A Tauberian theorem for ergodic averages, spectral decomposability, and the dominated ergodic estimate for positive invertible operators (Q1417872)

Let \((\Omega,\mathcal S,\mu)\) be a \(\sigma\)-finite measure space, let \(T\colon L^p(\mu)\to L^p(\mu),\;1<p<\infty\), be a bounded, invertible linear operator such that both \(T\) and \(T^{-1}\) are positive, and let \(\mathcal C_n(T):= {1\over n} \sum_{0 < |j|\leq n}T^j\) be the \((C,1)\)-means of the sequence \(\{T^j+T^{-j}\}^\infty_{j=1}\). Under these conditions on \(T\), one finds: (a) For each \(i\) there exists a positive function \(g_i\) such that \(T^i f=g_i S^i f,\;\forall f\in L^p(\mu)\), and \(g_{i+j}=g_i S^i g_j\), where \(S\) is a positive and multiplicative linear map acting on measurable functions. (b) For each \(i\) there exists a positive function \(J_i\) such that \(J_{i+j}=J_i S^i J_j\) and \(\int_\Omega J_i S^i fd\mu=\int_\Omega f\,d\mu,\;\forall f\in L^1(\mu)\). (c) For each \(i\), the adjoint of \(T^i\) is the operator given by \((T^i)^* f=(g_{-i})^{-1} J_{-i} S^{-i} f\). These, in turn, yield:\ (d) If \(h_i=\colon g_i^{-p} J_i\), then \(\int_\Omega |T^i f|^p \,d\mu=\int_\Omega |f|^p \,d\mu,\;\forall f\in L^p(\mu)\); and: (e) If \(q=(1-p^{-1})^{-1}\), then for every nonnegative \(f\) in \(L^q(\mu)\), \(h_i(T^if^{q-1})^{p-1}=(T^{-1})^* f\). \textit{F. J.~Martin-Reyes} and \textit{A.~de la Torre} [Proc.~Am.~Math.~Soc.~104, 69--75 (1988; Zbl 0679.47007)] showed that the following conditions are equivalent: (i) There exists a positive constant \(C\) such that, for the maximal operator \(M\) defined by the \(\mathcal C_n(T)\), one has \(\int_\Omega |Mf|^p d\mu\leq C\int_\Omega |f|^p \,d\mu\), \(\forall f\in L^p(\mu)\). (ii) There exists a positive constant \(C\) such that \(\forall f\in L^p(\mu),\;\sup_{n\geq 0}\int_\Omega |\mathcal C_n(T)|^p d\mu\leq C\int_\Omega |f|^p \,d\mu\). (iii) For almost all \(x\), the function defined on \(\mathbb Z\) by \(i\mapsto h_i (x)\) satisfies the \(A_p\) condition with a constant independent of \(x\). In the present work, the authors show that the strong convergence of the \((C,2)\)-means of \(\{T^j+T^{-j}\}^\infty_{j=1}\) implies (ii). This result, together with earlier work of the authors [Stud. Math.~144, 39--61 (2001; Zbl 0971.47004)] concerning the strong convergence of the \((C,2)\)-weighted ergodic averages for all trigonometrically well-bounded operators, yields a spectral-theoretic characterization of the maximal ergodic estimate, as well as the equivalence of the uniform boundedness of the rotated Hilbert averages of \(T\) and the uniform boundedness of \(\{\mathcal C_n (T)\}^\infty_{n=1}\).