4

Let $\mathcal{F}_1$ and $\mathcal{F}_2$ be two foliations of a manifold. We say that $\mathcal{F}_1\pitchfork \mathcal{F}_2$ if $T_p L^{(1)}+T_pL^{(2)}=T_p M$ for any $p\in M$, where $L^{(1)}$ and $L^{(2)}$ are the leaves trough $p$.

Now if we have that $\mathcal{F_1}\pitchfork \mathcal{F}_2$, we define $\mathcal{F}_1\cap \mathcal{F}_2$ to be the foliation where the leaves are the connected components of $L^{(1)}\cap L^{(2)}$. Now we want to check that is a foliation with codimension the sum of the codimensions of $\mathcal{F}_1$ and $\mathcal{F}_2$.

I think the idea is to use the implicit function theorem but I am getting nowhere in concrete. That is take foliated charts for $\mathcal{F}_1$, $(x_1,...,x_k,y_1,...,y_{m-k})$, and foliated charts for $\mathcal{F}^2$, $(w_1,...,w_k',z_1,...,z_{m-k'})$ such that $p\cap L^{(1)}\cap L^{(2)}=\{p\in U: y_1(p)=ct,...,y_{m-k}(p)=ct,z_1(p)=ct,...,z_{m-k'}(p)=ct\}$ . Now from this I have tried and construct a new coordinate chart so that we get the result bu I got nowhere.

Any help or hints with this are aprecciated. Thanks in advance.

Attempt at solution :

First thing we note is that since $\mathcal{F}_1$ and $\mathcal{F}_2$ are foliations we will have that the leaves of this new foliation will cover $M$, are disjoint since we are taking the connected components, and will be path connected since we are taking the connected components and $M$ is locally path-connected. Now we need to find the foliated charts of $\mathcal{F}$ and check it's dimension. Let $p\in M$, note that since $L^{(1)}$ and $L^{(2)}$ are transversal we will have that $L^{(2)}\cap L^{(1)}$ is a submanifold of $L^{(1)}$.Using the local normal form we know that there exists open set of $p\in L^{(2)}, U,$ and coordinate chart $(V,\phi)=(V,x_1,...,x_k)$ for $L^{(1)}$ such that $U\cap L^{(1)}=\{p\in V : x^{k'}(p)=...=x^k(p)=0\}$, where $k'=dim L_1 -(dim L_1 + dim L_2 -dim M)= codim \mathcal{F}_2$. Now note that the way we gave the manifold structure to $L^{(1)}$ was that the topology was generated by the plaques $L^{(1)}\cap U'$ where $U'$ was a foliated chart for $p$, and the coordinate chart was the restriction of the foliated chart to the non-constant components. So we have the $(x_1,...,x_k)$ are associated with a chart of $M$, $(x^1,...,x^k,y^1,...,y^{d-k})$ such that the connected components of $L^{(1)}\cap U'$ are of the form $\{p\in U' : y^1(p)=ct,...,y^{d-k}(p)=ct\}$. Now we have $U$ is a plaque for $U''\cap L^{(2)}$. We also know that $p$ will be in a connected component of $L^{(1)}\cap L^{(2)}$, wich we now denote by $L$. If we consider the foliated chart to be $(W:=U'\cap U'',x^1,...,x^k,y^1,...,y^{d-k})$ we will have that the connected components of $W\cup L$ are of the form $\{p\in W: y^1(p)=ct,...,y^{d-k}(p)=ct,x^{k-k'}(p)=,...,x^k(p)=0\}$. To see the statement about the dimension we note that $codim \mathcal{F}_1+ dim L^{(1)}-(dim L^{(1)}+dim L^{(2)}-dim M)=codim\mathcal{F}_1 +codim \mathcal{F}_2 $.

Someone
  • 5,059

1 Answers1

4

Here is, I think, an answer. Let $\mathcal{D}_1$ be the tangent ditribution of the foliation $\mathcal{F}_1$ and $\mathcal{D}_2$ those of $\mathcal{F}_2$. Theses distribution are totally integrable, that is $[\mathcal{D}_i,\mathcal{D}_i]\subset \mathcal{D}_i$. Let $\mathcal{D} = \mathcal{D}_1\cap \mathcal{D}_2$ be the intersection distribution. It has constant rank as at any $p\in M$, $\dim\mathcal{D}_p = \dim\left({\mathcal{D}_1}_p\cap{\mathcal{D}_2}_p\right)=\dim{\mathcal{D}_1}_p+\dim{\mathcal{D}_2}_p-\dim\left({\mathcal{D}_1}_p+{\mathcal{D}_2}_p\right)$, and then $\mathrm{codim}\mathcal{D}_p = \mathrm{codim}{\mathcal{D}_1}_p + \mathrm{codim}{\mathcal{D}_2}_p$ (recall that $\mathcal{F}_1\pitchfork \mathcal{F}_2$)

Now, let us show that $\mathcal{D}$ is integrable, that is $[\mathcal{D},\mathcal{D}]\subset \mathcal{D}$. Let $X,Y$ be vector fields in $\mathcal{D}$. As $\mathcal{D}\subset \mathcal{D}_1$, $[X,Y]$ is in a vector field in $\mathcal{D}_1$. Symmetrically, it is a vector field in $\mathcal{D}_2$. Then, it is a vector field in $\mathcal{D}_1 \cap \mathcal{D}_2 = \mathcal{D}$, and $[\mathcal{D},\mathcal{D}]\subset \mathcal{D}$. By Frobenius theorem, there exists a foliation $\mathcal{F}$ those tangent bundle is $\mathcal{D}$. The integral submanifolds are the intersections of the leaves of the previous foliations.

Edit Here is another strategy. Fist, show that the intersection of two transverse submanifolds is a submanifold. Let $N$ and $N'$ be two trasverse submanifold with $\mathrm{codim} N= p, \mathrm{codim} N' =q$. Let $x \in N\cap N'$ be fixed. Take a chart $\varphi : U \to \mathbb{R}^n$ centered at $x$. Then $\varphi (U\cap N)$ and $\varphi(U\cap N')$ are submanifolds of $\varphi(U)$ of codimension $p$ and $q$. There exist $f = (f_1,\ldots,f_p): \mathbb{R}^n \to \mathbb{R}^p$ and $g=(g_1,\ldots,g_q): \mathbb{R}^n\to \mathbb{R}^q$ smooth submersions such that $\varphi(U\cap N) = f^{-1}(\{0\})$ and $\varphi(U\cap N')= g^{-1}(\{0\})$. Let $h = (f_1,\ldots,f_p,g_1,\ldots,g_q) : \mathbb{R}^n \to \mathbb{R}^{p+q}$. The transversality condition assures that $h$ is a sumbersion, and thus, $\varphi(U\cap\left(N\cap N'\right)) = h^{-1}(\{0\})$ is a submanifold of $\varphi(U)\cap \mathbb{R}^n$ of codimension $p+q$, and $U\cap (N \cap N') = \varphi^{-1}(h^{-1}(\{0\}) = (h\circ\varphi)^{-1}(\{0\})$ is a smooth submanifold of codimension $p+q$.

Let $\mathcal{F}^1$ and $\mathcal{F}^2$ be two transverse smooth foliations on $M^n$. If $L^1$ and $L^2$ are leaves of $\mathcal{F}^1$ and $\mathcal{F}^2$, and if $L^1\cap L^2 \neq 0$, then the result above shows that $L^1\cap L^2$ is a smooth submanifold of $M$ with codimension $\mathrm{codim}L_1 + \mathrm{codim}L_2$. As the dimension of the leaves of a foliation is constant, if two transverse leaves intersect, their intersection is a submanifold of dimension $n - \mathrm{codim}\mathcal{F}^1 -\mathrm{codim}\mathcal{F}^2$.

It is staightforward to show that $\{L^1\cap L_2 ~|~ L_1 \in \mathcal{F}^1,L_2\in \mathcal{F}^2,L_1\cap L_2\neq 0\}$ is a smooth foliation.

Remark: this shows that the distribution $D_1\cap D_2$ associated is smooth.

user57
  • 828
Didier
  • 20,820
  • I aprecciate the answer and It's a nice application of the frobenius theorem but I was trying to do it in a simpler way. – Someone Oct 29 '20 at 10:25
  • 1
    Then I encourage you to read the proof of Frobenius theorem, I think all the ideas are inside. Or maybe the proof of transversality theorem – Didier Oct 29 '20 at 10:27
  • I just have a question. Frobenius theorem gives us the exsitence of a unique foliation, and we know that the intersection of the leaves are submanifolds that have this tangent space, but apriori we don't know that these form a foliation so that we can claim that this is exactly the foliation we want right ? @Dldier_ – Someone Oct 31 '20 at 07:24
  • 1
    The intersection of two transverse submanifolds is a submanifold. intersection is tangent to the intersection of the two distributions $\mathcal{D}_1\cap\mathcal{D}_2$, thus it is an integral submanifold of the distribution $\mathbb{D}$. By uniqueness in Frobenius theorem, all the intersections of the leaves form the foliation. – Didier Oct 31 '20 at 09:20
  • @Dldier_ doesnt the maximality property just gives me that the intersection of all the leaves is contained in the foliation ? Do you know any reference where I can look this up ? – Someone Oct 31 '20 at 11:31
  • Yes, but the transversality theorem gives you the good dimension (the codimension is the sum of the condimension), and by maximality, the intersections are the leaves of the foliation! – Didier Oct 31 '20 at 11:43
  • Yeah makes sense I was confusing something.Thanks ! – Someone Oct 31 '20 at 11:52
  • 1
    I am sorry to be only remembering this now, but why will $\mathcal{D}$ be a smooth distribution ? So that we can apply the frobenius theorem @Dldier – Someone Nov 09 '20 at 14:23
  • @Something That is a really interesting and important question. Right now, I am sorry, but I don't have an answer. I will try to figure out if there is a counter example or a simple proof. I encourage you to do the same – Didier Nov 09 '20 at 14:35
  • 1
    @Something in an eddit of my answer, I addd another strategy to prove your first question: the codimension od the intersection of the leaves is constant – Didier Nov 15 '20 at 14:02
  • 1
    Ah thanks @Dldier_. I think your idea is kinda what I am doing at my attempt at a solution , but I could be mistakened. – Someone Nov 15 '20 at 14:11
  • @Didier, hey! What gives the existence of "There exist $=(1,…,): \mathbb{R}^n \rightarrow \mathbb{R}^p$ and $=(1,…,):\mathbb{R}^n\rightarrow \mathbb{R}^$ smooth submersions such that $\varphi(\cap )=^{-1}({0})$ and $\varphi(\cap′)=^{-1}({0})$."? Thank you in advance – user57 Oct 31 '23 at 17:25
  • 1
    @user57 The characterization of submanifolds as being locally the zero-locus of a submersion. – Didier Oct 31 '23 at 17:30
  • @Didier how straightforward is it to show that the intersection of the leaves will be a smooth foliation? Having two charts for the two intersecting leaves, I'm having trouble creating a chart for the intersection such that it is a countable union of slices. – ImHackingXD Nov 04 '23 at 14:08
  • 1
    @ImHackingXD The fact that it forms a partition is clear. The fact that the connected components of the intersections are immersed submanifolds of constant dimension is a consequence of the assumption that the intersection is transverse – Didier Nov 05 '23 at 23:07
  • @Didier What I wanted to know is why the connected components of the intersections locally look like a countable union of slices, but I figured it out that's because of your proof that they are transversal: from the function $h$, you can complete it so that you get a foliated chart right? – ImHackingXD Nov 06 '23 at 14:05