Factorization of reduced density matrices and closed representations of the nearest-neighbour two-point function in integrable face models

For integrable vertex models such as the spin-1/2 Heisenberg (XXZ) chain, the reduced density matrix of any finite segment of the inhomogeneous model is known to factorize into products of a single nearest-neighbour two-point function times model-independent structure functions, and the two-point function admits closed representations through the fermionic basis and non-linear integral equations. Frahm and Westerfeld (2021) derived the analogous discrete functional equations (of reduced quantum Knizhnik-Zamolodchikov type) for integrable interaction-round-a-face (IRF) models and observed the factorization property for the r=4 and r=5 restricted solid-on-solid (RSOS) models in topological sectors of quantum dimension one, but left the theory incomplete: the factorization is conjectured, not proven, for general RSOS models and is only an open question for general IRF models; no closed model-independent representation (integral formulae or integral equations) of the two-point function exists at finite length or finite temperature; and the boundary contributions arising from the fusion-path structure of the RSOS Hilbert space are unclarified. Later work (2023-2025) extended the finite-size computations to four sites, produced conjectural thermodynamic-limit ground-state formulae for the two-point function, obtained non-linear integral equations for one special Heisenberg-related face model, and recast the universality as unproven Temperley-Lieb algebra conjectures. The problem asks to complete the face-model theory: prove the factorization conjecture, obtain closed representations of the physical part, and clarify the boundary/fusion-path contributions.

Background

Integrable lattice models come in two families. In vertex models (such as the six-vertex model) degrees of freedom live on edges and the local Boltzmann weights are encoded in an R-matrix satisfying the Yang-Baxter equation. In interaction-round-a-face (IRF or face) models, degrees of freedom are heights $a$ assigned to lattice sites, restricted by an adjacency rule, and the Boltzmann weight $W(a,b,c,d\,|\,\lambda)$ of a face depends on the four surrounding heights and on a spectral parameter $\lambda$; integrability again follows from a (dynamical) Yang-Baxter equation for these weights, which guarantees that the row-to-row transfer matrices for different spectral parameters commute. The paradigmatic examples are the restricted solid-on-solid (RSOS) models of Andrews, Baxter and Forrester, whose heights take the values $1,\dots,r-1$ with adjacency $|a-b|=1$: in the critical regime these models describe the unitary minimal conformal field theories, and their one-point local height probabilities (the probability of a prescribed height at a single site) were computed by Baxter's corner transfer matrix method (transfer matrices encoding the partition function of a lattice quadrant, from which one-point functions follow directly) and by the bosonization of the corresponding vertex operators, which yields multiple-integral representations for multi-point local height probabilities, i.e. joint probabilities of prescribed heights on several adjacent sites (Lukyanov and Pugai). The transfer matrices of these face models double as Hamiltonians of quantum chains of non-Abelian anyons: the RSOS chain with parameter $r$ is equivalent to a chain of $su(2)_k$ anyons with $k=r-2$, and $r=5$ realizes the Fibonacci anyons whose fusion rule $\tau\times\tau=1+\tau$ underlies topological quantum computation proposals (the 'golden chain' of Feiguin et al.). The local projectors of these chains generate the Temperley-Lieb (TL) algebra, defined by generators $e_j$ with relations $e_j^2=d\,e_j$ and $e_j e_{j\pm1} e_j=e_j$, where the loop weight $d=2\cos(\pi/r)$ for the critical RSOS representations.

For vertex models, a remarkably complete theory of correlation functions exists. In an inhomogeneous generalization of the spin-1/2 Heisenberg chain, where each site carries its own spectral parameter (inhomogeneity) $\lambda_i$, the reduced density matrix $D_N(\lambda_1,\dots,\lambda_N)$ of $N$ consecutive sites satisfies discrete functional equations of reduced quantum Knizhnik-Zamolodchikov (qKZ) type (the qKZ equations are difference equations for correlation functions of vertex operators; their 'reduced' version acts directly on density matrices). These equations, supplemented by an asymptotic reduction relating $D_N$ to $D_{N-1}$ when one spectral parameter is sent to infinity, determine $D_N$ uniquely. Boos, Jimbo, Miwa, Smirnov and Takeyama discovered that $D_N$ then factorizes: it can be written as a finite sum of products of the nearest-neighbour two-point function $\omega(\lambda_i,\lambda_j)$ (the 'physical part', which carries all dependence on the system size, the state, and the inhomogeneities) multiplied by recursively defined 'structure functions' $f_{N;I,J}(\lambda_1,\dots,\lambda_N)$ (the 'algebraic part') that are elementary functions of the spectral parameters only. Using the fermionic basis — a special basis of the operator space of the XXZ model with a hidden Grassmann structure — this factorization was proven to hold for arbitrary correlation functions at finite length and finite temperature (Jimbo, Miwa and Smirnov), and the physical part was characterized by integral formulae and by solutions of linear and non-linear integral equations (NLIE) (Boos and Göhmann; Aufgebauer and Klümper). This machinery makes short-distance correlation functions of the Heisenberg chain exactly computable.

Frahm and Westerfeld (2021) initiated the transfer of this program to face models. They solved the inverse problem for a class of local operators, expressing them through generalized transfer matrices of inhomogeneous IRF models, and derived the analogue of the discrete reduced-qKZ equations for the reduced density matrices. Applying these equations to the cyclic solid-on-solid (CSOS) model (whose heights differ by $\pm1$ modulo $r$, so that the adjacency graph is a cycle rather than a path), the RSOS models with $r=4,5$, and the Fibonacci anyon chain, they found that the density matrices are uniquely determined by the functional equations combined with the analyticity properties inherited from the Boltzmann weights. In topological sectors with quantum dimension $d_q=1$ they observed a further structure: the periodic RSOS Hilbert space decomposes into sectors labeled by a fixed boundary height $q$, whose degeneracy grows with the quantum dimension $d_q$ (the component of the Perron-Frobenius eigenvector of the adjacency graph at node $q$); the sectors with $d_q=1$ (boundary heights $1$ and $r-1$) contain the ground states. There, the two- and three-site density matrices of the $r=4$ and $r=5$ RSOS models could be expressed through a single nearest-neighbour two-site function together with elementary structure functions independent of the model parameters. However, the explicit solution of the discrete functional equations for the two-site function was available only for small systems or special eigenstates (the CSOS reference state), and the authors formulated the central open questions of the field-model program, which form the present problem.

Problem Statement

Complete the model-independent theory of reduced density matrices for integrable face models, bringing it to the level achieved for the spin-1/2 Heisenberg (XXZ) chain. The components serve this single objective — making the short-distance correlation functions of integrable face models and the equivalent anyon chains exactly computable — and are linked by the common framework of the discrete reduced-qKZ functional equations. Components (a), (b), and (c) below are jointly mandatory parts of this objective: each is an independent sub-question, and a resolution requires a definite result for all three; a partial result within a component (for example a single topological sector, a single RSOS regime, a single model class, or a single system size) does not by itself resolve the task.

(a) Prove or disprove the factorization conjecture: for every critical RSOS model (all $r\ge 4$), in topological sectors with quantum dimension $d_q=1$, the $N$-site reduced density matrix of the inhomogeneous model takes the factorized form

$D_N(\lambda_1,\dots,\lambda_N)=\sum_{m=0}^{\lfloor N/2\rfloor}\;\sum_{I,J}\Big(\prod_{p=1}^{m}\omega(\lambda_{i_p},\lambda_{j_p})\Big)\,f_{N;I,J}(\lambda_1,\dots,\lambda_N),$

where $I=(i_1<\dots<i_m)$ and $J=(j_1<\dots<j_m)$ are disjoint index sets with $i_p<j_p$, $\omega$ is the nearest-neighbour two-point function of the given sector and transfer-matrix eigenstate, and the structure functions $f_{N;I,J}$ are elementary functions of the spectral parameters that are independent of the system size, the choice of inhomogeneities, and the eigenstate. Determine the corresponding statement (expected to involve two nearest-neighbour functions instead of one) for the sectors with $d_q\ne 1$, for the massive RSOS regimes, and answer the source question of whether such factorization extends to general integrable IRF models beyond the RSOS class. Each of these three extension clauses — the $d_q\ne 1$ sectors, the massive regimes, and general IRF models — requires a definite result: the corresponding statement determined, or the extension question answered, with the argument establishing it, and not merely a restatement of the expected form.

(b) Obtain a closed, model-independent representation of the physical part: integral formulae, linear or non-linear integral equations, or a fermionic-basis analog, for the two-point function $\omega(\lambda_1,\lambda_2)$ at arbitrary finite length and finite temperature, including a proof or refutation of the conjectured thermodynamic-limit ground-state expressions $\omega_r(x)$ of Westerfeld et al. (2023) — where $x=\lambda_1-\lambda_2$ is the difference of the two spectral parameters, restricted to the physical analyticity strip — which are so far supported only by finite-size numerics. For the sectors with $d_q\ne 1$, where the physical part is expected to be described by two nearest-neighbour functions rather than the single $\omega$, obtaining the corresponding closed representations is a further open step, which Westerfeld et al. (2023) suggest studying along the same lines.

(c) Clarify the boundary and fusion-path contributions to RSOS correlation functions: determine how the peculiar fusion-path structure of the RSOS Hilbert space (the constraint that admissible height configurations form paths on the adjacency graph) manifests itself as boundary terms in the correlation functions of the corresponding vertex-model chains. Here the boundary terms are the explicit contributions to the RSOS correlation functions that originate in this fusion-path restriction and distinguish them from the corresponding root-of-unity Heisenberg-chain correlators. This component is resolved by a definite determination of these boundary terms for the critical RSOS models (all $r\ge 4$) and their corresponding anisotropic Heisenberg chains, together with the argument establishing it — not by a single worked example or a survey. As proposed in the source, a comparison of the density matrices of the critical RSOS models with the corresponding quantities for the related anisotropic Heisenberg chains at roots of unity can provide insights into these boundary contributions.

References

1. H. Frahm and D. Westerfeld, "Density matrices in integrable face models", SciPost Phys. 11, 057 (2021), doi:10.21468/SciPostPhys.11.3.057, https://arxiv.org/abs/2104.06269
2. D. Westerfeld, M. Großpietsch, H. Kakuschke, and H. Frahm, "Factorization of density matrices in the critical RSOS models", J. Stat. Mech. (2023) 083104, doi:10.1088/1742-5468/aceeef, https://arxiv.org/abs/2303.15252
3. T. S. Tavares and G. A. P. Ribeiro, "Correlation functions of the six-vertex IRF model and its quantum spin chain", Nucl. Phys. B 1008, 116715 (2024), doi:10.1016/j.nuclphysb.2024.116715, https://arxiv.org/abs/2409.05814
4. K. Fukai, R. Kleinemühl, B. Pozsgay, and E. Vernier, "On correlation functions in models related to the Temperley-Lieb algebra", SciPost Phys. 16, 003 (2024), doi:10.21468/SciPostPhys.16.1.003, https://arxiv.org/abs/2309.07472
5. M. Blakeney, L. Corcoran, M. de Leeuw, B. Pozsgay, and E. Vernier, "Temperley-Lieb integrable models and fusion categories", J. High Energy Phys. 04 (2026) 165, doi:10.1007/JHEP04(2026)165, https://arxiv.org/abs/2510.19902
6. M. Sinha, T. S. Tavares, A. Roy, and H. Saleur, "Integrability and lattice discretizations of all topological defect lines in minimal CFTs", Nucl. Phys. B 1023, 117308 (2026), doi:10.1016/j.nuclphysb.2026.117308, https://arxiv.org/abs/2509.04257
7. H. Boos, M. Jimbo, T. Miwa, F. Smirnov, and Y. Takeyama, "Reduced qKZ equation and correlation functions of the XXZ model", Commun. Math. Phys. 261, 245 (2006), doi:10.1007/s00220-005-1430-6, https://arxiv.org/abs/hep-th/0412191
8. M. Jimbo, T. Miwa, and F. Smirnov, "Hidden Grassmann structure in the XXZ model III: introducing the Matsubara direction", J. Phys. A: Math. Theor. 42, 304018 (2009), doi:10.1088/1751-8113/42/30/304018, https://arxiv.org/abs/0811.0439
9. H. Boos and F. Göhmann, "On the physical part of the factorized correlation functions of the XXZ chain", J. Phys. A: Math. Theor. 42, 315001 (2009), doi:10.1088/1751-8113/42/31/315001, https://arxiv.org/abs/0903.5043
10. B. Aufgebauer and A. Klümper, "Finite temperature correlation functions from discrete functional equations", J. Phys. A: Math. Theor. 45, 345203 (2012), doi:10.1088/1751-8113/45/34/345203, https://arxiv.org/abs/1205.5702
11. S. Lukyanov and Ya. Pugai, "Multi-point local height probabilities in the integrable RSOS model", Nucl. Phys. B 473, 631 (1996), doi:10.1016/0550-3213(96)00221-0, https://arxiv.org/abs/hep-th/9602074
12. G. E. Andrews, R. J. Baxter, and P. J. Forrester, "Eight-vertex SOS model and generalized Rogers-Ramanujan-type identities", J. Stat. Phys. 35, 193 (1984), doi:10.1007/BF01014383, https://doi.org/10.1007/BF01014383
13. V. V. Bazhanov and N. Yu. Reshetikhin, "Critical RSOS models and conformal field theory", Int. J. Mod. Phys. A 4, 115 (1989), doi:10.1142/S0217751X89000042, https://doi.org/10.1142/S0217751X89000042
14. A. Feiguin, S. Trebst, A. W. W. Ludwig, M. Troyer, A. Kitaev, Z. Wang, and M. H. Freedman, "Interacting anyons in topological quantum liquids: The golden chain", Phys. Rev. Lett. 98, 160409 (2007), doi:10.1103/PhysRevLett.98.160409, https://arxiv.org/abs/cond-mat/0612341
15. R. J. Baxter, "Corner transfer matrices of the eight-vertex model. I. Low-temperature expansions and conjectured properties", J. Stat. Phys. 15, 485 (1976), doi:10.1007/BF01020802, https://doi.org/10.1007/BF01020802