2
$\begingroup$

This is a follow up question to the insightful answer by Theo Johnson-Freyd to the question 105221.

This answer explained that the quantised enveloping algebra, $U_q(\mathfrak{g})$, (defined by a presentation due to Jimbo) could not be defined without a choice, say, of a Cartan subalgebra of $\mathfrak{g}$. The answer also made a comment to the effect that the category of (finite dimensional) representations of $U_q(\mathfrak{g})$ is a deformation quantisation of the category of (finite dimensional) representations of $\mathfrak{g}$. My question is whether this comment can be made precise?

The idea (as I understand it) is that we should put $q=\exp(h)$ and expand in power series over $h$. Then taking the degree zero part in $h$ gives the representation of $\mathfrak{g}$ and the linear part in $h$ gives some additional structure. If we think of a K-linear category as a K-algebra then this additional structure would be a Poisson structure. This would also be a replacement for Drinfeld's Poisson-Lie groups and Lie bialgebras. Then there would be further structure corresponding to the quasi-triangular structure.

Another line of thought is that Chern-Simons theory is canonically associated to $\mathfrak{g}$. Does this lead to a canonical construction of the representation theory of $U_q(\mathfrak{g})$? This would appear to work only for $q$ a root of unity.

My motivation for this question is that I would like to apply this to Vogel's universal Lie algebra where the Lie algebra is an object in a category and so any choice required for the construction of the quantised enveloping algebra would seem to be unavailable.

$\endgroup$
2
$\begingroup$

I haven't thought about this stuff in a while, but here's what I can remember. (Perhaps I should just read my notes for a talk on the subject instead.) Before I begin, let me emphasize that nothing I'm about to say is due to me --- much of it dates from my mathematical infancy.

Let me answer your questions a bit out of order.

Another line of thought is that Chern-Simons theory is canonically associated to $\mathfrak g$. Does this lead to a canonical construction of the representation theory of $U_q(\mathfrak g)$? This would appear to work only for q a root of unity.

My memory is that the category $\mathrm{Rep}^{\mathrm{fd}}(U_q\mathfrak g)$ of finite-dimensional $U_q\mathfrak g$-modules can be canonically associated to any semisimple Lie algebra $\mathfrak g$ without extra data, in a purely algebraic way. I don't remember how to do this, at least not slickly. When $q = e^\hbar$ is in an infinitesimal neighborhood of $1$, there is a method, probably due to Drinfeld, which works both for $\mathrm{Rep}^{\mathrm{fd}}(U_q\mathfrak g)$ and for the category of all representations $\mathrm{Rep}(U_q\mathfrak g)$. More generally, it works for any Lie algebra over a characteristic-$0$ field $\mathbb K$ with a chosen nondegenerate adjoint-invariant symmetric pairing (indeed, only a quadratic Casimir is needed, possibly degenerate). For semisimples, you can take the Killing form.

Here's the coarse outline. Take the category $\mathrm{Rep}_\hbar(U\mathfrak g) = \mathrm{Rep}(U\mathfrak g) \otimes_{\mathbb K} \mathbb K[[\hbar ]]$, which is the Kelly–Deligne tensor product of $\mathrm{Rep}(U\mathfrak g)$ and $\mathrm{Rep}(\mathbb K[[\hbar ]])$, along with its standard monoidal functor $\otimes : \mathrm{Rep}_\hbar(U\mathfrak g) \otimes \mathrm{Rep}_\hbar(U\mathfrak g) \to \mathrm{Rep}_\hbar(U\mathfrak g)$, but not its standard monoidal structure — I want to modify the associator, and also the braiding. The braiding is modified in some standard way, and the associator is modified by some $O(\hbar)$ term depending on a choice of Drinfeld associator. These is the main topic of the notes I linked above, so I won't go into it now.

Anyway, this category determines the Hopf algebra $U_q\mathfrak g$ (again with $q = e^\hbar$) up to isomorphism, but not up to canonical isomorphism. That was what I was trying to explain in the answer you linked to.

You are right that Chern–Simons theory also should give a canonical description of $\mathrm{Rep}(U_q\mathfrak g)$, at least when $q$ is a root of unity. Actually, you should choose a compact group $G$ integrating $\mathfrak g$. Then it's expected, but in general not yet proved, that the semisimplified version of $\mathrm{Rep}(U_q\mathfrak g)$ is described by the fusion rules for positive-energy representations of the loop group of $G$ (at a level determined by the order of $q$). This latter category is the "path integral / analytic / Chern–Simons" quantization of $\mathfrak g$.

The idea (as I understand it) is that we should put $q=\exp(h)$ and expand in power series over $h$. Then taking the degree zero part in h gives the representation of 𝔤 and the linear part in h gives some additional structure. If we think of a K-linear category as a K-algebra then this additional structure would be a Poisson structure. This would also be a replacement for Drinfeld's Poisson-Lie groups and Lie bialgebras. Then there would be further structure corresponding to the quasi-triangular structure.

You can think of the "infinitesimal braiding" determined by the Casimir as being analogous to a Poisson structure. Alternately, you can think of the Casimir itself as analogous to a Poisson structure. This has been made precise in the notion of "shifted symplectic structure" prevalent in the AKSZ literature (see e.g. work by Roytenberg). The category $\mathrm{Rep}(U\mathfrak g)$ "is" the category of quasicoherent sheaves on a "stack" $B \mathfrak g$, the "classifying space" of the formal group determined by $\mathfrak g$. I put a bunch of words in quotes because you should think of formal groups and their classifying spaces as infinitesimal objects (in particular, $B \mathfrak g$ is not a stack in the usual algebrogeometric meaning). The Killing form defines a "shifted symplectic structure" on $B\mathfrak g$, meaning a nondegenerate closed 2-form of some cohomological degree (remember that the tangent bundle to a stack is a bundle of chain complexes, not of vector spaces).

That said, there are some open questions relating general shifted-Poisson structures to structures on the category of quasicoherent sheaves.

The answer also made a comment to the effect that the category of (finite dimensional) representations of $U_q(\mathfrak g)$ is a deformation quantisation of the category of (finite dimensional) representations of $\mathfrak g$. My question is whether this comment can be made precise?

It is probably worth emphasizing, earlier than I am doing it, that for generic $q$ (or $q = e^\hbar$), $\mathrm{Rep}(U_q\mathfrak g)$ and $\mathrm{Rep}(U\mathfrak g)$ are equivalent as categories — the only thing deforming is their braided monoidal structures. Actually, I should be a bit careful: there can be multiply non-isomorphic equivalences, and depending on the context either one has been chosen or it hasn't. Compare the case of deformations of Poisson to associative algebras: we usually assume that the Poisson algebra $A_0$ and the deformation quantization $A_\hbar$ are equal as vector spaces (perhaps up to tensoring with $\mathbb K[[\hbar]]$), and that that equality is part of the data. Changing that equality by something $O(\hbar)$ is a "gauge transformation" in Kontsevich's sense.

On the other hand, extracting the Poisson structure doesn't require knowing the equality $A_\hbar = A_0 \otimes_{\mathbb K} \mathbb K[[\hbar]]$, just that $A_\hbar$ is a flat deformation (i.e. $A_\hbar$ is a sheaf of algebras over the formal disk $\mathrm{Spec}(\mathbb K[[\hbar]])$, and I want to know that as a sheaf of vector spaces it is a vector bundle). Similarly, you can extract the infinitesimal braiding just from $\mathrm{Rep}(U_q\mathfrak g)$. I state this more precisely in the "Proposition" in section 2 of my linked lecture notes.

But, in any case, yes: the point is that you can present (in many ways) the category $\mathrm{Rep}(U_q\mathfrak g)$ as having as its objects and morphisms the objects and morphisms of $\mathrm{Rep}(U\mathfrak g)$ (up to base changing), but with its algebraic structure (composition, associator, braiding, etc.) deformed.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.