ABINIT Discussion Forums

Posted: **Mon Dec 12, 2011 2:27 am**

I have calculated phonon dispersion for tungsten carbide (WC), which is of 187# space group. The calculation was done on 3x3x3 q-mesh (6 q-points) with 9x9x9 k-mesh. nshiftk = 1 and shiftk = (0, 0, 0.5).
After calculation through such processes:

Code: Select all

# Do the SCF calculation
getwfk1  0
kptopt1  1
nqpt1  0
tolvrs1  1.0d-19
rfphon1  0

# Do the d/dk calculation
iscf2  -3
kptopt2  2
rfphon2  0
rfelfd2  2
tolwfr2  1.0d-22

# Gamma point phonon
getddk3  2
kptopt3  2
rfelfd3  3

# Other q-points
getwfk  1
kptopt  3
rfphon  1
rfatpol  1  2
rfdir  1  1  1
tolvrs  1.0d-9

However, when I try to use MRGDDB to sum the 6 DDB files, the error occurred:

compare the current and input DDB information
chki8 : ERROR -
Comparing integers for variable nkpt.
Value from input DDB is 365 and
from transfer DDB is 729.

I checked my DDB files, and found that the DDB for Gamma point contented 365 kpoints. But the DDBs for other q-points contented 729 kpoints. According to the source code of chki8.f90:

Code: Select all

 if(inti/=intt) then
  write(message, '(a,a,a,a,a,a,a,i10,a,a,a,i10,a,a,a)' )&
&  ' chki8 : ERROR -',ch10,&
&  '  Comparing integers for variable',name,'.',ch10,&
&  '  Value from input DDB is',inti,' and',ch10,&
&  '        from transfer DDB is',intt,'.',ch10,&
&  '  Action : check your DDBs.'

This must be the reason.
But, I do not know why such a problem occurred, and whether there are some methods to avoid this problem.
Any suggestions and replies will be welcome.
Thanks.

Posted: **Mon Dec 12, 2011 9:18 am**

Dear Zhe,

I think that this condition is there to force you to use nsym 1 in all your calculation. The reason is that symmetries are not (?) yet implemented in the response part of abinit.

Best regards,

Samuel

Posted: **Sat Dec 17, 2011 8:09 am**

sponce wrote:Dear Zhe,

I think that this condition is there to force you to use nsym 1 in all your calculation. The reason is that symmetries are not (?) yet implemented in the response part of abinit.

Best regards,

Samuel

Thank you for your reply.
I have tried the same system with nsym=1. But the same problem happened as well:

ddbvrs= 100401
compare the current and input DDB information
chki8 : ERROR -
chki8 : ERROR -
Comparing integers for variable nkpt.
Comparing integers for variable nkpt.
Value from input DDB is 14 and
Value from input DDB is 14 and
from transfer DDB is 27.
from transfer DDB is 27.
Action : check your DDBs.
Action : check your DDBs.

However, when I calculated FCC structure ( as the official example showing for AlAs ), this process were without problem even the nkpt for Gamma point was also different with other q-points.

Posted: **Mon Dec 19, 2011 5:12 am**

Well~ I solve the problem by myself~
At first, I tested different k-mesh and q-mesh, and found 4x4x4 k + 4x4x4 q was the best one to pass anaddb program. The different k-mesh and q mesh cannot pass and with error about DDB nkpt problem. However, 3x3x3 k + 3x3x3 q cannot pass, too.
Since I am using Quantum-ESPRESSO as well, I know that 4x4x4 q-mesh is normally enough for phonon calculation, but 4x4x4 k-mesh is too rough to obtain good result. So I have to calculate the phonon on density k-mesh with 4x4x4 q-mesh.
Finally, I checked the successful *.ddb.out file, and found the function of mrgddb is just combinating the DDB files in some rule. Then I combided all of DDBs by hand~ it finally passed anaddb.

ABINIT Discussion Forums

[SOLVED] The different nkpt values in DDBs

[SOLVED] The different nkpt values in DDBs

Re: The different nkpt values in DDBs

Re: The different nkpt values in DDBs

Re: The different nkpt values in DDBs