By a theorem of Drinfeld, there is a one to one correspondence between Lie bialgebras and Poisson Lie groups. Therefore given a Lie cobracket $\delta: g \to \Lambda^2 g$, there is a Poisson bracket on $C[G]$ corresponding to $\delta$, where $G$ is a Lie group and $g$ the Lie algebra of $G$. I would like to figure out how to compute the Poisson bracket on $C[G]$ using $\delta$ explicitly.

We can recover the cobracket on $g$ using the Poisson bracket on a Poisson Lie group $G$ easily. For example, consider $G = SL_2$. There is a Poisson bracket on $C[SL_2]$: \begin{align} & \{x_{11}, x_{12}\} = -1/2 x_{11} x_{12}, \\ & \{x_{11}, x_{21}\} = -1/2 x_{11} x_{21}, \\ & \{x_{11}, x_{22}\} = - x_{12} x_{21}, \\ & \{x_{12}, x_{21}\} = 0, \\ & \{x_{12}, x_{22}\} = -1/2 x_{12} x_{22}, \\ & \{x_{21}, x_{22}\} = -1/2 x_{21} x_{22}. \end{align} The corresponding cobracket $\delta$ on $sl_2 = \text{Span}(e,f,h)$ is \begin{align} & \delta(e) = -1/2 h \wedge e, \\ & \delta(f) = -1/2 h \wedge f, \\ & \delta(h) = 0. \end{align} The procedure is as follows (I learned this from Theo Johnson-Freyd). Choose local coordinates on $G$ centered at the identity. For example, choose: \begin{align} y_{11} & = x_{11} - 1, \\ y_{12} & = x_{12}, \\ y_{21} & = x_{21}, \\ y_{22} & = x_{22}-1. \end{align} Then \begin{align} & \{ y_{11}, y_{12} \} = \{ x_{11} - 1, x_{12} \} = \{ x_{11}, x_{12}\} = -1/2 x_{11} x_{12} = -1/2 (y_{11} + 1) y_{12} = -1/2 y_{12} - 1/2 y_{11}y_{12}, \\ & \{ y_{11}, y_{21} \} = -1/2 y_{21} + (\text{quadratic or higher}), \\ & \{ y_{12}, y_{21} \} = \text{quadratic or higher}. \end{align} We obtain the Lie bracket on $g^*$ if we take only linear terms. Dualizing the Lie bracket $\Lambda^2 g^* \to g^*$ gives the Lie cobracket on $g$.

Now I try to recover the Poisson bracket on $C[G]$ using a given cobracket $\delta: g \to \Lambda^2 g$. For example, let $G = SL_2$ and suppose that \delta(e) = -1/2 h \wedge e, \delta(f) = -1/2 h \wedge f, \delta(h) = 0. How to recover the corresponding Poisson bracket on C[SL_2]? I found that there is a formula in the book Introduction to quantum groups by Prof. Etingof and Prof. Schiffmann: $$p(e^a) = \sum_{n \geq 0} \frac{(- ad(a))^n}{(n+1)!} \delta(a), \ a \in g.$$ But it seems that it is not easy to use this formula to compute {x_{11}, x_{12}} = -1/2 x_{11} x_{12}?

Any help will be greatly appreciated!


You have the Poisson bivector explicitly at every point $exp(a)$ and you know that it is multiplicative. When $\delta=dr$ was exact you wrote it like $exp(a).r-r.exp(a)$.

You have some functions and you need to decompose their differentials in terms of the basis of one forms that you wrote the bivector in. That way when you write $\pi ( df \wedge dg)$ you will just be reading off some coefficients. If you knew $f=a_{j_1}a_{j_2}\cdots$,$g=a_{i_1} \cdots$ were factorizations where the Poisson bivector was written with a sum of $\frac{\partial}{\partial a_k} \wedge \frac{\partial}{\partial a_l}$ kinds of terms, you would have written them in a dual pair of bases so the computation would be easier. See Gekhtman Shapiro Vainshtein.


Not an answer, but too long for a comment.

The Poisson bracket on the group can be written in $r$-matrix notations:

$$ \{ M \frac{ \otimes}{,} M \} = [ r^{Group} , M \otimes M ] $$,

Taking $M = 1 + \epsilon L$, $r^{Group} = 1 + \epsilon r $ and looking on the terms in $\epsilon^2$ we get:

$$ \{ L \frac{ \otimes}{,} L \} = [r , L \otimes 1 + 1 \otimes L ] $$,

That is $r$-matrix bracket on Lie algebra. The point is that it is given by the same $r$-matrix. The only difference is that one bracket is quadratic (on the group), the other is linear (on Lie algebra). ($1+r$ or $r$ gives same bracket, because $1$ dissappears in RHS since it is given by commutator).

As far as I understand your question is how to go back from bracket on Lie algebra to bracket on Lie group. Well, as far as I understand that is somehow just saying that bracket on the group satisfies what we need: invariance and and limit to Lie algebra, the bracket satisfying such properties is unique and that is why it is what we need.

  • $\begingroup$ thank you very much. If we have a Lie cobracket $\delta: g \to \Lambda^2 g$, how to compute $r^{Group}$? $\endgroup$ – Jianrong Li Feb 13 '16 at 2:32
  • $\begingroup$ @JianrongLi The point is that they are the same $r = r^{Group}$ , the transformations which I made $r = 1 + \epsilon r $ they do not change the bracket, because RHS of bracket is given by commutator $[r, M \otimes M]$ so it exactly equals to $ [ (1+ r) , M \otimes M]$ because $1$ commute with everything. At least that how I remember things - I have not thought on these for years, it might be I misremember something, but the idea is like that. $\endgroup$ – Alexander Chervov Feb 13 '16 at 9:53
  • $\begingroup$ thank you very much. I still have some questions. Using $\delta: L \to \Lambda^2 L$, we can obtain $[,]_{L^*}: \Lambda^2 L^* \to L^*$. But how to compute the r-matrix $r$ for the Lie algebra $L$? $\endgroup$ – Jianrong Li Feb 13 '16 at 14:25
  • $\begingroup$ @JianrongLi I guess you know that - if $\delta$ is coboundary it gives rise to $r$. The point is that in $r$-matrix notation corresponding bracket on $L^*$ will have the form $\{L \otimes L\} = [r , L\otimes 1 + 1 \otimes L ]$, $\endgroup$ – Alexander Chervov Feb 13 '16 at 15:52

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.