Convergence of a mass conserving Allen-Cahn equation whose Lagrange multiplier is nonlocal and local (Q399400)

The authors consider the solution \(u_\varepsilon= u_\varepsilon(t,x)\) of an Allen-Cahn equation with conservation of mass, namely, \[ \partial_t u_{\varepsilon}=\Delta u_\varepsilon+ {1\over \varepsilon^2} \Biggl( f(u_\varepsilon)- {\int_\Omega f(u_\varepsilon)\over \int_\Omega\sqrt{4W(u_\varepsilon)}} \sqrt{4W(u_\varepsilon)}\Biggr)\quad\text{in }\Omega\times(0, \infty),\tag{1} \] supplemented with the homogeneous Neumann boundary condition \[ {\partial u_\varepsilon\over \partial\nu} (x,t)= 0\quad\text{on }\partial\Omega\times (0,\infty)\tag{2} \] and the initial conditions \[ u_\varepsilon(x,0)= g_\varepsilon(x)\quad\text{in }\Omega.\tag{3} \] Here \(\Omega\) is a smooth bounded domain in \(\mathbb{R}^N(N\delta 2)\) and \(\nu\) is the Euclidean unit normal vector exterior to \(\partial\Omega\). The small parameter \(\varepsilon> 0\) is related to the thickness of a diffuse interfacial layer. Then the authors prove their first main theorem constructing an accurate approximate solution. Theorem 0.1 (Approximate solution) Let us fix an arbitrary integer \(k>\max\{N,4\}\). Then there exist \((u_{\varepsilon,k}(x,t), \lambda_{\varepsilon,k}(x, t))_{x\in\Omega, 0\leq t\leq T}\) such that \[ \begin{gathered}\partial_t u_{\varepsilon,k}-\Delta u_{\varepsilon,k}-{1\over \varepsilon^2} (f(u_{\varepsilon,k})- \lambda_{\varepsilon,k}(1- u^2_{\varepsilon,k}))= \delta_{\varepsilon,k}\quad\text{in }\Omega\times (0,T),\\ \|\delta_{\varepsilon, k}\|_{L^\infty(\Omega\times (0,T))}= o(\varepsilon^k)\quad\text{as }\varepsilon\to 0,\end{gathered} \] and \[ \begin{aligned} {\partial u_{\varepsilon, k}\over \partial\nu} (x,t)= 0\quad &\text{on }\partial\Omega\times (0,T),\\ {d\over dt} \int_\Omega u_{\varepsilon,k} dx= 0\quad &\text{for all }(0,T).\end{aligned} \] Moreover, the following estimate in the \(L^2\)-norm for the error between the approximate solution \(u_{\varepsilon,k}\) and the solution \(u_\varepsilon\) holds. Theorem 0.2 (Error estimate) Let us fix an arbitrary integer \(k>\max\{N,4\}\). Let \(u_\varepsilon\) be a solution of (1), (2), (3) with the initial condition satisfying \[ g_\varepsilon(x)= u_{\varepsilon,k}(x,0)+ \phi_\varepsilon(x)\in [0,1],\;\int_\Omega \phi_\varepsilon= 0,\;\|\phi_\varepsilon\|_{L^2(\Omega)}= 0(\varepsilon^{k-{1\over 2}}). \] Then there is \(C>0\) such that for \(\varepsilon>0\) small enough, \[ \| u_\varepsilon(\cdot, t)- u_{\varepsilon,k}(\cdot, t)\|_{L^\infty(\Omega)}\leq C\varepsilon^{k-{1\over 2}}. \]