Long-time behavior of a gradient system governed by a quasiconvex function (Q2026725)

This is an excellent paper on the long time behavior of the solutions to the second order equation \[ u^{\prime \prime}(t)=\nabla \phi (u(t)), \ \ t\ge 0, \] where \(\phi\) is a differentiable quasiconvex function defined on a real Hilbert space \(H\) with a nonempty set of minimizers. More precisely, if in addition \(\nabla \phi\) is Lipschitz continuous on bounded sets and \(\Vert u(t) \Vert \) is bounded, then \(u(t)\) converges weakly as \(t\longrightarrow \infty\) to a point \(p\in (\nabla \phi )^{-1}(0)\), and if \(p\notin \mathrm{argmin}\, \phi\), then the convergence is strong; on the other hand, if \(\Vert u(t)\Vert\) is is unbounded, then \(\Vert u(t)\Vert \longrightarrow \infty\) as \(t\longrightarrow \infty\). Thus a previous analysis of the second author and the reviewer [J. Convex Anal. 26, No. 4, 1175--1186 (2019; Zbl 1435.34061)] is here successfully extended from the pseudoconvex case to the quasiconvex one. The discrete version of the above equation is also investigated in detail.