🔢 mathematics

Quantitative stability for Bakry--Émery log-Sobolev and Talagrand inequalities

This paper establishes quantitative $L^1$ -stability estimates for Bakry--Émery log-Sobolev and Talagrand inequalities with a universal exponent of 1/19, which improves to the optimal 1/2 in the radial setting, by combining Maurey-type arguments with Prékopa--Leindler stability estimates to characterize equality cases and analyze hypercontractivity deficits.

Original authors: Alexandru Kristály, Alexandru Pîrvuceanu

Published 2026-08-11

📖 7 min read🧠 Deep dive

CC0 1.0

Original authors: Alexandru Kristály, Alexandru Pîrvuceanu

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

Imagine the universe as a giant, invisible landscape where everything from the flow of heat to the movement of particles follows a set of hidden rules. Mathematicians and physicists have spent centuries trying to map these rules, often using "inequalities" as their compass. Think of an inequality not as a strict law that says "you must do this," but more like a safety net or a speed limit sign. It tells us that no matter how you arrange things, certain outcomes can't get too far out of hand. For example, one famous rule says that if you have a cloud of gas, it can't get too spread out without using a certain amount of energy to keep it that way. These rules are crucial because they help us predict how complex systems behave, from how a drop of ink spreads in water to how information travels through a computer network.

But here's the tricky part: what happens when things are almost perfect? If a system is just a tiny bit off from the ideal, how far off is it really? This is the question of "stability." It's like asking, "If I build a tower of blocks that is almost perfectly straight, how much does it wobble?" For a long time, mathematicians knew the rules for the perfect, straight towers, but they were fuzzy on the wobbly ones. They wanted to know exactly how the "wobble" (the error) relates to the "straightness" (the ideal state). This paper dives into that fuzzy middle ground, trying to measure the wobble with extreme precision.

The Great Wobble Hunt

In this paper, the authors, Alexandru Kristály and Alexandru Pîrvuceanu, are on a mission to measure the "wobble" in some of the most important mathematical safety nets ever discovered. They are looking at two specific rules: the Log-Sobolev inequality and the Talagrand transport inequality.

To understand what they did, let's use a metaphor. Imagine you are trying to match a shape (let's call it a "blob") to a perfect template.

The Log-Sobolev inequality is like a rule about how much energy it takes to keep that blob from spreading out too much.
The Talagrand inequality is like a rule about how much "work" it takes to move that blob from one spot to another.

In the perfect world, if your blob matches the template exactly, the "deficit" (the amount of extra energy or work needed) is zero. But in the real world, blobs are rarely perfect. The big question is: If the deficit is tiny, how close is the blob to the perfect shape?

The authors found a way to answer this with a new, universal ruler. They proved that the "wobble" (the distance from the perfect shape) is always related to the "deficit" (the error) by a specific power. In the general case, they found that if you take the error and raise it to the power of 1/19, you get a good estimate of how far off the shape is.

Why 1/19? Think of it like a very sensitive scale. If you have a tiny amount of weight (the error), the scale tells you that the object is slightly off-center. The number 1/19 is the "sensitivity setting" of their new scale. It's not necessarily the most sensitive setting possible (the authors admit they suspect a setting of 1/2 might be better in some cases), but it is a setting that works for almost every situation they tested.

The Special Case: When Things Are Radial

The paper gets even more exciting when they look at a special type of blob: one that is radial. Imagine a perfectly round ball, like a marble, instead of a squashed potato. When the shape is perfectly round (radial), the authors discovered that their ruler becomes much more sensitive.

In this round-world scenario, the exponent jumps from 1/19 to 1/2. This is a huge deal because 1/2 is considered the "gold standard" or the "optimal" setting. It means that for round shapes, their estimate is as good as it possibly can be. They didn't just guess this; they proved it by building a specific counter-example (a "test blob") that showed you can't make the ruler any more sensitive than 1/2 without it breaking.

How They Did It: The Magic of "Maurey" and "Prekopa"

So, how did they measure this wobble? They didn't just guess; they used a clever combination of two existing mathematical tools.

The Prekopa-Leindler Inequality: Think of this as a master key that unlocks the relationship between the volume of shapes and their average positions. The authors used a recent, super-precise version of this key that tells you exactly how much a shape must wobble if its volume is slightly off.
The Maurey Argument: This is like a mathematical magic trick. It allows you to take a complex problem and break it down into smaller, simpler pieces that you can solve one by one. By combining this trick with the master key, they were able to translate the "wobble" of the Log-Sobolev and Talagrand rules into the "wobble" of the Prekopa-Leindler rule, which they already knew how to measure.

They also used a concept called optimal mass transport, which is like figuring out the most efficient way to move a pile of sand from one place to another. This helped them connect the "distance" between shapes to the "energy" needed to move them.

What This Means for the Real World

The authors didn't just stop at measuring the wobble; they used their new ruler to solve a few other puzzles.

First, they figured out exactly what the "perfect" shapes look like when the error is zero. In the past, mathematicians knew the perfect shapes for simple cases (like the Gaussian or "bell curve" distribution), but for more complex landscapes, the answer was a mystery. The authors showed that the perfect shapes are always related to the "flat" directions of the landscape. If the landscape has a flat spot where you can slide without going up or down, the perfect shape is a slide in that direction.

Second, they applied their findings to Hopf-Lax semigroups. This sounds like a mouthful, but it's essentially a mathematical machine that simulates how things evolve over time, like how a wave moves or how heat spreads. The authors used their stability results to estimate how much "error" builds up in this machine. They proved that if the machine starts with a nearly perfect input, the output stays very close to perfect, and they gave a precise formula for how close it stays.

The Bottom Line

This paper is a triumph of precision. The authors didn't just say, "It's close." They said, "It's close by exactly this much, and here is the exact formula."

They proved that for general shapes, the distance from perfection is bounded by the error raised to the power of 1/19.
They proved that for round shapes, this bound tightens to the power of 1/2, which is the best possible result.
They showed that these results hold true for a wide variety of mathematical landscapes, not just the simple ones.

While they suspect that the 1/19 number might be improved for non-round shapes in the future, their current work provides a solid, unshakeable foundation. They have turned a vague idea of "closeness" into a hard, measurable fact, giving mathematicians a new tool to understand how the universe behaves when things are almost, but not quite, perfect.

Technical Summary: Quantitative Stability for Bakry–Émery Log-Sobolev and Talagrand Inequalities

Problem Statement
The paper addresses the problem of quantitative stability for functional inequalities, specifically the Bakry–Émery log-Sobolev inequality and the Talagrand transport inequality, within the setting of probability measures $d\mu_V = e^{-V}dx$ on $\mathbb{R}^n$ . While sharp inequalities and the characterization of their equality cases (extremizers) are well-established, the authors focus on estimating the "deficit" (the difference between the two sides of the inequality) in terms of a "natural distance" from the manifold of extremizers.

The specific challenges addressed are:

Extending stability results beyond the classical Gaussian measure to general Bakry–Émery measures where the potential $V$ satisfies $\nabla^2 V - cI_n \geq 0$ for some $c > 0$ .
Establishing $L^1$ -stability estimates with explicit exponents governing the deficit.
Characterizing the equality cases for these inequalities in the general Bakry–Émery framework, a gap noted in the existing literature.
Determining the optimality of the stability exponents, particularly in the radial setting.

Methodology
The authors employ a unified approach combining two primary mathematical tools:

Maurey-type Arguments: Adapted from Bobkov and Ledoux, this technique relates the log-Sobolev and Talagrand inequalities to the Prékopa–Leindler inequality. By constructing specific families of functions based on the potential $V$ and the test functions, the authors reduce the stability of the functional inequalities to the stability of the Prékopa–Leindler inequality.
Stability of the Prékopa–Leindler Inequality: The core of the quantitative estimates relies on recent stability results for the Prékopa–Leindler inequality.
- For the general case, the authors utilize the result by Böröczky and De, which provides an $L^1$ -stability estimate with an exponent of $1/19$ for log-concave functions.
- For the radial case, they apply the improved result by Figalli and Ramos, which yields an optimal exponent of $1/2$ under radial symmetry assumptions.

The proofs involve careful limit arguments (taking $\lambda \to 0^+$ in the Prékopa–Leindler setup) to recover the deficits of the log-Sobolev and Talagrand inequalities. The characterization of equality cases is further refined using the HWI inequality (Otto and Villani) and the Donsker–Varadhan variational formula.

Key Contributions and Results

Quantitative $L^1$ -Stability (General Case):
The paper establishes that for a positive log-concave function $f$ with a sufficiently small Bakry–Émery log-Sobolev deficit $\delta_{LSI}^{\mu_V}(f)$ , there exists a translation vector $x_0$ such that the $L^1$ distance between the normalized function and a shifted potential profile is bounded by:
$\int_{\mathbb{R}^n} \left| \frac{f^2(x)}{\alpha} - e^{V(x) - V(x-x_0)} \right| d\mu_V(x) \leq C_n \cdot (\delta_{LSI}^{\mu_V}(f))^{1/19}$
A similar $L^1$ -stability estimate with the same exponent $1/19$ is proven for the Talagrand transport inequality deficit $\delta_{Tal}^{\mu_V}(\nu)$ .
Optimality in the Radial Setting:
When the potential $V$ and the functions (or measures) are radially symmetric, the authors improve the exponent from $1/19$ to $1/2$ . They prove that this exponent is optimal and cannot be improved for the family of Bakry–Émery measures. This confirms a conjecture by Kim regarding the optimal exponent for $L^1$ -deficit estimates in the Gaussian case, extending it to the radial Bakry–Émery setting.
Characterization of Equality Cases:
The stability results provide elementary characterizations for the equality cases in the general Bakry–Émery framework:
- Log-Sobolev: Equality holds if and only if $f(x) = a \cdot e^{\langle x_0, x \rangle}$ where $x_0 \in \text{Ker}(\nabla^2 V - cI_n)$ .
- Talagrand: Equality holds if and only if the measure $\nu$ is a translation of $\mu_V$ by a vector $x_0 \in \text{Ker}(\nabla^2 V - cI_n)$ .
  The authors note that this algebraic characterization (involving the kernel of the Hessian of $V$ ) is equivalent to, but distinct in formulation from, previous analytic characterizations found in the literature (e.g., Arnold et al.).
Hypercontractivity Deficit:
As an application, the paper derives an estimate for the hypercontractivity deficit of the Hopf–Lax semigroup. The authors establish a relationship linking the hypercontractivity deficit to an integral of the log-Sobolev deficit along the semigroup flow. This leads to a characterization of the equality case for hypercontractivity estimates in the Bakry–Émery setting.

Significance and Claims
The paper claims to provide a "novel, elementary approach" to characterizing equality cases for a broad class of functional inequalities by bypassing the regularity issues often associated with the Brenier map and Monge–Ampère equations in optimal transport theory.

The authors emphasize that their results fill a gap in the literature regarding stability for non-Gaussian Bakry–Émery measures. Specifically, they provide the first $L^1$ -stability estimates with explicit exponents for the Talagrand inequality in this general setting and offer a complete characterization of equality cases for both the Talagrand and log-Sobolev inequalities without requiring the reference measure to be Gaussian. The identification of the optimal exponent $1/2$ in the radial case is presented as a confirmation of existing conjectures and a sharpening of previous stability bounds.

The work relies on adapting Maurey's argument via the Prékopa–Leindler inequality, demonstrating that this method successfully extends to the Talagrand transport inequality and provides a unified framework for analyzing stability and rigidity in the Bakry–Émery setting.

The Great Wobble Hunt

The Special Case: When Things Are Radial

How They Did It: The Magic of "Maurey" and "Prekopa"

What This Means for the Real World

The Bottom Line

More like this