The lattice gas on the square lattice in which each particle excludes occupation of all sites up to its second-nearest neighbors (the 2NN model, equivalent to 2×22\times2 hard squares) undergoes a continuous transition between a disordered fluid and a fourfold-degenerate columnar phase. Its universality class has been contested for two decades: Monte Carlo studies have variously reported two-dimensional Ising exponents and Ashkin–Teller criticality, and an independent 2021 attack via exact solutions on sequences of Husimi lattices combined with the coherent anomaly method proved inconclusive, its exponent estimates swinging by roughly 10% for a 1% change in the assumed critical chemical potential. The problem is to determine this universality class — Ising, Ashkin–Teller (with the corresponding location on the Ashkin–Teller critical line), or another class — with error control sufficient to discriminate the competing candidates, resolving the documented sensitivities to the critical-point location and to lattice-neighborhood counting biases in cluster approximations. No study to date settles the question: the reported determinations remain in conflict, and none excludes the competing candidates with controlled uncertainties, so the problem remains open.

Background

Hard-core lattice gases are athermal models whose only interaction is a geometrical exclusion rule, so that any phase transition they display is driven entirely by entropy. In the kkNN family on the square lattice, a particle placed on a site forbids occupation of all sites up to its kk-th nearest neighbors. The 1NN model is equivalent to hard squares of side 2a\sqrt{2}a tilted by 45° with respect to the lattice (with aa the lattice spacing); its fluid–solid (checkerboard) transition is continuous and lies in the two-dimensional Ising universality class, with critical parameters known to very high precision from transfer-matrix calculations. The 2NN model is equivalent to 2×22\times2 hard squares: at the maximum density ρ2,max⁡=1/4\rho_{2,\max}=1/4 the ground state is not a commensurate solid but a fourfold-degenerate columnar phase, in which one of four sublattices A,…,DA,\dots,D is preferentially occupied. A standard order parameter for the fluid–columnar transition is Q2=(∣ρ2A−ρ2C∣+∣ρ2B−ρ2D∣)/ρ2,max⁡Q_2=(|\rho_{2A}-\rho_{2C}|+|\rho_{2B}-\rho_{2D}|)/\rho_{2,\max}, where ρ2S\rho_{2S} are sublattice densities, and Q2=0Q_2=0 in the fluid. By a universality class one means the set of critical exponents and scaling functions shared by all continuous transitions with the same symmetries and spatial dimension; for the two-dimensional Ising class the order-parameter exponent is β=1/8\beta=1/8 and the correlation-length exponent is ν=1\nu=1. The Ashkin–Teller model consists of two coupled Ising copies, and its critical line runs from the decoupled-Ising point to the four-state Potts point (the critical point of the q=4q=4 Potts spin model); along this line the exponents vary continuously while the magnetic scaling dimension stays fixed, so that β=ν/8\beta=\nu/8 with ν\nu between 2/32/3 (four-state Potts) and 11 (decoupled Ising). The position on this line is fixed by the coupling between the two Ising copies, conventionally written as a four-spin coupling.

For the 2NN model, most studies agree that the fluid–columnar transition is continuous, with critical chemical potential μ2,c≈4.58\mu_{2,c}\approx4.58–4.664.66 and critical density ρ2,c≈0.23\rho_{2,c}\approx0.23, but the universality class has been disputed. Monte Carlo simulations by Fernandes, Arenzon and Levin (2007) claimed Ising behavior, in contradiction to the naive Landau–Lifshitz expectation of an XY model with cubic anisotropy (the continuum field theory that symmetry considerations alone suggest for a fourfold-degenerate planar order parameter). High-precision simulations by Feng, Blöte and Nienhuis (2011) found exponents deviating from the Ising values and hinted at Ashkin–Teller criticality, and Ramola, Damle and Dhar (2015) reported that in mixtures of 2×12\times1 dimers and 2×22\times2 hard squares the critical exponents vary with composition in the manner predicted by Ashkin–Teller theory, including in the pure hard-square limit, where exponents close to β≈0.115\beta\approx0.115 and ν≈0.92\nu\approx0.92 were obtained; an interfacial-tension calculation by Mandal, Nath and Rajesh (2017) gave μ2,c≈4.66\mu_{2,c}\approx4.66, about 1% above the Monte Carlo values.

An independent, semi-analytical route is provided by Husimi lattices and the coherent anomaly method (CAM). A Husimi lattice is a treelike ('cactus') lattice obtained by replacing the sites of a Bethe lattice (an infinite loopless regular lattice) by finite clusters; solving a model exactly on a sequence of such lattices with growing building blocks (level LL) yields a systematically improvable family of mean-field approximations. Every member of the sequence has classical critical exponents (β=1/2\beta=1/2), but CAM extracts the true exponents from how the non-universal amplitudes of the sequence — the 'coherent anomalies' — diverge as the distance between the finite-level and true critical points shrinks. Rodrigues and Oliveira (2021) solved the 1NN and 2NN models exactly on square Husimi lattices built from diagonal blocks of 2L(L+1)2L(L+1) sites, up to L=7L=7. For the 1NN model the extrapolated critical parameters are extremely accurate and CAM yields β≈0.121\beta\approx0.121 and ν≈1.06\nu\approx1.06, within 6% of the Ising values. For the 2NN model the analysis is inconclusive: fixing the true critical chemical potential at 4.58, 4.629 or 4.66 gives β=0.146(14)\beta=0.146(14), 0.132(10)0.132(10) or 0.123(8)0.123(8), respectively — a swing of about 10% in β\beta for a 1% change in the assumed μ2,c∗\mu^*_{2,c} — and with their own estimate one finds ν≈1.33\nu\approx1.33, some 45% above the Ashkin–Teller value (and coincidentally close to two-dimensional percolation exponents, an agreement the authors consider accidental). The authors identify specific obstacles on this route: the definition of second neighbors on the Husimi blocks is ambiguous (one over- or under-counts the next-nearest neighbors of the corner sites, and the under-counting branch behaves non-systematically in LL), the critical free energy ϕ2,c(L)\phi_{2,c}(L) converges so poorly that no thermodynamic-limit value can be quoted, and exact enumeration grows explosively with LL — the recursion relations already involve of order 101810^{18} terms at L=7L=7 — capping the sequence at small levels. Their conclusion is that they are 'not in a position to draw any conclusion about the universality class of the 2NN model', specifically as regards the dispute between Ising and Ashkin–Teller criticality, while regarding Ashkin–Teller criticality as the more plausible scenario.

Problem Statement

Determine the universality class of the continuous fluid–columnar transition of the 2NN hard-square lattice gas on the square lattice. Specifically: establish whether its critical behavior is that of the two-dimensional Ising class (β=1/8\beta=1/8, ν=1\nu=1), of the Ashkin–Teller critical line (where β=ν/8\beta=\nu/8, with the Monte Carlo estimates β≈0.115\beta\approx0.115 and ν≈0.92\nu\approx0.92 near its decoupled-Ising end), or of some other universality class — and, if it is Ashkin–Teller, locate the model on that line (equivalently, determine ν\nu or the effective four-spin coupling). The determination must carry error control sufficient to discriminate between the competing candidates: with the leading candidates separated by only about 8% in ν\nu (Ising ν=1\nu=1 versus the decoupled-Ising-end Ashkin–Teller estimate ν≈0.92\nu\approx0.92), this requires uncertainties on ν\nu and β\beta small enough that one candidate is excluded at the two-sigma level — of order a few percent in ν\nu. New computation and error-controlled re-analysis of published data are both admissible routes to this standard; either way the exponent estimates must be robust against the residual uncertainty in the critical chemical potential (current estimates span μ2,c≈4.58\mu_{2,c}\approx4.58–4.664.66) and against systematic biases of the approximation used, such as the over/under-counting of second neighbors in cluster constructions. No particular method is prescribed: the determination may be pursued by pushing the exact Husimi-lattice/coherent-anomaly sequence to larger levels or to alternative systematic cluster families free of the neighbor-counting ambiguity, or by independent high-precision numerical or analytical determination of the critical exponents.

References

  1. N. T. Rodrigues and T. J. Oliveira, 'Husimi-lattice solutions and the coherent-anomaly-method analysis for hard-square lattice gases', Phys. Rev. E 103, 032153 (2021). DOI: 10.1103/PhysRevE.103.032153; arXiv:2103.00965, https://arxiv.org/abs/2103.00965
  2. H. C. M. Fernandes, J. J. Arenzon and Y. Levin, 'Monte Carlo simulations of two-dimensional hard core lattice gases', J. Chem. Phys. 126, 114508 (2007). DOI: 10.1063/1.2539141; arXiv:cond-mat/0612372, https://arxiv.org/abs/cond-mat/0612372
  3. X. Feng, H. W. J. Blöte and B. Nienhuis, 'Lattice gas with nearest- and next-to-nearest-neighbor exclusion', Phys. Rev. E 83, 061153 (2011). DOI: 10.1103/PhysRevE.83.061153, https://doi.org/10.1103/PhysRevE.83.061153
  4. K. Ramola, K. Damle and D. Dhar, 'Columnar order and Ashkin-Teller criticality in mixtures of hard-squares and dimers', Phys. Rev. Lett. 114, 190601 (2015). DOI: 10.1103/PhysRevLett.114.190601; arXiv:1408.4943, https://arxiv.org/abs/1408.4943
  5. D. Mandal, T. Nath and R. Rajesh, 'Estimating the critical parameters of the hard square lattice gas model', J. Stat. Mech. (2017) 043201. DOI: 10.1088/1742-5468/aa650e, https://doi.org/10.1088/1742-5468/aa650e
  6. N. T. Rodrigues, J. F. Stilck and T. J. Oliveira, 'Entropy of fully packed rigid rods on generalized Husimi trees: A route to the square-lattice limit', Phys. Rev. E 105, 024132 (2022). DOI: 10.1103/PhysRevE.105.024132, https://doi.org/10.1103/PhysRevE.105.024132
  7. A. A. A. Jaleel, J. E. Thomas, D. Mandal, Sumedha and R. Rajesh, 'Rejection-free cluster Wang-Landau algorithm for hard-core lattice gases', Phys. Rev. E 104, 045310 (2021). DOI: 10.1103/PhysRevE.104.045310; arXiv:2108.01402, https://arxiv.org/abs/2108.01402
  8. A. A. A. Jaleel, D. Mandal, J. E. Thomas and R. Rajesh, 'Freezing phase transition in hard-core lattice gases on the triangular lattice with exclusion up to seventh next-nearest neighbor', Phys. Rev. E 106, 044136 (2022). DOI: 10.1103/PhysRevE.106.044136, https://doi.org/10.1103/PhysRevE.106.044136
  9. S. Nyckees and F. Mila, 'Tensor network investigation of the hard-square model', Phys. Rev. B 106, 174433 (2022). DOI: 10.1103/PhysRevB.106.174433; arXiv:2205.02508, https://arxiv.org/abs/2205.02508
  10. X. Lyu, 'Lattice and PT symmetries in tensor-network renormalization group: Case study of a hard-square lattice gas model', arXiv:2603.25492 (2026), https://arxiv.org/abs/2603.25492