- Huge subject; excellent textbooks
- Many library subroutines exist
- But: "physical" matrices often simple in structure
- Specific algorithms that may (may!) be self-programmed
- We will concentrate on Relaxation Methods
Similarly, given
$f(x,y)$ or $f(x,t)$
$\Longrightarrow$ Matrix
$F \equiv [f_{i,j}] \equiv [f(x_{i},y_{j}) ; \; i=1,\dots M;\;j=1,\dots N]$
Approximate the various differentials by differences:
$\Longrightarrow$
Convert Partial Differential Equations (PDEs)
into Systems of Linear Equations
$ A \cdot x= b$
As a rule $A$
has a simple structure: sparse,
diagonally dominated, positive definite, etc.
Fundamental manipulations:
Invert a matrix:
$
\begin{eqnarray}
A & \Longleftrightarrow & A^{-1}
\end{eqnarray}
$
Find the solution to the system of equations:
$
\begin{eqnarray}
A \cdot x = b
\end{eqnarray}
$
Find the eigenvalues
$\lambda_{i}$ and the eigenvectors
$ a_{i}$ of a quadratic matrix:
$
\begin{eqnarray}
\left.
\begin{array}{r@{\quad=\quad}l}
\left| A - \lambda_{i} I \right| & = 0 \\
( A - \lambda_{i} I) \cdot a_{i} & = 0
\end{array} \right\} && \; \; \; i=1,\dots N
\end{eqnarray}
$