It is shown that mean-field theory fails to give a correct qualitative picture of the thermodynamic behavior of the q-state Potts model when the exchange interaction is anisotropic in spin space. The correct picture is recovered either by introducing a single-particle anisotropy or by taking correlations into account via a Bethe-Peierls approximation. This analysis helps the interpretation of previous renormalization-group results for asymmetric Potts models.
I. INTRODUCfiONThe Potts model 1 of a ferromagnet has been extensively studied 2 either in its original lattice formulation or in the continuum version 3 -6 as a Euclidean f{J 3 field theory. Extensions of the model have also been introduced 7 -9 that allow for anisotropy in spin space. These are used, for example, to describe structural phase transitions in perovskites. 7 • 8 In a recent paper, Barbosa, Gusmão, and Theumann 9 discussed the phase transitions in the continuum version of the q-state Potts model with symmetry breaking, using a form of the renormalization group (RG) suitable for studying the crossover behavior when some components of the order-parameter field remain massive through the transitiono 10 The purpose of this work is to complement the analysis of Ref. 9, where the interpretation of the RG results was based on a mean-field theory (MFT) which, as I shall discuss below, is not appropriate for the anisotropic case. In particular, the MFT predicts the existence of a disordered phase with zero magnetization even when the exchange anisotropy favors only one of the q states against ali the otherso The existence of such a phase is obviously not expected on physical groundso This failure can be explained by the fact that an asymmetry in the exchange interaction manifests itself through correlations, which are neglected in MFTo lndeed, when a single-particle (crystal-field) anisotropy is introduced (Seco 111) it appears as an effective magnetic field in the Landau free energy and, consequently, the disordered phase is not present. I also show that in the case of purely exchange asymmetry the introduction of correlations through a simple Bethe-Peierls approximation 11 (Seco IV) results in the absence of a disordered phase. This precludes a paramagnetic-to-ferromagnetic second-order phase transition, although a first-order transition between a weakly and a strongly magnetized phase is not ruled outo 8 It also explains the absence of a nontrivial fixed point of the RG for this case, 9 since the mean-field minimum at zero-order parameter, around which the perturbation expansion is performed, is no longer a minimum when the asymmetric interaction is taken into accounto A controversial point concerning the continuation of the results for small noninteger q that is important to understand the role of the RG fixed point and criticai exponents is also addressed (Seco V).
THE MODEL AND RG RESULTSThe Hamiltonian of the q-state Potts model in a lattice with only nearest-neighbor interactions is usually 2 written aswhere 8u 1 u 1 is the Kronecker 8 func...