Let $G$ be a Lie group acting on itself as $\phi(h)(g)= L_h(g)$ as a left translation. Then we can consider the cotangent lift of this action, namely $\Phi: G \times T^*G \rightarrow T^*G$ as $\Phi(h)(g,p) = (hg,(dL_{h^{-1}}(hg))^*p).$

It can now be shown that such a map induces a canonical Hamilton function with moment map on the cotangent bundle

$H_{\xi}(g,p) = J(q,p)(\xi):=(dR_g)^*(e)(p)(\xi)$ for some $\xi \in \mathfrak{g}.$

This is now my motivation for the question:

If we consider $J^{-1}(x)$ for $x \in \mathfrak{g}^*$ then this set is given by $$J^{-1}(x) = \{ (g , (dR_{g^{-1}})^*(g)(x));g \in G \}.$$

My question is: Why is it a manifold? (I admit that it looks very much like an application of the regular value theorem, but I don't see why it applies)

If there is anything unclear about my question, please let me know.

  • $\begingroup$ Do you want me to try an answer for that one, as well? Although I really think that you just need to follow your nose and try some simple examples. Start out with $G=S^1$, and after it $SO(3;\mathbb{R})$. $\endgroup$
    – CvZ
    Sep 9, 2015 at 2:07
  • $\begingroup$ @CvZ could you maybe briefly sketch in a comment how you want to show that this is a manifold? (maybe mention some theorems that you want to use)? $\endgroup$
    – user167575
    Sep 9, 2015 at 2:11
  • $\begingroup$ You wanted to see if this is an application of the regular value theorem, is it not? So try it for $G=S^1$, check if the momentum mapping satisfies the hypothesis of the theorem. You do understand what happens for the momentum mapping of an integrable system, do you not? $\endgroup$
    – CvZ
    Sep 9, 2015 at 2:16
  • 1
    $\begingroup$ ah, then $H_{\xi}(q,p) = \xi(q_1p_2-q_2p_1),$we can differentiate w.r.t to any coordinate and get $dJ(q,p) = (p_2,-p_1,-q_2,q_1).$ So not any $x$ is a regular value, as $x=0$ does not do it, as then $dJ(0,0)=(0,0,0,0) $ in particular. $\endgroup$
    – user167575
    Sep 9, 2015 at 2:45
  • $\begingroup$ I recommend you to change your title. Every smooth action of Lie group on any manifold generates smooth manifolds: the orbits. It is a misleading title and not that helpful for other users. Try something with the words 'smooth manifolds and preimage of momentum mappings'. $\endgroup$
    – CvZ
    Sep 9, 2015 at 16:46

1 Answer 1


Using the left multiplication of $G$ on itself one can show that $T^*G\cong \mathfrak{g}^*\times G$, and under that identification the lifted action is simply $h\cdot (\xi,g)=(\xi,h\cdot g)$, and the momentum mapping is $\mathfrak{g}^*\times G\ni (\xi,g)\mapsto(\mathrm{Ad}_{g^{-1}})^*(\xi)\in\mathfrak{g}^*$.

The preimage by the momentum mapping of a point $\zeta\in\mathfrak{g}^*$ is the set of points $(\xi,g)\in\mathfrak{g}^*\times G$ satisfying $(\mathrm{Ad}_{g^{-1}})^*(\xi)=\zeta$; thus, $\{((\mathrm{Ad}_g)^*(\zeta),g)\}_{g\in G}\subset\mathfrak{g}^*\times G$.

The mapping $G\ni g\mapsto ((\mathrm{Ad}_g)^*(\zeta),g)\in\mathfrak{g}^*\times G$ provides a diffeomorphism between the The preimage by the momentum mapping of a point $\zeta\in\mathfrak{g}^*$ and $G$.


In the Original Poster notation, $J^{-1}(x)=\{(g,(dR_{g^{-1}})^*(g)(x)); \ g\in G\}$ and the mapping $G\ni g\mapsto (g,(dR_{g^{-1}})^*(g)(x))\in T^*G$ provides a diffeomorphism between $G$ and $J^{-1}(x)$; therefore the preimage $J^{-1}(x)$ is a smooth manifold.

  • $\begingroup$ I am really tired right now, so I might be wrong about this. Morrow is another day and I shall revise this answer. $\endgroup$
    – CvZ
    Sep 9, 2015 at 4:42
  • $\begingroup$ so you say: Yes it is a manifold, because it is diffeomorphic to $G$? I think you can also see it from my explicit representation of the preimage in the question, right? $\endgroup$
    – user167575
    Sep 9, 2015 at 11:23
  • $\begingroup$ Sorry for not see a direct answer. I was really tired and had to rewrite things in a way that I was able to work it out. However, It is really useful to work with left and right trivialisations of $TG$ and $T^*G$. $\endgroup$
    – CvZ
    Sep 9, 2015 at 16:43

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.