Compactly-supported Wannier functions and algebraic $K$-theory

TL;DR

This paper explores the conditions under which compactly-supported Wannier functions can exist in various symmetry classes of band structures, revealing topological constraints linked to algebraic $K$-theory and extending previous results.

Contribution

It extends the topological analysis of compactly-supported Wannier functions to all ten symmetry classes using algebraic $K$-theory, identifying when such functions can exist.

Findings

Compactly-supported Wannier functions imply topologically trivial bundles in class A.

Non-trivial bundles can only have one-dimensional winding in each direction.

Results are derived using algebraic $K$-theory and polynomial rings in momentum space.

Abstract

In a tight-binding lattice model with $n$ orbitals (single-particle states) per site, Wannier functions are $n$-component vector functions of position that fall off rapidly away from some location, and such that a set of them in some sense span all states in a given energy band or set of bands; compactly-supported Wannier functions are such functions that vanish outside a bounded region. They arise not only in band theory, but also in connection with tensor-network states for non-interacting fermion systems, and for flat-band Hamiltonians with strictly short-range hopping matrix elements. In earlier work, it was proved that for general complex band structures (vector bundles) or general complex Hamiltonians---that is, class A in the ten-fold classification of Hamiltonians and band structures---a set of compactly-supported Wannier functions can span the vector bundle only if the bundle is topologically trivial, in any dimension $d$ of space, even when use of an overcomplete set of such functions is permitted. This implied that, for a free-fermion tensor network state with a non-trivial bundle in class A, any strictly short-range parent Hamiltonian must be gapless. Here, this result is extended to all ten symmetry classes of band structures without additional crystallographic symmetries, with the result that in general the non-trivial bundles that can arise from compactly-supported Wannier-type functions are those that may possess, in each of $d$ directions, the non-trivial winding that can occur in the same symmetry class in one dimension, but nothing else. The results are obtained from a very natural usage of algebraic $K$-theory, based on a ring of polynomials in $e^{\pm ik_x}$, $e^{\pm ik_y}$, . . . , which occur as entries in the Fourier-transformed Wannier functions.