Journal of Inequalities in Pure and Applied Mathematics http://jipam.vu.edu.au/ Volume 5, Issue 2, Article 25, 2004 ON HARDY-HILBERT’S INTEGRAL INEQUALITY W.T. SULAIMAN M OSUL U NIVERSITY C OLLEGE OF M ATHEMATICS AND C OMPUTER S CIENCE waadsulaiman@hotmail.com Received 17 April, 2003; accepted 18 January, 2004 Communicated by L. Pick A BSTRACT. In the present paper, by introducing some parameters, new forms of Hardy-Hilbert’s inequalities are given. Key words and phrases: Hardy-Hilbert’s integral inequality. 2000 Mathematics Subject Classification. 26D15. 1. I NTRODUCTION R∞ R∞ If p > 1, p1 + 1q = 1, f (t), g(t) ≥ 0, 0 < 0 f p (t)dt < ∞, and 0 < 0 g q (t)dt < ∞, then, the Hardy-Hilbert integral inequality is given by: Z ∞ p1 Z ∞ 1q Z ∞Z ∞ f (x)g(y) π p q (1.1) dxdy ≤ f (t)dt g (t)dt , x+y sin πp 0 0 0 0 where the constant sinπ π is the best possible (see [1]). p Yang [2] and [3] gave the following generalization of (1.1) Z ∞Z ∞ f (x)g(y) (1.2) dxdy (x + y − 2α)λ 0 0 Z ∞ p1 Z 1 1 p q 1−λ p (t − α) f (t)dt ≤ Kλ (p)Kλ (q) 0 where ∞ 1q g (t)dt , 1−λ q (t − α) 0 1 u r −1 1 1 1 Kλ (r) = du = B ,λ − 0 < λ ≤ 1, λ > > 0, λ (1 + u) r r r 0 B is the beta function defined by Z 1 B(p, q) = xp−1 (1 − x)q−1 dx, p, q > 0 Z ∞ 0 ISSN (electronic): 1443-5756 c 2004 Victoria University. All rights reserved. 053-03 2 W.T. S ULAIMAN and, for 0 < T < ∞, Z T T Z (1.3) 0 0 f (x)g(y) dxdy (x + y)λ ! 12 Z T" λ2 # 1 t λ λ t1−λ f 2 (t)dt ≤β , 1− 2 2 2 T α ! 12 λ2 # Z T" 1 t t1−λ g 2 (t)dt . × 1− 2 T α In the present paper, by introducing some parameters, new forms of Hardy-Hilbert’s inequalities are given. 2. N EW R ESULTS We state and prove the following: Lemma 2.1. Let λ > 0, p, q, r > 1, and assume that 1 p + Z 1 q + 1 r = 1, f (t), g(t), h(t) ≥ 0, λi (t) ≥ 0, i = p, q, r, b λpp (t)f p (t)dt < ∞, 0< a Z d λqq (t)g q (t)dt < ∞ 0< c and Z 0< d λrr (t)hr (t)dt < ∞. c Then the two following inequalities are equivalent Z bZ d Z k (2.1) a c e f (x)g(y)h(z) dxdydz hλ (x, y, z) Z b p1 Z p p ≤K λp (t)f (t)dt d 1q Z c a r1 k λqq (t)g q (t)dt λrr (t)hr (t)dt , e where K = K(λ, p, q, r) is a constant, and Z b qr q+r Z d Z λp (x) (2.2) a c e k qr q+r g(y)h(z) dydz dx hλ (x, y, z) r Z Z d q+r qr q+r ≤K λqq (t)g q (t)dt c J. Inequal. Pure and Appl. Math., 5(2) Art. 25, 2004 k q q+r λr (t)hr (t)dt . r e http://jipam.vu.edu.au/ O N H ARDY-H ILBERT ’ S I NTEGRAL I NEQUALITY 3 Proof. Suppose (2.2) is satisfied, then Z bZ dZ k f (x)g(y)h(z) dxdydz hλ (x, y, z) a c e Z b Z dZ k g(y)h(z) −1 = λp (x)f (x) λp (x) dydz dx λ a c e h (x, y, z) ! q+r qr q+r Z b p1 Z b qr Z d Z k qr g(y)h(z) q+r p p dydz dx ≤ λp (x)f (x)dx λp λ a c e h (x, y, z) a Z b p1 Z d 1q Z k r1 p p q q r r ≤K λp (t)f (t)dt λq (t)g (t)dt λr (t)h (t)dt . c a e Now, suppose that (2.1) is satisfied, then qr q+r Z b qr Z d Z k g(y)h(z) q+r λp dydz dx λ a c e h (x, y, z) qr q+r Z d Z k Z bZ dZ k −1 qr g(y)h(z) q+r g(y)h(z) = .λp dydz dxdydz λ λ a c e h (x, y, z) c e h (x, y, z) Z d 1q Z k r1 q q r r ≤K λq (y)g (y)dy λr (z)h (z)dz c e Z b Z d Z k g(y)h(z) dydz λ a c e h (x, y, z) Z d 1q Z k r1 q q r r =K λq (y)g (y)dy λr (z)h (z)dz qr −p q+r λpp (x)λp × (x) c ! p1 qr p( q+r −1) dx e Z b × qr q+r Z d k Z λp (x) c a e g(y)h(z) dydz hλ (x, y, z) ! p1 qr q+r dx , therefore Z b qr q+r Z d Z λp (x) c a e k g(y)h(z) dydz hλ (x, y, z) ! 1q + r1 qr q+r dx Z ≤k d 1q Z k λqq (t)g q (t)dt c , e and the desired equivalence is proved. Lemma 2.2. r1 λrr (t)hr (t)dt (a) Let 0 ≤ y ≤ 1, α > 0, f ≥ 0. If we define the function Z y −α g(y) = y f (x)dx, 0 then g(y) ≥ g(1). (b) Let y ≥ 1, α > 0, f ≥ 0. Defining the function, Z y −α h(y) = y f (x)dx, 0 we have h(y) ≥ h(1). J. Inequal. Pure and Appl. Math., 5(2) Art. 25, 2004 http://jipam.vu.edu.au/ 4 W.T. S ULAIMAN (a) Let x = 1t , then Proof. g(y) = y −α Z ∞ f y −1 1 t t2 dt . We observe also that g 0 (y) = y −α −y 2 f (y) + Z ∞ y −1 f 1 t t2 ! dt (−α)y −α−1 ≤ 0, therefore g is non-increasing, which implies g(y) ≥ g(1). (b) We obviously have Z y 0 α h (y) = y f (y) + f (x)dx αy α−1 ≥ 0, 0 therefore h is non-decreasing, and hence h(y) ≥ h(1). The following result may be stated as well. Theorem 2.3. Let f (t), g(t), h(t) ≥ 0, p, q, r > 1, p1 + 1q + 1r = 1, 2 < λ < 3, γ > µ max {p, q, r}, and 1 1 1 λ−1 λ−1 λ−1 max − , − , − < µ < min , , . p q r p q r T Z (t − α)2−λ f p (t)dt < ∞, 0< α T Z (t − α)2−λ g q (t)dt < ∞ 0< α and Z T (t − α)2−λ hr (t)dt < ∞, 0< α then Z T T Z T Z (2.3) α α α Z T f (x)g(y)h(z) dxdydz (x + y + z)λ Z T p1 ≤ φ(t, µ, λ, p)(t − α)2−λ f p (t)dt α × 1q Z g (t)dt T 2−λ q φ(t, µ, λ, q)(t − α) α r1 h (t)dt , 2−λ r φ(t, µ, λ, r)(t − α) α where φ(t, µ, λ, j) 2γ Z t−α γ 1 uλ−3 t−α = B(λ−µj−1, µj+1) B(λ − 2, 1 − µj) − ( ) du − λ−µj−1 T −α T −α 0 (1 + u) Z 1 λ−µj−2 Z 1 u uγ−µj−2 × du du, j = p, q, r. λ λ−µj−1 0 (1 + u) 0 (1 + u) J. Inequal. Pure and Appl. Math., 5(2) Art. 25, 2004 http://jipam.vu.edu.au/ O N H ARDY-H ILBERT ’ S I NTEGRAL I NEQUALITY 5 Proof. The proof is as follows. We have Z TZ TZ T f (x)g(x)h(z) dxdydz λ α α α (x + y + z) µ µ Z T Z T Z T f (x) z−α y−α µ h(z) x−α g(y) x−α y−α z−α dxdydz = λ/p (x + y + z)λ/q (x + y + z)λ/r α α α (x + y + z) µp p1 ! 1q µq Z TZ TZ T q Z T Z T Z T f p (x) z−α g (y) x−α y−α z−α dxdydz dxdydz ≤ λ (x + y + z)λ (x + y + z) α α α α α α Z T Z T Z T × α α 1 p 1 q α ! r1 y−α µr hr (z) x−α dxdydz (x + y + z)λ 1 r = F G H , say. Then we have Z T F = (x − α)2−λ f p (x)dx α Z T α (1 y−α −µp x−α y−α λ−µp−1 + x−α ) dy x−α Z T α −µp z−α x+y−2α z−α (1 + x+y−2α )λ dz . x + y − 2α Now by Lemma 2.2, we can state that µp z−α Z T x+y−2α dz z−α λ 0 (1 + x+y−2α ) x + y − 2α Z T −α x+y−2α uµp = du (1 + u)λ 0 Z T −α y−α uµp ≤ du (1 + u)λ 0 Z ∞ λ−µp−2 u = du y−α (1 + u)λ T −α γ −γ Z y−α ! λ−µp−2 Z ∞ T −α y−α y−α u = − du T −α T −α (1 + u)λ 0 0 γ Z 1 λ−µp−2 y−α u ≤ B(λ − µp − 1, µp + 1) − du . λ T −α 0 (1 + u) Therefore T Z F ≤ 2−λ p (x − α) Z T f (x)dx y−α −µp dy x−α y−α λ−µp−1 x−α + x−α ) γ Z 1 λ−µp−2 α (1 × B(λ − µp − 1, µp + 1) − α Z T = (x − α)2−λ f p (x)dx α J. Inequal. Pure and Appl. Math., 5(2) Art. 25, 2004 Z 0 T −α x−α y−α T −α 0 u du (1 + u)λ −µp u du (1 + u)λ−µp−1 http://jipam.vu.edu.au/ 6 W.T. S ULAIMAN γ Z 1 λ−µp−2 u x−α uγ du × B(λ − µp − 1, µp + 1) − λ T −α 0 (1 + u) " Z T Z T −α x−α u−µp 2−λ p = (x − α) f (x)dx× B(λ − µp − 1, µp + 1) − (1 + u)λ−µp−1 α 0 # γ Z 1 λ−µp−2 Z T −α x−α u uγ−µp x−α du − du λ (1 + u)λ−µp−1 T −α 0 (1 + u) 0 " Z T Z ∞ uλ−3 2−λ p = (x − α) f (x)dx× B(λ − µp − 1, µp + 1) − du x−α (1 + u)λ−µp−1 α T −α # γ Z 1 λ−µp−2 Z T −α x−α u uγ−µp x−α − du du λ T −α (1 + u)λ−µp−1 0 0 (1 + u) Z T = α (x − α)2−λ f p (x)dx " Z ∞ × B(λ − µp − 1, µp + 1) − − 0 − Z x−α T −α γ Z 0 1 uλ−µp−2 du (1 + u)λ Z Z 0 T −α x−α 0 x−α T −α ! uλ−3 du (1 + u)λ−µp−1 # uγ−µp du (1 + u)λ−µp−1 T = α (x − α)2−λ f p (x)dx " ( × B(λ − µp − 1, µp + 1) B(λ − 2, 1 − µp) − x−α T −α γ x−α T −α −γ x−α T −α 2γ Z 1 λ−µp−2 uλ−3 x−α u × du − du λ−µp−1 λ (1 + u) T −α 0 0 (1 + u) γ Z T −α x−α T −α uγ−µp × x−α (1 + u)λ−µp−1 0 γ Z T x−α 2−λ p ≤ (x − α) f (x)dx× B(λ − µp − 1, µp + 1) B(λ − 2, 1 − µp) − T −α α Z 1 Z Z 2γ 1 λ−µp−2 1 uλ−3 x−α u uγ−µp × du − du λ−µp−1 λ λ−µp−1 T −α 0 (1 + u) 0 (1 + u) 0 (1 + u) Z T = φ(x, λ, µ, p)(x − α)2−λ f p (x)dx, Z α where φ(x, λ, µ, p) γ Z 1 x−α uλ−3 = B(λ − µp − 1, µp + 1) B(λ − 2, 1 − µp) − du λ−µp−1 T −α 0 (1 + u) # 2γ Z 1 λ−µp−2 Z 1 x−α u uγ−µp − du du . λ λ−µp−1 T −α 0 (1 + u) 0 (1 + u) J. Inequal. Pure and Appl. Math., 5(2) Art. 25, 2004 http://jipam.vu.edu.au/ O N H ARDY-H ILBERT ’ S I NTEGRAL I NEQUALITY 7 Similarly T Z φ(y, λ, µ, q)(y − α)2−λ g q (y)dy, G= α and Z T φ(z, λ, µ, r)(z − α)2−λ hr (z)dz. H= α This completes the proof. Corollary 2.4. Let f (t), g(t), h(z) ≥ 0, p, q, r > 1, p1 + 1q + 1r = 1, 2 < λ < 3, and 1 1 1 λ−1 λ−1 λ−1 < µ < min , , . max − , − , − p q r p q r If ∞ Z (t − α)2−λ f p (t)dt < ∞, 0< Zα∞ 0< (t − α)2−λ g q (t)dt < ∞ α and Z ∞ 0< (t − α)2−λ hr (t)dt < ∞, α then we have the inequality Z ∞Z ∞Z ∞ f (x)g(y)h(z) dxdydz (2.4) (x + y + z)λ α α α Z ∞ p1 Z 2−λ p ≤K (t − α) f (t)dt α 1q g (t)dt ∞ 2−λ q (t − α) α Z ∞ × r1 (t − α)2−λ hr (t)dt , α where K= Y B 1/j (λ − µj − 1, µj + 1)B 1/j (λ − 2, 1 − µj) j=p,q,r and Z ∞ (x − α) (2.5) α qr q+r g(y)h(z) dxdydz dx (x + y + z)λ α α q r Z Z ∞ q+r q+r ∞ qr 2−λ q 2−λ r ≤ K q+r (t − α) g (t)dt (t − α) h (t)dt . qr(λ−2) p(q+r) Z ∞ ∞ Z α α The inequalities (2.4)and (2.5) are equivalent. Proof. Follows from Theorem 2.3 and Lemma 2.1, on choosing γ = 1, T = ∞, and λj (t) = 2−λ (t − α) j . We omit the details. J. Inequal. Pure and Appl. Math., 5(2) Art. 25, 2004 http://jipam.vu.edu.au/ 8 W.T. S ULAIMAN Corollary 2.5. Let f (t), g(t), h(t) ≥ 0, 2 < λ < 3, γ > 3µ, and − 13 < µ < Z T 0< (t − α)2−λ f 3 (t)dt < ∞, α Z T 0< (t − α)2−λ g 3 (t)dt < ∞ λ−1 . 3 If α and T Z (t − α)2−λ h3 (t)dt < ∞, 0< α then Z T T Z Z T (2.6) α α Z α f (x)g(x)h(z) dxdydz (x + y + z)λ 13 Z f (t)dt T T 2−λ 3 ≤ φ(t, λ, µ, 3)(t − α) α 13 g (t)dt 2−λ 3 φ(t, λ, µ, 3)(t − α) α Z T × 13 φ(t, λ, µ, 3)(t − α)2−λ h3 (t)dt , α and 12 g(y)h(z) φ (t, λ, µ, 3)(x − α) dydz dx λ α α α (x + y + z) Z T 13 Z T 13 ≤ φ(t, λ, µ, 3)(t − α)2−λ g 3 (t)dt φ(t, λ, µ, 3)(t − α)2−λ h3 (t)dt . Z (2.7) T − 12 λ −1 2 Z T Z T α α The inequalities (2.6) and (2.7) are equivalent. Proof. Follows from Theorem 2.3 and Lemma 2.1, by putting p = q = r = 3, and λj (t) = 2−λ 1 (t − α) 3 φ 3 (t, λ, µ, 3), j = p, q, r. Note. In Corollary 2.5, we may take as a special case µ = 1 − λ 3 to obtain φ(t, λ, 1 − λ/3, 3) γ 1 t−α = B(2λ − 4, 4 − λ)B(λ − 2, λ − 2) 1 − 2 T −α 2γ Z 1 2λ−5 Z 1 t−α u uγ+λ−3 − du du. λ 2λ−4 T −α 0 (1 + u) 0 (1 + u) Corollary 2.6. Let f (t), g(t), h(t) ≥ 0, 2 < λ < 3, and − 13 < µ < Z ∞ 0< (t − α)2−λ f 3 (t)dt < ∞, Zα∞ 0< (t − α)2−λ g 3 (t)dt < ∞ λ−1 . 3 If α and Z ∞ 0< (t − α)2−λ h3 (t)dt < ∞, α J. Inequal. Pure and Appl. Math., 5(2) Art. 25, 2004 http://jipam.vu.edu.au/ O N H ARDY-H ILBERT ’ S I NTEGRAL I NEQUALITY 9 then Z ∞ Z ∞ (2.8) α α Z ∞ f (x)g(y)h(z) dxdydz (x + y + z)λ α Z ∞ 13 Z 2−λ 3 ≤K (t − α) f (t)dt α 13 g (t)dt ∞ 2−λ 3 (t − α) α Z ∞ 13 h (t)dt , 2−λ 3 × (t − α) α where K = (B(λ − 3µ − 1, 3µ + 1)B(λ − 2, 1 − 3µ) and Z ∞ − 21 Z ∞ (x − α) (2.9) α α 12 g(y)h(z) dydz dx (x + y + z)λ α Z ∞ 12 3/2 2−λ 3 ≤K (t − α) g (t)dt Z ∞ α Z ∞ × 12 h (t)dt . 2−λ 3 (t − α) α The inequalities (2.8) and (2.9) are equivalent. Proof. Follows from Corollary 2.4 and Lemma 2.1, by putting p = q = r = 3, T = ∞. R EFERENCES [1] G.H. HARDY, J.E. LITTLEWOOD UK, (1952). AND G. POLYA, Inequalities, Cambridge Univ., Cambridge, [2] B. YANG, On generalization of Hardy-Hilbert’s integral inequality, Acta. Math. Sinica, 41 (1998), 839–844. [3] B. YANG, On Hilbert’s integral inequality, J. Math. Anal. Appl., 220 (2000), 778–785. J. Inequal. Pure and Appl. Math., 5(2) Art. 25, 2004 http://jipam.vu.edu.au/