Some comparisons (equality, strong density or complete characterizations) are made between the Clarke and limiting subdifferentials for the class of integral functionals defined on reflexive Lebesgue spaces LpL_p, without any locally Lipschitzian assumptions on these functionals. Let IfI_f be the integral functional associated with a normal integrand f:Ω×ER{±}f:\Omega\times E\to {\mathbb{R}}\cup\{\pm\infty\}, where EE is a finite dimensional space and Ω\Omega is endowed with an atomless σ\sigma-finite measure. At a given point xx such that almost everywhere the Clarke subdifferential of fωf_{\omega} at x(ω)x(\omega) coincide with the closed convex hull of the limiting subdifferential and such that the limiting subdifferential multifunction at xx, ωLfω(x(ω))\omega\rightrightarrows \partial^{L}f_{\omega}(x(\omega)), admits a q=p(p1)1q=p(p-1)^{-1} integrable selection, we show that the limiting subdifferential of the integral functional is strongly dense in the Clarke subdifferential. This property allows us to give an exact computation of the Clarke subdifferential by a legitimate disintegration formula. As a consequence, when for almost every ω\omega in Ω\Omega the function fω(.)f_{\omega}(.) is locally Lipschitz at x(ω)x(\omega) then the limiting subdifferential of IfI_f at xx is strongly dense in the Clarke subdifferential of IfI_f at xx. When the Lipschitz rate is qq-integrable, there is the coincidence of the Clarke subdifferential with the limiting subdifferential of IfI_f at xx and the disintegration formula is valid. Moreover such a disintegration formula is valid for the calculus of the Clarke normal (respectively tangent) cone at a given point to the set of pp-integrable selections of a measurable multifunction.

Contact details are reproduced from the original publication and may be historical.

Emmanuel Giner

Laboratoire MIP, Université Paul Sabatier, 118 route de Narbonne, 31062 Toulouse cedex 04, France

giner@math.univ-toulouse.fr

E. Giner. “Clarke and Limiting Subdifferentials of Integral Functionals.” Journal of Convex Analysis 24 (2017), No. 2, 661–678.