Green's function for Sturm-Louiville ODE

Join the discussion
Ask a follow-up here, or get your own question answered by working scientists, mathematicians and engineers — people, not an autocomplete.
Real named experts · corrections over time · the nuance an AI answer skips
3 replies · 2K views
member 428835
Hi PF!

Given the following ODE $$(p(x)y')' + q(x)y = 0$$ where ##p(x) = 1-x^2## and ##q(x) = 2-1/(1-x^2)## subject to $$y'(a) + \sec(a)\tan(a)y(a) = 0$$ and $$|y(b)| < \infty,$$ where ##a = \sqrt{1-\cos^2\alpha} : \alpha \in (0,\pi)## and ##b = 1##, what is the Green's function?

This is the associated Legendre ODE, which admits two linearly independent solutions: $$y_1 = P_1^1, \,\,\, y_1 = Q_1^1$$ where ##P_1^1## and ##Q_1^1## are associated Legendre polynomials first and second kind respectively. Things now get a little murky: notice ##y_1## automatically satisfies ##y_1'(a) + \sec(a)\tan(a)y_1(a) = 0##, but ##y_2(b) = \infty##. If ##y_2(b)## was bounded, the Green's function would be very simple to calculate. But it's not bounded, so I'm confused how to proceed?

Any help is greatly appreciated.
 
Physics news on Phys.org
If you want a solution bounded at 1, then you are limited to scalar multiples of [itex]P_1^1[/itex]: [tex] G(x;x_0) = \begin{cases}<br /> AP_1^1(x) & x \leq x_0 \\<br /> BP_1^1(x) & x > x_0\end{cases}.[/tex] The conditions you need to satisfy are [tex] \begin{split}<br /> (A - B)P_1^1(x_0) &= 0 \\<br /> (A - B){P_1^1}'(x_0) &= \frac{1}{p(x_0)}\end{split}[/tex] which cannot be done unless [itex]P_1^1(x_0) = 0[/itex] or you are prepared to relax the constraint that [itex]G(\cdot;x_0)[/itex] be continuous at [itex]x_0[/itex].
 
I appreciate your reply. So it looks like a Green's function doesn't exist for this problem, right?
 
joshmccraney said:
I appreciate your reply. So it looks like a Green's function doesn't exist for this problem, right?

Yes.

The way to solve the non-homogenous boundary value problem [tex] L(y) = (py')' + qy = -g[/tex] on the interval [itex][a,b][/itex] subject to arbitrary boundary conditions on [itex]y[/itex] is to first solve the eigenvalue problem [tex] L(\phi) = -\lambda w \phi[/tex] for a convenient weight function [itex]w[/itex] which is strictly positive on [itex](a,b)[/itex] with [itex]\phi[/itex] subject to a convenient self-adjoint boundary condition. This gives you a discrete spectrum of eigenvalues [itex]\lambda_n \neq 0[/itex] with corresponding eigenfunctions [itex]\phi_n[/itex] which are orthogonal with respect to the inner product [itex](f,g)_w = \int_a^b f(x)g(x) w(x)\,dx[/itex].

(In this case, choosing a suitable boundary condition and setting [itex]w = 1[/itex] will give [itex]\phi_n = P_n^1[/itex] with [itex]\lambda_n = n(n+1) - 2[/itex] for [itex]n \geq 2[/itex].)

Next, look for a particular solution [tex]y_P = \sum_n a_n \phi_n[/tex] so that [tex] L(y_P) = -\sum_na_n\lambda_n w \phi_n = -g[/tex] and then multiplying by [itex]\phi_m[/itex] and integrating you should find[tex] a_m = \frac{1}{\lambda_m \|\phi_m\|_w^2}\int_a^b g(x)\phi_m(x)\,dx.[/tex] A complimentary function [itex]y_C \in \ker L[/itex] can then be added to [itex]y_P[/itex] in order to satisfy the boundary conditions on [itex]y[/itex].