Creating a regular Voronoi tesselation about neper

[rquey]

Error : . Could not process '-n from_morpho'.

This comes from the fact that -n from_morpho determines n from -morpho and not -morphooptiini. Just provide the number of seeds in -n...

number of common seeds = 4 != 3
seeds = 7 -5 27 4

This comes from the fact that the seeds are regularly arranged on a grid, leading to a degenerate configuration where more than 4 cells intersect at a vertex. (This only happens for the discrete events where seeds are regularly arranged. This will never happen when there is some level of randomness in the seed coordinates.)

A workaround is to "randomize" the seed coordinates:

gsl-randist 1 32 flat 0 0.00001 > x
gsl-randist 2 32 flat 0 0.00001 > y
gsl-randist 3 32 flat 0 0.00001 > z
paste POSCAR-00002 x y z | awk '{print $1+$4,$2+$5,$3+$6}' > pos

and then

neper -T -n 32 -morphooptiini "coo:file(pos)" -o n32 -per 1
neper -V n32.tess -datacellcol id -datacelltrs 0.5 -dataedgerad 0.001 -cameracoo 8:6:4 -print n32

Still, a possible issue is that the cells will contain tiny edges and faces coming from the randomized seed coordinates, which you can see like this:

neper -T -n 32 -morphooptiini "coo:file(pos)" -o n32 -per 1 -statedge length -statface area
head n32.stedge
head n32.stface

This could be fixed using regularization (-reg 1), but it is not available yet for periodic tessellations.

Some regular tessellations are directly supported by Neper: cubes and truncated octahedra. If I'm correct, you are aiming for truncated dodecahedra here, which are not included.

[mbaeker]

Thanks a lot for the quick answer.
Yes, basically I am trying to create the Wigner-Seitz-cells in an fcc crystal (rhombic dodecahedron)

The tesselation does indeed work after adding some randomness, however after
neper -T -n 32 -morphooptiini "coo:file(POSCAR-00002)" -o n32 -per 1
meshing with
neper -M n32.tess -format 'inp'
fails:

Info : Loading input data...
Info : - Loading tessellation...
Info : [i] Parsing file n32.tess'... Info : [i] Parsed file n32.tess'.
Info : Meshing...
Info : - Preparing... (cl = 0.1575) 100%
Info : - 0D meshing... 100%
Info : - 1D meshing... 100%
Info : - 2D meshing... 100% (0|0.2/99%| 1%| 0%)
Info : - Fixing 2D-mesh pinches...
Info : - 3D meshing... 3% (0|0/ 0%|100%| 0%)Error : > Meshing of poly 1 (1) failed

I've tried playing around with the meshing algorithm, but so far without success.
I also tried to change the element size with -cl, but this ends with a core dump.

It works when I use hex elements, which is probably sufficient for what I need, but if you have any hints how I could create the tet mesh, that would of course be great.
Thanks again

[rquey]

Meshing struggles with the tiny edges and faces. You can increase the "randomization" distance:

gsl-randist 1 32 flat 0 0.01 > x
gsl-randist 2 32 flat 0 0.01 > y
gsl-randist 3 32 flat 0 0.01 > z
paste POSCAR-00002 x y z | awk '{print $1+$4,$2+$5,$3+$6}' > pos

and then

neper -T -n 32 -morphooptiini "coo:file(pos)" -o n32 -per 1
neper -M n32.tess -pl 10000 -rcl 0.5 -for msh,inp
neper -V n32.tess,n32.msh -dataelsetcol id -dataelt1drad 0.0025 -dataelt3dedgerad 0.001 -showelt1d all -cameracoo 8:6:4 -print n32m

You will notice that the geometry differs slightly from theory.

[mbaeker]

Thanks again, that's very helpful.