| 1 |
|
| 2 |
.. _demo_poisson_equation: |
| 3 |
|
| 4 |
Poisson equation |
| 5 |
================ |
| 6 |
|
| 7 |
This demo is implemented in a single Python file, |
| 8 |
:download:`demo_poisson.py`, which contains both the variational forms |
| 9 |
and the solver. |
| 10 |
|
| 11 |
This demo illustrates how to: |
| 12 |
|
| 13 |
* Solve a linear partial differential equation |
| 14 |
* Create and apply Dirichlet boundary conditions |
| 15 |
* Define Expressions |
| 16 |
* Define a FunctionSpace |
| 17 |
* Create a SubDomain |
| 18 |
|
| 19 |
The solution for :math:`u` in this demo will look as follows: |
| 20 |
|
| 21 |
.. image:: poisson_u.png |
| 22 |
:scale: 75 % |
| 23 |
|
| 24 |
|
| 25 |
Equation and problem definition |
| 26 |
------------------------------- |
| 27 |
|
| 28 |
The Poisson equation is the canonical elliptic partial differential |
| 29 |
equation. For a domain :math:`\Omega \subset \mathbb{R}^n` with |
| 30 |
boundary :math:`\partial \Omega = \Gamma_{D} \cup \Gamma_{N}`, the |
| 31 |
Poisson equation with particular boundary conditions reads: |
| 32 |
|
| 33 |
.. math:: |
| 34 |
- \nabla^{2} u &= f \quad {\rm in} \ \Omega, \\ |
| 35 |
u &= 0 \quad {\rm on} \ \Gamma_{D}, \\ |
| 36 |
\nabla u \cdot n &= g \quad {\rm on} \ \Gamma_{N}. \\ |
| 37 |
|
| 38 |
Here, :math:`f` and :math:`g` are input data and :math:`n` denotes the |
| 39 |
outward directed boundary normal. The most standard variational form |
| 40 |
of Poisson equation reads: find :math:`u \in V` such that |
| 41 |
|
| 42 |
.. math:: |
| 43 |
a(u, v) = L(v) \quad \forall \ v \in V, |
| 44 |
|
| 45 |
where :math:`V` is a suitable function space and |
| 46 |
|
| 47 |
.. math:: |
| 48 |
a(u, v) &= \int_{\Omega} \nabla u \cdot \nabla v \, {\rm d} x, \\ |
| 49 |
L(v) &= \int_{\Omega} f v \, {\rm d} x |
| 50 |
+ \int_{\Gamma_{N}} g v \, {\rm d} s. |
| 51 |
|
| 52 |
The expression :math:`a(u, v)` is the bilinear form and :math:`L(v)` |
| 53 |
is the linear form. It is assumed that all functions in :math:`V` |
| 54 |
satisfy the Dirichlet boundary conditions (:math:`u = 0 \ {\rm on} \ |
| 55 |
\Gamma_{D}`). |
| 56 |
|
| 57 |
In this demo, we shall consider the following definitions of the input |
| 58 |
functions, the domain, and the boundaries: |
| 59 |
|
| 60 |
* :math:`\Omega = [0,1] \times [0,1]` (a unit square) |
| 61 |
* :math:`\Gamma_{D} = \{(0, y) \cup (1, y) \subset \partial \Omega\}` |
| 62 |
(Dirichlet boundary) |
| 63 |
* :math:`\Gamma_{N} = \{(x, 0) \cup (x, 1) \subset \partial \Omega\}` |
| 64 |
(Neumann boundary) |
| 65 |
* :math:`g = \sin(5x)` (normal derivative) |
| 66 |
* :math:`f = 10\exp(-((x - 0.5)^2 + (y - 0.5)^2) / 0.02)` (source |
| 67 |
term) |
| 68 |
|
| 69 |
|
| 70 |
Implementation |
| 71 |
-------------- |
| 72 |
|
| 73 |
This description goes through the implementation (in |
| 74 |
:download:`demo_poisson.py`) of a solver for the above described |
| 75 |
Poisson equation step-by-step. |
| 76 |
|
| 77 |
First, the :py:mod:`dolfin` module is imported: :: |
| 78 |
|
| 79 |
from dolfin import * |
| 80 |
|
| 81 |
We begin by defining a mesh of the domain and a finite element |
| 82 |
function space :math:`V` relative to this mesh. As the unit square is |
| 83 |
a very standard domain, we can use a built-in mesh provided by the |
| 84 |
class :py:class:`UnitSquareMesh <dolfin.cpp.UnitSquareMesh>`. In order |
| 85 |
to create a mesh consisting of 32 x 32 squares with each square |
| 86 |
divided into two triangles, we do as follows :: |
| 87 |
|
| 88 |
# Create mesh and define function space |
| 89 |
mesh = UnitSquareMesh(32, 32) |
| 90 |
V = FunctionSpace(mesh, "Lagrange", 1) |
| 91 |
|
| 92 |
The second argument to :py:class:`FunctionSpace |
| 93 |
<dolfin.functions.functionspace.FunctionSpace>` is the finite element |
| 94 |
family, while the third argument specifies the polynomial |
| 95 |
degree. Thus, in this case, our space ``V`` consists of first-order, |
| 96 |
continuous Lagrange finite element functions (or in order words, |
| 97 |
continuous piecewise linear polynomials). |
| 98 |
|
| 99 |
Next, we want to consider the Dirichlet boundary condition. A simple |
| 100 |
Python function, returning a boolean, can be used to define the |
| 101 |
subdomain for the Dirichlet boundary condition (:math:`\Gamma_D`). The |
| 102 |
function should return ``True`` for those points inside the subdomain |
| 103 |
and ``False`` for the points outside. In our case, we want to say that |
| 104 |
the points :math:`(x, y)` such that :math:`x = 0` or :math:`x = 1` are |
| 105 |
inside on the inside of :math:`\Gamma_D`. (Note that because of |
| 106 |
rounding-off errors, it is often wise to instead specify :math:`x < |
| 107 |
\epsilon` or :math:`x > 1 - \epsilon` where :math:`\epsilon` is a |
| 108 |
small number (such as machine precision).) :: |
| 109 |
|
| 110 |
# Define Dirichlet boundary (x = 0 or x = 1) |
| 111 |
def boundary(x): |
| 112 |
return x[0] < DOLFIN_EPS or x[0] > 1.0 - DOLFIN_EPS |
| 113 |
|
| 114 |
Now, the Dirichlet boundary condition can be created using the class |
| 115 |
:py:class:`DirichletBC <dolfin.fem.bcs.DirichletBC>`. A |
| 116 |
:py:class:`DirichletBC <dolfin.fem.bcs.DirichletBC>` takes three |
| 117 |
arguments: the function space the boundary condition applies to, the |
| 118 |
value of the boundary condition, and the part of the boundary on which |
| 119 |
the condition applies. In our example, the function space is ``V``, |
| 120 |
the value of the boundary condition (0.0) can represented using a |
| 121 |
:py:class:`Constant <dolfin.functions.constant.Constant>` and the |
| 122 |
Dirichlet boundary is defined immediately above. The definition of the |
| 123 |
Dirichlet boundary condition then looks as follows: :: |
| 124 |
|
| 125 |
# Define boundary condition |
| 126 |
u0 = Constant(0.0) |
| 127 |
bc = DirichletBC(V, u0, boundary) |
| 128 |
|
| 129 |
Next, we want to express the variational problem. First, we need to |
| 130 |
specify the trial function :math:`u` and the test function :math:`v`, |
| 131 |
both living in the function space :math:`V`. We do this by defining a |
| 132 |
:py:class:`TrialFunction <dolfin.functions.function.TrialFunction>` |
| 133 |
and a :py:class:`TestFunction |
| 134 |
<dolfin.functions.function.TrialFunction>` on the previously defined |
| 135 |
:py:class:`FunctionSpace |
| 136 |
<dolfin.functions.functionspace.FunctionSpace>` ``V``. |
| 137 |
|
| 138 |
Further, the source :math:`f` and the boundary normal derivative |
| 139 |
:math:`g` are involved in the variational forms, and hence we must |
| 140 |
specify these. Both :math:`f` and :math:`g` are given by simple |
| 141 |
mathematical formulas, and can be easily declared using the |
| 142 |
:py:class:`Expression <dolfin.functions.expression.Expression>` class. |
| 143 |
Note that the strings defining ``f`` and ``g`` use C++ syntax since, |
| 144 |
for efficiency, DOLFIN will generate and compile C++ code for these |
| 145 |
expressions at run-time. |
| 146 |
|
| 147 |
With these ingredients, we can write down the bilinear form ``a`` and |
| 148 |
the linear form ``L`` (using UFL operators). In summary, this reads :: |
| 149 |
|
| 150 |
# Define variational problem |
| 151 |
u = TrialFunction(V) |
| 152 |
v = TestFunction(V) |
| 153 |
f = Expression("10*exp(-(pow(x[0] - 0.5, 2) + pow(x[1] - 0.5, 2)) / 0.02)", degree=2) |
| 154 |
g = Expression("sin(5*x[0])", degree=2) |
| 155 |
a = inner(grad(u), grad(v))*dx |
| 156 |
L = f*v*dx + g*v*ds |
| 157 |
|
| 158 |
Now, we have specified the variational forms and can consider the |
| 159 |
solution of the variational problem. First, we need to define a |
| 160 |
:py:class:`Function <dolfin.functions.function.Function>` ``u`` to |
| 161 |
represent the solution. (Upon initialization, it is simply set to the |
| 162 |
zero function.) A :py:class:`Function |
| 163 |
<dolfin.functions.function.Function>` represents a function living in |
| 164 |
a finite element function space. Next, we can call the :py:func:`solve |
| 165 |
<dolfin.fem.solving.solve>` function with the arguments ``a == L``, |
| 166 |
``u`` and ``bc`` as follows: :: |
| 167 |
|
| 168 |
# Compute solution |
| 169 |
u = Function(V) |
| 170 |
solve(a == L, u, bc) |
| 171 |
|
| 172 |
The function ``u`` will be modified during the call to solve. The |
| 173 |
default settings for solving a variational problem have been |
| 174 |
used. However, the solution process can be controlled in much more |
| 175 |
detail if desired. |
| 176 |
|
| 177 |
A :py:class:`Function <dolfin.functions.function.Function>` can be |
| 178 |
manipulated in various ways, in particular, it can be plotted and |
| 179 |
saved to file. Here, we output the solution to a ``VTK`` file (using |
| 180 |
the suffix ``.pvd``) for later visualization and also plot it using |
| 181 |
the :py:func:`plot <dolfin.common.plot.plot>` command: :: |
| 182 |
|
| 183 |
# Save solution in VTK format |
| 184 |
file = File("poisson.pvd") |
| 185 |
file << u |
| 186 |
|
| 187 |
# Plot solution |
| 188 |
import matplotlib.pyplot as plt |
| 189 |
plot(u) |
| 190 |
plt.show() |