A NEW WEBER TYPE INTEGRAL EQUATION RELATED TO THE WEBER-TITCHMARSH PROBLEM S.YAKUBOVICH 7 1 0 2 n ABSTRACT. Wederivesolvability conditions andclosed-formsolutionfortheWebertypeintegral equation, a related tothefamiliar Weber-Orrintegral transforms andtheoldWeber-Titchmarshproblem (posedinProc. J Lond.Math.Soc.22(2)(1924),pp.15,16),recentlysolvedbytheauthor.Ourmethodinvolvespropertiesofthe 5 inverseMellintransformofintegrablefunctions.TheMellin-Parsevalequalityandsomeintegrals,involvingthe 1 Gausshypergeometricfunctionareused. ] A C Recently,theauthorgavesolvabilityconditionsandclosed-formsolutionfortheclassicalWeberequation . [7] h t ¥ ma Cn (xx ,ax )g(x )dx = f(x), (1) Z0 [ where f(x)isagivenfunctionon[a,¥ ), a>0,g(x),x R+shouldbedeterminedandthekernel ∈ 1 Cn (a ,b )=Jn (a )Yn (b ) Yn (a )Jn (b ) (2) − v 8 involvesBesselfunctionsofthefirstandsecondkindJn (z),Yn (z), n C[1],Vol. II.Itwassolvedformally ∈ 0 byTitchmarshin1924andposedasanopenproblem(see[3],p. 15.)todescribeaclassofcomplex-valued 0 functionsg(x), x R ,whichcanbeexpandedintermsofthefollowingrepeatedintegral + 4 ∈ 0 x ¥ ¥ 1. g(x)= Jn2(ax)+Yn2(ax)Za Cn (xt,xa)tZ0 Cn (tx ,ax )g(x )dx dt, x>0. (3) 0 Expansion(3)is related to the familiar Weber-Orr integralexpansionsof an arbitraryfunction f(x) as re- 7 peatedintegrals 1 Xiv: f(x)=Z0¥ Jn2t(aCtn)(+xtY,na2t()at)Za¥ Cn (x t,at)x f(x )dx dt, (4) ar f(x)=Za¥ Cn (xt,xa)tZ0¥ Jn2(Caxn ()t+x ,Yan2x()ax )x f(x )dx dt, (5) which are differentfrom (3). Our method is based on the use of the Mellin transform [4]. Precisely, the MellintransformisdefinedinLm ,p(R+), 1<p 2bytheintegral ≤ ¥ f∗(s)= f(x)xs−1dx, (6) Z0 Date:January17,2017. 2000MathematicsSubjectClassification. Primary44A15,44A35,33C10;Secondary33C05,45E99. Keywordsandphrases. Weber-Orrintegraltransforms,Mellintransform,Besselfunctions,Gauss’shypergeometricfunction. 1 2 S.Yakubovich being convergentin mean with respect to the norm in L (m i¥ ,m +i¥ ), q= p/(p 1). Moreover, the q Parsevalequalityholdsfor f Lm ,p(R+), g L1 m ,q(R+) − − ∈ ∈ − ¥ 1 m +i¥ Z0 f(x)g(x)dx= 2p iZm i¥ f∗(s)g∗(1−s)ds. (7) − TheinverseMellintransformisgivenaccordingly 1 m +i¥ f(x)= f (s)x sds, (8) 2p iZm i¥ ∗ − − wheretheintegralconvergesinmeanwithrespecttothenorminLm ,p(R+) ¥ 1/p f m ,p= f(x) pxm p−1dx . (9) || || (cid:18)Z0 | | (cid:19) In particular, letting m =1/p we get the usual space L (R ). We will modify the definition of a special 1 + classoffunctionsrelatedtotheMellintransform(6)anditsinversion(8),whichwasintroducedin[5],[6]. Indeed,wehave Definition1. DenotebyM 1(L )thespaceoffunctions f(x), x R , beingrepresentablebyinverse − c + Mellintransform(8)ofintegrablefunctionsF(s) L (c)onthevertic∈allinec= s C:m =Res=c . 1 0 ThespaceM 1(L )withtheusualoperations∈ofadditionandmultiplicationb{y∈scalarisalinearvec}tor − c space. IfthenorminM 1(L )isintroducedbytheformula − c 1 +¥ f = F(c +it) dt, (10) M−1(Lc) 2p Z ¥ | 0 | (cid:12)(cid:12) (cid:12)(cid:12) − thenitbecomesaBanachspace. (cid:12)(cid:12) (cid:12)(cid:12) Definition2. Let m =0, c ,c R besuchthat2signc +signc 0. By M 1 (L )we denotethe 6 1 2∈ 1 2≥ c−1,c2 c spaceoffunctions f(x),x R+,representableintheform(8),wheresc2ep c1|s|F(s) L1(c). ∈ ∈ ItisaBanachspacewiththenorm 1 (cid:12)(cid:12)f(cid:12)(cid:12)Mc−11,c2(Lc)= 2p Zcep c1|s||sc2F(s)ds|. Inparticular,lettingc =c =0w(cid:12)(cid:12)eg(cid:12)(cid:12)etthespaceM 1(L ). Moreover,itiseasilyseentheinclusion 1 2 − c Md−1,1d2(Lc)⊆Mc−1,1c2(Lc) when2sign(d c )+sign(d c ) 0. 1 1 2 2 − − ≥ Usingthistechniqueweprovedthefollowing Theorem1[7]. Leta>0, n C, 0<Ren <1/2,g(x) M 1(L )withc= s C: 1<Res<0 . ∈ ∈ 0−,1 c { ∈ − } Then for almost all x>0 expansion (3) holds, where the inner and outer integralsare understoodin the impropersense. Theseresultswillbeappliedtosolvetheso-calledWebertypeintegralequation ¥ j (l )[Jn (xl )Yn +1(al ) Yn (xl )Jn +1(al )]dl = f(x), x>a>0 (11) Z0 − intheclassM 1(L ). ThekeyingredientwillbealsopropertiesforderivativeofBesselfunctions,namely, 0−,1 c (see[1],Vol. II) d n n x∓ x± Jn (x)= Jn 1(x), (12) (cid:20) dx (cid:21) ± ∓ Webertypeintegralequation 3 d n n x∓ x± Yn (x)= Yn 1(x), (13) (cid:20) dx (cid:21) ± ∓ andthefollowingintegral,whichisadirectconsequenceofrelation(2.13.15.4)in[2],Vol. 2,namely, ¥ 2 san +1 cos(pn ) Fn (x,s)=Z0 l −s[Jn (xl )Yn +1(al )−Yn (xl )Jn +1(al )]dl = p −x2−s+n G (s/2) G (−n −1)G (1+n −s/2) s s a2 2 sxn +sG (n +1)G ( s/2) ×2F1(cid:18)1−2, 1+n −2; 2+n ; x2(cid:19)− p−a1+n G (1+n +−s/2) s s a2 2 san +1 p s G (n +1 s/2)G (1 s/2) ×2F1(cid:18)−n −2, −2; −n ; x2(cid:19)−p −x1+n −scos(cid:16) 2 (cid:17) −G (2+n ) − s s a2 F 1+n , 1 ; 2+n ; , 1<Res<0, x>a, (14) ×2 1(cid:18) −2 −2 x2(cid:19) − where G (z) is Euler’s gamma-function and F (a,b;c;x) is Gauss’s hypergeometric function [1], Vol. 1, 2 1 havinganintegralrepresentationastheEulerintegral G (c) 1 F (a,b;c;x)= ub 1(1 u)c b 1(1 xu) adu, (15) 2 1 G (b)G (c b)Z0 − − − − − − − for instance, underconditionsRea>0, Rec>Reb>0 x [0,1). Moreover,representation(15)givesus ∈ thefollowinguniformestimatefortheGaussfunction,whichwillbeusedbelow G (c) 1 F (a,b;c;x) uReb 1(1 u)Re(c b) 1(1 xu) Readu |2 1 |≤(cid:12)(cid:12)G (b)G (c−b)(cid:12)(cid:12)Z0 − − − − − − (cid:12) (cid:12) (cid:12) (cid:12) G (c) (1 x) ReaB(Reb, Re(c b)) , (16) ≤ − − − (cid:12)G (b)G (c b)(cid:12) (cid:12) − (cid:12) whereB(a,b)isEuler’sbeta-function[1],Vol. 1.Let 1<Re(cid:12)n < 1/2.The(cid:12)n,denotingbythesameletter (cid:12) (cid:12) − − Cvariouspositiveconstants,whichcanoccur,weobtain s s a2 G (2+n ) F 1 , 1+n ; 2+n ; Cx2 Res(x2 a2)Res/2 1 , (cid:12)2 1(cid:18) −2 −2 x2(cid:19)(cid:12)≤ − − − (cid:12)G ((2(1+n ) s)/2)G ((2+s)/2)(cid:12) (cid:12) (cid:12) (cid:12) − (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12)2F1(cid:18)−n −2s, −2s; −n ; ax22(cid:19)(cid:12)(cid:12)≤Cx−2n −Res(x2−a2)n +R(cid:12)es/2(cid:12)G ( s/2G)G(−(n n)+s/2)(cid:12), (cid:12) (cid:12) (cid:12) (cid:12) − − (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) s s(cid:12) a2 (cid:12) (cid:12) F 1+n , 1 ; 2+n ; Cx2(1+n ) Res(x2 a2)Res/2 1 n (cid:12)2 1(cid:18) −2 −2 x2(cid:19)(cid:12)≤ − − − − (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) G (2+(cid:12) n ) . ×(cid:12)G ((2 s)/2)G ((2(1+n )+s)/2)(cid:12) (cid:12) − (cid:12) Theseestimatesallowtoprovetheco(cid:12)nvergenceoftheintegral(11)asa(cid:12)nimproperone.Infact,writingitas (cid:12) (cid:12) N lim j (l )[Jn (xl )Yn +1(al ) Yn (xl )Jn +1(al )]dl , N ¥ Z0 − → 4 S.Yakubovich wetakej fromthesubspaceM 1 (L ) M 1(L )withc= s C: 1<Res<0 . So, accordingto 1−/2,1 c ⊂ 0−,1 c { ∈ − } Definition2j isgivenbyintegral(8)ofsomefunctionF (s)fromtheweightedL -space. Thenchanging 1 theorderofintegrationbyFubini’stheoremforeachfixedN andusing(14),itbecomes N 1 m +i¥ Z0 j (l )[Jn (xl )Yn +1(al )−Yn (xl )Jn +1(al )]dl = 2p iZm i¥ F (s)Fn (x,s)ds − 1 m +i¥ ¥ −2p iZm i¥ F (s)ZN l −s[Jn (xl )Yn +1(al )−Yn (xl )Jn +1(al )]dl ds, x>a. − Wewillprovethat m +i¥ ¥ lim F (s) l −s[Jn (xl )Yn +1(al ) Yn (xl )Jn +1(al )]dl ds=0, x>a. (17) N ¥ Zm i¥ ZN − → − Todothis,weappealtotheasymptoticbehaviorofBesselfunctionsatinfinity[1],Vol. IItofindforfixed x>a 2 1 Jn (xl )Yn +1(al )−Yn (xl )Jn +1(al )=−pl √xa(cid:20)cos(l (x−a))+O(cid:18)l (cid:19)(cid:21), l →¥ . Substitutingthisexpressioninto(17)andintegratingbypartsintheinnerintegralwithrespecttol itgives ¥ l −s[Jn (xl )Yn +1(al ) Yn (xl )Jn +1(al )]dl =O (s +1)N−m −1 , ZN − | | (cid:0) (cid:1) and m +i¥ ¥ F (s) l −s[Jn (xl )Yn +1(al ) Yn (xl )Jn +1(al )]dl ds (cid:12)(cid:12)Zm −i¥ ZN − (cid:12)(cid:12) (cid:12) (cid:12) (cid:12) m +i¥ (cid:12) CN−m −1 F (s) (s +1)ds 0, N ¥ ≤ Zm i¥ | | | | | |→ → − underassumptionm +1>0.Henceweprovedtheequality ¥ 1 m +i¥ Z0 j (l )[Jn (xl )Yn +1(al )−Yn (xl )Jn +1(al )]dl = 2p iZm i¥ F (s)Fn (x,s)ds, (18) − wheretheintegralintheright-handsideof(18)convergesabsolutely. Indeed,from(14),(16)andStirling’s asymptoticformulaforthegamma-functionatinfinity[1],Vol. I,wehave F(x,s) Cxm Ren ep s/2 s m , x>a. − || − | |≤ | | Therefore, m +i¥ m +i¥ Zm i¥ |F (s)Fn (x,s)ds|≤Cxm −Ren Zm i¥ |F (s)|ep |s|/2|sds|=C||j ||M1−/12,1(Lc)xm −Ren . − − Moreover,ittendstozerowhenx ¥ whenm Ren <0. Let f M 1(L )withc= s→C:Res=g−>1/2 . Returningtointegralequation(11)andobserving ∈ 0−,1 c { ∈ } thatduetotheabsoluteanduniformconvergenceoftheintegral(8)anditsderivativewithrespecttox x > 0 0functions f,j arecontinuouslydifferentiableon[a,¥ )andR , respectively,weactwiththediffe≥rential + operator xn d x n onitsbothsides. Hence,employing(12),(13),weobtain dx − (cid:2) (cid:3) Webertypeintegralequation 5 ¥ d lj (l )[Yn +1(xl )Jn +1(al ) Jn +1(xl )Yn +1(al )]dl = xn x−n f(x). (19) Z0 − (cid:20) dx (cid:21) Itisallowedowingtotheuniformconvergencebyx a >aoftheintegralwithrespecttol in(19)ifwe 0 keepfunctionlj (l )inthesamespaceM 1(L ),i.e≥. 0−,1 c 1 m +i¥ lj (l )= 2p iZm i¥ Y (s)l −sds, (20) − where sY (s) L (c), c= s C: 1< m <0 . In fact, recalling the asymptotic behavior of Bessel 1 | | ∈ { ∈ − } functionsatinfinityandintegratingbyparts,wefind ¥ (cid:12)(cid:12)ZN lj (l )[Yn +1(xl )Jn +1(al )−Jn +1(xl )Yn +1(al )]dl (cid:12)(cid:12) 1 (cid:12)(cid:12) ¥ sin(l (x a)) m +i¥ Y (s)l s 1dsdl +CN m 1 m (cid:12)(cid:12)+i¥ Y (s)ds ≤ p 2√xa(cid:12)ZN − Zm i¥ −− (cid:12) − − Zm i¥ | | (cid:12) − (cid:12) − ≤O(N−m −1(cid:12)(cid:12))+p 2(a0−1a)√a0a(cid:12)(cid:12)ZN¥ cos(l (x−a))Zm(cid:12)(cid:12)m−+i¥i¥ Y (s)(s+1)l −s−2dsdl (cid:12)(cid:12) O(N m 1)+ N−m −1(cid:12)(cid:12) m +i¥ Y (s)(s +1)ds 0, N ¥ , (cid:12)(cid:12) ≤ − − p 2(a0 a)√a0aZm i¥ | | | | | |→ → − − wherethedifferentiationundertheintegralsignisallowedviatheabsoluteanduniformconvergence. Meanwhile,bythesamereasonsforx a,wehave(see(8)) ≥ (cid:20)xn ddxx−n (cid:21)f(x)=−21p iZgg+i¥i¥ (s+n )F(s)x−s−1ds, x≥a, andittendstozerowhenx ¥ viatheestimate − → d 1 xn x n f(x) Cx g 1 f , g > (cid:12)(cid:20) dx − (cid:21) (cid:12)≤ − − M0−,11(Lc) 2 (cid:12) (cid:12) (cid:12)(cid:12) (cid:12)(cid:12) aswellas (cid:12) (cid:12) (cid:12)(cid:12) (cid:12)(cid:12) (cid:12) (cid:12) (cid:12)ddx x−n f(x) (cid:12)≤Cx−g−n −1 f M0−,11(Lc)→0, x→¥ . (cid:12) (cid:2) (cid:3)(cid:12) (cid:12)(cid:12) (cid:12)(cid:12) This meansthat equations(cid:12)(11), (19)are(cid:12)equivalent. H(cid:12)e(cid:12)nc(cid:12)(cid:12)e employingTheorem1, the uniquesolution of (cid:12) (cid:12) equation(19)hastheform 1 ¥ d j (l )=−Jn2+1(al )+Yn2+1(al )Za Cn +1(l t,l a)tn +1dt(cid:2)t−n f(t)(cid:3)dt, l >0. (20) Itcan be written in a differentformwith the integrationby parts, eliminatingouterintegratedtermssince t1/2f(t)=o(1), t ¥ andtakingintoaccount(12),(13).Hence, → 1 ¥ d j (l )= Jn2+1(al )+Yn2+1(al )Za tf(t)(cid:20)t−1−n dtt1+n (cid:21)Cn +1(l t,l a)dt l ¥ = Jn2+1(al )+Yn2+1(al )Za tf(t)[Jn (l t)Yn +1(al )−Yn (l t)Jn +1(al )]dt, l >0. 6 S.Yakubovich Wesummarizeourresultsbythefollowing Theorem2. Let 1<Ren < 1/2, f M 1(L )withc= s C:Res=g >1/2 , j M 1 (L ) − − ∈ 0−,1 c { ∈ } ∈ 1−/2,1 c andlj (l ) M 1(L )with c= s C: 1<Res<Ren . Thenj is theuniquesolutionoftheWeber ∈ 0−,1 c { ∈ − } typeintegralequation(11),givenbytheformula l ¥ j (l )= Jn2+1(al )+Yn2+1(al )Za tf(t)[Jn (l t)Yn +1(al )−Yn (l t)Jn +1(al )]dt, l >0. Acknowledgments TheworkwaspartiallysupportedbyCMUP[UID/MAT/00144/2013],whichisfundedbyFCT(Portugal) with national (MEC) and European structural funds through the programs FEDER, under the partnership agreementPT2020.TheauthorthanksMarkCraddockforpointingouttheWebertypeequationforpossible applicationstohisattention. REFERENCES 1. A.Erde´lyi,W.Magnus,F.OberhettingerandF.G.Tricomi,HigherTranscendentalFunctions,Vols.IandII,McGraw-Hill,New York,LondonandToronto(1953). 2. A.P.Prudnikov,Yu.A.BrychkovandO.I.Marichev,IntegralsandSeries:Vol.1:ElementaryFunctions,GordonandBreach,New York(1986);IntegralsandSeries:Vol.2:SpecialFunctions,GordonandBreach,NewYork(1986);Vol.3:MoreSpecialFunctions, GordonandBreach,NewYork(1990). 3. E.C.Titchmarsh,Weber’sintegraltheorem,Proc.Lond.Math.Soc.,22(2),(1924),15-28. 4. E.C.Titchmarsh,AnIntroductiontotheTheoryofFourierIntegrals,ClarendonPress,Oxford(1937). 5. VuKimTuan, O.I.Marichev andS.Yakubovich, Compositionstructureofintegral transformations, J.Soviet Math., 33(1986), 166-169. 6. S. Yakubovich and Yu. Luchko, The Hypergeometric Approach to Integral Transforms and Convolutions. Mathematics and its Applications,287.KluwerAcademicPublishersGroup,Dordrecht(1994). 7. S.Yakubovich,OntheWeberintegralequationandsolutiontotheWeber-Titchmarshproblem.Arxiv.1612.05455. DEPARTMENTOFMATHEMATICS,FACULTYOFSCIENCES,UNIVERSITYOFPORTO,CAMPOALEGRESTR.,687;4169-007 PORTO,PORTUGAL E-mailaddress: [email protected]