A description of norm-convergent martingales on vector-valued \(L^p\)-spaces (Q852726)

Let \(X\) be a Banach space and \(1\leq p < \infty\). The main Theorem 1.1 states that a martingale \((f_n,\Sigma_n)_{n=1}^\infty\) in \(L^p(\Omega, \Sigma, \mu, X)\) is norm convergent if and only if there are convergent martingales \((x_i^{(n)}, \Sigma_n)_{n=1}^\infty\) in \(L^p(\Omega, \Sigma, \mu)\) and elements \(y_i\in X\), \(i\in\mathbb{N}\), such that \[ f_n(s)=\sum_{i=1}^\infty x_i^{(n)}(s)y_i \quad\text{for all } s\in\Omega \text{ and } n\in\mathbb{N}, \quad \text{where} \] \[ \quad\left\| \sum_{i=1}^\infty \left| \lim_{n\to\infty} x_i^{(n)}\right| \right\|_{L^p(\mu)}< \infty \quad \text{and}\quad \lim_{i\to\infty} \| y_i\|_X=0\,. \] The authors obtain this as a special case of more abstract results in the framework of the following measure-free notion of Banach space valued martingales. Definition: A sequence of contractive projections \(T_i:X\to X\) on a Banach space \(X\) is called a BS-filtration if \(T_iT_j=T_{\min(i,j)}\) for all \(i,j\in \mathbb{N}\); and \((f_i,T_i)\) is called a martingale on \(X\) if \(T_if_j=f_i\) whenever \(i\leq j\).