Necessary and sufficient conditions under which convergence follows from summability by weighted means (Q1348716)

Concerning weighted means, virtually all known Tauberian results (by the authors in part) allegedly are covered here in terms of conditions implied by those of slow decrease or oscillation, respectively (cf. [\textit{G. H. Hardy}, Divergent series (At the Clarendon Press, Oxford) (1949; Zbl 0032.05801)]). Write \((s_n)=s\in(\overline N,p)\), where \(p=(p_n)\) such that \(p_n\geq 0\), \(0<P_n=\sum^n_{k=0}p_k\to\infty\), to denote proper convergence of \(P_n^{-1}\sum^n_{k=0}p_ks_k\). With natural \(\rho(n)\to\infty\) increasing, write \((\rho(n))= \rho\in \Lambda_u\) if \(\liminf P_{\rho(n)}/P_n>1\) and \(\rho\in \Lambda_\ell\) if \(\liminf P_n/P_{\rho(n)}>1\). (Strictness of increase, as was supposed for \(\rho\) on p. 399 bottom, must be dropped to enable definition (2.2). The assumption above is being approved on the authors' part.) Such ``upper and lower \(\rho\) admitted for \(p\)'' make \[ U_n= (P_{\rho(n)}-P_n)^{-1} \sum^{\rho (n)}_{k=n+1} p_k(s_k-s_n),\quad L_n= (P_n-P_{\rho(n)})^{-1}\sum^n_{k= \rho(n)+1} p_k(s_n-s_k) \] defined for large \(n\). Then, in order that any real-valued \(s\in (\overline N,p)\) be convergent it is necessary and sufficient that (1) \(\inf \{\limsup U_n: \rho\in\Lambda_u\}\leq 0\) and (2) \(\inf \{\limsup L_n: \rho\in \Lambda_\ell\}\leq 0\). For another version of the theorem have the places of ``inf limsup'', ``\(\leq 0\)'' in (1), (2) be taken by ``sup liminf'', ``\(\geq 0\)'' and thus arrive at a set of conditions equivalent to (1), (2) for \(s\in (\overline N,p)\). As to the case of complex \(s_n\), make (1), (2) turn into \((1')\), \((2')\) with \(|U_n|, |L_n|\) instead of \(U_n,L_n\) and ``=0'' instead of ``\(\leq 0\)''. Now, either of \((1')\), \((2')\) is sufficient for \(s\in (\overline N,p)\) to be convergent, and both hold for convergent \(s\).