- Email: [email protected]
Reduced graphene oxide-based gas sensor array for pattern recognition of DMMP vapor
Sensors and Actuators B 234 (2016) 361–370
Contents lists available at ScienceDirect
Sensors and Actuators B: Chemical journal homepage: www.elsevier.com/locate/snb
Reduced graphene oxide-based gas sensor array for pattern recognition of DMMP vapor Taher Alizadeh ∗ , Leyla Hamed Soltani Department of Analytical Chemistry, Faculty of Chemistry, University College of Science, University of Tehran, P.O. Box 14155-6455, Tehran, Iran
a r t i c l e
i n f o
Article history: Received 27 October 2015 Received in revised form 27 April 2016 Accepted 28 April 2016 Available online 29 April 2016 Keywords: Sensor array Graphene oxide Discrimination Dimethy methylphosphonate Reducing agent
a b s t r a c t In this work, a new sensor array was designed for the discrimination of dimethyl methylphosphonate (DMMP) vapor from some other vapors, capable of interfering with DMMP vapor in the reduced graphene oxide (RGO)-based chemiresistor sensor. The referred sensor array was constructed using different RGOs, obtained by the application of various reducing agents in synthesis of the RGO. Three kinds of reducing agents including hydrazine hydrate, ascorbic acid and sodium borohydride were utilized for the reduction of graphene oxide, synthesized by chemical oxidation of graphite. It was found that the kind of reducing agent, used in the reduction of graphene oxide (GO) had huge effect on the DMMP sensitivity of the related RGO. Sodium borohydride and ascorbic acid led to the RGOs which were much sensitive to DMMP vapor than that was reduced by hydrazine hydrate. The electrical resistance changes of three kinds of RGOs were recorded upon exposing of them to various vapors and the obtained data were analyzed by principal component analysis. The obtained results showed that the RGOs, prepared by hydrazine hydrate and sodium borohydride, were the best combinations of different RGOs to efficiently discriminate DMMP from other vapors. © 2016 Published by Elsevier B.V.
1. Introduction Nerve agents such as sarin and soman, are among the most toxic organophosphate-based chemical warfare agents [1,2]. Toxicity of nerve agents is mainly because of their interfering in operation of acetyl cholinesterase. They inhibit irreversibly the activity of enzyme acetyl cholinesterase, leading to accumulation of neurotransmitter acetylcholine. The accumulation of acetylcholine in body manifests itself by stopping the function of some vital organs like heart, brain, muscules and respiratory system which can finally leads to death [3]. Unfortunately, these highly killing agents have been used many times in human history during wars and also terrorist attacks [4–6]. Therefore, it is urgently required to access an early warning nerve agent sensing system to detect and monitor the presence of these hazardous agents in different environment for military and security applications. Because of extremely high toxicity of nerve agents their handling in laboratory, when testing the related sensor, is very dangerous risk. Therefore, in place of the nerve agents an appropriate stimulant is usually utilized for the testing of sensing devices
∗ Corresponding author. E-mail addresses: [email protected], taa [email protected] (T. Alizadeh). http://dx.doi.org/10.1016/j.snb.2016.04.165 0925-4005/© 2016 Published by Elsevier B.V.
for nerve agents. Dimethyl methylphosphonate (DMMP), known as an appropriate simulant for the nerve agents of sarin or soman, has been generally used in developing various types of nerve agent sensors [2,7]. Various kinds of sensing devices including semiconducting metal oxides (SMO) [8–10], microcantilevers [11,12], quartzcrystal microbalance (QCM) [13] and surface acoustic wave (SAW) sensors [14,15] have been reported for the detection of DMMP. Carbon nanotube [16–18] and graphene [19–21] based materials have also been utilized as the new generation of chemiresistor gas sensor for DMMP detection, representing interesting sensing behavior. However, one major drawback of these sensors is the poor selectivity of them for DMMP recognition. In order to tackle with this disadvantage in most cases a sensor array, known as an electronic nose is applied in place of a single gas sensor [2,22,23]. Reduced graphene oxide (RGO) has shown to be high sensitive sensing material for various gases. In spite of the fact that the RGO material exhibits slight selectivity in sensing of some kinds of target vapors [20], however, it is not an ideally selective material for gas sensing. For instance, RGO has been applied as all-organic vapors sensitive material [24]. However, it has been shown that the selectivity of graphene can be managed by changing of the reducing agent used for the reduction of graphene oxide [19,25].
362
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
In this work, we managed the gas sensing characteristics of the RGO via reducing of graphene oxide by various reducing agents. It was found that the kind of reducing agent had crucial effect on the sensitivity of the resulting RGO to DMMP vapor. Different RGOs with various DMMP vapor sensitivities were used to design a simple sensor array, capable to discriminate DMMP vapor from other vapors.
2. Experimental 2.1. Reagents Hydrazine hydrate (HZ), ascorbic acid (A) and graphite powder were obtained from Merck (Germany). Dimethyl methylphosphonate and different solvents, applied as volatile organic vapors, were
also from Merck (Germany). Sodium borohydride (NaBH4 ) was purchased from Fluka (Buchs, Switzerland). All other chemicals were of analytical grade and purchased from Merck (Germany).
2.2. Synthesis of three kinds of graphene Graphene oxide (GO) was synthesized by chemical oxidation and exfoliation of natural graphite according to a modified Hummers method [26]. The recipe, utilized for the preparation of GO can be found elsewhere [27]. In a typical synthesis procedure, 2 g of natural graphite and 1 g of NaNO3 were mixed with H2 SO4 (95%, 48 mL) in a flask. The mixture was stirred (for 30 min) in an ice bath. Then, 6 g of KMnO4 was progressively added to the suspension under vigorous stirring. The mixture was removed from the ice bath and stirred further at
Fig. 1. scanning electron microscopy image of graphite sheets (a) and the graphene oxide materials reduced by hydrazine hydrate (b), sodium borohydride (c) and ascorbic acid (d).
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
35 ◦ C for 1 h. Then, water (60 mL) was slowly added to the paste, under stirring. The temperature was rapidly increased to 98 ◦ C and the stirring was continued for 2 h. Eventually, H2 O2 solution (30%, 20 mL) was poured in the mixture. The mixture was then washed with HCl (5% w/v) and distillated water (several times) to obtain the purified graphite oxide. The obtained sample was collected and dried in vacuum at 60 ◦ C for 6 h. After synthesis of GO, it was reduced via three kinds reducing agents including ascorbic acid, hydrazine hydrate and sodium borohydride. This led to three types of reduced graphene oxide materials, called as RGO-A, RGO-B and RGO-H, respectively. For producing of RGO-H, the obtained GO (30 mg) was dispersed in 20 mL of water. The GO was then reduced by adding hydrazine hydrate (0.5 mL) into the solution, stirred for 12 h at 50 ◦ C. At the end, the RGO-H was obtained by filtration and dried in vacuum. In order to obtain RGO-B, GO (30 mg) was dispersed in 20 mL of methanol via sonication. Afterwards, the pH of dispersion was increased to about 9 by the addition of Na2 CO3 solution (5%). Then, 330 mg of sodium borohydride was transferred to the dispersion and stirred for 5 h (70 ◦ C). After of completion of the reaction, the mixture was filtered and washed several times with methanol and water until a neutral pH of the filtrate was obtained. The solid samples were dried in a vacuum oven. The RGO-A was produced via reduction of the GO by ascorbic acid in water and at room temperature. For this purpose, 30 mg of GO was dispersed in distilled water (20 mL) and then 200 mg of ascorbic acid was added to the solution. The mixture was stirred for 12 h. At the end, the mixture was filtered, washed with water, dried and saved until use. 2.3. Preparation of sensor device and sampling system The sensor was based on a quartz substrate on which platinum electrodes (20 pairs) were deposited by custom-designed mask. The line widths of the platinum electrodes and the spacing between electrodes were about 0.1 and 0.02 mm, respectively. The determined mass of reduced graphene oxides (5 mg of RGO-H, RGO-B and RGO-A) were weighted and dispersed in a mixture of dimethyl formamide/water (9:1) via sonication for 2 h. Once the RGO was dispersed, little amount (L) of dispersed material was put directly above the electrodes by using a micropipette. The solution put on the electrodes was dried by transferring of the substrate to an oven (keeping there for 12 h at 60 ◦ C). A batch sensing setup was used to measure the RGOs-based sensor array response to the gases of interest. The prepared sensor array was inserted in the test barrel (total volume of 3.4 L). The vapor was generated by dropping of little volume of the target solvent (L, using microsyringe) on a small plate, situated inside the sensing chamber. The plate was electrically heated, resulting in rapidly evaporation of the solvent, placed there. In order to uniformly diffuse of the solvent vapor in the chamber inside area, a fan was also set inside the barrel. The concentration of the vapor generated was calculated with regard to the vapor volume (estimated respecting the injected solvent amount) and the barrel inside volume. After each sensing test, the sensor environment was cleaned by passing of clean and dry nitrogen (proving a relative humidity of below 8%) through the sensing chamber. The sensor array, placed in the sensing chamber was thermally controlled at a fixed temperature of 25 ◦ C using a temperature control processor of model WTC3224 (Wavelength electronic) and the TEC device (AT6-2.501, Marlow), connected to the sensor array. A computer interfaced multi-channel multimeter (Keithley, Model 2700) was used to measure the electrical resistance of the sensors. The resistance (R, ) was measured, and the relative resistance change R/R0 was applied for the evaluation of the vapor responses. R is the difference of the maximum and minimum
363
Fig. 2. Fourier transform infrared spectroscopy (FTIR) of GO and RGO materials obtained via treatment of GO with various reducing agents.
values in the resistance response, and R0 is the initial resistance of the RGO-based sensor (the resistance of a sensor in the absence of any vapor). All data were recorded, saved and then processed by Pentium 3.06 GHz and 1 GB of RAM having Window’s XP operation system. The Minitab 14 software was applied for principal component analysis and cluster analysis. 3. Result and discussion 3.1. Characterization of the synthesized reduced graphene oxide materials Fig. 1 illustrates the scanning electron microscopy (SEM) image of graphite and RGOs, prepared by various reducing agents. These images show that during chemical oxidation, graphite sheets are exfoliated into highly intertwined graphene sheets. This can provide great deal of surface area and facilitate thus the interaction of RGO surface with vapors of interest. Furthermore, according to the SEM images, some differences exist among the graphene oxide materials, created with the same oxidation procedure, but, reduced with various reducing agents. The graphene oxide material, reduced with hydrazine hydrate, is relatively rippled and exfoliated. However, the exfoliation severity in the case of RGOA is lower, compared to RGO-H; but, it is still higher than that of RGO-B. The RGO-B material has a multilayer structure and the flakes are overlapped and not segregated considerably. This can be attributed to the effect of oxygen functional groups existing in the RGO-B. The insufficient removal of the oxygen functional groups from the reduced graphene oxide, in the case of RGO-B, prevents the restacking of the flakes during the reduction stage. Fourier transform infrared spectroscopy (FTIR) of GO and RGO materials, obtained by the treatment of GO with various reducing agents, were illustrated in Fig. 2. In the spectrum of GO the peaks at wavenumbers of 3200, 1718, and 870 cm−1 are assigned to vibrations of oxygenated functional groups of O H, C O, and C O C (from epoxy groups), respectively. Moreover, the peaks at wavenumbers of 1380 cm−1 and 1050 cm−1 are attributed to C OH vibration mode. Upon reductive treatment with borohydride, although, the sharpness of the described peaks are considerably decreased (except C O related peak which is partially disappeared), all of them are still exist in the spectrum, indicating that sodium borohydrine is not capable of completely removing of the oxygen containing functional groups from the GO structure. The spectra related to the RGO-H and RGO-A indicates that hydrazine hydrate and ascorbic acid can effectively remove most of the
364
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
Fig. 3. the XPS spectra of C1s of GO before (a) and after reduction with hydrazine hydrate (b), sodium borohydride (c) and ascorbic acid (d).
oxygen containing functional groups. However, it seems that hydrazine treated material is slightly better vacated from the oxygenated groups, compared to that reduced with ascorbic acid. Fig. 3(a) shows the C 1s XPS spectrum of GO material. As can be seen, five different peaks centered at about 284.5, 285.6, 286.7, 288.3, and 289.5 eV, corresponding to C C/C C in aromatic rings, C OH, C O (epoxy and alkoxy), C O and COOH groups, respectively, are present in the XPS spectrum of GO. Fig. 3(b–d) illustrate the C 1s XPS spectra of RGOs, obtained by the treatment of the GO with hydrazine hydrate, sodium borohydride and ascorbic acid, respectively. As can be seen, although, in all cases the reduction process reduces the oxygenated functional groups of the GO (compared to the XPS spectrum of source GO, depicted in Fig. 2(a)), but, the reduction process are not complete, that is to say, a considerable quantities of different functional groups are still present in the RGOs. On the other hand, it is evident that the reduction intensities of the utilized reducing agents differ considerably, compared to each other. For instance, the hydrazine hydrate and ascorbic acid
reduces the GO more effectively than sodium borohydride, confirming the results of FT-IR. Furthermore, it can be clearly seen that various reducing agents leads to various quantities of oxygenated functional groups in various proportions. As expected, hydrazine hydrate leads to the RGO of more complicated structure, compared to the other tested reducing agents. For example, the peak observed at 285.9 eV, assigned probably to the C N bonding, is absent in two other RGOs. These finding show that really various kinds of the RGOs can be obtained using various kinds of the reducing agents. Thermogravimetric analysis (TGA) of the GO and its derived RGO (Fig. 4) confirms again the previous results, that is to say, different chemical structure is resulted from the same GO by using different reducing agents. Because, different chemical structures lead to various TGA behaviors. It can be seen that all the RGOs are much stable than the GO. A significant weight loss (∼15%) occurs below 100 ◦ C in GO; whereas, such a weight loss is insignificant in the cases of the RGOs. This is attributed to the evaporation of adsorbed water molecules due to the higher hydrophilic nature of GO. Moreover,
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
365
carbon atoms) and 1350 cm−1 (D-band, the symmetry A1 g mode). As a result of chemical oxidation of graphene, the conjugated structure of graphene is relatively lost and this leads to locate below 1 the ratio of D/G. however, with the reduction of graphene oxide its conjugated structure is partly restored and thus the ratio of D/G increases significantly [24,28]. According to the figure and regarding the D/G ratios of Raman spectra of RGOs (all being above 1), one can deduce that the reduction step has successfully restored the conjugated structure of source graphene oxide materials. Most importantly, the Raman spectra of three kinds of RGOs, exhibiting different D/G ratios, show that the described RGOs are really different, regarding their different defect intensities.
3.2. The effect reducing agent kind on the selectivity of the RGO to DMMP
Fig. 4. the TGA curves of GO and the related ROGs, obtained via treatment of the GO with different reducing agents including hydrazine hydrate, ascorbic acid and sodium borohydride.
Fig. 5. Raman spectra of reduced graphene oxide materials of RGO-H (a), RGO-A (b) and RGO-B (c).
the notable weight decrease (∼40%) occurs in the case of GO below 200 ◦ C, whereas, such a weight loss is little than 10% in the cases of the RGOs. However, at temperatures being higher than 200 ◦ C, there are considerable temperature-dependent weight loss differences among various RGOs, leading to really different TGA behaviors for the three examined RGOs. Fig. 5 shows the Raman spectra of different RGOs. The Raman spectrum of a graphene material is usually characterized by two distinct peaks occurring at 1580 cm−1 (G-band, the E2 g mode of sp2
It is well-established fact that different reducing agents possess different reduction powers. Thus, they can lead to different RGOs with various defect densities and remained functional groups. Also, some reducing agents transfer oxygen or nitrogen atoms or the related functional groups to the structure of graphene, altering further the electronic and adsorption characteristics of RGO [28]. For instance, oxygen-based functionalities and structure defects function as electron-withdrawing sources and thus increase the number of the hole-type carriers in the valence band of RGO. This, leads to p-type semiconducting behavior of the related RGO and have important effect in the sensing properties of the RGO [29,30]. Hydrazine is a strong reducing agent and thus most of the oxygen containing functional groups of GO are eliminated by hydrazine. Furthermore, nitrogen atoms of hydrazine tend to bind covalently to the corresponding RGO. Such residual C N groups function as n-type dopants and influence acutely the electronic structure of the resulting RGO [28,31]. Sodium borohydride is a strong reducing agent in nature; but, it is kinetically slow reducing agent. Furthermore, its reducing power is usually weakened via its reaction with water as the general solvent, utilized for the dispersion of GO for the generation of RGO. Therefore, it may result in different C/O ratio in the related RGO, compared to hydrazine [32,33]. Ascorbic acid is weak reducing agent, compared to both hydrazine and sodium borohydride. Thus, the density of oxygenbased defects and moieties in the case of RGO-A may be low, compared to RGO-B and specially RGO-H. Furthermore, no hetro atom like nitrogen is added to the structure of reduced graphene oxide when using ascorbic acid as reducing agent [24,28]. Hence, the p-type characteristic of RGO is guaranteed by applying ascorbic acid as reducing agent. Fig. 6 represents the time dependent response of the sensors fabricated using different kinds of RGOs. It can be seen that in all cases the electrical resistance of the RGO films increase as the result of adsorption of DMMP vapor on them. This is due to the p-type semiconducting behavior of the RGO material [34]. Since, DMMP is a strong electron donor compound; it can decrease the hole-type carrier population in the RGO sheets when approaching to it, increasing thus the electrical resistance of the sensing film. Other thing which is clear in the depicted response curves is the significant effect of the kind of the reducing agent on the sensitivity of the RGOs to DMMP vapor. It can be seen that the DMMP sensitivities of the RGOs obey the order of RGO-A > RGOB > RGO-H. Although, the response difference between RGO-A and RGO-B is small; but, the sensitivity of RGO-H to DMMP is considerably smaller than two other RGOs. Based on the described facts, we think that the defect density as well as the dopant atoms inserted to the reduced graphene oxides (from the reducing agents) has essential effect on the DMMP sensitivity of the RGO materials. As shown, the RGO-B has more oxygen
366
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
Fig. 6. the time dependent response of the sensors fabricated using different kinds of RGOs to DMMP vapor (20 ppm); base resistance of the sensors: RGO-H = 104 K, RGO-B = 1.1 M, RGO-A = 364 K.
containing functional groups, compared to RGO-H. Therefore, the high sensitivity of the RGO-B to DMMP, compared to RGO-H may be assigned to the presence of these groups on the structure of RGO-B which probably simplify the adsorption of basic compound of DMMP. However, comparing the response of RGO-A with that of RGO-H and regarding the fact that the oxygen containing functional groups of RGO-B is much higher of RGO-A, makes one to modify the previously stated mechanism. We think that the dopant atom, arisen from the reducing agent, play a significant role in the sensitivity of the reduced graphene oxide to DMMP vapor. Likely, the additional hetro atoms in the RGO structure (originating from the reducing agent) function as n-type dopants and thus influence negatively the p-type characteristics of the resulting RGO. The electron donating DMMP can decrease the electrical conductivity of the RGO. Thus, any destruction in the p-type characteristic of the RGO material can decrease its sensitivity to the adsorption of electron donating compound like DMMP. We think that the RGO-B and particularly RGO-A have higher p-type characteristic, compared to RGO-H and thus the DMMP sensitivity of RGO-B is considerably higher than that of RGO-H. Fig. 7 represents the responses of the RGO-based sensors to the injection of successive and additive concentration of DMMP and then cleaning of the sensing environment. It can be seen that after every injection step, the sensor response increases sharply as a result of sensor exposing to DMMP vapor and then decreases at relatively slow rate by purging a dry nitrogen to the sensing environment. Moreover, it is clear that the sensor can be completely recovered as a result of gas cleaning step. Such a completely backing of the sensor response to the initial base line, after every gas sensing experiment, is a main problem with most of the RGO-based gas sensors. 3.3. The response of RGOs to other vapors The RGOs, prepared using various reducing agents were also checked for other vapors including acetic acid, triethylamine, ethanol, methanol, n-hexane, acetonitrile, diethylether, dichloromethane, chloroform and acetone. The time dependent responses of the sensors, prepared by RGO-B, RGO-A and RGO-H, to DMMP and the mentioned vapors are shown in Fig. 8. It can be seen that the responses of all RGO-based chemiresistor sensors to DMMP vapor are considerably higher than their responses to other men-
Fig. 7. responses of the RGO-based sensors to the injection of successive and additive concentrations of DMMP (10, 20, 40 and 60 ppm, respectively) and then cleaning of the sensing environment; base resistance of the sensors: RGO-H = 144 K, RGOB = 2.1 M, RGO-A = 394 K.
tioned vapors. Adopting a p-type semiconducting characteristic for the RGOs, this observation can be assigned to the higher electron donating behavior of DMMP, compared to the other tested vapors. However, it is clear that among the vapor tested, the responses of the RGO-based sensors to vapors of thriethylamine and acetic acid are also high and comparable with the signals of the sensor to DMMP. This means that the signal of these vapors may be interfere with that of DMMP, particularly when their concentrations are slightly higher than that of DMMP. Furthermore, the responses of the sensors to the other tested vapors (marked in the figure as VOCs) are very smaller than either DMMP or triethylamine or acetic acid. Although, these vapors can not interfere with DMMP detection in the concentrations equal to that of DMMP, however, if their concentrations are very higher than that of DMMP, they will be considered as potential interfere agents. For this reason, we decided to make a sensor array using the as prepared RGOs to precisely discriminate DMMP from the other vapors. 3.4. Reduced graphene oxide-based sensor array for the differentiation of DMMP from acetic acid and triethylamine Principal component analysis is an effective way for decreasing the dimensionality of a data set. By this technique the original data matrix is decomposed onto a new coordinate base. As a result of the change in the coordinate, the dimensionality of the data of interest is reduced to the principal components. There are no correlations among the principle components. The results of a PCA are generally represented as scores and loads plots. When PCA is applied for the analysis of the sensor array response, the scores plot of the PCA is usually used for studying the classification of the data clusters, while the loads plot can be used for giving information about the relative importance of the sensors to each principal component and their mutual correlation [23]. In this work, the responses of three RGO-based chemiresistor sensors array were normalized towards the concentrations of three organic vapors including DMMP, acetic acid (AA), triethylamine (TEA). The described normalization reduces the dependency of the
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
367
Fig. 8. The responses of sensors of RGO-B(a), RGO-A(b) and RGO-H (c) to 20 ppm of DMMP, acitic acid, triethylamine and other organic compounds vapors; VOCs: ethanol, methanol, n-hexane, diethylether, acetonitrile, chloroform, dichloromethane and acetone; base resistance of the sensors: RGO-H = 144 K, RGO-B = 3.1 M, RGO-A = 394 K.
Fig. 9. The score plot in the PC1 –PC2 plane, obtained by performing of PCA on the data of three RGO-based sensors (including RGO-B, RGO-A and RGO-H) (I); loading factors associated to the first and second principal components of the data set of the three RGO-based sensor array (II).
368
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
Fig. 10. loading factors associated to the first and second principal components of the data set of two-sensor array (I); the score plot in the PC1 –PC2 plane, obtained by performing of PCA on the data of two RGO-based sensor array (including RGO-B and RGO-H) (II).
array response to vapors concentrations and also slightly reduces the effects of the sensors drift. Each vapor was sampled five times at different concentrations and the normalized responses, obtained from the sensors array, were used for statistical analysis. The results of the PCA for the data set with dimensions of 15 × 3, are represented in Fig. 9 as the so called scores (Fig. 9. (I)) and loads (Fig. 9 (II)) plots. Regarding the score plot of the first principal components of the three-sensor array, it is obvious that the senor array has discriminated the aimed vapors, as, each separated vapor data is placed in the distinct location in the PC1 -PC2 plane. As mentioned earlier, the loadings analysis, correlated to the PCA, gives the important information about the mutual correlation of the sensors. Using this analysis, the sensors are evaluated for their obligations for the vapors classification. Sensors with loading parameters near to zero for a particular principal component have an insignificant contribution to the total signal of the array, whereas, high values show discriminating sensors. Thus, sensors, located close to the center of the diagram (0, 0), have a minor responsibility for the distribution of pattern in the PCA plot. Such sensors are usually taken away from the sensor array; because, they may have a negative influence on the pattern resolution [23]. Furthermore, the sensors with almost identical loading parameters can be represented by just one sensor. The loading factors associated to the first and second principal components for each sensor are represented in Fig. 9 (II). As can be seen, for the first 2 principal components there is no sensor with loading parameters near to zero, indicating that all three tested sensors have significant effect on the vapors discrimination. However, it can be seen that sensors fabricated with RGO-B and RGO-A, have loading values partly similar to each other. Hence, we decided to remove the sensor of RGO-A from the array and design a new sensor array composed of only two kinds of RGO-based sensors. The normalized responses of two sensors to different concentrations of the previously mentioned organic vapors were again used for PCA in order to test the effect of removing of RGO-A sensor from the three-sensor array. The results of the principal components analysis for the data set with dimensions of 10 × 2, are represented as loading and score plots in Fig. 10(I) and (II), respectively. It can be seen that none of the sensors in the two-sensor array have loading parameters near to zero, indicating that both sensors have significant effect on the vapors discrimination. Furthermore, it can be
seen that both selected sensors have considerable difference, considering their loading parameters, suggesting that no overlapping exists between the roles of the sensors in vapors discrimination performance. On the other hand, the comparison of the score plots of the two-sensor array with that of the three-sensor array revels that the location of AA in the PC1 –PC2 plane has been exchanged with that of TEA as a result of change in the sensor number in the array. However, no significant variation is observed in the vapor classification of the new sensor array, compared to previous one. This result indicates that the removal of a RGO-A from the sensor array has no significant effect on the vapor discrimination capability of the sensor array. The repeatability as well as the reproducibility of the sensor array, was evaluated by calculation of coefficient of variation (D) for the responses of the various sensors to different vapors, according to Eq. (1): D(%) = ı/m × 100
(1)
where ı is the standard deviation of a set of responses of a sensor (or different sensors) towards a specific vapor concentration (20 ppm) and m is the average of the responses of a sensor (or different sensors) toward the same vapor concentration. The repeatability of each RGO-based sensor was evaluated by calculating the coefficient of variation of the replicated responses (n = 4) of the given RGO to the vapors of DMMP, TEA and AA. The average of the coefficient of variation values, obtained for the RGObased sensors, was considered as the repeatability of the sensor array. Furthermore, the reproducibility of each RGO based sensor, used for the fabrication of the sensor array, was evaluated by calculating the coefficient of variation of the responses of four separate RGO-based sensors, prepared with the same RGO, to a given vapor. The reproducibility of the sensor array can be represented as the average of the obtained reproducibility values. The described values are summarized in Table 1. In this table, the lower D value means the higher repeatability or reproducibility. It can be seen that the D maximum values, do not exceed 9.6%, in the case of inter-sensor variations (evaluating the repeatability characteristics). However, according to the represented data, the D average value obtained
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370 Table 1 coefficient of variation of the responses of various RGO-based gas sensors exposed to different vapors of DMMP, TEA and AA. Vapor tested (20 ppm)
DMMP inter-sensor intra-sensor TEA inter-sensor intra-sensor AA inter sensor Inter sensor total inter-sensor intra-sensor
coefficient of variation, D(%)
averaged D (%)
RGO-H
RGO-B
RGO-A
7.7 10.6
8.2 11.5
7.9 11.8
7.9 11.1
8.3 10.9
7.5 11.6
8.7 9.5
8.2 10.6
9.4 11.3
8.5 12.1
9.6 13.1
9.2 12.2 8.4 11.3
for inter-sensor variation is 8.4%. Furthermore, the D average value obtained for intra-sensors variation is 11.3%. In addition, the life-time of the sensor array was guessed to be at least 6 months; since, during the experimental period of 6 months no significant variation was observed in the responses of the sensors to the tested vapors, compared to their initial responses and regarding the inter-sensor variation values, recorded previously. The DMMP classification capability of the sensor array, described in this work, was tested in very low humidity levels (below 8%), provided by purging of the clean and dry nitrogen through the testing chamber, before the measurements. It was found that the humidity had significant effect on the responses of the sensors to the vapors tested. However, the discrimination of DMMP from the other vapors, examined in this study, was not disturbed up to humidity level of 30%; but, at such a condition two other classes (VOCs and TEA) of the vapors were overlapped significantly. Regardless of overlapping of interfering vapor classes in the PC1–PC2 plane, the discrimination of DMMP from other vapors tested was observed even at relative humidity level of 75%. However, at further humidity levels the DMMP discrimination capability of the developed sensor array started to demolish.
369
3.5. Discrimination of DMMP vapor from highly concentrated volatile organic compound vapors As mentioned previously, the response of the RGO-based sensors were found to be very smaller than that for DMMP as well as TEA and AA. For this aim at the same concentrations the effect of organic vapors like acetone, chloroform etc. on the sensors responses to the DMMP was negligible. However, the response magnitude of the sensors to the organic vapors was increased as their concentrations were increased. Therefore, the presence of high concentrations of the mentioned organic vapors could interfere with the discrimination of DMMP vapor. In order to examine this effect, the responses of the RGO-based sensors to some vapors including acetone, chloroform, ethanol methanol, acetonitrile, diethyl ether, n-hexane and dichloromethane were recorded at their concentration being 50 times higher than DMMP and then the obtained data matrix (23 × 2) was analyzed by PCA for recognition purpose. Fig. 11 shows the obtained result as the score plot (in PC1 –PC2 plane). As can be seen, the DMMP related signals are classified again in a distinct section of the plane which is considerably far from other vapors classes. The data of the newly added vapors is situated near the TEA class. However, magnification of the related zone in the plane (as shown in Fig. 11 (left)) reveals that even TEA class can be discriminated from the tested organic vapors. These results suggest that RGO-based sensor array, composed of only RGO-B and RGO-H, can be effectively utilized for the classification of DMMP from TEA and AA, as the compounds creating high signals in the sensor array as well as from chloroform, acetone, ethanol, methanol, dichlorometane, n-hexane, diethyl etaniline and acetonitrile, as the compounds producing low signals in the sensor array.
4. Conclusion It was demonstrated that the kind of reducing agent, used for the conversion of GO to RGO, had crucial effect on the DMMP sensitivity of the related RGO-based sensor. It was also shown that the GO, treated with reducing agent of ascorbic acid and sodium borohydride, responded to DMMP very better than that treated with hydrazine hydrate. This phenomenon was ascribed to the negative
Fig. 11. the score plot in the PC1 –PC2 plane, obtained by performing of PCA on the signals of two RGO-based sensor array (including RGO-B and RGO-H) to DMMP, AA, TEA and eight kinds of organic vapors (I) the magnified zone of the score plot of (I), (II); DMMP, AA, TEA were tested at five concentrations of 5, 10, 20, 30, 40 ppm; each of as called VOCs was tested at concentration of 500 ppm.
370
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
effect of nitrogen atom, inserted in the RGO when using hydrazine hydrate as reducing agent. Nitrogen atom seems to act as n-dopant agent and destroy the p-type semiconducting characteristic of the corresponding RGO. It was shown that an electronic nose, composed of two RGO-B and RGO-H, was capable of discrimination of DMMP from wide range of the organic vapors including either those with great signals in the sensor array or those creating small responses in sensor array. Although, our aim in this study was to develop a sensor array capable of discrimination of DMMP from other vapors, but, the designed sensor array was found to be capable of discrimination of some other vapors from each other. References [1] X. Du, Z. Ying, Y. Jiang, Z. Liu, T. Yang, G. Xie, Synthesis and evaluation of a new polysiloxane as SAW sensor coatings for DMMP detection, Sens. Actuators B 134 (2008) 409–413. [2] T. Alizadeh, S. Zeynali, Electronic nose based on the polymer coated SAW sensors array for the warfare agent simulants classification, Sens. Actuators B 129 (2008) 412–423. [3] N.B. Munro, S.S. Talmage, G.D. Griffin, L.C. Waters, A.P. Watson, J.F. King, V. Hauschild, Environ. Health Perspect. 107 (1999) 933–974. [4] S.M. Somani, Chemical Warfare Agents, Academic Press, London, 1992. [5] R.M. Black, R.J. Clarke, R.W. Read, M.T.J. Reid, Application of gas chromatography-mass spectrometry and gas chromatography-tandem mass spectrometry to the analysis of chemical warfare samples, found to contain residues of the nerve agent sarin, sulphur mustard and their degradation products, J. Chromatogr. A 662 (1994) 301–321. [6] H. Nozaki, N. Aikawa, Sarin poisoning in Tokyo subway, Lancet 345 (1995) 1446–1447. [7] D.B. Kim, B. Gweon, S.Y. Moon, W. Choe, Decontamination of the chemical warfare agent simulant dimethyl methylphosphonate by means of large-area low-temperature atmospheric pressure plasma, Curr. Appl. Phys. 9 (2009) 1093–1096. [8] C. Leea, H.Y. Choi, S.J. Lee, W.S. Lee, J.S. Huh, D.D. Lee, J.C. Kim, The development of SnO2-based recoverable gas sensors for the detection of DMMP, Sens. Actuators B 137 (2009) 239–245. [9] M. Yang, H.C. Kim, S.H. Hong, assenting behavior of ZnO-coated single-wall carbon nanotube network sensors, Mater. Lett. 89 (2012) 312–315. [10] S.M. Kanan, A. Wagh, B.L. Jensen, C.P. Tripp, Dual WO3 based sensors to selectively detect DMMP in the presence of alcohols, Talanta 72 (2007) 401–407. [11] I. Voiculescu, M.E. Zaghlou, R.A. Mcgil, E.J. Houser, K. Gary, Electrostatically actuated resonant microcantilever beam in cmos technology for the detection of chemical weapons, IEEE Sens. J. 5 (2005) 641–647. [12] C. Karnati, H.W. Du, H.F. Ji, X.H. Xu, Y. Lvov, A. Mulchandani, P. Mulchandani, W. Chen, Organophosphorus hydrolase multilayer modified microcantilevers for organophosphorus detection, Biosens. Bioelectron. 22 (2007) 2636–2642. [13] Z. Ying, Y. Jiang, X. Du, G. Xie, J. Yu, H. Wang, PVDF coated quartz crystal microbalance sensor for DMMP vapor detection, Sens. Actuators B 125 (2007) 167–172. [14] C. Tasaltin, I. Gurol, M. Harbeck, E. Musluoglu, V. Ahsen, Z.Z. Ozturk, Synthesis and DMMP sensing properties of fluoroalkyloxy and fluoroaryloxy substituted phthalocyanines in acoustic sensors, Sens. Actuators B 150 (2010) 781–787. [15] R.A. McGill, V.K. Nguyen, R. Chung, R.E. Shaffer, D. DiLella, J.L. Stepnowski, T.E. Mlsna, D.L. Venezky, D. Dominguez, A nose for toxic gases, Sens. Actuators B 65 (2000) 10–13. [16] Y. Wang, Z. Zhou, Z. Yang, X. Chen, D. Xu, Y. Zhang, Gas sensors based on deposited single-walled carbon nanotube networks for DMMP detection, Nanotechnology 20 (2009) 345502–345510. [17] Y. Wang, Z. Yang, Z. Hou, D. Xu, L. Wei, E.S.W. Kong, Y. Zhang, Flexible gas sensors with assembled carbon nanotube thin films for DMMP vapor detection, Sens. Actuators B 150 (2010) 708–714. [18] S. Mubeen, M. Lai, T. Zhang, J.H. Lim, A. Mulchandani, M.A. Deshussesd, N.V. Myung, Hybrid tin oxide-SWNT nanostructures based gas sensor, Electrochim. Acta 92 (2013) 484–490.
[19] N. Hu, Y. Wang, J. Chai, R. Gao, Z. Yang, E.S.W. Kong, Y. Zhang, Gas sensor based on p-phenylenediamine reduced graphene oxide, Sens. Actuators B 163 (2012) 107–114. [20] T. Alizadeh, L.H. Soltani, Graphene/poly(methyl methacrylate) chemiresistor sensor for formaldehyde odor sensing, J. Hazard. Mater. 248–249 (248) (2013) 401–406. [21] T. Alizadeh, L.H. Soltani, Graphene/graphite/molecularly imprinted polymer nanocomposite as the highly selective gas sensor for nitrobenzene vapor recognition, J. Environ. Chem. Eng. 2 (2014) 1514–1526. [22] A.J. Ricco, R.M. Crooks, G.C. Osbourn, Surface acoustic wave chemical sensor arrays: new chemically sensitive interfaces combined with novel cluster analysis to detect volatile organic compounds and mixtures, Acc. Chem. Res. 31 (1998) 289–296. [23] T. Alizadeh, Chemiresistor sensors array optimization by using the method of coupled statistical techniques and its application as an electronic nose for some organic vapors recognition, Sens. Actuators B 143 (2010) 740–749. [24] V. Dua, S.P. Asurwade, S. Ammu, S.R. Agnihotra, S. Jain, K.E. Roberts, S. Park, R.S. Ruoff, S.K. Manohar, All-organic vapor sensor using inkjet-printed reduced graphene oxide, Angew. Chem. Int. Ed. 49 (2010) 2154–2157. [25] X. Huang, N. Hu, L. Zhang, L. Wei, H. Wei, Y. Zhang, The NH3 sensing properties of gas sensors based on aniline reduced graphene oxide, Synth. Met. 185–186 (2013) 25–30. [26] S. William, J.R. Hummers, R. Offeman, Preparation of graphitic oxide, J. Am.Chem. Soc. 80 (1958), 1339–1339. [27] T. Alizadeh, S. Mirzagholipur, A Nafion-free non-enzymatic amperometric glucose sensor based oncopper oxide nanoparticles-graphene nanocomposite, Sens. Actuators B 198 (2014) 438–447. [28] S. Pei, H.M. Cheng, The reduction of graphene oxide, Carbon 50 (2012) 3210–3228. [29] S. Park, R.S. Ruoff, Chemical methods for the production of graphenes, Nat. Nanotechnol. 4 (2009) 217–224. [30] J.D. Fowler, M.J. Allen, V.C. Tung, Y. Yang, R.B. Kaner, B.H. Weiller, Practical chemical sensors from chemically derived graphene, ACS Nano 3 (2009) 301–306. [31] S. Park, J. An, J.R. Potts, A. Velamakanni, S. Murali, R.S. Ruoff, Hydrazine-reduction of graphite- and graphene oxide, Carbon 449 (2011) 3019–3023. [32] H.J. Shin, K.K. Kim, A. Benaya, S.M. Yoon, H.K. Park, I.S. Jung, M.H. Jin, H.K. Jeong, J.M. Kim, J.Y. Choi, Y.H. Lee, Efficient reduction of graphite oxide by sodium borohydride and its effect on electrical conductance, Adv. Funct. Mater. 19 (2009) 1987–1992. [33] C.K. Chua, M. Pumera, Reduction of graphene oxide with substituted borohydrides, J. Mater. Chem. A 1 (2013) 1892–1898. [34] F. Yavari, Z. Chen, A.V. Thomas, W. Ren, H.M. Cheng, N. Koratkar, High sensitivity gas detection using a macroscopic three-dimensional graphene foam network, Sci. Rep. 1 (2011) 166–171.
Biographies Taher Alizadeh received the BS degree from the University of Guilan, Rasht, Iran, in 1999. He received MS degree from the University of Tabriz, Iran, in 2002 and PhD degree from the University of Tehran, Iran, in 2006. He was with University of Mohaghegh Ardabili, from September 2006 to October 2013. Now, he is continuing his research activities as an associated professor in Department of Analytical Chemistry, Faculty of Chemistry, University Collage of Science, University of Tehran, Iran. His current research interests include the development of chemical sensors and electronic nose based on new materials, Electrochemistry and molecular/Ionic imprinting technology. Leyla Hamed Soltani received the BS degree from the University of Mohaghegh Ardabili in 2010. She received MS degree in analytical chemistry from Department of Applied Chemistry, University of Mohaghegh Ardabili, Ardabil, Iran, in 2013.
Contents lists available at ScienceDirect
Sensors and Actuators B: Chemical journal homepage: www.elsevier.com/locate/snb
Reduced graphene oxide-based gas sensor array for pattern recognition of DMMP vapor Taher Alizadeh ∗ , Leyla Hamed Soltani Department of Analytical Chemistry, Faculty of Chemistry, University College of Science, University of Tehran, P.O. Box 14155-6455, Tehran, Iran
a r t i c l e
i n f o
Article history: Received 27 October 2015 Received in revised form 27 April 2016 Accepted 28 April 2016 Available online 29 April 2016 Keywords: Sensor array Graphene oxide Discrimination Dimethy methylphosphonate Reducing agent
a b s t r a c t In this work, a new sensor array was designed for the discrimination of dimethyl methylphosphonate (DMMP) vapor from some other vapors, capable of interfering with DMMP vapor in the reduced graphene oxide (RGO)-based chemiresistor sensor. The referred sensor array was constructed using different RGOs, obtained by the application of various reducing agents in synthesis of the RGO. Three kinds of reducing agents including hydrazine hydrate, ascorbic acid and sodium borohydride were utilized for the reduction of graphene oxide, synthesized by chemical oxidation of graphite. It was found that the kind of reducing agent, used in the reduction of graphene oxide (GO) had huge effect on the DMMP sensitivity of the related RGO. Sodium borohydride and ascorbic acid led to the RGOs which were much sensitive to DMMP vapor than that was reduced by hydrazine hydrate. The electrical resistance changes of three kinds of RGOs were recorded upon exposing of them to various vapors and the obtained data were analyzed by principal component analysis. The obtained results showed that the RGOs, prepared by hydrazine hydrate and sodium borohydride, were the best combinations of different RGOs to efficiently discriminate DMMP from other vapors. © 2016 Published by Elsevier B.V.
1. Introduction Nerve agents such as sarin and soman, are among the most toxic organophosphate-based chemical warfare agents [1,2]. Toxicity of nerve agents is mainly because of their interfering in operation of acetyl cholinesterase. They inhibit irreversibly the activity of enzyme acetyl cholinesterase, leading to accumulation of neurotransmitter acetylcholine. The accumulation of acetylcholine in body manifests itself by stopping the function of some vital organs like heart, brain, muscules and respiratory system which can finally leads to death [3]. Unfortunately, these highly killing agents have been used many times in human history during wars and also terrorist attacks [4–6]. Therefore, it is urgently required to access an early warning nerve agent sensing system to detect and monitor the presence of these hazardous agents in different environment for military and security applications. Because of extremely high toxicity of nerve agents their handling in laboratory, when testing the related sensor, is very dangerous risk. Therefore, in place of the nerve agents an appropriate stimulant is usually utilized for the testing of sensing devices
∗ Corresponding author. E-mail addresses: [email protected], taa [email protected] (T. Alizadeh). http://dx.doi.org/10.1016/j.snb.2016.04.165 0925-4005/© 2016 Published by Elsevier B.V.
for nerve agents. Dimethyl methylphosphonate (DMMP), known as an appropriate simulant for the nerve agents of sarin or soman, has been generally used in developing various types of nerve agent sensors [2,7]. Various kinds of sensing devices including semiconducting metal oxides (SMO) [8–10], microcantilevers [11,12], quartzcrystal microbalance (QCM) [13] and surface acoustic wave (SAW) sensors [14,15] have been reported for the detection of DMMP. Carbon nanotube [16–18] and graphene [19–21] based materials have also been utilized as the new generation of chemiresistor gas sensor for DMMP detection, representing interesting sensing behavior. However, one major drawback of these sensors is the poor selectivity of them for DMMP recognition. In order to tackle with this disadvantage in most cases a sensor array, known as an electronic nose is applied in place of a single gas sensor [2,22,23]. Reduced graphene oxide (RGO) has shown to be high sensitive sensing material for various gases. In spite of the fact that the RGO material exhibits slight selectivity in sensing of some kinds of target vapors [20], however, it is not an ideally selective material for gas sensing. For instance, RGO has been applied as all-organic vapors sensitive material [24]. However, it has been shown that the selectivity of graphene can be managed by changing of the reducing agent used for the reduction of graphene oxide [19,25].
362
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
In this work, we managed the gas sensing characteristics of the RGO via reducing of graphene oxide by various reducing agents. It was found that the kind of reducing agent had crucial effect on the sensitivity of the resulting RGO to DMMP vapor. Different RGOs with various DMMP vapor sensitivities were used to design a simple sensor array, capable to discriminate DMMP vapor from other vapors.
2. Experimental 2.1. Reagents Hydrazine hydrate (HZ), ascorbic acid (A) and graphite powder were obtained from Merck (Germany). Dimethyl methylphosphonate and different solvents, applied as volatile organic vapors, were
also from Merck (Germany). Sodium borohydride (NaBH4 ) was purchased from Fluka (Buchs, Switzerland). All other chemicals were of analytical grade and purchased from Merck (Germany).
2.2. Synthesis of three kinds of graphene Graphene oxide (GO) was synthesized by chemical oxidation and exfoliation of natural graphite according to a modified Hummers method [26]. The recipe, utilized for the preparation of GO can be found elsewhere [27]. In a typical synthesis procedure, 2 g of natural graphite and 1 g of NaNO3 were mixed with H2 SO4 (95%, 48 mL) in a flask. The mixture was stirred (for 30 min) in an ice bath. Then, 6 g of KMnO4 was progressively added to the suspension under vigorous stirring. The mixture was removed from the ice bath and stirred further at
Fig. 1. scanning electron microscopy image of graphite sheets (a) and the graphene oxide materials reduced by hydrazine hydrate (b), sodium borohydride (c) and ascorbic acid (d).
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
35 ◦ C for 1 h. Then, water (60 mL) was slowly added to the paste, under stirring. The temperature was rapidly increased to 98 ◦ C and the stirring was continued for 2 h. Eventually, H2 O2 solution (30%, 20 mL) was poured in the mixture. The mixture was then washed with HCl (5% w/v) and distillated water (several times) to obtain the purified graphite oxide. The obtained sample was collected and dried in vacuum at 60 ◦ C for 6 h. After synthesis of GO, it was reduced via three kinds reducing agents including ascorbic acid, hydrazine hydrate and sodium borohydride. This led to three types of reduced graphene oxide materials, called as RGO-A, RGO-B and RGO-H, respectively. For producing of RGO-H, the obtained GO (30 mg) was dispersed in 20 mL of water. The GO was then reduced by adding hydrazine hydrate (0.5 mL) into the solution, stirred for 12 h at 50 ◦ C. At the end, the RGO-H was obtained by filtration and dried in vacuum. In order to obtain RGO-B, GO (30 mg) was dispersed in 20 mL of methanol via sonication. Afterwards, the pH of dispersion was increased to about 9 by the addition of Na2 CO3 solution (5%). Then, 330 mg of sodium borohydride was transferred to the dispersion and stirred for 5 h (70 ◦ C). After of completion of the reaction, the mixture was filtered and washed several times with methanol and water until a neutral pH of the filtrate was obtained. The solid samples were dried in a vacuum oven. The RGO-A was produced via reduction of the GO by ascorbic acid in water and at room temperature. For this purpose, 30 mg of GO was dispersed in distilled water (20 mL) and then 200 mg of ascorbic acid was added to the solution. The mixture was stirred for 12 h. At the end, the mixture was filtered, washed with water, dried and saved until use. 2.3. Preparation of sensor device and sampling system The sensor was based on a quartz substrate on which platinum electrodes (20 pairs) were deposited by custom-designed mask. The line widths of the platinum electrodes and the spacing between electrodes were about 0.1 and 0.02 mm, respectively. The determined mass of reduced graphene oxides (5 mg of RGO-H, RGO-B and RGO-A) were weighted and dispersed in a mixture of dimethyl formamide/water (9:1) via sonication for 2 h. Once the RGO was dispersed, little amount (L) of dispersed material was put directly above the electrodes by using a micropipette. The solution put on the electrodes was dried by transferring of the substrate to an oven (keeping there for 12 h at 60 ◦ C). A batch sensing setup was used to measure the RGOs-based sensor array response to the gases of interest. The prepared sensor array was inserted in the test barrel (total volume of 3.4 L). The vapor was generated by dropping of little volume of the target solvent (L, using microsyringe) on a small plate, situated inside the sensing chamber. The plate was electrically heated, resulting in rapidly evaporation of the solvent, placed there. In order to uniformly diffuse of the solvent vapor in the chamber inside area, a fan was also set inside the barrel. The concentration of the vapor generated was calculated with regard to the vapor volume (estimated respecting the injected solvent amount) and the barrel inside volume. After each sensing test, the sensor environment was cleaned by passing of clean and dry nitrogen (proving a relative humidity of below 8%) through the sensing chamber. The sensor array, placed in the sensing chamber was thermally controlled at a fixed temperature of 25 ◦ C using a temperature control processor of model WTC3224 (Wavelength electronic) and the TEC device (AT6-2.501, Marlow), connected to the sensor array. A computer interfaced multi-channel multimeter (Keithley, Model 2700) was used to measure the electrical resistance of the sensors. The resistance (R, ) was measured, and the relative resistance change R/R0 was applied for the evaluation of the vapor responses. R is the difference of the maximum and minimum
363
Fig. 2. Fourier transform infrared spectroscopy (FTIR) of GO and RGO materials obtained via treatment of GO with various reducing agents.
values in the resistance response, and R0 is the initial resistance of the RGO-based sensor (the resistance of a sensor in the absence of any vapor). All data were recorded, saved and then processed by Pentium 3.06 GHz and 1 GB of RAM having Window’s XP operation system. The Minitab 14 software was applied for principal component analysis and cluster analysis. 3. Result and discussion 3.1. Characterization of the synthesized reduced graphene oxide materials Fig. 1 illustrates the scanning electron microscopy (SEM) image of graphite and RGOs, prepared by various reducing agents. These images show that during chemical oxidation, graphite sheets are exfoliated into highly intertwined graphene sheets. This can provide great deal of surface area and facilitate thus the interaction of RGO surface with vapors of interest. Furthermore, according to the SEM images, some differences exist among the graphene oxide materials, created with the same oxidation procedure, but, reduced with various reducing agents. The graphene oxide material, reduced with hydrazine hydrate, is relatively rippled and exfoliated. However, the exfoliation severity in the case of RGOA is lower, compared to RGO-H; but, it is still higher than that of RGO-B. The RGO-B material has a multilayer structure and the flakes are overlapped and not segregated considerably. This can be attributed to the effect of oxygen functional groups existing in the RGO-B. The insufficient removal of the oxygen functional groups from the reduced graphene oxide, in the case of RGO-B, prevents the restacking of the flakes during the reduction stage. Fourier transform infrared spectroscopy (FTIR) of GO and RGO materials, obtained by the treatment of GO with various reducing agents, were illustrated in Fig. 2. In the spectrum of GO the peaks at wavenumbers of 3200, 1718, and 870 cm−1 are assigned to vibrations of oxygenated functional groups of O H, C O, and C O C (from epoxy groups), respectively. Moreover, the peaks at wavenumbers of 1380 cm−1 and 1050 cm−1 are attributed to C OH vibration mode. Upon reductive treatment with borohydride, although, the sharpness of the described peaks are considerably decreased (except C O related peak which is partially disappeared), all of them are still exist in the spectrum, indicating that sodium borohydrine is not capable of completely removing of the oxygen containing functional groups from the GO structure. The spectra related to the RGO-H and RGO-A indicates that hydrazine hydrate and ascorbic acid can effectively remove most of the
364
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
Fig. 3. the XPS spectra of C1s of GO before (a) and after reduction with hydrazine hydrate (b), sodium borohydride (c) and ascorbic acid (d).
oxygen containing functional groups. However, it seems that hydrazine treated material is slightly better vacated from the oxygenated groups, compared to that reduced with ascorbic acid. Fig. 3(a) shows the C 1s XPS spectrum of GO material. As can be seen, five different peaks centered at about 284.5, 285.6, 286.7, 288.3, and 289.5 eV, corresponding to C C/C C in aromatic rings, C OH, C O (epoxy and alkoxy), C O and COOH groups, respectively, are present in the XPS spectrum of GO. Fig. 3(b–d) illustrate the C 1s XPS spectra of RGOs, obtained by the treatment of the GO with hydrazine hydrate, sodium borohydride and ascorbic acid, respectively. As can be seen, although, in all cases the reduction process reduces the oxygenated functional groups of the GO (compared to the XPS spectrum of source GO, depicted in Fig. 2(a)), but, the reduction process are not complete, that is to say, a considerable quantities of different functional groups are still present in the RGOs. On the other hand, it is evident that the reduction intensities of the utilized reducing agents differ considerably, compared to each other. For instance, the hydrazine hydrate and ascorbic acid
reduces the GO more effectively than sodium borohydride, confirming the results of FT-IR. Furthermore, it can be clearly seen that various reducing agents leads to various quantities of oxygenated functional groups in various proportions. As expected, hydrazine hydrate leads to the RGO of more complicated structure, compared to the other tested reducing agents. For example, the peak observed at 285.9 eV, assigned probably to the C N bonding, is absent in two other RGOs. These finding show that really various kinds of the RGOs can be obtained using various kinds of the reducing agents. Thermogravimetric analysis (TGA) of the GO and its derived RGO (Fig. 4) confirms again the previous results, that is to say, different chemical structure is resulted from the same GO by using different reducing agents. Because, different chemical structures lead to various TGA behaviors. It can be seen that all the RGOs are much stable than the GO. A significant weight loss (∼15%) occurs below 100 ◦ C in GO; whereas, such a weight loss is insignificant in the cases of the RGOs. This is attributed to the evaporation of adsorbed water molecules due to the higher hydrophilic nature of GO. Moreover,
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
365
carbon atoms) and 1350 cm−1 (D-band, the symmetry A1 g mode). As a result of chemical oxidation of graphene, the conjugated structure of graphene is relatively lost and this leads to locate below 1 the ratio of D/G. however, with the reduction of graphene oxide its conjugated structure is partly restored and thus the ratio of D/G increases significantly [24,28]. According to the figure and regarding the D/G ratios of Raman spectra of RGOs (all being above 1), one can deduce that the reduction step has successfully restored the conjugated structure of source graphene oxide materials. Most importantly, the Raman spectra of three kinds of RGOs, exhibiting different D/G ratios, show that the described RGOs are really different, regarding their different defect intensities.
3.2. The effect reducing agent kind on the selectivity of the RGO to DMMP
Fig. 4. the TGA curves of GO and the related ROGs, obtained via treatment of the GO with different reducing agents including hydrazine hydrate, ascorbic acid and sodium borohydride.
Fig. 5. Raman spectra of reduced graphene oxide materials of RGO-H (a), RGO-A (b) and RGO-B (c).
the notable weight decrease (∼40%) occurs in the case of GO below 200 ◦ C, whereas, such a weight loss is little than 10% in the cases of the RGOs. However, at temperatures being higher than 200 ◦ C, there are considerable temperature-dependent weight loss differences among various RGOs, leading to really different TGA behaviors for the three examined RGOs. Fig. 5 shows the Raman spectra of different RGOs. The Raman spectrum of a graphene material is usually characterized by two distinct peaks occurring at 1580 cm−1 (G-band, the E2 g mode of sp2
It is well-established fact that different reducing agents possess different reduction powers. Thus, they can lead to different RGOs with various defect densities and remained functional groups. Also, some reducing agents transfer oxygen or nitrogen atoms or the related functional groups to the structure of graphene, altering further the electronic and adsorption characteristics of RGO [28]. For instance, oxygen-based functionalities and structure defects function as electron-withdrawing sources and thus increase the number of the hole-type carriers in the valence band of RGO. This, leads to p-type semiconducting behavior of the related RGO and have important effect in the sensing properties of the RGO [29,30]. Hydrazine is a strong reducing agent and thus most of the oxygen containing functional groups of GO are eliminated by hydrazine. Furthermore, nitrogen atoms of hydrazine tend to bind covalently to the corresponding RGO. Such residual C N groups function as n-type dopants and influence acutely the electronic structure of the resulting RGO [28,31]. Sodium borohydride is a strong reducing agent in nature; but, it is kinetically slow reducing agent. Furthermore, its reducing power is usually weakened via its reaction with water as the general solvent, utilized for the dispersion of GO for the generation of RGO. Therefore, it may result in different C/O ratio in the related RGO, compared to hydrazine [32,33]. Ascorbic acid is weak reducing agent, compared to both hydrazine and sodium borohydride. Thus, the density of oxygenbased defects and moieties in the case of RGO-A may be low, compared to RGO-B and specially RGO-H. Furthermore, no hetro atom like nitrogen is added to the structure of reduced graphene oxide when using ascorbic acid as reducing agent [24,28]. Hence, the p-type characteristic of RGO is guaranteed by applying ascorbic acid as reducing agent. Fig. 6 represents the time dependent response of the sensors fabricated using different kinds of RGOs. It can be seen that in all cases the electrical resistance of the RGO films increase as the result of adsorption of DMMP vapor on them. This is due to the p-type semiconducting behavior of the RGO material [34]. Since, DMMP is a strong electron donor compound; it can decrease the hole-type carrier population in the RGO sheets when approaching to it, increasing thus the electrical resistance of the sensing film. Other thing which is clear in the depicted response curves is the significant effect of the kind of the reducing agent on the sensitivity of the RGOs to DMMP vapor. It can be seen that the DMMP sensitivities of the RGOs obey the order of RGO-A > RGOB > RGO-H. Although, the response difference between RGO-A and RGO-B is small; but, the sensitivity of RGO-H to DMMP is considerably smaller than two other RGOs. Based on the described facts, we think that the defect density as well as the dopant atoms inserted to the reduced graphene oxides (from the reducing agents) has essential effect on the DMMP sensitivity of the RGO materials. As shown, the RGO-B has more oxygen
366
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
Fig. 6. the time dependent response of the sensors fabricated using different kinds of RGOs to DMMP vapor (20 ppm); base resistance of the sensors: RGO-H = 104 K, RGO-B = 1.1 M, RGO-A = 364 K.
containing functional groups, compared to RGO-H. Therefore, the high sensitivity of the RGO-B to DMMP, compared to RGO-H may be assigned to the presence of these groups on the structure of RGO-B which probably simplify the adsorption of basic compound of DMMP. However, comparing the response of RGO-A with that of RGO-H and regarding the fact that the oxygen containing functional groups of RGO-B is much higher of RGO-A, makes one to modify the previously stated mechanism. We think that the dopant atom, arisen from the reducing agent, play a significant role in the sensitivity of the reduced graphene oxide to DMMP vapor. Likely, the additional hetro atoms in the RGO structure (originating from the reducing agent) function as n-type dopants and thus influence negatively the p-type characteristics of the resulting RGO. The electron donating DMMP can decrease the electrical conductivity of the RGO. Thus, any destruction in the p-type characteristic of the RGO material can decrease its sensitivity to the adsorption of electron donating compound like DMMP. We think that the RGO-B and particularly RGO-A have higher p-type characteristic, compared to RGO-H and thus the DMMP sensitivity of RGO-B is considerably higher than that of RGO-H. Fig. 7 represents the responses of the RGO-based sensors to the injection of successive and additive concentration of DMMP and then cleaning of the sensing environment. It can be seen that after every injection step, the sensor response increases sharply as a result of sensor exposing to DMMP vapor and then decreases at relatively slow rate by purging a dry nitrogen to the sensing environment. Moreover, it is clear that the sensor can be completely recovered as a result of gas cleaning step. Such a completely backing of the sensor response to the initial base line, after every gas sensing experiment, is a main problem with most of the RGO-based gas sensors. 3.3. The response of RGOs to other vapors The RGOs, prepared using various reducing agents were also checked for other vapors including acetic acid, triethylamine, ethanol, methanol, n-hexane, acetonitrile, diethylether, dichloromethane, chloroform and acetone. The time dependent responses of the sensors, prepared by RGO-B, RGO-A and RGO-H, to DMMP and the mentioned vapors are shown in Fig. 8. It can be seen that the responses of all RGO-based chemiresistor sensors to DMMP vapor are considerably higher than their responses to other men-
Fig. 7. responses of the RGO-based sensors to the injection of successive and additive concentrations of DMMP (10, 20, 40 and 60 ppm, respectively) and then cleaning of the sensing environment; base resistance of the sensors: RGO-H = 144 K, RGOB = 2.1 M, RGO-A = 394 K.
tioned vapors. Adopting a p-type semiconducting characteristic for the RGOs, this observation can be assigned to the higher electron donating behavior of DMMP, compared to the other tested vapors. However, it is clear that among the vapor tested, the responses of the RGO-based sensors to vapors of thriethylamine and acetic acid are also high and comparable with the signals of the sensor to DMMP. This means that the signal of these vapors may be interfere with that of DMMP, particularly when their concentrations are slightly higher than that of DMMP. Furthermore, the responses of the sensors to the other tested vapors (marked in the figure as VOCs) are very smaller than either DMMP or triethylamine or acetic acid. Although, these vapors can not interfere with DMMP detection in the concentrations equal to that of DMMP, however, if their concentrations are very higher than that of DMMP, they will be considered as potential interfere agents. For this reason, we decided to make a sensor array using the as prepared RGOs to precisely discriminate DMMP from the other vapors. 3.4. Reduced graphene oxide-based sensor array for the differentiation of DMMP from acetic acid and triethylamine Principal component analysis is an effective way for decreasing the dimensionality of a data set. By this technique the original data matrix is decomposed onto a new coordinate base. As a result of the change in the coordinate, the dimensionality of the data of interest is reduced to the principal components. There are no correlations among the principle components. The results of a PCA are generally represented as scores and loads plots. When PCA is applied for the analysis of the sensor array response, the scores plot of the PCA is usually used for studying the classification of the data clusters, while the loads plot can be used for giving information about the relative importance of the sensors to each principal component and their mutual correlation [23]. In this work, the responses of three RGO-based chemiresistor sensors array were normalized towards the concentrations of three organic vapors including DMMP, acetic acid (AA), triethylamine (TEA). The described normalization reduces the dependency of the
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
367
Fig. 8. The responses of sensors of RGO-B(a), RGO-A(b) and RGO-H (c) to 20 ppm of DMMP, acitic acid, triethylamine and other organic compounds vapors; VOCs: ethanol, methanol, n-hexane, diethylether, acetonitrile, chloroform, dichloromethane and acetone; base resistance of the sensors: RGO-H = 144 K, RGO-B = 3.1 M, RGO-A = 394 K.
Fig. 9. The score plot in the PC1 –PC2 plane, obtained by performing of PCA on the data of three RGO-based sensors (including RGO-B, RGO-A and RGO-H) (I); loading factors associated to the first and second principal components of the data set of the three RGO-based sensor array (II).
368
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
Fig. 10. loading factors associated to the first and second principal components of the data set of two-sensor array (I); the score plot in the PC1 –PC2 plane, obtained by performing of PCA on the data of two RGO-based sensor array (including RGO-B and RGO-H) (II).
array response to vapors concentrations and also slightly reduces the effects of the sensors drift. Each vapor was sampled five times at different concentrations and the normalized responses, obtained from the sensors array, were used for statistical analysis. The results of the PCA for the data set with dimensions of 15 × 3, are represented in Fig. 9 as the so called scores (Fig. 9. (I)) and loads (Fig. 9 (II)) plots. Regarding the score plot of the first principal components of the three-sensor array, it is obvious that the senor array has discriminated the aimed vapors, as, each separated vapor data is placed in the distinct location in the PC1 -PC2 plane. As mentioned earlier, the loadings analysis, correlated to the PCA, gives the important information about the mutual correlation of the sensors. Using this analysis, the sensors are evaluated for their obligations for the vapors classification. Sensors with loading parameters near to zero for a particular principal component have an insignificant contribution to the total signal of the array, whereas, high values show discriminating sensors. Thus, sensors, located close to the center of the diagram (0, 0), have a minor responsibility for the distribution of pattern in the PCA plot. Such sensors are usually taken away from the sensor array; because, they may have a negative influence on the pattern resolution [23]. Furthermore, the sensors with almost identical loading parameters can be represented by just one sensor. The loading factors associated to the first and second principal components for each sensor are represented in Fig. 9 (II). As can be seen, for the first 2 principal components there is no sensor with loading parameters near to zero, indicating that all three tested sensors have significant effect on the vapors discrimination. However, it can be seen that sensors fabricated with RGO-B and RGO-A, have loading values partly similar to each other. Hence, we decided to remove the sensor of RGO-A from the array and design a new sensor array composed of only two kinds of RGO-based sensors. The normalized responses of two sensors to different concentrations of the previously mentioned organic vapors were again used for PCA in order to test the effect of removing of RGO-A sensor from the three-sensor array. The results of the principal components analysis for the data set with dimensions of 10 × 2, are represented as loading and score plots in Fig. 10(I) and (II), respectively. It can be seen that none of the sensors in the two-sensor array have loading parameters near to zero, indicating that both sensors have significant effect on the vapors discrimination. Furthermore, it can be
seen that both selected sensors have considerable difference, considering their loading parameters, suggesting that no overlapping exists between the roles of the sensors in vapors discrimination performance. On the other hand, the comparison of the score plots of the two-sensor array with that of the three-sensor array revels that the location of AA in the PC1 –PC2 plane has been exchanged with that of TEA as a result of change in the sensor number in the array. However, no significant variation is observed in the vapor classification of the new sensor array, compared to previous one. This result indicates that the removal of a RGO-A from the sensor array has no significant effect on the vapor discrimination capability of the sensor array. The repeatability as well as the reproducibility of the sensor array, was evaluated by calculation of coefficient of variation (D) for the responses of the various sensors to different vapors, according to Eq. (1): D(%) = ı/m × 100
(1)
where ı is the standard deviation of a set of responses of a sensor (or different sensors) towards a specific vapor concentration (20 ppm) and m is the average of the responses of a sensor (or different sensors) toward the same vapor concentration. The repeatability of each RGO-based sensor was evaluated by calculating the coefficient of variation of the replicated responses (n = 4) of the given RGO to the vapors of DMMP, TEA and AA. The average of the coefficient of variation values, obtained for the RGObased sensors, was considered as the repeatability of the sensor array. Furthermore, the reproducibility of each RGO based sensor, used for the fabrication of the sensor array, was evaluated by calculating the coefficient of variation of the responses of four separate RGO-based sensors, prepared with the same RGO, to a given vapor. The reproducibility of the sensor array can be represented as the average of the obtained reproducibility values. The described values are summarized in Table 1. In this table, the lower D value means the higher repeatability or reproducibility. It can be seen that the D maximum values, do not exceed 9.6%, in the case of inter-sensor variations (evaluating the repeatability characteristics). However, according to the represented data, the D average value obtained
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370 Table 1 coefficient of variation of the responses of various RGO-based gas sensors exposed to different vapors of DMMP, TEA and AA. Vapor tested (20 ppm)
DMMP inter-sensor intra-sensor TEA inter-sensor intra-sensor AA inter sensor Inter sensor total inter-sensor intra-sensor
coefficient of variation, D(%)
averaged D (%)
RGO-H
RGO-B
RGO-A
7.7 10.6
8.2 11.5
7.9 11.8
7.9 11.1
8.3 10.9
7.5 11.6
8.7 9.5
8.2 10.6
9.4 11.3
8.5 12.1
9.6 13.1
9.2 12.2 8.4 11.3
for inter-sensor variation is 8.4%. Furthermore, the D average value obtained for intra-sensors variation is 11.3%. In addition, the life-time of the sensor array was guessed to be at least 6 months; since, during the experimental period of 6 months no significant variation was observed in the responses of the sensors to the tested vapors, compared to their initial responses and regarding the inter-sensor variation values, recorded previously. The DMMP classification capability of the sensor array, described in this work, was tested in very low humidity levels (below 8%), provided by purging of the clean and dry nitrogen through the testing chamber, before the measurements. It was found that the humidity had significant effect on the responses of the sensors to the vapors tested. However, the discrimination of DMMP from the other vapors, examined in this study, was not disturbed up to humidity level of 30%; but, at such a condition two other classes (VOCs and TEA) of the vapors were overlapped significantly. Regardless of overlapping of interfering vapor classes in the PC1–PC2 plane, the discrimination of DMMP from other vapors tested was observed even at relative humidity level of 75%. However, at further humidity levels the DMMP discrimination capability of the developed sensor array started to demolish.
369
3.5. Discrimination of DMMP vapor from highly concentrated volatile organic compound vapors As mentioned previously, the response of the RGO-based sensors were found to be very smaller than that for DMMP as well as TEA and AA. For this aim at the same concentrations the effect of organic vapors like acetone, chloroform etc. on the sensors responses to the DMMP was negligible. However, the response magnitude of the sensors to the organic vapors was increased as their concentrations were increased. Therefore, the presence of high concentrations of the mentioned organic vapors could interfere with the discrimination of DMMP vapor. In order to examine this effect, the responses of the RGO-based sensors to some vapors including acetone, chloroform, ethanol methanol, acetonitrile, diethyl ether, n-hexane and dichloromethane were recorded at their concentration being 50 times higher than DMMP and then the obtained data matrix (23 × 2) was analyzed by PCA for recognition purpose. Fig. 11 shows the obtained result as the score plot (in PC1 –PC2 plane). As can be seen, the DMMP related signals are classified again in a distinct section of the plane which is considerably far from other vapors classes. The data of the newly added vapors is situated near the TEA class. However, magnification of the related zone in the plane (as shown in Fig. 11 (left)) reveals that even TEA class can be discriminated from the tested organic vapors. These results suggest that RGO-based sensor array, composed of only RGO-B and RGO-H, can be effectively utilized for the classification of DMMP from TEA and AA, as the compounds creating high signals in the sensor array as well as from chloroform, acetone, ethanol, methanol, dichlorometane, n-hexane, diethyl etaniline and acetonitrile, as the compounds producing low signals in the sensor array.
4. Conclusion It was demonstrated that the kind of reducing agent, used for the conversion of GO to RGO, had crucial effect on the DMMP sensitivity of the related RGO-based sensor. It was also shown that the GO, treated with reducing agent of ascorbic acid and sodium borohydride, responded to DMMP very better than that treated with hydrazine hydrate. This phenomenon was ascribed to the negative
Fig. 11. the score plot in the PC1 –PC2 plane, obtained by performing of PCA on the signals of two RGO-based sensor array (including RGO-B and RGO-H) to DMMP, AA, TEA and eight kinds of organic vapors (I) the magnified zone of the score plot of (I), (II); DMMP, AA, TEA were tested at five concentrations of 5, 10, 20, 30, 40 ppm; each of as called VOCs was tested at concentration of 500 ppm.
370
T. Alizadeh, L.H. Soltani / Sensors and Actuators B 234 (2016) 361–370
effect of nitrogen atom, inserted in the RGO when using hydrazine hydrate as reducing agent. Nitrogen atom seems to act as n-dopant agent and destroy the p-type semiconducting characteristic of the corresponding RGO. It was shown that an electronic nose, composed of two RGO-B and RGO-H, was capable of discrimination of DMMP from wide range of the organic vapors including either those with great signals in the sensor array or those creating small responses in sensor array. Although, our aim in this study was to develop a sensor array capable of discrimination of DMMP from other vapors, but, the designed sensor array was found to be capable of discrimination of some other vapors from each other. References [1] X. Du, Z. Ying, Y. Jiang, Z. Liu, T. Yang, G. Xie, Synthesis and evaluation of a new polysiloxane as SAW sensor coatings for DMMP detection, Sens. Actuators B 134 (2008) 409–413. [2] T. Alizadeh, S. Zeynali, Electronic nose based on the polymer coated SAW sensors array for the warfare agent simulants classification, Sens. Actuators B 129 (2008) 412–423. [3] N.B. Munro, S.S. Talmage, G.D. Griffin, L.C. Waters, A.P. Watson, J.F. King, V. Hauschild, Environ. Health Perspect. 107 (1999) 933–974. [4] S.M. Somani, Chemical Warfare Agents, Academic Press, London, 1992. [5] R.M. Black, R.J. Clarke, R.W. Read, M.T.J. Reid, Application of gas chromatography-mass spectrometry and gas chromatography-tandem mass spectrometry to the analysis of chemical warfare samples, found to contain residues of the nerve agent sarin, sulphur mustard and their degradation products, J. Chromatogr. A 662 (1994) 301–321. [6] H. Nozaki, N. Aikawa, Sarin poisoning in Tokyo subway, Lancet 345 (1995) 1446–1447. [7] D.B. Kim, B. Gweon, S.Y. Moon, W. Choe, Decontamination of the chemical warfare agent simulant dimethyl methylphosphonate by means of large-area low-temperature atmospheric pressure plasma, Curr. Appl. Phys. 9 (2009) 1093–1096. [8] C. Leea, H.Y. Choi, S.J. Lee, W.S. Lee, J.S. Huh, D.D. Lee, J.C. Kim, The development of SnO2-based recoverable gas sensors for the detection of DMMP, Sens. Actuators B 137 (2009) 239–245. [9] M. Yang, H.C. Kim, S.H. Hong, assenting behavior of ZnO-coated single-wall carbon nanotube network sensors, Mater. Lett. 89 (2012) 312–315. [10] S.M. Kanan, A. Wagh, B.L. Jensen, C.P. Tripp, Dual WO3 based sensors to selectively detect DMMP in the presence of alcohols, Talanta 72 (2007) 401–407. [11] I. Voiculescu, M.E. Zaghlou, R.A. Mcgil, E.J. Houser, K. Gary, Electrostatically actuated resonant microcantilever beam in cmos technology for the detection of chemical weapons, IEEE Sens. J. 5 (2005) 641–647. [12] C. Karnati, H.W. Du, H.F. Ji, X.H. Xu, Y. Lvov, A. Mulchandani, P. Mulchandani, W. Chen, Organophosphorus hydrolase multilayer modified microcantilevers for organophosphorus detection, Biosens. Bioelectron. 22 (2007) 2636–2642. [13] Z. Ying, Y. Jiang, X. Du, G. Xie, J. Yu, H. Wang, PVDF coated quartz crystal microbalance sensor for DMMP vapor detection, Sens. Actuators B 125 (2007) 167–172. [14] C. Tasaltin, I. Gurol, M. Harbeck, E. Musluoglu, V. Ahsen, Z.Z. Ozturk, Synthesis and DMMP sensing properties of fluoroalkyloxy and fluoroaryloxy substituted phthalocyanines in acoustic sensors, Sens. Actuators B 150 (2010) 781–787. [15] R.A. McGill, V.K. Nguyen, R. Chung, R.E. Shaffer, D. DiLella, J.L. Stepnowski, T.E. Mlsna, D.L. Venezky, D. Dominguez, A nose for toxic gases, Sens. Actuators B 65 (2000) 10–13. [16] Y. Wang, Z. Zhou, Z. Yang, X. Chen, D. Xu, Y. Zhang, Gas sensors based on deposited single-walled carbon nanotube networks for DMMP detection, Nanotechnology 20 (2009) 345502–345510. [17] Y. Wang, Z. Yang, Z. Hou, D. Xu, L. Wei, E.S.W. Kong, Y. Zhang, Flexible gas sensors with assembled carbon nanotube thin films for DMMP vapor detection, Sens. Actuators B 150 (2010) 708–714. [18] S. Mubeen, M. Lai, T. Zhang, J.H. Lim, A. Mulchandani, M.A. Deshussesd, N.V. Myung, Hybrid tin oxide-SWNT nanostructures based gas sensor, Electrochim. Acta 92 (2013) 484–490.
[19] N. Hu, Y. Wang, J. Chai, R. Gao, Z. Yang, E.S.W. Kong, Y. Zhang, Gas sensor based on p-phenylenediamine reduced graphene oxide, Sens. Actuators B 163 (2012) 107–114. [20] T. Alizadeh, L.H. Soltani, Graphene/poly(methyl methacrylate) chemiresistor sensor for formaldehyde odor sensing, J. Hazard. Mater. 248–249 (248) (2013) 401–406. [21] T. Alizadeh, L.H. Soltani, Graphene/graphite/molecularly imprinted polymer nanocomposite as the highly selective gas sensor for nitrobenzene vapor recognition, J. Environ. Chem. Eng. 2 (2014) 1514–1526. [22] A.J. Ricco, R.M. Crooks, G.C. Osbourn, Surface acoustic wave chemical sensor arrays: new chemically sensitive interfaces combined with novel cluster analysis to detect volatile organic compounds and mixtures, Acc. Chem. Res. 31 (1998) 289–296. [23] T. Alizadeh, Chemiresistor sensors array optimization by using the method of coupled statistical techniques and its application as an electronic nose for some organic vapors recognition, Sens. Actuators B 143 (2010) 740–749. [24] V. Dua, S.P. Asurwade, S. Ammu, S.R. Agnihotra, S. Jain, K.E. Roberts, S. Park, R.S. Ruoff, S.K. Manohar, All-organic vapor sensor using inkjet-printed reduced graphene oxide, Angew. Chem. Int. Ed. 49 (2010) 2154–2157. [25] X. Huang, N. Hu, L. Zhang, L. Wei, H. Wei, Y. Zhang, The NH3 sensing properties of gas sensors based on aniline reduced graphene oxide, Synth. Met. 185–186 (2013) 25–30. [26] S. William, J.R. Hummers, R. Offeman, Preparation of graphitic oxide, J. Am.Chem. Soc. 80 (1958), 1339–1339. [27] T. Alizadeh, S. Mirzagholipur, A Nafion-free non-enzymatic amperometric glucose sensor based oncopper oxide nanoparticles-graphene nanocomposite, Sens. Actuators B 198 (2014) 438–447. [28] S. Pei, H.M. Cheng, The reduction of graphene oxide, Carbon 50 (2012) 3210–3228. [29] S. Park, R.S. Ruoff, Chemical methods for the production of graphenes, Nat. Nanotechnol. 4 (2009) 217–224. [30] J.D. Fowler, M.J. Allen, V.C. Tung, Y. Yang, R.B. Kaner, B.H. Weiller, Practical chemical sensors from chemically derived graphene, ACS Nano 3 (2009) 301–306. [31] S. Park, J. An, J.R. Potts, A. Velamakanni, S. Murali, R.S. Ruoff, Hydrazine-reduction of graphite- and graphene oxide, Carbon 449 (2011) 3019–3023. [32] H.J. Shin, K.K. Kim, A. Benaya, S.M. Yoon, H.K. Park, I.S. Jung, M.H. Jin, H.K. Jeong, J.M. Kim, J.Y. Choi, Y.H. Lee, Efficient reduction of graphite oxide by sodium borohydride and its effect on electrical conductance, Adv. Funct. Mater. 19 (2009) 1987–1992. [33] C.K. Chua, M. Pumera, Reduction of graphene oxide with substituted borohydrides, J. Mater. Chem. A 1 (2013) 1892–1898. [34] F. Yavari, Z. Chen, A.V. Thomas, W. Ren, H.M. Cheng, N. Koratkar, High sensitivity gas detection using a macroscopic three-dimensional graphene foam network, Sci. Rep. 1 (2011) 166–171.
Biographies Taher Alizadeh received the BS degree from the University of Guilan, Rasht, Iran, in 1999. He received MS degree from the University of Tabriz, Iran, in 2002 and PhD degree from the University of Tehran, Iran, in 2006. He was with University of Mohaghegh Ardabili, from September 2006 to October 2013. Now, he is continuing his research activities as an associated professor in Department of Analytical Chemistry, Faculty of Chemistry, University Collage of Science, University of Tehran, Iran. His current research interests include the development of chemical sensors and electronic nose based on new materials, Electrochemistry and molecular/Ionic imprinting technology. Leyla Hamed Soltani received the BS degree from the University of Mohaghegh Ardabili in 2010. She received MS degree in analytical chemistry from Department of Applied Chemistry, University of Mohaghegh Ardabili, Ardabil, Iran, in 2013.