[CPMD-list] Power spectrum of the electronic degrees of freedom

[Łukasz Walewski]

Dnia 2008-05-15, czw o godzinie 01:12 +0200, Łukasz Walewski pisze:
> Dnia 2008-05-14, śro o godzinie 18:10 +0200, Łukasz Walewski pisze:
> > Hi,
> > 
> > This question has been raised already
> > (http://cpmd.org/mailman/htdig/cpmd-list/2006-November/003574.html) but
> > is still left unanswered: how to calculate the power spectrum of the
> > electronic degrees of freedom, i.e. Eq. (47) from
> > http://www.theochem.ruhr-uni-bochum.de/research/marx/marx.pdf
> > 
> > The problem is how to get the velocity autocorrelation function of the
> > electronic DOF from CPMD simulation ?
> > 
> > Regards,
> > Lukasz
> Lukasz,
> 
> you can easily do that in xmgrace. load your electronic kinetic energy
> as a 
> single set and do a Fourier transform of it. 
> this is easily done with the little skript attached.
> maybe you have to modify the timestep. i've based it on the
> data you have posted recently.
> you can use it as
> > xmgrace -batch power.xmb elec-ekin.dat
> or however your data file is called.
> have fun,
> gerald
> 
> PS: i've just seen that axel was quicker. i've trouble with the mailing
> list accepting my mails, so i address you directly. if you want, just
> forward this 
> mail to the list.

Dear Gerald, Axel,

Yes. The Fourier transform of EKINC gives some hints about the power
spectrum but has a lot of noise in the low frequency region (that is of
my particular interest). That's why I wanted to know how to get the
exact power spectrum or alternatively how to "temperature correct" it.
One can subtract the mean value of the signal in order to get rid of the
0th frequency but this does not take all noise away.

Autocorrelating EKINC before FT does not work for me since the ACF in
the case of systematic noisy drift is almost linear decay with only one
frequency close to 0 (or couple of them).

Regards,
Lukasz