Anal. Chem. 2002, 74, 1307-1315
Properties of Vapor Detector Arrays Formed through Plasticization of Carbon Black-Organic Polymer Composites Michael E. Koscho, Robert H. Grubbs,* and Nathan S. Lewis*
Division of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California 91125
Arrays of vapor detectors have been formed through addition of varying mass fractions of the plasticizer diethylene glycol dibenzoate to carbon black-polymer composites of poly(vinyl acetate) (PVAc) or of poly(N-vinylpyrrolidone). Addition of plasticizer in 5% mass fraction increments produced 20 compositionally different detectors from each polymer composite. Differences in vapor sorption and permeability that effected changes in the dc electrical resistance response of these compositionally different detectors allowed identification and classification of various test analytes using standard chemometric methods. Glass transition temperatures, Tg, were measured using differential scanning calorimetry for plasticized polymers having a mass fraction of 0, 0.10, 0.20, 0.30, 0.40, or 0.50 of plasticizer in the composite. The plasticized PVAc composites with Tg < 25 °C showed rapid responses at room temperature to all of the test analyte vapors studied in this work, whereas composites with Tg > 25 °C showed response times that were highly dependent on the polymer/analyte combination. These composites showed a discontinuity in the temperature dependence of their resistance, and this discontinuity provided a simple method for determining the Tg of the composite and for determining the temperature or plasticizer mass fraction above which rapid resistance responses could be obtained for all members of the test set of analyte vapors. The plasticization approach provides a method for achieving rapid detector response times as well as for producing a large number of chemically different vapor detectors from a limited number of initial chemical feedstocks. Arrays of chemically sensitive resistors, formed from composites of conductors and insulating organic polymers, provide an interesting approach to the classification, identification, and quantification of vapors.1-5 Each individual detector in such an (1) 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) Doleman, B. J.; Severin, E. J.; Lewis, N. S. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 5442-5447. (3) Lonergan, M. C.; Severin, E. J.; Doleman, B. J.; Beaber, S. A.; Grubbs, R. H.; Lewis, N. S. Chem. Mater. 1996, 8, 2298-2312. (4) Swann, M. J.; Glidle, A.; Cui, L.; Barker, J. R.; Cooper, J. M. Chem. Commun. 1998, 2753-2754. (5) Patel, S. V.; Jenkins, M. W.; Hughes, R. C.; Yelton, W. G.; Ricco, A. J. Anal. Chem. 2000, 72, 1532-1542. 10.1021/ac011054+ CCC: $22.00 Published on Web 02/14/2002
© 2002 American Chemical Society
array responds to a variety of analytes, and each analyte in turn elicits a response from a variety of detectors. Pattern recognition algorithms, in conjunction with a database of the array response signals produced by a variety of analytes, are then used to obtain information on the identity and concentration of test vapors.1,6-29 Improved detector array performance can often be achieved by increasing the chemical diversity of the detectors in the array.30-32 Although a subset (10 min to disperse the carbon black particles, and the resulting suspensions were sprayed onto detector substrates using an air brush. Films thicknesses range from 242 to 560 nm for the PVAc-DGD carbon black composite array and from 107 to 8624 nm for the PVPd-DGD carbon black composite array, as measured by profilometry (see Supporting Information). After fabrication, all of the detectors were placed in a stream of dry air for at least 24 h and were then exposed to a number of the test solvents until reproducible responses were obtained. B. Data Collection. The dc electrical resistance of each detector was measured using a Keithley model 2002 multimeter connected to a Keithley model 7001 multiplexer. The resistance of each detector was recorded approximately every 7-8 s during data collection. The apparatus used to control and deliver the vapor streams consisted of mass flow controllers, solvent bubblers, and solenoid valves controlled by LabView software, as described previously.42 A flow of laboratory air (1.10 ( 0.15 parts per thousand (ppth) of water vapor) through the solvent bubblers provided a saturated stream of vapor, which was then diluted to the desired concentration with background laboratory air and delivered to the detectors. Ten exposures to each of 11 test vapors were performed, with the test vapors arranged in two groups of 5 or 6 bubblers each. Exposures were performed in random order within each group of analytes at P/Po ) 0.050, where P is the partial pressure and Po is the vapor pressure of the analyte at room temperature. Group 1 consisted of (concentration in ppth) methylene chloride (22.8), tetrahydrofuran (8.5), toluene (1.4), hexane (9.2), ethanol (3.0), and acetone (12.1), whereas group 2 consisted of ethyl acetate (4.5), acetonitrile (4.6), cyclohexanone (0.22), chloroform (10.1), and methanol (6.2). To obtain response data at ambient temperature (21 ( 1 ° C), all 40 detectors of one of the arrays were placed in a 200-mL chamber constructed of Teflon and aluminum. Laboratory air was then passed for 180 s over the detectors, followed by a 300-s stream of analyte at P/Po ) 0.050 followed additionally by 600 s of air flow to ensure that the detectors had recovered to their initial resistance values. The total flow rate was 5 L min-1 during the experiments. The total elapsed time during this set of exposures to both groups of analytes was 36 h. Additional experiments were performed in which the concentration of chloroform was varied. In these runs (which were collected two months after completion of the initial two groups of analyte exposures at ambient temperature), a portion of the background laboratory air stream was passed through a water bubbler to (42) Severin, E. J.; Doleman, B. J.; Lewis, N. S. Anal. Chem. 2000, 72, 658668.
produce a gas stream that was saturated with water vapor. Using three mass flow controllers, the resulting gas stream was mixed with a separate chloroform-saturated vapor stream and with another background flow of laboratory air to produce the desired analyte mixture for presentation to the detectors. To obtain data at other temperatures, two detectors were placed onto a brass block that was connected to a 50-mL sample chamber. A K-type thermocouple was affixed with epoxy to the top of a piece of glass that was the same thickness as the detector substrates and that was placed between the detectors. Water from a circulating temperature bath was then passed through a hole that had been drilled through the brass block. Before being delivered to the detectors, the gas stream that led to the detectors was passed through a ∼3-m coil of copper tubing that was submerged in the temperature bath. The temperature reading of the thermocouple and the resistances of both detectors were monitored throughout each experiment using a Hewlett-Packard model 34970A meter that recorded the value of each quantity every 0.7-0.8 s. Digital scanning calorimetry (DSC) was performed using a Perkin-Elmer Pyris 1 digital scanning calorimeter. Each sample was prepared by spraying the polymer-plasticizer-carbon black suspensions onto a clean microscope slide and allowing the film to dry in ambient air for at least 24 h. To mimic the conditions under which the detectors were used, no additional drying or other processing was performed on these detector films. After deposition, the films were removed from the glass and placed into aluminum pans. Each of the samples was cycled between -100 and 150 °C at a rate of 10 °C min-1 to collect DSC data. C. Data Analysis. Although the resistance of each detector was sampled once every 7-8 s during each exposure, only the maximum relative differential resistance change, ∆Rmax/Rb, where ∆Rmax is the maximum resistance change of the detector during the exposure, taken as the average of five data points about the maximum measured differential resistance, and Rb is the baseline resistance of the detector before the exposure, taken as the average of the five data points immediately prior to exposure, was used in the analysis of the data. Statistical methods based on cluster analysis using the Fisher linear discriminant methodology were used to analyze detector array data, as described previously.30-32,43 RESULTS A. Resistance Responses of Plasticized Poly(vinyl acetate) and Plasticized Poly(N-vinylpyrrolidone) Carbon Black Composite Vapor Detectors to Various Analytes. Figure 1 presents ∆Rmax/Rb values for the plasticized PVAc chemiresistor array upon exposure for 300 s to methanol, acetone, and cyclohexanone at P/Po ) 0.050 in a laboratory air background. Clearly, the ∆Rmax/ Rb values did not vary monotonically with the mass fraction of plasticizer in the detector films. A summary of the data for all analytes and all detectors is included in the Supporting Information. Maximum ∆Rmax/Rb values for this series of detectors at 21 ( 1 °C were observed at DGD mass fractions, Mf (DVD/PVAc), equaling 0.25 for methylene chloride, 0.30 for acetonitrile, 0.30 for ethyl acetate, 0.30 for chloroform, 0.30 for ethanol, 0.30 for (43) Duda, R. O.; Hart, P. E. Pattern Classification and Scene Analysis; John Wiley & Sons: New York, 1973.
Figure 1. Relative differential resistance responses of PVAccarbon black composites at varying mole fractions of DGD to (a) methanol, (b) acetone, and (c) cyclohexanone, each at 5% of its vapor pressure at room temperature. Error bars represent one standard deviation of the responses for both detector copies at each composition to 10 exposures of each analyte vapor.
acetone, 0.30 for methanol, 0.30 for toluene, 0.35 for tetrahydrofuran, 0.35 for water, 0.50 for cyclohexanone, and 0.50 for hexane. Figure 2 presents the ∆Rmax/Rb responses of the plasticized PVPd detectors to methanol, acetone, and cyclohexanone. The maximum ∆Rmax/Rb values for each of these analytes occurred at different mass fractions of plasticizer. In addition, the maximal ∆Rmax/Rb values occurred at different plasticizer mass fractions for PVPd/DGD detectors than for PVAc/DGD detectors. For the plasticized PVPd detector, the maximum response to polar analytes such as methanol, acetonitrile, and water was obtained at Mf(DGD/PVPd) ) 0.05-0.15 (Figure 2a). In contrast, relatively Analytical Chemistry, Vol. 74, No. 6, March 15, 2002
1309
sis30,31,43 was performed to determine the pairwise resolution of various test analytes with the collection of plasticized polymeric vapor detectors. Normalization of the patterns produced by the response of the detector to each analyte exposure was performed according to eq 1, where the signal (Xi ) ∆Rmax/Rb) of the ith d
Xin ) Xi/
∑X
n
(1)
n)1
detector was divided by the sum of the Xi signals of all the detectors in the array. Normalization ensured that differences in analyte resolution arose from inherent differences in the response patterns produced by the array as opposed to arising only from differences the intensity of the responses. The mean response of the ith detector to each analyte was then computed by taking the average of the normalized ∆Rmax/Rb values that were recorded over multiple exposures for that detector to the analyte of interest. Each collection of mean normalized ∆Rmax/Rb values that describes the response of the detector array to an analyte of interest defined a vector in n-dimensional space, where n is the number of detectors in the array, for each analyte of interest. The distance between the mean responses to each analyte (the mean class response), and the standard deviation of given responses for each analyte with respect to a particular direction in n-dimensional space, are inherent properties of the data and were used to describe how well the responses of two analytes were separated statistically. The metric used to evaluate the separation between analyte clusters was the resolution factor (rf), which was calculated according to eq 2, where δ is the Euclidean distance between the
rf )
δ
xσ1
2
Figure 2. Relative differential resistance responses of PVPdcarbon black composites at varying mole fractions of DGD to (a) methanol, (b) acetone, and (c) cyclohexanone, each at 5% of its vapor pressure at room temperature. Error bars represent one standard deviation of the responses for both detector copies at each composition to 10 exposures of each analyte vapor.
nonpolar analytes, such as toluene, hexane, tetrahydrofuran, ethyl acetate, and cyclohexanone, produced a very small response for Mf(DGD/PVPd) < 0.10, but produced higher ∆Rmax/Rb values for detectors having Mf(DGD/PVPd) ) 0.40-0.95 (Figure 2c). Analytes of intermediate polarity, such as acetone, ethanol, methylene chloride, and chloroform, produced relatively small responses at low and high DGD mass fractions, with maximum ∆Rmax/Rb responses typically observed for detectors having Mf(DGD/PVPd) ) 0.40-0.50 (Figure 2b). B. Classification of Test Vapors Using Arrays of Plasticized Polymer Vapor Detectors. Fisher linear discriminant analy1310
Analytical Chemistry, Vol. 74, No. 6, March 15, 2002
(2)
+ σ22
two class means (i.e., the distance between the coordinates of the vectors that correspond to the mean normalized response values for each of the two analytes of interest) and σ1 and σ2 are the standard deviations of the two classes, with both δ and the standard deviations projected along the optimal discriminant vector between the two classes.30,31,43 Tables 1 and 2 present the pairwise resolution factors for the plasticized PVAc and PVPd arrays, respectively. The worstresolved analyte pair was ethanol/methanol for the plasticized PVAc array, and ethyl acetate/THF was worst-resolved for the plasticized PVPd array. Even in these cases, resolution factors under the experimental conditions used in this work were sufficient for robust classification of each of the test analytes on an array formed from different degrees of plasticization of a single polymer composite. C. Correlation of Response Rate and Glass Transition Temperatures of Plasticized Poly(vinyl acetate) Composite Detectors. Since DGD forms a compatible blend with PVAc, issues related to phase separation are minimized. Consequently, the behavior of the resistance versus time, temperature, extent of plasticization, and background analyte concentration was investigated in detail for this set of composites. Figure 3 shows the relative differential resistance responses, ∆R(t)/Rb, versus time, for a set of plasticized PVAc-carbon black composite
Table 1. Pairwise Resolution Factors for the Eleven Analytes Exposed to the PVAc/DGD Array
methylene chloride THF toluene hexane ethanol acetone ethyl acetate acetonitrile cyclohexanone chloroform
THF
toluene
hexane
ethanol
acetone
ethyl acetate
acetonitrile
cyclohexanone
chloroform
methanol
101
149 29.5
108 75.7 83.5
101 72.3 134 81.1
42.2 47.6 62.2 115 82.6
95.8 57.6 42.8 137 84.1 30.0
87.8 92.0 134 81.0 33.9 64.3 92.7
102 32.9 34.3 82.2 94.6 60.7 38.0 99.7
115 86.0 80.7 88.4 142 71.0 56.1 118 84.1
97.6 65.4 95.2 58.0 20.0 80.7 74.9 23.4 60.0 103
Table 2. Pairwise Resolution Factors for the Eleven Analytes Exposed to the PVPd/DGD Array
methylene chloride THF toluene hexane ethanol acetone ethyl acetate acetonitrile cyclohexanone chloroform
THF
toluene
hexane
ethanol
acetone
ethyl acetate
acetonitrile
cyclohexanone
chloroform
methanol
64.5
54.7 10.4
102 56.7 52.5
48.4 28.6 60.6 45.2
47.2 17.3 26.0 34.9 22.6
50.4 7.4 17.8 42.7 41.3 19.0
55.5 39.3 77.8 20.1 22.8 29.2 51.1
53.4 27.2 35.7 80.8 110 27.5 27.3 61.2
71.9 105 85.3 189 97.7 121 105 135 143
58.3 60.7 67.0 121 50.6 65.8 86.7 35.8 82.8 118
detectors exposed to methanol, acetone, and cyclohexanone at 21 ( 1 °C at P/Po ) 0.050 in a laboratory air background. The most polar analytes (methanol, acetonitrile, water) produced ∆R(t)/Rb responses that rapidly achieved steady state on all of the detectors that contained PVAc (Figure 3a). However, the other analytes showed slower ∆R(t)/Rb responses, with steady-state ∆R(t)/Rb values only being obtained within the 300-s exposure time on plasticized films having Mf(DGD/PVAc) g 0.45 (Figure 3b, c). The recovery times of the response to the baseline value after the analyte exposure had been terminated decreased as the degree of plasticization increased, although for some analyte/ polymer combinations, the recovery time was significantly slower than the rise time of the response (Figure 3). Similar behavior was observed for the plasticized PVPd-carbon black composites, in that the response and recovery times decreased at larger mass fractions of plasticizer in the composite. All plasticized PVPd composite detectors with Mf(DGD/PVPd) > 0.50 reached a steady-state ∆R(t)/Rb response within a 300-s exposure time for all of the test analytes investigated in this work. Figure 4 displays the DSC data for the plasticized PVAc composites with Mf(DGD/PVAc) e 0.50. The composite without added plasticizer had a Tg of 47 °C, whereas the composite with Mf(DGD/PVAc) ) 0.10 exhibited a very broad phase transition that began at -3 °C and ended at 44 °C. Composites with Mf(DGD/PVAc) ) 0.20, 0.30, and 0.40 displayed a minor transition near -20 °C, with the main transitions centered at ∼33, 16, and 8 °C, respectively. The transition for the Mf(DGD/PVAc) ) 0.50 composite was well below room temperature and was centered at ∼-11 °C. Thus, plasticized PVAc-carbon black composites that had Tg , 20 °C showed rapid response times at room temperature to all of the test analyte vapors studied in this work (Figure 3), whereas plasticized PVAc-carbon black composites with Tg > 20 °C showed ∆R(t)/Rb responses at room temperature that were
highly dependent on the polymer/analyte combination being evaluated. D. Temperature Dependence of the Response of Plasticized PVAc Detectors to Analyte Vapors. To evaluate further whether d(∆R(t)/Rb)/dt correlated with the Tg of the plasticized PVAc-carbon black composites, measurements of ∆R(t)/Rb for selected PVAc-carbon black composite detectors were performed at several temperatures. If the Tg of the composite were a significant factor in determining the rate at which a detector responded to a particular analyte vapor, then exposure to a vapor at temperatures on either side of the Tg should produce significant differences in d(∆R(t)/Rb)/dt. Figure 5 presents ∆R(t)/Rb responses for three PVAc composite detector films at four to five different temperatures when exposed to a relatively nonpolar analyte, chloroform. These temperatures were selected to span the Tg values revealed by the DSC measurements of these composites. Furthermore, chloroform was used because the data of Figure 3 show that plasticized PVAc composites with mass fractions of plasticizer less than 0.45 did not rapidly reach steady state when exposed to nonpolar analytes at room temperature. The PVAc-carbon black composite detector, with Tg ) 47 °C, showed a steady increase in the response rate as the temperature increased, but the detector did not produce steady-state ∆R/Rb values to this analyte in the 300-s exposure time for T e 45 °C. However, at 55 °C, a steady-state response was rapidly obtained. For the carbon black composites having Mf(DGD/PVAc) ) 0.30 (Tg ≈ 16 °C), steady-state responses were not observed in 300 s for T e 25 °C, while a steady-state response was rapidly obtained at T ) 35 °C. The responses for the detector with Mf(DGD/PVAc) ) 0.40 (Tg ) 8 °C) showed a similar trend, with steady-state responses being obtained in 15 °C. Analytical Chemistry, Vol. 74, No. 6, March 15, 2002
1311
Figure 4. Differential scanning calorimetry data from -100 to 150 °C at 10 °C min-1 for PVAc/DGD-carbon black composites. The mass fraction of DGD in each composite is shown to the right of each curve. The ordinate is in arbitrary units, and an arbitrary offset was added to facilitate comparison between the different data sets.
Figure 3. Effect of plasticizer mass fraction on the rate of the relative differential resistance response to analyte vapors. Exposures are (a) methanol, (b) acetone, and )c) cyclohexanone, each at 5% of its vapor pressure at room temperature, to the PVAc-DGD carbon black composite array. Values on each chart correspond to the mass fraction of DGD in the film. In (a), the response is rapid and reversible at 0% plasticizer because the diffusivity of highly polar analytes is quite large for this particular polymer-carbon black composite. Nonpolar analytes have lower diffusion coefficients, and hence slower response times, in this film and thus require plasticization of the polymer to obtain high diffusivities and rapid response times.
E. Correlation between Tg and the Temperature Dependence of the Baseline Resistance of Plasticized PVAcCarbon Black Vapor Detectors. Figure 6 displays the dc electrical resistance of unplasticized and plasticized Mf(DGD/ PVAc ) 0.30) PVAc-carbon black composites, respectively, as a function of temperature. Both samples showed a pronounced change in dRb/dT as the temperature was varied. Furthermore, 1312 Analytical Chemistry, Vol. 74, No. 6, March 15, 2002
the temperatures of these baseline resistance transitions correlated well with the abrupt changes observed in the value of d(∆R(t)/ Rb)/dt upon exposure to various test analytes. The transition in dRb/dT for the pure PVAc-carbon black composite detector occurred between 44 and 48 °C, whereas inspection of Figure 5a shows that this detector exhibited a slow d(∆R(t)/Rb)/dt response to chloroform at 45 °C and a rapid response at 55 °C. The plasticized PVAc detector having Mf(DGD/PVAc) ) 0.30 displayed a transition in dRb/dT between 18 and 30 °C, which is higher than the Tg obtained by DSC (16 °C) but is consistent with the vapor response data of Figure 5b. For these detectors, the discontinuity in dRb/dT therefore served as a good indicator of the temperature above which ∆R(t)/Rb values would rapidly reach steady state for all of the analytes evaluated in this work. F. Dependence of Detector Response on Background Humidity and Analyte Concentration. Figure 7 displays the responses of the plasticized PVAc detector with Mf(DGD/PVAc) ) 0.30 to 2 ppth of chloroform at 25 °C, as a function of the relative humidity in the background carrier gas. Each of the response curves has been normalized to its maximum differential resistance change during the given exposure period. Increasing the humidity, which increases the amount of water sorbed in the film, increased the response rate of the detector. In essence, the sorbed water is acting in much the same manner as a plasticizer. Figure 8 presents the response of this same detector, at 25 °C, to differing concentrations of chloroform at constant humidity. The response rate was clearly dependent on the concentration of analyte. At 22 ppth, the detector achieved a steady-state response within 180 s, but steady state was not reached for 2 ppth exposures to chloroform in 600-s exposure times. In contrast, when the background stream contained 8 ppth of chloroform, a steady-state response was achieved in ∼150 s when the detector was exposed to an additional 2 ppth of chloroform. Hence, the analyte itself is capable of acting as a plasticizer in the polymer with this combination of temperature, polymer composition, and analyte concentration. In contrast, a detector plasticized such that the Tg of the composite was much lower than room temperature, Mf(DGD/PVAc) ) 0.45, displayed rapid and reversible responses at 25 °C to chloroform. Additionally, such ∆Rmax/Rb responses
Figure 6. Resistance of (a) PVAc-carbon black and (b) 30% DGD/ PVAc-carbon black composite detectors as a function of temperature. The composites were heated from 5 to (a) 70 and (b) 60 °C at a rate of 1 °C min-1, held for 5 min at the upper temperature value, and then were cooled back to the initial temperature value at a rate of -1 °C min-1.
Figure 5. Responses of (a) PVAc-carbon black, (b) 30%DGD/ PVAc-carbon black, and (c) 40% DGD/PVAc-carbon black composite detectors to 2 ppth chloroform as a function of temperature. Each experiment consisted of a background flow of laboratory air for 180 s, a stream of analyte vapor for 300 s, and then a flow of laboratory air for the remainder of the exposure.
were linearly related to the chloroform concentration over the range 0.01e P/Poe 0.10. DISCUSSION A. Trends in Detector Response versus Mass Fraction of Plasticizer. The series of plasticized PVAc and PVPd composites investigated in this work spans two extremes: one in which the composite is predominantly plasticizer by mass and the other in which the composite contains no added plasticizer. In the former extreme, the relative partition coefficients of analyte vapors into the sorbent phase are largely dominated by interactions with the
Figure 7. Effect of humidity on the response rate of a 30% DGD/ PVAc-carbon black composite detector. The composite was exposed to 2 ppth of chloroform at 25 °C, with differing background humidity. The humidity in the background flow increased monotonically on the chart from low to high, with traces shown corresponding to 0, 3.6, 7.3, 15, 29, 44, and 58% relative humidity, respectively. The exposure to chloroform was arbitrarily initialized at 0 s and lasted for 180 s. The total flow rate was held constant throughout each experiment.
plasticizer, whereas in the latter extreme, the relative partition coefficients are dominated by the interactions with the polymer. The magnitudes and ratios of the ∆Rmax/Rb values observed for different analytes exposed to unplasticized PVAc-carbon black Analytical Chemistry, Vol. 74, No. 6, March 15, 2002
1313
Figure 8. Effect of chloroform concentration on the relative differential resistance response rate of the 30% DGD/PVAc-carbon black detector. Each exposure was conducted for 180 s, beginning at t ) 0, at a total flow rate of 5 L min-1, at 25 °C. The analyte concentration was increased monotonically from low to high values, with the traces corresponding to 2, 6, 10, 14, 18, and 22 ppth chloroform, respectively.
composites were clearly very different from those obtained when these same analytes were exposed to highly plasticized (e.g., Mf(DGD/PVAc) g 0.50) PVAc-carbon black detectors. Even more pronounced differences in relative ∆Rmax/Rb values were observed for exposure of the test analytes to unplasticized PVPdcarbon black composites relative to exposure to the highly plasticized (Mf(DGD/PVAc) g 0.50) PVPd-carbon black composites. DGD forms a compatible blend with PVAc,44 and the polymer and plasticizer are therefore similar in polarity. In contrast, PVPd is significantly more polar than PVAc. The composites obtained when PVPd is mixed with the relatively less polar DGD plasticizer should therefore differ significantly in their polarity as the mass fraction of DGD is varied; consequently, the polarity (and other chemical properties) of the analyte will strongly influence which plasticized DGD-PVPd carbon black composite will provide the maximum ∆Rmax/Rb value when exposed to that analyte. The variation in maximum ∆Rmax/Rb values for DGDplasticized PVPd-carbon black composite detectors with different analytes is similar to the trend in ∆Rmax/Rb response that has been observed previously when diblock copolymers, in which the two blocks were of very different polarity, were used in carbon blackpolymer composite vapor detectors.41 B. Time and Temperature Dependences of Detector Response Rate. A major effect of plasticization is to lower the glass transition temperature of the plasticized polymer relative to that of the unplasticized polymer, when the two form a compatible blend.44-46 At constant temperature, this produces an increase in the diffusion coefficient of analytes through the material, which produces more rapid ∆R/Rb responses from carbon blackpolymer composite vapor detectors exposed to such analytes. Above Tg, the material ought to be in the rubbery phase, which should enable relatively rapid diffusion of permeants,44-46 whereas (44) Brandup, J.; Immergut, E. H.; Grulke, E. A. Polymer Handbook; WileyInterscience: New York, 1999. (45) Allen, G.; Bevington, J. C. Comprehensive Polymer Science: The Synthesis, Characterization, Reactions & Applications of Polymers; Pergamon Press: New York, 1989. (46) Rodgers, C. E. In Polymer Permeability; Comyn, J., Ed.; Elsevier: Amsterdam, 1985.
1314 Analytical Chemistry, Vol. 74, No. 6, March 15, 2002
below Tg, diffusion of permeant should be impeded due to the presence of significant glassy regions in the film. These expectations are clearly reflected in the dependence of ∆R(t)/Rb on time and detector temperature, Tdet, for the plasticized PVAc-carbon black composite films evaluated in this work. In general, the DSC measurements of Tg correlated well with the response of ∆R(t)/Rb versus time. When Tdet > Tg, the responses rapidly reached steady state for all analyte/polymer combinations in our test set. Under these conditions, the trends in ∆Rmax/Rb response values with mass fraction of plasticizer predominantly reflect changes in the polymer/vapor partition coefficients for the analyte of interest. When the detectors were operated at a temperature, Tdet, that was less than Tg, some analytes showed rapid responses whereas other analytes did not produce steady-state ∆R/Rb signals during the 300-s exposure period. Addition of a plasticizer to a polymer is well known to affect both the diffusion coefficient (D) and the polymer/vapor partition coefficient (K) of a permeant.27,44-46 Variation in either D or K can be useful in producing distinctive analyte response patterns on an array of sorption-based vapor detectors and can thereby enable differentiation between analytes through subsequent signal processing algorithms. The ∆R/Rb data are useful in either situation for distinguishing between analytes. For the PVAc-carbon black composite detectors with Mf(DGD/ PVAc) ) 0.30, a rapid ∆R/Rb response to nonpolar analytes was not observed at 25 °C, which is above the Tg for this material as determined by DSC. However, the glass transition was broad and occurred over a range of ∼20 °C. If the composite film possesses both glassy and rubbery regions, the rate at which the detector reaches a steady-state response will likely be limited by the glassy regions. This may account for the need to increase the temperature by an additional 10 °C to produce steady-state responses for this particular analyte/detector combination. The Tg value of the nonplasticized PVAc-carbon black composite was greater than that of the pure polymer. The Tg of the pure polymer is 40 °C, whereas that of the carbon black composite was 47 °C. Prior work on bulk polymer composites has shown that fillers such as carbon black restrict the local mobility of the polymer, leading to an increase in Tg.45 This expectation is consistent with behavior of the thin-film carbon black composites used herein, both in DSC measurements and in their analyte ∆R(t)/Rb response properties. The use of an electrically conductive filler material such as carbon black facilitates a determination of the glass transition temperature of the polymer composite through monitoring the resistivity of the composite as a function of temperature. The discontinuity observed in the temperature dependence of the baseline resistivity of the plasticized PVAc- carbon black composites correlated well with Tg (Figures 4 and 6), as might be expected if the phase change produced a change in the connectivity of the percolative network that produces electrical conduction through the conductive filler particles. When this discontinuity in dRb/dt can be measured, it provides a convenient method of establishing the temperature above which the detector should be operated to obtain high diffusivities for most low molecular weight analytes. C. Implications for Use of Plasticized Polymeric Composites in Arrays of Vapor Detectors. To obtain rapid steady-
state responses to a diverse collection of analytes in a given detector film, the detector temperature should clearly be above the value of Tg for the composite film. Optimally, however, Tdet should not be far above Tg, because the polymer/vapor partition coefficient for a given analyte decreases as K(T) ) Ko exp(-Eact/ kT), where k is Boltzmann’s constant and Eact is the activation energy for the equilibration process.27,45 A decrease in K will produce a decrease in ∆Rmax/Rb and therefore will produce a decreased sensitivity of the detector toward the analyte of concern. A desirable operating condition would therefore involve maintaining each detector at a temperature slightly above its characteristic Tg value. Of course, reaching the optimal operating temperature when Tg is much lower than room temperature will require the use of a preconcentrator or similar approach to eliminate humidity from the analyte vapor stream to avoid condensation of liquid water onto the operational detector surfaces. Although polymers can lose plasticizers over time, thereby changing the properties of plasticized composite vapor detectors, the use of relatively nonvolatile plasticizers as described herein can provide reproducible vapor detectors for time periods that are of interest for many vapor detection applications. For example, little variation in response was observed over the 36-h period during which the responses of all the detectors discussed herein were exposed to the various different analytes. Furthermore, a detector having Mf(DGD/PVAc) ) 0.45 showed essentially the same ∆Rmax/Rb response to chloroform at P/Po ) 0.050 over a two-month period during which the detector had been exposed to ambient air as well as other types of vapor exposures between the exposures to chloroform. Another issue is related to the molecular weight of the analytes to be analyzed. Analytes with high molecular weights will have higher partition coefficients, but generally will have lower diffusion coefficients into a given polymer film than will volatile low molecular weight analytes.44,45,47 If a rapid response to steady state is desired for a given detector film, the detector temperature should generally be high enough relative to Tg to produce an acceptable diffusion rate of such analytes though the composite of interest. In some instances, classification can be potentially be improved through the use of the kinetic data as well as the steadystate response data; however, the kinetic data will often be sensitive to the details of the sampling fluidics that deliver analyte to the detectors48 as well as the identity and concentration of the background components of a complex analyte mixture. When the kinetic data are linear with analyte concentration and when the
kinetic response to mixtures of analytes is linearly related to the mole fraction of the components in a mixture, and when the signals vary very slowly with time or the exposure time to the analyte is well-controlled, robust analyte classification should be attainable as has been shown for PVAc and PVPd detectors in prior studies of chemiresistive vapor detector responses to these types of analytes.30 The approach described herein is particularly attractive when steady-state detector responses, which reflect primarily the variation in equilibrium partition coefficients between the analyte and the polymer composite phases, are desired as the descriptors for test vapors to be analyzed using chemometric data analysis methods.
(47) Doleman, B. J.; Severin, E. J.; Lewis, N. S. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 5442-5447. (48) Briglin, S. M.; Burl, M. C.; Freund, M. S.; Lewis, N. S.; Matzger, A.; Ortiz, D. N.; Tokumaru, P. Proc. SPIE 2000, 4038.
Received for review October 2, 2001. Accepted December 6, 2001.
CONCLUSIONS Variation in the mass fraction of plasticizer added to a carbon black-polymer composite chemiresistor provides a straightforward, convenient method of producing detector arrays having systematic variations in polymer polarity and analyte permeation rates. Differences in vapor sorption and permeability that affect the dc electrical resistance response of these compositionally different detectors allow robust identification and classification of various test analytes using standard chemometric methods. Plasticized polymer composites with glass transition temperatures significantly below the operating temperature of the detector showed rapid responses to all of the test analyte vapors studied in this work, so a desirable operating condition is achieved when each detector in the array is maintained at a temperature slightly above its characteristic Tg value. Carbon black-polymer composites exhibit a discontinuity in the temperature dependence of their resistance, providing a simple method for determining the Tg of the composite and for determining the temperature and plasticizer mass fraction above which rapid resistance responses can be obtained to organic vapors. The plasticization approach provides a method for achieving rapid detector response times as well as for producing a large number of chemically different vapor detectors from a limited number of initial chemical feedstocks. ACKNOWLEDGMENT We acknowledge the ARO, DOE, and NIH for support of this research. SUPPORTING INFORMATION AVAILABLE Additional data as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.
AC011054+
Analytical Chemistry, Vol. 74, No. 6, March 15, 2002
1315
