Characterizing topological pumping of charges in exactly solvable Rice-Mele chains of the non-Hermitian variety

TL;DR

This paper investigates the topological charge pumping in a non-Hermitian extension of the Rice-Mele model, analyzing how non-Hermiticity affects quantized transport and the associated topological invariants.

Contribution

It provides an analytical study of non-Hermitian Rice-Mele chains, introduces the use of generalized Brillouin zones, and examines the validity of topological charge pumping in non-Hermitian systems.

Findings

Chern numbers are well-defined in gapped regions for NH systems.

Deviations from quantized charge pumping occur when eigenvalues have large imaginary parts.

Non-Hermitian effects challenge the conventional understanding of adiabaticity and topological invariants.

Abstract

We address the nature of the Thouless charge-pumping for a non-Hermitian (NH) generalization of the one-dimensional (1d) Rice-Mele model, considering a variety which allows closed-form analytical solutions for the eigensystems. The two-band system is subjected to an ``adiabatic'' time-periodic drive, which effectively gives us a closed two-dimensional (2d) manifold in the momentum-frequency space, mimicking a 2d Brillouin zone (BZ) for periodic boundary conditions. For open boundary conditions, we formulate the non-Bloch generalized Brillouin zone (GBZ). All these allow us to compute the Chern numbers of the 2d manifolds emerging for various parameter values, using the BZ or GBZ, which are well-defined integers in the gapped regions of the spectra. If Thouless pumping still holds for NH situations, the expectation values of the net displacement for the particles occupying a given band within a cycle, calculated using the right and left eigenvectors of a biorthogonal basis for the complex Hamiltonians, would coincide with the Chern numbers. However, we find that there are deviations from the expected quantized values whenever the spectra exhibit strongly-fluctuating magnitudes of the imaginary parts of the (generically) complex eigenvalues. This implies that we need to address the question of how to meaningfully define time-evolution, as well as adiabaticity, in the context of NH systems.