Problem it solves
A recent investigation of stability issues when using adaptive dt with particle collisions indicates that the issue may be that adaptive time stepping does not take collisions into account. Because of this, collisions are not getting well resolved and the stiffness of collisions is often too low.
Proposed solution
We should add some subroutine that computes the the new dt based upon the collision time and if particles are colliding. The dt used should then be the minimum of the dt computed from CFL and collisions. Most-likely, we want the collisions to be resolved over some fixed number of time steps, and we want the dt to slowly increase if there are no collisions back to the time step from CFL.
Problem it solves
A recent investigation of stability issues when using adaptive dt with particle collisions indicates that the issue may be that adaptive time stepping does not take collisions into account. Because of this, collisions are not getting well resolved and the stiffness of collisions is often too low.
Proposed solution
We should add some subroutine that computes the the new dt based upon the collision time and if particles are colliding. The dt used should then be the minimum of the dt computed from CFL and collisions. Most-likely, we want the collisions to be resolved over some fixed number of time steps, and we want the dt to slowly increase if there are no collisions back to the time step from CFL.