Zinn-Justin's conjecture for the inhomogeneous eigenvector of the eight-vertex model on its combinatorial line and the fully inhomogeneous sum rule

The sum-rule program for the eight-vertex model on its combinatorial line (crossing parameter $\eta=\pi/3$ , odd size $L=2n+1$ , periodic boundary conditions) rests on an unproved conjecture of Zinn-Justin (2012): the inhomogeneous transfer matrix admits an eigenvector $\Psi_L$ whose entries are theta functions of degree $L-1$ in each inhomogeneity with prescribed pseudo-periodicity, with eigenvalue $\prod_i r(x_i-u)$ (which subsumes the Razumov-Stroganov inhomogeneous eigenvalue conjecture). Under this conjecture the source paper computed the associated infinite-cylinder partition function (the "sum rule" generalizing the squared norm of the XYZ ground state) only in a half-specialized regime. Later work proved the homogeneous eigenvalue (Hagendorf-Liénardy 2018), constructed theta-function qKZB solutions for the even-size face model (Finch-Weston-Zinn-Justin 2015), identified the relevant polynomials as Painlevé VI tau functions (Rosengren), and derived conditional sum rules, eigenvector components and nearest-neighbour correlations (Brasseur-Hagendorf 2021, Hagendorf-Rosengren 2024). The open core - prove or refute Conjecture 1, construct $\Psi_L$ rigorously, and determine the fully inhomogeneous partition function $Z_L(x_1,\dots,x_L)$ together with the functional relations it satisfies - remains open as of 2026.

Background

The eight-vertex model is an integrable two-dimensional lattice model of statistical mechanics whose configurations are orientations ("arrows") on the edges of the square lattice, with an even number of arrows pointing into each vertex; eight local configurations are allowed, with Boltzmann weights $a,b,c,d$ . Its row-to-row transfer matrix $T_L(u|x_1,\dots,x_L)$ acts on $(\mathbb{C}^2)^{\otimes L}$ , where $L$ is the number of vertical lines, $u$ is the spectral parameter and the $x_i$ are column inhomogeneities; differentiating the transfer matrix yields the Hamiltonian of the XYZ spin chain, so chain and model are two views of one object. In the elliptic parametrization the weights are built from Jacobi theta functions with nome $p$ ( $p=e^{i\pi\tau}$ , $|p|<1$ ) and depend on a crossing parameter $\eta$ . The "combinatorial line" is the one-parameter family $\eta=\pi/3$ : there the weights obey $(a^2+ab)(b^2+ab)=(c^2+ab)(d^2+ab)$ , the associated XYZ Hamiltonian acquires lattice supersymmetry (it can be written as an anticommutator of nilpotent local operators), and the model is called the supersymmetric eight-vertex model.

A striking phenomenon discovered by Stroganov in 2001 ("the importance of being odd") is that on this line, for an odd number of sites $L=2n+1$ with periodic boundary conditions, the transfer matrix possesses a remarkably simple, doubly degenerate eigenvalue $\Theta_n=(a+b)^{2n+1}$ , whose eigenstates are the ground states of the XYZ chain; the existence of this eigenvalue was proved by Hagendorf and Liénardy in 2018 using the lattice supersymmetry. The trigonometric cousin (the six-vertex model at $\Delta=-1/2$ , the limit $p\to 0$ ) is much better understood: there the ground-state components are polynomials of bounded degree, and the Razumov-Stroganov sum rules (exact identities for the squared norm and related scalar products of the ground state) were proved via polynomial solutions of the quantum Knizhnik-Zamolodchikov (qKZ) equation, an intertwining difference equation satisfied by correlation-type vectors in integrable models.

At the elliptic point, Bazhanov and Mangazeev conjectured exact expressions for certain components of the homogeneous ground state and for its squared norm in terms of polynomials $s_n,\bar s_n$ that they identified as tau functions of the Painlevé VI equation (the sixth of the classical nonlinear ordinary differential equations with the Painlevé property; tau functions are functions whose logarithmic derivatives solve the equation in Hamiltonian form). To attack the problem at the inhomogeneous level - the strategy that succeeded in the six-vertex case - Zinn-Justin (2012) introduced the conjectured ground-state eigenvector $\Psi_L(x_1,\dots,x_L)$ of the inhomogeneous transfer matrix at $\eta=\pi/3$ (his Conjecture 1), studied the partition function $Z_L(x_1,\dots,x_L)=\langle\Psi_L(-x_1,\dots,-x_L)\mid\Psi_L(x_1,\dots,x_L)\rangle$ (a quadratic form generalizing the squared norm of the XYZ ground state, interpretable as the partition function of the inhomogeneous model on an infinite cylinder), and computed it only under the half-specialization $x_{i+n}=-x_i$ , $x_L=0$ , obtaining Pfaffian and determinant (Izergin-Korepin-type) expressions and, in the homogeneous limit, bilinear differential recurrence relations for the squared norm. From Conjecture 1 one derives the exchange relations, the wheel condition (the vanishing of $\Psi_L$ whenever three inhomogeneities form an arithmetic progression $x,x+2\eta,x+4\eta$ ), and recursive reduction relations. The whole derivation is conditional: Conjecture 1 is not proved, no closed form for the fully inhomogeneous $Z_L$ was found, and the paper's prospects section adds the classification of the differential and divided-difference equations satisfied by $Z_L$ ("only the tip of the iceberg"), the relation to Rosengren's results for the eight-vertex-solid-on-solid (8VSOS, or three-colour) model, the connection to Painlevé VI and Bäcklund transformations, the computation of individual eigenvector components and correlation functions, and a combinatorial interpretation of the positive-integer-coefficient polynomials appearing in the homogeneous limit.

A key structural difficulty is elliptic. Unlike the six-vertex case, where the ground-state entries are polynomials and existence/uniqueness follow from qKZ techniques, at $\eta=\pi/3$ with $p\neq 0$ the entries are theta functions, and the natural framework - the quantum Knizhnik-Zamolodchikov-Bernard (qKZB) equation, the elliptic dynamical analogue of qKZ - was shown by Finch, Weston and Zinn-Justin (2015) to provide explicit theta-function eigenvectors only for the face (SOS) model at even size; the odd-size vertex model admits no direct vertex-IRF (interaction-round-a-face) correspondence, and its Conjecture 1 remains unproved.

Problem Statement

Work in the conventions of Zinn-Justin (2012): the eight-vertex model on the square lattice with periodic boundary conditions in the horizontal direction and an odd number $L=2n+1$ of vertical lines, with column inhomogeneities $x_1,\dots,x_L$ and Boltzmann weights

a(x)=\vartheta_4(2\eta,p^2)\,\vartheta_4(x,p^2)\vartheta_1(x+2\eta,p^2),\qquad b(x)=\vartheta_4(2\eta,p^2)\,\vartheta_1(x,p^2)\vartheta_4(x+2\eta,p^2),

c(x)=\vartheta_1(2\eta,p^2)\,\vartheta_4(x,p^2)\vartheta_4(x+2\eta,p^2),\qquad d(x)=\vartheta_1(2\eta,p^2)\,\vartheta_1(x,p^2)\vartheta_1(x+2\eta,p^2),

where $\vartheta_1,\vartheta_4$ are Jacobi theta functions of nome $p^2$ , $p=e^{i\pi\tau}$ with $\operatorname{Im}\tau>0$ , and the crossing parameter is fixed to $\eta=\pi/3$ (the combinatorial line). Let $R(x)$ be the corresponding eight-vertex $R$ -matrix acting on $(\mathbb{C}^2)^{\otimes2}$ ,

R(x)=\begin{pmatrix}a(x)&0&0&d(x)\\0&b(x)&c(x)&0\\0&c(x)&b(x)&0\\d(x)&0&0&a(x)\end{pmatrix},

and let the transfer matrix be

T_L(u|x_1,\dots,x_L)=\operatorname{Tr}_0\,R_{01}(x_1-u)R_{02}(x_2-u)\cdots R_{0L}(x_L-u),

acting on $(\mathbb{C}^2)^{\otimes L}$ , each operator acting on the tensor factors indicated by its subscripts. Let $F=\prod_i\sigma^x_i$ be the spin-reversal operator on $(\mathbb{C}^2)^{\otimes L}$ , and $r(x)=a(x)+b(x)=\vartheta_4(0,p^2)\vartheta_1(x+\eta,p^2)\vartheta_4(x+\eta,p^2)$ .

(1) Settle Conjecture 1 of Zinn-Justin (2012). Prove, or refute by an explicit counterexample at some odd size $L$ , that the eigenvector problem

T_L(u|x_1,\dots,x_L)\,\Psi_L=\Big(\prod_{i=1}^{L}r(x_i-u)\Big)\,\Psi_L,\qquad F\,\Psi_L=(-1)^n\,\Psi_L

admits a solution $\Psi_L(x_1,\dots,x_L)$ whose entries are theta functions of degree $L-1=2n$ and nome $p^2$ in each variable $x_i$ - holomorphic, generically non-zero, and without common factor - with the pseudo-periodicity laws

\Psi_L(\dots,x_i+2\pi\tau,\dots)=p^{-4n}\,z_i^{2n}\prod_{j\ne i}z_j^{-1}\,\Psi_L(\dots,x_i,\dots),\qquad \Psi_L(\dots,x_i+\pi,\dots)=\prod_{j\ne i}\sigma^z_j\,\Psi_L(\dots,x_i,\dots),

where $z_i=e^{-2ix_i}$ and $\sigma^z_j$ is the Pauli matrix acting on site $j$ . A proof settles in particular the conjecture of Razumov and Stroganov (2009) that $\prod_{i=1}^{L}r(x_i-u)$ is a doubly degenerate eigenvalue of the inhomogeneous transfer matrix at $\eta=\pi/3$ (only the homogeneous case is proved).

(2) Determine the fully inhomogeneous sum rule. With $\Psi_L$ normalized as in (1) (its overall multiplicative constant being fixed as in the source, so that the reduction recurrence for $Z_L$ holds without an extra constant), determine, for arbitrary inhomogeneities, the cylinder partition function

Z_L(x_1,\dots,x_L)=\langle\Psi_L(-x_1,\dots,-x_L)\mid\Psi_L(x_1,\dots,x_L)\rangle,

lifting the half-specialization $x_{i+n}=-x_i$ , $x_L=0$ under which it is currently known only conditionally; and investigate the structure of the set of functional relations (recurrence, divided-difference and differential equations) satisfied by $Z_L$ , the relations derived in the source being in its author's words "only the tip of the iceberg". This last part is settled by any non-trivial structural result on that relation set - for instance, exhibiting at least one functional relation for $Z_L$ beyond those derived in the source, or determining the complete set of relations of a specified type.

The common objective of (1) and (2) is to put the sum-rule program of the supersymmetric eight-vertex model on the same rigorous footing as its proved trigonometric ( $\Delta=-1/2$ ) counterpart. Resolving (1) in either direction settles the central conjecture: a proof constructs $\Psi_L$ , whereas an explicit counterexample refutes its existence, in which case (2) is void because $Z_L$ is defined only through $\Psi_L$ . A complete answer along the positive branch - a proof of (1) together with (2) - makes the currently conditional results of the program - the homogeneous sum rules, the individual eigenvector components, and the nearest-neighbour correlation functions of the supersymmetric XYZ chain with their Painlevé VI identities - unconditional.

References

P. Zinn-Justin, "Sum Rule for the Eight-Vertex Model on Its Combinatorial Line", in: Symmetries, Integrable Systems and Representations (Springer Proceedings in Mathematics and Statistics 40), Springer, 2013, pp. 599-637; arXiv:1202.4420; DOI 10.1007/978-1-4471-4863-0_26; https://arxiv.org/abs/1202.4420
Yu. G. Stroganov, "The importance of being odd", J. Phys. A: Math. Gen. 34 (2001) L179-L185; arXiv:cond-mat/0012035; DOI 10.1088/0305-4470/34/13/104; https://arxiv.org/abs/cond-mat/0012035
A. V. Razumov, Yu. G. Stroganov, "A possible combinatorial point for XYZ-spin chain", arXiv:0911.5030 (2009); https://arxiv.org/abs/0911.5030
V. V. Mangazeev, V. V. Bazhanov, "The eight-vertex model and Painlevé VI equation II: eigenvector results", J. Phys. A: Math. Theor. 43 (2010) 085206; arXiv:0912.2163; DOI 10.1088/1751-8113/43/8/085206; https://arxiv.org/abs/0912.2163
A. V. Razumov, Yu. G. Stroganov, P. Zinn-Justin, "Polynomial solutions of qKZ equation and ground state of XXZ spin chain at Delta = -1/2", J. Phys. A: Math. Theor. 40 (2007) 11827-11847; arXiv:0704.3542; DOI 10.1088/1751-8113/40/39/009; https://arxiv.org/abs/0704.3542
H. Rosengren, "The three-colour model with domain wall boundary conditions", Adv. Appl. Math. 46 (2011) 481-535; arXiv:0911.0561; DOI 10.1016/j.aam.2010.10.007; https://arxiv.org/abs/0911.0561
H. Rosengren, "Special polynomials related to the supersymmetric eight-vertex model. III. Painlevé VI equation", arXiv:1405.5318 (2014); https://arxiv.org/abs/1405.5318
H. Rosengren, "Special Polynomials Related to the Supersymmetric Eight-Vertex Model: A Summary", Comm. Math. Phys. 340 (2015) 1143-1170; arXiv:1503.02833; DOI 10.1007/s00220-015-2439-0; https://arxiv.org/abs/1503.02833
H. Rosengren, "Elliptic Pfaffians and solvable lattice models", J. Stat. Mech. (2016) 083106; arXiv:1605.02915; DOI 10.1088/1742-5468/2016/08/083106; https://arxiv.org/abs/1605.02915
P. E. Finch, R. A. Weston, P. Zinn-Justin, "Theta function solutions of the quantum Knizhnik-Zamolodchikov-Bernard equation for a face model", J. Phys. A: Math. Theor. 49 (2016) 064001; arXiv:1509.04594; DOI 10.1088/1751-8113/49/6/064001; https://arxiv.org/abs/1509.04594
C. Hagendorf, J. Liénardy, "On the transfer matrix of the supersymmetric eight-vertex model. I. Periodic boundary conditions", J. Stat. Mech. (2018) 033106; arXiv:1711.04397; DOI 10.1088/1742-5468/aab01d; https://arxiv.org/abs/1711.04397
S. Brasseur, C. Hagendorf, "Sum rules for the supersymmetric eight-vertex model", J. Stat. Mech. (2021) 023102; arXiv:2009.14077; DOI 10.1088/1742-5468/abda28; https://arxiv.org/abs/2009.14077
L. Hietala, "A Combinatorial Description of Certain Polynomials Related to the XYZ Spin Chain", SIGMA 16 (2020) 101; arXiv:2004.09924; DOI 10.3842/SIGMA.2020.101; https://arxiv.org/abs/2004.09924
L. Hietala, "A Combinatorial Description of Certain Polynomials Related to the XYZ Spin Chain. II. The Polynomials p_n", SIGMA 18 (2022) 036; arXiv:2104.04651; DOI 10.3842/SIGMA.2022.036; https://arxiv.org/abs/2104.04651
C. Hagendorf, H. Rosengren, "Nearest-Neighbour Correlation Functions for the Supersymmetric XYZ Spin Chain and Painlevé VI", Comm. Math. Phys. 405 (2024) 97; arXiv:2208.13533; DOI 10.1007/s00220-024-04977-w; https://arxiv.org/abs/2208.13533