18.07.2014 Views

John Stillwell - Naive Lie Theory.pdf - Index of

John Stillwell - Naive Lie Theory.pdf - Index of

John Stillwell - Naive Lie Theory.pdf - Index of

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

170 8 Topology<br />

Pick, say, the leftmost <strong>of</strong> the two intervals [0,1/2] and [1/2,1] not contained<br />

in a finite union <strong>of</strong> U i and divide it into halves similarly. By the same<br />

argument, one <strong>of</strong> the new subintervals is not contained in a finite union <strong>of</strong><br />

the U i , and so on.<br />

By repeating this argument indefinitely, we get an infinite sequence <strong>of</strong><br />

intervals [0,1]=I 1 ⊃ I 2 ⊃ I 3 ⊃···. Each I n+1 is half the length <strong>of</strong> I n<br />

and none <strong>of</strong> them is contained in the union <strong>of</strong> finitely many U i .Butthere<br />

is a single point P in all the I n (namely the common limit <strong>of</strong> their left and<br />

right endpoints), and P ∈ [0,1] so P is in some U j .<br />

This is a contradiction, because a sufficiently small I n containing P is<br />

contained in U j ,sinceU j is open. So in fact [0,1] is contained in the union<br />

<strong>of</strong> finitely many U i .<br />

□<br />

The general definition <strong>of</strong> compactness motivated by this theorem is the<br />

following. A set K is called compact if, for any collection <strong>of</strong> open sets<br />

O i whose union contains K, there is a finite subcollection O 1 ,O 2 ,...,O m<br />

whose union contains K . The collection <strong>of</strong> sets O i issaidtobean“open<br />

cover” <strong>of</strong> K , and the subcollection O 1 ,O 2 ,...,O m is said to be a “finite<br />

subcover,” so the defining property <strong>of</strong> compactness is <strong>of</strong>ten expressed as<br />

“any open cover contains a finite subcover.”<br />

The argument used to prove the Heine–Borel theorem is known as the<br />

“bisection argument,” and it easily generalizes to a “2 k -section argument”<br />

in R k , proving that any closed bounded set has the finite subcover property.<br />

For example, given a closed, bounded set K in R 2 ,wetakeasquare<br />

that contains K and consider the subsets <strong>of</strong> K obtained by dividing the<br />

square into four equal subsquares, then dividing the subsquares, and so on.<br />

If K has no finite subcover, then the same is true <strong>of</strong> a nested sequence <strong>of</strong><br />

subsets with a single common point P, which leads to a contradiction as in<br />

the pro<strong>of</strong> for [0,1].<br />

Exercises<br />

The bisection argument is also effective in another classical theorem about the<br />

unit interval: the Bolzano–Weierstrass theorem, which states that any infinite set<br />

<strong>of</strong> points {P 1 ,P 2 ,P 3 ,...} in [0,1] has a limit point.<br />

8.4.1 Given an infinite set <strong>of</strong> points {P 1 ,P 2 ,P 3 ,...} in [0,1], conclude that at least<br />

one <strong>of</strong> the subintervals [0,1/2], [1/2,1] contains infinitely many <strong>of</strong> the P i .<br />

8.4.2 (Bolzano–Weierstrass). By repeated bisection, show that there is a point P<br />

in [0,1], every neighborhood <strong>of</strong> which contains some <strong>of</strong> the points P i .

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!