0
$\begingroup$

In this question it is said that:

The best way to numerically work with continuous phase transitions is to study observables that have a vanishing length dimension (or mass dimension in the language of QFT)

I don't understand what that is (I don't know QFT). What is the length dimension in this context and why does $\xi/L$ has a vanishing length dimension in the Ising model?

$\endgroup$
  • 2
    $\begingroup$ Do you know what dimensionless variables are? Because I think that's what you're looking for. $\endgroup$ – Danu May 15 '15 at 18:58
  • $\begingroup$ I assume $\xi$ is the correlation length. So $\xi/L$ is dimensionless. It is generally good to work with dimensionless quantity, not just for the study of phase transitions. $\endgroup$ – Meng Cheng May 15 '15 at 18:58
  • $\begingroup$ Actually, $\xi$ diverges at the phase transition (and thus renormalization group methods work well around phase transitions, because the systems gets scale invariant). $\endgroup$ – Sebastian Riese May 15 '15 at 19:26
  • $\begingroup$ @Danu Well, I know what a dimesionless variable is. I just don't know how that applies. If a quantity is dimensionless, then in a scale invariant system, it should not depend on size. Is that it? Why? And if so, should any dimensionless quantity I could build behave that way? For example, if the susceptibility goes like $\chi\sim L^{\gamma/\nu}$, then $\chi/L^{\gamma/\nu}$ behaves that way although it has dimensions. $\endgroup$ – MyUserIsThis May 15 '15 at 23:23
  • $\begingroup$ @MengCheng I think it's the word length that confused me, sorry. For example the Binder cumulant as defined in the previous question ($\langle m^4\rangle/\langle m^2\rangle^2$) is dimensionless, why should it be better that $\xi/L$ to find the critical point? $\endgroup$ – MyUserIsThis May 15 '15 at 23:26
1
$\begingroup$

Perhaps in order to understand the previous answer you should follow the bellow first:

There are three different behaviours when $N\rightarrow \infty$ of $<\sigma_{0,0},\sigma,_{N,N}>$ for fixed temperature T. whether $T>T_c$, $T=T_c$ or $T<T_c$ the correlation function $<\sigma_{0,0},\sigma,_{N,N}>$ in each of this cases has a different representation those you can find them explicitly here. Now correlation length or $ξ$ is defined as $\ln t = ξ^{-1}$ where one can later use it as scale factor.

$\endgroup$
  • $\begingroup$ Hm, I had $\xi$ defined as $$\xi^2=\frac{\int_0^\infty r^2G(r)}{\int_0^\infty G(r)}$$ But anyway, I know the expression of the correlation function in the three regimes, I just fail to see why I should expect values that don't depend on the size of the system for dimensionless variables. Thank you for answering $\endgroup$ – MyUserIsThis May 15 '15 at 23:34
  • $\begingroup$ it's because your small t is defined to be $t=(T-T_{c})/T_{c}$ it measures the deviation from the critical temperature in a dimensionless way, and hence the correlation length diverges as a negative power of t namely $ξ=|t|^{-\nu}$ as $T\rightarrow T_{c}$ where $\nu$ is the critical exponent. $\endgroup$ – William May 16 '15 at 3:19

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.