-1
$\begingroup$

I have the Hamiltonian of an harmonic oscillator (with $\hbar=1$) $$ H = \omega \left(a^\dagger a + \dfrac{1}{2} \right) \;, $$ and the associated (canonical) partition function $$ Z = \text{Tr}\left[e^{-\beta H} \right] = \dfrac{1}{2} \text{csch}\left(\dfrac{\beta \omega}{2}\right) \;, $$ where $\beta = 1/T$ is the inverse temperature (with $k_B=1$).

For the operator $$ O = A(a+a^\dagger)+B(a+a^\dagger)^2 \;, $$ where $A$ and $B$ are real constants, I want to know its expectation value when evaluated on the harmonic oscillator at finite $\beta$.

Since this expectation value is given by $$ \langle O \rangle = \dfrac{\text{Tr}\left[O e^{-\beta H} \right]}{Z} \;, $$ the sum at the numerator can be easily computed if $O$ and $H$ can be diagonalised in the same basis. However, what should I do when this is not the case?

$\endgroup$
  • $\begingroup$ Welcome to Physics Stack Exchange. We support questions like this only when the author shows effort and identifies a specific aspect on which they are stuck. $\endgroup$ – DanielSank Oct 2 '17 at 16:54
  • $\begingroup$ @DanielSank, I'll add some more details. $\endgroup$ – m137 Oct 2 '17 at 20:46
1
$\begingroup$

You can always go straightforward and use that you know how $\hat{a}$ and $\hat{a}^\dagger$ act on the basis $|n\rangle$. Let's write down straightforwadly what $\mathrm{Tr}$ is \begin{equation} \mathrm{Tr}\Big(\hat{O}e^{-\beta\hat{H}}\Big)=\sum_n\langle n|\hat{O}e^{-\beta\hat{H}}|n\rangle=\sum_n e^{-\beta(n+1/2)}\langle n|\hat{O}|n\rangle \end{equation} Now, your operators can be written in the form, \begin{equation} \hat{O}=\sum_{lk}o_{lk}\Big(\hat{a}^\dagger\Big)^l \Big(\hat{a}\Big)^k \end{equation} As trace is linear that implies that, \begin{equation} \mathrm{Tr}\Big(\hat{O}e^{-\beta\hat{H}}\Big)=\sum_{nlk}e^{-\beta(n+1/2)}o_{lk}\langle n|\Big(\hat{a}^\dagger\Big)^l \Big(\hat{a}\Big)^k|n\rangle \end{equation} Now remember that, \begin{equation} a^\dagger|n\rangle=\sqrt{n+1}|n+1\rangle,\quad a|n\rangle=\sqrt{n}|n-1\rangle \end{equation} That means that, \begin{equation} \langle n|\Big(\hat{a}^\dagger\Big)^l \Big(\hat{a}\Big)^k|n\rangle\sim \langle n|n+l-k\rangle \end{equation} But as $\langle n|m\rangle=\delta_{nm}$ that means that only when $l=k$ this product is nonzero. So you need to take into account only such terms in $\hat{O}$. From here I think it's easy to do the rest.

$\endgroup$
0
$\begingroup$

If $\hat H$ admits of the basis $\hat H |n\rangle = E_n|n\rangle$ and this basis resolves the identity $\hat 1 = \sum_n |n\rangle\langle n|$, then we can insert this twice and use the cyclic permutativity rule of trace to rewrite $$\operatorname{Tr}\left(\hat O~e^{-\beta \hat H}\right) = \sum_{m,n}\langle m|\hat O|n\rangle~\langle n|e^{-\beta \hat H}|m\rangle=\sum_n\langle n|\hat O|n\rangle e^{-\beta E_n}.$$ This is not the only basis which makes things simple, but it's probably the most straightforward in general, and in your case where you can define $x = a^\dagger + a$ as one of the quadratures of your system (different people normalize this in different ways) and then have $O = A x+ Bx^2,$ this is likely to just be $$\sum_n e^{-\beta~\hbar\omega~n} \int_{-\infty}^\infty dx~ \psi_n^*(x)~ \psi_n(x)~(A x + B x^2).$$ The $A x$ term probably vanishes completely due to symmetry and the $x^2$ is probably not too complicated to work out.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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