Partial regularity for manifold constrained \(p(x)\)-harmonic maps (Q1725723)

The author considers local mimimizers of the functional \[ W^{1,p(\cdot)}(\Omega,\mathcal{M}) \ni w \mapsto \mathcal E(w,\Omega) := \int_\Omega k(x)\,|Dw|^{p(x)}\,dx, \] where \(p(\cdot)\) and \(k(\cdot)\) are Hölder-continuous functions satisfying the following conditions: (P1) \(p \in C^{0,\alpha}(\Omega)\), \(\alpha \in (0,1]\), (P2) \(1 < \gamma_1 \le p(x) \le \gamma_2 < \infty\) \text{ for all } \(x \in \Omega\), \(\gamma_1 := \underset{x\in\Omega}{\inf}, p(x)\) \text{ and } \(\gamma_2 := \underset{x\in\Omega}{\sup} p(x)\), (K1) \(k \in C^{0,\nu}(\Omega)\), \(\nu \in (0,1]\), (K2) \(0 < \lambda \le k(x) \le \Lambda < \infty\) \text{ for all } \(x \in \Omega\). Here \(\Omega \subset \mathbb{R}^n\), \(n \ge 2\), is a bounded open set and \(\mathcal{M} \subset \mathbb{R}^N\), \(N \ge 3\), is an \(m\)-dimensional compact submanifold endowed with a topology that satisfies the conditions: \((\mathcal{M}1)\) \(\mathcal{M}\) is a compact, \(m\)-dimensional, a \(C^3\) Riemannian submanifold without boundary of \(\mathbb{R}^N\), \((\mathcal{M}2)\) \(\mathcal{M}\) is \([\gamma_2]-1\) connected. \(W^{1,p(\cdot)} (\Omega,\mathcal{M})\) denotes the Musielak-Orlicz-Sobolev space of functions defined on \(\Omega\) into \(\mathcal{M}\), see, e.g. [\textit{V. V. Zhikov} et al., Homogenization of differential operators and integral functionals. Transl. from the Russian by G. A. Yosifian. Berlin: Springer-Verlag (1994; Zbl 0838.35001)]; as usual \([\cdot]\) denotes the integer part of \(x\). The author proves then that there exists a relatively open set \(\Omega_0 \subset \Omega\) of full \(n\)-dimensional Lebesgue measures such that \(u \in C_{\mathrm{loc}}^{1,\beta_0}(\Omega_0,\mathcal{M}\)) for some \(\beta_0 \in (0,1)\) and \(\Sigma_0(u) := \Omega\setminus \Omega_0\) has Hausdorff dimension at most equal to \(n - \gamma_1\)\,. Further, assuming extra conditions on the variable exponent \(p(\cdot)\) enables her to obtain an improved dimension estimate for the Hausdorff dimension of the singular set of \(\mathcal{M}\)-constrained minimizers of the \(p(\cdot)\)-energy, given by \[ w \mapsto \int_\Omega |Dw|^{p(x)}\,dx. \tag{0.2} \] The regularity theory of functionals has been studied, among several others, by Ragusa, Tachikawa and Usuba, see, e.g., [\textit{A. Tachikawa}, Calc. Var. Partial Differ. Equ. 50, No. 1--2, 145--169 (2014; Zbl 1305.49049)] and the references therein for a more recent presentation. The Tachikawa paper aforementioned generalizes several classical results, where \(p(\cdot)=\) constant, see, e.g. [\textit{M. Giaquinta} and \textit{E. Giusti}, Ann. Sc. Norm. Super. Pisa, Cl. Sci., IV. Ser. 11, 45--55 (1984; Zbl 0543.49018)]. The vectorial quasi convex case with standard \(p\)-growth, also in case \(p(\cdot)\) constant, was handled by \textit{C. P. Hopper} [Arch. Ration. Mech. Anal. 222, No. 1, 91--141 (2016; Zbl 1347.49055)]. Regularity results, for both scalar and vectorial minimizers, in the \(p\)-growth case, have recently been surveyed in [\textit{T. Kuusi} and \textit{G. Mingione}, Bull. Math. Sci. 4, No. 1, 1--82 (2014; Zbl 1315.35095)]. The main theorems proven in the paper, stated below, generalizes Tachikawa's approach, since it allows minimizers taking values into a manifold instead of taking values into a single chart. This non-trivial extension brings interesting new features to the subject. It is worth noticing that the functionals considered in the paper include as particular cases the functionals with \((p,q)\)-growth introduced in [\textit{P. Marcellini}, Arch. Ration. Mech. Anal. 105, No. 3, 267--284 (1989; Zbl 0667.49032)]. In particular the case of the so-called double phase energy introduced in [\textit{V. V. Zhikov}, Russ. J. Math. Phys. 3, No. 2, 249--269 (1995; Zbl 0910.49020)]. The main theorems of the paper, stated below, describe in part the ampleness of this new approach. The main results read: Theorem 1.\, Let \(u \in W^{1,p(\cdot)}(\Omega,\mathcal{M}) \) be a local minimizer of the functional in \((0,1)\), where \(p(\cdot)\) satisfies assumptions (P1)--(P2), \, \(k(\cdot)\) satisfies (K1)--(K2) and \(\mathcal{M}\) is as in \((\mathcal{M}1)\)--\((\mathcal{M}2)\). Then there exists a relatively open set \(\Omega_0 \subset \Omega\) such that \(u \in C_{\mathrm loc}^{1,\beta_0}(\Omega_0,\mathcal{M})\) for some \(\beta_0 \in (0,1)\) and \(\mathfrak{H}^{n-\gamma_1}(\Omega\backslash\Omega_0)=0\). By strengthening further the assumptions on the variable exponent \(p(\cdot)\), the author provides a better dimension estimate for the singular set. This is given in the following Theorem 2.\, Let \(u \in W^{1,p(\cdot)}(\Omega,\mathcal{M})\) be a constrained local minimizer of energy \((0.2)\), where \(p(\cdot) \in {\mathrm{Lip}}(\Omega)\), \(\gamma_1 \ge 2\) and \(\mathcal{M}\) is as in \((\mathcal{M}1)\)--\((\mathcal{M}2)\). Then, \begin{itemize} \item[i.] if \(n \le [\gamma_1]+1\), then \(u\) can have only isolated singularities; \item[ii.] if \(n > [\gamma_1]+1\), then the Hausdorff dimension of the singular set is at the most \(n-[\gamma_1]-1\). \end{itemize} The seven steps long proof of the first theorem is well-prepared by ten lemmas, some of which provide some extension to well-known results, see Lemmas 5--10, which additionally with further two Lemmas, 11 and 12, also prepare the proof for the second theorem. Despite the technical nature of the subject the clear author's writing style makes the paper for a good measure self-sufficient.