Internat. J. Math. & Math. Sci. VOL. 16 NO. 3 (1993) 435-448 435 ON A NEW APPROACH TO CONVOLUTION CONSTRUCTIONS S.B. YAKUBOVICH Department of Mathematics & Mechanics Byelorussian State University P. O. BOX 385, Hinsk 50 Byelorussia and SHYAM L. KALLA Research Center for Applied Mathematics Engineering Faculty Zulia University partado- 10182 Maracaibo- Venezuela. (Recelved October 3, 1991 and in revlsed form Aprll 7, 1992) ABSTRACT. In this paper we establish some new approach to constructing convolution for general Mellin type transforms. This method is based on the theory of double Hellin-Barnes integrals. Some properties of convolutions and several examples are given. KEY PHRASES. AND NORDS Mellin-Barnes Convolution, Integrals G-transforms, 1991AHS SUB3ECT CLASSIFICATION CODES. 44A35, 44A15, 33C40. 1. INTRODUCTION. As is [1], from known the following operators of typ Mellin convo I ut i on x k(---) (Kf)(x) f( u) du u (1.1) x>O o define a sufficiently particular, if k(x)=e Laplace transform large -x, then class formula of integral (1.1) defines operators. the In modified (/+f) [2] (A+f)(x) e-x/Uf(u) dUu (1.2) 0 which leads to Laplace classical transform (Lf) [3] with the following convolution x (f=g)(x) f(x-t)g(t)dt (1.3) 0 and the factorization equality L (L(f*g)) (x)=(Lf) (x) (Lo) (x). (z.4) S.B. YAKUBOVICH AND S.L. KALLA 436 Recently we [4] have developed a method of generalization of Laplace convolution (1.3) on Mellin transforms (1.1). LYI:> and general Bore 5 ]. G-transforms In case this the be in with G-function effectively extended to the integral transforms kernel It can the factorization equality of type (1.4) takes the following form (1.5) (K1 (f’g)) (x)= (Kzf) (x) (K3g) (x), here (fg)(x) is general convolution which is defined below for three (K1,K2,K3). operators Zn case of classical Laplace transform it is the (G1,G2,G 3) set (L,L,L), for G-transforms it is (see below). In this paper we consider some properties of these convolutions in special spaces and their various integral representations. By this method one can obtain the known convolutions and many new examples. 2. SOHE FUNCTIONAL SPACES, G-TRANSFORH AND ITS PARTICULAR CASES In this section we will consider generalization of the Mellin type [2] convolution transform (1.1) by Hellin-Parseval equality "+ ic where * k * (s) 1 (Kf)(x) 2=i f* (s)x ’ds, (2.1) denotes the Nellin transform of functions k(x),f(x) f(x)x-ldx, f (s) (2.2) 0 (-i,/i) is some vertical contour in the complex plane s. Thus the transform (1.1) can be studied with aid of asymiptotic of function k=(s) and f=(s) on the contour o (-i, /i=), in particular, hen {s, Re(s)=7=l/2}.Our o= main aim is to consider transforms of the form (2.2) which are convenient for our further studies. Thus the behaviour of the functions k=(s) and f=(s) on the contour o can be observed from the fact that their inverse Hellin transforms -I(L) [5]. belong to the space of functions As is shown below this space is very convenient for the studies of transform (2.2). DEFINITION (0, +), Denote 1. representable functions f*(s) by E L(O)=L(o) space --1 multiplication -1 (L) (L) by i th scalar space the Helln of functions transform f(x), x integrable of on the contour o: f*(s); f(x) The (L) inverse by the is x 2i usual linear addi tion of operations vector space. If the and norm in is introduced by the formula HfI-I(L) [If*(l/2+it)ldt, (2.4) -I(L) is a Banach space. Now we consider main properties of the space then the space 1) f(x) e This -(L) property if and only if x follows from the -1 f(x fact -1 e that " ’1(L). (L). the functions f(x) and NEW APPROACH TO CONVOLUTION CONSTRUCTIONS 437 (s)*and f(x- ) are the inverse Hellin transforms of the functions f * f (l-s) respectively, But the functions (s) and f * (1-s) either both belong to L(o) or none of them belong to L(o). x f* 2) If f(x) e -l(L) then x (x) is bounded uniformly, continuous on 1/2 (0, +=) and furthermore x f(x)= 0(1) when x-> and x--> O. This property follows from the Riemann Lebesgue lemma Z-(L) then x 1/2 f(x)g(x) -(L). 3;) If f(x), g(x) 1/2 This fact follows from that x f(x)g(x) is the inverse Mellin 1 g* (s-z/l/2)dz which belongs to transform of the function f* (t) 2i J L(o) by Fubi-ni theorem. Here g(x)= I-I(L) 4) Let f(x) e -1 (s); x and x -1/2g (x) e L(B+).Then g(u) x f(G) d__u belongs to u 0 - (L) In fact by the property of the Mellin convolution this integral is inverse the f=(s)E Mellin g(s) L(o) and hence f transform of (s)g*(s) f*(s)g*(s) the. function and since belongs to space of essential bounded functions, L(o). [1] As is known by the Mellin transform of G-function is the ratio of products of gamma-functions and according to asymptotic expansion of gamma-functions this ratio has power-exponential behavior on the contour o. Therefore it is necessary to take into consideration this fact in the spaces of, -(L) type. DEFINITION 2.[5]. Let c,7 I be such that + sgn ()_>0. 2sgn (c) Deno te ,(L) by the space of 2.5 representable by the inverse Mellin transform xe E f(x), functions x (2.3), where e (o, +,), (s)lsl Note that f=(s)lsleol’’"lelL()l f and only f f=(s)lslZe=l’le L(), and in this case the integraZ in (2.3) converges if c>O, 7 or c=O, 7>0, ehich is equivalent to (2.5). The sce (L) is Banach sce ith norm - s oous eth the sce that the - ecl’sl isZf *(s) lds fM-c,7 (L) Zt x ) L sce (L). c,7 (L) n case C:O, (2.6) :0 coincides DEFINITION 3. The G-transform of function f(x), x>O, is defined as (Gf)(x)= G m’n ()p P’ql() where o:{ s e , o f(u) (x): Re(s) o 1/2 }, O<_m<_p m (s) 2i i(s) (s)x-ds, (2.7) n P q T-T (+s) T r(z-p-s) r j=n+l x>O, O_<n<_q, r(z--s) T r (e +s) T-T J" j=m+l B, (2.8) 438 S.B. YAKUBOVICH AND S.L. KALLA Here, e assume that Re (/j)+l/2>O,,j 1/2-Re((x ) >0, ,.i=1 m; 1 (2.9) Re((xl)+l/2>O,, I/2-Re(.BI)>O, (c*,*), where * =re+n- = P/q 2 Re .1 q j=m+l p; n+l DEFINITION 4. The ordwred pair C n; (2.10) Gj- is called the index of the G-transform (2.7) 1.[5]. THEOREN I"c,(L) space case this exists on the if and only if 2sgn (c+c In (c*,?*) The G-transform with the index the ) +sgn ( + maps G-transform (2.11 ) ) >0. 1 (L) into isoorphically X+X e see that the inversion of G-transform G-transform and it can be written as follows No (2.7) also is -1 ,; / () xil (s)l g(u) (x)= () (o) q, q g* (s)x-eds (2.12) o The following theorem represents our G-transform defined by contour (2.7) in the traditional real form. The proof is obvious from integral type convolution (1.1). the Hellin THEOREH 2. Let + sgn 2sgn(c :’I’:) then the G-transform (?* (2.13) 1)>0, (2.7) ith the index (c*,*) can be represented in the real form (1.1) as follows (Gf)(x)= p, q x ,J’G; o "q (s)x- 2i (j) qj ((X)p "n f(U) du (2 14) ds is Heijer’s G-function [1] <7 Here the existence of the G-transform (2.7) is guaranteed by Definition 2 of I-1 c, (L) (see (2.13) and inequality (2.5) in (2.11) ). Theorem convergence G-function. of 2 shows integral inequali ty that which is used for (2.13) provides definition of also the Heifer’s REHRRK 1. If inequality (2.13) is replaced by 4sgnCc*) 2sgn(?*) + sgn IP-ql > O, then the statement in Theorem 2 is also true (see [5]). Further,since the Pleier’s G-function is rather a general function, its particular cases + lead to a number of corresponding transforms. Here e give a table of those transforms hich are required to study 439 NEW APPROACH TO CONVOLUTION CONSTRUCTIONS the convolutions. For convenience we introduce the following notations of transforms (see k(x) [2]). k(-G-)f(u)--G- f(u) k* 2i L o x=k (x) f(u) k x (----) f (u) du u x-f(x) k(x) x o (xk (x)x-) (f) (x), f(u) k(x) 2i L xCk(x) f(u) k*(s) 2i L Now we give G-transforms (k =(s)) is a table the in defined by k* (s+) the of forms (2.16) definitions (2.15)-(2.18), important of where (2.8) such that p+q_<2. the These simple function (s) transforms are special Bodifications of known classical integral transforms and their inversions (Laplace transform, Hankel transform, Stieltes transform, Riemann-Liouville differintegral operators, Heifer transform). The more general particular cases of G-transforms, which will be introduced in section 4, can be easily obtained from (2.15)- (2.18) by the table of Hellin transforms and representations of the using kernels k(x) through Heijer’s G-function [6]. TABLE 1 SIMPLE ZNTEGRL G-TRNSFORHS AND THEIR INVERSIONS Hodified Laplace transforms G ’0 0,1 _ s-uf(u) A+f(x)= f(u) f(u) (x): 0 o A_f(x)= f(u) f(u) (x)= GZ, o e -(/’) o f(u) e- x-ef(x) x u-el( (2.1 du (2.21) 0 ,1 0 (2.22) L Ae-() (u) ] 2i ) Hodified Riemann-Liouville differintegral and their inversions o, f(u) (x): GI’I 0 (_u)_ r() 0 u r () (u)du, Re()>O f(u) orators 440 S.B. YAKUBOVICH AND S.L. KALLA x ; (x-u)/ dx" d n-1 u -f (u)du r(=*n) -n<Re()<-n+l or Re(x): -n, o I(o)#O, n:t,2,3 G1 f(u) (x): 1 J" (u-x) =r(,) d u r(+n) f(u) r () 0 u-Xf (u)du, (u)du :z =_ (x-f(x)): Re (=) >0 (2.24) -n<Re(=)<-n+t or Re(=)-- n, x Im (=)0, n:1,2,3 r () }"I f(u) ] z (x-f(x)):z+/-(x (x)) (2.25) (2.26) Hodified Hankel transforms and their invers-ons 2,0 0 2 z, -u/z f(u) (x): 3p(2/-) f(u) (z. :9) (2.o) Sin- and Cos- transforms and their inversions GO,2 ,0 f(u) (x): :n(2)/’) f(u) [’jsin(2__)f(u) du o G2, 0 f(u) (x) in(2x (2.) 441 NEW APPROACH TO CONVOLUTION CONSTRUCTIONS ’sin(2) f (u)dU (2.33) o 112 cos 1 (2) f (u)dU (2.4) --1 sin(2V--) f(u) (x): x 0 ) sin(2x f(u) (x)=-=---cos(2y/---) /2 cos(2x x ) sin(2--) x (x)= f(u) I f(u) (2.3;6) f(u) f(u) cos(2x-l/2) cos(2V-) x (2.35) f(u) in(2x -2) f(u) (x) 0 f(u) (2.37) f(u) f(u) (2.30) Hodified Stieltjes transform and its inversion s,,, ) f(u) r()(+x)- (x) r(O) 0 ,1 ,1 ; 1 P f(u) 1 2{i (x) " uPf(u) (x+u)O r(p)(l+x) du (2.3;9) u -P-t r(s)P(o-s) f* (s)x Sds (2.40) L Plodified Pleijer transforBs and their inversions 2,0 GO,2 f(u)] S.B. YAKUBOVICH AND S.L. KALLA 442 2Kt)(2/ X/U "f(u)dUu (2.41) o 0,2 G2 -/z,Z./z f(u) (x)= 0 2 2K f(u) =2K=,(2 u /x 0 2Ku(2V’- f(u) (x) /2,-/2 ,2 1 1 F(s+/2)F(s-p/2) 2i )f (u) duu (2.42) f(u) (s)x-Sds (2.43;) L Z-p/z, Z+u/z 0 2 0 f(u) (x) G 1 2/[i 2K 2 f(u) 1 f* (s)x-’ds ]" r(s+/2)r(-s-u/2) (2.44) L DEFINITION AND RAIN PROPERTIES OF CONVOLUTIONS FOR MELLIN TYPE TRRNSFORS No e consider the Laplace convolution (l.3)if f(x), g(x) I-(L). By Definition 1 substituting representations (2.3) for f(x), g(x) (1.3), changing order of integration and using the in beta-integral we can represent (1.3) in its equivalent form3;. (f=L x g)(x) (2i) where o x o ={(s t) E z (s) F(2-s-t) (t)x -tdsdt, (3.1) o o z Re(s)= Re(t):1/2} For Laplace convolution (3.1) e have the folling result. THEOREM 3. The classical Laplace convolution (f=g(x) (1.3) exists for -(L) and it sseses the fatorization prorty all f(x), g(x) (1.4). Zn Lh case x -112 L (f.g)(x) E I- (L). POF. From (3.1) L fotlows LhaL the Laplace convoZuLon exists f and only f r(z-s)r (z-t) f* (s) g* (t) r(2-s-t) E L(osx Or). (fg)(x) (3.2) Using the representation for the beta-function we obtain that r(z-s)ro-t) r (2-s-t) o(I) hen s t-->(R) (s t) e o xo t" (3 3) f*(s) g*(t)E Bt by Definition 1 it follows that L(osxot).Hnce the condition (3.2) holds valid. Further by the substitution =s+t-I/2 representation (3.1) can be written in the following form: x/Z (f=Lg)(X) re 0={: x [ e , - r(3;/2-) 1 Re()=Z/2}. d I 2i r(1/2-+t)r(l-t) f* (-t+1/2) g* (t)dt, (3;.4) 443 NEW APPROACH TO CONVOLUTION CONSTRUCTIONS (.g)(x) Hence it follows that x (L). Now by an appeal to the classical Laplace transform e-XUf(u)du (Lf)(x) 0 (3.1) it is not difficult to obtain the property and the convolution (1.4) Theorem 3 is proved. The representation (3.t) for Laplace convolution can be modified and generalized on the G-transform (2.7). DEFZNITION 5. Integral convolution (f*g)(x) for two functions f(x) and g(x) for .the set of three G-transforms (G1,G2,G3)is defined as the following integral. Hl(S+t)Hz(SH3(t) ) 1 (f*g)(x) )’2ni’2 g* (t)x--tdsdt,x>O, f * (s) (3.5) where H (s), j 1,2,3 are the corresponding kernels of G-transforms (G f)(x)which are defined by the following relations: m. ( Hj (s):r -(.J +s, "j ( )+s PJ n,r(+s)n,r(-.-) ))-s -( qj k=n r(,s) +1 (.6) r(+1 Thus by Definition 5 e see that for each convolution defined by (3.5) there corresnds the ordered set (G,G,Ga)of G-transforms and vice Hence versa. e obtain corresnding G-transforms. s G-transform fami Zy the As of is known also G-Eransform. integral froa [5], convol utons for the coBsi[o Hence e can noEe some algebraic and sCructural prorties of these convolutions (see 1o). HEOREH 4. Let f(x), g(x) H.(s) 2 3 and =1 -1 (L) and for Nraaeters of the the corresndng conditions (2.9) hold. H(s+1/2) Let further, [1 t)lexp(-c(lsl+lt I)) Istl-7]<+, Sup (s E t) [Hl(S+t)]-H(s)H3(t).Then here (s,t) 1/2 and x t (f=g)(x) By PROOF. (3 7) (S o xo (L). cond ton (s,t)f*(s)g=(t) the convolution (3.5) exists (3.7) L(osxot). and Def n ton 2 Jt Hence e rJte the ntegral fol los tha (3.5) Jn the following form (see 3.4). -1/2 (f*g)(x) 2wi --[ H1(+1/2) H_(l/2+r-t)f* (1/2+;-t) H3(t)g*(t)dt, 2i (,:3.8) o and using Definition I the proof of Theorem 4 is completed. THEOREH 5. Let the conditions of Theorem 4 hold following inequalities hold 2sgn (c+c)+sgn (+1)> 0 j=2 3 and moreover the 444 S.B. YAKUBOVICH AND S.L. KALLA i=2,3- indices of G-transforms with corresponding (3.6). Then the factorization equality of type (1.5)for the G-transforms set (G1,G,G 3) is true c where kernels (G1 (f-g)) (x): (Gz f) (x) (G3g) (x), where def (Gf)(x) 112 - _(t1z f)(x) -lZ (L), f(x) x H(s)=H(s+I/2). case in kernel x G G-transform and the has (L) and Hellin-Parseval equality 1/z (G z f) (x) (G3g) (x) is the (2.1) we notice that the function h(x)=x Further the inverse Hellin transform of the t-integral of (3.8). PROOF. By the property (3) of condi Lions G ,j=2,3. to Gl-transform NOW we apply the provide 5 Theorem of (3.5) convolution - corresponding factorization equality. THEOREH 6. Let (s,t) (X l/2f(x), X- 112 g(x) L(osxot)and -- f(x), g(x) L(l*)). E L(x G-transforms of existence obtain and -112 the ,1+) Then the following real representation of convolution (3.5) holds- (f=g)(x) S 2 S(x,y) (s,t)x dudv f(u)g(v) y (.9) MV (3. 0) dsdt. can I: obtained by Fubini theoreM. convol utions in form (3.9), where the kernels S (x, y) depend on the sum and maximum o f The proof of Theorem Now consider corresponding the [7] variables x,y. As it is known from (s, t) and (s+t)k* (s+t)sHI(S)= Hz(S)=H3(s)=(s) where h(x)=x k =(s) -1/2 k (x) t -1 (for (s,t)= maximum). transform [,E], h*(s) O<-<E<,. Ih(x)x -dx, Hi(s)= s some of [8] belongs to the space I arbitrary segment (for sum) In the 1 ,H z (s)=H3(s)=s case first = k (s) and in the second case is. Hellin k=(s+tB(s,t) = k (s) function k(x) such that of integrable functions on Further the integral o, s (3.11) 0 converges, i.e.there exists a constant C>O, for which for all , E>O and t E R E k(x)x t-11 REHARK 2. In particular, if k(x) Zdx - < C. L(O,), then h(x)=x (3.12) 1/2 k (x) !. 445 NEW APPROACH TO CONVOLUTION CONSTRUCTIONS . (L), x tk (x) Let f(x),g(x) c the following integral representations of type (3.9) for convolutions(3.5) (f*g),(x) THEOREM 7. and (f*g)max(X) Then are true l" (f’g) (x) I k u+v x UV k x max (f*g)max(x) R 1 f(u)g(v) dudv {u (3.13) UV f(u)g(v) dudv uv } ,v 2 (3.14) The proof of Theorem 7 can be obtained by Fubini theorem and using of the following identities [7] B(s, t)k*(s+t)= k (x+y)x R S+t st y Thor s-1 dxdy, y (3.15) 2 k*(s+t)= k(max{x,y}x s- y -dxdy. it i (3.16) not (G+(f*g)/)(x)= (A+f)(x)(/+g)(x), (3.17) (G,ax(f*g)(x)=(Zlx-l)(f)(x)(!_ _lx’l)(g)(x), (3. le) where G-transform G and G have the kernels as indicated above and operators in right parts of (3.17), (3.18) are defined by Table I. Now we consider some estimations of norms for convolutions (3.13), (3.14) in traditional L-spaces. THEOREM 8.Let k(x) L(O,-) and f(x),g(x) L(x -112 ;O,-).(x -l/Zg (x (0,)). Then the convolutions (f*g)/(x),(fg),ax(X) L(O,) and following estimations of norms are true - L(R L the ;R L(x (.9) The proof of Theorem 8 can obtained folling chain of inequalities and equalities. ok R < x max 2 ,v- } Ik(x)Idx min{u 2uv lk(x) Idx u- 2m&u,v} f(u)g(v), uv v}lf(u)g(v)I dudv UV f(u)g(v) dudv UV <2 uv f (u)g(v) ’k(x)ldx -1 R R 446 S.B. YAKUBOVICH AND S.L. KALLA dUdVuv lf(u)g(v)l, ,/ Uv Ik(x) ldx _< R eualities R k(x)=2Ko(2 /)be EXAMPLE 1. Let corresponding dudv, Ik(x)ldx lf(u)l u dv. R MacDonald function (3.13), (3.14) convolutions lg(v) du and their [1]. Then the factorization can I ritten in the fors UV 2 /\ ( (f’g) (f’g) +) (x) (x)-- 2 2 2xt/2K (2-) (^/f) (x) (/\/g) (x), Ko 2 g)ma UV (U) max{u x (3.20) f(u)g(v) dudv v UV (Ilx 1) (f) (x) (11 -1 4. SOME EXAMPLES OF CONVOLUTIONS AND FACTORIZATION EQUALITIES section we give a table of examples of integral convolution by this In real representation (3.9) with S(x,y) as hypergeometric function of two variables (Horn functions)[9]. Some other Horns functions can be defined by the use of G-function of two variables [4]. The sets of G-transforms in factorization equalities follow from Table 1 and table of Mellin transforms in [1], [6]. TBLE 2 e W (g)(x). 1-e’-’ x’- x v (4.2) f o o (x-/_ lx( ( f*g ) (x) :’ () ) (u)g(v)dUdv uv (x){,F, Z’--x)} (’" (g) 447 NEW APPROACH TO CONVOLUTION CONSTRUCTIONS (f*g) (x)= 2 dudv F, (,,, (x -(o+)/2. 2 K_( x ,,z }- u x (o+)/2 )(f*g)(x)= r()r( ) r()r() -(f)(x) (x x(x x{a,_1(2x112)} (f*g)(x)= o o (4.4) )(g)(x). x x x U V ) f (u)g(v)dUdv MY [( z+x)-C}(f) (x) A+ (g)(x). @2 ()’’ " (x-XA_xX)(f*g)(x)=r() x v u (z+x) x v (f’g)(x)=r()){(].+x)-/} (f)(x)A/ x f -’ (g)(x). x v 3 (2x’/z ) } x F (],’-1)/ (g) (g)(x). (4.7 ) f (u)g (v,)dudv uv (x (1-’)/2 (4.8) (x). dudv )f(u)g(v) uv (x -(A-_ Ix (f*g) (x): r() (4.6) (u)g(v)dUdv My ---)I (/;7"-x)}(f)(x) (x-A-- Ix) (f’g) (x)-r () (4.5) ) f (u)g(v)dUdv My (f)(x) (1+x) x u xl ,_ (--7’ r() r() t (x1- I o-]’xi’-.)(f=g)(x) (x-’A_x ’) uv v (x (I-7)/2 S.B. YAKUBOVICH AND S.L. KALLA 448 x x(3 _ f3_l(2Xl/Z)t X(-I)/Z)(f)(x)(x(1-" t - )/2 x (2X1/2 ) X(’’--1)/2 ) (g) (X). ’=’1 (f*g)(x)= (’="P’ ;- o o (x-il’A_x ’) (f’g) (x)=[’( r(,/2+ uv v ;s )r(*+ ) r()r() (4.o) x No, p_a 2 ( ) (f)Cx) A+Cg)(x). (4.11) _ ACKNOLEENEMENT. This work was partially funded by a grant from the CONDES of th Zulia university. REFERENCES 1.- HRRICHEV, 0.I. :" Handbook o__f Integral Transforms o_f Higher Transcendental Functions", Ellis Hor)od, Chichester, 1983. Itogi Nauk__ 2.- BRYCHKOV, Yu.A. GLAESKE, H-Yu., HRRICHEV, O.I. English Tekhniki: Ha__t. Anal., re1. 2_1, VINITI, 1983, pp. 3-41. transl, in 3.Soviet Math. 30 (1985), no. 3. Itogi Nauki i Tekhniki: Mat. 3..- DITKIN, V.A., PRUDNIKOV, A.P. Anal., VINITI, 1967, pp. 7-82.; English transl, in Progress in Hath.vol. 4, Plenum Press, New York, 1969. 4.-YAKUBOVICH, S.B.: Dokl. A_N BSSR, 34, N 7, 1990, 588-591.(Russian). 5.-VU KIM TURN, MARZCHEV, O.I.; YAKUBOVICH, S.B.: Dokl. A__N_N SSSR 28__6, 1986, 786-790; English transl, in Soviet Math. Dokl. 33 (1986). 6.- PRUDNIKOV, A.P. BRYCHKOV, Yu.A. MARICHEV, O.I. :" Integrals and Series Here Special Functions", Gordon and Breach, New York, 1989. 7.- NGUYEN THANH HRI, YAKUBOVICH, S.B.: Dokl. AN BSSR, 34, N 5, 1990, 396-398. (Russian). 8.-VU KIM TUAN: Dokl. AN SSSR, 28__6, 1986, 521-524; English transl.in Soviet Math. Dokl. 33 (1986). Higher 9.- BATEHRN, H., ERDELYI, A. Transcendental. Functions, HcGraw--H11 New York, 1953.