Remarks on 2-Reps
Posted by Urs Schreiber
Here is a first refinement of some ideas related to the representation theory of the -2-group which I mentioned recently ().
Actually, I won’t talk about here at all, but instead try to make some general facts about 2-reps explicit. I don’t claim that the following is particularly deep. In fact, the considerations are quite elementary. The motivation for spelling this out is that by slightly enriching the following discussion (in particular by replacing vector spaces by Hilbert spaces) this should describe 2-reps of the -2-group in terms of bimodules of vonNeumann algebras.
Claim: Every faithful representation
of any group on a vector space (over some field) gives rise in a canonical way to a 2-representation
of the automorphism 2-group of on a 2-vector space (a module category over ).
This representation factors through the 2-category of bimodules.
The simple idea is this. Last time () I tried to indicate how every faithful representation of on the vector space gives rise to a 2-representation of on by
- sending an object of (an automorphism of ) to the autofunctor
- sending a morphism of to the natural isomorphism given by these naturality squares:
But we may want a representation on a proper 2-vector space. By this I shall mean, somewhat loosely, any module category over some monoidal category .
Here we can choose . There is a close relation between the 2-category of module categories and the 2-category of bimodules of algebras in (). So let’s first construct a representation in terms of bimodules from the above one.
This is obvious. We let be the algebra generated from the operators in – nothing but the category algebra of
(In the more general case where we have representations on infinite dimensional Hilbert spaces we’d take the double commutant of and obtain a vonNeumann algebra .)
The above autofunctors sending to extend to automorphisms
of this algebra. By a standard construction, from every morphism of algebras we obtain an - bimodule
which, as a vector space, is simply itself, where the right action by is simply the product in and where the left action by is obtained by first sending to using and then acting from the left on by multiplication.
Hence for any two objects and of we obtain two - bimodules and .
Now, one can easily check that for every morphism in we get a homomorphism of --bimodules
by multiplying from the left with . The condition for this to preserve the left action is precisely the commutativity of the above naturality squares. (The right action is preserved trivially.)
One also checks that horizontal and vertical composition of bimodule homomorhisms reproduces the horizontal and vertical composition in and .
Finally, we can regard homomorphisms of bimodules as 2-morphisms in by a standard construction ().
Spelled out, the 2-representation
obtained this way from works as follows.
The 2-group is represented on the 2-vector , which is the category of right modules over the algebra .
Every object is sent to the -linear functor
Every morphism \in \mathrm{Mor}(\mathrm{Aut}(H)) is sent to the obvious natural isomorphism of these functors. (See the diagram at the end of these notes.)