ON I-CONVERGENCE FIELD

TIBOR ŠALÁT, BINOD CHANDRA TRIPATHY AND MILOŠ ZIMAN

Abstract. In this paper we introduce the notion of cIA and mIA , the I-convergence
field and bounded I-convergence field of an infinite matrix A. We restrict our study to
diagonal matrices. We find necessary and/or sufficient conditions on the elements of A
for solidity of cIA and mIA .

1. Introduction
This paper is a generalization of previous paper of Maddox (cf. [7]). The second of
the authors came with an idea to extend the notion of summability fields of a matrix
A to statistical convergence field of A (using statistical convergence instead of usual
convergence) and the first of the authors noticed the possibility of extension of these
notions with the help of I-convergence.
The concept of I-convergence of sequences in a metric space is introduced in [5].
The statistical convergence is a special case of I-convergence. It was introduced in [2]
and [11] independently. Later on it was studied from sequence space point of view and
linked with summability in [1], [3], [4], [6], [8], [9], [10], [12], [13] and many others.

2. Definitions and Background

Let X be a normed linear space. We denote by w(X), c(X), c0 (X), l1 (X), l∞ (X)
the sets of all, convergent, null, absolutely summable and bounded sequences, respectively,
with elements from X. For X = C we will write simply w, c, c0 , l1 and l∞ .
We recall the concept of an ideal of subsets of N. A non-void class I ⊆ 2N is called the
ideal if I is additive (i.e., A, B ∈ I ⇒ A ∪ B ∈ I) and hereditary (i.e., A ∈ I, B ⊆ A ⇒
B ∈ I). An ideal I ⊆ 2N is called non-trivial if I 6= 2N .
Definition 2.1. A non-trivial ideal I of subsets of N is said to be admissible if I contains
every finite subset of N.
Throughout this paper the calligraphic letter I will denote a non-trivial admissible
ideal of subsets of N.
Definition 2.2. Let X be a normed linear space. We say that x = (xn ) ∈ w(X) is
I-convergent if there exists ξ ∈ X such that for all ε > 0 the following holds:

n ∈ N : kxn − ξk ≥ ε ∈ I.
(k · k denotes the norm in X.) In this case we say that x I-converges to ξ and write
ξ = I- limn→∞ xn (cf. [5]).

1991 Mathematics Subject Classification. 40D09, 40D20, 46A45.

Key words and phrases. density, solid space, matrix map, I-convergence field.
1
2 TIBOR ŠALÁT, BINOD CHANDRA TRIPATHY AND MILOŠ ZIMAN

Example 2.3. (a) The usual convergence is a special case of I-convergence. In this case:
I = I f (the ideal of all finite subsets of N).
(b) Taking I = I d (I δ ), where I d (I δ ) is the ideal of subsets of N of asymptotic
(logarithmic) density zero, we get the statistical convergence (the logarithmic statistical
convergence).
(c) For further examples of I-convergence see [5].
Remark. As the ideal I is admissible, I f ⊆ I. Hence every convergent sequence is
I-convergent.
For any ideal I there is a filter F(I) corresponding with I:

F(I) = K ⊆ N : (N r K) ∈ I .
Lemma 2.4 ([5, Proposition
 1.2]). Let X be a normed linear space and x = (xn ) ∈ w(X).
If there is a set K = n1 < n2 < . . . ∈ F(I) such that x converges along K to ξ ∈ X
(i.e., ξ = limi→∞ xni ), then x I-converges to ξ.
Lemma 2.5. Let D ∈ F(I) and M ⊆ N. If M ∈
/ I then M ∩ D ∈
/ I.
Proof. Assume that M ∈/ I and M ∩ D ∈ I. Then K = (N r D) ∪ (M ∩ D) ∈ I. But
M ⊆ K and so M ∈ I. This is a contradiction to the assumption.
Therefore M ∩ D ∈/ I. 
Definition 2.6. A subspace E of space w(X) is said to be solid if x = (xn ) ∈ E and
kyn k ≤ kxn k for all n ∈ N implies y = (yn ) ∈ E.
Definition 2.7. If A = (Ank ) is an infinite matrix of operators Ank ∈ B(X, Y ) (B(X, Y )
being the set of all bounded linear operators from X to Y ) and x = (xn ) ∈ w(X), then
we say that x is I-summable by A to ξ ∈ Y if
X∞
An x = Ank xk
k=1
I-converges to ξ in the norm of Y . Further x is said to be bounded I-summable by A to
ξ ∈ Y if and only if (An x) is bounded and I-converges to ξ in the norm of Y .

I
I
I
I
For a linear normed space X we will denote c(X) , c0 (X) , m(Y ) and m0 (Y )
the sets of I-convergent, I-null, bounded I-convergent and bounded I-null sequences of
points from X, respectively. In case X = C we will write simply cI , cI0 , mI and mI0 ,
respectively.
Definition 2.8. The I-convergence field of a matrix A is defined by

I n
I o
c(Y ) A = x = (xk ) ∈ w(X) : (An x) ∈ c(Y ) .
The bounded I-convergence field of a matrix A is defined by

I n
I o
m(Y ) A = x = (xk ) ∈ w(X) : (An x) ∈ m(Y )

I
I

Since c(Y ) ⊂ c(Y ) , so c(Y ) A is a natural generalization of c(Y ) A , i.e., the conver-
gence field of A (cf. [7]). We introduce the following notations:

n o
c0 (Y ) A = x = (xk ) ∈ w(X) : (An x) ∈ c0 (Y ) ,

I n
I o
c0 (Y ) A = x = (xk ) ∈ w(X) : (An x) ∈ c0 (Y )
 ON I-CONVERGENCE FIELD 3

and
I n
I o
m0 (Y ) A = x = (xk ) ∈ w(X) : (An x) ∈ m0 (Y ) .
If Y = C we will omit the argument C and we will use the notations cIA , mIA , (c0 )A ,
(c0 )IA and (m0 )IA , respectively.
Throughout the paper we assume that Ank = θ if n 6= k, where θ is the null operator
from X to Y , i.e., A is infinite diagonal matrix of operators. In this case An x = Ann xn
for x = (xn ) ∈ w(X). We also restrict ourselves to the case Y = C.

3. Main Results
The proof of the following proposition is obvious.
Proposition 3.1. (c0 )A , cIA , (c0 )IA , mIA and (m0 )IA are linear spaces.
Recall that a non-trivial ideal I is said to be maximal if there does not exist any
non-trivial ideal J 6= I containing I as a subset (cf. [5]).
Theorem 3.2. If mIA is solid and I is not maximal, then

(1) n ∈ N : Ann = θ ∈/ I.

Proof. Let us denote B = n ∈ N : Ann = θ and assume that B ∈ I.
Then for all n ∈ D = N r B there exists zn ∈ X such that Ann zn 6= 0.
Define the sequences x = (xn ), y = (yn ) ∈ w(X) as follows:

xn = 0 for n ∈ B, yn = 0 for n ∈ B,
zn zn
xn = Ann zn otherwise, yn = Ann zn for n ∈ M ∩ D,
yn = 0 otherwise,
where M ⊂ N is such that
(2) M∈
/ I nor (N r M ) ∈
/I
(such a set exists because I is not maximal).
One can see that

Ann xn = 0 for n ∈ B,
Ann xn = 1 otherwise.
Hence (Ann xn ) is bounded and I- limn→∞ Ann xn = 1, i.e., x ∈ mIA .
As kyn k ≤ kxn k for all n ∈ N and by the assumption mIA is solid, then y ∈ mIA too.
Let us enumerate the sequence (Ann yn ):

Ann yn = 0 for n ∈ B,
Ann yn = 1 for n ∈ M ∩ D,
Ann yn = 0 otherwise.
Hence the I- limn→∞ Ann yn can be only 0 or 1. 
Let I- lim n→∞ A nn y n = 0. Taking any ε ∈ (0, 1) we get M ∩ D = n ∈ N : |Ann yn | ≥
ε ∈ I. But this contradicts Lemma 2.5. 
Let I- limn→∞ Ann yn = 1. Taking any ε ∈ (0, 1) we get N r (M ∩ D) = n ∈ N :
|Ann yn − 1| ≥ ε ∈ I. As N r (M ∩ D) = (N r M ) ∪ (N r D) then (N r M ) ∈ I but this
contradicts (2).
4 TIBOR ŠALÁT, BINOD CHANDRA TRIPATHY AND MILOŠ ZIMAN

Hence I- limn→∞ Ann yn does not exist.

Therefore mIA is not solid and this completes the proof of the theorem. 
Throughout this section we will assume that ideal I is not maximal. By M we will
denote a subset of N satisfying condition (2). The case of maximal ideal will be discussed
later.
The following examples show that the condition (1) is not sufficient in general for the
solidity of mIA .
Example 3.3. Let X = c0 , the space of all complex sequences converging to zero with
sup-norm. Define Ann s for s = (sn ) ∈ c0 as follows:
Ann s = sn for n ∈ M,
Ann s = 0 otherwise.
The condition (1) is satisfied, but we show that mIA is not solid.
To prove this, construct the sequences x = (x(n) ) and y = (y (n) ), where x(n) =
(xn1 , xn2 , . . . ) and y (n) = (yn1 , yn2 , . . . ) are in c0 , in the following way:
xn1 = 1 for n ∈ N, ynn = 1 for n ∈ N,
xnn = n1 for n ∈ N, ynk = 0 for k 6= n.
xnk = 0 for k =6 n, k ≥ 2,
For all n ∈ N we have ky (n) k = kx(n) k.
One can easily verify that limn→∞ Ann x(n) = 0. Hence I- limn→∞ Ann x(n) = 0. But
I- limn→∞ Ann y (n) does not exist (Ann y (n) equals 1 for n ∈ M and 0 otherwise).
Therefore y ∈/ mIA and mIA is not solid.
Example 3.4. Let X = l1 , the space of all absolutely summable complex sequences.
Define Ann s for s = (sn ) ∈ l1 as follows:
Ann s = sn for n ∈ M,
Ann s = 0 otherwise.
The condition (1) is satisfied, but mIA is not solid.
Define the sequences x = (x(n) ) and y = (y (n) ) in w(l1 ) as follows: For all n ∈ N set
x(n) = (xn1 , xn2 , . . . ) and y (n) = (yn1 , yn2 , . . . ), where
n
xn1 = 1, ynn = (−1)2
,
xnn = n1 , ynk = 0 for k 6= n.
xnk = 0 for k 6= n, k ≥ 2,
We have ky (n) k = ∞
P 1 1
P∞ (n)
k=1 |y nk | = 2
≤ 1 + n
= k=1 |xnk | = kx k.
Because limn→∞ Ann x = 0 and I- limn→∞ Ann y does not exist, mIA is not solid.
(n) (n)

The proof of the following result is a routine work in view of Theorem 3.2.
Theorem 3.5. If cIA is solid then the condition (1) holds.
Proposition 3.6. Let X = c0 . Then cIA is solid if and only if

(3) n ∈ N : Ann = θ ∈ F(I).
 ON I-CONVERGENCE FIELD 5

Proof. Let us denote B = n ∈ N : Ann = θ .
If B ∈ F(I) then for all x = (x(n) ) ∈ w(c0 ) we have: I- limn→∞ Ann x(n) = 0 (see
Lemma 2.4). Therefore cIA = w(c0 ) and w(c0 ) is solid. So the sufficiency of condition (3)
is proved.
Now assume that cIA is solid and B ∈

/ F(I), i.e., D = N r B = n ∈ N : Ann 6= θ ∈ / I.
For every n ∈ N, Ann is a bounded linear functional. Hence it has the form
∞
X
Ann s = ank sk for all s = (sk ) ∈ c0 ,
k=1

where
∞
X
(4) |ank | < ∞.
k=1

Let n1 < n2 < . . . be the enumeration of D. Since Ani ni 6= θ, there is ki ∈ N such
that ani ki 6= 0 for i = 1, 2, . . . . Condition (4) implies that there exists an ri > ki such that
1
|ani ri | < |ani ki |.
i
Define the sequences x = (x(n) ) and y = (y (n) ), where x(n) = (xn1 , xn2 , . . . ) and y (n) =
(yn1 , yn2 , . . . ) are in c0 for all n ∈ N, as follows:

xnk = ian1 k for n = ni , k = ki , ynk = an1k for n = ni , k = ki ,

i i i i
xnk = an1k for n = ni , k = ri , ynk = 0 otherwise.
i i
xnk = 0 otherwise,
It is clear that kx(n) k = ky (n) k for all n ∈ N.
Let us look at the sequences (Ann x(n) ) and (Ann y (n) ): If n ∈ B then Ann x(n) = Ann y (n) =
0.
Let n = ni . Then
|Ani ni x(ni ) | = | P ∞ 2
P
k=1 ani k xni k | = |ani ki xni ki + ani ri xni ri | ≤ i ,
∞
|Ani ni y (ni ) | = | k=1 ani k yni k | = |ani ki yni ki | = 1.
It follows that I- limn→∞ Ann x(n) = 0. But I- limn→∞ Ann y (n) does not exist, since
Ann y (n) = 0 for n ∈ B and Ann y (n) = 1 for n ∈ D with B ∈
/ I (by Theorem 3.5) and
D∈ / I (by assumption).
Thus cIA is not solid.

The following example demonstrates that cIA in the previous proposition cannot be
replaced by mIA .
Example 3.7. Let X = c0 . Let D ∈ I be an infinite set. (If such a D does not exist,
then I = I f and in this case cIA = mIA .) Define Ann s for s = (sn ) ∈ c0 as follows:

Ann s = ns1 + ns2 , for n ∈ D,

Ann s = 0, otherwise.
Then the condition (3) is satisfied, but mIA is not solid.
6 TIBOR ŠALÁT, BINOD CHANDRA TRIPATHY AND MILOŠ ZIMAN

Take the sequences x = (x(n) ) and y = (y (n) ) in w(c0 ), where

x(n) = (1, −1, 0, 0, . . . ),
y (n) = (1, 0, 0, . . . )
for all n ∈ N. Then ky (n) k = kx(n) k and x ∈ mIA , but (Ann y (n) ) is unbounded, thus
y∈/ mIA .
Proposition 3.8. Let X = C. Then mIA (cIA ) is solid if and only if the condition (1)
holds.
Proof. We prove the statement of the proposition for mIA only. The proof for cIA runs in
a similar way.
If mIA is solid, then the condition (1) holds by Theorem 3.2. Hence the necessity of (1)
is proved.
Let us prove the sufficiency of (1).
Assume that (1) holds. Take x = (xn ) ∈ mIA and let y = (yn ) ∈ w be such that
kyn k ≤ kxn k for all n ∈ N. We claim that y ∈ mIA .
Every operator Ann has the form Ann s = an s for s ∈ C, where an ∈ C.
Let ξ = I- limn→∞ an xn . Then for all ε > 0 the following holds:

n ∈ N : |an xn − ξ| ≥ ε ∈ I.
Since (1) holds, ξ = 0.
By assumption for all n ∈ N we have:
|an yn | = |an ||yn | ≤ |an ||xn | = |an xn |.
This means that (an yn ) is bounded (x ∈ mIA ) and
 
n ∈ N : |an yn | ≥ ε ⊆ n ∈ N : |an xn | ≥ ε
Thus I- limn→∞ an yn = 0 and y ∈ mIA . 
Proposition 3.9. Let X = C. Then the spaces (c0 )A , (c0 )IA and (m0 )IA are solid for any
infinite diagonal matrix A.
Proof. Let A = (ank ) be a diagonal matrix. Further, let x = (xn ) ∈ (m0 )IA and y = (yn ) ∈
w be such that |yn | ≤ |xn | for all n ∈ N. Then we have
|ann yn | ≤ |ann xn | for all n ∈ N.
As I- limn→∞ ann xn = 0 then also I- limn→∞ ann yn = 0. The boundedness of (ann yn ) is
clear. Thus (yn ) ∈ (m0 )IA .
Similarly one can show that (c0 )A and (c0 )IA are solid too. 

4. Case of maximal ideal

Definition 4.1. Let X be a normed linear space. We say that x = (xn ) ∈ w(X) is
I-bounded if there is a number M > 0 such that the following holds:

n ∈ N : kxn k > M ∈ I.
Remark. Every bounded sequence is I-bounded.
Lemma 4.2. Let I be maximal. Then every I-bounded sequence x = (xn ) of complex
numbers is I-convergent.
 ON I-CONVERGENCE FIELD 7

Proof. Let x = (xn ) be I-bounded sequence of complex numbers. Then there is a closed
square U (1) in complex plane such that the set of indices of elements of x lying outside of
U (1) belongs to I, i.e.,
n ∈ N : xn ∈ U (1) ∈ F(I).

(1) (1) (1) (1)
Divide U (1) into four closed squares U1 , U2 , U3 , U4 with
(1) 1
diam Ui = diam U (1) (i = 1, 2, 3, 4).
2
The following holds at least for one of these squares:
 (1)
n ∈ N : xn ∈ Ui ∈ F(I).
Denote this square U (2) .
In this way, by induction, we construct a sequence
U (1) ⊃ U (2) ⊃ · · · ⊃ U (n) ⊃ . . .
of closed squares with the properties
k ∈ N : xk ∈ U (n) ∈ F(I)

for all n ∈ N
and
lim diam U (n) = 0.
n→∞
By Cantor’s Theorem there exists (the only)
\
ξ∈ U (n) .
n∈N

We claim that ξ = I- limn→∞ xn . 

Let ε > 0. Take a sphere B(ξ, ε) = z ∈ C : |z − ξ| < ε . Then for all sufficiently large
m ∈ N we have:
U (m) ⊆ B(ξ, ε).
Fix such an m. Then
n ∈ N : xn ∈ U (m) ⊆ n ∈ N : xn ∈ B(ξ, ε) .
 

As the left side belongs to F(I) so does the right side. Thus

n ∈ N : |xn − ξ| ≥ ε ∈ I.
Hence I- limn→∞ xn = ξ. 
Proposition 4.3. Let X = C and I be maximal. Then mIA (cIA ) is solid for every infinite
diagonal matrix A.
Proof. Let A = (aij ) be a diagonal matrix (i.e., aij = 0 for i 6= j) and x = (xn ) ∈ mIA .
Then (ann xn ) is bounded I-convergent sequence of complex numbers.
Let y = (yn ) ∈ w be such that
|yn | ≤ |xn | for all n ∈ N.
Hence
|ann yn | ≤ |ann xn | for all n ∈ N.
Thus (ann yn ) is bounded and so the solidity of mIA follows easily from Lemma 4.2.
The proof for cIA uses the fact that every I-convergent sequence is I-bounded.

8 TIBOR ŠALÁT, BINOD CHANDRA TRIPATHY AND MILOŠ ZIMAN

References
[1] Connor, J.S.: The statistical and p-Cesàro convergence of sequences, Analysis 8 (1998), 47-63.
[2] Fast, H.: Sur la convergence statistique, Colloq. Math. 2 (1951), 241-244.
[3] Fridy, J.A.: Statistical limit points, Proc. Amer. Math. Soc. 11 (4) (1993), 1187-1192.
[4] Kolk, E.: The statistical convergence in Banach spaces, Tartu Ul Toimetised 928 (1991), 41-52.
[5] Kostyrko, P., Šalát, T. and Wilczyński, W.: I-convergence, Real Anal. Exchange 26 (2000/2001),
669-686.
[6] Maddox, I.J.: Statistical convergence in locally convex spaces, Math. Proc. Camb. Phil. Soc. 104
(1988), 11-145.
[7] Maddox, I.J.: Solid summability fields, Math. Slovaca 42 no. 2 (1992), 211-214.
[8] Miller, H.I.: A measure theoretical subsequence characterization of statistical convergence, Tran.
Amer. Math. Soc. 347 (1995), 1811-1819.
[9] Rath, D. and Tripathy, B.C.: Matrix maps on sequence spaces associated with sets of integer se-
quences, Ind. Jour. Pure and Appl. Math. 27 (2) (1996), 197-206.
[10] Šalát, T.: On statistically convergent sequences of real numbers, Math. Slovaca 30 (1980), 139-150.
[11] Schoenberg, I.J.: The integrability of certain functions and related summability methods, Amer.
Math. Monthly 66 (1959), 361-375.
[12] Tripathy, B.C.: Matrix transformation between some classes of sequences, Jour. Math. Analysis and
Appl. 206 (2) (1997), 448-450.
[13] Tripathy, B.C.: On statistical convergence, Proc. Estonian Acad. Sci. Phys. Math. 47 (4) (1998),
299-303.

Department of Algebra and Number Theory, Faculty of Mathematics, Physics and

Informatics, Comenius University, 842 48 Bratislava, SLOVAKIA

Mathematical Sciences Division, Institute of Advanced Study in Science and Technol-

ogy, Khanapara, Guwahati - 718 022, INDIA
E-mail address: tripathybc@yahoo.com

Department of Algebra and Number Theory, Faculty of Mathematics, Physics and

Informatics, Comenius University, 842 48 Bratislava, SLOVAKIA
E-mail address: ziman@fmph.uniba.sk