[CP2K:1105] SIC does not converge
fa... at gmx.ch
Thu Jun 12 12:49:53 UTC 2008
On Jun 12, 2008, at 1:50 PM, jschmidt wrote:
> I'm trying to calculate partial charges for an ionic system (Cl and an
> organic cation) and as a first test I tried to find out what charges
> they have in a dissociated state (distance in my simulation about 10
> Angstroms). Unfortunately I always find a charge of about -0.7 on the
> Cl. My guess is that this might be due to self-interaction and so I
> added the keywords
are you aware that partial charges depend on the method you use to
Mulliken for example typically underestimate the charge for ionic
systems (and does not converge wrt. basis set).
Other methods are much better, I also have one that seem to work well
for ionic systems, but it isn't really ready for public consumption yet.
> SIC_METHOD MAURI_SPZ
> &END SIC
> to my input (using UKS).
As far as I know you need OT with RKS and rotations for SIC, if you
really need it probably Joost VandeVondele is the one that knows best
how to really use it.
> Unfortunately, the wavefunction does not
> converge anymore (at least not after about 500 steps), neither with
> DIIS or OT/CG or OT/SD nor without OT. I also tried different initial
> guesses (atomic, random, restart from calculation without SIC). I have
> the same problems with SIC_METHOD MAURI_US and with SIC_METHOD AD.
> Is there any severe mistake in my settings? Any ideas?
> Thanks for your help!
More information about the CP2K-user