Prove the inhomogeneous eigenvalue formula for the supersymmetric eight-vertex model on a strip (Hagendorf–Liénardy Conjecture 5.1)

The square-lattice eight-vertex model at the supersymmetric point — crossing parameter $\eta=\pi/3$, where the vertex weights automatically satisfy $(a^2+ab)(b^2+ab)=(c^2+ab)(d^2+ab)$ — admits, on a strip with open boundary conditions encoded by Sklyanin $K$-matrices, a remarkably simple transfer-matrix eigenvalue $\Lambda_L=(a+b)^{2L}\,\mathrm{tr}(K^+K^-)$, proven by Hagendorf and Liénardy in 2020 for the homogeneous strip. Their Conjecture 5.1 predicts the generalization to arbitrary inhomogeneity parameters $u_1,\dots,u_L$ on the vertical lines: the double-row transfer matrix then possesses the eigenvalue $\Lambda_L(u)=\mathrm{tr}(K^+(u)K^-(u))\prod_{j=1}^L(a(u+u_j)+b(u+u_j))(a(u-u_j)+b(u-u_j))$, with the $K$-matrix evaluated at the boundary parameter $t=\pi/6$. The conjecture was checked numerically for small $L$ in the trigonometric limit $p\to0$ and is compatible with the transfer matrix's functional equations; its trigonometric limit has since been proven rigorously via the boundary quantum Knizhnik–Zamolodchikov equations. A proof for the genuine elliptic regime $0<p<1$ remains open. The problem is to prove, or disprove, this conjectured inhomogeneous eigenvalue formula in the elliptic regime.

Background

Baxter's eight-vertex model is a statistical-mechanical model on the square lattice in which each vertex carries one of eight arrow configurations compatible with the ice rule (an even number of arrows points into every vertex), weighted by four numbers $a,b,c,d$. Its partition function is governed by the transfer matrix, built from the $R$-matrix — the $4\times4$ operator on two two-dimensional spin spaces whose non-zero entries are the vertex weights. In the standard parameterization the weights depend on a spectral parameter $u$, a crossing parameter $\eta$, and an elliptic nome $0<p<1$ through Jacobi theta functions; for instance $a(u)=\rho\,\vartheta_4(2\eta,p^2)\vartheta_4(u,p^2)\vartheta_1(u+2\eta,p^2)$ and $b(u)=\rho\,\vartheta_4(2\eta,p^2)\vartheta_1(u,p^2)\vartheta_4(u+2\eta,p^2)$. The Yang–Baxter equation (the fundamental quadratic consistency relation between $R$-matrices acting on three spin spaces, which makes the vertex weights compatible under rearrangement of the lattice) implies that transfer matrices with different spectral parameters commute, and in the limit $p\to0$ the theta functions degenerate to trigonometric ones and the model becomes the six-vertex model (the trigonometric or "massless" limit of the elliptic theory).

On a strip with $L$ vertical lines and open (fixed) boundary conditions, Sklyanin's construction replaces the periodic trace by a double-row transfer matrix $\mathcal{T}(u)=\mathrm{tr}_0(K^+_0(u)\,U_{0,[1,L]}\,K^-_0(u)\,\bar U_{0,[1,L]})$, where $U$ and $\bar U$ are ordered products of $R$-matrices coupling the auxiliary space $V_0=\mathbb{C}^2$ to the $L$ vertical lines, and where the $K$-matrices ($2\times2$ operators on $V_0$) encode the boundary conditions; they must solve the reflection equation, the boundary analogue of the Yang–Baxter equation. The choice $K^-(u)=K(u)$, $K^+(u)=K(u+2\eta)$ yields a commuting family whose logarithmic derivative at $u=0$ is the Hamiltonian of an open XYZ spin-$\frac12$ chain: a chain with fully anisotropic nearest-neighbour couplings $J_1\sigma^1_j\sigma^1_{j+1}+J_2\sigma^2_j\sigma^2_{j+1}+J_3\sigma^3_j\sigma^3_{j+1}$ and boundary magnetic fields, $\sigma^\alpha$ denoting the Pauli matrices.

A distinguished locus is the combinatorial or supersymmetric line, where the weights obey $(a^2+ab)(b^2+ab)=(c^2+ab)(d^2+ab)$; at the supersymmetric point $\eta=\pi/3$ this relation holds automatically for every $0<p<1$. There the associated XYZ Hamiltonian, with anisotropy parameters $J_1=1+\zeta$, $J_2=1-\zeta$, $J_3=(\zeta^2-1)/2$ and boundary fields parameterized by a complex number $y$, exhibits dynamical lattice supersymmetry in the sense of Witten: up to a shift and rescaling it is the anticommutator $\{Q,Q^\dagger\}$ of a nilpotent supercharge $Q$ (a local operator that annihilates and creates pairs of adjacent spins and obeys $Q^2=0$). States annihilated by both $Q$ and $Q^\dagger$ are supersymmetry singlets, and when they exist they are ground states. Stroganov (2001) conjectured that for periodic boundary conditions and an odd number $L=2n+1$ of vertical lines the transfer matrix on this line possesses the doubly degenerate eigenvalue $\Theta_n=(a+b)^{2n+1}$; Hagendorf and Liénardy proved this in 2018 using the supersymmetry. The line is a meeting point of several further structures: the enumerative combinatorics of alternating sign matrices (square arrays of $0,\pm1$ entries with unit row and column sums whose non-zero entries alternate in sign, which the model counts in special limits), functional equations for the transfer matrix, and tau-function solutions of the Painlevé VI equation (the sixth and most general of the classical nonlinear ordinary differential equations with the Painlevé property).

For open boundary conditions, the same authors proved in 2020 the companion statement: when $y$ solves the biquadratic equation $\zeta(1+y^4)-(3-\zeta^2)y^2=0$ — in particular at the point $t=\pi/6$ of the parametrization $y=\vartheta_1(t,p^2)/\vartheta_4(t,p^2)$, where $\zeta=(\vartheta_1(2\pi/3,p^2)/\vartheta_4(2\pi/3,p^2))^2$ — the strip transfer matrix possesses the non-degenerate eigenvalue $\Lambda_L=(a+b)^{2L}\,\mathrm{tr}(K^+K^-)$, which is the largest eigenvalue for positive weights, and whose eigenspace is the one-dimensional space of supersymmetry singlets.

All of these results concern the homogeneous model, in which every vertical line carries the same spectral parameter. Allowing distinct inhomogeneity parameters $u_1,\dots,u_L$ on the vertical lines is the standard device that makes eigenvectors — not only eigenvalues — accessible: analyticity and reduction properties in the $u_j$ support Izergin–Korepin-type induction arguments (an induction on the number of vertical lines built on determinant representations) and determinant formulae for sum rules (exact expressions for sums of eigenvector components). For periodic boundary conditions, Razumov and Stroganov (2010) conjectured, and Zinn-Justin (2013) sharpened into an eigenvalue-plus-eigenvector conjecture, that the inhomogeneous transfer matrix at $\eta=\pi/3$ with $L=2n+1$ vertical lines possesses the doubly degenerate eigenvalue $\Theta_n(u|u_1,\dots,u_L)=\prod_{i=1}^{2n+1}(a(u_i-u)+b(u_i-u))$; this periodic conjecture remains unproven — Brasseur and Hagendorf (2021) derive sum-rule consequences from it but must assume Zinn-Justin's conjecture. The open-boundary analogue is the conjecture at issue here. Hagendorf and Liénardy's Conjecture 5.1 states that the inhomogeneous double-row transfer matrix at $\eta=\pi/3$, with the $K$-matrix evaluated at $t=\pi/6$, possesses a simple product-form eigenvalue (written out in the problem statement). It was checked numerically for small $L$ in the trigonometric limit $p\to0$ and found compatible with the functional equations obeyed by the transfer matrix and with simplifications at special spectral parameters. The trigonometric limit has since been proven: Hagendorf and Liénardy (2021) constructed a contour-integral solution of the boundary quantum Knizhnik–Zamolodchikov equations — a $q$-deformed analogue of the Knizhnik–Zamolodchikov equations of conformal field theory whose solutions at roots of unity yield transfer-matrix eigenvectors — and showed that at $q=e^{2\pi i/3}$ it diagonalizes the inhomogeneous six-vertex strip transfer matrix with exactly the conjectured eigenvalue. For the genuine elliptic regime $0<p<1$ no proof is known, and the homogeneous supersymmetry argument does not extend, because the inhomogeneous eigenvector is no longer the supersymmetry singlet.

Problem Statement

Prove, or disprove, Conjecture 5.1 of Hagendorf and Liénardy (J. Stat. Mech. 2020, 033104), in the full elliptic regime.

Fix the crossing parameter $\eta=\pi/3$ and an elliptic nome $0<p<1$, and take the eight-vertex weights in the theta-function parameterization of the source paper: $a(u)=\rho\,\vartheta_4(2\eta,p^2)\,\vartheta_4(u,p^2)\,\vartheta_1(u+2\eta,p^2),\qquad b(u)=\rho\,\vartheta_4(2\eta,p^2)\,\vartheta_1(u,p^2)\,\vartheta_4(u+2\eta,p^2),$ $c(u)=\rho\,\vartheta_1(2\eta,p^2)\,\vartheta_4(u,p^2)\,\vartheta_4(u+2\eta,p^2),\qquad d(u)=\rho\,\vartheta_1(2\eta,p^2)\,\vartheta_1(u,p^2)\,\vartheta_1(u+2\eta,p^2),$ where $\rho$ is an arbitrary non-zero constant and $\vartheta_j(\cdot,q)$ denotes the Jacobi theta function of nome $q$. Let $R(u)$ be the corresponding $R$-matrix, which in the canonical basis $|{\uparrow\uparrow}\rangle,|{\uparrow\downarrow}\rangle,|{\downarrow\uparrow}\rangle,|{\downarrow\downarrow}\rangle$ of $V\otimes V$ with $V=\mathbb{C}^2$ reads $R(u)=\begin{pmatrix}a(u)&0&0&d(u)\\0&b(u)&c(u)&0\\0&c(u)&b(u)&0\\d(u)&0&0&a(u)\end{pmatrix}.$ Let $K(u)=1+\sum_{\alpha=1}^{3}\frac{\vartheta_1(u,p)}{\vartheta_{5-\alpha}(u,p)}\,\mu_\alpha\,\sigma^\alpha$ be the general reflection-equation $K$-matrix of the eight-vertex model, with Pauli matrices $\sigma^\alpha$ and with the coefficients $\mu_1=\frac{\vartheta_4(\eta,p)}{\vartheta_1(\eta,p)}\,\frac{2\,\mathrm{Re}\,y}{1+|y|^2},\qquad \mu_2=\frac{\vartheta_3(\eta,p)}{\vartheta_1(\eta,p)}\,\frac{2\,\mathrm{Im}\,y}{1+|y|^2},\qquad \mu_3=\frac{\vartheta_2(\eta,p)}{\vartheta_1(\eta,p)}\,\frac{1-|y|^2}{1+|y|^2},$ evaluated at the boundary parameter value $y=\vartheta_1(\pi/6,p^2)/\vartheta_4(\pi/6,p^2)$ — the point $t=\pi/6$ of the parametrization $y=\vartheta_1(t,p^2)/\vartheta_4(t,p^2)$ — and set $K^-(u)=K(u)$, $K^+(u)=K(u+2\eta)$. For $L\geqslant1$ vertical lines and arbitrary inhomogeneity parameters $u_1,\dots,u_L$ (the matrix entries are entire functions of these parameters, so the natural domain is all complex values), define the inhomogeneous double-row transfer matrix $\mathcal{T}(u|u_1,\dots,u_L)=\mathrm{tr}_0\!\left(K^+_0(u)\,U_{0,[1,L]}(u|\vec u)\,K^-_0(u)\,\bar U_{0,[1,L]}(u|\vec u)\right),$ with $U_{0,[1,L]}=R_{0L}(u+u_L)\cdots R_{01}(u+u_1)$ and $\bar U_{0,[1,L]}=R_{01}(u-u_1)\cdots R_{0L}(u-u_L)$.

Claim to be settled: for every $L\geqslant1$, every spectral parameter $u$, and all inhomogeneity parameters $u_1,\dots,u_L$, this transfer matrix possesses the eigenvalue $\Lambda_L(u)=\mathrm{tr}\!\left(K^+(u)K^-(u)\right)\prod_{j=1}^{L}\bigl(a(u+u_j)+b(u+u_j)\bigr)\bigl(a(u-u_j)+b(u-u_j)\bigr).$

The statement to be established is the full elliptic one, $0<p<1$; the already-proven homogeneous case $u_1=\cdots=u_L=0$ (where the formula reduces to $\Lambda_L=(a+b)^{2L}\,\mathrm{tr}(K^+K^-)$) and the already-proven trigonometric limit $p\to0$ must be recovered as special cases. A disproof consists of an explicit instance with $0<p<1$, a specified $L\geqslant1$ and specified $u,u_1,\dots,u_L$ for which the displayed number is demonstrably not an eigenvalue of the displayed finite matrix.

References

1. Christian Hagendorf and Jean Liénardy, 'On the transfer matrix of the supersymmetric eight-vertex model. II. Open boundary conditions', J. Stat. Mech. (2020) 033104. DOI: 10.1088/1742-5468/ab7748; arXiv:1911.03348. https://doi.org/10.1088/1742-5468/ab7748 (source of the open question)
2. Christian Hagendorf and Jean Liénardy, 'On the transfer matrix of the supersymmetric eight-vertex model. I. Periodic boundary conditions', J. Stat. Mech. (2018) 033106. DOI: 10.1088/1742-5468/aab01d; arXiv:1711.04397. https://doi.org/10.1088/1742-5468/aab01d
3. Christian Hagendorf and Jean Liénardy, 'The open XXZ chain at Δ = −1/2 and the boundary quantum Knizhnik–Zamolodchikov equations', J. Stat. Mech. (2021) 013104. DOI: 10.1088/1742-5468/abd028; arXiv:2008.03220. https://doi.org/10.1088/1742-5468/abd028
4. Sandrine Brasseur and Christian Hagendorf, 'Sum rules for the supersymmetric eight-vertex model', J. Stat. Mech. (2021) 023102. DOI: 10.1088/1742-5468/abda28; arXiv:2009.14077. https://doi.org/10.1088/1742-5468/abda28
5. Christian Hagendorf and Hjalmar Rosengren, 'Nearest-Neighbour Correlation Functions for the Supersymmetric XYZ Spin Chain and Painlevé VI', Comm. Math. Phys. 405, article 97 (2024). DOI: 10.1007/s00220-024-04977-w; arXiv:2208.13533. https://doi.org/10.1007/s00220-024-04977-w
6. Jean Liénardy and Christian Walmsley Hagendorf, 'The open XXZ chain at Δ = −1/2 and totally-symmetric alternating sign matrices', arXiv:2509.09209 (2025). https://arxiv.org/abs/2509.09209
7. A. V. Razumov and Yu. G. Stroganov, 'A possible combinatorial point for the XYZ spin chain', Theor. Math. Phys. 164 (2010) 977–991. DOI: 10.1007/s11232-010-0078-3. https://doi.org/10.1007/s11232-010-0078-3
8. Paul Zinn-Justin, 'Sum Rule for the Eight-Vertex Model on Its Combinatorial Line', in K. Iohara, S. Morier-Genoud, B. Rémy (eds.), Symmetries, Integrable Systems and Representations, Springer Proceedings in Mathematics & Statistics 40, Springer, London (2013), pp. 599–637. DOI: 10.1007/978-1-4471-4863-0_26. https://doi.org/10.1007/978-1-4471-4863-0_26
9. Yu. G. Stroganov, 'The importance of being odd', J. Phys. A: Math. Gen. 34 (2001) L179–L185. DOI: 10.1088/0305-4470/34/13/104. https://doi.org/10.1088/0305-4470/34/13/104