Speaker
Description
Topological Anderson Insulator (TAI) phase is often hard to characterize as it is driven by disorder. While Su-Schreiffer-Heeger (SSH) system is well understood but its long-ranged hopping and quasi-periodically disordered extensions offer a topologically richer landscape. Calculation of conventional topological invariants like real space winding numbers in such a system become both a numerically unstable and computationally expensive exercise as they lack translational symmetry. In this work we have proposed a machine learning framework for the identification of topological phases by training an Artificial Neural Network (ANN) that spots these phases by focusing on coalescence of central eigen modes into edge localized zero energy modes under open boundary conditions (OBCs). We have performed a comparative analysis which showcases superiority of ANN based approach over traditional winding number calculations in accurate classification of phases even in the regions where invariants fluctuate traditionally. Our results demonstrate that spectral feature learning is a robust method for identifying disordered topological phases, contributing to the growing intersection of machine learning and condensed matter physics.
Keywords: Topological Anderson Insulator, Quasi-Periodic Systems, Artificial Neural Networks, SSH Model, Phase Transitions.