Mosco-convergence of Cheeger energies on varying spaces satisfying curvature dimension conditions

This paper establishes the Mosco-convergence of Cheeger energies on Gromov-Hausdorff converging spaces satisfying curvature-dimension conditions by employing a Lagrangian approach that combines Wasserstein geodesic stability with nonsmooth calculus duality, thereby ensuring the continuity of Neumann eigenvalues even in infinite-dimensional settings.

The Gist

Here is an explanation of the paper "Mosco-convergence of Cheeger Energies on Varying Spaces Satisfying Curvature Dimension Conditions," translated into everyday language with creative analogies.

The Big Picture: The "Shape-Shifting" Universe

Imagine you are an architect designing a building. But instead of building on a single, solid piece of land, you are building on a series of landscapes that are slowly morphing into one another.

• The Landscapes ($X_n$): These are your starting spaces. They could be smooth hills, jagged mountains, or even digital grids.
• The Transformation: Over time, these landscapes are changing. A jagged mountain range might smooth out into a gentle hill, or a grid of pixels might blur into a continuous curve.
• The Goal: You want to know if the rules of physics (specifically, how heat flows, how things vibrate, or how much "effort" it takes to move across the land) stay consistent as the landscape changes.

This paper asks: If we have a function (like a temperature map or a vibration pattern) on a changing landscape, does the "energy" required to maintain that function behave predictably as the landscape transforms?

The Core Concepts, Simplified

1. The "Cheeger Energy" (The Effort Meter)

In math, we often talk about the "energy" of a function. Think of this as the friction or tension in a rubber sheet stretched over your landscape.

• If the sheet is flat, it has low energy.
• If the sheet is crumpled or has steep cliffs, it has high energy.
• The Cheeger Energy is a way to measure this "roughness" or "tension" even on weird, non-smooth shapes (like a crumpled piece of paper or a digital grid).

2. The "Curvature Dimension" (The Shape Rules)

The paper focuses on spaces that follow specific "shape rules" called Curvature-Dimension (CD) conditions.

• The Analogy: Imagine you are rolling a ball.
  • On a flat surface, it rolls straight.
  • On a sphere (positive curvature), the paths converge (like lines of longitude meeting at the pole).
  • On a saddle (negative curvature), the paths diverge.
• The authors are studying spaces that obey a specific "safety code" regarding how they curve and how many dimensions they have. They want to know: If we change the landscape but keep these safety codes, do the rules of friction (energy) stay the same?

3. The "Mosco-Convergence" (The Perfect Handoff)

This is the technical heart of the paper. It's a fancy way of saying: "Does the energy match up perfectly when we switch landscapes?"

Imagine you are passing a hot potato (the energy) from one person to another in a relay race.

• Bad Handoff: The potato drops, or the next runner has to run faster than necessary to catch it. The energy is lost or gained unpredictably.
• Mosco-Convergence: This is the perfect handoff.
  1. Lower Semicontinuity: You can't suddenly lose energy. If the first runner had a lot of tension, the second runner must have at least that much. (You can't magically smooth out a crumpled sheet without doing work).
  2. Recovery Sequence: You can always find a way to match the energy. If the second runner needs a specific amount of tension, you can find a way to set up the first runner to provide exactly that.

The paper proves that for these specific "shape-coded" landscapes, this perfect handoff always works.

The Two Main Discoveries

The authors proved two main things, depending on the type of "shape code" the landscape follows:

1. The "Gold Standard" Result (CD Spaces)
If the landscapes follow the strict CD(K, N) rules (a very robust set of curvature and dimension constraints), the energy handoff is perfect.

• The Result: The energy of the final shape is exactly the limit of the energies of the previous shapes. No loss, no gain.
• Why it matters: This allows mathematicians to solve problems on simple shapes (like grids) and be 100% sure the answer applies to complex, smooth shapes (like spheres) that those grids approximate.

2. The "Good Enough" Result (MCP Spaces)
If the landscapes follow a slightly looser rule called MCP (Measure Contraction Property), the handoff is almost perfect, but with a small "tax."

• The Result: The energy of the final shape is bounded by the limit of the previous energies, but multiplied by a factor (like $2^N$).
• The Analogy: It's like passing the potato, but the second runner has to run a little bit faster (pay a small tax) to catch it. It's not a perfect match, but it's still controlled and predictable.

The Secret Weapon: "Polygonal Interpolation"

How did they prove this? They used a clever trick called Polygonal Interpolation.

• The Problem: You can't just draw a straight line between two points on a weird, changing landscape. The path might get stuck or break.
• The Solution: Instead of one long, smooth path, they broke the journey into tiny, straight segments (like a polygon).
• The Magic: They showed that even as the landscape morphs, these tiny straight segments can be adjusted to "snap" into place, preserving the energy calculations. It's like using a flexible ruler made of tiny, rigid links that can bend to fit any shape while keeping the measurements accurate.

Why Should You Care? (The Real-World Impact)

The paper ends with a practical application: Neumann Eigenvalues.

• The Analogy: Imagine a drum. The shape of the drum determines the sound it makes (its "eigenvalues").
• The Application: If you have a drum made of a weird material that is slowly changing shape (melting, or being 3D printed), will the pitch of the drum change smoothly?
• The Conclusion: Yes! Because the "energy" rules (Cheeger energies) are stable, the sound of the drum will change continuously. You won't hear a sudden, jarring jump in pitch just because the material shifted slightly.

Summary in One Sentence

This paper proves that if you have a series of shapes that follow specific geometric "safety rules," the mathematical "friction" (energy) of functions on those shapes behaves perfectly and predictably as the shapes morph into one another, ensuring that physical properties like vibration frequencies change smoothly rather than chaotically.

Technical Summary

Here is a detailed technical summary of the paper "Mosco-convergence of Cheeger energies on varying spaces satisfying curvature dimension conditions" by Francesco Nobili, Federico Renzi, and Federico Vitillaro.

1. Problem Statement and Context

The paper addresses the stability of first-order calculus (Sobolev and BV spaces) and associated energy functionals (Cheeger energies) under the convergence of metric measure spaces.

• The Setting: The authors consider a sequence of pointed metric measure spaces $(X_n, d_n, m_n, x_n)$ converging to a limit space $(X_\infty, d_\infty, m_\infty, x_\infty)$ in the pointed measured Gromov-Hausdorff (pmGH) topology.
• The Curvature Condition: The spaces satisfy synthetic lower Ricci curvature bounds, specifically the Curvature-Dimension condition $CD(K, N)$ or the Measure Contraction Property $MCP(K, N)$, where $K \in \mathbb{R}$ and $N \in [1, \infty)$.
• The Core Question: If a sequence of functions $f_n$ converges weakly to $f_\infty$ in $L^p$, does the limit function belong to the corresponding Sobolev or BV space on the limit manifold, and does the energy satisfy the lower semicontinuity inequality?
  $$\text{Does } Ch_p(f_\infty) \leq \liminf_{n \to \infty} Ch_p(f_n) \quad \text{and} \quad |Df_\infty|(X_\infty) \leq \liminf_{n \to \infty} |Df_n|(X_n)?$$
  This property is known as Mosco-convergence of the Cheeger energies.
• The Challenge: While this stability is known for fixed spaces or specific classes (like RCD spaces), it is non-trivial for general $CD(K, N)$ spaces, especially when $K < 0$ or when the spaces are non-compact or infinite-dimensional. Standard counterexamples exist where zeroth-order convergence (metric measure) fails to preserve first-order structures without uniform second-order control (curvature).

2. Methodology

The authors employ a direct Lagrangian approach, avoiding the Eulerian perspective (e.g., heat flow or Fisher information) used in previous works like [36] and [10].

• Lagrangian Characterization: They utilize a characterization of Sobolev and BV functions via test plans (probability measures on the space of curves).
  • For $f \in W^{1,p}$, the Cheeger energy is equivalent to the existence of a constant $C$ such that for all $q$-test plans $\pi$:
    $$\int (f(\gamma_1) - f(\gamma_0)) \, d\pi \leq \text{Comp}(\pi)^{1/p} \text{Ke}_q(\pi)^{1/q} C$$
    where $\text{Comp}$ is the compression constant and $\text{Ke}_q$ is the kinetic energy.
  • A similar dual characterization exists for BV functions using $\infty$-test plans.
• Polygonal Interpolation on Varying Spaces: The core technical innovation is the construction of polygonal geodesic interpolations in the Wasserstein space of the varying sequence $X_n$.
  • Given a test plan $\eta$ on the limit space $X_\infty$, the authors construct a sequence of polygonal plans $\pi_n^M$ on $X_n$ (where $M$ is the number of segments).
  • Case $K \geq 0$: A single geodesic interpolation suffices.
  • Case $K < 0$: A single geodesic interpolation fails to preserve necessary density estimates due to mass spreading. The authors must construct polygonal geodesics (breaking the path into $M$ segments) and carefully discard mass along the geodesics to maintain control over the compression constants. They prove that as $M \to \infty$ and $n \to \infty$, these polygonal plans recover the properties of the original test plan $\eta$.
• Independence of Transport Exponent: For the $CD(K, N)$ case, they leverage a recent result (Theorem 2.15 from [1]) stating that the $CD(K, N)$ condition is independent of the transport exponent $q$ in essentially non-branching spaces with finite measures. This allows treating all $p$-Cheeger energies simultaneously.

3. Key Contributions and Results

A. Main Theorems on Mosco-Convergence

The paper establishes the lower semicontinuity of Cheeger energies and Total Variation under pmGH convergence for spaces satisfying curvature bounds.

1. Theorem 1.1 (CD(K, N) Spaces):

   • Assumptions: $X_n$ are $q$-essentially non-branching, satisfy $CD(K, N)$, and have finite reference measures.
   • Result: The Cheeger energies $Ch_p$ ($p \in (1, \infty)$) and Total Variation $|Df|$ satisfy the sharp Mosco-convergence inequality:
     $$Ch_p(f_\infty) \leq \liminf_{n \to \infty} Ch_p(f_n)$$
     $$|Df_\infty|(X_\infty) \leq \liminf_{n \to \infty} |Df_n|(X_n)$$
   • Significance: This is the first result covering all $p \neq 2$ for general $CD(K, N)$ spaces (previously known only for $p=2$ in [36] or for RCD spaces in [10]).

2. Theorem 1.2 (MCP(K, N) Spaces):

   • Assumptions: $X_n$ are $q$-essentially non-branching and satisfy the weaker $MCP(K, N)$ condition. No finiteness of measures is required.
   • Result: The stability holds but with a multiplicative constant $2^N$:
     $$Ch_p(f_\infty) \leq 2^N \liminf_{n \to \infty} Ch_p(f_n)$$
   • Significance: This extends stability to the broader MCP class, bypassing the need for the independence of $q$ in transport. The loss of sharpness ($2^N$) stems from weaker interpolation estimates available under MCP.

B. Applications to Spectral Geometry

• Theorem 1.4 (Continuity of Neumann Eigenvalues):
  • The authors prove that the first non-trivial Neumann eigenvalue $\lambda_p(X)$ of the $p$-Laplacian is continuous under measured Gromov-Hausdorff convergence for spaces satisfying $CD(K, N)$ with uniform diameter bounds.
  • $$\lambda_p(X_\infty) = \lim_{n \to \infty} \lambda_p(X_n)$$
  • This generalizes previous results known only for $p=2$ or restricted RCD classes.

C. Technical Innovations

• Polygonal Approximations: The construction of polygonal geodesics in varying spaces (Propositions 4.5 and 4.8) is a novel tool that allows handling negative curvature ($K < 0$) where standard geodesic interpolation fails to preserve density bounds.
• Direct Lagrangian Proof: By avoiding the heat flow/Fisher information link (which requires RCD structure), the authors extend results to non-Riemannian, non-branching spaces that do not necessarily satisfy the RCD condition.

4. Significance and Impact

• Robustness of Synthetic Geometry: The results reinforce the stability of first-order calculus in the synthetic theory of Ricci curvature, showing that the $CD(K, N)$ condition is strong enough to control the convergence of Sobolev structures even in infinite-dimensional or non-smooth settings.
• Generalization: The work significantly broadens the scope of known stability results from the restrictive RCD class to the general $CD(K, N)$ and $MCP(K, N)$ classes, including cases with negative curvature bounds.
• Methodological Shift: The reliance on a purely Lagrangian approach (test plans and Wasserstein geodesics) rather than Eulerian methods (heat flow) provides a more flexible toolkit for analyzing non-Riemannian spaces and local curvature constraints.
• Spectral Stability: The continuity of eigenvalues is a crucial application, ensuring that spectral properties of the Laplacian are preserved in the limit, which is fundamental for geometric analysis and PDEs on singular spaces.

In summary, this paper provides a comprehensive and rigorous framework for understanding how Sobolev and BV structures behave under geometric convergence in spaces with synthetic curvature bounds, utilizing advanced optimal transport techniques to overcome the difficulties posed by negative curvature and varying geometries.