[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [ESPResSo-users] Any limit of the lattice Boltzmann grid number?

From: Chunda Zhou
Subject: Re: [ESPResSo-users] Any limit of the lattice Boltzmann grid number?
Date: Thu, 1 Sep 2011 12:01:42 -0400

Hi Florian,

I am sorry. I thought gmail would automatically cc to the mail list. 

My current system has 384 ions. And the overall density is 0.01. As a result, the size of each edge is 37. The system might be expanded to a larger scale with more ions under different scenario. 


On Wed, Aug 31, 2011 at 4:38 AM, Florian Fahrenberger <address@hidden> wrote:
Hallo Chunda,

first of all: Please ALWAYS use the mailing list for questions! Otherwise everyone who encounters the same questions will not be able to find the answer already given and ask again.

1. But the basic idea from your last reply to
choose cell number is as many as computer can handle so that the whole
"resolution" of water is high enough to avoid unphysical behavior such as
fluctuation, right?

No. That is not what I would do. Since the Lattice-Boltzmann time step is usually larger than the MD time step, you will more probably end up with unphysical behaviour if you choose your grid too fine. Each LB time step, the fluid speeds can only propagate for one LB cell, and a very fine mesh results in very small cells, which results in a very slow propagation of hydrodynamics.
As I said before, grids of 16 or 32 cubed are usually big enough for a senseful fluid dynamics. If you have huge or unconform systems like tubes or pores, you might want to go with 64^3 or even bigger, but that is usually not necessary.

2. My system now is a 128-monomer polyelectrolyte in salt solution in an
external electric field. The contour length is 16microns. Is this system too
large for a good Lattice Boltzmann Simulation. If I take a 64^3-cell LBF,
does that means in a certain direction, only half of the monomers are
affected by fluid? If this true, then I need a way bigger number of cells to
have a water molecule like behavior?

This system is definitely not too large. :o) All particles are affected by the fluid, since the fluid speed in interpolated onto the particle position. Even with a 4^3 LB lattice, all particles would see the fluid, but the resolution of hydrodynamics would be lesser, i.e. effects between neighboring particles might be distorted.

You did not tell me your box size, but if it is the same as your contour length, then even if your polymer is completely stretched out (that will probably not happen in salt solution, unless you apply a very, very high electric field, and then you should be careful about linear response anyways), you will have useful hydrodynamics in your system, and the resolution will be enough. Probably even 32^3 would be sufficient.

3. Another question: about the units of lbf. The manual said “The parameters
dens and visc set up the density and viscosity of the LB fluid in (usual) MD
units.”  So my understand of this is all these units are same as MDs. If the
understand is yes, then the manual said: "  In one LB time step typically
several MD steps are performed.". What is the ratio between the LB iteration
and the MD iteration? Is that simple Tau/md_timeStep? Could  you provide a
reasonable set of parameters for the system like mine?

Yes, the ratio for the amount of steps really is Tau/time_step. For the reasonable set of parameters, you should ask the list again, since I have never worked with LB myself and it seems to be a complicated subject...

4. For the friction coefficient gamma. It describes the interaction between
small particles such as ions in the solution. My question is: does it affect
all the different objects in the fluid in the same way? Because the monomers
in my polymer are much heavier and bigger than ions. I expect ions gain very
high speed when interact with the polymer. As a result, the dispanssive
force due to speed difference will unfairly slow down ions more ( or slow
down monomers less), because temperature is not solely determined by speed.
Is there a way to avoid this or just choose a good common gamma value? Base
on your experience, could supply any good range of this friction coefficient
gamma for water and ions?  (By the way, the MD does not allow to set mass of
monomer other than 1, so I chose single monomer mass as reduced mass unit. I
am not sure whether this is fine or not.)

Again, you should pose this question to the mailing list. There are people far more experienced than me in the choosing of LB parameters, since I have never really used it.

ALWAYS write to the mailing list!



Thank you  very much.

On Mon, May 2, 2011 at 1:17 AM, Florian Fahrenberger<
address@hidden>  wrote:

Hallo Chunda,

My current box length is 690. And I want to have LB grid size at 0.004

Wow! Just a quick calculation with a D3Q19 implementation of Lattice
Boltzmann for this system:
(690/0.004)^3 * 19 = 9.7526e+16
This means you would need 100000000 GBytes of RAM just to store your
fluid in single precision!!! I don't think this is a senseful number for
whatever system you want to calculate.

Even your second try with 690 cells per dimension will leave you with 10
GBytes of memory and is probably too much. Depending on what kind of
system you want to do your simulation on.

Usual system sizes for LB are around 32^3 or 64^3 cells, maybe slightly
bigger if necessary.

So, to answer your question: There is no hard implemented limit for the
number of cells in LB, but it obviously depends on how much RAM your
system has, and you seem to be running out there. Your first try might
even have been caused by the problem that the cells could not be
numbered because your computer ran out of integers. :o)

What kind of system do you want to simulate anyways, with such a huge
amount of LB cells?



Florian Fahrenberger
Institute for Computational Physics
Universität Stuttgart
Pfaffenwaldring 27
D-70569 Stuttgart
phone: (+49)-(0)711 / 685 - 63594
fax:   (+49)-(0)711 / 685 - 63658

Chunda Zhou
Department of Physics
Office phone: 919-513-3914
Cell phone: 919-995-5631

reply via email to

[Prev in Thread] Current Thread [Next in Thread]