1

Let $A$ be a non-empty convex, unbounded set in $\mathbb R^n$. Prove that for each point $a \in A$, there is a non-zero vector $h \in \mathbb R^n$ such that $l = \{x \in \mathbb R^n \mid x=a+th,\ t\ge0 \} \subset A$.

Ashot
  • 4,753
  • 3
  • 34
  • 61
  • 3
    To get the best possible answers, you should explain what your thoughts on the problem are so far. That way, people won't tell you things you already know, and they can write answers at an appropriate level; also, people tend to be more willing to help you if you show that you've tried the problem yourself. Also, many would consider your post rude because it is a command ("Prove"), not a request for help, so please consider rewriting it. – Zev Chonoles Feb 24 '13 at 12:01
  • @ZevChonoles I am not requesting for help, I am solving a problem and I think there may be people who will be interested in it too, so they will solve it and answer. I will do in the same way to other peoples questions. Also I consider "Please help" , "Thank you" and such expressions not professional to this site. You can see my other questions. Most of them are written in the same way. They are answered and also liked by many people. – Ashot Feb 24 '13 at 12:27
  • 4
    You don't have to say please or thank you; you should however avoid the impression that you are giving commands to other users. Also, are you saying that you understand how to do this problem, but just wanted other people to have a chance to solve it too? If that is the case, I think it is inappropriate not to mention it in the question. – Zev Chonoles Feb 24 '13 at 12:31

4 Answers4

5

The claim is incorrect. Consider the following counter-example: $$A=\left\{ \left(x,y\right)\in\mathbb{R}^{2}:0<x<1\right\} \cup\left\{ \left(0,0\right)\right\}\\a=\left(0,0\right)$$ It is correct if you require $A$ to be closed or $a$ to be in $\text{rel int}A$. Since for every non-empty convex set we have $\text{rel int}A\ne\emptyset$, it implies that every unbounded convex set contains a ray (but not necessarily from each point).

Ido
  • 672
3

For $r>0$, let $$V_r=\{h\in \mathbf S^{n-1}\mid a+rh\notin A\}.$$ By convexity, $r<s$ implies $V_r\subseteq V_s$. Let $$ U_t=\bigcup_{0<r<t}V_r.$$ Then $U_t$ is open, see below.

Assume $S^{n-1}\subseteq\bigcup_{t>0}U_t$, then by compactness of $S^{n-1}$ a finite subcover suffices and in fact $S^{n-1}\subseteq U_t$ for a single (maximal among finitely many) $t$. As $A$ is unbounded, there is $b\in A$ with $|b-a|>t$. With $h=\frac{b-a}{|b-a|}$, we have $h\in U_t\subseteq V_t\subseteq V_{|b-a|}$ hence $b=a+|b-a|h\notin A$, contradiction. Therefore, there exists some $h\in S^{n-1}$ that is not in any $U_t$ and also not in any $V_t$, hence $a+th\in A$ for all $t>0$ (and trivially also for $t=0$). $_\square$


Why is $U_t$ open? Assume $h\in U_t$ and specifically $h\in V_r$ with $0<r<t$. By the convexity of $A$, there exists $k\in\mathbb R^n$ with $\langle (a+rh)-x,k\rangle >0$ for all $x\in A$. Especially, $\langle h,k\rangle>0$ when $x=a$ because $a\in A$. Select $s$ with $r<s<t$. Then $(s-r)\langle h,k\rangle>0$ and for all $h'$ sufficiently close to $h$, we have $(s-r)\langle h,k\rangle +s\langle h'-h,k\rangle>0$. Hence for such $h'$ and all $x\in A$ $$\begin{align}\langle (a+sh')-x,k\rangle &=\langle (a+rh)-x,k\rangle+(s-r)\langle h,k\rangle +s\langle h'-h,k\rangle >0,\end{align}$$ especially $h'\in V_s\subseteq U_t$ as was to be shown.

Hans
  • 9,804
  • 1
    You assume a strict separation property which is only true if $A$ is closed. – Ido Aug 27 '14 at 14:55
  • By $h\in \mathbb R^{n-1}$, I suppose you mean the set $h$ is in is isomorphic to $\mathbb R^{n-1}$. Would it not be better to just set $V_r={h\in \mathbf S^{n-1}\mid a+rh\notin A}$ where $\mathbf S^{n-1}$ is the unit sphere in $\mathbb R^n$? – Hans Aug 19 '18 at 18:42
2

Of course if A is closed the result holds.

In fact if $x_0\in A$, then for each natural number n, it is possible to chose a sequence $l_n$ of line segments of length greater then $n$, with one endpoint $x_0$. If the sequence of points on $l_n$ which lies on the unit sphere around $x_0$ converges to $x$, then the ray staring at $x_0$ passing through $x$ is entirely inside A.

What if A is not closed ? Can we have at least one point in $A$ from where we can have an infinite ray contained inside $A$

  • Good sketch of a proof. But the assertion that "then the ray staring at $x_0$ passing through $x$ is entirely inside $A$" needs proof. – Hans Aug 19 '18 at 20:09
0

Hint:

  • It suffices to consider vectors $h\in S^{n-1}$, the unit sphere of $\mathbb{R}^n$.
  • $S^{n-1}$ is compact.
  • Continuous functions from a compact space to $\mathbb{R}$ are bounded.
Zev Chonoles
  • 129,973