Completeness of the homogeneous $T$–$Q$ Bethe ansatz for the eight-vertex model with quasi-periodic boundary conditions on an odd number of sites

The inhomogeneous eight-vertex model (equivalently the spin-1/2 XYZ chain) with integrable quasi-periodic boundary conditions — periodic, or twisted by a Pauli matrix $\sigma^x$, $\sigma^y$, $\sigma^z$ — was completely solved by Niccoli and Terras (2016) using separation of variables and a vertex-IRF mapping to the antiperiodic dynamical six-vertex model. The reformulation of this solution in terms of Baxter's homogeneous $T$–$Q$ functional equation, with $Q$ an elliptic polynomial of quasi-periods fixed by the twist and hence ordinary Bethe-type equations, is proven complete only for an even number of sites $N$. Whether this ansatz gives a complete characterization of the transfer-matrix spectrum and eigenstates for odd $N$ — both in the twisted cases, where the spectrum is simple, and in the periodic case, where it is doubly degenerate — is the open problem stated here, drawn from the 2016 paper, whose authors identified it as an outstanding technical problem. The odd-$N$ case remains open: off-diagonal Bethe ansatz constructions cover odd $N$ only through inhomogeneous $T$–$Q$ relations, and later work by the authors and others treats adjacent models (open XYZ chains with constrained boundaries, the antiperiodic XXZ chain, rational degenerations), leaving the odd-site homogeneous $T$–$Q$ completeness question untouched.

Background

The eight-vertex model, introduced by Baxter in 1972, is a two-dimensional lattice statistical model whose local Boltzmann weights allow the eight arrow configurations with an even number of arrows pointing into each vertex; as a transfer-matrix problem it is equivalent to the spin-1/2 Heisenberg XYZ chain, the fully anisotropic relative of the XXZ chain. Baxter solved the periodic model by constructing a commuting family of transfer matrices $t(\lambda)$ and Q-operators satisfying the functional $T$–$Q$ equation $t(\lambda)Q(\lambda)=a(\lambda)Q(\lambda-\eta)+d(\lambda)Q(\lambda+\eta)$, where $a$ and $d$ are known coefficient functions, $\eta$ is the crossing (anisotropy) parameter, and the model is made inhomogeneous by site-dependent spectral parameters $\xi_n$ (a standard technical device that regularizes functional equations; the homogeneous chain is recovered in the limit $\xi_n\to 0$). By contrast with the six-vertex/XXZ case, the generic eight-vertex model admits no obvious reference state on which the algebraic Bethe ansatz (ABA), the standard eigenvector construction, can be iterated, so its spectrum and eigenstates are much harder to describe in Bethe-ansatz form. Two later tools are central here. First, Sklyanin's separation of variables (SoV) diagonalizes a transfer matrix by finding a basis in which it acts by multiplication, giving in favourable cases a complete characterization of all eigenvalues and eigenvectors at once. Second, the vertex-IRF (interaction-round-a-face) transformation is an invertible-at-the-operator-level change of representation that maps the eight-vertex model to the dynamical six-vertex (or SOS) model, a six-vertex model whose heights shift dynamically. Niccoli (2013) and Levy-Bencheton, Niccoli and Terras (2016) solved the dynamical six-vertex model with antiperiodic boundary conditions by SoV and, for an even number of sites $N$, proved that its spectrum and eigenstates are equivalently described by solutions of the homogeneous $T$–$Q$ equation of the above form, where $Q(\lambda)$ is an elliptic polynomial — a product of theta functions $Q(\lambda)=\prod_j\theta_X(\lambda-\lambda_j)$ with quasi-periods (shifts under which it picks up known exponential factors) tied to the boundary conditions. The zeros $\lambda_j$ then satisfy ordinary Bethe-type equations, and the eigenstates can be written in an ABA-like multiple-action form; completeness means that the solutions of this Bethe system are in bijection with the full set of eigenvalues, so that no state is missed. The source paper (Niccoli–Terras 2016) transported these results to the eight-vertex model with quasi-periodic boundary conditions, meaning periodic, or twisted by letting a Pauli matrix $\sigma^\alpha$ ($\alpha=x,y,z$) act on the auxiliary space of the monodromy matrix. In all twisted cases the vertex-IRF map is bijective, the spectrum is simple ($2^N$ distinct eigenvalues), and the SoV description is complete for any $N$; in the periodic case with odd $N$, two non-bijective vertex-IRF variants with non-intersecting kernels show the spectrum is doubly degenerate, consisting of $2^{N-1}$ eigenvalues of multiplicity 2 equal to half the spectrum of the antiperiodic dynamical six-vertex model. For these models the homogeneous $T$–$Q$ reformulation, however, is proven complete only for even $N$ (Theorem 4.3 of the source covers the completeness direction under the assumption that $N$ is even, and the underlying proof of Levy-Bencheton–Niccoli–Terras is even-$N$ only). For odd $N$ the authors expect the restriction to be purely technical, but no proof exists. Related but distinct: the off-diagonal Bethe ansatz (ODBA) of Cao, Cui, Yang, Shi and Wang (2014) describes the XYZ spectrum with arbitrary $N$ through an inhomogeneous $T$–$Q$ equation carrying an extra inhomogeneous term, without proving completeness of any homogeneous-ansatz description; and for the $Z_n$-Belavin generalization (Hao et al. 2017) the homogeneous $T$–$Q$ relation is recovered in that framework only when $N$ is a multiple of $n$, i.e. again only for even $N$ in the eight-vertex case. Odd-length eight-vertex chains were also studied early by Fabricius and McCoy (2005), but by Q-operator and degeneracy methods that do not supply a Bethe-equation completeness theorem. The odd-$N$ homogeneous $T$–$Q$ completeness question is thus the remaining gap in the Bethe-ansatz-type reformulation of the quasi-periodic eight-vertex model.

Problem Statement

Consider the inhomogeneous eight-vertex model (equivalently the spin-1/2 XYZ chain) on $N$ sites with generic inhomogeneity parameters and integrable quasi-periodic boundary conditions: transfer matrix twisted by $\sigma^x$, $\sigma^y$ or $\sigma^z$, or periodic. In all these cases the transfer-matrix spectrum and eigenstates are completely characterized by separation of variables through the vertex-IRF mapping to the antiperiodic dynamical six-vertex model: for the twisted cases the spectrum is simple with $2^N$ eigenvalues, and for the periodic case with odd $N$ it is doubly degenerate with $2^{N-1}$ eigenvalues of multiplicity two. The problem is to prove or disprove that, for an odd number of sites $N$, this characterization is equivalent to a description by Baxter's homogeneous $T$–$Q$ functional equation: that is, to determine whether every eigenvalue $t(\lambda)$ of the quasi-periodic eight-vertex transfer matrix with odd $N$ arises from a solution $Q(\lambda)$ of the homogeneous $T$–$Q$ equation, with $Q$ an elliptic polynomial of the quasi-periods fixed by the boundary condition (the ansatz of Levy-Bencheton–Niccoli–Terras, adjusted as needed in degree only), so that (i) the resulting system of Bethe-type equations for the zeros of $Q$ has exactly the right number of solutions — $2^N$ eigenvalues in the twisted cases, $2^{N-1}$ doubly degenerate eigenvalues in the periodic case — and (ii) the corresponding eigenstates are obtained in the associated algebraic-Bethe-ansatz-type construction. Both odd-$N$ settings left open by the source must be covered: the twisted cases and the periodic case, the latter requiring an account of how the two non-bijective vertex-IRF variants and the double degeneracy are reflected in the Bethe-equation solutions. The equivalence claim is established by a completeness proof for the ansatz at odd $N$ and refuted by an explicit counterexample (an eigenvalue not representable by any elliptic-polynomial solution of the homogeneous $T$–$Q$ equation); a corrected ansatz with proof likewise resolves it.

References

1. G. Niccoli, V. Terras, "The eight-vertex model with quasi-periodic boundary conditions", J. Phys. A: Math. Theor. 49, 044001 (2016). DOI: 10.1088/1751-8113/49/4/044001; arXiv:1508.03230; https://arxiv.org/abs/1508.03230
2. D. Levy-Bencheton, G. Niccoli, V. Terras, "Antiperiodic dynamical 6-vertex model by separation of variables II: functional equations and form factors", J. Stat. Mech. (2016) 033110. DOI: 10.1088/1742-5468/2016/03/033110; arXiv:1507.03404; https://arxiv.org/abs/1507.03404
3. R. J. Baxter, "Partition function of the eight-vertex lattice model", Annals of Physics 70, 193–228 (1972). DOI: 10.1016/0003-4916(72)90335-1; https://doi.org/10.1016/0003-4916(72)90335-1
4. G. Niccoli, "An antiperiodic dynamical six-vertex model: I. Complete spectrum by SOV, matrix elements of the identity on separate states and connections to the periodic eight-vertex model", J. Phys. A: Math. Theor. 46, 075003 (2013). DOI: 10.1088/1751-8113/46/7/075003; arXiv:1207.1928; https://arxiv.org/abs/1207.1928
5. K. Fabricius, B. M. McCoy, "New Developments in the Eight Vertex Model II. Chains of Odd Length", J. Stat. Phys. 120, 37–70 (2005). DOI: 10.1007/s10955-005-4410-5; https://doi.org/10.1007/s10955-005-4410-5
6. J. Cao, S. Cui, W.-L. Yang, K. Shi, Y. Wang, "Spin-1/2 XYZ model revisit: General solutions via off-diagonal Bethe ansatz", Nucl. Phys. B 886, 185–201 (2014). DOI: 10.1016/j.nuclphysb.2014.06.026; https://doi.org/10.1016/j.nuclphysb.2014.06.026
7. K. Hao, F. Wen, J. Cao, G.-L. Li, W.-L. Yang, K. Shi, Y. Wang, "On the solutions of the Z_n-Belavin model with arbitrary number of sites", Nucl. Phys. B 920, 419–441 (2017). DOI: 10.1016/j.nuclphysb.2017.04.019; arXiv:1609.00953; https://arxiv.org/abs/1609.00953
8. L. Pristyák, B. Pozsgay, "Current mean values in the XYZ model", SciPost Phys. 14, 158 (2023). DOI: 10.21468/SciPostPhys.14.6.158; arXiv:2211.00698; https://arxiv.org/abs/2211.00698
9. N. Slavnov, A. Zabrodin, A. Zotov, "Scalar products of Bethe vectors in the 8-vertex model", JHEP 2020, 123 (2020). DOI: 10.1007/JHEP06(2020)123; arXiv:2005.11224; https://arxiv.org/abs/2005.11224
10. G. Kulkarni, N. A. Slavnov, "Scalar products of Bethe vectors in the generalized algebraic Bethe ansatz", Theor. Math. Phys. 217, 1574–1594 (2023). DOI: 10.1134/S0040577923100100; arXiv:2306.12932; https://arxiv.org/abs/2306.12932
11. H. Pei, V. Terras, "On scalar products and form factors by separation of variables: the antiperiodic XXZ model", J. Phys. A: Math. Theor. 55, 015205 (2022). DOI: 10.1088/1751-8121/ac3b85; arXiv:2011.06109; https://arxiv.org/abs/2011.06109
12. G. Niccoli, V. Terras, "The open XYZ spin 1/2 chain: Separation of Variables and scalar products for boundary fields related by a constraint", SciPost Phys. 19, 090 (2025). DOI: 10.21468/SciPostPhys.19.4.090; arXiv:2402.04112; https://arxiv.org/abs/2402.04112
13. G. Niccoli, V. Terras, "On correlation functions of the open XYZ spin 1/2 chain with boundary fields related by a constraint", arXiv:2507.20299 (2025); https://arxiv.org/abs/2507.20299