Obata’s rigidity theorem in free probability

Charles-Philippe Diez † · 2026

Read Obata’s rigidity theorem in free probability by Charles-Philippe Diez † (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-05466.txt.

Introduction

Diez proves a free-probability version of Obata's rigidity theorem. In the old form, a sharp spectral gap forces a space to split off a 1D Gaussian factor. Under a non-commutative curvature-dimension rule, with Lipschitz conjugate variables, he shows that a non-zero saturator of the free Poincare inequality of Voiculescu must be affine in the generators. In short, the sharp case has one form. The von Neumann algebra must split off a freely complemented semicircular component, and the extremal directions form a free semicircular family.

Catalog record

CREATORCharles-Philippe Diez †
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE142725
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.05466

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Obata’s rigidity theorem in free probability Thanks: † Technion – Israel Institute of Technology, Faculty of Mathematics, Technion City, Haifa 3200003, Israel. [email protected] . Charles-Philippe Diez † Abstract . We establish a free analogue of Obata’s rigidity theorem. More precisely, Cheng and Zhou (2017) proved that on a weighted Riemannian manifold, the sharp spectral gap (Poincaré constant) is achieved only when the space splits isometrically off a one - dimensional Gaussian factor, providing an infinite - dimensional counterpart of Obata’s rigidity theorem. We obtain the corresponding phenomenon in free probability, extending it beyond the setting of analytic self - adjoint potentials: Assume a self-adjoint n n –tuple X = ( X 1 , … , X n ) X=(X_{1},\ldots,X_{n}) admits Lipschitz conjugate variables in the sense of Dabrowski (2014). Under a suitable non-commutative curvature–dimension condition, we show that any non-zero saturator of Voiculescu’s free Poincaré inequality must be an affine function of the generators. Consequently, we deduce that the von Neumann algebra M = W ∗ ​ ( X 1 , … , X n ) M=W^{\ast}(X_{1},\ldots,X_{n}) necessarily splits off a freely complemented semicircular component W ∗ ​ ( Y 1 ) ≃ L ∞ ​ ( [ − 2 , 2 ] , μ s ​ c ) W^{*}(Y_{1})\simeq L^{\infty}([-2,2],\mu_{sc}) , which is also maximal amenable in M M . More generally, whenever the first eigenspace of the free Laplacian Δ := ∂ ∗ ∂ ¯ \Delta:=\partial^{*}\bar{\partial} is finite-dimensional of rank r ≥ 1 r\geq 1 , our rigidity argument shows that these r r extremal directions form a free semicircular family, yielding a free product decomposition with an L ⁡ ( 𝔽 r ) L(\mathbb{F}_{r}) factor. This provides a free analogue of this classical Gaussian splitting phenomenon and reveals a rigidity mechanism under non-commutative curvature. 1. Introduction 1.1. Classical Rigidity and Stability in Spectral Geometry A central theme in geometric analysis is to understand the structure of spaces that maximize or minimize a given geometric or analytic quantity within a natural class. Extremal phenomena often force strong rigidity, revealing for example product structures or canonical geometric models. A nice example, which serves as a guiding analogy for our work, is the Cheeger–Gromoll splitting theorem [ 16 ] : a complete Riemannian manifold with non-negative Ricci curvature that contains a line (i.e. an infinite-length minimizing geodesic) must split isometrically as a product with \mathbb{R} . This result is the starting point of a broader principle: extremal geometric behavior, whether in diameter, curvature, isoperimetry, or spectral quantities, etc… typically forces the underlying space to exhibit strong structural constraints, and comparison with a reference space. Rigidity in spectral geometry begins with Obata’s theorem [ 61 ] , which identifies the round sphere as the unique extremal case of the Lichnerowicz estimate [ 51 ] . If a n -dimensional Riemannian manifold (M^{n},g) satisfies \mathrm{Ric}\geq(n-1)g , then the first non-zero eigenvalue \lambda_{1} of the Laplace–Beltrami operator obeys \lambda_{1}\geq n , and equality forces (M^{n},g) to be isometric to the standard sphere (\mathbb{S}^{n},g_{\mathrm{can}}) . This principle, saying that sharp eigenvalue bounds rigidly determine geometry, has driven a long line of developments: Cheeger’s inequality [ 15 ] links \lambda_{1} to isoperimetry, Payne–Weinberger [ 62 ] established sharp spectral gap estimates for convex domains, and Cheng’s comparison theorems [ 17 ] show how analytic inequalities encode global geometric information. The spectral gap is captured probabilistically by the Poincaré inequality: a probability measure \mu satisfies it with constant C_{P}>0 if \int f^{2}\,d\mu\;\leq\;C_{P}\int|\nabla f|^{2}\,d\mu\quad\text{whenever }\int f\,d\mu=0, and the optimal constant is C_{P}=1/\lambda_{1} . This inequality governs variance, mixing, and concentration, holds for many log-concave measures, and in one dimension is characterized by the Muckenhoupt criterion [ 58 ] . It sits within a hierarchy of functional inequalities: logarithmic Sobolev [ 35 ] , hypercontractivity, transportation inequalities, with deep consequences in convex geometry [ 9 , 48 ] , ergodic theory, and high-dimensional probability [ 57 ] . The Bakry–Émery \Gamma -calculus [ 2 ] unifies these ideas through the curvature–dimension condition CD(K,N) ; in particular, CD(1,\infty) implies an upper bound on the Poincaré constant C_{P}\leq 1 . This synthetic curvature framework extends to non-smooth metric measure spaces through the work of Sturm [ 69 , 70 ] , Lott–Villani [ 71 ] , and Ambrosio–Gigli–Savaré [ 1 ] . The rigidity phenomenon most relevant to our work is the theorem of Cheng and Zhou [ 18 ] , which shows that on spaces satisfying the curvature–dimension condition CD(1,\infty) , extremizers of the Poincaré inequality (equivalently, the sharp spectral gap \lambda_{1}=1 ) force a Gaussian splitting. Their result, extended to \mathrm{RCD}(1,\infty) spaces by Gigli, Ketterer, Kuwada, and Ohta [ 34 , 49 ] , identifies a one-dimensional Gaussian direction and yields an isometric and measure-theoretic product decomposition. For clarity, we recall the statement in the smooth setting: Theorem 1.1 (Cheng–Zhou [ 18 ] ; Gigli–Ketterer–Kuwada–Ohta [ 34 ] ) . Let (M,g,\mu) be a smooth weighted Riemannian manifold satisfying CD(1,\infty) . If there exists a centered non-zero f\in W^{1,2}(M,\mu) such that \int_{M}f^{2}\,d\mu\;=\;\int_{M}|\nabla f|^{2}\,d\mu, then (M,g,\mu) splits isometrically and measure-theoretically as (M,g,\mu)\ \simeq\ (\mathbb{R},|\cdot|,\gamma)\ \times\ (M^{\prime},g^{\prime},\mu^{\prime}), where \gamma is the standard Gaussian measure on \mathbb{R} . Recent work has also established the stability of this rigidity phenomenon: measures whose spectral gap is nearly optimal must be quantitatively close to Gaussian. This was shown by De Philippis–Figalli [ 28 ] , and improved by Courtade–Fathi [ 21 ] using Stein’s method, and then extended to \mathrm{RCD}(1,\infty) spaces by Bertrand–Fathi [ 4 ] using the extension of Klartag’s needle decomposition [ 50 ] to the RCD setting by Cavaletti–Mondino [ 13 ] . Altogether, these results show that extremal and near-extremal functional inequalities encode important geometric information. This classical picture, and especially the Cheng–Zhou splitting Theorem 1.1 therefore serves as the guiding analogy for our work, where we seek free analogues of such rigidity phenomena in the setting of free probability. 1.2. Free Probability Framework Free probability, introduced by Voiculescu [ 75 , 76 , 77 , 78 , 79 ] , provides a non - commutative analogue of classical probability in which independence is replaced by freeness . Random variables are modeled by operators in a tracial von Neumann algebra, and their joint distribution is encoded by mixed moments. Voiculescu’s discovery of asymptotic freeness [ 72 , 73 ] established the important bridge between random matrix theory and operator algebras: independent Gaussian or unitary ensembles behave, in the large - dimension limit, as freely independent non - commutative random variables. The semicircular law plays the role of the Gaussian distribution, and many classical analytic tools have free analogues, including the free Poincaré inequality [ 80 , 23 ] (and the free logarithmic Sobolev inequality of Biane and Speicher [ 7 , 8 ] ), expressed through the free difference quotient and closely tied to Voiculescu’s free entropy and free Fisher information [ 76 , 78 , 79 , 82 ] . Free entropy theory has had major consequences for the structure theory of free group factors L(\mathbb{F}_{n}),\>n\geq 2 . One of the main achievement of Voiculescu was the proof the absence of Cartan subalgebras [ 76 ] in free groups factors: the impossibility of realizing free group factors via measured

Read the complete text