Ni modified Fe3O4(001) surface as a simple model system for understanding the Oxygen Evolution Reaction

Electrochemical water splitting is an environmentally friendly technology to store renewable energy in the form of chemical fuels. Among the earth-abundant first-row transition metal-based catalysts, mixed Ni-Fe oxides have shown promising performance for effective and low-cost catalysis of the oxygen evolution reaction (OER) in alkaline media, but the synergistic roles of Fe and Ni cations in the OER mechanism remain unclear. In this work, we report how addition of Ni changes the reactivity of a model iron oxide catalyst, based on Ni deposited on and incorporated in a magnetite Fe3O4 (001) single crystal, using a combination of surface science techniques in ultra-high-vacuum such as low energy electron diffraction (LEED), x-rays photoelectron spectroscopy (XPS), low energy ion scattering (LEIS), and scanning tunneling microscopy (STM), as well as atomic force microscopy (AFM) in air, and electrochemical methods such cyclic voltammetry (CV) and electrochemical impedance spectroscopy (EIS) in alkaline media. A significant improvement in the OER activity is observed when the top surface presents an Fe:Ni composition ratio in the range 20-40%, which is in good agreement with what has been observed for powder catalysts. Furthermore, a decrease in the OER overpotential is observed following surface aging in electrolyte for three days. At higher Ni load, AFM shows the growth of a new phase attributed to an (oxy)-hydroxide phase which, according to CV measurements, does not seem to correlate with the surface activity towards OER. EIS suggests that the OER precursor species observed on the clean and Ni-modified surfaces are similar and Fe-centered, but form at lower overpotentials when the surface Fe:Ni ratio is optimized.


Introduction
Currently most energy sources used by our society are based on fossil fuels. Their combustion (coal, oil, and gas), together with large-scale deforestation, is causing massive emissions of greenhouse gases. Given the destructive environmental impact of these gases, effort has focused on the production, storage and transport of renewable energy (wind or sunlight) 1 . A promising technology to address this issue uses renewable energy to produce chemical energy through the splitting of water into hydrogen and oxygen (water electrolysis) 2 . However, the efficiency of the electrolysis process is hampered by the sluggish kinetics of water oxidation to O2, also known as oxygen evolution reaction (OER). This reaction has been described as the bottleneck of the water splitting and understanding its mechanism at the atomic scale could be a first step in addressing this challenge 2 . Many catalysts have been proposed to reduce the overpotential losses for OER and investigated in different pH conditions 3 , from acidic (2H2O → 4H + + O2 + 4e -) to alkaline (4OH -→ 2H2O + O2 + 4e -) media. In acidic media, noble metals such as Ru or Ir show promising OER stability and activity. However, due to their limited availability and high price many researchers are seeking alternative catalysts based on earth-abundant elements 3,4,5,6 . In an alkaline environment, oxides and hydroxides of late first-row transition metals (Mn, Fe, Co, Ni) have been found to have comparable performances to noble metals 3 . In particular, NiFe-based (oxy)hydroxide catalysts are reported to show the lowest overpotential for OER in alkaline conditions (pH 13 and 14) 7 , but the synergistic role of Fe and Ni is still under debate.
Comparing OER catalysts is complicated by many underlying factors, including differences in electrochemically active surface area, catalyst electrical conductivity, surface chemical stability, surface composition, and reaction mechanism. In this work, we describe our efforts to circumvent these issues by using a combined surface science/electrochemistry approach. We have prepared well-defined Nimodified Fe3O4(001) surfaces in ultra-high-vacuum (UHV) with different Fe:Ni ratios and, after characterization with surface science techniques, we have studied their electrochemical performances towards OER using cyclic voltammetry and electrochemical impedance spectroscopy. A significant increase in the OER activity is observed as the Ni content increases, and the optimum composition has ratios of Fe:Ni in the top surface layer in the range of 20-40%. These results are in good agreement with literature for the best OER powder catalysts 7 . Furthermore, based on the analysis of the surface morphological changes before and after reaction, together with adsorption capacitance measurements, we propose that the active sites responsible for the formation of the OER precursor are the same on the clean and on the Ni-modified magnetite. Nevertheless, the presence of the Ni on the surface shifts the formation of this precursor to lower overpotential.
Our study provides a well-defined model catalyst that is at the same time simple, highly active, and stable under operation conditions, and therefore ideal to be used as model system to gain atomic scale insights into the complicated OER mechanism.

Experimental Details
UHV preparation and characterization. The experiments were performed on a natural Fe3O4(001) single crystal (SurfaceNet GmbH) prepared in UHV by cycles of 1 keV Ar + sputtering and 900 K annealing. Every other annealing cycle was performed in an O2 environment ( 2 = 5×10 -7 mbar, 20 min) to maintain the stoichiometry of the crystal selvedge. Surface analysis was performed in a UHV system with a base pressure <10 -10 mbar, furnished with a commercial Omicron SPECTALEED rear-view optics and an Omicron UHV STM-1. XPS data were acquired using non-monochromated Al Kα x-rays and a SPECS PHOIBOS 100 electron analyser at grazing emission (70° from the surface normal). The same analyser was used to carry out the low-energy He + ion scattering (LEIS) experiments (1.225 keV He + , scattering angle 137°), an exquisitely surface-sensitive technique. Ni was deposited using a Focus electron-beam evaporator, for which the deposition rates were calibrated using a temperature-stabilized quartz crystal microbalances (QCM). One monolayer (ML) is defined as one atom per (√2 × √2)R45° unit cell, which corresponds to 1.42 × 10 14 atoms/cm 2 . Ni depositions higher than 2 ML were prepared by first depositing 2 ML Ni on the surface at room temperature, followed by mild annealing at 200°C for 10 min. This causes a transition from Ni being present as 2-fold coordinated adatoms to 6-fold coordinated "incorporated" cations 8 , see Fig. 1; the procedure was then repeated as many times as necessary to reach the desired coverage.
Characterization in-air. After UHV-preparation and characterization as well as after the electrochemical measurements, the samples were brought to air and imaged using an Agilent 5500 ambient AFM in intermittent contact mode with Si tips on Si cantilevers. Electrochemical measurements. Cyclic voltammetry and impedance spectroscopy were performed using a Metrohm-Autolab PGSTAT32 potentiostat and a custom-made electrochemical flow cell (made from perfluoroalkoxy alkane, PFA), mounted to the vacuum chamber. Prior to experiments, the chamber was filled with Ar (99.999%, Air Liquide, additionally purified with Micro-Torr point-of-use purifiers, SAES MC50−902 FV) to ambient pressure. The contact between sample and flow cell was sealed with Kalrez Orings. Prior to measurements, the electrolyte reservoir was evacuated and ultrasonicated to remove dissolved CO2. The flow cell was filled with electrolyte by increasing the pressure in the electrolyte compartment with Ar to slight overpressure. A glassy carbon counter electrode and a leak-free Ag/AgCl reference electrode (Innovative Instruments Inc.) were used. For impedance measurements, the latter was coupled to a glassy carbon quasi-reference electrode through a 100 nF capacitor. All electrochemical data were corrected for iRu drop; the uncompensated solution resistance Ru was determined from impedance Nyquist plots by extrapolating the minimum total impedance in the linear regime between 10 kHz and 100 kHz. All electrochemical potentials are referred to either the measured Ag/AgCl reference electrode EAg/AgCl or given as the overpotential η, which was determined via the equation η = EAg/AgCl+ERHE−1.229−iRu. ERHE is the potential of the reversible hydrogen electrode (RHE) vs a Ag/AgCl electrode. The potential of the RHE (Hydroflex) was measured before and after the electrochemical measurements to improve consistency of the results. The electrolyte was prepared from level-1 water (Merck Milli-Q, ρ= 18.2 MΩ·cm, 3 ppb total organic carbon), and reagent-grade NaOH (50 mass % in water, Sigma-Aldrich). Prior to use, all glassware and PFA parts where cleaned by boiling in 20% nitric acid and copious rinsing with Milli-Q water. Figure 1a shows a schematic model of the UHV-prepared Fe3O4(001) surface. The surface is oxidized with respect to the bulk Fe3O4 and is not a simple bulk truncation. Specifically, an interstitial tetrahedrally coordinated iron in the second layer (Fetet, light blue in the model) replaces two octahedrally coordinated iron atoms (Feoct, dark blue) in the third layer 9 , giving rise to a (√2 × √2)R45° periodicity. All surface Fe is in the 3+ state in the so-called subsurface cation vacancy (SCV) reconstruction, and it is the most stable termination of Fe3O4(001) over the range of oxygen chemical potentials encountered in UHV-based experiments 9 .

Characterization of the Catalyst Surface before Reaction
In the lower part of Figure 1a, a typical STM image of the UHV-prepared Fe3O4(001) surface is shown.
Undulating rows of surface Fe atoms appearing as protrusions run in the [110] direction. It is common to observe surface hydroxyl groups OsH (i.e. hydrogen atoms bonding to surface oxygen atoms, which are themselves not imaged) as bright protrusions on the Fe rows. This occurs because the hydroxyl modifies the density of states of the nearby Fe cations, causing them to appear brighter in empty-states STM images 10,11 . Figure 1a also displays other common defects visible on the clean surface, such as antiphase domain boundaries, which are imaged as meandering line defects, and unreconstructed unit cells, which appear similar to two neighboring hydroxyl groups. These are caused by two additional Fe atoms in the subsurface layer (instead of one interstitial Fe), which again modifies the density of states of the surface atoms 10,12 . It is not possible to image the surface oxygen in STM as they have no density of states in the vicinity of the Fermi level. However, their positions are exactly known from density functional theory calculations and quantitative low-energy electron diffraction (red in model in Figure 1a) 9 .
The surface reconstruction makes it possible to progressively modify the magnetite surface and accommodate foreign metal atoms (such as nickel) in specific positions. 8 Following Ni evaporation under the appropriate temperature conditions, it is possible to obtain two different Ni geometries: Ni adatoms 2-fold coordinated to surface oxygen atoms (model in Figure 1b, green) and incorporated Ni occupying octahedrally coordinated sites below the surface (model in Figure 1c) 8,13 . Ni deposition at room temperature leads to Ni adatoms in the 2-fold coordination, which are imaged in STM as isolated, bright protrusions appearing between the Fe rows (light blue circles in Fig. 1b). The transition from 2-fold to 6fold coordination is achieved by annealing the surface at 200 °C for 10 minutes. As the incorporated Ni are in the subsurface, they cannot be imaged directly in STM, but they modify the electronic structure of the nearby Fe cations, making them to appear brighter in empty-state images (red circles in Figure 1c) 8,13 .
Their appearance is similar to the unreconstructed cell discussed earlier (Figure 1a). Furthermore, the STM image in Figure 1c shows additional protrusions within the Fe rows (highlighted with yellow circles), which we previously assigned to Ni replacing Fe atoms in the 5-fold-coordinated position in the top-surface layer 13 .
The incorporation of Ni in the vacant subsurface octahedral site is only possible if the interstitial Fetet moves back into the other subsurface octahedral site of the unit cell. The resulting cation rearrangement closely resembles a bulk-truncated Fe3O4(001) surface 8,14 , and a (1 × 1) periodicity is observed in LEED. It is possible to recover clean (√2 × √2)R45° reconstructed surface by annealing to high temperatures, which causes the Ni atoms to diffuse into deep bulk layers. Hereafter, we deal exclusively with the incorporated Ni-doped magnetite shown in Fig. 1c, which resembles the structure of mixed spinel ferrite, i.e., a NixFe3-xO4-like system, suggested to be one of the active phases in OER 15,16 . After deposition of 1 ML, a small signal is observed in XPS at 855.5 eV, corresponding to the Ni 2p3/2 peak 17,7 . This is a higher binding energy than metallic Ni 17 , which, together with the strong satellite at ≈862 eV, indicates that the nickel is oxidized. Earlier DFT calculations predicted that incorporated Ni atoms are Ni(II) 8 , as in NiFe2O4.

Figure 1. Atomic models showing side views of the (a) Fe3O4(001) clean, (b) doped with 2-fold coordinated Ni adatoms, and (c) with 6-fold coordinated Ni incorporated, as well as corresponding STM
As the Ni deposition increases to 10 ML, the Ni 2p3/2 at 855.5 eV increases in intensity, together with the 861.9 eV and the 2p1/2 peak at 873 eV, which are harder to see at lower Ni coverage. These features increase in intensity as the Ni deposition increases up to 50 ML. At even higher Ni load (120 ML), two new signals at 853.1 eV and 870.2 eV emerge, indicating that metallic Ni is present on the surface 17 . Above 120 ML, the XPS spectrum changes shape to a peak with only two main features at 853.1 eV and 870.2 eV, indicating that the surface is fully covered with metallic Ni.
We imaged the Fe3O4(001) surface before and after Ni-doping using ambient AFM right after removing the crystal from the UHV chamber (Figure 3a Figure 3a, we suspect these to be residues originating from dust or carbonaceous species. The LEED pattern in the inset shows that the reconstruction spots are now absent and a (1 × 1) symmetry is observed (blue square), which is known to occur above 1 ML Ni atoms incorporated in the subsurface 8 .   Importantly, no systematic change in consecutive scans was observed, which rules out substantial damage to the surface by He + sputtering during LEIS measurements. In what follows, we will use the LEISdetermined Fe:Ni ratio to refer to our model catalysts.

Electrochemical Performance
The electrochemical performance of the clean and Ni-doped Fe3O4(001) surfaces was investigated using cyclic voltammetry. The overpotential required to reach a given current density is a key catalytic parameter to compare several catalysts and to estimate the energetic efficiency of integrated (photo-) electrochemical water splitting devices 3

(c) Overpotential for j = 5 mA·cm -2 (left axis) and Tafel slopes (right axis) values from (a) and (b) as a function of the surface Fe:Ni ratio (expressed in percent). This diagram shows that the lowest overpotential is obtained for an ideal surface Fe:Ni ratio between 15-40% range.
To check whether catalyst aging in electrolyte affects the activity, we performed cyclic voltammetry on the same surfaces after leaving the Ni-doped electrodes for three days in electrolyte. Figure 5a- Figure 3a-d after OER and three days aging in electrolyte. Before imaging, each surface was rinsed in milli-Q water several times, for several minutes and blow-dried using a gentle Ar flow to minimize the presence of salt residue from the electrolyte.

Figure 3a'-d' shows the AFM characterization of the surfaces imaged in
The morphology of the clean Fe3O4(001) remains unchanged after OER (a´), and shows a smooth appearance with the wide terraces and step bunches still visible, in agreement with earlier stability tests 18 .
The presence of small particles (white) is associated with residue from the electrolyte.

Electrochemical Impedance Spectroscopy
Electrochemical impedance spectroscopy (EIS) measurements were performed on the Ni-doped model catalysts electrochemically investigated in Figure 4. In the OER region the EIS Nyquist plots ( Figure S1a) exhibit two relaxation processes characterized by two semi-circles that can be assigned to two capacitances while the phase in Bode plots ( Figure S1b) exhibits two maxima eventually merging into a broad peak. This impedance behavior is consistent with previous measurements on metal transition oxides and perovskites during the OER. 25,26,27 The EIS response can be modelled by the equivalent circuit (EC) shown as an inset in Figure 6a with a double-layer capacitance (Cdl) in parallel with the combination of a polarization resistance (Rp) and an adsorption pseudo-capacitance (Cads) in parallel with a resistor Rs.
The Cdl element accounts for the charging of the electrified interface. Cads models the accumulation of an adsorbed intermediate involved in the rate-limiting step of the OER. The sum of the resistive elements Rs and Rp bear a physical meaning as the zero-frequency electron transfer resistance defined as Rf = Rp + Rs, i.e., the slope of the steady-state polarization curve after Ohmic-drop compensation. RΩ represents the electrolyte resistance. It has to be noted that both capacitors were modeled as constant phase elements (CPEs), defined as = =1 −1 ( ) − , where =1 −1 is the impedance of the capacitor without frequency dispersion, i.e., if the coefficient n = 1 which is the case for a perfect capacitor. The interpretation of the CPEs dispersion coefficient n is varied and complicated; its origin has been attributed to surface roughness, inhomogeneities, or inhomogeneous adsorption of ions 28 . In the double-layer region, prior to the onset of the OER, we will show in a separate work that the impedance response of the single-crystal magnetite electrode has to be modified by adding a Warburg element in series with Cads corresponding to a diffusion impedance that we attribute to electrolyte cations intercalating into the oxide ( Figure S2a). Of interest in this work is the impedance behavior in the OER region.
All the surfaces investigated in this work, with the exception of the one fully covered by metallic Ni clusters (light blue), show constant double layer capacitance values in the 10-25 F·cm -2 range prior to the OER onset ( Figure 6a). The exponent of the CPE element used for the fitting was equal to 1 in the double-layer region ( Figure S2d) and diverged from 1 at high current densities or when Ni is exposed such that Ni(OH)2 is oxidized to NiOOH. These values are comparable to a Cdl observed on metallic single crystals, suggesting that our catalysts have a perfect capacitor-like behavior. Figure 6c shows the value of this capacitance as a function of the Ni content: Cdl slightly increases from 10 to 15 μF cm -2 as the Ni loading increases, but a higher value is observed in the case of the surface fully covered with Ni metallic clusters (180 ML). The higher Cdl values observed for this surface may be explained by the formation of an irregular Ni(OH)2 layer upon oxidation of the metallic Ni by contact with the electrolyte. In this way, more active surface area is exposed to the electrolyte and polarized, leading to a higher Cdl.
The adsorption capacitance plot in Figure 6c shows that the Ni-doped Fe3O4(001) surfaces display a peak with similar Cads values independent of the Ni doping level, which however shifts to lower overpotential as the Ni load increases (Figure 6d). The surface fully covered with metallic Ni clusters appears to develop two additional capacitance peaks (Figure 6c). The overlay of the corresponding CV and Cads in Figure S3e, shows that the additional initial (pre-)peak is observed at the same potential as the Ni(OH)2 oxidation peak.
The group of Bandarenka 29,30 associated the observation of peaks or increase in Cads to the adsorption of OER reaction intermediates and reported them for various transition metal oxides. These observations suggest that the formation of the intermediate species before the onset of the OER involves similar mechanisms for pure and Ni-modified magnetite. This is also supported by the fact that value of Cads retains similar values at the maximum of the peak. From the capacitance data in Fig.6a and c we can draw the following conclusions: (i) given that the initial Cdl values hardly varies with Ni loading, there is no significant increase in electrochemically active surface area, and the catalytic effect of Ni shown in Fig.4 cannot be ascribed to an effective enhancement of the surface area; (ii) the fact that a similar peak in Cads is observed for all surfaces, also the one where Fe is expected to be the only active site (98:2), would be in agreement with the commonly held view that Fe is the active site in NiFe catalysts, but that it becomes more active in an Ni environment. The presence of two peaks in the EIS of the metallic Ni-decorated surface if not a noise effect can be interpreted as two types of adsorbates on Ni (and perhaps Fe) sites that are accessible due to the porosity and layered structure of Ni films that allow access to active sites down to 5 nm in depth 31 .

Discussion
The experimental data acquired on clean and Ni-doped Fe3O4(001) surfaces show that Ni doping enhances the OER activity of magnetite. Electrochemical voltammetric responses, in combination with surface sensitive techniques, suggest a strong dependence of the OER activity from the atomic structure of the surface exposed to electrolyte. In particular, LEIS measurements indicate that the catalyst with the best OER performances, with an overpotential of 340 mV vs RHE at 5 mA·cm -2 , exhibits a surface Fe:Ni ratio of 40:60.
In order to shed light on how the presence of Ni affects the magnetite atomic surface structure-activity relationship, the following observations have to be considered: We have previously shown that following In particular, Burke et al. 23 observed that electrochemical cycling leads to a transformation from nano-crystalline NiOx films to a layered (oxy)-hydroxide that correlates with an increase in OER activity. Similarly, Deng et al. 33 monitored the dynamic changes of single layered Ni(OH)2 using in situ electrochemical-AFM, and observed dramatic morphology changes already after one linear voltammetry sweep, as well as a direct relation between increase in OER activity and increase in volume and redox capacity of the layered oxy-hydroxide phase. Our results are, however, are not entirely in agreement with these observations.
The increase in volume and surface area of the hydroxide phase does not correlate directly with our catalysts' activity: the surface with the highest amount of (oxy)-hydroxide phase and redox capacity is ≈ 200 mV less active than the (almost) flat surface with Fe:Ni ratio 40:60. Based on these observations, we can rule out the possibility of the layered Ni-(oxy)-hydroxide as the active phase in our catalysts.
It is also important to mention that the activity exhibited by the surface prepared with a Fe:Ni ratio of 40:60, with an overpotential of 340 mV vs RHE is comparable to values reported for OER on (Fe)Ni based catalysts [34][35][36][37][38][39][40] . For comparison, the overview in Table 1 shows a selection of some of the best OER catalysts based on Ni-Fe oxides reported in literature. The lowest overpotential values measured on these catalysts at 5 mA·cm -2 vary typically in the 210 -347 mV vs RHE range (in 1 M KOH or NaOH electrolyte). Similar overpotential values were also obtained from our surface prepared with a higher Ni load (Fe:Ni = 15:85).
When comparing the latter surface with the one having an Fe:Ni ratio of 40:60 after electrochemical cycling and subsequent aging for three days in electrolyte, a different activity trend is observed. On the one hand, both surfaces show a significant increase in activity following voltammetric cycling and aging, in agreement with previous studies 3, 33 . On the other hand, their activity does not increase in the same way. Surprisingly, the surface with metallic Ni shows a much lower onset of the OER overpotential than the one with an Fe:Ni of 40:60 , despite the similar performances when freshly prepared. This surface is by far the most active with an overpotential of 247 mV vs RHE. However, it has to be taken into account that this surface, being characterized by the presence of metallic Ni, does not show a well-defined spinel structure and, therefore, cannot serve as a model system. Since one of the scopes of this work is to propose a working model system for the understanding of the OER mechanism, a compromise between activity and the ability to preserve atomic control has to be made. In this regard, the surface with a Fe:Ni ratio of 40:60, very well defined and highly active, fits the criteria to be used as model catalyst.  29,30 and the shift to lower overpotentials as the Ni doping increases and finally reaches a steady value whit the optimal Ni content (Ni content ≈ 60-80%).
To explain these observations, we propose the following scenario: the intermediate species that forms on the surface before the onset of the OER might be the same on the clean surface as well as on the Nimodified one, indicating Fe as the active sites. Accordingly, the right amount of Ni in the spinel surface does not cause the intermediates formation but facilitates it. Similar conclusions have been proposed by Bell and co-workers who used DFT to compared the OER activity of pure and Fe-doped γ-NiOOH and of pure and Ni-doped γ-FeOOH catalysts 41 . They showed that pure γ-NiOOH adsorbs the OER intermediates too weakly and pure γ-FeOOH too strongly. They found a considerable increase in activity for Fe sites that are surrounded by Ni next-nearest neighbours in both γ-NiOOH and γ-FeOOH. Similar results have also been obtained by Klaus  Moreover, it should be pointed out that the use of a single crystal enables an accurate determination of the electrochemically active surface area (ECSA) of these materials and provides reference values for the double-layer capacitance and adsorption capacitance on Fe-Ni based catalysts. The double-layer capacitance values are slightly affected by the Fe:Ni ratio and this should be taken into consideration for further determination of the ECSA of such electrodes 29,30 . Additionally, our results point out that, beyond the Fe:Ni ratio, the nature of the interface (spinel or separated NiOOH/Fe-Ni spinel) significantly affects the capacitance of the interface and its use as a reference for ECSA determination could be compromised.

Conclusions
The high intrinsic OER activity of mixed Fe-Ni oxides motivated our efforts to make further steps in the understanding of the fundamental roles of Fe and Ni in OER catalysis.
In this work, we show a combined surface science/electrochemistry approach for the preparation of welldefined Ni-modified Fe3O4(001) surfaces and the investigation of their electrochemical performances with respect to OER. We have found that the surface prepared with an Fe:Ni ratio of 40:60 shows performances comparable to those of the best powder catalysts reported in literature, and still maintains a well-defined structure. Being at the same time simple, highly active, and stable under operation conditions, this surface is an ideal candidate to serve as a working model system to gain atomic-scale insights into the complicated OER mechanism. Electrochemical impedance spectroscopy enables us to confirm that on our Fe-Ni catalyst, the active site for the OER is located on Fe atoms at the surface regardless of the Ni:Fe ratio in the structure.
Putting our results in the context of future perspective, a well-defined model system such as the Nimodified Fe3O4(001) presented in this work is desirable to address the fundamental aspects that are still controversial. With a limited variety of possible adsorption sites and being accessible to methods benefitting from on single-crystal surfaces, this model surface could thus be used for further investigations on the exact nature of the adsorbates involved in the rate limiting step, using in-situ surface science techniques, to shed more light on key parameters to improve the stability and activity of amorphous catalysts used in water splitting devices. We also believe that the good agreement of our results with what reported in the literature for powder or amorphous catalyst makes our model surface worthwhile to be used as a model to guide future computational studies.