06.08.2013 Views

11. More on locally compact Hausdorff spaces - Aarhus Universitet

11. More on locally compact Hausdorff spaces - Aarhus Universitet

11. More on locally compact Hausdorff spaces - Aarhus Universitet

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.

<str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong><br />

Klaus Thomsen matkt@imf.au.dk<br />

Institut for Matematiske Fag<br />

Det Naturvidenskabelige Fakultet<br />

<strong>Aarhus</strong> <strong>Universitet</strong><br />

November 2005<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


Additi<strong>on</strong>al material, comes from ’Noter og kommentarer’ til Rudins<br />

bog.<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


The support of a c<strong>on</strong>tinuous functi<strong>on</strong><br />

Let X be a <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> space. The support of a<br />

functi<strong>on</strong> f : X → C is the set<br />

supp f = {x ∈ X : f (x) = 0};<br />

i.e. the closure of the set of points where f is not zero.<br />

We say that f : X → C has <strong>compact</strong> support when supp f is<br />

<strong>compact</strong>.<br />

The following famous lemma, Urysohn’s lemma, guarantees a rich<br />

supply of c<strong>on</strong>tinuous functi<strong>on</strong>s of <strong>compact</strong> support <strong>on</strong> a <strong>locally</strong><br />

<strong>compact</strong> <strong>Hausdorff</strong> space.<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


Urysohn’s lemma - the statement<br />

Lemma<br />

(Urysohn’s lemma) Let X be a <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> space, K<br />

a <strong>compact</strong> subset of X and V an open subset of X such that<br />

K ⊆ V . It follows that there is a c<strong>on</strong>tinuous functi<strong>on</strong> f : X → [0, 1]<br />

of <strong>compact</strong> support such that f (k) = 1, k ∈ K, and supp f ⊆ V .<br />

If d is a metric for the topology (which in the generality of the<br />

lemma may not exist), <strong>on</strong>e can c<strong>on</strong>struct f relatively painlessly: For<br />

x ∈ X , set<br />

Dist(x, K) = inf {d(x, y) : y ∈ K} .<br />

Choose (!) ɛ > 0 so small that {x ∈ X : Dist(x, K) ≤ ɛ} ⊆ V , and<br />

set<br />

f (y) = max 1 − ɛ −1 Dist(y, K), 0 <br />

for all y ∈ X .<br />

It is not difficult to see that this functi<strong>on</strong> has the stated properties,<br />

at least when we assume, as we may, that V has <strong>compact</strong> closure.<br />

The general (n<strong>on</strong>-metric) case is a little more difficult.<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


Urysohn’s lemma - the proof<br />

Let r1, r2, r3, . . . be a list of the rati<strong>on</strong>al numbers in ]0, 1[, chosen<br />

such that r1 = 0 and r2 = 1, and without repetiti<strong>on</strong>s.<br />

It follows from a lemma from Additi<strong>on</strong>al material 1 that there are<br />

open sets V0 and V1 in X with V0 <strong>compact</strong> such that<br />

K ⊆ V1 ⊆ V1 ⊆ V0 ⊆ V0 ⊆ V .<br />

Note that V1 is then also <strong>compact</strong>. Suppose now that n ≥ 2, and<br />

that we have found open sets Vr1 , Vr2 , Vr3 , . . . , Vrn in X such that<br />

Vri is <strong>compact</strong> for all i, and<br />

ri < rj ⇒ K ⊆ Vrj<br />

⊆ Vrj ⊆ Vri ⊆ Vri ⊆ V .<br />

Let ri be the largest element of {r1, r2, . . . , rn} which is smaller<br />

than rn+1, and rj the smallest element of {r1, r2, . . . , rn} which is<br />

larger than rn+1.<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


Urysohn’s lemma - the proof<br />

We can then use the same lemma as above to find an open set<br />

in X such that<br />

Vrn+1<br />

Vrj<br />

⊆ Vrn+1 ⊆ Vrn+1 ⊆ Vri .<br />

C<strong>on</strong>tinuing in this way we obtain a collecti<strong>on</strong> {Vr } of open sets in<br />

X , <strong>on</strong>e for each rati<strong>on</strong>al number r ∈ (0, 1) such that Vr is<br />

<strong>compact</strong>, K ⊆ V1, V0 ⊆ V , and<br />

r < s ⇒ Vs ⊆ Vr . (1)<br />

When r, s ∈]0, 1[ are rati<strong>on</strong>als, define fr : X → [0, 1] and<br />

gs : X → [0, 1] such that<br />

<br />

fr (x) =<br />

r,<br />

0,<br />

if x ∈ Vr<br />

otherwise<br />

and<br />

gs(x) =<br />

<br />

1, if x ∈ Vs<br />

s, otherwise<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


Urysohn’s lemma - the proof<br />

Set<br />

and<br />

f (x) = sup fr (x)<br />

r<br />

g(x) = inf<br />

s gs(x).<br />

If x ∈ X is a point where fr (x) > gs(x), it follows from the<br />

definiti<strong>on</strong>s that x ∈ Vr and x /∈ Vs,<br />

and hence from (1) that s > r.<br />

But this is impossible since gs(x) ≥ s and fr (x) ≤ r.<br />

We see therefore that fr (x) ≤ gs(x) for all r, s, and hence that<br />

f (x) ≤ g(x) for all x ∈ X .<br />

If x ∈ X is a point such that f (x) < g(x), there are rati<strong>on</strong>al<br />

numbers r, s ∈]0, 1[ such that f (x) < r < s < g(x).<br />

The first inequality implies that x /∈ Vr and the sec<strong>on</strong>d that x ∈ Vs.<br />

This impossible by (1), and we c<strong>on</strong>clude therefore that f = g.<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


Urysohn’s lemma - the proof<br />

Since K ⊆ Vr for all r, we see that f (k) = 1 when k ∈ K and since<br />

fr (x) = 0 for all r when x /∈ V0, we see that supp f ⊆ V0 so that f<br />

has <strong>compact</strong> support.<br />

It remains now <strong>on</strong>ly to show that f is c<strong>on</strong>tinuous. To this end it<br />

suffices to show that f −1 (]α, β[) is open when α < β. Note first<br />

that<br />

and that<br />

g −1 (] − ∞, β[) = <br />

g −1<br />

r (] − ∞, β[) =<br />

This shows that g −1 (] − ∞, β[) is open.<br />

r<br />

g −1<br />

r (] − ∞, β[),<br />

⎧<br />

⎪⎨ X , when β > 1<br />

c<br />

Vr , when r < β ≤ 1, .<br />

⎪⎩<br />

∅, when β ≤ r<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


Urysohn’s lemma - the proof<br />

Similarly,<br />

where<br />

f −1 (]α, ∞[) = <br />

f −1<br />

r (]α, ∞[) =<br />

r<br />

f −1<br />

r (]α, ∞[),<br />

⎧<br />

⎪⎨ ∅, when α ≥ r,<br />

Vr ,<br />

⎪⎩<br />

X ,<br />

when 0 ≤ α < r,<br />

when α < 0.<br />

This shows that f −1 (]α, ∞[) is open. Since<br />

f −1 (]α, β[) = g −1 (] − ∞, β[) ∩ f −1 (]α, ∞[),<br />

we see that f −1 (]α, β[) is open, as desired. <br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


Partiti<strong>on</strong> of unity - the statement<br />

Theorem<br />

Let X be <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> space, K a <strong>compact</strong> subset of<br />

X , and V1, V2, . . . , Vn open subsets of X such that<br />

K ⊆ V1 ∪ V2 ∪ · · · ∪ Vn.<br />

It follows that there are c<strong>on</strong>tinuous functi<strong>on</strong>s<br />

hi : X → [0, 1], i = 1, 2, . . . , n, all of <strong>compact</strong> supports such that<br />

supp hi ⊆ Vi, for all i, and<br />

for all k ∈ K.<br />

n<br />

hi(k) = 1<br />

i=1<br />

This is often expressed by saying that ’{hi} is a partiti<strong>on</strong> of unity<br />

<strong>on</strong> K subordinate to the cover {Vi}’.<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


Partiti<strong>on</strong> of unity - the proof<br />

Let x ∈ K. It follows from the lemma from Additi<strong>on</strong>al material 1<br />

that there is an i ∈ {1, 2, . . . , n}, and an open neighborhood Wx of<br />

x such that Wx is <strong>compact</strong> and<br />

x ∈ Wx ⊆ Wx ⊆ Vi.<br />

By <strong>compact</strong>ness of K there are then points x1, x2, . . . , xm in K such<br />

that K ⊆ Wx1 ∪ Wx2 ∪ · · · ∪ Wxm. Let Hj be the uni<strong>on</strong> of the Wx k ’s<br />

that are subsets of Vj.<br />

It follows from Urysohn’s lemma that there are c<strong>on</strong>tinuous<br />

functi<strong>on</strong>s gj : X → [0, 1], j = 1, 2, . . . , n, of <strong>compact</strong> supports such<br />

that gj(y) = 1, y ∈ Hj and supp gj ⊆ Vj for all j.<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


Partiti<strong>on</strong> of unity - the proof<br />

Set<br />

h1 = g1,<br />

h2 = (1 − g1) g2,<br />

h3 = (1 − g1) (1 − g2) g3,<br />

.<br />

hn = (1 − g1) (1 − g2) . . . (1 − gn−1) gn.<br />

Then supp hi ⊆ supp gi is <strong>compact</strong> and a subset of Vi for all i.<br />

It is easy to prove, e.g. by inducti<strong>on</strong>, that<br />

h1 + h2 + · · · + hn = 1 − (1 − g1)(1 − g2)(1 − g3) . . . (1 − gn)<br />

so we see that n<br />

i=1 hi(k) = 1 when k ∈ K. <br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


The St<strong>on</strong>e-Weierstrass theorem - the classical theorem<br />

We prove here a versi<strong>on</strong> of the St<strong>on</strong>e-Weierstrass theorem. The<br />

point of departure is the classical theorem of Weierstrass:<br />

Theorem<br />

Let [a, b] be a <strong>compact</strong> interval in the real line R. Every c<strong>on</strong>tinuous<br />

functi<strong>on</strong> <strong>on</strong> [a, b] can approximated uniformly by polynomials, i.e.<br />

for every f ∈ C[a, b] and every ɛ > 0 there is a polynomial P such<br />

that<br />

sup |f (t) − P(t)| < ɛ.<br />

t∈[a,b]<br />

We assume that the reader is familiar with this classical theorem.<br />

We are <strong>on</strong>ly going to use it in the rather special case where<br />

f (t) = |t| <strong>on</strong> [−b, b], for some b > 0, so if the reader knows<br />

another way of approximating this functi<strong>on</strong> with polynomials, the<br />

following will give a self-c<strong>on</strong>tained proof of Weierstrass’ theorem.<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


The St<strong>on</strong>e-Weierstrass theorem - the general case<br />

Theorem<br />

Let X be a <strong>compact</strong> <strong>Hausdorff</strong> space. Let A ⊆ C(X ) be a<br />

subalgebra of C(X ) such that<br />

a) A is self-adjoint, i.e. f ∈ A ⇒ f ∈ A,<br />

b) A separates points <strong>on</strong> X , i.e. when x, y ∈ X and x = y, then<br />

there is an element g ∈ A such that g(x) = g(y),<br />

c) A vanishes at no point, i.e. for every x ∈ X there is an f ∈ A<br />

such that f (x) = 0.<br />

It follows that A is dense in C(X ), i.e. for all g ∈ C(X ) and ɛ > 0<br />

there is an element f ∈ A such that<br />

sup |g(x) − f (x)| < ɛ.<br />

x∈X<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


The St<strong>on</strong>e-Weierstrass theorem - the proof in general case<br />

Set AR = {f ∈ A : f (x) ∈ R ∀x ∈ X }. For every g ∈ C(X ),<br />

g = Re g + i Im g, where<br />

Re g = 1<br />

(g + g)<br />

2<br />

and<br />

Im g = 1<br />

(g − g) .<br />

2i<br />

It follows from a) that A = AR + iAR. It suffices therefore to show<br />

that a real-valued functi<strong>on</strong> f ∈ C(X ) can be approximated<br />

uniformly by elements of AR.<br />

To this end, note that AR separates points and does not not vanish<br />

at any point, since this is true for A by assumpti<strong>on</strong>.<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


The St<strong>on</strong>e-Weierstrass theorem - the proof in general case<br />

We prove first that<br />

∀ x1, x2 ∈ X , x1 = x2, ∀ r1, r2 ∈ R ∃g ∈ AR : g(xi) = ri, i = 1, 2.<br />

(2)<br />

Indeed, it follows from b) that there is a g ∈ AR such that<br />

g(x1) = g(x2). It follows from c) that there are functi<strong>on</strong>s<br />

h, k ∈ AR such that h(x1) = 0 and k(x2) = 0. Put<br />

u = gk − g(x1)k, v = gh − g(x2)h, and note that u, v ∈ AR.<br />

Since u(x1) = v(x2) = 0, while u(x2) = 0 and v(x1) = 0, we can<br />

set<br />

g = r1v r2u<br />

+<br />

v(x1) u(x2) .<br />

Then g ∈ AR and g(xi) = ri, i = 1, 2, proving (2).<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


The St<strong>on</strong>e-Weierstrass theorem - the proof in general case<br />

Next we prove that<br />

f ∈ AR ⇒ |f | ∈ AR. (3)<br />

(Recall that AR is the closure of AR in C(X ).)<br />

To prove (3), it suffices to take an ɛ > 0 and find h ∈ AR such that<br />

sup x∈X |h(x) − |f (x)|| < 2ɛ.<br />

To this end note that since f ∈ AR, there is an f1 ∈ AR such that<br />

sup ||f (x)| − |f1(x)|| ≤ sup |f (x) − f1(x)| < ɛ. (4)<br />

x∈X<br />

x∈X<br />

By Theorem 3 there is a polynomial P such that<br />

|P(t) − |t|| < ɛ (5)<br />

for all t ∈ [−f1, f1].<br />

Using 1<br />

<br />

2 P + P in place of P, we may assume that P is a<br />

real-valued polynomial. It follows then that h = P ◦ f1 is in AR<br />

because A is a subalgebra of C(X ).<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


The St<strong>on</strong>e-Weierstrass theorem - the proof in general case<br />

It follows from (5) that sup x∈X |h(x) − |f1(x)|| < ɛ. By combining<br />

this with (4), we c<strong>on</strong>clude that sup x∈X |h(x) − |f (x)|| < 2ɛ,<br />

completing the proof of (3).<br />

It follows from (3) that<br />

because max{f , g} =<br />

f , g ∈ AR ⇒ max{f , g}, min{f , g} ∈ AR, (6)<br />

f +g<br />

2<br />

+ |f −g|<br />

2<br />

and min{f , g} =<br />

f +g<br />

2<br />

− |f −g|<br />

2 .<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


The St<strong>on</strong>e-Weierstrass theorem - the proof in general case<br />

Next we prove that<br />

I) for any real-valued f ∈ C(X ), any x ∈ X and any ɛ > 0, there<br />

is a functi<strong>on</strong> gx ∈ AR such that gx(x) = f (x) and<br />

gx(y) > f (y) − ɛ for all y ∈ X .<br />

To this end, note that it follows from (2) that for every y ∈ X ,<br />

there is a hy ∈ AR such that hy (x) = f (x) and hy (y) = f (y).<br />

There is then an open neighborhood Jy of y such that<br />

hy (z) > f (z) − ɛ<br />

for all z ∈ Jy .<br />

Since X is <strong>compact</strong> there are finitely many Jyi , i = 1, 2, . . . , N, such<br />

that X ⊆ Jy1 ∪ Jy2 ∪ · · · ∪ JyN .<br />

It follows from (6) that<br />

gx = max {hy1 , hy2 , . . . , hyN }<br />

is in AR.<br />

Furthermore, by c<strong>on</strong>structi<strong>on</strong> gx(x) = f (x) and gx(y) > f (y) − ɛ<br />

for all y ∈ X , completing the proof of I).<br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>


The St<strong>on</strong>e-Weierstrass theorem - the proof in general case<br />

The proof of the theorem is then completed by using I) in the<br />

following way: When f ∈ C(X ) is real-valeud and ɛ > 0, it follows<br />

from I) that for every x ∈ X there is a functi<strong>on</strong> gx ∈ AR such that<br />

gx(x) = f (x) and gx(y) > f (y) − ɛ for all y ∈ X .<br />

For each x ∈ X there is then an open neighborhood Ix of X such<br />

that<br />

gx(y) < f (y) + ɛ<br />

for all y ∈ Ix.<br />

Since X is <strong>compact</strong> there are finitely many x1, x2, . . . , xM ∈ X such<br />

that X ⊆ Ix1 ∪ Ix2 ∪ · · · ∪ IxM .<br />

Set<br />

g = min {gx1 , gx2 , . . . , gxM } ,<br />

and note that g ∈ AR. Since<br />

f (z) − ɛ < g(z) < f (z) + ɛ<br />

for all z ∈ X , this completes the proof. <br />

Klaus Thomsen <str<strong>on</strong>g>11.</str<strong>on</strong>g> <str<strong>on</strong>g>More</str<strong>on</strong>g> <strong>on</strong> <strong>locally</strong> <strong>compact</strong> <strong>Hausdorff</strong> <strong>spaces</strong>

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

Saved successfully!

Ooh no, something went wrong!