• Nie Znaleziono Wyników

On convergence and Riesz means of a series of Bessel functions

N/A
N/A
Protected

Academic year: 2021

Share "On convergence and Riesz means of a series of Bessel functions"

Copied!
16
0
0

Pełen tekst

(1)

RO C ZN IK I POLSK IEG O TOWARZYSTWA MATEM ATYCZNEGO Séria I: PRACE MATEMATYCZNE XXII (1980)

S. R. Agrawal and С. M. Patel (Baroda, India)

On convergence and Riesz means of a series of Bessel functions

1. Introduction. Let Zf be the class of all Lebesgue integrable functions with p-th power (1 ^ p < oo) over [ a ,b ] , and let C be the class of all functions continuous over [ a ,b ] , 0 < a < b.

Let J v(t) and Y^it) be the Bessel functions of the first and the second kind respectively, for v ^ —1/ 2, and let

Cy{&, P) = -Ma) Yv(fi) — J v(fi) Yv(a).

Denote by j x < j 2 < j$ < ... the successive positive zeros of J v(t) and by 7i < Ï2 < Уз < • •• those of cv(at,bt),0 < a < b.

Define

(1.1) <pv(0 = (in t)ll2J v (t)

<pv(0) = lim <pv(t);

r-»0 +

for t > 0,

and

(1.2) C%4t) = y / t c v(tym,bym), a ^ t < b Consider the series

(1-3) £ am(pv{xjm), 0 x ^ 1,

m= 1 and

(1.4) £ d „ C ÿ (x ), a U x H b ;

m= 1 and the typical Riesz means:

(2)

(1.5) R*n(x) = ^ (*/«), 0 < x ^ l , and

(1.6) « W = î 1 - ^ i C W , a ü x ç b ,

m= 1 \ *>„ /

with a > 0, where j n < An < j n + l and y„ < Bn < y„ + 1. If the series (1.3) and (1.4) are Fourier-Bessel series corresponding to a function / (cf. [2]), then the Riesz-means (1.5) and (1.6) will be denoted by R f,(x ,f) and Q l ( x , f ) respectively, and we shall have

(1.7) K ( x , f ) = [f{t)<Pl{t,x)dt, O ^ x ^ l ,

where

and

(1.8)

where K ( t , x )

= 1 V A n / J v + l\ J m )

Qn(x, f ) = J / ( i ) ^ ( f , x)dt, a ^ x ^ b,

I 1

7m \ 71 Jv ( a y j Cv (xym, bym) cv (tym, bym)

B; 2 {v2(aym) - v 2(bym)}

Titchmarsh [6] has studied the convergence of Fourier-Bessel series (1.4) corresponding to a function of bounded variation in the neighbourhood of a point and of class L1. Gupta and Khoti [4] have considered the Riesz summability of (1.4). The authors of the present paper also have established certain properties regarding convergence of (1.4) in [1] and have studied certain Riesz means of (1.4) in [2].

Taberski [5] studied the-Riesz-means (1.5) of the series (1.3). The purpose of this paper is to study the typical Riesz-means (1.6) of the series (1.4).

2. Fundamental lemmas. Properties of 9*n(t,x). The first three lemmas follow from Lemmas 3.1 and 3.2 in [1] (K m and K m(oi) respectively denote suitable positive constants depending upon v and v, a, throughout the paper):

Lemma 2.1. For large values of n, Bn ~ n.

Lemma 2.2. On the rectangle R whose vertices are at ± B i, Bn± B i in the w-plane, where В will be made to tend to infinity and w = u + iv,

(3)

= О

= 0 {y/bjx e ~ ^ {b~x))

wcv(xw, aw)cv(tw, bw) ( ° x)

cv(aw, bw) for t > x and n sufficiently large.

Lemma 2.3.' On the rectangle R, cv (xw, aw) cv (aw, bw) for a < $ < b, and w = u + iv.

Lemma 2.4. Let v ^ —1/2. Then (i) for a > 0,

(2.1) \9a„(t, x)| ^ n K 1 (a), t, x e [ a , b ], n ^ 1;

(ii) for 0 < a ^ 1,

(2.2) \œH{t, x)| ^ » Г , х б [ а , Ч , о 1

, ri \t —

arcd (iii) /o r a > 1,

(2.3) lflj(*,x)| ^ . . ^ 3 ^ ,2 » t , x e [ a , b ] , x , n ^ 1.

n |t — x|

P ro o f. Let

4 wa \ nwcv(xw, aw)cv(tw ,bw)

G(w) = y /x t 1— — --- :---r -t---,

\ Bn J cv (aw, bw)

where a ^ x < t ^ b.

The residue of G(w) at w = ym is given by (cf. [2], Lemma 4) xt I j _ \ y i J hay„)c,(xym,bym)c.(tym,bym)

K ) 2{Jv2( a y J - J v2(i-Tj}

Taking R as the contour of integration, we shall have

j B r, + aoi

(2.4) f i ( t , x ) = - — J G(w)dw +

2 Я i g — cei

+- 1 2я l* — ooi because on account of Lemma 2.2

(u±Bi)a

J y/Xt wa + 1- cv(xw, aw)cv(tw ,bw)

в: cv (aw, bw) dw;

|G(w)| ^ K4 1 — в:

B(t - дс)

, w = u ± B i, 0 ^ и ^ Bn,

(4)

SO that

в„ ± Bi

lim J G(w)dw = 0.

B-* oD ± B i Also, for 0 < a < 1,

(B„ + vi)a (2.5)

and for a > 1,

m < K 5(oc)-— , - o o < v < o o ,

D„

(2.6) 1- (Bn + v if

Bl

к s (a)

< b„ ’

к , v

N ^ Bn, M > B„.

6 K ’ If 0 < a < 1, by (2.4) and (2.5), we have

a oo a

|0“(t,x)| ^ 2K1K M \ - W e-'>«-x4 v + 2K b \ - - e - ' ’(' - x4 v

0 0 D n

< K 2 ( a )

na (t — x)a +1 ’ by using Lemma 2.1.

Similarly, by an interchange of t and x, i f x > £ , 0 < a < l , K 2 (a)

n“( x - t) a This proves (2.2).

Again, if a > 1, by (2.4) and (2.6), we obtain by a similar reasoning, K ( t , x ) \ < *з(«)

n (x — t)2 ’

where n is so chosen that Bn > K/(t — x), for K , a constant.

This proves (2.3).

The case, a ^ x ^ t ^ b, follows as in [2], Lemma 4.

Lemma 2.5. For v > —1/2, v Ф 0, and any a > 0, f b

‘j |0 “(t,x)M r K 8 (a), x e [ f l,b ] , n > 1.

a

P ro o f. We have,

i K b . x W t = { Т ,П+ T " + / } 10Ж X)M( « K , ( a),

a a x — l /п x + 1 /n

(5)

by using Lemma 2.4.

Lemma 2.6. Let f x{t) = tv + 112 for te\_a,b~\. Then (i) for 0 < a ^ 2 , a < x < b,

(2.7) Ш * , / , ) - / , (x)| ^ K ,(«) + » 1,

and (ii) for a > 2 , a < x < b,

(2.8) Ш * . Л ) - Л М < « , . ( - ) - > ! •

P ro o f. 1Ву (1.8), for a < x < b,

n - y / x " / f m \ y2mJ 2(aym)cv(xym,bym) е л х . л ) - I ( ‘ “ И j H a y j - J H b y j x

x J tv + l cv(tym, b y j d t - / V ( l Ут \ ^v(fly«)Cv(xyw,b y J

L 1 к ) J h a y m) - J t ( b y m) X

X (fev Jv (aym) - a v J v {bym)}

(cf. [2], Lemma 5).

Considering

H(w) = 2 1- wa \ bvcv(xw,aw) — avcv(xw,bw) wcv(aw, few)

and Я as the contour of integration, we obtain (cf. [2], Lemma 5)

j B„ + ooi . oci

Qan( x , f 1) - x v+i/2 = -1— - j H(w)dw— -—г (P) J H(w)dw,

B „ - o o i 2 n i

since, by Lemma 2.3,

Bn ± B i

J H(w)dw-> 0 , as £-> oo

± B i

Hence,

(2.9) |Π( x ,/ i) - /i( * ) l

11

n 1 - (Д .+ 1ИГ fov + l/ 2 e ~ |t)|(b —x) _|_ flv + 1/2 M ( x - a )

V * î +*>2

dv + 2K n * r’ " 1

£ £*

+ ---— J ——- (fev +1/2 e~v(b~x) + av + 1/2 e~v(x~a)} dv,

= I ! + / 2, say.

(6)

Now, if 0 < a ^ 1, by (2.5)

2 К 1г K 5 (a) °? va bv + ll2e - v{b~x) + av + ll2e~v(x~a)

(2 .1 0 ) - ^ S r dv

c К и («) j 1 , 1

(2.11)

B* [{b — x f (x — a f Similarly, if a > 1, then by (2.6)

* 1 3 ( « )

h ^ Bt

1 1

(b —x )2 + -(x — a)2 Moreover, for any a > 0

(2.12)

2К п Г(а) f bv + l12 _ av + 1/2

12 = л ~ ; ~ + -

nBt l ( b - x f ( x - a f

^ K 14(«) j 1 t 1

b: (b — x f (x — a f

The lemma now follows from (2.9) to (2.12) and Lemma 2.1, by choosing n so large that

' 1 1

B„ > max

b — x x —a J

Lemma 2.7. The following estimates are valid for v ^ —1/2, v Ф 0:

(i) I f 0 < a < i, a < x < b, n ^ i,

(2.13) U f f ( t , x H f - l | X 15( « ) - j ^ r + . 1 1 хи* n’ f x - a f

(ii) if ol = 1, a < x < b, n > 1, ь

(2.14) + +

(iii) if 1 < a < 2 , a < x < b, n > 1, ь

(2.15) \$æn( t , x ) d t - l \

, 4 . log n i 1 1

^ --- 1---- :--- — H— r - :---~ + '

nx n(x — a)2 na (x — a f ri* (b — x f ( ’ and (iv) if a > 2 , a < x < b, n > 1,

(7)

(2.16) I jæ n( t , x ) d t - l \

r, / ч . log n 1 1 1

^ *18 (a) --- 1---- ;---ttH --- — + '

nx n(x — a)2 n2(x — a)2 n2(b — x)2 P ro o f. Let a < d < x < b, and let

f 2(t) — tv + 1/2 for a ^ t < d,

= dv + 112 for d < t ^ b . Then (cf. [2], Lemma 6),

(2.17) Ш х , / 2) - / 2( х ) К S K i t . x t Iv + 1,2<fc +

a

b

+ dv+1/2 J 11 - ( t/xy +ll2\\e*n(t , x)| d t+

a

+ (d/xy +1/2 m x J J - M x ) . Let, now, 0 < a < 1; then by (2.2),

K 2( a) (2.18) j W , x ) | C ^ d t «

a n a [ X I /

Also,

(2.19) J |l - ( t / x ) v + I'2| № ((,x)|dr

x — 1 /n x x + 1/n b

= { I + S + J + I Ï

d x —1 /n x x + l / n

= ^1 + ^2 + ^з + ^4? say.

Д + 1 an* (.x — d f

v + 1 / 2

|0S (L x)| dt

By (2.2),

(2.20)

Similarly,

(2.21)

, „ x 2(«) 'V '" 1-(« A ),+1' 2 ^ *!<>(«) 'l < . . J --- :--- T T T ^ d t <

i (x-()* +1

K „ ( « )

xn

xn

(2.22)

Again, by (2.1), in a similar manner,

K 2 0 (oi)

and 7, ^ K 20 (a)

n x nx

By (2.17) to (2.22) and using Lemma 2.6, we obtain

(8)

(2.23) \Ql(x, f 2)—f2(x)\ < <f + 1/2* 2i(a) 1 1 na(x —d f xri*

1 1

H— r:--- ^r + -

n* (x — à f t f (b — x f Also, by (1.8) and definition of / 2,

ь

(2.24) \&xn( t , x ) d t - \

a

= S 0 U t,x )d t + d - ’' - l >2 [{Q’„ ( x ,f2) - f2(x) } - f t' + 4 26 i(t,x)dt].

a a

Using (2.23), (2.24) and (2.18), and taking limit as d-+a, we have for a < x < b,

№ ( t , x ) d t - 1| < K „ (x )

a

1 1 1

\

xrf + n*(x — a f + n*(b —x f J for n > 1. This proves (2.13).

If a = 1, everything in (2.18) and (2.22) goes well by substituting a = 1.

In case of (2.20) and (2.21), (2.25) / 1 ^ K 19( a ) -lo g n

and JL ^ K 19(a)log n

nx nx

Using (2.17) to (2.19), (2.22), (2.24) and (2.25), and Lemma 2.6, it follows by taking limit as d->a, that (2.14) is proved.

In case 1 < a ^ 2, by (2.3) of Lemma 2.4,

(2.26) $ m , x ) \ t v+i/2dt *22 (* Kv+ 1/2 nx Also, by a similar analysis that is used in (2.19),

(2.27) f |1 — (0c)v + 1/2l \6^(t,x)\dt 2* i 9 (a) log n + 2K20 (a) nx

The proof of (2.15) is, now, completed by (2.17), (2.24), (2.26), (2.27) and (2.7).

Similarly, for a > 2, the proof of (2.16) follows from (2.17), (2.24), (2.26), (2.27) and (2.8).

3. Classes of functions and typical means. Let for / б С [ о ,Ь ] ,

0)(ô, / ) = sup \ f ( t ) - f ( x ) \ , t, x g [a, b ] .

I t - x \ i â

(9)

We say that / е Л я [а ,Ь ], if co(S,f) = 0 (S a) as <5->0 + , 0 < а < 1. The integral modulus of continuity of f e l? will be denoted by

('OpiàJ) = sup { J \ f ( x ± h ) - f ( x ) \ pdx}llp.

O^n^o a

Theorem 3.1. Let (1.4) be the Fourier-Bessel series of f e C , f ( a) = f(b)

— 0. I f v ^ —1/2, v ^ 0, then (i) for 0 < а < 1,

(3.1) max \Q*n(x, f ) - f ( x ) \ ^ K 23 (a) co(l/nf / ) , n ^ 1;

a^x^b

(ii) for a = 1,

( l o g n

(3.2) max \Q*„(x, f ) - f (x)\ < X 24m 1 , / , n > 1;

a^x^b \ ft

and (iii) /o r а > 1,

( logn

(3.3) max \Q*n(x, f ) - f (x)\ ^ X 25(a)col--- , / , n > 1.

a^x^b \ П

P ro o f. We write

(3.4) Qan(x, f ) - f ( x ) = $ { f{ t)-f(x )} 0 * n(t,x)d t+ f(x)[$ 0 * n( t , x ) d t - \ ]

= U„(x)+V„(x), say.

For 0 < а < 1,

\ f ( t ) - f ( x ) \ ^ со (1/n*, /) { n a|t - x | + l}, so that by Lemmas 2.4 and 2.5, we obtain,

(3.5) ш ( 1 Д Л /) [ К 8(а) + К 2(а )< ' ('" + J ’ -7Т~пг +

a x + l / n 11

x + l / n

+ nl+r Kiitx) J \t — x\dt~\

x — l In

^ К 26(а)со(1/пя, / ) for all х е [ а , Ь ] . Also, by Lemma 2.7, for x e [a + 1/n, b — 1/n], (3.6) |F„(*)I

^ / 4 f 1 + ^ _ a )na , 1 + (x —а)ия 1 +(b —x)n*

^ co(l/n , / ) X 15 (a) <--- ;--- + — --- — +-

xn~ n*(x —a)* n*(b — x )*

= K 21 (a) co (1/n*,/).

For a ^ x < a + l/n, b —1/n < x < b, by Lemma 2.5,

(3.7) |F„U) ^ {1 + К 8(а)}со(1/пя, / ) { 1/п1_я + 1} = К 28(а)со(1/пя, / ) By (3.4) to (3.7), we get (3.1) proved.

(10)

If a = 1, we take,

\ f ( t ) - f ( x ) \ ^ œ / ) | ÏQg n \*~*1 + 1 j . so that, by a similar treatment as in the previous case,

(3.8) ll/.M I K „ ( o t ) Л

Similarly, by (2.14), for a+ 1/n ^ x ^ b — 1/n, and by Lemma 2.5 for a ^ x < a + l/n , b — l/n < x ^ b,

(3.9) IК M l « / ) •

Estimation (3.2) is now proved by (3.4), (3.8) and (3.9). The treatment in case a > 1, is similar.

The following corollaries follow immediately from Theorem 3.1, Lemma 2.4 and Lemma 2.7:

Corollary 3.1.1. I f / е Л а [а ,Ь ], 0 < a < 1, /(a ) = f(b) = 0, then

Ш * , / ) - / М 1 = 0 ( " - 2), « > i .

Corollary 3.1.2. Let f e C [ a ,b ] and /ef f (a) = f (b) = 0. TTien /or v ^ —1/2, v # 0 and a > 0, tlte typical Riesz-means Q„{x, f ) converges uniformly to f (x) in [a, b].

Corollary 3.1.3. Let f e l l , 1 ^ p < oo. Then for v ^ —1/2, v # 0, and a > 0 ,

\ i m ] m x , f ) - f ( x ) \ pdx = 0.

л - o o a

Theorem 3.2. I f the series (1.4) is the Fourier-Bessel series of f e L p, 1 ^ p < oo, and if v ^ - 1/2 , v ^ 0 , 0 < a < l , f/ien

(3.10) { S m x , f ) - f ( x r d x } " '

a

a K 31(a) {cop(l/n " ,/)+ W * (l/n ,f)}, n > 1, where

w ; ( & , f) = { J i/(x )|M x } 1" + « ^ f

/ м

Ь- ô

a + <5J

/ ( * ) p 1 1 /p d x f +

b - 0 +<*1 1

a + <5

b — ô (x — a f

+ { J i / M I ' d * } 1" -

p ^ i / p

dx> + <5a<( J

J 1 e + à

f i x ) ( b - x f

p ) 1/p

dx> +

b

b — S1

(11)

P roof. As in (3.4), we have

(3.11) {J I «: [J !J|/(l)-/(x)| |(>;(r,x)|dr|'V x]1"’ +

a a a

+ [ J l / ( * ) l p | J 0£(f. x) d t - l \pdx]llp = / + /', say.

a a

b x — 1 fn x x + \ /n b

(3.12) / = [ J {( S +

s

+ 1 + 1 ) I / W - / M I №(1, x 'Ad t y d x ] 1"’

a a x — 1 /n x x + 1 /n

^ /1 + /2 + / 3 + / 4 , say.

By Lemma 2.4,

K 2{ a) (3.13) /1 <

n

К 2 ( a )

n*

Jl ! лгх

J { J | / ( x + a ) - / ( x ) | pd x }1/p

l / n u + a

1 /P

du Л + 1

< K 2(«) * у €0р(ц, / ) ,а+ 1

l / n

< /C2(a)cop(l/na, / ) . fa na u + l

,a + 1 du

l / n

Similarly, (3.14)

Again, by Lemma 2.4,

= i K 31(a)cop(l/na, / ) .

/4 ^ i X 31(a)û>p( l K , / ) .

I 2 < n K ^ U { J I f ( x - u ) - f ( x ) \ d u } pd x f lp ^ \ K 3l (a)ftip(l/n“, / ) .

a 0

Similarly, 73 can be evaluated and therefore, (3-15) I 2 + U < i X 31(a)cup(l/na, / ) .

Further,

a + l / n b — l / n b b

(3.16) '' = [ ( 1 + 1 + 1 ) I / (x)lp 11 0ü(t. x)dt —l\pdx] llp

a a + l / n b — l / n a

— + f 2 + f 3 » 8аУ•

a + l / n

By Lemma 2.5,

(3.17) /', + / , « K „ (a)[{ 1 |/(* )|'< fr} , " 4 { j I f i x r d x } 1"’].

Also, by (2.13),

b b - 1 /n1

(12)

(3.18) Г2 ^ * 3 i ( « ) b - l / n a + 1/n

f i x )

( b - 1 In

+ I (a + t/n^

X

f i x ) i x - a f

p ) t/p dx> +

» ~)1/P ( b - t / n

d x \ + <^ f

J U+1/и

f i x ) p ) t/p dx >

( b - x f The theorem is now proved by (3.11) to (3.18).

Theorem 3.3. A necessary and sufficient condition for the series (1.4) to be the Fourier-Bessel series of a function f e C , vanishing at a and b, is that

(3.19) lim Q*„ix) — f {x),

GO

uniformly in [ a ,b ].

P ro o f. The necessity of condition (3.19) follows from Corollary 3.1.2.

For the sufficiency part, we have, by using (1.6) and the orthogonality of sequence {C^(x)} (cf. Titchmarsh [6], p. xiv),

(3.20) ( :1 - dm = bm 1 Ql (X) CS> (x) d x , where

, = 7Г2 Ут Jv jaVm) 2 {J?(aym) - J v2(bym)} '

In view of (3.19), by taking limit as n->g o in (3.20), we obtain,

(3.21) dm = К

b

$ f i x ) C {ZHx)dx, m — 1 , 2 , 3 , . . .

(3.21) exhibit the coefficients dm in (1.4) as the Fourier-Bessel coefficients of / , hence the theorem is proved.

Theorem 3.4. Series (1.4) is the Fourier-Bessel series of a function f e l } , or f e U, 1 < p < oo, if and only if

(3.22) lim J \Qan( x ) - fi x ) \d x = 0,

n ~*cc a

or

b

(3.23) sup J \Ql(x)\pdx < oo

n > 1 a

respectively.

P ro o f. Let us consider series (1.4) to be the Fourier-Bessel series of / e L p, 1 < p < oo. Then

(13)

(3.24) { ï m x j r d x } 1»

a

« [ J {(’ Г + ’ T + î w m e r u t . x n d t y d x y »

a a x — 1/n x + 1/n

^ / l + / 2 + ^ 3 » S a Y -

If 0 < a < 1, then as in the proof of (3.13),

/1 ^ *Щ- J 1 If ( x - u ) \ pdx}llpdu ^ K 32( a ) ||/ ||p.

fl 1 /n W u + a

If a > 1, only a has to be replaced by 1, so that for all a > 0, we have,

(3.25) « K 32( « ) ||/||„ and h < К г2(и)\\/\\р.

In a similar way,

(3.26) I 2 ^ K 33(a) II/ ||p.

The necessity of condition (3.23), now, follows from (3.24) to (3.26).

The sufficiency part may be proved as in Bary ([3], Vol. I, Theorem 3, p. 165-167).

In case / e L1, the necessity of condition (3.22) follows from Corollary 3.1.3. The sufficiency part may be proved as in the case of Lp.

4. Convergence of Fourier-Bessel series. Throughout this section we shall assume that (1.4) is the Fourier-Bessel series of the function / .

Theorem 4.1. Let / e L 2, 0 < a < 1, v ^ —1/2, v Ф 0, and let

00 J

(4.1) £ - щ - {m2( l/n * ,/) + H ? ( l/n ,/) } < 00.

n = l n

Then the Fourier-Bessel series (1.4) of f (x) converges absolutely and uniformly in [a, b~\.

P ro o f. By the theorem of best approximation and the Parseval’s relation,

00 b

(4.2) £ \dm\2 ^ J \f(x)-Q *n( x , f ) \ 2dx.

m = n - h i a

00

Set r„ = £ \dm\2. Then by (4.1), (4.2), Theorem 3.2 and Bary ([3],

m = n + 1

Vol. II, p. 157),

I Id j K 34 E

n = 1 n = 1

(4.3)

n < 0 0.

(14)

Since,

C{m M = 0 ( 1/yJ, as m->co (cf. [1], Lemma 2), the theorem follows from (4.3).

Theorem 4.2. Let f e C be a function of bounded variation over [a ,b ] such that f (a) = f (b) = 0. I f moreover,

1

(4.4) £ — {co(l/na, f ) } l/2 < со, 0 < a < 1,

n= 1 П

then the Fourier-Bessel series (1.4) of f converges absolutely and uniformly in [a, b~\.

P ro o f. By (3.1) and (4.2), for 0 < a < 1,

(4-5) r„ ^ K 23( o t) c o ( l /n \f ) \\f ( x ) - Q : ( x , f) \ \1.

Since, / is of bounded variation, it is bounded, so that (4.6) ffli ( 5 ,/) = 0(0) and W?(<5 ,/) = 0(0).

Therefore, by Theorem 3.2, (4.5) and (4.6), we have as in (4.3),

(4.7) £ I4J « K 35 £

m = 1 n= 1 П

The theorem, now, follows from (4.4) and (4.7).

Following is a theorem of Lorentz type (cf. [3], Vol. I, p. 215-216):

Theorem 4.3. Let / е Л а [ а , Ь ] , 1/p \ < a < 1, 1 ^ p ^ 2, f (a) — f (b)

= 0. Then

( 4 -8 ) { m = n + 1

Î

\ П J a * Ю -

P ro o f. By (4.2) and Corollary 3.1.1,

00

(4.9) I I4.I2 = 0 ( n - 2-).

m = n + 1

Now, by Holder’s inequality (cf. [3], Vol. I, p. 216) and (4.9), (4.10)

к = 0, 1 , 2 , 3 , . . .

(4.8), now, follows from (4.10) and

00 00 2 k + i n

1 14.1'= I I 14.1'.

m = n + l k = 0 m = 2 * n + l

By putting p = 1 and p = 2, respectively, the following theorems of Bernstein type and Zygmund type (cf. [3], Vol. II, p. 154, 161) are established:

I К

= -L 1

\p ^

К3 6 p(a + l/2) - l

П {2 -p(a+l/2)+l

Г,

(15)

Theorem 4.4. I f 1/2 < a < 1 in the hypothesis of Theorem 4.3, then (1.4) converges absolutely and uniformly in [a ,b ].

Theorem 4.5. I f 0 < a < 1 in the hypothesis of Theorem 4.3, then the Fourier-Bessel coefficients of (1.4) have the order given by

d„ = 0 (n ~ a), as n-+cc.

In particular, (1.4) converges absolutely and uniformly in [a ,b ].

5. Appendix. We consider the special case v = 1/2. By the identity (cf.

[7], § 3.4), (5-1)

we can write

Ci/2(a, P) sin (p- a ) ,

(5.2) / ( * ) ~ m — 1

z

dmCXl2)(x).

Also, if v = ± | , we may choose ([1], Lemma 1),

(5.3) Ут mn

b — a for large n.

Using (5.1) to (5.3), we obtain,

v m u x , m ux

(5.4) f ( x ) ~ X 1 am cos --- + br, sin

m = i I b - a b - a Г a ^ x ^ b, where

(5.5) an

2 (b — a)dm sin я 2т ч/Ь

mub b — a

and bm =

2 (b — a) dm cos - и2т^/Ь

mub b — a

Hence, (5.4) represents, a Fourier series corresponding to / in [ a ,b ] , which is absolutely and uniformly convergent under the conditions of Theorem 4.5.

The following theorem is, thus, easily established:

Theorem 5.1. Let f e Aa [a, b ], 0 < a < 1, / ( a ) = / ( b ) = 0, and let its Fourier series be represented by (5.4). Then its Fourier coefficients (5.5) have the order given by

In particular, the Fourier series (5.4) of f converges absolutely and uniformly in [a, b ] .

(16)

References

[1] S. R. A g ra w a l and С. M. P a te l, On convergence of Fourier-Bessel series, Bull. Acad.

Polon. Sci., Sér. Sci. Math., Astronom. et Phys. 23 (1975), p. 1255-1263.

[2] —, — On the convergence and Riesz summability o f a Fourier-Bessel series, Jnânabha, Sect. A, 5 (1975), p. 145-160.

[3] N. K. Bary, A treatise on trigonometric series, Vols. I and II, Pergamon Press, 1964.

[4] D. P. G u p ta and B. P. K h o ti, On Riesz summability of a series o f Bessel functions, Publ. Math., Debrecen 17 (1970), p. 273-282.

[5] R. T a b e r sk i, Some properties o f Fourier-Bessel series, V, Bull. Acad. Polon. Sci., Sér.

Sci. Math., Astronom. et Phys. 15 (1967), p. 253-259.

[6] E. C. T itc h m a r sh , A class o f expansions in series o f Bessel functions, Proc. London Math. Soc. (2) 22 (1924), p. xiii-xvi.

[7] G. N. W a tso n , A treatise on the theory of Bessel functions, Cambridge, University Press, 1952.

D EPARTM ENT O F A PPLIED M ATHEMATICS FACULTY O F TEC H N O LO G Y AND EN G IN E ER IN G TH E M. S. UNIVERSITY O F BARODA

BARODA, INDIA

Cytaty

Powiązane dokumenty

The partitioning scheme is the induction step which shows that partitioning of Ak−1 into Ak i 4i+j , j ∈ {0, 1, 2, 3}, makes it possible to earn the suitable quantity so that

Every weakly sequentially complete locally convex vector space, in particular every Hilbert space, satisfies condition (0) (this follows easily from the Orlicz-

The difference between the number of partitions of n into an even number of parts from the multiset S and the number of partitions into an odd number of parts from S is bounded if

[r]

Patel, Reader in Applied Mathematics, Faculty of Technology and Engineering for his encouragement and valuable suggestion for the preparation of this

Bolewicz asked (oral communication) what is the shape of the set В for a power series in a general F-space, in particular, whether В must be connected.. We exhibit

In this paper, we give an answer to this question by showing the following theorem, which implies that there is no measurable g satisfying (1.4) for the function of Example B..

Returning to the general situation we recall the following estimate, due to H¨ormander [8] and Peetre [19], on the kernel of the Riesz mean associated with a dth order