Atomistic Picture of Electronic Metal Support Interaction and the Role of Water

Electronic metal support interaction in the Pt/Co3O4(111) model catalysts involves cation exchange yielding atomically dispersed Pt2+ and Pt4+ species. In the presence of water, these can be stabilized in the form of triaqua complexes.

Pt (Goodfellow, 99.99%) was deposited by PVD onto the Co3O4(111) film either in a stepwise or a single-step manner at 300 K in UHV.The nominal thickness of the Pt film was determined from the attenuation of the Co 2p1/2 intensity.The Pt thickness is expressed in terms of monolayers (ML)   considering 1 ML = 0.2266 nm.The core level spectra were acquired with photon energies of 180 eV (Pt 4f and Ir 4f), 380 eV (Pt 4f and C 1s), 650 eV (Pt 4f, O 1s), and 930 eV (Pt 4f, Ir 4f, Co 2p).Additionally, the valence band spectra were acquired with photon energies of 60 eV and 115 eV.The binding energies in the spectra were calibrated with respect to the Fermi level.All spectra were acquired at constant pass energy and at an emission angle of the photoelectrons of 0° with respect to the sample normal.The total spectral resolution was 200 meV (h = 60-180 eV), 350 meV (h = 380 eV), 650 meV (h = 650 eV), and 1 eV (h = 930 eV).All SRPES data were processed using the KolXPD fitting software.S4 The spectral components in the Pt 4f spectra were fitted with a Voigt profile after subtraction of a Shirley background.
During the experiments, the sample temperature was controlled by a DC power supply passing a current through Ta wires holding the sample.Temperature was monitored by a K-type thermocouple spot-welded to the back of the sample.

S1.2. Scanning Tunnelling Microscopy and Temperature Programmed Desorption (TPD)
Scanning tunneling microscopy (STM), X-ray photoelectron spectroscopy (XPS), and temperature programmed desorption (TPD) experiments were performed in an UHV system (base pressure 1×10 -9 mbar) at Charles University in Prague, Czech Republic.The chamber was equipped with a photoelectron spectrometer (Specs Phoibos 150), an Al Kα X-ray source, LEED optics with a CCD camera, a home-built STM setup, a quadrupole mass spectrometer (QMS, Pfeiffer PrismaPlus), a sputter gun (Ar + ), a dual-beam evaporator (Tectra), a gas inlet system, and a radiative heating.
Co (Alfa Aesar, 99.995%) was deposited onto Ir(100)-(2×1)-O in the background O2 atmosphere of 1.5×10 -6 mbar at 300 K from an electron-heated Ta crucible.After Co deposition, the layer was further oxidized in 2×10 -6 mbar of O2 for 60 min at 600 K and finally at 680 K. To achieve the final Co3O4(111) ordering, annealing in UHV at 720 K for 10 min was performed.The layers were prepared with a nominal thickness of approximately 5.0 nm as determined by a quartz crystal microbalance (QCM) and by attenuation of the Ir 4f XPS signal.The Pt/Co3O4(111) model catalysts were prepared by means of PVD of Pt on the as-prepared Co3O4(111) film.Pt deposition was performed from electron-heated Pt wire (diameter 0.5 mm, Goodfellow, 99.99%) in UHV at 300 K.The amount of deposited Pt was determined by means of XPS and QCM.
STM imaging was performed at 300 K using electrochemically etched W tips. STM images were obtained at tip bias and tunneling currents as follows: Co3O4(111) at -2.0 V, 0.3 nA; 0.02 ML Pt/Co3O4(111) at -1.8 V, 0.15 nA; 0.04 ML Pt/Co3O4(111) at +2.0 V, 0.35 nA; and 0.13 ML Pt/Co3O4(111) at +2.0V, 0.3 nA.TPD spectra were acquired using QMS, enclosed in a differentially pumped compartment with a nozzle facing the sample, in order to separate molecules desorbing directly from the sample surface from the background contributions.The samples were heated radiatively and the cooling was provided by a liquid-nitrogen-fed cryostat.H2O exposures took place at sample temperature 100 K.The sample temperature was ramped during TPD at 2 K/s rate from 100 to 650 K.The signal of hydrogen (2 amu) was recorded in parallel with H2O related signals (18 and 17 amu) and experimentally determined fragmentation ratios were used to subtract contribution of H2O fragments in H2 TPD.Other masses (namely 12-16, 28-32 and 44 amu) were recorded as well to verify an absence impurities in the desorption spectra.
LEED diffraction patterns were obtained at RT using Specs ErLEED 150 device equipped with AIDA electron energy controller and LEED image recorder.I-V LEED data were extracted from the experimental LEED snapshots using a software package ViPErLEED.S5

S1.3 Computational details
Density functional theory (DFT) calculations were performed using the plane waves pseudopotentials approach as implemented in the Quantum ESPRESSO code.S6,7 For Co and O we used the pseudopotentials from the standard solid-state pseudopotentials (SSSP Efficiency) library S8 (Co_pbe_v1.2.uspp.F.UPF and O.pbe-n-kjpaw_psl.0.1.UPF), while for Pt we used the ultrasoft pseudopotential for the PSLibrary 1.0 (Pt.pbe-n-rrkjus_psl.1.0.0.UPF).The plane wave cutoff was set to 50 Ry for the wave functions and 500 Ry for the charge density.Occupations were smeared with a Gaussian broadening of 0.01 Ry.We employed the PBE exchange and correlation functional and applied a Hubbard U correction of 3.0 eV to Co atoms, in agreement with our previous work on this system.S9 The Co3O4(111) surfaces were modeled using a slab geometry with periodic replicas separated by vacuum in the direction normal to the surface, in a (2×2) surface unit cell sampled at the Gamma point.We used a (4×2) surface unit cell to model larger Pt particles.The interface between Co3O4(111) and Pt(111) was modelled using a (1×1) surface unit cell of the oxide, sampling the system with 4×4×1 grid of k-points.Structural optimizations were performed with a convergence criterion for forces of 0.001 Ry/arb.units (0.025 eV/Å).
Adsorption energies of Pt atoms and small metallic Pt clusters supported on Co3O4(111) were computed as   = 1   (  −   −   ), where   is the total energy of the Co3O4(111) slab containing   atoms of Pt,   is the total energy of the Co3O4(111) slab without Pt, and   is the total energy of an isolated Pt atom.
Formation energies are defined in the framework of ab initio atomistic thermodynamics.S10 We treat the oxygen gas phase as a reservoir held at fixed chemical potential, set at values of temperature and partial pressure compatible with those of the experiments (T=300K, p(O2)= 2×10 - 6 mbar).The chemical potential of oxygen,   , depends on temperature and pressure according to: where   is the Boltzmann constant and  0 is the standard pressure, 1 bar.
The chemical potentials of Co and O atoms are related to the free energy per formula unit of Co3O4 via the following relation: ( 3  4 ) = 3  + 4  , and we approximate the free energy of the solid phases with their total energies.The formation free energy of a vacancy ∆   therefore reads: where   and  + are the total energies of the pristine surface and of the surface with a vacancy, respectively.The formation energy of a structure where Pt replaces a Co atom, ∆   , reads: where  @ is the total energy of the system where a Pt atom substitutes a Co atom and   is the chemical potential of Pt, that we take equal to the total energy of an isolated Pt atom.
The replaced Co atom is assumed to be incorporated into Co3O4, for example via diffusion to a step edge, and its chemical potential is therefore equal to   .
The STM images were computed using the Tersoff-Hamann approach, where the tunneling current measured with STM at point r is assumed to be proportional to the local density of states (LDOS) at point r integrated over the energy window between the Fermi energy EF and EF + eV, where V is the tip bias relative to sample and e is the charge of an electron: Consistent with experiment, we consider a tip bias of +2 V and -2 V, hence the STM images show a map of the occupied and empty states of the system, respectively, in an energy window of 2 eV below the Fermi energy.By considering the constant current mode, the plots show a map of the height above the surface required to maintain a set value of current.

S2. Analysis of STM images
Table S1: Classification of species I-IV observed in occupied and empty states of STM images (Figure1 e-h).Based on the presented statistics and the amount of deposited Pt, average number of Pt atoms in unresolved species X can be determined.For the sample 0.02 ML (Figure 1f, S1a), species X contain < 1 Pt atom indicating single-atom dispersion of Pt on the sample.This suggests that in this case species X reperesent a mixture of single-atom Pt features decorated by dissociated water molecules as identified in Figure S6.For sample 0.04 ML, the species X contain, on average 2 Pt atoms indicating that, besides unresolved single-atom features, small Pt aggregates appear on the surface.On the 0.1 ML sample, the species II, III and IV cannot be clearly resolved and all STM features are considered as species X. Species X are determined to contain, on average, 3 Pt atoms.
Still, up to 50% of species X in the STM images feature lateral sizes and apparent heights comparable to Pt single-atom features III or IV.Under assumption that the species X represent 50% of unresolved Pt single-atom features, and 50% Pt aggregates, average size of Pt aggregates is 2-3 atoms for the sample with 0.04 ML Pt, and 5-6 atoms for the sample with 0.1 ML Pt.

S3.1. Simulation of species I
In order to identify the origin of the species I, we simulated STM images resulting from the dissociated H2O yielding OHˉ at Co 2+ and H + at O 2 ˉ (Figures S2a-b), OHˉ at Co 2+ and H + at Co 2+ (Figures S2c-d), and isolated OHˉ at Co 2+ (Figures S2e-h) at 25% (e, f) and 100% (g, h) OHˉ coverage for both occupied (a, c, e, g) and empty states (b, d, f, h).
Additionally, we computed STM images resulting from the formation of triaqua complexes at Co 2+ (Figure S3a

S3.2. Simulation of species II
In order to identify the origin of the species II, we computed STM images for the defect-free (Figure S4a      With respect to the nature of species II appearing as dark depressions in the STM images (Figure  The R-factors obtained from the comparison of the IV-LEED curves for the structures of the asprepared Co3O4(111) film with respect to these obtained after the annealing at 650 K in UHV and exposure to 20 L of water at 300 K are summarized in Tables S4 and S5.
Generally, the IV-LEED data obtained from Co3O4(111) film before and after water exposure (Figure S9, red and black curves) exhibit slight differences with respect to these obtained from asprepared sample (Figure S9, green curve).We assign this effect to the adsorption of water from the background atmosphere.Based on our TPD studies, desorption of water from as-prepared and hydroxylated Co3O4(111) films corresponds to about 9% and 20% relative to the density of Co 2+ cations, respectively.A qualitative comparison with the data obtained by Meyer et al.S1 suggests that the differences are rather subtle.On the other hand, in the DFT calculations, H2O saturation of Co3O4 surface causes severe surface relaxations when Co atoms below OH move up from around 0.3 Angstroms relative to the oxygen layer in the clean surface to around 0.8 Angstroms in the presence of OH¯.We believe that this significant displacement of Co atoms in the presence of ordered (1×1) OH¯ layer would result in more substantial changes in the IV-LEED curves that are not evidenced in our control experiments.Therefore, we rule out formation of ordered (1×1) OH¯ layer.

S5. Adsorption of Pt atoms on Co3O4(111): DFT study
The adsorption of Pt atom on the Co3O4(111) surface at FCC site, where Pt is coordinated to three O 2-ions, yields the most favorable configuration with respect to on-top and bridge sites shown in  Looking at the Bader charges reported in Figure S10, we can see that the Pt atom transfers electronic charge to the oxide, and that the charge transfer correlates with the adsorption energy.
In order to assign the oxidation states of Pt, we computed the Bader charges in bulk Pt 2+ and Pt 4+ oxides.S11 We obtained the corresponding net Bader charges: 0.99e (PtO), 1.69e (α-PtO2), 1.73e (β-PtO2), 1.81e (β'-PtO2).Accordingly, we assign the oxidation states of Pt atom substituting Cot 2+ and Coo 3+ ions to Pt 2+ and Pt 4+ , respectively.We label Pt atoms chemisorbed on the surface as Pt δ+ , where δ is the partial charge which is significantly smaller than Pt 2+ state.

S6. Hydroxylation of Pt atoms
We have considered the adsorption of water on both the pristine Co3O4( 111) surface and on a Pt atom adsorbed on Co3O4(111).A single water molecule adsorbs on the pristine surface in a dissociated form, with the OH¯ group bound atop a Cot atom and a H atom bound to a surface O atom, with an adsorption energy of -1.00 eV.
In the presence of a Pt atom located in a FCC site, a single water molecule binds dissociatively, with the OH group bound atop the Pt atom and with an adsorption energy of -1.51 eV.Hence, water will preferentially bind to Pt atoms, if present, compared to surface Co atoms.The Bader charge on the Pt atom is +0.99e, compared to +0.79e in the absence of water.
We then considered the case when three water molecules adsorb at a Pt atom.Also in this case water binds in a dissociated form.The adsorption energy per water molecule is -1.In the case where Pt replaces a Cot site at the surface, we considered the adsorption of three water molecules at this site.We find that the molecular and dissociate forms of adsorption are energetically equivalent, with dissociated form favored by 0.01 eV per water molecule.The adsorption energy in this case is -1.27 eV per water molecule.The Bader charge on Pt changes from +1.13e in the absence of water to +1.87e in the presence of three water molecules.We therefore assign the Pt 4+ oxidation state in the latter case, showing that also in this case water adsorption drives a transfer of charge from the Pt atom to the Co3O4 slab.

S7. Dehydroxylated, hydroxylated, and moist Co3O4(111) substrates
In order to establish the role of water and OH¯ groups in the formation of species IV, we investigated the interaction of Pt with three well-defined Co3O4(111) surfaces denoted as (i) dehydroxylated, (ii) hydroxylated, and (iii) moist.The O 1s spectra obtained from the dehydroxylated, hydroxylated, and moist Co3O4(111) surfaces are shown in Figure S12.at 650 K in UHV followed by exposure to 20 L of water at 300 K (hydroxylated) and at 100 K followed by brief annealing at 170 K in UHV (moist).Besides the dominant contribution from the lattice oxygen in the Co3O4(111) film (red peak), the two peaks at higher binding energy arise from OH¯ groups and strongly chemisorbed H2O chem (blue peaks) and from physisorbed molecular water (green peak).Principally, the dehydroxylated surface should contain a single contribution from the lattice oxygen.However, a minor amount of OH¯ groups was also found on this surface.
We note that hydroxylation of Co3O4( 111) is a fast process in UHV and it is difficult to avoid a small amount of hydroxyls from background water on the timescale of the experiment.Typically, we observed some hydroxylation of Co3O4(111) at the present experimental conditions after 10 min in UHV at 300 K. On the hydroxylated sample, a strong contribution from OHˉ groups and strongly chemisorbed H2O chem was observed.Note that the presence of strongly chemisorbed water cannot be completely ruled out due to overlapping contributions.S12 On the moist sample, the O 1s spectrum contains contributions from the OHˉ groups and from physisorbed molecular water.

S8. Pt aggregates
In Figure S14, we show the optimized geometries of small Pt aggregates consisting of up to four Pt atoms adsorbed on the Co3O4(111) surface.We can see that the adsorption energy, normalized per Pt atom, drops as the cluster size increases, showing that the Pt-Pt interaction is stronger than the Pt-Co3O4(111) interaction.This trend is confirmed for one-dimensional structure consisting of one-atom thick and three-atom wide stripe of Pt atoms, and a rod consisting of three Pt layers discussed in Figure 3e and 3f, respectively (see main text).Among these, the latter structure can be considered a model representative of the metal/oxide interface found in large metal particles.S13 The starting geometries were selected from the atomic positions of bulk Pt.In the optimized geometry of the one-atom thick layer (Figure 3e), we find that the Pt atoms occupy positions on top of the surface O atoms.The adsorption energy per Pt atom of this system is -4.82 eV and all Pt atoms are positively charged, with an average charge of 0.27e.The three-layer rod has an adsorption energy of -5.02 eV and the first layer, in contact with the oxide surface, has positive charge of 0.22e, while the top two layers have a negligibly small charge.These results show that charge transfer between Pt and the oxide surface involves only the Pt atoms that are directly in contact with the surface.
These atoms acquire a partial positive charge, and are therefore labeled Pt δ+ .
Given the behavior of the adsorption energy per Pt atom as a function of cluster size reported in

S9. Dissociation of Pt δ+ aggregates in the presence of water
We considered the possibility of dissociation of ultra-small Pt δ+ aggregates driven by the reaction with water.We calculated the thermodynamics of such process at room temperature by varying the partial pressure of water.We find that a Pt4 cluster can adsorb 4 to 6 water molecules, depending on the pressure.We find that water always prefers to dissociate on Pt4.In several cases, when we start with molecular waters, these dissociate spontaneously.When they do not dissociate spontaneously, we find that the dissociated form is always more stable.Protons are in all cases bound to oxygen atoms of the support.The Pt1 atoms prefer to coordinate three water molecules.
To estimate whether water adsorption can drive the dissociation of a Pt cluster into Pt adatoms, in Figure S16 we plot the formation free energy of a Pt1 adatom adsorbing 3 dissociated water molecules (dashed black line) as well as the formation energy of a Pt4 cluster, normalized per Pt atom, with a varying number of adsorbed water molecules (we considered 1 to 9 water molecules, and we display 1 to 6 molecules in the figure), as a function of the water chemical potential, (H2O).Above a water chemical potential of -1.12 eV, corresponding to 10 -11 bar at T=300 K, the thermodynamically most favorable structure is an isolated Pt1 adatom with a triaqua complex.
At lower values of water chemical potential (i.e.lower water partial pressure) the most favorable structure is a Pt4 cluster covered with water.This analysis shows that the ability of Pt1 adatoms to bind more water molecules per Pt atom compared to Pt clusters can be the driving force for the dissociation of the clusters.
Note, however, that the error bar we have on the estimate of the pressure at which the transition takes place could be 3-4 orders of magnitude.

S11. Depth profiling of Pt oxidation state
In order to establish the origin of the Pt 4+ species re-appearing after annealing above 600 K in UHV, we performed the depth profiling of the Pt oxidation state using different photon energies.
The Pt 4f spectra obtained from 0.1 ML Pt/Co3O4(111) after annealing at 650 K in UHV are shown in Figure S20.The spectra were obtained with photon energies of 180 eV, 380 eV, and 650 eV.
The corresponding inelastic mean free path, IMFP, values in Co3O4 and Pt are listed in Table S6.
The spectra obtained with photon energy of 180 eV are the most surface sensitive.
Based on the evolution of the intensity ratio of the Pt 4+ contribution relative to Pt δ+ /Pt 2+ and Pt 0 , we conclude that the Pt 4+ species are located sub-surface.

Figure S1 .
Figure S1.Identification of observed features in STM images of Pt/Co3O4 systems with Pt coverages 0.02 ML (a, Figure 1f, empty states) and 0.04 ML (b, Figure 1g, occupied states).Clearly resolved species II and IV are marked with corresponding symbols II (red), III (green), and IV (cyan).Unresolved species are marked with X (black).The size of STM images in (a-b) is 15×20 nm 2 .
-b) and triaqua complexes at Co 2+ accompanied by one dissociated H2O molecule (Figure S3c-d) for both occupied (a, c) and empty states (b, d).

Figure S3 .
Figure S3.Simulated STM patterns for triaqua complexes at Co 2+ (a, b) and triaqua complexes at Co 2+ accompanied by one dissociated H2O molecule (c, d) for occupied (a, c) and empty states (b, d) obtained with tip bias voltages of +2.0 V and -2.0 V, respectively.

Figure S5 .
Figure S5.Simulated STM patterns for the Pt atoms substituting Co 2+ cations (a, b) for occupied (a) and empty states (b) obtained with tip bias voltages of +2.0 V (a) and -2.0 V (b).

Figure S6 .Figure S7 .
Figure S6.Simulated STM patterns for triaqua complexes at Pt atom substituting a Co 2+ ion (a, b), triaqua complexes at a Pt atom in FCC site (c, d), and triaqua complexes at a Pt atom in FCC site accompanied by one dissociated H2O molecule (e, f) for occupied (a, c, e) and empty states (b,d, f) obtained with tip bias voltages of +2.0 V and -2.0 V, respectively.The apparent height scale is adjusted to the apperent height scale of STM images in Figure2and TableS1.

Figure S8 .
Figure S8.Simulated STM patterns for Pt atom substituting Co 3+ cation (a, b) and Pt atom substituting Co 3+ cation with Co atom in FCC site (c, d) for occupied (a, c) and empty states (b, d) obtained with tip bias voltages of +2.0 V and -2.0 V, respectively.
FigureS4 (g, h).Therefore, we validated the assignment (i) by investigating the termination of Co3O4(111) film before and after the exposure to 20 L H2O at room temperature by means of IV-LEED.The results of the study are summarized in FigureS9.

Figure S9 .
Figure S9.IV-LEED curves obtained from Co3O4(111) after the preparation (green), annealing at 650 K in UHV (red), and exposure to 20 L H2O at 300 K (black).Experimental data are not normalized to the electron current on the screen.

Figure S10 .
Figure S10.We find, however, that a substitution of a surface tetrahedrally coordinated (t) Cot 2+ ion by Pt is much more energetically favorable.A large gain in the energy of formation (-0.92 eV) is achieved when Pt and an octahedrally coordinated (o) Coo 3+ ion just below the surface exchange places.

Figure S10 .
Figure S10.Side (top panels) and top (bottom panels) of adsorption geometries of Pt atoms in FCC (a) bridge (b), on-top (c) sites and Pt atom substituting surface Co 2+ cation (d).The Pt atoms are shown as grey balls.
49 eV, almost identical to the case where a single water molecule adsorbs at Pt.In this configuration, Pt is surrounded by an octahedron of O atoms and has a Bader charge of +1.75e, and is consequently in a Pt 4+ oxidation state.The adsorption of three water molecules therefore drives the oxidation of the Pt atom from Pt δ+ to Pt 4+ .The charge lost by Pt is transferred to the Co3O4 slab, where it is delocalized.

Figure S11 .
Figure S11.Side views of the optimized geometries of triaqua complexes formed at Pt atom in FCC site (a) and at Pt atom substituting Co 2+ cations (b).

Figure S14 .
Figure S14.Side views of the optimized geometries of small Pt aggregates on Co3O4(111).

Figures S14 and
Figures S14 and Figure 3e-f, our results indicate that there is a strong thermodynamic driving force for Pt atoms to coalesce into Pt particles.The adsorption energy drops as a function of cluster size, approaching the cohesive energy of bulk Pt which is -5.56 eV.The electronic structure of the Pt4/Co3O4 system is shown in FigureS15, which displays the projected density of states.It is clear from FigureS15that the presence of Pt atoms closes the Kohn-Sham gap, making the system metallic.The electronic structure of the oxide, however, is not perturbed strongly by the presence of the metal.

Figure S15 .
Figure S15.Projected density of states for the Pt4/Co3O4 system.The Fermi energy is set to zero.

Figure S16 .Figure S17 .
Figure S16.Formation free energy of Pt clusters with adsorbed water molecules plotted against the chemical potential of water.The horizontal bar at the top of the figure shows the values of water partial pressure at T=300 K at corresponding values of water chemical potential.The vertical grey line at (H2O)=-1.12 eV highlights the transition between Pt1 and Pt4.

Figure S18 .
Figure S18.TPD spectra of H2O obtained from moist Co3O4(111) (black) and moist Pt/Co3O4(111) (red) samples following the exposure to 20 L H2O at 100 K.The spectra are shown after subtraction of low-temperature desorption peak.

Figure S21 .
Figure S21.Side views of the optimized structures consisting of a Pt5, Pt4Co and Pt3Co2 clusters.

Table S3 .
Apparent heights of structural features in simulated STM patterns, relative to the top of terminating Co 2+ atom of clean Co3O4(111).

Table S4 .
The R-factors resulting from the comparison of the structures of the as-prepared Co3O4(111) film to the film after annealing at 650 K in UHV.

Table S5 .
The R-factors resulting from the comparison of the structure of the Co3O4(111) film to the film after exposure to 20 L H2O at 300 K.

Table S6 .
The inelastic mean free path, IMFP, of photoelectrons emitted from Pt 4f core level as a function of photon energy.The IMFP are given for Co3O4 and Pt materials.S14