info prev up next book cdrom email home

Green's Function

Let

\begin{displaymath}
\tilde L = \tilde D^n + a_{n-1}(t)\tilde D^{n-1} + \ldots + a_1(t)\tilde D + a_0(t)
\end{displaymath} (1)

be a differential Operator in 1-D, with $a_i(t)$ Continuous for $i=0$, 1, ..., $n-1$ on the interval $I$, and assume we wish to find the solution $y(t)$ to the equation
\begin{displaymath}
\tilde L y(t) = h(t),
\end{displaymath} (2)

where $h(t)$ is a given Continuous on $I$. To solve equation (2), we look for a function $g: C^n(I)\mapsto C(I)$ such that $\tilde L(g(h)) = h$, where
\begin{displaymath}
y(t)=g(h(t)).
\end{displaymath} (3)

This is a Convolution equation of the form
\begin{displaymath}
y=g*h,
\end{displaymath} (4)

so the solution is
\begin{displaymath}
y(t) = \int^t_{t_0} g(t-x)h(x)\,dx,
\end{displaymath} (5)

where the function $g(t)$ is called the Green's function for $\tilde L$ on $I$.


Now, note that if we take $h(t)=\delta(t)$, then

\begin{displaymath}
y(t)=\int_{t_0}^t g(t-x)\delta(x)\,dx = g(t),
\end{displaymath} (6)

so the Green's function can be defined by
\begin{displaymath}
\tilde Lg(t)=\delta(t).
\end{displaymath} (7)

However, the Green's function can be uniquely determined only if some initial or boundary conditions are given.


For an arbitrary linear differential operator $\tilde L$ in 3-D, the Green's function $G({\bf r},{\bf r}')$ is defined by analogy with the 1-D case by

\begin{displaymath}
\tilde LG({\bf r},{\bf r}') = \delta({\bf r}-{\bf r}').
\end{displaymath} (8)

The solution to $\tilde L\phi = f$ is then
\begin{displaymath}
\phi({\bf r}) = \int G({\bf r},{\bf r}')f({\bf r}')\,d^3{\bf r}'.
\end{displaymath} (9)

Explicit expressions for $G({\bf r},{\bf r}')$ can often be found in terms of a basis of given eigenfunctions $\phi_n({\bf r}_1)$ by expanding the Green's function
\begin{displaymath}
G({\bf r}_1,{\bf r}_2)=\sum_{n=0}^\infty a_n({\bf r}_2)\phi_n({\bf r}_1)
\end{displaymath} (10)

and Delta Function,
\begin{displaymath}
\delta^3({\bf r}_1-{\bf r}_2)=\sum_{n=0}^\infty b_n\phi_n({\bf r}_1).
\end{displaymath} (11)

Multiplying both sides by $\phi_m({\bf r}_2)$ and integrating over ${\bf r}_1$ space,
\begin{displaymath}
\int\phi_m({\bf r}_2)\delta^3({\bf r}_1-{\bf r}_2)\,d^3{\bf ...
...\infty b_n\int\phi_m({\bf r}_2)\phi_n({\bf r}_1)\,d^3{\bf r}_1
\end{displaymath} (12)


\begin{displaymath}
\phi_m({\bf r}_2)=\sum_{n=0}^\infty b_n \delta_{nm} = b_m,
\end{displaymath} (13)

so
\begin{displaymath}
\delta^3({\bf r}_1-{\bf r}_2)=\sum_{n=0}^\infty \phi_n({\bf r}_1)\phi_n({\bf r}_2).
\end{displaymath} (14)

By plugging in the differential operator, solving for the $a_n$s, and substituting into $G$, the original nonhomogeneous equation then can be solved.


References

Arfken, G. ``Nonhomogeneous Equation--Green's Function,'' ``Green's Functions--One Dimension,'' and ``Green's Functions--Two and Three Dimensions.'' §8.7 and §16.5-16.6 in Mathematical Methods for Physicists, 3rd ed. Orlando, FL: Academic Press, pp. 480-491 and 897-924, 1985.



info prev up next book cdrom email home

© 1996-9 Eric W. Weisstein
1999-05-25