A search for correlation-induced adiabatic paths between distinct topological insulators

TL;DR

This study investigates whether interactions can connect distinct topological insulators adiabatically, finding that trivializing surface states does not necessarily imply a bulk adiabatic path, highlighting complex interplay between surface and bulk properties.

Contribution

The paper provides a counterexample showing that surface trivialization via interactions does not guarantee an adiabatic connection between topological phases in the bulk.

Findings

No adiabatic path found between topological phases under studied interactions.

Extended symmetry-breaking region separates topological sectors.

Frustration influences phase transition nature and region size.

Abstract

Correlations in topological states of matter provide a rich phenomenology, including a reduction in the topological classification of the interacting system compared to its non-interacting counterpart. This happens when two phases that are topologically distinct on the non-interacting level become adiabatically connected once interactions are included. We use a quantum Monte Carlo method to study such a reduction. We consider a 2D charge-conserving analog of the Levin-Gu superconductor whose classification is reduced from $\mathbb{Z}$ to $\mathbb{Z}_4$. We may expect any symmetry-preserving interaction that leads to a symmetric gapped ground state at strong coupling, and consequently a gapped symmetric surface, to be sufficient for such reduction. Here, we provide a counter example by considering an interaction which (i) leads to a symmetric gapped ground state at sufficient strength and (ii) does not allow for any adiabatic path connecting the trivial phase to the topological phase with $w=4$. The latter is established by numerically mapping the phase diagram as a function of the interaction strength and a parameter tuning the topological invariant. Instead of the adiabatic connection, the system exhibits an extended region of spontaneous symmetry breaking separating the topological sectors. Frustration reduces the size of this long-range ordered region until it gives way to a first order phase transition. Within the investigated range of parameters, there is no adiabatic path deforming the formerly distinct free fermion states into each other. We conclude that an interaction which trivializes the surface of a gapped topological phase is necessary but not sufficient to establish an adiabatic path within the reduced classification. In other words, the class of interactions which trivializes the surface is different from the class which establishes an adiabatic connection in the bulk.