Above is one of Steve’s pictures of a typical solution with its characteristic ‘stripes’. Cheyne started trying to identify these stripes as locations where for some cutoff for example He made some nice 3d views of illustrating the problems with this. As he explains:
So there is the tradeoff that if I make the cutoff too high, the sections that look greenish are getting identified as stripes. If I make it too low, the points near where two stripes merge disappear. I attached some 3D plots showing the landscape of du/dx reiterating this point.
The disappearance of the stripes as they merge is unavoidable to a certain extent. The plots in the PowerPoint show how even when there is no cutoff, there is some gap between the merging stripes. There may be some characteristic length scale needed to qualify them as merging.
You can click on these pictures to make them bigger.