[CPMD-list] Energy components in PIMD

[Axel Kohlmeyer]

On Mon, 12 May 2008, ukasz Walewski wrote:

LW> Dnia 2008-05-09, piÿÿ o godzinie 10:34 -0400, Axel Kohlmeyer pisze:
LW> > lukasz,
LW> > 
LW> > here is the NC-input that was working well for me.
LW> 
LW> Dear Axel,
LW> 
LW> I repeated the simulation using your input
LW> (http://bioexploratorium.pl/ljw/cpmd/nvt/axel/2mdnvt.inp) and I did not
LW> get the adiabatic behavior - the EKINC drifts from around 0.004 Ha
LW> initially to 0.008 Ha after 7ps

since the machine i have available for these kinds of tests, 
i cannot run that long. please see the first attached graph. 
it does show a little slant now. 
BTW: based on its "spectrum" i have changed coupling frequency
for the N-H chains to 2500 cm-1, if you look at the second 
attached graph this seems a much more reasonable choice as it
couples to a wider selection of modes and i particularly don't 
think it is a good idea to mess with the mode that interests
you directly. it will of course be excited because its overtones
are included in the frequencies here.

LW> (http://bioexploratorium.pl/ljw/cpmd/nvt/axel/EKINC.png). - Did you run
LW> the simulation for that long time ?
LW> 
LW> When I reduce the electron mass to 300 a.u. the drift is slower
LW> (http://bioexploratorium.pl/ljw/cpmd/nvt/axel/EKINC-1.png) but reducing
LW> e-mass seems to be a poor strategy since from some point I would have to

when you turn on path-integrals, you _will_ have to reduce the time step, 
anyways (unless you run BO-dynamics).

LW> decrease also the time step which cancels the effect. It also seems to
LW> me that it is impossible to achieve exact adiabaticity for my system.

yes. it looks like your system is a somewhat pathological case.

LW> The only thing I can dream of is minimizing the EKINC drift. Is there
LW> any reasonable strategy to locate this minimum in the CPMD parameter
LW> space (at least dt, e-mass) ? Does this happen for every real system or
LW> is this only my case ?

one issue is that you have large vaccuum areas and those tend to 
produce "noise", particularly with gradient corrected functionals. 
this can in principle be reduced by increasing the wavefunction cutoff,
but i suspect that you need a ridiculously large cutoff to get totally
rid of this noise. it may also change with the choice of functional. 
some people claim, that BLYP is particularly sensitive.

the code already cuts off the gradient correction calculation when the 
density is too low (GC-cutoff), but that in itself can be a source of
noise that would feed the increase of EKINC.

cheers,
   axel.

LW> 
LW> Regards,
LW> Lukasz
LW> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ekinc.png
Type: image/png
Size: 14174 bytes
Desc: 
Url : http://cpmd.org/pipermail/cpmd-list/attachments/20080512/5a6d556f/attachment-0002.png 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: ekinc-spec.png
Type: image/png
Size: 9589 bytes
Desc: 
Url : http://cpmd.org/pipermail/cpmd-list/attachments/20080512/5a6d556f/attachment-0003.png