An explicit semi-Lagrangian, spectral method for solution of Lagrangian transport equations in Eulerian-Lagrangian formulations

TL;DR

This paper introduces an explicit high-order semi-Lagrangian spectral method for solving Lagrangian transport equations within Eulerian-Lagrangian frameworks, ensuring consistency with spectral element discretizations and enabling efficient, local, and parallel computations.

Contribution

The paper develops a novel explicit semi-Lagrangian spectral method that integrates particles at Gauss points and maps solutions back via least squares, maintaining consistency with spectral element discretizations.

Findings

Method converges exponentially in numerical tests.

Mass and energy constraints can enhance accuracy.

Particles remain within element bounds during integration.

Abstract

An explicit high order semi-Lagrangian method is developed for solving Lagrangian transport equations in Eulerian-Lagrangian formulations. To ensure a semi-Lagrangian approximation that is consistent with an explicit Eulerian, discontinuous spectral element method (DSEM) discretization used for the Eulerian formulation, Lagrangian particles are seeded at Gauss quadrature collocation nodes within an element. The particles are integrated explicitly in time to obtain an advected polynomial solution at the advected Gauss quadrature locations. This approximation is mapped back in a semi-Lagrangian fashion to the Gauss quadrature points through a least squares fit using constraints for element boundary values and optional constraints for mass and energy preservation. An explicit time integration is used for the semi-Lagrangian approximation that is consistent with the grid based DSEM solver, which ensures that particles seeded at the Gauss quadrature points do not leave the element's bounds. The method is hence local and parallel and facilitates the solution of the Lagrangian formulation without the grid complexity, and parallelization challenges of a particle solver in particle-mesh methods. Numerical tests with one and two dimensional advection equation are carried out. The method converges exponentially. The use of mass and energy constraints can improve accuracy depending on the accuracy of the time integration.