Skip to main content
Open Access Publications from the University of California


UCLA Electronic Theses and Dissertations bannerUCLA

A silicon-based self-programming synaptic resistor network for neuromorphic computing


Compared to modern supercomputers, which consume roughly 10^6 W of power, the human brain requires only 20 W to function, and still exceeds the performance of supercomputers in many creative tasks. This stark difference in energy requirements is caused by a fundamental difference in computing architecture. Modern computers follow the Von Neumann architecture, in which transistors dedicated to logic and memory functions are physically separated, and the time and energy required to communicate between the two units constitutes a bottleneck which impedes performance in machine learning and optimization problems. On the other hand, in the human brain, logic, memory, and learning functions are integrated together in a single element: the synapse. Without the Von Neumann bottleneck, the brain can achieve fast real-time learning, adaptability in complex environments, and massive parallelism. For the future of neuromorphic computing, it is important to develop an electronic device which mimics the synaptic function, so that large-scale circuits which mimic the neurobiological architecture can be developed. This work reports a silicon-based synaptic resistor (referred to as “synstor” hereinafter) which integrates logic, learning, and memory in a single device. The synstor is composed of a semiconducting silicon channel connected via Schottky contacts to titanium input and output electrodes, a thermal silicon dioxide, an aluminum oxide switching layer, and a tantalum oxide reference electrode. The large defect density in the switching layer attracts or repels charge carriers in the silicon channel to modify its conductance, and the defect density in turn can be modified by voltage pulses applied on its input and output electrodes (pre- and post-synaptic spikes). Synaptic resistor circuits could be scaled-up to facilitate mobile artificial intelligence systems with brain-like intelligence and adaptability in complex environments.

Main Content
For improved accessibility of PDF content, download the file to your device.
Current View