221 Analysis 2, 2008–09 Suggested solutions to exercise sheet 1 1. Show that if A is a ring of subsets of a set X, then A ∩ B ∈ A whenever A ∈ A and B ∈ A. Solution. We have A ∩ B = (A ∪ B) \ (B \ A) ∪ (A \ B) (to see this, draw a Venn diagram). Since A is closed under \ and ∪, we have A ∩ B ∈ A. Alternatively, observe that A∩B = A\(A\B); this shows that closure under the set difference is enough to ensure that A is closed under intersections. 2. (a) Suppose that X and I are sets, and that for every i ∈ I, we are given a ring Ai of subsets of X. Prove that \ A= Ai i∈I is a ring of subsets of X. (b) If S ⊆ P(X), let r(S) be given by \ r(S) = {A : A is a ring of subsets of X and S ⊆ A}. Prove that (i) r(S) is a ring containing S, and (ii) if B is any ring of subsets with S ⊆ B, then r(S) ⊆ B. [This shows that r(S) is the smallest ring containing S; we say that r(S) the ring generated by S.] (c) Show that if X = R and S = {(a, b] : a, b ∈ R} then r(S) is the interval ring I. [You may assume that I is a ring of subsets of R; this is proven in an exercise below.] Solution. (a) Since each Ai is a ring, we have ∅ ∈ Ai for all i ∈ I, so ∅ ∈ A. If A, B ∈ A then A, B ∈ Ai for all i ∈ I. Since each Ai is a ring, we have A \ B ∈ Ai and A ∪ B ∈ Ai for all i ∈ I. So A \ B ∈ A and A ∪ B ∈ A. Hence A is a ring. (b) (i) r(S) is an intersection of rings, so by (a) it is a ring; since each A in the intersection contains S, so does r(A). (ii) If B is a ring and S ⊆ B then B participates in the intersection defining r(S), so r(S) ⊆ B. 1 (c) Since S ⊆ I and I is a ring, by property (ii) we have r(S) ⊆ I. On the other hand, if S ⊆ A where A is any ring of subsets of R, then since A is closed under finite unions and I is the set of finite unions of elements of S, we must have I ⊆ A. So by the definition of r(S), it follows that I ⊆ r(S); the reverse inclusion was established in the last paragraph, so r(S) = I. 3. Let X be a set, let A ⊆ P(X) with ∅ ∈ A, and let m : A → [0, ∞] be a measure on A. Show that if A, B and A \ B are in A with B ⊆ A, (a) m(B) ≤ m(A) [we say that m is increasing], and (b) if m(B) < ∞, then m(A \ B) = m(A) − m(B). What happens in (b) if m(B) = ∞? Solution. Since B ⊆ A, we have A = B ∪ (A \ B), a disjoint union. Using the (countable) additivity of m, we see that m(A) = m(B ∪ (A \ B)) = m(B) + m(A \ B). Since m(A \ B) ≥ 0, this shows that m(A) ≥ m(B). If m(B) < ∞ then we can subtract m(B) from both sides to obtain (b). If m(B) = ∞ then m(A) = m(B) + m(A \ B) ≥ m(B), so m(A) = ∞ too. So the expression m(B) − m(A) is undefined in this case. Also, when m(B) = ∞, the measure of A \ B can take any value in [0, ∞]; for example, if m is Lebesgue measure on the line then m(R \ R) = m(∅) = 0 and m(R \ (−∞, 0)) = m[0, ∞) = ∞. 4. (a) Show that for any a, b, c, d ∈ R, (a, b] ∩ (c, d] = (max(a, c), min(b, d)], (a, b] ∩ (−∞, d] = (a, min(b, d)] and (a, b] ∩ (c, ∞) = (max(a, c), b]. [If b ≤ a then (a, b] is the empty set.] (b) Suppose that B = (a1 , b1 ] ∪ · · · ∪ (an , bn ] where a1 < b1 < a2 < b2 < · · · < bn−1 < an < bn . Write B ∁, the complement of B in R, as a union of n + 1 intervals. (c) Show that the interval ring I = {(a1 , b1 ] ∪ · · · ∪ (an , bn ] : n ≥ 1, ai , bi ∈ R, 1 ≤ i ≤ n} is a ring of subsets of R. [Hint: A \ B = A ∩ B ∁.] 2 Solution. (a) We have x ∈ (a, b] ∩ (c, d] ⇐⇒ ⇐⇒ ⇐⇒ ⇐⇒ ⇐⇒ x ∈ (a, b] & x ∈ (c, d] (x > a & x ≤ b) & (x > c & x ≤ d) (x > a & x > c) & (x ≤ b & x ≤ d) x > max(a, c) & x ≤ min(b, d) x ∈ (max(a, c), min(b, d)]. So (a, b] ∩ (c, d] = (max(a, c), min(b, d)]. We could prove the other two identities in a similar fashion; but observe that the argument above works just as well if any of a, b, c, d are ±∞. So (a, b] ∩ (−∞, d] = (max(a, −∞), min(b, d)] = (a, min(b, d)] and (a, b] ∩ (c, ∞) = (max(a, c), min(b, ∞)] = (max(a, c), b]. (b) A point x is in the complement of B iff x ∈ (bi , ai+1 ] where 1 ≤ i < n, or x ≤ a1 , or x > bn . So B ∁ = (−∞, a1 ] ∪ (b1 , a2 ] ∪ (b2 , a3 ] ∪ · · · ∪ (bn−1 , an ] ∪ (bn , ∞). (c) The empty set ∅ = (0, 0] is in I, and it is clear from the definition of I that it is closed under finite unions. Let A, B ∈ A; then we may write A = A1 ∪ · · · ∪ Am and B = (a1 , b1 ] ∪ . . . (an , bn ] where each Ai is of the form (a, b], and B is written as in (b) (since every element of I can be written in this way). Let D0 = (−∞, a1 ], Dj = (bj , aj+1] for 1 ≤ j < n and Dn = (bn , ∞) so that B ∁ = D0 ∪ D1 ∪ · · · ∪ Dn . Then ∁ A\B =A∩B =A∩ n [ Dj = j=0 = n [ m [ j=0 Ai ∩ Dj = i=1 n [ (A ∩ Dj ) j=0 n m [[ (Ai ∩ Dj ). j=0 i=1 By part (a), we have Ai ∩ Dj ∈ I for each 1 ≤ i ≤ m and 0 ≤ j ≤ n. Since I is closed under finite unions, the union A \ B of these (n + 1)m sets is also in I. 5. Let A = A1 ∪ · · · ∪ An where Ai = (ai , bi ] and a1 < b1 < a2 < · · · < bn−1 < an < bn , and suppose there are real numbers bj , cj such that A can also be written as ∞ [ A= Bj where Bj = (cj , dj ]. j=1 Prove that there is a partition N = S1 ∪S2 ∪· · ·∪Sn such that for i = 1, 2, . . . , n we have [ Ai = Bj . j∈Si [We write N = {1, 2, 3, . . . }; a partition is simply a disjoint union.] 3 Solution. For i = 1, 2, . . . , n − 1, let Si = {j ∈ N : Bj 6= ∅, Bj ⊆ Ai }, and let Sn = {j ∈ N : Bj ⊆ An }. If k 6= ℓ and j ∈ Sk ∩ Sℓ then Bj ⊆ (ak , bk ] ∩ (aℓ , bℓ ] = ∅, so Bj = ∅ which forces k = ℓ = n, a contradiction. So Sk ∩ Sℓ = ∅ if k 6= ℓ. If j ∈ N \ (S1 ∪ · · · ∪ Sn ) then ∅ = 6 Bj ⊆ A buy Bj is not entirely contained in any Si . So there exist k < ℓ and x ∈ Ak ∩ Bj and y ∈ Aℓ ∩ Bj . [Why?] Since Bj is an interval containing x and y and x < y, we have [x, y] ⊆ Bj . But x ≤ bk < aℓ < y, so aℓ ∈ [x, y] ⊆ Bj ⊆ A which is false, since aℓ 6∈ A. So S1 ∪ · · · ∪ Sn = N. S Let Ci = j∈Si Bj ; we claim that Ci = Ai . By construction, Bj ⊆ Ai for j ∈ SSi , so Ci ⊆ Ai . Hence Ci ∩ Ai = Ci and Ck ∩ Ai = ∅ ifSk 6= i. Since A = j∈N Bj and S1 , . . . , Sn is a partition of N, we have A = nk=1 Ci , so Ai = A ∩ Ai = n [ k=1 Ck ∩ Ai = n [ (Ck ∩ Ai ) = Ci . k=1 4