Numerics of ODEs on exterior algebra spaces 

Given a vector space V of dimension n,
there are a number of other vector spaces that can be
built on it: the dual space, the spaces of kvectors, and
kforms, for k=0,...,n. Given a linear ODE
on V it is often of interest to numerically
integrate the induced systems on exterior algebra spaces.
Such systems
arise in the linearization of nonlinear ODEs about trajectories or
nonlinear PDEs about solitary waves, where
V is a model for the tangent space of the phase space.
These lectures
discuss the theory behind such equations and the implementation of
numerical algorithms for their integration.
Applications of the theory to the stability of solitary waves,
solution of boundary value problems, and hydrodynamic stability
are presented. 




Geometric numerical integration for breaking water waves 

The widely used governing equations for modelling water waves are Hamiltonian,
and therefore one would expect that symplectic integrators would be appropriate
for time integration. In this lecture the use of symplectic or other
geometric integrators for water waves is discussed.
For a simple free surface (a graph) the Hamiltonian formulation is canonical, but still the use of symplectic integrators is not straightforward. For general surfaces, for example breaking waves, one needs a coordinatefree Hamiltonian formulation, and this was first proposed by Benjamin & Olver (1982). For breaking waves the Hamiltonian structure is no longer canonical, and new ideas
from geometric integration are needed.





Multisymplectic structures and geometric integration 

In these lectures, the concept of multisymplecticity and its role
in the discretization of partial differential equations is discussed.
The topics to be discussed include: overview of multisymplecticity,
variational integrators and the Cartan form, continuous and discrete
conservation of symplecticity, discrete multisymplectic structures, and
implications for waves. A new approach to multisymplectic structures
will be also introduced and its implications for numerics discussed.
The latter
idea is based on the observation that any Riemannian manifold has a
natural coordinatefree multisymplectic structure
on the total exterior algebra bundle, and this
``canonical multisymplectic structure'' turns out
to be useful for analysis and numerics of Hamiltonian PDEs.





