The Ginzburg-Landau system with general potential: maximum principle and gradient estimates

Technical Summary: The Ginzburg-Landau System with General Potential: Maximum Principle and Gradient Estimates

Problem Statement
The paper investigates the critical points of the Ginzburg-Landau energy functional $F_\varepsilon$ defined on a bounded domain $\Omega \subset \mathbb{R}^M$ ($M, N \ge 2$) with values in $\mathbb{R}^N$:
$$F_\varepsilon[u] = \int_\Omega \left[ \frac{1}{2}|\nabla u|^2 + \frac{1}{2\varepsilon^2} W(1 - |u|^2) \right] dx,$$
where $\varepsilon > 0$ is a small parameter and $W$ is a non-negative potential function. The study focuses on a family of critical points $(u_\varepsilon)$ satisfying Dirichlet boundary conditions of unit length ($u_\varepsilon|_{\partial\Omega} \in S^{N-1}$) that converge, as $\varepsilon \to 0$, to a stationary $S^{N-1}$-valued harmonic map $u_0$.

A central challenge addressed is the behavior of these critical points under general conditions on the potential $W$. Unlike classical literature which often assumes the standard quadratic potential $W(t) = \frac{1}{2}t^2$ or requires $u_\varepsilon$ to be energy minimizers, this work considers general critical points and potentials that may exhibit super-quadratic behavior near their zero set. This generality is motivated by physical models such as Landau-de Gennes theory for liquid crystals and Blume-Capel models in ferromagnetics, which may involve first-order phase transitions.

Methodology
The authors employ a combination of elliptic regularity theory, maximum principle arguments, monotonicity formulas, and barrier constructions.

1. Maximum Modulus Principle: To establish the global bound $|u_\varepsilon| \le 1$, the authors construct a family of entire radial solutions to the scalar Ginzburg-Landau equation. These solutions serve as barriers to deduce bounds for the modulus of the vectorial critical point $u_\varepsilon$. This approach avoids the standard requirement that $W'(t) \le 0$ for $t < 0$, which is often assumed in simpler settings.
2. Monotonicity Formula: The authors derive a monotonicity formula for the scaled energy density $\phi_{x_0}(r) = r^{-(M-2)} \int_{\Omega \cap B_r(x_0)} e_\varepsilon(u) dx$. This formula is adapted to handle boundary points, utilizing a Pohozaev-type identity and a boundary Rellich-type estimate to control the energy concentration near $\partial\Omega$.
3. Gradient and Laplacian Estimates: The proof of uniform gradient bounds relies on the continuity of the limit map $u_0$. The authors show that the continuity of $u_0$ implies small energy concentration at all scales. This smallness allows for the application of Bochner's inequalities and the control of the Laplacian of the modulus and the director field.
4. Differential Inequalities: To bound the Laplacian $\Delta u_\varepsilon$ (and consequently the potential term $\varepsilon^{-2}W'(1-|u_\varepsilon|^2)$), the authors analyze a semi-linear differential inequality satisfied by the potential energy density. They distinguish between the case $\alpha=1$ (linear, solvable via Bessel functions) and $\alpha > 1$ (semi-linear, requiring techniques analogous to the Loewner-Nirenberg equation in conformal geometry).

Key Assumptions
The potential $W \in C^2((-\infty, 1])$ is assumed to satisfy:

• $W(0) = 0$ and $W(t) \ge 0$ for $t \le 0$.
• $W(t) > 0$ for $t \in (0, 1]$.
• Local non-degeneracy near $t=0$: $W'(t) \sim c_0(\alpha+1)t|t|^{\alpha-1}$ as $t \to 0$ for some $\alpha \ge 1$, and $W''(t) \ge 0$ near $t=0$.
  These assumptions allow for potentials that are not necessarily convex globally and accommodate both second-order ($\alpha=1$) and first-order ($\alpha > 1$) phase transitions.

Main Results

1. Global Maximum Principle (Theorem 1.1):
   Under the weak assumption that $W(0)=0$ and $W(t) \ge 0$ for all $t \le 1$, every critical point $u_\varepsilon \in H^1 \cap L^\infty$ with unit-length boundary data satisfies the global uniform bound $|u_\varepsilon| \le 1$ in $\Omega$. This result holds even if $W$ is not convex or if $W'(t) > 0$ for some $t < 0$, ruling out counter-examples where the modulus exceeds 1.

2. Uniform Gradient and Laplacian Estimates (Theorem 1.3):
   Assuming the limit $u_0$ is a smooth stationary harmonic map and the boundary data is $C^{1,\beta}$, the authors prove that the family of critical points $(u_\varepsilon)$ satisfies uniform bounds in $C^{1,\beta'}(\Omega)$ for any $\beta' < \beta$. Specifically:
   $$\|\Delta u_\varepsilon\|_{L^\infty(\Omega)} + \|u_\varepsilon\|_{C^{1,\beta}(\Omega)} \le C,$$
   where $C$ is independent of $\varepsilon$. Furthermore, $u_\varepsilon$ converges to $u_0$ in $C^{1,\beta'}(\Omega)$ as $\varepsilon \to 0$.

Significance and Claims
The paper claims to provide the necessary analytical framework to study the existence and uniqueness of critical points for the Ginzburg-Landau system with general potentials, a topic addressed in the authors' companion paper [25].

• Robustness of the Maximum Principle: The authors highlight that their maximum principle is new in the context of general potentials. It demonstrates that the accuracy of the potential $W$ for large $|u|$ (where the Taylor expansion used to derive $W$ might deviate from physical reality) is irrelevant for critical points; only the behavior near the isotropic state ($u=0$) and the preferred states ($|u|=1$) matters.
• Topological Framework: The uniform $C^{1,\beta}$ estimates established in Theorem 1.3 are presented as the crucial topological framework required to apply implicit function theorem arguments for proving existence and uniqueness of solutions near a non-degenerate harmonic map.
• Generality: By moving beyond the standard quadratic potential and energy minimizers, the results accommodate a broader class of physical phenomena, including first-order phase transitions, while maintaining rigorous control over the solution's regularity and bounds.

The paper concludes that these estimates are essential for bridging the gap between the $\varepsilon$-dependent Ginzburg-Landau problem and the limiting harmonic map problem, particularly when dealing with non-minimizing critical points and complex potential landscapes.