The 'Class 5' and 'Class 6' models found in the classification of 4×44\times4 solutions of the Yang-Baxter equation are non-diagonalisable Jordanian Drinfeld twists of the Heisenberg XXX spin chain, and their continuum limits, constructed by de Leeuw, Fontanella and Nieto Garcia (SciPost Phys. 20 (2026) 041), are non-unitary deformations of the Landau-Lifshitz action. Classically these deformed models are integrable: an infinite tower of commuting charges is generated by a boost functional acting as a hereditary (strong) symmetry in the sense of Fuchssteiner's theorem. Yet their perturbative quantization exhibits a non-vanishing tree-level 1→21\to2 amplitude in which the produced particle is soft - it carries zero momentum and, by the dispersion ω=−p2\omega=-p^2, zero energy - so the models fail the usual 'no particle production' criterion of quantum integrability, whose standard proof assumes Lorentz invariance and massive excitations. The open problem is to settle the full integrable structure of these models: construct an explicit Lax pair for the deformed equations of motion; compute the multi-particle S-matrix beyond the tree-level 2→22\to2 sector and decide whether it factorizes once the soft mode is treated consistently; establish whether the classical boost/hereditary recursion has a quantum counterpart at the level of the twisted R-matrix; and determine whether the soft particle production is physically meaningful or removable by Faddeev-Kulish dressing.

Background

The Heisenberg XXX spin chain is the prototypical integrable quantum system: a one-dimensional lattice of spin-12\frac12 sites whose dynamics commute with a transfer matrix built from an RR-matrix, a solution of the Yang-Baxter equation - the consistency condition underpinning factorized scattering and the algebraic structures that generate an infinite tower of mutually commuting conserved charges. In a classification of all regular 4×44\times4 RR-matrices of difference form (those depending only on the difference of spectral parameters) acting on two-dimensional local Hilbert spaces, de Leeuw, Pribytok and Ryan (2019) found fourteen solutions, six of them new; the 'Class 5' and 'Class 6' models among the new ones have the unusual feature that the transfer matrix, and hence the Hamiltonian, is non-diagonalisable: it possesses Jordan blocks instead of a complete eigenbasis, so the Bethe ansatz (the standard exact diagonalisation method for such chains) does not retrieve the full Hilbert space. de Leeuw, Fontanella and Nieto Garcia (2026) showed that the Class 5 RR-matrix is a Jordanian Drinfeld twist of the XXX RR-matrix - a similarity transformation by a triangular matrix built from a spin-raising operator - and constructed the continuum (long-wavelength) limits of both models.

The continuum limit of the XXX chain is the Landau-Lifshitz (LL) model: a classical field theory for a unit spin field S⃗(x,t)∈S2\vec S(x,t)\in S^2, parametrized by angles θ(x,t)\theta(x,t) and ϕ(x,t)\phi(x,t), with a first-order-in-time kinetic term and gradient energy. Historically the same action appears both for fast-spinning strings in AdS5×S5\mathrm{AdS}_5\times S^5 and for large-spin anomalous dimensions in N=4\mathcal N=4 super Yang-Mills, making it a central meeting point of AdS/CFT integrability. The Class 5 twist survives the continuum limit as a non-unitary (complex) deformation with parameter α\alpha:

S5lead=−ε2∫dt dx[cos⁡θ ϕt+12((θx)2+sin⁡2θ (ϕx)2)+α e−iϕ2(θx−i2sin⁡2θ ϕx)].S^{\rm lead}_5=-\varepsilon^2\int dt\,dx\left[\cos\theta\,\phi_t+\frac12\left((\theta_x)^2+\sin^2\theta\,(\phi_x)^2\right)+\frac{\alpha\,e^{-i\phi}}{2}\left(\theta_x-\frac{i}{2}\sin 2\theta\,\phi_x\right)\right].

The first two terms are the standard LL action; the α\alpha-term is complex, so the theory is non-Hermitian, encoding the decay η→η×η\eta\to\eta\times\eta of its single complex excitation but not the inverse process. The Class 6 model similarly yields an LL theory in a complex potential α2e−iϕsin⁡2θ\frac{\alpha}{2}e^{-i\phi}\sin 2\theta:

S6lead=−ε2∫dt dx[cos⁡θ ϕt+12((θx)2+sin⁡2θ (ϕx)2)+α e−iϕ2sin⁡2θ].S^{\rm lead}_6=-\varepsilon^2\int dt\,dx\left[\cos\theta\,\phi_t+\frac12\left((\theta_x)^2+\sin^2\theta\,(\phi_x)^2\right)+\frac{\alpha\,e^{-i\phi}}{2}\sin 2\theta\right].

Classical integrability survives the deformation. de Leeuw et al. constructed the infinite tower of commuting charges and showed it is generated by a boost functional B[H]B[H] through Qr+1={B[H],Qr}Q_{r+1}=\{B[H],Q_r\}, with involutivity {Qr,Qs}=0\{Q_r,Q_s\}=0 following from Fuchssteiner's hereditary-recursion argument for the LL hierarchy: the boost acts as a strong (hereditary) symmetry of the equations of motion - a transformation mapping every solution to another solution whose repeated action generates the whole hierarchy - so the existence of Q3Q_3 guarantees the entire tower. Notably, this bypasses the usual route to such towers - a Lax pair, i.e. matrices LL and MM whose zero-curvature condition ∂tL−∂xM+[L,M]=0\partial_t L-\partial_x M+[L,M]=0 is equivalent to the equations of motion, a representation long known for the undeformed LL model - and no Lax pair for the deformed equations is known.

For quantum theories in 1+11+1 dimensions the standard integrability criterion is different: elastic, factorized scattering, meaning vanishing n→mn\to m amplitudes for n≠mn\ne m and factorization of n→nn\to n amplitudes into 2→22\to2 blocks. Parke's theorem establishing absence of particle production assumes Lorentz invariance and massive excitations, since its proof separates scattering centres into spacelike-separated wave packets; both assumptions fail here - the LL model is non-Lorentzian, with first-order dynamics and dispersion ω=−p2\omega=-p^2, and its excitations are massless. Indeed, classically integrable two-dimensional sigma models are known to exhibit particle production and non-factorization in their massless tree-level S-matrices (Hoare, Levine and Tseytlin; Georgiou), and cubic vertices producing particles were found by Borsato and Driezen in a light-cone-gauge-fixed Jordanian deformation of AdS5×S5\mathrm{AdS}_5\times S^5 that possesses a classical Lax connection (the chiral halves of a Lax pair). The deformed LL model sharpens this tension into a potential counterexample with a fully controlled lattice antecedent: quantizing the Class 5 action expanded about the vacuum in the complex fluctuation field η\eta (whose quadratic part matches that of the undeformed LL model), the deformation introduces cubic vertices and the tree-level 1→21\to2 amplitude is

S(k;p,p′)=iαk4∣k∣[δ(p) δ(k−p′)+δ(p′) δ(k−p)],\mathcal S(k;p,p')=\frac{i\alpha k}{4|k|}\left[\delta(p)\,\delta(k-p')+\delta(p')\,\delta(k-p)\right],

which is non-zero precisely when one of the produced particles is soft (zero momentum, and zero energy by the dispersion). The 2→22\to2 tree-level amplitude is undeformed. Because the produced particle is soft, the situation parallels soft photons in QED: Faddeev and Kulish proposed dressing asymptotic states with clouds of soft particles so that the scattering operator becomes well-defined; in gravity, Weinberg's soft theorem confines soft gravitons to a phase factor, which was later understood to be physically meaningful through its equivalence to supertranslations of the BMS (Bondi-Metzner-Sachs) group, the asymptotic symmetry group of gravity in asymptotically flat spacetime (He, Lysov, Mitra and Strominger). de Leeuw et al. observed that the dressing which removes the late-time cubic term is the continuum limit of the Drinfeld twist itself, suggesting - but not proving - that the soft production is an artifact of naive Fock states and that the dressed S-matrix reduces to the undeformed LL one, whose three-particle factorization was verified perturbatively by Melikyan, Pinzul, Rivelles and Weber. Whether this inference is correct, whether factorization survives at 3→33\to3 and beyond tree level, whether a quantum boost operator exists, and what the soft mode ultimately means are all open; for the Class 6 model even the vacuum is unidentified, since constant fields are not solutions and the naive field expansion contains a tadpole (a linear term showing one has expanded about a non-solution).

Problem Statement

Consider the deformed Landau-Lifshitz theories obtained as the continuum limits of the Class 5 and Class 6 non-diagonalisable (Jordanian Drinfeld-twisted) deformations of the Heisenberg XXX spin chain, with actions given by de Leeuw, Fontanella and Nieto Garcia (SciPost Phys. 20 (2026) 041), equations (2.13) for Class 5 (quoted in the background) and (A.9) for Class 6, with deformation parameter α\alpha. The problem is to settle the integrable structure of these models and the physical status of their soft particle production, through four jointly linked components:

  1. Lax pair. Construct an explicit Lax pair - matrices L(t,x)L(t,x) and M(t,x)M(t,x) built from the fields and finitely many spatial derivatives, possibly with a spectral parameter - whose zero-curvature condition ∂tL−∂xM+[L,M]=0\partial_t L-\partial_x M+[L,M]=0 is equivalent to the deformed equations of motion, expected to be a deformation of the known Lax pair of the undeformed Landau-Lifshitz model; or prove that no such pair exists within the specified class, which would demonstrate boost integrability without Lax integrability.

  2. Multi-particle S-matrix. Compute the S-matrix beyond the tree-level 2→22\to2 sector - at minimum the tree-level 3→33\to3 amplitude, and the 2→22\to2 amplitude at one loop - with a consistent treatment of the soft mode (for example via Faddeev-Kulish-dressed asymptotic states), and decide whether it factorizes into products of 2→22\to2 amplitudes as the undeformed LL S-matrix does, or exhibits genuine non-factorization. This component is completed by the factorization verdict that the stipulated minimum computations - the tree-level 3→33\to3 amplitude and the one-loop 2→22\to2 amplitude, with a consistent treatment of the soft mode - establish, together with an explicit statement of whether that verdict is expected to persist at higher nn or loop order; component 4 below is a separate sub-claim about the standing of the soft mode and does not by itself discharge this component.

  3. Quantum boost. Establish or rule out a quantum counterpart of the classical boost/hereditary recursion: an operator statement at the level of the twisted R-matrix or transfer matrix (for instance a quantum boost operator) that generates the full tower of mutually commuting quantum conserved charges, mirroring the classical relation Qr+1={B[H],Qr}Q_{r+1}=\{B[H],Q_r\}.

  4. Status of the soft mode. Determine whether the tree-level production of the zero-momentum particle is removable by dressing - i.e. whether the dressed S-matrix coincides with the undeformed LL S-matrix, possibly up to a computable phase - or whether it carries observable or symmetry significance, in the sense of an analogue of Weinberg's soft theorem and its BMS-supertranslation interpretation.

For the Class 6 model this includes the prerequisite task of identifying the correct classical vacuum (which is expected to be spacetime-dependent, since expansion about constant fields produces a tadpole), without which no perturbative S-matrix can be defined.

These components jointly discharge one scientific objective: deciding the question mark in the source title - whether these models are genuinely integrable at the quantum level, and in what sense their particle production is real. The classical Lax pair is the standard zero-curvature backbone; the multi-particle S-matrix is the quantum criterion in the regime where its usual proof fails; the quantum boost is the structural mechanism that would guarantee quantum integrability; and the soft-mode status fixes the interpretation of the production that motivated the question. Partial results - for instance the Lax pair alone, or the 3→33\to3 amplitude alone - constitute progress but do not complete the problem.

References

  1. M. de Leeuw, A. Fontanella and J. M. Nieto Garcia, An integrable deformed Landau-Lifshitz model with particle production?, SciPost Phys. 20 (2026) 041, DOI 10.21468/SciPostPhys.20.2.041, arXiv:2506.13598, https://arxiv.org/abs/2506.13598
  2. M. de Leeuw, A. Pribytok and P. Ryan, Classifying integrable spin-1/2 chains with nearest neighbour interactions, J. Phys. A 52 (2019) 505201, DOI 10.1088/1751-8121/ab529f, arXiv:1904.12005, https://arxiv.org/abs/1904.12005
  3. S. Parke, Absence of particle production and factorization of the S-matrix in 1+1 dimensional models, Nucl. Phys. B 174 (1980) 166, DOI 10.1016/0550-3213(80)90196-0, https://doi.org/10.1016/0550-3213(80)90196-0
  4. B. Fuchssteiner, On the hierarchy of the Landau-Lifshitz equation, Physica D 13 (1984) 387, DOI 10.1016/0167-2789(84)90139-8, https://doi.org/10.1016/0167-2789(84)90139-8
  5. A. Melikyan, A. Pinzul, V. O. Rivelles and G. Weber, On S-Matrix factorization of the Landau-Lifshitz model, JHEP 10 (2008) 002, DOI 10.1088/1126-6708/2008/10/002, arXiv:0808.2489, https://arxiv.org/abs/0808.2489
  6. P. Kulish and L. Faddeev, Asymptotic conditions and infrared divergences in quantum electrodynamics, Theor. Math. Phys. 4 (1970) 745, DOI 10.1007/BF01066485, https://doi.org/10.1007/BF01066485
  7. S. Weinberg, Infrared Photons and Gravitons, Phys. Rev. 140 (1965) B516, DOI 10.1103/PhysRev.140.B516, https://doi.org/10.1103/PhysRev.140.B516
  8. T. He, V. Lysov, P. Mitra and A. Strominger, BMS supertranslations and Weinberg's soft graviton theorem, JHEP 05 (2015) 151, DOI 10.1007/JHEP05(2015)151, arXiv:1401.7026, https://arxiv.org/abs/1401.7026
  9. B. Hoare, N. Levine and A. A. Tseytlin, On the massless tree-level S-matrix in 2d sigma models, J. Phys. A 52 (2019) 144005, DOI 10.1088/1751-8121/ab0b79, arXiv:1812.02549, https://arxiv.org/abs/1812.02549
  10. G. Georgiou, The massless S-matrix of integrable sigma-models, JHEP 01 (2025) 189, DOI 10.1007/JHEP01(2025)189, arXiv:2408.03673, https://arxiv.org/abs/2408.03673
  11. R. Borsato and S. Driezen, Particle production in a light-cone gauge fixed Jordanian deformation of AdS5 x S5, Phys. Rev. D 111 (2025) 086010, DOI 10.1103/PhysRevD.111.086010, arXiv:2412.08411, https://arxiv.org/abs/2412.08411
  12. S. Driezen and A. Molines, Jordanian spin chains for twisted strings in AdS5 x S5, Phys. Rev. D 112 (2025) 106001, DOI 10.1103/b1cg-6s5n, arXiv:2507.13911, https://arxiv.org/abs/2507.13911
  13. R. Borsato and T. Meier, Non-commutative deformations of gauge theories via Drinfel'd twists of the scale symmetry, arXiv:2512.04162, https://arxiv.org/abs/2512.04162
  14. S. Driezen, F. Levkovich-Maslyuk and A. Molines, Integrability for the spectrum of Jordanian AdS/CFT, arXiv:2511.11521, https://arxiv.org/abs/2511.11521
  15. F. He, A. Hutsalyuk, G. Mussardo and A. Stampiggi, Statistical signatures of integrable and non-integrable quantum hamiltonians, J. Stat. Mech. (2026) 023101, DOI 10.1088/1742-5468/ae337c, arXiv:2510.02440, https://arxiv.org/abs/2510.02440
  16. C. Paletta and T. Prosen, On the integrability structure of the deformed rule-54 reversible cellular automaton, J. Phys. A 59 (2026) 345201, DOI 10.1088/1751-8121/ae925e, arXiv:2603.25424, https://arxiv.org/abs/2603.25424
  17. A. Fontanella and J. M. Nieto Garcia, An introduction to string Newton-Cartan holography and integrability, J. Phys. A 59 (2026) 333001, DOI 10.1088/1751-8121/ae91a3, arXiv:2603.24657, https://arxiv.org/abs/2603.24657
  18. Y. F. Adans, M. de Leeuw and T. McLoughlin, On deforming and breaking integrability, arXiv:2603.17018, https://arxiv.org/abs/2603.17018
  19. K. Schouten and M. de Leeuw, The quantum group structure of long-range integrable deformations, arXiv:2604.27189, https://arxiv.org/abs/2604.27189