Weak compactness and variational characterization of the convexity (Q1949857)

The paper is concerned with two problems raised by Biagio Ricceri (private communication). The first one is: Must a weakly lsc function \(f\) defined on a reflexive Banach space \(X\) with values in \( \mathbb R\cup\{+\infty\} \) such that \(f-\ell\) attains its global minimum for every \(\ell\in X^*\) be convex? The author gives positive solutions to this problem, showing also that the reflexivity hypothesis can be dropped. For instance, if \(f\) is such that \(f-\ell\) has a unique global minimum for every \(\ell \in X^*,\) then \(f\) is strictly convex (Theorem 3.3). The convexity of \(f\) follows also under the hypothesis that \(f-\ell\) has a global minimum (not necessarily unique) for every \(\ell \in X^*\) (Theorem 3.6). The author shows that, under a supplementary hypothesis, the reflexivity is also necessary: if there exists a weakly lsc function \(f\) on \(X,\) with the effective domain \(D_f\) having nonempty interior, such that \(f-\ell\) attains its global minimum for every \(\ell\in X^*,\) then \(X\) is reflexive (Corollary 2.5). The author shows also that a \(C^1\) weakly lsc function on a reflexive Banach space with one-to-one derivative \(f':X\to X^*\) is strictly convex, answering thus positively another question of Biagio Ricceri. As a final consequence, he obtains a version of a result of \textit{L. N. H. Bunt} [Contributions to the theory of convex point sets (Dutch). Dissertation. Amsterdam: Noordhollandsche Uitgeversmaatschappij (1934; Zbl 0008.07801, JFM 60.0673.02)] and \textit{T. Motzkin} [Atti Accad. Naz. Lincei, Rend., VI. Ser. 21, 562--567 (1935; Zbl 0011.41105, JFM 61.0758.01); ibid., 773--779 (1935; Zbl 0012.27109, JFM 61.0758.01)], namely, that every weakly sequentially closed Chebyshev subset of a Hilbert space is convex.