T-2: Difference between revisions

From Disordered Systems Wiki
Jump to navigation Jump to search
No edit summary
Line 53: Line 53:
f= -\lim_{N \to \infty} \lim_{n \to 0}\frac{1}{\beta N n} \frac{\overline{Z^n}-1}{n}.
f= -\lim_{N \to \infty} \lim_{n \to 0}\frac{1}{\beta N n} \frac{\overline{Z^n}-1}{n}.
</math></center>
</math></center>
Therefore,
Therefore, to compute the free-energy we need to compute the moments <math>{\overline{Z^n}}</math> and then take the limit  <math>n \to 0</math>. This calculation can be done in 3 main steps.


<ol>
<ol>
<li> <em> Heavy tails and concentration.</em> ccc</li>
<li> <em> Step 1: average over the disorder.</em> ccc</li>
</ol>
</ol>
<br>
<br>


<ol start="2">
<ol start="2">
<li><em> Inverse participation ratio.</em> cccc
<li><em> Step 2: identify the order parameter.</em> cccc
</li>
</ol>
<br>
 
<ol start="3">
<li><em> Step 3: the saddle point.</em> cccc
</li>
</li>
</ol>
</ol>
<br>
<br>

Revision as of 23:10, 6 December 2023

In this set of problems, we use the replica method to study the equilibrium properties of a prototypical toy model of glasses, the spherical p-spin model. In the spherical p-spin model the configurations σ=(σ1,,σN) that the system can take satisfy the spherical constraint i=1Nσi2=N, and the energy associated to each configuration is

E(σ)=1i1i2ipNJi1i2ipσi1σi2σip,

where the coupling constants Ji1i2ip are independent random variables with Gaussian distribution with zero mean and variance p!/(2Np1), and p3 is an integer.



Problem 1: the annealed free energy

In TD1, we defined the quenched free energy density as the quantity controlling the scaling of the typical value of the partition function Z. The annealed free energy fann instead controls the scaling of the average value of Z. It is defined by

fann=limN1βNlogZ.

Let us compute this quantity.


  1. Energy correlations. At variance with the REM, in the spherical p-spin the energies at different configurations are correlated. Show that E(σ)E(τ)=Nq(σ,τ)p/2+o(1), where q(σ,τ)=1Ni=1Nσiτi is the overlap between the two configurations. Why an we say that for p this model converges with the REM discussed in the previous TD?


  1. Energy contribution. Show that computing Z boils down to computing the average eβJi1ipσi1σip. Compute this average. Hint: if X is a centered Gaussian variable with variance σ2, then eαX=eα2σ22.


  1. Entropy contribution. The volume of a sphere of radius N in dimension N is given by NN2πN2/(N2)!. Use the large-N asymptotic of this to conclude the calculation of the annealed free energy: the final result is only slightly different with respect to the free-energy of the REM in the high-temperature phase: can you identify the source of this difference?


Problem 2: the replica trick and the quenched free energy

The quenched free energy density is the quantity controlling the scaling of the typical value of the partition function Z. This means that

f=limN1βNlogZ.

This formula differs from the one above by the order in which the logarithm and the average over disorder are taken. Computing the average of the logarithm is in general a hard problem, which one can address by using a smart representation of the logarithm, that goes under the name of replica trick:

logx=limn0xn1n

which can be easily shown to be true by Taylor expanding xn=enlog(x)=1+nlogx+O(n2). Applying this to the average of the partition function, we see that

f=limNlimn01βNnZn1n.

Therefore, to compute the free-energy we need to compute the moments Zn and then take the limit n0. This calculation can be done in 3 main steps.

  1. Step 1: average over the disorder. ccc


  1. Step 2: identify the order parameter. cccc


  1. Step 3: the saddle point. cccc