ON A GENERALIZATION OF LIPSCHITZ’S CLASSES
XH. Z. KRASNIQI
DEPARTMENT OFMATHEMATICS ANDCOMPUTERSCIENCES, AVENUE"MOTHERTERESA" 5,
PRISHTINË, 10000, REPUBLIC OFKOSOVO
xheki00@hotmail.com
Received 26 March, 2008; accepted 06 August, 2008 Communicated by H. Bor
ABSTRACT. In this paper we obtain a generalization of Lipschitz’s classesΛm(β, p, r)defined in [1]. We give necessary conditions for even or odd functions with Fourier series to belong to the classesΛm(p, r, α). We also give sufficient conditions for even or odd functions with Fourier series to belong to the same classes.
Key words and phrases: Lipschitz classes, Fourier series.
2000 Mathematics Subject Classification. 42A20, 42A32.
1. DEFINITIONS ANDUSEFULSTATEMENTS
We consider the series
(1.1) a0
2 +
∞
X
n=1
ancosnx or
(1.2)
∞
X
n=1
ansinnx whereanare Fourier coefficients of integrable functionf.
Definition 1.1. We say that a functionf belongs toW Ap,(1< p <∞)if
∞
X
n=1
np−2
∞
X
k=n
|∆ak|
!p
<+∞
where∆ak =ak−ak+1 (see [1]).
We say that any function α(t) is a function of type σ (see [4]) if it is measurable in [0,1], integrable in[δ,1]for each δ∈ (0,1), and there exist real numbersC1,α >0, σandδ0 ∈(0,1) such that
(1) α(t)≥C1,α, for allt∈[0,1];
096-08
(2) Rδ
0 α(t)tsdt <∞for eachs > σandδ∈(0, δ0);
(3) Rδ
0 α(t)tsdt =∞for eachs < σandδ∈(0, δ0), and Z δ
0
α(t)tσdt≤C2δσ Z 2δ
δ
α(t)dt.
In [1], the classesΛm(β, p, r,)are defined in the following way Definition 1.2. f ∈Λm(β, p, r,), if
f
(m) β,p,r ≡
(Z 1
0
Z 2π 0
|∆mf(x, t)|p tβp dx
pr dt
t )1r
<+∞,
where1< p <+∞,1≤r <+∞, β >0, m∈Nand
∆mf(x, t) =
m
X
i=1
(−1)iCmi f[x+ (m−2i)t].
Now we define classesΛm(p, r, α)as follows:
Definition 1.3. We sayf ∈Λm(p, r, α), if f
(m) p,r,α ≡
(Z 1
0
α(t) Z 2π
0
|∆mf(x, t)|pdx pr
dt )1r
<+∞,
whereα(t)is function of the typeσ.
Forα(t) =t−rβ−1, β > 0, we get the classesΛm(β, p, r), considered in [1]. Therefore classes Λm(p, r, α)are generalizations of classesΛm(β, p, r).
We need some auxiliary statements.
Lemma 1.1 ([2]). Letaν,bνandβnbe numbers such thataν ≥0,bν ≥0andP∞
ν=naν =anβn: (1) For0< p≤1the following inequality is valid
∞
X
ν=1
aν
ν
X
µ=1
bµ
!p
≥pp
∞
X
ν=1
aν(bνβν)p; (2) For1≤p < ∞we have
∞
X
ν=1
aν
ν
X
µ=1
bµ
!p
≤pp
∞
X
ν=1
aν(bνβν)p.
Lemma 1.2 ([2]). Letaν,bνandγnbe numbers such thataν ≥0,bν ≥0andPn
ν=1aν =bnγn: (1) For0< p≤1, we have
∞
X
ν=1
aν
∞
X
µ=ν
bµ
!p
≥pp
∞
X
ν=1
aν(bνγν)p; (2) For1≤p < ∞, we have
∞
X
ν=1
aν
∞
X
µ=ν
bµ
!p
≤pp
∞
X
ν=1
aν(bνγν)p.
Lemma 1.3 ([3]). Letµ,τ andaν be numbers such that0< µ < τ <∞andaν ≥0. Then
∞
X
ν=1
aτν
!1τ
≤
∞
X
ν=1
aµν
!µ1 .
We denote by C a constant that depends only on m, p, r and may be different in different relations.
Theorem 1.4 ([1]). Iff ∈W Ap,1< p <+∞, then
ω(m)p (h;f) p ≤Chmp X
n≤[h1]
n(m+1)p−2
∞
X
k=n
|∆ak|
!p
+C X
n>[h1] np−2
∞
X
k=n
|∆ak|
!p
,
whereω(m)p (h;f)is the integral modulus of smoothness of orderm.
2. MAINRESULTS
Let us denote
A(n) :=
Z 1/n 1/(n+1)
α(t)dt,
b(n) := b1(n) +b2(n) =nmr Z 1/n
0
α(t)tmrdt+ Z 1
1/(n+1)
α(t)dt.
We have the following first main result.
Theorem 2.1. Letmbe any natural number and
f ∈AWp, 1< p <+∞, 1≤r <+∞.
If for the coefficients of series(1.1)or(1.2)we haveP∞
k=1|∆ak|<+∞, then:
(1) Forp≤rwe have f
(m) p,r,α ≤C
( ∞ X
n=1
nr 1−2p ∞
X
k=n
∆ak
!r
b(n)
b(n) A(n)
rp−1)1r
;
(2) Forp > rwe have f
(m) p,r,α ≤C
( ∞ X
n=1
nr 1−2p ∞
X
k=n
∆ak
!r
b(n) )1r
.
Proof. Using the characteristics of the integral modulus of smoothness we have n
f
(m) p,r,α
or
= Z 1
0
α(t) Z 2π
0
|∆mf(x, t)|pdx rp
dt
≤
∞
X
N=1
Z 1/N 1/(N+1)
α(t)
ω(m)p (f;t)r
dt
≤
∞
X
N=1
ω(m)p (f; 1/N)rZ 1/N 1/(N+1)
α(t)dt
=
∞
X
N=1
A(N)
ωp(m)(f; 1/N)r .
According to the Theorem 1.4, we have n
f
(m) p,r,α
or
≤C
∞
X
N=1
A(N)N−mr ( N
X
n=1
n(m+1)p−2
∞
X
k=n
|∆ak|
!p)rp
+C
∞
X
N=1
A(N) ( ∞
X
n=N+1
np−2
∞
X
k=n
|∆ak|
!p)rp
=I1+I2.
Now we estimateI1 andI2. Letr/p≥1. Then according to Lemma 1.1 we have I1 ≤C
∞
X
N=1
A(N)N−mr (
N(m+1)p−2
∞
X
k=N
|∆ak|
!p
βN )rp
.
Now we estimate the quantityβN: A(N)N−mrβN =
∞
X
i=N
A(i)i−mr
=
∞
X
i=N
i+ 1 i
mr
· 1 (i+ 1)mr
Z 1/i 1/(i+1)
α(t)dt
≤2mr Z 1/N
0
α(t)tmrdt, or
βN ≤Cb1(N) A(N). Consequently
(2.1) I1 ≤C
∞
X
N=1
A(N)Nr(1−2p)
∞
X
k=N
|∆ak|
!r
b1(N) A(N)
rp .
According to Lemma 1.2, forr/p≥1we have I2 ≤C
∞
X
N=1
A(N) (
Np−2
∞
X
k=N
|∆ak|
!p
γN
)rp .
We estimate the quantityγN: A(N)γN =
N
X
i=1
A(i) = Z 1
1/(N+1)
α(t)dt⇒γN = b2(N) A(N).
Consequently
(2.2) I2 ≤C
∞
X
N=1
A(N)Nr(1−2p)
∞
X
k=N
|∆ak|
!r
b2(N) A(N)
rp .
By (2.1) and (2.2) we get n
f
(m) p,r,α
or
≤C
∞
X
N=1
A(N)Nr(1−2p)
∞
X
k=N
|∆ak|
!r(
b1(N) A(N)
rp +
b2(N) A(N)
rp) .
Finally, according to Lemma 1.3, forr ≥pwe have f
(m) p,r,α ≤C
( ∞ X
N=1
A(N)Nr(1−2p)
∞
X
k=N
|∆ak|
!r
b(N) A(N)
rp)1r .
Now letr/p <1. Then, according to Lemma 1.3, we have I1 ≤C
∞
X
N=1
A(N)N−mr
N
X
n=1
n(m+1)r−2rp
∞
X
k=n
|∆ak|
!r
.
If we change the order of summation we get I1 ≤C
∞
X
n=1
n(m+1)r−2rp
∞
X
k=n
|∆ak|
!r ∞
X
N=n
A(N)N−mr
≤C
∞
X
n=1
nr(1−2p)
∞
X
k=n
|∆ak|
!r
b1(n).
(2.3)
Now we estimateI2. Using Lemma 1.3 and changing the order of summation we have:
I2 ≤C
∞
X
N=1
A(N)
∞
X
n=N
nr(1−p2)
∞
X
k=n
|∆ak|
!r
=C
∞
X
n=1
nr(1−2p)
∞
X
k=n
|∆ak|
!r n
X
N=1
A(N)
=C
∞
X
n=1
nr(1−2p)
∞
X
k=n
|∆ak|
!r
b2(n).
(2.4)
From (2.3) and (2.4) we deduce f
(m) p,r,α ≤C
( ∞ X
n=1
nr 1−2p ∞
X
k=n
∆ak
!r
b(n) )1r
, which fully demonstrates Theorem 2.1.
Theorem 2.2. Letmbe any natural number and
1< p ≤2, 1≤r <+∞, 1/p+ 1/q= 1.
Ifanare the coefficients of series(1.1)or(1.2), then:
(1) Forr≤qwe have f
(m) p,r,α≥C
( ∞ X
n=1
n−mr|an|rb1(n)
b1(n) A(n)
rq−1)1r
;
(2) Forr > qwe have f
(m) p,r,α≥C
( ∞ X
n=1
n−mr|an|rb1(n) )1r
.
Proof. Letf be an even function. Iff is an odd function then the proof of the theorem is anal- ogous to the even case. It is not difficult to see that the Fourier series of∆mf(x, t)is
∆mf(x, t)∼
(−1)m22m
∞
P
n=1
ancosnxsinmnt, formeven
(−1)m−12 −12m
∞
P
n=1
ansinnxsinmnt, formodd.
According to the well-known Hausdorf-Young’s theorem we find C
Z 2π 0
|∆mf(x, t)|pdx rp
≥
∞
X
n=1
|an|q|sinnt|mq
!rq , and then
n f
(m) p,r,α
or
≥C
∞
X
ν=1
Z 1/ν 1/(ν+1)
α(t)
ν
X
n=1
|an|q|sinnt|mq
!rq dt.
Using the well-known inequalitysinB ≥ π2Bfor0≤B ≤ π2, we get n
f
(m) p,r,α
or
≥C
∞
X
ν=1
A(ν)
ν
X
n=1
n−mq|an|q
!rq .
Letr ≤q, then according to Lemma 1.1 we have n
f
(m) p,r,α
or
≥C
∞
X
ν=1
A(ν)
ν−mq|aν|qβνrq .
It is easy to prove thatβν ≥ bA(ν)1(ν), from which we get
(2.5) n
f
(m) p,r,α
or
≥C
∞
X
ν=1
ν−mr|aν|rb1(ν)
b1(ν) A(ν)
rq−1
.
Let q < r, then according to Lemma 1.3 and with the change of the order of summation we have
n f
(m) p,r,α
or
≥C
∞
X
ν=1
A(ν)
ν
X
n=1
n−mr|an|r
=C
∞
X
n=1
n−mr|an|r
∞
X
ν=n
A(ν)
≥C
∞
X
n=1
n−mr|an|rb1(n).
(2.6)
Relations (2.5) and (2.6) prove Theorem 2.2.
We can deduce three corollaries from Theorem 2.1 and Theorem 2.2.
Corollary 2.3. Under the conditions of Theorem 2.1 and withb(n)≤CA(n), we have f
(m) p,r,α ≤C
( ∞ X
n=1
nr 1−2p ∞
X
k=n
∆ak
!r
b(n) )1r
.
Corollary 2.4. Under the conditions of Theorem 2.2 and withb1(n)≤CA(n), we have f
(m) p,r,α≥C
( ∞ X
n=1
n−mr|an|rb1(n) )1r
. As a special case, forα(t) =t−βr−1, it is easy to prove the estimates:
A(n)≤Cnβr−1 and b(n)≤Cnβr.
From Theorem 2.1 and the last estimates we can deduce the following result proved in [1].
Corollary 2.5 ([1]). Letmbe any natural number and
0< β≤m, 1< p <+∞, 1≤r <+∞, 1/p+ 1/q= 1.
If the coefficients of series (1.1) or (1.2) satisfyP∞
k=1|∆ak|<+∞, then f
(m) β,p,r ≤C
( ∞ X
n=1
nr(β+1q)−1
∞
X
k=n
∆ak
!r)1r .
REFERENCES
[1] T.Sh. TEVZADZE, Some classes of functions and trigonometric Fourier series, Some Questions of Function Theory, v. II, 31–92, Tbilisi University Press, 1981 (in Russian).
[2] M.K. POTAPOVANDM. BERISHA, Moduli of smoothnes and Fourier coefficients of functions of one variable, Publ. Inst. Math. (Beograd) (N.S.), 26(40) (1979), 215–228 (in Russian).
[3] B. HARDY, E. LITLEWOODAND G. POLYA, Inequalities, GIIL Moscow, 1948, 1–456 (in Rus- sian).
[4] M.K. POTAPOV, A certain imbedding theorem, Mathematica (Cluj), 14(37) (1972), 123–146.