0

I'm trying to compute the 2nd order correction to the energy spectrum of a 1D quantum harmonic oscillator when a perturbation of the form $\gamma\,\hat{x}^k$ (with $\gamma\ll1$) is added to the Hamiltonian. This is

$$E_n^{(2)}\equiv\langle \psi_n^{(0)} | \,\gamma\hat{x}^k\, | \psi_n^{(1)} \rangle = \sum_{n'\neq n} \frac{{\left|\, \langle \psi_{n'}^{(0)} | \,\gamma\hat{x}^k\, | \psi_n^{(0)} \rangle\,\right|}^{\,2}}{E_n^{(0)}-E_{n'}^{(0)}} = \left(\frac{\hbar}{2m\omega}\right)^k \sum_{n'\neq n} \frac{{ \gamma^2 \left|\, \langle n' | \,(\hat{a}^\dagger+\hat{a})^k\, | n\rangle \,\right|}^{\,2}}{\hbar\omega\,(n-n')}\;.$$

where $\,\hat{x}=\sqrt{\frac{\hbar}{2mw}}\,(\,\hat{a}^\dagger + \hat{a}\,)\;$, $\;E_n^{(0)}=\hbar\omega(n+1/2)\;$ and $\;| \psi_n^{(0)} \rangle\equiv|n\rangle$.


Now, the question is how to compute this term:

$$\alpha_{n'n}\equiv \langle n' | \,(\hat{a}^\dagger+\hat{a})^k\, | n\rangle =\left(\frac{\langle \,0\,|\hat{a}^{n'} (\hat{a}^\dagger+\hat{a})^k {\hat{a}^\dagger}^n\, |\,0\, \rangle}{\sqrt{n'\,!\; n!}}\right)\;.$$

I think it can be "easily" obtained uaing Wick's theorem but I'm not sure about how to apply it. I'm struggling with the part of counting all of the possible ways to fully-contract the operators for any $k$. Do you know any shortcut or some hint to help me compute it?. In fact, is it even manageable?.

JuanC97
  • 266
  • 3
  • 9

1 Answers1

-1

Use the definition of $a$ and $a^{\dagger}$: $$ a \left| n \right> = \sqrt{n} \left| n - 1 \right>, $$ $$ a^{\dagger} \left| n \right> = \sqrt{n+1} \left| n + 1 \right>. $$

Expand the $k$-th power in a combinatorical series, then apply the definitions above. Finally, use the orthonormality relation $$ \left< n | m \right> = \delta_{nm}. $$

  • 1
    For arbitrary $k$ that'd be pretty cumbersome as $\hat{a}$ and $\hat{a}^\dagger$ don't commute. The correct binomial expansion formula is just unmanageable. However, for $k$'s of order 1 one might try what you suggested. – JuanC97 Dec 06 '20 at 09:50
  • 1
    @JuanC97 it is still manageable for arbitrary $k$, you just have to do more work. – Prof. Legolasov Dec 06 '20 at 12:07
  • Would you like to elaborate by writing down which series you mean? The binomial series will not generally hold for non-commuting operators – ohneVal Dec 23 '20 at 11:47