2

I want to prove the following:

Show that for any one dimensional, time independent potential $U(x)$, where $\lim\limits_{x \rightarrow \infty}{U(x)} = 0$ and $\int_{-\infty}^{\infty}U(x) < 0$ there always exists at least one bound state with energy $E < 0$.

Hint: Approximate the expectation value of the energy $E(\alpha)$ in the state $\psi(x, \alpha) = \sqrt{\alpha} \exp(-\alpha|x|)$ for small $\alpha > 0$ and $\alpha \in \mathbb{R}$.

My question is if my proof below is correct and sound; I tried to also derive the approach given in the hint:

I argue that if $\int_{-\infty}^{\infty}U(x) < 0$ then there must exist an interval $[x_1, x_2]$ where $U(x \in [x_1, x_2]) < 0$. If I can now show that a bound state with negative energy exists even if this interval goes to zero, then we can deduct that such a state will always exist.

We reduce the interval $[x_1, x_2]$ to a point $x^{'}$; the potential at that point is then reduced to a delta distribution $U(x^{'}) = -C \delta(x - x^{'})$.

This leads to the following stationary Schrödinger Equation:

$$\frac {\partial^2} {\partial x^2} \psi(x) + \frac {2m} {\hbar^2} (C\delta(x - x^{'}) + E) \psi(x) = 0 $$

We know from a previous assignment that the normalized solution for the bound case is

$$\psi(x) = \sqrt{\alpha} \exp{(-\alpha |x|)}$$

where $\alpha = \sqrt{2mE} / \hbar$ and $\alpha \in \mathbb{C}$.

Let $\alpha = i \sqrt{2m|E|} = i \kappa$. Then the expectation value $\langle E \rangle = \langle \psi(x)| \hat{H} | \psi(x) \rangle$ with $\hat{H} = \frac{\hat{p}}{2m} + U(x)$ can be calculated as:

\begin{equation} \begin{split} \langle E \rangle & = \kappa^2 \int_{-\infty}^{\infty} \text{dx} \exp{(i \kappa |x|)} \left ( \frac{-\hbar^2}{2m} \frac {\partial^2} {\partial x^2} + U(x) \right) \exp{(- i \kappa |x|)} \\ & = \frac{\hbar^2 \kappa^4}{2m} \int_{-\infty}^{\infty} \text{dx} \exp{( i \kappa |x|)} \exp{(- i \kappa |x|)} + \kappa^2 \int_{-\infty}^{\infty} \text{dx} \exp{( i \kappa |x|)} \, U(x) \, \exp{(- i \kappa|x|)}. \\ \end{split} \end{equation}

The first integral must equal $1$ because bound states are normalizable. Since we're looking for bound states it's sufficient to consider small energies and thus $\alpha \approx 0$. Therefore, the first term vanishes. For the second integral we notice that the exponentials can now be approximated as $\exp(\pm i \kappa |x|) \approx \pm i \kappa |x|$ to first order. We retrieve:

\begin{equation} \begin{split} \langle E \rangle & = \kappa^2 \int_{-\infty}^{\infty} \text{dx} \exp{( i \kappa |x|)} \, U(x) \, \exp{(- i \kappa|x|)} \\ & \approx \kappa^4 \int_{-\infty}^{\infty} \text{dx} (i |x|) \, U(x) \, (- i |x|) \\ & = \kappa^4 \int_{-\infty}^{\infty} \text{dx} |x|^2 \, U(x). \end{split} \end{equation}

Now we remember that $\int_{-\infty}^{\infty}U(x) < 0$. Since $\alpha$ and $|x|^2$ are always positive and thus don't change the sign of the integral it follows that

$$\langle E \rangle < 0.$$

Finally, because every state $\psi(x)$ can be decomposed into eigenstates of the Hamiltonian as $\psi(x) = \sum_n c_n \psi_n(x)$ there must exist at least one bound energy eigenstate with eigenvalue $E < 0$.

Qmechanic
  • 201,751
  • 1
    The proof is given in my Phys.SE answer here. – Qmechanic Feb 07 '21 at 05:20
  • I am not sure that the proof is sound. At the end you have assumed the existence of a Hilbert basis of bound states (no continuous spectrum). That was actually part of the thesis. – Valter Moretti Feb 07 '21 at 08:35
  • @Qmechanic thanks for the link! I also went through the referenced paper and it helps a lot already. So to summarize is it correct that the hint given in the assignment is actually a test function and the task is to minimize $\bar{E}(\alpha)$?

    If yes, why is it not rather suggested to evaluate $d/d\alpha \bar{E}(\alpha) = 0$ but rather to approximate $\alpha \approx 0$?

    My suggestion is that $\alpha \approx 0$ is already enough to show that negative eigenvalues exist and the task is not to find its exact value. But maybe you can try to explain this in your own words? Thanks!

    – FunkyPeanut Feb 07 '21 at 12:33
  • @ValterMoretti My line of reasoning was that first I show that there exists a bound state with negative energy expectation value. Now if an eigenbasis of the Hamiltonian exists, then it can be used to represent that bound state. So the assumption is that such a basis exists but not that is is already a basis of bound states. I think it's safe to make that claim, or maybe I missed something? – FunkyPeanut Feb 07 '21 at 15:42
  • I think that this claim is the crucial claim and it is the principal fact one should prove. The remaining part is quite easy. – Valter Moretti Feb 07 '21 at 15:54
  • @ValterMoretti Ah, now that I think about it... are you saying the issue is that in my proof I'm choosing an arbitrary state $\psi(x)$ where I'd still have to show that this state can be represented with energy eigenstates? How is this problem solved by using the variation principle? I have been reading that the test function is defined to have a representation in $\hat{H}$; why can't I do the same here? Because as far as I have been reading the test function can be arbitrary, too, right? Thanks for the help! :) – FunkyPeanut Feb 07 '21 at 16:18
  • No, it cannot be solved using only the variational principle. One should prove that the subspace corresponding to the negative part of the spectrum is made of proper eigenvalues. I guess that it can be proved by demonstrating that the resolvent operator of your Schroedinger operator projected in that subspace is compact. Presumably exploiting some extension of Sturm Liouville theory. Compact operators have discrete spectrum up to, possibly, zero. – Valter Moretti Feb 07 '21 at 16:27
  • @ValterMoretti After some research I see now that you are right. Thanks for the input! By the way your book on spectral theory is very interesting. I plan to read into it once I'm a bit more familiar with the basics. – FunkyPeanut Feb 07 '21 at 23:57

0 Answers0