Fwd: [Pw_forum] constraints

[carlo sbraccia]

Dear Guido,

> Subject: [Pw_forum] constraints
> Date: Friday 19 May 2006 10:46
> From: Guido Roma <roma at srmp.saclay.cea.fr>
> To: pw_forum at pwscf.org
>
> Dear All,
>
> I have been using contrained relaxations with older version of PWSCF,
> I've seen that in version 3.1 many things are changed in that respect.
> The module coping with constraints seems to allow only certain types of
> constraints (distances, angles), not center of mass type constraints for
> examples, and is not any more so easy to write one's own constrain.f90
> subroutine.
It is true that the module implementing constrained MD and relaxations 
presently allows for few standard constraint-types only. It is also true 
that one shouldn't restrict the code to only those standard types. For 
this reason the module implementing SHAKE has been designed in order to 
be flexible and easily extensible. What you need is a piece of code 
implementing the constraint and its gradient. The procedure to port the 
code to QE is described step-by-step in the file 
Doc/constraints_HOWTO.tex. Of course, if you have a constraint-type that 
is not problem-specific and if you think it can be useful to other 
users, I'll be pleased to port it to QE.
> But, further more, I don't understand if the allowed
> constraints as specified in the CONSTRAINTS input card are working or
> not. Looking at the code I have the impression that they should work if
> relaxing by damped dynamics (ion_dynamics='damp'), but not with bfgs.
> Can someone confirm this or tell me if and where I'm wrong?
>   
As far as I know the constraints as specified in the CONSTRAINTS input 
card work fine. If you have any indication of the contrary, please be 
more specific.

Constrained relaxation is not allowed for Newton-Raphson relaxation 
scheme (BFGS) because of technical problems (if you want I can be more 
specific about the technical problems). Only 
preconditioned-damped-dynamics is available for constrained relaxations, 
but this is not a limitation since it doesn't do much worse than BFGS.
> Any don't see any hint in the user-guide.
>   
Maybe the documentation is not very detailed, but you can find a hint on 
how to perform a constrained MD calculation in the file Doc/INPUT_PW:

ion_dynamics   CHARACTER
                          specify the type of ionic dynamics.
                          For constrained dynamics or constrained 
optimisations add the
                          CONSTRAINTS card.
                          ....

Regards,

carlo