Probing phase transitions with correlations in configuration space: a Monte Carlo study on lattice models

While phases and phase transitions are mostly characterized by order parameters or physical quantities in real space, we propose that the correlation in Hilbert space is closely connected to phase transitions. Specifically, this correlation is quantified by the 1-norm distance between configurations, and the distribution of distances can be obtained from a small fraction of configurations in the exponentially large Hilbert space, with the help of importance sampling procedures of the Monte Carlo method. Considering the obtained distances as a data set, its distribution varies substantially in different phases, and the numerical results further suggest a universal critical behavior for the uncertainty and participation entropy extracted from it. For various classical spin models with different types of phases and phase transitions, the finite-size analysis based on these quantities successfully catches the phase transitions with accurate critical points. Moreover, in all cases for different systems and phase transitions, the critical exponent from the uncertainty of the distances is found to numerically equal the anomalous dimension that determines the decay of the correlation in real space. This implies a deeper connection between the correlation in the real space and configuration space, which deserves further investigation. In our proposal, the definition of distance covers various lattice models with different local degrees of freedom, e. g. , two levels for Ising-like models, discrete multi-levels for q-state clock models, and continuous local levels for the XY model, and the way handling the distances is very simple and robust. Our proposal provides an alternative way of understanding the complex phases and phase transitions in complicated systems where the order parameter is hard to compute or define.