The BEG model at the FAD triple point on the square lattice

Technical Summary: Uniqueness of the Gibbs Measure for the BEG Model at the FAD Triple Point

Problem Statement
The Blume-Emery-Griffiths (BEG) model is a spin system on a $d$-dimensional lattice with spins $\sigma_x \in \{-1, 0, +1\}$, governed by a Hamiltonian involving nearest-neighbor interactions and quadratic terms. The phase diagram is divided into three regions based on ground states: Disordered (D), Ferromagnetic (F), and Antiquadrupolar (A). These regions meet at the Ferromagnetic-Antiquadrupolar-Disordered (FAD) triple point, located at parameters $(X, Y) = (0, -1)$.

At the FAD point, the model exhibits an infinite number of ground states that cannot be organized into a finite number of equivalence classes, resulting in high residual entropy. While rigorous results exist for the model in the bulk regions (using Pirogov-Sinai theory) and at high temperatures, the behavior along the phase boundaries and specifically at the FAD point has remained elusive. Previous rigorous results for phase transitions along the AD line or at the FAD point were limited to sufficiently high dimensions ($d$). For the two-dimensional square lattice ($d=2$), no rigorous analysis existed for non-zero temperatures, leaving open the possibility of a phase transition at positive inverse temperature $\beta$.

Methodology
The authors establish the absence of phase transitions by proving the uniqueness of the Gibbs measure for the BEG model on the 2D square lattice at the FAD point for all temperatures. The proof relies on a specific graphical representation and coupling technique:

1. Random-Cluster Representation: The authors construct a random-cluster type representation for the BEG model at the FAD point. This generalizes the Fortuin-Kasteleyn (FK) representation used for the Potts model. The partition function is rewritten using bond configurations $\omega \in \{0,1\}^E$ where an edge is "open" ($\omega_e=1$) if the associated spin product $\sigma_e \neq -1$.
2. Coupling with Zero-Temperature Model: A joint probability measure is defined on the product space of spin configurations and bond configurations. This allows the finite-temperature expectation of any observable to be expressed as an expectation over the bond configurations of the zero-temperature (ground state) BEG model on the subgraph induced by the open bonds. Specifically, the magnetization is related to a connectivity event in the zero-temperature system.
3. Magnetization-Connectivity Identity: The authors derive an identity showing that the magnetization at a site $x$ under positive boundary conditions is bounded by the probability that $x$ is connected to the boundary via a path of non-zero spins in the zero-temperature model.
4. Stochastic Domination by Percolation: Leveraging previous results (specifically from [12]), the authors show that the zero-temperature BEG model at the FAD point can be coupled with Bernoulli site percolation with parameter $p=1/2$. This establishes that the probability of a spin connecting to the boundary in the zero-temperature model is stochastically dominated by the connectivity probability in Bernoulli site percolation with parameter $1/2$.

Key Results
The main theorem states that for the BEG model on $\mathbb{Z}^2$ at the FAD point and any inverse temperature $\beta \in [0, +\infty]$, the mean spin at the origin vanishes under both $+$ and $-$ boundary conditions:
$$\langle \sigma_x \rangle^+_\beta = \langle \sigma_x \rangle^-_\beta = 0$$
The proof proceeds by showing that the magnetization is bounded above by the probability of a site being connected to the boundary in Bernoulli site percolation with parameter $1/2$. Since the critical parameter for site percolation on the square lattice is $p_c = 1/2$, the system is at the critical point (or subcritical in the sense that infinite clusters do not form almost surely in the thermodynamic limit for $p \leq 1/2$ in this context), causing the connection probability to vanish as the volume goes to infinity.

Combined with the validity of FKG inequalities for this model (as established by Lebowitz and Monroe), the equality of local magnetizations under different boundary conditions implies the independence of all correlation functions from boundary conditions. Consequently, the Gibbs measure is unique for all temperatures.

Significance and Scope
The paper establishes the absence of phase transitions for the BEG model at the FAD triple point on the 2D square lattice at any temperature. This resolves the question of whether a phase transition could exist at positive $\beta$ in this specific regime.

The authors note that their argument relies on the fact that the critical parameter for Bernoulli site percolation on the graph is greater than or equal to $1/2$. This result extends to any infinite graph $G$ where $p_c(\text{site}, G) \geq 1/2$. However, the authors explicitly state that for dimensions $d \geq 3$, where $p_c(\text{site}, \mathbb{Z}^d) < 1/2$, the upper bound provided by $p=1/2$ percolation becomes supercritical. In these higher dimensions, the magnetization does not necessarily vanish, and Monte Carlo simulations suggest a non-trivial phase transition may exist. The paper concludes that proving uniqueness or non-uniqueness in intermediate dimensions remains an open problem requiring new ideas beyond the current stochastic domination approach.