NiO_x nanoparticles were synthesized via chemical co-precipitation. First, 0.125 mol of Ni(NO₃)₂•6H₂O was dissolved in 25 mL of deionized water and stirred for 30 min to obtain a dark green transparent solution. A 10 mol L^-1 KOH solution was added dropwise to the solution until the pH reached 10. After another 30 min of stirring, the bright green turbid solution was centrifuged and washed three times with deionized water at 1,500 rpm. The precipitate was redispersed in water, dried at 80 °C for 8 h, ground and sieved through a 300-mesh screen to yield Ni(OH)₂ nanopowder. This powder was calcined at 240 °C for 2 h to yield black NiO_x nanoparticles (top row, Supplementary Figure 1).

Bi₂Te₃ nanoparticles were synthesized hydrothermally. First, 0.96 g of polyvinyl pyrrolidone (PVP) was dissolved in 42 mL of ethylene glycol under stirring for 30 min. Subsequently, 1.18 mmol of Bi₂O₃ and 3.6 mmol of TeO₂ were introduced and stirred for another 45 min, followed by the addition of 6 mL of 4 mol NaOH solution, yielding a yellow suspension after stirring at 300 rpm at 25 °C for 30 min. This suspension was transferred to a hydrothermal reactor and kept at 200 °C for 4 h. After natural cooling, the product was centrifuged with ethanol at 2,000 rpm for 4 min; this process was repeated three times to ensure purity. The product was then dispersed in ethanol, vacuum-dried at 60 °C for 8 h, ground, and sieved through a 250-mesh sieve to obtain a grayish-black Bi₂Te₃ nanopowder.

Bi₂Te₃ nanopowder was mixed with NiO_x nanopowder at mass fractions of 0, 1, 5, 10, and 15 wt%. The mixtures were dispersed in solution at 20 mg/mL and ultrasonicated for 30 min. NiO_x sensors with varying Bi₂Te₃ doping levels were fabricated by drop-casting the respective dispersions onto the sensor chip using disposable syringes. The resulting sensors were labeled as 1%, 5%, 10%, and 15% Bi₂Te₃@NiO_x sensors.

Supplementary Figure 1 (bottom) presents the sensor testing system, in which the micro-electro-mechanical systems (MEMS) chip is wired to a four-pin in-line socket base for performance evaluation. The prepared powder was weighed, mixed with anhydrous ethanol, and sonicated to form a uniform dispersion, which was then drop-cast onto the MEMS chips using a syringe. The coated MEMS devices were placed in an eight-channel test chamber for simultaneous testing of up to eight chips. The chamber included temperature and humidity sensors, with an external direct current power supply driving the test electrodes. Heating was regulated via a wireless-controlled module, and output signals were processed by dedicated software and transmitted wirelessly to a computer for real-time data monitoring. The computer terminal automatically adjusted multiple mass flow controllers according to preset gas parameters to maintain stable target gas concentration. Relative humidity was regulated by passing dry air through a humidity generator [Supplementary Figure 1]. Sensor response to reducing gases is defined as the ratio of resistance in the target gas to that in air, and inversely for oxidizing gases. Response time is the duration required to reach 90% of the response value, while recovery time is the time needed to return to 10% of the response value.

The crystal structures of NiO_x and Bi₂Te₃@NiO_x were analyzed via wide-range X-ray diffraction (XRD) patterns (Bruker D8 ADVANCE) using Cu Kα radiation at 40 kV and 40 mA with θ-2θ scans from 20° to 70°. Surface morphology and microstructure were conducted using a ZEISS GeminiSEM 300 field emission scanning electron microscope (SEM). SEM imaging was performed at 10 kV and 100,000× magnification. Transmission electron microscopy (TEM) analysis was executed on a Talos F200X instrument (Thermo Fisher) at 200 kV. Carrier mobility was measured using an HMS-7000 Optical Hall Effect tester (Ecopia). Electrochemical impedance spectroscopy (EIS) and photocurrent intensity response measurements were carried out on a CHI660E electrochemical workstation using a three-electrode system. Test samples were coated onto clean fluorinated tin oxide glass as the working electrode, with a platinum plate as the counter electrode and a silver/silver chloride (Ag/AgCl) electrode as the reference. The Brunauer-Emmett-Teller (BET) specific surface areas were determined from nitrogen adsorption isotherms using a Micromeritics Accelerated Surface Area and Porosimetry 2460 instrument.

Powdered specimens were characterized by XRD. Figure 1A presents the XRD patterns for NiO_x and 10% Bi₂Te₃/NiO_x. Three distinct diffraction peaks appear at 2θ = 37.1°, 43.2°, and 62.5°, corresponding to the (111), (200), and (220) planes of cubic NiO_x, consistent with the reference values from the standard card (JCPDS No. 4-0835). In the XRD pattern of the 10% Bi₂Te₃/NiO_x heterojunction, diffraction peaks appearing at 2θ = 27.6°, 41.2°, 50.2°, 57.1°, and 66.9° correspond to the (015), (110), (205), (0210), and (125) planes of trigonal Bi₂Te₃ (JCPDS 82-0358), confirming the coexistence of both phases in the heterojunction. Notably, only NiO_x and Bi₂Te₃ diffraction peaks were observed, with no impurity phases detected, confirming the successful synthesis of the heterojunction.

TEM analysis was conducted on NiO_x and 10% Bi₂Te₃@NiO_x. Figure 1B reveals uniformly dispersed NiO_x nanoparticles with grain sizes of 10-22 nm. The high-resolution TEM (HRTEM) image in Figure 1C (corresponding to the red-boxed area in Figure 1B) reveals crystal lattice spacings of 0.21 and 0.24 nm, corresponding to the (200) and (111) planes of NiO_x, respectively. The TEM image in Figure 1D confirms the integration of Bi₂Te₃ nanorods with NiO_x nanoparticles in the 10% composite. The HRTEM image in Figure 1E illustrates a 0.22 nm lattice fringe assigned to Bi₂Te₃ (110), in contact with the NiO_x (200) plane, revealing the heterojunction microstructure of the 10% composite.

SEM images [Figure 2A] illustrate pristine NiO_x as flower-like aggregates of interconnected nanosheets, while energy-dispersive X-ray spectroscopy (EDS) mapping [Figure 2B] confirms uniform Ni distribution, indicating compositional homogeneity of the NiO_x scaffold. After Bi₂Te₃ introduction, the 10% Bi₂Te₃@NiO_x composite presents a distinctly different morphology. Figure 2C presents rod-shaped Bi₂Te₃ nanostructures (indicated by the yellow dashed boxes) dispersedly anchored onto NiO_x nanoflower surfaces. EDS Ni mapping [Figure 2D] reveals uniform Ni distribution in NiO_x regions, while Bi₂Te₃ nanorods appear as Ni-free dark-contrast areas. EDS mappings for Bi [Figure 2E] and Te [Figure 2F] reveal that both elements are predominantly concentrated in the nanorod regions, matching the dark areas observed in the Ni map. EDS spectrum [Supplementary Figure 2] verifies the presence of Ni, O, Bi, and Te in the 10% composite, with no detectable impurities beyond the Si substrate. Quantitative analysis yields a Bi:Te atomic ratio of 2:3, confirming compositional homogeneity.

To investigate the effects of Bi₂Te₃ concentration on interfacial charge transfer, EIS was performed on all samples (0-15 wt% Bi₂Te₃). The spectra were simulated using the equivalent circuit (Figure 3A; inset), including electrolyte resistance and charge transfer resistance (Rct) at the nanocomposite interface, to extract key electrical parameters. Nyquist analysis demonstrates that R_ct is minimized at 10% Bi₂Te₃@NiO_x (0.74 × 10⁶ Ω) compared with pure NiO_x (1.50 × 10⁶ Ω) and other ratios, indicating heterojunction formation for electron transfer. Higher Bi₂Te₃ loadings likely cause agglomeration, increasing interfacial resistance. Optical Hall measurements [Figure 3B] reveal that carrier mobility fluctuates with Bi₂Te₃ content rather than following a linear trend, suggesting it is more sensitive to localized synthesis variations and heterostructure quality than to Bi₂Te₃ loading alone. The enhanced gas-sensing performance of the 10% Bi₂Te₃@NiO_x sensor stems primarily from minimized interfacial charge transfer resistance and optimal heterojunction modulation, rather than bulk carrier mobility. This optimized interfacial property enables more efficient utilization of active sites, leading to superior sensing responses.

To investigate the effect of Bi₂Te₃ on NiO_x gas sensitivity, pristine NiO_x and the 10% Bi₂Te₃@NiO_x composite were employed as sensitive layers and systematically evaluated with a dynamic testing system. Given the influence of operating temperature on gas sensor responses, the 10% Bi₂Te₃@ NiO_x sensor was tested toward 10 ppm H₂S at 30% RH across a range of temperatures. The sensor responses were 1.09, 1.32, 5.37, 11.81, and 5.14 at 30, 50, 70, 90, and 110 °C, respectively, peaking at 11.81 at 90 °C [Figure 4A]. This temperature was selected for all subsequent measurements. This trend is attributed to insufficient activation energy at low temperatures, limiting both charge carrier excitation and surface reactions. As temperature increases, these reactions are progressively promoted. However, excessively high temperatures suppress adsorption and reduce efficiency, causing a decline in response^[17,18].

To examine humidity effects, the sensor’s dynamic response to 10 ppm H₂S at 90 °C was tested from 0% to 80% RH in 10% steps [Figure 4B]. The baseline resistance remains stable across humidity variations, reflecting satisfactory environmental stability. This notable humidity independence is attributed to the strong competitive adsorption of H₂S, which exhibits a significantly higher affinity for active sites than H₂O and dominates the adsorption process. The 90 °C operating temperature thermodynamically inhibits water physisorption. Based on this stability, 30% RH was selected as the baseline for all subsequent tests. Unlike idealized dry air (0% RH), 30% RH accurately simulates real ambient and industrial environments, making the assessment of selectivity and stability metrics more applicable to practical deployment.

The effect of Bi₂Te₃ loading on NiO_x sensing performance was investigated via dynamic flow measurements [Figure 4C]. At 90 °C and 30% RH, responses to 10 ppm H₂S for 0-15 wt% Bi₂Te₃ were 1.47, 2.74, 4.3, 11.81, and 6.03, respectively [Figure 4D]. The 10% Bi₂Te₃@NiO_x sensor demonstrated the highest response, attributed to an optimized p-n heterojunction effect. The Bi₂Te₃/NiO_x interface generates a built-in field that promotes carrier separation and migration upon gas adsorption^[19,20], while the 10% loading minimizes interfacial defects that could impede mobility^[21]. Bi₂Te₃ doping also lowers sensor resistance via its high conductivity^[21,22]; moderate doping enhances response, but excess doping over-enhances conductivity and suppresses gas adsorption, reducing sensitivity^[20].

The response-recovery curve for 10 ppm H₂S (first cycle, Figure 4C) is presented in Figure 4E, with response and recovery times of 970 s and 1,810 s, respectively. Dynamic transients of the 10% Bi₂Te₃@NiO_x sensor to 0.2-50 ppm H₂S were measured at 90 °C and 30% RH [Figure 4F]. The response scales positively with H₂S concentration, enabling resistance-based concentration estimation, with the corresponding response values summarized in Figure 4G. Power-law fitting gave the relationship between response (y) and the H₂S concentration (x): y = -11.76 + 17.78x^0.1 (R² = 0.97, where R² near 1 indicates good fit). At low concentrations (0.2-4 ppm), a near-linear fit yielded: y = 4.04 + 1.04x (R² = 0.927) (inset). The sensor response is defined as follows:

(1) $$ S_{{Bi}_{2} {Te}_{3} @ {NiO}_{x}}=R_{{H}_{2} {S}} / R_{a i r} $$

where $$ S_{{Bi}_{2} {Te}_{3} @ {NiO}_{x}} $$ signifies the response value, $$ R_{H_2S} $$ indicates the resistance in H₂S, and R_air denotes the resistance in air.

Sensor selectivity was evaluated against 10 ppm of acetone, ethanol, H₂S, NO₂, and formaldehyde at 90 °C and 30% RH [Figure 4H]. The 10% Bi₂Te₃@NiO_x sensor demonstrates a distinctly higher response to H₂S than to all interferents, exhibiting exceptional H₂S selectivity. Long-term stability tests [Figure 4I] reveal a response fluctuation below 5% over a month, confirming reliable performance for continuous H₂S monitoring. The initial slight increase in response over the first few days is due to surface activation and aging^[23-25], during which residual impurities desorb and chemisorbed oxygen species equilibrate, exposing additional active sites and marginally enhancing performance. After this brief activation, the sensor exhibits excellent long-term stability with negligible degradation, highlighting its practical applicability. Compared to recently reported low-temperature gas sensors, the Bi₂Te₃@NiO_x sensor exhibits markedly superior sensing performance [Supplementary Table 1].

These results indicate that Bi₂Te₃ loading modulates sensor performance. To elucidate this mechanism, density functional theory (DFT) calculations were performed to determine the adsorption energy (E_ads) and charge density difference for H₂S adsorption on NiO_x and Bi₂Te₃@NiO_x.

Adsorption energy (E_ads), which quantifies interaction strength and adsorption difficulty, is calculated as follows:

(2) $$ E_{a d s}=E_{t o t a l}-E_{s u b}-E_{g a s} $$

where E_total, E_sub, and E_gas present the free energies of the adsorption structure, substrate, and gas molecule, respectively. Negative E_ads denotes stable exothermic adsorption. Optimal H₂S adsorption models on NiO_x and Bi₂Te₃@NiO_x were constructed [Figure 5A]. Bi₂Te₃@NiO_x exhibited lower H₂S E_ads than NiO_x, suggesting stronger adsorption, which is key to enhanced sensing performance. Differential charge densities and Bader charges (Δq) of adsorbed H₂S were also evaluated. Differential charge density is given by:

(3) $$ \Delta \rho=\rho_{t o t a l}-\rho_{l a y e r}-\rho_{g a s} $$

where Δρ denotes the differential charge density; ρ_total signifies the total charge density of the adsorption system; and ρ_layer and ρ_gas represent the charge densities of the isolated substrate and gas molecule under identical conditions, respectively. The charge transfer between Bi₂Te₃@NiO_x and H₂S increased to 0.13 e [Figure 5B], enhancing adsorption and amplifying the sensing response.

Bi₂Te₃, a 3D topological insulator, exhibits Dirac points in its band structure. To investigate the impact of H₂S adsorption on its energy band, we calculated the band structures of pristine Bi₂Te₃ and H₂S-adsorbed Bi₂Te₃ using DFT on the (111) surface. Consistent with previous findings that reducing gas adsorption induces n-type doping in topological insulators, shifting Dirac points downward and raising the Fermi level^[15], our calculations reveal analogous behavior for H₂S on Bi₂Te₃.

Our DFT calculations [Figure 5C and D] exhibit a Dirac point at -0.1110 eV for pristine Bi₂Te₃, which downshifts to -0.1176 eV (Δ = -0.007 eV) upon H₂S physisorption, confirming n-type surface doping. However, at the 90 °C operating temperature, bulk bands are expected to dominate the electronic transport of Bi₂Te₃, rather than the topological surface state alone. This subtle surface-state shift is nonetheless critical, as gas sensing is fundamentally interface-dominated. By analogy with chemical doping (e.g., Sb-doping) that alters the electronic structure of Bi₂Te₃^[26], H₂S physisorption locally perturbs surface states and catalytically lowers the activation energy for electron transfer. Upon H₂S exposure, the enhanced response is driven synergistically by charge transfer across the p-NiO_x/n-Bi₂Te₃ heterojunctions, responsible for the primary resistance modulation, and Bi₂Te₃’s exceptionally high carrier mobility^[20]. Uniform heating of the ultra-thin film on the micro-ceramic substrate further ensures a macroscopic isothermal condition throughout. The accelerated electron transfer is thus attributable solely to the interfacial p-n junction effect and surface state modulation, with no thermoelectric contributions from thermal gradients.

N₂ adsorption-desorption measurements were performed on pristine NiO_x and the 10% Bi₂Te₃@NiO_x composite to ascertain whether performance enhancement stems simply from increased surface area. Both samples display type IV isotherms, confirming slit-like mesoporous structures. BET results [Supplementary Figure 2] yield specific surface areas of 12.81 m²/g for pristine NiO_x and 12.98 m²/g for the 10% Bi₂Te₃@NiO_x composite. The marginal difference demonstrates that the incorporation of Bi₂Te₃ exerts minimal influence on the physical surface area of the sensing material.

A plausible H₂S-sensing mechanism for Bi₂Te₃@NiO_x is illustrated in Figure 5E. Upon exposure to H₂S, the majority charge carriers (holes) on the NiO_x surface interact with H₂S gas molecules:

(4) $$ H_{2} S_{(g a s)} \rightarrow H_{2} S_{(a d s)} $$

(5) $$ H_{2} S_{(a d s)}+\mathrm{h}^{+} \rightarrow H_{2} S^{+} $$

Upon H₂S exposure, hole carriers at the NiO_x surface are neutralized, resulting in a reduced carrier concentration and increased resistance. Concurrently, H₂S adsorption on Bi₂Te₃ induces a downward shift of its Dirac point, implying that the composite experiences both hole depletion on NiO_x and n-type surface doping on Bi₂Te₃^[16]. This electron excitation further reduces the hole concentration in Bi₂Te₃@NiO_x, leading to higher resistance ($$ R_{H_2S} $$) than that of NiO_x under identical H₂S exposure. Consistent with (1), the optimal Bi₂Te₃ loading therefore maximizes the sensing response.

Furthermore, Bi₂Te₃’s topologically protected surface states, safeguarded by time-reversal symmetry during gas doping^[14,15], suppress carrier backscattering from internal defects and thus significantly enhance carrier mobility in Bi₂Te₃@NiO_x. This enhanced mobility promotes efficient charge transfer with H₂S molecules, further improving sensing performance.