Applied Mathematics Letters 20 (2007) 10461051

www.elsevier.com/locate/aml
Weighted L
p
-approximation with positive linear operators
on unbounded sets
A.D. Gadjiev
a
, A. Aral
b,
a
Institute of Mathematics and Mechanics, National Academy of Sciences of Azerbaijan, 9 F. Agayev Str., Baku, Azerbaijan
b
Kirikkale University, Department of Mathematics, 71450 Yahsihan, Kirikkale, Turkey
Received 10 November 2005; received in revised form 21 November 2006; accepted 15 December 2006
Abstract
In this work we state and prove a Korovkin type theorem for the weighted space L
p,
(R) and also its n-dimensional analogue
for the weighted space L
p,
(R
n
).
c 2007 Elsevier Ltd. All rights reserved.
Keywords: Positive linear operators; Korovkins theorem; Test functions; Weighted space
1. Introduction
The theorem of Banach and Steinhaus gives a rst result concerning sets of test functions for norm convergence of
the sequence of L
n
( f ; x) to f (x), where L
n
is the sequence of bounded linear operators. Korovkin showed that this
set of test functions can be reduced to a nite set if the operators under consideration are positive. The fundamental
Korovkin theorem delivers such conditions in the case of space C(a, b) which consists of continuous functions on
nite interval [a, b] and bounded on the whole real axis. This theorem has been extended in several directions (see [1]
and the literature cited there). These Korovkin type theorems for the functions in L
p
(, ) can be found in Curtis
Jr. [2] and Dzjadyk [3]. Recently, in [4,5], weighted Korovkin type theorems have been proven by Gadjiev et al.
in the space of locally integrable functions on the whole real axis R. Some weighted Korovkin type theorems can
also be found in the space of weighted continuous functions on the semi-real axis in [8]. Recently, in [6], some
applications of weighted Korovkin type theorems to the BernsteinChlodowsky polynomials were given in detail.
Moreover, Dzjadyk [3] solved a problem of approximation of the function f L
p
(a, b) by the sequence of positive
linear operators L
n
. It states that the norm convergence lim
n
L
n
f f
L
p
(a,b)
= 0 occurs for all f L
p
(a, b)
if and only if it does so for three particular test functions
m
(t ) = t
m
, m = 0, 1, 2, and there exists a constant K > 0
such that L
n
K for all n N.
Note that the paper [7] is devoted to Korovkin type approximation theorems in the spaces C(a, b) and L
p
(a, b),
via statistical convergence.

Corresponding author.
E-mail addresses: akif gadjiev@mail.az (A.D. Gadjiev), aliaral73@yahoo.com (A. Aral).
0893-9659/$ - see front matter c 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.aml.2006.12.007
A.D. Gadjiev, A. Aral / Applied Mathematics Letters 20 (2007) 10461051 1047
Here we mention a further interesting problem concerned with questions on the order of approximation of functions
by positive linear operators in L
p
(a, b). The quantitative form of the Korovkin theorem has shown that the rate of
convergence of L
n
( f ; x) to f (x) is estimated in terms of rates of convergence of L
n
(
m
; x) to
m
(x), m = 0, 1, 2
(see [9,10]).
At times we do not know the statement of the Korovkin type theorem in the space L
p
(R). Our goal in this work
is to obtain a Korovkin type approximation by positive linear operators result for the functions in the weighted space
L
p,
(R). First, in this work, we shall give an analogue of the Korovkin theorem in the weighted space L
p,
(R).
Secondly, we shall prove a Korovkin type theorem in the n-dimensional weighted space L
p,
(R
n
).
2. Main results
For a xed p [1, ), let be a positive continuous function on the whole real axis satisfying the condition
_
R
t
2p
(t ) dt < . (2.1)
We denote by L
p,
(R) (1 p < ) the linear space of measurable, p-absolutely integrable functions on R with
respect to the weight function , i.e.
L
p,
(R) =
_
f : R R; f
p,
:=
__
R
| f (t ) |
p
(t ) dt
_1
p
<
_
.
The minimum and maximum values of the function on nite intervals will be denoted by
min
and
max
respectively.
Theorem 1. Let (L
n
)
nN
be a uniformly bounded sequence of positive linear operators from L
p,
(R) into L
p,
(R),
satisfying the conditions
lim
n
_
_
_L
n
_
t
i
, x
_
x
i
_
_
_
p,
= 0, i = 0, 1, 2. (2.2)
Then for every f L
p,
(R), we have
lim
n
L
n
f f
p,
= 0.
Proof. For any positive number A, let
A
1
(t ) be characteristic function of the interval [A, A] and
A
2
(t ) =
1
A
1
(t ). Given > 0, we can choose a number A so large that
__
R
| f (t ) |
p
|
A
2
(t ) |
p
(t ) dt
_1
p
< ,
that is,
f
A
2

p,
< . (2.3)
On the other hand, linearity of the operators L
n
yields the inequality
L
n
f f
p,

_
_
_L
n
_

A
1
f
_

A
1
f
_
_
_
p,
+
_
_
_L
n
_

A
2
f
_

A
2
f
_
_
_
p,
= I

n
+ I

n
. (2.4)
Firstly, we estimate I

n
. Since L
n
is a uniformly bounded sequence, there exists a positive constant K independent of
n such that
L
n
f
p,
K f
p,
.
1048 A.D. Gadjiev, A. Aral / Applied Mathematics Letters 20 (2007) 10461051
This estimate and (2.3) imply
I

n

_
_
_L
n
_

A
2
f
__
_
_
p,
+
A
2
f
p,
(K +1)
A
2
f
p,
< (K +1) . (2.5)
Moreover for every function f L
p,
(R) the inequality

A
1
f
p

1/p
min
f
p,
implies L
p,
(R) L
p
(A, A). Since the continuous functions are dense in L
p
(A, A), for each

> 0, there exists

a continuous function on [A, A] satisfying the condition (x) = 0 for |x| > A such that
( f )
A
1

p
<

/(K +1)
1/p
max
.
We can write from these inequalities and (2.5)
I

n
L
n
( f )
A
1

p,
+
_
_
_L
n
_

A
1
_

A
1
_
_
_
p,
+ ( f )
A
1

p,

_
_
_L
n
_

A
1
_

A
1
_
_
_
p,
+

. (2.6)
Since
A
1
2

A
1
= 0 for some A
1
> A, one can write the following inequality for the rst term of (2.6):
_
_
_L
n
_

A
1
_

A
1
_
_
_
p,

_
_
_
_
L
n
_

A
1
_

A
1
_

A
1
1
_
_
_
p,
+
_
_
_
A
1
2
L
n
_

A
1
__
_
_
p,
. (2.7)
Dening M

= max
t R
| (t ) |
A
1
(t ), we can write
_
_
_
A
1
2
L
n
_

A
1
__
_
_
p,
=
__
|t |>A
1

L
n
_

A
1
; t
_

p
(t ) dt
_1
p
M

__
|t |>A
1
|L
n
(1; t ) 1|
p
(t ) dt
_1
p
+ M

__
R

A
1
2
(t ) (t ) dt
_1
p
.
According to the hypotheses of the theorem, since L
1
(R) we can choose the number A
1
such that
__
R

A
1
2
(t )(t )dt
_1
p
<

.
Therefore we have
_
_
_
A
1
2
L
n
_

A
1
__
_
_
p,
M

L
n
(1; x) 1
p,
+

. (2.8)
Substituting (2.8) into (2.7) we obtain
_
_
_L
n
_

A
1
_

A
1
_
_
_
p,

_
_
_
_
L
n
_

A
1
_

A
1
_

A
1
1
_
_
_
p,
+ M

L
n
(1; x) 1
p,
+

.
Using this inequality in (2.6), we get
I

n
2

+ M

L
n
(1; x) 1
p,
+
_
_
_
_
L
n
_

A
1
_

A
1
_

A
1
_
_
_
p,
(2.9)
and it is sufcient to show that the last term tends to zero as n (the condition (2.2) shows that L
n
(1; x)
1
p,
<

for a large n). Since the function

A
1
is continuous on [A, A], given any

> 0 there is a > 0 such

that
| (t )
A
1
(t ) (x)
A
1
(x) | <

+2M

(t x)
2

2
.
A.D. Gadjiev, A. Aral / Applied Mathematics Letters 20 (2007) 10461051 1049
Hence
_
_
_
_
L
n
_

A
1
_

A
1
_

A
1
1
_
_
_
p,

_
_
_
_
L
n
_
| (t )
A
1
(t ) (x)
A
1
(x) |; x
__

A
1
1
(x)
_
_
_
p,
+ (x)
A
1
1
(x) (L
n
(1; x) 1)
p,

L
n
(1, x)
p,
+
2M

2
_
_
_
_L
n
_
t
2
; x
_
x
2
_
_
_
p,
+ 2AL
n
(t ; x) x
p,
+ A
2
L
n
(1; x) 1
p,
_
+M

L
n
(1; x) 1
p,

K
1/p
1
+
2M

2
(1 + A)
2
max
i =0,1,2
L
n
(t
i
; x) x
i

p,
+

.
By the condition (2.2) for a large n
max
i =0,1,2
L
n
(t
i
; x) x
i

p,

2
2M

(1 + A)
2
.
Therefore, by (2.9), we have
I

n
<
_
5 + K
1/p
1
_

. (2.10)
Using (2.10) and (2.5) in (2.4) we obtain
L
n
f f
p,

_
5 + K
1/p
1
_

+(K +1)
and the proof is completed.
Now we establish an analogue of Theorem 1 for the space of function of several variables. Note that the n-
dimensional analogue of Korovkins theorem was proved by

Sa skin [11].
Let be a positive continuous function in R
n
, satisfying the condition
_
R
n
|t |
2p
(t ) dt <
and for 1 p <
L
p,
_
R
n
_
=
_
f : R
n
R; f
p,
=
__
R
n
| f (t ) |
p
(t ) dt
_
1/p
<
_
.
Theorem 2. Let (L
n
)
nN
be a uniformly bounded sequence of positive linear operators from L
p,
(R
n
) into itself,
satisfying the conditions
lim
n
L
n
(1; x) 1
p,
= 0,
lim
n
L
n
(t
i
; x) x
i

p,
= 0, i = 1, n,
lim
n
_
_
_L
n
_
|t |
2
; x
_
|x|
2
_
_
_
p,
= 0.
Then for any f L
p,
(R
n
)
lim
n
L
n
f f
p,
= 0.
Proof. As in the proof of Theorem 1 we can choose so large a number A that for each > 0
f
A
2

p,
< ,
1050 A.D. Gadjiev, A. Aral / Applied Mathematics Letters 20 (2007) 10461051
where
A
1
is the characteristic function of the ball |x| A and
A
2
(t ) = 1
A
1
(t ).
Also given
1
> 0 we can nd a continuous function on |x| A satisfying the condition (x) = 0 for |x| > A
such that
( f )
A
1

p,
<

1
(K +1)(max
|t |A
(t ))
1/p
where K is a constant satisfying L
n

p,
K. Thus we have
L
n
f f
p,
(K +1) +
1
+
_
_
_L
n
_

A
1
_

A
1
_
_
_
p,
(2.11)
and for a some A
1
> A
_
_
_L
n
_

A
1
_

A
1
_
_
_
p,

_
_
_
_
L
n
_

A
1
_

A
1
_

A
1
1
_
_
_
p,
+ M

L
n
(1; x) 1
p,
+ M

A
1
2

p,
.
Choosing A
1
such that

A
1
2

p,
<

1
M

and taking into account that for a large n

L
n
(1; x) 1
p,
<

1
M

by (2.11), we obtain
L
n
f f
p,
(K +1) +3
1
+
_
_
_
_
L
n
_

A
1
_

A
1
_

A
1
1
_
_
_
p,
. (2.12)
Furthermore for x, t R
n
|(t )
A
1
(t ) (x)
A
1
(x)| <
1
+2M

|t x|
2

2
and we obtain from (2.12)
L
n
f f
p,
(K +1) +4
1
+
1
K
1/p
1
+
2M

2
(1 + A)
2

_
_
_
_L
n
_
|t |
2
; x
_
|x|
2
_
_
_
p,
+
n

i =0
L
n
(t
i
; x) x
i

p,
+L
n
(1; x) 1
p,
_
.
By the conditions of the theorem, we have
L
n
(|t x|
2
, x)
p,
(1 + A)
2
_
_
_
_L
n
_
|t |
2
; x
_
|x|
2
_
_
_
p,
+
n

i =0
L
n
(t
i
; x) x
i

p,
+ L
n
(1; x) 1
p,
_
<

2
2M

(1 + A)
2
for a large n.
Therefore, for a large n, we have
L
n
f f
p,
(K +1) +
_
5 + K
1/p
1
_

1
and the proof is completed.
A.D. Gadjiev, A. Aral / Applied Mathematics Letters 20 (2007) 10461051 1051
References
[1] F. Altomare, M. Campiti, Korovkin-Type Approximation Theory and its Application, in: Walter De Gruyter Studies in Mathematics, vol. 17,
de Gruyter & Co., Berlin, New York, 1994.
[2] P.C. Curtis Jr., The degree of approximation by positive convolution operators, Michigan Math. J. 12 (2) (1965) 153160.
[3] V.K. Dzjadyk, Approximation of functions by positive linear operators and singular integrals, (Russ.) Mat. Sb. (N,S) 112 (70) (1966) 508517.
[4] A.D. Gadjiev, R.O. Efendiyev, E. Ibikli, On Korovkin type theorem in the space of locally integrable functions, Czechoslovak Math. J. 53
128 (1) (2003) 4553.
[5] A.D. Gadjiev, E. Ibikli, On non-existence of Korovkins theorem in the space of L
p
-locally integrable functions, Turkish J. Math. 26 (2002)
207214.
[6] A.D. Gadjiev, R.O. Efendiyev, E. Ibikli, Generalized Bernstein Chlodowsky polynomials, Rocky Mountain J. Math. 28 (4) (1998) 12671277.
[7] A.D. Gadjiev, C. Orhan, Some approximation theorems via statistical convergence, Rocky Mountain J. Math. 32 (1) (2002) 129138.
[8] A.D. Gadjiev, On Korovkin type theorems, Mat. Zametki 20 (5) (1976) 781786.
[9] J.J. Swetits, B. Wood, Quantitative estimates for L
p
approximation with positive linear operators, J. Approx. Theory 38 (1983) 8189.
[10] J.J. Swetits, B. Wood, On the degree of L
p
approximation with positive linear operators, J. Approx. Theory 87 (2) (1996) 239241.
[11] J.A.

Sa skin, Korovkin system in space of continuous functions, Amer. Math. Soc. Transl. Ser. 2 54 (1966) 125144.