Neuromorphic computing, aiming to mimic the human brain's energy efficiency, requires materials with tunable resistance states responsive to input signal intensity, frequency, and history. Transition metal oxides (TMOs) showing metal-to-insulator transitions (MITs) are promising candidates. La_1−xSr_xCoO_3−δ (LSCO) is a TMO exhibiting a significant resistivity change with varying pressure or doping. Recent studies suggest controlling the MIT by tuning oxygen vacancy concentration, resulting in a topotactic transition from a paramagnetic metallic perovskite to an antiferromagnetic (AFM) semiconducting brownmillerite (BM) structure. However, the underlying mechanisms remain unclear, hindering the design of efficient neuromorphic devices. This research uses first-principles calculations to investigate the MIT in LSCO and the semiconductor-to-insulator transition in LaCoO₃ (LCO), exploring the interplay between structural, electronic, and magnetic properties during the perovskite-to-BM phase transition as a function of oxygen vacancies.

Existing literature highlights the potential of TMOs, specifically LSCO, for neuromorphic computing due to their tunable resistivity. Previous studies demonstrated resistivity changes in LSCO with pressure and doping, while recent work explored oxygen vacancy control as a means to induce the MIT. However, the microscopic mechanisms responsible for this transition remain elusive. Studies have shown that oxygen vacancy concentration can be varied using different methods like depositing oxygen-scavenging metals, annealing in a reducing environment, applying electric fields, and using epitaxial strain. Despite extensive research, the fundamental mechanisms driving the MIT in cobaltites remain poorly understood, hindering the ability to design efficient neuromorphic devices based on these materials.

Density functional theory (DFT) with a Hubbard correction (U) was used. The PBE functional with U = 3 eV was selected after benchmarking against experimental data for LCO and BM SCO_2.5. Calculations were performed on a supercell with four layers of octahedral units to investigate structural distortions as a function of oxygen vacancy concentration (4.2%, 8.3%, 12.5%). Symmetry-inequivalent oxygen vacancy positions and different magnetic states (ferromagnetic, FM, and several AFM states) were considered. The lowest-energy structures were determined, revealing a transformation of CoO₆ units into CoO₅ and CoO₄ units, ultimately leading to the BM phase. The geometry changes, including supercell volume expansion and Co-O-Co tilt angle variation, were analyzed. A Hamiltonian model was developed to represent the atomic deformations in the perovskite phase as a combination of Jahn-Teller (JT), breathing, and rotation distortions. Ab initio phonon calculations were used to estimate the stiffness of these distortions. The Hamiltonian was rewritten to incorporate the coupling between JT/breathing mode and rotations, expressed in momentum space to represent phonon-phonon interactions. The changes in magnetic properties with increasing oxygen vacancy concentration were investigated. The electronic structure was studied using the density of states (DOS) analysis and Born effective charge calculations. The effect of structural distortions on the band gap was assessed by constraining the structure to specific magnetic states and angles, examining the role of electron correlation and spin polarization. Finally, a model was developed to predict the required electric field to trigger the MIT by considering the oxygen vacancy formation energy, computed using the electrochemical potential needed to spontaneously form vacancies. The dielectric constant and polarization were calculated using density functional perturbation theory (DFPT) and the Berry phase approach.

The study revealed a complex interplay between structural, electronic, and magnetic transitions in LSCO and LCO as a function of oxygen vacancy concentration. The MIT was found to be primarily driven by cooperative, global structural distortions, specifically the rotations of Co-oxygen units, rather than local bonding changes. The lowest-energy transition pathway from perovskite to BM structure was identified and involves a layer-by-layer nucleation process of oxygen vacancies. Upon introducing oxygen vacancies, the perovskite lattice expands, accompanied by a significant decrease in the Co-O-Co angle. The combination of Jahn-Teller and breathing mode distortions caused lattice expansion. The study demonstrated that the rotations of octahedra effectively counterbalance the increase in elastic energy caused by JT distortions. A transition from a non-magnetic (in LCO) or ferromagnetic (in LSCO) state to an antiferromagnetic state was observed with increasing oxygen vacancy concentration. The MIT was directly linked to this magnetic state transition. The density of states analysis revealed that the band gap opening is intimately connected to both the changes in the magnetic state and the development of structural distortions leading to pyramidal and tetrahedral coordination. The band gap increases with increasing structural deformations. The crystal field splitting of oxygen-deficient Co was found to be significantly influenced by the spin configuration of neighboring octahedral sites, indicating that the electronic effects of oxygen vacancies are not strictly local. A model predicting the electric field required to induce the MIT was developed and showed good agreement with experimental observations. The required voltage for the transition was estimated to be around 1.2V for LCO and 0.8V for LSCO, comparable to the operating voltages in other neuromorphic devices.

The findings address the research question by revealing the microscopic mechanism of the MIT in LSCO and LCO, clarifying the role of cooperative structural distortions, magnetic transitions, and electronic structure changes. The significance lies in providing a clear understanding of the MIT, enabling the rational design of cobaltite-based neuromorphic devices. The results challenge the existing notion that local bonding changes are the dominant factor, instead highlighting the importance of global cooperative phenomena. The agreement between the theoretical predictions of the required electric field and experimental measurements validates the model's accuracy and general applicability. The identification of the key parameters (volume expansion, rotations, and magnetic transitions) needed for the MIT provides valuable design rules for future neuromorphic devices. This research significantly contributes to the understanding of topotactic transformations in TMOs and their application in energy-efficient computing.

This study successfully characterized the metal-insulator transition in La_1−xSr_xCoO_3−δ, emphasizing the crucial role of cooperative structural distortions and magnetic transitions. The first-principles model accurately predicted the electric bias required for the transition. The findings offer a comprehensive understanding of the MIT mechanism, facilitating the design of improved neuromorphic devices. Future work could focus on exploring the impact of strain and interfacial effects on the MIT, further optimizing material properties for neuromorphic applications.

The study primarily focuses on specific oxygen vacancy concentrations and configurations. A more comprehensive investigation across a broader range of conditions could provide a more complete picture of the MIT. The Hamiltonian model used for atomic deformations relies on the harmonic approximation, which might not capture all the complexities of the system's behavior. Further experimental validation of the predicted electric field values and cooperative distortions is necessary. The study assumes a relatively uniform distribution of Sr dopants, and further investigations may be needed to fully understand the effects of varying Sr dopant distributions.