Acta Math. Univ.

Comenianae 281
Vol. LXXIX, 2(2010), pp. 281–292

MATRIX SUMMABILITY AND KOROVKIN TYPE

APPROXIMATION THEOREM ON MODULAR SPACES

S. KARAKUŞ and K. DEMIRCI

Abstract. In this paper, using a matrix summability method we obtain a Korovkin

type approximation theorem for a sequence of positive linear operators defined on
a modular space.

1. Introduction

Approximation theory has important applications in the theory of polynomial ap-

proximation, in various areas of functional analysis, in numerical solutions of dif-
ferential and integral equations [9], [10], [11]. Most of the classical approximation
operators tend to converge to the value of the function being approximated. How-
ever, at points of discontinuity, they often converge to the average of the left and
right limits of the function. There are, however, some sharp exceptions such as the
interpolation operator of Hermite-Fejer (see [7]). These operators do not converge
at points of simple discontinuity. For such misbehavior, the matrix summability
methods of Cesáro type are strong enough to correct the lack of convergence (see
[8]). Using a matrix summability method some approximation results were studied
in [1, 2, 18, 19, 21]. In this paper, using a matrix summability method we give
a theorem of the Korovkin type for a sequence of positive linear operators defined
on a modular space.
We now recall some basic definitions and notations used in the paper.
Let I = [a, b] be a bounded interval of the real line R provided with the
Lebesgue measure. Then, by X (I) we denote the space of all real-valued mea-
surable functions on I provided with equality a.e. As usual, let C (I) denote
the space of all continuous real-valued functions, and C ∞ (I) denote the space of
all infinitely differentiable functions on I. In this case, we say that a functional
ρ : X (I) → [0, +∞] is a modular on X (I) provided that the following conditions
hold:
(i) ρ (f ) = 0 if and only if f = 0 a.e. in I,
(ii) ρ (−f ) = ρ (f ) for every f ∈ X (I),

Received May 5, 2010; revised July 13,2010.

2000 Mathematics Subject Classification. Primary 40A30, 41A36, 47G10, 46E30.
Key words and phrases. Positive linear operators; modular space; matrix summability; Ko-
rovkin theorem.
282 S. KARAKUŞ and K. DEMIRCI

(iii) ρ (αf + βg) ≤ ρ (f ) + ρ (g) for every f, g ∈ X(I) and for any α, β ≥ 0 with
α + β = 1.
A modular ρ is said to be N -quasi convex if there exists a constant N ≥ 1 such
that the inequality
ρ (αf + βg) ≤ N αρ (N f ) + N βρ (N g)
holds for every f, g ∈ X (I), α, β ≥ 0 with α + β = 1. In particular, if N = 1, then
ρ is called convex.
A modular ρ is said to be N -quasi semiconvex if there exists a constant N ≥ 1
such that the inequality
ρ(af ) ≤ N aρ(N f )
holds for every f ∈ X (I) and a ∈ (0, 1].
It is clear that every N -quasi convex modular is N -quasi semiconvex. We should
recall that the above two concepts were introduced and discussed in details by
Bardaro et. al. [6].
We now consider some appropriate vector subspaces of X(I) by means of a
modular ρ as follows
 
ρ
L (I) := f ∈ X (I) : lim+ ρ (λf ) = 0
λ→0

and
E ρ (I) := {f ∈ Lρ (I) : ρ (λf ) < +∞ for all λ > 0} .
Here, Lρ (I) is called the modular space generated by ρ and E ρ (I) is called the
space of the finite elements of Lρ (I) . Observe that if ρ is N -quasi semiconvex,
then the space
{f ∈ X (I) : ρ (λf ) < +∞ for some λ > 0}
ρ
coincides with L (I). The notions about modulars were introduced in [17] and
widely discussed in [6] (see also [12, 16]).
Now we recall the convergence methods in modular spaces.
Let {fn } be a function sequence whose terms belong to Lρ (I) . Then, {fn } is
modularly convergent to a function f ∈ Lρ (I) iff
(1.1) lim ρ (λ0 (fn − f )) = 0 for some λ0 > 0.
n

Also, {fn } is F -norm convergent (or strongly convergent) to f iff

(1.2) lim ρ (λ (fn − f )) = 0 for every λ > 0.
n

It is known from [16] that (1.1) and (1.2) are equivalent if and only if the modular
ρ satisfies the ∆2 -condition, i.e., there exists a constant M > 0 such that ρ (2f ) ≤
M ρ (f ) for every f ∈ X (I).
In this paper, we will need the following assumptions on a modular ρ:
• if ρ(f ) ≤ ρ(g) for |f | ≤ |g| , then ρ is called monotone,
• if the characteristic function χI of the interval I belongs to Lρ (I) , ρ is
called finite,
 MATRIX SUMMABILITY 283

• if ρ is finite and, for every ε > 0, λ > 0, there exists δ > 0 such that
ρ (λχB ) < ε for any measurable subset B ⊂ I with |B| < δ, then ρ is called
absolutely finite,
• if χI ∈ E ρ (I) , then ρ is called strongly finite,
• ρ is called absolutely continuous provided that there exists α > 0 such that,
for every f ∈ X (I) with ρ (f ) < +∞, the following condition holds: for
every ε > 0 there is δ > 0 such that ρ (αf χB ) < ε whenever B is any
measurable subset of I with |B| < δ.
Observe now that (see [5]) if a modular ρ is monotone and finite, then we have
C(I) ⊂ Lρ (I) . In a similar manner, if ρ is monotone and strongly finite, then
C(I) ⊂ E ρ (I). Some important relations between the above properties may be
found in [4, 6, 14, 17].

2. Korovkin Type Theorems

 
Let A := (An )n≥1 , An = ankj be a sequence of infinite non-negative real
k,j∈N
matrices. For a sequence of real numbers, x = (xj )j∈N , the double sequence
n
Ax := {(Ax)k : k, n ∈ N}
∞
n
ankj xj is called the A-transform of x whenever the series
P
defined by (Ax)k :=
j=1
converges for all k and n. A sequence x is said to be A-summable to L if
∞
X
lim ankj xj = L
k→∞
j=1

uniformly in n ([3], [20]).

If An = A for a matrix A, then A-summability is the ordinary matrix summa-
bility by A. If ankj = k1 for n ≤ j ≤ k + n, (n = 1, 2, . . .) and ankj = 0 otherwise,
then A-summability reduces to almost convergence [13].
Let ρ be a monotone and finite modular on X (I). Assume that D is a set
satisfying C ∞ (I) ⊂ D ⊂ Lρ (I). We can construct such a subset D since ρ is
monotone and finite (see [5]). Assume further that T := {Tn } is a sequence of
positive linear operators from D into X (I) for which there exists a subset XT ⊂ D
containing C ∞ (I) such that
∞
X
(2.1) lim sup ankj ρ (λ (Tj h)) ≤ P ρ (λh) , uniformly in n.
k→∞ j=1

The inequality holds for every h ∈ XT , λ > 0 and for an absolute positive constant
P . Throughout the paper we use the test functions ei defined by

ei (x) = xi (i = 0, 1, 2, . . .).
284 S. KARAKUŞ and K. DEMIRCI

Theorem 2.1. Let A = (An )n≥1 be a sequence of infinite non-negative real

matrices such that
∞
X
(2.2) sup ankj < ∞
n,k j=1

and let ρ be a monotone, strongly finite, absolutely continuous and N -quasi semi-
convex modular on X (I). Let T := {Tj } be a sequence of positive linear operators
from D into X (I) satisfying (2.1). Suppose that
∞
X
(2.3) lim ankj ρ (λ (Tj ei − ei )) = 0, uniformly in n
k→∞
j=1

for every λ > 0 and i = 0, 1, 2. Now, let f be any function belonging to Lρ (I) such
that f − g ∈ XT for every g ∈ C ∞ (I). Then, we have
∞
X
lim ankj ρ (λ0 (Tj f − f )) = 0, uniformly in n
k→∞
j=1

for some λ0 > 0.

Proof. We first claim that

∞
X
(2.4) lim ankj ρ (µ (Tj g − g)) = 0, uniformly in n
k→∞
j=1

for every g ∈ C(I) and every µ > 0. To see this assume that g belongs to C (I)
and µ is any positive number. Then, there exists a constant M > 0 such that
|g (x)| ≤ M for every x ∈ I. Given ε > 0, we can choose δ > 0 such that
|y − x| < δ implies |g (y) − g (x)| < ε where y, x ∈ I. It is easy to see that for all
y, x ∈ I
2M 2
|g (y) − g (x)| < ε + 2 (y − x) .
δ
Since Tj is a positive linear operator, we get
|Tj (g; x) − g (x) |
= |Tj (g (·) − g (x) ; x) + g (x) (Tj (e0 (·) ; x) − e0 (x))|
≤ Tj (|g (·) − g (x)| ; x) + |g (x)| |Tj (e0 (·) ; x) − e0 (x)|
 
2M 2
≤ Tj ε + 2 (· − x) ; x + M |Tj (e0 (·) ; x) − e0 (x)|
δ
2M  2

≤ εTj (e0 (·) ; x) + 2 Tj (· − x) ; x + M |Tj (e0 (·) ; x) − e0 (x)|
δ
≤ ε + (ε + M ) |Tj (e0 (·) ; x) − e0 (x)|
2M
+ 2 [Tj (e2 (·) ; x) − 2e1 (x) Tj (e1 (·) ; x) + e2 (x) Tj (e0 (·) ; x)]
δ
 MATRIX SUMMABILITY 285

2M
≤ ε + (ε + M ) |Tj (e0 (·) ; x) − e0 (x)| + |Tj (e2 (·) ; x) − e2 (x)|
δ2
4M |e1 (x)| 2M e2 (x)
+ 2
|Tj (e1 (·) ; x) − e1 (x)| + |Tj (e0 (·) ; x) − e0 (x)|
δ δ2
2M c2
 
4M c
≤ ε+ ε+M + 2
|Tj (e0 (·) ; x) − e0 (x)| + 2 |Tj (e1 (·) ; x) − e1 (x)|
δ δ
2M
+ 2 |Tj (e2 (·) ; x) − e2 (x)|
δ
where c := max {|a| , |b|}. So, the last inequality gives, for any µ > 0 that
µ |Tj (g; x) − g (x)| ≤ µε + µK |Tj (e0 (·) ; x) − e0 (x)|
+ µK |Tj (e1 (·) ; x) − e1 (x)| + µK |Tj (e2 (·) ; x) − e2 (x)|

2M c2 4M c 2M
 
where K := max ε + M + , , . Applying the modular ρ in the
δ2 δ2 δ2
both-sides of the above inequality, since ρ is monotone, we have
ρ(µ(Tj (g; ·) − g(·)))
≤ ρ (µε + µK |Tj e0 − e0 | + µK |Tj e1 − e1 | + µK |Tj e2 − e2 |) .
So, we may write that
ρ (µ (Tj (g; ·) − g (·))) ≤ ρ (4µε) + ρ (4µK (Tj e0 − e0 ))
+ ρ (4µK (Tj e1 − e1 )) + ρ (4µK (Tj e2 − e2 )) .
Since ρ is N -quasi semiconvex and strongly finite, we have, assuming 0 < ε ≤ 1
ρ (µ (Tj (g; ·) − g (·))) ≤ N ερ (4µN ) + ρ (4µK (Tj e0 − e0 ))
+ ρ (4µK (Tj e1 − e1 )) + ρ (4µK (Tj e2 − e2 )) .
Hence
X∞ ∞
X ∞
X
ankj ρ (µ (Tj (g; ·) − g (·))) ≤ N ερ (4µN ) ankj + ankj ρ (4µK (Tj e0 − e0 ))
j=1 j=1 j=1
(2.5) ∞ ∞
X X
+ ankj ρ (4µK (Tj e1−e1 )) + ankj ρ (4µK (Tj e2−e2 ))
j=1 j=1

By taking limit superior as k → ∞ in the both-sides of (2.5), by using (2.3), we

get
X∞
lim ankj ρ (µ (Tj (g; ·) − g (·))) = 0 uniformly in n
k→∞
j=1

which proves our claim (2.4). Now let f ∈ Lρ (I) satisfying f − g ∈ XT for every
g ∈ C ∞ (I). Since |I| < ∞ and ρ is strongly finite and absolutely continuous, we
can see that ρ is also absolutely finite on X(I) (see [4]). Using these properties of
286 S. KARAKUŞ and K. DEMIRCI

the modular ρ, it is known from [6, 14] that the space C ∞ (I) is modularly dense
in Lρ (I) , i.e., there exists a sequence {gk } ⊂ C ∞ (I) such that
lim ρ (3λ0 (gk − f )) = 0 for some λ0 > 0.
k

This means that, for every ε > 0, there is a positive number k0 = k0 (ε) so that
(2.6) ρ (3λ0 (gk − f )) < ε for every k ≥ k0 .
On the other hand, by the linearity and positivity of the operators Tj , we may
write that
λ0 |Tj f − f | ≤ λ0 |Tj (f − gk0 )| + λ0 |Tj gk0 − gk0 | + λ0 |gk0 − f | .
Applying the modular ρ in the both-sides of the above inequality, since ρ is mono-
tone, we have
ρ (λ0 (Tj f − f )) ≤ ρ (3λ0 (Tj f − gk0 )) + ρ (3λ0 (Tj gk0 − gk0 ))
(2.7)
+ ρ (3λ0 (gk0 − f )) .
Then, it follows from (2.6) and (2.7) that
ρ (λ0 (Tj f − f )) ≤ ρ (3λ0 (Tj f − gk0 )) + ρ (3λ0 (Tj gk0 − gk0 )) + ε.
Hence, using the facts that gk0 ∈ C ∞ (I) and f − gk0 ∈ XT , we have
∞
X
ankj ρ (λ0 (Tj f − f ))
j=1
(2.8) ∞ ∞ ∞
X X X
≤ ankj ρ (3λ0 (Tj f − gk0 )) + ankj ρ (3λ0 (Tj gk0 − gk0 )) + ε ankj .
j=1 j=1 j=1

∞
ankj < B. So, taking
P
From (2.2), there exists a constant B > 0 such that sup
n,k j=1
limit superior as k → ∞ in the both-sides of (2.8), from (2.1) and (2.2) we obtain
that
∞
X
lim sup ankj ρ (λ0 (Tj f − f ))
k j=1
∞
X ∞
X
≤ ε lim sup ankj + P ρ (3λ0 (f − gk0 )) + lim sup ankj ρ (3λ0 (Tj gk0 − gk0 )) ,
k j=1 k j=1

which gives
∞
X
lim sup ankj ρ (λ0 (Tj f − f ))
k j=1
(2.9) ∞
X
≤ ε (B + P ) + lim sup ankj ρ (3λ0 (Tj gk0 − gk0 )) .
k j=1
 MATRIX SUMMABILITY 287

By (2.4), since
∞
X
lim ankj ρ (3λ0 (Tj gk0 − gk0 )) = 0, uniformly in n
k
j=1

we get
∞
X
(2.10) lim sup ankj ρ (3λ0 (Tj gk0 − gk0 )) = 0, uniformly in n.
k j=1

Combining (2.9) with (2.10), we conclude that

∞
X
lim sup ankj ρ (λ0 (Tj (f ; x) − f (x))) ≤ ε (B + P ) .
k j=1

Since ε > 0 was arbitrary, we find

X∞
lim sup ankj ρ (λ0 (Tj f − f )) = 0 uniformly in n.
k j=1
∞
ankj ρ (λ0 (Tj (f ; x) − f (x))) is non-negative for all k, n ∈ N,
P
Furthermore, since
j=1
we can easily show that
∞
X
lim ankj ρ (λ0 (Tj f − f )) = 0, uniformly in n
k
j=1

which completes the proof. 

If the modular ρ satisfies the ∆2 -condition, then one can get the following result
from Theorem 2.1 at once.
Theorem 2.2. Let A = (An )n≥1 be a sequence of infinite non-negative real
matrices such that
∞
X
sup ankj < ∞,
n,k j=1

and T := {Tn }, ρ be the same as in Theorem 2.1. If ρ satisfies the ∆2 -condition,

then the following statements are equivalent:
∞
ankj ρ (λ (Tj ei − ei )) = 0 uniformly in n for every λ > 0 and i =
P
(a) lim
k j=1
0, 1, 2,
∞
ankj ρ (λ (Tj f − f )) = 0 uniformly in n for every λ > 0 provided
P
(b) lim
k j=1
that f is any function belonging to Lρ (I) such that f − g ∈ XT for every
g ∈ C ∞ (I).
If An = I, identity matrix, then the condition (2.1) reduces to
(2.11) lim sup ρ (λ (Tj h)) ≤ P ρ (λh)
j
288 S. KARAKUŞ and K. DEMIRCI

for every h ∈ XT , λ > 0 and for an absolute positive constant P. In this case,
the next results which were obtained by Bardaro and Mantellini [5] immediately
follows from our Theorems 2.1 and 2.2.
Corollary 2.3. Let ρ be a monotone, strongly finite, absolutely continuous
and N -quasi semiconvex modular on X (I). Let T := {Tj } be a sequence of
positive linear operators from D into X (I) satisfying (2.11). If {Tj ei } is strongly
convergent to ei for each i = 0, 1, 2, then {Tj f } is modularly convergent to f
provided that f is any function belonging to Lρ (I) such that f − g ∈ XT for every
g ∈ C ∞ (I).
Corollary 2.4. T := {Tj } and ρ be the same as in Corollary 2.3. If ρ satisfies
the ∆2 -condition, then the following statements are equivalent:
(a) {Tj ei } is strongly convergent to ei for each i = 0, 1, 2,
(b) {Tj f } is strongly convergent to f provided that f is any function belonging
to Lρ (I) such that f − g ∈ XT for every g ∈ C ∞ (I).

3. Application

Take I = [0, 1] and let ϕ : [0, ∞) → [0, ∞) be a continuous function for which the
following conditions hold:
• ϕ is convex,
• ϕ (0) = 0, ϕ (u) > 0 for u > 0 and limu→+∞ ϕ (u) = ∞.
Hence, consider the functional ρϕ on X(I) defined by
Z 1
ρϕ (f ) := ϕ (|f (x)|) dx for f ∈ X (I) .
0
In this case, ρϕ is a convex modular on X (I) , which satisfies all assumptions listed
in Section 1 (see [5]). Consider the Orlicz space generated by ϕ as follows
Lρϕ (I) := {f ∈ X (I) : ρϕ (λf ) < +∞ for some λ > 0} .
Then, consider the following classical Bernstein-Kantorovich operator U := {Un }
on the space Lρϕ (I) (see [5]) which is defined by
j   Z (k+1)/(j+1)
X j k j−k
Uj (f ; x) := x (1 − x) (j + 1) f (t) dt for x ∈ I.
k k/(j+1)
k=0

Observe that the operators Uj map the Orlicz space Lρϕ (I) into itself. Moreover,
the property (2.11) is satisfied with the choice of XU := Lρϕ (I). Then, by Corol-
lary 2.3, we know that, for every function f ∈ Lρϕ (I) such that f − g ∈ XU for
every g ∈ C ∞ (I), {Uj f } is modularly
 convergent
 to f.
Assume that A := (An )n≥1 = ankj is a sequence of infinite matrices
k,j∈N
1
defined by = if n ≤ j ≤ k + n, (n = 1, 2, . . .), and ankj = 0 otherwise, then
ankj k
A−summability reduces to almost convergence. Define s = (sn ) of the form
(3.1) 0101 . . . 0101; 001001 . . . 001; 00010001 . . . 0001; . . .
→n1 terms← → n2 terms ← → n2 terms ←
 MATRIX SUMMABILITY 289

where n1 is a multiple of 2, n2 is a multiple of 6, n3 is a multiple of 1, 2, . . . and nk

is a multiple of k (k + 1). So s is almost convergent to zero (see [15]). However,
the sequence {sn } is not convergent to zero. Then, using the operators Uj , we
define the sequence of positive linear operators V := {Vn } on Lρϕ (I) as follows:
(3.2) Vj (f ; x) = (1 + sj ) Uj (f ; x) for f ∈ Lρϕ (I) , x ∈ [0, 1] and j ∈ N,
where s = {sj } is the same as in (3.1). By [5, Lemma 5.1], for every h ∈ XV :=
Lρϕ (I), all λ > 0 and for an absolute positive constant P, we get

ρϕ (λVj h) = ρϕ (λ (1+sj ) Uj h) ≤ ρϕ (2λUj h)+ρϕ (2λsj Uj h)

= ρϕ (2λUj h) + sj ρϕ (2λUj h) = (1+sj ) ρϕ (2λUj h) ≤ (1+sj ) P ρϕ (2λh) .
Then, we get  
n+k
1 X
lim sup sup ρϕ (λVj h) ≤ P ρϕ (2λh) .
k n k
j=n

So, the condition (2.1) works for our operators Vn given by (3.2) with the choice
of XV = XU = Lρϕ (I).
Now, we show that condition (2.3) in the Theorem 2.1 holds.
First observe that
Vj (e0 ; x) = 1 + sj ,
 
jx 1
Vj (e1 ; x) = (1 + sj ) + ,
j + 1 2 (j + 1)
!
j (j − 1) x2 2jx 1
Vj (e2 ; x) = (1 + sj ) 2 + 2 + 2 .
(j + 1) (j + 1) 3 (j + 1)
So, for any λ > 0, we can see, that
λ |Vj (e0 ; x) − e0 (x)| = λ |1 + sj − 1| = λsj ,
which implies
Z 1
ρϕ (λ (Vj e0 − e0 )) = ρϕ (λsj ) = ϕ (λsj ) dx = ϕ (λsj ) = sj ϕ (λ)
0

because of the definition of (sj ). Since (sj ) is almost convergent to zero, we get
 
n+k
1 X
lim sup sup ρϕ (λ (Vj e0 − e0 )) = 0 for every λ > 0,
k n k
j=n

which guarantees that (2.3) holds true for i = 0. Also, since

 
j jsj 1 sj
λ |Vj (e1 ; x) − e1 (x)| = λ x + −1 + +
j+1 j+1 2(j + 1) 2(j + 1)
 
j jsj 1 sj
≤ λ |x| −1 + + +
j+1 j+1 2(j + 1) 2(j + 1)
290 S. KARAKUŞ and K. DEMIRCI
 
1 2jsj sj 1
≤λ + + +
(j + 1) 2 (j + 1) 2 (j + 1) 2(j + 1)
  
3 2j + 1
≤λ + sj ,
2(j + 1) 2 (j + 1)
we may write that
    
2j + 1 3
ρϕ (λ (Vj e1 − e1 )) ≤ ρϕ λ sj +
2 (j + 1) 2(j + 1)
    
2j + 1 3λ
≤ sj ρϕ λ + ρϕ
j+1 j+1
 
by the definitions of (sj ) and ρϕ . Since 2j+1 is convergent, it is bounded. So
 j+1 
there exists a constant M > 0 such that 2j+1 j+1 ≤ M for every j ∈ N. Then using
ϕ
the monotonicity of ρ , we have
  
ϕ 2j + 1
ρ λ ≤ ρϕ (λM )
j+1
for any λ > 0, which implies

  
3λ 3λ
ρϕ (λ (Vj e1 − e1 )) ≤ sj ρϕ (λM ) + ρϕ = sj ϕ (λM ) + ϕ .
j+1 j+1
   
3λ 3λ
Since ϕ is continuous, we have lim ϕ j+1 = ϕ lim j+1 = ϕ(0) = 0. So, we get
j j
   
3λ 3λ
ϕ j+1 is almost convergent to zero. Using s and ϕ j+1 are almost convergent
to zero, we obtain
   
n+k n+k
X  
1 X 1 3λ
lim supsup ρϕ (λ (Vj e1 −e1)) ≤ lim sup sup sj ϕ (λM)+ϕ 
k n k k n k j+1
j=n j=n
   
n+k n+k
X  3λ 
1 X 1
= ϕ (λM ) lim sup sup sj  +lim sup sup ϕ  = 0.
k n k k n k j+1
j=n j=n

Finally, since
j (j − 1) 2jx 1
λ |Vj (e2 ; x) − e2 (x)| = λ x2 2 + 2 + 2
(j + 1) (j + 1) 3 (j + 1)
j (j − 1) x2 2jx 1
+ sj 2 + sj 2 + sj 2 − x2
(j + 1) (j + 1) 3 (j + 1)
j (j − 1) j (j − 1)
≤ λx2 2 − 1 + x2 sj 2
(j + 1) (j + 1)
!
2j 2j 1 1
+ |x| 2 + sj 2 + 2 +sj 2
(j + 1) (j + 1) 3 (j + 1) 3 (j + 1)
 MATRIX SUMMABILITY 291

(
3j + 1 j (j − 1) 2j 2j
≤λ 2 + sj 2 + 2 + sj 2
(j + 1) (j + 1) (j + 1) (j + 1)
)
1 1
+ 2 + sj 2
3 (j + 1) 3 (j + 1)
( !)
15j + 4 3j 2 + 3j + 1
≤λ 2 + sj 2 .
3 (j + 1) 3 (j + 1)
 
3j 2 +3j+1
Since 3(j+1)2
is convergent, it is bounded. So there exists a constant K > 0
3j 2 +3j+1
such that 3(j+1)2 ≤ K for every j ∈ N. Then using the monotonicity of ρϕ and
the definition of (sj ), we have
!! !!
ϕ ϕ 15j + 4 ϕ 3j 2 + 3j + 1
ρ (λ (Vj e2 − e2 )) ≤ ρ 2λ 2 +ρ 2λsj 2
3 (j + 1) 3 (j + 1)
!!
ϕ 30j + 8
≤ρ λ 2 + ρϕ (2λsj K) ,
3 (j + 1)
where which yields
!!
30j + 8
(3.3) ρϕ (λ (Vj e2 − e2 )) ≤ ϕ λ 2 + sj ϕ (2λK)
3 (j + 1)
   
30j+8 30j+8
Since ϕ is continuous, we have lim ϕ λ 3(j+1) 2 = ϕ λ lim 2 = ϕ(0) = 0.
j j 3(j+1)
   
30j+8 30j+8
So, we get ϕ λ 3(j+1) 2 is almost convergent to zero. Using s and ϕ λ 3(j+1) 2

are almost convergent to zero, it follows from (3.3) that

 
n+k
1 X
lim sup sup ρϕ (λ (Vj e2 − e2 )) = 0 uniformly in n for every λ > 0.
k n k
j=n

Our claim (2.3) holds true for each i = 0, 1, 2 and for any λ > 0. So, we can say that
our sequence V := {Vj } defined by (3.2) satisfy all assumptions of Theorem 2.1.
Therefore, we conclude that
 
n+k
1 X
lim sup sup ρϕ (λ0 (Vj f − f )) = 0 uniformly in n for some λ0 > 0
k n k
j=n

holds for every f ∈ Lρϕ (I) such that f − g ∈ XV = Lρϕ (I) for every g ∈ C ∞ (I).
However, since (sj ) is not convergent to zero, it is clear that {Vj f } is not
modularly convergent to f . So, Corollary 2.3 does not work for the sequence
V := {Vj }.
292 S. KARAKUŞ and K. DEMIRCI

References
1. Atlıhan Ö. G. and Orhan C., Matrix summability and positive linear operators, Positivity
11 (2007), 387–389.
2. , Summation process of positive linear operators, Computers & Mathematics with
Applications, 56 (2008), 1188–1195.
3. Bell H. T., Order summability and almost convergence, Proc. Am. Math. Soc. 38 (1973),
548–553.
4. Bardaro C. and Mantellini I., Approximation properties in abstract modular spaces for a
class of general sampling-type operators, Appl. Anal. 85 (2006), 383–413.
5. , Korovkin’s theorem in modular spaces, Commentationes Math. 47 (2007), 239–253.
6. Bardaro C., Musielak J. and Vinti G., Nonlinear Integral Operators and Applications, de
Gruyter Series in Nonlinear Analysis and Appl. Vol. 9, Walter de Gruyter Publ., Berlin,
2003.
7. Bojanic R. and Cheng F., Estimates for the rate of approximation of functions of bounded
variation by Hermite-Fejer polynomials, Proceedings of the conference of Canadian Math.
Soc. 3 (1983), 5–17.
8. Bojanic R. and Khan M. K., Summability of Hermite-Fejer interpolation for functions of
bounded variation, J. Nat. Sci. Math. 32(1) (1992), 5–10.
9. Devore R. A., The Approximation of Continuous Functions by Positive Linear Operators,
Lecture Notes in Mathematics, Springer-Verlag, 293, Berlin 1972.
10. Gadžiev A. D., The convergence problem for a sequence of positive linear operators on
unbounded sets, and theorems analogous to that of P. P. Korovkin, Soviet Math. Dokl.
15(5) (1974), 1433–1436.
11. Korovkin P. P., Linear Operators and Approximation Theory, Hindustan Publ. Corp., Delhi,
1960.
12. Kozlowski W. M., Modular Function Spaces, Pure Appl. Math., 122, Marcel Dekker, Inc.,
New York, 1988.
13. Lorentz G. G., A contribution to the theory of divergent sequences, Acta Math. 80 (1948),
167–190.
14. Mantellini I., Generalized sampling operators in modular spaces, Commentationes Math.,
38 (1998), 77–92.
15. Miller H. I. and Orhan C., On almost convergent and statistically convergent subsequences,
Acta Math. Hungar 93(1-2) (2001), 135–151.
16. Musielak J., Orlicz Spaces and Modular Spaces, Lecture Notes in Mathematics, Vol. 1034,
Springer-Verlag, Berlin, 1983.
17. Musielak J., Nonlinear approximation in some modular function spaces I, Math. Japon. 38
(1993), 83–90.
18. Nishishiraho T., Quantitative theorems on linear approximation processes of convolution
operators in Banach spaces, Tôhoku Math. J. 33 (1981), 109–126.
19. , Convergence of positive linear approximation processes, Tôhoku Math. J. 35
(1983), 441–458.
20. Stieglitz M., Eine verallgemeinerung des begriffs der fastkonvergenz, Math. Japon. 18 (1973)
53–70.
21. Swetits J. J., On summability and positive linear operators, J. Approx. Theory 25 (1979),
186–188.

S. Karakuş, Sinop University, Faculty of Sciences and Arts, Department of Mathematics, 57000,
Sinop, Turkey, e-mail: [email protected]

K. Demirci, Sinop University, Faculty of Sciences and Arts, Department of Mathematics, 57000,
Sinop, Turkey, e-mail: [email protected]