The electric field plays a key role in enzyme catalysis, with a significant contribution to the local electrostatics and protein dynamics that allowed slow and difficult chemical reactions to occur rapidly at the active centers1,2,3. To mimic such field effect in natural systems, oriented external electric fields (OEEFs) have been proposed4,5,6,7,8 for modulating catalytic activity and selectivity in single-molecule reactions9,10,11. Specifically, electric fields can influence electronic interactions (vibrational frequencies, adsorption structures, and orbital orientation) in adsorbate/catalytic material interfaces and alter thermodynamic and kinetic properties of elementary reactions12. Moreover, with the rapid development of micro/nano-device technology, the electrical transport spectroscopy (ETS), and similar in-device electrocatalysis have emerged as a versatile platform for quantitative analysis at the micro/nano-scale13,14,15, where the OEEFs can be easily implemented, providing precisely controlled states for the nanoscale catalysts, particularly the two-dimensional (2D) atomic crystals16,17,18,19,20.

Single-atom catalysts (SACs) have arguably become one of the most active frontiers in heterogeneous catalysis for its high atom utilization efficiency21,22, featuring highly tunable active centers with well-defined coordination geometric/electronic structures similar to typical homogeneous catalysts or nanozymes23,24,25,26. In this regard, SACs offer an attractive model catalyst for mimicking the natural biocatalytic systems and exploring the corresponding catalytic pathways27,28,29. Despite enormous research focusing on the compositional design and synthesis of SACs with different coordination environments or local electronic structures30,31,32,33,34, there is a lack of an effective approach to dynamically modulate the activity of SACs through external intervention during the catalytic process. Meanwhile, it has been suggested that the protrusion-shaped SAs can produce internal local electric fields (tip effect) in specific directions (determined by the fixed electronic structure of SACs) and may greatly influence the reactant activation35,36 and ion distribution in electrolyte37. To this end, it is of great significance to explore the electrostatic effect on the system of SACs in a controlled and quantitative manner, which may offer a precise and universal model for the mechanistic evaluations, and the development of high-performance catalysts.

Here we demonstrate the quantitative manipulation of local catalytic sites of single-atom catalyst anchored on 2D atomic crystals via a precisely controlled external electric field. A series of model SACs supported on different 2D substrates have been systematically tested in the microcell, and their electrocatalytic performance can be dynamically tailored by OEEFs. Specifically, the Pt SAs on n-type MoS2 (Pt SAs-MoS2) demonstrate significant improvement in HER performance under positive OEEFs (ηHER of 20 mV @10 mA cm−2, Tafel slope of 51 mV dec−1, at Vg of +40 V), and the Co SAs on p-type WSe2 (Co SAs-WSe2) achieve a dramatically promoted activity for OER under negative OEEFs (ηOER of 139 mV at 10 mA cm−2, Tafel slope of 64 mV dec−1, at Vg of −40 V). In situ investigations, electrokinetic studies and density functional theory (DFT) calculations reveal an “onsite electrostatic polarization” mechanism, where the vertical electric field effectively modulates the charge distribution at the SA sites, polarizing the frontier orbitals of metal atoms/adsorbates/intermediates and significantly modifying the activation energies and/or reaction pathways during the catalytic processes. This unique onsite modulation mode can dynamically promote the electrocatalytic performance of SACs, leading to record-high activities, and serves as a highly tunable physical model for the enzyme-mimic artificial modulation and in-depth understanding of general heterogeneous/electro-catalysis.

Results and discussion

Micro-device construction and characterization of model SACs on 2D atomic crystals

Electrolyte gating occurs spontaneously to make semiconductors more conductive during the electrochemical process38. As shown in Fig. 1a, Pt SAs was first deposited on a prefabricated micro-device of n-type MoS2 through UV light-assisted reduction. For the systematic study, a series of model SACs (combination of Pt/Co/Pd and n-type MoS2/p-type WSe2/metallic-graphene) were also prepared via similar approaches. The detailed synthesis steps of 2D SACs and fabrication of micro-device had been shown in “Methods” and Supplementary Information. Aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), which has been widely adopted in microstructure characterization39, demonstrated that Pt SAs were successfully introduced into the 2H-MoS2 lattice without disturbing its crystalline structure (Fig. 1b–e and Supplementary Figs. 14). The Mo:S atomic ratios (Supplementary Table 2) are determined to be 1.00:1.98 and 1.00:1.96 in bulk MoS2 and Pt SAs-MoS2, respectively, further confirming that the near-perfect lattice (few defects) was maintained. The area density of Pt SAs on mechanically exfoliated MoS2 is determined to be 0.42 atom·nm−2 (~4.84 wt%) via statistical analysis of the STEM images (Supplementary Table 1). Raman and photoluminescence spectroscopy (Fig. 1f and Supplementary Fig. 8) revealed the doping effect of Pt SAs on 2H-MoS2 nanosheets40,41,42,43, which also leads to a more positive threshold voltage in the electrical FET transfer curves (Fig. 1g and Supplementary Fig. 7) of n-type MoS2 micro-device modified by Pt SAs44,45,46.

Fig. 1: Synthesis and characterizations of Pt SAs-MoS2.
figure 1

a Proposed reaction mechanism for the preparation of Pt SAs-MoS2 micro-device. b, c HAADF-STEM images of mechanically exfoliated monolayer MoS2. Specifically, c is the magnified image of yellow dashed area in b. d, e HAADF-STEM images of Pt SAs-MoS2. Specifically, e is the magnified image of yellow dashed area in d, with red circles highlighting the Pt SAs distributed in the 2H-MoS2 lattice. f Raman spectroscopy of mechanically exfoliated monolayer MoS2 (black) and Pt SAs-MoS2 (red). g Transfer characteristics of FET devices fabricated from mechanically exfoliated monolayer MoS2 (black) and Pt SAs-MoS2 (red). h Peak deconvolution of Pt 4f XPS spectra for commercial Pt/C, Pt NPs-MoS2, and Pt SAs-MoS2. i Normalized Pt L3-edge XANES spectra of Pt foil, Pt SAs-MoS2, and PtO2. j First shell fitting of EXAFS spectra for Pt foil, Pt SAs-MoS2, and PtO2.

The chemical configuration and local coordination environment of Pt SAs-MoS2 were then investigated through the combination of X-ray photoelectron spectroscopy (XPS) and X-ray absorption spectroscopy (XAS). Compared to that of commercial Pt/C and Pt NPs-MoS2 located at around 71.0 eV (for Pt0), the Pt 4f7/2 XPS peak (Fig. 1h) of Pt SAs-MoS2 significantly moved towards higher binding energies (approximately 72.5-73.0 eV), with obvious alteration in peak shape. Moreover, the results of peak deconvolution further demonstrated that the Pt species in Pt SAs-MoS2 were partially oxidized to the valence state of 0-+4 (Ptδ+), mainly composed of Pt2+ and Pt4+ (72.3 and 73.4 eV, Supplementary Fig. 11), which should be attributed to the electronic interactions between Pt SAs and MoS2 nanosheets47. Moreover, in X-ray absorption near-edge spectroscopy (XANES) the Pt L3-edge analysis (Fig. 1i) showed that the white line intensity for Pt SAs-MoS2 was higher than that for Pt foil, indicating the oxidized Pt species were not in the metallic state47,48,49. As shown by EXAFS in R space (Fig. 1j), Pt SAs-MoS2 exhibited a peak at approximately 2.3 Å from the Pt-S shell with a coordination number of 4.0 (Supplementary Table 3). Besides, no typical peaks for Pt-Pt bond at longer distances (>2.7 Å) were detected, revealing the atomic dispersion of Pt atoms on MoS2 nanosheets. More importantly, the theoretical distance of Pt-S bond (2.291 Å, Supplementary Table 11) from the optimized model of (M-top)-Pt SAs-MoS2 (more details in Supplementary Fig. 88) was also consistent with the EXAFS results, further confirming the Mo-atop site of Pt SAs anchored on MoS2.

OEEF modulation in SA-catalyzed HER

On-chip electrochemical and in situ transport investigations were simultaneously conducted in the four-electrode microcell (Fig. 2a and Supplementary Figs. 2325), and the vertical electric fields can be applied through back-gate electrode50,51. To set a background, modulation of electrochemical behaviors under electric fields was first investigated on the pristine MoS2 with different thicknesses (Supplementary Fig. 27). As shown in Fig. 2b and Supplementary Figs. 2830, the HER performance of n-type MoS2 with less layers can be tailored more efficiently by OEEFs. This thickness-selectivity on pristine 2D TMDs20,38 originates from the Debye screening effect52,53, which provides a background behavior of 2D support for this study. On this basis, HER tests (with in situ transport measurement of Ids) were performed on Pt SAs-MoS2 with few-layer MoS2 (<5 nm) as the support. To quantify the modulation efficiency of OEEFs on electrocatalysis, we define a “tunable potential range (ΔηE)” (inset of Fig. 2b), with Vg (275 nm SiO2 as the dielectric layer) varying from −40 V to +40 V at 50 mA cm−2. Figure 2c shows that Pt SAs-MoS2 (red/blue solid lines) exhibits considerably higher HER performance over pristine MoS2 (gray dashed lines), with a further and significant promotion (ΔηE = 52 mV) under vertically applied positive electric field. Meanwhile, Tafel slopes (Fig. 2d) of Pt SAs-MoS2 also decline from 117 to 65 mV dec−1 with Vg applied from −60 V to +60 V, which is much smaller than the pristine 2D support without SA decorations (from 161 to 105 mV dec−1, consistent with the pristine 2H-MoS2 for HER54,55,56) and gradually approaches Pt-based electrocatalysts reported previously57,58,59. Note that negligible back-gate leakage current (Supplementary Figs. 86 and 87) was observed, indicating charge accumulated at the MoS2–SiO2 interface was kept stable during the top-gate electrochemical process, and no extra energy input is required by the introduction of an electrostatic field via the back-gate circuit. Such continuous promotion at the high-activity region demonstrates the potential of an externally applied electric field to further push the already effective SACs toward superior performance, presumably through alternation in catalytic kinetics.

Fig. 2: In situ electronic/electrochemical investigations on Pt-based SACs for HER under oriented external electric field (OEEF) regulation.
figure 2

a Schematic diagram of an on-chip four-electrode microcell for electronic/electrochemical measurements, OEEFs were introduced via vertically applied back gate voltage (Vg) through a dielectric layer. b Layer dependence on the HER activity modulation (tunable potential range, ΔηE) in pristine 2D MoS2 (n-type semiconductor). Inset: HER polarization curves and in situ drain-source current of monolayer MoS2 under OEEF regulation, with Vds = 100 mV and Vg from −40 V to +40 V. c, d HER polarization curves (c) and Tafel curves (d) of Pt SAs-MoS2 in response to OEEFs. e, f HER polarization curves of Pt SAs-WSe2 (e) and Pt SAs-graphene (f) in response to OEEFs. The two gray dashed lines in cf represent the tunable HER activity of pristine 2D substrates under OEEF regulation. The color bar in cf shows the direction and scope of Vg. gi Correlations between electrochemical current density (jHER) and in situ sheet conductance (Gs) of Pt SAs-MoS2 (g), Pt SAs-WSe2 (h), Pt SAs-graphene (i), and pristine 2D substrates under OEEF regulation. The arrows in gi indicate the OEEF-modulation directions of HER performance optimization. Note that in gi the unit of abscissa is μS sq−1 for Pt SAs-MoS2 and Pt SAs-WSe2, while mS sq−1 for Pt SAs-graphene, and jHER in gi were extracted at the potentials (vs. RHE) of: −0.15 V for Pt SAs-MoS2, −0.2 V for MoS2, −0.2 V for Pt SAs-WSe2, −0.4 V for WSe2, −0.2 V for Pt SAs-graphene, −0.4 V for graphene, to obtain values of similar level for better comparison.

To probe the mechanism of this electric-field modulation on SACs, we extracted in situ sheet conductance (Gs derived from Ids at different OEEFs, Supplementary Fig. 33) in Pt SAs supported on n-type MoS2 and studied its correlation with HER current density (jHER), as shown in Fig. 2g. Similar ETS signals have been proved effective to identify the operando charge carrier states in pristine semiconductor catalysts38,60, and are presumably informative for the more comprehensive single catalytic sites. First, a positive and close-to-linear jHER-Gs correlation (black dashed line in Fig. 2g) was established in pristine MoS2. In this case, carrier (electron) injection into n-type semiconductors can lead to the gradually declined overpotential originated from inner resistance (contact and/or sheet resistance of MoS2), and the rise of their Fermi level reduces the corresponding energy barrier in HER (i.e., change of H adsorption ability), thus promoting the HER kinetics17,18,61. In contrast, Pt SAs-MoS2 (red solid line in Fig. 2g) exhibited a distinct near-exponential jHER-Gs correlation, indicating a different modulation mechanism in single-atom electrocatalysis, where field-induced carrier injection (i.e., the gating effect) in 2D supports is not the major cause for HER modulation in SACs. Considering the structural and catalytic features of the SACs, we propose an alternative theory that the OEEF modulation of the catalytic performance could originate its direct influence on the SA catalytic sites. To further verify this hypothesis, OEEF-induced HER modulation was investigated in Pt SAs anchored on p-type WSe2 (Fig. 2h) and metallic-graphene (Fig. 2i), where electric gating will lead to different conductivity trends in the 2D supports. Indeed, pristine p-type WSe2 shows an opposite direction in both Gs-Vg and jHER-Gs relationships compared to n-type MoS2 (note that Gs increases at different Vg directions for n- and p-type TMDs, as indicated by the arrows in Fig. 2g, h), as a result of the same field modulation mechanism through charge carrier injection. In contrast, a negative jHER-Gs correlation (red solid line in Fig. 2h) was observed in Pt SAs-WSe2 (same as Pt SAs-MoS2), which confirms that SACs follow an alternative OEEF modulation mechanism aside from the carrier injection in 2D supports. More obviously, negligible electric-field response (ΔηE < 10 mV, Supplementary Fig. 43) was observed in pristine graphene due to the relatively smaller modification in its already large carrier concentrations under different OEEFs, whereas the HER performance of Pt SAs-graphene was again significantly enhanced by positive OEEFs (ΔηE = 84 mV, Fig. 2f), with almost no fluctuation in Gs (red solid line in Fig. 2i). Overall, it can be concluded that for HER performance under OEEFs, 2D supports follow the typical mechanism of carrier-concentration modulation in semiconductors that leads to different trends in n-type, p-type, and metallic supports, while the electrocatalytic performance promotion in Pt SAC systems was unanimously triggered by positive OEEFs regardless of the different behaviors in 2D supports. These results provide strong evidence that the OEEF modulation in SACs is originated from a distinctive mechanism, probably through the direct modification on the single-atom catalytic sites.

To rationalize the unique reaction pathway for the OEEF-induced catalytic modulation in Pt SACs, theoretical calculations were conducted to investigate their physicochemical characteristics under OEEFs. A 4×4 supercell of monolayer 2H-MoS2 with 3.1% sulfur vacancies, which are usually regarded as the main active sites in the basal plane of 2H-MoS2 for HER62,63,64, was set up for HER studies (see more details in “Methods”). The hydrogen adsorption free energy (ΔGH), which is widely recognized as a descriptor for HER activity65,66, was investigated by direct introduction of electric field gradients (Efield) into the supercell (Fig. 3a). For Pt SAs, the M-top site binding to S atom, which was confirmed by XAS results and STEM observations, was constructed as the optimal DFT model for subsequent theoretical analysis. As shown in Fig. 3b, ΔGH of (M-top)-Pt SAs-MoS2 becomes noticeably more negative (red solid line) when placed in a vertical electric field from −0.4 V/Å to +0.4 V/Å, whereas ΔGH of 2H-MoS2 slightly moves towards the zero point (black solid line) in response to Efield. Accordingly, the partial density of states (PDOS) projected onto the H 1s and the \({{{{{\rm{Pt}}}}}}\,5{{{{{{\rm{d}}}}}}}_{{{{{{{\rm{z}}}}}}}^{2}}\) orbitals show consistent response to OEEFs, and the energy gap between the centers of H 1s-\({{{{{\rm{Pt}}}}}}\,5{{{{{{\rm{d}}}}}}}_{{{{{{{\rm{z}}}}}}}^{2}}\) bonding states and antibonding states become smaller when applying positive OEEFs (Supplementary Fig. 92). In this case, less energy is required to form the hybridized orbitals67,68,69 and thus higher *H adsorption energy on Pt SAs-MoS2 (a more compact Pt-H bond) is observed. Solvation model of (M-top)-Pt SAs-MoS2 was further constructed to investigate the possible effect from solvent (water) molecules at the electrochemical interface (Supplementary Fig. 90). The values of hydrogen adsorption energy in presence of a water molecule exhibited a negative correlation with the introduced electric field gradients, which is similar to the ΔGH-Efield relationship observed in the gas-phase model (Supplementary Fig. 91). More importantly, the trend of OEEF-modulated charge distribution on the adsorbed *H (QH, Supplementary Table 15) is also close to the gas-phase model.

Fig. 3: “Onsite electrostatic polarization” in the Pt SAs-MoS2 catalyzed HER.
figure 3

a Schematic diagram of the electric-field introduction into Pt SAs-MoS2. b Hydrogen adsorption energy (ΔGH) in the models of 2H-MoS2 (black) and (M-top)-Pt SAs-MoS2 (red), with Efield from −0.4 V/Å to +0.4 V/Å. c Charge analysis of (M-top)-Pt SAs-MoS2 under OEEF regulation, where *H is adsorbed on Pt atom. Top: charge densities with interface values of 0.15 e/Bohr3. The zoomed-in area shows the charge distribution at the first adsorbed *H site. Bottom: deformation charge densities with interface values of 0.001 e/Bohr3. Red and blue areas denote electron accumulation and depletion, respectively. Efield from left to right is −0.4 V/Å, 0, and +0.4 V/Å. d Schematic diagram of the “onsite electrostatic polarization” mechanism on Pt SAs-MoS2 for HER. During the Heyrovsky reaction, H+ in electrolyte tends to move towards the first adsorbed *H site, where electron accumulation was triggered by positive OEEFs. e The positive-OEEF-induced breakthrough at HER volcano apex (jHERGH) in Pt SAs-MoS2. Inset: HER performance of pristine MoS2 associated with the OEEF-regulated carrier concentrations (consistent with the jHER-Gs correlation observed in Fig. 2g), rather than the slight change in ΔGH (black solid line in b).

The declining of ΔGH on Pt SAs-MoS2 is actually moving away from the volcano apex (near the zero point) that is favorable for HER70, thus inconsistent with the observed OEEF modulation on HER activities. These results indicate that the OEEF-modulation of HER in Pt SACs essentially went through a different mechanism, which is also indicated by comparing the Tafel characteristics observed in Pt SAs-MoS2 and MoS2 (Fig. 2d). In principle, a Tafel slope around 120 mV dec−1 or higher suggests that the first H adsorption (the Volmer reaction) limits the overall reaction rate in HER71. Therefore, the mediocre values of the Tafel slope (over the whole range of Vg) observed in MoS2 (105–161 mV dec−1) signifies that the HER optimization should be attributed to an increase in carrier concentrations (rather than a slight change in ΔGH) under electric field regulation17,18,19. Distinctively, the smaller Tafel slope of Pt-MoS2 (65–117 mV dec−1) demonstrates that the response in HER activities to OEEFs in the single-atom system goes beyond the limitation of the Volmer step and presumably originates from the modulation on the overall kinetics. It is also interesting to notice that Pt SAs-MoS2 exhibits a unique HER behavior compared to those of Pt NPs-MoS2 (Supplementary Fig. 51) and Pt microelectrode (Supplementary Fig. 53), with lower onset potential but slightly larger Tafel slope (Supplementary Table 9). This can be rationalized by that Pt SAs-MoS2 follows the Volmer-Heyrovsky mechanism71 with a theoretical Tafel slope of 40 mV dec−1 (rate-determining step is the Heyrovsky reaction), while other Pt species (with relative high Pt content) are usually limited by the Tafel reaction with theoretical Tafel slope of 30 mV dec−1. On these bases, OEEFs presumably act on the Heyrovsky reaction to further influence the electro-kinetics.

The charge analysis (Supplementary Figs. 93 and 94 and Supplementary Table 14) of H adsorption on 2H-MoS2 and Pt SAs-MoS2 was then performed to reveal the in-depth electrocatalytic regulation mechanism induced by OEEFs. As is shown in Fig. 3c, the outward electron density around the first adsorbed *H site exhibits a significant enhancement under the positive Efield, with increased QH (top) and less electron depletion (bottom). For comparison, injection of excess electrons equivalent to Vg from −60 V to 60 V (Supplementary Table 16) was also performed, which generated similar results (Supplementary Figs. 96 and 97). Note that such modulation was based on charge “polarization” directly on the *H site, rather than through the modulation of carrier concentrations as in the electric field-induced charge injections, which explains the universal trend in jHER-Vg correlations observed in Pt SACs anchored on different types of 2D supports (p-, n-, and metallic), as indicated by red arrows in Fig. 2g–i. This electrostatic polarization certainly benefits the approaching of a second proton (H+) to the Pt-H intermediate and thus significantly promotes the kinetics of the Heyrovsky reaction in HER, as summarized in Fig. 3d. Overall, positive OEEFs effectively lower the energy barrier through outward orbital polarization at the *H site to promote the Heyrovsky step, pushes the desorption of H2 molecules originally restricted by the strong ΔGH, and thus breaking through the limiting volcano plots of HER to achieve a superior activity, as illustrated in Fig. 3e. This OEEFs-triggered mechanism appears specific in SAC systems (regardless of the modulation on carrier concentrations that existed in 2D substrates), and can be understood as “onsite electrostatic polarization” in single-atom electrocatalysis, boosting SACs kinetics via an artificial enzyme-mimicking approach to achieve superior performance.

OEEF modulation of SACs for OER through “onsite electrostatic polarization”

The “onsite electrostatic polarization” mechanism discovered in SACs was further extended to the oxidation reaction. Co SAs-WSe2 was used as a model catalyst as it combines an OER favorable SAC and a p-type substrate. As confirmed by HAADF-STEM images (Supplementary Figs. 1214), Co SAs were successfully introduced into the mechanically exfoliated WSe2, with an area density of 0.41 atom·nm−2 (~0.78 wt%, Supplementary Table 4). The Co 2p XPS signals (Supplementary Fig. 20) also indicated that the oxidized Co species (Coδ+) in Co SAs-WSe2 is different from Co nanoparticles. As expected, Co SAs-WSe2 exhibits dramatically improved OER activity under the negative OEEF modulation. Both onset potential and Tafel slope of OER were optimized when negative Vg was applied on Co SAs-WSe2 (Fig. 4a, d), which was also observed in Co SAs-MoS2 (Fig. 4b) and Co SAs-graphene (Fig. 4c). More importantly, unique jOER-Gs correlations (Fig. 4e–g) were observed in Co SACs, distinct from the close-to-linear behaviors in pristine 2D substrates. For all substrates (p-WSe2, n-MoS2, and metallic-graphene), the OEEF modulation directions were identical to the alteration of carrier concentrations in response to the applied Vg (black arrows in Fig. 4e–g). As a result, p-type WSe2 and n-type MoS2 showed opposite modulation directions (indicated by black arrows in Fig. 4e, f). For all SACs, the trend of OER modulation is unanimously correlated to the OEEF direction regardless of the different trends in Gs (red arrows in Fig. 4e–g), which signifies the presence of “onsite electrostatic polarization” in anodic electrocatalysis.

Fig. 4: OEEF-induced “onsite electrostatic polarization” in the Co SAs-WSe2 catalyzed OER.
figure 4

ac OER polarization curves of Co SAs-WSe2 (a), Co SAs-MoS2 (b), and Co SAs-graphene (c) in response to OEEFs. d Tafel curves of Co SAs-WSe2 in response to OEEFs. The regions between two gray dashed lines in ad represent the tunable OER activity of pristine 2D substrates under OEEF regulation. The color bar in ad shows the direction and scope of Vg. eg Correlations between electrochemical current density (jOER) and in situ sheet conductance (Gs) of Co SAs-WSe2 (e), Co SAs-MoS2 (f), Co SAs-graphene (g), and pristine 2D substrates under OEEF regulation. The arrows in eg indicate the OEEF-modulation directions of performance optimization. Note that in eg the unit of abscissa is μS sq−1 for Co SAs-WSe2 and Co SAs-MoS2, while mS sq−1 for Co SAs-graphene, and the values of jOER in eg were obtained at the potentials (vs. RHE) of: 1.45 V for Co SAs-WSe2, 1.8 V for WSe2, 1.6 V for Co SAs-MoS2, 1.9 V for MoS2, 1.6 V for Co SAs-graphene, 1.8 V for graphene. h Adsorption free energies diagram of (M-top)-Co SAs-WSe2 for OER under OEEF regulation, where *OH, *O, and *OOH are successively absorbed on Co atom. i Comparisons of ΔGO (solid lines) and ΔG3 (dashed lines) in 2H-WSe2 (black) and (M-top)-Co SAs-WSe2 (red), with Efield from −0.4 V/Å to +0.4 V/Å. j Charge analysis of (M-top)-Co SAs-WSe2 under OEEF regulation, where *O is adsorbed on Co atom. Top: charge densities with interface values of 0.15 e/Bohr3. The zoomed-in area shows the charge distribution at the adsorbed *O site. Bottom: deformation charge densities with interface values of 0.01 e/Bohr3. Red and blue areas denote electron accumulation and depletion, respectively. Efield from left to right is −0.4 V/Å, 0, and +0.4 V/Å. k Schematic diagram of the “onsite electrostatic polarization” mechanism on Co SAs-WSe2 for OER. OH in electrolyte tends to move towards the adsorbed *O site, where electron depletion was triggered by negative OEEFs.

All possible configurations of Co atom anchored on 2H-WSe2 were considered and three DFT models were constructed (Supplementary Fig. 98). In line with the STEM observations, the two optimal models with lower formation energies (M-top and X-sub) were ultimately applied for subsequent theoretical analysis by introducing different Efield, with successive *OH, *O, and *OOH adsorptions. Figure 4h shows the adsorption free energies (ΔGOH, ΔGO, ΔGOOH) of three oxygen intermediates along the overall OER pathway on (M-top)-Co SAs-WSe2. Specifically, the adsorption free energies can be effectively tailored by the external electric field, accompanied by obvious changes in free energy differences (ΔG1, ΔG2, ΔG3, ΔG4). And the maximum of free energy difference (ΔG3, see “Methods” and Supplementary Tables 19 and 20) indicated that the conversion from *O to *OOH was the potential-determining step (pds) in Co SAs-WSe2 for OER. As shown in Fig. 4i, ΔGO of (M-top)-Co SAs-WSe2 moves towards the zero point when Efield goes negatively (red solid line), whereas WSe2 has no obvious change in ΔGO (black solid line). Meanwhile, ΔG3 of Co SAs-WSe2 appears to be significantly reduced under the negative electric fields, which rationalizes the boosted OER performance of Co SAs-WSe2 under negative Efield. The PDOS of O 2pz and Co \(3{{{{{{\rm{d}}}}}}}_{{{{{{{\rm{z}}}}}}}^{2}}\) orbitals (Supplementary Fig. 102) also confirms a weaker Co-O interaction under negative OEEFs (consistent with the results of ΔGO). More importantly, lower charge density (Fig. 4j) around the adsorbed *O site (Supplementary Tables 21 and 23) was observed in Co SAs-WSe2 under negative OEEFs, which presumably promotes the consequent approaching of a OH to *O and thus reduces ΔG3 for the formation of a O-O bond, as shown in Fig. 4k. These results verified the presence of “onsite electrostatic polarization” mechanism in SA-catalyzed OER. Notably, the solvation model of (M-top)-Co SAs-WSe2 (Supplementary Fig. 100) exhibited more positive free energies (Supplementary Table 19) of *OH, *O, *OOH after introducing OEEFs, which is reasonable as the stabilization effect from hydrogen bonding72,73 is weakened by OEEFs (regardless of direction). In addition, the potential-determining step (ΔG3, from *O to *OOH) remained unchanged and the trend of OEEF-modulated charge density on the adsorbed O* (QO, Supplementary Table 22) appeared to be identical to that of the gas-phase model, demonstrating that the solvent effects did not affect the conclusion of OEEF-induced “onsite electrostatic polarization” mechanism proposed here. It is also worth mentioning that the asymmetrical oxygen species (*OH and *OOH) exhibit reorientation under the external electric field (Supplementary Figs. 104107), which is projected on the irregular (non-linear) results of free energies (Supplementary Tables 19 and 20) in response to OEEFs. In contrast to a single sigma bond (Co-O bond in OER and Pt-H bond in HER) that is parallel to the electric field, the alteration of orientation angle leads to changes in spin states or valence states of Co SAs, contributes to the more sophisticated modulation of orbital interaction in oxygen electrocatalysis74,75,76 from an alternative perspective. Overall, we have proposed and verified the “onsite electrostatic polarization” mechanism in single-atom electrocatalysis (in both cathodic and anodic reactions) that principally mimics the function of electrostatic polarization in natural enzymes, thus achieving the superior electrocatalytic performance.

Superior electrocatalytic performance of OEEF-modulated 2D SACs

To verify the general OEEF modulation in single-atom electrocatalysis, a series of SACs with different SA species and 2D substrates were systematically tested in the micro-platform for HER and OER, as summarized in Fig. 5a, e. Particularly, Pt SAs-MoS2 (Vg = +40 V) exhibits excellent HER performance in acidic condition (0.5 M H2SO4) with the overpotential of 20 mV at 10 mA cm−2 (Fig. 5i and Supplementary Table 9) and an ultra-high mass activity of 1171.1 A \({{{{{{\rm{mg}}}}}}}_{{{{{{\rm{Pt}}}}}}}^{-1}\) (ηHER = −0.05 V), which is superior to several representative Pt-based electrocatalysts in recent works (see Fig. 5k and Supplementary Tables 5 and 6)32,33,37,47,57,59,77,78,79,80,81. Moreover, due to the relatively low atomic loading in Co SACs (quantitatively derived from STEM characterizations), Co SAs-WSe2 (Vg = −40 V) demonstrates a remarkable OER performance in alkaline condition (0.1 M KOH), with the overpotential of 139 mV at 10 mA cm−2 (lowest among other electrocatalysts in previous studies82,83,84,85,86,87,88,89,90,91,92, Fig. 5j and Supplementary Table 10) and record-high mass activity of 19,024.2 A \({{{{{{\rm{mg}}}}}}}_{{{{{{\rm{Co}}}}}}}^{-1}\) (ηOER = 0.2 V, Fig. 5k and Supplementary Tables 7 and 8). In addition, to verify the correspondence between on-chip electrocatalytic measurements and the actual chemical production by 2D SACs under OEEF regulation, quantitative product analysis was conducted with a custom-designed on-chip electrochemical reactor (Supplementary Fig. 26). HER and OER electrolysis were performed on SACs supported by centimeter-size graphene (Pt SAs-graphene for HER and Co SAs-graphene for OER), with application of varying Vg. In a 4-h potentiostatic electrolysis, the gas products (H2 and O2) were collected and quantified by a gas chromatograph (Supplementary Figs. 8285). As shown in Fig. 5d, h, the chemical production rates clearly demonstrated the significant modulation by the external electric fields, further confirming the OEEF-modulated electrocatalytic process observed in the micro-platform. Therefore, the superior performance via “onsite electrostatic polarization” mechanism exhibited in OEEF-modulated SACs offers an effective and universal strategy to tailor the active sites, thus achieving the enzyme-mimic function and high efficiency in SAC-based systems and corresponding on-chip reactors.

Fig. 5: Electrocatalytic performance and stability of model SACs through “onsite electrostatic polarization” modulation.
figure 5

a HER polarization curves of 2D SACs (at optimal OEEFs) and the control catalysts. b HER polarization curves of Pt SAs-MoS2 (Vg = +40 V) before (black) and after (red) 100 CV cycles, in the potential range of −0.1 V to 0.1 V (vs. RHE). Inset (with identical abscissa) depicts the recorded leakage current. c Electrochemical current density (jHER) of Pt SAs-MoS2 at E = −0.1 V (vs. RHE) recorded in a 100-cycle HER test, with Vg switching between −40 V and +40 V every ten cycles. d Chronoamperometric curves in a 4-h potentiostatic HER test of pristine graphene and Pt SAs-graphene under OEEF regulation. The formation rates of H2 are added alone each curve. Note that the ordinates in ad all correspond to HER current density (jHER). e OER polarization curves of 2D SACs (at optimal OEEFs) and the control catalysts. f OER polarization curves of Co SAs-WSe2 (Vg = −40 V) before (black) and after (red) 100 CV cycles, in the potential range of 1.1 V to 1.5 V (vs. RHE). Inset (with identical abscissa) depicts the recorded leakage current. g Electrochemical current density (jOER) of Co SAs-WSe2 at E = 1.45 V (vs. RHE) recorded in a 100-cycle OER test, with Vg switching between −40 V and +40 V every ten cycles. h Chronoamperometric curves in a 4-h potentiostatic OER test of pristine graphene and Co SAs-graphene under OEEF regulation. The formation rates of O2 are added alone each curve. Note that the ordinates in e-h all correspond to OER current density (jOER). i Summarized HER performance of Pt SAs-MoS2 (with varying Vg) and previously reported electrocatalysts. j Summarized OER performance of Co SAs-WSe2 (with varying Vg) and previously reported electrocatalysts. k Mass activity of 2D SACs (at optimal OEEFs) in this work and several representative up-to-date catalysts. The green bars for HER at ηHER = −0.05 V and the purple bars for OER at ηOER = 0.2 V.

Long-term cyclic voltammetry (CV) scans were further performed on the model SACs under the constant and optimal electric field. During a 100-cycle test, no obvious current decay was observed in Pt SAs-MoS2 (Vg = +40 V) for HER (Fig. 5b) and Co SAs-WSe2 (Vg = −40 V) for OER (Fig. 5f), manifesting the high stability of SACs constructed in this work, which is also consistent with the spectroscopic results and HAADF-STEM observations (Supplementary Figs. 4 and 15). Although an irreversible oxidation peak was observed in pristine WSe2 at around 1.4 V (vs. RHE) in the first few scans (Supplementary Fig. 81), which should be attributed to the inherent electrochemistry of TMDs93,94,95, the introduction of Co SAs effectively shifted the OER onset potential to ~1.3 V, and led to significantly enhanced stability. PL and XPS results (Supplementary Figs. 1922) showed that there was no significant change in the structure and chemical state of Co SAs-WSe2, despite a possible trace level conversion of WSe2 to WO3. The chemical mapping from energy-dispersive X-ray spectroscopy (STEM-EDS, Supplementary Fig. 16) further verified the predominant of selenides localized on the surface of Co SAs-WSe2 after OER stability test, validating the theoretical models employed in DFT calculations. Moreover, we tested the durability of altering external electric fields acting on Pt SAs-MoS2 and Co SAs-WSe2, by running linear sweep voltammetry (LSV) scans with dynamically switched Vg. Moreover, accompanied by Vg switching among 0 V, +40 V, and −40 V every ten cycles, the electrochemical current responds rapidly to the sudden change of Vg and keeps relatively steady in ten consecutive cycles (Fig. 5c, g), suggesting that the facile and quantitative strategy for introducing OEEFs, is non-destructive to electrocatalysts.

In summary, we have demonstrated the experimental evidence that the electrocatalytic performance of 2D SACs can be continuously and quantitatively modulated to a record-high level via an externally applied oriented electric field. Combined with theoretical investigations, such OEEF-induced performance enhancement is revealed to distinctively originate from “onsite electrostatic polarization” in SAC systems, where electric fields have significant influences on the charge redistribution (especially at the active sites) of SACs, leading to the optimization of chemical bonding between SA and key adsorbates, the spatial distribution of outward charges, and the reaction kinetics of the rate determine steps. This precise and easy-to-operate modulation mode reveals a universal mechanism in single-atom electrocatalysis, which applies to a broader range of electrochemical reactions. The combination of homogeneously distributed SACs with atomically precise geometric/electronic structure and the dynamic modulation through OEEFs, also presents a viable and promising functional model toward the development of artificial enzyme-mimicking systems, which holds great potential for future applications.


Preparation of Pt SAs-MoS2

2D MoS2 nanosheets mechanically exfoliated from bulk MoS2 (n-type, SPI Supplies) were transferred to the silicon wafer using magic tape (Scotch). The silicon wafer was immersed in 30 mL deionized water and ethylene glycol (≥99%, Sigma-Aldrich) mixed solution (9:1 volume ratio) followed by 1.46 mg K2PtCl6 (>99.95%, Admas-beta) was added into the solution with stirring, the concentration of K2PtCl6 was 0.1 mmol L−1. Then the silicon wafer immersed in the mixed solution was irradiated under the 365 nm ultraviolet light at room temperature for 2 h, which was taken out to heat at 150 °C for 15 min. Finally, Pt SAs-MoS2 samples were obtained on the silicon wafer, washed by deionized water three times, and dried by N2 purging. Principally, UV-light radiation induces the formation of EG radicals to reduce metal ions into single atoms96, which appeared to be more efficient to attach the single atoms on the exfoliated pristine 2D samples (with few defect sites), see more details in Supplementary Figs. 5 and 6.

Preparation of Co SAs-WSe2

2D WSe2 nanosheets mechanically exfoliated from bulk WSe2 (p-type, 2D Semiconductors) were transferred to the silicon wafer using magic tape. The silicon wafer was immersed in 30 mL deionized water and ethylene glycol (9:1 volume ratio) mixed solution, followed by 3.57 mg CoCl2·6H2O (AR, Aladdin) was added into the solution with stirring, the concentration of CoCl2 was 0.5 mmol L−1. Then the silicon wafer immersed in the mixed solution was irradiated under the 365 nm ultraviolet light at room temperature for 3 h, which was taken out to heat at 150 °C for 15 min. Finally, Co SAs-WSe2 samples were obtained on the silicon wafer, washed by deionized water three times, and dried by N2 purging.

Fabrication of micro-device (microcell) for electronic/electrochemical measurement

A silicon wafer with 275 nm SiO2 was cleaned and spun with a layer of photoresist (AZ5214, Nalgene). The as-prepared silicon wafer was patterned by ultraviolet lithography and deposited with Ti (20 nm)/Au (50 nm) as metal electrodes through physical vapor deposition (PVD), followed by the transfer of 2D materials and the wet impregnation for single-atom loadings. Then the silicon wafer was spun with a layer of polymethylmethacrylate (495 PMMA A8, Microchem), patterned by e-beam lithography (eBL), and deposited with Ti (20 nm)/Au (50 nm) as the source and drain electrodes, which directly contact the 2D SACs. After lifting off the layer of PMMA, we obtained the individual micro-device for electrical measurement. Based on the fabrication of the individual micro-device, a new PMMA film was spun on the silicon wafer as the passivation layer, where a window was patterned and developed to expose the fixed area (20–100 μm2) of 2D supports. The electrolyte was dripped in the window to directly contact the electrocatalysts. A graphite rod and a leak-free Ag/AgCl electrode (with 3.4 M KCl) inserted into the droplet were used as the counter electrode (CE) and the reference electrode (RE), respectively. And the 2D SACs connected with the source and drain electrodes were used as the working electrode (WE). Meanwhile, a self-reliant Au electrode (contacting the silicon directly), was fabricated as the gate electrode. To this end, the four-electrode microcell was set up for electronic/electrochemical measurement. See more details in Supplementary Figs. 2325.

On-chip in situ electronic/electrochemical measurement under OEEF regulation

A three-channel system consisting of three source measurement units (SMU) from two device analyzers (Agilent B2902A) was designed for the in situ measurements. SMU1 and SMU2 from one analyzer were used for collecting the electronic and electrochemical signals during the electrocatalytic process, and SMU3 from the other one was used for the control of gate voltages (Vg). In detail, SMU1 is regarded as an electrochemical workstation to control the potential of WE as to RE, and collect the electrochemical current (Ie) through CE. SMU2 channel is used for adding a small bias (Vds) between the source and drain electrodes to collect the electronic signals (Ids) through WE. SMU3 is used for applying variable Vg (through the gate electrode) to introduce OEEFs on 2D materials. Notably, Ie is generally several orders of magnitude smaller than Ids, indicating that the electronic signals cannot be affected by the electrochemical current. And the tiny leakage current (Ig < 1 nA) flowing gate electrode can be also neglected. In typical cyclic voltammetry (CV) and linear sweep voltammetry (LSV) measurements, 0.5 M H2SO4 was used for HER and 0.1 M KOH was used for OER, with the scan rate of 20 mV s−1. Ie was recorded by SMU1 and the electrochemical current density (j) was calculated through normalizing Ie with the exposed area of 2D materials. Ids was recorded by SMU2 and the sheet conductance (Gs) is given by,

$${G}_{{{{{{\rm{s}}}}}}}=\frac{1}{{R}_{{{{{{\rm{s}}}}}}}}=\frac{L}{W}\times \frac{{I}_{{{\mbox{ds}}}}}{{V}_{{{\mbox{ds}}}}}$$

where Rs is the sheet resistance, L is the channel length, and W is the channel width.

We express the electrochemical reference voltage with respect to RHE,


Notably, area density and mass density of 2D SACs were obtained through the statistical analysis of STEM observations, to determine their mass activities (more details in Supplementary Tables 1 and 4). Besides, to accurately evaluate the electrocatalytic performance of 2D SACs proposed in this work, electrochemical measurements for the determination of catalytic activities (in Fig. 5) were conducted without operating the in situ electrical transport measurement, to avoid the potential deviations of work electrodes (micro-devices) induced by small bias (Vds). The bias-dependent electrochemical potential deviations had been shown in Supplementary Figs. 77 and 78.

On-chip electrochemical reactor with OEEF modulation and gas product analysis

The CVD-grown large-size graphene (centimeter-level) was further employed to support single-atom catalysts and was fabricated into on-chip micro-device reactor, with a gas-tight PDMS placed on top as electrolyte cell and gas collection chamber. The electrolyte was injected into the cavity of PDMS to directly contact the exposed area (10 mm2) of 2D SACs. A graphite rod and a leak-free Ag/AgCl electrode (with 3.4 M KCl) were used as the counter electrode (CE) and the reference electrode (RE), respectively. The SACs (on 2D graphene) connected to the Au electrodes were regarded as the working electrode (WE), and a self-reliant Au electrode was also fabricated as the gate electrode, which contacts the silicon directly. In this three-electrode system, a series of electrochemical tests were performed through a CHI630E electrochemical workstation, with gas products collected by a syringe after electrolysis. In the potentiostatic HER tests, the applied potential (vs. RHE) is: −0.3 V for graphene, −0.1 V for Pt SAs-graphene. In the potentiostatic OER tests, the applied potential (vs. RHE) is: 1.8 V for graphene, 1.6 V for Co SAs-graphene. The detection of H2 and O2 collected in the electrochemical reactor was analyzed by a Shimadzu GC-2030 gas chromatograph. See more details in Supplementary Figs. 26 and 8285.

Material characterizations

The atomic resolution STEM images and STEM-EDS elemental mappings were recorded on an aberration-corrected STEM Titan cubed with a field emission gun at 300 kV. The beam current was 38 pA and the probe convergence semi-angle was 20.5 mrad and the angular range of the HAADF detector was from 60.5 to 200 mrad. SEM images were obtained on Hitachi S-4800 with samples deposited on carbon conductive tapes. The thickness characterizations of micro-devices were obtained on Bruker Dimension Icon atomic force microscope (AFM). The X-ray absorption data at the Pt L-edge of samples were recorded in the fluorescent mode at beamline 14W197 of the Shanghai Synchrotron Radiation Facility (SSRF), China. The station was operated with a Si (111) double-crystal monochromator at the energy of 3.5 GeV and the photon energy was calibrated with the first inflection point in Pt L-edge of Pt foil. X-ray photoelectron spectroscopy (XPS) was measured on a UlVAC-PHI 5000 Versa Probe spectrometer with Al-Kα as the radiation source. The binding energies in recorded spectra were calibrated by the C 1s peak, the internal standard reference at 284.6 eV. Raman and photoluminescence (PL) spectroscopies were performed on HORIBA XploRA Plus Raman microscope with a 532 nm laser. XRD patterns were recorded on a Shimadzu Lab X/XRD-6000 X-ray diffractometer equipped with a Cu-Kα radiation source (λ = 0.15418 nm) operating at 40 kV and 30 mA. The electrical measurements were performed on Lakeshore TTPX probe station. The detection of gas products collected in the electrochemical reactor was realized by a Shimadzu GC-2030 gas chromatograph.

Theoretical calculations

All the DFT calculations in this work were performed by using the Vienna ab initio simulation package (VASP)98,99. The projected augmented wave (PAW) method100 with an energy cutoff of 500 eV was selected to describe the ion-electron interactions. The electron exchange-correlation interactions were represented by generalized gradient approximation (GGA) with the Perdew–Burke–Ernzerhof (PBE) functional101. Dipole correction102 and spin polarization were also considered during all calculations. The k-point mesh for structural optimization was sampled using a 3 × 3 × 1 Monkhorst–Pack grid103. The structural optimization was terminated as the absolute force on each atom was smaller than 0.02 eV/Å and the electronic minimization reached the convergence criterion of 1×10−6 eV. Bader charge calculations were used to analyze the charge transfer104. ΔGH is commonly used to indicate the HER performance. A catalyst with a “|ΔGH| → 0” might be a good HER catalyst and the hydrogen adsorption free energy (ΔGH) is computed using the equation ΔGH = ΔEH + ΔEZPE + ΔU0→T – TΔS65. ΔEH is the hydrogen adsorption energy computed by the equation ΔEH = E(*H) – E(*) – 0.5E(H2), where E(*H) and E(*) are the total energies of the model with and without hydrogen adsorbed on the binding site *, respectively, and E(H2) is the total energy of the single hydrogen molecule. ΔEZPE is the zero-point energy difference and T is the temperature (298.15 K in this study). ΔU0→T is the internal energy at temperature T and ΔS means the entropy change caused by the adsorbates. Detailed calculation equations and values of ZPE, TS, and ΔU0→T are given in Supplementary Tables 12 and 17. The Gibbs free energy profiles including three oxygen intermediates (*OH, *O, and *OOH)105 were also calculated for OER systems using a similar method. The overall OER process (2H2O → O2 + 4H+ + 4e) can be divided into four electron reaction paths under standard conditions (pH = 0, U = 0 V) shown as follows,

$${{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}({{{{{\rm{l}}}}}})+{}^{\ast }\to {}^{\ast }{{{{{\rm{O}}}}}}{{{{{\rm{H}}}}}}+{{{{{{\rm{H}}}}}}}^{+}({{{{{\rm{aq}}}}}})+{{{{{{\rm{e}}}}}}}^{-}$$
$${}^{\ast }{{{{{\rm{O}}}}}}{{{{{\rm{H}}}}}}\to {}^{\ast }{{{{{\rm{O}}}}}}+{{{{{{\rm{H}}}}}}}^{+}({{{{{\rm{aq}}}}}})+{{{{{{\rm{e}}}}}}}^{-}$$
$${{{{{{\rm{H}}}}}}}_{2}{{{{{\rm{O}}}}}}({{{{{\rm{l}}}}}})+{}^{\ast }{{{{{\rm{O}}}}}}\to {}^{\ast }{{{{{\rm{O}}}}}}{{{{{\rm{OH}}}}}}+{{{{{{\rm{H}}}}}}}^{+}({{{{{\rm{aq}}}}}})+{{{{{{\rm{e}}}}}}}^{-}$$
$${}^{\ast }{{{{{\rm{O}}}}}}{{{{{\rm{OH}}}}}}\to {}^{\ast }+{{{{{{\rm{O}}}}}}}_{2}({{{{{\rm{g}}}}}})+{{{{{{\rm{H}}}}}}}^{+}({{{{{\rm{aq}}}}}})+{{{{{{\rm{e}}}}}}}^{-}$$

where * represents the adsorption site, l represents the liquid phase, aq represents the solution, and g represents the gas phase, respectively. The Gibbs free energies of intermediates were denoted as ΔGOH, ΔGO, and ΔGOOH, respectively. Then, the free energy differences of each step could be calculated by equations as ΔG1 = ΔGOH, ΔG2 = ΔGO − ΔGOH, ΔG3 = ΔGOOH − ΔGO, and ΔG4 = 4.92 − ΔGOOH. The overpotential η of OER process at U = 0 V can be calculated using the following equation,

$${\eta }_{{{{{{\rm{OER}}}}}}}={{{{{\rm{max }}}}}}\{{\Delta G}_{1},\,{\Delta G}_{2},\,{\Delta G}_{3},\,{\Delta G}_{4}\}/e\,-1.23{{{{{\rm{V}}}}}}$$

The potential-determining step is assigned to the step corresponding to the maximum of free energy difference. Dehydrogenation of *OOH is thermodynamically easy to occur in our research systems, so ΔG4 was not considered.

The explicit solvation model with adsorbed water molecules106 was constructed to investigate the OEEF-induced effect from solvents at the interface. For HER, one water molecule was put on top of the adsorbed *H in Pt SAs-MoS2. Furthermore, two water molecules were placed around the oxygen intermediates (*OH, *O, and *OOH) in Co SAs-WSe2 for OER to consider the influence of short-ranged intermolecular hydrogen bonding interactions, with more details shown in Supplementary Information.