Representing solutions to the radial Laplace equation in an eigenbasis

122 Views Asked by At

I am wondering if the solution $u(r) = \ln(r)/\ln(S)$ to the ODE

\begin{align} \Delta_r (u) = \frac{d^2 u}{dr^2} + \frac{1}{r}\frac{\partial u}{\partial r} = 0, \ \ \ u(S) = 1, \ \ \ u(1) = 0, \end{align}

can be expanded in some eigenbasis {$\phi_i$} for the radial Laplacian $\Delta_r$. I've tried considering the expansion $u(r) = \sum_{n=1}^\infty a_n\hat{\phi}_n(r)$, however substituting this into the ODE gives

\begin{align} -\sum_{n=1}^\infty a_n\lambda_n \hat{\phi}_n(r)= 0, \end{align}

which seems to imply that each $a_i = 0$, and therefore no such representation is possible.

Note that the eigenbasis corresponding to the above problem consists of linear combinations of Bessel functions of the first and second kind.

1

There are 1 best solutions below

0
On BEST ANSWER

The Sturm-Liouville form of your operator is \begin{align} Lf &= -\frac{d^2f}{dr^2}-\frac{1}{r}\frac{df}{dr}=-\frac{1}{r}\frac{d}{dr}\left(r\frac{df}{dr}\right). \end{align} This is a formally self-adjoint operator on $L^2_r[S,1]$ with respect to the following inner product $$ \langle f,g\rangle_r = \int_S^1f(r)g(r)rdr. $$ However, it is not actually self-adjoint without imposing endpoint conditions such as $f(S)=f(1)=0$. For such conditions imposed on $f,g$, it follows that \begin{align} \langle Lf,g\rangle_r-\langle f,Lg\rangle_r &=-\int_S^1\frac{d}{dr}\left(r\frac{df}{dr}\right)g-f\frac{d}{dr}\left(r\frac{dg}{dr}\right)dr \\ & =-\int_S^1\frac{d}{dr}\left[r\frac{df}{dr}g-rf\frac{dg}{dr}\right]dr \\ & = -r(f'g-fg')|_{S}^{1}=0. \end{align} Therefore $\langle Lf,g\rangle_r = \langle f,Lg\rangle_r$ for $f,g\in\mathcal{D}(L)$. For $S > 0$, this is a regular Sturm-Liouville problem; so the eigenvalues form a discrete set and the corresponding eigenfunctions, once normalized, form a complete orthonormal basis of $L^2[S,1]$. So you'll be able to use this set of eigenfunctions to expand any function in $L^2[S,1]$ using this set of eigenfunctions.