RE: [ESPResSo-devel] Book-keeping of bonded partners

[Limbach, Hans Joerg, LAUSANNE, NRC-FS]

Dear Tristan,

Can you explain, why you can not solve your problem with a bonded
interaction potential?

Concerning your other problems:
It is not possible to obtain the bond list of all particles in the
non-bonded interaction calculations because the bond list is not
communicated during the ghost communication. If you try you may access
memory that has not been initialized and you will get rubbish.
Do not implement anything that assumes that bonds just go from particle
i to particle i+1

Workaround:
(But this doubles the computational effort to calculate the bonded
interaction!)
If you want to have a bond between particle i and j. Create two bonds,
one from i to j and one from j to i and divide the strength of your
original interaction by 2.

Best regards,
Hanjo

> -----Original Message-----
> From: address@hidden 
> [mailto:address@hidden On 
> Behalf Of Tristan Bereau
> Sent: mercredi, 2. avril 2008 21:34
> To: address@hidden
> Subject: [ESPResSo-devel] Book-keeping of bonded partners
> 
> Dear all,
> 
> I am implementing a new non-bonded interaction that requires 
> knowledge about bonded partners.
> At the C level of Espresso, is there an easy way to determine 
> all the bonded partners of a particle ?
> 
> Say I have particles only connected by bond length potentials 
> (just an example). If I take one particle (let's call it 
> 'i'), I can get its bonded partner 'i+1' from the bond 
> length, but not particle 'i-1'
> (that's because it is particle 'i-1' that stores particle 'i' 
> as a bonded partner).
> 
> Right now I'm cheating by using the information given by bond 
> angles between neighboring partners. Particle 'i' has a bond 
> angle with 'i-1'
> and 'i+1'. This method works only if I use that potential... 
> Is there any way I can do something more elegant ?
> 
> As a side note, I get information about bonded partners by 
> using the properties of particle i :
> p_i+1 = local_particles[p_i->bl.e[...]]; I hope this is what 
> I should be doing.
> 
> Many thanks,
> 
> Tristan
> 
> _______________________________________________
> ESPResSo-devel mailing list
> address@hidden
> https://fias.uni-frankfurt.de/mailman/listinfo/espresso-devel
>