Plasticity Mechanisms and Neuromorphic Learning with Correlated Semiconductors
Abstract
Strongly correlated materials represent an emerging class of semiconductors with highly tunable electronic structure. Oxides in the perovskite family display functional properties closely dependent on the orbital occupancy. We have recently demonstrated that reversible and massive electronic phase change can be induced in perovskite nickelates by electron doping with gap opening of the order of 3eV accompanied by electrical conductivity change ranging nearly 10 orders of magnitude at ambient temperature. In this collaborative proposal, the PIs aim to understand fundamental mechanisms of plasticity in two and three-terminal devices and neural learning algorithms that implement controlled memory loss. Building on recent collaborative work demonstrating environmental habituation and forgetting phenomena in nickelates via electron doping, we will experimentally measure conductance profiles under electrical bias in devices to compare plasticity properties with that observed in electrolyte-gated FETs. We will incorporate these electrical characteristics in neural network simulations to understand the learning properties that emerge when synaptic connections that demonstrate forgetting similar to brain-like characteristics are utilized. The relaxation kinetics due to the inherent plasticity will enable designing varying leak rates for the synaptic weights. Dopaminergic modulation of synaptic weights to mimic novelty detection and learning will be carried out. The statement of objectives include understanding the potential of correlated semiconductors in neuromorphic learning and algorithm development. The results from the project have potential relevance to the DoD mission in low power electronics, adaptive programmable circuits and machine learning.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 24, 2019
- Source ID
- W911NF1910313
Entities
People
- Shriram Ramanathan
Organizations
- Army Contracting Command
- United States Army
- University of Virginia