Boundary-condition dependence of the six-vertex bulk free energy: extending the density-only (n-equivalence) classification beyond periodic matched patterns

For the homogeneous six-vertex model on a finite domain with fixed boundary line configurations, Minami (2008) argued, at physics level of rigor and via the n-equivalence relation, that the thermodynamic free energy is classified by the boundary arrow densities alone, $f=f(\rho_1,\rho_2)$, equal to the Lieb–Sutherland periodic value at $\rho_1=\rho_2=1/2$, provided the boundary patterns are strictly periodic and satisfy left–right and top–bottom matching. The open problem is to determine, rigorously, how far this density-only classification extends: (i) to non-periodic boundary patterns (slowly varying or fractal-like) with matching and vanishing boundary-site fraction; (ii) to regimes where the boundary-site fraction does not vanish, so the hypothesis $n=o(N/N')$ fails; (iii) to boundary conditions violating the matching hypotheses, of which the square domain-wall problem — rigorously solved in all phases by the Bleher school — is only the special archetype; and (iv) to a rigorous foundation for the n-equivalence argument itself. The modern variational framework of Lammers–Tassy covers six-vertex weights only for non-taut boundary profiles and proves strict convexity of the surface tension only in the monotone regime, so all four threads remain open.

Background

The six-vertex model places an arrow on every edge of the square lattice subject to the ice rule — at each vertex exactly two arrows point in and two point out — so each vertex realizes one of six local arrow patterns, weighted $a$, $b$, $c$ (two patterns of each weight). Its phase structure is governed by the anisotropy parameter $\Delta=(a^{2}+b^{2}-c^{2})/(2ab)$: $|\Delta|<1$ is the disordered regime, $\Delta>1$ ferroelectric, $\Delta<-1$ antiferroelectric. The central thermodynamic quantity is the bulk free energy per site, $f=-(\beta N)^{-1}\log Z$, in the limit where the number of sites $N\to\infty$. An arrow configuration decomposes into disjoint directed lines that enter and leave through the boundary; fixing which boundary bonds carry a line defines a fixed boundary condition, and the fractions of occupied bonds on the top/bottom and left/right edges are the boundary line densities $\rho_1$ and $\rho_2$.

With periodic (cyclic) boundary conditions the free energy was computed exactly by the Bethe ansatz — an exact diagonalization technique for transfer matrices, originally developed for one-dimensional quantum spin chains — in the classic 1967 works of Lieb and Sutherland; write $f_{LS}$ for the value at zero field. Boundary conditions genuinely matter at the level of the bulk free energy: Korepin and Zinn-Justin (2000) showed that with domain-wall boundary conditions (DWBC — all boundary arrows point inward along two adjacent sides and outward along the other two, the finite-volume condition underlying the Izergin–Korepin determinant formula, a closed determinant expression for the finite-volume partition function) the bulk free energy differs from the periodic one and bears no simple relation to it.

Minami (2006, published 2008) asked the general question: for a fixed boundary line configuration on an arbitrary finite domain $D$, how does the limiting free energy depend on the boundary pattern? He introduced the n-equivalence relation: two boundary conditions are n-equivalent if the sets of configurations possible on the n-boundary — the layer of sites reachable from the boundary in at most $n$ steps, together with the bonds between its layers — coincide. His Proposition 1 (physics-level) asserts that n-equivalence for every lattice in a sequence, with $n=o(N/N')$ where $N'$ is the number of boundary sites, forces the two limiting free energies to be equal. His Proposition 3 then derives, under (a) strictly periodic boundary line patterns with fixed periods $w_p,h_p$, (b) matching conditions (a line occupies a left-edge bond if and only if it occupies the same-row right-edge bond, and similarly bottom/top), and (c) a sequential-limit argument, that the free energy is classified by the densities alone: $f=f(\rho_1,\rho_2)$, with $f(1/2,1/2)=f_{LS}$; at the free-fermion point $a=b=1$, $c=\sqrt{2}$ (the weight choice at which the model maps to free fermions, i.e. to a dimer model) this reproduces the Kasteleyn–Temperley–Fisher domino-tiling value $-\beta f=2G/\pi$, with $G$ Catalan's constant. DWBC violates the hypotheses — consistently with Korepin–Zinn-Justin.

The rigorous counterpart is the variational principle. An arrow configuration determines a height function on the faces of the lattice (the height changes by $\pm1$ across each edge, with the sign fixed by the arrow direction); the surface tension $\sigma(u)$ is the negative entropy density of shift-invariant six-vertex Gibbs measures at fixed average slope $u$. A variational principle asserts that $-N^{-1}\log Z$ converges to the minimum of $\int\sigma(\nabla h)$ over height functions matching the boundary data, and a large deviation principle (LDP) refines this to a full rate function, giving the exponential decay rate of the probability of each macroscopic profile. Cohn–Kenyon–Propp (2001) proved it for domino tilings, and Sheffield's monograph Random surfaces built the general gradient-surface framework. Lammers–Tassy (2020, published 2024) proved the LDP and variational principle for a class of Lipschitz random surfaces that includes the six-vertex model at arbitrary weights (their Section 13.4), but only for non-taut asymptotic boundary profiles — profiles admitting a Lipschitz extension with slack; extremal (taut, frozen-facet-type) profiles are excluded — and they prove strict convexity of $\sigma$ only in the monotone regime $c^{+}c^{-}\geq\max(a^{+}a^{-},b^{+}b^{-})$ (for homogeneous weights $c^{2}\geq\max(a^{2},b^{2})$), recovering a result of Sheffield. On the other anchor, Duminil-Copin–Kozlowski–Krachun–Manolescu–Tikhonovskaia (2022) gave a fully rigorous Bethe-ansatz computation of the six-vertex free energy on the torus, including fixed arrow densities, for $\Delta<1$.

Between these anchors the picture is incomplete. The rigorous variational framework does not reach taut or non-matched boundary data; the exact DWBC asymptotics of the Bleher school cover the square (and special variants) but not general boundary patterns, since the determinant representation is specific to DWBC; and Minami's n-equivalence method — the only framework aimed directly at classifying boundary dependence on general domains — has had essentially no follow-up and rests on unproved reachability and sequential-limit assumptions.

Problem Statement

Determine, with mathematical rigor, how the thermodynamic (bulk) free energy of the homogeneous six-vertex model on finite domains depends on fixed boundary line configurations — that is, settle the boundary of the density-only classification $f=f(\rho_1,\rho_2)$ established by Minami under periodic, matched, vanishing-boundary-fraction hypotheses. The common objective is a precise delineation of which boundary statistics determine the bulk free energy and which can change it, through four linked regimes:

1. Non-periodic patterns. For boundary line configurations that satisfy the left/right and top/bottom matching conditions and have densities $(\rho_1,\rho_2)$ and vanishing boundary-site fraction ($N'/N\to 0$) but are not periodic — for instance slowly varying or fractal-like patterns — is the limiting free energy still $f(\rho_1,\rho_2)$, or can finer boundary statistics change the bulk free energy? Provide either a rigorous classification theorem (stating precisely the class of boundary patterns covered) or an explicit pair of matched, equal-density boundary sequences on explicit domains whose limiting free energies provably differ, identifying the boundary statistic responsible for the difference.

2. Non-vanishing boundary fraction. Establish or refute the extension of the n-equivalence principle when $N'/N\not\to 0$, i.e. when the hypothesis $n=o(N/N')$ of Proposition 1 fails: does equality of the n-boundary configuration sets still force equal limiting free energies for some nontrivial scaling of $n$, and on what boundary statistics does the free energy then depend?

3. Non-matched boundary data. Classify the free energy for boundary conditions violating the left/right and top/bottom matching hypotheses — beyond the square domain-wall case, which is exactly solved in all phases — at least for balanced non-matched data (densities $(\rho_1,\rho_2)=(1/2,1/2)$) on rectangles and general domains. In particular, settle whether the taut (extremal, frozen-facet-type) boundary profiles excluded from the Lammers–Tassy large-deviation principle admit a variational-principle description, and identify its minimizers. The balanced case is the mandatory floor of this part: resolving it is a necessary first step but does not by itself resolve Part 3, whose target is the general classification of non-matched data.

4. Foundation of the method. Give a rigorous formulation and proof — or a refutation — of the n-equivalence principle itself (Minami's Proposition 1, including its reachability and sequential-limit assumptions) for all boundary densities $0<\rho_1,\rho_2<1$ and all $\Delta<1$.

Resolution standard. The four regimes above are independent open problems, and a response need not close all of them. A regime is resolved in full when the response provides either a rigorous positive result — a classification theorem, or a proof of the criterion in question, covering a precisely stated class of boundary patterns — or a rigorous refutation — an explicit counterexample, or an explicit pair of matched, equal-density boundary sequences on explicit domains whose limiting free energies provably differ — matching that regime's own prove-or-refute wording. A submission states which regimes it resolves: resolving all four is a complete resolution, and resolving a proper subset is a partial resolution whose scope must be declared. Part 4 supplies a foundational tool but does not by itself discharge Parts 1–3, each of which requires its own explicit construction.

References

1. Kazuhiko Minami, 'The free energies of six-vertex models and the n-equivalence relation', Journal of Mathematical Physics 49 (2008) 033514. DOI: 10.1063/1.2890671. arXiv:cond-mat/0607513. https://arxiv.org/abs/cond-mat/0607513
2. V. Korepin and P. Zinn-Justin, 'Thermodynamic limit of the six-vertex model with domain wall boundary conditions', Journal of Physics A: Mathematical and General 33 (2000) 7053–7066. DOI: 10.1088/0305-4470/33/40/304. arXiv:cond-mat/0004250. https://arxiv.org/abs/cond-mat/0004250
3. Pavel M. Bleher and Vladimir V. Fokin, 'Exact Solution of the Six-Vertex Model with Domain Wall Boundary Conditions. Disordered Phase', Communications in Mathematical Physics 268 (2006) 223–284. DOI: 10.1007/s00220-2006-0097-y. arXiv:math-ph/0510033. https://arxiv.org/abs/math-ph/0510033
4. Pavel Bleher and Karl Liechty, 'Exact Solution of the Six-Vertex Model with Domain Wall Boundary Conditions. Ferroelectric Phase', Communications in Mathematical Physics 286 (2009) 777–801. DOI: 10.1007/s00220-008-0709-9. https://doi.org/10.1007/s00220-008-0709-9
5. Pavel Bleher and Karl Liechty, 'Exact Solution of the Six-Vertex Model with Domain Wall Boundary Conditions: Antiferroelectric Phase', Communications on Pure and Applied Mathematics 63 (2010) 779–829. DOI: 10.1002/cpa.20311. https://doi.org/10.1002/cpa.20311
6. Pavel Bleher and Karl Liechty, 'Exact Solution of the Six-Vertex Model with Domain Wall Boundary Conditions. Critical Line between Ferroelectric and Disordered Phases', Journal of Statistical Physics 134 (2009) 463–485. DOI: 10.1007/s10955-009-9688-2. https://doi.org/10.1007/s10955-009-9688-2
7. Pavel Bleher and Thomas Bothner, 'Exact Solution of the Six-Vertex Model with Domain Wall Boundary Conditions: Critical Line between Disordered and Antiferroelectric Phases', Random Matrices: Theory and Applications 1 (2012) 1250012. DOI: 10.1142/S2010326312500128. https://doi.org/10.1142/S2010326312500128
8. P. Zinn-Justin, 'The Influence of Boundary Conditions in the Six-Vertex Model', preprint (2002). arXiv:cond-mat/0205192. https://arxiv.org/abs/cond-mat/0205192
9. Henry Cohn, Richard Kenyon, and James Propp, 'A variational principle for domino tilings', Journal of the American Mathematical Society 14 (2001) 297–346. DOI: 10.1090/S0894-0347-00-00355-6. arXiv:math/0008220. https://arxiv.org/abs/math/0008220
10. Scott Sheffield, 'Random surfaces', Astérisque 304, Société Mathématique de France (2005). arXiv:math/0304049. https://arxiv.org/abs/math/0304049
11. Piet Lammers and Martin Tassy, 'Macroscopic behavior of Lipschitz random surfaces', Probability and Mathematical Physics 5 (2024) 177–267. DOI: 10.2140/pmp.2024.5.177. arXiv:2004.15025. https://arxiv.org/abs/2004.15025
12. Hugo Duminil-Copin, Karol Kajetan Kozlowski, Dmitry Krachun, Ioan Manolescu, and Tatiana Tikhonovskaia, 'On the Six-Vertex Model's Free Energy', Communications in Mathematical Physics 395 (2022) 1383–1430. DOI: 10.1007/s00220-022-04459-x. arXiv:2012.11675. https://arxiv.org/abs/2012.11675
13. Hugo Duminil-Copin, Alex M. Karrila, Ioan Manolescu, and Mendes Oulamara, 'Delocalization of the height function of the six-vertex model', Journal of the European Mathematical Society 26 (2024) 4131–4190. DOI: 10.4171/jems/1436. https://doi.org/10.4171/jems/1436
14. Filippo Colomo and Andrea Sportiello, 'Arctic Curves of the Six-Vertex Model on Generic Domains: The Tangent Method', Journal of Statistical Physics 164 (2016) 1488–1523. DOI: 10.1007/s10955-016-1590-0. arXiv:1605.01388. https://arxiv.org/abs/1605.01388
15. T.S. Tavares, G.A.P. Ribeiro, and V.E. Korepin, 'The entropy of the six-vertex model with variety of different boundary conditions', Journal of Statistical Mechanics (2015) P06016. arXiv:1501.02818. https://arxiv.org/abs/1501.02818
16. Filippo Colomo and Andrei G. Pronko, 'Thermodynamics of the six-vertex model in an L-shaped domain', Communications in Mathematical Physics 339 (2015) 699–728. arXiv:1501.03135. https://arxiv.org/abs/1501.03135
17. Kazuhiko Minami, 'Fractal structure of a solvable lattice model', preprint (2008). arXiv:0801.0186. https://arxiv.org/abs/0801.0186
18. Elliott H. Lieb, 'Residual Entropy of Square Ice', Physical Review 162 (1967) 162–172. DOI: 10.1103/PhysRev.162.162. https://doi.org/10.1103/PhysRev.162.162
19. Bill Sutherland, 'Exact Solution of a Two-Dimensional Model for Hydrogen-Bonded Crystals', Physical Review Letters 19 (1967) 103–104. DOI: 10.1103/PhysRevLett.19.103. https://doi.org/10.1103/PhysRevLett.19.103
20. P.W. Kasteleyn, 'The statistics of dimers on a lattice', Physica 27 (1961) 1209–1225. DOI: 10.1016/0031-8914(61)90063-5. https://doi.org/10.1016/0031-8914(61)90063-5
21. H. N. V. Temperley and Michael E. Fisher, 'Dimer problem in statistical mechanics — an exact result', Philosophical Magazine 6 (1961) 1061–1063. DOI: 10.1080/14786436108243366. https://doi.org/10.1080/14786436108243366