[challenge]: Does the standard 2D RP² model have a finite-temperature transition? An adversarial test of Z₂-vortex unbinding versus pseudocritical crossover

Released by

Youjin Deng

Contact email

yjdeng@ustc.edu.cn

Method

Classical cluster Monte Carlo / replica exchange or density-of-states Monte Carlo; tensor-network transfer matrix (stretch cross-check)

Challenge issue

This is a five-day adversarial test of a long-standing controversy. The challenge does not assume that the standard two-dimensional (RP^2) model has a finite-temperature phase transition. A reproducible negative result, a quantitative finite-size exclusion, and a statistically justified “inconclusive” result are all successful outcomes.

The dispute

The standard ferromagnetic (RP^2), or Lebwohl-Lasher, model has a three-component unit director on every site of a two-dimensional square lattice,

$$
\mathbf n_i\in S^2,\qquad \mathbf n_i\equiv-\mathbf n_i ,
$$

and a nearest-neighbor interaction

$$
H_{\rm LL}
=-J\sum_{\langle ij\rangle}P_2(\mathbf n_i!\cdot!\mathbf n_j),
\qquad
P_2(x)=\frac{3x^2-1}{2}.
$$

Its order-parameter space has

$$
\pi_1(RP^2)=\mathbb Z_2,
$$

so stable point defects exist. Their existence alone, however, does not establish a thermodynamic transition.

Four interpretations of the finite-size data remain to be distinguished:

1. Finite-(T), BKT-like topological transition. Early cluster Monte Carlo and conformal finite-size studies interpreted the anomaly near (T/J\sim0.5) as a defect-driven transition with an essential singularity and a low-temperature critical phase.
2. Finite-(T), non-BKT singularity. Density-of-states work reported two temperature scales without nematic quasi-long-range order. A recent preprint further proposes a finite-temperature third-order transition near (T_n/J\simeq0.585), followed by (Z_2)-defect pairing at a lower scale.
3. No finite-(T) transition. Probability-distribution, stiffness, tensor-renormalization, and symmetric-MPS studies instead support a sharp crossover into a pseudocritical regime with a very large but finite correlation length; the only true critical point is then at (T=0).
4. Weak first-order behavior. This is not the leading interpretation for the standard model, but it must be tested rather than excluded by assumption. First-order transitions known in modified, singular, gauged, or generalized (RP^2) models are not evidence for this hypothesis.

The challenge takes no one of these outcomes as ground truth. In particular:

• a peak or cusp at one lattice size is not a thermodynamic transition;
• (Z_2)-defect pairing, proliferation, or a density anomaly is not by itself a free-energy singularity;
• an apparently excellent BKT fit over correlation lengths below (O(10^2)) is not sufficient if it softens at larger scales;
• absence of conventional nematic long-range order does not by itself rule out a topological or other symmetry-preserving transition.

Model boundary and pinned conventions

Use an (L\times L) square torus with periodic boundary conditions, nearest-neighbor bonds only, (J=1), and (k_B=1):

$$
H=-\sum_{\langle ij\rangle}P_2(\mathbf n_i!\cdot!\mathbf n_j).
$$

The following algebraic equivalence must be tested in code:

$$
H=-\frac32\sum_{\langle ij\rangle}(\mathbf n_i!\cdot!\mathbf n_j)^2+\frac12N_b ,
$$

where (N_b=2L^2). Consequently, a paper using

$$
H_{\rm sq}=-\sum_{\langle ij\rangle}(\mathbf n_i!\cdot!\mathbf n_j)^2
$$

uses the reduced-temperature conversion

$$
T_{\rm LL}=\frac32T_{\rm sq}.
$$

Every quoted temperature must state which convention is used. Bare (P_2) and squared-interaction temperatures must never be mixed.

Out of scope for the core verdict:

• antiferromagnetic (J<0);
• a local gauge field or plaquette coupling;
• vector-isotensor mixed actions;
• defect-suppression or constraint terms;
• (P_4), high-power, singular-well, long-range, or off-lattice interactions;
• confined films, active nematics, or triangular-lattice Heisenberg antiferromagnets.

Results for such variants may be reported only as explicitly labeled context or stretch controls.

Gauge-invariant observables

Define the local traceless nematic tensor

$$
Q_i^{ab}=n_i^a n_i^b-\frac{\delta^{ab}}3,
\qquad
\overline Q=\frac1N\sum_iQ_i .
$$

The pinned nematic magnitude and Binder-like ratio are

$$
m_Q^2=\operatorname{Tr}\overline Q^2,
\qquad
R_Q=\frac{\langle m_Q^2\rangle^2}{\langle(m_Q^2)^2\rangle}.
$$

Define the tensor structure factor

$$
S_Q(\mathbf k)=\frac1N
\left\langle
\operatorname{Tr}!\left[Q(\mathbf k)Q(-\mathbf k)\right]
\right\rangle
$$

and the second-moment correlation length, averaged over the two lattice axes,

$$
\xi_Q=
\frac{1}{2\sin(\pi/L)}
\sqrt{\frac{S_Q(\mathbf 0)}{S_Q(\mathbf k_{\min})}-1},
\qquad
\mathbf k_{\min}\in{(2\pi/L,0),(0,2\pi/L)}.
$$

To match the defect construction used in the density-of-states literature, split every square along the same registered diagonal into a triangulation (\mathcal T). For each triangle (\triangle), define

$$
w_\triangle=\prod_{(ij)\in\partial\triangle}
\operatorname{sgn}(\mathbf n_i!\cdot!\mathbf n_j),
\qquad
v_\triangle=\frac{1-w_\triangle}{2},
\qquad
\rho_v=\frac1{2L^2}\sum_{\triangle\in\mathcal T}
\langle v_\triangle\rangle .
$$

Also report the net square-plaquette parity

$$
w_p=\prod_{(ij)\in\partial p}
\operatorname{sgn}(\mathbf n_i!\cdot!\mathbf n_j),
$$

which equals the product of its two triangular fluxes. Repeat the density analysis with the opposite diagonal convention as a lattice-discretization check. If a claimed topological temperature moves by more than its combined uncertainty, the topological verdict is inconclusive. Configurations for which a dot product is numerically indistinguishable from zero must follow a registered tie rule, and the fraction of such events must be reported.

For noncontractible loops, measure the row- and column-resolved parities

$$
W_x(y)=\prod_x\operatorname{sgn}(\mathbf n_{x,y}!\cdot!\mathbf n_{x+1,y}),
\qquad
W_y(x)=\prod_y\operatorname{sgn}(\mathbf n_{x,y}!\cdot!\mathbf n_{x,y+1}).
$$

In the presence of plaquette defects these parities may differ between parallel loops. Report their full distributions; assign one global winding sector only on configurations for which that assignment is path independent.

The primary thermodynamic observables are:

• (\xi_Q/L);
• (R_Q);
• the nematic correlation ratio (G_Q(L/2)/G_Q(L/4));
• the rotational stiffness (\Upsilon), defined as the second derivative of the free energy under a pinned (SO(3)) boundary twist;
• energy density, specific heat per site (c_V), energy cumulants through third order, and the energy histogram (P_L(e)).

The primary topological observables are:

• (\rho_v) and (d\rho_v/dT);
• row- and column-resolved noncontractible (Z_2) loop-parity distributions;
• one preregistered defect-binding diagnostic.

To claim “vortex binding” or “unbinding,” a team must measure at least one free-energy-sensitive diagnostic, for example:

• a constrained two-defect separation free energy (F_2(r)); or
• a (Z_2)-twisted-sector free-energy cost
  (\Delta F_{Z_2}=-T\log(Z_{\rm twist}/Z_{\rm periodic})).

A defect-density curve alone is insufficient for that claim.

Fairness and five-day contract

1. Day-0 scaffold is mandatory. Organizers must provide a tested sampler for the pinned Hamiltonian, input conventions, local-sign-flip tests, and small-size fixtures. A Wolff-type embedded-cluster implementation is preferred. Replica exchange, multicanonical, or Wang-Landau sampling may be added, but their equilibrium distributions must be cross-checked at small (L).
2. Raw-data escrow. Before thermodynamic extrapolation, deposit per-temperature/per-size time series or sufficient blocked statistics, random seeds, replica round-trip information where applicable, integrated autocorrelation estimates, effective sample sizes, code revision, compiler flags, and hardware metadata.
3. Locked analysis. By the end of day 3, register temperature windows, size exclusions, reweighting range, all correction ansätze, the definition of each pseudocritical temperature, and model-selection criteria. Later changes must be labeled exploratory.
4. Blind adjudication. A helpdesk-assigned adjudicator with no stake in any of the proposed scenarios receives anonymized dimensionless-ratio, cumulant, and defect curves and applies the registered rules.
5. No one-observable or one-fit verdict. A conclusion that changes after dropping the smallest size, narrowing the temperature window, changing the registered correction ansatz, or replacing one pseudocritical estimator is reported as inconclusive.

The release team must state any scientific conflict of interest. No releaser-side prediction or unpublished curve is used as hidden ground truth.

Validation

Exact and implementation anchors

• Local (Z_2) invariance: independently flip (\mathbf n_i\to-\mathbf n_i) at randomly selected sites. Energy, all (Q)-tensor observables, plaquette fluxes, and winding observables must remain unchanged to machine precision.

• Hamiltonian conversion: the (P_2) and squared-interaction implementations must agree configuration by configuration after applying the exact additive constant and (3/2) coupling conversion above.

• Zero-temperature anchor: the aligned/director-degenerate ground-state energy is

  $$
  e_0=H/L^2=-2 .
  $$

• Infinite-temperature anchor: (\langle P_2(\mathbf n_i!\cdot!\mathbf n_j)\rangle=0), hence (e\to0), and

  $$
  c_V=\frac{2}{5T^2}+O(T^{-3})
  $$

  in the (P_2,J=1) convention.

• For (L\le 8), compare cluster and local-update/replica-exchange implementations for energy, (R_Q), (S_Q(\mathbf k)), and (\rho_v) using statistically compatible confidence intervals.

Positive and negative analysis controls

• BKT positive control: with the same analysis code, reproduce the square-lattice XY transition (T_{\rm BKT}=0.89294(8)) for (H=-\sum\cos(\theta_i-\theta_j)), including the drift expected from logarithmic corrections. This validates that the pipeline can recognize a genuine BKT transition.
• Pseudocritical negative control: run the two-dimensional nearest-neighbor (O(3)) Heisenberg model on at least (L=32,64,128). The analysis must not turn its large low-temperature correlation length into a stable finite-(T) transition.

Published (RP^2) controls

• Reproduce the historical apparent BKT window near (T_{\rm LL}\simeq0.51). A restricted fit over (\xi_Q\lesssim10^2) should be able to recover the previously reported BKT-like scale before the full-size analysis is unblinded.
• In the squared-interaction convention, Burgelman et al. obtained an apparent fit (T_c=0.339(1)), equivalent to (T_{\rm LL}=0.5085(15)), followed by visible softening beyond (\xi\sim10^2). Reproducing the fitted scale is a validation target; reproducing or refuting the softening is part of the disputed measurement.
• At (L=128), explicitly test the recent density-of-states claims (T_{\rm BKT}\simeq0.413), (T_n\simeq0.585), and (T_p\simeq0.590) in the pinned (P_2) convention. These are disputed targets, not accepted critical temperatures.

Failure to reproduce a disputed target is scientifically meaningful only after the exact Hamiltonian, temperature conversion, equilibration, autocorrelation, and observable definitions have passed the anchors above.

Core Monte Carlo measurement

Perform a coarse scan over

$$
0.35\le T\le0.70
$$

and a dense scan or histogram-reweighted analysis over

$$
0.40\le T\le0.62 .
$$

The required size ladder is

$$
L=32,64,128,256,512.
$$

At least four consecutive sizes must pass the registered effective-sample-size and round-trip criteria. (L=1024) and (2048) are stretch targets after wall-clock and autocorrelation audits.

For every retained size:

• measure all pinned primary thermodynamic observables;
• measure (\rho_v), its temperature derivative, and the noncontractible loop-parity distributions;
• save energy and (m_Q^2) histograms in a format that permits independent reweighting;
• report integrated autocorrelation times and effective sample counts separately for energy, (m_Q^2), and (\rho_v);
• repeat at least three temperatures near each detected anomaly with independent seeds or independently initialized replica ladders.

Pseudocritical temperatures must be extracted independently from:

• extrema or crossings of (\xi_Q/L), (R_Q), and the correlation ratio;
• the (c_V) and third-energy-cumulant features;
• the stiffness drift;
• (\rho_v), loop-parity, or defect-binding features.

Do not force these temperatures to coincide. Their difference and its (L)-dependence are primary results.

Adjudication

The following are competing models, not a menu from which to select after viewing the answer.

Hypothesis A: finite-(T) BKT-like transition

Fit the disordered-side correlation length to

$$
\xi_Q(T)=A\exp!\left[
\frac{b}{(T-T_c)^{1/2}}
\right]
\times{\text{registered corrections}}.
$$

A BKT-like verdict requires all of:

• a stable (T_c>0) under the registered fit-window and (L_{\min}) variations;
• consistent logarithmic finite-size drifts in at least two dimensionless observables;
• algebraic nematic correlations or a nonvanishing thermodynamic stiffness below the same (T_c);
• a free-energy-sensitive defect-binding diagnostic consistent with the same scale.

No universal stiffness-jump value is assumed for (RP^2) unless it is derived for the pinned normalization before analysis.

Hypothesis B: finite-(T), non-BKT continuous or higher-order transition

Test a common nonzero limiting temperature across the dimensionless ratios and energy derivatives. A third-order claim additionally requires:

• no latent heat and no interface barrier growing as (L);
• a registered finite-size scaling form for the third free-energy derivative or equivalent energy cumulant;
• a sharpening nonanalytic signature whose location and amplitude extrapolate consistently;
• independent evidence that the feature is not merely a smooth crossover of a finite correlation length.

A cusp in a spline derivative at one (L) does not pass this gate.

Hypothesis C: weak first-order transition

A first-order verdict requires mutually consistent evidence from:

• an energy-density separation that extrapolates to nonzero latent heat;
• a bimodal energy histogram with an interface barrier growing proportionally to (L);
• (c_{V,\max}) scaling asymptotically as (L^2).

If these signals are absent, report the confidence-dependent upper bound on latent heat that the data exclude.

Hypothesis D: finite-(T) crossover with a (T=0) critical point

Test finite thermodynamic (\xi_Q(T)) at every sampled (T>0), vanishing stiffness with increasing (L), drifting or disappearing dimensionless-ratio crossings, and an analytic free energy. Compare the low-temperature growth to both:

$$
\xi_Q\sim \exp(a/T)
$$

with registered corrections and a finite-(T) essential singularity.

A crossover verdict must quantify the accessible crossover temperature and length scale. It must not be stated as a proof of analyticity for all (T>0).

Model-selection rules

• Fit the registered correction variants for every hypothesis and report AIC/BIC or another locked predictive score.
• Repeat after dropping (L=32), and again after dropping the smallest retained temperature or the edge of the fit window.
• Report goodness-of-fit, residual structure, parameter covariance, and bootstrap coverage; do not rank models by (R^2).
• Require agreement between at least two thermodynamic observables before claiming a thermodynamic transition.
• Treat topological and thermodynamic temperatures as separate parameters until the data support equality.
• If two scenarios remain statistically indistinguishable, or the verdict changes under a registered robustness test, the correct result is inconclusive at accessible scales.

Tensor-network / transfer-matrix stretch cross-check

This route is not required for the five-day core.

Construct the classical partition-function tensor using a spherical-harmonic expansion that preserves the even-(\ell) content and explicit (SO(3)) symmetry. Report convergence in:

• angular-momentum cutoff (\ell_{\max});
• boundary-MPS or environment bond dimension;
• free-energy density and energy;
• transfer-matrix correlation length;
• finite-entanglement scaling and the effective central charge.

The cross-check should determine whether the BKT-like correlation-length growth softens beyond (\xi\sim10^2) and whether the apparent low-temperature scaling is a true CFT regime or a finite-entanglement pseudocritical window. Local observables alone are not sufficient for this judgment.

Five-day floor

If the Day-0 cluster sampler or fixtures are unavailable, or if they fail validation and cannot be repaired in time, the accepted deliverable stops at:

• the pinned-model implementation and exact invariance tests;
• XY positive-control and (O(3)) negative-control analyses;
• two statistically cross-checked update schemes through (L=64);
• a documented finite-size drift study of the standard (RP^2) model through at least (L=128);
• released raw data, code, and a clear failure report.

No thermodynamic verdict is expected at this floor.

What is reproduction and what is new

Reproduction milestones:

• the apparent BKT scale near (T_{\rm LL}\sim0.51);
• the qualitative (Z_2)-defect anomaly;
• the (L=128) density-of-states markers claimed in 2018 and 2026;
• the initial BKT-like growth and its reported softening around (\xi\sim10^2);
• the absence of ordinary nematic long-range order at finite (T).

A publishable edge requires at least one of:

• an independently implemented, uncertainty-controlled multi-size confirmation or falsification of the proposed third-order singularity;
• a joint finite-size estimate showing whether the thermodynamic and (Z_2)-binding scales coincide or remain separated;
• stable evidence for or against a finite-(T) BKT-like phase at correlation lengths beyond previous Monte Carlo fits;
• a quantitative exclusion region for (T_c), latent heat, or a singular free-energy derivative;
• a rigorous demonstration that five-day-accessible sizes cannot distinguish the hypotheses, including the minimum length scale or precision required to do so.

A negative or inconclusive result is citable only with raw data, seeds, autocorrelation metadata, code, and the locked analysis record.

References

Key primary sources:

• P. A. Lebwohl and G. Lasher, “Nematic-liquid-crystal order—a Monte Carlo calculation,” Phys. Rev. A 6, 426.
• H. Kunz and G. Zumbach, “Numerical evidence for a topological phase transition in a two-dimensional spin model,” Phys. Lett. B 257, 299.
• H. Kunz and G. Zumbach, “Topological phase transition in a two-dimensional nematic (n)-vector model,” Phys. Rev. B 46, 662.
• A. I. Fariñas-Sánchez, R. Paredes, and B. Berche, “Evidence for a topological transition in nematic-to-isotropic phase transition in two dimensions,” arXiv:cond-mat/0301065.
• R. Paredes, A. I. Fariñas-Sánchez, and R. Botet, “No quasi-long-range order in the two-dimensional liquid crystal,” arXiv:0808.3071.
• A. I. Fariñas-Sánchez et al., “On the critical behavior of two-dimensional liquid crystals,” arXiv:0906.4079.
• Y. Tomita, “Finite-size scaling analysis of pseudocritical region in two-dimensional continuous-spin systems,” Phys. Rev. E 90, 032109.
• B. K. Latha and V. S. S. Sastry, “Two phase transitions in the two-dimensional nematic three-vector model with no quasi-long-range order,” Phys. Rev. Lett. 121, 217801.
• G. Delfino, Y. Diouane, and N. Lamsen, “Absence of nematic quasi-long-range order in two-dimensional liquid crystals with three director components,” arXiv:2005.06307.
• C. Bonati et al., “Asymptotic low-temperature behavior of two-dimensional (RP^{N-1}) models,” arXiv:2006.13061.
• A. Ueda and M. Oshikawa, “Tensor Network Renormalization Study on the Crossover in Classical Heisenberg and (RP^2) Models in Two Dimensions,” arXiv:2202.07042.
• L. Burgelman et al., “Contrasting pseudocriticality in the classical two-dimensional Heisenberg and (RP^2) models,” arXiv:2202.07597.
• B. Kamala Latha, V. S. S. Sastry, and S. R. Shenoy, “Entropic bottlenecks to nematic ordering in an (RP^2) apolar spin model,” arXiv:2503.12970v2.

Algorithm and control references:

• U. Wolff, “Collective Monte Carlo Updating for Spin Systems,” Phys. Rev. Lett. 62, 361.
• F. Wang and D. P. Landau, “Efficient, Multiple-Range Random Walk Algorithm to Calculate the Density of States,” Phys. Rev. Lett. 86, 2050.
• M. Hasenbusch, “The two-dimensional XY model at the transition temperature,” arXiv:cond-mat/0502556.

No repository files, tests, or code entry points are named; begin with the mandatory day-0 sampler scaffold and verify the pinned Hamiltonian using the listed local-invariance, conversion, and small-size anchors. Done means completing the registered Monte Carlo measurement and validation plan, depositing the required raw statistics, and reporting a statistically justified verdict.