[Pw_forum] pw.x not able to specity initial ion velocities

Konstantin Kudin konstantin_kudin at yahoo.com
Thu Aug 24 01:27:22 CEST 2006


--- Nicola Marzari <marzari at MIT.EDU> wrote:

> 
> A side comment: initializing velocities makes sense for a
> born-oppenheimer simulation, but not for a car-parrinello one.

 Actually, it can make sense for CP as well (obviously, with some help
from the operator!).

> In the latter case, the electronic wavefunctions must acquire a
> dynamics that follows quasi-adiabatically the ionic one. So, one
> wants
> to avoid an abrupt start (i.e. ionic velocities different from zero)
> since this would excite a number of frequencies in the
> electronic evolution (that we do not want, would lower the quality of
> the simulation, and that would take a lot of time to thermalize out).

 This is true, however, thanks to the conjugate gradient (CG) [ which I
think actually originated from Nicola's group :-) ] one can do a couple
of CG steps with proper CP timestep and whatever initial velocities,
and then restart the CP from that. The electronic velocities in this
case are fully consistent with the moving ions.

 Also note that sometime ago I added a CP option to change the timestep
as well, which makes the code rescale the ionic velocities
proportionally such that the temperature is preserved. Again, here one
goes CP(dt1)->CG(dt2, some rescaling)->CP(dt2), and everything is nice,
consistent, and at the same temperature as before :-)

 I use both these features all the time.

 Kostya

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 



More information about the Pw_forum mailing list