I'm on my way proving the Preimage Lemma for a transverse smooth map but I've encountered some problems with two passages:

Let $f\colon M \to N$ be a smooth map transverse to a submanifold $L$ of $N$ of codimension $k$ and $f^{-1}(L)$ is not empty. Then $f^{-1}(L)$ is a codimension $k$ submanifold of $M$

The proof given during lecture is the following:

Let $f(p)=q$, $q \in L$. For some neighbourhood $V$ of $q$, $L\cap V \simeq \mathbb{R}^{n-k}\cap V'$ where we see $\mathbb{R}^{n-k}$ as $(x_1,\dots,x_{n-k},0,\dots,0)$ in $\mathbb{R}^n$). Let $\pi\colon \mathbb{R}^n\to \mathbb{R}^k$ be the projection to the last $k$-coordinates. By transversality we have that $0 \in \mathbb{R}^k$ is a regular value of the composition $$ U \xrightarrow{f} V \simeq V' \xrightarrow{\pi} \mathbb{R}^k$$ Hence for an open neighbourhood $U$ in $M$, $f^{−1}(L)\cap U$ is a codimension $k$ submanifold of $U$ (by the Regular Value Theorem). It follows that $f^{−1}(L)$ is a codimension $k$ submanifold of $M$.

my doubts are:

1) why $0$ is a regular value. My opinion is that the preimage of $0$ via the projection and the diffeomorphism in the middle is (in the non trivial case) a point of $f(M)\cap L$, but I don't see why by transversality, the rank of $f$ in a preimage of such point should be maximal.

2) where does the $U$ come from and how the Regular Value Thm is applied. In my opinion I'd conclude that the preimage of such $0$ is a submanifold, but I don't know how to prove the more general assertion in the proof (why in the proof we go from $f^{-1}$ of a point to $f^{-1}(L)$?).

thanks in advance for any help!


1) The rank of $\pi\circ f$, not of $f$. Look at differentials. We may identify $\pi$ with its differential. The point is in showing that $\ker \pi\cap \mathrm{im}(d_pf)=0$. Once you have this you just apply the usual dimension formulas for kernel and image of a linear operator. Remember that you can use $\mathrm{rank} (\pi \circ d_pf)\le \mathrm{min}(\mathrm{rank}\pi,\mathrm{rank}d_p f)=k$.

2) Again, take care you are taking about $\pi\circ f$. Then the point is that $(\pi\circ f)^{-1}(0)=f^{-1}(L)$ (at least locally).

  • $\begingroup$ Hi, sorry for late commenting this answer, but I was quite busy trying figure out your reasoning. I don't see how to prove that $\ker \pi \cap Im(d_pf) = 0$ could you give me some more hints? $\endgroup$ – Luigi M Aug 22 '15 at 16:51
  • $\begingroup$ When you say that $L\cap V\simeq \mathbb R^{n-k} \cap V^\prime$ you get also, looking at tangent spaces, that $T_qL\simeq \mathbb R^{n-k}$, with ${\mathbb R}^{n-k}$ given by the first coordinates. This coincides precisely with the kernel of $\pi$, since $\pi$ projects on the last $k$ coordinates. On the other hand transversality implies $\mathrm{im}d_p f\oplus T_qL=\mathbb R^k$, and I think this is it. $\endgroup$ – N. Ciccoli Aug 23 '15 at 8:49
  • $\begingroup$ Nope, transversality means $\mathbb{R}^n = Im d_pf + T_qL$ i.e no direct sum. Anyway I managed to work out what I needed thanks to your hints :) $\endgroup$ – Luigi M Aug 24 '15 at 14:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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