[CP2K:902] Re: Variable cell optimization

[Nichols A. Romero]

Rad,

If you are planning to wait for the fluctuations to go away or to reach some
steady state you are always going to have to wait a while. In my opinion,
the only thing that matters is the mean pressure and the error in the mean.
That
having been said, you need to look at
1. mean pressure, make sure it is not drifting significantly
2. calculate the error in the mean taking into the autocorrelation

BTW, the pressure is only well-defined for an extended system in CP2K (as
far as I know).

On Thu, Mar 27, 2008 at 9:07 AM, Rad <rad.... at arl.army.mil> wrote:

>
> Nick,
>
> I have done cp2k calculations with pressure fluctuation of few bars
> (well after 4000 steps in an NPT ensemble). All my diamond
> calculations had a fluctuation of few kbars. This one I am trying is a
> molecular system and I am hoping
> to have fluctuations within few kbars. If not I will have to relax my
> convergence criteria. The volume of the cell also fluctuates and I am
> not sure it is because of the fluctuation in the pressure or due to
> other factors such as not taking
> into account dispersion in molecular crystals. I am hoping that the
> molcular dynamics based cell optimization will give better resulsts
> and my calculations are waiting in the queue.
>
> Rad
>
> On Mar 26, 5:16 pm, "Nichols A. Romero" <naro... at gmail.com> wrote:
> > Rad,
> >
> > For a material like diamond, a fluctuation of 50 kbar should be expected
> > (even larger). That's only 5 GPa.
> >
> > My experience with DFT is that converging the pressure to less than 1
> kbar
> > is unrealistic. So worrying about
> > bars is a waste of time. The story is probably different for interatomic
> > potential-based calculations. I cannot
> > comment on those.
> >
> > On Wed, Mar 26, 2008 at 4:36 PM, Teodoro Laino <teodor... at gmail.com>
> > wrote:
> >
> >
> >
> >
> >
> >
> >
> > > Rad,
> >
> > > On 26 Mar 2008, at 21:20, Rad wrote:
> >
> > > > Teo,
> >
> > > > Couple of feedback: the volume is printed in Angstrom in some places
> > > > and bohr in other places. Is it possible to sick to one unit (just
> > > > bohr) so that post processing scripts people set up can stay as they
> > > > are?
> > > the .cell file is always in ANGSTROM. the output on the screen is
> > > according the unit that you
> > > specified in your input file (if you use angstrom you will get
> > > angstrom, if you use bohr you will get bohr)..
> > > Use the .cell file for post-processing if you don't like the idea
> > > that the output depends on the unit that
> > > you decide to use.
> >
> > > > The output produces a bunch of restart files for wavefunction and
> > > > Restart*restart. (this is for the case of GEO_OPT option). When I
> > > > restart the calculation which one I should use?
> > > The files generated during the geometry optimization are tagged *-
> > > GEO_OPT-*..
> > > you should also have a normal restart file without the -GEO_OPT-..
> > > use that one!
> > > In case you do MD instead of GEO_OPT you will find MD.
> >
> > > > As usual convergence is always a tricky one. I am trying small
> systems
> > > > and having issues with pressure fluctuating wildly, starting with
> few
> > > > bars and reaching 50 kbars withing few steps. I will do few more
> runs
> > > > (QS, GPW) and get back to you on this. The second item is the one I
> > > > need a response before restarting my calculations.
> > > Rad, for material that are quite hard of course small variations of
> > > the cell have large fluctuations in the pressure.
> > > You can play with the bunch of parameters of the CELL_OPT section
> > > (specifically with the LS 2PNT for reducing
> > > the max_allowed_step) to get a safer optimization..
> >
> > > Teo
> >
> > > > Thanks a lot for this most anticipated module for our team.
> >
> > > > Rad
> >
> > > > On Mar 25, 9:03 am, Rad <rad.... at arl.army.mil> wrote:
> > > >> Thanks Teo,
> >
> > > >> I just got back from my vacation and it is a great news. I will
> > > >> get to
> > > >> it immediately as I am working on several systems for which this
> > > >> method would be ideal.
> >
> > > >> Rad
> >
> > > >> On Mar 20, 8:50 pm, Teodoro Laino <teodor... at gmail.com> wrote:
> >
> > > >>> Hi Rad,
> >
> > > >>> just a follow up about this message.
> > > >>> Me and Paolo Raiteri implemented the cell optimization algorithm
> in
> > > >>> cp2k both at 0K (optimizing the atomic positions) and at finite
> > > >>> temperature (averaging the stress tensor over an MD run). There
> > > >>> are a
> > > >>> couple of regtest added in Fist/regtest-12/argon* showing how to
> use
> > > >>> the cell optimization algorithm. At the moment the only scheme
> > > >>> available is the CG with a 2PNT line search (quite efficient in
> the
> > > >>> systems
> > > >>> I tested).
> > > >>> Comments and suggestion are, as usual, welcome.
> >
> > > >>> Ciao
> > > >>> Teo
> >
> > > >>> On 17 Oct 2007, at 21:51, Rad wrote:
> >
> > > >>>> Hi everybody,
> >
> > > >>>> Is it possible to do variable cell optimization using CP2K? Can
> > > >>>> somebody provide me with a sample input section for that?
> >
> > > >>>> Thanks
> >
> > > >>>> Rad- Hide quoted text -
> >
> > > >>> - Show quoted text -- Hide quoted text -
> >
> > > >> - Show quoted text -
> >
> > --
> > Nichols A. Romero, Ph.D.
> > DoD User Productivity Enhancement and Technology Transfer (PET) Group
> > High Performance Technologies, Inc.
> > Reston, VA
> > 443-567-8328 (C)
> > 410-278-2692 (O)- Hide quoted text -
> >
> > - Show quoted text -
> >
>


-- 
Nichols A. Romero, Ph.D.
DoD User Productivity Enhancement and Technology Transfer (PET) Group
High Performance Technologies, Inc.
Reston, VA
443-567-8328 (C)
410-278-2692 (O)
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20080327/b60a111d/attachment.htm>