Phase transitions in generalized XY models

Proving Phase Transitions in Generalized XY Models via Height Function Duality

Numerical Monte Carlo simulations have long suggested that certain magnetic materials undergo complex phase transitions. However, these simulations only offer hints. They lack the rigorous mathematical proofs required to confirm exactly when these transitions occur. This gap is particularly evident in generalized XY models, which describe systems of two-dimensional spins (vectors that can rotate freely in a plane).

Real-world systems often involve more complex interactions than the standard XY model. Many materials exhibit both ferromagnetic tendencies (favoring parallel alignment) and nematic tendencies (favoring either parallel or anti-parallel alignment). Predicting how these competing forces create new phases of matter is a difficult task. Researchers have struggled to move beyond numerical observations to achieve formal mathematical certainty.

The Gap Between Simulation and Rigor

The difficulty in studying generalized XY models lies in their complex phase diagrams. In a standard XY model, the system typically undergoes a Berezinskii–Kosterlitz–Thouless (BKT) transition. This involves moving from a disordered state to a state with slow, power-law decay of correlations. However, adding nematic interactions introduces a third regime: the nematic phase. In this phase, the system lacks traditional ferromagnetic order. Instead, it maintains nematic order, where spins align along a common axis even if they point in opposite directions.

As shown in the schematic phase diagram in, the interplay between the coupling parameter $\Delta$ and the inverse temperature $\beta$ creates a competitive landscape.

The disordered, nematic, and BKT phases all vie for dominance. Until recently, proving the existence of this intermediate nematic region was elusive. Proving that the nematic order parameter ($G_2$) does not decay exponentially requires a new way to connect spin dynamics to a manageable mathematical structure.

Mapping Spins to Heights

To bridge this gap, Fabio Plaga employs a technique called height function duality. The strategy is to move the problem from the "primal" spin model to a "dual" height function model. Instead of tracking individual spinning vectors, the researcher tracks a "height" assigned to the faces of the graph. Fluctuations in this height function act as a proxy for the correlations between the original spins.

The mechanism proceeds in several stages:

1. Random Current Expansion: The author introduces an enhanced random current expansion. This method decomposes the Hamiltonian (the function representing the total energy) into a sum of weights over "currents." These are mathematical objects representing the flow of interaction between vertices.
2. Defining the Dual Model: These currents define a height function $h$ on the dual graph. The height difference between adjacent faces relates to the net flow of these currents.
3. Generalized Loop Representation: The author develops a specialized loop representation. This links the variance of the height difference—how much the "surface" fluctuates—directly to the nematic order parameter $G_2$.
4. Connecting Delocalisation to Order: The crucial logical step connects "delocalisation" to order. If the height function delocalises (meaning its fluctuations grow without bound as the system size increases), then nematic order cannot decay exponentially.

By translating spin alignment into a problem of surface roughness, the author leverages existing theorems regarding height model stability.

Evidence of a Nematic Regime

The paper provides rigorous proof for the phases predicted by earlier simulations. Theorem 1 establishes that whenever the dual height function model is delocalised, the nematic order parameter $G_2$ is guaranteed to avoid exponential decay. This confirms that the "nematic" region in the phase diagram is a mathematically distinct state.

For the specific case of the $H_\Delta$ Hamiltonian on a square lattice ($\mathbb{Z}^2$), the author provides a quantitative lower bound. The paper reports that for sufficiently low $\Delta$ and high inverse temperature $\beta$, the nematic order parameter $G_2(v)$ is at least $\frac{1}{8|v|}$. This is a significant quantitative improvement over previous results by Pfister [19]. For a researcher, this means the order is not just "not exponential," but follows a specific, predictable power-law decay.

To validate this, the author compares the generalized XY model to the Ashkin–Teller model. As shown in, the phase diagram of the Ashkin–Teller model resembles the expected diagram for the generalized XY model.

By using the Ashkin–Teller model as a baseline, the author places rigorous bounds on the correlation functions of the generalized model.

Limitations of the Formalism

There are boundaries to what this mathematical framework achieves. The author notes that certain powerful tools, like the Messager–Miracle–Sole (MMS) and Simon–Lieb inequalities, are not universally applicable to all Hamiltonians in this class. This means the "sharpness" of the transition—whether other types of decay exist besides exponential and power-law—cannot be guaranteed for every model variation.

Additionally, the proof of the nematic regime (Theorem 2) relies on a comparison argument with the Ashkin–Teller model. This approach works well for the $H_\Delta$ family of Hamiltonians. However, it may not immediately generalize to more exotic Hamiltonians where the relationship to the Ashkin–Teller model is less direct. While the "bridge" between height functions and spins is robust, applying it to every possible interaction remains an open question.

The Verdict

This research provides a definitive mathematical foundation for a phenomenon previously supported only by simulation. By linking the delocalisation of a dual height function to the non-exponential decay of nematic order, Plaga turns a heuristic observation into a rigorous law.

For researchers in condensed matter physics or statistical mechanics, this work offers a vital analytical tool. It confirms that the "nematic" phase is a fundamental consequence of competing symmetries. The ability to provide specific lower bounds like $\frac{1}{8|v|}$ gives practitioners a benchmark to test against their own numerical models. Whether this approach can scale to three-dimensional systems remains the next frontier.

Figures from the paper