40
DISSERTATIONES MATHEMATICAE UNIVERSITATIS TARTUENSIS
40
NATALIA SAEALLE
TARTU UNIVERSITY
PRESS
Faculty of Mathematics and Computer Science, University of Tartu, Tartu,
Estonia
Supervisors:
PhD Heino Türnpu,
University of Tartu,
Tartu, Estonia
Opponents:
PhD Prof. Igor Brui,
Baranovichi University,
Baranovichi, Belarus
ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
SUMMARY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
PUBLICATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
5
CHAPTER II. Summability of orthogonal series with
speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
CURRICULUM VITAE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6
LIST OF PAPERS
CONTAINED IN THE THESIS
This thesis comprises the papers I, II, III, and IV. They will be presented,
respectively, as Chapters I, II, III, and IV of the thesis.
7
ACKNOWLEDGEMENTS
8
INTRODUCTION
(σn (t)). It is said that series (1) is A-summable if (σn ) is convergent. The
interest in summability methods is that they provide a way to understand
series, which are not convergent. An A-means sequence may have better
convergence properties than the sequence of partial sums. For example,
by well-known Fejer’s theorem, the arithmetic means of the Fourier’ series
of every continuous 2π-periodic function converge uniformly; but, on the
other hand, there exist continuous 2π-periodic functions whose Fourier series
diverge at infinitely many points.
Often it is important to estimate the speed of a convergence process.
In 1969, Kangro laid down foundations of the theory of λ-convergence (i.e.,
convergence with speed) of sequences in a context of topological sequence
spaces. This theory allows to apply methods of functional analysis in in-
vestigation of convergence speed. Let λ be a speed, i.e. a monotonically
increasing sequence of positive numbers. It is said that series (1) is λ-
convergent if the sequence (λn ∞
P
k=n+1 ξk ϕk (t))n converges. If the A-means
of series converge with a speed λ, then it is said that series is Aλ -summable.
In the thesis, maximal λ-convergence and P λ -summability almost every-
where (a.e.) of series (1) are investigated, where P is a Riesz method of
weighted means. For p ≥ 1, series (1) is said to be p-maximal convergent
9
a.e. on [a, b] if it is convergent a.e. and
n p
Z b X
sup ξk ϕk (t) dt < ∞.
a n
k=0
10
3) Let {ϕk } be a product system and {wk } the Walsh system. For
functions u from the classes C[0,1] and Lp[0,1] it is showed that some problems
concerning the λ-convergence or λ-summability of the series ∞
P
k=0 < u, wk >
ϕk can be reduced to the corresponding problems for the Walsh-Fourier
series ∞
P
k=0 < u, wk > wk .
In Chapter I we consider a specific problem. Applying a general com-
plicated theorem due to Türnpu in the case of Riesz method, we find when
the boundedness of according Lebesgue functions implies the maximal P λ -
summability a.e. of series (1) for all sequences (ξk ) from the Banach space `2λ .
In Chapter II, the same problem in the case of any regular λ2 -conservative
summability method A is investigated. For this purpose, we compare A
with a (suitably constructed) Riesz method P (A) using our results about
λ-inclusion of summability methods in the class of series (1).
In Chapters III and IV we consider series (1), where {ϕk } is any pro-
duct system (not necessarily orthogonal). Chapter III examines p-maximal
λ-convergence and p-maximal λ-boundedness of series (1), where a system
{ϕk } satisfies some
additional conditions. For example, it is proved that if
P∞ R b
k=0 a ϕk (t)dt < ∞ (i.e. the system {fk } is weakly multiplicative), then
series (1) is 2-maximally λ-convergent a.e. on [a, b] for every (ξk ) ∈ `2λ . Our
main attention is concentrated to the series
∞
X
< u, wk > ϕk (t), (2)
k=0
where < u, wk >:= 01 u(t)wk (t)dt are the Walsh-Fourier coefficients of any
R
integrable function u. In the case of u ∈ Lp[0,1] (1 < p < ∞), the problem of 1-
maximal λ-boundedness of this series might be reduced to examination of p-
maximal λ-boundedness of the Walsh-Fourier series ∞
P
k=0 < u, wk > wk (t).
Chapter IV investigates uniform convergence of series (2), uniform A-
summability, uniform Aλ -boundedness and uniform regular Aλ -summability,
where u is a continuous function on [0, 1]. It is shown that each of the above
mentioned properties is equivalent to the same property of the corresponding
Walsh-Fourier series.
11
SUMMARY
where (αnk ) is a triangular matrix (i.e. αnk = 0 for k > n). The correspond-
ing sequence-to-sequence method A = (ank ) is defined by the transformation
n
ηn0
X
:= ank ζk (n ∈ N)
k=0
with
n
X
ank := αnk − αn,k+1 , or, equivalently, αnk = anν (n, k ∈ N).
ν=k
A series ∞
P
k=0 uk (a sequence (ζk )) is called A-summable if the limit limn ηn
(limn ηn0 ) exists.
A series-to-sequence method A is said to be regular if
∞
X
lim ηn = uk
n
k=0
P∞
for every convergent series k=0 uk . It is well known (see, for example, [3,
Theorem 1.3]) that A is regular if and only if
∞
X
lim αnk = 1 (k ∈ N) and |αnk − αn,k+1 | = O(1).
n
k=0
In the thesis, the main attention has been paid to two certain summa-
bility methods: the Cesàro method and the Riesz method.
1
We will denote the set of non-negative integers by N.
12
1) The sequence-to-sequence Cesàro method (or the method of arithmetic
means) (C, 1) = (ank ) is given by the matrix A with
1
if k ≤ n,
ank = n + 1
0 if k > n,
k
1− if k ≤ n,
αnk = n+1
0 if k > n.
pk
if k ≤ n,
ank = Pn
0 if k > n,
where n
X
Pn := pk % ∞
k=0
where P−1 := 0.
Note that P is a regular method (cf. [3, Theorem 17.1]) and (C, 1) is a
special case of it.
13
If a sequence z is summable by a sequence-to-sequence summability
method A, then it is called Aλ -summable, provided that the limit
n
!
0
X
lim λn αnk ζk − lim ηm
n m
k=0
{1, cos t, sin t, cos 2t, sin 2t, cos 3t, sin 3t, . . .}
14
(b) Let r0 be the function defined on [0, 1) by
(
1 if t ∈ [0, 12 ),
r0 (t) :=
−1 if t ∈ [ 12 , 1)
w0 (t) := 1 and wn (t) := rn0 +1 (t)rn1 +1 (t) . . . rnk +1 (t) (t ∈ [0, 1]), (3)
where n = 2n0 + 2n1 + . . . + 2nk (n0 < n1 < . . . < nk ) is the dyadic
representation of n.
(d) The Haar system h = {hn (t)} is defined as follows. Set h0 := 1. For
n, k ∈ N with 0 ≤ k < 2n define hn on [0,1] by
n/2
2
if t ∈ [2−(n+1) 2k; 2−(n+1) (2k + 1)),
n/2
h2n +k (t) :=
−2 if t ∈ [2−(n+1) (2k + 1); 2−(n+1) (2k + 2)),
0 otherwise.
A classical result of Kaczmarz [9] states that series (1) converges a.e. on
[a, b] for each (ξk ) ∈ `2 if Ln (ϕ, t) = O(1) on [a, b]. Türnpu [23] showed that
the last condition can be replaced by
By Kaczmarz [9], [10], series (1) is (C, 1)-summable a.e. on [a, b] for each
(ξk ) ∈ `2 if Ln (A, ϕ, t) = O(1) on [a, b]. On the other hand, Móricz and Tan-
dori [13] showed that there exist a regular triangular summability method A
15
and an orthogonal system ϕ such that series (1) is not A-summable a.e. on
[a, b] for some (ξk ) ∈ `2 . Móricz [12] and Türnpu [23] found certain classes
of regular methods A for which the conditions Ln (A, ϕ, t) = O(1) on [a, b]
and Ln (A, ϕ, t) = Ot (1) (t ∈ [a, b]), respectively, imply the A-summability
a.e. on [a, b] of series (1) for each (ξk ) ∈ `2 .
Series (1) is Aλ -summable a.e. on [a, b] for all x ∈ `2λ if and only if the
following conditions hold:
1◦ series (1) is A-summable a.e. on [a, b] for every x ∈ `2λ ;
2◦ for each ε > 0 there exist a measurable subset Tε ⊂ [a, b] satisfying
mes Tε > b − a − ε and a constant Mε > 0 such that, for all measurable
decompositions
n
Nm := Nmn : n = 0, 1, . . . , m; Nmk ∩ Nmn = ∅ if k 6= n;
m
[ o
Nmn ⊂ [a, b] , (4)
n=0
16
one has
Z Z m−2 m−1 m
X X X
m
Am (ε) = χmn (t) χmp (τ ) ϕν (t)ϕν (τ )Dnpν dtdτ ≤ Mε ,
Tε Tε n=0 p=n+1 ν=0
If
pn = O(Pn−1 ), (6)
!
λ2n 1 1 1
↓ 0, − ↓ 0, (7)
Pn−1 pn λ2n λ2n+1
and Z b
sup Lk (P, ϕ, t)dt < ∞,
a k
then series (1) is maximally P λ -summable a.e. on [a, b] for each (ξk ) ∈ `2λ .
Example. Let λk = (k + 1)α , α > 0. Then, as an example of a
Riesz method satisfying (5) − (7), we may consider the method P with
pk = (k + 1)β , where β > 2α − 1.
For example, if 0 < α < 1/2, we can put β = 0, i.e. P = (R, 1) = (C, 1).
17
1
Remark. In [6], Kangro proved that if the sequence λ is a sequence
Pk∞ 1
of summability factors of type (A, Aλ ) (i.e. the series λ
k=0 λk uk is A -
summable for each A-summable series ∞
P
k=0 uk ), then the A-summability
a.e. on [a, b] of series (1), where (ξk ) ∈ ` , implies the Aλ -summability of the
2
series ∞ ξk
P
k=0 λk ϕk (t) a.e. on [a, b]. From [3, Theorem 29.3], it follows, that
1
λk is a sequence of summability factors (P, P λ ), provided that conditions
(5), (6), and (7) hold. Therefore the P -summability a.e. of series (1) for
each (ξk ) ∈ `2 implies the P λ -summability a.e. of the series ∞
P
k=0 ζk ϕk (t)
for each (ζk ) ∈ `2λ .
Note that the above argument does not imply the maximal P λ -summa-
bility a.e. of the series ∞ 2
k=0 ζk ϕk (t) for each (ζk ) ∈ `λ .
P
Let A and B be two summability methods. If for series (1) from the B-
summability a.e. follows the A-summability a.e. for every (ξk ) ∈ `2 , then we
say that A includes B in the class of series (1) (shortly B ⊂ A). Inclusion of
summability methods (in the class of orthogonal series) is well investigated.
A review of inclusion results may be found in [27].
The following result by Türnpu is the starting point in Chapter II
([17]), where we discuss the λ-inclusion Aλ ⊃ B λ in the class of series (1).
Theorem 3 (cf. [24]). Let A = (αnk ) and B = (βnk ) be regular trian-
gular summability methods. If
∞
sup |ank | (βkν − 1)2 = O(1),
X
k=ν n≥k
18
Our interest is focused to the inclusion with respect to maximal λ-
summability of summability methods. The main result of Chapter II is
the following
Theorem 5 (cf. [17]). Let A = (αnk ) and B = (βnk ) be regular λ2 -
conservative methods and let
m m
λ−2
X X
n (βkn − 1)2 sup λ2l |alk | = O(1) and (βkn − 1)2 sup |alk | = O(1).
k=n l≥k k=n l≥k
Ln (P (A), t) = O(1)
hold, then the orthogonal series (1) is maximally Aλ -summable a.e. on [a, b]
for every x ∈ `2λ .
Consider some examples of Riesz method P defined by regular summa-
bility methods A.
Example 1. Let A = (C, 2) be the Cesàro method of order 2, where
2(n − k + 1)
if k ≤ n,
ank = (n + 1)(n + 2)
0 if k > n.
19
Then
1
ak = & 0,
2k + 1
and for the Riesz method P ((C, 2)) we have
k
!
X 1 √
Pk = exp > 2k + 3
ν=0
2ν + 1
and
1
pk = Pk − Pk−1 = Pk−1 exp − 1 < e.
2k + 1
Let λn = (n + 1)α (0 < α < 1/4). In this case, by asymptotic formula
n
X 1 1
γ
∼ (n + 1)1−γ (0 < γ < 1),
k=0
(k + 1) 1 − γ
the methods (C, 2) and P ((C, 2)) are λ2 -conservative. Conditions (7) are
also fulfilled:
! !
λ2n λ2 1 1 1 √ 1 1
<√ n ↓ 0 and 2
− 2 < 2n + 1 2
− 2 ↓ 0.
Pn−1 2n + 1 pn λn λn+1 λn λn+1
As in the previous example, we get that the method P ((R∗ , 1, 2)) is λ2 -con-
servative. Conditions (7) are fulfilled if λn = (n + 1)α , (0 < α < 1/16), for
instance.
20
3 Convergence and summability with speed
of series with respect to product systems
g0 (t) := 1 and gn (t) := fn0 +1 (t)fn1 +1 (t) . . . fnk +1 (t) (t ∈ [a, b]),
where n = 2n0 + 2n1 + . . . + 2nk (n0 < n1 < . . . < nk ) is the dyadic
representation of n, is called the product system of {fk }. For example, the
Walsh system (3) is the product system of Rademacher system (2).
The system {fk } is said to be weakly multiplicative if the product system
g satisfies the condition
∞ Z b
X
gn (t)dt < ∞
a
n=0
21
In [20] it is proved that the series
∞
X
ξk fk (t)
k=0
converges a.e. on [a, b] for all rearrangements of {ξk fk } if (ξk ) ∈ `2 and {fk }
is p-weakly multiplicative for a number p with 1 < p < ∞.
The series ∞ X
ξk gk (t) (8)
k=0
The following two theorems are starting points for investigations in Chapter
III ([16]).
Theorem 7 (cf. [20]). Series (8) is 1-maximally convergent a.e. on
[a, b] if (ξk ) ∈ `2 and {gk } is the product system of a p-weakly multiplicative
system for 2 ≤ p < ∞.
Theorem 8 (cf. [19], [21, p. 292]). Series (8) is 2-maximally conver-
gent a.e. on [a, b] if (ξk ) ∈ `2 and {gk } is the product system of a weakly
multiplicative system.
22
Theorem 9 (cf. [16]).
(a) If (ξk ) ∈ `2λ and {gk } is the product system of a weakly multiplicative
system, then series (8) is 2-maximally λ-convergent a.e. on [a, b].
(b) If (ξk ) ∈ `2λ and {gk } is the product system of a 2-weakly multiplicative
system, then series (8) is 1-maximally λ-convergent a.e. on [a, b].
Beside the trigonometric and Haar systems, the Walsh system is one
of the most widely used complete orthonormal systems of functions in the
theory of functions of a real variable. It is very similar to the trigonometric
system, differing from the latter by its greater simplicity.
The Walsh-Fourier coefficients of an integrable function u are the num-
bers Z 1
< u, wn >:= u(t)wn (t)dt (n ∈ N),
0
and the Walsh-Fourier series of u is the series
∞
X
< u, wn > wn (t). (9)
n=0
Properties of this series are well studied. Mention some of them. For ex-
ample, it was shown in [26], in which Walsh introduced the system bearing
his name, that for every point τ0 ∈ [0, 1] there exists a continuous function
u, whose Walsh-Fourier series diverges at that point. On the other hand,
Walsh remarked, that
n −1
2X
lim < u, wν > wν (τ ) = u(τ ) uniformly on [0, 1]
n
ν=0
where u ∈ Lp[a,b] , and {gk } is the product system of a system {fk } of inte-
grable functions. By means of the Banach-Steinhaus theorem (in context of
sublinear operators), the following theorems are proved.
23
Theorem 10 (cf. [16]). Let 1 < p, q < ∞ be conjugate exponents
+ q = 1 and let u be a function in Lp[0,1] . If g is the product system of
1
p
1
of the series ∞
P
k=0 < u, gk > wk (t), by use of an estimate due to Balashov
and Rubinstein [2] we obtain that in context of Theorem 11
Z b
< u, gk >:= u(t)gk (t)dt =< h, wk > (k ∈ N),
a
24
Proofs of these results are essentially based on orhogonality of the Walsh
system and classical inequalities.
25
FUNKTSIONAALRIDADE KIIRUSEGA
KOONDUVUS JA SUMMEERUVUS
KOKKUVÕTE
Töö koosneb viiest osast: kokkuvõte ja peatükid I, II, III ja IV, mis
kujutavad endast teaduslikke artikleid (artiklite loetelu vt. lk. 7).
26
Töö I peatükis lahendatakse konkreetne ülesanne: lähtudes Türnpu
poolt tõestatud üldisest ja keerulisest teoreemist (vt. [25]), leitakse need
seosed Rieszi kaalutud keskmiste menetluse P ja kiiruse λ vahel, mille puhul
vastavate Lebesgue’i funktsioonide tõkestatusest järeldub rea (1) maksi-
maalne P λ -summeeruvus peaaegu kõikjal (p.k.) iga jada (ξk ) korral Banachi
ruumist `2λ .
Töö II peatükis lahendatakse sama ülesanne teatavate regulaarsete λ2 -
konservatiivsete menetluste A = (ank ) jaoks, võrreldes neid Rieszi menetlu-
sega P (A), mis on konstrueeritud järgmiselt:
k
!
X
Pk = exp sup |anν | (k = 0, 1, 2, . . .).
ν=0 n≥ν
kus k = 2k0 + 2k1 + . . . 2kn (k0 < k1 . . . < kn ). III peatükk uurib ridade
(1) p-maksimaalset λ-koonduvust ja p-maksimaalset λ-tõkestatust korru-
tissüsteemi {ϕk } korral. Seejuures nimetatakse rida (1) p-maksimaalselt
λ-koonduvaks (p-maksimaalselt λ-tõkestatuks), kui ta on λ-koonduv (λ-tõ-
kestatud) p.k. lõigus [a, b] ja
p
Z b
X∞
p
sup λn ξk ϕk (t) dt < ∞,
a n k=n+1
27
(sel juhul öeldakse, et lähtesüsteem {fk } on nõrgalt multiplikatiivne), siis
rida (1) on 2-maksimaalselt λ-koonduv p.k. lõigus [a, b] iga (ξk ) ∈ `2λ korral.
Erilise tähelepanu all on read
∞
X
< u, wk > ϕk (t), (2)
k=0
28
REFERENCES
29
14. R. Paley, A remarkable system of orthogonal functions, Proc. London
Math. Soc. 34 (1932), 241-279.
30
PUBLICATIONS
CHAPTER I
Acta et Commentationes Universitatis
Tartuensis de Mathematica 5 (2001), 3-14
Riesz summability with speed
of orthogonal series
1. Main result
35
The series ξk ϕk (t) is said to be Aλ -summable almost everywhere (a.e.)
P
and n
X
αnk = anν .
ν=k
The series ξk ϕk (t) is Aλ -summable a.e. on [a, b] for all x ∈ `2λ if and only
P
36
one has
Z Z m−2 m−1 m
X X X
m
Am (ε) =
χmn (t) χmp (τ ) ϕν (t)ϕν (τ )Dnpν dtdτ ≤ Mε ,
Tε Tε n=0 p=n+1 ν=0
λn λp
(αmν − αnν )(αmν − αpν ) 2 , if 0 ≤ v ≤ n < p < m,
λν
λ λ
m n p
Dnpν = α
mν (α
mν − αpν ) , if n < ν ≤ p < m,
λ2ν
2 λn λp
αmν , if n < p < ν ≤ m.
2λν
In the present paper we will mainly consider the case, when A is the
Riesz summability method P , i.e.
pk , k ≤ n,
ank = Pn
0, k > n,
then
λ2n λ2
= O(1) k for k ≤ n , k, n ∈ N. (6)
Pn Pk
Since by the Cauchy inequality
n n n
!1/2 n
!1/2
λn X pk λ2n X pk X λ2n X pk
≤ pk = ,
Pn k=0 λk Pn2 k=0 λ2k k=0 Pn k=0 λ2k
37
we have that if P is λ2 -convergence preserving, then P is also λ-convergence
preserving.
The main objective of this paper is to prove the following theorem.
Theorem 2. Let condition (5) hold, and let
λ2n
↓ 0 , pn = O(Pn−1 ), (7)
Pn−1
1 1
∆ 2 ↓ 0, (8)
pn λ n
where
1 1 1
∆ 2
= 2 − 2 .
λn λn λn+1
If
Z b
sup Lk (P, t)dt < ∞, (9)
a k
where
k
Z b X Pν−1
Lk (P, t) = 1− ϕν (t)ϕν (τ ) dτ,
a
ν=0
Pk
with P−1 = 0, are the Lebesgue functions of the method P , then the series
ξk ϕk (t) is maximally P λ -summable a.e. on [a, b] for every x ∈ `2λ .
P
gument does not imply the maximal P λ -summability of the series ξk ϕk (t)
P
38
2. Main Lemma
Proof. Denote
j
X
Rj (t, τ ) = αjν ϕν (t)ϕν (τ ),
ν=0
where
Pν−1
αjν = 1 − .
Pj
Then ν
X
ϕν (t)ϕν (τ ) = ηνk Rk (t, τ ),
k=0
and therefore
Z Z m−2X m−1
X Xm X ν
m
Am (ε) = χmn (t) χmp (τ ) ηνk Rk (t, τ )Dnpν dtdτ
Tε Tε n=0 p=n+1 ν=0 k=0
Z Z m−2 m−1 m m
∆Dnpν
X X X
= χmn (t) χmp (τ ) Rk (t, τ )Pk ∆ dtdτ
Tε Tε n=0 p=n+1 k=0
pk
Z Z m−2 m−1
" m
X X X
= χmn (t) χmp (τ )λn λp Rk (t, τ ) ∆1k (n, p, m)
Tε Tε
n=0 p=n+1 k=0
#
+∆2k (p, m) + ∆3k (p, m) dtdτ ,
39
where
h i
1
∆ (αmk − αnk )(αmk − α )
pk λ2
k
Pk ∆
if 0 ≤ k < n,
p
k
∆1k (n, p, m) = Pn αnn (αmn − αpn )
if k = n,
λ2n pn
0 if k > n,
h i
1
∆ αmk (αmk − αpk ) λk2
if n ≤ k < p,
Pk ∆
p
k
∆2k (p, m) = Pp αmp αpp
if k = p,
λ2p pp
0 if k < n, k > p,
and
h i
2 1
∆ αmk λk2
Pk ∆ if p ≤ k < m−1,
p
k
2
αm,m−1 2
αmm 2
αmm
∆3k (p, m) = Pm−1 2 − − if k = m − 1,
λm−1 pm−1 λ2m pm−1 λ2m pm
α 2
Pm 2mm
if k = m,
λ m pm
0 if k < p.
Observe, that
2
Pk−1
∆
1
Pk λ2k
∆k (n, p, m) ≤ ∆ for0 < k < n,
Pn Pp pk
by (4)
1 1
∆1n (n, p, m) = O(1) = O(1) ,
λ2n Pp λ n λ p Pn
and
1
∆2p (p, m) = O(1) .
λ2p
Denote
Pn 2 1
Mn = λ ∆ .
pn n λ2n
From (7) it follows that Mn ≤ 1 for all n ∈ N. Therefore
2
Pm−1 αm,m−1 pm 1 1
∆3m−1 (p, m) = ∆ 2 − 2 2 = O(1)Mm−1 2 + O(1) 2 ,
pm−1 λm−1 λ m Pm λm−1 λm
40
and
1
∆3m (p, m) = O(1) .
λ2m
Thus
P2
m−2 n−1 ∆ λk−1
λ2n
Z Z X X 2
k
Am (ε) ≤ χmn (t) Pk ∆ |Rk (t, τ )| dtdτ
Tε Tε n=0 Pn Pn−1
k=0
p k
Z Z m−1
"p−1
X X
+ χmp (τ )λ2p ∆2k (p, m) |Rk (t, τ )|
Tε Tε p=1 k=0
m−2
#
X 3
+ ∆k (p, m) |Rk (t, τ )| dtdτ
k=p
m−2 m−1
Z Z "
X X
+ O(1) χmn (t) χmp (τ ) |Rn (t, τ )| + |Rp (t, τ )|
Tε Tε n=0 p=n+1
#
+ |Rm−1(t, τ )| + |Rm (t, τ )| dtdτ.
Now we have
P2
n−1 ∆ λk−1
λ2n
Z X 2
k
Am (ε) ≤ sup Lk (P, t)dt sup Pk ∆
Tε k<m n<m Pn Pn−1
k=0
p k
Z p−1
X
sup Lk (P, τ )dτ sup λ2p
2
+ ∆k (p, m)
Tε k<m p<m
k=0
Z m−2
X
+ sup Lk (P, τ )dτ sup λ2p |∆3k (p, m)|
Tε k<m p
k=p
Z
+O(1) sup Lk (P, t)dt.
Tε k<m
41
By [4] (see p. 220) we have that, for any sequence (an ) ⊂ R,
∆ λak2
!
k
1 1 ∆ak
Pk ∆ = ∆ 2 +∆ 2 Pk (11)
pk λk λk+1 pk
!
1 ∆ak 1 1
+ 2 Pk ∆ + Pk ak+1 ∆ ∆ .
λk+2 pk pk λ2k
Consider the case when i = 1; then, by (11), we have
λ2n n−1
!
2
∆Pk−1
1
X 1 1
Vnpm ≤ ∆ 2 +∆ 2
Pn k=0 λk λk+1 pk
2
∆Pk−1
1 1 1
+ 2 ∆ + Pk2 ∆ ∆ 2 .
λk+2 pk pk λ k
Since by (7)
n−1
!
X 2 1 1
Pk ∆ ∆
pk λ2k
k=0
n−1
1 1 1 1 P2 1
∆ 2 − n−1 ∆ 2 ,
X
= P02 ∆ 2− ∆Pk2
p0 λ0 k=0 pk+1 λk+1 pn λn
we have
1
Vnpm
λ2 n−1
! !
1 1 1
≤ n
X
∆ 2 +∆ 2 (Pk−1 + Pk ) + (pk + pk+1 )
Pn k=0 λk λk+1 λ2k+2
λ2 n−1
λ2 X 1 1 λ2 P 2 1
+ n O(1) + n ∆Pk2 ∆ 2 + n2 n−1 ∆ 2 .
Pn Pn k=0 pk+1 λk+1 Pn pn λn
Now, by (5) and (6), we have
λ2 n−1
!
X pk pk+1 1
1
Vnpm ≤2 n 2 Mk + 2 Mk+1 + pk+1 ∆ 2
Pn k=0 λk λk+1 λk+1
λ2 n−1
!
X pk pk+1
+ n 2 + 2 + O(1)
Pn k=0 λk λk+1
λ2n n−1
X pk+1 λ2n Pn
+2 M k+1 + Mn + O(1)
Pn k=0 λ2k+1 Pn2 λ2n
n
λ2n X pk λ2n n−1
X pk+1
= O(1) + O(1) + O(1)
Pn k=0 λ2k Pn k=0 λ2k+1
= O(1).
42
Analogously we have
αmk Pk−1
λ2p p−1
X ∆ λ2
2 k
Vnpm ≤ Pk ∆
Pp k=0 pk
λ2p p−1
!
X 1 1 ∆(αmk Pk−1 ) 1 ∆(αmk Pk−1 )
≤ ∆ 2 +∆ 2 Pk + 2 Pk ∆
Pp k=0 λk λk+1 pk λk+2 pk
λ2p p−1
!
X 1 1
+ αmk Pk2 ∆ ∆ 2
Pp k=0 pk λ k
λ2p p−1
" ! #
1 1 Pk Pk Pk−1
X
= O(1) ∆ 2 +∆ 2 Pk + 2 ∆ − 1−
Pp k=0 λk λk+1 λk+2 Pm P m
3. Proof of Theorem 2
43
where Nm ranges over all decompositions defined by (3).
Proof of Theorem 2. By [6] (see p. 201) the condition
Denote
ᾱpk = αpk − αp−1,k ,
then
p
∞ X
X
βn (A, x, t) = λn ᾱpk ξk ϕk (t) = Bmn (x, t) + Cmn (x, t),
p=n+1 k=0
where p
m
X X
Bmn (x, t) = λn ᾱpk ξk ϕk (t),
p=n+1 k=0
∞
X p
X
Cmn (x, t) = λn ᾱpk ξk ϕk (t).
p=m+1 k=0
Therefore
Z m
X Z m
X
χmn (t)βn (A, x, t)dt = O(1) χmn (t)Bmn (x, t)dt
Tε n=0 Tε n=0
Z X m
+O(1) χmn (t)Cmn (x, t)dt.
Tε n=0
44
By orthogonality of ϕ we have
m
Z
X
χmn (t)Cmn (x, t)dt
Tε
n=0
m m
Z bX
X
≤ χmn (t)λn αmk ξk ϕk (t) − fx (t) dt
a n=0
k=0
1/2
m
!1/2 ∞
√ λm |αmk − 1| X 2 2 X
≤ b − a sup ξk λk + ξk2 λ2k .
k≤m λk k=0 k=m+1
and therefore Z m
X
χmn (t)Cmn (x, t)dt = O(kxkl2 ).
λ
Tε n=0
Denoting
p−1
X
Am
p (t) = χmn (t)λn ,
n=0
we have
Z m
X m
X Z m
X
χmn (t)Bmn (x, t)dt = ξk ᾱpk ϕk (t)Am
p (t)dt
Tε n=0 k=0 Tε p=k
Z Xm
+ χmn (t)λn (αm0 − αn0 )ξ0 ϕ0 (t)dt.
Tε n=0
Now by (13)
Z m
X
χmn (t)Bmn (x, t)dt
Tε n=0
X m Z m
X
= ξk ᾱpk ϕk (t)Am
p (t)dt + O(kxk`2 ). λ
k=0 Tε p=k
m m m
!1/2
ϕk (t)ϕk (τ ) X
Z Z X X
= O(1) ᾱpk Am
p (t) ᾱνk Am
ν (τ )dtdτ kxk`2.
Tε Tε k=0 λ2k p=k ν=k
λ
45
Finally
m m m
ϕk (t)ϕk (τ ) X
Z Z X X
ᾱpk Am
p (t) ᾱνk Am
ν (τ )dtdτ = Am (ε) + Em ,
Tε Tε k=0 λ2k p=k ν=k
where
m m 2
λn
Z Z X X
Em = χmn (t)χmn (τ ) ϕk (t)ϕk (τ ) (αmk − αnk ) dtdτ.
Tε Tε n=0 k=0
λk
By (13) and Bessel’s inequality we get
m−1 m 2 Z 2
X X λn
Em = (αmk − αnk ) ϕk (t)χmn (t)dt
n=0 k=0
λk Tε
X ∞ Z
m X 2
= O(1) ϕν (t)χmn (t)dt
n=0 ν=0 Tε
Xm Z
= O(1) χ2mn (t)dt
n=0 Tε
Z X m
= O(1) χmn (t)dt
Tε n=0
= O(1).
References
46
CHAPTER II
Analysis Mathematica (Szeged) 31 (2005), 63-73
Summability of orthogonal series
with speed
1. Introduction
The set of all λ-convergent sequences is denoted by cλ . The series (1) is said
to be convergent with speed λ or λ-convergent almost everywhere (a.e.) on
[a, b] if the limits
n
X
lim ξk ϕk (t) =: fx (t)
n
k=0
and
n
!
X
lim λn ξk ϕk (t) − fx (t)
n
k=0
49
Throughout this paper, we assume that A = (αnk ) is a triangular sum-
mability method and denote
50
exists a.e. on [a, b]), and the limit
lim βn (A, x, t),
n
51
implies that series (1) is A-summable a.e. on [a, b] for every x ∈ `2 .
For example, for the case of the Riesz method P from the boundedness of
Lebesgue functions a.e. on [a, b] it follows that the series (1) is P -summable
a.e. on [a, b] for every x ∈ `2 .
On the other hand, necessary and sufficient conditions for A-summability
of series (1) a.e. on [a, b] for all x ∈ `2 are founded in [12] as follows. It is
proved that from the conditions
lim αnk = 1
n
and
n
Z b X
sup αnk αpk ϕk (t)ϕk (τ ) dτ = Ot (1)
a p≥n k=0
a.e. on [a, b] it follows that the series (1) is A-summable a.e. on [a, b] for all
x ∈ `2 .
Necessary and sufficient conditions for Aλ -summability of series (1) a.e.
on [a, b] for all x ∈ `2λ
are found in [14]. We proved there that if A is λ2 -convergence preserving,
lim αnk = 1 and the series (1) is A-summable a.e. on [a, b] for every x ∈ `2λ ,
n
then the condition
m−1
Z b
X m
sup ϕν (t)ϕν (τ )Dnpν dτ = Ot (1)
a p≥n ν=0
implies that series (1) is Aλ -summable a.e. on [a, b] for every x ∈ `2λ .
Since the form of the above condition is very complicated, in [9] we
considered the case A = P , the Riesz summability method.
Theorem A (see [9]). Let
n
λ2n X pk
= O(1), (2)
Pn k=0 λ2k
λ2n
& 0 , pn = O(Pn−1 ), (3)
Pn−1
and !
1 1 1
− & 0. (4)
pn λ2n λ2n+1
52
If
Z b
sup Lϕ
k (P, t)dt < ∞, (5)
a k
then the orthogonal series (1) is maximally P λ -summable a.e. on [a, b] for
every x ∈ `2λ .
Ln (P (A), t) = O(1)
hold, then the orthogonal series (1) is maximally Aλ -summable a.e. on [a, b]
for every x ∈ `2λ .
In the proof of Theorem B, we will use the following
Theorem C. Let A = (αnk ) and B = (βnk ) be regular λ2 -convergence
preserving methods, and let
m m
λ−2
X X
n (βkn −1)2 sup λ2l |alk | = O(1) and (βkn −1)2 sup |alk | = O(1). (7)
k=n l≥k k=n l≥k
53
3. Proofs of Theorems B and C
if and only if for each ε > 0 there exists a measurable subset Tε ⊂ [a, b] such
that mes Tε > b − a − ε and
Z
|f (t)|dt < ∞.
Tε
if and only if for each ε > 0, there exist a measurable subset Tε ⊂ [a, b] with
mes Tε > b − a − ε and a constant Mε > 0 such that for all measurable
decompositions
n
Nm := Nmn : n = 0, 1, . . . , m; Nmk ∩ Nmn = ∅ if k 6= n;
m
[ o
Nmn ⊂ [a, b] , (8)
n=0
one has
m
Z
X
ε
Bm := χmn (t)fn (t)dt ≤ Mε , where χmn := χNmn . (9)
Tε
n=0
∞
!1/2
X
k x k `2 = ξk2 λ2k
λ
k=0
54
is a Banach space and the set {ei = (δki )∞ k=0 : i ∈ N}, where δki is the
Kronecker symbol , forms a total set in `2λ (that is, the linear combinations
of ei are everywhere dense in `2λ ), we can use the Banach theorem.
Lemma 3 (see [3], p. 361). Let (Dn : n ∈ N) be continuous linear oper-
ators from `2λ to the Frechet space M[a,b] of all functions totally measurable
on [a, b]. Suppose that the following conditions hold:
1◦ supn |Dn (x, t)| < ∞ a.e. on [a, b] for every x ∈ `2λ ;
2◦ the limit limn Dn (ei , t) exists a.e. on [a, b] for every i ∈ N.
Then the limit limn Dn (x, t) exists a.e. on [a, b] for all x ∈ `2λ .
Proof of Theorem C. Let the second equality in (7) hold. By [13, Corol-
lary], if the orthogonal series (1) is B-summable a.e. on [a, b] for x0 ∈ `2 ,
then it is A-summable a.e. on [a, b] for the same x0 . In Theorem C we
assume that series (1) is B λ -summable a.e. on [a, b] for every x ∈ `2λ ⊂ `2 ,
therefore it is B-summable a.e. on [a, b] for every x ∈ `2λ . So, by [13, Corol-
lary 1], series (1) is A-summable a.e. on [a, b] to some function fx for every
x ∈ `2λ . Furthermore, the operator p = fx (t) defined by
p : `2λ → M[a,b] , x 7→ fx
The operator Dn (x, t) from `2λ into M[a,b] is continuous and linear. We will
use Lemma 3 and show that conditions 1◦ and 2◦ are fulfilled.
By Lemma 2, for condition 1◦ it is sufficient to show that inequality (9)
with fn = Dn holds for every decompositions (8), that is, for every ε > 0
and fixed x ∈ `2λ there exists a measurable subset Tε = Tε (x) ⊂ [a, b] with
mes Tε (x) > b − a − ε and a constant Mε = Mε (x) > 0 such that for all
decomposition (8) one has
m
Z
X
ε
Bm = χmn (t)Dn (x, t)dt ≤ Mε (x).
Tε (x)
n=0
55
and by using the Cauchy-Bunyakovsky inequality, we have
m n k
Z !
X X X
ε
Bm = χmn (t)λn ank ξν ϕν (t) − fx (t) dt
Tε (x)
n=0 k=0 ν=0
m n k
Z
X X X
≤ χmn (t)λn ank βkν ξν ϕν (t) − fx (t) dt
Tε (x) n=0
k=0 ν=0
m n k
Z
X X X
+ χmn (t)λn ank (βkν − 1)ξν ϕν (t) dt
Tε (x) n=0
k=0 ν=0
m n k
Z !
X X X
≤ χmn (t)λn ank βkν ξν ϕν (t) − fx (t) dt
Tε (x) n=0
k=0 ν=0
m n
Z
X X
+ χmn (t)λn ank − 1 |fx (t)| dt
Tε (x) n=0
k=0
m
( n
)1
Z X X 2
+ χmn (t)λn |ank | ×
Tε (x) n=0 k=0
n k
!2 12
X X
× |ank | (βkν − 1)ξν ϕν (t) dt.
k=0 ν=0
Therefore, we have
m n
k
|ank |
Z X XX
ε
Bm = χmn (t)λn λk βkν ξν ϕν (t) − fx (t) dt
Tε (x) n=0 k=0
λ k
ν=0
n
Z !1/2
X b
2
+O(1) sup λn ank − 1 (fx (t)) dt
n a
k=0
!2 1
Z m n k 2
X X X
+O(1) χmn (t)λ2n |ank | (βkν − 1)ξν ϕν (t) dt .
Tε (x) n=0 k=0 ν=0
56
By (7), we get
k
Z
X
ε
Bm = O(1) sup λk βkν ξν ϕν (t) − fx (t) dt
Tε (x) k
ν=0
+O(1)kfx kL2 + O(1)||x||`2 .
λ
Since series (1) is B λ -summable a.e. on [a, b] for every x ∈ `2λ , by Lemma 1,
we have that for every ε > 0 a nd every fixed x ∈ `2λ there exist a measurable
set Eε (x) ⊂ [a, b] with mesEε (x) > b − a − ε and a constant Nε (x) > 0 such
that one has
k
Z
X
sup λk βkν ξν ϕν (t) − fx (t) dt = Nε (x).
Eε (x) k
ν=0
Thus there exist a measurable subset Tε (x) = Eε (x) and the constant Mε (x)
such that
Therefore, we have
ε
Bm ≤ Mε (x),
which means that condition 1◦ of Lemma 3 holds.
Let δki be the Kronecker symbol. Since the series ∞
P
k=0 δki is λ-conver-
gent, A is regular and λ-convergence preserving, the limit
n
X
lim λn αnk δki ϕk (t) − ϕi (t) = lim λn (αni − 1)ϕi (t)
n n
k=0
lim Dn (ei , t)
n
exists a.e. on [a, b] for every i ∈ N, which means that condition 2◦ of Lemma
3 also holds. From Lemma 3 it follows that if the series (1) is B λ -summable
a.e. on [a, b] for every x ∈ `2λ , then series (1) is also Aλ -summable a.e. on
[a, b] for every x ∈ `2λ .
Assume that the series (1) be maximally B λ -summable a.e. on [a, b] for
all x ∈ `2λ , then
n
Z b X
sup λn βnk ξk ϕk (t) − fx (t) dt = Ox (1).
a n
k=0
57
Now, by the above Remark we have
n
Z b X
max λn αnk ξk ϕk (t) − fx (t) dt
a n≤m
k=0
Z m n
bX X
≤ 2 sup χmn (t)λn αnk ξk ϕk (t) − fx (t) dt
Nm a n=0
k=0
k
Z b
X
= O(1) sup λk βkν ξν ϕν (t) − fx (t) dt + O(1)||fx ||L2 + O(1)
a k
ν=0
= Mε (x),
that is, the series (1) is maximally Aλ -summable a.e. on [a, b] for every
x ∈ `2λ . The proof of the theorem is now complete.
If P = P (A), then
pk = Pk−1 (eak − 1),
that is,
pk
≥ ak . (11)
Pk−1
We will show that condition (7) in Theorem C is satisfied with P = P (A)
in place of B. Using (3), (10) and (11) gives
m m 2
Pn−1 pk
λ−2 ) sup λ2m |amk | ≤ λ−2
X X
2
n (1 − βkn n 2 λ2k
k=n m≥k k=n
Pk Pk−1
2 2 m
λn Pn−1 X pk
≤
λ2n Pn−1 k=n Pk Pk−1
1 1
= Pn−1 − = O(1).
Pn−1 Pm
58
References
59
CHAPTER III
Proceedings of the Estonian Academy of Sciences.
Physics. Mathematics 53 (2004), No. 1, 13-25
Convergence and λ-boundedness of functional series
with respect to multiplicative systems
P
Abstract. The series ck gk (t), where {gk } is a product system de-
fined by a multiplicative system, is studied. Some sufficient conditions
for p-maximal convergence with speed of this series are found. Also
< f, wk > gk (t) with f ∈ Lp[0,1] , and {wk } being a Walsh
P
the series
system is considered. It is proved that this series converges almost
everywhere for various product systems. In the last section the λ-
boundedness of this series is discussed.
1. INTRODUCTION
Let f = {fk }∞
k=0 be a system of integrable functions on [a, b] satisfying
g0 (t) = 1 and gn (t) = fn0 +1 (t)fn1 +1 (t) . . . fnk +1 (t) (t ∈ [a, b]),
where n = 2n0 + 2n1 + . . . + 2nk (n0 < n1 < . . . < nk ) is the dyadic
representation of n. If {gn } is orthogonal, then {fk } is called orthogonal
multiplicative. If
Z b
gn (t)dt = 0 for n = 1, 2 . . . ,
a
then it is said that {fk } is strongly multiplicative system (see [1]). For
example, the Rademacher system is orthogonal multiplicative and the Walsh
Key words and phrases. Multiplicative systems, Walsh functions, convergence with
speed, convergence almost everywhere, λ-boundedness.
63
system {wn }∞
n=0 is their product system. If
∞ Z b
X
gn (t)dt < ∞,
a
n=0
then the system {fk } is called weakly multiplicative (see [5], p.292). If
m −1 p
1 2X
!
Z Z b
gn (τ )dτ wn (t) dt = O(1),
0
n=0 a
and ∞
X
ck gk (t). (2)
k=0
Notice that if the series (2) converges a.e. on [a, b] for all (ck ) ∈ `2 , then the
same statement is true for the series (1).
In [7] it is proved that the series (1) converges a.e. on [a,b] for all
rearrangements of {ck fk } if (ck ) ∈ `2 and {fk } is p-weakly multiplicative
system for a number p with 1 < p < ∞.
The series (2) is called p-maximally convergent a.e. on [a, b] if it is
convergent a.e. on [a, b] and
n p
Z b X
sup ck gk (t) dt < ∞.
a n
k=0
64
On the other hand Schipp in [6] proved
Theorem B ([6]). A series (2) is 2-maximally convergent a.e. on [a, b]
if (ck ) ∈ `2 and {gk } is the product system of a weakly multiplicative system.
in the sense of the convergence with speed. Let λ = (λk ) be a sequence such
that 0 < λk % ∞. The series (2), which is convergent a.e. on [a, b], is called
1)λ-convergent (or convergent with speed λ) a.e. on [a, b] if the limit
∞
X
lim λn ck gk (t)
n
k=n+1
then it is said that the series (2) is p-maximally λ-convergent a.e. on [a, b].
65
is a Banach space and the sequences ei := (δki )∞
k=0 (i = 0, 1, . . .) form a total
set in (`2λ , k k) (cf. [4], p. 138).
In Section 3 we will consider the series (2) where
Z 1
ck =< f, wk >:= f (t)wk (t)dt (f ∈ Lp[0,1] )
0
or Z b
ck =< f, gk >:= f (t)gk (t)dt (f ∈ Lp[a,b] )
a
and we have found some sufficient conditions for p-maximal convergence a.e.
(1 ≤ p < ∞) of these series.
In Section 4 we will characterize p-maximal λ-boundedness a.e. of the
∞
< f, gk > gk (t), where f ∈ Lp[a,b] .
X
series
k=0
2. p-MAXIMAL λ-CONVERGENCE
66
(see [5], p. 293) and the Walsh system is orthogonal, by the Cauchy-Schwartz
inequality we get
!2 1/2 !2 1/2
Z b m
X Z b Z m
1X
ck gk (t) dt = ck wk (τ )Km (t, τ )dτ dt
a k=0
a 0 k=0
!2 1/2
Z b Z 1 m
X Z 1
≤ ck wk (τ ) Km (t, τ )dτ Km (t, u)du dt
a 0 k=0 0
!2 1/2
m
!
Z 1 X Z b
= ck wk (τ ) Km (t, τ )dt dτ
0 k=0 a
( m
)1/2 (2m −1 Z )1/2 (m )1/2
X X b X
≤ c2k g (t)dt = O(1) 2
ck .
a ν
k=0 ν=0 k=0
Therefore
2 1/2 1/2
Z b ∞
X
X∞
ck gk (t) = O(1) c2k (ck ) ∈ `2λ
a k=m+1
k=m+1
67
and using the Minkowski inequality we have
2 1/2
Z b ∞
X
max λ2n ck gk (t) dt
a n≤m k=n+1
2 1/2
Z b m
X
≤ max λ2n ck gk (t) dt
a n≤m k=n+1
2 1/2
Z b ∞
X
+ λ2m ck gk (t) dt
a k=m+1
2 1/2 1/2
Z b m
X
∞
X
≤ max λ2n ck gk (t) dt + O(1) c2k λ2k .
a n≤m k=n+1
k=m+1
we obtain
2 1/2
Z b m
X
max λ2n ck gk (t) dt
a n≤m k=n+1
!2 1/2
b k m−1
1 1
Z X X
≤ O(1) max cν λν gν (t) dt max λn −
a k≤m ν=0
n≤m
k=n+1
λk λk+1
!2 1/2 !2 1/2
Z b n
X Z b m
X
+ max cν λν gν (t) dt + cν λν gν (t) dt
a n≤m ν=0
a ν=0
!2 1/2
Z b k
X
= O(1) max cν λν gν (t) dt .
a k≤m ν=0
Then by Theorem B
2 1/2
Z b ∞
X
max λ2n ck gk (t) dt = O(1) (ck ) ∈ `2λ (5)
a n≤m k=n+1
68
which gives
∞
X
a.e. on [a, b] for each (ck ) ∈ `2λ .
sup λn ck gk (t) < ∞ (6)
n k=n+1
are continuous and the statements 1◦ (cf. (6)) and 2◦ from Lemma are
fulfilled. By Lemma, the limit
∞
X
lim λn ck gk (t)
n
k=n+1
exists a.e. on [a, b] for every (ck ) ∈ `2λ . Hence the series (2) is λ-convergent
a.e. on [a, b] and by (5) it is 2-maximally λ-convergent. The proof of the
theorem is now complete.
69
By Abel’s transformation (4) and Theorem A we obtain
Z b
∞
X
max λn ck gk (t) dt
a n≤m k=n+1
Z b
X n
Z b ∞
X
= O(1) max ck λk gk (t) dt + λm ck gk (t) dt
a n≤m k=0 a k=m+1
n
Z b
X
= O(1) max ck λk gk (t) dt + O(1) k c kl2 = O(1) k c kl2 .
a n≤m λ λ
k=0
70
On the other hand from [5], p.103 it follows that
n
X
sup < f, wk > wk (τ ) ∈ Lp[0,1] . (8)
n
k=0
p )1/p
n
(Z
b X
max < f, wk > gk (t) dt
a n≤m
k=0
(Z Z n
p )1/p
b 1X
= max < f, wk > wk (τ )Km (t, τ )dτ dt
a n≤m 0 k=0
p
n
(Z Z
b 1 X
≤ max < f, wk > wk (τ ) Km (t, τ )dτ ×
a 0 n≤m k=0
p/q )1/p
Z 1
× Km (t, τ )dτ dt
0
(Z n p Z ! )1/p
1 X b
= max < f, wk > wk (τ ) Km (t, τ )dt dτ
0 n≤m k=0 a
(Z n p )1/p ( ∞ Z )1/p
1 X X b
= O(1) max < f, wk > wk (τ ) dτ gν (t)dt ,
0 n≤m a
k=0 ν=0
71
Set n
k
X
hn (t) := 1− < f, gk > wk (t),
k=0
n+1
n
Z 1 X k
vrai sup 1− wk (τ )wk (t) dτ = O(1),
n 0
k=0
n+1
72
Therefore h(t) := lim hn (t) ∈ Lp[0,1] and < f, gν > are the Walsh-Fourier
n
coefficients of h for every k = 0, 1, 2, . . .:
1 n
k
Z X
< h, wν > = wν (t) lim 1− < f, gk > wk (t)dt
0 n
k=0
n+1
n 1
k
X Z
= lim 1− < f, gk > wk (t)wν (t)dt
n
k=0
n+1 0
ν
= lim 1 − < f, gν >=< f, gν > .
n n+1
This yields the following result.
Theorem 5. If {gk } is a product system of a weakly multiplicative
system, then the series
∞
X
< f, gk > gk (t),
k=0
4. p-MAXIMAL λ-BOUNDEDNESS
∞
X
< f, wk > wk (t) (9)
k=0
is p-maximally λ-bounded a.e. on [0, 1], then the series (7) for the same f
is p-maximally λ-bounded a.e. on [a, b].
Proof. Let (sm ) be a sequence of natural numbers. Because the Walsh
73
system is orthogonal, by the Minkowski inequality we obtain
p )1/p
n
(Z
b X
Cm := max λpn
< f, wk > gk (t) − g(t) dt
a n≤m
k=0
(Z "Z n #p )1/p
b 1 X
p
≤ max λn < f, wk > wk (τ ) − f (τ ) Ksm (t, τ )dτ dt
a n≤m 0 k=0
(Z
b
Z 1 p )1/p
p
+ max λn f (τ )Ksm (t, τ )dτ − g(t) dt .
n≤m a 0
Thus
Cm
(Z n
p ! )1/p
1 X Z b
≤ max λpn < f, wk > wk (τ ) − f (τ ) Ksm (t, τ )dt dτ
0 n≤m
k=0
a
sm p 1/p
Z b 2 X−1
+ λpm < f, wν > gν (t) − g(t) dt
a
ν=0
(Z n
p )1/p
1 X
= O(1) max λpn
< f, wk > wk (τ ) − f (τ ) dτ
0 n≤m
k=0
sm p 1/p
Z b 2 X−1
+ λpm < f, wν > gν (t) − g(t) dt .
a
ν=0
74
Therefore we have
(Z n p )1/p
1 X
Cm = O(1) max λpn < f, wk > wk (τ ) − f (τ ) dτ + O(1).
0 n≤m
k=0
75
References
N.Saealle ja H. Türnpu
76
CHAPTER IV
Acta et Commentationes
Universitatis Tartuensis de Mathematica (to appear)
Uniform convergence and Aλ -boundedness
of series with respect to product systems
Natalia Saealle
Let {fk }∞
k=0 be a system of measurable functions such that
g0 (t) = 1 and gn (t) = fn0 +1 (t)fn1 +1 (t) . . . fnk +1 (t) (t ∈ [a, b]),
where n = 2n0 + 2n1 + . . . + 2nk (n0 < n1 < . . . < nk ) is the dyadic represen-
tation of n. For example, the product system of Rademacher system is the
Walsh system {wn }∞ n=0 (in the Paley enumeration), which is complete and
orthonormal (see e.g. [1], pp. 12, 60).
In this paper we consider the series
∞
X
< u, wk > gk (t), (1)
k=0
79
where Z 1
< u, wk >:= u(τ )wk (τ )dτ (k = 0, 1, . . .)
0
are the Walsh-Fourier coefficients of u. In our previous work [4], we con-
sidered p-maximal λ-boundedness of series (1) in the case of functions u ∈
Lp[0,1] (1 < p < ∞). Now, we suppose, that u is continuous on [0, 1] and
study the uniform convergence, the uniform A-summability, the uniform
Aλ -boundedness, and the uniform regular Aλ -summability of series (1).
Let λ = (λk ) be a sequence of real numbers such that 0 < λk % ∞ and
let A = (αnk ) be a triangular regular summability method. For a function
u ∈ C[0,1] we put
n m
!
X X
bn (A, t) := λn αnk < u, wk > gk (t) − lim αmk < u, wk > gk (t) .
m
k=0 k=0
The series (2) are well studied. For example, it was shown in [6] that for
every point τ0 ∈ [0, 1] there is a continuous function u, whose Walsh-Fourier
series diverges at that point. On the other hand, Walsh remarked, that
n −1
2X
lim < u, wν > wν (τ ) = u(τ ) uniformly in τ ∈ [0, 1] (3)
n
ν=0
80
for every u ∈ C[0,1] .
We prove the following
Theorem 1. Let u ∈ C[0,1] and let A be a regular triangular summability
method.
(a) Series (1) is convergent (A-summable, Aλ -bounded, regularly Aλ -
summable) uniformly on [a, b], if series (2) is convergent (A-summable, Aλ -
bounded, regularly Aλ -summable) uniformly on [0, 1].
(b) If {gk } is an orthogonal system, then series (1) is convergent (A-
summable, Aλ -bounded, regularly Aλ -summable) uniformly on [a, b], if and
only if series (2) is convergent (A-summable, Aλ -bounded, regularly Aλ -
summable) uniformly on [0, 1].
Let A be the summability method of arithmetic means, i.e. A = C 1 =
(γnk ), where
k
1− , if k ≤ n,
γnk := n+1
0, if k > n.
2. Proof of Theorem 1
81
From Lemma 3 it follows that for every speed (λp ) there exists a subsequence
(sp ) of (mn ) such that
sp
2 X −1
lim λp < u, wk > gk (t) − vu (t) = 0 uniformly on [a, b]. (4)
p
k=0
for all p > N . The right side of this inequality converges to zero, hence (4)
holds.
Proof of Lemma 3. First, note that the kernel
n −1
2X
Kn (t, τ ) := gk (t)wk (τ )
k=0
Bn : C[0, 1] −→ L∞
[a,b] (n = 0, 1, . . .)
defined by
2mn −1
X
Bn (u, t) = < u, wk > gk (t).
k=0
On the one hand, we have
Z 1 2mn −1
X
|Bn (u, t)| =
u(τ ) wk (τ )gk (t)dτ
0 k=0
Z 1
≤ kukC[0,1] |Kmn (t, τ )| dτ = kukC[0,1] (t ∈ [a, b], n = 0, 1, . . .),
0
82
for 2mn ≥ i + 1. Therefore (Bn (P, t)) is uniformly convergent on [0, 1] for
every P ∈ P, where P is the collection of finite linear combinations of Walsh
functions. It is known that P is dense in C[0,1] (cf. [1], p. 63).
The assertion of Lemma follows from the Banach-Steinhaus theorem.
Proof of Theorem 1. (a) Let {gn } be a product system. By the orthog-
onality of the Walsh system, we have
Z 1 2m
Xn −1
Consequently,
n
X
λn αnk < u, wk > gk (t) − vu (t)
k=0
n 2m n −1
Z
1X X
= λn αnk < u, wk > wk (τ ) wν (τ )gν (t)dτ − vu (t)
0
ν=0
Z k=0
n
!
1 X
= λn αnk < u, wk > wk (τ ) − u(τ ) + u(τ ) ×
0
k=0
2m n −1
X
× wν (τ )gν (t)dτ − vu (t)
ν=0
2mn −1
Z
1 X
= λn u(τ ) wν (τ )gν (t)dτ − vu (t)
0
ν=0
Z 1 X n
! 2mn −1
X
+ αnk < u, wk > wk (τ ) − u(τ ) wν (τ )gν (t)dτ .
0 k=0 ν=0
Since
2m
Xn −1 Z 1
gν (t) wν (τ )dτ = g0 (t) = 1,
ν=0 0
83
holds for every regular triangular matrix A = (αnk ) and speed (λn ). If
series (2) is uniformly Aλ -bounded or uniformly regularly Aλ -summable,
then, by (4) and (5), series (1) enjoys the same property. To prove that the
uniform convergence of series (2) on [0, 1] implies the uniform convergence
of series (1) on [a, b], we use (4) and (5), where λn = 1 (n = 0, 1, . . .) and
A = Σ. Similarly, using Lemma 3 and (5) we can prove the same statement
concerning the A-summability.
(b) Suppose that the product system {gn } is orthogonal. Then
n
X
λn αnk < u, wk > wk (τ ) − u(τ )
k=0
n
Z !
b X
= λn αnk < u, wk > gk (t) − vu (t) + vu (t) ×
a
k=0
2mn −1
X
× gν (t)wν (τ )dt − u(τ )
ν=0
mn
2 X−1
= λn < vu , gν > wν (τ ) − u(τ )
ν=0
n
! 2mn −1
Z b X X
+ αnk < u, wk > gk (t) − vu (t) gν (t)wν (τ )dt.
a k=0 ν=0
84
holds. Moreover, by Lemma 3, from the A-summability of (1) it follows that
n
X
lim αnk < u, wk > gk (t) = vu (t) uniformly on [a, b]. (7)
n
k=0
Now, to prove that series (2) converges uniformly on [a, b], if series (1)
converges uniformly on [0, 1], we use (3), (7), and inequality (6), where
λn = 1 (n = 0, 1, . . .), A = Σ. The converse assertion follows from part
(a) of this theorem. Similarly we can prove the statements concerning the
A-summability. The assertion concerning the Aλ -boundedness and regular
Aλ -summability follows from (3), (6), (7), and part (a) of this theorem. The
proof is complete.
References
85
CURRICULUM VITAE
Natalia Saealle
Education
Professional employment
87
CURRICULUM VITAE
Natalia Saealle
Haridus
Erialane teenistuskäik
88
DISSERTATIONES MATHEMATICAE
UNIVERSITATIS TARTUENSIS
89
18. Krista Fischer. Structural mean models for analyzing the effect of
compliance in clinical trials. Tartu, 1999. 124 p.
19. Helger Lipmaa. Secure and efficient time-stamping systems. Tartu,
1999. 56 p.
20. Jüri Lember. Consistency of empirical k-centres. Tartu, 1999. 148 p.
21. Ella Puman. Optimization of plastic conical shells. Tartu, 2000.
102 p.
22. Kaili Müürisep. Eesti keele arvutigrammatika: süntaks. Tartu, 2000.
107 p.
23. Varmo Vene. Categorical programming with inductive and coinductive
types. Tartu, 2000. 116 p.
24. Olga Sokratova. Ω-rings, their flat and projective acts with some ap-
plications. Tartu, 2000. 120 p.
25. Maria Zeltser. Investigation of double sequence spaces by soft and
hard analitical methods. Tartu, 2001. 154 p.
26. Ernst Tungel. Optimization of plastic spherical shells. Tartu, 2001.
90 p.
27. Tiina Puolakainen. Eest keele arvutigrammatika: morfoloogiline ühe-
stamine. Tartu, 2001. 138 p.
28. Rainis Haller. M (r, s)-inequalities. Tartu, 2002. 78 p.
29. Jan Villemson. Size-efficient interval time stamps. Tartu, 2002. 82 p.
30. Eno Tõnisson. Solving of expression manipulation exercises in com-
puter algebra systems. Tartu, 2002. 92 p.
31. Mart Abel. Structure of Gelfand-Mazur algebras. Tartu, 2003. 94 p.
32. Vladimir Kuchmei. Affine completeness of some Ockham algebras.
Tartu, 2003. 100 p.
33. Olga Dunajeva. Asymptotic matrix methods in statistical inference
problems. Tartu, 2003. 78 p.
34. Mare Tarang. Stability of the spline collocation method for Volterra
integro-differential equations. Tartu, 2004. 90 p.
35. Tatjana Nahtman. Permutation invariance and reparameterizations in
linear models. Tartu, 2004. 91 p.
36. Märt Möls. Linear mixed models with equivalent predictors. Tartu,
2004. 70 p.
37. Kristiina Hakk. Approximation methods for weakly singular integral
equations with discontinuous coefficients. Tartu, 2004. 137 p.
90
38. Meelis Käärik. Fitting sets to probability distributions. Tartu, 2005.
90 p.
39. Inga Parts. Piecewise polynomial collocation methods for solving
weakly singular integro-differential equations. Tartu, 2005. 140 p.