The \(q\)-concavity constants of Lorentz sequence spaces and related inequalities (Q1377066): Difference between revisions

Let \(d(w,p)\) denote the Lorentz sequence space determined by the weighting sequence \(w= (w_n)\). The main result identifies the \(q\)-concavity constant of \(d(w, 1)\) as \(\sup_{n\geq 1}\rho_{r, n}(w)/\rho_{1,n}(w)\), where \(r= q/(q- 1)\) and \(\rho_{r,n}(w)= n^{-1/r}(w^r_1+\cdots+ w^r_n)^{1/r}\). This is proved by considering the \(r\)-convexity constant and the dual space, and applies equally in the finite-dimensional case. The \(q\)-concavity constant of \(d(w, p)\) for \(1\leq p<q\) is then deduced easily, and the case \(q\leq p\) is elementary. We then address the problem of evaluating the above supremum for the two natural choices \(u\), \(v\) of weighting sequence defined by \(u_n= n^{-\alpha}\) and \(\sum^n_{i= 1}v_i= n^{1-\alpha}\). These problems are of some interest purely as questions on inequalities. In both cases, it is shown that the supremum of the above ratio equals its limit as \(n\to\infty\), which is easily seen to be \((1-\alpha)/(1-\alpha r)^{1/r}\). This is proved first for \(v\), then deduced for \(u\) using the majorization principle. In the case of \(v\), the ratio also increases with \(n\).