1
$\begingroup$

I'm interested in the existence of eigenfunctions and finding eigenvalues of the following operator

$$L(\varphi) = \varphi_{rr} - \frac{1}{r} \varphi_r - [V + \frac{m}{r^2}] \varphi$$ $$\varphi(0) = \varphi(1) = 0$$

where $\varphi: [0,1] \to \mathbb{R}$, $m$ is any nonnegative integer, and $V$ is say smooth and bounded and $-1 \leq V \leq 2$. This comes from a Schrödinger operator $L = \Delta + \overline{V}$ motivated by the Allen-Cahn equation (see here) on $D^2$ with some radial symmetry assumed.

I'm not sure if there are general sources about existence of eigenfunctions in this 1 dimensional case (or determining the index), especially since there is degeneracy as $r \to 0$. Any references would be appreciated!

$\endgroup$
10
  • 1
    $\begingroup$ If you don't specify anything about boundary conditions, it's hard to see how anything could be said here. $\endgroup$ Commented Jan 13 at 15:49
  • 1
    $\begingroup$ This operator is not symmetric on $L^2(0,1)$, so perhaps you are considering it on a weighted $L^2$ space? $\endgroup$ Commented Jan 13 at 15:51
  • 2
    $\begingroup$ It's symmetric on $L^2(I; dr/r)$. In Sturm-Liouville form the equation is $-(1/r\varphi')'+(V/r+m/r^3)\varphi =(\lambda/r)\varphi$. $\endgroup$ Commented Jan 14 at 0:58
  • 3
    $\begingroup$ The transformation $y=r^{-1/2}\varphi$ brings this to Schrodinger form $-y''+(V+(m+3/4)/r^2)y=\lambda y$, if I calculated correctly (unlikely, as far as the details are concerned). This is a perturbation of the $c/r^2$ potential, which is well studied. For $c>3/4$, as here, you have limit point case at $r=0$. $\endgroup$ Commented Jan 14 at 1:55
  • 2
    $\begingroup$ One final comment: the boundary condition you added at $r=0$ doesn't make sense because this endpoint is singular, so solutions/elements of the domain don't necessarily have continuous extensions to $r=0$. More to the point, my previous comment shows (if I got the details right) that you don't need a boundary condition at $r=0$. $\endgroup$ Commented Jan 14 at 1:56

1 Answer 1

5
$\begingroup$

This is a slightly expanded (and slightly more systematic) summary of my comments above. First of all, the equation $$ -\varphi''+\frac{1}{r}\varphi' + (V+\frac{m}{r^2})\varphi=\lambda\varphi \quad\quad\quad\quad (1) $$ fits into the standard Sturm-Liouville theory if written in the form $$ T\varphi\equiv-\left( \frac{1}{r}\varphi'\right)'+W\varphi =\frac{\lambda}{r}\varphi , $$ with $W=V/r+m/r^3$. Now $T$ defines a symmetric operator on $L^2(I,dr/r)$.

The equation can be brought to Schrodinger form by the substitution $y=r^{-1/2}\varphi$. Then (1) becomes $$ -y'' + \left( V + \frac{m+3/4}{r^2} \right) y = \lambda y .\quad\quad\quad\quad (2) $$ Now the LHS defines an operator $Sy$ on $L^2(I, dr)$ which is unitarily equivalent to $T$.

Near $r=0$, this is a bounded perturbation of the $c/r^2$ potential. For $c\ge 3/4$, as here, we have limit point case: For $\lambda=0$ (and $V=0$), we can solve explicitly by $y=r^{\alpha}$, with $\alpha=1/2\pm \sqrt{1/4+c}$, and for $c\ge 3/4$, only the solution with $+$ is in $L^2$ near $r=0$.

As a consequence, the minimal operator (the closure of $S$ on $C_0^{\infty}(0,1)$) is symmetric with deficiency $(1,1)$. The self-adjoint realizations are obtained by imposing a boundary condition at $r=1$ only on the elements of the domain of the adjoint. The general boundary condition is $y(1)\cos\beta +y'(1)\sin\beta =0$; this gives all self-adjoint realizations.

As for the spectrum, I doubt that one can find this explicitly even when $V=0$. If $V$ is analytic, then $r=0$ is a regular singular point of (2), so the solution that is in $L^2$ will be of the form $y(r,\lambda)=r^{\alpha}\sum_{n\ge 0}c_n(\lambda)r^n$, with $\alpha=1/2+\sqrt{1+m}$ from above. The eigenvalues will be the solutions of $y(1,\lambda)=0$. (In particular, the regularity of these solutions proves that the spectrum is purely discrete, which need not hold in general when there is a limit point endpoint.)

$\endgroup$
1
  • 1
    $\begingroup$ For $V=0$, this is the equation for a Bessel function. Of course, the zeros of the Bessel functions are not something one can write explicit expressions for (except in limits). $\endgroup$ Commented Jan 14 at 17:23

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .