We have $N$ oscillators and each of them is described by the Hamiltonian:

$$H = \frac{p^2}{2m} + \frac{Kq^4}{4} $$

I have to compute the average total energy $\langle E\rangle$ of the $N$ oscillators. But to do so, first I have to compute the one particle partition function and to do so I have to solve the following integral:

$$Z_1 (V,T) = \iint_{\mathbb{R}^2} e ^{-\beta \,H_1(p,q)}\,dp\,dq.$$

So in this case:

$$ Z_1 = \int_{-\infty}^{\infty} e^{-\beta\frac{p^2}{2m}}\,dp \int_{-\infty}^{\infty} e^{-\beta \frac{Kq^4}{4}}\,dq. $$

I know this integral can be solved by the Gauss method, knowing that:

$$\int_{-\infty}^{\infty} e^{-x^2} dx = \sqrt{\pi}.$$

For the first integral I got:

$$\int_{-\infty}^{\infty} e^{-\beta\frac{p^2}{2m}}\,dp = \sqrt{\frac{2m\pi}{\beta}}$$

I am having difficulties solving the second one:

$$\int_{-\infty}^{\infty} e^{-\beta \frac{Kq^4}{4}}\,dq.$$

I have tried to make a the change of variables: $q^4 = a^2,$ but this does not simplify the calculation. What method should I use?

Once you get $Z_1$:

$$Z_N = (Z_1)^N.$$

Then you just have to apply:

$$\langle E \rangle = -\frac{ \partial \log(Z_N)}{\partial \beta}.$$


$$\langle E \rangle = \frac{3N}{4\beta}.$$

  • $\begingroup$ Wolfram Dev Platform gives an answer of $\Gamma(1/4)/(\sqrt{2}\sqrt[4]{\beta K}).$ $\endgroup$ – Adrian Keister Sep 28 '18 at 13:12
  • 1
    $\begingroup$ For any $n > 0\, \int_0^\infty e^{-x^n}dx = \Gamma(1+\frac{1}{n})$ - perhaps there is an application here given the spiltting of the half-line may be achievable. $\endgroup$ – Kevin Sep 28 '18 at 13:40
  • 1
    $\begingroup$ @JD_PM: I doubt that. Check the definitions in your textbook or notes. $\endgroup$ – Adrian Keister Sep 28 '18 at 13:44
  • 1
    $\begingroup$ @FelixMarin: It wouldn't surprise me if that was textbook-dependent. $\endgroup$ – Adrian Keister Sep 28 '18 at 19:50
  • 1
    $\begingroup$ @AdrianKeister That's right. $\endgroup$ – Felix Marin Sep 28 '18 at 19:57

So, we have \begin{align*} Z_1&=\int_{-\infty}^{\infty} e^{-\beta\frac{p^2}{2m}}\,dp \int_{-\infty}^{\infty} e^{-\beta \frac{Kq^4}{4}}\,dq \\ &=\sqrt{\dfrac{2\pi m}{\beta}} \cdot \frac{\Gamma(1/4)}{\sqrt{2}\sqrt[4]{\beta K}} \\ &=\sqrt{\dfrac{\pi m}{\beta}} \cdot \frac{\Gamma(1/4)}{\sqrt[4]{\beta K}} \\ &=\frac{\sqrt{\pi m}\,\Gamma(1/4)}{\beta^{3/4}\,K^{1/4}} \\ &=\beta^{-3/4}\,\frac{\sqrt{\pi m}\,\Gamma(1/4)}{K^{1/4}}. \end{align*} Next, following your procedure, we do \begin{align*} Z_N&=\left(Z_1\right)^{\,N} \\ &=\left(\beta^{-3/4}\,\frac{\sqrt{\pi m}\,\Gamma(1/4)}{K^{1/4}}\right)^{\!N}, \\ \langle E\rangle&=-\frac{\partial}{\partial\beta} \, \ln\left[\left(\beta^{-3/4}\,\frac{\sqrt{\pi m}\,\Gamma(1/4)}{K^{1/4}}\right)^{\!N}\,\right] \\ &=-N\frac{\partial}{\partial\beta} \, \ln\left[\beta^{-3/4}\,\frac{\sqrt{\pi m}\,\Gamma(1/4)}{K^{1/4}}\right] \\ &=-N\frac{\partial}{\partial\beta} \left[-\frac34\ln(\beta)+\ln\left(\frac{\sqrt{\pi m}\,\Gamma(1/4)}{K^{1/4}}\right) \right] \\ &=\frac{3N}{4\beta}. \end{align*} Here you can see that the $\Gamma(1/4)$ just goes away because of the logarithm.

Lesson for you: Trust the intermediate results (at least temporarily), and keep going!

  • 1
    $\begingroup$ You are right, the definition of < E > has a negative sign! My bad. Now fixed. Thanks for your advise :) $\endgroup$ – JD_PM Sep 28 '18 at 14:21

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.