Introductory Jupyter notebooks¶
These notebooks provide an introduction to usage of Firedrake, and are designed to familiarise you with manipulating Firedrake objects to solve finite element problems.
To run the notebooks, you will need to install jupyter inside your activated Firedrake virtualenv.
These notebooks are maintained in the Firedrake repository, so all the
material is available in your Firedrake installation source
directory. If you installed in
Documents/firedrake, then the
notebooks are in the directory
Documents/firedrake/src/firedrake/docs/notebooks. The links to
the notebooks below are non-interactive renderings using Jupyter
A first example¶
In this notebook, we solve the symmetric positive definite “Helmholtz” equation, and learn about meshes and function spaces. A rendered version of this notebook is available here.
Incorporating strong boundary conditions¶
Next, we modify the problem slightly and solve the Poisson equation. We introduce strong (Dirichlet) boundary conditions and how to use them.
A vector-valued problem¶
Moving on from scalar problems, we look at our first vector-valued problem, namely the equations of linear elasticity. In this notebook, we learn about some of UFL’s support for tensor algebra, and start looking at configuring linear solvers.
A time-dependent, nonlinear, problem¶
This notebook looks at a simple nonlinear problem, the viscous Burgers’ equation, and also treats simple timestepping schemes. We learn about formulating nonlinear, as opposed to linear problems, and also a little bit about how to write efficient Firedrake code.
A mixed formulation of the Poisson equation¶
In this notebook., we look at our first mixed finite element problem. A dual formulation of the Poisson equation. This equation also appears in the context of flow in porous media, as Darcy flow. We introduce mixed function spaces and how to work with them. Equations with multiple variables are typically more challenging to precondition, and so we discuss some of the preconditioning strategies for such block systems, and how to control them using PETSc solver options.
PDE-constrained optimisation with dolfin-adjoint¶
Now that we’ve learnt how to solve some PDEs, we might want to consider optimisation subject to PDE constraints. This notebook introduces the use of dolfin-adjoint to solve PDE constrained optimisation problems. We solve the Stokes equations and minimise energy loss due to heat, controlling inflow/outflow in a pipe.
The next notebook looks a little bit at the support Firedrake has for geometric multigrid, and how you can configure complex multilevel solvers purely using PETSc options.
We next dive a little deeper into the advanced ways in which Firedrake and PETSc enable solvers and preconditioners to be composed in arbitrarily complex ways to create an optimal solution strategy for a particular problem.
Building on the theme of composable solvers, we now explore Firedrake’s capabilities in the area of static condensation and hybridisation.