Beruflich Dokumente
Kultur Dokumente
Natarajan
Classical Summability
Theory
123
P.N. Natarajan
Formerly of the Department of Mathematics
Ramakrishna Mission Vivekananda College
Chennai, Tamil Nadu
India
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
ix
x Contents
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
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
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
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
4 1 General Summability Theory and Steinhaus
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
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
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)
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
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
(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
8 1 General Summability Theory and Steinhaus
= 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
lim xk = s.
k
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
r
lim |ank | M.
n
k=0
1.2 The SilvermanToeplitz Theorem, Schurs Theorem, and Steinhaus Theorem 9
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
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
So,
|bmk | h, m
k=0
lim bmk = 0, k = 0, 1, 2, . . . .
m
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
and
k(2)
h
|bm(2),k | < .
k=0
10
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
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
12 1 General Summability Theory and Steinhaus
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
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),
1.2 The SilvermanToeplitz Theorem, Schurs Theorem, and Steinhaus Theorem 13
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.
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 ,
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
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,
14 1 General Summability Theory and Steinhaus
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
Note that,
while
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
/ 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
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.
16 1 General Summability Theory and Steinhaus
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 ) = ,
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).
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
18 1 General Summability Theory and Steinhaus
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, . . . .
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
p = {x = {xk } : |xk | p < }, p 1;
k=0
c0 = {x = {xk } : lim xk = 0}.
k
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
so that
lim (ank 1)xk = 0.
n
k=0
Consequently,
lim (Ax)n = xk
n
k=0
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.
So, we have
(, c; P ) (X, c) = ,
22 1 General Summability Theory and Steinhaus
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
(Ax)n = xk , x = {xk } . We also write A (, ; P) if A (, ; P) with
n=0 k=0
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
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
1.5 More Steinhaus Type Theorems 23
lim ank > 0, (1.30)
k
n=0
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
%&
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
24 1 General Summability Theory and Steinhaus
so that
n(m)
an,k(m) > 2
n=n(m1)+1 2 2
= .
)
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.
1.5 More Steinhaus Type Theorems 25
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
(, ; 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
References
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 *.
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
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
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]).
Natarajans theorem is
for any bounded sequence x if and only if A is regular and satisfies (2.3),
i.e., lim |ank | , 1.
n
k=0
|y z| lim |z (Ax)n |.
n
|y z| lim |z (Ax)n |
n
= lim ank (z xk )
n
k=0
lim |z xk |,
k
Conversely, let
()
K (A(x)) K (x).
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
30 2 Core of a Sequence and the Matrix Class (, )
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.
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
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
Theorem 2.3 The matrix class (, ) is a Banach algebra under the norm
A = sup |ank | , A = (ank ) (, ), (2.6)
k0 n=0
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.
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 .
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
< ,
so that
ank 1 < for all > 0.
n=0
Consequently,
ank = 1, k = 0, 1, 2, . . . .
n=0
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
so that
sup |cnk | <
k0 n=0
A B A B .
i.e., e0k = 1, k = 0, 1, 2, . . . ;
enk = 0, n = 1, 2, . . . ; k = 0, 1, 2, . . . .
n
cnk = aik bni,k
n=0 n=0 i=0
= ank bnk
n=0 n=0
= 1, k = 0, 1, 2, . . . ,
We close the present chapter by proving a Mercerian theorem for the Banach algebra
(, ) under the convolution product *.
Theorem 2.6 If
|| < 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
(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
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)
6. Maddox, I.J.: Elements of Functional Analysis. Cambridge University Press, Cambridge (1977)
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.
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
and
Pn , n . (3.2)
Remark 3.1
n
|Pn | | pk |
k=0
n+m
| pk |
k=0
L|Pn+m |,
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
Thus,
sm
0, m .
Qm
40 3 Special Summability Methods
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
3.1 Weighted Mean Method 41
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
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.
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, . . . ,
3.1 Weighted Mean Method 43
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
|Pn | L|Pn+m |,
|Q n | L|Q n+m |, (3.8)
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
44 3 Special Summability Methods
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 .
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,
3.1 Weighted Mean Method 45
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
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
46 3 Special Summability Methods
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
and
x A2 = sup |Bn |, x = {xk } A2 ,
n0
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
cn Pn1 tn1 ck ck+1
Bn = Qn + Qn Pk tk
pn k=n
pk pk+1
= ank tk ,
k=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,
48 3 Special Summability Methods
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
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
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.
n = 0 + 2 + 4 + + 2k , if n = 2k or 2k + 1.
Now,
|2k | |n | < ,
k=0 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
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
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.
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
lim wn = s,
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 ,
A B (or B A),
(M, n ) (M, n )
if and only if
52 3 Special Summability Methods
|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
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.
3.2 (M, n ) Method or Natarajan Method 53
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
54 3 Special Summability Methods
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.
u n = 0 an + 1 an1 + + n a0 , n = 0, 1, 2, . . . .
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
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
The Euler summability method is defined as follows (see [11, pp. 5657]).
56 3 Special Summability Methods
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.
lim e(r ) = 0, k = 0, 1, 2, . . .
n nk
if and only if
|1 r | < 1
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
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
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
58 3 Special Summability Methods
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
to conclude that
n1
(1) jk n c j+1 j ck = 1, 0 k n 1. (3.22)
j=k
n
n1
n
ck r k (1 r )nk xk1 = r (1 r )nk1 k . (3.23)
k=1 k=0
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 ,
60 3 Special Summability Methods
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
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
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)
6. Maddox, I.J.: Elements of Functional Analysis. Cambridge University Press, Cambridge (1977)
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.
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.
Let ((M, )) denote the set of all sequences x = {xk } such that (M, )(x) .
if
((M, )) ((M, ));
if
((M, )) ((M, ));
if
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
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.
((M, )) ((M, ))
if and only if
, i.e., (M, ) M.
((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.
66 4 More Properties of the (M, n ) Method and Cauchy
In view of Theorem 2.2, ((M, )) ((M, )) if and only if |n | < , i.e.,
n=0
, completing the proof.
((M, )) ((M, ))
if and only if
b = {bn } .
=b =
and so
((M, )) ((M, ))
((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 } .
Let {xn } . Since (M, ), (M, ) (, ), using Theorem 4.1,
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 *.
68 4 More Properties of the (M, n ) Method and Cauchy
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, . . .
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.
(x) = n x n
n=0
n1 + n
= xn
n=0
2
1+x
= (x).
2
It is now clear that
((M, )) ((M, )),
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 },
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
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
72 4 More Properties of the (M, n ) Method and Cauchy
= 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
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,
completing the proof of the theorem.
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)
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
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, . . . .
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
so that
lim u n = B an = AB,
n
n=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
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 ).
76 4 More Properties of the (M, n ) Method and Cauchy
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
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
n
where z n = xk ynk , n = 0, 1, 2, . . . .
k=0
78 4 More Properties of the (M, n ) Method and Cauchy
n
n = n
ck r k (1 r )nk yk , n = 0, 1, 2, . . . . (4.2)
k=0
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
we have,
n1
(1) jk n c j+1 j ck = 1, 0 k n 1. (4.5)
j=k
n
n1
n
ck r k (1 r )nk yk1 = r (1 r )nk1 k . (4.6)
k=1 k=0
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
80 4 More Properties of the (M, n ) Method and Cauchy
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
<
{x1 (1 r )n + x2 r (1 r )n1 + + xn r n1 (1 r )}
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
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
where we note that the series on the right-hand side converges, since,
|xk r k1 | |xk | < ,
k=1 k=0
82 4 More Properties of the (M, n ) Method and Cauchy
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
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.
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].
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.
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.
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
lim xm,n = 0.
m+n
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).
,
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
and
|am 1 ,n 1 ,k,0 | < , using (5.6).
k=0
r p1
|am p ,n p ,k,0 | < ; (5.7)
k=0
8
|am p ,n p ,k,0 | > (5.8)
k=0
|am p ,n p ,k,0 | < . (5.9)
k=r p +1
8
88 5 The SilvermanToeplitz, Schur, and Steinhaus Theorems
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 .
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
5.2 The SilvermanToeplitz Theorem for Four-Dimensional Infinite Matrices 89
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
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
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
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
5.2 The SilvermanToeplitz Theorem for Four-Dimensional Infinite Matrices 91
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
22 p+1
|am p ,n p ,k p , p | > , p = 0, 1, 2, . . . . (5.13)
p
,
|(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
92 5 The SilvermanToeplitz, Schur, and Steinhaus Theorems
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
where
Hence,
,
(Ax)m,n x = am,n,k, (xk, x) + rm,n x.
k,=0
5.2 The SilvermanToeplitz Theorem for Four-Dimensional Infinite Matrices 93
|am,n,k, | H, m, n, k, = 0, 1, 2, . . . , H > 0.
Let
|xk, x| L , k + = 0, 1, 2, . . . , L > 0.
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|
Consequently,
lim (Ax)m,n = x.
m+n
Let cds ,
ds , respectively, denote the spaces of convergent double sequences and
bounded double sequences.
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
is finite.
and
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
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, . . . ,
96 5 The SilvermanToeplitz, Schur, and Steinhaus Theorems
,
|(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),
|(Ax)m,n+v (Ax)m,n | 0, m + n , v = 0, 1, 2, . . . ,
,
(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
|am p +u 0 ,n p ,k,0 am p ,n p ,k,0 | > 2, p = 1, 2, . . . . (5.25)
k=0
1 1
r
7
|am 1 +u 0 ,n 1 ,k,0 am 1 ,n 1 ,k,0 | > > .
k=0
4
1 1
r
|am 2 +u 0 ,n 2 ,k,0 am 2 ,n 2 ,k,0 | < . (5.27)
k=0
4
Using (5.1),
|am 2 +u 0 ,n 2 ,k,0 am 2 ,n 2 ,k,0 | < ,
k=0
2 1
r
3
|am 2 +u 0 ,n 2 ,k,0 am 2 ,n 2 ,k,0 | > > .
k=r1
2
|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
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
= (am p +u 0 ,n p ,k,0 am p ,n p ,k,0 )xk,0
k=0
r p 1
+ (am p +u 0 ,n p ,k,0 am p ,n p ,k,0 )xk,0
k=r p1
+ (am p +u 0 ,n p ,k,0 am p ,n p ,k,0 )xk,0
k=r p
r p1 1
= (am p +u 0 ,n p ,k,0 am p ,n p ,k,0 )xk,0
k=0
r p 1
+ |am p +u 0 ,n p ,k,0 am p ,n p ,k,0 |
k=r p1
+ (am p +u 0 ,n p ,k,0 am p ,n p ,k,0 )xk,0
k=r p
5.3 The Schur and Steinhaus Theorems for Four-Dimensional Infinite Matrices 99
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
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
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.
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]).
let
m,n
Pm,n = pi, j , m, n = 0, 1, 2, . . . .
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 ).
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 ).
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
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
104 6 The Nrlund, Weighted Mean, and (M, m,n ) Methods for Double Sequences
and
m
lim m,n,i, j = 0, j = 0, 1, 2, . . . , n.
m+n
i=0
So = . Thus, the regular Nrlund methods (N , pm,n ) and (N , qm,n ) are consistent,
completing the proof of the theorem.
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
6.1 Nrlund Method for Double Sequences 105
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
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 .
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
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.
6.1 Nrlund Method for Double Sequences 107
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),
,
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.
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].
m,n
Pm,n = pi, j , m, n = 0, 1, 2, . . . .
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 ).
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,
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
if
sm,n (N , pm,n ) implies that sm,n (N , qm,n ).
Theorem 6.5 The Weighted Mean method (N , pm,n ) is regular if and only if
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
110 6 The Nrlund, Weighted Mean, and (M, m,n ) Methods for Double Sequences
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,
,
sup |am,n,k, | < .
m,n0 k,=0
6.2 Weighted Mean Method for Double Sequences 111
Thus, (5.1) holds. So, by Theorem 5.3, (N , pm,n ) is regular, completing the proof of
the theorem.
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
112 6 The Nrlund, Weighted Mean, and (M, m,n ) Methods for Double Sequences
,
qm,n
< ; (6.11)
m,n=0
pm,n
and
Pm,n = o(Q m,n ), m + n , (6.12)
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 ).
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
6.2 Weighted Mean Method for Double Sequences 113
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.
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
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,
6.2 Weighted Mean Method for Double Sequences 115
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
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
Theorem 6.8 If (N , pm,n ), (N , qm,n ) are two Weighted Mean methods such that
and
,
Pm,n
< , (6.14)
m,n=0
Q m,n
then
(N , pm,n ) (N , qm,n ).
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,
6.2 Weighted Mean Method for Double Sequences 117
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,
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
,
,
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
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,
< ,
Since A is regular,
lim am,n,k, = 0, k, = 0, 1, 2, . . .
m+n
lim bm,n,k, = 0, k, = 0, 1, 2, . . . .
m+n
6.2 Weighted Mean Method for Double Sequences 119
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
(N , pm,n ) A.
(N , pm,n ) (N , qm,n ).
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
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
if
sk, (M, m,n ) implies that sk, (M, m,n ) too.
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
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,
Consequently, = and so the methods (M, m,n ) and (M, m,n ) are consistent,
completing the proof of the theorem.
Let
,
,
(x, y) = m,n x m y n , (x, y) = m,n x m y n .
m,n=0 m,n=0
122 6 The Nrlund, Weighted Mean, and (M, m,n ) Methods for Double Sequences
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
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
Similarly,
,
(M, m,n )(s)x m y n = (x, y)s(x, y).
m,n=0
6.3 (M, m,n ) or Natarajan Method for Double Sequences 123
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
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
m,n
i.e., |ki, j | H, m, n = 0, 1, 2, . . . ,
i, j=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,
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
6.3 (M, m,n ) or Natarajan Method for Double Sequences 125
,
|bm,n | < . (6.18)
m>M1 ,n>N1
4M
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
+ 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).
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
,
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
6.3 (M, m,n ) or Natarajan Method for Double Sequences 127
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)).
Theorem 6.18 For given regular methods (M, m,n ), (M, m,n ), and (M, tm,n ),
if and only if
(M, tm,n )(M, m,n ) (M, tm,n )(M, m,n ).
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
128 6 The Nrlund, Weighted Mean, and (M, m,n ) Methods for Double Sequences
,
(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
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