HEJ, HU ISSN 1418-7108
Manuscript no.: ANM-981030-A
Frontpage previous next

   
Spaces

In the following, we always assume that $\Gamma=\partial\Omega$ and all interfaces between different materials are sufficiently smooth (e.g. piecewise from $C^2$). We use the standard Sobolev spaces (cf. e.g., [7]) $L^2(\Omega)$ and $L^2(\Omega)^3$ equipped with the norm $\vert\vert.\vert\vert _0$ and the scalar product $(.,.)_0$, and the space

\begin{displaymath}
L^2_\star(\Omega) = \left\{ \lambda \in L^2(\Omega) :
(\lambda,1)_0 = \int_\Omega \lambda \; dx = 0 \right\}.
\end{displaymath}

Next, we define the space $H^1(\Omega)^3$ with

\begin{displaymath}
\vert{\bf v}\vert _1^2 = \vert\vert { \bf grad }{\bf v} \vert\vert _0^2 =
\int_\Omega {\bf v}_{i,j} {\bf v}_{i,j} \;dx,
\end{displaymath}

where ${\bf v}_{i,j} = \frac{\partial {\bf v}_i}{\partial x_j}$ and Einstein's summation convention is applied to indices $i$ and $j$, and

\begin{displaymath}
\vert\vert{\bf v}\vert\vert _1^2 = \vert{\bf v}\vert _1^2 + \vert\vert{\bf v}\vert\vert _0^2.
\end{displaymath}

It is a standard result that $\vert.\vert _1$ is a norm of $H^1_0(\Omega)^3$. The following lemma (cf. [6]) allows us to represent the norm $\vert.\vert _1$ of the $H^1_0(\Omega)^3$.

Lemma 1   For all ${\bf v}\in H^1_0(\Omega)^3$ the equation
 \begin{displaymath}
\vert{\bf v}\vert _1^2 = \vert\vert{ \bf rot }{\bf v}\vert\vert _0^2 + \vert\vert\div {\bf v}\vert\vert _0^2
\end{displaymath} (16)

holds.

Proof. Assume ${\bf v}\in H^2(\Omega)^3$. We rewrite the ${ \bf rot }\;$ operator as follows

\begin{eqnarray*}
\int_\Omega { \bf rot }{\bf v} \cdot { \bf rot }{\bf v} \...
...t_\Omega {\bf v}_{i,j} {\bf v}_{i,j} - (\div {\bf v})^2
\;dx,
\end{eqnarray*}



and after having applied partial integration twice we arrive at
\begin{displaymath}
\vert\vert{ \bf rot }{\bf v}\vert\vert _0^2 = \vert{\bf v...
...quad \forall
{\bf v}\in H^1_0(\Omega)^3 \cap H^2(\Omega)^3.
\end{displaymath} (17)

Since $H^1_0(\Omega)^3 \cap H^2(\Omega)^3$ is dense in $H^1_0(\Omega)^3$, equation (16) holds for all ${\bf v}\in H^1_0(\Omega)^3$.
Let us discuss further spaces, cf. [4]. We define

\begin{eqnarray*}
H(\div ;\Omega) &=& \left\{{\bf v}\in L^2(\Omega)^3 : \div {\...
...ega) &=& \overline{{\cal D}(\Omega)}^{H({ \bf rot };\Omega)}.
\end{eqnarray*}



It is proved in [7] that

\begin{displaymath}
H^1_0(\Omega)^3 = H_0(\div ;\Omega) \cap H_0({ \bf rot };\Omega).
\end{displaymath}

We remark that $H_0(\div ;\Omega)$ has a zero trace of the normal component on $\Gamma=\partial\Omega$, whereas $H_0({ \bf rot };\Omega)$ has a zero trace of the tangential component. Therefore, the space $H_0({ \bf rot };\Omega)$ seems to be appropriate to the flux line boundary condition, cf. (12). Since we are interested in formulating a gauge condition that involves the divergence of the vector potential, we introduce the space
 \begin{displaymath}
{\bf Y}= H_0({ \bf rot };\Omega) \cap H(\div ;\Omega)
\end{displaymath} (18)

equipped with the norm
 \begin{displaymath}
\vert\vert{\bf v}\vert\vert _{\bf Y}^2 = \vert\vert{\bf v...
...f v}\vert\vert _0^2 + \vert\vert\div {\bf v}\vert\vert _0^2.
\end{displaymath} (19)

>From [7], we have the following results.

Lemma 2   Assume that $\Omega$ is bounded, Lipschitz-continuous, simply-connected and $\Gamma$ has just one component. Then there exists a positive constant $c$ such that

\begin{displaymath}
\vert\vert{\bf v}\vert\vert _0 \le c \left( \vert\vert{ \b...
...f v}\vert\vert _0 \right)
\quad \forall {\bf v} \in {\bf Y}
\end{displaymath}

holds.

Proof. See [7], Lemma 3.4.

Lemma 3   Assume that $\Omega$ is bounded, Lipschitz-continuous, simply-connected and $\Gamma$ has just one component. Then, the seminorm
 \begin{displaymath}
\vert{\bf v}\vert _{\bf Y}= \left( \vert\vert{ \bf rot ...
...rt _0^2 + \vert\vert\div {\bf v}\vert\vert _0^2 \right)^{1/2}
\end{displaymath} (20)

is a norm in Y, too.

Proof. Lemma 3 is a direct consequence of Lemma 2.
The next lemma requires more than the standard assumptions for technical magnetic field problems. Indeed, electromagnetic devices may have re-entrant corners. Therefore, we will not base our further investigations on Lemma 4.

Lemma 4   Assume further that either $\Omega$ has a ${\cal C}^{1,1}$ boundary, or $\Omega$ is a convex polyhedron. Then ${\bf Y}$ is continuously imbedded in $H^1(\Omega)^3$.

Proof. See [7], Theorem 3.7.
HEJ, HU ISSN 1418-7108
Manuscript no.: ANM-981030-A
Frontpage previous next