The proposed technology involves a synapse based on highly manufacturable Charge Trap Flash (CTF) Memory on Silicon-On-Insulator (SOI) technology. It demonstrates gradual symmetric Long-Term Potentiation (LTP) and Long-Term Depression (LTD) with arbitrarily tunable states, making it suitable for neuromorphic computing applications.
Type of IP
Faculty
Category of Patent
Department
The technology enables the development of efficient and scalable neuromorphic systems, contributing to advancements in AI and machine learning. This can lead to more energy-efficient computing systems and potentially transform various sectors by providing more intelligent and adaptive technologies.
- Gradual symmetric LTP and LTD demonstrated.
- Tunable number of states (101 to 104) by pulse characteristics.
- Spike-Timing Dependent Plasticity (STDP) demonstrated by FN tunneling.
- Synaptic unit cell size of 30F².
- Endurance of more than 109 LTP/LTD cycles.
- Energy-efficient with <660 aJ per essential energy spike
3
The technology is implemented with a single cell Charge Trap Flash (CTF) memory for ultra-low energy synapses (<2fJ). These synaptic cells effectively represent tunable and symmetric learning capabilities, demonstrating their utility in Spiking Neural Networks (SNN) and Deep Neural Networks (DNN) algorithms for neuromorphic computing.
The technology is highly manufacturable and compatible with the CMOS silicon industry, making it suitable for applications in neuromorphic computing, artificial intelligence, and advanced computing systems that require low-energy, high-reliability synaptic devices.
The primary application of this technology is in neuromorphic computing systems, where it can be used to build large-scale neural networks that mimic the human brain's learning and energy efficiency. It is particularly suited for real-time learning applications and advanced AI system.