Complex oxides for brain-inspired computing: A review.

TL;DR: The opportunities presented by complex oxides, a class of electronic ceramic materials whose properties can be elegantly tuned by doping, electron interactions and a variety of external stimuli near room temperature, are discussed.

Abstract: The fields of brain-inspired computing, robotics, and, more broadly, artificial intelligence seek to implement knowledge gleaned from the natural world into human-designed electronics and machines. Most advances to date have happened in software and algorithms. However, there is growing interest in designing hardware, i.e., new semiconductor materials and devices and their interconnections that can impart intelligence at the device or network level for software co-design. This requires a paradigm shift in the way one thinks of semiconductors where the intrinsic material or device itself can learn, be re-programmed, remember or forget information and, when connected, collectively create output signals that are non-trivial, emergent summation of individual components. In this review, we discuss the opportunities presented by complex oxides, a class of electronic ceramic materials whose properties can be elegantly tuned by doping, electron interactions and a variety of external stimuli near room temperature. Most importantly, oxides can host a vast number of electrically addressable metastable phases by carrier doping and subtle crystal structure distortions essential for adaptive devices. Concurrent changes in optical or magnetic properties create opportunities for multi-modal readout. Since neuroscience and evolutionary biology disciplines serve as an inspiration and scientific guiding disciplines for neuromorphic computing, we begin with a discussion of natural intelligence at the elementary level in the nervous system, followed by collective intelligence and learning at the animal colony level mediated by social interactions. An important aspect we highlight here is the vast spatial and temporal scales involved in learning and memory. We then focus on phenomena such as metal-to-insulator transitions, ferroelectricity, and related examples to highlight recent demonstrations of artificial neurons, synapses, and circuits and their learning behavior. First-principles theoretical treatments of the electronic structure; in-situ synchrotron spectroscopy of operating devices is then discussed. We review the implementation of the experimental characteristics into neural networks and algorithm design. Finally, we highlight outstanding materials challenges that require a microscopic understanding of the physical mechanisms which will be essential for advancing the frontiers of neuromorphic computing. This article is protected by copyright. All rights reserved.