Chloroplatinic acid (H₂PtCl₆), ascorbic acid (AA), ammonium molybdate tetrahydrate ((NH₄)₆Mo₇O₂₄·4H₂O), aniline, L-Cysteine, hydrochloric acid (HCl) and ethanol (C₂H₅OH) were analytical grade and purchased from Shanghai Chemical Corp. All reagents in the experiment were used without any further purification. All solutions were prepared in deionized water. Gases including hydrogen sulfide (H₂S), nitrous oxide (N₂O), xylene (C₈H₁₀), nitrogen dioxide (NO₂), nitric oxide (NO), hydrogen (H₂), ammonia (NH₃) and sulfur dioxide (SO₂) were obtained from Shanghai Shenkai Gases Technology Co., Ltd.

X-ray diffraction patterns (XRDs) were collected by using powder XRD (Rigaku D/MAX-2500 X-ray diffractometer, with Cu Kα₁ radiation λ = 0.154056 nm 40 kV and 40 mA, scanning from 10° to 70°). The microtopography of the synthesized materials was recorded with scanning electron microscopy (SEM, JSM-6700F) and high-resolution transmission electron microscopy (HRTEM, JEM-2100F, Japan). The high-resolution TEM and high-angle annular dark-field scanning TEM (HAADF-STEM) images were recorded on an FEI Tecnai G2 F20 S-Twin high-resolution transmission electron microscope (HRTEM; Hillsboro, OR, United States) set at 200 kV, and a JEOL JEM-ARM300F TEM/STEM (Tokyo, Japan), respectively, with a spherical aberration corrector worked at 300 kV. Through-focal HAADF series were acquired at nanometer intervals, with the first image under-focused (beyond the beam exit surface) and the final image over-focused (before the beam entrance surface). Then, the images were aligned manually to remove the sample drift effects. Electron paramagnetic resonance (EPR) spectra were recorded on a Bruker CW Elexsys E500 spectrometer by applying an X-band (9.4 GHz) microwave with sweeping magnetic field at room temperature.

The MoS₂ were firstly synthesized by an anion-exchange strategy by using MoO₃-ANE (aniline-functional MoO₃) as precursors (Fig. S1). The SEM, TEM and HRTEM images in Fig. S2 reveal a uniform hierarchical nanosheet configuration with a well-ordered lattice fringe spacing of 0.27 nm, which agrees well with the (100) facet of hexagonal molybdenum disulfide. Subsequently, the Pt species were introduced through a deposition-precipitation process. As seen in Fig. S3, the high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image shows the atomic dispersion of Pt species over the Pt₁-MoS₂ surface and no notable Pt clusters are observed. Driven by thermal treatment of Pt₁-MoS₂ in a reductive atmosphere, Pt₁-MoS₂-def catalyst was synthesized. No remarkable architectonic changes and Pt nanoparticles are found (Fig. S4a, b). Meanwhile, the energy-dispersive X-ray spectroscopy (EDS) mapping (Fig. S4c) further demonstrates the homogeneous distribution of Pt, Mo and S elements. Furthermore, HAADF-STEM (Fig. 2a) clearly shows that the individual Pt atoms atomically disperse over the whole support surface without agglomerate, which is further confirmed by the extractive X-Y intensity profiles and atom-overlapping Gaussian-function fitting images of selective area scan (Fig. 2b-d). The simulating configuration (Fig. 2f) clearly exhibits the distribution of single dispersed Pt atoms, it is in accordance with the STEM image (Fig. 2e). For comparison, Pt nanoparticle modified MoS₂ (Pt NPs-MoS₂) was also synthesized through the ascorbic acid reduction in Fig. S5. More representative SEM, TEM, XRD and BET images of Pt₁-MoS₂, Pt₁-MoS₂-def, Pt NPs-MoS₂ are displayed in Figs. S6-S9. Moreover, the Pt loading contents of Pt₁-MoS₂-def and Pt NPs-MoS₂ are measured as 0.1% by inductively coupled plasma optical emission spectrometry (ICP-OES) (Table S1).

Element-selective X-ray absorption fine structure (XAFS) was further performed to explore the coordination environment of Pt species. As seen in Fig. 2g, the X-ray absorption near edge structure (XANES) spectra reveal the white line intensities of Pt₁-MoS₂-def and Pt₁-MoS₂ peaks are located between the PtS₂ and Pt foil, suggesting the oxidation state of Pt^δ+ (0 <  δ < 4). The Fourier transferred extended X-ray absorption fine structure (FT-EXAFS) in Fig. 2h displays two notable peaks at ~ 1.90 Å, which are contributed by the Pt-S coordination path. No Pt-Pt characteristic peak (at ~ 2.57 Å) is detected for both Pt₁-MoS₂-def and Pt₁-MoS₂ samples, confirming the atomic dispersion of Pt on the supports. Moreover, the first-shell EXAFS fitting curves (Fig. 2i) and corresponding fitting results (Fig. S10 and Table S2) validate that the central Pt atom in both Pt1-MoS2 and Pt1-MoS2-def samples directly coordinates with three S atoms with the average bond lengths of 2.32 and 2.33 Å, respectively. Moreover, the wavelet transforms spectra of Pt EXAFS oscillations in Fig. 2j display only one peak at 7 Å⁻¹ for Pt₁-MoS₂-def and Pt₁-MoS₂ samples, regarded as the Pt-S coordination contribution, indicating the high-distributed status of single Pt species.

The first-principal calculations were further conducted to verify the Pt-site configuration on the support. Figure 3a simulates the diffusion path of the Pt atom on the MoS₂ surface, including ①Pt on the S site, ②Pt on the hollow site and ③Pt on the Mo site. The results of the DFT calculation (Fig. S11) show that the Pt atom can stabilize on the top of the Mo site with the lowest binding energy of − 3.265 eV in contrast to − 2.764 eV (Pt on the S site) and 6.024 eV (Pt on the hollow site). Also, this configuration of Pt atom bonding with three S atoms is consistent with the XAFS fitting results. After further treating the Pt₁-MoS₂ in a reduced atmosphere, the S atoms will spontaneously escape from the MoS₂ surface to generate sulfur vacancy sites (Pt₁-MoS₂-def). This process will induce five potential configurations in Fig. S12. Considering the adjudication of experimental investigations (detailed descriptions in supporting information), the defect 2 model (adjacent non-bonding S defect) was selected as the optimal configuration with the lowest vacancy formation energy of − 1.876 eV. This spontaneous reaction process (Fig. S13) indicates that the introduced single Pt atom can promote the evaporation of adjacent S species. Importantly, this three-coordinated structure of Pt₁-MoS₂-def is also well conforming by XAFS results. Furthermore, the theoretical deduction can be verified by the experimental electron paramagnetic resonance (EPR) analysis. As shown in (Fig. 3b), the signal intensity of sulfur vacancies (at g = 2.003) of Pt₁-MoS₂-def is significantly higher than that of MoS₂-def [29] (more synthetic process showed in supporting information), demonstrating single Pt atoms can induce the formation of sulfur vacancies and thus activate the whole S plane.

The XPS measurements are employed to unveil the chemical environment of Mo (Fig. 3c), S (Fig. 3d) and Pt (Fig. 3e) species. As shown in Fig. 3c, the peaks at 231.9 eV (Mo 3d_3/2) and 228.7 eV (Mo 3d_1/2) are attributed to Mo⁴⁺ and the peaks at 233.9 eV (Mo 3d_3/2) and 230.7 eV (Mo 3d_1/2) are assigned to Mo⁵⁺ [30], respectively. After introducing the Pt species, the doublet peaks of both Mo⁴⁺ (232.1 and 228.9 eV) and S²⁻ (161.9 and 163.2 eV) in the Pt₁-MoS₂ shift to higher binding energy (0.2 and 0.3 eV left shifts, respectively) compared with pure MoS₂. Meanwhile, the binding energy of Pt 4f peak is higher than Pt⁰ (71.2 eV, Fig. S14) but lower than that of Pt⁴⁺ (74.0 eV) [26] according to the XAFS results. For Pt₁-MoS₂-def catalysts, the double peaks of Mo⁵⁺ (233.9 and 230.7 eV) and the high valence Sⁿ⁺ peaks (164.5 and 169.2 eV) gradually disappear. The Pt peaks in Pt₁-MoS₂-def (72.8 eV) shift toward lower binding energy (0.3 eV) in contrast to the Pt₁-MoS₂ sample. The theoretical planar-averaged electron density difference ∆ρ(z) and Bader charge transfer in Fig. 3f further verify that the Pt atom in the as-established Pt₁-MoS₂-def model can reduce the electronic loss toward supports, according well to experimental XPS investigation. Both experimental and theoretical results reveal that activated S species will provide a feedback role in tailoring the electronic states of Pt atoms. In conclusion, the single Pt sites on the MoS₂ surface can induce easier volatilization of adjacent S species and the activated S species can inversely impact the electronic configuration of Pt atoms, thus leading an intact feedback-regulation system.

The sensing properties of the MoS₂, Pt₁-MoS₂ and Pt₁-MoS₂-def MEMS sensors were investigated at room temperature (RT). The dynamic response transitions curves of the sensors along with SO₂ concentrations in the range of 0.5-40 ppm are exhibited in Fig. 4a, b. Obviously, the Pt₁-MoS₂-def sensor displays the highest response values toward different concentrations of SO₂ compared with Pt₁-MoS₂ and MoS₂ sensors. Some baseline drift of the three sensors at room temperature could be attributed to the incomplete desorption of the SO₂ molecule [31]. Especially, the Pt₁-MoS₂-def sensors (25% to 5 ppm SO₂) exhibit an enhanced response approximately four times and five times higher than that of Pt₁-MoS₂ (7% to 5 ppm SO₂) sensors and MoS₂ sensors (5% to 5 ppm SO₂), respectively. Figure 4c shows the linear relationships between responses and SO₂ concentrations within 0.5-40 ppm for three sensors. Impressively, the Pt₁-MoS₂-def sensors exhibit an exceptional linear section with increased SO₂ concentrations, while the MoS₂ sensors endure the erratic response-concentration relationship. Additionally, the experimental limit of detection (LOD) of the Pt₁-MoS₂-def sensor can reach as low as 500 ppb, whereas the Pt₁-MoS₂ sensors and MoS₂ sensors cannot perceive SO₂ gas below 1 ppm. Such a distinguished response performance exceeds that of Pt NPs-MoS₂ sensors (Fig. S15) and most reported SO₂ sensors (Fig. 4d and Table S4). More dynamic response-recovery resistance curves of MoS₂, Pt₁-MoS₂ and Pt₁-MoS₂-def sensors are shown in Fig. S16. As shown in Fig. 4e, eight kinds of typical vaporous molecules including sulfur dioxide (SO₂), ammonia (NH₃), hydrogen (H₂), sulfuretted hydrogen (H₂S), nitric oxide (NO), xylene (C₈H₁₀), nitrogen dioxide (NO₂) and nitrous oxide (N₂O) were detected at the same concentration of 5 ppm at RT. It is found that the responses of interfering gases are significantly negligible. Furthermore, the calculated selectivity coefficients (K) of all sensors (Fig. S17) reveal relatively high K values, indicating excellent SO₂ selectivity for gas-sensing performances. The adsorption (k_ads) and desorption (k_des) rate constants of all sensors (Fig. 4f) were further calculated by fitting the response versus time curves (Fig. S18) using the equations below:

where R_max is the maximum response value, R₀ is the response in air, C_a is the concentration of SO₂ gas, t is the time and K is the equilibrium constant (k_ads/k_des) [32]. The results indicate that the kinetic adsorption and desorption of SO₂ molecules are promoted by Pt species and sulfur vacancies. In addition, the cycle stability (Fig. 4g) and long-term durability measurements (Fig. S19) show that the Pt₁-MoS₂-def sensor can remain almost unchanged responses during continuous eight-cyclical exposure of SO₂ and superior stability for up to 35 days. The effects of operating temperature for Pt₁-MoS₂-def sensor were conducted in Fig. S21, in which the sensors can realize optimal gas-sensing performance at room-temperature conditions. Besides, the resistances of the Pt₁-MoS₂-def sensor may be affected by the humidity environment (Fig. S22), which is mainly because water molecules might react with SO₂ molecules. A water filtration membrane to filter out water molecules was added on the device (Fig. S23a). As shown in Fig. S23b, c, after adjusting the device, the response of the material to 25 ppm SO₂ under different humidity conditions were tested. It was observed that under 90% humidity, the sensor’s response decreased by no more than 9% (before the improvement, the sensor’s response decreased by about 28%), effectively improving the normal operation of the device in high humidity environments.

To further clear the design superiority of the Pt₁-MoS₂-def catalysts in the gas-sensing reaction, SO₂ adsorption and desorption tests were implemented (experimental details are shown in Methods and Table S3). As Fig. 5a-c shows, the Pt₁-MoS₂-def reveals the maximum adsorption capacity (2.34 wt%) and the fastest average adsorption rate (0.52 wt% s⁻¹) for SO₂ compared with MoS₂ (1.32 wt%, 0.18 wt% s⁻¹), Pt₁-MoS₂ (1.62 wt%, 0.41 wt% s⁻¹) and Pt NPs-MoS₂ (1.75 wt%, 0.44 wt% s⁻¹) at room temperature, indicating the S vacancy-assisted single Pt sites (Pt-Vs) can induce superior adsorption and excitation ability for sulfur dioxide. Ex situ XPS spectra were further employed to investigate the reaction mechanism of different catalysts. As seen in Fig. 5d, the peaks of Pt⁴⁺ in Pt₁-MoS₂ suddenly appear in the SO₂ atmosphere while no obvious new peaks are found in the Mo and S species, confirming the electron transfer path is from the single Pt site to adsorbed SO₂ molecule. However, a small number of surface-adsorbed SO₂ molecules are not distinguished by the XPS detection. Same results can also be found in MoS₂-def catalysts in Fig. S24. Considering the resistances of MoS₂-based sensors significantly reduce in the oxidizing SO₂ atmosphere (Fig. S25), the sensing type can be identified as p-type semiconductors. Detailly, the SO₂ species can capture electrons from the conduction band of p-type semiconductors and induce the hole generation, resulting in decreased resistances (Fig. S26). For the Pt₁-MoS₂-def catalyst (Fig. 5e), two new higher-valence S peaks (168.83 and 170.1 eV, S⁶⁺) and Pt⁴⁺ peaks are discovered after SO₂ treatment while there is a slight high binding energy shift in the Mo 3d spectrum, revealing that both Pt and MoS₂ supports lose electrons during the gas-sensing process. The in situ Raman spectra display the 150 s exposure periods of Pt₁-MoS₂-def, Pt₁-MoS₂ (Fig. 5f) and MoS₂-def (Fig. S27) after treating with 1000 ppm SO₂. Clearly, the A_1g peak (at 402.25 cm⁻¹) of the Pt₁-MoS₂-def sample shifts to a higher wavenumber after SO₂ exposure for 60 s, while no peak changes for the Pt₁-MoS₂ and MoS₂-def samples. Therefore, this unequivocally corroborates that the electronic transform process in Pt₁-MoS₂-def involves the adsorbed SO₂ molecules and the whole supports. In conclusion, experimental results demonstrate there are two potential reaction paths dominant to the SO₂ sensing performance for Pt₁-MoS₂-def and Pt₁-MoS₂ sensors. For Pt₁-MoS₂ (Fig. 5g, type I), the electrons mainly locate around the single Pt site and the electron transfer only occurs within Pt-S (SO₂). In comparison, the Pt₁-MoS₂-def sensors contain synergistically single Pt sites and activated inert S plane, which prompts the whole supporting surface to participate in the SO₂ sensing process (Fig. 5g, type II). Meanwhile, the rapid electron transfer between Pt₁-MoS₂-def and the oxidizing SO₂ guarantees high response and low LOD. In addition, the in situ SO₂ adsorption breakthrough curves (Fig. 5h) were implemented to quantitatively measure the absorption of SO₂, in which the adsorbed SO₂ amounts on Pt₁-MoS₂-def is 281 μmol g⁻¹ more than that on Pt₁-MoS₂, and there is 292 μmol g⁻¹ more than that on MoS₂-def at 25 °C, according well with the SO₂ adsorption and desorption tests in Fig. 5a-c.

DFT calculations were conducted to explore the structure-activity relationships between single Pt sites and gas-sensing performance. As seen in Fig. 6a, the Pt 4d partial DOS (pDOS) results show that the d-band center in Pt₁-MoS₂-def (-2.58 eV) obviously upshifts compared with the Pt₁-MoS₂ (− 4.10 eV). The corresponding DOS images are displayed in Fig. S28. When interacting with SO₂ molecule, the Pt-S (SO₂) projects crystal orbital Hamilton population (pCOHP) is calculated in Fig. 6b, where the negative -pCOHP represents the antibonding contribution and the positive -COHP stands for bonding contribution [33,34,35]. Obviously, the antibonding contribution gradually moves upward (compared to E_f) from Pt₁-MoS₂-def to Pt₁-MoS₂, proving a decreased occupation state and a strengthening Pt-S bond. The -IpCOHP (the integral of -COHP up to the Fermi level) values are further calculated in Fig. 6c. The antibonding orbital occupation state of the Pt-S bond in Pt₁-MoS₂-def is smaller than that in Pt₁-MoS₂, indicating more stable adsorption of SO₂ for Pt₁-MoS₂-def, which is conformed to the d-band center theory. In addition, the density of states (DOS) analysis in Fig. 6d displays that a new peak around Fermi energy level could be observed in SO₂/Pt₁-MoS₂-def, suggesting more electrons transfer between SO₂ and Pt₁-MoS₂-def compared with SO₂/Pt₁-MoS₂ and SO₂/MoS₂ models. This can be further verified by the differential charge density comparisons in Fig. 6e, f. Also, the Pt₁-MoS₂-def reveals larger the adsorption energy (E_ads) of -2.40 eV than that of the Pt₁-MoS₂ (− 1.17 eV) and MoS₂ (− 0.035 eV) upon interaction with SO₂ molecules (Fig. S29), according well with gas-sensing performance. Overall, utilizing Pt species to promote adjacent S evaporation, the electronic state of S vacancy-assisted single Pt atom pronouncedly changes and thus induces a gradually decreased antibonding contribution state of Pt-S and a stronger SO₂ adsorption. Hence, the establish of feedback-regulation system is crucial for SO₂-sensing material with high response and low LOD (Fig. 6g).

As we all known, the real-time monitor, track and prediction of stressful circumstances (such as SO₂) is of vital importance to optimize and adjust plant growth [8]. For instance, an appropriate SO₂ environment can promote seed germination and stomatal movement, while excessive SO₂ concentrations will disturb the physiological and biochemical metabolism [36,37,38]. Therefore, engineering multifunctional SO₂ sensors array to timely collect and transmit the gas/temperature/humidity signals is desired for plant cultivation process (Fig. 7a). Based on the superior SO₂-sensing materials of Pt₁-MoS₂-def, a miniaturized, integrated and wireless sensing SO₂ sensors device is further developed in Fig. 7b, c. The disk-shaped device is composed of Bluetooth, Active Front End (AFE) and Micro Controller Unit (MCU) on one side, while a MEMS array (containing six sensors) and temperature/humidity modules are embedded at the back. Wireless data transmission from devices to smartphones can be achieved through the Bluetooth system. The corresponding block diagram is presented in Fig. 7d, including the power delivery, data conditioning and data transmission pathways. Impressively, synchronous six sensing signals can be received and revealed on the screen to calibrate the errors. More details of printed circuit board and the main interfaces of the custom mobile application device are shown in Figs. S30-S33. Furthermore, the portable sensing device displays constant responses for 5 and 10 ppm SO₂ gases (Fig. 7e) in a realistic greenhouse environment. More gas-sensing performance tests are shown in Fig. S34. Notably, the current design still stays in a proof-of-concept stage, it may require more optimizing process including signal handling, data integration, intellectual detection and so on in future work.