Ferroelectric AlScN Four-Mode Reconfigurable Transistors Enabling Unified Logic, Memory, and Neuromorphic Computing

Abstract

The increasing demand for data-centric and artificial intelligence workloads has intensified the need for device platforms that integrate logic, memory, and adaptive computing beyond conventional von Neumann architectures. Here, an aluminum scandium nitride (AlScN)-based nonvolatile four-mode reconfigurable field-effect transistor is demonstrated as a multifunctional device platform. By combining the large remanent polarization of AlScN with the ambipolar transport of the MoTe2 channel, the same transistor can be persistently programmed into logic nFET, logic pFET, memory-on, and memory-off states. These modes arise from polarization-controlled modulation of carrier injection and channel electrostatics, enabling a single device to function as a polarity-reconfigurable logic element, nonvolatile memory cell, and analog synaptic element. Programming is performed through voltage pulses applied to the source, drain, and back gate, while top-gate readout is decoupled from AlScN programming, enabling stable nonvolatile operation. Based on this device, a single-transistor ternary content-addressable memory cell concept and two-input reconfigurable multifunctional logic gates are demonstrated. The wide polarization-tunable memory window further supports multilevel cell memory and nearly linear long-term potentiation and depression for neuromorphic synaptic operation. These results establish ferroelectric AlScN as a robust platform for integrating reconfigurable logic, nonvolatile memory, associative search, and neuromorphic computing within a common device framework.