Strongly correlated perovskite lithium ion shuttles

Suppression of electronic transport in SNO upon lithiation can be seen in the evolution of film resistivity (Fig. 1D). The pristine SNO shows low resistivity of around 0.6 mΩ⋅cm at room temperature. A thermal insulator–metal transition point is observed at ∼140 °C, indicating good film quality. Lithiation for 1 min enhances the film resistivity by nearly three orders of magnitude to ∼10³ mΩ⋅cm, along with a visual color change. As lithiation proceeds, the film becomes much more transparent and the resistivity continues to increase to ∼10⁸ mΩ⋅cm and then nearly saturates, indicating significant suppression of electronic conduction (denoted as Li-SNO). This saturated low electronic conductivity indicates the Mott state. Direct real-space nanoimaging at terahertz frequencies also highlights this carrier suppression effect. (SI Appendix, Fig. S2).

Fig. 1E shows the Arrhenius conductivity plots of Li-SNO and other representative oxide Li⁺ conductors. Li-SNO has room temperature ionic conductivity of 3.1 × 10⁻³ S⋅cm⁻¹ with low activation barrier of ∼0.24 eV, comparable to even organic liquid systems. The cycling performance of the lithiation–delithiation process (Fig. 1F and SI Appendix, Fig. S3A) demonstrates an average coulombic efficiency of 96.8%. Upon cycling, the Li-SNO maintained transparency throughout the experiments, indicating that under galvanostatic lithiation, of up to 4.2 V, the material remains stable. These studies demonstrate the shuttling of Li ions across the perovskite lattice under electric fields.

Collective phenomena due to strong electron–electron correlations leads to these striking observations in the perovskite nickelate. In Li-SNO, the transport bottleneck from the flexible tilting of the NiO₆ octahedra controls the interstitial Li⁺ diffusion across adjacent hopping sites. The larger Ni²⁺ radius combined with the electron localization effect synergistically increase the Ni–O bond length and result in volume expansion of the perovskite unit cell, which significantly reduces the Li⁺ transport barrier.

To show that lithiation does indeed lead to reduction of the Ni sites, we measured X-ray photoelectron spectroscopy (XPS) spectra of Ni (2p_3/2) peak and Li (1s) peak (SI Appendix, Fig. S4). The Ni peak shifts to lower binding energy after lithiation, demonstrating the increase in the fraction of Ni(II) and decrease in the fraction of Ni(III) in Li-SNO. As the lithiation proceeded, the Li_1s peak intensity gradually increased, indicating increasing content of Li⁺ (Li⁺:Ni ∼ 0.86 for fully doped Li-SNO).

To further illustrate the detailed change in the electronic structure upon lithiation, we characterize the films with X-ray absorption near-edge spectroscopy (XANES). In the normalized Ni K-edge XANES (Fig. 2A), the preedge (feature A) points to the covalence between O_2p and Ni_3d orbital. The rest of absorption features essentially stem from two different coordination shells, for example, the first shell (in features B, D, E) is formed by octahedral oxygen coordination while the second shell (features C and C′) is due to rare-earth atoms in a pseudocubic coordination (29). The intensity of feature A (Fig. 2B) significantly decreased after lithiation, indicating the Ni valence decrease. Ni L₃-edge spectra of SNO also shows drastic changes after lithiation with diminished peak at ∼855 eV, stressing the transition of Ni³⁺ to Ni²⁺ (Fig. 2C) (30). A detailed comparison of preedge area (Ni K-edge) clearly shows that the lithiation process decreases the concentration of ligand hole (3d⁸L in the ground state of Ni, where L denotes a hole state generated in ligand p orbital) by 40% (29). Further study on the O K-edge spectra (Fig. 2D) indicates a reduced O-projected density of unoccupied states due to lithiation induced electron filling, suggesting the formation of Ni²⁺. The rest of the holes within O²⁻ could bond with Li⁺ and facilitate its transport. The first derivative plot (SI Appendix, Fig. S5 A and B) of the Ni K-edge XANES presents a shift of the preedge and absorption edge toward lower energy by 1.65 eV for the 2-h–lithiated sample. Previous studies have shown a slope shift of 1–2 eV/electron charge for Ni-containing oxides with octahedral coordination (31). An effective change in valence of −1 for the Ni ions is seen here. Such lithiation-induced phase transition phenomena can propagate through the film thickness (SI Appendix, Fig. S6).

The lithiation of SNO initially leads to the interstitial intercalation of Li⁺ and electron localization, which further results in structural distortion. To obtain quantitative information on the lattice parameter evolution of SNO upon lithiation, synchrotron X-ray diffraction (XRD) was performed and the corresponding pattern is shown in Fig. 3A. Pristine SNO has a pseudocubic lattice constant of 3.799 Å, which is quite close to that of the LaAlO₃ (LAO) substrate (3.790 Å) (32). As seen in curve I of Fig. 3A, the SNO exhibits a shoulder [(220) peak, orthorhombic notation] located at q_z = 3.29 Å⁻¹ close to LAO (002) peak (pseudocubic notation). Lithiation induces the appearance of peaks at q_z = 2.98 Å⁻¹ corresponding to ∼9.1% out-of-plane lattice dilatation (curve II in Fig. 3A). Noticeably, the XRD peak of SNO parent phase becomes narrower with Li intercalation, which may be attributed to electrochemical crystallization, which is sometimes referred to as “cold annealing” process (33, 34). Two-dimensional reciprocal space mapping (RSM) around the pseudocubic (002) reflections of pristine SNO and Li-SNO (Fig. 3 B and C) further confirm lattice expansion of ∼9% in Li-SNO. The lithiation induced thin film lattice expansion was further seen in synchrotron X-ray reflectivity (XRR) (Fig. 3D). Compared with the pristine state, a significantly decreased oscillation period of Li-SNO can be observed, due to the increase of film thickness. This significant lattice expansion due to NiO₆ octahedral distortion can then be correlated to the reduced diffusion activation energy (35).

It further can be seen in the first derivative plot of the Ni K-edge XANES spectra (SI Appendix, Fig. S5C) that the energy separation between A and B feature is ∼16.45 eV for pristine SNO, which is larger than that of Li-SNO (∼15.35 eV). The significant change in energy splitting suggests that lithiation enhances the lattice distortion and increases the Ni–O bond length. The fitted Fourier-transformed extended X-ray absorption fine structure (EXAFS) spectra (Fig. 3E and SI Appendix, Fig. S7) gives a local picture of the Ni coordination shells. The first peak (A) at ∼1.5 Å is from the Ni–O coordination shell, and the feature at 2.6 Å stems from the Ni–Sm shells. Lithiation not only leads to the decrease of the Ni–O peak intensity but also shifts the peak position toward larger radius, which is in part due to the larger Ni(II) ions in SNO (36). The fitting results demonstrate that the Ni–O bond length expands by 0.08 Å due to the Li doping (from 1.95 to 2.03 Å). Noticeably, a distinct feature around 2.2 Å was found in Li-SNO, which we ascribed to an additional Ni–Li shell between Ni–O and Ni–Sm, further strengthening our hypothesis that Li⁺ occupies interstitial sites. In situ synchrotron XRD experiments were performed to detect the structural evolution during cycling (Fig. 3A). The diffraction peak at q_z = 2.98 Å⁻¹ shifts to 3.03 Å⁻¹ (curve III in Fig. 3A). Multiple cycles make the SNO thin film disordered (weak Laue fringes between q_z = 2.8–3.1 Å⁻¹ in curves IV and V in Fig. 3A); the high electronic resistance of Li-SNO, however, is still maintained.

First-principles density functional theory (DFT) calculations are carried out on pristine SNO in a monoclinic cell with no breathing distortion, as such a distortion is not observed in room temperature SNO. As lithium is added to SNO, its valence electron localizes on a Ni site, thus converting it to Ni²⁺. The localization of the electron on the Ni³⁺ is clearly observed as the local magnetic moment increases from 0.8 μ_B (Ni³⁺) to 1.6 μ_B (Ni²⁺). The Li valence electron occupies a previously unoccupied Ni–O state and does not introduce a lithium state near the Fermi energy. In fact, once a Ni site is converted to Ni²⁺, the remaining unoccupied e_g states are pushed up in energy relative to the occupied e_g states due to electron–electron correlations, leading to a much larger band gap once all of the nickel are converted to Ni²⁺ (Fig. 4A).

Structural optimization calculations using DFT show that at a concentration of 1:4 Li⁺:Ni the lowest-energy intercalated Li⁺ ion is tetrahedrally coordinated by four oxygens from two adjacent corner-sharing NiO₆ octahedra. These octahedra are canted toward each other with an out-of-phase tilt angle between them such that the LiO₄ tetrahedron is face-sharing with one NiO₆ octahedron and edge-sharing with the other (Fig. 4B). For details, see SI Appendix, Fig. S8 J–M. The added Li valence electron causes an increase in the volume of the NiO₆ octahedron that contains a Ni²⁺ ion, which is not necessarily the closest Ni to the intercalated Li⁺, as the localized electron need not sit on one of the adjacent NiO₆ octahedra. At a concentration of 1:1 Li⁺:Ni, the lowest energy structure has intercalated Li⁺ tetrahedrally coordinated in the face–edge sharing configuration described above as well as Li⁺ between two octahedra canted toward each other with an in-phase tilt angle between them such that the LiO₄ tetrahedra is edge-sharing with both NiO₆ octahedra (SI Appendix, Fig. S8 G–I). Informed by the experimental characterization and the molecular dynamics calculations presented below, however, it seems that this structure is kinetically inaccessible. Restricting our attention to geometries where the Li⁺ remains in the face–edge sharing tetrahedra, the calculated overall volume expansion for 1 Li/Ni is ∼10%, when the cell is allowed to expand in the (110) direction matching the epitaxial condition of the experimental film, consistent with diffraction data (Fig. 4C). For the series of geometries shown in Fig. 4B, the average Ni–O bond length increases 0.07 Å (from 2.002 Å in SNO to 2.074 Å in Li-SNO), the coordination number decreases by around 1 (6 to 5), and the average Ni–Li bond length is ∼2.1 Å, in good agreement with the EXAFS results (SI Appendix, Table S1). The lattice expansion mechanism associated with donor doping from Li described here is similar to the lattice expansion caused by donor doping by oxygen vacancies in oxides (37). The larger the extent of localization of the electron on the redox active metal cation, the larger the expansion in the structure (38).

Ab initio molecular dynamics (AIMD) and DFT calculations were performed to understand Li migration pathways through the SNO lattice and the associated barriers. During a typical AIMD run at a modest temperature of 300 K, we observe frequent Li hopping from one tetrahedral site to a nearby one. Climbing-image nudged elastic band (CI-NEB) calculations within the framework of DFT+U for an isolated Li hop in an unstrained SNO lattice reveal an energetic barrier of ∼0.44 eV (Fig. 4D). The optimal diffusion pathway involves migration of Li between two tetrahedral sites above the ring formed by the four Ni, and four shared O corners. This hopping is facilitated by the rotation of a Li–O bond about two corner-shared NiO₆ octahedra, and consequent distortion of the initial LiO₄ tetrahedron (Fig. 4E, SI Appendix, Fig. S9, and Movie S1). Recent ab initio modeling suggests concerted migration of multiple Li ions can proceed with energy barriers that are much lower than isolated Li hops, owing to strong ion–ion interactions at high Li concentrations (39). Future studies therefore may focus on exploring the effects of concentration dependence of ion diffusivity coupled with strain.

The expanded Ni–O bond lengths facilitating faster Li mobility in this system is akin to the larger mobility and reduced migration barriers of oxygen atoms in tensile strained oxides (40, 41). Experimental studies of Na⁺ intercalation into SNO (SI Appendix, Fig. S10) as well as Li⁺ doping experiments in related rare-earth perovskite nickelate systems [EuNiO₃ (ENO) and NdNiO₃ (NNO)] further suggest the generality of the electrochemical poling strategy to design ion shuttles in strongly correlated oxides (SI Appendix, Fig. S11).

We demonstrate that cations such as Li⁺, Na⁺ can be reversibly incorporated into perovskite nickelate lattices under voltage bias in electrochemical cells. Strain due to ion incorporation in interstitial sites enables rapid diffusion with reduced activation barriers. The electron doping from intercalated Li, resulting in Ni²⁺, suppresses electronic conduction and enables selective ionic transport with tunable resistance states.

We acknowledge National Science Foundation Grant 1609898 and Air Force Office of Scientific Research Grant FA9550-16-1-0159 for support. M.K. and K.M.R. acknowledge support from Office of Naval Research Grant N00014-17-1-2770. D.L. and V.G.P. acknowledge support from Office of Naval Research Grant N00014-18-1-2397. This research used resources of the Argonne Leadership Computing Facility, which is a US Department of Energy (DOE) Office of Science User Facility supported under Contract DE-AC02-06CH11357. Use of the Center for Nanoscale Materials, an Office of Science User Facility, was supported by the US Department of Energy, Office of Science, Office of Basic Energy Sciences, under Contract DE-AC02-06CH11357. This research used resources of the National Energy Research Scientific Computing Center, a DOE Office of Science User Facility supported by the Office of Science of the US DOE under Contract DE-AC02-05CH11231. Z.Z. acknowledges support from Army Research Office Grant W911NF-16-1-0042. This research used resources of the APS, an Office of Science User Facility operated for the US DOE Office of Science by Argonne National Laboratory, and was supported by the US DOE under Contract DE-AC02-06CH11357, and the Canadian Light Source and its funding partners. Q.L. and B.Y. acknowledge funding support from the Massachusetts Institute of Technology Materials Research Science and Engineering Center (MRSEC) through the MRSEC Program of the National Science Foundation under Award DMR-1419807. This research used resources 23-ID-2 beamline of the National Synchrotron Light Source II, a US DOE Office of Science User Facility operated for the DOE Office of Science by Brookhaven National Laboratory under Contract DE-SC0012704. H.T. and A.H. acknowledge Japan Society for the Promotion of Science KAKENHI Grant 15KK0236. S.G. acknowledges support provided by National Science Foundation Grant 1553251. The work of Y.A. was supported by the Air Force Office of Scientific Research Grant FA9559-16-1-0172.