Discrete non-Abelian groups and asymptotically free models

Sergio Caracciolo, Andrea Montanari, Andrea Pelissetto

Abstract We consider a two-dimensional $\sigma$-model with discrete icosahedral/dodecahedral symmetry. Using the perturbative renormalization group, we argue that this model has a different continuum limit with respect to the $O(3)$ $\sigma$ model. Such an argument is confirmed by a high-precision numerical simulation.

Introduction

In the study of two-dimensional quantum field theories and statistical mechanics, the $O(3)$ $\sigma$-model is a fundamental pillar, widely believed to be asymptotically free. This property implies that interactions weaken at shorter distances, a behavior analogous to quarks in Quantum Chromodynamics (QCD). However, a long-standing controversy arose regarding models where spins are restricted to the vertices of Platonic solids, such as the icosahedron. Early numerical data suggested these discrete models behaved almost identically to the continuous $O(3)$ model. This similarity presented a theoretical challenge: discrete models are known to possess a phase transition at a finite temperature ($\beta$), whereas asymptotically free models should not. If both models shared the same continuum limit, it would imply the $O(3)$ model is not asymptotically free, contradicting decades of theoretical physics.

Theoretical Framework: Renormalization Group Analysis

The authors address this conflict by analyzing a Hamiltonian that interpolates between the continuous $O(3)$ model and a discrete-symmetry model. By treating the discrete symmetry as a perturbation, they employ Renormalization Group (RG) equations to determine the "relevance" of the perturbation.

The theoretical analysis demonstrates that any discrete-symmetry perturbation of the $O(3)$ model is mathematically relevant, with an RG eigenvalue of 2. In RG terms, a relevant perturbation grows as the system is observed at larger scales, eventually driving the system toward a different fixed point. This implies that the two models must belong to different universality classes, regardless of their apparent similarity at small scales.

Numerical Results and Finite-Size Scaling

To verify the RG predictions, high-precision Monte Carlo simulations were performed on the icosahedral model. The authors focused on the "step function"—the ratio of an observable measured at scale $L$ versus scale $2L$. If two models belong to the same universality class, their step functions should converge to the same universal curve in the continuum limit.

Finite-size scaling of the susceptibility ratio $\chi_1(2L)/\chi_1(L)$ as a function of $\xi(L)/L$. The data points for larger lattice sizes systematically deviate from the continuous $O(3)$ curve.

The numerical evidence confirms the theoretical divergence. While the second-moment correlation length showed only minor discrepancies, the spin-n susceptibilities ($\chi_n$) provided definitive evidence. As shown in the finite-size scaling (FSS) curves, for $\xi(L)/L \gtrsim 1$, the icosahedral data points systematically "peel away" from the $O(3)$ curve. This discrepancy increases with the lattice size $L$, confirming it as a physical effect of the continuum limit rather than a lattice artifact.

The higher-order susceptibility ratio $\chi_4(2L)/\chi_4(L)$ exhibits an even more pronounced divergence from the $O(3)$ prediction at larger scales.

Conclusion

The findings resolve the debate by proving that the icosahedral and $O(3)$ models have distinct continuum limits. The reason previous studies were misled is that the differences only become visible at extremely large correlation lengths ($\xi_\infty \gtrsim 10^5$). By establishing this distinction, the authors show that the existence of a phase transition in discrete models does not contradict the asymptotic freedom of the $O(3)$ model, thereby preserving the standard theoretical scenario.

Summary updated

Abstract

We consider a two-dimensional $σ$-model with discrete icosahedral/dodecahedral symmetry. Using the perturbative renormalization group, we argue that this model has a different continuum limit with respect to the O(3) $σ$ model. Such an argument is confirmed by a high-precision numerical simulation.
BibTeX
@inproceedings{caracciolo2001discrete,
  author = {Sergio Caracciolo and Andrea Montanari and Andrea Pelissetto},
  title = {{Discrete non-Abelian groups and asymptotically free models}},
  booktitle = {Proceedings of International Europhysics Conference on High Energy Physics — PoS(hep2001)},
  year = {2001},
  doi = {10.22323/1.007.0230},
}
Figure