A 3D conductive amorphous unsaturated-ligand MIL-68(Fe) MOF as room-temperature H₂S chemiresistive sensor

Abstract

Simultaneously achieving chemiresistive gas sensors with high selectivity, sensitivity, and stability remains a significant challenge. In this study, inspired by the H₂S-specific catalytic enzymes in biological systems, we construct an amorphous unsaturated-ligand MIL-68(Fe) 3D MOF via in situ thermal induction. Without the introduction of extrinsic catalytic or conductive components, this MOF modification strategy creates unsaturated Fe atoms within the framework itself. These atoms function dually as catalytic and conductive sites, thereby overcoming the inherent low conductivity of conventional 3D MOFs and enabling the direct construction of chemiresistive sensor for room-temperature molecular olfactory sensing. The targeted water-adsorption-coordinated catalytic mechanism bypasses the limitations of traditional oxygen adsorption processes to achieve unprecedented comprehensive performance, including fingerprint-level selectivity (against 29 interfering gases), superior sensitivity coefficient (~83/ppm), ultra-low measured lower limit (1 ppb), rapid gas sensing response/recovery characteristics (~30 s/73 s), and excellent long-term stability at room temperature.

In this study, authors construct amorphous Fe metal organic frameworks, where the Fe atoms act as catalytic and conductive sites. The resulting material can be exploited for the direct construction of chemiresistive gas sensors.

Introduction

Since the inception of the first ZnO gas sensor in 1962¹, heated metal oxide semiconductor (MOS) sensors have emerged as the dominant paradigm in chemiresistive gas sensors due to their high sensitivity and robust stability (Fig. 1a). However, iterative advancements have revealed intrinsic limitations^2–4—poor selectivity and sluggish recovery—stemming from their universal oxygen adsorption mechanism^5,6.

Fig. 1. Integrated catalytic-conductive site chemiresistive sensor design for H₂S molecular olfactory sensing.

a Comparison of gas-sensing performance among chemical sensors based on different gas-sensing material systems (MOS metal oxide semiconductor, MOF metal-organic framework). The arrow direction indicates a relatively better performance advantage among four properties corresponding to a specific gas sensing material system. b Schematic diagram of the working principle of chemiresistive sensors. A typical sensing process consists of gas molecules donating electrons ($\mathrm{\Delta }n$) to the sensing material (perception), the sensing material receiving these electrons through a gas-sensitive reaction pathway (conversion), and the migration of the material’s intrinsic carriers ($n$) (transport). Gas sensing sensitivity ($S$) is determined by the ratio of $\mathrm{\Delta }n$ to $n$. c Catalytic mechanism of conventional heated MOS H₂S sensors based on oxygen-adsorption processes. d Design and catalytic process of bioinspired H₂S molecular olfactory sensors employing targeted water-adsorption sensing mechanism.

Using hydrogen sulfide (H₂S) detection as a representative case, the oxygen-mediated mechanism catalyzes H₂S oxidation into higher-valence species (Fig. 1c). While oxygen’s high reactivity enables broad-spectrum gas detection, it also leads to irreversible byproduct accumulation, severely restricting the sensing cycle lifetime. Although extensive material modifications (e.g., doping, nanostructuring) have been implemented, these efforts fail to address the fundamental issue. This impasse necessitates a paradigm shift. Intriguingly, biological olfaction systems (e.g., porcine/drosophila olfactory receptors) achieve exceptional specificity through protein-targeted molecular recognition^7–10, inspiring our design. Unfortunately, proteins’ inherent instability in vitro and insulating properties preclude their use in chemiresistive gas sensors (CRS)¹¹. Then, metal-organic frameworks (MOFs), as hybrid organic-inorganic analogs¹², overcome stability issues while mimicking enzymatic functional groups for selective catalysis. However, conventional 3D MOFs are plagued by ultralow electrical conductivity, confining them exclusively to chemicapacitive sensors (CCS)^13,14, which inherently exhibit poor sensitivity. To apply 3D-MOF to CRS, it must be composited with conductive media (such as graphene/ionic liquids, etc.) to construct a conductive 3D MOF composite material^15,16. As a comparison, the emergence of 2D materials (including 2D MOFs, graphene, and transition metal dichalcogenides) has addressed this conductivity dilemma. Their exceptional charge transport properties make them ideal candidates for room-temperature CRS^17–20. However, their selectivity remains inferior to 3D MOFs, while their sensitivity significantly underperforms conventional MOS sensors (Fig. 1a).

The performance trade-offs in CRS fundamentally arise from their sensing mechanism, which consists of three consecutive processes (Fig. 1b). Gas molecules donate electrons ($\mathrm{\Delta }n$) to the sensing material, thereby modulating its intrinsic carrier concentration ($n$) and subsequently altering the charge transport signal. In this framework, selectivity is exclusively governed by molecular perception capability, charge transport capacity directly reflects electrical conductivity, while sensitivity ($S$)—defined as the ratio of gas-induced electrons to intrinsic carriers ($\mathrm{\Delta }n$/$n$)—depends on conversion efficiency. Conventional 2D materials exhibit an intrinsic limitation where improved charge transport (through abundant conductive sites) inevitably compromises molecular perception (by reducing catalytic sites), simultaneously degrading selectivity due to insufficient specific catalytic centers. Furthermore, enhanced material conductivity results in an excessive intrinsic carrier concentration ($n$), which significantly compromises the sensitivity of both 2D materials and conductive 3D MOF composites. This reveals the critical requirement for CRS active sites to simultaneously serve as both catalytic and conductive sites—a dual functionality successfully achieved in MOS sensors through oxygen vacancies (Fig. 1c), though at the cost of selectivity^2–4,21. Consequently, the central challenge in CRS development lies in constructing material systems with both selective sensing mechanisms and dual-functional active sites capable of concurrent molecular perception and charge transport.

The biological catalytic process of H₂S occurs in lymphatic fluid, where water molecules serve as the reaction medium with unsaturated Fe atoms acting as the core catalytic centers (Fig. 1d)^22–24. Table S1 presents the comparative advantages of the water molecule catalysis sensing mechanism over the conventional oxygen adsorption mechanism. Inspired by this natural mechanism, we have successfully mimicked the structure of H₂S-specific enzymatic proteins by constructing an amorphous unsaturated-ligand 3D Fe-MOF (AUL MIL-68(Fe)) through in situ thermal induction. Without relying on additional doping or assembly steps^15,16, this modification method directly transforms the pristine MOF, in a single step, into an active and conductive MOF where the incorporated unsaturated Fe atoms function simultaneously as catalytic and conductive sites. This material exhibits a remarkable four-order-of-magnitude enhancement in electrical conductivity (3.18 × 10⁻⁷ S cm⁻¹) compared to pristine MOF (1.28 × 10⁻¹¹ S cm⁻¹), enabling its direct application in CRS fabrication. Benefiting from the amorphous coordination configuration and exceptional water-mediated catalytic cycling capability of this MOF, the sensor achieves a humidity-dependent H₂S coordination catalytic cycle at room temperature under ambient conditions (Fig. 1d). This unique bioinspired design achieves selective molecular olfactory sensing at the level of a single resistive sensing element, thereby granting the sensor outstanding comprehensive performance without any external excitation including high selectivity (against 29 interfering gases), superior sensitivity (~83/ppm), ultra-low measured lower limit (1 ppb), rapid gas sensing response/recovery characteristics (~30 s/73 s), and excellent long-term stability at room temperature.

Results and discussion

Design, construction and characterization of AUL MIL-68(Fe)

In nature, there exists a class of bacteria specifically capable of catalyzing H₂S—the sulfate-reducing bacteria. Their catalytic enzyme, hydrogenase, features a core catalytic site known as the H-cluster, whose structure is shown in Fig. 2a. This H-cluster consists of unsaturated Fe atoms and several organic groups^22,24. Substituting the H-cluster with MOFs for constructing CRS is intriguing because of their structural similarity. The resulting CRS could potentially exhibit hydrogenase-like high selectivity for H₂S detection, but modifications to the MOF are necessary to achieve this biomimicry. In hydrogenase, H₂S catalysis occurs in the cellular fluid, where H₂S is progressively cleaved by water molecules into HS⁻ and S²⁻, which then coordinate with the unsaturated Fe atoms for subsequent catalysis (Fig. 2a). Inspired by this, selecting MOFs with high affinity for H₂O molecules is more promising for replicating this process in the gas phase environment. Based on multiple studies, the MIL series of MOFs emerged as our target due to their high affinity for H₂O and unique proton-conductive behavior^25–28. Next, to introduce catalytically active unsaturated Fe sites, post-synthetic excitation of the MOF is essential, as as-synthesized MOFs typically exhibit extremely low unsaturation²⁹. Another interesting phenomenon provided further insight: when subjected to high-temperature deep annealing, MOFs shed their organic ligands and transform into MOS^30,31, with a significant boost in electrical conductivity. This suggests the possible existence of an intermediate state where Fe atoms in MOF detach from ligands but do not fully dissociate, creating unsaturated Fe sites while establishing Fe-O bond-dependent conductive pathways that serve dual roles in catalysis and charge transport. In this paper, DFT calculations and experiments confirmed this hypothesis, leading to the development of an amorphous unsaturated-ligand (AUL) MOF suitable for direct CRS construction.

Fig. 2. Design and characterization of AUL MIL-68(Fe).

a Molecular structure of the H-cluster for [FeFe] hydrogenase. The shading indicates the products of H₂S degradation by hydrogenase. b Schematic illustration of the structural transformation from MIL-68(Fe) to AUL MIL-68(Fe) based on density functional theory (DFT) calculations. The dark blue, purple, red, gray, and white balls represent Fe, Na, O, C, and H atoms, respectively. c Fourier transform infrared (FT-IR) spectra of MIL-68(Fe) under various annealing conditions and d enlarged view of the low-frequency region in c. e Fourier transforms of experimental extended X-ray absorption fine structure (EXAFS) spectra, and f  X-ray absorption near edge structure (XANES) profiles at the Fe K-edge for MIL-68(Fe) and AUL MIL-68(Fe). g Powder X-ray diffraction (PXRD) patterns of MIL-68(Fe) under various annealing conditions. h High-angle annular dark field (HAADF) images and energy-dispersive X-ray spectroscopy (EDS) elemental mapping analysis. The microscopic morphology was similar in the results of the two independent experiments. i  Projected density of states (PDOS) and j high-resolution O 1 s X-ray photoelectron spectroscopy (XPS) spectra of MIL-68(Fe) and AUL MIL-68(Fe).

We selected MIL-68(Fe) as the initial MOF and achieved AUL MIL-68(Fe) through moderate in situ thermal induction. Because MIL-68(Fe) is a rigid MOF (Fig. S2a)³², its organic ligands resist complete detachment during thermal induction. For comparison, we also examined the flexible MIL-88(Fe) (Fig. S2b)³³, which shares identical ligands with MIL-68(Fe) and slowly converts to MOS under the same thermal treatment, making it difficult to obtain the AUL MOF intermediate.

Based on thermogravimetric and differential scanning calorimetry (TG-DSC) results (Fig. S3a), an appropriate temperature (350 °C) below the phase transition point was selected to thermally induce the transformation of MIL-68(Fe) into an unsaturated 3D conductive MOF. In contrast to conventional one-step synthetic approaches that enhance metal center unsaturation^29,34, thermal induction drives extensive BDC (1,4-benzenedicarboxylate) linker fragmentation coupled with framework amorphization, ultimately yielding the uniquely structured AUL MIL-68(Fe) material, density functional theory (DFT) simulations revealed its structural evolution (Fig. 2b).

Compared with MIL-68(Fe), AUL MIL-68(Fe) retained the BDC chains, but the Fe-O1-C bonds became prone to cleavage (Fig. 2b), as shown in Fourier transform infrared (FT-IR) spectra. The characteristic ν_as (COO⁻) and ν_s (COO⁻) vibration bands of BDC at 1557.4 cm⁻¹ and 1401.2 cm⁻¹ (Fig. 2c) were consistently showed³⁵. Meanwhile, in the low-frequency Fe-O region, the vibration band at 508.4 cm⁻¹ (corresponding to Fe-O2 bonds) remained stable, while the band at 547.2 cm⁻¹ (corresponding to Fe-O1 bonds) exhibited significant broadening (Fig. 2d)^36,37. The bond cleavage transformed the Fe atoms into highly catalytically active unsaturated Fe species (Fig. 2b), as evidenced by the clear decrease in Fe-O (1.35 Å) and Fe-C (2.01 Å) coordination numbers in the extended X-ray absorption fine structure (EXAFS) profile (Fig. 2e). Notably, the X-ray absorption near edge structure (XANES) spectra revealed a narrower and lower-energy white line peak (Fig. 2f), suggesting that the Fe 4p orbital reverted to a degenerate state after hybridization³⁸. The bond cleavage also disrupted the octahedral structure of the FeO₆ clusters, leading to the formation of a pre-edge peak in the XANES spectra (Fig. 2f) and driving the material’s transition to an amorphous state. This was further confirmed by the weakening and shifting of two prominent XRD peaks near 8˚ (corresponding to the largest d-spacing planes) as the temperature increased to 280 °C, followed by their complete disappearance at 350 °C without the emergence of any new diffraction peaks (Fig. 2g).

In this amorphous unsaturated-ligand state, the directional nature of original chemical bonds was disrupted, leading to atomic rearrangement and reduced interatomic distances that caused structural collapse (Fig. 2h). Notably shortened Fe-O bonds induced Fe-Fe coupling, as evidenced by the increased Fe-Fe (2.51 Å) shell intensities in the extended EXAFS profile (Fig. 2e). These changes induced modifications in both the electronic band structure and atomic configuration of the material. In the projected density of states (PDOS), the original bandgap disappeared, and orbital coupling between Fe 3d and O 2p orbitals was observed near the Fermi level (Fig. 2i). Characteristic Fe-O-Fe peaks were also detected in the Fe 2p X-ray photoelectron spectroscopy (XPS) spectra (Fig. 2j). These indicate the formation of a conductive pathway dependent on Fe-O bonds (Fig. 2b), which reduced the optical bandgap of the material from 2.77 eV to 1.98 eV (Fig. S4) and ultimately enhanced the electrical conductivity by four orders of magnitude (from 1.28 × 10⁻¹¹ S cm⁻¹ to 3.18 × 10⁻⁷ S cm⁻¹) (Fig. S5b, c and Table S2). Through this controlled approach, unsaturated Fe atoms were successfully introduced that simultaneously function as both highly active catalytic sites and conductive pathways.

For comparison, we synthesized relatively flexible spindle-shaped MIL-88(Fe) (Fig. S6a). The TG-DSC curves revealed that its mass loss rate was significantly faster than that of MIL-68(Fe) (Fig. S2). Upon annealing at 350 ˚°C, the material underwent MOS transformation, as confirmed by the following evidence. Unlike MIL-68(Fe), the complete disappearance of BDC-derived ν_as (COO⁻) and ν_s (COO⁻) vibration bands in FT-IR spectra indicates thorough removal of BDC linkers (Fig. S7a), resulting from carbon atom oxidation during thermal induction process. This was corroborated by the appearance of additional subpeaks in Fe 2p and C 1s XPS spectra (Figs. S8 and S9), potentially corresponding to carbon-oxygen compounds. Concurrent with BDC linker loss, the material underwent serious molecular structural transformation as shown in FT-IR results: the Fe-O2 vibration band at 547.2 cm⁻¹ broadened and redshifted, while the Fe-O1 band at 611.7 cm⁻¹ disappeared completely, accompanied by the emergence of a new vibration peak at 450 cm⁻¹ (Fig. S7b, c). Its structural collapse was more pronounced compared to MIL-68(Fe) (Figs. 2h and  S6). XRD analysis ultimately revealed weak and partially aligned characteristic peaks corresponding to Fe₂O₃ (Fig. S10). Due to the incomplete nature of this MOS transformation, we designated the resulting material as activated (AC) MIL-88 (Fe). Notably, its bandgap showed a significant reduction compared to pristine MIL-88(Fe) (from 2.79 eV to 1.95 eV, Fig. S4), accompanied by a remarkable conductivity enhancement (from 5.77 × 10⁻¹² S cm⁻¹ to 2.07 × 10⁻⁸ S cm⁻¹, Fig. S5d, e and Table S2). These comparative results demonstrate that AUL MIL-68(Fe) not only preserves the intrinsic MOF characteristics, but also achieves MOS-level conductivity improvement—a unique advantage attributable to its highly symmetric rigid framework structure.

Water adsorption behavior of AUL MIL-68(Fe)

In conventional oxygen-adsorption mechanisms, oxygen molecules must be adsorbed and activated by the MOS sensor to form reactive oxygen species (O₂⁻, O²⁻, O⁻) for catalytic processes (Fig. 1c)³⁹. In the current material system, a highly analogous water-adsorption-driven catalytic mechanism was identified, where ambient water molecules spontaneously undergo electrooxidation on the material surface to generate reactive hydroxyl species for subsequent H₂S catalysis (Fig. 1d). This section systematically investigates and analyzes this water adsorption behavior.

Water molecules are readily adsorbed by both MIL-68(Fe) and AUL MIL-68(Fe), as evidenced by the broad hydroxyl peak (ν (OH) band around 3400 cm⁻¹) in FT-IR results (Fig. 2c). Subsequently, sensor was fabricated by drop-casting ethanol slurries of the materials onto planar interdigitated electrodes with Al₂O₃ substrates for electrochemical characterization and subsequent gas-sensing tests (Figs. 3a and  S11). In situ cyclic voltammetry (CV) revealed that the adsorbed H₂O undergoes electrooxidation on both MIL-68(Fe) and AUL MIL-68(Fe), related to the oxygen evolution reaction (OER)^34,40. Distinct oxidation peaks were observed at 0.19 V for MIL-68(Fe) and at 0.47 V and 2.08 V for AUL MIL-68(Fe) (Fig. 3b). The oxidation peaks at 0.19 V and 0.47 V both surpass the 0 V potential and exhibit symmetric reduction peaks, indicating spontaneous and reversible electrooxidation processes. DFT calculations illustrate this spontaneous electrooxidation process, generating protons and hydroxyl species (Fig. 3c). The hydroxyl species formation requires only one transition state, with AUL MIL-68(Fe) demonstrating a lower energy barrier than MIL-68(Fe) (Figs. 3c and  S12), attributable to its amorphous unsaturated-ligand characteristics. Unlike MIL-68(Fe), AUL MIL-68(Fe) enables coordinated H₂O molecules to undergo free conformational torsion following the principle of energy minimization, overcoming directional coordination constraints. This confirms that the amorphous and unsaturated nature of AUL MIL-68(Fe) provides superior catalytic kinetic advantages.

Fig. 3. Water adsorption behavior of AUL MIL-68(Fe).

a Schematic diagram of the CRS device used for testing. b In situ cyclic voltammetry (CV) curves of MIL-68(Fe) and AUL MIL-68(Fe) measured in air. c DFT-calculated schematic of H₂O molecule decomposition processes catalyzed by MIL-68(Fe) and AUL MIL-68(Fe). The dark blue, purple, red, gray, and white balls represent Fe, Na, O, C, and H atoms, respectively. The yellow shading illustrates the evolution of species during the catalytic process. d In situ electrochemical impedance spectroscopy (EIS) fitting results for AUL MIL-68(Fe). The illustration shows the enlarged view at low impedance. e Constant potential curves under varying humidity conditions.

Variable-humidity in situ electrochemical impedance spectroscopy (EIS) fitting results confirmed the persistent presence of spontaneous water electrooxidation. The impedance data were perfectly fitted using a two-time-constant model (Fig. 3d and Table S3). Both proton (R₁) and electron (R₂) resistances decreased significantly with increasing humidity, indicating enhanced electrooxidation under humid conditions. These results suggest that AUL MIL-68(Fe) maintains a hydroxylated coordination state under ambient humidity. Under these conditions, since gas sensing requires operation at a constant potential, the selection of applied voltage inevitably influences the sensing performance. Therefore, we selected 0.3 V within the low oxidation potential range common to both materials as the working potential. Current-time (I–t) characteristics at 0.3 V demonstrated typical mixed conduction behavior for AUL MIL-68(Fe) and MIL-68(Fe) (Figs. 3e and  S13a)^41,42. The electrooxidation-generated protons rapidly migrated under the applied electric field, producing a substantial proton current. As the proton current depleted, the current stabilized with dominant electronic conduction. Given the humidity’s influence on absolute current values, we maintained stable environmental humidity during subsequent gas sensing tests and initiated measurements only after baseline current stabilization.

Notably, broad hydroxyl peaks (ν (OH) band ~3400 cm⁻¹) were also found for MIL-88(Fe) and AC MIL-88(Fe) in FT-IR results (Fig. S7a), and identical I–t curve characteristics were observed for MIL-88(Fe) and AC MIL-88(Fe) (Fig. S13b, c), indicating their similar water adsorption behaviors.

Hydrogen sulfide gas sensing properties of AUL MIL-68(Fe)

In gradient concentration testing, AUL MIL-68(Fe) exhibited an extensive detection range from 1 ppb to 166.67 ppm H₂S (Fig. 4a), the sensor’s response value to 1 ppb H₂S (0.044) is greater than three times the noise level (RMS, Root Mean Square) (0.038), demonstrating stable gas sensing responses even at high concentrations while overcoming the gas sensing response delays characteristic of MOS sensors⁴³. The material showed rapid and consistent gas sensing response times across different concentration steps (Fig. S14) with complete recovery at the end (Fig. S15), indicating highly efficient H₂S catalytic cycling. Unlike the linear gas sensing response-concentration relationship typical of MOS sensors, AUL MIL-68(Fe) displayed a distinct logarithmic correlation (Fig. 4b) and progressively decreasing sensitivity during single-exposure tests (Fig. 4c), mimicking the olfactory desensitization phenomenon observed in biological systems exposed to high H₂S levels⁴⁴. Remarkably, the sensor achieved an exceptional sensitivity of 83/ppm at low concentrations, highlighting its trace-detection capability. After six months of operation, the sensor maintained high sensitivity (Figs. 4c and  S16).

Fig. 4. Gas sensing properties for AUL MIL-68(Fe) to H₂S.

a Gas sensing response curve of AUL MIL-68(Fe) to H₂S concentrations ranging from 1 ppb to 166.67 ppm at room temperature and 50% RH, with an inset highlighting the measured lower limit; b Gas sensing response-concentration relationship fitted using the gas sensing response values from a. c Time-dependent sensitivity calculated for single gas injection concentration. d Repeatability for eight consecutive 1 ppm H₂S tests of MIL-68(Fe) and AUL MIL-68(Fe), and e corresponding gas sensing response and recovery time for a single process in d. Performance comparison of AUL MIL-68(Fe) CRS with reported H₂S sensors: f Gas sensing recovery-response time. g Measured lower limit -sensitivity. Blue, yellow, green, red and purple in f and g respectively represent different gas-sensitive material systems: MOS, 3D MOFs, Conductive 3D MOF composite material, 2D material and AUL MIL-68(Fe) in this work. h Selectivity and i calculated selective ratio reciprocal for MIL-68(Fe) and AUL MIL-68(Fe). The gases in h from left to right are hydrogen sulfide, methanethiol, carbon disulfide, ammonia, triethylamine, methanol, ethanol, acetone, benzene, methylbenzene, chlorobenzene, hydrogen and methane.

Comparative studies revealed AUL MIL-68(Fe)’s superior reproducibility in both gas sensing response magnitude and response/recovery kinetics relative to MIL-68(Fe) (Fig. 4d, e). Benchmarking against H₂S sensors based on alternative material systems—including MOS^45–48, conventional 3D MOFs^49–51, conductive 3D MOF composite materials^15,16 and 2D materials^52–55—demonstrated the comprehensive performance advantages of bioinspired 3D conductive AUL MIL-68(Fe) resistive sensor in measured lower limit, sensitivity coefficient (sensitivity/ppm), and gas sensing response/recovery speed (Fig. 4f, g). Notably, the sensor exhibited fingerprint-level selectivity for H₂S, showing no anysensitivity to 12 different gases at 5 ppm concentrations (Fig. 4h). Extended testing against 29 interfering gases (Figs. S17 and S18) with calculated selectivity ratio (Rsp (H₂S)/Rsp (Gas)) at the same gas concentration) (Table S4) further confirmed this exceptional specificity. To visualize the results more clearly, the selective ratio reciprocal (Rsp (Gas)/Rsp (H₂S)) image for 30 gases is presented (Fig. 4i and Table S5). Building on the intrinsic classification capability of MIL-68(Fe) (specific to thiols, carboxylic acids, and amines), the AUL-modified variant eliminated its amine sensitivity while markedly enhancing its characteristic response to H₂S (Fig. 4i). When challenged with 29 interfering gases, it achieved selectivity on par with biomass-based chemical sensors^9,10. This performance not only matches that of biomimetic sensors but also obviously surpasses reported high-selectivity abiotic systems^10,56. We therefore propose to define this exceptional characteristic as fingerprint-level selectivity.

Comparative analysis revealed that thermal induction under other conditions showed limited effectiveness in improving the gas-sensing performance of both MIL-68 (Fe) and MIL-88 (Fe). The sensors exhibited unsatisfactory detection limits and poor resolution (Figs. S19 and S20).

Furthermore, the humidity dependence of the sensors was evaluated (Fig. S21). As humidity increases, both MIL-68(Fe) and AUL MIL-68(Fe) exhibit a significant improvement in response signal-to-noise ratio, and the response current also rises with increasing humidity. In contrast to the recovery rate, the response rate shows a strict positive correlation with humidity. Overall, elevated humidity contributes to enhanced gas-sensing performance of the materials. In the following section, the water-adsorption-catalyzed sensing mechanism of the materials toward H₂S is discussed in detail.

Hydrogen sulfide molecular olfactory sensing mechanism of AUL MIL-68(Fe)

The high selectivity of AUL MIL-68(Fe) for H₂S sensing is established on a water-adsorption-coordinated catalytic mechanism. In the CV curves, the overlapping oxidation peaks of H₂O, H₂S, and CH₂SH at the 0.3 V test potential reflect the co-oxidation behavior between the two gases and H₂O (Fig. 5a), which also explains why increasing humidity can effectively enhance the response signal/current to them (Fig. S22). Under low-humidity conditions, AUL MIL-68(Fe) exhibits very low efficiency in catalyzing H₂S. When the reddish-brown AUL MIL-68(Fe) powder was placed in a 100 ppm H₂S atmosphere at 20% RH, no significant color change was observed, and the powder remained its original reddish-brown color. However, in a 70% RH environment with 100 ppm H₂S, the powder quickly turned black (Fig. 5c). This color change originates from ligand-to-metal charge transfer (LMCT) from HS^*/S^* species to the Fe atom, more specifically to its 3 d orbital⁵⁷. We compared MIL-68(Fe), MIL-88 (Fe), and AC MIL-88 (Fe) and did not observe the same color change as AUL MIL-68 (Fe) under high humidity conditions (Fig. S23). Although they exhibit similar water adsorption behaviors to AUL MIL-68(Fe) (Figs. 3e and  S13), the observed differences confirm the critical role of MOF selection and optimized thermal processing in generating unsaturated Fe active sites, collectively validating the success of the biomimetic material design strategy.

Fig. 5. Characterization and explanation of the H₂S sensing mechanism of AUL MIL-68(Fe).

a In situ cyclic voltammetry (CV) curves of AUL MIL-68(Fe) measured under different atmospheres. b Gas sensing responses of AUL MIL-68(Fe) to H₂S, CH₃SH, and NH₃ at different potentials. c Colorimetric experiments of AUL MIL-68(Fe) for H₂S catalysis under varying humidity conditions. d In situ electrochemical impedance spectroscopy (EIS) fitting results and equivalent circuit models at different H₂S concentrations. e X-ray absorption near edge structure (XANES) profiles at the Fe K-edge for MIL-68(Fe) and AUL MIL-68(Fe). f High-resolution Fe 2p X-ray photoelectron spectroscopy (XPS) spectra of MIL-68(Fe), MIL-68(Fe) (H₂S), AUL MIL-68(Fe), AUL MIL-68(Fe) (H₂S), AUL MIL-68(Fe) (2 V), and AUL MIL-68(Fe) (2 V) (NH₃). DFT-calculated schematic of H₂S catalytic process by AUL MIL-68(Fe) g under anhydrous conditions and h under humid conditions. Dark blue, purple, red, yellow, gray, and white balls represent Fe, Na, O, S, C, and H atoms, respectively. The yellow shading illustrates the evolution of species during the catalytic process. MIL-68(Fe) (H₂S): MIL-68(Fe) aged in 100 ppm H₂S at 0.3 V for 2.5 h. AUL MIL-68(Fe) (H₂S): AUL MIL-68(Fe) aged in 100 ppm H₂S at 0.3 V for 2.5 h. AUL MIL-68(Fe) (2 V): AUL MIL-68(Fe) aged in 50% RH air at 2 V for 2.5 h. AUL MIL-68(Fe) (2 V) (NH₃): AUL MIL-68(Fe) (2 V) aged in 100 ppm NH₃ at 0 V for 2.5 h.

Additional experiments were performed to further investigate the interaction between H₂S and AUL MIL-68(Fe). In situ EIS measurements revealed a significant decrease in electron transfer resistance (R₂) with increasing H₂S concentration, demonstrating the redox reaction between H₂S and AUL MIL-68(Fe) (Fig. 5d and Table S6). The narrower white-line peak in the XANES spectrum indicates a further reduced splitting of the Fe 4p orbitals (Fig. 5e). Simultaneously, the narrower pre-edge features suggest coordination between Fe atoms and S species (Fig. 5e). This interaction induces a rightward shift in the Fe 2p binding energy for AUL MIL-68(Fe) (H₂S) (Fig. 5f), implying an increased electron density within the system. Based on the amplified edge profile (Fig. S24), the Fe valence state in AUL MIL-68(Fe) is close to +3⁵⁸. Upon H₂S exposure, the valence state of AUL MIL-68(Fe) (H₂S) increases slightly (shifting ~0.3 eV toward the MIL-68(Fe) direction), accompanied by a reduction in the occupancy of Fe 3d orbitals. Integrating the XPS and XAS results, it is demonstrated that electron delocalization within the Fe–O–Fe bridges leads to the formation of electron-rich, high-valent Fe species. This electronic state enables Fe atoms to function as both Lewis acid and Lewis base sites, facilitating facile electron transfer between the ligands and Fe centers, thereby conferring superior catalytic activity to the Fe atoms. These results demonstrate that the unsaturated Fe atoms, serving as conductive sites, also provide abundant catalytic centers, thereby endowing the sensor with ultrahigh sensitivity and an ultralow measured lower limit. The key to achieving integrated catalytic conductivity lies in the Fe atoms possessing both the ability to coordinate and catalyze water molecules and H₂S molecules, as well as a high degree of orbital coupling with other atoms to facilitate electron delocalization. On one hand, this electron delocalization enhances the intrinsic carrier concentration of the material, improving its electrical conductivity. On the other hand, it also facilitates the migration of valence electrons generated from the catalytic oxidation of H₂S within the material. The initial MOF is in a + 3 oxidation state, and its lattice integrity is high, making it difficult for H₂S to coordinate with Fe atoms, thereby suppressing the catalytic activity of Fe. Simultaneously, electrons are confined within the metal/organic ligand nodes, making it challenging for them to overcome the high energy barrier for effective migration¹². After thermal induction, AUL MIL-68(Fe) exhibits a mixed-valence state slightly lower than +3, indicating that the MOF lattice integrity has been compromised. When this disruption reaches a certain extent, unsaturated metal atoms are introduced into the material system. In a weak-field environment with the presence of water molecules, Fe³⁺ tends to form a high-coordination octahedral structure with six coordination sites⁵⁷. The unsaturated mixed-valence Fe atoms can readily associate with H₂S molecules, demonstrating catalytic activity as an electron acceptor for H₂S, while also exhibiting electron-donor characteristics that promote electron delocalization. Furthermore, this symmetry breaking reduces the directional covalent bonding in the MOF, enhancing orbital coupling between metal-metal and metal-organic ligands. This results in a reduced band gap and enhanced electron delocalization.

Similar to H₂O molecules (Fig. 3b), since the overlapping oxidation potential region of H₂S in the CV curves is close to the zero potential point and highly symmetric with the reduction potential (Fig. 5a), the catalytic oxidation of H₂S in the presence of water molecules is actually a highly reversible spontaneous process. So, when the black AUL MIL-68(Fe) (H₂S) powder was divided into two portions and placed in air environments with different humidity levels, it was found that the powder in the 20% low-humidity environment remained black for an extended period, while the powder in the 70% high-humidity environment quickly reverted to its reddish-brown color (Fig. 5c). This indicates that H₂O molecules accelerate the catalytic cycle of H₂S.

To provide a clearer explanation of this water-adsorption-coordinated catalytic mechanism, we performed DFT calculations on this process. Under anhydrous conditions, H₂S molecules are gradually decomposed into HS^* and S^*, with two barrier transition states (TS1 and TS2) existing in between (Figs. 5g and  S25). In humid environments, based on previous results, AUL MIL-68(Fe) exists in a hydroxylated coordination state, where hydroxyl groups accelerate the decomposition of H₂S coordinated with Fe atoms into HS^* and S^* without any intermediate barrier states (Fig. 5h). Assuming that the energy barriers for the two-step decomposition of H₂S into HS^* and S^* can be approximated as zero, the enhancement of the reaction rate due to water-molecule catalysis can be quantified. According to the Arrhenius equation⁵⁹:

$$k=A{e}^{-\frac{E_a}{RT}}$$ 1

where $k$ is the rate constant, $A$ is the pre-exponential factor, $E_a$ is the activation energy, $R$ the ideal gas constant (8.314 J mol⁻¹ K⁻¹), and $T$ the absolute temperature (293.15 $K$ here). Let $k_1$ and $k_2$ represent the rate constants for the first and second decomposition steps of H₂S on AUL MIL-68(Fe) under dry conditions, and $k_3$ and $k_4$ represent the corresponding constants under humid conditions. The ratio of rate constants is given by:

$$\frac{k_2}{k_1}=e^{\frac{E_{a1}}{RT}}$$ 2

$$\frac{k_4}{k_3}=e^{\frac{E_{a2}}{RT}}$$ 3

with $E_{a1}$ = 0.16 eV and $E_{a2}$ = 0.27 eV (see Fig. S25). Substituting these values shows that the presence of water accelerates the two-step H₂S decomposition by factors of ~564 and 43,904, respectively. It is evident that this water-assisted process significantly accelerates the overall catalytic cycle of H₂S. In this process, H₂O molecules act as catalysts, similar to oxygen in the oxygen-adsorption mechanism, providing OH^* as active catalytic mediators. Unlike the MOS-catalyzed generation of high-valence S species from H₂S, this process exhibits a far superior cycling efficiency compared to conventional oxygen-adsorption mechanisms. This high efficiency directly accounts for its fast response/recovery characteristics in gas sensing.

The high selectivity of AUL MIL-68(Fe) is strongly correlated with this water-adsorption-coordinated catalytic mechanism. In contrast, we found that MIL-68(Fe) catalyzes H₂S more dominantly via oxygen atoms, leading to a left-shift of the Fe 2p XPS peak for MIL-68(Fe) (H₂S) (Fig. 5f) along with the formation of sulfur oxides (Fig. S26a). Compared with MIL-68(Fe), the Fe 2p and O 1s XPS of MIL-68(Fe) (NH₃) shows highly similar evolution characteristics to that of MIL-68(Fe) (H₂S) (Fig. S27), while nitrogen oxides are simultaneously generated (Fig. S26b). This explains why MIL-68(Fe) is susceptible to interference from amine substances. In contrast, no N signal was detected in AUL MIL-68(Fe) (NH₃) (Fig. S26b). Furthermore, based on Fe 2p XPS results, we found that oxidation of AUL MIL-68(Fe) at the second H₂O oxidation potential (actual test potential of 2 V) (Fig. 5a) generates high-valence-state Fe (AUL MIL-68(Fe) (2 V)) (Fig. 5f), under which condition the material’s selectivity relationship between H₂S and NH₃ is reversed (Figs. 5b and  S28). After catalyzing NH₃, this high-valence state completely disappears (binding energy shifts right), with Fe atoms acting as electron acceptors. Therefore, the catalysis of NH₃ by AUL MIL-68(Fe) (2 V) still belongs to the water-adsorption-coordinated catalytic mechanism. Notably, this water-adsorption-coordinated catalytic mechanism of AUL MIL-68(Fe) provides a controllable selectivity regulation method, enabling electrochemical selectivity modulation of gas-sensing responses on CRS.

Methods

Sample synthesis and sensor fabrication (Fig. S1)

Materials

Iron chloride hexahydrate (FeCl₃·6H₂O, 99.99%) and nickel nitrate hexahydrate (Ni(NO₃)₂·6H₂O, 99.99%) were purchased from Shanghai Aladdin Biochemical Technology Co., Ltd. Terephthalic acid, N, N-dimethylformamide (DMF) and NaOH (95%, AR grade) were purchased from Shanghai Macklin Biochemical Co., Ltd. All chemicals were used directly without further purification.

Synthesis of MIL-68(Fe)

32.0 mg FeCl₃·6H₂O and 66.4 mg terephthalic acid were dissolved in 20 ml DMF ultrasonic treatment 10 min to get a clear and transparent yellow solution (solution A), stirring for another 10 min. Meanwhile, 160.0 mg NaOH was dissolved in 10 ml ultrapure water ultrasonic treatment 10 min to get an even solution. Then 1.0 ml NaOH solution was added into solution A drop by drop, the solution was kept stirring during the process and for  another 30 min, and finally got a tan emulsion (solution B). Subsequently, solution B was transferred into Teflon-sealed autoclave and heated for 12 h at 100 °C. The as-prepared samples were washed and centrifugated with DMF, absolute alcohol, absolute alcohol respectively for three times. Finally, the washed samples were dried at 65 °C for 8 h and collected.

Synthesis of MIL-88(Fe)

Except that  the content of FeCl₃·6H₂O was changed to 115.7 mg, the other parameters and procedures were consistent with the synthesis of MIL-68(Fe).

Thermal induction for MOF

Synthetic MOF samples were simulated through annealing to induce structure response. Samples were put in ceramic crucible and located in muffle furnace to keep at different temperature for a certain time with a heating rate of 3 °C/min, and then cooled in furnace.

Amorphous unsaturated-ligand (AUL) MIL-68(Fe) and activated (AC) MIL-88(Fe)

MIL-68(Fe) and MIL-88(Fe) powders were annealed at 350 °C for 12 h in air with a heating rate of 3 °C/min and then cooled in furnace.

Sensor fabrication

The resulting powder sample was dispersed in ethanol to form a suspension, which was then dropped onto screen-printed Au planar electrodes on a 100 °C hotplate to fabricate room-temperature sensors.

Characterization

Powder X-ray diffraction (PXRD) patterns were carried out on a Rigaku D/Max 2550 diffractometer using Cu Kα radiation (λ = 1.5418 Å) with a scanning speed of 5°/min.

Scanning electron microscopy (SEM) images were obtained by a JEOL JSM-7500F electron microscope. Transmission electron microscopy (TEM), and high-resolution TEM (HRTEM) images were obtained by Japan JEOL JEM-F200 electron microscope. Energy-dispersive X-ray spectroscopy (EDS) images were collected on matched spectroscopy system of SEM and TEM microscopes.

Thermogravimetric and differential scanning calorimetry (TG-DSC) test was conducted on STA449F3 TG/DSC integrated analyzer from Germany in air atmosphere with increasing temperature rate of 10 °C/min.

UV-visible absorption spectra were performed on an ultraviolet-visible spectrophotometer (SolidSpec-3700, Japan).

Fourier transform infrared spectra (FT-IR) was analyzed with Fourier transform infrared spectrometer (Nicolet iS50R, American).

X-ray photoelectron spectroscopy (XPS) was performed on AXIS Supra⁺ X-ray photoelectron spectrometer from Shimadzu Corporation, Japan.

Gas sensing measurements. The gas sensors were placed in a static testing system with a 20 L testing chamber for gas sensing measurements and connected to electrochemical workstation. Constant potential module of workstation was used to apply a constant voltage on sensor, the current flowing through the sensor is monitored, and in gas sensing, the “Sensitivity” is defined as:

$$\frac{I_{gas}-I_{air}}{I_{air}}$$ 4

where $I_{\mathrm{gas}}$ represents the current of gas sensor exposed in air, and $\frac{I_{\mathrm{gas}}-I_{\mathrm{air}}}{I_{\mathrm{air}}}$ is the current of gas sensor exposed in target gas. The calculated selectivity is a dimensionless value. In particular, to facilitate direct comparison with results reported in the literature, we define the “Sensitivity coefficient” explicitly as the sensitivity per unit part-per-million (Sensitivity/ppm). Additionally, the gas sensing response/recovery time is based on definition of time that the current value of sensors reaches 90% of final equilibrium value. Different from fitting LOD (Limit of detection)^52,60,61, the term “Measured Lower Limit” is used here to indicate the actual lower concentration value that the sensor can reliably detect, with its response value being greater than three times noise RMS (Root Mean Square).

$$RMS=\sqrt{\sum {\left(x_i-\overline{x}\right)}^2/N}$$ 5

where $x_i$ is the measured data point, $\overline{x}$ is the mean of the baseline data, and $N$ is the number of measured data points. The measured lower limit values compared in Fig. 4g from relevant literature are also all based on the actual measured lower concentration. During the whole testing process, the actual concentration of injected target gas was determined based on the equation:

$$V_1\times X_1{=V}_2\times X_2$$ 6

where the $X_1$ and $X_2$ represent the injected and actual gas concentrations, and $V_1$ and $V_2$ represent volume of injected gas and chamber, respectively. Dried target gases (H₂S, SO₂, NH₃, CH₄, H₂, Methanol, Ethanol, Acetone, Formaldehyde, Benzene) were purchased from Wuhan Zhongxin Ruiyuan Gas Co., Ltd in the form of compressed gas cylinders. Each cylinder contains a certified mixture of the target gas in nitrogen at a fixed concentration, as detailed in Table S7. Other interfering gases were obtained by sampling the saturated vapor above the corresponding liquid reagent. Their concentrations (in ppm) were calculated under standard testing conditions (atmospheric pressure, 20 °C). Chamber humidity is controlled via two adjustable methods. Evaporation from a water-filled beaker serves as the humidification source, with the rate fine-tuned by altering the beaker’s opening area and water volume. Conversely, humidity is reduced by introducing a nitrogen flow from an external compressed cylinder, with its rate controlled via the flow regulator. By balancing these inputs, a stable humidity level is maintained. Internal conditions are monitored in real time using a high-precision sensor (0.1 °C, 0.1% RH) to ensure testing specifications are met. Specifically, the H₂S testing under 0% RH (Fig. S22a, e) was conducted using a dynamic test system. In the dynamic test, a fixed concentration of H₂S (from a compressed gas cylinder) is passed through the reaction chamber at a certain flow rate. The chamber is also connected to another line of N₂, and the concentration of H₂S is controlled by adjusting the flow ratio between N₂ and H₂S. Since both gases are dry compressed gases, the humidity in the reaction chamber remains 0% RH.

Electrochemical measurements were performed on DH7006 electrochemical workstation from Jiangsu Donghua Analytical Instrument Co., Ltd, China at two electrode system and counter electrode and reference electrode were short contact to evaluate the property of sensor directly.

X-ray absorption fine structure (XAFS) was collected in the transmission mode using inhouse laboratory-based X-ray absorption spectrometer. All XAFS data were analyzed using the program Demeter. For all samples, the EXAFS oscillations were extracted from the normalized XAFS spectra by subtracting the atomic background using a cubic spline fit to k³-weighted data, where k is the photoelectron wave number. The χ(k) functions were then Fourier transformed into R-space. The Fourier-transform window was in the k range of 2–10 Å⁻¹.

Conductivity measurements were conducted with 2-contact probe method⁶² aided by B1500A semiconductor Device Analyzer from KEYSIGHT. The MOF powder was filled into a stainless steel die with a measured inner diameter of 2.96 mm, and a flat press was used to apply a fixed pressure of 1 MPa on both sides of the die to obtain a dense and uniform cylindrical pellet. The diameter of pellet was the same as the inner diameter of the mold, 2.96 mm, and the thickness was measured with a vernier caliper. Conductive silver paste was applied evenly on the upper and lower surfaces of the pellet to improve the electrical contact between the MOF particle and the electrode. Then, a thin copper sheet is adhered to the silver paste horizontally as a current collector. The two copper probes of the probe table of the semiconductor analyzer are in contact with two copper sheets for testing, and the I–V module tests the voltage range of −10 V – 10 V. The conductivity can be obtained from the formula

$$\sigma =G\mathrm{\cdot }\frac{L}{A}=\frac{I}{V}\mathrm{\cdot }\frac{L}{A}$$ 7

and the conductivity of MOF can be obtained directly by fitting the curve of

$$\sigma \mathrm{\cdot }V=\frac{L}{A}\mathrm{\cdot }I$$ 8

Density functional theory (DFT) calculations were performed using the Vienna ab initio simulation package (VASP) with the Perdew-Burke-Ernzerhof (PBE) form of generalized gradient approximation (GGA) and projector-augmented wave (PAW) method. The period (110) MIL-68(Fe) surface slab models were constructed. Valence electrons considered for crystalline and amorphous MIL-68(Fe) included 1 s for H, 2s²2p⁴ for O, 2s²2p² for C, 3s¹ for Na, and 3 d⁶4s² for Fe. The strongly correlated 3d electron of Fe was treated with the GGA + U correction with U^d = 5.3 eV, which is suggested in other theoretical works⁶³. Amorphous MIL-68(Fe) structures were obtained using Ab initio molecular dynamics (AIMD) simulations including annealing, cooling and relaxation. For the structural relaxations, the convergence criteria of 1.0 × 10⁻⁵ eV atm^-−1 and 0.02 eV Å⁻¹ were used for the electronic self-consistent iteration and the maximum force on each atom, respectively. A plane-wave cutoff of 500 eV was used and the Brillouin zone was sampled using the Monkhorst-Pack method with 2 × 1 × 1 mesh of k-point for stacked structures, respectively. The adsorption energies (ΔE_ads) are defined as:

$$\mathrm{\Delta }{E}_{ads}=E_{adsorbate/slab}-\left(E_{slab}+E_{adsorbate}\right)$$ 9

in which $E_{\mathrm{adsorbate}/\mathrm{slab}}$ is the total energy of the surface with the adsorbed H₂O and H₂S molecules, $E_{\mathrm{adsorbate}/\mathrm{slab}}$ and $E_{\mathrm{slab}}$ are the energies of stacked structure surface (crystalline and amorphous MIL-68(Fe)) and the isolated H₂O and H₂S molecules, respectively. In addition, the climbing-image nudged elastic band method (CI-NEB) was used to further explore the transition states of H₂S dissociation on the crystalline and amorphous MIL-68(Fe) surface.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

G.D. conceived and supervised the project. G.L. designed the experiment and participated in the entire project. L.Z. and J.Z. carried out the X-ray absorption spectroscopy experiment. Y.L. performed the Density Functional Theory calculations. Y.Z., Z.Z., J.W., L.Q. and P.T. aided in the experimental results discussion. G.D. and G.L. prepared the manuscript. All authors reviewed and approved the final version of the paper.

Peer review

Peer review information

Nature Communications thanks Matteo Valt, Osama Shekhah, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Funding

This study was supported by the National Key R&D Program of China (2020YFB2008701), the National Science Foundation of China (Grant No. 22227809) and Shanghai Science and Technology Innovation Action Plan (22142200300).

Data availability

The data that support the findings of this study are available from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-76704-z.