Algebraic Bethe ansatz for 19-vertex models with upper triangular K-matrices

By means of an algebraic Bethe ansatz approach we study the Zamolodchikov-Fateev and Izergin-Korepin vertex models with non-diagonal boundaries, characterized by reflection matrices with an upper triangular form. Generalized Bethe vectors are used to diagonalize the associated transfer matrix. The eigenvalues as well as the Bethe equations are presented.

💡 Research Summary

The paper addresses the long‑standing problem of applying the algebraic Bethe ansatz (ABA) to nineteen‑vertex integrable models with non‑diagonal, upper‑triangular boundary reflections. Specifically, the authors consider the Zamolodchikov‑Fateev (ZF) and Izergin‑Korepin (IK) models, whose bulk interactions are encoded in a 19×19 R‑matrix that satisfies the Yang‑Baxter equation. For both models the R‑matrix is given explicitly in terms of trigonometric functions (equations (10) and (11)), and its fundamental properties—PT symmetry, unitarity and crossing‑unitarity—are verified.

Boundary conditions are introduced via K‑matrices K⁻(u) and K⁺(u) that are upper‑triangular (equation (14)). The right‑hand matrix K⁻(u) contains three diagonal entries k⁻₁₁, k⁻₂₂, k⁻₃₃ and three non‑zero off‑diagonal elements k⁻₁₂, k⁻₂₃, k⁻₁₃, whose functional forms depend on free parameters ξ±, β± and an auxiliary sign ε (equations (15) for ZF and (16) for IK). The left‑hand matrix K⁺(u) is obtained from K⁻ by the standard crossing transformation (equation (17)).

The authors then construct the double‑row monodromy matrix Uₐ(u)=Tₐ(u)K⁻(u)T̂ₐ(−u) (equation (3)) and decompose it into operators Aⱼ(u), Bⱼ(u), Cⱼ(u) (equation (19)). Using the reflection equations they derive the full set of commutation relations among these operators (Appendix A). A reference (pseudo‑vacuum) state Ψ₀ is chosen as the tensor product of the local highest‑weight vector (1,0,0)ᵀ at each site (equation (20)). Acting on Ψ₀, the diagonal operators Dⱼ(u) (shifted versions of Aⱼ) produce scalar eigenvalues Δⱼ(u) (equations (21)–(23)), while the C‑operators annihilate Ψ₀. Consequently the double‑row transfer matrix t(u)=tr