Math 318 HW #8 Solutions 1. (a) Prove Chebyshev’s inequality, which says that if f is nonnegative and measurable on the bounded, measurable set A, then Z 1 f dm. m{x ∈ A : f (x) ≥ c} ≤ c A Proof. Let c ∈ R and define Ac := {x ∈ A : f (x) ≥ c}, which is measurable since f is a measurable function. Then f (x) ≥ c on Ac , so, by Theorem 25.4(2), Z f dm ≥ cm{x ∈ A : f (x) ≥ c)}. Ac Therefore, 1 m{x ∈ A : f (x) ≥ c} ≤ c Z 1 f dm ≤ c Ac Z f dm, A where the last inequality follows from Theorem 25.3 and the fact that f is nonnegative. (b) Show that if R A |f | dm = 0, then f = 0 a.e. on A. Proof. For any n ∈ N, Chebyshev’s inequality implies that Z m{x ∈ A : |f (x)| > 1/n} ≤ n |f | dm = 0. A Letting Bn = {x ∈ A : |f (x)| > 1/n} and B = {x ∈ A : |f (x)| > 0}, we have that B= ∞ [ Bn , n=1 and B1 ⊂ B2 ⊂ · · · . Therefore, Corollary 10.9 implies that m(B) = lim m(Bn ) = lim 0 = 0. n→∞ n→∞ But of course B = {x ∈ A : f (x) 6= 0}, so f = 0 a.e. 2. Exercise 26.24. If f is a measurable function such that B ⊂ A, show that f = 0 a.e. on A. R B Proof. Let B+ := {x ∈ A : f (x) > 0} B− := {x ∈ A : f (x) < 0}. 1 f dm = 0 for all measurable sets Since f is measurable, both B+ and B− are measurable sets. Note that B+ ∪ B− = B, where B = {x ∈ A : f (x) 6= 0}. The goal is to show that B has measure zero, so it suffices to show that m(B+ ) = m(B− ) = 0. Now, for each n ∈ N, define B+n := {x ∈ A : f (x) > 1/n} B−n := {x ∈ A : f (x) < −1/n}, which again are measurable sets since f is measurable. Hence, Z 1 0= f dm ≥ m(B+n ) n B Z +n 1 0= f dm ≤ − m(B−n ), n B−n which can only be true if m(B+n ) = m(B−n ) = 0. Moreover, B+ = ∞ [ B+n and B− = n=1 ∞ [ B−n n=1 and we have B+1 ⊂ B+2 ⊂ · · · and B−1 ⊂ B−2 ⊂ · · · . Therefore, by Corollary 10.9, m(B+ ) = lim m(B+n ) = lim 0 = 0 n→∞ n→∞ m(B− ) = lim m(B−n ) = lim 0 = 0. n→∞ n→∞ Thus, we conclude that m(B) = m(B+ ) + m(B− ) = 0 + 0 = 0, so f = 0 a.e. 3. Exercise 26.36. Let g ∈ L[a, b], h(x) = [a, b]. R [a,x] g dm for x ∈ [a, b]. Show that h is continuous on Proof. Let An := {x ∈ [a, b] : n − 1 ≤ |g(x)| < n}. Then the An are pairwise disjoint, so the additivity of the Lebesgue integral implies Z ∞ Z X |g| dm = |g| dm. [a,b] n=1 An Let > 0. Then the definition of convergence implies there exists N ∈ N such that Z ∞ X |g| dm < /2. n=N +1 An For R notational convenience, define BN = CN |g| dm < /2. SN n=1 An 2 and CN = [a, b]\BN ; the above says that 2N < x ≤ c. Then Z Z g dm g dm − |h(c) − h(x)| = [a,c] [a,x] Z g dm = [x,c] Z |g| dm ≤ [x,c] Z Z = |g| dm + |g| dm [x,c]∩BN [x,c]∩CN Z Z ≤ N dm + |g| dm Let x, c ∈ [a, b] such that c − [x,c]∩BN CN <N + 2N 2 = . Similar reasoning shows that |h(x) − h(c)| < when c ≤ x < 2N . Hence, we see that for any x, c ∈ [a, b] such that |x − c| < δ = 2N , we have |h(x) − h(c)| < , so h is uniformly continuous on [a, b]. 3