Site-specific electrodeposition enables self-terminating growth of atomically dispersed metal catalysts

The growth of atomically dispersed metal catalysts (ADMCs) remains a great challenge owing to the thermodynamically driven atom aggregation. Here we report a surface-limited electrodeposition technique that uses site-specific substrates for the rapid and room-temperature synthesis of ADMCs. We obtained ADMCs by the underpotential deposition of a non-noble single-atom metal onto the chalcogen atoms of transition metal dichalcogenides and subsequent galvanic displacement with a more-noble single-atom metal. The site-specific electrodeposition enables the formation of energetically favorable metal–support bonds, and then automatically terminates the sequential formation of metallic bonding. The self-terminating effect restricts the metal deposition to the atomic scale. The modulated ADMCs exhibit remarkable activity and stability in the hydrogen evolution reaction compared to state-of-the-art single-atom electrocatalysts. We demonstrate that this methodology could be extended to the synthesis of a variety of ADMCs (Pt, Pd, Rh, Cu, Pb, Bi, and Sn), showing its general scope for functional ADMCs manufacturing in heterogeneous catalysis.

voltammograms (CVs; Supplementary Fig. 9), at which only the UPD process occurs. The resulting product was transferred into an argon-filled glovebox for storage at room temperature.

Large-scale synthesis of Pt-SAs/MoS 2
Typically, single-atom catalysts synthesized in the laboratory significantly outperform commercial catalysts. However, most of these nanoscale studies are fundamental, and the technology is not scaled up for widespread adoption. Therefore, we designed a site-specific electrodeposition (SSED) device for potential macroscale production. The electrolysis system was constructed from a "U-type" electrochemical cell with a proton-exchange membrane ( Supplementary Fig. 11). In this three-electrode configuration, a graphite rod and Ag/AgCl Then, a solution of 0.05 M H2SO4 containing 5 mM K2PtCl4 (5 mL; degassed with bubbling argon for 30 min) was immediately injected into the cell containing the black powder, and the reaction was allowed to stir for 30 min. The resulting product was washed several times with water, frozen by liquid nitrogen, and lyophilized overnight. All the synthetic procedures were conducted at ambient temperature. The resulting product was then transferred into an argonfilled glovebox for storage.

Operando Raman Spectroscopy of ce-MoS 2 during the UPD of Cu
An in-house-built electrochemical cell was used for the operando Raman spectroscopy experiments ( Supplementary Fig. 17). In a typical experiment, an indium tin oxide (ITO) substrate spin-coated (acceleration rate: 1000 rpm; rotation speed: 3000 rpm) with ce-MoS2 nanosheets was utilized as the working electrode. A polydimethylsiloxane (PDMS) membrane, which was synthesized according to standard procedures, was used to control the exposed area of the working electrode. A platinum wire and a Ag/AgCl electrode served as the counter and reference electrodes, respectively. A potentiostat was used to apply potentials of +0.10 V to the working electrode while the Raman spectra were collected. A confocal Raman microscope (FTRaman Spectrometer, Bruker) was used to acquire in situ and operando Raman spectra using a 532 nm laser with an acquisition time of 15 s for each spectrum. During the reaction, electrolyte (2 mM CuSO4 containing 0.1 M H2SO4, approx. 200 μL) was added. A CHI 660E potentiostat was used to establish the electrolysis conditions for the Cu UPD.

Supplementary Note 1. Conceiving a new electrochemical approach for ADMC preparation
Conductive substrates with high surface areas (e.g., graphene and carbon nanotubes) have often been used to electrodeposit metal nanoparticles or metallic thin films ( Supplementary Fig. 1).
However, this process typically leads to the formation of a multilayer bulk phase.
Underpotential deposition (UPD) is an electrochemical phenomenon by which typically a metal cation (e.g., Cu 2+ , Ag + , Bi 3+ , and Pb 2+ ) is deposited onto a solid metal (for example, Au, Pd and Pt) at a potential more positive than its equilibrium potential (the potential at which it deposits onto itself) 1 . This phenomenon usually arises from the strong interaction between the depositing metal and the substrate: the metal-substrate interaction is energetically favorable compared to the metal-metal interaction in the crystal lattice of the bare metal. UPD can then be understood to be when a metal can deposit onto another material more easily than it can deposit onto itself 2 . Adzic et al. have reported various single-layer core-shell nanostructures through underpotentially depositing a foreign metal (e.g., Cu) onto a naked metal support (e.g., Au, Pd, and Pt), followed with metal exchange by the desired Pt-group metal [3][4][5][6] . However, single-layer Au@Pt core-shell structures are almost completely and rapidly deactivated because of the ease of surface rearrangements between the Pt islands and exposed Au substrate at the electrode-electrolyte interface. Inspired by these findings, we propose that the synthesis of ADMCs might be realized through SSED on a supporting substrate with isolated active sites for the UPD.

Supplementary Note 2. Estimation of Cu coverage
According to the calculation shown below, we can estimate the coverage of Cu atoms on the ce-MoS2 nanosheets ( ). The amount of Cu can be obtained by integration of the cathodic peak corresponding to Cu UPD ( Supplementary Fig. 4b). The value of 96500 is the Faraday constant in C mol -1 . The factor 1/2 reflects that two electrons are required to reduce one Cu 2+ ion to Cu (Cu 2+ + 2e − → Cu).
The amount of the ce-MoS2 nanosheets on GCE: The coverage of Cu atoms on the ce-MoS2 nanosheets is thus: The factor 1/2 reflects that one mole of MoS2 contains two moles of S atoms.
According to our experimental characterizations and DFT simulations, Pt atom is attached on the Mo top site through three Pt-S coordinate bond. The theoretical coverage of Cu on MoS2 should be 33.3%. However, owing to the existence of many S defects and localized structural distortion, the real coverage of Cu atoms (2.76%) on the ce-MoS2 nanosheets is far less than the theoretical one (33.3%), which indirectly confirms that SSED Cu on ce-MoS2 is truly less than a monolayer.

Supplementary Note 3. Pt SA loading amount
Owing to a local one-to-one galvanic exchange of Cu atoms, we can theoretically estimate the amount of Pt loaded onto Pt SA (nPt). This estimation is based on the stoichiometric conversion from Cu to Pt on the ce-MoS2 nanosheets.
Assuming that no ce-MoS2 nanosheets detach from the GCE and that the Faradic efficiency and the conversion efficiency from Cu to Pt are 100%, then: The molar amount of Pt is equal to that of Cu (Cuupd + Pt 2+ → Cu 2+ + Pt).
The theoretical loading amount of Pt on the ce-MoS2 nanosheets is thus: 6.25×10 −4 × 100% = 6.73 wt% (6) The experimental loading amount of Pt in Pt-SAs/MoS2 (approximately 5.1 wt% analyzed by ICP-OES), though slightly lower than the nominal loading (approximately 6.73 wt%), has been reasonably considered as a first approximation to the theoretical amount. The small discrepancy can be ascribed to many complicated experimental factors, such as metal loss during the preparation process, Faradic efficiency/conversion efficiency less than 100%, etc..

Supplementary Note 4. Deep understanding of the site-specific UPD process on TMD materials
We obtained cyclic voltammograms (CVs) of the process of UPD of Cu adatoms on the TMDs (MoS2, MoSe2, WS2 and WSe2), and this process is reversible and surface-dominate ( Supplementary Fig. 13). In principle, the anodic and cathodic currents will become nearly symmetric about the potential axis in the CVs if the sweep rate of the potential is sufficiently slow ( Supplementary Fig. 14). The shapes, positions, and number of the UPD peaks largely depend on the substrate and the property of the electrolyte. On a single-crystal metal electrode, a very dense monolayered metal film can be formed via underpotential deposition. Such atomby-atom metal film structure leads to the strong mutual interaction between metal atoms. In that case, the energy level of each deposited metal tends to be equal, and thus a very sharp current peak can be usually achieved.  Supplementary Fig. 13).
These CVs have a pronounced current maximum at a distinct potential Up. This means that the majority of adsorbed atoms are deposited at this potential with the least variation of adsorption energy as a function of coverage. This current peak therefore seems most suitable for characterizing the properties of the adatoms on different substrates 7 .
Since electrochemical measurements show only the relative energy values, we can relate the free energy of deposition onto the supporting substrate to the free energy of deposition onto the bulk crystal 8 . The chemical potentials ( ) of bulk metal (M bulk ) and supported single-atom metal (M SA ) can be obtained from the equilibrium conditions for reactions: with ̃M solv + + ̃e bulk − = bulk (8) and with ̃M solv + + ̃e SA − = SA (10) where electrochemical potential, ̃= + 0 .
Considering the equal activity of the metal ions in solution (equal electrolyte composition, equal ̃M solv + ), the difference in electron energies at the respective electrode potentials (∆ p = p − 0 ) arises from the difference in the free energy of bulk metal atoms (M) and the single metal atoms (S) adsorbed on a supporting substrate: where is the chemical potential in atomic units, 0 is the electronic charge, 0 is the equilibrium potential of the electrode M bulk , and p is the peak potential for adsorption on the supporting substrate.
We have tried to correlate the ∆ p value with physical parameters pertinent to the system that might semiquantitatively explain the energy gain for the energetically favored deposition (UPD) compared to bulk deposition. Because (in atomic units) is defined as the Gibbs free energy that can be absorbed/released owing to the change of a metal atom, one might assume that the binding energy (Δ BE ), which also reflects the energy required to dissociate a metal atom from the substrate atom, should have a positive correlation with : Thus,  Table   7).
The linear relationship connects the difference in the binding energy (metal-support) between a metal adatom bound to a substrate atom and one bound in bulk material with the difference in electron energies at the respective electrode potentials, in order to establish equilibrium conditions for the reactions (Supplementary Fig. 15b). The site-specific UPD of single-atom metals can be influenced by factors such as potential window, nature of substrate and adsorption of ions. The process can be reasoned as follows: (i) solvated metal ions move from the diffuse layer to the reaction zone, getting rid of the solvation sheath, and (ii) electron transfer from the substrate to the metal ions, leading to the subsequent formation of metalsubstrate bond ( Supplementary Fig. 15c). During this process, the formation of metal-substrate bond involves knocking off the adsorbed solvent dipoles from the deposition sites of the substrate. Note that the substrate is solvated, we should also consider the nature of the arrangement of solvent dipoles at the substrate surface ( Supplementary Fig. 15c).

Supplementary Note 5. Plausible mechanism for the interaction between copper ions and ce-MoS 2
When discussing the growth mechanism of UPD of Cu on ce-MoS2, we first exclude the possibility of functionalization by physisorption ( Supplementary Fig. 18). Instead, we postulate that surface S atoms of ce-MoS2 first coordinate the Lewis acidic metal prior to UPD ( Supplementary Fig. 19a As shown by XPS analysis (Supplementary Fig. 19d), a characteristic peak for Cu 2p appears, which can be unambiguously ascribed to the adsorption of Cu 2+ ions on the ce-MoS2 nanosheets 9 . The binding energies (953.5 and 933.8 eV) for the adsorbed Cu ions are located in the range between Cu 0 and Cu II , 10,11 suggesting partial electron donation from the lone pair electrons of the S atom. As a result of electronic interaction with the Lewis acidic metal atoms, the surface S atom donors become more electropositive, resulting in a slightly broader band and higher photoelectron emission energy ( Supplementary Fig. 19c), which in turn confirms this effect. In the UV-visible spectrum, the peak in the near-UV region corresponding to ce-MoS2 is also redshifted ( Supplementary Fig. 19e). These changes are consistent with the dominant mechanism as proposed above, in which the formal positive charge on the S atoms facilitate the electron extraction.
Then, an applied potential served as an electron donor to reduce in situ the Cu 2+ ions to Cu 0 adatoms on MoS2. Owing to the driving force of UPD that is determined by the specific affinity between metal adatoms and the substrate atoms, the reaction was terminated, and no additional Cu 2+ ions were able to bind to MoS2 because the interfacial S atoms had fully reacted, leading to the construction of a stable single-atom model (identified by DFT calculations, Supplementary Table 6, Supplementary Fig. 19b).

Supplementary Note 6. HER mechanism
It is well recognized that there are two possible mechanisms for the HER in acidic solution (Supplementary Fig. 22) 12 , both starting with reductive proton adsorption (S-14). The Volmer-Heyrovsky mechanism (S-14 and S-15) follows that with reduction of a second proton at the same site and release of a hydrogen molecule (two-electron process). In the Volmer-Tafel mechanism (S-14 and S-16), proton adsorption is followed immediately by the surface combination of two adsorbed hydrogen atoms and the release of a hydrogen molecule (oneelectron process).

Supplementary Note 8. Proposed HER mechanism on Pt-SAs/MoS 2
The Tafel slope is an inherent property of electrocatalytic materials and is a useful indicator of the rate-limiting step for reactions involving electron transfer. The Pt-SAs/MoS2 catalyst yields a Tafel slope of 31 mV dec −1 , which is close to the value of commercial Pt/C (32 mV dec −1 , theoretical value: 30 mV dec −1 ). The seemingly identical Tafel behavior of commercial Pt/C and Pt-SAs/MoS2 is not necessarily an indicator that the exact same HER pathway is followed.
Our experimental and theoretical data (XANES, XPS, and Bader charges analysis) have shown that single Pt atoms are positively charged and that the S atom obtains the electron, leading to a much higher total unoccupied density of Pt 5d states. During H chemisorption, the 5d orbitals of the Pt atoms interact strongly with the 1s orbital of the H atoms, leading to electron pairing and hydride formation 13,14 . In this case, we reason that atomic-scale tailoring should unconventionally modulate the adsorption state of hydrogen atoms on the Pt single atom of Pt- To demonstrate the underlying type of mechanism dominant in the Pt-SAs/MoS2 system, we further performed a computational study on single-Pt-atom catalysts to gain more detailed insights into the HER process ( Supplementary Fig. 29d Although a systematic experimental kinetic investigation is beyond the scope of the current work, the likely HER mechanism of hydrogen recombination and desorption for single-atom Pt on the ce-MoS2 is proposed using DFT calculations, with an attempt to obtain insights into the kinetics of HER.

Supplementary Note 9. Comparison between the previously reported electrodeposition method and SSED for ADMC synthesis
Electrodeposition offers a facile, controllable and room-temperature method for reducing metal ions into their elemental states. In this case, an efficient electrodeposition method has been widely used for single-atom synthesis in recent years [16][17][18][19][20] . The cathodic deposition of singleatom metals on a working electrode can be achieved by anodic dissolution of a bulk metal foil electrode as a counter electrode under acidic conditions using a three-electrode configuration ( Supplementary Fig. 30a). However, the potential cycling process is usually completed within hours (at least 10 h) and is less controllable, which results in the formation of nanoclusters or nanoparticles at longer cycling times.
Recently, Zeng's group have reported a universal and rapid electrodeposition approach for the fabrication of single-atom metals 21 . The depositions can be both cathodically and anodically conducted (C, A-ED) for synthesis of single-atom metals with distinct electronic states, which holds great promises for various catalytic reactions. They proposed that the electrodeposition process resembles the molecular nucleation mechanism. The upper limit of mass loading for single-atom metals cannot exceed the minimum supersaturation level, otherwise single-atom metal tends to nucleate ( Supplementary Fig. 30b). Thus, single-atom synthesis can be realized by controlling the metal precursor concentration and deposition time.
In our work, we developed an "intelligent" methodology for the single-atom growth by sitespecific electrodeposition which is capable of automatically terminating the aggregation of metal atoms ( Supplementary Fig. 30c). Such intrinsically self-terminating effect distinguishes our site-specific UPD method from other previously reported electrodeposition method for single-atom synthesis. We show that the site-specific UPD method can be used to produce high-loading single-atom metals without the consideration of the high metal precursors or long deposition time (Supplementary Table 10), which might seriously cause atom aggregation in other electrodeposition methods. After the formation of thermodynamically favorable metalsupport bonds, the sequential formation of metal-metal bond is forbidden at the UPD potential, restricting it to the single-atom metal. In our design, two requisite factors should be considered: (1) identifying electrically conductive substrate materials, which consist of isolated active sites that possess lone pair electrons and suitable electronegativity for the UPD of single atoms; (2) choosing suitable applied electrodeposition potential restricted to UPD region, at which metalsupport bonding predominates over metallic bonding. Our site-specific UPD method for single-atom synthesis can be rapidly completed on a timescale of seconds to minutes. Additionally, we confirm that single-atom metals are straddled atop Mo by coordinating with three nearest neighboring S of Mo (see HADDF-STEM image, EXAFS data, and DFT simulations), which shows distinct from depositing site of vacancies/edges/defects into the lattice reported by previous works [16][17][18][19][20][21] . However, during the following Tafel reaction, for traditional Pt nanocrystals, the two protons bind to two adjacent Pt atoms, combine, and generate a H2 molecule ( Supplementary Fig. 29a), whereas for Pt SAs, the two protons bind to a single Pt atom, combine and produce H2

Supplementary
(Supplementary Fig. 29b; step III and IV in Supplementary Fig. 29c To remove the residual charge on the ce-MoS2 nanosheets, we carried out an iodine treatment experiment by immersing ce-MoS2 in 0.15 M iodine in acetonitrile 27 . After this treatment, the zeta potential of ce-MoS2 increased significantly (-23 mV) owing to the suppression of charge by mild oxidation 27 . We found that the UPD of Cu atoms is also possible on iodine-treated ce-MoS2 ( Supplementary Fig. 18), which strongly suggests that the S atoms with a lone electron pair are the active sites for the UPD process.

Representation
* The red rows represent the optimal structures and data for further DFT simulation. The initial position of Pt atoms at Mo1/S1 in Pt-SAs/MoS2, W1/S1 in Pt-SAs/WS2, and Mo1/Se2 in Pt-SAs/MoSe2 were stabilized to the same structure after geometry optimization.