next up previous
: Derivation of an exact : Difference Notation : Difference operators

Differentiation Formulas:

In order to obtain formulas for numerical differentiation, by operational methods, it is necessary to relate D (the differential operator) to the operators $ \Delta,$ $ \nabla$ and $ \delta$ (delta operators). For this purpose, we notice that the familiar formula of the Taylor-series expansion

$\displaystyle p(x+h)=p(x)+\frac{h}{1!}\quad
p'(x)+\frac{h^{2}}{2!}\quad p''(x)+.......+\frac {h^{n}}{n!}\quad
p^{n}(x)+........$

(which certainly is valid for any polynomial) can be written in the operational form

$\displaystyle E
p(x)=(I+\frac{hD}{1!}+\frac{h^{2}D^{2}}{2!}+......+\frac{h^{n}D^{n}}{n!}+......)p(x)$

$\displaystyle =exp(hD)p(x)$

and thus we deduce that

$\displaystyle E=exp(hD)$

which is to be interpreted as an abbreviation of the statement that the operators $ E$ and $ \displaystyle
I+\frac{hD}{1!}+\frac{h^{2}D^{2}}{2!}+......+\frac{h^{n}D^{n}}{n!}$ are equivalent when applied to any polynomial $ p_{{n}}(x)$ of degree $ n$ for any n.
Further, we obtain the additional relations

$\displaystyle hD = log E = log(I+\Delta)$

$\displaystyle =-log(I-\nabla)$

$\displaystyle = 2\quad
log\left[\left(I+\frac{1}{4}\delta^{2}\right)^{\frac{1}{2}}+\frac{1}{2}\delta\right]$

$\displaystyle =2\quad sinh^{-1}\frac{\delta}{2}$

Classification of second order PDES':
The general linear second order partial differential equation in two independent variables is given by

$\displaystyle A u_{xx}+2B u_{xy}+Cu_{yy}+Du_{x}+Eu_{y}+Fu+G=0$

where the coefficients are functions of x and y and the subscripts denote partial derivatives with respect to the independent variables. The above equation is called

$\displaystyle elliptic, \quad if \quad B^{2}-AC<0 $

$\displaystyle parabolic, \quad if \quad B^{2}-AC=0 $

$\displaystyle and \quad hyperbolic,\quad if \quad B^{2}-AC>0 $

This classification depends in general on the region of the $ (x,y)$ plane under consideration. The differential equation
     $ xu_{xx}+u_{yy}=0,$     for instance,    is elliptic for $ x>0,$
$\displaystyle hyperbolic for \quad x<0 \quad and $
$\displaystyle parabolic for \quad x=0 \quad $

The well known examples of the three types are:
Heat Flow equation:
\begin{displaymath}
\begin{array}{ccc}
\displaystyle \frac{\partial u}{\partial...
...&.......& \texttt{which is of parabolic type} \\
\end{array}%
\end{displaymath}
Wave equation:
\begin{displaymath}
\begin{array}{ccc}
\displaystyle \frac{\partial^{2}u}{\part...
..........& \texttt{which is of hyperbolic type} \\
\end{array}%
\end{displaymath}
Laplace equation:
\begin{displaymath}
\begin{array}{ccc}
\displaystyle \frac{\partial^{2}u}{\part...
...0&.......& \texttt{which is of elliptic type} \\
\end{array}%
\end{displaymath}
The parabolic and hyperbolic type of equations are either initial value problems or initial boundary value problems whereas the elliptic type equation is always a boundary value problem. The boundary conditions can be one of the following three types:
  1. The Disichlet or first boundary condition: Here, the solution is prescribed along the boundary
  2. The Neumann or Second boundary condition: Here, the derivative of the solution is specified along the boundary.
  3. The third a mixed boundary condition: Here, the solution and its derivative are prescribed along the boundary.

next up previous
: Derivation of an exact : Difference Notation : Difference operators