Finite-size effects and root structure of the inhomogeneous Bethe ansatz equations for the half-filled Hubbard chain with nonparallel boundary magnetic fields

The one-dimensional Hubbard chain with integrable boundary magnetic fields pointing in different directions at the two ends is exactly solvable by the off-diagonal Bethe ansatz (ODBA), but the resulting Bethe ansatz equations (BAEs) are inhomogeneous: they contain an extra source term, proportional to a constant measuring the non-parallelism of the two boundary fields, that blocks the standard logarithmic route to the thermodynamic limit. Sun, Qiao, Yang, Cao and Yang (Phys. Rev. B 110, 035153 (2024)) computed the exact surface energy at half filling from the reduced (homogeneous) BAEs, justified dropping the inhomogeneous term by fitting density-matrix-renormalization-group data to $\delta e=|E-E_{\mathrm{hom}}|=\gamma N^{\tau}$ with $\tau\in[-0.96,-0.78]$ depending on the boundary-parameter region, and explicitly listed the finite-size effect and the patterns of the inhomogeneous BAEs as open problems. This problem asks for a controlled finite-$N$ theory of the inhomogeneous term for this model: the true asymptotic behavior of $\delta e$ (leading term, corrections, and proof or disproof of its vanishing) in each of the five boundary-parameter regions, and the finite-$N$ solution structure of the full inhomogeneous BAEs, including root patterns, boundary strings, and the modification of excited-state and finite-size spectra.

Background

The Hubbard model describes interacting electrons hopping on a lattice with hopping amplitude $t$ and on-site repulsion $U$; in one dimension with periodic boundaries it is exactly solvable by the Bethe ansatz (Lieb and Wu, 1968) and is the paradigmatic model of correlated electrons. On a chain with open ends one may add boundary magnetic fields $\mathbf{h}_1$ and $\mathbf{h}_N$ acting on the spins of the two end sites; integrable choices combine a Zeeman term with spin-flip terms. When the two fields are parallel, a global spin rotation makes the boundary terms diagonal, and the nested Bethe ansatz — a two-level Bethe ansatz for particles carrying spin, which yields coupled equations for charge and spin rapidities — applies, starting from a reference state, i.e. a trivially known eigenstate from which all others are generated. When they are nonparallel, no reference state exists; the model is still integrable, and the off-diagonal Bethe ansatz (ODBA) of Cao, Yang, Shi and Wang (2013) — which builds the eigenvalues of the transfer matrix (the generator of the model's commuting conserved quantities) from functional $T$-$Q$ relations, identities expressing those eigenvalues through simple functions of the rapidities, instead of from a reference state — yields exact Bethe ansatz equations (BAEs) for the charge rapidities $\{k_j\}$ and spin rapidities $\{\lambda_l\}$ (Li et al., 2014). These BAEs are inhomogeneous: alongside the usual logarithmically additive terms sits an extra source proportional to $c=2(\varepsilon|h_1||h_N|-\mathbf{h}_1\cdot\mathbf{h}_N)$, where $\varepsilon=\pm 1$ is the sign carried by $\mathbf{h}_1\cdot\mathbf{h}_N$, so that $c$ vanishes only when the two fields are parallel or antiparallel (the source paper sets $\varepsilon=1$ for convenience, which restricts it to fields with $\mathbf{h}_1\cdot\mathbf{h}_N>0$). Such equations cannot be treated by the standard route of taking logarithms and passing to root densities — the continuum description of the Bethe roots, on which the thermodynamic Bethe ansatz, the standard method for extracting thermodynamics from Bethe equations, is built — so thermodynamic-limit quantities were long inaccessible.

Sun, Qiao, Yang, Cao and Yang (2024) — the source of this problem — analyzed the half-filled chain (one electron per site). The ground-state root configurations of the reduced (inhomogeneous-term-dropped) BAEs fall into five regions of the boundary-parameter plane $\alpha=|h_1|$, $\beta=|h_N|$, delimited by the critical values $h_0$ (solving $(h_0^2-1)/(2h_0)+U/4=0$) and $1$; some regions carry imaginary roots, called boundary strings, which signal boundary bound states localized at the chain ends. From these patterns they computed the root densities, the ground-state energy density, and the surface energy — the $O(1)$ boundary contribution to $E_N=N e_{\mathrm{bulk}}+e_{\mathrm{surf}}+o(1)$ — exactly. Dropping the inhomogeneous term is justified numerically: with $E$ the exact ground-state energy (obtained by the density matrix renormalization group, DMRG, a variational tensor-network method for one-dimensional quantum systems) and $E_{\mathrm{hom}}$ that of the chain with the equivalent parallel fields of equal magnitudes, they fitted $\delta e=|E-E_{\mathrm{hom}}|=\gamma N^{\tau}$ with $\tau$ between $-0.96$ and $-0.78$ across the five regions, concluding that $\delta e\to 0$ so the reduced BAEs give exact thermodynamic-limit quantities. For the isotropic Heisenberg (XXX) spin chain the analogous vanishing of the inhomogeneous term has been proved (Qiao et al., 2023); for the Hubbard chain it remains a numerical observation, and the fitted exponents — close to, but not equal to, $-1$ — have no derivation or explanation.

For open chains with parallel boundary fields, finite-size corrections to the ground-state energy are classical (Asakawa and Suzuki, 1996, via boundary conformal field theory, the conformal description of a critical system with edges, which fixes the universal $1/N$ corrections). For nonparallel fields the only available results are for spin chains: Frahm, Grelik and Seel (2012) computed the finite-size corrections, including the boundary-boundary correlation term, of the open XXX chain with nonparallel fields from functional relations (identities among commuting transfer matrices); Dong et al. (2023) obtained its surface energy and elementary excitations; Ji, Sun, Xu, Qiao, Cao and Yang (2025) determined the large-$N$ root patterns of the inhomogeneous ODBA equations of the same chain by tensor networks; and Frahm, Klümper, Wagner and Zhang (2026) derived nonlinear integral equations for it, in which a third source function captures the inhomogeneous ODBA contribution. None of these constructions covers the nested fermionic Hubbard case, whose two coupled rapidity sets and boundary strings make each of them harder. The source paper's conclusion explicitly leaves 'the finite size effect and patterns of the inhomogeneous BAEs' — together with a separate finite-temperature extension — as open future research topics.

Problem Statement

For the half-filled one-dimensional Hubbard chain with integrable nonparallel boundary magnetic fields $\mathbf{h}_1$, $\mathbf{h}_N$ satisfying $\mathbf{h}_1\cdot\mathbf{h}_N>0$ (equivalently $\varepsilon=1$, the convention in which the source states its equations and its five boundary-parameter regions; the perpendicular case $\mathbf{h}_1\cdot\mathbf{h}_N=0$ is excluded) — the Hamiltonian and the inhomogeneous Bethe ansatz equations (BAEs) exactly as published by Sun, Qiao, Yang, Cao and Yang (Phys. Rev. B 110, 035153 (2024); arXiv:2401.14356), whose inhomogeneous term is proportional to $2(\varepsilon|h_1||h_N|-\mathbf{h}_1\cdot\mathbf{h}_N)$ with $\varepsilon=\pm 1$ the sign carried by $\mathbf{h}_1\cdot\mathbf{h}_N$ in that paper — characterize what that paper left open, namely the finite-$N$ structure of the inhomogeneous term.

(1) Determine the asymptotic behavior of its ground-state contribution $\delta e=|E-E_{\mathrm{hom}}|$ as $N\to\infty$ in each of the five boundary-parameter regions: the leading term (the source's fits suggest approximate power laws $\delta e=\gamma N^{\tau}$ with $\tau\in[-0.96,-0.78]$, but no exponent has been derived and the fitted values are unexplained), the corrections to the leading term, and whether $\delta e\to 0$ can be established by proof rather than by numerical fit.

(2) Characterize the finite-$N$ solution structure of the full inhomogeneous BAEs: the configurations of the charge rapidities $\{k_j\}$ and spin rapidities $\{\lambda_l\}$ — real roots and boundary strings — for finite $N$ in each region, and how the inhomogeneous term modifies the excited states and the finite-size spectrum.

These two deliverables form one objective: a controlled finite-$N$ theory of the ODBA inhomogeneous term for this model, replacing the source paper's numerical power-law fit. Deliverable (2) is scoped to leading order in $1/N$: it suffices to classify the charge and spin roots $\{k_j\}$, $\{\lambda_l\}$ — real roots and boundary strings — in each region at that order and to state how the inhomogeneous term modifies the ground state and the lowest excited states, not the full excitation tower or higher orders in $1/N$; and (2)'s spectral requirement does not re-request the proof-level treatment of the ground-state $\delta e$ already demanded in (1). The source paper's separately listed finite-temperature extension is not part of this problem.

References

1. Pei Sun, Yi Qiao, Tao Yang, Junpeng Cao, Wen-Li Yang, "Exact surface energy of the Hubbard model with nonparallel boundary magnetic fields", Phys. Rev. B 110, 035153 (2024); DOI 10.1103/PhysRevB.110.035153; arXiv:2401.14356, https://doi.org/10.1103/PhysRevB.110.035153
2. Elliott 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, 1445-1448 (1968); DOI 10.1103/PhysRevLett.20.1445, https://doi.org/10.1103/PhysRevLett.20.1445
3. Junpeng Cao, Wen-Li Yang, Kangjie Shi, Yupeng Wang, "Off-diagonal Bethe ansatz and exact solution of a topological spin ring", Phys. Rev. Lett. 111, 137201 (2013); DOI 10.1103/PhysRevLett.111.137201; arXiv:1305.7328, https://arxiv.org/abs/1305.7328
4. Yuan-Yuan Li, Junpeng Cao, Wen-Li Yang, Kangjie Shi, Yupeng Wang, "Exact solution of the one-dimensional Hubbard model with arbitrary boundary magnetic fields", Nucl. Phys. B 879, 98-109 (2014); DOI 10.1016/j.nuclphysb.2013.12.004; arXiv:1311.0432, https://arxiv.org/abs/1311.0432
5. Hitoshi Asakawa, Masuo Suzuki, "Finite-size corrections in the XXZ model and the Hubbard model with boundary fields", J. Phys. A: Math. Gen. 29, 225-245 (1996); DOI 10.1088/0305-4470/29/2/004, https://doi.org/10.1088/0305-4470/29/2/004
6. H. Frahm, J. H. Grelik, A. Seel, "Persistent currents in open spin chains", Theor. Math. Phys. 171, 715-724 (2012); DOI 10.1007/s11232-012-0069-7, https://doi.org/10.1007/s11232-012-0069-7
7. Yi Qiao, Junpeng Cao, Wen-Li Yang, Kangjie Shi, Yupeng Wang, "Off-diagonal approach to the exact solution of quantum integrable systems", Chin. Phys. B 32, 117504 (2023); DOI 10.1088/1674-1056/ad0774; arXiv:2312.04153, https://arxiv.org/abs/2312.04153
8. Jia-Sheng Dong, Pengcheng Lu, Pei Sun, Yi Qiao, Junpeng Cao, Kun Hao, Wen-Li Yang, "Exact surface energy and elementary excitations of the XXX spin-1/2 chain with arbitrary non-diagonal boundary fields", Chin. Phys. B 32, 017501 (2023); DOI 10.1088/1674-1056/ac981f, https://doi.org/10.1088/1674-1056/ac981f
9. Zhouzheng Ji, Pei Sun, Xiaotian Xu, Yi Qiao, Junpeng Cao, Wen-Li Yang, "Tensor-Network Analysis of Root Patterns in the XXX Model with Open Boundaries", arXiv:2512.24182 (2025), https://arxiv.org/abs/2512.24182
10. Holger Frahm, Andreas Klümper, Dennis Wagner, Xin Zhang, "Non-linear integral equations for the XXX spin-1/2 quantum chain with non-diagonal boundary fields", SciPost Phys. 20, 012 (2026); DOI 10.21468/scipostphys.20.1.012, https://doi.org/10.21468/scipostphys.20.1.012