Dynamics of Electrochemical Ionic Synapses to Emulate Spike Timing Depending Plasticity and other Bio-Realistic Learning Rules

In our research, we aim to reduce the energy consumption in analog, brain-inspired computing, by focusing on designing materials and devices that use ions to perform data storage and computation in a single architecture. In this talk, I will share our work on the ionic electrochemical synapses, whose electronic conductivity we can control deterministically by electrochemical insertion/extraction of dopant ions across the active device layer. We are exploring protons and magnesium ions, which bring different advantages to the operation of the device. The protons present very low energy consumption, on par with biological synapses in the brain. Magnesium ions present with better stability without the need for encapsulation. The modeling results indicate the desirable material properties, such as ion conductivity and interface charge transfer kinetics, that we must achieve for fast (ns), low energy (< fJ) and low voltage (1V) performance of these devices. We are also exploring ion dynamics in these materials to emulate bio-realistic learning rules deduced from neuroscience studies. These include spike timing dependence achieved by electrochemical ionic synapses, as well as a local learning rule describing how song birds learn to sing. Our findings provide pathways towards brain-inspired hardware that has high yield and consistency and uses significantly lesser energy as compared to current computing architectures.