ROCZNIKI POLSKIEGO TOWARZYSTWA MATEMATYCZNEGO Séria I: PRACE MATEMATYCZNE XXX (1990)
W l o d z im ie r z L e n s k i (Poznan)
On the rate of almost strong summability of Fourier series
Abstract. We estimate the rates of almost strong summability of Fourier series of functions belonging to the spaces Lp (1 < p < oo) and C, by matrix means. As corollaries, the norm and pointwise approximation of functions from the Lipschitz classes is examined. We also present connections of our results with some earlier ones.
1. Introduction. Let X = Lp (1 ^ p ^ oo) [respectively X = C] be the space of all measurable real-valued functions/= /( • ) , 27t-periodic, Lebesgue- integrable with pth power [continuous] and let Lip (a, X) denote the class of all functions f e X whose moduli of continuity
co(S)x = co(S,f)x = sup ||<jp.(t)||x, where (px(t) = f ( x + t)+f(x — t) — 2f(x), satisfy the condition
co(< 5 )x = 0{Sa) as <5->0+.
Consider the Fourier series
S[/](*) = a0(f)/2 + Z (ak(f)cos kx + bk(f) sin kx)
k = 1
and denote by Я Д /] the partial sums of S [ / ] , and by
i n + m
*«СЛ = гтт I Ш] rrl "Г 1 n = n
the generalized de la Vallée Poussin means of the sequence {ЯД/]}.
The aim of this paper is to estimate the quantities Я М [/]Х,7>== { Z tm,k\Wm, n i n - f \ \ x Y lr,
m = 0
H k , n \ f \ x , T , q , r — A Z *m,k m — 0
1 n + m \ 1 lq
J Z IS„[ n - f n
r ] 1/r
x\
ниЛтЛ*) = { Z Wk^/](x)-/(*)lr}1/r>
m = 0
о / 1
n + m\r /<0 1
l rU u M \ t , , , ( x ) = •! Z «„.»(— T I |S „ [/]W -/(x )|«
^m = 0 уШ + m + 1 д = n 1 /
94 W. Le ns ki
where T = (rm k) is an arbitrary non-negative matrix and q, r > 0, by means of the quantity
, г 1 _ f v , Ш У \ lqV '
where g — g(-) will be either the modulus of continuity co(n/{- + 1 ),/)* , or the best approximation E.+n{f)x of / by trigonometric polynomials of degree at most n in the space X , or the function wJC(7i/(- + l))x, where
r sup
0 <u«<5
1 « ") 1 /P -$ \< P x(t)\pdt>
U 0 )
wx(<5)x = wx{ô,f)x = < ess sup {|<рх( 01 }
0 <t=$<
sup {\(pM}
Lo<t$<5
if X = Lp (1 ^ p < oo), if X = L°°,
if X = C.
We also show how some earlier results follow from ours.
By convention, the letter M will mean either an absolute constant or a constant depending on the parameters q, r, p, a, not necessarily the same at each occurrence.
2. Norm approximations. First we present a general estimate basing on the well-known results of Dahmen [1] and Steckin [ 8 ].
T heorem 1. I f f e X , then
H k , n L f ^ X , T , r ^ - M ^ f c , n [ 2 i - + n ( / ) x ] r , l , r , l '
P ro o f. The above mentioned results of Dahmen and Steckin may be written in the form
n + 2m
I K „ [ / ] - / l l * « ; M I (n + m + l ) - ' E r+n(f)x .
f i =
0
Therefore, using the monotonicity of the best approximation En(f), we immediately obtain our assertion.
The next theorem concerns a stronger estimate.
T heorem 2. I f f e X (here X = C when p = oo) and q(q—l) 1 ^ p ^ q (q ^ 2 ), then
H k , n \ f \ x , T , q , r ^ M h k ,n CO
+ 1 /X + Mhkf „
_ T , l , r , l
CO + 1 X j T , p , r , p / q
P ro o f. It is clear that
j n + m 'J l/q
m + Y I Is „[/](x )-/(*)l* < m +
+
1 n + m A ц = п i n-fm
- I l
j п/(и + т+1)
Я о
в') l/q
m + ц = п i n + m
J я/(т+1) я/(n + m+ 1)
q") l/q
ц = п
= + Л2 + у 43, where D (0 denotes the Dirichlet kernel.
Since p ^ n + m, we have |/)Д01 ^ n + m + l, whence , n + m+ 1 n/(" + "I + 1>
< --- J |фх(01^
, 71 0
and so
ii . || ( n WAJx^ co
1 я
~ \ <Px(t)D,i(t)dt Пп/(т + 1)
q') i/q
n + m+ 1
To estimate A2 we use the inequality \D (t)\ ^ (7i/2) |t| 1 and obtain n/(m + 1 )
A , ^ f
1 ф х ( 0 1dt,
whence n/(n + m+ 1)
я/(т+1) fn(t\ n + m \ CO I 1
I /Jr
cu i i /1+1 я/(и + т+1) 1 » = m A*+l ^ p = 0 P + ™+l
Next, we apply the following theorem of G. Hardy and J. E. Littlewood (cf. [10], p. 126 (5.20II)): I f /Д 0 = t~Qf( t ) , f ee X with 6 = 1/p + l / g - l and q i q - i y 1 ^ p ^ q (q> 2), then
Hence
In ( f\\9 °° ] !/«
Li4 r L+ I ( M / r + I M / r ) « M ||/B(-)ll*-
^ k= 1 J
1 m +
^ M and so
n + m i i A ^ = n 1
1 <P,(2t) :
- f n ti/(2(m + l)) 2Sint sin ((2/i+l )t)dt
q") i/q
| 4 3 | | *
(m+ 1)р/9я/(Ш +1) 1
i Ш l +p/9 ь ]и’
(m + 1 )p/4 w = о
I
ay /1+1
^ ( / i + l ) 1 р/Ч/1 + т + 1 ) p/q /1+1
i/p
l/q
96 W. Le ns ki
Thus, in the notation of Section 1, we obtain the desired inequality.
3. Pointwise approximations. The first theorem of this section is similar to Theorem 1:
T heorem 3. I f f e X , then
Нк,„[ЛтЛх) < МЛ*.„ w.
+ 1
x _ T, l , Г лP ro o f. By the standard computation,
t t
n sin(m +l)-sin(2n + m + 1)- 0m.,,[ /] ( * ) - / ( * ) = T" 2 k 0 I --- :--- dt
(m + l)sin2-
, ./(.+m+ 1 ) „ sin(m+l)Um(2n + m + l ) -
~ 2n( I + J + J )<P*W---: dt
0 я/(п + т + 1 ) 1 \ «//».4-14 я / ( т + 1 ) . T 1
(m + l)sk+
— B1 + B2 + B3.
The terms B{(i = 1, 2, 3) will be estimated successively. Using partial integra
tion we obtain
ID . 2n + m + l" /("+fm + 1\ /4IJ ( n
\Bt \ ^ ---- —---- J \(px(t)\ dt < w 2 k
*/(m+l) I I s 2l « i f
и + т + 1 /у ’
я/(и + m + 1 ) t m +
w .
ï h + i . Z
/i + l fi = m /1+1
m +
n + m w vx
î L + . l .
/i+ l
X f i ~ m /i + m+1
,B3| < —J î — î Af- Ц . £ wv
2(m+l )n/(w+D f m + l M=0 V/^+l/x
7Ü
a i
W
m *'/1 + 1
„ = o Z^ + m+1
Hence, by the monotonicity of we obtain our assertion.
In the case of stronger almost summability we have T heorem 4. I f f e X and q ( q - 1)-1 ^ p ^ q (q ^ 2), then Hk.n\f\T.q,r(X) ^ M Kn wv + 1 + Mh k,m
T , l , r , l
W, + 1 T, p, r . p/ q
P roof. Under the notations of the proof of Theorem 2,
{ J n + m I | S , m ( * ) - / M I , j ■) 1 tq « A t + A 2 + A 3 .
It is clear that
A x ^ w x .
\n + m+ l /x and, by the estimate of B2 from the previous proof,
/4, ^ ) л !
к
/* + l / x
p + m + 1 Next, integrating by parts, we get
A3 ^ M 1
( m + l ) ^ o Z (n + 1Y19' 1 wt
jU + 1
i I
pf * W* \ u + l j ) 1/p
< 2 llqM { У _______ У1! 1'* ____ I { Д ( / i + l ^ - ^ + m + l ^ j
Summing up our partial estimates we get the desired result.
4. Corollaries. In this section we confine our attention to non-negative triangular matrices T which lead to regular methods of summability.
First, we give corollaries to the theorems of Section 2.
The following result is an improved version of [3] and [7] for 0 < a < 1.
C o r o l l a r y 1. Suppose that {£„,,*}„ = 0 is a non-decreasing sequence with respect to m and that /e L ip (a , X). Then, for ae(0, 1),
Г if 0 < ar < 1,
Hk,nLf]x,T,r ^ "S M(tM log(2 + Tfc))“ if otr = 1,
^ M (tk,k)llr if ar > 1, for all natural n, where xk II g rn ?r 1 1 _ 1
Proof. This is a consequence of Theorem 1. Indeed, the assumption /eL ip (a, X) implies, by Jackson’s theorem, that E^(f)x ^ M( g+ 1)_“,
whence
n + m