I am asking the question because in the classical mechanics book by Arnold, he states that there is a distinguished $1$-form on $T^*V$. It seems that there is no such distinguished $1$-form on a general even-dimensional manifold.

So, not every even-dimensional manifold can be the phase space of a system, or not every even-dimensional manifold can be a cotangent bundle?

Is $S^2$ a cotangent bundle (intuitively it is not)? How about $S^4$?

  • 11
    $\begingroup$ Vector bundles are non-compact. $\endgroup$ Aug 19, 2015 at 15:50
  • $\begingroup$ Great! I feel stupid to miss this simple point! $\endgroup$
    – kaiser
    Aug 19, 2015 at 15:59
  • 2
    $\begingroup$ The argument in the first paragraph is a red herring. While it's true that a manifold isn't equipped with a distinguished $1$-form, it's also true that a manifold that admits the structure of a cotangent bundle isn't equipped with a distinguished such structure. $\endgroup$ Aug 19, 2015 at 18:34

2 Answers 2


As I alluded to in the comments, vector bundles of positive rank are necessarily non-compact. Therefore, no compact manifold can be the total space of a vector bundle of positive rank over any topological space. In particular, neither $S^2$ nor $S^4$ can be a cotangent bundle.

On the other hand, any space can be viewed as a rank zero vector bundle over itself.


The answer to your main question is no. Here is an argument (perhaps there are simpler arguments, but I came out with this one).

You can show that no sphere $S^{2n}$ for $n>1$ admits as symplectic structure. This was asked before here.

Let me elaborate a bit on the "distinguished" one-form to show that the cotangent bundle always admits a natural symplectic structure.

Local coordinates in the cotangent bundle are given by a pair $(x,a)$ where $p$ stands for a point on the manifold and $a$ for the coefficients of a basis in the cotangent space. In other words, $(p,a)$ stands for the point $a_idx^i$ with local coordinates centered at $p$. Consider the projection map $\pi\colon T^*M \to M$ which assings $(p,a)$ to $p$. These map induces a pullback $\pi\colon T^*M \to T^*T^*M$. You can identify $T^*T^*M$ with $T^*M$ itself. At each point $(p,a)$, define a one-form taking the value $\pi^*(p,a)$. This defines a one-form called the tautological one-form $\theta$. It is easy to prove that $d\theta$ defines a symplectic structure on $T^*M$.

This argument would prove that $S^4$ cannot be the cotangent bundle of a manifold, since it admits no symplectic structure.

I admit this is not the neatest presentation of the tautological one-form, but there is plenty about it on wikipedia.

EDIT: The other answer is definitely better. But I hope this illustrates a bit of the so-called distinguished form. The symplectic structure, after all, is import for the study of phase spaces.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .