[gmx-developers] Morse as non bonded (?)

Marcelo A. Carignano cari at purdue.edu
Wed Mar 11 19:59:54 CET 2009


Than you very much David,
I will try with the nb_generic code to see it I can make it work.

But, if you say the tables are more efficient, then I will try again
with them.
I have a problem though, the simulation gives a segmentation fault as
soon as
two atoms are within the cut-off distance (r < r_cutoff).
This happens for a system that consists of chain molecules (just 5
beads, bonded
with fourth power bonds and harmonic angles, exclusion among the five).
One chain runs fine, two chains (interacting via table) dont. The
program crashes.
If instead of the chain molecules I have only point particles
interacting with the same table,
everything goes well.
I could't find a solution by modifying the parameters (temp, tau, dt, etc).
Do you have any idea where to look for the problem?

I am using v4.0.4; and try both single and double precision, with the
respective
recommended table spacing.

Thanks again.
Marcelo.
--
David van der Spoel wrote:
> Marcelo A. Carignano wrote:
>> Salut,
>>
>> Morse potentials are implemented in gromacs as a bond type.
>> However, the potential has the interesting feature that allows an
>> independent control
>> of the depth, core size and range of the interactions (I need something
>> like that).
>> Therefore, using Morse potential as non-bonding will be useful for me,
>> and possible others.
>>
>> The question is how difficult would be to modify the code, using as base
>> the Morse code
>> for bonds, to write a Morse code for non-bonding. I am willing to
>> attempt this, if I get some
>> general indications (I can use also very detailed indications) from
>> those of you who know
>> the code with sufficient detail to provide some advise.
>>
>> I know that I can use the tables, but that is not the solution I am
>> looking for.
> Why not? It is very efficient (much more efficient than analytical
> code). The bonded Morse code is quite simple (src/gmxlib/bondfree.c).
> There is also a generic nonbonded kernel (in C) for you to play with
> in src/gmxlib/nonbonded/nb_generic.c.
>
> The real interesting thing in my opinion would be to have non-bonded
> three-body interactions...
>
>>
>> Many thanks,
>>
>> Marcelo Carignano
>> -- 
>> _______________________________________________
>> gmx-developers mailing list
>> gmx-developers at gromacs.org
>> http://www.gromacs.org/mailman/listinfo/gmx-developers
>> Please don't post (un)subscribe requests to the list. Use the www
>> interface or send it to gmx-developers-request at gromacs.org.
>
>




More information about the gromacs.org_gmx-developers mailing list