Atomic-precision Pt6 nanoclusters for enhanced hydrogen electro-oxidation

The discord between the insufficient abundance and the excellent electrocatalytic activity of Pt urgently requires its atomic-level engineering for minimal Pt dosage yet maximized electrocatalytic performance. Here we report the design of ultrasmall triphenylphosphine-stabilized Pt6 nanoclusters for electrocatalytic hydrogen oxidation reaction in alkaline solution. Benefiting from the self-optimized ligand effect and atomic-precision structure, the nanocluster electrocatalyst demonstrates a high mass activity, a high stability, and outperforms both Pt single atoms and Pt nanoparticle analogues, uncovering an unexpected size optimization principle for designing Pt electrocatalysts. Moreover, the nanocluster electrocatalyst delivers a high CO-tolerant ability that conventional Pt/C catalyst lacks. Theoretical calculations confirm that the enhanced electrocatalytic performance is attributable to the bifold effects of the triphenylphosphine ligand, which can not only tune the formation of atomically precise platinum nanoclusters, but also shift the d-band center of Pt atoms for favorable adsorption kinetics of *H, *OH, and CO.

A lkaline exchange membrane fuel cells (AEMFCs) have been nominated as one of the most promising energy suppliers due to their green attributes and high energy efficiency 1,2 . Although AEMFCs are able to deliver Pt-like performance in economic electrocatalysis of cathodic oxygen reduction reaction (ORR), the same cannot be said for the anodic hydrogen oxidation reaction (HOR) [3][4][5][6][7][8][9][10][11][12] . Indeed, the sluggish HOR kinetics in alkaline media (in comparison to that in acidic media) demands a much higher Pt loading (almost ten times higher than that in acids) and thus inevitably results in high costs [13][14][15][16][17] . Despite intensive investigations in cost-effective alternative electrocatalysts in the past several decades, Pt remains the best HOR electrocatalyst, inherently owing to its near-optimal hydrogen binding energy (HBE). However, the conventional Pt/C electrocatalyst is suffering from high cost, poor anti-poisoning capability, and insufficient stability, which bring about additional obstacles for industrial applications of Pt-based catalysts in alkaline HOR 18 . Therefore, it is highly desirable to engineer the Pt catalysts at the atomic level for higher atomic utilization, enhanced anti-poisoning capability, and better stability.
Atomically precise metal nanoclusters (NCs), which bridge metal single atoms (SAs) and large-sized metal nanoparticles (NPs, >2 nm), have emerged as functional materials for addressing various challenges in diverse fields, due to their unique molecule-like structural properties (e.g., ultrasmall size (<2 nm), ultrahigh surface-to-volume ratio, atomic-precision structure, and rich surface chemistry) and physicochemical properties (e.g., luminescence, chirality, and discrete HOMO-LUMO transition) [19][20][21][22][23][24][25][26][27][28][29] . For example, ultrasmall metal NCs (e.g., Au, Pt, Pd, and their alloy analogues) have been employed as model catalysts for CO 2 reduction reaction 30 , alcohols oxidation 31 , hydrogen evolution reaction (HER) 32 , and ORR 33,34 , which makes the understanding on the catalytic mechanism as well as the design of nanocatalysts stepping from nanoscale into the molecular level 35 . Specifically, a precise metal cluster containing only a few atoms could have a discrete energy band structure, which would impel the d-band center to shift with the number of metal atom for favorable adsorption kinetics modulation of intermediates with purpose 36 . From this viewpoint, atomically precise metal NCs may provide a good opportunity for us to address the aforementioned challenges in the design of Pt catalysts, especially considering NC's powerful capabilities in manipulating atom-precision structure and regulating the interfacial catalytic behavior.
However, the synthesis of Pt NCs with definite Pt atom number and decent surface chemistry is a long-standing challenge in the cluster research community. This is most probably due to the vibrant catalytic activity of freshly generated Pt atoms that induce in-situ deposition and catalytic decomposition of reducing agents, like NaBH 4 in most reduction-growth strategies, giving rise to wild reduction kinetics that makes the controllable growth of Pt NCs unsuccessful. In addition, the surface chemistry of Pt NCs should also be engineered for the design of Pt NCs electrocatalysts, because the organic ligands not only influence the stability as well as the catalytic activity of Pt NCs through regulating the atomic packing structure and dictating the diffusion of reactants on the surface of Pt NCs, but also impact the adsorption/desorption competition between reactants and poisoning species on the catalytic active sites of Pt NCs 37 .
Herein, we report the design of Pt 6 (PPh 3 ) 4 Cl 5 NCs toward the alkaline HOR catalysis by utilizing a mild reducing agent, boranetert-butylamine (TBAB for short), in a less polar medium, and choosing triphenylphosphine (PPh 3 ) molecules as a benign protecting agent. Benefiting from the self-optimized ligand effect and precise atomic structure, the catalyst exhibits excellent catalytic activity and high stability and outperforms Pt SAs and Pt NPs analogues. Moreover, it also exhibits high tolerance against CO impurity by weakening the binding energy of CO intermediates. Theoretical simulations highlight the structural significance of the atomic-precision Pt 6 NCs and further uncover the distinctive catalytic mechanism of molecular-structure Pt 6 NCs for enhancing the whole HOR performance.

Results and discussion
Synthesis and characterization of catalyst. Regarding the NCs synthesis, it is well-documented that a mild reduction environment is crucial to the precise size control of metal NCs 38,39 , which may be even more pivotal in the case of Pt NC synthesis. This is due to the superior catalytic performance of freshly formed Pt, which could accelerate the decomposition of NaBH 4 , yielding too fast reduction kinetics. Once introducing the strong reducing agent NaBH 4 into the solution containing Pt precursor (Supplementary Fig. 1 and Supplementary Movie 1), abundant bubbles of H 2 , which is the product of NaBH 4 decomposition and acts as active species in the subsequent reduction reaction, could be observed. A similar phenomenon was not observed during the synthesis of Au and Ag NCs within commonly used NaBH 4reduction systems 19,38 , evidencing the self-catalytic feature of Pt in the NaBH 4 reduction system. Owing to the wild and thus less controllable reduction kinetics inherently induced by the selfcatalytic habit of Pt, no molecularly pure Pt NCs but polydisperse Pt particles (40 to 90 nm) could be obtained by typical NaBH 4 reduction methods ( Supplementary Fig. 2).
In order to achieve atomic precision in Pt NC synthesis, as shown in Fig. 1a, we employed a mild reducing agent TBAB to replace the NaBH 4 to minimize the catalytic effect of Pt species (Route I), thus slowing down the reduction kinetics for Pt NC growth. The reduction kinetics was further retarded by using less polar solvent (e.g., toluene and ethanol) to replace polar water, thereby creating a mild reduction environment 39 . Besides, PPh 3 ligands with the proper affinity to Pt are selected for the formation of atomically precise Pt NCs. Here, PPh 3 can serve as σ donors and π acceptors, leading to the formation of pπ-dπ P-Pt bonds with moderate binding energy to avert poisoning of Pt active sites 40,41 . Such a three-pronged approach results in PPh 3protected Pt NCs with atomic precision. By simply tuning the experimental parameters, the PPh 3 -protected Pt SAs and largesized Pt NPs are also obtained as references for the latter electrocatalytic test.
Upon acquiring Pt NCs, ultraviolet-visible (UV-vis) absorption spectroscopy, as well as electrospray ionization mass spectrometry (ESI-MS) were firstly performed to analyze their size. As shown in Fig. 1b, two absorption bands at 335 and 385 nm are observed in the UV-vis spectrum of Pt NCs, which implies the formation of ultrasmall Pt NCs with molecule-like optical absorption features 39 . The ESI-MS result of Pt NCs shows an intense peak centered at m/z = 2395 (Fig. 1c), which can be assigned to Pt 6 (PPh 3 ) 4 Cl 5 carrying +1 charge. The experimentally obtained isotope pattern of Pt 6 (PPh 3 ) 4 Cl 5 is in good agreement with the simulated one (inset in Fig. 1c), corroborating the formation of Pt 6 (PPh 3 ) 4 Cl 5 NCs (denoted as Pt 6 NCs) 42 . Such a result also highlights the importance of precisely controlling the reduction kinetics in dictating the size monodispersity of atomically precise Pt NCs. By contrast, the sample prepared without introducing the reducing agent, TBAB, presents a distinctive peak at 390 nm in the UV-vis spectrum (Supplementary Fig. 3), and the intense peak centered at m/z = 754 in the ESI-MS spectrum ( Supplementary Fig. 4) corresponding to Pt 1 (PPh 3 ) 2 Cl, evidencing the formation of single-atom Pt catalyst (denoted as Pt 1 SAs). In addition, the as-fabricated Pt 6 NCs display good stability, and there is no obvious change in the ESI-MS spectra for the Pt 6 NCs after three weeks of storage at 4°C without N 2 protection ( Supplementary Fig. 5).
Prior to the electrochemical test, the Pt 6 NCs were firstly deposited on commercially available carbon black by impregnation adsorption, and the as-formed Pt 6 NCs/C was characterized by transmission electron microscopy (TEM). As shown in Fig. 1d, individual NC is visualized as a separated dark dot with a typical diameter of~1 nm (inset in Fig. 1e). The formation of~1 nm NCs was also supported by high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), where NCs appear as bright dots (Fig. 1e). Aberration-corrected HAADF-STEM images ( Fig. 1f and inset) confirm the atom-precise synthesis of clusters based on the expected structural models. The above-mentioned morphological characterizations demonstrate that Pt 6 NCs are well dispersed (without notable aggregation or agglomeration), which should be attributed to the steric effects of PPh 3 ligands and the immobilization effects of the carbon support. Elemental maps of the Pt 6 NCs/C (Fig. 1g) confirm the uniform distribution of Pt and P elements across the surface of the carbon support. The Pt loading of Pt 6 NCs/C was measured to be 2.5 wt.% by thermogravimetric analysis (TGA; Supplementary  Fig. 6a). In comparison, TEM and HAADF-STEM images of carbon-supported Pt 1 SAs (denoted as Pt 1 SAs/C) reveal the formation of atomically dispersed Pt 1 atoms with only a few sporadic NCs or NPs ( Supplementary Fig. 7). TEM image of carbon-supported PtNPs (denoted as PtNPs/C) shows that the average size of NPs reaches up to 4.4 nm ( Supplementary Fig. 8a). The lattice fringe of the Pt NPs ( Supplementary Fig. 8b) has a regular spacing of 0.22 nm, corresponding well to the interplanar spacing of (1 1 1) plane of crystalline Pt. The Pt loading of Pt 1 SA/ C and PtNPs/C were calculated as 2.2 wt.% and 5.4 wt.% by TGA ( Supplementary Fig. 6), respectively. Furthermore, we also attempted to synthesize Pt NCs by using thiolate ligands (e.g., glutathione or GSH for short), which gives rise to Pt NCs with a much larger diameter of 1.9 nm ( Supplementary Fig. 9), implying that the interaction between ligands and Pt ions is crucial in determining the size and monodispersity of Pt NCs.
Powder X-ray diffraction (PXRD) analysis ( Fig. 2a) was carried out to confirm the crystalline structure of Pt-based catalysts. No characteristic peak associated with Pt species can be detected in the spectra of Pt 6 NCs/C and Pt 1 SA/C, thereby ruling out the Notably, the binding energy of Pt 4f in Pt 6 NCs/C presents a remarkable positive core-level shift of 0.4 eV as compared with that in PtNPs/C, inferring that more electron transfer from Pt atoms to P atoms due to the stronger orbital interactions 43 . This assertion is supported by our control experiments, where ligand removal by traditional calcination treatment can lower the binding energy of Pt 4f in Pt 6 NCs/C catalysts ( Supplementary  Fig. 10a). This readily means that the PPh 3 ligands have strong interactions with Pt centers and thus can significantly alter the electron distribution of the latter.
Evaluation of HOR performance. The HOR activity of Pt-based catalysts was evaluated in H 2 -saturated 0.1 M KOH aqueous solution. Pt/C com (an optimal loading of 10 μg cm −2 , on the Pt basis thereafter) was introduced as the benchmark for comparison ( Supplementary Fig. 11a). Generally, more Pt usage induces higher catalytic activity before reaching the optimum loading amount, and here it is demonstrated that further increasing the catalyst loading of Pt 6 NCs/C over 5 μg cm −2 would not lead to additional performance gain ( Supplementary Fig. 11b). Therefore, the mass loading of Pt 6 NCs/C was set to be 5 μg cm −2 to maximize the catalytic performance with the minimized Pt loading. As shown in Fig. 3a, the polarization curve of Pt 6 NCs/C quickly rises to the saturation current density of 2.7 mA cm −2 at an overpotential of 50 mV (vs. Reversible hydrogen electrode (RHE)). By contrast, the anodic current density of Pt/C com increases more mildly (1.8 mA cm −2 at an overpotential of 50 mV (vs. RHE)). Meanwhile, Pt 6 NCs/C displays a much higher anodic current than Pt 1 SAs/C and PtNPs/C at the kinetic control regions, implying that the atomically precise six-Pt-atom structure promotes their electrocatalytic activity on HOR. In contrast to the inconspicuous anodic current across the whole potential range in the Ar-saturated 0.1 M KOH (Supplementary Fig. 12), Pt 6 NCs/C shows a fast anodic current response in the presence of H 2 above 0 V (vs. RHE), suggesting that the anodic current is derived from the oxidation of H 2 .
The mass-specific activity of Pt 6 NCs/C was derived to quantitatively analyze the HOR activity of the catalysts. Figure 3b displays the HOR polarization curves under different electrode rotation rates. By plotting and fitting the inverse of the current density (j −1 ) at an overpotential of 50 mV (vs. RHE) versus the inverse of the square root rotation rate (ω −1/2 ), a straight line with a slope of 5.08 cm 2 mA −1 s −1/2 is obtained, agreeing well with the Koutecky-Levich equation (see details in the "Electrochemical measurements"). The as-calculated slope is approaching to the theoretical value of 4.87 cm 2 mA −1 s −1/2 for the twoelectron HOR process 44,45 . By extrapolating the ω −1/2 to 0, the kinetic current density (j k ) of 18.2 mA cm −2 and the massspecific kinetic current (j k,m ) of 3658 A g −1 are obtained. Remarkably, the mass-specific activity of Pt 6 NCs/C is almost 9.1 times higher than that of Pt/C com (402 A g −1 ) and 7.4 times higher than that of PtNPs/C (496 A g −1 ), as shown in Supplementary Fig. 13 and Supplementary Table 1. Assynthesized Pt 6 NCs/C catalysts also exhibit superior activity than most of the previously reported platinum group metal (PGM) catalysts (Supplementary Table 1). To the best of our knowledge, this is the highest attainable value among all monometallic HOR electrocatalysts under the tested conditions.
Tafel plots of Pt-based catalysts originated from the j k were depicted according to the Bulter-Volmer equation in Fig. 3c. The j k at an overpotential of 20 mV (vs. RHE) is 3.62 mA cm disk −2 for Pt 6 NCs/C, which is much higher compared to those of Pt 1 SAs/C, PtNPs/C, and Pt/C com catalysts. Since the HER and HOR branches are symmetrical with each other, the HOR reaction mechanisms of all catalysts are determined to be a Tafel (H 2 + 2* ⇋ 2*H, * represents adsorption site) -Volmer (*H + OH − ⇋ H 2 O + e − + *) process and the Volmer step is the rate-determining step. The exchange current density (j 0 ) is also determined from the linear fitting of the micro-polarization region (Fig. 3d). The massnormalized j 0 (j 0,m ) of Pt 6 NCs/C (646.3 A g −1 ) is about 29.4, 3.9, and 5.7 times higher than those of Pt 1 SAs/C (22.0 A g −1 ), PtNPs/C (164.5 A g −1 ), and Pt/C com (113.1 A g −1 ), respectively. The areaspecific exchange current density (j 0,s ) values are obtained by normalization of the j 0,m according to the corresponding a b  Table 1 and Supplementary Fig. 14). Pt 6 NCs/C catalyst exhibits a much higher j 0,s value (1.546 mA cm −2 ) compared to those of Pt 1 SAs/C (0.149 mA cm −2 ), PtNPs/C (0.446 mA cm −2 ), and Pt/C com (0.351 mA cm −2 ) catalysts (Fig. 3e), implying that Pt 6 NCs/C possesses better catalytic activity toward HOR than Pt 1 SAs/C and PtNPs/C. These results unambiguously indicate Pt 6 NCs as the optimal size for electrocatalytic HOR, in comparison to their smaller (SAs) and larger (NPs) counterparts.
To shed more light on the effect of PPh 3 ligands on HOR catalytic activity, we performed a thermal annealing treatment on Pt 6 NCs/C at 550°C under Ar atmosphere to remove the ligands (denoted as Pt 6 NCs/C-550). It is found that Pt 6 NCs/C-550 can largely retain the discrete and tiny morphology of Pt NCs, with only negligible agglomeration observed ( Supplementary Fig. 15). The calcination process eliminated PPh 3 ligands ( Supplementary  Fig. 10b), excluding any possible impact from ligands on the catalytic activity of NCs. Pt 6 NCs/C-550 possesses inferior HOR activity (j k,m = 2128 A g −1 and j 0,m = 462.2 A g −1 ) than Pt 6 NCs/ C ( Supplementary Fig. 16), which may be attributed to the slight aggregation of Pt NCs and the removal of the PPh 3 ligands. These results indicate that instead of blocking the active sites of Pt nanomaterials as previously reported, the PPh 3 ligands anchored on the surface of Pt 6 NCs can not only maintain the molecular structure of the Pt NCs, but also improve their catalytic activity probably by regulating the molecule-like catalytic behavior (e.g., adsorption/desorption kinetics of both reactants and the asproduced reaction intermediates, and reactivity of active sites). We termed the phenomenon above as the self-optimizing ligand effect, which is derived from the interaction between inorganic Pt and organophosphorus ligand structures, including steric repulsion and electronic effects 37 . In contrast to the PPh 3 ligand, the hydrosoluble GSH ligand containing the toxicant S element presents an adverse effect on the catalytic activity of Pt NCs c Representative HOR/HER Tafel plots of kinetic current density (j k ) for Pt 1 SAs/C, Pt 6 NCs/C, PtNPs/C, and Pt/C com catalysts. Symbols are calculated j k values, and lines are fitted results based on the Volmer-Butler equation. d Linear current potential region around the equilibrium potential of HOR/HER of Pt 1 SAs/C, Pt 6 NCs/C, PtNPs/C, and Pt/C com catalysts conducted at a scan rate of 5 mV s −1 with the rotation rate at 2500 rpm. The dotted lines indicate the linear fitting of the data. e Comparison of the mass-specific kinetic current (j k,m ) at an overpotential of 50 mV (vs. RHE) and the area-specific exchange current (j 0,s ) with those of previous reports. f HOR polarization curves for Pt 6 NCs/C and Pt/C com before (solid lines) and after (dashed lines) accelerated durability test, respectively.  Supplementary Fig. 17), indicating that the ligand attributes play a crucial role in catalyzing HOR.
Catalytic durability is another important attribute dictating the usefulness of Pt NCs in HOR. Firstly, an accelerated durability test (ADT) was performed on Pt 6 NCs/C catalyst by 2000 repetitive cyclic voltammetry (CV) scans in H 2 -saturated 0.1 M KOH aqueous solution (Supplementary Fig. 18a). It is found that the ECSA of Pt 6 NCs/C decreases by 3.6% and 15.2% after 1000 and 2000 cycles (Supplementary Fig. 18b). After 2000 cycles, the Pt 6 NCs/C manifests a 22.5 mV increase in half-wave potential, versus 34.5 mV observed on Pt/C com catalyst (Fig. 3f), indicating the robustness of Pt 6 NCs/C catalyst. Subsequently, the chronoamperometry was carried out to further evaluate the stability of Pt 6 NCs/C (Fig. 3g). After a continuous operation at an overpotential of 60 mV (vs. RHE) for 10 h, the Pt 6 NCs/C catalyst showed an inconspicuous degradation and retained 98.3% of the activity (current retention rate). By sharp contrast, a serious deterioration of more than 42% was observed in the catalytic activity of Pt/C com catalyst under the same conditions, indicating the superior durability of Pt 6 NCs/C to Pt/C com . The surface chemistry and size change of Pt 6 NCs/C after the HOR test were also characterized ( Supplementary Fig. 19). The similar Pt 4f and P 2P XPS spectra of Pt 6 NCs before and after ADT corroborate the good structural stability of Pt 6 NCs in the long-term HOR process ( Supplementary Fig. 19a, b). HAADF-TEM images reveal that no marked size change could be observed for the Pt 6 NCs/C catalyst after ADT ( Supplementary Fig. 19c, d). These results unambiguously manifest the superior HOR durability of Pt 6 NCs/C catalyst, which originates from the strong interactions of PPh 3 ligands with Pt as well as the proper bulkiness of PPh 3 ligands, effectively inhibiting the agglomeration of Pt NCs during the electrochemical process.
CO-tolerance evaluation. Apart from the activity and long-term durability, CO tolerance is another equally important aspect determining the industrial acceptance of HOR electrocatalysts. Pt is prone to CO poisoning even at low levels of CO (10 ppm) due to the preferential CO binding on Pt, which can block the active sites for hydrogen adsorption and dissociation. The CO tolerance of Pt 6 NCs/C, Pt 6 NCs/C-550, and Pt/C com catalysts was evaluated by feeding diluted (5%) CO to the H 2 -saturated electrolyte for a short period during the chronoamperometric test at an overpotential of 100 mV (vs. RHE) 46 . As depicted in Fig. 3h, Pt/C com catalyst suffered from a rapid decrease in activity and its anodic relative current abruptly dropped from 100% to 0 with the introduction of CO. The current cannot be recovered even the fuel stream is reverted to pure H 2 , confirming that CO poisoning is irreversible under the tested HOR potential window. By contrast, Pt 6 NCs/C catalyst can retain a high HOR activity (current retention rate: 77%) with the same level of CO contamination, evidencing its superior CO-tolerant ability. It should be noted that Pt 6 NCs/C-550 presents a lower CO tolerance than Pt/C com , suggesting the PPh 3 ligands play a crucial role in anti-poisoning. Additionally, the position of the CO stripping peak of Pt 6 NCs/C exhibits 170 mV negative shifts compared with Pt/C com (Supplementary Fig. 20), which further indicates that Pt 6 NCs/C possesses a much lower CO adsorption ability, resulting in the enhanced CO-tolerant ability. We rationalize that the selfoptimized ligand effect and the molecular structure of Pt 6 NCs offer efficient electron penetration and give rise to sluggish CO adsorption kinetics during the HOR 1 .
Mechanism investigation and computational studies. The alkaline HOR catalytic mechanism is still under severe debate, but HBE and/or OH binding energy (OHBE) are generally considered to be the activity descriptor based on theoretical and experimental observations 11 . It is well-documented that the alkaline HOR process follows a Tafel − Volmer or Heyrovsky (H 2 + OH − + * ⇋ H 2 O + *H + e − ) -Volmer route, and the Volmer step is regarded as the rate-determining step owing to the strong HBE of PGM. On the basis of the Sabatier principle, weakening HBE in alkaline media could favor higher HOR activity 11,14,18 . One of the salient features of Pt 6 NCs/C is that its HBE could be manipulated by both the molecule-like structure and the PPh 3 ligands of Pt 6 NCs. The HBE information of Pt 6 NCs/C and Pt/C com could be obtained directly from their CV curves. As shown in Fig. 4a, the CV curves exhibit that the hydrogen desorption peak of Pt 6 NCs/C is negatively shifted compared with that of Pt/C com (0.233 vs. 0.293 V (vs. RHE)), implying the smaller HBE as well as the result better HOR activity of Pt 6 NCs/C than that of Pt/C com .
The molecular structure of the PPh 3 -protected Pt 6 NCs could be largely responsible for the optimization of HBE. Owing to the difficulties in discriminating the molecular structure of Pt 6 NCs by ESI-MS analysis, we resorted to density functional theory (DFT) simulations to predict the possible structure of Pt 6 NCs. Our DFT simulations reveal a stable structure of Pt 6 NCs with the lowest energy consisting of partially P(CH 3 ) 3 − and/or Cl-protected six-Pt-atom octahedron, wherein three Pt atoms form Pt-Cl (2.382 Å) and Pt-P (2.236 Å) bonds simultaneously, while the other three Pt atoms form Pt-P (2.223 Å) and Pt-Cl (2.305 Å) bonds individually ( Supplementary Fig. 21). The average Pt-Pt bond length amounts to 2.717 Å ( Supplementary Fig. 22), which is in good agreement with the reported data in the literature for similar clusters 47 . In addition, the structural models of Pt 1 SAs and PtNPs were also simulated by DFT to clarify the correlation of cluster structure and HBE ( Supplementary Figs. 23 and 24). However, it should be mentioned that the structures of the Pt 6 NCs as well as Pt 1 SAs and PtNPs discussed here are simulated based on DFT, and their actual structures still rely on the singlecrystal X-ray crystallography, which will be realized with future research efforts. Nevertheless, the DFT simulation could provide useful information to deepen our fundamental understanding of this issue. Of note, based on DFT analysis, the electron gathering area in Pt-P bonds is close to the P atom, while the electron densities around the Pt atom are decreased accordingly (Fig. 4b  and c, Supplementary Figs. 25 and 26), indicating that the PPh 3 coordination can tune the electronic structure of Pt.
After modulating the electronic configuration, the d-band center of Pt sites would show a significant shift. According to the d-band theory, the lower d-band center far from the Femi level (E F ) could be corresponding to the weaker binding strength of intermediate adsorbates, owing to the lower band filling of the antibonding states 48,49 . It is shown that the d-band center of the Pt in Pt 6 NCs (−2.842 eV) downshifted far away relative to the E F compared to that of the Pt 1 SAs (−0.540 eV) and PtNPs (−2.689 eV) (Fig. 4d), which may be attributed to the intensive d-orbital interaction of Pt 6 NCs, thus revealing the weakened binding strength of *H and CO. In addition, it is found that the Gibbs free energy of *H (ΔG *H for short) of Pt 6 NCs (−0.07 eV; Fig. 4e) on the optimal adsorption site (Pt-P bonding) is very close to the ideal value (ΔG *H = 0) for HOR ( Supplementary  Fig. 27). By contrast, the ΔG *H values of Pt 1 SAs (0.71 eV) and PtNPs (−0.21 eV) demonstrate larger deviation from the ideal value. In parallel, the formation mechanism and tunable nucleophilicity of Pt 6 NCs can be well elucidated by the limited surface negative electrostatic potential obstructed by ligands bonded to Pt atoms ( Supplementary Fig. 28) 50 , consistent with the DFT analysis. Experimental and simulation results manifest that Pt 6 NCs possess the near-optimal HBE due to the selfoptimized ligand effect, which can tremendously facilitate the Volmer step for alkaline HOR. Indeed, the attribute of the rich surface chemistry of metal NCs may offer broad flexibility in regulating the d-band center of Pt NCs. For example, DFT simulation reveals that different N-heterocyclic carbene-derivative ligands ranging from strong to weak σ-donors can be employed to tune the d-band center of the Pt 6 NCs, thereby regulating the HBE (Supplementary Fig. 29) 51 . This strategy may provide a feasible research direction in the electronic regulation of catalysts for alkaline HOR catalysis.
Since the adsorbed * OH plays a crucial role to remove the adsorbed * H in the Heyrovsky and Volmer steps, increasing *OH coverage on Pt would greatly benefit alkaline HOR 52,53 . The Gibbs free energy of *OH (ΔG *OH ) was analyzed on the *OH adsorption models of Pt 1 SAs, Pt 6 NCs, and PtNPs ( Supplementary  Fig. 30). In alkaline media, the slow HOR rate on the Pt catalyst is closely associated with the inherent weak OH adsorption on Pt (1 1 1) due to weak Pt-OH interaction 54 . However, we found that Pt 6 NCs possess a much more negative ΔG *OH (−2.57 eV) than those of PtNPs (−2.06 eV) and Pt 1 SAs (−1.69 eV) (Fig. 4f), suggesting its enhanced adsorptions of *OH to the references. Furthermore, we conducted CO-stripping tests to reflect the strength of the OHBE on the Pt surface because more *OH can promote the removal of *CO 12 . CO-stripping results (Supplementary Fig. 20) show that Pt 6 NCs/C oxidizes *CO at much lower potential compared with Pt/C com (−170 mV), implying the stronger OHBE on Pt 6 NCs/C. Such an enhanced * OH adsorptive behavior would facilitate the capture of *OH species on the surface of Pt 6 NCs, leading to acceleration of the Volmer step during the alkaline HOR process.
Based on the results above, we reason that the enhanced *OH adsorption and decreased *H adsorption on Pt 6 NCs would synergistically accelerate alkaline HOR. The reaction between surface adsorbed * OH and * H to form H 2 O is the ratedetermining step of alkaline HOR (Fig. 4g). Energy barriers of this reaction are calculated to be 0.47, 0.82, and 1.53 eV on the surface of Pt 6 NCs, Pt 1 SAs, and PtNPs, respectively ( Fig. 4h and Supplementary Fig. 31). The much lower energy barrier of Pt 6 NCs contributes to its remarkable HOR activity in alkaline media. In terms of CO adsorption, DFT simulations reveal that due to the downshift of the d-band center based on the ligand effect, Pt 6 NCs possess a much lower CO adsorption strength (−1.1 eV) than that of PtNPs (−1.46 eV) and Pt 6 NCs/C-550 (−1.87 eV) ( Fig. 4i and Supplementary Fig. 32), suggesting its much lower binding energy to CO. These results explain the marked COtolerance ability of Pt 6 NCs/C and agree well with above COstripping trend. Besides, stronger OHBE on Pt 6 NCs surface than PtNPs assists in the oxidation of adsorbed CO, which also leads to the enhanced CO-tolerance ability and recoverable catalytic activity.
In summary, we have successfully synthesized atomically precise PPh 3 -stabilized Pt 6 NCs leveraging on creating a mild reduction environment and achieved superior HOR catalytic performance in the alkaline medium. Mechanism studies demonstrate that the Pt 6 NCs structure and PPh 3 ligand may play decisive roles in tuning the HBE, OHBE, H 2 O formation energy, and CO absorption energy. Benefiting from the selfoptimized ligand effect and atomically precise six-Pt-atom structure, the Pt 6 NCs/C catalyst exhibits much better HOR catalytic performance in terms of activity, durability, and COtolerant ability. This study exemplified the application of atomically precise Pt NCs for high-performance HOR catalysis, and the knowledge conveyed here may be inspirable for the future design of metal NCs-based electrocatalysts for energy conversion.

Methods
Materials and chemicals. Chloroplatinic acid hexahydrate (H 2 PtCl 6 ·6H 2 O, 37.5% Pt basis), triphenylphosphine (PPh 3 , 99%), toluene (99.5%), borane-tertbutylamine complex (TBAB, ≥95.0%), sodium borohydride (NaBH 4 , 98%), L-reduced glutathione (GSH, 98%), ethanol (≥99.5%), Nafion solution (~5% in a mixture of lower aliphatic alcohols and water), and commercial 20% Pt/C catalyst were purchased from Sigma-Aldrich. Carbon black (Vulcan XC-72R) was bought from Carbot Co. All reagents were used as received without further purification. Synthesis of GSH-PtNCs/C. Typically, 5 mL of 50 mM H 2 PtCl 6 ·6H 2 O aqueous solution and GSH (0.94 mmol) were added into a beaker. 2.0 mL of NaBH 4 (1.9 mmol) was added dropwise under vigorous stirring. After stirring for 30 min, 20 mg of XC-72R was added into the above suspension and kept stirring for 2 h. Finally, the final product was collected by vacuum filtration and denoted as GSH-PtNCs/C. Physical characterizations. UV-vis absorption spectra were recorded on an Agilent 8453 spectrometer. ESI-MS was conducted on a Bruker Impact II time-offlight MS system. TEM images and HAADF-STEM images were obtained with a JEM-2100UHR field emission electron microscope at an accelerating voltage of 200 kV. Atomic resolution high magnification STEM image was obtained using an aberration-corrected transmission electron microscope (JEOL, JEM-ARM200F) operated at 200 kV. XRD patterns were collected on a D8 Advance X-ray powder diffractometer equipped with a Cu Kα radiation source (λ = 0.15405 nm) at 40 kV. All the diffraction data were collected in a 2θ range from 5°to 75°at a scanning rate of 8°min −1 . XPS analyses were acquired with a PHI 5000 Versa Probe spectrometer from ULVACPHI using an Al Kα (1486.6 eV) photon source. All binding energies were calibrated using the C(1s) carbon peak (284.8 eV), which was applied as an internal standard. TGA was performed on SDT650 from NETZSCH. The starting temperature was 20°C with a 5°C min −1 ramp rate to 800°C under 100 mL min −1 airflow 55 .
Electrochemical measurements. To prepare the electrode, 2.0 mg of catalyst, 490 μL of ethanol, and 10 μL of 5 wt.% Nafion solution were mixed and sonicated for 1 h to form a homogeneous catalyst ink. A certain amount of catalyst ink was dropped on the surface of the glassy carbon rotating disk electrode (RDE) in batches (diameter: 4 mm). For HOR, the loading was 5 μg cm −2 (on the basis of the weight of Pt species in the whole experiment) to have a wide kinetic-diffusion mix controlled region so that Koutecky-Levich equation could be applied. Before any electrochemical measurement, three quick LSVs from −0.02 V to 0.5 V (vs. RHE) with a scan rate of 10 mV s −1 were performed to activate the catalyst. 20 wt.% Pt/C was used as a reference and the loading amount was 10 μg cm −2 on the RDE, and then dried naturally at room temperature. All electrochemical measurements were performed in a standard three-electrode system at room temperature by the RRDE-3A connected to CHI Electrochemical Station (Model 660d). All reference potentials have been converted to a reversible hydrogen electrode and were iRcorrected. The RDE coated with the catalyst, a Hg/HgO with 1.0 M KOH solution filled, and a graphite rod were employed as the working electrode, reference electrode, and counter electrode, respectively 55 .
All electrochemical tests were performed in a 0.1 M KOH solution. The linear sweep voltammetry (LSV) was carried out in the KOH solution saturated with H 2 at various rotation rates from 400 to 3600 rpm to evaluate the catalytic activity of HOR. The test parameters of LSV were set to a voltage window from −0.04 V to 0.5 V (vs. RHE) and a sweep speed of 5 mV s −1 . The cyclic voltammetry (CV) curves were obtained in an Ar-saturated 0.1 M KOH solution by scanning the potential from 0.005 V to 1.1 V (vs. RHE) at 50 mV s −1 . The ECSAs were evaluated from the integral area of hydrogen underpotential deposition (H upd ) peaks with the subtraction of the double layer in the CV curve and a charge density of 210 μC cm Pt −2 . Electrochemical impedance spectroscopy (EIS) measurement was conducted with a frequency ranging from 0.1 to 10 5 Hz and an amplitude of 5 mV under an overpotential of 30 mV. The accelerated durability tests were carried out in 0.1 M KOH between 0.005 V and 0.6 V (vs. RHE) at a scan rate of 100 mV s −1 for 2000 cycles. CO stripping was performed by holding the RDE at 0.1 V (vs. RHE) for 10 min in the purged CO to adsorb CO. Then RDE was quickly moved to a fresh 0.1 M KOH solution and recorded the two CV cycles in a potential region from 0 to 1.2 V (vs. RHE) at a sweep rate of 20 mV s −1 . The kinetic current density (j k ) can be obtained by the Koutecky-Levich (K-L) equation: where j is the measured current, which can be deconvoluted into j k and diffusional current (j d ) components; B is the Levich constant; C 0 is the solubility of H 2 in the electrolyte; ω is the angular velocity of rotating disc electrode during measurements. The exchange current density (j 0 ) can be calculated by fitting j k into the Butler-Volmer (B-V) equation: where α, R, T, and η represent the transfer coefficient, the universal gas constant, the operating temperature, and the overpotential, respectively. In a small potential window of the micro-polarization region near the equilibrium potential, j k approximately equals to j. In this case, the B-V equation can be expanded by Taylor's formula and simplified as Eq. (3): By linearly fitting the polarization curve in the micro-polarization region, the j 0 can be obtained.
Computational models and methods. To comprehensively consider the accuracy and efficiency of the calculations, the P(C 6 H 5 ) 3 ligand was represented by the P(CH 3 ) 3 group in this work. Therefore, according to the experimental results, the Pt 1 species was built by a Pt coordinated by two P(CH 3 ) 3 groups (Pt(P(CH 3 ) 3 ) 2 ), while the Pt 6 species was modeled as a Pt 6 cluster coordinated by five Cl ions and four P(CH 3 ) 3 groups (Pt 6 (P(CH 3 ) 3 ) 4 Cl 5 ). To study PtNPs, Pt(1 1 1) was built using a periodic four-layer slab with a p(4 × 4) unit cell. To study Pt 6 NCs/C-550, the Pt octahedron was constructed by using six Pt atoms. A vacuum of 20 Å in thickness was used to separate the catalyst surface from its periodic image in the direction along the surface normal.
All the calculations were performed by the spin-polarized density functional theory (DFT) method. Each model was optimized by minimizing the energy by the DFT calculation. The metal ion cores were treated by density functional semi-core pseudopotential (DSPP) 56 , while the valence electrons were represented by the double numerical plus polarization (DNP) basis set. The exchange-correlation energy was calculated by the Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation 57 . Moreover, the Grimme' PBE+D2 correction was adopted to consider the long-range dispersion forces, which exhibits comparable accuracy with the PBE0 hybrid XC functional including an ab initio van der Waals correction 58,59 . According to the Monkhorst-Pack (MP) scheme, the Brillouin zone was sampled with a 4 × 4 × 1 k-mesh for the slab calculation. The convergence tolerance of the geometry optimization was set as 1 × 10 −5 Ha for energy, 2 × 10 −3 Ha Å −1 for maximum force, and 5 × 10 −3 Å for maximum displacement [60][61][62] . The transition state was searched by the linear synchronous transition/quadratic synchronous transit (LST/QST) method, confirmed by the frequency calculations. The geometry optimization, transition state search, and frequency calculations were performed by the DMol 3 code 59,61 .
The density of states (DOS) and electron density difference were calculated by the CASTEP code with a cutoff energy of 750 eV 63 . In calculations, the normconserving pseudopotential was chosen 64 . The Brillouin zone was sampled using 8 × 8 × 8 k-points. The other details were consistent with the above DMol 3 computations.
The change of free energy for the reaction was calculated by Eq. (4): where ΔE, ΔZPE, and ΔS are the change of electronic energy, zero-point energy, and entropy at the temperature of 298 K, respectively. The adsorption energy of an adsorbate (E ads ) was calculated by Eq. (5): where E adsorbate+sub is the total energy of the adsorbed system, and E adsorbate is the energy of an adsorbate. The d-band center (ε d ) of the catalysts was calculated by Eq. (6): where ρ d (E) is the density of d states at an energy level E.