I am quite confused on the definition of a lagrangian submanifold $L$ of a symplectic manifold $(M,\omega)$.

In particular, I read that $L \subset M$ is lagrangian iff the symplectic form field $\omega(x)$ evaluated on every point $p\in L$ gives zero. How is it possible that $\omega$ assumes the value zero on a subset $L$ of $M$, but still it is not degenerate (it is never zero) over the whole $M$?!

Furthermore, consider $M=\mathbb{R}^2$ with the standard symplectic form $\omega=dq\wedge dp$. It appears to me that there are no lagrangian submanifolds at all, since the symplectic form is everywhere constant and never zero. However, I am reading that for example all the submanifolds of the tipe $q=const$ are lagrangian, in this case.

Clearly, I am not understanding well the definition of Lagrangian submanifold.

Where am I going wrong?

up vote 7 down vote accepted

The condition of being a lagrangian submanifold $L\subset M$ means that for every point $p\in L$ and every pair of tangent vectors $X,Y\in T_pL$, $$\omega_p(X,Y)=0\,.$$ This does not violate the non-degeneracy of $\omega$, since $T_pL$ is merely a subspace of $T_pM$. However, non-degeneracy implies that any submanifold satisfying this condition is of dimension at most $\frac12\dim M$. Lagrangian submanifolds are submanifolds satisfying the condition above of maximal possible dimension, i.e. $\dim L=\frac12\dim M$.

Another was of saying the same thing is that the pullback $i^*\omega$ of $\omega$ under the inclusion (smooth map) $i:L\hookrightarrow M$ of the $2$-form $\omega$ is zero: $$i^*\omega=0\,.$$

  • 2
    Ok, now I understand. Thank you very much. The point is that since the dimension of $L$ is smaller than the dimension on $M$, $\omega(p)$ can be zero on $T_pL$ but still nonzero on $T_pM$, cause $T_pL$ is just a vector subspace of $T_pM$, preserving thus the nondegeneracy condition. I feel really stupid now, it was easy! – Federico Carta Dec 19 '14 at 18:49

In particular, I read that $L\subset M$ is lagrangian iff the symplectic form field $\omega(x)$ evaluated on every point $p\in L$ gives zero.

More precisely, the restriction of $\omega$ to $T_p L$ is zero. As an element of $\wedge^2 T_p M$, $\omega(p)$ is nonzero and nondegenerate.

For an example, consider $\mathbb{R}^2$ with the symplectic form $\omega=dx\wedge dy$. This is surely nondegenerate, but the 1-dimensional submanifold spanned by $L=\langle 1,0 \rangle$ (the $x$-axis) satisfies that $\omega$ vanishes on the tangent space of $L$.

It is, in fact, an exercise in linear algebra that a quadratic form of signature $(n,n)$ admits a decomposition of the vector space into two parts $V_1\oplus V_2$ s.t. the restriction of the form to $V_i$ is trivial. (This decomposition is not unique in general.) For example, the matrix $$ \begin{pmatrix} 0 & 1\\ -1 & 0 \end{pmatrix} $$ is nondegenerate but the quadratic form it represents restricted to $\langle 1,0\rangle$ is just the zero in the upper left corner.

Nonetheless, there is in general not a simple answer to the question stated in the title of your question.

Your Answer


By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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