Internat. J. Math. & Math. Scl. Vol. 8 No. 2 (1985) 345-354 345 THE SECOND CONTINUOUS JACOBI TRANSFORM E. Y. DEEBA Department of Applied Mathematical Sciences University of Houston-Downtown Houston, Texas 77002 and E. L. KOH Department of Mathematics and Statistics University of Regina Regina, Canada $4S OA2 (Received February 14, 1984) ABSTRACT: This paper continues the work started in [I]; a second continuous Jacobi transform is defined for suitable functions f(x). Properties of the transform are studied. In particular, the first continuous Jacobi transform in [i] and the second continuous Jacobi transform are shown to be inverse to each other. an extension of The paper concludes with Campbell’s sampling theorem [2]. KEY WORDS AD PHRASES: Couinuous Jocobi traJform, inverse Jacobi trs form, discrete Jacobi trans form, sampli,g theorem. 1980 ,THEMATICS SUBJECT CLASSIFICATION CODES. I. 44A15, 33A30, 94A65. INTRODUCTION. In this paper, the second continuous Jacobi transform, ^f(a ,B), of a function f(x) is developed along similar lines of Butzer, Stens, and Wehrens [3]. Debnath [4 The results generalize the work in [3] as well as the work of on the discrete Jacobi transform. Basic properties of ^f(a,B) will be derived including an inversion formula tersely given by (^f(a,)(.))^(a,)=f(.). The results are then applied to an extended form of Campbell’s sampling theorem [2]. The paper is divided as follows. Section two includes basic notations 346 E. Y DEEBA AND E. L. KOH Section three and results obtained in [I] that will be used in the sequel. is devoted to the study of the second continuous Jacobi transform. In this section, the first and the second continuous Jacobl transforms are shown to be inverse to each other. Section four is devoted to a sampling theorem based on Jacobi transforms and an estimate of a truncation error. PRELIMINARIES. 2. In this section we recall all the necessary background material on Jacobi functions and the first continuous Jacobi transform as studied in [i]. For the sake of completeness, we repeat some of the basic notions of hypergeometric functions. For each a, b, c real numbers with c0,-i,-2 the hypergeometric function is given by F(a,b;c;z)=k=Ol (a) k(b) k k (C)kkW. The above series converges absolutely and uniformly on each compact subset of (-i,I). Also, lim F(a,b;c;z)--F(a,b;c;l)-zl r(c) r(c-a-b) F-(c-a)-F(c-b) where the ganna function is always assumed to be a well-defined function of its argument. The Jacobi function is defined by p a,) (x)-where a, 8>-1. I’(+a+l F(a+l)F(+l) F(-k, l+a+B+l c+l;) The following are some of the properties of derived in [i] and will be used in the sequel. xe (-1,1] pa,B)(x) We refer the reader to [I] for proofs. [Lemma 2.2, [1], page 148]. LEMMA 2.1. i_> a++l 2 (i) For any x(-1,1) and we have for (x) -l<_<0, (ii) g>O, (+a+l) <r-f(a+l) r(1) +M(,n,) log( _) 2 r(+a+) ) kiP a’a)Cx)[-< rc+)r(+) +M’(k,a,B)log(-) where M(,a,) and M’(,a,B) are constants depending on %, a, and We will denote the weighted w(x)--(l-x)a(l+x) by LP(-I W 1). LP(-I,I) The norm on B. (p>l) space with welgbt LP(-I W I) is given by 34? SECOND CONTINUOUS JACOBI TRANSFORM I 1 i Eli p=(2+B+I _lW(X)I f(x) IPdx) l/p It was shown in [4] that (I.emma 2.4, [i], p.150). LEMMA 2.2. and for all for all p>l such that p+l>O and -I /p< 8< I /p B and P’B)(x)ELwP(-I,I) Jacobi polynomials with %=nP, P is the set of all positive integers, satisfy the following orthogonality relation I I w(x) e(U,6) (x) e(a,6) (x) dx= m n u++l 2 -i f where I n#m 0 n n=m r(n+a+l) r(n+B+l) n!(2n+a+B+l) r(n++B+l) n However, Jacobi functions do not satisfy such an orthogonality relation. Instead, the following result holds: (;.emma 2.5, [i], p.150). LE4A 2.3. than a+B+l -v 2 and X-(u+a++l). +I #-Iw(x)P 2 a,g) r_(+a+l) r(++l) (X-) (X++B+I) Let and u both be greater For u, Be(-,-), we have (x) (-x) dx sin r(,+l)r(+a+B+l) sin. r(+l)r(v+a++l) We also recall the definition of the discrete Jacobi transform as tudled by Debnath in [4]. (x) f(x)dx and under appropriate conditions n=O (2.2) n geL2w(-1,1) Now, for any f, and for appropriate (2.3) iJ-iw(x) f(x)g(x)dx n--O 6 -l(a B) (n)g^(,B) (n). Lemma together P(na’B)(-x)=(-l)nP(n8’a) (x) 1 2 n ad+l From (2.3) and with the identity 2.3 we ob rain n (n)= r(x+a+l) r(n.+_+_l)sin% (-n) (k+n+a++l) n r (+a+B+I) %@n (2.4) F (nd<x+l) r (n+B+l) 2n+cx+B+l) n r (n d<x+B+l) =n E. Y. DEEBA AND E. L. KOH 348 pC,B) (x) The following estimates on will be used. for large For X, (Lemma 2.6, [I], page 152). LEMMA 2.4. ++I >-------, >-q and -1/2<<1/2, we have for each [a,b](-l,l), there holds for all xc[a,b] (i) IP (il) ’{) liP (x)I=O(X -) )(x)ll I (iii) for each X -*o. as (x) 2 o(x -h) xo. as ,oo), there exists a constant M>O such c,d]c_[- ++I 2 that for all X, v[c,d] I1P :’)<x)-P <’(3)<x)il v 2 <MI,-vl The continuous Jacobi transform of the first kind is defined as in [1] by Z) (c (i for every I 1 +i 2 fL2w(-1,1) with 1 --1 w(x)P X (2.5) (x) f (x)dx >-1/2 and -1/2<8<1/2. We recall that if =8-0, we X-nEP, 3] and if obtain the results of Butzer, Stens and Wehrens we obtain the results of Debnath [4]. Again, it was shown in [i] that for any If (c’ ) (x)l--o(x-) as fEL.2.(-I,I) we have (2.6) X -,’. and (’)(’- a+l+l +) nLP(IR+) 2 )Co(IR (2.7) p>2. The following Lemma will be essential for our work. LEMMA 2.5. IR+=[0, o) (Lemma 3.2, such that X G(x)-0 belongs to ++ F(X)LI(IR+). Let F(x) be defined on Then the function F(X)P (’) .(-x) H() XsinnXdX X +B+i 2 C(-I,I)nL__(-I,I) H(X) 1], page 155). where F2tX" + +B+I. 2 r(x+ a-B+l B-+I. From now on we will assume throughout the paper that and satfsfy SECOND CONTINUOUS JACOBI TRANSFORM 349 (2.8) If 3 and 8 satisfy (2.8), then it was shown in [I] that _(a, B) (x)=4 ek 0 P a,) (’-) (,c’O (-X)Ho() sSnn)’d (2.9) PX- where r2(+) H0(R) =r(+a+)r(++)’ x(-,], kP. The following is also true. LEMA 2.6. (Theorem 3.1, [I], page 158). that X(’B)(-)LI(IR+). f(x) - (a,) Let fL2(-l,l) be such Then for almost every xE(-1,1) (2.1:0) (.j)F_ (_X)Ho()slnwd. Moreover, if f(x) is continuous then (2.D)holds everywhere on (-I,I). SECOND CONTINUOUS JACOBI TRANSFORM 3. In this section we define the second continuous Jacobi transform ^f(a,) of a function f with a and 8 satisfying ^fCa, S) and f^(a,) are inverse of each other. For each f defined on IR (x) + (2.8). We also show that we associate the integral ffi4I(x)v,a)(_x) r(),+,). 0 proosltlon 3.1. e ve + For each f defied on R Ith f and (x)C(-,)nL2(-,). thgh the proo of Proposition . alg the sa lines as Lena 2.5, we preset the proof with the proper modificati for the sake of selfontaient. PROOF of PrOposition 3.1. We first show that l(x) is well defined. rcx+) that, for large l, r(l+8+) behaves llke (see [i]). T>O, we have by Lemma 2.4 (il) and the hypothesis that 1-8 l(x)l-<l 4 ITof(X)P_Ba) (-x) r(x+) r(x++) Xsinxdxl (,) (-x) r(+) +l 4, f(),)’X_.t r(/) slnXdl -<Cl+C2 T[-+lfcx)ldX< Observe Thus for any fixed 350 E. Y. DEEBA AND E. L. KOH where C 1 and C 2 are some positive constants. We show that l(x)C(-l,l). x+/-61c(-l,l). Hence I is well-defined. For any xc(-l,l), there exists a By lemma 2.4 (i) and for all lyl<l, such that we have (-x) (-x-y) f 61>0 O f(l+8) 0 by hypothesis and C is some positive constant. Thus for 4 0 sufficiently large I r(+) (,a)(_x_y)_pa), c be made sufficiently all. Thus for >0 given, there Ists a 0 sufficiently large such that the above difference is less than /2. Fix By t. oi.ity o P)-_ (-). ., h.e o >0 tt t.,. i,t,. 0" 2>0 such tt -(,) (-x-y)p) (-x)[< whenever lyl<6 2. [X c=2 0&%[f()ldX and 6=min(gl,2). With this choice. oose 0 [I(x+y)-I(x)[ whenever [y[<. We finally show that Therefore I(x)C(-1,1). I(x)L2w(-1,1). and by Bardy-Littleed inequality (see [1], page 148) d La 2.4 ue have llx(x)2a x-+lf(X)ld( -1 (X-x) by hypothesis. Bence Z+x)IP_ I(x)cC(-1,1)L2w(-I,i). (-x)12dx) This completes the proof of Proposition 3.1. Ne sll call I(x) the second continuous Jacobi transfo and ee will denote this by f(a,). f(a’) (x)=4 Thus, ee define x) (-x) r(+) SECOND CONTINUOUS JACOBI TRANSFORM ^f(’8)eC(-l,l)OL2w(-l,l) with -<u, 8< From Proposition 3.1 we deduce that and 351 In the next theorem, we will prove an inversion formula for u+=O. ^f(u,8) the transform The inversion formula under appropriate choice of f. proof is analogous to the proof given in [3]. We will employ the Fourier cosine transform of a function f which is given by If()costd.. Fc(f)(t)= feLl(IR +) Assume that THEOREM 3.1. is such that A-8+1/2f(A)eLl(IR +) and that Fc(f)(t)--O for all -<t< Then we have for almost all heiR FCI+1/2) 1/2 (++1/2) Moreover, if fC(IR+), I 1 . + w(x)^f (,8) (x) -1 P ) (x)dx=f (). then (3.1) holds for all AelR (3.} +. By employing Fublni’s theorem together with Lemma 2.3 PROOF. e+=O), we obtain (with 1/2 r (++1/2) 2r(+) r (++:) F(+1/2) =4 F (%4<x+1/2) w(x){4 f(t)P o -1 tf(t)sinwt 0 I (8 t- )(-x) r r(t+1/2) (t+S+%) r (t+1/2) r(t+8+1/2) {I r(t+%) r(t+s+1/2) t f (t)slnt I w(x) _I P_ t sintdt Pa_) (x) dx (x) -(8,e) vt_1/2 (-x)dx}dt r(,+a+%) r(t+) w(12-t2) sin(t-1/2) -sin (l-l}) }dt tf(t)sint{ cst-csWl} dtQ(l). =4 w(12-t2) 0 Set qt cost-cos 2 -t 2 ( We claim that Q()gLI(IR+). Indeed, an application of Fubini’s theorem together with the assumption that fgLI(IR +) implies that 0 Now, by another application of Fublni’s theorem, we obtain 352 E. Y. DEEBA AND E. L. KOH FC[Q] (s)-- (k)coslsd 0 (t) slnt( =4 =4 tf(t)sintqt(%)dt)cos%sd% f (t) sint (k) cos%sdk)dt Fc(qt) (s)dt. - We employ the result of [3] and proceed analogously and determine that Fc(qt)(s)= Thus FC( Q)(s)=2 sin([-s)t, O-<s<1, t>O. (t)sint sin(-s)tdt FC (Q) (s)=Fc(f) (s)-Fc(f) (2-s)= Fc(f) (s) Moreover, Fc(Q)(s)=0 for -<s< . The assumption O-<s-<. Fc(f)(s)--O for w<s< together with the uniqueness of the Fourler-cosine transform implies that Q(%)=f(%) for almost every -F(+) % IR +" Thus _lW(X) ^f (’8) (x)P ) (x) dx. The continuity of f will imply that (3.1) holds for all %IR + From Theorem 3.1 and Lemma 2.6 we deduce that (f(a,S) (.))^(a,8) <z)=2 r(%qx+) f(%). r(z+) (3 2) and r(%+) (r(++5) f^(,8)(.))(a,8)(x)__f(x). Equations (3.2) and (3.3) reduce to the formulas obtained in 4. (3 3) [3] whenever e=8=O. A SAMPLING THEOREM In this section we give a proof of a sampling Theorem of Campbell [2] by employing results on the continuous Jacobi transform of the first and second kinds. Moreover, we will obtain an error estimate for a function f that is band-llmlted in the sense of the Fourler-coslne transform. SECOND CONTINUOUS JACOBI TRANSFORM If FC(IR THEOREM 4.1. i F(>,) t +) 353 is given by IB) w(x) f(x)p (c’ (x) dx I),-% (4.i) -1 p for some >0 fCL(-l,l) and then one obtains for all heir (2n+l) f(n+l) f(A+a+%) F()= n=O -(n+1/2) ) F(n,+P n (2n+l) r(n+l) n=O r(n+c+l) r(n+lB+l) o nn r(v+a+) r(n++l)sin(A-) ( 212- (n+) 2)n (X22-(n+)2) r(n+a+l) n=O <2n+l) F(n+l) r(,V+a+)s.in(Xp-(n+)) 2) 7 (n++J) r (1+) ( . 212-(n+) for s>q for some n, F(n.+) v >0. Let F satisfy hypotheses of Theorem 4.1 for some fixed F(%)-(Sn,F) () <C(%+I ;=En(f) where En (f)=InfP EP f-Pn2 is the set of all algebraic polynomials of degree n. Denote by S f the n-th partial sum of f(x)~ We give In particular, we show that Then there exists a constant C>O such that PROOF. and F denote the n-th partial sum of the series in (4.2). THEOREM 4.2. Pn F(-) FLI(IR+), %-B+FLI(IR +) below an error estimate for approximating F. >0. (a,) (n) Theorem 3.1 when applied to the above situation yields the series representation (4.2) for Let S sinr(),l-t) n=O REMARK 4.1. Fc(F)(s)=0 (4.2) (a,) (n)(a,) (n) [ (-l)n(2n+l) r(n+l) r<x++s) F(),)= sin(-(n+)) F(n4+l) F(%W+) (-i) (2n+l) r(n+l) L n=O r(nd<x+l) r(n+3+l) or that From (4.1), (2.3) and (2.4), we obtain (with +--0) PROOF. F(,)-- 22 ,n 2 + I -l(x,)(n)p(Ct,)(x) n n n= 0 354 E. Y. DEEBA AND E. L. KOH 2 where 6 !) (2n.+l -1= (n r (n+a+l) F (n++l) n Thus (S n (2k+l r (k+l) r(x++) F) (X) k=O ( 22 -(k+) 2 F(k++l) F(+) - By Cauchy-Schwartz inequality and Lemma 2.4, we have F(1)-(S n, <_%[i F)(I) w(x) -i < f-S n ii 12=1% -iw (x)(f(x) (S n f) (x))P If(x)-(Sn f)(x)l 2 dx’ 2< 2.f)lp fI122 p(a,B) 2-En llJ-1/2 LI 1 a,) (x)dxl 2 w(x) P (a’s) (x) 2 dx ’v-’ -1 2" Thus F(1)-(S F)(1) I-<CEn(f)(I+])-1/2 for some constant C. The estimate obtained in Theorem 4.2 shows that the error becomes smaller for large ACKNOWLEDGEMENT. The work of the first author was partially supported by a grant from the University of Houston-Downtown while the work of the second author was partially supported by NSERC of Canada under Grant A-7184. References I. 2. 3. 4. 5. DEEBA, E.Y. and KOH, E.L. The Continuous Jacobi Transform, Internat. J. Math. & Math. Sci., Vol. 6, No. 1 (1983), 145-160. CAMPBELL, L.L. A Comparison of the Sampling Theorems of Kramer and Whittaker, J. Soc. Induct. Appl. Math. 12 (1964), 117-130. BUTZER, P.L., STENS, R.L. and WEHRENS, M. Continuous Legendre Transform, Internat. J. Math. & Math. Sci., Vol. 3, No. I (1980), 47-67. DEBNATH, L. On Jacobi Transform, Bull. Cal. Math. Soc., Vol. 55 (1963) 113-120. ERDELYI, A. MAGNUS, W., OBERHETTINGER F. and TRICOMI, F.G. Higher Transcendental Functions, Vol. I, McGraw Hill, New York 1953.