I have the quantum anharmonic oscillator:

$$H = \frac{\hat{p}^2}{2m} + \frac{1}{2}m\omega^2 \hat{x}^2 + \frac{\lambda}{4!}\hat{x}^4$$

and I want to find the free energy (so essentially $\log Z(\beta)$) up to 2nd order in $\lambda$ using Feynman diagrams.

I don't know how to do this. I know that the logarithm means one only needs to look at connected diagrams. The first order would be

$$A_1 = 3 \left(-\frac{\lambda}{\hbar}\right) \frac{1}{4!} \int_0^{\hbar \beta} d\tau \left< x^2(\tau)\right>_0 \left< x^2(\tau)\right>_0 $$

and since $\left< x^2(\tau)\right>_0 = \frac{\hbar}{m}G(0)$, where $G$ is the propagator, then

$$A_1 = 3 \left(-\frac{\lambda}{\hbar}\right) \frac{1}{4!} \hbar \beta \left(\frac{\hbar}{m}G(0)\right)^2 $$

For the second order there's two kinds of connected diagrams, with combinatorial factors $2{4\choose 2}{4\choose 2}=72$ and $4\cdot 3 \cdot 2 =24$. So there's two terms in this order:

$$A_2^1 = 72 \frac{1}{2!}\left(-\frac{\lambda}{\hbar}\right)^2 \frac{1}{4!^2} \int_0^{\hbar \beta} d\tau \int_0^{\hbar \beta} d\tau' \left[ \left< x(\tau)x(\tau') \right>_0 \right]^2\left< x^2(\tau)\right>_0 \left< x^2(\tau')\right>_0 $$

$$A_2^2 = 24 \frac{1}{2!}\left(-\frac{\lambda}{\hbar}\right)^2 \frac{1}{4!^2} \int_0^{\hbar \beta} d\tau \int_0^{\hbar \beta} d\tau' \left[ \left< x(\tau)x(\tau') \right>_0 \right]^4$$

which involve double integrals of powers of the propagator. For the harmonic oscillator, I found that it's given by a very complicated expression:

$$G(\tau - \tau') = \frac{\cosh(\hbar\beta\omega/2 - \omega |\tau - \tau'|)}{2\omega \sinh (\hbar \omega \beta /2)}$$

All in all, the free energy to second order would be $$F = -kT(1+A_1 + A_2^1 + A_2^2)$$

Now it seems that to get the $A_2^i$s I need to integrate $G^2(\tau - \tau')$ and $G^4(\tau - \tau')$ in order to get the result, and I'd really like to know that this is the right way to go before I even attempt it. The integrals look absolutely hellish so even if this is right, how do I calculate them? I don't have access to mathematica so I don't even know if they give a reasonable result.


If the expression for the propagator you provided is correct, the integral is not daunting as it may at first seem. We essentially are faced with integrating,

$$\int_0^c d\tau \int_0^c d\tau' \, \cosh^n(a-b|\tau-\tau'|)$$

for $a,b, c>0$ and $n \in \mathbb Z$. I will demonstrate the $n=2$ case. We can expand the integrand as,

$$\cosh^2(a-b|\tau-\tau'|) = \frac12 + \frac14 e^{2a} e^{-2b|\tau-\tau'|} + \frac14e^{-2a}e^{2b|\tau-\tau'|}.$$

Thus, the only tricky part is figuring out how to integrate,

$$\int_0^c d\tau \int_0^c d\tau' \, e^{\gamma |\tau-\tau'|}, \quad \gamma \in \mathbb R.$$

We see we are integrating over a square $[0,c]\times[0,c]$. We can split this square diagonally into two regions, the top triangle with $\tau-\tau' <0$ and the bottom half with $\tau-\tau' >0$, and then we can sum the contributions. For the bottom half, we would have,

$$\int_0^c d\tau \int_0^\tau d\tau' \, e^{\gamma(\tau-\tau')} = \frac{1}{\gamma}\int_0^c d\tau \, (e^{\gamma \tau} -1) = \frac{1}{\gamma^2}(e^{\gamma c} - \gamma c - 1).$$

Now for the top region, $|\tau-\tau'| = \tau'-\tau$ and we have,

$$\int_0^c d\tau' \int_0^{\tau'} d\tau \, e^{\gamma(\tau'-\tau)} = \frac{1}{\gamma^2}(e^{\gamma c} - \gamma c - 1)$$

and so we have that,

$$\int_0^c d\tau \int_0^c d\tau' \, e^{\gamma |\tau-\tau'|} = \frac{2}{\gamma^2}(e^{\gamma c} - \gamma c - 1).$$

Going back to the original $n=2$ integral, using this result and simplifying yields,

$$\int_0^c d\tau \int_0^c d\tau' \, \cosh^2(a-b|\tau-\tau'|) = \frac{1}{4b^2} \left[ \cosh(2(a-bc)) - \cosh 2a + 2bc(bc+\sinh 2a)\right].$$

It is tempting to now generalise the result for all even powers. Using the generalised binomial theorem and the expansion of $\cosh$ in terms of exponentials, we have,

$$\cosh^{2n}(x) = \frac{1}{2^{2n}}\sum_{k=0}^{2n} \binom{2n}{k} e^{(2n-2k)x} = \frac{1}{2^{2n}} \binom{2n}{n} + \sum_{k\neq n} \binom{2n}{k} e^{(2n-2k)x}.$$

Hence the integral can be written as,

$$\int_0^c d\tau \int_0^c d\tau' \, \cosh^{2n}(a-b|\tau-\tau'|) = \frac{c^2}{2^{2n}}\binom{2n}{n}+\sum_{k\neq n} \binom{2n}{k} \int_0^c d\tau \int_0^c d\tau' \, e^{2(n-k)(a-b|\tau-\tau'|)}.$$

This integral we already know how to compute:

$$\int_0^c d\tau \int_0^c d\tau' \, e^{2(n-k)(a-b|\tau-\tau'|)} = e^{2a(n-k)}\int_0^c d\tau \int_0^c d\tau' \, e^{2b(k-n)|\tau-\tau'|}$$ $$=\frac{1}{2b^2(k-n)^2}e^{2a(n-k)} \left[ e^{2bc(k-n)} + 2bc(n-k)-1\right].$$

Thus, the integral reduces to the finite sum from $k=0,\dots, 2n$ (without $k=n$), as

$$\frac{c^2(2n)!}{2^{2n} (n!)^2} + \frac{1}{2b^2}\sum_{k \neq n} \frac{1}{(k-n)^2}\binom{2n}{k}e^{2a(n-k)}\left[ e^{2bc(k-n)} + 2bc(n-k)-1\right].$$

Returning the propagator, plugging in the constants, the integral of $G^{2n}(\tau-\tau')$ is the finite sum,

$$\boxed{\frac{(2n)!}{(2\omega)^{2n}\sinh^{2n} \hbar \omega \beta /2} \left[ \frac{\hbar^2 \beta^2}{2^{2n}(n!)^2 } + \frac{1}{2\omega^2} \sum_{k\neq n}\frac{1}{(k-n)^2(2n-k)! k!} e^{\hbar\omega \beta (n-k)} \left( e^{2\hbar\omega\beta(k-n)}+2\hbar\omega\beta(n-k)-1\right)\right]}$$

  • $\begingroup$ Thanks a lot. So I take it that my solution is otherwise correct? $\endgroup$ Dec 18 '16 at 21:58
  • $\begingroup$ also i think there's a power of 2n missing in the sinh in the final expression $\endgroup$ Dec 18 '16 at 22:21
  • $\begingroup$ @SpineFeast You mean $2\omega$? I pulled that into the brackets. $\endgroup$
    – JamalS
    Dec 18 '16 at 22:25
  • $\begingroup$ isnt the sinh raised to 2n since its the denominator of $G^{2n}$ ? To be honest I'm not seeing the $\omega^{2n}$ either $\endgroup$ Dec 18 '16 at 22:32
  • $\begingroup$ @SpineFeast Yes, I forgot. I'll fix it. $\endgroup$
    – JamalS
    Dec 18 '16 at 22:33

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.