This is a heavy question, that contains many topics in it that are worthy of their own questions, so I'm not going to give a complete answer. I am relying mainly on this excellent review paper by Nayak, Simon, Stern, Freedman and Das Sarma. The first part can be skipped by anyone already familiar with anyons.
Abelian and non-Abelian anyons
Anyons are emergent quasiparticles in two dimensional systems that have exchange statistics which are neither fermionic nor bosonic. A system that contains anyonic quasiparticles has a ground state that is separated by a gap from the rest of the spectrum. We can move the quasiparticles around adiabatically, and as long as the energy we put in the system is lower than the gap we won't excite it and it will remain in the ground state. This is partly why we say the system is topologically protected by the gap.
The simpler case is when the system contains Abelian anyons, in which case the ground state is non-degenerate (i.e. one dimensional). When two quasiparticles are adiabatically exchanged we know the system cannot leave the ground state, so the only thing that can happen is that the ground state wavefunction is multiplied by a phase $e^{i \theta}$. If these were just fermions or bosons than we would have $\theta=\pi$ or $\theta=0$ respectively, but for anyons $\theta$ can have other values.
The more interesting case is non-Abelian anyons where the ground state is degenerate (so it is in fact a ground space). In this case the exchange of quasiparticles can have a more complicated effect on the ground space than just a phase, most generally such an exchange applies a unitary matrix $U$ on the ground space (the name 'non-Abelian' comes from the fact that these matrices do not in general commute with each other).
The quantum dimension
So we know that the ground space of a system with non-Abelian anyons is degenerate, but what can we say about its dimension? We expect that the more quasiparticles we have in the system, the larger the dimension will be. Indeed it turns out that for $M$ quasiparticles, the dimension of the ground space for large $M$ is roughly $\sim d_a^{M-2}$ where $d_a$ is a number that depends on $a$ - the type of the quasiparticles in the system. This scaling law is reminiscent of the scaling of the dimension of a tensor product of multiple Hilbert spaces of dimension $d_a$, and for this reason $d_a$ is called the quantum dimension of a quasiparticle of type $a$. You can think of it as the asymptotic degeneracy per particle. For Abelian anyons we have a one-dimensional ground space no matter how many quasiparticles are in the system, so for them $d_a=1$.
Although we used the analogy to a tensor product of Hilbert spaces, note that in that case the dimension of each Hilbert space is an integer, while the quantum dimension is in general not an integer. This is an important property of non-Abelian anyons that differentiates them from just a set of particles with local Hilbert spaces - the ground space of non-Abelian anyons is highly nonlocal.
More details on anyons and the quantum dimension can be found in the review paper cited above. The quantum dimension can be generalized to other systems with topological properties, maintaining the same intuitive meaning of asymptotic degeneracy per particle. It is in general very hard to calculate the quantum dimension, and there is only a handful of papers that do (most of them cited in the paper by Kitaev and Preskill that inspired this question).
Relation to entanglement
I can also try and give a handwaving argument for why the quantum dimension would be related to entanglement. First of all, the fact that the entanglement entropy of a bounded region depends only on the length of the boundary $L$ and not on the area of the region is very clearly explained in this paper by Srednicki, which is also cited by Kitaev and Preskill. Basically it says that the entanglement entropy can be calculated by tracing out the bounded region, or by tracing out everything outside the bounded region, and the two approaches will yield the same result. This means the entanglement has to depend only on features that both regions have in common, and this rules out the area of the regions and leaves only the boundary between them.
Now for a system with no topological order the entanglement would go to zero when the size of bounded region goes to zero. However for a topological system there is intrinsic entanglement in the ground space which yields the constant term $-\gamma$ in the entanglement. The maximal entanglement entropy a system with dimension $D$ has with its environment is $\log D$, so in an analogous manner the topological entanglement is $\gamma=\log D$ where $D$ is the quantum dimension. Again this last argument relies heavily on handwaving so if anyone can improve it please do.
I hope this answers at least the main concerns in the question, and I welcome any criticism.
This post imported from StackExchange Physics at 2014-04-05 03:59 (UCT), posted by SE-user Joe
Q&A (4871)
