This image by Gyorgy Doczi gave me an idea, that maybe a conformal mapping from a logarithmic diagonal mesh from its circle boundary to a square could be interesting to try.
I thought it was interesting that relaxing the mesh gets the circles pretty close to tangent throughout. Adding the isothermic goal solves it for the centermost cells as well.
Happy with the symmetry after remapping to the square, but this is only a start. Would be nice to find a way to pack the seams and remove the obvious hole in the center in favor of something more distributed across a longer rectangle - closer to the image in the original post. The radii do matter - I’d like to be able to control the minimum and maximum, with the former being more important. Any ideas are welcome and appreciated!
