6

According to Landau's symmetry breaking theory, there is a symmetry breaking when phase transition occurs.

  1. What is the symmetry breaking of superfluid-Mott insulator transition in Bose-Hubbard model?

  2. Why metallic state to Mott insulator state transition in Fermi-Hubbard model is not a phase transition, but a crossover.

Qmechanic
  • 201,751
Timothy
  • 2,469
  • 4
  • 25
  • 37

1 Answers1

5

The Mott transition in the Bose-Hubbard model is a quantum phase transition. From the point of view of field theory, that does not change much compare to standard (finite-temperature) phase transitions. The main difference is that you now have to take into account the quantum fluctuations which correspond to the "imaginary time" direction in addition to the d dimensions of space. It also means that there are at least two control parameters (i.e. parameters that have to be fine-tuned to have the transition), a non-thermal control parameter (such as the hopping amplitude or the density) and the temperature (which must be zero by definition).

Other than that, you can use Landau theory to understand the transition (which is second-order) at zero-temperature. The disordered phase is the Mott insulator, and the ordered one is the superfluid, where the non-zero order parameter is the condensate density (I will only talk about the 3D case, which is the simplest, as I won't have to deal with BKT phases). The broken symmetry is the usual one for Bose-Einstein condensate : the U(1) symmetry. One can then show that there is two universality classes, depending on the way the transition is made (at constant density or with a change of density at the transition).

Now, at finite temperature, things are different. First, the Mott insulator does not exist anymore, as a finite temperature can excite particles and one gets a finite compressibility (or conductivity). That might correspond to the cross-over you're talking about in the fermionic case. On the other hand, the superfluid exists at least up to a critical temperature.

Adam
  • 11,843
  • thanks a lot. The superfulid oder parameter $\left<b^{\dagger}\right>$ breaks the U(1) symmetry:$b\rightarrow b e^{i\phi}$, when MI-SF transition happens. But I still cannot understand why there is no phase transition in Fermi Hubbard model? What is the essential difference between cross-over and phase transition? – Timothy Sep 23 '13 at 20:52
  • Previous answer by @Adam is mostly correct except for the statement that there are no Mott insulators at finite temperatures. See for example: http://prl.aps.org/abstract/PRL/v99/i12/e120405. Indeed optical lattice experiments have to conduct measurements at low enough temperatures that the gapped quasiparticles are well defined. And in a cross-over, there are no divergences associated with thermodynamic quantities or their derivatives. – MaviPranav Sep 25 '13 at 13:52
  • @MaviPranav : Strictly speaking, there is no Mott insulator at finite temperature, because the conductivity is non-zero. Of course, at very low temperature (much smaller than the gap), the system looks like an insulator (exponentially small compressibility), but you won't find a real transition, which exists only at $T=0$. Of course, experimentally one is happy with an exponentially vanishing compressibility (in the same way that insulators "experimentally exists" at finite temperature in solids) but I don't see how what I said in my answer is false. – Adam Oct 01 '13 at 03:57
  • @Jeremy : I know less about the fermionic case, but I think that in that case, the order parameter is less easy to define. For instance, you can say that the conductivity is the order parameter (zero in the insulator, finite otherwise), but there is no real symmetry breaking (but as I said, I might be wrong on that). Nevertheless, you don't need a symmetry breaking to get a phase transition. It's just that when there is a symmetry, it makes everything much easier. – Adam Oct 01 '13 at 04:03
  • @MaviPranav : it might be a matter of taste, but I think people usually describe a phase by its thermodynamical properties, and the hallmark of the Mott phase is its vanishing compressibility, hence the $T=0$. Just to say that I don't think my answer was wrong, as you suggested in your first comment. – Adam Oct 05 '13 at 19:59
  • @Adam: As remarked earlier, it is the gapped nature of doublon/holon quasiparticles in the Mott phase that I consider one its defining hallmarks (compressible matter implies gapless excitations as you mentioned); see for e.g. Figure 2(a) in http://www.nature.com/nature/journal/v487/n7408/full/nature11255.html. In the 'strict' sense, you are correct clearly, which I didn't know you were going for in your first reply ... Because well, on a slightly philosophical note, if not experimentally I'm not sure what else the quality of something's 'existence' can possibly signify. – MaviPranav Oct 05 '13 at 20:17
  • @MaviPranav: I was just trying to show that my answer wasn't wrong in itself as I thought you were arguing. I'm not against discussing what should be the hallmarks, but I think I was giving the mainstream definition. And I agree that this definition is really theoretical and that measuring the gap and an almost vanishing compressibility is what experimentalists should look for. On the other hand, we, as physicists, always use a modelized description of Nature (what about the thermodynamic limit, or the existence of real phase transition and discontinuities ?). – Adam Oct 05 '13 at 20:27
  • Just one last comment: seeing a gap in the one-particle excitation spectrum in not sufficient to have an insulator, you need the system to be fully gapped, which is very difficult to show experimentally. It is then easier to see a vanishing compressibility. But because we understand so well the Bose-Hubbard model and the cold atoms experiments, having a gap in the single particle spectrum and a compressibility vanishing in the error bars are sufficient proofs to say that we have observed a Mott insulator. – Adam Oct 05 '13 at 20:40