Design of SH-Surface Acoustic Wave Sensors for

Detection of ppb Concentrations of BTEX in Water

Florian Bender and Fabien Josse Rachel E. Mohler Antonio J. Ricco

Department of EECE Chevron Energy Technology Co. Department of Electrical
Marquette University 100 Chevron Way Engineering
Milwaukee, WI 53233 Richmond, CA 94801 Center for Integrated Systems
fabien.josse@mu.edu Stanford University
Stanford, CA 94305

Abstract—To address the need to protect public health from approach for this application. SH-SAWs combine the
contamination of drinking water and water for recreational use, advantages of high surface confinement of the acoustic energy
a compact sensor system for in-situ detection of fuel and oil (which facilitates high sensitivity to the analyte) and the ability
components in water is being investigated. The system makes use to propagate along a solid-liquid interface without suffering
of shear-horizontal surface acoustic wave (SH-SAW) sensors excessive acoustic wave attenuation [6]. In our previous
coated with thin chemically sensitive polymer films. For this studies, the performance of the sensors and various coating
work, the BTEX compounds (benzene, toluene, ethylbenzene, materials was investigated under varying environmental
and xylenes) were selected as target analytes because they are conditions, including changes in temperature, acidity, and
good indicators of fuel and oil releases, but also because they
salinity [7]. It was demonstrated that by evaluating both
include known carcinogens (benzene, ethylbenzene). Chemical
selectivity is achieved by combining sensors with various polymer
steady-state and transient response information, this type of
coatings into a sensor array, and by evaluating both steady-state sensor can be used to quantify the concentration of benzene in
and transient response information. This work focuses on the binary analyte mixtures in the low ppm concentration range
influence of interdigital transducer (IDT) design on signal [8].
distortion and rms (root-mean-square) noise level. It is The BTEX compounds (benzene, toluene, ethylbenzene,
demonstrated that for suitable IDT design and experimental and xylenes) were selected as target analytes for this study
approach, it is possible to detect all BTEX compounds at
because they are good indicators of fuel and oil releases [3],
concentration levels of 100 ppb or below, and to quantify benzene
[4], [9], but also because they include known carcinogens
concentration in binary analyte mixtures at concentrations well
below 1 ppm.
(benzene, ethylbenzene) [10]. Due to their hazard potential for
public health, legal concentration limits for the BTEX
Keywords—groundwater, surface acoustic wave sensors, liquid- compounds in drinking water are low, with MCL (maximum
phase sensing, benzene, rms noise level, multi-electrode interdigital contaminant level) values in the US ranging from 5 ppb (μg/L)
transducer design to 10 ppm (mg/L) [5], [11]. Detection of these contaminants at
ppb levels is a challenging task, especially if the concentration
of one BTEX compound has to be quantified in the presence of
I. INTRODUCTION
similar aromatic compounds.
The processing, storage and transport of fuel and oil are
often associated with potential environmental hazards. In The BTEX detection challenge is addressed in our work by
particular, places where large amounts of fuel or oil are stored, evaluating both the steady-state frequency shift and the time-
such as gas stations, present significant potential threats to dependent behavior of the sensor response. We investigated
public health through unintentional releases into groundwater various polymer materials (≤ 1 μm thick) as sensor coatings
and surface water [1]–[4]. Timely detection of such releases based on their partial selectivities for the analytes of interest
can minimize the impact on public health and reduce clean-up [8]. In order to achieve detection limits for BTEX compounds
costs. However, current monitoring practice is largely based on in the ppb range, every aspect of the sensor design has to be
collection of samples in the field, transport and subsequent evaluated for its potential to minimize rms (root-mean-square)
analysis in the laboratory [3], procedures which can become noise and signal distortion. For this study, 36°-rotated YX-
very time-consuming for high monitoring frequencies. To LiTaO3 was selected as piezoelectric substrate because of its
enable frequent or continuous, in-situ monitoring of high dielectric constants [12], permitting immersion of the
groundwater and surface water at critical sites, a number of entire device surface—including the polymer-coated
suitable sensor systems are under development [5]. interdigital transducers (IDTs)—in water, thus obviating the
need for a gasket on the acoustic delay line, which could lead
Sensor systems based on SH-SAW (shear-horizontal to signal distortion. LiTaO3 also has high piezoelectric
surface acoustic wave) sensor platforms represent a promising constants [12], restricting the number of electrode fingers per

978-1-4799-0342-9/13/$31.00 ©2013 IEEE 628 2013 Joint UFFC, EFTF and PFM Symposium
IDT if overcoupling (and consequent signal distortion) is to be
avoided. One solution would be to use a wideband IDT, but
this could lead to overlap between the SH-SAW and
neighboring bulk modes with further adverse consequences for
signal distortion. Therefore, another approach is investigated
here, based on the use of multi-electrode IDTs [13]. In this
way, overcoupling can be avoided by placing many electrode
fingers in antiphase with the acoustic wave. It is demonstrated
that a suitable multi-electrode IDT design leads not only to
better bulk mode suppression and lower phase distortion, but
also has a beneficial effect on rms noise, resulting in lower
detection limits and higher accuracy in analyte quantification
in binary mixtures. In addition, multi-electrode IDTs can be
designed to generate multiple SH-SAWs at different
frequencies in a single delay line, permitting investigation of
frequency-dependent contributions in sensor responses.

II. EXPERIMENTAL
The first of the SH-SAW sensor platforms used in this Fig. 1: Mode spectra (in air) for SH-SAW sensors with Se = 4 (a) and
Se = 12 (b). Sensors were coated with 0.5 μm PMMA. Insets are
work has been described previously [14], [15]. The sorbent schematics of the IDT electrode patterns (only part of the IDT is shown).
polymers selected as sensor coatings are poly(ethyl acrylate)
(PEA), poly(epichlorohydrin) (PECH), and poly(isobutylene) passband. The latter design has a lower bandwidth, but the
(PIB) (all from Sigma-Aldrich). Sorbent polymers were larger number of electrode fingers leads to stronger acoustic
deposited from solution by spin coating and baked for 15 min wave reflections and phase distortions, and the coupling of
at 55 °C, resulting in thicknesses ranging from 0.55 to 1.0 μm unwanted bulk modes is also enhanced. The mode spectrum of
as indicated below. All sensor platforms utilize a dual-delay- the LIDT = 44P design is shown in Fig. 1a for a polymer-coated
line design, with the second line coated with poly(methyl device. The highest peak (at 103 MHz) is the SH-SAW; it is
methacrylate) (Scientific Polymer Products) and baked for 120 followed by a series of bulk modes.
min at 180 °C, resulting in a glassy, non-sorbent coating. The
purpose of this reference line is to compensate for the influence A solution of the above dilemma is offered by multi-
of temperature drifts and other secondary effects. All BTEX electrode IDT designs [13] where a number of the electrode
analytes had purities of 98.5% or higher. fingers are deliberately placed in anti-phase with the acoustic
wave in order to reduce overall mode coupling. This allows the
The experimental set-up used a network analyzer (Agilent combination of a narrow IDT bandwidth with low bulk mode
E5061B; the 16-fold data averaging function was activated) coupling and modest acoustic reflections. As an example, Fig.
and a switch/control system (Agilent 34980A) to switch 1b shows the mode spectrum and IDT pattern for a SH-SAW
between the two SH-SAW delay lines. The sensor was placed device with Se = 12. In this case, a series of SH-SAW modes is
inside a flow cell made in-house, and the peristaltic pump obtained with frequencies which are multiples of 34 MHz. This
(Ismatec Reglo Digital MS-CA4/12-100) was set to a sample IDT design also leads to a lower wave reflection and better
flow rate of 7 μL/s. Switching between different samples was triple-transit echo suppression, a more symmetrical passband,
accomplished without interruption of sample flow in order to and lower phase distortion. The latter is illustrated in Fig. 2.
avoid any resulting signal distortions. The flow cell and
samples were set up inside a box, simulating an environment While the relationship between phase linearity and rms
similar to a groundwater monitoring well, and measurements noise level is not a straightforward one, it was experimentally
were conducted at room temperature (22.0 ± 0.1 °C). All data observed that the IDT design giving the lowest phase distortion
shown are corrected for linear baseline drift. also gave the lowest rms noise level. This is shown in Fig. 3.

III. RESULTS AND DISCUSSION

In previous studies, a 103-MHz SH-SAW sensor device has
successfully been used to detect BTEX compounds in the ppm
concentration range [7], [8], [14], [15]. This device used an
IDT pattern with four electrode fingers per electrical period (Se
= 4), schematically illustrated in the inset of Fig. 1a. Two
versions of this device were fabricated which differed in their
bandwidths; one used an IDT length of 29 electrical periods
(LIDT = 29P), the other design used LIDT = 44P. While the
former shows relatively low phase distortion and good bulk Fig. 2: Phase deviation (from linear phase) for SH-SAW sensors with
mode and sidelobe rejection, the large bandwidth leads to Se = 4 (—) and Se = 12 (- -). Sensors were coated with 0.5 μm PMMA and
immersed in water.
larger overlap between the SH-SAW and adjacent bulk modes
and results in a significant asymmetry of the SH-SAW

629 2013 Joint UFFC, EFTF and PFM Symposium

 TABLE I. EXPERIMENTALLY DETERMINED DETECTION LIMITS (3 × RMS
NOISE LEVEL / SENSITIVITY) FOR SH-SAW SENSORS WITH SE = 12 FOR
SIX COATING/ANALYTE COMBINATIONS.

Solubility Detection Limit (ppb)

Analyte
(ppm) [16] 1.0 μm PEA 0.6 μm PECH
Benzene 1780 100 100

Toluene 531 35 33

Ethylbenzene 161 11 9

are additive in the concentration ranges investigated [8] and

that the transient response to a step change in analyte
concentration is approximately exponential with time, the
response to a binary mixture can be modeled using a dual-
exponential fit, whereby the response times and sensitivities for
the two analytes in the mixture can be taken from single-
analyte measurements. This allows an estimation of the
Fig. 3: Baseline noise levels for the 103-MHz SH-SAW mode for concentrations of two analytes in a mixture using a single
devices with Se = 4 (a) and Se = 12 (b). Sensors were coated with 1.0 μm
PEA and immersed in water (flow rate: 7 μL/s). sensor device simply by evaluating the transient response
information. As an example, for the responses to binary
For the Se = 4 IDT design, a rms noise level of 25.9 Hz was mixtures in Fig. 4, analyte concentrations estimated using dual-
calculated for a 1.0 μm PEA coating and the same flow rate exponential fits of the measured data are presented in Table II
used in BTEX detection. (Note that such a relatively thick and compared to nominal (actual) concentrations. Note that the
coating of a rubbery polymer will slightly increase the rms manual sample mixing procedure will introduce slight error in
noise level but also will enhance the quantity of BTEX sorbed). analyte concentrations. Nevertheless, there is a good agreement
For the Se = 12 IDT design, under the same conditions an rms between nominal and estimated concentration for benzene.
noise level of only 10.7 Hz was found. For a 0.6 μm PECH Even better agreement is observed for toluene (because of the
coating and Se = 12 IDT design, the rms noise level was further higher sensitivity for this analyte).
reduced to 7.1 Hz (data not shown). The above results illustrate the detection limit in single-
In order to evaluate the performance of the improved SH- analyte detection and the accuracy in binary mixture analysis
SAW sensor devices in BTEX detection, measurements were that can be achieved using the multi-electrode sensor platform
conducted on detection of single analytes and binary analyte designs. Another advantage of multi-electrode IDT designs is
mixtures in the concentration range of 100 ppb to 1 ppm. the presence of multiple SH-SAW passbands [13], which can
Sample data are shown in Fig. 4; it clearly demonstrates that be used for analysis of frequency-dependent effects in a single
benzene can be detected at a concentration as low as 200 ppb. measurement with a single device. As an example, Fig. 5
From these measurements, the detection limits listed in Table I shows the response of a SH-SAW device with Se = 20 to 5 ppm
have been calculated using the relation: detection limit = 3 × benzene. The basic SH-SAW frequency for this device is 41
rms noise / sensitivity. Note that xylene and ethylbenzene are MHz; the responses of the 2nd to 4th SH-SAW harmonics to the
chemical isomers and thus will give nearly identical same sample are shown at 82, 123, and 164 MHz.
sensitivities; therefore, only ethylbenzene is listed. As To facilitate a more detailed analysis, the exact steady-state
expected, the detection limits approximately follow the trend in frequency shifts for this measurement are listed in Table III.
analyte solubilities. In all cases, analytes can be detected at less The dependence of the observed frequency shifts on mode
than 10-4 of their saturation concentration in water. frequency contains a contribution from Sauerbrey-type mass
Fig. 4 also shows several responses to binary mixtures. loading [17]. Since all SH-SAW modes are propagating along
Assuming the sensor responses to multiple analytes in a sample the same delay line and will be exposed to the same amount of
mass loading, this contribution is expected to give a Δf ∝ f 2
dependence, where f is the device operating frequency. But the
last column of Table III reveals significant additional
contributions for the higher harmonics. This can be explained
TABLE II. NOMINAL ANALYTE CONCENTRATIONS FOR THE FIRST
THREE SENSOR RESPONSES (TO BINARY MIXTURES) IN FIG. 4, AND
CONCENTRATIONS ESTIMATED USING DUAL-EXPONENTIAL FITS.

Benzene Conc. (ppb) Toluene Conc. (ppb)

Sample
nominal fit nominal fit
#1 1000 1200 500 690
Fig. 4: Response of a SH-SAW device (103-MHz mode) with Se = 4 #2 500 810 1000 1110
and a 0.6 μm PECH coating to various benzene samples and binary
mixtures. Analyte concentrations are given in the graph. #3 1000 1280 1000 990

630 2013 Joint UFFC, EFTF and PFM Symposium

 ACKNOWLEDGMENT
The authors would like to thank Karen A. Synowiec and
Urmas Kelmser for helpful discussion.

REFERENCES
[1] U.S. Environmental Protection Agency, FY 2011 Annual Report On The
Underground Storage Tank Program (2012); EPA-510-R-12-001.
[2] U.S. Environmental Protection Agency, Underground Storage Tank
Flood Guide (2010); EPA-510-R-10-002.
[3] Z. Wang, S.A. Stout, and M. Fingas, “Forensic Fingerprinting of
Biomarkers for Oil Spill Characterization and Source Identification”,
Environ. Forensics, vol. 7, pp. 105–146, 2006.
Fig. 5: Response of a SH-SAW device with Se = 20 to 5 ppm benzene,
shown for three different SH-SAW modes with frequencies indicated in [4] R. Johnson, J. Pankow, D. Bender, C. Price, and J. Zogorski, “MTBE—
the graph. The device was coated with 0.55 μm PEA. To What Extent Will Past Releases Contaminate Community Water
Supply Wells?” Environ. Sci. Technol., vol. 34, pp. 210A–217A, 2000.
by the higher ratio of waveguide thickness to wavelength for [5] C.K. Ho, A. Robinson, D.R. Miller, and M.J. Davis, “Overview of
the higher modes, resulting in more efficient waveguiding and Sensors and Needs for Environmental Monitoring”, Sensors, vol. 5, pp.
a corresponding increase in mass sensitivity for SH-SAW 4–37, 2005.
devices [15], [18]. Another contribution is from viscoelastic [6] E. Gizeli, “Design considerations for the acoustic waveguide biosensor”,
effects, which would also be expected to result in increased Smart Mater. Struct., vol. 6, pp. 700–706, 1997.
frequency shifts for higher ratios of waveguide thickness to [7] F. Bender, F. Josse, and A.J. Ricco, “Influence of Ambient Parameters
on the Response of Polymer-Coated SH-Surface Acoustic Wave Sensors
wavelength [7], [14], [19]. to Aromatic Analytes in Liquid-Phase Detection”, 2011 Joint
Conference of the IEEE IFCS and the EFTF, pp. 422–427, 2011.
IV. CONCLUSIONS [8] F. Bender, R. Mohler, A.J. Ricco, and F. Josse, “Quantification of
Benzene in Groundwater Using SH-Surface Acoustic Wave Sensors”,
In chemical sensor applications, for a high-coupling IMCS 2012 Proc., pp. 473–476, 2012.
piezoelectric material like LiTaO3, use of the standard double- [9] J.S. Arey and P.M. Gschwend, “Estimating Partition Coefficients for
finger (Se = 4) IDT leads to a dilemma: use of a wideband IDT Fuel-Water Systems: Developing Linear Solvation Energy Relationships
will result in significant overlap of adjacent mode passbands; Using Linear Solvent Strength Theory To Handle Mixtures”, Environ.
use of a narrowband IDT will lead to overcoupling and Sci. Technol., vol. 39, pp. 2702–2710, 2005.
increased acoustic wave reflections. Thus, both alternatives [10] www.sigmaaldrich.com: material safety data sheets for benzene and
will lead to significant signal distortions. It was demonstrated ethylbenzene, accessed 7/9/2013.
that the use of multi-electrode IDTs solves this issue, and [11] U.S. Environmental Protection Agency, National Primary Drinking
narrowband SH-SAW modes can be obtained with high phase Water Regulations (2009); EPA-816-F-09-004.
linearity, low acoustic wave reflection, and good sidelobe and [12] G. Kovacs, M. Anhorn, H.E. Engan, G. Visintini, and C.C.W. Ruppel,
“Improved Material Constants for LiNbO3 and LiTaO3” Proc. 1990
bulk mode suppression. IEEE Ultrason. Symp., pp. 435–438, 1990.
It was observed that properly designed multi-electrode [13] H. Engan, “Surface Acoustic Wave Multielectrode Transducers”, IEEE
IDTs also result in significantly lower rms noise levels, with Trans. Son. Ultrason., vol. 22, pp. 395–401, 1975.
beneficial effects for BTEX detection in water including lower [14] Z. Li, Y. Jones, J. Hossenlopp, R. Cernosek, and F. Josse, “Analysis of
Liquid-Phase Chemical Detection Using Guided Shear Horizontal-
detection limits and higher accuracy in binary mixture analysis, Surface Acoustic Wave Sensors”, Anal. Chem., vol. 77, pp. 4595–4603,
as well as improved reproducibility. In addition, multi- 2005.
electrode IDTs produce spectra of SH-SAW harmonics which [15] F. Josse, F. Bender, and R.W. Cernosek, “Guided Shear Horizontal
permit investigation of frequency-dependent contributions in Surface Acoustic Wave Sensors for Chemical and Biochemical
sensor responses using a single device in a single Detection in Liquids”, Anal. Chem., vol. 73, pp. 5937–5944, 2001.
measurement. This option can also be useful in a sensor array [16] D.L. Lide, Handbook of Chemistry and Physics, 82nd ed., CRC Press:
since it permits collection of information from a larger number Boca Raton, FL, 2001–2002, pp. 8–95.
of input parameters using a compact sensor platform. [17] G. Sauerbrey, “Verwendung von Schwingquarzen zur Wägung dünner
Schichten und zur Mikrowägung” (in German), Z. Phys., vol. 155, pp.
206–222, 1959.
TABLE III. STEADY-STATE FREQUENCY SHIFTS, -Δf, FOR THE SENSOR [18] E. Gizeli, N.J. Goddard, C.R. Lowe, and A.C. Stevenson, “A Love plate
RESPONSES SHOWN IN FIG. 5. A FREQUENCY DEPENDENCE STRONGER biosensor utilising a polymer layer”, Sens. Actuators, B, vol. 6, pp. 131–
2
THAN Δf ∝ f IS OBSERVED. 137, 1992.
[19] S.J. Martin, G.C. Frye, and S.D. Senturia, “Dynamics and Response of
Mode -Δf (Hz) (f/f2)2 Δf/Δf2 Polymer-Coated Surface Acoustic Wave Devices: Effect of Viscoelastic
f2 = 82 MHz 280 1 1 Properties and Film Resonance”, Anal. Chem., vol. 66, pp. 2201–2219,
1994.
f3 = 123 MHz 930 2.25 3.32

f4 = 164 MHz 2600 4 9.29

631 2013 Joint UFFC, EFTF and PFM Symposium