ARTICLE IN PRESS
Journal of Approximation Theory 123 (2003) 300–304 http://www.elsevier.com/locate/jat
Note
On the absolute ðC; aÞ convergence for functions of bounded variation$ Wang Kunyang and Yu Dan Department of Mathematics, Beijing Normal University, Beijing 100875, China Received 21 December 2002; accepted in revised form 7 May 2003 Communicated by Ranko Bojanic
Let f be a 2p-periodic function integrable on ½p; p: For a4 1; the Cesa`ro means of order a of the Fourier series of f are defined by Z 1 p a sn ð f ÞðxÞ ¼ f ðx þ tÞKna ðtÞ dt; p p where Kna ðtÞ ¼
n 1 X Aa1 Dk ðtÞ; a An k¼0 nk
Aan ¼
Gðn þ a þ 1Þ ; Gðn þ 1ÞGða þ 1Þ
Dk ðtÞ ¼
sinðk þ 12Þt : 2 sin 12t
Define Ran ð f ; xÞ ¼
N X
jsak ð f ÞðxÞ sak1 ð f ÞðxÞj:
k¼nþ1
The following result was presented in [1]: Theorem HB. Let xA½0; p and f a 2p-periodic function of bounded variation on ½p; p: Then, for a40 and nX2; we have n p 4a X Ran ð f ; xÞp V k ðj Þ; np k¼1 0 x $
Supported by NSFC under the Grant 10071007. Corresponding author. Current address (until June 12, 2003): School of Mathematics and Statistics, The University of Sydney, NSW 2006, Australia. E-mail address:
[email protected] (W. Kunyang).
0021-9045/03/$ - see front matter r 2003 Elsevier Science (USA). All rights reserved. doi:10.1016/S0021-9045(03)00102-3
ARTICLE IN PRESS W. Kunyang, Y. Dan / Journal of Approximation Theory 123 (2003) 300–304
301
where jx ðtÞ :¼ f ðx þ tÞ þ f ðx tÞ f ðx þ 0Þ f ðx 0Þ and Vab ð f Þ denotes the total variation of f on ½a; b: But this is incorrect when 0oao1: Our result is the following theorem. Theorem. Suppose 0oao1; xA½0; p and f is a 2p-periodic function of bounded variation on ½p; p: Then for nX2 n p 100 X Ran ð f ; xÞp 2 a ka1 V0k ðjx Þ; a n k¼1 and there exists a 2p-periodic function f of bounded variation on ½p; p and a point xA½0; p such that n X p 1 Ran ð f ; xÞ4 ka1 V0k ðjx Þ ðnX8Þ: a 20000an k¼1
We need two lemmas. Lemma 1. Let 0oao1: Define a gan ðtÞ ¼ Kna ðtÞ Kn1 ðtÞ
cos½ðn þ a2Þt p2a Aan ð2 sin 2t Þa
0otpp:
Then for pnptpp; nX2 16a : jgan ðtÞjp ðntÞaþ1
ð1Þ
Proof. It follows from formula (5.15) of [2, p. 95] that Kna ðtÞ ¼ Dan ðtÞ þ Ena ðtÞ
ð0otppÞ;
where p 1 sin½ðn þ 1þa 2 Þt 2a ; aþ1 a An ð2 sin 2t Þ ( ) N iðnþ12Þt iðnþ1Þt X e e a 1 Aa2 : Ena ðtÞ :¼ R þ n þ 1 ð2 sin t Þ2 Aan ð2 sin 2t Þ n¼nþ1 nþ1 1 eit
Dan ðtÞ :¼
2
Since Dan ðtÞ Dan1 ðtÞ ¼
pa cos½ðn þ a2Þt pa a sin½ðn þ a1 2 2 Þt 2 ; a t Aan ð2 sin 2Þ nAan ð2 sin t Þaþ1 2
ARTICLE IN PRESS W. Kunyang, Y. Dan / Journal of Approximation Theory 123 (2003) 300–304
302
we have pa a sin½ðn þ a1 a 2 Þt 2 gan ðtÞ ¼ þ Ena ðtÞ En1 ðtÞ: nAan ð2 sin 2t Þaþ1
Hence a a jgan ðtÞjp þ jEna ðtÞ En1 ðtÞj: nAan ð2 sin 2t Þaþ1
ð2Þ
A straightforward calculation gives ( ) it N a2 X Anþn Aa2 e 2 nþ1þn a a int En ðtÞ En1 ðtÞ ¼ R e 2 sin 2t ð1 eit Þ n¼1 Aan1 Aan
a 1 : nðn þ 1Þ ð2 sin t Þ2 2
Then we obtain jEna ðtÞ
a 2að1 aÞ þ 3 n sin 2t ð2 sin 2t Þ2 n2 3a p ptpp : o n 2ð2n sin t Þaþ1
a En1 ðtÞjp
1
ð3Þ
2
Substituting (3) into (2) and noticing
1 a Aan pn
(when 0oao1) we get (1).
&
Lemma 2. Let 0oao1: Define Z t a gan ðtÞ :¼ ½Kna ðyÞ Kn1 ðyÞ dy: 0
Then
8 < 54t; 0otop; jgan ðtÞjp p : 40a1 na1 ta ; otop: n
Proof. By the definition of Kna ðtÞ we easily obtain a jKna ðtÞ Kn1 ðtÞjp1 þ
This shows jgan ðtÞjp
Z
We have gan ðtÞ ¼
n1 aGðn þ 1Þ X ðn kÞGða þ kÞ p54: Gða þ n þ 1Þ k¼1 Gðk þ 1Þ
t a jKna ðyÞ Kn1 ðyÞj dyp54 t ð0otopÞ:
0
Z t
p a ½Kna ðyÞ Kn1 ðyÞ dy ¼ In ðtÞ
Z
p t
gan ðyÞ dy;
ð4Þ
ARTICLE IN PRESS W. Kunyang, Y. Dan / Journal of Approximation Theory 123 (2003) 300–304
303
where gan is defined in Lemma 1 and Z p cos½ðn þ a2Þy p2a 1 In ðtÞ ¼ a dy: An t ð2 sin y2Þa Integrating by parts we derive 1 2 jIn ðtÞjp þ a a ðn þ 2ÞAn ð2 sin 2t Þa
Z
cos y2
p t
ð2 sin y2Þ
! dy o aþ1
2 þ 1a 1 : ðn þ a2ÞAan ð2 sin 2t Þa
ð5Þ
By (4) and (5), applying Lemma 1 we get 40 jgan ðtÞjp aþ1 a an t
p when otop: n
&
Proof of the Theorem. We have N Z p N Z 1 X 1 X a a Rn ð f ; xÞ ¼ jx ðtÞ dgj ðtÞ ¼ p p j¼nþ1
p
N X
1 p
0
Z
j¼nþ1
0
0
j¼nþ1
p
jgaj ðtÞj dV0t ðjx Þ ¼:
p
gaj ðtÞ djx ðtÞ
1 a ðR þ Ran2 Þ; p n1
where Ran1 ¼
N Z X k¼nþ1
0
p k
jgak ðtÞj dV0t ðjx Þ;
Ran2 ¼
Z N X k¼nþ1
By Lemma 2 we have Z p X Z N nþ1 twð0;pkÞ ðtÞ dV0t ðjx Þp54p Ran1 p54 0
k¼nþ1
0
p nþ1
p p k
jgak ðtÞj dV0t ðjx Þ:
p
dV0t ðjx Þ ¼ 54pV0nþ1 ðjx Þ;
where wða;bÞ denotes the characteristic function of ða; bÞ: Also by Lemma 2 we have Z p Z p Z p! N N nþ1 40 X 1 40 X 1 a t dV0 ðjx Þ ¼ þ dV0t ðjx Þ Rn2 p aþ1 a aþ1 a p p p a k¼nþ1 k k t a k¼nþ1 k k t nþ1 0 1 Z p X Z p X N 40@ nþ1 1 1 ¼ dV0t ðjx Þ þ dV0t ðjx ÞA aþ1 ta aþ1 ta p a k p k 0 k4 t nþ1 k¼nþ1 ! Z p p 40 1 1 p 2 V0nþ1 ðjx Þ þ a dV0t ðjx Þ p ta a n nþ1 ! n Z pk X p 40 1 1 ¼ 2 V0nþ1 ðjx Þ þ a dV0t ðjx Þ p ta a n k¼1 kþ1 p
n p 120 X a1 k k V ðjx Þ: 0 a2 na k¼1
ARTICLE IN PRESS 304
W. Kunyang, Y. Dan / Journal of Approximation Theory 123 (2003) 300–304
Combining the estimates for Ran1 and Ran2 we obtain n p 100 X ka1 V0k ðjx Þ ðnX2Þ: Ran ð f ; xÞp 2 a a n k¼1 On the other hand, the function 8 N
when 0oxo2p; when x ¼ 0
gives the second conclusion of the Theorem at point x ¼ p2:
&
Acknowledgments The authors thank Dr. Dai Feng for the valuable discussion.
References [1] N. Humphreys, R. Bojanic, Rate of convergence for the absolutely ðC; aÞ summable Fourier series of functions of bounded variation, J. Approx. Theory 101 (1999) 212–220. [2] A. Zygmund, Trigomometric Series, Vol. I, Cambridge University Press, Cambridge, 1959.