P.N. Natarajan-Classical Summability Theory-Springer (2017)

P.N.
Natarajan
Classical Summability
Theory
123
P.N. Natarajan
Formerly of the Department of Mathematics
Ramakrishna Mission Vivekananda College
Chennai, Tamil Nadu
India
ISBN 978-981-10-4204-1 ISBN 978-981-10-4205-8 (eBook)

DOI 10.1007/978-981-10-4205-8
Library of Congress Control Number: 2017935546
Springer Nature Singapore Pte Ltd. 2017

This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microlms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specic statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, express or implied, with respect to the material contained herein or
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictional claims in published maps and institutional afliations.
Printed on acid-free paper
This Springer imprint is published by Springer Nature

The registered company is Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore
Preface
The study of convergence of innite series is a very old art. In ancient times, people
were interested in orthodox examination for convergence of innite series.
Divergent series, i.e., innite series which do not converge, was of no interest to
them until the advent of L. Euler (17071783), who took up a serious study of
divergent series. He was later followed by a galaxy of very great mathematicians.
Study of divergent series is the foundation of summability theory. Summability
theory has many utilities in analysis and applied mathematics. An engineer or
physicist, who works on Fourier series, Fourier transforms or analytic continuation,
can nd summability theory very useful for his/her research.
In the present book, some of the contributions of the author to classical
summability theory are highlighted, thereby supplementing, the material already
available in standard texts on summability theory.
There are six chapters in all. The salient features of each chapter are listed below.
In Chap. 1, after a very brief introduction, we recall well-known denitions and
concepts. We state and prove SilvermanToeplitz theorem, Schurs theorem and
then deduce Steinhaus theorem. We introduce a sequence space r , r 1 being a
xed integer and make a detailed study of the space r , especially from the point of
view of sequences of zeros and ones. We prove a Steinhaus type result involving
the space r , which improves Steinhaus theorem. We prove some more Steinhaus
type theorems too.
Chapter 2 deals with the core of a sequence. We present an improvement of
Sherbakhoffs result, which leads to a short and very elegant proof of Knopps core
theorem. We also present some nice properties of the class ; of innite matrices.
Chapter 3 is devoted to a detailed study of some special methods of summability,
viz., the Abel method, the Weighted mean method, the Euler method and the
M; n or Natarajan method. We bring out the connection between the Abel
method and the Natarajan method. Some product theorems involving certain
summability methods are also proved.
In Chap. 4, some nicer properties of the M; n method are established. Further,
we prove a few results on the Cauchy multiplication of certain summable series.
vii
viii Preface
In Chap. 5, a new denition of convergence of a double sequence and a double

series is introduced. In the context of this new denition, SilvermanToeplitz
theorem for 4-dimensional innite matrices is proved. We also prove Schurs and
Steinhaus theorems for 4-dimensional innite matrices.
Finally in Chap. 6, we introduce the Nrlund, the Weighted mean and the
M; m;n or Natarajan methods for double sequences and double series and study
some of their properties.
I thank my mentor Prof. M.S. Rangachari for initiating me to the topic of
summability theoryboth classical and ultrametric. I thank Mr. E. Boopal for
typing the manuscript.
Chennai, India P.N. Natarajan

Contents
1 General Summability Theory and Steinhaus Type Theorems . . .... 1

1.1 Basic Denitions and Concepts . . . . . . . . . . . . . . . . . . . . . . . .... 1
1.2 The SilvermanToeplitz Theorem, Schurs Theorem,
and Steinhaus Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 A Steinhaus Type Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.4 The Role Played by the Sequence Spaces r . . . . . . . . . . . . . . . . . 16
1.5 More Steinhaus Type Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2 Core of a Sequence and the Matrix Class ; . . . . . . . . . . . . . . . . . 27
2.1 Core of a Sequence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2.2 Natarajans Theorem and Knopps Core Theorem . . . . . . . . . . . . . 28
2.3 Some Results for the Matrix Class ; . . . . . . . . . . . . . . . . . . . . 31
2.4 A Mercerian Theorem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3 Special Summability Methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1 Weighted Mean Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.2 M; n Method or Natarajan Method . . . . . . . . . . . . . . . . . . . . . . . 48
3.3 The Abel Method and the M; n Method . . . . . . . . . . . . . . . . . . . 54
3.4 The Euler Method and the M; n Method . . . . . . . . . . . . . . . . . . 55
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4 More Properties of the M; n Method and Cauchy Multiplication
of Certain Summable Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1 Some Nice Properties of the M; n Method . . . . . . . . . . . . . . . . . 63
4.2 Iteration of M; n Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.3 Cauchy Multiplication of M; n -Summable Series . . . . . . . . . . . . 73
4.4 Cauchy Multiplication of Euler Summable Series . . . . . . . . . . . . . . 77
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
ix
x Contents
5 The SilvermanToeplitz, Schur, and Steinhaus Theorems

for Four-Dimensional Innite Matrices . . . . . . . . . . . . . . . . . . . . . . . . 83
5.1 A New Denition of Convergence of a Double Sequence
and a Double Series . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.2 The SilvermanToeplitz Theorem for Four-Dimensional Innite
Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.3 The Schur and Steinhaus Theorems for Four-Dimensional Innite
Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
6 The Nrlund, Weighted Mean, and M; m;n Methods
for Double Sequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.1 Nrlund Method for Double Sequences . . . . . . . . . . . . . . . . . . . . . 101
6.2 Weighted Mean Method for Double Sequences . . . . . . . . . . . . . . . 107
6.3 M; m;n or Natarajan Method for Double Sequences . . . . . . . . . . 120
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
Chapter 1
General Summability Theory and Steinhaus
Type Theorems
The present chapter is devoted to a study of some basic concepts in summability

theory and Steinhaus type theorems. This chapter is divided into 5 sections. In the
first section, we introduce some basic definitions and concepts. In the second section,
we prove the SilvermanToeplitz theorem, Schurs theorem, and Steinhaus theorem.
In the third section, we introduce the sequence space r , r 1 being a fixed integer
and prove a Steinhaus type result improving Steinhaus theorem. The fourth section
is devoted to a study of the sequence spaces r . In the final section, we prove more
Steinhaus type theorems.
1.1 Basic Definitions and Concepts
The study of convergence of infinite series is an ancient art. In ancient times, people
were more concerned with orthodox examinations of convergence of infinite series.
Series, which did not converge, was of no interest to them until the advent of L. Euler
(170783), who took up a series study of divergent series, i.e., series which did
not converge. Euler was followed by a galaxy of great mathematicians C.F. Gauss
(17771855), A.L. Cauchy (17891857), and N.H. Abel (180229). The interest in
the study of divergent series temporarily declined in the second half of the nineteenth
century. It was rekindled at a later date by E. Cesro, who introduced the idea of (C, 1)
convergence in 1890. Since then, many other mathematicians have been contributing
to the study of divergent series. Divergent series have been the motivating factor for
the introduction of summability theory.
Summability theory has many uses in analysis and applied mathematics. An engi-
neer or a physicist, who works with Fourier series, Fourier transforms, or analytic
continuation can find summability theory very useful for his/her research.
Consider the sequence
{sn } = {1, 0, 1, 0, . . . },
Springer Nature Singapore Pte Ltd. 2017 1
P.N. Natarajan, Classical Summability Theory, DOI 10.1007/978-981-10-4205-8_1
2 1 General Summability Theory and Steinhaus
which is known to diverge. However, let
s0 + s1 + + sn
tn = , n = 0, 1, 2, . . . ,
n+1
k+1
i.e., tn = , if n = 2k;
2k + 1
k+1
= , if n = 2k + 1,
2k + 2
proving that
1
tn , n .
2
In this case, we say that the sequence {sn } converges to 21 in the sense of Cesro or
{sn } is (C, 1) summable to 21 . Similarly, consider the infinite series

an = 1 1 + 1 1 + .
n=0
The associated sequence {sn } of partial sums is {1, 0, 1, 0, . . . }, which is (C, 1)

summable to 21 . In this case, we say that the series 1 1 + 1 1 + is (C, 1)
summable to 21 .
We now recall the following definitions and concepts.
Definition 1.1 Given an infinite matrix A = (ank ), n, k = 0, 1, 2, . . . and a sequence
x = {xk }, k = 0, 1, 2, . . . , by the A-transform of x = {xk }, we mean the sequence
Ax = {(Ax)n },

(Ax)n = ank xk , n = 0, 1, 2, . . . ,
k=0
where we suppose that the series on the right converge. If lim (Ax)n = s, we say
n
that the sequence x = {xk } is A-summable or summable A to s.

Given an infinite series xk , define
k=0

n
sn = xk , n = 0, 1, 2, . . . .
k=0

If {sn } is A-summable to s, we say that xk is A-summable to s.
k=0
1.1 Basic Definitions and Concepts 3
Definition 1.2 Let X, Y be sequence spaces. The infinite matrix A = (ank ), n, k

= 0, 1, 2, . . . is said to transform X to Y , written as A (X, Y ) if whenever
the sequence x = {xk } X , (Ax)n is defined, n = 0, 1, 2, . . . and the sequence
{(Ax)n } Y .
Definition 1.3 Let c be the Banach space consisting of all convergent sequences with
respect to the norm x = sup |xk |, x = {xk } c. If A = (ank ) (c, c), A is said to
k0
be convergence preserving or conservative. If, in addition, lim (Ax)n = lim xk ,
n k
x = {xk } c, A is called a regular matrix or a Toeplitz matrix. If A is regular, we
write
A (c, c; P),
the letter P denoting Preservation of limit.
1.2 The SilvermanToeplitz Theorem, Schurs Theorem,

and Steinhaus Theorem
We now prove a landmark theorem in summability theory, due to SilvermanToeplitz,

which characterizes a regular matrix in terms of the entries of the matrix (see [13]).
Theorem 1.1 (SilvermanToeplitz) A = (ank ) is regular, i.e., A (c, c; P) if and
only if

sup |ank | < ; (1.1)
n0 k=0
lim ank = 0, k = 0, 1, 2, . . . ; (1.2)

n
and

lim ank = 1. (1.3)
n
k=0
Proof Proof of the sufficiency part. Let (1.1), (1.2), and (1.3) hold. Let x =
{xk } c with lim xk = s. Since {xk } converges, it is bounded and so |xk | M,
k
k = 0, 1, 2, . . . , M > 0. Now,

|ank xk | M |ank | < , in view of (1.1)
k=0 k=0
and so

(Ax)n = ank xk is defined, n = 0, 1, 2, . . . .
k=0
Now,

(Ax)n = ank (xk s) + s ank , n = 0, 1, 2, . . . . (1.4)
k=0 k=0
Since lim xk = s, given > 0, there exists n N, where N denotes the set of all
k
positive integers, such that

|xk s| < , k > N, (1.5)
2L
where L > 0 is such that

|xn s| L , |ank | L , n = 0, 1, 2, . . . . (1.6)
k=0
Thus,

N

ank (xk s) = ank (xk s) + ank (xk s),
k=0 k=0 k=N +1

N

ank (xk s) |ank ||xk s| + |ank ||xk s|.

k=0 k=0 k=N +1

|ank ||xk s| |ank |, using (1.5)
k=N +1
2L k=0

L , using (1.6)
2L

= ;
2
Using (1.2), we have

|ank | < , k = 0, 1, . . . , N ,
2L(N + 1)
so that,

N

|ank ||xk s| < L(N + 1)
k=0
2L(N + 1)

= .
2
1.2 The SilvermanToeplitz Theorem, Schurs Theorem, and Steinhaus Theorem 5
Consequently,

ank (xk s) < +
2 2
k=0
= , for every > 0.
Thus,

lim ank (xk s) = 0. (1.7)
n
k=0
Taking limit as n in (1.4), we have
lim (Ax)n = s, using (1.3) and (1.7)

n
and so A is regular, completing the proof of the sufficiency part.

Proof of the necessity part. Let A be regular. For every fixed k = 0, 1, 2, . . . ,
consider the sequence x = {xn }, where

1, n = k;
xn =
0, otherwise.
For this sequence x, (Ax)n = ank . Since lim xn = 0 and A is regular, it follows that
n
lim ank = 0, k = 0, 1, 2, . . .
n
so that (1.2) holds. Again consider the sequence x = {xn }, where xn = 1, n =

0, 1, 2, . . . . Note that lim xn = 1. For this sequence x, (Ax)n = ank . Since
n
k=0
lim xn = 1 and A is regular, we have
n

lim ank = 1,
n
k=0

so that (1.3) also holds. It remains to prove (1.1). First, we prove that |ank |
k=0
converges, n = 0, 1, 2, . . . . Suppose not. Then, there exists N N such that

|a N k | diverges.
k=0

In fact, |a N k | diverges to . So, we can find a strictly increasing sequence k( j)
k=0
of positive integers such that

k( j)1
|a N k | > 1, j = 1, 2, . . . . (1.8)
k=k( j1)
Define the sequence x = {xk }, where

|a N k |
, if a N k = 0 and k( j 1) k < k( j), j = 1, 2, . . . ;
xk = ja N k
0, if k = 0 or a N k = 0.

Note that lim xk = 0 and ank xk converges, n = 0, 1, 2, . . . . In particular,
k
k=0

a N k xk converges. However,
k=0

k( j)1
|a N k |
ank xk =
k=0 j=1 k=k( j1)
j

k( j)1
1
= |a N k |
j=1
j k=k( j1)

1
> .
j=1
j

1
This leads to a contradiction since diverges. Thus,
j=1
j

|ank | converges, n = 0, 1, 2, . . . .
k=0
To prove that (1.1) holds, we assume that

sup |ank | =
n0 k=0
and arrive at a contradiction.

We construct two strictly increasing sequences {m( j)} and {n( j)} of positive
integers in the following manner.

Let m(0) = 0. Since |am(0),k | < , choose n(0) such that |am(0),k | <
k=0 k=n(0)+1
1. Having chosen the positive integers m(0), m(1), . . . , m( j 1) and n(0), n(1), . . . ,
n( j 1), choose the positive integers m( j) > m( j 1) and n( j) > n( j 1) such
that

|am( j),k | > j 2 + 2 j + 2; (1.9)
k=0

n( j1)
|am( j),k | < 1; (1.10)
k=0
and

|am( j),k | < 1. (1.11)
k=n( j)+1
Now, define the sequence x = {xk }, where

|a
m( j),k |
, if n( j 1) < k n( j), am( j),k = 0, j = 1, 2, . . . ;
xk = jam( j),k
0, otherwise.
Note that lim xk = 0. Since A is regular, lim (Ax)n = 0. However,

k n

(Ax)m( j) =
am( j),k xk

k=0

n( j1)

n( j)

= am( j),k xk + am( j),k xk + am( j),k xk
k=0 k=n( j1)+1 k=n( j)+1

n( j) n( j1)

am( j),k xk |am( j),k xk | |am( j),k xk |
k=n( j1)+1 k=0 k=n( j)+1
1
n( j)
> |am( j),k | 1 1, using (1.10) and (1.11)
j k=n( j1)+1

1
n( j1)
= |am( j),k | |am( j),k | |am( j),k | 2
j k=0 k=0 k=n( j)+1
1 2
> [( j + 2 j + 2) 1 1] 2, using (1.9), (1.10), (1.11)
j
= j +22
= j, j = 1, 2, . . . .
Thus, {(Ax)m( j) } diverges, which contradicts the fact that {(Ax)n } converges. Con-
sequently, (1.1) holds. This completes the proof of the theorem.
Exercise. Prove that A = (ank ) is conservative, i.e., A (c, c) if and only if (1.1)
holds and
(i) lim ank = k , k = 0, 1, 2, . . . ; (1.12)
n
and

(ii) lim ank = . (1.13)
n
k=0
In such a case, prove that

lim (Ax)n = s + (xk s)k , (1.14)
n
k=0
lim xk = s.
k
Definition 1.4 A = (ank ) is called a Schur matrix if A ( , c), i.e., {(Ax)n } c,

whenever, x = {xk } .
The following result gives a characterization of a Schur matrix in terms of the

entries of the matrix (see [13]).
Theorem 1.2 (Schur) A = (ank ) is a Schur matrix if and only if (1.12) holds and

|ank | converges uniformly in n. (1.15)
k=0

Proof Sufficiency part. Let (1.2) and (1.15) hold. (1.15) implies that |ank | con-
k=0
verges, n = 0, 1, 2, . . . . Note that (1.12) and (1.15) imply that

sup |ank | = M < .
n0 k=0
Thus, for each r = 0, 1, 2, . . . , we have

r
lim |ank | M.
n
k=0
Hence,

r
|k | M, r = 0, 1, 2, . . . ,
k=0
and so

|k | < .
k=0

Thus, if x = {xk } , it follows that ank xk converges absolutely and uniformly
k=0
in n. Consequently,

lim (Ax)n = lim ank xk
n n
k=0

= k x k ,
k=0
proving that {(Ax)n } c, i.e., A ( , c), proving the sufficiency part.

Necessity part. Let A = (ank ) ( , c). Then, A (c, c) and so (1.12) holds.
Again, since A (c, c), (1.1) holds,

i.e., sup |ank | < .
n0 k=0

As in the sufficiency part of the present theorem, it follows that |k | < . We
k=0
write
bnk = ank k , n, k = 0, 1, 2, . . . .

Then, bnk xk converges for all x = {xk } . We now claim that
k=0 n=0

|bnk | 0, n . (1.16)
k=0
Suppose not. Then,

lim |bnk | = h > 0.
n
k=0
So,

|bmk | h, m
k=0
through some subsequence of positive integers. We also note that
lim bmk = 0, k = 0, 1, 2, . . . .
m
We can now find a positive integer m(1) such that

h
|bm(1),k | h <
10
k=0
and
h
|bm(1),0 | + |bm(1),1 | < .
10

Since |bm(1),k | < , we can choose k(2) > 1 such that
k=0

h
|bm(1),k | < .
k=k(2)+1
10
It now follows that

k(2)

|bm(1),k | h = |bm(1),k | h |bm(1),0 | + |bm(1),1 |

k=2 k=0

|bm(1),k |
k=k(2)+1
h h h
< + +
10 10 10
3h
= .
10
Now choose a positive integer m(2) > m(1) such that

h
|bm(2),k | h <
10
k=0
and

k(2)
h
|bm(2),k | < .
k=0
10
Then, choose a positive integer k(3) > k(2) such that

h
|bm(2),k | < .
k=k(3)+1
10
It now follows that

k(3)
3h
|b | <
m(2),k h 10 .
k=k(2)+1
Continuing this way, we find m(1) < m(2) < . . . and 1 = k(1) < k(2) < k(3) <
. . . such that
k(r )
h
|bm(r ),k | < ; (1.17)
k=0
10

h
|bm(r ),k | < ; (1.18)
k=k(r +1)+1
10
and
k(r +1)
3h
|bm(r ),k | h < . (1.19)

k=k(r )+1 10
We now define a sequence x = {xk } as follows: x0 = x1 = 0 and
xk = (1)r sgn bm(r ),k ,
if k(r ) < k k(r + 1), r = 1, 2, . . . . Note that x = {xk } and x = 1. Now,
k(r )

k(r +1)

bm(r ),k xk (1)r h = bm(r ),k xk + bm(r ),k xk

k=0 k=0 k=k(r )+1

r
+ bm(r ),k xk (1) h
k=k(r +1)+1

k(r+1)

= |bm(r ),k | h (1)r
k=k(r )+1

k(r )

+ bm(r ),k xk + bm(r ),k xk
k=0 k=k(r +1)+1
3h h h
< + + , using (1.17), (1.18) and (1.19)
10 10 10
h
= .
2

Consequently, bnk xk is not a Cauchy sequence and so it is not convergent,
k=0 n=0
which is a contradiction. Thus, (1.16) holds. So, given > 0, there exists a positive
integer n 0 such that

|bnk | < , n > n 0 . (1.20)
k=0

Since |bnk | < for 0 n n 0 , we can find a positive integer M such that
k=0

|bnk | < , 0 n n 0 . (1.21)
k=M
In view of (1.20) and (1.21), we have

|bnk | < for all n = 0, 1, 2, . . . ,
k=M

i.e., |bnk | converges uniformly in n. Since |k | < , it follows that |ank |
k=0 k=0 k=0
converges uniformly in n, proving the necessity part. The proof of the theorem is
now complete.
Using Theorems 1.1 and 1.2, we can deduce the following important result.
Theorem 1.3 (Steinhaus) An infinite matrix cannot be both a regular and a Schur
matrix. In other words, given a regular matrix, there exists a bounded, divergent
sequence which is not A-summable.
Proof Let A be a regular matrix. Then, (1.2) and (1.3) hold. Using (1.15),

lim ank = lim ank
n n
k=0 k=0
= 0, using (1.2),
which contradicts (1.3). This establishes our claim.
Exercise. Try to prove Steinhaus theorem without using Schurs theorem, i.e., given
a regular matrix, construct a bounded, divergent sequence x = {xk } such that {(Ax)n }
diverges.
1.3 A Steinhaus Type Theorem
In the context of Steinhaus theorem, the author introduced the sequence space r ,
r 1 being a fixed integer as follows (see [4]).
Definition 1.5 r is defined as the set of all sequences x = {xk } such that
|xk+r xk | 0, k ,
r 1 being a fixed integer.

It is easily proved that r is a closed subspace of with respect to the norm
defined for elements in .
The following result, improving Steinhaus theorem, was proved in [4] (it is worth
noting that a constructive proof was given).
Theorem 1.4
1
r
(c, c; P) (r i , c) = .
i=1
Proof Let A = (ank ) be a regular matrix. We can now choose two sequences of
positive integers {n(m)}, {k(m)} such that if
m = 2 p, n(m) > n(m 1), k(m) > k(m 1) + (2m 5)r,
then

k(m1)+(2m5)r
1
|an(m),k | < ,
k=0
16

1
|an(m),k | < ;
k=k(m1)
16
and if
m = 2 p + 1, n(m) > n(m 1), k(m) > k(m 1) + (m 2)r,
then,

k(m1)+(m2)r
1
|an(m),k | < ,
k=0
16

k(m)
7
|an(m),k | > ,
k=k(m1)+(m2)r +1
8

1
|an(m),k | < .
k=k(m)+1
16
Define the sequence x = {xk } as follows:

if k(2 p 1) < k k(2 p), then

2 p2
, k = k(2 p 1) + 1,

2 p1

k(2 p 1) + 1 < k k(2 p 1) + r,

1,

22 p1

p3
, k = k(2 p 1) + r + 1,

k(2 p 1) + r + 1 < k k(2 p 1) + 2r,

1,

..

.

1, k(2 p 1) + (2 p 4)r + 1 < k k(2 p 1) + (2 p 3)r,
xk =

1
, k = k(2 p 1) + (2 p 3)r + 1,

2 p1

2 p2
, k(2 p 1) + (2 p 3)r + 1 < k k(2 p 1) + (2 p 2)r,

2 p1

0, k = k(2 p 1) + (2 p 2)r + 1,

..

.

2 p1 , k(2 p 1) + (4 p 6)r + 1 < k k(2 p 1) + (4 p 5)r,
1

0, k(2 p 1) + (4 p 5)r < k k(2 p),
and if k(2 p) < k k(2 p + 1), then

1
, k(2 p) < k k(2 p) + r,

2p

2
, k(2 p) + r < k k(2 p) + 2r,

2p
.
xk = ..

2 p1

, k(2 p) + (2 p 2)r < k k(2 p) + (2 p 1)r,

2p
1, k(2 p) + (2 p 1)r < k k(2 p + 1).
1.3 A Steinhaus Type Theorem 15
Note that,
if k(2 p 1) < k k(2 p),

1
|xk+r xk | < ,
2p 1
while
if k(2 p) < k k(2 p + 1),

1
|xk+r xk | < .
2p
Thus,
|xk+r xk | 0, k
and so x = {xk } r .
However,
2p 2
|xk+1 xk | = , if k = k(2 p 1) + (2 p 3)r, p = 1, 2, . . . .
2p 1
Hence,
|xk+1 xk | 0, k
and consequently x = {xk }

/ 1 . In a similar fashion, we can prove that
/ i , i = 2, 3, . . . , (r 1).
x
Thus,
1
r
x
/ i
i=1
and so
1
r
x r i .
i=1
Further,
|(Ax)n(2 p) | < 16
1
+ 16
1
= 18 ,
|(Ax)n(2 p+1) | > 8 16 16
7 1 1
= 3 , p = 1, 2, . . . ,
4
which shows that {(Ax)n }

/ c, completing the proof of the theorem.
Remark 1.1 We note that (c, c; P) (r , c) = . Since ( , c) (r , c), it fol-
lows that (c, c; P) ( , c) = , which is Steinhaus theorem.
We call such results Steinhaus type theorems. More explicitly, whenever there is
some notion of limit or sum in the sequence spaces X, Y , we denote by (X, Y ; P)
that subclass of (X, Y ) consisting of all infinite matrices which preserve this limit or
sum. Then, results of the form,
(X, Y ; P) (Z , Y ) = ,
X Z , are called Steinhaus type theorems.
1.4 The Role Played by the Sequence Spaces r
Let us now study in detail the role played by the sequence spaces r . It is well
known that an infinite matrix which sums all sequences of 0s and 1s sums all

bounded sequences (see [5]). It is clear that any Cauchy sequence is in r so that
r =1
each r is a sequence space containing the space C of Cauchy sequences. It is to be
noted that

C r .
r =1
Though r do not form a tower between C and , they can be deemed to reflect
the measure of non-Cauchy nature of sequences contained in them. It is also easy to
prove that r s if and only if s is a multiple of r and that r r +1 = 1 . It
is worthwhile to observe the nature of location of sequences of 0s and 1s in these
spaces r . In the first instance, we note that a sequence of 0s and 1s is in r if and
only if it is periodic with period r eventually. Consequently, any sequence of 0s and

1s is in r if and only if it is non-periodic. Let N P denote the set of all
r =1

sequences of 0s and 1s in r , i.e., the set of all sequences of 0s and 1s
r =1
which are non-periodic. The following sequences
ei(r ) = {eik
(r )
}
k=0

= 1, 1, . . . , 1, 0, 0, . . . , 0, 1, 1, . . . , 1, 0, 0, . . . , 0, . . . , i = 1, 2, . . . , r,

i r i i r i
(1.22)
have a role to play in the structure of r . Note that there are sequences of 0s and
1s in r which are not necessarily of the form (1.22),
1.4 The Role Played by the Sequence Spaces r 17

e.g., 1, 0, 1, 0, 0, . . . , 0, 1, 0, 1, 0, 0, . . . , 0, . . . .

r 3 r 3
We also note that A = (ank ) sums the sequences in (1.22) if and only if

lim an, j+kr exists, j = 0, 1, 2, . . . , r 1. (1.23)
n
k=0
The role of the sequences in (1.22) is illustrated by the following theorem (see
[6], Theorem 1.1).
Theorem 1.5 A = (ank ) sums every sequence of 0s and 1s in r if and only if

(1.1) and (1.23) hold.
In view of Schurs version of Steinhaus theorem [5], viz. given a regular matrix
A, there exists a sequence of 0s and 1s not summable A, the following questions
arise:

(1) Given a regular matrix A, does there exist a sequence of 0s and 1s in r ,
r =1
i.e., non-periodic which is not summable A?

(2) Given a regular matrix A, does there exist a sequence of 0s and 1s in r ,
r =1
i.e., eventually periodic which is not summable A?
The following example provides the negative answer to question (2). Consider
the infinite matrix
1 0 0 0 0 0 ...
0 1 1 0 0 0 . . .
2 2

A = 0 0 1 1 1 0 . . . .
3 3 3

... ... ... ... ... ... ...

We can prove that A sums all sequences of 0s and 1s in r , using Theorem 7 of
r =1
[7], noting that an eventually periodic sequence of 0s and 1s is almost convergent.
We now record some of the structural properties of , vis-a-vis, the set of all
sequences of 0s and 1s. It can be proved that the closed linear span of the set of all
sequences of 0s and 1s in the supremum norm, i.e., x = sup |xk |, x = {xk }
k0
is (see [8] for details). The following result is an improvement of this assertion.
Theorem 1.6 The closed linear span of N P is .
Proof It suffices to show that periodic sequences of 0s and 1s, which are periodic
from the beginning, are in the closed span of N P. In fact, they are in the linear
span of N P. To this end, we show that any such sequence is the difference of two
sequences of N P. It is clear that any sequence of 0s and 1s which converges to 0 or
1 can be expressed as the difference of two sequences of N P. Hence, we shall take
a divergent sequence x = {xk }, xk+r = xk , k = 0, 1, 2, . . . . Let {k(i)} be a strictly
increasing sequence of positive integers such that k(i + 1) k(i) , i ,
xk(i) = 1, xk(i)+1 = 0, i = 1, 2, . . . . We now construct two sequences x (1) = {xk(1) },
x (2) = {xk(2) }, using x as follows:

= 1, k = k(i) + 1;
xk(1) = xk , k = k(i) + 1 , i = 1, 2, . . . ;

= 1, k = k(i) + 1;
xk(2) = 0, k = k(i) + 1 , i = 1, 2, . . . .
By construction, x = x (1) x (2) . It is easy to observe that x (2) is non-periodic. To

prove that x (1) is non-periodic, suppose x (1) has period p. Since both x (1) and x have
period pr , x (2) = x (1) x has period pr , a contradiction. This completes the proof
of the theorem.
The following is an important result due to Schur [5].
Theorem 1.7 Any matrix A which sums all sequences of 0s and 1s is necessarily
a Schur matrix, i.e., A sums all bounded sequences.
In view of Theorem 1.6, we have an improvement of Theorem 1.7.
Theorem 1.8 Any matrix A which sums all sequences in N P is necessarily a Schur
matrix.
Theorem 1.8 provides an affirmative answer to question (1) already recorded and
is an improvement of Steinhaus theorem.
Thus, it turns out that the probability of success in our search for a sequence of
0s and 1s not summable by a given regular matrix A is more when we concentrate
on non-periodic than on eventually periodic sequences.
Remark 1.2 In the context of Theorem 1.4, one may enquire whether a matrix which

sums all sequences of 0s and 1s in r sums all sequences in r . The answer
r =1 r =1
to this query is, however, in the negative. For, if x = {xk } is a sequence of 0s and

1s in r , then xk+r = xk , k k0 and for some integer r 1, k0 being a positive
r =1
integer, i.e., x is eventually periodic and hence (C, 1) summable as is seen directly
or using the idea of almost convergence (see [9, p. 12]). The sequence
# $
1 2 1 1 2 3 4 3 2 1
0, 1, 0, , , , 0, , , , , , , , 0, . . .
2 2 2 4 4 4 4 4 4 4
1.4 The Role Played by the Sequence Spaces r 19
is in 1 but it is not (C, 1) summable.

In the context of r , i.e., the closure of r in , we introduce the

r =1 r =1
notion of generalized semiperiodic sequences (for the definition of semiperiodic
sequences, one can refer to [10]).
Definition 1.6 x = {xk } is called a generalized semiperiodic sequence, if for any

> 0, there exist positive integers n, k0 such that
|xk xk+sn | < , k k0 , s = 0, 1, 2, . . . .
Let Q denote the set of all generalized semiperiodic sequences. One can prove
that Q is a closed linear subspace of . Further,

Q r .

r =1
1.5 More Steinhaus Type Theorems
In this section, we prove some more Steinhaus type theorems.

The sequence spaces p , p 1, c0 are defined as usual:

p = {x = {xk } : |xk | p < }, p 1;
k=0
c0 = {x = {xk } : lim xk = 0}.
k
Note that p c0 c , where p 1. Just for convenience, we write 1 =

. Let (, c; P ) denote the class of all infinite matrices A (, c) such that

lim (Ax)n = xk , x = {xk } . The following result is known ([11], p. 4, 17).
n
k=0
Theorem 1.9 A = (ank ) (, c) if and only if
sup |ank | < , (1.24)

n,k
and (1.1) holds.
We now prove the following (see [12, Theorem 2.1]).

Theorem 1.10 A = (ank ) (, c; P ) if and only if (1.24), (1.1) hold with k = 1,
k = 0, 1, 2, . . . .
Proof Let A (, c; P ). Then, (1.24) holds. Let ek be the sequence in which 1
occurs in the kth place and 0 elsewhere,
i.e.,
ek = {xi(k) }i=0

,
where
1, i = k;
xi(k) =
0, otherwise,

k = 0, 1, 2, . . . . Then ek , k = 0, 1, 2, . . . and xi(k) = 1, k = 0, 1, 2, . . . .
i=0
Now, (Aek )n = ank so that lim ank = 1, k = 0, 1, 2, . . . , i.e., k = 1, k = 0, 1, 2, . . . .
n
Conversely, let (1.24), (1.1) hold and k = 1, k = 0, 1, 2, . . . . Let x = {xk } . In
view of (1.24), (Ax)n is defined, n = 0, 1, 2, . . . . Now,

(Ax)n = ank xk
k=0

= (ank 1)xk + xk ,
k=0 k=0

this being true since ank xk and xk both converge. Since |xk | < , given
k=0 k=0 k=0
> 0, there exists a positive integer N such that

|xk | < , (1.25)
k=N +1
2A
where A = sup |ank 1|. Since lim ank = 1, k = 0, 1, 2, . . . , N , we can choose a

n,k n
positive integer N > N such that

|ank 1| < , n N , k = 0, 1, 2, . . . , N , (1.26)
2(N + 1)M
where M > 0 is such that |xk | M, k = 0, 1, 2, . . . . Now, for n N ,

1.5 More Steinhaus Type Theorems 21

N

(ank 1)xk |ank 1||xk | + |ank 1||xk |

k=0 k=0 k=N +1

< (N + 1) M+A ,
2(N + 1)M 2A
in view of (1.25) and (1.26)
= ,
so that

lim (ank 1)xk = 0.
n
k=0
Consequently,

lim (Ax)n = xk
n
k=0
and so A (, c; P ), completing the proof of the theorem.
Using Theorem 1.10, we have the following Steinhaus type theorem (see [12,
Theorem 2.2]).
Theorem 1.11
(, c; P ) ( p , c) = , p > 1.
Proof Let A = (ank ) (, c; P ) ( p , c), p > 1. It is known ([11], p. 4, 16) that
A ( p , c), p > 1, if and only if (1.1) holds and

sup |ank |q < , (1.27)
n0 k=0

where 1
p
+ 1
q
= 1. It now follows that |k |q < , which contradicts the fact that
k=0

k = 1, k = 0, 1, 2, . . . , since A (, c; P ) and consequently |k |q diverges.
k=0
This proves our claim.
Remark 1.3 Since p c0 c ,
( , c) (c, c) (c0 , c) ( p , c), p > 1.
So, we have
(, c; P ) (X, c) = ,
when
X = , c, c0 , p , p > 1.
Remark 1.4 is a Banach space with respect to the norm x = sup |xk |, x =
k0
{xk } . c0 and c are closed subspaces of . p , p > 1, is a Banach space with
respect to the norm

1p

x = |xk | p ,
k=0
x = {xk } p . Let B L(, c) denote the space of all bounded linear mappings of into
c. We note that if A (, c; P ), then A is bounded and A = sup |ank |. However,
n,k
(, c; P ) is not a subspace of B L(, c), since lim 2ank = 2, k = 0, 1, 2, . . . and
n
/ (, c; P ) when A (, c; P ).
consequently 2 A
Let

cs = x = {xk } : xk converges .
k=0
Note that cs c0 c . We write A = (ank ) (, ; P) if A (, ) and

(Ax)n = xk , x = {xk } . We also write A (, ; P) if A (, ; P) with
n=0 k=0
lim ank = 0, n = 0, 1, 2, . . . . (1.28)

k
It is well known (see, for instance, [2, p. 189]) that A (, ; P) if and only if
A (, ) and

ank = 1, k = 0, 1, 2, . . . . (1.29)
n=0
Maddox [13] noted that

(, ; P) (cs, ) = .
We now prove that given a (, ; P) matrix A, there exists a sequence x = {xk }
cs whose A-transform is not in . In particular, a lower triangular (, ; P) matrix
cannot belong to the class (cs, ). We need the following lemma, the proof of which
is modeled on that of Fridys (see [14]).
Lemma 1.1 (see [15, pp. 140142, Lemma]) If A = (ank ) (, ) and satisfies
(1.28) and if

lim ank > 0, (1.30)
k
n=0
then there exists a sequence x = {xk } cs, Ax = {(Ax)n }

/ .
Proof By hypothesis, for some > 0, there exists a strictly increasing sequence
{k(i)} of positive integers such that

an,k(i) 2, i = 1, 2, . . . .

n=0
In particular,

an,k(1) 2.

n=0
We then choose a positive integer n(1) such that
%&
1
|an,k(1) | < min , ,
n>n(1)
2 2

this being possible since |ank | < , k = 0, 1, 2, . . . , in view of the fact that
n=0
A (, ). Now, it follows that
n(1)

an,k(1) > .

n=0
In general, having chosen k( j), n( j), j m 1, choose a positive integer k(m) >
k(m 1) such that

an,k(m) 2,

n=0

n(m1) % &
1
|an,k(m) | < min , ,
n=0
2 2
and then choose a positive integer n(m) > n(m 1) such that
&%
1
|an,k(m) | < min m , ,
n>n(m)
2 2
so that

n(m)

an,k(m) > 2

n=n(m1)+1 2 2
= .
Let the sequence x = {xk } be defined by

(1)i+1
, k = k(i);
xk = i
0, k = k(i), i = 1, 2, . . . .
It is clear that x = {xk } cs. Defining n(0) = 0, we have
)
n(N
N
n(m)
|(Ax)n | |(Ax)n |
n=0 m=1 n=n(m1)+1

N
n(m)

= an,k(i) xk(i)

m=1 n=n(m1)+1 i=1

N
n(m)
(1)i+1

= an,k(i)
i
m=1 n=n(m1)+1 i=1

N
n(m)
(1) m+1
(1) i+1
a a
m
n,k(m) n,k(i)
m=1 n=n(m1)+1 i=1
i
i =m

1

N
n(m) 1
= |an,k(m) | |an,k(i) |
m
m=1 n=n(m1)+1
i=1
i
i =m

1
N N n(m)
1
> |an,k(i) |, since 1.
m=1
m m=1 n=n(m1)+1 i=1 i
i =m
Now,

n(m)
1
|an,k(i) | m
m=1 n=n(m1)+1 i<m m=1
2
= 1.
Similarly,

n(m) 1
|an,k(i) | .
m=1 n=n(m1)+1 i>m
2
So,
)
n(N N
1 3
|(Ax)n | > .
n=0 m=1
m 2

1
Since diverges, Ax
/ , completing the proof of the lemma.
m=1
m
We now prove a Steinhaus type result.

Theorem 1.12 ([15, p. 143, Theorem])
(, ; P) (cs, ) = .

Proof Let A = (ank ) (, ; P) (cs, ). Since ank = 1, k = 0, 1, 2, . . . ,
n=0

lim ank = 1 > 0.
k
n=0
In view of the preceding Lemma 1.1, there exists a sequence x = {xk } cs such that
Ax / , which is a contradiction. This completes the proof.
Remark 1.5 If the condition (1.28) is dropped, the above theorem fails to hold as the
following example illustrates. The infinite matrix

1 1 1
0 0 0
A = (ank ) =
0

0 0

is in (, ; P) (cs, ) but a0k 0, k .

For more Steinhaus type theorems, the reader can refer to [1619].
References
1. Hardy, G.H.: Divergent Series. Oxford University Press, Oxford (1949)

2. Maddox, I.J.: Elements of Functional Analysis. Cambridge University Press, Cambridge (1970)
3. Peyerimhoff, A.: Lectures on Summability. Lecture Notes in Mathematics, vol. 107. Springer,
Berlin (1969)
4. Natarajan, P.N.: A Steinhaus type theorem. Proc. Amer. Math. Soc. 99, 559562 (1987)
5. Schur, I.: Uber lineare Transformationen in der Theorie der unendlichen Reihen. J. Reine
Angew. Math. 151, 79111 (1921)
6. Natarajan, P.N.: On certain spaces containing the space of Cauchy sequences. J. Orissa Math.
Soc. 9, 19 (1990)
7. Lorentz, G.G.: A contribution to the theory of divergent sequences. Acta Math. 80, 167190
(1948)
8. Hill, J.D., Hamilton, H.J.: Operation theory and multiple sequence transformations. Duke Math.
J. 8, 154162 (1941)
9. Zeller, K., Beekmann, W.: Theorie der Limitierungverfahren. Springer, Berlin (1970)
10. Berg, I.D., Wilansky, A.: Periodic, almost periodic and semiperiodic sequences. Michigan
Math. J. 9, 363368 (1962)
11. Stieglitz, M., Tietz, H.: Matrixtransformationen von Folgenramen Eine Ergebnisbersicht.
Math. Z. 154, 116 (1977)
12. Natarajan, P.N.: Some Steinhaus type theorems over valued fields. Ann. Math. Blaise Pascal
3, 183188 (1996)
13. Maddox, I.J.: On theorems of Steinhaus type. J. London Math. Soc. 42, 239244 (1967)
14. Fridy, J.A.: Properties of absolute summability matrices. Proc. Amer. Math. Soc. 24, 583585
(1970)
15. Natarajan, P.N.: A theorem of Steinhaus type. J. Anal. 5, 139143 (1997)
16. Natarajan, P.N.: Some more Steinhaus type theorems over valued fields. Ann. Math. Blaise
Pascal 6, 4754 (1999)
17. Natarajan, P.N.: Some more Steinhaus type theorems over valued fields II. Commun. Math.
Anal. 5, 17 (2008)
18. Natarajan, P.N.: Steinhaus type theorems for (C, 1) summable sequences. Comment Math.
Prace Mat. 54, 2127 (2014)
19. Natarajan, P.N.: Steinhaus type theorems for summability matrices. Adv. Dev. Math. Sci. 6,
18 (2014)
Chapter 2
Core of a Sequence and the Matrix Class
(, )
In this chapter, we present a study of the core of a sequence and properties of the
matrix class (, ). This chapter is divided into 4 sections. In the first section, we
introduce the core of a sequence and study some of its properties. In the second
section, we prove an improvement of Sherbakoffs result. This result leads to a short
and elegant proof of Knopps core theorem. The third section is devoted to a study of
the matrix class (, ) in the context of a convolution product *. In the final section,
we prove a Mercerian theorem for the Banach algebra (, ) under the convolution
product *.
2.1 Core of a Sequence
The core of a complex sequence is defined as follows.
Definition 2.1 If x = {xk } is a complex sequence, we denote by K n (x), n =

0, 1, 2, . . . , the smallest closed convex set containing xn , xn+1 , . . . .

K (x) = K n (x)
n=0
is defined as the core of x.
It is known [1] that if x = {xk } is bounded,

K (x) = C lim |zx | (z),
n
zC n
where Cr (z) is the closed ball centered at z and radius r . Sherbakhoff [1] generalized
the notion of the core of a bounded complex sequence by introducing the idea of the
generalized -core K () (x) of a bounded complex sequence as

28 2 Core of a Sequence and the Matrix Class (, )

()
K (x) = C (z), 1. (2.1)
lim |z xn |
zC n
When = 1, K () (x) reduces to the usual core K (x). Sherbakhoff [1] showed that
under the condition

lim |ank | = , 1, (2.2)
n
k=0
()
K (A(x)) K (x).
Natarajan [2] improved Sherbakhoffs result by showing that his result works with
the less stringent precise condition

lim |ank | , 1, (2.3)
n
k=0
(2.3) being also necessary besides the regularity of A for

()
K (A(x)) K (x),
for any bounded complex sequence x. This result for the case = 1 yields a simple
and very elegant proof of Knopps core theorem (see, for instance, [3]).
2.2 Natarajans Theorem and Knopps Core Theorem
Natarajans theorem is
Theorem 2.1 ([2, Theorem 2.1]) A = (ank ) is such that

()
K (A(x)) K (x), 1,
for any bounded sequence x if and only if A is regular and satisfies (2.3),

i.e., lim |ank | , 1.
n
k=0
Proof Let x = {xn } be a bounded sequence. If y K (A(x)), for any z,
|y z| lim |z (Ax)n |.
n
If A is a regular matrix satisfying (2.3), then

2.2 Natarajans Theorem and Knopps Core Theorem 29
|y z| lim |z (Ax)n |
n

= lim ank (z xk )
n
k=0
lim |z xk |,
k
i.e., y C (z) for any z,

lim |z xk |
k
which implies that

()
K (A(x)) K (x).
Conversely, let
()
K (A(x)) K (x).
Then, it is clear that A is regular by considering convergent sequences for which

()
K (x) = lim xn .
n
It remains to prove (2.3). Let, if possible,

lim |ank | > .
n
k=0
Then,

lim |ank | = + h, for some h > 0.
n
k=0
Using the hypothesis and the fact that A is regular, we can choose two strictly
increasing sequences {n(i)} and {k(n(i))} of positive integers such that

k(n(i1))
h
|an(i),k | < ,
k=0
8

k(n(i))
h
|an(i),k | > +
k=k(n(i1))+1
4
and

h
|an(i),k | < .
k=k(n(i))+1
8
Define the sequence x = {xk } by
xk = sgn(an(i),k ), k(n(i 1)) k < k(n(i)), i = 1, 2, . . . .
Now,

k(n(i))
k(n(i1))
|(Ax)n(i) | |an(i),k | |an(i),k |
k=k(n(i1))+1 k=0

|an(i),k |
k=k(n(i))+1
h h h
>+
4 8 8
= , i = 1, 2, . . . . (2.4)

By the regularity of A, {(Ax)n(i) }i=1 is a bounded sequence. It has a convergent
subsequence whose limit cannot be in C (0), in view of (2.4). Using (2.1), we have,
K () (x) C (0) for the sequence x chosen above. This leads to a contradiction of
the fact that K (A(x)) K() (x), completing the proof of the theorem.
Remark 2.1 Condition (2.3) cannot be relaxed if K(A(x)) were to be contained in

K() (x). The infinite matrix

1 0 0 0
0 1 0 0
,
0 0 1 0

where || > , transforms the sequence {1, 0, 1, 0, . . .} into the sequence {, 1

, , 1 , . . .}. K() (x) C (0) while K (A(x)) and / C (0).
Remark 2.2 For a regular matrix A = (ank ), note that

lim |ank | 1
n
k=0
is equivalent to

lim |ank | = 1.
n
k=0
Remark 2.3 The proof of Theorem 2.1, for the case = 1, yields a very simple and
elegant proof of Knopps core theorem. This proof is much simpler than the proofs
of Knopps core theorem known earlier (for instance, see [3, p.149]).
2.3 Some Results for the Matrix Class (, ) 31
2.3 Some Results for the Matrix Class (, )
We recall that

= x = {xk } : |xk | < .
k=0
Note that is a linear space with respect to coordinatewise addition and scalar
multiplication and it is a Banach space with respect to the norm defined by

x = |xk |, x = {xk } .
k=0
(, ; P) denotes the set of all infinite matrices A = (ank ) (, ) such that

(Ax)n = xk , x = {xk } .
n=0 k=0
We recall the following results (see [46])
Theorem 2.2 A = (ank ) (, ) if and only if

sup |ank | < . (2.5)
k0 n=0
Further, A (, ; P) if and only if A (, ) and (1.29) holds.
Theorem 2.3 The matrix class (, ) is a Banach algebra under the norm

A = sup |ank | , A = (ank ) (, ), (2.6)
k0 n=0
with the usual matrix addition, scalar multiplication, and multiplication.
We shall now prove a few results for the matrix class (, ) (see [7]).
Theorem 2.4 The class (, ; P), as a subset of (, ), is a closed convex semigroup
with identity, the multiplication being the usual matrix multiplication.
Proof Let A = (ank ), B = (bnk ) (, ; P) and + = 1, , being nonnegative

real numbers. Then, there exists M > 0 such that

sup |ank | , sup |bnk | M.
k0 n=0 k0 n=0
Now,

sup |ank + bnk | sup |ank | + sup |bnk |
k0 n=0 k0 n=0 k0 n=0
( + )M
= M, since + = 1.
Also,

(ank + bnk ) = ank + bnk
n=0 n=0 n=0
= (1) + (1)
=+
= 1, k = 0, 1, 2, . . . ,

since ank = bnk = 1, k = 0, 1, 2, . . . , using (1.29). In view of Theorem 2.2,
n=0 n=0
A + B (, ; P) so that (, ; P) is a convex subset of (, ).
(m)
Let, now, A = (ank ) (, ; P). Then, there exist A(m) = (ank ), m = 0, 1, 2, . . .
such that
A(m) A 0, m .
Thus, given > 0, there exists a positive integer N such that
A(m) A < , m N ,

(m)
i.e., sup |ank ank | < , m N . (2.7)
k0 n=0
Now,

(N )
(N )
sup |ank | sup |ank ank | + sup |ank |
k0 n=0 k0 n=0 k0 n=0

(N )
< + sup |ank | , using (2.7)
k0 n=0
< ,
since A(N ) (, ; P), so that A (, ). Again,


(N )
ank 1 = ank ank , since A(N ) (, ; P)

n=0 n=0 n=0

(N )
= (ank ank )

n=0

(N )
|ank ank |
n=0

(N )
sup |ank ank |
k0 n=0
< , k = 0, 1, 2, . . . , in view of (2.7)
so that

ank 1 < for all > 0.

n=0
Consequently,

ank = 1, k = 0, 1, 2, . . . .
n=0
Thus, A (, ; P) and so (, ; P) is a closed subset of (, ).

It is clear that the unit matrix is in (, ; P), and it is the identity element of
(, ; P).
To complete the proof, it suffices to check closure under matrix multiplication.
If A = (ank ), B = (bnk ) (, ; P), using Theorem 2.3, AB (, ). In fact, AB
(, ; P), since

cnk = ani bik
n=0 n=0 i=0

= bik ani
i=0 n=0

= bik , since ani = 1, i = 0, 1, 2, . . .
i=0 n=0

= 1, since bik = 1, k = 0, 1, 2, . . . .
i=0
This completes the proof of the theorem.

Remark 2.4 (, ; P) is not an algebra since the sum of two elements of (, ; P) is
not in (, ; P).
We now introduce a convolution product (see [5]).
Definition 2.2 For A = (ank ), B = (bnk ), define

n
(A B)nk = aik bni,k , n, k = 0, 1, 2, . . . . (2.8)
i=0
A B = ((A B)nk ) is called the convolution product of A and B.

We keep the usual norm structure in (, ) as defined by (2.6) and replace matrix
product by the convolution product as defined by (2.8) and prove the following result.
Theorem 2.5 (, ) is a commutative Banach algebra, with identity, under the con-
volution product * as defined by (2.8). Furthermore, (, ; P), as a subset of (, ),
is a closed convex semigroup with identity.
Proof Recall that it was proved in Theorem 2.4 that (, ; P) is a convex subset of
(, ). We will first prove closure under the convolution product *. Let A = (ank ),
B = (bnk ) (, ), and A B = (cnk ). Then,
n

|cnk | = aik bni,k

n=0 n=0 i=0

n
|aik ||bni,k |
n=0 i=0

= |ank | |bnk |
n=0 n=0

sup |ank | sup |bnk |
k0 n=0 k0 n=0
= A B , k = 0, 1, 2, . . . ,
so that

sup |cnk | <
k0 n=0
and so A B (, ). Also,
A B A B .
It is clear that A B = B A. The identity element is the matrix E = (enk ), whose

first row consists of 1s and which has 0s elsewhere,
i.e., e0k = 1, k = 0, 1, 2, . . . ;
enk = 0, n = 1, 2, . . . ; k = 0, 1, 2, . . . .
We note that E (, ; P) and E = 1. It now suffices to prove that (, ; P) is

closed under the convolution product *. Now,

n

cnk = aik bni,k
n=0 n=0 i=0

= ank bnk
n=0 n=0
= 1, k = 0, 1, 2, . . . ,
where A, B (, ; P). This completes the proof of the theorem.
2.4 A Mercerian Theorem
We close the present chapter by proving a Mercerian theorem for the Banach algebra
(, ) under the convolution product *.
Theorem 2.6 If
yn = xn + (cn x0 + cn1 x1 + + cxn1 + xn ),
|c| < 1 and if {yn } , then {xn } , provided
|| < 1 c.
Proof Since (, ) is a Banach algebra under the convolution product *, if || < A
1
,
A (, ), then E A, where E is the identity element of (, ) under *, has an
inverse in (, ). We recall that

1 1 1
0 0 0
E = (enk ) =
0
.
0 0

We note that the equations
yn = xn + (cn x0 + cn1 x1 + + cxn1 + xn ), |c| < 1, n = 0, 1, 2, . . .
can be written in the form

(E + A) x = y ,
where
1 0 0
c 0 0

A= 2 ,
c 0 0

x0 0 0
x1 0 0
x =

x2
,
0 0

y0 0 0
y1 0 0
y =

y2
.
0 0

It is clear that A (, ) with A = 1c 1

. So, if || < 1 c, (E + A) has an
inverse in (, ). Consequently, it follows that
x = (E + A)1 y .
Since y (, ) and (E + A)1 (, ), we have, x (, ). In view of Theorem
2.2, it follows that {xn } , completing the proof of the theorem.
References
1. Sherbakoff, A.A.: On cores of complex sequences and their regular transform. Mat. Zametki 22,
815828 (1977) (Russian)
2. Natarajan, P.N.: On the core of a sequence over valued fields. J. Indian. Math. Soc. 55, 189198
(1990)
3. Cooke, R.G.: Infinite Matrices and Sequence Spaces. Macmillan, New York (1950)
4. Fridy, J.A.: A note on absolute summability. Proc. Amer. Math. Soc. 20, 285286 (1969)
5. Knopp, K., Lorentz, G.G.: Beitrge zur absoluten Limitierung. Arch. Math. 2, 1016 (1949)
7. Natarajan, P.N.: On the algebra (1 , 1 ) of infinite matrices. Analysis (Mnchen) 20, 353357
(2000)
Chapter 3
Special Summability Methods
In the current chapter, some special summability methods are introduced and their
properties are studied. In the first section, we introduce the Weighted Mean method
and study some of its properties. We prove a result, which gives an equivalent for-
mulation of summability by Weighted Mean methods. The result of Hardy [3] and
that of Mricz and Rhoades [4] are particular cases of the above result. The second
section is devoted to a detailed study of the (M, n ) method or the Natarajan method.
The connection between the Abel and the Natarajan methods is brought out in the
third section. In the final section, the Euler method is introduced and its properties are
studied. We prove an interesting product theorem involving the Euler and Natarajan
methods.
3.1 Weighted Mean Method
The Weighted Mean method is defined as follows.
Definition 3.1 ([1, p. 16]) The Weighted Mean method or (N , pn ) method is defined
by the infinite matrix A = (ank ), where

pk
, k n;
ank = Pn
0, k > n,

n
Pn = pk , n = 0, 1, 2, . . . , Pn = 0, n = 0, 1, 2, . . . .
k=0
Theorem 3.1 ([1, p. 16]) The Weighted Mean method (N , pn ) is regular if and only if

n
| pk | = O(Pn ), n ; (3.1)
k=0

38 3 Special Summability Methods
and
Pn , n . (3.2)
Remark 3.1

n
|Pn | | pk |
k=0

n+m
| pk |
k=0
L|Pn+m |,
for some L > 0, m = 0, 1, 2, . . . ; n = 0, 1, 2, . . . .
We now prove a result, which gives an equivalent formulation of summability by

Weighted Mean methods.
Theorem 3.2 (see [2]) Let (N , pn ), (N , qn ) be two regular Weighted Mean methods.

For a given series xk , let
k=0

xk
bn = qn , n = 0, 1, 2, . . . .
k=n
Qk

Let bn converge to s. Then xk is (N , pn ) summable to s if and only if
n=0 k=0
n
1 pk Q k+1 pk1 Q k1
sup q < . (3.3)
n0 |Pn | k=1 k+1 qk

n
Proof Let bn converge to s. Then Bn = bk s, n .
n=0 k=0
Now,

bn bn+1 xk xk
=
qn qn+1 k=n
Q k k=n+1
Q k
xn
= ,
Qn
so that
bn bn+1
xn = Q n , n = 0, 1, 2, . . . .
qn qn+1
3.1 Weighted Mean Method 39
Consequently,

m
sm = xk
k=0
m
bk bk+1
= Qk
k=0
qk qk+1

m
bk
m+1
bk
= Qk Q k1
k=0
qk k=1
q k
b0
m
bk bm+1
= Q0 + (Q k Q k1 ) Q m
q0 k=1 qk qm+1

m
bk bm+1
= b0 + qk Qm
k=1
qk qm+1
m
bm+1
= b0 + bk Q m
k=1
qm+1

m
bm+1
= bk Q m
k=0
qm+1
bm+1
= Bm Q m . (3.4)
qm+1

xk
By hypothesis, converges so that
k=0
Qk
xk
bn
= 0, n .
qn k=n
Qk
Now,
sm Bm bm+1
= , using (3.4).
Qm Qm qm+1
Since {Bn } converges, it is bounded so that |Bn | M, n = 0, 1, 2, . . . , for some

M > 0. Since (N , qn ) is regular, |Q n | , n so that

Bm M

Q |Q | 0, m .
m m
Thus,
sm
0, m .
Qm
Now, for n = 0, 1, 2, . . . ,

xk
bn = qn
k=n
Qk
m
xk
= qn lim
m
k=n
Q k
m
sk sk1
= qn lim , s1 = 0
m
k=n
Qk
m
sk sk
m1
= qn lim
m
k=n
Q k k=n1 Q k+1
m1
sk sm sk
m1
sn1
= qn lim +
m
k=n
Qk Qm k=n
Q k+1 Qn
m1
1 1 sm sn1
= qn lim sk +
m
k=n
Qk Q k+1 Qm Qn

sn1 1 1 sm
= qn + qn sk , since lim =0
Qn k=n
Qk Q k+1 m Q m

sn1 1 1
= qn + qn ck sk , where ck = , k = 0, 1, 2, . . . . (3.5)
Qn k=n
Q k Q k+1
We now have,

n1
Bn = bk + bn
k=0

n1
bk
= qk + bn
k=0
qk

n1 xu
= qk + bn
k=0 u=k
Qu

xu xu xu xu
= q0 + q1 + q2 + + qn1 + bn
u=0
Qu u=1
Qu u=2
Qu u=n1
Qu

n2
n2
n2
xu xu xu xu
= (q0 + q1 + + qn1 ) + q0 + q1 + q2
u=n1
Qu u=0
Qu u=1
Qu u=2
Qu
xn2
+ + qn2 + bn
Q n2

xu xn2 xn3 x0
= Q n1 + bn + Q n2 + Q n3 + + Q0
u=n1
Q u Q n2 Q n3 Q 0

n2
xu
= Q n1 + bn + xk
u=n1
Qu k=0

xu
= sn2 + bn + Q n1
u=n1
Qu

xu xu
= sn2 + qn + Q n1
u=n
Qu u=n1
Qu

xu xu
= sn2 + (Q n Q n1 ) + Q n1
u=n
Q u u=n1
Q u

xu xn1
= sn2 + Q n + Q n1
u=n
Qu Q n1

xu
= sn2 + Q n + xn1
u=n
Q u

xu
= sn1 + Q n
u=n
Qu
bn
= sn1 + Q n
qn

sn1
= sn1 + Q n + ck sk , using (3.5)
Qn k=n

= sn1 sn1 + Q n ck sk
k=n

= Qn ck sk ,
k=n
so that

Bn
= ck sk .
Qn k=n
Consequently,
Bn Bn+1
cn sn = , n = 0, 1, 2, . . . . (3.6)
Qn Q n+1
If {Tn } is the (N , pn ) transform of {sk }, then

1
n
Tn = p k sk
Pn k=0

1 1 Bk
n
Bk+1
= pk , using (3.6)
Pn k=0 ck Q k Q k+1
n
1 p0 B0 pk pk1 Bk pn Bn+1
= +
Pn c0 Q 0 k=1 ck ck1 Q k cn Q n+1

= ank Bk ,
k=0
where

Pn c0 Q 0 ,
1 p0
k = 0;

1 pk pk1
1
, 1 k n;
ank = Pn ck ck1 Qk

1 pn
, k = n + 1;

0,
P c Q
n n n+1
k n + 2.
Note that lim ank = 0, k = 0, 1, 2, . . . . Also,

n

n+1
ank = ank
k=0 k=0
n
1 p0 pk pk1 1 pn
= +
Pn c0 Q 0 k=1 ck ck1 Q k cn Q n+1

1 p0 p1 p0 1 p2 p1 1
= + +
Pn c0 Q 0 c1 c0 Q 1 c2 c1 Q 2

pn pn1 1 pn
+ +
cn cn1 Q n cn Q n+1

1 p0 1 1 p1 1 1 p2 1 1
= + +
Pn c0 Q 0 Q1 c1 Q 1 Q2 c2 Q 2 Q3

pn 1 1
+ +
cn Q n Q n+1

1 p0 p1 pn
= c0 + c1 + + cn
Pn c0 c1 cn
1
= [ p0 + p1 + + pn ]
Pn
1
= Pn
Pn
= 1, n = 0, 1, 2, . . . ,

so that lim ank = 1. By hypothesis, Bk s, k . In view of Theorem 1.1,
n
k=0

Tn s, n , i.e., xn is (N , pn ) summable to s if and only if
n=0
n
1 p0 pn 1 pk pk1
sup c Q + c Q + Q
ck1
< . (3.7)
n0 |Pn | 0 0 n n+1 k=1 k ck
In view of Remark 3.1, we have,
|Pn | L|Pn+m |,
|Q n | L|Q n+m |, (3.8)
for some L > 0, m = 0, 1, 2, . . . ; n = 0, 1, 2, . . . . However,

pn
L pn , using (3.8)
P c Q P c Q
n n n+1 n n n

L pn
=
|Pn Q n | cn
n
L pk pk1 p0
= +
|Pn Q n | k=1 ck ck1 c0
n
L 2 1 pk pk1 p0
+ ,
|Pn | k=1 Q k ck ck1 c0 Q 0
since |Q k | L|Q n |, k n, using (3.8) again
n
L 2 1 pk pk1 p0
+ .
|Pn | k=1 Q k ck ck1 c0 Q 0
Thus, (3.7) is equivalent to

n
1 1 pk pk1
sup < . (3.9)
n0 |Pn | k=1 Q k ck ck1
Now,
pk pk1 pk pk1
=
ck ck1 1
Qk
Q 1k+1 1
Q k1
Q1k
pk Q k Q k+1 pk1 Q k Q k1
=
qk+1 qk
so that (3.9) can now be written as

n
1 1 pk Q k Q k+1 pk1 Q k Q k1
sup < ,
n0 |Pn | k=1 Q k qk+1 qk
n
1 pk Q k+1 pk1 Q k1
i.e., sup q < ,
n0 |Pn | k=1 k+1 qk
completing the proof of the theorem.
Remark 3.2 The result of Hardy [3] and that of Mricz and Rhoades [4] are particular
cases of Theorem 3.2. In the context of Theorem 3.2, it is worthwhile to note that the
result of Mricz and Rhoades is valid even without the assumption Ppnn 0, n .
In the context of Theorem 3.2, we have another interesting result.
Theorem 3.3 (see [5]) Let (N , pn ), (N , qn ) be two regular Weighted Mean methods
and
Pn = O( pn Q n ), n , (3.10)

Pn
i.e., M, n = 0, 1, 2, . . . , for some M > 0.

p Q
n n

Let xn be (N , pn ) summable to s. Then bn converges to s if and only if
n=0 n=0

Pk q q
sup |Q n | k

k+2 < , (3.11)
Q pk Q k pk+1 Q k+2
n0 k=n k+1
where

xk
bn = qn , n = 0, 1, 2, . . . .
k=n
Qk
Proof Let

n
sn = xk ,
k=0
p 0 s0 + p 1 s1 + + p n sn
tn = , n = 0, 1, 2, . . . .
Pn
Then,
s 0 = t0 ,
1
sn = (Pn tn Pn1 tn1 ), n = 1, 2, . . . .
pn

Let xn be (N , pn ) summable to s, i.e., lim tn = s.
n
n=0
Now,
sn 1
= (Pn tn Pn1 tn1 )
Qn pn Q n
1
= [Pn (tn s) Pn1 (tn1 s) + s(Pn Pn1 )]
pn Q n
1
= [Pn (tn s) Pn1 (tn1 s) + spn ]
pn Q n
Pn Pn1 s
= (tn s) (tn1 s) + ,
pn Q n pn Q n Qn
so that

sn
M[|tn s| + L|tn1 s|] + |s| ,
Q |Q n |
n
since |Pn1 | L|Pn |, using Remark 3.1
0, n , since lim tn = s and lim |Q n | = ,
n n
(N , qn ) being regular, using Theorem 3.1.
Now, in view of (3.5),

qn sn1
bn = + qn ck sk ,
Qn k=n
where
1 1
ck = , k = 0, 1, 2, . . . .
Qk Q k+1
We now have,

Bn = Q n ck sk as worked out in Theorem 3.2
k=n

m
= Q n lim ck sk
m
k=n

m
1
= Q n lim ck {Pk tk Pk1 tk1 }
m
k=n
pk

cn Pn1 tn1
m1
cm Pm tm ck ck+1
= Q n lim + Pk tk . (3.12)
m pm pn k=n
pk pk+1
Let

A1 = {xk } : xk is (N , pn ) summable ;
k=0

A2 = {xk } : bk converges .
k=0
We note that A1 , A2 are BK spaces with respect to the norms defined by
x A1 = sup |tn |, x = {xk } A1 ;

n0
and
x A2 = sup |Bn |, x = {xk } A2 ,
n0
respectively. Appealing to BanachSteinhaus theorem (see [6]), we have,
x A2 U x A1 , for some U > 0. (3.13)
For every fixed k = 0, 1, 2, . . . , define the sequence x = {xk }, where

1, if n = k;
xn = 1, if n = k + 1;

0, otherwise.
For this sequence x,

pk
x A1 = and x A2 = |Q k ck |.
Pk
Using (3.13), we have, for k = 0, 1, 2, . . . ,

pk
|Q k ck | U ,
Pk
so that

ck Pk U

p |Q |
k k
0, k , since lim |Q k | = , in view of Theorem 3.1.
k
Thus,
ck Pk
lim = 0. (3.14)
k pk
Using (3.14) in (3.12), we have,

cn Pn1 tn1 ck ck+1
Bn = Qn + Qn Pk tk
pn k=n
pk pk+1

= ank tk ,
k=n
where (ank ) is defined by

if 0 k < n 1;
0,
Q n cn Pn1
ank = pn ,
if k = n 1;

Q n Pk ck ck+1
, if k n.
pk pk+1
First we note that lim ank = 0, k = 0, 1, 2, . . . . We also note that

n

ank = 1, n = 0, 1, 2, . . . ,
k=0
so that

lim ank = 1.
n
k=0

Thus, appealing to Theorem 1.1, bn converges to s if and only if
n=0

cn Pn1 c c
sup |Q n | + Pk
k k+1
< . (3.15)
n0 pn k=n
pk pk+1
However,

Q n cn Pn1
L Q n cn Pn ,
pn pn
since |Pn1 | L|Pn |, using Remark 3.1

ck+1
ck
= L|Q n | Pn , using (3.14)
pk pk+1
k=n

c
k ck+1
L 2 |Q n | Pk ,
pk pk+1
k=n
since |Pn | L|Pk |, k n, using Remark 3.1 again

L 2 |Q n | Pk ck ck+1 . (3.16)
pk pk+1
k=n
Using (3.16), it is clear that (3.15) is equivalent to

c c
sup |Q n | Pk k

k+1 < .
pk pk+1
n0 k=n
Now,

ck ck+1 1 1 1 1 1 1
=
pk pk+1 pk Qk Q k+1 pk+1 Q k+1 Q k+2
qk+1 qk+2
= .
pk Q k Q k+1 pk+1 Q k+1 Q k+2

So bn converges if and only if
n=0

qk+1 qk+2
sup |Q n | < ,
Pk p Q Q p Q Q
n0 k=n k k k+1 k+1 k+1 k+2

Pk qk+1 qk+2

i.e., sup |Q n | < ,
Q pk Q k pk+1 Q k+2
n0 k=n k+1
3.2 (M, n ) Method or Natarajan Method
Natarajan introduced the (M, n ) method and studied some of its properties in [79].
3.2 (M, n ) Method or Natarajan Method 49

Definition 3.2 Let {n } be a sequence such that |n | < . The (M, n ) method
n=0
or the Natarajan method is defined by the infinite matrix (ank ), where

nk , k n;
ank =
0, k > n.
Remark 3.3 In this context, we note that the (M, n ) method reduces to the well-
known Y -method, when 0 = 1 = 21 and n = 0, n 2.
The Natarajan method (M, n ) is a non-trivial summability method, i.e., it is not

equivalent to convergence. Take any (M, n ) method. Then, we have, |n | <
n=0
. Consider the sequence {1, 0, 1, 0, . . . }, which is not convergent. If {n } is the
(M, n )-transform of {1, 0, 1, 0, . . . }, then,
n = 0 + 2 + 4 + + 2k , if n = 2k or 2k + 1.
Now,

|2k | |n | < ,
k=0 n=0
so that {n } converges to s (say). Thus, {1, 0, 1, 0, . . . } is (M, n )-summable to s.

Similarly, the series 11+11+ , whose partial sum sequence is {1, 0, 1, 0, . . . },
is (M, n )-summable. In particular, if 0 = 1 = 21 , n = 0, n 2, the (M, n )
method reduces to the Y -method. The non-convergent sequence {1, 0, 1, 0, . . . } and
the non-convergent series 1 1 + 1 1 + are Y -summable to 21 . The reader
should give other examples of non-convergent sequences and non-convergent series
which are (M, n )-summable.
Theorem 3.4 The (M, n ) method is regular if and only if

n = 1.
n=0

n
n
Proof Since |n | < , sup |ank | = sup |ank | = sup |nk | =
n=0 n0 k=0 n0 k=0 n0 k=0

n
sup |k | < and lim ank = lim nk = 0, k = 0, 1, 2, . . . , since
n0 k=0 n n
lim n = 0. Thus, (M, n ) (ank ) is regular if and only if

n

n
1 = lim ank = lim ank
n n
k=0 k=0
n
= lim nk
n
k=0
n
= lim k
n
k=0

= k ,
k=0
completing the proof.
Definition 3.3 Two matrix methods A = (ank ), B = (bnk ) are said to be consistent
if whenever x = {xk } is A-summable to s and B-summable to t, then s = t.
We now have,
Theorem 3.5 Any two regular methods (M, n ), (M, n ) are consistent.
Proof Let (M, n ), (M, n ) be two regular methods. Let
u n = 0 xn + 1 xn1 + + n x0 s, n
and
vn = 0 xn + 1 xn1 + + n x0 t, n .
Let
n = 0 n + 1 n1 + + n 0 , n = 0, 1, 2, . . . .
Now,
wn = 0 xn + 1 xn1 + + n x0
= (0 0 )xn + (0 1 + 1 0 )xn1 + + (0 n + 1 n1 + + n 0 )x0
= 0 (0 xn + 1 xn1 + + n x0 ) + 1 (0 xn1 + 1 xn2 + + n1 x0 )
+ + n (0 x0 )
= 0 vn + 1 vn1 + + n v0 .
Thus, {wn } is the (M, n )-transform of the sequence {vk }. Since (M, n ) is regular
and lim vk = t, it follows that
k
lim wn = t.
n
In a similar manner, we can prove that

lim wn = s,
n
so that s = t, completing the proof.

Definition 3.4 Let s = {s0 , s1 , s2 , . . . }, s = {0, s0 , s1 , s2 , . . . } and s = {s1 , s2 , . . . }.
The summability method A is said to be translative if s, s are A-summable to t
whenever s is A-summable to t.
Theorem 3.6 Every (M, n ) method is translative.
Proof Writing A (M, n ),
(As)n = n 0 + n1 s0 + n2 s1 + + 0 sn1
= n1 s0 + n2 s1 + + 0 sn1
= u n1 ,
where
u n = n s0 + n1 s1 + + 0 sn , n = 0, 1, 2, . . . .
So, if u n t, n , then
(As)n t, n .
Also,
(As )n = n s1 + n1 s2 + + 0 sn+1
= (n+1 s0 + n s1 + + 0 sn+1 ) n+1 s0
= u n+1 n+1 s0
t, n ,
since u n t, n and n 0, n . Thus, (M, n ) is translative.

Definition 3.5 Given two summability methods A, B, we say that A is included in
B (or B includes A), written as
A B (or B A),
if whenever x = {xk } is A-summable to s, then it is also B-summable to s.

We now have
Theorem 3.7 (Inclusion theorem) Given the methods (M, n ), (M, n ),
(M, n ) (M, n )
if and only if

|kn | < and kn = 1,
n=0 n=0

(x)
where = k(x) = kn x n , (x) = n x n , (x) = n x n .
(x) n=0 n=0 n=0
Proof As in Hardy [10, pp. 6568], we can work out as follows:

Let (x) = n x n , (x) = n x n . Both the series on the right converge for
n=0 n=0
|x| < 1. Let {u n }, {vn } be the (M, n ), (M, n ) transforms of {sn }, respectively. If
|x| < 1, then

vn x n = (0 sn + 1 sn1 + + n s0 )x n
n=0 n=0

= n x n
sn x n
n=0 n=0
= (x)s(x).
Similarly,

u n x n = (x)s(x), if |x| < 1.
n=0
Now,
k(x)(x) = (x),
k(x)(x)s(x) = (x)s(x),

i.e., k(x) un x n = vn x n .
n=0 n=0
Thus,
vn = k0 u n + k1 u n1 + + kn u 0

= an j u j ,
j=0
where
kn j , j n;
an j =
0, j > n.
If (M, n ) (M, n ), then the infinite matrix (an j ) is regular. So, appealing to
Theorem 1.1,

sup |an j | < ,
n0 j=0

n
i.e., sup |an j | < ,
n0 j=0

n
i.e., sup |kn j | < ,
n0 j=0

n
i.e., sup |k j | < ,
n0 j=0

i.e., |k j | < .
j=0

Also, lim an j = 1 implies that kn = 1. Conversely, if |kn | <
n
j=0 n=0 n=0

and kn = 1, then it follows that (an j ) is regular and so lim u j = t implies
j
n=0
that lim vn = t. Thus, (M, n ) (M, n ). The proof of the theorem is now
n
complete.
Consequently, we have the following result.
Theorem 3.8 (Equivalence theorem) The methods (M, n ), (M, n ) are equivalent,
i.e., (M, n ) (M, n ) and vice versa if and only if

|kn | < , |h n | < ;
n=0 n=0
and

kn = 1, h n = 1,
n=0 n=0
where

(x)
= k(x) = kn x n ,
(x) n=0

(x)
= h(x) = hn x n .
(x) n=0
3.3 The Abel Method and the (M, n ) Method
We recall the following (see [11]).
Definition 3.6 The sequence {sn } is Abel summable to s if

lim (1 x) sn x n exists and = s.
x1
n=0
The following result is well known (see [11]).
Theorem 3.9 The Abel method is regular.
Remark 3.4 It is worthwhile to note that the Abel method cannot be described by
an infinite matrix.
The next result gives the connection between the Abel method and the Natarajan
method (see [7]).
Theorem 3.10 ([7, Theorem 4.2]) If {an } is (M, n )-summable to s, where (M, n )
is regular, then {an } is Abel summable to s.
Proof Let {u n } be the (M, n )-transform of the sequence {ak } so that
u n = 0 an + 1 an1 + + n a0 , n = 0, 1, 2, . . . .
Then lim u n = s. Now,

n

n x n
an x n
= un x n ,
n=0 n=0 n=0

(1 x) an x n n x n = (1 x) un x n ,
n=0 n=0 n=0

(1 x) an x n
(1 x) n x n
x n
= (1 x) un x n
,
n=0 n=0 n=0 n=0

(1 x) an x n
(1 x) n x n
= (1 x) un x n
,
n=0 n=0 n=0
(3.17)

n
where n = k , n = 0, 1, 2, . . . . Taking limit as x 1 in (3.17), we have,
k=0
3.3 The Abel Method and the (M, n ) Method 55

lim (1 x) an x n
= lim (1 x) un x n
, (3.18)
x1 x1
n=0 n=0

noting that in view of Theorem 3.9, lim (1 x) = 1, since lim n = n x n
x1 n

n=0

n = 1, (M, n ) being regular. Since lim u n = s, lim (1x) u n x = s,
n
n x1
n=0 n=0
using Theorem 3.9 once again. Thus, from (3.17), it follows that

lim (1 x) an x n
= s,
x1
n=0
i.e., {an } is Abel summable to s, completing the proof of the theorem.
Remark 3.5 The converse of Theorem 3.10 is not true. Consider the (M, n )
method with 0 = 1 = 21 , n = 0, n 2, i.e., the Y -method. The sequence
{1, 1, 2, 2, 3, 3, . . . } is Abel summable to 41 but not (M, n ) summable.
Remark 3.6 Theorem 3.10 implies that any Tauberian theorem for the Abel method
is a Tauberian theorem for any regular (M, n ) method. It is worthwhile to have a
Tauberian theorem for a regular (M, n ) method, which is not a Tauberian theorem
for the Abel method.
We now recall that a product theorem means the following: given regular methods
A, B, does x = {xk } (A) imply B(x) (A), limits being the same, where (A)
denotes the convergence field of A? i.e., does A(x) converges imply A(B(x))
converges to the same limit?
In view of (3.18), we have the following product theorem.
Theorem 3.11 (Product theorem) If {an } is Abel summable to s, then (M, n )({an })
is also Abel summable to s.
Proof If {an } is Abel summable to s, using (3.18),

lim (1 x) un x n
= s,
x1
n=0
where {u n } is the (M, n )-transform of {an }, completing the proof.
3.4 The Euler Method and the (M, n ) Method
The Euler summability method is defined as follows (see [11, pp. 5657]).
Definition 3.7 Let r C {1, 0}, C being the field of complex numbers. The Euler
(r )
method of order r or the (E, r ) method is defined by the infinite matrix (enk ), where

(r )
n
ck r k (1 r )nk , k n;
enk =
0, k > n.
If r = 1,

(1) 1, k = n;
enk =
0, k = n.
If r = 0,
(0)
enk = 0, n = 0, 1, 2, . . . ; k = 1, 2, . . . ;
(0)
en0 = 1, n = 0, 1, 2, . . . .
The following result is well known (see [11, p. 57, Theorem 3.15]).
Theorem 3.12 The (E, r ) method is regular if and only if r is real and 0 < r 1.
Remark 3.7 In the context of Theorem 3.12, we note that
lim e(r ) = 0, k = 0, 1, 2, . . .
n nk
if and only if
|1 r | < 1
(see [11, p. 57, proof of Theorem 3.15]).
Theorem 3.13 ([11, pp. 5859, Corollary 3.17]) If r = 0, the (E, r ) matrix is
invertible and its inverse is the (E, r1 ) matrix.
The following result is very useful (see [10, p. 234, Theorem 176]).

Theorem 3.14 If lim an = 0 and |bn | < , then
n
n=0
lim (a0 bn + a1 bn1 + + an b0 ) = 0.

n
We now prove an interesting product theorem involving the (E, r ) and (M, n )
methods.
Theorem 3.15 ([8, Theorem 2.3]) Given a sequence x = {xk }, if (E, r )(x) con-
verges to s, then (E, r )((M, n )(x)) converges to
3.4 The Euler Method and the (M, n ) Method 57

s 0 + n r n1
,
n=1
where we suppose that the (E, r ) method is regular.
Proof Let

n
n = n
ck r k (1 r )nk xk , (3.19)
k=0
tn = n x0 + n1 x1 + + 0 xn , n = 0, 1, 2, . . . .

By hypothesis, lim n = s and |n | < . Let (E, r )({tn }) = {n } so that
n
n=0

n
n = n
ck r k (1 r )nk tk
k=0
= (1 r )n t0 + n c1r (1 r )n1 t1 + n c2 r 2 (1 r )n2 t2 + + r n tn
= (1 r )n (0 x0 ) + n c1 r (1 r )n1 (0 x1 + 1 x0 )
+ n c2 r 2 (1 r )n2 (0 x2 + 1 x1 + 2 x0 ) +
+ r n (0 xn + 1 xn1 + + n x0 )
= 0 [(1 r )n x0 + n c1r (1 r )n1 x1 + n c2 r 2 (1 r )n2 x2 + + r n xn ]
+ 1 [n c1r (1 r )n1 x0 + n c2 r 2 (1 r )n2 x1 + + r n xn1 ]
+ 2 [n c2 r 2 (1 r )n2 x0 + n c3r 3 (1 r )n3 x1 + + r n xn2 ]
+ + n r n x 0
n n

= 0 n
ck r (1 r ) xk + 1
k nk n
ck r (1 r ) xk1
k nk
k=0 k=1

n
+ 2 n
ck r k (1 r )nk xk2 + + n r n x0
k=2
n

= 0 n + 1 n
ck r (1 r )
k nk
xk1
k=1

n
+ 2 n
ck r (1 r )
k nk
xk2 + + n r n x0 . (3.20)
k=2
Now,

n
n
ck r k (1 r )nk xk1
k=1

n1
= n
c j+1 r j+1 (1 r )n j1 x j
j=0
k

n1
j
1 1 jk
= n
c j+1 r j+1
(1 r ) n j1 j
ck 1 k ,
j=0 k=0
r r
using (3.19) and Theorem 3.13

n1 n1
= r (1 r )nk1 k (1) jk n c j+1 j ck ,

k=0 j=k
interchanging the order of summation. (3.21)
We now use the identity

n1 n1
n1
(1) jk n c j+1 j ck z k = zk ,
k=0 j=k k=0
to conclude that

n1
(1) jk n c j+1 j ck = 1, 0 k n 1. (3.22)
j=k
Using (3.21) and (3.22), we have,

n
n1
n
ck r k (1 r )nk xk1 = r (1 r )nk1 k . (3.23)
k=1 k=0
Using (3.23) and similar results, (3.20) can now be written as

n1

n = 0 n + 1 r (1 r )nk1 k
k=0
n2

+ 2 r (1 r )
2 nk2
k + + n r n 0
k=0
n1

= 0 (n s) + 1 r (1 r ) nk1
(k s)
k=0
n2

+ 2 r (1 r )
2 nk2
(k s) + + n r n (0 s)
k=0
n1

+ s 0 + 1 r (1 r ) nk1
k=0
n2

+2 r (1 r )
2 nk2
+ + n r n
k=0
n1

= 0 (n s) + 1 r (1 r ) nk1
(k s)
k=0
n2

+ 2 r 2 (1 r )nk2 (k s) + + n r n (0 s)
k=0

1 (1 r )n
+ s 0 + 1 r
1 (1 r )
n1
2 1 (1 r )
+2 r + + n r n
1 (1 r )
n1

= 0 (n s) + 1 r (1 r )nk1 (k s)
k=0
n2

+ 2 r (1 r )
2 nk2
(k s) + + n r n (0 s)
k=0
#
+ s 0 + 1 {1 (1 r )n }
$
+2 r {1 (1 r )n1 } + + n r n1 {1 (1 r )}
n1

= 0 (n s) + 1 r (1 r ) nk1
(k s)
k=0
n2

+ 2 r (1 r )
2 nk2
(k s) + + n r n (0 s)
k=0
#
+ s (0 + 1 + 2 r + + n r n1 )
$
{1 (1 r )n + 2 r (1 r )n1 + + n r n1 (1 r )} . (3.24)
Since the (E, r ) method is regular, r is real and 0 < r 1, using Theorem 3.12. So,

|n r n1 | |n | <
n=0 n=0
and
|(1 r )n | = |1 r |n 0, n ,
since |1 r | < 1, in view of Remark 3.7. Using Theorem 3.14, we have,

# $
lim 1 (1 r )n + 2 r (1 r )n1 + + n r n1 (1 r ) = 0.
n

Let n = n r n . Then |n | |n | < , since 0 < r 1. Let
n=0 n=0

n1
n = (1 r )nk1 (k s), n = 0, 1, 2, . . . , 0 = 0.
k=0

lim (n s) = 0 and |(1 r )n | = |1 r |n < , since |1 r | < 1,
n
n=0 n=0
using Remark 3.7 again. Thus, lim n = 0, using Theorem 3.14 once more. Since
n

|n | < and lim n = 0, appealing to Theorem 3.14 again, we see that
n
n=0
n1

lim 1 r (1 r ) nk1
(k s)
n
k=0
n2

+2 r 2 (1 r )nk2 (k s) + + n r n (0 s) = 0.
k=0
Taking limit as n in (3.24), we have,

lim = s 0 + k r k1
,
n n
k=1

noting that the series on the right converges, since |k r k1 | |k | < . In
k=1

k=0

other words, (E, r )((M, n )(x)) converges to s 0 + k r k1 , completing the
k=1
proof of the theorem.
Corollary 3.1 (Product theorem) If we want to get the same limit s, we have to
choose {n } such that

0 + k r k1 = 1,
k=1
an example being the Y -method.

Corollary 3.2 Any regular (E, r ) method and (M, n ) method for which 0 +

k r k1 = 1 are consistent.
k=1
In particular, any regular (E, r ) method and the Y -method are consistent.
References
1. Peyerimhoff, A.: Lectures on Summability. Lecture Notes in Mathematics, vol. 107. Springer,
Berlin (1969)
2. Natarajan, P.N.: A generalization of a theorem of Mricz and Rhoades on Weighted means.
Comment. Math. Prace Mat. 52, 2937 (2012)
3. Hardy, G.H.: A theorem concerning summable series. Proc. Cambridge Philos. Soc. 20, 304
307 (1920-21)
4. Mricz, F., Rhoades, B.E.: An equivalent reformulation of summability by weighted mean
methods. revisited. Linear Algebra Appl. 349, 187192 (2002)
5. Natarajan, P.N.: Another theorem on weighted means. Comment. Math. Prace Mat. 50, 175181
(2010)
7. Natarajan, P.N.: On the (M, n ) method of summability. Analysis (Mnchen) 33, 5156 (2013)
8. Natarajan, P.N.: A product theorem for the Euler and the Natarajan methods of summability.
Analysis (Mnchen) 33, 189195 (2013)
9. Natarajan, P.N.: New properties of the Natarajan method of summability. Comment. Math.
Prace Mat. 55, 915 (2015)
10. Hardy, G.H.: Divergent Series. Oxford (1949)
11. Powell, R.E., Shah, S.M.: Summability Theory and Applications. Prentice-Hall of India (1988)
Chapter 4
More Properties of the (M, n ) Method
and Cauchy Multiplication of Certain
Summable Series
In this chapter, we study some more properties of the Natarajan method. Some results
on the Cauchy multiplication of certain summable series are also proved. This chapter
is divided into 4 sections. The first section is devoted to a study of the (M, n ) method.
For instance, we prove that the set M of all (M, n ) methods is an ordered abelian
semigroup and there exist infinite chains of (M, n ) methods. In the second section,
we study iteration of (M, n ) methods. In the third section, we prove a few results
on the Cauchy multiplication of (M, n )-summable series, while in the last section,
we prove a couple of results on the Cauchy multiplication of Euler summable series.
4.1 Some Nice Properties of the (M, n ) Method
In the present section, following Defranza [1], we record some nice properties of the
Natarajan method. In this context, we recall that

= x = {xk } : |xk | < .
k=0
In the sequel, for convenience, we write (M, ) for (M, n ). Let M denote the set
of all (M, ) methods.
We note the following.
Theorem 4.1 For any method (M, ) M, (M, ) (, ).
Let ((M, )) denote the set of all sequences x = {xk } such that (M, )(x) .
Definition 4.1 Given the methods (M, ), (M, ) M, we say that
(M, ) is stronger than (M, )

64 4 More Properties of the (M, n ) Method and Cauchy
if
((M, )) ((M, ));
(M, ) is strictly stronger than (M, )
if
((M, )) ((M, ));
(M, ) and (M, ) are equivalent
if
((M, )) = ((M, )).
Given the methods (M, ), (M, ) M, we formally define

(x) = n x n , (x) = n x n
n=0 n=0
and

(x)
a(x) = = an x n ;
(x) n=0

(x)
b(x) = = bn x n .
(x) n=0
Following an argument similar to the one in [2, Theorem 18], we can prove the
following results.

Theorem 4.2 If (M, ) M, then the series n x n converges for |x| < 1.
n=0

Theorem 4.3 If (M, ), (M, ) M, then an x n , bn x n have positive radii
n=0 n=0
of convergence and
(i) n = an 0 + an1 1 + + a0 n ;
(ii) n = bn 0 + bn1 1 + + b0 n , n = 0, 1, 2, . . . .

Further, if s = {sn } ((M, )), then the series s(x) = sn x n has positive radius
n=0
of convergence.
4.1 Some Nice Properties of the (M, n ) Method 65
Given the sequences = {n }, = {n }, let = {gn } denote their Cauchy

product, i.e.,
n
gn = k nk , n = 0, 1, 2, . . . .
k=0
Definition 4.2 Given (M, ), (M, ) M, we say that (M, ) is the symmetric
product of (M, ) and (M, ).

Since |n | < and |n | < , in view of Abels theorem on Cauchy
n=0 n=0

multiplication of absolutely convergent series, it follows that |gn | < so that
n=0
(M, ) M, whenever (M, ), (M, ) M.
We need the following result in the sequel.
Lemma 4.1 For given sequences = {n }, = {n }, let = . Let the methods

(M, ), (M, ) M. Then,
((M, )) ((M, ))
if and only if
, i.e., (M, ) M.
Proof For any sequence x = {xk },
((M, )(x))n = n x0 + n1 x1 + + 0 xn
= (n 0 + n1 1 + + 0 n )x0
+ (n1 0 + n2 1 + + 0 n1 )x1
+ + (0 0 )xn
= (n x0 + n1 x1 + + 0 xn )0
+ (n1 x0 + n2 x1 + + 0 xn1 )1
+ + (0 x0 )n
= ((M, )(x))n 0 + ((M, )(x))n1 1
+ + ((M, )(x))0 n

= tnk ((M, )(x))k ,
k=0
where
nk , k n;
tnk =
0, k > n.

In view of Theorem 2.2, ((M, )) ((M, )) if and only if |n | < , i.e.,
n=0
, completing the proof.
We now deduce an inclusion result.
Theorem 4.4 Given the methods (M, ), (M, ) M,
((M, )) ((M, ))
if and only if
b = {bn } .
Proof In Lemma 4.1, we replace the sequence by the sequence b so that
=b =
and so
((M, )) ((M, ))
if and only if b = {bn } .
Corollary 4.1 Let (M, ), (M, ) M. Then,

(i) ((M, )) = ((M, )) if and only if {an } and {bn } ;
(ii) ((M, )) ((M, )) if and only if {an }
/ and {bn } .
Corollary 4.2 For (M, ) M, let h(x) = 1

(x)
. Then,
((M, )) =
if and only if {h n } .

Proof Let I be the identity map so that (I ) = . Now, I (x) = i n x n = 1, i.e.,
n=0
i 0 = 1, i n = 0, n 1. So a(x) = (x)
I (x)
= (x) and b(x) = I (x)
(x)
= (x)
1
= h(x).
Using Theorem 4.4, it follows that {h n } .
Corollary 4.3 Let (M, ), (M, ) M and = . Then,
((M, )) ((M, )).
Theorem 4.5 Let (M, ) M. Let = {0 , 1 , 2 , . . .}, 0 = 0 . Then (M, )

M and
((M, )) ((M, )) = .
Proof Since {n } , it follows that {n } so that the method (M, ) M.

Now, for any sequence x = {xk },
((M, )(x))n = 0 xn + 1 xn1 + + n x0

= 0 xn + 1 xn1 + + n x0
= (0 xn + 1 xn1 + + n x0 ) + (0 0 )xn
= ((M, )(x))n + (0 0 )xn . (4.1)
Let {xn } . Since (M, ), (M, ) (, ), using Theorem 4.1,
{((M, )(x))n } and {((M, )(x))n }
so that
{xn } ((M, )) ((M, )).
Consequently,
((M, )) ((M, )).
Let, now, {xn } ((M, )) ((M, )). Then, {((M, )(x))n } and
{((M, )(x))n } . Now, using (4.1), it follows that {xn } . Thus,
((M, )) ((M, )) .
Consequently,
((M, )) ((M, )) = ,
We now recall that M denotes the set of all (M, ) methods. We now prove
that M is an ordered abelian semigroup, the order relation being the set inclusion
between summability fields of the type ((M, )) and the binary operation being the
symmetric product *.
Lemma 4.2 If (M, ), (M, ) M and = , then (M, ) M and
((M, )) ((M, )) ((M, )).

Proof Since |n | < and |n | < , |n | < so that (M, ) M
n=0 n=0 n=0
as noted earlier. Using Corollary 4.3, we have,
((M, )) ((M, )) ((M, )).

Lemma 4.3 Let (M, ), (M, ), (M, ) M. Let = , = .

(i) if ((M, )) ((M, )), then ((M, )) ((M, ));
(ii) if ((M, )) ((M, )), then ((M, )) ((M, )).
Proof Let b(x) = (x)(x)

and c(x) = (x)
(x)
. By Theorem 4.4, {bn } . We claim that
{cn } . First, we note that
(x) = (x)(x)
and
(x) = (x)(x).
Now,

(x)
cn x n = c(x) =
n=0
(x)
(x)(x)
=
(x)(x)
(x)
=
(x)
= b(x)

= bn x n .
n=0
Thus,
cn = bn , n = 0, 1, 2, . . .
and so {cn } . Using Theorem 4.4 again, we have,
((M, )) ((M, )).
The second part of the theorem follows using Corollary 4.1(ii).
We now have
Theorem 4.6 With respect to strictly -weaker than as the order relation and
symmetric product * as the binary operation and symmetric product * as the binary
operation, M is an ordered abelian semigroup.
Lemma 4.4 Let (M, ) M. Define 1 = 0, = n12+n , n 0. Then (M, )

M and
((M, )) ((M, )).
Proof Since {n } , {n } so that (M, ) M. Now,


(x) = n x n
n=0

n1 + n
= xn
n=0
2
1+x
= (x).
2
It is now clear that
((M, )) ((M, )),
using Corollary 4.1(ii).
We conclude this section with the following interesting result.
Theorem 4.7 There are infinite chains of (M, ) methods from M.

Proof Let {(1)
n } be a sequence such that |(1) (1)
n | < . Let (x) = (1) n
n x .
n=0 n=0
Then, (M, (1) ) M. For n 2, define
n1

1+x
(n)
(x) = (n1)
k xk.
2 k=0
Then, (M, (n) ) M, n 2. Now, applying Lemma 4.4 repeatedly, we have,
((M, (1) )) ((M, (2) )) ((M, (n) )) .
This completes the proof of the theorem.
4.2 Iteration of (M, n ) Methods
In this section, we prove a few theorems on the iteration product of Natarajan methods
(see [3]).
For convenience, we denote the methods (M, pn ), (M, qn ), and (M, tn ) by (M, p),
(M, q), and (M, t), respectively.
Theorem 4.8 Let (M, p), (M, t) be regular methods. Then (M, p)(M, t) is also
regular, where we define, for x = {xk },
((M, t)(M, p))(x) = (M, t)((M, p)(x)).

Proof Let {n } be the (M, p)-transform of x = {xk } and {n } be the (M, t)(M, p)-
transform of x = {xk }. Then,

n
n = pk xnk ,
k=0
n
n = tk nk , n = 0, 1, 2, . . . .
k=0
Now,

n
n = tk nk
k=0
= t0 n + t1 n1 + + tn 0
n n1

= t0 pk xnk + t1 pk xn1k
k=0 k=0
+ + tn1 [ p0 x1 + p1 x0 ] + tn [ p0 x0 ]
= t0 [ p0 xn + p1 xn1 + + pn x0 ]
+ t1 [ p0 xn1 + p1 xn2 + + pn1 x0 ]
+ + tn1 [ p0 x1 + p1 x0 ] + tn [ p0 x0 ]
= ( p0 t0 )xn + ( p0 t1 + p1 t0 )xn1
+ + ( p0 tn + p1 tn1 + + pn t0 )x0

= cnk xk ,
k=0
where

k
p xk , k n;
cnk = =0

0, k > n.
Now,

n
|cnk | = |cnk |
k=0 k=0

n
k
= p tk

k=0 =0
4.2 Iteration of (M, n ) Methods 71

n
k
| p ||tk |
k=0 =0

1
= | p0 ||t0 | + | p ||t1 |
=0

2
n
+ | p ||t2 | + + | p ||tn |
=0 =0
= | p0 ||t0 | + (| p0 ||t1 | + | p1 ||t0 |)

+ (| p0 ||t2 | + | p1 ||t1 | + | p2 ||t0 |)
+ + (| p0 ||tn | + | p1 ||tn1 | + + | pn ||t0 |)
n n1

= | p0 | |tk | + | p1 | |tk |
k=0 k=0
+ + | pn |(|t0 |)

|tk | | pk | , n = 0, 1, 2, . . . ,
k=0 k=0
so that

sup |cnk | |tk | | pk | < ,
n0 k=0 k=0 k=0

noting that |tk | < and | pk | < .
k=0 k=0

Using lim pn = 0, |tn | < and Theorem 3.14, it follows that
n
n=0
lim cnk = 0, k = 0, 1, 2, . . . .
n
Now,

n
cnk = cnk
k=0 k=0

n
k
= p tk
k=0 =0
n n1

= p0 tk + p 1 tk + + pn (t0 )
k=0 k=0
= p0 Tn + p1 Tn1 + + pn T0 , where

n
Tn = tk , n = 0, 1, 2, . . .
k=0
= p0 (Tn 1) + p0 (Tn1 1) + + pn (T0 1)

n
+ Pn , where Pn = pk , n = 0, 1, 2, . . . .
k=0
Since (M, t), (M, p) are regular, lim Tn = lim Pn = 1, in view of Theorem 3.4.
n n

Using lim (Tn 1) = 0, | pn | < and Theorem 3.14, we have,
n
n=0
lim [ p0 (Tn 1) + p1 (Tn1 1) + + pn (T0 1)] = 0.

n
Thus,

lim cnk = 0 + 1 = 1.
n
k=0
In view of Theorem 1.1, (cnk ) is regular. In other words, (M, t)(M, p) is regular,
Theorem 4.9 Let (M, p), (M, q), (M, t) be regular methods. Then
(M, p) (M, q)
if and only if
(M, t)(M, p) (M, t)(M, q).
Proof We write

n
n
rn = pk tnk , rn = qk tnk , n = 0, 1, 2, . . . .
k=0 k=0
Let

r = {rn }, r = {rn }, r (x) = rn x n ,
n=0

r (x) = rn x n , p(x) = pn x n , q(x) = qn x n
n=0 n=0 n=0
4.2 Iteration of (M, n ) Methods 73
and

t (x) = tn x n .
n=0
Since (M, p), (M, q), and (M, t) are regular, (M, t)(M, p), and (M, t)(M, q) are
regular too in view of Theorem 4.8. To prove the present theorem, it suffices to show
that
r (x) q(x)
= .
r (x) p(x)
We first note that
r (x) = p(x)t (x) and r (x) = q(x)t (x)
so that
r (x) q(x)
= .
r (x) p(x)
We now use Theorem 2.1 of [4] to arrive at the conclusion, thus completing the
proof of the theorem.
In view of Theorem 2.1 of [4], we can reformulate Theorem 4.9 as follows:
Theorem 4.10 For given regular methods (M, p), (M, q) and (M, t), the following
statements are equivalent:
(i) (M, p) (M, q);
(ii) (M, t)(M, p) (M, t)(M, q);
and

(iii) |kn | < and kn = 1,
n=0 n=0

q(x)
where = k(x) = kn x n .
p(x) n=0
4.3 Cauchy Multiplication of (M, n )-Summable Series
In this section, we prove a few results on the Cauchy multiplication of (M, n )-

summable series (see [5]).

Theorem 4.11 If |ak | < and {bk } is (M, n )-summable to B, then {ck } is
k=0

n

(M, n )-summable to AB, where cn = ak bnk , n = 0, 1, 2, . . . and ak = A.
k=0 k=0
Proof Let
tn = 0 bn + 1 bn1 + + n b0 ,
u n = 0 cn + 1 cn1 + + n c0 , n = 0, 1, 2, . . . .
By hypothesis, lim tn = B. Now,

n
u n = 0 cn + 1 cn1 + + n c0
= 0 (a0 bn + a1 bn1 + + an b0 )
+ 1 (a0 bn1 + a1 bn2 + + an1 b0 )
+ + n (a0 b0 )
= a0 (0 bn + 1 bn1 + + n b0 )
+ a1 (0 bn1 + 1 bn2 + + n1 b0 )
+ + an (0 b0 )
= a0 tn + a1 tn1 + + an t0
= [a0 (tn B) + a1 (tn1 B) + an (t0 B)]
+ B[a0 + a1 + + an ].

Since |an | < and lim (tn B) = 0, using Theorem 3.14,
n
n=0
lim [a0 (tn B) + a1 (tn1 B) + + an (t0 B)] = 0,

n
so that

lim u n = B an = AB,
n
n=0
i.e., {ck } is (M, n )-summable to AB, completing the proof.
It is now easy to prove the following theorem on similar lines.

Theorem 4.12 If |ak | < and ak = A and bk is (M, n )-summable to
k=0 k=0 k=0

n
B, then ck is (M, n )-summable to AB, where cn = ak bnk , n = 0, 1, 2, . . . .
k=0 k=0
4.3 Cauchy Multiplication of (M, n )-Summable Series 75
Following Mears [6], we prove the following result.

Theorem 4.13 If ak is (M, n )-summable to A, bk is (M, n )-summable to
k=0 k=0

k
B, where |k k+1 | < , {n } is the (M, n )-transform of {Bk }, Bk = bj,
k=0 j=0

n
k = 0, 1, 2, . . . , then ck is (M, n )-summable to AB, where cn = ak bnk ,
k=0 k=0

n
n = k nk , n = 0, 1, 2, . . . .
k=0

Proof First, we note that since |n | < and |n | < , |n | < so
n=0 n=0 n=0
that the method (M, n ) is defined. Let

n
n
n
An = ak , Bn = bk , Cn = ck , n = 0, 1, 2, . . . .
k=0 k=0 k=0
Let

n
n
n
n = k Ank , n = k Bnk , n = k Cnk , n = 0, 1, 2, . . . .
k=0 k=0 k=0
We now claim that

n
n1
n = k nk k nk1 .
k=0 k=0
We first note that

Cn = a0 Bn + a1 Bn1 + + an B0 ,
so that
n = 0 Cn + 1 Cn1 + + n C0
= 0 (a0 Bn + a1 Bn1 + + an B0 )
+ 1 (a0 Bn1 + a1 Bn2 + + an1 B0 )
+ + n (a0 B0 )
= a0 (0 Bn + 1 Bn1 + + n B0 )
+ a1 (0 Bn1 + 1 Bn2 + + n1 B0 )
+ + an (0 B0 ).
One can prove that
0 Bn + 1 Bn1 + + n B0 = 0 n + 1 n1 + + n 0 , n = 0, 1, 2, . . . ,
so that
n = a0 (0 n + 1 n1 + + n 0 )
+ a1 (0 n1 + 1 n2 + + n1 0 )
+ + an (0 0 )
= A0 (0 n + 1 n1 + + n 0 )
+ (A1 A0 )(0 n1 + 1 n2 + + n1 0 )
+ (A2 A1 )(0 n2 + 1 n3 + + n2 0 )
+ + (An An1 )(0 0 )
= n (0 A0 ) + n1 [1 A0 + 0 (A1 A0 )]
+ n2 [2 A0 + 1 (A1 A0 ) + 0 (A2 A1 )]
+ + 0 [n A0 + n1 (A1 A0 ) + n2 (A2 A1 )
+ + 0 (An An1 )]
= n [0 A0 ] + n1 [(0 A1 + 1 A0 ) 0 A0 ]
+ n2 [(0 A2 + 1 A1 + 2 A0 ) (0 A1 + 1 A0 )]
+ + 0 [(0 An + 1 An1 + + n A0 )
(0 An1 + 1 An2 + + n1 A0 )]
= n 0 + n1 [1 0 ] + n2 [2 1 ]
+ + 0 [n n1 ]
= (0 n + 1 n1 + + n 0 )
(0 n1 + 1 n2 + + n1 0 )

n
n1
= k nk k nk1 ,
k=0 k=0
proving our claim. Thus, for n = 0, 1, 2, . . . ,

n
n1
n = (k A)nk (k A)nk1
k=0 k=0

n
n1
+A nk nk1
k=0 k=0

n
n1
= (k A)nk (k A)nk1 + An
k=0 k=0
4.3 Cauchy Multiplication of (M, n )-Summable Series 77

n1
= (k A)(nk nk1 ) + (n A)0 + An .
k=0

Since lim (k A) = 0 and |k k+1 | < , in view of Theorem 3.14,
k
k=0

n1
lim (k A)(nk nk1 ) = 0,
n
k=0
so that
lim n = AB,
n

i.e., ck is (M, n )-summable to AB, completing the proof of the theorem.
k=0

Remark 4.1 If (M, n ), (M, n ) are regular, then n = n = 1, in view of
n=0 n=0
Theorem 3.4. By Abels theorem on Cauchy multiplication of absolutely convergent
series,

n = n n = 1.
n=0 n=0 n=0
Again appealing to Theorem 3.4, it follows that (M, n ) is regular too.
4.4 Cauchy Multiplication of Euler Summable Series

Theorem 4.14 |xk | < and {yk } is (E, r )-summable to , where the (E, r )
k=0
method is regular, then {z k } is (E, r )-summable to

x0 + xk r k1
,
k=1

n
where z n = xk ynk , n = 0, 1, 2, . . . .
k=0
Proof Let {n } be the (E, r )-transform of {yk }. Then,

n
n = n
ck r k (1 r )nk yk , n = 0, 1, 2, . . . . (4.2)
k=0
By hypothesis, lim n = . Let {n } be the (E, r )-transform of {z k } so that

n

n
n = n
ck r k (1 r )nk z k
k=0
= (1 r )n z 0 + n c1 r (1 r )n1 z 1 + n c2 r 2 (1 r )n2 z 2
+ + r n zn
= (1 r )n (x0 y0 ) + n c1r (1 r )n1 (x0 y1 + x1 y0 )
+ n c2 r 2 (1 r )n2 (x0 y2 + x1 y1 + x2 y0 )
+ + r n (x0 yn + x1 yn1 + + xn y0 )
= x0 [(1 r )n y0 + n c1r (1 r )n1 y1 + + r n yn ]
+ x1 [n c1r (1 r )n1 y0 + n c2 r 2 (1 r )n2 y1 + + r n yn1 ]
+ + xn [r n y0 ]
n

= x0 n
ck r (1 r ) yk
k nk
k=0
n

+ x1 n
ck r (1 r )
k nk
yk1
k=1
n

+ x2 n
ck r (1 r )
k nk
yk2 + + xn [r n y0 ]
k=2
n

= x 0 n + x 1 n
ck r (1 r )
k nk
yk1
k=1
n

+ x2 n
ck r (1 r )
k nk
yk2 + + xn [r n y0 ]. (4.3)
k=2
Now,

n
n1
n
ck r k (1 r )nk yk1 = n
c j+1 r j+1 (1 r )n j1 y j
k=1 j=0

n1
n
= c j+1 r j+1 (1 r )n j1
j=0
4.4 Cauchy Multiplication of Euler Summable Series 79
k

j
1 1 jk
j
ck 1 k ,
k=0
r r
using Theorem 3.13 and (4.2)

n1 n1
= r (1 r )nk1 k (1) jk n c j+1 j ck ,

k=0 j=k
interchanging the order summation. (4.4)
Using the identity

n1 n1
n1
(1) jk n c j+1 j ck z k = zk ,
k=0 j=k k=0
we have,

n1
(1) jk n c j+1 j ck = 1, 0 k n 1. (4.5)
j=k
Thus, using (4.4) and (4.5), we have,

n
n1
n
ck r k (1 r )nk yk1 = r (1 r )nk1 k . (4.6)
k=1 k=0
Using (4.6) and similar results, (4.3) can now be written as
n1

n = x 0 n + x 1 r (1 r ) nk1
k
k=0
n2

+ x2 r (1 r )
2 nk2
k + + xn [r n 0 ]
k=0
n1

= x0 (n ) + x1 r (1 r ) nk1
(k )
k=0
n2

+ x2 r (1 r )
2 nk2
(k ) + + xn [r n (0 )]
k=0
n1

+ x0 + x1 r (1 r ) nk1
k=0
n2

+x2 r (1 r )
2 nk2
+ + xn {r }
n
k=0
n1

= x0 (n ) + x1 r (1 r ) nk1
(k )
k=0
n2

+ x2 r (1 r )
2 nk2
(k ) + + xn [r n (0 )]
k=0

1 (1 r )n 1 (1 r )n1
+ x0 + x1r + x2r 2 + + xn r n
1 (1 r ) 1 (1 r )
n1

= x0 (n ) + x1 r (1 r ) nk1
(k )
k=0
n2

+ x2 r (1 r )
2 nk2
(k ) + + xn [r n (0 )]
k=0
+ [x0 + x1 {1 (1 r )n } + x2 r {1 (1 r )n1 }
+ + xn r n1 {1 (1 r )}]
n1

= x0 (n ) + x1r (1 r ) nk1
(k )
k=0
n2

+ x2r 2 (1 r )nk2 (k ) + + xn r n {0 }
k=0
+ [(x0 + x1 + x2 r + + xn r n1 )
{x1 (1 r )n + x2 r (1 r )n1 + + xn r n1 (1 r )}].
(4.7)
Since the (E, r ) method is regular, 0 < r 1, using Theorem 3.12. Now,

|xn r n1 | |xn |, since 0 < r 1
n=1 n=0
<
and |(1 r )n | = |1 r |n 0, n , using Remark 3.7, since the (E, r ) method

is regular. Note that the sequence
{x1 (1 r )n + x2 r (1 r )n1 + + xn r n1 (1 r )}
is the Cauchy product of the sequences {xn r n1 } and {(1 r )n }. In view of

Theorem 3.14,
4.4 Cauchy Multiplication of Euler Summable Series 81
lim {x1 (1 r )n + x2 r (1 r )n1 + + xn r n1 (1 r )} = 0.

n

Let n = xn r n . |n | = |xn r n | |xn | < , since 0 < r 1, using
n=0 n=0 n=0
Theorem 3.12 again. Let

n1
n = (1 r )nk1 (k ).
k=0
Note that {n } is the Cauchy product of the sequences {(1 r )n } and {n }, where
lim (n ) = 0 and
n

|(1 r )n | = |(1 r )|n < , since |1 r | < 1,
n=0 n=0
in view of Remark 3.7. Using Theorem 3.14,
lim n = 0.
n

Since |n | < and lim n = 0, appealing to Theorem 3.14 once again, it
n
n=0
follows that
n1

lim x1r (1 r ) nk1
(k )
n
k=0
n2

+x2 r 2 (1 r )nk2 (k ) + + xn r n (0 ) = 0.
k=0
Thus, taking limit as n in (4.7), we have,

lim n = x0 + xk r k1
,
n
k=1
where we note that the series on the right-hand side converges, since,

|xk r k1 | |xk | < ,
k=1 k=0
0 < r 1, the (E, r ) method being regular, using Theorem 3.12. In other words,
{z k } is (E, r )-summable to

x0 + xk r k1
,
k=1
The following result can be proved in a similar fashion.

Theorem 4.15 If |xk | < andyk is (E, r )-summable to , where the
k=0

k=0

(E, r ) method is regular, then z k is (E, r )-summable to x0 + xk r k1
,
k=0 k=1

n
zn = xk ynk , n = 0, 1, 2, . . . .
k=0
References
1. Defranza, J.: An ordered set of Nrlund means. Int. J. Math. Math. Sci. 4, 353364 (1981)
2. Natarajan, P.N.: New properties of the Natarajan method of summability. Comment. Math. Prace
Mat. 55, 915 (2015)
3. Natarajan, P.N.: Iteration product of Natarajan methods of summability. Int. J. Phys. Math. Sci.
3, 2529 (2013)
4. Natarajan, P.N.: On the (M, n ) method of summability. Analysis (Mnchen) 33, 5156 (2013)
5. Natarajan, P.N.: Cauchy multiplication of (M, n ) summable series. Adv. Dev. Math. Sci. 3,
3946 (2012)
6. Mears, F.M.: Some multiplication theorems for the Nrlund means. Bull. Amer. Math. Soc. 41,
875880 (1935)
Chapter 5
The SilvermanToeplitz, Schur,
and Steinhaus Theorems
for Four-Dimensional Infinite Matrices
In this chapter, we prove the SilvermanToeplitz, Schur and Steinhaus theorems for
four-dimensional infinite matrices. This chapter is divided into 3 sections. In the first
section, we introduce a new definition of convergence of a double sequence and a
double series and study some of its properties. In the second section, we prove the
SilvermanToeplitz theorem for four-dimensional infinite matrices, while, in the last
section, we establish the Schur and Steinhaus theorems for such matrices.
5.1 A New Definition of Convergence of a Double Sequence

and a Double Series
In this section, we introduce a new definition of limit of a double sequence and a dou-
ble series and record a few results on convergent double sequences. The motivation
for the introduction of this definition in classical analysis is the paper of Natarajan
and Srinivasan [1], in which the authors had already introduced this definition in the
context of a non-archimedean field. In this context, it is worthwhile to refer to the
papers of Kojima [2] and Robison [3]. For the concepts and theorems presented in
this chapter, one can refer to the papers of Natarajan [4, 5].
Definition 5.1 Let {xm,n } be a double sequence. We say that
lim xm,n = x,
m+n
if for every > 0, the set {(m, n) N2 : |xm,n x| } is finite, N being the set of
positive integers. In such a case, x is unique and x is called the limit of {xm,n }. We
also say that {xm,n } converges to x.

84 5 The SilvermanToeplitz, Schur, and Steinhaus Theorems
Definition 5.2 Let {xm,n } be a double sequence. We say that

,

xm,n = s,
m,n=0
if
lim sm,n = s,
m+n
where

m,n
sm,n = xk, , m, n = 0, 1, 2, . . . .
k,=0
,

In such a case, we say that the double series xm,n converges to s.
m,n=0
,

Definition 5.3 The double series xm,n is said to converge absolutely if
m,n=0
,

|xm,n | converges.
m,n=0
,

Remark 5.1 If xm,n converges absolutely, it converges. However, the converse
m,n=0
is false.
Remark 5.2 If lim xm,n = x, then the double sequence {xm,n } is automatically
m+n
bounded.
It is easy to prove the following results.
Theorem 5.1
lim xm,n = x
m+n
if and only if
(i) lim xm,n = x, n = 0, 1, 2, . . . ;
m
(ii) lim xm,n = x, m = 0, 1, 2, . . . ;
n
and
(iii) for any > 0, there exists N N such that
|xm,n x| < , m, n N ,
5.1 A New Definition of Convergence of a Double Sequence and a Double Series 85
which we write as
lim xm,n = x
m,n
(Note that this is Pringsheims definition of limit of a double sequence).

,

Theorem 5.2 If the double series xm,n converges, then
m,n=0
lim xm,n = 0.
m+n
However, the converse is not true.
5.2 The SilvermanToeplitz Theorem

for Four-Dimensional Infinite Matrices
In this section, we prove SilvermanToeplitz theorem for double sequences. We need

the following definition in the sequel.
Definition 5.4 Given the four-dimensional infinite matrix A = (am,n,k, ), m, n,

k, = 0, 1, 2, . . . and a double sequence {xk, }, k, = 0, 1, 2, . . . , by the
A-transform of x = {xk, }, we mean the double sequence A(x) = {(Ax)m,n }, where
,

(Ax)m,n = am,n,k, xk, , m, n = 0, 1, 2, . . . ,
k,=0
assuming that the series on the right converge. If lim (Ax)m,n = s, we say that
m+n
the double sequence x = {xk, } is A-summable or summable A to s, written as
xk, s(A).
If lim (Ax)m,n = s, whenever lim xk, = s, we say that the four-dimensional

m+n k+
infinite matrix A = (am,n,k, ) is regular.
We now prove the following important theorem.
Theorem 5.3 (SilvermanToeplitz) The four-dimensional infinite matrix A =

(am,n,k, ) is regular if and only if
,

sup |am,n,k, | < ; (5.1)
m,n0 k,=0
lim am,n,k, = 0, k, = 0, 1, 2, . . . ; (5.2)

m+n
,

lim am,n,k, = 1; (5.3)
m+n
k,=0

lim |am,n,k, | = 0, = 0, 1, 2, . . . ; (5.4)
m+n
k=0
and

lim |am,n,k, | = 0, k = 0, 1, 2, . . . . (5.5)
m+n
=0
Proof Proof of necessity part. Define the double sequence {xk, } as follows: For any
fixed p, q = 0, 1, 2, . . . , let

1, if k = p, = q;
xk, =
0, otherwise.
Then,
(Ax)m,n = am,n, p,q .
Since lim xk, = 0 and A is regular, it follows that (5.2) is necessary. Define the
k+
double sequence {xk, }, where xk, = 1, k, = 0, 1, 2, . . . . Now,
,

(Ax)m,n = am,n,k, , m, n = 0, 1, 2, . . . ,
k,=0
where the double series on the right converges. Since lim xk, = 1 and A is
k+
regular,
,

lim am,n,k, = 1,
m+n
k,=0
so that (5.3) is necessary. We now prove that

lim |am,n,k, | = 0, = 0, 1, 2, . . . .
m+n
k=0
5.2 The SilvermanToeplitz Theorem for Four-Dimensional Infinite Matrices 87
Suppose not. Then, there exists 0 N such that

lim |am,n,k,0 | = 0
m+n
k=0
does not hold. So, there exists an > 0, such that

(m, n) N :
2
|am,n,k,0 | > (5.6)
k=0
is infinite. Let us choose m 0 = n 0 = r0 = 1. Choose m 1 , n 1 N such that m 1 +n 1 >

m 0 + n0,
r0

|am 1 ,n 1 ,k,0 | < , using (5.2)
k=0
8
and

|am 1 ,n 1 ,k,0 | < , using (5.6).
k=0
We then choose r1 N such that r1 > r0 and

|am 1 ,n 1 ,k,0 | < , using (5.3).
k=r1 +1
8
Inductively, we choose m p + n p > m p1 + n p1 such that

r p1

|am p ,n p ,k,0 | < ; (5.7)
k=0
8

|am p ,n p ,k,0 | > (5.8)
k=0
and then choose r p > r p1 such that

|am p ,n p ,k,0 | < . (5.9)
k=r p +1
8
In view of (5.7)(5.9), we have,

rp

|am p ,n p ,k,0 | >
k=r p1 +1
8 8
3
= .
4
We can now find k p N, r p1 < k p r p such that
3
|am p ,n p ,k p ,0 | > .
4
Define the double sequence {xk, } as follows:

sgn(am p ,n p ,k p ,0 ), if = 0 , k = k p , p = 0, 1, 2, . . . ;
xk, =
0, if = 0 .
We first note that lim xk, = 0. Now,

k+
r p1 r p1

am p ,n p ,k,0 xk,0 |am p ,n p ,k,0 |

k=0 k=0

< , using (5.7);
8

a x |am p ,n p ,k,0 |
m p ,n p ,k,0 k,0
k=r p +1 k=r p +1

< , using (5.9);
8
and

rp

a x = |am ,n ,k , |
m ,n
p p ,k, 0 k, 0 p p p 0
k=r p1 +1
3
> , using (5.8).
4
Thus, using the above inequalities,

,

|(Ax)m p ,n p | = am p ,n p ,k, xk,

k,=0

= am p ,n p ,k,0 xk,0

k=0

rp

am p ,n p ,k,0 xk,0
k=r p1 +1
r p1

am p ,n p ,k,0 xk,0

k=0

am p ,n p ,k,0 xk,0
k=r p +1
3
>
4 8 8

= , p = 0, 1, 2, . . . .
2
Thus, lim (Ax)m,n = 0 does not hold, which is a contradiction. Thus, (5.4) is
m+n
necessary. The necessity of (5.5) follows in a similar fashion. To prove (5.1), we
shall suppose that (5.1) does not hold and arrive at a contradiction. Let 0 < < 1.
Choose m 0 = n 0 = 1. Using (5.2) and (5.3), choose m 1 + n 1 > m 0 + n 0 such that
0 +n 0
m
|am 1 ,n 1 ,k, | < 2, using (5.2);
k+=0

6
2
|am 1 ,n 1 ,k, | > ;
k+=0

and

|am 1 ,n 1 ,k, | < 22
k+=m 0 +n 0 +1
using (5.3), Theorems 5.1 and 5.2. It now follows that

|am 1 ,n 1 ,k, | < 22 .
k+=m 1 +n 1 +1
Choose m 2 + n 2 > m 1 + n 1 such that
1 +n 1
m
|am 2 ,n 2 ,k, | < 22 ;
k+=0

8
2
|am 2 ,n 2 ,k, | > ;
k+=0

and

|am 2 ,n 2 ,k, | < 24 .
k+=m 2 +n 2 +1
Inductively, choose m p + n p > m p1 + n p1 such that
m p1 +n p1

|am p ,n p ,k, | < 2 p1 ; (5.10)
k+=0

2 p+2
2
|am p ,n p ,k, | > ; (5.11)
k+=0

and

|am p ,n p ,k, | < 22 p2 . (5.12)
k+=m p +n p +1
Using (5.10)(5.12), we have,
m p +n p

|am p ,n p ,k, |
k+=m p1 +n p1 +1
2 p+2
2
> 22 p2 2 p1

2 p+2 2 p2 p1
2 2 2 1
, since >1

p1 p+3 p1
2 2 2
= 1

p1 p+3 p1 p1 p1
2 2 2 2 2
, since 1

p1 4 p1 p1
2 2 2 2
= 2

p1 4 p1 p1
2 2 2 2 2 2
> , since > 2

2 p1 3
2 2
= 1

2 p1
2 2
> [23 1], since > 2

2 p1
2
=7

2 p1
2
>4

2 p+1
2
= 2 p1

22 p+1 1
> , since > 1.
p
Thus, there exist k p , p , m p1 + n p1 < k p + p m p + n p such that
22 p+1
|am p ,n p ,k p , p | > , p = 0, 1, 2, . . . . (5.13)
p
Now, define the double sequence {xk, } as follows:

p sgn(am p ,n p ,k p , p ), if k = k p , = p , p = 0, 1, 2, . . . ;
xk, =
0, otherwise.
Note that lim xk, = 0. Now,

k+
,

|(Ax)m p .n p | = am p ,n p ,k, xk,

k,=0

m p +n p

am p ,n p ,k, xk,
k+=m p1 +n p1 +1
m p1 +n p1

am p ,n p ,k, xk,

k+=0

am p ,n p ,k, xk,
k+=m p +n p +1
m p1 +n p1

p |am p ,n p ,k p , p | |am p ,n p ,k, |
k+=0

|am p ,n p ,k, |
k+=m p +n p +1
22 p+1 p
> 22 p2 2 p1 , using (5.10), (5.12) and (5.13)
p
= 22 p+1 22 p2 2 p1
= 22 p2 (23 1) 2 p1
= 22 p2 (7) 2 p1
= 2 p1 [7.2 p1 1]
2 p1 [7.2 p1 2 p2 ]
= 2 p1 [2 p2 (14 1)]
= 2 p1 [13.2 p2 ]
= 13.22 p3 ,
i.e., |(Ax)m p ,n p | > 13.22 p3 , p = 0, 1, 2, . . . ,
i.e., lim (Ax)m,n = 0 does not hold, which is a contradiction. Thus, (5.1) is
m+n
necessary too.
Proof of the sufficiency part. Let lim xm,n = x. Then,
m+n
,

(Ax)m,n x = am,n,k, xk, x.
k,=0
Using (5.3), we have,

,

am,n,k, rm,n = 1,
k,=0
where
lim rm,n = 0. (5.14)

m+n
Hence,
,

(Ax)m,n x = am,n,k, (xk, x) + rm,n x.
k,=0
|am,n,k, | H, m, n, k, = 0, 1, 2, . . . , H > 0.
Given > 0, we can choose sufficiently large p and q such that

|xk, x| < . (5.15)
k+= p+q+1
5H
Let
|xk, x| L , k + = 0, 1, 2, . . . , L > 0.
We now choose N N such that, whenever m + n N , the following are satisfied:

p,q

|am,n,k, | < , using (5.2); (5.16)
k,=0
5L

|am,n,k, | < , = 0, 1, 2, . . . , q, using (5.4); (5.17)
k=0
5L

|am,n,k, | < , k = 0, 1, 2, . . . , p, using (5.5); (5.18)
=0
5L
and

|rm,n | < , using (5.14). (5.19)
5|x|
So, whenever m + n N , we thus have,

,

|(Ax)m,n x| = am,n,k, (xk, x) + rm,n x

k,=0
p,q

am,n,k, (xk, x)

k,=0

p,

+ am,n,k, (xk, x)
k=0,=q+1

,q

+ am,n,k, (xk, x)
k= p+1,=0

,

+ am,n,k, (xk, x)
k= p+1,=q+1
+ |rm,n ||x|

< L+ L+ L+ H+ |x|,
5L 5L 5L 5H 5|x|
using (5.15)(5.19)
= .
Consequently,
lim (Ax)m,n = x.
m+n
Thus, A is regular, completing the proof of the theorem.

5.3 The Schur and Steinhaus Theorems for

Four-Dimensional Infinite Matrices
Let cds ,
ds , respectively, denote the spaces of convergent double sequences and
bounded double sequences.
Definition 5.5 A is called Schur matrix if {(Ax)m,n } cds , whenever x = {xk, }

ds .
In this section, we obtain necessary and sufficient conditions for A = (am,n,k, )

to be a Schur matrix and then deduce Steinhaus theorem.
Definition 5.6 The double sequence {xm,n } is called a Cauchy sequence if for every
> 0, there exists N N (the set of all positive integers) such that the set
{(m, n), (k, ) N2 : |xm,n xk, | , m, n, k, N }
is finite.
It is now easy to prove the following result.
Theorem 5.4 The double sequence {xm,n } is Cauchy if and only if
lim |xm+u,n xm,n | = 0, u = 0, 1, 2, . . . ; (5.20)

m+n
5.3 The Schur and Steinhaus Theorems for Four-Dimensional Infinite Matrices 95
and
lim |xm,n+v xm,n | = 0, v = 0, 1, 2, . . . . (5.21)

m+n
Definition 5.7 If every Cauchy double sequence of a normed linear space X

converges to an element of X , X is said to be double sequence complete or
ds-complete.
Note that R (the set of all real numbers) and C (the set of all complex numbers)
are ds-complete.
For x = {xm,n }
ds , define

x
= sup |xm,n |.
m,n0
One can easily prove that

ds is a normed linear space which is ds-complete. With
the same definition of norm for elements of cds , cds is a closed subspace of
ds .
Theorem 5.5 (Schur) The necessary and sufficient conditions for a four-dimensional
infinite matrix A = (am,n,k, ) to be a Schur matrix, i.e., {(Ax)m,n } cds , whenever
x = {xk, }
ds are as follows:
,

|am,n,k, | < , m, n = 0, 1, 2, . . . ; (5.22)
k,=0
,

lim |am+u,n,k, am,n,k, | = 0, u = 0, 1, 2, . . . ; (5.23)
m+n
k,=0
and
,

lim |am,n+v,k, am,n,k, | = 0, v = 0, 1, 2, . . . . (5.24)
m+n
k,=0
Proof Proof of the sufficiency part. Let (5.22)(5.24) hold and x = {xk, }
ds .
First, we note that in view of (5.22),
,

(Ax)m,n = am,n,k, xk, , m, n = 0, 1, 2, . . .
k,=0
is defined, the double series on the right being convergent. Now, for n = 0, 1, 2, . . . ,
,

|(Ax)m+u,n (Ax)m,n | = (am+u,n,k, am,n,k, )xk,

k,=0
,

M |am+u,n,k, am,n,k, |
k,=0
0, m + n , using (5.23),
where |xk, | M, k, = 0, 1, 2, . . . , M > 0. Similarly, it follows that
|(Ax)m,n+v (Ax)m,n | 0, m + n , v = 0, 1, 2, . . . ,
using (5.24). Thus, {(Ax)m,n } is a Cauchy double sequence. Since R (or C) is

ds-complete, {(Ax)m,n } converges, i.e., {(Ax)m,n } cds , completing the sufficiency
part of the proof.
Proof of the necessity part. Let A be a Schur matrix. For m, n = 0, 1, 2, . . . ,
consider the double sequence {xk, }, where xk, = sgn(am,n,k, ), k, = 0, 1, 2, . . . .
Then, {xk, }
ds so that, by hypothesis,
,

(Ax)m,n = |am,n,k, |, m, n = 0, 1, 2, . . .
k,=0
is defined. Since the series on the right converges, (5.22) holds. Suppose (5.23) does
not hold. So, there exist 0 , u 0 N such that

lim |am+u 0 ,n,k,0 am,n,k,0 | = 0
m+n
k=0
does not hold. So, there exists > 0 such that the set

(m, n) N :2
|am+u 0 ,n,k,0 am,n,k,0 | > 2
k=0
is infinite. Thus, we can choose pairs of integers m p , n p N such that m 1 + n 1 <

m 2 + n 2 < < m p + n p < . . . and

|am p +u 0 ,n p ,k,0 am p ,n p ,k,0 | > 2, p = 1, 2, . . . . (5.25)
k=0

|am 1 +u 0 ,n 1 ,k,0 am 1 ,n 1 ,k,0 | < .
k=0
Consequently, there exists r1 N such that

|am 1 +u 0 ,n 1 ,k,0 am 1 ,n 1 ,k,0 | < . (5.26)
k=r1
4
In view of (5.25) and (5.26), we have,
1 1
r
7
|am 1 +u 0 ,n 1 ,k,0 am 1 ,n 1 ,k,0 | > > .
k=0
4
By hypothesis, (5.2) holds so that we can suppose that
1 1
r

|am 2 +u 0 ,n 2 ,k,0 am 2 ,n 2 ,k,0 | < . (5.27)
k=0
4

|am 2 +u 0 ,n 2 ,k,0 am 2 ,n 2 ,k,0 | > 2. (5.28)
k=0
Using (5.1),

|am 2 +u 0 ,n 2 ,k,0 am 2 ,n 2 ,k,0 | < ,
k=0
so that there exists r2 N, r2 > r1 such that

|am 2 +u 0 ,n 2 ,k,0 am 2 ,n 2 ,k,0 | < . (5.29)
k=r2
4
From (5.27)(5.29), we have,
2 1
r
3
|am 2 +u 0 ,n 2 ,k,0 am 2 ,n 2 ,k,0 | > > .
k=r1
2
Inductively, we can choose a strictly increasing sequence {r p } of positive integers

such that
r p1 1

|am p +u 0 ,n p ,k,0 am p ,n p ,k,0 | < ; (5.30)
k=0
4

|am p +u 0 ,n p ,k,0 am p ,n p ,k,0 | < ; (5.31)
k=r p
4
and
r p 1

|am p +u 0 ,n p ,k,0 am p ,n p ,k,0 | > . (5.32)
k=r p1
Now, define {xk, }

ds , where

sgn(am p +u 0 ,n p ,k,0 am p ,n p ,k,0 ), if = 0 , r p1 k < r p , p = 1, 2, . . . ;
xk, =
0, otherwise.
Then,
,

(Ax)m p +u 0 ,n p (Ax)m p ,n p = (am p +u 0 ,n p ,k, am p ,n p ,k, )xk,
k,=0

= (am p +u 0 ,n p ,k,0 am p ,n p ,k,0 )xk,0
k=0
r p1 1

k=0
r p 1

+ (am p +u 0 ,n p ,k,0 am p ,n p ,k,0 )xk,0
k=r p1

k=r p
r p1 1

k=0
r p 1

+ |am p +u 0 ,n p ,k,0 am p ,n p ,k,0 |
k=r p1

k=r p
so that
r p 1

|am p +u 0 ,n p ,k,0 am p ,n p ,k,0 |
k=r p1
= {(Ax)m p +u 0 ,n p (Ax)m p ,n p }
r p1 1

(am p +u 0 ,n p ,k,0 am p ,n p ,k,0 )xk,0
k=0

(am p +u 0 ,n p ,k,0 am p ,n p ,k,0 )xk,0 .
k=r p
In view of (5.30)(5.32), we have,
r p 1

< |am p +u 0 ,n p ,k,0 am p ,n p ,k,0 |
k=r p1

(Ax)m p +u 0 ,n p (Ax)m p ,n p + + ,
4 4
from which it follows that

(Ax)m +u ,n (Ax)m ,n > , p = 1, 2, . . . .
p 0 p p p
2
Consequently,
{(Ax)m,n }
/ cds ,
which is a contradiction. Thus, (5.23) holds. Similarly, (5.24) holds too. This com-
pletes the proof of the theorem.

We now deduce the following result.
Theorem 5.6 (Steinhaus) A four-dimensional infinite matrix A = (am,n,k, ) cannot
be both a regular and a Schur matrix, i.e., given a four-dimensional regular matrix
A, there exists a bounded, divergent double sequence which is not A-summable.
Proof Since A is regular, (5.2) and (5.3) hold. If A were a Schur matrix too, then
{am,n,k, },
m,n=0 is uniformly Cauchy with respect to k, = 0, 1, 2, . . . . Since R
(or C) is ds-complete, {am,n,k, },
m,n=0 converges uniformly to 0 with respect to k,
= 0, 1, 2, . . . . Consequently, we have,
,

lim am,n,k, = 0,
m+n
k,=0
a contradiction of (5.3), completing the proof.

References
1. Natarajan, P.N., Srinivasan, V.: Silverman-Toeplitz theorem for double sequences and series and
its application to Nrlund means in non-archimedean fields. Ann. Math. Blaise Pascal 9, 85100
(2002)
2. Kojima, T.: On the theory of double sequences. Thoku Math. J. 21, 314 (1922)
3. Robison, G.M.: Divergent double sequences and series. Trans. Amer. Math. Soc. 28, 5073
(1926)
4. Natarajan, P.N.: A new definition of convergence of a double sequence and a double series and
Silverman-Toeplitz theorem. Comment. Math. Prace Mat. 54, 129139 (2014)
5. Natarajan, P.N.: The Schur and Steinhaus theorems for 4-dimensional infinite matrices.
Comment. Math. Prace Mat. 54, 159165 (2014)
Chapter 6
The Nrlund, Weighted Mean, and
(M, m,n ) Methods for Double Sequences
The present chapter is devoted to a study of the Nrlund, Weighted Mean, and
(M, m,n ) or Natarajan methods for double sequences. This chapter is divided into 3
sections. In the first section, the Nrlund method for double sequences is introduced
and its properties are studied in the context of the new definition of convergence of a
double sequence introduced earlier in Chap. 5. In the second section, we introduce the
Weighted Mean method for double sequences and study its properties in the context of
the new definition. In the final section, the (M, m,n ) method or the Natarajan method
for double sequences is introduced and its properties are studied in the context of the
new definition.
6.1 Nrlund Method for Double Sequences
We now introduce Nrlund methods for double sequences and double series and
study them in the context of Definitions 5.1 and 5.2 (see [1]).
Definition 6.1 Given a doubly infinite set of real numbers pm,n , m, n = 0, 1, 2, . . . ,

where
pm,n 0, (m, n) = (0, 0) and p0,0 > 0,
let

m,n
Pm,n = pi, j , m, n = 0, 1, 2, . . . .
i, j=0
For any double sequence {sm,n }, we define

102 6 The Nrlund, Weighted Mean, and (M, m,n ) Methods for Double Sequences
m,n = (N , pm,n )({sm,n })

Sm,n
=
Pm,n

1
m,n
= pmi,n j si, j , m, n = 0, 1, 2, . . . .
Pm,n i, j=0
If lim m,n = , we say that the double sequence {sm,n } is summable (N , pm,n )
m+n
to , written as
sm,n (N , pm,n ).
(N , pm,n ) is called a Nrlund method (or Nrlund mean).

,

Definition 6.2 A double series u m,n is said to be (N , pm,n ) summable to , if
m,n=0
the double sequence {sm,n }, where

m,n
sm,n = u i, j , m, n = 0, 1, 2, . . . ,
i, j=0
is (N , pm,n ) summable to .
Definition 6.3 The Nrlund methods (N , pm,n ) and (N , qm,n ) are said to be consis-
tent if
sm,n (N , pm,n )
and
sm,n (N , qm,n )
imply that = .
Definition 6.4 We say that (N , pm,n ) is included in (N , qm,n ) (or (N , qm,n ) includes
(N , pm,n )), written as
(N , pm,n ) (N , qm,n )
if
sm,n (N , pm,n ) implies that sm,n (N , qm,n ).
The two Nrlund methods (N , pm,n ), (N , qm,n ) are said to be equivalent if
(N , pm,n ) (N , qm,n ) and vice versa.
In view of Theorem 5.3, it is easy to prove the following result.

6.1 Nrlund Method for Double Sequences 103
Theorem 6.1 The Nrlund method (N , pm,n ) is regular if and only if

n
lim pmi,n j = 0, i = 0, 1, 2, . . . , m; (6.1)
m+n
j=0
and

m
lim pmi,n j = 0, j = 0, 1, 2, . . . , n. (6.2)
m+n
i=0
In the sequel, let (N , pm,n ), (N , qm,n ) be regular Nrlund methods such that each
row and each column of the two-dimensional infinite matrices ( pm,n ), (qm,n ) is a
regular Nrlund method for simple sequences.
Theorem 6.2 Any two such regular Nrlund methods are consistent.
Proof Given regular Nrlund methods (N , pm,n ), (N , qm,n ), where each row and
each column of the two-dimensional infinite matrices ( pm,n ), (qm,n ) is a regular
Nrlund method for simple sequences, we define a third method (N , rm,n ) by the
equation

m,n
rm,n = pi, j qmi,n j , m, n = 0, 1, 2, . . . .
i, j=0
Then, for s = {sm,n }, we can check that

,

(N , rm,n )(s) = m,n,i, j (N , qi, j )(s),
i, j=0
where

pmi,n j Q i, j
, if 0 i m and 0 j n;

m,n
m,n,i, j = pm,n Q ,

,=0

0, otherwise.
Using the fact that p0,0 > 0, q0,0 > 0, pm,n 0, qm,n 0, (m, n) = (0, 0)
and the fact that each row and each column of the two-dimensional infinite matrices
( pm,n ), (qm,n ) is a regular Nrlund method for simple sequences, it follows that

n
lim m,n,i, j = 0, i = 0, 1, 2, . . . , m
m+n
j=0
and

m
lim m,n,i, j = 0, j = 0, 1, 2, . . . , n.
m+n
i=0
Appealing to Theorem 5.3, the four-dimensional infinite matrix (m,n,k, ) is regular.

Consequently,
sm,n (N , qm,n ) implies that sm,n (N , rm,n ).
Similarly, it can be shown that
sm,n (N , pm,n ) implies that sm,n (N , rm,n ).
So = . Thus, the regular Nrlund methods (N , pm,n ) and (N , qm,n ) are consistent,
Let (N , pm,n ), (N , qm,n ) be regular Nrlund methods. Then,

P(x, y) = Pm,n x m y n ,

Q(x, y) = Q m,n x m y n ,

p(x, y) = pm,n x m y n ,

q(x, y) = qm,n x m y n
all converge for |x|, |y| < 1. Also,

q(x, y) Q(x, y)
k(x, y) = km,n x m y n = = ,
p(x, y) P(x, y)
p(x, y) P(x, y)
(x, y) = m,n x m y n = =
q(x, y) Q(x, y)
converge for |x|, |y| < 1.

Further, we have

m,n
ki, j pmi,n j = qm,n ;
i, j=0

m,n
ki, j Pmi,n j = Q m,n ;
i, j=0

m,n
i, j qmi,n j = pm,n ;
i, j=0
and

m,n
i, j Q mi,n j = Pm,n .
i, j=0
We now have
Theorem 6.3 Let (N , pm,n ), (N , qm,n ) be regular Nrlund methods. Then
(N , pm,n ) (N , qm,n )
if and only if

m,n
|ki, j |Pmi,n j
i, j=0
= O(1), m + n ,
Q m,n
i.e.,

m,n
|ki, j |Pmi,n j H Q m,n , m, n = 0, 1, 2, . . . ,
i, j=0
H > 0 is independent of m, n; (6.3)

n
|ki, j |Pmi,n j
j=0
lim = 0, 0 i m; (6.4)
m+n Q m,n
and

m
|ki, j |Pmi,n j
i=0
lim = 0, 0 j n. (6.5)
m+n Q m,n
Proof Let
s(x, y) = sm,n x m y n .
Then, for small x and y,


m,n
(q)
Q m,n Nm,n (s)x m y n = qmi,n j si, j x m y n
i, j=0
= q(x, y)s(x, y);
Similarly,

( p)
Pm,n Nm,n (s)x m y n = p(x, y)s(x, y).
Thus,

(q)
Q m,n Nm,n (s)x m y n q(x, y)
=
( p)
Pm,n Nm,n (s)x m y n p(x, y)
= k(x, y),

(q) ( p)
i.e., Q m,n Nm,n (s)x m y n = k(x, y) Pm,n Nm,n (s)x m y n

( p)
= km,n x m y n Pm,n Nm,n (s)x m y n
and so

m,n
( p)
(q)
Q m,n Nm,n (s) = kmi,n j Pi, j Ni, j (s), m, n = 0, 1, 2, . . . .
i, j=0
Consequently,
,

(q) ( p)
Nm,n (s) = cm,n,i, j Ni, j (s),
i, j=0
where the four-dimensional infinite matrix (cm,n,i, j ) is defined by

k
mi,n j
, if 0 i m and 0 j n;
cm,n,i, j = Q m,n
0, otherwise.
Now,
(N , pm,n ) (N , qm,n )
if and only if the four-dimensional infinite matrix (cm,n,i, j ) is regular. Rest of the
proof follows using Theorem 5.3. This completes the proof of the theorem.
Theorem 6.4 The regular Nrlund methods (N , pm,n ), (N , qm,n ) are equivalent if
and only if
,
,

|km,n | < and |m,n | < . (6.6)
m,n=0 m,n=0
Proof Necessity part. Let (N , pm,n ), (N , qm,n ) be equivalent. Note that k0,0 > 0 and
0,0 > 0, since p0,0 > 0 and q0.0 > 0. Since (N , pm,n ) (N , qm,n ), in view of (6.3),
k0,0 Pm,n H Q m,n , m, n = 0, 1, 2, . . . ,

so that QPm,n
m,n
is bounded. Similarly, (N , qm,n ) (N , pm,n ) implies that QPm,n
m,n
is
bounded too. Using (6.3) again, we have
,

|ki, j | H L ,
i, j=0
,

m,n
where QPm,n L, m, n = 0, 1, 2, . . . . Thus, |km,n | < . Similarly,
m,n=0
,

|m,n | < .
m,n=0
,
,

Sufficiency part. Let (6.6) hold. |km,n | < and |m,n | < in
m,n=0 m,n=0
,
,

the sense of Definition 5.2 implies |km,n | < and |m,n | < in
m,n=0 m,n=0
Pringsheims sense, in view of Theorem 5.1. It now follows that the regular Nr-
lund methods (N , pm,n ) and (N , qm,n ) are equivalent, using the sufficiency part of
Theorem III of [2]. The proof of the theorem is now complete.
6.2 Weighted Mean Method for Double Sequences
We introduce Weighted Mean method for double sequences and double series and
study them in the context of the new definition of convergence of a double sequence
and a double series introduced earlier (see Definitions 5.1 and 5.2). For details about
this study, one can refer to [3, 4].
Definition 6.5 Given a doubly infinite set of real number pm,n , m, n = 0, 1, 2, . . . ,

where pm,n 0, (m, n) = (0, 0) and p0,0 > 0, let

m,n
Pm,n = pi, j , m, n = 0, 1, 2, . . . .
i, j=0
For any double sequence {sm,n }, define
m,n = (N , pm,n )({sm,n })

Sm,n
=
Pm,n

1
m,n
= pi, j si, j , m, n = 0, 1, 2, . . . .
Pm,n i, j=0
If lim m,n = , we say that the double sequence {sm,n } is summable (N , pm,n )
m+n
or (N , pm,n )-summable to , written as
sm,n (N , pm,n ).
(N , pm,n ) is called a Weighted Mean method.
Note that the (N , pm,n ) method is defined by the four-dimensional infinite matrix
A = (am,n,k, ), where
p
k,
, if 0 k m and 0 n;
am,n,k, = Pm,n
0, otherwise,
where pm,n 0, (m, n) = (0, 0), p0,0 > 0 and

m,n
Pm,n = pi, j , m, n = 0, 1, 2, . . . .
i, j=0
,

Definition 6.6 A double series u m,n is said to be (N , pm,n )-summable to if
m,n=0
the double sequence {sm,n }, where

m,n
sm,n = u i, j , m, n = 0, 1, 2, . . . ,
i, j=0
is (N , pm,n )-summable to .
Definition 6.7 We say that (N , pm,n ) is included in (N , qm,n ) (or (N , qm,n ) includes
(N , pm,n )), written as
6.2 Weighted Mean Method for Double Sequences 109
(N , pm,n ) (N , qm,n ) (or (N , qm,n ) (N , pm,n ))
if
sm,n (N , pm,n ) implies that sm,n (N , qm,n ).
In view of Theorem 5.3, we have the following result.
Theorem 6.5 The Weighted Mean method (N , pm,n ) is regular if and only if
lim Pm,n = ; (6.7)

m+n

m
pk,
k=0
lim = 0, = 0, 1, 2, . . . ; (6.8)
m+n Pm,n
and

n
pk,
=0
lim = 0, k = 0, 1, 2, . . . . (6.9)
m+n Pm,n
Proof Proof of the necessity part. Let (N , pm,n ) be regular. Using (5.2),
lim am,n,0,0 = 0,
m+n
p0,0
i.e., lim = 0,
m+n Pm,n
i.e., lim Pm,n = ,
m+n
since p0,0 = 0.
Using (5.4),

lim |am,n,k, | = 0, = 0, 1, 2, . . . ,
m+n
k=0
m
pk,
i.e., lim = 0, = 0, 1, 2, . . . ,
m+n P
k=0 m,n

m
pk,
k=0
i.e., lim = 0, = 0, 1, 2, . . . ,
m+n Pm,n
proving (6.8). Proof of (6.9) is similar.

Proof of the sufficiency part. Let (6.7)(6.9) hold. For every fixed k, =
0, 1, 2, . . . ,
pk,
lim am,n,k, = lim
m+n m+n Pm,n
= 0,
in view of (6.7) so that (5.2) holds. Now,

m,n
pk,
,
k,=0
lim am,n,k, = lim
m+n
k,=0
m+n Pm,n
Pm,n
= lim
m+n Pm,n
= 1,
proving (5.3). Also, for every fixed = 0, 1, 2, . . . ,

m
lim |am,n,k, | = lim |am,n,k, |
m+n m+n
k=0 k=0
m
pk,
k=0
= lim
m+n Pm,n
= 0, using (6.8).
This proves (5.4). Proof of (5.5) is similar. Now,

, m,n
k,=0 pk,
|am,n,k, | =
k,=0
Pm,n
Pm,n
=
Pm,n
= 1, m, n = 0, 1, 2, . . . ,
,

sup |am,n,k, | < .
m,n0 k,=0
Thus, (5.1) holds. So, by Theorem 5.3, (N , pm,n ) is regular, completing the proof of
the theorem.
Theorem 6.6 (Limitation theorem) If {sm,n } is (N , pm,n )-summable to s, then

Pm,n
|sm,n s| = o , m + n , (6.10)
pm,n
in the sense that pm,n

(sm,n s) 0, m + n .
Pm,n
Proof Let {tm,n } be the (N , pm,n )-transform of {sm,n }. Now,

pm,n pm,n sm,n pm,n s

P (sm,n s) = Pm,n
m,n

Pm,n tm,n Pm,n1 tm,n1 Pm1,n tm1,n + Pm1,n1 tm1,n1

(Pm,n Pm,n1 Pm1,n + Pm1,n1 )s
=

Pm,n

Pm,n1 Pm1,n
= (tm,n s) (tm,n1 s) (tm1,n s)
Pm,n Pm,n

Pm1,n1
+ (tm1,n1 s)
Pm,n
|tm,n s| + |tm,n1 s| + |tm1,n s| + |tm1,n1 s|,
Pm,n1 Pm1,n Pm1,n1
since , , 1
Pm,n Pm,n Pm,n
0, m + n , since lim tm,n = s.
m+n
So,
pm,n
lim (sm,n s) = 0,
m+n P m,n

Pm,n
i.e., |sm,n s| = o , m + n ,
pm,n
We now prove a few inclusion theorems involving Weighted Mean methods for
double sequences.
Theorem 6.7 (Comparison theorem for two Weighted Mean methods for double
sequences) If (N , pm,n ), (N , qm,n ) are two Weighted Mean methods such that
,
qm,n
< ; (6.11)
m,n=0
pm,n
and
Pm,n = o(Q m,n ), m + n , (6.12)
in the sense that

Pm,n
0, m + n ,
Q m,n
where

m,n
m,n
Pm,n = pk, , Q m,n = qk, , m, n = 0, 1, 2, . . . ,
k,=0 k,=0
then,
(N , pm,n ) (N , qm,n ).
Proof Given a double sequence {sm,n }, let

m,n
pi, j si, j
i, j=0
tm,n = ;
Pm,n

m,n
qi, j si, j
i, j=0
u m,n = , m, n = 0, 1, 2, . . . .
Q m,n
Then,
p0,0 s0,0 = P0,0 t0,0 ;
pm,n sm,n = Pm,n tm,n Pm1,n tm1,n Pm,n1 tm,n1 + Pm1,n1 tm1,n1 ,
so that
Pm,n tm,n Pm1,n tm1,n Pm,n1 tm,n1 + Pm1,n1 tm1,n1
sm,n = ,
pm,n
Pm,0 tm,0 Pm1,0 tm1,0
sm,0 = ,
pm,0
P0,n t0,n P0,n1 t0,n1
s0,n = ,
p0,n
where we suppose that
P1,n = 0, Pm,1 = 0, P1,1 = 0.
Now,

1
m,n
u m,n = qi, j si, j
Q m,n i, j=0

Pi, j ti, j Pi1, j ti1, j Pi, j1 ti, j1

1
m,n
+Pi1, j1 ti1, j1
= qi, j
Q m,n i, j=0
pi, j

,

= cm,n,k, tk, ,
k,=0
where

qk, qk,+1 qk+1, qk+1,+1 Pk,
pk, pk,+1 pk+1, +

pk+1,+1 Q m,n
, if k < m, < n;

qk, qk,+1 Pk,
pk, pk,+1 Q m,n ,
if k = m, < n;
cm,n,k, = qk, qk+1, Pk,
pk, pk+1, Q m,n ,

if k < m, = n;

qk, Pk,
, if k = m, = n;

pk, Q k,
0, if k > m or > n.
In view of (6.11) and (6.12), we have
lim cm,n,k, = 0, k, = 0, 1, 2, . . . .
m+n
If sm,n = 1, m, n = 0, 1, 2, . . . , then

m,n
pi, j
i, j=0
tm,n = = 1,
Pm,n
noting that Pm,n = 0, m, n = 0, 1, 2, . . . . Similarly, u m,n = 1, m, n = 0, 1, 2, . . . ,

so that
,

u m,n = cm,n,k, tk,
k,=0
yields
,

1= cm,n,k, (1).
k,=0
Thus,
,

cm,n,k, = 1, m, n = 0, 1, 2, . . .
k,=0
and consequently,
,

lim cm,n,k, = 1.
m+n
k,=0
,

m,n
|cm,n,k, | = |cm,n,k, |
k,=0 k,=0
qk,
m1,n1
qk,+1 qk+1,

qk+1,+1 Pk,

=
p p
pk+1,
+
pk+1,+1 Q m,n
k,=0 k, k,+1
n1

qm, qm,+1 Pk,
+
p pm,+1 Q m,n
=0 m,

qk,n
m1
qk+1,n Pk,n
+
p p Q
k=0 k,n k+1,n m,n

qm,n Pm,n
+
pm,n Q m,n
,

Pm,n qk,
4
Q m,n k,=0 pk,
,
Pm,n qk,
+2
Q m,n k,=0 pk,
,
Pm,n qk,
+
Q m,n k,=0 pk,
,
Pm,n qk,
=7
Q m,n k,=0 pk,
qk,
,
7M , m, n = 0, 1, 2, . . . ,
k,=0
pk,
Pm,n
where Q m,n
M, m, n = 0, 1, 2, . . . , in view of (6.12). Thus,
,

sup |cm,n,k, | < .
m,n0 k,=0
We will now prove that for each fixed = 0, 1, 2, . . . ,

lim |cm,n,k, | = 0.
m+n
k=0
For such a fixed , we consider three cases, viz., < n, = n and > n.
Case 1. When < n,

m1
|cm,n,k, | = |cm,n,k, | + |cm,n,m, |
k=0 k=0

qk,
m1
qk,+1 qk+1,

qk+1,+1 Pk,

=
p p
pk+1,
+
pk+1,+1 Q m,n
k=0 k, k,+1

qm, qm,+1 Pm,
+
pm, pm,+1 Q m,n
,
Pm,n qk,
6
Q m,n k,=0 pk,
0, m + n , in view of (6.11) and (6.12),
proving that

lim |cm,n,k, | = 0, = 0, 1, 2, . . .
m+n
k=0
in this case.
Case 2. The case = n can be proved similarly.
Case 3. If > n, then cm,n,k, = 0, m, n = 0, 1, 2, . . . and so

lim |cm,n,k, | = 0, = 0, 1, 2, . . .
m+n
k=0
in this case too. Consequently,

lim |cm,n,k, | = 0, = 0, 1, 2, . . .
m+n
k=0
in all cases. Similarly, we can prove that

lim |cm,n,k, | = 0, k = 0, 1, 2, . . . .
m+n
=0
Appealing to Theorem 5.3, it follows that the four-dimensional infinite matrix

(cm,n,k, ) is regular and so
(N , pm,n ) (N , qm,n ),
Following the proof of Theorem 6.7, we can prove
Theorem 6.8 If (N , pm,n ), (N , qm,n ) are two Weighted Mean methods such that
qm,n = O( pm,n ), m + n , (6.13)
i.e., there exists H > 0 such that

qm,n
H, m, n = 0, 1, 2, . . .
pm,n
and
,
Pm,n
< , (6.14)
m,n=0
Q m,n
then
(N , pm,n ) (N , qm,n ).
We now prove another inclusion theorem.
Theorem 6.9 (Comparison theorem for a (N , pm,n ) method and a regular matrix
method) Let (N , pm,n ) be a Weighted Mean method and A = (am,n,k, ) be a regular
four-dimensional infinite matrix method. If
,
Pk,
sup |am,n,k, | < ; (6.15)
m,n0 k,=0 pk,

Pk,
lim |am,n,k, | = 0, = 0, 1, 2, . . . ; (6.16)
m+n
k=0
pk,
and

Pk,
lim |am,n,k, | = 0, k = 0, 1, 2, . . . , (6.17)
m+n
=0
pk,
then,
(N , pm,n ) A.
Proof Let {sm,n } be a double sequence and {tm,n }, {m,n } be its (N , pm,n ) and A-
transforms, respectively. Then,
m,n
i, j=0 pi, j si, j
tm,n = ,
Pm,n
,

m,n = am,n,k, sk, , m, n = 0, 1, 2, . . . .
k,=0
Now,
Pm,n tm,n Pm1,n tm1,n Pm,n1 tm,n1 + Pm1,n1 tm1,n1

sm,n =
pm,n
m, n = 0, 1, 2, . . . , where we suppose that
t1,n = tm,1 = t1,1 = 0.
We now have
,

m,n = am,n,k, sk,
k,=0
,
am,n,k,
= Pk, tk, Pk1, tk1, Pk,1 tk,1 + Pk1,1 tk1,1
k,=0
pk,
am,n,k,
,
am,n,k+1, am,n,k,+1 am,n,k+1,+1

= + Pk, tk, .
k,=0
pk, pk+1, pk,+1 pk+1,+1
Let lim tm,n = s. Since {tm,n } converges, it is bounded and so |tm,n | M,

m+n
m, n = 0, 1, 2, . . . , M > 0. Now,
,
,

Pk, Pk,
|am,n,k, | |tk, | M |am,n,k, |
k,=0
pk, k,=0
pk,
< , in view of (6.15);
,
,

Pk, Pk+1,
|am,n,k+1, | |tk, | M |am,n,k+1, | ,
k,=0
pk+1, k,=0
pk+1,
since Pk, Pk+1,
< , in view of (6.15);
Similarly,
,
Pk,
|am,n,k,+1 | |tk, | <
k,=0
pk,+1
and
,
Pk,
|am,n,k+1,+1 | |tk, | < ,
k,=0
pk+1,+1
in view of (6.15). We now write

,

m,n = bm,n,k, tk, , m, n = 0, 1, 2, . . . ,
k,=0
where

am,n,k, am,n,k+1, am,n,k,+1 am,n,k+1,+1
bm,n,k, = + Pk, ,
pk, pk+1, pk,+1 pk+1,+1
m, n, k, = 0, 1, 2, . . . . Now,
,
,
Pk,
|bm,n,k, | 4 sup |am,n,k, |
k,=0 m,n0 k,=0 pk,
< ,
m, n = 0, 1, 2, . . . , in view of (6.15), so that

,

sup |bm,n,k, | < .
m,n0 k,=0
Since A is regular,
lim am,n,k, = 0, k, = 0, 1, 2, . . .
m+n
in view of Theorem 5.3, from which it follows that
lim bm,n,k, = 0, k, = 0, 1, 2, . . . .
m+n
Let sm,n = 1, m, n = 0, 1, 2, . . . . Then, tm,n = 1, m, n = 0, 1, 2, . . . . Now,

,
,

bm,n,k, = am,n,k, , m, n = 0, 1, 2, . . . ,
k,=0 k,=0
so that
,
,

lim bm,n,k, = lim am,n,k,
m+n m+n
k,=0 k,=0
= 1,
since A is regular, again appealing to Theorem 5.3. Also, for every fixed =
0, 1, 2, . . . ,

lim |bm,n,k, |
m+n
k=0

am,n,k, am,n,k+1, am,n,k,+1 am,n,k+1,+1
= lim + Pk,
m+n p pk+1, pk,+1 pk+1,+1
k=0 k,

Pk, Pk,
lim |am,n,k, | + |am,n,k+1, |
m+n
k=0
p k, k=0
p k+1,

Pk, Pk,
+ |am,n,k,+1 | + |am,n,k+1,+1 |
k=0
pk,+1 k=0 pk+1,+1

Pk,
4 lim |am,n,k, |
m+n
k=0
pk,
= 0, using (6.16).
Thus,

lim |bm,n,k, | = 0, = 0, 1, 2, . . . .
m+n
k=0
Similarly, we can prove that

lim |bm,n,k, | = 0, k = 0, 1, 2, . . . ,
m+n
=0
using (6.17). Consequently, the four-dimensional infinite matrix (bm,n,k, ) is regular,

again using Theorem 5.3. So lim tk, = s implies that lim m,n = s. In other
k+ m+n
words,
(N , pm,n ) A.
The proof of the theorem is now complete.
Corollary 6.1 If we choose a regular Weighted Mean method (N , qm,n ) for A, we

get the corresponding sufficient conditions, using (6.15)(6.17), for
(N , pm,n ) (N , qm,n ).
6.3 (M, m,n ) or Natarajan Method for Double Sequences
In this section, we introduce the (M, m,n ) method or the Natarajan method for double
sequences and extend some of the results of the (M, n ) method (or the Natarajan
method) for simple sequences presented in Chap. 5 (for details, see [5]).
,

Definition 6.8 Let {m,n } be a double sequence such that |m,n | < . The
m,n=0
(M, m,n ) method is defined by the four-dimensional infinite matrix (am,n,k, ), where

mk,n , k m, n;
am,n,k, =
0, otherwise.
Definition 6.9 The methods (M, m,n ), (M, m,n ) are said to be consistent, if
sk, (M, m,n ) and sk, (M, m,n ) imply that = .
Definition 6.10 We say that (M, m,n ) is included in (M, m,n ) (or (M, m,n )
includes (M, m,n )), written as
(M, m,n ) (M, m,n ) (or (M, m,n ) (M, m,n ))
if
sk, (M, m,n ) implies that sk, (M, m,n ) too.
The methods (M, m,n ), (M, m,n ) are said to be equivalent if
(M, m,n ) (M, m,n ) and vice versa.
It is easy to prove the following result.

6.3 (M, m,n ) or Natarajan Method for Double Sequences 121
Theorem 6.10 The method (M, m,n ) is regular if and only if

,

m,n = 1.
m,n=0
In the sequel, let (M, m,n ), (M, m,n ) be regular methods such that each row
and each column of the two-dimensional infinite matrices (m,n ), (m,n ) is a regular
Natarajan method for simple sequences.
Theorem 6.11 Any two such regular Natarajan methods are consistent.
Proof Let (M, m,n ), (M, m,n ) be two regular methods such that each row and each
column of the two-dimensional infinite matrices (m,n ), (m,n ) is a regular Natara-
jan method for simple sequences. We now define a third method (M, m,n ) by the
equation

m,n
m,n = i, j mi,n j , m, n = 0, 1, 2, . . . .
i, j=0
Now, for s = {sm,n }, we get

,

(M, m,n )(s) = u m,n,i, j (M, i, j )(s),
i, j=0
where
mi,n j Q i, j , if i m, j n;
u m,n,i, j =
0, otherwise,

i, j
Q i, j = k, , i, j = 0, 1, 2, . . . . Using Theorem 6.8, we can verify that the
k,=0
four-dimensional infinite matrix (u m,n,i, j ) is regular. Thus,
sk, (M, m,n ) implies that sk, (M, m,n ).
Similarly, we can prove that
sk, (M, m,n ) implies that sk, (M, m,n ).
Consequently, = and so the methods (M, m,n ) and (M, m,n ) are consistent,
Let
,
,

(x, y) = m,n x m y n , (x, y) = m,n x m y n .
m,n=0 m,n=0
It is clear that these series converge for |x|, |y| < 1. Let
,
(x, y)
k(x, y) = km,n x m y n = ;
m,n=0
(x, y)
,
(x, y)
h(x, y) = h m,n x m y n = .
m,n=0
(x, y)
We note that

m,n
m,n = ki, j mi,n j ;
i, j=0

m,n
m,n = h i, j mi,n j ,
i, j=0
m, n = 0, 1, 2, . . . .
We now have
Theorem 6.12 If (M, m,n ), (M, m,n ) are regular, then
(M, m,n ) (M, m,n )
if and only if
,
,

|km,n | < and km,n = 1.
m,n=0 m,n=0
Proof Let
,

s(x, y) = sm,n x m y n .
m,n=0
Then,

,
,

m,n
(M, m,n )(s)x m y n = mi,n j si, j x m y n
m,n=0 m,n=0 i, j=0
= (x, y)s(x, y).
Similarly,
,

(M, m,n )(s)x m y n = (x, y)s(x, y).
m,n=0
Thus,
,
,

(M, m,n )(s)x m y n = k(x, y) (M, m,n )(s)x m y n
m,n=0 m,n=0
! , " ! , "

= km,n x y m n
(M, m,n )(s)x y
m n
,
m,n=0 m,n=0
which implies that

m,n
(M, m,n )(s) = kmi,n j (M, i, j )(s)
i, j=0
,

= cm,n,i, j (M, i, j )(s),
i, j=0
where
kmi,n j , if i m, j n;
cm,n,i, j =
0, otherwise.
If (M, m,n ) (M, m,n ), then (cm,n,i, j ) is regular. In view of Theorem 5.3,
,

lim cm,n,i, j = 1,
m+n
i, j=0

m,n
i.e., lim kmi,n j = 1,
m+n
i, j=0

m,n
i.e., lim ki, j = 1,
m+n
i, j=0
,

i.e., km,n = 1.
m,n=0
Again, by Theorem 5.3, there is H > 0 such that

,

|cm,n,i, j | H, m, n = 0, 1, 2, . . . ,
i, j=0

m,n
i.e., |kmi,n j | H, m, n = 0, 1, 2, . . . ,
i, j=0

m,n
i.e., |ki, j | H, m, n = 0, 1, 2, . . . ,
i, j=0
from which it follows that

,

|km,n | < .
m,n=0
,
,

Conversely, if |km,n | < and km,n = 1, it is easy to verify that
m,n=0 m,n=0
(cm,n,i, j ) is regular and consequently,
(M, m,n ) (M, m,n ),
As a consequence of Theorem 6.12, we have

Theorem 6.13 The regular methods (M, m,n ), (M, m,n ) are equivalent if and
only if
,
,

|km,n | < , km,n = 1
m,n=0 m,n=0
and
,
,

|h m,n | < , h m,n = 1.
m,n=0 m,n=0
Analogous to Theorem 3.14, we have the following result in the context of double
sequences and double series:
,

Theorem 6.14 If lim am,n = 0 and |bm,n | < , then,
m+n
m,n=0

m,n
lim amk,n bk, = 0.
m+n
k,=0
Proof Since {am,n }, {bm,n } are convergent, they are bounded and so, there exits M > 0
such that
|am,n | M, |bm,n | M, m, n = 0, 1, 2, . . . .
,

Since |bm,n | < , given > 0, there exist positive integers M1 , N1 such that
m,n=0
,

|bm,n | < . (6.18)
m>M1 ,n>N1
4M
Since, for fixed k, = 0, 1, 2, . . . ,
lim amk,n = 0,
m+n
we can choose positive integers M2 > M1 , N2 > N1 such that for m > M2 , n > N2 ,
we have
sup |amk,n | < ; (6.19)
0kM1 4M
0N1

sup |amk,n | < ; (6.20)
0kM1 4M
N1 +1n
and
sup |amk,n | < . (6.21)
M1 +1km 4M
0N1
Then, for m > M2 , n > N2 ,

m,n

amk,n bk,

k,=0

= amk,n bk, + amk,n bk,
0kM
0N 1 0kM1
>N

1 1

+ amk,n bk, + amk,n bk,

k>M1
0N1
k>M1
>N1

= |amk,n ||bk, | + |amk,n ||bk, |
0kM1 0kM1
0N1 >N1

+ |amk,n ||bk, | + |amk,n ||bk, |
k>M1 k>M1
0N1 >N1

<M +M +M +M
4M 4M 4M 4M
= , using (6.18)(6.21).
It now follows that


m,n
lim amk,n bk, = 0,
m+n
k,=0
We now have the following results on the Cauchy multiplication of (M, m,n )-
summable double sequences and double series (see [6]).
,

Theorem 6.15 If |am,n | < and {bm,n } is (M, m,n )-summable to B, then
m,n=0
{cm,n } is (M, m,n )-summable to AB, where

m,n
cm,n = amk,n bk, , m, n = 0, 1, 2, . . .
k,=0
,

and am,n = A.
m,n=0
Proof Let {tm,n }, {u m,n } be the (M, m,n )-transforms of {bm,n }, {cm,n }, respectively,

m,n
i.e., tm,n = mk,n bk, ,
k,=0

m,n
u m,n = mk,n ck, , m, n = 0, 1, 2, . . . .
k,=0
We can work to see that

! "

m,n
m,n
u m,n = amk,n (tk, B) + B ak, , m, n = 0, 1, 2, . . . ,
k,=0 k,=0
,

where lim tk, = B. Since |am,n | < and lim tm,n = B, using Theorem
k+ m+n
m,n=0
6.14, it follows that

m,n
lim amk,n (tk, B) = 0,
m+n
k,=0
so that ! , "

lim u m,n = B am,n = AB,
m+n
m,n=0
It is easy to prove the following result on similar lines:

,
,
,

Theorem 6.16 If |am,n | < with am,n = A and bm,n is (M, m,n )-
m,n=0 m,n=0 m,n=0
,

summable to B, then cm,n is (M, m,n )-summable to AB, where
m,n=0

m,n
cm,n = amk,n bk, , m, n = 0, 1, 2, . . . .
k,=0
As in the case of the Natarajan method (M, n ) for simple sequences, we can
prove the following result, using Theorem 6.14.
Theorem 6.17 Let (M, m,n ), (M, m,n ) be regular methods. Then (M, m,n )
(M, m,n ) is also regular, where we define, for x = {xm,n },
((M, m,n )(M, m,n ))(x) = (M, m,n )((M, m,n )(x)).
We can prove the following results too.
Theorem 6.18 For given regular methods (M, m,n ), (M, m,n ), and (M, tm,n ),
(M, m,n ) (M, m,n )
if and only if
(M, tm,n )(M, m,n ) (M, tm,n )(M, m,n ).
In view of Theorem 6.12, Theorem 6.18 can be reformulated as under:
Theorem 6.19 Given the regular methods (M, m,n ), (M, m,n ) and (M, tm,n ), the
following statements are equivalent:
(i) (M, m,n ) (M, m,n );
(ii) (M, tm,n )(M, m,n ) (M, tm,n )(M, m,n );
and
,
,

(iii) |km,n | < and km,n = 1,
m,n=0 m,n=0
where
,

(x)
= k(x) = km,n x m y n ,
(x) m,n=0
,

(x) = m,n x m y n ,
m,n=0
,

(x) = m,n x m y n .
m,n=0
In this context, we remark that we can extend many more results for the Nrlund,
the Weighted Mean, and the Natarajan methods for simple sequences to the corre-
sponding summability methods for double sequences. It is left to the reader to try
and solve such problems.
While concluding the present book, the author wishes to point out that the sum-
mability of trigonometric and WalshFourier series has been studied intensively in
the literature in the past 2030 years (see, for instance, [712]). The readers, who are
interested in summability theory, can acquaint themselves in this part of the literature
too for further studies.
References
1. Natarajan, P.N.: Nrlund means for double sequences and double series (communicated for
publication)
2. Moore, C.N.: On relationships between Nrlund means for double series. Proc. Amer. Math.
Soc. 5, 957963 (1954)
3. Natarajan, P.N.: Weighted means for double sequences and double series. Indian J. Math. 58,
3142 (2016)
4. Natarajan, P.N.: An inclusion theorem for weighted mean methods for double sequences. Indian
J. Math. (Supplement) Proceedings Sixth Dr. George Bachman Memorial Conference, 57, 39
42 (2015)
5. Natarajan, P.N.: Natarajan method of summability for double sequences and double series. An.
Stiint. Univ. Al. I. Cuza Iasi Math. (N.S.) Tomul LXII, 2, 547552 (2016)
6. Natarajan, P.N.: New properties of the Natarajan method of summability for double sequences.
Int. J. Phys. Math. Sci. 6(3), 2833 (2016)
7. Butzer, P.L., Nessel, R.J.: Fourier Analysis and Applications. Birkhauser, Basel (1971)
8. Schipp, F., Wade, W.R., Simon, P., Pal, J.: Walsh Series: An Introduction to Dyadic Harmonic
Analysis. Adam Hilger, Bristol (1990)
9. Shawyer, B., Watson, B.: Borels Methods of Summability. Oxford (1994)
10. Weisz, F.: Summability of Multidimensional Fourier series and Hardy spaces. Mathematics
and its Applications. Kluwer Academic Publishers, Dordrecht (2002)
11. Zygmund, A.: Trigonometric Series, 3rd edn. Cambridge, London (2002)
12. Grafakos, L.: Classical and Modern Fourier Analysis. Pearson Education, New Jersey (2004)
Index
A Double sequence complete, 83, 95

Abel method, 37, 54, 55 ds-complete, 95, 96, 99
Absolutely-convergent series
convergent double series, 83
Analytic continuation, 1 E
A-transform, 2, 22, 85, 117 Equivalence theorem, 37, 53
Equivalent, 30, 37, 38, 43, 48, 49, 53, 73,
102, 107, 120, 124, 127
B Euler method, 37, 55
Banach-space
algebra, 27, 31, 34, 35
Bounded-sequence
double sequence, 94, 99 F
Fourier-series
transforms, 1
C
Cauchy-sequence
double sequence, 9496 G
multiplication, 63, 65, 73, 77, 80, 81, 126 Generalized -core of a sequence, 27
product, 65 Generalized semiperiodic sequence, 1, 19
Closed-subspace
convex semigroup , 27, 31
convex set, 27 H
Comparison theorem, 111, 116 Hardy, 37, 44, 52
(c, 1) method, 1, 2, 18, 19
Consistent, 37, 50, 61, 101104, 120, 121
Convergent-sequence
double sequence, 83, 94 I
double series, 83, 95 Inclusion, 66, 67, 101
Convolution product, 27, 34, 35 Inclusion theorem, 37, 51, 111, 116
Core of a sequence, 27 Infinite chain, 63, 69
Invertible, 37, 56
Iteration product, 63, 69
D
Defranza, 63
Double-sequence K
series, 8386, 102, 108, 124 Knopps core theorem, 27, 28, 30
P.N. Natarajan, Classical Summability Theory, DOI 10.1007/978-981-10-4205-8
130 Index
L space, 3
-equivalent, 64 Sequence space r , 1, 13
Linear span, 1, 17, 18 Series-convergent
-stronger, 63 absolutely convergent, 65, 83
divergent, 1
Fourier, 1
M Sherbakhoff, 27, 28
Matrix-infinite signum (sgn.), 11, 30, 88, 91, 96, 98
conservative, 1, 3 Silverman-Toeplitz theorem, 1, 3, 83, 85
convergence preserving, 1, 3 Silverman-Toeplitz theorem for 4-
lower triangular, 22 dimensional infinite matrices, 83,
regular, 1, 3, 12, 13, 17, 18 85
Schur, 8, 12, 18, 96, 99 Steinhaus theorem, 1, 3, 13, 15, 17, 18, 46,
Mercerian theorem, 27, 35 83, 94
(M, n ) method, 37, 48, 49, 51, 5456, 120 Steinhaus theorem for 4-dimensional infinite
Mricz and Rhoades, 37, 44 matrices, 83, 94
Steinhaus type theorem, 1, 13, 16, 19, 21, 25
Strictly -stronger, 64
N Summability field, 67
Natarajans method, 37, 48, 49, 54, 63, 69, Summability method-(c, 1)
101, 120, 121, 127, 128 Abel, 37, 54, 55
Natarajans theorem, 28 Euler, 37, 55
Non-trivial method, 49 (M, n ) or Natarajan, 37
Nrlund method, 101105, 107 non-trivial, 49
Norm, 3, 17, 22, 31, 34, 46, 95 Nrlund, 128
translative, 37, 51
weighted mean, 37, 38, 108, 128
P Y , 3, 37, 49, 55
Pringshiem, 83, 85, 107 Symmetric product, 63, 65, 67, 68
Product theorem, 37, 55, 56, 60
S T
Schurs theorem, 1, 3, 13, 83 Tower, 16
Schurs theorem for 4-dimensional infinite Translative, 37, 51
matrices, 83, 94
Sequence-convergent
almost convergent, 17 W
Cauchy, 12 Weighted Mean method, 37, 38, 44, 101,
divergent, 18 107109, 111, 116, 120
eventually periodic, 1, 17
generalized semiperiodic, 1
non-periodic, 1 Y
of zeros and ones, 17 Y -method, 37, 49, 55, 61

P.N. Natarajan-Classical Summability Theory-Springer (2017)

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

P.N. Natarajan-Classical Summability Theory-Springer (2017)

Hochgeladen von

Copyright:

Verfügbare Formate

P.N.

ISBN 978-981-10-4204-1 ISBN 978-981-10-4205-8 (eBook)

Springer Nature Singapore Pte Ltd. 2017

Printed on acid-free paper

This Springer imprint is published by Springer Nature

In Chap. 5, a new denition of convergence of a double sequence and a double

Chennai, India P.N. Natarajan

1 General Summability Theory and Steinhaus Type Theorems . . .... 1

5 The SilvermanToeplitz, Schur, and Steinhaus Theorems

The present chapter is devoted to a study of some basic concepts in summability

1.1 Basic Definitions and Concepts

which is known to diverge. However, let

The associated sequence {sn } of partial sums is {1, 0, 1, 0, . . . }, which is (C, 1)

Definition 1.2 Let X, Y be sequence spaces. The infinite matrix A = (ank ), n, k

the letter P denoting Preservation of limit.

1.2 The SilvermanToeplitz Theorem, Schurs Theorem,

We now prove a landmark theorem in summability theory, due to SilvermanToeplitz,

lim ank = 0, k = 0, 1, 2, . . . ; (1.2)

Taking limit as n in (1.4), we have

lim (Ax)n = s, using (1.3) and (1.7)

and so A is regular, completing the proof of the sufficiency part.

so that (1.2) holds. Again consider the sequence x = {xn }, where xn = 1, n =

Define the sequence x = {xk }, where

To prove that (1.1) holds, we assume that

and arrive at a contradiction.

Now, define the sequence x = {xk }, where

Note that lim xk = 0. Since A is regular, lim (Ax)n = 0. However,

In such a case, prove that

Definition 1.4 A = (ank ) is called a Schur matrix if A ( , c), i.e., {(Ax)n } c,

The following result gives a characterization of a Schur matrix in terms of the

Thus, for each r = 0, 1, 2, . . . , we have

proving that {(Ax)n } c, i.e., A ( , c), proving the sufficiency part.

Suppose not. Then,

through some subsequence of positive integers. We also note that

We can now find a positive integer m(1) such that

It now follows that

Then, choose a positive integer k(3) > k(2) such that

It now follows that

We now define a sequence x = {xk } as follows: x0 = x1 = 0 and

xk = (1)r sgn bm(r ),k ,

In view of (1.20) and (1.21), we have

which contradicts (1.3). This establishes our claim.

1.3 A Steinhaus Type Theorem

r 1 being a fixed integer.

m = 2 p, n(m) > n(m 1), k(m) > k(m 1) + (2m 5)r,

Define the sequence x = {xk } as follows:

and if k(2 p) < k k(2 p + 1), then

if k(2 p 1) < k k(2 p),

if k(2 p) < k k(2 p + 1),

and consequently x = {xk }

which shows that {(Ax)n }

X Z , are called Steinhaus type theorems.

1.4 The Role Played by the Sequence Spaces r

Theorem 1.5 A = (ank ) sums every sequence of 0s and 1s in r if and only if

By construction, x = x (1) x (2) . It is easy to observe that x (2) is non-periodic. To

is in 1 but it is not (C, 1) summable.

Definition 1.6 x = {xk } is called a generalized semiperiodic sequence, if for any

|xk xk+sn | < , k k0 , s = 0, 1, 2, . . . .

1.5 More Steinhaus Type Theorems

In this section, we prove some more Steinhaus type theorems.

Note that p c0 c , where p 1. Just for convenience, we write 1 =

Theorem 1.9 A = (ank ) (, c) if and only if

sup |ank | < , (1.24)

and (1.1) holds.

is in (, ; P) (cs, ) but a0k 0, k .