Completeness of the Bethe ansatz and rationality of Baxter polynomials for twisted transfer-matrices of quantum affine algebras

The transfer-matrices of quantum integrable lattice models associated with an untwisted quantum affine algebra $U_q(\widehat{\mathfrak{g}})$ form a commuting family whose joint eigenvalues encode the spectra of the models. Frenkel and Hernandez (Duke Math. J. 164 (2015) 2407-2460) proved, in a twist-deformed setting, that every eigenvalue on a weight subspace is obtained from the q-character of the auxiliary representation by a substitution involving certain Baxter polynomials $Q_i(z,u)$, and they derived the corresponding generalized Bethe ansatz equations. Two analytic questions posed there remain open. First, the completeness conjecture (their Conjecture 5.15): for generic $q$ and generic highest $\ell$-weights there is a one-to-one correspondence between the eigenvalues of the twisted transfer-matrices on a weight subspace and the solutions of the generalized Bethe ansatz equations. Second (their Remark 5.12(ii)): the coefficients of the Baxter polynomials, a priori algebraic elements over the ring of formal Taylor series in the twist variables, are expected to be expansions of rational functions in those variables, with systematically describable poles. The subsequent literature (Baxter 2002; Feigin-Jimbo-Miwa-Mukhin 2017; Tarasov 2018; Frenkel-Koroteev-Sage-Zeitlin 2024; Varagnolo-Vasserot 2023; and later work of Frenkel-Hernandez through 2025) establishes only special cases and one direction of the correspondence in a related framework, so both questions remain open at the source's generality.

Background

A quantum affine algebra $U_q(\widehat{\mathfrak{g}})$ is a Hopf-algebra deformation of the universal enveloping algebra of an untwisted affine Kac-Moody algebra, where $\widehat{\mathfrak{g}}$ is built from a finite-dimensional complex simple Lie algebra $\mathfrak{g}$ by adjoining a loop variable; $I=\{1,\dots,n\}$ indexes the simple roots $\alpha_i$ of $\mathfrak{g}$, $C=(C_{i,j})$ is the Cartan matrix, $q$ is a complex parameter taken not to be a root of unity, and $q_i=q^{d_i}$ for the standard symmetrizing integers $d_i$. The irreducible finite-dimensional representations of $U_q(\widehat{\mathfrak{g}})$ are classified by Drinfeld polynomials: an $n$-tuple $P_i(z)$, $i\in I$, of polynomials with constant term 1, equivalently by the highest $\ell$-weight, a tuple of rational functions $\Psi_i(z)$ determined by the $P_i$. The archetype of the associated quantum integrable lattice models is the six-vertex (XXZ) chain. Its transfer-matrices are built from the universal R-matrix, the canonical intertwiner solving the Yang-Baxter equation: for each finite-dimensional auxiliary representation $V$ and spectral parameter $z$, the transfer-matrix $t_V(z)$ acts on the quantum space $W$, and the family $\{t_V(z)\}$ commutes; its joint spectrum determines the thermodynamics of the model. The q-character $\chi_q(V)$, introduced by Frenkel and Reshetikhin, is an injective map assigning to a representation a formal sum of monomials in variables $Y_{i,a}$ ($i\in I$, $a\in\mathbb{C}^\times$) encoding its weight decomposition; it serves as a generating function for transfer-matrix eigenvalues. Frenkel and Reshetikhin conjectured in 1999 that every eigenvalue of $t_V(z)$ is obtained from $\chi_q(V)$ by a uniform substitution of the variables $Y_{i,a}$ — the analytic Bethe ansatz — with Baxter's classical TQ-relation (expressing the product of a transfer-matrix eigenvalue $T(z)$ with a polynomial $Q(z)$ as a linear combination of $Q$ at shifted arguments) as the rank-one shadow.

Frenkel and Hernandez (Duke Math. J. 164 (2015) 2407-2460) proved this conjecture in a deformed setting. Their construction rests on the Borel subalgebra $U_q(\mathfrak{b})$ of $U_q(\widehat{\mathfrak{g}})$ and the category $\mathcal{O}$ of its representations introduced by Hernandez and Jimbo, which contains infinite-dimensional limits of Kirillov-Reshetikhin modules (a distinguished family of finite-dimensional modules) called prefundamental representations $L^{\pm}_{i,a}$, one for each simple root $i$ and parameter $a\in\mathbb{C}^\times$; the transfer-matrix of $L^{+}_{i,a}$ is a generalized Baxter Q-operator, and its eigenvalue is the analogue of Baxter's polynomial $Q$. Frenkel and Hernandez replace the ordinary trace by a twisted trace functional $\operatorname{Tr}_{V,u}$ depending on twist parameters $u_i\in\mathbb{C}^\times$, $i\in I$ (reducing to the ordinary trace when all $u_i=1$), obtaining twisted transfer-matrices $t_V(z,u)$ that act on every finite-dimensional $W$, commute with one another (their Theorem 5.3), and have coefficients in formal Taylor series in the variables $v_i=\prod_{j\in I}u_j^{C_{j,i}}$. Their main theorem (Theorem 5.11) states: for $W=\otimes_{j=1}^N L(\boldsymbol{\Psi}_j)$ a tensor product of finite-dimensional irreducibles of highest weight $\omega$, every eigenvalue of $t_V(z,u)$ on the weight subspace $(W)_\lambda$ equals $\chi_q(V)$ in which each variable $Y_{i,a}$ is replaced by

$$q_i^{ht_i(\omega-\lambda)}a_i u_i\frac{f_i(azq_i^{-1})Q_i(azq_i^{-1},u)}{f_i(azq_i)Q_i(azq_i,u)},$$

where $ht_i(\omega-\lambda)$ is the multiplicity of $\alpha_i$ in $\omega-\lambda$, the scalars $a_i$ depend only on $W$, $f_i(z)$ is an explicit formal power series, and $Q_i(z,u)$ is a polynomial in $z$ of degree $ht_i(\omega-\lambda)$, depending on the eigenvalue — the Baxter polynomial. Writing $Q_i(z,u)=\prod_{k=1}^{m_i}(w^{(i)}_k-z)$ with $m_i=ht_i(\omega-\lambda)$ and demanding that the poles at the roots $w^{(i)}_k$ (which are extraneous, i.e. not forced by the Drinfeld polynomials of $W$) cancel uniformly in $V$ yields the generalized Bethe ansatz equations, equation (5.23) of the source, displayed in the problem statement. Two steps of this derivation are conditional in the source: the cancellation is only expected, at least for generic $q$, to occur between adjacent monomials of the q-character, and each root of $Q_i(z,u)$ is expected to be simple. The resulting open questions are the completeness conjecture (Conjecture 5.15, described by the authors as a version of the completeness of the Bethe ansatz at the level of eigenvalues) and the rationality of the coefficients of $Q_i(z,u)$ in the twist variables (Remark 5.12(ii)); in contrast, rationality of the eigenvalues in the spectral variable $z$ is proved in the same paper (Remark 5.12(i)).

For context on completeness in special cases: Baxter argued at physics level of rigor for the six- and eight-vertex models; Tarasov established a bijection between eigenvalues and solutions of an algebraic system that reduces to the standard Bethe equations under a coprimality condition, for the periodic isotropic (rational) Heisenberg chain with arbitrary spins and inhomogeneities, via discrete Wronskians; Feigin-Jimbo-Miwa-Mukhin established the eigenvalue-to-solution direction in a related framework of finite-type modules over the Borel subalgebra. On the geometric side, Frenkel-Koroteev-Sage-Zeitlin parametrize nondegenerate Bethe solutions by $(G,q)$-opers (a $q$-difference analogue of Drinfeld-Sokolov opers on flag varieties), Hernandez and Leclerc interpret Baxter relations as cluster-algebra mutations in the Grothendieck ring of category $\mathcal{O}$, and Varagnolo-Vasserot and Pushkar-Smirnov-Zeitlin realize prefundamental modules and Q-operators in critical and quantum K-theory of quiver varieties; none of these yields the conjectured bijection.

Problem Statement

Let $\widehat{\mathfrak{g}}$ be an untwisted affine Kac-Moody algebra with finite part $\mathfrak{g}$, Cartan matrix $C=(C_{i,j})_{i,j\in I}$, and $q_i=q^{d_i}$. Let $q$ be generic ($q$ is not a root of unity, and the parameters below avoid a proper closed algebraic locus of degenerate values). Let $W=\otimes_{j=1}^N L(\boldsymbol{\Psi}_j)$ be a tensor product of finite-dimensional irreducible $U_q(\widehat{\mathfrak{g}})$-modules whose highest $\ell$-weights are given by

$$\boldsymbol{\Psi}_j=(\Psi_{j,i}(z))_{i\in I},\qquad \Psi_{j,i}(z)=q_i^{\deg P_{j,i}}\frac{P_{j,i}(zq_i^{-1})}{P_{j,i}(zq_i)},$$

for generic Drinfeld polynomials $P_{j,i}$. Let $\omega$ be the highest weight of $W$, $\lambda$ a weight of $W$, and $m_i=ht_i(\omega-\lambda)$ the multiplicity of $\alpha_i$ in $\omega-\lambda$. Let $t_V(z,u)$ be the twisted transfer-matrices of Frenkel-Hernandez acting on $W$, with twist parameters $u_i$ and $v_i=\prod_{j\in I}u_j^{C_{j,i}}$, and let $(W)_\lambda$ be the $\lambda$-weight subspace. Consider the generalized Bethe ansatz equations for unknowns $w^{(i)}_k$, $k=1,\dots,m_i$, $i\in I$:

$$v_i\prod_{j=1}^N q_i^{\deg P_{j,i}}\frac{P_{j,i}(q_i^{-1}/w^{(i)}_k)}{P_{j,i}(q_i/w^{(i)}_k)}=\prod_{s\neq k}q_i^2\frac{w^{(i)}_k-w^{(i)}_s q_i^{-2}}{w^{(i)}_k-w^{(i)}_s q_i^2}\prod_{l\neq i}\prod_{s=1}^{m_l}q^{C_{li}}\frac{w^{(i)}_k-w^{(l)}_s q^{-C_{li}}}{w^{(i)}_k-w^{(l)}_s q^{C_{li}}}.$$

(a) Completeness (Conjecture 5.15 of Frenkel-Hernandez). Prove, or disprove, that for generic $q$ and generic polynomials $P_{j,i}$ there is a one-to-one correspondence between the simultaneous eigenvalues of the transfer-matrices $t_V(z,u)$ on $(W)_\lambda$ and the solutions of the Bethe ansatz equations above, the correspondence being mediated by the Baxter polynomials $Q_i(z,u)=\prod_{k=1}^{m_i}(w^{(i)}_k-z)$ appearing in the eigenvalue formula of their Theorem 5.11. A complete proof must establish both directions: (i) the roots of the Baxter polynomials of every eigenvalue satisfy the Bethe ansatz equations, and (ii) conversely, every solution of the equations arises from an eigenvalue. Since the statement is generic, a disproof must refute the correspondence at the level of generic parameters; a failure confined to a proper closed degenerate locus of the parameters would not suffice.

(b) Rationality in the twist variables (Remark 5.12(ii) of Frenkel-Hernandez). Prove, or disprove, that for each $i\in I$ and each $z^m$-coefficient of the Baxter polynomial $Q_i(z,u)$ — which a priori is an element of an algebraic closure of the field of fractions of the ring of formal Taylor power series in the $v_j$, $j\in I$ — the coefficient is the expansion near $v_j=0$ of a rational function in the $v_j$; and, if true, describe systematically the locations and orders of the poles of these rational functions.

The common objective of (a) and (b) is to complete the analytic Bethe ansatz for models associated with arbitrary untwisted quantum affine algebras at generic $q$: (a) makes the Baxter-root parametrization of the twisted transfer-matrix spectrum complete (bijective, with correct counting), and (b) establishes that the parametrizing coefficients are rational functions of the twist variables with controlled poles, which is what allows the twist to be specialized and the ordinary transfer-matrix spectrum to be recovered. Both questions concern the same objects and were posed together in Section 5 of the source.

References

1. Edward Frenkel, David Hernandez, Baxter's Relations and Spectra of Quantum Integrable Models, Duke Math. J. 164 (2015) no. 12, 2407-2460, DOI 10.1215/00127094-3146282, arXiv:1308.3444, https://arxiv.org/abs/1308.3444
2. David Hernandez, Michio Jimbo, Asymptotic representations and Drinfeld rational fractions, Compositio Math. 148 (2012) no. 5, 1593-1623, DOI 10.1112/S0010437X12000267, arXiv:1104.1891, https://arxiv.org/abs/1104.1891
3. Edward Frenkel, Nicolai Reshetikhin, The q-characters of representations of quantum affine algebras and deformations of W-algebras, Contemp. Math. 248 (1999) 163-205, DOI 10.1090/conm/248/03823, arXiv:math/9810055, https://arxiv.org/abs/math/9810055
4. Boris Feigin, Michio Jimbo, Tetsuji Miwa, Eugene Mukhin, Finite type modules and Bethe ansatz equations, Ann. Henri Poincaré 18 (2017) no. 8, 2543-2579, DOI 10.1007/s00023-017-0577-y, arXiv:1609.05724, https://arxiv.org/abs/1609.05724
5. V. Tarasov, Completeness of the Bethe Ansatz for the Periodic Isotropic Heisenberg Model, Rev. Math. Phys. 30 (2018) no. 8, 1840018, DOI 10.1142/S0129055X18400184, https://doi.org/10.1142/S0129055X18400184
6. R. J. Baxter, Completeness of the Bethe ansatz for the six and eight vertex models, J. Statist. Phys. 108 (2002) no. 1-2, 1-48, DOI 10.1023/A:1015437118218, arXiv:cond-mat/0111188, https://arxiv.org/abs/cond-mat/0111188
7. Edward Frenkel, Peter Koroteev, Daniel S. Sage, Anton M. Zeitlin, q-Opers, QQ-Systems, and Bethe Ansatz, J. Eur. Math. Soc. 26 (2024) no. 1, 355-405, DOI 10.4171/JEMS/1268, arXiv:2002.07344, https://arxiv.org/abs/2002.07344
8. Edward Frenkel, David Hernandez, Extended Baxter relations and QQ-systems for quantum affine algebras, Comm. Math. Phys. 405 (2024) no. 8, Paper No. 190, DOI 10.1007/s00220-024-05051-1, arXiv:2312.13256, https://arxiv.org/abs/2312.13256
9. Edward Frenkel, David Hernandez, Extremal monomial property of q-characters and polynomiality of the X-series, arXiv:2504.00260 (2025), https://arxiv.org/abs/2504.00260
10. Michela Varagnolo, Eric Vasserot, Quantum loop groups and critical convolution algebras, arXiv:2302.01418 (2023), https://arxiv.org/abs/2302.01418
11. Michela Varagnolo, Eric Vasserot, Non symmetric quantum loop groups and K-theory, arXiv:2308.01809 (2023, revised 2025), https://arxiv.org/abs/2308.01809
12. Petr P. Pushkar, Andrey V. Smirnov, Anton M. Zeitlin, Baxter Q-operator from quantum K-theory, Adv. Math. 360 (2020) 106919, DOI 10.1016/j.aim.2019.106919, arXiv:1612.08723, https://arxiv.org/abs/1612.08723
13. Anton M. Zeitlin, Geometric realizations of the Bethe ansatz equations, arXiv:2410.19997 (2024), https://arxiv.org/abs/2410.19997
14. David Hernandez, Bernard Leclerc, Cluster algebras and category O for representations of Borel subalgebras of quantum affine algebras, Algebra Number Theory 10 (2016) no. 9, 2015-2052, DOI 10.2140/ant.2016.10.2015, arXiv:1603.05014, https://arxiv.org/abs/1603.05014
15. David Hernandez, Representations and characters of quantum affine algebras at the crossroads between cluster categorification and quantum integrable models, arXiv:2510.06437 (2025), https://arxiv.org/abs/2510.06437
16. S. Gautam, V. Toledano-Laredo, An abelian formula for the quantum Weyl group action of the coroot lattice, arXiv:2501.02365 (2025, revised 2026), https://arxiv.org/abs/2501.02365