ebook img

Real zeros of 2F1 hypergeometric polynomials PDF

0.12 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 Real zeros of 2F1 hypergeometric polynomials

Real zeros of F hypergeometric polynomials 2 1 DDominici1 DepartmentofMathematics,StateUniversityofNewYorkatNewPaltz,1HawkDr.Suite9,NewPaltz,NY12561-2443,USA SJJohnston DepartmentofMathematicalSciences,UniversityofSouthAfrica,POBox392,UNISA0003,SouthAfrica KJordaan2 3 DepartmentofMathematicsandAppliedMathematics,UniversityofPretoria,LynnwoodRoad,Pretoria,0002,SouthAfrica 1 0 2 n a J Abstract 1 Weuseamethodbasedonthedivisionalgorithmtodetermineallthevaluesoftherealparametersbandcforwhichthe 2 hypergeometricpolynomials2F1( n,b;c;z)haven real, simplezeros. Furthermore,we usethe quasi-orthogonality − ] ofJacobipolynomialstodeterminetheintervalsonthereallinewherethezerosarelocated. A C Keywords: Orthogonalpolynomials,zeros,hypergeometricpolynomials. 2000MSC:33C05,33C45,42C05. . h t a m [ 1. Introduction 1 The F hypergeometricfunctionisdefinedby(cf. [1]) v 2 1 1 7 F (a,b;c;z)=1+ ∞ (a)k(b)k zk, z <1, 7 2 1 (c) k! | | 4 Xk=1 k 1. wherea, bandcarecomplexparameters, c<N0 = 0,1,2,... and − { } 0 3 α(α+1)...(α+k 1) , k N, 1 (α)k =1 − , k =∈ 0, α,0 v:  Xi isPochhammer’ssymbol.Thisseriesconvergeswhen z <1andalsowhenz=1providedthatRe(c a b)>0and | | − − whenz = 1providedthatRe(c a b+1) > 0. Whenoneofthenumeratorparametersisequaltoanonpositive ar integer,say−a= n, n N ,thes−eries−terminatesandthefunctionisapolynomialofdegreeninz. 0 − ∈ The problem of describing the zeros of the polynomials F ( n,b;c;z) when b and c are complex arbitrary 2 1 − parameters, has not been solved. Even when b and c are both real, the only cases that have been fully analyzed imposeadditionalrestrictionsonbandc. Recentpublications(cf.[4],[6],[7],[8],[11]and[13])consideredthezero locationofspecialclassesof F ( n,b;c;z)withrestrictionsonthe parameters bandc. Resultsonthe asymptotic 2 1 − zerodistributionofcertainclassesof F ( n,b;c;z)havealsoappeared(cf. [5],[10],[14],[15]and[27]). 2 1 − Emailaddresses:[email protected](DDominici),[email protected](SJJohnston),[email protected](K Jordaan) 1ResearchbythisauthorwassupportedbyaHumboldtResearchFellowshipforExperiencedResearchersfromtheAlexandervonHumboldt Foundation. 2ResearchbythisauthorwaspartiallysupportedbytheNationalResearchFoundationundergrantnumber2054423. PreprintsubmittedtoJournalofComputationalandAppliedMathematics January22,2013 Differenttypes of F ( n,b;c;z)have well-established connectionswith classical orthogonalpolynomials, no- 2 1 − tably the Jacobi polynomialsand the Gegenbauerorultrasphericalpolynomials(cf. [1]). Forthe rangesof the pa- rameterswherethesepolynomialsareorthogonal,informationaboutthezerosof F ( n,b;c;z)followsimmediately 2 1 − fromclassicalresults(cf. [1], [28]). Theasymptoticzero distributionof F ( n,b;c;z)whenbandc dependonn 2 1 − canbededucedfromrecentresultsbyKuijlaars,Mart´ınez-Finkelshtein,Mart´ınez-Gonza´lezandOrive(cf.[20],[21], [22], [23])onthe asymptoticzerodistributionofJacobipolynomialsP(α,β)(x) whentheparameters αandβdepend n onn. Conversely,ifthedistributionofthezerosof F ( n,b;c;z)isknown,thisleadstoinformationaboutthezero 2 1 − distribution of other special functions (cf. [6]). This makes knowledge of the zero distribution of F ( n,b;c;z) 2 1 − extremelyvaluable. Theorthogonalityofthepolynomials F ( n,b;c;z)giveninthenexttheoremfollowsfromtheorthogonalityof 2 1 − theJacobipolynomials(cf. [25, p. 257-261])andcanalso be proveddirectlyusingtheRodrigues’formulaforthe polynomials F ( n,b;c;z)(cf. [1,p. 99])aswasdonein[9]and[21]. 2 1 − Theorem1(cf.[9]). Letn N ,b, c Rand c<N . Then F ( n,b;c;z)isthenthdegreeorthogonalpolynomial 0 0 2 1 ∈ ∈ − − forthen-dependentpositiveweightfunction zc 1(1 z)b c n ontheintervals − − − | − | (i) ( ,0)forc>0andb<1 n; −∞ − (ii) (0,1)forc>0andb>c+n 1; − (iii) (1, )forc+n 1<b<1 n. ∞ − − Asaconsequenceoforthogonality,weknowthatforeachn,thenzerosof F ( n,b;c;z)arereal,simpleandlie 2 1 − intheintervaloforthogonalityforthecorrespondingrangesoftheparameters(see,forexample,[12],Theorem4)as illustratedinFigure1. b G 1 b=c+n 1 − c b=1 n − G 3 G 2 Figure1:Valuesofbandcforwhich2F1( n,b;c;z)isorthogonalandhasnrealsimplezerosintheintervals(0,1),( ,0)and(1, )areindicated − −∞ ∞ byregions 1, 2and 3respectively. G G G In his classical paper (cf. [19]), Felix Klein obtained results on the precise number of zeros of F (a,b;c;z) 2 1 thatlie in eachof the intervals( ,0),(0,1)and (1, )by generalizingearlierresultsofHilbert(cf. [17]). These −∞ ∞ Hilbert-Kleinformulasarevalidforhypergeometricfunctionsandnotonlyforpolynomials. Szego¨ recapturedthese results for the special case of Jacobi polynomials P(α,β)(x), which have a representation as F ( n,b;c;z), in the n 2 1 − intervals( , 1),( 1,1)and(1, )(cf. [28],p.145,Theorem6.72). Thenumberandlocationof therealzerosof −∞ − − ∞ F ( n,b;c;z)forbandcrealcanbededucedasfollows. 2 1 − 2 Theorem2(cf. [11],Theorem3.2). Letn N,b, c Randc>0. Then, ∈ ∈ (i) Forb>c+n,allzerosof F ( n,b;c;z)arerealandlieintheinterval(0,1). 2 1 − (ii) Forc<b<c+n,c+ j 1<b<c+ j, j=1,2,...,n, F ( n,b;c;z)has jrealzerosin(0,1).Theremaining 2 1 − − (n j) zeros of F ( n,b;c;z) are all non-realif (n j) is even, while if (n j) is odd, F ( n,b;c;z) has 2 1 2 1 − − − − − (n j 1)non-realzerosandoneadditionalrealzeroin(1, ). − − ∞ (iii) For0 < b < c, allthezerosof F ( n,b;c;z)arenon-realifniseven,while ifnisodd, F ( n,b;c;z)has 2 1 2 1 − − onerealzeroin(1, )andtheother(n 1)zerosarenon-real. ∞ − (iv) For n < b < 0, j < b < j+1, j = 1,2,...,n, F ( n,b;c;z)has j realnegativezeros. The remaining 2 1 − − − − (n j) zeros of F ( n,b;c;z) are all non-realif (n j) is even, while if (n j) is odd, F ( n,b;c;z) has 2 1 2 1 − − − − − (n j 1)non-realzerosandoneadditionalrealzeroin(1, ). − − ∞ (v) Forb< n,allzerosof F ( n,b;c;z)arerealandnegative. 2 1 − − Thevaluesoftheparametersbandcforwhich F ( n,b;c;z)hasexactlynrealsimplezerosin(0,1)givenin 2 1 − Theorem2 (i)and(ii) correspondto thosein Theorem1(ii) whilethe parametervaluesinTheorem1(i) thatensure thatallthezerosof F ( n,b;c;z)arereal,simpleandnegativearethesameasthoseinTheorem2(iv)and(v).The 2 1 − valuesofbandcinTheorem1(iii)forwhichnzerosarein(1, )canalsobeobtainedfromTheorem2(iv)and(v) ∞ usingthetransformation(cf. [1,p. 79,(2.3.14)]) (c b) F ( n,b;c;z)= − n F ( n,b;1 n+b c;1 z) 2 1 2 1 − (c) − − − − n duetoPfaff. AnaturalquestiontoaskiswhethertheparameterrangesinTheorems1and2aretheonlyvaluesofb,c Rfor ∈ which F ( n,b;c;z)havenrealsimplezeros. Inthispaper,weuseamethodthatdoesnotrelyonorthogonalityto 2 1 − determinealltherealvaluesoftheparametersbandcforwhich F ( n,b;c;z)havenrealsimplezeros. Weapply 2 1 − analgorithmwhichcountsthezerosofpolynomialswithrealcoefficientsandtheirmultiplicities. Wealsodetermine theintervalswheretherealzerosarelocatedforthesevaluesofbandc. 2. Thealgorithm Recallthatgiventwopolynomials f(x)andg(x),withdeg(f) deg(g),thereexistuniquepolynomialsq(x)and ≥ r(x) suchthat f(x) = q(x)g(x)+r(x) with deg(r) < deg(g). We willdenotethe leadingcoefficientofa polynomial f(x)=a xn+a xn 1+ +a bylc(f)=a . n n 1 − 0 n − ··· Weusethefollowingalgorithm(cf. [24]). Let f(x)bearealpolynomialwithdeg(f)=n 2. Define ≥ f (x):= f(x) and f (x):= f (x) 0 1 ′ andproceedfork Nasfollows: ∈ Ifdeg(f )>0performthedivisionof f by f toobtain k k 1 k − f (x)=q (x)f (x) r (x). k 1 k 1 k k − − − Define r (x) ifr (x).0 f (x)= k k k+1 f (x) ifr (x) 0  k′ k ≡ andgeneratethesequenceofnumbersc1,c2,...where lc(f ) k+1 ifr (x).0 k ck =lc(fk 1) . 0 − ifrk(x) 0 ≡  3 When f isconstant,thealgorithmterminates. k Notethatthealgorithmmustterminate,sincethedegreesofthepolynomials f (x)decreaseoneachstep. k Thenwehavethefollowingtheoremwhichwewillapplyto F ( n,b;c;z). 2 1 − Theorem3(cf. [24],Theorem10.5.7,p.339). Let f beapolynomialofdegreenwithrealcoefficients. Then f has onlyrealzeros ifandonlyifthe abovealgorithmproducedn 1non-negativenumbersc ,...,c . Moreover, the 1 n 1 zerosof f areallrealandsimpleifandonlyifthenumbersc ,−...,c areallpositive. − 1 n 1 − 3. Mainresults We shallassumethroughoutourdiscussionthat b,c R with b,c , 0, 1,..., n+1. Theassumptiononb is ∈ − − madetoensurethat F ( n,b;c;z)isapolynomialofdegreen. 2 1 − Proposition4. Letb,c R. Then, ∈ 1. Thezerosof F ( 2,b;c;z)arerealandsimpleifandonlyifeither(seeFigure2): 2 1 − (i) c< 1andc<b<0. − (ii) 1<c<0andb>0orb<c. − (iii) c>0 andb<0orc<b. 2. Thezerosof F ( 3,b;c;z)arerealandsimpleifandonlyifeither(seeFigure3): 2 1 − (i) c< 2 and1+c<b< 1. − − (ii) 2<c< 1 and 1<b<1+c. − − − (iii) c> 1,c,0andb< 1orb>c+1. − − b b b=c+1 b=c -1 c -2 -1 c -1 Fhaigsuornely2:reVaallusiemspolfebzearnodscforwhich2F1(−2,b;c;z) Fhaigsuornely3:reVaallusiemspolfebzearnodscforwhich2F1(−3,b;c;z) Theorem5. Foranyintegern 4,thepolynomial F ( n,b;c;z)hasonlyrealandsimplezerosifandonlyif(c,b) 2 1 ≥ − belongstooneofthefourn-dependentregions ,..., definedby 1 4 R R = c+n 2<b<2 n , 1 R { − − } = c> 1, b<2 n , 2 R { − − } = c> 1, b>n 2, b>c+n 2 , 3 R { − − − } = 1<c<0, c+n 2<b<n 2 . 4 R {− − − } 4 b c= 1 − b=n 2 − 1 0 c − b=c+n 1 c=0 − b=c+n 2 − b=2 n − b=1 n − Figure4:Valuesofbandcforwhich2F1( n,b;c;z),n=4,5,... hasnrealsimplezeros − Theparametervalues(c,b) describedinTheorem5forwhich F ( n,b;c;z),n = 4,5,... 1 2 3 4 2 1 ∈ R ∪R ∪R ∪R − has n real simple zeros are illustrated by the grey and diagonally shaded regionsin Figure 4 with the grey regions indicatingthoseparametervaluesthatextendtheresultsinTheorems1and2. Next, we turn ourattention to the location of the zeros of F ( n,b;c;z)for those parametervalueslocated in 2 1 − thegreyshadedregionsinFigure4wherethepolynomialsarenolongerorthogonalandthelocationoftherealzeros cannotbeobtainedusingTheorems1and2. Forthesevaluesof bandc,thepolynomials F ( n,b;c;z)arequasi- 2 1 − orthogonalof order1 and, in some cases, order2 (cf. [3] and [2]). Theorem3 in [2] and Theorem6 in [18] yield informationonthezerolocationofquasi-orthogonalpolynomialswithnon-varyingweightfunctions.However,these resultscannotbeappliedto F ( n,b;c;z)sincetheirweightfunctiondependsonn. Weuseinformationaboutthe 2 1 − zerosofJacobipolynomialstoobtainthefollowingthreeresults. Theorem6. Letn Nandb,c R. Then, F ( n,b;c;z)hasallitszerosrealandsimpleand(n 2)ofthemliein 2 1 ∈ ∈ − − (i) (0,1)for 1<c<0andc+n 2<b<c+n 1. Oneoftheremainingzerosliesin(1, )andtheotherone − − − ∞ in( ,0). −∞ (ii) (1, )for1 n< b< 2 nandb n+1<c <b n+2. Oneoftheremainingzerosliesin( ,0)andthe ∞ − − − − −∞ otheronein(0,1). (iii) ( ,0)for 1 < c < 0and1 n < b < 2 n. Oneoftheremainingzerosliesin(1, )andtheotheronein −∞ − − − ∞ (0,1). Theorem6appliestotheparametervaluesillustratedinFigure5. Theorem7. Letn Nandb,c R. Then, F ( n,b;c;z)hasallitszerosrealandsimpleand(n 1)ofthemliein 2 1 ∈ ∈ − − (i) (0,1)for 1<c<0andb>c+n 1. Theremainingzeroisnegative. − − (ii) (1, )for1 n<b<2 nandc<b n 1. Theremainingzeroisnegative. ∞ − − − − (iii) ( ,0)for 1<c<0andb<1 n. Theremainingzeroliesin(0,1). −∞ − − TheparametervaluesdescribedinTheorem7areillustratedinFigure6. Theorem8. Letn Nandb,c R. Then, F ( n,b;c;z)hasallitszerosrealandsimpleand(n 1)ofthemliein 2 1 ∈ ∈ − − (i) (0,1)forc>0andc+n 2<b<c+n 1. Theremainingzeroisintheinterval(1, ). − − ∞ 5 b c= 1 − b=n 2 − 1 0 c − b=c+n 1 c=0 − b=c+n 2 − b=2 n − b=1 n − Figure5:ValuesofbandccorrespondingtothosedescribedinTheorem6 (ii) (1, )forb<1 nandc+n 2<b<c+n 1. Theremainingzeroliesin(0,1). ∞ − − − (iii) ( ,0)forc>0and1 n<b<2 n. Theremainingzeroliesin(1, ). −∞ − − ∞ Figure7illustratestherangeoftheparametersbandcreferredtoinTheorem8. 4. Proofs ProofofProposition4. 1. Since 2b b(b+1) F ( 2,b;c;z)=1 z+ z2, 2 1 − − c c(c+1) weseethat F ( 2,b;c;z)=0ifandonlyif 2 1 − b(c+1) √b(c+1)(b c) z= ± − . b(b+1) Hence,thezerosof F ( 2,b;c;z)arerealandsimpleifandonlyifb(c+1)(b c)>0. 2 1 − − 2. Thediscriminantof 3b 3b(b+1) b(b+1)(b+2) F ( 3,b;c;z)=1 z+ z2 z3 2 1 − − c c(c+1) − c(c+1)(c+2) b2(b+1)(b c 1)(b c)2 isgivenby∆ =108 − − − (cf. [16])andtherefore F ( 3,b;c;z)hasrealsimpleroots 3 c4(c+1)3(c+2)2 2 1 − ifandonlyif∆ >0. 3 Thefollowingtwolemmaswillbeusedintheproofofourmainresult. 6 b c= 1 − 1 0 c − b=c+n 1 c=0 − b=c+n 2 − b=2 n − b=1 n − Figure6:ValuesofbandccorrespondingtothosedescribedinTheorem7 Lemma9. Let k n 2k n b 3 k b 1+c αk,l = (cid:16) −2 (cid:17)l(cid:16) −4− − (cid:17)l(cid:16) − −2 (cid:17)l k b 2 2k b 1 n k n 1 c −2− l −4− − l − −2 − l (cid:16) (cid:17) (cid:16) (cid:17) (cid:16) (cid:17) andletthesequenceθ berecursivelydefinedby k θ =α θ , (1) k+1 k,1 k 1 − fork 2,...,n 2 ,with ∈{ − } nb (b c)(n 1) θ = , θ = − − . 1 − c 2 c(b+n 1) − Then, (b+1)(n+c) n 1 θ = α , k=1,2,... − , (2) 2k c(n+b+1) 1,k $ 2 % and nb n 2 θ2k+1 = α2,k , k=0,1,... − . (3) − c $ 2 % ProofofLemma9. Weprovetheresultbyinductiononk. Whenk =1,theright-handsideof(2)is (b+1)(n+c)(cid:16)1−2n(cid:17)(cid:16)−n+4b+1(cid:17)(cid:16)c−2b(cid:17) = (b−c)(n−1) c(n+b+1) b+1 n+b 1 n+c c(b+n 1) − 2 − 4− − 2 − (cid:16) (cid:17)(cid:16) (cid:17)(cid:16) (cid:17) whichisθ asrequired. 2 We now assume the result is true for k = t and prove the result true for k = t +1. If we let k = t+1 on the 7 b c= 1 − 1 0 c − b=c+n 1 c=0 − b=c+n 2 − b=2 n − b=1 n − Figure7:ValuesofbandccorrespondingtothosedescribedinTheorem8 right-handsideof(2),weobtain (b+1)(n+c) RHS = α c(n+b+1) 1,t+1 = (bc(+n+1)b(n++1c)) (cid:16)b1+−21n ++tt(cid:17)(cid:16)c−2nb+c++t(cid:17)t(cid:16)−n+nb4++b11++t(cid:17)t α1,t since (a)k+1 =(a+k)(a)k − 2 − 2 − 4− (1 n+2t(cid:16))(c b+(cid:17)2(cid:16)t)( n (cid:17)b(cid:16) 1+4t) (cid:17) =θ − − − − − bytheinductivehypothesis 2t( b 1+2t)( n c+2t)( n b+1+4t) − − − − − − =α θ 2t+1,1 2t =θ from(1) 2t+2 andtheresultfollowsbyinduction. Thesecondrelation(3)maybeprovedbyinductioninasimilarway. Lemma10. Letn 4. Then,forallk 2,...,n 1 , ≥ ∈{ − } (n k)(n+c k)(b+1 k)(b c+1 k) − − − − − (4) (n+b+2 2k)(n+b 2k)(n+b+1 2k)2 − − − and (n 1)(n+c 1)(b c) − − − (5) (n+b 2)(n+b 1)2 − − arepositiveifandonlyif(c,b) . 1 2 3 4 ∈R ∪R ∪R ∪R ProofofLemma10. Sincen 1>0and(n+b 1)2 >0foralln N,b R,weseethat(5)ispositiveifandonly − − ∈ ∈ 8 if (c,b) c<1 n, b>c, b<2 n = or 1 1 ∈{ − − } A ⊃R c>1 n, b<c, b<2 n = or 2 2 ∈{ − − } A ⊃R c>1 n, b>c, b>2 n = ( )or 3 3 4 ∈{ − − } A ⊃ R ∪R c<1 n, b<c, b>2 n = . ∈{ − − } ∅ (n k) Clearly − >0forallk 2,...,n 1 ,n Nandb R,b,,3 n,5 n,...,n 5,n 3. (n+b+1 2k)2 ∈{ − } ∈ ∈ − − − − − Furthermoreb>n 2ifandonlyifn+b+2 2k>0,n+b 2k>0andb+1 k>0forallk 2,...,n 1 . − − − − ∈{ − } Hence,whenb>n 2,(4)willbepositiveforallk 2,...,n 1 ifandonlyif − ∈{ − } (c,b) b>c+k 1, c+n k >0, k =2,...,n 1 = b>c+n 2, c> 1 = or 3 ∈{ − − − } { − − } R b<c+k 1, c+n k <0, k =2,...,n 1 = b<c+1, c<2 n = or ∈{ − − − } { − } ∅ b<c+k 1, c+n k <0, k =2,...,l b>c+k 1, c+n k >0, k =l+1,...,n 1 = or ∈{ − − }∩{ − − − } ∅ b>c+k 1, c+n k >0, k =2,...,l ∈{ − − }∩ b<c+k 1, c+n k<0, k=l+1,...,n 1 b>n 2 = . { − − − }∩{ − } ∅ Similarly,b<2 nifandonlyifb+1 k<0,n+b+2 2k<0andn+b 2k<0fork 2,...,n 1 .Hence, − − − − ∈{ − } whenb<2 n,(4)willbepositiveforallk 2,...,n 1 ifandonlyif − ∈{ − } (c,b) b>c+k 1, c+n k <0, k =2,...,n 1 = b>c+n 2, c<2 n = or 4 1 ∈{ − − − } { − − } A ⊃R b<c+k 1, c+n k >0, k =2,...,n 1 = b<c+1, c> 1 = or 5 2 ∈{ − − − } { − } A ⊃R b>c+k 1, c+n k <0, k =2,...,l b<c+k 1, c+n k >0, k =l+1,...,n 1 = or ∈{ − − }∩{ − − − } ∅ b<c+k 1, c+n k >0, k =2,...,l b>c+k 1, c+n k <0, k =l+1,...,n 1 = . ∈{ − − }∩{ − − − } ∅ Fortheremainingcasewhere2 n<b<n 2or,morespecifically, 2<b+n 2k<0withb k+1<0forall − − − − − k 2,...,n 1 ,theonlynon-emptypossibilityisthat(4)ispositivefork =n 1ifandonlyifc> 1,b>c+n 2 ∈{ − } − − − andn 4<b<n 2whereas(4)ispositivefork 2,3,...,n 2 ifandonlyifc> 1,b>c+n 2andb>n 3. − − ∈{ − } − − − Hence,when2 n<b<n 2,(4)ispositiveforallk 2,...,n 1 ifandonlyif(c,b) . 4 − − ∈{ − } ∈R Since = and = ,theresultfollows. 1 4 1 2 5 2 A ∩A R A ∩A R ProofofTheorem5. Weapplythealgorithmtothepolynomial f(z)= F ( n,b;c;z). 2 1 − Wehave(cf.[25],p.69,ex.1) nb f (z)= f (z)= F ( n+1,b+1;c+1;z). 1 ′ 2 1 − c − UsingRaimundasVidu˜nas’Maplepackageforcontiguousrelationsof F hypergeometricseries(cf. [29],[30]),we 2 1 obtain 1 c+n 1 (b c)(n 1) f (z)= z − f (z) − − F ( n+2,b;c+1;z). 0 n − b+n 1! 1 − c(b+n 1) 2 1 − − − Thisrelationcaneasilybeverifiedbycomparingcoefficients.Thus, (b c)(n 1) f (z)=r (z)= − − F ( n+2,b;c+1;z). 2 1 c(b+n 1) 2 1 − − Inthenextstep(k=2),weget f (z)=q (z)f (z) r (z), 1 1 2 2 − 9 with n(b+n 1)2(b+n 2) (n 2)(c+n 2) (n 1)(c+n 1) q (z)= − − z+ − − − − 1 (n 1)(c+n 1)(b c)" b+n 3 − b+n 1 # − − − − − and (b+n 1)(b 1 c)n(n 2) r (z)= − − − − F ( n+3,b 1;c+1;z). 2 c(c+n 1)(b+n 3) 2 1 − − − − Setting f (z)=θ F ( n+k,b+2 k;c+1;z), k 1,...,n 1 , k k 2 1 − − ∈{ − } weseethatingeneralweneedacontiguousrelationoftheform F ( n+k 1,b+3 k;c+1;z) 2 1 − − − θ θ =q (z) k F ( n+k,b+2 k;c+1;z) k+1 F ( n+k+1,b+1 k;c+1;z), k−1 θk 1 2 1 − − − θk 1 2 1 − − − − fork 2,...,n 2 ,with ∈{ − } nb (b c)(n 1) θ = , θ = − − . 1 − c 2 c(b+n 1) − Using Vidu˜nas’package,we obtain(1) fork = 2,3,... andfromLemma9we concludethat θ iswell defined k andnon-zeroforallk 1,...,n 1 whenc,0and(c,b) .Thus, 1 2 3 4 ∈{ − } ∈R ∪R ∪R ∪R lc(f ) c = k+1 , k N, k lc(f ) ∈ k 1 − whichimpliesthat (b c)(n 1)( n+2) (b) (c) n! (n 1)(c+n 1)(b c) c1 = c(−b+n−−1) (c−+1)n−n2−(2n−n2−)2!(−n)nn(b)n = (b−+n−2)(b−+n−1−)2 and,fork 2,...,n 1 , ∈{ − } θ ( n+k+1) (b+1 k) (n k+1)!(c+1) ck = θk+1 − (c+1) n−k−1(n k −1)!n−k−1( n+k− 1) (b+3n−k+k1) (kn−1 k)(n k+cn)−(kb−1 k−+1)−( c+1+b− k) − n−k+1 − n−k+1 = − − − − − . (n 2k+b+2)(n 2k+b)(b 2k+1+n)2 − − − FromLemma10,weknowthatc >0forallk 1,...,n 1 when(c,b) . Theresultnow k 1 2 3 4 ∈{ − } ∈R ∪R ∪R ∪R followsfromTheorem3. ProofofTheorem6. From[2],Corollary4(i),weknowthatfor 1<α<0and 1<β<0,theJacobipolynomials P(nα−1,β−1)(x)haverealsimplezerosand(n 2)ofthemareinthein−terval( 1,1).T−hesmallestzeroissmallerthan 1 andthelargestzeroislargerthan1. Equiv−alently,thesame istrueforthe−zerosofJacobipolynomialsP(α,β)(x)wh−en n 2<α< 1and 2<β< 1. − − − − (i) OneoftheconnectionsbetweenJacobipolynomialsandthepolynomials F ( n,b;c;z)isgivenby(cf. [25], 2 1 − p. 254,eq. 3) ( 1)n(1+β) x+1 P(α,β)(x)= − n F n,1+α+β+n;1+β; (6) n n! 2 1 − 2 ! whereα=b n candβ=c 1. Theconditions 2<β< 1and 2<α< 1areequivalentto 1<c<0 − − − − − − − − and c+n 2 < b < c+n 1. Furthermore, the intervals ( 1,1), (1, ) and ( , 1) are transformed to − − − ∞ −∞ − (0,1), (1, ) and ( , 1) respectively under the linear mapping x = 2z 1. Thus, when c ( 1,0) and ∞ −∞ − − ∈ − b (c+n 2,c+n 1), F ( n,b;c;z)hasn 2real,simplezerosintheinterval(0,1),onezeroin(1, ) 2 1 an∈donezer−oin( ,−0)foreach−n N. − ∞ −∞ ∈ 10

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.