0

I recently trying to solve the questions from John Lee's Smooth Manifold. And for the question 2.5, I gave this proof. But I felt something is not right in my proof, is there any modification that I need to fix?

Let $P:\mathbb{R}^{n+1}\setminus \{0\}\rightarrow\mathbb{R}^{k+1}\setminus \{0\}$ be a smooth function, and suppose that for some $d\in\mathbb{Z}$, $P(\lambda x)=\lambda^d P(x)$ for all $\lambda \in \mathbb{R}\setminus \{0\}$ and $x\in \mathbb{R}^{n+1}\setminus \{0\}$. Show that the map $\tilde{P}:\mathbb{RP}^n\rightarrow\mathbb{RP}^k$ defined by $\tilde{P}([x])=[P(x)]$ is well defined and smooth.

Below is my proof.
Let $x,y\in [x]$, then $y=tx$ for $t\in \mathbb{R},t\neq 0$. Then \begin{equation*} \begin{aligned} &\tilde{P}([x])=[P(x)]\\ &\tilde{P}([y])=[P(y)]=[P(tx)]=[t^dP(x)]=[P(x)]\\ \end{aligned} \end{equation*} Hence $\tilde{P}$ is well defined.
For the next step, note that there are natural injection from $\mathbb{R}^{n+1}$ to $\mathbb{RP}^n$ and projection from $\mathbb{R}^{k+1}$ to $\mathbb{RP}^k$, i.e, \begin{equation*} \begin{aligned} i:\mathbb{RP}^n\rightarrow\mathbb{R}^{n+1}\text{ by }i([x])=x\\ \pi:\mathbb{R}^{k+1}\rightarrow \mathbb{RP}^k\text{ by }\pi(x)=[x], \end{aligned} \end{equation*} And since they are injection and projection, they are both smooth functions locally around each point of x. Then by the well define of $\tilde{P}$, we obtain \begin{equation*} \tilde{P}([x])=(\pi\circ P\circ i)([x]) \end{equation*} Then let $[x]\in\mathbb{RP}^n$, there is $x\in \mathbb{R}^{n+1}\setminus \{0\}$ such that $\pi_1(x)=[x]$. By smoothness of $i$, there are two smooth charts $(U',\varphi'),(U,\varphi)$ containing $x,[x]$ such that $\varphi'\circ i \circ (\varphi)^{-1}$ is smooth. Similarly, there are two charts $(V',\psi'),(V,\psi)$ containing $P(x),[P(x)]$ such that $\psi\circ\pi\circ(\psi')^{-1}$ is smooth. Hence for $x\in \varphi(U\cap \tilde{P}^{-1}(V))$, we have \begin{equation*} \begin{aligned} \psi\circ \tilde{P}\circ \varphi^{-1}(x)&=\psi\circ (\pi\circ P\circ i)\circ \varphi^{-1}(x)\\ &=(\psi\circ \pi \circ (\psi')^{-1})\circ (\psi'\circ P\circ (\varphi')^{-1})\circ (\varphi' \circ i\circ \varphi^{-1})(x)\\ \end{aligned} \end{equation*} The first and the third part is smooth, then the second part is smooth since $P$ is smooth. Hence $\tilde{P}$ is smooth.

Any suggestion would be greatly appreciated.

Paul Frost
  • 87,968
  • 1
    Your $i:\Bbb{RP}^n\to\Bbb{R}^{n+1}$ is not well-defined. Because for example, if $\xi=[x]$, then $\xi=[2x]$ as well, so do you send $\xi$ to $x$ or $2x$ (or $-x$ or $3.14x$, or any other multiple for that matter)? – peek-a-boo May 23 '25 at 20:21
  • Also, when it comes to dealing with quotients, this will be very helpful. If you’re doing these sorts of things “by hand” then it’s morally sort of the wrong way to go about things, and you’re expending too much effort (perhaps when first dealing with projective spaces, it may be good to go through this repetition, but very quickly things get annoying in this manner). – peek-a-boo May 23 '25 at 20:27
  • Just check smoothness in a coordinate patch, then it will be obvious. – Chris May 23 '25 at 20:39
  • @peek-a-boo thank you so much for your comments! I found that the first theorem in your common is super useful, then I just solve the question by proving $/tilde{p}\circ\pi_1=\pi_2\circ p$. And then use the theorem. – mathcal-K May 23 '25 at 21:26

0 Answers0