Researchers develop monolithic 3D memristor stack for neuromorphic computing

Researchers at Seoul National University and Yonsei University have published a technical paper detailing a monolithic three-dimensionally integrated memristor-thin-film transistor stack, Semiconductor Engineering reported. The research introduces an electrically programmable device designed for multimode reservoir computing to improve temporal information processing. [1]

According to Semiconductor Engineering, conventional hardware reservoirs used in temporal processing often suffer from fixed or narrowly tunable functionality. The new monolithic stack aims to provide electrically programmable hardware to overcome these limitations in neuromorphic computing architectures. [1]

A three-dimensional integrated circuit stacks multiple integrated circuits vertically, using connections such as through-silicon vias or copper-copper interfaces. This design approach allows stacked components to behave as a single device, offering reduced power consumption and a smaller physical footprint compared with traditional two-dimensional manufacturing processes. [2]

The research team published their findings under the title Monolithic three-dimensionally integrated memristor-thin-film transistor for electrically programmable multimode reservoir computing, Semiconductor Engineering reported. The architecture represents an effort to improve hardware flexibility for dynamic neuromorphic workloads. [1]