I'm referencing this thesis which should be open-access.

In Appendix D.1 "High temperature expansion in general", the author writes the high temperature expansion in the following way:

$$ \begin{align*} \langle \hat{O} \rangle &= \frac{\sum_i \langle i| \hat{O} e^{-\beta \hat{H}} |i\rangle}{\sum_i \langle i| e^{-\beta \hat{H}} |i\rangle} \\ &= \beta^0 \Bigl[ \frac{1}{\Theta} \mathrm{Tr}(\hat{O})\Bigr] -\beta^1 \Bigl[ \frac{1}{\Theta} \mathrm{Tr}(\hat{O}\hat{H}) - \frac{1}{\Theta^2} \mathrm{Tr}(\hat{O})\mathrm{Tr}(\hat{H}) \Bigr] \\ &\qquad + \beta^2 \Bigl[ \frac{1}{2}\frac{1}{\Theta} \mathrm{Tr}(\hat{O}\hat{H}^2) - \frac{1}{\Theta^2} \mathrm{Tr}(\hat{OH})\mathrm{Tr}(\hat{H}) \\ &\qquad\quad - \frac{1}{2}\frac{1}{\Theta^2} \mathrm{Tr}(\hat{O})\mathrm{Tr}(\hat{H}^2) + \frac{1}{\Theta^3} \mathrm{Tr}(\hat{O})\mathrm{Tr}(\hat{H})^2 \Bigr] + \mathcal{O}(\beta^3) \end{align*} $$ where $\hat{O}$ is some operator and trace is over multiparticle states $|i\rangle$; $\Theta \equiv \mathrm{Tr}(I)$ is the dimension of the problem.

My question is: How did they do this expansion? (My attempt:) Clearly there has been an expansion of the exponential in the numerator in terms of $\beta$, $$ \sum_i \langle i| \hat{O} e^{-\beta \hat{H}} |i\rangle = \sum_m \frac{(-\beta)^m}{m!} \sum_i \langle i| \hat{O} \hat{H}^m |i\rangle \tag{1} $$

but I'm not sure 1) how or where $\Theta$ comes from, and also 2) why there is a split of the traced terms: $\mathrm{Tr}(\hat{O}\hat{H})$, $\mathrm{Tr}(\hat{O})\mathrm{Tr}(\hat{H})$ in $\beta^1$ for example. And also 3) how to formally divide out the denominator $\sum_i \langle i| e^{-\beta \hat{H}} |i\rangle$, like after substituting (1) back into the original equation:

$$ \frac{\sum_i \langle i| \hat{O} \hat{H}^m |i\rangle}{\sum_i \langle i| e^{-\beta \hat{H}} |i\rangle} = \text{terms for each $\beta^m$} $$

Can someone enlighten me on this?


To first order in $\beta$, the numerator reads $$\eqalign{ \sum_i<i|Oe^{-\beta H}|i> &=\sum_i <i|O|i>-\beta\sum_i<i|OH|i>\cr &={\rm Tr}\ \!O-\beta\ \!{\rm Tr}\ \!OH\cr }$$ while the denominator is $$\eqalign{ \sum_i<i|e^{-\beta H}|i> &=\sum_i <i|i>-\beta\sum_i<i|H|i>\cr &=\Theta-\beta\ \!{\rm Tr}\ \!H\cr &=\Theta\Big(1-{\beta\over\Theta}\ \!{\rm Tr}\ \!H\Big)\cr }$$ Since $\beta$ is small, the inverse is (to first order) $${1\over\Theta}\Big(1-{\beta\over\Theta}\ \!{\rm Tr}\ \!H\Big)^{-1}={1\over\Theta}\Big(1+{\beta\over\Theta}\ \!{\rm Tr}\ \!H\Big)$$ The average is finally to first order in $\beta$ $$\eqalign{ \langle O\rangle&={1\over\Theta}\Big({\rm Tr}\ \!O-\beta\ \!{\rm Tr}\ \!OH\Big)\Big(1+{\beta\over\Theta}\ \!{\rm Tr}\ \!H\Big)\cr &={1\over\Theta}{\rm Tr}\ \!O -{\beta\over\Theta}{\rm Tr}\ \!OH +{\beta\over\Theta^2}{\rm Tr}\ \!O\ {\rm Tr}\ \!H\cr }$$ I let you extend the calculation to higher orders.

| cite | improve this answer | |
  • $\begingroup$ There's a typo for your denominator line, but otherwise I get the gist of the method. Thank you! ($\sum_i<i|e^{-\beta H}|i>$, not $\sum_i<i|Oe^{-\beta H}|i>$.) $\endgroup$ – ksgj1 Nov 7 '18 at 2:36
  • $\begingroup$ Thanks! I have updated the answer. I am happy to have been helpful! $\endgroup$ – Christophe Nov 7 '18 at 19:55

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.