Reservoir Computing-Based Real-Time Prediction for Quantized Conductance of Au Atomic Junctions

Abstract

Feedback-controlled electromigration method has been known to fabricate the atomic junction by controlling the quantized conductance on the Au nanowires. We demonstrated a real-time prediction of the quantized conductance on the Au nanowires dynamically inferred by reservoir computing on a real-time operating system. As a result, the root mean squared error in ballistic regime of the proposed system is 0.414. We believe that our system makes a valuable contribution to the fabrication of nanoscale devices in that it introduces the real-time prediction under running processes.