[Pw_forum] Constrained VC relaxation

[Serge Nakhmanson]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Dear Ben and Paolo,

I don't know the exact answer either, but here are my two cents. Below
is a copy of my e-mail to Carlo and his reply. This was a private
(i.e., not via forum) communication. I asked Carlo about a possibility
of simulating pseudomorphic growth in a single run. Carlo's general
answer was that it is not possible but you can use certain tricks to
speed up your fixed volume calculations.

On a personal note, I would add that I care greatly about vc-relaxations
of both pseudomorphic (constrained) and "general" type. I'm currently
doing some vc-relaxations of polymer crystals with PWscf 2.1.2 and
with Paolo's help they start to make sence. I think that such calculations
will be a good test for PWscf vc-relax capabilities and fine tuning
because these systems are very anisotropic (floppy in x and y, and rigid
in z directions). I'm not sure, though, that I'm quite ready to hack into
the code just yet (which is a question of available time and not lack of
desire).

WBW,

Serge

> Dear Serge,
> 
> the procedure of restricting the forces to the [001] direction will not work because
> the cell (which is moved according to the stress tensor) will still be free to deform.
> I think that the only thing you can do is to perform several (6-8 points should be
> enough) structural relaxations for different values of c.
> 
> If you do this with the BFGS, consider that the approximate inverse hessian matrix
> obtained in the previous relaxation can be used as starting guess for the new one. This
> should considerably speed-up the convergence. To do this you have to go in the outdir.
> There you will find a file "old_prefix.hess_out" (old_prefix is the prefix of the old
> relaxation). Change it to "new_prefix.hess_in" (new_prefix is the prefix of the new
> relaxation). The program will automatically look for this file in the outdir and will
> read it if present.
> 
> Good luck,
> 
> carlo
> 
> Serge Nakhmanson wrote:
> 
>> -----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
>> 
>> Dear Carlo,
>> 
>> I apologize for contacting you directly, but I have a small question that is probably
>> not interesting to the rest of the forum members and I hope you can help me out. If
>> you remember I was doing some calculations of pseudomorphically grown films when the
>> "in plane" lattice const (a) of my (tetragonal) system are fixed and the [001]
>> lattice const (c) needs to be relaxed. I was doing a series of runs with various
>> fixed c and then using equations of state to find the c corresponding to the minimum
>> energy. This technique is precise but requires a lot of work, which is OK for smaller
>> systems and a real pain for large ones. I'm wondering if in the latter case I could
>> do a single "vc-relax" simulation? If in my input file I restrict my forces to [001]
>> direction only (by placing 0 0 1 after x y z coords of every atom) will this
>> automatically constrain the x and y dimensions of my cell during the variable-cell
>> relaxation? I would really appreciate your advice here. Of course, if you know a
>> better way to do what I need to do, I would be very thankful if you show me how to do
>> this.
>> 
>> Thanks,
>> 
>> Serge



Paolo Giannozzi wrote:
> On Monday 31 January 2005 13:34, Ben Hope wrote:
> 
> 
>> Does the lack of replies indicate that this is not possible with pwscf?
> 
> 
> it indicates that nobody knows the answer, or that nobody who knows the answer is
> taking the time to answer. I have never heard that constrained variable-cell relaxation
> is implemented, but this is more or less all I know.
> 
> 
>> This would seem to be quite a significant disadvantage for the code when using it to
>> study 1-dimensional systems.
> 
> 
> if you implement it, the disadvantage will be removed
> 
> Paolo
> 

- --
******************************************************************************
 Serge M. Nakhmanson                                 phone: +1 (732) 445-4603
 Postdoctoral RA                                       fax: +1 (732) 445-4400
 Department of Physics & Astronomy
 Rutgers, The State University of New Jersey
 136 Frelinghuysen Road                          nakhmans at physics.rutgers.edu
 Piscataway, NJ 08854-8019          http://www.physics.rutgers.edu/~nakhmans/

 My GnuPG public key is here:
 http://www.physics.rutgers.edu/~nakhmans/PGP/nakhmans_pubkey_current.asc
******************************************************************************
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.1 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFB/nNSS2Ygfv2TcL8RAj/vAJ4zWMwTbt4W3AoU0wEkkuT+6td7XACeNBil
Tyh0Hp7oSRp8MwYtW012LeI=
=+WEI
-----END PGP SIGNATURE-----