Nature of the phase-transition line of the three-dimensional Ashkin–Teller model: non-universality, wide crossover, or emergent-O(2) universality?

The Ashkin–Teller model — two Ising spins per lattice site coupled by a nearest-neighbor four-spin interaction — is a canonical model of coupled order parameters. In two dimensions it is exactly understood, with critical exponents varying continuously along its critical line; in three dimensions the nature of its phase-transition line remains unsettled. Monte Carlo studies have variously reported Ising-like transitions, tricritical points, possible non-universal exponent variation near tricritical points, a wide crossover covering the whole line between tricritical points in the negative-coupling region, and an uncommon coexistence of continuous and first-order transitions on a single phase boundary, while hierarchical-lattice renormalization-group calculations give a qualitatively distinct picture in which first- and second-order boundaries are separated by tricritical and critical endpoints. Recent large-scale Monte Carlo evidence for emergent O(2) symmetry with XY exponents along the paramagnetic–(Baxter/mixed) critical line, together with the field-theoretic weak irrelevance of the cubic anisotropy, narrows but does not settle the question. The question remains open: no published work reconciles the Monte Carlo and renormalization-group phase diagrams or establishes the universality class of every continuous segment.

Background

The Ashkin–Teller (AT) model places two Ising spins, $\sigma_i=\pm1$ and $\tau_i=\pm1$, on every site of a lattice and couples them through a nearest-neighbor two-spin coupling $J$ and a four-spin coupling $U$:

$$\mathcal{H}=-J\sum_{\langle ij\rangle}\left(\sigma_i\sigma_j+\tau_i\tau_j\right)-U\sum_{\langle ij\rangle}\sigma_i\tau_i\sigma_j\tau_j,$$

so the model interpolates between two independent Ising models (at $U=0$) and the four-state Potts model (at $J=U$; the Potts model is the $q$-state generalization of the Ising model, here with $q=4$), and its physics is organized by the two-parameter plane $(J,U)$.

In two dimensions the AT model is one of the rare interacting systems with an exactly understood critical line along which the critical exponents vary continuously with the couplings — explicit non-universality, which follows from Baxter's (1972) exact solution of the equivalent eight-vertex model (a two-dimensional lattice model whose degrees of freedom live on the edges, to which the planar AT model maps exactly) and whose critical properties were analyzed by Fan (1972). A natural, long-standing question is what survives of this behavior in three dimensions.

On the simple cubic lattice the model has (at least) three phases: the paramagnetic phase; the Baxter phase, in which both spin components order ferromagnetically; and the mixed phase — discovered by Ditzian, Banavar, Grest and Kadanoff (1980) and absent in two dimensions — in which, for negative four-spin coupling, one component orders ferromagnetically while the other orders antiferromagnetically. The order parameter therefore has up to three independently behaving components, and the phase boundaries can be continuous or first-order. A tricritical point is a point where a line of continuous transitions meets a first-order line; a critical endpoint is a second-order point terminating on a first-order line; a bifurcation point is where phase boundaries split as an additional ordering channel turns on. By a universality class one means the set of models sharing the same critical exponents and scaling functions; a symmetry is called emergent when it appears at criticality even though the Hamiltonian does not possess it. Field-theoretically, two coupled Ising order parameters are described by an $O(2)$-symmetric $\phi^4$ theory supplemented by a cubic anisotropy (a $D_4$-symmetric term allowed by the microscopic $Z_2\times Z_2$ symmetry); the general theory of such multicomponent $O(N)$ $\phi^4$ models is reviewed by Pelissetto and Vicari (2002).

The three-dimensional situation is contested. Monte Carlo work by Musiał, Wojtkowiak and co-workers confirmed Ising-like continuous transitions along two lines (2002, 2008), located tricritical points and a bifurcation point (2004, 2022), at one point announced possible non-universal exponent variation along the transition line near tricritical points (2006), and later concluded — for the paramagnetic–mixed–Baxter transitions in the negative-coupling region — that a wide crossover covers the whole line between tricritical points, reporting coexistence of continuous and first-order transitions on the same phase boundary for different components of the order parameter (2019). Renormalization-group (RG) calculations that are exact on a hierarchical lattice (a self-similar lattice on which real-space RG recursions can be evaluated exactly, taken as an approximation to the cubic lattice) and approximate on the cubic lattice via the improved Migdal–Kadanoff scheme (a real-space bond-moving approximation) give instead a d=3 phase diagram in which first- and second-order boundaries are separated by tricritical and critical endpoints (Keçoğlu and Berker, 2023). Very recently, embedding-cluster Monte Carlo simulations (an efficient cluster-update Monte Carlo method) of very large systems found the distribution of the two-component magnetization vector to display emergent O(2) symmetry along the critical line separating the paramagnetic phase from the Baxter and mixed phases, with exponents close to the three-dimensional XY values (Zhang, Hu, Sun and Lv, 2025), and a field-theoretic Monte Carlo study of the two-component $\phi^4$ theory with cubic perturbation found that perturbation to be irrelevant but only weakly so, with RG exponent $Y_4=-0.1118(10)$, implying slow flows and wide crossover (Hasenbusch, 2025). These three lines of work have not been reconciled, and the unresolved target is the definitive characterization of the three-dimensional AT transition line.

Problem Statement

Determine, on the basis of the available Monte Carlo, field-theoretic, and renormalization-group evidence, the nature of the phase-transition line(s) of the three-dimensional Ashkin–Teller model on the simple cubic lattice, as a function of the two-spin coupling $J$ and the four-spin coupling $U$ (including the negative-$U$ region of the standard phase diagram). This single objective comprises three jointly linked deliverables: (1) the topology and order of all phase boundaries — paramagnetic–Baxter, paramagnetic–mixed, and mixed–Baxter — including the existence, number, and locations of tricritical points, any bifurcation points and critical endpoints, and which segments are first-order versus continuous; (2) the universality class of every continuous segment — Ising, XY with emergent O(2) symmetry, or genuinely varying exponents — and in particular whether the previously reported continuous drift of critical exponents along the line between tricritical points reflects genuine non-universality or a wide crossover governed by the weakly irrelevant cubic anisotropy; and (3) a reconciliation of the Monte Carlo phase diagram (wide crossover; coexistence of continuous and first-order transitions on the same boundary for different order-parameter components) with the hierarchical-lattice renormalization-group phase diagram (first- and second-order boundaries separated by tricritical and critical endpoints). Published reference points relevant to any resolution include the decoupled-Ising point at $U=0$, the four-state-Potts point at $J=U$ (where Monte Carlo and field-theoretic evidence indicates that transitions become very weakly first-order), and the published tricritical-point locations.

References

1. C. Fan, 'On critical properties of the Ashkin-Teller model', Physics Letters A 39, 136 (1972). DOI: 10.1016/0375-9601(72)91051-1, https://doi.org/10.1016/0375-9601(72)91051-1
2. R. J. Baxter, 'Partition function of the Eight-Vertex lattice model', Annals of Physics 70, 193 (1972). DOI: 10.1016/0003-4916(72)90335-1, https://doi.org/10.1016/0003-4916(72)90335-1
3. R. V. Ditzian, J. R. Banavar, G. S. Grest, L. P. Kadanoff, 'Phase diagram for the Ashkin-Teller model in three dimensions', Physical Review B 22, 2542 (1980). DOI: 10.1103/PhysRevB.22.2542, https://doi.org/10.1103/PhysRevB.22.2542
4. P. Arnold, Y. Zhang, 'Monte Carlo study of very weakly first-order transitions in the three-dimensional Ashkin-Teller model', Nuclear Physics B 501, 803 (1997). DOI: 10.1016/S0550-3213(97)00405-7, arXiv:hep-lat/9610032, https://arxiv.org/abs/hep-lat/9610032
5. G. Musiał, L. Dębski, G. Kamieniarz, 'Monte Carlo simulations of Ising-like phase transitions in the three-dimensional Ashkin-Teller model', Physical Review B 66, 012407 (2002). DOI: 10.1103/PhysRevB.66.012407, https://doi.org/10.1103/PhysRevB.66.012407
6. G. Musiał, 'Monte Carlo analysis of the tricritical behavior in a three-dimensional system with a multicomponent order parameter: The Ashkin-Teller model', Physical Review B 69, 024407 (2004). DOI: 10.1103/PhysRevB.69.024407, https://doi.org/10.1103/PhysRevB.69.024407
7. G. Musiał, J. Rogiers, 'On the possibility of nonuniversal behavior in the 3D Ashkin–Teller model', physica status solidi (b) 243, 335 (2006). DOI: 10.1002/pssb.200562438, https://doi.org/10.1002/pssb.200562438
8. G. Szukowski, G. Kamieniarz, G. Musiał, 'Verification of Ising phase transitions in the three-dimensional Ashkin-Teller model using Monte Carlo simulations', Physical Review E 77, 031124 (2008). DOI: 10.1103/PhysRevE.77.031124, https://doi.org/10.1103/PhysRevE.77.031124
9. Z. Wojtkowiak, G. Musiał, 'Wide crossover in the 3D Ashkin–Teller model', Physica A: Statistical Mechanics and its Applications 513, 104 (2019). DOI: 10.1016/j.physa.2018.08.166, https://doi.org/10.1016/j.physa.2018.08.166
10. G. Musiał, D. Jeziorek-Knioła, Z. Wojtkowiak, 'Monte Carlo examination of first-order phase transitions in a system with many independent order parameters: Three-dimensional Ashkin-Teller model', Physical Review E 103, 062124 (2021). DOI: 10.1103/PhysRevE.103.062124, https://doi.org/10.1103/PhysRevE.103.062124
11. Z. Wojtkowiak, G. Musiał, 'The Behavior of the Three-Dimensional Askin–Teller Model at the Mixed Phase Region by a New Monte Carlo Approach', Journal of Statistical Physics 189, 3 (2022). DOI: 10.1007/s10955-022-02964-w, https://doi.org/10.1007/s10955-022-02964-w
12. İ. Keçoğlu, A. N. Berker, 'Global Ashkin–Teller phase diagrams in two and three dimensions: Multicritical bifurcation versus double tricriticality—endpoint', Physica A: Statistical Mechanics and its Applications 630, 129248 (2023). DOI: 10.1016/j.physa.2023.129248, https://doi.org/10.1016/j.physa.2023.129248
13. D. Zhang, M. Hu, Y. Sun, J.-P. Lv, 'XY criticality arising from emergent symmetry in the three-dimensional Ashkin-Teller model', Physical Review E 111, 044128 (2025). DOI: 10.1103/PhysRevE.111.044128, https://doi.org/10.1103/PhysRevE.111.044128
14. M. Hasenbusch, 'Monte Carlo study of the $O(2)$-invariant $\varphi^4$ theory with a cubic perturbation in three dimensions', arXiv:2510.19473 (2025). https://arxiv.org/abs/2510.19473
15. A. Türkoğlu, A. N. Berker, 'XY-Ashkin–Teller phase diagram in d=3', Physica A: Statistical Mechanics and its Applications 674, 130715 (2025). DOI: 10.1016/j.physa.2025.130715, arXiv:2503.23739, https://arxiv.org/abs/2503.23739
16. A. Pelissetto, E. Vicari, 'Critical phenomena and renormalization-group theory', Physics Reports 368, 549 (2002). DOI: 10.1016/S0370-1573(02)00219-3, https://doi.org/10.1016/S0370-1573(02)00219-3