\documentstyle[12pt]{article} \setlength{\textwidth}{6in} %\setlength{\textheight}{8in} \renewcommand{\baselinestretch}{1.2} \newcounter{local} \newcounter{locallocal} \newcommand{\scl}{\stepcounter{local}} %\pagestyle{empty} \setcounter{local}{0} \setcounter{locallocal}{1} \setcounter{section}{0} \renewcommand{\theequation}{\arabic{section}.\arabic{local}} \newcommand{\g}{\nabla \times } \newcommand{\vh}{ {\bf H}} \newcommand{\wpo}{W_{p}^{2,1}(Q_{T})} \newcommand{\ca}{C^{\alpha, \frac{\alpha}{2}}(\bar{Q}_{T})} \newcommand{\caa}{C^{1+\alpha, \frac{1+\alpha}{2}}(\bar{Q}_{T})} \newcommand{\caaa}{C^{2+\alpha, 1+\frac{\alpha}{2}}(\bar{Q}_{T})} \newcommand{\cax}{C^{\alpha}(\bar{\Omega})} \newcommand{\caax}{C^{1+\alpha}(\bar{\Omega})} \newcommand{\caaax}{C^{2+\alpha}(\bar{\Omega})} \newcommand{\cao}{C^{\alpha, \frac{\alpha}{2}}({Q}_{T})} \newcommand{\caao}{C^{1+\alpha, \frac{1+\alpha}{2}}({Q}_{T})} \newcommand{\caaao}{C^{2+\alpha, 1+\frac{\alpha}{2}}({Q}_{T})} \newcommand{\z}{\varepsilon} \newcommand{\h}{\hspace{1cm}} \newcommand{\hh}{\hspace{2cm}} \newcommand{\hhh}{\hspace{3cm}} \newcommand{\lp}{L^{p}(Q_{T})} \newcommand{\by}{\begin{eqnarray}} \newcommand{\ey}{\end{eqnarray}} \newcommand{\bys}{\begin{eqnarray*}} \newcommand{\eys}{\end{eqnarray*}} \newcommand{\B}{ {\bf H}} \newcommand{\E}{ {\bf E}} \newcommand{\F}{ {\bf F}} \newcommand{\J}{ {\bf J} } \newcommand{\Ph}{ {\bf \Phi}} \newcommand{\Ps}{ {\bf \Psi}} \newcommand{\W}{{\bf W}} \newcommand{\y}{{\bf c}} \newcommand{\e}{{\bf e}} \newcommand{\Y}{{\bf Y}} \begin{document} \begin{center} \begin{bf} On a Singular Limit Problem for Nonlinear Maxwell's Equations \end{bf} \ \\ \vspace{.5cm} Hong-Ming Yin\\ Department of Mathematics, University of Notre Dame\\ Notre Dame, IN 46556. \end{center} \vspace{1cm} \vspace{3cm} {\bf Abstract}: In this paper we study the following nonlinear Maxwell's equations \\ $\z \E_{t}+\sigma(x,|\E|)\E= \g \vh +\F,\, \vh_{t}+\g \E=0$, where $\sigma(x,s)$ is a monotone graph of $s$. It is shown that the system has a unique weak solution. Moreover, the limit of the solution as $\z\rightarrow 0$ converges to the solution of quasi-stationary Maxwell's equations. \ \\ {\bf AMS(MOS)} Subject Classifications: 35K20, 35Q20. \ \\ {\bf Key Words and Phases}: Nonlinear Maxwell's Equations, Singular limit. \newpage \begin{center} {\bf 1. Introduction} \end{center} Let $\Omega$ be a bounded domain in $R^3$ and $Q_{T}=\Omega\times (0,T]$ for any fixed $T>0$. Let $\E$ and $\vh$ be the electric and magnetic fields, respectively, in $\Omega$ (here and thereafter a bold letter represents a vector in $R^3$). Let $\sigma$ be the electric conductivity in the field, which is assumed to be a function of $x$ and $|\E|$. Consider the following Maxwell's equations (see Landau-Lifschitz \cite{LL}): \setcounter{section}{1} \setcounter{local}{1} \begin{eqnarray} & & \z \E_{t}+\sigma(x, |\E|)\E=\g \vh + \F, \h (x,t)\in Q_{T},\\ & & \vh_{t}+\g\E= 0, \hhh (x,t)\in Q_{T},\scl\\ & & \nabla \times {\E}=0, \hhh (x,t)\in \partial \Omega\times (0,T],\scl \\ & & \E(x,0)=\E_0(x), {\bf H}(x,0)={\bf H}_{0}(x), \h x\in \Omega, \scl \end{eqnarray} where $\z$ is the dielectric parameter and other physical parameters are normalized. In some applications (\cite{DL,MM}), the electric conductivity, $\sigma$, strongly depends on the electric field $|\E|$, hence the electric current density. Particularly, the electric conductivity may act like a switch-like function in some electromagnetic fields. On the other hand, for many types of micron devices and other industrial problems (such as microwave heating \cite{MM,GY,Y3}) etc.) the experiment shows that the displacement current, $\z \E_{t}$, is often negligible since it is small in comparison of the eddy current, $\J=\sigma \E$. This motivates us to investigate the nonlinear problem (1.1)-(1.4) and the singular limit problem as $\z\rightarrow 0$. It is shown that there exists a unique global solution to (1.1)-(1.4). Moreover, the limit of the solution converges to the solution of the quasi-stationary system (i.e., the system (1.1)-(1.4) with $\z=0$ in (1.1)). This limit solution provides new existence result for the quasi-stationary system. Indeed, when $\z=0$, the system (1.1) becomes \by \sigma(x, |\E|)\E =\g \B +\F,\scl \ey Thus, one can solve Eq. (1.5) for $|\E|$ in terms of $|\g \vh|$ and known data, \[ |\E|=g(x,|\g \B|),\] where $g(x,s)$ is the inverse function of $\sigma(x,s)s$. It follows from (1.2) that $\B$ satisfies \by \vh_t+\g [ \rho(x,|\g \vh|)\g\B ]=0,\scl \ey where \[ \rho(x,|\g \vh||)=\frac{1}{\sigma(x, |\E|)}= \frac{1}{\sigma(x, g(x,|\g\vh|))}\] represents the electric resistivity in the field. The research on Maxwell's equations is of great interesting because of the important applications in plasma physics, semiconductor-superconductor modeling and other industrial problems (\cite{GY,H,LL,Y3} etc.). The study on the system (1.1)-(1.4) as well as the quasi-stationary form (1.6) received considerable attention recently. In \cite{MP}, the authors established the well-posedness for a quasi-stationary system, where a constitutive relation between the magnetic field $\vh$ and the magnetic induction ${\bf B}$ is assumed to be nonlinear. In \cite{Y1}, the author studied the regularity of weak solution to a linear system of (1.6) with minimal requirement on coefficients. There is a special interest when \[ \rho(|\g \vh|)=|\g \vh|^{p-2}, \h p>2.\] On one hand, if $\B$ is restricted in one direction (scalar field) then the evolution system (1.6) becomes the p-Laplacian which has been studied extensively (see \cite{D} and the references therein). On the other hand, in a recent work \cite{Y2} (also see \cite{AEW,BP} for the scalar case), it is shown that the limit of the solution to (1.6) as $p\rightarrow \infty$ is the unique solution to Bean's critical-state model in the superconductivity theory (\cite{B}). Thus, for large $p$ the system (1.6) provides a good approximation to Bean's model. More recently, the author of \cite{JO} studied the similar problem to this paper in a domain with a bounded complement in $R^3$. The conditions on $\sigma$ in \cite{JO} is quite different from ours here. Like many nonlinear problems, the major difficulty is how to pass the weak limit of an approximate solution for a nonlinear function $\sigma(x,s)$. This is done by employing a monotonicity argument (\cite{E}). The monotonicity of $\sigma(x, s)$ in $s$ is essential in the proof. In \S 2, we use the finite element method to establish the well-posedness of the system (1.1)-(1.4) for fixed $\z>0$. In \S 3, we show that the singular limit of the solution to (1.1)-(1.4) has a unique limit. Moreover, the limit solution solves the quasi-stationary Maxwell's equations. Some examples are also discussed in this section. %The monotonicity of $\sigma(x, s)$ in $s$ plays an essential role. %In \S 4, we study the regularity of the limit solution. \begin{center} { \bf 2. Existence and Uniqueness for fixed $\z>0$ } \end{center} Introduce some standard spaces (see \cite{DL,GR}). \bys & & H(div, \Omega)=\{ {\bf U} \in L^2(\Omega)^3, div {\bf U}\in L^2(\Omega)\};\\ & & H(curl,\Omega)=\{ {\bf U}\in L^2(\Omega)^3, curl {\bf U}\in L^2(\Omega)^3\}\, ;\\ & & H(div0, \Omega)=\{ {\bf U}\in H(div, \Omega): div {\bf U}=0 \mbox{ in $\Omega$} \},\\ & & H_0(curl, \Omega)=\{ {\bf U}\in H(curl, \Omega): {\bf N}\times {\bf U}=0\, \mbox{on $\partial \Omega$} \}, \eys where ${\bf N}$ is the exterior unit normal on $\partial \Omega$. Note that the trace of a function in $H(curl, \Omega)$ is well defined (see \cite{DL} for example). We shall assume the following conditions on $\sigma(x,s)$ and data $\E_0(x), {\bf H}_{0}(x)$ and ${\bf F}(x,t)$. \ \\ H(2.1): Let $\sigma(x, s)$ be measurable in $\Omega\times [0, \infty)$ and monotone increasing in $s$. Moreover, \bys & & \int_{0}^{s^{2} } \sigma(x,\sqrt{s}) dx\geq a_0 s^{p+2}-a_1, \mbox{ if $s$ is sufficiently large},\\ & & 0\leq \sigma(x,s)\leq b_0 (1+s^{p}), s\in [0,\infty),\mbox{for some $p\geq 0$}, \eys where the constants $a_0>0, a_1\geq 0$ and $b_0\geq 0$. \ \\ H(2.2): Assume that $ {\bf H}_{0}\in H(curl, \Omega)\bigcap H(div0,\Omega), {\bf F}\in H^1(0,T; H^1(\Omega))$. \ \\ {\bf Definition 2.1}: {\em A pair of vector fields ($\E(x,t), \B(x,t))$ is said to be a weak solution of the problem (1.1)-(1.4), if \[ \E\in L^{2}(0,T; H_0(curl, \Omega))\bigcap L^{p+2}(0,T;\Omega), {\bf H}\in L^2(0,T;H(div0,\Omega) )\bigcap H(0,T, L^2(\Omega))\] which satisfy the following integral identities: \setcounter{section}{2} \setcounter{local}{1} \by & & \int \int_{Q_{T}}\left[-\z \E\cdot {\bf \Phi}_{t}+ \sigma(x,|\E|)\E\cdot \Ph \right] dxdt\nonumber\\ & & = \int\int_{Q_{T}}{\bf H}\cdot (\g \Ph)dxdt +\z \int_{\Omega} \E_0\cdot \Ph(x,0)dx,\\ & & \int\int_{Q_{T}}\left[- \B \cdot \Ps_{t} + \E \cdot (\g \Ps)\right] dxdt= \int_{\Omega} \left[ {\bf H}_{0}(x) \cdot \Ps(x,0)\right]dx \scl \ey for all test functions $\Ph\in H^1(0,T; H_0(curl, \Omega)), \Ps\in H^1(0,T; H(curl, \Omega))$ with \\$\Ph(x,T)=\Ps(x,T)=0$.} First of all, we derive some energy estimates. A special attention is paid on how various constants depend on $ \z$ since we will study the singular limit problem in section 3. \ \\ {\bf Lemma 2.1}: {\em Under the assumptions H(2.1)-H(2.2) there exist constants $C_1, C_2$ and $C_3$ such that \setcounter{local}{2} \bys & &\sup_{[0,T]}\int_{\Omega}\left[\z |\E|^2+ |{\bf H}|^2 +|\E|^{p+2}\right] dx + \int_{0}^{T}\int_{\Omega} \left[\z |\E_{t}|^2+|\vh_{t}|^2\right]dxdt \nonumber\\ & & \leq C_1\int_{\Omega} \left[|\E_0|^2+ |{\bf H}_0|^2+ |\g \B_0|^2 \right]dx +C_2\int\int_{Q_{T}}[|\F|^2+|\F_t|^2] dxdt+ C_3, \eys where $C_1, C_2$ and $C_3$ depend only on known data.}\\ {\bf Proof}: Note that for any vector fields ${\bf A}, {\bf B}\in H(curl, \Omega)$ with either ${\bf A}$ or ${\bf B}$ in $H_0(curl, \Omega)$, the following identity holds: \[ \int _{\Omega} {\bf A}\cdot ({\bf \g B}) dx= \int _{\Omega} {\bf B}\cdot ({\bf \g A}) dx.\] Taking inner product to the system (1.1) and (1.2) by ${\bf E}$ and $\B$, respectively, we add up the resulting equations to obtain \by & & \sup_{[0,T]}\int_{\Omega}\left[ \z |\E|^2+|{\bf H}|^2\right] dx + \int\int_{Q_{T}}\sigma(x, |\E|) |\E|^{2}dxdt\nonumber \\ & & \leq C\int_{\Omega}\left[|\E_0|^2+{\bf H}_{0}|^2\right]dx +\int\int_{Q_{T}}[|\E\cdot\F|]dxdt\scl \ey where the constant $C$ depends only on known data, but not on $\z$. We first assume that $\sigma(x,s)$ is differentiable with respect to $s$. Then we formally differentiate Eq.(1.1) and Eq.(1.2) with respect to $t$ to obtain \bys & &\z \E_{tt}+\sigma(x,|\E|)\E_t+\sigma_{s}(x,|\E|)(|\E|)_{t}\E=\g \B_t+\F_t,\\ & & \B_{tt}+\g\E_t=0. \eys It is clear that \[ \int\int_{Q_{T}}(\g \E_t)\cdot \B_t dxdt= \int\int_{Q_{T}} (\g\B_t)\cdot \E_tdxdt.\] We take the inner product by $\E_t$ for the first equation and by $\B_t$ for the second equation and add up the resulting equations to obtain: \bys & & \sup_{[0,T]}\int_{\Omega}\left[ \z |\E_t|^2+|\B_t|^2\right]dxdt+ \int\int_{Q_{T}}\left[\sigma|\E_{t}|^2+\sigma_s (|\E|)_{t}\E\cdot \E_{t}\right] dxdt\\ & & \leq C, \eys where $C$ depends only on known data. Note that $\sigma_{s}\geq 0$, we see that \bys & & \int\int_{Q_{T}}\sigma_{s}(|\E|)_{t}\E\cdot \E_{t} dxdt\\ & & =\int\int_{Q_{T}}\sigma_{s}(|\E|)_{t} \frac{d}{dt}\frac{|\E|^{2}}{2}dxdt\\ & & =\int\int_{Q_{T}}\sigma_{s}|\E|[ |\E|_{t}]^2dxdt\geq 0. \eys It follows that \bys & & \sup_{[0,T]}\int_{\Omega}\left[ \z |\E_t|^2+|\B_t|^2\right]dxdt+ \int\int_{Q_{T}}[\sigma|\E_{t}|^2 dxdt \leq C. \eys Since the above estimate does not depend on the differentiability of $\sigma$ with respect to $s$, therefore the above estimate holds as long as $\sigma$ is monotone increasing with respect to $s$. Now we take the inner product by $\E_t$ to (1.1) and by $\B_t$ to (1.2) and then add up the resulting equations to obtain \bys & & \int\int_{Q_{T}}\left[ \z |\E_{t}|^2+|\vh_{t}|^2\right]dxdt+ \frac{1}{2}\int_{\Omega}\int_{0}^{|\E(x,T)|^{2}}\sigma(x, \sqrt{s})dsdx\\ & &\leq \int\int_{Q_{T}}\E_{t}\cdot \F dxdt +C\int_{\Omega}\int_{0}^{|\E_{0}(x)|^{2}}\sigma(x, \sqrt{s})dsdx+\int_{\Omega} |\g \B_0|^2dx+C. \eys Now \bys & & \int\int_{Q_{T}}\E_{t}\cdot \F dxdt\\ & & =\int_{\Omega} \left[\E(x,T)\cdot \F(x,T)-\E_{0}(x)\cdot\F(x,0)\right]dx -\int\int_{Q_{T}}\E\cdot \F_{t}dxdt\\ & & \leq \int_{\Omega}\left[\frac{a_{0}}{4}\E(x,T)|^{p+2}+\frac{16}{a_{0}} |\F_{t}|^{\frac{p+2}{p+1}}\right] dxdt. \eys On the other hand, by the assumption H(2.1) we may assume that the growth condition of $\sigma(x,s)$ on $s$ holds for all $s\geq M_0$, i.e., \[ \int_{0}^{s^{2} } \sigma(x,\sqrt{s}) dx\geq a_0 s^{p+2}-a_1, \mbox{ if $s\geq M_0$},\] where $M_0$ is a fixed constant. It follows that \by & & \int_{\Omega}\int_{0}^{|\E(x,T)|^{2}}\sigma(x,s)dsdx\nonumber\\ & & \geq a_0 \int_{\Omega \bigcap \{x:|\E(x,T)|\geq M_{0}\}} |\E(x,T)|^{p+2}dx -C.\scl \ey Combining (2.3)-(2.4) yields \bys & & \sup_{0\leq t\leq T}\int_{\Omega}[\z |\E|^2+|\vh|^2]dx+\sup_{0\leq t\leq T} \int_{\Omega}|\E(x,t)|^{p+2}dx+ \int_{0}^{T}\int_{\Omega}[\z |\E_t|^2+|\vh_t|^2]dxdt\\ & & \leq C\int\int_{Q_{T}} [|\F|^{2}+|\F_{t}|^{2} ]dxdt+\int_{\Omega}[|\E_0|^2+|\B_0|^2+|\g\B_0|^2]dx+C. \eys \hfill Q.E.D. \ \\ {\bf Theorem 2.2}: {\em Under the assumptions H(2.1)-H(2.2) the problem (1.1)-(1.4) has a unique weak solution. Moreover, \[ curl \E\in L^2(Q_{T}), \E_t\in L^2(Q_{T})\] and} \[ {\bf H}_{t}\in L^{2}(Q_{T}), \g\B\in L^{\frac{p+2}{p+1}}(Q_{T}).\] {\bf Proof}: The proof is based on the finite element method (see \cite{LSU} for parabolic equations). The monotonicity of $\sigma(x, s)$ on $s$ plays an important role. We shall first deal with the case where $\sigma(x, s)$ is continuous on $s$. For convenience, we rewrite the system (1.1)-(1.4) to the following form: \by & &\z \W_{tt}+\sigma(x,|\W_t|)\W_{t}=\g[ \B_0-\g \W] + \F, \h (x,t)\in Q_{T}, \scl \\ & &{\bf N}\times (\W_{t})=0, \hhh x\in \partial \Omega, 0\leq t\leq T, \scl \\ & & \W(x,0)=0, \W_t(x,0)=\E_0(x), \hh x\in \Omega,\scl \ey where $\W$ is defined as follows: \[ \W(x,t)=\int_{0}^{t}\E(x,\tau)d\tau.\] It is clear that if $\W$ is a solution of the system (2.5)-(2.7) then a pair of functions defined by \[ \E(x,t)=\W_t(x,t), \B(x,t)=\B_0(x)-\g \W(x,t)\] will be a weak solution of (1.1)-(1.4). Let $\{ {\bf e}_{k}\}=\{ e_{k}^{(1)}, e_{k}^{(2)}, e_{k}^{(3)}\} $ be a smooth basis of $H_0(curl, \Omega)$ and orthonormal in $L^2(\Omega)^3$, i.e. \[ < {\bf e}_{i}, {\bf e}_{j}>=\delta_{ij},\] where $\delta_{ij}=1$ if $i=j$ and $\delta_{ij}=0$ if $i\neq j$. Now we expand the known data as follows: \bys & & \B_0(x)=\sum_{k=1}^{\infty}diag[{\bf a}_{k}\circ {\bf e}_{k}],\\ & & \E_{0}(x)=\sum_{k=1}^{\infty}diag[ {\bf b}_{k}\circ {\bf e}_{k}],\\ & & \F(x,t)=\sum_{k=1}^{\infty}diag[{\bf g}_{k}(t)\circ {\bf e}_{k}], \eys where ${\bf a}_{k}, {\bf b}_{k}$ and ${\bf g}_{k}$ are $3\times 1$ matrices, the symbol $\circ$ is the matrix product and $diag[ ]$ represents the diagonal vector of a matrix. Let \[ \W_n(x,t)=\sum_{k=1}^{n}diag[ \y_{n}^{(k)}\circ {\bf e}_{k}],\] where $\y_{n}^{(k)}(t)$ is a $3\times 1$ vector which is determined by the following ordinary differential system: \by & & \z \frac{d^{2}}{dt^{2}}\y_{n}^{(k)}+ \sigma(x, |\W_{nt}|) \frac{d}{dt} \y_{n}^{(k)} =A_{k}(\W_{n}^{(k)}, {\bf e}_{k})+B_{k}(t),\scl \\ & & \y_{n}^{(k)})(0)=0,\scl \\ & & \frac{d}{dt}\y_{n}^{(k)}(0)={\bf b}_{k}, \scl \ey where \bys & & \W_{n}^{(k)}=diag[ \y_{n}^{(k)}\circ {\bf e}_{k}],\\ & & A_{k}(\W_{n}^{(k)}, {\bf e}_{k})=\int_{\Omega}diag\{ [\g \W_{n}^{(k)}] \circ [\g {\bf e}_{k}]\}dx,\\ & & B_{k}(t) =\int_{\Omega}diag\{ (\g \B_0+\F)\circ {\bf e}_{k}\} dx, \h k=1,2,\cdots n. \eys Now we define the approximate solution $(\E_n, \B_{n})$ as follows: \[ \E_{n}(x,t)=\W_{n t}(x,t), \B_n(x,t)=\B_{0n}-\g \W_{n}(x,t),\] where \[ \B_{0n}(x)=\sum_{k=1}^{n}diag[{\bf a}_{k}\circ\e_{k}].\] Equivalently, then $(\E_n, \B_n)$ satisfies the following system in the weak sense: \by & & \z \E_{n t}+\sigma(x,|\E_{n}|)\E_{n}=\g \B_{n}, \h (x,t)\in Q_{T},\scl \\ & & \B_{n t}+\g \E_{n}=0, \hh (x,t)\in Q_{T}.\scl \ey Similar to Lemma 2.1, one can easily derive the following energy estimates : \bys & &\sup_{[0,T]}\int_{\Omega}\left[\z |\E_n|^2+ |{\bf H}_n|^2 +|\E_n|^{p+2}\right]dx+ \int_{0}^{T}\int_{\Omega} \left[\z |\E_{nt}|^2+|\vh_{nt}|^2\right]dxdt \nonumber\\ & & \leq C_1\int_{\Omega} \left[|\E_{0n}|^2+ |{\bf H}_{0n}|^2+ |\g \B_{0n}|^2 \right]dx +C_2\int\int_{Q_{T}}[|\F_n|^2+|\F_{n t}|^2] dxdt+ C_3, \eys where $C_1,C_2$ and $C_3$ are independent of $n$ and $\z$. By the weak compactness property, we can extract a subsequence (still denoted by $(\E_{n}, \vh_{n}))$ such that \bys & & \E_{n}\rightarrow \E, \E_{n t}\rightarrow \E_t, \B_{nt} \rightarrow \B_t, \, \mbox{weakly in $L^2(Q_{T})$,}\\ & & \vh_{n}\rightarrow \vh, \mbox{ weakly in $L^2(0,T; W^{1,\frac{p+2}{p+1}} (\Omega))$},\\ & & \vh_n \rightarrow \vh, \mbox{ a.e. in $Q_{T}$}. \eys Moreover, \bys \E_{n}\rightarrow \E \h \mbox{ weakly in $L^{p+2}(Q_{T})$}. \eys Next we claim that the sequence $\E_{n}$ converges to $\E$ strongly in $L^2(Q_{T})^3$. To prove the claim we only need to show that $\{\E_{n}\}$ is a Cauchy sequence in $L^2(Q_{T})^3$. Let \[ {\E}_{n}^*(x,t)= \E_{n}(x,t)-\E_{m}(x,t), \vh_{n}^*(x,t)=\vh_{n}(x,t)-\vh_{m}(x,t).\] By energy estimates, we see \bys & & \sup_{0\leq t\leq T}\int_{\Omega}[ {|\E}_{n}^{*}|^2+|\vh_{n}^{*}|^2]dx +\\ & & \int\int_{Q_{T}}\left\{[ \sigma(x, |\E_{n}|)\E_{n}-\sigma(x, |\E_{m}|)\E_{m}] \cdot [\E_{n}-\E_{m}]\right\} dxdt\\ & & \leq C\int_{\Omega}[ |\E_{0n}-\E_{0m}|^2+|\vh_{0n}-\vh_{0m}|^2]dx+ C \int\int_{Q_{T}}[ |\F_n-\F_m|^2 ]dxdt, \eys where $C$ is a constant independent of $n$ and $m$. Note that $\sigma(x,s)$ is monotonic increasing in $s$, then \bys & & [\sigma(x, |\E_{n}|)\E_{n}-\sigma(x, |\E_{m}|)\E_{m}] \cdot [\E_{n}-\E_{m}]\\ & & \geq \frac{ \sigma(x, |\E_{n}|)-\sigma(x, |\E_{m}|)][|\E_{n}|^2- |\E_m|^2]}{ 2}\geq 0. \eys It follows that \bys & & \sup_{0\leq t\leq T}\int_{\Omega}[ |{\E}_{n}^{*}|^2+|\vh_{n}^{*}|^2dx\\ & & \leq C\int_{\Omega}[ |\E_{0n}-\E_{0m}|^2+|\vh_{0n}-\vh_{0m}|^2] dx+ C\int\int_{Q_{T}}|\F_n-\F_m|^2dxdt. \eys This implies that $ \E_{n}, \vh_{n}$ are Cauchy sequences since both $\E_{0n}, \vh_{0n}$ and $\F_n$ are Cauchy sequences in $L^2(Q_{T})^3$. Hence, \[ \E_n, \B_n\rightarrow \E, \B \, \mbox{strongly in $L^2(Q_{T})$}.\] After taking a subsequence if necessary, we see that \[ \E_{n}\rightarrow \E, \h a.e. \,in \, Q_{T}.\] To show the existence of a weak solution to (1.1)-(1.4), we only need to show \[ \sigma(x, |\E_n|)\E_n\rightarrow \sigma(x, |\E|) \E \h \mbox{ in $L^1(Q_{T})$}.\] As $\sigma(x,s)$ is continuous on $s$ and $\E_{n}$ converges to $\E$ almost everywhere in $Q_{T}$, we know \[ \sigma(x, |\E_n|)\E_n\rightarrow \sigma(x, |\E|) \E \h a.e. \mbox{ in $Q_{T}$.} \] We now show that $\sigma(x, |\E_n|)\E_n$ is equip-integrable in $Q_{T}$. We adopt a technique used for scalar elliptic and parabolic equations. Let $A$ be any measurable subset of $Q_{T}$. For any large $m>0$, \bys & & \int\int_{A}\sigma(x, |\E_{n})|\E_{n}|dxdt\\ & & \leq \int\int_{A\bigcap \{|\E_{n}|\leq m\}}\sigma(x, |\E_{n}|)|\E_n|dxdt+ \int\int_{A\bigcap \{|\E_{n}|\geq m\}} \sigma(x, |\E_{n}|)|\E_n|dxdt\\ & & \equiv I_1+ I_2. \eys The assumption on $\sigma(x,s)$ yields \bys & & I_1\leq C\int\int_{A\bigcap \{|\E_{n}|\leq m\}}[1+|\E_{n}|^{p} |\E_n|]dxdt, \eys which can be arbitrarily small if $|A|$ is small since $\E_{n}\in L^{p+2}(Q_{T})$. On the other hand, \bys & & I_2 \leq \frac{1}{m} \int\int_{A\bigcap \{|\E_{n}|\geq m\}} \sigma(x, |\E_{n}|)|\E_{n}|^2dxdt\leq \frac{C}{m}, \eys which is also small if $ m$ is sufficiently large. This concludes that $\sigma(x, |\E_n|) \E_{n}$ is equip-integrable in $Q_{T}$. It follows by Vitali's theorem that \[ \sigma(x, |\E_n|) \E_{n}\rightarrow \sigma(x, |\E|)\E \h \mbox{ in $L^1(Q_{T})$}.\] Finally, we show that $(\E,\B)$ is a weak solution of (1.1)-(1.4). By multiplying Eq.(2.11) and Eq.(2.12) by test functions $\Ps$ and $\Ph$, respectively, and then taking integration over $Q_{T}$, after some routine calculations and taking the limit, we see that $(\E,\B)$ is a weak solution to the system (1.1)-(1.4). Now we consider the case where $\sigma(x,s)$ is discontinuous on $s$ at some points. Without loss of generality, we may assume that $\sigma(x,s)$ has a jump only at one point $s=1$. In this case $\sigma(x,s)$ is not uniquely defined at $s=1$. We shall understand the value of $\sigma(x,1)$ in the following sense: \[ \sigma(x,1)\in [\sigma(x,1-),\sigma(x,1+)],\] where $\sigma(x, 1\pm)$ represents the right or left limit as $s\rightarrow 1$. By the standard approximation, we can construct a smooth approximation sequence $\sigma_m(x,s)$ such that \bys (i) & & \sigma_m(x,s)\, \mbox{ is monotonic increasing for all $s\geq 0$},\\ (ii) & & \sigma_m(x,s)=\sigma(x,s), \mbox{ if $|s-1|\geq \frac{1}{m}$}. \eys Let $(\E_m,\vh_m)$ be a solution of (1.1)-(1.4) in which $\sigma(x,s)$ is replaced by $\sigma_m(x,s)$. By the same energy estimate we see that there exists a measurable function $\beta(x,t)\in L^{\frac{p+2}{p+1}}(Q_{T})$ such that \[ \sigma_m(x,|\E_m|)\rightarrow \beta(x,t), \mbox{weakly in $L^{\frac{p+2}{p+1}}(Q_{T})$}.\] Define \bys & & A_m=\{ (x,t): 1-\frac{1}{m}\leq |\E|\leq 1+\frac{1}{m}\},\\ & & A=\{ (x,t): |\E(x,t)|=1\}. \eys Since $\sigma(x,s)$ is continuous except at $s=1$, we see \[ \beta(x,t)=\sigma(x, |\E|)\E, \mbox{if $(x,t)\in Q_{T}\backslash A$}.\] Now it is clear that \[ A=\bigcap_{m=1}^{\infty} A_m.\] Recall that $\sigma_m(x,s)=\sigma(x,s)$ if $|s-1|\geq \frac{1}{m}$. It follows that for all $(x,t)\in A_m$ \[ \sigma(x, 1-\frac{1}{m})\leq \sigma_m(x,|\E|)\leq \sigma(x,1+\frac{1}{m}).\] Consequently, as $m\rightarrow 0$, \[ \sigma(x, 1-)\leq \beta(x,t)\leq \sigma(x,1+), (x,t)\in A.\] Thus, $(\E, \vh)$ is a weak solution of (1.1)-(1.4). Finally, we show the uniqueness. Suppose $(\E, \vh)$ and $(\E^*, \vh^*)$ are two solutions of (1.1)-(1.4). Let \[ \hat{\E}=\E-\E^*, \hat{\vh}=\vh-\vh^*.\] Similar to the calculation in deriving energy estimates, we find \bys && \sup_{0\leq t\leq T}\int_{\Omega} [|\hat{\E}|^2+|\hat{\vh}|^2] dx+\int\int_{Q_{T}}\left[ \sigma(x, |\E|)\E-\sigma(x, |\E^*|)\E^*\right]\cdot \left[\E-\E^*\right] dxdt\\ & & \leq 0. \eys The monotonicity of $\sigma(x, s)$ implies that the second term in the above inequality is nonnegative. It follows that \[ \sup_{0\leq t\leq T}\int_{\Omega} [|\hat{\E}|^2+|\hat{\vh}|^2] dx\leq 0.\] Therefore, the uniqueness follows immediately. \hfill Q.E.D. \begin{center} {\bf 3. Singular Limit Problem} \end{center} In this section we shall show that the solution of (1.1)-(1.4) has a limit as $\z \rightarrow 0$, which solves Maxwell's equations in quasi-stationary fields, i.e. the system (1.1)-(1.4) with $\z=0$. A weak solution of the quasi-stationary system is defined as in Definition 2.1 with $\z=0$. From now on we denote by $(\E_{\z}, \B_{\z})$ the weak solution of the system (1.1)-(1.4). \ \\ {\bf Theorem 3.1}: {\em The limit of $(\E_{\z}, \B_{\z})$ as $\z\rightarrow 0$ solves the quasi-stationary system (1.1)-(1.4) with $\z=0$ in the weak sense. Moreover, the weak solution is unique if $\sigma(x,s)>0$ for all $(x,s)\in \Omega\times R^+$.}\\ {\bf Proof}: The crucial step in proving the convergence is to show \[ \sigma(|\E_{\z}|)\E_{\z}\rightarrow \sigma(|\E|)\E,\, a.e.\, \mbox{in $Q_{T}$ as $\z\rightarrow 0$}.\] The monotonicity of $\sigma(x,s)$ in $s$ plays a key rule. First of all, from Lemma 2.1 and the weak compactness we see \bys & & \E_{\z}\rightarrow \E,\, \B_{\z t}\rightarrow \B_{t}, \mbox{weakly in $L^2(Q_{T})$},\\ & & \g \B_{\z} \rightarrow \g \B,\, \mbox{weakly in $L^{\frac{p+2}{p+1}}(Q_{T})$},\\ & & \sigma(x,|\E_{\z}|)\E_{\z}\rightarrow {\bf J}(x,t), \h \mbox{ weakly in $L^{ \frac{p+2}{p+1}}(Q_{T})$}, \eys where ${\bf J}(x,t)\in L^{\frac{p+2}{p+1}}(Q_{T})$. Moreover, as $div \B_{\z}(x,t)=0$, by the decomposition property of $H^1(\Omega)$ property, after extracting a subsequence if necessary we see that \[ \B_{\z}\rightarrow \B,\, \mbox{strongly in $L^2(Q_{T})$ }\] and \[ \B_{\z}\rightarrow \B, \, a.e.\, \mbox{ in $Q_{T}$}.\] Next we show \[ {\bf J}(x,t)=\sigma(x,|\E|)\E, \h a.e. \mbox{in} \, Q_{T}.\] We use a monotonicity argument. As a first step, we show \[ \lim_{\z\rightarrow 0}\int\int_{Q_{T}}\sigma(x,|\E_{\z}|)|\E_{\z}|^2dxdt =\int\int_{Q_{T}}\J\cdot \E dxdt.\] Here we adopt an idea from \cite{JO}. Let $\lambda(t)$ be a nonnegative smooth function and \[ \lambda'(t)\leq 0, \lambda(0)=1, \lambda(T)=0.\] Define an operator ${\cal}{L}$ in $L^{p+2}(Q_{T})^3$ as follows: \[ {\cal}{L} [\E]=\sigma(x,|\E|)\E .\] Since $\sigma(x, s)$ is monotonic increasing in $s$, then the operator ${\cal}{L}$ is monotonic increasing, that is, \bys & & <{\cal}{L}[\E_{\z}]-{\cal}{L}[\E], \E_{\z}-\E>\geq 0. \eys It is clear that \bys & & <{\cal}{L}[\E_{\z}]-{\cal}{L}[\E], \E_{\z}-\E>\\ & & =< {\cal}{L}[\E_{\z}], \E_{\z}>-<{\cal}{L}[\E_{\z}], \E>-<{\cal}{L} [\E],\E_{\z}>+<{\cal}{L}[\E],\E>\\ \eys It follows that \setcounter{section}{3} \setcounter{local}{1} \by & & \lim_{\z\rightarrow 0}inf <{\cal}{L}[\E_{\z}], \E_{\z}> \geq <{\bf J}, \E>. \ey % \by % & & \lim_{\z\rightarrow 0}inf <\lambda(t){\cal}{L}[\E_{\z}, \E_{\z}> % \geq <\lambda(t){\bf J}, % \E>. % \eys On the other hand, from the system (1.1)-(1.2) we have \by & & \int_{0}^{T}\int_{\Omega} \sigma(\E_{\z})|\E_{\z}|^2\lambda(t)dxdt\nonumber \\ & & =- \int_{0}^{T}\int_{\Omega}\left[ \z \lambda(t)\E_{\z t}\cdot \E_{\z}+\lambda \vh_{\z t}\cdot \vh_{\z}\right]dxdt +\int\int_{Q_{T}}\F \cdot \E_{\z}dxdt\nonumber\\ & & =-\int_{0}^{T}\int_{\Omega} \left\{ \frac{\partial}{\partial t}\left[ \frac{1}{2}(\z |\E_{\z}|^2+|\vh_{\z}|^2)\lambda(t)\right]-\lambda'(t)\left[ \frac{1}{2}(\z |\E_{\z}|^2+|\vh_{\z}|^2) \right]\right\}dxdt+\nonumber\\ & & \int\int_{Q_{T}}\F \cdot \E_{\z}dxdt \nonumber \\ & & \leq \int_{0}^{T}\int_{\Omega} \left\{ \lambda'(t)\left[ \frac{1}{2}|\vh_{\z}|^2 \right]\right\}dxdt+\frac{1}{2} \int_{\Omega}\left[ \z |\E_{\z}(x,0)|^2+|\vh_{\z}(x,0)|^2\right] dx+\nonumber\\ & & \int\int_{Q_{T}}\F \cdot \E_{\z}dxdt.\scl % & & \leq \frac{1}{2}\int_{0}^{T}\int_{\Omega} % \lambda'(t)|\vh|^2 % dxdt+\frac{1}{2} \int_{\Omega}|\vh_{0}(x)|^2dx+\int\int_{Q_{T}}\F \cdot % \E dxdt,\scl \ey % as $\z\rightarrow 0$. Since $\lambda'(t)\leq 0$, it follows that \by & & \lim_{\z\rightarrow 0}sup \int_{0}^{T}\int_{\Omega} \sigma(|\E_{\z}|)|\E_{\z}|^2 \lambda'(t)dxdt\nonumber\\ & & \leq \frac{1}{2}\int_{0}^{T}\int_{\Omega} \lambda(t)|\vh|^2 dxdt+\frac{1}{2} \int_{\Omega}|\vh_{0}(x)|^2dx+\int\int_{Q_{T}} \E\cdot \F dxdt.\scl \ey Recall from Definition 2.1 that $(\E_\z, \B_{\z})$ satisfies the following integral equations: \bys & & \int_{0}^{T} \int_{\Omega}\left[-\z \E_{\z}\cdot {\bf \Phi}_{t}+ \sigma(x,|\E_{\z}|)\E_{\z} \cdot \Ph \right] dxdt\\ & & = \int_{0}^{T}\int_{\Omega}{\bf H_{\z}}\cdot (\g \Ph)dxdt +\int\int_{Q_{T}}\F \cdot \Ph_{\z}dxdt+\int_{\Omega}\z \E_0\cdot \Ph(x,0)dx,\\ & & \int_{0}^{T}\int_{\Omega}\left[- \B_{\z} \cdot \Ps_{t} + \E_{\z} \cdot (\g \Ps)\right] dxdt= \int_{\Omega} \left[ {\bf H}_{0}(x) \cdot \Ps(x,0)\right]dx. \eys Note by Lemma 2.1 that \bys & & |\int_{0}^{T}\int_{\Omega} \z \E_{\z t}\cdot \Ph dxdt|\\ & & \leq \left(\int_{0}^{T}\int_{\Omega} \z |\E_{\z t}|^2dxdt\right)^{1/2} \left(\int_{0}^{T}\int_{\Omega} \z |\Ph|^2 dxdt\right)^{1/2}\\ & & \rightarrow 0 \h \mbox{as $\z\rightarrow 0$}. \eys We take the limit as $\z\rightarrow 0$ to obtain \by & & \int_{0}^{T} \int_{\Omega}\left[{\bf J} \cdot \Ph \right] dxdt = \int_{0}^{T}\int_{\Omega}{\bf H}\cdot (\g \Ph)dxdt+ \int\int_{Q_{T}}\F \cdot \Ph dxdt\scl\\ & & \int_{0}^{T}\int_{\Omega}\left[- \B \cdot \Ps_{t} + \E\cdot (\g \Ps)\right] dxdt= \int_{\Omega} \left[ {\bf H}_{0}(x) \cdot \Ps(x,0)\right]dx. \scl \ey Now by choosing $\Ph =\lambda(t) \E$ and $\Ps=\lambda(t) \vh$ (note that the condition at $t=T$ is satisfied since $\lambda(T)=0$), we obtain \[ \int_{0}^{T}\int_{\Omega}[\lambda(t) {\bf J}\cdot \E] dxdt =\frac{1}{2}\int_{0}^{T}\int_{\Omega} [ \lambda'(t)|\vh|^2] dxdt+\frac{1}{2} \int_{\Omega}|\vh_{0}(x)|^2] dx+\int\int_{Q_{T}} \E\cdot\F dxdt.\] It follows by (3.3) that \by \lim_{\z \rightarrow 0}sup <\lambda(t){\cal}{L}[\E_{\z}], \E_{\z}> \leq <\lambda(t){\bf J}, \E>.\scl \ey Combination of (3.1) and (3.6) yields \by \lim_{\z \rightarrow 0}<\lambda(t){\cal}{L}[\E_{\z}], \E_{\z}> =<\lambda(t){\bf J}, \E>.\scl \ey Consequently, by choosing $\lambda(t)$ properly we have \by & & \lim_{\z \rightarrow 0}<{\cal}{L}[\E_{\z}], \E_{\z}> =<{\bf J}, \E>.\scl \ey % Now for any test vector fields $\Ps$ and $\Ph$, we have % \bys % & & \int\int_{Q_{T}}{\bf J}\cdot \Ph dxdt=\int\int_{Q_{T}}\B % \cdot (\g \Ph)dxdt,\\ % & & \int\int_{Q_{T}}\B_t\cdot\Ps +\E\cdot \Ps dxdt=0. % \eys For any vector field $\W\in L^2(Q_{T})\bigcap L^{p+1}(Q_{T})$, the monotonicity of $\sigma(x,s)$ in $s$ implies \by \int\int_{Q_{T}}[\sigma(x, |\E_{\z}|)\E_{\z}-\sigma(x, |\W|)\W] \cdot [\E_n-\W]dxdt\geq 0.\scl \ey We rewrite the above inequality to the following form: \by & & \int\int_{Q_{T}}\left\{ \sigma(|\E_{\z}|)|\E_{\z}|^2-\sigma(|\E_{\z}|)\E_{\z} \cdot \W\right\}dxdt \nonumber\\ & & \geq \int\int_{Q_{T}}[\sigma(x,|\W|)\W]\cdot [ \E_{\z}-\W]dxdt.\scl \ey We take the limit as $\z\rightarrow 0$ and use (3.8) for the first term in (3.10) to obtain \bys & & \int\int_{Q_{T}}\left\{ {\bf J}\cdot \E-{\bf J}\cdot \W\right\}dxdt\nonumber\\ & & \geq \int\int_{Q_{T}}[\sigma(x, |\W|)\W]\cdot [ \E-\W]dxdt. \eys Equivalently, \by & & \int\int_{Q_{T}}\left\{ [{\bf J}- \sigma(x, |\W|)\W]\cdot [ \E-\W]\right\} dxdt\geq 0.\scl \ey Set $\W=\E+\delta \Y $, where $\delta>0$ is small parameter and $\Y \in L^2(Q_{T})\bigcap L^{p+1}(Q_{T})$ is arbitrary. With the above choice of $\W$ in the equality (3.10), we obtain \bys & & \int\int_{Q_{T}}\left\{ \Y \cdot [{\bf J}-\sigma(|\E+\delta \Y |)[\E+\delta \Y ] \right\} dxdt\geq 0. \eys When $\sigma(x,s)$ is continuous in $s$, then we let $\delta\rightarrow 0$ to obtain \[ {\bf J}(x,t)=\sigma(x,|\E|)\E,\] since $\Y(x,t)$ is arbitrary in $L^2(Q_{T})\bigcap L^{p+1}(Q_{T})$. When $\sigma(x,s)$ has a jump at a point, say, $s=1$. Then as in \S 2 we understand that the value of $\sigma(x, s)$ at $s=1$ is \[ \sigma(x,1-)\leq \sigma(x,1)\leq \sigma(x,1+).\] By using the same procedure as in \S 2, we can derive the above inequality. Finally, by taking limit for (2.1)-(2.2) we see that $(\E, \vh)$ is a weak solution of the quasi-stationary system. To prove the uniqueness, we assume that $(\E,\B)$ and $(\E^*,\B^*)$ are two weak solutions to the quasi-stationary system. Let $\hat{\E}=\E-\E^*$ and $\hat{\B}= \B-\B^*$. Then the energy estimate implies \bys & & \sup_{0\leq t\leq T}\int_{\Omega}|\hat{\B}|^2dx+\int\int_{Q_{T}} [\sigma(x,|\E|)\E-\sigma(x, |\E^*|)\E^*]\cdot [\E-\E^*]dxdt\leq 0. \eys The monotonicity of $\sigma(x,s)$ in $s$ implies the second term in the above inequality is nonnegative. It follows that \[ \hat{\B}=0,\h a.e. \mbox{in $Q_{T}$}.\] From the definition of weak solution, we have $\hat{\E}=0$ as long as $\sigma>0$. \hfill Q.E.D. To conclude this section, we consider two special classes of electric conductivity $ \sigma(x, s)=s^{p}$ and \[ \sigma(x,s)=\left\{ \begin{array}{ll} & a, \,\mbox{if $|s|\leq 1$};\\ & b, \,\mbox{if $|s|>1$}, \end{array} \right. \] where $01\}$. Note that \[ div\B(x,t)=div\B_0(x)=0,\] we see that \[ \g\g\B=-\Delta \B.\] It follows that $\B$ satisfies the parabolic equation: \[ \B_{t}-\frac{1}{a}\Delta \B=-\frac{1}{a}\g\F, \h (x,t)\in Q_{T}^{-}\] and \[ \B_{t}-\frac{1}{b}\Delta \B=-\frac{1}{b}\g\F, \h (x,t)\in Q_{T}^{+}.\] The regularity theory of parabolic equations implies that $\B(x,t)$ is smooth in $Q_{T}^{\pm}$. The interface between $Q_{T}^{-}$ and $Q_{T}^{+}$ is defined by \[ \Gamma=\{(x,t): |\E|=1\},\] which is a free boundary. \ \\ {\bf Remark 3.1}: {\em We may allow that $\sigma(s)=0$ if $|\E|<1$ in the above example. In this case, one must consider the full system (1.1)-(1.2) in order to define a weak solution. However, in this case the uniqueness of the weak solution does not hold.} \ \\ {\bf Remark 3.2}: {\em It is not clear whether or not $\Gamma$ is indeed a hypersurface in $R^3\times (0,\infty]$. It would be of great interesting to study the smoothness of the interface $\Gamma$ and to find the free boundary conditions for $\B$.} \ \\ {\bf Acknowledgment}: {\em This work was done when author was visiting University of California and MSRI at Berkeley. The author would like to thank them for the hospitality. The author is also grateful to Professor L.C. Evans and Professor Guy David for many helpful discussions.} \begin{thebibliography}{99} \begin{small} \bibitem{AEW} G. Aronsson, L.C. Evans and Y. Wu, Fast/Slow diffusion and growing sandpiles, J. Diff. equations, 131(1996), 304-335. \bibitem{B} C.P. Bean, Magnetization of high-field superconductors, Rev. Mod. Phys., 36(1964), 31-39. \bibitem{BP} J.W. Barrett and L. Prigozhin, Bean's critical-state model as the $p\rightarrow \infty$ of an evolutionary $p-$Laplacian equation, preprint, December 10, 1997. \bibitem{DL} R. Dautray and J.-L. Lions, Analyse Mathematique et Calcul Numerique, Masson, 1988. \bibitem{D} E. DiBenedetto, Degenerate Parabolic Equations, Springer-Verlag, New York, 1993. \bibitem{E} L.C. Evans, Weak Convergence methods for Nonlinear Partial Differential Equations, AMS publication, Providence, R.I., 1990. %\bibitem{GT} D. Gilbarg and N. Trudinger, Elliptic Partial Differential Equations of Second Order, % Springer-Verlag, Berlin, 2nd ed., 1983. \bibitem{GR} V. Girault and P-A., Raviart, Finite Element methods for Navier-Stokes Equations, Springer-Verlag, New York, 1986. \bibitem{GY} R. Glassey and H.M. Yin, On Maxwell's equations with a temperature effect, II, Comm. in Mathematical Physics, 1998. \bibitem{H} D.W. Hewett, Low-frequency electromagnetic (Darwin) applications in plasma simulation, Computational Phys. Comm., 84(1994), 243-277. \bibitem{JO} F. Jochmann, The semistatic limit for Maxwell's Equations in an exterior domain, preprint. \bibitem{MM} A.C. Metaxas and R.J. Meredith, Industrial Microwave Heating, P. Peregrimus Ltd., London, 1983. \bibitem{MP} A. Milani and R. Picard, Weak solution theory for Maxwell's equations in the semistatic limit, J. Math. Anal. Appl., 191(1995), 77-100. \bibitem{LL} L. D. Landau and E.M. Lifshitz, Electrodynamics of Continuous Media, Pergamon Press, New York, 1960. \bibitem{LSU} O.A. La\,dyzenskaja, V. A. Solonnikov and N. N. Ural'ceva, Linear and Quasi--linear Equations of Parabolic Type, AMS Trans. 23, Providence., R.I., 1968. \bibitem{Y1} H.M. Yin, Regularity of solutions to Maxwell's system in quasi-stationary electromagnetic fields and applications, Comm. in P.D.E., 22(1997), 1029-1053. \bibitem{Y2} H.M. Yin, On a p-Laplacian type of evolution system and applications to the Bean model in the type-II superconductivity theory, MSRI preprint No. 1998-001, University of California at Berkeley. \bibitem{Y3} H. M. Yin, On Maxwell's equations with a temperature effect, SIAM J. on Math. Anal., 1998. \end{small} \end{thebibliography} \end{document}