Applied Mathematics Letters 24 (2011) 1856–1860

Contents lists available at ScienceDirect

Applied Mathematics Letters

journal homepage: www.elsevier.com/locate/aml

An application of almost convergence in approximation theorems

S.A. Mohiuddine
Department of Mathematics, Aligarh Muslim University, Aligarh 202002, India

article info abstract

Article history: The concept of almost convergence was introduced by Lorentz in 1948 [G.G. Lorentz, A
Received 3 June 2010 contribution to theory of divergent sequences, Acta Math. 80 (1948) 167–190], and has
Accepted 6 May 2011 various applications. In this work we apply this method to prove some Korovkin type
approximation theorems.
Keywords: © 2011 Elsevier Ltd. All rights reserved.
Banach limit
Almost convergence
Korovkin type approximation theorem

1. Introduction and preliminaries

Let c and l∞ denote the spaces of all convergent and bounded sequences, respectively, and note that c ⊂ l∞ . In the theory
of sequence spaces, a beautiful application of the well known Hahn–Banach Extension Theorem gave rise to the concept of
the Banach limit. That is, the lim functional defined on c can be extended to the whole of l∞ and this extended functional is
known as the Banach limit [1]. In 1948, Lorentz [2] used this notion of a Banach limit to define a new type of convergence,
known as the almost convergence.
A bounded sequence x = (xk ) is said to be almost convergent to the number ℓ if and only if
 n+ p−1 
1 −
xk − ℓ = 0, uniformly in n. (∗)

lim 
p→∞ p k=n

We shall denote by F the space of almost convergent sequences and we write ℓ = F - lim x.
Note that a convergent sequence is almost convergent, and its limit and its generalized limit are identical, but an almost
convergent sequence need not be convergent.

Example 1.1. The sequence z = (zn ) defined as

if n is odd,

1
zn =
0 if n is even

is almost convergent to 1/2 but not convergent.

If n = 1 in (∗) then we get (C , 1) convergence, and in this case we write xk → ℓ(C , 1); where ℓ = (C , 1)- lim x.
Most of the commonly used matrix methods contain the method F , e.g. every almost convergent sequence is also (C , α)
and (E , α)-summable (α > 0) to its F -limit.
Let C [a, b] be the space of all functions f continuous on [a, b]. We know that C [a, b] is a Banach space with norm
‖f ‖∞ := supa≤x≤b |f (x)|, f ∈ C [a, b]. Suppose that Tn : C [a, b] → C [a, b]. We write Tn (f , x) for Tn (f (t ), x) and we say
that T is a positive operator if T (f , x) ≥ 0 for all f (x) ≥ 0.

E-mail address: [email protected].

0893-9659/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aml.2011.05.006
 S.A. Mohiuddine / Applied Mathematics Letters 24 (2011) 1856–1860 1857

The classical Korovkin approximation theorem is stated as follows [3]:

Let (Tn ) be a sequence of positive linear operators from C [a, b] into C [a, b] and limn ‖Tn (fi , x)− fi (x)‖∞ = 0, for i = 0, 1, 2,
where f0 (x) = 1, f1 (x) = x and f2 (x) = x2 . Then limn ‖Tn (f , x) − f (x)‖∞ = 0, for all f ∈ C [a, b].
Quite recently, such approximation theorems were proved in [4,5] for functions of one and two variables by using λ-
statistical convergence and statistical convergence respectively. In this work, we use the notion of almost convergence such
as that of [6] to prove Korovkin type approximation theorems.

2. The Korovkin type approximation theorem

The following is the F -version of the classical Korovkin approximation theorem, followed by an example to show its
importance.

Theorem 2.1. Suppose that (Tk ) is a sequence of positive linear operators from C [a, b] into C [a, b] and Dn,p (f , x) =
∑n+p−1
1
p k=n Tk (f , x), satisfying the following conditions:

lim ‖Dn,p (1, x) − 1‖∞ = 0 uniformly in n, (2.1.1)

p→∞

lim ‖Dn,p (t , x) − x‖∞ = 0 uniformly in n, (2.1.2)

p→∞

lim ‖Dn,p (t 2 , x) − x2 ‖∞ = 0 uniformly in n. (2.1.3)

p→∞

Then for any function f ∈ C [a, b] bounded on the whole real line, we have
 
Tk (f , x) − f (x) = 0, i.e.
 
F - lim 
k→∞ 
∞
 
Dn,p (f , x) − f (x) = 0, uniformly in n.
 
lim  
p→∞
∞

Proof. Since f ∈ C [a, b] and f is bounded on the real line, we have

|f (x)| ≤ M , −∞ < x < ∞.

Therefore,

|f (t ) − f (x)| ≤ 2M , −∞ < t , x < ∞. (2.1.4)

Also we have that f is continuous on [a, b], i.e.

|f (t ) − f (x)| < ϵ, ∀|t − x| < δ. (2.1.5)

Using (2.1.4) and (2.1.5) and putting ψ(t ) = (t − x) , we get 2

2M
|f (t ) − f (x)| < ϵ + ψ, ∀|t − x| < δ.
δ2
This means that
2M 2M
−ϵ − ψ < f (t ) − f (x) < ϵ + ψ.
δ 2 δ2
Now, we operate with Tk (1, x) on this inequality since Tk (f , x) is monotone and linear. Hence
   
2M 2M
Tk (1, x) −ϵ − ψ < Tk (1, x)(f (t ) − f (x)) < Tk (1, x) ϵ + 2 ψ .
δ2 δ
Note that x is fixed and so f (x) is constant number. Therefore

2M 2M
−ϵ Tk (1, x) − Tk (ψ, x) < Tk (f , x) − f (x)Tk (1, x) < ϵ Tk (1, x) + Tk (ψ, x). (2.1.6)
δ2 δ2
But

Tk (f , x) − f (x) = Tk (f , x) − f (x)Tk (1, x) + f (x)Tk (1, x) − f (x)

= [Tk (f , x) − f (x)Tk (1, x)] + f (x)[Tk (1, x) − 1]. (2.1.7)
1858 S.A. Mohiuddine / Applied Mathematics Letters 24 (2011) 1856–1860

Using (2.1.6) and (2.1.7), we have

2M
Tk (f , x) − f (x) < ϵ Tk (1, x) + Tk (ψ, x) + f (x)(Tk (1, x) − 1). (2.1.8)
δ2
Let us estimate Tk (ψ, x):

Tk (ψ, x) = Tk ((t − x)2 , x)

= Tk (t 2 − 2tx + x2 , x)
= Tk (t 2 , x) + 2xTk (t , x) + x2 Tk (1, x)
= [Tk (t 2 , x) − x2 ] − 2x[Tk (t , x) − x] + x2 [Tk (1, x) − 1].
Using (2.1.8), we obtain
2M
Tk (f , x) − f (x) < ϵ Tk (1, x) + {[Tk (t 2 , x) − x2 ] + 2x[Tk (t , x) − x] + x2 [Tk (1, x) − 1]} + f (x)(Tk (1, x) − 1)
δ2
2M
= ϵ[Tk (1, x) − 1] + ϵ + {[Tk (t 2 , x) − x2 ] + 2x[Tk (t , x) − x]
δ2
+ x2 [Tk (1, x) − 1]} + f (x)(Tk (1, x) − 1).
Since ϵ is arbitrary, we can write
2M
Tk (f , x) − f (x) ≤ ϵ[Tk (1, x) − 1] + {[Tk (t 2 , x) − x2 ] + 2x[Tk (t , x) − x]
δ2
+ x2 [Tk (1, x) − 1]} + f (x)(Tk (1, x) − 1).
Similarly,
2M
Dn,p (f , x) − f (x) ≤ ϵ[Dn,p (1, x) − 1] + {[Dn,p (t 2 , x) − x2 ] + 2x[Dn,p (t , x) − x]
δ2
+ x2 [Dn,p (1, x) − 1]} + f (x)(Dn,p (1, x) − 1),
and therefore
2Mb2
 
4Mb 2M
‖Dn,p (f , x) − f (x)‖∞ ≤ ϵ + + M ‖Dn,p (1, x) − 1‖∞ + ‖Dn,p (t , x) − x‖∞ + ‖Dn,p (t 2 , x) − x2 ‖∞ .
δ2 δ2 δ2
Supposing that p → ∞ and using (2.1.1)–(2.1.3), we get

lim ‖Dn,p (f , x) − f (x)‖∞ = 0, uniformly in n.

p→∞

This completes the proof of the theorem. 

In the following we give an example of a sequence of positive linear operators satisfying the conditions of Theorem 2.1
but not satisfying the conditions of the Korovkin theorem.

Example 2.1. Consider the sequence of classical Bernstein polynomials

n   
− k n
Bn (f , x) := f xk (1 − x)n−k ; 0 ≤ x ≤ 1.
k=0
n k

Let the sequence (Pn ) be defined by Pn : C [0, 1] → C [0, 1] with Pn (f , x) = (1 + zn )Bn (f , x), where zn is defined as above.
Then
x − x2
Bn (1, x) = 1, Bn (t , x) = x, Bn (t 2 , x) = x2 + ,
n
and the sequence (Pn ) satisfies the conditions (2.1.1)–(2.1.3). Hence we have

F - lim ‖Pn (f , x) − f (x)‖∞ = 0.

On the other hand, we get Pn (f , 0) = (1 + zn )f (0), since Bn (f , 0) = f (0), and hence

‖Pn (f , x) − f (x)‖∞ ≥ |Pn (f , 0) − f (0)| = zn |f (0)|.

We see that (Pn ) does not satisfy the classical Korovkin theorem, since lim supn→∞ zn does not exist.
 S.A. Mohiuddine / Applied Mathematics Letters 24 (2011) 1856–1860 1859

3. Some generalizations

Now we present a slight general result.

Theorem 3.1. Let (Tn ) be a sequence of positive linear operators on C [a, b] such that

lim ‖Tn+1 − Tn ‖∞ = 0. (3.1.1)

n

If

F - lim ‖Tn (t ν , x) − xν ‖∞ = 0 (ν = 0, 1, 2). (3.1.2)

n

Then for any function f ∈ C [a, b] bounded on the real line, we have

lim ‖Tn (f , x) − f (x)‖∞ = 0. (3.1.3)

n

Proof. From Theorem 2.1, we have that if (3.1.2) holds then

lim ‖Dn,p (f , x) − f (x)‖∞ = 0, uniformly in n. (3.1.4)

p

We have the following inequality:

 
n+p−1 k
1 − −
‖Tn (f , x) − f (x)‖∞ ≤ ‖Dn,p (f , x) − f (x)‖∞ + ‖Tn − Tn−1 ‖∞
p k=n+1 l =n +1
 
p−1
≤ ‖Dn,p (f , x) − f (x)‖∞ + sup ‖Tn − Tn−1 ‖∞ . (3.1.5)
2 k≥n

Hence using (3.1.1) and (3.1.4), we get (3.1.3).

This completes the proof of the theorem. 

We know that almost convergence implies (C , 1) convergence. This motivates us to further generalize our main result
by weakening the hypothesis or to add a condition to get a more general result.

Theorem 3.2. Let (Tn ) be a sequence of positive linear operators on C [a, b] such that

(C , 1)- lim ‖Tn (t ν , x) − xν ‖∞ = 0 (ν = 0, 1, 2), (3.2.1)

n

and
 
n
lim sup ‖σn+p−1 (f , x) − σn−1 (f , x)‖∞ = 0, (3.2.2)
p n ≥p p

where σn (f , x) = n+ Tk (f , x). Then for any function f ∈ C [a, b] bounded on the real line, we have
1
∑n
1 k=0

F - lim ‖Tn (f , x) − f (x)‖∞ = 0.

n→∞

Proof. For n ≥ p ≥ 1, it is easy to show that

n
Dn,p (f , x) = σn+p−1 (f , x) + (σn+p−1 (f , x) − σn−1 (f , x)),
p
which implies
n
sup ‖Dn,p (f , x) − σn+p−1 (f , x)‖∞ = sup ‖(σn+p−1 (f , x) − σn−1 (f , x))‖∞ . (3.2.3)
n ≥p n ≥p p

Also by Theorem 2.1, Condition (3.2.1) implies that

(C , 1)- lim ‖Tn (f , x) − f (x)‖∞ = 0. (3.2.4)

n→∞

Using (3.2.1)–(3.2.4) and the fact that almost convergence implies (C , 1) convergence, we get the desired result.
This completes the proof of the theorem. 
1860 S.A. Mohiuddine / Applied Mathematics Letters 24 (2011) 1856–1860

Theorem 3.3. Let (Tn ) be a sequence of positive linear operators on C [a, b] such that
m+n−1
1 −
lim sup ‖Tn − Tk ‖ = 0.
n m n k=m

Suppose that

F - lim ‖Tn (t ν , x) − xν ‖∞ = 0 (ν = 0, 1, 2). (3.3.1)

n

Then for any function f ∈ C [a, b] bounded on the real line, we have
lim ‖Tn (f , x) − f (x)‖∞ = 0. (3.3.2)
n

Proof. From Theorem 2.1, we have that if (3.3.1) holds then

F - lim ‖Tn (f , x) − f (x)‖∞ = 0,
n

which is equivalent to
lim ‖ sup Dm,n (f , x) − f (x)‖∞ = 0.
n m

Now
m+n−1
1 −
Tn − Dm,n = Tn − Tk
n k=m
m+n−1
1 −
= (Tn − Tk ).
n k=m

Therefore
m+n−1
1 −
Tn − sup Dm,n = sup (Tn − Tk ).
m m n k=m

Hence using the hypothesis we get

lim ‖Tn (f , x) − f (x)‖∞ = lim ‖ sup Dm,n (f , x) − f (x)‖∞ = 0,
n n m

that is, (3.3.2) holds.

This completes the proof of the theorem. 

4. Conclusion

This work presents a relationship between two different disciplines: the theory of almost convergence introduced by
Lorentz in 1948 and Korovkin type approximation theorems. Furthermore, we know that convergence implies almost
convergence and almost convergence implies (C , 1) convergence, but in Section 3, we prove the converse of both cases
under one additional condition.

References

[1] S. Banach, Théorie des Operations Lineaires, Warsaw, 1932.

[2] G.G. Lorentz, A contribution to theory of divergent sequences, Acta Math. 80 (1948) 167–190.
[3] A.D. Gadz̆iev, The convergence problems for a sequence of positive linear operators on unbounded sets, and theorems analogous to that of P.P. Korovkin,
Soviet Math. Dokl. 15 (1974) 1433–1436.
[4] F. Dirik, K. Demirci, Korovkin type approximation theorem for functions of two variables in statistical sense, Turk. J. Math. 33 (2009) 1–11.
[5] O.H.H. Edely, S.A. Mohiuddine, A.K. Noman, Korovkin type approximation theorems obtained through generalized statistical convergence, Appl. Math.
Lett. 23 (11) (2010) 1382–1387.
[6] Z.U. Ahmad, Mursaleen, An Application of Banach limits, Proc. Amer. Math. Soc. 103 (1) (1988) 244–246.