🔢 mathematics

Measure preserving maps with bounded total variation

This paper proves that piecewise affine Lipschitz maps with locally bounded variation gradients, which satisfy a specific injectivity condition on their gradient flows, must be locally convex, thereby confirming a conjecture by Liu and Pego for this broader class of measure-preserving maps.

Original authors: Stefano Bianchini, Luca Talamini

Published 2026-03-20

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Stefano Bianchini, Luca Talamini

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

The Big Picture: The "Perfectly Elastic Sheet"

Imagine you have a giant, stretchy rubber sheet (this is your mathematical space, $\Omega$ ). On this sheet, you have drawn a grid. Now, imagine you push or pull different parts of the sheet.

The paper asks a very specific question: If you push the sheet in a way that never squashes or stretches the total amount of "stuff" (mass) in any region, does the sheet have to be shaped like a bowl (convex) rather than a saddle or a crumpled mess?

The authors, Bianchini and Talamini, prove that yes, under certain conditions, the sheet must be shaped like a bowl.

The Cast of Characters

The Map ( $\phi$ ): Think of this as the "instruction manual" for how to push the sheet. It tells every point on the sheet where to go.
The Flow ( $x + t\nabla\phi(x)$ ): Imagine you start pushing the sheet. At time $t=0$ $t = 0$ , everything is in its original spot. As time $t$ $t$ passes, every point moves in a straight line based on the instructions.
- The Rule: The paper assumes that as you push, the points never crash into each other. If you look at the sheet at any future time, the points are still distinct and spread out.
The "Measure Preserving" Condition: This is the most important rule. It means no compression and no expansion.
- Analogy: Imagine the sheet is made of a fluid that is perfectly incompressible (like water). If you push a square inch of water, it moves, but it doesn't shrink into a tiny dot or spread out into a giant puddle. The "volume" of every little patch remains exactly the same.

The Conjecture (The Guess)

In a previous paper (by Liu and Pego), mathematicians looked at a specific type of map (piecewise affine, meaning the sheet is made of flat triangular pieces) and guessed:

"If you push this sheet without squashing it, and the points never collide, the underlying shape must be locally convex."

What does "Locally Convex" mean?
Imagine a bowl. If you put a marble on it, it rolls down. That's convex.
Imagine a saddle (like a Pringles chip). If you put a marble in the middle, it might roll up or down depending on the direction. That is not convex.
The conjecture says: You cannot have a saddle shape if you are pushing it without squashing it. The sheet must always look like a bowl.

The Problem: The "Rough" Sheet

The previous guess was hard to prove for "rough" sheets.

Smooth sheets: Easy to handle.
Rough sheets: Imagine a sheet that is crinkled, has sharp folds, or jagged edges. In math terms, this is when the gradient (the slope of the sheet) is in a class called BV (Bounded Variation). It means the sheet can have jumps or kinks, but not infinite chaos.

The authors say: "We can prove the conjecture is true, but only if the sheet isn't too rough. Specifically, if the 'roughness' is controlled (Bounded Total Variation)."

How They Proved It (The Two-Step Dance)

The proof is like a two-step dance to show the sheet is a bowl.

Step 1: The "Sub-harmonic" Check (The Slope Test)

First, they proved that the sheet must be "sub-harmonic."

The Metaphor: Imagine the sheet is a landscape. A "sub-harmonic" landscape is one where, on average, the ground is always sloping upward or staying flat, never dipping down into a valley.
They showed that if you push the sheet without squashing it, the landscape cannot have deep valleys. It must be "bowl-like" in a general sense.
Note: This step works even for very rough sheets.

Step 2: The "Roughness" Filter (The BV Step)

This is where they use the special condition (Bounded Variation).

The Metaphor: Imagine the sheet is made of two layers:
1. The Smooth Layer: The gentle, flowing curves.
2. The Kink Layer: The sharp folds and jumps.
The authors analyzed the "Smooth Layer" first. They used a clever trick involving the "Area Formula" (a way to count how much space a shape covers). They proved that for the sheet to preserve volume, the smooth part cannot have any curvature at all. It must be perfectly flat.
Then, they looked at the "Kink Layer." Because the sheet preserves volume, the kinks can only bend in one direction (upward). They cannot bend downward.
The Result: Since the smooth part is flat and the kinked part only bends up, the entire sheet is a bowl.

The "Optimal Transport" Alternative (The Delivery Driver)

In the last section, the authors offer a different way to prove the same thing, using Optimal Transport.

The Metaphor: Imagine you are a delivery driver. You have packages (mass) scattered on the sheet. You want to move them to new locations using the least amount of energy (shortest distance).
There is a famous theorem (Brenier's Theorem) that says the most efficient way to move these packages is always to push them along a "convex" path (like rolling a ball down a bowl).
The authors showed that their "measure preserving" map is so close to this "most efficient" delivery path that it must share the same bowl-shaped property.

The Conclusion

In simple terms:
If you have a shape that you can deform (push and pull) without ever changing the size of any piece of it, and the shape isn't infinitely chaotic, then that shape must be convex (bowl-shaped).

You cannot have a "saddle" or a "dip" in the middle if you are moving the material without squashing it. The math proves that nature (or at least, this specific mathematical universe) forbids "saddles" in volume-preserving flows.

Why does this matter?
This helps mathematicians understand how fluids, gases, and even the universe itself might evolve over time. If we know that certain flows must be convex, we can predict how they will behave, which is crucial for physics, engineering, and economics.

1. Problem Statement and Context

The paper addresses a conjecture proposed by Liu and Pego (2025) regarding the geometric structure of specific measure-preserving maps.

The Setup: Let $\Omega \subset \mathbb{R}^d$ be an open set. Consider a piecewise affine Lipschitz map $\phi: \Omega \to \mathbb{R}$ .
The Condition: The map $T_t(x) := x + t\nabla\phi(x)$ is assumed to be injective for almost every $t > 0$ . In the context of optimal transport, this relates to the flow of particles with initial positions $x$ and velocities $\nabla\phi(x)$ .
The Conjecture (Liu-Pego): If such a map exists, $\phi$ must be locally convex.
The Gap: While the conjecture was proven for the case where $\nabla\phi$ is continuous ( $C(\Omega)$ ), the general case where $\nabla\phi$ lacks regularity remained open.
The Goal: The authors aim to prove the conjecture under the assumption that the gradient $\nabla\phi$ belongs to the space of functions of Bounded Variation ( $BV_{loc}$ ).

2. Methodology and Framework

The authors generalize the problem to a broader class of measure-preserving maps and employ a two-step proof strategy combining geometric measure theory (BV functions) and optimal transport theory.

A. Generalization to Measure Preserving Maps

The authors define a map $\phi \in W^{1,1}(\Omega)$ as satisfying the Measure Preserving Condition (MPC) if there exists a full measure set $F \subset \Omega$ such that the transport map $T_t(x) = x + t\nabla\phi(x)$ preserves the Lebesgue measure on $F$ for almost every $t \geq 0$ .

B. Step 1: Subharmonicity (Without BV Assumption)

The first major technical hurdle is establishing that $\phi$ is subharmonic ( $\Delta \phi \geq 0$ in the distributional sense).

Approach: The authors introduce the concept of expanding maps (maps that do not increase measure). They prove that if $T_t$ is expanding for a.e. $t$ , then $\phi$ is subharmonic.
Technique: They utilize a variational argument involving the time derivative of the pushforward measure $\rho_t = (T_t)_\sharp \mathcal{L}^d$ . By analyzing the function $g_\psi(t) = \int \psi d\rho_t$ for non-negative test functions $\psi$ , they show that the derivative at $t=0$ implies $\int \nabla\phi \cdot \nabla\psi \leq 0$ , which is the definition of subharmonicity.
Significance: This step holds even if $\nabla\phi \notin BV$ , showing that the geometric constraint of measure preservation inherently forces the potential to be subharmonic.

C. Step 2: Convexity via BV Theory

The second step uses the $BV_{loc}$ assumption to upgrade subharmonicity to local convexity.

Decomposition: Since $\nabla\phi \in BV_{loc}$ , the distributional derivative $D(\nabla\phi)$ (the Hessian of $\phi$ ) can be decomposed into an absolutely continuous part ( $D_a$ ) and a singular part ( $D_s$ ) with respect to the Lebesgue measure.
Analysis of the Singular Part: Using Alberti's Rank-1 Theorem, the singular part of the derivative of a BV function is shown to be of the form $\lambda \vec{n} \otimes \vec{n}$ . The subharmonicity result from Step 1 forces the scalar $\lambda \geq 0$ , implying the singular part of the Hessian is a positive semi-definite measure.
Analysis of the Absolutely Continuous Part: The authors use Lusin-Lipschitz properties of BV functions. They approximate $\nabla\phi$ $\nabla ϕ$ by Lipschitz functions $f_\Lambda$ $f_{Λ}$ on sets $G_\Lambda$ $G_{Λ}$ where the approximation is good.
- They apply the Area Formula to the transport map $T_t$ restricted to these sets.
- The measure-preserving condition implies that the determinant of the Jacobian of the transport map must be 1 almost everywhere: $|\det(I + t D_a(\nabla\phi))| = 1$ .
- By analyzing the eigenvalues $\lambda_i$ of the matrix $D_a(\nabla\phi)$ , the equation $\prod (1 + t\lambda_i) = 1$ for all $t$ forces all $\lambda_i = 0$ .
- Result: The absolutely continuous part of the Hessian vanishes ( $D_a(\nabla\phi) = 0$ ).
Conclusion: The total Hessian $D(\nabla\phi) = D_s(\nabla\phi)$ is a positive semi-definite matrix-valued measure. A distribution with a positive semi-definite Hessian is locally convex.

D. Alternative Proof (Optimal Transport)

In Section 4, the authors provide an alternative proof of subharmonicity using Brenier's Polar Factorization Theorem.

They decompose the transport map $T_t$ into a gradient of a convex function $\nabla f_t$ and a measure-preserving map $\Phi_t$ .
By analyzing the limit as $t \to 0$ , they relate the convex potential $f_t$ to the original $\phi$ .
They show that the convexity of $f_t$ (which implies $\Delta f_t \geq 0$ ) transfers to $\phi$ in the limit, reinforcing the subharmonicity result without relying on the "expanding map" variational argument.

3. Key Contributions

Proof of the Liu-Pego Conjecture (BV Case): The paper confirms that if $\nabla\phi \in BV_{loc}$ and the flow $x + t\nabla\phi(x)$ preserves measure, then $\phi$ is locally convex.
Decoupling of Regularity: The authors demonstrate that subharmonicity is a consequence of the measure-preserving property alone (valid for $W^{1,1}$ ), while convexity requires the additional $BV$ regularity to control the singularities of the Hessian.
Rigorous BV Analysis: The paper provides a detailed analysis of the absolutely continuous part of the derivative of BV functions in the context of transport maps, proving that the measure-preserving constraint forces the absolutely continuous Hessian to vanish.
Dual Approaches: The paper offers two distinct proofs for the subharmonicity of $\phi$ : one based on the geometry of expanding maps (Section 3) and one based on optimal transport and polar factorization (Section 4).

4. Results

Theorem 1.2 / Theorem 2.3: Let $\phi \in W^{1,1}(\Omega)$ satisfy the measure-preserving condition (MPC). If $\nabla\phi \in BV_{loc}(\Omega)$ , then $\phi$ is locally convex.
Proposition 2.4: Under the MPC (without BV assumption), $\phi$ is subharmonic ( $\Delta \phi \geq 0$ ).
Corollary 4.3: In the context of polar decomposition, the gradient of the convex potential in the decomposition converges to $\nabla\phi$ in $L^2$ as $t \to 0$ .

5. Significance

Resolution of an Open Problem: This work resolves a specific case of a conjecture regarding "rigidly breaking potential flows," a topic relevant to fluid dynamics and the structure of solutions to the Euler equations.
Bridge Between Fields: The paper effectively bridges Geometric Measure Theory (BV functions, Alberti's theorem) and Optimal Transport (Brenier's theorem, Monge-Ampère equations).
Limitations and Future Work: The authors explicitly note that the $BV$ assumption is currently essential. The case where $\nabla\phi$ is not in $BV$ (but perhaps only in $L^1$ or continuous) remains an open problem, as the subharmonicity alone does not guarantee convexity without controlling the singular part of the Hessian.
Methodological Insight: The proof highlights that measure-preserving flows impose extremely strong constraints on the potential function, effectively forcing the "linear" part of the deformation to vanish and leaving only convex singularities.

In summary, Bianchini and Talamini prove that for a broad class of regular potentials ($BV$ gradients), the only way a flow generated by a gradient can preserve Lebesgue measure is if the potential is convex. This deepens the understanding of the rigidity of measure-preserving transformations in optimal transport.