[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [ESPResSo-users] [ESPResSo-devel] Magnetic particle dynamics in LB f
Re: [ESPResSo-users] [ESPResSo-devel] Magnetic particle dynamics in LB fluid flow
Wed, 29 Jan 2014 15:11:56 +0100
Dear Markus, dear users-list,
as this question can be of interest to many, I have moved it to the Espresso
users mailing list.
On Jan 28, 2014, at 2:00 PM, Markus Gusenbauer <address@hidden> wrote:
> Dear developers,
> I would like to incorporate magnetic particle dynamics (particles have its
> volume, collision detection between particles) in the lattice Boltzmann flow.
> My plan would be to connect the features
> - Dipolar all-with-all and no replicas (DAWAANR)
> - object-in-fluid or/and virtual sites
> - LB
I have no objections, but I'd assume that the dipolar P3M reaches a higher
speed. I'd also assume that DAWAANR needs atomic decomposition (every CPU is
responsible for a certain chunk of particles) and CPU LB does not like that.
CPU LB only likes domain decomposition, as it also decomposes the computational
domain. That can in principle be fixed, but surely takes a day or two. GPU LB
should in principle work with that method, but I'm not sure it is actually
implemented. (The particle collection must work somewhat different then…).
The guy to ask about magnetic interactions is Rudolf Weeber btw.
From my perspective the strength of the OIF is that you can have soft
particles, but probably your magnetic particles are supposed to be solid.
Therefore I'd go for virtual site. Joost de Graaf made tests with "rigid
raspberries" and it looked very good. Before using them, I'd however calculate
the pair mobility of two of them and check it against analytical results (See
for example Ngyuen and Ladd 2004 (I guess), Duenweg and Ladd 2009, Brady and
Bossis, and other people).
The question here is: How important is their "leakiness" for their hydrodynamic
interactions -- and how accurate does it have to bee.
Furthermore it is probably possible to include a rotational coupling in the
same fashion as the translational Duenweg-Ahlrichs-point coupling. One could
introduce a frictional force proportional to the difference between the angular
velocity and the local vorticity of the fluid. This makes sense if you have a
very dilute system of particles.
A very similar effect to that can probably be achieved with very small rigid
particles. If you \eg just use a small tetrahedron of particles you already get
rotational coupling. It however will be necessary to check if there is a unique
"radius" that describes both rotational and translational coupling.
> The magnetic particles should interact with the fluid and vice verca. And of
> course the particles should interact via dipolar forces.
> Would it be best to use virtual sites or the object-in-fluid implementation?
> Using virtual sites the fluid interaction of the virtual nodes need to be
> implemented. And the particle in the center of mass could handle the dipole
> interaction. Object-in-fluid would have already the fluid-particle
> interaction. Maybe a combination of both methods would be sufficient.
> Is there anyone who is already working on the subject or a similar one? Maybe
> without the magnetic part, just virtual sites in the LB flow. Is it maybe
> already in developement?
> I would be happy to get comments on the topic and to share ideas.
|[Prev in Thread]
||[Next in Thread]|
- Re: [ESPResSo-users] [ESPResSo-devel] Magnetic particle dynamics in LB fluid flow,
Stefan Kesselheim <=