1. Substrate Preparation

The device fabrication process began with a 4-inch, p-type silicon wafer (100 orientation), thermally oxidized to form a 100 nm-thick SiO₂ insulating layer. The oxide layer served both as a dielectric barrier and as a platform for channel deposition. Prior to channel formation, the wafer was cleaned using a standard cleaning procedure followed by deionized (DI) water rinse and N₂ blow-drying.

2. Channel Formation with SWCNTs

SWCNTs, due to their high surface-to-volume ratio and excellent electrical properties, were selected as the core sensing medium. A dispersion solution was prepared by mixing 1 mg of SWCNT powder with 10 ml of DI water. The sensing channels were formed by SWCNTs using a drop-casting method. A solution of SWCNTs and deionized water, mixed in a specific ratio, was dropwise applied onto the Si substrate and then dried on a hot plate. The solvent evaporated, leaving the SWCNTs deposited as a thin film on the substrate surface. The resultant network provided continuous percolation pathways between electrode contacts, forming a semiconducting conduction channel sensitive to surface interactions. Excess or loosely bound nanotubes were gently removed by rinsing with DI water to improve device consistency.

3. Electrode Deposition

The metal electrodes were patterned using standard photolithography and lift-off process. Contact exposure was performed with a Suss MicroTec MA6Gen4 Large-area high-resolution mask aligner for photolithography, and all processing was carried out at Gangneung-Wonju National University. A positive photoresist was spin coated onto the SWCNT-coated substrate and exposed through a photomask defining the electrode geometry. After development, a Ti/Au bilayer was deposited by a RF sputtering system. Titanium served as an adhesion layer to improve bonding between gold and the SiO₂ surface. The subsequent lift-off process produced well-defined electrodes in direct contact with the underlying SWCNTs layer, forming the sensor’s electrical contacts.

4. Lithium-Ion Decoration Process

For Li⁺ surface decoration, a 0.1 M lithium hydroxide (LiOH) solution was prepared by dissolving 0.01 mol of LiOH in 100 ml of methanol. Given the low solubility of LiOH in methanol, the mixture was subjected to ultrasonic dispersion for 30 minutes to fully dissolve the LiOH particles and produce a homogenous solution. This step was essential to prevent localized Li⁺ accumulation and ensure uniform distribution across the device.

Spin coating was employed to deposit the LiOH solution onto the SWCNTs surface. A two-step spin process was implemented: 1000 rpm for 5 seconds (spread), followed by 4000 rpm for 30 seconds (uniform film formation). Due to the high volatility of methanol, rapid spinning helped achieve thin, even LiOH layers. After each spin-coating cycle, the device was annealed at 100℃ for 2 minutes to evaporate residual methanol and promote adhesion of the Li⁺-layer. This coating-annealing process was repeated 3-5 times to build up a consistent Li⁺ network. The deposited LiOH formed a snowflake-like morphology, enabling enhanced CH₄ adsorption via induced dipole interactions between CH₄ molecules and Li⁺ ions.

Fig. 2 presents scanning electron microscopy (SEM) images of the sensor before and after Li⁺ decoration. Fig. 2(a) shows the bare SWCNTs channel, while Fig. 2(b) highlights the Li⁺-decorated sensor surface at the same scale. Closer inspection reveals clear morphological differences: Fig. 2(c) shows a uniform, featureless film for the bare SWCNTs at 30 µm scale, while Fig. 2(d) shows dendritic LiOH growth with distinct snowflake-like branches at 200 µm scale, validating the successful formation of the Li⁺ network.

The deposited LiOH formed a nanostructured snowflake-like morphology, as confirmed via optical microscopy, enabling enhanced CH₄ adsorption via induced dipole interactions between CH₄ molecules and Li⁺ ions. The morphology is critical because it increases the effective surface area and provides multiple adsorption sites, contributing to improved sensing performance.

Fig. 2. SEM images of the sensor before and after decorating with Li⁺: (a) Bare SWCNTs sensor device, (b) Li⁺/SWCNTs sensor device, (c) zoomed-in image of bare SWCNTs at 30 µm, (d) zoomed-in image of Li⁺/SWCNTs at 200 µm.

5. Measurement Setup

After fabrication and Li⁺ surface decoration, the electrical characteristics of the sensor were measured using a gas probe station with micro positioned probes inside a sealed chamber. Measurements were performed at Gangneung-Wonju National University using the MS TECH MST-5000 PROBE STATION (gas probe station). The sensor was exposed to CH₄ at concentrations ranging from 0.1 ppm to 50 ppm. A constant bias voltage of 1 V was applied across the electrodes, and the resulting current was recorded using a source measurement unit (SMU). Humidity levels during gas exposure were controlled at approximately 5%, 20%, 40%, 60%, and 80%, verified with a digital hygrometer. All measurements were performed at room temperature.

Resistance changes were monitored under a constant bias of 1 V using a SMU. The sensor’s response was quantified as sensitivity, calculated using the following expression ^[16]:

where $R_{air}$ is the baseline resistance in ambient air and $R_{gas}$ is the resistance upon CH₄ exposure.

The selected CH₄ concentrations (0.1, 1, 10, and 50 ppm) cover sub-ppm detection, intermediate values relevant to industrial leakage monitoring, and concentrations approaching safety critical levels, well below the lower explosive limit of methane (∼5% in air) ^[2]. This range enables performance evaluation under realistic and safety relevant conditions.

1. Sensitivity to Methane Gas

Table 1 summarizes the measured sensitivity values of both sensors at different CH₄ concentrations. While the bare SWCNTs sensor exhibited weak sensitivity values the Li⁺/SWCNTs sensor demonstrated consistently positive and enhanced sensitivities: 1.1% at 1 ppm, 1.46% at 10 ppm, and 1.57% at 50 ppm. Fig. 3 compares the I-T characteristics of both sensors under 10 ppm CH₄. The values shown in the upper-left corner of each panel (e.g., 13.3 ℃ / 33% and 15.6 ℃ / 32%) indicate the ambient temperature and relative humidity measured during the respective tests. In Fig. 3(a), the bare sensor exhibits a modest current change and almost no recovery. In contrast, Fig. 3(b) shows a stronger current shift for the Li⁺/SWCNTs device, along with clear evidence of recovery behavior (0.90%). The Li⁺/SWCNTs device showed a maximum relative increase in sensitivity of over 157% at 10 ppm compared to the undecorated sensor. This performance is attributed to induced dipole interactions between Li⁺ ions and CH₄ molecules, which promote greater adsorption and enhanced charge transfer to the SWCNTs channel.

Fig. 3. I-T curves for CH₄ exposure at 10 ppm: (a) Bare SWCNTs sensor shows 0.46% sensitivity and 0.01% recovery, (b) Li⁺/SWCNTs sensor shows 1.46% sensitivity and 0.90% recovery. Upper-left values: temperature (℃) and relative humidity (%).

The trend in sensitivity across concentrations from 0.1 ppm to 50 ppm is further visualized in Fig. 4. The bare SWCNTs sensor displays limited or even negative sensitivity at 0.1 ppm and 50 ppm, with a modest peak at 10 ppm. In contrast, the Li⁺/SWCNTs sensor consistently shows high sensitivity values above 1% across all tested concentrations. This consistent performance highlights the efficacy of Li⁺ decoration in enabling low concentration CH₄ detection with greater signal strength.

Fig. 4. Comparison of sensitivity between bare SWCNTs and Li⁺/SWCNTs sensors across CH₄ concentrations (0.1-50 ppm) at consistent relative humidity.

2. Recovery Behavior

The recovery performance of the sensor was also markedly improved with Li⁺ decoration. While the bare SWCNTs sensor exhibited negligible recovery (near 0%) across all tested concentrations, the Li⁺/SWCNTs sensor achieved recoveries of 0.58%, 0.90%, and 0.98% at 1 ppm, 10 ppm, and 50 ppm, respectively. Fig. 5 visualizes the recovery trend for both sensors across multiple CH₄ concentrations. The bare SWCNTs sensor shows consistently negligible recovery, close to 0% at all concentrations. In contrast, the Li⁺/SWCNTs sensor exhibits a clear increase in recovery percentage as CH₄ concentration rises, confirming its ability to desorb gas molecules and return toward baseline levels. The improvement is particularly evident at 10 and 50 ppm. Notably, a recovery of 1.00% was observed at 0.1 ppm, indicating responsiveness even at extremely low concentrations.

The improved recovery can be explained by the enhanced desorption dynamics due to the Li⁺ network, which facilitates more efficient CH₄ molecule detachment during the purge phase. This is consistent with prior theoretical insights that Li⁺ creates shallow binding states for gas molecules, enabling reversible adsorption.

Fig. 5. Recovery comparison between SWCNTs and Li⁺/SWCNTs sensors across CH₄ concentrations. Li⁺ decoration enhances desorption and baseline return across all cases.

3. Response Time and Humidity Effects

The response time of the Li⁺/SWCNTs sensor was measured to be approximately 15 seconds, a significant improvement over the 120 seconds observed for the bare SWCNTs sensor at 10 ppm. This improvement highlights the role of Li⁺ in accelerating adsorption kinetics.

The effect of humidity on the Li⁺/SWCNTs sensor’s CH₄ response was evaluated at 5%, 20%, 40%, 60%, and 80%. As summarized in Fig. 6, the sensitivity increases with humidity across all concentrations, with the highest value at 80%. At 50 ppm CH₄, the sensor exhibited a sensitivity of 8.25% at 80% humidity, compared to 0.60% under 20% humidity, indicating 13.75 times increase. This enhancement is attributed to water-assisted charge transfer: adsorbed H₂O donates carriers and modulates the SWCNTs surface potential, thereby facilitating CH₄ adsorption. Consequently, humidity-induced redistribution of surface charge further amplifies carrier modulation in the SWCNTs network, yielding higher sensitivity.

Device-to-device variability was evaluated by fabricating two nominally identical sensors (S1, S2) and testing them under identical conditions (CH₄ = 1, 10, 50 ppm; humidity = 20%, 40%, 80%; 24±1 ℃). Fig. 6 reports the mean standard deviation across S1 and S2 at 20%, 40%, and 80% humidity. By contrast, the 5% and 60% points are single-device measurements from the same process. Across concentrations, the two sensors exhibit comparable sensitivity levels, and minor offsets–including a crossover at 50 ppm–lie within the plotted spread. Overall, sensitivity increases with humidity and peaks at 80%. However, it was noted that in some cases, resistance did not fully return to the baseline after gas exposure. This may be attributed to residual oxidation or minor secondary reactions with ambient air upon opening the chamber. More stable performance may be achievable with nitrogen purging during the recovery phase.

Fig. 6. Humidity-dependent CH₄ sensitivity (1, 10, 50 ppm) of Li⁺/SWCNTs sensors at 5, 20, 40, 60, and 80% humidity. Error bars indicate the standard deviation. Points without error bars represent a single measurement. Measurements were performed at 24±1 ℃.

4. Summary of Sensor Performance

The Li⁺/SWCNTs sensor exhibited clear improvements in all evaluated metrics. At 10 ppm CH₄, sensitivity increased from 0.46% (bare) to 1.46%, while recovery improved from near 0% to 0.90%. At 1 ppm and 50 ppm, the bare sensor showed negative sensitivities (−0.04% and −0.08%, respectively), whereas the Li⁺/SWCNTs sensor maintained positive values of 1.10% and 1.57%. Response time was reduced from 120 seconds to 15 seconds. Under 60% humidity, sensitivity at 50 ppm improved by 2.44% compared to ambient conditions, confirming enhanced humidity tolerance.