Minima of quasisuperminimizers (Q2357199)

Let \(p, \, Q \geq 1\) and \(\Omega \subset \mathbb{R}^n\) an open set. A \(Q\)-quasiminimizer \(u \in W^{1,p}_{\mathrm{loc}}(\Omega)\) is an almost minimizer of the \(p\)-Dirichlet integral in the sense that \[ \int_{\varphi \neq 0}|\nabla u|^p \,dx \leq Q \int_{\varphi \neq 0}|\nabla (u+ \varphi)|^p \,dx \text{ for all } \varphi \in W^{1,p}_{0}(\Omega).\eqno{(\ast)} \] Many results concerning the solutions of the \(p\)-Laplace equation, i.e., \(Q= 1\), have been generalized to quasiminimizers. Much less is known on \(Q\)-quasisuperminimizers which satisfy (*) for all \(\varphi \in W^{1,p}_{0}(\Omega)\), \(\varphi \geq 0\). In [\textit{J. Kinnunen} and \textit{O. Martio}, Ann. Acad. Sci. Fenn., Math. 28, No. 2, 459--490 (2003; Zbl 1035.31007)] it was shown that if \(u_i\) is a \(Q_i\)-quasisuperminimizer, \(i=1,2\), then \(\min(u_1,u_2)\) is a \(\min(Q_1Q_2, Q_1 + Q_2)\)-quasisuperminimizer. The authors improve this bound, present several examples, obtain lower bounds for \(Q\) and consider more functions than two. In particular, it is shown that \(\min(u_1,u_2)\) is a \(\tilde{Q}\)-quasisuperminimizer with \(\tilde{Q} = (Q_1+ Q_2-2)Q_1Q_2(Q_1Q_2 -1)^{-1}\) whenever \(\max(Q_1Q_2 ) > 1\). The lower bounds are constructed in the one-dimensional case where the structure of quasiminimizers is better understood than in higher-dimensional spaces, see [the reviewer and \textit{C. Sbordone}, Ann. Mat. Pura Appl. (4) 186, No. 4, 579--590 (2007; Zbl 1150.49016)]. It is also shown that if \(\Omega_1 \subset \Omega_2\) and \(u_i\) is a \(Q_i\)-quasisuperminimizer in \(\Omega_i\), then the function \(u = u_2\) in \(\Omega_2 \setminus \Omega_1\) and \(u = \min(u_1,u_2)\) in \(\Omega_1\) is a \(Q_1Q_2\)-quasisuperminimizer in \(\Omega_2\) provided that \(u \in W^{1,p}_{\mathrm{loc}}(\Omega_2)\) and the bound \(Q_1Q_2\) is optimal in this case. This a counterpart of the so-called pasting lemma well known for superharmonic functions. The results in the paper extend to metric measure spaces where \(|\nabla u|\) is replaced by the minimal upper gradient.