Mimicry of Biological Neurons in an Array of MoS ₂ Field Effect Transistors with Ionic Liquid Gates

Neuromorphic computing has recently emerged as a biologically inspired scheme for power-efficient data processing. Memory devices based on field-effect transistors (memFETs) of two-dimensional materials, such as monolayer MoS2, have been explored for such applications due to their high extinction ratio, switching speed, and endurance. Herein, we further explore memFETs with ionic liquid gates (ILG) to mimic biological neurons in an array of nine MoS2 memFETs. Unlike a memFET with a regular dielectric gate, characteristics similar to the leaky integrate-and-fire (LIF) mechanism of biological neurons have been demonstrated in the MoS2 memFETs with an ILG. Microscopically, this characteristic is attributed to the complex dynamic response of the molecular dipoles in the ILG under applied gate voltages and the consequent impact on the electric double layer on the monolayer MoS2 memFET channel. As a demonstration for neuromorphic circuits, pattern-recognition with an accuracy of 96% and 97% were obtained for memFET encoded and nonencoded data, respectively, achieving a factor of ∼3 decrease in training time with a 1% loss in image recognition accuracy. This result illustrates that monolayer MoS2 memFETs arrays are promising for neuromorphic circuit applications, with unique advantages in mimicking biological neurons via a LIF mechanism.