Quantum mechanics in an evolving Hilbert space

TL;DR

This paper introduces a differential geometric framework for understanding evolving basis sets in quantum mechanics, clarifying their effects and connecting to Berry phases, with applications to time-dependent electron dynamics.

Contribution

It formalizes the evolution of basis sets using differential geometry and fibre bundles, providing new insights and methods for quantum dynamics simulations.

Findings

New geometric formalism for basis set evolution in quantum mechanics

Generalized Berry connection and curvature expressions for non-orthogonal bases

Proposed novel finite-difference time integrators based on geometric insights

Abstract

Many basis sets for electronic structure calculations evolve with varying external parameters, such as moving atoms in dynamic simulations, giving rise to extra derivative terms in the dynamical equations. Here we revisit these derivatives in the context of differential geometry, thereby obtaining a more transparent formalisation, and a geometrical perspective for better understanding the resulting equations. The effect of the evolution of the basis set within the spanned Hilbert space separates explicitly from the effect of the turning of the space itself when moving in parameter space, as the tangent space turns when moving in a curved space. New insights are obtained using familiar concepts in that context such as the Riemann curvature. The differential geometry is not strictly that for curved spaces as in general relativity, a more adequate mathematical framework being provided by fibre bundles. The language used here, however, will be restricted to tensors and basic quantum mechanics. The local gauge implied by a smoothly varying basis set readily connects with Berry's formalism for geometric phases. Generalised expressions for the Berry connection and curvature are obtained for a parameter-dependent occupied Hilbert space spanned by non-orthogonal Wannier functions. The formalism is applicable to basis sets made of atomic-like orbitals and also more adaptative moving basis functions (such as in methods using Wannier functions as intermediate or support bases), but should also apply to other situations in which non-orthogonal functions or related projectors should arise. The formalism is applied to the time-dependent quantum evolution of electrons for moving atoms. The geometric insights provided here allow us to propose new finite-differences time integrators, and also better understand those already proposed.