Mechano-Driven Logic-in-Memory with Neuromorphic Triboelectric Charge-Trapping Transistor

In the post-Moore's Law era, there is a growing trend towards the development of advanced electronic devices that combine sensory perception, data storage, and computation for various applications. Two-dimensional semiconductor transistors, which utilize charge storage mechanisms, present a promising avenue for future information devices. Here, we introduce a neuromorphic triboelectric charge-trapping MoTe2 transistor with stacked high-k dielectric structure, aiming to facilitate mechano-driven logic-in-memory for neuromorphic computation. By gating through triboelectric potential, the device demonstrates superior electrical performance, including an impressive switching ratio (>105), minimal off-state current (~0.6 pA), and robust cyclic stability. By modulating the trapped charges in the stack gate structure via tribopotential modulation, the conductivity state of the MoTe2 channel can be readily controlled, realizing an exceptional mechano-driven nonvolatile memory with a retention time of up to 104seconds, consistent switching behavior over 100 cycles, and multi-level data storage capabilities at 8 levels. Furthermore, a mechano-driven programmable inverter can be achieved by connecting a load resistor in series. The triboelectric charge-trapping transistor also possesses the capacity to emulate typical synaptic characteristics at low energy levels (~147 fJ). Leveraging the finely tunable conductivity through tribopotential, we demonstrate a mechano-assisted artificial neural network capable of recognizing handwritten digits with an accuracy rate of approximately 88.59%. These findings underscore the significant potential of the triboelectric charge-trapping transistor in mechanical-assisted real-time interaction, energy-efficient data storage, and neuromorphic computing.