Splitting in a degenerate quasilinear parabolic equation (Q923826)

In \(\Omega =\{x|\) \(0<x_ i<\ell_ i\), \(i=1,2\}\), with \(Q_ T=(0,T)\times \Omega\), the author considers the nonlinear partial differential equation \[ \frac{\partial \phi (u)}{\partial t}=\mu \Delta u,\quad \mu =const.>0,\quad u|_{t=0}=u_ 0;\quad u|_{\partial \Omega}=0 \] under the hypotheses \(\phi \in C^{\infty}\), \(\phi '(u)>0\), \(u\neq 0\), \(\phi '(0)=0\), \(and\) (\(\phi\) (a)-\(\phi\) (b))(a-b)\(\geq \eta (a-b)^ P\), \(\eta >0\) and \(p\geq 2\). \(\phi '(u)\) is even and \(\exists C^{\infty}\) function \(\upsilon_ 0\) which is periodic in each variable separately of period \(\ell_ i\) and extends \(u_ 0\) to all of \(R^ 2\). The author constructs a penalty method approximation by solving \[ \epsilon \frac{\partial u}{\partial t}+\phi '(u)\frac{\partial u}{\partial t}=\mu \Delta u \] and linearizes that equation further by \[ \epsilon \frac{\partial u}{\partial t}+\phi '(u(t-\theta))\frac{\partial}{\partial t}=\mu \Delta u \] with the same initial conditions as above. This latter problem is decomposed by the method of weak approximation into two one dimensional problems \[ \epsilon \frac{\partial u_{\tau}}{\partial t}+\phi '(u_{\tau}(t- \theta))\frac{\partial u_{\tau}}{\partial t}=2\mu \frac{\partial^ 2u_{\tau}}{\partial x^ 2_ 1},\quad m\tau <t\leq (m+\frac{1}{2})\tau, \] where \(m=0,1,...,N-1\), and \(N\tau =T\), and \[ \epsilon \frac{\partial u_{\tau}}{\partial t}+\phi '(u_{\tau}(t-\theta))\frac{\partial u_{\tau}}{\partial t}=3\mu \frac{\partial^ 2u_{\tau}}{\partial x^ 2_ 2},\quad (m+\frac{1}{2})\tau \leq T\leq (m+1)\tau. \] He proves that \[ \| u-u_{\tau}^{\epsilon,\theta}\|_{L_ p(Q_ T)}\leq c\epsilon^{\frac{1}{p- 1}}+c(\epsilon)\theta^{\frac{1}{2p}}+c(\epsilon,\theta)\tau, \] where the constants c(\(\epsilon\)), c(\(\epsilon\),\(\theta\)) depend only on \(\epsilon\), and \(\epsilon\),\(\theta\), and c is independent of the parameters \(\epsilon\),\(\theta\), and \(\tau\).