Remarks on the heat flow of harmonic maps into CAT(0)-spaces

This paper provides an elementary proof, inspired by Korevaar and Schoen, of the local Lipschitz regularity for suitable weak solutions of the heat flow of harmonic maps from a complete Riemannian manifold with positive injectivity radius and bounded curvature into any CAT(0)-metric space.

The Gist

Here is an explanation of the paper "Remarks on the Heat Flow of Harmonic Maps into CAT(0)-Spaces" using simple language, analogies, and metaphors.

The Big Picture: Smoothing Out a Crumpled Sheet

Imagine you have a piece of fabric (this is your domain, like a sphere or a flat sheet) and you are trying to stretch it over a strange, bumpy, or curved landscape (this is your target space).

In mathematics, a "Harmonic Map" is the most efficient, relaxed way to stretch that fabric over the landscape without wrinkling it unnecessarily. It's like the fabric finding its natural, lowest-energy resting position.

Now, imagine the fabric is currently crumpled or twisted. The Heat Flow is a process where you gently heat the fabric and let it relax over time. As time passes, the crumples smooth out, and the fabric settles into that perfect, relaxed shape.

The Problem:
For a long time, mathematicians knew how to smooth out fabric if the landscape was a nice, smooth surface (like a sphere or a flat plane). But what if the landscape is weird? What if it has sharp corners, is made of a grid of triangles, or has "negative curvature" (like a saddle shape that curves away in all directions)? These are called CAT(0) spaces.

In these weird landscapes, the fabric might get stuck, tear, or behave unpredictably. A few years ago, the authors (Lin, Segatti, Sire, and Wang) proved that the fabric does eventually smooth out, but their proof was like a heavy, complex machine: it used a very complicated method called "elliptic regularization" to get there.

The Goal of This Paper:
Lin and Wang (the authors of this specific paper) wanted to show that the fabric smooths out using a much simpler, more direct method. They wanted to prove that the fabric doesn't just smooth out, but that it becomes locally Lipschitz.

• What does "Locally Lipschitz" mean? In plain English, it means the fabric becomes smooth enough that it doesn't have any sudden, infinite spikes or jagged tears. If you zoom in on any small part of the fabric, it looks like a gentle slope, not a cliff.

---

The Two Main Ingredients of the Proof

The authors use two clever tricks to prove this smoothness. Think of these as two different tools in a toolbox.

Tool 1: The "Rubber Band" Trick (The EVI)

The authors use a concept called the Evolution Variational Inequality (EVI).

• The Analogy: Imagine you have a rubber band stretched between two points. If you pull one point slightly, the tension changes in a predictable way.
• How it works: The authors look at the "distance" between the fabric's current shape and a slightly shifted version of itself. By using the special geometry of the CAT(0) landscape (which acts like a giant, perfect rubber sheet that always pulls things together), they can write down a rule that says: "The energy of the fabric cannot jump up suddenly; it must flow down smoothly."
• The Result: This rule allows them to track how the "slope" of the fabric (how steep it is) changes over time.

Tool 2: The "Heat Detector" (Sub-Caloricity)

The second part of the proof involves looking at how fast the fabric is moving as it smooths out.

• The Analogy: Imagine the fabric is a hot metal plate. If you touch a spot, you can feel how hot it is. In math, we look at the "speed" of the fabric changing ($|\partial_t u|^2$).
• The Discovery: The authors proved that this "speed" behaves like heat. If a spot is moving fast, the heat (energy) around it will naturally spread out and cool down. It can't stay super-hot (super-fast) forever in one tiny spot without leaking energy to the neighbors.
• The Result: Because the "speed" is controlled by this heat-like spreading, the fabric can't suddenly jerk or tear. It has to move smoothly.

Putting It Together: The Domino Effect

Here is how the authors connect these two tools to prove the fabric is smooth:

1. Step 1: They prove that the speed of the fabric's movement is controlled (it doesn't explode). This is like saying the fabric isn't vibrating violently.
2. Step 2: Because the speed is controlled, they can use a famous mathematical technique (Moser's method, named after a mathematician who studied heat) to prove that the slope (how steep the fabric is) is also controlled.
3. Conclusion: If the speed is calm and the slope is gentle, the fabric is Lipschitz continuous. In everyday terms: The fabric is smooth.

Why Does This Matter?

Before this paper, we knew the fabric would eventually settle, but we didn't have a simple, elementary way to prove it wouldn't tear or develop sharp corners along the way.

• The "Old Way": Was like building a massive, complex factory to bake a cake. It worked, but it was hard to understand why the cake rose.
• This Paper: Is like showing that if you just put the batter in a warm oven, the physics of heat and the shape of the pan guarantee the cake will rise perfectly.

This new, simpler proof works for any weird landscape (CAT(0) space) and any starting shape, as long as the starting shape isn't infinitely crumpled. It confirms that nature (or mathematics) has a built-in tendency to smooth things out, even in the most bizarre geometries.

Summary in One Sentence

The authors found a simpler, more intuitive way to prove that when you try to smooth out a crumpled sheet over a weird, bumpy landscape, the sheet will naturally settle into a perfectly smooth shape without tearing, using the natural laws of "heat" and "distance" to guarantee the result.

Technical Summary

Here is a detailed technical summary of the paper "Remarks on the Heat Flow of Harmonic Maps into CAT(0)-Spaces" by Fang-Hua Lin and Changyou Wang.

1. Problem Statement

The paper addresses the regularity theory for the heat flow of harmonic maps where the target space $(X, d)$ is a CAT(0)-metric space (a complete, separable metric space with non-positive curvature in the sense of Alexandrov) and the domain $(M, g)$ is a complete Riemannian manifold with positive injectivity radius and bounded curvature.

While the existence of global "suitable weak solutions" to this flow was recently established by the authors (with Segatti and Sire) via an elliptic regularization approach (minimizing a Weighted Energy Dissipation functional), the local Lipschitz regularity of these solutions remained a challenging open problem. Previous results on regularity were limited to specific targets (like Euclidean buildings) or required different, more complex methods (such as viscosity solutions and sup-convolutions used by Zhang and Zhu).

The primary goal of this paper is to provide an alternate, elementary proof of the local Lipschitz regularity for suitable weak solutions in the general setting of CAT(0) targets and Riemannian domains, without relying on the heavy machinery of viscosity theory.

2. Methodology

The authors' approach is inspired by the work of Korevaar and Schoen on minimizing harmonic maps but is adapted for the parabolic (heat flow) setting. The core methodology relies on deriving parabolic inequalities for the energy densities of the solution.

A. Evolution Variational Inequality (EVI)

The proof is built upon the definition of a "suitable weak solution" $u$, which satisfies the Evolution Variational Inequality (EVI):
$$\frac{1}{2}\frac{d}{dt}d^2(u(t), v) + E(u(t)) \leq E(v) \quad \text{in } \mathcal{D}'(0, \infty)$$
for all $v \in H^1(M, X)$. This inequality serves as the fundamental tool to derive differential inequalities for the energy.

B. Derivation of Parabolic Inequalities

The authors establish two key differential inequalities in the weak sense:

1. Sub-caloricity of Time Derivative ($|\partial_t u|^2$):
   By considering the time-shifted solution $v(x,t) = u(x, t+\delta)$ and applying the EVI, the authors show that $|\partial_t u|^2$ satisfies:
   $$(\partial_t - 2\Delta)|\partial_t u|^2 \geq 0$$
   This implies that $|\partial_t u|^2$ is a sub-caloric function. Consequently, standard parabolic Harnack inequalities (Moser's iteration) can be applied to obtain local $L^\infty$ bounds for $|\partial_t u|^2$.

2. Parabolic Inequality for Spatial Gradient ($|\nabla u|^2$):
   Using the EVI and Reshetnyak's quadrilateral comparison property (a defining feature of CAT(0) spaces), the authors derive a coupled inequality for the spatial energy density $|\nabla u|^2$. Specifically, they show that for any non-negative test function $\eta$:
   $$\iint |\nabla u|^2 (\partial_t \eta + 2\Delta \eta + C\eta + C|\nabla \eta|) \, dV_g dt \geq -C \iint \eta |\partial_t u|^2 \, dV_g dt$$
   This inequality indicates that $|\nabla u|^2$ behaves like a subsolution to a parabolic equation, provided the source term $|\partial_t u|^2$ is controlled.

C. Moser Iteration

With the $L^\infty$ bound on $|\partial_t u|^2$ established in Step 1, the authors apply Moser's iteration technique (originally developed for the heat equation) to the inequality derived in Step 2. This allows them to bootstrap the $L^2$ bound of $|\nabla u|^2$ to an $L^\infty$ bound, thereby proving Lipschitz continuity.

3. Key Contributions

• Elementary Proof of Regularity: The paper provides a self-contained, "elementary" proof of local Lipschitz regularity that avoids the complex sup-convolution and viscosity solution techniques used in recent works (e.g., Zhang-Zhu). It relies instead on the geometric properties of CAT(0) spaces and variational inequalities.
• General Domain and Target: The results hold for any CAT(0) metric space as the target and any complete Riemannian manifold (with positive injectivity radius and bounded curvature) as the domain. This generalizes previous results that were often restricted to Euclidean domains or specific target geometries.
• Uniform Lipschitz Estimates for Regularization: The authors prove a uniform Lipschitz estimate for the minimizers $u_\varepsilon$ of the elliptic regularization functional (Theorem 1.4). This answers a specific question regarding the extension of previous results to general CAT(0) spaces and provides explicit bounds depending on the regularization parameter $\varepsilon$.
• Parabolic Frequency Function Connection: While the main focus is on the new proof, the paper contextualizes its results within the framework of parabolic frequency functions introduced in their previous work, linking the monotonicity of these functions to regularity.

4. Main Results

Theorem 1.3 (Local Lipschitz Regularity):
Let $(X, d)$ be a CAT(0)-metric space and $(M, g)$ be a complete Riemannian manifold with positive injectivity radius and bounded curvature. Let $u$ be any suitable weak solution of the heat flow of harmonic maps with initial data $u_0 \in H^1(M, X)$.

• Then $u$ is locally Lipschitz continuous on $M \times (0, \infty)$.
• Furthermore, for any $\delta > 0$, there exists a constant $C$ (depending on $\delta$, the initial energy $E(u_0)$, the injectivity radius, and curvature bounds) such that:
  $$|\partial_t u|^2 + |\nabla u|^2 \leq C \quad \text{in } M \times [\delta, \infty).$$

Theorem 1.4 (Uniform Lipschitz Estimate for Regularization):
For the domain $(M, g) = (\mathbb{R}^n, dx^2)$, let $u_\varepsilon$ be the minimizer of the Weighted Energy Dissipation (WED) functional. The authors establish a uniform bound:
$$\sup_{B_r(x_0) \times (t_0-r^2, t_0+r^2)} |\nabla u_\varepsilon|^2 \leq c \left[ \frac{\varepsilon}{r^{n+2}} + \frac{1}{r^n} \right] E(u_0)$$
for all $0 < \varepsilon \leq r^2$. This confirms that the regularized solutions converge to a Lipschitz solution as$\varepsilon \to 0$.

5. Significance

This paper is significant for several reasons in the field of geometric analysis and partial differential equations:

1. Unification of Methods: It demonstrates that the regularity of harmonic map heat flows into singular spaces (CAT(0)) can be understood through classical parabolic PDE techniques (sub-caloricity and Moser iteration) combined with metric geometry (Reshetnyak's inequality), rather than requiring entirely new analytical frameworks like viscosity solutions.
2. Robustness: The proof is robust enough to handle general Riemannian domains, not just Euclidean ones, which is crucial for applications in geometric group theory and the study of manifolds with curvature bounds.
3. Foundational for Future Work: By establishing a clean, elementary proof of Lipschitz regularity, the paper lays a stronger foundation for studying further properties of these flows, such as uniqueness, long-time behavior, and singularity formation, in the context of non-smooth target spaces.
4. Resolution of Open Questions: It positively answers the question of how to extend the regularity results from the authors' previous work (which relied on frequency functions) to a broader class of spaces using a more direct variational approach.

In summary, Lin and Wang successfully bridge the gap between the existence theory of weak solutions and their regularity in the context of CAT(0) targets, offering a streamlined and powerful analytical toolset for future research in geometric flows.