Optimal boundary regularity of harmonic maps from $RCD(K,N)$-spaces to $CAT(0)$-spaces

Hui-Chun Zhang, Xi-Ping ZhuarXiv 2025 · DOI · arXiv

TL;DR

This paper establishes optimal boundary regularity estimates for harmonic maps between generalized metric spaces (RCD(K,N)-spaces) and spaces with non-positive curvature (CAT(0)-spaces). The key contribution is a new gradient estimate that is "optimal" in the sense that it matches known results for simpler cases, even when the domain's boundary is not smooth and the boundary data is only Lipschitz continuous. This work extends fundamental results in geometric analysis to a broader class of spaces.

What Problem Does This Paper Address?

The paper tackles the problem of understanding how "smooth" harmonic maps can be near the boundary of their domain, especially when the domain itself is not perfectly smooth (e.g., has corners or rough edges) and the prescribed boundary values are not perfectly smooth either (e.g., only Lipschitz continuous).

Harmonic maps are generalizations of harmonic functions and play a crucial role in geometry and physics. For maps between smooth manifolds with smooth boundaries and smooth boundary data, their regularity near the boundary is well-understood (Schoen-Uhlenbeck, 1983). However, extending these results to more general settings, particularly to "singular" spaces (like those with Ricci curvature bounds but not necessarily smooth) and with less regular data, is a significant challenge. This paper focuses on the specific case of RCD(K,N)-spaces (a synthetic notion of lower Ricci bounds) mapping into CAT(0)-spaces (spaces with non-positive curvature). The goal is to find the best possible bound on how quickly the gradient of the harmonic map can grow as it approaches the boundary.

Methods

The authors develop several key technical tools:

  • Gauss-Green Formula for RCD Spaces: A central part of the work is establishing a version of the Gauss-Green formula that holds for RCD(K,N)-spaces. This formula relates integrals over a domain to integrals over its boundary and is crucial for relating the map's behavior inside to its behavior at the boundary. They develop a new version that is particularly useful for domains with "one-sided approximations."
  • Heat Kernel and Green Function Estimates: They derive sharp upper bounds for the Dirichlet heat kernels and Green functions on these RCD(K,N)-spaces, especially near the boundary. These estimates are essential for controlling the behavior of solutions to related parabolic and elliptic equations.
  • Barrier Functions: The proof heavily relies on constructing appropriate "barrier functions" for heat equations, which are then used to control the growth of solutions near the boundary.
  • Approximation Techniques: To handle the lack of smoothness in the domain and boundary data, they employ sophisticated approximation techniques, including using sequences of smooth functions to approximate functions with bounded variation and to approximate the domain itself.
  • Extension of Existing Results: The paper builds upon and extends existing results on harmonic maps and regularity theory in metric spaces, particularly the work of Serbinowski, Jost, and others, adapting them to the RCD(K,N) framework.

Datasets

Not discussed in the paper. This is a theoretical mathematics paper, not an empirical study.

Key Insights and Ideas

  • Optimality of the Logarithmic Growth: The paper proves that the logarithmic growth of the gradient near the boundary, observed in classical examples for harmonic functions, is indeed the optimal bound for harmonic maps in this generalized setting. This is a significant result as it suggests that even with rougher spaces and data, the fundamental behavior of harmonic maps near boundaries remains similar to the well-understood Euclidean case.
  • Bridging Smooth and Singular Settings: The work successfully bridges the gap between the well-established regularity theory for smooth spaces and the more challenging theory for singular metric spaces. The development of a Gauss-Green formula and heat kernel estimates in the RCD(K,N) framework is a major step in this direction.
  • The Power of Synthetic Notions: The use of RCD(K,N) spaces highlights the power of synthetic notions of curvature in extending geometric analysis results to a very broad class of spaces, including spaces that are not necessarily smooth manifolds.
  • Technical Sophistication: The paper showcases a high level of technical sophistication in dealing with measure theory, PDEs on metric spaces, and geometric analysis. The construction of approximations and the careful handling of inequalities are crucial.

Key Results

The main result is Theorem 1.8, which provides an optimal boundary gradient estimate for harmonic maps $u: \Omega \to Y$ where $\Omega$ is a bounded domain in an RCD(K,N)-space and $Y$ is a CAT(0)-space. The estimate states that the Lipschitz constant of $u$ near the boundary grows at most logarithmically with the distance to the boundary:

$$ \mathrm{Lip}_u(x) := \limsup_{y \to x, y \neq x} \frac{d_Y(u(x), u(y))}{d(x, y)} \leq c_1 L \cdot \ln\left(\frac{2e \cdot \mathrm{diam}(\Omega)}{\delta(x)}\right) $$

where $L$ is the Lipschitz constant of the boundary data $u_0$, $\delta(x)$ is the distance from $x$ to the boundary $\partial \Omega$, and $c_1$ is a constant depending on the parameters of the spaces.

This result is shown to be optimal by an example analogous to the Hardy-Littlewood example for harmonic functions.

Limitations and Future Work

  • Assumptions on the Domain: The results rely on the domain $\Omega$ being perimetrically regular and satisfying a uniformly exterior ball condition. While these are more general than Lipschitz domains, they still impose structural constraints.
  • Target Space: The target space is assumed to be a CAT(0)-space. The paper briefly discusses extensions to CAT(κ)-spaces for κ > 0, but this is not the main focus.
  • Data Regularity: The boundary data $u_0$ is assumed to be Lipschitz continuous. Extending the results to less regular data would be a natural next step.
  • Generalization of Gauss-Green: The paper develops a specific version of the Gauss-Green formula. Further exploration of its properties and applications in RCD spaces could be fruitful.
  • Further Applications: The techniques developed in this paper could potentially be applied to other problems in geometric analysis on RCD spaces, such as the regularity of other types of geometric flows or solutions to PDEs.

Summary updated

Abstract

In 1983, Schoen-Uhlenbeck \cite{SU83} established boundary regularity for energy-minimizing maps between smooth manifolds with the Dirichlet boundary condition under the assumption that both the boundary and the data are of $C^{2,α}$. A natural problem is to study the qualitative boundary behavior of harmonic maps with rough boundary and/or non-smooth boundary data. For the special case where $u$ is a harmonic function on a domain $Ω\subset \mathbb R^n$, this problem has been extensively studied (see, for instance, the monograph \cite{Kenig94}, the proceedings of ICM 2010 \cite{Tor10} and the recent work of Mourgoglou-Tolsa \cite{MT24}). The $W^{1,p}$-regularity ($1
BibTeX
@misc{zhang2025optimal,
  author = {Hui-Chun Zhang and Xi-Ping Zhu},
  title = {{Optimal boundary regularity of harmonic maps from $RCD(K,N)$-spaces to $CAT(0)$-spaces}},
  year = {2025},
  eprint = {2505.02442},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2505.02442},
}
Figure