0
$\begingroup$

I have found in the literature that the critical dynamical exponent $z$ of an Ising model simulated with a local algorithm (such as Metropolis) is something around 2 near the critical temperature, where the critical slowing-down occurs and, luckily, that is also true for my personal simulations near the critical temperature 2.269..?

Still, when I go at temperature higher than the critical one, I find an higher exponent z. Is that normal? And if so, why's that?

$\endgroup$
2
  • $\begingroup$ It seems to me that there is no reason anymore for the autocorrelation time to increase algebraically with the lattice size as $L^z$ outside the critical point. I would expect that it saturates at some limiting value in the paramagnetic phase. $\endgroup$
    – Christophe
    Feb 22, 2019 at 9:21
  • $\begingroup$ @Christophe hi and thanks for the reply! The autocorrelation time should actually decrease while going away from the critical region to higher temperatures so yeah, I agree! And what about z in your opinion? $\endgroup$ Feb 22, 2019 at 10:31

1 Answer 1

1
$\begingroup$

Like any other critical exponent, the dynamical exponent $z$ can be extracted from the scaling of the autocorrelation time $\tau$ with

  • the reduced temperature as $\tau\sim |T-T_c|^{-z}$ for $T$ sufficiently close to the critical temperature
  • the lattice size $L$ as $\tau\sim L^z$ only at $T=T_c$.

Away from the critical point, the time $\tau$ does not diverge anymore. Therefore, there is no point in fitting $\tau$ as $\tau\sim L^z$ for $T\ne T_c$ in the Ising model. Only in a critical phase (as in the 2D-XY model), or in glasses, there exists a temperature-dependent dynamical exponent.

$\endgroup$
1
  • $\begingroup$ What I tought, thank you very much! Any book that discusses this? $\endgroup$ Feb 23, 2019 at 1:27

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.