Investigation of LiO₂ Adsorption on LaB_1−xB′_xO₃(001) for Li-Air Battery Applications: A Density Functional Theory Study

Article information

Abstract

1. Introduction

The use of fossil fuels has exacerbated various environmental concerns, particularly global warming. To reduce the dependence on fossil fuels, nowadays Li-air batteries, due to their high theoretical specific energy, have attracted interest for future electric vehicles, electrical energy storage devices, and other energy applications.1–3) However, many scientific issues such as low round-trip efficiency, short cycle life, poor rate capability, and electrolyte instability must be solved for the realization of affordable Li-air battery technology.4–9) Especially, the slow kinetic rate of the oxygen reduction reaction (ORR) or the oxygen evolution reaction (OER) on the cathode side has been widely known as a main obstacle to high energy efficiency.10) To overcome this problem, numerous attempts have been made to design new electrocatalysts with high activity for ORR or OER.11) In the past, perovskite oxides (ABO₃) have been nominated as potential ORR or OER catalysts12,13) due to their high catalytic activity, fast electronic and ionic conductivity, and low cost.12,14,15) For these reasons, many researchers have made efforts to enhance the cathode kinetics using LaBO₃-type perovskite oxides such as LaSrMnO₃,16,17) LaFeO₃,18) and LaSrCoO_3−δ 19) in Li-air batteries. However, in spite of these constant efforts, it is still difficult to find optimal combinations of components in perovskite oxides due to changes in catalytic properties with partial or full substitution at A- and B-sites.

Previous studies have provided efficient ways to improve the catalytic properties. Introducing an activity descriptor enables us to facilitate fast screening of perovskite catalysts for Li-air batteries.14,15) Using density functional theory (DFT) calculations, Nørskov et al.20) suggested a method that can assess the thermochemistry of the electrochemical reactions. They also verified the origin of the overpotential for oxygen reduction over Pt(111). Based on these results, Man et al.21) derived a descriptor (ΔG_O*–ΔG_HO*) for oxygen reduction using different kinds of oxide materials. These results has been widely applied in various electrochemical reactions including Li-ORR and OER to introduce useful descriptors. Choi et al.22) used the DFT calculated adsorption energy database of lithium oxide intermediates to examine the theoretical overpotentials for ORR and OER on Pd, Cu, and PdCu alloy surfaces with an ordered body centered cubic (B2-type) structure. They reported that the LiO₂ adsorption energy shows a good correlation with the ORR and/or OER overpotentials. Moreover, they demonstrated that controlling the LiO₂ adsorption energy by alloying the material with Pd and Cu can enhance the catalytic activity. A similar argument was also made by Kim et al.,23) who suggested that the overpotential of Li-OER decreases with decreasing of the strength of LiO₂ adsorption on Pt(111), Co(0001), and Pt₃Co(111). These results motivated us to estimate the catalytic activity on perovskite oxides by using only a simple descriptor, the LiO₂ adsorption energy.

In this study, we first use DFT calculations to perform a structural investigation of the LiO₂ adsorption on LaB_1−xB′_xO₃(001) (B, B′ = Mn, Fe, Co, and Ni, x = 0.0, 0.5). Then, charge analysis of the adsorption behavior is performed to understand the phenomenon in detail. Finally, the energetics of LiO₂ adsorption is investigated to rapidly predict promising candidate perovskite catalysts. In order to tune the adsorption energy, which is closely related to the catalytic activity, we dope the B-site metal site with other 3d-transition metal species.24–26)

2. Computational Methods

All periodic density functional theory (DFT) calculations were performed using the Vienna Ab initio Simulation Package (VASP).27,28) The Perdew-Burke-Ernzerhof (PBE) functional, based on the generalized gradient approximation (GGA), was employed.29) In addition, the DFT+U method30) with U_eff = 4.0 eV (Mn), 4.0 eV (Fe), 3.3 eV (Co), and 6.4 eV (Ni)31) was applied to reduce the self-interaction error. All calculations included spin polarization and were performed with a cutoff of 400 eV. Geometries were optimized until the forces on each atom were below 0.03 eV Å⁻¹. The optimized lattice constants were obtained using 2 × 2 × 2 supercells. An asymmetric slab structure was constructed by cleaving the bulk structure along the (001) plane. For the calculations of (2 × 2) surface unit cells with 4 atomic layers, an 8 × 8 × 1 Monkhorst-Pack32) grid k-point mesh was used. The slab models have two fixed bottom layers as the bulk positions and two fully relaxed top layers. A 15 Å vacuum layer was used to prevent interaction between the two periodic slabs. In all surface calculations, a dipole correction was applied in the perpendicular direction.33) We used BO₂-terminated (001) surfaces, which often show high catalytic activity due to B-site metals.14,15) Fig. 1 shows the bare and doped perovskite surface structures used in this study.

Fig. 1

Unit cell of (a) LaBO₃ and (b) LaB₅₀B′₅₀O₃ perovskite surfaces. La (lanthanum) atoms are green spheres, B or B′ (transition metal) atoms are purple or brown spheres, and O (oxygen) atoms are red spheres.

The adsorption energy of LiO₂ was calculated by

where EadsLiO2 is the adsorption energy of LiO₂ on the perovskite surfaces. E_LiO2/surf, E_LiO2(g), and E_surf are the total energies of surfaces with adsorbed LiO₂, an LiO₂ molecule in the gas phase, and the bare surface, respectively.34) For the calculations of LiO₂ in the gas phase, a large periodically repeating cubic box, approximately 10 Å on a side, was used as a unit cell. Table 1 indicates the geometric information of the optimized LiO₂ molecule, which shows good agreement with previous experimental measurements.35)

Table 1

Comparison between Calculated and Measured LiO₂ Geometric Information

Quantitative charge transfer analysis between adsorbates and surfaces allows us to deeply understand the adsorption phenomenon. Thus, an electron density analysis using the Bader charge model36,37) was carried out. A 12 × 12 × 1 Monkhorst-Pack32) grid k-point mesh was used for the Bader charge analysis.

3. Results and Discussion

3.1 LiO₂ adsorption on LaBO₃(001)

We first examined the adsorption phenomenon of LiO₂ molecules on undoped perovskite (LaBO₃) surfaces because these surfaces have relatively simple structures. Dathar et al. previously reported that the first e⁻ transfer step of Li- ORR on a metal surface can be divided into associative and dissociative mechanisms depending on the activation barrier (E_a) for dissociating oxygen molecules.38) They suggested that a Li atom reacts with an oxygen molecule on metal surfaces with high E_a (LiO₂ generation), whereas it reacts with dissociated oxygen atoms on surfaces with low E_a (OLiO generation). Unlike the case of metal surfaces, the initial adsorption mechanism on LaBO₃ -type perovskite surfaces is still unclear. However, the O₂ dissociation reaction often shows a high activation barrier (~ 1 eV) on A-site undoped LaBO₃-type perovskites,39,40) which would have few oxygen vacancies due to the good balance of charge. This means that the first e⁻ transfer step of Li-ORR on LaBO₃ perovskite surfaces may be associative adsorption. Hence, we here examined the adsorption energies and site preferences of the LiO₂ molecule, not the OLiO molecule.

Figure 2 shows the DFT-optimized LiO₂ adsorbed LaMnO₃ surface structure as a representative case. The O₂ part in LiO₂ is adsorbed on the top layer Mn atom. This is similar to previous reports in which the adsorption site of an O₂ molecule on an ABO₃ perovskite surface is the B-transition metal site.39–42) The Li atom in LiO₂ is adsorbed on the 4-fold site surrounded by four lattice oxygen atoms. This adsorption site preference of LiO₂ on the LaMnO₃ surface was almost identical for the other LaBO₃ perovskite surfaces examined in this study. However, the adsorption strength of LiO₂ shows clear differences. The trend in the adsorption energies on LaBO₃(001) (B = Mn, Fe, Co, and Ni) follows the order of atomic number of the B-transition metals in the periodic table (Table 2). This implies that the LiO₂ adsorption energy is mainly affected by the properties of transition metals in the B-site.

Fig. 2

(a) Side and (b) top views of LiO₂-adsorbed LaMnO₃ surface. La atoms are green spheres, Mn atoms are purple spheres, O atoms are red spheres, and Li atoms are blue-green spheres.

Table 2

Adsorption Energies of LiO₂ on LaBO₃(001) (B = Mn, Fe, Co, and Ni)

3.2 LiO₂ adsorption on LaB_0.5B′_0.5O₃(001)

Our results for the undoped perovskites in Section 3.1 provide useful insight into how we determined the binding site of LiO₂ on the B-site doped perovskite surfaces, LaB_0.5B′_0.5O₃(001). As can be seen in Fig. 1(b), there are two possible adsorption sites according to the two types of transition metals exposed on the surface (B and B′). Based on the results of adsorption on the undoped LaBO₃(001), we can reasonably assume that the O₂ part of LiO₂ is adsorbed on the top of B or B′ cations, while the Li atom in LiO₂ binds only on the 4-fold site surrounded by four lattice oxygen atoms. We thus calculated the adsorption energies by considering those two possible adsorption sites for all LaB_0.5 B′_0.5O₃(001) (Fig. 3 and Table 3).

Fig. 3

LiO₂ adsorption on (a), (b) Mn atoms and (c), (d) Co atoms in LaMn_0.5Co_0.5O₃(001). (a), (c) indicate the side view while (c), (d) indicate the top view. La atoms are green spheres, Mn atoms are purple spheres, Co atoms are deep blue spheres, O atoms are red spheres, and Li atoms are blue-green spheres.

Table 3

Adsorption Energies and Site Preferences of LiO₂ on LaB_0.5B′_0.5O₃(001) (B, B′ = Mn, Fe, Co, and Ni). The B- and B′-sites Mean the Adsorption Sites of O₂ in LiO₂

It is found that the adsorption energy and site preference of LiO₂ on LaB_0.5B′_0.5O₃ surfaces are not dependent on those characteristics of the undoped LaBO₃ or LaB′O₃ surfaces. For example, the adsorption strength of LiO₂ (−2.05 eV) on the Co atom of LaCoO₃(001) is larger than that (−1.75 eV) on the Mn atom of LaMnO₃(001) (Because an Li atom is always absorbed at a site surrounded by four oxygen atoms, here we only mention the adsorption sites of the O₂ part in LiO₂). On the other hand, in LaMn_0.5Co_0.5O₃(001), the adsorption strength of LiO₂ (−2.21 eV) on the Mn atom is larger than that (−2.06 eV) on the Co atom. Furthermore, the adsorption strength on LaMn_0.5Co_0.5O₃(001) becomes much stronger than that on LaMnO₃(001) or LaCoO₃(001). These unpredictable tendencies of the site preferences and adsorption energies of LiO₂ imply that it is hard to minutely tune the adsorption energy (further, the activity) by doping or mixing of different metals in perovskite oxides. As a result, determining the fundamental origin of the characteristic adsorption phenomena must be definitively resolved and is a subject for further study.

3.3 Charge density difference analysis

Adsorption is often accompanied by charge density redistribution. Thus, to understand the adsorption phenomenon in detail, we analyzed the charge density difference using Bader charge analysis.36,37) Here, we demonstrate the results on LaMnO₃(001) as a representative case. Fig. 4(a) indicates that charges of −0.20e, −0.06e, and −0.06e transfer from the LaMnO₃ surface to the O_t, O_b, and Li in the LiO₂ molecule, respectively (Here, O_t, and O_b indicate the atomic oxygens located at the top and bottom, respectively, of the adsorbed LiO₂ molecule). Overall, −0.32e transfers from the surface to the adsorbate (LiO₂). This tendency was also observed in the other perovskite surfaces we examined in this study. O_t has the most negative charge state and O_b and Li have relatively lower negative charge states after adsorption. Moreover, the total charge density differences of adsorbed LiO₂ always have negative values; these values are related to the adsorption energies of LiO₂. Fig. 4(b) shows the adsorption energies as a function of the changes in electron density on the ten bare or doped perovskite surfaces. Overall, the adsorption strength increases with increasing transfer of electrons from the perovskite surface to the adsorbed LiO₂ molecule. These results indicate that the adsorption strength of the LiO₂ molecule is mainly determined by electron charge transfer between the adsorbate and the perovskite oxide surface.

Fig. 4

(a) Changes in the electron density upon LiO₂ adsorption on LaMnO₃(001). Yellow and cyan colors indicate decreased and increased electron densities, respectively. O_t and O_b indicate the atomic oxygens located at the top and bottom of the adsorbed LiO₂ molecule, respectively. The isosurface contour is plotted with a charge density value of 0.002 e/ų. (b) Adsorption energies as a function of the charge density differences on the LiO₂ adsorbed LaB_1−xB′_xO₃(001) (B, B′ = Mn, Fe, Co, and Ni, x = 0.0, 0.5). Each point indicates the pure or mixed B-site cations used in this system.

3.4 Significance of LiO₂ adsorption analysis for Li-air battery cathodes

As mentioned earlier, LiO₂ adsorption energy can be a useful descriptor for the activity because the weak adsorption strength of LiO₂ is closely associated with the low overpotentials of ORR or OER.22,23) To estimate the activity of electrocatalysts for Li-air batteries, we assessed the adsorption energies of LiO₂. Fig. 5 shows the DFT-calculated adsorption energies found in this study (reorganized from Tables 2 and 3). More red color means a weaker strength of LiO₂ adsorption, while more blue color indicates stronger strength. Our results suggest that among the lanthanum based perovskites LaMnO₃, LaMn_0.5Fe_0.5O₃, or LaFeO₃ would be good candidate materials as Li-air battery cathodes in terms of the reaction kinetics due to the weak adsorption strength of LiO₂. Indeed, some of these materials such as LaSrMnO₃ or LaFeO₃ have already shown high electrochemical performance in Li-air batteries.16–18) This implies that our DFT results at the atomic scale provide a plausible explanation of the experimental results. To elucidate the detailed mechanism of ORR or OER in Li-air battery electrodes, a more thorough investigation for each elementary step will be required.

Fig. 5

Adsorption energies of LiO₂ on LaB_1−xB′_xO₃ (B, B′ = Mn, Fe, Co, and Ni, x = 0.0, 0.5) perovskite surfaces.

In addition, unlike the case on metal surfaces,22,23,38,43) the adsorption phenomenon of Li-ORR or OER intermediates on perovskite surfaces have not been precisely determined at the atomic scale. Therefore, adsorption analysis of LiO₂, a key intermediate of Li-ORR or OER, on a wide range of perovskite surfaces may provide useful insight into the adsorption study of the other intermediates.

4. Conclusions

In this study, we carried out structural, charge, and energetic investigations of LiO₂ adsorption behaviors on LaB_1−xB′_xO₃(001) (B, B′ = Mn, Fe, Co, and Ni, x = 0.0, 0.5) using DFT+U calculations. Unlike that of metal catalysts, the adsorption of lithium oxide intermediates on perovskite surfaces has hardly been examined. Our structural analysis showed that the adsorbed LiO₂ has a similar adsorption site preference and structure in all BO₂-terminated LaB_1−xB′_xO₃(001). Charge analysis demonstrated that the charge transfer from the perovskite surface to the LiO₂ molecule upon adsorption is a crucial factor for LiO₂ adsorption. LiO₂ adsorption energy was used as an activity descriptor to estimate the catalytic activities on LaB_1−xB′_xO₃(001). Our results showed that LaMnO₃, LaMn_0.5Fe_0.5O₃, or LaFeO₃ are promising candidate catalysts due to their weak LiO₂ adsorption, even though the cathode performance is dependent on a number of other factors. The screening results we present here also offer useful insight for experimental research into a rational design of perovskite catalysts for Li-air batteries.

Article information Continued