Curvature-dimension condition, rigidity theorems and entropy differential inequalities on Riemannian manifolds

This paper employs an information-theoretic approach to establish the equivalence between the ${\rm CD}(K, m)$ curvature-dimension condition and entropy differential inequalities on Riemannian manifolds, while deriving new rigidity theorems for $K$-Einstein and $(K, m)$-Einstein manifolds through the monotonicity of $W$-entropy along Wasserstein geodesics.

The Gist

Imagine you are a landscape architect trying to understand the shape of a vast, invisible terrain. This terrain isn't made of dirt and rocks, but of probability—the likelihood of finding something in a certain place. In mathematics, this is called a Wasserstein space. Think of it as a giant map where every point represents a different "cloud" of probability (like a fog spreading out over a city).

The paper by Xiang-Dong Li is like a new set of tools for measuring the curvature (how curved or flat the terrain is) and the dimension (how many directions you can move) of this invisible landscape.

Here is the breakdown of the paper's big ideas using simple analogies:

1. The Two Ways to Measure a Mountain

Traditionally, mathematicians (like Lott, Sturm, and Villani) have tried to measure the curvature of this probability landscape by looking at how "clouds" of probability move along the shortest paths (geodesics) between them. They use a very complex, heavy formula involving integrals and trigonometric functions. It's like trying to measure the steepness of a mountain by calculating the exact path of every single raindrop falling on it. It works, but it's messy and hard to use.

The Paper's Innovation:
Li proposes a much simpler way. Instead of tracking every raindrop, he looks at Entropy.

• Entropy is a measure of "disorder" or "spread." Imagine a drop of ink in water. At first, it's a tight, ordered drop. As it spreads, it becomes disordered (high entropy).
• Li shows that if you watch how fast this "ink" spreads (changes entropy) as it moves along the shortest path on the map, you can tell exactly how curved the terrain is.

The Analogy:
Instead of measuring the mountain by walking every possible path, you just drop a ball and watch how fast it rolls. If it rolls fast, the ground is flat. If it slows down or curves, the ground is shaped differently. Li found that the "speed of spreading" (entropy) tells you the same thing as the complex mountain measurements, but in a much simpler way.

2. The "Rigid" Shapes (Rigidity Theorems)

The paper also asks: "What does the landscape look like if the rules are perfectly followed?"

In geometry, there are special shapes called Einstein manifolds. Think of these as the "perfect spheres" or "perfect cylinders" of the mathematical world. They are so symmetrical that their curvature is the same everywhere.

• The Discovery: Li proves that if the "spreading of ink" (entropy) follows a specific, perfect mathematical rhythm (an equality rather than an inequality), then the landscape must be one of these perfect Einstein shapes.
• The Metaphor: Imagine you are listening to a drum. If the drum vibrates with a slightly messy, irregular sound, it could be any shape. But if it vibrates with a perfect, pure tone (a specific mathematical equality), you know immediately that the drum is a perfect sphere. Li found the "perfect tone" for these probability landscapes.

3. The "W-Entropy" (The Magic Compass)

The paper introduces a concept called W-entropy, which is a special kind of compass invented by the famous mathematician Grisha Perelman (who solved the Poincaré Conjecture).

• What it does: It measures how the "disorder" of the system changes over time as it moves along the shortest path.
• The Result: Li proves that on these special landscapes, this compass always points in one direction (it is "monotonic"). It never wiggles back and forth; it only goes one way.
• Why it matters: This gives mathematicians a powerful new way to prove that a shape is "rigid" (unbreakable/perfect). If the compass stops moving or behaves strangely, the shape isn't perfect.

4. Why This Matters (The Big Picture)

Why do we care about the "curvature of probability"?

1. Simplifying the Complex: The old way of defining these shapes was like trying to describe a symphony by listing the frequency of every single sound wave. Li's method is like describing the melody. It's much easier to understand and use.
2. Connecting Fields: This paper bridges Information Theory (how we measure data and uncertainty) with Geometry (the study of shapes). It shows that the way information spreads is deeply connected to the shape of the universe.
3. New Definitions: It suggests a new, simpler way to define what it means for a space to have a certain "curvature" and "dimension," which could help mathematicians study spaces that aren't smooth (like fractals or digital grids) in the future.

Summary in One Sentence

Xiang-Dong Li discovered that by watching how "disorder" (entropy) spreads along the shortest paths in a probability landscape, we can easily measure the shape of that landscape and identify its most perfect, symmetrical forms, replacing complex, heavy formulas with elegant, simple rules.

Technical Summary

Here is a detailed technical summary of the paper "Curvature-dimension condition, rigidity theorems and entropy differential inequalities on Riemannian manifolds" by Xiang-Dong Li.

1. Problem Statement

The paper addresses the characterization of the Curvature-Dimension condition ($CD(K, m)$) on Riemannian manifolds (specifically weighted manifolds with the Bakry-Emery Ricci curvature). While the synthetic geometry framework developed by Lott, Sturm, and Villani characterizes $CD(K, N)$ via the displacement convexity of entropy functionals along geodesics in the Wasserstein space, the defining inequalities (Sturm's inequality) are often complex and involve intricate distortion coefficients.

The author aims to:

1. Establish simpler, equivalent characterizations of the $CD(K, m)$ condition using Shannon and Rényi entropy differential inequalities along geodesics on the Wasserstein space.
2. Investigate rigidity theorems: Determine the geometric structures (rigidity models) that satisfy the equality cases of these enhanced entropy inequalities.
3. Extend the concept of Perelman's $W$-entropy to geodesic flows on the Wasserstein space and prove monotonicity and rigidity results for these entropies.

2. Methodology

The paper employs an information-theoretic approach combined with optimal transport theory and geometric analysis.

• Geometric Setting: The study is set on a complete Riemannian manifold $(M, g)$ with a weighted volume measure $d\mu = e^{-V}dv$. The relevant curvature is the $m$-dimensional Bakry-Emery Ricci curvature:
  $$Ric_{m,n}(L) = Ric + \nabla^2 V - \frac{\nabla V \otimes \nabla V}{m-n}$$
  where $L = \Delta - \nabla V \cdot \nabla$ is the Witten Laplacian.
• Wasserstein Space: The analysis takes place on the $L^2$-Wasserstein space $P_2(M, \mu)$ equipped with Otto's infinite-dimensional Riemannian metric. Geodesics $\gamma(t) = \rho(t)\mu$ are described by the Benamou-Brenier formulation, satisfying the continuity equation and the Hamilton-Jacobi equation:
  $$\partial_t \rho - \nabla^*_\mu(\rho \nabla \phi) = 0, \quad \partial_t \phi + \frac{1}{2}|\nabla \phi|^2 = 0$$
• Entropy Functionals: The author defines Shannon entropy ($H$), Rényi entropy ($H_p$), and their corresponding entropy powers ($N_m, N_{m,p}$).
• Differential Calculus: The core technique involves computing the first and second time derivatives of these entropy functionals along the geodesic flow. The author utilizes the generalized Bochner formula and the $\Gamma_2$-calculus to derive exact formulas for the second derivatives, separating terms involving the curvature, the Hessian of the potential $\phi$, and the Fisher information.
• Rigidity Analysis: By analyzing the equality cases in the derived differential inequalities, the author identifies the necessary conditions on the manifold geometry (Einstein metrics) and the potential function (Hessian solitons).

3. Key Contributions and Results

A. Equivalent Characterizations of $CD(K, m)$

The paper proves that the condition $Ric_{m,n}(L) \geq Kg$ is equivalent to a family of differential inequalities along geodesics in the Wasserstein space. These provide simpler alternatives to Sturm's original definition.

• Shannon Entropy: The $CD(K, m)$ condition holds if and only if the Shannon entropy $H(\rho(t))$ satisfies:
  $$-H'' \geq \frac{1}{m}(H')^2 + K W_2^2(\rho(0), \rho(1))$$
  Equivalently, the Shannon entropy power $N_m$ satisfies a concavity inequality involving $K$.
• Rényi Entropy: Similar equivalences are established for the Rényi entropy $H_p$ and entropy power $N_{m,p}$ for $p \geq 1 - 1/m$.
• Sturm's Inequality: The author demonstrates that these differential inequalities are equivalent to Sturm's integral inequality (1.3) for the $CD(K, N)$ condition, thereby bridging the gap between synthetic geometry definitions and differential calculus on smooth manifolds.

B. Enhanced Entropy Inequalities and Rigidity Theorems

The paper derives "enhanced" differential inequalities that include non-negative remainder terms (variance of the Laplacian, Hessian norms, etc.).

• Enhanced Inequality:
  $$H''_p + \frac{1}{m}(H'_p)^2 + \text{Non-negative Terms} \leq -K \int |\nabla \phi|^2 d\gamma$$
• Rigidity Models:
  • Equality Case: The equality holds for all geodesics if and only if the manifold is a $(K, m)$-Einstein manifold (i.e., $Ric_{m,n}(L) = Kg$).
  • Soliton Condition: If the manifold is $(K, m)$-Einstein, the equality holds for a specific geodesic if and only if the potential $\phi$ satisfies the Hessian soliton equation:
    $$\nabla^2 \phi = \frac{I}{m}g, \quad L\phi + \frac{m}{m-n}\nabla V \cdot \nabla \phi = 0$$
    where $I$ is the Fisher information.
  • Special Case ($m=n, V=0$): The rigidity models reduce to standard Einstein manifolds ($Ric = Kg$) and Ricci solitons.

C. $W$-Entropy and Monotonicity

The author introduces a $W$-entropy functional associated with the geodesic flow on the Wasserstein space, analogous to Perelman's $W$-entropy for the Ricci flow.

• NIW Formula: A new formula (NIW) is derived linking the second derivative of the entropy power, the Fisher information, and the $W$-entropy.
• Monotonicity: Under the $CD(0, m)$ condition, the $W$-entropy is proven to be non-increasing along the geodesic flow.
• Rigidity of $W$-Entropy: If the $W$-entropy is constant (derivative is zero) at some time $t > 0$, the manifold must be isometric to Euclidean space $\mathbb{R}^n$, $m=n$, $V$ is constant, and the flow is the standard Gaussian heat kernel flow.

4. Significance

1. Simplification of Synthetic Geometry: The paper provides a more accessible, differential-calculus-based characterization of the $CD(K, m)$ condition on smooth manifolds, avoiding the complex distortion coefficients used in the general metric measure space setting.
2. New Rigidity Results: It establishes a direct link between the equality cases of entropy inequalities and specific geometric structures (Einstein and quasi-Einstein manifolds) and soliton solutions, offering a new perspective on rigidity in optimal transport.
3. Information-Theoretic Insight: The work deepens the understanding of Perelman's $W$-entropy and the Ricci flow by interpreting them through the lens of information theory and optimal transport on the Wasserstein space.
4. Foundation for Future Work: The results lay the groundwork for extending these concepts to non-smooth metric measure spaces (RCD spaces) and time-dependent geometries, as suggested in the open problems section.

In summary, this paper unifies concepts from information theory, optimal transport, and Riemannian geometry to provide a robust framework for understanding curvature bounds and rigidity phenomena through the evolution of entropy along geodesics.