First-Principles Study of Mo Segregation in MoNi(111): Effects of Chemisorbed Atomic Oxygen

Segregation at metal alloy surfaces is an important issue because many electrochemical and catalytic properties are directly correlated to the surface composition. We have performed density functional theory calculations for Mo segregation in MoNi(111) in the presence of chemisorbed atomic oxygen. In particular, the coverage dependence and possible adsorption-induced segregation phenomena are addressed by investigating segregation energies of the Mo atom in MoNi(111). The theoretical calculated results show that the Mo atom prefers to be embedded in the bulk for the clean MoNi(111), while it segregates to the top-most layer when the oxygen coverage is thicker than 1/9 monolayer (ML). Furthermore, we analyze the densities of states for the clean and oxygen-chemisorbed MoNi(111), and see a strong covalent bonding between Mo d-band states and O p-states. The present study provides valuable insight for exploring practical applications of Ni-based alloys as hydrogen evolution electrodes.


Introduction
Adding a second metal into a pure metal catalyst can provide a great opportunity to tailor the properties of the catalyst [1][2][3][4]. Therefore, bimetallic systems have attracted considerable attention due to their importance in basic science and industry [5][6][7][8]. Bimetallic systems are much more complicated than pure metal catalysts since one element may segregate to the surface and lead to surface composition enrichment/depletion compared to the bulk. The adsorbate-induced surface segregation of metallic alloys under the reaction conditions and, thus, the changes in local atomic composition and surface structure have been predicted and demonstrated to occur for a number of bimetallic systems [9,10]. So for a given bimetallic configuration which exhibits a desired property, it is crucial to know whether the particular configuration is stable under the operating environment for a specific application.

Computational Methods
The first-principles calculations were performed within a density functional theory using the Vienna Ab-initio Simulation Package (VASP) [37][38][39]. The electron-ion interaction was described using the projector augmented wave method (PAW) [40,41] and the exchange correlation potential using the Perdew-Burke-Ernzerhof (PBE) functional method [42]. The energy cutoff for the plane wave basis set was 450 eV for all the calculated systems. Spin polarization was taken into account in the calculations and the Methfessel and Paxton [43] was employed to determine electron occupancies with a smearing parameter σ of 0.14 eV. The convergence criteria for the electronic self-consistent iteration and the ionic relaxation loop were set to 10´5 eV and 0.02 eV/A, respectively.
To simulate metallic surfaces, a slab supercell was employed. All the calculations presented in this work were based on slabs of 54 atoms, containing six atomic layers representing a 3ˆ3 supercell, separated by 15 A of vacuum space. MoNi(111) alloy systems corresponding to the substitution of one Ni atom by one Mo atom in the first, second, third or fourth nickel layer are presented in Figure 1, respectively. The Mo atom plays a role as a prober to determine whether it prefers to stay on the surface or in the bulk. As pointed out above, the drawback of this method is that it cannot predict the equilibrium phases and structures. However, investigation of the equilibrium phases and structures under certain given conditions is beyond the scope of the present study. Chemisorbed atomic oxygen was located on only one side of the slab. The influence of the resulting electric dipole on the computed energy values was estimated to be very small according to standard methods [44], and had thus been neglected. The atoms in the top four layers and the chemisorbed atomic oxygen were allowed to relax, while the atoms in the bottom two layers were fixed at the bulk geometry positions. Brillouin zone integrations were performed using Monkhorst-Pack grids [45] of 4ˆ4ˆ1 for slab calculations. The gas-phase oxygen molecule was simulated through a large supercell with dimensions of 12ˆ12ˆ12 A 3 .
In order to characterize the segregation behavior of the Mo atom in nickel, the segregation energy (E segr ) was defined as the energy difference between the states with the Mo atom located at the upper surface layer and in the bulk. According to this definition, the segregation energies were calculated according to the following equation: where E MoNipMo,x´layer) represents the total energy of the MoNi alloy system with the Mo atom located in the upper x nickel layers (x = 1, 2 or 3), and E MoNipMo,4th-layerq represents the total energy of the MoNi alloy system with the Mo atom located in the fourth nickel layer, which corresponds to the presence of the Mo atom in the "bulk" nickel matrix. For the oxygen adsorption cases, E MoNipMo,x-layerq and E MoNipMo,4th-layerq are the total energies of the slabs with O atoms adsorbed on the top layer and the Mo atom located at the corresponding atomic layer. According to our calculations, a chemisorbed atomic oxygen prefers to occupy a three-fold fcc hollow site on MoNi(111) and Ni(111). For higher oxygen coverage cases (such as 2/9, 3/9, and 4/9 ML oxygen coverages corresponding to the adsorption of two, three and four oxygen atoms, respectively), we considered all possible oxygen adsorption configurations and calculated the segregation behavior by using the most stable adsorption configurations. However, the mistake does not change any calculated results. For the preciseness of academic logic and the right of academic ethics, the authors must point out their mistake, and correct it. The correct Figure 1 is shown below. And the authors are responsible for these errors, they regret any inconvenience or misunderstanding caused by them.

Results and Discussion
We have done an exhaustive and extensive search on the lowest-energy O adsorption patterns (or arrangements) for each oxygen coverage with the Mo atom located at various atomic layers. The most stable O adsorption configurations for Mo located in the top-most atomic layer are shown in Figure 2. For the 1/9 ML oxygen coverage case, the O atom prefers to occupy a fcc site near the surface Mo atom. The two O atoms occupy a fcc site and an hcp site next to the Mo atom, respectively, for the oxygen coverage of 2/9 ML It is less stable by 0.16 eV for the two O atoms to occupy the two fcc sites. For the higher oxygen coverage cases (such as 3/9 ML and 4/9 ML) O atoms first occupy the fcc sites next to the surface Mo atom and then those fcc sites far away from the Mo atom. We find that the subsurface Mo atom prefers to stay away from the adsorbed O atoms; for instance, when the Mo atom is located at the second atomic layer, O atoms occupying the fcc sites far away from the top position of the Mo atom are energetically more favorable. When the Mo atom is located at the third or lower atomic layer, it imposes a negligible effect on the O adsorption, as we can see from Figure 3, that the segregation energies do not rely on the oxygen coverage.

Results and Discussion
We have done an exhaustive and extensive search on the lowest-energy O adsorption patterns (or arrangements) for each oxygen coverage with the Mo atom located at various atomic layers. The most stable O adsorption configurations for Mo located in the top-most atomic layer are shown in Figure 2. For the 1/9 ML oxygen coverage case, the O atom prefers to occupy a fcc site near the surface Mo atom. The two O atoms occupy a fcc site and an hcp site next to the Mo atom, respectively, for the oxygen coverage of 2/9 ML It is less stable by 0.16 eV for the two O atoms to occupy the two fcc sites. For the higher oxygen coverage cases (such as 3/9 ML and 4/9 ML) O atoms first occupy the fcc sites next to the surface Mo atom and then those fcc sites far away from the Mo atom. We find that the subsurface Mo atom prefers to stay away from the adsorbed O atoms; for instance, when the Mo atom is located at the second atomic layer, O atoms occupying the fcc sites far away from the top position of the Mo atom are energetically more favorable. When the Mo atom is located at the third or lower atomic layer, it imposes a negligible effect on the O adsorption, as we can see from Figure 3, that the segregation energies do not rely on the oxygen coverage.  The calculated segregation energies for clean and oxygen-adsorbed surfaces are given in Table 1 and also plotted in Figure 3. In absence of adsorbed gas, i.e., under vacuum conditions, the segregation energy for Mo located at the top-most layer is positive and has a value of 0.79 eV. This indicates that Mo does not segregate to the top-most surface layer. We can attribute this behavior to the smaller surface energy of nickel compared to that of molybdenum (experimental values are 2.45 eV/atom for Ni and 3.00 eV/atom for Mo [46]). Interestingly, the Mo atom prefers to occupy a site below the top-most surface layer. This oscillatory phenomenon is quite common in alloying systems [47][48][49][50][51]. That the Mo atoms located at the second layer are energetically favorable can be understood by two facts: (1) since it is not on the top-most layer it can avoid the higher surface energy of Mo; and (2) by locating at the second layer, the system can largely release the elastic energy due to the atomic size mismatch.  The calculated segregation energies for clean and oxygen-adsorbed surfaces are given in Table 1 and also plotted in Figure 3. In absence of adsorbed gas, i.e., under vacuum conditions, the segregation energy for Mo located at the top-most layer is positive and has a value of 0.79 eV. This indicates that Mo does not segregate to the top-most surface layer. We can attribute this behavior to the smaller surface energy of nickel compared to that of molybdenum (experimental values are 2.45 eV/atom for Ni and 3.00 eV/atom for Mo [46]). Interestingly, the Mo atom prefers to occupy a site below the top-most surface layer. This oscillatory phenomenon is quite common in alloying At the oxygen coverage of 1/9 ML, although the segregation energy drops to 0.23 eV from the value of 0.79 eV for the clean surface, the Mo atom still prefers to locate inside of the bulk. With increasing the oxygen coverage to 2/9 ML or thicker, the Mo atom becomes more stable in the top-most layer than in the lower layer and in the bulk. This result indicates that the presence of chemisorbed oxygen on the surface of a MoNi alloy electrode, such as during the decarbonation of the electrode, may cause the surface to be Mo-rich, i.e., having a concentration higher than the nominal concentration in the alloy. This explains our recent experimental observation that the electrode surface is Mo-rich [17]. We can understand the O adsorption-driven segregation of Mo as follows. There are two factors that favor Mo to segregate onto the top-most layer. The first one is that segregation of the Mo atom onto the top-most layer helps to release the elastic energy caused by the size mismatch. Another factor is that the Mo-O bond is stronger than the Ni-O bond (from the chemical rubber company (CRC) handbook [52], the bond strengths for diatomic Mo-O and Ni-O molecules are 145.1 kcal/mol and 93.6 kcal/mol at 298 K, respectively). There exists a factor that is against the segregation of Mo onto the first surface layer. This unfavorable factor is that the Mo surfaces have a larger surface energy than the corresponding Ni surfaces. The final segregation behavior is determined by the competition between the two favorable factors and the unfavorable factor. The d-band densities of states (DOS) for Mo and its nearest neighboring nickel atoms in the top-most surface layer with and without oxygen adsorption were calculated. Figure 4 shows the d-band DOS for both molybdenum and nickel atoms in the MoNi alloy compared to their corresponding DOS Materials 2016, 9, 5 6 of 10 in pure metals. The d-band DOS for Ni(111) and Mo(110) are presented because they are the most stable surfaces for the metals. Compared to the pure metal surface atom, the nickel d-band DOS center is slightly shifted up towards the Fermi level in the alloy case (Figure 4a), while the molybdenum d-band DOS center is shifted towards the lower energy region, away from the Fermi level (Figure 4b).
According to the d-band center model developed by Hammer, et al. [53], these d-band center shifts closely correlate to oxygen adsorption-induced Mo surface segregation as we have observed above. There exists an additional peak at the higher edge of the d-band DOS for the alloy case (Figure 4a). This peak indicates the formation of a covalent Ni-Mo bond [12]. From Figure 4b we can see that the Mo d-band not only shifts to the lower energy region, it also becomes narrower than in the pure metal, and the empty states are fewer. This means that the Mo atom gains some electrons from its neighboring nickel atoms, which is consistent with the d-band DOS of Ni shifting up and having more empty states in the alloy case (Figure 4a).  The d-band DOS for the Mo atom located at the top-most surface layer, the second layer and the third layer without oxygen adsorption are presented in Figure 5a. From Figure 5a, we see the shape and position of the Mo d-band DOS do not significantly change no matter which layer the Mo atom is located at. This indicates that the preference of Mo atoms to locate in the Ni bulk is mainly governed by surface energy considerations and atomic size effects (2.39 A for Mo and 2.22 A for Ni [54]) [12]. Figure 5b shows that the d-band DOS of Mo in MoNi(111) interacting with oxygen are drastically modified compared to those for the clean surface (i.e., without oxygen adsorption). The d-band DOS for the oxygen adsorption case are largely widened. The narrower DOS for the surface atoms are easy to understand, given they have a smaller coordination number relative to the bulk atoms. For the adsorption cases, it can be considered that as the adsorbate increases the coordination number for the surface atoms, it therefore broadens the DOS [55]. From Figure 5b, we can see that the broadening of the Mo d-band DOS causes some d-states to move to the lower energy region upon O adsorption. This effect lowers the system energy and leads to better stability of the system. Furthermore, we also see that the O-Mo anti-bonding states, located above the Fermi level, are largely unoccupied and that the O p-states and Mo d-states have a large overlap, thus allowing a strong covalent bonding and hybridization interaction between molybdenum and the oxygen atom. The d-band DOS for the Mo atom located at the top-most surface layer, the second layer and the third layer without oxygen adsorption are presented in Figure 5a. From Figure 5a, we see the shape and position of the Mo d-band DOS do not significantly change no matter which layer the Mo atom is located at. This indicates that the preference of Mo atoms to locate in the Ni bulk is mainly governed by surface energy considerations and atomic size effects (2.39 A for Mo and 2.22 A for Ni [54]) [12]. Figure 5b shows that the d-band DOS of Mo in MoNi(111) interacting with oxygen are drastically modified compared to those for the clean surface (i.e., without oxygen adsorption). The d-band DOS for the oxygen adsorption case are largely widened. The narrower DOS for the surface atoms are easy to understand, given they have a smaller coordination number relative to the bulk atoms. For the adsorption cases, it can be considered that as the adsorbate increases the coordination number for the surface atoms, it therefore broadens the DOS [55]. From Figure 5b, we can see that the broadening of the Mo d-band DOS causes some d-states to move to the lower energy region upon O adsorption. This effect lowers the system energy and leads to better stability of the system. Furthermore, we also see that the O-Mo anti-bonding states, located above the Fermi level, are largely unoccupied and that the O p-states and Mo d-states have a large overlap, thus allowing a strong covalent bonding and hybridization interaction between molybdenum and the oxygen atom. 6 number for the surface atoms, it therefore broadens the DOS [55]. From Figure 5b, we can see that the broadening of the Mo d-band DOS causes some d-states to move to the lower energy region upon O adsorption. This effect lowers the system energy and leads to better stability of the system. Furthermore, we also see that the O-Mo anti-bonding states, located above the Fermi level, are largely unoccupied and that the O p-states and Mo d-states have a large overlap, thus allowing a strong covalent bonding and hybridization interaction between molybdenum and the oxygen atom.

Conclusions
We have performed density functional theory calculations to investigate the effects of chemisorbed atomic oxygen on the segregation behavior of the Mo element in MoNi(111). In particular, the coverage dependence and possible adsorption-induced segregation phenomena are addressed by calculating the segregation energies of the Mo atom in the upper layers of MoNi(111). The theoretical calculated results show that the Mo atom prefers to be embedded in the bulk in the clean MoNi(111), while it segregates to the top-most layer when the oxygen coverage is thicker than 1/9 ML For the clean MoNi(111) we see that the d-band center of Ni atoms surrounding the Mo atom shifts up to the Fermi energy and the Mo d-band becomes narrower with its center shifted down away from the Fermi energy. The shape and position of the Mo d-band DOS do not significantly change, no matter which layer the Mo atom is located at. This indicates that the preference of Mo atoms to be embedded in the Ni bulk is mainly governed by surface energy considerations and atomic size effects. The Mo d-band DOS for the oxygen adsorption cases are largely widened and lead to a strong covalent bonding between the molybdenum atom and the oxygen atom. The present study provides valuable insight for exploring practical applications of Ni-based alloys as hydrogen evolution electrodes.