Convergence of polynomial ergodic averages of several variables for some commuting transformations (Q610628)

Let \(T_1,\dots, T_l\) be commuting invertible measure preserving transformations on the probability space \((X,{\mathcal B},\mu)\). The main theorem of this article is: Theorem 1.3. Let \((T_1,\dots, T_l)\) be a totally ergodic generating set, and let \(p_{ij}: \mathbb{Z}^d\to \mathbb{Z}\) for \(1\leq i\leq r\), \(1\leq j\leq l\) be polynomials. For any \(f_1,\dots, f_r\in L^\infty(\mu)\) and any Følner sequences \(\{\Phi_N\}^\infty_{N=1}\) in \(\mathbb{Z}^d\), the averages \[ {1\over|\Phi_N|} \sum_{u\in \Phi_N}\,\prod^r_{i=1} f_i(T^{p_{i1}(n)}_1\cdots T^{p_{il}(n)}x) \] converges in \(L^2(\mu)\) as \(N\to\infty\). Here, we call \((T_1,\dots, T_l)\) a totally ergodic generating set if \(T^{c_1}_1\cdots T^{c_l}_l\) is ergodic for any choice of \((c_1,\dots, c_l)\neq (0,\dots,0)\), and a sequence of finite subsets \(\{(\Phi_N\}^\infty_{N=1}\) of a countable discrete group \(G\) is a Følner sequence if for all \(g\in G\), \[ \lim_{n\to \infty} {|g\Phi_n \Delta\Phi_n|\over |\Phi_n|}= 0, \] where \(\Delta\) is the symmetric difference of operators. The main theorem is a consequence of the following proposition. Proposition 2.10. Let \((T_1,\dots, T_l)\) be a totally ergodic generating set, and \(P= \{p_{ij}: \mathbb{Z}^d\to\mathbb{Z}\) for \(1\leq i\leq r\), \(1\leq j\leq l\}\) be an ED-set of polynomials. Then there exists \(k\in\mathbb{N}\) such that for any \(f_1,\dots, f_r\in L^\infty(\mu)\) with \(|||f_m|||_k= )\) for some \(1\leq m\leq r\), we have \[ \limsup_{N\to\infty}\,\Biggl\|{1\over |\Phi_N|} \sum_{u\in \Phi_N}\,\Biggl(\prod^r_{i=1} T^{p_{i1}(n)}_1\cdots T^{p_{il}(n)}_l f_i\Biggr)\Biggr\|_{L^2(\mu)}= 0 \] for any Følner sequence \(\{\Phi_N\}^\infty_{N=1}\) in \(\mathbb{Z}^d\). Here, \(P= \{p_{ij}: \mathbb{Z}^d\to\mathbb{Z}\) for \(1\leq i\leq r\), \(1\leq j\leq l\}\) be an ED-set if all of the following hold: 1. Each \(p_{ij}\) in \(P\) is not equal to a nonzero constant. 2. No two polynomials \(p_{i_1,j}\), \(p_{i_2,j}\) in \(P\) differ by a nonzero constant for any \(j= 1,\dots,l\). 3. For each \(i= 1,\dots, r\), there is some \(j\in\{1,\dots,l\}\), where \(p_{ij}\) is nonzero. The norm \(|||\cdots|||_k\) is defined as follows. Let \(\mu^{[k]}_T\) be an invariant probability measure on \(X^{2^k}\) with respect to \(T^{2^k}\) defined inductively. \(\mu^{[0]}_T= \mu\) and let \({\mathcal I}^{[k]}_T\) be the \(\sigma\)-algebra of \(T^{[k]}\) of \(X^{2^k}\), and \(\mu^{[k+1]}_T= \mu^{[k]}_T\times_{{\mathcal I}^{[k]}_k} T^{[k]}\) to be the relatively independent square of \(\mu^{[k]}_T\) over \({\mathcal I}^{[k]}_T\). Then for \(f\in L^\infty(\mu)\), \[ |||f|||^{2^k}_k= \int_{X^{2^k}}\,\prod^{2^k- 1}_{j=0} f(x_j)\,d\mu^{[k]}_T(x). \] They use PET induction introduced by Bergelson for the proof of Proposition 2.10.