Given the following Lie transformation:

$$ \exp(\lbrace H, \cdot \rbrace):=\sum_{n=0}^{\infty} \frac{(\lbrace H, \cdot \rbrace)^n}{n!} $$ and apply it to a Poisson Bracket $\lbrace g_1, g_2 \rbrace$. I would like to show that $$ \exp(\lbrace H, \lbrace g_1, g_2 \rbrace\rbrace)=\sum_{n=0}^{\infty} \frac{(\lbrace H, \lbrace g_1, g_2 \rbrace \rbrace)^n}{n!}=\dots=\lbrace \sum_{n=0}^{\infty} \frac{(\lbrace H, g_1\rbrace)^n}{n!}, \sum_{n=0}^{\infty} \frac{(\lbrace H, g_2\rbrace)^n}{n!}\rbrace=\lbrace\exp(\lbrace H, g_1\rbrace),\exp(\lbrace H, g_2\rbrace)\rbrace $$ I put some dots in the Passage I am not able to do...(I tried expressing the power using the binomial theorem)

I am sure this property holds, since I found this here (page 25): http://www.aps.anl.gov/Science/Publications/lsnotes/content/files/APS_1418211.pdf

but I have difficulties in proving this in an elegant way. Could someone give me a good reference?

This is the way I tried: $$\exp(\lbrace H, \lbrace g_1, g_2 \rbrace\rbrace)=\sum_{n=0}^{\infty} \frac{(\lbrace H, \lbrace g_1, g_2 \rbrace \rbrace)^n}{n!}= \sum_{n=0}^{\infty} \frac{( \{\{H,g_1\},g_2\}+\{g_1,\{H,g_2\}\} )^n}{n!}= \sum_{n=0}^{\infty} \frac{1}{n!} \sum_{k=0}^n {n \choose k}(\{\{H,g_1\},g_2\})^{n-k}(\{g_1,\{H,g_2\}\})^k = $$ $$\sum_{n=0}^{\infty} \sum_{k=0}^n \frac{1}{(n-k)!}(\{\{H,g_1\},g_2\})^{n-k}\frac{1}{k!}(\{g_1,\{H,g_2\}\})^k $$

  • $\begingroup$ H,g1,g2 are function of which you take the Poisson Bracket. I get stuck in the second equality above $\endgroup$
    – Chaos
    Dec 25, 2015 at 19:11
  • $\begingroup$ Use the Leibniz rule on the Poisson brackets and the result should easily follow. $\endgroup$
    – gented
    Dec 25, 2015 at 19:12
  • $\begingroup$ I am sorry but I cannot see where to use it.. $\endgroup$
    – Chaos
    Dec 25, 2015 at 19:14
  • 2
    $\begingroup$ Are you sure $H$ is not the Hamiltonian and $f,g$ functions of $q,p$? If so, the Poisson bracktes can de reduced to derivatives. $\endgroup$
    – gented
    Dec 25, 2015 at 19:22
  • $\begingroup$ Yes in my case H is the Hamiltonian and g1,g2 functions of q,p...however I tried to write all in terms of derivatives but it is a mess....I am looking for a clever way to show it...I think this should be a standard property that should be demonstrated in Lie algebras books but I have no idea whare to search...googling it did not provide me satisfactory results... $\endgroup$
    – Chaos
    Dec 25, 2015 at 19:28

1 Answer 1


Although OP strictly speaking writes something different in the question formulation (v4), it seems OP essentially wants to prove the following Lemma.

Lemma. Given a (possibly infinite-dimensional$^1$) Lie algebra $L$ over a field $\mathbb{F}\supseteq\mathbb{R}$ with Lie bracket $[\cdot,\cdot]:L\times L \to L$ and Lie algebra derivation$^2$ $D:L\to L$, which satisfies Leibniz rule $$ \forall x,y\in L:~~ D[x,y]~=~ [Dx,y]+[x,Dy].\tag{1}$$ Assume that the exponential map $$e^D:= {\bf 1}_L+ D + \frac{1}{2}D^2+ \ldots ~:~L\to L\tag{2}$$ is pointwise convergent. Then $$ \forall x,y\in L:~~ e^D[x,y]~=~[e^Dx,e^Dy].\tag{3}$$

Proof of eq. (3): The identity (3) follows by setting $t=1$ in the following identity:

$$ \forall t\in [0,1]\forall x,y\in L:~~ e^{tD}[x,y]~=~[e^{tD}x,e^{tD}y].\tag{4}$$

To prove eq. (4), first note that eq. (4) is trivially true for $t=0$. Next show that the lhs. and the rhs. of eq. (4) independently satisfy the same first-order ODE $$ \frac{d}{dt}(\ldots)~=~D(\ldots). \tag{5}$$ Hence they must be equal.


$^1$ Note that the Poisson algebra $(C^{\infty}(M);\{\cdot,\cdot\}_{PB})$ of smooth functions over a finite-dimensional phase space $M$ is an infinite-dimensional Lie algebra.

$^2$ In OP's case the derivation $D=[z,\cdot]$ is an inner derivation.

  • $\begingroup$ what do you mean with OP? $\endgroup$
    – Chaos
    Dec 25, 2015 at 21:21
  • 2
    $\begingroup$ OP=Original Poster. In this case: You. $\endgroup$
    – Qmechanic
    Dec 25, 2015 at 21:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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