A convergent string method: Existence and approximation for the Hamiltonian boundary-value problem

TL;DR

This paper introduces a convergent string method based on Jacobi's principle and Birkhoff's curve shortening to prove the existence and approximate long-time solutions of Hamiltonian boundary value problems, relevant in molecular dynamics.

Contribution

It develops a new convergent string method for Hamiltonian boundary value problems, enabling existence proofs and numerical approximations of long-time trajectories.

Findings

Proves existence of long-time solutions connecting configurations.

Develops a string method inspired by Birkhoff's curve shortening.

Provides a framework for detecting transition paths in molecular dynamics.

Abstract

This article studies the existence of long-time solutions to the Hamiltonian boundary value problem, and their consistent numerical approximation. Such a boundary value problem is, for example, common in Molecular Dynamics, where one aims at finding a dynamic trajectory that joins a given initial state with a final one, with the evolution being governed by classical (Hamiltonian) dynamics. The setting considered here is sufficiently general so that long time transition trajectories connecting two configurations can be included, provided the total energy $E$ is chosen suitably. In particular, the formulation presented here can be used to detect transition paths between two stable basins and thus to prove the existence of long-time trajectories. The starting point is the formulation of the equation of motion of classical mechanics in the framework of Jacobi's principle; a curve shortening procedure inspired by Birkhoff's method is then developed to find geodesic solutions. This approach can be viewed as a string method.