Consequences of local gauge symmetry in empirical tight-binding theory

TL;DR

This paper develops a gauge-invariant tight-binding model incorporating electromagnetic fields, using group theory to construct a basis of atomic-like orbitals, and applies it to semiconductors Ge and Si.

Contribution

It introduces a gauge-invariant tight-binding approach with a new basis construction based on group theory, improving the description of electromagnetic interactions in semiconductors.

Findings

Accurately models valence and conduction bands of Ge and Si.

Yields a good density of states but underestimates oscillator strength.

Imposes strong restrictions on the Hamiltonian range due to gauge symmetry.

Abstract

A method for incorporating electromagnetic fields into empirical tight-binding theory is derived from the principle of local gauge symmetry. Gauge invariance is shown to be incompatible with empirical tight-binding theory unless a representation exists in which the coordinate operator is diagonal. The present approach takes this basis as fundamental and uses group theory to construct symmetrized linear combinations of discrete coordinate eigenkets. This produces orthogonal atomic-like "orbitals" that may be used as a tight-binding basis. The coordinate matrix in the latter basis includes intra-atomic matrix elements between different orbitals on the same atom. Lattice gauge theory is then used to define discrete electromagnetic fields and their interaction with electrons. Local gauge symmetry is shown to impose strong restrictions limiting the range of the Hamiltonian in the coordinate basis. The theory is applied to the semiconductors Ge and Si, for which it is shown that a basis of 15 orbitals per atom provides a satisfactory description of the valence bands and the lowest conduction bands. Calculations of the dielectric function demonstrate that this model yields an accurate joint density of states, but underestimates the oscillator strength by about 20% in comparison to a nonlocal empirical pseudopotential calculation.