Journal of Inequalities in Pure and Applied Mathematics http://jipam.vu.edu.au/ Volume 7, Issue 4, Article 151, 2006 ON AN INEQUALITY OF OSTROWSKI TYPE JOSIP PEČARIĆ AND ŠIME UNGAR FACULTY OF T EXTILE T ECHNOLOGY U NIVERSITY OF Z AGREB pecaric@hazu.hr D EPARTMENT OF M ATHEMATICS U NIVERSITY OF Z AGREB ungar@math.hr Received 22 August, 2006; accepted 28 August, 2006 Communicated by W.-S. Cheung A BSTRACT. We prove an inequality of Ostrowski type for p-norm, generalizing a result of Dragomir [1]. Key words and phrases: Ostrowski’s inequality, p-norm. 2000 Mathematics Subject Classification. 26D15, 26D10. 1. I NTRODUCTION For a differentiable function f : [a, b] → R, a · b > 0, Dragomir has in [1] proved, using Pompeiu’s mean value theorem [3], the following Ostrowski type inequality: Z b a + b f (x) 1 (1.1) · − f (t) dt ≤ D(x) · kf − ι f 0 k∞ , 2 x b−a a where ι(t) = t, t ∈ [a, b], and (1.2) D(x) = b − a 1 + |x| 4 a+b 2 x− b−a !2 . We are going to prove a general estimate with the p-norm, 1 ≤ p ≤ ∞, which will for p = ∞ give the Dragomir result. ISSN (electronic): 1443-5756 c 2006 Victoria University. All rights reserved. 225-06 2 J OSIP P E ČARI Ć AND Š IME U NGAR 2. T HE M AIN R ESULT Theorem 2.1. Let the function f : [a, b] → R be continuous on [a, b] and differentiable on (a, b) with 0 < a < b. Then for p1 + 1q = 1, with 1 ≤ p, q ≤ ∞, and all x ∈ [a, b], the following inequality holds: Z b 1 a + b f (x) (2.1) · − f (t) dt ≤ P U (x, p) · kf − ι f 0 kp 2 x b−a a where ι(t) = t, t ∈ [a, b], and (2.2) P U (x, p) = (b − a) 1 −1 p " · 1 x2−q − a1+q x1−2q q a2−q − x2−q + (1 − 2q)(2 − q) (1 − 2q)(1 + q) 2−q 1 # b − x2−q x2−q − b1+q x1−2q q + + . (1 − 2q)(2 − q) (1 − 2q)(1 + q) Note that in cases (p, q) = (1, ∞), (∞, 1), and (2, 2) the constant P U (x, p) has to be taken as the limit as p → 1, ∞, and 2, respectively. Proof. Define F : [1/b, 1/a] → R by F (t) := t f ( 1t ). The function F is continuous and is differentiable on (1/b, 1/a), and for all x1 , x2 ∈ [1/b, 1/a] we have Z x1 F (x1 ) − F (x2 ) = F 0 (t) dt x Z 2x1 1 1 0 1 1 (2.3) = f − f dt u := t t t t x2 Z 1/x1 1 (2.4) =− (f (u) − uf 0 (u)) 2 du . u 1/x2 Denote x1 =: 1/x and x2 =: 1/t. Then for all x, t ∈ [a, b] from (2.4) we get Z t 1 1 1 f (x) − f (t) = (f (u) − uf 0 (u)) 2 du, x t u x i.e., Z tf (x) − xf (t) = x t (2.5) x t (f (u) − uf 0 (u)) 1 du . u2 Integrating on t and dividing by x, we obtain Z b Z b Z t b2 − a2 f (x) 1 0 · − f (t) dt = t (f (u) − uf (u)) 2 du dt , 2 x u a a x and therefore (2.6) Z b b2 − a2 f (x) · − f (t) dt 2 x a Z b Z t t ≤ (f (u) − uf 0 (u)) 2 du dt u a x Z x Z x Z b Z t t t 0 = (f (u) − uf (u)) 2 du dt + (f (u) − uf 0 (u)) 2 du dt . u u a t x x J. Inequal. Pure and Appl. Math., 7(4) Art. 151, 2006 http://jipam.vu.edu.au/ O N AN I NEQUALITY OF O STROWSKI 3 TYPE First, consider the case 1 < p, q < ∞. Applying Hölder’s inequality, the sum in the last line of (2.5) is Z x Z x 1q p1 Z x Z x q t p 0 ≤ (2.7) |f (u) − uf (u)| du dt · du dt 2q a t a t u 1q Z b Z t p1 Z b Z t q t p + |f (u) − uf 0 (u)| du dt du dt · 2q x x x x u p1 Z b Z b p ≤ |f (u) − uf 0 (u)| du dt a a " Z x Z × a t x tq du u2q 1q Z b Z + dt x x t tq du u2q 1q # . dt The first factor in (2.7) equals Z b Z b p1 1 p (2.8) |f (u) − uf 0 (u)| du dt = (b − a) p kf − ι f 0 kp , a a and for the second factor, for p, q 6= 2, we get Z x Z x (2.9) a t tq du u2q 1q Z b Z dt + x = t x tq du u2q 1q dt 1 a −x x2−q − a1+q x1−2q q + (1 − 2q)(2 − q) (1 − 2q)(1 + q) 2−q 1 b − x2−q x2−q − b1+q x1−2q q + + . (1 − 2q)(2 − q) (1 − 2q)(1 + q) 2−q 2−q Putting (2.8) and (2.9) into (2.7) and dividing (2.6) and (2.7) by (b − a) gives the required inequality (2.1) in the case 1 < p, q < ∞, p, q 6= 2. For p = q = 2, instead of (2.9) we obtain Z x Z (2.10) a t x t2 du u4 12 t2 du dt 4 x x u 12 21 ! 3 3 3 3 x x a b ln + 3 −1 + ln + 3 −1 a x b x 12 Z b Z dt + = 1 3 t which is easily shown to be equal to the limit of the right hand side in (2.9) for q → 2, i. e. 12 12 ! x 3 a3 x 3 b3 1 ln (2.11) lim P U (x, p) = + 3 −1 + ln + 3 −1 . 1 p→2 a x b x 3(b − a) 2 Now consider the case p = ∞, q = 1. The last line in (2.6) is Z x Z x Z b Z t t t 0 (2.12) ≤ sup |f (u) − u f (u)| du dt + du dt 2 2 a≤u≤b a t u x x u 2 a + b2 0 +x−a−b . = kf − ι f k∞ · 2x J. Inequal. Pure and Appl. Math., 7(4) Art. 151, 2006 http://jipam.vu.edu.au/ 4 J OSIP P E ČARI Ć AND Š IME U NGAR Putting (2.12) into (2.6) and dividing by (b − a) gives 2 Z b a + b f (x) 1 1 a + b2 · − f (t) dt ≤ + x − a − b · kf − ι f 0 k∞ 2 x b−a a b−a 2x which reproves the Dragomir’s result [1]. It is easy to see that 2 a + b2 1 + x − a − b = lim P U (x, p) p→∞ b−a 2x proving (2.1) in the case p = ∞, q = 1. Finally consider the case p = 1, q = ∞. In this case the last line of (2.6) is Z x Z x t (2.13) ≤ |f (u) − uf 0 (u)| max 2 du dt t≤u≤x u a t a≤t≤x Z b Z t t + |f (u) − uf 0 (u)| max 2 du dt x≤u≤t u x x x≤t≤b Z bZ b 1 b 0 ≤ |f (u) − u f (u)| du dt · + a x2 a a 1 b = (b − a) + · kf − ι f 0 k1 . a x2 Appending (2.13) to (2.6) and dividing by (b − a) gives Z b a + b f (x) 1 1 b (2.14) · − f (t) dt ≤ + · kf − ι f 0 k1 . 2 x b−a a a x2 It is not too difficult to show that 1 b (2.15) lim P U (x, p) = + 2 , p→1 a x so (2.14) proves formula (2.1) for p = 1, q = ∞, proving the theorem. Remark 2.2. We have considered only the positive case, 0 < a < b. In the case a < b < 0, a similar but more cumbersome formula holds, where most a’s, b’s and x’s have to be replaced by their absolute values. 3. T HE W EIGHTED C ASE In the weighted case we have the following result: Theorem 3.1. Let the function f : [a, b] → R be continuous on [a, b] and differentiable on (a, b) with 0 < a < b, and let w : [a, b] → R be a nonnegative integrable function. Then for p1 + 1q = 1, with 1 ≤ p, q ≤ ∞, and all x ∈ [a, b], the following inequality holds: Z Z b f (x) b (3.1) t w(t) dt − f (t) w(t) dt x a a " 1 1q Z x Z x p (b − a) q q 0 1−2q q 1−q x t (w(t)) dt − t (w(t)) dt ≤ kf − ιf kp · 1 · a a (1 − 2q) q Z b 1q # Z b . + t1−q (w(t))q dt − x1−2q tq (w(t))q dt x J. Inequal. Pure and Appl. Math., 7(4) Art. 151, 2006 x http://jipam.vu.edu.au/ O N AN I NEQUALITY OF O STROWSKI 5 TYPE Proof. Multiplying (2.5) by w(t)/x and integrating on t, we get Z t Z Z b Z b f (x) b 1 0 t w(t) dt − f (t) w(t) dt = t w(t) (f (u) − u f (u)) 2 du dt , x a u a a x and as in the proof of Theorem 2.1 we have Z Z b f (x) b t w(t) dt − f (t) w(t) dt x a a Z b Z t Z x Z x t w(t) t w(t) 0 ≤ (f (u) − uf (u)) du dt + (f (u) − uf 0 (u)) du dt 2 u u2 a t x x Z x Z x p1 Z x Z x q 1q t (w(t))q p 0 ≤ |f (u) − u f (u)| du dt · du dt u2q a t a t Z b Z t p1 Z b Z t q 1q t (w(t))q p 0 + |f (u) − u f (u)| du dt du dt · u2q x x x x Z b Z b p1 p 0 ≤ |f (u) − u f (u)| du dt a a " Z x Z · a t t (w(t))q du u2q x q 1q Z b Z dt + x which gives (3.1). x t (w(t))q du u2q t q 1q # dt R EFERENCES [1] S.S. DRAGOMIR, An inequality of Ostrowski type via Pompeiu’s mean value theorem, J. of Inequal. in Pure and Appl. Math., 6(3) (2005), Art. 83. [ONLINE: http://jipam.vu.edu.au/ article.php?sid=532]. [2] A. OSTROWSKI, Über die Absolutabweichung einer differentienbaren Funktionen von ihren Integralmittelwert, Comment. Math. Hel., 10 (1938), 226–227. [3] D. POMPEIU, Sur une proposition analogue au théorème des accroissements finis, Mathematica (Cluj, Romania), 22 (1946), 143–146. J. Inequal. Pure and Appl. Math., 7(4) Art. 151, 2006 http://jipam.vu.edu.au/