Polynomial ergodic averages of measure-preserving systems acted by \(\mathbb{Z}^d\) (Q6118068)

Many questions in multiple recurrence or infinite combinatorics reduce to the following sort of question (here called the Furstenberg-Bergelson-Leibman conjecture from [\textit{V. Bergelson} and \textit{A. Leibman}, Invent. Math. 147, No. 2, 429--470 (2002; Zbl 1042.37001)]): Given a collection of invertible measure-preserving transformations \(T_1,\dots,T_d\) of a Lebesgue probability space \((X,\mu)\) that together generate a nilpotent group, polynomials \(p_{i,j}\in\mathbb{Z}[x]\) for \(1\leqslant i\leqslant d\), \(1\leqslant j\leqslant m\), and any \(f_1,\dots,f_m\in L^{\infty}(X)\), do the polynomial ergodic average \[\frac1N\sum_{n=0}^{N-1}\prod_{j=1}^{m}f_j (T_1^{p_{1,j}(n)}\cdots T_d^{p_{d,j}(n)}x)\] converge as \(N\to\infty\)? Multiple authors contributed to progress on special cases (that is, with bounds on \(d\), \(m\), or the complexity of the group generated by the transformations) culminating in a proof of convergence in \(L^2\) by \textit{M. N. Walsh} [Ann. Math. (2) 175, No. 3, 1667--1688 (2012; Zbl 1248.37008)] that used a decomposition into random and structured components of the system, the latter being handled via an inductive argument. Here work of \textit{J.-M. Derrien} and \textit{E. Lesigne} [Ann. Inst. Henri Poincaré, Probab. Stat. 32, No. 6, 765--778 (1996; Zbl 0868.60028)] is extended from actions of \(\mathbb{Z}\) to actions of \(\mathbb{Z}^d\), showing that pointwise convergence almost everywhere under this hypotheses of commuting transformations reduces to the zero entropy case. This gives the pointwise convergence almost everywhere result for systems of completely positive entropy. A key technical part of the argument involves choosing an appropriate algebraic past in \(\mathbb{Z}^d\) and using this to describe the structure of the Pinsker \(\sigma\)-algebra.