Ionic Memristors with Gate Control for Low-Power Artificial Synapses

Modern computers are built on CMOS (Complementary Metal–Oxide–Semiconductor) technology, where memory and processing are separate. This design causes huge energy losses because data must move back and forth between the two units. For example, supercomputers use nearly 7.8 MW of power, while the human brain performs far more complex tasks at only ~20 W. To solve this problem, researchers are turning to neuromorphic devices, which mimic the brain by combining memory and processing in a single unit. These devices can learn and adapt while using far less energy. This project will develop a voltage-gated ionic memristor using very narrow channels in layered vermiculite, with a height of only about 3 Å. Inside these channels, calcium ions (Ca²⁺), similar to those used in biological synapses, will move and create memory effects. This project will explore all the symmetric and asymmetric designs of channels with voltage gating. By applying a gate voltage, the device can be tuned in real time to strengthen or weaken its memory states. This allows the device to reproduce brain-like functions such as short-term memory (learning and forgetting quickly) and long-term memory (stable learning over time), which is the fundamental principle for the fabrication of neuromorphic computing, which will be exploited in this project with periodic potential. To understand how memory forms in these tiny channels, COMSOL simulations will be used to study ion movement and the role of the gate voltage on it. At the system level, NeuroSim simulations will test how well these devices can recognize patterns, such as images. This combination of experiments and simulations will provide both a proof of concept and design rules for scaling the technology for the fabrication of the next generation of neuromorphic computing. The project is expected to deliver a new type of neuromorphic hardware that is far more energy efficient than CMOS.