Controlled finite-bond-dimension extrapolation for the Kosterlitz–Thouless point of the triangular Ising antiferromagnet in a field and the approach of its Potts critical line

The antiferromagnetic Ising model on the triangular lattice in a uniform field is a canonical frustrated magnet. In reduced coordinates $H=h/T$ and $K=J/T$, the zero-temperature line $K^{-1}=0$ carries a field-induced Kosterlitz–Thouless (KT) transition separating a critical phase from the ordered two-up-one-down ($m=1/3$) plateau phase. Coulomb-gas and solid-on-solid theory fixes the universality (correlations decaying with exponent $\eta=4/9$ at the transition) but not the location $H_{KT}$, whose determinations disagree: transfer-matrix finite-size scaling with logarithmic corrections gives $H_{KT}=0.266\pm0.01$ (reanalyzed to $0.26\pm0.02$), Monte Carlo Yang–Lee zeros (zeros of the partition function in a complex magnetic field, which pinch the real axis at the transition) give $0.29(3)$, and a $C_3$-symmetric corner-transfer-matrix renormalization group (CTMRG) study gives $0.305\pm0.006$ as an upper bound. The CTMRG estimate cannot presently be reconciled with the transfer-matrix value because no controlled scheme of finite-bond-dimension corrections analogous to the finite-size logarithmic and power-law corrections has been established; existing finite-$\chi$ and finite-entanglement scaling analyses at KT-type transitions do not yet supply one, and it remains unclear how to carry those corrections over to finite bond dimension. The problem is to develop and validate such a controlled finite-$\chi$ extrapolation framework for tensor-network contractions at KT transitions, to use it to decide the location of $H_{KT}$ with an uncertainty small enough to discriminate the competing estimates, and to determine the asymptotic square-root approach of the finite-temperature three-state Potts critical line to the KT point at reduced temperatures below $K^{-1}=0.25$.

Background

The antiferromagnetic Ising model on the triangular lattice in a uniform magnetic field (the triangular Ising antiferromagnet, TIAF) is defined by the Hamiltonian $\mathcal{H}=J\sum_{\langle i,j\rangle}\sigma_i\sigma_j-h\sum_i\sigma_i$ with Ising spins $\sigma_i=\pm1$ and antiferromagnetic coupling $J>0$. Because no spin configuration can satisfy all three bonds of a triangular plaquette, the zero-field ground state is macroscopically degenerate: it obeys the two-up-one-down (UUD) rule on every triangle and has a finite residual entropy. At $h=0$ and $T=0$ the model sits at a critical point (the Villain–Stephenson point) where spin–spin correlations decay algebraically with exponent $\eta=1/2$. This point admits an exact mapping onto a triangular solid-on-solid (SOS) model — a height model whose spin configuration is encoded in a discrete interface height — whose long-distance physics is captured by a Coulomb-gas description, i.e. a representation of the long-wavelength height modes as a gas of integer topological charges interacting through a logarithmic potential. It is conventional to work with the reduced field $H=h/T$ and coupling $K=J/T$, so that the phase diagram in the $(H,K^{-1})$ plane has the $K^{-1}=0$ axis (the constrained model) as its zero-temperature limit.

Introducing a finite reduced field at the Villain–Stephenson point perturbs the height model. Renormalization-group analysis of the SOS/Coulomb-gas mapping (Nienhuis, Hilhorst and Blöte) shows that this perturbation becomes relevant when the correlation-decay exponent reaches $\eta=4/9$, where a Kosterlitz–Thouless (KT) transition occurs. A KT transition is an infinite-order transition driven by the binding of topological vortices: it has no local order parameter, and on one side of it lies a critical phase in which correlations decay algebraically with an exponent that varies continuously along the phase. The theory fixes the universality class of this field-induced transition but not the critical reduced field $H_{KT}$ at which it occurs. For $0<H<H_{KT}$ the constrained model is critical; beyond it the system orders into the $\sqrt{3}\times\sqrt{3}$ state — a three-sublattice pattern with two spins up and one down on every triangle, breaking the lattice's threefold rotational symmetry — which forms the $m=1/3$ magnetization plateau, the regime in which the magnetization per site stays pinned at one third of its saturation value; this plateau persists up to $h=6J$, above which the ground state is fully polarized. At finite temperature the ordered phase melts along a critical line in the three-state Potts universality class, and the same renormalization-group arguments predict that this line meets the KT point with a leading square-root shape, $K_c^{-1}\propto(H_c-H_{KT})^{1/2}+o(H_c-H_{KT})$.

The location of $H_{KT}$ has a thirty-year history of conflicting determinations. Transfer-matrix calculations with phenomenological renormalization (comparing gaps of finite-width transfer matrices) established the KT character of the transition and gave $H_{KT}=0.266\pm0.01$; crucially, this number emerged only after fitting finite-size data with both logarithmic and power-law corrections, $H_{KT}(L)=H_{KT}+a/(b+\ln L)+cL^{-2}$, for system sizes $L\le27$. An independent transfer-matrix analysis without such corrections gave the inconsistent $0.211(7)$, and a later reanalysis of the same data with more conservative error treatment gave $0.26\pm0.02$. A Monte Carlo determination via Yang–Lee zeros (zeros of the partition function in a complex field, which pinch the real axis at the transition) gave $0.29(3)$. Most recently, Nyckees, Rufino, Mila and Colbois introduced a $C_3$-symmetric corner transfer matrix renormalization group algorithm on the honeycomb lattice, adapted to the threefold rotational symmetry of the triangular lattice, and located the KT transition via the criterion $\eta=4/9$ applied to finite-bond-dimension data, obtaining $H_{KT}=0.305\pm0.006$ — which the authors themselves state should be taken as an upper bound, because the fitted exponent drifts with the range of bond dimensions used in the fit.

The corner transfer matrix renormalization group (CTMRG, introduced by Nishino and Okunishi) is a deterministic contraction method for infinite two-dimensional tensor networks: corner and edge tensors of the environment are iteratively grown and truncated to a fixed bond dimension $\chi$ through an isometry, so that $\chi$ acts as a cutoff on the correlation length, playing a role analogous to entanglement truncation in matrix product states (one-dimensional wavefunctions built from site-local matrices of bounded dimension). Extrapolation in $\chi$ (finite-entanglement scaling) is comparatively well understood for ordinary power-law critical points, where truncation acts like an effective finite size. At KT transitions, however, even finite-size scaling itself requires logarithmic corrections, and the analogue of these corrections for finite $\chi$ has not been developed — which is exactly why the CTMRG estimate and the transfer-matrix estimate of $H_{KT}$ cannot presently be reconciled. The same study could not reach reduced temperatures $K^{-1}<0.25$, because the physical correlation length diverges exponentially as $K^{-1}\to0$ and would demand exponentially large $\chi$; its fits of the finite-temperature critical line turned out to be compatible with both candidate values of $H_{KT}$.

Problem Statement

Determine the zero-temperature limit of the phase diagram of the triangular Ising antiferromagnet in a field, in reduced coordinates $(H,K^{-1})$, anchored at the field-induced Kosterlitz–Thouless point of the constrained model at $K^{-1}=0$. Concretely, the following linked deliverables define completion:

1. Develop a controlled extrapolation and correction framework for finite-bond-dimension ($\chi$) tensor-network contractions (CTMRG or an equivalent contraction method) at Kosterlitz–Thouless transitions — the analogue for finite $\chi$ of the logarithmic and power-law finite-size corrections used in transfer-matrix finite-size scaling.

2. Apply this framework to the constrained model ($K^{-1}=0$) to determine the critical reduced field $H_{KT}$ with a total uncertainty (statistical plus systematic) small enough to discriminate between the transfer-matrix estimate $H_{KT}=0.266\pm0.01$ (reanalyzed to $0.26\pm0.02$), the Monte Carlo Yang–Lee-zero value $0.29(3)$, and the CTMRG upper bound $0.305\pm0.006$, thereby deciding which, if any, of the existing estimates is correct.

3. Using the same controlled framework, map the finite-temperature three-state Potts critical line down to reduced temperatures below $K^{-1}=0.25$ and test the predicted asymptotic approach $K_c^{-1}\propto(H_c-H_{KT})^{1/2}+o(H_c-H_{KT})$, so that the line shape and the KT point are determined consistently with one another.

These deliverables form one scientific objective — the location of the KT point and the approach of the Potts critical line to it, with the finite-$\chi$ correction framework as the enabling methodology whose acceptance runs through its validated application to the model. Algorithmic improvements to the honeycomb CTMRG — the choice of isometry, multisite or arbitrary-unit-cell formulations, automatic-differentiation variants, or adaptation to contract two-dimensional tensor-network wavefunctions such as iPEPS (infinite projected entangled-pair states) and PESS — are admissible means toward this objective, not acceptance criteria in themselves; likewise, extending the critical line to reduced temperatures arbitrarily far below the $K^{-1}<0.25$ range that deliverable 3 requires is welcome but not required for acceptance.

References

1. Samuel Nyckees, Afonso Rufino, Frédéric Mila, Jeanne Colbois, "Critical line of the triangular Ising antiferromagnet in a field from a $C_3$-symmetric corner transfer matrix algorithm", Physical Review E 108, 064132 (2023). DOI: 10.1103/PhysRevE.108.064132, arXiv:2306.09046, https://doi.org/10.1103/PhysRevE.108.064132
2. H. W. J. Blöte, M. P. Nightingale, X. N. Wu, A. Hoogland, "Field-induced Kosterlitz-Thouless transition in the zero-temperature triangular Ising antiferromagnet", Physical Review B 43, 8751 (1991). DOI: 10.1103/PhysRevB.43.8751, https://doi.org/10.1103/PhysRevB.43.8751
3. H. W. J. Blöte, M. P. Nightingale, "Antiferromagnetic triangular Ising model: Critical behavior of the ground state", Physical Review B 47, 15046 (1993). DOI: 10.1103/PhysRevB.47.15046, https://doi.org/10.1103/PhysRevB.47.15046
4. S. L. A. de Queiroz, T. Paiva, J. S. de Sá Martins, R. R. dos Santos, "Field-induced ordering in critical antiferromagnets", Physical Review E 59, 2772 (1999). DOI: 10.1103/PhysRevE.59.2772, https://doi.org/10.1103/PhysRevE.59.2772
5. Xiaofeng Qian, Maarten Wegewijs, H. W. J. Blöte, "Critical frontier of the triangular Ising antiferromagnet in a field", Physical Review E 69, 036127 (2004). DOI: 10.1103/PhysRevE.69.036127, arXiv:cond-mat/0312031, https://doi.org/10.1103/PhysRevE.69.036127
6. Chi-Ok Hwang, Seung-Yeon Kim, "Field-induced Kosterlitz–Thouless transition in critical triangular-lattice antiferromagnets", Monte Carlo Methods and Applications 20, 217 (2014). DOI: 10.1515/mcma-2013-0027, https://doi.org/10.1515/mcma-2013-0027
7. B. Nienhuis, H. J. Hilhorst, H. W. J. Blöte, "Triangular SOS models and cubic-crystal shapes", Journal of Physics A: Mathematical and General 17, 3559 (1984). DOI: 10.1088/0305-4470/17/18/025, https://doi.org/10.1088/0305-4470/17/18/025
8. John Stephenson, "Ising-Model Spin Correlations on the Triangular Lattice. III. Isotropic Antiferromagnetic Lattice", Journal of Mathematical Physics 11, 413 (1970). DOI: 10.1063/1.1665154, https://doi.org/10.1063/1.1665154
9. Tomotoshi Nishino, Kouichi Okunishi, "Corner Transfer Matrix Renormalization Group Method", Journal of the Physical Society of Japan 65, 891 (1996). DOI: 10.1143/jpsj.65.891, https://doi.org/10.1143/jpsj.65.891
10. Hiroshi Ueda, Kouichi Okunishi, Kenji Harada, Roman Krčmár, Andrej Gendiar, Seiji Yunoki, Tomotoshi Nishino, "Finite-m scaling analysis of Berezinskii-Kosterlitz-Thouless phase transitions and entanglement spectrum for the six-state clock model", Physical Review E 101, 062111 (2020). DOI: 10.1103/PhysRevE.101.062111, https://doi.org/10.1103/PhysRevE.101.062111
11. Atsushi Ueda, Masaki Oshikawa, "Finite-size and finite bond dimension effects of tensor network renormalization", Physical Review B 108, 024413 (2023). DOI: 10.1103/PhysRevB.108.024413, https://doi.org/10.1103/PhysRevB.108.024413
12. Luke Hodgkiss, Laurens Lootens, Atsushi Ueda, Bram Vanhecke, Frank Verstraete, "On the origin of finite entanglement scaling", arXiv:2607.15124 (2026), https://arxiv.org/abs/2607.15124
13. Sing-Hong Chan, Pochung Chen, "Extracting conformal data from finite-size tensor-network flow in critical two-dimensional classical models", arXiv:2604.15749 (2026), https://arxiv.org/abs/2604.15749
14. Qi Yang, Philippe Corboz, "Efficient iPEPS simulation on the honeycomb lattice via QR-based corner transfer matrix renormalization group", Physical Review B 113, 085109 (2026). DOI: 10.1103/9gmp-byx8, arXiv:2509.05090, https://doi.org/10.1103/9gmp-byx8
15. Yutaka Okabe, Hiromi Otsuka, "Berezinskii-Kosterlitz-Thouless phase transitions of the antiferromagnetic Ising model with ferromagnetic next-nearest-neighbor interactions on the kagome lattice", Physical Review B 113, 014441 (2026). DOI: 10.1103/599c-896h, arXiv:2602.07401, https://doi.org/10.1103/599c-896h
16. Wooseop Kwak, Sanghyun Kim, Chi-Ok Hwang, "Phase transitions of 2D triangular antiferromagnetic Ising model in a uniform magnetic field near h=0 and T=0", Journal of Physics: Conference Series 750, 012012 (2016). DOI: 10.1088/1742-6596/750/1/012012, https://doi.org/10.1088/1742-6596/750/1/012012