Anal. Chem. 2003, 75, 4551-4557
Frequency Encoding of Resonant Mass Sensors for Chemical Vapor Detection Shenheng Guan*
Palo Alto Sensor Technology Innovation, 879 Newell Place, Palo Alto, California 94303
Chemical vapors can be detected by a resonant mass sensor array with selective absorption coatings implementing a frequency encoding method. The sensor array consists of sensor elements with different frequencies for their identifications in the frequency response obtained with a pulse Fourier transform detection scheme. Zeroloading resonance frequencies are chosen so that frequency shift due to absorption is bounded within a predefined region so that there is no overlap of peaks and all peaks can be assigned to the correct elements at any operation conditions. Mechanical oscillations of all or selected numbers of the sensor elements are excited by application of an excitation signal. Free oscillation decay signals from all or selectively excited sensor elements are detected and digitized. The free oscillation decay signal is subjected to a spectral analysis routine converting into a frequency spectrum, in which frequency shifts due to absorption of chemical vapors can be obtained. The implementation of the frequency encoding method with pulse Fourier transform detection to resonant mass sensors allows simultaneous multisensor detection, fast data acquisition speed, high signal-to-noise ratio by coaddition of raw data, flexible excitation, reduced complexity of electronic hardware, application of advanced data/spectral analysis algorithms, and realization of many other advantages by the introduction of the pulse Fourier transform method. A practical chemical vapor sensing system is demonstrated experimentally by use of nine frequency-encoded and polymer-coated sensors. Chemical vapor detection is a field of analytical chemistry undergoing two paradigm shifts. The first is a miniaturization of analytical instruments from large, laboratory or desktop instruments to portable or wearable units. The second is the proliferation of sensor-based detection systems. Those paradigm shifts are catalyzed by development of powerful and decreasing sized computers and by deeper understanding of the sensing mechanism at an increasingly interdisciplinary level. A portable chemical vapor analyzer can be based on direct miniaturization of conventional gas-phase chemical instruments, such as microscale gas chromatography and miniaturized mass spectrometers. However, the major development of chemical vapor sensors is based on detecting physical changes after a vapor is absorbed onto a selective coating on a sensory element. The physical change due to the absorption can be an increase of volume, a change in 10.1021/ac034228r CCC: $25.00 Published on Web 07/17/2003
© 2003 American Chemical Society
resistivity, an increase in mass, a change in optical properties, and so on. The field of chemical sensors has been reviewed.1 Detection based on mass measurement has many advantages over measurement of other properties. There are more complicated relationships between the amount of absorbed material and changes in other properties. Mass measurement based on acoustic resonators is very sensitive. Common acoustic sensors are thickness-shear mode (TSM), also known as quartz crystal microbalance (QCM); surface acoustic wave (SAW); and flexural plate wave (FPW) sensors. The field of acoustic sensors is the subject of extensive reviews.2 For all the above acoustic devices, a polymer film is deposited on the device surface. The deposition adds additional mass to the device and causes a decrease in resonance frequency. Absorption of organic vapor in the film causes a further frequency decrease, and this relates to the amount of vapor absorption. The relation between the concentration of the analyte in the film and that in the vapor phase at equilibrium is described by K ) Cs/Cv, where Cs is the concentration in the film, Cv is the concentration in the gas phase, and K is the partition constant dependent on the nature of the polymer and the analyte. The frequency shift due to absorbing analyte is given by3
∆fv ) ∆fsCvK/F
(1)
where ∆fv, ∆fs, Cv, K, and F are the frequency shifts caused by the mass of the vapor, the frequency shift by the polymer coating, the vapor concentration in the gas phase, the partition coefficient, and the density of the polymer, respectively. In a SAW device, only a thin layer of solid is excited, and therefore, it also exhibits high mass sensitivity. Because of operating at high frequency, absorption of an analyte causes a change of modulus in the absorbing polymer. The frequency shift due to absorption of a vapor is too complicated to be explained by the simple equation. In principle, the frequency encoding and Fourier transform detection method developed in this paper can be applied to any resonant mass sensors realized by use of an acoustic device, such as surface acoustic wave (SAW), thickness-shear mode (TSM), (1) (a) Janata, J.; Josowicz, M.; Vanysek, P.; DeVaney, D. M. Anal. Chem. 1998, 70, 179R and references therein. (b) Albert, K. J.; Lewis, N. S.; Schauer, C. L.; Sotzing, G. A.; Stitzel, S. E.; Vaid, T. P.; Walt, D. R. Chem. Rev. 2000, 100, 2595-2626. (2) Ballantine, D. S., Jr.; White, R. M.; Martin, S. J.; Ricco, A. J.; Frye, G. C.; Wohltjen, H. Acoustic wave sensors: theory, design, and physicochemical applications; Academy Press: San Diego, 1996. (3) (a) Grate, J. W.; Martin, S. J.; White, R. M. Anal. Chem. 1993, 65, 940A, 987A. (b) Grate, J. W. Chem. Rev. 2000, 100, 2627-2648.
Analytical Chemistry, Vol. 75, No. 17, September 1, 2003 4551
and flexural plate wave (FPW) devices. However, it is very convenient to demonstrate it on a quartz tuning fork-based sensor array. Quartz tuning forks, found in most electronic wristwatches nowadays, have a high Q-value and high temperature stability, and they can be efficiently machined or microfabricated.4 Quartz tuning forks have been used for sensing minute forces in the field of force microscopy5 and for characterization of materials and combinatorial libraries.6 Other work related to this investigation includes vapor sensing by use of microcantilevers7 and Fourier transform detection of conductive gas sensors.8 Frequency Encoding Scheme. Introduction of the fast Fourier transform (FFT) method has revolutionized many fields of chemical instrumentation. FTIR, FTNMR, and FTICRMS have superior performance over their frequency scanning counterparts.9,10 The advantages include superior spectral resolution, versatile operation procedures, high analysis speed, and simplification of hardware construction. Many of the advantages of the Fourier transform technique can be realized when applied to resonant mass sensors. The purpose of this investigation is to introduce a robust and inexpensive chemical vapor detection system. Piezoelectric tuning forks are selected as a mass sensor for chemical vapor detection. Because of the high q-value and high stability of acoustic mass resonators, the pulsed Fourier transform (FT) detection method allows a frequency shift be detected with high resolution and high precision. The FT methods also simplify detection electronics, making the sensor system more affordable. By directly measuring time domain data, one can apply advanced data reduction methods, such as linear prediction and maximum entropy method for spectral analysis, further increasing the accuracy of frequency determination. The major advantage of using a tuning fork resonator as the sensor element is that it is easy to implement the so-called frequency encoding method to take an advantage of simultaneous multichannel detection of the FT method to measure multisensors simultaneously. By making sensors with predefined frequencies, one can distinguish individual sensors by detecting a peak in a certain frequency range. Although it is easy to implement the present frequency encoding method on tuning fork sensors, the method can be adapted to any resonance-based sensors with high q-factors, such as cantilevers, TSM, etc. As shown in Figure 1, a sensor array is formed by connecting N sensors in parallel. Such a sensor configuration works well at low resonance frequency and for devices having low shunt capacitance, as is the case for tuning forks. In this case, a high-q or well-localized mechanical equivalent circuit dominates the electronic behavior, and little cross-talk results. For other higher (4) Staudte, J. H. Proc. 27th Annu. Symp. Freq. Control 1973, 50. (5) Karrai, K.; Grober, R. D. Appl. Phys. Lett. 1995, 66, 1842. (6) (a) Matsiev, L. F.; Bennett, J. W.; McFarland, E. W. IEEE Ultrason. Symp. 1998, 459. (b) Matsiev, L. F. IEEE Ultrason. Symp. 1999, 457. (c) Matsiev, L. F. IEEE Ultrason. Symp. 2000, 427. (7) Thundat, T.; Chen, G. Y.; Warmack, R. J.; Allison, D. P.; Wachter, E. A. Anal. Chem. 1995, 67, 519. (8) Nakata, S.; Akakabe, S.; Nakasuji, M.; Yoshikawa, K. Anal. Chem. 1996, 68, 2067. (9) Ernst, R. R.; Bodenhausen, G.; Wokaun, A. Principles of Nuclear Magnetic Resonance in One and Two Dimensions; Clarendon: Oxford, 1987. (10) Marshall, A. G.; Verdun, F. R. Fourier Transforms in Optical, NMR, and Mass Spectrometry: A User’s Handbook; Elsevier: Amsterdam; 1990.
4552 Analytical Chemistry, Vol. 75, No. 17, September 1, 2003
Figure 1. Excitation/detection configuration for frequency encoded piezoelectric resonant sensors. Vs(t) is the excitation voltage signal and V0(t) is the detection voltage signal.
Figure 2. Excitation and detection sequence of pulse Fourier transform resonant sensor system. The temporal separation of excitation and detection allows free oscillation decay to be acquired without interference of the excitation source.
frequency devices, such as SAW and TSM, matching networks are required to reduce shunt capacitance.11 The excitation source is connected to one end of the sensor array, and the other end is connected to the electrical ground through a resistor and to the acquisition preamplifier. In an ideal situation, the output impedance of the excitation amplifier should be infinitely small and the input impedance of the preamplifier should be infinitely large. Although a conventional frequency scanning method can be used to construct the spectral response of the sensor array, it is more effective to apply a pulsed Fourier transform detection method, in which all frequencies are detected simultaneously. In a pulsed Fourier transform detection process, excitation and detection are temporally separated (Figure 2), and all spectral data is obtained from a single data acquisition. Data acquisition can also be accumulated or coadded to increase signal-to-noise ratio. There are two generic applications for a frequency encoded sensor array. First, to identify unknown chemical vapors in a single medium, each of the N sensors is coated with different absorption materials, and all sensors in the array are exposed to the same medium. Devices used for this application are also known as electronic noses. Second, each of the N sensors is coated with the same absorption material. Each sensor is exposed to a different medium. Those situations are common in the field of combinatorial chemistry and high-throughput screening in drug discovery. Since the frequency encoding method described below is not limited to any one of the application areas, it can be used for either situation or any combination of the two cases. Typically, a polymer layer is coated on a sensor, and differentiation of vapor molecules comes from discrimination of the (11) Guan, S. unpublished results.
was constructed to demonstrate the capability of the frequency encoding and Fourier transform detection method.
Figure 3. One possible implementation of the frequency encoding method for resonant sensor detection. The zero loading frequencies f10, f20, ..., fN0 were chosen so that there would be no overlap at any operating conditions. A peak detected in the frequency range of [fi+10, fi0] will be assigned to sensor no. i (1 e i e N).
absorbing processes. From the frequency shift of the coated sensors from the virgin sensors, the coating thickness can be determined (see Appendix). Since frequency shift due to mass load caused by absorption of chemical vapors on polymer coatings is quite small (,1% of resonance frequency), it is possible to fabricate sensors with different resonance frequencies to utilize the frequency spectral space, as long as they do not overlap at any mass loading (operating) conditions. Figure 3 illustrates one of the possible implementations of the frequency encoding method. Sensor no. 1 has the highest zero-load resonance frequency (f10) and sensor no. N has the lowest zero-loading resonance frequency (fN0). The zero-loading resonance frequency (fi0) of sensor no. i should be chosen so that at any analysis conditions, its resonance frequency (fi1) will not overlap with its neighbor’s zero-loading resonance frequency (fi+10). If this condition holds for all sensors (fi1 > fi+10 for 1 e i
