🔢 mathematics

Interior Hessian estimates for the quadratic Hessian equation

This paper establishes interior Hessian estimates and smoothness for viscosity solutions of the quadratic Hessian equation on the positive branch in arbitrary dimensions.

Original authors: Zhisu Li, Ke Wu

Published 2026-08-25

📖 5 min read🧠 Deep dive

CC0 1.0

Original authors: Zhisu Li, Ke Wu

Original paper dedicated to the public domain under CC0 1.0 (http://creativecommons.org/publicdomain/zero/1.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the vast landscape of mathematics, there is a branch dedicated to understanding how shapes bend and curve in space. This field, known as the study of partial differential equations, provides the language for describing everything from the flow of heat to the shape of a soap bubble. At the heart of this inquiry are equations that describe the curvature of a surface at every single point. Imagine a smooth hill; the way it curves up or down in different directions is captured by a specific mathematical object called the Hessian. For decades, mathematicians have been trying to solve a particular puzzle involving a type of equation that depends on the product of these curvatures. The challenge has been to prove that if a solution to this equation exists and behaves reasonably well on the outside, it must also remain smooth and well-behaved on the inside, without developing sudden, sharp spikes or breaks. This question is not just a theoretical curiosity; it is a fundamental test of whether the mathematical models we use to describe the physical world are robust enough to handle complex, real-world shapes without falling apart.

For a long time, the answer to this question depended heavily on the number of dimensions the shape existed in. In two dimensions, the problem was solved nearly a century ago. In three dimensions, it took until the late 2000s to find a solution, but only for shapes that were strictly convex, meaning they curved outward everywhere like a perfect sphere. As mathematicians moved into higher dimensions, the problem became significantly harder. Previous attempts to solve it required adding extra, restrictive conditions, such as assuming the shape was convex or that it curved in a specific way. These assumptions acted like training wheels, allowing mathematicians to prove the result, but they left a gap: no one knew if the smoothness held true for the most general, unrestricted cases in higher dimensions. The question remained open: could a solution to this specific curvature equation develop a sudden, infinite spike in its interior, or was it guaranteed to stay smooth?

A team of researchers has now closed this chapter, proving that for this specific type of curvature equation, the solution is always smooth inside, regardless of the dimension or the shape's complexity, provided it stays on a specific "positive branch" where the overall curvature is positive. They demonstrated that if a solution is bounded and well-behaved on the boundary of a region, its internal curvature cannot blow up. This means that the solution is guaranteed to be smooth and predictable everywhere inside, without needing any extra assumptions about convexity. The proof is a rigorous mathematical certainty, not a simulation or a suggestion, establishing a new standard for how these equations behave in any number of dimensions.

To reach this conclusion, the researchers had to navigate a tricky obstacle. In the past, proofs of this nature relied on a specific inequality that worked well for convex shapes but failed when the shape became more complex. The authors realized they could not use the old tools. Instead, they invented a new strategy that involved creating a "comparison" shape. They imagined a second, simpler solution that sat just above the original, unknown solution. The key to their success was proving that these two shapes never touched or got too close to each other in the interior of the region. They showed that there is always a guaranteed, measurable gap between the original solution and this comparison shape.

The researchers achieved this separation by first finding a tiny spot where the gap was known to exist. From there, they used a clever propagation technique to show that if the gap existed in one small area, it had to exist in a slightly larger area nearby. By linking a chain of these overlapping areas together, they proved that the gap extended all the way across the entire interior of the region. This uniform separation was the critical missing piece. It allowed them to construct a specific mathematical barrier that prevented the curvature of the original solution from becoming infinite. By combining this barrier with a known estimate that controls how fast a solution can change near the edge, they were able to derive a strict limit on the curvature everywhere inside.

This work removes the need for the restrictive assumptions that had limited previous results. It confirms that the smoothness of the solution is an inherent property of the equation itself, not a result of the shape being simple or convex. The findings apply to any dimension, from the two-dimensional plane to spaces with many more dimensions. The researchers also showed that their method works for a broader class of equations where the right-hand side is not just a constant number but can vary smoothly. Furthermore, they extended their result to "viscosity solutions," a type of generalized solution that allows for less smooth initial data, proving that even these rougher solutions become perfectly smooth once they enter the interior of the domain.

The implications of this work are foundational. By proving that these solutions cannot develop singularities or infinite spikes in their interior, the researchers have solidified the mathematical framework used to describe complex geometric structures. Their method, which relies on constructing a specific auxiliary function to maintain a separation between solutions, offers a new pathway for tackling similar problems in other areas of geometry and analysis. The result stands as a complete proof, resolving a long-standing open problem and providing a clear, dimension-independent understanding of how these curvature equations behave. It confirms that in the world of these specific mathematical surfaces, order and smoothness are guaranteed, even in the most complex and high-dimensional scenarios.

Technical Summary: Interior Hessian Estimates for the Quadratic Hessian Equation

Problem Statement
This paper addresses the interior regularity of viscosity solutions to the quadratic Hessian equation:
$\sigma_2(D^2u) = \sum_{1 \le i < j \le n} \lambda_i \lambda_j = 1$
in dimension $n \ge 2$ , where $\lambda_1, \dots, \lambda_n$ are the eigenvalues of the Hessian matrix $D^2u$ . The study focuses on solutions lying on the "positive branch," defined by the admissibility condition $\lambda(D^2u) \in \Gamma_2$ , where $\Gamma_2 = \{ \lambda \in \mathbb{R}^n : \sigma_1(\lambda) > 0, \sigma_2(\lambda) > 0 \}$ .

Historically, interior Hessian estimates for this equation have been established under restrictive conditions:

For $n=2$ , the equation reduces to the Monge–Ampère equation, with classical estimates by Heinz.
For $n=3$ , Warren and Yuan utilized special Lagrangian structures.
For arbitrary dimensions, previous results required convexity (Guan–Qiu, McGonagle–Song–Yuan) or semiconvexity (Shankar–Yuan).
Recent works by Qiu, Shankar–Yuan, and Fan have extended these results to specific dimensions (e.g., $n=4$ ) or under dynamic semiconvexity conditions.

The primary objective of this work is to remove both the dimension restriction and the additional convexity assumptions, establishing interior Hessian estimates for arbitrary $n \ge 2$ relying solely on the admissibility condition on the positive branch.

Methodology
The authors employ a novel approach that avoids relying on Jacobi inequalities or doubling arguments for the Hessian, which were central to previous proofs. Instead, the proof strategy centers on constructing a suitable auxiliary function $w$ to which the Chou–Wang Pogorelov estimate applies. The core of the method involves establishing a "uniform interior separation" between the solution $u$ and a comparison solution $v$ .

The technical execution proceeds in four main stages:

Comparison and Seed Separation:
The authors introduce a comparison solution $v$ satisfying $\sigma_2(D^2v) = 1/4$ in a larger ball $B_{3/2}$ with $v=u$ on the boundary. By the comparison principle, $u \le v$ . Using a localized integration by parts argument involving the linearized operator of the average function $\omega = (u+v)/2$ , they establish a "seed separation": there exists a small ball $B_r(x^*)$ where $v - u \ge \varepsilon_0 > 0$ .
Propagation of Separation via Auxiliary Functions:
To extend this lower bound from a small ball to the entire interior domain $B_{1/2}$ , the authors construct a family of auxiliary functions. For any point $y \in B_{1/2}$ , they define a phase function:
$\phi_y(x) = 2((x-y) \cdot Dv(x) - v(x) + v(y)) + \frac{\alpha}{2}|x-y|^2 - 2\beta|Dv(x)|^2$
and an exponential barrier $w_y = e^{(c_y - \phi_y)/\gamma} - 1$ .
The term $-2\beta|Dv|^2$ is critical; it generates a positive quadratic term in the eigenvalues that compensates for cases where a negative eigenvalue might otherwise cause cancellation in the linearized operator. By carefully choosing parameters $\alpha, \beta, \gamma$ , they prove that $v - u \ge \delta$ in $B_r(y)$ implies $v - u \ge \theta \delta$ in $B_{2r}(y)$ . Iterating this along a chain of intersecting balls yields a uniform lower bound $v - u \ge \varepsilon_1$ in $B_{1/2}$ .
Boundary Decay and Domain Selection:
The authors establish a quantitative decay estimate for $v - u$ near the boundary of the domain, showing that $(v-u)(x) \le C d_x(1 + |\log d_x|)$ , where $d_x$ is the distance to the boundary. This decay, combined with the uniform interior separation, allows the selection of a regular value $\delta_1$ and a subdomain $\Omega$ such that $w := v - \delta_1$ satisfies $w > u$ in $\Omega$ , $w = u$ on $\partial \Omega$ , and $L_u w > 0$ .
Application of Pogorelov Estimate:
With the barrier function $w$ constructed, the authors apply the Chou–Wang Pogorelov estimate:
$\sup_{\Omega} (w-u)^4 |D^2u| \le C(n, \|Du\|_{L^\infty}, \|Dw\|_{L^\infty})$
Since $w-u$ is uniformly bounded away from zero in the interior, this yields the desired interior Hessian bound.

Key Results
The paper establishes two main theorems:

Theorem 1.1 (Smooth Solutions): For $n \ge 2$ , if $u \in C^\infty(B_2)$ satisfies $\sigma_2(D^2u)=1$ in $B_2$ , then the Hessian is bounded in the interior:
$\|D^2u\|_{L^\infty(B_{1/2})} \le C(n, \|u\|_{L^\infty(B_2)}).$
This result extends to smooth solutions of $\sigma_2(D^2u) = f(x, u, Du) > 0$ with $f \in C^{1,1}$ .
Theorem 1.2 (Viscosity Solutions): For $n \ge 2$ , if $u \in C(B_2)$ is a viscosity solution to (1.1) with $\Delta u > 0$ in the viscosity sense, then $u \in C^\infty(B_2)$ and satisfies the same interior Hessian estimate.

Significance and Claims
The authors claim that their results remove the dimension restriction and the need for convexity or semiconvexity assumptions that characterized previous literature. The proof is self-contained regarding the construction of the auxiliary function, which handles the interaction between positive and negative eigenvalues without requiring the solution to be convex.

The paper further notes that these estimates imply:

Regularity: Viscosity solutions on the positive branch are automatically smooth ( $C^\infty$ ).
Liouville-type Theorem: Every entire solution to (1.1) with quadratic growth must be a quadratic polynomial.
Hessian Quotient Equations: The results directly apply to the Hessian quotient equation $\sigma_2(D^2u)/\sigma_1(D^2u) = 1$ under the condition $\Delta u > 0$ .

The methodology is presented as a distinct alternative to the doubling arguments used in recent works, offering a unified approach for arbitrary dimensions based on the propagation of separation via specific barrier constructions.

More like this