🔢 mathematics

Convexity inequalities for eigenvalues and log-concavity of eigenfunctions

This paper presents simple new proofs for the Brunn–Minkowski inequality regarding Dirichlet eigenvalues and the log-concavity of the first Dirichlet eigenfunction for the Schrödinger operator, extending the former to a broader class of domains and potentials.

Original authors: Paul Bryan, Julie Clutterbuck, Cale Rankin

Published 2026-05-05

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Paul Bryan, Julie Clutterbuck, Cale Rankin

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ 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

Imagine you have a musical instrument, like a drum or a bell. When you strike it, it vibrates at a specific pitch. In mathematics, this "pitch" is called an eigenvalue, and the shape of the drum is called a domain. The paper you're reading is about how these pitches change when you mix two different drum shapes together.

The authors, Paul Bryan, Julie Clutterbuck, and Cale Rankin, have found a new, simpler way to prove two famous mathematical rules about these shapes and their sounds. They use a clever trick called a "sup-convolution," which is a bit like blending two recipes to see what the new flavor tastes like.

Here is a breakdown of their work using simple analogies:

1. The Main Idea: Mixing Two Drums

Imagine you have two different drum shapes, let's call them Drum A and Drum B.

Drum A has a low, deep hum (a low eigenvalue).
Drum B has a high, sharp ring (a high eigenvalue).

Now, imagine you want to create a "Drum C" that is a perfect mix of the two. You don't just glue them together; you use a mathematical recipe called a Minkowski sum. Think of this as taking every point on Drum A and every point on Drum B and blending them together in a specific ratio (say, 50% A and 50% B) to create a new, smooth shape in the middle.

The big question the paper answers is: What is the pitch of this new mixed drum?

2. The First Discovery: The "Average Pitch" Rule

The first result (Theorem 1.2) is about the Brunn–Minkowski inequality.

The Old Way: Previously, proving this required very strict conditions. The drums had to be perfectly convex (like a smooth ball, not a starfish) and the math had to be incredibly complex.
The New Proof: The authors show that you can mix almost any two connected shapes (even if they are a bit lumpy, as long as they are smooth enough) and the pitch of the new mixed drum will never be higher than the average of the two original pitches.

The Analogy: Think of the pitch as the "height" of a hill. If you have two hills, one low and one high, and you blend their shapes to make a new hill, the new hill's peak will never be taller than the average height of the two original peaks. The authors proved this rule holds true even for shapes that aren't perfectly round, as long as the "potential" (the material the drum is made of) behaves nicely.

3. The Second Discovery: The "Smooth Hill" Shape

The second result (Theorem 1.3) is about the shape of the vibration itself, known as the eigenfunction.

When a drum vibrates, the air moves up and down. The "first eigenfunction" describes the main, simplest wave pattern.

Log-Concavity: This is a fancy math term that essentially means the wave looks like a smooth, single hill. It goes up, reaches a peak, and goes down. It doesn't have weird bumps, dips, or multiple peaks.
The Analogy: Imagine pouring water into a bowl. If the bowl is convex (curved outward), the water surface forms a smooth, single mound. The authors proved that if your drum shape is convex, the main vibration wave will always look like that smooth, single mound. It will never be bumpy or have multiple humps.

4. The Secret Weapon: "Sup-Convolution"

How did they prove this so simply? They used a tool called sup-convolution.

The Metaphor: Imagine you have two maps of terrain, Map A and Map B. You want to create a new map that represents the "best possible view" from a point that is a mix of the two locations.
The authors take the vibration patterns of Drum A and Drum B and "blend" them together using a specific formula. They create a new, hybrid vibration pattern.
They then show that this hybrid pattern is a valid "candidate" for the new mixed drum. By checking the energy of this hybrid pattern, they can mathematically prove the rules about the pitch and the shape without needing the heavy machinery used in older proofs.

5. Why This Matters (According to the Paper)

Simplicity: The authors didn't invent a new law of physics; they just found a much shorter, clearer path to prove laws that were already known. It's like finding a shortcut through a forest that everyone else was walking around.
Flexibility: Their method works for a wider variety of shapes than previous methods. You don't need the shapes to be perfect spheres; they just need to be connected and reasonably smooth.
The "Bonus" Result: The proof for the "Average Pitch" rule is so strong that the "Smooth Hill" rule (log-concavity) falls out as a free bonus. If you prove the pitch rule for a drum mixed with itself, you automatically prove that the vibration must be a smooth hill.

What They Didn't Do

The paper is purely mathematical. They did not:

Apply this to real-world engineering or medicine.
Suggest how to build better musical instruments.
Predict future uses in technology.

They simply took two existing mathematical theorems, stripped away the unnecessary complexity, and showed that the core logic is much simpler and more robust than anyone realized. They proved that when you mix shapes, the resulting "sound" and "shape" follow predictable, smooth rules.

Technical Summary: Convexity Inequalities for Eigenvalues and Log-Concavity of Eigenfunctions

Problem Statement
The paper addresses two fundamental results concerning the Schrödinger operator $-\Delta + V$ on bounded domains $\Omega \subseteq \mathbb{R}^n$ :

The Brunn–Minkowski inequality for Dirichlet eigenvalues: Establishing that the first Dirichlet eigenvalue $\lambda(\Omega_t)$ of a Minkowski sum domain $\Omega_t = (1-t)\Omega_0 + t\Omega_1$ satisfies the concavity inequality $\lambda_t \leq (1-t)\lambda_0 + t\lambda_1$ .
The log-concavity of the first eigenfunction: Proving that for a convex domain $\Omega$ and a convex potential $V$ , the first Dirichlet eigenfunction is log-concave.

While these results were originally established by Brascamp and Lieb [1], this paper aims to provide "simple new proofs" that relax certain regularity assumptions and clarify the logical relationship between the two theorems.

Methodology
The core methodology relies on the sup-convolution technique, an idea attributed to Colesanti [2]. The authors define a function $u: \Omega_t \to \mathbb{R}$ via the sup-convolution of the first eigenfunctions $u_0$ and $u_1$ of the domains $\Omega_0$ and $\Omega_1$ :
$u(z) = \sup \{ u_0(x_0)^{1-t} u_1(x_1)^t : z = (1-t)x_0 + tx_1, x_i \in \Omega_i \}.$

The proof strategy proceeds as follows:

Regularity Analysis: The authors rigorously establish the regularity properties of the sup-convolution $u$ . They prove that $u$ is locally semi-convex and globally Lipschitz on $\Omega_t$ , even when the boundary behavior of $u_0$ and $u_1$ (vanishing on $\partial \Omega_i$ ) complicates the analysis. This involves showing that $u$ possesses Alexandrov second derivatives almost everywhere and satisfies specific integration by parts inequalities (Lemmas 3.6–3.11).
Optimality Conditions: By analyzing the points $(x_0, x_1)$ where the supremum is attained, the authors derive a gradient matching condition: $\frac{\nabla u_0}{u_0}(x_0) = \frac{\nabla u_1}{u_1}(x_1)$ .
Differential Inequality: Using the semi-convexity of $u$ , the authors compute the Alexandrov second derivatives of $u$ at a point $z$ . They show that the Laplacian of the sup-convolution satisfies:
$\frac{\Delta u}{u} \geq (1-t)\frac{\Delta u_0}{u_0} + t\frac{\Delta u_1}{u_1}.$
Substituting the eigenvalue equations $-\Delta u_i + V u_i = \lambda_i u_i$ and utilizing the convexity of the potential $V$ , they derive a lower bound for $\frac{\Delta u}{u}$ involving a convex combination of $\lambda_0$ and $\lambda_1$ .
Rayleigh Quotient: The authors integrate this inequality using a justified integration by parts argument (valid due to the semi-convexity and Lipschitz continuity of $u$ ). By substituting the resulting bound into the Rayleigh quotient for $\lambda_t$ , they establish the desired inequality.

Key Contributions and Results

Relaxed Regularity Assumptions: Unlike previous proofs (specifically Colesanti's [2]) which required the domains to be convex and the eigenfunctions to be log-concave a priori, this proof applies to a broader class of domains. Specifically, it requires only that $\Omega_0$ and $\Omega_1$ are $C^{1,1}$ connected domains (or sufficiently regular to ensure $C^1$ eigenfunctions up to the boundary) and that the potential $V$ is convex.
Logical Hierarchy: The paper demonstrates that the log-concavity of the first eigenfunction (Theorem 1.3) is a direct corollary of the Brunn–Minkowski inequality for eigenvalues (Theorem 1.2). By setting $\Omega_0 = \Omega_1 = \Omega$ , the Minkowski sum $\Omega_t$ remains $\Omega$ , and the optimality of the sup-convolution implies the log-concavity of the eigenfunction itself.
Rigorous Regularity Proofs: The authors provide self-contained proofs for the regularity of the sup-convolution (semi-convexity, Lipschitz continuity, and Sobolev regularity), addressing technical gaps regarding boundary behavior and variational definitions that are often glossed over in the literature.

Significance and Claims
The authors claim that their primary contribution is the simplicity and generality of the new proofs.

The proof of the Brunn–Minkowski inequality for eigenvalues is shown to hold for connected domains with $C^{1,1}$ boundaries, rather than strictly convex domains.
The log-concavity of eigenfunctions is derived as a "simple corollary" rather than a prerequisite.
The authors explicitly note a limitation: their method does not capture the conditions for equality, which were established in Colesanti's work [2].
They also clarify a subtle point regarding the convexity of the function $h(t) = \lambda_t$ , noting via Remark 2.1 that the Brunn–Minkowski inequality $\lambda_t \leq (1-t)\lambda_0 + t\lambda_1$ does not imply that $h(t)$ is convex, as the Minkowski deformation of non-convex domains does not necessarily preserve the inclusion structure required for such convexity.

In summary, the paper offers a streamlined, self-contained derivation of these classical results, extending their applicability to less regular domains while clarifying the logical dependency between eigenvalue inequalities and eigenfunction properties.