Midterm Solutions Math 461, Spring 2016 Problem 1. If S and T are sets, the symmetric difference of S and T is the set S 4 T = (S ∩ T c ) ∪ (S c ∩ T ). If {En } is a sequence of measurable sets in [0, 1], we say that {En } converges to a measurable set E if lim m(En 4 E) = 0. n→∞ Part (a) Proposition. Let E1 , E2 , . . . and E be measurable sets in [0, 1] and suppose that χEn → χE pointwise almost everywhere on [0, 1]. Then {En } converges to E. Proof. Since χEn → χE pointwise almost everywhere, we know that |χEn −χE | → 0 pointwise almost everywhere. But |χEn − χE | ≤ 1 for all n, so by the bounded convergence theorem Z lim m(En 4 E) = lim n→∞ n→∞ χEn 4E dm X |χEn − χE | dm = = lim n→∞ Z Z Z X lim |χEn − χE | dm = X n→∞ 0 dm = 0. X Part (b) One simple example is the following sequence {En } of closed intervals: [0, 1], 0, 21 , 12 , 1 , 1 1 2 2 0, 3 , 3 , 3 , 3 , 1 , 1 0, 4 , . . . Since m(En ) → 0 as n → ∞, this sequence converges to the empty set. However, for each number x ∈ [0, 1], there exist infinitely many terms of this sequence that contain x, as well as infinitely many that do not contain x, and therefore {χEn (x)} does not converge. 1 Part (c) Proposition. Let {En } be a sequence of measurable sets in [0, 1], and suppose that ∞ X m(En 4 En+1 ) < ∞. n=1 Then the sequence {χEn } converges pointwise almost everywhere. Proof. For each N ∈ N, let [ ZN = (En 4 En+1 ). n≥N and let Z = T N ∈N ZN . Since m(ZN ) ≤ X m(En 4 En+1 ), n≥N we know that m(ZN ) → 0 as N → ∞, and hence m(Z) = 0. But if x ∈ Z c , then x ∈ ZNc for some N ∈ N, and hence x ∈ / En 4 En+1 for all n ≥ N . It follows that χn (x) = χn+1 (x) for all n ≥ N , so the sequence {χn (x)} converges. Thus {χEn } converges pointwise on Z c . Part (d) Proposition. Let {En } be a sequence of measurable sets in [0, 1], and suppose that for every > 0, there exists an N ∈ N so that i, j ≥ N ⇒ m(Ei 4 Ej ) < . Then {En } converges to some measurable set E. Proof. Let N1 < N2 < · · · be an increasing sequence so that i, j ≥ Nk ⇒ m(Ei 4 Ej ) < 1 2k for each k, and let Fk = ENk . Then m(Fk 4 Fk+1 ) < 1/2k for each k, so by part (c) the sequence {χFk } converges pointwise almost everywhere. The limit is a measurable function with range {0, 1}, so χFk → χE pointwise almost everywhere for some measurable set E. By part (a), it follows that {Fk } converges to E. To prove that {En } converges to E, observe that for n ≥ Nk we have m(En 4 E) ≤ m(En 4 Fk ) + m(Fk 4 E) ≤ 1 + m(Fk 4 E). 2k But the quantity on the right goes to 0 as k → ∞, so it follows that m(En 4 E) → 0 as n → ∞. 2 Problem 2. If f : R → R is a bounded measurable function and g : R → R is an L1 function, the convolution of f and g is the function f ∗ g : R → R defined by Z (f ∗ g)(x) = f (x − t) g(t) dm(t). R (Note that this integral always exists, so f ∗ g is a well-defined function.) Part (a) Proposition. Let fn : R → R be a uniformly bounded sequence of measurable functions converging pointwise to a function f : R → R, and let g : R → R be an L1 function. Then fn ∗ g converges pointwise to f ∗ g. Proof. Let x ∈ R, and let M > 0 so that |fn | ≤ M for all M . Then |fn (x − t)g(t)| ≤ |M g(t)| for all n ∈ N and all t ∈ R. But |M g| is an L1 function, so by the dominated convergence theorem Z lim (fn ∗ g)(x) = lim fn (x − t) g(t) dm(t) n→∞ n→∞ R Z Z lim fn (x − t) g(t) dm(t) = = f (x − t) g(t) dm(t) = (f ∗ g)(x). R n→∞ R Part (b) Proposition. If f : R → R is a bounded continuous function and g : R → R is an L1 function, then f ∗ g is continuous. Proof. Let x ∈ R, and let {xn } be a sequence in R that converges to x. We will prove that (f ∗ g)(xn ) converges to (f ∗ g)(x). Since f is continuous, we know that lim f (xn − t) = f (x − t) n→∞ for all t ∈ R. Since f is bounded, there exists a constant M > 0 so that |f | ≤ M . Then |f (xn − t) g(t)| ≤ |M g(t)| for all n ∈ N and all t ∈ R. But |M g| is an L1 function, so by the dominated convergence theorem Z lim (f ∗ g)(xn ) = lim n→∞ n→∞ f (xn − t) g(t) dm(t) R Z = Z lim f (xn − t) g(t) dm(t) = f (x − t) g(t) dm(t) = (f ∗ g)(x). R n→∞ R 3