MWNTs used in this paper were purchased from the Chengdu Institute of Organic, Chinese Academy of Sciences and were grown by chemical vapor deposition (CVD) method. The tube diameter is 20∼30 nm, length 10∼30 μm, purity >95%, and catalyst residue (ash) <1.5 wt%. Due to the cluster effect, they required pretreatment before modification so that the cluster MWNTs can be spread out evenly and achieve better modification effect. First, the MWNTs were placed into a beaker containing the appropriate ethanol solution, after which the beaker was placed in an ultrasonic bath for an hour. Finally, MWNTs was filtered out of the solution using filtration paper with a pore size of 0.1 μm. This step was performed to separate the MWNTs from the solution. Through this process, MWNTs can be spread better after pretreatment.

In this paper, MWNTs were treated by atmospheric pressure DBD plasma for surface modification. Air was used as precursor gas for plasma in the laboratory, and the experimental temperature was 25 °C. The scheme of the surface modification experiment setup is shown in Figure 1. The frequency of power excitation ranges from 16 kHz to 30 kHz, and the voltage amplitude was adjusted continuously in a range from 0 kV to 20 kV. Supply voltage waveform was collected by the high voltage probe P6015A (attenuation ratio 1,000). Transimission charge in discharge space was obtained by 2,000 pF capacitor in series in the circuit indirectly. The oscilloscope was a Tektronix model DPO4054.

The main body of the reactor in the modifying experiment setup was a cylindrical quartz tube. A 200 mm long copper strip was used as grounding electrode and wrapped around the outer wall of the quartz glass tube. A fixed coaxial high voltage copper rod was placed in the quartz glass tube and needle-shaped copper prick electrodes were placed in an array on the copper rod.

Resonance occurs when the power frequency is 21 kHz in Figure 1. At this frequency, cycle transmission charge, discharge power and efficiency of energy injection are all at maximum. Therefore, the frequency and peak to peak voltage of the power supply were chose to be 21 kHz and 16.4 kV, respectively.

The load characteristic of DBD reactor was capacitive, and the discharge process can be modeled as capacitor charge-discharge process. The voltage U_m across the capacitor C_m was proportional to tansimission charge in discharge space Q_m. The supply voltage and U_m measured by high voltage probe were added to the y-x axis of the oscilloscope, so the Lissajous curve [23] can be obtained (in Figure 2). d_Q in the figure was transimission charge in half cycle, and U_b was starting discharge voltage. DBD cycle transimission charge and discharge power can be calculated approximately by the following equations:

The measured voltage and current waveform, voltage instantaneous power waveform and Lissajous curve in modified experimental conditions were shown in Figure 3. According to the equations (1) to (2), the cycle transmission charge was 1,159 nC and discharge power was 114.3 W.

After treatment, the MWNTs were distributed as thinly as possible on the bottom of the cylindrical quartz glass tube. All MWNTs were evenly distributed in the region wrapped up by the copper strip. This is because the discharge in this region is more intense, thus improving the plasma modification effect. The MWNTs were treated for 30, 60 and 120 s. In the end, we obtained MWNTs processed for different duration.

The substrate where the MWNTs were deposited on, was a interdigital electrodes printed circuit board with area 5 mm × 10 mm, thickness of electrodes 30 μm, space between the electrodes 1 mm. The MWNTs modified by DBD plasma were put into a beaker containing the appropriate ethanol solution, after which they were made to undergo ultrasonic treatment for 1 h. Drops of the mixed solution were dropped on the surface of the substrate. Finally, the substrates coated with MWNTs were placed in oven and baked at 80 °C for 2 h. The process was repeated for several times until uniform MWNTs film was prepared on the surface. According to this method, we fabricated four kinds of MWNTs-based gas sensors, including plasma-modified MWNTs (i.e., those prepared at three different times of 30, 60, and 120 s) and untreated MWNTs. The homemade system mainly includes gas chamber, impedance and test gas. The scheme of the detection chamber is shown in Figure 4. The gas amount is controlled by gas flow meter, and the gas flow rate is about 20 sccm.

The resistance of the MWNTs gas sensor was measured in the gas experiment. The concentrations of H₂S and SO₂ used in the experiment were both 50 ppm. The experiment was performed at approximately 25 °C. The sensor sensitivity S is defined as: S = | R − R 0 | R 0 × 100 %where R and R₀ are the values of resistance measured in the presence of gas and vacuum, respectively.

The morphologies of the plasma-modified MWNTs were characterized by field emission scanning electron microscope (FESEM). Figure 8(a) is SEM image of the unmodified MWNTs. Figure 8(b,c) are SEM images of modified MWNTs, and the two figures come from the different areas of modified MWNTs.

It can be seen from Figure 8(a) that there are many amorphous carbons and residual catalysts on the surface of untreated MWNTs. Figure 8(b) shows that the amorphous carbon and residual catalysts on the surface of MWNTs are apparently removed after plasma modification. Figure 8(c) shows that part of the MWNTs has been slightly damaged after plasma treatment. There are a small amount of surface defects, but the structural integrity has not been destroyed. Both the removal of amorphous carbons and residual catalysts on the surface of MWNTs and a small amount of surface defects may contribute to gas adsorption. Thus, the sensitivity of the modified MWNTs is enhanced greatly.

Important chemical information regarding the incorporated functional groups was characterized by Fourier transform infrared spectroscopy. Figure 9 shows the infrared spectra of the untreated MWNTs and MWNTs plasma-treated MWNTs.

Comparing Figure 9(a) to 9(d), we can see that the adsorption bands they have in common are 3,445 cm⁻¹, 2,922 cm⁻¹, and 2,852 cm⁻¹, where 3,445 cm⁻¹ corresponds with the stretching vibration of hydroxyl, and 2,922 cm⁻¹, 2,852 cm⁻¹ correspond to the -CH₂ asymmetric stretching vibration and symmetric stretching vibration, respectively. The appearance of these peaks resulted from the MWNT growth process with the chemical vapor deposition method.

In Figure 9(a), the adsorption bands 1,578 cm⁻¹ correspond to the stretching vibration of C=C, and the adsorption bands 1,636 cm⁻¹ and 1,385 cm⁻¹ correspond with the sp²-C sheet structure of MWNTs and D-band caused by defects. These adsorption bands can prove that infrared spectrum of unmodified MWNTs is a typical one for MWNTs [26].

Figure 9(b) shows the infrared spectrum of MWNTs modified by plasma for 30 s. In Figure 9(b), there are absorption bands at 1,560 cm⁻¹ and 1,535 cm⁻¹ which do not appear in Figure 9(a). Figure 9(c) shows the infrared spectrum of MWNTs treated by plasma for 60 s. In Figure 9(c), there are absorption bands at 1,620 cm⁻¹ and 1,540 cm⁻¹ which do not appear in Figure 9(a) too. The four absorption bands above are in the range of 1,620∼1,540 cm⁻¹ which is the asymmetric stretching of carboxylic ion. So we can infer that carboxyl is introduced in MWNTs modified by plasma for 30 s and 60 s.

Compared with Figure 9(a), the new absorption bands appeared in Figure 9(b) to (d) also include 1,168 cm⁻¹ [in Figure 9(b)], 1,165 cm⁻¹ [in Figure 9(c)], 1,120 cm⁻¹ [in Figure 9(d)], 1,117 cm⁻¹ [in Figure 9(b)] and 1,115 cm⁻¹ [in Figure 9(c)]. These absorption bands range from 1,140 to 1,110 cm⁻¹ which is the characteristic stretching vibration of C-N. Thus, we can infer that the nitrogen containing groups are introduced in the treated MWNTs.

From the above analysis, some carboxyl and nitrogen-containing groups are introduced on the surface of plasma-modified MWNTs. These groups interact with the gas molecules and become the active center of gas adsorption, thus increasing the sensitivity of MWNTs.

The modified MWNTs showed higher sensitivity to H₂S and greatly reduced response time, while they showed no sensitivity to SO₂. This is probably due to the following reason: it is known that SO₂ shows both oxidizing and reducing behavior. According to the analysis above some carboxyl and nitrogen-containing groups are introduced onto the surface of plasma-modified MWNTs. When SO₂ interacts with modified MWNTs [seen in Figure 10(a)], there are electrons transferring from SO₂ to MWNTs because of the weak carboxyl oxidizability. There are also electrons transferring from MWNTs to SO₂, because the N atoms are electron rich. When the two processes above achieve dynamic equilibrium, there are no electrons transferring between SO₂ and MWNTs in general. So the MWNTs modified by plasma show no sensitivity to SO₂.

The addition of nitrogen-containing groups has almost no effect on the MWNT adsorbtion of H₂S [27]. When H₂S interacts with the modified MWNTs [seen in Figure 10(b)], it doesn't interact with the nitrogen-containing group. However, the carboxyl on the MWNTs surface may improve MWNTs gas sensitivity to H₂S. The reason is that these carboxyl groups can provide more adsorption sites for H₂S molecules. In addition, carboxyl shows weak oxidizability, so the amount of the charges transferring from H₂S to MWNTs will increase greatly, leading to the resistance of the MWNTs decreasing much more. Therefore, the modified MWNTs show higher sensitivity to H₂S.