For a school project, I'm implementing SPH fluid simulation, according to the paper by Müller et al in 2003: "Particle-Based Fluid Simulation for Interactive Applications". I've implemented the calculation of density and pressure, the pressure forces, viscosity forces and gravity. Now that I'm adding a bounding box, things start going wrong.

My response to a collision is to move the particle back to the contact point, reflect its velocity around the normal of the box at the contact point, and damp the magnitude a bit by some bounce factor.

Now I have the following scenario. Two particles, p2 above p1, start by floating somewhere in the bounding box. They are too far away from each other for the pressure or viscosity forces to work on them. So gravity starts pulling them both down. p1 reaches the bottom of the bounding box, bounces a little bit and then stays on the bottom. Now, p2 is still too far away for pressure/viscosity forces and then, within one timestep, p2 hits the bounding box as well and is placed at the bottom. Now p1 and p2 are incredibly close to each other, causing the pressure force to be incredibly large. This makes the particles propel away from each other with extreme speed.

What kind of solution would you suggest? Just decrease the timestep? Use penalty forces instead of projection?

Thanks in advance!