1
$\begingroup$

Here is Example 2, Sec. 28, in the book Topology by James R. Munkres, 2nd edition:

... Consider the minimal uncountable well-ordered set $S_\Omega$, in the order topology. The space $S_\Omega$ is not compact, since it has no largest element. However, it is limit point compact. Let $A$ be an infinite subset of $S_\Omega$. Choose a subset $B$ of $A$ that is countably infinite. Being countable, the set $B$ has an upper bound $b$ in $S_\Omega$; then $B$ is a subset of the interval $\left[ a_0, b \right]$ of $S_\Omega$, where $a_0$ is the smallest element of $S_\Omega$. Since $S_\Omega$ has the least upper bound property, the interval $\left[ a_0, b \right]$ is compact. By the preceding theorem, $B$ has a limit point $x$ in $\left[ a_0, b \right]$. The point $x$ is also a limit point of $A$. Thus $S_\Omega$ is limit point compact.

Now my question is related to the following part of this example.

The space $S_\Omega$ is not compact, since it has no largest element.

In general, we can prove the following:

Let $X$ be a simply ordered set, in the order topology. If $X$ has no smallest element or if $X$ has no largest element, then $X$ is not compact. In other words, if a simply ordered set $X$ in the ordered topology is compact, then $X$ must have a smallest element and a largest element.

Proof:

Suppose that $X$ has a smallest element $a$ but that $X$ has no largest element. Then the collection $$ \mathscr{A} \colon= \left\{ \ [a, x) \ \colon \ x \in X, x > a \ \right\} $$ is an open covering of $X$. If there were a finite sub-collection of $\mathscr{A}$ that also covered $X$, say, the collection $$ \mathscr{A}_0 \colon= \left\{ \ \left[ a, x_1 \right), \ldots, \left[ a, x_n \right) \ \right\},$$ then we would have $$ X = \bigcup_{i=1}^n \left[ a, x_i \right) = \left[a, x_* \right), \tag{1}$$ where $$ x_* \colon= \max \left\{ \ x_1, \ldots, x_n \ \right\}. $$ But this would yield a contradiction because $x_*$ is an element of $X$, by our construction, but from (1) we would obtain $x_* \not\in X$.

Hence $X$ cannot be compact if $X$ has no largest element even if $X$ does have a smallest element.

By analogous reasoning we can also show that if $X$ has no smallest element, then $X$ cannot be compact, even if $X$ does have a largest element.

Am I right?

Now what about the converse? That is, if $X$ is a simply ordered set having a smallest element and a largest element, then can we prove that $X$ is always compact?

$\endgroup$
  • $\begingroup$ @JoséCarlosSantos what have you edited in my post? $\endgroup$ – Saaqib Mahmood Feb 24 at 11:52
  • $\begingroup$ I have added the proof-verification tag. $\endgroup$ – José Carlos Santos Feb 24 at 11:54
  • 1
    $\begingroup$ The empty set might be considered a counterexample. It is a compact space, and it has no largest or smallest element. I don't know, does the definition of an ordered set require the set to me nonempty? $\endgroup$ – bof Feb 24 at 12:08
2
$\begingroup$

Your proof is correct.

On the other hand, $[-1,1]\setminus\{0\}$ is not compact with respect to the order topology induced by its usual order, in spite of the fact that it has both a smallest and a largest element.

$\endgroup$
1
$\begingroup$

If $X$ is a (LOTS) linearly ordered topological space and $X$ is compact, then for any subset of $A$, $\sup A$ exists in $X$. This in fact characterises compact LOTSes. In particular $\sup X = \max X$ and and $\sup \emptyset = \min X$ exist.

A corollary to the above is that an order complete LOTS $X$ is compact iff it has a maximum and a minimum. José's example shows we cannot completeness; endpoints alone are not enough,there shouldn't be any "holes".

As to your proof, if $X$ has a minimum $a$ and if $X$ has no maximum then your family is indeed an open cover without a finite subcover, contradicting the compactness. So for the case at hand this suffices. But I also wanted to point out the general result for LOTSes.

$\endgroup$
  • $\begingroup$ What if $X=\emptyset$? Does the definition of a LOTS require it to be nonempty? $\endgroup$ – bof Feb 24 at 12:09
  • $\begingroup$ @bof of course, all my spaces are non-empty. $\endgroup$ – Henno Brandsma Feb 24 at 12:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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