In-Materio Reservoir Computing Using Tailored Epitaxial Complex Oxides

The ever-increasing power consumption of electronics necessitates the development of new energy-efficient device paradigms for data storage and computation. In this pursuit, approaches aimed at mimicking the functionality of the human brain in artificial devices, so-called neuromorphic engineering, have attracted a lot of attention. Reservoir computing relying on a random, untrained reservoir provides a powerful platform to exploit materials for brain-inspired computing without strict requirements on the device-to-device variability. Implementing reservoir computing in-materio is, however, challenging because of difficulties to adapt the two main reservoir properties – nonlinearity and memory capacity - to computational tasks. Epitaxial complex oxides offer a wealth of nonlinear material properties with tunable memory characteristics, making them ideal candidates for task-adaptive physical reservoir computing. With RECOMPUTE, I aim to establish in-materio reservoir computing using epitaxial complex oxides to efficiently perform computational tasks. I propose to achieve this by combining in thin-film epitaxial heterostructures (i) rare-earth nickelates with a tunable nonlinear negative differential resistance and (ii) relaxor ferroelectrics with a memory capacity that is tunable via the chemical composition. By fabricating prototypical thin-film devices, characterizing and optimizing their electrical transport characteristics, and simulating their computational performance, I intend to shed light on the fundamental relationship between the intrinsic material properties of reservoir devices and their performance in solving different cognitive tasks – a crucial insight that has been elusive so far.Ultimately, the successful implementation of the proposed research will establish novel reservoir computing devices based on complex oxides and serve as a template for the study of other nonlinear systems for unconventional computing among and beyond complex oxides.