mean-field theory of a phase transition is supposed to give an approximate
description of the behaviour of macroscopically large systems. For a macroscopic system, the fully symmetrical state is realized only at sufficiently high temperatures, i.e., above the highest ordering temperature. Upon cooling, interactions cause the system to break one, or several symmetries of the Hamiltonian6. “Breaking a s y m m e t y ” means that there are symmetry operations under which the Hamiltonian is invariant but the state of the system i s not. The symmetry-breaking nature of the state is signalled by non-zero results for certain expectation values which should exactly vanish if the state were symmetrical. In our present case, the symmetry breaking is associated with the appearance of a non-vanishing magnetization (Sj) in a system whose Hamilton operator is spin-rotationally invariant. Rotating all the spins by the same angle, the Hamiltonian is taken into itself, but the vector (Sj)turns into a different direction, i.e., the state has been changed. It may be mentioned that the phenomenon we are discussing is often called spontaneous symmetry breaking. It is “spontaneous” because we did not put into the Hamiltonian any symmetry-breaking term which could have reduced the symmetry of the energy expression to that of the ordered state. There are cases in which even a macroscopic system retains full symmetry at all temperatures, or at least at all non-zero temperatures. We mention two basic possibilities. First, it may seem obvious that a non-interacting system does not order, since it looks like an assembly of independent finite systems. However, if we think a bit more carefully, we realize that “non-interacting” is not quite the same as “independent”: quantum statistics also induces correlations, It turns out that a free Fermi system is indeed non-ordering but a three-dimensional free Bose gas condenses at a finite transition temperature. Second, in low dimensions, thermal and/or quantum fluctuations can pre- vent long-range order even in strongly interacting systems. The breaking of continuous symmetries at finite temperatures is ruled out for one- and two- dimensional systems by the powerful Mermin-Wagner theorem, assuming that the interaction falls off sufficiently rapidly with increasing interparticle distance [269]. The statement covers the isotropic spin models, as well as interacting electron models which do not include spin-orbit coupling or dipole-dipole inter- action and thus cannot describe magnetic anisotropy. The theorem is certainly valid for the Hubbard model which has a zero-range interaction, thus we can be sure that one- and two-dimensional Hubbard models are not magnetically ordered at any finite temperature7. The same holds for the extended Hubbard
‘See [22] for a thorough discussion of the concept of symmetry breaking.
7We have to be very cautious in interpreting this result: the non-existence of magnetic order at finite T does not necessarily mean that there cannot be a magnetic