Design of Pd-Based Bimetallic Catalysts for ORR : A DFT Calculation Study

Developing Pd-lean catalysts for oxygen reduction reaction (ORR) is the key for large-scale application of proton exchange membrane fuel cells (PEMFCs). In the present paper, we have proposed a multiple-descriptor strategy for designing efficient and durable ORR Pd-based alloy catalysts. We demonstrated that an ideal Pd-based bimetallic alloy catalyst for ORR should possess simultaneously negative alloy formation energy, negative surface segregation energy of Pd, and a lower oxygen binding ability than pure Pt. By performing detailed DFT calculations on the thermodynamics, surface chemistry and electronic properties of Pd-M alloys, Pd-V, Pd-Fe, Pd-Zn, Pd-Nb, and Pd-Ta, are identified theoretically to have stable Pd segregated surface and improved ORR activity. Factors affecting these properties are analyzed. The alloy formation energy of Pd with transition metals M can be mainly determined by their electron interaction.This may be the origin of the negative alloy formation energy for Pd-M alloys.The surface segregation energy of Pd is primarily determined by the surface energy and the atomic radius of M. The metals M which have smaller atomic radius and higher surface energy would tend to favor the surface segregation of Pd in corresponding Pd-M alloys.


Introduction
Fuel cells have received much attention in recent years as the global effort to reduce our reliance on fossil fuels has increased.In particular, low-temperature proton exchange membrane fuel cells (PEMFCs) possess great potential as substitutes for conventional combustion engines in future mobile applications, which can efficiently convert chemical energy into electrical energy.However, the development of PEMFCs is severely hampered by the fact that the oxygen reduction reaction (ORR) is slow, even when using Pt as a catalyst.The sluggish kinetics of the ORR requires a substantial amount of this precious metal within their cathodes in real electrochemical systems, which has limited the scaling-up of corresponding renewable energy technologies.Therefore, the PEMFCs still leave room for reducing cost and improving efficiency that could be achieved by finding more active and stable electrocatalytic cathode catalysts for ORR.Because Pt is expensive and scarce, alloys have attracted interest in the request to develop advanced electrocatalysts in recent years [1][2][3][4].Several Pt alloys, such as Pt-Fe, Pt-Co, Pt-Ni, and Pt-Cu [5][6][7][8][9][10][11][12][13][14], are considerably more active than Pt and have been studied extensively toward advanced ORR electrocatalysts.For example, Stamenkovic et al. found that on Pt 3 Ni the ORR is 90 times faster than on pure Pt [8].Alloying Pt with nonnoble transition metals, Fe, Co, and Cu, can increase the catalytic activity and the stability of these catalysts, where their Pt-based alloy catalysts have been reported to be 2-10 times more active than polycrystalline Pt for the ORR [6,9,13,[15][16][17][18][19][20][21][22].However, one of the major drawbacks of PEMFCs commercialization is the high cost of Pt-based catalysts.Thus, alternative cost-effective catalysts must be developed to eliminate the Pt-based ones.Recent studies on ORR electrocatalysis have attempted to replace Pt with lessexpensive materials.
In recent years, some Pd-M (M is transition metals) alloys have been used as ORR electrocatalysts [23][24][25][26][27][28][29][30], in which alloying with Fe and Co was found to increase strongly the ORR electrocatalytic activity of Pd.This represents significant progress in the research on non-Pt ORR catalysts for PEMFCs.For example, Adzic [28,29] studied Pd-M alloy electrocatalysts and found that Pd-Fe and Pd-Co alloys become very active for ORR, and Pd atoms segregate to the surface to form a pure Pd skin on bulk alloys.The results showed that the active ORR electrocatalysts can be devised without Pt, and their activity can surpass those of pure Pt electrocatalysts [29].Further enhancement of the ORR catalytic activity and stability of Pd is of considerable interest, and replacing Pt with the less-expensive metal Pd alloy can considerably decrease the operational costs, thereby facilitating the faster and broader application of PEMFCs.In the search for improved alloy catalyst materials, different strategies and schemes have been used, from experimental high-throughput catalyst screen [31] to the study of idealized model systems [8].But to date only limited efforts have been made to use computational techniques based on density functional theory (DFT) to screen for improved ORR catalysts [32].The pace of materials discovery for alloy electrocatalysts could be accelerated in principle by the development of efficient computational screening methods.Some studies have proposed thermodynamic guidelines for design of Pdbased alloy electrocatalysts for ORR.For example, Bard et al. [30,33,34] suggested that, for Pd-M alloys, the site for O-O bond breaking is constituted by the transition metal M and then formed O ads migrates to hollow sites formed by Pd atoms, where it is easily reduced to water.Based on this mechanism, the alloy surface should compose of a relatively reactive metal such as Co, and the atomic ratio of this transition metal should be 10% to 20% so that sufficient sites exist for reactions of O-O bond breaking on M and for O ads reduction at hollow sites dominated by Pd atoms.Wang and Balbuena [35,36] proposed a similar thermodynamic guideline for design of Pd-based bimetallic alloy catalysts.For Pd with fully occupied valence d-orbitals, alloying with transition metals such as Co with unoccupied valence d-orbitals significantly reduces the Gibbs free energy both for the first charge-transfer step and for the steps involving the reduction of intermediates.Although these studies have convincingly showed that first-principles catalyst design is now a reality for the ORR and corresponding progress has been made in this respect, these arguments for ORR electrocatalysis on Pdbased bimetallic surfaces cannot explain the relatively good activity and stability of Pd-M alloys in acidic medium [37].On the other hand, the Pd-enriched skin can account for both the good activity and stability of these alloys [38][39][40].Hence, a fully DFT-based, high-throughput, computational design and screening of Pd-based alloy catalysts has yet to be realized, and further development of screening schemes is still necessary for design of Pd-based bimetallic catalysts.
As we all know, the ORR occurs in electrochemical environments under conditions of high electrode potential and is, thus, expected to be susceptible to the stability issues mentioned above.Indeed, recent reports have highlighted the importance of stability considerations in the identification of improved ORR catalysts [8,41].Pd materials are known to be kinetically stable in electrochemical systems over relatively long periods of operation [42,43].It is also one of the metallic elements that are stable under the high potentials and acidic conditions of a PEMFC [44].Additionally, core-shell particles that possess an inexpensive core surrounded by a Pd shell are extensively discussed as promising PEMFC catalyst materials and can account for both the good activity and the stability of Pd-based alloys at present.However, the stability issues have almost been not considered in previous computational design and screen of alloy catalysts based on DFT.Therefore, in the present paper we will focus our attention on Pd-based alloy catalysts that could form a protective Pd shell on the topmost surface layer and that possess a high tendency to segregate, a property closely dependent on the surface segregation energy.
As electrocatalytic reactions on catalysts are far too complex for a complete theoretical description, the complexity is first reduced.Consequently, in the present study, we restrict our calculations to close packed surfaces and, in this contribution, we search for a DFT-based, computational catalyst screening procedure including the adsorption energy of O atom, the alloy formation energy, and the surface segregation energy of Pd that are considered sufficiently able to predict the catalytic activity and stability of alloy catalysts.The overall goal is to find alloys of the composition Pd 3 M (M is transition metals) with improved ORR activity and stability.We illustrate the use of this scheme on the ORR and successfully use these strategies to identify several promising new catalysts for this reaction, and some alloy catalysts have been synthesized and tested experimentally and show improved ORR performance compared with pure Pt.Although our results were not always quantitatively accurate, the conclusions on trends were reasonably accurate qualitatively because the present experimental phenomenon of the ORR confirmed the computational screening results.

Models and Methods
Calculations were carried out based on the periodic DFT slab approach by using the generalized gradient approximation with the Perdew-Burke-Ernzerhof exchange correlation functional [45].The nuclei and core electrons were described by using PAW potentials [46].The Kohn-Sham equations were solved self-consistently by using a plane-wave basis set.A kinetic energy cutoff of 26 Ry and a charge-density cutoff of 260 Ry were used to make the plane-wave basis set finite.The Fermi surface has been treated by the smearing technique of Methfessel-Paxton with a parameter of 0.02 Ry [47].All calculations in this study have been implemented by using the PWSCF codes included in the Quantum ESPRESSO distribution [48], while figures of the chemical structures were produced with the XCRYSDEN graphical package [49][50][51].
Crystal structure of Pd-based alloys is calculated with L1 2 cubic system structure in which the stoichiometric ratio is 3 : 1 [52], in which the Pd atoms occupy the face-centered positions and the M (M is 3d, 4d, and 5d transition metals) atoms are located at the corners in the unit cell.For each Pd 3 M structure, the lattice constant was obtained after a bulk optimization.In all calculations, the (2 × 2) five-layer fcc(111) slabs with theoretical equilibrium lattice constant to model the Pd 3 M(111) surface were used.In the super cell, each slab layer contains 3 Pd atoms and 1 M atom.
The segregation structures were reached by exchanging M atoms at the first layer and Pd atoms at the second layer.Therefore, the topmost layer is fully covered by Pd atoms and the second layer contains 50% Pd atoms for each Pd segregation structure.A (4 × 4) uniformly shifted k-mesh was used to describe the first Brillouin-zone for (2 × 2) slab.Vacuum layers 16 > in thickness were added above the top layer of slabs in all cases, which is sufficiently large to ensure that the interactions between repeated slabs are negligible.In the slab calculations, the structure of the bottom two layers of the slab was fixed at the theoretical bulk positions, whereas the geometry of the top three layers and all other structural parameters on a (2 × 2) five-layer slab was allowed to fully relax to minimize the total energy of the system.The criteria for the total energy and the Cartesian force components acting on each atom convergence used were within 10 −5 Ry and below 10 −3 Ry/Bohr with regard to structural optimization, respectively.
The -band center of surface atoms,   , is a key parameter that influence surface adsorption characteristics [53,54].Therefore,   was calculated in various catalytic materials, and the relationship between the values of   and the adsorption energy of O atom was examined.The -band center was calculated as the first moment of the projected -band density of states on the surface atoms referenced to the Fermi level.

Stability Descriptor of Pd-Based Bimetallic Catalysts.
It could be imagined that formation of Pd-based bimetallic catalysts with Pd surface segregated structure can be divided into two stages.Primarily, bulk alloys must be formed by the liquid phase or solid phase reactions between Pd and transition metals M. Thus, the fact that formation energies, Δ alloy , of the alloys of reaction nPd + mM → Pd n M m are negative is required.Otherwise, separate phase is formed between Pd and transition metals M. Δ alloy is calculated by the formula Δ alloy = (   1− ; ) −   ( 0  ) − (1 − )  ( 0  ), where (   1− ; ) is the total energy of the alloy,   ( 0  ) and   ( 0  ) are the total energy of constituents  and  under equilibrium lattice constant, respectively.Then, there is surface segregation of interior Pd atoms to form Pd segregated surface via some special processes such as high temperature annealing or surface inducing adsorption.In general, the surface segregation energy, Δ segr , is the energy of moving an atom from the interior to the surface of a crystal.In this work, it is defined as the difference of total free energies between the slabs with the Pd surface segregated structure and that with the bulk structure.Negative surface segregation energy indicates that the interior Pd atoms are able to segregate to the surface, while positive segregation energy suggests that no Pd segregation occurs on the surface.Although the structure of Pd segregated surface can be achieved by depositing Pd on transition metals or dealloying of Pd-based alloy catalysts, this do not influence the fact that surface segregation energy Δ segr is a negative act as stability descriptor of alloy catalysts with Pd segregated surface.

Activity Descriptor of Pd-Based Bimetallic Catalysts.
In our previous study on ORR mechanisms [55], the results indicated that the protonation of adsorbed O atom to form OH is the rate-determining step (rds) on the Pd(111) surface; such finding about the rds of ORR indicates that the ORR activity of the Pd segregated surface can be assessed by their binding ability to oxygenated species and Pd-based catalysts that more weakly bind atomic oxygen than pure Pt may have better ORR activity.It has been well-established in the literature that Pd and Pt are somewhat too oxophilic so that the surface oxygenated species formed from dissociation of O 2 tend to block the surface sites for ORR [56].Thus, oxygen adsorption energy can be a good descriptor for the ORR catalytic activity of Pd-based bimetallic alloys.If the adsorption energy of O atom on stable alloy catalysts with Pd segregated surface is close to or slightly weaker than that of pure Pt, the alloy catalysts should have enhanced ORR performance compared with pure Pt.The adsorption energies are calculated with the energies of the bare Pd 3 M(111) slab ( slab ) with Pd segregated surface and the isolated oxygen atom  O as the references according to where  total refers to the total energies of the optimized slab/O system.
To sum up, three criteria should be met simultaneously for ideal Pd-based ORR alloy catalysts: (1) Δ alloy < 0; (2) Therefore, we applied DFT method to calculate above mentioned three criteria for alloy catalysts based on Pd 3 M model and then expect that possible ORR catalysts with high activity and stability could be revealed theoretically.

Results and Discussion
According to the sequence of periodic table of elements, the DFT-calculated lattice constants () and the values for three criteria of various Pd 3 M alloys are shown in Tables 1, 2, and 3 when M is 3d, 4d, and 5d transition metals, respectively.Valence electron configuration (VEC) of transition metals M is also given.

Stability Analysis of Pd-Based Bimetallic Catalysts.
According to the formation energies of Pd 3 M alloys in Tables 1, 2, and 3, it can be predicted that 3d transition metals Sc, Ti, V, Cr, Mn, Fe, Cu, and Zn, 4d transition metals Nb, Mo, Ru, and Rh, and 5d transition metals Ta, W, Re, Ir, and Pt can form stable alloys with Pd since the formation energy Δ alloy of these Pd-based bimetallic alloys is negative.
To ascertain the effect of electronic structure on stability of Pd-based alloys, the variation trend of the DFT-calculated Δ alloy of Pd 3 M (M is 3d, 4d, and 5d transition metals) alloys with the valence electrons number (sums of outmost s and d electrons) of M is shown in Figure 1.The results show that the values of Δ alloy become more and more positive firstly and then become more and more negative as valence electrons number increases.There are the most positive Δ alloy in Pd 3 Co (valence electrons number of Co is 9), Pd 3 Ru (valence electrons number of Ru is 8), and Pd 3 Re site (valence electrons number of Re is 7) for the alloys that formed by Pd and 3d, 4d, and 5d transition metals, respectively, which are approximately in the center of periodic table of elements.Therefore, the fact that whether they can form stable alloys between Pd and transition metals is related to valence electrons number of transition metals.This should be due to Δ alloy values that can be affected by the electronic interaction between Pd and transition metals M. For metals which have very different valence electrons numbers from that of Pd, the electronic interaction should be strong due to the large difference in electronegativity.This may be the origin of the largely negative Δ alloy for alloys of Pd with transition metals M. Relation between the DFT-calculated segregation energy, Δ segr , of Pd in Pd 3 M alloys and the valence electrons number of M is shown in Figure 2.For 3d transition metals, the Δ segr in Pd 3 M alloys exhibit an oscillating variation with increasing the valence electrons number as moving from the left to right in the periodical table.The Δ segr for 4d and 5d metals show first decrease then increase with increasing valence electrons number.There are the most negative Δ segr in Pd 3 V and Pd 3 Co (valence electrons number of V and Co is 5 and 9, resp.),Pd 3 Mo (valence electrons number of Mo is 6), and Pd 3 Re site (valence electrons number of Re is 7) when alloying Pd with 3d, 4d, and 5d transition metals, respectively,  which are also approximately in the center of periodic table of elements.This should be also due to Δ segr values that can be affected by the electronic interaction between Pd and M. For metals which have very different valence electrons numbers from that of Pd, the electronic interaction should be strong due to the large difference in electronegativity [57].This may be also the origin of the largely negative Δ segr for alloys of Pd with transition metals.Thus, we may conclude that electronic interactions play the important role in determining the alloy formation energy and segregation energy of Pd of Pd-based bimetallic alloys for ORR.According to the criteria of Δ segr < 0, Pd-V, Pd-Fe, Pd-Co, Pd-Ni, Pd-Zn, Pd-Nb, Pd-Mo, Pd-Ru, Pd-Rh, Pd-Ta, Pd-W, Pd-Re, Pd-Ir, and Pd-Pt alloys are possible to form Pd segregated surface by surface segregation of Pd.It has been widely reported experimentally and theoretically that surface segregation is primarily determined by the surface energy and the atomic size [58,59].Metal atoms with larger radii and lower surface energies usually tend to segregate to the surface.The correlation between Δ segr and both the M metallic radius  [60] and the experimental surface energy  surf [61] can be found in Figures 3 and 4, respectively.One can find that no matter when 3d, 4d, or 5d metals are the alloying elements, Pd is predicted to segregate to the surface for those metals with higher  surf than Pd except for the Ti and Cr.This demonstrates that Pd surface segregation can be mainly determined by the surface energy.M with more positive  surf than Pd does not always correspond to a negative Δ segr in the corresponding Pd 3 M alloy since values of Δ segr are also affected by atomic radius, such as Ti and Cr, which have atomic radii close to and lower than that of Pd, respectively.As far as we know, segregating Pd atom to the surface alleviates the large strain in the alloy and thus decreases the total energy.This leads to the conclusion that the atomic size may be also an important factor for surface segregation.It can be found that Pd tends to segregate to the surface for those metals with atomic radii smaller than Pd except for Mn and Cr.In some cases, it can even determine the direction of surface segregation.For example, Pd-Ti, Pd-Cu, and Pd-Zn alloys, in which Ti possesses a much slight larger atomic radius than Pd, are still predicted to have positive segregation energy of Pd although its slightly higher surface energies than Pd, Cu, and Zn possess much smaller atomic radii than Pd and are still predicted to have a slight trend of Pd surface segregation, in spite of their very low   surf relative to that of Pd.As mentioned above, metal Cr alloy has higher surface energy and lower atomic radius than Pd.However, Pd 3 Cr alloy has positive segregation energy of Pd, indicating that electronic interaction may play the major role in determining the segregation energy of Pd in Pd 3 Cr alloy.For other metals M, such as 3d transition metal, Sc, 4d transition metals, Y, Zr, Ag, and Cd, 5d transition metal, Au, which have larger atomic radii and lower surface energy than Pd, alloys of Pd 3 Sc, Pd 3 Y, Pd 3 Zr, Pd 3 Ag, Pd 3 Cd, and Pd 3 Au exhibit considerably positive segregation energy Δ segr .Thus, it will be difficult for Pd 3 Sc, Pd 3 Y, Pd 3 Zr, Pd 3 Ag, Pd 3 Cd, and Pd 3 Au alloys to form Pd surface segregated structure.The segregation energy Δ segr of Pd in Pd 3 Mn alloy is positive and mainly determined by its surface energy because 3d transition metal Mn has smaller atomic radius and lower surface energy than Pd.Since 4d transition metal, Mo, and 5d transition metals W, Re, Ir, and Pt mostly have similar atomic radii to Pd, values of segregation energy Δ segr for the Pd 3 Mo, Pd 3 W, Pd 3 Re, Pd 3 Ir, and Pd 3 Pt alloys are mainly determined by their surface energy.W and Re especially have much more positive surface energy than Pd.Therefore, both of those alloys exhibit considerably negative segregation energy Δ segr .
To form a stable Pd surface segregated structure in Pd-M bimetallic alloys, it is necessary that both corresponding Δ alloy and Δ segr are negative.Otherwise, separate phase may tend to be formed between Pd and transition metals M. As shown in Figure 5, alloys of Pd with 3d transition metals V, Fe, and Zn simultaneously have negative Δ alloy and Δ segr .Alloys of Pd with early 4d transition metals have negative Δ alloy but positive Δ segr , while those with late 4d transition metals have negative Δ segr but positive Δ alloy .Only middle 4d transition metals such as Nb and Mo can form stable alloys with Pd, which simultaneously have negative Δ alloy and Δ segr .Thus, they should be able to form stable Pd surface segregated.Among Pd-5d alloys, Pd-Ta and Pd-W alloys simultaneously have negative Δ alloy and Δ segr .Thus, it is expected that Pd, Ta, and W could form stable alloys with Pd surface segregated structure.
The segregation process of Pd was also simulated in systems containing atomic oxygen adsorbed in the fcc (without transition metal atoms) and hcp (with a transition metal atom on the no Pd-segregated surface) positions, respectively.The most stable adsorption site for atomic oxygen is fcc on the Pdsegregated surface, and the differences of adsorption energy between fcc and hcp sites are relatively small.However, on no Pd-segregated surfaces, the differences of adsorption energy are large between both sites.Therefore, it is important to include the hcp case in the evaluation of surface segregation of Pd.The DFT-calculated surface segregation energies of Pd, Δ segr , for various Pd 3 M (M is 3d, 4d, and 5d transition metals) alloys in systems containing atomic oxygen adsorbed in the fcc and hcp positions are given in Table 4.The results showed that the segregation energies of Pd change when atomic oxygen adsorbs in fcc site, but the segregation trend of Pd is almost not changed compared to the clean surfaces.It is noticeable that in the hcp site the segregation behavior of Pd in most Pd 3 M alloys turns into nonsegregation (large position value of the segregation energy).It is observed that the presence of oxygen in hcp position causes the largest change in segregation energy of Pd, even leading to the reversion of segregation trends of Pd.It may be due to the fact that the high oxygen philicity of some M transition metal elements prohibit the surface segregation of Pd.Although the presence of oxygen in hcp position can revert the segregation behavior of Pd, strong adsorbed oxygenated species in hcp site may cause surface poisoning, which is unfavorable for ORR.In recent study conducted by Ramírez-Caballero and coworkers [62], DFT was used to evaluate the surface segregation of Pt, and two groups of cores are investigated, namely, a monolayer Pt over monometallic or bimetallic core.It was found that some 4d and 5d pure cores may serve as stable cores, and bimetallic Pd 3 X cores could enhance the stability of the surface Pt atoms both in vacuum and under adsorbed oxygen.Our present studies also give similar results for the surface segregation of Pd.It can be observed that Pd still can segregate to surface when Pd alloys with 4d transition metals Mo, Ru, and Rh and 5d transition metals Re, Ir, Pt, and Au, even when oxygen adsorbs in hcp site, as shown in Table 4. Simultaneously, both of our present and previous studies from Ramírez-Caballero and coworkers all showed that surface segregation can be primarily determined by the surface energy and the atomic size, and metal atoms with larger radii and lower surface energies usually tend to segregate to the surface regardless of being the segregation process of Pd or Pt.

ORR Catalytic Activity Analysis of Pd-Based Bimetallic
Catalysts.In this section, the adsorption energies of O atom are calculated on the all Pd-based bimetallic catalysts for comparison although some alloys cannot form stable Pd surface segregated structures.As shown in Tables 1, 2, and 3, the DFT-calculated O adsorption energies on Pd-segregated Pd-Sc, Pd-Ti, Pd-V, Pd-Cr, Pd-Mn, Pd-Fe, Pd-Co, Pd-Ni, Pd-Cu, Pd-Zn, Pd-Y, Pd-Zr, Pd-Nb, Pd-Mo, Pd-Ru, Pd-Rh, Pd-Ag, Pd-Cd, Pd-Ta, Pd-W, and Pt-Re alloys are more positive than that on pure Pt, indicating that these alloy catalysts may have enhanced ORR performance compared with pure Pt.
In order to verify -band center model that Hammer and Nørskov proposed, that is, the interaction between surface atoms of metal and adsorbates weakens as downshift of   , there is an approximate linear relationship between adsorption energy and   [53,54].The relationship between O adsorption energies and   is given in Figure 6.For all alloys considered in the present study, they mostly have  O alloy values which are close to or more positive than that on the pure Pt surface except for Pd 3 Ag, Pd 3 Ir, Pd 3 Pt, and Pd 3 Au alloys.Based on the volcano model of the ORR activity versus  O alloy values [56],  O alloy values of alloy surfaces with segregated structure are more positive than that for the pure Pt surface within 0.50 eV; the ORR activity would be close or superior to that of pure Pt.Thus, the Pd-based bimetallic alloys falling between two dash lines in Figure 6 will be potential ORR electrocatalysts.Meanwhile, it can be observed from Figure 6 that the calculated Δ O values between Pd 3 M and Pt roughly linearly correlated with the calculated   differences between Pd-segregated Pd 3 M and Pd.In general, more negative Δ  corresponds to more positive Δ O , indicating that adsorption characteristics of O on the Pd 3 M alloy surfaces meet approximately the abovementioned -band center model.Therefore,   also could be a significant parameter for design of Pd-based alloy catalysts.
To understand the effect of the alloying elements on the electronic structure of the Pd segregated surface, the relationship between the calculated -band center,   , and number of valence electrons in M is shown in Figure 7.The results show that Δ  values first decrease and then increase as the number of valence electrons of alloy elements (M) increase from the left to right in the periodical table when M is 3d or 4d transition metals and there is V-shaped change of Δ  values, whereas, for 5d transition metals, the calculated Δ  values linearly correlated with the number of valence electrons of alloy elements.
Alloying elements M may be able to affect the electronic structures of the Pd segregated surface by lattice strain (strain effect) or electronic interaction (ligand effect).In general, compressive strain results in downshift of -band center,   ; namely, the value of   becomes more negative than that of the pure Pd.As shown in Figure 8, the values of -band center,   , as a function of the optimized lattice constant () values are given.One can see from Figure 8 that compressive strain leads to downshift of -band center,   , as compared with that of pure Pd among all Pd 3 M alloys that we have considered, regardless of M being 3d, 4d, or 5d transition metals; tensile strain results in upshift of -band  center,   .However, some alloys such as Pd 3 Sc, Pd 3 Y, Pd 3 Zr, Pd 3 Nb, and Pd 3 Ta even cause downshift of -band center,   , although they have larger lattice constants than pure Pd (tensile strain).Therefore, it can be speculated that ligand effect plays important role in determining the downshift or upshift of -band center,   , of various Pd-M alloys, and, in some cases, it even can dominate the variation of -band center.For example, in Pd 3 Sc, Pd 3 Y, Pd 3 Zr, Pd 3 Nb, and Pd 3 Ta alloys, the downshift of -band center,   , due to the ligand effect is much beyond the upshift caused by lattice expansion.

Screening of Pd-Based Bimetallic Alloy Catalysts.
Based on the DFT-calculated results (As shown in Figure 9), Pd-V, Pd-Fe, Pd-Zn, Pd-Nb, Pd-Mo, Pt-Ta, and Pt-W alloys can meet simultaneously three criteria of design of ORR catalysts; that is, Δ alloy < 0, Δ segr < 0, and Δ O > 0, and the Δ O values are within 0.50 eV.However, the formation energy of Pd-Mo and Pd-W alloys is nearly 0; thus, the stability of these two alloys may be relatively poor.The fact that Pd 3 Fe have excellent ORR catalytic activity has been reported experimentally [29], which indicated that these three criteria based on DFT calculations are feasible in design of ORR catalysts of Pd-based alloys.The ORR catalytic property of Pd-V, Pd-Zn, Pd-Nb, and Pd-Ta alloys is rarely reported experimentally, and we think that Pd-V, Pd-Zn, Pd-Nb, and Pd-Ta alloys as ORR electrocatalysts in PEMFCs deserve further careful study in the future.In the recent experimental studies by Adzic et al. [27,39,40], Pd-Co bimetallic catalyst was found to have enhanced ORR electrocatalytic activity than pure Pt.Based on the Δ segr value obtained in the present study, Pd-Co alloy can form a stable Pd surface segregated structure because Co has smaller atomic radius and larger surface energy than Pd.However, according to the Δ alloy value obtained here, the Pd-Co alloy may have problems in forming stable Pd-Co alloy due to slightly positive Δ alloy value.

Conclusions
Developing catalysts for ORR with low noble metal contents, high catalytic activity, and high stability is the key for largescale application of PEMFCs.We have proposed a multipledescriptor strategy for rational design of efficient and durable ORR alloy catalysts with low precious metal content based on DFT calculations.We argued that good Pd-based bimetallic catalysts for ORR should simultaneously have negative alloy formation energy, negative surface segregation energy of Pd, and lower oxygen binding strength than pure Pt.By performing detailed DFT calculations on the thermodynamics, surface chemistry, and electronic properties of Pd-M alloys (M is 3d, 4d, and 5d transition metals), Pd-V, Pd-Fe, Pd-Zn, Pd-Nb, and Pd-Ta alloys were identified theoretically to have stable Pd segregated surface and improved ORR activity, among which Pd-Fe alloy has indeed been reported to have excellent ORR catalytic activity.Pd-V, Pd-Zn, Pd-Nb, and Pd-Ta alloys as ORR electrocatalysts in PEMFCs deserve further experimental study in future.
Factors affecting these properties are analyzed carefully.The alloy formation energy Δ alloy of Pd with transition metals M can be mainly determined by their electron interaction, which makes Δ alloy values of the most of Pd-M alloys negative due to the large difference in electronegativity except for those in which the alloying transition metals have valence electrons numbers close to Pd.This may be the origin of the negative Δ alloy for Pd-M alloys.The surface segregation energy of Pd is primarily determined by the surface energy and the atomic radius of M. The metals M which have smaller atomic radius and higher surface energy would tend to favor the surface segregation of Pd in corresponding Pd-M alloys.In some cases, atomic radius can determine the direction of surface segregation of Pd.For instance, Pd-Ti, Pd-Cu, and Pd-Zn alloys, in which Ti possesses a much slight larger atomic radius than Pd, are still predicted to have positive segregation energy of Pd although its slightly higher surface energies than Pd, Cu, and Zn possess much smaller atomic radii than Pd and are still predicted to have a slight trend of Pd surface segregation, in spite of their very low  surf relative to that of Pd.Noticeably, although metal Cr has higher surface energy and lower atomic radius than Pd, Pd 3 Cr alloy has positive segregation energy of Pd, indicating that electronic interaction may play the major role in determining the segregation energy of Pd in Pd 3 Cr alloy.
The calculated oxygen adsorption energy on the Pd segregated surface roughly linearly correlated with the calculated -band center of surface Pd atoms.Alloying elements M affect the -band center by strain effect and ligand effect.Compressive strain generally causes downshift of -band center, and tensile strain leads to upshift of -band center.However, Pd 3 Sc, Pd 3 Y, Pd 3 Zr, Pd 3 Nb, and Pd 3 Ta alloys exhibit downshift of -band center although they have larger lattice constants than pure Pd, indicating that ligand effect can dominate the variation of -band center of surface Pd atoms in some cases.In these five alloys, the downshift of -band center due to ligand effect is much beyond the upshift caused by lattice expansion.

Figure 1 :
Figure 1: The relationship between the calculated formation energies of Pd 3 M alloys, Δ alloy , and number of valence electrons.

Figure 2 :
Figure 2: The relationship between the calculated segregation energy (Δ segr ) of Pd 3 M alloys and number of valence electrons.

Figure 3 :Figure 4 :
Figure 3: Plots of DFT-calculated segregation energy (Δ segr ) of Pd in Pd 3 M alloys as functions of atomic radii of alloying metals (M).Atomic radii are given with respect to that of Pd; that is, Δ = (M)− (Pd).

28 −Figure 5 :
Figure 5: Relation between the DFT-calculated segregation energy (Δ segr ) of Pd and formation energies in Pd 3 M alloys: stability analysis.