Completing the exact solution of the eight-vertex model with quasi-periodic boundary conditions: odd-$N$ completeness of the homogeneous $T$-$Q$ ansatz, thermodynamic limit, and form factors

The inhomogeneous eight-vertex (XYZ) model with integrable quasi-periodic boundary conditions ($\sigma^x$-, $\sigma^y$-, $\sigma^z$-twisted transfer matrices, and the periodic transfer matrix with an odd number $N$ of sites) has a complete finite-$N$ spectral solution by Sklyanin's separation of variables, obtained by mapping it through the vertex-IRF transformation to the antiperiodic dynamical six-vertex model (Niccoli-Terras, 2016). Three linked parts of the solution program remain open: (i) the homogeneous Baxter-type $T$-$Q$ characterization, in which $Q$ is an elliptic polynomial with twist-dependent quasi-periods, is proven complete only for even $N$; (ii) no efficient route from the resulting inhomogeneous Bethe-type equations to the thermodynamic (large-$N$) limit is known for all quasi-periodic sectors; (iii) no compact finite-$N$ form factors or correlation functions exist for the elliptic quasi-periodic model, the obstruction being the non-local vertex-IRF dressing of the monodromy-matrix elements. The problem is to close these three gaps, thereby converting the finite-$N$ spectral solution into a complete solution (spectrum at all parities, thermodynamics, and observables) of the quasi-periodic eight-vertex model.

Background

The eight-vertex model is a two-dimensional statistical lattice model in which arrows on the edges of a square lattice obey an even-in/even-out rule at each vertex, giving eight local Boltzmann weights; as a one-dimensional quantum problem it is equivalent to the spin-1/2 XYZ Heisenberg chain, the most general nearest-neighbour integrable spin chain. Its solution is encoded in the transfer matrix $T(\lambda)$, the trace of a monodromy matrix built from Baxter's elliptic $R$-matrix; the commutativity of transfer matrices at different spectral parameters generates the conserved charges. Baxter solved the periodic model on a lattice with an even number $N$ of sites via his homogeneous functional $T$-$Q$ equation,

$t(\lambda)\,Q(\lambda)=a(\lambda)\,Q(\lambda-\eta)+d(\lambda)\,Q(\lambda+\eta),$

where $t(\lambda)$ is a transfer-matrix eigenvalue, $a(\lambda)$ and $d(\lambda)$ are known coefficient functions, $\eta$ is the crossing (anisotropy) parameter, and $Q(\lambda)$ is an elliptic polynomial: a meromorphic function fixed by its $N$ zeros together with quasi-periods inherited from the periods $\pi$ and $\pi\omega$ of the Jacobi theta functions in the $R$-matrix. In the algebraic Bethe ansatz (ABA) the zeros of $Q$ are the Bethe roots, and the functional equation becomes a system of Bethe equations; the eigenstates are built by multiple action of monodromy-matrix elements on a reference (pseudo-vacuum) state.

This route fails for integrable quasi-periodic boundary conditions: a twist of the monodromy matrix by a Pauli matrix $\sigma^\alpha$, $\alpha=x,y,z$, before taking the trace (antiperiodic and twisted boundary conditions for the equivalent spin chain), or periodic boundary conditions on a lattice with an odd number of sites. In these sectors there is no usable pseudo-vacuum and no global $U(1)$ charge symmetry, and Baxter's original vertex-IRF argument collapses: the space of states of the periodic solid-on-solid (SOS) model is zero-dimensional when $N$ is odd at generic $\eta$. Niccoli and Terras (J. Phys. A 49 (2016) 044001) nevertheless solved the finite-$N$ spectral problem in all these sectors. Their tool is the vertex-IRF transformation, a generalized gauge transformation relating the eight-vertex model to the dynamical six-vertex (SOS) model — an integrable face model in which a height variable lives on lattice faces and the dynamical parameter of the weights shifts with the height — with antiperiodic boundary conditions, a model diagonalized by Sklyanin's separation of variables (SOV). In SOV the transfer matrix is represented on a basis in which its eigenvalues are encoded in the values of a separated function at $N$ discrete inhomogeneity points (the SOV spectrum), bypassing Bethe roots altogether. The vertex-IRF map is bijective in all twisted cases, and non-bijective (two variants with non-intersecting kernels) in the periodic odd-$N$ case, where the eight-vertex spectrum is doubly degenerate and coincides with half of the antiperiodic dynamical six-vertex spectrum. This gives the complete spectrum and explicit eigenstates of the inhomogeneous quasi-periodic eight-vertex transfer matrix at finite $N$.

Two reformulations of this spectral characterization exist. The first is an inhomogeneous $T$-$Q$ equation, Baxter's equation with an extra inhomogeneous term, proven complete in all the sectors studied; its Bethe-type equations contain extra inhomogeneous contributions whose behaviour at large $N$ is hard to control. The second, proposed by Levy-Bencheton, Niccoli and Terras (2016) for the antiperiodic dynamical six-vertex model and transferred to the eight-vertex case, is the ansatz that $Q(\lambda)$ is an elliptic polynomial of degree $N$ with twist-dependent quasi-periods, leading to a homogeneous Baxter-type $T$-$Q$ equation, ordinary Bethe equations, and ABA-type eigenstates; its completeness has been proven only for even $N$. In parallel, the off-diagonal Bethe ansatz (ODBA) of Cao, Yang, Shi and Wang constructs inhomogeneous $T$-$Q$ relations for the periodic and antiperiodic XYZ chains, treating even and odd $N$ simultaneously, and proves that such inhomogeneous functional relations characterize the complete spectrum.

Form factors, the matrix elements of local operators between transfer-matrix eigenstates, are the elementary building blocks of correlation functions. Computing them requires the solution of the quantum inverse problem (expressing local spin operators through monodromy-matrix elements) and control of the vertex-IRF dressing of the off-diagonal monodromy elements, a non-local action on the whole space of states that produces complicated combinatorial relations (Eqs. (3.45)-(3.46) of the source paper). Since 2016, determinant form factors and correlation functions have been obtained by SOV for the antiperiodic XXZ and XXX chains and for open chains with constrained boundary fields, including the open XYZ chain, and by ODBA for the twisted XXZ chain; the antiperiodic XYZ thermodynamic limit has been extracted from the ODBA equations via degenerate crossing parameters. None of these delivers the corresponding results for the elliptic quasi-periodic eight-vertex model itself, whose solution program the source paper leaves explicitly incomplete in its conclusion.

Problem Statement

Consider the inhomogeneous eight-vertex (XYZ) transfer matrix $T^{(8V)}_{(x,y)}(\lambda)$ on $N$ sites at generic crossing parameter $\eta$ (away from roots of unity, where the transfer-matrix spectrum develops extra degeneracies), with integrable quasi-periodic boundary conditions: a twist of the monodromy matrix by $K_{(x,y)}=(\sigma^x)^y(\sigma^z)^x$ with $(x,y)\in\{0,1\}^2$, where $\sigma^x$, $\sigma^y$, $\sigma^z$ are the Pauli matrices ($(x,y)=(1,0)$, $(0,1)$, $(1,1)$ give the $\sigma^z$-, $\sigma^x$- and, up to a phase, $\sigma^y$-twisted transfer matrices), or the periodic case $(x,y)=(0,0)$ with $N$ odd. Take inhomogeneity parameters $\xi_1,\dots,\xi_N$ satisfying the genericity conditions

$\xi_j+\eta/2\notin\pi\mathbb{Z}+\pi\omega\mathbb{Z}\ \text{for all }j,\qquad \xi_a-\xi_b+\epsilon\eta\notin\pi\mathbb{Z}+\pi\omega\mathbb{Z}\ \text{for all }\epsilon\in\{-1,0,1\},\ a\neq b,$

(conditions (2.10)-(2.11) of Niccoli-Terras (2016)), where $\pi$ and $\pi\omega$, with $\operatorname{Im}\omega>0$, are the quasi-periods of the Jacobi theta functions entering Baxter's $R$-matrix; the homogeneous model (all $\xi_j$ equal) is reached as a degenerate limit of these generic inhomogeneities. The finite-$N$ spectrum and SOV eigenstates are known, and the spectrum admits a complete reformulation by an inhomogeneous $T$-$Q$ equation with extra inhomogeneous term. Complete the solution program by resolving the following three linked questions.

(1) Completeness of the homogeneous ansatz for odd $N$. Prove, or disprove, that the homogeneous Baxter-type functional $T$-$Q$ equation with $Q(\lambda)$ an elliptic polynomial of degree $N$ with twist-dependent quasi-periods, together with the associated system of Bethe equations and ABA-type eigenstate construction, completely characterizes the spectrum and eigenstates of the quasi-periodic eight-vertex transfer matrix when $N$ is odd - in the periodic odd-$N$ sector (where the spectrum is doubly degenerate and both vertex-IRF variants are non-bijective) and in the twisted odd-$N$ sectors. A proof must establish the correspondence between solutions of the Bethe-type equations and the full SOV spectrum, including the degeneracy structure; a disproof must exhibit a transfer-matrix eigenvalue or eigenstate not representable within the ansatz.

(2) Thermodynamic limit. Develop an efficient and controlled analysis of the large-$N$ (thermodynamic) limit starting from the inhomogeneous $T$-$Q$ equations and their Bethe-type equations with extra inhomogeneous terms, valid across the quasi-periodic sectors (including the periodic odd-$N$ sector, not only the antiperiodic one). For each such sector the analysis must deliver, with controlled error terms, the large-$N$ (thermodynamic-Bethe-ansatz-type) equations that determine the Bethe-type roots or their root densities, the ground-state energy density, and the elementary excitation spectrum, and must state, with justification, whether the homogeneous $T$-$Q$ ansatz is preferable for such analyses; these are the required outputs of this sub-question. The known degenerate-crossing-parameter extrapolation with $O(N^{-2})$ corrections covers only the antiperiodic sector at spectral level; a solution must either extend such techniques or replace them.

(3) Finite-$N$ form factors and correlation functions. Derive compact formulas for the form factors of local operators (matrix elements between transfer-matrix eigenstates) and for the correlation functions of the quasi-periodic eight-vertex model at finite $N$, overcoming the non-local vertex-IRF dressing of the off-diagonal monodromy-matrix elements (relations (3.45)-(3.46) of the source paper). The accepted output form is a determinant representation of the form factors, from which the correlation functions follow as finite sums over the elementary blocks, or equivalently a finite multiple-sum representation whose number of summed terms grows at most polynomially in $N$ in the homogeneous limit (the compactness bar met by the antiperiodic XXZ and open constrained XYZ analogues); an exact finite-$N$ closed form whose complexity grows faster than polynomially in $N$ (for example a sum over exponentially many configurations) does not meet this bar.

The common objective is to convert the existing finite-$N$ spectral solution into a complete solution of the quasi-periodic eight-vertex model: a spectral characterization valid for all parities and boundary sectors, controlled thermodynamics, and computable observables. Answers may address the three questions jointly or separately, but each answer must address its question at the stated generality (generic elliptic anisotropy, arbitrary $N$ within the stated parity class), not only in rational or trigonometric limits or at special parameter points.

References

1. G. Niccoli and V. Terras, The eight-vertex model with quasi-periodic boundary conditions, J. Phys. A: Math. Theor. 49 (2016) 044001, doi:10.1088/1751-8113/49/4/044001, arXiv:1508.03230, https://arxiv.org/abs/1508.03230
2. D. Levy-Bencheton, G. Niccoli, and 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. 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 (2013) 075003, doi:10.1088/1751-8113/46/7/075003, arXiv:1207.1928, https://arxiv.org/abs/1207.1928
4. R. J. Baxter, Eight-Vertex Model in Lattice Statistics, Phys. Rev. Lett. 26 (1971) 832, doi:10.1103/PhysRevLett.26.832, https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.26.832
5. R. J. Baxter, Completeness of the Bethe ansatz for the six and eight-vertex models, J. Stat. Phys. 108 (2002) 1-48, doi:10.1023/A:1015437118218, arXiv:cond-mat/0111188, https://arxiv.org/abs/cond-mat/0111188
6. N. Grosjean and G. Niccoli, The tau_2-model and the chiral Potts model revisited: completeness of Bethe equations from Sklyanin's SOV method, J. Stat. Mech. (2012) P11005, doi:10.1088/1742-5468/2012/11/P11005, https://iopscience.iop.org/article/10.1088/1742-5468/2012/11/P11005
7. J. Cao, S. Cui, W.-L. Yang, K. Shi, and Y. Wang, Spin-1/2 XYZ model revisit: general solutions via off-diagonal Bethe ansatz, Nucl. Phys. B 886 (2014) 185-201, doi:10.1016/j.nuclphysb.2014.06.026, arXiv:1307.0280, https://arxiv.org/abs/1307.0280
8. J. Cao, W.-L. Yang, K. Shi, and Y. Wang, On the complete-spectrum characterization of quantum integrable spin chains via the inhomogeneous T-Q relation, arXiv:1409.5303, https://arxiv.org/abs/1409.5303
9. Z. Xin, Y. Cao, X. Xu, T. Yang, J. Cao, and W.-L. Yang, Thermodynamic limit of the spin-1/2 XYZ spin chain with the antiperiodic boundary condition, J. High Energy Phys. 2020 (2020) 146, doi:10.1007/jhep12(2020)146, arXiv:2008.13398, https://arxiv.org/abs/2008.13398
10. Z. Xin, Y. Qiao, K. Hao, J. Cao, W.-L. Yang, K. Shi, and Y. Wang, Thermodynamic limit and twisted boundary energy of the XXZ spin chain with antiperiodic boundary condition, Nucl. Phys. B 936 (2018) 501-519, doi:10.1016/j.nuclphysb.2018.10.002, arXiv:1804.06144, https://arxiv.org/abs/1804.06144
11. Y. Qiao, Z. Xin, X. Xu, K. Hao, T. Yang, J. Cao, and W.-L. Yang, Correlation functions of the XXZ spin chain with the twisted boundary condition, J. Math. Phys. 61 (2020) 033502, doi:10.1063/1.5113651, arXiv:1901.09604, https://arxiv.org/abs/1901.09604
12. H. Pei and V. Terras, On scalar products and form factors by separation of variables: the antiperiodic XXZ model, J. Phys. A: Math. Theor. 55 (2022) 015205, doi:10.1088/1751-8121/ac3b85, arXiv:2011.06109, https://arxiv.org/abs/2011.06109
13. G. Niccoli, H. Pei, and V. Terras, Correlation functions by separation of variables: the XXX spin chain, SciPost Phys. 10 (2021) 006, doi:10.21468/SciPostPhys.10.1.006, arXiv:2005.01334, https://arxiv.org/abs/2005.01334
14. G. Niccoli and 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 (2025) 090, doi:10.21468/SciPostPhys.19.4.090, arXiv:2402.04112, https://arxiv.org/abs/2402.04112
15. G. Niccoli and V. Terras, On correlation functions of the open XYZ spin 1/2 chain with boundary fields related by a constraint, arXiv:2507.20299, https://arxiv.org/abs/2507.20299
16. G. Niccoli and V. Terras, On the gauge invariant boundary condition in open XXZ chains: SoV bases, elementary blocks and overlaps, arXiv:2608.22365, https://arxiv.org/abs/2608.22365
17. N. Slavnov, A. Zabrodin, and A. Zotov, Scalar products of Bethe vectors in the 8-vertex model, J. High Energy Phys. 2020 (2020) 123, doi:10.1007/JHEP06(2020)123, arXiv:2005.11224, https://arxiv.org/abs/2005.11224
18. L. Pristyák and B. Pozsgay, Current mean values in the XYZ model, SciPost Phys. 14 (2023) 158, doi:10.21468/SciPostPhys.14.6.158, arXiv:2211.00698, https://arxiv.org/abs/2211.00698
19. K. Hao, F. Wen, J. Cao, G.-L. Li, W.-L. Yang, K. Shi, and Y. Wang, On the solutions of the Z_n-Belavin model with arbitrary number of sites, Nucl. Phys. B 920 (2017) 419-441, doi:10.1016/j.nuclphysb.2017.04.019, arXiv:1609.00953, https://arxiv.org/abs/1609.00953
20. H. Frahm and D. Westerfeld, Density matrices in integrable face models, arXiv:2104.06269, https://arxiv.org/abs/2104.06269