Phase transitions by means of information theory

Sequential files for observables show different compression capabilities depending on the temperature T at which the simulation is done. Compressed files reach maximum size at a T coinciding with the critical temperature TC; these TCs can be independently determined by well of established methods (Binder cumulants or time autocorrelation functions). This behavior can be explained by information theory: near the critical temperature values for the observable span a large universe and the sequence is chaotic; under TC the system is trapped within few states yielding monotonous sequences of values; over TC values tend to the corresponding thermal noise. Compression is maximal in presence of repeated information which is to be found away from TC. This new method to obtain TC is successfully applied to the Edwards-Anderson model where a ferromagnetic 2D Ising system is progressively changed by randomly introducing antiferromagnetic interactions in concentration x. Analysis is done for classical Monte Carlo simulations on square lattice of different sizes, increasing x, varying T. Possible extensions of this treatment are discussed.