3D can be rescaled by maximum. 1D seems perfect up until material conservation kicks in at boundary.

Probably need proper dependence on phi_tot instead of direct dependence on x.

But only up to a factor between different partitionings/diffusion coefficients outside.
Problem: Maybe need different concentration scaling for radial problem?

Take care of volume elements everywhere! Needs to be in all integrals (also t) and mass conservation check.

This reverts commit 9e00834c.