Lecture #7 MATH 321: Real Variables II University of British Columbia Lecture #7: Instructor: Scribe: January 21, 2008 Dr. Joel Feldman Peter Wong See handout: Reduction to the Riemann Integral Theorem. (Reduction to the Riemann Integral Theorem) Proof. (Continued) |S(P, T, f α0 , x) − S(P, T, f, α)| = n X |f (ti )| |α0 (ti ) − α0 (vi )| |xi − xi−1 | i=1 Let ε > 0, f bounded. Then ∃M > 0 such that |f (ti )| ≤ M, ∀i and α0 is continuous on [a, b] =⇒ α0 is uniformly continuous on [a, b] ε 2M (b − a) ε =⇒ ∃δ such that kP k < δ =⇒ |α0 (ti ) − α0 (vi )| < 2M (b − a) =⇒ ∃δ such that |t − v| < δ =⇒ |α0 (t) − α0 (v)| < =⇒ ∃δ such that kP k < δ =⇒ |S(P, T, f α0 , x) − S(P, T, f, α)| ≤ n X M i=1 ε 2M (b − a) [xi − xi−1 ] = ε . 2 Suppose limP S(P, T, f α0 , x) = I. Then ∃Pε0 such that P ⊃ Pε0 =⇒ |S(P, T, f α0 , x) − I| < ε . 2 Choose Pε = Pε0 ∪ {finitely many points chosen such that kPε k < δ}. Then P ⊃ Pε =⇒ |S(P, T, f, α) − I| ≤ |S(P, T, f, α) − S(P, T, f α0 , x)| + |S(P, T, f α0 , x) − I| < Existence (Rudin p.120–127) b U (P, f, α) S(P, T, f, α) b L(P, f, α) Strategy. L(P, f, α) ≤ S(P, T, f, α) ≤ U (P, f, α) We shall show that as P gets finer, L(P, f, α) % and U (P, f, α) & ; and if f is continuous, U (P, f, α) − L(P, f, α) → 0. Until further notice, assume f is bounded and α is monotonic increasing. ε ε + = ε. 2 2 2 MATH 321: Lecture #7 Definition. For any partition P = {x0 , x1 , . . . , xn } on [a, b]. Set L(P, f, α) = U (P, f, α) = n X i=1 n X mi [α(xi ) − α(xi−1 )] where mi = inf f (x) sup f (x) xi−1 ≤x≤xi Mi [α(xi ) − α(xi−1 )] where Mi = xi−1 ≤x≤xi i=1 Lemma. 1. For any partition P inf f (x)[α(b) − α(a)] ≤ L(P, f, α) ≤ S(P, T, f, α) ≤ U (P, f, α) ≤ sup f (x)[α(b) − α(a)] a≤x≤b a≤x≤b 2. If P 0 ⊃ P , then L(P, f, α) ≤ L(P 0 , f, α) and U (P, f, α) ≥ U (P 0 , f, α) 3. If P, P 0 are partitions (not necessarily related), then L(P, f, α) ≤ U (P 0 , f, α) Proof. 1. Trivial since inf f (x) ≤ mi ≤ f (ti ) ≤ Mi ≤ sup f (x), x∈[a,b] x∈[a,b] and we can multiply this by α(xi ) − α(xi−1 ) and sum. 2. Easy: P 0 = P ∪ {a finite # of points}. It suffices, by induction, to consider P 0 = P ∪ {y} with y ∈ (xj−1 , xj ). Then n X Mi [α(xi ) − α(xi−1 )]. U (P 0 , f, α) = i=1 i6=j More on this next time. . . Lecture #8 MATH 321: Real Variables II University of British Columbia Lecture #8: Instructor: Scribe: January 23, 2008 Dr. Joel Feldman Peter Wong Existence Theory Until further notice, we assume f is bounded and α is non-decreasing. Notation. P = {x0 , x1 , . . . , xn }, L(P, f, α) = n X mi = inf xi−1 ≤x≤xi mi [α(xi ) − α(xi−1 )] f (x), Mi = supxi−1 ≤x≤xi f (x) and U (P, f, α) = n X Mi [α(xi ) − α(xi−1 )] i=1 i=1 Lemma. (1) inf a≤x≤b f (x)[α(b) − α(a)] ≤ L(P, f, α) ≤ S(P, T, f, α) ≤ U (P, f, α) ≤ supa≤x≤b f (x)[α(b) − α(a)] (2) P 0 ⊃ P =⇒ L(P, f, α) ≤ L(P 0 , f, α) and U (P, f, α) ≥ U (P 0 , f, α) (3) For any P, P 0 , L(P, f, α) ≤ U (P 0 , f, α). Proof. (1) Done last class. (2) It suffices to consider P 0 = {y} for y ∈ (xj−1 ,j ). For U , U (P 0 , T, α) = ≤ ≤ n X i=1 i6=j n X i=1 i6=j n X Mi ∆αi + sup f (x)[α(y) − α(xj−1 )] + sup f (x)[α(xj ) − α(y)] xj−1 ≤x≤y y≤x≤xj Mi ∆αi + Mj ∆αj Mi ∆xi = U (P, T, α). i=1 The proof for L is similar. (2) (1) (2) (3) Set P ∗ = P ∪ P 0 . Then L(P, f, α) ≤ L(P ∗ , f, α)) ≤ U (P ∗ , f, α) ≤ U (P 0 , f, α). Definition. Define Z b f dα = inf{ U (P, T, α | P is a partition of [a, b] } a Z b f dα = sup{ L(P, T, α | P is a partition of [a, b] } a Note. For all bounded f ’s, increasing α’s, Z a b f dα and Z Theorem. The following are equivalent: a b f dα always exist, and Z a b f dα ≥ Z a b f dα. 2 MATH 321: Lecture #8 (i) f ∈ R(α) on [a, b]. (In-class definition) (ii) ∀ε > 0, ∃Pε such that 0 ≤ U (Pε , f, α) − L(Pε , f, α) < ε. (iii) Rb af dα = Rb af dα. (Rudin’s definition) Proof. (i)=⇒(ii) Assume f ∈ R(α). Let ε > 0. Choose Pε so that Z b ε f dα < . S(Pε , T, f, α) − 3 a Since we have Z b ε f dα < U (Pε , f, α) = sup S(Pε , T, f, α) =⇒ U (Pε , f, α) − 3 T a Z ε b L(Pε , f, α) = inf S(Pε , T, f, α) =⇒ L(Pε , f, α) − f dα < , T 3 a 0 ≤ U (Pε , f, α) − L(Pε , f, α) " # Z " b = U (Pε , f, α) − f dα − L(Pε , f, α) − a ≤ Z b f dα a ε ε + < ε. 3 3 (ii)=⇒(iii) Let ε > 0. We know that L(P, f, α) ≤ Z f dα ≤ Z f dα ≤ U (P, f, α) ≤ ε =⇒ 0 ≤ =⇒ Z Z f dα ≤ f dα = Z Z f dα < ε, f dα # ∀ε > 0 Lecture #9 MATH 321: Real Variables II University of British Columbia Lecture #9: Instructor: Scribe: January 25, 2008 Dr. Joel Feldman Peter Wong Until further notice, f is bounded and α is monotone increasing. Theorem. The following are equivalent: (i) f ∈ R(α) on [a, b]. (The in-class definition) (ii) ∀ε > 0, ∃Pε such that 0 ≤ U (Pε , f, α) − L(Pε , f, α) < ε. (iii) Rb a f dα = Rb a f dα. (Rudin’s definition of integrability) Proof. (i)=⇒(ii): Done last class. (ii)=⇒(iii): Done last class. (iii)=⇒(i): Let ε > 0. Recall that we have b Z f dα = inf U (P, f, α) =⇒ ∃Pε0 such that 0 ≤ U (Pε0 , f, α) − P a Z b f dα < ε a b f dα = sup L(P, f, α) =⇒ P a Z ∃Pε00 such that 0 ≤ Z b f dα − L(Pε00 , f, α) < ε a Choose Pε = Pε0 ∪ Pε00 . Let P ⊃ Pε . Then b ≤ Z b S(P, T, f, α) ≥ L(P, f, α) ≥ L(Pε00 , f, α) ≥ Z S(P, T, f, α) ≤ U (P, f, α) ≤ U (Pε0 , f, α) f dα + ε a f dα − ε a Hence, f ∈ R(α) on [a, b]. Theorem. If f is continuous and α is monotone increasing on [a, b], then f ∈ R(α) on [a, b]. Proof. Let ε > 0. Given that f is continuous on a compact set [a, b], it follows that f is uniformly continuous on [a, b], which means that ∃δ > 0 such that for any a ≤ x, y ≤ b, |x − y| < δ =⇒ |f (x) − f (y)| < See Below (1) Choose Pε to be See Below (2). Then U (Pε , f, α) − L(Pε , f, α) = n X (Mi − mi )∆αi i=1 where M i − mi = sup xi−1 ≤x≤xi f (x) − inf xi−1 ≤x≤xi f (x) = sup xi−1 ≤x,y≤xi [f (x) − f (y)] ≤ See Below (1) 2 MATH 321: Lecture #9 Choose (2) any partition with kPε k < δ. Then U (Pε , T, α) − L(Pε , T, α) = n X (Mi − mi )∆αi ≤ n X See Below (1) ∆αi i=1 i=1 = See Below (1) n X ∆αi < ε i=1 | {z } α(b)−α(a) Choosing (1) ε 2[α(b)−α(a)] completes the proof. Theorem. Let f : [a, b] → [−M, M ], α : [a, b] → R be increasing, and ϕ : [−M, M ] → R be continuous. If f ∈ R(α) on [a, b], then ϕ ◦ f ∈ R(α) on [a, b]. Hence ϕ ◦ f (x) = ϕ(f (x)). Before we prove this theorem, we would like to see two corollaries. Corollary 1. If f ∈ R(α) on [a, b], then |f |, f n ∈ R(α) for all n ∈ N. Corollary 2. If f, g ∈ R(α) on [a, b], then f g ∈ R(α) on [a, b]. Proof. Suppose f, g ∈ R(α). By linearity of integrals, f + g ∈ R(α). By Corollary 1, (f + g)2 , f 2 , and g 2 ∈ R(α). Note that f g = 21 [(f + g)2 − f 2 − g 2 ]. So by linearity, f g ∈ R(α). Proof of Main Theorem. Let ε > 0. Given that 1. ϕ is continuous on the compact set [−M.M ]. This means that ϕ is bounded, so ∃Mϕ such that |ϕ(y)| < M ϕ for all |y| ≤ M . Set ε0 = See Below (3), then ∃δ > 0 such that |x − y|δ =⇒ |ϕ(x) − ϕ(y)| < ε0 , for all |x| ≤ M. 2. f ∈ R(α) on [a, b]. Set η = See Below (4). Then ∃Pη such that U (Pη , f, α) − L(Pη , f, α) < η. Choose Pε = Pη . We will show that U (Pε , ϕ ◦ f, α) − L(Pε , ϕ ◦ f, α) < ε. We know that U (Pη , f, α) − L(Pη , f, α) = n X (Mi − mi )∆αi < η i=1 where Mi − mi = sup [f (x) − f (y)]. xi−1 ≤x,y≤xi We wish to show that U (Pη , ϕ ◦ f, α) − L(Pη , ϕ ◦ f, α) = n X i=1 (Mi∗ − m∗i )∆αi < ε where Mi∗ − m∗i = sup [ϕ(f (x)) − ϕ(f (y))]. xi−1 ≤x,y≤xi We know that if Mi − mi < δ, then Mi∗ − m∗i < ε0 . Next lecture, we shall reveal that (3) ε0 = (4) η = εδ 4Mϕ ε 2[α(b)−α(a)] and