0
$\begingroup$

I initially asked this question at math.stackexchange.com but there was no reaction, so I thought this may be a good idea to transfer it to mathoverflow.net

Let $\langle\mathrm{r}\mathscr{O},\mathord{\subseteq}\rangle$ be the complete Boolean algebra of open domains (regular open sets, these that are equal to the interior of their closure: $a=\mathrm{int}(\mathrm{cl}(a))$) of a topological space $\langle \mathbb{R}^2,\mathscr{O}\rangle$ (i.e. the Euclidean plane with the standard topology). Let $\mathcal{C}$ be the family of all element of $\mathrm{r}\mathscr{O}$ which are convex in the standard sense, i.e. for any points $p_1,p_2\in c$, the closed interval $[p_1,p_2]$ is a subset of $c$. Let for any regular open set $\mathrm{hull}(x)$ be the convex hull of $x$, which is the smallest convex set $c\in\mathrm{r}\mathscr{O}$ such that $x\subseteq c$. We define $\mathrm{hull}$ as the special case of the closure operator on the Boolean algebra: $$\mathrm{hull}(x)=\mathrm{int}\bigcap\{c\in\mathcal{C}\mid x\subseteq c\}\,.$$

My question concerns the existence of euclidean half-planes with respect to elements of the Boolean algebra. Let $H$ be the set of all half-planes in $\mathbb{R}^2$ without $0$ and $1$ ($\emptyset$ and $\mathbb{R}^2$ itself). $H$ may be defined in the following way: $$H=\{x\mid x\in\mathcal{C}\wedge-x\in\mathcal{C}\}$$ where $-$ is the complement operation, in this case for $x$ it is $\mathrm{int}\,(\mathbb{Real}^2\setminus x)$. Let $+$ be the Boolean sum in the algbera $\mathrm{r}\mathscr{O}$: $x+y=\mathrm{int}(\mathrm{cl}(x\cup y))$, $\cdot$ the Boolean product: $x\cdot y=\mathrm{int}(x\cap y)$

It seems obvious that in case we take $a,b,c\in\mathcal{C}\setminus\{0\}$ such that $a\cdot \mathrm{hull}(b+c)\neq 0$ then there is a half-plane $h\in H$ such that for all $x\in\{a,b,c\}$: $$h\cdot x\neq 0\neq -h\cdot x\,.$$ But can we replace in the above convex sets with arbitrary regular open sets? I.e. is it true that for all non-zero $x,y,z\in\mathrm{r}\mathscr{O}$ such that $x\cdot\mathrm{hull}(y+z)\neq 0$ there is a half-plane $h$ such that both it and its complement meets each one of $x,y,z$? I will be grateful for any hints of a solution.

$\endgroup$
1
$\begingroup$

It seems that if you replace $x,y,z$ by their hulls and apply the `obvious' claim to them then you get the desired half-plane $h$. Namely, if $h$ or its complement contains, say, $x$ then it surely contains $\mathop{\rm hull} x$.

$\endgroup$
  • $\begingroup$ I think this is not an answer since I am looking for a half-plane $h$ which meets every one of $x,y,z$, i.e. such that $h\cdot x\neq 0\neq-h\cdot x$ and $h\cdot y\neq 0\neq-h\cdot y$ and $h\cdot z\neq 0\neq-h\cdot z$. If you replace the sets with their convex hulls in the 'obvious' claim that you only now that there is $h$ which meets each one of the hulls, not the regions themselves. $\endgroup$ – Mad Hatter Oct 24 '14 at 8:37
  • $\begingroup$ As I told, if a half-plane contains $x$ then it contains $\mathop{rm hull} x$, due to your definition of a hull. Thus neither the obtained half-plane nor its complement contains $x$ which is what you need. The same holds for $y$ and $z$. $\endgroup$ – Ilya Bogdanov Oct 24 '14 at 11:07
  • $\begingroup$ Yes, you are right. I misunderstood your answer. Thanks a lot. $\endgroup$ – Mad Hatter Oct 24 '14 at 11:42
  • $\begingroup$ Ilya, I would like to ask you one more question. What do you think about my 'obvious' assumption? Do you think it is true? $\endgroup$ – Mad Hatter Oct 24 '14 at 12:02
  • 1
    $\begingroup$ Surely it is. If a common point $A$ of $a$ and $\mathop{\rm hull}(b+c)$ lies in $b$ then the boundary line of $h$ is a line connecting $A$ with some point of $c$; othersise $A$ lies on a line containing some points of $b$ and $c$, and this is a desired boundary line. $\endgroup$ – Ilya Bogdanov Oct 24 '14 at 14:52

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.