Effects of O2 plasma and UV-O3 assisted surface activation on high sensitivity metal oxide functionalized multiwalled carbon nanotube CH4 sensors

The authors present a comparative analysis of ultraviolet-O 3 (UVO) and O 2 plasma-based surface activation processes of multiwalled carbon nanotubes (MWCNTs), enabling highly effective functionalization with metal oxide nanocrystals (MONCs). Experimental results from transmission electron microscopy, scanning electron microscopy, x-ray photoelectron spectroscopy, and Raman spectroscopy show that by forming COOH (carboxyl), C-OH (hydroxyl), and C ¼ O (carbonyl) groups on the MWCNT surface that act as active nucleation sites, O 2 plasma and UVO-based dry pretreatment techniques greatly enhance the afﬁnity between the MWCNT surface and the functionalizing MONCs. MONCs, such as ZnO and SnO 2 , deposited by the atomic layer deposition technique, were implemented as the functionalizing material following UVO and O 2 plasma activation of MWCNTs. A comparative study on the relative resistance changes of O 2 plasma and UVO activated MWCNT functionalized with MONC in the presence of 10 ppm methane (CH 4 ) in air is presented as well. V C 2017 American Vacuum Society . [http://dx.doi.org/10.1116/1.4993579]

The authors present a comparative analysis of ultraviolet-O 3 (UVO) and O 2 plasma-based surface activation processes of multiwalled carbon nanotubes (MWCNTs), enabling highly effective functionalization with metal oxide nanocrystals (MONCs). Experimental results from transmission electron microscopy, scanning electron microscopy, x-ray photoelectron spectroscopy, and Raman spectroscopy show that by forming COOH (carboxyl), C-OH (hydroxyl), and C¼O (carbonyl) groups on the MWCNT surface that act as active nucleation sites, O 2 plasma and UVO-based dry pretreatment techniques greatly enhance the affinity between the MWCNT surface and the functionalizing MONCs. MONCs, such as ZnO and SnO 2 , deposited by the atomic layer deposition technique, were implemented as the functionalizing material following UVO and O 2 plasma activation of MWCNTs. A comparative study on the relative resistance changes of O 2 plasma and UVO activated MWCNT functionalized with MONC in the presence of 10 ppm methane (CH 4

I. INTRODUCTION
Methane (CH 4 ) gas has a 100 year global warming impact factor of 24-36 compared to CO 2 . 1 With the emergence of the U.S. as the world's leading producer of natural gas, it is important to enable widespread monitoring of CH 4 emission from natural gas infrastructures. Metal oxide chemoresistive sensors are widely used to sense CH 4 . [2][3][4] Continuous heating is necessary for these sensors to initiate the surface chemisorption of O 2 , a prerequisite to detect CH 4 , often requiring 100 s of milliwatts of power. [2][3][4][5] Carbon nanotube (CNT)based chemoresistor sensors have demonstrated ppm levels of gas sensing at room temperature, with a power consumption of only a few milliwatts. 5 This is a direct outcome of CNT's high surface-to-volume ratio and outstanding modulation of electrical conductance during interaction with gas species. However, functionalizing particles (ranging from metal, 5 metal oxides, 6 and polymer coating 7 to biomolecules 8 ) must be deposited uniformly on the surface of pristine carbon nanotubes in order to enable effective and reversible electrical modulation in the presence of target gas species. Unfortunately, in general, the surface of CNTs shows poor affinity with the functionalizing materials. [9][10][11][12][13] Consequently, before applying the functionalization materials, activation of the inert graphitic surface of the CNTs is necessary. [9][10][11] We present here a comparative analysis of novel UV-O 3 (UVO) and O 2 plasma-based surface activation processes, which enables highly effective functionalization of multiwalled carbon nanotubes (MWCNTs) with metal oxide nanocrystals (MONCs). O 2 plasma and UVO-based dry surface activation techniques have not been applied in CNTbased CH 4 chemoresistor sensors before. 5 Traditionally, the surface of bare CNTs is activated by exposing them to high temperature vapors 10 and/or using wet chemistry. 11 High temperature or air exposure may actually destroy or excessively damage the CNTs. 10 Acid treatments used in wet chemistry can considerably reduce the mechanical and electric performance of the tubes by introducing large numbers of defects. 21 Wet chemistry also involves additional steps, such as dissolution, sonication, mixing, and drying, which often causes undesirable agglomeration of treated CNTs. 10 To increase the efficiency of CNT functionalization, two alternative dry activation processes have been proposed: (1) gas plasma 9-11 and (2) UVO treatment. 22 Due to the interaction of surface C atoms with active O atoms during O 2 plasma or UVO activation and subsequent exposure to the atmosphere, chemical groups such as COOH, C¼O, C-OH, and ether (C-O-C) are formed on the MWCNT surface. [11][12][13]23,24 These groups act as active sites for the nucleation of MONCs. [11][12][13]22 Both the O 2 plasma and UVO exposure have no effect on the aspect ratio (i.e., length to diameter ratio) of the MWCNTs. 21

A. Sensor fabrication
We fabricated surface activated MONC (ZnO or SnO 2 ) functionalized MWCNT chemoresistive CH 4 sensors using the following fabrication steps: (1) lift-off based photolithography, (2) O 2 plasma or UVO based surface activation, and (3) ALD based functionalization. The sensor concept is illustrated in Fig. 1(a). Figure 1(b) shows an SEM image of the sensor where functionalized MWCNTs can be seen deposited between a pair of Au electrodes.
Interdigitated Au electrodes were fabricated by photolithography. Details on the fabrication process can be found elsewhere. 25 MWCNT (98% pure) with an average diameter of 12 nm, an average length of 10 lm, and a specific surface area of 220 m 2 /g was purchased from Sigma Aldrich. Using a microsyringe, an aliquot of 50 ll from a 1 g/50 ml solution of MWCNT-ethanol was deposited on an active area of 1 mm 2 on the fabricated Au electrodes. It was followed by baking the devices at 75 C to remove ethanol and improve adhesion.
The deposited MWCNTs were O 2 plasma activated in a reactive ion etching chamber (March plasma CS-1701). The base pressure of the plasma chamber was almost 40 mTorr. O 2 was introduced at a flow rate of 20 sccm, while the pressure was maintained at 160 mTorr during the process. O 2 plasma was generated by applying a radio frequency of 13.56 MHz with a power of 100 watts. The duration of the plasma treatment was 5 min. A UVO cleaner (Nanonex Ultra 100) was used for the UVO treatment of the MWCNT surface where a 185 nm UV was radiated to atmosphere for generating O 3 and activating the MWCNT surface. The process duration was 20 min.
Using diethylzinc [(C 2 H 5 ) 2 Zn] as a precursor, ALD of ZnO on the surface activated MWCNTs was performed using an Arradiance Gemstar ALD tool (details can be found in Ref. 25).
ALD was also used to deposit SnO 2 nanocrystals (NCs) onto the surface activated MWCNTs. The growth was carried out using an Ultratech Savannah S200 with tetrakis(dimethylamino)tin(IV) as a precursor (details can be found in Ref. 26).

B. TEM sample preparation and imaging
Holey carbon films on Cu grids were used to prepare the TEM sample. Using a microsyringe, an aliquot of 50 ll from a 1 mg/50 ll solution of MWCNT-ethanol solution was deposited on the TEM grid. It was followed by baking the devices at 75 C to remove ethanol and improve adhesion. MWCNTs were surface activated in a similar manner as described in Sec. II A (5 min O 2 plasma or 20 min UVO).
ALD of ZnO or SnO 2 was performed following a similar approach described in Subsec. II A.
A JEOL 2100F TEM operating at 200 kV was used to characterize the atomic scale morphology and crystal quality of the MONCs deposited on MWCNT surfaces.

C. XPS sample preparation
Au of 20 nm was deposited on a clean Si wafer using electron beam evaporation. Using a microsyringe, an aliquot of relatively higher density solution (1 mg/1 ll) of MWCNTethanol was deposited on the Au-coated Si wafers. It was followed by baking the devices at 75 C to remove ethanol and improve adhesion. MWCNTs were surface activated in a similar manner as described in Sec. II A (5 min O 2 plasma or 20 min UVO).

D. Test setup and approach
Sensors were exposed to a 10 ppm mixture of CH 4 in synthetic air (20.81% of O 2 and 79.19% of N 2 ; prepared by Praxair Inc.). The flow rate was maintained at 0.94 l/min with a residence time of 4.5 min inside the plastic test chamber (details elsewhere 25 ). After the CH 4 exposure, the sensors were flushed with N 2 (the same flow rate as CH 4 , 0.94 l/min). The electrical signal obtained from the sensors was recorded using a custom interface circuit connected to a computer. A onset HOBO U12 series data logger temperature and RH sensor was used to continuously monitor and record the RH and temperature inside the plastic test chamber during the test.

A. Sample characterization
High resolution XPS (h ¼ 650 eV) was carried out using beamline 4-ID-C at the Advanced Photon Source, Argonne National Laboratory. MWCNTs were deposited on goldcovered silicon substrates and subsequently activated by O 2 plasma or UVO. Binding energies were calibrated to the Au 4f binding energy of 84.0 eV. Quantification was performed using XPS data analysis software CASAXPS. Figure  The normalized peak areas (NPA) of various components of C 1s and O 1s spectra were calculated with respect to the area of their respective C1 component (sp 2 ) ( Table I).
Comparison among the NPA of C5 (COOH group) in several samples suggests that COOH is the primary chemical group created by the surface activation process. The NPA of C-OH components (C2 and O2) are significantly larger in the surface activated sample compared to the pristine sample, suggesting the strong presence of C-OH in the surface activated MWCNT as well. The NPA of C4 and O1, representing the C¼O group, were found to be highest in the 5 min plasma activated MWCNT but insignificant in pristine and 20 min UVO activated MWCNT. On the other hand, the NPA of the C3 and O3, representing the C-O-C functional group, were found to be highest in the pristine MWCNT (Table I).
It is well known that active p bonds in C¼C are dissociated during plasma/UVO activation and -C. free radicals are produced. 11,23,24 Subsequently, -C. free radicals are oxidized by active O atoms present in the O 2 plasma and UVO, resulting in C-O and C¼O bonds. 23,24 After prolonged interaction with plasma/UVO, C¼O is further oxidized and O-C¼O is formed. 23,24 Due to atmospheric exposure, C-O and O-C¼O stabilize by reacting with ambient H 2 O and generate C-OH and COOH, respectively. 23 This is the probable cause of the strong presence of COH and COOH groups in our plasma/ UVO activated MWCNTs. Surface C atoms of pristine MWCNT react with atmospheric H 2 O to create C-O, a probable cause of the presence of the C-O-C group in pristine MWCNT. 24 In summary, the XPS results corroborate that the surface activation process produces the COOH functional group along with C-OH and C¼O. In later steps, these groups help in nucleating the functionalizing MONCs on the surface of the MWCNTs.
The TEM micrographs show that MONCs are not visible on the surface of the nonactivated but ALD processed MWCNTs [ Fig. 3(a)]. Uniform deposition of ZnO-MONCs was found on the surface of the activated MWCNTs [ Fig.  3(b)]. The clearly visible lattice fringes in the higher resolution TEM (HRTEM) image in Fig. 3(d) illustrate the wurtzite structure of the ZnO MONC and its good crystalline quality. The interplanar spacings of 2.8, 2.68, and 2.48 Å correspond to the h100i, h002i, and h101i planes of ZnO, respectively. 30 The HRTEM image in Fig. 3(c) shows the atomic scale morphology of rutile SnO 2 MONCs deposited on the MWCNT surface. The interplanar spacings of 2.6 and 3.3 Å correspond to h101i and h110i planes of SnO 2 , respectively. 6 TEM results validate the hypothesis that surface activation of the MWCNTs is essential for effective functionalization, i.e., nucleation and stronger binding of the MONCs to the surfaces of the MWCNTs.
Room temperature Raman spectroscopy was performed using a Renishaw Invia micro-Raman system with a 514 nm  (2), while none of these peaks were visible on the untreated MWCNT samples (3), consistent with the hypothesis that the ZnO NC functionalization is enhanced in surface-activated MWCNTs. The peaks described in Fig. 4(b) 31 Consequently, Raman characterization results also validate the hypothesis that surface activation of the MWCNTs is essential for effective functionalization, i.e., stronger nucleation and binding of the MONCs onto the MWCNT surfaces.
The G peak represents the movement in the opposite direction of two neighboring carbon atoms in a graphitic sheet, hence indicating the presence of crystalline graphitic carbon in MWCNTs, 10 while the D peak represents the defects in the curved graphite sheet, sp 3 carbon, or other impurities. 33 The I D /I G ratio, where I corresponds to the peak area of the Lorentzian functions, is an estimate of the relative structural defects. Our preliminary characterization suggests that due to O 2 plasma activation, the relative intensity of the D-peak with respect to the G-peak (I D /I G ratio) of the MWCNT increases 13.5%. The results are presented in the supplementary material. 49 The probable reason for the increase in the intensity of the D-peak with respect to the G peak is the presence of reactive sites on the surface of the MWCNTs created by O 2 . These sites are supposed to  enhance the uniform distribution of the metal oxide nanoparticles on the MWCNT surface.

B. Methane sensing
Our hypothesis is that MONCs facilitate chemisorption of methane gas molecules on the surface of the activated MWCNTs. We found that the resistance of the MONC-MWCNT sensors increases in the presence of a mixture of methane in dry air. The relative resistance change has been defined as DR=R ¼ ðR methane À R air Þ=R air : A series of experiments were conducted to evaluate the effect of UVO and O 2 plasma treatments on the performance of the MONC-MWCNT sensor. To decouple the sensor response to CH 4 from the interference of variable RH, the RH inside the test chamber was kept constant and monitored in real time during each test period. The relative resistance of the sensor monotonically increased at room temperature when 10 ppm CH 4 in air was introduced to the test chamber. While maintaining a constant flow rate, when the incoming gas was switched from CH 4 to N 2 , the relative resistance of the sensor decreased and returned to the baseline [Figs. 5(a)-5(c)]. The sensing mechanism could be elucidated from this phenomenon: the monotonic increase in the sensor's relative resistance is a result of absorption of CH 4 molecules on the MONC functionalized MWCNT surface. Figure 5 also corroborates the assumption that surface activation is essential for effective functionalization of the MWCNT by MONCs and for the sensor to act reversibly in the presence and absence of 10 ppm CH 4 in air. Figures 5(a) and 5(b) show reproducible changes in the relative resistance of the surface activated ZnO-MWCNT sensor during alternating exposure to CH 4 and N 2 . Figure 5(c) illustrates a surface activated SnO 2 -MWCNT chemoresistor sensor alternatively exposed to CH 4 and N 2 showing similarity to the ZnO-MWCNT results. No discernible signals were observed in the untreated (but ZnO NC deposited) MWCNT sensor [ Fig. 5(d)].
At room temperature, the average relative resistance change ½DR=R ¼ ðR methane À R air Þ=R air was found to be 1.91 6 0.98% for UVO activated and 10.5 6 1.01% for O 2 plasma activated ZnO-MWCNT sensors. The results show that the O 2 plasma activation significantly enhances the affinity of the MONCs (in this case ZnO NCs) to the MWCNT surface in comparison to UVO activation. This enhanced affinity causes stronger electron transport through the ZnO-MWCNT junction, i.e., a larger resistance change in the presence of CH 4 at room temperature [Figs. 5(a) and 5(b)]. This is likely due to the better crystal quality of the ZnO NCs on O 2 plasma activated MWCNTs compared to UVO activated MWCNTs (as also observed from the Raman results in Fig. 4).
A novel UV-based recovery technique was recently presented by our group. 34 The sensor was first exposed to 10 ppm CH 4 in air for 30 min, and without interrupting the flow of CH 4 , the sensor was irradiated with UV light until the sensor returned to its baseline resistance. No N 2 flow was used during the recovery. A recovery time of about 3 min was observed. The improvement in the recovery time, we believe, was due to the UV induced reduction of the desorption energy barrier of the CH 4 molecules at the sensor surface. 5 To verify the sensor response to methane, the ZnO-MWCNT sensor was tested at varying CH 4 concentrations (2, 5, and 10 ppm in dry air) at room temperature. Methane of 2 and 5 ppm was obtained from the dilution of 10 ppm methane in synthetic air. The sensors were exposed to CH 4 for 10 min (gas phase) and then to nitrogen for 10 min (desorbing phase), repeating this protocol in the entire set of experiments. The response from a representative sensor is shown in Fig. 6. The sensors were also tested and displayed zero cross-sensitivity to O 2 . The cross-sensitivity to O 2 was determined to be insignificant comparing the sensitivity of the sensor with a variable dilution of O 2 in N 2 , i.e., synthetic air (results not shown here).
The sensitivity of a semiconducting oxide gas sensor is defined as follows: 5,6 (1) for reducing gas DR=R ¼ ðR gas À R air Þ=R air ; (2) for oxidizing gas where R air is the resistance of the sensor in air and R gas is the resistance of the sensor in the presence of gas and air.
We found that the resistance of the MONC-MWCNT sensors changes in the presence of a mixture of methane in air. The change in resistance is in accordance with the change in resistance of the SnO 2 -MWCNT nanohybrid reported by Lu et al., 6 where they show that the resistance of the SnO 2 -MWCNT nanohybrid decreases in the presence of oxidizing NO 2 . A possible sensing mechanism has been reported by Lu et al. 6 Target molecules, in this case NO 2 , get directly adsorbed onto the SnO 2 -MWCNT surface, facilitate electron transfer, and change the electrical conductivity of the hybrid nanostructure.
One can find references that describe the electronic properties of MWCNTs as metallic 35,36 or semiconductive. 6,[37][38][39] However, ZnO and SnO 2 are widely known as n-type materials, 6,32 and the presence of a reducing gas, such as CH 4 , alters their charge concentration, resulting in a change in the resistance of the MONC-MWCNT conglomerate, which was observed experimentally.
A deeper investigation on the methane-functionalized CNT surface interaction is in order; however, this is beyond the scope of this paper. Furthermore, we are currently conducting experiments to study the electronic properties of surface pretreated metal oxide nanocrystal functionalized MWCNTs, which will help us to understand their methane gas sensing mechanism more thoroughly.
Understanding the effect of RH is important in estimating the outdoor performance of microfabricated gas sensors. To examine the effect of RH on sensor performance, the change in the baseline relative resistance of the surface activated ZnO-MWCNT sensors was measured at room temperature due to a change in RH. Humidity was provided by a controlled flow of moist air (flow rate: 0.94 l/min) into a plastic test chamber (residence time: 4.5 min). The baseline relative resistance of the sensor increased by about 4% as the RH was increased from 10% to 91% and returned back to the original baseline once the RH was reduced back to 10%. This suggests a strong electron transfer between the MONC functionalized MWCNTs and water molecules (Fig. 7). Our ongoing research involves fabricating a network of MWCNTs selectively functionalized with various metal oxide nanoparticles with different sets of sensitivities to CH 4 and H 2 O. By deconvoluting the constructive/destructive interference at various RH levels, the RH contribution can be effectively determined and separated from the device's response to CH 4 .
We explored the relative resistance change for 10 ppm methane at two different RH levels. 26 While exposing the SnO 2 -MWCNT sensor to 10 ppm of CH 4 in dry air at the higher RH (approximately 70%), a monotonic resistance increment was observed which was similar to the low RH tests where RH was held constant at 5%, as seen in Fig. 8. The sensor also equilibrated to its original response in a similar fashion when the chamber was purged with N 2 . The response to CH 4 and signal-to-noise ratio reduced in comparison to those at lower RH, which we believe was a result of adsorbed H 2 O molecules on the SnO 2 -MWCNT sensor. Although the sensor showed a reduced response, it was still capable of detecting 10 ppm CH 4 in air at 70% RH. We observed a monotonic increase in the sensor resistance, while it was exposed to CH 4 at low and high RH, as well as a monotonic decrease (return to baseline) when CH 4 was purged with N 2 (Fig. 8). The sensor behavior is similar at both low and high humidities although the sensitivity is reduced at high RH. This is likely the result of absorbed H 2 O molecules on the sensor surface. Water molecules are found to behave as electron donors on the surface of carbon  4 in the dry air mixture at room temperature. RH was kept constant at 2% during the experiment. A different metal electrode/CNT configuration was used, resulting in a smaller relative resistance change compared to Fig. 5(a).  nanotubes. [40][41][42] It was reported that a hydrogen-bonded water monolayer forms around the nanotube at a fully water covered condition. 40 Na et al. 40 presented the change in electrical resistance as a function of relative humidity, which agrees with the result presented in Fig. 8, i.e., a decrease in the relative change in resistance (DR) at a high RH condition. This can be attributed to electron donation by the H 2 O molecules on the sensor surface. 40,42 Our ongoing work focuses on studying the response of the MONC-MWCNT sensor at a fixed ppm methane for multiple RH%. The CH 4 sensor described here uses MWCNTs functionalized by MONCs to sense methane. We used two well known MONCs, ZnO and SnO 2 , that are widely used methane sensing materials and are inexpensive. ZnO and SnO 2 promote energetically favorable electron transport at the MO-MWCNT junction. 6,20 The work function of ZnO was reported to be almost 4.64 (Ref. 43) or 5.2 eV, 44 while SnO 2 has a work function of 4.7 eV. 6 The work functions of these MONCs are almost equal to the work function of MWCNTs (4.7-4.9 eV). 6,45 Therefore, the Schottky barrier height at the MONC-MWCNT junction is low, facilitating electron transfer between MWCNTs and MONCs. The low Schottky barrier improves the overall sensitivity of the sensor (i.e., high DR/R at low ppm), making the hybrid MONC-MWCNT system a potentially superior sensing element to either of its constituent components. 6 Table II compares the performance of our sensor with other published CNT CH 4 sensors. Note that although the sensor presented in Ref. 46 shows an equivalent performance to our sensors, it used wet chemically treated single walled carbon nanotubes (SWCNTs). Wet chemical treatment may be undesirable in CNT sensor fabrication as it is well known that acid treatments used in wet chemistry can considerably reduce the mechanical and electric performance of the tubes by introducing large numbers of defects, 21 which might limit its reproducibility and reliability, as well as increase the cost of the sensor. Wet chemistry also involves additional steps, such as dissolution, sonication, mixing, and drying, which often causes undesirable agglomeration of treated CNTs.

IV. CONCLUSION
In summary, O 2 plasma activation has a stronger impact than UVO activation on enhancing the MONC functionalization of MWCNTs and thus on the response of the chemoresistive sensors to 10 ppm CH 4 in air. The strong relative resistance change in the presence of 10 ppm of CH 4 at room temperature is a consequence of: (1) strong electron transfer to the MONCs from CH 4 molecules, (2) energetically favorable electron transport at the MONC-MWCNT junction, and (3) enhanced affinity of the dry surface activated MWCNT to MONCs as a result of formation of active chemical groups. The O 2 plasma and UVO-based activation processes give rise to COOH, C¼O, and C-OH functional groups on the MWCNT surface and hence enhance the nucleation and bonding of MONCs with the MWCNT. These treatments produce a strong reversible relative resistance change in the chemoresistors under iterative exposure to 10 ppm CH 4 in air and a relatively reduced response to lower concentrations. The response varies with RH, with a lower response at FIG. 8. (Color online) Relative resistance change of the SnO 2 -MWCNT chemoresistor sensor while exposed to 10 ppm of CH 4 in dry air at (a) a lower RH (5%) and (b) a higher RH (70%) and was recovered by N 2 . The circle symbol plot represents the relative resistance change ½DR=R ¼ ðR RH À R air Þ= R air of the chemoresistor sensor (left hand, y-axis), while the triangle symbol plot represents the RH inside the test chamber recorded by a commercial RH data logger (right hand, y-axis).  higher RH as well as a lower detection limit. At low RH, the detection limit is between 2 and 5 ppm.

ACKNOWLEDGMENTS
The work performed at the Center for Nanoscale Materials and the Advanced Photon Source, Office of Science user facilities was supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract No. DE-AC02-06CH11357. The U.S. Environmental Protection Agency, through its Office of Research and Development, collaborated in the research described here. It has been subjected to Agency review and approved for publication. The authors would like to thank Scienta Omicron for loan of the Argus electron energy analyzer.