Chemical Sensors Based on Micromachined Transducers with

Anal. Chem. 2006, 78, 5633-5638

Articles

Chemical Sensors Based on Micromachined Transducers with Integrated Piezoresistive Readout Radislav A. Potyrailo,*,† Andrew Leach,† William G. Morris,† and Sisira Kankanam Gamage‡

Materials Analysis and Chemical Sciences, General Electric Global Research Center, Niskayuna, New York 12309, and General Electric Industrial Sensing, Fremont, California 94539

We demonstrate an approach for the development of chemical sensors utilizing silicon micromachined physical transducers with integrated piezoresistive readout. Originally, these transducers were developed and optimized as sensitive accelerometers for automotive applications. However, by applying a chemically responsive layer onto the transducer, we convert these transducers into chemical sensors. These transducers are attractive for chemical sensing applications for several key reasons. First, the required sensitivity of the chemical sensor can be achieved by choosing the right spring constant of the transducer. Second, the integrated piezoresistive readout of the transducer is already optimized and is very straightforward, providing a desired reproducibility in measurements, while not requiring bulky equipment. Third, chemically responsive film deposition is simple due to the ease of access to the transducer’s surface. Fourth, such transducers are already available for another (automotive) application, making these sensors very cost-effective. The applicability of this approach is illustrated by the fabrication of highly sensitive CO2 sensors. To study hysteresis effects, we selected high CO2 concentrations (10-100% CO2) to provide the worst-case scenario for the sensor operation. These sensors demonstrate a hysteresis-free performance over the concentration range from 10 to 100% vol CO2, have detection limits of 160-370 ppm of CO2, and exhibit a relatively rapid response time, T90 ) 45 s. Importantly, we demonstrate a simple method for cancellation of vibration effects when these physical transducers, initially developed as accelerometers, are applied as chemical sensors. Existing micromachined sensors typically have long timelines from the concept through the evolution and cost reduction to commercial products. For example, micromachined pressure sensors first were contemplated in mid-1950s and became common in automotive applications only by the late 1980s. Similarly, * Corresponding author. E-mail: [email protected]. † General Electric Global Research Center. ‡ General Electric Industrial Sensing. 10.1021/ac052086q CCC: $33.50 Published on Web 07/15/2006

© 2006 American Chemical Society

micromachined accelerometers were invented in mid-1970s; however, it was not until the early 1990s when they became common in automotive applications.1 To date, numerous micromachined structures have been specifically designed and coated with responsive materials for the detection of chemical and biological species in water and air.2-4 Often, the resonance frequency of cantilevers is measured in air on an atomic force microscope5 or using an optical deflection readout.6-8 These costly, complicated, and bulky readout schemes make the whole sensor systems difficult to miniaturize and preclude them from being efficiently integrated into low-cost and small-size packaged sensor suites. It is well recognized that an ideal chemical/biochemical sensor should have the right combination of a physical transducer coupled with a responsive layer. Such a sensor recognizes changes in a chemical parameter and converts this information into an analytically useful signal.9 Taking advantage of previously developed, optimized, mass-produced, and thus cost-effective microfabricated physical transducers10,11 and optoelectronic components12-16 is an attractive avenue for chemical and biological detection. Interestingly, widely deployed and accepted commodity (1) Hagleitner, C.; Hierlemann, A.; Brand, O.; Baltes, H. In Sensors Update; Baltes, H., Go¨pel, W., Hesse, J., Eds.; VCH: Weinheim, 2002; Vol. 11, pp 101-155. (2) Semancik, S.; Cavicchi, R. Acc. Chem. Res. 1998, 31, 279-287. (3) Hagleitner, C.; Hierlemann, A.; Lange, D.; Kummer, A.; Kerness, N.; Brand, O.; Baltes, H. Nature 2001, 414, 293-296. (4) Lavrik, N. V.; Sepaniak, M. J.; Datskos, P. G. Rev. Sci. Instrum. 2004, 75, 2229-2253. (5) Su, M.; Li, S.; Dravid, V. P. Appl. Phys. Lett. 2003, 82, 3562-3564. (6) Hansen, K. M.; Ji, H.-F.; Wu, G.; Datar, R.; Cote, R.; Majumdar, A.; Thundat, T. Anal. Chem. 2001, 73, 1567-1571. (7) Dutta, P.; Tipple, C. A.; Lavrik, N. V.; Datskos, P. G.; Hofstetter, H.; Hofstetter, O.; Sepaniak, M. J. Anal. Chem. 2003, 75, 2342-2348. (8) Savran, C. A.; Knudsen, S. M.; Ellington, A. D.; Manalis, S. R. Anal. Chem. 2004, 76, 3194-3198. (9) Potyrailo, R. A. Angew. Chem., Int. Ed. 2006, 45, 702-723. (10) Boussaad, S.; Tao, N. J. Nano Lett. 2003, 3, 1173-1176. (11) Ren, M.; Forzani, E. S.; Tao, N. Anal. Chem. 2005, 77, 2700-2707. (12) Vo-Dinh, T.; Alarie, J. P.; Isola, N.; Landis, D.; Wintenberg, A. L.; Ericson, M. N. Anal. Chem. 1999, 71, 358-363. (13) Ivanisevic, A.; Yeh, J.-Y.; Mawst, L.; Kuech, T. F.; Ellis, A. B. Nature 2001, 409. (14) Manzano, J.; Filippini, D.; Lundstro ¨m, I. Sens. Actuators, B 2003, 96, 173179. (15) Cho, E. J.; Bright, F. V. Anal. Chem. 2001, 73, 3289-3293.

Analytical Chemistry, Vol. 78, No. 16, August 15, 2006 5633

consumer products also present an attractive set of capabilities applicable for advanced sensors. Examples include cell phones with additional functionality for alcohol detection17 and conventional CD or DVD optical disks with an array of chemical and biological sensor regions and conventional optical disk drives in laptop and desktop computers as quantitative optical readers.18 These examples conclusively demonstrate the power of implementation of existing electronic capabilities to accelerate the development of sensors with previously unavailable capabilities, functionality, or level or acceptance. In this work, we demonstrate that high-sensitivity micromachined transducers with an integrated piezoresistive readout designed, optimized, and mass-produced as automotive accelerometers at low cost can be coupled with the relevant sensor materials to construct chemical sensors. The accelerometer structures include a frame in which a seismic mass is disposed and coupled to the frame by two hinges (beams). A layer responsive to chemical species of interest is deposited onto the seismic mass. The concentration of the chemical species is sensed as a function of the flexure of the beams. Accelerometer structures are attractive as physical transducers in these chemical sensor applications for several key reasons. First, beams with varied different spring constants are available to match the required sensitivity of the chemical sensor. Second, the integrated readout of the transducer is optimized and very straightforward, not requiring external lasers and detectors that the user must manually adjust,19 yet providing a desired reproducibility in measurements. Third, easy access to the seismic mass surface allows for the deposition of chemically responsive films. Fourth, such transducers are indeed available as automotive accelerometers at low cost. EXPERIMENTAL SECTION Transducer Fabrication. The transducers employed in this study are a part of automotive accelerometer OEM components (GE Industrial Sensing, Fremont, CA). Electrical characteristics of the piezoresistive readout included the recommended Wheatstone bridge bias voltage of 5 V and sensitivity of the piezoresistive readout of 26 ( 2.2 µV V-1 g-1 at room temperature. Fabrication steps of the micromachined structure are schematically depicted in Figure 1.20 First, a single-crystal n-doped silicon substrate 1 was used and a recessed region 2 was formed as shown in Figure 1A. Next, p-type silicon substrate 3 was used to form an n-type layer 4 and to etch a recess region 5 from the n-type layer 4 as shown in Figure 1B. Substrate 1 and layer 4 were further aligned to match an edge of recess region 5 with an edge of recessed region 2 as shown in Figure 1C and bonded without the use of an intermediate glue material.21 Further, as shown in Figure 1D, (16) Cho, E. J.; Tao, Z.; Tehan, E. C.; Bright, F. V. Anal. Chem. 2002, 74, 61776184. (17) Seju Engineering. Micro alcohol sensors for cell phone; Presented at Transducers ‘05, Seoul, Korea, www.safe-drive.com, 2005. (18) Potyrailo, R. A.; Morris, W. G.; Leach, A. M.; Sivavec, T. M.; Wisnudel, M. B.; Boyette, S. M. Anal. Chem. 2006, same issue. (19) Baselt, D. R.; Lee, G. U.; Hansen, K. M.; Chrisey, L. A.; Colton, R. J. Proc. IEEE 1997, 85, 672-680. (20) Wu, G.; Mirza, A. R. Micromechanical device with thinned cantilever structure and related methods. U.S. Patent Application 20050172717, 2005 (available from www.uspto.gov) (21) Petersen, K. E.; Gee, D.; Pourahmadi, F.; Craddock, R.; Brown, J.; Christel, L. Proc. Transducers 91 1991, 397-399.

5634 Analytical Chemistry, Vol. 78, No. 16, August 15, 2006

Figure 1. Fabrication steps (A-F) of a micromachined structure20 for further deposition of a chemically or biologically sensing layer. Key: 1, n-doped silicon substrate; 2, recessed region; 3, p-type silicon substrate; 4, n-type layer; 5, recess region; 6, hinge; 7, p-type piezoresistive elements; 8, insulating oxide/nitride layer; 9, metal interconnects; and 10, seismic mass.

the p-type silicon substrate 3 was removed, leaving layer 4 on n-doped silicon substrate 1 and, thus, forming a base of beam hinge 6. Next, as shown in Figure 1E, two p-type piezoresistive elements 7 were formed near the hinge 6 in layer 4. The formation of piezoresistive elements 7 was performed by boron implantation and subsequent diffusion at 1100 °C. Further, an insulating oxide/ nitride layer 8 was formed over the layer 4 followed by the metallization of metal interconnects 9 coupled to the piezoresistive elements 7. Last, a deep reactive ion etch22 was performed to “release” hinge 6 and a seismic mass 10 as shown in Figure 1F. Thus, the seismic mass acquired the freedom for a rotational movement out of the horizontal plane of the substrate about an (22) Petersen, K. E.; Maluf, N.; McCulley, W.; Logan, J.; Klaasen, E.; Noworolski, J. M. Single-crystal silicon sensor with high aspect ratio and curvilinear structures. U.S. Patent 6,316,796, 2001.

Figure 2. Dimensions (in micrometers) of the micromachined transducer.

axis through the flexure region in response to an acceleration force applied in a direction generally perpendicular to the horizontal plane of the substrate. The finite element analysis (FEA) calculations of the spring constant of transducers were performed using ANSYS software. Test Setup for Chemical Sensing. For carbon dioxide detection, we employed polycarbonate (GE Global Research) polymer. Polycarbonate has been employed earlier for the reversible CO2 detection at room temperature.23 The seismic mass suspended over the recessed region of the substrate was coated with a polycarbonate sensing film using a spray coating method from a 5 wt % polycarbonate solution in chloroform. The coated device was positioned into a gas flow cell. Carbon dioxide gas was introduced into the flow cell at different ratios with dry nitrogen as a blank carrier gas. Different ratios of carbon dioxide to nitrogen were produced using a computer-controlled gas delivery system. Flexure of the seismic mass was sensed as a function of concentration of chemical species by monitoring an output voltage from the device. The recommended excitation voltage (the Wheatstone bridge bias voltage) of the piezoresistive transducer was 5 V. We explored two different levels of excitation voltage (2.5 and 7.5 V) to test transducer performance in the chemical sensing application with the expectation that if the total noise will be the same for different excitation voltages, the overall performance (detection limit) of the sensor will be improved at a higher excitation voltage. Gas flow control and data acquisition were performed using programs written in LabVIEW (National Instruments, Austin, TX). Data processing was performed using KaleidaGraph (Synergy Software, Reading, PA). RESULTS AND DISCUSSION Operation of Sensors. Transducers were fabricated, wirebonded, and mounted into TO-8 packages at GE Industrial Sensing. For chemical detection, transducers were selected with hinge thicknesses of 5 µm, seismic mass thicknesses of 25 µm, (23) Ahuja, A.; James, D. L.; Narayan, R. Sens. Actuators, B 1999, 72, 234-241.

and other dimensions as shown in Figure 2. The spring constant k of the transducer can be calculated as24

k)

RwEt3 2{(L + l)3 - L3}

(1)

where w is the width of the hinge, E is the Young’s modulus of silicon, t is the thickness of the hinge, R is a constant related to the number of employed hinges, l is the length of the hinge, and L is the length of the seismic mass. Using eq 1 and dimensions of the transducer (Figure 2), the spring constant of the transducer was calculated to be 3.68 N/m. Response of the sensor is measured as an output voltage V, which is related to the spring constant k as19

V)

FVb ∆R 4k R

(2)

where F is the force exerted on the transducer, Vb is the Wheatstone bridge bias voltage, and ∆R/R is the resistance change of the transducer per unit deflection (∆R) divided by the resistance of the transducer R. Equation 2 illustrates that the sensor response is inversely proportional to the transducer spring constant. FEA-based sensor simulations using ANSYS software were utilized for the optimization of the seismic mass and hinge dimensions of the transducer. A FEA-based optimization approach has been demonstrated as an accurate method of sensitivity optimization of piezoresistive transducers.25,26 The first three resonance modes were calculated to be 0.751, 21.662, and 38.957 kHz, respectively. The FEA-calculated spring constant was 3.14 (24) Rossel, C.; Bauer, P.; Zech, D.; Hofer, J.; Willemin, M.; Keller, H. J. Appl. Phys. 1996, 79, 8166-8173. (25) Nagler, O.; Trost, M.; Hillerich, B.; Kozlowski, F. Sens. Actuators, A 1998, 66, 15-20. (26) Lin, L.; Chu, H.-C.; Lu, Y.-W. J. Microelectromech. Syst. 1999, 8, 514-522.

Analytical Chemistry, Vol. 78, No. 16, August 15, 2006

5635

Figure 3. Micromachined transducer applied for chemical detection of CO2: (A) TO-8 wire bond diagram; (B) general view. An outer diameter of TO-8 package is 12 mm.

N/m, in agreement with the results obtained using eq 1.26 This achieved spring constant of the transducer was adequate for highsensitivity detection of CO2. Transducers with much higher spring constants of 20-80 N/m have been applied for detection of trace levels of explosives.27 Nanofabricated cantilevers with small spring constants typically require a bulky complicated optical readout.6-8 As can be seen from eq 1, the spring constant k is proportional to the cube of the thickness of the hinge t. The lower limit of the spring constant mainly depends on the microfabrication limit of the thickness of these hinges. In the present fabrication mode, the hinge thickness can be varied between 4.5 and 5.5 µm, which translates to the spring constant range of 2.68-4.9 N/m. However, the present fabrication technology can be easily extended to produce hinges with t ) 3 µm that will result in the spring constant of 0.79 N/m. Thus, because sensor response is inversely proportional to the spring constant (see eq 2), if needed, a more sensitive sensor can be straightforwardly produced with the spring constant k as low as 0.79 N/m. More sensitive sensors can potentially provide an improved detection limit by assuming that the sensor noise remains constant. Noise contributions in these sensors can be divided into phenomena intrinsic to the device and those related to interactions with its environment.4 As shown above, sensors with lower spring constants should provide an enhanced sensitivity. The noise of these lower spring constant sensors is not expected to increase because the main noise contributions are Johnson noise and 1/f noise, which are independent of the spring constant and the thermomechanical noise, which is directly proportional to the spring constant.28 Thus, given the same amount of noise and an increased sensitivity of response for smaller spring constant sensors, the detection limit is expected to be improved if only the noise contributions outlined above are considered. However, in practical sensors, additional noise contributions (for example, from the more pronounced imperfections in fabrication of small spring constant sensors) may become important, potentially making the sensor more noisy and degrading the detection limit. These noise contributions should require additional research. (27) Pinnaduwage, L. A.; Yi, D.; Tian, F.; Thundat, T.; Lareau, R. T. Langmuir 2004, 20, 2690-2694. (28) Harley, J. A.; Kenny, T. W. J. Microelectromech. Syst. 2000, 9, 226-235.

5636 Analytical Chemistry, Vol. 78, No. 16, August 15, 2006

Electrical connections to power the devices and to collect respective signals were made as shown in Figure 3A. For operation as a chemical sensor, Vsub and Vdd were connected together and a constant excitation voltage was applied across Vdd and GND. The sensor output voltage was measured through S+ and S-. An example of a packaged device for gas chemical sensor applications is illustrated in Figure 3B. Upon deposition of a polycarbonate sensing film, the sensor’s response to carbon dioxide was linearly proportional to the polymer film thickness. Polycarbonate polymer is a glassy polymer and has been employed earlier for the reversible CO2 detection at room temperature.23 The mechanism of the polymer response to CO2 is based on the diffusion of the gas into the “microvoids” of the polymer film resulting in an increase of the film mass. This polymer-CO2 interaction causes no detectable change in the mechanical properties of the film and minimal relaxation of the polymer matrix.23 Upon CO2 sorption into polycarbonate, the mass increase of the film was detected with the transducer. Other materials for CO2 determinations at room temperature include tetramethylammonium fluoride tetrahydrate,29 aminopropylsiloxane-octadecylsiloxane copolymer,30 polyethyleneimine,31 fluorinated polyimide,32 and tetrakis(hydroxyethyl)ethylenedlamine.33 Sensor Reproducibility. Reproducibility and hysteresis-free sensor response are among the key figures of merit of a chemical sensor. These metrics depend on the performance of both the transducer and sensing film. Reproducibility of the sensor response was evaluated by repetitively exposing sensor to different CO2 concentrations in both increasing and decreasing concentration sequences. We intentionally selected high CO2 concentrations (10-100% vol CO2) to provide the worst-case scenario for the sensor operation to study hysteresis effects. (29) Gomes, M. T. S. R.; Rocha, T. A.; Duarte, A. C.; Oliveira, J. A. B. P. Anal. Chim. Acta 1996, 335, 235-238. (30) Oprea, A.; Henkel, K.; Oehmgen, R.; Appel, G.; Schmeisser, D.; Lauer, H.; Hausmann, P. Mater. Sci. Eng. C 1999, 8-9, 509-512. (31) Korsah, K.; Ma, C. L.; Dress, B. Sens. Actuators, B 1998, 50, 110-116. (32) Hoyt, A. E.; Ricco, A. J.; Bartholomew, J. W.; Osbourn, G. C. Anal. Chem. 1998, 70, 2137-2145. (33) Fatibello-Filho, O.; de Andrade, J. F.; Suleiman, A. A.; Guilbault, G. G. Anal. Chem. 1989, 61, 746-748.

Figure 4. Reproducibility of sensor response evaluated by repetitively exposing sensor to different (0, 10, 30, 50, 75, and 100% vol) CO2 concentrations at two excitation voltages of (A) 2.5 and (B) 7.5 V.

Figure 5. Quantification of sensor hysteresis at two excitation voltages of 2.5 and 7.5 V. Sensor response at 2.5 V is 20.89 ( 0.08 mV/1% vol CO2. Sensor response at 7.5 V is 57.50 ( 0.38 mV/1% vol CO2.

Figure 6. Calibration results for the micromachined CO2 sensor for part-per-million CO2 concentrations. Inset, sensor response from 0 to 8000 ppm of CO2.

The piezoresistive readout technique requires an electrical current to flow through the transducer.4 Thus, while an increased high-excitation voltage should result in higher sensitivity of sensor response, it may also potentially add a thermally induced sensor noise4 (see previous section). The sensor was operated at two different excitation voltages, 2.5 and 7.5 V, to determine the effects of the excitation voltage on the sensor reproducibility. An initial sensor response was offset by negative 2000-3000 mV in order to operate over the full 10-V measurement range. Results of these experiments are presented in Figure 4. At these testing conditions, the sensor demonstrated an excellent reproducibility. Repetitive exposures to 100% vol CO2 provided signal reproducibility of 0.34 and 0.17% relative standard deviation (RSD) upon excitation with 2.5 and 7.5 V, respectively. Thus, operation of the sensor at 7.5 V provided better signal reproducibility and was further used for the evaluation of the detection limit. Quantification of sensor hysteresis was performed over the wide concentration range as shown in Figure 5. We have found that the response of the sensor to increasing and decreasing concentrations of CO2 had no detectable hysteresis effect. The sensitivity of the CO2 sensor operating at two excitation voltages of 2.5 and 7.5 V was 20.89 ( 0.08 and 57.50 ( 0.38 mV/1% vol CO2, respectively. The standard deviations for the mean values (0.08 and 0.38 mV/1% vol CO2) were calculated from the slopes of the calibration curves measured with the increasing and decreasing CO2 concentrations. Detection Limit and Dynamic Response. Evaluation of the sensor response to part-per-million concentrations of CO2 was performed to determine the detection limit of the sensor. Figure 6 demonstrates response of the sensor to low concentrations of CO2. At low concentrations of CO2, the sensor had a sensitivity of

response of 5.63 mV/1000 ppm CO2 as determined from the fit of the response over the 0-8000 ppm concentration range. This sensitivity was very similar to that obtained for high concentrations (57.50 mV/1% vol CO2). To determine the sensor detection limit, we evaluated sensor noise. We have found that sensor response had a noise level of ∼0.7 mV as shown in Figure 7A. However, the sensor noise contained periodic signal spikes that were less than 4 s in duration. From the analysis of the dynamic data collected during the reproducibility studies (see Figure 4), we evaluated the response time of the sensor to a chemical stimulus. We measured the time T90 required to achieve a 90% signal change upon a step change of CO2 gas concentration to be ∼45 s. The time scale of these periodic signal spikes was inconsistent with that of chemically relevant sensor response and likely was related to the electronic readout. Based on the temporal criteria of these signals, periodic signal spikes with durations less than 4 s were electronically removed. Removal of these signal spikes reduced the noise down to ∼0.3 mV (see Figure 7B). The calculated detection limit at the signal-to-noise ratio of 3 was 370 ppm for the noise shown in Figure 7A and was improved down to 160 ppm for the noise shown in Figure 7B. Effects of Vibration. Any accelerometer structure will be sensitive to vibrations if a prescribed data-sampling rate is maintained. In chemical detection, vibration-induced response is not desirable and will constitute an additional noise source. We have developed and tested several methods for reduction of such noise. Figure 8 demonstrates the successful removal of vibrationinduced noise to the sensor response. The vibrations were produced by placing the sensor on a high-speed laboratory shaker and operating the shaker at four settings with the relative vibration Analytical Chemistry, Vol. 78, No. 16, August 15, 2006

5637

Figure 7. Noise contributions of the micromachined sensor. (A) before and (B) after electronic removal.

Figure 8. Vibration-induced noise removal in response of the micromachined sensor. (A) Initial sensor response; (B) noise-reduced sensor response.

speeds set at 0, 2, 5, and 10, where the highest setting corresponded to the vibration speed of 3200 rpm. To cancel the vibration-induced noise, we sampled the sensor response with ∼1kHz sampling rate, followed by the electronic averaging of about 1000-2000 data points. Our results conclusively demonstrate that this physical transducer, initially developed as an accelerometer, can be successfully applied as a chemical sensor without effects of the vibrations. CONCLUSIONS A chemical sensor has been proposed and experimentally demonstrated based on the integration of a low-cost micromachined physical transducer with an appropriate sensor material. The development and optimization of microelectromechanical system (MEMS) structures often take more than a decade from their first invention to commercial implementations.1 Our work should encourage more analytical chemistry uses of previously developed MEMS devices for other demanding applications. We employ mass-produced silicon accelerometers for sensitive chemical detection by coating the transducers with a polymer film (34) Lemmo, A. V.; Fisher, J. T.; Geysen, H. M.; Rose, D. J. Anal. Chem. 1997, 69, 543-551. (35) de Gans, B.-J.; Kazancioglu, E.; Meyer, W.; Schubert, U. S. Macromol. Rapid Commun. 2004, 25, 292-296. (36) Turcu, F.; Hartwich, G.; Scha¨fer, D.; Schuhmann, W. Macromol. Rapid Commun. 2005, 26, 325-330. (37) Singh, B. K.; Hillier, A. C. Anal. Chem. 2006, 78, 2009-2018.

5638

Analytical Chemistry, Vol. 78, No. 16, August 15, 2006

responsive to carbon dioxide. Such an approach can be easily adapted for detection of other chemical and biological species. These chemical or biological sensors also have advantages of simplicity of integration with other types of existing micromachined sensors to achieve multiparameter measurements and production of such sensor platforms at low cost. Of course, micromachined chemical sensors reported here should have effects similar to that of interfering species because these effects originate from the contributions of the sensing material deposited onto the transducer. Fabrication of our sensors in an array format is quite straightforward and will include only (1) a new wire bonding layout to incorporate excitation of multiple transducers and their readout and (2) a microspot deposition of sensor materials on individual transducers. Numerous ink-jet and microspotter systems are available.34-37 Such sensors should be attractive not only for long-term unattended monitoring applications but also as detectors in more complicated microanalytical systems. ACKNOWLEDGMENT We gratefully acknowledge Dr. Greg Chambers for the partial support of this work from GE Corporate long-term research funds. Received for review November 27, 2005. Accepted June 6, 2006. AC052086Q