Pointwise and {\(L^1\)} mixing relative to a sub-sigma algebra (Q1765975)

Let \(T\) be a measure preserving transformation on a probability space with invariant measure \(\mu\), and \(\mathcal H\) be a \(T\)-invariant sub \(\sigma\)-algebra (factor-algebra). Define \(L^{2,\infty}_{\mathcal H}\) to be the set of all functions \(f\) such that the conditional expectation of \(| f| ^2\) with respect to \(\mathcal H\) is in \(L^\infty(\mu)\). The author defines \(T\) to be \(L^1\)-relatively mixing with respect to a factor-algebra \, \(\mathcal H\) if as \(| n| \to\infty\), \[ \| E(f\cdot g\circ T^n| \mathcal H)- E(f| \mathcal H)E(g\circ T^n| \mathcal H)\| _1\to 0 \] for all \(f\) and \(g\) in \(L^2\), and pointwise-relatively mixing with respect to \, \(\mathcal H\) if as \(| n| \to\infty\), \[ | E(f\cdot g\circ T^n| \mathcal H)- E(f| \mathcal H)E(g\circ T^n| \mathcal H)| \to 0 \] pointwise a.s. for all \(f\) and \(g\) in \(L^{2,\infty}_\mathcal H\). These are shown to be different notions: a map that is \(L^1\)-, but not pointwise-mixing is presented. It is shown that, to prove either type of mixing, it is enough to consider \(g=f\) and \(E(f| \mathcal H)=0\), thus study the convergence to zero of \(| E(f\cdot f\circ T^n| \mathcal H)| \). The author proves rather easily that, to verify if \(T\) is \(L^1\)-relatively mixing, it is enough to check the above convergence on an \(L^2\)-dense family of functions \(f\). The central result of the paper is that the same is true for the pointwise-mixing property. As an application, it is shown that if \(T\) is mixing and \(S\) is ergodic, then the skew-product \(S\times T\) is pointwise-relatively mixing w.r.t.~the first coordinate sub \(\sigma\)-algebra.