kmesh_get_bvector: Not enough bvectors found

[eliephys78]

Dear all,

It seems I manage to resolve one error but get into another. Th error I am getting is related to the W90 calculations I believe:

*---------------------------------- K-MESH ----------------------------------*
+----------------------------------------------------------------------------+
| Distance to Nearest-Neighbour Shells |
| ------------------------------------ |
| Shell Distance (Ang^-1) Multiplicity |
| ----- ----------------- ------------ |
| 1 0.051068 2 |
| 2 0.102137 2 |
| 3 0.153205 2 |
| 4 0.204273 2 |
| 5 0.255341 2 |
| 6 0.306410 2 |
| 7 0.357478 2 |
| 8 0.383296 6 |
| 9 0.386683 12 |
| 10 0.396671 12 |
| 11 0.408546 2 |
| 12 0.412780 12 |
+----------------------------------------------------------------------------+
| The b-vectors are chosen automatically |
Exiting.......
kmesh_get_bvector: Not enough bvectors found.

I have looked on the internet where the error could be easily solved by lowering kmesh_tol in the win.in file. However since EPW uses W90 indirectly to produce the wannier functions, this "option" is not present in the epw.in input. i tried to force it in the win file that the code automatically produces but in vain since the code overwrites the win file. Is there any way to overcome this problem in EPW?

thanks and sorry to bombard you with too many questions each now and then

regards

[giustino]

Hi there,

In order to pass additional runtime options to Wannier90 from within EPW you can specify the
array

Code: Select all

wdata(:)

, please take a look at the example in the page describing the
input variables:

http://epw.org.uk/Documentation/Inputs#wdata

FG

[eliephys78]

Thanks for your reply.

However I still have a problem I hope you can help me with. I calculated the phonon frequencies and obtained the dyn and dvscf files by applying ph.x using ldisp=.true. These points at which the frequencies were calculated are:

28 cartesian
0.0000000 0.0000000 0.0000000 0.0011574
0.0000000 0.0000000 0.0256410 0.0023148
0.0000000 0.0000000 0.0512821 0.0023148
0.0000000 0.0000000 -0.0769231 0.0011574
0.0000000 0.1924501 0.0000000 0.0069444
0.0000000 0.1924501 0.0256410 0.0138889
0.0000000 0.1924501 0.0512821 0.0138889
0.0000000 0.1924501 -0.0769231 0.0069444
0.0000000 0.3849002 0.0000000 0.0069444
0.0000000 0.3849002 0.0256410 0.0138889
0.0000000 0.3849002 0.0512821 0.0138889
0.0000000 0.3849002 -0.0769231 0.0069444
0.0000000 -0.5773503 0.0000000 0.0034722
0.0000000 -0.5773503 0.0256410 0.0069444
0.0000000 -0.5773503 0.0512821 0.0069444
0.0000000 -0.5773503 -0.0769231 0.0034722
0.1666667 0.2886751 0.0000000 0.0069444
0.1666667 0.2886751 0.0256410 0.0138889
0.1666667 0.2886751 0.0512821 0.0138889
0.1666667 0.2886751 -0.0769231 0.0069444
0.1666667 0.4811252 0.0000000 0.0138889
0.1666667 0.4811252 0.0256410 0.0277778
0.1666667 0.4811252 0.0512821 0.0277778
0.1666667 0.4811252 -0.0769231 0.0138889
0.3333333 0.5773503 0.0000000 0.0023148
0.3333333 0.5773503 0.0256410 0.0046296
0.3333333 0.5773503 0.0512821 0.0046296
0.3333333 0.5773503 -0.0769231 0.0023148
The problem I am having lies in the epw part. I have done a scf calculation and an nscf calculation on a 15x15x15 0 0 0 grid which in this case reduces to 216 points due to symmetry. Running the epw.x executable produces erros like "kmesh_get_bvector". i have tried increasing the number of shells and reducing and increasing kmesh_tol but in vain. I tried producing the nscf points using kpoints.x as instructed in README but got problems like "kpoint not symmetric!". I am not sure what to do. i know that kmesh.pl works fine but in this case I will get into another problem like "q point not commensurate" and i spent three weeks to calculate the 28 phonons. Any advice of how to overcome this "kmesh_get_bvector" error. Am i doing something wrong?

Any help would be really appreciated.

thanks

[sponce]

Dear eliephys78,

You should have the same ph and nscf grid.

If you want to do a 15x15x15 scf+nscf calculations, you should then do a 15x15x15 q-point ph.x grid.
You should therefore have

Code: Select all

216 cartesian
...

in your epw input file.

I would however keep the 28 (maybe a 6x6x6 q-grid? ) and do a 15x15x15 k-grid scf and a 6x6x6 nscf calculation.

The nscf calculation should contain 216 (for a 6x6x6 grid) points that should all be [0,1] positive.

Best,

Samuel

[eliephys78]

I actually used a grid of 12 12 12 0 0 0 for scf
6 6 6 0 0 0 (=216 ) for nscf

and the frequencies were calculated in the first part( 6x6x6)=28 points. However the total number of equivalent q points summed up to 144 (not 216!) and the code stopped with an error:

"' error : elphon_shuffle_wrap' nqc ne nq1 nq2 nq3". What has gone wrong this time?

Thank you for your help and patience.

[sponce]

Hello,

Can you make sure that you are providing the correct list of q-points to EPW.

It should be the same as the one of the ouptut of the phonon calculation. It should look like:

Code: Select all

     Dynamical matrices for ( 3, 3, 3)  uniform grid of q-points
     (   4q-points):
       N         xq(1)         xq(2)         xq(3)
       1   0.000000000   0.000000000   0.000000000
       2  -0.333333333   0.333333333  -0.333333333
       3   0.000000000   0.666666667   0.000000000
       4   0.666666667  -0.000000000   0.666666667

but with your 28 q-points.

Best,

Samuel