On negative results concerning Hardy means (Q313436): Difference between revisions

A function \(\mathfrak A:\bigcup_{i=1}^{\infty}I^n \mapsto \mathbb R_+\), where \(I\subset \mathbb R_+\) is an interval with \(\inf I = 0\), is called a Hardy mean on \(I\) if there is a positive \(C\) such that for any convergent sequence \(a_1, a_2, \ldots\) we have that \( \sum_{n=1}^{\infty} \mathfrak A(a_1, \ldots, a_n)<C\sum_{n=1}^{\infty}a_n> \) The name arises from a 1920 result of Hardy that a \(p\)-power mean is Hardy if and only if \(p<1\); see [\textit{G. H. Hardy}; Math. Zeitschr. 6, 314-317 (1920; JFM 47.0207.01); \textit{Zs. Páles, L.-E. Persson}, Bull. Aust. Math. Soc. 70, No. 3, 521--528 (2004; Zbl 1058.26015); \textit{Y.-M. Chu, W.-F. Xia}, J. Inequal. Appl. 2009, Article ID 763252, 7 p. (2009; Zbl 1187.26014); \textit{J. A. Oguntuase, L.-E. Persson}, Aust. J. Math. Anal. Appl., No. 7, 1-19 (2011)]. In this paper, the following result is proved: if for some sequence \(c_n, 1\leq n<\infty\), in \(I\) with \(\sum_{i=1}^{\infty}c_n=\infty\) we have that \(\lim c_n^{-1}\mathfrak A(c_1, \ldots, c_n) = \infty\) then \(\mathfrak A\) is not Hardy. The proof proceeds by assuming \(\mathfrak A\) is Hardy with constant \(C\) and using the sequence \(c_n, 1\leq n<\infty\), to obtain a sequence \(a_1, a_2, \ldots\) with \( \sum_{n=1}^{\infty} \mathfrak A(a_1, \ldots, a_n)>C\sum_{n=1}^{\infty}a_n\). Using this result it is shown that a Gini mean \(\mathfrak G_{p,q}\) is Hardy if and only if \(\min(p,q) \leq 0\) and \( \max(p,q)<1\) and that no symmetric polynomial mean is Hardy.