espressomd-users
[Top][All Lists]
Advanced

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

Re: system time step appropriate values


From: Jean-Noël Grad
Subject: Re: system time step appropriate values
Date: Tue, 29 Jun 2021 12:23:44 +0200
User-agent: Mozilla/5.0 (X11; Linux i686; rv:68.0) Gecko/20100101 Thunderbird/68.12.0

Hi,

If you have indeed a channel that is several orders of magnitude longer than its diameter, you could consider setting up an infinite LB channel using periodic boundary conditions along the main axis.

Best,
JN

On 6/29/21 11:21 AM, Martin Kaiser wrote:
Hey Benyamin,

in my humble experience, it is extremely difficult to set-up the system exactly how you described it. From what I understand, you are trying to simulate a nanometer sized object in some sort of channel that is in the centimetre range that has a flow inside. Therefore, the characteristic lengths (size of your particle and size of the channel) are separated by several orders of magnitude, which is also true for the time scales of those systems. This poses a huge issue on its own and usually requires special techniques to achieve, depending also on the effects you want to capture in your system. The fact that you want to simulate several seconds, paired with the fact that typical time-scales in simulation on that particle scale are, as you already found out, picoseconds or even smaller, means that you would need insanely long simulations.

Apart from the time-scale, if you want to stick to the correct ratio between your particle and channel size, you will get technically unmanageable system sizes. In my opinion it would be worth to reconsider what the effects are that you want to measure and to rescale the system to a more manageable size.

Regarding the simulation parameters, I remember a sentence about the parameters in the tutorials saying something along the lines of “we use those parameters for no other reason than them being stable”. I have not looked at the tutorials in a while, so I don’t know if this is still true, but keep that in mind. It is always advised to triple check and calculate your parameters, which can be troublesome when using LB. The guys from reference [1] did a great job with LB and it helped us a lot when setting up our own systems. There is only a limited range of parameters for with LB is stable, so you not being able to increase your tilmestep beyond your given value does not surprise me. I can also recommend book [2], especially sections about stability, to get an overview over the possibilities within LB.

I hope that helps.

Best,
Martin


[1] Kreissl, Patrick, Christian Holm, and Rudolf Weeber. "Frequency-dependent magnetic susceptibility of magnetic nanoparticles in a polymer solution: a simulation study." /arXiv preprint arXiv:2010.00299/ (2020). https://arxiv.org/abs/2010.00299

[2] Krüger, Timm, et al. "The lattice Boltzmann method." /Springer International Publishing/ 10.978-3 (2017): 4-15.


===========================
Martin Kaiser, M.Sc.
Computational and Soft Matter Physics
Dipolar Soft Matter Group
Faculty of Physics, University of Vienna
Kolingasse 14-16, 1090, Vienna, Austria

On 28.06.2021, at 18:38, Benyamin Naranjani <bnaranjani@gmail.com <mailto:bnaranjani@gmail.com>> wrote:

Hi,

I intend to simulate transport of one particle representing an insulin molecule in a macro-scale fluid flow ~cm using lattice-Boltzmann coupling. I notice that in tutorials values for friction coefficient and particle mass are close to unity. In this case, the user is able to vary time step value in a reasonable range. I assume this is because the friction coefficient and particle mass which are on both sides of the coupling equation would be roughly of the same order of magnitude. This would allow the user to vary the value for time step without causing any numerical issues for the integration procedure.

In my simulation, I set values for Boltzmann constant times temperature parameter, gamma, and particle mass based on Stokes-Einstein's relation in SI units. I calculate the gamma according to the diffusivity value of the insulin molecule. For the particle's mass, I set it to be equal to the mass of one insulin molecule. In this manner, the value of gamma and particle's mass would be significantly different as they are 2.85e-11 [kg/s] and 9.64e-24 [kg], respectively.

Under this condition, I cannot increase the time step value to be over 5e-13 [s], otherwise, the jupyter notebook kernel would die. Since I am interested in studying transport of this molecule in larger time/length scales for instance in tens of seconds, I need to be able to increase the time step value significantly. I would really appreciate your thoughts on the applicability/possibility of increasing the time step.


All the best,
Benyamin


----------------------------------------------------------
Benyamin Naranjani; PhD candidate
Department of Pharmacy
Uppsala University
Sweden




reply via email to

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