The University of Sydney MATH 3901

Metric Spaces

2004

Tutorial 9 PROBLEM SET 9 1. Use the least upper bound axiom to prove the nested interval property: Let (In ), where In = [an , bn ], be a nested sequence of closed (and bounded) intervals in R. Prove that there exists at least one point common to every interval. Moreover, if limn |In | = 0, where |In | denotes the length of In , then prove that there exist exactly one point common to every interval.

Solution. Since In In+1 , it follows that a1 a2 . . . and . . . b2 b1 .

We claim that, for all m, n = 1, 2, . . . , am < bn . In fact, if m > n, then am < bm bn and if m n, then am an < bn . Thus each bn is an upper bound for the set A = {a1 , a2 , . . . } of left end points. By the least upper bound axiom of R, sup A exists; say, p = sup A. Since each bn is an upper bound for A and p is the least upper bound, it follows that p bn , for all n = 1, 2, . . . . Next, since p is an upper bound for A, we see that an p, for all n = 1, 2, . . . . Hence an p bn , for all n = 1, 2, . . . . Since each In is an interval, we conclude that p In = [an bn ], for all n. Hence p is common to every interval. Suppose that limn |In | = limn (bn an ) = 0. Then for any > 0, there exists a positive integer K such that whenever n > K , bn an < . Suppose that p1 and p2 belong to every interval. If p1 = p2 , then |p1 p2 | = > 0. Since limn |In | = limn (bn an ) = 0 , there exists an interval IK = [aK , bK ] such that the length of IK is less that the distance |p1 p2 | = between p1 and p2 . Thus p1 and p2 cannot both belong to IK , a contradiction. Hence p1 = p2 , that is, there exists exactly one point common to every interval.

2 2. In R, let An = (0, n1 ), for n = 1, 2, . . . . Show that d(An ) 0 as n ,

and that
n=1

An = .

Solution. Note that (An ) is a nested sequence. We have for n = 1, 2, . . . that d(An ) = sup d(x, y ) =
x, y An

1 , n

and so d(An ) 0 as n . If there is a p common to every interval An , then 0 < p < 1/n, for all n. Since 0 < p, there is a positive integer K such that 0 < 1/K < p which implies that

p / AK , a contradiction. Hence
n=1

An = .

3.

In R, let An = [n, ), for n = 1, 2, . . . . Show that

An = .

Solution. If p is a point common to every interval An , then n < p < , for all n = 1, 2, . . . . But p is not greater than p + 1 so that p / Ap+1 , a contradiction.

Hence
n=1

An = .

4.

Give an example of a nested sequence of open intervals in R with one point common to every interval.

Solution. In R, let A1 = (1, 1), A2 = (1/2, 1/2), . . . , Ak = (1/k, 1/k ), . . . . Then (An ) is a nested sequence of open intervals in R. We have shown in

Question 3/Problem Set 1 that

Ak = {0}. Hence 0 is a point common to

every interval. 5. Let X = (0, 1) with the Euclidean metric d. Give an example of a nested sequence (An ) of non-empty closed sets in X with d(An ) 0 as n , but

An = .
n=1

Solution. Note that (X, d) is not a complete metric space. Let An = (0, n1 ], for n = 2, 3, . . . . Then (An ) is a nested sequence of non-empty closed sets in X . A similar arguments as in Question 2 shows that d(An ) 0 as n , and

An = .
n=1

3 6. Let (X, d) be a metric space. Prove that (X, d) is complete if and only if every nested sequence (An ) of nonempty closed sets in X , with diameters d(An )

tending to zero, has exactly one point in the intersection

An .

Solution. Suppose that X is complete and let (An ) be a nested sequence of nonempty closed sets in X such that diameters d(An ) 0 as n . Since each An is non-empty, we can choose a sequence (xn ) such that xn An for each n. Then (xn ) is a Cauchy sequence. In fact, let > 0. Then since lim d(An ) = 0,
n

there exists a positive integer K such that d(AK ) < . Since (An ) is a nested sequence, it follows that for any n, m > K , An AK and Am AK so that xn , xm AK which implies that d(xn , xm ) < . Hence (xn ) is a Cauchy sequence. Since X is complete (xn ) converges to a point p in X . We claim

that p
n=1

An . Suppose not, that is, suppose there is K such that p / AK .

Then p X \ AK . Since AK is closed, X \ AK is open and so there is an open ball B (p; ) such that B (p; ) AK = . But n > K implies xn AK which

implies xn / B (p; ). This is impossible, since xn p. Hence p

An .

Suppose that p1 and p2 are common to every An . If p1 = p2 , then d(p1 , p2 ) = > 0. Let > 0 be such that < and choose K such that d(AK ) < . Since p1 , p2 AK , it follows that d(p1 , p2 ) < < , a contradiction. Hence p1 = p2 . Suppose that the condition holds and let (xn ) be a Cauchy sequence in X . Set A1 = {x1 , x2 , . . . }, A2 = {x2 , x3 , . . . }, . . . . Let > 0. Since (xn ) is a Cauchy sequence, there exists K > 0 such that for all m, n > K , d(xn , xm ) < . Thus, for n > K , d(An ) < and so lim d(An ) = 0. Since d(A) = d(A), where
n

A is the closure of A, it follows that A1 A2 . . . is a (nested) sequence of nonempty closed sets whose diameters tend to zero. An = ; say p An . We claim that the Cauchy Hence by the hypothesis,
n n n

sequence (xn ) converges to p. Let > 0. Since lim d(An ) = 0, there is N > 0 such that d(AN ) < ,. Then for n > N , we see that an , p AN so that d(xn , p) < . Hence (xn ) converges to p. This completes the proof.

4 7. Prove that if X is complete and Y is isometric with X , then Y is also complete.

Solution. Suppose that X is complete and Y is isometric with X . Let f : X Y be an isometry. Let (yn ) be a Cauchy sequence in Y and let xn X be such that f (xn ) = yn , for all n. Since f is an isometry, it follows that d(xn , xm ) = d f (xn ), f (xm ) = d(yn , ym ) so that (xn ) is a Cauchy sequence in X , since (yn ) is. Since X is complete, the Cauchy sequence (xn ) converges to a limit x in X ; that is, lim d(xn , x) = 0.
n

Let y = f (x). Then y Y and d(yn , y ) = d f (xn ), f (x) = d(xn , x) since f is an isometry. Thus lim d(yn , y ) = 0; that is (yn ) converges to y in n Y . Hence Y is complete. 8. Let X = (X, d) be a metric space and CS (X ) the collection of all Cauchy sequences in X . For (xn ) and (yn ) in CS (X ), dene (xn ) (yn ) if and only if
n

lim d(xn , yn ) = 0.

Show that is an equivalence relation CS (X ). Solution. Since d(xn , xn ) = 0, for all n, we see that for all (xn ) CS (X ), (xn ) (xn ) . Next, since d(x, y ) = d(y, x) for all x, y X , it follows that (xn ) (yn ) = (yn ) (xn )

for all (xn ), (yn ) CS (X ). If (xn ) (yn ) and (yn ) (zn ), for (xn ), (yn ), (zn ) CS (X ), then
n

lim d(xn , yn ) = 0

and
n

lim d(yn , zn ) = 0.

Since d(xn , zn ) d(xn , yn ) + d(yn , zn ), it follows that

lim d(xn , zn ) lim d(xn , yn ) + lim d(yn , zn ) = 0 + 0 = 0.

Therefore (xn ) (zn ) . Hence is an equivalence relation on CS (X ).

5 9. Prove that two convergent sequences in a metric space (X, d) are equivalent if and only if they have the same limit.

Solution. Let (xn ) and (yn ) be two convergent sequences with limits x and y, respectively; that is, lim d(xn , x) = 0 and lim d(yn , y ) = 0.
n n

Suppose that (xn ) and (yn ) are equivalent. Then

lim d(xn , yn ) = 0.

Now d(x, y ) d(x, xn ) + d(xn , yn ) + d(yn , y ) and as n , we see that the right hand side tends to 0. Thus d(x, y ) 0 so that d(x, y ) = 0 and hencex = y. Now suppose that x = y . Since d(xn , yn ) d(xn , x) + d(x, y ) + d(y, yn ) = d(xn , x) + d(yn , y ), it follows that
n

lim d(xn , yn ) = 0,

and so (xn ) and (yn ) are equivalent. 10. Show that (xn ) (yn ) if and only if they are both subsequences of some Cauchy sequence (zn ). Solution. Suppose that (xn ) (yn ); that is, lim d(xn , yn ) = 0. Dene a sequence (zn ) n by xn/2 , if n is even, zn = y(n+1)/2 , if n is odd. That is, (zn ) = (y1 , x1 , y2 , x2 , . . . ). We claim that (zn ) is a Cauchy sequence. Let > 0. Then there exist positive integers K1 , K2 , K3 , such that m, n > K1 = d(xm , xn ) < /2; m, n > K2 = d(ym , yn ) < /2; n > K3 = d(xn , yn ) < /2. Set K = max(K1 , K2 , K3 ). Then we will show that m, n > 2K = d(zm , zn ) < .

6 Note that, if m > 2K , then m/2 > K1 , K2 and (m + 1)/2 > K2 , K3 . Thus we see that m, n even zm = xm/2 , zn = xn/2 d(zm , zn ) < /2 < ; m, n odd zm = y(m+1)/2 , zn = y(n+1)/2 d(zm , zn ) < /2 < ; m even, n odd zm = xm/2 , zn = y(n+1)/2 d(zm , zn ) < d(xm/2 , ym/2 ) + d(ym/2 , y(n+1)/2 ) < /2 + /2 = ; and so (zn ) is a Cauchy sequence; that is, (xn ) and (yn ) are both subsequences of some Cauchy sequence (zn ). Conversely, suppose that (xn ) and (yn ) are both subsequences of some Cauchy sequence (zn ); say, (xn ) = (zjn ) and (yn ) = (zkn ). Then
n

lim d(xn , yn ) = lim d(zjn , zkn ) = 0,

since (zn ) is Cauchy and n implies jn , kn . Hence (xn ) (yn ).