ResNet CNN for Majorana Zero Mode Detection

Abstract

Majorana zero modes (MZMs) are the enabling phenomenon for topological quantum computing, but distinguishing them from trivial quasi-Majorana modes in the presence of disorder remains a central bottleneck. While most topological invariants fail to make this distinction, the recent periodic disorder invariant (PDI) is theoretically capable even in noisy systems. This work provides the first real evaluation of the PDI in an MZM prediction pipeline: a ResNet CNN is trained on simulated nanowire conductance data to predict PDI phase maps across experimental parameter ranges, providing a robust, noise-tolerant protocol for MZM detection.

This work was supported by the Clemson University Creative Inquiry + Undergraduate Research Program and the South Carolina Quantum Association.