Take the 2-minute tour ×
MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.


According to Arnol'd, a contact structure on a smooth manifold $M$ is given by a corank 1 tangent distribution $C$ which is maximally non-integrable; this means that, for any local $1$-form $\alpha$ on $M$, $$\ker\alpha=C\Longrightarrow d\alpha\textrm{ is non-degenerate on }C.$$ A contact structure is said to be co-orientable if there exists a global $1$-form $\alpha$ on $M$ such that $\ker\alpha=C$.
Then a contact manifold $(M,C)$ is constituted by a smooth manifold $M$ equipped with a contact structure $C$. Necessarily we have that

  • $\dim M=2n+1$, and
  • if $N\subset M$ is an integral manifold of $C$, i.e. $TN\subset C$, then $\dim M\le n$.

In particular $n$-dimensional integral manifolds are called Legendrian submanifolds of $(M,C)$.
A diffeomorphism $\phi:M_1\to M_2$ is said to a contactomorphism of $(M_1,C_1)$ onto $(M_2,C_2)$ if it satisfies $$(T\phi)C_1=C_2.$$


Let us remark that the current context is slightly different than the one adopted in the previous question, where only co-orientable contact manifolds are examined.
So I am wondering myself if, in our case, it is possible to get an analogous Legendrian Tubular Neighbourhood Theorem.

Let $N$ be a legendrian submanifold of $(M,C)$. Is again possible to find open neighborhoods $U$ and $V$ of $N$ respectively in $M$ and $J^1(N,\mathbb R)$, such that there exists a contactomorphism $$\phi:(U,C|_U)\to(V,\mathscr C|_V),\textrm{ with }\phi|_N=\operatorname{id}_N?$$ Above $\mathscr C$ is the Cartan distribution on $J^1(N,\mathbb R)$,and $N$ is canonically identified with $j^1 0$.

Actually, this question should admit some equivalent reformulations like as:

  • is the line bundle $(TM)/C$ trivial over $N$?
  • is there a local contact form for $(M,C)$ which is defined on a whole neighborhood of $N$ in $M$?

Probably, in the current context, there does not exist something like a Legendrian Tubular Neighborhood Theorem, but until now I have not been able to point out a counter-example. In such a negative case, I would also ask

there exists a classification of legendrian embeddings?

As usual any feedback is welcome.

share|improve this question
Consider the natural contact structure on the projectivization $PT^*M$ of cotangent bundle to a manifold $M$. Any fiber of the natural projection $PT^*M \to M$ is Legendian submanifold. The contact structure restricted to any neighborhood of a fiber is non-coorientable, i.e. could not be a kernel field of a contact form. –  Petya Nov 16 '13 at 21:14
Dear @Petya, could you give more details about the non-coorientability of the contact structure of $\mathbb P(T^\ast M)$ in any neighbourhood of the fibers of $\mathbb P(T^\ast M)\to M$? You could also consider to post your comment as an answer. –  Giuseppe Tortorella Nov 17 '13 at 22:38
Dear GiuseppeTortorella, here's my own point of view on non-coorientability: if $N=\mathbb{P}T^*_xM$, then the line bundle $(T(\mathbb{P}T^*M))/C$, restricted to $N$, is nothing but the tautological line bundle $\ell$ over the projective space $\mathbb{P}^{n-1}$, which is non-trivial (by $n$ I mean the dimension of $M$). (If $\ell$ is trivial, then it admits a nowhere zero section which, in turn, allows to embed $\mathbb{P}^{n-1}$ into $\mathbb{R}^n$.) P.S. Thanks to @Petya for suggesting the counterexample. –  G_infinity Nov 18 '13 at 8:49
Dear @G_infinity, so, on a neighbourhood of $\mathbb P_x(T^\ast M)$ in $\mathbb P(T^\ast M)$, there doesn't exist any contact form $\alpha$, otherwise it would determine a global section of the line bundle. Thank you very much. –  Giuseppe Tortorella Nov 18 '13 at 9:58
add comment

1 Answer 1

This is really a comment, but I don't have enough reputation to leave one (or to vote the question up, which I like indeed!). As pointed out by the comments of Petya and G_infinity, you should not expect a local normal form theorem of the kind that is asked for in the question, as the line bundle restricted to N need not be trivial.

Let me, therefore, suggest a different take to the question, by suggesting that the model contact structure proposed is not the correct one. Let $p: L := TM/C|_N \to N$ be the restriction of the line bundle to the Legendrian submanifold $N$. Denote by $J^1 L$ the space of 1-jets of sections of $L \to N$, that is $$ J^1 L : = \{ j^1_x \sigma \mid x \in N \text{ and } \sigma \in \Gamma_{\mathrm{loc}}(L) \}. $$

There are natural projections $\mathrm{pr} : J^1 L \to L$ and $p \circ \mathrm{pr}: J^1 L \to N$ given by $\mathrm{pr}(j^1_x \sigma) = \sigma(x)$, and $p \circ \mathrm{pr}(j^1_x \sigma) = x$ respectively. The smooth manifold $J^1 L$ carries a canonical contact structure $H$, which is defined as the kernel of a 1-form $\theta_{\mathrm{can}} \in \Omega^1(J^1 L ; \mathrm{pr}^* p^* L)$, which is known in the literature on exterior differential systems and the geometric theory of PDEs as the Cartan form. Before defining $\theta_{\mathrm{can}}$, observe that the line bundle $p^* L \to L$ can be identified as the vertical tangent bundle $\ker Dp \subset TL \to L$, by identifying the tangent space to each fibre $L_x$ with the fibre itself (recall that the fibre is a vector space). Therefore, given a point $q = j^1_x \sigma$ and a vector $X_q \in T_q J^1 L$, set $$ \theta_{\mathrm{can}}(X_q) := D_q\mathrm{pr}(X_q) - D_x \sigma \circ D_{\sigma(x)}p \circ D_q \mathrm{pr} (X_q); $$ this is easily seen to take value in the appropriate line bundle, and its kernel $H$ is a contact distribution. The Cartan form $\theta_{\mathrm{can}}$ generalises the canonical contact 1-form on $J^1(N;\mathbb{R})$; in particular, a (locally defined) section $s \in \Gamma_{\mathrm{loc}}(J^1L)$ satisfies $s^*\theta_{\mathrm{can}} = 0$ if and only if $s = j^1 \sigma$, for some $\sigma \in \Gamma_{\mathrm{loc}}(L)$.

Thus it makes sense to modify the question slightly to allow an open neighbourhood of the zero section of $p \circ \mathrm{pr} : J^1 L \to N$ as a local model for an open neighbourhood of the Legendrian submanifold $N \subset (M,C)$, with contact structure $H$ defined above. That is, to look for open neighbourhoods $U$ of $N \subset M$ and $V$ of the zero section of $J^1L \to N$, and a contactomorphism $\phi: (U, C|_U) \to (V,H|_V)$ with $\phi|_N = id_N$. I don't know the answer to this different question, but I will think about it and see whether I can get anywhere with it.

share|improve this answer
Dear Daniele Sepe, first of all, thank for your attention to my question, and sorry for the delay in my comment; initially, some activities forced me to reduce my presence on the site, then, I forgot of you answer. Now, I do not know if, in the meanwhile, you went somewhere with the modified question. But I have found a 1998 paper by F. Loose which gives a positive answer exactly to that question after having chosen the graph of the $1$-jet of the zero section of $p:L|_N\to N$ as the normal form of contact manifolds on a neighborhood of a legendrian submanifold $N$. –  Giuseppe Tortorella Apr 30 at 18:29
Dear @GiuseppeTortorella, thanks for the reference, I was not aware of it! –  Daniele Sepe May 1 at 12:57
add comment

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.