step-6. Adaptive Refinement and Hanging nodes. On hyper cube mesh looks like hanging nodes are not constrained.

[Oleg Kmechak]

Hello everyone, 

I found such behaviour on step-6: if I use hyper_ball triangulation as it is by default in step-6, then solution is ok and continuosus: 

But if I change triangulation to hyper_cube, it looks like some hanging nodes are not constraied. Or it is just Paraview artefact?

Best,

Oleg Kmechak

[Oleg Kmechak]

UPD: Looks like bug, AFAIK paraview has some strange visualization of quadrangle if their nodes are not on same surface, but I tried some processing of results: contour plot and it is discontinuous:

Best, 

Oleg

[Wolfgang Bangerth]

On 3/6/20 11:55 AM, Oleg Kmechak wrote:
>
> But if I change triangulation to hyper_cube, it looks like some hanging
> nodes are not constraied. Or it is just Paraview artefact?

It's an artifact. The solution is piecewise quadratic, but what you
output (and what paraview visualizes for you) is a piecewise linear
interpolation of the solution.

You might want to read the documentation of the DataOut::build_patches()
function for more information about how you can get closer to seeing
what actually happens in the program.

Best
W.

--
------------------------------------------------------------------------
Wolfgang Bangerth email: bang...@colostate.edu
www: http://www.math.colostate.edu/~bangerth/