Introduction

Gas sensing technology, which converts gas components into light and electrical signals, holds promise in environmental monitoring, safety alert systems, and disease diagnosis, offering significant potential in enhancing human safety and quality of life1,2,3. Despite their high sensitivity, low cost, and fast response times, metal oxide semiconductor gas sensors face challenges in miniaturization and performance due to high energy consumption and detection limits4,5. Halide perovskites, with their numerous active sites, high mobility, long diffusion lengths, and excellent tunability, are ideal for preparing high-performance chemical resistive gas sensors, offering significant improvements over traditional materials like metal oxides in detecting low-concentration gases6,7,8. Moreover, lead-free halide perovskites stand out for their environmentally friendly characteristics compared to traditional lead-based counterparts. By eliminating lead, these materials significantly mitigate health and environmental risks, making them a sustainable choice for advanced sensor applications.

Halide perovskites are widely recognized to have intrinsic defect tolerance; however, moderate temperature, humidity, and sunlight can degrade bonds and lead to perovskite crystal decomposition9,10,11. Unsurprisingly, early perovskite gas sensors could only operate in air for a few days12,13. Since then, significant progress has been made, with some sensors demonstrating operational stability for 180 days exposure to the atmosphere and negligible loss of sensing performance14. Despite these improvements, current perovskite gas sensors still do not fully meet the requirements for gas sensing applications. Lead-free halide perovskites, by addressing the toxicity concerns associated with traditional perovskites, offer broader applicability, particularly in environments where environmentally responsible solutions are essential.

In this study, we prepared three lead-free two-dimensional hybrid double perovskites (2D HDPs) with varying cation sizes, all of which demonstrated remarkable environmental stability, remaining operational even after >270 days of exposure to the atmosphere. The characteristic feature of 2D HDPs is a large hydrophobic layer sandwiched between the perovskite inorganic layers, which effectively prevents water molecules from attacking the perovskite, thereby enhancing its environmental stability (Fig. 1). However, this structural improvement reduces the interaction between gas molecules and the perovskite surface. Notably, (BA)4AgBiBr8 (BA = CH3(CH2)3NH3+) can effectively detect 20 ppb CO while maintaining stability, marking a breakthrough in the detection of ultra-low concentration CO. This finding highlights the potential of 2D HDPs to overcome the limitations of traditional halide perovskites in gas sensor technology, paving the way for the practical application of halide perovskite-based gas sensors.

Fig. 1: Schematic diagram of dimensional reduction in alkylamine-assisted 3D double perovskite to achieve long-term stable perovskite-based CO sensing.
figure 1

The black, white, purple, blue, yellow, green, and pink balls represent C, H, N, Cs, Bi, Ag, and Br atoms, respectively. T denotes temperature and hv represents incident light.

Results

Preparation and structural characterization of the two-dimensional hybrid double perovskite sensors

The 3D A2BIBIIIX6 (X = halogen) double perovskite lattice consists of corner-sharing octahedrons, with small ions (Cs+ and MA+) occupying the cuboctahedral cavities at the A-site15,16. However, if larger organic cations replace the ions at the A-site, the 3D structure is transformed into a 2D HDP lattice of A4BIBIIIX8. This lattice comprises single sheets of corner-sharing B − X octahedra, which are templated by ordered arrays of organic cations17,18. To enhance the environmental stability of halide double perovskites, we employed structural modifications by reducing the dimensions of double perovskites. Specifically, we synthesized three 2D HDPs, namely (BA)4AgBiBr8 (BA = CH3 (CH2)3NH3+), (HA)4AgBiBr8 (HA = CH3(CH2)5NH3+), and (OA)4AgBiBr8 (OA = CH3(CH2)7NH3+), by utilizing cations of different sizes (Fig. 2a).

Fig. 2: Structure and characterization of two-dimensional hybrid double perovskites (2D HDPs).
figure 2

a Schematic diagram of the 2D HDPs structure. The yellow, green, and pink balls refer to Bi, Ag, and Br atoms, respectively. bd Representative scanning electron microscopy (SEM) images of (BA)4AgBiBr8, (HA)4AgBiBr8, and (OA)4AgBiBr8 from three independently prepared samples (n = 3), showing similar morphological features. eg High-resolution transmission electron microscopy (HRTEM) diffraction patterns of the corresponding samples. h Powder X-ray diffraction (PXRD) spectra. i Raman shift spectra. j Ultraviolet–visible (UV–vis) absorption spectra. k Tauc plot of (Aν)1/2 versus photon energy (ν) for Cs2AgBiBr6 and (BA/HA/OA)4AgBiBr8 Here, A is the absorption coefficient and ν is the photon energy in eV. The linear region was extrapolated to estimate the optical band gap. ln Attenuated total reflection Fourier transform infrared (ATR-FTIR) spectra of (BA)4AgBiBr8, (HA)4AgBiBr8, and (OA)4AgBiBr8. Source data are provided as a Source Data file.

First, BA/HA/OA, AgBr, and BiBr3 were dissolved in a heated hydrobromic acid solution using the proper stoichiometry. The solution was then slowly cooled to facilitate the formation of (BA)4AgBiBr8, (HA)4AgBiBr8, and (OA)4AgBiBr8 crystals. Their morphologies, crystal phases, and elemental compositions were characterized by scanning electron microscopy (SEM) and transmission electron microscopy (TEM). SEM imaging showed that the three 2D HDPs exhibited flake-shaped structures with smooth surfaces and uniform thickness (Fig. 2b–d). High-resolution TEM (HRTEM) analysis revealed an inter-plane spacing of 3.2 Å in two vertical directions for (BA)4AgBiBr8, corresponding to the (200) plane of (BA)4AgBiBr8 (Supplementary Fig. 7). The presence of clear diffraction spots in the electron diffraction pattern further confirmed the high crystallinity of the three perovskites (Fig. 2e–g). Furthermore, energy-dispersive X-ray spectroscopy demonstrated the uniform distribution of Ag, Bi, and Br elements within the perovskite structures (Supplementary Figs. 810).

Next, we synthesized three different perovskite powders using a rapid cooling method and conducted a comprehensive analysis of their structure, thermal properties, and optical properties. To verify the phase and purity of the powder samples, we performed powder X-ray diffraction (PXRD) measurements. For the layered compounds, we observed the first prominent 2θ peaks corresponding to (001) at 6.5°, 5.0°, and 4.3° for (BA)4AgBiBr8, (HA)4AgBiBr8 and (OA)4AgBiBr8, respectively. This shift in the position of the (001) diffraction peak toward smaller angles is attributed to the increase in spacing between octahedral sheets as the length of the carbon chain increases. The materials were further characterized by attenuated total reflection Fourier transform infrared (ATR-FTIR) spectroscopy and Raman spectroscopy. ATR-FTIR analysis revealed changes in the characteristic peaks of (BA)4AgBiBr8 compared with pure n-butylamine. The N-H δ peak shifted from 1571–1575 cm−1, and the CH2 δ peak shifted from 1463–1472 cm−1, indicating a clear interaction between the butylamine ligands within the perovskite and the metal halide (Fig. 2l). Similar shifts in characteristic peaks were observed for (HA)4AgBiBr8 and (OA)4AgBiBr8 (Fig. 2m, n). Raman spectroscopy results for Cs2AgBiBr6 exhibited three well-defined peaks at 177.9, 135.2, and 77.4 cm−1, corresponding to the A1g, Eg, and T2g vibrational modes, respectively (Fig. 2i)19. When introducing the organic layer, the A1g peak associated with the stretching of [AgBr6]5- and [BiBr6]3- octahedra shifted to lower energy, the Eg peak shifted to higher energy, and the T2g peak became weaker. These alterations are attributed to the distortion and separation of the octahedral sheets, leading to a softening of the phonon modes19,20,21. Thermogravimetric analysis results indicated that Cs2AgBiBr6 remained stable up to 410 °C, retaining over 99% of its initial weight (Supplementary Fig. 11). Two distinct weight loss steps were observed after exceeding 400 °C, with complete material decomposition occurring at 900 °C. The onset of thermal degradation for the 2D hybrid perovskite occurred around 200 °C, regardless of the length of the organic chain. The reduced thermal stability is attributed to the decomposition of the organic portion19.

The optical properties were determined by examining the ultraviolet–visible absorption spectrum. Cs2AgBiBr6 demonstrated absorption around 600 nm, and the absorption peaks of all 2D hybrid perovskites were blue-shifted (Fig. 2j). The corresponding band gaps increased from 1.93 eV for Cs2AgBiBr6 to over 2.1 eV (Fig. 2k). This shift in absorption and increase in band gap is ascribed to the differences in confinement effects and interactions between the constituent atoms of the crystal structures22. The introduction of organic cations increased the band gap by indirectly contributing to the asymmetry in the octahedral sheets, leading to intra- and inter-octahedral distortions. Such distortions are expected to affect the overlap of participating orbitals, consequently increasing the band gap19.

To gain deeper insights into the structural changes occurring in the perovskite after dimensionality reduction, we employed density functional theory (DFT) calculations to simulate the crystal structure, energy band structure, and projected density of states (PDOS) for Cs2AgBiBr6 and (BA)4AgBiBr8, based on their crystallographic information files. The n-butylamine cation was used to transform the 3D parent structure of Cs2AgBiBr6 into a Ruddlesden–Popper-type layered perovskite, denoted as (BA)4AgBiBr8 (Fig. 3a, b). A noticeable difference was observed in the bond lengths between the two structures. In Cs2AgBiBr6, the Ag-Br bond length remains uniform at 2.80 Å throughout the 3D structure (Fig. 3c). However, in (BA)4AgBiBr8, significant bond length deviations were observed (Fig. 3d). The inorganic layer within (BA)4AgBiBr8 undergoes substantial distortion, particularly at the Ag site. In this structure, the Ag-Br octahedron undergoes tetragonal distortion, with an unusually short bond length between Ag and the axial terminal bromine (Ag-Brax = 2.68 Å), whereas the bond length between Ag and the bridging equatorial bromides is longer (Ag-Breq = 3.00 Å; Fig. 3d). This distortion is primarily attributed to the mixing of filled transition metal nd orbitals with empty (n + 1)s orbitals, which stabilizes the linear coordination geometry of Ag. In Cs2AgBiBr6, the Br-4p, Ag-4d, and Bi-6s orbitals contribute to the valence band maximum (VBM), and the conduction band minimum (CBM) is primarily generated by Bi-6p orbitals (Fig. 3g)17,23,24,25. In this case, the orbitals of the A-site cation do not contribute to the VBM or CBM. However, upon introducing organic cations, a significant increase in the band gap is observed (Fig. 3h), resulting from the distortion of octahedral structures and changes in the overlap of participating orbitals19.

Fig. 3: Crystal and energy band structure.
figure 3

Crystal structures of (a) Cs2AgBiBr6 and (b) (BA)4AgBiBr8. Ball and stick models showing the bond lengths (in Å) and in-plane connectivity of [BiBr6] and [AgBr6] octahedra for (c) Cs2AgBiBr6 and (d) (BA)4AgBiBr8. Bond angles of MI − X − MIII for (e) Cs2AgBiBr6 and (f) (BA)4AgBiBr8. The black, white, purple, blue, yellow, green, and pink balls refer to C, H, N, Cs, Bi, Ag, and Br atoms, respectively. Calculated electronic band structure and atomic orbital projected density of states (PDOS) of (g) Cs2AgBiBr6 and (h) (BA)4AgBiBr8. Source data are provided as a Source Data file.

Gas sensing of the two-dimensional hybrid double perovskite-based sensors

Perovskites possess advantageous properties in various chemical applications owing to the active sites exposed by point defects and the ionic nature of the crystal lattice. To investigate this further, three 2D HDPs were employed as chemiresistive gas-sensing materials, and their performance was evaluated using carbon monoxide (CO) as the analyte. Figure 4a–c illustrate that the conductance of the three 2D hybrid perovskites increased rapidly upon exposure to CO and quickly recovered when air was reintroduced. A detailed comparison of response times, recovery times, and detection limits revealed that the incorporation of alkylamines significantly enhanced the performance of the Cs2AgBiBr6 sensor (Fig. 4d, Supplementary Figs. 1217). Among the tested sensors, (BA)4AgBiBr8 exhibited the best overall performance, with a positive linear response to CO across concentrations ranging from 10 ppb to 400 ppb, achieving a detection limit as low as 20 ppb (Fig. 4d and Supplementary Fig. 17b). Using the root mean square deviation (RMSD) method26,27, the theoretical detection limit of the (BA)4AgBiBr8 sensor was calculated as 29.74 ppb, aligning closely with the experimental findings (See Supplementary Information for detailed calculation process). Additionally, (HA)4AgBiBr8 and (OA)4AgBiBr8 were tested under identical conditions. The (HA)4AgBiBr8 sensor achieved an experimental detection limit of 80 ppb and a theoretical detection limit of 103.59 ppb, while the (OA)4AgBiBr8 sensor exhibited an experimental detection limit of 120 ppb and a theoretical detection limit of 140.37 ppb (Supplementary Table 5). Real-time sensing curve cycling experiments for CO concentrations of 0.02, 0.2, 4, and 10 ppm were conducted to evaluate the stability and reproducibility of (BA)4AgBiBr8, (HA)4AgBiBr8, and (OA)4AgBiBr8 (Supplementary Figs. 1820). All three sensors demonstrated similar initial response values across three consecutive assays, highlighting their stability and reproducibility. Importantly, replacing Cs+ with alkyl chains significantly enhanced the environmental stability. The three perovskite-based sensors remained stable for over 270 days (Fig. 4f), marking a new record for perovskite-based sensors. As shown in Fig. 4g, this comparison is made using stability data compiled from previously published literature12,13,14,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59. The complete list of sources is also provided in Supplementary Table 4. The low standard deviation in sensor responses further demonstrates the reliability of all three sensors (Supplementary Figs. 2123).

Fig. 4: Sensing performance.
figure 4

Real-time response of (a) (BA)4AgBiBr8, (b) (HA)4AgBiBr8, and (c) (OA)4AgBiBr8 sensors with increasing concentrations of CO (from 10 ppb to 10 ppm). The insets show the corresponding perovskite crystal structures of each material. The black, white, purple, yellow, green, and pink balls refer to C, H, N, Bi, Ag, and Br atoms, respectively. d Comparison of response times, recovery times, and lower limits of detection (LOD) between Cs2AgBiBr6 and the three two-dimensional hybrid double perovskites (2D HDPs) is presented, with detailed measurements of response time, recovery time, and lower limit of detection provided in Supplementary Figs. 1317. e Schematic diagram of the perovskite sensing structure. f Comparison of long-term sensing stability among the four perovskites. g Long-term time stability of CO sensors based on various perovskite materials. Reference numbers next to the data points correspond to sources listed in the main text reference list12,13,14,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59. The compiled data are detailed in Supplementary Table 4, which uses a separate reference list. Source data are provided as a Source Data file.

To evaluate the effects of O2 and H2O concentrations on sensor performance, we adjusted the carrier gas composition and tested responses under varying oxygen (N2, 20%, 60%, 100% O2) and relative humidity (RH). The sensor’s response decreased at 20% O2 due to O2 molecule occupation of exposed metal sites but improved at higher concentrations due to the formation of reactive oxygen species (Supplementary Figs. 2426)60. Under low humidity, water adsorption hindered CO access to metal sites, reducing response, while high humidity enhanced the response of (BA)4AgBiBr8 (Supplementary Fig. 27), possibly due to material decomposition or denaturation, unlike (HA)4AgBiBr8 and (OA)4AgBiBr8 (Supplementary Figs. 28 and 29), which were protected by longer alkyl chains. XRD analysis confirmed structural changes in (BA)4AgBiBr8 with atmospheric exposure (Supplementary Fig. 30), while no significant changes were observed for (HA)4AgBiBr8 and (OA)4AgBiBr8 (Supplementary Figs. 31 and 32).

Mechanistic inferences

The sensing response of perovskite can be attributed to its semiconductor properties and the interactions between CO molecules and exposed active sites. The ionic lattice of perovskite facilitates CO adsorption at point defects through acid-base interactions, resulting in electron extraction and transfer from CO to (BA)4AgBiBr8, which alters the material’s resistivity (Fig. 5a). To identify the optimal adsorption sites for CO molecules, we calculated the system’s energy following CO adsorption, with the optimal adsorption configuration depicted in Supplementary Fig. 33. Charge difference density maps reveal distinct charge redistributions between CO and adsorption sites on (BA)4AgBiBr8 (Fig. 5a), indicating specific interactions. When CO adsorbs at exposed Bi sites, formed due to the absence of Br ions, a thicker electron accumulation layer is generated, leading to changes in carrier concentration. This phenomenon was further confirmed by sum frequency generation (SFG) spectroscopy (Fig. 5b), where the SFG signal intensity of (BA)4AgBiBr8 increases significantly after CO treatment, suggesting charge transfer. Additionally, the presence of CO characteristic peaks in the SFG spectra indicates that CO molecules exhibit high orientation uniformity on the surface of (BA)4AgBiBr8 at room temperature.

Fig. 5: Characterization of the sensing mechanism.
figure 5

a The charge difference density of adsorption of CO on BA4AgBiBr8 (200) surface. The electron accumulation and depletion are represented with yellow and cyan contours, with iso-surfaces being 0.003 e/Å3. The black, white, purple, yellow, green, and pink balls refer to C, H, N, Bi, Ag, and Br atoms, respectively. Corresponding atomic coordinates are available in Supplementary Data 1. b Sum frequency generation (SFG) spectra of (BA)4AgBiBr8 film before and after CO adsorption. c The Electron Localization Function (ELF) diagram for adsorption of CO on BA4AgBiBr8 surface. The black, white, purple, yellow, green, and pink balls refer to C, H, N, Bi, Ag, and Br atoms, respectively. Corresponding atomic coordinates are available in Supplementary Data 2. d The Crystal Orbital Hamilton Population (COHP) and the corresponding Integrated Crystal Orbital Hamilton Population (ICOHP) value for C-Bi bond in the configuration of CO absorbed on BA4AgBiBr8 (200). The Fermi level (Ef) was set to 0.00 eV. e Total density of states (TDOS) of BA4AgBiBr8 (200) surface, and BA4AgBiBr8 (200) surface with CO molecule. The Ef is set to 0.00 eV. f Projected density of states (PDOS) of 6p orbitals of Bi in the BA4AgBiBr8 (200) surface with a CO molecule absorbed and Bi in the clean BA4AgBiBr8 (200) surface. The Ef is set to 0.00 eV. g PDOS of p orbitals of a CO molecule on the BA4AgBiBr8 (200) surface and an isolated CO molecule. The Ef is set to 0.00 eV. The insets are the partial charge densities with isosurface level set to 0.02 e/Å. The electron accumulation is represented with yellow contour. Source data are provided as a Source Data file.

To further clarify the interaction between CO and (BA)4AgBiBr8, we analyzed the density of states (DOS) before and after CO adsorption. Figure 5e shows negligible changes in the total density of states (TDOS) near the Fermi level of (BA)4AgBiBr8 after CO adsorption, suggesting weak orbital interactions. The projected density of states (PDOS) indicates that these weak interactions primarily originate from the p orbital of CO and the 6p orbital of Bi (Fig. 5f, g). Electron Localization Function (ELF) analysis further supports this conclusion, showing discontinuous electron distributions between CO and (BA)4AgBiBr8 (Fig. 5c), which is characteristic of physical adsorption. Additionally, Crystal Orbital Hamilton Population (COHP) analysis confirms weak orbital coupling (Fig. 5d), as evidenced by minimal antibonding characteristics (few positive COHP values) and a low Integrated Crystal Orbital Hamilton Population (ICOHP) value (-0.53 eV). Although these results indicate that CO molecules undergo physical adsorption on the surface of (BA)4AgBiBr8, the observed significant charge transfer and excellent sensing performance may arise from the strong localized charge redistribution around adsorption sites, which facilitates rapid response and recovery (Fig. 4a).

Discussion

This study advances the field of gas sensing by addressing the limitations of halide perovskite sensors through innovative material modifications. By replacing the cation Cs+ in Cs2AgBiBr6 with long-chain alkylamines, we have developed perovskite materials that effectively mitigate structural decomposition and performance degradation induced by environmental factors. The resultant sensor, (BA)4AgBiBr8, demonstrates exceptional performance with a CO detection limit of 20 ppb and maintains its stability after 270 days of continuous exposure to ambient air. The sensing performance was excellent compared to all other reported electric CO sensors as summarized in Supplementary Table 5. This achievement marks a significant improvement over perovskite sensors and highlights the effectiveness of using two-dimensional hybrid double perovskites to enhance sensor durability and sensitivity. This breakthrough paves the way for practical applications in environmental monitoring, safety systems, and disease diagnosis, thereby significantly advancing the potential of halide perovskite-based gas sensing technology.

Methods

Synthesis of (BA)4AgBiBr8

In the synthesis of the product, 0.9 mmol of AgBr, 0.9 mmol of BiBr3, and 3.6 mmol of n-butylamine solution were added to a 5 mL round bottom flask. Then, 2.6 mL of 48% HBr was added to the flask. The mixture was stirred at 100 °C until all the reactants were completely dissolved. The resulting solution was allowed to cool to room temperature naturally, and the product was obtained. Subsequently, the product was washed multiple times with diethyl ether and dried under vacuum at 60 °C for 10 h.

Synthesis of (HA)4AgBiBr8

In the synthesis of the product, 0.9 mmol of AgBr, 0.9 mmol of BiBr3, and 3.6 mmol of n-hexylamine solution were added to a 5 mL round bottom flask. Then, 2.6 mL of 48% HBr was added to the flask. The mixture was stirred at 130 °C until all the reactants were completely dissolved. The resulting solution was allowed to cool to room temperature naturally, and the product was obtained. Subsequently, the product was washed multiple times with diethyl ether and dried under vacuum at 60 °C for 10 h.

Synthesis of (OA)4AgBiBr8

In the synthesis of the product, 0.9 mmol of AgBr, 0.9 mmol of BiBr3, and 3.6 mmol of n-octylamine solution were added to a 5 mL round bottom flask. Then, 2.6 mL of 48% HBr was added to the flask. The mixture was stirred at 140 °C until all the reactants were completely dissolved. The resulting solution was allowed to cool to room temperature naturally, and the product was obtained. Subsequently, the product was washed multiple times with diethyl ether and dried under vacuum at 60 °C for 10 h.

Fabrication of the sensors

Ultrasonically disperse 20 mg of (BA)4AgBiBr8 in 1 mL of chlorobenzene. Then, apply 20 μL of the dispersion onto an Al2O3 substrate that has been printed with Ag-Pd interdigitated electrodes. The channel width of the electrodes is 200 μm (MJ-10, Beijing Elite Technology Co. Ltd, China). Subsequently, dry the coating for 5 min under an infrared drying lamp.

Gas sensing measurements

The performance of the (BA)4AgBiBr8 sensor was assessed by measuring the current under varying concentrations of CO/Air gas flow at 25 °C. A sealed chamber with gas input, gas exit, and electrical feed, having a capacity of 1200 cm3, was utilized to power the sensor. To monitor the current, DC current-time (I-T) measurements were conducted using a Keithley 4200-SCS instrument. Each I-T measurement lasted for 4 min to ensure accurate and precise experimental data. Prior to introducing the target gas, the chamber was purged with dry air until the sensor’s current stabilized. During the experiments, the airflow rate was maintained at 100 mL/min to minimize its influence on the measurements.

Material characterization and spectroscopic investigation

Scanning electron microscopy (SEM) and energy-dispersive X-ray spectroscopy (EDS) spectra were acquired using a HITACHI Regulus 8230 cold field emission SEM at an acceleration voltage of 15 kV. High-resolution transmission electron microscopy (HRTEM), and selected area electron diffraction (SAED) were performed on a Talos F200X G2 electron microscope (Thermo Scientific) at 200 kV. The structural properties of the perovskite were analyzed using X-ray diffraction (XRD) with a Bruker D8 Advance and Fourier-transform infrared spectroscopy (FTIR) with an IRTracer-100. Sensor performance was evaluated using a Keithley 4200-SCS instrument. The sum frequency generation (SFG) spectrometer laser system was configured by EKSPLA.

Theoretical calculation

All of the structural relaxation and electronic structure calculations were performed using the density functional theory (DFT), as implemented in the Vienna Ab initio Simulation Package (VASP)29,30,61,62. The exchange-functional was treated within the generalized gradient approximation (GGA) of the Perdew-Burke-Ernzerhof (PBE) functional63,64,65. The core and valence electronic interactions were described using the frozen-core projector augmented-wave (PAW) potentials66. The (BA)4AgBiBr8 (200) surface model was constructed for theoretical calculations. The convergence criteria of the force and energy change were 0.01 eV Å–1 and 1 × 10−4 eV per atom in the geometry optimization. The cut-off energy of the plane-wave basis was set as 400 eV to optimize the calculations for atoms and cell optimization. The k-point sampling was modeled from the Gamma scheme with a 3 × 1 × 1 mesh. To avoid interactions between periodic slabs, lattice parameters in the direction perpendicular to the surface were set as 25 Å. The DFT-D3 method was used to evaluate the long-range van der Waals (vdW) contributions67.

Reporting summary

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