Logarithmic diffusion ($u_t=\kappa \Delta \log u$)

For a positive density $u$, logarithmic diffusion is the nonlinear diffusion equation $u_t=\kappa \Delta \log u$, equivalently $u_t=\kappa \mathrm{div}(u^{-1}\nabla u)$. Its diffusion coefficient is inversely proportional to the density. In one dimension, a mass-$M$ self-similar solution isIts integral is $M$, and substitution verifies the equation. This is $M$ times a Cauchy distribution density of scale $2\pi \kappa t/M$. It tends weakly to $M$ times the Dirac delta distribution as $t\downarrow 0$; the slow algebraic tails illustrate the enhanced diffusion where density is small.