Math 516 Professor Lieberman April 10, 2009 HOMEWORK #6 SOLUTIONS Chapter 11 5. (a) We just have to check countable S additivity. If (En ) is a disjoint sequence of measurable sets, then we set E = En . It follows that λE = µE + νE X X X = µEn + νEn = (µEn + νEn ) X = λEn . (b) First, if νE < ∞, we set λE = µE − νE. If νE = ∞, then we need to be more careful. Specifically, we set λE = sup{µ(E 0 ) − ν(E 0 ) : E 0 ⊂ E and νE 0 < ∞}. Again, we just have to check that λ is countably additive, so let (En ) be a S sequence of disjoint measurable sets and set E = E . Suppose first that n P λEn = ∞ for all n. Then for any M > 0, there is a positive integer N (M ) such that N (M ) X λEn > M, n=1 and hence, for each n ≤ N (M ), there is a subset En0 of En with νEn < ∞ and N (M ) X SN (M ) µ(En0 ) − ν(En0 ) > M. n=1 Setting E 0 = n=1 En0 , we see that ν(E 0 ) < ∞ and µ(E 0 ) − ν(E 0 ) > M , so λE = ∞. P Next, suppose that λEn = A < ∞. Then, for any ε > 0, there is a positive integer N (ε) such that N (ε) X ε λEn > A − . 2 n=1 Then, for each n ≤ N (ε), there is a set En0 ⊂ En such that νEn0 < ∞ and ε λEn ≥ λEn0 + . 2N (ε) It follows that N (ε) X λE ≥ λEn0 ≥ A − ε. n=1 1 2 P Hence λE ≥ λEn . 0 0 On thePother hand, if E 0 ⊂ E and νE 0 < ∞, we set E n = E ∩ En . Then P νE 0 = νEn0 so this sum converges. In addition, µE 0 = µEn0 , so we have X λE 0 = λEn0 . P 0 0 From the definition λEn , P of λ, we know that λEn ≤ λEn , so we have λE ≤ and hence λE ≤ λEn . Combined with our previous inequality, we conclude that X λE = λEn , and hence λ is a measure. (c) If ν is σ-finite, let (Xn ) be a sequence of disjoint sets with X = ∪Xn . If λ is a measure such that µ = ν + λ, then we must have λ(E ∩ Xn ) = µ(E ∩ Xn ) − ν(E ∩ Xn ) for any measurable set E. Then countable additivity implies that λE = ∞ X µ(E ∩ Xn ) − ν(E ∩ Xn ). n=1 (d) To show that λ is not unique, take B to be the σ-algebra {∅, X} and ( 0 if E = ∅, µE = νE = ∞ if E = X. We then note that any λ with λ∅ = 0 and λX ≥ 0 satisfies µ = λ + ν. 14. Suppose first that f is measurable with respect to B0 . For each α ∈ R, we set Aα = {x ∈ X : f (x) ≥ α}. From property (iii) from Proposition 11.4, there are sets Bα , Cα , and Dα with Bα and Cα in B, Dα ⊂ Cα , µCα = 0 and Aα = Bα ∪ Dα . Since Q is a dense subset of R, Lemma 11.9 gives a function g, which is measurable with respect to B such that g ≥ α a.e. in Bα and g ≤ α a.e. in X \ Bα for all α ∈ Q. If we take [ E= Cα , α∈Q it follows that f = g off of E. Conversely, suppose there are g which is measurable with respect to B and E ∈ B such that µE = 0 and f = g off of E. Then g is measurable with respect to B0 (because it’s a larger σ-algebra) and f = g a.e. [µ0 ]. Proposition 11.8 implies that f is measurable with respect to B0 . 21. (a) For each positive integer n, Set An = {x : |f (x)| ≥ n1 }. Then Z 1 |f | ≥ µAn , n so An has finite measure. On the other hand, {x : f (x) 6= 0} is the union of the An ’s, so it is of σ-finite measure. 3 (b) From part (a), we have that E = {x : f (x) 6= 0} is σ-finite, so Proposition 11.7 gives a sequence (ϕn ) of simple functions which increases to f such that ϕn vanishes outside a set of finite measure. (c) From the definition of the integral, there are simple functions ϕ1 and ϕ2 such that 0 ≤ ϕ1 ≤ f + , 0 ≤ ϕ2 ≤ f − and Z Z ε ε + (f − ϕ1 ) dµ < , (f − − ϕ2 ) dµ < . 2 2 We then take ϕ = ϕ1 − ϕ2 and observe that ϕ1 = 0 wherever f ≤ 0 and ϕ2 = 0 wherever f ≥ 0. Hence |f − ϕ| = (f + − ϕ1 ) + (f − − ϕ2 ), It follows that Z |f − ϕ| dµ < ε. 24. As usual, we only have to check countable additivity. We write (En ) for a sequence S of disjoint measurable sets and E = En . Then, for any k, we have X X µk E = µk En ≤ µEn . Sending k → ∞, we see that µE ≤ X µEn . Now let N be a positive integer. Then, for any k, we have µk E = ∞ X µk En ≥ n=1 N X µk En . n=1 Since the sum on the right is finite, we have N X µEn = n=1 N X lim µk En = lim n=1 N X µk E n . n=1 Hence µE ≥ N X µEn n=1 for any positive integer N , so µE ≥ ∞ X µEn . n=1 (Note that this inequality is true without assuming that (µn ) is increasing.) Combining the two inequalities shows that µ is countably additive. 4 25. Write m for Lebesgue measure on the real line and B for the σ-algebra of all Lebesgue measurable subsets of R. We then define µn on B by µn E = µ(E ∩ (n, ∞)). Then µn converges setwise to the function µ defined by ( ∞ if m(E ∩ [0, ∞)) = ∞, µE = 0 otherwise. It’s easy to see that (µn ) is decreasing. On the other hand, we can write [1, ∞) = ∪[n, n + 1). Since µ[1, ∞) = ∞ but µ[n, n + 1) = 0, it follows that µ is not countably additive and hence it isn’t a measure. 32. (a) We use the formula in the hint and set µ ∧ ν = 21 (µ + ν − |µ − ν|). We now derive an alternative formula for µ ∧ ν. Let A be A is a positive set for µ − ν such that X \ A is a negative set for µ − ν. (In other words, A and X \ A form a Hahn Decomposition for µ − ν.) Then, for any measurable set E, we have |µ − ν|(E) = (µ − ν)(E ∩ A) − (µ − ν)(E \ A), so 1 (µ ∧ ν)E = (µ(E ∩ A) + µ(E \ A) + ν(E ∩ A) + ν(E \ A) 2 + (ν − µ)(E ∩ A) + (µ − ν)(E \ A)) = ν(E ∩ A) + µ(E \ A). Since A is a positive set for µ − ν, it follows that ν(E ∩ A) ≤ µ(E ∩ A), so (µ ∧ ν)E ≤ µ(E ∩ A) + µ(E \ A) = µE, and a similar argument shows that (µ ∧ ν)E ≤ νE. If λ ≤ µ and λ ≤ ν, we have that, for any measurable set E, λE = λ(E ∩ A) + λ(E \ A) ≤ ν(E ∩ A) + µ(E \ A), and hence λ ≤ µ ∧ ν. (b) We define µ ∨ ν = 12 (µ + ν + |µ − ν|). The proof that µ ∨ ν is the smallest measure larger than µ and ν is similar to part (a), and the formulas show that µ ∨ ν + µ ∧ ν = µ + ν. (c) If µ and ν are mutually singular, let A and B be disjoint measurable sets with A ∪ B = X and µ(A) = ν(B) = 0. Fix a set E and note that µE = µ(E ∩ B) and νE = ν(E ∩ A). Hence 1 (µ ∧ ν)(E ∩ A) = (µ(E ∩ A) + ν(E ∩ A) − |µ − ν|(E ∩ A)). 2 But E ∩ A is a null set for µ and hence it’s a positive set for ν − µ, which means that |µ − ν|(E ∩ A) = (ν − µ)(E ∩ A) = ν(E ∩ A). Hence (µ ∧ ν)(E ∩ A) = 0. A similar argument shows that (µ ∧ ν)(E ∩ B) = 0, so (µ ∧ ν)E = 0. 5 Conversely, if µ ∧ ν = 0, then µ + ν = |µ − ν|. Taking A as in part (a), we see that µ(A) + ν(A) = µ(A) − ν(A), so ν(A) = 0. In addition, if we set B = X \ A, then B is a negative set, so µ(B) + ν(B) = ν(B) − µ(B) and hence µ(B) = 0. Hence µ and ν are mutually singular.