W lodzimierz Lenski, Bogdan Szal
Pointwise strong and very strong approximation by matrix means of Fourier series
Abstract. We generalize and extend the some results of the paper [6]. Considering a wider class of function and more general means we obtain the results of the V.
Totik type [8, 9].
2000 Mathematics Subject Classification: 42A24, 41A25.
Key words and phrases: Strong and very strong approximation, Rate of pointwise summability, Matrix means.
1. Introduction. Let Lp (1 < p < ∞) [resp.C] be the class of all 2π–periodic real–valued functions integrable in the Lebesgue sense with p–th power [continuous]
over Q = [−π, π] and let X = Xp where Xp = Lp when 1 < p < ∞ or Xp = C when p = ∞. Let us define the norm of f ∈ Xp as
kfkXp = kf(x)kXp =
R
Q | f(x) |pdx1/p
when 1 < p <∞ , supx∈Q| f(x) | when p =∞.
Consider the trigonometric Fourier series
Sf (x) =ao(f) 2 +X∞
k=0
(ak(f) cos kx + bk(f) sin kx)
and denote by Skf the partial sums of Sf . Let
Hkq0,krf (x) :=
( 1 r + 1
Xr ν=0
|Skνf (x)− f (x)|q )1q
, (q > 0) .
where 0 ≤ k0< k1< k2< ... < kr (≥ r) , and
Hn,Aq f (x) :=
( n X
k=0
ank|Skf (x)− f (x)|q )1q
,
where A := (ank) (k, n = 0, 1, ...) be a lower triangular infinite matrix of real num- bers satisfying the following conditions:
(1) ank≥ 0 (k, n = 0, 1, ...) , ank= 0, k > n and Xn k=0
ank= 1.
Now we define two classes of sequences (see [4]).
A sequence c := (cn) of nonnegative numbers tending to zero is called the Rest Bounded Variation Sequence, or briefly c ∈ RBV S, if it has the property
(2) X∞
k=m
|cn− cn+1| ≤ K (c) cm
for all natural numbers m, where K (c) is a constant depending only on c.
A sequence c := (cn) of nonnegative numbers will be called the Head Bounded Variation Sequence, or briefly c ∈ HBV S, if it has the property
(3)
m−1X
k=0
|cn− cn+1| ≤ K (c) cm
for all natural numbers m, or only for all m ≤ N if the sequence c has only finite nonzero terms and the last nonzero terms is cN.
Therefore we assume that the sequence (K (αn))∞n=0 is bounded, that is, that there exists a constant K such that
0 ≤ K (αn) ≤ K
holds for all n, where K (αn) denotes the sequence of constants appearing in the in- equalities (2) or (3) for the sequence αn:= (ank)∞k=0.Now we can give the conditions to be used later on. We assume that for all n and 0 ≤ m ≤ n
(4)
X∞ k=m
|ank− ank+1| ≤ Kanm
and
(5)
mX−1 k=0
|ank− ank+1| ≤ Kanm
hold if αn:= (ank)∞k=0 belongs to RBV S or HBV S, respectively.
As a measure of approximation by the above quantities we use the pointwise characteristic
wx,β(δ; f)Lp:=
1 δ1+βp
Zδ
0
|ϕx(t)|
sin t
2
β!p
dt
1 p
,
where β ≥ 0 and
ϕx(t) = f (x + t) + f (x − t) − 2f (x) .
The quantity wx,β(δ; f)Lpis constructed on the base of the definition of the Lebesgue (Lp-points), introduced in [5] (cf.[1]) and every point x for which
1 δ1+βp
Zδ
0
|f (x ± t) − f (x)|
sin t
2
β!p
dt = o(1) as δ→ 0+
we will call the Lpβ-Lebesgue point.
We can observe that for f ∈ Xpeand ep ≥ p ≥ 1, by the Minkowski inequality, kw·,βf (δ)LpkX ep≤ ωX epf (δ) ,
where ωXf is the modulus of continuity of f in the space X = Xpedefined by the formula
ωXf (δ) := sup
0<|h|≤δkϕ·(h)kX .
We shall write I1 I2 if there exists a positive constant C such that I1≤ CI2. Let us consider a function wx of modulus of continuity type on the interval [0, 2π], i.e. a nondecreasing continuous function having the following properties:
wx(0) = 0, wx(δ1+ δ2) ≤ wx(δ1) + wx(δ2) for any 0 ≤ δ1≤ δ2≤ δ1+ δ2≤ 2π and let for β ≥ 0
Lp(wx)?β = {f ∈ Lp:
"
1 δ1+βp
Z δ
0 |ϕx(t) − ϕx(t ± γ)|
sin t
2
β!p
dt
#1/p
wx(γ) and wx,βf (δ)Lp wx(δ) , where γ, δ > 0} .
Since, for β > α ≥ 0,
wx,β(δ; f)Lp=
δαp δβp
1 δ1+αp
Zδ
0
|ϕx(t)|
sint
2
αp sin2tβ sint2α
!p dt
1 p
≤πα 2β
δp(α−β) 1 δ1+αp
Zδ
0
|ϕx(t)|
sin t
2
αp
tp(β−α)dt
1 p
≤πα
2βwx,α(δ; f)Lp
and
"
1 δ1+βp
Z δ 0
|ϕx(t) − ϕx(t ± γ)|sins 2
βp
dt
#1/p
≤ πα 2β
"
1 δ1+αp
Z δ 0
|ϕx(t) − ϕx(t ± γ)|sins 2
αp dt
#1/p
,
we have Lp(wx)∗α⊂ Lp(wx)∗β. Let also
Lp(ω) = {f ∈ Lp: ωLpf (δ)≤ ω (δ)} , where ω is a modulus of continuity type function.
It is well-known that H0,rq f (x)− means tend to 0 (as r → ∞) at the Lp−point x of f ∈ Lp (1 < p ≤ ∞) . In [2] this fact was by G. H. Hardy, J. E. Littlewood proved as a generalization of the Fej´er classical result on the convergence of the (C, 1) -means of Fourier series. In [6] W. Lenski obtained an estimation of the Hkq0,krf (x) means as an approximation version of the Totik type (see [8, Na1]) generalization of the result of G. H. Hardy, J. E. Littlewood, considered by N. L.
Pachulia in [7]. Namely, he proved the following theorem:
Theorem 1.1 [6] If f ∈ Lp(wx)?0 (1 < p ≤ 2) and 0 < k0 < k1 < k2 < ... < kr
(≥ r), then
(6) Hkq0,krf (x) wx
π
k0+ 1
log kr+ 1 r + 1/2, where p1+1q = 1.
In the present paper we shall generalize this result taking a wider class of func- tions and more general means. We also give some corollaries on norm approxima- tion.
2. Main results. We start with the following theorem:
Theorem 2.1 If f ∈ Lp(wx)?β (p > 1, β ≥ 0) and 0 < k0 < k1 < k2 < ... < kr
(≥ r), then
(7) Hkq0,krf (x) wx
π
k0+ 1
1 + logkr+ 1/2 r + 1/2
,
where q is such that 1 < q (q − 1)−1 ≤ p ≤ q.
Remark 2.2 If β = 0 and p = q (q− 1)−1, then we have (6) as a corollary of (7).
Similarly, we have estimate for norms.
Theorem 2.3 If f ∈ Lp(ω) (p ≥ 2) and 0 < k0< k1< k2< ... < kr (≥ r), then
Hkq0,krf (·)
Lp ω
π
k0+ 1
1 + logkr+ 1/2 r + 1/2
, where q is such that 0 < q ≤ p.
Basing on Theorem 2.1 and 2.3, we can formulate the next theorems.
Theorem 2.4 Let (1) and (4) hold. If f ∈ Lp(wx)?β (p > 1, β ≥ 0), then
Hn,Aq f (x) ( n
X
k=0
ank
wx
π
k + 1
q)1q
,
where q is such that 1 < q (q − 1)−1 ≤ p ≤ q.
Theorem 2.5 Let (1) and (5) hold. If f ∈ Lp(wx)?β (p > 1, β ≥ 0), then
Hn,Aq f (x)
[Xn+14 ]
k=1
an,4k
wx
π k + 1
q +
wx
π
n + 1
log ((n + 1) ann)q
1 q
,
where q is such that 1 < q (q − 1)−1 ≤ p ≤ q.
Theorem 2.6 Let (1) and (4) hold. If f ∈ Lp(ω) (p ≥ 2), then
Hn,Aq f (·)
Lp ( n
X
k=0
ank
ω
π k + 1
q)1q , where q is such that 0 < q ≤ p.
3. Lemmas. With the notation
Φxf (δ, γ) :=1 δ
Z γ+δ γ
ϕx(t) dt, wxf (δ, γ)Lp:=
"
1 δ
Z γ+δ
γ |ϕx(t)|pdt
#1/p
we can formulate the following lemma:
Lemma 3.1 If f ∈ Lp(wx)?β (p ≥ 1) , then
|Φxf (δ, γ)| ≤ wxf (δ, γ)Lp ≤ πβ(1 + βp)1/p(wx(δ) + wx(γ)) for any positive γ, δ.
Proof The first inequality is evident, then we prove the second one only. Using integration by parts we obtain
wxf (δ, γ)Lp=
"
1 δ
Z δ
0 |ϕx(t + γ)|pdt
#1/p
≤
"
1 δ
Z δ
0 |ϕx(t + γ) − ϕx(t)|pdt
#1/p
+
"
1 δ
Z δ
0 |ϕx(t)|pdt
#1/p
=
1 δ
Z δ 0
|ϕx(t + γ) − ϕx(t)|sint2βp
sin2tβp dt
1/p
+
1 δ
Z δ 0
|ϕx(t)|sin2tβp
sint2βp dt
1/p
≤ πβ
"
1 δ
Z δ 0
1 tβp
d dt
Z t 0
|ϕx(s + γ) − ϕx(s)|sins 2 β
p
ds
dt
#1/p
"
1 δ
Z δ 0
1 tβp
d dt
Z t 0
|ϕx(s)|sins 2 β
p
ds
dt
#1/p
= πβ ("
1 δ
1 tβp
Z t 0
|ϕx(s + γ) − ϕx(s)|sins 2 β
p ds
δ
0
βp δ
Z δ 0
1 t1+βp
Z t 0
|ϕx(s + γ) − ϕx(s)|sins 2
βp
dsdt
#1/p
+
"
1 δ
1 tβp
Z t 0
|ϕx(s)|sins 2
βp
ds
δ
0
βp δ
Z δ 0
1 t1+βp
Z t 0
|ϕx(s)|sins 2 β
p
dsdt
#1/p
. Since f ∈ Lp(wx)?β, we have
wxf (δ, γ)Lp≤ πβ
"
(wx(γ))p+βp δ
Z δ 0
(wx(γ))pdt
#1/p
"
(wx(δ))p+βp δ
Z δ 0
(wx(t))pdt
#1/p
≤ πβ(1 + βp)1/p(wx(δ) + wx(γ))
and this ends of our proof.
Lemma 3.2 If f ∈ Lp(ω) (p ≥ 1), then
kΦ·f (δ, γ)kLp≤ kw·f (δ, γ)LpkLp≤ 2 (ω (δ) + ω (γ)) and
"
1 δ1+βp
Z δ
0 |ϕ·(t) − ϕ·(t ± γ)|sin t 2
β!p
dt
#1/p Lp
≤ 2ω (γ)
for any γ, δ > 0 and β ≥ 0.
Proof The first inequality is proved in [6]. Therefore we prove the second one, only. If we change the order of integration, then we obtain
"
1 δ1+βp
Z δ
0 |ϕ·(t) − ϕ·(t ± γ)|sin t 2
β!p dt
#1/p
Lp
≤
"
1 δ
Z δ
0 |ϕ·(t) − ϕ·(t ± γ)|pdt
#1/p Lp
≤ (1
δ Z δ
0
Z π
−π|ϕx(t) − ϕx(t ± γ)|pdx
dt
)1/p
≤ (1
δ Z δ
0
[2ωLpf (γ)] dt )1/p
= 2ωLpf (γ)
and our lemma follows.
Lemma 3.3 [N1, Theorem 5.20 II, Ch. XII] Suppose that 1 < q (q − 1)−1 ≤ p ≤ q and ξ = p1+1q − 1. Ift−ξg (t)
∈ Lp then
(8)
(|a0(g)|q 2 +X∞
k=0
(|ak(g)|q+ |bk(g)|q) )1q
Zπ
−π
t−ξg (t)pdt
1 p
.
4. Proofs of the Results. We prove the theorems on pointwise approxima- tion only. The estimations of the norm of our means are an immediate consequences of these which will be proved.
4.1. Proof of Theorem 2.1. Let as usual
Hkq0,kr(x) = ( 1
r + 1 Xr ν=0
1
π Z π
0
ϕx(t) Dkν(t) dt
q)1/q
≤ Ar+ Br+ Cr,
where
Ar= ( 1
r + 1 Xr ν=0
1 π
Z 2δν
0
ϕx(t) Dkν(t) dt
q)1/q
Br= ( 1
r + 1 Xr ν=0
1
π Z 2γr
2δν
ϕx(t) Dkν(t) dt
q)1/q
Cr= ( 1
r + 1 Xr ν=0
1
π Z π
2γr
ϕx(t) Dkν(t) dt
q)1/q
,
with Dkν(t) = sin((kν+12)t)
2 sin2t , δν= k π
ν+1/2 and γr= r+1/2π , and
Ar≤ π 2
( 1 r + 1
Xr ν=0
"
kν+ 1/2 π
Z 2δν
0 |ϕx(t)| dt
#q)1/q .
Using integration by parts we obtain Z 2δν
0 |ϕx(t)| dt = Z 2δν
0
|ϕx(t)|sin2tβ sin2tβ dt
Z 2δν
0
1 tβ
d dt
Z t
0 |ϕx(s)|sins 2 βds
dt
= 1 tβ
Z t
0 |ϕx(s)|sins 2
β
ds
2δν
0
+βZ 2δν
0
1 t1+β
Z t
0 |ϕx(s)|sins 2
β
dsdt
= 2δνwx,βf (2δν)L1+ βZ 2δν
0
wx,βf (t)L1dt
≤ 2δνwx,βf (2δν)Lp+ βZ 2δν
0
wx,βf (t)Lpdt
(9) ≤ 2δνwx(2δν) + βZ 2δν
0
wx(t) dt δνwx(2δν) , whence
Ar ( 1
r + 1 Xr ν=0
[wx(2δν)]q )1/q
wx(δ0).
The quantities Brand Cr we will estimate by the Totik method [9].
At the begin, we divide the term Brinto the two parts and use (9). Thus
Br= ( 1
r + 1 Xr ν=0
1 π
Z 2γr
2δν
ϕx(t) Dkν(t) dt
q)1/q
≤ ( 1
r + 1
νX0−1 ν=0
+ Xr ν=ν0
! 1 π
Z 2γr
2δν
ϕx(t) Dkν(t) dt
q)1/q
≤ ( 1
r + 1
νX0−1 ν=0
1 π
Z 2γr
2δν
ϕx(t) Dkν(t) dt
q)1/q
+ ( 1
r + 1 Xr ν=ν0
1
π Z 2γr
2δν
ϕx(t) Dkν(t) dt
q)1/q
( 1
r + 1
νX0−1 ν=0
"
kν+ 1/2 π
Z 2δν
2γr
|ϕx(t)| dt
#q)1/q + Br,ν0
wx(δ0) + Br,ν0,
where the index ν0 is such that kν0−1 < r≤ kν0. Next, we divide the term Br,ν0
into the three following parts
Br,ν0. = ( 1
r + 1 Xr ν=ν0
1
π Z 2γr
2δν
ϕx(t) Dkν(t) dt
q)1/q
= 1
2 ( 1
r + 1 Xr ν=ν0
1 π
Z 2γr
2δν
+Z 2γr−δν
δν
+Z 2γr
2γr−δν
− Z 2δν
δν
!
ϕx(t) Dkν(t) dt
q)1/q
≤ Br,ν1 0+ Br,ν2 0+ Br,ν3 0, where the first term
B1r,ν0 = 1 2
( 1 r + 1
Xr ν=ν0
1 π
Z 2γr
2δν
+Z 2γr−δν
δν
!
ϕx(t) Dkν(t) dt
q)1/q
= 1 2
( 1 r + 1
Xr ν=ν0
1 π
Z 2γr
2δν
[ϕx(t) Dkν(t) + ϕx(t − δν) Dkν(t − δν)] dt
q)1/q
≤ 1 2
( 1 r + 1
Xr ν=ν0
1 π
Z 2γr
2δν
(ϕx(t) − ϕx(t − δν)) Dkν(t) dt
q)1/q
+1 2
( 1 r + 1
Xr ν=ν0
1
π Z 2γr
2δν
ϕx(t − δν) (Dkν(t) + Dkν(t − δν)) dt
q)1/q
≤ 1 2
( 1 r + 1
Xr ν=ν0
Z 2γr
2δν
1
t |ϕx(t) − ϕx(t − δν)| dt
q)1/q
+1 2
( 1 r + 1
Xr ν=ν0
1 π
Z 2γr
2δν
ϕx(t − δν) 1
2 sin2t − 1 2 sint−δ2ν
!
sin(2kν+ 1) t
2 dt
q)1/q
( 1
r + 1 Xr ν=ν0
Z 2γr
2δν
1
t |ϕx(t) − ϕx(t − δν)| dt
q)1/q
+ ( 1
r + 1 Xr ν=ν0
δν
Z 2γr
2δν
|ϕx(t − δν)|
t (t− δν) dt
q)1/q
≤ ( 1
r + 1 Xr ν=ν0
Z 2γr
2δν
1
t |ϕx(t) − ϕx(t − δν)| dt
q)1/q
+ ( 1
r + 1 Xr ν=ν0
δν
Z 2γr
δν
|ϕx(t)|
t2 dt
q)1/q
. Integrating by parts and applying our assumption we get
B1r,ν0 ( 1
r + 1 Xr ν=ν0
Z 2γr
2δν
1 t1+β
d dt
Z t
0 |ϕx(s) − ϕx(s − δν)|sins 2 βds
dt
q)1/q
+ ( 1
r + 1 Xr ν=ν0
δν
Z 2γr
δν
1 t2+β
d dt
Z t
0 |ϕx(s)|sins 2 βds
dt
q)1/q
≤ ( 1
r + 1 Xr ν=ν0
"
1 (2γr)1+β
Z 2γr
0 |ϕx(s) − ϕx(s − δν)|sins 2
β
ds
#q)1/q
+ ( 1
r + 1 Xr ν=ν0
"
1 (2δν)1+β
Z 2δν
0 |ϕx(s) − ϕx(s − δν)|sins 2
β
ds
#q)1/q
+ ( 1
r + 1 Xr ν=ν0
Z 2γr
2δν
1 t2+β
Z t
0 |ϕx(s) − ϕx(s − δν)|sins 2
β
dsdt
q)1/q
+ ( 1
r + 1 Xr ν=ν0
"
δν
1 (2γr)2+β
Z 2γr
0 |ϕx(s)|sins 2 βds
#q)1/q
+ ( 1
r + 1 Xr ν=ν0
"
1 (δν)1+β
Z δν
0 |ϕx(s)|sins 2
β
ds
#q)1/q
+ ( 1
r + 1 Xr ν=ν0
δν
Z 2γr
δν
1 t3+β
Z t
0 |ϕx(s)|sins 2
β
dsdt
q)1/q
≤ wx(δ0) + ( 1
r + 1 Xr ν=ν0
wx(δν)Z 2γr
2δν
1 tdt
q)1/q
( 1 r + 1
Xr ν=ν0
δν
wx(2γr) 2γr
q)1/q
+ wx(δ0) + ( 1
r + 1 Xr ν=ν0
δν
Z 2γr
δν
wx(t) t2 dt
q)1/q
wx(δ0)
1 + logkr+ 1/2 r + 1/2
+
( 1 r + 1
Xr ν=ν0
[wx(δν)]q )1/q
+ ( 1
r + 1 Xr ν=ν0
wx(δν)Z 2γr
δν
1 tdt
q)1/q
wx(δ0)
1 + logkr+ 1/2 r + 1/2
. Consequently, by Lemma 3.1,
Br,ν2 0 =1 2
( 1 r + 1
Xr ν=ν0
1 π
Z 2γr
2γr−δν
ϕx(t) Dkν(t) dt
q)1/q
≤1 4
( 1 r + 1
Xr ν=ν0
Z 2γr
2γr−δν
1
t |ϕx(t)| dt
q)1/q
≤1 4
( 1 r + 1
Xr ν=ν0
1
2γr− δν Z 2γr
2γr−δν
|ϕx(t)| dtq)1/q
≤ 1 4
( 1 r + 1
Xr ν=ν0
"
δν
2γr− δν 1 δν
Z 2γr−δν+δν
2γr−δν
|ϕx(t)| dt
#q)1/q
≤1 4
( 1 r + 1
Xr ν=ν0
δν
2γr− δν[wx(2γr− δν) + wx(δν)]q)1/q
≤ 1 4
( 1 r + 1
Xr ν=ν0
δνwx(2γr− δν)
2γr− δν + wx(δν)
q)1/q
( 1
r + 1 Xr ν=ν0
δνwx(δν) δν
+ wx(δν)q)1/q
wx(δ0)
and
B3r,ν0 = 1 2
( 1 r + 1
Xr ν=ν0
1 π
Z 2δν
δν
ϕx(t) Dkν(t) dt
q)1/q
≤1 4
( 1 r + 1
Xr ν=ν0
Z 2δν
δν
|ϕx(t)|1 tdt
q)1/q
( 1
r + 1 Xr ν=ν0
[wx(δν)]q )1/q
wx(δ0) . Thus
Br wx(δ0)
1 + logkr+ 1/2 r + 1/2
. For the estimate of Cr we use our notation in the form Φxf (δ0, t) := δ10 Rt+δ0
t ϕx(u) du with δ0= k0+1/2π . Then
Cr= ( 1
r + 1 Xr ν=0
1 π
Z π 2γr
ϕx(t)
2 sint
2
−1
sin
kν+1 2
t
dt
q)1/q
≤ ( 1
r + 1 Xr ν=0
1 π
Z π 2γr
Φxf (δ0, t)− ϕx(t) 2 sint2
sin
kν+1 2
t
dt
q)1/q
+ ( 1
r + 1 Xr ν=0
1
π Z π
2γr
Φxf (δ0, t) 2 sin2t sin
kν+1 2
t
dt
q)1/q
= Cr1+ Cr2 and, by inequality (8),
Cr1 1 (1 + r)1/q
(1 π
Z π 2γr
[Φxf (δ0, t)− ϕx(t)] 2 sin t
2
−1
t1−1/p−1/q
p
dt )1/p
1
(1 + r)1/q
Z π 2γr
|Φxf (δ0, t)− ϕx(t)|p
t1+p/q dt
1/p
= 1
(1 + r)1/q (Z π
2γr
1 δ0t1/p+1/q
Z t+δ0
t
[ϕx(u) − ϕx(t)] du
p
dt )1/p
≤ 1
(1 + r)1/q (Z π
2γr
"
1 δ0t1/p+1/q
Z δ0
0 |ϕx(u + t) − ϕx(t)| du
#p
dt )1/p
. Applying the generalized Minkowski inequality and the partial integration we obtain
Cr1 1 δ0(1 + r)1/q
Z δ0
0
Z π 2γr
|ϕx(u + t) − ϕx(t)|p t1+p/q dt
1/p
du 1
δ0(1 + r)1/q
· Z δ0
0
Z π 2γr
1 t1+p/q+βp
d dt
Z t 0
|ϕx(u + v) − ϕx(v)|sinv 2
βp
dv
dt
1/p du
= 1
(1 + r)1/q 1 δ0
Z δ0
0
( 1
t1+p/q+βp Z t
0
|ϕx(u + v) − ϕx(v)|sinv 2
βp
dv
π
t=2γr
+ (1 + p/q + βp)Z π 2γr
1
t2+p/q+βp Z t
0
|ϕx(u + v) − ϕx(v)|sinv 2 β
p dv
dt
1/p du.
Hence, since f ∈ Lp(wx)?β,we have
Cr1 1 (1 + r)1/q
1 δ0
Z δ0
0
1
π1+p/q+βp Z π
0
|ϕx(u + v) − ϕx(v)|sinv 2
βp
dv
+
"
1 (2γr)1+p/q+βp
Z 2γr
0
|ϕx(u + v) − ϕx(v)|sinv 2 β
p dv
#
+Z π 2γr
1
t1+p/q(wx(u))pdt
1/p
du
≤ 1
(1 + r)1/q 1 δ0
Z δ0
0
(
[wx(u)]p 1
πp/q + 1
(2γr)p/q +Z π 2γr
1 t1+p/qdt
!)1/p
du
1
(1 + r)1/qwx(δ0) (
1 + 1
(γr)p/q +Z π 2γr
t−(1+p/q)dt )1/p
1
(1 + r)1/qwx(δ0)
( 1
(γr)p/q +t−p/q
−p/q
π
t=2γr
)1/p
≤ 1
(1 + r)1/qwx(δ0)
( 1
(γr)p/q
1 +q
p
)1/p
wx(δ0) .
For the estimate of Cr2 we will consider the two cases r ≤ k0 and r ≥ k0 . In the first case, using the partial integration and inequality (8), we obtain
Cr2= 1 2 (r + 1)1/q
( r X
ν=0
1 π
Z π 2γr
Φxf (δ0, t) sin2t
d dt
cos kν+12 t kν+12
! dt
q)1/q
= 1
2π (r + 1)1/q
Xr ν=0
"
Φxf (δ0, t) sint2
cos kν+12 t kν+12
#π
2γr
− Z π
2γr
d dt
Φxf (δ0, t) sin2t
cos kν+12 t kν+12 dt
q)1/q
1 (r + 1)1/q
( r X
ν=0
"
Φxf (δ0, 2γr) sin γr
cos kν+122γr kν+12
+
Z π 2γr
d dt
Φxf (δ0, t) sin2t
cos kν+12 t kν+12 dt
#q)1/q
and analogously
Cr2 1 (r + 1)1/q
((r + 1)1/q|Φxf (δ0, 2γr)|
γr(k0+ 1)
+ 1
k0+ 1
Z π 2γr
d
dt
Φxf (δ0, t) sint2
t1−1/p−1/q
p
dt
1/p)
1
γr(k0+ 1)wx(δ0) + 1 k0+ 1
wx(2γr) 2γr
+ 1
(k0+ 1) (r + 1)1/q
Z π 2γr
d
dt
Φxf (δ0, t) sint2
t1−1/p−1/q
p
dt
1/p
wx(δ0) + 1 k0+ 1
wx(2δ0) 2δ0
+ 1
(k0+ 1) (r + 1)1/q
"Z π 2γr
d
dtΦxf (δ0, t)
sin2t −Φxf (δ0, t) cos2t 2 sin2t2
!
t1−1/p−1/q
p
dt
#1/p
wx(δ0)
+ 1
(k0+ 1) (r + 1)1/q
"Z π 2γr
1
δ0|ϕx(δ0+ t) − ϕx(t)|
t1/p+1/q +wx(δ0) + wx(t) t1+1/p+1/q
p
dt
#1/p
wx(δ0) + 1
(k0+ 1) (r + 1)1/q
· 1 δ0p
Z π 2γr
t−1−p/q−βpd dt
Z t 0
|ϕx(δ0+ u) − ϕx(u)|sinu 2 β
p
du
dt
1/p
+ 1
(k0+ 1) (r + 1)1/q (Z π
2γr
wx(δ0) t1+1/p+1/q
p
dt
1/p +Z π
2γr
wx(t) t1+1/p+1/q
p
dt
1/p)
wx(δ0)
+ 1
(k0+ 1) (r + 1)1/qδ0
1
π1+p/q+βp Z π
0
|ϕx(δ0+ u) − ϕx(u)|sinu 2
βp
du
− 1
(2γr)1+p/q+βp Z 2γr
0
|ϕx(δ0+ u) − ϕx(u)|sinu 2
βp
du
+ (1 + p/q + βp)Z π 2γr
t−2−p/q−βp Z t
0
|ϕx(δ0+ u) − ϕx(u)|sinu 2
βp
dudt
1/p
+ wx(δ0) (k0+ 1) (r + 1)1/q
Z π 2γr
t−p−1−p/qdt
1/p
+ 1
(k0+ 1) (r + 1)1/q
Z π 2γr
wx(t) /t t1/p+1/q
p
dt
1/p
wx(δ0)+ wx(δ0) (k0+ 1) (r + 1)1/qδ0
"
1 πp/q + 1
γrp/q
+ (1 + p/q + βp)Z π 2γr
t−1−p/qdt
#1/p
+(γr)−1−1/qwx(δ0)
(k0+ 1) (r + 1)1/q + (γr)−1/qwx(γr) (k0+ 1) (r + 1)1/qγr
wx(δ0) + wx(δ0) (k0+ 1) (r + 1)1/qδ0
"
1 γrp/q
+ (1 + p/q + βp)(2γr)−p/q p/q
#1/p
+ wx(δ0)
(k0+ 1) (r + 1)1/q(γr)1+1/q + wx(γr) /γr
(k0+ 1) (r + 1)1/qγr1/q
wx(δ0) + wx(δ0)
(k0+ 1) δ0(r + 1)1/qγr1/q
+ wx(δ0)
(k0+ 1) (r + 1)1/qγr1+1/q
wx(δ0) . In the case r ≥ k0we will divide the interval of integrability into two parts. Thus, by inequality (8), Lemma 3.1 and the partial integration,
Cr2= 1 2 (r + 1)1/q
( r X
ν=0
1
π Z π
2γr
Φxf (δ0, t) sin2t sin
kν+1 2
t
dt
q)1/q
1
(r + 1)1/q (Z 2δ0
2γr
Φxf (δ0, t)
sint2 t1−1/p−1/q
p
dt )1/p
+ 1
2 (r + 1)1/q ( r
X
ν=0
1 π
Z π 2δ0
Φxf (δ0, t) sin2t
d dt
cos kν+12 t kν+12
! dt
q)1/q
≤ 1
(r + 1)1/q (Z 2δ0
2γr
Φxf (δ0, t)
sin2t t1−1/p−1/q
p
dt )1/p
+ 1
2π (r + 1)1/q ( r
X
ν=0
"
Φxf (δ0, t) sint2
cos kν+12 t kν+12
#π
2δ0
− Z π
2δ0
d dt
Φxf (δ0, t) sint2
cos kν+12 t kν+12 dt
q)1/q
1 (r + 1)1/q
(Z 2δ0
2γr
wx(δ0) + wx(t) t1/p+1/q
p
dt )1/p
+Φxf (δ0, 2δ0) sin δ0
1 k0+12
+ 1
(k0+ 1) (r + 1)1/q ( r
X
ν=0
Z π 2δ0
d dt
Φxf (δ0, t) sin2t
cos
kν+1 2
t
dt
q)1/q
. Using once more inequality (8) and Lemma 3.1
Cr2 wx(δ0) (r + 1)1/q
(Z 2δ0
2γr
t−1−p/qdt )1/p
+wx(δ0) + wx(2δ0) (k0+ 1) δ0
+ 1
(k0+ 1) (r + 1)1/q
Z π 2δ0
d
dt
Φxf (δ0, t) sint2
t1−1/p−1/q
p
dt
1/p
wx(δ0)
+ 1
(k0+ 1)1+1/q
"Z π 2δ0
d
dtΦxf (δ0, t)
sint2 −Φxf (δ0, t) cos2t sin2t2
!
t1−1/p−1/q
p
dt
#1/p . Similarly to the first case
Cr2 wx(δ0) + 1 (k0+ 1)1+1/q
· 1 δp0
Z π 2δ0
t−1−p/q−βpd dt
Z t 0
|ϕx(δ0+ u) − ϕx(u)|sinu 2
βp
du
dt
1/p
+ 1
(k0+ 1)1+1/q (Z π
2δ0
wx(δ0) t1+1/p+1/q
p
dt
1/p +Z π
2δ0
wx(t) t1+1/p+1/q
p
dt
1/p)
wx(δ0) + 1 (k0+ 1)1+1/qδ0
1
π1+p/q+βp Z π
0
|ϕx(δ0+ u) − ϕx(u)|sinu 2 p
du
+ 1
(2δ0)1+p/q+βp Z 2δ0
0
|ϕx(δ0+ u) − ϕx(u)|sinu 2
βp
du
+ (1 + p/q + βp)Z π 2δ0
t−2−p/q−βp
Z t 0
|ϕx(δ0+ u) − ϕx(u)|sinu 2 β
p
du
dt
1/p
+ wx(δ0) (k0+ 1)1+1/q
Z π 2δ0
t−p−1−p/qdt
1/p
+ 1
(k0+ 1)1+1/q
Z π 2δ0
wx(t) /t t1/p+1/q
p
dt
1/p
wx(δ0) + wx(δ0) (k0+ 1)1+1/qδ0
"
1 πp/q + 1
δp/q0 + (1 + p/q + βp)Z π 2δ0
t−1−p/qdt
#1/p
+(δ0)−1−1/qwx(δ0)
(k0+ 1)1+1/q +(δ0)−1/qwx(δ0) (k0+ 1)1+1/qδ0