Optimized Pt-Based Catalysts for Oxygen Reduction Reaction in Alkaline Solution: A First Principle Study

The combined density functional theory (DFT) and kinetic Monte Carlo (KMC) is employed to understand the capability of Pt alloys via Pt-X-Pt sandwich motif in tuning the oxygen reduction reaction activity and stability of Pt catalysts in alkaline solution. For X = Fe, Co, Ni, Cu, the structure of Pt-X-Pt alloy is likely to stay under reaction condition for lowering the surface energy. Both Co and Ni are identiﬁed as promoters, being able to facilitate the removal of hydroxyl from the surface and the ORR; while all four systems show the enhancement in the stability of surface Pt compared to pure Pt. For X = Ti, Mn, Ce, the alloyed X metal is too active, which tends to anti-segregates to the Pt surface and forms oxides due to the strong X-O interaction. Wherein, the decoration of Ce oxide shows a promoting effect for the ORR on Pt, which induces strain on the neighboring Pt-Pt bonds and helps in release of hydroxyl species; yet it destabilizes the interacted Pt atom and can lead to the deactivation of the catalyst.

The proton exchange membrane fuel cell (PEMFC) is the most popular fuel cell which can work at a low temperature in atmosphere. 1,2 However, due to the high corrosive electrolyte, the PEMFC highly relies on Pt electrodes. Though many efforts have been made, there is still no viable substitute for Pt. 3 Alkaline fuel cell (AFC) uses alkaline solution as the electrolyte, which is less corrosive than PEMFC, making it possible to use some non-Pt catalyst as electrode materials. 4,5 With the recent development of anion-exchange membranes, 6-10 AFC shows great potential of a competent clean energy solution. Yet, like the PEMFC, Pt is still the best cathode material for AFC. 11 Our previous study 12 built a detailed reaction network for the oxygen reduction reaction (ORR) on Pt (111) surface in alkaline solution. The DFT calculations and kinetic Monte Carlo (KMC) simulations were combined, being able to well reproduce the experimental measured polarization curve. 13 In addition, the hydroxyl species ( * OH) binding and oxygen ( * * O 2 ) binding were identified as the key descriptors. To enhance the ORR activity of Pt (111), the binding property should be tuned to provide the weakened binding to * OH and the strengthened binding to * * O 2 . Such mechanistic understanding provides the basis for the catalyst optimization from theoretical calculations.
Here, built on the key binding descriptors identified previously, the combined DFT and KMC study was performed to modify the Pt catalysts using cheap metal X (X = Fe, Co, Ni, Cu, Ti, Mn, Ce), aiming to enhance the ORR activity in alkaline solution, elongate the life cycle and reduce the precious metal loading. For decades, extensive efforts have been made to optimize the Pt-based ORR catalysts in PEMFC. [14][15][16][17][18][19][20] One of the commonly used methods is to form Pt alloys. Although alloying with cheap metals, such as Ni and Co, was predicted to improve the ORR activity via either X@Pt core-shell or Pt-X-Pt sandwich motifs, 21,22 the low stability of Ni or Co under the ORR conditions in acid solution has long been an issue. To stabilize the X metal, the formations of metal alloy with noble metals, e.g. Au and metal compound with light element, e.g. N, were found to be effective. [23][24][25][26] In the case that the alloyed X metal is too active, it is easily oxidized during the reaction. It was reported that the formed metal oxides on Pt can improve its stability. 27,28 By comparison, much less attention has been paid to the catalysis study in AFC. So far, most of the studies focus on Ag-based core-shell, e.g. AgCo@Ag, catalysts. 4 In the current study, the optimization of Pt alloys via the Pt-X-Pt sandwich motif was considered to improve the ORR performance z E-mail: pingliu3@bnl.gov of Pt catalysts in alkaline solution. For X = Fe, Co, Ni, Cu, the structure of Pt-X-Pt was assumed to stay under reaction condition for maintaining the low surface energy. 29 Both the extended Pt-X-Pt(111) slab model and the Pt-X@Pt core-shell particle model were considered to describe the size effect, which was shown to have significant impact on both activity and stability. 26,30 In the case of X = Ti, Mn, Ce, the alloyed X metal is too active, which likely anti-segregates to the Pt surface and forms oxide (TiO x , MnO x and CeO x ) due to the strong X-O interaction. 27 Our study enables the identification of the possible promoters to tune the activity and stability of Pt-based catalysts on the mechanistic understanding.

Methods
DFT calculations.-All the DFT calculations were carried out with VASP 31-36 code with projector augmented wave method. 37 The exchange-correlation effects were described with Perdew-Wang functional (GGA-PW91) pseudo potential. 34 The convergence of energy and force were set to 1 × 10 −5 eV and 0.01 eV/Å, respectively, and the cutoff energy was set to 500 eV. All the calculations were spin polarized.
For the Pt-X-Pt(111) (X = Fe, Co, Ni, Cu) sandwich model, a five-layer of (111) slab model with 3 × 3 array in each layer was built (Figure 1a). A 3 × 3 × 1 k-point sampling method was used for the supercells. 38 A 20 Å vacuum was included in the supercell between slabs. In the DFT calculations, the bottom two layers were fixed in the Pt bulk positions while the top three layers and adsorbates were fully relaxed. To describe the Pt-X@Pt core-shell nanoparticle, the truncated octahedron models were built (Figure 1b). Like the slab model, the nanoparticle model included 3 parts: a Pt monolayer shell, a X interlayer and a Pt core. Due to the calculation limit, only half of the truncated octahedron was considered in DFT calculations with particle size around 2.05nm. For the nanoparticles, the atoms of the bottom layer were allowed to relax only in-plane, rather than perpendicular to the plane, while the other atoms were fully relaxed with adsorbates. For comparison, the (111) facet was chosen to test the binding energies of the ORR intermediates. Our previous studies 24, [39][40][41] showed that such half-sphere structure and binding on the (111) facet successfully described the size and shape effects on the ORR performance of the Pt-based catalysts. We also tried active metals, such as Ti, Ce and Mn, to tune the Pt catalysts. To simulate the effect of metal oxide, the hydroxylated metal oxide trimer (Ti 3 O 7 H 7 , Mn 3 O 9 H 9 , Ce 3 O 7 H 7 ) was deposited on Pt (111) using a 4-layer slab and 5 × 5 array in each layer (Figures 1c-1e). The 2 × 2 × 1 k-point sampling was used. The bottom two layers were fixed in the Pt bulk positions. The top two layers of Pt, the hydroxylated metal oxides and the adsorbates were fully relaxed. DFT+U method was used for Ce, Ti and Mn and the U parameters were set as 4.5, 4.5 and 4, respectively. [42][43][44][45][46][47] Four ORR intermediates on surface: * * O 2 , * * OOH, * O and * OH. The binding energies were calculated as E binding = E(Adsorbate/Surface) -E(Adsorbate) -E(Surface), where E(Adsorbate/Surface), E(Surface) and E(Adsorbate) represent the total energy for the surface in interaction with the adsorbate, bare surface and the adsorbate alone, respectively.
KMC simulations.-In our previous study, the KMC package we developed for the ORR on Pt(111) and Ag(111) simulated the polarization curves that agreed well with experimental results. 12,13,48 According to the reaction network proposed previously for Pt(111), 12 multiple reaction pathways were considered for the ORR overall modified Pt systems. It included the 4e − mechanism via the new chemisorbed * H 2 O-mediated and non-potential dependent reaction pathway in addition to the conventional potential-dependent dissociative and associative pathways. Besides, we also considered the OOH − dissolution step to describe the 2e − mechanism. The rate constant for each elementary step was calculated with Arrhenius equation k = υ Exp(-E a /RT), where υ is the pre-exponential factor, E a is the activation energy, R is the gas constant and T is the temperature. The pre-exponential factor υ was set to 1 × 10 9 for potential dependent reactions and 1 × 10 13 for potential independent surface reactions. 49 The activation energies for surface reaction steps were estimated based on the DFT-calculated reaction energies via the BEP relationships. 12,50 The temperature of all simulations was set to 298K.

Results and Discussion
Our previous study 12 showed that the Pt catalysts can be promoted by weakening the * OH binding to facilitate the dissolution from the surface and strengthening the * * O 2 binding to hinder the desorption and promote the sequential activations. In order to achieve this goal, the Pt alloys via the Pt-X-Pt sandwich motif was used. As indicated above, for X = Fe, Co, Ni, Cu, the structure of Pt-X-Pt alloy slabs and nanoparticles were adopted, while for the more active X (X = Ti, Mn, Ce), a likely oxide (TiO x , MnO x and CeO x ) cluster formed over the Pt(111) surface was constructed. Our goal is to not only shift the binding energies of ORR intermediates, especially * OH and * * O 2 in our reaction network, to the desired value, but also tune the stability of surface Pt layer.
Pt-X-Pt (111) slab motif.-It has been shown previously that the big Pt nanoparticles (>4 nm) are believed to have similar properties with Pt bulk. 29,40,41,51 Thus, the Pt-X-Pt(111) slab model was used to represent the Pt-X-Pt core-shell particles with relatively big size. Such structure is likely to form in Pt-rich PtX alloys. Due to the lower surface energy of Pt than Co, Ni, Fe and Cu, Pt atoms prefer to segregate to the surface forming a Pt monolayer, which results in the X-rich is the sublayer beneath. 52 Within this structure, the supercell kept the lattice of Pt bulk. Thus, different from the X@Pt core-shell structure, there was no surface strain imposed on Pt monolayer due to the presence of X underneath. 53 The formation of Pt-X-Pt(111) slab model system has been shown to be an effective way for Pt alloys to improve the catalytic activities of Pt alone. [53][54][55][56][57][58] For Pt-X-Pt(111), the Pt-Pt bond length on the surface is 2.82 Å, which equals to that of pure Pt bulk. By comparison, the Pt-X bond length (2.62 Å for Pt-Ni; 2.63 Å for Pt-Co; 2.64 Å for Pt-Fe and Pt-Cu) is shorter. The induced compressive strain perpendicular to the surface can tune the binding energies of the ORR intermediates on Pt surface, which has been also observed previously. 51 Indeed, as shown in Figure 2a, the interaction between the ORR intermediates and Pt surface layer is weakened going from Pt(111) to Pt-X-Pt(111). This is due to the fact that the d-band center of surface Pt is down-shifted by the perpendicular compression. 59 The only exception is the binding energy of * O on Pt-Cu-Pt(111), which is slightly stronger than that on Pt (111). Our previous study 29 showed that the Cu-Pt interaction is weaker than the other Pt-X, and the surface Pt atoms above the Cu monolayer have higher flexibility. When interacting with the strong adsorbate like * O, the Pt atoms shift significantly to provide the strong binding. In this case, the Pt-Pt bond length is elongated from 2.82 Å to 3.03 Å to form the strong Pt-O bond. In general, the formation of all Pt-X-Pt(111) helps to weaken the binding of Pt with the ORR intermediates, where the weakening effect increases in a sequence of Cu < Co < Ni < Fe (Figure 2a).
We also considered the effect of solvation on the binding energy, which was previously identified as an essential factor to model the ORR, 12,48 where one ice-like bilayer of water molecules was found to . KMC-simulated ORR polarization curve on Pt-X-Pt(111) (X = Fe, Co, Ni, Cu) sandwich motif. The pure Pt data was adopted from our previous work. 12 be suitable to simulate the solvation effect at the electrolyte-electrode interface. When the solvation effect is included (Figure 2b), the weakening effect in Pt-adsorbate interaction still exists; however, the difference from one Pt-X-Pt to the next decreases significantly. This can be attributed to the direct interaction between water molecules and the ORR intermediates. In this case, the calculated binding energies include not only the adsorbate-Pt interactions but also adsorbate-water interactions via hydrogen bonds. In addition, the water-water interaction and water-Pt interaction also contribute. Among all the ORR intermediates, the presence of water affects the binding energy of * O the most significantly (Figures 2a, 2b). This can be explained by the destructive adsorption of * O at the water/Pt(111) interface. According to our previous study, 12 the adsorption of * O partially breaks the ice-like structure of water bilayer on Pt(111), which decreases the stability of overall structure drastically.
According to our previous study, 12 a weakened Pt-OH binding is one of the criteria to improve the ORR on Pt(111). Fortunately, all Pt-X-Pt(111) model systems tested here show the weaker * OH binding ( Figure 2). Besides, the strengthening in Pt-O 2 binding is also necessary, which cannot be achieved by any of Pt-X-Pt(111). In addition, the bindings of O-containing species on the surface also vary from one Pt-X-Pt(111) to the next. Although it may not affect the ORR activity as significantly as * OH and * * O 2 , the bindings of different intermediates should be balanced to facilitate the reaction. Due to the complexity of ORR mechanism, it is not straightforward to predict the overall activity merely based on the DFT calculated binding energies. To tackle that, we performed the KMC simulations to estimate the polarization curve based on the DFT calculated energies, which is an important experimental observable to evaluate the ORR catalysts. Here we used the E binding with one bilayer of water (Figure 2b) to calculate the reaction energy E and then estimate the barriers following the methodology employed previously for Pt(111). 12 According to the KMC simulated polarization curve (Figure 3), it is clear that Pt-Ni-Pt(111) and Pt-Co-Pt(111) display the higher halfwave potential and current density at the high potential (U) region (U > 0.8 V vs. RHE) than Pt(111), where Pt-Ni-Pt(111) is slightly better than Pt-Co-Pt(111). The promotion effect is associated with the reduced * OH binding on Pt-Ni-Pt and Pt-Co-Pt compared to that on Pt, which enables the higher current density at U > 0.8 V vs. RHE. However, at the low potential region, the lower maximum current densities are observed for Pt-Ni-Pt(111) and Pt-Co-Pt(111) than that of Pt(111) (Figure 3). This is associated with the weakened * * O 2 binding on Pt-Ni-Pt(111) and Pt-Co-Pt(111), which makes it easier for * * O 2 to desorb from the surface rather than being reduced to * * OOH and contributing to the current. Pt-Cu-Pt(111) is an extreme case, where the Pt-O 2 binding is overweakened ( Figure 2b) and the ORR is completely terminated (Pink line in Figure 3). Pt-Fe-Pt(111) provides a moderate * * O 2 binding as Pt-Co-Pt(111) and Pt-Ni-Pt(111), while the weakening in * OH binding is more significant (Figure 2b). Consequently, the corresponding current density drops at a lower potential than Pt(111) (Green line in Figure 3). According to the KMC simulation, in this case neither * OH nor * * O 2 binding is important. Instead, the * * OOH binding is important, which is weakened too much and the reduction of * * O 2 to * * OOH is slowed down. The potential dependent * * O 2 reduction is very sensitive to potential, and the current density drops rapidly as potential increases. The comparison in the polarization curve indicates that the weaker * OH and * * O 2 bindings than Pt are necessary to promote the ORR on Pt. However, the corresponding magnitude should be moderate and the over-destabilization for * OH and/or * * O 2 can hinder or completely terminate the reaction.
To estimate the effect on stability, the dissolution potential (U D ) were estimated following the previous study by Greeley and Nørskov. 60 The results show that all Pt-X-Pt(111) systems are able to raise U D on Pt(111) and therefore the stability of surface Pt is enhanced. Wherein, Pt-Fe-Pt(111) displays the highest capability corresponding to an increase U D from Pt (111)  Pt-X@Pt core-shell nanoparticle motif.-The small nanoparticles (<4 nm) has a significant effect on the ORR activities, according to our previous studies. 41,51 Since the Pt-X-Pt(111) slab model is not sufficient to describe such nanoscale effect, cluster models were built using half of a truncated octahedron (∼2 nm, Figure 1b) to reduce the computational cost. A Pt nanoparticle in same size and shape was also included as the reference. The central site on the (111) facet was chosen to study the binding energies of the ORR intermediates, which successfully described the measured size and shape effects on the ORR activity of core-shell nanoparticles experimentally. 24,26,[39][40][41] Here we chose X = Ni, Co to build the Pt-X@Pt particle models, as the corresponding Pt-X-Pt (111) slab models showed higher ORR activity than Pt (111). Like the slab model, both Pt-X@Pt core-shell nanoparticles show the weaker interaction with the ORR intermediates than the Pt nanoparticle ( Figure 4). In term of magnitude, the weakening effect by nanoparticles is more significant than those of slabs (Figures 2a  and 5). According to our previous study, 26,30 this is due to the additional surface contraction introduced by the subsurface X atoms, when adopting the nanoparticle motif. The Pt-Pt bond length on the top (111) facet is 2.69 Å on both Pt-Ni@Pt and Pt-Co@Pt nanoparticles, which is shorter than 2.72 Å on Pt nanoparticle. Such shortened Pt-Pt bond can result in further downward shift in Pt d band and the weakening in binding energy, as shown previously. 26  In general, the size and shape effect at nanoscale does not reverse the trend in binding energy going from Pt to Pt-X@Pt, however it does increase the magnitude of the weakening effects. The binding of Pt to * OH and * * O 2 , may be over-weakened by the formation of small Pt-X@Pt (X = Co, Ni) nanoparticles (∼2 nm), which hinders the overall reaction as the cases of Pt-Cu-Pt(111) and Pt-Fe-Pt(111). Instead, increasing the particle size likely promotes the ORR activity of Pt. For nanoparticles, the solvation effects were not included due to the computational demand, which only mitigates the magnitude of weakening effect, rather than the overall trend.
In term of stability, similar promotion by the formation of Pt-X@Pt core-shell nanoparticles as that of Pt-X-Pt(111) slabs is observed. For the (111) facet, for instance, both Pt-Co@Pt ( U D = 0.45 V) and Pt-Ni@Pt ( U D = 0.42 V) can stabilize the Pt shell more significantly than the Pt nanoparticle, which results in an increase in U D .
Deposited oxides model.-When the alloyed X metal in Pt-X-Pt sandwich motif are active metals (X = Ti, Mn, Ce), it likely antisegregates to the surface and forms metal oxides due to the strong interaction with * O. 27,28 To describe this phenomenon, we built a model of Pt (111) slab deposited by a metal oxide trimer, Ti 3 O 7 H 7 /Pt(111), Mn 3 O 9 H 9 /Pt(111), Ce 3 O 7 H 7 /Pt(111), to describe the multifunctional sites available at the metal-oxides interface (Figures 1c-1e). In addition, the possible hydroxylation of oxide under the electrochemical conditions was also included. 61 Such inverse model of oxide/metal model successfully described the promoting effect of metal-oxide catalysts in water-gas shift, CO 2 hydrogenation and methanol oxidation reactions previously. [61][62][63][64][65] All three oxide clusters interact with Pt(111) via the direct interaction between Pt and the X ion in the cluster (Figures 1c-1e). However, the effect of deposition of oxide clusters on the structure of Pt(111) is different. For Ti 3 O 7 H 7 /Pt(111), the Pt-Pt bonds beneath the cluster are either compressed or stretched, ranging from 2.78Å to 2.88Å. As for the uncovered Pt surface, the Pt-Pt bonds are not affected by the cluster deposition. The distortion of Pt surface beneath Mn 3 O 9 H 9 is similar with that of Ti 3 O 7 H 7 /Pt(111), and the uncovered surface Pt atoms around the oxide cluster are slightly shifted by the interaction with Mn 3 O 9 H 9 , where the Pt-Pt bond length ranges from 2.75 Å to 2.84 Å. This indicates a stronger Mn 3 O 9 H 9 -Pt(111) interaction than Ti 3 O 7 H 7 -Pt(111). The Ce 3 O 7 H 7 has the strongest effect on the Pt atoms beneath the cluster, where the Pt-Pt bond length ranges from 2.73 Å to 2.90 Å. For the uncovered Pt atoms, the shift is even bigger with the Pt-Pt bond lengths ranging from 2.73Å to 3.06Å.
The structural difference for the surface Pt that is introduced by the oxide deposition varies the Pt-OH and Pt-O 2 bindings, which are important to evaluate the ORR activity. Different from the case of alloying, the oxides/Pt interface is able to strengthen the binding, especially for * OH, by forming hydrogen bonds with the hydroxylated metal oxides, which hinders the ORR on Pt. 12 The multiple sites at the metal-oxide interface has been found as the active sties to promote the activity and selectivity of the catalysts toward various heterogeneous reactions. [61][62][63][64][65] For the ORR, however, the interfacial sites bind the reaction intermediates too strongly to allow the ORR.
When * OH (Figure 5a) and * * O 2 (Figure 5b) were located slightly away from the cluster, the interactions were weakened as the case of Pt-X-Pt and Pt-X@Pt ( Figure 6). The bond weakening does not involve the direct interaction with the oxide, and the corresponding magnitude is associated with the structural changes of uncovered surface Pt. As demonstrated above, Ce 3 O 7 H 7 introduces the most significant changes among the three oxide/Pt(111) systems studied. It leads to the most decrease in * OH binding by 0.27 eV and * * O 2 binding by 0.13eV, which is much less than that of Pt-X-Pt. By comparison, the effect of Mn 3 O 9 H 9 and Ti 3 O 7 H 7 is smaller (≤0.05 eV for * OH and ≤ 0.03 eV for * * O 2 ). Overall, the bond-tuning via the Pt-X-Pt and Pt-X@Pt sandwich motifs seems more effective than that via the oxide/Pt motif. According to our calculations, the decoration of Ce oxide is likely more active than the oxides of Mo and Ti to promote the ORR activity of Pt, though none of them may be able to compete with Pt-Co-Pt and Pt-Ni-Pt.
We also evaluated the effect of oxide clusters on the stability of surface Pt atoms using U D . For the Pt atoms uncovered by the oxide, there is no change in U D compared to Pt(111). The difference is observed for the Pt atoms interacted directly with the oxide. Ti 3 O 7 H 7 /Pt(111) is the only system, being able to stabilize the underneath Pt more and enhance the stability of Pt on Pt(111). Compared to Pt(111), U D is raised by 0.40 V when the Pt atom is located under the center of Ti 3 O 7 H 7 , while it is only 0.01 V when the Pt atom is under the cluster edge. For Mn 3 O 9 H 9 /Pt(111) (≤ −0.07 V) and Ce 3 O 7 H 7 /Pt(111) (≤−0.34 V), the Pt dissolution potentials are lower than that on Pt(111). It means that the formation of Mn 3 O 9 H 9 or Ce 3 O 7 H 7 on Pt(111) does not stabilize the interacted Pt atoms on the surface, but resulting in the destabilization and lowered U D . Looking into the structures, we found that after removing a Pt atom, the clusters have strong interactions with the Pt atoms around the Pt vacancy, which leads to the reconstruction, and thus the gained stability of defected structures. This effect is especially significant on Ce 3 O 7 H 7 , which may be explained by the strong interaction between the Ce 3 O 7 H 7 cluster and Pt (111) surface. The strong oxide-metal interaction has been reported as the key to promote various heterogeneous catalytic processes, 62,63,66 while the present study show that it may facilitate the degradation of the Pt catalysts under the ORR condition.
In general, the decoration of metal oxides can affect both the ORR activity and stability of Pt(111). The multiple sites at the metaloxide interface bind the reaction intermediates too strongly to allow the ORR. The active sites are the surface Pt atoms which are close to the oxide, but not directly interacted. The strong oxide-Pt(111) interaction imposes the strain on these sites, enabling the variation in binding property. Among the three oxides studied, both Ti 3 O 7 H 7 and Mn 3 O 9 H 9 have a relatively mild effect on tuning the binding energy toward the ORR intermediates and therefore the ORR activity of Pt. By comparison, the promotion by Ce 3 O 7 H 7 is more significant. In term of stability, only Ti 3 O 7 H 9 helps to enhance the stability of Pt directly interacted, while Mn 3 O 9 H 9 and Ce 3 O 7 H 9 can facilitate the dissolution of interacted Pt due to the reconstruction induced by the strong oxide-Pt interaction.

Conclusions
The combined DFT and KMC study was performed, aiming to use cheap metals to improve the ORR activity and stability of Pt catalysts in alkaline solution. The study was based on our previous study 12 on deep mechanistic understanding of the ORR on Pt (111), indicating that the ORR activity of Pt can be promoted by weakening the * OH binding to facilitate the dissolution and strengthening the binding energy of * * O 2 to hinder the desorption from the surface.
The Pt-X-Pt (X = Co, Ni, Cu, Fe, Ti, Mn, Ce) sandwich motif was employed to tune the binding property and therefore the ORR activity of Pt catalysts. For X = Fe, Co, Ni, Cu, the structure of Pt-X-Pt alloy likely stays under reaction condition for lowering the surface energy. According to the estimated polarization curves, Pt-Ni-Pt and Pt-Co-Pt display the higher ORR activity than Pt by weakening the * OH binding and promoting the * OH removal. A decrease in activity is observed when going from Pt to Pt-Cu-Pt and Pt-Fe-Pt. This is due to the low stability of * * O 2 and the difficulty in reduction to * * OOH. For both Pt-Ni-Pt and Pt-Co-Pt catalysts the decrease in particle size (<4 nm), may also result in the over-weakened * * O 2 binding and the hindered ORR. With X = Ti, Mn and Ce, the oxide of X can be formed on the surface of Pt and form oxides due to the anti-segregation of X driven by the strong X-O interaction under the reaction condition. The decoration of Ce oxide helps in the removal of * OH and therefore promotion in the ORR activity of Pt, though the degree is not as significant as Pt-Co-Pt and Pt-Ni-Pt. By comparison the effect of Ti and Mn oxides is rather mild.
In term of stability, the Pt-X-Pt systems (X = Co, Ni, Cu, Fe) stabilize the surface Pt better than pure Pt under the ORR condition, where Pt-Fe-Pt show the most significant enhancement. The formation of oxide only affects the stability of directly interacted Pt atoms on the surface, which are enhanced for the presence of Ti oxide, and decreased by forming Mn oxide or Ce oxide Pt.