[Pw_forum] Calculating modulated structures with PWscf

[Paolo Giannozzi]

-------- Forwarded Message --------
From: Herman van Midden <Herman.vanMidden at ijs.si>
To: mailman-owner at democritos.it
Subject: Calculating modulated structures wit PWscf
Date: Thu, 19 Dec 2013 15:09:16 +0100

Hi
 
I am rather new in the field of PWscf calculations and I am hoping
someone can help me to answer some, maybe obvious questions.
 
I would like to perform total energy calculations on structures
modulated with a change density wave (CDW).
 
These CDW's usually extend over several unit cells. So my approach is to
first calculate the equilibrium structure for one unit cell, repeat
these unit cell in the desired direction and modulate it with the CDW.
If my model gives a good description of the material, the total energy
of the modulated structure should be lower than that of the unmodulated
structure.
 
The problem of this approach is that the repeated unmodulated structure
contains a large number of atoms. How can I calculate the total energy
of this unmodulated structure? 
I expected I would be able to use the total energy of one unit cell,
calculated with N times more k-points in the direction where I am going
to extend the unit cell N-times. Repeating the unit cell N times in real
space, should be equivalent to reducing the k point sampling by a factor
N in reciprocal space. Or Not?
This approach worked for some potentials, but now I am running into
problems. I am currently using the GBRV potentials which use ecutwfc=40
and ectrho=200 Ry. Using this approach results in a total energy which
is larger than N times the unit cell energy, and subsequent vc-relax
steps decrease the energy. So it is not the desired unmodulated
equilibrium structure.
The calculations also take much longer than I expected. I hoped that
simultaneously increasing the number of atoms and reducing the number of
k-points would result in a comparable computer load. When I compare the
outputs I noticed that the fft mesh  increases with the number of atoms.
Can I reduce n1,n2 or n3 to optimize calculation speed? Or will this
significantly reduce the reliability of the calculation?
 
Does anyone have any suggestions how I can perform total energy
calculations on CDW modulated structures.
 
Any help is welcome
 
Herman van Midden
Institute Jozef Stefan
Ljubljana Slovenia

-- 
 Paolo Giannozzi, Dept. Chemistry&Physics&Environment, 
 Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
 Phone +39-0432-558216, fax +39-0432-558222