Re: [ESPResSo-users] DPD

[Stefan Kesselheim]

Hi,

On Sep 11, 2014, at 11:42 AM, Dudo <address@hidden> wrote:

> Hi Ulf,
> 
> thanks for replying. Some nice papers you've got there, I'm going to see if I 
> find the solution in there.
> 
> Well, I'm not sure about the physical picture though. Of course, the picture 
> is correct as for the conservation of energy,
> but the periodic box now does not represent a sample of an infinite system, 
> it's more like a limited amount of
> solvent taking an infinite energy, what is not very physical.
> 
> Also, I very agree on the word 'pump'. The molecule really acts here like a 
> pump, what is not the picture
> I'd like to have. 
> 
> I like the idea to dissipate the energy on the walls as Chris suggested, 
> please, would I implement it with the 
> "thermostat inter_dpd ignore_fixed_particles 0"? Or should I make the 
> boundary from additional particles? I'd like 
> to avoid explicit particles as possible, due to the size of the problem.

For walls you can either use the "tuneable slip boundary condition" in 
conjunction with "constraints" (Smiatek, around 2008), or the "reflecting" 
property of the constraints. If you set that property to two, and apply only a 
"dummy" interaction between constraint and particle, the velocity of particles 
is inverted, when they cross the boundary. If you use one, only the 
perpendicular component is reflected (like a mirror, creates full slip), but 
with two all velocity components are reflected. 


The wall is one solution, the other possible solution is using a counterforce 
on the solvent molecules so that the net force on the system is zero. In both 
cases, you obtain correction of O(1/L), where L is the box length, compared to 
the physical situation you are interested in. You can correct for that by 
calculating D for different L and extrapolate to L->infinity. For the 
counterforce case, there is an analytical correction available (1-2.81 a /L, 
Hasimoto 1959, see e.g. Duenweg, Ladd 2009), but as far as I know not for the 
case with walls. You need to know the solvent viscosity though. This you can 
obtain e.g. by doing the extrapolation for a single chain length (e.g. 1), or 
other methods. The simplest possible is probably creating a Poisseuille flow, 
by applying a body force to a fluid between two walls, measuring the velocity 
profile and calculating the viscosity from the curvature. 

Cheers
Stefan