Pointwise Convergence for Random Ergodic Averages in Non-commutative L p L^{p} -spaces

Christian Le Merdy, Safoura Zadeh · 2026

Read Pointwise Convergence for Random Ergodic Averages in Non-commutative L p L^{p} -spaces by Christian Le Merdy, Safoura Zadeh (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-2604-25029.txt.

Introduction

Le Merdy and Zadeh prove pointwise limits for random ergodic averages in non-commutative settings. They consider a semifinite algebra and a positive contraction on two spaces. Averages run along a random sparse subset set by Bernoulli draws. They prove the averages converge bilaterally to the ergodic projection, almost surely, for all powers above 1. This carries a theorem of Bourgain into the non-commutative setting.

Catalog record

CREATORChristian Le Merdy, Safoura Zadeh
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE51755
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.25029

Browse the full library · About the Tower of Babel

Full text

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

Pointwise Convergence for Random Ergodic Averages in Non-commutative L p L^{p} -spaces Christian Le Merdy, Safoura Zadeh Abstract . Let M M be a semifinite von Neumann algebra and T T a positive contraction on both L 1 ​ ( M ) L^{1}(M) and L ∞ ​ ( M ) L^{\infty}(M) . We consider ergodic averages along a random sparse subsequence determined by independent Bernoulli variables ( X n ) n ≥ 1 (X_{n})_{n\geq 1} with ℙ ⁡ ( X n = 1 ) = n − α \mathbb{P}(X_{n}=1)=n^{-\alpha} , and set W N = ∑ n = 1 N 𝔼 ⁡ [ X n ] W_{N}=\sum_{n=1}^{N}\mathbb{E}[X_{n}] . We prove that, almost surely, the averages 1 W N ​ ∑ n = 1 N X n ​ T n ​ ( x ) \frac{1}{W_{N}}\sum_{n=1}^{N}X_{n}\,T^{n}(x) converge bilaterally almost uniformly to the ergodic projection for all 1 < p < ∞ 1<p<\infty . This extends a theorem of Bourgain to the non-commutative setting. 2020 Mathematics Subject Classification: Primary 46L53, 46L55; Secondary 46L51, 37A99 Key words: random sets; individual ergodic theorems; non-commutative L^{p} -spaces. 1. Introduction In the 1930s, von Neumann proved a monumental mean ergodic theorem, establishing convergence in L^{2} , while Birkhoff obtained a pointwise ergodic theorem, which guarantees almost everywhere convergence and is often referred to as the individual ergodic theorem. Both results admit a common formulation: for a measure-preserving transformation T , the ergodic averages A_{N}(f):=\frac{1}{N}\sum_{n=1}^{N}T^{n}(f) converge almost everywhere for all f\in L^{1} , and converge in L^{p} for 1<p<\infty ; in general, L^{1} -norm convergence fails. The almost everywhere convergence result was later established within the Dunford–Schwartz ergodic theorem for a broad class of positive operators T , namely positive contractions on both L^{1} and L^{\infty} (see, for example, [ DS , Section 3, Theorem 6] and [ EFHN , Section 11.1] ). A celebrated theorem of Akcoglu and Sucheston [ AS ] extends this result to all positive contractions on L^{p} , 1<p<\infty , via a dilation argument. For further details, see Chapters 1 and 5 of [ Kre ] , along with the notes at the end of those chapters. A major direction of research since Birkhoff has been the study of ergodic averages along sparse arithmetic sets. A fundamental example arises when considering polynomial iterates such as \frac{1}{N}\sum_{n=1}^{N}T^{n^{2}}(f). The increasing gaps and arithmetic rigidity of such sequences prevent direct application of classical ergodic methods. In a seminal work, Bourgain [ B4 ] (see also [ B1 , B2 , B3 ] ) established almost everywhere convergence for polynomial sequences with integer coefficients using harmonic-analytic techniques based on discrete Fourier multipliers given by oscillatory exponential sums. These ideas revealed that cancellation in arithmetic phases compensates for sparsity and allows maximal estimates to control pointwise behaviour. A result of comparable nature is the theorem of Bourgain [ B3 ] and Wierdl [ W ] , which establishes the pointwise ergodic theorem along prime numbers for functions in L^{p} , p>1 . Given these developments, it is natural to ask what happens when the averaging set is no longer deterministic but random. A standard model is provided by sparse random sequences, for instance independent Bernoulli variables (X_{n}) with \mathbb{P}(X_{n}=1)=n^{-1/2} , and the associated random increasing sequence of hitting times n_{k}:=\min\{N:X_{1}+\cdots+X_{N}=k\}. In this case, the counting function \sum_{n=1}^{N}X_{n} grows on the order of \sqrt{N} . One may then ask whether the corresponding ergodic averages along this random subsequence, \frac{1}{m}\sum_{k=1}^{m}T^{n_{k}}(f), still converge almost surely. Bourgain [ B1 , Appendix 1] showed that, under suitable conditions, almost sure convergence persists even along sufficiently sparse random subsequences. The connection between ergodic theory and von Neumann algebras goes back to the early development of operator algebras. In the non-commutative framework, functions are replaced by operators affiliated with a von Neumann algebra, and classical L^{p} -spaces by their non-commutative counterparts. The first ergodic theorem here is due to Lance, who proved an ergodic theorem for trace-preserving * -automorphisms on a von Neumann algebra. This was sharpened by Yeadon in 1977, who established a non-commutative version of Birkhoff’s pointwise ergodic theorem. Building on these developments, Junge and Xu [ JX ] obtained a non-commutative extension of the Dunford–Schwartz ergodic theorem. As pointwise convergence no longer makes sense without an underlying space, it is replaced by bilateral almost uniform (b.a.u.) convergence [ Y ] , in the spirit of Egorov’s theorem. Roughly speaking, this means convergence is uniform outside a projection of arbitrarily small trace. As in the classical theory, maximal inequalities remain central to the study of ergodic behaviour. In the non-commutative setting, however, the absence of a total order rules out any direct notion of a pointwise supremum. To circumvent this, Junge [ Ju ] , extending Pisier’s construction [ Pi ] , introduced \ell^{\infty} -valued non-commutative L^{p} -spaces, which provide a workable substitute for pointwise suprema and a natural setting for maximal inequalities. The connection with b.a.u. convergence is crucial: much as maximal inequalities can lead to almost everywhere convergence in the commutative case, bounds in these \ell^{\infty} -valued spaces can yield b.a.u. convergence in the non-commutative setting. This principle, introduced in [ JX ] , is reflected in the closed subspace L^{p}(M;c_{0}) of L^{p}(M;\ell^{\infty}) , consisting of sequences that vanish at infinity. Not every result or method carries over to the non-commutative setting. As noted above, in the commutative case Akcoglu’s dilation theorem allows one to extend the Dunford–Schwartz theorem to general positive contractions. In the non-commutative framework, however, such dilation results fail in general; a counterexample is given in [ JLM ] . Notably, this example still satisfies a maximal inequality [ HRW , 7.2] . A dilation is nevertheless available for a restricted class of positive contractions in the noncommutative setting [ HRW ] , namely those that are separating (i.e. disjointness-preserving), also called Lamperti-type. In the commutative setting these coincide with weighted composition operators, and in the non-commutative case they admit an analogous rigid structural description [ HRW , LMZ ] . For this class, the dilation method continues to yield both maximal inequalities and norm convergence of the ergodic averages. With non-commutative analogues of the Birkhoff and Dunford–Schwartz ergodic theorems in place, it is natural to look for a counterpart of Bourgain’s theorem on ergodic averages along polynomial times. In recent work [ CHW ] , a non-commutative sampling principle is introduced, allowing discrete Fourier multiplier estimates to be transferred to the operator-valued setting. This leads to a maximal inequality for ergodic averages along certain polynomial (arithmetic) sequences in non-commutative L^{p} -spaces for p>\frac{\sqrt{5}+1}{2} , extending the harmonic-analytic component of Bourgain’s argument [ B2 ] to the non-commutative framework. However, unlike in the commutative case, it is unknown whether these maximal inequalities imply b.a.u. convergence. Consequently, obtaining a full noncommutative pointwise ergodic theorem by this method remains open. The results of the present paper can be viewed as a step in this direction, establishing b.a.u. convergence for a class of random sparse sequences. Let M be a semifinite von Neumann algebra. In this paper, we consider the same random subsequences described above, namely those given by the hitting times n_{k}:=\min\{N:X_{1}+\cdots+X_{N}=k\} . For a positive contraction on both M(=L^{\infty}(M)) and L^{1}(M) , we study the corresponding ergodic averages (1.1)

Read the complete text