🔢 mathematics

Stability of the Markov-Krein Transform Under Classical Convolutions

This paper proves that the generalized Markov-Krein transform preserves closure under both classical additive and Mellin convolutions for probability measures with finite logarithmic moments, thereby resolving the positivity conjecture for the additive high-temperature convolution in random matrix theory.

Original authors: Pierre Mergny

Published 2026-10-14

📖 6 min read🧠 Deep dive

CC0 1.0

Original authors: Pierre Mergny

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

In the vast landscape of probability, where scientists study how random events combine to shape the world, there are two fundamental ways that independent quantities can merge. The most familiar is simple addition: if you roll two dice, the total is the sum of their faces. In mathematics, this is called classical convolution, a rule that governs everything from the spread of heat to the movement of stock prices. But in the last few decades, a different kind of addition has emerged from the study of massive, complex systems like the nuclei of atoms or the behavior of huge networks. This is known as free convolution. While classical addition describes how independent things pile up, free convolution describes how things interact when they are part of a giant, interconnected whole where the usual rules of independence break down. For a long time, these two worlds seemed separate, governed by different laws. However, researchers have recently discovered a bridge between them, a mathematical tool that can translate between the familiar world of simple sums and the exotic world of complex interactions. This tool, known as the Markov-Krein transform, acts like a lens that reveals the hidden structure of random systems, but for years, a critical question remained unanswered: does this lens work reliably for every possible type of random data, or does it crack under certain conditions?

A researcher has now answered that question with a definitive proof, showing that this mathematical lens is robust and stable across the entire spectrum of random data. The study focuses on a specific operation that creates a new kind of probability distribution by combining two existing ones. In the past, this operation was only known to work for very specific, well-behaved types of data, leading to a lingering doubt among experts: if you tried to combine two random distributions that were messy or had extreme outliers, would the result still be a valid probability distribution, or would the math break down and produce nonsense? The author of this new work has demonstrated that the answer is a firm no; the math does not break. They proved that for any two random distributions that have a reasonable amount of information about their spread, the process of combining them through this specific lens always yields a valid, positive result. This resolves a long-standing conjecture that had been debated in the physics and mathematics communities, confirming that the bridge between classical and free probability is solid for all practical cases.

The researcher arrived at this conclusion by treating the problem as a puzzle of shapes and functions rather than just numbers. They focused on a special class of mathematical functions that behave in a very orderly way, known as prestarlike functions. These functions have a unique property: when you combine them using a specific multiplication rule, the result remains just as orderly as the originals. By translating the problem of combining random distributions into the language of these orderly functions, the team was able to show that the combination always works. They first proved this for distributions that are confined to a small, finite range, where the math is easier to handle. Then, using a clever trick involving stretching and shrinking the data, they extended the proof to cover all possible distributions, including those that stretch out to infinity. This step-by-step approach allowed them to show that the stability holds true regardless of how wild or complex the original data might be.

The significance of this finding extends beyond pure mathematics. The operation they studied is not just a theoretical curiosity; it describes the behavior of systems in the "high-temperature" limit, a state where random fluctuations are so intense that they smooth out the details of individual components. This scenario appears frequently in the study of random matrices, which are used to model everything from the energy levels of heavy atoms to the connectivity of the internet. In these systems, scientists often look for a parameter that controls the transition from classical behavior to free behavior. The new proof confirms that this transition is smooth and predictable for all inputs, removing a major obstacle for physicists and mathematicians who rely on these models to understand complex systems. It means that when they use these tools to predict the behavior of massive networks or quantum systems, they can do so with the confidence that the underlying mathematics will not fail, even when the data is imperfect or extreme.

Perhaps most surprisingly, the researcher also uncovered a similar stability for a different kind of combination: multiplication. Just as random variables can be added together, they can also be multiplied, a process that is crucial for understanding how quantities grow or shrink in tandem. While the additive case had been the subject of intense debate, the multiplicative version had remained largely unexplored in this specific context. The team showed that the same stability holds here as well, proving that the lens works just as effectively for products as it does for sums. This suggests a deep, underlying symmetry in how random systems interact, whether they are piling up or multiplying. The work does not just solve a specific problem; it opens the door to a broader understanding of how randomness behaves at the edges of mathematical possibility, providing a solid foundation for future discoveries in fields ranging from statistical physics to representation theory.

In the end, the paper offers a quiet but powerful reassurance to the scientific community. It confirms that the tools used to navigate the complex seas of random interactions are reliable, even in the roughest waters. By proving that the Markov-Krein transform is stable under both addition and multiplication for all reasonable inputs, the author has closed a chapter of uncertainty that had persisted for years. Their work stands as a testament to the power of connecting abstract mathematical structures to concrete physical problems, showing that even in the realm of pure chance, there are rules that hold firm. For anyone studying the hidden patterns of the universe, this result provides a clear path forward, ensuring that the bridge between the known and the unknown remains sturdy and traversable.

Technical Summary: Stability of the Markov-Krein Transform Under Classical Convolutions

Problem Statement
The paper addresses the stability of the Markov-Krein transform (MKT), denoted $M_c$ , under classical additive and multiplicative (Mellin) convolutions. The MKT maps a probability measure $\mu$ with finite logarithmic moment to the law of the mean of a Dirichlet process with concentration parameter $c > 0$ and base measure $\mu$ .

In the context of random matrix theory and representation theory, specifically the high-temperature limit where $N \to \infty$ and $N\beta_N/2 \to c$ , a binary operation $\mu \star_c \nu$ was proposed to interpolate between classical convolution ( $c \to 0$ ) and free additive convolution ( $c \to \infty$ ). Previous works (e.g., [MP22], [BGCG22]) established that the MKT of this operation corresponds to the classical convolution of the transformed measures: $M_c(\mu \star_c \nu) = (M_c \mu) * (M_c \nu)$ . However, a critical "positivity conjecture" remained unresolved: it was not proven that for arbitrary input measures $\mu$ and $\nu$ , the resulting distribution $\mu \star_c \nu$ (or its associated moment sequence) actually corresponds to a valid probability measure. Furthermore, the multiplicative counterpart (high-temperature product) had not been constructed in the literature.

Methodology
The author resolves these issues by establishing a rigorous connection between the Markov-Krein transform and the theory of universally prestarlike functions. The proof proceeds in three main stages:

Reduction to Prestarlike Functions: The paper utilizes the class of universally prestarlike functions, denoted $R^{(u)}_\alpha$ , and the Ruscheweyh convolution operator $\Delta_a$ . A key theorem by Ruscheweyh, Stankus, and others ([RSS09], [RS08]) states that the Hadamard product (coefficient-wise product of power series) of two universally prestarlike functions remains in the same class.
- The author links the MKT to these functions by defining a holomorphic function $H_\mu(z) = z \mathbb{E}[(1-zY_\mu)^{-1}]$ derived from the Dirichlet mean $Y_\mu$ .
- It is shown that $H_\mu$ belongs to the class $R^{(u)}_{1-c/2}$ if and only if the quotient of specific Ruscheweyh derivatives of $H_\mu$ belongs to the Stieltjes class $T$ (functions representable as Stieltjes transforms of probability measures).
Multiplicative Case (Compact Support): For compactly supported measures on $\mathbb{R}_+$ , the independence of Dirichlet means $Y_\mu$ and $Y_\nu$ implies that the generating function of their product $Y = Y_\mu Y_\nu$ corresponds to the Hadamard product of their individual generating functions: $H_Y = H_\mu \odot H_\nu$ . By the closure theorem for universally prestarlike functions, $H_Y$ remains in the class, implying the existence of a unique probability measure $\rho$ such that $M_c \rho = (M_c \mu) \odot (M_c \nu)$ .
Extension to General Measures:
- Additive Case: The additive stability is derived from the multiplicative result via an affine contraction argument. By mapping measures to a compact interval, applying the multiplicative result to a transformed product, and taking the limit as the contraction parameter vanishes, the additive convolution stability is established.
- Unbounded Measures: The results are extended from compactly supported measures to measures with finite logarithmic moments ( $P_{\log}(\mathbb{R})$ ) using a standard weak-convergence argument. The author proves the continuity of the MKT and its inverse under weak convergence, ensuring the existence of the limit measure.

Key Contributions and Results
The paper establishes Theorem 1.4, which asserts that for any $c > 0$ :

(i) Additive Stability: For any $\mu, \nu \in P_{\log}(\mathbb{R})$ , there exists a unique probability measure $\mu \star_c \nu \in P_{\log}(\mathbb{R})$ such that $M_c(\mu \star_c \nu) = (M_c \mu) * (M_c \nu)$ .
(ii) Multiplicative Stability: For any $\mu, \nu \in P_{\log}(\mathbb{R}_+)$ , there exists a unique probability measure $\mu \boxtimes_c \nu \in P_{\log}(\mathbb{R}_+)$ such that $M_c(\mu \boxtimes_c \nu) = (M_c \mu) \odot (M_c \nu)$ .

The proof relies on a convolution theorem for universally prestarlike functions, providing a new analytic tool for studying the MKT.

Significance and Claims
The paper claims to resolve the positivity conjecture for the additive high-temperature convolution proposed in [MP22] and [BGCG22]. Prior to this work, the existence of the measure $\mu \star_c \nu$ for arbitrary inputs was only conjectured or proven for specific cases (e.g., symmetric Bernoulli measures or convolutions with Gaussian laws). This result extends the construction to the entire space of probability measures with finite logarithmic moments.

Additionally, the paper constructs the multiplicative counterpart (high-temperature product) for the first time, proving that the Markov-Krein transform of the new measure equals the Mellin convolution of the transformed measures (i.e., $M_c(\mu \boxtimes_c \nu) = (M_c \mu) \odot (M_c \nu)$ ). While the paper notes that the correspondence between this construction and the high-temperature limit of rank-one multiplicative spherical integrals (suggested by [MP21], [Hus25]) is strongly indicated, it explicitly leaves the rigorous proof of this specific physical correspondence for future work, focusing instead on the probabilistic existence and uniqueness of the measure.

The work bridges complex analysis (prestarlike functions) with probability theory (Dirichlet processes and random matrix limits), offering a unified framework for understanding the stability of the Markov-Krein transform under classical operations.

Technical Summary: Stability of the Markov-Krein Transform Under Classical Convolutions

More like this