Split ODE Solvers

Split ODE Solvers

The solvers which are available for a SplitODEProblem depend on the input linearity and number of components. Each solver has functional form (or many) that it allows.

Implicit-Explicit (IMEX) ODE

The Implicit-Explicit (IMEX) ODE is a SplitODEProblem with two functions:

\[\frac{du}{dt} = f_1(t,u) + f_2(t,u)\]

where the first function is the stiff part and the second function is the non-stiff part (implicit integration on f1, explicit integration on f2).

Recommended Methods

The recommended method in most cases is KenCarp4. In cases of extreme stiffness or for high tolerances, KenCarp3 can be a good choice. The ARKODE methods are generally inefficient and diverge unless the options are tweaked to match the problem, though for large enough PDEs the ARKODE method with linear_solver=:GMRES is a good choice.

OrdinaryDiffEq.jl

Sundials.jl

Semilinear ODE

The Semilinear ODE is a split ODEProblem with one linear operator and one function:

\[\frac{du}{dt} = Au + f(t,u)\]

where the first function is a constant (not time dependent)AbstractDiffEqOperator and the second part is a (nonlinear) function. ../../features/diffeq_operator.html.

The appropriate algorithms for this form are:

OrdinaryDiffEq.jl

These methods utilize caching of the exponential operators and are thus are faster than Krylov-based methods but are only suited for smaller systems where expm(dt*A) can fit in memory.

Note that the generic algorithms allow for a choice of nlsolve.

Additional Krylov-based methods which allow for lazy calculation of expm(dt*A)*v are in development.