4

I'm having some problems to prove the following assertion:

Let $\mathscr{D}$ be an involutive distribution of dimension $k$ in a manifold $N$. Let $(M,\varphi)$ be a integral connected submanifold, such that $\varphi(M)\subseteq N$ is a closed subset. Show that $M$ is a maximal connected integral submanifold of $\mathscr{D}$ (a leaf, say).

I've tried using the local version of Frobenius Theorem (each involutive distribution is integrable and locally its integral connected submanifolds are slices). However, the problem talks of something that is rather global (?), and I can't manage to complete the idea. I'd thank any kind of help.

1 Answers1

3

My solution uses two main ingredients.

1) Frobenius theorem ensures complete integrability. Let $N$ be the ambient manifold for the distribution $\mathcal{D}$. Then for each $p \in N$, there is a cubical (its image in $\mathbb{R}^{n}$ is an open cube) coordinate chart $(U,(x^{1},\dots,x^{n}))$ containing $p$, such that for all $q \in U$, $\mathcal{D}_{q} = span\{ \partial_{1}|_{q}, \dots, \partial_{k}|_{q} \}$. The slices $x^{k+1} = c^{k+1}$, $\dots$, $x^{n} = c^{n}$ then form embedded integral submanifolds of $\mathcal{D}$. These charts are called flat for $\mathcal{D}$.

2) Local structure of integral submanifolds: For every integral submanifold $M$ of $N$, and any chart flat for $\mathcal{D}$, the intersection $U \cap M$ is a countable disjoint union of open subsets of parallel $k$-dimensional slices of $U$, each of which is open in $M$ (in the "inner" topology of $M$) and embedded in $N$.

Note that I identify immersed submanifold $M$ with its image in $N$ (to save letters).

Now to the actual proof of the statement. Let $M$ be any integral submanifold, and let $M \subseteq M'$ for some connected integral submanifold of the distribution $\mathcal{D}$. We will now prove that $M$ is both closed and open in $M'$:

i.) $M$ is closed in $M'$:

By assumption $M$ is closed in $N$, hence it is closed in the subspace topology of $M'$ and consequently also in the "inner" topology of $M'$. We conclude that $M$ is a closed subset of $M'$.

ii.) $M$ is open in $M'$:

First, find a flat chart $(U,(x^{1},\dots,x^{n}))$ for $\mathcal{D}$ containing the point $p$. By 2) above, this means that both $M \cap U$ and $M' \cap U$ are disjoint union of some open subsets (open in the subspace topology of the slices) of parallel $k$-dimensional slices.

Let $S$ be the slice of $U$ containing $p$. As $M \subseteq M'$, from 2) we get the two open subsets $W,W' \subseteq S$, such that $M \cap S = W$ and $M' \cap S = W'$. As $W$ is open in $S$, it is also open in $W'$. But by 2) $W'$ is open in $M'$. This implies that $W$ is also open in $M'$. This statement is in fact quite subtle, using the fact that $W' \subseteq S$ with subspace topology is homeomorphic to the connected component of the open submanifold $M' \cap U \subseteq M'$. Note that by construction $W \subseteq M$.

To summarize, to every $p \in M$, we have found its neighborhood $W$ open in $M'$, such that $W \subseteq M.$ This proves that $M$ is open in $M'$.

As $M'$ is connected, we see that $M = M'$. If $M$ is assumed connected, we have just proved that it is a maximal connected integral submanifold. Q.E.D.

Observe that I have in fact proved a slightly stronger statement - any integral submanifold which is closed as a subset of $N$ and it is contained in some connected integral submanifold $M' \subseteq N$ must be necessarily the whole $M$'.

Jan Vysoky
  • 565
  • 3
  • 13
  • Might be good to explain also why $W$ is open in $W'$. I think it follows because $S$ and $W'$ are embedded in $N$ so they have the subspace topology. Then we can write $W = V \cap S$ and $W' = V' \cap S$ for some $V, V'$ open in $N$ and so $W = W' \cap W = (V' \cap S) \cap (V \cap S) = (V' \cap S) \cap V = W' \cap V$ which shows that $W$ is open in the subspace topology of $W'$. – Tob Ernack May 22 '23 at 03:50
  • I think this is not necessary - both W and W' are open in S (with S indeed having a subspace topology induced from N, but this is not really important). And every open subset W of S, which is a subset of another open subset W' of S is also open in W' - this is because $W = W' \cap W$. – Jan Vysoky May 23 '23 at 07:14
  • This shows that $W$ is open in $W'$ with the subspace topology inherited from $S$, but we want to know it is open in the topology inherited from $M'$. Probably the subspace topology from $S$ is coarser than that from $M'$ so it works. Maybe you already addressed the point in your answer in the sentence following "This statement is in fact quite subtle, [...]" and I misunderstood. – Tob Ernack May 23 '23 at 17:53
  • To be honest, I have no idea what I meant by the quoted sentence. Probably something similar to what you say. – Jan Vysoky May 24 '23 at 18:41