On the interior differentiability of weak solutions of parabolic systems with quadratic growth nonlinearities (Q1813282)

The author discusses the results of the title for the system \[ {\partial \over {\partial t}} u_ i (t)- {\partial \over {\partial x_ \alpha}} A_ i^ \alpha (x,t, \nabla u)= B_ i (x,t, \nabla u) \quad \text{in } Q, \qquad i=1,\dots, N, \] where \(A_ i^ \alpha\), \(B_ i\) are Carathéodory functions on \(\mathbb{Q}\times \mathbb{R}^ n \times \mathbb{R}^{nM}\) satisfying \(\forall M>0\), \(\exists c=c(M)> 0\) s.t. \[ | A^ \alpha_ i (x,t,u, \xi)|\leq c(1+ | \xi|), \qquad B_ i (x,t,u, \xi)|\leq c(1+| \xi|^ 2), \] \(\forall (x,t, u,\xi)\in \mathbb{Q} \times \mathbb{R}^ n \times \mathbb{R}^{nN}\) with \(| u| \leq M\). He defines the appropriate notion of weak solution. It is known that bounded weak solutions of elliptic systems need not have locally square integrable second derivatives. However, Hölder continuous weak solutions of strongly elliptic systems with quadratic growth nonlinearities have locally square integrable second derivatives. The author studies the parabolic case, noting first that any bounded weak solution is partially Hölder continuous (i.e., Hölder continuous in an open set \(Q_ 0\) with \(\text{meas} (Q\setminus Q_ 0) =0\)). He makes further assumptions on the coefficients \[ A_ i^ \alpha\in C^ 1 (\mathbb{Q} \times \mathbb{R}^ n \times \mathbb{R}^{nN}); \qquad B_ i, {{\partial B_ i} \over {\partial t}}, {{\partial B_ i} \over {\partial u_ j}}, {{\partial B_ i} \over {\partial \chi_ \beta^ j}} \in C(\mathbb{Q} \times \mathbb{R}^ n \times \mathbb{R}^{nN}); \] \[ {{\partial A_ i^ \alpha} \over {\partial \xi_ \beta^ j}} (x,t,u, \xi) \eta^ i_ \alpha \eta^ j_ \beta\geq \nu |\eta |^ 2 \qquad \forall (x,t,u, \xi)\in \mathbb{Q}\times \mathbb{R}^ N \times \mathbb{R}^{nN}, \qquad \forall \eta\in \mathbb{R}^{nN} \] and \(\nu\) is a positive constant. He also assumes that \(A^ \alpha_ i\) and its derivatives with respect to each of \(x_ \beta, t, u^ j\) grow not faster than the first power \((1+ |\xi |)\), that the derivative of \(A^ \alpha_ i\) with respect to \(\xi^ j_ \beta\) is bounded and that the derivatives of \(B_ i\) with respect to each of \(x_ \beta, t, u^ j\) grow not faster than the second power \((1+| \xi|^ 2)\) while its derivative with respect to \(\xi^ j_ \beta\) grows not faster than \((1+| \xi|)\) when \(| u|\leq M\). Under these hypotheses, he shows that a weak solution of the system which is Hölder continuous of order \(\gamma\) with \(1/2< \gamma<1\) has a gradient which is in \(L^ 4_{\text{loc}} (\mathbb{Q}; \mathbb{R}^{nN})\). He notes that standard methods then give the fact that \(u_{x_ \alpha x_ \beta}, u_ t\in L^ 2_{\text{loc}} (\mathbb{Q}; \mathbb{R}^{nN})\) and that \(\nabla u\) is partially Hölder continuous in \(\mathbb{Q}\).