Dear all,
I had been looking forward to use level sets to keep track of the surface in a simulation, but one thing has me stumped. Since many here work with XFEM, I was hoping someone might have some valuable input.

I want to simulate flows driven by a surface tension. This will typically yield holes, which will close rapidly. At the point at which the hole disappears, the signed distance function will have a discontinuity.
I would expect this to happen in pretty much any fluid simulation, but I can't see any way to handle it.

I have read J.A. Sethian's work, which mostly involves the fast matching method to construct an extension velocity. I can't see how it would handle a very common situation as this (any kind of situation where a free surface disappears). It all builds up from a well defined velocity on the boundary, which doesn't exist in these cases.

There scenarios are much to common to be ignored, so how should I handle it? Any input is appreciated.

/ Mikael