[CPMD-list] BFGS update

[Erik Santiso]

Hi,

I thought that it may be useful if one wants to run a constrained
optimization for a structure that is at (or close to) a transition state.
I was trying to optimize a 1,3-butadiene molecule in the cis conformation
(which is a transition state) by constraining the torsion angle between
the carbons, but could not get it to converge. I guess the reason is that
in this case the Hessian for the optimum structure is not positive
definite and hence the BFGS update doesn't work.

I guess the solution in this case is to use RFO ORDER=1. Thanks again for
all your help!

Erik.

>
>
>
>
> Hi,
>
> for the nuclear Hessian, both BFGS and Powell updates are implemented.
> However, Powell is only called if RFO (not P-RFO which uses its own
> Hessian
> and its own implementation of the Powell formula) is used and a saddle
> point is searched for. In all other cases, the BFGS update is used because
> it preserves the positive definiteness of the Hessian which is a
> prerequisite of a energy minimum. Why do you want to use Powell with
> GDIIS?
>
> Best regards,
> Salomon
>
>