Thermal responsive Rh₁-Pd single-atom catalyst for controlling activity in direct formic acid fuel cells

Abstract

Single-atom catalysts are highly active for specific reactions; however, their sinter-resistance and clean surfaces make it difficult to impart them with a “smart” property. Herein we design a thermosensitive single-atom Rh₁-Pd nanosheets (NSs) by employing amino functionalized poly(N-isopropylacrylamide) as thermal responsive gate for the “open”/“closed” control of electro-catalytic activity. Using formic acid electro-oxidation as model reaction, the thermosensitive Rh₁-Pd NSs exhibit high mass activity (2.29 A mg_PGM^-1), strong poisoning resistance, and reversible thermo-responsibility at a lower critical solution temperature (LCST) of 35 ^oC. In-situ spectroscopy and theoretical investigations reveal that the d-electron deficient Rh₁-Pd NSs favor the direct formate pathway and weaken the binding of self-poisonous species, contributing to the high activity and anti-self-poisoning talent. Most importantly, for the direct formic acid fuel cells (DFAFCs), the thermosensitive Rh₁-Pd NSs exhibit a high power density operated below the LCST, while the power drops sharply once above the LCST, realizing the intelligent battery thermal protection for DFAFCs.

Thermal protection is critical to extend the service life and avoid overheating in electrochemical devices. Here, the authors report a thermosensitive single-atom Rh1-Pd electrode that enables high activity towards formic acid electrooxidation under 35 °C, which drop down when the temperature is over 35 °C.

Introduction

Direct liquid fuel cells have attracted wide interest as a type of promising portable-power device in consumer electronics and unmanned electric machines due to their easy storage, high conversion efficiency and environmental benign characteristics^1,2. It is well established that the battery performance and safety strongly hinge upon the operation temperature. An over-high operation temperature will bring safety risks and cause the loss in service life^3–5. In this circumstance, effective battery thermal protection of direct liquid fuel cells is critical to ensure the high-performance operation and avoid the thermal runaway^6,7. However, the existing battery thermal protection methods highly rely on installing temperature sensor or adding thermally responsive materials at cell-to-cell interlayers^8,9. The increased weight, extra electricity consumption cost, and battery assembly complexity leave a large space for further developing new-type battery thermal protection techniques.

Among various liquid fuel cells, direct formic acid fuel cells (DFAFCs) stand out due to their renewable nature, high energy density (1740 Wh kg⁻¹, 2086 Wh L⁻¹), and low membrane crossover^10,11. Pd has long been regarded as the most promising DFAFCs anode electro-catalyst, which follows the so-called direct pathway to proceed formic acid oxidation reaction (FAOR) without the formation of carbon monoxide (CO_ads) intermediates (three main pathways illustrated in Supplementary Fig. 1)¹². Even so, the FAOR stability of Pd catalysts is still criticized due to the easy dissolution of Pd in acidic electrolyte and the self-poison of Pd through surface accumulation of “CO*-like” species (i.e. formate, (COOH)_ad, (bi)sulfate)¹³. For alternatives to substitute Pd, single-atom Rh catalysts start to shine due to their maximum atom utilization, high anti-corrosion capability, and appropriate binding ability with FAOR intermediates¹⁴. For example, Xiong et al. reported a single-atom Rh/N-doped carbon as a highly efficient FAOR electro-catalyst in 2020¹⁵. Currently, the FAOR kinetics of single-atom Rh catalysts are restricted by the lack of a larger surface metal ensemble to facilitate the successive bond cleavage of HCOOH. Therefore, we assume that coupling single-atom Rh sites with Pd catalysts is promising for achieving better FAOR performances.

For the fabrication of single-atom catalysts, high temperature thermal treatments are usually demanded for yielding sinter-resistant and thermodynamically stable single-atom sites¹⁶. During the thermal treatments, surface species (i.e. hydroxyls, carbonates) are desorbed at high temperatures, leading to clean surfaces for the resultant single-atom catalysts¹⁷. Therefore, almost all of the single-atom catalysts exhibit fixed surface structures and stable interface properties, resulting in their insensitivity toward external environment stimulus (e.g. pH, heat)¹⁸. Still now, although single-atom catalysts have achieved satisfactory performances in various fields, giving them a “smart” property is still a blank.

Herein, we design a smart single-atom catalyst, named thermosensitive Rh₁-Pd nanosheets (NSs), by employing amino functionalized poly(N-isopropylacrylamide) (PNIPAM-NH₂) as a thermal responsive gate for the “opening”/“closing” control of electro-catalytic activity (Fig. 1a). The in-situ variable-temperature transmission electron microscopy (TEM), dynamic light scattering (DLS), and near ambient pressure X-ray photoelectron spectroscopy (NAP-XPS) verified the thermos-sensitivity node and temperature-dependent properties of the Rh₁-Pd NSs. Electrochemical measurements, in-situ FT-IR spectroscopy, and density functional theory (DFT) calculations clarified the high FAOR activity and poisoning resistance of the Rh₁-Pd NSs. Then the DFAFCs tests verified the battery thermal protection capability of thermosensitive Rh₁-Pd NSs electrode, which exhibited high power density under the lower critical solution temperature (LCST), but automatically switched to deactivation once above the LCST.

Fig. 1. Microstructures of the thermosensitive Rh₁-Pd NSs.

a Schematics of temperature response mechanism. b FT-IR spectra. Upper part: structural model of the PNIPAM-NH₂. c HRTEM image. d AFM. e EDX mapping images. f–h In-situ variable-temperature microscope images showing the temperature-dependent morphology change. Upper part: schematic illustrations of the surface states of Rh₁-Pd NSs at different temperatures. Blue, orange, red balls and pink spiral line represent Pd, Rh, O atoms and PNIPAM-NH₂, respectively. i–k AC-HADDF-STEM images.

Results

The thermosensitive Rh₁-Pd NSs were prepared by decorating single-atom Rh sites on thermosensitive Pd NSs, as the schematic illustrated Supplementary Fig. 2. First, the thermosensitive Pd NSs were synthesized by a one-pot hydrothermal method, using amino functionalized poly(N-isopropylacrylamide) (PNIPAM-NH₂) as surfactant to avoid the metal particles aggregation. Like PNIPAM, which is well known as a temperature-responsive polymer applied in biological fields^19,20, the PNIPAM-NH₂ is also thermosensitive and amphiphilic (structural formula displayed in Fig. 1b). Below the LCST, the PNIPAM-NH₂ molecules are stretched and expansive. The amide bonds (CONH) in PNIPAM-NH₂ form secondary structure through the interactions of hydrogen bonds and van der Waals forces, generating a hydrated layer and exhibiting the hydrophilicity. Upon above the LCST, the PNIPAM-NH₂ molecules undergo serious aggregation and form a swollen coil structure, resulting in the collapse of hydrated layer and exhibiting the hydrophobicity^21,22. Then the atomic level decoration of Rh sites on thermosensitive Pd NSs was achieved by uniformly ultrasonic dispersion of a small amount of Rh(acac)₃ on thermosensitive Pd NSs in an organic solution and heating at 250 °C to thermally decompose the metal–ligand bonds in Rh(acac)₃. This condition has been proven to prevent the clustering of Rh atoms because of the high surface free energy and interatomic bond energy (93 kJ mol⁻¹) of Rh-Rh²³. Therefore, single-atom Rh sites were expected to form on the surface of Pd NSs. Moreover, the thermos-sensitive PNIPAM-NH₂ molecules were expected to retain on the surface of resultant Rh₁-Pd NSs, taking charge of making Rh₁-Pd NSs a “smart” single-atom catalyst.

Fourier transform infrared spectroscopy (FT-IR) measurements were conducted to verify the reservation of PNIPAM-NH₂ on the surface of Rh₁-Pd NSs. As illustrated in Fig. 1b, the as-obtained Rh₁-Pd NSs exhibit typical stretching vibrational peaks at 3310–3436, 1650, 1546, and 1387 cm⁻¹, attributed to the CH₂NH₂, C = O (amide I band), the deformation of N-H bond and CH(CH₃)₂ groups of PNIPAM-NH₂, respectively. These characteristic peaks are similar to those of pure PNIPAM-NH₂ (grey line), indicating that PNIPAM-NH₂ molecules are well retained on the surface of Rh₁-Pd NSs. Typical TEM image demonstrates the porous and freestanding sheet-like structure of Rh₁-Pd NSs with a lateral size up to ≈ 600 nm (Fig. 1c). Atomic force microscope (AFM) shows a bumpy height profile with averaging thickness of ca. 5.6 nm, demonstrating a rough surface for the Rh₁-Pd NSs (Fig. 1d). Energy dispersive X-ray spectroscopy (EDS) elemental mapping images represent the Pd, Rh, N, O signals evenly distributed over the Rh₁-Pd NSs, suggesting the successful decoration of Rh on Pd nanosheets (Fig. 1e). Meanwhile, the N signal refers to the existence of PNIPAM-NH₂ on the surface of Rh₁-Pd NSs, consistent with the FT-IR analysis. The content of Rh in Rh₁-Pd NSs was determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES), suggesting a value of approx. 0.98 wt% for Rh.

In-situ variable-temperature TEM images were performed to capture the morphological alterations in thermosensitive Rh₁-Pd NSs upon the application of temperature (Fig. 1f–h). At 20 ^oC, the PNIPAM-NH₂ molecules are linear and hard to be captured by electron microscopy. TEM image of the thermosensitive Rh₁-Pd NSs shows numerous interlaced nanowires/nanoparticles with clear edge lines (Fig. 1f). These nanowires/nanoparticles with an averaging diameter of ≈3.2 nm tend to bend and interweave together to form the 2D knitted Rh₁-Pd NSs with high porosity. At 40 °C, the PNIPAM-NH₂ molecules started to become swollen and agglomerate on the surface of the thermosensitive Rh₁-Pd. Therefore, TEM image shows a misty surface for the projection area of the Rh₁-Pd NSs, indicating their slightly marginal gelatinization (Fig. 1g). Upon heating up to 60 °C, TEM image (Fig. 1h) shows an unclear surface for the projection area of the thermosensitive Rh₁-Pd NSs, pointing to the coverage of a thick mass of PNIPAM-NH₂ polymer that causing a pronounced gelatinization. In comparison with the TEM images out of focus (Supplementary Fig. 3), the TEM images of the thermosensitive Rh₁-Pd NSs taken at 60 °C exhibit indefinable sample surface but clear background stripes, as marked by red dash circles, which make an apparent difference. Furthermore, in order to clarify that the gelatinization of thermosensitive Rh₁-Pd NSs at 60 °C is induced by the surface PNIPAM-NH₂, the thermosensitive Rh₁-Pd NSs were underwent an ultraviolet/ozone system treatment to remove the surface PNIPAM-NH₂. As displayed in Supplementary Fig. 4, TEM images show that the morphology of the Rh₁-Pd NSs after the removal of PNIPAM-NH₂ was almost unchanged along with the temperature raising from 20 °C to 60 °C, indicating the crucial role of PNIPAM-NH₂ for inducing the gelatinization effect.

The aberration-corrected high-angle annular dark field scanning transmission electron microscopy (AC-HAADF-STEM) images show the atomic arrangements of Rh₁-Pd NSs (Fig. 1i–k), with polycrystalline nature evidenced by selected area electron diffraction (Supplementary Fig. 5). Each wavy nanowire is highly twisted with obvious grain boundaries between adjacent particle units (Fig. 1i). A large number of defect-rich structures (lattice distortions, atomic steps, and edges) are observed in Rh₁-Pd NSs, which could serve as efficient active sites for electro-catalysis^24,25. Possibly due to the similar atomic radius between Rh and Pd, it is hard to distinguish Rh single atom from Pd NSs^26,27. Due to the high distortion of Pd atoms, the lattice fringes are discontinuous, with lattice spacing measured to be 0.230 and 0.237 nm at different regions of an individual particle, respectively, assignable to the (111) planes of face-centered cubic (fcc)-structured Pd. Such lattice misfit leads to a strong surface tensile strain along the grain boundary of an individual particle, which reaches the value of +0.20% and −0.20%, respectively, as revealed by the geometric phase analysis of atomic columns (Fig. 1j). Figure 1k shows the atomic dislocations at the junction of adjacent nanoparticles, together with intermittent lattice streaks, atomic vacancies, and dark/bright spots, further confirming the rich-defects feature of the thermosensitive Rh₁-Pd NSs.

The thermos-sensitivity node and temperature-dependent property of Rh₁-Pd NSs were real-time monitored by a series of in-situ variable-temperature techniques. First, DLS measurements, which could describe the dispersion behavior of particles in the fluid^21,28, were performed to evaluate the hydrodynamic diameter of Rh₁-Pd NSs at different temperatures. As shown in Fig. 2a, a sharp diameter increase emerges at 35.0 °C, indicating an exact LCST of 35.0 °C for Rh₁-Pd NSs. Below the LCST, the hydrodynamic diameter is stabilized at ca. 284 nm. Above the LCST, the hydrodynamic diameter increases dramatically from 284 nm to 832 nm, suggesting the serious aggregation of Rh₁-Pd NSs. In addition, the temperature-reversible tests between 20 °C and 60 °C for 4 cycles also suggest the temperature sensitivity and reproducibility of hydrodynamic diameter for Rh₁-Pd NSs (Fig. 2b). In situ zeta potential tests were performed to investigate the variation of surface charge with temperature for Rh₁-Pd NSs (Fig. 2c and Supplementary Fig. 6). Below 35 °C, the Rh₁-Pd NSs present a positive ζ-potential value at ca. +0.51, due to the existence of ammonia groups from PNIPAM-NH₂^29,30. Above 35 °C, the ζ-potential of Rh₁-Pd NSs exhibits a sharp increase to +5.05, indicating a high temperature sensitivity of the surface electrostatic charges on Rh₁-Pd NSs. This could be explained by the aggregation of Rh₁-Pd NSs above the LCST, which leads to the decrease of total specific surface area. In this condition, particle population in per unit volume is decreased and more crowded positive charge is gathered on the surface of Rh₁-Pd NSs (schematic illustrated in Fig. 2c).

Fig. 2. In-situ variable-temperature measurements for the thermosensitive Rh₁-Pd NSs.

a DLS diameter as a function of temperature. b Plot showing the reversible thermal expansion and shrinkage. The error bars were obtained by three parallel experiments. c In-situ zeta potentials. d In-situ NAP-XPS spectra at N 1 s, Pd 3 d, and Rh 3 d orbits. e–g Variation of binding energy as a function of temperature. Inserts in (a, c, e–g): schematic illustrations of the thermosensitive behaviors of Rh₁-Pd NSs below and above the LCST. Blue, orange, red balls and pink spiral line represent Pd, Rh, O atoms and PNIPAM-NH₂, respectively.

In-situ variable-temperature NAP-XPS measurements were performed to investigate the variation of valence state with temperature for Rh₁-Pd NSs (Fig. 2d). First, the wide range XPS spectra were provided to show the intensity variations of Rh, Pd, and N signals along with the temperatures (Supplementary Fig. 7). The Rh and Pd signals at 20 and 30 °C are sharper than those at 40–60 °C, suggesting the easier detection of Rh and Pd on the surface of Rh₁-Pd NSs below the LCST. In contrast, the N signal at 40–60 °C is stronger than those at 20 and 30 °C This could be attributed to the agglomeration and coverage of PNIPAM-NH₂ on the surface of Rh₁-Pd NSs above the LCST. For the deconvoluted N 1 s spectra, a constant peak is observed at 400.3 eV from the entire temperature range, corresponding to the -N-C = O bond of PNIPAM-NH₂ in Rh₁-Pd NSs. Meanwhile, below the LCST, a main peak at 399.4 eV is observed for Rh₁-Pd NSs, attributed to the -N-H-O bond. While above the LCST, the binding energy at 399.4 eV is downshifted to 399.1 eV, pointing to the dissolution of -N-H-O bond and the predomination of -N-H bond³¹. This result indicates the breakage of hydrogen bonds and the disappearance of hydration layer above the LCST, resulting in a hydrophobic surface for Rh₁-Pd NSs, as illustrated in Fig. 2e. The high-resolution Pd 3 d XPS spectra also suggest the switch of hydrophilic and hydrophobic surface for Rh₁-Pd NSs by employing LCST as a temperature node (middle region in Fig. 2d). Below the LCST, the deconvoluted Pd 3 d spectra exhibit two coupled peaks assignable to the main Pd⁰ state (3d_5/2 = 335.8 eV, 3d_3/2 = 341.0 eV) and the secondary Pd^II state (3d_5/2 = 336.8 eV, 3d_3/2 = 342.0 eV), corresponding to the slight surface oxidation of Pd³². Above the LCST, the binding energies of Pd 3 d spectra exhibit an obviously negative shift (Pd⁰ 3d_5/2 = 335.2 eV, Pd⁰ 3d_3/2 = 340.4 eV), which could be attributed to a wetting transition from hydrophilicity to hydrophobicity on the surface of Rh₁-Pd NSs. The hydrophobic fragments adsorbed on Pd surface could decrease the surface energy, resulting in the downshift of Pd 3 d XPS spectra, as illustrated in Fig. 2f³³. For the Rh 3 d spectra of Rh₁-Pd NSs, two main peaks located at 312.8 and 308.7 eV are observed below the LCST, corresponding to the oxidation state of Rh^34,35. In contrast, above the LCST, no signal assignable to Rh species is detected, which could be attributed to the coverage of shrunken PNIPAM-NH₂ that blocks the Rh sites of Rh₁-Pd NSs, as illustrated in Fig. 2g. Overall, above in situ measurements verify the thermos-sensitivity of Rh₁-Pd NSs in aspects of hydrodynamic diameter, charge density, and valence state, making the Rh₁-Pd NSs promising for serving as a “smart” electro-catalyst.

Figure 3a shows the X-ray diffraction (XRD) pattern of the thermosensitive Rh₁-Pd NSs. All the diffraction peak locations are similar to those of the Pd NSs, corresponding to the standard fcc Pd (JCPDS-PDF 87-0641). The broadened diffraction peaks imply that the Rh₁-Pd NSs might be constructed by nanoscale substructures with small grain size, matching the morphological observations in Fig. 1^36,37. Meanwhile, no characteristic peak referring to Rh species is detected. The Brunauer-Emmett-Teller (BET) surface area and pore-size distribution of thermosensitive Rh₁-Pd NSs were investigated by N₂ adsorption-desorption measurements. As shown in Fig. 3b, the hysteresis loop exhibits a type IV curve with a high relative pressure (p/p₀) range of 0.8-1.0, corresponding to the mesoporous structures (2–50 nm). The BJH pore-size distribution inset of Fig. 3b indicates the presence of multimodal porosity with an averaging diameter of 27.8 nm. The single-atomic dispersion form of Rh in thermosensitive Rh₁-Pd NSs was clarified by X-ray absorption fine structure spectroscopy (XAFS). The X-ray absorption near edge structure (XANES) profiles for Rh K-edge show that the adsorption threshold position and white line intensity of Rh₁-Pd NSs closely resemble those of Rh₂O₃ (Fig. 3c), indicating that the valence state of Rh in Rh₁-Pd NSs is near +3. Through fitting the white-line peak area, the averaging oxidation state of Rh in Rh₁-Pd NSs was +2.85 (Fig. 3d and Supplementary Fig. 8). The extended XAFS (EXAFS) demonstrates a prominent peak at ca. 1.54 Å for Rh₁-Pd NSs, attributed to first-shell Rh-O bonds (Fig. 3e and Supplementary Fig. 9). Meanwhile, a minor signal at ca. 2.30 Å is observed for Rh₁-Pd NSs, which is slightly lower than the Rh-Rh bond at 2.45 Å and can be assigned to the Rh-Pd signal at other shells. The absence of Rh-Rh/Rh-Pd bonds at the first-shell excludes the formation of Rh particles and RhPd alloy^15,38. Then, quantitative EXAFS curve-fitting analysis (Fig. 3f, Supplementary Fig. 10 and Supplementary Table 1) reveals that the coordination number of the Rh-O bond in Rh₁-Pd NSs is 3.2. Therefore, a structural model with each Rh atom coordinated with 3 oxygen atoms is illustrated for Rh₁-Pd NSs (insert in Fig. 3f). The averaging bond length of Rh-O is fitted as 2.02 Å, which is much shorter than that of Rh-Rh (2.68 Å), confirming the much more stable Rh-O bond than Rh-Rh in thermodynamics. The XAFS spectra for Pd K-edge were performed to further confirm the non-existence of Pd-Rh alloy in Rh₁-Pd NSs. As shown in Fig. 3g, similar absorption threshold positions are observed for Rh₁-Pd NSs, Pd NSs and Pd foil, indicating the dominant metallic Pd in both Rh₁-Pd NSs and Pd NSs. The corresponding FT-EXAFS spectra (Fig. 3h) exhibit a prominent peak at 2.65 Å for both the Rh₁-Pd NSs and Pd NSs, attributed to the Pd-Pd metal coordination, which is in line with that of Pd foil³⁹. Importantly, the characteristic peak of Rh₁-Pd NSs does not show any shift in comparison with that of Pd NSs, indicating the non-existence of Pd-Rh metallic bond in Rh₁-Pd NSs. The quantitative EXAFS curve-fitting analysis (Supplementary Fig. 11, 12 and Supplementary Table 2) also suggest the similar coordination parameters between Rh₁-Pd NSs and Pd NSs, which could exclude the formation of Pd-Rh alloy in Rh₁-Pd NSs. Wavelet transform (WT) analyses were performed to better resolve the K space feature and radial distance for Rh₁-Pd NSs (Fig. 3i, j). The WT plot for Rh k-edge of Rh₁-Pd NSs exhibits only a single intensity maximum at about 5.1 Å⁻¹, consistent with the intensity maximum associated with Rh-O backscattering in Rh₂O₃ (Supplementary Fig. 13). Notably, no intensity maximum at 10.5 Å⁻¹ associated with the Rh-Rh coordination is detected, further confirming the atomic Rh-O dispersion in Rh₁-Pd NSs. The WT plot for Pd k-edge of Rh₁-Pd NSs exhibits only a single intensity maximum at about 10.1 Å⁻¹, which does not show any shift in comparison with the Pd-Pd backscattering in Pd NSs and Pd foil, further excluding the non-existence of PdRh alloy related species in Rh₁-Pd NSs (Supplementary Fig. 14).

Fig. 3. Atomic structure analysis of thermosensitive Rh₁-Pd NSs.

a XRD pattern. b N₂ absorption/desorption isotherms. c XANES spectra, d Valence state fitting, e Fourier transformed EXAFS spectra, and f Atomic configuration fitting for Rh K-edge and the corresponding structural model. Blue, orange and red balls represent Pd, Rh, O atoms, respectively. g XANES spectra and h Fourier transformed EXAFS spectra for Pd K-edge. i, j Wavelet transformations.

The FAOR performance of Rh₁-Pd NSs was first assessed in three-electrode system at a constant operation temperature of 25 °C. To shed lights on the synergistic effects between single-atom Rh sites and Pd NSs, pure Pd NSs and commercial Pd/C were chosen as reference catalysts. Cyclic voltammogram (CV) curves tested in 0.5 M H₂SO₄ electrolyte exhibit butterfly peaks at the H_upd adsorption/desorption regions (0–0.2 V, H^- +e^- =H_upd), which are typical characteristics for Pd-based catalysts (Fig. 4a). The electrochemical active surface area (ECSA, Supplementary Fig. 15) of Rh₁-Pd NSs was measured to be 7.10 cm², which was much larger than that of Pd NSs (5.78 cm²) and commercial Pd/C (0.98 cm²). The mass-normalized FAOR curves (Fig. 4b) exhibit a large peak current density for Rh₁-Pd NSs catalyst (2.294 A mg_PGM⁻¹), which is 2.5-fold and 6.6-fold higher than that of Pd NSs (0.933 A mg_PGM⁻¹) and Pd/C (0.349 A mg_PGM⁻¹), respectively. Meanwhile, the Rh₁-Pd NSs catalyst shows a peak oxidation potential of 0.401 V vs. RHE, which is much lower than that of Pd NSs (0.418 V) and Pd/C (0.440 V), indicating the enhanced selectivity for direct FAOR pathway on Rh₁-Pd NSs⁴⁰. The corresponding Tafel curves suggest the preferential FAOR kinetics for Rh₁-Pd NSs at electrochemical control region (Supplementary Fig. 16). The ECSA-normalized FAOR curves displayed in Supplementary Fig. 17 also show the largest oxidation peak current on Rh₁-Pd NSs (6.45 mA cm⁻²) than that of Pd NSs (3.23 mA cm⁻²) and Pd/C (1.42 mA cm⁻²), suggesting the high intrinsic activity of Rh₁-Pd NSs. Overall, as summarized in Fig. 4c, the promotions of specific activity (SA) and mass activity (MA) for Rh₁-Pd NSs confirm the pivotal role of synergistic effects between single-atom Rh sites and Pd nanosheets, which could strongly enhance the FAOR activity. Furthermore, the Rh₁-Pd NSs catalyst is also competitive in state-of-art FAOR catalysts, as displayed in Supplementary Fig. 18 and Supplementary Table 3. In order to explore the role of the single-atom form of Rh in enhancing the electro-catalytic performance, two reference samples were synthesized by using the standard protocol of Rh₁-Pd NSs except for adding 5 mL and 10 mL of Rh(acac)₃, respectively. The resulted samples were named as 10% Rh-Pd NSs and 20% Rh-Pd NSs, respectively. TEM images (Supplementary Fig. 19a, b), XRD pattern (Supplementary Fig. 19c), and XPS spectra (Supplementary Fig. 19d–f) show that the two samples are sheet-like structure with clustered/metallic Rh(0) species on Pd NSs. The CV curves of 10% Rh-Pd NSs and 20% Rh-Pd NSs (Supplementary Fig. 20a) exhibit two reduction peaks at 0.72 V and 0.45 V, corresponding to the reduction of Pd-O and Rh-O, respectively, indicating the different electrochemical behaviors between metallic Rh-Pd NSs and single-atom Rh₁-Pd NSs²³. The FAOR curves (Supplementary Fig. 20b) show that the mass activity of Rh₁-Pd NSs is much larger than those of the 10% Rh-Pd NSs and 20% Rh-Pd NSs, confirming the pivotal role of single-atom Rh sites, which are more conducive to enhance the FAOR performance of Pd NSs than the clustered/metallic Rh species.

Fig. 4. Electro-catalytic FAOR performance of thermosensitive Rh₁-Pd NSs at a constant temperature of 25 ^oC.

a CV curves in N₂-saturated 0.5 M H₂SO₄ solution. b FAOR curves in N₂-saturated 0.5 M H₂SO₄ + 0.5 M HCOOH solution (without iR correction, scan rate: 50 mV s⁻¹, mass loading: 20 μg). c Peak mass and specific activities of different catalysts. d Chronoamperometry tests at 0.25 V vs. RHE. Insert: ADTs for 5000 cycles. e Comparisons of the j_θ=0 and k_ads for different catalysts. f CO-stripping tests. In-situ FTIR spectra in 0.5 M H₂SO₄ + 0.5 M ¹³C isotopes-labelled HCOOH solution for g–i Rh₁-Pd NSs and j–l Pd NSs.

The FAOR stability of Rh₁-Pd NSs was evaluated by chronoamperometry (CA) and accelerated durability tests (ADTs). The Rh₁-Pd NSs show much higher initial current density with slow decay than that of Pd NSs and commercial Pd/C over the entire time course (Fig. 4d). After 24 h, the steady-state FAOR current density on Rh₁-Pd NSs decreases to 76.7% of its initial value, indicating the long-term durability of Rh₁-Pd NSs. The accelerated durability tests (Supplementary Fig. 21 and histograms in Fig. 4d) also suggest the high FAOR stability of Rh₁-Pd NSs, which retain merely ≈86.8% of the initial oxidation peak current after 5,000 cycles, surpassing that of Pd NSs (53.8%) and commercial Pd/C (23.9%). AFM space diagram, TEM image and XRD pattern (Supplementary Figs. 22, 23) of Rh₁-Pd NSs after ADTs show the well retention of morphology and crystalline structure, verifying their structural robustness after long-term FAOR operation. The FT-IR spectra show that the stretching vibrational peaks are almost unchanged after ADTs, indicating the well retention of PNIPAM-NH₂ on the surface of Rh₁-Pd NSs after long-term FAOR operation (Supplementary Fig. 24).

The FAOR kinetic data for direct and indirect pathway were investigated by pulsed voltammetry, which could obtain the current transients at a series of constant potentials (Supplementary Fig. 25). The extrapolated current in the transient at t = 0 (j_θ=0) was used to determine the current density for direct pathway, and the rate constant for the CO formation (k_ads) was used to evaluate the indirect pathway rate⁴¹. As listed in Fig. 4e and Supplementary Table 4, the maximum j_θ=0 of the Rh₁-Pd NSs catalyst reaches a value of 24.32 mA cm⁻², which is much higher than that of the Pd NSs (6.06 mA cm⁻²) and Pd/C (12.77 mA cm⁻²). Meanwhile, the maximum k_ads for the Rh₁-Pd NSs catalyst show a value of 4.75 s⁻¹, which is lower than that of the Pd NSs (19.74 s⁻¹) and Pd/C (4.84 s⁻¹). The high j_θ=0 but low k_ads suggest the preferential direct dehydrogenation pathway on Rh₁-Pd NSs than Pd NSs, which could be attributed to the synergistic effects between single-atom Rh sites and Pd NSs. The anti-CO ability of Rh₁-Pd NSs was demonstrated by CO stripping tests, through bubbling excess CO_gas into the electrolyte to simulate the toxic environment (Fig. 4f). The Rh₁-Pd NSs catalyst exhibits a much smaller oxidation peak area than that of Pd NSs and Pd/C, validating its weakened CO adsorption. Meanwhile, the CO oxidation peak current of Rh₁-Pd NSs is negatively shifted by 87 and 50 mV than that of Pd NSs and Pd/C, indicating the much easier removal of poisoning species on Rh₁-Pd NSs.

The in situ FTIR spectra (Supplementary Fig. 26) were employed to compare the dynamic FAOR processes between Rh₁-Pd NSs (Fig. 4g-i) and Pd NSs (Fig. 4j-l). In order to eliminate the interference of CO₂ from the environment, isotope ¹³C labelled HCOOH was used for tracing the evolved carbon-containing species during the FAOR. A positive band at ca. 2270 cm^-1 is detected for both two catalysts, which is attributed to the asymmetric stretching vibration of O − ¹³C − O in ¹³CO₂ (the ultimate product for FAOR)^42,43. Obviously, the ¹³CO₂ signal on Rh₁-Pd emerges at a much lower potential (0.15 V, Fig. 4h) than that of Pd NSs (0.50 V, Fig. 4k), indicating the much earlier FAOR kinetics on Rh₁-Pd NSs. Meanwhile, for the Rh₁-Pd NSs, no band associated with ¹³CO_ads species is resolved across the entire potential range, suggesting a dominant direct dehydrogenation pathway and a rapid elimination of poisoned species on Rh₁-Pd NSs (Fig. 4i). In contrast, the pure Pd NSs catalyst exhibits the stretching vibration of bridged-¹³CO species (¹³CO_B) at 1858 cm^-1, which remains un-oxidized even at 1.15 V, suggesting a dual-pathway reaction and a limited poisoned species removal ability (Fig. 4l). Based on above analysis, we attribute the high FAOR performance of Rh₁-Pd NSs catalyst to the introduction of single-atom Rh sites on Pd NSs, which greatly enhances the selectivity of direct dehydrogenation pathway and inhibits both the formation and coverage of poisoned species.

DFT calculation was carried out to understand the origin of the high FAOR activity and CO tolerance on Rh₁-Pd NSs (Rh₁-O₃-Pd(111) model), with the same analysis for Pd NSs (Pd(111) model, Supplementary Fig. 27). Three main reaction pathways were considered, namely direct formate pathway (HCOO*, red line), direct carboxyl pathway (COOH*, green line), and indirect pathway (sequentially generating COOH* and CO*, blue line), respectively. For the Pd NSs, the free energy profiles of FAOR demonstrate that the direct carboxyl pathway is the most favorable, with C-H cleavage of HCOOH* to form COOH* as rate-determining step and an endothermic energy barrier of +0.21 eV is required (Fig. 5a and Supplementary Fig. 28). In contrast, for the Rh₁-Pd NSs, the most favorable pathway is the direct formate pathway, with C-H cleavage of HCOO* as the rate-determining step and an exothermic energy barrier of −0.22 eV is required, which is much lower than that of direct carboxyl pathway with endothermic requirement (Fig. 5b and Supplementary Fig. 29). Moreover, the Gibbs free energies change (ΔG) of HCOOH(g) → HCOO* was used to evaluate the priority of direct formate pathway. The Rh₁-Pd NSs exhibit much lower ΔG_{HCOOH(g)→HCOO*} (-0.24 eV) than that of Pd NSs ( + 0.38 eV), demonstrating the significantly boosted O-H cleavage ability of HCOOH* to form HCOO* on Rh₁-Pd NSs. This result indicates the pivotal role of Rh in Rh₁-Pd NSs, which is constant with the references that the single-atom Rh sites favor the direct formate route and reduce the FAOR energy barrier¹⁵.

Fig. 5. Theoretical FAOR mechanism of thermosensitive Rh₁-Pd NSs.

Free energy diagrams for FAOR on a Pd NSs and b Rh₁-Pd NSs, respectively. Below: optimized structures of the intermediates. c Potential energy profiles of HCOOH decomposition to CO. d Potential energy profiles of H₂O dissociation. e Adsorption energy of different self-poisoning species. f, g Charge density difference of CO* adsorption with an isosurface value of 0.0001 e bohr^-3. The blue and yellow regions represent electron accumulation and depletion, respectively. h PDOS for Pd-4d, Rh-4d and O-2p of Rh₁-Pd NSs. Colors in a, b, f, g: blue, orange, red, brown and white balls represent Pd, Rh, O, C and H atoms, respectively.

To elucidate the origin of poisoning resistance of Rh₁-Pd NSs, the activation barrier of formic acid decomposition to produce CO, the activation barrier of H₂O activation to accelerate CO oxidation, and the adsorption energy of poisoned species were considered (Fig. 5c–e). As shown in Fig. 5c, the adsorption energy of HCOOH is −0.48 eV and −0.92 eV on Rh₁-Pd NSs and Pd NSs, respectively, suggesting the much weaker HCOOH adsorption on Rh₁-Pd NSs. Then, the subsequent decomposition of HCOOH to produce CO occurs hardly readily on Rh₁-Pd NSs, with an activation barrier as high as 1.28 eV. In contrast, on the Pd NSs, the activation barrier to produce CO is only 0.52 eV, suggesting the easy formation of CO on Pd NSs. Moreover, the activation barrier of H₂O dissociation to produce OH^- is found to be much lower on Rh₁-Pd NSs (0.72 eV) than that on Pd NSs (2.57 eV), suggesting the accelerated H₂O activation on Rh₁-Pd NSs, which can easily oxide and remove the CO species (Fig. 5d). This result is constant with the ΔG of CO* + OH* + H⁺+e^-→ CO_2(g) + 2(H⁺+e^-) illustrated in Fig. 5a, b, which indicates the more favorable CO oxidation on Rh₁-Pd NSs (exothermic, −0.35 eV) than Pd NSs (exothermic, −0.31 eV). Also, the easier CO oxidation on Rh₁-Pd NSs can be demonstrated by the more negative onset potential of CO stripping in Fig. 4f. Since self-toxicity is a key factor to affect the performance of Pd catalysts, the adsorption energy of typical self-toxic species (CO*, HCOO*, COOH*) on Rh₁-Pd NSs and Pd NSs were compared (Fig. 5e). In comparison with Pd NSs, all these poisoned species exhibit the weakened adsorption on both the Rh site and Pd site of Rh₁-Pd NSs, indicating the infrequent self-poison and surface passivation on Rh₁-Pd NSs. The weakened adsorption of CO*, HCOO*, COOH* at both of Rh site and Pd site of Rh₁-Pd NSs can be attributed to the downward shift of the d orbital below the Fermi level (Fig. 5h, more details in Supplementary Fig. 30, 31). Moreover, the weakened adsorption could also be attributed to the d-electron deficiency of the single-atom Rh sites of Rh₁-Pd NSs, as evidenced by the charge density difference analyses (Fig. 5f–g). A pronounced yellow area is expanded at Rh sites of the Rh₁-Pd NSs, indicating the electron deficient property, which could eliminate the d-electron back donation to poisoned species. Meanwhile, Bader charge analyses suggest that the electron transfer between CO* and Rh₁-Pd NSs is much smaller than CO* and Pd NSs, further confirming the weakened toxicant adsorption on Rh₁-Pd NSs. Hence, the high barrier to produce CO, the easy H₂O activation to accelerate CO oxidation, and the unfavorable binding with poisoned species contribute to the high CO tolerance of Rh₁-Pd NSs.

The “smart” behaviors of thermosensitive Rh₁-Pd NSs were verified by examining their FAOR activity at different temperatures. Generally, a high temperature is beneficial for fast proton-electron transport and easy electrolyte penetration of electro-catalysis, resulting in positive correlation between temperature and FAOR activity⁴⁴. As the FAOR curves in Fig. 6a, the temperature-activity relationship of commercial Pd/C matches well with this trend. In contrast, for the Rh₁-Pd NSs catalyst, a completely different temperature-activity relationship is observed (Fig. 6b). From 20 °C to 35 °C, the first temperature range, the FAOR activity of Rh₁-Pd NSs catalyst is proportional to the temperature, reaching a maximized mass activity of 2.58 A mg⁻¹ at 35 °C (LCST). From 35 °C to 60 °C, the second temperature range, the FAOR activity of Rh₁-Pd NSs catalyst gradually attenuates, indicating an entirely different activity trend against conventional FAOR catalysts, as the broken line graph shown in Fig. 6c. This result manifests the key role of PNIPAM-NH₂ as a quick-witted gate keeper to control the FAOR activity of Rh₁-Pd NSs catalyst by using temperature as a trigger. Below the LCST, PNIPAM-NH₂ chains are stretched and hydrophilic, allowing sufficient and easy connection between electrolyte/reactants and Rh₁-Pd NSs catalyst. Therefore, the FAOR activity of Rh₁-Pd NSs is proportional to temperature. Above the LCST, PNIPAM-NH₂ chains are swollen and hydrophobic, burying the Rh₁-Pd active sites and resulting in the catalyst inactivation. Moreover, the activation and deactivation of the thermosensitive Rh₁-Pd NSs are recyclable, as evidenced by monitoring the chronoamperometry curves under a cycling temperature regime (Fig. 6d). When the temperature is dipping and heaving, the Rh₁-Pd NSs catalyst can fully regain its previous catalytic activity. After more than 10 cycles, no obvious decay in activity is observed for Rh₁-Pd NSs, demonstrating a fine reversibility of the Rh₁-Pd NSs as a temperature-responsive electro-catalyst. To investigate the practicability of thermos-sensitive Rh₁-Pd NSs catalyst for DFAFCs, the electrode membrane assembly (MEA) was prepared with Rh₁-Pd NSs as the anode and commercial Pt/C catalyst as the cathode (Supplementary Figs. 32 and 33). The polarization and power density curves reveal that the optimal operation temperature is 35 °C, well matching the LCST (Fig. 6e, f). At 35 °C, the open circuit voltage of single-cell with Rh₁-Pd NSs anode reaches a value of 0.92 V, which is much higher than that with commercial Pd/C (0.73 V, Supplementary Fig. 34). Meanwhile, the corresponding maximum power density of Rh₁-Pd NSs anode reaches 263.4 mW cm⁻², which is 2.8 times higher than that of commercial Pd/C (93.1 mW cm⁻²). Furthermore, taking common operation temperature of 30 °C as a benchmark for comparison, the power density of single-cell with Rh₁-Pd NSs electrode is still higher than those of the state-of-art single-cells operated at 80 °C (Fig. 6g and Supplementary Table 5)^15,45–47. The mass normalized power density curves show a maximum mass power density of 52.7 mW mg^-1 for the Rh₁-Pd NSs anode at 35 °C, which is still higher than that of commercial Pd/C (18.6 mW mg^-1, Supplementary Fig. 35). Moreover, as listed in Supplementary Table 6, the mass power density of Rh₁-Pd NSs is also higher than those of the control samples. The DFAFC stability tests at different temperatures (Supplementary Fig. 36) show that the Rh₁-Pd NSs electrode exhibits long-term durability below the LCST (i.e. 30, 35 °C). At 35 °C, it retains about 86.9% of the initial current density after more than 100 h, suggesting its promising future in stably catalyzing anodic fuel oxidation. TEM image and XRD pattern of the Rh₁-Pd NSs after life tests indicate the well retention of morphology and crystalline structure, verifying the stable state of Rh₁-Pd NSs electrode for DFAFCs (Supplementary Fig. 37).

Fig. 6. Temperature-dependent FAOR and DFAFC performance of thermosensitive Rh₁-Pd NSs.

a, b FAOR activity of Pd/C and Rh₁-Pd NSs at different temperature (without iR correction, scan rate: 50 mV s⁻¹, mass loading: 20 μg, temperature: 20–60 °C). c FAOR activity trend as a function of temperature. d Controllable on/off conversion of FAOR activity by periodic fluctuations in temperature. e, f Steady-state polarization and power density curves of Rh₁-Pd NSs for a single cell at different temperatures (anode: Rh₁-Pd NSs, catalyst loading: 5 mg cm^-2; cathode: Pt/C (60 wt%), 2.5 mg cm⁻²; HCOOH flow rate: 0.3 mL min⁻¹, O₂ flow rate: 200 sccm, temperature: 20–80 °C, solution resistance: 19.15  ±  1.5 mΩ). g Power comparison of the Rh₁-Pd NSs at 30 °C against state-of-art catalysts at 80 °C. h TOFs of Rh₁-Pd NSs at different temperature. i Thermal protection mechanism of Rh₁-Pd NSs for DFAFCs.

Interestingly, when the operation temperature is raising from 35 °C to 80 °C, the power density dramatically decreases to a much lower level, especially at 60 and 80 °C, which is unable to maintain the effective operation of DFAFCs device. This is attributed to the “closed” control of the FAOR activity from thermo-responsive Rh₁-Pd NSs electrode, realizing the automatically thermal protection when the DFAFCs device is over-heating. Furthermore, turnover frequency (TOF) also suggests the abnormal temperature-activity relationship of thermo-responsive Rh₁-Pd NSs catalyst, which exhibits first increasing and then decreasing trend at any cell voltage (0.8–0.2 V, Fig. 6h). Assuming all metal atoms are active, the maximum TOF for Rh₁-Pd NSs is calculated as 71.86 s⁻¹ (0.2 V, 35 °C). Overall, above results prove the “smart” feature of thermosensitive Rh₁-Pd NSs catalyst, which exhibits competitive power density below the LCST and automatically switches into devitalization above the LCST (schematic in Fig. 6i).

Discussion

In summary, we demonstrate the thermosensitive Rh₁-Pd NSs as a highly efficient and “smart” electro-catalyst for FAOR and DFAFCs. The smartness is realized by using PNIPAM-NH₂ as a thermal responsive gate for the “open”/“closed” control of FAOR activity. The high FAOR activity and strong CO resistance are attributed to the synergistic effects between single-atom Rh sites and Pd NSs, which promote the direct formate pathway, increase the CO production barrier, and weaken the toxicant adsorption. For the DFAFCs, the thermosensitive Rh₁-Pd NSs electrode exhibits high power density below the LCST, while the power drops sharply once above the LCST, taking charge of the battery thermal protection. This work realizes the fabrication of a “smart” single-atom catalyst, which takes full advantage of its high intrinsic activity, meanwhile inspires its potentiality in intelligent fields.

Methods

Chemicals

Potassium tetrachloropalladite(II) (K₂PdCl₄, 98.0%) was purchased from Shanghai D&B Biological Sci-Tech Co., Ltd. Hexamethylenetetramine (HMTA, 99.0%) was supplied by Lingfeng Chemical Reagent Co., Ltd. Amine-terminated poly(N-isopropylacrylamide) (PNIPAM-NH₂, M_n = 2,500), Rhodium(III) acetylacetonate (Rh(acac)₃, 97.0%), benzyl ether (C₁₄H₁₄O, 98.0%) were purchased from Sigma-Aldrich trading Co., Ltd. Oleylamine (OAm, 80.0–90.0%) was purchased from Aladdin Industrial Co., Ltd. Commercial Pd/C (20.0 wt%) and Pt/C (60.0 wt%) were obtained from Johnson Matthey Corporation. Sulfuric acid (H₂SO₄, 95.0–98.0%) and formic acid (CH₂O₂, analytical reagent, 88.0%) were supplied by Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). ¹³C-labelled formic acid (H¹³COOH, 98%) was purchased from Wuhan Isotope Technology Co., Ltd. All the reagents were used without further purification.

Synthesis of thermosensitive Rh₁-Pd NSs

The thermosensitive Rh₁-Pd NSs were synthesized by decorating single-atom Rh sites on thermosensitive Pd NSs. First, the Pd NSs were synthesized by a one-pot hydrothermal method. Typically, 1 mL of K₂PdCl₄ solution (0.05 M), 2 mL of HMTA solution (0.05 M), 2 mL of PNIPAM-NH₂ solution (0.025 M) were added to 5 mL of deionized water and stirred at room temperature for 1 h. Then, the solution was transferred into a Teflon-lined stainless steel autoclave and kept at 140 ^oC for 6 h to obtain the thermosensitive Pd NSs. Subsequently, the products were cooling down, collected by centrifugation (12,578 g, 10 min), and washed with ethanol to remove the excessive reactants.

Then, 30 mg of above fresh-made Pd NSs, 6 mL of OAm, and 6 mL of benzyl ether were added in a glass vial. After ultrasound to form a uniform suspension, the mixture was heated to 250 °C under Ar atmosphere. Subsequently, 1 mL of Rh(acac)₃ (0.007 mmol, dissolved in benzyl ether) was slowly injected into the mixture, and the reaction was kept at 250 °C for 1 h. Then the product was cooled down to room temperature, centrifuged with toluene and ethanol (12,578 g, 10 min) for three times, and vacuum dried to obtain the thermosensitive Rh₁-Pd NSs.

Physicochemical characterizations

All the characterizations were performed at 25 ^oC if not specifically marked: AFM (Bruker Dimension Icon system); TEM, HRTEM, EDX spectra (JEOL JEM-2100F, 200 kV); AC-HAADF-STEM (Thermo Fisher Scientific Spectra 300 S/TEM), FT-IR spectrometer (Nicolet380, Thermo); XRD pattern (Model D/max-rC, using Cu K_α radiation source, λ = 1.5406 Å); ICP-AES (PerkinElmer, OPTIMA 7000DV); N₂ adsorption/desorption isotherms (Micromeritics ASAP 2050 system, 473 K); XAFS spectra (beamline 1W1B station of Beijing Synchrotron Radiation Facility). Data reduction, analysis and EXAFS fitting were performed with the Athena and Artemis software packages. Rh foil and Rh₂O₃ were used as references.

In-situ variable-temperature measurements

In-situ TEM experiments were acquired by using a DENS solutions DH30 double tilt in situ thermoelectric holder fitted to the optical TEM microscope. The temperature was monitored at 30, 40, 60 °C to observe the size/shape change of thermosensitive Rh₁-Pd NSs sample. The samples were held isothermally for over 2 min at each temperature before images were taken. For the removal of surface PNIPAM-NH₂, the sample was underwent an ultraviolet/ozone system (PSDPUV4T, Novascan) for 40 min.

In-situ DLS and zeta potential measurements were performed at the Malvern Nanosizer (Zetasizer Nano ZSP) with a temperature controller. The Rh₁-Pd NSs sample was first diluted using DI water and equilibrated at a given temperature (20–60 °C) for 10 min prior to data acquisition. The hydrodynamic diameter of the material was calculated using the Stokes−Einstein equation: d_h= k_bT/(3πŋ_oD), where k_b is the Boltzmann’s constant, T is the temperature (K), ŋ_o is the dynamic viscosity of the solvent (μ_water = 8.9× 10⁻⁴ Pa s⁻¹), and D is the bulk diffusion coefficient obtained from DLS measurements.

In-situ NAP-XPS measurements were performed on a Thermo VG Scientific ESCALAB 250 spectrometer with an Al K_α radiator, using a temperature-controllable laser heating device and thermocouple equipped. Before tests, the sample was fixed in the holder, pretreated using Ar etching (5 × 10⁻⁶ mbar), and transferred to the analysis chamber. Then the heat-treatment was implemented using a laser in a hydrogen atmosphere (0.1 mbar) for 10 min prior to data acquisition. All results were calibrated to C 1 s (284.6 eV).

Electrochemical measurements

All the electrochemical measurements were performed at 25 °C if not specifically marked. The CHI 760E CH analyzer with a three-electrode cell was employed for the measurements (reference electrode: a saturated calomel electrode (SCE); counter electrode: a graphite rod; working electrode: a glassy carbon (GC, 5 mm in diameter). The SCE was converted to reversible hydrogen electrode (RHE) according to the following equation, $\mathrm{E}\left(vs\mathrm{RHE}\right)=\mathrm{E}\left(vs\mathrm{SCE}\right)+0.242+0.0591\times \mathrm{pH}$.

For the preparation of 0.5 M H₂SO₄ electrolyte (pH = 0.26 ± 0.01), 13.6 mL of concentrated H₂SO₄ was diluted to a final volume of 0.5 L with deionized water. For the preparation of 0.5 M H₂SO₄  +  0.5 M HCOOH electrolyte (pH = 0.27 ± 0.01), 13.6 mL of concentrated H₂SO₄ and 10.9 mL of HCOOH was diluted to a final volume of 0.5 L with deionized water. The resulting electrolytes were freshly prepared before tests.

For the preparation of catalyst-coated working electrode, 2 mg of the catalyst was dispersing in a mixed solvent containing 0.5 mL of DI water, 0.5 mL of ethanol, and 5 μL of Nafion (5%). After ultrasound, 10 μL of above catalyst ink was casted on galssy carbon electrode and vacuum dried to acquire the working electrode. The mass loading of all electrocatalysts was 20 µg, including Rh₁-Pd NSs, Pd NSs, and Pd/C. The noble metal loading of Rh₁-Pd NSs, Pd NSs, and Pd/C was 20 µg, 20 µg, and 4 µg, respectively.

The CV curves were conducted between −0.242–0.958 V (vs. SCE) at a scan rate of 50 mV s⁻¹ in N₂ saturated 0.5 M H₂SO₄ solution. The ECSAs were calculated according to the following equation, $\mathrm{ECSA}=\mathrm{Q}/{\mathrm{Q}}^0$, where the Q is obtained by integrating the area of the hydrogen adsorption–desorption peak, Q⁰ is the areal charge needed for single layer oxygen atom adsorption (420 μC cm⁻²). The specific ECSAs were calculated according to the following equation: $\mathrm{specific\; ECSA}=\mathrm{Q}/\left({\mathrm{m}}_{\mathrm{Pd}}\times {\mathrm{Q}}^0\right)$, where m_Pd is the loading amount of Pd.

The FAOR measurements were conducted in N₂ saturated 0.5 M H₂SO₄ + 0.5 M HCOOH solution at a scanning rate of 50 mV s⁻¹. By normalizing the peak oxidation current to the loading amount of noble metals, the MA of each catalyst can be calculated for FAOR. Chronoamperometry tests were performed in 0.5 M H₂SO₄ + 0.5 M HCOOH solution at a constant potential of 0.25 V vs. RHE for 24 h. The CO-stripping curves were tested in a CO-saturated 0.5 M H₂SO₄ solution with two turns at a sweep rate of 100 mV s⁻¹. The pulsed voltammetry experiments were performed in N₂ saturated 0.5 M H₂SO₄ + 0.5 M HCOOH solution at a scanning rate of 50 mV s⁻¹. 0.85 V (vs. RHE) was chosen as the upper potential to ensure the complete oxidation of adsorbed CO species. The potential steps were continued for 1 s, both at the measuring potential and the stripping of the adsorbed CO molecules. The j_θ=0 and k_ads were calculated according to the reference⁴⁸.

Single cell tests

An electrode membrane assembly (MEA) was prepared by using a Rh₁-Pd NSs electrode with metal loading of 5 mg cm⁻² as the anode and a commercial Pt/C (60.0 wt%) electrode with metal loading of 2.5 mg cm⁻² as the cathode. The MEA was prepared by hot-pressing the anode and cathode electrode onto the two sides of Nafion 212 membrane (Du Pont, thickness: 50.8 μm, size: 5 × 5 cm², washed by deionized water once) at 7 MPa and 130 °C for 120 s. The single cells (25 cm²) were assembled in a fuel cells fixture (850 Fuel Cell Test System, Scribner Associates Inc., USA) for cell testing. 1.0 M H₂SO₄ and 1.0 M HCOOH solution were fed to the anode at a rate of 300 μL min⁻¹ and humidified oxygen (99.99%) gas was fed to the cathode at 200 sccm. A fuel cell test system (Arbin Instruments Company) was used to obtain the steady-state polarization and power density curves for Rh₁-Pd NSs from 0.95–0.20 V at different temperatures (20, 25, 30, 35, 40, 60 and 80 °C). The resistance was measured to be 19.15  ±  1.5 mΩ using Scribner 850 Fuel Cell Test System.

In-situ electro-chemical FT-IR measurements

The in situ FT-IR spectroscopy was performed on IS 50 FTIR spectrometer equipped with a liquid N₂-cooled system and MCT-A detector. The glassy carbon electrode loading with catalysts was used as the working electrode, which was assembled into a homemade spectra electrochemical cell, and pressed vertically on the CaF₂ window plate to form a thin liquid layer with a thickness of about 20 μm. The SCE was used as reference, which was introduced near the working electrode via a Luggin capillary. The graphite rod was serving as the counter electrode. Before the test, argon was purged into the electrolyte for over 30 min to eliminate the environmental CO₂. All spectra were shown in $\frac{\mathrm{\Delta }\mathrm{R}}{\mathrm{R}}=\left(E_s-E_R\right)/E_R$, with E_s and E_R representing the sample and reference spectra, respectively. The spectroscopy were collected at a resolution of 4 cm^-1 for 32 scans during the positive scanning of LSV in 0.5 M H₂SO₄ + 0.5 M H¹³COOH solution.

DFT calculation

Density functional theory computations were performed using the Vienna ab initio simulation package (VASP). The projector-augmented wave (PAW) method was served to describe the ion-electron interactions⁴⁹. The Perdew–Burke–Ernzerhof (PBE) functional was employed as a parametrization to the general gradient approximation (GGA) for the description of exchange-correlation energies⁵⁰. The DFT-D3 scheme was used to correct the long-range dispersion interactions⁵¹. The relevant kinetic cut-off energy was set to 500 eV in the plane-wave basis. In iterative solution of the Kohn-Sham equation, the energy criterion was established as 10⁻⁵eV. Taking consideration of the sizes of slab models, the Brillouin zone sampling in all calculations was performed with a 3 × 3 × 1 k-point mesh. To restrain the spurious mirror image interactions between adjacent unit cells, a vacuum space of 25 Å was employed along the z direction. The geometrical structures were permitted to be relaxed until the forces subject to each atom have declined to less than 0.03 eV/Å.

For evaluating the FAOR mechanism, the ΔG value of each elementary step was calculated as the following equation: $\mathrm{\Delta }G=\mathrm{\Delta }E+\mathrm{\Delta }{E}_{\mathrm{ZPE}}-T\mathrm{\Delta }S+\mathrm{\Delta }{G}_{\mathrm{U}}+\mathrm{\Delta }{G}_{\mathrm{p}\mathrm{H}}$. In this equation, ΔE represents the reaction energy, which can be directly obtained from DFT computations. ΔE_ZPE and ΔS represent the difference of zero–point energy and entropy, respectively, which can be derived from the computations on the vibrational frequencies and the standard thermodynamic data. ΔG_U represents the ΔeU, where U is the applied potential. ΔG_pH represents the correction on the pH in the electrolyte, which can be computed by $\mathrm{\Delta }{G}_{\mathrm{pH}}=-kT\ln 10\times \mathrm{pH}$, and pH was set to 0 here.

Author contributions

X.Q. conceived the idea and designed this study. K.S., S.Y., and Y.L. conducted the materials synthesis, structural characterizations of the prepared catalysts and electrochemical measurements. M.G. performed the HAADF-STEM imaging and assisted with the in situ experimental investigations. D.S. and Y.T. provided suggestions on the manuscript. K.S. and X.Q. wrote the paper. All authors discussed the results and edited the paper.

Peer review

Peer review information

Nature Communications thanks Ivano Alessandri and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The source data generated in this study are provided in the Source Data file. DFT models generated in this study have been deposited in Figshare under accession code 10.6084/m9.figshare.30031585⁵². 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 10.1038/s41467-025-66532-y.