SOLUTIONS OF HW7 MINGFENG ZHAO April 04, 2011 (Problem 46, in Page 68) Let X = Y = [0, 1], M = N = B[0,1] , and µ=Lebesgue mea- 1. If D = {(x, x) : sure, and ν=counting measure. RR χD dµdν, RR χD dνdµ, and R x ∈ [0, 1]} is the diagonal in X × Y , then χD d(µ × ν) are all unequal. Proof. Consider the function f : [0, 1] × [0, 1] → R given by f (x, y) = x − y. So f is continuous on [0, 1] × [0, 1], so f is (B[0,1] × B[0,1] , BR )-measurable. In particular, D = {(x, x) : x ∈ [0, 1]} = f −1 (0) is in B[0,1] × B[0,1] . By Proposition 2.34,in Page 65, we know for any x, y ∈ [0, 1], Dx = {(x, x)}, Dy = {(y, y)} are in B[0,1] . Hence for any y ∈ [0, 1], we have Z Z χD (x, y)dµ(x) = dµ(x) {y} [0,1] = µ({y}) = 0. Therefore, we have Z Z Z χD dµdν Z = χD (x, y)dµ(x)dν(y) [0,1] [0,1] Z = 0dν(y) [0,1] = 0. 1 2 MINGFENG ZHAO And also for any x ∈ [0, 1], we know Z Z χD (x, y)dν(y) = dν(y) {x} [0,1] = ν({x}) = 1. Therefore, we have Z Z Z Z χD (x, y)dν(y)dµ(x) χD dνdµ = [0,1] [0,1] Z = 1dµ(x) [0,1] = For R 1. χD d(µ × ν), we know χD is the simple function on D, then Z χD d(µ × ν) = (µ × ν)(D). By the definition of µ × ν, we know ( (µ × ν)(D) = inf ∞ X (µ × ν)(Ak × Bk ) : Ak , Bk ∈ B[0,1] , D⊂ k=1 For any D ⊂ S∞ k=1 (Ak ×Bk ), ∞ [ ) (Ak × Bk ) . k=1 which means that for any 0 ≤ x ≤ 1, we have (x, x) ∈ S∞ k=1 (Ak ×Bk ). In particular, we get x∈ ∞ [ x∈ Ak , k=1 So we get x ∈ S∞ k=1 (Ak ∞ [ Bk . k=1 ∩ Bk ). By the arbitrary of x ∈ [0, 1], we get [0, 1] ⊂ ∞ [ (Ak ∩ Bk ). k=1 S∞ P∞ Hence µ([0, 1]) ≤ µ ( k=1 (Ak ∩ Bk )) ⊂ k=1 µ(Ak ∩ Bk ). So at least there exists some k0 such that µ(Ak0 ∩ Bk0 ) > 0. SOLUTIONS OF HW7 3 Hence we get µ(Ak0 ) > 0, and µ(Bk0 ) > 0, which implies that Bk0 is an infinity set, so ν(Bk0 ) = ∞. So we have (µ× = 6 )(Ak0 × Bk0 ) = µ(Ak0 ) · ν(Bk0 ) = ∞. So we get P∞ k=1 (µ × ν)(Ak × Bk ) = ∞. Since D ⊂ S∞ k=1 (Ak × Bk ) is arbitrary, then by the definition of (µ × ν)(D), we get (µ × ν)(D) = ∞. This implies that Z χD d(µ × ν) = (µ × ν)(D) = ∞. 2. (Problem 50, in Page 69) Suppose (X, M, µ) is a σ-finite measure space, and f ∈ L+ (X). Let Gf = {(x, y) ∈ X × [0, ∞] : y ≤ f (x)}. Then Gf is M ⊗ BR -measurable, and (µ × m)(Gf ) = R f dµ; the same is also true if the inequality y ≤ f (x) in the definition of Gf is replaced by y < f (x). This is the definitive statement of the familiar theorem from calculus, “ the integral of a function is the area under its graph”. Proof. Without loss of generality, we assume 0 ≤ f (x) < ∞ for all x ∈ X. Consider the function F : X × [0, ∞) −→ R2 given by F (x, y) = (f (x), y). For any A ∈ BR , since f is (BR , BR )-measurable, then f −1 (A) ∈ BR . So for any B ∈ B[0,∞) , we get F −1 (A × B) = f −1 (A) × (B) ∈ BR × B[0,∞) ⊂ BR ⊗ B[0,∞) Since BR ⊗ B[0,∞) = σ(BR × B[0,∞) ), then F is (M ⊗ B[0,∞) , R ⊗ B[0,∞) )-measurable. 4 MINGFENG ZHAO Let the function G : R × [0, ∞) −→ R given by G(x, y) = x − y. Then G is continuous on R × [0, ∞), then G is (BR×[0,∞) , BR )-measurable. Therefore, we have G ◦ F is (M ⊗ B[0,∞) , BR )-measurable. In particular, we have Gf = {(x, y) ∈ X × [0, ∞) : y ≤ f (x)} = (G ◦ F )−1 ((−∞, 0]) ∈ M ⊗ B[0,∞) G̃f = {(x, y) ∈ X × [0, ∞) : y ≤ f (x)} = (G ◦ F )−1 ((−∞, 0)) ∈ M ⊗ B[0,∞) Consider χGf (x, y), χG̃f (x, y), so they are (M ⊗ B[0,∞) , BR )-measurable. Since M ⊗ B[0,∞) is σfinite, therefore we have Z (µ × m)(Gf ) χGf (x, y)d(µ × m) = X×[0,∞) Z Z = χGf (x, y)dm(y)dµ(x) X By Tonelli’s theorem [0,∞) Z Z = dm(y)dµ(x) X [0,f (x)] Z = f (x)dµ(x) X Z (µ × m)(G̃f ) = X×[0,∞) χG̃f (x, y)d(µ × m) Z Z = X [0,∞) χG̃f (x, y)dm(y)dµ(x) By Tonelli’s theorem Z Z = dm(y)dµ(x) X [0,f (x)) Z = f (x)dµ(x) X 3. (Problem 7, in Page 88) Suppose that ν is a signed measure on (X, M) and E ∈ M. Show that a. ν + (E) = sup {ν(F ) : E ∈ M, F ⊂ E}, and ν − (E) = − inf {ν(F ) : E ∈ M, F ⊂ E}; b. |ν|(E) = sup nP n j=1 o |ν(Ej )| : n ∈ N, E1 , · · · , En are disjoint, and ∪nj=1 Ej = E . SOLUTIONS OF HW7 5 Proof. a. For any F ⊂ E, by the Jordan Decomposition Theorem, in Page 87, we know ν + and ν − are two positive measure on X. So by the monotonicity of positive measure, we know ν + (F ) ≤ ν + (E), and ν − (F ) ≤ ν − (E). Hence ν(F ) ν(F ) = ν + (F ) − ν − (F ) ≤ ν + (F ) ≤ ν + (E) = ν + (F ) − ν − (F ) ≥ −ν − (F ) ≥ −ν − (E) By the arbitrary of F ⊂ E, we know ν + (E) ≥ sup {ν(F ) : E ∈ M, F ⊂ E} −ν − (E) ≤ inf {ν(F ) : E ∈ M, F ⊂ E} On the other hand, by the Hahn Decomposition Theorem, in Page 86, we know that there exists sets P and N such that ν(P ) ≥ 0, ν(N ) ≤ 0, and P ∪ N = X, P ∩ N = φ. And also by the definition of ν + , ν − in the Jordan Decomposition Theorem, in Page 87, we know ν + (E) = ν(E ∩ P ), ν − (E) = −ν(E ∩ N ). But E ∩ P ⊂ E, E ∩ N ⊂ E. Therefore, we get ν + (E) = ν − (E) = − inf {ν(F ) : E ∈ M, F ⊂ E} sup {ν(F ) : E ∈ M, F ⊂ E} 6 MINGFENG ZHAO b. For any E1 , · · · , En are disjoint, and ∪nj=1 Ej = E. By the Jordan Decomposition Theorem, in Page 87, we know n X n X |ν(Ej )| = j=1 |ν + (Ej ) − ν − (Ej )| j=1 n X ≤ |ν + (Ej )| + |ν − (Ej )| j=1 n X = |ν|(Ej ) j=1 |ν|(E) = Since Ej ’s are disjoint, and |ν| is a measure on X Hence we get |ν|(E) ≥ sup n X |ν(Ej )| : n ∈ N, E1 , · · · , En are disjoint, and ∪nj=1 Ej = E j=1 On the other hand, by the Hahn Decomposition Theorem, in Page 86, we know that there exists sets P and N such that ν(P ) ≥ 0, ν(N ) ≤ 0, and P ∪ N = X, P ∩ N = φ. So we can get |ν|(E) = ν + (E) + ν − (E) = ν(E ∩ P ) − ν(E ∩ N ) By the definition of ν + , ν − in the Jordan Decomposition Theorem, in Page 87 = |ν(E ∩ P )| + |ν(E ∩ N )| Therefore, we get |ν|(E) = sup n X |ν(Ej )| : n ∈ N, E1 , · · · , En are disjoint, and ∪nj=1 Ej = E j=1 Department of Mathematics, University of Connecticut, 196 Auditorium Road, Unit 3009, Storrs, CT 06269-3009 E-mail address: mingfeng.zhao@uconn.edu