1. Introduction
Global warming and climate change represent some of the most critical environmental challenges facing modern society. As part of global carbon neutrality initiatives, there is an urgent demand for technologies capable of precisely and continuously monitoring atmospheric carbon dioxide (CO
2) concentrations in real time [
1]. Although electronic gas sensors, including semiconductor- and electrochemical-based devices, dominate the current commercial market due to their high sensitivity and fabrication simplicity, they inherently suffer from vulnerability to electromagnetic interference (EMI) and limited long-term stability in harsh chemical environments or hazardous industrial settings with explosion risks [
2,
3]. In contrast, optical fiber sensors employ light as the signal transmission medium and are therefore completely immune to electromagnetic interference [
4]. In addition, they offer unique advantages such as compact size, remote sensing capability over long distances, and intrinsic safety in electrically hazardous environments. Owing to the excellent chemical resistance and thermal stability of silica-based optical fibers, these sensors can maintain high reliability even under corrosive and extreme operating conditions, making them highly attractive as next-generation platforms for environmental monitoring applications [
5,
6]. To further enhance the sensing performance of optical fiber sensors, extensive research efforts have been devoted to functionalizing the fiber cladding surface with various nanostructured and functional materials [
7,
8,
9]. Among various functional nanomaterials, two-dimensional transition metal dichalcogenides (TMDCs), exemplified by molybdenum disulfide (MoS
2), have attracted considerable attention owing to their atomically thin structures combined with exceptionally high specific surface areas, which significantly enhance the adsorption probability of gas molecules [
9,
10,
11,
12,
13]. When a 2D MoS
2 layer is coated onto the surface of an optical fiber, gas adsorption induces variations in surface charge density through charge transfer interactions. In addition, these changes subsequently modulate the complex refractive index of the MoS
2 layer, leading to measurable attenuation or phase shifts in the light propagating through the optical fiber. Conventional optical fiber-based gas sensors are often fabricated by chemically etching the cladding to expose the evanescent field and subsequently coating functional nanomaterials onto the etched region. However, such fully etched structures frequently suffer from non-uniform control of the cladding removal depth and mechanical fragility, as the thinned fiber can readily fracture during handling or operation [
12]. To overcome these limitations, we employed a D-shaped optical fiber architecture. Unlike approaches that remove the cladding circumferentially, the D-shaped configuration is formed by selectively polishing/etching only one side of the cladding, thereby preserving most of the cylindrical geometry and providing improved mechanical robustness, durability, and etching uniformity [
8,
14]. As a result, subtle refractive index variations induced by gas molecule adsorption can be sensitively transduced into measurable optical signals, enabling the realization of high-performance gas sensors. However, a critical issue addressed in this study arises from the intrinsic structural defects of MoS
2 thin films deposited by conventional sputtering processes. Due to the high-energy particle bombardment inherent to sputtering, the resulting films typically exhibit poor crystallinity. Consequently, an undesirable coexistence of the metastable metallic 1T (octahedral) phase and the semiconducting 2H (trigonal prismatic) phase is commonly observed [
15,
16]. In particular, sulfur vacancies (S-vacancies), which inevitably form during the deposition process, generate a high density of trap sites within the MoS
2 thin film. These trap states impede charge transport and promote irreversible interactions with gas molecules, thereby severely degrading the recovery characteristics of the sensor [
17,
18]. A sensor surface containing a high density of the metallic 1T phase and sulfur vacancies may exhibit a rapid initial response upon gas adsorption; however, incomplete desorption of gas molecules during the recovery process leads to a gradual baseline drift. This drift behavior originates from irreversible adsorption and trapped charge states, which prevent full restoration of the sensor to its initial state. Therefore, to realize high-performance CO
2 sensors with stable and repeatable responses, it is essential to convert the unstable 1T phase into the thermodynamically stable semiconducting 2H phase and to heal sulfur-related defects through sulfur-rich annealing processes [
19]. The high crystallinity of the 2H phase enhances charge carrier mobility, thereby improving the signal-to-noise ratio (SNR), while the healed surface enables reversible electron transfer with gas molecules, leading to a marked improvement in sensor reproducibility and long-term stability [
20].
In this work, MoS2 thin films sputter-deposited on SiO2/Si substrates were thermally treated under a sulfur-rich atmosphere to form a highly crystalline 2H-MoS2 phase. This process suppresses the metallic 1T phase and compensates sulfur-related defects, thereby optimizing physicochemical interactions with gas molecules. The treated films were then integrated onto the D-shaped optical fiber through a precise transfer process to construct a CO2 sensor. Comprehensive structural and chemical characterizations by Raman spectroscopy) were performed before and after phase engineering, together with practical gas sensing evaluations. Ultimately, this study aims to elucidate how sulfur-vacancy suppression promotes reversible adsorption of CO2 molecules and how such adsorption dynamics influence the evanescent-field-based optical sensing performance.
3. Results and Discussion
Figure 3 shows the MoS
2-coated D-shaped optical fiber sensor. The surface morphology and deposition uniformity of the sputter-deposited MoS
2 sensing layer were examined using high-resolution scanning electron microscopy (SEM). Unlike solution-based processes or mechanical exfoliation methods, the DC sputtering technique employed in this study produced a highly uniform and continuous film on both SiO
2/Si substrates and the polished surface of the D-shaped optical fiber as shown in
Figure 3b. Such a continuous film morphology is essential to ensure uniform interaction between the evanescent field of the optical fiber and the sensing layer, thereby contributing critically to signal stability and reproducibility. As clearly observed in
Figure 3a,b and
Supplementary Figure S1, a large-area, continuous MoS
2 film was successfully and uniformly transferred onto the region above the fiber core. Raman spectroscopy was performed to investigate the crystal structure of the MoS
2 thin films deposited at different substrate temperatures. As shown in
Figure 3c, characteristic in-plane (E
12g) and out-of-plane (A
1g) vibrational modes of MoS
2 are clearly observed at approximately 380 cm
−1 and 405 cm
−1, respectively, indicating successful deposition of MoS
2 thin films via sputtering. In addition, the peak separation between the E
12g and A
1g Raman modes exceeds approximately 30 cm
−1, indicating that the deposited MoS
2 film possesses a thickness on the approximately 65 ± 15 nm (
Supplementary Figure S2). However, an additional Raman feature is detected near 200 cm
−1, which is commonly associated with the presence of the metallic 1T phase coexisting with the semiconducting 2H phase [
17,
21]. Although previous studies have reported that increasing growth temperature can promote partial conversion of the 1T phase into the 2H phase, the sputtered films in this work still exhibit residual 1T-related signals, even at elevated deposition temperatures [
18]. This behavior can be attributed to differences in sputtering yields between molybdenum and sulfur atoms from the sputtering target. During sputtering, accelerated argon ions preferentially displace sulfur atoms due to their lower atomic mass, leading to sulfur deficiency and local stoichiometric imbalance within the film. Such non-stoichiometry induces lattice distortion, which manifests as 1T-related features in Raman spectra [
22]. The metallic 1T phase and sulfur vacancies are known to hinder reversible adsorption–desorption processes of moisture and reactive gas species, thereby degrading sensor response speed, recovery behavior, and overall sensitivity. To verify this effect, both mixed-phase (1T@2H) and phase-pure 2H MoS
2 thin films were evaluated in terms of their electrical CO
2 response and optical fiber sensor performance. To eliminate sulfur vacancies and stabilize the 2H crystal structure, an annealing process was introduced. This process proceeds via two dominant mechanisms. First, thermal energy induces atomic rearrangement, driving a complete phase transition from the metastable 1T structure to the thermodynamically stable 2H phase. Second, sulfur atoms supplied from the sulfur-rich atmosphere effectively compensate sulfur vacancies within the lattice, resulting in defect healing. To investigate the elimination of sulfur vacancies through atomic recrystallization induced by high-temperature annealing, thermal treatments were performed under an N
2 atmosphere while varying the annealing temperature from 500 to 750 °C. As shown in
Figure 3d, the intensities of the A
1g and E
12g Raman modes gradually increase with increasing annealing temperature, indicating enhanced atomic rearrangement and recrystallization at elevated temperatures. However, despite the improvement in crystallinity, the Raman peak associated with the metallic 1T phase remains clearly observable even after high-temperature N
2 annealing. This result suggests that thermal energy alone is insufficient to fully suppress the 1T phase and eliminate sulfur-related defects. Therefore, an additional supply of sulfur is required during the recrystallization process to effectively compensate sulfur vacancies. Based on this observation, subsequent annealing was conducted under a sulfur-rich atmosphere to simultaneously promote phase conversion and defect healing. As shown in
Figure 3d, Raman spectra after sulfur annealing exhibit the complete disappearance of the 1T-related peak near 200 cm
−1, accompanied by significantly sharpened and intensified E
12g and A
1g peaks, confirming the formation of a highly crystalline 2H-MoS
2 thin film. The resulting high-quality 2H-MoS
2 sensing layer is expected to promote reversible charge transfer interactions with CO
2 molecules, thereby markedly enhancing sensor responsiveness and stability. X-ray photoelectron spectroscopy (XPS) analysis was performed to verify the reduction of sulfur (S) vacancies and the improvement of MoS
2 crystallinity (
Supplementary Figure S3). As illustrated in the spectra, the characteristic peaks of Mo and S were significantly enhanced after the sulfur-rich annealing process. Furthermore, the stoichiometric S/Mo ratio, calculated from the peak area ratios, increased from 1.62 to 2.00 following the treatment. These results confirm that annealing in a sulfur-rich atmosphere is a highly effective method for reducing sulfur defects and restoring the stoichiometry of the sputtered MoS
2 films.
To investigate the influence of sulfur-annealing-induced crystal structure evolution on CO
2 sensing behavior, electronic sensors were fabricated using MoS
2 thin films with mixed 1T@2H and phase-pure 2H crystal structures, and their sensing responses were systematically evaluated. This comparative study was conducted to analyze the intrinsic limitations of electronic sensors and to benchmark their performance against that of optical fiber-based sensors. As shown in
Figure 4a,b, the chemiresistor-type electronic sensors exhibit a pronounced dependence of sensing behavior on the structural state of the MoS
2 thin films. Because reliable sensing responses were difficult to resolve at low CO
2 concentrations, the sensor performance was initially evaluated at a fixed CO
2 concentration of 40,000 ppm to clearly assess its intrinsic sensing behavior. In the as-sputtered sensor (1T@2H MoS
2), although a current response is observed upon CO
2 exposure, the baseline current fails to recover to its initial level after gas removal, resulting in a severe baseline drift. This behavior is attributed to the high density of sulfur vacancies distributed throughout the film, which act as high-energy trap sites and promote irreversible chemical interactions with CO
2 molecules. Charges trapped at these defect sites cannot be efficiently released during the desorption process, thereby significantly impairing signal reversibility. For the sulfur-annealed electronic sensor, the conversion of the mixed-phase structure into the semiconducting 2H phase leads to an improvement in the magnitude of the CO
2 response. However, due to the nanoscale thickness of the conductive channel, substantial electrical noise is introduced by thermal fluctuations on the highly crystalline surface and contact resistance at the electrode–film interfaces [
23]. Such electrical noise hinders the precise extraction of small current variations associated with subtle changes in CO
2 concentration, indicating that the intrinsic advantages of highly crystalline MoS
2 cannot be fully exploited in electronic sensor configurations. In addition, accurately resolving the sensor response at low CO
2 concentrations remains challenging due to the limited signal variation. To overcome these limitations, optical fiber-based CO
2 sensors were fabricated and evaluated, as shown in
Figure 4c,d. In the case of the non-annealed MoS
2 sample, reliable CO
2 responses were difficult to resolve, which is consistent with the electronic sensor results and can be attributed to suppressed desorption caused by sulfur-vacancy-mediated irreversible adsorption. In contrast, the D-shaped optical fiber sensor incorporating sulfur-annealed MoS
2 exhibited a remarkable enhancement in sensing performance and excellent operational stability. The highly crystalline 2H-MoS
2 structure formed through sulfur annealing effectively suppresses sulfur vacancies, which plays a crucial role in ensuring reversible gas adsorption–desorption processes. The defect-minimized 2H-phase surface provides an optimized surface energy landscape that facilitates reversible adsorption and desorption of CO
2 molecules, thereby maintaining highly consistent and repeatable responses without baseline fluctuation during repeated gas exposure cycles. Notably, because the optical fiber sensor does not rely on direct charge transport measurements but instead exploits light–matter interactions mediated by the evanescent field, it is inherently immune to the electrical noise that critically limits electronic sensors. Stable dielectric property variations associated with the 2H phase are directly converted into optical intensity changes, resulting in a substantially higher signal-to-noise ratio (SNR) and enabling reliable detection of even subtle CO
2 concentration changes. The observed CO
2 sensing mechanism can be explained by charge transfer interactions between the MoS
2 thin film and gas molecules. MoS
2 exhibits typical n-type semiconductor behavior, while CO
2 molecules act as electron acceptors. In electronic sensors, adsorption of CO
2 extracts electrons from the MoS
2 channel, reducing carrier density and electrical conductivity, which manifests as a decrease in current. In optical fiber sensors, this carrier depletion induces changes in the complex refractive index (n + ik) of the MoS
2 sensing layer. Variations in carrier concentration affect both the refractive index (
n) and extinction coefficient (
k), thereby modulating absorption and scattering of the evanescent field exposed outside the fiber core. Sulfur-annealing-induced phase conversion from the metastable 1T phase to the stable 2H phase, together with effective suppression of sulfur vacancies, significantly enhances the physicochemical stability of MoS
2 thin films. When integrated with the evanescent field of a D-shaped optical fiber, this strategy overcomes the inherent limitations of electronic sensors and enables the realization of CO
2 sensors with both high sensitivity and excellent long-term stability. The high-quality 2H-MoS
2-based D-shaped optical fiber sensor developed in this work demonstrates superior gas responsiveness and recovery characteristics compared with conventional technologies. In contrast, as-sputtered mixed-phase (1T@2H) films inevitably form irreversible bonds with CO
2 molecules at sulfur-deficient sites due to differences in sputtering yields between molybdenum and sulfur, leading to persistent baseline drift during gas desorption. The optical fiber sensor subjected to sulfur annealing the central strategy of this study clearly demonstrates effective defect healing. Based on the structural stability of the 2H phase, reversible charge transfer interactions with CO
2 molecules are achieved, resulting in complete baseline recovery without fluctuation even at room temperature.
Figure 5a summarizes the change in optical output as a function of CO
2 concentration, based on the data obtained from
Figure 4d. A monotonic increase in optical signal is observed with increasing CO
2 concentration, confirming that the proposed sensor responds continuously over a wide concentration range. This concentration-dependent optical modulation indicates that variations in the optical properties of the sensing layer are directly correlated with CO
2 adsorption, providing a reliable basis for quantitative gas sensing.
Figure 5b presents the temporal response of a commercial non-dispersive infrared (NDIR) CO
2 sensor measured under identical conditions for comparison. Although the NDIR sensor exhibits a detectable response upon CO
2 exposure, a pronounced delay in both response and recovery is clearly observed. In particular, the recovery process is significantly prolonged, requiring several tens of seconds for the signal to return toward its baseline after CO
2 removal. This slow recovery behavior originates from the diffusion-limited sensing mechanism of NDIR sensors, in which gas exchange within the optical chamber governs the overall response dynamics. In contrast, the sulfur-annealed MoS
2-coated D-shaped optical fiber sensor exhibits a comparable response time and a significantly faster recovery behavior compared to conventional NDIR technology as shown in
Figure 5c. The optical signal returns to its baseline within only 9 s after CO
2 removal, demonstrating near-complete and fluctuation-free recovery even under repeated gas exposure cycles. This rapid recovery is a direct consequence of reversible charge transfer interactions between CO
2 molecules and the highly crystalline 2H-MoS
2 sensing layer, enabled by effective sulfur-vacancy healing.
The dramatic difference in recovery time highlights a critical advantage of the proposed optical fiber sensor over conventional NDIR technology. While NDIR sensors inherently suffer from slow desorption and gas exchange processes, the evanescent-field-based sensing mechanism allows immediate transduction of adsorption and desorption events into optical signals without reliance on bulk gas diffusion. As a result, the proposed sensor achieves a substantially higher signal-to-noise ratio (SNR) and enables reliable detection of rapid CO2 concentration fluctuations, which is particularly advantageous for real-time environmental monitoring and industrial safety applications.