The six-vertex model with domain-wall boundary conditions (DWBC) is exactly solved at the level of its partition function and arctic-curve limit shape for all anisotropies Δ\Delta, but its correlation functions are known analytically in the thermodynamic limit only at the free-fermion point Δ=0\Delta=0, where the height field is a Gaussian free field and the two-point function is a Dirichlet Green's function of the Laplacian. Belov and Reshetikhin (2022) computed the two-point height correlator by Markov chain Monte Carlo at Δ=1/2\Delta=1/2 (logarithm-like decay) and Δ=−7/2\Delta=-7/2 on a 90×9090\times90 lattice (exponential decay), and posed the analytic problem: determine these correlators for generic Δ≠0\Delta\neq0 — in the disordered regime ∣Δ∣<1|\Delta|<1, where an inhomogeneous Gaussian free field with position-dependent coupling is conjectured but not derived, and in the antiferroelectric regime Δ<−1\Delta<-1, where only fitted exponentials exist — together with the mesoscopic crossover as the antiferroelectric droplet size approaches the correlation length and the limit Δ→−1−\Delta\to-1^- toward the Kosterlitz–Thouless point. The 2019–2026 literature shows: the free-fermion side is now completely determinantal (Johnston–Shiatis 2025), Gaussian-free-field convergence has been proven for the homogeneous model on Z2\mathbb{Z}^2 for −1≤Δ≤−1/2-1\le\Delta\le-1/2 (Duminil-Copin, Kozlowski, Lammers, Manolescu 2026), and exact finite-lattice integral representations for nonlocal DWBC correlators exist (Colomo, Di Giulio, Pronko 2021) — but no analytic thermodynamic-limit two-point height correlator at any Δ≠0\Delta\neq0, no closed-form coupling, no antiferroelectric correlation length, and no mesoscopic crossover analysis has been published. The problem remains open.

Background

The six-vertex model places arrows on the edges of a square lattice with Pauling's ice rule at every vertex: exactly two of the four incident edges carry arrows pointing into the vertex. Up to rotations there are six allowed vertex configurations, carrying positive weights aa, bb, cc, and the physics is organized by the anisotropy parameter Δ=(a2+b2−c2)/(2ab)\Delta=(a^2+b^2-c^2)/(2ab). For Δ>1\Delta>1 the model is ferroelectric (ordered); for ∣Δ∣<1|\Delta|<1 it is disordered (critical); for Δ<−1\Delta<-1 it is antiferroelectric, with two degenerate staggered ground states and correlations that decay exponentially with a finite correlation length. The points Δ=±1\Delta=\pm1 are Kosterlitz–Thouless transition points at which the correlation length diverges.

This problem concerns the model with domain-wall boundary conditions (DWBC): all arrows on the left and right sides of the N×NN\times N square point inward and all arrows on the top and bottom point outward. These boundaries force strong spatial inhomogeneity: in the thermodynamic limit the height function — the integer-valued function on faces whose value changes by ±1\pm1 across each edge according to the arrow direction — concentrates on a deterministic limit shape, with frozen, ferroelectrically ordered corners separated from an inner liquid region by smooth "arctic" curves.

The free-fermion point Δ=0\Delta=0 (equivalently a2+b2=c2a^2+b^2=c^2) is special: the model maps to a dimer model on a decorated square lattice, so both the partition function and all correlation functions reduce to determinants and minors of the Kasteleyn matrix. Belov and Reshetikhin (2022) used this to show that at Δ=0\Delta=0 the thermodynamic-limit two-point correlation function of the centered height field is the Dirichlet Green's function of the Laplacian on the upper half-plane, written in the conformal coordinate determined by the limit shape (their Eq. (13)); at short distances it behaves as −(1/2π)ln⁡∣r⃗i−r⃗j∣-(1/2\pi)\ln|\vec r_i-\vec r_j|, and multipoint functions follow from Wick's rule. In other words, the height field is a Gaussian free field (GFF) — the massless Gaussian scalar field of central charge c=1c=1.

Away from Δ=0\Delta=0 this determinantal structure disappears. The DWBC partition function is still given by Izergin's determinant for arbitrary weights, but correlation functions are nonlocal observables for which no determinant representation is known, and their thermodynamic asymptotics at generic Δ\Delta have resisted analysis. The main physics-level input is the conjecture of Granet, Budzynski, Dubail and Jacobsen (2019), based on Bethe-ansatz computations with an imaginary extensive twist: inside the interacting arctic curve the height field is an inhomogeneous GFF — still Gaussian, but with a position-dependent coupling K(x)K(x), so that the two-point function is the Green's function of the generalized Laplacian ∇⋅(K(x)−1∇)\nabla\cdot(K(x)^{-1}\nabla) rather than of the ordinary Laplacian. The coupling K(x)K(x) is not known in closed form and the claim is unproven. For Δ<−1\Delta<-1 an antiferroelectric droplet (a region ordered in one of the two staggered ground states) sits at the center of the liquid region, and correlations inside it are expected to decay exponentially.

Belov and Reshetikhin approached these questions numerically by Markov chain Monte Carlo (MCMC): a Markov chain with local, detailed-balance updates that drives an arbitrary initial configuration toward the Boltzmann distribution. Their numerics at Δ=1/2\Delta=1/2 show the logarithm-like behavior of the critical regime (extrapolated log-coefficient −0.1588±0.0058-0.1588\pm0.0058, consistent with −1/(2π)-1/(2\pi)), and at Δ=−7/2\Delta=-7/2 on a 90×9090\times90 lattice an exponential decay fitted by 0.484+0.516 e(−99±4)∣r⃗∣0.484+0.516\,e^{(-99\pm4)|\vec r|} on top of a flat "pillow" background that they attribute to mesoscopic (Airy-scale) effects near the boundary of the disordered region. Two practical obstacles are emphasized in the paper: local Glauber-type dynamics provably mix exponentially slowly in the ordered phases (Fahrbach–Randall 2019), and no effective criterion for thermalization — recognizing when the Markov chain has reached equilibrium — is known, so the authors monitor the drift of the normalized volume under the height function empirically. Near the arctic curve, boundary fluctuations are expected to be governed by Airy-process/Tracy–Widom statistics (proven only at free-fermion points; Monte Carlo evidence elsewhere), and these "Airy waves" contaminate finite-lattice correlators.

The unresolved target is the analytic determination of these correlators at generic Δ≠0\Delta\neq0 — in the disordered and antiferroelectric regimes, through the mesoscopic crossover as Δ→−1−\Delta\to-1^- where the antiferroelectric droplet shrinks to the scale of the diverging correlation length — together with the finite-size and boundary-layer asymptotics needed to connect finite lattices to the thermodynamic limit.

Problem Statement

Determine, analytically, the thermodynamic-limit two-point correlation function of the height field of the six-vertex model with domain-wall boundary conditions on the N×NN\times N square, at generic anisotropy Δ≠0\Delta\neq0 — beyond the free-fermion point Δ=0\Delta=0 where the answer is the Laplacian Green's function of Belov and Reshetikhin — together with its multipoint (Gaussian) structure and asymptotic regimes. Concretely, the source asks for, and later work has not supplied:

(a) In the disordered regime ∣Δ∣<1|\Delta|<1: an explicit continuum formula, or exact asymptotics, for the two-point function inside the liquid region — in particular a derivation (or refutation) of the conjectured inhomogeneous Gaussian-free-field structure with an explicitly determined position-dependent coupling K(x)K(x), which would also fix all multipoint functions through Wick's rule.

(b) In the antiferroelectric regime Δ<−1\Delta<-1: the analytic form of the exponentially decaying two-point function in the interior of the antiferroelectric droplet, including the correlation length and prefactor as functions of Δ\Delta.

(c) The mesoscopic crossover when the linear size of the antiferroelectric droplet is comparable to the antiferroelectric correlation length, and the approach to the Kosterlitz–Thouless point Δ→−1−\Delta\to-1^- where the droplet shrinks and finite-size effects dominate — a regime the authors explicitly promised to analyze "both numerically and from the exact solution" in a separate publication that has not appeared.

(d) The finite-size and boundary-layer (arctic-curve, Airy-scale) asymptotics of the correlators at generic Δ\Delta.

Components (a)–(c) form the core scientific objective: a correlation-function theory of the domain-wall six-vertex model away from free fermion, spanning the phase diagram and its crossovers — (a) and (b) are the bulk answers in the two phases, and (c) connects them at the phase boundary; (d) states the complementary finite-size and boundary-layer asymptotic questions that connect the bulk answers to simulations on finite lattices. Resolution of this problem is constituted by supplying (a)–(c), the multipoint (Gaussian) structure being the Wick-rule content of (a) rather than a separate deliverable; (d) is a complementary extension and not a required deliverable, and rigorously establishing any single one of (a)–(c) counts as partial progress. The methodological question of controlling thermalization and mixing times of local Markov-chain samplers, or accelerating them (for example by GPU implementations or nonlocal updates), which the same paper flags as its practical obstacle, is an auxiliary direction with an independent literature in Markov-chain theory and is not part of the completion boundary of this problem.

References

  1. P. Belov, N. Reshetikhin, The two-point correlation function in the six-vertex model, J. Phys. A: Math. Theor. 55 (2022) 155001, DOI 10.1088/1751-8121/ac578e; arXiv:2012.05182, https://arxiv.org/abs/2012.05182
  2. E. Granet, L. Budzynski, J. Dubail, J. L. Jacobsen, Inhomogeneous Gaussian Free Field inside the interacting arctic curve, J. Stat. Mech. (2019) 013102, DOI 10.1088/1742-5468/aaf71b; arXiv:1807.07927, https://arxiv.org/abs/1807.07927
  3. F. Colomo, G. Di Giulio, A. G. Pronko, Six-vertex model on a finite lattice: integral representations for nonlocal correlation functions, Nucl. Phys. B 972 (2021) 115535, DOI 10.1016/j.nuclphysb.2021.115535; arXiv:2107.13358, https://arxiv.org/abs/2107.13358
  4. F. Colomo, A. G. Pronko, Evaluation of integrals for the emptiness formation probability in the square-ice model, Nucl. Phys. B 1004 (2024) 116565, DOI 10.1016/j.nuclphysb.2024.116565; arXiv:2405.04358, https://arxiv.org/abs/2405.04358
  5. S. G. G. Johnston, R. Shiatis, Explicit correlation functions for the six-vertex model in the free-fermion regime, arXiv:2512.10199 (2025), https://arxiv.org/abs/2512.10199
  6. A. Glazman, R. Peled, On the transition between the disordered and antiferroelectric phases of the 6-vertex model, Electron. J. Probab. 28 (2023), DOI 10.1214/23-ejp980; arXiv:1909.03436, https://arxiv.org/abs/1909.03436
  7. A. Glazman, P. Lammers, Delocalisation and Continuity in 2D: Loop O(2), Six-Vertex, and Random-Cluster Models, Commun. Math. Phys. 406 (2025) 108, DOI 10.1007/s00220-025-05259-9, https://doi.org/10.1007/s00220-025-05259-9
  8. H. Duminil-Copin, K. K. Kozlowski, D. Krachun, I. Manolescu, M. Oulamara, Rotational invariance in critical planar lattice models, arXiv:2012.11672 (2020), https://arxiv.org/abs/2012.11672
  9. H. Duminil-Copin, K. K. Kozlowski, P. Lammers, I. Manolescu, Gaussian free field convergence of the six-vertex model with −1≤Δ≤−12-1\le\Delta\le-\frac{1}{2}, arXiv:2603.06268 (2026), https://arxiv.org/abs/2603.06268
  10. I. Lyberg, V. Korepin, J. Viti, Fluctuation of the phase boundary in the six-vertex model with domain wall boundary conditions: a Monte Carlo study, J. Phys. A: Math. Theor. 56 (2023) 495002, DOI 10.1088/1751-8121/ad0a43, https://doi.org/10.1088/1751-8121/ad0a43
  11. M. Prähofer, H. Spohn, Domain wall fluctuations of the six-vertex model at the ice point, J. Phys. A: Math. Theor. 57 (2024) 025001, DOI 10.1088/1751-8121/ad13b4, https://doi.org/10.1088/1751-8121/ad13b4
  12. M. Fahrbach, D. Randall, Slow Mixing of Glauber Dynamics for the Six-Vertex Model in the Ordered Phases, in Proceedings of the 23rd International Conference on Randomization and Computation (RANDOM 2019); arXiv:1904.01495, https://arxiv.org/abs/1904.01495
  13. Z. Fu, J. Li, X. Yang, Beyond Windability: An FPRAS for The Six-vertex Model, arXiv:2202.02999 (2022), https://arxiv.org/abs/2202.02999
  14. R. Keesman, J. Lamers, Numerical study of the F model with domain-wall boundaries, Phys. Rev. E 95 (2017) 052117, DOI 10.1103/PhysRevE.95.052117, https://doi.org/10.1103/PhysRevE.95.052117
  15. J.-M. Stéphan, Two dimensional inhomogeneous classical systems at criticality, arXiv:2608.02903 (2026), https://arxiv.org/abs/2608.02903