[Pw_forum] nr fixed in q_gen subroutine : give wrong total_weight

[Antoine JAY]

Dear Paolo Giannozzi and community,

This 4*4*4 is important because it modifies the "total_weight" variable.

Example:
I did an interpolation  of a 5^3 q-mesh to calculate the phonon DOS.
I get the error message:

frc_blk, wrong total_weight

because the condition
ABS(total_weight-nr1*nr2*nr3).GT.1.0d-8)   of matdyn.f90/subroutine frc_blk
is not fulfilled.
in frc_blk, nr1, nr2 and nr3 have been changed through read_ifc_param and
are each 5 as asked in ph.in.

But the total_weight was 124.1 and not 125.
This is due to the fact that total_weight is calculated through variables
that are in the subroutine q_gen which uses fixed nr:
INTEGER, PARAMETER:: nr1=4, nr2=4, nr3=4

When I change in q_gen
INTEGER, PARAMETER:: nr1=5, nr2=5, nr3=5

the total_weight changes and is exactly 125
and the condition was fulfilled so that the phonon DOS run without any error.

What hypothesis did you do to keep these variables fixed?
Am I doing something wrong?


Antoine JAY






I think 4x4x4 is not directly related to the q-point grid, but to the
maximum size of the supercell (if you want to build a supercell with force
constants computed in the unit cell. It was used once upon a time in
GaAs/AlAs heterostructure, I think)


Paolo

On Tue, Nov 15, 2016 at 9:00 AM, Antoine JAY <antoine.... at polytechnique.edu>
wrote:

> Dear community and developers,
>
> Looking at the code of qe_6.0
>
> I am wondering why in PHonon/PH/matdyn.f90
> subroutine q_gen
>
> the values nr1, nr2 and nr4 are fixed to be 4:
> INTEGER, PARAMETER:: nr1=4, nr2=4, nr3=4
>
> whereas they should be fixed as nr1=nq1, nr2=nq2, and nr3=nq3.
> as done by the first call of read_ifc_param(nr1, nr2, nr3).
>
> Is there a physical meaning?
> Is this subroutine right if someone don't whant a 4^3 q-mesh?
>
> Antoine Jay