General Mathematics Vol. 18, No. 1 (2010), 119–129 Some evaluations of the remainder term 1 Daniel Florin Sofonea Abstract We present some representations of the remainder f (x)−(L n f )(x), where Ln is defined in (1). Using Lupaş operators (4), we prove Theorem 3 and one finds a lower-bound for (L n e2 )(x) (see Theorem 4). 2000 Mathematics Subject Classification: 41A36 Key words and phrases: Bernstein operator, Lupaş operators, positive linear operator. 1. Let f : [0, 1] → R. The Bernstein polynomials of f is (Bn f )(x) = n X n k=0 k k x (1 − x) n−k k f , n with n = 1, 2, . . . , . We consider the operator 00 (Ln f )(x) = (Bn f )(x) − αn (x)(Bn f )(x), 1 Received 17 November, 2009 Accepted for publication (in revised form) 3 January, 2010 119 120 D. F. Sofonea x(1 − x) results from the following condition (Ln e0 ) = e0 , 2(n − 1) (Ln e1 ) = e1 , (Ln e2 ) = e2 , with ej (t) = tj , j = 0, 1, . . .. We obtain the next where α(x) = operator (see [2]) (1) (Ln f )(x) = (Bn f )(x) − x(1 − x) 00 (B f )(x), 2(n − 1) n where (Bn f )(x) is Bernstein polynomials. We use the notations K = [a, b], ∞ < a < b < +∞, Ωj (t, x) = Ωj (t) = |t − x|j , (2) j = 0, 1, . . . , x ∈ K, ω(f ; δ) = sup |f (t) − f (x)|, |t−x|<δ t, x ∈ K, δ ≥ 0. From [3] we obtain Theorem 1 If L : C(K) → C(K1 ), K1 = [a1 , b1 ] ⊆ K, is a linear positive operator, then for all f ∈ C(K) and δ > 0 we have ||f −Lf ||K1 ≤ ||f ||·||e0 −Le0 ||K1 + inf {||Le0 ||K1 +δ −m ||LΩm ||K1 }ω(f ; δ), m=1,2,... where || · || = max | · | and Ωm are defined in (2). K Theorem 2 Let f ∈ C[0, 1]. If Ln are linear operator defined as in (1), then 19 1 n 1 ||f − Ln f || ≤ ω f ; √ + ω f; 16 4 n n Some evaluations of the remainder term 121 Proof. We consider m = 4. Results Ω4 (t; x) = |t − x|4 = (t − x)4 . But for Bernstein operator we know (Bn Ω4 )(x) = 1 2 2 3(n − 2)x (1 − x) + x(1 − x) n3 (3) n−2 2n(n − 1) X n − 2 k n−2−k k k + 1 k + 2 (Bn f )(x) = x · (1 − x) , , ;f n2 k n n n k=0 00 and using theorem 1 we obtain 1 19 ||f − Bn f || ≤ ω f ; √ 16 n We have x(1 − x) x(1 − x) 00 00 ||f − Ln f || = f − Ln f − B f+ B f 2(n − 1) n 2(n − 1) n x(1 − x) 00 ≤ ||f − Bn f || + B f . 2(n − 1) n But from (3) the theorem is proved because n−2 x(1 − x) 00 x(1 − x) 2!n(n − 1) X n − 2 k B f= x · 2(n − 1) n 2(n − 1) n2 k k=0 k k+1 k+1 k+2 , ;f − , ;f n n n n n−2−k · (1 − x) 2 n n−2 x(1 − x) X n − 2 k = x (1 − x)n−2−k 2 k k=0 122 D. F. Sofonea · f k+2 n −f 1 n k+1 n − f k+1 n −f 1 n k n ! n−2 x(1 − x) X n − 2 k 1 1 n−2−k ≤ n x (1 − x) · ω f; + ω f; 2 k n n k=0 n−2 1 X n−2 k = nx(1 − x) · ω f ; x (1 − x)n−2−k n k k=0 1 n 1 = nx(1 − x)ω f ; ≤ ω f; . n 4 n On the other hand we observe k+2 f nk f k+1 f k k+1 k+2 n n , , ;f = 2 − + 1 2 n n n 2 2 2 n n n n2 k k+1 k+2 n2 2 k = f − 2f +f = ∆1 f; , 2 n n n 2 n n r X r r ∆h (f ; x) = (−r)r−k f (x + kh), k k=0 with r = 1, 2, . . . , and h ∈ R. Using the following definition ωr (f, δ) = sup0<h≤δ ||∆rh (f ; ·)||, with δ > 0, results and k k+1 k+2 n2 1 , , ; f ≤ ω2 f ; , n n n 2 n x(1 − x) 00 nx(1 − x) 1 n 1 B f≤ ω2 f ; ≤ ω2 f ; . 2(n − 1) n 2 n 8 n from above formula the following proposition is proved Corollary 1 Let f ∈ C[0, 1]. If Ln are linear operator defined as in (1), then 19 1 n 1 ||f (x) − (Ln f )(x)|| ≤ ω f ; √ + ω2 f ; 16 8 n n Some evaluations of the remainder term 123 2. For approximation of the continuous functions f : [0, ∞) → R and which satisfy an inequality like |f (x)| ≤ AeBx , (x > 0), A > 0, B > 0 on [0, ∞) with A and B independent of f (that is f is of exponential type) we use linear positive operators as (see [4]) ∞ X k (4) (Ln f )(x) = an,k (x)f , x ∈ [0, ∞) n k=0 where an,k : [0, ∞) → [0, ∞), and i) the series ∞ X k=0 ii) ∞ X an,k (x)z k are convergent for |z| < r, where r > 1 B , x ∈ [0, ∞) ∀ n ≥ n0 := 1 + ln r an,k (x) = 1 iii) k=0 ∞ X k=0 k · an,k (x) = nx Examples: (nx)k - Ln = Favard – Szasz operators k! n+k xk = - Ln = Lupaş - Baskakov operators n (1 + x)n+k I. an,k = e−nx II. an,k (Ln ei = ei , i = 0, 1) Our purpose: a representation of the remainder (Rn f )(x) := f (x) − (Ln f )(x). Known methods: 124 D. F. Sofonea (A) D.D. Stancu - with the help of divided differences. (B) Peano0 s method. We suppose that f ∈ C 2 [0, ∞). (C) The present method witch is actually a following of those in (A) which where studied on particular cases. The idea is to use A. Lupa’s operators defined as ∞ X with k (Sn f )(x) = En,k (f ; x)f n k=0 j 2 k−1 k k+1 k (Sn f ) = , , ; |t − x| f , n n n n n n t 0 , k = 0, . . . , k0 1 − {nx} , k = k0 2 k−1 k k+1 En,k (f ; x) = , , ; |t − x|+ = n n n n {nx} 0 , k = k0 − 1 , k ≥ k0 + 2 where k0 = [nx]. Properties: • They are positive linear operator (h(t) = |t − x| convex); • Ln Sn f = Ln f (*) Proof. We have (Ln Sn f )(x) = ∞ X k=0 k , an,k (Sn f ) n Some evaluations of the remainder term 125 with (5) we obtain ∞ X k (Ln Sn f )(x) = an,k f = Ln f n k=0 j j (Sn f ) =f , n n (5) (6) S n ei = e i , i ∈ {0, 1}, j = 0, 1, . . . e0 (t) = 1, e1 (t) = t We write successive X ∞ j k k=k0 =[nx], (Sn f ) = En,k f ; = n n k=0 x= nj j 1 + n · n j j = 1− n· + n· f n n n but n · j n = 0, and from that results j j (Sn f ) =f n n we have (Sn f )(x) = (1 − {nx})f [nx] n + {nx}f 1 + [nx] n But 1 |t − x|+ = (|t − x| + t − x)[x1 , x2 , x3 ; |t − x|+ ]t = 2 1 1 [x1 , x2 , x3 ; |t − x|]t + [x1 , x2 , x3 ; t − x]t {z } 2 2| 0 . 126 D. F. Sofonea Theorem 3 We have the following representation: zn (x)(1 − zn (x)) [nx] 1 + [nx] (Rn f )(x) = − ,x ;f + n2 n n ∞ 2 X k−2 k−1 k + , , ; f ϕk,n(x) , n k=2 n n n where zn = {nx} and ϕk,n(x) = Ωk,n(x) − (Ln Ωk,n )(x) and Ωk,n(t) = k − 1 . − t n Proof. Using the next representation: ∞ k − 1 2 X k−2 k−1 k (7) (Sn f )(x) = a0,n (f )x+b0,n (f )+ , , ; f · − x n k=2 n n n n and ∞ k − 1 2 X k−2 k−1 k (8) (Sn f )(t) = a0,n (f )t + b0,n (f ) + , , ; f · − t . n k=2 n n n n We calculate (7)-(8): (Sn f )(x) − (Sn f )(t) = a0,n (f )(x − t)+ (9) ∞ k − 1 k − 1 2 X k−2 k−1 k + , , ;f · − x − − t n k=2 n n n n n Applied on (9) linear operator Ln , refer to t and we take the result on x. Some evaluations of the remainder term 127 From that result: (Sn f )(x) − (Ln Sn f )(x) = a0,n (f )(x − (Ln e1 )(x)) | {z } 0 ∞ k − 1 2 X k−2 k−1 k + , , ; f · − x − (Ln Ωk,n)(x) . n k=2 n n n n f (x) − f (x) + (Sn f )(x) − (Ln f )(x) ∞ 2 X k−2 k−1 k = , , ; f · (Ωk,n(x) − (Ln Ωk,n)(x)) n k=2 n n n From (*) result ∞ 2 X k−2 k−1 k , , ; f ϕk,n(x) f (x) − (Ln f )(x) = [f (x) − (Sn f )(x)] + n k=2 n n n where ϕk,n(x) = Ωk,n (x) − (Ln Ωk,n )(x). But f (x) − (Sn f )(x) is a representation introduce by A. Lupaş in ([3], p23). And now we have zn (x)(1 − zn (x)) [nx] 1 + [nx] (Rn )(x) = − ,x ;f n2 n n ∞ 2 X k−2 k−1 k + , , ; f ϕk,n(x) n k=2 n n n where zn (x) = {nx}, and ϕk,n(x) = Ωk,n (x) − (Ln Ωk,n )(x). If Ωk,n are convergent, and if L is linear operator and positive with Lej = ej , j = 0, 1, . . . then ∀ h convex h(x) ≤ (Lh)(x) ⇒ ϕk,n(x) ≤ 0 (see [3]). 128 D. F. Sofonea Theorem 4 If Ln , verifies the hypothesis, then (Ln e2 )(x) ≥ nx[nx] + {nx}(1 + [nx]) n2 , x≥0 with e2 (t) = t2 . Proof. Let f (t) = e2 (t). From Theorem 3 we have: ∞ zn (1 − zn ) 2 X (Rn e2 )(x) = − + ϕk,n(x). n2 n k=2 But (Rn e2 )(x) = x2 − (Ln e2 )(x), then ∞ zn (1 − zn ) 2X ϕk,n(x) = x2 − (Ln e2 )(x) + , n k=2 n2 and because ϕk,n ≤ 0, we have ∞ X n 2 {nx}(1 − {nx}) ϕk,n(x) = x − (Ln e2 (x) + , 2 n2 k=2 and ∞ X n ([nx] + {nx})2 + {nx} − {nx}2 − (Ln e2 )(x) ϕk,n(x) = 2 2 n |k=2 {z } ≤0 n ([nx]2 + 2[nx]{nx}) + {nx} = − (Ln e2 )(x) ≤ 0 2 n2 From above we have (Ln e2 )(x) ≥ and e2 (t) = t2 . nx[nx] + {nx}(1 + [nx]) , n2 x≥0 Some evaluations of the remainder term 129 References [1] S.N. Bernstein, Démonstration du théorème de Weierstrass fondée sur la calcul des probabilités, Comm. Soc. Math. Charkow Sér. 2 t. 13, 1912, 1-2. [2] Franchetti C., Sull’innalzamento dell’ordine di aprossimazione dei polinomi di Bernstein, Rend. Sem. Mat. Univ. Politec. Torino, 28, 19681969, 1-7. [3] A. Lupaş, ”Contribuţii la teoria aproximării prin operatori liniari”, Ph. D. Thesis, Cluj-Napoca: Babeş-Bolyai University, 1975. [4] A. Lupaş, ”On the Remainder Term in some Approximation Formulas”, General Mathematics 3, no.1-2, 1995. [5] A. Lupaş, ”Classical polynomials and Approximation Theory.”, Colloqiumvortrag, Angewandte Analysis, Univ. Duisburg / Germany, 1996. [6] F. Sofonea, ”Remainder in approximation by means of certain linear operators”, Mathematical Analysis and Aproximation Theory, Burg Verlag (Sibiu), 2002, 255-258. [7] F. Sofonea, ”On a Class of Linear and Positive Operators”, WSEAS Transactions on Mathematics, Issue 12, Volume 5, ISSN 1109-2769, 2006, 1263-1268. [8] F. Sofonea, ”On a Sequence of Linear and Positive Operators”, Results in Mathematics, ISSN 0378-6218, Birkhuser Verlag, 2009, 435-444. [9] D.D. Stancu, ”Evaluation of the remainder term in approximation formulas by Bernstein polynomials”, Math. Comp. 83 1963, 270-278. Florin Sofonea University ”Lucian Blaga” of Sibiu Department of Mathematics Str. I. Raţiu, No. 5-7, 550012, Sibiu, Romania e-mail: florin.sofonea@ulbsibiu.ro