Sometimes in Monte Carlo simulations we need to compute the correlation length, but this is a hard task without a formula. However, for instance, in an periodic cubic lattice of $L^3$ spins, some papers present a formula of the correlation length as follows:

$$ \xi = \frac{1}{2 \sin(\pi/L)} {\left (\frac{\chi}{F}-1 \right)}^{1/2} $$

where $\chi$ is the susceptibility, which is $\hat{G}(0)$ (i.e. the Fourier transformed two-point correlation function at $\vec{k} = (0,0,0)$), and $F$ is $\hat{G}(\vec{k}_{min})$, where $\vec{k}_{min} = (2\pi/L,0,0)$. My guess is that it comes from the following definition $$ {\xi}^2 = \frac{\sum_{\vec{r}} r^2 G(\vec{r})}{\sum_{\vec{r}}G(\vec{r})}, $$

but I can't get to the formula.


1 Answer 1


To me this looks like a rearranged version of the Ornstein-Zernike form of the Fourier-transformed correlation function, which reads $$ \hat{G}(k) = \frac{\hat{G}(0)}{1+(k\xi)^2} $$ for any sufficiently small $k$. Obviously, this form applies to an isotropic system rather than a cubic lattice.

The assumptions behind that equation are discussed in this question and my answer to it; both the question and the answer give further pointers to the literature. Your equation reduces to this one, if we make the assumption that $k=k_\text{min}$ is small, so that $\sin\frac{1}{2}k_\text{min}\approx \frac{1}{2}k_\text{min}$. (I note that the lattice spacing is taken to be unity, making both $L$ and $k$ dimensionless).

I guess that the lattice structure of the problem is the reason your equation involves a $\sin$ function, while mine does not, but I don't know enough about the context to say for sure.

In any case, I hope that this sets you along the right lines.

  • $\begingroup$ Hi LonelyProf. Thank you for your thoughts. The fomula was used in computing the spin-glass correlation length in the cubic lattice. Nevertheless, I have also seen the same formula for the two-dimensional Ising model. For instance, this is shown in Eq. (2.8) in Journal of Statistical Physics, Vol. 98, Nos. 3/4, 2000 by J. Salas et al. $\endgroup$
    – O. Daniel
    Commented Jun 6, 2019 at 1:45
  • $\begingroup$ ....the appearance of that $\sin(\pi/L)$ is still a mystery for me. $\endgroup$
    – O. Daniel
    Commented Jun 6, 2019 at 1:52
  • $\begingroup$ I see this formula in earlier papers e.g. Ballesteros et al J Phys A, 32, 1 (1999) and references therein. Clearly it is a standard one which the specialists should recognize; some of those people are active on this site for sure, so you should get a better answer from one of them before long, if those papers are not enough to track it down. $\endgroup$
    – user197851
    Commented Jun 6, 2019 at 10:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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