Nanoporous carbon-composites as gas sensors: Importance of the specific adsorption forces for ammonia sensing mechanism

Carbon 121 (2017) 114e126

Contents lists available at ScienceDirect

Carbon journal homepage: www.elsevier.com/locate/carbon

Nanoporous carbon-composites as gas sensors: Importance of the specific adsorption forces for ammonia sensing mechanism Nikolina A. Travlou a, b, Teresa J. Bandosz a, b, * a b

Department of Chemistry, The City College of New York, New York, NY, 10031, USA Ph.D. Program in Chemistry, The Graduate Center of the City University of New York, New York, NY, 10016, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 7 April 2017 Received in revised form 12 May 2017 Accepted 22 May 2017 Available online 24 May 2017

Wood-based/polymer-derived nanoporous carbon composites were tested as ammonia sensors for gas concentrations from 45 to 500 ppm. A synergy between the composite components affected their electrical response. Incorporation of only a 1% polymer-derived carbon phase was sufficient to convert the conduction type of the carbon matrix and to increase the sensitivity. An increase in the amount of the polymer-derived carbon phase (10 and 25%) in the composites decreased the electrical signals. This was linked to a decrease in the porosity and thus to a decrease in the contribution of physical adsorption to a sensing mechanism. The reversible sensing of the carbon-coated chips was governed by weak physical interactions of ammonia with surface functional groups and by a charge transport through ionic conductivity (NHþ 4 ). The results suggested that the nature of the weak interactions rather than the electronic properties of the carbon matrix is responsible for the electrical performance in gas sensing. Based on the results obtained, a priority order of the importance of various surface functional groups for the ammonia sensing capability of nanoporous carbons is proposed for the first time. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Ammonia gas sensors Nanoporous carbons Specific adsorption forces Surface chemistry Conduction type

1. Introduction The presence of ammonia (NH3) in the environment, its usage in various industrial processes and its toxicity create a need for highsensitivity sensors that work at low gas concentrations [1]. According to the Occupational Safety and Health Administration (OSHA), NH3 has a permissible exposure limit of 50 ppm and a maximum short-term exposure tolerance of 500 ppm [2]. Various polymers and metal oxides (WO3, ZnO, SnO2, Cr2O3, etc.) have been used as NH3 gas solid-state sensors [3e5]. The latter are sensitive at high temperatures (200e500  C). However, high power consumption makes their practical application disadvantageous. In recent years, carbon-based sensors (carbon nanotubes, graphene, and reduced graphite oxide) [6e12] have been of particular importance because of their relatively low cost, semiconducting electrical properties, potential for modifications and ability to operate at room temperature. To improve their selectivity additional functionalization with metals [13,14], metal oxides [15,16] and conducting polymers [8,9,17] is usually required. For instance, modification of

* Corresponding author. Department of Chemistry, The City College of New York, New York, NY, 10031, USA. E-mail address: [email protected] (T.J. Bandosz). http://dx.doi.org/10.1016/j.carbon.2017.05.081 0008-6223/© 2017 Elsevier Ltd. All rights reserved.

MWCNT with polyaniline led to 4 times higher sensitivity (32%) when exposed to10 ppm of NH3 than that on pure carboxylated MWCNT (7.2%) [8]. Functionalization of CNTs with metals was also shown to increase the sensitivity. Abdelhalim and co-workers reported 92% sensitivity (100 ppm of NH3) for CNTs functionalized with Au nanoparticles (NPs) [18]. Polypyrroleegraphene nanocomposites decorated with titania NPs were also found to be very sensitive (~100%) to 50 ppm of NH3 [17]. Even though such functionalization processes may improve the sensors' response and selectivity, they drastically increase production costs. Activated carbons, on the other hand, have several advantages over the above-mentioned carbon-based materials when used as toxic gas sensors. We recently showed that the performance of nanoporous carbon-coated chips is comparable to those reported in the literature for modified graphene or CNT-based chips [19e22]. This is related to their high surface areas and pore volumes. These features strongly favor the adsorption and retention of gas molecules in the carbon matrix, leading to a high sensitivity. Furthermore, a diversity of the surface chemistry of nanoporous carbons determines the degree of surface reactivity, which is an important feature that enhances the sensor selectivity towards specific species without a need for additional high-cost doping processes (metal/metal oxides).

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

An introduction of heteroatoms (such as O, S or N) to a carbon matrix is an effective way to alter its chemistry and electronic properties [19,21,23e29]. An improved electrochemical performance is related to doping-induced charge redistributions around the heteroatoms (dopants) [30]. It was previously shown that by applying a specific surface chemical modification a predominant type of charge carriers (electrons or holes) can be targeted [20,21]. This enables one to control the extent of the electronic properties of nanoporous carbons and also the nature of specific interactions when electron withdrawing or donating gas molecules are the sensor's targets. It has been shown that oxidation of O-containing wood-based activated carbon with HNO3 caused the conversion of the conduction type from predominantly ep to predominantly en [20]. The former conduction type was attributed mainly to the presence of specific O-functional groups such as carboxyl, carbonyl, ether, epoxy and sp3-hydroxyl groups in the carbon matrix [31]. On the other hand, the -n conduction type was related to an introduction of electron acceptor defects (NOx groups). In other studies, polymer-derived nanoporous carbons doped with only S [22] or codoped with both S and N [19] were tested as NH3 gas sensors. The results showed that a surface chemical heterogeneity played a crucial role in the chips' electrical performance. A chemical synergy between S and N- heteroatoms in the form of specific groups led to an enhanced chip sensitivity compared to the carbons doped solely with S [19]. Both oxidized and reduced S-species (such as sulfoxides, sulfones and thiols, respectively) were involved in the sensing mechanism by participating in weak interactions with ammonia. The electronic properties of carbon-based materials are known to be governed by nitrogen incorporated to the sp2 carbon lattice in specific configurations [32,33]. DFT calculations indicated that N with a coordination number of 3 induces n-type doping, while

pyridinic and pyrrolic-type nitrogen induces p-type doping [33e35]. The results of testing the NH3 sensing capability of nanoporous carbons doped with solely N [19,21] suggested that surface acidity, by enhancing the affinity towards NH3 adsorption, is also an important factor determining the sensitivity/selectivity of the chips. The objective of this paper is to examine the NH3 sensing capability of wood-based/polymer-derived nanoporous carbon composites. We investigate whether or not, and if so, to which extent, a synergistic effect of the composite formation on the carbon surface features can affect the electrical response of the sensors. Even though we have previously addressed the role of porosity and surface chemistry on sensing, the results did not lead to a clear answer which surface feature (chemistry or porosity) plays a predominant role. The relationship between the sensitivity of the chips and the volume of micro/ultramicropores was found to be either direct or indirect. The contribution of the specific configuration/ arrangement of S and N heteroatoms on sensing is still hardly distinguishable and needs further attention. An additional goal of this study is to investigate effective means of increasing the sensitivity of the detection devices based on the adsorption

115

phenomenon in the pore system of carbons.

2. Experimental 2.1. Preparation of materials Two polymers and wood-based commercial activated carbon (BAX-1500; (MeadWestvaco)) (apparent density, g/cc: 0.26e0.35, BWC, g/100 mL: 14.4 min, moisture, wt%: 10 max, pellet strength: 50 min, mean particle diameter, mm: 2.10 min) were used for the composite preparation. Poly (sodium 4-styrene sulfonate) (SigmaAldrich, MW~70,000, Relative density 0.801 g/mL at 25  C) and poly (4-ammonium styrene-sulfonic acid) ammonium salt solution (Sigma-Aldrich, Mw ~200,000, 30 wt % in H2O, solid content: 28.0e32.0%) were used as the polymer precursors. To prepare the composites with a polymer-derived carbon content of 1, 5, 10 and 20% in the final material, BAX was first mixed with the predetermined amounts of two polymers dissolved in water. After their drying at 120  C for 24 h, the mixtures were carbonized at 800  C for 40 min. The flow of N2 was 300 mL/min with a heating rate of 50  C/min. To modify their structure and chemistry, the soobtained carbons were oxidized in air at 350  C for 3 h. The resulting materials are referred to as BAX-CS-X% or BAX-CSN-X% where X-represents 1, 5, 10 and 25% of the polymer-derived carbon phase. CS refers to the sulfur containing carbon that was obtained from poly (sodium 4-styrene sulfonate), and CSN refers to the sulfur and nitrogen dual-doped carbon obtained from poly (4ammonium styrene-sulfonic acid). The same treatment (heating at 800  C followed by air oxidation) was also applied to BAX carbon, and the resulting material is referred to as BAX-AO. A brief description of the treatments applied is illustrated below:

2.2. Characterization of the materials and electrochemical measurements The initial and exhausted materials after NH3 exposure were extensively characterized using XPS analysis, sorption of N2, potentiometric titration and SEM. The suffix eED is added to the name of the exhausted samples after NH3 exposure in dry conditions. The reversible sensing was carried out by exposing the carbon chips to NH3 concentrations 45e500 ppm in dry air, with a total flow rate of 500 mL/min, at room temperature, and applying the bias potential of 1 V. The changes in the resistance were monitored and are discussed in terms of a normalized resistance, (DR/Ro) ¼ (Rt-Ro)/Ro. For the sensing test, thin film gold interdigitated finger electrodes on 8  8 mm alumina substrate, with 50 micron lines/spaces, and without insulation layer were used. The electrodes were purchased from Electronic Design Center, Case Western Reserve University, Cleveland, OH. The sensing chips were coated with the active materials, and placed into a closed home-made gas chamber (20 cm3). To establish dry conditions, before ammonia dilution, the air passed through a column packed with Drierite. To establish humid

116

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

conditions, air was passing through water. Details on the preparation, characterization, and electrochemical measurements are provided in the Supporting Information (SI). Fig. S1 of Supplementary Information illustrates the experimental set-up. 3. Results and discussion 3.1. Sensing response and surface characterization A chamber containing the chip was initially purged with 500 ppm of NH3 until the electrical signal reached a plateau. This is illustrated in Fig. S2 of Supplementary Information for the three best performing chips (BAX-AO, BAX-CS-1% and BAX-CSN-1%). This signal stabilization indicates a complete passivation of the carbon's active sites by the NH3 molecules. This process involves both physical and chemical adsorption. The chips were subsequently recovered in air until a stable signal was reached again. During this air purging stage, physically adsorbed and thus weakly retained gas molecules were removed from the carbon's surface. Assuming that physisorption is responsible for the reversibility of the electrical signal, based on the % of the sensor response change at this stabilization stage the predominant interaction type between NH3 and our carbons' surface (chemisorption or physisorption) could be estimated. The % recovery of the normalized resistance upon the air-purging stage (Fig. S2) suggests that chemisorption is more pronounced for BAX-AO, while physisorption predominates for BAX-CS-1% and BAX-CSN-1%. The reversible and repeatable sensing operation of our chips

was examined by exposing them to various NH3 concentrations (45e500 ppm). They were selected following the OSHA permissible exposure guidelines [2]. The electrical responses of the chips are illustrated in Fig. 1A and B. In this stage the sensing is considered as fully reversible and is based on weak reversible interactions between NH3 and the carbons' surface. Linear relationships between the responses and the NH3 concentrations for two best performing carbon samples from both composite groups (BAX-CS-1% and BAXCSN-1%) are presented in Fig. 1C. The response time of both carbon series upon exposure to the highest gas concentration was defined as 40 s, and the recovery time - as 17 min. A relatively long recovery time in our case is likely linked to the developed micro/ultramicroporosity of the carbons. Even though this feature is responsible for the good sensor sensitivity, during the air purging stage, it is also responsible for the slow desorption of the gas molecules. The recovery time could be shortened by either decreasing the time of exposure of the chips to the target gas or by purging the system using a high air flow. This however would have a negative effect on both the sensitivity and stability/repeatability of the sensors. This is because an arbitrary decrease in the recovery time, will result in less marked changes in the normalized resistance. Using a high air flow for purging the system during the sensor recovery stage would cause instabilities to the flow rate. This would bring complexity to stable/reproducible sensor operation. The response times of the chips at each ammonia concentration were also analyzed, and they are presented in Fig. S3(A) of the SI. A decrease in the ammonia concentrations, slightly slows the adsorption kinetics, which leads to an increased response time. Furthermore, the comparison of the

Fig. 1. (A, B): Response curves for the BAX-CS and BAX-CSN composite series. (C): Dependence of DR/Ro (%) of the two best performing materials from each composite series, on various NH3 concentrations. (A colour version of this figure can be viewed online.)

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

chips' sensing time for two different samples series (BAX-CS and BAX-CSN) Fig. S3(B)) shows small but not significant differences in the sensing cycles. Such differences are related to differences in the surface chemical and structural features of the two composite series, which affect their reactivity with ammonia in different ways. The response curves of the carbons tested using the actual sensor resistance were also analyzed, and are illustrated in Fig. S4 of Supplementary Information. Important differences in the initial resistivities of the chips were found. Thus the resistivities of BAXCS-1%, BAX-CS-5%, BAX-CS-10% and BAX-CS-20% are 2.5 kU, 0.6 kU, 64U and 90 U,respectively. For the BAX-CSN series the resistivity varies from 0.2 to 1 kU (Fig. S4(C). The stability of the best performing carbon composite (BAX-CS-1%) was tested by exposing it to repeatable NH3 (500 ppm)/air cycles. The data collected in Fig. S5 of the Stability of the sensor response. Fig. 2 compares the absolute values of the normalized resistance change upon exposure to 500 ppm of NH3 for all carbon chips addressed in this study. As seen, BAX-AO exhibits an 18% change in the normalized resistance when exposed to 500 ppm. Interestingly, the modification of BAX-AO by introducing only 1% of the S-containing polymer-derived carbon phase led to a 21% improvement in the chip sensitivity. On the other hand, adding 1% of the S,Ncontaining polymer-derived carbon phase caused a 10% sensitivity decrease. Upon the addition of more polymer-based carbons, more pronounced decreases in sensitivity were recorded. The comparison of the sensitivity of the BAX-CS series to that of the BAX-CSN series shows that the composites consisting of 5% of the polymer-derived carbon phase (BAX-CS-5% and BAX-CSN-5%) exhibit the biggest variation in this parameter. While BAX-CS shows 17% change in the normalized resistance upon ammonia exposure, for BAX-CSN this change is 12%. The sensitivity of the carbon-chips upon exposure to 500 ppm of ammonia are 18, 21, 17, 10, and 7% for BAX-O, BAX-CS-1%, BAX-CS-5%, BAX-CS-10%, and BAC-CS-20%, respectively, and 18, 16, 12, 13, and 10% for BAX-O, BAX-CSN-1%, BAX-CSN-5%, BAX-CSN-10%, and BAC-CSN-20%, respectively. To comprehend the observed trends in the sensitivity of the carbons tested and to understand whether or not the synergistic

Fig. 2. Comparison of the normalized resistance changes of all carbon samples tested upon exposure to 500 ppm of NH3. (A colour version of this figure can be viewed online.)

117

effect on porosity and surface chemistry affects their electrical response, NH3 breakthrough tests were carried out [36]. The surface structural and chemical features of the initial and exhausted samples were extensively characterized. To clearly see the trends, we have arbitrarily chosen to examine only the composites that show the highest electrical sensitivity and biggest differences in this parameter. Thus the same counterparts of each series (BAX-AO, BAX-CS-1% with BAX-CS-5%, and BAX-CSN-1% with BAX-CSN-5%) are analyzed in details. The parameters of the porous structure calculated from the N2 adsorption isotherms and the pore size distributions (PSD) of the carbons tested are presented in Table S1 and Fig. 3, respectively. Analysis of the data indicates that there is no direct relationship between their sensitivity order and the surfaces areas. This is expected since only the smallest pores should be involved in ammonia adsorption at ambient conditions. To comprehend the role of porosity in sensing, the porosity data for the initial composites with 10 and 25% of the polymer derived carbon phase (before ammonia exposure) are also, exceptionally, included in Table S1. All samples exhibit similar volumes of mesopores, micropores, and especially ultramicropores (V < 1 nm). Only for the composite with 10 and 25% of the polymer derived carbon phase the unaltered or slightly increased volumes of ultramicropores (volume < 0.7 nm) and decreased volumes of mesopores are found compared to these features for BAX. The former ones are known to be the most active in NH3 adsorption [37], and based on our previous studies they are expected to play the most important role in sensing [19,20,22]. After NH3 exposure, BAX-AO shows an ~35% decrease in the volume of ultramicropores, which is linked to the contribution of NH3 reactive adsorption. This is also supported by the change in the electrical response of this carbon during its initial stabilization (Fig. S2). For the composites of BAX-AO with 1% and 5% of the polymer-derived carbon phase, on the other hand, the unaltered volumes of meso-, micro- and ultramicropores are linked to physisorption being a predominant surface interaction type. This is also in agreement with the results illustrated in Fig. S2. The PSDs (Fig. 3) indicate that the volumes of ultramicropores (V < 0.7 nm) increased for the composites with 10 and 25% of the polymer derived carbon phase. On the other hand, the volumes of micropores and mesopores show opposite trends. For the composites with 1 and 5% of the polymer derived carbon phase, the volumes of all pores remain practically unaltered. The comparison of the hypothetical and experimental surface areas, volumes of mesopores, micropores and ultramicropores (V < 0.7 nm) for the materials studied is presented in Fig. S6 and discussed in details in Supplementary Information. The results suggest that the higher than the hypothetical volumes of microand ultramicropores must be a consequence of the composite formation and interactions of the polymer with the BAX-AO carbon functional groups. BAX-CS-1% that exhibits the highest electrical sensitivity, also exhibits the greatest deviation from the expected hypothetical volume of ultramicropores. It has also the highest volume of these pores. Thus its superior electrical performance might be related to the developed ultramicroporosity that favors the adsorption and retention of the NH3 molecules in the carbon matrix [37]. SEM images of the composites with the highest percentage (25%) of the polymer-derived carbon show two different phases originating either from the polymer (smooth) or commercial carbon (particles) (Fig. 4). The content of elements on the surface in atomic % and the results of the deconvolution of the C1s, O1s, N1s, and S 2p core energy level spectra are summarized in Fig. 5. Further details are provided in Tables S2eS6. Prior to the XPS analysis and N2 adsorption a high vacuum was applied and weakly adsorbed NH3 molecules were

118

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

Fig. 3. Pore size distributions for the materials studied. (A colour version of this figure can be viewed online.)

Fig. 4. SEM images of BAX-CS-25% and BAX-CSN-25%.

removed. Thus, using these two characterization methods we are able to examine only the irreversible surface chemistry changes caused by the reactive NH3 adsorption. Considering that, the active surface sites chemically reacted with ammonia during the chip stabilization, the reversible sensing process strongly depends on the new surface chemical features of the exhausted samples. Thus, the data obtained for the samples after NH3 exposure, by both XPS and N2 adsorption, are of a great importance to comprehend the reversible sensing mechanism. While BAX-AO has the highest content of oxygen, its amount in other samples tested is smaller and practically the same (Fig. 5A). Both BAX-CSN-1% and BAX-CSN-5% have N on their surfaces. The latter sample and BAX-CNS-5% also contain small amounts of S. The deconvolution of the S 2p core energy level spectra indicates that S is incorporated to the carbon matrix mainly in thiols and bisulfides. In BAX-CS-1% and BAX-CSN-1% S is not detectable. After NH3

adsorption the content of N increased in all samples owing to its chemical reactions with the surface functional groups (Figs. 5 and 6). The content of S does not show any marked differences after NH3 exposure which indicates that even assuming that S is involved in chemical reactions, it remains on the carbons' surface in the form of stable surface compounds. The deconvolution of the C 1s, O 1s, and N 1s core energy level spectra for the best performing carbons from each composite series (BAX-CS-1% and BAX-CSN-1%), along with the deconvolution of the S 2p core energy level spectra of the composites with the 5% of the polymer drived carbon phase (BAXCS-5% and BAX-CSN-5%), before and after ammonia exposure, are presented in Figs. 6 and 7. The deconvolution of the S 2p core energy level spectra provides important information about S-type groups that play the most important role in the sensing mechanism of our carbon composites (Figs. 5 and 6). For BAX-CSN-5%, NH3 exposure decreased the

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

119

Fig. 5. Content of elements and surface contributions of the specific carbon, oxygen, nitrogen and sulfur groups. (A colour version of this figure can be viewed online.)

contribution of oxidized S-species (sulfoxides and sulfones) and increased the contribution of reduced S-forms (thiols, bisulfides). This is linked to the electron donating nature of NH3, and is in

contrast with our previously reported results for S- and N- codoped polymer-derived carbons where the opposite trend was found [19]. The reaction of thioesters with NH3, where the former is

120

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

Fig. 6. C 1s, O 1s, and N 1s core energy level spectra for BAX-CS-1% and BAX-CSN-1% before and after ammonia exposure. (A colour version of this figure can be viewed online.)

converted into the corresponding thiol with a simultaneous formation of an amide group [38], is a plausible explanation of an increase in the relative concentration of the reduced S-species. This is in agreement with the deconvolutions of the S 2p and N 1s core energy level spectra, which indicate an increase in the contribution of thiols and amides. For BAX-CS-5%, on the other hand, the

contribution of the reduced S-forms decreased and an increase in sulfoxides and sulfones is found. This trend is once again opposite to the one we reported in our previous study for the S-containing polymer-derived carbon [22]. Comparison of the surfaces of the exhausted BAX-CS-5% and BAX-CSN-5% indicates that the former composite that exhibits a higher sensitivity, also exhibits a lower

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

121

Fig. 7. S 2p core energy level spectra for BAX-CS-5% and BAX-CSN-5% before and after ammonia exposure. (A colour version of this figure can be viewed online.)

contribution of the reduced S-species. The contribution of the oxidized S-species is similar in these two samples. This suggests that the oxidized sulfur species, rather than the reduced ones, play an important role in the sensing mechanism. The deconvolutions of the N 1s core energy levels indicate that exposure to NH3 increased the contribution of N in NHþ 4 (ammonium ions). Such ions can be formed in the acid-base reactions of NH3 with acidic O-containing groups (carboxyl and sulfonic). Carboxylic acids were detected on the surface of all carbons samples, and sulfonic acids on the surface of the composites with 5% of the polymer derived carbon phase (BAX-CS-5% and BAX-CSN-5%). BAXCS-5% and BAX-CSN-5% show the lowest contribution of NHþ 4 and the highest contribution of amines. This might be related to a greater involvement of carbonyl groups (ketones and aldehydes) rather than carboxyl and sulfonic in reactions with NH3. The amine formation is also supported by the deconvolutions of the N 1s and O 1s core energy levels. The N 1s core energy level spectra also show that the initial

BAX-CSN-1% sample has a higher contribution of amines and amides compared to those in BAX-CSN-5%. After NH3 exposure however, the increases in the contributions of amines and amides were greater for BAX-CSN-5% than for BAX-CSN-1%. Their appearance is likely related to the involvement of carboxylic, epoxy, thioester and anhydride groups in reactions with NH3. The reactivity of anhydrides with NH3 is demonstrated by an increase in the contribution of -NH2 and carboxyls on the surface of the exhausted samples. The distributions of the pKa values of the species present on the initial and exhausted carbon surfaces along with the surface pH values are discussed in details in Supplementary Information. (Fig. S7). Both XPS and potentiometric titration results show that BAX-AO, BAX-CS-1% and BAX-CSN-1%, which are the best performing carbons, have the highest amounts of NHþ 4 species as a result of the reactive NH3 adsorption on their surfaces (Fig. 6). In the case of BAX-AO this reaction considerably decreased the ultramicropore volume.

122

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

Fig. 8. The Mott-Schottky plots.

3.2. Sensing mechanism A conduction type of our materials was evaluated using the MottSchottky approach [39]. For BAX-AO (Fig. 8A), linear trends with similar positive and negative slopes indicate the coexistence of both n- and p- conduction types, respectively, and thus both negatively charged electrons (e-) and positively charge holes (hþ) act as charge carriers. O-containing electron-withdrawing groups, such as sp3-bonded hydroxyl, carboxyl, carbonyl, ethers, and epoxides would likely induce a p-type conductivity, while O-containing electron-donating groups such as sp2 -bonded hydroxyl,

ether and epoxide groups induce the n-conduction type [31]. For both composite series, the introduction of only 1% of the polymer-derived carbon was sufficient to alter the surface chemical features at such extent that the p-conduction type became predominant (Fig. 8B and D). Indeed, XPS analysis indicated that for BAX-CS-1% and BAX-CSN-1% the contribution of sp3-OH, ethers, and epoxy groups increased in comparison to those BAX-AO. On the surface of BAX-CSN-1% amides and quaternary N-species were also detected. Such groups due to electron-withdrawing properties would induce a p-type conductivity. In BAX-CS-5% and BAX-CSN5%, -SO3H groups would also act as p-type doping impurities

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

because of their strong electron withdrawing nature [40]. The MottSchottky plots for the BAX-CS and BAX-CSN series (Fig. 8) suggest that the former ones exhibits a slightly higher contribution of a typical p-conduction type than the latter one. This could be explained by the high contribution of amines and eNHCOR groups on the surface of the BAX-CSN composites. They are considered as n-type dopants and are responsible for a more pronounced n-type character. Based on the predominant conduction type of our materials (ptype), exposure to NH3 was expected to cause an increase in the normalized resistance due to the depletion of the positively charged hþ by e- donated by NH3. This is not, however, the trend that we observe here. In our previous studies, we have shown that a decrease instead of an expected increase in the normalized resistance of p-type materials suggests a partial oxidation of NH3 to NO2 by superoxide ions (O 2 ) [19]. These ions are usually generated on the surface of heteroatom-containing nanoporous carbons. NO2, owing to its electron withdrawing properties would cause an increase in the concentration of hþ, resulting in a decrease in the normalized resistance. Here, the formation of O 2 is not supported by XPS, which would otherwise show a contribution at a high binding energy (536.9 eV) on the O 1s core energy level spectrum (Table S4). Moreover, the N 1s core energy level spectrum lacks a contribution at ~402.5 eV, corresponding to NOx, which would be formed as a result of the partial NH3 oxidation. The lack of these features suggests that the nature of the interactions between the surface functional groups and NH3, rather than the electronic configuration of the surface functional groups (electron donating or withdrawing), which determines the type of the charge carriers, is responsible for the electrical response of the carbons. It is the first time that we observed such a phenomenon. The interactions that occur between the surface functional groups and the NH3 molecules are weak in nature and thus are responsible for the reversibility of the electrical signal. They include: a) specific intermolecular forces such as hydrogen bonding between the NH3 molecules and the surface functional groups, b) electrostatic interactions such as dipole-dipole interactions, and c) liquid-liquid interactions due to the presence of quasi-liquid NH3 molecules in the small micropores where functional groups do not exist due to steric effects and thus specific interactions do not take place. The latter interactions, by allowing e- transport across the ultramicropores through charge hopping mechanisms [41], are responsible for a decrease in the normalized resistance upon NH3 exposure. Since the composites with 1, 5 and 10% of the polymer derived carbon phase have similar porosities (Table S1), the trend found in their sensitivity indicates that surface chemistry is predominantly responsible for the differences in their electrical performance. The results reveal that by introducing O, S and N heteroatoms and thus altering the samples' surface chemical features, we are able to control the conduction type and nature of specific interactions with NH3. The comparison of the porosity of the composites with 5% of the polymer derived carbon phase with that of the composites with 10 and 25% of that phase (Table S1 and Fig. 3) reveals that while the ultramicropore volumes of the two latter composites are unaltered or slightly increased compared to former composites (5% of the polymer derived phase), their micro- and especially mesopore volumes decreased. This leads to the following conclusions: a) surface chemistry is a crucial factor determining the sensing response of the carbons when samples of similar porosity are compared; b) even though ultramicropores are the most crucial for NH3 retention and thus govern the sensing capability of nanoporous carbons, mesopores also play an important role. A higher volume of these pores leads to a more NH3 molecules in ultramicropores, and thus to a higher sensitivity. Furthermore,

123

functional groups in mesopores participate in specific interactions with NH3 enhancing e- transport. Based on this, the small mesopore volumes of the 10 and 25% BAX-CS and BAX-CSN composites would be one potential reason for their low sensitivity. Both oxidized and reduced (R-S) S-species participate in the above-mentioned weak specific interactions with NH3, and thus promote e- transport through the hopping mechanisms. However, based on the electrical performances of BAX-CS-5% and BAX-CSN5% (Fig. 1), and on the XPS analysis (Fig. 5), BAX-CSN-5% that exhibits the highest contribution of the reduced S-species, also shows the less marked electrical signal changes compared to those for BAX-CS-5%. This suggests that the oxidized S-forms (sulfones and sulfonic acids) rather than the reduced ones (thiols and bisulfides) play the most important role in the sensing capability of our carbons. Unlike in the case of sulfur doping, where oxidized species are the most important for sensing, oxygen as a heteroatom plays an important role in its various configurations. Highly acidic carboxyl and sulfonic groups, by participating in acid-base interactions with NH3 promote the charge transport through ionic conductivity. The same groups also participate in the hydrogen bonding with NH3 and thus promote e- transport through hopping mechanisms. Carbonyl (ketones, aldehydes or C]O of carboxylic groups) and epoxy groups are also important. They react with NH3 (initial stabilization stage) and -OH, -NH2 and -CONH2 groups are formed. They can then weakly interact with the NH3. Finally, weakly acidic phenols and alcohol groups, which can also weakly interact with NH3, are the second most important O-containing groups. BAX-CS-1% that has the highest contribution of these species also shows the highest sensitivity. BAX-AO on the other hand, which exhibits the second highest sensitivity, shows the highest contribution of carboxyl groups and thus the highest surface acidity. The extent of NH3 adsorption on the carbons' surface strongly depends on the latter feature. We have previously found a direct relationship between the sensitivity of carbons tested as NH3 sensors and their surface acidity [21]. That was related to the presence of specific surface functional groups which enhanced NH3 adsorption. As mentioned above, surface acidity plays an important role in promoting the charge transport through an ionic conductivity. The potentiometric titration results presented here indicate that our best performing samples (BAXCS-1% and BAX-AO), after exposure to NH3, have the largest number of species of pKa 9e10 (assigned to NHþ 4 ) (Fig. S7B). This indicates the crucial role of an ionic conductivity in governing the electrical response of our carbon chips. Based on the XPS results, the BAX-CSN series shows the higher surface contribution of carbonyl groups in ketones, aldehydes and amides than does the AX-CS series. As a result of their reaction with NH3, amines and hydroxyl groups, rather than NHþ 4 are formed. Even though they may contribute to the reversible sensing through the weak interactions with NH3, they do not participate in a charge transport via an ionic conductivity. This partially explains the small signal changes measured on the BAX-CSN composites. The best performing composite, BAX-CSN-1%, has a high contribution of amines and amides in its initial form. After ammonia exposure, the contributions of quaternary-N and protonated amines increased. Considering that the reversible sensing strongly depends on the new surface chemical features of the exhausted/initially exposed samples, the above results indicate that quaternary-N rather than amines and amides are more important for sensing. Based on the electrical performance of the carbon tested and their surface features the order of importance of specific surface groups for the ammonia sensing capability of nanoporous carbons is proposed. It is as follows:

124

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

The higher conductivity of the BAX-CS composite series than that of the BAX-CSN series can be an additional reason for the higher sesnitivity of the former samples. The initial resistivity of the CS and CSN coated chips was found as 17 U and 5.44 kU, respectively [19,22]. Since the resistance of pure BAX-AO is 2.93 kU, it is expected that the incorporation of the more conductive carbonaceous phase leads to enhanced e- transport capabilities, and thus to its higher sesnitivity. The mechanisms and factors that govern the electrical response of nanoporous carbons as ammonia sensors are summarized in Fig. 9. The effect of relative humidity on the NH3 sensing response of two best performing carbon chips from both carbon series (BAXCS-1% and BAX-CSN-1%) was examined. A bar graph in Fig. 10 compares the changes in the normalized resistance for BAX-CS1% and BAX-CSN-1% upon exposure to 500 ppm of ammonia, at various humidity levels. As seen, an enhanced electrical response is measured at RH 50e55%. Such a behavior is expected since not only ammonia but also water molecules contribute to the electrical response of the carbons through the donation of electrons, participation in weak interactions with the surface functional groups, and the ionization process. This mechanism was addressed in details in our previous study, which focused on the effect of humidity on the sensing capability of nanoporous carbons [42]. As more water adsorbs on the carbon surface (65e70% RH), the signal increase becomes suppressed, and the sensitivity is similar to that in dry

conditions. This decrease in the sensor response at higher humidity levels can be explained by filling of the pores by water and the dissolution of NH3 in the water film formed on the carbons' surface and inside the pores. At these conditions the sites on the carbons' surface that would be otherwise available to weakly interact with ammonia and contribute to the reversible sensing are “blocked”, and thus a decrease in the electrical signal is recorded. This sensor behavior is also in agreement with the previously reported results where the sensing in moist air was analyzed in details [42]. A summary of the sensing performance of other carbon-based materials reported in the literature, along with the results on our best performing samples from each carbon composite series, is provided in Table 1. As seen our nanoporous carbon-composites show sensitivities that are comparable to other carbon-based ammonia sensors. Moreover, in some reported cases, even though functionalized processes were applied, aiming to increase the sensitivity of the graphene-based sensors', relatively low resistance changes or even smaller than those observed for our samples, were reported. It is worth to mention that the advantage of our conductometric ammonia gas sensors compared to other electrochemical gas sensors (amperometric, potentiometric) or other types of solid-state gas sensors (catalytic, optical, semiconductors, Field-Effect Transistors (FET)), is not in the low detection limit. It is rather in other features, which are related to the fabrication and operation of

Fig. 9. Schematic illustration of the different factors that govern the electrical response of nanoporous carbons. (A colour version of this figure can be viewed online.)

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

Fig. 10. Changes in the normalized resistance for the best performing carbon chips from both series tested upon exposure to 500 ppm of ammonia, at different humidity levels. (A colour version of this figure can be viewed online.)

125

controlled. When these interactions are weak, porosity is a crucial factor determining the extent of the sensor response. The composite consisting of 1% of the S-containing polymer-derived carbon phase showed the highest sensitivity. More of the polymer derived carbon in the composites (10 and 25%) decreased the porosity and thus the contribution of physical adsorption to the sensing mechanism. This decreased the electrical signal measured. Even though ultramicropores are the most crucial for the extent of NH3 physical adsorption, in this study mesopores were also found as playing a determining role in sensing by promoting NH3 penetration to the pore system. Functional groups located in these pores were also involved in sensing mechanism. The reversible sensing capability of the carbon chips is governed by weak interactions between the surface functional groups and NH3. A charge transport through an ionic conductivity (NHþ 4 ) also determines the electrical response and thus sensitivity of our sensors. Such sensing mechanisms strongly depend on the specific surface chemistry of the carbons. Based on the electrical performance and specific surface chemistry of the samples studied, a priority order of various chemical groups on the NH3 sensing capability of nanoporous carbons has been proposed. The results

Table 1 Comparative table of the sensing performance of various carbon-based ammonia gas sensors. Carbon-based ammonia sensors

Name

NH3 concentration

Response of the sensor (%)

References

BAX-CS1% BAX-CSN-1% sc-SWCNT Graphene-PEDOT:PSS PEDOT:PSS Graphene SWCNTs AuNPs-rGO VA-MWCNTs RGOePANI Graphene-PtNPs GF RGO/3eCuPc SWCNTs-PANI Graphene-AuNPs

Carbon composites in this study

500 ppm

e

Semiconducting single-walled carbon nanotube grapheneepoly(3,4-ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS)

0.03e0.6% 500 ppm

Single-walled carbon nanotubes (-COOH) functionalized gold NPs-reduced graphene oxide Vertically aligned multi-walled carbon nanotubes Reduced graphene oxideepolyaniline hybrid Graphene films decorated with platinum NPs Graphene foam Reduced graphene oxide/copper phthalocyanine Single-walled carbon nanotubes functionalized with polyaniline Graphene decorated with gold nanoparticles

500 ppm 60 ppm 500 ppm 50 ppm 58 ppm 1000 ppm 3200 ppm 100 ppm 58

21 16 7.5e54.6 9.6 4.4 2.4 27.3 2.5 1.9 59.2 12 30 15.4 14.26 8

sensors. They include low fabrication costs, which is lower than for most other gas detection technologies, flexibility in production, simplicity of usage and detection, and excellent reproducibility and accuracy. In our previous study [20] we had reported the selectivity of the initial BAX carbon towards 500 ppm of H2S, which similarly to ammonia is an electron donating gas but of a different chemistry. A very low sensitivity of the sensor was recorded (4% for H2S vs 29.2% for NH3), which was linked to the small H2S adsorption capacity of BAX, due to its surface acidity. Owing to an acidic surface pH (~3) of the composites addressed in this paper, where BAX is the predominant carbonaceous phase, small H2S adsorption capacities are expected to lead to low sensor sensitivities. 4. Conclusions The results obtained show that the synergistic effect on the porosity and surface chemistry between the components of woodbased/polymer-derived nanoporous carbon composites affects the electrical response of these materials as NH3 sensors. By incorporating small amounts of polymer-derived carbon phases, (1 and 5%) the type of the charge carriers and the nature of the specific interactions between NH3 and the surface functional groups can be

[43] [44]

[45] [46] [47] [9] [48] [49] [50] [51] [14]

collected strongly suggest that the nature of the interactions between the surface functional groups and NH3, rather than the electronic nature of the carbon matrix, is responsible for the observed electrical response of the carbon chips tested. The effect of humidity was also examined on two best performing carbon chips from both sample series (BAX-CS-1% and BAX-CSN-1%). 50e55% RH enhanced electrical response, while a 65e70% resulted in the sensitivity similar to that in dry conditions. Consistent differences in the response between the chips tested were measured. Acknowledgements This work was supported by the ARO (Army Research Office) grant W911NF-13-1-0225 and NSF collaborative CBET Grant no. 1133112. Special thanks are addressed to Dr. Mykola Seredych (currently Research Associate II at Drexel University) for his help with XPS measurement and to Christopher Ushay for help with sensing measurements. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.carbon.2017.05.081.

126

N.A. Travlou, T.J. Bandosz / Carbon 121 (2017) 114e126

References [1] B. Timmer, W. Olthuis, A. Van Den Berg, Ammonia sensors and their applicationsda review, Sens. Actuators, B 107 (2005) 666e677. [2] Toxicological Profile for Ammonia, U.S. Department of Health and Human Services, Public Health Service, Agency for Toxic Substances and Disease Registry, 2004. [3] C.S. Rout, M. Hegde, A. Govindaraj, C.N.R. Rao, Ammonia sensors based on metal oxide nanostructures, Nanotechnology 18 (2007) 205504e205512. [4] S.M. Kanan, O.M. El-Kadri, I.A. Abu-Yousef, M.C. Kanan, Semiconducting metal oxide based sensors for selective gas pollutant detection, Sensors 9 (2009) 8158e8196. [5] H. Bai, G. Shi, Gas sensors based on conducting polymers, Sensors 7 (2007) 267e307. [6] R. Ghosh, A. Midya, S. Santra, S.K. Ray, P.K. Guha, Chemically reduced graphene oxide for ammonia detection at room temperature, ACS Appl. Mater. Interfaces 5 (2013) 7599e7603. [7] Y.Y. Wang, L. Zhang, N. Hu, Y.Y. Wang, Y. Zhang, Z. Zhou, Y. Liu, S. Shen, C. Peng, Ammonia gas sensors based on chemically reduced graphene oxide sheets self-assembled on Au electrodes, Nanoscale Res. Lett. 9 (2014) 251e263. [8] S. Abdulla, T.L. Mathew, B. Pullithadathil, Highly sensitive, room temperature gas sensor based on polyaniline-multiwalled carbon nanotubes (PANI/ MWCNTs) nanocomposite for trace-level ammonia detection, Sens. Actuators, B 221 (2015) 1523e1534. [9] X. Huang, N. Hu, R. Gao, Y. Yu, Y. Wang, Z. Yang, E. Siu-Wai Kong, H. Wei, Y. Zhang, Reduced graphene oxideepolyaniline hybrid: preparation, characterization and its applications for ammonia gas sensing, J. Mater. Chem. 22 (2012) 22488e22495. [10] R. Zhang, V. Alecrim, M. Hummelgård, B. Andres, S. Forsberg, M. Andersson, H. Olin, Thermally reduced kaolin-graphene oxide nanocomposites for gas sensing, Sci. Rep. 5 (2015) 7676e7682. [11] J. Wu, K. Tao, J. Miao, L.K. Norford, Improved selectivity and sensitivity of gas sensing using a 3D reduced graphene oxide hydrogel with an integrated microheater, ACS Appl. Mater. Interfaces 7 (2015) 27502e27510. [12] X.L. Huang, N.T. Hu, Y.Y. Wang, Y.F. Zhang, Ammonia gas sensor based on aniline reduced graphene oxide, Adv. Mater. Res. 669 (2013) 79e84. [13] T. Kavinkumar, S. Manivannan, Uniform decoration of silver nanoparticle on exfoliated graphene oxide sheets and its ammonia gas detection, Ceram. Int. 42 (2016) 1769e1776. [14] M. Gautam, A.H. Jayatissa, Ammonia gas sensing behavior of graphene surface decorated with gold nanoparticles, Solid-State Electron. 78 (2012) 159e165. [15] S. Hazra, S.J. Basu, Graphene-oxide nano composites for chemical sensor applications, Carbon Res. 2 (2016) 1e18. [16] R. Ghosh, A.K. Nayak, S. Santra, D. Pradhan, P.K. Guha, Enhanced ammonia sensing at room temperature with reduced graphene oxide/tin oxide hybrid film, RSC Adv. 5 (2015) 50165e50173. [17] C. Xiang, D. Jiang, Y. Zou, H. Chu, S. Qiu, H. Zhang, F. Xu, L. Sun, L. Zheng, Ammonia sensor based on polypyrroleegraphene nanocomposite decorated with titania nanoparticles, Ceram. Int. 41 (2015) 6432e6438. [18] A. Abdelhalim, A. Abdellah, G. Scarp, P. Lugli, Metallic nanoparticles functionalizing carbon nanotube networks for gas sensing applications, Nanotechnology 25 (2014) 055208e055217.  n, T.J. Bandosz, Insight into [19] N.A. Travlou, M. Seredych, E. Rodríguez-Castello ammonia sensing on heterogeneous S- and N- co-doped nanoporous carbons, Carbon 96 (2016) 1014e1021.  n, T.J. Bandosz, Activated [20] N.A. Travlou, M. Seredych, E. Rodríguez-Castello carbon-based gas sensors: effects of surface features on the sensing mechanism, J. Mater. Chem. A 3 (2015) 3821e3831. n, T.J. Bandosz, Ni[21] N.A. Travlou, C. Ushay, M. Seredych, E. Rodríguez-Castello trogen-doped activated carbon-based ammonia sensors: effect of specific surface functional groups on carbon electronic properties, ACS Sens. 1 (2016) 591e599.  n, T.J. Bandosz, [22] K. Singh, N.A. Travlou, S. Bashkova, E. Rodríguez-Castello Nanoporous carbons as gas sensors: exploring the surface sensitivity, Carbon 80 (2014) 183e192. [23] T. Zhang, S. Mubeen, N.V. Myung, M.A. Deshusses, Recent progress in carbon nanotube-based gas sensors, Nanotechnology 19 (2008) 332001e332014. [24] Y.H. Zhang, Y.B. Chen, K.G. Zhou, C.H. Liu, J. Zeng, H.L. Zhang, Y. Peng, Improving gas sensing properties of graphene by introducing dopants and defects: a first principles study, Nanotechnology 20 (2008) 185504e185511. [25] D. Hulicova-Jurcakova, M. Seredych, G.Q. Lu, T.J. Bandosz, Combined effect of nitrogen- and oxygen-containing functional groups of microporous activated carbon on its electrochemical performance in supercapacitors, Adv. Funct. Mater 19 (2009) 438e447. [26] D.W. Wang, F. Li, L.C. Yin, X. Lu, Z.G. Chen, I.R. Gentle, G.Q. Lu, H.M. Cheng, Nitrogen-doped carbon monolith for alkaline supercapacitors and understanding nitrogen-induced redox transitions, Chem. - Eur. J. 18 (2012) 5345e5351.

[27] B.K. Ostafiychuk, I.M. Budzulyak, B.I. Rachiy, M.M. Kuzyshyn, L.O. Shyyko, Nanoporous nitrogen-containing coal for electrodes of supercapacitors, Nanosci. Nanotechnol. Res. 1 (2013) 17e22. [28] V. Sahu, S. Grover, B. Tulachan, M. Sharma, G. Srivastava, M. Roy, M. Saxena, N. Sethy, K. Bhargava, D. Philip, H. Kim, G. Singh, S.K. Singh, M. Das, R.K. Sharma, Heavily nitrogen doped, graphene supercapacitor from silk cocoon, Electrochim. Acta 160 (2015) 244e253. [29] H.L. Guo, P. Su, X. Kang, S.K. Ning, Synthesis and characterization of nitrogendoped graphene hydrogels by hydrothermal route with urea as reducingdoping agents, J. Mater. Chem. A 1 (2013) 2248e2255. [30] J. Zhang, L. Dai, Heteroatom-doped graphitic carbon catalysts for efficient electrocatalysis of oxygen reduction reaction, ACS Catal. 5 (2015) 7244e7253. [31] N.D.K. Tu, J. Choi, C.R. Park, H. Kim, Remarkable conversion between n- and ptype reduced graphene oxide on varying the thermal annealing temperature, Chem. Mater. 27 (2015) 7362e7369. [32] K.P. Prathish, M.M. Barsan, D. Geng, X. Sun, C.M.A. Brett, Chemically modified graphene and nitrogen-doped graphene: electrochemical characterisation and sensing applications, Electrochim. Acta 114 (2013) 533e542. [33] D. Usachov, O. Vilkov, A. Grüneis, D. Haberer, A. Fedorov, V.K. Adamchuk, A.B. Preobrajenski, P. Dudin, A. Barinov, M. Oehzelt, C. Laubschat, D.V. Vyalikh, Nitrogen-doped graphene: efficient growth, structure, and electronic properties, Nano Lett. 11 (2011) 5401e5407. [34] S. Ni, Z.Y. Li, J.L. Yang, Oxygen molecule dissociation on carbon nanostructures with different types of nitrogen doping, Nanoscale 4 (2012) 1184e1189. [35] Y. Wang, Y. Shao, D.W. Matson, J. Li, Y. Lin, Nitrogen-doped graphene and its application in electrochemical biosensing, ACS Nano 4 (2010) 1790e1798. [36] T.J. Bandosz, C. Petit, On the reactive adsorption of ammonia on activated carbons modified by impregnation with inorganic compounds, J. Colloid Interface Sci. 338 (2009) 329e345. [37] C. Petit, K. Kante, T.J. Bandosz, The role of sulfur-containing groups in ammonia retention on activated carbons, Carbon 48 (2010) 654e667. [38] I. Otomo, C. Kuroda, Effect of the position of reaction-site in amphipathic-type thioester in aqueous amidation reaction, Adv. Chem. Eng. Sci. 5 (2015) 311e316. [39] M.G.S. Ferreira, G. Engineering, Semiconducting properties of oxide and passive films formed on AISI 304 stainless steel and alloy, J. Braz. Chem. Soc. 13 (2002) 1e14. [40] T. Yamamoto, T.A. Koizumi, Synthesis of p-conjugated polymers bearing electronic and optical functionalities by organometallic polycondensations and their chemical properties, Polymer 48 (2007) 5449e5472. [41] J.S. Im, S.C. Kang, S.H. Lee, Y.S. Lee, Improved gas sensing of electrospun carbon fibers based on pore structure, conductivity and surface modification, Carbon 48 (2010) 2573e2581.  n, T.J. Bandosz, Sensing of NH3 on hetero[42] N.A. Travlou, E. Rodríguez-Castello geneous nanoporous carbons in the presence of humidity, Carbon 100 (2016) 64e73. [43] X. Zhang, J. Zhao, M. Tange, W. Xu, W. Xu, K. Zhang, W. Guo, T. Okazaki, Z. Cui, Sorting semiconducting single walled carbon nanotubes by poly(9,9dioctylfluorene) derivatives and application for ammonia gas sensing, Carbon 94 (2015) 903e910. [44] D.C. Tiwari, P. Atri, R. Sharma, Sensitive detection of ammonia by reduced graphene oxide/polypyrrole nanocomposites, Synth. Met. 203 (2015) 228e234. [45] P. Teerapanich, M.T.Z. Myint, C.M. Joseph, G.L. Hornyak, J. Dutta, Development and improvement of carbon nanotube-based ammonia gas sensors using inkjet printed interdigitated electrodes, IEEE Trans. Nanotechnol. 12 (2013) 255e262. [46] X. Xia, S. Guo, W. Zhao, P. Xu, H. Yu, T. Xu, X. Li, Carboxyl functionalized gold nanoparticles in situ grown on reduced graphene oxide for micro-gravimetric ammonia sensing, Sens. Actuators, B 202 (2014) 846e853.  ek, J. Peka rek, V. Svatos, A. Bannov, O. Jasek, [47] P. Majzlíkov a, J. Sedla cek, J. Pra , Hub P. Synek, M. Eli as, L. Zají ckova alek, j. Sensing properties of multiwalled carbon nanotubes grown in MW plasma torch: electronic and electrochemical behavior, gas sensing, Field emission, IR absorption, Sensors 15 (2015) 2644e2661. [48] M. Gautam, A.H. Jayatissa, Adsorption kinetics of ammonia sensing by graphene films decorated with platinum nanoparticles adsorption kinetics of ammonia sensing by graphene films decorated with platinum nanoparticles, J. Appl. Phys. 111 (2012) 094317. [49] 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) 166e170. [50] X. Zhou, X. Wang, B. Wang, Z. Chen, C. He, Y. Wu, Preparation, characterization and NH3-sensing properties of reduced graphene oxide/copper phthalocyanine hybrid material, Sens. Actuators, B 193 (2014) 340e348. [51] T. Zhang, M.B. Nix, B.Y. Yoo, M.A. Deshusses, N.V. Myung, Electrochemically functionalized single-walled carbon nanotube gas sensor, Electroanalysis 18 (2006) 1153e1158.