. 1994, 66, 2201−2219. (20) Grate, J. W.; Zellers, E. T. Anal. Chem. 2000, 72, 2861−2868. (21) Jones, Y. K. Spectroscopic and Computational Investigation of Polymer Coatings and Analyte Systems for use with Guided Shear Horizontal Surface Acoustic Wave (SH-SAW) Sensors for Liquid Phase Detection. Ph.D. Dissertation; Marquette University, 2005. (22) Hierlemann, A.; Zellers, E. T.; Ricco, A. J. Anal. Chem. 2001, 73, 3458−3466. (23) Mensah-Brown, A. K.; Wenzel, M. J.; Josse, F. J.; Yaz, E. E. IEEE Sens. J. 2009, 9, 1817−1824. (24) Wenzel, M. J.; Mensah-Brown, A.; Josse, F.; Yaz, E. E. IEEE Sens. J. 2011, 11, 225−232.
Figure 5. Response of a SH-SAW sensor coated with 1.0 μm PEA to various single analytes and binary mixtures. All concentrations used are 10 ppm (binary mixtures: 10 ppm benzene + 10 ppm ethylbenzene).
BTEX detection in groundwater, the responses of the polymercoated SH-SAW sensors to multiple analytes in a mixture are additive, and the response time is specific to each coating/ analyte pair. On the basis of these findings, it has been demonstrated that evaluation of transient responses can help identify and quantify benzene in a mixture. For example, the response to ethylbenzene in Figure 4 shows about the same frequency shift as the response to the binary mixture of benzene and ethylbenzene, but benzene concentration was still successfully extracted from the different transient response characteristics. Within experimental error, concentrations for ethylbenzene and toluene can usually be predicted with high accuracy; Table 5 shows average errors in Cfit of 0.8 and 0.5 ppm for toluene and ethylbenzene, respectively. For benzene, a slightly larger error of 1.0 ppm was observed. This is due to its higher solubility in water and lower partition coefficients for the coatings; as discussed above, the limited data-sampling rate and a slight mechanical disturbance of the system when switching between sample vials also affect quantification of benzene more than the other BTEX compounds. With the present system, benzene is measurable at concentrations down to 320 ppb in water. In order to improve this detection limit, the eventual approach will involve the use of a sensor array utilizing both the transient and the steady-state responses. Other improvements currently underway include a new sensor design using a higher frequency of operation, a higher data sampling rate, and improved data processing using Extended Kalman Filter (EKF) techniques.23,24 The latter will also help to extract information on analyte concentrations from the sensor responses before steady-state is reached, thus permitting the reduction of the time required for sample analysis below the approximately 15 min observed in this work.
■
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Fax: +1 414 288 3951. Tel: +1 414 288 6789. Notes
The authors declare no competing financial interest.
■ ■
ACKNOWLEDGMENTS The authors thank Karen A. Synowiec and Urmas Kelmser from Chevron Energy Technology Co. for helpful discussions. REFERENCES
(1) FY 2011 Annual Report On The Underground Storage Tank Program. EPA 510-R-12−001, U.S. Environmental Protection Agency: 1799
dx.doi.org/10.1021/ac403724f | Anal. Chem. 2014, 86, 1794−1799
