Exploiting Bias Temperature Instability for Reservoir Computing in Edge Artificial Intelligence Applications

In this paper, we utilize Negative Bias Temperature Instability (NBTI) effect in pFETs as a computing mechanism. Specifically, NBTI is capable of implementing a physical reservoir computer that can do one-shot learning. In both simulation (Comphy framework) and hardware (foundry pFETs), we demonstrate the feasibility of our algorithm. Notably, our implementation has no requirements for specialized device engineering, enabling a straightforward use of CMOS technology for reservoir computing.