Closure of the quantum separation-of-variables program for twisted $gl_{\mathcal{M}|\mathcal{N}}$ spin chains and the inhomogeneous Hubbard model

Maillet, Niccoli and Vignoli (SciPost Phys. 9, 060 (2020)) constructed quantum separation-of-variables (SoV) bases for the fundamental inhomogeneous $gl_{\mathcal{M}|\mathcal{N}}$ supersymmetric spin chains and for the inhomogeneous Hubbard model with quasi-periodic twisted boundary conditions defined by twist matrices with simple spectrum, but proved a complete SoV characterization of the transfer-matrix spectrum only for the $gl_{1|2}$ chain with a special non-invertible twist, and for two-site lattices for any simple-spectrum twist. They conjectured (Conjecture 3.1) that for general twists the spectrum is exactly characterized by the inner-boundary quantum-Berezinian relation together with the fusion/interpolation formulae and the null out-boundary conditions in the separate variables $x_a=t_1(\xi_a)$ , verified this up to three sites, and deferred the general proof, the complete spectral solution for both model classes, and the derivation of the Hubbard fusion/closure functional relations to future publications that never appeared. The open problem is to prove this closure characterization for arbitrary $\mathcal{M},\mathcal{N}$ and arbitrary simple-spectrum twists (including non-invertible ones), to derive the corresponding complete set of functional relations for the transfer matrix of the inhomogeneous Hubbard model built from the Shastry $R$ -matrix, and thereby to obtain the complete SoV spectral solution — eigenvalues and factorized eigenvectors, with Bethe-ansatz completeness as a corollary — for both model classes under general twisted boundary conditions.

Background

Quantum integrable lattice models are solved by the quantum inverse scattering method: a local $R$ -matrix obeying the Yang–Baxter equation (the consistency condition for factorized scattering on the lattice) is used to build a monodromy matrix, and its trace — the transfer matrix $T_1(\lambda)$ — generates a one-parameter commuting family of conserved quantities. Quasi-periodic (twisted) boundary conditions are imposed by a twist matrix $K$ inserted on the auxiliary space; throughout, $K$ is an integrable twist matrix with simple spectrum, i.e. distinct eigenvalues — it may be non-diagonal in the natural basis and may be non-invertible (a zero eigenvalue), and both generalizations go beyond what standard Bethe-ansatz completeness theorems cover. The full spectral problem — simultaneous eigenvalues and eigenvectors of $T_1(\lambda)$ — encodes all conserved charges of the model.

Two model classes are involved. First, the fundamental inhomogeneous $gl_{\mathcal{M}|\mathcal{N}}$ chains: $gl_{\mathcal{M}|\mathcal{N}}$ is the general linear Lie superalgebra on the graded vector space $\mathbb{C}^{\mathcal{M}|\mathcal{N}}$ , its rational $R$ -matrix defines super-Yangian-invariant lattice models whose $N$ local quantum spaces carry the fundamental (defining) representation, and the inhomogeneities $\xi_a$ shift the spectral parameter from site to site. Second, the one-dimensional Hubbard model — the paradigmatic model of correlated electrons, exactly solved at the level of the Hamiltonian spectrum by Lieb and Wu's coordinate Bethe ansatz — is integrable in the same framework through Shastry's $R$ -matrix, whose dependence on the spectral parameter and on the coupling $U$ (through $\eta=-2iU$ ) is intricate and not of difference type.

Separation of variables (SoV), introduced by Sklyanin, solves such spectral problems by reducing an $N$ -body question to $N$ one-variable questions. Maillet and Niccoli realized a basis-free version: the SoV covector basis is generated by the multiple actions of the transfer matrix on a generic co-vector $\langle S|$ , the separate variables being the values $x_a=t_1(\xi_a)$ , and in this basis any eigenvector's wavefunction factorizes into products of the eigenvalue evaluated at the separated points, so that the eigenvalue problem alone yields the full spectral solution. The method is not an Ansatz: completeness of the spectrum description is built in once the basis and the closure relations — the identities reducing arbitrary powers of the transfer matrices acting on the basis back to the finite basis — are established. For higher-rank models the natural closure identities come from the fusion hierarchy of transfer matrices acting in symmetric and antisymmetric (for superalgebras, 'fat-hook') auxiliary representations, related by determinant-type functional equations of Bazhanov–Reshetikhin type, together with an inner-boundary condition supplied by the quantum Berezinian (the super-analogue of the determinant, generating the conserved charges of the super Yangian) and by null out-boundary conditions that truncate the hierarchy.

What the source paper proved and what it left open. Maillet, Niccoli and Vignoli proved that SoV bases of this type exist for both model classes under general simple-spectrum twists — implying the transfer matrices are diagonalizable with simple spectrum — and solved the $gl_{1|2}$ spectral problem completely for a special non-invertible twist class ( $k_1=0$ with a generic invertible $2\times2$ block), including a quantum-spectral-curve reformulation and completeness of the nested algebraic Bethe ansatz (the nested Bethe ansatz is the standard construction of eigenvectors as sums over partitions of auxiliary Bethe roots; its completeness — producing all eigenvectors — is a separate theorem). For general twists they stated Conjecture 3.1: the spectrum of the $gl_{1|2}$ transfer matrix is exactly the set of admissible solutions of the closure system — the interpolation formulae expressing the fused-transfer-matrix eigenvalues in terms of the separate variables $x_a$ , the inner-boundary quantum-Berezinian relation, and the null out-boundary conditions $t^{(M+n)}_{\mathcal{N}+m}=0$ — excluding the trivial solution $x_a=0$ . They proved it for two-site lattices (any simple-spectrum twist), verified it numerically for three sites, derived the form of the $gl_{\mathcal{M}|\mathcal{N}}$ closure relation in an appendix, and deferred the general proofs and the Hubbard functional-relations analysis to future publications.

Problem Statement

Prove the closure-conjecture characterization of the transfer-matrix spectrum in the separation-of-variables framework, and thereby obtain the complete spectral solution, for the two model classes of the source paper:

(1) For the fundamental inhomogeneous $gl_{\mathcal{M}|\mathcal{N}}$ chain of arbitrary lattice size $N$ and arbitrary $\mathcal{M},\mathcal{N}$ , with quasi-periodic boundary conditions defined by an arbitrary integrable twist matrix having simple spectrum (including non-diagonal and non-invertible twists): prove that the transfer-matrix spectrum is exactly the set of admissible solutions of the closure system consisting of the interpolation formulae for the fused transfer-matrix eigenvalues in the separate variables $x_a=t_1(\xi_a)$ , the inner-boundary quantum-Berezinian relation, and the null out-boundary conditions, with the trivial solution excluded — that is, Conjecture 3.1 of the source for $gl_{1|2}$ and its $gl_{\mathcal{M}|\mathcal{N}}$ extension — with the solution count matching the Hilbert-space dimension $(\mathcal{M}+\mathcal{N})^N$ .

(2) For the inhomogeneous Hubbard model with a general simple-spectrum twist: derive the complete set of fusion/truncation functional relations for the transfer matrix built from the Shastry $R$ -matrix, identify the corresponding inner-boundary and null conditions, and prove that they completely characterize its spectrum in the SoV basis.

(3) As direct corollaries of (1) and (2): the complete SoV spectral solution — eigenvalues together with the factorized representation of eigenvectors in the SoV bases constructed in the source — under general twisted boundary conditions, and a proof (or a precise delimitation of failure) of the completeness of the nested algebraic Bethe ansatz for these models for generic twists, including the non-invertible twist classes not covered by existing Wronskian-algebra treatments.

These deliverables share one scientific objective: a closed, non-Ansatz algebraic characterization of the full spectrum (and eigenvectors) of both model classes under general twisted boundary conditions, with the closure conjecture as the common technical core and the Hubbard analysis expected, per the source, to build directly on the $gl_{\mathcal{M}|\mathcal{N}}$ results. Deliverables (1) and (2) are independent claims, both required for a non-refutation resolution, while (3) is their corollary; a refutation of (1) or of (2) therefore removes only the parts of (3) that depend on the refuted claim. A resolution may equally take the form of a disproof: an explicit finite-lattice counterexample to the closure characterization for some admissible simple-spectrum twist refutes the characterization asserted in deliverable (1) — and the analogous counterexample in the Hubbard model refutes deliverable (2) — leaving the remaining claim to be established or refuted on its own. Such a counterexample is by itself a complete resolution of the claim it refutes; exhibiting a corrected conjecture or corrected functional relations is a natural follow-up but is not a required deliverable.

References

J. M. Maillet, G. Niccoli, L. Vignoli, Separation of variables bases for integrable $gl_{\mathcal{M}|\mathcal{N}}$ and Hubbard models, SciPost Phys. 9, 060 (2020), doi:10.21468/SciPostPhys.9.4.060, arXiv:1907.08124, https://arxiv.org/abs/1907.08124
J. M. Maillet, G. Niccoli, On quantum separation of variables, J. Math. Phys. 59, 091417 (2018), doi:10.1063/1.5050989, arXiv:1807.11572, https://arxiv.org/abs/1807.11572
J. M. Maillet, G. Niccoli, L. Vignoli, On Scalar Products in Higher Rank Quantum Separation of Variables, SciPost Phys. 9, 086 (2020), doi:10.21468/SciPostPhys.9.6.086, arXiv:2003.04281, https://arxiv.org/abs/2003.04281
L. Vignoli, Quantum Separation of Variables in Higher Rank and Supersymmetric Integrable Models, PhD thesis, ENS Lyon (2021), tel-03550854, https://tel.archives-ouvertes.fr/tel-03550854
D. Chernyak, S. Leurent, D. Volin, Completeness of Wronskian Bethe Equations for Rational $gl(m|n)$ Spin Chains, Commun. Math. Phys. 391, 969 (2022), doi:10.1007/s00220-021-04275-9, arXiv:2004.02865, https://arxiv.org/abs/2004.02865
P. Ryan, D. Volin, Separation of Variables for Rational $gl(n)$ Spin Chains in Any Compact Representation, via Fusion, Embedding Morphism and Bäcklund Flow, Commun. Math. Phys. 383, 311 (2021), doi:10.1007/s00220-021-03990-7, arXiv:2002.12341, https://arxiv.org/abs/2002.12341
N. Gromov, F. Levkovich-Maslyuk, P. Ryan, Determinant form of correlators in high rank integrable spin chains via separation of variables, JHEP 05 (2021) 169, doi:10.1007/JHEP05(2021)169, arXiv:2011.08229, https://arxiv.org/abs/2011.08229
N. Gromov, N. Primi, P. Ryan, Form-factors and complete basis of observables via separation of variables for higher rank spin chains, JHEP 11 (2022) 039, doi:10.1007/JHEP11(2022)039, arXiv:2202.01591, https://arxiv.org/abs/2202.01591
N. Gromov, F. Levkovich-Maslyuk, New Compact Construction of Eigenstates for Supersymmetric Spin Chains, JHEP 09 (2018) 085, doi:10.1007/JHEP09(2018)085, arXiv:1805.03927, https://arxiv.org/abs/1805.03927
K. Lu, E. Mukhin, Jacobi–Trudi Identity and Drinfeld Functor for Super Yangian, Int. Math. Res. Not. 2021, no. 21, 16751 (2021), doi:10.1093/imrn/rnab023, arXiv:2007.15573, https://arxiv.org/abs/2007.15573
K. Lu, Completeness of Bethe ansatz for Gaudin models associated with $gl(1|1)$ , Nucl. Phys. B 980, 115790 (2022), doi:10.1016/j.nuclphysb.2022.115790, arXiv:2202.08162, https://arxiv.org/abs/2202.08162
K. Lu, On Bethe eigenvectors and higher transfer matrices for supersymmetric spin chains, JHEP 04 (2023) 120, doi:10.1007/JHEP04(2023)120, arXiv:2209.14416, https://arxiv.org/abs/2209.14416
F. Levkovich-Maslyuk, Separation of variables for higher rank integrable models: review of recent progress, J. Phys. A 58, 243001 (2025), doi:10.1088/1751-8121/ade103, arXiv:2503.15398, https://arxiv.org/abs/2503.15398
K. Fukai, Proof of Completeness of the Local Conserved Quantities in the One-Dimensional Hubbard Model, J. Stat. Phys. 191, 70 (2024), doi:10.1007/s10955-024-03267-y, arXiv:2309.09354, https://arxiv.org/abs/2309.09354
Y. Yi, Y. Qiao, J. Cao, W.-L. Yang, A novel Bethe ansatz scheme for the one-dimensional Hubbard model, Nucl. Phys. B 977, 115732 (2022), doi:10.1016/j.nuclphysb.2022.115732, https://doi.org/10.1016/j.nuclphysb.2022.115732
S. Ekhammar, D. Volin, Monodromy bootstrap for $SU(2|2)$ quantum spectral curves: from Hubbard model to $AdS_3/CFT_2$ , JHEP 03 (2022) 192, doi:10.1007/JHEP03(2022)192, arXiv:2109.06164, https://arxiv.org/abs/2109.06164
C. Rylands, B. Bertini, P. Calabrese, Integrable quenches in the Hubbard model, J. Stat. Mech. 2210 (2022) 103103, doi:10.1088/1742-5468/ac98be, arXiv:2206.07985, https://arxiv.org/abs/2206.07985
E. H. Lieb, F. Y. Wu, Absence of Mott transition in an exact solution of the short-range, one-band model in one dimension, Phys. Rev. Lett. 20, 1445 (1968), doi:10.1103/PhysRevLett.20.1445, https://doi.org/10.1103/PhysRevLett.20.1445
B. Sriram Shastry, Exact Integrability of the One-Dimensional Hubbard Model, Phys. Rev. Lett. 56, 2453 (1986), doi:10.1103/PhysRevLett.56.2453, https://doi.org/10.1103/PhysRevLett.56.2453