Do not submit problems in blue, but at least do them.
(1) Prove that the only subsets of Rn which are both open and closed are Rn and the empty set. (Hint: If X is another such a set, pick a X, b 6 X and consider sup{t [0, 1]|a+t(b a) X}. In other words, imitate what we did in class.) Since X 6= Rn and X 6= , we can find a 6= b as in the hint. Let c(t) = a + t(b a) for t [0, 1]. Then the set S = {t [0, 1]|c(t) X} is bounded above and non-empty, since 0 S, and thus has a supremum, say u. Consider c(u). Then it has two choices, either c(u) X or 6 X. We will arrive at a contradiction in both the cases. So, first assume that c(u) 6 X. Then, u 6= 0, since c(0) = a X. So, 0 < u 1. Since Rn X is open and c(u) Rn X, we can find a B(c(u), r) Rn X for some r > 0. r If t [0, 1] satisfies u ||ab|| < t < u, then I will show that r c(t) B(c(u), r) and hence c(t) 6 X. Thus, supS u ||ab|| , contradicting the fact that u = supS. Notice that such t exists r since u > 0. So, assume that t [0, 1] with u ||ab|| < t < u. ||c(t) c(u)|| = (u t)||a b|| r < ||a b|| = r. ||a b|| This proves that c(t) B(c(u), r) in this range, leading to a contradiction. Next assume that c(u) X. Then u 6= 1 since c(1) = b 6 X. So, 0 u < 1. Since X is open, we can find a B(c(u), r) X for some r > 0. As before, we will show that if t [0, 1] r with u < t < u + ||ab|| , then c(t) B(c(u), r) and which will contradict the fact that u = supS, since u < 1. ||c(t) c(u)|| = (t u)||a b|| r < ||a b|| = r. ||a b|| This yields the required contradiction. (2) Let (M, d) be a metric space. If S, T are subsets of M , define d(S, T ) = inf{d(s, t)|s S, t T }, which makes sense, since this set is bounded below by zero. If S = {a}, a singleton set, we will write d(a, T ) instead of d({a}, T ). (a) Prove that if S is a closed subset and a M , then d(a, S) = 0 if and only if a S. 1
Since d(a, S) = inf{d(a, s)|s S}, if a S, we have
0 = d(a, a) in this set and thus its infimum (which is nonnegative) is zero. Conversely assume that d(a, S) = 0 and S is closed. For any r > 0, B(a, r) S can not be empty, since there exists an s S with d(a, s) < r, by our hypothesis that the infimum is zero. So, a is an adherent point of S and since S is closed, a S. (b) If S is compact and T is closed with S T = , prove that d(S, T ) > 0. Let a S and then since a 6 T , by the previous result d(a, T ) = ra > 0. Then B(a, ra ) S = . Consider the open cover {B(a, ra /2)} of S. This is a covering, since any a S is in B(a, ra /2). Since S is compact, it has a finite subcover, say B(ai , rai /2) with 1 i m. I claim that d(S, T ) is at least r = min{rai /2|1 i m} > 0. For this it suffices to prove that for any s S, t T , d(s, t) r. Let s B(ai , rai /2) for some i. Since B(ai , rai ) T = , we have d(ai , t) rai . Thus, rai d(ai , t) d(ai , s) + d(s, t) < rai /2 + d(s, t) So, d(s, t) > rai /2 r. (c) Give an example of two closed subsets S, T of R2 with S T = , but d(S, T ) = 0. Take S to be the hyperbola xy = 1 and T , the line x = 0. (Remember asymptotes from Calculus). (d) Prove that every closed subset of M is the intersection of countably many open sets. Let S be closed. Let Un be the open set defined as {a M |d(a, S) < 1/n} for n N. (Why is Un open?) Then Un = S. (3) Prove that a collection of disjoint open sets in Rn is necessarily countable. Give an example of a collection of disjoint closed sets which is not countable. The first part is just Lindelofs theorem. If {U } is a collection of disjoint open sets, then consider X = U , an open set. By Lindelof, since X is covered by the U s, it must have a countable subcover, say Ui for i N, where each Ui = U for some . Now for any , U X = Ui , but these are disjoint
and so easy to see that U = Ui for some i, showing that there
are only countably many U s. The second part is easy. Take the set R which is uncountable. Then R = xR {x}. Each {x} is closed and they are disjoint. (4) If X, Y are connected subsets of a metric space and X Y 6= , prove that X Y is connected. Let Z = X Y and fix a point c X Y . If Z is not connected, we will have open sets U, V such that Z U V , U Z 6= = 6 V Z, but U V Z = . Since c Z U V , c is in one of U, V . If c is in both U, V , then c U V Z = . So, c belongs to precisely one of U, V say U . Now we will show that Z U and then = U V Z = V Z 6= , leading to a contradiction. For this, it suffices to show that X U and Y U . We will do one of these, the other being identical. If X is not contained in U , we know that X U V and thus X V 6= . Of course c U and thus X U 6= . On the other hand X U V Z U V = , contradicting the fact that X is connected. (5) If S is a subset of Rn such that for every point x S has an open neighbourhood B(x, r) (r > 0 may depend on x) which intersects S in a countable set, prove that S is countable. So, we have a positive number rx such that B(x, rx ) S is countable for each x S. These form an open cover of S and thus by Lindelof, we can extract a countable subcover, which for ease of notation, let me call Bi , i N. Then S i S Bi . But, S Bi is countable and since a countable union of countable sets is countable and a subset of a countable set is countable, we see that S is countable. (6) We say that a subset S of Rn is convex, if for any two points a, b S and for any t [0, 1], ta + (1 t)b S. (a) Prove that any open ball in Rn is convex. Consider B = B(p, r) the open ball of radius r > 0 with center p and let a, b B and let t [0, 1]. Then, ||ta + (1 t)b p|| = ||t(a p) + (1 t)(b p)|| t||a p|| + (1 t)||b p|| < tr + (1 t)r = r So ta + (1 t)b B.
(b) Prove that if S is convex, so is its closure. (Remember
that the closure of S is its union with all its accumulation points.) Let a, b S and let t [0, 1]. Then there exists sequences {an }, {bn } in S such that lim an = a and lim bn = b. So, ta + (1 t)b = t lim an + (1 t) lim bn = lim(tan + (1 t)bn ) Each term inside the paranthesis is in S, since S is convex and thus the limit which exists must be in S. (c) For a set S we define the interior of S, denoted by int S to be {x S|B(x, r) S} (again, r > 0 may depend on x). Prove that if S is convex, so is int S. (Hint: If U, V are open in Rn , so is U + V defined as {u + v|u U, v V }.) Again, let a, b int S and let t [0, 1]. We want to show that ta + (1 t)b int S. If t = 0, 1, this is obvious. So, we may assume that t > 0, 1 t > 0. We can find an r > 0 so that B(a, r) and B(b, r) are subsets of S. Consider the set tB(a, r) + (1 t)B(b, r) defined as {tx + (1 t)y|x B(a, r), B(b, r)}. By convexity of S, this set is contained in S. If we show that this set is open, we would be done. Noticing that tB(a, r) = B(ta, tr) and similarly one has (1t)B(b, r) = B((1t)b, (1t)r) and tr > 0, (1t)r > 0, we see that it suffices to prove the following. If U, V are open sets in Rn , so is the set U + V = {u + v|u U, v V }. Since U + V = vV (U + v), and union of open sets is open, suffices to prove that U + v is open where U is open and v Rn . This easily follows since for any u U , B(u, r) + v = B(u + v, r).