Lineability of non-differentiable Pettis primitives (Q2354033)

The main result of the paper is the very interesting discovery that the set \textbf{ND}(\(X\)) of strongly measurable Pettis integrable functions on \([0,1]\) into a Banach space \(X\), with nowhere weakly differentiable primitives (w.r.t. Lebesgue measure), is indeed lineable, i.e., \textbf{ND}(\(X\)) contains all non-zero elements of an infinite-dimensional vector space. A piece of history: Recall that the indefinite Bochner-integral is almost everywhere differentiable. \textit{B. J. Pettis} proved in [Trans. Am. Math. Soc. 44, 277--304 (1938; Zbl 0019.41603; JFM 64.0371.02)] that this is not so for the Pettis-integral of a strongly measurable function, but asked if such a Pettis-integral is weakly almost everywhere differentiable. The question was quite dramatically solved in 1995 when \textit{S. J. Dilworth} and \textit{M. Girardi} [Quaest. Math. 18, No. 4, 365--380 (1995; Zbl 0856.28006)] proved that \textbf{ND}(\(X\)) is always non-empty. The paper is very technical. It uses the ideas of Dilworth and Girardi [loc.\,cit.]\ together with some interesing applications of Dvoretzky's theorem and quite a bit of infinite combinatorics. In order to help the reader, the result is first shown for \(X=\ell_2\) (Section 3) and then in general (Section 4). The problem whether \textbf{ND}(\(X\)) is spaceable, i.e., contains all non-zero elements of an infinite-dimensional vector space closed in the norm of \(X\), is left open.