Georgian Mathematical Journal Volume 16 (2009), Number 4, 693–704
ON THE RATES OF CONVERGENCE OF CHLODOVSKY–DURRMEYER OPERATORS AND THEIR ´ BEZIER VARIANT HARUN KARSLI AND PAULINA PYCH-TABERSKA
Abstract. We consider the B´ezier variant of Chlodovsky–Durrmeyer operators Dn,α for functions f measurable and locally bounded on the interval [0, ∞). By using the Chanturia1 modulus of variation we estimate the rate of pointwise convergence of (Dn,α f ) (x) at those x > 0 at which the one-sided limits f (x+), f (x−) exist. In the special case α = 1 the recent result of [14] concerning the Chlodovsky–Durrmeyer operators Dn is essentially improved and extended to more general classes of functions. 2000 Mathematics Subject Classification: 41A25, 41A35, 41A36. Key words and phrases: Rate of convergence, Chlodovsky–Durrmeyer operator, B´ezier basis, Chanturia modulus of variation, p-th power variation.
1. Introduction It is well-known that the classical Bernstein–Durrmeyer operators (see [8]) Mn applied to f are defined as (Mn f ) (x) = (n + 1)
n X
Z1 f (t)Pn,k (t) dt, 0 ≤ x ≤ 1,
Pn,k (x)
k=0
(1)
0
µ ¶ n k where Pn,k (x) = x (1−x)n−k . Several researchers have studied approximak tion properties of the operators Mn (see, e.g., [7], [10]) for functions of bounded variation defined on the interval [0, 1]. Not long ago Zeng and Chen [22] defined the B´ezier variant of Durrmeyer operators as (Mn,α f ) (x) = (n + 1)
n X k=0
(α)
Z1 (α) Qn,k
(x)
f (t)Pn,k (t) dt, 0
α α where Qn,k (x) = Jn,k (x)−Jn,k+1 (x) and Jn,k (x) =
n P
Pn,j (x) for k = 0, 1, . . . , n,
j=k
Jn,n+1 (x) = 0 are the B´ezier basis functions (introduced by P. B´ezier [4]), and estimated the rate of convergence of Mn,α f for functions of bounded variation on the interval [0, 1]. 1
The proper form of the family name sometimes written as Chanturiya.
c Heldermann Verlag www.heldermann.de ISSN 1072-947X / $8.00 / °
694
H. KARSLI AND P. PYCH-TABERSKA
Let Xloc [0, ∞) be the class of all complex valued functions measurable and locally bounded on the interval [0, ∞). For f ∈ Xloc [0, ∞) the Chlodovsky– Durrmeyer operators Dn are defined as µ ¶ Zbn µ ¶ n n+1X x t (Dn f ) (x) = Pn,k f (t)Pn,k dt, 0 ≤ x ≤ bn , (2) bn k=0 bn bn 0
where (bn ) is a positive increasing sequence with the properties bn lim bn = ∞ and lim = 0. (3) n→∞ n→∞ n Ibikli and Karsli [14] estimated the rate of convergence of Chlodovsky–Durrmeyer operators for functions in BV [0, ∞). We should also mention the recent articles which devoted to the investigation of Chlodovsky operators (for a Voronovskaya type theorem see an earlier paper by Albrycht and Radecki [2]), the B´ezier variant of Chlodovsky operators and Chlodovsky–Durrmeyer operators for functions in BV [0, ∞) and DBV [0, ∞) are due to Karsli [15], Karsli and Ibikli [16, 17], Butzer and Karsli [5]. For f ∈ Xloc [0, ∞) and α ≥ 1, we introduce the B´ezier variant of Chlodovsky– Durrmeyer operators Dn,α as follows: n
n + 1 X (α) Q (Dn,α f ) (x) = bn k=0 n,k
µ
x bn
¶ Zbn
µ f (t)Pn,k
0
t bn
¶ dt,
0 ≤ x ≤ bn .
(4)
Obviously, Dn,α is a positive linear operator and (Dn,α 1) (x) = 1. In particular, when α = 1 the operators (4) reduce to the operators (2). Recently Agratini [1], Aniol and Pych-Taberska [3], Pych-Taberska [21] and Gupta [11, 12] have investigated the rate of pointwise convergence for Kantorovich and Durrmeyer type Baskakov–B´ezier and B´ezier operators using a different approach. They have proved their theorems in terms of the Chanturia modulus of variation, which is a generalization of the classical Jordan variation. It is useful to point out that a deeper analysis of the Chanturia modulus of variation can be found in [6], but actually the modulus of variation was introduced for the first time by Lagrange [18]. Although the Chanturia modulus of variation was defined as a generalization of the classical variation nearly four decades years ago, it was not used to a sufficient extent to solve the problem mentioned above. The paper is concerned with the rate of pointwise convergence of the operators (4) when f belong to Xloc [0, ∞). Using the Chanturia modulus of variation defined in [6], we examine the rate of pointwise convergence of (Dn,α f ) (x) at the points of continuity and at the first kind discontinuity points of f . For some important papers on different operators related to the present study we refer the readers to Gupta et al. [9, 20] and Zeng and Piriou [23]. It is necessary to point out that in the present paper we extend and improve the result of [14] for Chlodovsky–Durrmeyer operators. We begin by giving
ON THE RATES OF CONVERGENCE
695
Definition 1. Let f be a bounded function on a compact interval I = [a, b]. The modulus of variation νn (f ; [a, b]) of a function f is defined for nonnegative integers n as ν0 (f ; [a, b]) = 0 and for n ≥ 1 as n−1 X νn (f ; [a, b]) = sup |f (x2k+1 ) − f (x2k )| , Πn
k=0
where Πn is an arbitrary system of n disjoint intervals (x2k , x2k+1 ), k = 0, 1, . . . , n − 1, i.e., a ≤ x0 < x1 ≤ x2 < x3 ≤ · · · ≤ x2n−2 < x2n−1 ≤ b. The modulus of variation of any function is a non-decreasing function of n. Some other properties of this modulus can be found in [6]. If f ∈ BVp (I), p ≥ 1, i.e., if f is of p-th bounded power variation on I, then for every k ∈ N, νk (f ; I) ≤ k 1−1/p Vp (f, I), (5) where Vp (f, I) denotes the total p-th power variation of f on I, defined as ³P ´1/p the upper bound of the set of numbers |f (kj ) − f (lj )|p over all finite j
systems of non-overlapping intervals (kj , lj ) ⊂ I. p We also consider the class BVloc [0, ∞), p ≥ 1, consisting of all functions of bounded p-th power variation on every compact interval I ⊂ [0, ∞). In the sequel it will be always assumed that the sequence (bn ) satisfies the fundamental conditions (3). The symbol [a] will denote the greatest integer not greater than a. Theorem 1. Let f ∈ Xloc [0, ∞) and let the one-sided limits f (x+), f (x−) exist at a fixed point x ∈ (0, ∞). Then, for all integers n such that bn > 2x and 4bn ≤ n one has ¯ ¯ p ¯ ¯ f (x+) + αf (x−) ¯(Dn,α f ) (x) − ¯ ≤ 2v1 (gx ; Hx (x bn /n)) ¯ ¯ α+1 # p ¶ "m−1 µ µ ¶ X vj (gx ; Hx (jx bn /n)) vm (gx ; Hx (x)) 32α x bn + 2 x 1− + + x bn n j3 m2 j=1 µ ¶q µ µ ¶ ¶q bn x bn 2α |f (x+) − f (x−)| 2αcq q x 1− + + + 2q µ(bn ; f ) , nx x x n bn n (1 − ) bn bn p where m := [ n/bn ], Hx (u) = [x−u, x+u] for 0 ≤ u ≤ x, µ(b; f ) := sup |f (t)| , f (t) − f (x+) if t > x, gx (t) := 0 if t = x, f (t) − f (x−) if 0 ≤ t < x,
0≤t≤b
(6)
q is an arbitrary positive integer and cq is a positive constant depending only on q.
696
H. KARSLI AND P. PYCH-TABERSKA
From Theorem 1 and inequality (5) we get p Theorem 2. Let f ∈ BVloc [0, ∞), p ≥ 1, and let x ∈ (0, ∞). Then, for all integers n such that bn > 2x and 4bn ≤ n we have ¯ ¯ p ¯ ¯ f (x+) + αf (x−) ¯(Dn,α f ) (x) − ¯ ≤ 2Vp (gx ; Hx (x bn /n)) ¯ ¯ α+1 2 ¶ (m+1) µ µ ¶ X −1 Vp (gx ; Hx ( √xk )) x bn √ x 1− + bn n ( k)1−1/p k=1 µ ¶q µ µ ¶ ¶q 2αcq bn x bn 2α |f (x+) − f (x−)| q + 2q µ(bn ; f ) x 1− + + . nx x n bn n (1 − x )
27+1/p α + 2 1+1/p xm
bn
bn
Note that Theorems 1 and 2 are approximation theorems. In order to show this it is necessary to prove that the right-hand sides of the inequalities mentioned in the hypotheses hp iof the theorems tend to zero as n → ∞. In view of (3) we have m = n/bn → ∞ as n → ∞. So, in Theorem 1 it is enough to consider only the term Λm (x) =
m−1 X j=1
vj (gx ; Hx (jxdn )) , j3
where dn =
p
bn /n.
Clearly, Λm (x) ≤
m−1 X j=1
v1 (gx ; Hx (jxdn )) ≤ 4dn j2
mdn Z
v1 (gx ; Hx (xt)) dt t2
dn m+1 Z
≤ 4dn 1
m ³ x ´´ ³ ³ x ´´ 4 X ³ ds ≤ v1 gx ; Hx . v1 g x ; H x s m k=1 k
Since the function gx is continuous at x and v1 (gx ; Hx ( xk )) denotes the oscillation of gx on the interval Hx ( xk ), we have ³ ³ x ´´ =0 lim v1 gx ; Hx k→∞ k and, consequently, lim Λm (x) = 0. m→∞
As regards Theorem 2, it is easy to verify that in view of the continuity of gx at x, 2 −1 µ µ ¶¶ m X x 1 1 √ Vp gx ; Hx √ lim = 0. 1−1/p m→∞ m1+1/p ( k) k k=1 Thus we get the following approximation theorem.
ON THE RATES OF CONVERGENCE
697
p Corollary. Suppose that f ∈ Xloc [0, ∞) (in particular, f ∈ BVloc [0, ∞), p ≥ 1) and that there exists a positive integer q such that µ ¶q bn lim µ(bn ; f ) = 0. n→∞ n
Then at every point x ∈ (0, ∞) at which the limits f (x+), f (x−) exist we have lim (Dn,α f ) (x) =
n→∞
f (x+) + αf (x−) . α+1
Obviously, the above relations hold true for every measurable function f bounded on [0, ∞), in particular for every function f of bounded p-th power variation (p ≥ 1) on the whole interval [0, ∞). Remark. Now, let us consider the special case α = 1, p = 1, and let us suppose that a function f is of bounded variation in the Jordan sense on the whole interval [0, ∞) (f ∈ BV [0, ∞)). Then, for all integers n such that bn > 2x and 4bn ≤ n, we have the following estimation for the rate of convergence of the Chlodovsky–Durrmeyer operators (2): ¯ ¯ p ¯ ¯ f (x+) + f (x−) ¯ ≤ 2V (gx ; Hx (x bn /n)) ¯(Dn f ) (x) − ¯ ¯ 2 µ µ ¶ ¶ 2[n/bn ] µ ¶¶ µ 210 bn x bn X x + x 1− + V gx ; Hx √ nx2 bn n k k=1 µ µ ¶ ¶ 4M bn x 2 |f (x+) − f (x−)| bn + , x 1− + q + 2 nx bn n n bxn (1 − bxn ) where M = sup |f (x)| and V (gx ; H) denotes the Jordan variation of gx on 0≤x 1, µ Wn,2q (x) =
bn n
¶q X q
µ µ ¶¶q−j µ ¶j x bn βj,q x 1 − , bn n j=0
(8)
where βj,q are real numbers independent of x and bounded uniformly in n. Moreover, for n ≥ 2 µ µ ¶ ¶ bn x bn x 1− + (9) Wn,2 (x) ≤ 2 n bn n and, for q > 1, µ Wn,2q (x) ≤ cq
bn n
¶q µ µ ¶ ¶q x bn x 1− + , bn n
(10)
where cq is a positive constant depending only on q. Proof. Formulas for Wn,0 , Wn,1 , Wn,2 and inequality (9) follow by simple calculation. Suppose q > 1 and put y := x/bn . Then y ∈ [0, 1] and n
n+1X Pn,k (y) Wn,2q (x) = bn k=0 = (n +
1) b2q n
n X
Z1 Pn,k (y)
k=0
µ
Zbn 2q
(t − ybn ) Pn,k 0
t bn
¶ dt
¡ ¢ 2q (s − y)2q Pn,k (s) ds = b2q M ψ (y) , n n y
(11)
0
where Mn is the classical Bernstein–Durrmeyer operator (1). The representation formula (8) follows at once from the known identity µ ¶q−j q X ¡ ¢ y(1 − y) 2q Mn ψy (y) = n−2j , βj,q,n n j=0 where βj,q,n are real numbers independent of y and bounded uniformly £ 1in ¤ n (see £[13], Lemma 4.8 with c = −1). Now, let us observe that for y ∈ 0, or n ¤ 1 n−1 y ∈ 1 − n , 1 , n ≥ 2, one has y(1 − y) ≤ n2 and ¡
Mn ψy2q
¢
(y) ≤
q X j=0
µ |βj,q,n |
n−1 n3
¶q−j −2j
n
≤
q X j=0
ηj,q n−2q ,
ON THE RATES OF CONVERGENCE
699
£1
¤ where ηj,q are non-negative numbers depending only on j and q. If y ∈ n , 1 − n1 1 n then ny(1−y) ≤ n−1 ≤ 2 and µ µ ¶q q ¶q q ¡ ¢ y(1 − y) X y(1 − y) X 1 2q Mn ψy (y) ≤ |βj,q,n | ηj,q 2j . j ≤ n n (ny(1 − y)) j=0 j=0 Consequently, ¡ ¢ cq Mn ψy2q (y) ≤ q n
µ
1 y(1 − y) + n
¶q with cq =
q X
ηj,q 2j .
j=0
Taking in (11) and in the above inequality y = x/bn we easily obtain estimation (10). ¤ Lemma 2. Let x ∈ (0, ∞) and let µ ¶ µ ¶ µ ¶ n x t n + 1 X (α) x t Kn,α , := Qn,k Pn,k . bn bn bn k=0 bn bn Then
µ
Zbn Kn,α t
and
µ
Zt Kn,α 0
x u , bn bn
x u , bn bn
¶ du ≤
¶
α Wn.2 (x) if (t − x)2
α Wn.2 (x) if (x − t)2
du ≤
x 0, we write Ix (u) := [x + u, x] ∩ [0, ∞) if u < 0 Ix (u) := [x, x + u]
if u > 0.
Lemma 3. Let f ∈ Xloc [0, ∞) and let the one-sided limits f (x+), f (x−) exist at a fixed p point x ∈ (0, ∞). Consider the function gx defined by (6) and write dn := bn /n. If h = −x or h = x, then for all integers n such that dn ≤ 1/2 we have
700
H. KARSLI AND P. PYCH-TABERSKA
¯ Z µ ¶ ¯¯ ¯ x t ¯ ¯ gx (t)Kn,α , dt¯ ≤ v1 (gx ; Ix (hdn )) ¯ ¯ ¯ bn bn Ix (h)
"m−1 # 8αWn,2 (x) X vj (gx ; Ix (jhdn )) vm (gx ; Ix (h)) , + + 3 2 h2 d2n j m j=1
where m = [1/dn ] and Wn,2 (x) is estimated in (9). Proof. Restricting the proof to h = −x we define the points tj = x + jhdn for j = 0, 1, 2, 3, . . . , m and we denote tm+1 = 0. Put Tj = [tj , x] for j = 1, 2, 3, . . . , m + 1 and write µ
Z gx (t)Kn,α Ix (h)
+
m X
gx (tj )
Kn,α
tj+1
µ
Zx dt =
gx (t)Kn,α t1
µ
Ztj
j=1
¶
x t , bn bn
x t , bn bn
¶ dt +
t m Zj X j=1 t
x t , bn bn
¶ dt µ
(gx (t) − gx (tj )) Kn,α
x t , bn bn
¶ dt
j+1
=: I1 (n, x) + I2 (n, x) + I3 (n, x), say. Clearly, µ
Zx |I1 (n, x)| ≤
|gx (t) − gx (x)| Kn,α t1
µ
Zbn ≤ v1 (gx ; T1 )
Kn,α 0
x t , bn bn
x t , bn bn
¶ dt
¶ dt = v1 (gx ; T1 ).
By the Abel lemma on summation by parts and by (13) we have |I2 (n, x)| tj+1 µ ¶ Z x t Kn,α |gx (tj+1 ) − gx (tj )| ≤ |gx (t1 )| Kn,α dt + , dt bn bn j=1 0 0 Ã ! m−1 X 1 αWn,2 (x) |gx (tj+1 ) − gx (tj )| ≤ |gx (t1 ) − gx (x)| + h2 d2n (j + 1)2 j=1 Ã µ j m−2 XX 1 αWn,2 (x) |gx (t1 ) − gx (x)| + |gx (tk+1 ) − gx (tk )| = 2 2 h dn (j + 1)2 j=1 k=1 ! ¶ m 1 1 X − + 2 |gx (tk+1 ) − gx (tk )| (j + 2)2 m k=1
Zt1
µ
x t , bn bn
¶
m−1 X
ON THE RATES OF CONVERGENCE
701
m−2 X
vj+1 (gx ; Tj+1 ) vm (gx ; Tm ) + 3 m2 (j + 1) j=1 Ã m−1 ! X vj (gx ; Tj ) vm (gx ; Tm+1 ) αWn,2 (x) ≤ 2 + . h2 d2n j3 m2 j=1
≤
αWn,2 (x) h2 d2n
Ã
v1 (gx ; T1 ) + 2
!
Next, in view of (13) and the Abel transformation, |I3 (n, x)| ≤
m X
v1 (gx ; [tj+1, tj ])
j=1
≤
µ
Ztj Kn,α
tj+1
x t , bn bn
¶ dt
m αWn,2 (x) X v1 (gx ; [tj+1, tj ])
h2 d2n
αWn,2 (x) = h2 d2n
j=1
j2
à m X v1 (gx ; [tk+1, tk ]) Ã
k=1
m2
+
j m−1 XX
µ v1 (gx ; [tk+1, tk ])
j=1 k=1
1 1 − 2 j (j + 1)2
¶!
! m−1 X vj (gx ; Tj+1 ) vm (gx ; Tm+1 ) +6 m2 (j + 1)3 j=1 ! Ã m X vj (gx ; Tj ) αWn,2 (x) vm (gx ; Tm+1 ) +6 . ≤ h2 d2n m2 j3 j=2 αWn,2 (x) ≤ h2 d2n
Combining the results and observing that Tj = Ix (jhdn ) we get the desired estimation for h = −x. In the case of h = x the proof runs analogously; we use inequality (12) instead of (13). ¤ 3. Proof of the Main Results Now we are ready to prove our main theorems. Proof of Theorem 1. We decompose f (t) into four parts as · ¸ f (x+) + αf (x−) f (x+) − f (x−) α−1 f (t) = + sgnx (t) + α+1 2 α+1 · ¸ f (x+) + f (x−) + gx (t) + δx (t) f (x) − , 2 where gx (t) is defined as (6) and sgnx (t) := sgn(t − x), ½ 1, x = t, δx (t) = 0, x 6= t. From (14) we have f (x+) + αf (x−) f (x+) − f (x−) (Dn,α f ) (x) = + (Dn,α gx ) (x) + α + 1¸ · · ¸ 2 α−1 f (x+) + f (x−) × (Dn,α sgnx ) (x) + + f (x) − (Dn,α δx ) (x). α+1 2
(14)
(15)
702
H. KARSLI AND P. PYCH-TABERSKA
For operators Dn,α , using (15) we can observe that the last term on the right hand side vanishes. In addition it is obvious that (Dn,α 1) (x) = 1. Hence we have ¯ ¯ ¯ ¯ f (x+) + αf (x−) ¯(Dn,α f ) (x) − ¯ ≤ |(Dn,α gx ) (x)| ¯ ¯ α+1 ¯ ¯¯ ¯ ¯ f (x+) − f (x−) ¯ ¯ α − 1 ¯¯ ¯ ¯ ¯ +¯ ¯ ¯(Dn,α sgnx ) (x) + α + 1 ¯ . (16) 2 In order to prove our theorem we need the estimates for (Dn,α gx ) (x) and α−1 . (Dn,α sgnx ) (x) + α+1 To estimate (Dn,α gx ) (x), with the help of the fixed points x and 2x, we decompose it into three parts as follows: ¯ Zbn µ µ ¶ ¯¯ ¯¯ Zx ¶ ¯¯ ¯ x x t t ¯ ¯ ¯ ¯ , dt¯ ≤ ¯ gx (t)Kn,α , dt¯ ¯ gx (t)Kn,α ¯ ¯ ¯ ¯ bn bn bn bn 0
0
¯ Z2x µ ¶ ¯¯ ¯¯ Zbn ¶ ¯¯ µ ¯ x t x t ¯ ¯ ¯ ¯ + ¯ gx (t)Kn,α , dt¯ + ¯ gx (t)Kn,α , dt¯ ¯ ¯ ¯ ¯ bn bn bn bn x
2x
=: |E1,α (n, x)| + |E2,α (n, x)| + |E3,α (n, x)| , (17) where Kn,α ( bxn , btn ) is defined in Lemma 2. The estimations for |E1,α (n, x)| and |E2,α (n, x)| are given in Lemma 3 in which we put h = −x and h = x, respectively. Using the obvious inequality vj (gx ; Ix (−u)) + vj (gx ; Ix (u)) ≤ 2vj (gx ; Hx (u)), where Hx (u) = [x − u, x + u], 0 < u ≤ x, we obtain p |E1,α (n, x)| + |E2,α (n, x)| ≤ 2v1 (gx ; Hx (x bn /n)) "m−1 # p 16αWn,2 (x)n X vj (gx ; Hx (jx bn /n)) vm (gx ; Hx (x)) + + . (18) 3 2 x2 b n j m j=1 Now, we estimate |E3,α (n, x)| . Clearly, given any q ∈ N, we have µ ¶ Zbn µ ¶ n n + 1 X (α) x t |E3,α (n, x)| ≤ 2µ(bn ; f ) Qn,k Pn,k dt bn k=0 bn bn 2x
n
n + 1 X (α) Q ≤ 2µ(bn ; f ) 2q x bn k=0 n,k n
µ
2αµ(bn ; f ) n + 1 X ≤ Pn,k x2q bn k=0 =
2αµ(bn ; f ) Wn,2q (x). x2q
x bn
µ
¶ Zbn
x bn
µ 2q
(t − x) Pn,k 2x
¶ Zbn
µ 2q
(t − x) Pn,k 0
t bn
¶
t bn
dt ¶ dt (19)
ON THE RATES OF CONVERGENCE
703
Finally, replacing x by bxn in the result of X. M. Zeng and W. Chen [22] (Sect. 3, pp. 9–11) we immediately get ¯ ¯ ¯ ¯ ¯(Dn,α sgnx ) (x) + α − 1 ¯ ≤ q 4α . (20) ¯ α + 1¯ n bxn (1 − bxn ) Putting (17), (18), (19) and (20) into (16), we get the required result. Thus the proof of Theorem 1 is complete. ¤ p Proofp of Theorem 2.pLet f ∈ BVloc [0, ∞), p ≥ 1. In view of (5) and the notation dn = bn /n, m = [ n/bn ], we have m−1 X j=1
mdn Z m−1 vj (gx ; Hx (jxdn )) X Vp (gx ; Hx (jxdn )) Vp (gx ; Hx (xt)) 1+1/p ≤ ≤ (2dn ) dt 3 2+1/p 2+1/p j j t j=1
µ ≤
dn
2 m
¶1+1/p
2 (m+1) Z
1
and
2 √ µ ¶1+1/p (m+1) √ X −1 Vp (gx ; Hx (x/ k) 2 Vp (gx ; Hx (x/ s) ds ≤ ³√ ´1−1/p √ 1−1/p m ( s) k=1 k
vm (gx ; Hx (x)) Vp (gx ; Hx (x)) ≤ . 2 m m1+1/p
Moreover,
p p v1 (gx ; Hx (x bn /n)) ≤ Vp (gx ; Hx (x bn /n)). The estimation given in Theorem 2 now immediately follows from Theorem 1. ¤ References 1. O. Agratini, An approximation process of Kantorovich type. Math. Notes (Miskolc) 2(2001), No. 1, 3–10. 2. J. Albrycht and J. Radecki, On a generalization of the theorem of Voronovskaya. Zeszyty Nauk. Uniw. Mickiewicza No. 25 (1960), 3–7. 3. G. Aniol and P. Pych-Taberska, Some properties of the Bezier-Durrmeyer type operators. Comment. Math. Prace Mat. 41(2001), 1–11. ´zier, Numerical Control Mathematics and Applications, Wiley, London, 1972. 4. P. Be 5. P. L. Butzer and H. Karsli, Voronovskaya-type theorems for derivatives of the Bernstein–Chlodovsky polynomials and the Sz´asz–Mirakyan operator. Comment. Math. 49(2009), No.1, 33–57. 6. Z. A. Chanturiya, The moduli of variation of a function and its applications in the theory of Fourier series. (Russian) Dokl. Akad. Nauk SSSR 214(1974), 63–66; English transl.: Sov. Math., Dokl. 15(1974), 67–71. 7. M. M. Derriennic, Sur l’approximation de fonctions int´egrables sur [0, 1] par des polynˆomes de Bernstein modifi´es. J. Approx. Theory 31(1981), No. 4, 325–343. 8. J. L. Durrmeyer, Une formule d’inversion de la transform´ee de Laplace-applications `a la th´eorie des moments. Th`ese de 3e cycle, Facult´e des Sciences de l’Universit´e de Paris, 1967.
704
H. KARSLI AND P. PYCH-TABERSKA
9. N. K. Govil and V. Gupta, Simultaneous approximation for the Bezier variant of Baskakov-beta operators. Math. Comput. Modelling 44(2006), No. 11-12, 1153–1159. 10. S. S. Guo, On the rate of convergence of the Durrmeyer operator for functions of bounded variation. J. Approx. Theory 51(1987), No. 2, 183–192. 11. V. Gupta, An estimate on the convergence of Baskakov-Bezier operators. J. Math. Anal. Appl. 312(2005), No. 1, 280–288. 12. V. Gupta, Rate of convergence of Durrmeyer type Baskakov-Bezier operators for locally bounded functions. Turkish J. Math. 28(2004), No. 3, 271–280. 13. M. Heilmann, Direct and converse results for operators of Baskakov-Durrmeyer type. Approx. Theory Appl. 5(1989), No. 1, 105–127. 14. E. Ibikli and H. Karsli, Rate of convergence of Chlodowsky type Durrmeyer operators. JIPAM. J. Inequal. Pure Appl. Math. 6(2005), No. 4, Article 106, 12 pp. (electronic). 15. H. Karsli, Order of convergence of Chlodowsky type Durrmeyer operators for functions with derivatives of bounded variation. Indian J. Pure Appl. Math. 38(2007), No. 5, 353– 363. 16. H. Karsli and E. Ibikli, Convergence rate of a new Bezier variant of Chlodowsky operators to bounded variation functions. J. Comput. Appl. Math. 212(2008), No. 2, 431–443. 17. H. Karsli and E. Ibikli, Rate of convergence of Chlodowsky-type Bernstein operators for functions of bounded variation. Numer. Funct. Anal. Optim. 28(2007), No. 3-4, 367– 378. 18. M. R. Lagrange, Sur les oscillations d’ordre sup´erieur d’une fonction num´erique. Ann. ´ Sci. Ecole Norm. Sup. (3) 82(1965), 101–130. 19. G. G. Lorentz, Bernstein polynomials. Mathematical Expositions, No. 8. University of Toronto Press, Toronto, 1953. 20. H. M. Srivastava and V. Gupta, Rate of convergence for the Bezier variant of the Bleimann-Butzer-Hahn operators. Appl. Math. Lett. 18(2005), No. 8, 849–857. 21. P. Pych-Taberska, Some properties of the Bezier-Kantorovich type operators. J. Approx. Theory 123(2003), No. 2, 256–269. 22. X. M. Zeng and W. Chen, O On the rate of convergence of the generalized Durrmeyer type operators for functions of bounded variation. J. Approx. Theory 102(2000), No. 1, 1–12. 23. X. M. Zeng and A. Piriou, On the rate of convergence of two Bernstein-Bezier type operators for bounded variation functions. J. Approx. Theory 95(1998), No. 3, 369–387.
(Received 4.05.2009) Authors’ addresses: H. Karsli Abant Izzet Baysal University, Faculty of Science and Arts Department of Mathematics, 14280 Bolu, Turkey E-mail: karsli
[email protected] P. Pych-Taberska Faculty of Mathematics and Computer Science Adam Mickiewicz University, Umultowska 87, 61-614 Pozna´ n, Poland E-mail:
[email protected]