[CPMD-list] about hessian maatrix

[Juerg Hutter]

Hi

> I have a few questions about hessian calculations in CPMD.
>
> 1) Is it calculated by analytical means or numerically?

If you run a "VIBRATIONAL ANALYSIS" you can choose between
analytic computation (LR) or finite differences (FD).
See the manual for more options. I would say the finite difference
code is more robust.

> 2) Is it calculated on a Born-Oppenheimer surface? OR
>   is there any contribution from "fictious" electron dynamics?

If you refer with 'it' to the harmoonic second derivatives calculated
with "VIBRATIONAL ANALYSIS" then the answer is BO, no effects of
electron dynamics.
However, if you calculate the vibrational spectra from the Fourier 
transform of the velocity auto-correlation or the dipole auto-correlation
out of a CP run (this goes beyond the harmonic approximation)
then you have 'strong' effects from the electron mass. See the
literature about this.


> 3) Can we calculate a hessian while we are doing a MD simulation?
>   (I mean, without minimization) In this case, we, of course,
>   have imaginary frequencies but we sometimes use such quantities
>   in MD simulations ...
>

This option (e.g. instantenous normal modes) is not implemented
in CPMD, but see the possibility mentioned above.

regards

Juerg Hutter



> I appreciate your comments.
>
> Best wishes,
>
> -- 
> Hiroshi Fujisaki <fujisaki at bu.edu>
>
> _______________________________________________
> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://cpmd.org/mailman/listinfo/cpmd-list
>