2
$\begingroup$

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.

$\endgroup$

1 Answer 1

2
$\begingroup$

The gradient flow $$\frac{\partial u}{\partial t}=\epsilon^2\Delta u+f(u),\;\;\text{with}\;\;f(u)=u(1-u^2),$$ of the Allen-Cahn functional has no inertia (there is no second order derivative in time), so you cannot associate it with a "force". The $\epsilon$ term governs the width of the interface between two regions where $u=+1$ and $u=-1$, for example, in one dimension $u(x)=\tanh( 2^{-1/2} x/\epsilon).$ This has no interpretation as a "balance of forces" in the Allen-Cahn equation.

An alternative way to formulate the gradient flow, where this interpretation becomes possible, is to introduce inertia with a friction term (see section 2.4 of The Calculus of Variations). Then the evolution equation becomes of second order in time, $$\frac{\partial^2 u}{\partial t^2}+a\frac{\partial u}{\partial t}=\epsilon^2\Delta u+f(u),$$ with $a$ a friction coefficient. Now you can interpret $f(u)$ as a force which attracts towards the fixed points $u=\pm 1$. For the tanh profile it decays as $\exp(-2^{1/2}|x|/\epsilon)$ away from the interface.

$\endgroup$
4
  • $\begingroup$ There may be no force in a rigorous, physical sense, but the interaction between the phase interfaces that I describe is well-established. There's a ton of literature on the subject, but I'll just quote a short passage from the paper of del Pino, Kowalczyk and Wei entitled 'The Toda system and clustering interfaces in the Allen-Cahn equation: 'Considering interfaces, we observe that two nearly parallel interfaces attract [their emphasis], in the sense that as they get closer to one another, energy decreases [...].' I hope this clarifies this point. $\endgroup$
    – Leo Moos
    Commented Jan 5, 2023 at 18:59
  • $\begingroup$ I'm not sure what else to say... Clearly you didn't understand the question - which is fine -, but I think it's quite rude to ignore my previous comment, where I attempted to clear up the misunderstanding. $\endgroup$
    – Leo Moos
    Commented Jan 6, 2023 at 10:42
  • $\begingroup$ apologies, I don't understand the "rude" part; you ask for the dependence of the attractive force on $\epsilon$, doesn't $\exp(-2^{1/2}|x|/\epsilon)$ answer your question? $\endgroup$ Commented Jan 6, 2023 at 21:40
  • $\begingroup$ Sorry, I didn't mean to offend. You misunderstood what I asked - that's fine, no problem. But when I tried to clear up the misunderstanding, you ignored what I wrote and decided to add a paragraph that's even more off-topic that what you'd written originally. That, I find rude. $\endgroup$
    – Leo Moos
    Commented Jan 7, 2023 at 6:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .