John Creighto
Sep17-09, 03:37 PM
According to wikipedia the greens function is defined as:
L G(x,s) = - \delta(x-s)\,
http://en.wikipedia.org/wiki/Green%27s_function#Definition_and_uses
when L is a differential equation then the greens function is the impulse response of the differential equation.
If a Hilbert space can be found for the operator then the greens function is given as follows:
K(x,y)=\sum_n \frac{\psi_n^*(x) \psi_n(y)} {\omega_n}
http://en.wikipedia.org/wiki/Fredholm_theory#Homogeneous_equations
Where \phi are the eigen vectors and \omega_n are the eigenvalues of the operator. (Not sure how unbounded basis are dealt with).
For ODEs we can find the eigenvalues by finding the poles of the Laplace transform. I'm wondering if there is some generalization of the Laplace transform for partial differential equations. The form of the resolvent:
R(z;A)= (A-zI)^{-1}.\,
http://en.wikipedia.org/wiki/Resolvent_formalism
Looks strangely similar to part of the solution when solving for S the Laplace transform of a system of first order linear differential equations. Also with regards to generalizing with respect to partial differential equations, I presume a convolution with a greens function turns into a multiple convolution over several variables.
Thinking in terms of ODEs the poles of the resultant should be the eigenvalues.
L G(x,s) = - \delta(x-s)\,
http://en.wikipedia.org/wiki/Green%27s_function#Definition_and_uses
when L is a differential equation then the greens function is the impulse response of the differential equation.
If a Hilbert space can be found for the operator then the greens function is given as follows:
K(x,y)=\sum_n \frac{\psi_n^*(x) \psi_n(y)} {\omega_n}
http://en.wikipedia.org/wiki/Fredholm_theory#Homogeneous_equations
Where \phi are the eigen vectors and \omega_n are the eigenvalues of the operator. (Not sure how unbounded basis are dealt with).
For ODEs we can find the eigenvalues by finding the poles of the Laplace transform. I'm wondering if there is some generalization of the Laplace transform for partial differential equations. The form of the resolvent:
R(z;A)= (A-zI)^{-1}.\,
http://en.wikipedia.org/wiki/Resolvent_formalism
Looks strangely similar to part of the solution when solving for S the Laplace transform of a system of first order linear differential equations. Also with regards to generalizing with respect to partial differential equations, I presume a convolution with a greens function turns into a multiple convolution over several variables.
Thinking in terms of ODEs the poles of the resultant should be the eigenvalues.