Defect Engineering in Cobaltite Perovskites for Neuromorphic Applications

Conspectus

The rapidly growing energy cost of artificial intelligence presents an urgent challenge to conventional computing architectures based on separate processing and memory units. Neuromorphic devices, which integrate memory and logic into a single unit and operate in an event-driven manner, offer a promising alternative. A potential avenue to realize neuromorphic architectures is based on materials that can switch between distinct resistance states reversibly, at low voltage, and at room temperature. Transition metal oxides with oxygen-vacancy driven metal–insulator transitions (MITs) are compelling candidates, but the microscopic mechanisms governing these transitions have remained poorly understood, limiting the ability to design and optimize materials for device applications.

In this Account, we describe the authors’ effort at developing, and validating experimentally, a quantitative, first-principles understanding of ion migration and oxygen vacancy engineering in cobaltite perovskites, using La1–xSrxCoO3-δ (LSCO) as the primary model system. We started by asking a basic mechanistic question: what drives the MIT in LSCO? Using density functional theory calculations, we found that the MIT is not simply determined by local bonding changes at oxygen-deficient cobalt sites, but rather by cooperative, global structural distortions.

Building on this mechanistic foundation, we developed a model to predict the electrical bias required to trigger the MIT electrochemically. The model expresses the threshold voltage in terms of the oxygen vacancy formation free energy, the high-frequency dielectric constant, and the permanent polarization. The identification of these key quantities enables the prediction of concrete design guidelines for minimizing the operating voltage.

We also conducted experimental work to map the topotactic phase landscape of LSCO thin films, where we observed the emergence of the Ruddlesden–Popper phase under the most reducing conditions, extending well beyond the well-known perovskite–brownmillerite transition. This result was also confirmed thermodynamically by first-principles calculations. To further refine experimental work on oxygen-deficient phases, we developed a combined experimental and theoretical protocol for quantifying oxygen stoichiometry in thin films through the analysis of X-ray absorption spectra, providing a practical diagnostic tool.

Our investigations were further extended in three directions. First, we discovered that direct hydrogenation using H2 gas provides a low-temperature route to phase control, with hydroxyl bond formation catalyzing oxygen vacancy creation at neighboring lattice sites. Second, we predicted that a metallic interface could emerge between two coexisting insulating phases of LSCO as a result of orbital reconstruction driven by charge transfer and symmetry mismatch. This metallic state, which can be switched on or off by strain or gate voltage, opens a new route to low-energy resistive switching. Third, we showed that the cooperative distortion framework established for LSCO generalizes beyond cobaltites: Grenier-phase iron perovskites R1/3A2/3FeO2.67 exhibit tunable ferroelectricity by the same mechanism of concerted octahedral and tetrahedral unit distortions.

Further, combining theory and operando synchrotron X-ray diffraction, we identified “hidden” Grenier phases that form transiently during the perovskite-to-brownmillerite transition in LSCO; their metallic or insulating character depends nonmonotonically on Sr doping concentration, a trend that explains the nonmonotonic dependence of the threshold voltage on Sr content observed experimentally.