Closed luiswirth closed 2 months ago
In commit 3f2549bdf2d6abc6b4dd4d2fb763c44578704f73 this was done. Unfortunatly in dimensions higher than 1D the convergence rate is terrible. Not clear why.
Here we can see that we converge to the wrong solution. It looks like at x=1 a wrong boundary condition is being enforced.
As suspected there was a bug in the boundary condition enforcement. It is now fixed.
The 2D solution now correctly converges with order 2 in L2 norm as predicted by theory. 🥳
convergence in 3D is however still terrible... Unclear why.
----------------------------------------------
| k | nsubdivisions | error | conv_rate |
----------------------------------------------
| 0 | 1 | 0.000e0 | inf |
| 1 | 2 | 4.078e-6 | -inf |
| 2 | 4 | 3.781e-1 | -16.50 |
| 3 | 8 | 3.231e-1 | 0.23 |
| 4 | 16 | 2.966e-1 | 0.12 |
| 5 | 32 | 2.776e-1 | 0.10 |
----------------------------------------------
Beginning at 3 subdivisions the solution in 3D is completly spurious.
closed in favor of #28
I should convert the poisson{1d,2d} examples into a poisson example that is generic over the dimension.