04.08.2013 Views

Statistical Mechanics - Physics at Oregon State University

Statistical Mechanics - Physics at Oregon State University

Statistical Mechanics - Physics at Oregon State University

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

8.5. CRITICAL TEMPERATURE IN DIFFERENT DIMENSIONS. 177<br />

In the previous sections we have derived the same result in two different<br />

ways, and therefore have shown th<strong>at</strong> these two approxim<strong>at</strong>ions are equivalent.<br />

In Mean Field Theory we replace the quadr<strong>at</strong>ic coupling terms in the Hamiltonian<br />

by the product of an average and a variable. This is equivalent to coupling<br />

each spin to the average field due to the neighbors. In the Bragg-Williams<br />

approxim<strong>at</strong>ion we replace the density m<strong>at</strong>rix by a form th<strong>at</strong> does not contain<br />

coupling between neighboring sites. As a result, in the Hamiltonian again the<br />

average field of the neighbors plays a role only.<br />

8.5 Critical temper<strong>at</strong>ure in different dimensions.<br />

In the mean-field solution of the Ising model the critical temper<strong>at</strong>ure does not<br />

explicitly depend on the dimensionality of the problem. Indirectly it does, of<br />

course, because the possible values of q depend on the number of dimensions.<br />

For a mono-<strong>at</strong>omic, periodic l<strong>at</strong>tice in one dimension q has to be two, in two<br />

dimensions q can be two, four, or six, and in three dimensions q can be two,<br />

four, six, eight, or twelve. Hence the higher the number of dimensions, the<br />

larger the critical temper<strong>at</strong>ure can be. For closed packed systems the number<br />

of neighbors is maximal, and the critical temper<strong>at</strong>ure would increase according<br />

to the p<strong>at</strong>tern 2:6:12 going from one to three dimensions. This is also true<br />

in general. When the number of neighbors increases, the total strength of the<br />

interactions driving an ordered st<strong>at</strong>e increases and the ordered st<strong>at</strong>e can persist<br />

up to a higher temper<strong>at</strong>ure.<br />

The mean-field solution of the Ising model is only an approxim<strong>at</strong>ion. In<br />

general, the errors in the mean-field solutions become larger when the number<br />

of dimensions decreases. In one dimension, for example, there are no phase<br />

transitions in the Ising model! Wh<strong>at</strong> causes the error in mean-field theory?<br />

In essence, we did something dangerous by interchanging two limits. We first<br />

assumed th<strong>at</strong> N → ∞ in order to deduce th<strong>at</strong> 〈Si〉 is the same for all positions<br />

i. Or th<strong>at</strong> ρi is independent of i. Then we used this common value to solve the<br />

equ<strong>at</strong>ions for finite values of N. The results made of this procedure made sense<br />

in the thermodynamic limit, since m did not depend on N. The error is actually<br />

in the first step. We have to solve the whole problem <strong>at</strong> finite N and calcul<strong>at</strong>e<br />

mi(N, T ). In this expression we should take the limit N → ∞. The results will<br />

be different in th<strong>at</strong> case. We cannot interchange these limits!<br />

There is an old argument, due to Landau (who else?), th<strong>at</strong> the Ising chain<br />

does not show a phase transition. Consider a finite chain of N sites, numbered<br />

one through N sequentially. Assume th<strong>at</strong> h = 0. There are two completely<br />

ordered st<strong>at</strong>es, either all spins are up or down. Next we consider all st<strong>at</strong>es<br />

which include a so-called Bloch wall. The first m spins have one direction, while<br />

the remaining N − m spins have the opposite direction. The number of these<br />

st<strong>at</strong>es is 2(N −1). The energy of a completely ordered chain is E0 = −J(N −1).<br />

The energy of a Bloch wall st<strong>at</strong>e is different only due to the contribution of the<br />

two spin variables <strong>at</strong> opposite sites of the Bloch wall. Therefore the energy of<br />

a Bloch wall st<strong>at</strong>e is −J(N − 2) + J = E0 + 2J. At a fixed temper<strong>at</strong>ure and <strong>at</strong>

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!