Evaluation of Surface Cleaning Procedures in Terms of Gas Sensing Properties of Spray-Deposited CNT Film: Thermal- and O2 Plasma Treatments

The effect of cleaning the surface of single-walled carbon nanotube (SWNT) networks by thermal and the O2 plasma treatments is presented in terms of NH3 gas sensing characteristics. The goal of this work is to determine the relationship between the physicochemical properties of the cleaned surface (including the chemical composition, crystal structure, hydrophilicity, and impurity content) and the sensitivity of the SWNT network films to NH3 gas. The SWNT networks are spray-deposited on pre-patterned Pt electrodes, and are further functionalized by heating on a programmable hot plate or by O2 plasma treatment in a laboratory-prepared plasma chamber. Cyclic voltammetry was employed to semi-quantitatively evaluate each surface state of various plasma-treated SWNT-based electrodes. The results show that O2 plasma treatment can more effectively modify the SWNT network surface than thermal cleaning, and can provide a better conductive network surface due to the larger number of carbonyl/carboxyl groups, enabling a faster electron transfer rate, even though both the thermal cleaning and the O2 plasma cleaning methods can eliminate the organic solvent residues from the network surface. The NH3 sensors based on the O2 plasma-treated SWNT network exhibit higher sensitivity, shorter response time, and better recovery of the initial resistance than those prepared employing the thermally-cleaned SWNT networks.


Introduction
Since 2000, considerable progress has been made in the development of carbon nanotube (CNT)-based chemical gas sensors, both theoretically and experimentally. The high surface-to-volume ratio of CNTs can catalyze the chemisorption or physisorption between CNT surfaces and various gases or vapors. This eventually enhances the sensitivity of the CNT substrate to any target materials, even at room temperature. Indeed, numerous articles focusing on different aspects of CNT-based chemical sensors have been published, including a summary of the progress made thus far in gas sensing performance and potential applications in other fields. Although the working mechanism Sensors 2017, 17, 73 3 of 10 t-SWNT or p-SWNT films were deposited on the interdigitated electrodes to prepare two different NH 3 gas sensors.

Characterization and Gas Sensing Measurement
The morphology and structure of the heat or O 2 plasma-cured SWNTs were investigated using scanning electron microscopy (SEM; S-4300, Hitachi Ltd., Tokyo, Japan). To approximately identify the removal of organic materials and the surface restructuring on the modified film, contact angle measurements were taken before and after the cleaning of the SWNT networks using a commercially-available contact angle system (Phoenix-10, SEO Co., Seoul, Korea). Surface characterization of the modified electrode was analyzed using Raman (LabRAM, Horiba KOREA, Bucheon, Korea) and XPS (SIGMA PROBE, ThermoVG, London, UK). The electrochemical measurements were performed using 263A Potentiostat/Galvanostat (Princeton Applied Research, Oak Ridge, TN, USA) at room temperature. To record the square wave voltammograms, the following instrumental parameters were used: step potential = 1 mV; square-wave amplitude = 2 mV; frequency = 10 kHz; and scan rate = 50 mV·s −1 . The NH 3 sensing performance of the SWNT-based gas sensors was evaluated in a laboratory-prepared gas sensing system composed of a furnace, the gas control system (control of NH 3 gas concentration in the detection chamber can be accomplished by adjusting the ratio of gas flow rate of NH 3 to that of N 2 ), a Keithley 236 source meter, a PC, and gas cylinders of NH 3 and N 2 at an operating temperature of 250 • C. Figure 1 shows SEM images of the variously prepared SWNT bundles: (a) before cleaning; (b) after thermal cleaning; and (c) after plasma cleaning. As shown in Figure 1a, the pristine SWNT was susceptible to aggregation, and each individual SWNT was coated by the unwanted, DCB-involved residual sonopolymers [17][18][19][20]. Both cleaning processes can effectively remove the residual organic surfactants to provide a naked SWNT bundle. However, the cleaned surface state strongly depends on the cleaning process employed. The contact angle of the water droplet on the DCB-coated glass substrate which is not treated with heat was about 62.51 • , owing to the hydrophobic nature of DCB. The contact angle remained almost unchanged with the DCB-coated glass substrates annealed at a temperature below 250 • C, but the angle decreased significantly with the glass substrates that were heat treated at a temperature higher than 250 • C and up to 400 • C. This means that the DCB molecules are thermally decomposed at a temperature above 250 • C, eventually enabling the hydrophilic property of the glass substrate to be recovered. Figure 2 shows the static contact angle data for the t-SWNT and p-SWNT prepared at various conditions. The pristine SWNT surface is more hydrophobic than the DCB-coated glass substrate, and the observed contact angle is about 84.91 • , because both the SWNT and DCB molecules are hydrophobic. In the case of the t-SWNT annealed at temperatures higher than 300 • C, the contact angle decreased by only about 6.9% with the removal of the DCB molecules. Conversely, the p-SWNT showed a dramatic contact angle decrease by 93.2% or more. This is attributed to the removal of organic residuals and to the hydrophilic modification of the SWNT's surface; i.e., O 2 plasma chemically and more effectively oxidizes the hydrophobic SWNT surface than the thermal treatment, so the O 2 plasma-treated surface becomes more hydrophilic, since the oxygen-containing functional groups (mainly carboxyl) are more effectively introduced to the SWNT surface.

Results and Discussion
The plasma treatment condition for preparing p-SWNT is optimized mainly in terms of the applied plasma power and the plasma treatment time. The peak current (i p ) obtained from the cyclic voltammograms of the variously prepared p-SWNT electrodes in a 3 M KCl solution containing 10 mM K 3 [Fe(CN) 6 ] at 100 mV·s −1 (inset of Figure 3) is plotted as a function of the plasma treatment time, as shown in Figure 3. The results indicate that the i p of the p-SWNT electrodes treated with O 2 plasma at 20 W were larger than those obtained at 10 W; the larger i p means an increase in the conductivity of the p-SWNT electrode prepared at 20 W power O 2 plasma. The enhanced conductivity is probably due to the larger number of carboxyl or carbonyl substituents on the p-SWNT surface. However, a long plasma treatment time can reduce the conductivity of the p-SWNT electrode, giving a decreased i p , which results from a decreased concentration of the oxygen-containing chemical functional groups on the SWNT surface due to the over-etched SWNTs [21]. Finally, the optimal condition for p-SWNT is obtained at 20 W for 20 s. hydrophilic property of the glass substrate to be recovered. Figure 2 shows the static contact angle data for the t-SWNT and p-SWNT prepared at various conditions. The pristine SWNT surface is more hydrophobic than the DCB-coated glass substrate, and the observed contact angle is about 84.91°, because both the SWNT and DCB molecules are hydrophobic. In the case of the t-SWNT annealed at temperatures higher than 300 °C, the contact angle decreased by only about 6.9% with the removal of the DCB molecules. Conversely, the p-SWNT showed a dramatic contact angle decrease by 93.2% or more. This is attributed to the removal of organic residuals and to the hydrophilic modification of the SWNT's surface; i.e., O2 plasma chemically and more effectively oxidizes the hydrophobic SWNT surface than the thermal treatment, so the O2 plasma-treated surface becomes more hydrophilic, since the oxygen-containing functional groups (mainly carboxyl) are more effectively introduced to the SWNT surface.  The plasma treatment condition for preparing p-SWNT is optimized mainly in terms of the applied plasma power and the plasma treatment time. The peak current (ip) obtained from the cyclic voltammograms of the variously prepared p-SWNT electrodes in a 3 M KCl solution containing 10 mM K3[Fe(CN)6] at 100 mV•s −1 (inset of Figure 3) is plotted as a function of the plasma treatment time, as shown in Figure 3. The results indicate that the ip of the p-SWNT electrodes treated with O2 plasma at 20 W were larger than those obtained at 10 W; the larger ip means an increase in the conductivity of the p-SWNT electrode prepared at 20 W power O2 plasma. The enhanced conductivity is probably due to the larger number of carboxyl or carbonyl substituents on the p-SWNT surface. However, a long plasma treatment time can reduce the conductivity of the p-SWNT electrode, giving a decreased ip, which results from a decreased concentration of the oxygencontaining chemical functional groups on the SWNT surface due to the over-etched SWNTs [21]. Finally, the optimal condition for p-SWNT is obtained at 20 W for 20 s. The plasma treatment condition for preparing p-SWNT is optimized mainly in terms of the applied plasma power and the plasma treatment time. The peak current (ip) obtained from the cyclic voltammograms of the variously prepared p-SWNT electrodes in a 3 M KCl solution containing 10 mM K3[Fe(CN)6] at 100 mV•s −1 (inset of Figure 3) is plotted as a function of the plasma treatment time, as shown in Figure 3. The results indicate that the ip of the p-SWNT electrodes treated with O2 plasma at 20 W were larger than those obtained at 10 W; the larger ip means an increase in the conductivity of the p-SWNT electrode prepared at 20 W power O2 plasma. The enhanced conductivity is probably due to the larger number of carboxyl or carbonyl substituents on the p-SWNT surface. However, a long plasma treatment time can reduce the conductivity of the p-SWNT electrode, giving a decreased ip, which results from a decreased concentration of the oxygencontaining chemical functional groups on the SWNT surface due to the over-etched SWNTs [21]. Finally, the optimal condition for p-SWNT is obtained at 20 W for 20 s.  Figure 4 shows the difference in the core level spectra of SWNTs obtained by thermal-and plasma-cleaning treatments. The XPS spectrum shows distinct carbon, oxygen, and chlorine peaks, representing the major constituents of the SWNT surface. Here, the peak of chlorine (200.34-200.40 eV) is associated with the organic chlorocarbons from the residual DCB molecules. The XPS peak deconvolution of the C ls and Cl 2p peaks for the modified SWNTs is illustrated in Figure 5. As shown in Figure 5a,b,c, the C 1s spectra for the modified SWNTs have five characteristic peaks, including the elemental carbon atoms (sp 2 and sp 3 ) and the oxidized carbon atoms (C-O, C=O, and -  representing the major constituents of the SWNT surface. Here, the peak of chlorine (200.34-200.40 eV) is associated with the organic chlorocarbons from the residual DCB molecules. The XPS peak deconvolution of the C ls and Cl 2p peaks for the modified SWNTs is illustrated in Figure 5. As shown in Figure 5a-c, the C 1s spectra for the modified SWNTs have five characteristic peaks, including the elemental carbon atoms (sp 2 and sp 3 ) and the oxidized carbon atoms (C-O, C=O, and -COO).
The concentrations of C, O, and Cl for all the samples were estimated, and the corresponding results are summarized in Table 1. Compared with the pristine SWNT, the ratio of oxygen-bound carbons to carbon atoms ([CO x ]/[C]) increases in both the heat treatment and O 2 plasma treatment. Especially, a larger number of oxidized species on the SWNTs is observed after the O 2 plasma treatment. The p-SWNT shows a decrease in sp 2 bonding and an increase in sp 3 bonding, revealing that the sp 2 hybridization of SWNTs is altered to sp 3 hybridization, indicating that the characteristics of SWNT (such as electrical conductivity and mechanical strength) gradually disappear [21,22]. The chlorine Cl 2p core level spectra are depicted in Figure 5d results are summarized in Table 1. Compared with the pristine SWNT, the ratio of oxygen-bound carbons to carbon atoms ([COx]/[C]) increases in both the heat treatment and O2 plasma treatment. Especially, a larger number of oxidized species on the SWNTs is observed after the O2 plasma treatment. The p-SWNT shows a decrease in sp 2 bonding and an increase in sp 3 bonding, revealing that the sp 2 hybridization of SWNTs is altered to sp 3 hybridization, indicating that the characteristics of SWNT (such as electrical conductivity and mechanical strength) gradually disappear [21,22]. The chlorine Cl 2p core level spectra are depicted in Figure 5d,     results are summarized in Table 1. Compared with the pristine SWNT, the ratio of oxygen-bound carbons to carbon atoms ([COx]/[C]) increases in both the heat treatment and O2 plasma treatment. Especially, a larger number of oxidized species on the SWNTs is observed after the O2 plasma treatment. The p-SWNT shows a decrease in sp 2 bonding and an increase in sp 3 bonding, revealing that the sp 2 hybridization of SWNTs is altered to sp 3 hybridization, indicating that the characteristics of SWNT (such as electrical conductivity and mechanical strength) gradually disappear [21,22]. The chlorine Cl 2p core level spectra are depicted in Figure 5d,       Figure 6 shows the Raman spectra of the modified SWNT bundles: (a) pristine SWNT; (b) t-SWNT; and (c) p-SWNT. The intensity ratios of the D-band to the G-band (I D /I G ) for the modified SWNT bundles were estimated in order to evaluate the degree of structural deformation [18]. Generally, the G and D bands are characterized as ordered carbon in a tangential mode and as disordered sp 2 carbon, respectively. The I D /I G value of pristine SWNT was about 0.0518, which means that pristine SWNT exhibited some structural damage owing to acid treatment. It is known that most commercially available SWNTs are purified by acid and heat treatments during synthesis in order to improve purity. For the p-SWNT sample, the I D /I G was approximately 0.0717, which is greater than the pristine SWNT, attributed to the defect generation or chemical functionalization on the end edge and even on the sidewalls of the SWNT surface by the O 2 plasma treatment. In contrast, the I D /I G of t-SWNT was about 0.0451, which is smaller than that of the pristine SWNT, indicating that the thermal treatment of the SWNT does not generate significant defects on the SWNT surface. The Raman spectra of the modified SWNTs in the spectral region of the radial breathing mode (RBM) at 170-183 cm −1 are demonstrated in the insets of Figure 6. The diameter of SWNT can be estimated using Equation (1).
where ω is the RBM frequency; d is the diameter; c 1 and c 2 (as the experimentally derived parameters) are 223.5 cm −1 and 12.5 cm −1 , respectively, for SWNT bundles. For all samples, two resolved Radial Breathing Mode peaks were observed, and their corresponding diameters are summarized in Table 2. It has been reported that the radical species react more selectively with SWNTs of a smaller diameter [18]. Based on this, it can be concluded that p-SWNT shows excellent reactivity in comparison to pristine SWNT and t-SWNT.
Sensors 2017, 17, 73 6 of 9 Figure 6 shows the Raman spectra of the modified SWNT bundles: (a) pristine SWNT; (b) t-SWNT; and (c) p-SWNT. The intensity ratios of the D-band to the G-band (ID/IG) for the modified SWNT bundles were estimated in order to evaluate the degree of structural deformation [18]. Generally, the G and D bands are characterized as ordered carbon in a tangential mode and as disordered sp 2 carbon, respectively. The ID/IG value of pristine SWNT was about 0.0518, which means that pristine SWNT exhibited some structural damage owing to acid treatment. It is known that most commercially available SWNTs are purified by acid and heat treatments during synthesis in order to improve purity. For the p-SWNT sample, the ID/IG was approximately 0.0717, which is greater than the pristine SWNT, attributed to the defect generation or chemical functionalization on the end edge and even on the sidewalls of the SWNT surface by the O2 plasma treatment. In contrast, the ID/IG of t-SWNT was about 0.0451, which is smaller than that of the pristine SWNT, indicating that the thermal treatment of the SWNT does not generate significant defects on the SWNT surface. The Raman spectra of the modified SWNTs in the spectral region of the radial breathing mode (RBM) at 170-183 cm −1 are demonstrated in the insets of Figure 6. The diameter of SWNT can be estimated using Equation (1).
where ω is the RBM frequency; d is the diameter; c1 and c2 (as the experimentally derived parameters) are 223.5 cm −1 and 12.5 cm −1 , respectively, for SWNT bundles. For all samples, two resolved Radial Breathing Mode peaks were observed, and their corresponding diameters are summarized in Table 2. It has been reported that the radical species react more selectively with SWNTs of a smaller diameter [18]. Based on this, it can be concluded that p-SWNT shows excellent reactivity in comparison to pristine SWNT and t-SWNT.    As shown in Figure 7a, the cyclic voltammograms of all electrodes in a 3 M KCl solution containing 10 mM K 3 [Fe(CN) 6 ] at 100 mV·s −1 exhibit well-defined and quasi-reversible redox reactions, indicating that effective electron transfer occurs between the electrode and the electrolyte containing the redox active species. The pristine SWNT electrode showed a relatively lower peak current (i p ) and a wider peak-to-peak separation potential (∆E p ) than the t-SWNT and p-SWNT electrodes. The p-SWNT exhibited the smallest ∆E p (about 0.0648 V, as shown in Figure 7a), and the highest i p (ca. 127.01 µA, as shown in Figure 7b) than the other SWNTs. The magnitude of the electrochemical response for these electrodes is in the following order: p-SWNT > t-SWNT > pristine SWNT (Figure 7b). The excellent electrochemical properties of the p-SWNT electrode due to a sufficient amount of carbonyl/carboxyl substituents result in enhanced conductivity and electron transfer acceleration, which play important roles in electronic devices. indicating that effective electron transfer occurs between the electrode and the electrolyte containing the redox active species. The pristine SWNT electrode showed a relatively lower peak current (ip) and a wider peak-to-peak separation potential (ΔEp) than the t-SWNT and p-SWNT electrodes. The p-SWNT exhibited the smallest ΔEp (about 0.0648 V, as shown in Figure 7a), and the highest ip (ca. 127.01 μA, as shown in Figure 7b) than the other SWNTs. The magnitude of the electrochemical response for these electrodes is in the following order: p-SWNT > t-SWNT > pristine SWNT ( Figure 7b). The excellent electrochemical properties of the p-SWNT electrode due to a sufficient amount of carbonyl/carboxyl substituents result in enhanced conductivity and electron transfer acceleration, which play important roles in electronic devices. Note that the p-SWNT shows the smallest peak potential separation and the biggest ip. This indicates the best surface condition of the p-SWNT electrode. Note that the modified electrodes are prepared under optimal treatment conditions.  Figure 8 shows the dynamic response of the modified SWNT-based sensors at various NH3 concentrations. The sensing response is estimated using Equation (2): where Rg and R0 are the electrical resistances before and after exposure to NH3, respectively. The response time and the recovery time are defined as the times to reach 90% of total change in electrical resistance upon exposure to target gas and to return to 10% of the original resistance in N2 gas after the test gas is released, respectively. Among the three samples, the p-SWNTs sensor exhibited the highest sensitivity and the fastest response time (Figure 8), because the gas sensing performance depends mostly on the concentration of oxygen-containing functional groups and on the defect sites on the SWNT surface. Moreover, the pristine and t-SWNTs sensors showed incomplete recovery of their initial resistance for all concentrations. This might be attributed to a buildup of chemisorbed NH3 on the SWNT surface, resulting in irreversible change in the resistance [23]. The sensing performances of the modified SWNTs sensors are listed in Table 3. Note that the p-SWNT shows the smallest peak potential separation and the biggest i p . This indicates the best surface condition of the p-SWNT electrode. Note that the modified electrodes are prepared under optimal treatment conditions. Figure 8 shows the dynamic response of the modified SWNT-based sensors at various NH 3 concentrations. The sensing response is estimated using Equation (2): where R g and R 0 are the electrical resistances before and after exposure to NH 3 , respectively. The response time and the recovery time are defined as the times to reach 90% of total change in electrical resistance upon exposure to target gas and to return to 10% of the original resistance in N 2 gas after the test gas is released, respectively. Among the three samples, the p-SWNTs sensor exhibited the highest sensitivity and the fastest response time (Figure 8), because the gas sensing performance depends mostly on the concentration of oxygen-containing functional groups and on the defect sites on the SWNT surface. Moreover, the pristine and t-SWNTs sensors showed incomplete recovery of their initial resistance for all concentrations. This might be attributed to a buildup of chemisorbed NH 3 on the SWNT surface, resulting in irreversible change in the resistance [23]. The sensing performances of the modified SWNTs sensors are listed in Table 3.

Conclusions
We demonstrated an effective strategy to remove the residual DCB from SWNTs, and showed the influence of surface cleaning procedures on the sensing performance of a SWNT-based NH3 gas sensor. Thermal-and O2 plasma-cleaning treatments were introduced to remove DCB from the SWNTs, and the surface characteristics and electrochemical properties of the cleaned SWNTs were analyzed. The smaller contact angle and lower chlorine percentage after cleaning verified the removal of the DCB from the SWNTs. Although both the O2 plasma and thermal treatments exhibited similar DCB removal, the electrochemical property and reactivity to NH3 gas of the p-SWNTs were superior to those of the t-SWNTs, and this was attributed to the smaller diameter of p-SWNTs and larger number of carboxyl/carbonyl substituents on the p-SWNT surface. More importantly, the p-SWNTbased gas sensor showed more enhanced sensing performance to detect NH3 gas, such as higher response variations and shorter response/recovery times than the pristine-and t-SWNT-based sensors. While thermal cleaning of the SWNT is to remove organic adsorbents from the nanotube surface, plasma cleaning includes cleaning and functionalization of the SWNT at the same time. Therefore, O2 plasma treatment is a smart cleaning technique and can be widely applied to various SWNT-based electrodes or devices.

Conclusions
We demonstrated an effective strategy to remove the residual DCB from SWNTs, and showed the influence of surface cleaning procedures on the sensing performance of a SWNT-based NH 3 gas sensor. Thermal-and O 2 plasma-cleaning treatments were introduced to remove DCB from the SWNTs, and the surface characteristics and electrochemical properties of the cleaned SWNTs were analyzed. The smaller contact angle and lower chlorine percentage after cleaning verified the removal of the DCB from the SWNTs. Although both the O 2 plasma and thermal treatments exhibited similar DCB removal, the electrochemical property and reactivity to NH 3 gas of the p-SWNTs were superior to those of the t-SWNTs, and this was attributed to the smaller diameter of p-SWNTs and larger number of carboxyl/carbonyl substituents on the p-SWNT surface. More importantly, the p-SWNT-based gas sensor showed more enhanced sensing performance to detect NH 3 gas, such as higher response variations and shorter response/recovery times than the pristine-and t-SWNT-based sensors. While thermal cleaning of the SWNT is to remove organic adsorbents from the nanotube surface, plasma cleaning includes cleaning and functionalization of the SWNT at the same time. Therefore, O 2 plasma treatment is a smart cleaning technique and can be widely applied to various SWNT-based electrodes or devices.