Vector-valued local minimizers of nonconvex variational problems (Q1176417)

A method due to \textit{R. V. Kohn} and the author [Proc. R. Soc. Edinb., Sect. A 111, No. 1/2, 69-84 (1989; Zbl 0676.49011)] for obtaining local minimizers of perturbed variational problems is used to establish the existence of non-constant vector-valued local minimizers \(u_{\varepsilon}: \Omega \to \mathbb{R}^ 2\) of \(H^ 1(\Omega)\), \(\Omega\subset \mathbb{R}^ n\) an open and bounded domain, for the problem \[ \inf F_{\varepsilon}(u)\hbox{ s.t. }u\in H^ 1(\Omega)\hbox{ with }F_{\varepsilon}(u):=\int_ \Omega {1 \over \varepsilon} W(u)+\varepsilon|\nabla u|^ 2 dx.\leqno(1) \] Here \(\partial\Omega\) is Lipschitzian, \(W\) is a nonnegative double-well potential vanishing at the two points \(a\) and \(b\) in \(\mathbb{R}^ 2\) and \(\varepsilon\in (0,\varepsilon_ 0)\) with sufficiently small \(\varepsilon_ 0\). Such variational problems arise in the gradient theory of phase transitions and in studies of pattern selection. Defining \(F_ 0(u)\) and using geodesics with respect to the potential \(W\) it is shown that \(F_ \varepsilon\to F_ 0\) in the sense of epi-convergence, that \(F_ \varepsilon(v_ \varepsilon)\leq C<+\infty\) implies compactness of \((v_ \varepsilon)_{\varepsilon > 0}\) in \(L^ 1(\Omega)\), and that (1) for \(\varepsilon =0\) has an isolated local minimizer under certain ``neck''-conditions on the domain \(\Omega\). The application of Theorem 2.1 of the above-mentioned paper immediately implies the existence of local minimizers \(u_ \varepsilon\) for (1) for all sufficiently small \(\varepsilon >0\). Since the Euler-Lagrange- equation of (1) is valid with the natural boundary condition \(\partial u_ \varepsilon/\partial n=0\) the solution \(u_ \varepsilon\) is non- constant. The reader should be familiar with another paper of the author [Arch. Ration. Mech. Anal. 101, No. 3, 209-260 (1988; Zbl 0647.49021)].