Some necessary conditions for absolute matrix summability factors (Q699630)

Let \(\sum_{n\geq 0} a_n= S\) be a series of real numbers and \(s= (s_n)_{n\geq 0}\) be the sequence of its partial sums. Let us denote by \(C_\alpha\) the Cesàro infinite matrix of order \(\alpha\) and by \(\sigma^\alpha_n\) the \(n\)th term of the \(C_\alpha\)-transform of \(s\), that is \(C_\alpha s\). In 1957 \textit{T. M. Flett} [Proc. Lond. Math. Soc., III. Ser. 7, 113-141 (1957; Zbl 0109.04402)] introduced the notion of \(|C_\alpha|_k\)-absolute summability \((k\geq 1)\) as follows. \(S\) is \(|C_\alpha|_k\)-absolutely summable if \[ \sum_{n\geq 1} n^{k-1}|\sigma^\alpha_n- \sigma^\alpha_{n-1}|^k< \infty.\tag{1} \] Using a general lower triangular matrix \(T\) and denoting by \(t_n\) the \(n\)th term of the \(T\)-transform of \(s\), the condition (1) can be extended and we say that \(S\) is \(|T|_k\)-absolutely summable if \[ \sum_{n\geq 1} n^{k-1}|t_n- t_{n-1}|^k< \infty.\tag{2} \] In the seventies, another notion of absolute summability not equivalent to (1) or (2) was introduced and used in the setting of the problems of summability theory. Let \((p_n)_{n\geq 0}\) be a sequence of positive numbers such that \(P_n= \sum^n_{i=0} p_i\to \infty\) as \(n\to\infty\) and consider the finite lower triangular matrix \(P= (p_{n,i})_{i,n\geq 0}\) where \(p_{n,i}= {p_i\over P_n}\) and let us denote by \(\beta_n\) the \(n\)th term of the \(P\)-transform of \(s\). The series \(s\) is \([\overline N,p_n]_k\)-absolutely summable if \[ \sum_{n\geq 1} \Biggl({P_n\over p_n}\Biggr)^{k- 1}|\beta_n- \beta_{n-1}|^k< \infty. \] We say that \(s\) is \([\overline N,p_n]_k\)-bounded if \(\sum_{n\geq 1} p_i|s_i|^k= O(P_n)\) as \(n\to\infty\). In particular, given the lower triangular infinite matrix \(T= (t_{n,i})_{n,i\geq 0}\), taking \(p_n= 1\) for \(n= 0,1,\dots\) we have the notion of \(|T|_k\)-boundedness. In the paper, given a \(|T|_k\)-bounded series \(s\), the authors obtain that if the series \(\sum_{n\geq 0} a_n\lambda_n\) is \(|T|_k\)-summable then the following conditions are necessary: \hskip 10mm (1) \(\lambda_i= O((it_{i,i})^{{1\over k}- 1})\). \hskip 10mm (2) \(\sum_{n\geq i+1} n^{k-1}|\widehat t_{n,i}\lambda_i|^k= O(1)\) where \(\widehat t_{n,i}=\overline t_{n,i}-\overline t_{n-1,i}\) and \(\overline t_{n,i}= \sum^n_{j=i} t_{n,j}\). \hskip 10mm (3) \(\sum^\infty_{n= i+1} n^{k- 1}|\widehat t_{n,i} \lambda_i-\widehat t_{n-1,i}\lambda_i|^k= O(t_{i,i})\). This result extends others previously obtained in the setting of the \([\overline N,p_n]_k\)-summability. Taking as \(T\) the identity matrix, \(k= 1\) and considering the series \(\sum_{n\geq 0}(- 1){1\over n+1}\), it is also proved that the former conditions are not sufficient. Currently an open question is to find necessry and sufficient conditions for the series \(\sum_{n\geq 0} a_n\lambda_n\) to be \(|T|_k\)-absolutely summable.