On the second-order regularity of solutions to the parabolic \(p\)-Laplace equation (Q2115100)

The weak solutions \(u = u(x,t)\) of the equation \[ \frac{\partial u}{\partial t}\,=\,\mathrm{div}\bigl(|\nabla u|^{p-2}\nabla u \bigr), \quad 1<p<\infty, \] in the domain \(\Omega_T = \Omega \times (0,T)\) belong by definition to the space \(L^p(0,T;W^{1,p}(\Omega))\), where \(\Omega\) is a bounded domain in \(\mathbb{R}^n\). They are continuous in the range \(p > 2n/(n+2)\), but when \(p < 2n/(n+2)\) some precaution is called for. The authors prove the following sharp result. Theorem. Suppose that \(u\) is a continuous weak solution in the domain \(\Omega_T\). Then the Sobolev derivative \[ \nabla \left(|\nabla u|^{\frac{p-2+s}{2}} \nabla u \right) \] exists and belongs to \(L^2_{\mathrm{loc}}(\Omega_T)\), whenever \(s> -1\). The theorem comes with a bound for the local \(L^2\)-norm. A simple counterexample shows that the result fails for \(s \leq -1\). This sharp theorem improves parabolic results by \textit{H. Dong} et al. [Adv. Math. 370, Article ID 107212, 39 p. (2020; Zbl 1442.35195)] and by \textit{A. Cianchi} and \textit{V. G. Maz'ya} [J. Geom. Anal. 30, No. 2, 1565--1583 (2020; Zbl 1439.35204)]. The special cases \(s = 0\) and \(s = p-2\) provide a new proof of the reviewer's theorem about the time derivative \(u_t\): it exists in Sobolev's sense and belongs locally to \(L^2\). The key point is the fundamental inequality \[ |\nabla u|^4|D^2u|^2\,\geq\, 2\,|\nabla u|^2|D^2u \nabla u|^2\,-\,\left(\sum\frac{\partial u}{\partial x_i}\frac{\partial u}{\partial x_j}\frac{\partial ^2 u}{\partial x_i \partial x_j}\right)^2 \] valid at least for smooth functions. An earlier variant of this inequality appeared in [\textit{H. Koch} et al., J. Math. Pures Appl. (9) 132, 457--482 (2019; Zbl 1435.35143)]. The proof of the theorem is based on the classical method. First, the equation is properly regularized. Then the regularized equation is differentiated and estimates are derived for the norm of the gradient. The use of the above inequality now requires some technically sophisticated arguments.