0

Prove that $$S=\{(x_1,x_2,...,x_{n+1})\in \mathbb{R}^{n+1}:x_1^{2}+x_2^{2}+...+x_{n+1}^2=1\}$$ is connected in $\mathbb{R}^{n+1}$

Definition:- Let $U \subset \mathbb{R}^{n+1}$. If for every pair of points $p$ & $q$ on $U$ , there is continuous function $\alpha:[a,b] \to U$ such that $\alpha (a)=p$ and $\alpha (b)=q$ then $U$ is called connected set in $\mathbb{R}^{n+1}$.

Please help me to prove this.

Yanko
  • 13,758
  • 3
    Look at $\alpha(t)=((1-t)p+tq)/|(1-t)p+tq|$ for points that are not antipodes. When they are antipodes, join them both to a third. –  Apr 18 '18 at 17:18
  • 1
    @totoro You mean $(1-t)p + tq$ instead of $(1-t)a + tb$ ? – Kroki Apr 18 '18 at 17:19
  • 2
    Show that $\mathbb R^{n+1}\setminus{0}$ is connected (this is not hard) and define the continuous map $x \mapsto x/|x|$, so it is the image of a connected set. – Andres Mejia Apr 18 '18 at 17:26
  • 1
    @AndresMejia Is showing that $\mathbb{R}^{n+1}/{0}$ is connected really easier than $S^{n}$? – Yanko Apr 18 '18 at 17:30
  • I m not getting any hint sorry. – user539977 Apr 18 '18 at 17:32
  • @Yanko I think so. Straight line homotopies suffice so if you don’t have the idea present in the first hint, you can do this, and then normalize. It’s just the same thing in two steps. So formally, no. Conceptually, I think so. – Andres Mejia Apr 18 '18 at 17:41
  • Is anyone here who can give me a continuous function satisfying the properties? – user539977 Apr 18 '18 at 17:44
  • 1
    Be aware that to most mathematicians, your definition is actually known as "path connected" and plain "connected" means something slightly more general. – Daniel Schepler Apr 18 '18 at 17:44

3 Answers3

1

Let $p$ and $q$ be arbitrary points on $S$. Write $p=(p_1,p_2,\ldots, p_n)$ and $q=(q_1,q_2,\ldots, q_n)$. So $\sum_{i=1}^n p^2_i = \sum_{i=1}^n q^2_i = 1$.

Let $\alpha: [0,1] \mapsto \mathbb{R}^n$ be the following function:

$\alpha(t) = y(t)$, where $y(t)$ is defined $y_i(t) = \sqrt{tq^2_i + (1-t)p^2_i}$ for each $i=1,\ldots, n$. As $\sum_i y^2_i(t)$ is 1 for all such $t \in [0,1]$, it follows that $\alpha$ is indeed a function from $[0,1]$ to $S$. Furthermore, $\alpha$ is clearly continuous. Finally, $\alpha(0) = p$ while $\alpha(1) = q$.

Mike
  • 20,434
1

The accepted answer given by Mike looks great, but upon closer examination you have to ask how to define $y_i(t)$ when, say, $p_i > 0$ and $q_i < 0$.

Here is a simple proof that, like Mike's idea, only uses the square root function; it just applies the technique found here

Prove that the unit circle is path-connected

Sketch:

We can use the mappings

$\tag 1 t \mapsto \left(\,(1-t)x_0,\, \pm\sqrt{r^2 - [(1-t)x_0]^2}\,\right) \text{ with } 0 \lt r \le 1$

to continuously connect (push), in turn, all the $x_k$ coordinates, for $1 \le k \le n$, to $0$. After this you are positioned at either $(0,0,\dots,-1)$ or $(0,0,\dots,+1)$. But using

$\tag 2 t \mapsto \left(\,\sqrt{1-t^2},\, t\,\right)$

we can ultimately connect any point on the $n\text{-sphere}$ of $\mathbb R^{n+1}$ to $(0,0,\dots,+1)$.

CopyPasteIt
  • 11,366
1

Yes @CopyPasteIt you are right. Implicitly assumed in my argument is that $p_iq_i$ is nonnegative for each $i$. And this clearly is not the case. We can get around this though: Let $S^+$ be the set of $i$ s.t. both $p_i$ and $q_i$ $\ge 0$ and let $S^-$ be the set of $i$ s.t. both $p_i$ and $q_i$ $\le 0$. Two cases:

Case 1: $|S^+ \cup S^-| = a \ge 1$. Then let $m$ be the point s.t. $m_i = \frac{1}{\sqrt{a}}$ if $i \in S^+$; $m_i = -\frac{1}{\sqrt{a}}$ if $i \in S^-$, and 0 otherwise. Use my line of reasoning in my previous Answer about to conclude that $p$ is connected to $m$ and $m$ is connected to $q$.

Case 2: $|S^+ \cup S^-|$ is empty. Then let $i_1$ and $i_2$ be two distinct integers. Let $m$ be the vector such that $m_{i_1} = $sign$(q_{i_1})(\sqrt{2})^{-1}$ and $m_{i_2} = $sign$(p_{i_2})(\sqrt{2})^{-1}$ and $m_i = 0$ for every other $i$. Use the line of reasoning in Case 1 to conclude there is a path from $q$ to $m$, and then from $p$ to $m$.

Mike
  • 20,434