Anal. Chem. 2010, 82, 66–68
Letters to Analytical Chemistry Dual-Channel Sensing of Volatile Organic Compounds with Semiconducting Nanoparticles Da Liu,† Mingyang Liu,†,‡ Guohong Liu,§ Sichun Zhang,*,† Yayan Wu,† and Xinrong Zhang† Department of Chemistry, Key Laboratory for Atomic and Molecular Nanosciences of the Education Ministry, Tsinghua University, Beijing 100084, P. R. China, Technical Centre, Liaoning Entry-Exit Inspection and Quarantine Bureau, Dalian Liaoning 116001, P.R. China, and Analytical Centre, Institute of Chemical Defence, Beijing 102205, P.R. China Extracting multidimensional information from an individual transducer simultaneously is a promising alternative sensing strategy to traditional sensors. Here, we proposed a novel dual channel sensing method with simultaneously recording conductivity change of sensing material and chemiluminescence emission during catalytic oxidation of volatile organic compounds on tin oxide nanoparticles. The orthogonal and complementary electrical and optical signals have been obtained for each compound, which have been applied to discriminate 20 volatile organic compounds using hierarchical cluster analysis (HCA). Unknown samples from three groups at concentrations of 0.2%, 0.6%, and 1.0% have been successfully classified using linear discriminant analysis (LDA) with accuracies of 98.3%, 96.7%, and 98.3%, respectively. This dual channel sensing mode is a complement of semiconducting type gas sensors and quite promising for the development of chemical sensor arrays with multimode transducing principles. Sensor devices with more than one transduction principle in chemical sensing allows the target analytes to be identified and quantified even in the presence of unknown interfering agents.1 There are mainly two ways to accomplish multitransduction sensor devices: one way is to combine several different transducers, such as mass-sensitive, capacitive, and calorimetric, on a single-chip chemical microsensor system;2 the other way is to extract more than one signal from single sensing material, which should be as orthogonal or complementary as possible to increase the ability to detect and discriminate the analytes.3 For the latter case, a multimode sensing system based on only one sensor element can be considered as a virtual sensor array extracting multidimensional * To whom correspondence should be addressed. E-mail: sczhang@ mail.tsinghua.edu.cn. Phone: (+86) 10-6277-6888. Fax: (+86) 10-6278-2485. † Tsinghua University. ‡ Liaoning Entry-Exit Inspection and Quarantine Bureau. § Institute of Chemical Defence. (1) Hierlemann, A. Chem. Rev. 2008, 108, 563–613. (2) (a) Brand, O. Proc. IEEE 2006, 94, 1160–1176. (b) James, D.; Scott, M. S.; Ali, Z.; O’Hare, T. W. Microchim. Acta 2005, 149, 1–17. (c) Joo, S.; Brown, R. B. Chem. Rev. 2008, 108, 638–651. (d) Mitrovics, J.; Ulmer, H.; Weimar, U.; Go ¨pel, W. Acc. Chem. Res. 1998, 31, 307–315. (3) De Silva, A. P. Nature 2007, 445, 718–719.
66
Analytical Chemistry, Vol. 82, No. 1, January 1, 2010
information through modulating parameters (e.g., temperature)4 or detecting various parameters (mass-sensitive, capacitive, electric, optic, and calorimetric) simultaneously.5 Thus far, it is still a challenge to extract more complementary information from a single sensing material utilizing different transduction principles for gas-phase molecule sensing. The conductivity-change of metal oxide semiconductors has been studied extensively, which has been matured as the transducing mode for detection of gas-phase molecules.6 Compared with conductivity responses, chemiluminescence (CL) on catalytic nanomaterials has been studied over the last several years for developing sensors/sensor arrays for the detection of volatile organic compounds (VOCs) and characterizing catalytic activities of nanocatalysts.7,8 However, simultaneous study of the two kinds of electron-related phenomena during catalytic reactions has not yet been reported before. Here we found that the electrical and optical signals obtained during catalytic reaction of VOCs on semiconducting nanomaterials are orthogonal and complementary, which can be used for the discrimination of the analytes. We implemented a system for simultaneously detecting the changes of conductivity and emission of CL with nanosized SnO2 as an example (because it is one of the most frequently used sensing materials for gas sensors) (Figure 1 and Figures S1-S3 in the Supporting Information). The changes of conductivity and CL intensities of four compounds including n-propanol, sec-butanol, butanone, and acetic acid were investigated in the first phase. Because of the different structures of these molecules and their unique physical and chemical properties, which leads to different adsorption properties and reactivities, the observed responses of the CL intensity and changes in conductivity for different chemical (4) Meier, D. C.; Taylor, C. J.; Cavicchi, R. E.; White, E.; Semancik, S.; Ellzy, M. W.; Sumpter, K. B. IEEE Sens. J. 2005, 5, 712. (5) Schmittel, M.; Lin, H. W. Angew. Chem., Int. Ed. 2007, 46, 893–896. (6) Kolmakov, A.; Moskovits, M. Annu. Rev. Mater. Res. 2004, 34, 151–180. (7) (a) Wang, X.; Na, N.; Zhang, S. C.; Wu, Y. Y.; Zhang, X. R. J. Am. Chem. Soc. 2007, 129, 6062–6063. (b) Na, N.; Zhang, S. C.; Wang, X.; Zhang, X. R. Anal. Chem. 2009, 81, 2092–2097. (c) Na, N.; Zhang, S. C.; Wang, S.; Zhang, X. R. J. Am. Chem. Soc. 2006, 128, 14420–14421. (d) Wu, Y. Y.; Na, N.; Zhang, S. C.; Wang, X.; Liu, D.; Zhang, X. R. Anal. Chem. 2009, 81, 961–966. (8) Nakagawa, M.; Yamashita, N. Springer Ser. Chem. Sens. Biosens. 2005, 3, 93–132. 10.1021/ac902422s 2010 American Chemical Society Published on Web 12/11/2009
Figure 1. Illustration showing the conception of simultaneous detection of CL and the conductivity responses during the nanocatalysis of the analyte on nanosized SnO2. Figure 4. Dendrogram of the dual-channel sensing of CL and conductivity to 20 common VOCs at 0.6% showing the relative relationship among gas samples. Averages of six trials were used for each analyte. The sorting of the analyte was classified due to the proximity of the characteristics of optical-electrical signatures.
Figure 2. CL and conductivity responses of 0.6% n-propanol, secbutanol, butanone, and acetic acid detected simultaneously.
Figure 3. Schematic of the two electron-related processes of catalytic oxidation on nanosized SnO2.
molecules are diverse and differentiable (Figure 2). For example, the electrical responses of n-propanol and sec-butanol vapor are similar, but their optical responses are different, while the optical responses of butanone and acetic acid vapor are nearly the same, but their electrical responses are distinct. The irrelevant or quasiindependent response of electrical and optical signatures may be caused by different electron processes during catalytic reactions, as shown descriptively in Figure 3. The electrons of oxygen transfer back to the SnO2 crystal after reacting with analytes, which increase the amount of the electrons and reduce the intergranular potential barrier to increase the conductivity. Simultaneously, the intermediate or product molecules are at the excited state as soon as they were formed during the catalytic oxidation, which decay from this excited state to the ground state with light emission.8,9 Because the electrical response mainly depends on the interaction between the reactant molecules and the nanocatalyst while the optical response mainly depends on the reaction activity and quantum yield of excited state intermediates or product molecules, the two signals are independent. (9) Breysse, M.; Claudel, B.; Faure, L.; Guenin, M.; Williams, R. J. J.; Wolkenstein, T. J. Catal. 1976, 45, 137–144.
The irrelevant or quasi-independent response of electrical and optical signatures has been used to discriminate 20 compounds, including alkanes, alcohols, ketones, aldehydes, acids, esters, ethers, amines, acetonitrile, and toluene in the second phase (Table S1 in the Supporting Information). To examine the multivariate distances between the responses to 20 chemicals in this two-dimensional sensing space, a hierarchical cluster analysis (HCA) was performed to generate dendrograms with Euclidean distance (Figure 4).10 For this HCA, the optical and electrical responses were standardized logarithmically so as to weigh both response channels equally. The analyte was separated into each class tolerably in accord with the integrated effect of carbon numbers, polarities, and molecular structures. In contrast to sensing responses of either optical or electrical individually (Figure S4 in the Supporting Information), the dual-channel sensing system is able to cluster the analytes more logically as well as discrimate similar compounds, even for structural isomers without misclassification. Also, PCA10 was performed, and the analytes were clustered correctly (Figures S5 and S6 in the Supporting Information). To further demonstrate the discrimination ability of the dualchannel sensing system for samples at different concentrations, 20 compounds at 3 concentrations were investigated. The CLconductivity response patterns of 120 samples (20 chemical species × 6 replicates) at 0.2%, 0.6%, and 1.0%, respectively, as well as 60 unknown samples (20 chemical species × 3 replicates) at concentrations of 0.2%, 0.6%, and 1.0%, respectively, were subjected to linear discriminant analysis (LDA) separately (Figure 5).10 After logarithmic normalization, LDA converted the patterns of the three training matrices to canonical scores. The canonical patterns of 120 samples were clustered into 20 different groups. According to the classification training matrices in the LDA results, 99.2%, 100%, and 100% accuracies were given, respectively. Of 60 samples in each testing matrix, 59, 58, and 59 unknown samples were correctly classified resulting in accuracies of 98.3%, 96.7%, and 98.3%, respectively. Moreover, to simulate the real world situation, detection and discrimination of chemicals with various concentra(10) (a) Scott, M. S.; James, D.; Ali, Z. Microchim. Acta 2007, 156, 183–207. (b) Jurs, P. C.; Bakken, G. A.; McClelland, H. E. Chem. Rev. 2000, 100, 2649–2678.
Analytical Chemistry, Vol. 82, No. 1, January 1, 2010
67
Figure 5. CL and conductivity responses of the nanosized SnO2 based dual-channel sensing against 20 chemicals (see Table S1 in the Supporting Information) at (a) 0.2%, (b) 0.6%, and (c) 1.0%. Canonical score plot by LDA for discrimination of 20 kinds of compounds at (d) 0.2%, (e) 0.6%, and (f) 1.0%.
tions is required. Of 360 samples in a combined group of the above three training sets, 343 were correctly classified using LDA for the training matrix (Figure S7 in the Supporting Information), affording a discrimination accuracy of 95.3%, and of 180 unknown samples in the combined blind group for the testing matrix, 166 were correctly classified affording an accuracy of 92.2%. Detailed discrimination results and statistical data are presented in the Supporting Information(Tables S2-S8).
68
Analytical Chemistry, Vol. 82, No. 1, January 1, 2010
In conclusion, the simultaneous detection of CL and the conductivity change on SnO2 provides quasi-independent and complementary optical-electrical signatures for each compound, which have been successfully used for the discrimination of 20 VOCs. The optical-electrical mode sensing strategy is expected to develop a hybrid sensor array for recognition and discrimination of complicated samples. The advantage of this proposed sensing strategy to sensor arrays is that the differentiability could be improved with the same number of sensor units or the number of sensing units could be reduced while keeping the same differentiability, compared to sensor arrays with an individual tranducing mode. We regard this nanomaterials-based dual channel sensing as our first step in the development of a multiplex, multiple-channel sensing system for improved discrimination and recognition power. ACKNOWLEDGMENT This work is supported by grants from the MOST (Grant Numbers 2008IM040600 and 2009AA03Z321) and NSFC (Grant Number 20875053). SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review November 27, 2009. AC902422S
October
26,
2009.
Accepted
