Exact solution theory of the non-diagonalisable spin chains from the non-difference-form regular 4×4 Yang–Baxter solutions

The completed classification of regular $4\times4$ solutions of the Yang–Baxter equation (Corcoran–de Leeuw, SciPost Phys. Core 7, 045 (2024)) exhibited four new integrable spin-chain models of non-difference form, all non-Hermitian and non-diagonalisable, one of them a non-diagonalisable integrable deformation of the XXX chain. The open problem left by that work is the exact solution theory of these models: identify the symmetry algebra underlying each model; compute the Jordan-block spectra of the transfer matrix and Hamiltonian at arbitrary chain length, where direct symbolic computation becomes very difficult for $L>4$; construct an algebraic Bethe ansatz, or an equivalent systematic method, that yields generalised eigenvectors and is complete for non-diagonalisable chains; and decide whether the deformed XXX model is the dilatation operator of a logarithmic conformal field theory. Later work settled parts of the difference-form specialisation of the deformed model — it is a Jordanian Drinfeld twist of XXX, with transfer-matrix Jordan form known at all lengths — but the generic non-difference models, a general non-diagonalisable Bethe ansatz, and the field-theory interpretation remain open.

Background

The Yang–Baxter equation (YBE) is the consistency condition

$$R_{12}(u,v)\,R_{13}(u,w)\,R_{23}(v,w)=R_{23}(v,w)\,R_{13}(u,w)\,R_{12}(u,v)$$

for a matrix $R(u,v)$ acting on the tensor product of two two-dimensional vector spaces, where $R_{ij}$ acts non-trivially on factors $i$ and $j$ of a larger tensor product. A solution $R$ (an R-matrix) generates, through the trace of the monodromy matrix — the ordered product of R-matrices along the chain, whose trace is the transfer matrix $T(u)$ — a commuting family of operators on a spin chain, which is the operational meaning of quantum integrability used here. A solution is called regular if $R(u,u)$ is proportional to the permutation operator; regularity guarantees a local nearest-neighbour Hamiltonian $H\propto\dot R(u,u)\big|_{u=0}$, and the solution is of difference form if $R$ depends on the spectral parameters only through $u-v$ (the XXX and XXZ chains are of difference form).

Since Sutherland's 1970 work, an R-matrix can be reconstructed from the Hamiltonian through what are now called the Sutherland equations, an infinite tower of conditions whose lowest order is the Reshetikhin condition $[Q_2,[B[Q_2],Q_2]]=0$, where $Q_2$ is the total Hamiltonian (the second conserved charge) and $B[Q_2]$ is the boost operator that translates charges along the chain. This turns classification into a computation: parametrise a general Hamiltonian density, impose the Reshetikhin condition, solve the resulting equations, and reconstruct $R$. de Leeuw–Pribytok–Ryan (2019) classified integrable nearest-neighbour spin-$1/2$ Hamiltonians this way, and de Leeuw–Paletta–Pribytok–Retore–Ryan (2021) classified the Hermitian difference-form solutions of six- and eight-vertex type. Corcoran–de Leeuw (2024) completed the classification of all regular $4\times4$ solutions (local Hilbert-space dimension two). Besides previously known models, they found four new solutions of non-difference form, depending on free functions $F(u)$, $G(u)$ and constants: two nilpotent models (Models 1 and 2), a third model (Model 3), and a non-diagonalisable integrable deformation of the XXX chain (Model 4), which reduces to XXX when the deformation is switched off. The difference-form specialisation of Model 4 had in fact already appeared as the 'Class 5' model of the 2019 classification; the new content of Model 4 is the general non-difference-form family with arbitrary free functions.

All four new models are non-Hermitian and, at generic values of the free functions and constants, non-diagonalisable: their transfer matrices and Hamiltonians are then not similar to diagonal matrices, so their canonical form is a Jordan normal form — a direct sum of Jordan blocks $J^\ell_\lambda$, meaning $\ell\times\ell$ upper-triangular blocks with eigenvalue $\lambda$ on the diagonal and ones on the superdiagonal — and the Hilbert space is organised by generalised eigenvectors forming Jordan chains. Non-diagonalisability defeats the standard solution machinery: the algebraic Bethe ansatz (ABA) constructs eigenvectors of the transfer matrix from a pseudo-vacuum (a simple reference state) using the commutation relations of the monodromy matrix, and its completeness argument relies on diagonalisability; when $T(u)$ is defective, its eigenvectors do not span the Hilbert space. This machinery is normally organised by a symmetry algebra: the Yangian is a Hopf-algebraic deformation of the universal enveloping algebra of the chain's $\mathfrak{sl}_2$ spin symmetry (a Hopf algebra being an algebra equipped with a coproduct, the rule for how its generators act on tensor products), the quantum group $U_q(\mathfrak{sl}_2)$ is its trigonometric $q$-deformation, and both admit deformations by a Drinfeld twist, which deforms the coproduct while preserving the generators, as well as dynamical generalisations whose structure constants depend on extra auxiliary dynamical variables. The only ABA-type construction that handles a defective transfer matrix is that of Gainutdinov–Nepomechie for the quantum-group-invariant open XXZ chain (free-end boundary conditions) at roots of unity; for the 'eclectic' spin chain — a three-state chain arising from the $\gamma$-deformed ('fishnet') version of $\mathcal{N}=4$ super Yang–Mills theory — the Jordan spectrum was instead obtained combinatorially, with multiplicities given by $q$-binomial coefficients (the $q$-analogues of the binomial coefficients) and Pólya counting (enumeration of configurations up to the symmetries of the chain), and by degenerate limits of diagonalisable models. Corcoran–de Leeuw computed the Jordan-block spectra of the new models at chain length $L=4$, with refinements at special parameter choices, and noted that the symbolic computation becomes very difficult for $L>4$, especially in the presence of free functions. As the natural continuation they posed: identifying a symmetry algebra for the new models (which may hint at physical applications), finding an integrability-based method for the Jordan-block spectra and a Bethe ansatz for non-diagonalisable models, and the question whether the deformed XXX model is the dilatation operator of a deformed conformal field theory. This last question is motivated by AdS/CFT (the correspondence relating string theory on anti-de Sitter spacetimes to conformal gauge theories): the XXX chain is the one-loop dilatation operator of the $\mathfrak{su}(2)$ sector of planar $\mathcal{N}=4$ super Yang–Mills — its spectrum is the set of anomalous dimensions — the interaction-induced shifts of the scaling dimensions of the gauge theory's single-trace operators — and a deformed theory whose dilatation operator is non-diagonalisable would be a logarithmic conformal field theory, i.e. a CFT whose dilatation operator has Jordan blocks and whose correlation functions contain logarithms, as happens for the $\gamma$-deformation of $\mathcal{N}=4$.

Problem Statement

Let $R(u,v)$ be one of the four regular, non-difference-form $4\times4$ R-matrices found by Corcoran–de Leeuw (Models 1–4 of arXiv:2306.10423, Section 4.1, depending on free functions $F(u)$, $G(u)$ — and, for Model 2, a third free function $H(u)$ — and constants), together with the associated periodic spin chains: the Hamiltonian $H$, the transfer matrix $T(u)$, and the higher commuting charges $Q_r(u)$, at chain length $L$ and generic choices of the free functions and constants. The problem is to construct the exact solution theory of these models, comprising four jointly linked results.

(1) Algebraic structure. For each of the four models, identify the symmetry algebra — a Yangian- or quantum-group-type, twisted, dynamical, or genuinely new structure — that underlies the model and organises its commuting charges and spectrum, or prove that no such structure of the recognised types exists.

(2) Jordan-block spectra. Determine, for arbitrary chain length $L$ (the source's symbolic computations become very difficult for $L>4$), the Jordan normal forms of $T(u)$ and of $H$: eigenvalues, block sizes, and their multiplicities, including how these refine under the specialisations of the free functions and constants, and whether the multiplicities admit closed $q$-deformed or combinatorial formulae as they do for the eclectic chain.

(3) Non-diagonalisable Bethe ansatz. Construct a version of the algebraic (or coordinate) Bethe ansatz applicable to non-diagonalisable integrable chains that produces the generalised eigenvectors together with the eigenvalues and is provably complete on the Hilbert space, at least for these models.

(4) Field-theory interpretation. Decide whether the non-diagonalisable integrable deformation of the XXX chain (Model 4 at general non-difference form) is the dilatation operator of a deformed conformal field theory — in particular a logarithmic CFT — and identify the would-be theory whose spectrum and Jordan structure it reproduces.

These four results jointly constitute the solution theory of the new models: the algebra of (1) is expected to organise the spectrum determined in (2), the construction of (3) is the method expected to deliver it, and (4) is the physical interpretation of the solution of Model 4. For the difference-form specialisation of Model 4 — the Class 5 model of de Leeuw–Pribytok–Ryan (2019) — items (1) and (2) have been settled (a Jordanian Drinfeld twist of XXX; transfer-matrix Jordan form at all $L$); the problem asks for the corresponding results at the full non-difference generality and for Models 1–3, together with the generally applicable construction (3), while (4) is open even for the specialisation.

References

1. Luke Corcoran and Marius de Leeuw, 'All regular 4×4 solutions of the Yang–Baxter equation', SciPost Phys. Core 7, 045 (2024), doi:10.21468/SciPostPhysCore.7.3.045, arXiv:2306.10423, https://arxiv.org/abs/2306.10423
2. Marius de Leeuw, Anton Pribytok and Paul Ryan, 'Classifying integrable spin-1/2 chains with nearest neighbour interactions', J. Phys. A 52, 505201 (2019), doi:10.1088/1751-8121/ab529f, arXiv:1904.12005, https://arxiv.org/abs/1904.12005
3. Marius de Leeuw, Chiara Paletta, Anton Pribytok, Ana L. Retore and Paul Ryan, 'Yang-Baxter and the Boost: splitting the difference', SciPost Phys. 11, 069 (2021), doi:10.21468/SciPostPhys.11.3.069, arXiv:2010.11231, https://arxiv.org/abs/2010.11231
4. Marius de Leeuw and Vera Posch, 'All 4×4 solutions of the quantum Yang-Baxter equation', arXiv:2411.18685 (2024), https://arxiv.org/abs/2411.18685
5. Bill Sutherland, 'Two-Dimensional Hydrogen Bonded Crystals without the Ice Rule', J. Math. Phys. 11, 3183 (1970), doi:10.1063/1.1665111, https://doi.org/10.1063/1.1665111
6. Joseph A. Minahan, 'Review of AdS/CFT Integrability, Chapter I.1: Spin Chains in N=4 Super Yang-Mills', Lett. Math. Phys. 99, 33 (2012), doi:10.1007/s11005-011-0522-9, arXiv:1012.3983, https://arxiv.org/abs/1012.3983
7. Changrim Ahn and Matthias Staudacher, 'The Integrable (Hyper)eclectic Spin Chain', JHEP 02, 019 (2021), doi:10.1007/JHEP02(2021)019, arXiv:2010.14515, https://arxiv.org/abs/2010.14515
8. Changrim Ahn, Luke Corcoran and Matthias Staudacher, 'Combinatorial Solution of the Eclectic Spin Chain', JHEP 03, 028 (2022), doi:10.1007/JHEP03(2022)028, arXiv:2112.04506, https://arxiv.org/abs/2112.04506
9. Juan Miguel Nieto García and Leander Wyss, 'Jordan blocks and the Bethe ansatz I: The eclectic spin chain as a limit', Nucl. Phys. B 981, 115860 (2022), doi:10.1016/j.nuclphysb.2022.115860, arXiv:2112.13883, https://arxiv.org/abs/2112.13883
10. Juan Miguel Nieto García, 'Jordan blocks and the Bethe ansatz II: The eclectic spin chain beyond K=1', JHEP 12, 106 (2022), doi:10.1007/JHEP12(2022)106, arXiv:2206.08348, https://arxiv.org/abs/2206.08348
11. Juan Miguel Nieto García, 'Jordan blocks and the Bethe ansatz III: Class 5 model and its symmetries', Nucl. Phys. B 998, 116419 (2024), doi:10.1016/j.nuclphysb.2023.116419, arXiv:2309.10044, https://arxiv.org/abs/2309.10044
12. Azat M. Gainutdinov and Rafael I. Nepomechie, 'Algebraic Bethe ansatz for the quantum group invariant open XXZ chain at roots of unity', Nucl. Phys. B 909, 796 (2016), doi:10.1016/j.nuclphysb.2016.06.007, arXiv:1603.09249, https://arxiv.org/abs/1603.09249
13. Marius de Leeuw, Andrea Fontanella and Juan Miguel Nieto García, 'An integrable deformed Landau-Lifshitz model with particle production?', SciPost Phys. 20, 041 (2026), doi:10.21468/SciPostPhys.20.2.041, arXiv:2506.13598, https://arxiv.org/abs/2506.13598
14. Riccardo Borsato and Miguel García Fernández, 'Jordanian deformation of the non-compact and $\mathfrak{sl}_2$-invariant XXX$_{-1/2}$ spin-chain', JHEP 08, 074 (2025), doi:10.1007/JHEP08(2025)074, arXiv:2503.24223, https://arxiv.org/abs/2503.24223
15. Riccardo Borsato and Miguel García Fernández, 'Groenewold-Moyal twists, integrable spin-chains and AdS/CFT', arXiv:2604.07291 (2026), https://arxiv.org/abs/2604.07291
16. Ilya Vilkoviskiy and Kirill Matirko, 'Properties of the temporal transfer matrix in integrable Floquet circuits', arXiv:2508.13883 (2025), https://arxiv.org/abs/2508.13883
17. Shailesh Lal, Suvajit Majumder and Evgeny Sobko, 'Deep Learning based discovery of Integrable Systems', arXiv:2503.10469 (2025), https://arxiv.org/abs/2503.10469
18. Luke Corcoran, Marius de Leeuw and Balázs Pozsgay, 'Integrable models on Rydberg atom chains', SciPost Phys. 18, 139 (2025), doi:10.21468/SciPostPhys.18.4.139, arXiv:2405.15848, https://arxiv.org/abs/2405.15848
19. Matthew Blakeney, Luke Corcoran and Marius de Leeuw, 'Constrained integrability and anyonic chains', arXiv:2605.28946 (2026), https://arxiv.org/abs/2605.28946