Energetics of Nanoparticle Exsolution from Perovskite Oxides - The

Jun 18, 2018 - The presence of active metal nanoparticles on the surface significantly increases the electrochemical performance of ABO3 perovskite ox...
0 downloads 0 Views 5MB Size
Download PDF
Letter Cite This: J. Phys. Chem. Lett. 2018, 9, 3772−3778

pubs.acs.org/JPCL

Energetics of Nanoparticle Exsolution from Perovskite Oxides Yang Gao,†,‡ Ziheng Lu,† Tsam Lung You,† Jian Wang,† Lin Xie,§ Jiaqing He,§ and Francesco Ciucci†,∥,* †

J. Phys. Chem. Lett. 2018.9:3772-3778. Downloaded from pubs.acs.org by IOWA STATE UNIV on 01/28/19. For personal use only.

Department of Mechanical and Aerospace Engineering, The Hong Kong University of Science and Technology, Hong Kong, SAR, China ‡ College of Materials Science and Engineering, Hunan University, Changsha 410082, China § Department of Physics, Southern University of Science and Technology, Shenzhen 518055, China ∥ Department of Chemical and Biological Engineering, The Hong Kong University of Science and Technology, Hong Kong, SAR, China S Supporting Information *

ABSTRACT: The presence of active metal nanoparticles on the surface significantly increases the electrochemical performance of ABO3 perovskite oxide materials. While conventional deposition methods can improve the activity, in situ exsolution produces nanoparticles with far greater stability. The migration of transition metal atoms toward the surface is expected to affect the exsolution process. To study the energetics, we use ab initio computations combined with experiments in a SrTiO3-based model system. Our calculations show that Ni preferentially segregates toward the (100)-oriented and SrTiOterminated surfaces, note that this orientation is identical to one reported by the Irvine and Gorte groups. Vacancies in the Sr-site and O-site promote the segregation of Ni, while placing La on the Sr-site has an opposite effect. The corresponding experiments are in agreement with the computational predictions. Fast nanoparticle growth and activity enhancement are found in STO system with Sr vacancies and without La. The approach developed in this Letter could be used to study the mechanism of exsolution in other material systems, and possibly lead to the development of new compositions capable of nanoparticle exsolution with higher activity and stability.

M

materials capable of exsolving nanoparticles have been developed following these design principles, including La0.52Sr0.28Ti0.94 Ni0.06O3, La0.1Sr0.9Sc0.9Ni0.1O3‑δ, Sr0.95Ag0.05Nb0.1Co0.9O3‑δ, Sr 2 FeMo 0.65 Ni 0.35 O 6−δ , La 0.43 Ca 0.37 Ni 0.06 Ti 0.94 O 3‑δ , and La0.9Mn0.9Pt0.075Ni0.025O3‑δ.10,24,25,27−29 Related experimental work indicates that the nanoparticle growth is closely linked to the composition of the surface. For example, in the La0.4Sr0.4 Ti0.97Ni0.03O3‑δ system, it was expected that the exsolution of Ni is accompanied by an increased concentration of surface La3+.11 In a previous article, we have proposed a thermodynamic guideline to select the A- and B-site elements based on the redox stability of their oxide against the metal phase.24 However, this simple thermodynamic approach has severe limitations, and a systematic theoretical study of the role of defects and surface termination is needed to optimize the perovskites’ exsolution capabilities. The Hamada and Tian groups have used density functional theory (DFT) simulations to understand the mechanism of nanoparticle exsolution,32,33 with focus on Pd, a widely employed catalyst for the control of automotive emissions.12,13 Using LaFe1−xMxO3‑δ (M = Pd, Pt, and Rh) as a model system, the Hamada and Tian groups concluded that

etal nanoparticles are widely used in energy conversion and storage devices due to their high activity toward many catalytic reactions.1,2 Combining the high activity of metal nanoparticles with perovskites oxides has proven to be a successful strategy for the development of novel and efficient catalysts.3−7 The methods conventionally used to prepare perovskite/nanoparticle composites involve either chemical or physical deposition.8,9 However, at the high operating temperatures typical of solid oxide fuel cells (SOFCs) and solid oxide electrolysis cells (SOECs), deposited nanoparticles agglomerate.10,11 The exsolution of nanoparticles directly from the perovskite host can efficiently solve this problem because they are anchored to the perovskite host material.10,12,13 If an active transition metal M is placed in the B-site of an ABO3 perovskite parent material, forming AB1−xMxO3‑δ, the nanoparticles of M can be generated, under reducing conditions, by exsolution from the perovskite host. Such metal nanoparticles have shown an excellent catalytic activity and high stability in SOFCs, SOECs, and even low-temperature oxygen electrodes.14−29 Also, perovskites with nanoparticles exsolved in situ have improved the tolerance to sulfur poisoning and carbon coking compared to state-of-the-art electrodes.30,31 In a breakthrough article, the Irvine group has shown that A-site deficiency in the AB1−xMxO3‑δ perovskite facilitates the egress of transition metal nanoparticles.10 Many novel catalytic © 2018 American Chemical Society

Received: May 1, 2018 Accepted: June 18, 2018 Published: June 18, 2018 3772

DOI: 10.1021/acs.jpclett.8b01380 J. Phys. Chem. Lett. 2018, 9, 3772−3778

Letter

The Journal of Physical Chemistry Letters

Figure 1. Relaxed (100)-oriented and SrTiO-terminated STO slabs with Ni substitution at the surface (a), one layer below the surface (b), or in the bulk (c). The red arrows show the position of Ni, and the numbers close to the transition atoms (Ti and Ni) indicate their effective Bader charges.

corresponding surface system. Each slab model contained five layers of Ti atoms with 2 Ti in each layer. A vacuum layer of 20 Å was used to prevent slab−slab interactions. Three different orientations were considered where each orientation has two possible plane terminations. The unit cell we used to determine the orientations is given in Figure S1a. Calculation details are given in the Supporting Information. To calculate Eseg, we substituted one of the Ti atoms from the three uppermost Ti-containing layers with Ni. For illustration purposes, we show in Figure 1 the (100)-oriented, and SrTiO-terminated STO where Ni is substituted at the surface (a), in the bulk (c), and in between the two (b). An analogous method was employed to study the impact of substitution of Sr with La and the influence of A-site and O-site vacancies on the Eseg of Ni. To understand the physical mechanism of segregation, we also calculated the effective Bader charge of the various atoms present in the simulated materials. For a given atom, the Bader region boundary is found by computing the charge density minimizing surface, and the corresponding Bader charge is obtained by integrating the charge density within the obtained region.37 The Bader charge can be used to identify the charge redistribution and the columbic interactions within the crystal lattice studied. We first studied the Ni segregation in pristine STO systems (without introducing A-site and O-site vacancies). Following the Kröger−Vink notation,38 we indicate the Ni defect as NiTi ′′ because Ni2+ takes the place of Ti4+ and its charge is −2 compared to that of Ti4+. We calculated the total energy of the slabs with the Ni′Ti′ defect placed in one of the three uppermost positions. The slab model with Ni′Ti′ at layer 1 is regarded as the surface Ni configuration (Figure 1a), and its energy is Esurf. The ′′ at layer 3 corresponds instead to “bulk” Ni slab with NiTi (Figure 1c) whose energy is Ebulk. Figure 2a shows the energy of each individual configuration minus Ebulk, where we note that position 1 in Figure 2a corresponds to Eseg.24,32,33 The total slab energy decreases with NiTi ′′ moving from layer 3 to layer 1 with Eseg = −1.25 eV. Since the surface may have different orientations and terminations, we also calculated the Eseg for the 6 different configurations shown in Figure S1. All slab models show Eseg < 0. In particular, the (100)-oriented and SrTiO-terminated surface (see Figure 1) has the smallest Eseg. The result suggests that, among the cases studied, Ni segregation is particularly favored on the (100) surface. Therefore, in the following sections, we will use this particular orientation to investigate the impact of defects. Our results are in agreement with the experiments of the Irvine and Gorte groups, where the Ni nanoparticles were mostly found on (110) surface of the primitive cell in La0.52Sr0.28Ti0.94Ni0.06O3‑δ.11

oxygen vacancies enhance the surface segregation of Pd (as well as those of Pt and Ph).32,33 However, the surface termination, composition, and defect configurations were not considered in these studies. In principle, optimizing at least a few of these factors could facilitate the generation of nanoparticles. In this work, we focus on the exsolution of Ni nanoparticles because of the activity of Ni toward hydrogen reactions in SOFCs and SOECs.24,26−28 Starting from SrTiO3 (STO) as the parent material, we studied the LaxSr1−xTi1−yNiyO3‑δ composition.10,11,34−36 We used first-principle DFT calculations on a set of slab models to investigate the Ni segregation energy (Eseg), which is (somewhat loosely) defined as the energy difference between the system with Ni at the surface (Esurf) and the one with Ni in the bulk (Ebulk): Eseg = Esurf − E bulk

(1)

When Eseg > 0, Ni is energetically favored in the bulk. However, if Eseg < 0, Ni is preferably found on the surface. The calculations show that, due to a significant charge redistribution, the Ni segregation energy is particularly low (relative to the other cases studied) toward the (100)-oriented and SrTiOterminated surface. We should note that according to our notation, the (100) orientation is identical to the (110) orientation reported in the article by Irvine and Gorte groups (see the Supporting Information document for more details).11 More importantly, we determined that, while A-site and O vacancies promote the segregation of Ni, the substitution of Sr with La hinders it. We also validated the computations against experiments. In these experiments, we particularly focused on the role of A-site vacancies and on the impact of substituting Sr with La since these two properties can be controlled by adjusting the composition of the materials. In particular, we synthesized four different materials with or without A-site vacancy and with or without substituting La on Sr site. We examined the phase structure and near-surface Ni concentration in these materials and tracked the evolution of the area specific resistance (ASR) as a function of the reduction time. Consistently with the computations, the experiments confirm that the substitution of Sr with La impedes the nanoparticle egress while A-site vacancy promotes it. The experiments also suggest that the surface segregation of Ni might affect the nanoparticle growth and the SOFC performance, where the materials with smaller segregation energy show a higher surface Ni content, enhanced nanoparticle egress upon reduction, and faster enhancement of the activity. We used DFT+U to study the surface segregation of Ni in the STO-based systems with slab models to simulate the 3773

DOI: 10.1021/acs.jpclett.8b01380 J. Phys. Chem. Lett. 2018, 9, 3772−3778

Letter

The Journal of Physical Chemistry Letters

Figure 2. (a) Relative energy of Ni′Ti′, V′Sr′ and La•Sr at various layers of the (100)-oriented SrTiO-terminated STO slabs. (b) Eseg of Ni in different STO slabs. (c) Relaxed (100)-oriented and O2-terminated STO slabs with Ni substitution. (d) Relaxed (100)-oriented and SrTiO-terminated STO slabs with V′Sr′. (e) Relaxed (100)-oriented and SrTiO-terminated STO slabs with La•Sr. Only the three topmost layers are shown in (c−e). The top layer of each slab is labeled as “surface” and the bottom as “bulk”. The numbers close to the atoms indicate the effective Bader charges of Ti and Ni (c) or Sr and La (d, e).

slab, the O2-terminated one has more O atoms on the surface and it is therefore more negative. Consequently, the valence state of Ti (2.26 vs 2.29) and Ni (1.43 vs 1.54) are similar at the surface and in the bulk. As a result, Eseg is lower for the O2-terminated surface than for the SrTiO-terminated one. The A-site vacancies are expected to promote the exsolution of nanoparticles.32,33 To investigate the impact of an A-site vacancy (V′Sr′) on Eseg, we removed one Sr atom from each of the three uppermost layers of the STO slab models. Consistently with what we reported above, we computed the relative energy of the slabs with the V′Sr′ placed in each layer shown in Figure 2a. Similar to what we found for Ni′Ti′, the simulations indicate that ′′ is energetically favored on the surface. In fact, the presence VSr ′′ on the surface compensates for of the negatively charged VSr the positive surface charge and leads to a more effective charge redistribution (see Figure 2d). We also computed the slab ′′ and VSr ′′ arrangements in the system energy of all possible NiTi ′′ (see Figure S2). The lowest energy configuration with NiTi in the bulk, which corresponds to VSr ′′ at the surface, is regarded as the Ebulk. The most stable configuration with Ni′Ti′ at the surface has instead Eseg = −1.78 eV. It is worth noting that ′′ at the surface. This such arrangement corresponds to VSr ′′ (−1.25 eV, value is 0.53 eV smaller than the one without VSr Figure 2a), suggesting that the introduction of V′Sr′ is beneficial to the segregation of Ni. The introduction of V′Sr′ decreases the postive charge on the surface. As a result, the effective Bader charge of the Ni (see Figure S2a) increases to

As already outlined above, the (110) orientation of the primitive cell used in the reference is the same as the (100) orientation of the unit cell we used in this work. Detailed clarifications are given in the Supporting Information. The energetics of Ni segregation are strongly affected by the surface configuration.39,40 In particular, the dependence of Eseg on specific orientations and terminations appears to be correlated to the metal/oxygen atomic ratio on the surface. For example, for the (100)-oriented slab models, the SrTiOterminated surface has more metal atoms at the surface compared with the O2-terminated one and a smaller Eseg. The smaller Eseg is linked to a significant surface charge redistribution, which we examined using Bader charge analysis.37 In the (100)-oriented and SrTiO-terminated slab, the metal/oxygen atomic ratio on the surface layer is higher than it in the bulk. If all atoms remain at their nominal valence state, the surface will be positively charged. However, such positive charge is compensated by the reduction of the surface transition metal atoms. This is consistent with effective Bader charges shown in Figure 1. Ni exhibits a lower effective Bader charge than Ti (1.15 vs 1.81 at the surface and 1.30 vs 2.29 in the bulk) because it is capable of better compensating the high metal/ oxygen ratio and the corresponding excess surface charge. This effect contributes to favoring Ni segregation and, in turn, reducing the energy of the slab. As reported in Figure 2c, we also calculated the effective charges of Ti and Ni in a (100)-oriented and O2-terminated slab. In comparison to the SrTiO-terminated 3774

DOI: 10.1021/acs.jpclett.8b01380 J. Phys. Chem. Lett. 2018, 9, 3772−3778

Letter

The Journal of Physical Chemistry Letters

•• Figure 3. (a) Relative energy of NiTi ′′ at different layers with V•• O as shown in Figure S2. (b) ΔGseg of Ni with VO as a function of the O2 chemical . The O partial pressures at 900 and 1400 °C are shown as calculated from eq 3. potential difference and Eseg without V•• O 2

temperature (T) and the oxygen partial pressure (p) can be approximated as

1.55, a value higher than the corresponding value shown in Figure 1a. If a single La is introduced in the STO supercell to replace Sr, the bulk La•Sr has the lowest energy (Figure 2a). La appears to be favored in the bulk because of its higher valence and higher effective Bader charge compared to Sr (see Figure 2e). The surface La•Sr will further introduce excess positive charge on the surface. When both NiTi ′′ and La•Sr are present, as shown in Figure S2c, the most stable surface and bulk configurations of Ni′Ti′ are obtained when La•Sr is in the bulk, and the corresponding energy is indicated as Ebulk. Eseg is calculated to be −0.97 eV (Figure S 2d). The value is 0.28 eV larger than the one of the slab without La•Sr. The effective Bader charge of the surface Ni shown in Figure S2c is further reduced to 0.61 compared with the pristine case shown in Figure 1a. This is because the introduction of La•Sr increases the overall postive charge in the system. The charge redistribution and the columbic attraction between the bulk La•Sr and Ni′Ti′ may explain the weakened tendency for Ni to be on the surface. This result suggests that the introduction of La may prove detrimental to the surface Ni segregation. It has been observed experimentally that La and Sr also diffuse toward the surface during Ni exsolution, where a higher ratio of La atoms diffuses compared with that of Sr.11 A higher La concentration fills the surface A-site and may, in turn, limit the diffusion of Ni, hereby hindering the exsolution process. We also considered the effect of oxygen vacancies (V•• O ) on the Eseg. We obtained the most stable V•• O site for each of the three different Ni′Ti′ positions, such as Figure S3, where the most stable V•• O is the one closest to Ni′Ti′, probably because of the Coulombic attraction between the positively charged V•• O and negatively charged NiTi ′′. The presence of one V•• O in the slab only slightly promotes the Ni segregation, as the Eseg decreases from −1.25 eV to −1.28 eV (Figure 3a). However, we also need to consider the formation of V•• O during the Ni segregation process because of the following reaction: Ni″Ti,bulk + OOx F Ni″Ti,surf + V •• O +

1 O2 + 2e− 2

ij p yz ΔμO (T , p) = ΔμO0 (T , p0 ) + RT lnjjj 0 zzz jp z 2 2 k {

(3)

where Δμ0O2 can obtained from thermodynamic tables and p0 = 1 atm.41 The formation of V•• O during surface segregation requires extremely low O2 partial pressures, e.g., 10−11 atm at 900 °C (exsolution temperature) and 10−3.5 atm at 1400 °C (calcination temperature), as shown in Figure 3b. However, in a reducing atmosphere, e.g., in 97% H2 with 3% H2O at 900 °C, the O2 partial pressure can be as low as 10−19 atm according to thermodynamic equilibrium of H2, O2, and H2O. Under such conditions and in the presence of V•• O , Ni segregation is easier and ΔGseg is much smaller (−3 eV). These values suggest that reducing conditions facilitate the formation of V•• O , which in turn benefits the segregation of Ni toward the surface. While the terminations and orientations discussed above are difficult to control in bulk samples, the presence of VSr ′′ and La•Sr can be adjusted by modifying the materials’ composition. Therefore, we have synthesized four materials starting from STO, i.e., SrTi 0.9 Ni 0.1 O 3‑δ (STN10), Sr 0.8 Ti 0.9 Ni 0.1 O 3‑δ ( S T N8 ), L a 0 . 1 S r 0 . 9 T i 0 . 9 Ni 0 . 1 O 3 ‑ δ ( L S T N 1 0 ) , a n d La0.08Sr0.72Ti0.9Ni0.1O3‑δ (LSTN8). Experimental details are given in the Supporting Information. Based on our computations, the presence of VSr ′′ in STN8 and LSTN8 may promote the segregation of Ni compared with STN10 and LSTN10. Similarly, the presence of La•Sr in LSTN8 and LSTN10 is expected to hinder the segregation compared with STN8 and STN10. As a result, STN8 is expected to segregate Ni on the surface and exsolve nanoparticles most easily, while LSTN10 is projected be the worst among the four. To evaluate the computational results, we have investigated the lattice structure, Ni ratio at the surface and in the bulk, surface morphology after reduction, and symmetric cell performance of these four materials. The samples show similar tetragonal phases in agreement with previous reports, and small peaks in the XRD spectrum can be attributed to a rutile phase impurity (Figure S4a).10 The surface and bulk compositions are examined with XPS. Figure S5 shows the XPS spectra of the materials measured on the surface and in the bulk (after etching). The compositional ratios can be calculated from the peak areas and the corresponding sensitivity factors which are defined as the ratios between the atomic

(2)

where the left-hand side corresponds to the stoichiometric material with Ni in the bulk (Figure 1c). The right-hand side of eq 2 corresponds to a system with Ni at the surface and the formation of V•• O as well as the release of O2. The variation of the O2 chemical potential (ΔμO2) from (0 K, 0 atm) to 3775

DOI: 10.1021/acs.jpclett.8b01380 J. Phys. Chem. Lett. 2018, 9, 3772−3778

Letter

The Journal of Physical Chemistry Letters

Figure 4. (a) SEM image of STN8 after exsolution. The inset shows a high-resolution SEM image. (b−e) HAADF-STEM images of STN8 after exsolution and the corresponding EDX mappings of Ni, Ti, and Sr, respectively. (f,g) HAADF-STEM images of STN8 after exsolution. (i,j) Atomic scale HAADF-STEM images and EDX mapping of Ni.

compositions and the peak areas.42 The Ni/(Ni+Ti) and La/(La+Sr) ratios are given in Figure S4b. STN8 and LSTN8 with lower computed Eseg show higher surface Ni ratios compared with STN10 and LSTN10, and La deficiency is also observed at the surface. The results are in agreement with the DFT calculations. The Ni nanoparticle growth usually takes 12−70 h for La0.52Sr0.28Ti0.94Ni0.06O3‑δ and La0.4Sr0.4Sc0.9Ni0.1O3‑δ in a reducing H2-rich environment.1,10,11 We reduced all four materials in H2 at 900 °C for 15 h and examined the surface morphology after reduction. The surface of STN8 has nanoparticles with a diameter of ∼80 nm (Figure 4a). The lattice structure and composition of the nanoparticles were also examined with HAADF-STEM and EDX mapping, respectively. The elemental mappings in Figure 4b−e show that the nanoparticles are composed of pure Ni. From Figure 4f−i, a representative exsolved nanoparticle shows a plane distance of 2.04 Å, which agrees with the (111) interplane of metal Ni. We also found nanoparticles on all reduced materials except for LSTN10 (see Figure S6). To calculate the nanoparticle densities, we analyzed the SEM images of three different samples for each material with ImageJ. The densities are determined to be 0.37 ± 0.12, 1.2 ± 0.31, 0, and 0.51 ± 0.14 particles μm−2 for STN10, STN8, LSTN10, and LSTN8, respectively. The material with a lower segregation energy

shows larger particle density in accordance with the calculation results. We employed an additional indicator to assess the Ni nanoparticle exsolution, i.e., the ASR relaxation. Nonreduced perovskite materials are used as the electrodes in symmetric cells. Ni nanoparticles were exsolved during the EIS measurements in humid H2 with 3% H2O, leading to decreased ASRs (Figure S7a). The ASR decrease rate can be linked to the increase in the catalytic activity, which is, in turn, correlated to the exsolution rate of the Ni nanoparticles which contribute most of the activity. The four different materials tested show different ASR relaxations during reduction. The ASRs of STN10, STN8, and LSTN8 decrease as a function of time; by contrast, the ASR of LSTN10 increases. The ASR relaxation is consistent with the SEM results where there is no Ni nanoparticle for LSTN10. We then fitted the ASR as an exponentially decaying relaxation with the reduction time t: i ty ASR(t ) = ASR ∞ + (ASR 0 − ASR ∞)· expjjj− zzz k τ{

(4)

where ASR0 is the initial ASR of the electrode and ASR∞ is the final one. The relaxation time scale τ is regarded as an indicator of the elctrocatalytic time scale of the exsolution process. A more direct view of the ASR decrease is shown in Figure S7b, where the ASR is normalized. The fitted τ is 3.2, 0.10, and 0.41 h 3776

DOI: 10.1021/acs.jpclett.8b01380 J. Phys. Chem. Lett. 2018, 9, 3772−3778

Letter

The Journal of Physical Chemistry Letters

Figure 5. ASR as a function of reduction time of (a) STN10 and STN8 and (b) LSTN10 and LSTN8 measured at 800 °C in humid (3%) H2.

for STN10, STN8, and LSTN8, respectively, giving τSTN8 < τLSTN8 < τSTN10. These results suggest that the V′Sr′ is beneficial to the growth of Ni nanoparticles, while La•Sr may have a negative effect. This is also consistent with the conclusions obtained from the calculations illustrated above. We should note that, although STN8 has a faster decrease rate, the absolute ASR values of LSTN10 and LSTN8 are much lower than those of STN10 and STN8, as shown in Figure 5. This is attributed to the enhanced conductivity due to La doping.43−45 We have studied the energetics of Ni segregation in the STO-based perovskite framework. Our DFT calculations show that Ni preferentially segregates toward the (100)-oriented and SrTiO-terminated surfaces. The Ni surface segregation is energetically promoted by the presence of A-site and O-site vacancies, while the substitution of Sr2+ with La3+ hinders it. The insight from the ab initio computations was also validated experimentally by studying the impact of A-site vacancies and the substitution of Sr with La. We found a strong correlation between the calculated surface segregation energy and experimentally observed surface composition, nanoparticle density, and ASR relaxation time scales. More broadly, our work suggests that the DFT calculations on slab models could be used for designing novel compositions extending currently used defectchemical and thermodynamic design principles.

■

Province, China (No.2016A050503042), and the Science and Technology Program of Nansha District (No.2015CX009). L.X. and J.H. gratefully acknowledge the Science, Technology, and Innovation Commission of Shenzhen Municipality (JCYJ20150831142508365 and KQTD2016 022619565991).

■

(1) Aricò, A. S.; Bruce, P.; Scrosati, B.; Tarascon, J.-M.; Van Schalkwijk, W. Nat. Mater. 2005, 4, 366−377. (2) Zhang, Q.; Uchaker, E.; Candelaria, S. L.; Cao, G. Chem. Soc. Rev. 2013, 42, 3127−3171. (3) Zheng, N.; Stucky, G. D. J. Am. Chem. Soc. 2006, 128, 14278− 14280. (4) Jin, Z.; Xiao, M.; Bao, Z.; Wang, P.; Wang, J. Angew. Chem., Int. Ed. 2012, 51, 6406−6410. (5) Farmer, J. A.; Campbell, C. T. Science 2010, 329, 933−936. (6) Kim, J.-S.; Wieder, N. L.; Abraham, A. J.; Cargnello, M.; Fornasiero, P.; Gorte, R. J.; Vohs, J. M. J. Electrochem. Soc. 2011, 158, B596−B600. (7) Gorte, R.; Vohs, J. Curr. Opin. Colloid Interface Sci. 2009, 14, 236−244. (8) Madsen, B. D.; Kobsiriphat, W.; Wang, Y.; Marks, L. D.; Barnett, S. A. J. Power Sources 2007, 166, 64−67. (9) Bierschenk, D. M.; Potter-Nelson, E.; Hoel, C.; Liao, Y.; Marks, L.; Poeppelmeier, K. R.; Barnett, S. A. J. Power Sources 2011, 196, 3089−3094. (10) Neagu, D.; Tsekouras, G.; Miller, D. N.; Menard, H.; Irvine, J. T. Nat. Chem. 2013, 5, 916−923. (11) Neagu, D.; Oh, T.-S.; Miller, D. N.; Ménard, H.; Bukhari, S. M.; Gamble, S. R.; Gorte, R. J.; Vohs, J. M.; Irvine, J. T. Nat. Commun. 2015, 6, 8120. (12) Nishihata, Y.; Mizuki, J.; Akao, T.; Tanaka, H.; Uenishi, M.; Kimura, M.; Okamoto, T.; Hamada, N. Nature 2002, 418, 164−167. (13) Tanaka, H.; Uenishi, M.; Taniguchi, M.; Tan, I.; Narita, K.; Kimura, M.; Kaneko, K.; Nishihata, Y.; Mizuki, J. i. Catal. Today 2006, 117, 321−328. (14) Yang, L.; Xie, K.; Xu, S.; Wu, T.; Zhou, Q.; Xie, T.; Wu, Y. Dalton Transactions 2014, 43, 14147−14157. (15) Wei, H.; Xie, K.; Zhang, J.; Zhang, Y.; Wang, Y.; Qin, Y.; Cui, J.; Yan, J.; Wu, Y. Sci. Rep. 2015, 4, 5156. (16) Qin, Q.; Wu, G.; Chen, S.; Doherty, W.; Xie, K.; Wu, Y. Electrochim. Acta 2014, 127, 215−227. (17) Batlle-Vilanova, P.; Puig, S.; Gonzalez-Olmos, R.; VilajeliuPons, A.; et al. Int. J. Hydrogen Energy 2014, 1297−1305. (18) Xie, K.; Zhang, J.; Xu, S.; Ding, B.; Wu, G.; Xie, T.; Wu, Y. Fuel Cells 2014, 14 (6), 1036−1045. (19) Tsekouras, G.; Neagu, D.; Irvine, J. T. S. Energy Environ. Sci. 2013, 6, 256−266. (20) Xiao, G.; Wang, S.; Lin, Y.; Zhang, Y.; An, K.; Chen, F. ACS Appl. Mater. Interfaces 2014, 6, 19990−19996.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.jpclett.8b01380. Computational methods and details, materials preparation and characterizations, notation of the unit cell. (PDF)

■

REFERENCES

AUTHOR INFORMATION

ORCID

Ziheng Lu: 0000-0003-2239-8526 Francesco Ciucci: 0000-0003-0614-5537 Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS F.C., Y.G., Z.L., T.L.Y., and J.W. gratefully acknowledge the Research Grants Council of Hong Kong for support through the projects (16207615, 16227016, and 16204517), the Guangzhou Science and Technology Program (No.2016201604030020), the Science and Technology Planning Project of Guangdong 3777

DOI: 10.1021/acs.jpclett.8b01380 J. Phys. Chem. Lett. 2018, 9, 3772−3778

Letter

The Journal of Physical Chemistry Letters (21) Huang, Y.-H.; Dass, R. I.; Xing, Z.-L.; Goodenough, J. B. Science 2006, 312, 254−257. (22) Yang, C.; Yang, Z.; Jin, C.; Xiao, G.; Chen, F.; Han, M. Adv. Mater. 2012, 24, 1439−1443. (23) Liu, Q.; Dong, X.; Xiao, G.; Zhao, F.; Chen, F. Adv. Mater. 2010, 22, 5478−5482. (24) Gao, Y.; Chen, D.; Saccoccio, M.; Lu, Z.; Ciucci, F. Nano Energy 2016, 27, 499−508. (25) Zhu, Y.; Zhou, W.; Ran, R.; Chen, Y.; Shao, Z.; Liu, M. Nano Lett. 2016, 16, 512−518. (26) Sun, Y.-F.; Zhang, Y.-Q.; Chen, J.; Li, J.-H.; Zhu, Y.-T.; Zeng, Y.-M.; Amirkhiz, B. S.; Li, J.; Hua, B.; Luo, J.-L. Nano Lett. 2016, 16, 5303−5309. (27) Du, Z.; Zhao, H.; Yi, S.; Xia, Q.; Gong, Y.; Zhang, Y.; Cheng, X.; Li, Y.; Gu, L.; Świerczek, K. ACS Nano 2016, 10, 8660−8669. (28) Myung, J.-h.; Neagu, D.; Miller, D. N.; Irvine, J. T. Nature 2016, 537, 528. (29) Gao, Y.; Wang, J.; Lyu, Y.-Q.; Lam, K.; Ciucci, F. J. Mater. Chem. A 2017, 5, 6399−6404. (30) Haile, S. M. Acta Mater. 2003, 51, 5981−6000. (31) Yang, L.; Cheng, Z.; Liu, M.; Wilson, L. Energy Environ. Sci. 2010, 3, 1804−1809. (32) Hamada, I.; Uozumi, A.; Morikawa, Y.; Yanase, A.; KatayamaYoshida, H. J. Am. Chem. Soc. 2011, 133, 18506−18509. (33) Tian, Z.-x.; Uozumi, A.; Hamada, I.; Yanagisawa, S.; Kizaki, H.; Inagaki, K.; Morikawa, Y. Nanoscale Res. Lett. 2013, 8, 203. (34) Park, B. H.; Choi, G. M. Electrochim. Acta 2015, 182, 39−46. (35) Park, B. H.; Choi, G. M. J. Power Sources 2015, 293, 684−691. (36) Dayaghi, A. M.; Kim, K. J.; Kim, S.; Park, J.; Kim, S. J.; Park, B. H.; Choi, G. M. J. Power Sources 2016, 324, 288−293. (37) Henkelman, G.; Arnaldsson, A.; Jónsson, H. Comput. Mater. Sci. 2006, 36, 354−360. (38) Kröger, F. A.; Vink, H. J. In Solid State Physics; Seitz, F., Turnbull, D., Eds.; Academic Press: Cambridge, MA, 1956, Vol. 3, pp 307−435. (39) Kung, H. H. Transition metal oxides: surface chemistry and catalysis; Elsevier: Amsterdam, 1989. (40) Yu, R.; Hu, L.; Cheng, Z.; Li, Y.; Ye, H.; Zhu, J. Phys. Rev. Lett. 2010, 105, 226101. (41) Stull, D. R.; Prophet, H. JANAF Thermochemical Tables, DTIC Document, 1971. DOI: 10.6028/NBS.NSRDS.37 (42) Moulder, J. F.; Chastain, J.; King, R. C. Handbook of X-ray photoelectron spectroscopy: a reference book of standard spectra for identification and interpretation of XPS data; Physical Electronics: Eden Prairie, MN, 1995. (43) Okuda, T.; Nakanishi, K.; Miyasaka, S.; Tokura, Y. Phys. Rev. B: Condens. Matter Mater. Phys. 2001, 63, 113104. (44) Li, X.; Zhao, H.; Xu, N.; Zhou, X.; Zhang, C.; Chen, N. Int. J. Hydrogen Energy 2009, 34, 6407−6414. (45) Marina, O. A.; Canfield, N. L.; Stevenson, J. W. Solid State Ionics 2002, 149, 21−28.

3778

DOI: 10.1021/acs.jpclett.8b01380 J. Phys. Chem. Lett. 2018, 9, 3772−3778