4

Prove that $N(Z;Y)$ is trivial if and only if there exists a set of $k$ independent global defining functions $g_1,…,g_k$ for $Z$ on some set $U$ in $Y$. That is $Z=\{y∈U:g_1 (y)=0,…,g_k (y)=0\}$

Local submersion theorem: Suppose that $f:X\to Y$ is submersion at $x$ and $y=f(x)$. Then there exist local coordinate around $x$ and $y$ such that $f(x_1,...x_k)=(x_1,...,x_k)$.

$\epsilon$ neighborhood theorem: For a compact boundaryless manifold $Y$ in $R^M$ and $\epsilon >0$, let $Y^\epsilon$ be the open set of points in $R^k$ with distance less than $\epsilon$ from $Y$. If $\epsilon$ is sufficiently small then each point $w\in Y^\epsilon$ possesses a unique closest point in $Y$, denoted $\pi(w)$. Moreover the map $\pi : Y^\epsilon \to Y$ is a submersion.

Tubular neighborhood theorem there exist a diffeomorphism from an open neighborhood $Z$ in $N(Z;Y)$ onto an open neighborhood of $Z$ in $Y$

Trivial normal bundle proposition: Let $Z$ be a submanifold of codimension $k$ in $Y$, then $N(Z;Y)$ is trivial if there exists a diffeomorphism $\phi :N(Z;Y) \to Z\times R^k$ that restrict to a linear isomorphism $N_z (Z;Y)\to {z}\times R^k$ for each point $z\in Z$.

Here is what I got so far.

=>

Assume that $N(Z;Y)$ , the normal bundle of $Z$ in $Y$, is trivial, then by Tubular neighborhood theorem, there exist a diffeomorphism from an open neighborhood $Z$ in $N(Z;Y)$ onto an open neighborhood of $Z$ in $Y$. Let $U$ be an open neighborhood in $Y$, then $g:U \to R^k $ is submersion, thus $U$ can be cut out by many independent functions $g_1, \dots, g_k$

I feel like I'm saying something really stupid here.

<=

Assume that there exists a submersion $g:U \to R^k $ with $g^{-1} (0)=Z$. For this part, the book give me a hint that for each $z\in Z$ the transpose map $dg_z ^t:R^k \to T_z(Y)$ carries $R^k$ isomorphically onto the orthogonal complement of $T_z(Z) $ in $T_z(Y)$. But I don't know how they get this.

  • 1
    You need to define your terms more clearly. $N(Z;Y)$ is the normal bundle of $Z$ in $Y$? Independent defining functions are ones whose differentials are linearly independent at each point? And most crucially, what is the map $g$ in your forward proof? – Jack Lee Oct 19 '14 at 18:17
  • independent functions is one of the properties of the local submersion theorem, that say any submanifold can be cut out by many independent function. Maybe I understand it incorrectly, but I think $g$ is the projection of $U$ down to $R^k$, which is submersion by the $\epsilon$ neighborhood theorem. – Diane Vanderwaif Oct 19 '14 at 18:46
  • 1
    Just the fact that the terms are used in the statements of some theorems does not imply that everyone here will understand exactly what they mean. In the book you're reading, these terms were probably defined before the theorem statements. You'll probably get better results if you state the definitions clearly, and give the statements of the theorems that you're using. Different authors might have slightly different theorems that they call the "local submersion theorem" or the "$\epsilon$ neighborhood theorem." – Jack Lee Oct 19 '14 at 18:53
  • For your very last question, it's standard linear algebra. If $A$ is an $m\times n$ matrix, the column space is the orthogonal complement of the kernel of $A^\top$. – Ted Shifrin Oct 19 '14 at 23:43

1 Answers1

0

Let $S$ be an embedded submanifold of $\mathbb{R}^n$ of codimension $k$. Let $NS$ denote the normal bundle and $\pi_{NS}:NS \rightarrow S$ the projection.

Suppose first that there is a smooth defining function $\Phi:U \rightarrow \mathbb{R}^k$ on a neighborhood $U$ of $S$ in $\mathbb{R}^n$, say $S = \Phi^{-1}(c)$. Consider the map $\Psi: NS \rightarrow S \times \mathbb{R}^k$ (identifying $T_c\mathbb{R}^k$ with $\mathbb{R}^k$) given by $\Psi(v) = (p, d\Phi_p v)$ for $v \in N_pS$ and $p \in S$. If $\delta: NS \rightarrow NS \times NS$ is the diagonal map, then $\Psi = (\pi_S \times d\Phi|_{NS}) \circ \delta$, so $\Psi$ is smooth. Note that $\pi_S \circ \Psi = \pi_{NS}$, where $\pi_S:S \times \mathbb{R}^k \rightarrow S$ is the projection, so $\Psi$ is a smooth bundle homomorphism over $S$. For $p \in S$, $T_pS = \text{Ker}(d\Phi_p)$, so $d\Phi_p|_{N_pS} = (d\Phi|_{NS})_p:N_pS \rightarrow \mathbb{R}^k$ is a linear isomorphism. The map $\Gamma: S \times \mathbb{R}^k \rightarrow NS$ defined by $\Gamma(p,w) = (d\Phi|_{NS})_p^{-1} w$ is thus an inverse for $\Psi$. We now show that $\Gamma$ is smooth. Let $(E_1, \ldots, E_k)$ be a smooth local frame for $NS$ on an open set $V$ in $S$. Then $d\Phi|_{NS} \circ E_i = A_i^j \frac{\partial}{\partial x^i}\big|_c$ for some smooth functions $A^j_i$ on $V$, namely \begin{equation*} A^j_i(p) = \big((d \Phi|_{NS})_p \circ E_i|_p \big)(x^j) = E_i|_p(\Phi^j) = (E_i \Phi^j)(p) \end{equation*} for $p \in V$. Since $(d\Phi|_{NS})_p$ is an isomorphism for $p \in V$, the matrix $A^j_i$ is (smoothly) invertible. Let $B$ denote the inverse matrix. Then, for $p \in V$, \begin{equation*} \Gamma(p,w) = w^i B_i^j(p) E_j|_p, \end{equation*} so $\Gamma$ is smooth. Since $\Gamma$ is also a bundle homomorphism over $S$, $NS$ is trivial.

Suppose conversely that $NS$ is trivial. By the tubular neighborhood theorem, there is a neighborhood $U$ of $S$ in $\mathbb{R}^n$ and a diffeomorphism $E:V \rightarrow U$, where $V \subset NS$ is an open set of the form \begin{equation*} V = \{(p,v) \in NS : |v|< \delta(p)\} \end{equation*} for some continuous function $\delta:M \rightarrow \mathbb{R}$. By construction, $E(p,0) = p$ for $p \in S$, so $E^{-1}(q) \in N_{\pi_{NS}(E^{-1}(q))}S$ is nonzero for any $q \in U \smallsetminus S$. Let $\Theta: NS \rightarrow S \times \mathbb{R}^k$ be a smooth bundle isomorphism, and let $\pi_k:S \times \mathbb{R}^k \rightarrow \mathbb{R}^k$ denote the projection onto $\mathbb{R}^k$. Then $\Phi = \pi_k \circ \Theta \circ E^{-1}:U \rightarrow \mathbb{R}^k$, as a composition of smooth submersions, is a smooth submersion. For $p \in S$, since $\Theta|_{N_pS}: N_pS \rightarrow \{p\} \times \mathbb{R}^k$ is a linear isomorphism and $E^{-1}(q) \in N_{\pi_{NS}(E^{-1}(q))}S$ is zero if and only if $q \in S$, $S = \Phi^{-1}(0)$.