ANNALES SOCIETATIS MATHEMATICAE POLONAE Series I: COMMENTATIONES MATHEMATICAE X I (1967)
P. P ych (Poznań)
Approximation of functions in L- and O-metrics
1. Preliminaries. Let L[G] be the class of all 27r-periodic functions / Lebesgue-integrable (continuous) in every finite interval. Write
= sup max \f{x+t)—f{x)\,
(o2(6jf) = sup max \f{x+t)—2 f(x )+ f(x —t)\
О — rc<a;<Tt
for any feC, and
TT
= sup f \f( x+ t) —f{x)\dx, T U
<A l)(S,f) = sup f \f ( x+ t) —2f{x )+f (x —t)\dx 0 <(<<5
for any f e L . These quantities are called the moduU of continuity and the moduli of smoothness of / in G and L, respectively.
A 1, Z 1, Z 2 will denote the classes of all functions feG with moduli satisfying the inequalities
» , № / ) < « , юг(<5,/) < 2 й , ( « , / ) < <5*, respectively. The classes of functions /«X for which
«А'Ча.Л <-«, f> < 23,.
will be denoted by /1^, Z['K Z^K It is easy to see that Л 2 >= Z ,, Л*)* cz r/J'' and that these inclusions cannot be reversed.
By Wp and W$\ with an integer p > 1 , we denote the classes of 2 Tr-periodic functions f having the derivative f (p~l) absolutely continuous in any finite interval, such that
П
ess sup |/(р)(ж)| < 1 and J | f p){%) \dx < 1 ,
— л:<ж<гс _ —
respectively.
It is well known ([5], p. 216) that if f e Wp or f e W$, then vo cos{1c(t— х) + ^р-к} oo
/ W = T + T ... у
— tu k = 1
where a0 is the initial Fourier coefficient of / , xe( — oo, oo). Also, the function / conjugate to a function / of class Wv or (p ^ 2 ), defined by the formula
x 71
/ 0 ») = — ■ — lim f { f ( x + t ) - f ( x —t)}coątdt, A7Z e ->0 j
can be represented in the form 1
71
* t u u
?о») = - -X f f (p)(t) ^
—n fe= 1
sin{ft(/—ж )+ |р 7 т:}
for xe( — oo, oo). Moreover, J / (p)(/)d/ = 0.
In this paper we consider the linear operators
TC
f{x-, i) = — f f ( t ) 0 ( t —x, £)dt
TC
J
— TC
and
Я
f i x ; £) = —-X- j f ( t ) 0 ( t - x , S)dt,
— 7Z
with the kernels
O O O O
( 1 ) 0{t, £) = ł + ^ Qk(£)cosM, 0(t, i) = £ e&(l)sinfc«;
fc=i fc=i
we assume that the factors g*(£) are defined in a set E of real numbers having an accumulation point £0, and that series ( 1 ) are uniformly con
vergent in t for any fixed £.
In Section 2 we find the connections between the quantities
TV
sup /I/O»; £)—f{oo)\dx, sup max \f(x-, £)—/(a?)|
f _n 1 — 7и<Ж <ТГ and
sup £)—/ 0 »)|<й», sup max I/O»; £)—/ 0 »)l,
/ „тх f --n^X^TC
where the supremum is taken over all functions of the corresponding
classes mentioned above. In Section 3 the exact and asymptotic formulae
for some concrete operators are given.
The following remarks will be needed below.
1° If / is of class L, geE, then T C
( 2 ) f ( x ; |) - /( ж ) = — f{f (x -\ -t )- 2 f { x )+ f{ x -t )} 0 ( t, £)dt n J 0
for every x , and
7C
(3) /(ж ; 1 )-/(ж ) = — f { f { x + t ) - f ( x ~ t ) } { l c o t ^ t - 0 ( t , £)}dt 7C
J0
for almost every x.
2° Under the restriction / e Wp or /e Wp\ £еЯ, we have
(4) /(ж )-/(ж ; I) and
(5) /(»; Ś)-f(oc)
T / / (P) (*) ^ 1 jp * ^ c°s (< -a?) ^
— n fc= 1
1
7ТГ
T t 0 0
J / (p) (0 ^ 1 ■ J p -— sin {A: (< -ж ) + |ртс}dt for an arbitrary x.
3° Let Tp be the class of all trigonometric polynomials g snch that j \g^(x)\dx < 1. Obviously T x с Л^, T 2 a Z (2 \ Tv c= Wp* (see [11],
— Tt
p. 180, ex. 9). Applying Theorem 3 given in [ 6 ], we easily obtain the fun
damental inequality
(
6
)TT T t
sup ГI f g(p)(t)K (t—x)dt\ dx o^Tp
_ n _ n^ j \K(x)—K ( x Jr ii)\dx о
for any 27r-periodic function K(t) Lebesgue-integrable over < —тс, 7 i>.
2. General theorems. Let us start with L em m a . Given an integer p ^ 1, we have
(i! sup
feTp J f I№ ; — 7Г
oo fc =0
1 ) 1 £ 2 fc+l(£) (2&+l)p+1 ’
(ii)
T T
|/(ж; £)-/(ж )| 4
^ж > -—
тс
00
2 V 1)
fc = 0
kp 1 @ 2 fc+l(£)
( 2 & + l f +1
for every fixed £eE.
Proof. Denote the left-hand side of (i) and (ii) by Sp(ś) and Sp{łj), respectively. Representation (4) and inequality ( 6 ) imply
7C TC
&p(|) = sup-*- J| J сов {к (t—x)+$piz}dt
— TT — TC k = 1
utv 2
dx
— TT — T t
T t o o
I 1 @ 2 fc+l(£)
tvj
0 fc =0
( 2 Л + 1 Гv '
cos{(2ft+l)a? + |p 7 i} dx.
Therefore, in the case of p oddJ
TC o o
0 k = 0 v '
1
k = 0
1 Q 2 k + l ( £ )
(2Tc+l)p+1 ’ and in the case of p even,
TC o o
c . 2 Г I V-! l — Q2k+i{£) /07 ,
$«(£) > — > --- ,, cos( 2 fe-fl)a?
pV ' 7Г J \ ^ J ( 2 fe+l)p
0
&=o ' ' n/2 oo
> — f cos( 2 fe + l)xdx dx
к f 2 (21c + l f
= ± V ( - D
0 fc=0
4 y ^ ^ y f c 1 —&*+i(£)
TV
k = 0(2fc+l) p +P'*
Analogously, by (5) and ( 6 ),
71 TC OO
,V,,(s) = s u p -1 f f > W У L .
s i n { k ( t — x ) + $ p i r } d t>’ tp 71 A A H 1
dx
/ 1 2
1 (? 2 fc+l(£) (2Й + 1У
0 k = 0 ' 7
sin {(2 A: +1) x + \p tv } dx.
If p is even, Spd) > 2
TV ' II
0 fc= 01 £? 2 fc+l(£)
( 2 k + l ) p sin( 2 &+l)#da? 4 TV
OO
1 @ 2 fc+l (£) (2Jc+l)v+1 ’
if p is odd,
Sp(S)
1 £ * 2 f c + l ( £ )(2fc+l)p cos ( 2 ^ + 1 ) a? I dx > 4 TV
k = о1 @ 2 f c + l ( £ )
T 2 fe+T f+r *
Thus, estimates (i) and (ii) are proved.
T heorem 1. I f Ф(/, £) > 0 for t e ( 0, тг>, £eS, then
7Г oo
sup Г|/(ж; £ ) - / И 1 ^ = — V 1 f 2fc+-1^ = Slip max |/(a?; £ ) - / И 1 *
/eZ(i) _■£ 7Г ^ ( 2 A;+ 1)2 /eZl
P roof. In view of (2),
тс я
sup Г |/(ж; I ) —/(ж)|йя? < — Гt<J>{t, £)dt = Л ( 0 ; I)—Л ( 0 )
И Р Л " o'
= sup max |/(я?; f ) —/(a?)|, /eZj — тт<а:<7г
where / х(ж) = \x\ for a?e< — тс, тс>. Since тс 4 cos(
OO2 &+l)a>
Л И - Y “ 7 Ż ( 2 /c + l )2 ’
Л И £) = @2k+l(£) cos( 2 & +l)#
2 ( 2 * + 1)a
for же<-тс, тс>, л\ге have
oo
Л(0; ^)-Л(О) = — У -1-~ g2fc+l(^-.
J J У ’ iz (2 & + 1 )2
On the other hand,- by Lemma,
n
TC oo
sup Г I f ( x ; £ ) —f ( x ) \ d x > sup f| f ( x - , l ) —f ( x ) \ d x У 1
/.zW J , Ti Z j (2 * + l)>
and the required formula follows now immediately.
Also, a related result holds. Namely:
T heorem 2. I f W ( t , £) = ± c o t% t — 0 { t , £) > 0 f o r t e ( 0, тг>, f e S , th e n
7Г OO
sup Г |/(ж; £) — /(a>)|<&& = — V (— 1)*-—g2fc+1^
M ?)- k тс A J k= 0 (21 c +1)2
= sup max \f(x-,£)-f(oo)\.
fеЛj — тг<х<тс
Proof. Identity (3) can be rewritten in the form
tt/2
/И i)-f(oo) =— f {/(a?+t)— TC J /0»— £)d$ —
о
-- Г {/(®+*+"-<)-Л*+я-я+«)} TC J тт 9 ,((, {)й«
тт /2
5 — Prace Matematyczne XI (1967)
for almost every x. Therefore,
тс тт/2 тс
sup Г |/(ж; £) — /(ж)|йя?< — I Г |)й < + Г(тг— £)й/1
К ) Л 71 У- тс% j
= /
2(
0; l ) - / 2(0) = sup max |/(ж; £)-/(ж)1, /еЛ^ —
where / 2(ж) is 27u-periodic, odd, such that
/ 2 ( ж ) =
ж if 0 < x < 7 г /2, ти—ж if 7 u /2 ^ a? ^ 7u.
Since
4 v i / 2 (ж) = — > ( - 1 )
7U fc =0
& sin(2&-f l)a?
( 2 &+ 1)2 ’
~ 4
f 2(x) = ---> ( - 1 ) 7Г fc =0
vA. cos(2&+l)&
(2fc +l)2~ ’ and
^ fc
= 0cos(2&+l)a?
(2 & + l)2_
for each x e ( — ж, тс}, £ e 3 , we have
/ г ( 0 ; ! ) —/ 2 ( 0 ) = — Y ( - l ) 4
7U k=o
к
4 £ 2 f c + l ( £ )( 2 &+ 1 )2 Applying Lemma, we obtain
sup Г|/(ж; £ ) - f ( x ) \ d x ^sup Г|/(ж; £)-/(ж)| V ( - 1 ) к—О
&4 @ 2 fc+l(£) ( 2 &+ 1 )2~’
and the conclusion is evident.
T heorem 3. I f <P(t, £) > 0 /or U<0,7u>, £еЯ, i f
fc=i
&+i 4fc(&+l) —(2&+1 _ _ _ _ _ )2 {>i(£) + £ 2 fc+i(£)
0 { 3 —4ei(^ )+ e,(f)}
as £ -> £0, Йе?г
7Г
sup / |/(*; = l - ffl(f) + 0 { 3 - 4 ffl(f) + ffs(f)}.
*44 - i
Proof. In view of ( 2 ),
TC * TC
sup f I/(ж; I ) —f(oc)\dx < — Г^ 2 Ф( 1 , £)dt
uz\f) Л 71 ^
7Г T U
= — Г ( 2 в т ^ ) 2 Ф(«, £)d<+— f {«2 — (2 sini«)a} Ф(«, £)dt
TC
J Tt Jо 0
TC TC
= — J*(l—cost)<P(t, £)dt-\-0 j J ( s i n i)dt j
1 — gi(£) + ^{3 — 4^ i (^) + ^2(^)}- By Lemma,
snp f \f(x-, £)- f(x )\d x / 4 1}-
sup f I/(ж; ^)—/(а?)|йя? V ( - l ) ;
*T 2 _i те «
1 g 2 fc+l(£) ( 2 fc + l )3 4
7T
00
I ’ f - u fc = 0
l - g l ( g )
( 2 &+ 1)3 4
7Г oo
fc= 0
g l(£ ) — g2fc+l(£)
( 2 &+ 1)3
4
7Г oo
к— о
l - g i ( D
2&+1
~\---- 4 7C fc
= 0
■ 1 ) fc-f-l 4ft(fe+l)-(2fc+l)2gi(g) + g2ft+i(g)
( 2 &+ 1)3
= l - e i( f ) + 0 { 3 - 4 e i(f) + gs(f)}.
Thus, the proof is completed.
B e mark. Under the assumption <f>(t, |) > 0 , the asymptotic rela
tion
sup max |/(ж ; |)- /(ж ) | = l - M ! ) + 0 { 3 —4e1(£) + M £)}
ftZ^ — tt <£C< t c
can easily be deduced, too (cf. [ 1 ]).
How, writting
Ap(£) = snp max |/(ж; £)—/(a?)|
and
= S lip Г|/(ж; £)—f(x)\dx, feW§)-n
we shall demonstrate the following T heorem 4. Suppose that
00
Qi(t} i) = —
vanishes only at the points t = hn (Tc = 0, ± 1 , ± 2 , ...) for any £eE.
Then, given an integer p ^ 1 ,
4 4 (f) = ± y (__ i )'''(>'+
fe=o
1 {?2fc+l ( I)
(2*4-1 f +1 P roof. By identity (4),
with
4 4 (f) = sup uwm *
7Z П
J | j f v){x+t)Qp{t,
— TC — TC
dx
0 0
Qp{t, i) = ^ COS (Jct + $pn).
k= l
The function Qp (t, £) is continuous in t it p ^ 2 , and Q'p{t, £) = —Qp^ x(t, £);
hence it can easily be shown that Qp(t, £) possesses at most two roots in every half-open interval of length 2n.
In the case of p odd (p = 2 r + l, r — 0,1, 2, ...) the function
Qp(t, £) = ( - l )r+1 ^ sin Jet k= l
vanishes only at the points t = Jen (Jc = 0, ± 1 , ± 2 ,...) . Then
TC
4 l,(f) < — TV f I Qp(t, 01 dt J
2
TV 0 fc= 1
sin Mdt
oo
d ~ g 2 fc+i(l)
( 2 f c + l f +1 ■
If p is even (p — 2r, r = 1, 2, ...), then
OO
Q„(t, £ ) = ( - ! ) ' £ с о Ш,
and the difference Qp{t, £)—Qp{n/2, f) vanishes only at the points t — ^ic-f-hic (J c = 0 , ± 1 , ± 2 , ...). Consequently,
71 TC
4 4 (f) = sup — f f f v){x+t){Qp{^ £)—QMiz, £)}dt uw$ n J J
dx
n/2 n
~ j I Qp(t, f ) - e » ( K f)|d« = -=-| (J - fj{Qp(t, D - Q p O tz , l)}dt
n/2 oo
0 /Г A =0
1 £? 2 fc+l(£)
( 2 k + l f " ( Я + 1 ) , Й
0 n/2
4 TC TC fc
= 0
fc 1 g 2 fc+l(f) (2Jc+l)P+1 '
On the other hand, by Lemma, 4 4 ( f ) > sup
utv
TC
*/ TU
— тс
oo
Jfc=
1 @ 2 fc+l(£) (2A;+l )J,+1 ‘ Thus the first part of Theorem is established.
Analogously,
1 "
4 (f) = SUP 1 / ( 0 ) —/ ( 0 ; f)| = sup — f f p){t)Qp(t, £)dt
uwp uwp 1 тс
sint)Qp(t, £)dtI for p odd,
— 7T
71
sign (cos t){Qp(t, £)—Q p {\ tc , $)}dt\ for p even, i.e.,
4(f)
k-= 0
1 ^ 2 fe+l(f) T 2 fc + l ) pTr ? and the second part is proved.
Applying identity (5) instead of (4) and arguing as previously, we get T h eorem 5. I f the function
O O
- V l l — Qk(£)
Qiih i) = 2 , — с о Ш
fc= 1
has at most one root in the interval ( 0 , тс), then
тс 00
4 V"! . ' . .ftp 1 g 2 fc+l(£) sup f |/(ж; i ) - f ( x ) \ d x = — V ( - 1 )A
(2&+l): г >+1 sup max |/(®; £)—/(®)|, /еИ7», — r:<a:<7T
where p is a positive integer, p > 2 , £eS.
S. Examples. 1 . It is well known that for Abel-Poisson integrals
7C OO
f(x-, r) = — J /(<) fa + J ^ r fccos&(ź-a?)jdż (0
^ -n ' fc= 1 *
< Г < 1 )
the assumptions of Theorems 1 and 4 are satisfied. Hence
sup I \j(ar,r)-f{x)\dx = — V
71 ^
1 — r2k+1 2 1
--- 2 = — (1— r )ln --- [-0(1—r) тс ( 2 &+ 1 ) Ti 1 —r
as r -> 1 —. This is an analogue of a formula given in [9]. Further,
r 4 ^ l _ r2fc+1
sup I \ f ( x - , r ) - f ( x ) \ d x = — у ( - 1 ) (i>+1) = 0 ( 1 —r)
/.irWjl 71 t i <M+1)
if p > 1 . The asymptotic formula for the last series is known (see [9]).
Similar results for biharmonic operators can be obtained as in [3].
2. Theorems 2 and 5 apply to the conjugate Abel-Poisson integrals
7Г OO
f(oc-,r) = --- j f ( t ) ^ r ksinic( t - x ) d t (0 < r < 1 ).
-ТГ f t = l
Therefore,
Л OO n J ,
i X
r ~ ~ 4 v i Д — r + 4 rarctgo;
sup I \f (x -,r )- f(x )\d x = — У ( _ 1 ) /ox. 7 i \ a = ~ ■ — “— dx
, ,(n J 7 Г ^ (2 a ;+ 1 )2 7 z J x
/ел9)Л (cf. [4]), and
it oo
sup Г \ f{x-, r)—f(x)\dx = — 'S1 ( —l ) kp 1 — r 2 *+l
(23fc+l> , 21 + 1 = 0(1 —r).
In this case the asymptotic formula can be deduced, too.
3. Now, we shall verify the second condition of Theorem 3 for the Jackson integrals (see [ 6 ])
f { X )
n) = ---3--- [• ( Sini,,;, 2 тг%( 2 % 2 + 1 ) J j \ sm |(2 — x) I
— TC
T U oo
ł i 4 ł + 2 Qk (n ) cos Jc {t — a?) j dt, where
Qk{n)
i ( 2 n — i ( 2 % —
—fc+l)! (n-Tc+1) 2 %( 2 % 2 + l ) [ {In—1c— 2 )!
{In—&+ 1 ) !
- 4 i!
2 %( 2 % 2 + l ) { 2n -T i-2 )\
0
' { n - J c - 2 )
when 0 < 1c < %— 2 , when n —2 < 1c < 2 n —2, when A: > 2 %— 2 .
Clearly, 3 —4g 1 (w) + £ 2 (w) = 0 { l / n 3), and
2 i ^
fc+i 4fe(fe+l) — (2fc4-l) 2 gi(^)4-g2fc+i(^)
( 2 & + 1 )3
\n-2j
2 %( 2 % 2 + l ) V ( - i ) * +il i --- -— 1 + f - i l (2 J t+ l)« |T
f t = l П— 1
+ 2n{2n
3
1 fc+i | 4 %(%2 — 1) 2(6%2+% — 1) 12% j
»a+ l ) (_ 1 ) \ ( 2 ft+ 1)5 ( 2 ft+ij* h 2 f t + l _ 1 ) +
* - F i V
+ 3 V ( - p ^ - J - - V ( - p * + - - j— = о Ш
2 » 2+ l Z j 2 f t + l Z j (2 ft+ l)» \ я 3/
Hence
n
sup I |/(#; w) —/(a?) I da?
j */
3 / 1
---
(-0
—2%2 \%3 = sup max |/(a?; %)—/(a?)|
/е ^2 — п^Ж^я
4. Consider, next, the Cesar о means
п П
Г (я i п) = - i j ^ A yn_kcoslc(t-x)^dt (у > 0 ),
’k = l
where
A l ( T ) (y+ l)(y+ 2 ) ...
( y + v )V
I
Write е*(тг) = An_kjAn for Те < тг, р£(тг) = 0 for Tc > n, and
00
0 y ( t ,
w) = £ + ^ e f c ( w) co s f c < . fc=l
If у ^ 1 , the kernel Фу {t , w) is non-negative ([11], pp. 78, 88 , 94).
Then, our Theorem 1 and Theorem 1 of [2] yield
r 2 1 п п I 1 \
sup \\f(x-j n ) - f ( x ) \ d x = — у ---\-0\ — )
fczwJn ■ 7z n \ n j as n -> oo.
Now, we shall show that the last formula remains true when 0 < у < 1.
Clearly, 1 0 y{t, n)\ < n-\- 1 for у > 0, źe < 0, л:>. We also have 0 y(t, n) — + sm{Nt — ъу-к) 2 y { l —y )0 { t,n ,y )
n -\-1 ( 2 sin ^)2 А уп{ 2 &тЩ\V +1 [n + 1 ) (n + 2 ) ( 2 sin -|£)£
for 0 < у < 1, te(Tz/n, 7u>, where iV = w+-J + |y , \0(t, n, y)\ < 1 (see [11], pp. 94-95); hence, by (2),
f l f ( ^ ; n ) —f(x)\dx
— 7V
TT TC
= ~ / I / { /( ^ + ^ ) - 2 /(ж )+/(ж -«)}Ф у(«, n)dt f ( x + t ) —2 f( x )+ f( x —t)
- n 2Tcjn
Y
dx-\-0 ( i )
- П — Г 7C I I W+l J W+l
— tc 2 n / n
( 2 sin ^ )2
+
7T
■Jy j { f { x + t ) - 2 f ( x ) + f { x - t ) }
n 2njn
dt +
sin(jW—%уп)
( 2 sin \t) UW+1 dt dx-\-0 ( i )
TC
1 f n -\-1 n , f ) + — 12(x, n , f )
-An. dx-\-0
Ш-
uniformly in f e Z ^ . Obviously, we can write 412( ж , n , f ) = J 2 { f{ x +t )—2 f{ x )+ f( x- t)}
2 ir/łl
s i n ( B -
( 2 sin -^)y+1 dt —
■n—n/n
- f \ f [ x + t + — \ — 2 f { x ) + f ( x — t — —
J [ \ n I \ n
njn
Я + Я / П
— J |/(ж+г- — ) - 2 f { x ) + f [ x - t + —
sin JV< + ^ (l+ y )- ---± y
tvn
Znjn П
h*K)]
sin
НН)Г dt
STZjn
я
= 2 i f + / ) t f ( » + « ) - 2 / W + /(» -» )} dt-
2я/п я— njn'
Зя/ft.
- - " Г / \ / u sin ^ + l ( l + y ) ---|77T
/ {/ e + * + T - 2/(e)+/(e _ ,_ ^ f— 2 s i n | l * + — n r — * -
TC + n/n
я— я/m /I/
я - я In
+ I 3 jt // i
sin(jV<— | ( l + y ) — - | y + ж + ^ - ^ - ) - 2 /( ж ) + / |ж - / + ^ -)[--- —--- j--- -d« +
M-C
2 № + / ) — 2 /(a>) +/(o?—<)] sin(JVi —^y 7 r)
/ ( Ж+ Н — I —2f(x) -\-f\x—t ---
n j \ n
( 2 sin ±*)y+1
8 ш | ж £+-| 2 ( 1 + 7 )— — 1 у ъ n
f 2 sin 4 +^)r
Bi n m - * (i + 7 ) - - i y7c
- |/ |® + * - - ) - 2 / ( » ) + / |® - « + - ] j -
П TC \ y +1
t — WJ
dt
= 2( + iK *b/) + Y a(a?, n , f ) ) - Y 3{x, w , / ) - Y 4 (a?, w,/) + Y 5 (£t?, n ,f) .
It is easily seen that
7X
Л f I Yi{x, n, / ) Idx = o l —
ny J \ n
for i = 1 , 2 , 3, 4, uniformly in feZ^K Further, TZ—TCjn
n, f) = J 2 /(^ + ^ )—/ |^ + ^ + f | Ж-Н — I \ l sin {Nt — \y, /J ( 2 sin |/ ) y+1
+ 1 2 / ( ^ 0 / ( * - * - " ) ] 7 a S r ,łt) +
/ | а ? + < + - ^ - | — 2 / ( ж ) + / | ж — < — X
7Г)
• / лт-ł 1 ч s i n m + i ( l + y ) — - | y :
sm(NY—£y7r) \ n
( 2 s in i( r [ 2 gi n l ( « + ^ ) ]
+ [/ (ж+(_^ )_2/<а:)+/(ж_(+^ )]x
+
X s in ( ^ —■JyTC) ( 2 sin ^)y+1
s in J ft—$ (l+ y) — ---iv ТГ 71 i V
n f
[2sini(*-^)] +
>dt.
Next, an argument similar to that of [10], p. 422, leads to 7t
"T f \Y ^ X’ ny J — rc О l—), \ n j uniformly in feZ^K
Consequently,
П T U
sup Г [ f (ж; п ) - / ( ж )| а Ь = - . - - sup \ { I ^ x , n , f ) \ d x + 0
2 / 71
/
* (w + 1 >J»<2sinW , dtf + O Ш
2 Inn
= — у---\-0
71 П
1
П J
Now let деТг. Then,
7t
gv(x’,n ) - g ( x ) = — fg' (t ) K ( t- x )d t,
TC J
with
к м = — + 2 j - j ~ ~ i 0 { t >n)dt-
k=l
ОIt can easily be observed ([11], p. I l l ) that
тг, x У 1 , i . 2 0 !(a>, n, y) B ( x , n , y ) К (x) --- cot \ x +
n -\-1 2 n Avn(2sm %x)y+l n2x2
and
K {x - \ -tz) =
У 1 ,! 2 в г{у,п,у) B { y ,n , y )
п -\-1 2 соЦу n A yn{2$m%y) 14ЛУ +1 п 2у 2
for ж е < 7 г /п, тс—n /пу, where у = л —ж, |0 х(ж, п, у )| < 1, |В(х, п, у )| < С, С being an absolute constant. Thus, by ( 6 ), we get
sup I 1/У(ж5 n )—f{x)\dx ^ sup j Igv(x\ n ) —g{x)\
An geTi
dx
— TC
Ti—nln
- f \K{x)-K{x+ic) TU J
тс jn
dx
1 у ТС—П/П
tu т г - | - 1
J c o t ^ ^ + o | —
n/n ' ^
2 hm / 1
— y — ---- bO — i, 7U П n
and the desired formula is estabhshed.
5. Finally, let / у(ж; n) be the conjugate Cesaro means, that is,
TC 0 0
f v(x\n) = ----— ff(t) ^ g l ( n ) s i n k ( t —x)dt.
-n k=l
If у > 3, the difference
П n) = |c o t \ t —
is non-negative for <e(0, 7 r>, n = 1, 2 , .. .([ 8 ], [11], p. 78). Then, Theo
rem 3.1 of [7] leads to
(cf. [ 8 ]). In view of Theorem 2, the left-hand side of the last formula can be replaced by
[1] Л. И. Б а у с о в , О приближении функций класса Za положительными методами суммирования рядов Фурье, Успехи Мат. Наук 16, 3 (99), (1961), рр. 143-149.
[2] Guo Z h u -ru i, Approximation of a continuous function by Cesaro means of its Fourier series, Scientia Sinica 11 (1962), pp. 1625-1634.
[3] G. К ан и ев , Об уклонении бигармонических в круге функций от их гра
ничных значений, Доклады Акад. Наук СССР 153 (1963), рр. 995-998.
[4] В. Sz. N a g y , 8ur Vordre de Vapproximation Tune fonction par son integrate de Poisson, Acta Math. Acad. Sci. Hung. 1 (1950), pp. 183-188.
[5] С. M. Н и к о л ь ск и й , Приближение функций тригонометрическими поли
номами в среднем, Известия Акад. Наук СССР, сер. матем., 10 (1946), рр. 207-256.
[6] Г. П. С аф р он ов а, О методе суммирования расходящихся рядов, связанном с сингулярным интегралом Джексона, Доклады Акад. Наук СССР 73 (1950), рр. 277-278.
[7] R. T a b e r sk i, Approximation to conjugate function, Bull. Acad. Polon.
Sci., Ser. sci. math. astr. et phys., 10 (1962), pp. 255-260.
[8] — Asymptotic formulae for Cesaro singular integrals, ibid., 10 (1962), pp. 637-640.
[9] А. Ф. Тиман, Точная оценка остатка при приближении периодических дифференцируемых функций интегралами Пуассона, Доклады Акад. Наук СССР 74 (1950), рр. 17-20.
[10] П. Л. У л ь я н о в , О приближении функций, Сибирский Матем. Журнал 5 (1964), рр. 418-437.
[11] A. Z y g m u n d , Trigonometric series, I, Cambridge 1959.
TU
sup f |/ y(^; n )— f(x)\dx.
M ( / ) - 7 Г