Re: [ESPResSo] constraints ; interactions

[Florian Ruehle]

Dear Guillaume,

I never experienced your segmentation fault (maybe a particle is put
exactly onto a constraint?), neither your "bond type unknown". Maybe for
the latter, you have to check if your system is set up correctly or
maybe send the relevant parts of your script.

But of course I have often seen the "bond broken" message. :o) In
general, I can tell you that systems with constraints (especially very
dense systems) have to be dealt with very carefully. The most common
problem stems from doing the warmup like for unconstrained systems with
a "force capped lennard jones" potential. The problem with this method
is, that ALL LJ-potentials are capped, including the constraint
potential, which is very likely to be Lennard-Jones, too. So the
particles can actually violate the constraints and afterwards experience
very high forces.

The solution for me was, to do two more things manually:
1. Place the particles in the box manually per TCL random generator.
This lets you choose the region you place them in easily and you can
leave a small gap between your particle region and the constraint.
2. The warmup. For me it turned out to be the most clever way to put
particles in the box and slowly grow the LJ-sigma from a very small
(significantly smaller than the analye_mindist smallest particles
distance) value to the size I want my particles to have. That way, you
can warm up your system without giving the particles the chance to
violate your constraints. I also slowly grow the timestep during my
warmup, starting with a small value.

This might already solve your problems, hopefully.

Greetings and I hope I could help,

Floh
(Stuttgart)

Am 26.10.10 11:17, schrieb Guillaume Larsen:
> Dear  all,
> 
> I am implementing cell constraints with associated LJ and Tunable slip
> interactions. For information, the system is composed of 3 different
> types of polymers plus 1 type of particle.
> When I choose the standard 12-6 LJ potential, I get "segmentation fault"
> error at the beginning of the integration. Instead, if I choose a
> generic LJ potential (like 9-3LJ with appropriate parameters), I get the
> following message: background_errors 0 {082 add_bonded_force: bond type
> of atom 1 unknown} and also 1 bond broken between particles error
> message. I precise that the warm-up stage is processed correctly.
> Basically, the whole definition of the system - particles, polymer,
> thermostat, particles/polymers interactions... - works fine for an
> analysis without the use of cell constraints.
> Does anybody have a clue about the meaning of errors I get?
> 
> Thank you very much in advance,
> 
> Guillaume Larsen
> 
> 
> 
> _______________________________________________
> ESPResSo mailing list
> address@hidden
> https://fias.uni-frankfurt.de/mailman/listinfo/espresso