Strategies for simulating long duration experiments

Hello fellow LAMMPS-users,

In the lab people have done experiments investigating the compaction of salt by imposing a fixed load onto a sample consisting of many thousands of 1-100 micron sized particles, and measure the axial displacement over a period of many days/weeks. For my PhD research I’d like to simulate these experiments using DEM, but this creates a problem: if I enter realistic values for particle radius, stiffness, density, etc., the critical time step reduces to a value somewhere in the range of 1e-8 seconds. You can imagine that if I want to simulate a week-long experiment, it’d take quite a while for the simulation to finish (back-of-the-envelope estimate: 2600 years and a bit).

What are the strategies that one could employ to tackle this problem? Is there a tried & tested method to increase the time step to somewhere closer to 1e-3 s? I’d love to hear your suggestions!


Martijn van den Ende

This is a generic issue with DEM for stiff particles. I would start

by looking in the literature for how people handle it. You might
also post to the LIGGGHTS group and see what they say, since
they have users doing a variety of DEM modeling.


I am not familiar with many people doing very long time simulations for DEM (though many researchers will have similar interests as there is an obsession with relating grains to glass and vice-versa). We can usually get speed up using a few non-dimensional arguments. For example, (at least in dense systems with well controlled geometries) it is usually found that if stiffness is high enough stress is not significantly effected by it. This is really due to the fact that force statistics and force chain structures are fairly universal. So we can get away with using softer particles, which are faster to simulate. Another argument for dilute systems often used is that the stiffness/collision time-scale is MANY orders of magnitude smaller than any deformation time-scale, so you may use softer particles.

You lack any such inherent time-scale. From the long-time nature of your question, I imagine you are more interested in critical/rupture type compaction and ‘aging’. These, I believe, are less studied despite the focus on the topic. I imagine these events are sensitive to particle stiffness and friction, since you are in a very jammed state. So the usual arguments are probably off the table.

Shot in the dark, but you might look into how people who do things like glasses, gels, and other quenched type systems. You may get some insight as to how to speed stuff up from their strategies.

Thanks for you suggestions and explanation. I’ll look some more into glass and gels and post their solutions to this mailing list if I find anything interesting…


Martijn van den Ende