[molpro-user] Total energy unreasonable

[Gerald Knizia]

On Wednesday 26 November 2008 11:44, Vincent Liégeois wrote:
> I would like to run a simple test case consisting of a chain of H atoms.
>
> Here is the input file: [...]

Are you sure this is the correct input file? Because this ->

>   INTERPOLATION TYPE:            NONE
>   INTERPOLATION STEPS:              2 (START)      1 (STEP)
>   LEVEL SHIFTS:                  0.00 (CLOSED)  0.00 (OPEN)

Is not specified in the input. Setting IPOL to None disables DIIS
convergence control, and usually leads to the alternating fixed points you
see in your iteration. This cannot happen with DIIS enabled.
I guess that the default level shifts (also not seen here) would usually also
help on rather than break the convergence.

> I have try to play on different threshold like twoint or throvl
> without succes.

The problem is in the non-existing interpolation specified. Twoint or throvl 
are unlikely to become a problem with H atoms and minimal basis sets...
Also notice the comment of Prof. Knowles on {hf; noenest} which applies if you 
see the unreasonable total energy warning with DIIS enabled.
-- 
Gerald Knizia