MCF for a dumbbell

Uwe F. Mayer

System:	Dell Dimension 3000 Intel Pentium 4 @ 2.8GHz
Computation time:	About 6 minutes to neck singularity; about an hour afterwards
Scale:	Main radius = 0.6307, neck radius = 0.1, length = 1.7561, time step = starting with 1E-9, then set to 1E-11 just before the neck singularity, and various settings from 1E-12 to 1E-9 after the neck singularity
Shape:	Neck is modeled on y=x^5+0.1, at x=0.75 slope is matched to have a circular finish
Vertices:	Initially 96 points on 1/2 of generating curve of radially symmetric surface, at about t=.005075 (which is after the neck singularity) the coarseness was reset to speed up calculations, and only about 40 points remained, which ultimately all were combined with their neighbors until the component vanished
Segment threshold:	Eliminated segment if length is less than 0.001 before the neck singularity and for a while thereafter, after coarseness resetting if length is less than 0.05
Comment:	All numerics were done on the generating curve, which is significantly faster than working on a triangulated surface; the generating curve was manually separated at the singularity (y-value of point on y-axis was set from 5E-5 to 0; the computation used the symmetry and only worked with half of the generating curve; the re-setting of the computations to run fewer points likely will have increased the numeric error but should not invalidate the fundamental simulation results