Smooth-iso (Elliptic)

We show that it is a very bad idea to approximate smooth solutions using low-order elements.

Model problem

Equation solved: Poisson equation

(1)-\Delta u - f = 0.

Domain of interest: Square (0, \pi)^2.

Right-hand side:

(2)f(x, y) = 2\sin(x)\sin(y).

Boundary conditions: Zero Dirichlet.

Exact solution

(3)u(x, y) = \sin(x)\sin(y).

Sample solution

Solution.

Convergence comparisons

Below we show meshes obtained using various types of adaptivity. Note the tremendous differences in their performance. The meshes do not correspond to the same level of accuracy since the low-order methods could not achieve the same error as hp-FEM. Therefore, compare not only the number of DOF but also the error level. Convergence graphs for all cases are shown at the end of this section.

Final mesh (h-FEM, p=1): 27469 DOF, error 0.39173795799476 %

Final mesh

Final mesh (h-FEM, p=2): 39185 DOF, error 0.0022127484879974 %

Final mesh

Final mesh (hp-FEM): 49 DOF, error 4.2775412425017e-05 %

Final mesh

DOF convergence graphs:

DOF convergence graph.

CPU time convergence graphs:

CPU convergence graph.

Table Of Contents

Previous topic

Benchmarks (Problems with Known Exact Solutions)

Next topic

Smooth-aniso-x (Elliptic)