Applied Mathematics and Computation 244 (2014) 683–694
Contents lists available at ScienceDirect
Applied Mathematics and Computation journal homepage: www.elsevier.com/locate/amc
Pointwise approximation by Bernstein type operators in mobile interval Hee Sun Jung a, Naokant Deo b,⇑, Minakshi Dhamija b a b
Department of Mathematics Education, Sungkyunkwan University, Seoul 110-745, Republic of Korea Department of Applied Mathematics, Delhi Technological University (Formerly Delhi College of Engineering), Bawana Road, Delhi 110042, India
a r t i c l e
i n f o
a b s t r a c t In the present paper, we study pointwise approximation by Bernstein–Durrmeyer type h i 1 , with use of Peetre’s K-functional and operators in the mobile interval x 2 0; 1 nþ1
Keywords: Bernstein operators Pointwise approximation Rate of convergence
x2uk ðf ; tÞ ð0 6 k 6 1Þ, we give its properties and obtain the direct and inverse theorems for these operators. Ó 2014 Published by Elsevier Inc.
1. Introduction and axillary results ~ n , were introduced and studied by Deo et al. [6] and defined as: In the year 2008, the operators B
~ n ðf ; xÞ ¼ B
n X k ; pn;k ðxÞf n k¼0
ð1:1Þ
where
n nk n k 1 1 x 1 pn;k ðxÞ ¼ 1 þ x n nþ1 k
ð1:2Þ
h i 1 . If n is sufficient large then operators (1.1) convert in the classical Bernstein operators: and x 2 0; 1 nþ1
Bn ðf ; xÞ ¼
n X n k k x ð1 xÞnk f : n k k¼0
ð1:3Þ
Now we consider Durrmeyer type operators
V n ðf ; xÞ ¼
n ðn þ 1Þ2 X pn;k ðxÞ n k¼0
Z
n nþ1
pn;k ðtÞf ðtÞdt:
ð1:4Þ
0
In 2003, a very interesting general sequence of linear positive operators was introduced by Srivastava and Gupta [18] and investigated as well as estimated the rate of convergence. Then the faster rate of convergence was studied by Deo [2] for ⇑ Corresponding author. E-mail addresses:
[email protected] (H.S. Jung),
[email protected] (N. Deo),
[email protected] (M. Dhamija). http://dx.doi.org/10.1016/j.amc.2014.07.034 0096-3003/Ó 2014 Published by Elsevier Inc.
684
H.S. Jung et al. / Applied Mathematics and Computation 244 (2014) 683–694
these operators and similar type modification have given and studied simultaneous approximation for these operators (1.1) in [4]. Ditzian [8] used x2uk ðf ; tÞ and gave an interesting direct estimate for the Bernstein polynomials. Felten [10] studied local and global approximation theorems for positive linear operators, later on similar type results studied for Durrmeyer operators by Guo et al. [15] and gave the direct and inverse theorems for pointwise approximation by Bernstein–Durrmeyer operators via Ditzian–Totik moduli x2uk ðf ; tÞ. In the year 1998, Guo et al. [14] studied pointwise estimate for Szász–Durrmeyer operators with the help of Ditzian–Totik modulus of smoothness xruk ðf ; tÞ in the interval ½0; 1Þ. Deo [3] studied pointwise estimate for modified Baskakov type operator. Recently Abel et al. [1] used properties of the Jacobi polynomials in order to give a new proof of geometric series of Bernstein operators. Very recently, Gupta and Agarwal [12] have given last two decades, literature on positive linear operators in their book and some interesting results were given by researchers [5,11,13,17] on approximation operators. In a similar manner, in this research work, we give the direct and the inverse theorem for pointwise approximation by h i 1 Bernstein type operators by Ditzian–Totik modulus of smoothness x2uk ðf ; tÞ in the mobile interval 0; 1 nþ1 . h i h i 1 1 be the set of continuous and bounded functions on 0; 1 nþ1 and First we give some notations. Let C 0; 1 nþ1
2 sup Dhuk f ðxÞ; 0 0 such that c1 y 6 x 6 cy). Now we give some basic properties of the operators (1.1) as follow: h i 1 ~ n verify the following: and n 2 N. The operators B Lemma 1.1. Let ei ðtÞ ¼ t i ; i ¼ 0; 1; 2, then for x 2 0; 1 nþ1
~ n ðe0 ; xÞ ¼ 1; B
ð1:9Þ
~ n ðe1 ; xÞ ¼ x þ 1 x; B n
ð1:10Þ
2 ~ n ðe2 ; xÞ ¼ ðn þ 1Þ ðn 1Þ x2 þ n þ 1 x: B n3 n2
ð1:11Þ
We give next Lemma along the line of proposition 1:2 p. 326 of Derriennic [7]. h i 1 and n 2 N we have Lemma 1.2 [7]. Let es ðtÞ ¼ t s ; s ¼ 0; 1; 2; . . . with the properties s 6 n, then for x 2 0; 1 nþ1
ðV n es ÞðxÞ ¼
s X s s! ðn!Þ2 nsr xr : ðn þ s þ 1Þ! r¼0 r r! ðn rÞ!ðn þ 1Þsr1
ð1:12Þ
Proof. In the account of (1.4), we obtain
Z 0
n nþ1
n Z n nk n nþsþ1 nþ1 n n 1 1 1 n pn;k ðtÞts dt ¼ 1 þ t kþs 1 t dt ¼ 1 þ bðk þ s þ 1; n k þ 1Þ n nþ1 n nþ1 k k 0 sþ1 n n! ðk þ sÞ! ¼ : nþ1 ðn þ s þ 1Þ! k!
H.S. Jung et al. / Applied Mathematics and Computation 244 (2014) 683–694
685
Therefore from (1.4), we obtain
ns
ðV n es ÞðxÞ ¼
s1
ðn þ 1Þ
n X n! ðk þ sÞ! p ðxÞ : ðn þ s þ 1Þ! k¼0 n;k k!
ð1:13Þ
For any x; y 2 R and any s; n 2 N satisfying the inequality s 6 n we have n
X n k nk ðk þ sÞ! @s s xy x ðx þ yÞn ¼ : s k! @x k k¼0
ð1:14Þ
By Leibnitz formula, the left side of (1.14), we have s sr s r X X s @ s s! @s s n! n n s @ x ðx þ yÞ x ðx þ yÞ ¼ xr ðx þ yÞnr : ¼ @xs r @xsr @xr r r! ðn rÞ! r¼0 r¼0
ð1:15Þ
1 From (1.14) and (1.15), with y ¼ 1 nþ1 x s X s s! r¼0
r
nr n X n n! n n ðk þ sÞ! xr ¼ p ðxÞ : r! ðn rÞ! nþ1 n þ 1 k¼0 n;k k!
ð1:16Þ
Now from (1.13), with the help of (1.16), we obtain
ðV n es ÞðxÞ ¼
s X s s! ðn!Þ2 nsr xr : ðn þ s þ 1Þ! r¼0 r r! ðn rÞ!ðn þ 1Þsr1
h i 1 and n 2 N. The operators V n verify the following: Lemma 1.3. Let ei ðtÞ ¼ ti ; i ¼ 0; 1; 2, then for x 2 0; 1 nþ1
V n ðe0 ; xÞ ¼ 1; V n ðe1 ; xÞ ¼
V n ðe2 ; xÞ ¼
ð1:17Þ
n n þ x; ðn þ 1Þðn þ 2Þ ðn þ 2Þ 2n2 2
ðn þ 1Þ ðn þ 2Þðn þ 3Þ
þ
ð1:18Þ
4n2 nðn 1Þ xþ x2 : ðn þ 2Þðn þ 3Þ ðn þ 1Þðn þ 2Þðn þ 3Þ
Proof. From (1.16), we obtain these assertions for s ¼ 0; 1; 2.
ð1:19Þ
h
h i 1 , we have Lemma 1.4. For x 2 0; 1 nþ1
V n ððe1 xÞ; xÞ ¼
n 2ðn þ 1Þx ; ðn þ 1Þðn þ 2Þ
ð1:20Þ
h i 2 n2 þ ðn þ 1Þ2 ðn 3Þu2 ðxÞ 2nðn 3Þ 2ðn 3Þ V n ððe1 xÞ2 ; xÞ ¼ x x2 ¼ : þ ðn þ 2Þðn þ 3Þ ðn þ 1Þ2 ðn þ 2Þðn þ 3Þ ðn þ 1Þ2 ðn þ 2Þðn þ 3Þ ðn þ 1Þðn þ 2Þðn þ 3Þ 2n2
ð1:21Þ Proof. From Lemma 1.3, we obtain these assertions.
h
2. Direct estimates In this section we give rate of convergence. h i 1 Let f 2 C 0; 1 nþ1 , the modulus of continuity of f, defined as:
xðf ; dÞ ¼
sup
jf ðtÞ f ðxÞj:
ð2:1Þ
1 ;jtxj6d x;t2½0;1nþ1
In the year 1990, Lenze [16] introduced the Lipschitz type maximal functions of order a as follows:
^ a ðf ; dÞ ¼ x
lim 1 t–x;t2½0;1nþ1
jf ðtÞ f ðxÞj ; j t xj a
x 2 0; 1
1 ; 0 < a 6 1: nþ1
686
H.S. Jung et al. / Applied Mathematics and Computation 244 (2014) 683–694
^ a ðf ; dÞ is equivalent to f 2 LipM ðaÞ. The remarkable is that the roundedness of x h i 1 ; V n ðf ; xÞ be given by (1.4), then we have Theorem 2.1. Let f 2 C 0; 1 nþ1
pffiffiffi pffiffiffiffiffi kV n ðf ; :Þ f k 6 1 þ b x f ; kn ;
where b ¼
ð2:2Þ
pffiffiffiffiffiffi ffi n . nþ1
Proof. By Popoviciu’s technique we can write
jf ðtÞ f ðxÞj 6 xðf ; dÞ
j t xj þ1 ; d
for any d > 0:
ð2:3Þ
h i 1 , we have Using positivity and linearity properties of the operators V n ðf ; xÞ and for n 2 N and x 2 0; 1 nþ1
jV n ðf ; xÞ f ðxÞj 6 V n ðjf ðtÞ f ðxÞj; xÞ;
ð2:4Þ
From (2.3) and (2.4), we get
jV n ðf ; xÞ f ðxÞj 6 xðf ; dÞ
V n ðjt xj; xÞ þ1 ; d
ð2:5Þ
Now using Cauchy–Schwartz inequality and from Lemma 1.4, we get
jV n ðf ; xÞ f ðxÞj 6 xðf ; dÞ 6 xðf ; dÞ If we choose d ¼
1=2 ðV n ððtxÞ2 ;xÞÞ d
pffiffiffiffiffi kn x d
þ1 ; ð2:6Þ
þ1 :
h i pffiffiffiffiffi 1 , we get the required result (2.2). h kn in (2.6) and take maximum over x 2 0; 1 nþ1
Motivated from Guo et al. [15], now we give direct theorem. h i 1 ; 0 6 k 6 1, then we have Theorem 2.2. For f 2 C 0; 1 nþ1
2 jV n ðf ; xÞ f ðxÞj 6 C x2uk f ; n1=2 d1k n ðxÞ þ x f ; 1=n :
ð2:7Þ
Proof. From Lemma 1.4, we have
2
V n ððt xÞ ; xÞ ¼
h i 2 n2 þ ðn þ 1Þ2 ðn 3Þu2 ðxÞ ðn þ 1Þ2 ðn þ 2Þðn þ 3Þ
6 Cn1 d2n ðxÞ:
Let
n 2ðn þ 1Þx Mn ðf ; xÞ ¼ f ðxÞ f x þ ðn þ 1Þðn þ 2Þ and
Ln ðf ; xÞ ¼ V n ðf ; xÞ þ M n ðf ; xÞ then we obtain
jn þ 2ðn þ 1Þxj 6 x ðf ; 1=nÞ: jM n ðf ; xÞj 6 x f ; ðn þ 1Þðn þ 2Þ Ln ð1; xÞ ¼ 1; Ln ððt xÞ; xÞ ¼ 0; Ln ðt xÞ2 ; x 6 Cn1 d2n ðxÞ and kLn k 6 3: 2 , for a fixed x and k, we get From (1.7) and (1.8), we choose g ¼ g n;x;k 2 D k
kf g k 6 C x2uk f ; n1=2 dn1k ðxÞ ;
2 2k 00
u g 6 C x2 k f ; n1=2 d1k ðxÞ ; n1=2 d1k n ðxÞ n u
1=2 1k 4=ð2kÞ 00
dn ðxÞ n kg k 6 C x2uk f ; n1=2 d1k n ðxÞ :
ð2:8Þ
H.S. Jung et al. / Applied Mathematics and Computation 244 (2014) 683–694
687
From ([9], p. 141), for t < u < x, we have
jt uj
u2k ðuÞ
6
j t xj
u2k ðxÞ
jt uj
and
d2k n ðuÞ
6
jt xj d2k n ðxÞ
ð2:9Þ
:
So that
jLn ðf ; xÞ f ðxÞj 6 jLn ððf gÞ; xÞj þ jf ðxÞ gðxÞj þ jLn ðg; xÞ gðxÞj 6 4kf g k þ jLn ðg; xÞ gðxÞj Using Taylor expansion with integral reminder
f ðtÞ ¼ f ðxÞ þ f 0 ðxÞðt xÞ þ
Z
t
ðt uÞf 00 ðuÞdu
ð2:10Þ
x
(see [9], p. 134) and by (2.8) and (2.9), we have
Z t Z t ðt uÞg 00 ðuÞdu; x 6 V n ðt uÞgðuÞdu; x jLn ðg; xÞ gðxÞj ¼ Ln x
x
Z nþ2ðnþ1Þx ! xþðnþ1Þðnþ2Þ n þ 2ðn þ 1Þx ðt xÞ2 00 þ u gðuÞdu 6 C d2k xþ g ; x V n n x ðn þ 1Þðn þ 2Þ d2k n ðxÞ
2k 00 n þ 2ðn þ 1Þx 2 2 2k 00 d g þ n2 d2k ðxÞd2k g 00 6 C n1=2 d1k þ d2k n ðxÞ dn g n ðxÞ n n n ðn þ 1Þðn þ 2Þ
2 00 : 6 C n1=2 dn1k ðxÞ d2k n g
h i 1 k ; d2k For x 2 0; 1 nþ1 n ¼ 1=n and from (1.8) we get
ð2:11Þ
2 2
n1=2 dn1k ðxÞ u2k g 00 6 CK uk f ; n1=2 d1k x2uk f ; n1=2 d1k n ðxÞ n ðxÞ :
From (2.11), we have
jLn ðf ; xÞ f ðxÞj 6 4kf g k þ Cn1 dn2ð1kÞ ðxÞu2k g 00 6 C kf g k þ n1 dn2ð1kÞ ðxÞu2k g 00 6 C x2uk f ; n1=2 d1k n ðxÞ : h i 1 , we obtain Hence, for f 2 C 0; 1 nþ1
2 jV n ðf ; xÞ f ðxÞj 6 jLn ðf ; xÞ f ðxÞj þ jLn ðf ; xÞj 6 C x2uk f ; n1=2 d1k n ðxÞ þ x f ; 1=n :
ð2:12Þ
This completes the proof. h
3. Inverse theorem In this section we give inverse theorem. h i 1 2 ; 0 < a < 2k Theorem 3.1. For f 2 C 0; 1 nþ1 ; 0 6 k 6 1, then the following statements are equivalent:
a ; jV n ðf ; xÞ f ðxÞj ¼ O n1=2 d1k n ðxÞ
ð3:1Þ
x2uk ðf ; tÞ ¼ Oðta Þ and x ðf ; tÞ ¼ O tað1k=2Þ ; pffiffiffi pffiffiffi where dn ðxÞ ¼ uðxÞ þ 1= n max uðxÞ; 1= n . To prove the inverse theorem we need the following notations. Let us denote
k f k0 ¼
C a;k ¼
sup 1 x2½0;1nþ1
n o daðk1Þ ðxÞf ðxÞ ; n
f 2 C 0; 1
k f k1 ¼
1 ; k f k0 < 1 ; nþ1 o n ð 2 aÞð1kÞ 2k sup ðxÞf 0 ðxÞ ; dn
1 x2½0;1nþ1
ð3:2Þ
688
H.S. Jung et al. / Applied Mathematics and Computation 244 (2014) 683–694
C 1a;k ¼ f 2 C a;k ; kf k1 < 1 ; n o d2þaðk1Þ ðxÞ f 00 ðxÞ ; k f k2 ¼ sup n 1 x2½0;1nþ1
C 2a;k
¼ f 2 C a;k ; f 0 2 A:C:loc ; kf k2 < 1 ;
K 1a;k ðf ; tÞ ¼ inf kf g k0 þ t kg k1 ; g2C 1a;k
K 2a;k ðf ; tÞ ¼ inf kf g k0 þ t kg k2 : g2C 2a;k
Lemma 3.2. If 0 6 k 6 1; 0 < a < 2, then
kV n f k1 6 Cn1=2 k f k0 ðf 2 C a;k Þ;
ð3:3Þ
kV n f k1 6 C k f k1 f 2 C 1a;k ;
ð3:4Þ
kV n f k2 6 Cnk f k0 ðf 2 C a;k Þ;
ð3:5Þ
kV n f k2 6 C k f k2 f 2 C 2a;k :
ð3:6Þ
Proof. The proof of Lemma 3.2 will be given in Section 4. Lemma 3.3 [15]. For 0 < t < 18 ;
Z
t uk ðxÞ=2
t uk ðxÞ=2
t 2
h
h i 1 1 ; b < 2, we have 6 x 6 1 nþ1 2t ; x 2 0; 1 nþ1
b db n ðx þ uÞdu 6 CðbÞtdn ðxÞ:
h i 1 1 Lemma 3.4 [15]. For 0 < t < 14 ; t 6 x 6 1 nþ1 t; x 2 0; 1 nþ1 ; 0 6 b 6 2, we have
Z
t uk ðxÞ=2
t uk ðxÞ=2
Z
t uk ðxÞ=2
tuk ðxÞ=2
2 b db n ðx þ u þ v Þdudv 6 Ct dn ðxÞ:
Proof. First, we let a new K-functional as
K auk ðf ; t r Þ ¼ infr kðf ðxÞ gðxÞÞk0 þ tr kg ðrÞ ðxÞkr : g2C k
By this definition we may choose g 2 C rk such that
kðf ðxÞ gðxÞÞk0 þ nr=2 kg ðrÞ ðxÞkr 6 CK auk f ; nr=2 :
Suppose that for f 2 C½0; 1,
kM n ðf ; xÞ f ðxÞk0 6 Cna=2 : Then we will prove
xruk ðf ; tÞ ¼ Oðta Þ: ðrÞ a=2 K auk ðf ; t r Þ 6 kðf ðxÞ M n ðf ; xÞÞk0 þ t r MðrÞ þ t r M ðrÞ n ðf ; xÞ 6 Cn n ðf g; xÞ þ M n ðg; xÞ r r r
a=2
r=2 tr a
r a=2 r=2 þ t n kðf ðxÞ gðxÞÞk0 þ kgðxÞkr 6 C n þ r=2 K uk f ; n ; 6C n n which implies
K auk ðf ; t r Þ 6 Ct a :
689
H.S. Jung et al. / Applied Mathematics and Computation 244 (2014) 683–694
On the other hand, notice that for i ¼ 1; 2; . . . ; r; x rt uk ðxÞ=2 2 ð0; 1Þ. Let 0 6 r 6 2. For g 2 C 0k ,
! !að1kÞ=2r r r X X r að1kÞ r 2r r k r k r d d x þ j t u ðxÞ x þ j tu ðxÞ 6 CkgðxÞk0 Dtuk gðxÞ 6 kgðxÞk0 2 2 j n j n j¼0 j¼0 6 CkgðxÞk0 dnað1kÞ ðxÞ: aðk1Þ For g with g ðr1Þ 2 AC loc ; kg ðrÞ ðxÞdrþ ðxÞk < 1 and for 0 < tuk ðxÞ < 1=8r and rtuk ðxÞ=2 6 x 6 1 rt uk ðxÞ=2 n
Z tuk ðxÞ Z tuk2ðxÞ 2 r ðrÞ . . . g ðx þ u þ þ u Þdu du Dhuk gðxÞ 6 1 r 1 r t uk ðxÞ tuk ðxÞ 2 2 k k Z tu ðxÞ Z tu ðxÞ 2 2 rþað1kÞ 6 Ct r dðrþaÞð1kÞ ðxÞkgðxÞk : 6 kgkr . . . d ðx þ u þ þ u Þdu du 1 r 1 r r n n tuk ðxÞ tuk ðxÞ 2 2
From the above, for 0 < tuk ðxÞ < 1=8r and rtuk ðxÞ=2 6 x 6 1 rt uk ðxÞ=2 and choosing appropriate g we obtain
r r r aÞð1kÞ ðxÞkgkr Dhuk f ðxÞ 6 Dhuk ðf ðxÞ gðxÞÞ þ Dhuk gðxÞ 6 C kf gk0 dnað1kÞ ðxÞ þ tr dðrþ n ! tr 6 Ct a : 6 Cdnað1kÞ ðxÞ kf gk0 þ t r dnrð1kÞ ðxÞkgkr 6 Cdnað1kÞ ðxÞK auk f ; rð1kÞ dn ðxÞ
4. Proof of Lemma 3.2 n To prove Lemma 3.2, we let a ¼ nþ1 and we will use the following notations, in this section: Let
~n;k ðxÞ :¼ ð nk Þxk ð1 xÞnk : q Then we know for pn;k ðxÞ defined in (1.2) that
pn;k ðxÞ ¼ an
n k ~n;k ðyÞ; x ða xÞnk ¼ q k
x y¼ ; a
0 6 y 6 1:
Let us denote
n o k f kr;½0;1 ¼ sup dnrþaðk1Þ ðxÞf ðrÞ ðxÞ ; x2ð0;1Þ
n
o C a;k;½0;1 ¼ f 2 C r ½0; 1; k f kr;½0;1 < 1 ; r
and
pffiffiffi ~dn ðxÞ ¼ u ~ ðxÞ þ 1= n;
~ ðxÞ ¼ u
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi xð1 xÞ;
Then we know
pffiffiffi ~dn ðxÞ max u ~ ðxÞ; 1= n ; and
Z n X ~ n ðf ; xÞ ¼ ðn þ 1Þ q ~n;k ðxÞ Q k¼0
1
~n;k ðtÞf ðtÞdt: q
0
Then we have the following results: Lemma 4.1. Let r 2 N. If 0 6 k 6 1; 0 < a < r, then we have for f 2 C 0a;k;½0;1
~rþaðk1Þ ~ ðrÞ ðxÞQ n ðf ; xÞ 6 Cnr=2 kf k0;½0;1 ; dn
that is,
~ Q n ðf ; xÞ
r;½0;1
6 Cnr=2 kf k0;½0;1 ;
ð4:1Þ
690
H.S. Jung et al. / Applied Mathematics and Computation 244 (2014) 683–694
Lemma 4.2. Let r 2 N. If 0 6 k 6 1; 0 < a < r, then we have for f 2 C ra;k;½0;1
~rþaðk1Þ ~ ðrÞ ðxÞQ n ðf ; xÞ 6 Ckf kr ; dn that is,
~ Q n ðf ; xÞ 6 Ckf kr : r
Proof of Lemma 3.2. Let ~f ðxÞ ¼ f ðaxÞ. Then we see the followings;
Z a n n x Z 1 X n þ 1X ~ n ~f ; x ¼ ðn þ 1Þ q ~n;k ~n;k ðtÞ~f ðtÞdt ¼ q pn;k ðxÞ pn;k ðsÞf ðsÞdt ¼ V n ðf ; xÞ; Q a a 0 a k¼0 0 k¼0 d d ~ 0 ~ 1 V n ðf ; xÞ ¼ Q f;y ; dx dy n a and
r r r d d ~ 0 ~ 1 f ; y V ðf ; xÞ ¼ ; Q r n r a dx dy n
x y¼ : a
Moreover, since
uðxÞ ¼
x pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~ ; xða xÞ ¼ au a
we have
x x pffiffiffi pffiffiffi ~ þ 1= n ~dn ¼ ~dn ðyÞ: dn ðxÞ ¼ uðxÞ þ 1= n ¼ au a a Now, we prove Lemma 3.2. For 0 6 x 6 a and y ¼ x=a,
r r d ~ ~ 1 rþaðk1Þ ~rþaðk1Þ ðrÞ ðxÞV n ðf ; xÞ dn ðyÞ r Q n f ; y : dn a dy From Lemma 4.1, we have
r r r rþaðk1Þ aðk1Þ aðk1Þ d ~ ~ d ~ d r=2 ~ r=2 r=2 r ~ ~d : Cn Q d f ; y f ðyÞk f ðyÞ ð y Þ 6 Cn k ¼ Cn sup ðyÞ a sup d ðxÞ f ðxÞ n r r r 0;½0;1 n n n dy dy dx y2ð0;1Þ y2ð0;1Þ Therefore, we have for 0 6 x 6 a
rþaðk1Þ r=2 ðxÞV ðrÞ kf k0 ; dn n ðf ; xÞ 6 Cn that is,
kV n ðf ; xÞkr 6 Cnr=2 kf k0 : Similarly, we can prove using Lemma 4.2
rþaðk1Þ ðxÞV ðrÞ dn n ðf ; xÞ 6 Ckf kr ; that is,
ðrÞ V n ðf ; xÞ 6 Ckf kr : r
In the following we give the proof of Lemma 4.1. Proof of Lemma 4.1. From (4.1) and differentiation r-times on x, we know
~ ðrÞ ðf ; xÞ ¼ ð1Þr Q n
Z 1 nr ðn þ 1Þ!n! X ~nr;k ðxÞ ~ðrÞ q q nþr;kþr ðtÞf ðtÞdt: ðn rÞ!ðn þ rÞ! k¼0 0
ð4:2Þ
H.S. Jung et al. / Applied Mathematics and Computation 244 (2014) 683–694
691
Then we obtain
~ ðrÞ Q n ðf ; xÞ 6
Z 1 nr ðn þ 1Þ!n! X ðrÞ ~nr;k ðxÞ ~nþr;kþr ðtÞ~dnað1kÞ ðtÞdtkf k0;½0;1 : q q ðn rÞ!ðn þ rÞ! k¼0 0
Since we easily see ðrÞ ~n;k q ðtÞ ¼
r r n! X ~nr;krþi ðtÞ; q ð1Þi ðn rÞ! i¼0 i
we have the following relation: ðrÞ ~nþr;kþr q ðtÞ ¼
r r ðn þ rÞ! X ~n;kþi ðtÞ q ð1Þi n! i i¼0
ð4:3Þ
Using (4.3), we know
Z 1 r X X r nr ~ ðrÞ r ~nr;k ðxÞ ~n;kþi ðtÞ~dnað1kÞ ðtÞdtkf k0;½0;1 : q ðn þ 1Þq Q n ðf ; xÞ 6 Cn i k¼0 0 i¼0 Let
p¼
2r
and
að1 kÞ
1 2r að1 kÞ ¼ : q 2r
ð4:4Þ
Then since 1=p þ 1=q ¼ 1, using Hölder inequality, we have
!1=q !1=p Z 1 Z 1 r nr nr X X X r ~ ðrÞ r 2r ~ ~ ~ ~ ~ qn;kþi ðtÞdt qn;kþi ðtÞdn ðtÞdt ðn þ 1Þ qnr;k ðxÞ ðn þ 1Þ qnr;k ðxÞ : Q n ðf ; xÞ 6 Cn kf k0;½0;1 i 0 0 i¼0 k¼0 k¼0 Then we first know
Z nr X ~nr;k ðxÞ ðn þ 1Þ q
!1=q
1
~n;kþi ðtÞdt q
6
nr X ~nr;k ðxÞ ðn þ 1Þ q
0
k¼0
k¼0
1 nþ1
!1=q ¼ 1:
Since we see
n n k k nk 2r 2r ~n;k ðtÞ ¼ u ~ ðtÞq ~ ðtÞ ð n þ 2rk þ r Þt kþr ð1 tÞnþrk t ð1 tÞ u ¼ n þ 2r k kþr ðk þ 1Þ ðk þ rÞðn k þ 1Þ ðn k þ rÞ ~nþ2r;kþr ðtÞ; q ¼ ðn þ 1Þðn þ 2Þ ðn þ 2rÞ then we know
Z
1
~n;k ðtÞdt ¼ ~ 2r ðtÞq u
0
ðk þ 1Þ ðk þ rÞðn k þ 1Þ ðn k þ rÞ :¼ aðn; k; rÞ: ðn þ 1Þðn þ 2Þ ðn þ 2rÞðn þ 2r þ 1Þ
Then secondly we have
Z nr X ~nr;k ðxÞ ðn þ 1Þ q 0
k¼0
1
!1=p ~n;kþi ðtÞ~d2r q n ðtÞdt
nr X ~nr;k ðxÞ aðn; k þ i; rÞ þ 6 C ðn þ 1Þ q k¼0
!1=p 1 : nr ðn þ 1Þ
For 0 6 k 6 2r or n 3r 6 k 6 n r, 2r X k¼0
þ
nr X
! ~nr;k ðxÞaðn; k þ i; rÞ 6 2 ðn þ 1Þq
k¼n3r
ð4rÞr nr
and for 2r þ 1 6 k 6 n 3r 1
nr n3r1 n3r1 X X k ~nr;k ðxÞaðn; k þ i; rÞ 6 xr ð1 xÞr ~n3r;kr ðxÞðn þ 1Þaðn; k þ i; rÞ 6 CðrÞxr ð1 xÞr ; q ðn þ 1Þq n 3r k¼2rþ1 k¼2rþ1 kr
692
H.S. Jung et al. / Applied Mathematics and Computation 244 (2014) 683–694
because
nr
!
ðn 3r þ 1Þ ðn rÞ k ! ðn þ 1Þaðn; k þ i; rÞ ¼ n 3r ðk r þ 1Þ kðn k 2r þ 1Þ ðn k rÞ kr ðk þ i þ 1Þ ðk þ i þ rÞðn k i þ 1Þ ðn k i þ rÞ ðn þ 2Þ ðn þ 2rÞðn þ 2r þ 1Þ rþi rþi rþi 1þ 1 þ 6 1þ krþ1 krþ2 k 2r i 2r i 2r i 1þ 1þ 1 þ n k 2r þ 1 n k 2r þ 2 nkr r r rþi 2r i 1þ 6 CðrÞ: 6 1þ k nkr
Therefore, we have for x 2 ð0; 1Þ,
Z nr X ~nr;k ðxÞ ðn þ 1Þ q
1 0
k¼0
!1=p ~n;kþi ðtÞ~d2r q n ðtÞdt
6 C nr þ xr ð1 xÞr ~d2r n :
ð4:5Þ
Thus, for x 2 ð0; 1=nÞ [ ð1 1=n; 1Þ
~ ðrÞ að1kÞ : Q n ðf ; xÞ 6 Ckf k0;½0;1 nr ~dnað1kÞ kf k0;½0;1 nr=2 ~drþ n This implies that
~rþaðk1Þ ~ ðrÞ ðxÞQ n ðf ; xÞ 6 Ckf k0;½0;1 nr=2 : dn pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ~n;k ðxÞ, we have the following expressions: ~ 2 ðxÞ. Differentiating q xð1 xÞ ¼ u
dn ðxÞ Let x 2 ð1=n; 1 1=nÞ. Then ~ r ~ðrÞ q n;k ðxÞ ¼ ðxð1 xÞÞ
i r X k ~n;k ðxÞ; Q i ðr; n; xÞni x q n i¼0
ð4:6Þ
where Q i ðx; nÞ is a polynomial in nxð1 xÞ of degree ½ðr iÞ=2 with polynomial coefficients such that
ðxð1 xÞÞr Q i ðx; nÞni 6 C
n xð1 xÞ
rþi 2 ð4:7Þ
:
Using the expression (4.6), we know
~ ðrÞ ðf ; xÞ ¼ ðn þ 1Þ Q n
Z n X ~ðrÞ q n;k ðxÞ
1
~n;k ðtÞf ðtÞdt ¼ ðxð1 xÞÞr q
0
k¼0
i Z 1 r n X X k ~n;k ðxÞ ~n;k ðtÞf ðtÞdt: q Q i ðr; n; xÞni ðn þ 1Þ x q n 0 i¼0 k¼0
Then we have by (4.7)
i rþi Z 1 X r n X 2 n k ~ ðrÞ ~n;k ðxÞ ~n;k ðtÞ~dnað1kÞ ðtÞdtkf k0;½0;1 : q ðn þ 1Þ x q Q n ðf ; xÞ 6 xð1 xÞ n 0 i¼0 k¼0 Using Hölder inequality with p and q defined in (4.4), we obtain
i Z n X k ~n;k ðxÞ ðn þ 1Þ x q n k¼0
0
1
!1=q qi Z 1 n X k ~n;k ðxÞ ~n;k ðtÞdt q ðn þ 1Þ x q n 0 k¼0 !1=p Z 1 n X 2r ~ ~n;k ðxÞ ~n;k ðtÞdn ðtÞdt q ðn þ 1Þq :
~n;k ðtÞ~dnað1kÞ ðtÞdt 6 C q
k¼0
0
}lder inequality for 2m=iq and 2m=ð2m iqÞ, we have For the first term in the above equation, using Ho
!1=q !1=q !i=2m qi qi 2m Z 1 n n i=2m X X k k k x q ~n;k ðxÞ ~n;k ðtÞdt ~ ~ q q ðn þ 1Þ x q ¼ ðxÞ 6 x ðxÞ 6 nm ~d2m ¼ ni=2 ~din : n;k n;k n n n n 0 k¼0 k¼0 k¼0
n X
693
H.S. Jung et al. / Applied Mathematics and Computation 244 (2014) 683–694
For the second term, we know by the same procedure as the proof of (4.5)
Z n X ~n;k ðxÞ ðn þ 1Þq
6 C ~d2r=p ~dnað1kÞ : n
~n;k ðtÞ~d2r q n ðtÞdt
0
k¼0
!1=p
1
Then since we know
n xð1 xÞ
rþi 2
!rþi 2
n ~d2
;
n
we have
r X n ~ ðrÞ Q n ðf ; xÞ 6 Ckf k0;½0;1 ~d2 n i¼0
!rþi 2 að1kÞ ni=2 ~din ~dnað1kÞ 6 Ckf k0;½0;1 nr=2 ~drþ ; n
that is,
~rþaðk1Þ ~ ðrÞ ðxÞQ n ðf ; xÞ 6 Cnr=2 kf k0;½0;1 : dn
Proof of Lemma 4.2. From (4.2), we know
Z nr X ~ ðrÞ ~nr;k ðxÞ Q ðf ; xÞ 6 C ðn þ 1Þq n
0
k¼0
1
~nþr;kþr ðtÞ~dnrþað1kÞ dtkf kr;½0;1 : q
Then we see
Z nr X ~nr;k ðxÞ ðn þ 1Þq
1
0
k¼0
að1kÞ ~nþr;kþr ðtÞ~drþ q dt 6 n
Z n X ~nr;k ðxÞ ðn þ 1Þq k¼0
6
nr X ~nr;k ðxÞ ðn þ 1Þq k¼0
~nþr;kþr ðtÞ~dn2r dt q
0
Z
!1=p
1
!1=p
1 r
r
r
~nþr;kþr ðtÞ minfn ; t ð1 tÞ gdt q
;
0
where p ¼ 2r=ðr að1 kÞÞ. Here, we first know
Z nr X ~nr;k ðxÞ ðn þ 1Þq
1
~nþr;kþr ðtÞnr dt ¼ nr q
0
k¼0
nþ1 nþrþ1
and we secondly have
Z nr X ~nr;k ðxÞ ðn þ 1Þq
1
~nþr;kþr ðtÞtr ð1 tÞr dt ¼ q
0
k¼0
Z nr X ~nþr;kþr ðxÞxr ð1 xÞr ðn þ 1Þq k¼0
because
~nþr;kþr ðtÞtr ð1 tÞr ¼ q
nþr
ðn þ rÞ! k!ðn k rÞ! kþr ~nr;k ðtÞ ¼ ~nr;k ðtÞ: q q nr ðk þ rÞ!ðn kÞ! ðn rÞ! k
Thus, we have
Z nr X ~nr;k ðxÞ ðn þ 1Þq 0
k¼0
1
að1kÞ ~nþr;kþr ðtÞ~drþ q dt 6 C ~dnrþað1kÞ : n
This implies
~ ðrÞ Q n ðf ; xÞ 6 C ~dnrþað1kÞ kf kr;½0;1 ; that is,
~ Q n ðf ; xÞ
r;½0;1
6 Ckf kr;½0;1 :
0
1
~nr;k ðtÞdt ¼ q
n þ 1 r x ð1 xÞr ; nrþ1
694
H.S. Jung et al. / Applied Mathematics and Computation 244 (2014) 683–694
Acknowledgments The authors are extremely thankful to the referees for valuable suggestions that greatly improved this paper. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18]
U. Abel, M. Ivan, R. Pa˘lta˘nea, Geometric series of Bernstein operators revisited, J. Math. Anal. Appl. 400 (1) (2013) 22–24. N. Deo, Faster rate of convergence on Srivastava–Gupta operators, Appl. Math. Comput. 218 (2012) 10486–10491. N. Deo, Pointwise estimate for modified Baskakov type operator, Lobachevskii J. Math. 31 (2010) 36–42. N. Deo, N. Bhardwaj, S.P. Singh, Simultaneous approximation on generalized Bernstein–Durrmeyer operators, Afrika Mat. 24 (1) (2013) 77–82. N. Deo, H.S. Jung, R. Sakai, Degree of approximation by hybrid operators, J. Appl. Math. (2013) 8. Article ID 732069. N. Deo, M.A. Noor, M.A. Siddiqui, On approximation by a class of new Bernstein type operators, Appl. Math. Comput. 201 (2008) 604–612. M.M. Derriennic, Surl’approximation des fonctions integrables sur [0, 1] par des polynomes de Bernstein modifies, J. Approx. Theory 31 (1981) 325– 343. Z. Ditzian, Direct estimate for Bernstein polynomials, J. Approx. Theory 97 (1) (1994) 165–166. Z. Ditzian, V. Totik, Moduli of smoothness, Springer Series in Computational Mathematics, 9, Springer-Verlag, New York, 1987. M. Felten, Local and global approximation theorems for positive linear operators, J. Approx. Theory 94 (1) (1998) 396–419. S.G. Gale, Voronovskaja-type results in compact disks for quaternion q- Bernstein operators, q P 1, Complex Anal. Oper. Theory 6 (2) (2012) 515–527. V. Gupta, R.P. Agarwal, Convergence Estimates in Approximation Theory, XIII, Springer-Verlag, New York, 2014. V. Gupta, O. Duman, Bernstein–Durrmeyer type operators preserving linear functions, Filomat. Univ. Niš, 64 (4) (2010) 259264. S. Guo, C. Li, Y. Sun, G. Yang, S. Yue, Pointwise estimate for Szasz-type operators, J. Approx. Theory 94 (1998) 160–171. S. Guo, L. Liu, X. Liu, The pointwise estimate for modified Bernstein Opeators, Stud. Sci. Math. Hung. 37 (1) (2001) 69–81. B. Lenze, BernsteinBaskakovKantorovich operators and Lipschitz-type maximal functions, Approx. Th. Kecskemt, Hungary, Colloq. Math. Soc. Jnos Bolyai, 58 (1990) 469496. N.I. Mahmudov, Approximation by genuine q-Bernstein–Durrmeyer polynomials in compact disks in the case q > 1, Abstr. Appl. Anal. (2014) 11. Article ID 959586. H.M. Srivastava, V. Gupta, A certain family of summation integral type operators, Math. Comput. Model. 37 (12–13) (2003) 1307–1315.