Abstract
We revisit the classical problem due to Weyl, as well as its generalisations, concerning the isometric immersions of $\mathbb{S}^2$ into simply-connected $3$-dimensional Riemannian manifolds with non-negative Gauss curvature. A sufficient condition is exhibited for the existence of global $C^{1,1}$-isometric immersions. Our developments are based on the framework à la Labourie (Immersions isométriques elliptiques et courbes pseudo-holomorphes, J. Diff. Geom. 30 (1989), 395--424) of studying isometric immersions using $J$-holomorphic curves. We obtain along the way a generalisation of a classical theorem due to Heinz and Pogorelov.
Introduction
The "Weyl Problem" stands as a cornerstone of differential geometry. Originally posed by Hermann Weyl in 1916, it asks whether a given Riemannian metric $g$ on a sphere $\mathbb{S}^2$ with positive Gauss curvature $K$ can be realized as a surface in 3D Euclidean space. While the case for strictly positive curvature ($K > 0$) was resolved by Nirenberg, Heinz, and others using elliptic PDE theory, the "degenerate" case—where the Gauss curvature is non-negative ($K \ge 0$) but may vanish at certain points—presents a much steeper challenge.
When the curvature vanishes, the equations governing the immersion lose their ellipticity, which can lead to singularities or a loss of regularity in the resulting surface. This paper revisits this problem, extending the scope to general 3D Riemannian manifolds and providing a new sufficient condition for the existence of global $C^{1,1}$-isometric immersions.
The Framework of J-holomorphic Curves
To tackle the non-negative curvature case, the author utilizes the framework established by Labourie (1989). This approach translates the geometric problem of isometric immersions into the language of $J$-holomorphic curves (pseudo-holomorphic maps).
By viewing the 1-jet of the immersion as a map into a fiber bundle equipped with a specific almost complex structure, the problem of finding an immersion becomes equivalent to finding a $J$-holomorphic curve. The author specifically focuses on the mean curvature $H$ of the surface. By differentiating a first-order PDE for $H$ derived by Labourie, the author obtains a second-order elliptic PDE in divergence form for the inverse mean curvature, $W = 1/H$.
Technical Innovation: The Harnack Inequality
The core technical advancement in this work is the application of a Harnack inequality to the function $W$. In the study of PDEs, a Harnack inequality provides a way to bound the maximum value of a solution by its minimum value over a local domain.
In this geometric context, the inequality prevents the mean curvature from blowing up in an uncontrolled manner across the surface. This allows the author to establish a "curvature dichotomy" (Theorem 3.1): for a sequence of approximate immersions, either the mean curvature remains bounded, or the product of the mean curvature and the square root of the Gauss curvature ($H \cdot \sqrt{K}$) remains bounded. This dichotomy is essential for proving that the limit of these approximations is a well-behaved $C^{1,1}$ surface.
Key Results and Principal Curvature Balancing
The paper's primary contribution is Theorem 1.1, which establishes a precise relationship between the two principal curvatures, $\kappa_1$ (the smaller) and $\kappa_2$ (the larger). The author proves that a $C^{1,1}$ immersion is guaranteed to exist if the principal curvatures satisfy a specific blow-up rate:
$$\kappa_2(z) \leq \frac{C_0}{\sqrt[3]{\kappa_1(z)}}$$
This condition, known as principal curvature balancing, ensures that as the intrinsic curvature approaches zero (and $\kappa_1$ potentially vanishes), the larger curvature $\kappa_2$ does not grow so fast that it destroys the regularity of the surface.
Furthermore, the paper generalizes a classical theorem by Heinz and Pogorelov. While previous versions of this theorem required the ambient manifold to be a "space form" (a manifold with constant sectional curvature) or required high regularity ($C^{3,\alpha}$), this work demonstrates that $C^3$ regularity is sufficient even for general Riemannian manifolds, provided the manifold's sectional curvature is bounded above by the surface's intrinsic curvature.
Conclusion and Future Directions
By achieving $C^{1,1}$ regularity—where the surface has a continuous first derivative that is also Lipschitz—this paper provides a robust sufficient condition for solving the Weyl problem in the degenerate $K \ge 0$ case.
However, a significant gap remains. A long-standing question in the field is determining the necessary and sufficient conditions for a smooth ($C^\infty$) isometric embedding when the Gauss curvature is non-negative. While this work provides a path to $C^{1,1}$ regularity, the jump to $C^\infty$ remains an open challenge due to potential geometric obstructions. Future research may look to extend these $J$-holomorphic methods to different surface topologies, such as the torus, or to higher-order regularity estimates.
Summary updated