A note on the relation between ordinary and strong approximation of orthogonal series (Q1320478)

For orthonormal expansions with real coefficients of a square-integrable function \(f(x)\) on a finite interval two summability methods are considered. Given a double infinite matrix \(T=(\alpha_{ik})\) \((i,k=0,1,\ldots)\) of real numbers, ordinary \(T\)-means and strong \(T(p)\)- means are defined, respectively, as \[ t_ n(x): = \sum_{k=0}^ \infty \alpha_{nk} S_ k(x) \text{ and } T_ n (p,f;x): = \Bigl\{ \sum_{k=0}^ \infty \alpha_{nk} | S_ k (x) - f(x) |^ p \Bigr\}^{1/p}, \] where \(p>0\) and \(S_ n(x)\) is the \(n\)-th partial sum of the orthogonal series. A quantitative extension of an F. Móricz's result -- that, generally speaking, \(T\)-summability a.e. does not imply strong \(T(p)\)-summability - - is obtained. Namely, without going into details, for any positive nondecreasing sequence \(\gamma_ n\) satisfying some additional restrictions there exist an orthonormal series and a matrix \(T\) of the type considered so that \[ \lim_ n \gamma_ n^{-1} \bigl( f(x) - t_ n (x) \bigr) = 0, \text{ and } \lim_ n \gamma_ n^{-1} T_ n(p,f;x) = \infty \] a.e. and for any \(p>0\).