Almost uniform and strong convergences in ergodic theorems for symmetric spaces (Q2416462)

A space $X\subset L^0_\nu$ is fully symmetric on $((0,\infty),\nu)$ if $f\in X$, $g\in L^0_\nu$, and the decreasing rearrangement of $f$ dominates that of $g: g^\ast\leq f^\ast$ pointwise (resp., $\int_0^s g^\ast (t)\, dt\leq \int_0^s f^\ast(t)\, dt$ for all $s>0$) implies that $g\in X$ and $\Vert g\Vert_X\leq \Vert f\Vert_X$. \par The first main result extends the Dunford-Schwartz pointwise ergodic theorem in characterizing $\mathcal{R}_\mu=\{f\in L^1+L^\infty: \forall\lambda>0,\, \mu(\vert f\vert >\lambda)<\infty\}$ as a space on which pointwise ergodic limits converge uniformly. To be precise, let $(\Omega,\mathcal{A},\mu)$ be a measure space and $X$ a fully symmetric space on $(\Omega,\mathcal{A},\mu)$ such that the constant $1\notin X$. If $T\in DS$ (that is, $T$ is bounded on both $L^1$ and $L^\infty$) and $f\in X$, then the averages $M_n(T)(f)=\frac{1}{n}\sum_{k=0}^{n-1} T^k(f)$ converge a.u. to some $\hat{f}\in X$. In particular, $M_n(T)(f) =\frac{1}{n}\sum_{k=0}^{n-1} T^k (f)\, \to\hat{f}\in \mathcal{R}_\mu$ a.u. when $f\in \mathcal{R}_\mu$. \par The proofs involve a reduction to the $L^1$ case and the Hopf maximal theorem ($\int_{M(T)^\ast(f)>0} f\, d\mu>0$), where $M(T)^\ast(f)(x)=\sup \vert M_n(T)(f)(x)\vert $, and the weak type inequalities $$\mu(M(T)^\ast(\vert f\vert)>\lambda)\leq \bigl(2\frac{\Vert f\Vert_p}{\lambda}\bigr)^p,\quad\lambda>0\, .$$ \par Theorem 3.4 then states that, if $\mu$ is $\sigma$-finite, then $\mathcal{R}_\mu$ is the largest subspace of $L^1+L^\infty$ on which convergence is almost uniform, that is, if $f\in (L^1+L^\infty)\setminus \mathcal{R}_\mu$, then there is a $T\in DS$ such that the sequence $M_n(T)(f)$ fails to converge almost everywhere. In fact, the maximality of $\mathcal{R}_\mu$ for a subspace $X$ is equivalent to constants not belonging to $X$. Orlicz spaces are used to illustrate the condition that constants are not members. Strong convergence of Cesàro means is also discussed in the context of characterizing validity of mean ergodicity for fully symmetric spaces.