[gmx-developers] Question about the QM/MM methods

[Gerrit Groenhof]

Hi,

The gamess-uk interface is based on the 6.x version of Gamess-UK. It
worked via function calls from a linked games-uk library.

Huub van Dam has been working on a new interface, based on file
communication. The gaussian interface works like that as well.I was not
aware there were problems.

>From you emails I infer you have difficulties with the way the
information is passed on between the mdp file and the qm routines.

I have hard coded the possibilities for basissets and methods. The
problem being that basisset specification was different in different
programs. Thus now the user's input has to be recognized, stored as an
integer (enum) and then used to create the correct input for a QM routine.

For writing an interface, I would look at the gaussian.c, focus only on
the write_gaussian_input, read_gaussian_output, can call_gaussian, whcih
are used for 'normal' qmmm calculations.

Hope it helps,

Gerrit


Ake Sandgren wrote:
> On Fri, 2009-11-13 at 10:56 +1100, Mark Abraham wrote:
>   
>> Åke Sandgren wrote:
>>     
>>> Hi!
>>>
>>> In src/gmxlib/names.c there is the
>>> const char *eQMmethod_names[eQMmethodNR+1] = {
>>>   "AM1", "PM3", "RHF",
>>>   "UHF", "DFT", "B3LYP", "MP2", "CASSCF","B3LYPLAN",
>>>   "DIRECT", NULL
>>> };
>>>
>>> Can someone of you translate that to gamess-uk lingo?
>>> I.e. what is expected of gamess-uk for the various gromacs inputs that
>>> correspond to those?
>>>       
>> src/mdlib/qm_gamess.c is probably your best source of information...
>>     
>
> That's just the point. The input file that qm_gamess.c is writing is
> currently incorrect from a gamess point of view in some cases.
>
> So what I'd like to know is what gromacs expects to happen from a qm/mm
> point of view when those methods are requested by the users
> gromacs-input.
>
> I.e. for DFT, is it ok to just use gamess-uk's default or should it be
> more specific and so on.
>
> Or in other words is the gaussian or mopac interface known to be working
> well. If so then one should be able to work out what to generate in the
> gamess-uk case from that info. Although that is a bit hard since one
> would have to understand those two also...
>
>