Dual PIC: a structure preserving method for discretizing Lie-Poisson brackets

Abstract

We consider a general discretization strategy for Hamiltonian field theories generated by Lie-Poisson brackets which we call dual PIC (DPIC). This method involves prescribing two different discrete representations of the dynamical variable which are constrained as a Casimir invariant of the flow to coincide with one another via an L2 projection throughout the entire simulation. This allows one to leverage the relative advantages of each discrete representation. We begin by describing DPIC as applied to a general Lie-Poisson system and then provide illustrative examples: the discretization of the two-dimensional vorticity equations and the Vlasov-Poisson equation.