4
$\begingroup$

When solving the Schrodinger equation of the harmonic oscillator in one dimension you encounter the hermite differential equation:

\begin{equation} \left[\frac{d^{2}H}{d\xi^{2}}-2\xi\frac{d H}{d \xi }+\left(\lambda-1\right) H\right]e^{-\xi^{2}/2}=0. \end{equation}

We can find a solution to this equation in the form of a series:

\begin{equation} H(\xi)=\sum_{n=0}^{\infty} a_{n}\xi^{n} \ \end{equation}

with the following coefficients: \begin{equation} a_{n+2}=\frac{\left[2n-\left(\lambda-1\right)\right]a_{n}}{(n+2)(n+1)}. \label{coefficient_série} \end{equation} Since we look for a square integrable solution we usually truncate the series by imposing that $\lambda=2n+1$. My question is how do you assess that$\langle\psi|\psi\rangle$ diverges, with $\psi$= $H(\xi)e^{-\frac{\xi^2}{2}}$?

Usually in textbooks it's said that $H(\xi)$ behave assymptotically for large $\xi$ as $e^{\xi^2}$. Does someone as the actual derivation of this behaviour? From my understanding the only thing we can say is that for large value of $n$ we have:

\begin{equation} \frac{a_ {n+2}}{a_{n}} \simeq \frac{2}{n} \end{equation}

\begin{equation} e^{\xi^{2}}=\sum_{n=0}^{\infty}\frac{\xi^{2n}}{n!}=\sum_{n=0}^{\infty}c_{n}\xi^{2n} \end{equation}

\begin{equation} \left|\frac{c_{n+1}}{c_{n}}\right|=\frac{n!}{(n+1)!}=\frac{1}{n+1}. \end{equation}

So we see that both series are convergent with an infinite radius of convergence and that in both series the coefficients decreases in $\frac{1}{n}$. How is it sufficient to say that $H(\xi)$ behave as $e^{\xi^2}$ when $\xi$ goes to infinity?

$\endgroup$
2
  • $\begingroup$ This is an interesting question to me too. I have always heard that the power series method of solution for the harmonic oscillator lacks rigor and that the correct way to handle the problem is through Sturm-Liouville theory rather than make the $e^{-\xi^2}$ ansatz. I have never looked into the matter more though. $\endgroup$ Commented Jan 3 at 21:46
  • 1
    $\begingroup$ I think this is commented on in Galindo & Pascual's book (IIRC, they point to an appendix which discusses some properties of confluent hypergeometric functions). $\endgroup$ Commented Jan 3 at 22:00

1 Answer 1

3
$\begingroup$

For large $n$, $$\frac{a_{n+2}}{a_{n}}\approx\frac{2}{n}$$ means $$a_n\approx\frac{C}{(n/2)!}$$ for some constant $C$ and therefore $$H(\xi)\approx C\sum_n \frac{1}{(n/2)!} \xi^n\approx C\sum_l\frac{1}{l!}\xi^{2l}\approx C\exp(\xi^2)$$

Edit: To make this rigorous, it's enough to get a lower bound $H(\xi)\ge\epsilon \exp(\xi^2)$ for some $\epsilon>0$ and large enough $\xi$. That's enough to make the solution non-square-integrable.

$\endgroup$
2
  • $\begingroup$ So truncating the series ensures that $H(\xi)$ is a polynomial and hence does not blow up faster than $e^{-\xi^2}$ goes to zero? $\endgroup$ Commented Jan 4 at 1:52
  • $\begingroup$ @AlbertusMagnus indeed it explains why you have to truncate the series since multiplying $H(\xi)$ by $exp(-\frac{\xi^{2}}{2})$ is like $Cexp(\xi^{2})exp(-\frac{\xi^{2}}{2})=Cexp(\frac{\xi^{2}}{2})$ and this function is obviously not square integrable. $\endgroup$ Commented Jan 5 at 12:34

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.