Critical energy density of O$(n)$ models in $d=3$

TL;DR

This study investigates the critical energy densities of O(n) models in three dimensions, comparing them to the Ising model, and finds that their differences are very small for n<8, supporting a conjectured relation between these models.

Contribution

The paper provides new Monte Carlo estimates of critical energy densities for O(n) models in 3D and introduces an interpolation formula quantifying their similarity to the Ising model.

Findings

Critical energy densities for O(2), O(3), and O(4) models are precisely determined.

Differences between O(n) and Ising models' critical energies are less than 1% for n<8.

An interpolation formula shows the small variation of critical energies across different n.

Abstract

A relation between O$(n)$ models and Ising models has been recently conjectured [L. Casetti, C. Nardini, and R. Nerattini, Phys. Rev. Lett. 106, 057208 (2011)]. Such a relation, inspired by an energy landscape analysis, implies that the microcanonical density of states of an O$(n)$ spin model on a lattice can be effectively approximated in terms of the density of states of an Ising model defined on the same lattice and with the same interactions. Were this relation exact, it would imply that the critical energy densities of all the O$(n)$ models (i.e., the average values per spin of the O$(n)$ Hamiltonians at their respective critical temperatures) should be equal to that of the corresponding Ising model; it is therefore worth investigating how different the critical energies are and how this difference depends on $n$. We compare the critical energy densities of O$(n)$ models in three dimensions in some specific cases: the O$(1)$ or Ising model, the O$(2)$ or $XY$ model, the O$(3)$ or Heisenberg model, the O$(4)$ model and the O$(\infty)$ or spherical model, all defined on regular cubic lattices and with ferromagnetic nearest-neighbor interactions. The values of the critical energy density in the $n=2$, $n=3$, and $n=4$ cases are derived through a finite-size scaling analysis of data produced by means of Monte Carlo simulations on lattices with up to $128^3$ sites. For $n=2$ and $n=3$ the accuracy of previously known results has been improved. We also derive an interpolation formula showing that the difference between the critical energy densities of O$(n)$ models and that of the Ising model is smaller than $1\%$ if $n<8$ and never exceeds $3\%$ for any $n$.