ebook img

A sharp weighted Wirtinger inequality PDF

0.11 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview A sharp weighted Wirtinger inequality

A sharp weighted Wirtinger inequality Tonia Ricciardi∗ Dipartimento di Matematica e Applicazioni 5 Universit`a di Napoli Federico II 0 Via Cintia, 80126 Naples, Italy 0 2 fax: +39 081 675665 [email protected] n a J 4 Abstract ] P We obtain a sharp estimate for the best constant C > 0 in the Wir- A tinger typeinequality . h 2π 2π at γpw2 ≤C γqw′2 m Z0 Z0 [ where γ is bounded above and below away from zero, w is 2π-periodic and such that 2πγpw = 0, and p+q ≥ 0. Our result generalizes an 1 0 inequality of Piccinini and Spagnolo. v R 4 Let C(a,b) > 0 denote the best constant in the following weighted Wirtinger 4 type inequality: 0 1 0 2π 2π 5 (1) aw2 C(a,b) bw′2, ≤ 0 Z0 Z0 h/ where w H1 (R) is 2π-periodic and satisfies the constraint ∈ loc t a 2π m (2) aw=0, v: Z0 Xi and a,b with ∈B r a = a L∞(R) : a is 2π periodic and infa>0 . B { ∈ − } Hereandinwhatfollows,foreverymeasurablefunctionawedenotebyinfaand supatheessentiallowerboundandtheessentialupperboundofa,respectively. For every L>1, we denote (L)= a L (0,2π) : a is 2π periodic, infa=1 and supa=L . ∞ B { ∈ − } Our aim in this note is to prove: ∗Supported in part by Regione Campania L.R. 5/02 and by the MIUR National Project Variational Methods and Nonlinear DifferentialEquations. 1 Theorem 1. Suppose a=γp and b=γq for some γ (M), M >1, and for some p,q R such that p+q 0. Then ∈B ∈ ≥ 1 2πγ(p q)/2 2 (3) C(γp,γq) 2π 0 − . ≤ π4 arctaRn M−(p+q)/4 ! (cid:0) (cid:1) If p+q>0, then equality holds in (3) if and only if γ(θ)=γ¯ (θ+ϕ) for some p,q ϕ R, where ∈ 1, if 0 θ <c π, π θ <π+c π γ¯ (θ)= ≤ p,q2 ≤ p,q2 , p,q (M, if cp,qπ2 ≤θ <π, π+cp,qπ2 ≤θ <2π with 2 c = . p,q 1+M (p q)/2 − − Furthermore,equalityholdsin (1)–(2)witha(θ)=γ¯p (θ+ϕ)andb(θ)=γ¯q (θ+ p,q p,q ϕ) if and only if w(θ)=w¯ (θ+ϕ) where p,q w¯ (θ)= p,q sin √µ c−p,1qθ− π4 , if 0 θ <c π  (cid:2) (cid:0) (cid:1)(cid:3) ≤ p,q2 =M−s−in(p+√q)µ/4cπo+s(cid:2)c√−p,µ1q((cid:0)θπ2−+πc)−p−,1qM54π(p−ifq,)c/p2,(qθπ2−≤cpθ,q<π2)π− 34π(cid:1)(cid:3), , if π θ <π+c π (cid:2) (cid:0) (cid:1)(cid:3) ≤ p,q2 and µ=−(4M/π−)(pa+rcqt)/a4ncMos(cid:2)√(p+µq(cid:0))322π.+c−p,1qMi(fp−πq+)/2c(pθ,q−π2 π≤−θc<p,q2ππ2)− 74π(cid:1)(cid:3), − If p+q = 0, then (3) is an equality for any weight function γ. Equality is (cid:0) (cid:1) attained in (1)–(2) with a=γp and b=γ p if and only if − 2π θ w(θ)=Ccos γp+ϕ , 02πγp Z0 ! for some C =0 and ϕ R. R 6 ∈ Note that when p = q = 0, Theorem 1 yields C(1,1) = 1 according to the classicalWirtinger inequality. When p=q =0, the estimate (3) reduces to the 6 estimate obtained by Piccinini and Spagnolo in [4]. More related results may be found in [1, 2, 3] and in the references therein. We begin by recalling in the following lemma the Wirtinger inequality of Piccinini and Spagnolo [4]. Lemma 1 ([4]). Suppose b=a (L). Then, ∈B 2 (4) C(a,a) 4 arctanL−1/2 − . ≤ π (cid:18) (cid:19) 2 Equality holds in (4) if and only if a(θ)=a¯(θ+ϕ) for some ϕ R, where a¯ is ∈ defined by 1, if 0 θ < π, π θ < 3π (5) a¯(θ)= ≤ 2 ≤ 2 (L, if π θ <π, 3π θ <2π 2 ≤ 2 ≤ and equality holds in (1)–(2) with a(θ)=b(θ)=a¯(θ+ϕ) if and only if w(θ)= w¯(θ+ϕ), where sin √λ θ π , if 0 θ < π − 4 ≤ 2 (6) w¯(θ)=L−−s1ihn/2c√o(cid:0)λsh√θλ−(cid:0)(cid:1)5θi4π− 3,4π(cid:1)i, iiff ππ2 ≤≤θθ<<3π2π , where λ= 4π 1arcta−nLL−11h//22co2(cid:0)s.h√λ(cid:0)θ(cid:1)i− 74π(cid:1)i, if 32π ≤θ <2π − − In order(cid:0)to prove Theorem(cid:1)1, we need the following lemma, which yields an estimate for C(a,b) for arbitrary weight functions a,b. Lemma 2. Let a,b . The following estimate holds: ∈B 2 1 2π√ab 1 (7) C(a,b) 2π 0 − . ≤ 1/4 4 arctaRn infab π supab   (cid:16) (cid:17)  If √ab (L), L>1, then ∈B (8) 2 1 C2π(a√,ba)b 1 2 =√a′bs′up(L) 1 C2π(a√′,ab′b) 1 2 =(cid:18)π4 arctanL−1/2(cid:19)− 2π 0 − ∈B 2π 0 ′ ′− (cid:16) (cid:17) (cid:16) (cid:17) R R if and only if the following equation is satisfied: (9) a(θ(τ))b(θ(τ)) =a¯2(τ +ϕ) a.e. τ (0,2π), for some ϕ R, ∈ ∈ where θ(τ) is the homeomorphism of R defined by 1 θ a(θ˜) (10) τ(θ)= dθ˜, c Z0 sb(θ˜) c is defined by 1 2π a(θ˜) (11) c= dθ˜, 2π Z0 sb(θ˜) and a¯ is the function defined in Lemma 1. 2 If b = a 1, then C(a,a 1) = (2π) 1 2πa and equality is attained in − − − 0 (1)–(2) with b=a−1 if and only if w(cid:16)(θ)=CRcos(2(cid:17)π( 02πa)−1 0θa+ϕ) for some C =0 and ϕ R. 6 ∈ R R 3 Proof. Under the change of variables θ = θ(τ) defined by (10)–(11), setting α(τ)=a(θ(τ)), β(τ)=b(θ(τ)), ξ(τ)=w(θ(τ)), we obtain αθ′ =c αβ, βθ′−1 =c−1 αβ, and therefore: p p 2π 2π aw2dθ = αθ ξ2dτ =c αβξ2dτ ′ Z0 Z0 Z 2π 2π p awdθ = αθ ξdτ =c αβξdτ =0 ′ Z0 Z0 Z 2π 2π p bw′2dθ = βθ′−1ξ′2dτ =c−1 αβξ′2dτ. Z0 Z0 Z p Upon substitution, (1)–(2) takes the form: 2π C(a,b) 2π (12) Z0 αβξ2dτ ≤ 1 2π√ab 1 2 Z0 αβξ′2dτ, p 2π 0 − p with constraint (cid:16) R (cid:17) 2π (13) αβξdτ =0. Z0 p If √ab (L), in view of Lemma 1 we obtain ∈B 2 − C(a,b) 4 inf√αβ (14) =C( αβ, αβ) arctan 1 2π√ab 1 2 ≤π ssup√αβ 2π 0 − p p   (cid:16) R (cid:17) 4 infab 1/4 −2 = arctan . π supab (cid:18) (cid:19) ! This yields (7). Moreover, we have C(√αβ,√αβ) = (4/π)arctanL−1/2 −2 if and only if α(τ)β(τ) = a¯(τ +ϕ), for some ϕ R. That is, (8) holds if and ∈ (cid:0) (cid:1) only if (9) holds. p If b=a 1, then (12)–(13) takes the form − 2π C(a,a 1) 2π ξ2dτ − ξ2dτ Z0 ≤ 1 2πa 2 Z0 ′ 2π 0 (cid:16) (cid:17) with constraint R 2π ξdτ =0. Z0 Therefore, by the classical Wirtinger inequality, 1 2π 2 C(a,a−1)= a 2π (cid:18) Z0 (cid:19) andequality holdsin(1)–(2)withb=a 1 ifandonlyif ξ(τ)=Ccos(τ+ϕ)for − someC =0andϕ R,thatis,ifandonlyifw(θ)=Ccos(2π( 2πa) 1 θa+ϕ), 6 ∈ 0 − 0 as asserted. R R 4 Lemma 3. Suppose a,b satisfy √ab (L), L > 1, and (9), where θ(τ) is ∈ B defined in (10) and c is defined by (11). Suppose (15) a=γp, b=γq for some γ (M), with M = L2/(p+q), and for some p,q R such that p+q >0. T∈henBγ(θ)=γ¯ (θ+ϕ) for some ϕ R, where γ¯ ∈is the function p,q p,q ∈ defined in Theorem 1. Proof. When p+q > 0, we have γ(p+q)/2 (L). In view of (9) and (15) we ∈ B have γ(θ(τ))=a¯2/(p+q)(τ +ψ), τ R ∀ ∈ for some ψ R. It follows that ∈ τ b(θ(τ¯)) τ (16) θ(τ)=c dτ¯=c a¯ (p q)/(p+q)(τ¯+ψ)dτ¯ − − Z0 sa(θ(τ¯)) Z0 and, in view of the 2π-periodicity of a and b, 1 1 2π b(θ(τ¯)) − 1 2π −1 c= dτ¯ = a¯ (p q)/(p+q)(τ¯)dτ¯ . − − 2π Z0 sa(θ(τ¯)) ! (cid:18)2π Z0 (cid:19) Setting τ h (τ)=c a¯ (p q)/(p+q)(τ¯)dτ¯, p,q − − Z0 we have θ(τ ψ)=h (τ) h (ψ) for every τ R, and consequently τ(θ)= p,q p,q − − ∈ h 1(θ+h (ψ)) ψ. Inviewofthedefinitionofa¯ withL=M(p+q)/2,wehave: −p,q p,q − τ a¯ (p q)/(p+q)(τ¯)dτ¯= − − Z0 τ, if 0 τ < π ≤ 2 π +M (p q)/2(τ π), if π τ <π =2 − − − 2 2 ≤ . π2(1+M−(p−q)/2)+τ −π, if π ≤τ < 32π π(2+M (p q)/2)+M (p q)/2(τ 3π), if 3π τ <2π 2 − − − − − 2 2 ≤ In particular, we derive 2 c= =c . 1+M (p q)/2 p,q − − It follows that h (τ) is the piecewise linear homeomorphism of R defined in p,q [0,2π) by c τ, if 0 τ < π p,q ≤ 2 c π +M (p q)/2(τ π) , if π τ <π h (τ)= p,q 2 − − − 2 2 ≤ p,q cp,q(cid:2)π2(1+M−(p−q)/2)+τ −(cid:3) π , if π ≤τ < 32π c π(2+M (p q)/2)+M (p q)/2(τ 3π) , if 3π τ <2π p,q(cid:2)2 − − − (cid:3)− − 2 2 ≤  (cid:2) (cid:3) 5 andby h (τ+2πn)=2πn+h (τ), for any τ [0,2π)and for any integern. p,q p,q ∈ Inversion yields c 1θ, if 0 θ <c π −p,q ≤ p,q2 π +c 1M(p q)/2(θ c π), if c π θ <π h−p,1q(θ)=π2+c−p−p,,1qq(θ−−π), − p,q2 if πp,≤q2θ≤<π+cp,qπ2 , 3π +c 1M(p q)/2(θ π c π), if π+c π θ <2π 2 −p,q − − − p,q2 p,q2 ≤ for θ [0,2π) and h 1(θ + 2πn) = 2πn + h 1(θ) for any τ [0,2π) and ∈ −p,q −p,q ∈ for any integer n. Substitution yields γ(θ) = a¯2/(p+q) h 1(θ+h (ψ)) = −p,q p,q a¯2/(p+q) h 1(θ+ϕ) =γ¯ (θ+ϕ), with ϕ=h (ψ). −p,q p,q p,q (cid:0) (cid:1) Now(cid:0)we can prov(cid:1)e Theorem 1. Proof of Theorem 1. Estimate (7) with a = γp and b = γq yields (3). Suppose p+q >0. In view of Lemma 2 and Lemma 3 we have 1 C2(πγγp(,pγqq))/2 2 =(cid:18)π4 arctanM−(p+q)/4(cid:19)−2 2π 0 − (cid:16) (cid:17) R if and only if γ(θ)=γ¯ (θ+ϕ) for some ϕ R. Equality is attainedin (1)–(2) p,q ∈ witha(θ)=γ¯p (θ+ϕ)andb(θ)=γ¯q (θ+ϕ) ifandonlyif w(θ)=w¯ (θ+ϕ). p,q p,q p,q If p + q = 0, then the conclusion follows by Lemma 2 with a = γp and b=γ p. − Acknowledgements I am grateful to Professor Carlo Sbordone for many useful and stimulating discussions. References [1] P.R.Beesack,IntegralinequalitiesoftheWirtingertype,DukeMath.Jour. 25 (1958), 477–498. [2] G.CroceandB.Dacorogna,OnageneralizedWirtingerinequality,Discrete Cont. Dynam. Systems 9 No. 5 (2003), 1329–1341. [3] B. Dacorogna, W. Gangbo and N. Sub´ia, Sur une g´en´eralisation de l’in´egalit´e de Wirtinger, Ann. Inst. H. Poincar´e Anal. Non Lin´eaire 9 (1992), 29–50. [4] L.C. Piccinini and S. Spagnolo, On the Ho¨lder continuity of solutions of second order elliptic equations in two variables, Ann. Scuola Norm. Sup. Pisa 26 No. 2 (1972), 391–402. 6

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.