Re: [ESPResSo] Re: analyze nbhood & Bond-angle Interactions

[duenweg]

> Dear Peter,
> I actually tried out your suggestion and I liked what I saw.
> Before delving into the (controversial) choice of the expression of
> the interaction potential for nanoparticles (most likely I will have
> to use a tabulated potential; BTW: can it be capped?) there are
> several other questions I need to address (of course I am asking them
> to the whole Espresso list).
> Some of them go beyond numerics and Espresso, but probably someone
> knows the topical reference for that.
> Let us say that I want to study the dynamics of N particles in the
> Langevin thermostat, all of them interacting with a very deep
> Lennard-Jones (LJ) potential.
> 1)At the beginning, I will have to cap the LJ potential after randomly
> locating the particles.
> Is there a recommendation for the choice of the cap value? Is it
> arbitrary?

Read my paper with Achim Kopf. Reference should be
available on my web page (either preprints, or
publication list).

> 2)In some of the tests I was running, VMD was showing the particles to
> "stick" together and form a large aggregate. In some cases, the
> aggregate would cross the box boundary (periodic boundary conditions
> assumed) and part of it would re-enter the box somewhere else.
> Is this a conceptual problem? I can easily cope with a single particle
> exiting and being re-introduced, but I am not sure this is sound for a
> large aggregate made up of tens of particles strongly bound (whose
> structure is precisely what I want to study). In other words: what is
> the recommendation for the box size? Is it simply fixed by the density
> I want to reach?

Sounds all reasonable to me. I don't see a fundamental
difference between a re-enetering particle and a re-entering
cluster.

> 3)How does one know that the solution has "converged"?
> In other words: what are the indications that the time step, the box
> size and the total time duration of the simulation have been properly
> chosen?
> Is it perhaps the fact that some statistics of the results is
> insensitive to e.g. halving the time step, doubling the box size while
> keeping the density constant, and so on and so forth?

short answer: yes. you should be particularly careful that
the statistical averages are insensitive to the runtime
(and if you are not sure, you should better vary it by
at least one, better a few, orders of magnitude.

> 4)In the tutorial, I saw that Espresso actually creates a mesh. When
> should one start being concerned with it? In case it matters, I
> re-state my physical problem: diesel exhaust nanoparticles in hot air
> are "kicked" randomly by air molecules and only once they get within a
> certain distance (particle radius) from each other they "stick"
> together (so far the easiest thing to do was to use a deep LJ
> potential).
>

can't answer this honestly, but I believe the mesh is
nothing but a computational trick to find neighbors
efficiently - ie has no physical significance.

Regards Burkhard


> Many thanks
>
> Lorenzo
>
>
>
> On 23/10/2007, Peter Kosovan <address@hidden> wrote:
>> Dear Lorenzo
>>
>> In my opinion, using the bond-angle potential to represent non-spherical
>> shape and angle-dependent interactions in the aggregation of colloidal
>> particles is not the most natural choice. Moreover, using the FENE-type
>> of
>> bond for a colloidal system seems wierd to me as well. I would suggest a
>> different workaround.
>>
>> First of all, I would suggest using lennard-jones type of a potential
>> between the particles should suffice to stick them together rather than
>> introducing a true bond between them. If you set epsilon_LJ>>kT, then
>> once
>> the particles find each other, they stick together. If you set
>> epsilon_LJ
>> comparable to kT, you also get a finite probability that a particle
>> escapes from the cluster which seems to me a more physical picture of a
>> colloidal system when compared to introducing non-breakable FENE bonds.
>>
>> The second suggestion is that you do not have to represent a single
>> colloidal particle by a single particle in Espresso. In the simplest
>> case,
>> you can represent it by two espresso-particles forming a dumbbell
>> (dimer),
>> or you can build more complicated representations such as A-B-A trimer,
>> with the A-B-A bond angle fixed. In this way, a colloidal
>> particle of any shape can be represented and the individual
>> espresso-particles play the role of interaction sites on the surface of
>> the colloidal particle.
>>
>> With regards
>>
>> peter
>>
>> Peter Kosovan
>> Department of Physical and Macromolecular Chemistry
>> Faculty of Science
>> Charles University in Prague
>> Czech Republic
>> address@hidden
>> Tel. +420 221 951 290
>>
>
> _______________________________________________
> ESPResSo mailing list
> address@hidden
> https://fias.uni-frankfurt.de/mailman/listinfo/espresso
>
> This email was Anti Virus checked by Astaro Security Gateway.
> http://www.astaro.com
>
>