2

Suppose that $S_1, S_2, S_3, . . .$ is a sequence of connected, compact subsets of the plane and $S_1 ⊃ S_2 ⊃ . . . .$ Is $S = \bigcap Sn$ connected? Give a proof or counterexample.

I understand that if $S_i$ was simply closed that $S$ need not be connected.

I understand intuitively that making them compact ensures that they are connected (mostly because I realize that the cases where $S$ is disconnected when $S_i$ is simply closed does not work when $S_i$ is also bounded).

I am having a really hard time proving it though.

I am not sure how to proceed.

jackson5
  • 1,674
  • HAve you looked at answers to http://math.stackexchange.com/q/1238027/4280 e.g.? Also http://math.stackexchange.com/a/1491816/4280 is good. – Henno Brandsma Jan 26 '17 at 05:18
  • The only issue is that we have not been introduced to Hausdorff spaces yet, so I don't understand fully why those answers work or if I can use them. We've dealt almost exclusively in metric spaces so far – jackson5 Jan 26 '17 at 05:24
  • you need taht compact sets are closed, which holds in Hausdorff spces, an in metric spaces too (As these are Hausdroff) and that they're normal too. – Henno Brandsma Jan 26 '17 at 06:24

1 Answers1

3

You have to suppose also that each $S_i$ is non empty.

We prove the more general :

Theorem The intersection $S$ of any decreasing sequence $(S_i)_{i\ge0}$ of nonempty compact and connected subsets of some metric space $(E,d)$ is connected.

(whose proof can be read while thinking at $E$ as beeing $\mathbb{R}^2$).

Proof Let us prove first that, for any open set $\Omega$ containing $S$, there exists an integer $i$ such that $S_i\subset\Omega$ (which implies that the inclusion is also true for any index $>i$).

Suppose the contrary ...

Then, there exists une sequence $(x_i)_{i\ge0}$ such that $x_i\in S_i-\Omega$ for all $i$.

All the $x_i$ belong to $S_0$, which is compact, and therefore we can extract a sequence $(x_{i_p})_{p\in\mathbb{N}}$ which converges to some $y\in E$.

For any $N\in\mathbb{N}$, the truncated sequence $(x_{i_p})_{p\ge N}$ has all its terms in $S_{i_N}$ thus in $S_N$. So $y\in S_N$, because $S_N$ is closed. This proves that $y\in S$.

But $E-\Omega$ is closed in $E$, and therefore $y\not\in\Omega$, a contradiction.

Now, suppose $S=(S\cap \Omega_1)\cup(S\cap\Omega_2)$, where $\Omega_1$, $\Omega_2$ are disjoint subsets of $E$, such that $S\cap\Omega_1\neq\emptyset$ and $S\cap\Omega_2\neq\emptyset$.

Put $\Omega=\Omega_1\cup\Omega_2$. We can see that $\Omega$ is an open set, which contains $S_i$ for some $i\in\mathbb{N}$. But $S_i$ beeing connected, there exists $\alpha\in\{1,2\}$ such that $S_i\cap\Omega_\alpha=\emptyset$. And so $S\cap\Omega_\alpha=\emptyset$ for the same $\alpha$. Contradiction !

Adren
  • 8,184
  • What is a connex subset? I don't think I have learned this – jackson5 Jan 29 '17 at 02:01
  • @jakson5 - Sorry, I meant "connected" ("connex" was a wrong translation from the french "connexe"). Given a topological space $X$ (say an metric space if you prefer a known special case), a subset $A$ is said to be connected if, equipped with the induced topology, is itself a connected space. I will now edit my answer to replace "connex" by "connected" ...

    And remind that a connected topological is a space $X$ such that there is no way to obtain it as the union of two disjoint nonempty open subsets.

    – Adren Jan 29 '17 at 06:36