This is actually just a linear algebra problem, but this text is taken from Guillemin and Pollack, Differential Topology on page 100.

Real Problem: Let $f: X \to Y$ be smooth, $S = f^{-1}(Z)$ where $T_x(S) = (df)^{-1}(T_z(Z))$, $f(x) = z$ and $N(S;X)$ is the orthogonal complement to $T_x(S)$. Then since $T_x(S)$ contains the whole kernel of $df_x$, $df_x$ maps $N(S;X)$ isomorphically into $df N(S;X)$.

If anyone is confused as to why $T_x(S)$ contains the full kernel, it is because $f^{-1}(z) \subset f^{-1}(Z)$ and it is shown previously that $\ker df = T_x(f^{-1}(z))$. Also just a side note, does the tangent space come equipped with an inner product? How do they talk about orthogonal complements without it?

What this problem actually say in Linear Algebra: We linear map $g: V \to W$ where $V = S \oplus S^{\perp}$. $\ker g \subset S $, then $g : S^{\perp} \to g(S^\perp)$ isomorphically.

I can't quite pinpoint why $g|_{S^{\perp}}$ is isomorphism. Is it because $\ker g|_{S^{\perp}} = 0$ since the whole kernel is contained in $S$, but $S \cap S^{\perp} = \{0\}?$

Sorry this post appeared longer than it should have.

  • 1
    $\begingroup$ A typo in the translation to linear algebra, $\text{ker } g \subset V$ should be $\text{ker } g \subset S$. $\endgroup$ – hunter Aug 10 '18 at 3:18
  • $\begingroup$ As to the inner product, if I recall the setup in Guillemin and Pollack's textbook correctly, each tangent space lives in some giant $\mathbb{R}^N$ and then you can just restrict the standard inner product to it. $\endgroup$ – hunter Aug 10 '18 at 3:25
  • $\begingroup$ @hunter, I think they defined it as the image of the chart $\phi$. $\endgroup$ – Hawk Aug 10 '18 at 3:34
  • $\begingroup$ Fixed thank you. $\endgroup$ – Hawk Aug 10 '18 at 3:38

To say that $g|_{S^\perp}: S^\perp \to g(S^\perp)$ is an isomorphism is just to say that it's injective, since it's onto its image by definition.

But if $v \in \ker(g|_{S^\perp})$, then $v \in S \cap S^\perp = \{0\}$.

  • $\begingroup$ So I was right about it? $\endgroup$ – Hawk Aug 10 '18 at 3:33
  • $\begingroup$ Yes, you were. It is simple. $\endgroup$ – amsmath Aug 10 '18 at 3:35

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.