Graphite Microparticles as Coatings for Quartz Crystal Microbalance

Anal. Chem. 2000, 72, 5981-5987

Graphite Microparticles as Coatings for Quartz Crystal Microbalance-Based Gas Sensors Ruth Shinar,* Guojun Liu,† and Marc D. Porter*

Microanalytical Instrumentation Center, Ames Laboratory-USDOE, and Department of Chemistry, Iowa State University, Ames, Iowa 50011

The use of graphite particles (1-2 µm) as coatings on quartz crystal microbalances (QCMs) for detection and monitoring of toluene and other volatile organic compounds (VOCs) is described. Unlike the more commonly used polymeric coatings with low glass transition temperatures (Tg), particulate graphitic coatings are not as susceptible to loss of acoustic energy when coating thickness or operational temperature increases. This situation enables the use of relatively thick coatings, which increases the absolute amount of vapor sorbed in the coating and, consequently, lowers the level of detection and enhances operation over a wide temperature range. The use of small size particles also results in a coating with a more porous structure, which facilitates uptake and release of VOCs in comparison to coatings made from high Tg polymers, which have a lower porosity. These attributes, coupled with the inherent stability of graphitic materials, make particulate graphite coatings especially suitable for applications at high temperatures. The advantages of using particulate graphite as a coating on QCMs are demonstrated by comparison to the performance of a few low-Tg polymers [i.e., poly(isobutylene) and poly(diphenoxyphosphazene)] and high-Tg polymers (i.e., polystyrene). Mass-sensitive chemical sensors, such as quartz crystal microbalances (QCMs) and surface acoustic wave (SAW) devices coated with polymeric films or other materials, are often used for the detection and monitoring of volatile organic compounds (VOCs) and other analytes.1-9 To date, coating development has focused * To whom correspondence should be addressed. R.S.: (515)-294-5898; Fax, (515)-294-3254; E-mail, [email protected]. M.D.P.: (515)-294-6433; Fax, (515)294-3254; E-mail, [email protected]. † Current address: Photonic Sensor, 2930 Amwiler Court, Suite A, Atlanta, GA 30360. (1) Grate, J. W.; Martin, S. J.; White, R. M. Anal. Chem. 1993, 65, 987A996A. (2) Zellers, E. T.; Batterman, S. A.; Han, M.; Patrash, S. J. Anal. Chem. 1995, 67, 1092-1106. (3) Patrash, S. J.; Zellers, E. T. Anal. Chem. 1993, 65, 2055-2066. (4) Zellers, E.; Han, M. Anal. Chem. 1996, 68, 2409-2418. (5) Martin, S. J.; Frye, G. C.; Senturia, S. D. Anal. Chem. 1994, 66, 22012219. (6) (a) Grate, J. W.; Klusty, M.; McGill, R. A.; Abraham, M. H.; Whiting, G.; Andonian-Haftvan, J. Anal. Chem. 1992, 64, 610-625. (b) Grate, J. W.; Zellers, E. T. Anal. Chem. 2000, 72, 2861-2868. (7) Ballantine, D. S. Anal. Chem. 1992, 64, 3069-3076. 10.1021/ac0009548 CCC: $19.00 Published on Web 11/18/2000

© 2000 American Chemical Society

primarily on polymers that have glass transition temperatures (Tg) below room temperature because vapor sorption in these softer materials is more reversible and much more rapid than in the stiffer polymers that have higher Tg values.1-3 Ideally, the sensitivity of an acoustic sensor is directly proportional to the polymer film thickness. That is, as the thickness increases, larger amounts of vapor can be sorbed, which results in enhanced sensitivity. However, efforts to increase sensitivity when using a low Tg polymer through increases in film thickness are often compromised by the low shear modulus of the material, which results in a detrimental attenuation of the acoustic energy.1,5,10,11 This attenuation manifests itself through a reduction in signal-tonoise ratio (SNR) and, consequently, a degradation in the limit of detection (LOD). The attenuation of the acoustic energy that accompanies an increase in coating thickness is generally associated with a phase lag in the motion of the outer portion of the coating with respect to that located closer to the QCM surface.1,5 This attenuation is amplified by increases in temperature and vapor sorption. Increases in temperature result in the thermal expansion of the coating, and vapor sorption can cause the polymer to swell. Both situations can also soften the coating,1,5,6,12,13 which increases the phase lag that arises from nonsynchronous film motion. Furthermore, the phase lag can be accompanied by a decay in the displacement of the outer portions of the coating.1 The increased attenuation of the acoustic wave that results from increasing coating thickness, thermal expansion, or swelling restricts the upper limit of coating thickness. This problem is particularly acute when an oscillator circuit is used to drive a QCM because the device will cease to oscillate when the insertion loss exceeds the gain of the oscillator circuit.1 The situation is different with coatings of stiff materials, including high-Tg polymers.11,14-16 (8) Crooks, R. M.; Ricco, A. J. Acc. Chem. Res. 1998, 31, 219-227. (9) Buttry, D. A.; Ward, M. D. Chem. Rev. 1992, 92, 1355-1379. (10) Lucklum, R.; Behling, C.; Cernosek, R. W.; Martin, S. J. J. Phys. D: Appl. Phys. 1997, 30, 346-356. (11) (a) Lucklum, R.; Behling, C.; Hauptmann, P. Anal. Chem. 1999, 71, 24882496. (b) Lucklum, R.; Behling, C.; Hauptmann, P. Sens. Actuators 2000, B65, 277-283. (12) Grate, J. W.; Kaganove, S. N.; Bhethanabotla, V. R. Faraday Discuss. 1997, 107, 259-283. (13) Martin, S. J.; Frye, G. C. Appl. Phys. Lett. 1990, 57, 1867-1869. (14) Shinar, R.; Liu, G.; Porter, M. D. Unpublished results, 1999. (15) Ricco, A. J.; Staton, A. W.; Crooks, R. M.; Kim, T. Faraday Discuss. 1997, 107, 247-258. (16) Hager, H. E.; Escobar, G. A.; Verge, P. D. J. Appl. Phys. 1986, 59, 33283331.

Analytical Chemistry, Vol. 72, No. 24, December 15, 2000 5981

High-Tg polymers exhibit a smaller attenuation of the acoustic wave energy. However, their response to the uptake/loss of an analyte is generally much slower. A thinner coating can be employed to decrease the response time but only by compromising sensitivity. As a consequence, the vast majority of earlier research has focused on the use of low-Tg polymers.1-3 In an attempt to combine the more rapid response that is an attribute of soft coatings with reduced attenuation of the acoustic energy that is an advantageous feature of stiffer coatings, we tested the performance of QCMs coated with graphitic microparticles. We hypothesized that graphite, which is a mechanically stiff material, will not be as lossy as a low-Tg polymer. Furthermore, a coating of microparticles should result in a more open, porous structure, which would facilitate the uptake/release of gaseous analytes in comparison to high-Tg polymers. Coating selection for acoustic wave-based sensors is often influenced by drawing an analogy between sensor performance and the separation capabilities of the stationary phases used in both gas and liquid chromatography.3 Porous graphitic carbon has been employed, for example, as a stationary phase in highperformance liquid chromatographic separations.17,18 Other carbonbased materials have been used in various analytical applications; for example, modified carbon adsorbents in gas chromatography,19 carbon fibers in biosensors,20 modified glassy carbon electrodes21,22 in electroanalytical applications, and carbon black/polymer composite films in gas detectors based on resistance measurements.23 However, the potential advantages24 of utilizing such materials as coatings in acoustic wave-based gas sensors has been exploited in only a few instances. In an earlier study, a QCM coated with soot resulting from the combustion of chlorobenzoic acid was employed to detect selectively low levels (i.e., 1-60 ppm) of hydrogen sulfide.25 More recently, QCMs coated with pyrolyzed graphite films, which were prepared from mixtures of carbon powder, graphite particles, and a resin material,26,27 were employed to detect low levels (∼1 ppm) of ethanol.26 This paper describes the performance of particulate graphite coatings exposed to toluene and other VOCs. These coatings, which were deposited on both 9 and 10 MHz QCMs from a particulate dispersion containing a cellulosic additive, exhibited a significantly faster mass uptake and release in comparison to high Tg polymer coatings. Moreover, particulate graphite coatings could be used without significant attenuation of the acoustic wave energy at elevated temperatures (i.e., 50 °C). The results of (17) Bassler, B. J.; Hartwick, R. A. J. Chromatogr. Sci. 1989, 27, 162-165. (18) Forgacs, E.; Cserhati, T. Trends Anal. Chem. 1995, 14, 23-29. (19) Gavrilova, K. B.; Vlasenko, E. V.; Petsev. N.; Topalova, I.; Dimitriv, Chr.; Ivanov, S. Liq. Chromatogr. 1988, 454, 73-81. (20) Pantano, P.; Kuhr, W. G. Anal. Chem. 1993, 65, 623-630. (21) Ryan, M. D.; Bowden, E. F.; Chambers, J. Q. Anal. Chem. 1994, 66, 360R427R. (22) Ratcliff, B. B.; Klancke, J. W.; Koppang, M. D.; Engstrom, R. C. Anal. Chem. 1996, 68, 2010-2014. (23) Severin, E. J.; Doleman, B. J.; Lewis, N. S. Anal. Chem. 2000, 72, 658668. (24) Capps, R. N. J. Acoust. Soc. Am. 1985, 78, 406-413. (25) Webber; L. M.; Karmarkar, K. H.; Guillbault, G. G. Anal. Chim. Acta 1978, 97, 29-35. (26) Kim, J.-M.; Chang, S.-M.; Suda, Y.; Muramatsu, H. Sens. Actuators 1999, A72, 140-147. (27) Ikarashi, Y.; Blank, C. L.; Suda, Y.; Kawakubo, T.; Maruyama, Y. J. Chromatogr. 1995, 718, 267-272.

5982

Analytical Chemistry, Vol. 72, No. 24, December 15, 2000

experiments demonstrating these attributes are described, drawing in large part on comparisons to the performance of a few of the commonly employed low-Tg polymers [i.e., poly(isobutylene) (PIB) and poly(diphenoxyphosphazene) (PDPP)]. EXPERIMENTAL SECTION Coatings. Both 9 and 10 MHz QCMs were used. The 9 MHz QCMs were purchased from Standard Crystals, with Cr/Au (15/ 150 nm) contacts formed using a resistive heating evaporator (Edwards Coating System E306A). The 10 MHz crystals were obtained from International Crystal Manufacturing, and were already coated with Cr/Au electrodes. Porous particulate graphite coatings were prepared from a commercially available dispersion (Acheson) of 1-2 µm graphite particles in a 2-propanol matrix containing a cellulosic additive.28 Thin graphitic films were applied onto the QCMs using several conventional application procedures, including spray coating, solvent evaporation, and spin coating. The sensors were then dried in air at 55-100 °C for 5-24 h. However, higher drying temperatures (i.e., 80-100 °C) usually resulted in coatings with lower sensitivities. The data presented herein are for coatings dried at 55-65 °C. Coatings were also formed using PIB (molecular weight 420 000, Polysciences) and PDPP (molecular weight 379 000, Scientific Polymer Products). The PIB and PDPP have static Tg5,10 values of -73 °C and -20 °C, respectively. These coatings were cast by spin coating (2000-5000 rpm) solutions 0.3-3 wt % in toluene, chloroform, or tetrahydrofuran and then annealed at 6585 °C for 5-24 h. Instrumentation. The vapor generation system consisted of a gas stream module and a pair of three-way switchable valves that led into a test fixture that can house six separate sensors.29 The gas stream module included a reference module, which served to establish the baseline response, and an analyte module. The reference gas was typically dry nitrogen. The analyte vapor was generated by means of bubblers and calibrated mass-flow controllers (Tylan General FC-280AV), which were used to manipulate the analyte concentration. The bubbler was composed of two connected compartments, each containing the liquid analyte. The dry nitrogen carrier gas was introduced into the first compartment; the second compartment served as a headspace equilibrator. This process resulted in a gas stream that was saturated with analyte vapor, the concentration of which was manipulated by changing the flow rate and by mixing with dry nitrogen. Two measurement protocols were employed. In the first measurement protocol, termed “additive flow”, the reference dry nitrogen gas flowed through the test fixture during the entire experiment. The saturated analyte vapor was added at consecutively increasing levels (i.e., flow rates), with the reference flow adjusted to maintain a constant total flow rate, typically 100 or 200 sccm. In the second measurement protocol, termed “alternate flow”, the reference gas was passed through the test fixture before and after each exposure of a sensor to analyte vapor. Both measurement protocols yielded similar results. (28) We attribute the coating performance primarily to sorption on the graphite particles. Although we were unable to rule out a contribution from the cellulosic binder, experiments using particulate Teflon that were prepared from solutions of the same additive support our conclusion. (29) Liu, G.; Shinar, R.; Porter, M. D. Manuscript in preparation.

The sensors were exposed to either the analyte vapor or the reference gas by means of the computer-controlled three-way valves and an MKS multi-gas controller (model 147B) that was connected to the mass-low controllers. The gas generation system provided accurate and reproducible VOC levels, as verified by tests using a flow cell and UV-visible spectrophotometry. The sensor temperature was controlled by placing the sensor housing and associated tubing in an environmental chamber (Sun Electronic System, EC 127). Equivalent electrical circuit parameters of the sensors,9,12,30 before and following exposure to the vapor, were determined using an HP8753C network analyzer interfaced to a PC running HP85165A resonator-measurement software.30 In a one-port representation, the parameters include the equivalent series resistance, R1, the motional inductance, L1, the motional capacitance, C1, and the static capacitance, C0. The motional parameters were computed from the admittance characteristics of the device, which was obtained from network analyzer measurements of the reflection coefficient near resonance. The value of C0 was calculated from measurements at frequencies off-resonance. We used only the series resonant frequency, f, and R1 to describe sensor performance. The value of f was calculated by mathematically shifting the admittance points to present the locus of the series resonant circle. The shifted points were then converted to an impedance, and the reactance was fit to a polynomial function of the frequency. By definition, f is the frequency at zero reactance. The value of R1 was obtained from the admittance plane locus. The data obtained from the measurements were analyzed to yield the resonant frequency shifts (∆f) relative to dry nitrogen and the changes in R1. The error bars for R1 are typically ∼5% for measurements made within a few hours. This level of precision applies to all the reported data, except that in Figure 1 (see caption for details). Chemicals. Xylene (a mixture of isomers), octane, toluene, trichoroethylene, methyl ethyl ketone, 1,1,1 trichloroethane, hexane (a mixture of isomers), acetone, and chloroform were purchased from Fisher Scientific. Tetrahydrofuran was obtained from Mallinckrodt and ethanol and methanol, from Aldrich. RESULTS AND DISCUSSION In evaluating the particulate graphite coatings, measurements focused on the determination of f, R1, vapor uptake and release times, and the detection sensitivity (S) for toluene and other VOCs as key figures of merit. The effects of coating thickness and measurement temperature were also examined. R1 is a key indicator of energy loss, with increased energy dissipation revealed by its increase.9,11 As R1 increases, the noise in the measurement and, therefore, the LOD increase. We further note that coating thicknesses are given in terms of the change in resonant frequency that is caused by application of the coating. For example, a 100 kHz PDPP coating on a 9 MHz QCM is about 4.5 µm thick, as calculated by using the Sauerbrey equation.31 Series Resistance. Figure 1 plots the dependence of R1 on film thickness for QCMs that are coated with PIB, PDPP, and particulate graphite at 23 and 50 °C. For comparison, R1 values of an uncoated QCM at 23 and 50 °C are also presented. PIB and (30) O’Toole, R. P.; Burns, S. G.; Bastiaans, G. J.; Porter, M. D. Anal. Chem. 1992, 64, 1289-1294. (31) Sauerbrey, G. Z. Phys. 1959, 155, 206-222.

Figure 1. The series resistance of 9 and 10 MHz QCMs coated with PIB, PDPP, and particulate graphite as a function of coating thickness at 23 and 50 °C. The lines are a guide to the eye. Uncoated 9 MHz QCM at 23 and 50 °C (4), PIB on 9 MHz QCM at 23 °C (b), PIB on 9 MHz QCM at 50 °C (9), PDPP on 9 MHz QCM at 23 °C (2), particulate graphite on 9 MHz QCM at 23 °C (0), particulate graphite on 10 MHz QCM at 23 °C (]), particulate graphite on 10 MHz QCM at 50 °C (3). The uncertainty of the value of R1 for these data is ∼15%, which is a reflection of the reproducibility of the measurement over a time period of ∼1 year.

PDPP are used to illustrate the behavior typical of low-Tg polymers. As can be seen, the R1 values of the PIB-coated sensors at 23 °C do not change significantly up to a coating thickness of ∼70 kHz. Above ∼70 kHz, however, R1 undergoes a notable increase, approaching 250 Ω at ∼130 kHz. A comparable dependence of R1 on thickness was observed for the PDPP-coated QCMs at 23 °C. The increase of R1 with film thickness is indicative of an increase in energy dissipation that results when the movement of the outer portion of the coating, which may be accompanied by a decay in displacement, lags behind that of the coating that is closer to the film/QCM boundary.1,5 Increasing the temperature to 50 °C has a dramatic impact on R1 and its dependence on PIB thickness. At small thicknesses (e.g., 50 kHz), R1 is roughly twice that at 23 °C. Above ∼70 kHz, the PIB coating exhibits a marked increase in R1. Between ∼70 kHz and ∼110 kHz, R1 increases from ∼100 to ∼400 Ω. At this level of R1, the performance of the sensor deteriorates due to an overall increase in the noise level of the measurement. This behavior is attributed to a decrease in the shear modulus of the coating with increasing temperature, reflecting a softening of the film by thermal expansion. The decreased modulus enhances nonsynchronous film motion, which is accompanied in lossy films (e.g., PIB and PDPP) by an increasing decay in the physical displacement of the coating as the distance from the QCM surface increases.1,5 In contrast to QCMs coated with PIB and PDPP, those coated with particulate graphite show only a small increase in R1 as a function of thickness or temperature. The value of R1 for graphitecoated 9 MHz QCMs does not show an identifiable increase up to a thickness of ∼150 kHz. At larger thicknesses, the increase is Analytical Chemistry, Vol. 72, No. 24, December 15, 2000

5983

Table 1. Values of the Equivalent Series Resistance (R1) in Dry Nitrogena and upon Exposure to 1000 ppm Toluene at -10, 25, and 50 °C and Temperature Coefficients (r) for 10 MHz QCMs Coated with PIB, PDPP, and Particulate Graphite 93.3 kHz PIB temp (°C) -10 25 50 R (ppm/°C) a

R1 (Ω) 77.8 kHz PDPP

dry N2 toluene dry N2 toluene 19 89 318 595

29 95 329

10 64 312 850

18 70 321

180 kHz graphite dry N2

toluene

56 59 67 75

69 61 68

Total gas flow rate: 200 sccm.

relatively small, that is, the R1 value is only 50-60 Ω for a thickness of 250 kHz. Similarly, the values of R1 for 140-217 kHz graphite-coated 10 MHz QCMs are less than 80 Ω at both 25 and 50 °C. The notable increases in R1 with increasing temperature for the QCMs coated with the low-Tg polymers are also evident from the data presented in Table 1. This Table compares the R1 values at -10, 25, and 50 °C for QCMs coated with 93.3 kHz PIB, 77.8 kHz PDPP, and 180 kHz particulate graphite. For the PIB coating, R1 increased from 19 Ω at -10 °C to 89 Ω at 25 °C, and to 318 Ω at 50 °C. Similar increases were observed for a 77.8 kHz PDPP coating, for which R1 increased from 10 Ω at -10 °C to 312 Ω at 50 °C. As mentioned earlier, we attribute these increases to a softening of the coating, which is due to its thermal expansion, because there is no detectable change in R1 for an uncoated QCM upon heating from 25 to 50 °C. In sharp contrast, R1 values for the QCMs that were coated with thicker particulate graphite coatings are only marginally affected by temperature, with R1 for the 180 kHz coating equal to 56 Ω at -10 °C, 59 Ω at 25 °C, and 67 Ω at 50 °C. As another assessment of performance, the effect of exposure to 1000 ppm toluene on R1 is presented in Table 1. For the 180 kHz particulate graphite coating, there was no detectable change in R1 at 50 °C. An increase from 56 to 69 Ω, however, was observed at -10 °C. This increase is attributed to the large amount of vapor sorbed on the QCM at -10 °C. The ratio of the amounts sorbed at -10 and 50 °C for a given coating and exposure level can be roughly estimated from the relative magnitudes of the saturation vapor pressures of the analyte (P0) at the two temperatures. This ratio is ∼24 for toluene. Larger increases (i.e., increases relative to R1 in the absence of toluene) were observed for the PIB and PDPP coatings at -10, 25, and 50 °C. These increases, which indicate greater attenuations in acoustic energy, are ascribed to an increased swelling of the softer, less porous coatings, and therefore, a decreased shear modulus. To summarize briefly, the larger increases in R1 with increased thickness, temperature, and vapor sorption for QCMs coated with PIB and PDPP reflect a more substantial loss of the acoustic wave energy (which is accentuated when the dynamic Tg of the polymer is within the measurement temperature range) that is associated with decreasing shear moduli.1,5,13 In contrast, the smaller increases in R1 for QCMs that were coated with particulate graphite indicate significantly lower losses of acoustic wave energy, which is expected for less lossy films such as those formed using 5984 Analytical Chemistry, Vol. 72, No. 24, December 15, 2000

Figure 2. Toluene response of 55.5 kHz PIB (b) and 124 kHz particulate graphite (0) coatings on 9 MHz QCMs at 23 °C. The final frequency rise is due to the purging of the sample chamber with dry nitrogen.

metallic, crystalline, and very thin polymer materials.5 As mentioned, smaller increases in R1 were also observed for stiff, highTg polymer coatings.11,14 However, only very thin coatings of the latter are typically usable because their response and recovery times are long. Temperature Coefficients. In addition to exhibiting smaller changes in R1 with temperature, the frequencies of QCMs coated with particulate graphite show a smaller dependence on temperature. As can be seen in Table 1, the temperature coefficients (R), which, after normalization to coating thickness, are expressed in terms of the change in the frequency with respect to the change in temperature (from 25 to 50 °C), of the soft coatings are larger than that of particulate graphite. Thus, the smaller temperature coefficient of particulate graphite reduces the temperature-induced drift in the baseline frequency. Response and Recovery Times. Figure 2 shows the response of a 124 kHz particulate graphite coating upon exposure to increasing levels of toluene vapor at 23 °C. For comparison, the response of a 55.5 kHz PIB coating is also presented. As can be seen, the response and recovery times of the thicker particulate graphite coating are slower than those for the thinner PIB coating, but are significantly faster in comparison to thinner coatings of stiff polymers [e.g., the response time of a 2.2 kHz polystyrene (Tg ∼ 100 °C) coating to ∼55 ppm toluene exceeded 30 min].14 We believe that the faster responses of the particulate graphite coatings reflect a more porous morphology. The same general behavior has been observed for other coatings that were prepared from small-size polystyrene particles.29 Figure 3 shows the responses of a 140 kHz particulate graphite coating upon exposure to increasing levels of toluene vapor at 50 °C. As evident, the loss of analyte at the end of the exposure sequence (i.e., recovery time) is faster than that of a coating with a similar thickness at 23 °C (see Figure 2). Faster responses are expected at 50 °C because of enhanced mass transfer. However, the amounts of sorbed analyte at 50 °C are reduced, reflecting, in

Figure 3. Response of a 140 kHz particulate graphite coating on a 10 MHz QCM to toluene at 50 °C. The final frequency rise is due to the purging of the sample chamber with dry nitrogen.

part, the increase in the vapor pressure of toluene with temperature.3,4 Nevertheless, the frequency shifts of the particulate graphite coating at 50 °C are still appreciable in comparison to those of low-Tg polymeric coatings (see Figure 2). This situation, together with the inherent stability of the material at elevated temperatures and its low temperature coefficient, make particulate graphite coatings particularly suitable for high-temperature applications, without the limitations of either a high- or low-Tg material. We also note that the responses become slower as the temperature decreases because of a decrease in mass transfer. For example, the recovery time at -10 °C for a 10 MHz QCM coated with 152 kHz particulate graphite was ∼45 min, and that of a 100 kHz PIB coating was ∼60 min, whereas the recovery times at 25 °C were ∼9 and ∼1.5 min, respectively. Thinner particulate graphite (