Professional Documents
Culture Documents
Bafykbzaced 2 Hyfbtdkw 7 I 6 F 7 Abzgdr 642 Pu 7 WRLR 2 X 4 SMQFXSHQ 45 Zwy 3 N 4 Ry
Bafykbzaced 2 Hyfbtdkw 7 I 6 F 7 Abzgdr 642 Pu 7 WRLR 2 X 4 SMQFXSHQ 45 Zwy 3 N 4 Ry
Bafykbzaced 2 Hyfbtdkw 7 I 6 F 7 Abzgdr 642 Pu 7 WRLR 2 X 4 SMQFXSHQ 45 Zwy 3 N 4 Ry
Sequence Spaces
and Measures of
Noncompactness with
Applications to Differential
and Integral Equations
Sequence Spaces and Measures
of Noncompactness with Applications
to Differential and Integral Equations
Józef Banaś Mohammad Mursaleen
•
Sequence Spaces
and Measures
of Noncompactness
with Applications
to Differential and
Integral Equations
123
Józef Banaś Mohammad Mursaleen
Department of Mathematics Department of Mathematics
Rzeszów University of Technology Aligarh Muslim University
Rzeszow Aligarh, Uttar Pradesh
Poland India
Poland J. Banaś
India M. Mursaleen
v
Contents
1 Introduction to FK Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Classical Sequence Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Linear Metric Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Paranormed Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 FK and BK Spaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Matrix Domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.6 Sequence Spaces of Matrix Domains . . . . . . . . . . . . . . . . . . . . 23
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2 Matrix Transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.1 Continuous Duals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2 Köthe–Toeplitz Duals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.3 Other Duals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.4 Multiplier Spaces. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.5 Matrix Classes of Some FK and BK Spaces . . . . . . . . . . . . . . . 52
2.6 Conservative, Regular, and Schur Matrices . . . . . . . . . . . . . . . . 57
2.7 Matrix Transformations for Matrix Domains. . . . . . . . . . . . . . . 64
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
vii
viii Contents
Appendix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
About the Authors
xi
xii About the Authors
faculty members of the host institutions. Prof. Mursaleen is member of the edi-
torial board of various scientific journals and has served as member of various
international scientific bodies and organizing committees that include International
Council of Scientists ‘‘Global World Communicator Education and Science’’. He
has worked on a number of joint projects of international collaboration. His main
research interests are sequence spaces, summability theory, approximation theory,
functional equations, measures of non-compactness and fixed-point theory.
Chapter 1
Introduction to FK Spaces
In this chapter, we shall present the concepts of linear metric and paranormed spaces
which play an important role in our studies on sequence spaces. The concept of a
linear space involves an algebraic structure given by the definition of two operations,
namely the sum of any two of its vectors and the product of any scalar with any
vector. On the other hand, a topological structure of a set may be given by a metric. If
a set is both a linear and metric space, then it will be natural to require the algebraic
operations to be continuous with respect to the metric. We also give here a short
introduction into the general theory of FK and BK spaces (c.f. [1, 2]).
c := x ∈ w: there exists L ∈ C such that |xk − L| → 0 (k → ∞), x∞ = sup |xk | ,
k
the space of all convergent sequences;
k k
bs := x ∈ w: sup xn < ∞, xbs = sup xn ,
k k
n=1 n=1
the space of all bounded series;
k
k
cs := x ∈ w: lim xn exists, xbs = sup xn ,
k
n=1 k
n=1
the space of all convergent series;
1 := x ∈ w: |xk | < ∞, x1 = |xk | ,
k k
the space of all absolutely convergent series;
1/ p
p := x ∈ w: |xk | < ∞, x p =
p
|xk | p
(1 ≤ p < ∞) ,
k k
the space of all absolutely p-summable series;
bv := x ∈ w: |xk − xk+1 | < ∞, xbv = | lim x| + |xk − xk+1 | ,
k k
the space of all sequences of bounded variation.
bv0 = bv ∩ c0 , xbv = |xk − xk+1 |.
k
Let p = ( pk )∞
k=0 be a bounded sequence of positive real numbers with supk pk =
H and M = max{1, H }. The following spaces were introduced and studied by
Lascarides and Maddox [3] and Simons [4]. These spaces will be studied in details
in Chap. 5.
∞ ( p) := x ∈ w: sup |xk | pk < ∞ ;
k
1.1 Classical Sequence Spaces 3
c( p) := x ∈ w: lim |xk − L| pk = 0, L ∈ C ;
k
c0 ( p) := x ∈ w: lim |xk | pk
=0 ;
k
( p) := x ∈ w: |xk | pk < ∞;
k
1
n
w( p) := x ∈ w: |xk − L| pk = 0.
n
k=1
We wish to endow a linear space X with a metric structure that is compatible with
the algebraic/geometric structure of X. We thus need to choose a distance function
on X which goes well with the linear structure of X. That is, we need such a metric
structure to be imposed on a linear space which can be “compatible” with the inherent
algebraic structure of that space. This leads us to the notion of linear metric space
(or metric linear space).
We start with the following definition:
Definition 1.1 Let X be a linear space and d a metric on X . Then d is called trans-
lation invariant if
We observe from (1.1) that the translation invariance is thus a concept which is
partly metric and partly algebraic. That is, it connects in a particular way the distance
function on a linear space with the operation of vector addition. In fact, if a linear
space is endowed with a translation invariant metric, then it makes the vector addition
operation continuous. That is, we have the following basic observation: Let X be a
linear space which is also a metric space. If the metric d of X is translation invariant,
then the map (x, y) → x + y is a continuous function from X × X into X.
Remark 1.1 Note that the translation invariance alone does not guarantee the conti-
nuity of the other operation of linear space, that is, scalar multiplication operation. As
an extreme example, consider metrizing a given nontrivial metric space X by the dis-
crete metric (which is, obviously, translation invariant). Then the map (λ, x) → λx
4 1 Introduction to FK Spaces
(from R×X into X ) is not continuous for any x ∈ X \ {0}. For, if x = 0, then k1 x is
a sequence which is not eventually constant, so endowing X with the discrete metric
d yields d k1 x, 0 = 1 for each k. Thus, we need the scalar multiplication operation
on X to be continuous, and then, we arrive at a class of spaces which are both metric
and linear, and in which the metric and linear structures of the space are naturally
compatible.
Definition 1.2 Let X be a linear space which is also a metric space with the transla-
tion invariant metric d on X . Then (X, d), or X for short, is said to be a linear metric
space, if the algebraic operations on X are continuous functions. That is, that X is a
linear metric space if and only if it is both a linear and a metric space such that
(i) the distance between any two points are preserved under the identical transla-
tions of these points,
(ii) the vector addition map (x, y) → x + y is a continuous function from X × X
into X, and
(iii) the scalar multiplication map (λ, x) → λx is a continuous function from R× X
into X.
Definition 1.3 A complete linear metric space is called a Fréchet space (some
authors call a complete linear metric space an F space and a locally convex F
space a Fréchet space). A linear metric space X is said to be nontrivial if X = {0},
finite-dimensional if dim(X ) < ∞, and infinite-dimensional if dim(X ) = ∞.
Examples
1.1 Rn is a Fréchet space, for any n ∈ N. This is easily proved by using the fact
that a sequence converges in a Euclidean space if and only if each of its coordinate
converges in R.
1.2 If we metrize R by the discrete metric, we do not obtain a linear metric space
even though this metric is translation invariant. By contrast, if we metrize R by
d: R × R → R+ with d(a, b) := |a 3 − b3 |, then we guarantee that the scalar
multiplication operation on R is continuous, but do not make R a metric linear space
because the metric d is not translation invariant.
1.3 Consider the linear space w of all real sequences which is metrized by means of
1 |xk −yk |
the metric: d(x, y) = ∞ k=0 2k 1+|xk −yk | . It is easy to see that w is a Fréchet space.
1.4 p (1 ≤ p < ∞) and ∞ are Fréchet spaces with respect to their usual metrics.
Now, we shall demonstrate some nontrivial examples [5, pp. 141].
Example 1.5 The space ( p) is a linear metric space with the metric d(x, y) =
k | x k − yk | , where 0 < pk ≤ 1 for all k = 1, 2, 3, . . .
pk
It is easy to see that d is translation invariant and the operation of addition is contin-
uous. We shall show that the scalar multiplication is continuous. Let xk → a in the
linear metric space X = ( p), and λk , λ0 ∈ C such that λk → λ0 (k → ∞). Then
1.2 Linear Metric Spaces 5
d(λk xk , λ0 a) = | λk x k − λ0 a | pk
k
= | (λk − λ0 )(xk − a) + λ0 (xk − a) + (λk − λ0 )a | pk
k
≤ | λk − λ0 | pk | x k − a | pk + | λ0 | pk | x k − a | pk
k k
+ | λk − λ 0 | | a | =
pk pk
(1) + (2) + (3), say.
k
m ∞
(3) ≤ | λk − λ0 | pk | a | pk + | a | pk = A + B, say.
k=1 k=m+1
ε ε ε 2ε
d(λk xk , λ0 a) ≤ (1 + M)d(xk , a) + + < + = ε,
3 3 3 3
i.e., scalar multiplication is continuous. Therefore, ( p) is a linear metric space.
Moreover, it is complete, hence Fréchet space.
Similarly, we can show that
Example 1.6 The space c0 ( p) is a linear metric space with the metric d(x, y) =
supk | xk − yk | pk .
We present here another version of linear metric space. The concept of paranorm is
closely related to linear metric space, and in fact, total paranormed space and linear
metric space both are same. But some time the notion of paranorm is easier to handle
the situation of continuity of algebraic operations (c.f. [6]).
Definition 1.4 Let X be a linear space. A paranorm is a function p: X → R defined
on X such that
(P.1) p(x) = 0 if x = θ (zero element of X )
(P.2) p(x) ≥ 0 for all x ∈ X
6 1 Introduction to FK Spaces
Then:
(a) p is a paranorm on X and satisfies
1
and the series k 2k converges, the series
1 pk (xn )
2k 1 + pk (xn )
k
1 pk (xn )
≤ p(xn )
2k 1 + pk (xn )
2k p(xn )
pk (xn ) ≤
1 − 2k p(xn )
for all sufficiently large n, which implies that pk (xn ) → 0 (n → ∞) for each k.
Finally to prove (P.5), let λn → λ and p(xn − x) → 0 as n → ∞. Then by
(1.2), p(xn − x) → 0 (n → ∞) for all k. Also pk (λn xn − λx) → 0 (n → ∞)
for all k, since pk is paranorm. Again using (1.2), we get p(λn xn − λx) → 0
(n → ∞).
(b) Let q be a paranorm which is stronger than every pk . Then q(xn ) → 0 (n →
∞) ⇒ pk (xn ) → 0 (n → ∞) for all k. Now, by using statement (1.2), we get
p(xn ) → 0 (n → ∞). Hence q is stronger than p.
(c) It is trivial.
This completes the proof.
Example 1.7 The set C of complex numbers with the usual algebraic operations and
p =| . |, the modulus, is a totally paranormed space. If we put d(z 1 , z 2 ) =| z 1 − z 2 |
for all z 1 , z 2 ∈ C, then (C, d) is a Fréchet space.
Example 1.8 ([8, pp. 44]) The space ( p) is total paranormed space for any bounded
sequence ( pk ).
8 1 Introduction to FK Spaces
⎪ 1
M
p(x) = d(x, θ) = | xk | pk
.
k
k
⎪ 1 ⎪ 1
M M
≤ | xk | pk
+ | yk | pk
k k
= p(x) + p(y),
i.e., P.4 holds. Condition P.5 can be shown easily as in Example 1.5. Further, p(x) =
0 ⇒ d(x, θ) = 0 ⇒ x = θ.
Hence ( p) is total paranormed space.
Example 1.9 ([8, pp. 45]) The spaces ∞ ( p) and c( p) are not linear metric spaces
(scalar multiplication is not continuous) with the paranorm g(x) = supk | xk | pk /M .
Moreover, they turn out to be a linear metric spaces if and only if inf k pk > 0.
We consider ∞ ( p). Let pk = k1 and xk = 1 for each k. Then x ∈ ∞ ( p). Let
0 <| λ |< 1. Then | λ | pk =| λ |1/k < 1 for each k and | λ |1/k → 1 as k → ∞.
So that g(λx) = supk | λxk | pk /M = supk | λ |1/k = 1. Hence g(λx) 0 as
λ → 0 and thus the scalar multiplication is not continuous, i.e., ∞ ( p) is not a
linear metric space. Now, suppose that inf k pk = β > 0 and x ∈ ∞ ( p). Then
g(λx) = supk | λxk | pk /M ≤ max{| λ |, | λ |β/M }g(x). Thus g(λx) → 0 as
λ → 0 ⇒ ∞ ( p) is a linear metric space. Conversely, if inf k pk = 0, then the above
example shows that there is an x ∈ ∞ ( p) for which g(λx) 0 as λ → 0 which
contradicts that g is a paranorm.
Theorem 1.2 ([7, Theorem 1.5]) The set w is a Fréchet space with respect to the
metric d defined by
∞
1 | xk − yk |
dw (x, y) = for all x, y ∈ w, (1.3)
2k 1+ | xk − yk |
k=0
1.3 Paranormed Spaces 9
(where the 21k can be replaced by any positive summable sequence ak ). Furthermore,
convergence in (w, dw ) and coordinate wise convergence are equivalent, that is
(n)
x (n) → x (n → ∞) in (w, dw ) if and only if xk → xk (n → ∞) for every k.
Here we recall another notion which has certain relation with the paranorm.
Note that ν is always nonnegative: By (i) and (ii), we have 0 = ν(θ) = ν(x − x) ≤
ν(x) + ν(−x) = ν(2x).
Note that every norm is seminorm but not conversely. For converse, define ν(x) =
| limn xn | on c. Take xn = n1 , n ∈ N. Then ν(x) = 0 while x = θ. Hence ν is not
a norm, while it is seminorm on c.
We have the following important relation.
Theorem 1.3 [6, pp. 92] Let X be a linear space. Then each seminorm ν on X is
also a paranorm but not conversely.
and
ν(λk xk − λ0 a)
= ν([λk xk − λk a − λ0 xk + λ0 a] + [λ0 xk − λ0 a] + [λk a − λ0 a])
≤ ν((λk − λ0 )(xk − a)) + ν(λ0 (xk − a)) + ν(λk − λ0 )a)
= | λk − λ0 | ν(xk − a)+ | λ0 | ν(xk − a)+ | λk − λ0 | ν(a).
One of the main features of FK space theory is that it provides easy and short proofs
of many classical results of summability theory. It was initiated by Zeller [2] which
is the most powerful and widely used tool in the characterization of matrix mappings
between sequence spaces, and the most important result was that matrix mappings
between FK spaces are continuous [1, Theorem 4.2.8].
A more general case of linear metric spaces is linear topological space/topological
vector space.
Definition 1.7 A linear topological space (or topological vector space) X is a linear
space (vector space) over K (C or R) which is endowed with a topology such that
vector addition X × X → X and scalar multiplication K × X → X are continuous
functions.
All normed vector spaces, and therefore all Banach spaces and Hilbert spaces, are
examples of topological vector spaces.
Definition 1.8 Let (X, T ) be a topological vector space and x ∈ X. Then, a set U
is called a neighborhood of x if there is an open set G with x ∈ G ⊂ U. Thus any
open set G containing x is a neighborhood of x.
(i) Convex if for each x and y in E, t x + (1 − t)y is in E for all t in the unit
interval, that is, whenever 0 ≤ t ≤ 1. In other words, E contains all line
segments between points in E.
(ii) A cone (when the underlying field is ordered) if for every x ∈ E and 0 ≤ λ ≤ 1,
λx ∈ E. That is E is cone if λE ⊂ E for every 0 ≤ λ ≤ 1.
(iii) Balanced if for all x in E, λx is in E if |λ| ≤ 1. If the underlying field K is the
real numbers, this means that if x is in E, E contains the line segment between
x and −x. For a complex linear space X , it means for any x ∈ E, E contains
the disk with x on its boundary, centered on the origin, in the one-dimensional
complex subspace generated by x.
(iv) Absorbent or absorbing if for every x ∈ X , there exists ε > 0 such that λx ∈ E,
whenever λ is a scalar satisfying | λ |≤ ε. The set E can be scaled out to absorb
every point in the space.
(v) Absolutely convex if it is both balanced and convex. That is, if x, y ∈ E,
| λ | + | μ | ≤ 1 imply λx + μy ∈ E.
Remark 1.3 Let (X, p) be a paranormed space. Then the open neighborhoods of 0,
Nr (0) = {x ∈ X : p(x) < r }, are absorbing for all r > 0.
1.4 FK and BK Spaces 11
Proof We assume that Nr (0) is not absorbing for some r > 0. Then there are x ∈ X
and a sequence (λn )∞ n=0 of scalars with λn → 0 (n → 0) and λn x ∈ / Nr (0) for all
n = 0, 1, . . . But this means p(λn x) ≥ r for all n contradicting condition (P.5) in
Definition 1.4.
Definition 1.10 A linear topological space is called locally convex if and only if
every neighborhood U contains an absolutely convex set V. That is, a locally convex
space is a linear topological space in which the origin has a local base of absolutely
convex absorbent sets. Because translation is (by definition of “linear topological
space”) continuous, all translations are homeomorphisms, so every base for the
neighborhoods of the origin can be translated to a base for the neighborhoods of
any given vector.
A locally convex space is then defined to be a vector space X along with a family of
seminorms { pα }α∈A on X. Note that a seminormed space is locally convex.
Definition 1.11
(i) A sequence space X with linear topology is called a K space if each of the
maps Pn : X → C defined by Pn (x) = xn is continuous, for all x = (xk ) ∈ X
and every n ∈ N.
(ii) A Fréchet space is a complete linear metric space or equivalently, a complete
total paranormed space. In other words, a locally convex space is called a Fréchet
space if it is metrizable and the underlying metric space is complete.
(iii) K space X is called an FK space if X is a complete linear metric space, that
is, X is an FK space if X is Fréchet space with continuous coordinates (i.e., if
(n)
x (n) → x (n → ∞) in the metric of X then xk → xk (n → ∞) for each k).
Note that some authors include local convexity in the definition of FK space. But
much of the theory can be developed without local convexity.
Remark 1.4 A Fréchet sequence space (X, d X ) is an FK space if its metric d X is
stronger than the metric d | X of w on X. Hence, an K space X is continuously
embedded in w , that is the inclusion map ι: (X, d X ) → (w, d) defined by ι(x) = x
(x ∈ X ) is continuous.
Theorem 1.4 [7, Theorem 1.14] Let (X, d X ) be a Fréchet space, (Y, dY ) an FK
space and f : X → Y a linear map. Then f : (X, d X ) → (Y, d |Y ) is continuous if
and only if f : (X, d X ) → (Y, dY ) is continuous.
Proof First we assume that f : (X, d X ) → (Y, dY ) is continuous. Since Y is an FK
space its metric dY is stronger than the metric d |Y of w on Y. So f : (X, d X ) →
(Y, d |Y ) is continuous.
Conversely we assume that f : (X, d X ) → (Y, d |Y ) is continuous. Since (Y, dY ) is a
Hausdorff space and f is continuous, the graph of f, graph( f ) = { f (x, f (x)): x ∈
X }, is a closed set in (X, d X ) × (Y, d |Y ) by the Closed Graph Lemma, hence a
closed set in (X, d X ) × (Y, dY ), since the FK metric dY is stronger than d |Y . By the
Closed Graph Theorem, the map f : (X, d X ) → (Y, dY ) is continuous.
This completes the proof.
12 1 Introduction to FK Spaces
Proof For each n ∈ N0 , we define the linear functional f a,n : X → C by f a,n (x) =
n
k=0 ak x k for all x ∈ X . Since X is an FK space, the coordinates Pk : X → C
are continuous on X for all k = 0, 1, . . . , and so are the functionals f a,n (x) =
n
k=0 ak Pk (x) (n = 0, 1, 2, . . .). For each x ∈ X, f a (x) = lim n→∞ f a,n (x) exists,
and so by the Banach-Steinhaus Theorem f a : X → C is continuous.
This completes the proof.
Definition 1.12 A normed FK space is called a BK space, that is, a BK space is a
Banach sequence space with continuous coordinates.
Example 1.9
(i) The famous example of an FK space which is not a BK space is the space (w, dw ).
(ii) The spaces ∞ , c, c0 , p (1 ≤ p < ∞),bs,cs, and bv are BK spaces with their
natural norms.
Now we introduce the concept of a Schauder basis. For finite-dimensional spaces,
the concepts of Schauder and algebraic bases coincide. In most cases of interest,
however, the concepts differ. Every linear space has an algebraic basis. But there are
linear metric spaces without a Schauder basis, as we shall see later in this subsection.
A Schauder basis or countable basis is similar to the usual (Hamel) basis of a vector
space, the difference is that Hamel bases use linear combinations that are finite sums,
while for Schauder bases they may be infinite sums. This makes Schauder bases more
suitable for the analysis of infinite-dimensional topological vector spaces including
Banach spaces. A Hamel basis is free from topology while a Schauder basis depends
on the metric in question since it involves the notion of “ convergence” in its definition
and hence topology.
Definition 1.13 A sequence (bk )∞ k=0 in a linear metric space (X, d) is called a
Schauder basis (or briefly basis) for X if for every x ∈ X there exists a unique
∞
sequence (αk )∞
k=0 of scalars such that x =
[n]
k=1 αk bk , that is d(x, x ) → 0
(n → ∞), where x [n] n
= k=0 αk bk is known as the n-section of x. The series
k αk bk which has the sum x is called the expansion of x, and (αk ) is called the
sequence of coefficients of x with respect to the basis (bk ).
1.4 FK and BK Spaces 13
Example 1.10
(i) In φ, if we get, say, the metric of c0 , (e(k) )∞
k=1 is both a Schauder basis as well
as a Hamel basis, where e(k) is the sequence in which 1 occurs at the k-th place
and zeros elsewhere.
(ii) In c0 , (e(k) )∞
k=1 is a Schauder basis but not a Hamel basis. Since each x = (x k )
has the representation k xk e(k) which is unique . For if there is any other
representation, say, k bk e(k) which of course diverges for b = (bk ) ∈ / c0 , and
converges for b ∈ c0 , to b, not x, if (bk ) = (xk ). But this is not a Hamel basis,
since its span is φ, a proper subset of c0 . On the other hand, any Hamel basis of
c0 is uncountable and hence is automatically not a Schauder basis.
Let us demonstrate the case ( p). Take any x = (xk ) ∈ ( p). Write yn = x −
(x1 , x2 , . . . , xn , 0, 0, . . .), then
n
yn = x − xk e(k) = (0, 0, . . . , 0, xn+1 , xn+2 , . . .).
k=1
Hence
∞
p(yn ) = | xk | pk → 0(n → ∞),
k=n+1
i.e., x = (k)
k x k e . This representation for x is unique. For, if there is another
representation, say, x = k λk e(k) , then
⎪ n
(k)
p (λk − xk )e → 0(n → ∞).
k=1
l = limk→∞ xk .
(iv) The spaces w, c0 , and p (1 ≤ p < ∞) have AK, but the spaces c and ∞ do
not have AK.
Let us consider the case w. Let for each n = 0, 1, 2, . . . , e(n) be the sequence
(n) (n)
with en = 1 and ek = 0 (k = n). Then (e(n) )∞ n=1 is a Schauder basis for w.
More precisely, every sequence x = (xk )∞k=0 ∈ w has a unique representation
∞
x= xk e(k) ,
k=0
i.e.,
lim x [n] = x.
n→∞
m ∞
xn = (xk − xk+1 ) + xm+1 → (xk − xk+1 )
k=n k=n
and so |xn | ≤ xbv for each n. Hence bv0 is a BK space. Therefore bv = bv0 ⊕e
is also a BK space. Moreover bv0 is an AK space, since
∞
x − x [n] = |xk − xk+1 | → 0.
k=m+1
Moreover, bv is an AB space.
The notation AK arises from the German words Abschnittskonvergenz (for Sec-
tional Convergence), AD from Abschnittsdicht (for Sectionally Dense), and AB
from Abschnittsbeschrankte (i.e., Sectional Boundedness).
1.4 FK and BK Spaces 15
c A := {x ∈ w: Ax ∈ c}
is an FK space.
Theorem 1.7 [7, Theorem 1.21] Let X, Y , and Z be F H spaces with X ⊂ Y ⊂ Z .
(i) Then the topology T X is larger than the topology TY | X of Y on X.
(ii) They are equal if and only if X is a closed subspace of Y.
In particular, the topology of an F H space is unique.
(iii) If X is closed in Z , then X is closed in Y.
Proof Since X is an F H space, the inclusion map ι: X → H is continuous by
Remark 1.5(a), hence ι: X → Y is continuous by Theorem 1.4. This implies T X ⊃
TY | X . Now let T and T be F H topologies for an FK space. Then it follows by what
we have just shown that T ⊂ T ⊂ T .
If X is closed in Y, then X becomes an F H space with TY | X . It follows from the
uniqueness that T X = TY | X . If T X = TY | X , then X is a complete, hence closed,
subspace of Y.
Since X is closed in (Y, T Z |Y ), it is closed in (Y, TY ).
This completes the proof.
Remark 1.6 The class of FK spaces is fairly large.
Example 1.14 A Banach sequence space which is not a BK space.
We consider the spaces (c0 , . ∞ ) and (2 , . 2 ). Since they have the same
algebraic dimension, there is an isomorphism f : c0 → 2 . We define a second norm
. on c0 by x = f (x) 2 for all x ∈ c0 . Then (c0 , . ) becomes a Banach
space. But c0 and 2 are not linearly homeomorphic, since 2 is reflexive, and c0 is
not. Therefore the two norms on c0 are incomparable. By Theorem 1.7, (c0 , . ) is
a Banach sequence space which is not a BK space.
16 1 Introduction to FK Spaces
Example 1.15 (i) Since c0 and c are closed in ∞ , their BK topologies are the same;
since 1 is not closed in ∞ , its BK topology is strictly stronger than that of ∞ on
1 (Theorem 1.7).
Proof We prove here only (iii). Since the set E := {x ∈ ∞ : x j ∈ {0, 1}, j ∈ N} ⊂
∞ is uncountable, and for every distinct x, y ∈ E, x − y ∞ = 1. We have to
show that E is not dense in ∞ . Let if possible, E be dense in ∞ . Then there exists
z ∈ ∞ such that x − z ∞ < 1/4 (= ζ) for x ∈ E. Now
1 = x − y ∞ ≤ x − z ∞ + z − y ∞ < 1/4+ z − y ∞ .
This implies that z − y ∞ > 3/4, i.e., E is not dense in ∞ . Hence ∞ cannot be
separable.
Theorem 1.8 If a linear metric space X has Schauder basis, then it is separable.
Proof Let X be a Banach space which has a Schauder basis (bk )∞ k=1 (suppose that
bk = 1, for all k), i.e., every x ∈ X has a unique representation
∞
x= αk bk , αk ∈ R.
k=1
N
x− αk bk < ε/2.
k=1
For each αk ∈ R, we can find βk ∈ Q such that | αk − βk |< ε/2k+1 . Then, by the
triangle inequality
N
N
N N N
x− βk bk < x− αk bk + αk bk − βk bk < ε/2+ ε/2k+1 < ε.
k=1 k=1 k=1 k=1 k=1
1.4 FK and BK Spaces 17
and
X + π X := t = (tk )∞
k=1 : tk = a k + (−1)k
bk ∈ X ,
a = (ak )∞ ∞
k=1 , b = (bk )k=1 ∈ X. Now bs ⊂ X ⊂ ∞ implies that bs + πbs ⊂
X + π X ⊂ ∞ . But bs + πbs = ∞ , because if t ∈ ∞ then for
tk − tk−1 tk + tk−1
ak = , bk = ,
2 2
we have a ∈ bs, b ∈ π bs. Hence, X + π X = ∞ which is not separable. Hence X
is not separable.
This completes the proof.
Example 1.17 We define here the space ˆ p which is due to Das, Kuttner, and Nanda
[9].
Given an infinite series n an ,which will be denoted by a, let xn = a1 +a2 +· · ·+an .
We write ψ0,n = an , and
1 m
ψm,n = ian+i (m ≥ 1).
m(m + 1)
i=1
We say that the series “a” (or the sequence x) is absolutely almost convergent if
∞
m=0 | ψm,n |< ∞, uniformly in n. For p > 0, we write
∞
ˆ p := a : | ψm,n | < ∞, uniformly in n .
p
m=0
m
n+m
n+m
n−1
sup | an+i |= sup | ai |= sup | ai − ai |
m,n m,n m,n
i=0 i=n i=0 i=0
m
≤ 2 sup | ai |= 2M.
m
i=0
1 1
m m i m
| ian+i |= | (−1) an+ j + m an+i |
m m
i=0 i=0 j=0 i=0
≤ 2M + 2M = 4M.
Thus
m
1
| ψm,n |≤ ian+i
m(m + 1)
i=1
4M
≤ .
m+1
So
∞
∞
1
| ψm,n | p ≤ 4M < ∞.
(m + 1) p
m=0 m=0
In this section, we shall deal with the matrix domains related to the concepts of
ordinary and strong summability. We shall prove some results about matrix domains
of triangles to be BK spaces.
Definition 1.17 If A is an infinite matrix with complex entries ank (n, k ∈ N), then
we may write A = (ank ) instead of A = (ank )∞ n,k=0 . Also, we write An for the
∞
sequence in the n row of A, i.e., An = (ank )k=0 for every n ∈ N. In addition, if x =
th
(xk ) ∈ w then we define the A−transform of x as the sequence Ax = (An (x))∞ n=0 ,
where
∞
An (x) = ank xk (n ∈ N)
k=0
1.5 Matrix Domains 19
provided the series on the right converges for each n ∈ N. Further, the sequence x
is said to be A−summable to the complex number if An (x) → (n → ∞); we
shall write x → (A). This means that An ∈ X β for all n ∈ N and Ax ∈ c, where
X β = x −1 ∗ cs := a = (ak ) ∈ w: ax = (ak xk ) ∈ cs for all x = (xk ) ∈ X
is known as the β−dual of the sequence space X (which will be discussed in the
later chapter).
X A := {x ∈ w: Ax ∈ X }
and ⎪
X [A] := x ∈ w: A(| x |) = ank | xk | ∈X
k n
are called the (ordinary) matrix domain and strongly matrix domain of A. If X = c
then the sets c A and c[A] are called the Convergence domain and strong convergence
domain of A, respectively.
Remark 1.7 An infinite matrix T = (tnk )∞ n,k=0 is called a triangle (or normal) if
tnn = 0 and tnk = 0 for all k < n (n ∈ N). The study of matrix domains of triangles
in sequence spaces has a special importance due to the various properties which they
have, and most of these properties are immediate by the fact that every triangle has a
unique inverse S = (snk )∞ n,k=0 which is also a triangle, and x = T (S(x)) = S(T (x))
for all x ∈ w . For example, let X be a subset of w. If (X, d) is a linear metric
space, then X T is also a linear metric space with the metric dT given by dT (x, y) =
d(T x, T y) for all x, y ∈ X T , and X T has a Schauder basis if and only if X has a
Schauder basis. Further, if (X, d) is an FK space then so is (X T , dT ).
Lemma 1.10 ([7, Lemma 3.2]) Let X be a linear space, (Y, . Y ) a normed linear
space and T : X → Y a linear one-to-one map. Then X becomes a normed linear
space with x X = T x Y . If, in addition, Y is a Banach space and T is onto,
then (X, . X ) is a Banach space.
20 1 Introduction to FK Spaces
Theorem 1.11 ([7, Theorem 3.3, pp. 178]) Let T be a triangle and (X, . X ) be
a BK space. Then X T is also a BK space with the norm x T = T x X for all
x ∈ XT .
Proof Let us define the map L T : X T → X by L T (x) = T x for all x ∈ X T . Then
L T is linear, one-to-one (since T is triangle, and onto X, since X T = L −1
T (X ) and
L T is one-to-one). By Lemma 1.10, X T is a Banach space.
Now we show that the coordinates are continuous in X T . Let x (n) → x in X T . Then
(n)
yk = Tk (x (n) ) → yk = Tk (x), since X is a BK space. Let S be the inverse of T ,
(n) (n)
also a triangle. Then xk = kj=0 sk j y j → kj=0 sk j y j = xk . This shows that
the coordinates are continuous on X T .
This completes the proof.
We deduce the following:
Corollary 1.12 ([1, Theorem 4.3.13, pp. 64]) Let T be a triangle. Then cT is a BK
space with the norm x T,∞ = T x ∞ for all x ∈ cT .
Remark 1.8 A result similar to Theorem 1.11 also holds for the strong matrix domain
of triangles.
Definition 1.18 Let X be BK space. Then X is said to have monotone norm if (i)
x (m) ≥ x (n) for m > n, and (ii) x = sup x (m) . We call a norm
. of a sequence space X monotone, if | x̃k |≤| xk | (k = 0, 1, 2, . . .) implies
x̃k ≤ xk .
Example 1.18 (a) Let X = 1 ⊕ e. Then x (m) = m k=1 | x k | e
(n) = n, so
this norm satisfies condition (i) but not (ii). It is not a monotone norm.
(b) The spaces ∞ , c, c0 ,cs,bs have monotone norms.
(c) Let A = (C, 1). Then c A has monotone norm.
Theorem 1.12 ([7, Theorem 3.6]) Let X be BK space with monotone norm . ,
T a triangle and B a positive triangle. Then X [B] is a BK space with x X [B] =
B(| x |) for all x ∈ X [B] .
Proof For typographical convenience, we shall write . for x X [B] . Obviously,
. is a norm on X [B] . Further, since X is a BK space,
implies
n
Bn (| x (m) − x |) = bnk | x (m) − x |→ 0 (m → ∞)
k=0
for all n.
Hence the norm . is stronger than the metric of w on X [B] . Let (x (m) )∞
m=0 be
a Cauchy sequence in X [B] , hence in w by what we have just shown. Then there is
y ∈ w such that
x (m) → y in w. (1.4)
(m)
From (1.4), we conclude xk → yk (m → ∞) for each fixed k. Hence
B(| x (m) |) → B(| y |) (m → ∞) for all n, and so
Finally, (1.5) and (1.6) together imply z = B(| y |) ∈ X, that is, y ∈ X [B] .
Theorem 1.13 ([7, Theorem 3.7]) (a) Let X be a BK space with basis (b(k) )∞ k=0 ,
U = {u ∈ w: u k = 0 for all k}, u ∈ U and c(k) = u1 .b(k) (k = 0, 1, . . .) where
1 ∞ (k) ∞ −1 ∗ X.
u = ( u k )k=0 . Then (c )k=0 is a basis for Y = u
1
1
n
x (c0 )T = sup | xk |, by Theorem 1.11.
n un
k=0
1
n
x − x [m] (c0 )T = sup | xk |
n≥m+1 un
k=m+1
≤ sup | Tn (x) |< ε,
n≥m+1
hence
∞
x= xk e(k) , (1.7)
k=0
We have the following results concerning the Schauder bases of matrix domains of
triangles. Throughout, let T denote a triangle and S its inverse.
Proposition 1.14 Let (bn )∞ n=0 be a basis of the linear metric sequence space (X, d),
then, (S(b(n) ))∞
n=0 is a basis of X T with the metric dT given by dT (x, y) = d(T x, T y)
for all x, y ∈ X T .
∞
z= Tn (z)c(n) . (1.8)
n=0
∞
u = le + Tn (u − le)c(n) . (1.9)
n=0
∞
v = le(−1) + (Tn (v) − l)c(n) . (1.10)
n=0
First we give here some special and important matrices of triangles. The most impor-
tant summability methods are given by Hausdorff matrices and their special cases
c.f. [10, Theorem 4.3.13, pp. 64], [6, Theorem 4.3.13, pp. 64], [11, Theorem 4.3.13,
pp. 64] , [12, Theorem 4.3.13, pp. 64]).
Let μ = (μn )∞ ∞
n=0 be a given complex sequence, M = (m nk )n,k=0 be the diagonal
∞
matrix with m nn = μn (n = 0, 1, . . .), and D = (dnk )n,k=0 be the matrix with
dnk = (−1)k nk . Then, the matrix H = H (μ) = D M D is called the Hausdorff
matrix associated with the sequence μ; i.e.,
⎛ ⎝⎛ ⎝
n j
(−1) j+k nj k , 0 ≤ k ≤ n,
h nk = j=k
0, k > n,
for all k, n ∈ N0 .
24 1 Introduction to FK Spaces
for all k, n ∈ N0 .
Let r > −1 and define Arn by
(r +1)(r +2)...(r +n)
, n = 1, 2, . . . ,
An =
r n!
1, n = 0.
Then, the Cesàro matrix of order r is defined by the following matrix Cr = (cnk
r )
r−1
An−k
cnk =
r Arn , 0 ≤ k ≤ n,
0 , k > n,
for all k, n ∈ N0 .
for all k, n ∈ N0 whose generalization E q of order q > 0 was defined by the matrix
p
E q = (bnk ), where
n
p (q + 1)−n q n−k , 0 ≤ k ≤ n,
bnk = k
0, k > n,
for all k, n ∈ N0 .
Let 0 < r < 1 and nk = k!(n−k)!n!
for all k, n ∈ N0 . Then, the Euler matrix E r of
order r is defined by the matrix E = (enk
r r ), where
1.6 Sequence Spaces of Matrix Domains 25
n (1 − r )n−k r k , 0 ≤ k ≤ n,
r
enk = k
0, k > n,
Let t = (tk ) be a sequence of nonnegative real numbers with t0 > 0 and write
Tn = nk=0 tk for all n ∈ N0 . Then the Riesz matrix with respect to the sequence
t = (tk ) is defined by the matrix R t = (rnkt ) which is given by
tk
, 0 ≤ k ≤ n,
t
rnk = Tn
0, k > n,
for all k, n ∈ N0 . For t = e the Riesz matrix R t is reduced to the matrix C1 . The
inverse matrix S t = (snk
t ) of the matrix R t = (r t ) is given by
nk
(−1) n−k Tk
, n − 1 ≤ k ≤ n,
t
snk = tn
0, 0 ≤ k ≤ n − 2 or k > n,
Let q = (qk ) be a sequence of nonnegative real numbers with q0 > 0 and write
Q n = nk=0 qk for all n ∈ N0 . Then the Nörlund matrix with respect to the sequence
q
q = (qk ) is defined by the matrix N q = (ank ) which is given by
qn−k
q , 0 ≤ k ≤ n,
ank = Q n
0, k > n,
for all k, n ∈ N0 . For q = e the Nörlund matrix N q is reduced to the matrix C1 . The
inverse M q of N q is then given by (M q )n j = kn− j Q j for j ≤ n. For tn = Arn for
all n ∈ N0 , the method N q is reduced to the Cesàro method Cr of order r > −1.
26 1 Introduction to FK Spaces
1.6.6 Ar Matrix
Let 0 < r < 1. Then the Ar matrix, introduced by Başar [14], is given by Ar = (ank
r ),
where
1+r k
, 0 ≤ k ≤ n,
ank = n+1
r
0, k > n,
for all k, n ∈ N0 .
1.6.7 Matrix
Let U and U0 be the sets of all sequences with nonzero terms and nonzero first
terms, respectively. Let r, t ∈ U and s ∈ U0 . We define the infinite matrix Ā(r, s, t)
of generalized means (see [16]) by
⎧
⎨sn−k tk /rn ; (0 ≤ k ≤ n),
( Ā(r, s, t))nk =
⎩0; (k > n)
for all n, k ∈ N.
1.6 Sequence Spaces of Matrix Domains 27
For u ∈ U , let 1/u = (1/u k ). Let u, v, z ∈ U and define the difference matrix
(1) (m)
χ(1) = (χnk ), the difference matrix χ(m) = (χnk ) of order m, the generalized
difference matrix B(r ; s) = {bnk (r, s)}, the generalized weighted means or factorable
matrix G(u, v) = (gnk ), Aru = (ankr ), the summation matrix S m = (s m ) of order m,
nk
z
and A z = (ank ) by
(1) (−1)n−k , n − 1 ≤ k ≤ n,
χnk =
0, 0 ≤ k < n − 1 or k > n.
(m) (−1)n−k m , max{0, n − m} ≤ k ≤ n,
χnk = n−k
0, 0 ≤ k < max{0, n − m} or k > n.
⎧
⎨ r, k = n,
bnk (r, s) = s, k = n − 1, r, s ∈ R\{0}
⎩
0, 0 ≤ k < n − 1 or k > n.
u n vk , 0 ≤ k ≤ n,
gnk =
0, k > n.
1+r k
r
ank = n+1 u k , 0 ≤ k ≤ n,
0, k > n.
⎛ ⎝
m+n−k−1
, 0 ≤ k ≤ n,
m
snk = n−k
0, k > n.
z (−1)n−k z k , n − 1 ≤ k ≤ n,
ank =
0, 0 ≤ k < n − 1 or k > n.
The approach of constructing a new sequence space by means of the matrix domain
of a particular limitation method has recently been employed by several authors, e.g.,
Wang [17], Ng and Lee [18], Malkowsky [19], Başar and Altay [20], Malkowsky
and Sava ş [21], Aydın and Başar [22–25], Altay and Başar [26], Altay, Başar and
Mursaleen [27], Mursaleen, Başar and Altay [28], and Mursaleen and Noman [15,
29, 30], respectively. They introduced the sequence spaces (∞ ) Nq and c Nq in [17],
(∞ )C1 = X ∞ , and ( p )C1 = X p in [18], (∞ ) R t = r∞ t , c t = r t and (c ) t = r t
R c 0 R 0
in [19], ( p )χ(1) = bv p in [20], μG = Z (u, v; μ) in [21], (c0 ) Ar = a0r and c Ar = acr
in [22], [c0 (u, p)] Ar = a0r (u, p) and [c(u, p)] Ar = acr (u, p) in [23], (a0r )χ = a0r (χ)
28 1 Introduction to FK Spaces
X A XA References
p , ∞ Nq X a( p) , X a(∞) [17]
p (1 ≤ p ≤ ∞) C1 X p, X∞ [18]
X p (1 ≤ p ≤ ∞) χm C p (χm ), C∞ (χm ) [33]
c0 , c, ∞ Rq ( N̄ , q)0 , ( N̄ , q), ( N̄ , q)∞ [19]
c0 , c, ∞ χ c0 (χ), c(χ), ∞ (χ) [34]
c0 , c, ∞ χ2 c0 (χ2 ), c(χ2 ), ∞ (χ2 ) [35]
c0 , c, ∞ uχ2 c0 (u; χ2 ), c(u; χ2 ), ∞ (u; χ2 ) [36]
p , c0 , c, ∞ B(r, s) ˆ p , ĉ0 , ĉ, ˆ∞ [37]
c0 , c, p G(u, v) Z (u, v; c0 ), Z (u, v; c), Z (u, v; p ) [21]
c0 , c C1 c̃0 , c̃ [38]
c0 , c Er e0r , ecr [26]
e0t , ect , e∞
t B (m) e0t (B (m) ), ect (B (m) ), e∞ t (B (m) ) [39]
c0 , c G(u, v) (c0 )G(u,v) , cG(u,v) [40]
c0 , c Ar a0r , acr [22]
p (1 ≤ p ≤ ∞) Ar a rp , a∞r [25]
p (1 ≤ p ≤ ∞) Er e p , e∞
r r [27]
a0r , acr χ(1) a0r (χ), acr (χ) [24]
p
p (1 ≤ p < ∞) G(u, v) A [41]
p (1 ≤ p < ∞) χ(1) bv p [20, 42]
c0 , c, ∞ χm c0 (χm ), c(χm ), ∞ (χm ) [43, 44]
p (1 ≤ p < ∞) χ(m) p (χ(m) ) [45]
c0 , c, ∞ χ(m) c0 (χ(m) ), c(χ(m) ), ∞ (χ(m) ) [46]
e0r , ecr χ(m) e0r (χ(m) ), ecr (χ(m) ) [47]
∞ ( p) S bs( p) [14, 48]
c0 ( p), c( p), ∞ ( p) χ χc0 ( p), χc( p), χ∞ ( p) [49]
c0 ( p), c( p), ∞ ( p) uχ c0 (u, χ, p), c(u, χ, p), ∞ (u, χ, p) [50]
c0 ( p), c( p), ∞ ( p) uχ2 c0 (u, χ2 , p), c(u, χ2 , p), ∞ (u, χ2 , p) [51]
c0 ( p), c( p), ∞ ( p) G(u, v) c0 (u, v; p), c(u, v; p), ∞ (u, v; p) [52]
( p) G(u, v) (u, v; p) [53]
( p), ∞ ( p) Az bv(z, p), bv∞ (z, p) [54]
c0 (u, p), c(u, p) Ar a0r (u, p), acr (u, p) [23]
( p) Rt r t ( p) [55]
c0 ( p), c( p), ∞ ( p) Rt r0t ( p), rct ( p), r∞t ( p) [56]
r q ( p) χ r q ( p, χ) [57, 58]
r q ( p) Bm r q ( p, B m ) [59]
( p) Er er ( p) [60]
c0 ( p), c( p), ∞ ( p) χm χm c0 ( p), χm c( p), χm ∞ ( p) [61]
(m) (m) (m)
c0 ( p), c( p), ∞ ( p) uχ(m) χu c0 ( p), χu c( p), χu ∞ ( p) [62]
c0 , c β c0λ , cλ [15]
c0λ , cλ χ(1) c0λ (χ(1) ), cλ (χ(1) ) [29]
p , ∞ β λp , λ∞ [30, 31]
c0 , c, ∞ , p Ā(r, s, t) c̄0 , c̄, ¯∞ , ¯ p [16, 63]
and cβ = cλ in [15] , (c0λ )χ(1) = c0λ (χ(1) ) and (cλ )χ(1) = cλ (χ(1) ) in [29], λp =
( p )β and λ∞ = (∞ )β in [30, 31], where Nq , C1 , R t , and E r denote the Nörlund,
Cesàro, Riesz, and Euler means, respectively, χ denotes the band matrix defining
1.6 Sequence Spaces of Matrix Domains 29
the difference operator, G, Ar and β are defined in [21, 22] and [29], respectively,
μ ∈ {c0 , c, p } and 1 ≤ p < ∞. Also c0 (u, p) and c(u, p) denote the sequence
spaces generated from the Maddox’s spaces c0 ( p) and c( p) by Başarır [32].
We give here a useful table in continuation of the table given by [11].
Exercises
p(x) = sup | xk | pk /M .
k
1 | xk |
p(x) = .
2k 1+ | xk |
k
6. Prove that a paranormed space (X, p) which has a Schauder basis is separable.
7. Show that ν(x) =| lim x | is a seminorm on c but not a norm.
8. Show that d(x, y) =| x 3 − y 3 | is a metric on R but it is not translation invariant.
9. Show that every AK space has AD but converse need not be true.
10. Prove that if an FK space has AK, then
z −1 · X := {x ∈ w: x · z ∈ X }
c0 (χ) := {x ∈ w: χx = (χxk )∞ ∞
k=1 = (x k − x k+1 )k=1 ∈ c0 }
x χ = sup | χxk | .
k
References
25. Aydın, C., Başar, F.: Some new sequence spaces which include the spaces p and ∞ . Demon-
stratio Mathematica 38(3), 641–656 (2005)
26. Altay, B., Başar, F.: Some Euler sequence spaces of non-absolute type. Ukrainian Math. J.
57(1), 1–17 (2005)
27. Altay, B., Başar, F., Mursaleen, M.: On the Euler sequence spaces which include the spaces p
and ∞ . Inf. Sci. 176(10), 1450–1462 (2006)
28. Mursaleen, M., Başar, F., Altay, B.: On the Euler sequence spaces which include the spaces p
and ∞ II. Nonlinear Anal. 65(3), 707–717 (2006)
29. Mursaleen, M., Noman, A.K.: On some new difference sequence spaces of non-absolute type.
Math. Comput. Modell. 52, 603–617 (2010)
30. Mursaleen, M., Noman, A.K.: On some new sequence spaces of non-absolute type related to
the spaces p and ∞ I . Filomat 25(2), 33–51 (2011)
31. Mursaleen, M., Noman, A.K.: On some new sequence spaces of non-absolute type related to
the spaces p and ∞ I I . Math. Commun. 16(2), 383–398 (2011)
32. Başarır, M.: On some new sequence spaces and related matrix transformations. Indian J. Pure
Appl. Math. 26(10), 1003–1010 (1995)
33. Et, M.: On some generalized Cesàro difference sequence spaces. İstanbul Üniv. Fen Fak. Mat.
Derg. 55–56, 221–229 (1996–1997)
34. Kızmaz, H.: On certain sequence spaces. Canad. Math. Bull. 24(2), 169–176 (1981)
35. Et, M.: On some difference sequence spaces. Turkish J. Math. 17, 18–24 (1993)
36. Mursaleen, M.: Generalized spaces of difference sequences. J. Math. Anal. Appl. 203(3), 738–
745 (1996)
37. Kirişçi, M., Başar, F.: Some new sequence spaces derived by the domain of generalized differ-
ence matrix. Comput. Math. Appl. 60(5), 1299–1309 (2010)
38. Şengönül, M., Başar, F.: Some new Cesàro sequence spaces of non-absolute type which include
the spaces c0 and c. Soochow J. Math. 31(1), 107–119 (2005)
39. Kara, E.E., Başarır, M.: On compact operators and some Euler B (m) -difference sequence spaces.
J. Math. Anal. Appl. 379, 499–511 (2011)
40. Rhoades, B.E.: Some sequence spaces which include c0 and c. Hokkaido Math. J. 35, 587–599
(2006)
41. Khan, F.M., Rahman, M.F.: Matrix transformations on Ces àro sequence spaces of nonabsolute
type. J. Anal. 4, 97–101 (1996)
42. Çolak, R., Et, M., Malkowsky, E.: Some Topics of Sequence Spaces. Lecture Notes in Mathe-
matics, Fırat University Elâzıg, Turkey, 2004, pp. 1–63. Fırat University Press (2004), ISBN:
975-394-038-6
43. Et, M., Çolak, R.: On some generalized difference sequence spaces. Soochow J. Math. 21(4),
377–386 (1995)
44. Çolak, R., Et, M.: On some generalized difference sequence spaces and related matrix trans-
formations. Hokkaido Math. J. 26(3), 483–492 (1997)
45. Altay, B.: On the space of p-summable difference sequences of order m, (1 ≤ p < 1). Stud.
Sci. Math. Hungar. 43(4), 387–402 (2006)
46. Malkowsky, E., Parashar, S.D.: Matrix transformations in space of bounded and convergent
difference sequences of order m. Analysis 17, 87–97 (1997)
47. Polat, H., Başar, F.: Some Euler spaces of difference sequences of order m. Acta Math. Sci.
27B(2), 254–266 (2007)
48. Başar, F., Altay, B.: Matrix mappings on the space bs( p) and its α, β and γ−duals. Aligarh
Bull. Math. 21(1), 79–91 (2001)
49. Ahmad, Z.U., Mursaleen, M.: Köthe-Toeplitz duals of some new sequence spaces and their
matrix maps. Publ. Inst. Math. (Beograd) 42, 57–61 (1987)
50. Asma, Ç., Çolak, R.: On the Köthe-Toeplitz duals of some generalized sets of difference
sequences. Demonstratio Math. 33, 797–803 (2000)
51. Bektas, Ç.A.: On some new generalized sequence spaces. J. Math. Anal. Appl. 277, 681–688
(2003)
32 1 Introduction to FK Spaces
52. Altay, B., Başar, F.: Some paranormed sequence spaces of non-absolute type derived by
weighted mean. J. Math. Anal. Appl. 319(2), 494–508 (2006)
53. Altay, B., Başar, F.: Generalization of the sequence space ( p) derived by weighted mean. J.
Math. Anal. Appl. 330(1), 174–185 (2007)
54. Başar, F., Altay, B., Mursaleen, M.: Some generalizations of the space bv p of p−bounded
variation sequences. Nonlinear Anal. 68(2), 273–287 (2008)
55. Altay, B., Başar, F.: On the paranormed Riesz sequence spaces of non-absolute type. Southeast
Asian Bull. Math. 26(5), 701–715 (2002)
56. Altay, B., Başar, F.: Some paranormed Riesz sequence spaces of non-absolute type. Southeast
Asian Bull. Math. 30(5), 591–608 (2006)
57. Başarır, M.: On the generalized Riesz B-difference sequence spaces. Filomat 24(4), 35–52
(2010)
58. Başarır, M., Öztürk, M.: On the Riesz difference sequence space. Rend. Circ. Mat. Palermo
57(2), 377–389 (2008)
59. Başarır, M., Kayıkçı, M.: On the generalized B m -Riesz sequence space and property. J. Inequal.
Appl. Article ID 385029, 18 pp (2009)
60. Kara, E.E., Öztürk, M., Başarır, M.: Some topological and geometric properties of generalized
Euler sequence spaces. Math. Slovaca 60(3), 385–398 (2010)
61. Et, M., Başarır, M.: On some new generalized difference sequence spaces. Period. Math. Hung.
35(3), 169–175 (1997)
62. Malkowsky, E., Mursaleen, M., Suantai, S.: The dual spaces of sets of difference sequences of
order m and matrix transformations. Acta Math. Sin. Engl. Ser. 23(3), 521–532 (2007)
63. Mursaleen, M., Noman, A.K.: Applications of Hausdorff measure of noncompactness in the
spaces of generalized means. Math. Ineq. Appl. 16, 207-220 (2013)
Chapter 2
Matrix Transformations
The theory of matrix transformations deals with establishing necessary and sufficient
conditions on the entries of a matrix to map a sequence space X into a sequence space
Y. This is a natural generalization of the problem to characterize all summability
methods given by infinite matrices that preserve convergence.
If A is an infinite matrix with complex entries ank (n, k ∞ N), then we may write
A = (ank ) instead of A = (ank )∈ n,k=0 . Also, we write An for the sequence in the nth
row of A, i.e., An = (ank )∈ k=0 for every n ∞ N. In addition, if x = (xk ) ∞ w, then
we define the A-transform of x as the sequence Ax = (An (x))∈ n=0 , where
∈
An (x) = ank xk (n ∞ N)
k=0
provided the series on the right converges for each n ∞ N. Further, the sequence x is
said to be A-summable to the complex number if An (x) → (n → ∈); we shall
write x → (A) where is called the A- limit of x.
Let X and Y be subsets of w and A an infinite matrix. Then, we say that A defines
a matrix mapping from X into Y if Ax exists and is in Y for every x ∞ X . By (X, Y ),
we denote the class of all infinite matrices that map X into Y . Thus, A ∞ (X, Y ) if
and only if An ∞ X φ for all n ∞ N and Ax ∞ Y for all x ∞ X.
The study of sequence spaces is much more profitable when we consider them
equipped with certain topologies. In this chapter, we shall study various duals of
some sequence spaces and characterize several matrix classes (c.f. [1–3]).
Let X and Y be normed linear spaces. Then, B(X, Y ) denotes the set of all bounded
linear operators L: X → Y . If Y is complete, then B(X, Y ) is a Banach space
with the operator norm defined by →L→ = sup →L(x)→ for all L ∞ B(X, Y ).
x∞S X
By X ≤ = B(X, C), we denote the continuous dual of X , that is, the set of all
continuous linear functionals on X . If X is a Banach space, then we write X ∩ for X ≤
with its norm given by → f → = sup | f (x)| for all f ∞ X ≤ .
x∞S X
We shall write X ∩ Y if Y is the continuous dual of X , i.e. X ∩ is isomorphically
isometric to Y .
Theorem 2.1 ([3, pp. 92]) ∩1 ∈ .
Proof Let f ∞ ∩1 . Since e(k) is a Schauder basis for 1 , each x = (xk ) ∞ 1 can be
represented as x = xk e(k) . Let f ∞ ∩1 and define λk = f (e(k) ). Then, for every
k
x = (xk ) ∞ 1
f (x) = xk f (e(k) ) = xk λk . (2.1)
k k
|λn | ≥ → f →→x→ ≥ → f →
→λ→∈ ≥ → f →. (2.2)
where M = supk |λk |. Hence f is bounded and f ∞ ∩1 . Thus T is onto. Finally, by
(2.1) we get
(k) (k)
| f (x)| ≥ | xk f (e )| ≥ sup | f (e )| |xk | = sup | f (e(k) )| →x→1
k k k k
2.1 Continuous Duals 35
Therefore
f (x) = f (e) + (xk − ) f (e(k) ), for all x ∞ c.
k
Define x = (xk ) by
sgn f (e(k) ), 1 ≥ k ≥ n,
xk =
0, k > n.
n
| f (x)| = | f (e(k) )| ≥ → f →.
k=1
is absolutely convergent.
Put
f (x) = a + ak x k , (2.4)
k
where
a = f (e) − f (e(k) ), ak = f (e(k) ).
k
We have
36 2 Matrix Transformations
| f (x)| ≥ |a| + |ak | →x→∈ ,
k
Also, we have
r ∈
| f (x)| = a sgn a + an sgn an + an sgn a
n=1 n=r +1
r ∈
= |a| + |an | + an sgn a ≥ → f → by (2.6). (2.7)
n=1 n=r +1
Definition 2.1 For a non-empty set X , let R denote a ring of subsets of X . The symbol
ε E (E ∞ R) denotes a partition E 1 , . . . , E n of E such that E i ∞ R, E i ⊂ E j = ◦,
n E = E. By a charge μ on R, we mean a K-valued finitely additive set function
⊃i=1 i
with |μ(E)| < ∈ for each E ∞ R, and triplet (X, R, μ) is called a charged space;
further (X, R, μ) is called a completely charged space, if X ∞ R and |μ(x)| < ∈.
Let ba(X, R) denote the space of all charges on a ring R, equipped with the norm
n
→μ→ = sup sup |μ(E i )|.
E∞R ε E i=1
If X ∞ R, then
n
→μ→ = sup |μ(E i )|.
εX
i=1
n
→μ→ = sup |μ(E i )|.
εN
i=1
N
f ε = f (ε;n 1 , n 2 , . . . , n N ) = f (n i )μ(E i ).
i=1
n
f dμ ≥ sup | f (x)| sup |μ(E i )| ≥ →μ→ sup | f (x)|
x∞X εX x∞X
i=1
f dμ ≥ → f →.→μ→.
For our convenience, we will write M for ba(N, α∈ ), i.e., the space of bonded
finitely additive set functions (or measures) μ defined on subsets of the set of positive
integers N. It is to be remarked that the continuous dual of ∈ is not isomorphic to
a sequence spaces ([3, Example 6.4.8, pp. 93-94]. This can be attributed to the fact
that ∈ has no Schauder basis (it is not separable).
Theorem 2.3 ([1, Theorem 7.22; pp. 130]) Let F: ∩∈ → M be such that
F( f ) = μ f , μ f (E) = f (δ E ), E ∞ α∈ , (2.10)
→F( f )→ = → f →, f ∞ ∩∈ .
f μ ∞ ∩∈ (2.12)
i.e., F is surjective.
Choose a εN := {E 1 , E 2 , . . . , E N } for θ > 0 such that
N
|μ f (E i )| > →μ f → − θ,
i=1
2.1 Continuous Duals 39
N
N
f (x) = xdμ f = xdμ f = |μ f (E i )|
i=1 E i=1
i
n
→ f → ⇒ | f (x)| ⇒ |μ f (E i )| ⇒ →μ f → − θ.
i=1
z −1 ∩ Y := {x ∞ β: zx ∞ Y } .
and
⎛
n
ν
X := a ∞ β: sup ak xk < ∈ for all x ∞ X
n
k=1
⎝
:= (x −1 ∩ bs)
x∞X
are called the Köthe–Toeplitz dual (or λ-dual), generalized Köthe–Toeplitz (or φ-
dual), and bounded dual (or ν-dual) of X , respectively.
Proof (i) Let a ∞ 1 and x ∞ ∈ . Then →x→∈ = supk |xk | < ∈. Therefore
|ak xk | ≥ (sup |xk |) |ak | = →x→∈ |ak | < ∈,
k k k k
|ak xk | = |ak | < ∈,
k k
|ak xk | ≥ →a→∈ |xk | < ∈,
k k
Let a ∞ ∈ . Then there exists a strictly increasing sequence (n i ) with |an i | > i 3 . If
we define x = (xn ) by
i −2 , n = n i , i = 1, 2, . . .
xn =
0, n = n i ,
then x ∞ 1 but k |ak xk | = ∈. Thus a ∞ λ1 and so λ1 ⊕ ∈ .
Hence λ1 = ∈ and this completes the proof.
Theorem 2.5 ([2, Theorem 2.5.4]) Let 1 < p, q < ∈, with p −1 + q −1 = 1. Then
λp = q .
Proof Suppose that a ∞ λp and x ∞ q . Then k |ak xk | < ∈ for all x ∞ p . Also
let a ∞ q . Then we can find a sequence n 1 < n 2 < · · · such that
|an k |q > k 2 , k = 1, 2, . . . .
Take
1/an k , n = n k , k = 1, 2, . . .
xn =
0, n = nk .
Then x ∞ p but
|ak xk | = |an k ||an k |−1 = 1 + 1 + · · · = ∈.
k k
1/q 1/ p
|ak xk | ≥ |ak | q
|xk | p
= →a→q →x→ p < ∈,
k k k
Proof We prove the case cs λ = 1 , and the rest can be obtained similarly.
Let x ∞ cs and a ∞ 1 . Then
|ak xk | ≥ →x→bs |ak | < ∈,
k k
42 2 Matrix Transformations
where
n
→x→bs = sup xk .
n
k=1
Define
(−1)k 2−i/2 , n i < k ≥ n i+1 , i ⇒ 1
xk =
0, otherwise.
Then x = (xk ) ∞ cs but k |ak xk | = ∈. This contradicts that a ∞ cs λ , and so
cs λ ⊕ 1 .
Hence cs λ = 1 and we finish the proof.
|an k | > k, k = 1, 2, . . . .
Put
1/an k , k = n k , k = 1, 2, . . .
xk =
0, k = nk .
Then x ∞ c0 but
1
ak x k = an k = 1 + 1 + ···
an k
k k
diverges to ∈.
2.2 Köthe–Toeplitz Duals 43
φ
This contradiction shows that we must have a ∞ 1 , and so c0 ⊕ 1 .
Conversely, suppose that x ∞ c0 and a ∞ 1 . Then
ak xk ≥ →x→∈ |ak | < ∈.
k k
φ φ
This implies that a ∞ c0 , and so 1 ⊕ c0 .
φ
Hence c0 = 1 .
φ
(ii) Let a ∞ 1 . Then k ak xk converges for every x ∞ 1 . We have to show that
a ∞ ∈ . For, we may define
n
f n (x) = ak x k
k=1
Now put
sgn an , k = n
xk =
0, k = n.
ak xk ≥ →a→∈ →x→1 < ∈.
k
φ φ
Therefore a ∞ 1 and so ∈ ⊕ 1 .
φ
Hence 1 = ∈ . This completes the proof.
Theorem 2.8 ([2, Theorem 2.5.6 (a)]) Let 1 < p, q < ∈, with p −1 + q −1 = 1.
φ
Then p = q .
Proof It is similar to that of Theorem 2.5.
φ
Theorem 2.9 ([3, Theorem 7.3.5]) (i) cs φ = bv, (ii) bv φ = cs, (iii) bv0 = bs,
(iv) bs φ = bv0 .
44 2 Matrix Transformations
Proof We prove here (i), the other statements can be proved similarly.
Let a = (ak ) ∞ cs φ and z = (z k ) ∞ c0 . Then the sequence x = (xk )defined
by xk = z k − z k−1 , k ⇒ 1, where z 0 = 0, belongs to cs. Therefore, k ak xk
converges, but
n
n−1
(z k − z k−1 )ak = z k (ak − ak−1 ) + z n an
k=1 k=1
φ
and a ∞ cs φ ⊂ 1 = ∈ (since cs ∗ 1 ) imply that
∈
∈
(z k − z k−1 )ak = z k (ak − ak+1 ).
k=1 k=1
φ
Hence (ak − ak+1 )k ∞ c0 = c0λ = 1 , i.e., a ∞ bv. Therefore cs φ ⊕ bv.
let a ∞ bv. Then (ak − ak+1 )k ∞ 1 . Further, if x ∞ cs,the sequence
Conversely,
(βn ), βn = nk=1 xk , n ⇒ 1, is an element of c. As cλ = 1 , the series ∈k=1 βk (ak −
ak+1 ) is absolutely convergent. Also, we have
n
n−1
(βk − βk−1 )ak ≥ βk (ak − ak+1 ) +|βn an − βm−1 am |.
k=m k=m
As (βk ) ∞ c and (ak ) ∞ bv ⊂ c, the right-hand side of the above inequality converges
to zero as m, n → ∈. Hence the series
(βk − βk−1 )ak or ak x k
k k
n
ak βk ≥ K for all n ⇒ 1.
k=1
n
n+1
z k (ak − ak+1 ) ≥ (z k − z k−1 )ak +|z n+1 an+1 | ≥ K + M.
k=1 k=1
ν
Hence (ak − ak+1 ) ∞ c0 = 1 (by Theorem 2.10), i.e., (ak ) ∞ bv. Therefore
cs ν = bv.
X ⊂ Y, then Y † ⊂ X † ,
where † = λ, φ or ν.
(iii) X ††† = X † .
Example 2.1 (i) c is not monotone and hence not normal, since (1, 0, 1, 0, . . .) ∞ c.
(ii) c0 is normal but not perfect, since c0λλ = λ1 = ∈ .
(iii) M0 is monotone but not normal, since we can find {1/n} ∞ M0 although
1/n ≥ 1 for n ⇒ 1, and (1, 1, 1, . . .) ∞ M0 .
(iv) The spaces α, β, p (1 ≥ p < ∈) and ∈ all are perfect.
n
sup ak xk < ∈ for all x ∞ X.
n
k=1
Observe that
a¯k x¯k
xk ∞ X for any x ∞ X.
ak x k
Hence k |xk yk | < ∈ for each x ∞ X , so that a ∞ X λ , i.e., X ν ⊂ X λ . But
X λ ⊂ X ν , so that X λ = X ν , where x¯k denotes the canonical preimage of xk .
This completes the proof.
∈
∈
f (x) = xk f (e(k) ) = ak xk = (Ta )(x).
k=1 k=1
For the sake of completeness, we also include here some other type of duals. But
our emphasis is given on continuous duals and φ-duals because of their applications
in particular φ-duals are very helpful to study the
toward matrix transformations,
convergence of the series k ank xk , that is, for the existence of the A-transformed
sequence.
X f := {(g(e(k) )): g ∞ X ≤ }.
If X is a K space then α ⊂ X f .
Note that α ⊂ X ζ ⊂ X λ .
Theorem 2.15 ([1, Proposition 2.7]) For a sequence space X , X λ = X ζ if X is
symmetric.
Proof Let X be symmetric and a ∞ X λ . Then for each x ∞ X and ε ∞ ι, xε−1 ∞ X ,
so that
|ak xε−1 (k) | = |aε( j) x j | < ∈
k j
M(X, Y ) ⊂ M(M(M(X, Y ), Y ), Y ).
Xφ ⊂ X f . (2.14)
Corollary 2.22 The λ-, φ-, and ν-duals of a B K⎩space X are B K spaces with
∈
→ a →λ =→ a → X,λ = supx∞S X → ax →1 = supx∞S X | ak xk | for all a ∞ X λ ,
⎩
k=0
and → a →φ =→ a → X,φ = supx∞S X → a →bs = supx∞S X supn | nk=0 ak xk | for all
a ∞ X φ , X ν . Furthermore, X φ is a closed subspace of X ν .
Proof The first part is an immediate consequence of Theorem 2.21. Since the B K
norms on X φ and X ν are the same and X φ ⊂ X ν by Remark 2.4 (i), the second part
follows from Theorem 1.7.
Remark 2.6 Let X be any of the spaces ∈ , c, c0 , and p (1 ≥ p < ∈). Then,
→ . →∩X , → . → X,λ and → . → X,φ are equivalent on X φ , where
the norms → . → X φ ,
→ a → X = supx∞S X | ∈
∩
k=0 ak x k | .
Remark 2.8 The condition that X has AD is not necessary for X ⊂ X f f , in general.
For example, in Example 2.2, X does not have AD.
In this section, we apply the results of the theory of F K and B K spaces to characterize
the matrix classes.
Let (X, d) be a metric space, ζ > 0, and x0 ∞ X. Then, we write Bζ [x0 ] =
{x ∞ X : d(x, x0 ) ≥ ζ} for the closed ball of radius ζ with its center in x0 . If X ⊂ w
is a linear metric space and a ∞ w, then we write
∈
→ a →∩ζ =→ a →∩ζ,X = sup | ak x k | (2.15)
x∞Bζ [x0 ] k=0
provided the expression on the right-hand side exists and is finite which is the case
whenever a ∞ X φ ; if X is a normed space, we write
∈
→ a →∩X = sup | ak x k | (2.16)
x∞S X k=0
M A∩ (X, ∈ ) = sup → An →∩ .
n
φ φ
Remark 2.9 Let 1 < p < ∈ and q = p/( p−1). Then, we have ∈ = cφ = c0 = 1
φ
and p = q . Furthermore, let X denote any of the spaces ∈ , c, c0 , 1 , or p . Then,
we have → a →∩X =→ a → X φ for all a ∞ X φ , where → . → X φ is the natural norm on the
dual space X φ .
Now we give here results on matrix transformations using the theory of F K and
B K spaces.
2.5 Matrix Classes of Some FK and BK Spaces 53
is a bounded subset of X.
Proof Let X contain 1 . Then the inclusion map γ: 1 → X is continuous. Since
e(k) : k = 0, 1, 2, . . . is bounded in 1 , it is bounded in X.
Conversely, suppose that (2.17) holds. Let x = (xk ) ∞ 1 and let q be a continuous
seminorm on X. Then
n
n
(k)
q xk e ≥ | xk | q(e(k) ).
k=m k=m
∈
Since x = (xk ) ∞ 1 , i.e., k=0 |⎫ x k | is convergent,
⎬ we have
n (k)
n−1 (k)
k=m | x k | q(e ) → 0 (m, n → ∈). Thus k=0 x k e is a Cauchy sequence
n
convergent to x in X. Moreover, x ∞ X since X is complete. Hence 1 ⊂ X.
This completes the proof.
We derive the following corollary:
Corollary 2.26 Let A = (ank ) be an infinite matrix and X an F K space. Then,
A ∞ (1 , X ) if and only if the columns of A belong to X and form a bounded subset
of X.
Now from this corollary, we easily deduce the following classical results of sum-
mability theory.
If we put X = c, then we get:
Corollary 2.27 A ∞ (1 , c) if and only if (i) supn,k | ank |< ∈, and (ii) limn ank
exists for all k.
If we put X = 1 , then we get:
Corollary 2.28 A ∞ (1 , 1 ) if and only if supk n | ank |< ∈.
If we put X = p , then we get:
Corollary 2.29 A ∞ (1 , p ) if and only if supk ( n | ank | p )1/ p < ∈.
If we put X = ∈ then we get:
Corollary 2.30 A ∞ (1 , ∈ ) if and only if supn,k | ank |< ∈.
The following result is one of the most important in matrix transformations:
Theorem 2.31 [3, Theorem 4.2.8] Any matrix map between F K spaces is
continuous.
Proof Let X and Y be FK spaces, A ∞ (X, Y ) and the map f A : X → C be defined
by f A (x) = Ax for all x ∞ X . Since the maps Pn ◦ f A : X → C are continuous for
all n ∞ N0 by Theorem 1.6, the linear map f A is continuous by Corollary 1.5.
This completes the proof.
54 2 Matrix Transformations
where An = (ank )∈
k=1 denotes the sequence in the n-th row of the matrix A.
Proof First, we assume that (2.18) is satisfied. Then the series An (x) converge for
all x ∞ Bζ [0] and for all n, and Ax ∞ ∈ for all x ∞ Bζ [0]. Since Bζ [0] is absorbing
by Remark 1.3, we conclude that the series An (x) converge for all n and all x ∞ X ,
and Ax ∞ ∈ for all x ∞ X, i.e., A ∞ (X, ∈ ).
Conversely, suppose that A ∞ (X, ∈ ). Then the map L A : X → ∈ defined
by L A (x) = Ax for all x ∞ X is continuous by Theorem 2.31. Hence there exist a
neighborhood U of 0 in X and a real ζ > 0 such that Bζ [0] ⊂ U and → L A (x) →∈ < 1
for all x ∞ X . This implies (2.18).
This completes the proof.
Theorem 2.33 [7, Theorem 3.20] Let X and Y be B K spaces. Then, we have
(a) (X, Y ) ⊂ B(X, Y ), that is, every matrix A ∞ (X, Y ) defines an operator
L A ∞ B(X, Y ) by L A (x) = Ax for all x ∞ X .
(b) If X has AK , then B(X, Y ) ⊂ (X, Y ), that is, for every operator L ∞ B(X, Y )
there exists a matrix A ∞ (X, Y ) such that L(x) = Ax for all x ∞ X .
(c) A ∞ (X, ∈ ) if and only if
if A ∞ (X, ∈ ), then
→ A →(X,∈ ) =→ L A → . (2.20)
for all x ∞ X with → x →= 1. Thus | An (x) |≥→ L A → for all n and for all x ∞ X with
→ x →= 1, and, by the definition of the norm → A →(X,∈ ) ,
2.5 Matrix Classes of Some FK and BK Spaces 55
Hence A ∞ (X, Y1 ).
(b) First, we assume A ∞ (X 1 , Y ). Then X ⊂ X 1 implies A ∞ (X, Y ), and e ∞ X 1
implies Ae ∞ Y.
Conversely, we assume A ∞ (X, Y ) and Ae ∞ Y. Let x1 ∞ X 1 be given. Then
there are x ∞ X and λ ∞ C such that x1 = x + λe, and it follows that Ax1 =
A(x + λe) = Ax + λ Ae ∞ Y.
This completes the proof.
If A ∞ (X, 1 ), then
→A→(X,1 ) ≥ →L A → ≥ 4.→A→(X,1 ) , (2.24)
where
→A→(X,1 ) = sup → An →∩X ,
N ∞F n∞N
m
| An (x) |≥ | An (x) |≥ →L A →,
n∞N n=0
m(x)
ε
| An (x) |⇒ →A(x)→1 − .
2
n=0
Consequently
m(x)
| An (x) |⇒ →L A → − ε.
n=0
Remark 2.10 Since the B K spaces c0 and c are closed subspaces of ∈ , the matrix
classes (X, c0 ) and (X, c) can be characterized by combining Theorem 2.33 (c) and
Theorem 2.34 (a), where X is a B K space with Schauder basis. On the other hand,
we may note that if X , in Theorem 2.33 (c) or Theorem 2.34 (a), is any of the classical
sequence spaces, then any of (2.19), (2.20), or (2.23) implies the condition ‘ An ∞ X φ
for all n ∞ N’ by Remark 2.9. Thus, this condition is redundant in such cases. Also,
if X is a B K space with AK , then we obtain the following result which is immediate
by Propositions 3.2 and 3.3 of [9].
Remark 2.11 The results of the previous sections yield the characterizations of the
classes (X, Y ) where X and Y are any of the spaces p (1 ≥ p < ∈), c0 , c with the
exceptions of ( p , r ) where both p, r = 1, ∈ (the characterizations are unknown),
and of (∈ , c) (Schur’s theorem) and (∈ , c0 ) for which no functional analytic proofs
seem to be known. The class (2 , 2 ) was characterized in [10].
Theorem 2.38 [3, Theorem 8.3.8, pp. 124 ] Let X be an F K space and Y be any
set of sequences. If A ∞ (X, Y ) then A T ∞ (Y φ , X f ). If X and Y are B K spaces
and Y φ has AD then we have A T ∞ (Y φ , cl X f (X φ )).
Now we assume that X and Y are B K spaces and Y has AD. Then Y φ ⊂ X f
by Theorem 2.20 (a), and X f is a B K space by Theorem 2.23. Also cl X f (X φ ) is
a closed subspace of X f . Since A ∞ (X, Y ), we have An = (ank )∈ φ
k=0 ∞ X for all
(n)
∈ (k) ∈ ∈ φ
n, but A e = ( j=0 a jk e j )k=0 = (ank )k=0 = An ∞ X for all n. So we have
T
Theorem 2.39 [3, Theorem 8.3.9, pp. 124 ] Let X and Z be B K spaces with AK
and Y = Z φ . Then we have (X, Y ) = (X φφ , Y ); furthermore A ∞ (X, Y ) if and
only if A T ∞ (Z , X φ ).
Theorem 2.40 [7, Example 5.4] (a) We have (c0 , ∈ ) = (c, ∈ ) = (∈ , ∈ );
furthermore A ∞ (∈ , ∈ ) = B(∈ , ∈ ) if and only if
∈
→ A →(∈ ,∈ ) = sup | ank |< ∈. (2.26)
n
k=0
If A ∞ (c0 , c) then
∈
lim An (x) = ank xk . (2.28)
n→∈
k=0
2.6 Conservative, Regular, and Schur Matrices 59
(c) A ∞ (c0 , c0 ) if and only if (2.26) and (2.27) with λk = 0 for all k hold.
Proof (a) We have A ∞ (c0 , ∈ ) if and only if (2.26) holds by Theorem 2.33, and
φ
since c0 = 1 and c0∩ and 1 are norm isomorphic.
Furthermore c0 ⊂ c ⊂ ∈ implies (c0 , ∈ ) ∗ (c, ∈ ) ∗ (∈ , ∈ ). Also
φφ
(c0 , ∈ ) = (c0 , ∈ ) = (∈ , ∈ ) by the first part of Theorem 2.39. Also for all
x ∞ ∈ , we have
→ A(x) →∈ ≥→ A →(∈ ,∈ ) → x →∈ ,
i.e., A is a bounded linear operator. The last part is obvious from Theorem 2.33.
(b) Since c is a closed subspace of ∈ , the characterization of the class (c0 , c) is
an immediate consequence of Theorem 2.34 (a) and Part (a).
Now we assume A ∞ (c0 , c) and write → A →=→ A →(∈ ,∈ ) , for short. Let
m
mbe a given non-negative minteger. Then it follows from (2.27) and (2.26) that
k=0 | λk |= lim n→∈ k=0 | ank |≥→ A → . Since m was arbitrary, we have
(λk )∈k=0 ∞ 1 ,
∈
∈
| λk |≥→ A → and | λk xk |≥→ A →→ x →∈ for all x ∞ c. (2.29)
k=0 k=0
Now let x ∞ c0 and ε > 0 be given. Then we can choose an integer k(ε) such that
| xk |≥ ε/(4 → A → +1) for all k > k(ε), and by (2.27) we can choose and integer
k(ε)
n(ε) such that k=0 | ank − λk || xk |< ε/2. Let n > n(ε). Then (2.26) and (2.29)
imply
∈
k(ε) ∈
| An (x) − λk xk |≥ | ank − λk | + (| ank | + | λk |) | xk |
k=0 k=0 k=k(ε)+1
∈ ∈
ε ε ε ε
≥ + ( | ank | + | λk |) ≥ + = ε.
2 4 → A → +1 2 2
k=0 k=0
Theorem 2.41 [7, Example 5.5] We have (1 , 1 ) = B(1 , 1 ) and A ∞ (1 , 1 ) if
and only if
∈
→ A →(1 ,1 ) = sup | ank |< ∈. (2.30)
k n=0
Proof Since 1 has AK , Theorem 2.33 (b) yields the first part.
60 2 Matrix Transformations
Theorem 2.43 [7, Theorem 6.11] We have L ∞ B(c, c) if and only if there exists a
matrix A ∞ (c0 , c) and a sequence b ∞ ∈ with
∈
lim bn + ank = λ̃ exists (2.33)
n→∈
k=0
such that
L(x) = b lim xk + Ax for all x ∞ c. (2.34)
k→∈
2.6 Conservative, Regular, and Schur Matrices 61
Furthermore, we have
∈
→ L →= sup | bn | + | ank | . (2.35)
n
k=0
with
∈
bn = L n (e) − L n (e(k) ) and ank = L n (e(k) ) for k = 0, 1, . . . .
k=0
and
∈
→ L n →=| bn | + | ank | . (2.37)
k=0
Now (2.36) yields (2.34). Since L(x0 ) = Ax0 for all x0 , we have A ∞ (c0 , c), and
so → A →= supn ∈ k=0 | ank |< ∈ by Theorem 2.40 (b). Also L(e) = b + Ae
implies (2.33), and we obtain → b →∈ ≥→ L(e) →∈ + → A →, that is, b ∞ ∈ .
Consequently we have C= supn (| bn | + ∈ k=0 | ank |) < ∈. Now → L(x) →∈ =
supn | bn limk→∈ xk + ∈ a x
k=0 nk k |≥ (supn (| bn | + ∈
k=0 | ank |)) → x →∈ < ∈
implies → L →≥ C. We also have | L n (x) |≥→ L(x) →∈ ≥ → L → for all x ∞ B̄c and
all n, and so supn → L n →= C ≥→ L →. Thus (2.35) is proved.
Conversely we assume that A ∞ (c0 , c) and b ∞ ∈ satisfy (2.33). Since A ∞
(c0 , c) and b ∞ ∈ , we obtain C < ∈ by (2.26), and so L ∞ B(c, ∈ ). Finally
let x ∞c be given and lim k→∈ x k = ξ. Then we have x − ξe ∞ c0 , L n (x) =
bn ξ + ∈ a x
k=0 nk k = (b n + ∈
k=0 ank )ξ + An (x − ξe) for all n, and it follows from
(2.33) and A ∞ (c0 , c) that limn L n (x) exists. Since x ∞ c was arbitrary, we have
L ∞ B(c, c).
This completes the proof.
First we state the following lemma which is needed in proving Schur’s theorem.
To the best of our knowledge it seems that the functional analytic proof of Schur’s
theorem does not exist yet.
Lemma
B = (bnk )n,k be an infinite matrix
2.44 [2, Theorem 3.3.7] Let such that
k | b nk |< ∈ for each n and k | bnk |→ 0 (n → ∈). Then k | bnk |
converges uniformly in n.
Proof k | bnk |→ 0 (n → ∈) implies that k | bnk |< ∈ for n ⇒ N (ε).
Since k | bnk |< ∈ for 0 ≥ n ≥ N (ε), there exists m = M(ε, n) such that
62 2 Matrix Transformations
k⇒M | bnk |< ∈ for all n, which means that k | bnk | converges uniformly in
n.
This completes the proof.
Theorem 2.45 (Schur) [2, Theorem 3.3.8] A ∞ (∈ , c) if and only if (2.27) holds
and
| bnk | converges uniformly in n. (2.38)
k
Proof Suppose that the conditions (2.27) and (2.38) hold, and x ∞
∈ . Then,
k a nk x k is absolutely and uniformly convergent in n ∞ N. Hence, k ank x k →
k λk x k (n → ∈) which gives that A ∞ (∈ , c).
Conversely, suppose that A ∞ (∈ , c) and x ∞ c. Then necessity of (2.27) follows
easily
by taking x = e(k) for each k. Define bnk = ank − λk for all k, n ∞ N. Since
k | λk |< ∈, ( k bnk x k )n converges whenever x = (x k ) ∞ ∈ . Now if we can
show that this implies
lim | bnk |= 0, (2.39)
n
k
then by using
Lemma 2.44, we shall get the desired result. Suppose to the contrary
that limn k | bnk | = 0. Then, it follows that limn k | bnk |= l > 0 through some
subsequence of the positive integers. Also we have bmk → 0 as m → ∈ for each
k ∞ N. Hence we may determine m(1) such that
| | bm(1),k | −l |< l/2 and bm(1),1 < l/2.
k
Since k | bm(1),k |< ∈ we may choose k(2) > 1 such that
∈
| bm(1),k |< l/2 .
k=k(2)+1
It follows that
k(2)
| | bm(1),k | −l |< l/2.
k=2
q
For our convenience we use the notation k= p | bmk |= B(m, p, q).
Now we choose m(2) > m(1) such that | B(m(2), 1, ∈) − l |< l/10 and
B(m(2), 1, k(2)) < l/10. Then choose k(3) > k(2) such that | B(m(2), k(3) +
1, ∈) − l |< l/10. It follows that | B(m(2), k(2) + 1, k(3)) − l |< 3l/10. Con-
tinuing in this way and find m(1) < m(2) < · · · , 1 = k(1) < k(2) < · · · such
that
B(m(r ), 1, k(r )) < l/10
B(m(r ), k(r + 1) + 1, ∈) < l/10 (2.40)
| B(m(r ), k(r ) + 1, k(r + 1)) − l |< 3l/10
2.6 Conservative, Regular, and Schur Matrices 63
which isa multiplicative linear functional. The numbers limn→∈ ank and
limn→∈ k ank are called the characteristic numbers of A. A matrix A is called
coregular if δ(A) = 0 and is called conull if δ(A) = 0.
Remark 2.13 The Silverman–Toeplitz theorem yields for a regular matrix A that
δ(A) = 1 which leads us to the fact that regular matrices form a subset of coregular
matrices. One can easily see for a Schur matrix A that δ(A) = 0 which says us that
coercive matrices form a subset of conull matrices. Hence, we have the following
result which is known as Steinhaus’s theorem.
Theorem 2.47 (Steinhaus) [2, Theorem 3.3.14] For every regular matrix A, there
is a bounded sequence which is not summable by A.
Theorem 2.48 [12, Corollary, pp. 225] Weak and strong convergence coincide in 1 .
Proof We assume that the sequence (x (n) )∈ n=0 is weakly convergent to x in 1 , that
is, | f (x (n) ) − f (x) |→ 0 (n → ∈) for every f ∞ ∩1 . Since ∩1 and ∈ are
norm isomorphic, to every f ∞ ∩1 there corresponds a sequence a ∞ ∈ such that
(n)
f (y) = ∈ ∈
k=0 ak yk . We define the matrix B = (bnk )n,k=0 by bnk = x k −x k (n, k =
(n) ∈
0, 1, . . .). Then we have f (x (n) ) − f (x) = ∈ k=0 ak (x k − x k ) = k=0 bnk ak → 0
(n → ∈) for all a ∞ ∈ , that is, B ∞ (∈ , c0 ), and it follows from Corollary 2.46
(n) ∈
that → x (n) − x →1 = ∈ k=0 | x k − x k |= k=0 | bnk |→ 0 (n → ∈).
This completes the proof.
In this section, we characterize the classes (X, YT ) and (X, Y[T ] ) for triangles T.
Theorem 2.49 [5, Theorem 3.8] Let T be a triangle.
(a) Then, for arbitrary subsets X and Y of w, A ∞ (X, YT ) if and only if B =
T A ∞ (X, Y ).
(b) If X and Y are B K spaces and A ∞ (X, YT ), then
→ L A →=→ L B → . (2.43)
For the characterization of the class (X, Y[T ] ), we need the following lemma due
to Peyerimhoff [13].
Lemma 2.50 Let a0 , a1 , . . . , an ∞ C . Then
n
| ak |≥ 4. max | ak | .
N ⊂{0,1,...,n}
k=0 k∞N
n
| ak |=| ak | + | ak |≥ 2. max | ak | . (2.47)
N ⊂{0,1,...,n}
k=0 k∞N + k∞N − k∞N
n
| ak |≥ | λk | + | | φk |≥ 2(| x Nr | + | y Ni |)
k=0 k∞N + k∞N −
≥ 4. | z N∩ |= 4. max | ak | .
N ⊂{0,1,...,n}
k∞N
∈
∈
| SmN (x) |≥| N
smk xk |=| bmn ank xk |≥| ym | (m = 0, 1, . . .)
k=0 n∞Nm k=0
for all N ∞ N together imply S N (x) ∞ Y for all N ∞ N , since Y is normal. Hence
S N ∞ (X, Y ) for all N ∞ N .
Conversely, let S N ∞ (X, Y ) for all N ∞ N . Then SmN ∞ X φ for all m and for
all N ∞ N , in particular, for N = (m)∈ φ
m=0 , Sm = bmm Am ∞ X , hence Am ∞ X ,
N φ
since bmm = 0. Further, let x ∞ X be given. For every m = 0, 1, . . ., choose the set
(0)
Nm ⊂ {0, . . . , } such that
| bmn An (x) |=| max bmn An (x) | .
Nm ⊂{0,1,...,m}
(0)
n∞Nm
(0)
Hence by hypothesis, S N (x) ∞ Y, and the normality of Y implies y =
B(| A(x) |) ∞ Y, that is, A ∞ (X, Y[B] ).
This completes the proof.
(X T )φ ⊂ (X φ ) R .
Theorem 2.53 ([14, Theorem 29]) (a) Let X be an F K space with AK and Z = X T .
We write R = S t for the transpose of S. Then, a ∞ (X T )φ if and only if
Theorem 2.54 ([14, Remark 30]) We have a ∞ (cT )φ if and only if a ∞ (1 ) R and
W ∞ (c, c). Moreover, if a ∞ (cT )φ , then we have
∈
∈
ak z k = Rk (a)Tk (z) − ξλ for all z ∞ cT , (2.50)
k=0 k=0
m
where ξ = limk→∈ Tk (z) and λ = limm→∈ k=0 wmk .
Moreover, if A ∞ (X T , Y ) then,
Theorem 2.56 [14, Remark 32] Let Y be a linear subspace of β. Then, we have
A ∞ (cT , Y ) if and only if
and
m
(n)
Â(e) − (λn )∈
n=0 ∞ Y, where λn = lim wmk for all n. (2.53)
m→∈
k=0
Theorem 2.57 [14, Theorem 33] Let T be a triangle. Let X and Y be B K spaces
and X have AK . If A ∞ (X T , Y ), then
→ L A →=→ L Â →, (2.55)
Exercises
1. Prove that ∩p q , i.e., the continuous dual of p is q for 1 < p < ∈,
p −1 + q −1 = 1.
2. Show that (i) cs ∩ bv, (ii) bv ∩ bs.
3. Show that (i) β λ = χ, (ii) χλ = β, where
β := x ∞ β: lim |xn |1/n = 0 ,
n
⎧
χ := x ∞ β: sup |xn | 1/n
<∈ .
n
φ
4. Prove (i) bv φ = cs, (ii) bv0 = bs, and (iii) bs φ = bv0 .
φ ν
5. Show that m λ0 = 1 = m 0 = m 0 , where m 0 = sp{E}, E is the set of sequences
of zeros and ones.
6. Show that P λ = 1 , P φ = cs, P ν = bs, where
References
1. Kamthan, P.K., Gupta, M.: Sequence Spaces and Series. Marcel Dekker Inc., New York (1981)
2. Başar, F.: Summability Theory and Its Applications. Bentham Science Publishers, Istanbul
(2011)
3. Wilansky, A.: Summability Through Functional Analysis, North-Holland Mathematics Studies,
vol. 85. Elsevier Science Publishers, Amesterdam (1984)
4. Boos, J.: Classical and Modern Methods in Summability. Oxford University Press, New York
(2000)
5. Malkowsky, E., Rakočević, V.: An introduction into the theory of sequence spaces and measures
of noncompactness. Zbornik Radova, Mat. Institut SANU (Beograd) 9(17), 143–234 (2000)
6. Mursaleen, M.: Elements of Metric Spaces. Anamaya Publishers, New Delhi (2005)
7. Malkowsky, E.: Modern functional analysis in the theory of sequence spaces and matrix trans-
formations. Jordan J. Math. Stat. 1(1), 1–29 (2008)
8. Malkowsky, E.: Klassen von matrix abbildungen in paranormierten F K -Raumen. Analysis 7,
275–292 (1987)
9. Malkowsky, E., Rakočević, V., Zivkovic, S.: Matrix transformations between the sequence
space bv p and certain B K spaces. Bull. Acad. Serbe Sci. Arts Math. 123(27), 33–46 (2002)
10. Crone, L.: A characterization of matrix mappings on 2 . Math. Z. 123, 315–317 (1971)
11. Cooke, R.G.: Infinite Matrices and Sequence Spaces. Dover Publication Inc., New York (1955)
12. Maddox, I.J.: Elements of Functional Analysis, 2nd edn. The University Press, Cambridge
(1988)
13. Peyerimhoff, A.: Über ein Lemma von Herrn Chow. J. London Math. Soc. 32, 33–36 (1957)
14. Malkowsky, E.: Compact matrix operators between some B K spaces. In: Mursaleen, M. (ed.)
Modern Methods of Analysis and Its Applications, pp. 86–120. Anamaya Publishers, New
Delhi (2010)
Chapter 3
Some New Sequence Spaces of Non-absolute
Type
In the present chapter, we study some sequence spaces of non-absolute type as matrix
domains of classical sequence spaces c0 , c, ∞ and p (1 ∈ p < ∞). We establish
some inclusion relations concerning with those spaces and determine their φ-, λ-,
and ε-duals. Finally, we characterize some related matrix classes. Some of the results
of this chapter appeared in [1–4].
1
n
lim (αk − αk−1 )|xk − a| = 0.
n→∞ αn
k=0
Finally, we define the associated sequence y(α) = {yk (α)}, which will be fre-
quently used, as the β-transform of a sequence x = (xk ), i.e., y(α) = β(x) and so
we have
k
αj − αj−1
yk (α) = xj ; (k → N). (3.5)
αk
j=0
And hence,
k
αj
xk (α) = (−1)k−j yj ; (k → N). (3.6)
αk − αk−1
j=k−1
Here and in the sequel, we shall use the convention that any term with a negative
subscript is equal to zero, e.g., α−1 = 0 and x−1 = 0.
Now we define the sequence spaces α∞ , cα , c0α , and αp as the sets of all
α-bounded, α-convergent, α-null sequences, and α( p)-absolutely convergent series,
respectively, as follows:
α∞ = x → w: sup |βn (x)| < ∞ ,
n
cα = x → w: lim βn (x) exists ,
n→∞
c0α = x → w: lim βn (x) = 0 .
n→∞
3.1 α–sequence Spaces 73
and
∞
αp = x = (xk ) → w: |βn (x)| p < ∞ ; (0 < p < ∞)
n=0
We can redefine the spaces α∞ , cα , c0α , and αp as the matrix domains of the triangle
β in the spaces ∞ , c, c0 , and p , respectively, that is,
α∞ = (∞ )β , cα = cβ , c0α = (c0 )β and αp = ( p )β (0 < p < ∞), (3.7)
for all n, k → N. Then, for any sequence x = (xk ) → w, the β-transform of x is the
sequence β(x) = {βn (x)} , where βn (x) is given by (3.2) for all n → N.
Remark 3.1 It can easily be seen that the absolute property does not hold on the
spaces α∞ , cα , c0α , and αp , that is, ∩x∩α∞ = ∩ | x | ∩α∞ for at least one sequence x
in each of these spaces, where |x| = (|xk |). Thus, the spaces α∞ , cα , c0α , and αp are
BK spaces of non-absolute type.
We observe the following which are easy to prove.
Lemma 3.1 Every convergent sequence is α-convergent to the same ordinary limit.
Lemma 3.2 If a α-convergent sequence converges in the ordinary sense, then it must
converge to the same α-limit .
Lemma 3.3 A α-convergent sequence x converges in the ordinary sense if and only
if S(x) → c0 .
Lemma 3.4 Every bounded sequence is α -bounded.
Lemma 3.5 A α-bounded sequence x is bounded in the ordinary sense if and only
if S(x) → ∞ .
Lemma 3.6 The sequence spaces α∞ , cα , and c0α are BK spaces with the same norm
given by ∩x∩α∞ = ∩β(x)∩∞ , i.e.,
The space αp (1 ∈ p ∈ ∞) is a BK space with the norm ∩x∩αp = ∩β(x)∩ p , that is,
1/ p
∩x∩αp = |βn (x)| p ; (1 ∈ p < ∞) (3.10)
n
74 3 Some New Sequence Spaces of Non-absolute Type
Lemma 3.7 The sequence spaces α∞ , cα , c0α , and αp are norm isomorphic to the
spaces ∞ , c, c0 , and p , respectively, that is, α∞ ∼
= ∞ , c α ∼
= c, c0α ∼
= c0 , and
α ∼
p = p (0 < p ∈ ∞).
Proof Let X denote any of the spaces ∞ , c, or c0 and X α be the respective one of the
spaces α∞ , cα , or c0α . Since the matrix β is a triangle, it has a unique inverse which
is also a triangle [6, Proposition 1.1]. Therefore, the linear operator L β :X α → X ,
defined by L β (x) = β(x) for all x → X α , is bijective and is norm preserving by
(3.9). Hence, X α ∼ = X . To prove αp ∼ = p , we should show the existence of a linear
bijection between the spaces αp and p , where 0 < p ∈ ∞. For that, let 0 < p ∈ ∞
and consider the transformation T defined, with the notation of (3.5), from αp to p
by x ≥ −→ y(α) = T x. Then, we have T x = y(α) = β(x) → p for every x → αp .
Also, the linearity of T is trivial. Further, it is easy to see that x = 0 whenever
T x = 0, and hence, T is injective.
Furthermore, let y = (yk ) → p and define the sequence x = {xk (α)} by (3.6).
Then, we have for every n → N that
n
1
βn (x) = (αk − αk−1 )xk (α)
αn
k=0
n k
1
= (−1)k−j αj yj
αn
k=0 j=k−1
n
1
= (αk yk − αk−1 yk−1 ) = yn .
αn
k=0
This shows that β(x) = y, and since y → p , we obtain that β(x) → p . Thus,
we deduce that x → αp and T x = y. Hence, T is surjective.
Moreover, let x → αp . Then, we have by (3.10) that
This shows that T is p-norm and norm preserving in the cases of 0 < p < 1 and
1 ∈ p ∈ ∞, respectively. Hence, T is a linear bijection. Consequently, the spaces
αp and p are linearly isomorphic for 0 < p ∈ ∞.
This completes the proof.
3.1 α–sequence Spaces 75
(k)
Lemma 3.8 Define the sequence eα → c0α for every fixed k → N by
αk
n−k
(−1)
αn − αn−1
; (k ∈ n ∈ k + 1),
(k)
(eα )n = (n → N) (3.11)
⎪
0; (otherwise).
Then, we have
(0) (1)
(a) The sequence (eα , eα , . . .) is a Schauder basis for the space c0α , and every
(k)
x → c0α has a unique representation x = ∞ k=0 βn (x) eα .
(0) (1)
(b) The sequence (e, eα , eα , . . .) is a Schauder basis for the space cα , and
(k)
every x → cα has a unique representation x = le + ∞ k=0 (βk (x) − l)eα , where
l = limk→∞ βk (x).
(0) (1)
(c) The sequence (eα , eα , . . .) is a Schauder basis for the space αp , and every
x → αp has a unique representation of the form
x= βk (x) eα(k) . (3.12)
k
(k)
Proof (a) and (b) follow immediate, since β(e) = e and β(eα ) = e(k) for all k.
Let 1 ∈ p < ∞. Then, it is obvious by (3.11) that β(eα(k) ) = e(k) → p (k → N),
and hence, eα(k) → αp for all k.
Further, let x → αp be given. For every non-negative integer m, we put
m
(k)
x (m) = βk (x) eα .
k=0
m
m
β(x (m) ) = βk (x) β(eα(k) ) = βk (x) e(k)
k=0 k=0
and hence
0; (0 ∈ n ∈ m),
(m)
βn (x − x )= (n, m → N)
⎪
βn (x); (n > m).
∞
δ p
|βn (x)| p ∈ .
2
n=m 0 +1
n=m+1
⎛ ⎧1/ p
∞
∈⎝ |βn (x)| p ⎨
n=m 0 +1
δ
∈ <δ
2
and hence limm→∞ ∩x − x (m) ∩αp = 0. This shows that x → αp is represented as in
(3.12).
Finally, let us show the uniqueness of the representation (3.12) of x → αp . For
this, suppose that (k)
x= φk (x) eα .
k
In this section, we establish some inclusion relations concerning with the spaces c0α ,
cα , and α∞ , and we may begin with the following basic result:
Theorem 3.1 The inclusions αp ⇒ c0α ⇒ cα ⇒ α∞ strictly hold, where 0 < p < ∞.
Proof It is clear that the inclusions c0α ⇒ cα ⇒ α∞ hold. Further, since the inclusion
c0 ⇒ c is strict, it follows by Lemma 3.1 that the inclusion c0α ⇒ cα is also strict.
It is trivial that the inclusion αp ⇒ c0α holds for 0 < p < ∞, since x → αp implies
β(x) → p and hence β(x) → c0 which means that x → c0α . Further, to show that this
inclusion is strict, let 0 < p < ∞ and consider the sequence x = (xk ) defined by
3.2 Some Inclusion Relations 77
1
xk = ; (k → N).
(k + 1)1/ p
Then x → c0 and hence x → c0α , since the inclusion c0 ⇒ c0α holds. On the other hand,
we have for every n → N that
n
1 αk − αk−1
|βn (x)| =
αn (k + 1)1/ p
k=0
n
1
≤ (αk − αk − 1)
αn (n + 1)1/ p
k=0
1
= .
(n + 1)1/ p
This shows that β(x) → / p and hence x →/ αp . Thus, the sequence x is in c0α but
α α α
not in p . Therefore, the inclusion p ⇒ c0 is strict for 0 < p < ∞.
Similarly, it is also clear that the inclusion cα ⇒ α∞ holds. To show that this
inclusion is strict, we define the sequence y = (yk ) by
⎩
αk + αk−1
yk = (−1)k ; (k → N).
αk − αk−1
This shows that β(y) → ∞ \c. Thus, the sequence y is in α∞ but not in cα and hence
cα ⇒ α∞ is a strict inclusion.
This completes the proof.
Now, the following result is immediate by the regularity of the matrix β and by
Lemma 3.3.
Theorem 3.2 The inclusions c0 ⇒ c0α and c ⇒ cα hold. Furthermore, the equalities
hold if and only if S(x) → c0 for every sequence x in the spaces c0α and cα , respectively.
Proof The first part is obvious by Lemma 3.1. Thus, we turn to the second part. For
this, suppose firstly that the equality c0α = c0 holds. Then, we have for every x → c0α
that x → c0 and hence S(x) → c0 by Lemma 3.3.
Conversely, let x → c0α . Then, we have by the hypothesis that S(x) → c0 . Thus,
it follows, by Lemma 3.3 and then Lemma 3.2, that x → c0 . This shows that the
inclusion c0α ⇒ c0 holds. Hence, by combining the inclusions c0α ⇒ c0 and c0 ⇒ c0α ,
we get the equality c0α = c0 .
78 3 Some New Sequence Spaces of Non-absolute Type
Similarly, one can show that the equality cα = c holds if and only if S(x) → c0
for every x → cα .
This concludes the proof.
Moreover, the following result can be proved similarly by means of Lemmas 3.4
and 3.5.
Theorem 3.3 The inclusion ∞ ⇒ α∞ holds. Furthermore, the equality α∞ = ∞
holds if and only if S(x) → ∞ for every x → α∞ .
Now, it is obvious by Theorem 3.2 that c0 ⇒ c0α ⊂ c. Conversely, it follows by
Lemma 3.2 that c0α ⊂ c ⇒ c0 . This yields the following result:
Theorem 3.4 The equality c0α ⊂ c = c0 holds.
It is worth mentioning that the equality cα ⊂∞ = c need not be held. For example,
let αk = k + 1 and xk = (−1)k for all k. Then x → cα ⊂ ∞ while x → c. Now, let
x = (xk ) → w and n ≤ 1. Then,
n
1
Sn (x) = αk−1 (xk − xk−1 )
αn
k=1
1
n n
= αk−1 xk − αk−1 xk−1
αn
k=1 k=1
1
n n−1
= αk−1 xk − αk x k
αn
k=0 k=0
1
n−1
= αn−1 xn − (αk − αk−1 )xk
αn
k=0
αn−1
= xn − βn−1 (x)
αn
αn−1
= Sn (x) + βn (x) − βn−1 (x) .
αn
αn−1
Sn (x) = βn (x) − βn−1 (x) ; (n → N). (3.13)
αn − αn−1
On the other hand, by taking into account the definition of the sequence α given
by (3.1), we have αk+1 /αk > 1 for all k → N . Thus, there are only two distinct cases
of the sequence α, either lim inf k→∞ αk+1 /αk > 1 or lim inf k→∞ αk+1 /αk = 1.
Obviously, the first case holds if and only if lim inf k→∞ (αk+1 −αk )/αk+1 > 0 which
is equivalent to say that the sequence (αk /(αk − αk−1 ))∞ k=0 is a bounded sequence.
Similarly, the second case holds if and only if the above sequence is unbounded.
3.2 Some Inclusion Relations 79
Theorem 3.5 For any sequence α = (αk )∞ k=0 satisfying (3.1), we have
αk ∞ αk+1
(a) →
/ ∞ if and only if lim inf = 1.
α −α k=0 k→∞ αk
k α k−1 ∞ αk+1
k
(b) → ∞ if and only if lim inf > 1.
αk − αk−1 k=0 k→∞ αk
It is clear that Theorem 3.5 still holds if the sequence (αk /(αk − αk−1 ))∞ k=0 is
replaced by (αk /(αk+1 − αk ))∞ k=0 .
Now, we are going to prove the following result which gives the necessary and
sufficient condition for the matrix β to be stronger than convergence and boundedness
both, i.e., for the inclusions c0 ⇒ c0α , c ⇒ cα and ∞ ⇒ α∞ to be strict.
Theorem 3.6 The inclusions c0 ⇒ c0α , c ⇒ cα and ∞ ⇒ α∞ strictly hold if and
only if lim inf n→∞ αn+1 /αn = 1.
Proof Suppose that the inclusion ∞ ⇒ α∞ is strict. Then, Theorem 3.3 implies
the existence of a sequence x → α∞ such that S(x) = (Sn (x))∞ n=0 → ∞ . Since
x → α∞ , we have β(x) = (βn (x))∞ n=0 → ∞ and hence (βn (x) − βn−1 (x))n=0 →
∞
∞
∞ . Therefore, we deduce from (3.13) that (αn−1 /(αn − αn−1 ))n=0 → ∞ and
hence (αn /(αn − αn−1 ))∞ n=0 → ∞ . This leads us with Theorem 3.5 (a) to the
consequence that lim inf n→∞ αn+1 /αn = 1. Similarly, by using Theorem 3.2 instead
of Theorem 3.3, it can be shown that if the inclusions c0 ⇒ c0α and c ⇒ cα are strict,
then lim inf n→∞ αn+1 /αn = 1. This proves the necessity of the condition.
To prove the sufficiency, suppose that lim inf n→∞ αn+1 /αn = 1. Then, we have
by Theorem 3.5 (a) that (αn /(αn − αn−1 ))∞n=0 → ∞ . Let us now define the sequence
x = (xk ) by xk = (−1)k αk /(αk − αk−1 ) for all k. Then, we have for every n → N
that
n
1 n
k 1
|βn (x)| = (−1) αk ∈ (αk − αk−1 ) = 1
αn αn
k=0 k=0
which shows that β(x) → ∞ . Thus, the sequence x is in α∞ but not in ∞ . There-
fore, by combining this with the fact that the inclusion ∞ ⇒ α∞ always holds by
Theorem 3.3, we conclude that this inclusion is strict.
Similarly, if lim inf k→∞ αk+1 /αk = 1 then we deduce from Theorem 3.5 (a)
that lim inf k→∞ (αk − αk−1 )/αk = 0. Thus, there is a subsequence (αkr )r∞=0 of the
sequence α = (αk )∞ k=0 such that
α − α
kr kr −1
lim = 0. (3.14)
r →∞ αkr
for all k → N. Then y → c. On the other hand, we have for every n → N that
αn −αn−1
αn ; (n = kr ),
βn (y) = (r → N).
0; (n = kr )
This and (3.14) together imply that β(y) → c0 and hence y → c0α . Therefore, the
sequence y is in the both spaces c0α and cα but not in any one of the spaces c0 or c.
Hence, by combining this with Theorem 3.2, we deduce that the inclusions c0 ⇒ c0α
and c ⇒ cα are strict.
This concludes the proof.
Now, as a consequence of Theorem 3.6, we have the following result which gives
the necessary and sufficient condition for the matrix β to be equivalent to convergence
and boundedness both.
Corollary 3.7 The equalities c0α = c0 , cα = c and α∞ = ∞ hold if and only if
lim inf n→∞ αn+1 /αn > 1.
Proof The necessity is immediate by Theorem 3.6. For, if the equalities hold then
the inclusions in Theorem 3.6 cannot be strict and hence lim inf n→∞ αn+1 /αn = 1
which implies that lim inf n→∞ αn+1 /αn > 1.
Conversely, suppose that lim inf n→∞ αn+1 /αn > 1. Then, it follows by part (b) of
Theorem 3.5 that (αn /(αn − αn−1 ))∞ ∞
n=0 → ∞ and hence (αn−1 /(αn − αn−1 ))n=0 →
∞ .
Now, let x → cα be given. Then, we have β(x) = (βn (x))∞ n=0 → c and hence
(βn (x) − βn−1 (x))∞ n=0 → c 0 . Thus, we obtain by (3.13) that (Sn (x))∞
n=0 → c0 .
This shows that S(x) → c0 for every x → c and hence for every x → c0α . Con-
α
Proof Part (a) is immediate by Theorem 3.4. To prove (b), it is obvious by Theorem
3.2 that c ⇒ cα ⊂ ∞ , that is, the spaces cα and ∞ overlap. Furthermore, due to the
Steinhaus theorem (essentially saying that any regular matrix cannot sum all bounded
sequences), the regularity of the matrix β implies the existence of a sequence x → ∞
which is not β-summable, i.e., β(x) → c. Thus, such a sequence x is in ∞ but not
in cα . Hence, the inclusion ∞ ⇒ cα does not hold.
This concludes the proof.
Theorem 3.9 If lim inf n→∞ αn+1 /αn = 1, then the following hold:
(a) Neither of the spaces c0α and c includes the other.
(b) Neither of the spaces c0α and ∞ includes the other.
(c) Neither of the spaces cα and ∞ includes the other.
Proof For (a), it has been shown in Theorem 3.8 (a) that the inclusion c ⇒ c0α does
not hold. Further, if lim inf n→∞ αn+1 /αn = 1 then the converse inclusion is also
not held. For example, the sequence y defined by (3.15), in the proof of Theorem
3.6, belongs to the set c0α \c. Hence, part (a) follows.
To prove (b), we deduce from Theorem 3.8 that the inclusion ∞ ⇒ c0α does not
hold. Moreover, we are going to show that the converse inclusion does not hold if
lim inf n→∞ αn+1 \αn = 1. For this, suppose that lim inf n→∞ αn+1 \αn = 1. Then,
as we have seen in the proof of Theorem 3.6, there is a subsequence (αkr )r∞=0 of the
sequence α = (αn )∞ n=0 such that (3.14) holds and kr +1 − kr ≤ 2 for all r → N.
Now, let 0 < φ < 1 and define the sequence x = (xk )∞ k=0 by
φ
αk
; (k = kr ),
αk − αk−1
xk = − αk−1 − αk−2 ; (k = k + 1), (r → N)
αk − αk−1
r
⎪
0; (otherwise)
This, together with (3.14), implies that β(x) → c0 . Thus, the sequence x is in c0α
but not in ∞ . Consequently, the inclusion c0α ⇒ ∞ fails.
Finally, part (c) is immediate by combining part (b) and Theorem 3.8 (b).
This completes the proof.
Proof Obviously, the spaces p and αp overlap, since (α1 − α0 , −α0 , 0, 0, . . .) →
p ⊂ αp for 0 < p < ∞.
Now, suppose that 1/α → / p , where 0 < p < ∞ , and consider the sequence
x = e(0) = (1, 0, 0, . . .) → p . Then, we have for every n → N that
n
1 α0
βn (x) = (αk − αk−1 )ek(0) =
αn αn
k=0
which shows that β(x) → / p and hence x →/ αp . Thus, the sequence x is in p but not
α α
in p . Hence, the inclusion p ⇒ p does not hold when 1/α → / p (0 < p < ∞).
On the other hand, let 1 ∈ p < ∞ and define the sequence y = (yk ) by
1
α ; (k is even),
k
yk =
1 αk−1 − αk−2
⎪− ; (k is odd)
αk−1 αk − αk−1
and hence
|βn (y)| p = |β2n (y)| p
n n
1 α2n − α2n−1 p
= p
n α2n
α2n
1 1 α2n − α2n−2 p
∞
∈ p + p
α0 α α2n
n=1 2n−2
3.2 Some Inclusion Relations 83
1
∞ 1 α2n − α2n−2
p p
∈ p + p p
α0 α
n=1 2n−2
α2n
1
∞ 1 1
= p + p − p
α0 n=1
α2n−2 α2n
2
= p < ∞.
α0
This shows that β(y) → p and so y → αp . Thus, the sequence y is in αp but not in
p , where 1 ∈ p < ∞.
Similarly, one can construct a sequence belonging to the set αp \ p for 0 < p < 1.
Therefore, the inclusion αp ⇒ p also fails when 1/α →/ p (0 < p < ∞). Hence,
if 1/α → α
/ p then neither of the spaces p and p includes the other one, where
0 < p < ∞.
This concludes the proof.
Theorem 3.11 (a) If the inclusion p ⇒ αp holds, then 1/α → p , where 0 < p <
∞.
(b) If 1/α → 1 , then
∞
1
sup (αk − αk−1 ) < ∞.
k αn
n=k
Proof Suppose that the inclusion p ⇒ αp holds, where 0 < p < ∞, and consider
the sequence x = e(0) = (1, 0, 0, . . .) → p . Then x → αp and hence β(x) → p .
Thus, we obtain that
p
1 p
α0 = |βn (x)| p < ∞
n
α n n
(b) It is easy to prove once, since by the definition of the sequence α = (αk ), we
find that
αk − αk−1
0< < 1; (0 ∈ k ∈ n).
αn
Theorem 3.12 The inclusion p ⇒ αp holds if and only if 1/α → p , where 1 ∈
p < ∞.
p
Conversely, suppose that 1/α → p , where 1 ∈ p < ∞. Then 1/α p = (1/αk ) →
1 . Thus, it follows by Theorem 3.11 (b) that
∞
1 ∞
1
p p
sup (αk − αk−1 ) p p ∈ sup (α k − α k−1 ) p < ∞.
k α
n=k n k α
n=k n
∞
1
∩e(k) ∩α = (αk − αk−1 ) p
p
p < ∞; (k → N)
p
n=k
α n
which yields that e(k) → αp for every k → N, i.e., every basis element of the space
p is in αp . This shows that the space αp contains the Schauder basis of the space p
such that
Hence, we deduce that the inclusion p ⇒ αp holds, where 1 ∈ p < ∞, and this
concludes the proof.
Now, in the following example, we give an important special case of the space
αp , where 1 ∈ p < ∞.
Example 3.1 Consider the special case of the sequence α = (αk ) given by αk = k+1
/ 1 , while 1/α → p for 1 < p < ∞. Hence, the inclusion
for all k → N. Then, 1/α →
1 ⇒ α1 does not hold.
On the other hand, by applying the well-known inequality (see [7, p. 239])
∞
n p ∞
|xk | p
< |xn | p ; (1 < p < ∞),
n+1 p−1
n=0 k=0 n=0
holds for every x → p , where 1 < p < ∞. This shows that the inclusion p ⇒ αp
holds for 1 < p < ∞. Furthermore, this inclusion is strict; for example, the sequence
y = {(−1)k } is not in p but in αp , since
1
n p
1
|βn (y)| p = (−1)k = < ∞; (1 < p < ∞).
n n
n+1 n
(2n + 1) p
k=0
Remark 3.2 In the special case of the sequence α = (αk ) given in Example 3.1,
i.e., αk = k + 1 for all k → N, we may note that the spaces αp and α∞ reduce with
respect to the Cesàro sequence spaces X p and X ∞ of non-absolute type, which are
defined as the spaces of all sequences whose C1 -transforms are in the spaces p and
∞ , respectively, where 1 ∈ p < ∞ (see [8]).
In this section, we determine the φ-, λ-, and ε-duals of the sequence spaces c0α , cα ,
and α∞ .
Throughout, let F denote the collection of all non-empty and finite subsets of
N = {0, 1, 2, . . .}. For ready reference, we recall the following results of Chap. 2
(Theorem 2.36 and Theorem 2.39) which are fundamental for our investigation.
(ii) For p = 1,
86 3 Some New Sequence Spaces of Non-absolute Type
sup |ank | < ∞. (3.19)
k n
sup |ank |q < ∞. (3.21)
n
k
lim ank − lim ank = 0. (3.24)
n n
k
For the class ( p , c0 ) necessary and sufficient conditions can be obtained by putting
k
αj
xk = (−1)k−j yj ; (k → N). (3.25)
αk − αk−1
j=k−1
3.3 Duals of the Spaces of α–sequences 87
Now, we may begin the following result which determines the φ-dual of the spaces
c0α , cα , and α∞ .
Theorem 3.13 The φ-dual of the spaces c0α , cα , and α∞ is the set
∞
αn
a1α = a = (an ) → w: |an | < ∞ .
αn − αn−1
n=0
Proof For any fixed sequence a = (an ) → w, we define the matrix B = (bnk )∞
n,k=0
by
(−1)n−k αk
an ; (n − 1 ∈ k ∈ n),
bnk = αn − αn−1
⎪
0; (k < n − 1 or k > n)
for all n, k → N. Also, for every x → w we put y = β(x) . Then, it follows by (3.25)
that
n
αk
an x n = (−1)n−k an yk = Bn (y); (n → N). (3.26)
αn − αn−1
k=n−1
On the other hand, let n → N be given. Then, we have for any K → F that
0; (n − 1 → K and n → K ),
αn−1 |a |;
n (n − 1 → K and n → K ),
αn − αn−1
bnk =
αn
k→K
|an |; (n − 1 → K and n → K ),
α n − αn−1
⎪|a |;
n (n − 1 → K and n → K ).
φ
which shows that (c0α ) = a1α . Finally, we have by Lemma A that (c0 , 1 ) = (c, 1 ) =
φ φ
(∞ , 1 ). Thus, it can similarly be shown that (cα ) = (α∞ ) = a1α .
This completes the proof.
Remark 3.3 Let μ = (μn )∞ n=0 be defined by μn = (αn − αn−1 )/αn for all n.
Then, we have by Theorem 3.13 that (c0α )φ = (cα )φ = (α∞ )φ = 1μ , where 1μ
denotes the space of de Malafosse [9] which is defined as the set of all sequences
x = (xn ) → w such that x/μ = (xn /μn ) → 1 . On the other hand, we may note by
Theorem 3.5 (b) that if lim inf n→∞ αn+1 /αn > 1, then there is M > 1 such that
1 ∈ αn /(αn − αn−1 ) ∈ M for all n. In this special case, we obtain by Theorem 3.13
that (c0α )φ = (cα )φ = (α∞ )φ = 1 which is compatible with the fact that c0α = c0 ,
cα = c, and α∞ = ∞ by Corollary 3.7.
Now, let x, y → w be connected by the relation y = β(x). Then, by using (3.25),
we can easily derive that
n
n−1
ak αn
ak x k = χ̄ αk y k + an yn ; (n → N), (3.28)
αk − αk−1 αn − αn−1
k=0 k=0
where
ak ak ak+1
χ̄ = − ; (k → N).
αk − αk−1 αk − αk−1 αk+1 − αk
∞
ak
a2α = a = (ak ) → w: χ̄ αk < ∞ ,
αk − αk−1
k=0
αk
a3α = a = (ak ) → w: sup ak < ∞ ,
k αk − αk−1
αk
a4α = a = (ak ) → w: lim ak exists ,
k→∞ αk − αk−1
and
αk
a5α = a = (ak ) → w: lim ak = 0 .
k→∞ αk − αk−1
λ λ λ
Then, we have (c0α ) = a2α ⊂ a3α , (cα ) = a2α ⊂ a4α and (α∞ ) = a2α ⊂ a5α .
Proof This result is an immediate consequence of [10, Theorem 2].
Remark 3.4 Let us consider the special case x = y = e of the equality (3.28).
Then, it follows by Theorem 3.14 that the inclusions (c0α )λ ⇒ bs, (cα )λ ⇒ cs and
(α∞ )λ ⇒ cs hold.
3.3 Duals of the Spaces of α–sequences 89
Finally, we conclude this section with the following result concerning with the
ε-dual of the spaces c0α , cα , and α∞ .
Theorem 3.15 The ε-dual of the spaces c0α , cα , and α∞ is the set a2α ⊂ a3α .
αk q
dqα = a = (ak ) → w: ak < ∞
αk − αk−1
k
and
α αk
d∞ = a = (ak ) → w: sup ak < ∞ .
k αk − αk−1
Then (α1 )φ = d∞
α and (α )φ = d α , where 1 < p ∈ ∞.
p q
Proof Let a = (ak ) → w and 1 < p ∈ ∞. Then, by using (3.5) and (3.6), we
immediately derive for every n → N that
n
αk
an x n = (−1)n−k an yk = Bn (y), (3.29)
αn − αn−1
k=n−1
k+1
bα = αk
nk an .
n
αn − αn−1
n=k
Remark 3.5 We may note that if lim inf αk+1 /αk > 1, then there is a constant b > 1
such that 1 ∈ αk /(αk − αk−1 ) ∈ b for all k → N. This yields that dqα = q and
d∞α = , i.e., (α )φ = for 1 ∈ p ∈ ∞ which is compatible with the fact that
∞ p q
αp = p in this particular case (see [4, Corollary 4.19]).
and
αk
e0α = a = (ak ) → w: lim ak = 0 ,
k αk − αk−1
where
ak ak ak+1
χ̄ = − for all k → N.
αk − αk−1 αk − αk−1 αk+1 − αk
Then (α1 )λ = d∞
α , (α )λ = d α ⊂ eα and (α )λ = eα ⊂ eα , where 1 < p < ∞.
p ∞ q ∞ 0 1
n
n
k
αj
ak x k = (−1)k−j yj a k
αk − αk−1
k=0 k=0 j=k−1
n−1
ak αn
= χ̄ αk y k + a n yn
αk − αk−1 αn − αn−1
k=0
= Tn (y), (3.32)
Then, it is clear that the columns of the matrix T are in the space c, since
ak
α
lim tnk = χ̄ αk (3.33)
n αk − αk−1
for all k → N. Thus, we deduce from (3.32) with Lemma D that ax = (ak xk ) → cs
whenever x = (xk ) → αp if and only if T y → c whenever y = (yk ) → p . This yields
that a = (ak ) → (αp )λ if and only if T → ( p ,c), where 1 ∈ p ∈ ∞.
Let us firstly begin with the case 1 < p < ∞. Then, we derive from (3.21) that
ak q
χ̄ αk < ∞
αk − αk−1
k
and
αn
sup an < ∞. (3.34)
n αn − αn−1
92 3 Some New Sequence Spaces of Non-absolute Type
But it is obvious that the condition (3.35) is redundant, since it is obtained from
(3.34). Hence, we conclude that (α1 )λ = d∞
α .
∞
α ak
t α − lim t α = tnk − χ̄ αk
nk nk
n αk − αk−1
k k=n
αn ∞
an ak
= an − χ̄ αn + χ̄ αk .
αn − αn−1 αn − αn−1 αk − αk−1
k=n+1
αn
lim an = 0.
n αn − αn−1
Finally, we end this section with the following theorem which determines the
ε-dual of the space αp , where 1 ∈ p ∈ ∞.
Proof This can be proved similarly as the proof of Theorem 3.17 with Lemma E
instead of Lemma D.
3.4 Certain Matrix Mappings on α–sequence Spaces 93
In this final section, we state some results which characterize various matrix mappings
on the spaces c0α , cα , and α∞ and between them. Most of these results are immediate
by those of Malkowsky and Rakočević [6], and some of them are the improved
versions.
For an infinite matrix A = (ank ), we shall write for brevity that
ank ank an,k+1
ãnk = χ̄ αk = − αk (n, k → N).
αk − αk−1 αk − αk−1 αk+1 − αk
m
m−1
αm
ank xk = ãnk yk + anm ym ; (n, m → N). (3.37)
αm − αm−1
k=0 k=0
λ
In particular, let x → cα and An = (ank )∞ α
k=0 → (c ) for all n → N. Then, we obtain,
by passing to the limits in (3.37) as m → ∞ and using Theorem 3.14, that
∞
∞
ank xk = ãnk yk + lan
k=0 k=0
∞
∞
= ãnk (yk − l) + l ãnk + an
k=0 k=0
where l = limk→∞ yk and an = limk→∞ (αk ank /(αk − αk−1 )) for all n → N.
Now, let us consider the following conditions:
∞
sup |ãnk | < ∞, (3.39)
n
k=0
⎩ ∞
αk
ank → c0 for every n → N, (3.40)
αk − αk−1 k=0
94 3 Some New Sequence Spaces of Non-absolute Type
⎩ ∞
αk
ank → c for every n → N, (3.41)
αk − αk−1 k=0
⎩ ∞
αk
ank → ∞ for every n → N, (3.42)
αk − αk−1 k=0
∞
sup ank < ∞, (3.43)
n
k=0
∞
lim ank = a, (3.44)
n→∞
k=0
∞
lim ank = 0, (3.45)
n→∞
k=0
∞
∞
ank < ∞, (3.46)
n=0 k=0
∞
∞
p
ank < ∞; (1 < p < ∞), (3.47)
n=0 k=0
∞
lim |ãnk − ãk | = 0, (3.49)
n→∞
k=0
∞
lim |ãnk | = 0, (3.50)
n→∞
k=0
∞
sup ãnk < ∞, (3.52)
N →F k=0 n→N
3.4 Certain Matrix Mappings on α–sequence Spaces 95
∞
p
sup ãnk < ∞; (1 < p < ∞), (3.53)
K →F n=0 k→K
∞
|ãnk | converges for every n → N. (3.54)
k=0
Then, by combining Theorem 3.14 with the results of Stieglitz and Tietz [12], we
immediately deduce the following results by using (3.37) and (3.38).
∞
lim An (x) = ãk βk (x). (3.55)
n→∞
k=0
(b) A → (cα , c) if and only if (3.39), (3.41), (3.44), and (3.49) hold. Further, if
A → (cα , c) then we have for every x → cα that
∞
lim An (x) = ãk (βk (x) − l) + la,
n→∞
k=0
Finally, we conclude our work with the following corollaries which are immediate
by [6, Proposition 3.3].
Corollary 3.24 Let α◦ = (α◦k ) be a strictly increasing sequence of positive reals
tending to infinity, A = (ank ) an infinite matrix and define the matrix B = (bnk ) by
n
1 ◦
bnk = (αj − α◦j−1 )ajk ; (n, k → N).
α◦n
j=0
Then, the necessary and sufficient conditions for the matrix A to belong to any of
◦ ◦ ◦ ◦ ◦ ◦ ◦
the classes (α∞ , α∞ ), (cα , α∞ ), (c0α , α∞ ), (α∞ , cα ), (cα , cα ), (c0α , cα ), (α∞ , c0α ),
◦ ◦
(cα , c0α ), or (c0α , c0α ) are obtained from the respective one in Theorems 3.19, 3.20,
or 3.21 by replacing the entries of the matrix A by those of the matrix B.
Corollary 3.25 Let A = (ank ) be an infinite matrix and define the matrix B = (bnk )
by
n
bnk = ajk ; (n, k → N).
j=0
Then, the necessary and sufficient conditions for the matrix A to belong to any of the
classes (α∞ , bs), (cα , bs), (c0α , bs), (α∞ , cs), (cα , cs), (c0α , cs), (α∞ , cs0 ), (cα , cs 0 ),
or (c0α , cs 0 ) are obtained from the respective one in Theorems 3.19, 3.20, or 3.21 by
replacing the entries of the matrix A by those of the matrix B.
Corollary 3.26 Let 0 < r < 1, A = (ank ) an infinite matrix and define the matrix
B = (bnk ) by
n ⎩
n
bnk = (1 − r )n−j r j ajk ; (n, k → N).
j
j=0
Then, the necessary and sufficient conditions for the matrix A to belong to any of the
classes (α∞ , erp ), (cα , erp ), (c0α , erp ), (α∞ , ecr ), (cα , ecr ), (c0α , ecr ), (α∞ , e0r ), (cα , e0r ), or
(c0α , e0r ) are obtained from the respective ones in Theorems 3.19–3.23 by replacing
the entries of the matrix A by those of the matrix B, where 1 ∈ p ∈ ∞ and e0r ,
ecr , and erp denote the Euler sequence spaces which have been studied by Altay and
Başar [13] and by Altay et al. [14].
Remark 3.6 By following the same technique used in Corollaries 3.24, 3.25, and
3.26, we can deduce the characterization of matrix operators that map any of the
spaces c0α , cα , and α∞ into the sequence spaces defined in [8, 15, 16].
3.4 Certain Matrix Mappings on α–sequence Spaces 97
Now we characterize the matrix classes (αp , ∞ ), (αp , c), (αp , c0 ), (αp , 1 ),
(α1 , p ), and (α∞ , p ), where 1 ∈ p ∈ ∞. Further, we deduce the characteriza-
tions of some other classes by means of a given basic lemma.
The following lemmas (see [12, pp. 4–9]) will be needed in the proofs of our main
results on matrix transformations.
Lemma F Let 1 ∈ p < ∞. Then A → (1 , p ) if and only if
sup |ank | p < ∞.
k n
Now, we prove the following results characterizing the matrix mappings on the spaces
αp for 1 ∈ p ∈ ∞. Because the cases p = 1 and p = ∞ can be proved by analogy,
we shall omit the proof of these cases and only consider the case 1 < p < ∞ in the
proofs of Theorems 3.27–3.30, below. Also, these results will be proved by applying
the same technique used in [11, 16, 17].
Theorem 3.27 (i) A → (α1 , ∞ ) if and only if
⎩ ∞
αk
ank → ∞ for every n → N, (3.56)
αk − αk−1 k=0
(ii) Let 1 < p < ∞. Then A → (αp , ∞ ) if and only if (3.56) holds and
sup |ãnk |q < ∞. (3.58)
n
k
αk
lim ank = 0 for all n → N, (3.59)
k αk − αk−1
sup |ãnk | < ∞. (3.60)
n
k
Proof Suppose that conditions (3.56) and (3.58) hold and take any x = (xk ) → αp ,
⎫ α ⎬λ
where 1 < p < ∞. Then, we have by Theorem 3.17 that (ank )∞ k=0 → p for all
98 3 Some New Sequence Spaces of Non-absolute Type
n → N, and this implies the existence of the A-transform of x, i.e., Ax exists. Further,
it is clear that the associated sequence y = (yk ) is in the space p and hence y → c0 .
Let us now consider the following equality derived by using relations (3.5) and
(3.6) from the mth partial sum of the series k ank xk :
m
m−1
αm
ank xk = ãnk yk + anm ym (n, m → N), (3.61)
αm − αm−1
k=0 k=0
Further, since the matrix à = (ãnk ) is in the class ( p , ∞ ) by (3.58) and Lemma
E, we have Ãy → ∞ . Therefore, we deduce from (3.62) that Ax → ∞ and hence
A → (αp , ∞ ).
Conversely, suppose that A → (αp , ∞ ), where 1 < p < ∞. Then (ank )∞ k=0 →
α λ
( p ) for all n → N, and this, with Theorem 3.17, implies both (3.56) and
|ãnk |q < ∞ for each n → N,
k
which together imply that the relation (3.62) holds for all sequences x → αp and
y → p which are connected by the relation (3.5).
Let us now consider the continuous linear functionals f n (n → N) defined on αp
by the sequences An = (ank )∞k=0 as follows:
f n (x) = ank xk .
k
Then, since αp and p are norm isomorphic, it should follow with (3.62) that
1/q
∩ f n ∩ = ∩ Ãn ∩q = |ãnk |
q
Similarly, parts (i) and (iii) can be proved by means of Theorem 3.17 and Lemma
E, and so we leave the details to the reader.
Theorem 3.28 (i) A → (α1 , c) if and only if (3.56) and (3.57) hold and
(ii) Let 1 < p < ∞. Then A → (αp , c) if and only if (3.56), (3.58), and (3.63)
hold.
(iii) A → (α∞ , c) if and only if (3.59), (3.60), and (3.63) hold and
lim |ãnk − φk | = 0.
n
k
Proof Suppose that A satisfies the conditions (3.56), (3.58), and (3.63) and take any
x → αp , where 1 < p < ∞. Then Ax exists. Also, by using (3.63), we have for
every k → N that |ãnk |q → |φk |q as n → ∞. Thus, we deduce from (3.58) that the
inequality
k
|φj |q ∈ sup |ãnj |q = M < ∞
n
j=0 j
holds for every k → N which yields that (φk ) → q . Further, since x → αp , we have
y → p . Consequently, we obtain by applying the Hölder inequality that (φk yk ) → 1 .
Now, for any given β > 0, choose a fixed k0 → N such that
⎡ ⎤1/ p
∞
⎣ p⎦ β
|yk | < .
4M 1/q
k=k0 +1
⎡ ⎤1/q ⎡ ⎤1/ p
∞ ∞
β ⎣
< + (|ãnk | + |φk |)q ⎦ ⎣ |yk | p ⎦
2
k=k0 +1 k=k0 +1
⎡⎛ ⎧1/q ⎛ ⎧1/q ⎤
∞ ∞
β β ⎢⎝ ⎥
< + ⎣ |ãnk |q ⎨ +⎝ |φk |q ⎨ ⎦
2 4M 1/q
k=k0 +1 k=k0 +1
β β
< + 2M 1/q = β
2 4M 1/q
for all sufficiently large n ≤ n0 . This leads us to the consequence that An (x) →
α
k φk yk as n → ∞ which means that Ax → c and hence A → ( p , c).
Conversely, suppose that A → (αp , c), where 1 < p < ∞. Then A → (αp , ∞ ).
This leads us with Theorem 3.27 to the necessity of the conditions (3.56) and (3.58)
which together imply that (3.62) holds for all sequences x → αp and y → p which
are connected by the relation y = β(x).
Now, let y → p be given and let x be the sequence defined by (3.6). Then
y = β(x) and hence x → αp . Further, since Ax → c by the hypothesis, we obtain
by (3.62) that Ãy → c which shows that à → ( p , c), where à = (ãnk ). Hence, the
necessity of (3.63) is immediate by (3.20) of Lemma D. This concludes the proof of
part (ii).
Since parts (i) and (iii) can be proved similarly, we omit their proofs.
Theorem 3.29 (i) A → (α1 , c0 ) if and only if (3.56) and (3.57) hold and
(ii) Let 1 < p < ∞. Then A → (αp , c0 ) if and only if (3.56), (3.58), and (3.64)
hold.
(iii) A → (α∞ , c0 ) if and only if (3.59) holds and
lim |ãnk | = 0. (3.65)
n
k
Proof This theorem can be proved by the same technique used in the proof of The-
orem 3.28 with Lemma D, and by using the fact that (3.65) implies both (3.60) and
(3.64). Thus, we leave the proof to the reader.
(ii) Let 1 < p < ∞. Then A → (αp , 1 ) if and only if (3.56) holds and
3.4 Certain Matrix Mappings on α–sequence Spaces 101
q
sup ãnk < ∞. (3.66)
F→F
k n→F
Proof Suppose that conditions (3.56) and (3.66) hold and take any x → αp , where
1 < p < ∞. Then y → p . Also, it is obvious by (3.66) that (3.58) holds. Therefore,
we have by Theorem 3.17 that (ank )∞ α λ
k=0 → ( p ) for all n → N and hence Ax exists.
Further, it follows by combining (3.66) and Lemma C that the matrix à = (ãnk ) is
in the class ( p , 1 ) and hence Ãy → 1 . Moreover, we deduce by (3.56) and (3.58)
that the relation (3.62) holds which yields that Ax → 1 and hence A → (αp , 1 ).
Conversely, suppose that A → (αp , 1 ), where 1 < p < ∞. Then A → (αp , ∞ ).
Thus, Theorem 3.27 implies both (3.58) and the necessity of (3.56), which together
imply that (3.62) holds for all x → αp and y → p such that y = β(x). Therefore, the
necessity of (3.66) can be deduced similarly as the necessity of (3.63) in the proof
of Theorem 3.28 with Lemma C instead of Lemma D.
This completes the proof of part (ii).
Similarly, one can prove the other two parts by means of Theorems 3.17, 3.27,
and Lemma C.
Theorem 3.31 Let 1 ∈ p < ∞. Then A → (α1 , p ) if and only if (3.56) holds and
sup |ãnk | p < ∞. (3.67)
k n
Proof Suppose that A satisfies the conditions (3.56) and (3.67), and take any x → α1 .
Then y → 1 . Further, we have by Theorem 3.17 that (ank )∞ α λ
k=0 → (1 ) for all n → N
and hence Ax exists. Moreover, we obtain by (3.67) that
1/ p
sup |ãnk | ∈ sup |ãnk | p < ∞ for each n → N.
k k n
Therefore, the series k ãnk yk converges absolutely for each fixed n → N. Thus,
if we pass to the limits in (3.61) as m → ∞, then it follows by (3.56) that (3.62)
holds. Hence, by applying the Minkowski’s inequality and using (3.62) and (3.67),
we derive that
1/ p ⎛ p ⎧1/ p ⎡ 1/ p ⎤
|An (x)| p = ⎝ ãnk yk ⎨ ∈ ⎣|yk | |ãnk | p ⎦< ∞
n n k k n
102 3 Some New Sequence Spaces of Non-absolute Type
Exercises
1. Show that the space αp for p = 2 is not an inner product space and hence not a
Hilbert space for 1 ∈ p < ∞.
2. Prove that the inclusion αp ⇒ αs strictly holds for 0 < p < s < ∞.
3. Prove that the inclusion αp ⇒ p holds if and only if S(x) → p for every sequence
x → αp , where 0 < p ∈ ∞.
4. Show that the inclusion 1 ⇒ α1 holds if and only if 1/α → 1 .
5. Show that the equality αp = p holds if and only if lim inf n→∞ αn+1 /αn > 1,
where 1 ∈ p < ∞.
6. The sequence space αp of non-absolute type is separable for 1 ∈ p < ∞.
7. Let A = (ank ) be an infinite matrix and define the matrix C = (cnk ) by
n ◦ ◦
cnk = α1◦ j=0 (αj − αj−1 )ajk for all n, k → N. Obtain the necessary and suffi-
n
◦ ◦
cient conditions in order that A belongs to any of the classes (αp , α∞ ), (αp , cα ),
◦ ◦ ◦ ◦
(αp , c0α ), (αp , α1 ), (α1 , αp ), and (α∞ , αp ).
8. Let 1 < p < ∞. Then, prove that A → (α∞ , p ) if and only if (3.59) holds and
|ãnk | converges for every n → N,
k
p
sup ãnk < ∞.
K →F n
k→K
References
1. Mursaleen, M., Noman, A.K.: On the spaces of α-convergent and bounded sequences. Thai J.
Math. 8(2), 311–329 (2010)
2. Mursaleen, M., Noman, A.K.: On some new difference sequencespaces of non-absolute type.
Math. Comput. Model. 52, 603–617 (2010)
3. Mursaleen, M., Noman, A.K.: On some new sequence spaces of non-absolute type related to
the spaces p and ∞ I . Filomat 25(2), 33–51 (2011)
References 103
4. Mursaleen, M., Noman, A.K.: On some new sequence spaces of non-absolute type related to
the spaces p and ∞ II. Math. Commun. 16(2), 383–398 (2011)
5. Maddox, I.J.: Elements of Functional Analysis, 2nd edn. The University Press, Cambridge
(1988)
6. Malkowsky, E., Rakočević, V.: Measure of noncompactness of linear operators between spaces
of sequences that are ( N̄ , q) summable or bounded. Czech. Math. J. 51(3), 505–522 (2001)
7. Hardy, G.H., Littlewood, J.E., Polya, G.: Inequalities. Cambridge University Press, Cambridge
(1952)
8. Ng, P.N., Lee, P.Y.: Cesàro sequence spaces of non-absolute type. Comment. Math. Prace Mat.
20(2), 429–433 (1978)
9. de Malafosse, B.: The Banach algebra B(X ), where X is a B K space and applications. Mat.
Vesnik 57, 41–60 (2005)
10. Malkowsky, E., Savaş, E.: Matrix transformations between sequence spaces of generalized
weighted means. Appl. Math. Comput. 147(2), 333–345 (2004)
11. Başar, F., Altay, B.: On the space of sequences of p-bounded variation and related matrix
mappings. Ukrainian Math. J. 55(1), 136–147 (2003)
12. Stieglitz, M., Tietz, H.: Matrixtransformationen von folgenr äumen eine ergebnisübersicht.
Math. Z. 154, 1–16 (1977)
13. Altay, B., Başar, F.: Some Euler sequence spaces of non-absolute type. Ukrainian Math. J.
57(1), 1–17 (2005)
14. Altay, B., Başar, F., Mursaleen, M.: On the Euler sequence spaces which include the spaces p
and ∞ . Inf. Sci. 176(10), 1450–1462 (2006)
15. Aydın, C., Başar, F.: On the new sequence spaces which include the spaces c0 and c. Hokkaido
Math. J. 33(2), 383–398 (2004)
16. Aydın, C., Başar, F.: Some new sequence spaces which include the spaces p and ∞ . Demon-
stratio Math. 38(3), 641–656 (2005)
17. Mursaleen, M., Başar, F., Altay, B.: On the Euler sequence spaces which include the spaces p
and ∞ II. Nonlinear Anal 65(3), 707–717 (2006)
Chapter 4
Some Non-classical Sequence Spaces
We determine the β duals of Maddox sequence spaces which will be further applied
to characterize the matrix transformations of these spaces.
Theorem 4.1 ([5, Theorem 1]) Let 1 < pk for all k and
qk −qk / pk
M( p) := a = (ak ): |ak | N <∞
N >1 k
we get
i.e., ak xk is absolutely convergent and so convergent. Thus M(P) → β ( p).
k
Conversely, let a ∈ β ( p). Then k ak x k converges for all x ∈ ( p). This
implies that a ∈ M( p), and for otherwise, we can determine integers 0 = n(0) <
n(1) < n(2) < · · · such that
Ms = |ak |qk (s + 1)−qk / pk , s = 1, 2, . . .
I (s)
where I (s) := {k: n(s − 1) + 1 ≤ k ≤ n(s)}. Now for k ∈ I (s), define x = {xk } by
Then we have
ak xk = (s + 1)−1 ,
I (s)
and
− pk
|xk | pk = |ak |qk (s + 1)− pk −qk Ms
I (s) I (s)
≤ Ms−1 (s + 1)−1 |ak |qk (s + 1)−qk
I (s)
= Ms−1 (s + 1) −2
|ak |qk (s + 1)−qk / pk
I (s)
−2
= (s + 1) .
Thus x ∈ ( p), but k ak xk diverges. Hence we must have a ∈ M( p), i.e., β ( p) →
M( p). Therefore we have β ( p) = M( p).
This completes the proof.
−1/ pk
M0 ( p) := a = (ak ): |ak |N <∞ .
N >1 k
β
Then c0 ( p) = M0 ( p).
β
(b) For every p = ( pk ) we have cβ ( p) = c0 ( p) ≤ cs.
4.1 Sequence Spaces of Maddox 107
Proof (a) Let a ∈ M0 ( p) and x ∈ c0 ( p). Then k |ak |N −1/ pk < ∞ for some N > 1
and so |xk | pk < N −1 for all sufficiently large k. Thus for such k, |ak xk | ≤ |ak |N −1/ pk ,
β
and so k ak xk converges. Hence M0 ( p) → c0 ( p).
β
Conversely, let a ∈ c0 ( p). Then k ak xk converges for all x ∈ c0 ( p), which
implies that a ∈ M0 ( p). For otherwise, we can easily construct a sequence x ∈ c0 ( p)
β
such that k ak xk diverges. Hence c0 ( p) → M0 ( p).
β
Finally we have c0 ( p) = M0 ( p).
β
(b) Let a ∈ c0 ( p) ≤ cs and | xk − l | pk ∩ 0. Then l k ak and k ak (xk − l)
β
are well defined, and therefore, k ak xk converges. Hence c0 ( p) ≤ cs → cβ ( p).
On the other hand, let a ∈ cβ ( p). Then e = (1, 1, . . .) ∈ c( p) implies a ∈ cs.
β β
Also c0 ( p) → c( p) implies c0 ( p) ⊃ cβ ( p). Hence a ∈ c0 ( p) ≤ cs. we have
β
cβ ( p) = c0 ( p) ≤ cs.
This completes the proof.
β
Theorem 4.3 ([3, Theorem 2]) Let pk > 0 for every k. Then ∞ ( p) = M∞ ( p),
where
M∞ ( p) := a = (ak ): |ak |N 1/ pk < ∞ .
N >1 k
Proof Let a ∈ M∞ ( p) and x ∈ ∞ ( p). Choose an integer N > max{1, supk |xk | pk }.
Then |xk | ≤ N 1/ pk , and so
ak xk ≤ |ak ||xk | ≤ |ak |N 1/ pk < ∞.
k k k
β β
Therefore a ∈ ∞ ( p) and that is M∞ ( p) → ∞ ( p).
β
Conversely, let a ∈ ∞ ( p) \ M∞ ( p). Then there is an integer N > 1 such that
k |ak |N
1/ pk = ∞. Define x = {x } by x = N 1/ pk sgn a . Then x ∈ ( p), but
k k k ∞
ak x k = |ak |N 1/ pk = ∞.
k k
β
Hence a contradiction, and we must have a ∈ M∞ ( p), i.e., ∞ ( p) → M∞ ( p).
β
Therefore ∞ ( p) = M∞ ( p).
This completes the proof.
Theorem 4.4 ([3, Theorem 4]) Let 0 < pk ≤ 1 for every k. Then w β ( p) = M,
where
∞
r −1 1/ pk
M := a = (ak ): max{(2 N ) |ak |} < ∞ for some integer N > 1 .
r
r =0
∞ Let a r ∈ −1M.
Proof Then there exists an integer N > 1 such that
r =0 maxr {(2 N ) 1/ pk |a |} < ∞. Take x ∩ l(w( p)). Then there exists an inte-
k
ger R > 0 such that
2−r |xk − l| pk < 1/2N
r
and
2−r max{1, | l |} < 1/2N ,
for every r > R. Hence 2−r |xk | pk ≤ 1/N , and since pk ≤ 1, we have for every
r > R,
| ak xk |≤ max (2r N −1 )1/ pk |ak | N g(x),
r
r
where g(x) = supr {2−r r |xk | pk }. Hence | ak xk |< ∞, which implies that
β β
a ∈ w ( p). Therefore M ⊆ w ( p).
To prove the reverse inclusion, suppose that a ∈ w β ( p) \ M. Then
∞
max (2r N −1 )1/ pk |ak | = ∞
r
r =0
for every integer N > 1, and therefore, there exists a sequence Nr ∩ ∞ such that
∞
max (2r Nr−1 )1/ pk |ak | = ∞.
r
r =0
Write A(r, k) = 2r/ pk |ak | for 2r ≤ k < 2r +1 , and suppose kr is such that
maxr A(r, k) = A(r, kr ). Put
2r/ p(kr ) |ak | sgn a(kr )Nr−1 ,
x(kr ) =
0, k ≥ = kr .
Then 2−r r |xk | pk = (Nr−1 ) p(kr ) ≤ (Nr−1 ) M for
sufficiently large r since (Nr−1 )
β −1
sequence. Hence xk ∩ 0(w ( p)), but r ak xk = Nr maxr A(r, k), so
is a null
that k ak xk diverges.
This completes the proof.
Theorem 4.5 ([5, Theorem 2]) Let 1 < pk ≤ sup pk < ∞, for all k. Then ⇒ ( p) is
isomorphic to M( p).
p) as ek , k = 1, 2, .⇒. .. Then x = k xk ek , for
Proof Take the unit vectors in (
every x ∈ ( p). Hence f (x) = k ak xk for any f ∈ ( p), where f (ek ) = ak . By
Theorem 4.1, the convergence of k ak xk for every x ∈ ( p) implies that a ∈ M( p).
4.1 Sequence Spaces of Maddox 109
Now fix x ∈ ( p) and take any a ∈ M( p), then by Theorem 4.1, k ak xk
converges and defines a linear functional on ( p). It is easy to check that whenever
g(x) ≤ 1
ak x k ≤ |ak |qk N −qk / pk + N g(x),
k k
where g(x) = ( |xk |qk )1/H is the paranorm on ( p). Hence k ak xk defines an
element of ⇒ ( p). It is now clear that the map T : ⇒ ( p) ∩ M( p), given by T ( f ) = a,
is a linear bijection. Therefore ⇒ ( p) is isomorphic to M( p).
This completes the proof.
Theorem 4.6 ([5, pp. 432]) M( p) = (q) if and only if p = inf pk > 1.
Proof Let p > 1. Then we have for every N > 1, 1 < N qk / pk ≤ N 1/( p−1) , and
hence, M( p) = (q).
Conversely, if M( p) = (q), but inf pk = 1, then there exists k1 < k2 < · · ·
such that pkn < 1 + 1/n. Define x = (xk ) by
1 if k = kn ,
xk =
0 otherwise.
Then
|xk |qk /2qk / pk < 1,
k
Theorem 4.7 ([5, Theorem 3]) If 1 < pk ≤ sup pk < ∞, then M( p) and ⇒ ( p) are
linearly homeomorphic.
Theorem 4.8 ([5, Theorem 6]) If sup pk < ∞, then c0⇒ ( p) is isomorphic to M0 ( p).
Proof sup pk < ∞ implies that c0 ( p) is a topological linear space and each f ∈
c0⇒ ( p) has the representation
110 4 Some Non-classical Sequence Spaces
f (x) = ak xk , x ∈ c0 ( p).
k
By Theorem 4.2, it follows that a ∈ M0 ( p) and k ak xk is linear on c0 ( p), whenever
a∈ M0 ( p). Now g(x) < N −1/M ∞, where N −1/ > 1, implies |xk | ≤ g(x), and so
n
k |a k x k | ≤ g(x) k=1 |a k | + k=n+1 |a k |N pk .
For a ∈ M0 ( p) we can choose n so large and g(x) so small that |ak xk | is
as small as we please. It follows that k ak xk is continuous on c0 ( p), whenever
a ∈ M0 ( p). Hence the map f ∩ a is linear and bijective.
This completes the proof.
Theorem 4.9 ([5, Theorem 6]) If 0 < inf pk ≤ sup pk < ∞, then c0⇒ ( p) is isomet-
rically isomorphic to 1 .
Proof Let p = inf pk > 0 and the norm on 1 be the usual norm. If g(x) ≤ A and
k is such that |xk | < 1, then |xk | ≤ A, and if |xk | ⊂ 1, then |xk | p ≤ |xk | pk . Hence
and so that
|ak xk | ≤ B |ak |.
k
Hence
|ak |A1/ pk ≤ 1,
k
and k |ak | ≤ A−1/M , since A > 1. Consequently, the map f ∩ a is continuous.
Now we may observe that c0⇒ ( p) is normable with
◦ f ◦ = sup{| f (x)|: g(x) ≤ 1} = |ak |.
k
Maddox studied the sequence spaces ∞ ( p), c0 ( p), and c( p) within the framework
of paranormed sequence spaces. In each of the spaces, he considered the function
g(x) = sup |xk | pk /M with M = max(1, supk pk ) and introduced a topology τg via
the corresponding metric d(x, y) = g(x − y) with varying success.
In ∞ ( p), g is a paranorm (and τg a linear topology) only in the trivial case
inf pk > 0, when ∞ ( p) = ∞ . In fact, the natural topology of ∞ ( p) is not
metrizable and hence not paranormable unless ∞ ( p) = ∞ .
In c0 ( p), g is a paranorm and τg is an FK topology, so that by the uniqueness of
FK topologies, τg coincides with the projective limit for c0 ( p).
In c( p), again g is a paranorm (and τg a linear topology) only if inf k pk > 0,
when c( p) = c. The natural topology of c( p) can be induced by a paranorm. A
convenient one is g̃(x) = supk |xk − | pk /M + || where is the unique number with
x − e ∈ c0 ( p).
Recently, Grosse-Erdmann [6, 7] gave another approach to study these sequence
spaces and derived the fundamental properties.
(n)
Definition 4.1 Let X 0 = c0 and X p = p for 1 ≤ p ≤ ∞. Let A = (ak )n,k be
(n+1) (n)
a Köthe matrix, i.e., a matrix with ak ⊂ ak > 0 for n, k ∈ N. The associated
(n) (n) (n)
matrix V = (vk )n,k is given by vk = 1/ak for n, k ∈ N. Then, the spaces
(n)
λ p (A) := x ∈ w: (xk ak )k ∈ X p for all n ∈ N
∞
(n)
:= x ∈ w: (xk ak )k ∈ X p
n=1
and
K p (V ) := x ∈ w: (xk /ak(n) )k ∈ X p for some n ∈ N
∞
(n)
:= x ∈ w: (xk /ak )k ∈ X p
n=1
are called echelon and coechelon spaces, respectively, of order p. Echelon spaces
of order 1 and coechelon spaces of order ∞ are known as the echelon and coechelon
spaces of Köthe, respectively. λ p (A) is endowed with the projective limit topology
that is induced by the seminorms qn (x) = ◦(xk ak(n) ◦ p for n ∈ N with ◦.◦ p being
the norm in X p . K p (V ) is endowed with the obvious inductive limit topology. Thus,
λ p (A) becomes an FK space and K p (V ) an LBK space.
Definition 4.2 Besides the spaces ∞ ( p) c0 ( p), M∞ ( p), and M0 ( p), we also use
the following sequence spaces
112 4 Some Non-classical Sequence Spaces
∞
∞ ( p) := x ∈ w: sup |xk |n 1/ pk < ∞
n=1 k
:= x ∈ w: lim |δk xk | pk
= 0 for each δ ∈ c0
k
∞
−1/ pk
c0 ( p) := x ∈ w: lim |xk |n =0
k
n=1
:= x ∈ w: sup |xk /δk | pk < ∞
k
for some δ ∈ c0 with δk ≥ = 0 for all k ,
and
c( p) := c0 ( p) + {e}
⎪
:= x ∈ w: supk |(xk − )/δk | pk < ∞
for some ∈ C and some δ ∈ c0 with δk ≥ = 0 for all k} .
Remark 4.1 Let A( p) denote the Köthe matrix (n 1/ pk ) )n,k and V ( p) the associ-
ated matrix (n −1/ pk ) )n,k . Then, we have c0 ( p) = λ0 (A( p)), M∞ ( p) = λ1 (A( p)),
∞ ( p) = λ∞ (A( p)), c0 ( p) = K 0 (V ( p)), M0 ( p) = K 1 (V ( p)), ∞ ( p) =
K ∞ (V ( p)), c( p) = λ0 (A( p)) ⊃ {e}, and c( p) = K 0 (A( p)) ⊃ {e} (if p ∈
/ c0 ).
Theorem 4.10 [7] (i) Each of the spaces c0 ( p), c( p), M∞ ( p), and ∞ ( p) is
normable if and only if inf pk > 0.
(ii) Each of the spaces c0 ( p), c( p), M∞ ( p), and ∞ ( p) is metrizable if and only
if inf pk > 0.
is normable with norm ◦.◦. Then there are n ∈ N and M > 0 with ◦x◦ ≤
M supk |xk |n 1/ pk for x ∈ c0 ( p). By continuity of the inclusion map c0 ( p) ∩ E n+1
there is M̄ > 0 with supk |xk |(n + 1)1/ pk ≤ M̄◦x◦ ≤ M M̄ supk |xk |n 1/ pk for
x ∈ c0 ( p). Taking x = ek (k ∈ N) shows that ((n + 1)/n)1/ pk is bounded in
k; hence, that inf k pk > 0. Conversely, inf k pk > 0 implies that c0 ( p) = c0 is
normable. Similar arguments work for M∞ ( p) and ∞ ( p); for c( p) apply the results
of c0 ( p).
(ii) Assume that c0 ( p) = ⊕∞ n=1 E n , E n := {x ∈ w: lim |x k |n
−1/ pk = 0}, is
k
metrizable and hence an FK space. Then there is some n ∈ N with c0 ( p) = E n .
This forces inf k pk > 0. Conversely, if inf k > 0, then c0 ( p) = c0 is metrizable. A
similar argument applies for the other three spaces.
This completes the proof.
4.2 Echelon and Coechelon Spaces 113
∞ ( p)† = M∞ ( p),
c( p)γ = M0 ( p) ≤ bs;
c( p)αα = ∞ , c( p)ββ = c( p)γγ = ∞ ( p) + bv;
c( p)⇒ M0 ( p) × C, c( p)⇒⇒ ∞ ( p) × C.
(iv) M0 ( p), M∞ ( p), and ∞ ( p) are †-perfect for any p. Each of the spaces c0 ( p),
c0 ( p), and c( p) is †-perfect iff pk ∩ 0. c( p) is †-perfect for no p.
114 4 Some Non-classical Sequence Spaces
(v) Each of the spaces c0 ( p), c( p), M∞ ( p) , ∞ ( p), c0 ( p), M0 ( p), and ∞ ( p)
is reflexive iff pk ∩ 0.
In this section, we shall determine the matrices of the classes (( p), ∞ ), (( p), c),
(∞ ( p), ∞ ), (∞ ( p), c), (c( p), c), and (w( p), c).
As in [3], we shall use the notation C(n, N , m, s) = sk=m | ank |qk N −qk for
all integers n, m ⊂ 1, 1 ≤ s ≤ ∞ and N > 1. We shall simply write C(n, N ) for
C(n, N , 1, ∞) and put C(N ) = supn C(n, N ) for every integer N > 1. It is observed
that for each n, C(n, N ) decreases as N increases. The following inequality will be
used in proving the results: For any B and any two complex numbers x and y,
| x y |= B(| x |q B −q + | y | p ), (4.1)
Theorem 4.12 ([3, Theorem 1]) (i) Let 1 < pk ≤ H = supk pk < ∞ for every
k. Then A ∈ (( p), ∞ ) if and only if there exists an integer B > 1 such that
C(B) < ∞.
(ii) Let 0 < pk < 1 for every k. Then A ∈ (( p), ∞ ) if and only if
supn,k | ank | pk < ∞.
Proof (i) By using the inequality (4.1) we see that for x ∈ ( p) and for every n,
for every n and every x ∈ S[θ, δ], where S[θ, δ] denotes the closed sphere in ( p)
with center at the origin θ = (0, 0, 0, . . .) and radius δ.
Now choose an integer Q > 1 such that
Qδ H > G.
4.3 Matrix Transformations of Maddox Sequence Spaces 115
By our supposition we have C(Q) = ∞ and so two cases are possible: Either
C(n, Q) < ∞ for every n ⊂ 1, or there exists n ⊂ 1 such that C(n, Q) = ∞. In the
first case there exists n ⊂ 1 such that C(n, Q) > 2 and there exists k0 > 1 such that
C(n, Q, k0 + 1, ∞) < 1,
and hence, C(n, Q, 1, k0 ) > 1. In the second case we may choose k0 > 1 such that
C(n, Q, 1, k0 ) > 1, so that in either case there exist n ⊂ 1 and k0 > 1 such that
where S is given by (4.3). Then it is easy to see that g(x) ≤ δ and that
| An (x) |= S −1 Q | ank |qk Q −qk δ H/ pk Qδ H > G
k
Theorem 4.13 ([3, Corollary, pp. 101]) Let p = ( pk ) be as in Theorem 4.12. Then
A ∈ (( p), c) if and only if, together with the conditions of the theorem, we have
Proof We consider the case (i) as in Theorem 4.12, and case (ii) can be followed
similarly. The necessity of (4.4) is trivial so that we prove only the sufficiency. For
every integer m > 1 and every n we have
Hence
lim C(n, B, 1, m) ≤ C(B),
m,n∩∞
i.e.,
| αk |qk B −qk ≤ C(B).
k
116 4 Some Non-classical Sequence Spaces
β ∞ β
(ak ) ∈ ( p). Therefore, since An = (ank )k=0 ∈ ( p), the series
Hence k αk xk
and k ank xk converge for every n and every x ∈ ( p).
For each x ∈ ( p) we can choose an integer m > 1 such that
∞
| xk | pk < 1.
k=m+1
Then by the proof of Theorem 4.5 and the inequality (4.1) we have
∞
∞
⎛1/H
| ank − αk || xk |≤ 2B(2C + 1) | xk | pk
.
k=m+1 k=m+1
∞
Hence limn∩∞ ∞ k=0 ank x k = k=0 αk x k .
This completes the proof.
Corollary 4.14 If pk = p (constant) >1 for every k in Theorem 4.12, then we get
the well-known condition for A ∈ ( p , ∞ ) given by Hahn, i.e., A ∈ ( p , ∞ ) if and
only if
sup | ank |q < ∞
n
k
where p −1 + q −1 = 1.
Remark 4.2 (i) Note that (( p), ∞ ) ⊆ (1 , ∞ ) if and only if inf pk > 0.
(ii) The classes (( p), ∞ ) and (1 , ∞ ) are identical for 0 < inf pk < pk < 1.
Theorem 4.15 ([3, Theorem 3]) If pk > 0 for every k, then A ∈ (∞ ( p), ∞ ) if
and only if for every integer N > 1
D(N ) = sup | ank | N 1/ pk < ∞. (4.5)
n
k
Theorem 4.16 ([3, Corollary, pp. 102]) If pk > 0 for every k, then A ∈ (∞ ( p), c)
if and only if (4.4) holds and for all integers N > 1
| ank | N 1/ pk converges uniformly in n. (4.6)
k
Proof Since by (4.6) k ank xk converges uniformly in n for each x ∈ ∞ ( p), we
have
lim n ank xk = αk xk .
k k
Theorem 4.17 ([8, Theorem 9, Corollary 1–2]) Let p ∈ ∞ . Then (a) A ∈ (c( p), c)
if and only if
(i) there exists an absolute constant B > 1 such that
C = sup | ank | B −1/ pk < ∞, (4.7)
n
k
Proof (a) Let the conditions (i)–(iii) hold and | xk − l | pk ∩ 0. It is easy to check
that (αk ) ∈ cβ ( p). Given ε > 0, there exists k0 = k0 (ε, x) such that
every k >−1/
for k0 . Putting bnk = ank − αk we have bnk ∩ 0 (n ∩ ∞, k fixed) and
k | bnk | B pk ≤ 2C. Therefore
| (ank − αk )(xk − l) |≤| bnk (xk − l) | +ε
k k≤k0
and
lim ank xk = lα + αk (xk − l),
n
k k
M
where Q = R k≤k0 | xk | . For the term S2 we have
S2rn
= ank xk ≤ | ank | xk | B −rk ≤ T rn ,
k>k0 k>k0
Necessity. Let A ∈ (c0 ( p), ∞ ( p)). Then N (x) = supn | An (x) | pn < ∞ for
every x ∈ c0 ( p). Put f n (x) =| An (x) | pn . Then for every n, f n is a continuous
function on c0 ( p). Since c0 ( p) is a complete metric space and f n (x) ≤ N (x) for
every n, we have, by the uniform boundedness principle, that there exists a sphere
S[θ, δ] → c0 ( p) with δ < 1, θ = (0, 0, . . .) and an absolute constant K such that
pn
T = ank xk ≤ K (4.9)
k
for every n and for every x ∈ S[θ, δ]. For every integer m > 0 we define a sequence
(x (m) ) of elements of c0 ( p) as follows:
(m) δ Mrk sgnank (1 ≤ k ≤ m),
xk =
0 (k > m).
4.4 Applications
Here, we define the matrix domains of Maddox sequence spaces and apply the above
results to determine their duals.
Let m be a positive integer throughout. The operators β(m) , (m) : w ∩ w are
defined by
⎨
(1)
k
(β(1) x)k = xk − xk−1 , ⎩ x = x j (k = 0, 1 . . .),
k j=0
(m)
(1)
(m−1)
β(m) = β(1) ◦ β(m−1) , = ◦ (m ⊂ 2).
Throughout, we shall use the convention that a term with a negative subscript is
equal to zero.
120 4 Some Non-classical Sequence Spaces
(m)
In this section, we give the α−, β− and γ− duals of the sets βu X for Maddox
sequence spaces X = c0 ( p), c( p), and ∞ ( p) and characterize some matrix trans-
formations between them. Some of these results generalize those given in special
cases by Kizmaz [9] for m = 1 and p = u = e, by Malkowsky and Parashar [10]
for m ⊂ 1 and p = u = e, by Ahmad and Mursaleen [11] and by Malkowsky [12]
for m = 1, p = ( pk )∞ k=0 , and u = e, by Malkowsky, Mursaleen, and Qamaruddin
[13] for m = 1, p = ( pk )∞ k=0 , and u ∈ U and by Başarir and Et [14] for m ⊂ 1,
p = ( p k )∞
k=0 , u = e, and X = ∞ ( p). We shall use the following notations:
m ⎣ ⎤
k ⎣ ⎤
(m)
(β x)k = (−1) j mj xk− j = (−1)k− j m
k− j xj, (4.10)
j=0 j=max{0,k−m}
⎨
(m) m ⎣
⎤
⎩ x =
m+k− j−1
x j (k = 0, 1, 2, . . .), (4.11)
k− j
k j=0
(m)
(m)
(m) (m)
◦β =β ◦ = id, the identity on w, (4.12)
k ⎣
⎤ ⎣ ⎤
m+ j−1 m+k
j = k (k = 0, 1, 2, . . .). (4.13)
j=0
⎨ −1
(m)
M α,(m) ( p; B/ | u |) = ⎩ B(1/ p)/ | u | ⇒ 1
⎫
∞
k ⎣
⎤ B 1/ p j ⎬
m+k− j−1
:= a ∈ w: | ak | < ∞ ,
k− j
| uj | ⎡
k=0 j=0
α,(m)
M0 ( p; u) = ⊕ N ∈N\{1} M α,(m) ( p; N −1 / | u |),
α,(m)
M∞ ( p; u) = ≤ N ∈N\{1} M α,(m) ( p; N / | u |),
and
⎨ −1
(m)
M (m) ( p; u) = ⎩ 1/u ⇒ 1
⎫
∞
k ⎣
⎤ ⎬
m+k− j−1 1
:= a ∈ w: | ak || |< ∞ .
k− j
uj ⎡
k=0 j=0
Then we have
(a) (β(m) α
u c0 ( p)) = M0
α,(m)
( p; u),
(m) α
(b) (βu c( p)) = M α,(m) ( p; u) = M (m) ( p; u) ≤ M0α,(m) ( p; u),
(m) α,(m)
(c) (βu ∞ ( p))α = M∞ ( p; u).
(m) (m)
Let x ∈ βu c0 ( p). Then we may assume supk |(βu x)k | pk ≤ 1/N for all k.
(m)
For otherwise there is an integer k0 such that supk⊂k0 |(βu x)k | pk ≤ 1/N , and
(m)
putting K = max0≤k≤k0 |(βu x)k | pk , p̃ = min0≤k≤k0 pk , L = (K + 1)N , and
(m)
y = L −1/ p̃ .x, we obtain |(uβ(m) y) j | p j ≤ 1/N and |(βu y) j | p j ≤ N −1/ p j for all
(m) (m)
j. Writing z = β(m) y, we have y = z= (uz)/u and
(m)
(m)
|ak yk | ≤ |ak | | uz | / | u |≤ |ak | N −1/ p / | u |= |ak |sk(m) (N −1 /u), (4.15)
(m)
and consequently, ay ∈ 1 by (4.14). This implies ax ∈ 1 . Since x ∈ βu c0 ( p) is
(m)
arbitrary, we have a ∈ (βu c0 ( p))α .
Conversely, we assume a ∈ / M0α,(m) ( p; u). Then we can determine a strictly
∞
increasing sequence (k(s))s=0 of integers with k(0) = 0 such that
122 4 Some Non-classical Sequence Spaces
k(s+1)−1
(m)
|ak |sk ((s + 1)−1 / | u |) > 1 (s = 0, 1, 2, . . .).
k=k(s)
s−1 ⎣m+k− j−1 ⎤
k(l+1)−1
1 k ⎣
m+k− j−1
⎤ 1
= k− j + k− j
l=0 j=k(l)
| u j | (l + 1)1/ p j
j=k(s)
| u j | (s + 1)1/ p j
hence, x = (m) y, that is, β(m) x = y by (4.12), and so x ∈ β(m)
u c0 ( p). On the
other hand, for each s we have
k(s+1)−1
k(s+1)−1 k ⎣
⎤
m+k− j−1 1
|ak xk | ⊂ |ak | k− j
k=k(s) k=k(s) j=0
| u j | (s + 1)1/ p j
k(s+1)−1
(m)
= |ak |sk ((s + 1)−1 /u) > 1,
k=k(s)
hence, a ∈/ (β(m) α
u c0 ( p)) .
(b) First we assume that a ∈ M α,(m) ( p; u). Then condition (4.14) holds and
⎨ −1
(m)
a∈⎩ 1/u ⇒ 1 . (4.16)
(m)
i.e., a ∈ (βu c( p))α .
(m) (m) (m)
Conversely, we assume a ∈ (βu c( p))α . Since βu c0 ( p) → βu c( p), this
(m) α α,(m)
implies that a ∈ (βu c0 ( p)) , and so a ∈ M0 ( p; u) by Part (a). Further-
(m) (m) (m)
more, e ∈ c( p) implies 1/u ∈ βu c( p), and a ∈ (βu c( p))α implies
(m) −1 (m)
a∈( 1/u) ⇒ 1 , that is, a ∈ M ( p; u).
α,(m)
(c) First we assume that a ∈ M∞ ( p; u). Then
(m) (m)
Let x ∈ βu ∞ ( p) be given. Then there is an N ∈ N \ {1} such that |(βu x)k | ≤
N 1/ pk for all k, and so by (4.17), we get
∞
∞
(m)
|ak xk | ≤ |ak |sk (N /u) < ∞,
k=0 k=0
and
⎣ ⎤
β,(m) β,(m) β,(m)
M∞ ( p; u) := ≤ N ∈N\{1} M1 ( p; N / | u |) ≤ M2 ( p; N / | u |) .
Then
β,(m)
(β(m) β
u ∞ ( p)) = M∞ ( p; u). (4.18)
Since s(N /u) is a non-decreasing sequence of positive reals, it follows from (4.20)
by Lemma 4.19 that
R ∈ (s(N /u))−1 ⇒ c0 . (4.21)
n−1
n
ak x k = Rk (βx)k − Rn xn (n = 0, 1, . . .). (4.22)
k=o k=o
Since x ∈ βu ∞ (
p), there is N ∈ N \ {1} such that |(βx)k | ≤ N 1/ pk /|u k | for all
k, and so |xn | ≤ nk=o |(βx)k | ≤ sn (N /u) (n = 0, 1, . . .), and condition (4.21)
implies
Rx ∈ c0 . (4.23)
Finally, (4.22), (4.23), and (4.24) together imply ax ∈ cs. Thus a ∈ (βu ∞ ( p))β .
Conversely, we assume a ∈ (βu ∞ ( p))β . Then ax ∈ cs for all x ∈ βu ∞ ( p).
First e ∈ βu ∞ ( p) implies a = ae ∈ cs; hence, the sequence R is defined.
Furthermore, for each N ∈ N \ {1}, we have x = s(N /u) ∈ βu ∞ ( p); hence,
β
a ∈ (s(N /u))−1 ⇒ cs = M1 ( p; N /|u|) for all N ∈ N \ {1}, and condition (4.20)
holds again. As above, this implies (4.21) and then (4.23) for all x ∈ βu ∞ ( p).
Finally, from (4.22), we conclude that Rβx ∈ cs for all x ∈ βu ∞ ( p), that is,
R ∈ (u −1 ⇒ ∞ ( p))β and (4.19) holds.
4.4 Applications 125
and
a ∈ ≤ N ∈N\{1} (s (m+1) (N /u))−1 ⇒ cs. (4.26)
n
(m)
n−1
(m+1)
sk (N /u)Rk = ak sk (N /u) + Rn sn(m+1) (N /u). (4.27)
k=o k=o
for all n. It is easy to see that s (m+1) (N /u) is a non-decreasing sequence of positive
reals. Therefore, by Lemma 4.19, condition (4.26) implies
R ∈ (β(m) β
u ∞ ( p)) . (4.30)
(m+1) (m+1)
Let x ∈ βu ∞ ( p) be given. Since x ∈ βu ∞ ( p) if and only if y = βx ∈
β(m)
u ∞ ( p), condition (4.30) implies
(m+1)
Furthermore, there is N ∈ N \ {1} such that |(βu x)k ≤ N 1/ pk (k = 0, 1, . . .),
(m+1)
and so |xk | ≤ sk (N /u) for all k. Therefore (4.28) implies
Rx ∈ c0 . (4.32)
Finally, (4.22), (4.31), and (4.32) together imply ax ∈ cs. Thus we have shown
β,(m+1) (m+1)
M∞ ( p; u) → (βu ∞ ( p))β .
(m+1) (m+1)
Conversely, let a ∈ (βu ∞ ( p))β . Then ax ∈ cs for all x ∈ βu ∞ ( p).
(m+1)
First e ∈ βu ∞ ( p) implies a = ae ∈ cs; hence, the sequence R is defined.
(m+1)
Furthermore, for each N ∈ N \ {1}, we have x = s (m+1) (N /u) ∈ βu ∞ ( p).
Therefore condition (4.26) is satisfied. By Lemma 4.19, we have (4.28), and again,
126 4 Some Non-classical Sequence Spaces
(m+1)
this yields (4.32) for all x ∈ βu ∞ ( p). From (4.22), we conclude that Rβx ∈ cs
(m+1) (m)
for all x ∈ βu ∞ ( p), and consequently, R ∈ (βu ∞ ( p))β . This implies
R (m+1) = R (m) (R) ∈ ≤ N ∈N\{1} (N (1/ p)/ | u |)−1 ⇒ 1 by assumption. Thus
(m+1) β,(m+1) (m+1)
we have shown (βu ∞ ( p))β → M∞ ( p; u). Hence (βu ∞ ( p))β =
β,(m+1)
M∞ ( p; u).
This completes the proof.
From (4.22) and (4.23) the following can be shown easily by induction.
Remark 4.3 Let m ∈ N, p = ( pk )∞ k=0 be a sequence of positive reals and u ∈ U .
(m+1)
Then, a ∈ (βu ∞ ( p))β implies
∞
∞
ak x k = Rk(m) (β(m) (m)
u x)k for all x ∈ βu ∞ ( p). (4.33)
k=o k=o
and
n ∞ ⎣
⎤
m+k− j−1 N −1/ p j
sup | ak | < ∞. (4.35)
n
k− j
| uj |
j=0 k=n
Moreover, if a ∈ (β(m) β
u c0 ( p)) , then
⎨
∞
∞
∞ ⎣
⎤
⎩ 1
ak (β(m) (m)
m+k− j−1
ak x k = k− j u x) j for all x ∈ βu c0 ( p).
uj
k=o j=0 k= j
(4.36)
(b) Then a ∈ (β(m) β
u c( p)) if and only if conditions (4.33) and (4.34) hold and
⎨ −1
(m)
a∈⎩ 1/u ⇒ cs. (4.37)
Proof Given n ∈ N0 and a ∈ w, we write b(n,(m)) = b(n,(m)) (a) for the sequence
(n,(m)) n ⎣ m+k− j−1 ⎤
with b j = k= j k− j ak ( j = 0, 1, . . .). We also write y = β(m) x
and observe that x ∈ β(m)
u X if and only if y ∈ u
−1 ⇒ X . Interchanging the order of
summation, we obtain
4.4 Applications 127
⎨
n
n (m) k ⎣
n ⎤
ak ⎩ y =
m+k− j−1
ak x k = k− j yj
k=o
k=0 k k=0 j=0
n n ⎣
⎤
n
m+k− j−1 (n,(m))
= yj k− j ak = bj yj.
j=0 k= j j=0
and
(m) (n,(m))
1 ⎣ m+k− j−1 ⎤
bj bj ∞
= lim dn j = lim = k− j ak exists for each j.
uj n∩∞ n∩∞ uj uj
k= j
(4.39)
Let n ∈ N0 be given. By (4.38), there is N ∈ N \ {1} such that
n
(s,(m)) N −1/ p j
| bj | ≤ K D for all s ⊂ n;
| uj |
j=0
n
(m) N −1/ p j (s,(m)) N −1/ p j
n
| bj | = lim | bj | ≤ K D.
| uj | s∩∞ | uj |
j=0 j=0
(m)
Since n is arbitrary, we obtain (4.34), and so b(m) = (b j )∞
j=0 ∈ (u
−1 ⇒ c ( p))β .
0
We define the matrix W (m) = (wn(m) ∞
j )n, j=0 by
⎝ ⎣ ⎤
∞ m+k− j−1
ak (0 ≤ j ≤ k),
wn(m)
j = k=n k− j
0 ( j > k),
128 4 Some Non-classical Sequence Spaces
(n−1,(m))
Interchanging the order of summation and noting that bn = 0, we have
n−1
n−1
(n−1,(m))
n
(m) (m)
n
(m)
ak x k = bj yj = (b j −wn j )y j = b j y j −Wn(m) y, (4.40)
k=0 j=0 j=0 j=0
(m)
and a ∈ (βu c0 ( p))β and b(m) ∈ (u −1 ⇒ c0 ( p))β imply W (m) ∈ (u −1 ⇒ c0 ( p), c).
By Theorem 4.17, W (m) ∈ (u −1 ⇒ c0 ( p), c) implies
N −1/ p j
sup | wn(m)
j | < ∞ for some N ∈ N \ {1}, (4.41)
n | uj |
(m)
which is (4.35). Thus we have shown that a ∈ (βu c0 ( p))β implies conditions
(4.34) and (4.35). Moreover
wn(m)
j
lim = 0 for each j, (4.42)
n∩∞ uj
(m)
and hence, if a ∈ (βu c0 ( p))β , then b(m) ∈ (u −1 ⇒ c0 ( p))β , and conditions (4.41)
and (4.42) together imply W (m) ∈ (u −1 ⇒ c0 ( p), c0 ) by Theorem 4.17 and ([17],
8.3.6, p. 123). Therefore (4.36) follows from (4.40).
Conversely, we assume that conditions (4.34) and (4.35) hold. Then again (4.42)
holds, and this and condition (4.35) together imply W (m) ∈ (u −1 ⇒ c0 ( p), c0 ). Fur-
thermore, condition (4.34) implies b(m) ∈ (u −1 ⇒ c0 ( p))β . Now it follows from
(m) (m)
(4.40) that ax ∈ cs for all x ∈ βu c0 ( p), that is, a ∈ (βu c0 ( p))β .
(m) (m)
(b) We assume a ∈ (βu c( p))β . Then a ∈ (βu c0 ( p))β , and conditions (4.34)
and (4.35) follow from Part (a). Furthermore, we have seen in the proof of Part (a) that
(m)
a ∈ (βu c0 ( p))β implies b(m) ∈ (u −1 ⇒ c0 ( p))β and W (m) ∈ (u −1 ⇒ c0 ( p), c0 ). Let
(m) (m)
x ∈ βu c( p). Then there exists a complex number l such that βu x − le ∈ c0 ( p).
(m) (m)
We put y = x − l 1/u. Then y ∈ βu c0 ( p), and we obtain as in (4.40) with
z = β(m) y ∈ u −1 ⇒ c0 ( p)
⎨
n−1
n−1
n−1 (m)
ak x k = ak yk + l ak ⎩ 1/u
k=0 k=0 k=0 k
⎨
n
n−1 (m)
(m)
= b j z j − Wn(m) (z) + l ak ⎩ 1/u . (4.43)
j=0 k=0 k
Conversely we assume that conditions (4.34), (4.35), and (4.37) hold. Conditions
(4.34) and (4.35) together imply W (m) ∈ (u −1 ⇒ c0 ( p), c0 ) and b(m) ∈ (u −1 ⇒c0 ( p))β
as in the converse part of the proof of Part (a). Finally it follows from (4.43) and
condition (4.37) that ax ∈ cs for all x ∈ β(m) (m)
u c( p), that is, a ∈ (βu c( p)) .
β
n n ⎣
⎤
m+k− j−1 N −1/ p j
sup | a k | < ∞; (4.44)
n
k− j
| uj |
j=0 k= j
(m)
(b) a ∈ (βu c( p))γ if and only if condition (4.44) holds and
n k ⎣
⎤
m+k− j−1 1
sup | ak k− j | < ∞; (4.45)
n uj
k=0 j=0
(c) a ∈ (β(m) γ
u ∞ ( p)) if and only if for all N ∈ N \ {1}
n n ⎣
⎤
m+k− j−1 N 1/ p j
sup | ak | < ∞. (4.46)
n
k− j
| uj |
j=0 k= j
(m)
Proof We have a ∈ (βu X )γ if and only if
D ∈ (X, ∞ ) (4.47)
(m)
If X = βu c0 ( p), then, by Corollary 4.18 (a), condition (4.47) is equivalent to
∞
sup | dn j | N −1/ p j < ∞ for some N ∈ N \ {1}, (4.48)
n
j=0
which is condition (4.44). If X = β(m)u c( p), then, by Corollary 4.18 (b), condition
(4.47) is equivalent to condition (4.48) and
130 4 Some Non-classical Sequence Spaces
1 ⎣ m+k− j−1 ⎤
∞
n n
sup | dn j | = sup | k− j ak |
n n uj
j=0 j=0 k= j
n k ⎣
⎤
m+k− j−1 1
= sup | ak k− j | < ∞.
n uj
k=0 j=0
(m)
If X = βu ∞ ( p), then, by Theorem 4.15, condition (4.47) is equivalent to
∞
sup | dn j | N 1/ p j < ∞ for all N ∈ N \ {1},
n
j=0
which is (4.46).
This completes the proof.
(m)
Now, we characterize matrix transformations between the spaces βu X and Y and
(s)
βu Y , where X is any of the spaces c0 ( p), c( p), and ∞ ( p) and Y is any of the spaces
c0 (q), c(q), and ∞ (q) for bounded sequences p = ( pk )∞ ∞
k=0 and q = (qk )k=0 .
∞
k ⎣
⎤ 1/ p j
m+k− j−1 N
ank conver ges for all N ∈ N \ {1}
k− j
| uj |
k=0 j=0
and for all n = 0, 1, . . . . (4.49)
and
D (A,(m)) ∈ (u −1 ⇒ ∞ ( p), Y ), (4.50)
(A,(m))
where D (A,(m)) is the matrix with rows Dn = R (m) (An ) for n = 0, 1, . . . .
Proof We write D (m) = D (A,(m)) and R (n,(m)) = R (m) (An ) for short. First we
(m) (m)
assume that A ∈ (βu ∞ ( p), Y ). Then An ∈ (βu ∞ ( p))β for all n; hence,
condition (4.49) holds and
(m)
by Theorem 4.21, and An ∈ (βu ∞ ( p))β implies An (x) = R (n,(m)) (β(m) x) =
(m) (m)
Dn (β(m) x) for all x ∈ βu ∞ ( p) by Remark 4.3. Thus
l ∞ ⎣
⎤ −1/ p j
m+k− j−1 Nn
sup | a nk | <∞ (4.53)
l
k− j
| uj |
j=0 k=l
and
B (A,u,(m)) ∈ (c0 ( p), Y ), (4.54)
(m)
Proof We write B = B (A,u,(m)) for short. First we assume A ∈ (βu c0 ( p), Y ).
(m)
Then An ∈ (βu c0 ( p))β for all n = 0, 1, . . ., and this implies condition (4.53) by
(m)
Theorem 4.22(a). Let z ∈ c0 ( p) be given. Then x = (m) z/u ∈ βu c0 ( p). By
(m)
Theorem 4.22(a), An ∈ (βu c0 ( p))β also implies
Bn (z) = Bn (β(m)
u x) = An (x)for n = 0, 1, . . . . (4.55)
∞
∞ ⎣ ⎤ −1/ p j
m+k− j−1 Nn
| a nk | < ∞, (4.56)
k− j
| uj |
j=0 k= j
132 4 Some Non-classical Sequence Spaces
(m)
this and condition (4.53) together imply An ∈ (βu c0 ( p))β for all n by Theorem
(m) (m)
4.22(a). Let x ∈ βu c0 ( p) be given. Then z = βu x ∈ c0 ( p) and (4.55) holds
(m)
again. Now B(z) ∈ Y implies A(x) ∈ Y , and so A ∈ (βu c0 ( p), Y ).
This completes the proof.
Remark 4.5 From Theorem 4.25 and the well-known results that can be found in
[6, , Theorem 5.1, (5), (9) and (13)], we directly obtain the characterization for the
(m) (m) (m)
class (βu c0 ( p), c0 (q)), (βu c0 ( p), c(q)), and (βu c0 ( p), ∞ (q)).
Theorem 4.26 ([16, Theorem 5.3]) Let m ∈ N, p = ( pk )∞ k=0 be a bounded sequence
of positive reals, u ∈ U , and Y be any subset of w. Then A ∈ (β(m)
u c( p), Y ) if and
only if conditions (4.53) and (4.54) hold and
⎨
(m)
A⎩ 1/u ∈ Y. (4.57)
∞
∞ ⎣ ⎤ −1/ p j
m+k− j−1 Nn
lim ( sup ( | a nk | )qn ) = 0, (4.58)
M∩∞ n
k− j
| uj |
j=0 k= j
1 ⎣ m+k− j−1 ⎤
∞
lim | k− j ank |qn = 0 for all j, (4.59)
n∩∞ u j
k= j
and
∞
k ⎣
⎤
m+k− j−1 1 qn
lim | ank k− j | = 0; (4.60)
n∩∞ uj
k=0 j=0
(m)
4.5 Matrix Transformations Between the Spaces βu X 133
(m)
(b) A ∈ (βu c( p), c(q)) if and only if condition (4.53) holds and
∞
∞ ⎣ ⎤
m+k− j−1 N −1/ p j
sup | ank | < ∞ for some N ∈ N \ {1}, (4.61)
n
k− j
| uj |
j=0 k= j
there is a sequence (β j )∞
j=0 such that
1 ⎣ m+k− j−1 ⎤
∞
lim | k− j ank − β j |qn = 0 for all j, (4.62)
n∩∞ uj
k= j
there is a sequence (γ j )∞
j=0 such that
1 ⎣ m+k− j−1 ⎤
∞
∞
lim ( sup ( | k− j ank − γ j |M −1/ p j )qn ) = 0, (4.63)
M∩∞ n uj
j=0 k= j
and
∞
k ⎣
⎤
m+k− j−1 1
lim | ank k− j − β|qn = 0 for some complex number β;
n∩∞ uj
k=0 j=0
(4.64)
(m)
(c) A ∈ (βu c( p), ∞ (q)) if and only if condition (4.53) holds
∞
∞ ⎣ ⎤ −1/ p j
m+k− j−1 Nn
sup ( | a nk | )qn < ∞ for some N ∈ N \ {1}, (4.65)
n
k− j
| uj |
j=0 k= j
and
∞
k ⎣
⎤
m+k− j−1 1 qn
sup | ank k− j | < ∞. (4.66)
n uj
k=0 j=0
In this section, we shall study the sequence spaces m(φ) and n(φ) introduced by
Sargent [18] which are closely related to p spaces. We present here some properties
of these spaces and their relations with p spaces. We characterize the matrix classes
(X, m(φ)) and (X, n(φ)), where X is any FK space. We also give necessary and
sufficient conditions for infinite matrices A to belong to any of the classes (m(φ), p ),
(n(φ), p ), ( p , m(φ)), and ( p , n(φ)) for 1 ≤ p ≤ ∞ and establish estimates for
the norms of the corresponding operators L A . Finally, we characterize the compact
134 4 Some Non-classical Sequence Spaces
operators L A defined by the matrices A in the above classes and for the classes
(n(φ), c0 ), (n(φ), c), and (n(φ), ∞ ), except for the case (1 , m(φ)).
Sargent [18] defined the following sequence spaces.
Let C denote the space whose elements are finite sets of distinct positive integers.
Given any set σ ∈ C, we denote by c(σ) the sequence with
⎝
1 (n ∈ σ)
cn (σ) =
0 (n ≥ ∈ σ).
Further ⎝ ∞
⎧
Cs = σ ∈ C: cn (σ) ≤ s ,
n=1
that is, Cs is the set of those σ whose support has cardinality at most s. The set χ
consists of all real sequences (φk )∞
k=1 such that
φk
φ1 > 0, βφk ⊂ 0 and β ≤ 0 (k = 1, 2, . . . ),
k
i.e.,
For φ ∈ χ, the following sequence spaces were defined and studied by Sargent
[18] and further studied in [19].
⎝ ⎛ ⎧
1
m(φ) = x ∈ w: sup sup |xk | < ∞
s⊂1 σ∈Cs φs k∈σ
⎝ ∞ ⎛ ⎧
n(φ) = x ∈ w: sup |u k |βφk < ∞
u∈S(x) k=1
where βφk = φk − φk−1 , φ0 = 0, and S(x) denotes the set of all sequences that are
rearrangements of x.
Remark 4.6 (i) If φn = 1 for all n ∈ N, then m(φ) = 1 , n(φ) = ∞ , and if φn = n
for all n ∈ N, then m(φ) = ∞ , n(φ) = 1 .
(ii) If x ∈ m(φ) [x ∈ n(φ)] and u ∈ S(x), then u ∈ m(φ) [u ∈ n(φ)] and
◦u◦ = ◦x◦.
4.6 Sequence Spaces of Sargent 135
(iii) If x ∈ m(φ) [x ∈ n(φ)] and | u n |≤| xn | for every positive integer n, then
u ∈ m(φ) [u ∈ n(φ)] and ◦u◦ ≤ ◦x◦.
(iv) If φn = 1, then m(φ) = 1 , while m(φ) = ∞ for φn = n (n ∈ N).
(v) (m(φ))β and n(φ) ((n(φ))β and m(φ)) are norm isomorphic, i.e., (m(φ))β
n(φ) and (n(φ))β m(φ).
Proof Let s be an arbitrary positive integer, let σ ∈ Cs , and let τ consist of the
elements of σ which exceed s. Then, since φn ≤ φn+1 and (n + 1)φn ⊂ nφn+1 , we
find that
s
|βφn | ≤ βφn + (φn − φn−1 ) ≤ φs + {φn−1 /(n − 1)} ≤ 2φs ,
n∈σ n=1 n∈τ n∈τ
n
n
|br xr | ≤ ◦x◦ | cr | βφr . (4.67)
r =1 r =1
Proof In view of Remark 4.6 (ii), it is sufficient to consider the case when br = cr
(r = 1, 2, . . . , n). Then X r = ri=1 | xi |,
n
n
|br xr | ≤ {| cr | − | cr +1 |} X r + | cn | X n
r =1 r =1
n−1
≤ ◦x◦ {| cr | − | cr +1 |} φr + ◦x◦ | cn | φn
r =1
n
= ◦x◦ | cr | βφr .
r =1
Lemma 4.30 In order that m(φ) ⊆ m(ψ) [n(φ) ∼ n(ψ)], it is necessary and
sufficient that sups⊂1 ψφss < ∞.
Proof Inclusion follows immediately from the condition (by Remark 4.6 (v) the
second version follows from the first). Now suppose that m(φ) ⊆ m(ψ). Since
βφ ∈ m(φ), it follows that βφ ∈ m(φ), and hence, we find that for every positive
integer s,
136 4 Some Non-classical Sequence Spaces
s
φs = βφr ≤ ψs ◦βφ◦m(ψ) ,
r =1
From Lemma 4.30, we have m(φ) = m(ψ), where ψ = (n 1/q ). By Lemma 4.28, we
have βψ ∈ m(ψ), but βψ ∈ p , which is a contradiction. Hence, (a) is proved.
(b) Let p → m(φ). Then the arguments similar to those used in the proof of (a)
show that
4.6 Sequence Spaces of Sargent 137
provided the expression on the right-hand side exists and is finite which is the case
whenever a ∈ X β ([17, Theorem 7.2.9, p. 107]).
First we give the characterization of the class (X, m(φ)) where X is any BK space
and establish an estimate for the operator norm of L A .
Theorem 4.33 ([20, Theorem 1], [21, Theorem 2.1]) Let X be a BK space. Then
(a) we have A ∈ (X, m(φ)) if and only if
138 4 Some Non-classical Sequence Spaces
⎦ ⎦ ⎛
⎦ ⎦⇒
1 ⎦ ⎦
◦A◦(X,m(φ)) = sup sup ⎦ An ⎦ < ∞. (4.69)
t⊂1 τ ∈Ct φt ⎦ n∈τ ⎦
X
Proof (a) If X and Y are sequence spaces and A ∈ (X, Y ), we shall define the maps
f A : X ∩ Y and f A,n : X ∩ C (n = 1, 2, . . .) by
1
| gt,τ (x) |≤ | f A,n (x) |≤◦ f A (x) ◦m(φ) (x ∈ X ).
φt n∈τ
This implies
1
| An (x) |≤ 4◦A◦(X,m(φ)) , (4.73)
φt n∈τ
and hence, A(x) ∈ m(φ) for all x ∈ S X . Since S X is absorbing, A(x) ∈ m(φ) for all
x ∈ X.
4.7 Matrix Transformations on and into m(φ) and n(φ) 139
(b) If A ∈ (X, m(φ)), then L A ∈ B(X, m(φ)), and so for all x ∈ S X , τ ∈ Ct and
t ⊂1
∞ ⎛
1
1
ank xk ≤ |An (x)| ≤ ◦L A (x)◦m(φ) ≤ ◦L A ◦
φt n∈τ
φt n∈τ
k=1
Proof (a) Let A ∈ (X, n(φ)), B ∈ S(A), and N → N, N finite. Using the same
notations and argument as in the proof of Theorem 4.33, we conclude that
g B,N := f B,n βφn ∈ X .
n∈N
Again an application of the uniform boundedness principle yields that for all B ∈
S(A), and N → N, N finite, and x ∈ S X we have
n0
| Bn (x) | βφn ≤ 4 max | Bn (x)βφn |≤ 4◦A◦(X,n(φ)) , (4.76)
N →{1,...,n 0 }
n=1 n∈N
If A ∈ (X, n(φ)), then L A ∈ B(X, n(φ)), and so for all m ∈ N, all subsets Nm of
{1, 2, . . . , m}, all B ∈ S(A), and all x ∈ S X
⎨
∞
m
⎩ bnk βφn xk ≤ | Bn (x)βφn |≤ ◦L A (x)◦n(φ) ≤◦ L A ◦,
k=1 n∈Nm n=1
for all m ∈ N, all subsets Nm of {1, 2, . . . , m} and all B ∈ S(A), and consequently,
⎦ ⎦⇒
⎦ ⎦
⎦ ⎦
◦A◦(X,n(φ)) = sup sup⇒N ⎦ Bn βφn ⎦ ≤◦ L A ◦,
B∈S(A) ⎦ ⎦
n∈N X
the first inequality in (4.75). Furthermore, as in the proof of Theorem 4.33, by (4.76)
we get
m
| Bn (x)βφn |≤ 4 max | Bn (x)βφn |
Nm →{1,...,m}
n=1 n∈Nm
⎦ ⎦⇒
⎦ ⎦
⎦ ⎦
≤ 4 max ⎦ ⎦ Bn βφ ⎦
n ⎦ ≤ 4◦A◦(X,n(φ)) ,
Nm →{1,...,m} ⎦ ⎦
n∈Nm X
4.7 Matrix Transformations on and into m(φ) and n(φ) 141
(A;N )
bk = ank (k = 1, 2, . . . ).
n∈N
Theorem 4.35 ([21, Theorem 2.3]) (a) We have A ∈ (m(φ), ∞ ) if and only if
∞
⎛
◦A◦(m(φ),∞ ) = sup sup |u k |βφk < ∞; (4.77)
n u∈S(An ) k=1
◦L A ◦ = ◦A◦(m(φ),∞ ) . (4.78)
◦L A ◦ = ◦A◦(n(φ),∞ ) . (4.80)
Proof Since m(φ) and n(φ) are BK spaces, so are (m(φ))β with ◦·◦⇒m(φ) and (n(φ))β
with ◦·◦⇒n(φ) ([17, Theorem 4.3.15, p. 64]); also since (m(φ))β = n(φ) and (n(φ))β =
m(φ), the norms ◦·◦⇒m(φ) and ◦·◦n(φ) and the norms ◦·◦⇒n(φ) and ◦·◦m(φ) are equivalent
on (m(φ))β ) and on (n(φ))β ([17, Corollary 4.2.4, p. 56]).
Thus Parts (a) and (b) are an immediate consequence of [22, Theorem 1.23, p.
155], and Parts (c) and (d) are an immediate consequence of [23, Satz 1].
The proof is complete.
Proof Since ⇒p and q are norm isomorphic for 1 ≤ p < ∞, and ◦ · ◦⇒∞ = ◦ · ◦1
β
on ∞ , Parts (a) and (b) follow from Theorems 4.33 and 4.34 by replacing the norm
◦ · ◦⇒X in (4.69), (4.70), (4.74), and (4.75) by ◦ · ◦q (1 ≤ q ≤ ∞).
4.7 Matrix Transformations on and into m(φ) and n(φ) 143
Finally we characterize the classes (m(φ), p ) and (n(φ), p ) for 1 < p < ∞
and give estimates for the norms of L A . Given any infinite matrix A, we denote its
transpose by At .
Proof The necessity and sufficiency of the conditions in (4.89) and (4.91) follow
from [18, Lemma 14] and those in (4.87) and (4.85) in Corollary 4.36, respectively.
Also the estimates in (4.90) and (4.92) follow from [4, Lemma 2] and those in (4.88)
and (4.86) in Corollary 4.36, respectively.
Remark 4.7 The characterizations of the classes ( p , m(φ)) in Corollary 4.36(a) and
of (n(φ), p ) in Corollary 4.36(b) can be found in [18, Theorems 1 and 2].
Remark 4.8 The proof of Corollary 4.37 extends to the case p = 1; hence, we obtain
alternative characterizations for the classes (m(φ), 1 ) and (n(φ), 1 ) and estimates
for the operator norms from those given in Theorem 4.35(c) and (d).
Remark 4.9 We observe that ◦A◦(X, p ) = ◦At ◦(q ,X β ) for X = m(φ) or X = n(φ)
and 1 ≤ p ≤ ∞ (by [18, Lemma 14] and [4, Lemma 2 (4)]).
Exercises
1. If 0 < pk ≤ 1, then show that β ( p) = ∞ ( p).
2. Show that ββ ( p) = M∞ ( p), 0 < pk < 1.
144 4 Some Non-classical Sequence Spaces
9. Let p ∈ ∞ . Prove that A ∈ (c0 ( p), ∞ ( p)) if and only if there exists an absolute
constant B > 1 such that
⎝ ⎧ pn
−1/ pk
sup | ank | B < ∞.
n
k
11. Let 0 < pk ≤ 1. Then prove that A ∈ (w( p), c) if and only if
16. Characterize the classes (m(φ), λp ) and (n(φ), λp ) for 1 ≤ p ≤ ∞ and establish
estimates for the norms of the corresponding operators L A .
References
1. Nakano, H.: Modulared sequence spaces. Proc. Jpn. Acad. 27(2), 508–512 (1951)
2. Simons, S.: The sequence spaces ( pν ) and m( pν ). Proc. Lond. Math. Soc. 3(1), 422–436
(1965)
3. Lascarides, C.G., Maddox, I.J.: Matrix transformations between some classes of sequences.
Proc. Camb. Phil. Soc. 68, 99–104 (1970)
4. Sargent, W.L.: On compact matrix transformations between sectionally bounded B K -spaces.
J. Lond. Math. Soc. 41(1), 79–87 (1966)
5. Maddox, I.J.: Continuous and Köthe-Toeplitz duals of certain sequence spaces. Proc. Camb.
Phil. Soc. 65, 431–435 (1969)
6. Grosse-Erdmann, K.G.: Matrix transfomations between the sequence spaces of Maddox. J.
Math. Anal. Appl. 180, 223–238 (1993)
7. Grosse-Erdmann, K.G.: The structure of the sequence spaces of Maddox. Canad. Jour. Math.
44, 298–307 (1992)
8. Lascarides, C.G.: A study of certain sequence spaces of Maddox and a generalization of a
theorem of Iyer. Pacific J. Math. 38(2), 487–500 (1971)
9. Kızmaz, H.: On certain sequence spaces. Canad. Math. Bull. 24(2), 169–176 (1981)
10. Malkowsky, E., Parashar, S.D.: Matrix transformations in space of bounded and convergent
difference sequences of order m. Analysis 17, 87–97 (1997)
11. Ahmad, Z.U., Mursaleen, M.: Köthe-Toeplitz duals of some new sequence spaces and their
matrix maps. Publ. Inst. Math. (Beograd) 42, 57–61 (1987)
12. Malkowsky, E.: Absolute and ordinary Köthe-Toeplitz duals of certain sequence spaces. Publ.
Inst. Math. (Beograd) 46(60), 97–104 (1989)
13. Malkowsky, E., Mursaleen, M.: Qamruddin, Generalized sets of difference sequences, their
duals and matrix transformations. In: Jain, P.K., Malkowsky, E. (eds.) Sequence Spaces and
Applications, pp. 68–83. Narosa, New Delhi (1999)
14. Başarır, M., Et, M.: On some new generalized difference sequence spaces. Period. Math. Hung.
35(3), 169–175 (1997)
15. Malkowsky, E.: A note on the Köthe-Toeplitz duals of generalized sets of bounded and con-
vergent difference sequences. J. Anal. 4, 81–91 (1995)
16. Malkowsky, E., Mursaleen, M., Suantai, S.: The dual spaces of sets of difference sequences of
order m and matrix transformations. Acta Math. Sin. Engl. Ser. 23(3), 521–532 (2007)
17. Wilansky, A.: Summability through Functional Analysis, Mathematics Studies 85. North-
Holland, Elsevier, Amsterdam (1984)
18. Sargent, W.L.: Some sequence spaces related to the p spaces. J. Lond. Math. Soc. 35, 161–171
(1960)
19. Mursaleen, M.: On some geometric properties of a sequence space related to p . Bull. Australian
Math. Soc. 67, 343–347 (2003)
20. Malkowsky, E., Mursaleen, M.: Matrix transformations between F K -spaces and the sequence
spaces m(φ) and n(φ). J. Math. Anal. Appl. 196, 659–665 (1995)
146 4 Some Non-classical Sequence Spaces
21. Malkowsky, E., Mursaleen, M.: Compact matrix operators between the spaces m(φ), n(φ) and
p . Bull. Korean Math. Soc. 48(5), 1093–1103 (2011)
22. Malkowsky, E., Rakočević, V.: An introduction into the theory of sequence spaces and measures
of noncompactness. Zbornik Radova. Mat. Institut SANU (Beograd) 9(17), 143–234 (2000)
23. Malkowsky, E.: Klassen von Matrix abbildungen in paranormierten F K -Raumen. Analysis 7,
275–292 (1987)
Chapter 5
Measures of Non-compactness
5.1 Preliminaries
Let us recall some well-known definitions and results. If M and S are subsets of a
metric space (X, d) and α > 0, then the set S is called α-net of M if for any x ∞ M
there exists s ∞ S, such that d(x, s) < α. If the set S is finite, then the α-net S of M
is called finite α-net. The set M is said to be totally bounded if it has a finite α-net for
every α > 0. It is well known that a subset M of a metric space X is compact if every
sequence (xn ) in M has a convergent subsequence, and in this case the limit of that
subsequence is in M. The set M is said to be relatively compact if the closure M̄ of
The function dH is called the Hausdorff distance, and dH (S, Q) (S, Q ∞ M X ) is the
Hausdorff distance of sets S and Q.
Let us remark that if ∅ → = F ≤ X, r > 0 and
B(F, r ) = B(x, r ) = {y ∞ X : d(y, F) < r }
x∞F
is the open ball with center in F and radius r , then (5.1) is equivalent to
Let X and Y be infinite-dimensional complex Banach spaces and denote the set
of bounded linear operators from X into Y by B(X, Y ). We put B(X ) = B(X, X ).
For T in B(X ; Y ), N (T ) and R(T ) will denote, respectively, the null space and the
range space of T . A linear operator L from X to Y is called compact (or completely
continuous) if D(L) = X for the domain of L, and for every sequence (x n ) ∞ X
such that ∩ x n ∩≤ C, the sequence (L(xn )) has a subsequence which converges
in Y . A compact operator is bounded. An operator L in B(X, Y ) is of finite rank if
dim R(L) < ∞. An operator of finite rank is clearly compact. Let F(X, Y ), C(X, Y )
denote the set of all finite rank and compact operators from X to Y , respectively. Set
F(X ) = F(X, X ) and C(X ) = C(X, X ).
Let X be a vector space over the field R. A subset E of X is said to be convex if
δx + (1 − δ)y ∞ E for all x, y ∞ E and for all δ ∞ (0, 1). Clearly the intersection of
any family of convex sets is a convex set. If F is a subset of X , then the intersection
of all convex sets that contain F is called convex cover or convex hull of F denoted
by co(F).
The vector subspace linF is the set of all linear combinations of elements in F.
We shall prove that there is an analogous representation of the set co(F). Let us
5.1 Preliminaries 149
mention that a convex combination of elements of the set F is an element of the form
n
δ1 x1 + δ2 x2 + · · · + δn xn , xi ∞ F, δi ≥ 0 (i = 1, . . . , n), i=1 δi = 1 (n ∞ N).
Let us write cvx(F) for the set of all convex combinations of elements of the set F.
Let Q be a non-empty and bounded subset of a normed space X . Then the convex
closure of Q, denoted by Co(Q), is the smallest convex and closed subset of X that
contains Q. Note that Co(Q) = co(Q).
Lemma 5.1 If X is a vector space over the field F and E, E 1 , . . . , E n are convex
subsets of X and F ≤ X , then
cvx(E) ≤ E, (5.2)
n
n
xi ∞ E, δi ≥ 0 (i = 1, . . . , n), δi = 1 ⇒ δi xi ∞ E, (5.5)
i=1 i=1
Let us remark
150 5 Measures of Non-compactness
n
n
n
θi = δ φi + (1 − δ) εi = δ + (1 − δ) = 1. (5.7)
i=1 i=1 i=1
n
By (5.6) and (5.7) we have δx + (1 − δ)y = i=1 θi z i ∞ S.
This completes the proof.
Proof The statements in (5.11) and (5.13) follow from Definition 5.1. Clearly
φ(Q) ≤ φ( Q̄). Let α > 0, Si be a bounded subset of X with diam(Si ) < α
n S . Then Q̄ ≤ ⊂n S = ⊂n S̄ . Since
for i = 1, 2, . . . , n, and Q ≤ ⊂i=n i i=1 i i=1 i
¯ we conclude φ(Q) ≤ (Q). This proves equality (5.12). From
diam(Si ) =diam( Si),
(5.13), we have φ(Q 1 ) ≤ φ(Q 1 ⊂ Q 2 ) and φ(Q 2 ) ≤ φ(Q 1 ⊂ Q 2 ), and so
Let max{φ(Q 1 ), φ(Q 2 )} = s and α > 0. By Definition 5.1 we know that Q 1 and
Q 2 can be covered by a finite number of subsets of diameter smaller than s + α.
Obviously, the union of these covers is a finite cover of Q 1 ⊂ Q 2 . Hence we have
φ(Q 1 ⊂ Q 2 ) ≤ s + α, and now we obtain (5.14) from (5.16). From Q 1 ◦ Q 2 ≤ Q 1
and Q 1 ◦ Q 2 ≤ Q 2 we obtain φ(Q 1 ◦ Q 2 ) ≤ φ(Q 1 ) and hence φ(Q 1 ◦ Q 2 ) ≤
min{φ(Q 1 ), φ(Q 2 )}. This proves inequality (5.15).
This completes the proof.
for each n. The assumption of our theorem and (5.17) together imply φ(X 1 ) =
0, hence X 1 is a relatively compact set. Thus the sequence (xn ) has a convergent
subsequence (xkn ) with x = lim xkn ∞ X, say. Since Fn is closed in X, we get x ∞ Fn
for all n = 1, 2, . . ., that is, x ∞ F∞ .
This completes the proof.
If X is a normed space, then the function φ has some additional properties con-
nected with the vector (linear) structure of a normed space.
Lemma 5.7 Let Q, Q 1 and Q 2 be bounded subsets of a normed space X. Then:
Q 1 + Q 2 ≤ ⊂i=1
n
⊂mj=1 (Si + G j ) and diam(Si + G j ) < d + p. (5.22)
It follows from (5.22) that φ(Q 1 + Q 2 ) < d + p. This shows inequality (5.18). Let
x ∞ X. By (5.18) it follows that
and
Now we prove (5.21). Clearly φ(Q) ≤ φ(co Q), and it suffices to show φ(co Q) ≤
φ(Q). Let Si be a bounded subset of X with diam(Si ) < d for each i = 1, . . . , n
n
and Q = i=1 Si .
By Lemma 5.1 it follows that
n
n
co(Q) = δi xi : δi ≥ 0, δi = 1, xi ∞ co(Si ) (i = 1, . . . , n) . (5.25)
i=1 i=1
n
Let α > 0 and S = (δ1 , . . . , δn ) : δi = 1, δi ≥ 0 (i = 1, . . . , n) . Then S is
i=1
a compact subset of (R , ∩.∩∞ ), where ∩(δ1 , . . . , δn )∩∞ = max1≤i≤n |δi |. We put
n
α
∩(δ1 , . . . , δn ) − (t j,1 , . . . , t j,n )∩∞ < n. (5.26)
M
Since
n
n
n
δi xi = t j,i xi + (δi − t j,i )xi , (5.27)
i=1 i=1 i=1
Remark 5.1 We gather below some more properties that follow immediately from
the definition. Let X be an infinite-dimensional normed space and B X the closed unit
ball in X.
(i) Non-singularity: φ is equal to zero on every single-element set;
(ii) Lipschitzity: | φ(Q 1 ) − φ(Q 2 ) |≤ 2dH (Q 1 , Q 2 ), where dH denotes the
Hausdorff distance defined in Sect. 5.1.
(iii) Continuity: for any Q ≤ X and any α > 0, there is a β > 0 such that
| φ(Q) − φ(Q 1 ) |< α for all Q 1 satisfying ν(Q, Q 1 ) < β.
Proof The first assertion follows by the definition of φ. To prove the second assertion
we use the well-known Ljusternik-Šnirelman-Borsuk theorem (If Sn is the unit sphere
of an n-dimensional real normed space E n , Fi a closed subset of E n for each i =
1, . . . , n and Sn ≤ ⊂i=1n F , then there exists i ∞ {1, . . . , n} such that the set
i 0
Sn ◦ Fi0 contains a pair of antipodal points, that is, there exists x0 ∞ Sn ◦ Fi0 , such
that {x0 , −x0 } ≤ Sn ◦ Fi0 ). Clearly φ(B X ) ≤ 2. If φ(B X ) < 2, then there exist
bounded and closed subsets Q i of X with diam(Q i ) < 2 for i = 1, . . . , n such that
B X ≤ ⊂i=1 n Q . Let {x , . . . x } be a linearly independent subset of X and E be the
i 1 n n
set of all linear combinations of elements of the set {x1 , . . . , xn } with real coefficients.
Clearly E n is a real n-dimensional normed space (the norm on E n , of course, being
the restriction of the norm on X ). By Sn = {x ∞ E n : ∩x∩ = 1}, we denote the
unit sphere of E n . Let us mention that Sn ≤ ⊂i=1 n S ◦ Q , diam(S ◦ Q ) < 2 and
n i n i
Sn ◦ Q i is a closed subset of E n for each i = 1, . . . n. This is a contradiction to the
Ljusternik-Šnirelman-Borsuk theorem and the proof is complete.
Usually, it is complicated to find the exact value of φ(Q). Another measure of non-
compactness, which is more applicable in many cases, is given in the next definition.
Definition 5.2 Let (X, d) be a metric space and Q a bounded subset of X. Then
the Hausdorff measure of non-compactness (λ-measure or ball measure of non-
compactness) of the set Q, denoted by λ(Q), is defined to be the infimum of the set
of all reals α > 0 such that Q can be covered by a finite number of balls of radii < α,
that is,
n
λ(Q) = inf{α > 0 : Q ≤ B(xi , ri ), xi ∞ X, ri < α (i = 1, . . . , n) n ∞ N}.
i=1
(5.29)
The function λ is called the Hausdorff measure of non-compactness.
5.3 The Hausdorff Measure of Non-compactness 155
Q̄ ≤ ⊂nk=0 B(xk , λ(Q) + ι/2) = ⊂nk=0 B(xk , λ(Q) + ι/2) ≤ ⊂nk=0 B(x0 , λ(Q) + ι).
Now let ζ = max{λ(Q 1 ), λ(Q)} and ι > 0 be given. Then by the definition of λ, Q 1
and Q 2 can be covered by finite unions of open balls of radius ζ + ι. Obviously the
union of these covers is a finite cover of Q 1 ⊂ Q 2 . This implies λ(Q 1 ⊂ Q 2 ) ≤ ζ + ι,
and since ι > 0 was arbitrary, it follows that λ(Q 1 ⊂ Q 2 ) ≤ ζ, which together with
(5.31) imply (iv).
(v) It follows from (iii) that λ(Q 1 ◦ Q 2 ) ≤ λ(Q i ) for i = 1, 2, hence
λ(Q 1 ◦ Q 2 ) ≤ min{λ(Q 1 ), λ(Q 2 )}.
This completes the proof.
Lemma 5.10 Let Q, Q 1 and Q 2 be bounded subsets of the normed space (X, ∩ . ∩).
Then
(i) λ(Q 1 + Q 2 ) ≤ λ(Q 1 ) + λ(Q 2 ),
(ii) λ(Q + x) = λ(Q) for all x ∞ X,
(iii) λ(δQ) = |δ|λ(Q) for all δ ∞ F,
(iv) λ(Q) = λ(co(Q)).
Proof (i) Let ζ j = λ(Q j ) for j = 1, 2, ζ = ζ1 + ζ2 , and ι > 0 be given. Then there
( j)
are n j = n j (ι) ∞ N0 and xk ∞ X (0 ≤ k ≤ n j ) for j = 1, 2 such that
156 5 Measures of Non-compactness
( j)
Q j ≤ ⊂nk=0 B(xk , ζ j + ι/2) for j = 1, 2. (5.32)
(iii) Since it is trivial for δ = 0, we assume that δ → = 0. Let ι > 0 be given. Then
we have
Q ≤ ⊂nk=0 B(xk , λ(Q) + ι). (5.33)
Let y ∞ δQ be given. Then there are x ∞ Q such that y = δx, and k0 ∞ {0, 1, . . . n}
such that x ∞ B(xk , λ(Q) + ι). We put yk = δxk for k = 0, 1, . . . and obtain
∩ y − yk0 ∩ =| δ |∩ x − xk ∩<| δ | (λ(Q) + ι). This implies δQ ≤
⊂∞k=0 B(yk , | δ | (λ(Q) + ι)). Since ι > 0 was arbitrary, we conclude that
λ(δQ) ≤| δ | λ(Q). Furthermore, it follows by what we have just shown that
λ(Q) = λ(δ−1 (δQ)) ≤| δ−1 | λ(δQ), hence | δ | λ(Q) ≤ λ(δQ).
(iv) Since Q ≤ co(Q), we obtain λ(Q) ≤ λ(co(Q)). We have to show
Let ι > 0 be given. Then we have (5.33), and every ball Bk = B(xk , λ(Q) + ι) is a
convex set. To see this, let x, y ∞ Bk and 0 ≤ δ ≤ 1. Then we have
Putting ψk = tδk + (1 − t)μk , γk = tδk /ψk for ψk > 0 and γk = 0 for ψk = 0, and
z k = γk xk + (1 − γ)yk (k = 0, 1, . . . , n), we obtain
n
n
n
ψk z k = (γk xk + (1 − γ)yk ) = (tδk xk + (1 − t)μk yk ) = z.
k=0 k=0 k=0
Remark 5.2 Lemmas 5.9 and 5.10 restate the properties contained in Lemmas 5.5 and
5.7. Moreover, the properties gathered in Remark 5.1 hold for θ except Lipschitzity,
i.e., in case θ we have | θ(Q 1 ) − θ(Q 2 ) |≤ dH (Q 1 , Q 2 ).
The following result gives the relation between the Hausdorff measure of non-
compactness and the Hausdorff distance.
Proof Let α > 0 and d = dH (Q 1 , Q 2 ). Then it follows from (5.29) and (5.1) that
there exists a finite set S ≤ X, such that
k
BX ≤ {xi + (q + α)B X }. (5.43)
i=1
Since q + α < 1, by (5.44) we have q = 0, that is, B X is a totally bounded set. But
this is impossible since X is an infinite-dimensional space. Hence λ(B X ) = 1.
This completes the proof.
Remark 5.4 The sharpness of the second inequality in (5.42) follows from Theorems
5.8 and 5.14. From the following example we see that the first inequality is also sharp.
Indeed, take X = c0 with sup-norm. Let Q = (ek )∞ k=1 be its basis. Since the
diameter of any set containing more than one element of Q is equal to 1, we have
that φ(Q) = 1. On the other hand, λ(Q) = 1 because the distance from any infinite
subset of Q to any element of c0 is not smaller than 1.
Now we shall point out the well-known result of Goldenštein et al.
[3, Theorem 1] concerning the Hausdorff measure of non-compactness in Banach
spaces with Schauder basis. Let X be a Banach space with a Schauder ∞ basis
{e1 , e2 , . . .}. Then each element x ∞ X has a unique representation x = i=1 φi (x)ei
where the functions φi are the basis functionals. Let Pn: X ⊕ ∈ X be the projector
n
onto the linear span of {e1 , e2 , . . . , en }, that is Pn (x) = i=1 φi (x)ei . Then in view
of the Banach–Steinhaus theorem, all operators Pn and I − Pn are equibounded.
Now we have the following important result.
Q ≤ Pn Q + (I − Pn )Q. (5.46)
λ(Q) ≤ λ(Pn Q) + λ((I − Pn )Q) = λ((I − Pn )Q) ≤ sup ∩(I − Pn )(x)∩. (5.47)
x∞Q
Now we obtain
λ(Q) ≤ inf sup ∩(I − Pn )(x)∩ ≤ lim sup sup ∩(I − Pn )(x)∩ , (5.48)
n x∞Q n∈∞ x∞Q
Hence it suffices to show the first inequality in (5.45). Let α > 0 and {z 1 , . . . , z k } be
a [λ(Q)+α]-net of Q. Since Q ≤ {z 1 , . . . , z k }+[λ(Q)+α]B X ., this implies that for
any x ∞ Q there exist z ∞ {z 1 , . . . , z k } and s ∞ B X such that x = z + [λ(Q) + α]s,
and so
This yields
lim sup sup ∩(I − Pn )(x)∩ ≤ (λ(Q) + α) lim sup ∩I − Pn ∩.
n∈∞ x∞Q n∈∞
Hence
1
lim sup sup ∩(I − Pn )(x)∩ ≤ (λ(Q) + α),
a n∈∞ x∞Q
Proof For Q ∞ M X and a fixed n ∞ N let us put μn = μn (Q) = supx∞Q ∩(I −Pn )x∩.
Since X is a monotone B K space with AK , it follows that
5.3 The Hausdorff Measure of Non-compactness 161
Since α > 0 was arbitrary, we have μn (Q) ≥ μn+1 (Q) ≥ 0 for all n, and so
limn∈∞ μn (Q) exists for all Q ∞ M X , and we can replace lim sup by lim in (5.45).
It remains to show that lim supn∈∞ ∩I − Pn ∩ = 1. Since ∩.∩ is monotone,
∩(I − Pn )(x)∩ = ∩x − x [n] ∩ ≤ ∩x∩ for all x ∞ X and all n, hence
Remark 5.5 Let us mention that concerning the number a in Theorem 5.15, if X =
c0 , we have a = 1, but if X = c, then a = 2.
1
lim sup ∩(I − Pn )(x)∩ ≤ λ(Q) ≤ lim sup ∩(I − Pn )(x)∩ .
2 n∈∞ x∞Q n∈∞ x∞Q
Let x = (xk )∞
k=0∞ c be given and l = limk∈∞ xk . Then x has a unique repre-
sentation x = le + ∞ (k)
k=0 (x k − l)e , and so
∞
(I − Pn )(x) = (xk − l)e(k) for all n,
k=n+1
and
∩(I − Pn )(x)∩ = sup | xk − l | for all n. (5.54)
k≥n+1
As in the proof of Theorem 5.16, we can show that limn∈∞ μn (Q) exists for
all Q ∞ Mc . It remains to show that limn∈∞ ∩I − Pn ∩ = 2. Since | l |=
limk∈∞ | xk |≤ supk | xk |=∩ x ∩, then we get
Hence
∩I − Pn ∩ ≤ 2. (5.55)
Hence
∩I − Pn ∩ ≥ 2 for all n. (5.56)
The identity in (5.49) yields an explicit expression for the Hausdorff measure of non-
compactness of any bounded subset of an arbitrary monotone B K space with AK .
This indicates why the Hausdorff measure of non-compactness is the most suitable
measure of non-compactness for our purposes. Since p (1 ≤ p < ∞) and c0 are
B K spaces with AK with respect to their natural monotone norms ∩.∩ p and ∩.∩∞ ,
from Theorems 5.15, 5.16, and 5.17, we have the following result.
(b) Let Q be a bounded subset of the normed space X = c0 . Since {e(1) , e(2) . . .} is
a Schauder basis for c0 , by (5.45) we have
⎫ ⎬
λ(Q) = lim sup max | xk | .
n∈∞ x∞Q k≥n
(c) Let Q be a bounded subset of the normed space X = c. Since {e, e(1) , e(2) , . . .}
is a Schauder basis for c, by (5.45) we have the estimates
1
(Q) ≤ λ(Q) ≤ μ(Q), (5.57)
2
where ⎫ ⎡ ⎡⎬
⎡ ⎡
μ(Q) = lim sup sup ⎡⎡xk − lim xk ⎡⎡ (5.58)
n∈∞ x∞Q x≥n k∈∞
is the so-called regular measure of non-compactness in the space c (cf. Sect. 5.8).
However, Theorems 5.15 and 5.16 cannot be applied to the space ∞ which has
no Schauder basis. Concerning the space ∞ (R) of all bounded real sequences we
have the following result.
Theorem 5.19 (Benavides [17, Proposition 3.5]) Let ∞ be the real normed space of
bounded sequences with sup-norm and Q be a bounded subset of ∞ . Then φ(Q) =
2λ(Q).
Proof We know that φ(Q) ≤ 2λ(Q). Let α > 0 and Q 1 , . . . , Q n be subsets of
∞ (R) such that Q ≤ ⊂i=1 n Q . and diamQ < φ(Q) + α. For any k ∞ N we
i i
put φk,i = inf{xk : (x j ) ∞ Q i }, εk,i = sup{xk : (x j ) ∞ Q i }, ck,i = (φk,i +
εk,i) /2), Bi = B((ck,i )∞
k=1 , (φ(Q) + α)/2) for i = 1, 2, . . . , n. It is easy to prove
that Q i ≤ Bi . Hence λ(Q) ≤ (φ(Q) + α)/2, that is, 2λ(Q) ≤ φ(Q).
This completes the proof.
Proof Let α > 0. Construct a finite [λ(Q) + α]−net E of the set Q. Let x ∞ Q and
y ∞ E be such that ∩ x − y ∩≤ λ(Q) + α. Finally, let β > 0 and r ∞ [0, β]. Then
∩ x − xr ∩≤∩ x − y ∩ + ∩ y − yr ∩ + ∩ yr − xr ∩
≤ 2 ∩ x − y ∩ + ∩ y − yr ∩
⎣ ⎤
≤ 2λ(Q) + 2α + max max ∩ y − yr ∩ .
y∞Q 0≤r ≤β
Hence
⎣ ⎤ ⎣ ⎤
sup max ∩ x − xr ∩ ≤ 2λ(Q) + 2α + max max ∩ y − yr ∩ .
x∞Q 0≤r ≤β y∞Q 0≤r ≤β
Letting β ∈ 0 and taking into account that the finite family E is equicontinuous, we
obtain
⎣ ⎤
lim sup max ∩ x − xr ∩ ≤ 2λ(Q) + 2α,
β∈0 x∞Q 0≤r ≤β
which yields
⎣ ⎤
1
lim sup max ∩ x − xr ∩ ≤ λ(Q). (5.60)
2 β∈0 x∞Q 0≤r ≤β
1
(Rh x)(t) = (max{x(s) : s ∞ [t − h, t + h]} + min{x(s) : s ∞ [t − h, t + h]}),
2
and
⎦t+h
1
(Ph x)(t) = x(s)ds.
2h
t−h
It is easy to see that Ph Rh (Q) is relatively compact in C([a, b]). We claim that it
constitutes a (q2h /2)−net of the set Q, where q2h = supx∞Q max0≤r ≤β ∩ x − xr ∩ .
In fact,
5.5 The Hausdorff Measure of Non-compactness for Some Function Spaces 165
∩ Ph Rh x − x ∩
⎦t+h ⎦t+h
1 1
= max | (Rh x)(s)ds − x(t)ds|
0≤r ≤β 2h 2h
t−h t−h
⎦t+h
1
≤ max | (Rh x)(s) − x(t) | ds. (5.61)
2h a≤t≤b
t−h
If | t − s |≤ h, then we have
where B is a basis of neighborhoods of some point of K and rad x(V ) denotes the
infimum of the radii of all balls in Rm that contain x(V ).
Theorem 5.22 Let L ∞ ([a, b], Rm ) be the space of equivalence classes x of mea-
surable, essentially bounded functions f : [a, b] ∈ Rm , endowed with the
norm ||x|| = ess supt∞[a,b] | f (t)| = inf f ∞x supt∞[a,b] | f (t)|. Then (5.62) holds in
L ∞ ([a, b], Rm ) for an appropriate interpretation of the notations involved, namely,
B stands for an arbitrary maximal filter of measurable sets in [a, b] and x(V ) stands
for ◦ f (Ṽ ), where f runs through x and Ṽ runs through the set of all subsets of V
of full measure.
Theorem 5.23 Let Q be a bounded subset of the space L p ([a, b]) of equivalence
classes x of measurable functions f : [a, b] ∈ R which are p−integrable, endowed
⎢ 1/ p
b
with the norm ∩ x ∩= a | x(t) | p dt . Then
166 5 Measures of Non-compactness
1
μ(Q) ≤ λ(Q) ≤ μ(Q). (5.63)
2
The function μ appearing above is defined by the formula
⎣ ⎤
μ(Q) = lim sup max ||x − x h || ,
α∈0 x∞Q 0≤h≤α
⎦t+h
1
x h (t) = x(s)ds
2h
t−h
(here we put x(t) = 0 outside of the interval [a, b]). It may be shown [18] that the
estimates (5.63) are sharp.
Definition 5.3 Let (X, d) be a metric space and Q a bounded subset of X. Then
the inner Hausdorff measure of non-compactness of the set Q, denoted by λi (Q) is
defined to be the infimum of the set of all reals α > 0 such that Q can be covered by
a finite number of balls of radii < α and centers in Q, that is,
n
λi (Q) = inf α > 0 : Q ≤ B(xi , ri ), xi ∞ Q, ri < α (i = 1, . . . n), n ∞ N .
i=1
Remark 5.6 This “measure of non-compactness” does not have some properties
of the Hausdorff and Kuratowski measures of non-compactness. For example, the
inner Hausdorff
⊃ measure of non-compactness still satisfies λi (B X ) = 1; however
λi (S2 ) = 2. So λi is not invariant under passing to the convex hull.
We have the following properties of λi .
Lemma 5.24 If Q, Q 1 and Q 2 are bounded subsets of the metric space (X, d), then
5.6 Inner Hausdorff Measure of Non-compactness 167
Lemma 5.25 Let Q, Q 1 and Q 2 be bounded subsets of the normed space X. then
(i) λi (Q 1 + Q 2 ) ≤ λi (Q 1 ) + λi (Q 2 ),
(ii) λi (Q + x) = λi (Q) for each x ∞ X,
(iii) λi (δQ) = |δ|λi (Q) for each δ ∞ F.
But in general
λi (Q) → = λi (co(Q)).
In the fixed point theory in normed space (or more generally in locally convex
spaces) the relation φ(Q) = φ(co(Q)) is of great importance. Let us remark that
O. Hadžić [19], among other things, studied the inner Hausdorff measure of non-
compactness in paranormed spaces. She proved under some additional conditions
the inequality λi (co(Q)) ≤ φ[λi (Q)], where φ : [0, ∞) ⊕ ∈ [0, ∞), and, then got
some fixed point theorems for multivalued mappings in general topological vector
spaces.
Comparing the usual and the inner Hausdorff measure of non-compactness one
easily sees that
λ(Q) ≤ λi (Q) ≤ φ(Q).
Example 5.1 Let X = C[0, 1] be the Banach space of all continuous real functions
on [0, 1], equipped with the usual maximum norm. By Theorems 5.8 and 5.12, for
M = B X , say, we have then
1
λ(M) = λi (M) = φ(M) = .
2
168 5 Measures of Non-compactness
Definition 5.4 (Istrǎţescu [5]) Let (X, d) be a complete metric space and Q a
bounded subset of X. Then the Istrǎtescu measure of non-compactness (ε-measure
or lattice measure of non-compactness) of Q, is denoted by ε(Q), and defined by
Theorem 5.26 (Daneš [20]) Let (X, d) be a metric space and Q be a bounded subset
of X. Then
λ(Q) ≤ λi (Q) ≤ ε(Q) ≤ φ(Q) ≤ 2λ(Q).
Remark 5.7 From the well-known Riesz lemma it follows that ε(B X ) = ε(S X ) > 1
in every infinite-dimensional Banach space X . Unfortunately, there is no “universal”
formula for calculating the characteristic ε(B X ), but only formulas or estimates
which require individual arguments in every space. For example, one knows that
ε(B p ) = 21/ p for 1 ≤ p < ∞, ε(B L p ) = 21/ p for 1 ≤ p ≤ 2, and ε(B L p ) =
21−1/ p for 2 ≤ p < ∞. Loosely speaking, one could say that, in contrast to the
measures of non-compactness φ and λ, the measure of non-compactness ε “feels
the geometry” of the underlying space.
||x|| = sup{|x(t)| : t ∞ R+ }.
It turns out that in the space BC(R+ ) the Arzéla-Ascoli criterion fails to work (cf.
[9]) and only some sufficient conditions for relative compactness are known. By these
regards, we are not in a position to provide a formula for the Hausdorff measure of
non-compactness in the space BC(R+ ) or even to give an estimate of that measure.
The above presented reasons initiated an axiomatic approach to the concept of
a measure of non-compactness. Starting from seventies of the past century mathe-
maticians working in non-linear analysis created a few axiomatic definitions of that
concept. Some of those definitions are very general and not convenient in practice.
More precisely, in order to apply such a definition, we are mostly forced to impose
some additional conditions on a measure of non-compactness involved (cf. [6, 21]).
By these reasons, it seems that the axiomatic definition of the concept of a measure
of non-compactness should be not very general and should require satisfying such
conditions which enable the convenience of their use in concrete applications.
170 5 Measures of Non-compactness
The family ker μ described in axiom 1o is called the kernel of the measure of non-
compactness μ. Observe that from axiom 5o , we infer that μ(X ∞ ) ≤ μ(X n ) for
n = 1, 2, . . .. This implies that μ(X ∞ ) = 0. Thus X ∞ belongs to the kernel ker μ
described in axiom 1o . This fact plays a very important role in applications.
Let us notice that deleting some axioms in Definition 5.5, we can formulate the
definition of a measure of non-compactness in the setting of a metric space. We will
not develop this direction of investigations, since in practice, we mostly use measures
of non-compactness in Banach spaces.
Further on, let us observe that the system of axioms accepted in Definition 5.5 is
based on some properties of Kuratowski, Hausdorff, and Istratescu measures of non-
compactness. On the one hand, Definition 5.5 is sufficiently general to be applicable
in several branches of non-linear analysis. On the other hand, this definition admits
several natural realizations and overcomes the difficulties indicated before.
In order to indicate the usefulness of a measure of non-compactness in the sense of
Definition 5.5, let us notice that each measure μ is locally Lipschitzian with respect
to the Hausdorff metric (cf. Sect. 5.1). This fact was proved in [9].
Now we distinguish a few important classes of measures of non-compactness [9].
λ(X ) = dH (X, N E ),
(cf. Theorem 5.11). By the above-listed reasons, we will always compare any regular
measure of non-compactness μ with the Hausdorff measure λ. From this point of
view, we may assert that each regular measure μ is subordered to the Hausdorff
measure λ. Indeed, we have the following result [9].
for any X ∞ M E .
Proof Take an arbitrary number α > 0. Let us cover the set X by a finite family of
balls B(xk , r ), where k = 1, 2, . . . , n and r = λ(X ) + α. Now we have
n
μ(X ) ≤ μ B(xk , r ) = max{μ(B(xk , r ) : k = 1, 2, . . . , n}
k=1
μ2 (X ) = lim sup {sup[|xn − xm | : n, m ≥ k]} , (5.65)
k∈∞ x∞X
but has not maximum property. The kernel ker μ3 is the family of all non-empty and
bounded sets X such that the thickness of the boundle formed by sequences from X
tends to zero at infinity. Obviously, ker μ3 is not a full kernel.
The measures of non-compactness defined by formulas (5.64)–(5.66) will be used
in considerations concerning infinite systems of differential and integral equations
(cf. Chaps. 7 and 8).
Now we are going to describe some measures of non-compactness in the function
space BC(R+ ) discussed previously. Let us briefly recall that BC(R+ ) denotes the
space of all real functions defined, continuous, and bounded on R+ with the standard
supremum norm.
Assume that X is a fixed non-empty and bounded subset of the space BC(R+ ).
Equivalently, this means that X ∞ M BC(R+ ) . Fix numbers α > 0, T > 0 and a
function x ∞ X . Let us define the following quantity
This quantity represents the modulus of continuity of the function x on the interval
[0, T ], while the quantity
is the modulus of continuity of the set X . Since the function α ∈ ξ T (X, α) is non-
decreasing, we infer that there exists a finite limit lim ξ T (X, α). We denote this limit
α∈0
by ξ0T (X ), i.e., we put
ξ0T (X ) = lim ξ T (X, α).
α∈0
ξ0 (X ) = lim ξ0T (X ).
T ∈∞
where X (t) = {x(t) : x ∞ X } and the symbol diamX (t) stands for the diameter of
the set X (t), i.e.,
μb (X ) = ξ0 (X ) + b(X ),
μc (X ) = ξ0 (X ) + c(X ).
It can be shown [23] that the above-defined quantities are measures of non-
compactness in the space BC(R+ ). The measures μa and μb are sublinear and with
maximum property while the measure μc is sublinear but has no maximum property.
The kernel ker μa consists of all bounded sets X such that functions from X are
locally equicontinuous on R+ and tend to zero at infinity with the same rate. Similarly,
the kernel ker μb contains of bounded sets X such that functions from X are locally
equicontinuous on R+ and tend to limits at infinity with the same rate, i.e., uniformly
with respect to the set X . Finally, the kernel ker μc contains of all bounded subsets
X of the space BC(R+ ) such that functions from X are locally equicontinuous on
R+ and the thickness of the boundle formed by graphs of functions from X tends to
zero at infinity.
Let us also pay attention to the fact that no of the considered measures of non-
compactness μa , μb , and μc is full. Obviously, we have that ker μa ≤ ker μb and
ker μa ≤ ker μc but there is no inclusion between ker μb and ker μc .
Now we pay our attention to the algebraic properties of the kernel ker μ of a
measure of non-compactness μ. First of all, let us observe that the family ker μ of
an arbitrary measure (in the sense of Definition 5.5) is convex with respect to the
algebraic operations on sets. Indeed, this property is a consequence of axiom 4o of
Definition 5.5. Moreover, the family ker μ has the following property being important
in some applications (cf. [24]):
X, Y ∞ ker μ ⇒ X ⊂ Y ∞ ker μ.
X, Y ∞ ker μ ⇒ X + Y ∞ ker μ,
X ∞ ker μ, δ ∞ R ⇒ δX ∞ ker μ.
Observe that if X is a member of the family ker μ, then all singletons belonging to
X are elements of the kernel set E μ . Further notice that if μ is a regular measure of
non-compactness in E then E μ = E. But such an equality is not always true. For
example, if we take the measure μ in a Banach space E defined as the norm of a set,
i.e., μ(X ) = ||X || for X ∞ M E then E μ = {θ}. On the other hand for the measure
μ(X ) = diamX we have E μ = E.
In general, the kernel set E μ of a measure of non-compactness μ is a non-empty
and convex subset of E. This set is also closed in view of the earlier-mentioned fact
that μ is locally Lipschitzian with respect to the Hausdorff distance.
If μ is a sublinear measure of non-compactness, then the kernel set E μ is a linear
closed subspace of E. As we announced before, the concept of the kernel set will
play some role in characterization of the flow of solutions of differential equations
in Banach spaces (cf. Chap. 7).
hence we have by (5.69) and Theorem 5.14 that λ(L(B)) ≤ ∩L∩λ . Now we show
that ∩L∩λ ≤ λ(L(B)). Let X ∞ M and {xi }i=1
n be a finite r -net of X. Then X ≤
⊂i=1 B(xi , r ) and obviously we get
n
n
L(X ) ≤ L(B(xi , r )). (5.72)
i=1
∩ L ∩λ ≤∩ L ∩, (5.74)
∩ L̃ ◦ L ∩λ ≤∩ L̃ ∩λ ∩ L ∩λ . (5.76)
extension of Darbo’s fixed point theorem and used it to study the problem of existence
of solutions for a general system of nonlinear integral equations. Here we present
Darbo’s theorem.
We have
μ(Mn+1 ) = μ(Co(T (Mn ))) = μ(T (Mn )) ≤ kμ(Mn ),
Example 5.1 Let E = 2 be the space of all square-summable sequences, and let
T : B E ∈ B E be defined by
#
T (x) = T (x 1 , x2 , . . .) = ( 1− ∩ x ∩2 ,x1 , x2 , . . .).
Exercises
1
φ(M) = .
1 + min x∞M n(x)
Show that φ is (i) set additive, and (ii) invariance under passage to the convex
hull. But φ is not (i) algebraic subadditive, and (ii) regular.
4. Let E be any infinite-dimensional Banach space and A : E ∈ E defined by
A(x) = ||x||x. Show that λ(Sr (E)) = r , and λ(A(Sr (E))) = λ(Sr 2 (E)) = r 2 .
5. On the space E = L p (1 ≤ p < ∞), let us define
1 ⎝1/ p
⎦ ⎧
λ∗ (M) = lim sup | u(t + h) − u(t) | p dt
h∈0 u∞M ⎨
0
1
∩ T ∗ ∩λ = λ(B1 ) = 1 > ⊃ =∩ T ∩λ .
2
180 5 Measures of Non-compactness
∞
∞
'
lim sup X n = Xm .
n∈∞ m=n
n=1
where the symbol B(x, r ) denotes the closed ball centered at x and with radius
r and B(A, ψ) denotes the ball with center at a set A(A →= ∅) and with radius ψ,
i.e.,
B(A, ψ) = B(z, ψ).
z∞A
for all x, y ∞ X . Prove the following Nadler fixed point theorem: If (X, d) is
complete then every multivalued contraction F : X ∈ McX has at least one
fixed point in X , i.e., there exists x ∞ X such that x ∞ F(x).
16. Let G : E ∈ E be a contraction (in the classical sense) in the Banach space
(E, || · ||), i.e., there exists a constant L ∞ [0, 1) such that
G r (x) = B(G(x), r ),
where the symbol B(G(x), r ) was defined in Exercise 14. Show that G r has a
fixed point in E.
Hint. Show that G r is a multivalued contraction with the constant L (use Exer-
cises 13 and 14).
17. Let μ be an arbitrary measure of non-compactness in a Banach space (E, || · ||).
Show that for any t ∞ [0, 1) and for an arbitrary X ∞ M X the following
inequality holds
18. Let X be a subset of a Banach space (E, || · ||) such that ||X || < 1. Show that
μ(t, X ) ≤ tμ(X )
holds.
21. Let μ be a measure of non-compactness in a Banach space E and let x0 ∞ E μ
(∼ {x0 } ∞ ker μ). Show that
μ(x0 + t X ) ≤ tμ(x0 + X ),
n
for arbitrary real numbers ti ≥ 0 (i = 1, 2, . . . , n) such that ti ≤ 1.
i=1
23. Let (E k , || · ||k ) be a sequence of Banach spaces. For a fixed p ∞ [1, ∞) denote
by l p (E k ) the space of all sequence x = (xk ) such that xk ∞ E k for k = 1, 2, ....
Denote by Pn the projector of the space l p (E k ) on E n . Finally, let us put
λ(X ) ≤ μ2 (X ) ≤ 2λ(X )
for each set X ∞ Mc . Show also that the measure μ2 is equivalent to the measure
of non-compactness μ defined in the space c by formula (5.58).
References 183
References
1. Kuratowski, K.: Sur les espaces complets. Fund. Math. 15, 301–309 (1930)
2. Darbo, G.: Punti uniti in transformazioni a condominio non compatto. Rend. Sem. Math.
Univ. Padova 24, 84–92 (1955)
3. Goldenštein, L.S., Gohberg, I.T., Markus, A.S.: Investigations of some properties of bounded
linear operators with their q-norms, Učen. Zap. Kishinevsk. Univ. 29, 29–36 (1957)
4. Goldenštein, L.S. Markus, A.S.: On a measure of noncompactness of bounded sets and linear
operators. In: Studies in Algebra and Mathematical Analysis, Kishinev, pp. 45–54 (1965)
5. Istrăţescu, V.: On a measure of noncompactness, Bull. Math. Soc. Sci. Math. R.S. Roumanie
(N.S) 16(1972), 195–197 (1972)
6. Akhmerov, R.R., Kamenskij, M.I., Potapov, A.S., Rodkina, A.E., Sadovskii, B.N.: Measures
of Noncompactness and Condensing Operators, Operator Theory: Advances and Applications,
vol. 55. Birkhäuser, Basel (1992)
7. Appell, J.: Measures of noncompactness, condensing operators and fixed points: an
application-oriented survey. Fixed Point Theor. 6(2), 157–229 (2005)
8. Ayerbe Toledano, J.M., Domínguez Benavide, T., López Azedo, G.: Measures of Noncom-
pactness in Metric Fixed. Birkhäuser, Basel (1997)
9. Banaś, J., Goebel, K.: Measures of Noncompactness in Banach Spaces. Lecture Notes in Pure
and Applied Mathematics. Marcel Dekker, New York (1980)
10. Malkowsky, E.: Modern functional analysis in the theory of sequence spaces and matrix
transformations. Jordan J. Math. Stat. 1(1), 1–29 (2008)
11. Malkowsky, E.: Compact matrix operators between some B K spaces. In: Mursaleen, M. (ed.)
Modern Methods of Analysis and Its Applications, pp. 86–120. Anamaya Publication, New
Delhi (2010)
12. Malkowsky, E., Rakočević, V.: An introduction into the theory of sequence spaces and mea-
sures of noncompactness. Jordan J. Math. Stat. 1(1), 1–29 (2008)
13. Furi, M., Vignoli, A.: On a property of the unit sphere in a linear normed space. Bull. Pol.
Acad. Sci. Math. 18, 333–334 (1970)
14. Nussbaum, R.G.: The radius of the essential spectrum. Duke Math. J. 38, 473–478 (1970)
15. Benavides, T.D., Ayerbe, J.M.: Set-contractions and ball contractions in L p -spaces. J. Math.
Anal. Appl. 159, 500–506 (1991)
16. Webb, J.R.L., Zhao, W.: On connections between set and ball measures of noncompactness.
Bull. London Math. Soc. 22, 471–477 (1990)
17. Benavides, T.D.: Set-contractions and ball contractions in some class of spaces. J. Math. Anal.
Appl. 136, 131–140 (1988)
18. Dronka, J.: Note on the Hausdorff measure of noncompactness in L p -spaces. Bull. Pol. Acad.
Sci. Math. 41, 39–41 (1993)
19. Hadžić, O.: Some properties of measures of noncompactness in paranormed spaces. Proc.
Am. Math. Soc. 102(4), 843–849 (1988)
20. Daneš, J.: On the Istrǎţesku’s measure of noncompactness. Bull. Math. Soc. Sci. Math. RSR
16(4), 403–406 (1972)
21. Daneš, J.: On densifying and related mappings and their application in nonlinear functional
analysis. Akademic, Berlin (1974)
22. Banaś, J., Martinon, A.: Measures of noncompactness in Banach sequence spaces. Math.
Slovaca 42, 497–503 (1992)
23. Banaś, J.: Measures of noncompactness in the space of continuous tempered functions.
Demonstratio Math. 14, 127–133 (1981)
24. Banaś, J., Martinon, A.: Some properties of the Hausdorff distance in metric spaces. Bull.
Aust. Math. Soc. 42, 511–516 (1990)
25. Banaś, J., Hajnosz, A., Wȩdrychowicz, S.: Some generalization of Szufla’s theorem for ordi-
nary differential equations in Banach space. Bull. Acad. Pol. Sci., Sér. Sci. Math. 29, 459–464
(1981)
184 5 Measures of Non-compactness
26. Banaś, J., Hajnosz, A., Wȩdrychowicz, S.: On the equation x = f (t, x) in Banach spaces.
Comment. Math. Univ. Carol. 23, 233–247 (1982)
27. Banaś, J., Hajnosz, A., Wȩdrychowicz, S.: On existence and local characterization of solutions
of ordinary differential equations in Banach spaces, Diff. Eqs. Appl., Proc. Second Conf.,
Rousse, 55–58 (1982)
28. Sadovskii, B.N.: On a fixed point principle [in Russian]. Funkts. Analiz Prilozh. 1(2), 74–76
(1967)
29. Aghajani, A., Allahyari, R., Mursaleen, M.: A generalization of Darbo’s theorem with appli-
cation to the solvability of systems of integral equations. J. Comput. Appl. Math. 260, 68–77
(2014)
Chapter 6
Application to Compact Matrix Operators
The most effective way in the characterization of compact operators between the
Banach spaces is applying the Hausdorff measure of non-compactness. In this
chapter, we present some identities or estimates for the operator norms and the
Hausdorff measures of non-compactness of certain operators given by infinite matri-
ces that map an arbitrary BK space into the sequence spaces c0 , c, ∞ , and 1 , and
into the matrix domains of triangles in these spaces. Many linear compact operators
may be represented as matrix operators in sequence spaces or integral operators in
function spaces. For example, see [1–4, 6–15].
The general method for the characterization of compact matrix operators between
BK spaces X and Y consists of the following steps and can be applied whenever Y
has a Schauder basis:
Step 1: Determine the β-duals of X and Y .
Step 2: Characterize the class (X, Y ).
Step 3: Compute the operator norm (or find an estimate for the operator norm) of
the matrix operators from X into Y .
Step 4: Use the Hausdorff measure of non-compactness to find necessary and suffi-
cient conditions on the entries of the matrix to be a compact operator between
X and Y .
This method cannot be applied, in general, in the case when the final space Y has no
Schauder basis; then, it only yields sufficient conditions [16, Example 3.25, p. 195].
A different method for that case can be found in [17, p. 85] where compact matrix
operators in ( p , ∞ )(1 < p ∈ ∞), (c0 , ∞ ), and (c, ∞ ) were characterized. The
method given there was also applied in [16, Theorem 4.7]. We will not deal with this
approach here.
In this section, we give the characterization for the classes C(1 , 1 ) and C(c, c).
In continuation of Theorem 2.41 we prove the following result concerning the com-
pact operators for B(1 , 1 ).
Theorem 6.1 (a) Let L ∈ B(1 , 1 ), and A denote the infinite matrix such that
L(x) = Ax for all x ∈ 1 . Then, we have
∞
→ L →χ = lim sup | ank | . (6.1)
r ≤∞ k n=r
(b) Let L ∈ B(∞ , ∞ ), and A denote the infinite matrix such that L(x) = Ax for
all x ∈ ∞ . Then, we have
∞
→ L →χ = lim sup | ank | . (6.1∩ )
r ≤∞ n
k=r
Proof (a) By Theorem 2.33 (b), every L ∈ B(X, Y ) can be represented by a matrix
A ∈ (X, Y ). Writing S = S1 , by Theorems 5.18, 5.28 and 2.41, it follows that
∞ ∞
→ L →χ = χ (L(S)) = lim sup | ank xk | . (6.2)
r ≤∞ x∈S
n=r k=0
∞
χ (L(e(k) )) = lim sup | ank |∈ χ (L(S)),
r ≤∞ k
n=r
(b) This can be proved similarly by using Theorem 2.40 (a) instead of Theorem 2.41.
This completes the proof.
As an immediate consequence of Theorem 6.1, we have
Corollary 6.2 (a) Let L ∈ B(1 , 1 ), and A denote the infinite matrix such that
L(x) = Ax for all x ∈ 1 . Then, we have L ∈ C(1 , 1 ) if and only if
∞
lim sup | ank | = 0.
r ≤∞ k n=r
(b) Let L ∈ B(∞ , ∞ ), and A denote the infinite matrix such that L(x) = Ax for
all x ∈ ∞ . Then, we have L ∈ C(∞ , ∞ ) if and only if
∞
lim sup | ank | = 0.
r ≤∞ n
k=r
Remark 6.1 It is not true that every operator L ∈ B(1 , 1 ) is compact (cf. Remark
6.10 [18] and Remark 19 [19]). For example, let I ∈ B(1 , 1 ) be the identity
operator on 1 . Then, it is obvious that I cannot be compact. To see this, we know
that χ (B1 ) = 1. Hence, it follows from Theorems 5.18 and 5.28 that →I →χ =
χ (I (B1 )) = χ (B1 ) = 1.
The following theorem is in continuation of Theorems 2.42 and 2.43.
∞
∞
∞
lim An (x − ξ e) = αk (xk − ξ ) = αk xk − ξ αk . (6.6)
n≤∞
k=0 k=0 k=0
Now, we apply (5.49) from Theorem 5.16. First we note that → L →χ = χ (L(Bc ))
by Theorem 5.28. Since every sequence z = (z k )nk=0 ∈ c has the representation
z = ζe+ ∞ k=0 (z
k − ζ )e
(k) with ζ = lim
k≤∞ z k , we define the projector Pr : c ≤ c
by Pr (z) = ζ e + k=0 (z k − ζ )e(k) , and it follows that the sequence z̃ = (I − Pr )(z)
r
lim → I − Pr →= 2. (6.8)
r ≤∞
∗
∞
∞ see that f n ∈ c by Theorem 2.2, and → f n → = | bn − α̃ +
and k=0 αk | +
k=0 | ank − αk | . Therefore, we have
∞ ∞
sup →(I − Pr )(L(x))→ = sup → f n → = sup |bn − α̃ + αk | + |ank − αk |
x∈Bc n≥r +1 n≥r
k=0 k=0
and (6.4) now follows from (6.8), Theorems 5.17 and 5.28.
This completes the proof.
Remark 6.2 It is obvious from the Corollary 6.5 that if A is a regular matrix then
L A cannot be compact.
If A is a conservative matrix and L A is compact then A is
conull, that is, α̃ − ∞
k=0 αk = 0.
Corollary 6.6 [5] Let L ∈ B(c, c) be regular, that is, limn≤∞ (L(x)) = limk≤∞ xk
for all x ∈ c. Then, L A is compact if and only if
∞
lim | bn − 1 | + | ank | = 0. (6.11)
n≤∞
k=0
and ∞
lim (L(e))n = 1 = lim bn + ank = α̃. (6.13)
n≤∞ n≤∞
k=0
Conversely if L(x) = b limk≤∞ xk + A(x) and (6.12) and (6.13) are ∞satisfied,
then
∞ it follows from (6.7) that lim n≤∞ (L(x)) n = lim k≤∞ x k ( α̃ − k=0 αk ) +
k=0 αk x k = lim k x k for all x ∈ c, and L is regular.
This completes the proof.
In the present section, we establish some identities or estimates for the Hausdorff
measures of non-compactness of certain operators given by infinite matrices that
map an arbitrary BK space (with AK, sometimes) into the spaces c0 , c, ∞ , and 1 .
Further, by using the Hausdorff measure of non-compactness, we apply our results
to characterize some classes of compact operators on the BK spaces.
We may begin with the following lemma:
Lemma 6.7 Let X ≥ φ be a BK space with AK or X = ∞ . If A ∈ (X, c), then the
following hold:
αk = lim ank exists for every k ∈ N, (6.14)
n≤∞
190 6 Application to Compact Matrix Operators
α = (αk ) ∈ X β , (6.15)
∞
lim An (x) = αk xk for all x = (xk ) ∈ X. (6.17)
n≤∞
k=0
Proof Let us begin with the first case when X ≥ φ is a BK space with AK. We write
→ · → = → · → X , for short. Since A ∈ (X, c), we obtain by Theorem 2.37 that
Further, since X has AK, we have e(k) ∈ X and hence Ae(k) ∈ c for all k ∈ N.
Consequently, the limits αk in (6.14) exist for all k ∈ N.
Now, let x ∈ X be given. Since X has AK, there is a positive constant K such that
→x [m] → ∈ K →x→ for all m ∈ N. We thus derive that
m
ank xk = An (x [m] ) ∈ Ax [m] = L A (x [m] ) ∈ K →L A →→x→
∞ ∞
k=0
m m
αk k
x = lim ank xk ∈ K →L A →→x→; (m ∈ N).
n≤∞
k=0 k=0
This implies that αx = (αk xk ) ∈ bs, and since x ∈ X was arbitrary, we deduce that
α ∈ X γ . But X has AK which yields that X γ = X β by Theorem 2.13 and hence
(6.15) holds. Moreover, since X ≥ φ is a BK space, (6.15) implies →α→∗X < ∞ by
Theorem 2.14. Therefore, we get (6.16) from (6.18) by using the definition of →a→∗X .
Now, define the matrix B = (bnk ) by bnk = ank − αk for all n, k ∈ N. Then, it is
obvious that Bn ∈ X β for all n ∈ N. Also, it follows by (6.16) that
that is, Be(k) ∈ c0 for all k ∈ N. This leads us to the consequence that B ∈ (X, c0 )
by [16, Theorem 1.23 (c)]. Hence, limn≤∞ Bn (x) = 0 for all x ∈ X which yields
(6.17).
6.2 Compact Matrix Operators on Some BK Spaces 191
Now, we turn to the second case X = ∞ , i.e., A ∈ (∞ , c). Hence, by Schur’s
Theorem (6.14) holds [20, Proposition 2.2] and
∞
sup |ank | < ∞, (6.19)
n
k=0
∞
lim |ank − αk | = 0. (6.20)
n≤∞
k=0
k
∞
|α j | ∈ sup |an j | < ∞.
n
j=0 j=0
β
This implies that α = (αk ) ∈ 1 . Thus, α ∈ ∞ , and so (6.15) holds.
Moreover, it is immediate by (6.20) that
∞
sup →An − α→1 = sup |ank − αk | < ∞
n n
k=0
∞
∞
∞
An (x) − αk xk ∈ |ank − αk ||xk | ∈ →x→∞ |ank − αk | .
k=0 k=0 k=0
Now, by combining Lemma 6.7 and Theorems 2.33, 2.34, and 2.37, we deduce
the following result on the Hausdorff measure of non-compactness.
1
· lim sup →An − α→∗X ∈ →L A →χ ∈ lim sup →An − α→∗X , (6.22)
2 r ≤∞ n≥r r ≤∞ n≥r
192 6 Application to Compact Matrix Operators
Proof Let us remark that the limits in (6.21), (6.22), and (6.23) exist by Lemma
6.7 and Theorem 2.37, for example, if X is as in part (b) and A ∈ (X, c), then
∞
the sequence supn≥r →An − α→∗X r =0 of non-negative reals is non-increasing and
bounded by Lemma 6.7.
We write S = S X , for short. Then, we have by Theorems 5.28 and 2.33 (a) that
→L A →χ = χ (AS). (6.24)
1
· μ(A) ∈ →L A →χ ∈ μ(A), (6.27)
2
where
6.2 Compact Matrix Operators on Some BK Spaces 193
μ(A) = lim sup sup →(I − Pr )(Ax)→∞ .
r ≤∞ x∈S
∞
→(I − Pr )(Ax)→∞ = sup An (x) − αk xk
n≥r
k=0
∞
= sup (ank − αk )xk
n≥r
k=0
Hence, we get (6.22) from (6.27), since the limit in (6.22) exists.
Finally, to prove (c) we define Pr : ∞ ≤ ∞ (r ∈ N) as in the proof of part (a)
for all x = (xk ) ∈ ∞ . Then, it is clear that
Example 6.1 Let X ≥ φ be a BK space and choose a fixed m ∈ N such that xm∩ ⇒ = 0
for some x ∩ = (xk∩ ) ∈ S X . Now, we define the matrix A = (ank ) by anm = 1 and
ank = 0 for all k ⇒ = m (n ∈ N). Then, we have Ax = xm e for all x = (xk ) ∈ w,
hence A ∈ (w, ∞ ) ⊂ (X, ∞ ). Also, since L A is of finite rank, L A is compact. On
the other hand, we have An = e(m) and hence → An →∗X = supx∈S X |xm | for all n ∈ N.
This implies that
lim sup →An →∗X = sup |xm | ≥ |xm∩ | > 0.
r ≤∞ n>r x∈S X
Remark 6.3 Now, it is obvious that Theorem 6.8 and Corollary 6.9 have several
consequences, some of them are known results. For instance, consider the particular
case X = p (1 ∈ p < ∞). Then, we know by Theorems 2.33 (b) that every operator
L in B( p , c0 ), B( p , c), or B( p , ∞ ) is given by an infinite matrix A belonging to
the respective one of the classes ( p , c0 ), ( p , c), or ( p , ∞ ). Hence, by Theorem
6.8, we get an identity or estimate for →L→χ . Further, by means of Corollary 6.9,
we deduce the characterization of the classes C( p , c0 ) and C( p , c) of compact
operators, and the sufficient condition for an operator L ∈ B( p , ∞ ) to be compact.
Moreover, we may note that if X = p (1 ∈ p < ∞) in Theorem 6.8 and Corollary
6.9, then → · →∗ p is replaced by → · →q , where q is the conjugate of p.
On the other hand, let 1 < p < ∞. Then, we write bv p for the space of all
sequences of p-bounded variation, that is,
⎪
bv p = x = (xk ) ∈ w: (xk − xk−1 ) ∈ p ; (1 < p < ∞).
6.2 Compact Matrix Operators on Some BK Spaces 195
It is known that bv p is a BK space with its natural norm [21]. Further, we have for
every a = (ak ) ∈ (bv p )β that
∞
∞ q 1/q
→a→∗bv p = aj , (6.29)
k=0 j=k
Then, we have
(a) If A ∈ (bv p , c0 ), then
(r )
→L A →χ = lim →A→(bv p , ∞ )
r ≤∞
and
(r )
L A is compact if and only if lim → A→(bv p , ∞ ) = 0.
r ≤∞
(r )
0 ∈ →L A →χ ∈ lim →A→(bv p , ∞ )
r ≤∞
and
(r )
L A is compact if lim → A→(bv p , ∞ ) = 0.
r ≤∞
We may note that part (b) of Theorem 6.8 cannot be applicable for X = bv p , since
bv p does not have AK.
Now, let us recall that the upper limit (limit superior) of a bounded real sequence
x = (xn ) can be defined by
lim sup xn = lim sup xn .
n≤∞ r ≤∞ n≥r
With the above notations, Theorem 6.8 and Corollary 6.9 can equivalently be refor-
mulated as follows:
and
L A is compact if and only if lim →An →∗X = 0.
n≤∞
1
· lim sup → An − α→∗X ∈ →L A →χ ∈ lim sup →An − α→∗X
2 n≤∞ n≤∞
and
L A is compact if and only if lim →An − α→∗X = 0,
n≤∞
and
L A is compact if lim →An →∗X = 0. (6.30)
n≤∞
As we have seen in Example 6.1, the converse implication in (6.30) does not hold,
in general.
Remark 6.4 The special cases of Theorem 6.8 when A ∈ (c0 , c0 ), A ∈ (c, c0 ), or
A ∈ (c0 , c) are Theorem 6.3 and its corollaries.
We also have the following consequence of Theorem 6.11:
Corollary 6.12 We have (∞ , c0 ) ⊂ C(∞ , c0 ) and (∞ , c) ⊂ C(∞ , c), that is, for
every matrix A ∈ (∞ , c0 ) or A ∈ (∞ , c), the operator L A is compact.
∞
lim |ank | = 0,
n≤∞
k=0
that is, limn≤∞ →An →1 = 0. This implies that limn≤∞ → An →∗∞ = 0, since An ∈
β
∞ = 1 for all n ∈ N. Hence, we deduce from Theorem 6.11 (a) that L A is compact.
6.2 Compact Matrix Operators on Some BK Spaces 197
Then, we have
∞
sup xn ∈ |xn | ∈ 4 · sup xn
N ∈Fr n∈N n=r +1 N ∈Fr n∈N
Proof Let r ∈ N be given. The first inequality on the left is trivial, since
∞
xn ∈ |xn | ∈ |xn | < ∞
n∈N n∈N n=r +1
for all N ∈ Fr . To prove the other inequality, we have for every m > r that [16,
Lemma 3.9]
m
|xn | ∈ 4 · max xn
N ⊂{r +1,...,m}
n=r +1 n∈N
∈ 4 · sup xn < ∞.
N ∈Fr n∈N
and ∗
L A is compact if and only if lim sup An = 0.
r ≤∞ N ∈Fr X
n∈N
(r ) ∞
Since F ≥ F0 ≥ F1 ≥ · · · , the sequence → A→(X,1 ) r =0 of non-negative reals is
non-increasing and bounded by Theorem 2.35. Therefore, the limit in (6.31) exists.
Now, let S = S X . Then, we have by Theorem 2.33 (a) that L A (S) = AS ∈ M1 .
Thus,
→L A →χ = χ (AS) = lim sup →(I − Pr )(Ax)→1 ,
r ≤∞ x∈S
∞
→L A →χ = lim sup | An (x)| . (6.32)
r ≤∞ x∈S n=r +1
∞
sup An (x) ∈ |An (x)| ∈ 4 · sup An (x) (6.33)
N ∈Fr n∈N n=r +1 N ∈Fr n∈N
∗ ∞
An = sup ank xk = sup An (x)
X x∈S k=0 n∈N x∈S n∈N
n∈N
∗
∞ ∗
sup An ∈ sup |An (x)| ∈ 4 · sup An (6.34)
N ∈Fr X x∈S N ∈Fr X
n∈N n=r +1 n∈N
Finally, it is obvious that the second part of the theorem is immediate by (6.31).
This concludes the proof.
Remark 6.5 Now, it is clear that Theorem 6.14 has many consequences with any
particular case of the space X , for example let 1 ∈ p < ∞. Then, it follows
from Theorem 2.33 (b) that every operator L ∈ B( p , 1 ) is given by an infinite
matrix A ∈ ( p , 1 ). Therefore, by means of Theorem 6.14, we get estimates for the
Hausdorff measures of non-compactness of operators in B( p , 1 ) and deduce the
necessary and sufficient condition for an operator L ∈ B( p , 1 ) to be compact.
On the other hand, if L ∈ B(1 , 1 ) is given by the matrix A ∈ (1 , 1 ), then we
know that
∞
→L→χ = lim sup |ank | (6.35)
r ≤∞ k n=r +1
and
∞
L is compact if and only if lim sup |ank | = 0,
r ≤∞ k n=r +1
which are compatible with Theorem 6.14 by Lemma 6.13, but (6.35) is more exact.
Remark 6.6 Finally, by using 6.29 in Theorem 6.14 for X = bv p (1 < p < ∞), we
obtain Theorem 4.2 and Corollary 4.2 of [21].
In this final section, we apply our previous results to establish some identities or
estimates for the operator norms and the Hausdorff measures of non-compactness
of certain operators given by infinite matrices that map an arbitrary BK space (with
AK, sometimes) into the matrix domains of triangles in the spaces c0 , c, ∞ and
1 . Further, we deduce the necessary and sufficient (or only sufficient) conditions
for such operators to be compact. Moreover, some important special cases are also
deduced.
Throughout, we assume that A = (ank ) is an infinite matrix, T = (tnk ) is a
triangle and we define the matrix B = (bnk ) by
n
bnk = tnm amk ; (n, k ∈ N), (6.36)
m=0
n
n ∞
Bn = tnm Am = tnm amk ; (n ∈ N). (6.37)
k=0
m=0 m=0
200 6 Application to Compact Matrix Operators
Now, by using (6.36) and (6.37), we have the following result on the operator norms.
n
∗
→L A → = →A→(X,(∞ )T ) = sup tnm Am < ∞.
n X
m=0
→ A→(X,(1 )T ) ∈ →L A → ∈ 4 · →A→(X,(1 )T ) ,
where
n ∗
→A→(X,(1 )T ) = sup tnm Am < ∞.
N ∈F X
n∈N m=0
Proof This is immediate by combining Theorem 2.49 (b) with Theorems 2.37 and
6.14.
With the notations of (6.36) and (6.37), we define the sequence a = (ak )∞
k=0 by
n
ak = lim tnm amk ; (k ∈ N) (6.38)
n≤∞
m=0
provided the limits in (6.38) exist for all k ∈ N which is the case whenever A ∈
(X, cT ) by Theorem 2.49 (a) and Lemma 6.7, where X ≥ φ is a BK space with AK
or X = ∞ .
Now, we have the following result on the Hausdorff measure of non-compactness.
n
∗
→L A →χ = lim sup tnm Am
n≤∞ X
m=0
and
n
∗
L A is compact if and only if lim tnm Am = 0.
n≤∞ X
m=0
1
n
∗
n
∗
· lim sup tnm Am − a ∈ →L A →χ ∈ lim sup tnm Am − a
2 n≤∞ X n≤∞ X
m=0 m=0
6.3 Applications to Some Matrix Domains 201
and
n
∗
L A is compact if and only if lim tnm Am − a = 0.
n≤∞ X
m=0
n
∗
0 ∈ →L A →χ ∈ lim sup tnm Am
n≤∞ X
m=0
and
n
∗
L A is compact if lim tnm Am = 0. (6.39)
n≤∞ X
m=0
Proof This result follows from Theorem 6.11 by using Theorem 2.49 (a).
As we have seen in Example 6.1, it can be easily seen that the equivalence in
(6.39) does not hold.
On the other hand, by means of Theorem 2.49 (a), the following result is immediate
by Corollary 6.12.
Corollary 6.17 Let T be a triangle. If either A ∈ (∞ , (c0 )T ) or A ∈ (∞ , cT ),
then L A is compact.
Theorem 6.18 Let X ≥ φ be a BK space and T a triangle. If A ∈ (X, (1 )T ), then
(r ) (r )
lim →A→(X,(1 ) ∈ →L A →χ ∈ 4 · lim →A→(X,(1 )
r ≤∞ T) r ≤∞ T)
and
(r )
L A is compact if and only if lim → A→(X,(1 ) = 0,
r ≤∞ T)
where
(r )
n ∗
→A→(X,(1 ) = sup tnm Am ; (r ∈ N).
T) X
N ∈Fr n∈N m=0
Proof This is obtained from Theorem 6.14 by using Theorem 2.49 (a).
It is worth mentioning that Theorems 6.15, 6.16, and 6.18 have several conse-
quences with any particular case of the triangle T or any particular BK space X
[with AK in part (b) of Theorem 6.16]. For instance, we have the following results:
Corollary 6.19 Let X ≥ φ be a BK space and A an infinite matrix. If A is in any
of the classes (X, cs 0 ), (X, cs) or (X, bs), then
(n)
→L A → = sup →A→(X,bs) < ∞,
n
where
202 6 Application to Compact Matrix Operators
(n)
n
∗
→A→(X,bs) = Am ; (n ∈ N).
X
m=0
Furthermore, we have
(a) If A ∈ (X, cs 0 ), then
and
(n)
L A is compact if and only if lim →A→(X,bs) = 0.
n≤∞
(n)
0 ∈ →L A →χ ∈ lim sup →A→(X,bs)
n≤∞
and
(n)
L A is compact if lim →A→(X,bs) = 0.
n≤∞
1
n
∗
n
∗
· lim sup Am − a ∈ →L A →χ ∈ lim sup Am − a
2 n≤∞ X n≤∞ X
m=0 m=0
and
n
∗
L A is compact if and only if lim Am − a = 0,
n≤∞ X
m=0
n
where a = (ak ) with ak = limn≤∞ ( m=0 amk ) for all k ∈ N.
→A→(X,bv) ∈ →L A → ∈ 4 · →A→(X,bv) ,
where ∗
→A→(X,bv) = sup (An − An−1 ) < ∞.
N ∈F X
n∈N
(r ) (r )
lim →A→(X,bv) ∈ →L A →χ ∈ 4 · lim →A→(X,bv)
r ≤∞ r ≤∞
6.3 Applications to Some Matrix Domains 203
and
(r )
L A is compact if and only if lim →A→(X,bv) = 0,
r ≤∞
where ∗
(r )
→A→(X,bv) = sup (An − An−1 ) ; (r ∈ N).
N ∈Fr X
n∈N
Obviously, there are many special cases of Corollaries 6.19 and 6.20. For example,
by using (6.30) in Corollary 6.19, we have
Corollary 6.21 Let A be an infinite matrix, 1 < p < ∞ and q = p/( p − 1). If A
is in any of the classes (bv p , cs0 ), (bv p , cs) or (bv p , bs), then
→L A → = sup →A→(n)
(bv p ,bs) < ∞,
n
where
∞
∞
n q 1/q
→A→(n)
(bv p ,bs) = am j ; (n ∈ N).
k=0 j=k m=0
Furthermore, we have
(a) If A ∈ (bv p , cs 0 ), then
and
(n)
L A is compact if and only if lim →A→(bv p ,bs) = 0.
n≤∞
(n)
0 ∈ →L A →χ ∈ lim sup →A→(bv p ,bs)
n≤∞
and
L A is compact if lim →A→(n)
(bv p ,bs) = 0.
n≤∞
Remark 6.7 A similar result concerning the class (bv p , bv) can be obtained from
Corollary 6.20 by using (6.30), where 1 < p < ∞ [21, Theorems 3.2 (b), 4.3;
Corollary 4.3].
Now, for brevity, we shall confine ourselves to apply our main results to the spaces
(c) p
sα0 , sα , sα , and α of de Malafosse [6, 22] which are defined as follows:
∞
Let α = (αk )k=0 be a sequence of positive reals and define the diagonal matrix
D1/α by dnn = 1/αn for all n ∈ N. Then, the spaces sα0 , sα(c) , sα , and α are defined
p
204 6 Application to Compact Matrix Operators
1 a ∞
nk
An = ; (n ∈ N) (6.40)
βn βn k=0
1 ∗ 1
→L A → = sup An = sup →An →∗X < ∞.
n βn X n βn
1 ∗ 1 ∗
sup An ∈ →L A → ∈ 4 · sup An .
N ∈F n∈N βn X N ∈F n∈N βn X
1
→L A →χ = lim sup → An →∗X
n≤∞ βn
and 1
L A is compact if and only if lim →An →∗X = 0.
n≤∞ βn
(c)
(b) If X has AK and A ∈ (X, sβ ), then
1 1 ∗ 1 ∗
· lim sup An − γ ∈ →L A →χ ∈ lim sup An − γ
2 n≤∞ βn X n≤∞ βn X
6.3 Applications to Some Matrix Domains 205
and
1 ∗
L A is compact if and only if lim An − γ = 0.
n≤∞ βn X
and 1
L A is compact if lim → An →∗X = 0.
n≤∞ βn
1 ∗ 1 ∗
lim sup An ∈ →L A →χ ∈ 4 · lim sup An
r ≤∞ N ∈Fr βn X r ≤∞ N ∈Fr βn X
n∈N n∈N
and
1 ∗
L A is compact if and only if lim sup An = 0.
r ≤∞ N ∈Fr βn X
n∈N
Remark 6.8 Many new results can be obtained from Corollaries 6.22 and 6.23 with
any particular case of the space X , e.g., we may use (6.30) for X = bv p (1 < p < ∞)
in the above corollaries [except part (b) of Corollary 6.23].
(c) p
Now, let us turn to the β-duals of the spaces sα0 , sα , sα , and α . Then, it can be
β β β p β
easily seen that (sα0 ) = (sα(c) ) = (sα )β = 11/α , (1α ) = s1/α , and (α ) = 1/α ,
q
where 1 < p < ∞, q = p/( p − 1), and 1/α = (1/αk )∞ k=0 . Furthermore, if a = (ak )
p
is in the β-dual of any of the spaces sα , 1α or α , then we have
∞
→a→∗s 0 = →a→∗(c) = →a→∗sα = αk |ak |, (6.42)
α sα
k=0
and
∞ 1/q
→a→∗ p = |αk ak |q , (6.44)
α
k=0
respectively, where 1 < p < ∞ and q = p/( p − 1). Therefore, by using these
notations in Theorems 6.11, 6.14, 6.15, 6.16, 6.18 and Corollaries 6.19 and 6.20, we
get some applications as special cases of these results when X is any of the spaces
206 6 Application to Compact Matrix Operators
(c) p (c)
sα0 , sα , sα , 1α or α (except the cases X = sα and X = sα when X has AK). For
instance, we have the following consequence of Corollaries 6.19 and 6.20.
Corollary 6.24 Let 1 < p < ∞ and q = p/( p − 1). Then, we have
p p
(a) If L ∈ B(α , cs0 ) is given by the matrix A ∈ (α , cs0 ), then
p (n)
and L ∈ C(α , cs0 ) if and only if limn≤∞ →A→( p ,bs) = 0, where
α
∞
n q 1/q
(n)
→A→( p ,bs) = αk amk ; (n ∈ N).
α
k=0 m=0
p p
(b) If L ∈ B(α , bs) is given by the matrix A ∈ (α , bs), then
p (n)
and L ∈ C(α , bs) if limn≤∞ → A→( p ,bs) = 0.
α
p p
(c) If L ∈ B(α , cs) is given by the matrix A ∈ (α , cs), then
1
· lim sup →A→(n)
p
(α ,cs)
∈ →L→χ ∈ lim sup →A→(n)
p
(α ,cs)
2 n≤∞ n≤∞
p (n)
and L ∈ C(α , cs) if and only if limn≤∞ →A→( p ,cs) = 0, where
α
∞
n q 1/q
(n)
→ A→( p ,cs) = αk amk − ak ; (n ∈ N)
α
k=0 m=0
and ak = limn≤∞ ( nm=0 amk ) for all k ∈ N.
p p
(d) If L ∈ B(α , bv) is given by the matrix A ∈ (α , bv), then
(r ) (r )
lim → A→( p ,bv) ∈ →L→χ ∈ 4 · lim →A→( p ,bv)
r ≤∞ α r ≤∞ α
p (r )
and L ∈ C(α , bv) if and only if limr ≤∞ →A→( p ,bv) = 0, where
α
∞
q 1/q
→A→(r )
p
( ,bv)
= sup αk (ank − an−1,k ) ; (r ∈ N).
α
N ∈Fr k=0 n∈N
p
Remark 6.9 The conclusions of Corollary 6.24 still hold for sα0 instead of α with
q = 1.
6.3 Applications to Some Matrix Domains 207
Similarly, several results and consequences can be obtained from Corollaries 6.22
and 6.23 by using (6.42), (6.43), and (6.44). The most of those results can be found
(c)
in [6], for example, if A ∈ (sα0 , sβ ), then we have by Corollary 6.23 (b) that [6]
a
∞ a ∞
1 nk nk
· lim sup αk − γk ∈ →L A →χ ∈ lim sup αk − γk
2 n≤∞ βn n≤∞ βn
k=0 k=0
and
∞ a
nk
L A is compact if and only if lim αk − γk = 0,
n≤∞ βn
k=0
Corollary 6.25 Let A be an infinite matrix, 1 < p < ∞ and q = p/( p − 1). Then,
we have
(c)
(a) If A is in any of the classes (sα0 , 1β ), (sα , 1β ), or (sα , 1β ), then
∞
αk ank
∞
αk ank
sup ∈ →L A → ∈ 4 · sup .
N ∈F βn N ∈F βn
k=0 n∈N k=0 n∈N
p
(b) If A ∈ (α , 1β ), then
∞
αk ank q 1/q
∞
αk ank q 1/q
sup ∈ →L A → ∈ 4 · sup .
N ∈F βn N ∈F βn
k=0 n∈N k=0 n∈N
Remark 6.10 We may note that the matrix classes in part (a) of Corollary 6.25 are
equal, which is immediate by Proposition 4.3 of [21] and Theorem 2.49 (a), above.
Corollary 6.26 Let 1 < p < ∞ and q = p/( p − 1). Then, we have
p
(a) If A ∈ (α , sβ0 ), then
1
∞ 1/q
→L A →χ = lim sup |αk ank |q
n≤∞ βn
k=0
1
∞ 1/q
lim |αk ank |q = 0. (6.45)
n≤∞ βn
k=0
208 6 Application to Compact Matrix Operators
p (c)
(b) If A ∈ (α , sβ ), then
∞
q 1/q ∞
q 1/q
1 αk ank αk ank
·lim sup −αk γk ∈ →L A →χ ∈ lim sup −αk γk
2 n≤∞ βn n≤∞ βn
k=0 k=0
and
∞
q
αk ank
L A is compact if and only if lim − αk γk = 0.
n≤∞ βn
k=0
p
(c) If A ∈ (α , sβ ), then
1
∞ 1/q
0 ∈ →L A →χ ∈ lim sup |αk ank |q
n≤∞ βn
k=0
1 ∞
0 ∈ →L A →χ ∈ lim sup αk |ank |
n≤∞ βn
k=0
and
1 ∞
L A is compact if lim αk |ank | = 0.
n≤∞ βn
k=0
p
Corollary 6.27 Let 1 < p < ∞ and q = p/( p − 1). If A ∈ (α , 1β ), then
(r ) (r )
lim →A→( p ,1 ) ∈ →L A →χ ∈ 4 · lim →A→( p ,1 ) (6.46)
r ≤∞ α β r ≤∞ α β
and
(r )
L A is compact if and only if lim →A→( p ,1 ) = 0,
r ≤∞ α β
where
∞
αk ank q 1/q
→A→(r )
p = sup ; (r ∈ N).
( ,1 )
α β N ∈Fr βn
k=0 n∈N
6.4 Compact Operators Between the Spaces Related to p 209
Here we give necessary and sufficient conditions for a matrix A to define a compact
operator L A between the spaces p , m(φ), and n(φ).
We note that the norms of the BK spaces p for 1 ∈ p ∈ ∞, m(φ), and n(φ)
satisfy the condition
this is trivial for p , and the result for m(φ) and n(φ) can be found in [23, p. 64].
First we establish necessary and sufficient conditions on the entries of a matrix
A ∈ (m(φ), 1 ) or A ∈ (n(φ), 1 ) for L A to be a compact operator.
Given an infinite matrix A = (ank )∞ [m] = (a [m] )∞
n,k=1 and m ∈ N , we write A nk n,k=1
for the matrix with the rows A[m] [m]
n = An for 1 ∈ n ∈ m and An = 0 for n ≥ m + 1;
also let C [m] [m] ∗
= A − A . We denote by sup Nm the supremum taken over all finite
subsets of integers greater than or equal to m + 1.
∗ [m] ;N )
lim C [m] = lim sup b(C =0 (6.50)
m≤∞ (X,1 ) m≤∞ N Xβ
by [17, Theorem 2 (c), (8)] and (5.98)–(5.101) in Parts (c) and (d) of Theorem 4.35.
Let m ∈ N be given, N be a finite subset of N and Nm∩ = {n ∈ N : n ≥ m + 1}.
Then, we obviously have
(C [m] ;N )
[m]
(A;Nm )
bk = cnk = ank = bk for all k,
n∈N n∈Nm∩
hence
210 6 Application to Compact Matrix Operators
∗ [m] ;N ) ∗
sup b(C = sup b(A;Nm )
N Xβ Nm Xβ
We write D [m] = (C [m] )t . Then, D [m] is the matrix with the columns (D [m] )(k) = 0
for 1 ∈ k ∈ m and (D [m] )(k) = Ak = (akn )∞ n=1 for k ≥ m + 1. Now, the conditions
in (6.51) and (6.52) follow from (6.53) by Remark 4.8 and (4.91) and (4.92) in
Corollary 4.37 for X = m(φ), and (4.89) and (4.90) in Corollary 4.37 for X = n(φ).
This completes the proof.
Remark 6.11 It is obvious from Remark 6.3 that the result of Theorem 6.29 extends
to p = 1 and so we obtain alternative characterisations for the classes (m(φ), 1 )
and (n(φ), 1 ) from those given in Theorem 6.28.
Now, we establish necessary and sufficient conditions for the entries of a matrix
A ∈ ( p , m(φ)) or A ∈ ( p , n(φ)) (1 < p ∈ ∞) for L A to be a compact operator.
Given an infinite matrix A = (ank )∞ n,k=1 and m ∈ N, we write A
<m> =
<m> ∞
(ank )n,k=1 for the matrix with the columns (A <m> (k) (k)
) = A for 1 ∈ k ∈ m and
(A <m> (k)
) = 0 for k ≥ m + 1; also let C <m> = A − A<m> .
lim C <m> ( p ,Y )
= 0. (6.56)
m≤∞
Now, the conditions in (6.54) and (6.55) are immediate consequences of (4.85)–(4.88)
in Corollary 4.36. The proof is complete.
Remark 6.12 Let 1 < p < ∞ and X = m(φ) or X = n(φ). It follows from [24,
Lemma 14] and [17, Theorem 3] by [17, Lemma 2 (4)] that if A ∈ (X, p ) then
L A ∈ C(X, p ) if and only if L At ∈ C(q , X β ); also →L At → = →L A → by Remark 4.8.
Thus, the conditions in (6.51) and (6.52) can immediately be obtained from those in
(6.55) and (6.54), respectively, and vice versa.
In the sequel, we always assume that
k
φk ≤ ∞ and ≤ ∞ (k ≤ ∞), (6.57)
φk
Proof (a) Since we assume (6.57), (m(φ))β = n(φ) has AK by [23, Theorem 8
(c)], and it follows from [17, Corollary, p. 84] and (4.78) in Theorem 4.35 (a) that
L A ∈ C(m(φ), ∞ ) if and only if
Now, (6.58) is an immediate consequence of (6.60) and (4.77) in Theorem 4.35 (a).
(b) Let A ∈ (1 , n(φ)). As in Remark 6.12 it follows that L A ∈ C(1 , n(φ)) if and
only if L At ∈ C(m(φ), ∞ ); also →L A → = →L At →. So (6.59) is obtained from (6.58)
with A replaced by At and n and k interchanged.
This completes the proof.
Next, we obtain the necessary and sufficient (or only sufficient) conditions for
the classes of compact matrix operators C(n(φ), c0 ), C(n(φ), c), and C(n(φ), ∞ ).
These results can be found in [9, Corollary, p. 84].
First we require the following lemma which is a special case of Lemma 6.7.
Lemma 6.32 If A ∈ (n(φ), c), then the following hold
∞
lim An (x) = αk xk f or all x = (xk ) ∈ n(φ). (6.64)
n≤∞
k=0
1
· lim sup →An − α→∗n(φ) ∈ →L A →χ ∈ lim sup →An − α→∗n(φ) , (6.66)
2 r ≤∞ n≥r r ≤∞ n≥r
→L A →χ = χ (AS). (6.68)
1
· μ(A) ∈ →L A →χ ∈ μ(A), (6.71)
2
where
μ(A) = lim sup sup →(I − Pr )(Ax)→∞ .
r ≤∞ x∈S
Now, it is given that A ∈ (n(φ), c). Thus, it follows from Lemma 6.32 that the limits
αk = limn≤∞ ank exist for all k, α = (αk ) ∈ n(φ)β = m(φ) and
∞
lim An (x) = αk xk
n≤∞
k=0
∞
→(I − Pr )(Ax)→∞ = sup An (x) − αk xk
n≥r
k=0
∞
= sup (ank − αk )xk
n≥r
k=0
Remark 6.13 It is worth mentioning that the condition in (6.74) is only a suffi-
cient condition for the operator L A to be compact, where A ∈ (n(φ), ∞ ). In
the following
example, we show that it is possible for L A to be compact while
limr ≤∞ supn>r →An →∗n(φ) ⇒ = 0.
Choose a fixed m ∈ N such that xm∩ ⇒ = 0 for some x ∩ = (xk∩ ) ∈ Sn(φ) . Now, we
define the matrix A = (ank ) by anm = 1 and ank = 0 for all k ⇒ = m (n ∈ N). Then,
we have Ax = x m e for all x = (xk ) ∈ w, hence A ∈ (w, ∞ ) ⊂ (n(φ), ∞ ). Also,
since L A is of finite rank, L A is compact. On the other hand, we have An = e(m) and
6.4 Compact Operators Between the Spaces Related to p 215
Exercises
p ⎪
1. Let w0 : = x = (xk ): limn≤∞ n1 nk=1 | xk | p = 0 ,
p ⎪
w0 : = x = (xk ): limn≤∞ n1 nk=1 | xk − L | p = 0 for some L ∈ C , and
p ⎪
w∞ : = x = (xk ): supn≥1 n1 nk=1 | xk | p = 0 .
p p
Then, find the estimates or identities for → L →χ where L ∈ B(w0 , c), B(w0 , c0 ),
p
B(w0 , ∞ ), B(w p , c), B(w p , c0 ) and B(w p , ∞ ). Also deduce the conditions
for the operator L to be compact in each case.
2. Prove the following
(i) If A ∈ (λp , c0 ), then
→L A →χ = lim sup → Ãn →q .
r ≤∞ n>r
1
· lim sup → Ãn − α̃→q ∈ →L A →χ ∈ lim sup → Ãn − α̃→q ,
2 r ≤∞ n≥r r ≤∞ n≥r
where α̃ = (α̃k )∞ k=0 with α̃k = lim n≤∞ ãnk for all k.
(iii) If A ∈ (λp , ∞ ), then
0 ∈ →L A →χ ∈ lim sup → Ãn →q .
r ≤∞ n>r
Further write the conditions for L A to be compact in each case; where the spaces
c0λ , cλ , λ∞ and λp (1 ∈ p < ∞) are defined in Chap. 4.
3. Let 1 ∈ p < ∞. If A ∈ (λ1 , p ), then show that
∞ 1/ p
→L A →χ = lim sup |ãnk | p
r ≤∞ k n=r
and
∞
L A is compact if and only if lim sup |ãnk | p = 0.
r ≤∞ k n=r
216 6 Application to Compact Matrix Operators
4. Let 1 < p < ∞ and q = p/( p − 1). If A ∈ (λp , 1 ), then show that
(r ) (r )
lim → A→(λ , ) ∈ →L A →χ ∈ 4 · lim →A→(λ ,
r ≤∞ p 1 r ≤∞ p 1)
and
(r )
L A is compact if and only if lim →A→(λ , ) = 0,
r ≤∞ p 1
where
∞
q 1/q
→A→(r )
(λ , )
= sup ãnk ; (r ∈ N).
1
p
N ∈Fr k=0 n∈N
∞
inf sup | ai j |q = 0.
k,m j≥m
i=k
∞
∩
inf sup | ai j | p = 0,
k,m i≥k
j=m
10. Obtain the necessary and sufficient (or only sufficient) conditions for the classes
of compact matrix operators C(m(φ), c0 ) and C(m(φ), c).
11. Let Q be a bounded subset of n(φ). Prove that
6.4 Compact Operators Between the Spaces Related to p 217
∞
χ (Q) = lim sup sup | un | φn .
k≤∞ x∈Q u∈S(x) n=k
12. Characterize the classes (m(φ), λp ) and (n(φ), λp ) for 1 ∈ p ∈ ∞, and establish
estimates for the norms of the corresponding operators L A . Further characterize
the compact operators L A defined by the matrices A in the above classes.
13. Characterize the compact operators L A defined by the matrices A for the classes
(λp , m(φ)) and (λp , n(φ)) for 1 ∈ p ∈ ∞.
14. Characterize the classes of compact operators C(m(φ), bv p ) and C(n(φ), bv p )
for 1 ∈ p < ∞.
15. Characterize the classes of compact operators C(n(φ), c0λ ) and C(n(φ), cλ ).
Show that
L A is compact if and only if lim |ānk − āk | = 0,
n
k
References
1. Başarır, M., Kara, E.E.: On compact operators on the Riesz B m -difference sequence spaces.
Iran. J. Sci. Tech. Trans. A 35(A4), 279–285 (2011)
2. Başarır, M., Kara, E.E.: On the B-difference sequence space derived by generalized weighted
mean and compact operators. J. Math. Anal. Appl. 391, 67–81 (2012)
3. Başarır, M., Kara, E.E.: On compact operators on the Riesz B m -difference sequence spaces-II.
Iran. J. Sci. Tech. Trans. A 36(A3), 371–377 (2012)
4. Başar, F., Malkowsky, E.: The characterization of compact operators on spaces of strongly
summable and bounded sequences. Appl. Math. Comput. 217, 5199–5207 (2011)
5. Cohen, L.W., Dunford, N.: Transformations on sequence spaces, Duke Math. J. 3(4), 689–701
(1937)
6. de Malafosse, B., Rakočević, V.: Applications of measure of noncompactness in operators on
(c) p
the spaces sα , sα0 , sα , α . J. Math. Anal. Appl. 323(1), 131–145 (2006)
7. Malkowsky, E., Rakočević, V.: Measure of noncompactness of linear operators between spaces
of sequences that are ( N̄ , q) summable or bounded. Czech. Math. J. 51(3), 505–522 (2001)
8. Mursaleen, M., Karakaya, V., Polat, H., Simsek, N.: Measure of noncompactness of matrix
operators on some difference sequence spaces of weighted means. Comput. Math. Appl. 62,
814–820 (2011)
9. Mursaleen, M., Latif, A.: Applications of measure of noncompactness in matrix operators on
some sequence spaces. Abstr. Appl. Anal. 2012(378250), 10 p
10. Mohiuddine, S.A., Mursaleen, M., Alotaibi, A.: The Hausdorff measure of noncompactness
for some matrix operators. Nonlinear Anal. 92, 119–129 (2013)
11. Mursaleen, M., Noman, A.K.: Applications of the Hausdorff measure of noncompactness in
some sequence spaces of weighted means. Comput. Math. Appl. 60, 1245–1258 (2010)
12. Mursaleen, M., Noman, A.K.: Compactness by the Hausdorff measure of noncompactness.
Nonlinear Anal. 73, 2541–2557 (2010)
218 6 Application to Compact Matrix Operators
13. Mursaleen, M., Noman, A.K.: The Hausdorff measure of noncompactness of matrix operators
on some B K spaces. Oper. Matrices 5(3), 473–486 (2011)
14. Mursaleen, M., Noman, A.K.: Compactness of matrix operators on some new difference
sequence spaces. Linear Algebra Appl. 436(1), 41–52 (2012)
15. Mursaleen, M., Noman, A.K.: Applications of Hausdorff measure of noncompactness in the
spaces of generalized means. Math. Ineq. Appl. 16, 207–220 (2013)
16. Malkowsky, E., Rakočević, V.: An introduction into the theory of sequence spaces and mea-
sures of noncompactness. Zbornik Radova Mat. Inst. SANU (Beograd) 9(17), 143–234 (2000)
17. Sargent, W.L.: On compact matrix transformations between sectionally bounded B K spaces.
J. London Math. Soc. 41(1), 79–87 (1966)
18. Malkowsky, E.: Modern functional analysis in the theory of sequence spaces and matrix
transformations. Jordan J. Math. Stat. 1(1), 1–29 (2008)
19. Malkowsky, E.: Compact matrix operators between some B K spaces. In: Mursaleen, M. (ed.)
Modern Methods of Analysis and its Applications, pp. 86–120. Anamaya Publishers, New
Delhi (2010)
20. Stieglitz, M., Tietz, H.: Matrixtransformationen von folgenräumen eine ergebnisübersicht.
Math. Z. 154, 1–16 (1977)
21. Malkowsky, E., Rakočević, V., Zivkovic, S.: Matrix transformations between the sequence
space bv p and certain B K spaces. Bull. Acad. Serbe Sci. Arts Math. 123(27), 33–46 (2002)
22. de Malafosse, B.: The Banach algebra B(X ), where X is a B K space and applications. Mat.
Vesnik 57, 41–60 (2005)
23. Sargent, W.L.: On sectionally bounded B K spaces. Math. Z. 83, 57–66 (1964)
24. Sargent, W.L.: Some sequence spaces related to the p spaces. J. London Math. Soc. 35,
161–171 (1960)
Chapter 7
Applications to Infinite Systems of Differential
Equations
As we announced in the title, this section is devoted to present some essential results
concerning the theory of differential equations in Banach spaces. Thus, if we denote
by E a Banach space (with a norm || · ||), we consider the ordinary differential
equation
x ∞ = f (t, x) (7.1)
first result in this direction was obtained in the papers [21–23]. That result was an
immediate translation of the well-known Kamke comparison test to the case of an
arbitrary Banach space. It is an important key information that in the mentioned
result, the authors assumed that the function f = f (t, x) appeared in (7.1) satisfies
the condition of the type
for t → [0, T ] (or for t → (0, T ]), where w: [0, T ] × R+ ∈ R+ = [0, ∩) (or
w: (0, T ]×R+ ∈ R+ ) is the so-called Kamke comparison function (cf. [6, 24]). The
quoted result ensures that problem (7.1)–(7.2) has exactly local solution. From this
point of view, the result of Kisyński–Olech–Ważewski was not sufficiently good since
it guaranteed the existence and uniqueness. By this regard, it is not an appropriate
analogue of Peano’s existence theorem
Summing up, the mentioned result of Kisyński–Olech–Ważewski initiated the
quest for conditions ensuring only the existence of solutions of (7.1)–(7.2). The
first conditions of such a type, with the use of the technique of measures of non-
compactness, were received subsequently by Ambrosetti, Szufla, Goebel and Rzy-
moski, Sadowskii [25–28], among others. Below we present some of those results
in a generalized form obtained in [29–31], among others.
To this end, we will give first few auxiliary results which will be used in our
considerations.
At the beginning, let us fix an interval [0, T ] and a Banach space E with a norm
|| · ||. Denote by C = C([0, T ], E) the space of all continuous functions acting from
[0, T ] into E with the maximum norm. A set X → MC will be called regular if all
functions belonging to X are equicontinuous on the interval [0, T ].
For an arbitrary function x → C we will denote by φ(x, λ), its modulus of continuity
(cf. Sect. 5.8) defined by the equality
where d H denotes the Hausdorff metric defined in Chap. 5. This inequality in con-
junction with the obviously estimate
Lemma 7.2 Let X → MC be a regular set and let x0 → E be arbitrarily fixed. Then
t t
μ x0 + X (s)ds ≤ μ(x0 + X (s))ds.
0 0
Hence, applying the result from Exercise 22 (Chap. 5), we derive the following
estimates
⎛ ⎝
t n
μ x0 + X (s)ds ≤ μ x0 + X (εi )(ti − ti−1 )
0 i=1
n
+ λμ(B(L , 1)) ≤ (ti − ti−1 )μ(x0 + X (εi )) + λμ(B(L , 1)).
i=1
Finally, densifying the partition of the interval [0, T ] and keeping in mind that the
number λ was chosen arbitrarily, we obtain
t t
μ x0 + X (s)ds ≤ μ(x0 + X (s))ds.
0 0
decreasing with respect to u, having the property w(t, 0) = 0 and such that u(t) ≡ 0
is the only continuous function on J satisfying the inequality
t
u(t) ≤ w(s, u(s))ds
0
Proof Denote by X 0 the set of all functions x → C([0, T ], E) such that x(0) = x0
and ||x(t) − x(s)|| ≤ A|t − s| for t, s → [0, T ]. Obviously, the set X 0 is non-empty,
bounded, closed, convex, and regular. Apart from this, observe that the transformation
F defined by the formula
t
(F x)(t) = x0 + f (s, x(s))ds
0
maps continuously the set X 0 into itself. Thus, our problem is equivalent to the
existence of a fixed point of F.
Further on, let us denote X i+1 = CoF X i for i = 0, 1, 2, . . .. It is easily seen
that all these sets are of the same type as X 0 and X i+1 ⊂ X i (i = 0, 1, 2, . . .).
Let us put u i (t) = μ(X i (t)) for t → [0, T ] and i = 0, 1, 2, . . .. Then, we have
7.1 Ordinary Differential Equations in Banach Spaces 225
0 ≤ u i+1 (t) ≤ u i (t) for all i = 1, 2, . . .. In view of Lemma 7.1, we deduce that
all functions u i (t) are equicontinuous on the interval [0, T ]. The above-established
facts allow us to infer that the function sequence (u i (t)) converges uniformly to a
function u ∩ (t) = lim u n (t).
n∈∩
Further, let us consider the tangent segment y(t) = x0 + t f (0, x0 ) for t → [0, T ].
We show that y(t) → E μ for t → [0, T ]. To do this observe first that in view of (7.5),
if we put t = 0 and X = {x0 }, we have
μ(x0 + f (0, x0 )) = μ(x0 + f (0, {x0 })) ≤ w(0, μ({x0 })) = w(0, 0) = 0.
Hence, taking into account a result contained in Exercise 21 (Chap. 5), we obtain
and consequently
u ∩ (t)
lim = 0.
t∈0 t
Next, let us observe that the transformation (F x)(t) = f (t, x(t)) maps regular sets
into regular ones, which is a consequence of uniform continuity of the function f .
Hence, on the basis of Lemma 7.2, we get
t
u n+1 (t) = μ(Conv(F X n )(t)) = μ x0 + (F X n )(s)ds
0
t t
≤ μ(x0 + (F X n )(s))ds ≤ w(s, u n (s))ds.
0 0
⎧t
This implies that u ∩ (t) ≤ 0 w(s, u ∩ (s))ds. Hence, keeping in mind our assump-
tions, we conclude that u ∩ (t) ≡ 0. Finally, let us observe that the above-obtained
facts allow us to infer that
for almost all t → [0, T ] and for an arbitrary set X → MC . Obviously, we assume
here that μ is an arbitrary fixed measure of non-compactness in the Banach space E
and {x0 } → ker μ.
Notice that if we take g(t, x) = x0 + f (t, x) then for any x → E μ , in view of
(7.7), we obtain
7.1 Ordinary Differential Equations in Banach Spaces 227
for almost all t → [0, T ]. Thus, in view of continuity of f , we derive that g: [0, T ] ×
E μ ∈ E μ . In particular, x0 + f (0, x0 ) → E μ . Hence, by the same reasoning as in the
proof of Theorem 7.3, we have that the tangent segment {x0 + t f (0, x0 ): t → [0, 1]}
is a subset of the kernel set E μ .
Let us assert that under condition (7.7), the theorem on the existence of solutions
of problem (7.1)–(7.2) follows easily from the above-proved Theorem 7.3. In fact, it
is a consequence of the fact that the function w(t, u) = p(t)u is a particular case of
Kamke comparison function appearing in Theorem 7.3. Nevertheless, we formulate
and prove such a theorem since the proof, we are going to provide, is quite different
than that of Theorem 7.3.
t
(F x)(t) = x0 + f (s, x(s))ds, t → [0, T ].
0
Obviously, the operator F maps continuously the set X 0 into itself. Thus, our problem
is equivalent to the existence of a fixed point of the operator F.
Further on, fix a number δ > 1 and for a set X → MC being regular let us put
t
μδ (X ) = sup μ(X ) exp −δ p(s)ds : t → [0, T ] .
0
One can show that the set function μδ satisfies the axioms of a measure of non-
eq
compactness (cf. Chap. 5) on the family MC consisting of all non-empty and regular
sets belonging to the family MC .
eq
Now, in virtue Lemma 7.2, we get for an arbitrary set X → MC :
228 7 Applications to Infinite Systems of Differential Equations
t t
μ((F X )(t)) = μ x0 + f (s, X (s))ds ≤ μ(x0 + f (s, X (s)))ds
0 0
t t s
≤ p(s)μ(X (s))ds ≤ μδ (X ) p(s) exp δ p(θ )dθ ds
0 0 0
t
1
≤ exp δ p(s)ds μδ (X ).
δ
0
⎧
t
After dividing both sides of the above-obtained inequality by exp δ 0 p(s)ds and
taking supremum on the left-hand side, we derive
1
μδ (F X ) ≤ μδ (X ).
β
eq
The above-obtained inequality means that F is a contraction on the family MC
with respect to the measure of non-compactness μδ (with the constant 1/δ). Thus,
applying the fixed-point theorem of Darbo type (Theorem 5.30), we conclude that
the operator F has at least one fixed point in the space C (more precisely, in the set
X 0 ), which is a solution of problem (7.1)–(7.2). From remarks made in Sects. 5.8
and 5.9, it follows that x(t) → E μ for all t → [0, T ]. Thus, the proof is complete.
In this section, we would like to present some special theorems concerning the
existence of solutions of the initial value problem (7.1)–(7.2). These theorems will
describe the possibility of applying the theory of ordinary differential equations to
the theory of infinite systems of ordinary differential equations which we are going
to discuss in details in the next sections of this chapter.
First of all we discuss a special case of Theorem 7.4. Starting from now, we will
assume that μ is a sublinear measure of non-compactness defined on a Banach space
E and x0 is a fixed element of E such that x0 → E μ , where E μ denotes the so-called
kernel set associated with the measures μ and defined in Chap. 5. Then, we can
reformulate Theorem 7.4 in the following way.
Theorem 7.5 Suppose that the function f : [0, T ] × B(x0 , r ) ∈ E is uniformly
continuous and bounded, i.e., || f (t, x)|| ≤ A, where AT ≤ r . Further, assume that
for any non-empty set X , X ⊂ B(x0 , r ), and for almost all t → [0, T ] the inequality
is satisfied for almost all t → [0, T ], where p(t) is a Lebesgue integrable function on
the interval [0, T ]. Then, the problem (7.1)–(7.2) has at least one solution x = x(t)
on the interval [0, T ] such that x(t) → E μ for t → I .
To prove this theorem, it is sufficient to compare it with Theorem 7.4 and to
observe that our condition (7.8) is equivalent to condition (7.7) imposed in Theorem
7.4. To this end notice that since x0 → E μ (≥ {x0 } → ker μ) and μ is the sublinear
measure of non-compactness, for almost all t → [0, T ] and for X ⊂ B(x0 , r ), X ⇒ = ⊂,
we have
But on the other hand, using the sublinearity of the measure μ again, we have
Keeping in mind the above facts, we see that the assumptions of Theorem 7.5 are
satisfied. This shows that Theorem 7.6 is a special case of Theorem 7.5.
In this section, we will work in the Banach sequence space c0 consisting of real
sequences x = (xi ) converging to zero with the standard maximum norm
where pi (t) and qi (t) are real functions defined and continuous on I such that the
sequence ( pi (t)) converges uniformly on I to the function vanishing identically
and the sequence (qi (t)) is equibounded on I .
7.3 Infinite Systems of Differential Equations in the Space c0 231
Theorem 7.7 Under the above assumptions, initial value problem (7.9)–(7.10) has
at least one solution x = x(t) = (xi (t)) defined on the interval I1 = [0, T1 ], where
T1 < T and QT1 < 1. Moreover, x(t) → c0 for any t → I1 .
Hence, we get
|| f (t, x)|| ≤ P + Q||x||. (7.11)
Next, let us take the ball B(x0 , r ), where r is chosen according to Theorem 7.6. Then,
for an arbitrarily fixed non-empty subset X of the ball B(x0 , r ) and for t → I1 , we
derive the following estimate:
β( f (t, X )) = lim sup {sup{| f i (t, x)|: i ◦ n}}
n∈∩ x→X
= lim sup {sup{| f i (t, x1 , x2 , . . .)|: i ◦ n}}
n∈∩ (xi )→X
≤ lim sup sup{ pi (t) + qi (t) sup[|x p |: p ◦ ki ]
n∈∩ (xi )→X i◦n
≤ lim sup pi (t) + q(t) lim sup sup{sup[|x p |: p ◦ ki ]}
n∈∩ i◦n n∈∩ (xi )→X i◦n
≤ q(t)β(X ). (7.12)
Now, taking into account the assumptions and inequalities (7.11) and (7.12), in view
of Theorem 7.6 and Remark 7.1, we deduce that there exists a solution x = x(t) of
initial value problem (7.9)–(7.10) such that x(t) → c0 for any t → I1 .
This completes the proof.
Example 7.1 Let (kn ) be an increasing sequence of natural numbers. Consider the
infinite system of differential equations of the form
∩
xi∞ = f i (t, x1 , x2 , . . . , xkn ) + ai j (t)x j (7.13)
j=ki +1
∩ functions ai j (t) are defined and continuous on I and the function series
(iii) The
j=ki +1 ai j (t) converges absolutely and uniformly on I (to a function ai (t))
for any i = 1, 2, . . ..
(iv) The sequence (ai (t)) is equibounded on I .
(v) QT < 1, where Q = sup{ai (t): i = 1, 2, . . . , t → I }.
It can be shown that under assumptions (i)–(v) listed above, the assumptions of
Theorem 7.7 are satisfied. This implies that problem (7.13)–(7.14) has a solution
x(t) = (xi (t)) on the interval I which belongs to the space c0 for each fixed t → I .
We omit the standard details.
Let us mention that the above-considered problem (7.13)–(7.14) contains as a
special case the infinite system of differential equations occurring in the theory of
dissociation of polymers (cf. [4, p. 93], [9]). That system was investigated in [9] in
the sequence space l ∩ under very strong assumptions. The existence result proved
in [4] requires rather restrictive assumptions. This means that our result generalizes
those quoted above.
It is also worthwhile mentioning that the choice of the space c0 for the study of
the problem (7.13)–(7.14) enables us to obtain partial characterization of solutions
of this problem since we have xn (t) ∈ 0 when n ∈ ∩, for each fixed t → [0, T ].
Moreover, let us also notice that in the study of the heat conduction problem via
the method of semidiscretization, we can obtain the infinite system of form (7.13)
(cf. [5], for details).
Example 7.2 Now, we will consider some special case of problem (7.13)–(7.14) and
assume that ki = i for i = 1, 2, . . . and ai j ≡ 0 on I for all i, j. Then, system (7.13)
has the form
7.3 Infinite Systems of Differential Equations in the Space c0 233
x1∞ = f 1 (t, x1 ),
x2∞ = f 2 (t1 , x1 , x2 ),
····················· (7.15)
xi∞ = f i (t, x1 , x2 , . . . , xi ),
·····················
Putting Pi (t) = max{ p1 (t), p2 (t), . . . , pi (t)}, we can write the above estimate in
the following form:
Observe that from our assumptions, it follows that the initial value problem u ∞ =
Pi (t), u(0) = xi0 has a unique solution on the interval I . Hence, applying a result
from [4], we deduce that Cauchy problem (7.15)–(7.14) has a solution on the interval
I . Obviously, from the result contained in Theorem 7.7 and Example 7.1, we deduce
additionally that the mentioned solution belongs to the space c0 .
It is worthwhile mentioning that the result described above for row-finite system
of the type (7.15) can be obtained under more general assumptions. In fact, instead
of inequality (7.16), we may require that the following estimate holds to be satisfied:
where the functions pi (t) and qi (t)(i = 1, 2, . . .) satisfy the hypotheses analogous
to those imposed in Theorem 7.7.
We omit standard details.
Remark 7.2 Notice that considering the so-called birth process, one can obtain a
special case of the infinite system (7.15) which is lower diagonal linear infinite system
[4, 9]. Thus, the result obtained in the above example generalizes that from [4, 6].
234 7 Applications to Infinite Systems of Differential Equations
Theorem 7.8 Let assumptions (i)–(iv) be satisfied. If QT < 1, then the initial value
problem (7.17)–(7.18) has a solution x(t) = (xi (t)) on the interval I such that
x(t) → c for any t → I .
Keeping in mind assumptions (iii) and (iv), from the above estimate, we infer that
( f i (t, x)) is a real Cauchy sequence.
This yields that ( f i )(t, x) → c.
Next, we obtain the following inequality:
| f i (t, x)| ≤ |ai (t)| · |xi | + |gi (t, x)| ≤ Q|xi | + bi ≤ Q||x|| + B,
In what follows, let us consider the mapping f (t, x) on the set I × B(x0 , r ), where
r is chosen according to the assumptions of Theorem 7.6, i.e.,
Next, fix arbitrarily t, s → I and x, y → B(x0 , r ). Then, keeping in mind our assump-
tions, for a fixed i, we get
Hence, taking into account that the sequence (ai (t)) is equicontinuous on the interval
I and g is uniformly continuous on I × c, we conclude that the operator f (t, x) is
uniformly continuous on the set I × B(x0 , r ).
Further, let us take a non-empty subset X of the ball B(x0 , r ) and fix t → I ,
x → X .Then, for arbitrarily fixed natural numbers n and m, we have
Finally, joining (7.19), (7.20) and the fact that f is uniformly continuous on I ×
B(x0 , r ), in view of Theorem 7.6, we infer that problem (7.17)–(7.18) is solvable in
the space c.
Thus, the proof is complete.
Remark 7.3 In view of the fact that the measure of non-compactness μ used in
this section is regular and equivalent to the Hausdorff measure of non-compactness
β in the space c (cf. the beginning of this section), we infer that assumption (ii)
of Theorem 7.8 on the uniform continuity of the function g(t, x) on the set I × c
can be replaced by the weaker assumption requiring the continuity of g(t, x) on I ×c.
Remark 7.4 The infinite systems of differential Eqs. (7.17)–(7.18) considered above
contain as special case the systems studied in the theory of neutral sets (cf. [4] , pp.
86-87] and [9], for example). It is easy to observe that the existence results proved
in [4, 9] were obtained under stronger and more restrictive assumptions than our one.
The result contained in Theorem 7.8 comes from the paper [13]. In what follows,
we give a result of other type which was obtained in [18].
We will consider the following infinite system of differential equations
n
xn∞ = ai (t)xi + gn (t, xn+1 , x x+2 , . . .) (7.21)
i=1
while mentioning that systems of the above type can be encountered in the theory of
neutral nets and in discretization of some boundary problems for partial differential
equations (cf. [1, 4, 9, 11, 37], for example).
At the beginning, we formulate the assumptions under which the problem (7.21)–
(7.22) will be studied. First of all, we assume that the functions gn (n = 1, 2, . . .)
are defined on the set I × R∩ and take real values. Moreover, we assume that the
following conditions are satisfied:
(a) x0 = (xn0 ) → c.
(b) The mapping g = (g1 , g2 , . . .) : I × c ∈ c is uniformly continuous.
(c) There exists a sequence (bi ) converging to zero such that
n
sn (t) = |ai (t)|,
i=1
∩
s(t) = sup{sn (t): n = 1, 2, . . .} = lim sn (t) = |ai (t)|.
n∈∩
i=1
Observe that from assumption (e), it follows the uniform continuity of the function
s(t) on the interval I . Indeed, taking arbitrary u, v → I we obtain:
∩
∩
|s(u) − s(v)| ≤ ||ai (u)| − |ai (v)|| ≤ |ai (u) − ai (v)|
i=1 i=1
∩ ∩
≤ di w(|u − v|) = w(|u − v|) di .
i=1 i=1
238 7 Applications to Infinite Systems of Differential Equations
Q = sup{s(t): t → I }
is finite.
∩ Next, let us notice that from assumption (e), we can deduce that the series
i=1 |ai (t)| is uniformly convergent on I . This fact follows easily from the well-
known Dini theorem applied to the sequence (si (t)).
Now, we can formulate the announced existence result.
Theorem 7.9 Let assumptions (a)–(e) be satisfied and let QT < 1. Then, the initial
value problem (7.21)–(7.22) has at least one solution x = x(t) = (xi (t)) defined on
the interval I and such that x(t) → c for any t → I .
n
f n (t, x) = ai (t)xi + gn (t, xn+1 , xn+2 , . . .),
i=1
At first we show that the mapping f = f (t, x) transforms the set I × c into c. To
do this, fix arbitrarily natural numbers n and m such that m < n. Then, based on our
assumptions, we derive the following estimate:
⎪ n ⎪
⎪
m ⎪
⎪ ⎪
| f n (t, x) − f m (t, x)| ≤ ⎪ ai (t)xi − ai (t)xi ⎪
⎪ ⎪
i=1 i=1
+ |gn (t, xn+1 , xn+2 )| + |gm (t, xm+1 , xm+2 , . . .)|
n
n
≤ |ai (t)||xi | + bn + bm ≤ ||x|| |ai (t)| + bn + bm .
i=m+1 i=m+1
Hence, taking into account that the sequence (bi ) converges to zero and the series
∩
i=1 |ai (t)| is convergent for t → I , we infer that the sequence ( f i (t, x)) satisfies
the Cauchy condition which yields that this sequence belongs to the space c.
Further, for x → c and t → I and for fixed n → N we get
n
| f n (t, x)| ≤ |ai (t)| · |xi | + |gn (t, xn+1 , xn+2 , . . .)|
i=1
n
≤ ||x|| |ai (t)| + bn = ||x||sn (t) + bn ≤ Q||x|| + B,
i=1
7.4 Infinite Systems of Differential Equations in the Space c 239
In what follows, we will consider the mapping f = f (t, x) (defined above) on the
set I × B(x0 , r ), where r is chosen according to Theorem 7.6, i.e.,
Now, we show that f is uniformly continuous on the set I × B(x0 , r ). To this end,
fix t, s → I and x, y → B(x0 , r ). Then, in virtue of our hypotheses, for a fixed n, we
obtain:
n
| f n (t, x) − f n (s, y)| ≤ |ai (t)xi − ai (s)yi |
i=1
+ |gn (t, x n+1 , xn+2 , . . .) − gn (s, yn+1 , yn+2 , . . .)|
n
≤ (|ai (t)xi − ai (t)yi | + |ai (t)yi − ai (s)yi |)
i=1
n
+ |gn (t, x) − gn (s, y)| ≤ |ai (t)||xi − yi |
i=1
n
+ |ai (t) − ai (s)||yi | + |gn (t, x) − gn (s, y)|
i=1
n
≤ ||x − y||sn (t) + ||y|| |ai (t) − ai (s)| + |gn (t, x) − gn (s, y)|
i=1
n
≤ Q||x − y|| + (||x 0 || + r ) di w(|t − s|) + |gn (t, x) − gn (s, y)|
i=1
n
= Q||x − y|| + (||x 0 || + r )w(|t − s|) di + |gn (t, x) − gn (s, y)|.
i=1
n
|| f (t, x) − f (s, y)|| ≤ Q||x − y|| + (||x0 || + r )w(|t − s|) di
i=1
+ ||g(t, x) − g(s, y)||.
In virtue of assumptions (b) and (e) from the above estimate, we conclude that
the mapping f = f (t, x) is uniformly continuous on the set I × B(x0 , r ).
240 7 Applications to Infinite Systems of Differential Equations
Further on, let us take a non-empty subset X of the ball B(x0 , r ) and an arbitrary
number t → I . Fix x → X and n, m → N such that m < n.
Then, keeping in mind our assumptions, we obtain:
n
| f n (t, x) − f m (t, x)| ≤ |ai (t)||xi | + bn + bm
i=m+1
n
≤ |ai (t)|(|xi − xi+1 | + |xi+1 |) + bn + bm
i=m+1
n
n
= |ai (t)||xi − xi+1 | + |ai (t)||x I +1 | + bn + bm
i=m+1 i=m+1
n
≤ (sup[|xk − xl |: k, l = m + 1, m + 2, . . . , n]) |ai (t)|
i=m+1
+ ||x||(sn (t) − sm (t)) + bn + bm ≤ s(t) sup[|xk − xl |: k, l ◦ p]
+ ||x||(sn (t) − sm (t)) + bn + bm ,
∩
Hence, in view of the facts that the series i=1 |ai (t)| is convergent and bi ∈ 0 as
i ∈ ∩, we deduce the following estimate
μ( f (t, X )) ≤ s(t)μ(X )
for every t → I .
The above estimate means that there is satisfied condition (7.8) of Theorem 7.6.
Moreover, from (7.23) and other facts proved above, we see that all assumptions
of Theorem 7.6 are satisfied. Thus, in view of this theorem, we conclude that there
exists at least one solution x = x(t) of the problem (7.21)–(7.22) defined on the
interval I and such that x(t) → c for any t → I .
This completes the proof.
Remark 7.5 Observe that the method used in the above-given proof allows us to
replace assumption (b) by the following weaker assumption:
(b∞ ) g = (g1 , g2 , . . .): I × c ∈ c and for every non-empty and bounded subset X of
c, the mapping g is uniformly continuous on the set I × X .
Moreover, keeping in mind Remark 7.3, we infer that assumption (b∞ ) can be
replaced by the following weaker one:
(b∞∞ ) g = (g1 , g2 , . . .): I × c ∈ c and for every non-empty and bounded subset X of
c, the mapping g is continuous on the set I × X .
The main tool used in our considerations will be the Hausdorff measure of non-
compactness which is expressed in the space β1 by the formula from Theorem 5.18
(a), i.e.,
∩
β(X ) = lim sup |xk | , (7.24)
n∈∩ x=(xi )→X k=n
242 7 Applications to Infinite Systems of Differential Equations
where X → Mβ1 .
We will consider the infinite system of differential equations
(i = 1, 2, . . .).
We are interested in the existence of solutions x = x(t) = (xi (t)) of problem
(7.25)–(7.26) which are defined on the interval I = [0, T ] and are such that x(t) → β1
for any t → I .
It turns out that an existence theorem for problem (7.25)–(7.26) in the space β1 can
be formulated under slightly weaker assumptions that in the space c0 (cf. Sect. 7.3).
Hence, we will assume that the following hypotheses are satisfied:
(i) x0 = (xi0 ) → β1 .
(ii) f i : I × R∩ ∈ R (i = 1, 2, . . .) maps continuously the set I × β1 into β1 .
(iii) There exist nonnegative functions pi (t) and qi (t) defined on I such that
Theorem 7.10 Under the above assumptions, problem (7.25)–(7.26) has at least
one solution x(t) = (xi (t)) defined on the interval I = [0, T ] whenever QT < 1,
where Q is defined as the number
∩
∩
|| f (t, x)|| = | f i (t, x1 , x2 , . . .)| ≤ [ p1 (t) + qi (t)|xi |]
i=1 i=1
∩
∩
≤ pi (t) + sup{qi (t): i = 1, 2, . . .} |xi |
i=1 i=1
≤ P + Q||x||,
where ∩
P = sup pi (t): t → I .
i=1
Further, choose the number r defined in Theorem 7.6, i.e., r = (P T + QT ||x0 ||)/(1−
QT ). Consider the operator f = ( f i ) on the set I × B(x0 , r ). In view of Remark 7.1,
we have only to check that the operator F satisfies condition (7.8) from Theorem
7.6. Thus, let us take a set X such that X → Mβ1 . Then, we get
∩
β( f (t, X )) = lim sup | f k (t, x1 , x2 , . . .)|
n∈∩ x=(xi )→X ) k=n
∩
≤ lim sup ( pk (t) + qk (t)|xk |): x = (xi ) → X
n∈∩
k=n
⎫ ∩ ∩
⎬
≤ lim pk (t) + sup{qk (t): k ◦ n} |xk | .
n∈∩
k=n k=n
Hence, in virtue of assumptions (iv) and (v), we deduce the following estimate
β( f (t, X )) ≤ q(t)β(X ).
Let us observe that the above theorem can be applied to the perturbed diagonal
system of differential equations of the form
where i = 1, 2, . . . and t → I (cf. Remark 7.4). In this case, we may assume that the
following conditions are satisfied:
(a) (xi0 ) → β1 .
244 7 Applications to Infinite Systems of Differential Equations
(b) The sequence (|ai (t)|) is defined and equibounded on the interval I = [0, T ].
Moreover, the function
a(t) = lim sup |ai (t)|
i∈∩
∩
∩
|| f (t, x)|| p = | f i (t, x)| p ≤ [qi (t) + ri (t)|xi | p ]
i=1 i=1
∩
∩
≤ qi (t) + sup{ri (t): i = 1, 2, . . .} |xi | p
i=1 i=1
≤ Q + R||x|| p ,
∩
where Q = sup qi (t): t → I .
i=1
Further on, let us choose the number r = (QT + RT ||x0 || p )/(1 − RT ) being
defined in Theorem 7.6. Next, consider the operator f = ( f i ) on the set I × B(x0 , r ).
Then, for an arbitrary set X → Mβ p , using the mentioned formula for the Hausdorff
measure of non-compactness in the space l p , we obtain
∩
β( f (t, X )) = lim sup | f i (t, x1 , x2 , . . .)| p
n∈∩ x=(xi )→X i=n
⎫∩ ∩
⎬
≤ lim qi (t) + sup{ri (t): i ◦ n, t → I } |xi | p .
n∈∩
i=n i=n
β( f (t, X )) ≤ r (t)β(X )
which means that the condition (7.8) of Theorem 7.5 is satisfied. Hence, keeping
in mind Theorem 7.6 and Remark 7.1, we conclude that there exists a solution
x = x(t) = (xi (t)) of problem (7.25)–(7.26) such that x(t) → β p for any t → I .
The proof is complete.
Let us observe that, likewise as in the case of the space l1 , we can apply the
above-proved Theorem 7.11 to the perturbed diagonal infinite system of differential
equations of the form
xi∞ = ai (t)xi + gi (t, x1 , x2 , . . .)
where t → I and i = 1, 2, . . ..
An existence theorem for the above initial value problem in the space β p can be
formulated by imposing the following assumptions.
246 7 Applications to Infinite Systems of Differential Equations
(A) x0 = (xi0 ) → β p .
(B) The sequence (|ai (t)|) is defined and equibounded on the interval I = [0, T ].
Moreover, the function
a(t) = lim sup |ai (t)|
i∈∩
is integrable over I .
(C) The mapping g = (gi ) transforms continuously the set I × β p into β p .
(D) There exist nonnegative functions bi (t) such that
Now, we will study infinite systems of differential equations in the setting of the
sequence space l∩ . It is known (cf. Sect. 5.8) that in this space, formulas expressing
the Hausdorff measure of non-compactness are not known and even formulas for
regular measures of non-compactness [29]. However, we will use the measures of
non-compactness μ1 and μ3 defined in Sect. 5.8 by formulas (5.64) and (5.66),
respectively. Those measures are sufficiently convenient for our purposes.
At the beginning, we will consider the infinite system of differential equations
Theorem 7.12 Under hypotheses (i)–(vi), the initial value problem (7.27)–(7.28)
has at least one solution x = x(t) = (xi (t)) such that x(t) → l∩ for any t →
I1 = [0, T1 ], where T1 ≤ T and T1 is a number chosen according to Theorem 7.4.
Moreover, lim xi (t) = a uniformly with respect to t → I1 .
i∈∩
This means that the function f appearing in assumption (iv) is bounded on the ball
B(x0 , r ).
Further, let us choose a number T1 in such a way that T1 ≤ T , T1 ≤ 1 and
[P(r + ||x0 ||) + D]T1 ≤ r . Denote I1 = [0, T1 ]. Then, for an arbitrary non-empty
set X ⊂ B(x0 , r ) and for t → I1 , we obtain
μ1 (x0 + f (t, X ))
= lim sup sup |xi0 + ai (t)xi + f i (xi , xi+1 , xi+2 , . . .) − a|
i∈∩ x=(x0 )→X
≤ lim sup sup [|ai (t)||xi | + | f i (xi , xi+1 , xi+2 , . . .) + xi0 − a|]
i∈∩ x=(xi )→X
≤ lim sup sup [ p(t)|xi − a| + |ai (t)||a|
i∈∩ x=(xi )→X
+ | f i (xi , xi+1 , xi+2 , . . .)| + |xi0 − a|] .
248 7 Applications to Infinite Systems of Differential Equations
Hence, we get
μ1 (x0 + f (t, X )) ≤ p(t)μ1 (X ).
Finally, we assert that all assumptions of Theorem 7.4 are satisfied. In view of that
theorem, problem (7.27)–(7.28) has at least solution x = x(t) = (xi (t)) in the space
l∩ , where the sequence (xi (t)) converges uniformly to a on the interval I1 .
The proof is complete.
In what follows, we will consider the following semilinear infinite system of
differential equations
∩
xi∞ = ai j (t)xi + gi (t, x1 , x2 , . . .) (7.30)
j=1
Taking into account assumptions (d), (e), for an arbitrarily fixed i → N, we can
consider the functions Ai (t), Ai (t) and Ai (t) defined on the interval I in the following
way:
∩
Ai (t) = ai j (t),
j=1
i−1
Ai (t) = ai j (t),
j=1
∩
Ai (t) = ai j (t).
j=i
Obviously, the above formula defines the function Ai (t) for i ◦ 2. We can extend
this definition by putting A1 (t) = 0 for t → I . Moreover, let us observe that the
7.6 Infinite Systems of Differential Equations in the Space β∩ 249
functions Ai (t), Ai (t) and Ai (t) are nonnegative and non-decreasing on the interval
I.
In what follows, we will additionally impose the following assumptions.
(f) The sequence (Ai (t)) converges uniformly to zero on I .
(g) The sequence (Ai (t)) is equicontinuous and equibounded on the interval I .
Remark 7.6 Let us observe that in assumption (b), it is sufficient to require that the
mapping g = (g1 , g2 , . . .) is uniformly continuous on the set I × B(x0 , r ) for an
arbitrary fixed r > 0. It is an immediate consequence of Theorem 7.6 which will be
used in the proof of the below-presented result.
For our further purposes, let us define the following constants:
Notice that in view of the imposed assumptions, we have A < ∩ and B < ∩.
Finally, let us mention that in our further considerations, we will use the measure
of non-compactness μ3 defined by formula (5.66). It is a sublinear measure, and its
kernel ker μ3 is described in Sect. 5.8.
Now, we formulate our next result [12].
Theorem 7.13 Let assumptions (a)–(g) be satisfied and let AT < 1. Then, problem
(7.30)–(7.31) has at least one solution x = x(t) = (xi (t)) defined on the interval I
and such that x(t) → β∩ for t → I .
where the symbol || · || stands for the norm in the space β∩ . From the above estimate,
we deduce that the operator f = f (t, x) transforms the set I × β∩ into β∩ .
Now, let us take the number r = (BT + AT ||x0 ||)/(1 − AT ) (cf. Theorem 7.6).
We will consider the operator f on the set I × B(x0 , r ).
Next, fix arbitrarily t, s → I and x, y → B(x0 , r ). Without loss of generality, we
may assume that s < t (cf. assumption (d)). Then, in virtue of our assumptions, for
a fixed natural number i, we get
Hence, keeping in mind assumptions (b) and (g), we infer that the operator f (t, x)
is uniformly continuous on the set I × B(x0 , r ).
Further on, take a non-empty subset X of the ball B(x0 , r ) and fix x, y → X, t → I .
Then, for arbitrarily fixed natural number i, i ◦ 2, we obtain
7.6 Infinite Systems of Differential Equations in the Space β∩ 251
⎪ ⎪
⎪∩ ∩ ⎪
⎪ ⎪
⎪
| f i (t, x) − f i (t, y)| ≤ ⎪ ai j (t)x j − ai j (t)y j ⎪⎪
⎪ j=1 j=1 ⎪
+ |gi (t, x1 , x2 , . . .) − gi (t, y1 , y2 , . . .)|
⎪ ⎪
⎪ i−1 ∩ ∩ ⎪
⎪
i−1 ⎪
⎪
≤⎪ ai j (t)x j + ai j (t)x j − ai j (t)y j − ai j (t)y j ⎪⎪
⎪ j=1 j=i j=1 j=i ⎪
+ |gi (t, x1 , x2 , . . .)| + |gi (t, y1 , y2 , . . .)|
⎪ ⎪ ⎪ ⎪
⎪ i−1 ⎪ ⎪∩ ⎪
⎪ ⎪ ⎪ ⎪
⎪
≤⎪ ⎪
ai j (t)(x j − y j )⎪ + ⎪ ⎪ ai j (t)(x j − y j )⎪⎪ + 2bi
⎪ j=1 ⎪ ⎪ j=i ⎪
i−1 ∩
≤ ai j |x j − y j | + ai j (t)|x j − y j | + 2bi
j=1 j=i
i−1 ∩
≤ ||x − y|| ai j (t) + ai j (t) sup{|x j − y j |: j ◦ i} + 2bi
j=1 j=i
The above estimate and assumptions (c) and (e)–(g) allow us to deduce the following
inequality
μ3 ( f (t, X )) ≤ p(t)μ3 (X ), (7.33)
where the function p(t) is defined on the interval I in the following way:
Finally, taking into account (7.32), (7.33) and other facts established in the above-
conducted proof, in view of Theorem 7.6, we infer that there exists a solution x(t) =
252 7 Applications to Infinite Systems of Differential Equations
(xi (t)) of problem (7.30)–(7.31) such that x(t) → β∩ for each t → I . The proof is
complete.
Remark 7.7 Notice that on the basis of Theorem 7.6, it can be shown [29, 38] that
all solutions x = x(t) = (xi (t)) of problem (7.30)–(7.31) belonging to the ball
B(x0 , r ), i.e., x(t) → B(x0 , r ) for t → I , are such that x(t) → ker μ3 for t → I , where
μ3 is the measure of non-compactness in β∩ defined by formula (5.66).
given in Sect. 5.8, we conclude that all solutions of the infinite system of differ-
ential equations (7.30), satisfying the initial conditions (7.31) and belonging to the
ball B(x0 , r ) for all t → I are asymptotically coordinable stable according to the
following definition accepted in [38].
We say that solutions of problem (7.30)–(7.31) are asymptotically coordinable
stable if for any λ > 0 and t → I there exists i 0 → N such that for arbitrary solutions
x(t), y(t) of (7.30)–(7.31) with x, y → B(x0 , r ), we have |xi (t) − yi (t)| ≤ λ for
i ◦ i0 .
In what follows, we are going to present the corrected version of the result obtained
in the paper [38], where the following perturbed semilinear upper diagonal infinite
system of differential equations was investigated
∩
xi∞ = ai j (t)x j + gi (t, x1 , x2 , . . .) (7.34)
j=i
Theorem 7.14 Assume that the hypotheses (a)–(c) of Theorem 7.13 are satisfied and
additionally the following ones:
(d∞ ) For all pairs of natural numbers (i, j) such that j ◦ i the function ai j : I ∈ R+
is non-decreasing on the interval I .
∩
(e∞ ) For any i → N, the function series is uniformly convergent on the interval I .
j=i
7.6 Infinite Systems of Differential Equations in the Space β∩ 253
∩
(f∞ ) The sequence (Ai (t)), where Ai (t) = ai j (t), is equicontinuous and equi-
j=i
bounded on I .
If AT < 1, then problem (7.34)–(7.35) has at least one solution x = x(t) =
(xi (t)) defined on the interval I such that x(t) → β∩ for any t → I .
tarctan(xi + xi+1 )
gi (t, x1 , x2 , . . .) = ,
i + xi2 + xi+1
2
for t → I . This implies that the sequence (Ai (t)) appearing in assumption ( f ) is
uniformly convergent to zero on the interval I . Indeed, in view of the inequality
1
Ai (t) ≤ Ai (t) ≤ − ln(1 − T )
i
we derive our assertion.
Further, observe that in view of equality (7.36), we infer that there is satisfied
assumption (g). Moreover, we have
Next, let us notice that the following inequality holds for an arbitrary i → N and for
x = (xi ) → l∩ .
T ν/2 Tν
|gi (t, x1 , x2 , . . .)| ≤ ≤ .
i + xi + xi+1
2 2 2i
254 7 Applications to Infinite Systems of Differential Equations
This yields immediately that assumption (c) is satisfied with bi = T ν/2i for i =
1, 2, . . .. Obviously, assumption (i) is satisfied provided we impose that x0 = (xi0 ) →
β∩ .
Finally, let us fix arbitrarily a number r > 0 and consider the ball B(x0 , r ) in the
space β∩ . Then, it is easy to verify that the function gi (t, x1 , x2 , . . .) is uniformly
continuous on the set I × B(x0 , r ). This statement is a simple consequence of the
fact that the mapping g1 = g1 (t, x1 , x2 , . . .) has the largest modulus of continuity
among the functions gi (i = 1, 2, . . .) on the set I × B(x0 , r ). On the other hand,
all functions gi are uniformly continuous on the set I × B(x0 , r ) since the function
gi (t, x1 , x2 , . . .) depends upon three variables only. Thus, keeping in mind Remark
7.6, we conclude that assumption (b) of Theorem 7.13 is satisfied.
Now, we deduce that the semilinear infinite system of differential equations con-
sidered here has at least one solution x = x(t) = (xi (t)) defined on the interval
I = [0, T ], where T < 1 and T satisfy the following inequality [cf. (7.37)]
−T ln(1 − T ) < 1.
Apart from this, we have that (xi (t)) → β∩ for each t → I and solutions of the studied
infinite system are asymptotically coordinable stable (cf. Remark 7.7).
The proof is complete.
Starting from now, we will consider some particular cases of the perturbation
term g = g(t, x) = (g1 (t, x), g2 (t, x), . . .) appearing in infinite system (7.30). At
the beginning, let us take into account the following semilinear infinite system of
differential equations
∩
xi∞ = ai j (t)x j + gi (t, xi , xi+1 , . . .) (7.38)
j=1
Next, observe that keeping in mind assumptions (b∞ ), (c∞ ), and (c∞∞ ), we can define
the following finite constants:
Further on, let us notice that assumptions (c∞ ) and (c∞∞ ) imply that the function g =
g(t, x) satisfies the Lipschitz condition with the constant k with respect to the variable
x. Indeed, for arbitrarily fixed x = (xi ), y = y(yi ) → β∩ and for any fixed t → I ,
we obtain
||g(t, x) − g(t, y)|| = sup 1{|gi (t, xi , xi+1 , . . .) − gi (t, yi , yi+1 )|: i = 1, 2, . . .}
≤ sup{ki sup{|x j − y j |: j ◦ i}: i = 1, 2, . . .}
≤ sup{ki ||x − y||: i = 1, 2, . . .} ≤ k||x − y||. (7.40)
Apart from this, observe that the function g = g(t, x) is uniformly continuous on
the set I × B(x0 , r ), where r > 0 is arbitrarily fixed.
To prove this assertion, fix arbitrary t1 , t2 → I and x1 , x2 → B(x0 , r ). Then, in
view of (7.40), we get
Theorem 7.15 Suppose that assumptions (a), (b∞ ), (c∞ ), (c∞∞ ), and (d)–(g) are sat-
isfied and T (A + k) < 1. Then, problem (7.38)–(7.39) has at least one solution
x = x(t) = (xi (t)) defined on the interval I = [0, T ] and such that x(t) → β∩ for
t → I.
Proof We proceed similarly as in the proof of Theorem 7.13. Thus, fix i → N. Then,
for arbitrarily chosen x = (xi ) → β∩ and t → I , in view of assumptions and the facts
established above, we get
∩
| f i (t, x)| ≤ ai j (t)|x j | + |gi (t, xi , xi+1 , . . .)|
j=1
≤ Ai (t)||x|| + |gi (t, xi , xi+1 ) − gi (t, 0, 0, . . .)| + |gi (t, 0, 0, . . .)|
≤ Ai (t)||x|| + ki sup{|x j |: j ◦ i} + G ≤ A||x|| + ki ||x|| + G.
The above estimate allows us to infer that the operator f = f (t, x) transforms the
set I × β∩ into β∩ .
Further, let us take the number
Consider the operator f (t, x) on the set I × B(x0 , r ). In view of the earlier-stated
uniform continuity of the operator g = g(t, x) on the set I × B(x0 , r ) and the
reasoning conducted in the proof of Theorem 7.13, we deduce that the operator
f (t, x) is uniformly continuous on the set I × B(x0 , r ).
In what follows, let us fix a non-empty subset X of the ball B(x0 , r ) and take
arbitrary x, y → X and t → I . Then, for arbitrarily fixed i → N, we obtain
⎪ ⎪
⎪∩ ∩ ⎪
⎪ ⎪
⎪
| f i (t, x) − f i (t, y)| ≤ ⎪ ai j (t)x j − ai j (t)y j ⎪⎪
⎪ j=1 j=1 ⎪
+ |gi (t, xi , x j+1 , . . .) − gi (t, yi , yi+1 , . . .)|
⎪ ⎪
⎪ i−1 ∩ ∩ ⎪
⎪
i−1 ⎪
⎪
≤⎪ ai j (t)x j + ai j (t)x j − ai j (t)y j − ai j (t)y j ⎪⎪
⎪ j=1 j=i j=1 j=i ⎪
+ ki sup{|x j − y j |: j ◦ i}
⎪ ⎪ ⎪ ⎪
⎪ i−1 ⎪ ⎪∩ ⎪
⎪ ⎪ ⎪ ⎪
⎪
≤⎪ ⎪ ⎪
ai j (t)(x j − y j )⎪ + ⎪ ai j (t)(x j − y j )⎪⎪
⎪ j=1 ⎪ ⎪ j=i ⎪
+ ki sup{diamX j : j ◦ i}
i−1 ∩
≤ ai j (t)|x j − y j | + ai j (t)|x j − y j |
j=1 j=i
+ k sup{diamX j : j ≤ i}
i−1 ∩
≤ ||x − y|| ai j (t) + ai j (t) sup{|x j − y j |: j ◦ i}
j=1 j=i
+ k sup{diamX j : j ◦ i}
≤ Ai (t)diamX + Ai (t) sup{diamX j : j ◦ i}
+ k sup{diamX j : j ◦ i}.
i
(c) For each i → N, we have ki = kij < 1. Moreover, k < 1, where k =
j=1
sup{ki : i = 1, 2, . . .}.
Then, we can formulate the following existence result.
Theorem 7.16 Under assumptions (a), (b), (c), (d)–(g), if additionally T(A+k)<1,
problem (7.42)–(7.43) has at least one solution x = x(t) = (xi (t)) defined on the
interval I and such that x(t) → β∩ for any t → I .
The proof can be conducted similarly as the proof of Theorem 7.15 and is left to
a reader (see Exercise 16).
Exercises
x1∞ = x1
x2∞ = x1 + x2
x3∞ = x1 + x2 + x3
··················
xn∞ = x1 + x2 + . . . + xn
··················
x1 (0) = x2 (0) = . . . = 1.
for t → [0, T ] and x, y → B(x 0 , r ), where B(x0 , r ) denotes the closed ball in a
Banach space E centered at x0 and with radius r and w(t, u) is a Kamke com-
parison function according to the definition given in Sect. 7.1 (before Theorem
7.3). Show that the initial value problem x ∞ = f (t, x), x(0) = x0 has exactly
one local solution.
6. Let A, B be classes of functions defined below.
t
u(t) ≤ w(s, u(s))ds
0
Show that A = B.
7. Consider the following classes of comparison functions:
Class C. This class comprises all functions w = w(t, u) such that w: [0, T ] ×
R+ ∈ R+ satisfies Carathéodory’s conditions (i.e., w is Lebesgue measurable
in t for each u and continuous in u for any fixed t). Moreover, for each t0 → (0, T ]
and u 0 > 0, there exists a Lebesgue integrable function h = h(t) on the interval
[t0 , T ] such that w(t, u) ≤ h(t) for (t, u) → [t0 , T ] × [0, u 0 ]. Next, w(t, 0) = 0
and the only absolutely continuous function u = u(t) such that u ∞ = w(t, u)
almost everywhere on [0, T ] and lim u(t)/t = u(0) = 0 is u(t) ≡ 0.
t∈0
260 7 Applications to Infinite Systems of Differential Equations
Class D. This class contains functions w: [o, T ]×R+ ∈ R+ such that w(t, 0) =
0, satisfying Carathéodory’s conditions and locally integrable (cf. class C ).
Moreover, the only continuous function on [0, T ] which satisfies the inequality
t
u(t) − u(t) ≤ w(s, s(s))ds
t
for any non-empty subset X of the ball B(x0 , r ) and for almost all t → [0, T ],
where β is the Hausdorff measure of non-compactness in the Banach space E
(cf. Chap. 5) and w is a Kamke comparison function from class E (cf. Exercise
12). Consider the function w: [0, T ] × R+ ∈ R+ defined by the equality
References
1. Persidskii, K.P.: Countable systems of differential equations and stability of their solutions.
Izv. Akad. Nauk Kazach. SSR 7, 52–71 (1959)
2. Persidskii, K.P.: Countable systems of differential and stability of their solutions III: Funda-
mental theorems on stability of countable many differential equations. Izv. Akad. Nauk Kazach.
SSR 9, 11–34 (1961)
3. Persidskii, K.P.: Infinite systems of differential equations. Izdat. NAUKA Kazah. SSR, Alma-
Ata, Differ. Equ. Nonlinear Spaces (1976)
4. Deimling, K.: Ordinary differential equations in banach spaces. Lect. Notes Math. 596.
Springer, Berlin (1977)
5. Voigt, A.: Line method approximations to the cauchy problem for nonlinear parabolic differ-
ential equations. Numer. Math. 23, 23–36 (1974)
6. Walter, W.: Differential and Integral Inequalities. Springer, Berlin (1970)
7. Bellman, R.: Methods in Nonlinear Analysis II. Academic Press, New York (1973)
8. Hille, E.: Pathology of infinite system of linear first order differential equations with constant
coefficients. Ann. Mat. Pura Appl. 55, 135–144 (1961)
9. Oguztöreli, M.N.: On the neutral equations of Cowan and Stein. Utilitas Math. 2, 305–315
(1972)
10. Zautukov, O.A.: Countable systems of differential equations and their applications. Diff. Uravn.
1, 162–170 (1965)
11. Zautukov, O.A., Valeev, K.G.: Infinite systems of differential equations. Izdat. “Nauka” Kazach.
SSR, Alma-Ata (1974)
12. Banaś, J., Krajewska, M.: On solutions of semilinear infinite system of differential equations.
Dyn. Syst. Appl. 22, 301–316 (2013)
262 7 Applications to Infinite Systems of Differential Equations
13. Banaś, J., Lecko, M.: Solvability of infinite systems of differential equations in Banach sequence
spaces. J. Comput. Appl. Math. 137, 363–375 (2001)
14. Banaś, J.: Applications of measures noncompactness to various problems. Zeszyty Nauk. Polit.
Rzeszow., Mat.-Fiz., z.s, Rzeszów (1987)
15. Deimling, K.: Nonlinear Functional Analysis. Springer, Berlin (1985)
16. Mursaleen, M., Alotaibi, A.: Infinite systems of differential equations in some BK spaces. Abst.
Appl. Anal. 2012, 20 pages (2012). Article ID 863483
17. Mursaleen, M., Mohiuddine, S.A.: Applications of measures of noncompactness to the infinite
system of differential equations in l p spaces. Nonlin. Anal. 75, 2111–2115 (2012)
18. Banaś, J., Lecko, M.: On solutions of an infinite system of differential equations. Dyn. Syst.
Appl. 11, 221–230 (2002)
19. Mursaleen, M.: Application of measure of noncompactness to infinite system of differential
equations. Can. Math. Bull. 56, 388–394 (2013)
20. Dieudonné, J.: Deux examples singuliér d’équations différentielles. Acta Sci. Math. Szeged
12, 38–40 (1950). Leopoldo Fejér et Frederico Riesz LXX annos natis dedicatus, par B
21. Kisyński, J.: Sur les équations différentielles dans les espaces de Banach. Bull. Acad. Polon.
Sci., Sér. Sci. Math. Astronom. Phys. 7, 381–385 (1959)
22. Olech, C.: On the existence and uniqueness of solutions of an ordinary differential equation in
the case of Banach space. Bull. Acad. Polon. Sci., Sér. Sci. Math. Astronom. Phys. 8, 667–675
(1960)
23. Ważewski, T.: Sur l’existence et l’unicité des intégrales des équations différentielles ordinaries
au cas de l’espace de Banach. Bull. Acad. Polon. Sci., Sér. Sci. Math. Astronom. Phys. 8,
301–305 (1960)
24. Banaś, J., Hajnosz, A., Wȩdrychowicz, S.: Relations among various criteria of uniqueness for
ordinary differential equations. Commentationes Math. Univ. Carol. 22, 59–70 (1981)
25. Ambrosetti, A.: Un teorem di esistenza per le equazioni differenziali negli spazi di Banach.
Rend. Sem. Mat. Univ. Padova 33, 349–361 (1967)
26. Goebel, K., Rzymowski, W.: An existence theorem for the equation x ∞ = f (t, x) in Banach
space. Bull. Acad. Polon. Sci., Sér. Sci. Math. Astronom. Phys. 18, 367–370 (1970)
27. Sadovskii, B.N.: Differential equations with uniformly continuous right hand side. Tr. Nauch.-
Issl. Inst. Mat. Voronez. Gos. Univ. 1, 128–136 (1970)
28. Szufla, S.: Measures of noncompactness and ordinary differential equations in Banach spaces.
Bull. Acad. Polon. Sci., Sér. Sci. math. Astronom. Phys. 19, 831–835 (1971)
29. Banaś, J., Goebel, K.: Measures of Noncompactness in Banach Spaces. Lecture Notes in Pure
and Applied Mathematics, vol. 60. Marcel Dekker, New York (1980)
30. Banaś, J., Hajnosz, A., Wȩdrychowicz, S.: On the equation x ∞ = f (t, x) in Banach spaces.
Comment. Math. Univ. Carolinae 23, 233–247 (1982)
31. Banaś, J., Hajnosz, A., Wȩdrychowicz, S.: On existence and local characterization of solutions
of ordinary differential equations in Banach spaces. Diff. Eqs. Appl., Proc. Second Conf.
Rousse, 55–58 (1982)
32. Banaś, J.: On existence theorems for differential equations in Banach spaces. Bull. Austral.
Math. Soc. 32, 73–82 (1985)
33. Bompiani, E.: Un teorema di confronto ed un teorema di unicita per l’equazioné differenziale
y ∞ = f (x, y). Atti Acad. Naz. Lincei Rend. Cl. Fis. Mat. Nat. 6, 298–302 (1925)
34. Coddington, E.A., Levinson, N.: Theory of Differential Equations. McGraw-Hill Book Co.,
New York (1955)
35. Olech, C.: Remarks concerning criteria for uniqueness of solutions of ordinary differential
equations, Bull. Acad. Polon. Sci., Sér. Sci. Math. Astronom. Phys. 8, 661–666 (1960)
36. Mönch, H., von Harten, G.H.: On the Cauchy problem for ordinary differential equations in
Banach spaces. Arch. Math. 39, 153–160 (1982)
37. Leung, K.V., Mangeron, D., Oguztöreli, M.N., Stein, R.B.: On the stability and numerical
solutions of two neutral models. Utilitas Math. 5, 167–217 (1974)
38. Banaś, J., Rzepka, B.: On existence and asymptotic behavior of solutions of infinite systems
of differential equations. Panamer. Math. J. 14, 105–115 (2004)
Chapter 8
Applications to Integral Equations
The aim of this chapter is to show how the technique associated with measures
of non-compactness can be applied in order to prove existence results concerning
integral equations of various types as well as infinite systems of those equations.
Apart from the mentioned technique of measures of non-compactness, we also intend
to apply completely continuous operators and Schauder fixed-point principle. In
fact, any completely continuous operator (i.e., continuous and compact) may be
considered as an operator satisfying the Darbo condition with the constant equal to
zero. This fact justifies our treating of the use of Schauder fixed-point principle as a
special case of Darbo fixed theorem formulated in Chap. 5 as Theorem 5.30.
The chapter is organized as follows: In Sect. 8.1, we will consider a quadratic
integral equation of Volterra type and we will investigate its solvability in the space
BC(R+ ) described in Sect. 5.8. The solvability of the equation in question will be
considered with the help of a measure of non-compactness which will enable us
to assert additionally the attractivity and asymptotic stability of its solutions. Next
section, Sect. 8.2 is devoted to the existence of solutions of infinite system of Volterra
integral equations.
Section 8.3 will treat on the solvability of an integral equation of Volterra–Stieltjes
type. Equations of such a type contain, as a special case, integral equations of
the so-called fractional order, which are recently very often considered in several
applications.
In the last Sect. 8.4, we will consider an infinite system of integral equations of
Volterra–Stieltjes type.
This chapter may also be considered as the challenge to consider infinite systems
of integral equations with the use of the technique of measures of non-compactness.
Let us pay attention to the fact that such an approach was not realized up to now, as
far as we know.
The considerations of the present chapter are based on the papers [1–4].
This section is devoted to the study of the solvability of the following quadratic
Volterra integral equation
t
x(t) = p(t) + f (t, x(t)) v(t, s, x(s))ds, t ∞ 0. (8.1)
0
Some particular cases of Eq. (8.1) were investigated in the vehicular traffic theory,
biology, and queuing theory and in the theories of radiative transfer and neutron
transport and in the kinetic theory of gases (cf. [5–8], for example).
We will study solvability of Eq. (8.1) in the space BC(R+ ) (cf. Sect. 5.8)
consisting of real functions defined, continuous and bounded on the interval R+ =
[0, ∈), and endowed with the supremum norm. In our considerations, we will use
the measure of non-compactness μc in the space BC(R+ ) defined in Sect. 5.8 by the
formula
μc (X ) = φ0 (X ) + c(X ) (8.2)
for X ∈ M BC(R+ ) , where the components φ0 (T ) and c(X ) are defined subsequently
in the below described way: For a fixed λ > 0 and T > 0 and for X ∈ Mc and
x ∈ X , we put:
To explain these ideas, assume that is a non-empty subset of the space BC(R+ )
and Q is an operator acting from into BC(R+ ). Let us consider the following
operator equation:
x(t) = (Qx)(t), t ∈ R+ . (8.3)
Definition 8.1 The solution x = x(t) of Eq. (8.3) is said to be globally attractive if
for each solution y = y(t) of Eq. (8.3), we have that
Definition 8.2 We say that solutions of Eq. (8.3) are locally attractive if there exists
a ball B(x0 , r ) in the space BC(R+ ) such that for arbitrary solutions x(t) and y(t)
of Eq. (8.3) belonging to B(x0 , r ) ∩ , condition (8.4) is satisfied.
In the case when the limit (8.4) is uniform with respect to the set B(x0 , r ) ∩ , i.e.,
when for each λ > 0, there exists T > 0 such that |x(t) − y(t)| → λ for all solutions
x(t), y(t) of Eq. (8.3) from the set B(x0 , r ) ∩ and for any t ∞ T , we say that
solutions of Eq. (8.3) are uniformly locally attractive ( asymptotically stable).
In the similar way, we can introduce the concept of uniform global attractivity
(cf. [2]).
In what follows, we will assume that functions involved in Eq. (8.1) satisfy the
below-listed assumptions:
(i) p ∈ BC(R+ ).
(ii) The function f : R+ × R ≤ R is continuous, and there exists a continuous
function m : R+ ≤ R+ such that
t
a(t) = m(t) h(t, s)ds,
0
266 8 Applications to Integral Equations
t
b(t) = | f (t, 0)| h(t, s)ds,
0
t
c(t) = m(t) |v(t, s, 0)|ds,
0
t
d(t) = | f (t, 0)| |v(t, s, 0)|ds,
0
are bounded on R+ , and the functions a(t) and b(t) vanish at infinity, i.e.,
limt≤∈ a(t) = limt≤∈ b(t) = 0.
Taking into account the above assumption, we may define the following finite
constants:
A = sup{a(t) : t ∈ R+ },
B = sup{b(t) : t ∈ R+ },
C = sup{c(t) : t ∈ R+ },
D = sup{d(t) : t ∈ R+ }.
|| p|| + Ar H (r ) + B H (r ) + Cr + D → r
Proof Let us consider the operator V defined on the space BC(R+ ) by the formula
t
(V x)(t) = p(t) + f (t, x(t)) v(t, s, x(s))ds, t ∈ R+ .
0
t
|(V x)(t)| → | p(t)| + [| f (t, x(t)) − f (t, 0)| + | f (t, 0)|] |v(t, s, x(s))|ds
0
t
→ | p(t)| + [m(t)|x(t)| + | f (t, 0)|] [|v(t, s, x(s)) − v(t, s, 0)| + |v(t, s, 0)|]s
0
t
→ | p(t)| + [m(t)|x(t)| + | f (t, 0)|] [h(t, s)H (|x(s)|) + |v(t, s, 0)|]ds
0
t t
→ || p|| + ||x||H (||x||)m(t) h(t, s)ds + H (||x||)| f (t, 0)| h(t, s)ds
0 0
t t
+ ||x||m(t) |v(t, s, 0)|ds + | f (t, 0)| |v(t, s, 0)|ds
0 0
= || p|| + ||x||H (||x||)a(t) + H (||x||)b(t) + ||x||c(t) + d(t)
→ || p|| + A||x||H (||x||) + B H (||x||) + C||x|| + D.
In view of the above estimate, we infer that the function V x is bounded on the
interval R+ . This implies that V x ∈ BC(R+ ), which means that the operator V
transforms the space BC(R+ ) into itself.
Apart from this, keeping in mind assumption (v), we deduce that there exists
a number r0 > 0 such that the operator V transforms the ball Br0 into itself and
k = AH (r0 ) + C < 1.
Now, let us take a non-empty subset X of the ball Br0 . Fix arbitrarily x, y ∈ X .
Then, for a fixed t ∈ R+ , we get
t t
|(V x)(t) − (V y)(t)| → f (t, x(t)) v(t, s, x(s))ds − f (t, y(t)) v(t, s, x(s))ds
0 0
t t
+ f (t, y(t)) v(t, s, x(s))ds − f (t, y(t)) v(t, s, y(s))ds
0 0
t
→ | f (t, s(t)) − f (t, y(t))| |v(t, s, x(s))|ds
0
t
t
+ | f (t, y(t))| v(t, s, x(s))ds − v(t, s, y(s))ds
0 0
t
→ m(t)|x(t) − y(t)| [|v(t, s, x(s)) − v(t, s, 0)| + |v(t, s, 0)|]ds
0
268 8 Applications to Integral Equations
t
+ [| f (t, y(t)) − f (t, 0)| + | f (t, 0)|] |v(t, s, x(s)) − v(t, s, y(s))|ds
0
t
→ m(t)|x(t) − y(t)| [h(t, s)H |x(s)| + |v(t, s, 0)|]ds
0
t
+ [m(t)|y(t)| + | f (t, 0)|] h(t, s)H (|x(s) − y(s)|)ds
0
t
→ m(t)|x(t)| + |y(t)| h(t, s)H (|x(s)|)ds + m(t)|x(t)
0
t t
− y(t)| |v(t, s, 0)|ds + m(t)|y(t)| h(t, s)H (|x(s)| + |y(s)|)ds
0 0
t
+ | f (t, 0)| h(t, s)H (|x(s)| + |y(s)|)ds
0
t t
→ 2r0 H (r0 )m(t) h(t, s)ds + m(t)|x(t) − y(t)| |v(t, s, 0)|ds
0 0
t t
+ r0 H (2r0 )m(t) h(t, s)ds + H (2r0 )| f (t, 0)| h(t, s)ds
0 0
= 2r0 H (r0 )a(t) + c(t)|x(t) − y(t)| + r0 H (2r0 )a(t) + H (2r0 )b(t).
(8.5)
Hence, we obtain
diam(V X )(t) → 2r0 H (r0 )a(t) + r0 H (2r0 )a(t) + H (2r0 )b(t) + c(t)diamX (t).
Joining the above estimate and assumption (iv), we derive the following inequality
→ φ T ( p, λ)
t
2
+ | f (t2 , x(t2 )) − f (t1 , x(t1 ))| h(t2 , s)H (|x(s)|)ds
0
t2
+ |v(t2 , s, 0)|ds
0
t
1
+ | f (t1 , x(t1 ))| |v(t2 , s, x(s)) − v(t1 , s, x(s))|ds
0
t2
+ |v(t2 , s, x(s))|ds
t1
→ φ T ( p, λ)
+ [| f (t2 , x(t2 )) − f (t2 , x(t1 ))| + | f (t2 , x(t1 ))
t2
− f (t1 , x(t1 ))|H (r0 )] h(t2 , s)ds
0
+ [| f (t2 , x(t2 )) − f (t2 , x(t1 ))| + | f (t2 , x(t1 ))
t2
− f (t1 , x(t1 ))|] |v(t2 , s, 0)|ds
0
t
1
+ [| f (t1 , x(t1 )) − f (t1 , 0)| + | f (t1 , 0)|] φ1T (v, λ, r0 )ds
0
270 8 Applications to Integral Equations
t2
+ |v(t2 , s, x(s))|ds
t1
t2
→ φ T ( p, λ) + [m(t2 )|x(t2 ) − x(t1 )| + φ1T ( f, λ)]H (r0 ) h(t2 , s)ds
0
t2
+ [m(t2 )|x(t2 ) − x(t1 )| + φ1T ( f, λ)] |v(t2 , s, 0)|ds
0
+ [m(t1 )|x(t1 )| + | f (t1 , 0)|]{T φ1T (v, λ, r0 ) + λv(T, r0 )}
t2
→ φ T ( p, λ) + H (r0 )φ T (x, λ)m(t2 ) h(t2 , s)ds
0
T
+ H (r0 )φ1T ( f, λ) sup h(t, s)ds : t ∈ [0, T ]
0
t2
+ φ T (x, λ)m(t2 ) |v(t2 , s, 0)|ds
0
T
+ φ1T ( f, λ) sup |v(t, s, 0)|ds : t ∈ [0, T ]
0
⎪
+ m(T )r0 + f (T ) T φ1T (v, λ, r0 ) + λv(T, r0 )
where we denoted
Keeping in mind that the function f = f (t, x) is uniformly continuous on the set
[0, T ] × [−r0 , r0 ] and the function v = v(t, s, x) is uniformly continuous on the set
[0, T ] × [0, T ] × [−r0 , r0 ], from estimate (8.7), we infer the following one:
Hence, we get
φ0 (V X ) → (AH (r0 ) + C)φ0 (X ) = kφ0 (X ). (8.8)
Now, taking into account formula (8.2) expressing the measure of non-
compactness defined above in the space BC(R+ ) and joining (8.6) and (8.8), we
obtain
μc (V X ) → kμc (X ). (8.9)
In what follows, let us put Br10 = CoV (Br0 ), Br20 = CoV (Br10 ), and so on. Consider
the sequence of sets (Brn0 ). Notice that Brn+1
0
⊂ Brn0 for all n = 1, 2, . . .. Moreover,
the sets of this sequence are non-empty, bounded, closed, and convex. Apart from
this, in view of (8.9), we get
for n = 1, 2, . . .. Since k < 1 [cf. assumption (v)], from (8.10), we deduce that
lim μc (Brn0 ) = 0. Hence, taking into account Definition 5.5, we deduce that the
n≤∈
set Y = ∈ n
n=1 Br0 is non-empty, bounded, closed, and convex. Moreover, the set Y
belongs to the kernel ker μc (see remarks made after Definition 5.5). Particularly, we
have that
lim sup diamY (t) = lim diamY (t) = 0. (8.11)
t≤∈ t≤∈
On the other hand, let us notice that the operator V maps the set Y into itself.
Now, we show that the operator V is continuous on the set Y . To this end, fix a
number λ > 0 and take arbitrary functions x, y ∈ Y such that ||x − y|| → λ. Using
(8.11) and taking into account the fact that V Y ⊂ Y , we deduce that there exists
T > 0 such that for an arbitrary t ∞ T , we obtain
272 8 Applications to Integral Equations
Further, let us choose t ∈ [0, T ]. Then, evaluating similarly as in the proof of estimate
(8.5) and keeping in mind assumptions (ii)–(iv), we get
t
|(V x)(t) − (V y)(t)| → m(t)|x(t) − y(t)| [h(t, s)H (|x(s)|) + |v(t, s, 0)|]ds
0
t
+ [m(t)|y(t)| + | f (t, 0)|] h(t, s)H (|x(s) − y(s)|)ds
0
t t
→ λH (r0 )m(t) h(t, s)ds + λm(t) |v(t, s, 0)ds|
0 0
t t
+ H (λ)r0 m(t) h(t, s)ds + H (λ)| f (t, 0)| h(t, s)ds
0 0
→ λH (r0 )a(t) + λc(t) + H (λ)r0 a(t) + H (λ)b(t)
→ λAH (r0 ) + λC + H (λ)r0 A + H (λ)B.
The above estimate in combination with estimate (8.12) allows us to conclude that
the operator V transforms continuously the set Y into itself.
Finally, let us observe that keeping in mind all above-established facts concerning
both the set Y and the operator V : Y ≤ Y , and applying the classical Schauder
fixed-point principle, we infer that V has at least one fixed point x belonging to the
set Y . Obviously, the function x = x(t) is a solution of Eq. (8.1). Apart from that,
taking into account the construction of the set Y and the fact that Y ∈ ker μc , we
conclude that solutions of Eq. (8.1) are asymptotically stable (cf. Definition 8.2).
This assertion is an immediate consequence of the fact that if x, y are solutions of
Eq. (8.1) belonging to the ball Br0 , then x, y ∈ Y .
This completes the proof.
x(t) = t exp(−2t)
t
1 √ 2(x(s))2/3 + x(s) 1
+ arctan( t + t x(t)) + ds.
2ε (s + 1)(t 2 + 1) 10(t 2 + 1)
0
(8.13)
8.1 The Existence and Attractivity of Solutions 273
p(t) = t exp(−2t),
1 √
f (t, x) = arctan( t + t x),
2ε
2x 2/3 + x 1
v(t, s, x) = + ,
(s + 1)(t + 1) 10(t + 1)
2 2
then we can check that the assumptions of Theorem 8.1 are satisfied. Indeed, √ we
have that || p|| = p(1/2) = (1/2) exp(−1) = 0.183939 . . ., f (t, 0) = 2ε 1
arctan t
and v(t, s, 0) = 10(t12 +1) . Further, we have that m(t) = 2ε
1
. Moreover, we obtain
1 2/3
|v(t, s, x) − v(t, s, y)| → 2x + x − 2y 2/3
− y
(s + 1)(t 2 + 1)
1 ⎪
→ 2(x − y)2/3 + |x − y| .
(s + 1)(t 2 + 1)
1
h(t, s) = , H (r ) = 2r 2/3 + r.
(s + 1)(t 2 + 1)
Further, we deduce that the function a(t) appearing in assumption (iv) has the fol-
lowing form:
t
t ln(t + 1)
a(t) = m(t) h(t, s) = .
2ε(t 2 + 1)
0
3 1
B→ · √ = 0.14246919 . . . .
16 4 3
274 8 Applications to Integral Equations
t
1 1 1 t2
c(t) = ds = .
2ε 10(t 2 + 1) 20ε t 2 + 1
0
Hence, C = 0.01591549 . . ..
Finally, we get √
t arctan t t
d(t) = → .
20ε(t + 1)
2 40(t + 1)
2
In this section, we are going to present an existence theorem for the infinite system
of nonlinear integral equations of Volterra type having the form:
t
xi (t) = ai (t) + f i (s, x1 (s), x2 (s), . . .)ds, (8.14)
0
||x||C = max{||x(t)|| E : t ∈ I }.
If X is a set in C(I, E), then for a fixed t ∈ I , we denote by X (t) the cross-section
set defined in the following way:
X (t) = {x(t) : x ∈ X }.
Further on, we will use the generalized theorem of Arzéla–Ascoli [9] containing a
criterion of compactness in the space C(I, E).
Theorem 8.2 A bounded subset X of the space C(I, E) is relatively compact if and
only if all functions belonging to X are equicontinuous on I and the set X (t) is
relatively compact in E for each t ∈ I .
In what follows, we will work in the Banach space c0 consisting of real sequences
converging to zero with the standard norm ||x||c0 = max{|xi | : i = 1, 2, . . .} for
x = (x1 , x2 , . . .) ∈ c0 . Let us recall (cf. [10] and Sect. 5.8) that a bounded subset X
of c0 is relatively compact if and only if
⎧ ⎨
lim sup {max{|xk | : k ∞ 1}} = 0.
i≤∈ x=(xi )∈X
Now, let us return to the infinite system of integral Eq. (8.14). This system will
be investigated under the following hypotheses:
(i) The functions f i are defined on the set I × R∈ and take real values (i =
1, 2, . . .). Moreover, the operator f defined on the space I × c0 in the following
way
(t, x) ≤ ( f x)(t) = ( f 1 (t, x), f 2 (t, x), . . .)
transforms the space I × c0 into c0 and is such that the family of functions
{( f x)(t)}t∈I is equicontinuous at every point of the space c0 .
(ii) There exist non-negative functions αi (t) and δi (t) defined, integrable, and uni-
⎩T
formly bounded on I and such ⎛⎩ that 0 δ1⎝(s)ds → m, where m ∈ (0, 1). More-
t
over, the function sequence 0 αi (s)ds converges monotonically to zero at
⎛⎩ ⎝
t
each point t ∈ I , while the function sequence 0 δi (s)ds is non-increasing
at each point t ∈ I and the following estimate is satisfied:
Remark 8.1 Notice that assumption (i) on the equicontinuity of the family of func-
tions {( f x)(t)}t∈I at each point of the space c0 means that for any x0 ∈ c0 and for
any arbitrarily fixed λ > 0, there exists θ > 0 such that ||( f x)(t) − ( f x0 )(t)||c0 → λ
for each t ∈ I and for each x ∈ c0 such that ||x − x0 ||c0 → θ (see [11]).
⎩t
Remark 8.2 Observe that the functions 0 αi (s)ds and |ai (t)| (i = 1, 2, . . .) appear-
ing in assumptions (ii) and (iii)
⎛⎩ are continuous
⎝ on I . Thus, in view of Dini theorem
t
[11], the function sequences 0 αi (s)ds and (|ai (t)|) converge uniformly on I to
the function vanishing identically on I .
Now, we are prepared to formulate our main result.
Theorem 8.3 Under assumptions (i)–(iii), the infinite system (8.14) has at least one
solution x(t) = (xi (t)) such that x(t) ∈ c0 for each t ∈ I .
Proof Denote by X 0 the subset of the space C = C(I, c0 ) consisting of all functions
x(t) = (xi (t)) such that
for i = 1, 2, . . . and t ∈ I , where u i (t) and vi (t) are defined in the following way:
⎩t
αi (s)ds
0
u i (t) = ,
⎩t
1− δi (s)ds
0
sup{|ai (s)| : 0 → s → t}
vi (t) = ,
⎩t
1 − δi (s)ds
0
for i = 1, 2, . . .. Observe that the functions u i (t) and vi (t) are non-decreasing on the
interval I . Moreover, from the assumptions, it follows that the function sequences
(u i (t)) and (vi (t)) converge uniformly on I to the function vanishing identically on
I (cf. Remark 8.2).
Further, let us consider the operator F defined on the space C(I, c0 ) as follows:
t
(F x)(t) = ((F x)i (t)) = ai (t) + f i (s, x1 (s), x2 (s), . . .)ds .
0
Notice that the operator F maps the set X 0 into itself. In fact, fix arbitrarily i and
x0 ∈ X 0 . Then, for k ∞ i, we have
8.2 An Infinite System of Integral Equations of Volterra Type in the Space c0 277
t
|(F x)k (t)| → |ak (t)| + f k (s, x1 (s), x2 (s), . . .)ds
0
t
→ |ai (t)| + [αk (s) + δk (s) · sup{|xn (s)| : n ∞ k}]ds
0
t t
→ |ai (t)| + αk (s)ds + δk (s)[u k (s) + vk (s)]ds
0 0
t t
→ |ai (t)| + αi (s)ds + δi (s)[u i (s) + vi (s)]ds
0 0
t t
→ sup{|ai (s)| : 0 → s → t} + αi (s)ds + (u i (t) + vi (t)) δi (s)ds
0 0
→ u i (t) + vi (t).
Now, we show that the operator F is continuous on the set X 0 . To this end, let us
fix arbitrarily λ > 0 and x0 ∈ X 0 . Next, choose θ = θ(x0 , λ) according to assumption
(i) (see Remark 8.1), i.e., for x ∈ X 0 such that ||x − x0 ||c0 → θ, we have
Taking into account that the function sequences (αi (t)), (δi (t)), (u i (t)), and (vi (t))
are uniformly bounded on I and the function sequence (ai (t)) is equicontinuous on
I , from the above estimate, we deduce that the set X 1 = F X 0 is equicontinuous
on I .
Further, let us consider the set Y = CoX 1 (i.e., the closed convex hull of the set
X 1 ). Obviously, Y is closed, bounded, and equicontinuous on I . Moreover, Y ⊂ X 0
and FY ⊂ Y . On the other hand, for x ∈ X 0 , we have
for each i ∞ n 0 and for each t ∈ I . Hence, applying the criterion for compactness in
the space c0 quoted after Theorem 8.2, we deduce that for any t ∈ I , the set X 1 (t)
is relatively compact in the space C(I, c0 ) (cf. Theorem 8.2). Apart from this, the
closedness of Y implies that it is compact. Finally, keeping in mind that F maps
continuously the set Y into itself, by the Schauder fixed-point principle, we infer
that the operator F has at least one fixed point in the set Y being a solution of our
problem.
Thus, the proof is complete.
In order to illustrate the above result, we provide an example.
Example 8.2 Consider the semilinear upper diagonal infinite system of integral equa-
tions of the form:
t ∈
xi (t) = ai (t) + Hi bi j (s)x j (s) ds, (8.15)
0 j=i
We will assume that functions involved in Eq. (8.15) satisfy the below-listed
hypotheses:
1. The function bi j (t) maps continuously the interval I into R (1 → i → j < ∈).
⎫
∈
2. The series bi j (t) is absolutely uniformly convergent on the interval I for any
j=i
i = 1, 2, . . .. Moreover, the function sequence (Bi (t)), where
∈
Bi (t) = |bi j (t)|
j=i
T
Bi (t)dt → Q,
0
|( f i x)(t) − ( f i y)(t)| → λ,
which implies the desired continuity. This shows that assumption (i) of Theorem 8.3
is satisfied.
Now, let us fix a natural number i and take t ∈ I and x = (xi ) ∈ c0 . Then, in
view of the assumed hypotheses, we get
∈
∈
| f i (t, x1 , x2 , . . .)| = Hi
bi j (t)x j → A bi j (t)x j
j=i j=i
∈
∈
→A |bi j (t)||x j | → A sup{|x j | : j ∞ i} |bi j (t)|
j=i j=i
= ABi (t) sup{|x j | : j ∞ i}.
Keeping in mind the above estimate, we see that assumption (ii) of Theorem 8.3 is
satisfied if we put αi (t) ≥ 0 and δi (t) = ABi (t) (i = 1, 1, . . .).
Indeed, based on assumptions 2 and 4, we obtain
T T
δi (t)dt = A Bi (t)dt → AQ < 1.
0 0
8.2 An Infinite System of Integral Equations of Volterra Type in the Space c0 281
This completes the proof of the announced assertion and shows that system (8.15)
has at least one solution in the space c0 .
It is worthwhile mentioning that the result contained in Theorem 8.3 can be applied
to the infinite system of differential equations of the form:
for i = 1, 2, . . . and t ∈ I .
In fact, under assumptions (i)–(iii) of Theorem 8.3, the initial valued problem
(8.16)–(8.17) has at least one solution x(t) = (x 1 (t), x2 (t), . . .) on the interval I
such that x(t) ∈ c0 for any t ∈ I .
Finally, let us observe that systems of such a type were investigated in many
papers and monographs (see [6, 12–18], for example).
Let us also mention that the result contained in this section is taken from the
paper [1].
Now, we intend to present an existence result for the nonlinear integral equation of
Volterra–Stieltjes type having the form:
t
x(t) = h(t) + f (t, x(t)) u(t, s, x(s))ds g(t, s), (8.18)
0
where t ∈ I = [0, 1]. Obviously, in our considerations, we may replace the interval
[0, 1] by any interval [a, b].
In order to explain all facts connected with Eq. (8.18), we provide firstly a few
auxiliary definitions and lemmas related mainly to the concept of bounded variation
and the Riemann–Stieltjes integral [19, 20].
At the beginning, take a real function x defined on the interval [a, b]. Then, the
symbol Vab x will denote the variation of the function x on the interval [a, b]. If
Vab x < ∈, we say that x is of bounded variation on [a, b]. If we have a function
q
u(t, s) = u : [a, b] × [c, d] ≤ R, then we denote by Vt= p u(t, s) the variation of
the function t ≤ u(t, s) on the interval [ p, q] ⊂ [a, b]. Similarly, we define the
q
quantity Vs= p u(t, s).
For the properties of functions of bounded variation, we refer to [19, 20].
282 8 Applications to Integral Equations
If x and ν are two real functional defined on the interval [a, b], then under some
additional conditions [19, 20], we can define the Stieltjes integral (in the Riemann–
Stieltjes sense)
b
x(t)dν(t)
a
of the function x with respect to the function ν. In this case, we say that x is Stieltjes
integrable on the interval [a, b] with respect to ν.
It is worthwhile mentioning that several conditions are known which guarantee
Stieltjes integrability [10, 19]. One of the most frequently used requires that x is
continuous and ν is of bounded variation on [a, b].
In what follows, we will utilize a few properties of the Stieltjes integral contained
in below-given lemmas [19, 20].
Lemma 8.4 If x is Stieltjes integrable on the interval [a, b] with respect to a function
ν of bounded variation, then
b ⎬ t ⎣
b ⎡
x(t)dν(t) → |x(t)|d ν .
a
a a
Lemma 8.5 Let x1 and x2 be Stieltjes integrable functions on the interval [a, b] with
respect to a non-decreasing function ν such that x1 (t) → x2 (t) for t ∈ [a, b]. Then,
b b
x1 (t)dν(t) → x2 (t)dν(t).
a a
b
x(s)ds g(t, s),
a
where g : [a, b] × [a, b] ≤ R and the symbol ds indicates the integration with
respect to s. Details concerning the integral of this type will be given later.
Now, let us recall (cf. Sect. 5.8) that if x : [a, b] ≤ R is a given function,
then the symbol φ(x, λ) will denote the modulus of continuity of x. Similarly, if
p(t, s) = p : [a, b] × [c, d] ≤ R, then we define modulus of continuity of the
function p(t, s) with respect to each variable separately. For example,
1
ι(X ) = φ0 (X ),
2
where φ0 (X ) = lim φ(X, λ) and φ(X, λ) = sup{φ(x, λ) : x ∈ X } for X ∈ MC[a,b] .
λ≤0
Next, we recall a few facts concerning the so-called superposition operator [21].
To this end, assume that I = [a, b] and f : I × R ≤ R is a given function. Then,
to every function x acting from I into R, we may assign the function F x defined by
the formula:
(F x)(t) = f (t, x(t)), t ∈ I.
The operator F defined in such a way is called the superposition operator generated
by the function f = f (t, x). The properties of the superposition operator may be
found in [21], but for our purposes, we will only need the following result concerning
the operator F in the space C(I ).
Lemma 8.6 The superposition operator F generated by the function f : I ×R ≤ R
maps the space C(I ) into itself and is continuous if and only if the function f is
continuous on the set I × R.
Finally, let us collect some basic facts concerning the fractional calculus [22, 23].
To do this, let us recall (cf. Sect. 5.5) that the symbol L 1 (a, b) denotes the space of
all real functions defined and Lebesgue integrable on the interval (a, b). This space
is equipped with the standard norm.
Next, fix a number ζ > 0 and take an arbitrary function x ∈ L 1 (a, b). The
Riemann–Liouville fractional integral of order ζ of the function x = x(t) is defined
by the formula:
t
ζ 1 x(s)
I x(t) = ds, t ∈ (a, b),
β(ζ) (t − s)1−ζ
a
χ = {(t, s) : 0 → s → t → 1}.
We formulate the assumptions under which we will consider Eq. (8.18). Namely,
we assume that the following conditions are satisfied:
(i) h ∈ C(I ), where I = [0, 1].
284 8 Applications to Integral Equations
⎡
t1
[g(t2 , s) − g(t1 , s)] → λ.
s=0
Indeed, this lemma is an easy consequence of assumptions (iii)–(v) and the properties
of the variation of functions (cf. [19, 20]).
Corollary 8.8 Let assumptions (iii)–(v) be satisfied. Then, for an arbitrarily fixed
number t2 ∈ I (t2 > 0) and for any λ > 0, there exists θ > 0 such that if t1 ∈ I ,
t1 < t2 , and t2 − 11 → θ; then,
⎡
t2
p≤ g(t2 , s) → λ.
s=t1
Proof Fix t2 ∈ I , t2 > 0, and λ > 0. Consider the function ξ defined on the interval
[0, t2 ] by the formula:
⎡p
ξ( p) = g(t2 , s).
s=0
8.3 Solvability of a Class of Nonlinear Integral Equations 285
Then, according to Lemma 8.7, the function ξ is continuous at the point t2 . Hence,
we deduce that there exists θ > 0 such that if t1 ∈ I , t1 < t2 , and t2 − t1 → θ, and
then, we have |ξ(t2 ) − ξ(t1 )| → λ. On the other hand, we have
t
⎡ 2 ⎡t1
|ξ(t2 ) − ξ(t1 )| = g(t2 , s) − g(t2 , s)
s=0 s=0
t
⎡ 1 ⎡t2 ⎡
t1
= g(t2 , s) + g(t2 , s) − g(t2 , s)
s=t
s=0 1 s=0
⎡
t2
= g(t2 , s) → λ.
s=t1
The next result describing the properties of the function g(t, s) is contained in the
following lemma:
Lemma 8.9 Under assumptions (iii)–(v), the function
⎡
t
t≤ g(t, s)
s=0
Proof Fix arbitrary t1 , t2 ∈ I , t1 < t2 . Then, keeping in mind the properties of the
variation, we get
t t
⎡ 2 ⎡t1 ⎡ 1 ⎡ t2 ⎡t1
g(t2 , s) − g(t1 , s) = g(t2 , s) + g(t2 , s) − g(t1 , s)
s=t1
s=0 s=0 s=0 s=0
t
⎡ 1 ⎡
t1 ⎡ t2
→ g(t2 , s) − g(t1 , s) + g(t2 , s). (8.19)
s=t1
s=0 s=0
On the other hand, in view of the fact that the variation is subadditive [19], we have
⎡
t1 ⎡
t1
g(t2 , s) = {[g(t2 , s) − g(t1 , s)] + g(t1 , s)}
s=0 s=0
⎡t1 ⎡
t1
→ [g(t2 , s) − g(t1 , s)] + g(t1 , s).
s=0 s=0
286 8 Applications to Integral Equations
This implies
⎡
t1 ⎡
t1 ⎡
t1
g(t2 , s) − g(t1 , s) → [g(t2 , s) − g(t1 , s)]. (8.20)
s=0 s=0 s=0
⎡
t1 ⎡
t1
g(t1 , s) = {[g(t1 , s) − g(t2 , s)] + g(t2 , s)}
s=0 s=0
⎡t1 ⎡
t1
→ [g(t1 , s) − g(t2 , s)] + g(t2 , s)
s=0 s=0
⎡t1 ⎡t1
= [g(t2 , s) − g(t1 , s)] + g(t2 , s).
s=0 s=0
This yields
⎡
t1 ⎡
t1 ⎡
t1
g(t1 , s) − g(t2 , s) → [g(t2 , s) − g(t1 , s)] (8.21)
s=0 s=0 s=0
Now, taking into account (8.19) and (8.22), we derive the following estimate:
t
⎡ 2 ⎡
t1 ⎡ t1 ⎡
t2
g(t2 , s) − g(t1 , s) → [g(t2 , s) − g(t1 , s)] + g(t2 , s). (8.23)
s=0 s=0 s=0 s=t1
Further, let us assume that t2 is fixed. Take an arbitrary number λ > 0, and choose
θ > 0 to the number λ/2 according to assumption (v) and Corollary 8.8. This means
that there exists θ > 0 such that if t2 − t1 → θ; then,
⎡
t1
λ
[g(t2 , s) − g(t1 , s)] →
2
s=0
and
⎡
t2
λ
g(t2 , s) → .
s=t1
2
8.3 Solvability of a Class of Nonlinear Integral Equations 287
Then, we are prepared to formulate the last assumption needed in our considerations:
(viii) There exists a positive solution r0 of the inequality
Now, let us fix arbitrarily λ > 0 and choose t1 , t2 ∈ I such that t1 < t2 and t2 −t1 → λ.
Then, taking into account our assumptions, for a fixed x ∈ C(I ), we obtain
t
2 t1
|(U x)(t2 ) − (U x)(t1 )| → u(t2 , s, x(s))ds g(t2 , s) − u(t2 , s, x(s))ds g(t2 , s)
0 0
t
1 t1
+ u(t2 , s, x(s))ds g(t2 , s) − (t1 , s, x(s))ds g(t2 , s)
0 0
t
1 t1
+ u(t1 , s, x(s))ds g(t2 , s) − u(t1 , s, x(s))ds g(t1 , s)
0 0
t2 ⎡s
→ |u(t2 , s, x(x))|ds g(t2 , p)
t1 p=0
t1 ⎡
s
+ |u(t2 , s, x(s)) − u(t1 , s, x(s))|ds g(t2 , p)
0 p=0
t1 ⎡
s
+ |u(t1 , s, x(s))|ds [g(t2 , p) − g(t1 , p)]
0 p=0
t2 ⎡
s
→ α(||x||) ds g(t2 , p)
t1 p=0
t1 ⎡
s
+ |u(t2 , s, x(s)) − u(t1 , s, x(s))|ds g(t2 , p)
0 p=0
t1 ⎡
s
+ α(||x||) ds [g(t2 , p) − g(t1 , p)]
0 p=0
⎤ ⎦
⎡
t2 ⎡
t1
→ α(||x||) g(t2 , s) − g(t2 , s)
s=0 s=0
⎡
t1 ⎡
t1
+ φ(λ) g(t2 , s) + α(||x||) [g(t2 , s) − g(t1 , s)]
s=0 s=0
⎡
t2 ⎡
t2
→ α(||x||) g(t2 , s) + φ(λ) g(t2 , s) + α(||x||)M(λ)
s=t1 s=0
8.3 Solvability of a Class of Nonlinear Integral Equations 289
⎡
t2
→ α(||x||) g(t2 , s) + K φ(λ) + α(||x||)M(λ), (8.24)
s=t1
where we denoted
Notice that in view of the uniform continuity of the function u on the set χ ×
[−||x||, ||x||], we have that φ(λ) ≤ 0 as λ ≤ 0. Combining this fact with Corollary
8.8 and the property of the function M(λ) mentioned previously, we deduce from
(8.24) that the function U x is continuous on the interval I .
On the other hand, the function F x is continuous on I which is an easy conse-
quence of Lemma 8.6.
Thus, keeping in mind the above-established facts, we conclude that the function
Qx is continuous on the interval I . In other words, the operator Q transforms the
space C(I ) into itself.
In the sequel, we prove that the operator Q is continuous on the space C(I ). To
this end, let us first observe that in view of the properties of the superposition operator
F (cf. Lemma 8.6), it is sufficient to show that U is continuous on C(I ). To do this,
fix λ > 0 and x ∈ C(I ). Next, take an arbitrary function y ∈ C(I ) with ||x − y|| → λ.
Then, in virtue of Lemma 8.4, for an arbitrarily fixed t ∈ I , we obtain
t ⎡
s
|(U x)(t) − (U y)(t)| → |u(t, s, x(s)) − u(t, s, y(s))|ds g(t, p) .
0 p=0
Hence, if we denote
P = ||x|| + λ,
φ P (u, λ) = sup{|u(t, s, q) − u(t, s, v)| : (t, s) ∈ χ, q, v ∈ [−P, P], |q − v| → λ},
t ⎬ ⎣
⎡
s
|(U x)(t) − (U y)(t)| → φ P (u, λ)ds g(t, z)
0 z=0
⎡
t
→ φ P (u, λ) g(t, s) → K φ P (u, λ).
s=0
Thus, in view of the uniform continuity of the function u on the set χ × [−P, P],
we deduce that U is continuous on the space C(I ).
290 8 Applications to Integral Equations
Further, fix arbitrarily x ∈ C(I ). Then, taking into account the imposed assump-
tions and applying Lemmas 8.4 and 8.5, for a fixed t ∈ I , we obtain
t ⎡
s
|(Qx)(t)| → |h(t)| + | f (t, x(t))| |u(t, s, x(s))|ds g(t, p)
0 p=0
t ⎡
s
→ ||h|| + [| f (t, x(t)) − f (t, 0)| + | f (t, 0)|] α(||x||)ds g(t, p)
0 p=0
⎡
t
→ ||h|| + (k||x|| + F1 )α(||x||) g(t, s).
s=0
Now, keeping in mind assumption (viii), we deduce that there exists a number r0 > 0
such that Q transforms the ball Br0 into itself and k K α(r0 ) < 1.
In what follows, let us take a non-empty subset X of the ball Br0 and x ∈ X .
Further, fix λ > 0 and take t1 , t2 ∈ I such that t1 , t2 ∈ I and t2 − t1 → λ. Then,
keeping in mind the previously obtained estimate (8.24), we get
|(Qx)(t2 ) − (Qx)(t1 )| → |h(t2 ) − h(t1 )| + |(F x)(t2 )(U x)(t2 ) − (F x)(t2 )(U x)(t1 )|
+ |(F x)(t2 )(U x)(t1 ) − (F x)(t1 )(U x)(t1 )|
→ φ(h, λ) + |(F x)(t2 )||(U x)(t2 ) − (U x)(t1 )|
+ |(U x)(t1 )||(F x)(t2 ) − (F x)(t1 )|
→ φ(h, λ) + [| f (t2 , x(t2 )) − f (t2 , 0)|
⎡
t2
+ | f (t2 , 0)|] α(||x||) g(t2 , s) + K φ(λ) + α(||x||)M(λ)
s=t1
t
1
+ u(t1 , s, x(s))ds g(t1 , s) {| f (t2 , x(t2 )) − f (t2 , x(t1 ))|
0
+ | f (t2 , x(t1 )) − f (t1 , x(t1 ))|}
⎡
t2
→ φ(h, λ) + (k||x|| + F1 ){α(||x||) g(t2 , s) + K φ(λ)
s=t1
+ α(||x||)M(λ)}
t1 ⎡
s
+ |u(t1 , s, x(s))|ds g(t1 , s) {k|x(t2 ) − x(t1 )|
0 p=0
+ φr10 ( f, λ)},
8.3 Solvability of a Class of Nonlinear Integral Equations 291
where we denoted
Hence, we have
⎧ ⎨
⎡
t2
φ(Qx, λ) → φ( f, λ) + (kr0 + F1 ) α(r0 ) g(t2 , s) + kφ(λ) + α(r0 )M(λ)
s=t1
Now, keeping in mind that φ(λ) ≤ 0 and M(λ) ≤ 0 as λ ≤ 0 and taking into
account Corollary 8.8 as well as the fact that the function f is uniformly continuous
on the set I × [−r0 , r0 ] , we deduce the following estimate:
φ0 (Q X ) → k K α(r0 )φ0 (X ).
From the above estimate, assumption (viii), and Theorem 5.30, we conclude that
there exists at least one point x ∈ Br0 such that the function x = x(t) is a solution
of Eq. (8.18). This completes the proof.
In what follows, we pay our attention to the discussion concerning assumption
(v) of the above-proved theorem. That assumption plays an essential role in our
considerations. First of all, we formulate a condition which is handy in applications
292 8 Applications to Integral Equations
and guarantees that the function g appearing in Eq. (8.18) satisfies the mentioned
assumption (v).
To this end, assume, as previously, that g(t, s) = g : χ ≤ R, where χ =
{(t, s) : 0 → s → t → 1}. Moreover, we will assume that the function g satisfied
assumption (vi) and the following one:
(v⇒ ) For arbitrary t1 , t2 ∈ I such that t1 < t2 , the function s ≤ g(t2 , s) − g(t1 , s)
is non-increasing on the interval [0, t1 ].
Now, we can formulate the following lemma:
Lemma 8.12 Under assumptions (v⇒ ) and (vi), for any fixed s ∈ I , the function
t ≤ g(t, s) is non-increasing on the interval [s, 1].
Proof Fix an arbitrary s ∈ I , s < 1, and take arbitrary t1 , t2 ∈ [s, 1] such that
t1 < t2 . Then, in view of assumption (v⇒ ), we get
g(t2 , s) − g(t1 , s) → 0,
In view of Lemma 8.12, the above inequality can be written equivalently in the form:
Now, suppose that t1 and t2 are fixed. Take a partition 0 = s0 < s1 < . . . < sn = t1
of the interval [0, t1 ]. Then, keeping in mind our assumption and Lemma 8.12, we
obtain
n
|[g(t2 , si ) − g(t1 , si )] − [g(t2 , si−1 ) − g(t1 , si−1 )]|
i=1
n
= {[g(t2 , si−1 ) − g(t1 , si−1 )] − [g(t2 , si ) − g(t1 , si )]}
i=1
= g(t1 , t1 ) − g(t2 , t1 ).
8.3 Solvability of a Class of Nonlinear Integral Equations 293
⎡
t1
[g(t2 , s) − g(t1 , s)] = g(t1 , t1 ) − g(t2 , t1 ).
s=0
t
f 1 (t, x(t)) u(t, s, x(s))
x(t) = h(t) + ds, (8.26)
β(ζ) (t − s)1−ζ
0
where t ∈ I = [0, 1] and ζ is a fixed number, ζ ∈ (0, 1). Apart from this, β(ζ)
denotes the gamma function.
Observe that Eq. (8.26) represents the singular integral equation of Volterra type.
Equations of such a kind were considered recently in numerous papers and mono-
graphs (cf. [22–34] and references therein). Those equations can be applied to several
real-world problems.
Now, we show that the functional integral equation of fractional order (8.26) can
be considered as a special case of the Volterra–Stieltjes functional integral Eq. (8.18)
considered precedingly.
To this end, considered the function g(t, s) = g : χ ≤ R defined by the formula:
1 ζ
g(t, s) = [t − (t − s)ζ ]. (8.27)
ζ
It is easily seen that this function satisfies assumptions (iii), (iv), (v⇒ ), and (vi) for-
mulated earlier. Indeed, observe that the fact that the function g satisfies assumptions
(iii) and (vi) is obvious.
Next, let us notice that
πg 1
= >0
πs (t − s)1−ζ
for 0 → s < t. This implies that the function s ≤ g(t, s) is increasing on the interval
[0, t] for each t ∈ I . Thus, the function g satisfies assumption (iv).
In order to show that g satisfies assumption (v⇒ ), let us fix arbitrarily t1 , t2 ∈ I
such that t1 < t2 . Consider the function
1 ζ
G(s) = g(t2 , s) − g(t1 , s) = [t − t1ζ − (t2 − s)ζ + (t1 − s)ζ ],
ζ 2
defined on the interval [0, t1 ]. Then, we have
294 8 Applications to Integral Equations
1 1
G ⇒ (s) = (t2 − s)ζ−1 − (t1 − s)ζ−1 = − .
(t2 − s)1−ζ (t1 − s)1−ζ
Hence, we see that G ⇒ (s) < 0 for s ∈ [0, t1 ), which implies that the function g(t, s)
satisfies assumption (v’).
Now, keeping in mind Theorem 8.13, we conclude that the function g defined by
the formula (8.27) satisfies assumptions (iii)–(vi) formulated in Theorem 8.11.
Further, let us observe that
1
ds g(t, s) = ds.
(t − s)1−ζ
t
f 1 (t, x(t))
x(t) = h(t) + u(t, s, x(s))dsg(t, s).
β(ζ)
0
Hence, putting f (t, x) = f 1 (t, x)/ β(ζ), we see that Eq. (8.26) is a particular case
of Eq. (8.18).
In order to formulate an existence theorem concerning Eq. (8.26), we estimate first
the constant K used in assumption (viii) of Theorem 8.11 and defined in Corollary
8.10. To do this, let us take into account the fact that the function s ≤ g(t, s) is
non-decreasing on the interval [0, t] for any fixed t ∈ I . Then, we get
⎡
t
1 ζ
g(t, s) = g(t, t) − g(t, 0) = g(t, t) = t .
ζ
s=0
Keeping in mind the above-established facts, we can formulate the following exis-
tence result concerning Eq. (8.26).
Theorem 8.14 Assume that the function h involved in Eq. (8.26) satisfies assumption
(i), the function f 1 = f 1 (t, x) satisfies assumption (ii), and the function u = u(t, s, x)
satisfies assumption (vii) of Theorem 8.11. Apart from this, we assume the following
hypothesis:
(viii⇒ ) There exists a positive solution r0 of the inequality
1
||h|| + (kr + F 1 )α(r ) → r
β(ζ + 1)
8.3 Solvability of a Class of Nonlinear Integral Equations 295
such that kα(r0 ) < ζ, where the constant F 1 is defined by the formula:
Then, there exists at least one solution x = x(t) of Eq. (8.26) belonging to the
space C(I ).
The proof is an easy consequence of the fact that Eq. (8.26) is a particular case
of Eq. (8.18) with the function g(t, s) defined by the formula (8.27). In fact, it is
sufficient to apply Theorem 8.11 and the above-established properties of the function
g(t, s), showing that it satisfies assumptions (iii)–(vi).
Now, we are going to illustrate our results by an example.
Example 8.3 Consider the following functional integral equation of fractional order
t √
−t sin(t 2 + x(t)) |x(s)|
x(t) = te + ds, (8.28)
β(2/3) (4 + t 2 + s 2 )(t − s)1/3
0
h(t) = te−t ,
f 1 (t, x) = sin(t 2 + x),
√
|x|
u(t, s, x) = ,
4 + t + s2
2
then we can easily verify that the assumptions of Theorem 8.14 are satisfied.
In fact, the function h is continuous on I and ||h|| = 1/e . Thus, assumption (i)
is satisfied. Further notice that f 1 is continuous on I × R and satisfies the Lipschitz
condition with the constant k = 1. Moreover, F 1 = sin 1. Next, let us note that the
function u = u(t, s, x) is continuous on the set χ × R and the following inequality
is satisfied: √
|x| 1⎢
|u(t, s, x)| → → |x|.
4+t +s
2 2 4
√
This implies that the estimate from assumption (vii) is satisfied with α(r ) = 41 r .
Finally, let us pay attention to the fact that the first inequality from assumption
(viii⇒ ) of Theorem 8.14 has the form:
1 1 1√
+ (r + sin 1) r → r. (8.29)
e β(5/3) 4
296 8 Applications to Integral Equations
Keeping in mind that β(5/3) > 0.8856 [35] and sin 1 = 0.8415 . . . , it is easily
seen that the number r0 = 1 satisfies inequality (8.29). Moreover, the inequality
kα(r0 ) < ζ from assumption (viii⇒ ) is also satisfied since kα(r0 ) = 41 < 23 = ζ.
Thus, based on Theorem 8.14, we deduce that the functional integral equation of
fractional order (8.28) has at least one solution belonging to the ball B1 of the space
C(I ).
It is worthwhile mentioning that our result contained in Theorem 8.11 and adopted
with the help of Theorem 8.14 to the functional integral equation of fractional order
of the type (8.26) can also be applied to the other class of functional integral equations
(cf. Exercise 6).
The approach applied in Sects. 8.2 and 8.3 can be transferred to considerations
connected with the solvability of the following infinite system of nonlinear integral
equation of Volterra–Stieltjes type
t
xi (t) = pi (t) + vi (t, s, x1 (s), x2 (s), . . .)ds gi (t, s) (8.30)
0
for all x ∈ S.
Keeping in mind the above-formulated assumption (iii), we can define the function
N S = N S (λ) by putting
⎡
t1
[gi (t2 , s) − gi (t1 , s)] → λ
s=0
for all i = 1, 2, . . ..
(vii) gi (t, 0) = 0 for t ∈ I and i = 1, 2, . . ..
Before formulating the remainder of our assumptions, we pay our attention to
some consequences of assumptions (iv)–(vi). First of all, let us observe that based
on assumption (vi), we may define the function M : R+ ≤ R+ by the following
formula (cf. the proof of Theorem 8.11):
⎧
⎡
t1
M(λ) = sup [gi (t2 , s) − gi (t1 , s)] : t1 , t2 ∈ I, t1 < t2 ,
s=0
⎥
t2 − t1 → λ, i = 1, 2, . . . . (8.32)
Lemma 8.15 Let assumptions (iv)–(vi) be satisfied. Then, for any λ > 0, there exists
θ > 0 such that if t1 , t2 ∈ I , t1 , t2 , and t2 − t1 → θ; then,
⎡
t2
gi (t2 , s) → λ
s=t1
for all i = 1, 2, . . ..
Based on the above lemma, we can define the function G = G(λ) by putting
⎧ ⎨
⎡
t2
G(λ) = sup gi (t2 , s) : t1 , t2 ∈ I, t1 < t2 , t2 − t1 → λ, i = 1, 2, . . . .
s=t1
(8.33)
Obviously, we have that G(λ) ≤ 0 as λ ≤ 0.
Lemma 8.16
# Under assumptions (iv)–(vi), for any fixed natural number i, the func-
tion t ≤ ts=0 (t, s) is continuous on the interval I .
Now, taking into account Lemma 8.16, we infer that there exists a finite positive
constant K i (i = 1, 2, . . .) such that
⎧ ⎨
⎡
t
K i = sup gi (t, s) : t ∈ I . (8.34)
s=0
Our next and simultaneously final assumptions needed in our investigations are
formulated below.
(ix) There exist non-negative functions ζi = ζi (t, s) and αi = αi (t, s) defined
on the triangle χ for each i = 1, 2, . . . such that the sequences (ζi (t, s)) and
(αi (t, s)) are equibounded on χ, and for each fixed t ∈ I , there exist the
integrals
t t
ζi (t, s)ds gi (t, s), αi (t, s)ds gi (t, s)
0 0
8.4 Solvability of an Infinite System 299
⎩t ⎛# ⎝
s
Remark 8.3 Notice that the functions | pi (t)| and 0 ζi (t, s)ds p=0 gi (t, p) (i =
1, 2, . . .) appearing in assumptions (i) and (x), respectively, are continuous on I (cf.
[20]).
⎛⎩ Thus, in⎛ view of Dini ⎝⎝ theorem [36], the function sequences (| pi (t)|) and
t #s
0 ζi (t, s)ds p=0 gi (t, p) converge uniformly on I to the function vanishing
identically on I .
Remark 8.4 Observe that assumption (ii) on the equicontinuity of the family of
functions {(V x)(t, s)}(t,s)∈χ in the space c0 means that for any x0 ∈ c0 and for any
fixed λ > 0, there exists θ > 0 such that ||(V x)(t, s) − (V x0 )(t, s)||c0 → λ for each
(t, s) ∈ χ and for each x0 ∈ c0 such that ||x − x0 ||c0 → θ (see [11]).
Now, we can formulate our main result.
Theorem 8.17 Under assumptions (i)–(x), the infinite system of Volterra–Stieltjes
integral Eq. (8.18) has at least one solution x(t) = (xi (t)) such that x(t) ∈ c0 for
each t ∈ I .
Proof Denote by X 0 the subset of the space C = C(I, c0 ) consisting of all functions
x(t) = (xi (t)) such that
for i = 1, 2, . . . and t ∈ I , where the functions u i (t) and wi (t) are defined on the
interval I in the following way:
300 8 Applications to Integral Equations
⎬ ⎣
⎩t #
s
ζi (t, s)ds gi (t, p)
0 p=0
u i (t) = ⎬ ⎣,
⎩t #
s
1− αi (t, s)ds gi (t, p)
0 p=0
sup{| pi (s)| : 0 → s → t}
wi (t) = ⎬ ⎣
⎩t #s
1 − αi (t, s)ds gi (t, p)
0 p=0
for i = 1, 2, . . .. Observe that the functions u i (t) and wi (t) are non-decreasing on the
interval I . Apart from this, in virtue of assumptions (i), (ix), and (x), we deduce that
the function sequences (u i (t)) and (wi (t)) converge uniformly on I to the function
vanishing identically on I (cf. Remark 8.3). By these regards, we can define the finite
constants U and W by putting
Now, let us consider the operator F defined on the space C(I, c0 ) in the following
way:
t
(F x)(t) = ((Fi x)(t)) = pi (t) + vi (t, s, x1 (s), x2 (s), . . .)ds gi (t, s) .
0
Observe that the operator F maps the set X 0 into itself. In fact, to prove this assertion,
let us fix arbitrarily a natural number i and x ∈ X 0 . Then, for k ∞ i, in view of
Lemmas 8.4 and 8.5, we get
t
|(Fk x)(t)| → | pk (t)| + vk (t, s, x 1 (s), x2 (s), . . .)ds gk (t, s)
0
t
⎡s
→ | pi (t)| + |vk (t, s, x 1 (s), x2 (s), . . .)|ds gk (t, p)
0 p=0
t ⎡
s
→ | pi (t)| + [ζk (t, s) + αk (t, s) sup{|x n (s)| : n ∞ k}]ds gk (t, p)
0 p=0
t ⎡
s
→ | pi (t)| + ζk (t, s)ds gk (t, p)
0 p=0
8.4 Solvability of an Infinite System 301
t ⎡
s
+ αk (t, s)[u k (s) + wk (s)]ds gk (t, p)
0 p=0
t ⎡
s
→ | pi (t)| + ζi (t, s)ds gi (t, p)
0 p=0
t ⎡
s
+ [u i (t) + wi (t)] αi (t, s)ds gi (t, p)
0 p=0
t ⎡
s
→ sup{| pi (s)| : 0 → s → t} + ζi (t, s)ds gi (t, p)
0 p=0
t ⎡
s
+ [u i (t) + wi (t)] αi (t, s)ds gi (t, p) → u i (t) + wi (t).
0 p=0
Hence, in view of Lemmas 8.4 and 8.5, we derive the following estimates:
t
||(F x)(t)−(F x0 )(t)||c0 → max |vi (t, s, x1 (s), x2 (s), . . .)
0
⎡
s
− vi (t, s, x10 (s), x20 (s), . . .)|ds gi (t, p) : i = 1, 2, . . .
p=0
302 8 Applications to Integral Equations
t
⎡
s
→ sup λds gi (t, p) : i = 1, 2, . . .
0 p=0
⎧ ⎨
⎡
t
→ λ sup gi (t, s) : i = 1, 2, . . . = λK ,
s=0
where K is the constant defined by (8.35) (cf. also Lemma 8.16 and (8.34)).
The above estimate proves the desired assertion.
In the sequel, let us consider the set X 1 = F X 0 . We shows that the set X 1 consists
of functions being equicontinuous on the interval I . To do this, let us fix arbitrarily
x = (xi ) ∈ X 0 and a number λ > 0. Next, choose arbitrary t, s ∈ I such that
|t − s| → λ. Without the loss of generality, we may assume that s < t. Then, for a
fixed natural i, we obtain
⎡
t ⎡
t ⎡
t
=ζ gi (t, p) + α(U + W ) gi (t, p) = [ζ + α(U + W )] gi (t, p),
p=s p=s p=s
where the finite constants ζ, α, U, and W were defined by (8.36) and (8.37).
Hence, taking into account Lemma 8.15 and the function G(λ) defined by (8.33),
we obtain the following inequality:
t
s
vi (t, β , x1 (β ), x2 (β ), . . .)dβ gi (t, β ) − vi (t, β , x1 (β ), x2 (β ), . . .)dβ gi (t, β )
0 0
→ [ζ + α(U + W )]G(λ). (8.39)
Next, taking again into account Lemmas 8.4 and 8.5, we deduce the following esti-
mates:
s
s
vi (t, β , x1 (β ), x2 (β ), . . .)dβ gi (t, β ) − vi (t, β , x1 (β ), x2 (β ), . . .)dβ gi (s, β )
0 0
s
= vi (t, β , x1 (β ), x2 (β ), . . .)dβ [gi (t, β ) − gi (s, β )]
0
s ⎡ β
→ |vi (t, β , x1 (β ), x2 (β ), . . .)|dβ [gi (t, p) − gi (s, p)]
0 p=0
s β
⎡
→ [ζi (t, β ) + αi (t, β ) sup{|xk (β )| : k ∞ i}]dβ [gi (t, p) − gi (s, p)]
0 p=0
s β
⎡
→ [ζi (t, β ) + αi (t, β )(u i (β ) + wi (β ))]dβ [gi (t, p) − gi (s, p)]
0 p=0
304 8 Applications to Integral Equations
s β
⎡
→ [ζ + α(U + W )] dβ [gi (t, p) − gi (s, p)]
0 p=0
⎡
s
→ [ζ + α(U + W )] [gi (t, p) − gi (s, p)] = [ζ + α(U + W )]M(λ), (8.40)
p=0
where the function N X 0 (λ) was defined by (8.31) and K is the constant defined by
(8.34).
Finally, combining estimates (8.39)–(8.41) and taking into account inequality
(8.38), we conclude that the set X 1 consists of functions being equicontinuous on
the interval I .
Further on, we consider the set Y = CoX 1 , i.e., the closed convex hull of the set
X 1 . Notice that Y is closed, convex, and bounded and consists of functions being
equicontinuous on I . Apart from this, we have that Y ⊂ X 0 and FY ⊂ Y . On the
other hand, for arbitrarily chosen x ∈ X 0 , we have
(i = 1, 2, . . .). Keeping in mind that the sequence (u i (t) + wi (t)) converges uni-
formly on the interval I to the function vanishing identically on I , we deduce that
for every λ > 0, there exists a natural number n 0 such that |(Fi x)(t)| → λ for any
i ∞ n 0 and t ∈ I . This statement and the criterion of compactness in the space c0
quoted in Sect. 8.2 allow us to infer that the set X 1 (t) is relatively compact in the
space c0 for each fixed t ∈ I . Gathering all the above-established facts, we obtain
in light of Theorem 8.2 that the set Y is relatively compact in the space C(I, c0 ).
Moreover, the set Y is compact since it is closed.
Finally, taking into account the fact that the operator F maps continuously the set
Y into itself, by the Schauder fixed-point principle, we conclude that the operator F
8.4 Solvability of an Infinite System 305
has a fixed point in the set Y which is a solution of Eq. (8.30). Thus, the proof is
complete.
Now, we present an application of the result contained in Theorem 8.17 to the
infinite system of nonlinear Volterra integral equations of fractional orders having
the form:
t
1 vi (t, s, x1 (s), x2 (s), . . .)
xi (t) = pi (t) + ds, (8.42)
β(δi ) (t − s)1−δi
0
1 δi
gi (t, s) = [t − (t − s)δi ].
δi
πgi 1
= >0
πs (t − s)1−δi
for 0 → s < t. This implies that the function s ≤ gi (t, s) is increasing on the interval
[0, t] for each fixed t ∈ I . This means that the function gi satisfies assumption (v)
for i = 1, 2, . . .. Moreover, we get
⎡
t
1 δi
gi (t, s) = gi (t, t) − gi (t, 0) = gi (t, t) = t .
δi
s=0
Hence, we deduce that in this case, the constant K i defined by formula (8.34) has
the form: ⎧ t ⎨
⎡ 1
K i = sup gi (t, s) : t ∈ I = (8.43)
δi
s=0
306 8 Applications to Integral Equations
1 1
G i⇒ (s) = (t2 − s)δi −1 − (t1 − s)δi −1 = − .
(t2 − s)1−δi (t1 − s)1−δi
Hence, we see that G i⇒ (s) < 0 for s ∈ [0, t1 ). This implies that the function G i (s) is
decreasing on the interval [0, t1 ] for any i = 1, 2, . . ..
Now, taking into account the above-established fact, we get
⎡
t1
[gi (t2 , s) − gi (t1 , s)] = [gi (t2 , 0) − gi (t1 , 0)] − [gi (t2 , t1 ) − gi (t1 , t1 )]
s=0
1 δ δ
= gi (t1 , t1 ) − gi (t2 , t1 ) = [(t2 − t1 )δi + t1 i − t2 i ]. (8.45)
δi
Hence, keeping in mind the definition of the function M(λ) given by formula (8.32),
in view of (8.45), we obtain
$ ⎥
1 δ δ
M(λ) = sup [(t2 − t1 )δi + t1 i − t2 i ] : t1 , t2 ∈ I, t1 < t2 , t2 − t1 → λ, i = 1, 2, . . .
δi
$ ⎥
1 δi
= sup [λ + 1 − (1 + λ)δi ] : i = 1, 2, . . .
δi
1 δ⊂
= [λ + 1 − (1 + λ)δ⊂ ].
δ⊂
From the above equality, we conclude that the functions g = gi (t, s) satisfy assump-
tion (vi) of Theorem 8.17.
We omit the simple verification of assumption (vii). Obviously, assumption (viii)
is also satisfied, which is a simple consequence of (8.44).
Finally, let us observe that
ds
ds gi (t, s) =
(t − s)1−δi
8.4 Solvability of an Infinite System 307
t
1
xi (t) = pi (t) + vi (t, s, x1 (s), x2 (s), . . .)ds gi (t, s) (8.46)
β(δi )
0
for i = 1, 2, . . ..
Exercises
1. Show that the following quadratic integral equation of Volterra type
t ⎢ ⎪
1 x(t)
x(t) = + exp(s − t) |x(s)| + exp(−x 2 (s)) + exp(t − s − 1) ds
t 4 + 30 t + 1
0
has a solution in the space BC(R+ ) belonging to the ball B1/2 . Moreover, show
that solutions of this equation are asymptotically stable.
308 8 Applications to Integral Equations
t
x(t) = p(t) + v(t, s, x(s))ds, t ∞ 0.
0
Prove that under assumption (i) of Theorem 8.1 and under assumption that v :
R+ × R+ × R ≤ R is continuous and satisfies the conditions given in Exercise
2, the above integral equation is solvable in the space BC(R+ ) and solutions of
this equation are asymptotically stable.
4. Assume that the function g = g(t, s) satisfies assumption (vi) of Theorem 8.11
and the following one:
(v⇒⇒ ) For arbitrarily fixed t1 , t2 ∈ I such that t1 < t2 , the function s ≤ g(t2 , s) −
g(t1 , s) is non-decreasing on the interval [0, t1 ].
Prove that then, for each fixed s ∈ I = [0, 1], the function t ≤ g(t, s) is
non-decreasing on the interval [s, 1].
5. Assume that the function g(t, s) = g : χ × R ≤ R (χ = {(t, s) : 0 → s → t →
1}) satisfies the assumptions (iii) and (vi) of Theorem 8.11 and assumption (v”)
from Exercise 4. Prove that g satisfies assumptions (v) of Theorem 8.11.
6. Applying Theorem 8.11 and the results contained in Exercises 4 and 5, we show
that the functional integral equation of the form
t
t
x(t) = h(t) + f (t, x(t)) u(t, s, x(s))ds, t ∈ I = [0, 1],
t +s
0
Hint. Show that the above equation is a special case of Eq. (8.18) if we put
$
t ln t+s
for 0 < s → t → 1
g(t, s) = t
0 for t = 0.
References 309
References
1. Banaś, J., Lecko, M.: An existence theorem for a class of infinite systems of integral equations.
Math. Comput. Model. 34, 533–539 (2001)
2. Banaś, J., Rzepka, B.: On local attractivity and asymptotic stability of solutions of a quadratic
Volterra integral equation. Appl. Math. Comput. 213, 102–111 (2009)
3. Banaś, J., Rzepka, B.: The technique of Volterra-Stieltjes integral equations in the application
to infinite systems of nonlinear integral equations of fractional order. Comput. Math. Appl.
64, 3108–3116 (2012)
4. Banaś, J., Zaja̧c, T.: A new approach to the theory of functional integral equations of fractional
order. J. Math. Anal. Appl. 375, 375–387 (2011)
5. Chandrasekhar, S.: Radiative Transfer. Dover, New York (1960)
6. Deimling, K.: Nonlinear Functional Analysis. Springer, Berlin (1985)
7. Hu, S., Khavanin, M., Zhuang, W.: Integral equations arising in the kinetic theory of gases.
Appl. Anal. 34, 261–266 (1989)
8. Kelly, C.T.: Approximation of solutions of some quadratic integral equations in transport
theory. J. Integral Equations 4, 221–237 (1982)
9. Banaś, J., Goebel, K.: Measures of Noncompactness in Banach Spaces. Lecture Notes in Pure
and Applied Mathematics, vol. 60. Marcel Dekker, New York (1980)
10. Dunford, D., Schwartz, J.T.: Linear Operators I. Interscience Publishers, Leyden (1963)
11. Sikorski, R.: Real Functions. PWN, Warsaw (1958). (in Polish)
12. Deimling, K.: Ordinary Differential Equations in Banach Spaces. Lect. Notes in Math. 596,
Springer, Berlin (1977)
13. Mlak, W., Olech, C.: Integration of infinite system of differential inequalities. Ann. Polon.
Math. 13, 105–112 (1968)
14. Moszyński, K., Pokrzywa, A.: Sur les systémes infinis d’équations différentielles ordinaires
dans certain espaces de Fréchet. Dissert. Math. 115, 29 (1974)
15. Persidskii, K.P.: Countable systems of differential equations and stability of their solutions.
Izv. Akad. Nauk Kazach. SSR 7, 52–71 (1959)
16. Persidskii, K.P.: Countable systems of differential and stability of their solutions III: Fun-
damental theorems on stability of countable many differential equations. Izv. Akad. Nauk
Kazach. SSR 9, 11–34 (1961)
17. Rzepecki, B.: On infinite systems of differential equations with deviated argument II. Ann.
Polon. Math. 34, 251–264 (1977)
18. Zautukov, O.A.: Countable systems of differential equations and their applications. Diff.
Uravn. 1, 162–170 (1965)
19. Appell, J., Banaś, J., Merentes, N.: Bounded Variation and Around, De Gruyter Series in
Nonlinear Analysis and Applications 17. Walter de Gruyter, Berlin (2014)
20. Natanson, I.P.: Theory of Functions of a Real Variable. Ungar, New York (1960)
21. Appell, J., Zabrejko, P.P.: Nonlinear Superposition Operators. Cambridge University Press,
Cambridge (1990)
22. Kilbas, A.A., Srivastava, H.M., Trujillo, J.J.: Theory and Applications of Fractional Differ-
ential Equations. Elsevier Science B.V, Amsterdam (2006)
23. Miller, K.S., Ross, B.: An Introduction to the Fractional Calculus and Differential Equations.
John Wiley, New York (1992)
24. Srivastava, H.M., Saxena, R.K.: Operators of fractional integration and their applications.
Appl. Math. Comput. 118, 1–52 (2001)
25. Benchohra, M., Graef, J.R., Hamani, S.: Existence results for boundary value problems with
nonlinear fractional differential equations. Appl. Anal. 87, 851–863 (2008)
26. Cichoń, M., El-Sayed, A.M.A., Salem, H.A.H.: Existence theorems for nonlinear integral
equations of fractional orders. Comment. Math. 41, 59–67 (2001)
27. Darwish, M.A.: On quadratic integral equation of fractional orders. J. Math. Anal. Appl. 311,
112–119 (2005)
310 8 Applications to Integral Equations
28. Darwish, M.A., Henderson, J.: Existence and asymptotic stability of solutions of a perturbed
quadratic fractional integral equations. Fract. Calc. Appl. Anal. 12, 71–86 (2009)
29. Garth, M., Rao, A., Kalla, S.I.: Fractional generalization of temperature fields problem in oil
strata. Mat. Bilten 30, 71–84 (2006)
30. Hilfer, R.: Applications of Fractional Calculus in Physics. World Scientific, Singapore (2000)
31. Podlubny, I.: Fractional Differential Equations. Academic Press, San Diego (1999)
32. Samko, S., Kilbas, A.A., Marichev, O.: Fractional Integrals and Derivatives: Theory and
Applications. Gordon and Breach, Amsterdam (1993)
33. Saxena, R.K., Kalla, S.I.: On a fractional generalization of tree electron laser equation. Appl.
Math. Comput. 143, 89–97 (2003)
34. Saxena, R.K., Mathai, A.M., Haubold, A.L.: On generalized fractional kinetic equations.
Physica A 344, 657–664 (2004)
35. Fichtenholz, G.M.: Differential and Integral Calculus II. PWN, Warsaw (1980). (in Polish)
36. Lojasiewicz, S.: An Introduction to the Theory of real Functions. John Wiley, Chichester
(1988)
37. Rzepka, B., Sadarangani, K.: On solutions of an infinite system of singular integral equations.
Math. Comput. Modelling 34, 533–539 (2001)
Appendix
In this section, we state well-known theorems which are frequently used in this book
and can be found in any standard book of Functional Analysis (e.g., [1–3]).
Closed Graph Lemma. Any continuous map into a Hausdorff space has closed
graph.
Open Mapping Theorem. Let X and Y are Banach spaces, and suppose that
T ∞ B(X, Y ) is surjective. Then T is an open mapping.
References
1. Kreyszig, E.: Introductory Functional Analysis with Applications. Wiley, New York (1978)
2. Maddox, I.J.: Maddox, Elements of Functional Analysis, 2nd edn. The University Press, Cam-
bridge (1988)
3. Wilansky, A.: Functional Analysis. Blaisdell, New York (1964)
B
D
λ-bounded sequence, 73
Darbo fixed point theorem, 177
Balanced set, 10
Domain
BK space, 12
convergence, 19
Borel matrix, 70
matrix, 19
Bounded variation, 281
strongly matrix, 19
Dual
functional, 47
C symmetric, 47
λ-convergent sequence, 73
Carathéodory’s conditions, 259
Cauchy problem, 220 E
Cesàro matrix, 24 Echelon space, 111
Characteristic of a matrix, 63 β-net, 147
Charge, 37 Euler matrix, 24
Charged space, 37
Co-echelon space, 111
Compact matrix operator on BK space , 190 F
Compact matrix operators on some BK FK space, 4
spaces, 189 Fréchet combination of a sequence of para-
Compact operator, 148 norms, 6
M
G χ matrix, 26
Generalization of Cantor intersection theo- Maddox sequence spaces, 105
rem, 151 Matrix
Generalized mean matrix, 26 associated, 112
Globally attractive solution, 265 Borel, 70
Cesàro, 24
conservative, 57
H conull, 63
Hamel basis, 12 coregular, 63
Hausdorff distance, 148 Euler, 24
Hausdorff matrix, 23 Hausdorff, 23
Hausdorff measure of noncompactness, 150, infinite triangle, 19
154, 161, 162 Köthe, 111
Nörlund, 25
regular, 63
I Riesz, 25
Infinite system of differential equations, 230, Schur, 57
232, 234, 242 Toeplitz, 60
Infinite system of nonlinear integral equa- Matrix domain, 19
tions of Volterra type, 274 Matrix mapping, 33
Infinite system of nonlinear integral equa- Matrix transformation between spaces, 130
tions of Volterre–Stieltjes type, Measure of noncompactness
296 Hausdorff, 154, 161–163
Infinite system of nonlinear Volterra inte- inner Hausdorff, 166
gral equations of fractional or- Istrǎţescu’, 168
ders, 305 Kuratowski, 150
Infinite triangle matrix, 19 regular, 171
Inner Hausdorff measure of noncompact- Measure of noncompactness (axiomatic de-
ness, 166 finition), 169, 170
Integral equation of fractional order, 293, Measure of noncompactness of an operator,
295 175
Modulus of continuity, 173, 221
Monotone norm, 20, 160
K Monotone norm on a sequence, 160
Köthe–Toeplitz dual, 40 Monotone sequence space, 45
Kamke comparison function, 221, 223, 259
Kernel of a measure of noncompactness, 170
Kernel set, 224 N
Kojima–Schur theorem, 60 Nadler fixed point theorem, 181
Köthe matrix, 111 Nagumo uniqueness criterion, 260
Kuratowski measure of noncompactness, Nonlinear integral equation of Volterra–
150 Stieltjes type, 281
Normal (solid) sequence space, 45
L
λ-limit of a sequence, 73 O
Linear metric space, 4 Operator
Linear topological space, 10 compact, 148
Ljusternik-Šnirelman-Borsuk theorem, 154 completely continuous, 148
Index 315
T
S Tangent segment, 225
Sargent sequence space, 133, 134 Theorem
Schauder basis, 12, 13 Arzéla–Ascoli, 275
Schur matrix, 57 Kojima-Schur, 60
Schur theorem, 61 Ljusternik-Šnirelman-Borsuk, 154
Semilinear infinite system of differential Peano, 220
equations, 248, 254, 257 Schur, 61
Semilinear upper diagonal infinite system of Steinhaus, 63
integral equations, 278 Toeplitz matrix, 60
Seminorm, 9 Totally bounded set, 147
Separable space, 16 Translation invariant metric, 3
Sequence, 1, 3
Sequence space, 1, 2
Maddox, 105 U
normal (solid), 45 Upper closed limit of a sequence of sets, 180
perfect, 45 Upper limit of a real sequence, 195
Sargent, 133, 134 Upper limit of a sequence of sets, 180