Sensors and Actuators B 121 (2007) 365–371 Analysis of transient response of single quartz crystal nanobalance for determination of volatile organic compounds A. Mirmohseni a,∗ , H. Abdollahi b,1 , K. Rostamizadeh a,2 a Polymer Research Technology Laboratory, Department of Applied Chemistry, Faculty of Chemistry, University of Tabriz, Iran b Department of Chemistry, Institute for Advanced Studies in Basic Sciences (IASBS), Zanjan, Iran Received 15 January 2006; accepted 31 March 2006 Available online 23 May 2006 Abstract Quartz crystal nanobalance (QCN) sensors are considered as powerful mass sensitive sensors. In this study, a single quartz crystal nanobalance modified with polystyrene has been employed to determine benzene, toluene, ethylbenzene and xylene (BTEX compounds). The frequency shift of QCN sensor was found to be linear against the BTEX compound concentrations in the range about 0–27 mg l−1 . The correlation coefficients for benzene, toluene, ethylbenzene, and xylene were 0.9917, 0.9904, 0.9904 and 0.9925, respectively. The principal component analysis was also utilized to process the transient response data of the single quartz crystal, considering to the different adsorption–desorption dynamics of BTEX compounds. Using principal component analysis, it was found that over 75% of the data variance could still be explained by use of three principal components (PC1, PC2 and PC3). The score plot of the three first PCs data were used to classify and identify organic vapors. The results showed that the polystyrene modified QCN had favorable identification and quantification performances for the BTEX compounds. © 2006 Elsevier B.V. All rights reserved. Keywords: Quartz crystal nanobalance (QCN); BTEX compounds; Polystyrene; Principal component analysis 1. Introduction (m): Polymers, whose range of useful physical and chemical properties promise many potential applications, have received considerable attention in a variety of fields, including chemical sensors [1], membrane separation technology [2], solid phase extraction techniques [3]. In the last few years, there has been an increasing amount of attention paid to the application of polymer-coated quartz crystal nanobalance (QCN) sensors [4–6]. The QCN comprises a thin vibrating AT-cut quartz wafer sandwiched between two metal excitation electrodes. When a small amount of mass is adsorbed on the quartz electrode surface, the frequency of the quartz is changed. Sauerbrey’s equation is used to relate the frequency change (F) to the mass loading F = −2.26 × 10−6 F02 ∗ Corresponding author. Tel.: +98 411 3393171; fax: +98 411 3340191. E-mail addresses: mirmohseni@tabrizu.ac.ir, mirmohseni@uow.edu.au (A. Mirmohseni), abd@iasbs.ac.ir (H. Abdollahi), rostamizadeh@tabrizu.ac.ir (K. Rostamizadeh). 1 Tel.: +98 241 4152064; fax: +98 241 4248925. 2 Tel.: +98 411 3393151; fax: +98 411 3340191. 0925-4005/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.snb.2006.03.054  m A  (1) where F0 is the fundamental resonant frequency of the unloaded quartz crystal and A is the electrode area. The theoretical detection limit of QCN is reported to be as small as 10−12 g. When the surface of a quartz crystal electrode is coated by a polymer capable of interaction with the environment of interest, a sensor sensitive to the component can be constructed. The performance characteristics of the QCN sensor will depend on the chemical nature and physical properties of the polymeric coating. Polymer-coated quartz crystal has been studied as a sensor for volatile organic compounds [7–9]. For the development of a chemical sensor selective, sensitive and stable sensing materials must be employed. However, the construction of an entirely selective chemical sensor is almost impossible, since along with the analyte other compounds usually interfere. As a matter of fact, the major drawback of the polymer modified QCN sensors is a lack of selectivity since there is no discrimination between the sources of the mass changes. Using pattern recognition is 366 A. Mirmohseni et al. / Sensors and Actuators B 121 (2007) 365–371 2.3. Procedures Fig. 1. Schematic diagram of the experimental set-up for QCN sensor: (1) PC, (2) frequency counter, (3) electronic circuit, (4) quartz crystal electrode, (5) syringe, and (6) desorbing gas. a very promising technology in the data processing of the signals of a non-selective sensor. In this regard, the characteristic information contents of the signal have to be recorded. Different chemometric tools could be applied for this purpose. Fuzzy clustering [10], partial least squares (PLS) [11,12], principal component analysis (PCA) [13,14] and artificial neural network (ANN) [14,15] have been widely used for the classification of analytes. In recent years, considerable interest has risen on the use of quartz crystal as a sensor in conjunction with associated pattern recognition analysis in discriminating various kinds of compounds [10–15]. In this study, construction of a QCN sensor having good character to analysis volatile organic compounds in atmospheric media was investigated. Benzene, toluene, ethylbenzene and xylene (BTEX compounds) are classified as the most important volatile organic compounds that cause environmental pollution. We developed a method of thin polystyrene-coated QCN sensor in conjunction with PCA for determination of BTEX compounds. To perform PCA and identify BTEX compounds, the parameters extracted from the transient responses of sensor were used. 2. Experimental All measurements were carried out in a glass cell with an internal volume of 28.5 ml. Hamilton microliter syringes (Hamilton Bonaduz AG, Switzerland) were used for analyte injections and dispensing polymer solution cast on the gold electrode of the quartz crystal. A certain amounts of analytes were injected to the cell, and the frequency shift of crystal was measured every 2 s. The concentration of injected analyte in the cell was calculated in mg l−1 considering its density, purity percent and volume [16]. A flow of dried air was purged through the cell to desorb analyte and recover the electrode. All measurements were carried out at room temperature (23 ◦ C). 2.4. Coating preparation The crystal was coated with polystyrene via the casting method. The gold-coated quartz electrode was coated using a 4 ␮l solution of polystyrene in chloroform (various concentrations) and the thin film of polymer was formed by solvent evaporation. The film thickness varied corresponding to the polymer solution concentration. In order to regenerate the electrode, the polymeric coating was dissolved in chloroform and dried by acetone. 2.5. Data processing The principal component analysis (PCA) method is widely used to classify compounds and mixtures. PCA contains an orthogonalization procedure such as singular-value decomposition (SVD) that decomposes the primary data matrix by projecting the multi-dimensional dataset onto a new coordinates base formed by the orthogonal directions with data maximum variance. The data matrix consists of a number of experiments, each include a number of variables. The eigenvectors of the data matrix are called principal components (PCs). Generally, the kth principal component, PCk , is a linear combination of the n response vectors (Xn,j ) for the analyte under study: PCk = n  an,k Xn,j (2) n=1 2.1. Reagents and materials Benzene, toluene, ethylbenzene and xylene (BTEX) were all purchased from Merck (Germany) and used without any further purification. Polystyrene was supplied by Tabriz Petrochemical Co. 2.2. Instrumentation Ten megahertz AT-cut quartz crystal with gold electrodes on both sides was purchased from International Crystal Manufacturer (ICM, Oklahoma, USA). A homemade QCN apparatus was used for frequency measurements as shown in Fig. 1. A detail of measuring cell and device was described in our previous report [8]. where n is the number of the variables, j indicates different samples and the coefficients (an,k ) are called loading. The magnitude of each eigenvector is expressed by its own eigenvalue, which gives a measure of the variance related to that principal component. The PCs are ordered so that the first few PCs retain most of the variation present in all of the original data. By elimination of the less important eigenvectors, it is possible to achieve fewer vectors without any considerable information loss. So, during data processing, the results are transformed in a plane or in a space of the first two or three eigenvectors. The coordinates of the data in the new base are called their score. The scores plot is usually used for the classification of the data clusters. In this study, the input data of the primary matrix are the autoscaled data extracted from transient response of single QCN sensor. The score plots were used to classify and identify organic vapors. A. Mirmohseni et al. / Sensors and Actuators B 121 (2007) 365–371 367 Fig. 2. The effect of polymer concentration of casting solution (polymer thickness) on frequency changes of polystyrene modified quartz crystal electrode exposed to 15 mg l−1 toluene. Fig. 3. The effect of polymer concentration of casting solution (polymer thickness) on response time of polystyrene modified quartz crystal electrode exposed to 15 mg l−1 toluene. PCA calculations used a singular-value decomposition algorithm and were performed with the MATLAB software version 6.5. The operating system was Microsoft Windows 2000XP. of the crystal decreased upon exposure to the analyte according to Eq. (1). A plateau is reached after a period of time, indicating maximum adsorption onto the QCN electrode surface. After turning back to the purge cycle, the frequency goes back to a value approximating the basic line. k is considered as a value for reversibility of the sensor response [17]: 3. Results and discussion 3.1. Development and amount of coating material k= To obtain the optimum thickness, the amount of polystyrene that has to be coated on the electrode of QCN was investigated. The coating thickness depends on the polymer concentration in the casting solution. The crystal was coated with 4 ␮l of diverse concentrations of polystyrene solutions. The QCN coated with different amounts of polystyrene was exposed to the constant concentration of toluene. Fig. 2 is a frequency shift response of the crystal coated with different amounts of polystyrene upon exposure to 15 mg l−1 toluene. It was found that as the coating thickness increased the magnitude of the response (frequency shift) was also increased. On the other hand, response time obtained for crystals having different thickness of polystyrene showed that the QCN crystal with a thicker coating exhibit a high response time for the constant concentration of toluene as well (Fig. 3). The shortest response time obtained was for crystal coated by 0.1% (w/v) polystyrene solution. Besides, too much coating would disturb the resonance stability and occasionally leads to the failure in the quartz crystal oscillation. Based on these observations, it can be concluded that the QCN should be modified using 4 ␮l of 0.3% (w/v) polystyrene solution. So it had significant response and its response time was less than 12 min. The mass of the coating caused a frequency shift of about 6 kHz. t90,desorp t90,sorp (3) where t90 is the frequency value at 90% after starting the adsorption or desorption process. For an ideal fast reversible response, k must be 1. Table 1 shows the reversibility of polystyrene-coated QCN sensor at two different concentrations of analytes. It can be 3.2. Sensor response and sensitivity A typical response for a crystal coated with a solution of 0.3% (w/v) polystyrene/chloroform is shown in Fig. 4. The frequency Fig. 4. Typical frequency changes of the polystyrene modified quartz crystal electrode exposed to 27 mg l−1 toluene in air: (a) sample injection; (b) sample desorption by purging air through the cell. 368 A. Mirmohseni et al. / Sensors and Actuators B 121 (2007) 365–371 Table 1 Reversibility of polystyrene-coated QCN towards analytes (c = 15 mg l−1 ) k k (c = 27 mg l−1 ) Benzene Toluene Ethylbenzene Xylene 0.39 4.75 1.09 6.62 2.13 5.83 0.91 7.01 concluded that the polystyrene shows good reversibility, especially at low concentrations, with all analytes. This implies that the polystyrene-coated QCN crystal sensor can be repeatedly used for the detection of examined organic vapors. A response obtained by a sensor should be linear against concentration of species of interest. Fig. 5 shows the frequency shift (F) as a function of time for different concentrations of xylene. It is obvious from graph that F increases monotonically with increasing gas concentration. This result strongly suggests that it is possible to prepare the gas sensor using polystyrenecoated QCN. In this regard, the calibration curve was constructed by plotting the frequency shifts against the concentration of analytes. The frequency shifts used for plotting the calibration curves corresponded to the frequency shift of the crystal after 10 min of exposure. Plots of the frequency shifts against various concentrations of BTEX compounds are shown in Fig. 6. As shown there was a good linear relationship in the range about 0–27 mg l−1 . The values of 0.9917, 0.9904, 0.9904 and 0.9925 were calculated for correlation coefficients of benzene, toluene, ethylbenzene and xylene, respectively. The lower limit of detection (LLD) is the lowest concentration of analyte that can be detected. LLD of the analytes for the designed QCN sensor was investigated. The study showed that the base line noise of applied system was 1 Hz. The frequency change limit necessary to detect an analyte is set at a signal-tonoise ratio of 3, three times the base line noise. The LLD of the Fig. 5. Frequency shift of polystyrene modified quartz crystal electrode as a function of time exposed to various concentrations of xylene: (a) 3.0 mg l−1 , (b) 6.0 mg l−1 , (c) 9.0 mg l−1 , (d) 12.0 mg l−1 , (e) 15.0 mg l−1 , (f) 18.0 mg l−1 , (g) 21.0 mg l−1 , (h) 24.0 mg l−1 , and (i) 27.0 mg l−1 . Fig. 6. Calibration curves for determination of benzene (), toluene (), ethylbenzene () and xylene (△) using polystyrene-coated quartz crystal electrode. QCN for BTEX was calculated and the value of 2.74, 3.01, 3.02 and 2.66 mg l−1 were obtained, respectively. The sensitivity of sensor is expressed as the slope of the calibration curve. The bar graph (Fig. 7) illustrates the sensivity of the QCN sensors towards examined analytes. As expected the sensors showed different sensitivities to the BTEX gases. The maximum and minimum sensivity of sensor was towards xylene and benzene, respectively. To investigate the repeatability of the results, the sensor coated with polystyrene was alternatively exposed five times to the presence of BTEX compounds at the concentration of 12.0 mg l−1 . The frequency-shift responses of the sensor for toluene are shown in Fig. 8. The relative standard deviations (R.S.D.s) for BTEX compounds were 3.03%, 3.07%, 1.89% and 2.95%, respectively, indicating that the Fig. 7. Sensivity of polystyrene-coated QCN towards BTEX compounds. 369 A. Mirmohseni et al. / Sensors and Actuators B 121 (2007) 365–371 Fig. 8. Reproducibility test for frequency changes of the polystyrene modified quartz crystal electrode exposed to 12 mg l−1 toluene: (a) toluene injection; (b) toluene desorption by purging air through the cell. Fig. 9. The adsorption–desorption frequency responses of polystyrene-coated QCN sensor exposed to the concentration of 21 mg l−1 of BTEX compounds: (a) benzene, (b) toluene, (c) ethylbenzene, and (d) xylene. sensor showed a good repeatability for the detection of all compounds. 3.3. Identification of gases using principal component analysis It is necessary the gas sensor also possesses an acceptable selectivity towards BTEX compounds. To establish the ability of the polystyrene-coated QCN to discriminate between the BTEX gases a principal component analysis was carried out. The adsorption–desorption dynamics of the polystyrene-coated QCN exposed to the BTEX compounds at the concentration of 21 mg l−1 are shown in Fig. 9. As it can be seen, however, the outline of the adsorption and desorption process for all analytes are the same, but there are some differences in the shape of sensor response exposed to various organic vapors. The difference in the response shape of the sensor can be related to the type of analytes and their different affinity for adsorption and desorption of polystyrene coating. So attention to the difference shape of each response can be useful in order to distinguish gases using PCA [18]. In this regard, fourteen parameters extracted from transient responses of sensor which could reflect the characteristics of the components dynamic behavior were considered as input matrix for PCA analysis. The parameters used as primary data matrix for PCA are shown in Table 2. To extract some of these parameters such as integral at different times the response curves of crystal were modeled using the polynomial of eighth degree, according to the following equation: F (x) = c1 x8 + c2 x7 + c3 x6 + c4 x5 + c5 x4 + c6 x3 + c7 x2 + c8 x + c 9 (4) All compounds were exposed to the cell for three different concentrations including 12.0, 15.0 and 18 mg l−1 ; therefore, 12 sets of responses altogether were collected. So all input data can be presented by a (12 × 14) matrix. To account for the different response magnitudes from the different concentrations, all data were autoscaled by subtracting the mean and dividing by the standard deviation for each sample. PCA results indicated that most of the information (up to 75%) was provided by PC1, PC2 and PC3. the PCA results obtained have been displayed in Fig. 10. The graphs illustrate the score plot of the PC1–PC2 and PC1–PC3 for different three concentrations of BTEX compounds. It is clear that an easy discrimination of BTEX compounds was possible because they were specimen data clustered in four different regions, independent of their concentrations. Table 2 Parameter extracted from the transient response curve Parameters Description P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 P13 P14 Maximum value Time spent to reach the maximum value Minimum value Time spent to reach the minimum value First derivation at maximum value Second derivation at maximum value Integral at maximum value from gas exposure First derivation at minimum value Second derivation at minimum value Integral at minimum value from air exposure Response after 5 min Response after 3 min Desorption period Sorption period 370 A. Mirmohseni et al. / Sensors and Actuators B 121 (2007) 365–371 clearly discriminated from each other. It can be concluded that the polystyrene modified QCN can be utilized to determine BTEX compounds in atmospheric media. Acknowledgment We are most grateful for the continuing financial support of this research project by the University of Tabriz. References Fig. 10. PCA scores plot of benzene (), toluene (), ethylbenzene () and xylene (䊉) in the (a) PC1–PC2 plane and (b) PC1–PC3 plane of the normalized data matrix obtained form the transient responses of QCN sensor for three different concentrations. 4. Conclusion A sensor was constructed based on quartz crystal nanobalance modified with a thin layer of polystyrene to detect benzene, toluene, ethylbenzene and xylene (BTEX compounds) in atmospheric media. Frequency shifts versus concentration of analytes exhibited satisfactory linear correlation within the concentration range of 0–27 mg l−1 . The correlation coefficient of calibration curves was 0.9917, 0.9904, 0.9904 and 0.9925, respectively. The results showed that the sensor is sensitive enough to detect analytes. Principal component analysis was performed using transient response of single sensor. The score plot of the three first PCs showed that BTEX compounds were specimen data clustered in thoroughly different regions, independent of their concentrations. The results indicated that the analytes can be [1] A. Mirmohseni, A. Oladegaragoze, Detection and determination of CrVI in solution using polyaniline modified quartz crystal electrode, J. Appl. Polym. Sci. 85 (2001) 2772–2780. [2] A. Mirmohseni, A. Saeedi, Application of conducting polymer membranes (2): separation of H2 SO4 /H3 PO4 and HNO3 /H2 SO4 using dialysis, electrodialysis and elctrodynamics methods, Iran. Polym. J. 7 (1) (1998) 15–21. [3] J.B. Yates, K.R. Temsamani, O. Ceylan, S. Oztemiz, Electrochemical control of solid phase micro-extraction: conducting polymer coated film material applicable for preconcentration/analysis of neutral species, Talanta 58 (2002) 739–745. [4] L. Cui, M.J. Swann, A. Glidle, J.R. Barker, J.M. Cooper, Odour mapping using microresistor and piezo-electric sensor pairs, Sens. Actuators B: Chem. 66 (2000) 94–97. [5] V. Syritski, J. Reut, A. Opik, K. Idla, Environmental QCM sensors coated with polypyrrole, Synth. Met. 102 (1999) 1326–1327. [6] A. Mirmohseni, A. Oladegragoze, Application of the quartz crystal microbalance for determination of phenol in solution, Sens. Actuators B: Chem. 98 (2004) 28–36. [7] Y. Fu, H.O. Finklea, Quartz crystal microbalance sensor for organic vapor detection based on molecularly imprinted polymers, Anal. Chem. 75 (2003) 5387–5393. [8] A. Mirmohseni, A. Oladegaragoze, Determination of chlorinated aliphatic hydrocarbons in air using a polymer coated quartz crystal microbalance sensor, Sens. Actuators B: Chem. 102 (2004) 261–270. [9] K. Henkel, A. Oprea, I. Paloumpa, G. Appel, D. Schmeiber, P. Kamieth, Selective polypyrrole electrodes for quartz microbalances: NO2 and gas flux sensitivities, Sens. Actuators B: Chem. 76 (2001) 124–129. [10] G. Barkó, J. Abonyi, J. Hlavay, Application of fuzzy clustering and piezoelectric chemical sensor array for investigation on organic compounds, Anal. Chim. Acta 398 (1999) 219–226. [11] R. Ni, X.B. Zhang, W. Liu, G.L. Shen, R.-Q. Yu, Piezoelectric quartz crystal sensor array with optimized oscillator circuit for analysis of organic vapors mixtures, Sens. Actuators B: Chem. 88 (2003) 198–204. [12] F.L. Dickert, O. Hayden, M.E. Zenkel, Detection of volatile compounds with mass-sensitive sensor arrays in the presence of variable ambient humidity, Anal. Chem. 71 (1999) 1338–1341. [13] G. Barko, J. Hlavay, Application of principal component analysis for the characterisation of a piezoelectric sensors array, Anal. Chim. Acta 367 (1998) 135–143. [14] P. Chang, J.S. Shih, Multi-channel piezoelectric quartz crystal sensor for organic vapours, Anal. Chim. Acta 403 (2000) 39–48. [15] C.W. Kuo, J.S. Shih, Cryptand/metal ion coated piezoelectric quartz crystal sensors with artificial back propagation neural network analysis for nitrogen dioxide and carbon monoxide, Sens. Actuators B: Chem. 106 (2005) 468–476. [16] A. Mirmohseni, A. Oladegragoze, Construction of a sensor for determination of ammonia and aliphatic amines using polyvinylpyrrolidone coated quartz crystal microbalance, Sens. Actuators B: Chem. 89 (2003) 164–172. [17] S. Rosler, R. Lucklum, R. Borngraber, J. Hartmann, P. Hauptmann, Sensor system for the detection of organic pollutants in water by thickness shear mode resonators, Sens. Actuators B: Chem. 48 (1998) 415–424. [18] A.K.M. Shafiqul Islam, Z. Ismail, M.N. Ahmad, B. Saad, A.R. Othman, A.Y.Md. Shakaff, A. Daud, Z. Ishak, Transient parameters of a coated quartz crystal microbalance sensor, for the detection of volatile organic compounds (VOCs), Sens. Actuators B: Chem. 109 (2005) 238–243. A. Mirmohseni et al. / Sensors and Actuators B 121 (2007) 365–371 Biographies Abdolreza Mirmohseni received PhD degree in chemistry from the University of Wollongong, Australia in 1974. After a year of postdoc at this university, he was appointed as an assistant professor of chemistry at the University of Tabriz, Iran (second largest university in Iran) where he is now a full professor. He is the corresponding author of over 27 scientific papers in international journals. Prof. A. Mirmohseni has supervised more than 23 MSc and 2 PhD students. He is the author of one chapter in Polymeric Materials Encyclopedia. He has also conducted 7 national research projects. Hamid Abdollahi received PhD degree in chemistry from the University of Shiraz, Iran in 1999. He is assistant professor of analytical chemistry at the 371 Institute for Advanced Studies in Basic Sciences. His interest researches are focused on the area of simultaneous determination of chemical species by chemometric methods and chemometrics studies of chemical equilibria and kinetics. He is the author of over 26 scientific papers in the international journals. Kobra Rostamizadeh is a PhD student in applied chemistry at the University of Tabriz, Iran. She received a BSc degree in 2000 and an MSc degree in 2002 in applied chemistry from the University of Tabriz, Iran. Her current research interests are in selective chemical sensors based on the QCN and she also has research interests in pattern recognition techniques.