Problem with neighbour list

Submitted by masood on Tue, 11/17/2015 - 11:23

Hi,

I am trying to simulate mixing of two different particles (the small one is 5 micron and the large one is 100 micron) and also the simulation domain is in the order of a couple of centimetre.

I am not sure about the skin distance and bin size that I should use in the "neighbor" and "neigh_modify" commands.
As Andreas said during the course I should set the skin distance based on the large particle, so I set it like that:
neighbor "100 micron" bin
neigh_modify every 1000 delay 1000 check yes
But it eventually crashes after nearly 1000 time steps and the error is "Too many neighbor bins"
Then I tried to decrease the skin distance, but it doesn't work too.
Would you please help me with this case?

Best
Masood

ckloss's picture

ckloss | Thu, 12/10/2015 - 22:31

Hi Masood,

5/100 is a ratio of 20 - which is quite a lot and performance will suffer. We have an alpha-alpha version of the so called H-grid neigh list algorithm that can deal with it, but it's too early/buggy to release it

best wishes
Christoph

iluvatar | Mon, 11/26/2018 - 03:25

Hi, following the previous comment I would like to ask if currently there is a "recommended" way, like the h-grid, to handle systems with large polydispersities. In my case I am exploring systems with size ratio up to 32 (=max diameter/min diameter). For now I am working with size ratio up to 16, and I had to increse the parameter "one" for neigh_modify, but maybe I am not using liggghts efficiently for this problem. Thanks in advance.