An inequality for equimeasurable rearrangements and its application in the theory of differentiation of integrals (Q1079680)

In the first part of the note, an important theorem is proved, namely: Let \(\Phi\) be a nonnegative, nondecreasing function, and let f be nonnegative measurable functions on \(I^ N=[0,1]^ N\). 1. If \(\Phi\) is convex, then \[ \int_{I^ N}\Phi (\sum^{\infty}_{k=1}f_ k(x))dx\leq \int^{1}_{0}\Phi (\sum^{\infty}_{k=1}f^*_ k(t))dt. \] 2. If \(\Phi\) is concave, then \[ \int_{I^ N}\Phi (\sum^{\infty}_{k=1}f_ k(x))dx\geq \int^{1}_{0}\Phi (\sum^{\infty}_{k=1}f^*_ k(t))dt, \] where, \(f^*\) is an equimeasurable rearrangement of \(f\in S(R^ N)\)- the space of Lebesgue measurable functions. The above result is used for the demonstration of Theorem 2 concerning the differentiability of some integrals: For every positive function \(\sigma\) with \(\lim_{t\to \infty}\sigma (t)=0\), there exists a nonnegative, measurable function f such that for every orthogonal basis \(\theta\) in the Euclidean space \(R^ N\), \(\bar D_{B(s,\theta)}(\int f,x)=\infty \quad a.e.\quad in\quad I^ N,\) and \(\int_{I^ N}\sigma (f)f(\ln^+_ nf)^{N-s}dx<\infty,\quad s=1,2,...,N-1.\) \((B_{(s,\theta)}\) is a differential basis in \(R^ N)\). Mention must be made of \textit{B. López Melero's} paper [Stud. Math. 72, 173-182 (1982; Zbl 0531.28006)] which contains a generalization of Theorem 2.