Failure in kmesh_get for dense k-grid

[lwoncheol]

Dear Wannier90 developers and users,

I've been doing Wannier calculations for monolayer h-BN. It went well up to 36x36x1 k-grid, but as I increase the k-mesh to 48x48x1, it gives the following errors:

Exiting....... kmesh_get: something wrong, found too many nearest neighbours

Changing kmesh_tol or search_shells didn't work. I attached the input and output files below.

BN.win.txt
BN.wout.txt

[jryates]

The challenge here is caused because your 'c' lattice parameter is an integer multiple of your 'a' lattice parameter. That has produced an accidental degeneracy in the k-point shell. There are two options:

1. Given the 'c' is arbitrary, set it to something slightly different (40 Bohr for example).
2. You would need to specify the b-vectors by hand. Instructions for how to do this (for version 3.1 and earlier) are given below

You will see that the 12th k-point shell has a degeneracy:
| Shell: 12 Multiplicity: 8 | | b-vector 1 : 0.156829 0.271635 0.000000 (Ang^-1) | | b-vector 2 : 0.000000 0.000000 0.313657 (Ang^-1) | | b-vector 3 : 0.000000 0.000000 -0.313657 (Ang^-1) | | b-vector 4 : 0.156829 -0.271635 0.000000 (Ang^-1) | | b-vector 5 : 0.313657 0.000000 0.000000 (Ang^-1) | | b-vector 6 : -0.313657 0.000000 0.000000 (Ang^-1) | | b-vector 7 : -0.156829 0.271635 0.000000 (Ang^-1) | | b-vector 8 : -0.156829 -0.271635 0.000000 (Ang^-1) |
You need the 2nd and 3rd of these b-vectors.

From the list further down the file you will see that these are the 92 and 93 vectors. So create a file called BN.kmesh which looks like
1 2 3 4 5 6 92 93
and set devel_flag=kmesh_degen

And then it works ok!

Note: for @JeromeCCP9 this won't work in v4 due to the fact we removed devel_code. For this reason I will leave the issue open for the moment.