ABINIT Discussion Forums

Dear all,

To my knowledge, I didn't see this problem reported before:
When doing a DFPT phonon calculation in a cubic supercell (here a 2x2x2 cubic perovskite) there is an error message in mrgddb and anaddb saying that the maximum number of symmetry is 192 while in the supercell it is 384 (which is normal: 192*2 with tnons).

Isn't it possible to analyse the phonons with anaddb in "cubic" supercell?
My guess is that we can increase the maximum number of nsym when chkprim = 0 since I already did supercell calculations with non-cubic space group giving rise at the end to a total number of nsym smaller than 192 and it worked!

So the test of maximum nsym = 192 has to be considered differently when chkprim = 0 or did I miss something?

What do you think?

Best,
Eric

The error message:

inprep8 : An error has been found in the positive n-integers contained in the DDB :
Expected Found
natom , larger than 0 natom = 40
nkpt , larger than 0 nkpt = 32
nsppol, either 1 or 2 nsppol = 1
nsym , lower than 192 nsym = 384
ntypat , larger than 0 ntypat = 3
occopt, equal to 0,1 or 2 occopt = 1
inprep8 : ERROR -
See the error message above.

leave_new : decision taken to exit ...

OK guys,

I come back a bit later on this small problem and it's easy to solve:
The maximum number of symmetry (msym in mrgddb.F90 and anaddb.F90) have to be set up to a larger number.
It works fine when done.
So this is something that have to be changed in future versions when using super-cell (chkprim = 0).

Best,

Eric

Hi Eric,

In previous versions, abinit would indeed complain about the number of symmetries being greater than msym. So one had to change it by hand in the source code and compile again.

But it seemed to me that this problem had been solved a few versions ago. I guess they didn't change it in mrgddb and anaddb subroutines.

Boris

Indeed, it is not longer fixed in abinit, but:

../src/98_main/anaddb.F90: integer,parameter :: msym=192
../src/98_main/mrgddb.F90: integer,parameter :: mddb=5000,msym=192,ddbun=2

We'll see. There might be a simple way to lift this limitation, anaddb will only use the symops, never have to generate or check them...

newsp (not used at all) is even worse:
../src/98_main/newsp.F90: integer,parameter :: msppol=2,msym=48

ciao

Matthieu