Weighted Hardy inequalities for decreasing sequences and functions (Q818560)

This is a wonderful paper. The authors study the discrete weighted Hardy inequality \[ \left(\sum_{n=1}^{\infty}a_n\left(\sum_{k=1}^{n}x_k\right)^ q\right)^{1/q} \leq C \left(\sum_{n=1}^{\infty}b_nx_n^ p\right)^{1/p} \] for nonnegative non-increasing sequences \(\left\{x_n\right\}\), where \(\left\{a_n\right\}\) and \(\left\{b_n\right\}\) are given nonnegative weight sequences, \(0<p,q<\infty\) and \(C>0\) is independent of the sequence \(\left\{x_n\right\}\). They give a~complete characterization of the weights for which the inequality holds for all possible cases of \(p\) and \(q\). The discrete version of Hardy's inequality turns out to be considerably more difficult than its quite well-known and thoroughly studied integral counterpart, and all the results the authors obtain for the sums can be immediately translated to the language of integrals. Interestingly and rather surprisingly, this is not true vice versa. What goes for integrals, does not necessarily hold in the world of sums, where the absence of the comfort of a derivative makes things considerably more subtle (even though \textit{G. H. Hardy, J. E. Littlewood} and \textit{G. Pólya} claimed the opposite [``Inequalities'' (1934; Zbl 0010.10703), p. 11]). Despite the extensive investigation of the integral Hardy inequality by many authors, the case when \(0<q<p\leq1\) remained open. The authors obtain the characterization of the Hardy inequality in this missing case as a consequence of their discrete result. It is perhaps of interest to note that the very same result was independently obtained by different techniques, namely reduction methods, by Carro, Gogatishvili, Martín and the reviewer in a paper recently accepted to Journal of Operator Theory (see the journal website). The paper contains three main theorems: a discrete analogue of the famous Ariño and Muckenhoupt characterization of weighted Hardy inequality restricted to nonincreasing functions and to the case \(1\leq p=q\), a general discrete theorem involving all possible cases of \(p\) and \(q\), and a characterization of the integral inequality in the missing case. The second-named theorem is the most important one, it is a wonderful, comprehensive result, which brings unexpected surprises compared to the known integral analogues, and which required a great mathematical innovation not only for its proof but even for its formulation. Proofs are very deep, innovative and ingenious. Interesting connections are pointed out (for example to the Sawyer duality principle) and applications are considered (for example to the characterization of the boundedness of the maximal operator on weighted classical Lorentz spaces). Finally, integral counterparts of the results are formulated and a surprising counterexample is found to a claim made by \textit{V. D. Stepanov} [Soviet Math., Dokl. 43, No.~2, 620--623 (1994); translation from Dokl. Akad. Nauk SSSR 317, No.~6, 1308--1311 (1991; Zbl 0749.46019)]. The paper is very deep but it is written in an extraordinary reader-friendly way. The reader will enjoy even the technical part of the proofs. Truly a delightful stuff.