Arxiv:1712.02156V3 [Cond-Mat.Mtrl-Sci] 22 Feb 2018

Interfacial Strain Eﬀects on Lithium Diﬀusion Pathways in the Spinel Solid Electrolyte Li-Doped MgAl2O4

Conn O’Rourke∗ and Benjamin J. Morgan† Department of Chemistry, University of Bath, Bath, BA2 7AX (Dated: February 23, 2018)

(Li,Al)–co-doped magnesium spinel (LixMg1−2xAl2+xO4) is a solid lithium-ion electrolyte with potential use in all-solid-state lithium-ion batteries. The spinel structure means that interfaces with spinel electrodes, such as LiyMn2O4 and Li4+3zTi5O12, may be lattice-matched, with potentially low interfacial resistances. Small lattice parameter differences across a lattice-matched interface are unavoidable, causing residual epitaxial strain. This strain potentially modifies lithium diffusion near the electrolyte–electrode interface, contributing to interfacial resistance. Here we report a density functional theory study of strain effects on lithium diffusion pathways for (Li,Al)–co-doped magnesium spinel, for xLi = 0.25 and xLi = 0.5. We have calculated diffusion profiles for the unstrained materials, and for isotropic and biaxial tensile strains of up to 6%, corresponding to {100} epitaxial interfaces with LiyMn2O4 and Li4+3zTi5O12. We find that isotropic tensile strain reduces lithium diffusion barriers by as much as 0.32 eV, with typical barriers reduced by ∼ 0.1 eV. This effect is associated with increased volumes of transitional octahedral sites, and broadly follows qualitative changes in local electrostatic potentials. For biaxial (epitaxial) strain, which more closely approximates strain at a lattice-matched electrolyte–electrode interface, changes in octahedral site volumes and in lithium diffusion barriers are much smaller than under isotropic strain. Typical barriers are reduced by only ∼ 0.05 eV. Individual effects, however, depend on the pathway considered and the relative strain orientation. These results predict that isotropic strain strongly affects ionic conductivities in (Li,Al)–co-doped magnesium spinel electrolytes, and that tensile strain is a potential route to enhanced lithium transport. For a lattice-matched interface with candidate spinel-structured electrodes, however, epitaxial strain has a small, but complex, effect on lithium diffusion barriers.

INTRODUCTION highly tortuous, resulting in large interfacial resistances [6]. Lithium-ion batteries are widely used for energy stor- One strategy for developing all–solid-state batteries age. Historically, the principal applications have been with low interfacial resistances is to choose electrolyte– in consumer electronics, but lithium-ion batteries are in- electrode combinations that are lattice-matched [6, 7]. If creasingly finding use in other sectors, such as electric the cathode, electrode, and anode have mutually com- vehicles and grid-scale storage. This success has been patible crystal structures that allow epitaxially coherent tempered by concerns about the stability and safety of interfaces, the diffusion pathways across these solid–solid the liquid electrolytes found in commercial lithium-ion interfaces are expected to be relatively unobstructed, batteries. Organic liquid electrolytes are flammable, and giving low interfacial resistances. Lattice matching be- present an explosion risk if a cell were to short-circuit or tween electrodes and a solid lithium-ion electrolyte was suffer mechanical failure. One proposed solution is to re- first described by Thackeray and Goodenough in 1985 place the liquid electrolytes with electrochemically inert [7], who suggested using spinel-structured materials for solid lithium-ion conductors. This would eliminate the each of the anode, cathode, and electrolyte. Spinel explosion risk, remove the need for costly protection cir- structured electrodes, such as LiyMn2O4 lithium man- cuitry, and allow possible battery miniaturization [1–3]. ganate cathodes and Li4+3zTi5O12 lithium titanate an- odes are well known, and are already used in commercial A number of promising lithium-ion solid electrolytes batteries. Spinel-structured electrolytes, however, have exist; some with ionic conductivities comparable to those long proved elusive. In 2013, Rosciano et al. synthe- of liquid electrolytes [2–4]. For a solid lithium-ion elec- sised a new class of spinel-structured lithium-ion electrolyte to be commercially viable it must be electrically arXiv:1712.02156v3 [cond-mat.mtrl-sci] 22 Feb 2018 trolytes based on (Li,Al)–co-doped magnesium spinel and chemically stable under cell-operating conditions, (Li Mg Al O ) [6, 8, 9]. This development has re- and have high ionic and negligible electronic conductiv- x 1−2x 2+x 4 vived interest in the possibility of all-solid-state lithium- ities. In an all–solid-state lithium-ion battery, device ion batteries constructed to have a coherent face-centred– performance also depends on the interfacial resistance cubic (fcc) oxide lattice shared across the anode, elec- between the electrolyte and electrodes [5]. If a solid trolyte, and cathode. electrolyte–electrode pair have crystal structures or lat- tices that are incommensurate, the electrolyte–electrode Lattice matched electrode–electrolyte interfaces are interface is expected to be mismatched. Lithium-diffusion also interesting in the context of protective coatings for pathways across these mismatched interfaces may be cathodes used with conventional electrolyte chemistries. 2

˚ ˚ Coating cathode materials with solid electrolyte thin System aexpt. [A] Ref. aPBEsol [A] films can enhance cell performance and increase oper- Li0.28Mn2O4 8.043 [18] ating lifetimes, by inhibiting electrolyte–electrode side- Li0.74Mn2O4 8.144 [18] LiMn O 8.221 [18] reactions, and improving the mechanical resilience of the 2 4 Li4+3zTi5O12 8.360 [19] cathode during lithium insertion and extraction [10–13]. MgAl2O4 (xLi = 0) 8.04 [6] 8.051 Solid thin film coatings have also been used in solid- Li0.25Mg0.5Al2.25O4 (xLi = 0.25) 7.96 [6] 7.996 state lithium-ion batteries as a separating layer between Li0.5Al2.5O4 (xLi = 0.5) 7.89 [6] 7.907 the cathode and a solid electrolyte, to reduce interfacial resistance [14–16]. In both these cases, the interfacial TABLE I. Lattice parameters for lithium manganate and resistance between the cathode and the thin-film coat- lithium titanate spinels, and for LixMg1−2xAl2+xO4, from ex- periment and as calculated (PBEsol) for this work. ing can limit cell performance. Using lattice-matched solid electrolytes as thin film surface layers is one strategy to design low resistance cathode–thin-film interfaces. Epitaxial matching between a spinel-structured cathode Carlo (kMC) simulations to model long time-scale diffu- and a spinel-structured surface coating has been demon- sion. These kMC simulations predicted that increasing strated by Li et al. who have coated spinel LiMn2O4 with xLi from 0.25 to 0.50 causes the lithium diffusion coeffi- 2 Li4Ti5O12, giving a coherent epitaxial interface between cient to increase by ×10 , with a corresponding decrease the two materials [17]. in activation energy from 0.61 eV to 0.45 eV.1 Perfect lattice-matching between structurally compat- Despite this predicted increase in lithium diffusion co- ible electrode–electrolyte pairs is not achievable in prac- efficient with (Li,Al)-dopant concentration, Mees et al. tice, and any coherent lattice-matched interface will ex- suggested that the optimal composition for use with hibit some residual strain. For example, the lattice pa- the electrodes LiyMn2O4 or Li4+3zTi5O12 is xLi ≈ 0.3. rameter of MgAl2O4 differs from those of LiyMn2O4 This effective limit was proposed with two considera- and Li4+3zTi5O12 by 0.04%–2.3% and 4.0%, respectively tions in mind. First, that high lithium content may [18, 19]. (Li,Al)-codoping of MgAl2O4 causes the lat- promote phase separation to poorly conducting LiAl5O8 tice parameter to contract as the lithium concentration and MgAl2O4. Second, because the bulk lattice pa- is increased (see Table I). The degree of interfacial strain rameter of (Li,Al)–co-doped spinel decreases as xLi in- therefore depends not only on the choice of interfacing creases, a higher lithium content corresponds to a larger electrode, but also on the electrolyte stoichiometry. lattice-parameter mismatch with the spinel electrolytes Residual strain at electrolyte–electrode interfaces may LiyMn2O4 and Li4+3zTi5O12. This, in turn, was sug- not be without consequences. Lattice strains modify lo- gested to be likely to increase interfacial resistance, and cal atomic geometries and distort the potential energy reduce overall device performance. Subsequent exper- surface that defines lithium diffusion pathways. This af- imental work has shown that (Li,Al)–co-doped spinel fects diffusion barrier heights, and hence ionic conductiv- can be synthesised at high dopant concentrations (up ities. The effect of interfacial strain on ionic conductivi- to xLi = 0.4) without phase separation [9]. The first ties has been widely studied in fuel cell materials [20–27], concern regarding phase stability then, may be (at least motivated by the possibility of straining thin film elec- partially) avoided through careful synthesis. The second trolytes to enhance their oxide-ion conductivities, and question, however; the effect of epitaxial strain on lithium hence reduce device operating temperatures. More re- diffusion, in particular as xLi increases; remains open. cently, the concept of “strain engineering” has also been Here we report density functional theory (DFT) considered as a strategy for enhancing lithium conductiv- climbing-image nudged–elastic-band (CI-NEB) [39] cal- ity in lithium-ion electrodes and electrolytes [28–37]. The culations of lithium diffusion pathways in unstrained, effect of interfacial strain is particularly pertinent for an isotropically strained, and epitaxially (biaxially) strained electrolyte such as (Li,Al)–co-doped MgAl2O4, where in- (Li,Al)–co-doped MgAl2O4. We have performed these terest is motivated by the possibility of lattice-matching calculations to better understand the consequences with spinel-structured electrolytes. of strain at a hypothetical lattice-matched interface Computational modelling provides a powerful tool for between (Li,Al)–co-doped spinel and the electrodes studying how factors such as stoichiometry or strain can LiyMn2O4 and Li4+3zTi5O12. Our calculations assume affect lithium diffusion. To date, the only computational an implicit coherent interface across the full range of study of lithium transport in (Li,Al)–co-doped magnesium spinel was performed by Mees et al. [38]. These au- thors modelled lithium diffusion at dopant concentrations 1 of x = 0.125, x = 0.25, and x = 0.50, by calculat- Interestingly, the local DFT-NEB diffusion barriers for lithium Li Li Li motion calculated by Mees et al. [38] are broadly independent of ing lithium-ion diffusion barriers using density-functional dopant concentration, suggesting that the dominant effect of sto- theory (DFT) nudged–elastic-band (NEB) calculations, ichiometry on the lithium diffusion coefficient is due to changes These diffusion barriers were then used in kinetic Monte in the connectivity of available diffusion pathways. 3 dopant stoichiometries considered. Increasing xLi in- (a) x = 0 creases the degree of lattice mismatch with respect to both candidate electrodes. At high mismatch values it (Mg8)[Al]16O32 becomes increasingly likely that dislocations or other extended defects form, which relieve the interfacial strain, and which may have additional effects on lithium transport. In practice, high quality interfaces may require thin films of electrolyte, which are more able to preserve epitaxy under large lattice misfits [40]. We find that in (Li,Al)–co-doped spinel, isotropic tensile strain reduces lithium diffusion barriers, from ∼ (b) x = 0.25 0.4 eV to ∼ 0.25 eV under strains of up to 6%. This effect is correlated with changes in the volume of the octahedral (Li2Al2Mg4)[Al]16O32 site at the midpoint of each diffusion pathway, and differences in diffusion barriers are approximately correlated with differences in electrostatic potential along specific paths. For anisotropic strain, which approximates an ideal coherent electrolyte–electrode interface, the transi- tion site volume changes are much smaller, and the corresponding effect on the lithium-diffusion barrier height is weaker. While the effect of epitaxial strain on most barriers is a small reduction (∼ 0.05 eV at 6% strain) (c) x = 0.5 the quantitative details are more complex than in the isotropic strain case: changed to barrier heights depend (Li4Al4)[Al]16O32 on the choice of diffusion path being considered, and the orientation relative to the applied strain. These results indicate that isotropic strain is expected to have a large impact on ionic conductivities in (Li,Al)–co-doped spinel electrolytes, with tensile strain being a potential route to enhanced lithium transport. For “lattice-matched” electrolyte–electrode systems, however, providing interfaces remain coherent, the residual epitaxial strain is not FIG. 1. Example 56 atom unit cells for undoped and (Li,Al)-doped spinel structures with compo- expected to strongly affect lithium diffusion barriers. sition (AlxMg1−2xLix)[Al2]O4. (a) The undoped parent spinel (xLi = 0): (Mg8)[Al16]O32, (b) 50% doped (xLi = 0.25): (Al2Li2Mg4)[Al16]O32, (c) Fully doped (xLi = 0.5): METHODS (Al4Li4)[Al16]O32. Mg atoms are orange, Al atoms are blue, Li atoms are green, and O atoms are red. All calculations were performed using the plane-wave DFT code VASP [41, 42]. Exchange and correlation ef- 2+ 3+ fects were approximated by the revised Perdew-Burke- ichiometry (A )[B2 ]O4. In the normal magnesium- 2+ Ernzerhof generalized gradient approximation (GGA) spinel structure, (Mg)[Al2]O4, Mg cations occupy the PBEsol functional [43]. The pseudopotential method was tetrahedral 8a Wycoff positions, Al3+ cations occupy the used in the form of projector augmented wave (PAW) octahedral 16d positions, and O2− anions occupy the 32e pseudopotentials to treat core electrons [44] with 3 va- positions [45, 46]. Upon co-doping with {Li,Al}, pairs of lence electrons for Al (3s23p1), 1 valence electron for Li Mg2+ cations are substituted in equal proportion by Li+ (2s1), 6 valence electrons for O (2s22p6), and 2 valence and Al3+, to maintain charge neutrality, giving a compo- 2 electrons for Mg (3s ). All calculations were performed sition of (AlxMg1−2xLix)[Al2]O4 [6, 38]. The maximum on 56 atom cells, and k-space sampling used a (3 × 3 × 3) possible dopant content is xLi = 0.5 (Fig. 1). Monkhorst-Pack grid. The plane-wave cut-off energy was For our calculations, we consider dopant concentra- 600 eV for calculations with variable cell shape and vol- tions of xLi = 0.25 and xLi = 0.5. Substitutional dop- ume, and 550 eV for calculations with fixed cell shape. ing at the Mg sites lowers the crystal symmetry, making Geometry optimisations, including the CI-NEB calcula- the A-site cations non-equivalent. To account for this tions were, were deemed converged when the forces on A-site disorder we identified all symmetry inequivalent −1 ions were less than 1 × 10−2 eV A˚ . structures at each dopant concentration using the bsym The conventional spinel structure belongs to the Fd3m¯ code [47]. In a 56 atom cell, this gives 7 structures at spacegroup, and has 56 ions in the unit cell, with sto- xLi = 0.25 and 4 structures at xLi = 0.5. Fig. 1 shows ex- 4

The different strain protocols; unstrained, isotropic, and epitaxial; are illustrated schematically in Fig. 3. In all cases, we consider strains corresponding to model {100}A|{100}B coherent interfaces between the doped (Mg,Al)-spinel and the target electrode (lithium manganate or lithium titanate spinel). Li4+3zTi5O12 is a “zero-strain” material, whose lattice parameters change by < 0.1% upon lithium intercalation and extraction [19, 48, 49]. The lattice parameters of LiyMn2O4, however, expand and contract on lithium insertion and FIG. 2. Li+ diffusion pathway between neighbouring A cation extraction, varying from 8.04 A˚ (0.04% strain versus sites via the 16c octahedral vacancy (yellow). MgAl2O4) at xLi = 0.28, to 8.36 A˚ (2.3% strain versus MgAl2O4) at xLi = 1.0 [18]. For our calculations we consider an intermediate LiyMn2O4 lattice parame- amples of these doped structures alongside the undoped ter of 8.144 A˚ (1.3% strain versus MgAl2O4), which cor- + magnesium-spinel structure. Li diffusion between the responds to a stoichiometry of xLi ≈ 0.74. This gives tetrahedral A-sites proceeds via 8a–16c–8a paths, with three strain values that span our range of interest, for the vacant 16c octahedra connecting the end-point tetra- each diffusion pathway. The behaviour at intermediate hedra (Fig. 2). strains, such as those corresponding to fully inserted or In this study we focus on vacancy–mediated lithium- delithiated LiyMn2O4, can be extrapolated by interpo- ion transport, following the approach of Mees et al. [38]. lating the calculated data. For each inequivalent doped structure, we generated the full set of possible lithium-diffusion pathways by remov- (a) unstrained (b) isotropic strain (c) anisotropic strain ing, in turn, each of the A-site lithium ions. At a dopant (epitaxial) content of xLi = 0.25, three of the seven initial symmetry inequivalent structures have viable lithium diffusion pathways, where a lithium ion can move from an occupied A site, through an octahedral 16c site, into a previously unoccupied destination A-site. These three structures have one unique pathway each, which can be traversed in either direction. At a dopant content of xLi = 0.5, three of the four symmetry inequivalent structures have FIG. 3. 2D schematic illustrating the unstrained, isotropi- viable lithium-diffusion pathways. Two of these struc- cally strained, and anisotropically strained calculations. (a) Unstrained: all lattice parameters are relaxed during the DFT tures have a single pathway, and the third structure has DFT two. calculation (a, b, c = a ) (b) Isotropic strain: all three lattice parameters are scaled to match those of the target elec- To sample these different diffusion pathways, we con- trode (VCell = Velectrode) (c) Anisotropic strain: the two in- sider all three inequivalent paths at xLi = 0.25, and three plane lattice parameters are scaled to match those the target of the available paths at xLi = 0.5. For each lithium con- electrode, and the perpendicular lattice parameter is relaxed centration, this gives a set of pathways that includes all ([a, b = aelectrode, c = cDFT], [b, c = aelectrode, a = aDFT], electrode DFT possible combinations of initial and final A-site coordina- [a, c = a , b = b ]) tion environments within a 56 atom unit cell. For each candidate pathway, we have performed a series of CI- In addition to potential energy profiles from our CI- NEB calculation to evaluate the potential energy profile NEB calculations, we also present electrostatic potential of a diffusing lithium ion [39]. We denote specific path- profiles along each path. These are defined as the aver- p ways using Sx, where S describes the target strain (U = age electrostatic potential at the site of the mobile ion, unstrained, LM = lithium manganate lattice parameters, evaluated at the optimised geometry of each NEB im- LT = lithium titanate lattice parameters), x is the dopant age. If a diffusing lithium ion is approximated as a +1 concentration xLi, and p enumerates the pathways. An point charge, then the potential energy surface is charac- additional superscript R indicates a “reversed” pathway terised entirely by the electrostatic interactions between in cases where the diffusion profile is not symmetric. To the mobile lithium ion and the surrounding lattice. Pre- model isotropic strain, all three lattice parameters are vious computational studies of lithium-ion diffusion in equally scaled to match that of the target electrode. For electrodes have shown that electrostatic potential pro- anisotropic epitaxial strain, only two of the lattice pa- files often give a good approximate description of diffu- rameters are adjusted to match the target lattice. The sion barrier profiles [50–52]. In real systems, however, perpendicular lattice parameter is adjusted to minimise the interactions between mobile lithium ions and the the total cell energy. host lattice are not purely electrostatic. Lithium ions 5 also experience short-ranged repulsion from nearby lat- RESULTS tice ions, due to overlapping valence electron densities. The detailed shape of a diffusion potential energy profile Unstrained (Li,Al)–co-doped spinel therefore depends on the balance of the electrostatic and short-ranged repulsive interactions.2 In general, short- We first consider lithium diffusion path- ranged repulsion is expected to be reduced under ex- ways in unstrained (Li,Al)–co-doped spinel, pansive strain. Our inclusion of electrostatic potential (Al Mg Li )[Al ]O , using DFT-optimised lat- energy profiles is motivated, in part, by an interest in x 1−2x x 2 4 tice parameters for each dopant concentration. The the extent to which this conceptually simple metric de- potential energy profiles for the minimum-energy path- scribes lithium-ion diffusion in (Li,Al)–co-doped spinel ways for lithium diffusion are shown in Fig. 4, and electrolytes [51, 52]. the barrier heights are collected in Fig. 5. The energy To better understand the role of A-site cation disorder barriers for diffusion between adjacent A-sites range on lithium diffusion profiles, we also parametrise how the from 0.38 eV to 0.45 eV at xLi = 0.25, and from 0.27 eV A-site coordination environments differ from those in un- to 0.48 eV at xLi = 0.50. These energies are broadly doped MgAl2O4. We have followed the analysis of Mees consistent with those from the previous DFT NEB study et al., who noted that the energy of different co-doped of Mees et al. [38], and with values from NMR analysis spinel structures is correlated with the “average” local of lithium jumps between lattice sites in these materials oxidation state for the A-site cations, µi [38]. For a spe- [6]. cific A-site, µ is defined as the average formal oxidation i (Li,Al)-codoping of MgAl O introduces disorder state; O = +1, O = +2, O = +3, O = 0; 2 4 Li Mg Al vacancy across the A-sites, and makes the lithium-occupied A- of the ions occupying the central A-site, and the four sites non-equivalent. From the perspective of a cation nearest-neighbour A-sites: occupying a particular A-site, the local effect of this disorder is to introduce variation into the nearest-neighbour   A-site occupation: each neighbouring A-site may now 1  X  contain Li, Mg, or Al, or be vacant. For each lithium dif- µi = Oi + Oj (1) 5 fusion pathway, the nearest-neighbour A-site occupations j∈{nnA} i for the end-point configurations allows paths to be clas- sification as symmetric or asymmetric. Pathways with Structures with lower energies tend to have average µ identical nearest-neighbour A-site coordination environ- values closer to +2, i.e. the value obtained for A-site Mg ments at both end-points are symmetric, while those with ions in the undoped parent structure. Deviations from different coordination environments are asymmetric. this “optimal” average nearest-neighbour oxidation state For the symmetric pathways, the diffusion barrier can be quantified by calculating the root-mean-square height is necessarily equal for diffusion in both direc- difference between µi and the optimal value of +2 for all tions. For asymmetric pathways, however, the end-point occupied A-sites in each structure: energies may differ, and the potential energy profile relative to the starting structure depends on the direction of r P (µ − 2)2 lithium diffusion. For the asymmetric pathways consid- σ = i i . (2) A 7 ered here, we present data for lithium diffusion in both forward and reverse directions, with the latter indicated with a superscript R. Inputs and outputs for our VASP calculations are The relative energies of the two endpoints for an asym- openly available under the CC-BY-SA-4.0 license [53]. metric path can be analysed in terms of σ (Eqn. 2), This dataset includes Python scripts for extracting the A which describes the average A-site oxidation-state devia- NEB profiles, electrostatic potential profiles, and vol- tion from that of MgAl O . For the U3 path, the lower umes of transitional octahedral sites from the DFT calcu- 2 4 0.25 energy end-point gives σ = 0.1069, and the higher en- lation outputs, and for calculating σ (Eqn. 2). Analysis A A ergy end-point gives σ = 0.3207. Similarly, for the U3 codes that produce Figs. 4–10 are available as a Jupyter A 0.50 path, the lower energy end-point gives σ = 0.2, and notebook [54–56], published under the MIT license. Figs. A the higher energy end-point gives σ = 0.4721. In both 1 and 2 were generated using VESTA [57]. A cases, the lower energy end-point gives the lower value of σA, indicating a more uniform arrangement of formal charges on the spinel lattice. The electrostatic potential profiles approximately fol- 2 For this analysis, we neglect lithium ion polarisation and disper- low the potential energy profiles. For each path, the cal- sion interactions. Both contributions are expected to be small, culated electrostatic potential barrier typically overesti- due to the compact 1s2 electron configuration of Li+. mates the NEB barrier. This is consistent with lithium 6

0.6

0.4 ]

V 0.2 e [

B 0.0 E N

E 1 2 3 3R −0.2 U0.25 U0.25 U0.25 U0.25 Δ LM1 LM2 LM3 LM3R −0.4 0.25 0.25 0.25 0.25 1 2 3 3R LT0.25 LT0.25 LT0.25 LT0.25 −0.6

1.0

0.5 ] V [

i 0.0 L Φ

Δ U1 U2 U3 U3R −0.5 0.25 0.25 0.25 0.25 1 2 3 3R LM0.25 LM0.25 LM0.25 LM0.25 1 2 3 3R LT0.25 LT0.25 LT0.25 LT0.25 −1.0

0.6

0.4 ]

V 0.2 e [

B 0.0 E N

E 1 2 3 3R −0.2 U0.50 U0.50 U0.50 U0.50 Δ LM1 LM2 LM3 LM3R −0.4 0.50 0.50 0.50 0.50 1 2 3 3R LT0.50 LT0.50 LT0.50 LT0.50 −0.6

1.0

0.5 ] V [

i 0.0 L Φ

Δ U1 U2 U3 3R −0.5 0.50 0.50 0.50 U0.50 1 2 3 3R LM0.50 LM0.50 LM0.50 LM0.50 1 2 3 3R −1.0 LT0.50 LT0.50 LT0.50 LT0.50

FIG. 4. cNEB energy profiles and corresponding electrostatic potential profiles for unstrained and isotropically strained struc- p tures (xLi = 0.25 (top) and xLi = 0.5 (bottom)). Notation is as follows: Sx, where S is strain (U = unstrained, LM = strained to lithium manganate lattice parameters, LT = strained to lithium titanate lattice parameters), x is the dopant concentration and p enumerates the pathways (where R refers to the same pathway in reverse). The inset schematic in each panel shows the coordination environments of the initial and final A-site positions. Li ions are green, Mg are orange and Al are blue. Electrostatic potential profiles (including the contribution from the mobile Li+ ion) at the Li+ site along the diffusion pathway are shown by the corresponding coloured dashed lines. ions experiencing stronger short-ranged repulsion at the Isotropically-strained (Li,Al)–co-doped spinel tetrahedral A sites than at the intermediate octahedral sites. This increases the energies of lithium at each end- 3 We now consider the effect of isotropic strain on point relative to the NEB path maximum. For the U0.5 3R lithium diffusion. We have calculated NEB barriers for (and U0.5) paths, there is quantitative disagreement between the electrostatic potential profile and the NEB po- both dopant concentrations with the unit cell strained to tential profile. The electrostatic potential predicts the the LiyMn2O4 and Li4+3zTi5O12 lattice volumes (Figs. 4 incorrect energy ordering of the two end-points. This and 5). Isotropic strain has a large impact on the dif- illustrates that the electrostatic potential alone is not fusion barriers. For the symmetric pathways the effect guaranteed to give a good description of the potential of this strain is straightforward: increased tensile strain energy surface for mobile lithium ions. decreases the diffusion barriers at both dopant concentrations. This result mirrors the effects seen in many 7 solid oxide fuel cell electrolytes [22–24, 26], where tensile tions on isotropically strained systems to consider the strain typically increases oxide ion diffusion. effect of anisotropic strain on the CI-NEB diffusion path- For the asymmetric pathway the situation is more com- ways. For these calculations, we still consider strain plicated. Because asymmetric paths necessarily have in- as arising from coherent {100}A|{100}B heterointerfaces. equivalent endpoints, the applied strain can affect their Under anisotropic strain, the three {100} directions be- relative energies. The change in the diffusion barrier come inequivalent, and we consider each possible strain height therefore depends on the direction of motion along orientation. For each calculation, the two in-plane lattice the path. At xLi = 0.25 the barrier is reduced for the for- parameters are strained to match experimental values of ward pathway (3), but slightly increases for the reverse either LiyMn2O4 or Li4+3zTi5O12, and the third lattice path (3R). At xLi = 0.5 this asymmetry is even more parameter is optimised to minimise the total system en- pronounced, with a large barrier increase for the forward ergy. Because we consider only {100} oriented interfaces, path (3), but a large decrease for the reverse path (3R). this is equivalent to biaxial strain. Fig. 6 plots the relative change in volume for the interme- The effect of anisotropic strain on diffusion barrier diate octahedral site, ∆Voct, against the change in bar- height is much smaller than in the isotropic case, with rier height, ∆ENEB. For both xLi = 0.25 and xLi = 0.50, changes from −0.182 eV to +0.074 eV. Individual val- the average trend of decreasing the NEB barrier height ues depend on the exact path being considered, and on is apparent. The outliers (showing both increased and the relative orientation of the applied strain. Individ- strongly decreased barrier heights) correspond to the two ual NEB profiles are shown in Fig. 8 (xLi = 0.25) and asymmetric paths described above. Fig. 9 (xLi = 0.50), and the full set of barrier heights The effect of strain on the NEB profiles is mirrored are collected in Fig. 5. The NEB pathway again depends by the electrostatic potential profiles, suggesting that on the symmetry of the start- and end-point A cation changes in electrostatic potential do provide at least a sites. In some cases, however, the application of planar qualitative metric for predicting the effects of strain on strain breaks the symmetry found in the isotropic and lithium ion diffusion. Fig. 7 plots the NEB barrier against unstrained structures. This point is illustrated by the A 1 A 1 the electrostatic barrier for each path. Linear–least- LM0.25 and LT0.25 pathways (Fig. 8) where the energy squares fits for each strain protocol show that the elec- difference between the start- and end-points depends on trostatic potential barrier becomes an increasingly good the orientation of the applied strain. predictor of the NEB barrier at larger strains. This is For the symmetric isotropically strained pathways, dis- consistent with short-ranged repulsion interactions being cussed above, the effect on the electrostatic potential un- less significant as the lattice volume increases under ten- der strain gave a good indication of the effect on the sile strain. NEB diffusion barrier. This simple electrostatic model is The effect of isotropic strain on the asymmetric paths not applicable, however, for the antisymmetric pathways. can also be examined from the perspective of the σA Under anisotropic strain, we find that the effect on the values for each end-point. For both asymmetric paths, electrostatic potential is no longer a useful predictor of 3 3 S0.25 and S0.50, tensile strain produces a relative stabili- the change in NEB barrier: in some cases the electro- sation of the end-point structure that more strongly de- static potential barrier decreases while the NEB barrier viates from the average +II oxidation state of the parent increases. The electrostatic potential profiles also show MgAl2O4 spinel, i.e. the structure with the larger σA. At the effect of the applied strain lowering the cell symmetry, xLi = 0.5, this effect is large enough that the relative sta- with relative potentials of path end-points now varying bilities of the two end-points are reversed, contradicting with the orientation of the applied strain. For isotropic the general trend that higher σ values correspond to less strain, we observed a direct relationship between the vol- stable structures. ume of the transitional octahedral 16c site, and the electrostatic potential at the diffusing Li+ site. For these anisotropically strained calculations, we find that the oc- Anisotropically-strained (Li,Al)–co-doped spinel tahedral site volume changes are much smaller than those due to the equivalent isotropic strain, because the cell is In the previous section, we have presented results allowed to contract along the free axis. As a result of showing the effect of isotropic strain on lithium-ion this reduced volume change, the effect on diffusion barri- diffusion barriers. For a hypothetical lattice-matched ers typically is much smaller than that of the isotropically electrolyte–electrode interface, however, lattice strain strained equivalents (Figs. 8, 9, & 10). due to electrolyte–electrode epitaxy is not isotropic, but instead is anisotropic. Specifically, the lattice will be strained parallel to the interface, but is free to relax in SUMMARY & DISCUSSION the perpendicular direction. To model the strain at a hypothetical lattice-matched (Li,Al)-codoped MgAl2O4 represents a new class of electrolyte–electrode interface, we extended our calcula- spinel-structured lithium-ion solid electrolytes, with po- 8

(a) isotropic, xLi = 0.25 (b) isotropic, xLi = 0.50 (c) anisotropic, xLi = 0.25 (d) anisotropic, xLi = 0.50

0.6 0.6 0.6 0.6

0.5 0.5 0.5 0.5

0.4 0.4 0.4 0.4 ] V e [ 0.3 0.3 0.3 0.3 B E N E 0.2 0.2 0.2 0.2

0.1 0.1 0.1 0.1

0.0 0.0 0.0 0.0 U LM LT U LM LT U LM LT U LM LT strain strain strain strain

FIG. 5. Variation in NEB barrier heights, ENEB, with strain. U = unstrained, LM = strained to LiyMn2O4 lattice parameters, LT = strained to Li4+3zTi5O12 lattice parameters. Panels (a) and (b) show data for isotropic strain, at xLi = 0.25 and xLi = 0.50, respectively. Panels (c) and (d) show data for anisotropic strain, at xLi = 0.25 and xLi = 0.50, respectively. Closed circles correspond to data for symmetric paths, and open circles to data for asymmetric paths. tential applications in an all-spinel solid-state lithium- of potential energy barriers for all paths, particularly in ion battery [6], or as an electrode–buffer-layer [58], with cases where paths are asymmetric. As the lattice volume lattice-matched electrolyte–electrode interfaces. Any increases under tensile strain, however, we find that the lattice-matched (or epitaxial) interface between an elec- electrostatic potential profile gives an increasingly good trode and electrolyte will exhibit some residual intrin- approximation of the diffusion potential energy profile. sic strain. In the case of (Li,Al)-codoped MgAl2O4, For a realistic lattice-matched interface, the resulting hypothetical lattice-matched (100) interfaces with the strain will be anisotropic (epitaxial). From our CI-NEB spinel electrodes LiyMn2O4 or Li4+3zTi5O12 require ten- calculations of anisotropically strained cells, in general sile strains of up to 3.2 or 6.0 % respectively, at xLi = we predict a much smaller change in diffusion barriers 0.5. Because the lattice parameter of (Li,Al)-codoped than for equivalent isotropic strain. Individual changes, MgAl2O4 decreases with higher doping levels, It has however, show a complex dependency on the particular previously been suggested that high dopant concentra- path being considered, and the orientation of the dif- tions (xLi > 0.3) should be avoided, to minimise possi- fusion pathway relative to the applied strain, and some ble corresponding increases in interfacial resistance [38]. barriers increase, by up to +0.074 eV at 6.0% strain. The High dopant concentrations, however, have been pre- smaller effect on diffusion barriers under epitaxial strain dicted to have higher lithium conductivities, making compared to equivalent isotropic strain is attributed to them more appealing as solid electrolytes. These issues the capacity of a system under epitaxial strain to re- have prompted us to conduct this computational study lax in the perpendicular direction. This difference is can into the effects of tensile strain on the lithium-ion diffu- be seen in the relative volume changes of the intermedi- sion pathways in (Li,Al)-doped MgAl2O4, to assess the ate octahedral sites. Under anisotropic strain the transi- extent to which the increased strain expected at higher tional octahedral sites undergo a smaller volume expan- dopant concentrations affects lithium-ion transport. sion than under equivalent isotropic strain. We have performed a series of climbing-image nudged– Because the distribution of barrier heights is predicted elastic-band (CI-NEB) calculations to evaluate the po- to not change significantly under anisotropic strain, par- tential energy profile for lithium vacancy diffusion, which ticularly at xLi = 0.5, we postulate that epitaxial strain proceeds via tetrahedron–octahedron–tetrahedron paths arising from lattice-matching will not significantly af- through the spinel structure. Our calculations show that fect lithium diffusion in the region close to an electrode– isotropic strain, for most paths, decreases diffusion bar- electrolyte interface. The effect of epitaxial strain on in- riers; with an average reduction of ∼ 0.1 eV for a lattice dividual barriers is complex, however, causing both small strained to match Li4+3zTi5O12 (6% strain). This change decreases and small increases in barrier heights. A precise is correlated with changes in the electrostatic potential quantitative prediction of the effect of strain on ensem- profile along each diffusion path. The electrostatic po- ble transport properties (diffusion coefficients and ionic tential profile, however, does not give a good prediction conductivities), would therefore require statistically sam- 9

(a) xLi = 0.25 (a) xLi = 0.25

1.0 0.10

0.8 0.00 ] ] V V 0.6 e e [

-0.10 [

B B E E N N E E 0.4 Δ -0.20 Δ

-0.30 0.2

-0.40 0.0

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.2 0.4 0.6 0.8 1.0 3 ΔVoct [Å ] ΔΦNEB [V]

(b) xLi = 0.50 (b) xLi = 0.50

1.0 0.10

0.8 0.00 ] ] V V 0.6 e e [

-0.10 [

B B E E N N E E 0.4 Δ -0.20 Δ

-0.30 0.2

-0.40 0.0

0.0 0.5 1.0 1.5 2.0 2.5 3.0 0.0 0.2 0.4 0.6 0.8 1.0 3 ΔVoct [Å ] ΔΦNEB [V]

FIG. 6. Change in NEB barrier height, ∆ENEB, as a func- FIG. 7. Electrostatic potential barrier, ∆ΦNEB, versus tion of volume change for the 16c octahedra, ∆Voct, under potential energy barrier, ∆ENEB, for the diffusing lithium isotropic strain, for (a) xLi = 0.25 (b) xLi = 0.5. Red ion, for pathways under zero-strain (black), and isotropically and orange circles correspond to strain to LiyMn2O4 and strained to the LiyMn2O4 (red) and Li4+3zTi5O12 (orange) Li4+3zTi5O12 lattice parameters, respectively. Dashed lines cell volumes. (a) xLi = 0.25. (b) xLi = 0.50. The solid lines link equivalent pathways under different strains. The solid show linear best fits at each strain (U, LM, LT). The diagonal black points show mean values under each strain condition. dashed line corresponds to an exact 1:1 relationship between Closed circles correspond to data for symmetric paths, and ∆ΦNEB and ∆ENEB. Closed circles correspond to data for open circles to data for asymmetric paths. symmetric paths, and open circles to data for asymmetric paths. pling all relevant paths, using, for example, molecular dy- namics or kinetic Monte Carlo simulations (cf. ref. [38]). ther redistribution of ions can occur, producing more Changes to local diffusion barriers are not the only complex space-charge profiles [60]. The change in local possible mechanism by which interfacial strain might af- lithium-ion concentration associated with space-charge fect interfacial resistance. At a heterointerface between formation can affect the local conductivity (and resistiv- an electrode and electrolyte, the standard chemical po- ity) either positively or negatively, and hence contributes tential of lithium ions may differ between the two ma- to interfacial resistance. Because strain shifts the lo- terials. This can drive a spontaneous redistribution of cal electrostatic potential, interfacial strain changes the the mobile ions across the interface to form space-charge standard electrochemical potential offset at an electrode– layers on each side [59]. Under an applied voltage, fur- electrolyte interface. This consequently affects space- 10

0.6

0.4 ] V

e 0.2 [

B E

N 0.0 E Δ A 1 A 2 A 3 A 3R −0.2 LM0.25 LM0.25 LM0.25 LM0.25 1 2 3 3R U0.25 U0.25 U0.25 U0.25 −0.4

1.0

] 0.5 V [

i L

Φ 0.0 Δ

ALM1 ALM2 ALM3 ALM3R −0.5 0.25 0.25 0.25 0.25 1 2 3 3R U0.25 U0.25 U0.25 U0.25

0.6

0.4 ] V

e 0.2 [

B E

N 0.0 E Δ A 1 A 2 A 3 A 3R −0.2 LT0.25 LT0.25 LT0.25 LT0.25 1 2 3 3R U0.25 U0.25 U0.25 U0.25 −0.4

1.0

] 0.5 V [

i L

Φ 0.0 Δ

ALT1 ALT2 ALT3 ALT3R −0.5 0.25 0.25 0.25 0.25 1 2 3 3R U0.25 U0.25 U0.25 U0.25

FIG. 8. xLi = 0.25 cNEB energy profiles and corresponding electrostatic potential profiles for unstrained and anisotropically p strained structures. Notation is as follows: Sx, where S is strain (U = unstrained, LM = strained to lithium manganate lattice parameters (top two rows), LT = strained to lithium titanate lattice parameters (bottom two rows)), x is the dopant concentration and p enumerates the pathways (where R refers to the same pathway in reverse). A superscript A indicates anisotropic strain along only two axes. The inset schematic in each panel shows the coordination environments of the initial and final A-site positions. Li ions are green, Mg are orange and Al are blue. Electrostatic potential profiles (including the contribution from the mobile Li+ ion) at the Li+ site along the diffusion pathway are shown by the corresponding coloured dashed lines. charge formation, and the associated space-charge con- strain are small, even for relatively large in-plane strains. tribution to interfacial resistance. Our study also implic- Extrapolating from this result, we posit that lithium itly assumes that any relevant epitaxial heterointerface transport close to lattice-matched electrolyte–electrode between electrode and electrolyte can be formed with- interfaces is not significantly affected by the residual lat- out dislocations or other extended defects. A complete tice strain. Accurately resolving the transport behaviour model of lattice-matched electrode–electrolyte interfaces for a specific electrode–electrolyte pair, however, requires should, therefore, also consider these issues. Assuming a going beyond the local diffusion barrier modelling de- coherent lattice-matched interface, however, and neglect- scribed here, and explicitly calculating lithium transport ing shifts in electrostatic potentials, our results suggest coefficients under the appropriate strain. that changes in local diffusion barriers due to epitaxial 11

0.6

0.4 ] V

e 0.2 [

B E

N 0.0 E Δ A 1 A 2 A 3 A 3R −0.2 LM0.50 LM0.50 LM0.50 LM0.50 1 2 3 3R U0.50 U0.50 U0.50 U0.50 −0.4

1.0 ]

V 0.5 [

i L

Φ 0.0 Δ

A 1 A 2 A 3 A 3R −0.5 LM0.50 LM0.50 LM0.50 LM0.50 1 2 3 3R U0.50 U0.50 U0.50 U0.50

0.6

0.4 ] V

e 0.2 [

B E

N 0.0 E Δ A 1 A 2 A 3 A 3R −0.2 LT0.50 LT0.50 LT0.50 LT0.50 1 2 3 3R U0.50 U0.50 U0.50 U0.50 −0.4

1.0 ]

V 0.5 [

i L

Φ 0.0 Δ

A 1 A 2 A 3 A 3R −0.5 LT0.50 LT0.50 LT0.50 LT0.50 1 2 3 3R U0.50 U0.50 U0.50 U0.50

FIG. 9. xLi = 0.50 cNEB energy profiles and corresponding electrostatic potential profiles for unstrained and anisotropically p strained structures. Notation is as follows: Sx, where S is strain (U = unstrained, LM = strained to lithium manganate lattice parameters (top two rows), LT = strained to lithium titanate lattice parameters (bottom two rows)), x is the dopant concentration and p enumerates the pathways (where R refers to the same pathway in reverse). A superscript A indicates anisotropic strain along only two axes. The inset schematic in each panel shows the coordination environments of the initial and final A-site positions. Li ions are green, Mg are orange and Al are blue. Electrostatic potential profiles (including the contribution from the mobile Li+ ion) at the Li+ site along the diffusion pathway are shown by the corresponding coloured dashed lines.

DATA ACCESS STATEMENT 10.5281/zenodo.1069417), published under the MIT license. The DFT dataset supporting this study is available from the University of Bath Research Data Archive (doi: 10.15125/BATH-00438) [53], published under the CC- ACKNOWLEDGEMENTS BY-SA-4.0 license. This dataset contains input parameters and output ﬁles for all VASP calculations, and a This work was funded by EPSRC grant series of Python scripts for collating relevant data from EP/N004302/1. B. J. M. acknowledges support from the the VASP outputs. A Jupyter notebook containing code Royal Society (UF130329). Calculations were performed to produce Figs. 4–10 is also available (Ref 54, doi: using the Balena High Performance Computing Service 12

(a) xLi = 0.25

0.10 ∗ c.o’[email protected] † 0.05 [email protected] [1] Philippe Knauth, “Inorganic solid Li ion conductors: An overview,” Sol. Stat. Ionics 180, 911–916 (2009). 0.00

] [2] John Christopher Bachman, Sokseiha Muy, Alexis Gri- V

e maud, Hao-Hsun Chang, Nir Pour, Simon F. Lux, [

B -0.05 Odysseas Paschos, Filippo Maglia, Saskia Lupart, Peter E N

E Lamp, Livia Giordano, and Yang Shao-Horn, “Inorganic Δ solid-state electrolytes for lithium batteries: Mechanisms -0.10 and properties governing ion conduction,” Chem. Rev. 116, 140–162 (2016). -0.15 [3] Arumugam Manthiram, Xingwen Yu, and Shaofei Wang, “Lithium battery chemistries enabled by solid-state elec- -0.20 trolytes,” Nat. Rev. Mater. 2, 1–16 (2017). 0.0 0.2 0.4 0.6 0.8 1.0 1.2 [4] Noriaki Kamaya, Kenji Homma, Yuichiro Yamakawa, ΔV [Å3] Masaaki Hirayama, Ryoji Kanno, Masao Yonemura, oct Takashi Kamiyama, Yuki Kato, Shigenori Hama, Koji Kawamoto, and Akio Mitsui, “A lithium superionic con- (b) xLi = 0.50 ductor,” Nat. Mater. 10, 682–686 (2011). 0.10 [5] Dhamodaran Santhanagopalan, Danna Qian, Thomas McGilvray, Ziying Wang, Feng Wang, Fernando Camino, 0.05 Jason Graetz, Nancy Dudney, and Ying Shirley Meng, “Interface limited lithium transport in solid-state batteries,” J. Phys. Chem. Lett. 5, 298–303 (2014). 0.00

] [6] Fabio Rosciano, Paolo P. Pescarmona, Kristof V

e Houthoofd, Andre Persoons, Patrick Bottke, and [

B -0.05 Martin Wilkening, “Towards a lattice-matching solid- E N state battery: synthesis of a new class of lithium-ion E

Δ conductors with the spinel structure,” Phys. Chem. -0.10 Chem. Phys. 15, 6107–6112 (2013). [7] M. M. Thackeray and J. B. Goodenough, “Solid state -0.15 cell wherein an anode, solid electrolyte and cathode each comprise a cubic-close-packed framework structure,” -0.20 (1985), US Patent 4,507,371. 0.0 0.2 0.4 0.6 0.8 1.0 1.2 [8] E. S. (Merijn) Blaakmeer, Fabio Rosciano, and Ernst 3 R. H. van Eck, “Lithium doping of MgAl2O4 and ΔVoct [Å ] ZnAl2O4 investigated by high-resolution solid state NMR,” J. Phys. Chem. C 119, 7565–7577 (2015). FIG. 10. Change in NEB barrier height, ∆ENEB, as a func- [9] Ruzica Djenadic, Miriam Botros, and Horst Hahn, “Is tion of volume change for the 16c octahedra, ∆Voct, under Li-doped MgAl2O4 a potential solid electrolyte for an anisotropic strain, for (a) xLi = 0.25 (b) xLi = 0.5. Red and all-spinel Li-ion battery?” Sol. Stat. Ionics 287, 71–76 orange circles show strain to LiyMn2O4 and Li4+3zTi5O12 (2016). lattice parameters, respectively. Dashed lines link equivalent [10] Ting-Feng Yi, Yan-Rong Zhu, Xiao-Dong Zhu, J. Shu, pathways under diﬀerent strains. The solid black points show Cai-Bo Yue, and An-Na Zhou, “A review of recent de- mean values under each strain condition. Closed circles cor- velopments in the surface modiﬁcation of LiMn2O4 as respond to data for symmetric paths, and open circles to data cathode material of power lithium-ion battery,” Ionics for asymmetric paths. 15, 779–784 (2009). [11] Zonghai Chen, Yan Qin, Khalil Amine, and Y. K. Sun, “Role of surface coating on cathode materials for lithium- ion batteries,” J. Mater. Chem. 20, 7606–7 (2010). [12] Muratahan Aykol, Soo Kim, Vinay I. Hegde, David Sny- dacker, Zhi Lu, Shiqiang Hao, Scott Kirklin, Dane Mor- gan, and C. Wolverton, “High-throughput computational design of cathode coatings for Li-ion batteries,” Nature Comm. 7, 1–12 (2016). [13] Daxian Zuo, Guanglei Tian, Xiang Li, Da Chen, and Kangying Shu, “Recent progress in surface coating of at the University of Bath, and using the ARCHER cathode materials for lithium ion secondary batteries,” supercomputer, with access through membership of the J. All. Com. 706, 24–40 (2017). UK’s HPC Materials Chemistry Consortium, funded by [14] N. Ohta, K. Takada, L. Zhang, R. Ma, M. Osada, and EPSRC grant EP/L000202. T. Sasaki, “Enhancement of the high-rate capability of 13

solid-state lithium batteries by nanoscale interfacial mod- [30] Fanghua Ning, Shuai Li, Bo Xu, and Chuying Ouyang, ification,” Adv. Mater. 18, 2226–2229 (2006). “Strain tuned Li diffusion in LiCoO2 material for Li ion [15] Narumi Ohta, Kazunori Takada, Isao Sakaguchi, Lianqi batteries: A first principles study,” Sol. Stat. Ionics 263, Zhang, Renzhi Ma, Katsutoshi Fukuda, Minoru Osada, 46–48 (2014). and Takayoshi Sasaki, “LiNbO3-coated LiCoO2 as cath- [31] Jie Wei, Daisuke Ogawa, Tomoteru Fukumura, Ya- ode material for all solid-state lithium secondary batter- sushi Hirose, and Tetsuya Hasegawa, “Epitaxial strain- ies,” Electrochem Comm. 9, 1486–1490 (2007). controlled ionic conductivity in Li-ion solid electrolyte [16] Kazunori Takada and Takahisa Ohno, “Experimental Li0.33La0.56TiO3 thin films,” Crys. Growth & Design 15, and computational approaches to interfacial resistance 2187–2191 (2015). in solid-state batteries,” Front. Energy Res. 4, 014101–7 [32] Cristina Tealdi, Jennifer Heath, and M. Saiful Islam, (2016). “Feeling the strain: enhancing ionic transport in olivine [17] Jili Li, Youqi Zhu, Lin Wang, and Chuanbao Cao, phosphate cathodes for Li- and Na-ion batteries through “Lithium titanate epitaxial coating on spinel lithium strain effects,” J. Mater. Chem. A 4, 6998–7004 (2016). manganese oxide surface for improving the performance [33] Ashkan Moradabadi and Payam Kaghazchi, “Effect of of lithium storage capability,” ACS Appl. Mater. Int. 6, strain on polaron hopping and electronic conductivity in 18742–18750 (2014). bulk LiCoO2,” Phys. Rev. Appl. 7, 064008–5 (2017). [18] H. Berg and J. O. Thomas, “Neutron diffraction study of [34] Mingzhen Jia, Hongyan Wang, Zhandong Sun, electrochemically delithiated LiMn2O4 spinel,” Sol. Stat. Yuanzheng Chen, Chunsheng Guo, and Liyong Ionics 126, 227 – 234 (1999). Gan, “Exploring ion migration in Li2MnSiO4 for [19] M. Wagemaker, D.R. Simon, E.M. Kelder, J. Schoon- Li-ion batteries through strain effects,” RSC Adv. 7, man, C. Ringpfeil, U. Haake, D. Lützenkirchen-Hecht, 26089–26096 (2017). R. Frahm, and F.M. Mulder, “A kinetic two-phase and [35] Nitin Muralidharan, Casey N. Brock, Adam P. Cohn, equilibrium solid solution in spinel Li4+xTi5O12,” Adv. Deanna Schauben, Rachel E. Carter, Landon Oakes, Mater. 18, 3169–3173 (2006). D. Greg Walker, and Cary L. Pint, “Tunable [20] N. Schichtel, C. Korte, D. Hesse, and J. Janek, “Elastic mechanochemistry of lithium battery electrodes,” ACS strain at interfaces and its influence on ionic conductiv- Nano 11, 6243–6251 (2017). ity in nanoscaled solid electrolyte thin films—theoretical [36] Ashkan Moradabadi, Payam Kaghazchi, Jochen Rohrer, considerations and experimental studies,” Phys. Chem. and Karsten Albe, “Influence of elastic strain on the Chem. Phys. 11, 3043 (2009). thermodynamics and kinetics of lithium vacancy in bulk [21] Jennifer L. M. Rupp, “Ionic diffusion as a matter of LiCoO2,” arXiv (2017), 1706.01709v1. lattice-strain for electroceramic thin films,” Sol. Stat. [37] Arun K. Sagotra and Claudio Cazorla, “Stress-mediated Ionics 207, 1–13 (2012). enhancement of ionic conductivity in fast-ion conduc- [22] Kechun Wen, Weiqiang Lv, and Weidong He, “Interfa- tors,” ACS Appl. Mater. Int. 9, 38773–38783 (2017). cial lattice-strain effects on improving the overall perfor- [38] Maarten J. Mees, Geoffrey Pourtois, Fabio Rosciano, mance of micro-solid oxide fuel cells,” J. Mater. Chem. Brecht Put, Philippe M. Vereecken, and AndréStes- A 3, 20031–20050 (2015). mans, “First-principles material modeling of solid-state [23] Bilge Yildiz, ““Stretching” the energy landscape of electrolytes with the spinel structure,” Phys. Chem. oxides—Effects on electrocatalysis and diffusion,” MRS Chem. Phys. 16, 5399–5406 (2014). Bull. 39, 147–156 (2014). [39] Graeme Henkelman, Blas P. Uberuaga, and Hannes [24] Halit Aydin, Carsten Korte, Marcus Rohnke, and Jónsson,“A climbing image nudged elastic band method JürgenJanek, “Oxygen tracer diffusion along interfaces for finding saddle points and minimum energy paths,” J. of strained Y2O3/YSZ multilayers,” Phys. Chem. Chem. Chem. Phys. 113, 9901–9904 (2000). Phys. 15, 1944–1955 (2013). [40] B. S. Allimi, M. Aindow, and S. P. Alpay, “Thickness [25] Weida Shen, Jun Jiang, and Joshua L. Hertz, “Reduced dependence of electronic phase transitions in epitaxial ionic conductivity in biaxially compressed ceria,” RSC V2O3 films on (0001) LiTaO3,” Appl. Phys. Lett. 93, Adv. 4, 21625–21630 (2014). 112109 (2008). [26] Aline Fluri, Daniele Pergolesi, Vladimir Roddatis, [41] G. Kresse and J. Furthmüller, “Efficient iterative schemes Alexander Wokaun, and Thomas Lippert, “In situ stress for ab initio total-energy calculations using a plane-wave observation in oxide films and how tensile stress influ- basis set,” Phys. Rev. B 54, 11169–11186 (1996). ences oxygen ion conduction,” Nat. Comm. 7 (2016), [42] G. Kresse and J. Furthmüller,“Efficiency of ab-initio to- 10.1038/ncomms10692. tal energy calculations for metals and semiconductors us- [27] Chiara Ferrara, Christopher Eames, M. Saiful Islam, and ing a plane-wave basis set,” Comp. Mater. Sci. 6, 15–50 Cristina Tealdi, “Lattice strain effects on doping, hydra- (1996). tion and proton transport in scheelite-type electrolytes [43] John P. Perdew, Adrienn Ruzsinszky, Gábor I. Csonka, for solid oxide fuel cells,” Phys. Chem. Chem. Phys. 18, Oleg A. Vydrov, Gustavo E. Scuseria, Lucian A. Con- 29330–29336 (2016). stantin, Xiaolan Zhou, and Kieron Burke, “Restoring [28] Hui-Jun Yan, Zhi-Qiang Wang, Bo Xu, and Chuying the density-gradient expansion for exchange in solids and Ouyang, “Strain induced enhanced migration of polaron surfaces,” Phys. Rev. Lett. 100, 136406 (2008). and lithium ion in λ-MnO2,” Funct. Mater. Lett. 05, [44] P. E. Blöchl, “Projector augmented-wave method,” Phys. 1250037–4 (2012). Rev. B 50, 17953–17979 (1994). [29] Jaekwang Lee, Stephen J. Pennycook, and Sokrates T. [45] Kurt E. Sickafus, John M. Wills, and Norman W. Pantelides, “Simultaneous enhancement of electronic and Grimes, “Structure of spinel,” J. Am. Ceram. Soc. 82, + Li ion conductivity in LiFePO4,” Appl. Phys. Lett. 101, 3279–3292 (1999). 033901–5 (2012). 14

[46] Roderick J. Hill, James R. Craig, and G. V. Gibbs, https://doi.org/10.15125/BATH-00438 (2017). “Systematics of the spinel structure type,” Phys. Chem. [54] Conn O’Rourke and Benjamin J. Morgan, “Data analysis Mater. 4, 317–339 (1979). for “Interfacial strain effects on lithium diffusion path- [47] Benjamin J. Morgan, “bsym: A basic symmetry mod- ways in the spinel solid electrolyte Li-doped MgAl2O4”,” ule,” The Journal of Open Source Software 2 (2017), https://github.com/bjmorgan/data NEB spinel (2017). 10.21105/joss.00370. [55] J. D. Hunter, “Matplotlib: A 2D graphics environment,” [48] F. Ronci, P. Reale, B. Scrosati, S. Panero, V. Rossi Al- Computing In Science & Engineering 9, 90–95 (2007). bertini, P. Perfetti, M. di Michiel, and J. M. Merino, [56] Thomas Kluyver, Benjamin Ragan-Kelley, Fernando “High-resolution in-situ structural measurements of the Pérez,Brian Granger, Matthias Bussonnier, Jonathan Li4/3Ti5/3O4“zero-strain” insertion material,” J. Phys. Frederic, Kyle Kelley, Jessica Hamrick, Jason Grout, Syl- Chem. B 106, 3082–3086 (2002). vain Corlay, Paul Ivanov, DamiánAvila, Safia Abdalla, [49] Benjamin J. Morgan, Javier Carrasco, and Gilberto and Carol Willing, “Jupyter notebooks – a publishing Teobaldi, “Variation in surface energy and reduction format for reproducible computational workflows,” in drive of a metal oxide lithium-ion anode with stoichiom- Positioning and Power in Academic Publishing: Play- etry: a DFT study of lithium titanate spinel surfaces,” ers, Agents and Agendas, edited by F. Loizides and J. Mater. Chem. A 4, 17180–17192 (2016). B. Schmidt (IOS Press, 2016) pp. 87 – 90. [50] Benjamin J. Morgan and Graeme W. Watson, “GGA+U [57] Koichi Momma and Fujio Izumi, “VESTA3 for three- description of lithium intercalation into anatase TiO2,” dimensional visualization of crystal, volumetric and mor- Phys. Rev. B 82, 144119 (2010). phology data,” J. Appl. Cryst. 44, 1272–1276 (2011). [51] David A. Tompsett, Steve C. Parker, Peter G. Bruce, [58] Brecht Put, Philippe M. Vereecken, Maarten J. Mees, and M. Saiful Islam, “Nanostructuring of β-MnO2: The Fabio Rosciano, Iuliana P. Radu, and Andre Stes- important role of surface to bulk ion migration,” Chem. mans, “Characterization of thin films of the solid elec- Mater. 25, 536–541 (2013). trolyte LixMg1−2xAl2+xo4 (x = 0, 0.05, 0.15, 0.25),” [52] Ziqin Rong, Daniil Kitchaev, Pieremanuele Canepa, Phys. Chem. Chem. Phys. 17, 29045–29056 (2015). Wenxuan Huang, and Gerbrand Ceder, “An efficient [59] Benjamin J. Morgan and Paul A. Madden, “Frenkel algorithm for finding the minimum energy path for polarisation of coherent interfaces in fluorite het- cation migration in ionic materials,” J. Chem. Phys. 145, erostructures,” Phys. Rev. B 89 (2014), 10.1103/Phys- 074112–9 (2016). RevB.89.054304. [53] Conn O’Rourke and Benjamin J. Morgan, “DFT dataset [60] Kevin Leung and Andrew Leenheer, “How voltage drops for “Interfacial strain effects on lithium diffusion path- are manifested by lithium ion configurations at interfaces ways in the spinel solid electrolyte Li-doped MgAl2O4”,” and in thin films on battery electrodes,” J. Phys. Chem. C 119, 10234–10246 (2015).