particles go out of domain

kunzhang's picture
Submitted by kunzhang on Wed, 02/19/2014 - 19:42

Hi All,

Could anyone please discuss the reasons that particles go out of domain during simulation? The volume could increase quickly if particles are not in the domain.

Thanks!

Kun

Forums:

RjwV | Thu, 02/20/2014 - 09:46

Hello Kun,

The only reason I've seen this phenomenon before was high particle velocities. If the particle velocities are higher than a certain value, the collision list update of the walls cannot detect the particles anymore as collision partner, because they moved past the 'skin' (detection zone for collision partners) before being detected.

I.e. a solution to this problem would be decreasing particle velocities, or, if these high particle velocities are inherent to your system, decrease the skin value.

Kind regards,
Ramon

kunzhang's picture

kunzhang | Thu, 02/20/2014 - 17:26

Hi Ramon,

Thanks for your replay. One question is how to decrease the skin value in the in.** files?

Also, probably, another reason for my case is that the time step is large, I tried to decrease the time step, and it works! However, lower time step will take more time to run the simulation. I doubt that if large time step is used, the updated the coordinate of particle might be out of the domain, resulting in this problem.

kunzhang

RjwV | Thu, 02/20/2014 - 17:57

I am not sure what large and small time-steps are according to you, just make sure your Courant number is below 1, preferrably even lower (personally I try to stay below 0.1 is possible, but if necessary the treshold can be higher).

I am not sure how exactly to adjust the skin value, I do know you can do it in the liggghts input file. The documentation should be clear on how exactly to implement this. Just make sure your particles do not travel further than the skin distance in one time step.

Hope this helps!

Kind regards,
Ramon

kunzhang's picture

kunzhang | Fri, 02/21/2014 - 17:40

Hi Ramon,

Sure, I agree with that particles should not travel further than the skin distance in one time step. I use 0.0005 for small time step and 0.005 for large time step.

Also, for insertation of particles, overlap check is "yes" probably would be better choice. If it is "no" and insert too much particles in region, particles will "explode" at first.

Thanks for sharing your ideas!
Kun

kunzhang