Let $d\in\mathbb N$, $x\in M\subseteq\mathbb R^d$ and $\psi^{(i)}:\Omega_i\to\psi^{(i)}(\Omega_i)$ be a diffeomorphism with $x\in\Omega_i$, $$\psi^{(1)}(M\cap\Omega_1)=\psi^{(1)}(\Omega_1)\cap(\mathbb R^k\times\{0\})\tag1,$$ $$\psi^{(2)}(M\cap\Omega_2)=\psi^{(2)}(\Omega_2)\cap(\mathbb H^k\times\{0\})\tag2,$$ where $\mathbb H^k:=\{u\in\mathbb R^k:u_k\ge0\}$, and $\psi^{(2)}_k(x)=0$.

I want to conclude that both $\psi^{(i)}$ cannot exist simultaneously.

Let $\Omega:=\Omega_1\cap\Omega_2$. One argument that I found started with observing that $\phi^{(1)}(M\cap\Omega)$ is open (in $\mathbb R^d$). But I don't get that. Why is that necessarily the case? By definition of a diffeomorphism, all we know should be that $\Omega_i$ and $\psi^{(i)}(\Omega_i)$ are open.

  • $\begingroup$ What is $M$ here? $\endgroup$ Jul 5 '20 at 18:36
  • $\begingroup$ @mathcounterexamples.net An arbitrary subset of $\mathbb R^d$. $\endgroup$
    – 0xbadf00d
    Jul 5 '20 at 18:50
  • $\begingroup$ I think that $M$ should be open in order to be able to define a diffeomorphism. Or a manifold? $\endgroup$ Jul 5 '20 at 18:57
  • $\begingroup$ @mathcounterexamples.net The diffeomorphisms are defined on the open subsets $\Omega_i$. So there's no problem. $\endgroup$
    – 0xbadf00d
    Jul 5 '20 at 19:13
  • $\begingroup$ Just to be sure: When I say that $\psi^{(i)}:\Omega_i\to\psi^{(i)}(\Omega_i)$ is a diffeomorphism, this implicitly involves that $\Omega_i$ and $\psi^{(i)}(\Omega_i)$ are open subsets of $\mathbb R^d$. $\endgroup$
    – 0xbadf00d
    Jul 5 '20 at 19:34

This has been asked and answered before on MSE.

Why can there not be a diffeomorphism $\phi$ from an open neighborhood of a boundary point—say the origin—in $\Bbb H^k$ to an open neighborhood of a point in $\Bbb R^k$? View $\Bbb H^k$ as a subset of $\Bbb R^k$, and suppose $\phi(0)=a\in\Bbb R^k$. By the inverse function theorem, $\phi^{-1}$ maps some open neighborhood $U$ of $a$ onto a open neighborhood of $0\in\Bbb R^k$. (This follows from the fact that $d\phi^{-1}(a)$ is nonsingular.) So its image cannot be contained in $\Bbb H^k$.

  • $\begingroup$ Thank you for your answer. Do you know why we can conclude that $\phi^{(1)}(M\cap\Omega)$ is open? $\endgroup$
    – 0xbadf00d
    Jul 6 '20 at 4:26
  • 1
    $\begingroup$ You need to explain your notation a lot more carefully. Your intersection with $\Bbb R^k$ comes out of nowhere. $\endgroup$ Jul 6 '20 at 4:38
  • $\begingroup$ Sorry, I don't understand what you mean. I'm intersecting with $\mathbb R^k\times\{0\}\subseteq\mathbb R^d$, where $0\in\mathbb R^{d-k}$. $\endgroup$
    – 0xbadf00d
    Jul 6 '20 at 7:29
  • $\begingroup$ I guess I've figured it out. If we denote the canonical projection of $\mathbb R^d\cong\mathbb R^k\times\mathbb R^{d-k}$ to $\mathbb R^k$ by $\pi$ and $N_1:=\Omega_1\cap M$, then $U_1:=\pi(\psi^{(1)}(\Omega_1)\cap(\mathbb R^k\times\{0\}))$ is open and $\phi_1:=\pi\circ\left.\psi^{(1)}\right|_{\Omega_1\cap M}$ is a homeomorphism from $N_1:=\Omega_1\cap M$ onto $U_1$. Since $\Omega:=\Omega_1\cap\Omega_2$ is open, $\Omega\cap M$ is $N_1$-open and so $\phi_1(\Omega\cap M)=\pi(\psi^{(1)}(\Omega\cap M))$ is $U_1$-open and hence (since $U_1$ is $\mathbb R^k$-open) open. $\endgroup$
    – 0xbadf00d
    Jul 6 '20 at 8:46
  • $\begingroup$ Please take a look at the supplementary answer I provided. Do you agree? $\endgroup$
    – 0xbadf00d
    Jul 12 '20 at 18:32

Ted Shifrin's answer, but let me rephrase the claim in a version which is more useful for my purposes and adapt Ted Shifrin's arguments for the proof:

What I was trying to show is that if $M$ is a $k$-dimensional embedded $C^1$-submanifold of $\mathbb R^d$ with boundary$^1$ and $(\Omega_i,\phi_i)$ is a $k$-dimensional $C^1$-chart$^2$ of $M$, then

  1. $\psi_2(x)\in\partial\mathbb H^k=\mathbb R^{k-1}\times\{0\}$; and
  2. $\phi_1(\Omega_1)$ is $\mathbb R^k$-open

cannot hold simultaneously.

Proof: $\Omega:=\Omega_1\cap\Omega_2$ is $M$-open and $\phi_i$ is an open map. Thus, $V_i:=\phi_i(\Omega)$ is $\mathbb H^k$-open. Assume $U_1:=\phi_1(\Omega_1)$ is $\mathbb R^k$-open. Since $U_1\subseteq\mathbb H^k$ is $\mathbb R^k$-open and $V_1\subseteq U_1$ is $\mathbb H^k$-open, $V_1$ must be $\mathbb R^k$-open. By assumption, $$f:=\phi_2\circ\phi_1^{-1}:V_1\to V_2$$ is a $C^1$-diffeomorphism from $V_1$ onto $V_2$. In particular, it is an immersion.

This will yield that $V_2$ is $\mathbb R^k$-open. In fact, let $v_2\in V_2$ so that there is a $y\in\Omega$ with $v_2=\phi_2(y)$. Let $v_1:=\phi_1(y)$. Then $f(v_1)=v_2$ and since ${\rm D}f(v_1)$ is injective, the inverse function theorem implies that there is an $\mathbb R^k$-open neighborhood $W_i\subseteq V_i$ of $v_i$ with $f(W_1)=W_2$. Since $f(W_1)\subseteq V_2$, this shows that $V_2$ is $\mathbb R^k$-open.

If now $u_2:=\phi_2(x)\in\partial\mathbb H^k$, then there is no $\mathbb R^k$-open neighborhood of $u_2$. But since $u_2\in V_2$, this implies that $V_2$ cannot be open.

$^1$ i.e. each point of $M$ is locally $C^1$-diffeomorphic to $\mathbb H^k:=\mathbb R^{k-1}\times[0,\infty)$.

If $E_i$ is a $\mathbb R$-Banach space and $B_i\subseteq E_i$, then $f:B_1\to E_2$ is called $C^1$-differentiable if $f=\left.\tilde f\right|_{B_1}$ for some $E_1$-open neighborhood $\Omega_1$ of $B_1$ and some $\tilde f\in C^1(\Omega_1,E_2)$ and $g:B_1\to B_2$ is called $C^1$-diffeomorphism if $g$ is a homeomorphism from $B_1$ onto $B_2$ and $g$ and $g^{-1}$ are $C^1$-differentiable.

$^2$ A $k$-dimensional $C^1$-chart of $M$ is a $C^1$-diffeomorphism from an open subset of $M$ onto an open subset of $\mathbb H^k$.


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.