MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Hi, I am reading "Introduction to symplectic topology" by McDuff and salamon. At some point I cant go further. My question is: Let $(M,g)$ be a Riemannian manifold and consider the cotangent bundle $T^{\star}M$. My question is now, how one can show that there is an bundle isomorphism $\varphi : TM \oplus T^{\star}M \rightarrow T(T^{\star}M)$, s.t. $d\pi \circ \varphi$ is the identily on $TM$ and it preserves the symplectic forms. In the book there is a hint giving, by saying that one can choose $\varphi$ via a connection on the cotangen bundle.

share|cite|improve this question
What is $T^M$?? – Olivier Bégassat May 12 '11 at 6:18
Sorry I did some mistake. Its the cotangent bundle. Now I have changed it. – marco May 12 '11 at 7:27
The proposed isomorphism can't be right. Let $M$ be $n$-dimensional; then $\mathrm T M$ is rank-$n$, as is $\mathrm T^\ast M$, so $\mathrm T \oplus \mathrm T^\ast$ is rank $n+n$, so the total space of $\mathrm TM \oplus_M \mathrm T^\ast M$ has dimension $3n$. On the other hand, the total space of $\mathrm T(\mathrm T^\ast M)$ has dimension $2(2n) = 4n > 3n$ if $n\neq 0$. I agree that there is a map, however. (Side remark, at the risk of being a little rude: please correct spelling, capitalization, and punctuation; MO should be more professional than emails.) – Theo Johnson-Freyd May 12 '11 at 16:28

Your Maths does not add up. Let $n$ be the dimensional of $M$. The LHS is a vector bundle of dimension $2n$ on $M$. The RHS is a vector bundle of dimension of $2n$ on $T^*M$. If you just consider it as a bundle on $M$ , its dimension is $3m$, so no iso possible.

share|cite|improve this answer
But is it possible to decompose $T(T^{\star}M)$ in such a way ? – marco May 12 '11 at 10:47

I guess what you want to prove is the following: for a given vector bundle $E$ over $M$ there is a canonical subbundle, the vertical bundle $\mathrm{Ver}(E) = \ker T\pi \subseteq TE$ of the tangent bundle of the total space., where $\pi\colon E \longrightarrow M$ is the bundle projection. Then the choice of a connection gives a complementary subbundle $\mathrm{Hor}(E)$, the horizontal bundle. Depending on your favorite definition of a connection, this is just the definition ;) Now what is true is that the horizontal bundle is isomorphic to the pull-back bundle $\pi^\sharp TM$ of the tangent bundle of the base $M$. This is intrinsically defined as vector bundle over $E$ but it is not a subbundle of $TE$ in an intrinsic way. Here you need the connection. But then you have $\pi^\sharp TM \oplus \ker T\pi \cong TE$.

Now you can unwind this for $E = T^*M$ and get your result (stated correctly). I hope that helps.

EDIT: for $T^*M$ one then has a canonical pairing of the horizontal and vertical spaces at every point. This you can either symmetrize to get a pseudo-riemannian metric or antisymmetrize to get a symplectic form (guess which). Interestingly enough, the metric depends on the connection while the antisymmetric form does not (you may need torsion-free, I forgot). Also as a side remark: the Laplacian (better d'Alembertian) of the metric plays an important role in quantization theory when passing from standard to Weyl ordering on a cotangent bundle.

You can find some background info in the book of Yano&Ishihara or, if you prefer german, in Sect 5.4 of my book ;)

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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