Metal Core–Shell Nanoparticle Supercrystals: From Photoactivation of Hydrogen Evolution to Photocorrosion

Gas nanobubbles are directly linked to many important chemical reactions. While they can be detrimental to operational devices, they also reflect the local activity at the nanoscale. Here, supercrystals made of highly monodisperse Ag@Pt core–shell nanoparticles are first grown onto a solid support and fully characterized by electron microscopies and X‐ray scattering. Supercrystals are then used as a plasmonic photocatalytic platform for triggering the hydrogen evolution reaction. The catalytic activity is measured operando at the single supercrystal level by high‐resolution optical microscopy, which allows gas nanobubble nucleation to be probed at the early stage with high temporal resolution and the amount of gas molecules trapped inside them to be quantified. Finally, a correlative microscopy approach and high‐resolution electron energy loss spectroscopy help to decipher the mechanisms at the origin of the local degradation of the supercrystals during catalysis, namely nanoscale erosion and corrosion.


Introduction
Nanoparticles (NPs) made of plasmonic materials have emerged as very promising photocatalysts for a large number of heterogeneous chemical reactions such as H 2 S decomposition [1] or DOI: 10.1002/adma.202305402CO 2 reduction. [2]These NPs strongly interact with light through the collective oscillation of conduction electrons, a phenomenon known as the localized surface plasmon resonance. [3]During plasmon relaxation, energetic hot carriers (electrons and holes) are generated at the NP surface and local heating is produced by energy dissipation in the metal via the Joule effect. [4]Both can promote chemical transformations.Indeed, the photonic excitation of the NPs can generate a chemical potential sufficient to drive a redox reaction.As detailed by Yu and Jain, the hot carriers can be seen as reagents in a chemical reaction. [5]Therefore, plasmonic photocatalysis and photothermal catalysis can be used for triggering chemical reactions, [2] enhancing the turnover frequency (TOF) [6] or modifying the reaction selectivity. [7,8]fter studying model systems (i.e., well separated and monometallic plasmonic NPs), researchers are now turning their attention to more complex NP structures with heterogeneous chemical compositions with a view to tuning their physicochemical properties.For instance, bimetallic core-shell NPs [9] are promising for plasmonic catalysis to achieve higher TOFs based on the synergistic effects between the two constituents, one being plasmonic (e.g., Ag or Au) and the other being highly catalytic (e.g., Pt or Pd) but having weaker interactions with the irradiating light.In this way, the plasmon energy can be rapidly transferred to the catalytic metal through electron-electron interactions, limiting the e/h recombination step. [10]o counteract the low light-harvesting efficiency of single plasmonic NPs, a strategy consists of assembling them into superstructures to promote crosstalk (both chemical and physical communication) between the building blocks. [11]In order to enhance the local electromagnetic field that strongly depends on inter-NP distance, one can grow supercrystals (SCs) of NPs in a bottom-up manner to create multifunctional materials. [12,13]Recent advancements have also demonstrated the rational control of NP interfacial chemistry to tune inter-NP couplings and thereby modulate the electronic conductivity of SCs. [14]Such SCs have shown great promise, particularly for surface enhanced Raman scattering [15,16] while their potential in (photo)catalysis has been little explored.The aim of the present work is to investigate this subject by constructing and evaluating the potential of SCs composed of plasmonic and catalytically active NPs for photocatalytic chemical transformation, particularly in the context of the water splitting reaction.
In this work, SCs of Ag@Pt core-shell NPs were first synthetized and fully characterized.Then, their performance for the photocatalytic hydrogen evolution reaction (HER) was evaluated.The Ag core was chosen for its highly efficient plasmonic lightharvesting properties, while the Pt shell will confer both corrosion resistance and catalytic properties, as it is the best catalytic material for HER. [17]he extent of the photocatalytic reaction has been thoroughly scrutinized concerning both its impact on the SC structure and the products formed in the solution phase.The product formation is monitored operando and in real time, at the single SC level, by high-resolution label-free optical microscopy. [18]The strategy uses the production of gas as a proxy of the local catalytic activity of SCs, thanks to the large variation of refractive index between gas and water.[21][22][23][24] In addition, the structural changes in the SCs associated with the photocatalytic reaction are probed through a multiscale imaging strategy.First, sub-microscale analysis is provided by a correlative (identical location) optical and scanning electron microscopy (SEM) approach. [25]Then, shape and chemical nanoscale analysis (with sub-NP resolution) is obtained by scanning transmission electron microscopy using a high-angle annular dark-field detector (STEM-HAADF) analysis coupled with electron energy loss spectroscopy (EELS) elemental maps.

NP Synthesis
The Ag and Ag@Pt NPs with 1-2 Pt atomic layers were synthesized according to a previously published protocol. [26]Briefly, spherical Ag NPs of controlled size were synthesized by a hightemperature process in which AgNO 3 in the presence of oleylamine transforms to Ag(0) NPs whose sizes are controlled by the reaction time, heating ramp, and temperature.Here, the temperature was fixed at 240 °C, the heating ramp at 450 °C h −1 , and the reaction time at 1 h.Next, Ag@Pt NPs were synthesized via a seed-mediated growth process in which the Ag NPs are used as seeds in presence of a Pt precursor (Pt(acac) 2 ).The two-step synthesis process is depicted in Figure S1 in Section S1 (Supporting Information).The Pt shell thickness is controlled by adjusting the concentration ratio between the Pt precursor and the silver seed.An optimum Pt thickness between 1 and 2 layers was selected.Transmission electron microscopy (TEM) images of the resulting Ag NPs and Ag@Pt NPs are presented in Figure 1a,b, respectively.Both types of NPs show spherical shapes with sizes of 12.7 and 13.4 nm for Ag and Ag@Pt NPs.The shell thickness grown on the silver core is thus around ≈0.35 nm.The size dispersion graphs reported in Figure 1c evidence a size polydispersity equal to 9% for the two colloidal systems, which is narrow enough for 3D organization, [27] thereby enabling the formation of NP SCs, see below.
The Ag core, possessing a high extinction cross-section as well as a low imaginary dielectric constant in the visible, can strongly interact with light through electron-photon interaction.As a result, the extinction spectrum of the NP core (i.e., before Pt deposition) shows an intense plasmonic band centered at 403 nm, as noted in Figure 1d (blue curve).In the presence of the Pt shell, a material with a larger imaginary dielectric constant, a significant damping of the plasmon band is measured (Figure 1d, red curves), even though the shell is ultrathin.Nevertheless, considerable light-matter coupling persists and energy can be transferred via electron-electron interactions from the Ag plasmonic core to the Pt shell.The dotted lines correspond to the theoretical extinction spectra for NPs of equivalent size and composition obtained by Mie theory.Different configurations have been simulated in the framework of the Mie theory, described in Section S2 (Supporting Information), by varying the core diameters (Figure S2a, Supporting Information) and shell thickness (Figure S2b-d, Supporting Information).The simulated results highlight that the extinction spectra of the core-shell NPs are both sensitive to the core diameter and the shell thickness.It likely explains the slight differences between the simulated results and the experimental spectra reported in Figure 1d.Overall, the fairly good agreement found shows that the damping of the electromagnetic field enhancement upon introduction of 1-2 monolayers of atoms can be understood by Mie theory, which can thus assume a predictive role for other materials synthesis.The idea is that the e − generated during the light excitation process should be harvested via the electrocatalytic Pt shell to drive the reduction of protons into molecular H 2 .This hypothesis was verified by irradiating colloidal solutions of Ag, Pt, and Ag@Pt NPs in the presence of acid for a few minutes using a tungstenhalogen white light source.While no change is observed for Ag nor Pt NPs, a significant amount of gas is produced in presence of the bimetallic NPs (see Figure S3 in Section S3 of the Supporting Information).This illustrates qualitatively the synergistic effect of the core-shell structure.Similar results have been published previously for the reduction of para-nitrophenol. [26]

Self-Assembly into SCs
In order to develop a supported nanocatalyst platform enabling the photoactivation of HER, which could produce H 2 at high mass transfer, for instance in fluidic systems, different chemical engineering solutions have been proposed (fluidized particle beds, etc.). [28]Here, we wish to take advantage of the ability of the NPs to assemble into dense and mechanically stable SCs.This furnishes several advantages: i) the SCs can be assembled or disassembled upon chemical stimuli (herein wetting or solvent evaporation), ii) the NPs assemble into submicrometer-to micrometer-sized crystals which can be easily immobilized on surfaces, iii) the assembly further provides a higher cross-section for light absorption (leading to enhanced chemical transformation).
Ag@Pt NPs initially synthesized in oleylamine were then transferred to toluene to induce their 3D self-assembly.Toluene allows SCs to grow homogeneously by promoting van der Waals interactions between the NPs. [29]Ag@Pt NPs, coated with oleylamine and dispersed in toluene were deposited by drop-casting 20 μL of a 0.1 mm solution of Ag@Pt NPs onto flat substrates in an inert atmosphere.The deposition is first performed on a Step-by-step formation of colloidal supercrystals from the self-assembly of core-shell Ag@Pt NPs.a,b) TEM images correspond to silver seeds and Ag@Pt NPs, respectively.c) The corresponding NP size dispersion graphs and d) UV-vis spectra of Ag and Ag@Pt NPs dispersed in toluene (experiments and Mie predictions in solid and dotted lines, respectively).Data related to silver seeds appear in blue while those corresponding to core-shell architecture are presented in red.e) TEM and f) field-emission gun SEM (SEM-FEG) images of single crystals of Ag@Pt NPs obtained by drop-casting 5 μL of the corresponding colloidal solution.g) SEM images and the corresponding Ag (blue) and Pt (red) elemental EDX mapping of crystals grown on an ITO substrate.
TEM grid, using anticapillary tweezers, in order to visualize the SCs.An image of the resulting SCs is presented in Figure 1e.The Ag@Pt NPs spontaneously self-assemble into individual quasitriangles after the evaporation of the solvent.The inset in the TEM image of Figure 1e shows the fast Fourier transform (FFT) image of the individual SC, that displays well-defined spots characteristic of well-organized NPs and is typical of a face centered cubic (fcc) arrangement. [30]imilarly, the depositing of the colloidal solution onto the surface of an indium tin oxide (ITO) coated glass coverslip yields the self-assembly of the Ag@Pt NPs into individual SCs with sizes of ≈1 μm, as observed in the SEM images of Figure 1f,g.The deposit was also analyzed by energy-dispersive X-ray spectroscopy (EDX) and confirmed the elemental composition of the SCs (see Figure S4 in Section S4 of the Supporting Information) agreeing nicely with that deduced from TEM image analysis.EDX mapping (Figure 1g, blue and red for Ag and Pt, respectively) also shows a perfectly homogeneous repartition of Ag and Pt in the 3D structures, confirming the uniform composition of the building blocks.Indeed, in a previous article, we gave evidence that the core-shell structure is homogeneous in composition. [26]No seg-regation has been observed.The atomic composition of Ag@Pt NPs deduced from EDX analysis agrees with the one calculated from TEM images (see Table S1 in Section S4 of the Supporting Information).
Finally, the crystalline structure of the formed SCs was investigated by grazing incidence small angle X-ray scattering (GISAXS), detailed in Section S5 (Supporting Information).The GISAXS pattern (Figure S5 in Section S5 of the Supporting Information) confirms that the NPs are well-ordered in 3D.The diffraction spots of the GISAXS pattern can be associated with a fcc lattice.The clear equatorial reinforcement of the (111) and (222) diffraction lines also suggests that a large proportion of SCs present the (111) face parallel to the substrate.This is consistent with the SEM images (Figure 1f,g) that show a large number of triangles lying on their larger faces.One can also point out the presence of elongated spots that reflect the presence of out-ofplane stacking faults.(111) facets have the highest coordination number and should have the highest number of hot spots.In addition, they present the highest total surface area.Altogether, (111) facets should exhibit the highest catalytic activity.From the positions of the Bragg reflections, the NP-NP separation distance is estimated as 3.6 ± 0.4 nm, in agreement with the presence of oleylamine, a capping ligand carrying a long alkyl chain (length 2.04 nm), and with low chain interdigitation (Figure S5 in Section S5 of the Supporting Information).

Optical Monitoring of Gas NB Formation
The plasmonic photocatalytic activity of the supported SCs was scrutinized in real time using high-resolution optical microscopy.As shown in Figure 2, the SCs are illuminated from the backside of the transparent ITO substrate through a high numerical aperture (NA = 1.4,oil immersion) objective.The light is then reflected by the substrate surface (red arrows in Figure 2a) and scattered by any object present at the ITO surface, for instance, the SC (blue lines in Figure 2a) and further collected through the same objective by a digital camera acquiring images at 20 Hz.These particular illumination and imaging conditions, named interference reflection or interference scattering, boost the microscope sensitivity, since the optical signal collected on each pixel, I(x,y), is mainly due to the interference between the reflected and backscattered fields. [31,32][35] The light source of the optical microscope can also be used to trigger photochemical processes. [24]In this work, it appears that the white light source (light-emitting diode (LED), surface power density < 5 W cm −2 ) is sufficient to trigger the photoactivation of the cathodic semireaction corresponding to Equation (1) at the SC surface thereby enabling the evolution of H 2 gas, possibly at sufficient levels (above saturation) for the formation of H 2 gas NBs The experimental methodology then consists of monitoring dynamically, the formation (nucleation or appearance) of gas NBs, from their early stages to their detachment from a single SC of Ag@Pt NP immobilized on a transparent ITO-coated glass coverslip.Two examples of videos, showing, respectively, the intermittent blinking due to NB nucleation (Video S1, Supporting Information) and the motion upon detachment of a single SC (Video S2, Supporting Information), are provided as supporting materials and are described in Section S6 (Supporting Information).The conducting ITO layer coating facilitates SEM imaging, thereby enabling identical location correlative imaging of the SC structure and distribution (see examples of colocated SEM and optical images in Figure S6 and S7A,B in Section S7 of the Supporting Information).
The HER at these single immobilized crystal catalysts is then actuated and dynamically probed upon contact with a 5 mm H 2 SO 4 aqueous solution.In particular, the HER reaction is performed within an ≈10 3 μm 2 area of the surface, defined by a meniscus cell confined between a micropipet filled with the aqueous solution and the crystal-coated ITO substrate (see Figure 2a).This confinement technique, adapted from the scanning electrochemical cell microscope, [36,37] allows us to target optically multiple regions of interest (ROI) on the same substrate. [38,39]Note that a rather low H + concentration (10 mm) has been used to avoid damaging the ITO surface that has been shown to be unstable in too acidic conditions. [20]igure 2b presents a typical wide field (16.5 × 21 μm 2 ) optical image of the crystal-coated ITO substrate contacted with the meniscus of solution.It shows 44 crystals, appearing as bright contrasted features, with sizes often larger than the diffraction limit.Images are taken successively at a frequency of 20 Hz, over several seconds, from which the dynamics of H 2 NB production can be evaluated.The strong scattering from the crystals hampers the direct visualization of single NBs (compare the crystal, f 0 , and crystal + NB, f 1 , images in Figure 2c).However, by image reconstruction, one can obtain a differential contrast image revealing, in Figure 2c (image named NB), the presence of a NB captured on the edge of the crystal.Indeed, the overall optical intensity slightly increases when a NB is nucleating on the crystal (image labeled as f 1 ) in comparison with the intensity related to the same crystal without a NB (image labeled as f 0 ).The point spread function (PSF) observed in the image f 1 of Figure 2c is then a combination of the contribution of both the crystal and the NB.Zhou and co-workers already addressed this issue to monitor the decomposition of formic acid at PdAg nanoplates. [23]They proposed to separate the two contributions (material and NB) by image subtraction to retrieve the NB PSF (see Figure 2c) that can be further used to superlocalize its center of mass and therefore its anchoring site within ±10 nm spatial resolution.From a previous optical simulation of light scattering by NBs, [19] it has been shown that when NBs appear as bright contrast features, they already have rather large diameters, i.e., more than 380 nm.This suggests that NBs can be directly sized by fitting their PSF with a Gaussian function and measuring its full width at half maximum (FWHM), taking into account that the microscope, with a NA objective of 1.4 and a mean detection wavelength  = 510 nm possesses a PSF ≈ 1.22 × /NA ≈ 220 nm.The example presented in Figure 2 suggests that a NB with diameter < 500 nm was formed by the crystal, based on the Gaussian fit shown in Figure 2d.

Quantification of H 2 Gas Production
To describe the dynamics of NB production at the same individual crystal (Figure 2c), the optical signature arising from NBs was visualized dynamically by using a slightly different image reconstruction procedure named differential rolling average method. [40]Briefly, it consists of comparing the image frames (f) by dividing two consecutive stacks of (N = 5) frames and to roll the operation over the whole image sequence.Then, for each pixel (x,y) in the frame f S taken at time (S) within the image sequence, a contrast C S is obtained following Equation ( 2) This image reconstruction process removes static features in the images (i.e., the SCs) and increases the signal to noise ratio in order to distinguish single NB signals from the background intensity.
The production of NBs can then be dynamically probed during the substrate illumination.The dynamics of a NB is visualized through the evolution of local contrast with time (optical transient) extracted from a small ROI.Such an optical transient recorded in a single crystal region is shown in Figure 2e.The events of NB formation or disappearance in this single crystal region are detected instantaneously as contrast C spikes, the formation being associated with rise in C S , particularly above a given threshold value (displayed in red in Figure 2e), typically >2% above the noise level (≈0.5%).Similarly, the NB detachment from the crystal surface is associated with a fall in C S (going back to the baseline C S ≈ 0).
Figure 3 summarizes 6.5 s of NB tracking over a rather large support surface (area ≈ 350 μm 2 , Figure 3a) including 44 SCs.Data over a longer period of time are available in Figure S8A of Section S8 (Supporting Information).The positions of all the NBs detected during the optical monitoring are represented by the red dots in Figure 3a.It can be noted that all the NBs seem to grow onto or close to single crystals and therefore act as reporters of the local catalytic activity of individual SCs.Based on the NB coordinates, less than ≈40% of SCs are active during this time lapse.In addition, counting the number of NBs formed over the experiment yields the rate of NB production events.As attested by the NB counting and the linear evolution of the cumulative count shown in Figure 3b, the rate of NB production is constant over time and is evaluated to ≈2 × 10 −2 NBs μm −2 s −1 .However, the rate is not constant everywhere on the support, as detailed in Figure S8B (Supporting Information), and mostly depends on the structural characteristics (shape or NP stacking faults, for instance) of the drop-casted SCs and their surface density.
Next, we analyzed the stochastic appearance and detachment of NBs to evaluate the amount of H 2 associated with NB production over the imaged region.The PSF of every probed NB, just before they detached, was fitted by a Gaussian function to extract its FWHM.An example of such a PSF and the corresponding fit is shown in Figure 2d.Here, the FWHM is used as a proxy of the final NB size and to calculate the final NB volume, V NB , following Equation (3) assuming the NBs detach as spheres from the nanostructured SC The amount of H 2 gas contained inside the NB, n H 2 , can further be estimated from V NB and Equation (4) assuming the ideal gas situation where k B is the Boltzmann constant and T is the temperature.The internal pressure (P L ) for spherical NBs is calculated using the Young-Laplace equation where P 0 is the ambient pressure (1 atm),  is the aqueous solution surface tension (72.8 mN m −1 ), and r is the NB radius (r ≈ FWHM/2).Typically, a 0.5 μm NB would hold a 7 atm Laplace pressure, carrying 1.2 × 10 7 molecules or 0.02 fmol of H 2 gas.From the NB size dispersion graph presented in Figure 3c, one can reconstruct the evolution of n H 2 over time presented in Figure 3d.As the NBs have rather similar sizes (0.5 μm in average) and as the NB nucleation frequency is constant over time, H 2 production also appears constant over time.From the 0.9 fmoles of H 2 gas detected during the 6.5 s experiment presented in Figure 3d, the production rate of H 2 molecules rises to ≈2.6 × 10 5 molecules μm −2 s −1 .Considering that the crystals occupy a fraction  ≈ 7% of the area in the ROI, the equivalent photocurrent j resulting from proton reduction amounts to j = 2 × rate × q e / ≈ 0.12 mA cm −2 of active surface, with q e = 1.6 × 10 −19 C, the elementary charge of the electron.

Plasmonic Photocatalysis at Single SCs
We then turned our attention to the production of molecular hydrogen by individual SCs. Figure 4 summarizes the three distinct behaviors in photocatalytic activity observed at the single SC level.From the typical contrast intensity transients in Figure 4a: i) some SCs show a rather constant activity, giving a production rate of ≈0.2 NB s −1 , that does not fluctuate much over time; while other crystals: ii) exhibit intermittent activity, or iii) appear not to be active at all in the course of the experiment.In the latter case (54% of the SCs in the ROI in Figure 3a), the intensity transients in the SC region do not show any spikes.
As previously stated, the occurrence of a spike in the optical transient reveals the formation and subsequent detachment of a single NB whose size is inferred either from its FWHM or equivalently from its intensity (the NB volume being proportional to its intensity, as shown in Figure S9 in Section S9 of the Supporting Information).It is used similarly to calculate from the contrast time trajectory in Figure 4a the total number of H 2 molecules generated by the single SC over time.This production rate per SC is analogous to the TOF for an individual SC, which can be further determined for all the imaged crystals, as reported in Figure 4b.The distribution of the TOFs for the probed active SCs is in the range of around 0.6 × 10 6 H 2 molecules SC −1 s −1 or equivalently 0.2 pA SC −1 .In comparison, for an ultra-microelectrode of similar size (1 μm in diameter), the mass-transfer-limited reduction of protons would amount to ≈2 nA, which is about 10 4 times higher than the equivalent photocurrent supported by a SC.

Reaction Mechanism and Degradation
SC Erosion: Many studies have reported that structured substrates frequently reorganize during gas evolution [41] and that the catalysts often undergo reaction-induced chemical degradation.Such effects should be problematic in photocatalytic systems where an e − /h + pair is produced, since the e − is harvested for HER, the h + may have a deleterious effect if not carefully harvested into another useful reaction.This is not usually a problem, as photocatalytic reactions normally employ a sacrificial donor to evacuate the photogenerated h + .Here, the aqueous solution water oxidation may be one possible path for h + harvesting.We discuss here the other possible paths and more particularly those leading to the physical and chemical degradations of the catalytic SCs.As the optical monitoring clearly shows that many SCs are indeed active in hydrogen evolution, restructuring processes and material damage were investigated, from the single SC to the single NP scale.It is first inspected by a correlative microscopy approach that couples optical microscopy with postmortem SEM in the same ROI, as exemplified in Figure 5a (larger field of view of the images in Section S7 in the Supporting Information).Here, we chose a substrate region with larger SCs as they are likely to produce a larger amount of molecular hydrogen and therefore to cause more visible damage.However, with larger and then thicker SCs, the optical imaging tracks a smaller amount of NB formation, as it is likely be blind to NBs nucleating atop the SC and only NBs forming near the edges of the SCs are mostly visible.
The most evident changes associated with the photocatalytic reaction are primarily observed in the microscale motion of the SCs on the substrate.This motion can be directly detected through optical image sequences (it is illustrated in Video S2 in the Supporting Information, described in Section S6 in the Supporting Information).Figure S10 in Section S10 of the Support- ing Information provides specific examples of SCs in motion on the ITO support.These images demonstrate that the production of gas induces movement in multiple SCs, as previously observed in the case of individual nanoscale electrocatalysts. [42]The motion of SCs implies that the mechanical forces, such as convection, resulting from hydrogen evolution, do not cause the complete disintegration of SCs into individual NPs.Consequently, these forces are insufficient to break a significant number of attractive interactions between NPs.
Next, the effect of gas evolution can be seen within the single SC from same location SEM and optical images.Particularly, localizing the footprint of the droplet cell at the substrate surface in SEM images, as in Figure 5b, allows to compare, at high spatial resolution, the morphological changes induced by the gas evolution reaction at the single crystal level.The wider field identical location optical image, in the presence of the acidic droplet, and SEM image corresponding to Figure 5b are provided in Section S7 (Figure S7A,B, Supporting Information).In Figure 5b, the SCs located inside of the cell (Figure 5b(ii)) clearly show holes and cracks, while the crystals found outside remain unaffected (Figure 5b(iii)).A control experiment was performed in the absence of light irradiation on SCs exposed to 5 mm H 2 SO 4 during 10 min while SEM images were acquired before and after the exposition.The SEM images, provided in Section S11 (Figure S11A,B, Supporting Information), do not show the cracking or holes observed in the presence of light irradiation as in Figure 5.This control experiment shows that there is little effect of the SC exposure to dilute acidic solution in absence of light irradiation.We note that the size of the holes (≈100 nm) is compatible with the NB diameters found at the beginning of their growth and we believe this is evidence that H 2 evolution also leads to the erosion of the SC.The erosion of the SC suggests the existence of weak points in the crystal structure, which may be due to physical defects, such as stacking faults or grain boundaries.It may also arise from chemical weakening of the SC, such as the corrosion induced by the HER reaction.Both causes would lead to the detachment of NPs from the SC and their release into solution.These structural rearrangements are likely to affect the SC catalytic activity over time and could be at the origin of the intermittent activity evidenced in Figure 4.
SC and NP Corrosion: Beyond mechanical erosion of the SCs, corrosion can also be associated with a chemical transformation of the SC and of its individual structural building block, the Ag@Pt NP.To apprehend the possible chemical corrosion of the SC or NP during HER, the reaction mechanism is first inspected in colloidal solution of Ag@Pt NPs.Colloidal Ag or Ag@Pt NP solutions (containing 0.01 m metal precursor) in the presence of 0.05 m H 2 SO 4 were irradiated by a polychromatic white light source and the optical properties of the colloidal dispersions were evaluated in situ by UV-vis spectroscopy, as reported in Figure 6a.While the plasmon characteristics of Ag NPs remain almost constant over 2 h, the plasmon intensity of the Ag@Pt core-shell NPs rapidly decreases with time (linearly to a first approximation; decay rate of about 10% h −1 of irradiation).The Ag@Pt NPs' plasmon spectral changes are also associated, as mentioned above, with gas-bubble formation (see Section S3 in the Supporting Information) attesting to the HER occurring under photoactivation.Moreover, this is also associated with an instantaneous and significant deposition of materials at the bottom of the spectroscopy cell upon addition of chloride ions (KCl).The deposit was further analyzed by EDX spectroscopy and is found to be silver chloride salt, as shown in the EDX elemental spectrum of Figure S12 in Section S12 of the Supporting Information.
The bulk experiment described above suggests that the reductive HER step is accompanied by the oxidation of Ag into Ag + .Indeed, when the NPs are irradiated by light, an e − /h + pair is generated during the damping of the plasmonic excitation.The separation of this e − /h + pair is then involved in an overall redox process schematized in Figure 6b.The excess of protons and presence of catalytic Pt atoms enables the fast consumption of the e − and H + into molecular H 2 through the HER (Equation ( 1)).The counterpart anodic reaction (Equation ( 6)) appears to be the oxidation of the less noble metal of the Ag@Pt NPs, i.e., the Ag atoms, by the h + .This corrosion of Ag from the NP core is at the origin of the progressive disappearance of the plasmonic band and, in the presence of Cl − ions, to the concomitant precipitation of AgCl following Equation ( 7) From a simplified theoretical model which considers a H 2 /Ag ratio of 0.5 and a plasmonic intensity decay strictly linked to the symmetrical dissolution of the Ag core of the NPs, a H 2 production rate on the order of 3 × 10 2 n m s −1 can be estimated for the colloidal solution of Ag@Pt NPs (calculation detailed in Section S13 in the Supporting Information).At the single NP level, this would correspond to an average TOF of ≈2 H 2 molecules NP −1 s −1 .This value can be compared to the HER activity of single SCs probed by optical microscopy in Section 2.3.3, even though both experiments were not performed under the same light irradiance.If we consider a SC as a compact fcc 3D arrangement of NPs, one can evaluate a compactness constant of p =  √ 2∕6 ≈ 0.74.For a 1 μm 3 SC and inter-NP separation distance of 3.6 nm, the number of NPs constituting the SC amounts to ≈3 × 10 5 .Based on the TOF for a single NP, the SC would be able to produce 6 × 10 5 molecules s −1 .This actually compares well to the 6 × 10 5 molecules s −1 TOF measured by the optical microscopy monitoring.
The mechanism described above then suggests that most of the Ag@Pt NPs of the catalytically active crystals (or at least at the crystal surface) would be available for oxidative corrosion.This may explain the mechanical erosion observed at the level of individual SCs, induced by a chemical degradation of individual NPs.The oxidative corrosion of the NPs may also affect the integrity of the NPs and of the SCs.For example, in the absence of Cl − ions, one expects a dissolution of the core of the Ag@Pt NPs, while in the presence of Cl − , the Ag core would be converted into a less dense AgCl, yielding a size increase of the NP [43,44] ).In the absence of Cl − ions, H 2 evolution should be associated with dissolution of NPs from the SC.From the 6 × 10 5 H 2 molecules SC −1 s −1 , a twice larger amount of Ag atoms dissolved would correspond the dissolution of ≈20 NPs (assuming ≈5.5 × 10 4 Ag atom per NP) within each SC.It is definitely small compared to the number of NP per SC and compared to the sizes of holes and cracks suggesting that the aforementioned erosion has both chemical (dissolution) and mechanical origins.
The origin of the chemical corrosion of the Ag@Pt NP requires atomic scale imaging of the individual NPs.The core-shell architecture and the integrity of the Pt layer were then analyzed by STEM in the high-angle annular dark-field (HAADF) mode, and using EELS for chemical mapping.Figure 6c,e shows typical STEM-HAADF images of Ag@Pt NPs with Ag core and Pt shell in a dark and bright contrast, respectively.This can be verified by the corresponding EELS maps (Figure 6d,f) with Ag core in yellow and Pt shell in blue.
The Ag@Pt NP surface seems rather homogeneous (Figure 6c-f).A contrario, the EELS mapping reveals discontinuities in the Pt shell and pitting that could indeed explain that the Ag core, not fully protected by the Pt layer, may corrode.The arrows in Figure 6d also highlight a very thin atomic layer of silver deposited on the Pt shell for a few NPs in a discontinuous manner.The outer Ag layer may be attributed to the possible competition of Pt ion reduction during the NP synthesis.The direct reduction of Pt ions by the oleylamine ligand at the Ag NP may compete with galvanic exchange of the Ag metal (producing Ag + ions and Pt metal deposit).The dissolved Ag + may then be subsequently reduced by oleylamine during the synthesis step, which may lead to local deposition of Ag at the NP surface.The resulting structure is detrimental during the photocatalytic HER as it will promote access to the Ag NP core, thus accelerating the overall corrosion of the NP.
The low leakage of Ag + (proportional to the amount of hydrogen produced) should also result in a non-negligible layer of AgCl at the NP surface in presence of chloride ions that could be detrimental to the HER rate over time through a loss of efficiency of the light absorption and a poorer access to charge carriers in the Pt shell.In real applications, hole scavengers would be used to limit the corrosion process, such as alcohol or amines.However, silver oxidizes also quite easily.It is delicate evaluating the competition between Ag core corrosion and, e.g., isopropanol oxidation at the Pt shell.The onset of the oxidative dissolution of Ag NPs, which depends on the NP size, was reported for 8 nm Ag NPs around 0.4-0.45V vs standard hydrogen electrode (SHE). [45,46]By comparison, the oxidation of isopropanol at highly reactive single crystal Pt electrodes was shown to depend on the crystal orientation, ranging from 0.3 to 0.5 V vs SHE in 0.5 m H 2 SO 4 . [47]Similar onset oxidation potential was recently reported for Pt nanoparticles. [48]Herein, the Pt-functionalized Ag nanoparticles are coated with an oleylamine capping agent, which might shift Pt oxidation to more positive values compared to those reported ones.It is then likely that Ag will oxidize at comparable rate than isopropanol at the Pt shell.Based on those literature data, we believe that for practical applications, a more noble metal (such as gold) might be required.In addition, one should remember that for efficient HER, a sacrificial reducer is required to consume the h + , at least easier to oxidize than the plasmonic core of the NP.

Conclusion
Plasmonic SCs with fcc lattice and made of Ag@Pt core-shell NPs were successfully grown on solid support, as proved by TEM, SEM-EDX, and GISAXS analyses.High-resolution optical microscopy technique was then employed to evaluate quantitatively their photocatalytic activity toward HER in acidic condition at the single SC level by detecting, counting, and sizing operando single hydrogen NBs generated following light irradiation.Analyzing local optical intensity fluctuation also revealed three distinct SC behaviors: constant gas production, intermittent activity, and total inactivity.Degradation mechanisms were finally deciphered, in a topdown multi-imaging approach from the single SC level by correlative microscopy (same location optical and electron microscopy analysis) and then at the single NP level by high-resolution STEM-HAADF coupled with EELS elemental mappings.Results show that SCs are subjected to mechanical erosion during H 2 evolution suggesting the presence of weaker points in the SC structure.NPs also corrode in absence of h + scavenger, after carriers' generation.Indeed, the excess of protons in solution and the presence of the Pt shell lead to the fast consumption of e − through HER.The counterpart anodic reaction is likely to be the oxidative dissolution of Ag core accessible because of the inhomogeneity of the Pt shell.To go further and to study the relationship between the bubble formation and the light intensity, one would make sure that the hydrogen production is only driven by the cathodic reaction.Monitoring NB formation on gold-coated Pt NPs together with a hole scavenger, for instance, could be a solution.

Experimental Section
Materials and Chemicals: Chemicals were used as received without any further purification and stored under inert atmosphere.Sulfuric acid, silver nitrate, dioctyl ether, oleic acid (90%), chloroform (≥99%), and Pt(acac) 2 were purchased from Merck.Oleylamine (80-90%) was from ACROS Organics.Ultrapure water with a resistivity of 18.2 MΩ cm was used to dilute the sulfuric acid.
ITO coverslips (thickness #1.5) were purchased from SPI.The thickness of the conductive layer deposited on glass was 350 nm, corresponding to a resistivity of about 15-30 Ω cm.
Pipettes were made by pulling borosilicate glass capillaries (outer and inner diameters of 1.0 and 0.5 mm, respectively) using a P-2000 laser puller from Sutter Instruments.The pipette tips were then polished using aluminum oxide tape (3 μm, from Precision Surfaces International) to obtain a final diameter < 50 μm before being filled with dilute sulfuric acid.
UV-Vis Spectroscopy: Extinction spectra were recorded using a Varian Cary 5000 spectrometer.The NP concentration (C NPs ) was evaluated using the following equation where C M is the total atomic concentration after synthesis, V M is the molar volume of metal, D is the diameter of the NPs, and Na is the Avogadro number.C M was determined using the Beer-Lambert law.Optical Microscopy: Optical imaging was performed using a motorized inverted microscope (Zeiss, Axio Observer 7) equipped with a high numerical aperture objective (Zeiss Plan-Apochromat 63×/1.4NA with oil immersion).The substrate was illuminated from the backside (glass side) with a white light source and the reflected light was collected by a complementary metal oxide semiconductor (CMOS) camera (UI-3080CP Rev. 2, IDS) operating at 20 Hz.
The local photochemical experiments were conducted using an electrochemical probe scanner from Heka with a three-axis piezostage that allowed one to approach a pipette tip close to the surface and to create a chemical microdroplet cell.The whole setup was mounted on an isolation table that actively reduced the mechanical noise.
SEM-EDX and Correlative Microscopy Approach: SEM analyses on ITO substrates were performed on a Gemini SEM 360 from Zeiss equipped with an energy-dispersive X-ray detector from Oxford Instruments.For elemental analyses, the microscope aperture size was expanded to 60 μm.The data were processed using the Aztec software.
Correlative microscopy analyses (same location in optical and electron microscopy) were achieved using a specific marked substrate holder from Zeiss allowing retrieval of the same position in optical and electron microscopies for imaging.The data were processed using the Zeiss ZEN Blue software and the ZEN Connect module.
TEM: TEM images were recorded with a JEOL JEM1011 electron microscope with an acceleration voltage of 100 kV.Prior to the visualization, NPs were drop-cast from a diluted solution onto a carbon-coated copper grid.For observing crystals, NPs were deposited using anticapillary tweezers.
STEM-EELS: STEM images were acquired using a HAADF detector in a Nion Ultrastem 200 Cs-corrected microscope operating at 100 kV with a probe size of about 0.07 nm.The probe convergence and HAADF detector minimum collection semiangles were 35 and 75 mrad, respectively, meaning that HAADF images were dominated by Z-contrast, although some diffraction contrast was possible.Elemental maps were obtained using EELS in the spectrum-image collection mode using the Ag M-and Pt Medges.
GISAXS: GISAXS was performed in Orsay with a rotating copper anode generator operated at 40 kV and 20 mA (small focus: 0.1 Å-0.1 mm 2 in cross-section).The optics consisted of two parabolic multilayer graded mirrors in K-B geometry, which delivered a well-defined and intense parallel monochromatic beam.Photostimulable imaging plates were used as detector.The reading of the exposed imaging plate was performed by a STORM 820 scanner (Molecular Dynamics).

Figure 1 .
Figure 1.Step-by-step formation of colloidal supercrystals from the self-assembly of core-shell Ag@Pt NPs.a,b) TEM images correspond to silver seeds and Ag@Pt NPs, respectively.c) The corresponding NP size dispersion graphs and d) UV-vis spectra of Ag and Ag@Pt NPs dispersed in toluene (experiments and Mie predictions in solid and dotted lines, respectively).Data related to silver seeds appear in blue while those corresponding to core-shell architecture are presented in red.e) TEM and f) field-emission gun SEM (SEM-FEG) images of single crystals of Ag@Pt NPs obtained by drop-casting 5 μL of the corresponding colloidal solution.g) SEM images and the corresponding Ag (blue) and Pt (red) elemental EDX mapping of crystals grown on an ITO substrate.

Figure 2 .
Figure 2. Operando optical imaging by interference reflection microscopy of NB appearance over the supercrystals supported ITO substrate.a) Schematic representation of the microscopy configuration used to probe H 2 NBs produced by photoreduction of a 5 mm H 2 SO 4 aqueous solution at single SCs of Ag@Pt NPs.b) Optical image recorded in a region of interest on the ITO support.c) Optical image before (f 0 ) and after (f 1 ) the growth of a single NB on a catalytic crystal.The optical image in (c) labeled as NB corresponds to a contrast image, representing (f 1 − f 0 )/f 0 .d) Gaussian fitting of a line profile of the resulting feature revealed in the contrast image (NB).e) Optical intensity profile evidencing the transient generation of 3 NBs (orange crosses) at the same single crystal.

Figure 3 .
Figure 3. Mapping and quantification of H 2 production at SCs supported on a solid substrate.a) Optical mapping of NB nucleation events over 44 SCs during 6.5 s of white light irradiation.The superlocalization of the nucleation events is represented by red dots.b) NB counting over time in the region imaged in (a).c) Size dispersion graph for gas NBs obtained by systematic fitting of each NB PSF, as shown in Figure 2d.d) H 2 gas production calculated from (b) and (c).

Figure 4 .
Figure 4. Hydrogen production at the single SC level.a) Optical intensity transients in three single SC regions summarizing three distinct behaviors: i) constant photocatalytic activity, ii) intermittent H 2 NB production, and iii) absence of photoactivity.b) Distribution of turnover frequency at single crystals.

Figure 5 .
Figure 5. Correlative microscopy approach for visualizing the effects of the catalytic reaction at the single SC level.a) Identical location optical and electron microscopy images of SCs immobilized on an ITO coverslip.b) SEM images: i) recorded at the border (dashed line) of the droplet cell confining the photocatalytic reaction (see identical location optical image in Section S7 in the Supporting Information) and ii,iii) zoom over two single SCs inside (ii) and outside (iii) of the cell.

Figure 6 .
Figure 6.Unraveling the mechanism of plasmon-assisted hydrogen production at the core-shell NP surface.a) Monotonic decrease of the maximum intensity of the plasmonic band as a function of irradiation time for 0.01 m of Ag NPs (blue dots) and Ag@Pt NPs dispersed in 0.01 m sulfuric acid solution (orange dots).b) Scheme representing the redox process occurring at the NP surface.c,e) High-resolution STEM image of Ag@Pt NPs together with d,f) the corresponding EELS elemental mapping highlighting the discontinuity of the Pt (blue) shell and the presence of silver (yellow) at the Pt outer surface.