The heuristic explanation of the behavior of phase transition in the Allen–Cahn model describes two 'forces' at play:
- the curvature of the phase interfaces—they each 'want to' minimize length;
- and an attractive 'force' between the interfaces, which decays exponentially with the distance $d$ between them. This interaction is modelled by a Toda system of particles on the real line. (Accounting for it is crucial in the construction (or exclusion) of clustered phase interfaces, starting with the work of del Pino, Kowalczyk and Wei.)
For a function $u: D \to \mathbf{R}$ that is a stationary point of the $\epsilon$-Allen–Cahn functional (working on the unit disc $D \subset \mathbf{R}^2$ for example), so a solution of the PDE \begin{equation} \epsilon^2 \Delta u + u(1-u^2) = 0, \end{equation} these two forces should be of the same order to roughly balance each other out.
How does the attractive force depend on the parameter $\epsilon > 0$? I assumed it was something like $\mathrm{e}^{-d\ln \epsilon}$ or $\mathrm{e}^{-\frac{d}{\epsilon} \ln \epsilon}$, but neither matches the asymptotics obtained by rescaling a fixed solution $u: \mathbf{R}^2 \to \mathbf{R}$ by homotheties.