Strong Convergence: A Short Survey

Basic Information

• Paper ID: 2510.12520
• Title: Strong Convergence: A Short Survey
• Author: Ramon van Handel (Princeton University)
• Classification: math.PR math.CO math.DG math.OA math.SP
• Submission Date/Conference: Submitted to arXiv on October 14, 2025; Conference paper for the International Congress of Mathematicians (ICM 2026)
• Paper Link: https://arxiv.org/abs/2510.12520

Abstract

A family of random matrices is said to strongly converge to a family of limit operators when the operator norms of every noncommutative polynomial of the random matrices converge to the corresponding norms of the limit operators. Recent developments surrounding the strong convergence phenomenon have yielded new progress on important problems in random graphs, geometry, operator algebras, and applied mathematics. This paper reviews classical and recent results in this field, as well as their applications across various areas of mathematics.

Research Background and Motivation

Core Problem

The core problem addressed by strong convergence research is: Given a sequence of random matrices $X^N = (X^N_1, \ldots, X^N_r)$ and a family of bounded operators $x = (x_1, \ldots, x_r)$, when does $\lim_{N\to\infty} \|P(X^N)\| = \|P(x)\|$ hold in probability for every *-polynomial $P$? We then say that $X^N$ strongly converges to $x$.

Importance of the Problem

1. Theoretical Significance: Voiculescu observed in 1993 that the existence of any model that strongly converges to a free limit model would resolve a long-standing conjecture in C*-algebra theory
2. Broad Applicability: The strong convergence phenomenon has important applications in random graphs, hyperbolic surfaces, minimal surfaces, operator algebras, and applied mathematics
3. Technical Breakthrough: In 2005, Haagerup and Thorbjørnsen first proved the existence of such random matrix models, initiating rapid development in this field

Limitations of Existing Methods

Traditional random matrix theory primarily focuses on weak convergence of spectral distributions, whereas strong convergence requires the property to hold for all *-polynomials, which is an extremely strong condition requiring new theoretical tools and proof techniques.

Core Contributions

1. Systematic Survey: The first comprehensive review of classical and recent results in the strong convergence field
2. Methodological Summary: Summarizes four main approaches to proving strong convergence: Schwinger-Dyson equations, moment methods, interpolation methods, and polynomial methods
3. Application Demonstration: Showcases the broad applications of strong convergence in random graphs, geometry, operator algebras, and other fields
4. New Theory Introduction: Provides detailed exposition of new developments including the intrinsic freeness principle and polynomial methods
5. Open Problems: Identifies important open problems and future research directions in the field

Detailed Methods

Task Definition

Definition of Strong Convergence: Let $X^N = (X^N_1, \ldots, X^N_r)$ be a family of random matrices for each $N \geq 1$, and let $x = (x_1, \ldots, x_r)$ be a family of bounded operators on a Hilbert space. If for every *-polynomial $P$, $\lim_{N\to\infty} \|P(X^N)\| = \|P(x)\| \text{ in probability}$ then we say that $X^N$ strongly converges to $x$.

Limit Models

Free Haar Unitary Operators

Let $F_r$ be the free group with $r$ free generators, and define the regular representation $\lambda: F_r \to B(\ell^2(F_r))$. Then $u_k = \lambda(g_k)$ are called free Haar unitary operators.

Free Semicircular Family

The free semicircular family $s_1, \ldots, s_r$ is the free probability analog of independent Gaussian random variables, and can be constructed via $s_k = \Phi(u_k + u_k^*)$, where $\Phi$ is an appropriately chosen continuous function.

Main Theorems

Theorem 2.2 (Haagerup-Thorbjørnsen; Schultz)

Let $X^N = (X^N_1, \ldots, X^N_r)$ be independent GUE/GOE/GSE matrices, and let $s = (s_1, \ldots, s_r)$ be a free semicircular family. Then $X^N$ strongly converges to $s$.

Theorem 2.3 (Collins-Male)

Let $U^N = (U^N_1, \ldots, U^N_r)$ be independent Haar-distributed random unitary matrices, and let $u = (u_1, \ldots, u_r)$ be free Haar unitary operators. Then $U^N$ strongly converges to $u$.

Theorem 2.4 (Bordenave-Collins)

Let $\Pi^N = (\Pi^N_1, \ldots, \Pi^N_r)$ be independent uniformly distributed random permutation matrices, and let $U^N_k = \Pi^N_k|_{1^\perp}$ be the restriction to the orthogonal complement of $1$. Then $U^N$ strongly converges to free Haar unitary operators.

Classification of Proof Methods

1. Schwinger-Dyson Equation Method: Uses the variational form of classical random matrix theory
2. Moment Method: Based on matrix-valued extensions of the non-backtracking method
3. Interpolation Method: Constructs continuous interpolation between random matrices and limit operators
4. Polynomial Method: Based on the observation that spectral statistics are regular functions of $1/N$

Experimental Setup

Theoretical Verification

The paper is primarily a theoretical survey without numerical experiments, but provides numerous theorems and application examples to verify the validity of the theory.

Application Cases

1. Random Graphs: Friedman's theorem as a special case of strong convergence
2. Hyperbolic Surfaces: Hide-Magee theorem constructing hyperbolic surfaces with optimal spectral gaps
3. Operator Algebras: Resolving the Peterson-Thom conjecture
4. Minimal Surfaces: Song's theorem constructing minimal surfaces with constant curvature

Experimental Results

Main Results

Applications to Random Graphs

• Theorem 4.2: For any fixed base graph $G$, its random $N$-lift $G^N$ satisfies $\lim_{N\to\infty} \|A^N|_{\text{new}}\| = \rho$, where $\rho$ is the spectral radius of the universal cover
• Theorem 4.3: Random Schreier graphs have optimal spectral gaps, even using only polynomial amounts of randomness

Applications to Geometry

• Theorem 4.5 (Hide-Magee): There exist sequences of closed hyperbolic surfaces $X^N$ such that $\lambda_1(X^N) \geq 1/4 - o(1)$
• Theorem 4.6: For any closed orientable hyperbolic surface $X$, in its random $N$-fold covers, a $1-o(1)$ fraction satisfy all new eigenvalues greater than $1/4 - o(1)$

Applications to Operator Algebras

• Theorem 4.8: $\text{Ext}(C^*_{\text{red}}(F_2))$ is not a group
• Theorem 4.9 (Hayes): Any diffuse amenable subalgebra of $L(F_r)$ is contained in a unique maximal amenable von Neumann subalgebra

Intrinsic Freeness Results

Theorem 3.2

For Gaussian random matrices $X$ and the corresponding free limit $X^{\text{free}}$, $P\left[d_H(\text{sp}(X), \text{sp}(X^{\text{free}})) > Cv(X)^{1/2}\|X^{\text{free}}\|^{1/2}((\log D)^{3/4} + t)\right] \leq e^{-t^2}$ where $v(X) = \|\text{Cov}(X)\|^{1/2}$ is the intrinsic parameter.

Related Work

Historical Development

1. 1993: Voiculescu establishes the connection between strong convergence and C*-algebra conjectures
2. 2005: Haagerup-Thorbjørnsen first prove the existence of strong convergence models
3. 2019: Bordenave-Collins prove strong convergence for random permutation matrices
4. 2023: Introduction and development of polynomial methods

Related Theories

1. Free Probability Theory: Voiculescu's theory provides the theoretical foundation for strong convergence
2. Random Matrix Theory: Classical Wigner matrices and Gaussian unitary ensemble theory
3. Spectral Graph Theory: Alon-Boppana bounds and Ramanujan graph theory
4. Operator Algebras: C*-algebras and von Neumann algebras theory

Conclusions and Discussion

Main Conclusions

1. Strong convergence is a fundamental phenomenon in random matrix theory with deep mathematical significance
2. The development of multiple proof methods makes strong convergence theory applicable to increasingly broad classes of models
3. Strong convergence has important applications across multiple mathematical branches, demonstrating its unifying value
4. The intrinsic freeness principle provides new tools for handling random matrices with arbitrary structure

Limitations

1. Computational Complexity: Computing the norms of complex *-polynomials remains difficult
2. Non-free Cases: Understanding beyond free groups remains limited
3. Quantitative Results: Many results lack precise convergence rate estimates
4. Constructiveness: Most existence results lack explicit constructions

Future Directions

1. Open Problems:
   • Which groups admit strong convergence permutation representations?
   • Can strong convergence be achieved through deterministic constructions?
   • How broadly applicable is the polynomial method?
2. New Applications: Potential applications in quantum information, machine learning, network science, and other fields
3. Theoretical Development:
   • Further development of non-asymptotic strong convergence theory
   • Generalization and refinement of polynomial methods
   • Deeper connections with other mathematical branches

In-Depth Evaluation

Strengths

1. Comprehensiveness: This is the first comprehensive survey of the strong convergence field, covering all aspects from foundational theory to cutting-edge applications
2. Systematicity: Clearly organizes different proof methods, application domains, and theoretical development trajectories
3. Forward-Looking: Not only summarizes existing results but also identifies important open problems and research directions
4. Technical Depth: Provides clear exposition and explanation of complex mathematical concepts and theorems
5. Interdisciplinary Impact: Demonstrates the unifying role of strong convergence theory across multiple mathematical branches

Weaknesses

1. High Technical Threshold: Requires readers to have background knowledge in random matrix theory, operator algebras, free probability, and other fields
2. Proof Details: As a survey paper, most theorems are only stated rather than fully proved
3. Numerical Verification: Lacks numerical experiments to intuitively demonstrate theoretical results
4. Application Examples: While many applications are listed, concrete computational examples are relatively sparse

Impact

1. Theoretical Contribution: Provides an authoritative reference and research guide for the strong convergence field
2. Educational Value: Offers excellent introductory material for graduate students and researchers in this field
3. Research Advancement: Through summarizing open problems, may inspire new research directions
4. Cross-disciplinary Communication: Promotes exchange and collaboration between different mathematical branches

Applicable Scenarios

1. Theoretical Research: Researchers in random matrix theory, operator algebras, and free probability
2. Applied Research: Researchers in random graphs, quantum information, network science, and other application domains
3. Teaching Reference: Reference material for advanced mathematics courses and graduate seminars
4. Interdisciplinary Collaboration: Researchers seeking mathematical tools to solve practical problems

References

The paper contains 95 references covering major works in the strong convergence field, including:

• Foundational Works: Voiculescu (1993), Haagerup-Thorbjørnsen (2005)
• Major Breakthroughs: Bordenave-Collins (2019), Collins-Male (2014)
• Recent Developments: Chen-Garza-Vargas-Tropp-van Handel (2025), Magee-Puder-van Handel (2025)
• Application Works: Hide-Magee (2023), Hayes (2022), Song (2025)

These references constitute the complete trajectory of strong convergence theory development and provide abundant resources for further research.