counter charge

[Jun]

Thanks, Teo.


> Most of the time, Jun, it is enough to read carefully the few
> documentation you have!
> This keyword makes that task for you:
>
> http://cp2k.berlios.de/manual/CP2K_INPUT/FORCE_EVAL/SUBSYS/
> KIND.html#desc_CORE_CORRECTION
>
> the value of this keyword is summed-up to the core..

I did read the input reference. Sometimes, it is not easy to catch the
key idea
of the keyword, especially when I didn't know where should I focus.
If I understand it correctly, I can combine the keywords
CORE_CORRECTION (fractional or integer)
and BASIS_SET NONE to create a point charge with any amount of charge.
Two questions
come up as follows,
(i) Suppose we have a H2 molecule consisting of one normal H atom and
one fractional charged
(say 0.1) H nucleus without basis set, the system still contains 2
electrons (I saw it from output,
It seems that the number of electrons is decided by the kind of
element). Does it mean that the overall
system carry a net charge (-1-1+1+0.1=-0.9)?
(ii) I tried to assign a negative charge (say -0.1) to it, and cp2k
did allow that. Then, for the above-mentioned
example, the system shoudl consists of one normal H nucleus, one
negative point charge, and two electrons,
in total carrying a net charge of 1-0.1-1-1=-1.1. Am I right?

Thanks in advance.


> Jun, I think there's a misunderstanding.. or quite probably you are
> not used to calculations with plane waves.

Now I understand the difference between plane wave code and gaussian
orbital based code. In plane wave code, it is not possible to freely
switch
off basis sets as in gaussian orbital based code. Plane waves can
always
find the point charge. One have to put a high potential around the
point charge
to keep electrons away by modifying pseudopotential. Whereas, in cp2k
we
can conveniently switch on BASIS_SET NONE to do the same thing.


> Of course you can reinvent the wheel, but if you want the energy of
> an isolated charged system then I would highly recommend you to use a
> decoupling for the coulomb interactions (Martyna-Tuckermann or
> others..).

Thank you for kindly and constantly warning me about the calculation
of charged systems.
Martyna-Tuckermann method is certainly useful for cluster models.
However, we have to use
periodic boundary conditions. We will take into account the finite
size effects due to the spurious
interactions among charge, counter charge and their image.

Cheers,
Jun