On the absolute convergence of a lacunary Fourier series

On the absolute convergence of a lacunary Fourier series

JOURNAL OF MATHEMATICAL ANALYSIS On the Absolute AND 65, 391-398 (1978) APPLICATIONS Convergence of a Lacunary Fourier Series J. R. PATADIA l...

269KB Sizes 6 Downloads 168 Views

JOURNAL

OF MATHEMATICAL

ANALYSIS

On the Absolute

AND

65, 391-398 (1978)

APPLICATIONS

Convergence

of a Lacunary Fourier Series

J. R. PATADIA llefiavtment

of Mathematics, Faculty of Science, The M.S. University

of Baroda,

Bauoda-390 002, Gzljarat State, India Submitted by R. P. Boas

1 Let cos nlzx + b,, sin nkx) be the Fourier gaps

(ni,

numbers

, nk :.l),

series of a 2r-periodic where

(nk)

(k

E

N)

(1.1)

function f EL[--rr, n] with an infinity of is a strictly increasing sequence of natural

such that (ri~.+~- nk) + co

Steckin has proved the following gap condition.

as

/2 -

co.

theorem for the Fourier

(1.2) series without

any

THEOREM [I; Vol. 11,~. 1961. 1ff OYa g iven increasing sequence {nfi} of natural numbers we haze

then for the Fourier series off E L2[-rr,

n] we obtain

il (I a,, I + I bnkI) < ~0,

(1.3)

where W”‘(l/Tz, ,f)

=

sup O
[(j-" if(x + h) -f@ - h)12~$'2/ . -w

Observe that with nk = k for all k, SteCkin’s theorem reduces to the theorem of S&s2 [I; Vol. II, p. 1551.

391 0022-247X/78/0652-0391$02.00/0 Copyright 0 1978 by Academic Press, Inc. A11 rights of reproduction in any form reserved.

392

J. R. PATADIA

In this paper we propose to prove that if the sequence (Q} satisfies (1.2) and if (1.1) is the Fourier series off then (1.3) holds, that is, (1.1) converges absolutely, even when the hypothesis in the SteEkin’s theorem is satisfied only in a subinterval of [-r, z-1. We then generalize our result by considering the higher order differences in Theorem 2. Let x0 be a fixed point and 0 < 6 < n. Let I = {x: j x - s,, i < Sj be a subinterval of [-rr, rr] and put

W e prove the following theorems. THEOREM

1. 1f

% I_____ (~‘Yllnk ,fY c k6/2

< *

(0
(1.5)

k=l

and if (1.2) holds then for the Fourier series (1.1) off sL2(1) we have

where ~(~)(l/n, THEOREM

,f) is us in (1.4).

2.

Theorem 1 holds if (1.5) is replaced by the more general condition

where wj2)(l/nk ,f)

=

sup

[(J, 1 i.

(-l)“-’

(4, f (x + (2j - I) hf

dxjl;p/

.

04h41/7Ll,

THEOREM

3.

Theorem 1 holds ;f (1.5) is replaced by the condition

where EL::(f) is the best approximation given by

off(x)

in which T%,<(x) is a trigonometric polynomial

in the space L2 over the interval I

of order not higher than n,. .

CONVERGENCE

OF A FOURIER

SERIES

393

2 We need the following lemmas. Lemma 1 is a special case of a very general theorem due to Paley and Wiener [3, Theorem XLII’]. Lemma 2 is due to SteCkin [4, Lemma 21. The inequality (2.2) of Lemma 3 is a simple consequence of the more general lemma quoted by Kennedy [2, Lemma 11. LEMMA

(1.1) then

I.

f

If f E L2(J), EL2[--a, r].

where J is an interval,

and a. (nli+l - nk) --t co in

LEMMA 2. If u, >, 0 (n E N), u, S+ 0 and let the function F(u) be concave, increasing, and such that F(0) = 0. Then

LEMMA

3. Let f EL2(I);

n, = 0, n, = -nek

(k < 0);

CnO = 0, C711c =

B(arLiG- ib,> (k > 0) and C,,, = (?,-, (k < 0). If (nk+l - nl;) > %7W,

i

for all k,

I C,, I2 < 86-l 1 If(x I

dx;

(2.1)

(2.2)

C

I G, I2< C(a)(J2Ylln, ,f))”

(2.3)

C

I C,, I2< WI (d%ln, ,f)Y,

(2.4)

lnk12n,, OY more generally

Ing >n,, and

(2.5) where C(6) is some constant depending on 6. Proof of Lemma 3.

We have

(2.6)

394

J. R. PATADIA

and if we put $(r, x) = f C,,r’“k’ --p.

exp(in,x)

(0
< 1)

for all real x, then its existence is assured by (2.6) and we obviously

(2.7) get

cos nhx + bnk sin nkx) rnk. Since feL2(1) and (1.2) holds, therefore fgL2[-n, hence by a known theorem [5, p. 871 it follows that f(x)

= L2 -

lir$t 4(r, x)

n] by Lemma

1; and

(1x1 G4*

(2.9)

Now, on account of (2.1), (2.6), (2.7), and (2.9), we can apply Lemma in [2] to obtain the inequality (2.2). Now put Ax> =f(x

(2.8)

+ h) -f(x

-

1 quoted

(2.10)

4

and Czk = 2iC,, sin n,h.

(2.11)

Then 1 Czk 1 < 2 1Cnk j and hence by (2.6) we have gC;Jrf”“:
(2.12)

(O
Put g(r, x) = f Czkkne exp(in,x) -cc

(0 < Y < 1).

We then get the identity g(r, x) = +(r, x + h) - 4(~, x together with (2.9) and (2.10) we obtain g(x) = L2 -

liII;t g(r, x)

this

(2.14)

(I x I G n).

It follows from (2.1), (2.12), (2.13) (2.14), and the inequality

g I Ck I2G 8q

h), and from

(2.13)

(2.2) that

I &>I” &

(2.15)

Hence by (2.10) and (2.11) we get 4 ‘f 1 C,, I2 sin2 1nk / h < 86-l sI lf(x -m

+ h) -f(z

- h)j2 dx.

(2.16)

CONVERGENCE

OF A FOURIER

395

SERIES

Integrating both the sides of (2.16) with respect to h over (0, r/n,) (p E N), we get

If(x+h)-f(x- A)/2 d”).

(2.17)

We see that if j nk j > np then nlmp

sin2 ] nk j h dh = m

1

s0

(Ink/lnph o sin2 t dt s [ln~l/n,l7T

sin2 t dt

.i0

1 %?

[I nk (1

+

hbl

[I nk

l/%1>

. +

using 1 < [ / nk l/n,] < / nk i/n, < 1 + [I nk l/nJ, where [ ] denotes the integral part. Therefore, from (2.17), using w(2yh/n,f) < C(h) . wy1/?2, f) (A > 0, C(h) is a constant depending on A), we get 2 < 86-l % (w’~‘(+z, ,f))2 < 86-l $

C(W’~‘(l/n, ,f))2

from which (2.3) follows. Further, if we put

m = j$o(-I>“-j(;)A, + (2j - I)h)

(2.10)’

and Cck = Cn, exp(--in,&)

(exp(2in,h) - 1)’

= 2V,, exp(-&2,/z) (- 1)” exp(iZ(n,h - a/2)) sinz n,h,

(2.11)’

396

J. R. PATADIA

then / Czk 1 < 2z / Cmk ( and hence by (2.6) we have (2.12)’ Put g(r, x) = 2 Czkr’nk’ exp(in,x) --co

(2.13)’

(0 < r < 1).

We shall then get the identity g(r, x) = i

(- l)l-j

13 I#@, x + (2j - 1) h).

GO

This together with (2.9) and (2.10)’ gives g(x) = L2 -

l&tit g(r, X)

(2.14)’

(I x I < 4.

It follows from (2.1), (2.12)‘, (2.13)‘, (2.14)‘, and the inequality

(2.2) that

g I CZk I2d W’ ‘I[ Ig(4l”ax. Further,

as in the foregoing nlnp s0

proof of (2.3), analogously

(2.15)’

we get

1 Iin*, x dx sinzzI flk I h dh 3 x l +[I ‘,[, n, ,,n,l soa sin21 1

Using this along with (2.3), (2.4) is proved. Finally, let

(2.15)’ and proceeding

Tn,(x) = be any trigonometric

21-l

polynomial

Tnp(r, x) =

1

m=-nn

21-3 *21-2”‘T’T

analogously

1

as in the proof of

(2.18)

am exp(imx)

of order not higher than nz, (p E N). If we write

C a,rl”l m=-78,

exp(imx)

(0
< 1)

(2.19)

CONVERGENCE

OF A FOURIER

for all real X, then since T”, EI?[-,, T&v)

= L2 -

397

SERIES

~1 it obviously

lirmtit Tn,(r, x)

follows that (I x I ,< 4

(2.20)

Put g(x) = fb-4 -

(2.21)

T&)

and if

k + n,s and

= Cns- an,,

if

k = n,

and

kl
zzz0,

if

k fn,

and

kl >s,

= Csn,

if

k = n,

and

kI >n,.

A, = -ak,

Then, for /k j >n,, / ,4,1 =0 Hence by (2.6) we have

if k fn,

41 C 1A, IY~~I < co --3c

1k 1
and /A,

(0
(2.22)

= j Cn, I if k = n, .

(2.23)

< 1)

and if we put g(r, x) = f A,rlrcl exp(ikx) -B

(0
< 1)

(2.24)

for all real X, then its existence is assured by (2.23) and we get the identity g(r, X) = $(Y, X) - Tn,(y, x). This together with (2.9) (2.20), and (2.21) implies g(x) = L’ -

liff{t

g(r, x)

(I * I G q).

(2.25)

Now, on account of (2.1) (2.23), (2.24), and (2.25) we can apply the Lemma quoted in [2] to obtain

f I A, I2< 86-l s, I g(x)[” dx. --m Therefore

1

(2.26)

we get (2.27)

Since (2.27) holds for arbitrary trigonometric polynomial of order not higher than n, , we get the inequality (2.5). This completes the proof of Lemma 3. Proof 409/65/2-11

of Theorem

1.

Let

n, = 0, nk = -nMk

(k < 0); CnO = 0, C,* =

398

J. R. PATADIA

S(%k- ib,,) (k > 0) and C,* = Cn-, (k < 0). We assume throughout, without loss of generality, that (2.1) holds. In view of (1.2) this can be achieved, if necessary, by adding to f(x) a polynomial in exp(in,x), a process which affects neither the hypothesis nor the conclusion of the theorem. Then, putting m, = YL+, I Cnk I2 in th e inequality (2.3) of Lemma 3, we get r;a” < C(w(a)(lIn, ,f))“, where C is some constant depending on 6. Now, applying the Lemma 2 with uk = / C,& I2 (k E 2) and F(u) = uej2, we obtain using (2.29)

= g1 (rnklv’2

G c g1 ((~Yll%

,f)Y-P’“)

< 00

on account of (1.5). Therefore (1.6) holds and this completes the proof of Theorem 1. Applying the inequalities (2.4) and (2.5) instead of the inequality (2.3) in the proof of Theorem I and proceeding analogously we get Theorems 2 and 3, respectively. ACKNOWLEDGMENTS The author is thankful tion of this paper.

to Professor V. M. Shah for his help and guidance in the prepara-

REFERENCES 1. N. K. BARY, York, 1964.

“A

Treatise

on Trigonometric

Series,”

Vols.

I and II,

Pergamon,

New

2. P. B. KENNEDY, Fourier series with gaps, Quart. J. Math. Oxford Ser. 2 7 (1956), 224-230. 3. R. E. A. C. PALEY AND N. WIENER, “Fourier Transforms in the Complex Domain,” Vol. XIX, Amer. Math. Sot. Colloq. Publ., 1934. 4. S. B. STECKIN, On the absolute convergence of orthogonal series, I, Amer. Math. Sot. Trunsl. Ser. 1 3 (1962), 271-280. 5. A. ZYGMUND, “Trigonometrical Series,” Warsaw, 1934.