Main

In recent energy research, a significant focus has been placed on understanding the mechanism of catalytic reactions at the atomic level. The direct operando monitoring of surface catalytic reactions has always been a ‘holy grail’ in electrochemistry and heterogeneous catalysis, and will aid considerably in the design and development of more highly efficient catalysts1,2. As a classical catalytic reaction, the process and mechanism of the oxygen reduction reaction (ORR) at platinum surfaces have been a focus of attention in the literature for a long time3,4. Although many research groups have carried out experimental and theoretical studies to reveal the ORR mechanism, the detailed surface process is still not clear.

Generally, the mechanism of the ORR process at platinum electrodes in acidic conditions is considered to occur by two main pathways: one involves oxygen being reduced directly via a four-electron pathway into H2O; the other first reacts oxygen via a two-electron pathway to hydrogen peroxide, followed by a two-electron transfer reduction of the latter to water; hydrogen peroxide can also directly diffuse into the solution as a final product, which then quickly decomposes. However, some essential questions and uncertainties remain about the ORR process, including slow kinetics, the origin of observed high overpotentials, and the rate-determining step5,6,7,8,9,10,11. The main reason is that, as a multi-electron reaction, there are a variety of intermediates (for example, OH*, O22−, O2−, HO2*) that are generated during the ORR process, and most of the intermediates have a short lifetime, low coverage and are also influenced by other co-adsorbed species. Thus, the key factor to unravel the ORR mechanism is to develop an in situ method to identify the various reaction intermediates and their adsorbed configurations at platinum surfaces during the ORR process. With their well-defined surface structures, optical and electric field properties, and the ability to be modelled at the atomic level, single-crystal surfaces play a key role in probing catalytic reaction mechanisms in surface science12. However, most current spectroscopic methods are not suitable for single-crystal studies in aqueous solution, especially for the ORR reaction at Pt(hkl) electrode surfaces13,14,15,16,17,18,19.

Surface-enhanced Raman scattering (SERS) is a powerful fingerprint spectroscopy that can be used for in situ investigation of trace chemical species and identification with single-molecule sensitivity20,21,22. However, its applications are generally restricted to ‘free-electron-like’ metals such as Au, Ag and Cu that have non-smooth surfaces. To overcome the long-term limitation of SERS on morphology and material generality, we previously developed a surface vibrational spectroscopic method that was named shell-isolated nanoparticle-enhanced Raman spectroscopy (SHINERS)23. In SHINERS, an ultrathin and uniform silica shell coated onto a gold nanoparticle can efficiently enhance the Raman signal of molecules that are located near the nanoparticle surface without any interference. It is possible to obtain Raman signals from any substrate and any material surface. A unique advantage of SHINERS is its particular applicability to explore the adsorption configuration and catalytic processes of probe molecules at single-crystal surfaces24,25,26,27,28,29,30,31,32.

Here, we employ in situ electrochemical (EC)-SHINERS coupled with density functional theory (DFT) calculations to study the ORR process at Pt(hkl) electrode surfaces. We obtain direct spectral evidence that allows the ORR mechanism at these surfaces to be elucidated at a molecular and atomic level.

SHINERS enhancement at Pt(hkl) surfaces

For a clear understanding of the relationship between the shell-isolated nanoparticles (SHINs) enhancement and the electric field distribution, a 2 × 2 Au@SiO2 nanoparticles (NPs) array was modelled on a perfectly smooth platinum substrate surface and simulated using a 3D-finite-difference time-domain (3D-FDTD) theoretical system. Figure 1a shows the schematic diagram of in situ EC-SHINERS at low-index Pt(hkl) surfaces. The SHINs used in this experiment had a gold nanoparticle core (~55 nm) with a SiO2 shell (~2 nm) (Fig. 1b and Supplementary Fig. 1), with the coverage of SHINs at the Pt(hkl) electrode surface at around 30% (Fig. 1c). The 3D-FDTD technique has been employed to model the SHINERS system effectively33,34,35,36. The hot spots are mainly located around the particle–surface junctions under 638 nm excitation (Fig. 1d), and the average SERS enhancement factor of this configuration is about 1.0 × 105 on the Pt(hkl) surface24.

Fig. 1: Schematic illustration of the SHINERS study of the ORR process and correlated characterization and 3D-FDTD results at Pt(hkl) surfaces.
figure 1

a, Model of shell-isolated nanoparticles (Au@SiO2 NPs, SHINs) at a Pt(111) surface and the mechanism of the ORR process revealed by the EC-SHINERS method. The silver-white, red and white spheres represent Pt, O and H atoms, respectively. The large golden spheres with transparent shells represent SHINs. The SHINs, when being excited by a laser, can generate strong electromagnetic fields to enhance the Raman signals of molecules adsorbed at the Pt(hkl) single-crystal surface. b, Transmission electron microscope (TEM) image of Au@SiO2 nanoparticles. c, Scanning electron microscope (SEM) image of a Pt(111) single-crystal electrode surface modified with SHINs. d, 3D-FDTD simulations of four SHINs NPs with a model of a 2 × 2 array on a Pt substrate. E and E0 represent the localized electric field and the incident electric field, respectively.

Full size image

ORR processes at Pt(hkl) surfaces in acidic conditions

First, we obtained the polarization curves of the ORR process at three Pt(hkl) rotating disk electrode (RDE) surfaces in a 0.1 M HClO4 electrolyte solution saturated with O2; the rotation rate was 1,600 r.p.m.. We found that the ORR activity of Pt(hkl) decreased in the sequence (111) > (110) > (100) in the HClO4 solution (Fig. 2a). The O2 reduction current begins around 1.0 V and then quickly achieves its limiting diffusion current around 0.7 V when following the negative sweep direction (Fig. 2a and Supplementary Fig. 5). In the limiting diffusion potential range (0.3 V ~ 0.7 V), the ORR activity of Pt(hkl) is completely controlled by the mass transfer of oxygen. When the potential decreases below 0.3 V, the limiting diffusion current begins to decrease due to hydrogen adsorption at the Pt(hkl) electrode surface. The adsorbed hydrogen increases the difficulty of breaking the O–O bond of oxygen molecules, as there are not enough adjacent vacancies for O2 adsorption at the Pt(hkl) surfaces. As a result, the oxygen molecules will partly form H2O2(ref. 37). As can be observed in Fig. 2a, the onset potentials of the ORR for Pt(111) and Pt(110) are similar, whereas the onset potential is much lower for Pt(100). In previous works it was observed that the ORR activity of the surfaces containing both (111) terraces and (110) steps would increase with the increase of the number of (110), therefore indicating that Pt(110) is more active than Pt(111)38. In this work this difference is less noticeable since negatively directed sweeps from 1.0 V are shown (to compare them with the spectroscopic results), and Pt(110) initially presents PtO species (Supplementary Note 1 and Supplementary Fig. 2). Therefore, in the negatively directed sweeps the activity of Pt(110) is partially inhibited by the presence of these surface oxides. In the positively directed sweeps the surface oxides are not present, and the previously reported activity is recovered. To explore the ORR mechanism at Pt(hkl) surfaces, the in situ EC-SHINERS method was employed to evaluate the ORR system in a 0.1 M HClO4 solution. Since the electrode in the Raman cell is not able to rotate during the ORR experiment, the limiting ORR diffusion current is slightly different compared to the RDE system. Nevertheless, the starting potential and the potential range of the ORR without rotation were almost the same as the RDE system (Supplementary Note 4 and Supplementary Figs. 5,6).

Fig. 2: Electrochemical results of the ORR process at Pt(hkl) surfaces in acidic conditions and correlated EC-SHINERS and DFT results of the ORR at a Pt(111) surface.
figure 2

a, Polarization curves of the ORR process at three Pt(hkl) RDEs in oxygen-saturated 0.1 M HClO4 solutions; the rotation rate was 1,600 r.p.m. and the scan rate was 50 mV s−1. j and E represent the current density and potential, respectively. b, EC-SHINERS spectra of the ORR system at a Pt(111) electrode surface in a 0.1 M HClO4 solution saturated with O2. c, Normalized EC-SHINERS intensities of the stretching mode of O–OH around 732 cm–1 at different potentials. The polarization curve of the ORR process at the Pt(111) surface in a 0.1 M HClO4 solution saturated with O2, the rotation rate was 1,600 r.p.m. and the scan rate was 50 mV s−1. The arrows in panels a, b and c represent the potential scanning direction, and all the potentials are relative to the reversible hydrogen electrode (RHE). d, Side-view illustrations of HO2* at different stable adsorption configurations at a Pt(111) surface on top site and bridge site adsorption structures. The silver-grey, red and white spheres represent Pt, O and H atoms, respectively.

Full size image

The EC-SHINERS spectra of the ORR at Pt(111) electrodes were obtained over the potential range from 1.1 V to 0.5 V. During the negative potential excursion, there was no observable Raman signal in the range from 400 to 1,200 cm−1 until 0.8 V, except for the peak at 933 cm−1 (Fig. 2b). The peak at 933 cm−1 was attributed to the symmetric stretching mode of the perchlorate ion, \({\rm{\nu}}_{{\rm{s}}({\rm{CIO}}4^{-})}\). As the potential decreased, another obvious Raman band around 732 cm−1 in the acidic solution appeared when the potential reached 0.8 V, which further increased as the potential reached 0.6 V (Fig. 2c). Furthermore, a deuterium isotopic substitution measurement was carried out (Supplementary Note 5 and Supplementary Fig. 7). In the deuterium isotopic experiment, the peaks around 732 cm−1 were shifted to lower wavenumbers of around 705 cm−1, which implied that the intermediates should be correlated with a ‘H’ atom. Although the first candidate considered for 732 cm−1 involves the O–O stretching vibration of H2O2, this molecule is unlikely to be stable at the Pt(111) surface, and would be immediately oxidized or reduced further to oxygen or water. According to electrochemical results39,40, we can confirm that the peak around 732 cm−1 belongs to the O–O stretching vibration of adsorbed HO2* on Pt(111), which can also be considered as an important intermediate species of the ORR process40. The DFT method was also employed to calculate the vibrational frequencies of HO2* species at Pt(111) (Supplementary Note 12 and Supplementary Fig. 21). From the DFT results, we found that there were two different stable adsorption configurations of HO2* at the Pt(111) surface, on top (t-b) site and bridge (b-b) site (Fig. 2d) (the Pt–O distances were 2.008 Å and 2.020 Å, respectively) adsorption structures, and the correlated Raman frequencies of the O–O stretching vibration for these two different structures were 839 cm−1 and 726 cm−1, respectively. This means that the peaks around 732 cm−1 in our experiment can be assigned to the O–O stretching vibration of the b-b adsorption structure of HO2*.

In general, for the ORR process at the Pt(111) surface in an acidic solution, at high potentials, the oxygen reduces to water through particular intermediates, and at low potentials the oxygen reduction to water is inhibited and stops at the peroxide stage (perhaps involving the same intermediates). To further understand the ORR mechanism at the Pt(111) surface in acidic conditions, we tried to look at the ORR process from a higher to a lower potential range compared to the range from 1.1 V to 0.5 V at Pt(111) (Supplementary Note 6 and Supplementary Fig. 8a). At a potential of 1.2 V, there was an obvious Raman peak at around 571 cm−1, the frequency of which shifted to lower wavenumbers until at 0.9 V it was 567 cm−1. This peak was attributed to the Pt–O stretching mode41. Meanwhile, the peak of HO2* at 732 cm−1 was observed when the potential reached 0.75 V, and its intensity increased until 0.6 V, after which it remained stable following a further potential decrease. Finally, the 732 cm−1 peak intensity decreased when the potential was below 0.4 V, which correlated well with the ORR current results (Supplementary Fig. 8b).

According to electrochemical research, Pt(111) is saturated with about a third of a monolayer of OH at 0.8 V, and then decreases to zero at the upper end of the double-layer region42,43. However, we did not find the OH species at approximately 0.8 V during the ORR process at the Pt(111) surface using SHINERS. Recent work44 shows that the O–H bond is almost parallel with the Pt(111) surface at 0.8 V, whereas the SHINERS method requires a vibrational dipole component normal to the surface to undergo light absorption. This special structure of OH at the Pt(111) surface may be the reason why we do not detect the OH adsorption at Pt(111) using SHINERS around 0.8 V. For the ORR process, in the kinetic potential region, there will be a low concentration of available sites for forming adsorbed HO2*, rendering the signal too weak to be observed. Following the potential sweeps in the negative direction, the coverage of OH drops rapidly, opening up more sites for O2 to approach and form adsorbed HO2*. Thus, the SHINERS spectra clearly show the formation of adsorbed HO2* at 0.8 V. On reaching about 0.6 V, there is no further adsorbed OH and the potential continues through the double-layer region until it is slightly less than 0.4 V, when underpotential deposited H starts to block sites. The dependence on potential of the current density, in Supplementary Fig. 8, illustrates that shortly after entering the diffusion-limited region and double-layer region, there is surface congestion with adsorbed HO2* intermediates.

In addition, we have carefully compared the electrochemical behaviours with and without SHINs on Pt(111), Pt(100) and Pt(110) single-crystal surfaces during the ORR process (Supplementary Notes 2,3, Supplementary Figs. 3,4,9). From the experimental results, we find that the SHINs only affect the spectroscopy, but not the activity of Pt(hkl) surfaces for the ORR reaction. From the theoretical and experimental results, we can confirm that SHINERS method can identify different adsorption structures of surface adsorbate species. Combined with the electrochemical result39, we assert that the HO2* species is an important intermediate of the ORR process under acidic conditions. Moreover, we will further discuss the ORR mechanism at Pt(hkl) surfaces in the section ‘Mechanism of the ORR process at Pt(hkl) surfaces’.

The crystallographic orientation and the surface structure of the single-crystal electrode surface will greatly influence the reaction mechanism and reaction kinetics. Moreover, the ORR activity is also highly sensitive to the surface structure of the Pt(hkl) electrode. We therefore investigated the ORR processes at the other two low-index Pt(hkl) surfaces (that is, Pt(110) and Pt(100)) in a 0.1 M HClO4 solution saturated with O2. Interestingly, we observed different phenomenon at the three low-index Pt(hkl) surfaces from the SHINERS experimental results. There were two Raman peaks, around 1,030 cm−1 and 1,080 cm−1, that appeared at Pt(100) following a decrease in potential, and the phenomenon at the Pt(110) surface was similar to that with Pt(100) (Fig. 3), but their relative Raman intensities and onset potentials were different. At the same time, the deuterium isotopic substitution measurement was taken into consideration as before (Supplementary Note 7 and Supplementary Fig. 10). We did not observe any obvious shift in the 1,030 cm−1 feature, but the peak at 1,080 cm−1 shifted to a lower wavenumber around 717 cm−1 during the D2O experiment. Meanwhile, in the 18O2 isotopic substitution experiment at the Pt(110) surface (Supplementary Note 8 and Supplementary Fig. 11), we found that the peak around 1,030 cm−1 did not show any obvious shift, whereas the peak around 1,080 cm−1 was shifted to a lower wavenumber around 1,072 cm−1, which further implied that the intermediates of 1,080 cm−1 correlated with oxygen-related species. According to the literature14,45, the peak around 1,030 cm−1 can be assigned to the symmetric stretching vibrational mode of ClO3 in the HClO4 molecule, and the band around 1,080 cm−1 can be assigned to the Pt–OH bending mode δPtOH of OH* (Fig. 3).

Fig. 3: In situ EC-SHINERS results of the ORR at Pt(100) and Pt(110) surfaces in acidic conditions and DFT result of OH* at a Pt(110) surface.
figure 3

a, EC-SHINERS spectra of the ORR at a Pt(100) electrode surface in a 0.1 M HClO4 solution. b, EC-SHINERS spectra of the ORR at a Pt(110) electrode surface in a 0.1 M HClO4 solution. The arrows in panels a and b represent the potential scanning direction, and all the potentials are relative to the RHE. c, Side-view illustrations of OH* and O* at a Pt(110) surface. The silver-grey, red and white spheres represent Pt, O and H atoms, respectively.

Full size image

The DFT-calculated results showed that if only OH* adsorbed at the Pt(110) surface (Supplementary Note 13 and Supplementary Fig. 22), the Pt–OH bending would appear at 875 cm−1. But if OH* adsorbed at an atop site with an atomic oxygen on the nearest neighbour, the adsorbed O* atom plays a constructive role in bending the H atom (Fig.3c). In this case, the Pt–OH bending vibration increases to 1,078 cm−1, which correlates well with the experimental and reference results. From DFT calculation results, we also found that the HO2* species were not so stable at Pt(110) and Pt(100) surfaces, being easily dissociated to Pt–O and Pt–OH because of the lower coordination number of Pt in Pt(110) and Pt(100). On the other hand, the different ORR activity of Pt(100) and Pt(110) compared to Pt(111) may be due to the presence of OH* on the surfaces, which can block the active sites of platinum surfaces.

ORR processes at Pt(hkl) surfaces in alkaline conditions

Under electrochemical conditions, the interfacial state of the Pt(hkl) electrode surface should undergo some changes following the increase in the pH value. For example, the charge distribution and the adsorption state at the interface will change substantially. Since the ORR process is a typical electrode reaction to consume protons and generate OH*, the interface pH values will be changed as the reaction proceeds. During this process, the ORR reaction pathway, intermediates and their surface coverage rate at the electrode surface will be changed. Therefore, it is necessary to study the ORR mechanism at different pH values, which will give us more important information about the relationship between the interface structures and the reaction mechanism15,17,46. We investigated the ORR process at a Pt(110) electrode surface in an alkaline solution, similarly to the acidic conditions experiment (Fig. 4a, Supplementary Notes 9,10 and Supplementary Figs. 13–16). There was a broad Raman band around 1,150 cm−1 that appeared when the potential was decreased to 0.65 V. With a further decrease in the potential, this peak became stronger until 0.35 V, and then decreased.

Fig. 4: EC-SHINERS study of the ORR at Pt(hkl) surfaces in alkaline conditions.
figure 4

a–c, EC-SHINERS spectra of ORR at Pt(110) (a), Pt(111) (b) and Pt(100) (c) surfaces in a 0.1 M NaClO4 solution (pH of approximately 10.3) saturated with O2. The arrows represent the potential scanning direction, and all the potentials are relative to the RHE.

Full size image

To further investigate the effect of crystallographic orientation, comparative experiments were conducted at two other low-index Pt(hkl) single-crystal surfaces, Pt(111) and Pt(100), under identical conditions (0.1 M NaClO4 in H2O with a pH of approximately 10.3). As Fig. 4b, c shows, there was an almost identical phenomenon to Fig. 4a, with just a slight difference in the starting potentials of the peaks around 1,150 cm−1. This behaviour indicated that there should be the same intermediate species at the three low-index Pt(hkl) surfaces during the ORR process in alkaline conditions. A deuterium isotopic substitution measurement was also carried out, and we found that the peaks around 1,150 cm −1 did not exhibit obvious shifts in alkaline conditions (Supplementary Figs. 17,18). Thus, the intermediate species around 1,150 cm−1 should be without ‘H’. Furthermore, the peak around 1,150 cm−1 was obviously shifted to a lower wavenumber (around 1,120 cm−1) in an 18O2 isotopic substitution experiment at a Pt(111) surface (Supplementary Note 11 and Supplementary Fig. 19), which confirmed that the intermediates around 1,150 cm−1 could be attributed to oxygen-related species. In the previous studies, researchers had found that the characteristic Raman peak of the superoxide ion was around 1,150 cm−1 in the alkaline solution during the ORR process15. Our DFT calculation also proved that the peak around 1,150 cm−1 can be assigned to the O–O stretching vibration of the superoxide ion O2− (Supplementary Tables 1,2, Supplementary Note 14 and Supplementary Fig. 24). In the calculation results, the adsorption of O2− with the t-b site at Pt(110), Pt(100) and Pt(111), and their Raman frequencies were located at 1,162 cm−1, 1,177 cm−1 and 1,182 cm−1, respectively, which correlate very well with our experimental results. From the above information, we confirm that the superoxide species has been identified as an important intermediate of the ORR reaction at Pt(hkl) surfaces in our research system.

An important challenge for the ORR is that the reaction intermediates have a short lifetime and are thus difficult to detect. Consequently, large concentrations of these intermediates are necessary for their spectroscopic observation. Such conditions are not accomplished until sufficiently low potentials are reached, when the reaction is fast enough to form large amounts of such intermediates, commensurate with their rate of consumption. It is not unusual that this situation should coincide with when the reaction starts to be diffusion-controlled, indicating that the reaction is very fast and therefore the mass transport limits the reaction. However, it should be mentioned that in the experiments reported in the present study, no obvious ORR intermediate species are observed in the kinetic region. Evidently, more experimental evidence and theoretical calculations to categorically show that the detailed ORR mechanism found in the diffusion-controlled region maps directly onto the kinetic region will be required in future works.

Mechanism of the ORR process at Pt(hkl) surfaces

Based on EC-SHINERS experiments and theoretical calculations, taking into consideration previous research, the mechanism of ORR at the Pt(hkl) electrode surface in a 0.1 M HClO4 solution can be explained as follows: after being adsorbed at the Pt(hkl) electrode surface, O2- formed HO2* via a proton and an electron transfer, which then quickly dissociated to form a pair of OH* and O* on the neighbouring Pt atoms. The OH* species further combined with ‘H’ to generate H2O. The detailed schematic diagram of the ORR mechanism at a Pt(hkl) surface is shown in Fig. 5a,b (Supplementary Fig. 25). However, because of the different Gibbs free energies and dissociation barriers of the same intermediates on different crystallographic planes, there are differences between the Pt(111) facet and the Pt(110) and Pt(100) facets (Supplementary Tables 3–6). (Based on the previous reports, the dissociation barrier for HO2* on Pt(111) is about 0.59 eV higher than that on Pt(100).)47,48,49 During the ORR process, after protonation to form HO2* at the Pt(111) surface, the adsorbed HO2* species is stable and needs a higher activation energy to proceed to the next step (Fig. 5a). However, the instability of adsorbed HO2* at Pt(110) and Pt(100) surfaces, coupled simultaneously with a proton and an electron transfer process, leads to the O–O bond of HO2* quickly breaking, and forming a pair of OH* and O* on the neighbouring Pt atoms in acidic conditions (Fig. 5b). This should be correlated with the structure–activity relationship of different surfaces and the rate-determining step of the ORR. It is also in good agreement with previous references9,10,47,48,49,50.

Fig. 5: The proposed mechanism of the ORR at Pt(hkl) surfaces in a 0.1 M HClO4 solution and relevant Gibbs free energy (eV) of different intermediates at Pt(hkl) surfaces.
figure 5

a,b, Free energies are given relative to gas-phase H2 and O2 and a metal surface to simulate the reaction of H2 + O2 → H2O on the Pt surface. Different intermediates at Pt(111) (a) and Pt(100) (b) surfaces. The white, red and blue spheres represent Pt, O and H, respectively. The main difference between a and b is the fourth step. For a, it is OH* + O* + 3(H+ + e−) → H2O* + O* + 2(H+ + e−), whereas for b, it is OH* + O* + 3(H++ e−) → OH* + OH* + 2(H++ e−).

Full size image

Meanwhile, one idea that we have to have in mind is that, despite the fact that adsorbed species are detected or included in the calculations, solution species may exist and play a role. In general the detected adsorption processes such as those shown in equation (1) are fast and reversible:

$${\mathrm{Pt}}\,{\mathrm{ + }}\,{\mathrm{H}}_{\mathrm{2}}{\mathrm{O}} \rightleftharpoons {\mathrm{PtOH}}\,{\mathrm{ + }}\,{\mathrm{H}}^{\mathrm{ + }} + {\mathrm{e}}^ -$$
(1)

This may also happen with intermediates such as HO2*, which can go to the solution side40, interact strongly (H bonding) with the water layer and, eventually, become re-adsorbed before further reaction. In this respect the ORR, once started, would involve intermediates that combine hydrogen and oxygen atoms, which should be high-mobility species in aqueous solution.

Conclusions

In this work, we employed the in situ EC-SHINERS method to systematically investigate the ORR process at Pt(hkl) single-crystal surfaces and obtained direct spectral evidence of OH*, HO2* and O2−. We found during the ORR process, the adsorbed HO2* is stable at the Pt(111) surface, but there is just adsorbed OH* at the Pt(110) and Pt(100) surfaces. The steps to form HO2* and OH* species at the Pt(hkl) surfaces will directly affect the ORR activity of different single-crystal surfaces. Meanwhile, in alkaline conditions, there only O2− species were found on the three single-crystal surfaces. We therefore conclude that the protonation process noticeably affects the ORR activity and its mechanism. Combined with the theoretical calculation results and previous research, we further explained the ORR mechanism at the Pt(hkl) surface in acidic conditions, and produced a reasonable interpretation and inference from the EC-SHINERS measurements.

Methods

Reagents

Sodium citrate (99.0%), chloroauric acid (99.99%), sodium perchlorate (98.0% ~ 102.0%) and (3-aminopropyl)trimethoxysilane(APTMS) (97%) were purchased from Alfa Aesar; sodium hydroxide (97%, GR) and perchloric acid (70% ~ 72%, GR) were purchased from Sinopharm Chemical Reagent Co. Ltd.; sodium silicate solution (27% SiO2) was purchased from Sigma-Aldrich. Deuterium oxide (for NMR 99.8 atom % D) was purchased from ARMAR AG. All chemicals were used as received without further purification. Argon (99.999%), hydrogen (99.999%) and oxygen (99.999%) were purchased from Linde Gas. 18O2 (99.8%) was purchased from LION Biology Company. Milli-Q water (~18.2 MΩ·cm) was used throughout the study.

Equipment

High-resolution TEM (JEOL, cat. no. JEM 2100 EX) and SEM (HITACHI S-4800) were used to characterize the morphology of SHINs and single-crystal surfaces. The electrode potential was controlled with an Autolab PGSTAT30 (Metrohm).

Synthesis of SHINs

We consider 55 nm Au@ 2 nm SiO2 SHINs as an example to introduce the detailed preparation process23. The 55 nm Au NPs were prepared according to the Frens method. First, 200 ml of 0.01% HAuCl4 solution was placed in a 500 ml round-bottom flask and heated to boiling under stirring. Following that, 1.4 ml of 1% sodium citrate solution was quickly added into the above solution and the reaction continued for 40 min; then the solution was allowed to cool down to ambient conditions for the next step in preparing SHINs. SHINs were synthesized as follows: 30 ml of 55 nm Au NPs solution was added to a round-bottom flask under stirring without heat, and then 0.4 ml (3-Aminopropyl)trimethoxysilane (APTMS) (1 mM) was added. After 15 min reaction at room temperature under stirring, 3.2 ml of 0.54% Na2SiO3 solution (the pH was about 10.3) was added to the above solution. After 3 min, the mixed sample was transferred to a 98 °C bath and stirred for 20 min. Then the solution was quickly cooled down in an ice bath and centrifuged three times. Finally, the concentrated SHINs were diluted with pure water for further measurements.

Electrochemistry

The single-crystal electrodes were Clavilier-type Pt(hkl) electrodes (the diameter is ~2 mm). Before the experiment, the Pt(hkl) electrodes were annealed in a butane flame and cooled down in an Ar + H2 atmosphere. Electrochemical tests were conducted in a three-compartment glass cell with a Pt wire as the counter electrode and an RHE reference electrode (all potentials are reported with respect to the RHE electrode in this paper). All solutions in the electrochemical ORR measurements were saturated with oxygen. Electrochemical measurements were carried out with an Autolab PGSTAT30 (Metrohm) and the ORR electrochemical experiments were carried out in a hanging meniscus rotating disk electrode (HMRDE) configuration system, using a radiometer (EDI-101). The pH value of the 0.1 M NaClO4 electrolyte was adjusted by means of a NaOH solution.

The cleaning process for the SHINs on Pt(hkl) surfaces

The Pt(hkl) electrode (modified with SHINs) was placed in an electrochemical cell filled with a 0.1 M NaClO4 solution (pH~9), and polarized at −1.2 V (versus saturated calomel electrode; SCE) for about 1–2 min (the generation of tiny hydrogen gas bubbles could be observed). The hydrogen evolution reaction (HER) proceeded vigorously, and the impurities adsorbed on the electrode or SHINs surface were desorbed and diffused into the solution. The electrode surface was washed carefully and the solution was changed. These processes was repeated 3–5 times. Finally, the electrode was transferred to another clean electrochemical cell or Raman cell for CV or in situ Raman tests.

In situ EC-SHINERS

Raman spectra were recorded with an Xplora confocal microprobe Raman system (HORIBA JobinYvon). A 50× long working distance (8 mm) objective was used. The wavelength of the excitation laser was 637.8 nm from a He-Ne laser (the power was about 6 mW). Raman frequencies were calibrated using a Si wafer and ClO4− solution spectra. Each Raman spectrum shown here was acquired over a collection time of 120 s and is the average of two measurements.

3D-FDTD numerical method

The 3D-FDTD method was used to study the electromagnetic field enhancement. The fundamental principle of the FDTD method can found in the literature33. The FDTD method has been widely used to investigate optical properties, such as light scattering, absorption and electromagnetic field distributions. In the simulation, a perfectly matched layer was used. The simulation time was set to 1,000 fs, which was enough to ensure convergence of the calculation. We adopted a non-uniform mesh size in the junctions of the investigated structures. In detail, the Yee cell size in the junctions of the particle–particle and particle–Pt film was 0.25 nm × 0.25 nm × 0.25 nm and for the remaining regions was 0.5 nm × 0.5 nm × 0.5 nm. The dielectric functions of Pt and Au, which are wavelength-dependent, were taken from a multi-coefficient fitting model offered by Lumerical FDTD.

Computational details

All theoretical simulations were performed using the Perdew–Burke–Ernzerhof (PBE) functional of the generalized gradient approximation (GGA)51 to simulate the periodic boundary condition (PBC) model implemented in the Vienna ab initio simulation package (VASP)52. The projector-augmented wave (PAW) method was applied to describe the electron–ion interactions. A plane-wave basis cutoff of 400 eV was used for the wavefunctions and energies were converged to 10–5 eV. The Paxton and Methfessel method with a broadening factor of 0.1 eV was used, and a Γ-centered k-point sampling grid of 12 × 12 × 12 was adopted for the primitive cell calculation. A Γ-centered k-point sampling grid of 6 × 6 × 1 was adopted for all single-crystal facets concerned in this work. Vibrational frequencies of adsorbed molecules on the surface metal were calculated with density-functional perturbation theory (DFPT). The bottom two layers of the five-layer 2 × 2 Pt surface are fixed, while the top three layers are relaxed in all calculations. In addition, a spin-polarized calculation was performed in the geometry optimization. For the DFT calculation of O2−, the electron in our system is simulated by adding one electron, while charge neutrality is maintained by a compensating uniform charge background53. The calculated lattice constant of Pt was 3.977 Å, which agrees with the experimental value of 3.909 Å, and vacuum spaces of 15 Å were used to describe the five-layer 2 × 2 Pt surface. All thermodynamic energies were calculated at 298.15 K and 1 atm using the Atomic Simulation Environment suite of programs (equation (2))54.

$$G = H - TS = E_{\mathrm{DFT}} + E_{\mathrm{ZPE}} + {\int}_0^{298.15K} {C_v{\rm d}T - TS}$$
(2)

where EDFT is the total energy from DFT geometry optimization, EZPE is the zero-point vibrational energy (ZPE), \({\int}_0^{298.15K} {C_v{\rm d}T}\) is the thermal energy, Cv is heat capacity, T is the temperature and the S is entropy. The ideal gas approximation was used for O2 and H2, and the harmonic approximation was used for adsorbates.