REWE Ð'ANALYSE NUMÉRIQUE ET DE THÉORIE DE L'APPROXIMATION Tome XXVIÍ, No
l,
199g, pp. 99_106ON SOME PROPERTIES OF STANCU OPERATOR
ZOLTÁN FINTA
I.INTRODUCTION The positive linear polynomial operator defined by
(l) P:
(.f,")= É f ftln)wn,t(x,a), neN, xe
[0,1],a>0,
k=0
where/is a real function on [0,
l]
andk-l n-k-l
w,,t(x,*)=f,1 \k) (l+cr) ,'! t'..*' (t+2o")...(t+ ,--4 lt ,"-'^' (n_t)a)'
was introduced by stancu [6], who studied, among other properties, the conver- gence
of
Pff tofas n-)
@ and 0 <s
= c¿(z) _+ 0.In the case cx, =
0,
pro is the Bemstein operatorB,
given by(2) B,(î ,')
=É r&
tn)(:).r (r- *),-o
,where
f eCl\,ll.
Lorentz[4,p.|}2]proved
rhat forf eCl},ll:
(3)
lB,(-f,x)--f(x)l=M,'0
n^*)
ifandontyif a2çf ,h)=o(hr).
Here r¡ 2
(f
,t)
is the classical modulus of smoothness defined by(4)
úJ'(f
,r) =o:ï!,
lllr'r.f @)llrp,,t , wherenr.f (*) -{r rr
+h)-2f
(x)+f
(x -h), if
x+ he[o,r];
¡. 0,
otherwise.AMS Subject Classification: 41436.
100 Zoltán Finta
In
this paper, we pfovide further approximation propertiesfor P,". Let
rp denotethe following function:
q(.¡;)={rO-ù,xe[0,1]' For a
functionge c[0,1] we
denotethe
uniform norfnon a
subinterval[a,blc[o,\]
byllgll,,t",ol = sup { lg(¡) l: x efa, b)]'
2. MAIN RESULTS
3. LEMMAS
For proving Theorem
I
we need the following lemmas:Lsvnr,le
|
(the localization theorem).If f ec[0,1]
vanisheson a
sub-interval
la,blç10,1)
anda(n)=o(n't),
then(7)
P:(f,x)= o(n-t),
xe(a,b)'
Proof'
suppose thatf
has second derivativef ,,(x) for
some xe[0,l]
Then, by lheorem 7.1 [6,
p. ll92]
and a(n) = o(n_t), we have(B)
]Y*"tP;
(-f , x)-
.f (x)1 ={l jr"f - 't
"(x),From Lemma 5.2
[1,p.
ßa]we ger (7), Lnn¿¡¿e 2.If f
e C[0,l],
u(n) =o(n-t)
andtï* {r[p,"
(.f , x)- f (x)]]
>0,
x e(a,å)c
[0,l],
The theorems in question can be stated as follows:
Tneonpu
I'
Let u(n) = o(n-t ),a(n)'
n 3 r,for
n =l'2' "'
andf
ec[0'
1]'Then
for
each M > 0,(s)
lP:(f ,x)-,f(x)l< ,ÉP,x
e[0, rf, n=t,2'.'.
holds exactly when a2
(Í
,h) < lvIh2 , h> 0'Trmon¡v
2. Let O<o(n)-+0 (n->ø)
and.f .40,1)
with oz("f 'h)<l'¡hþ,h>0,and0<P<2,Then
lp.
(f ,x)-,r(x)l
=r(Él-)u'' ,
x e[0,1], n=1,2,'..
T¡ggREM 3.
Let
cr =cr(n) = O(n-1 ), o?., =(l+nu)l(n(l+o)), n=1,2,,,.
and
f
e C[0,ll.
Then there exists M > 0 such that(6)
llP:f - f
llrro,r¡l M'arfl(f ,a,)r¡o,,¡,
where cll{.f
,/)cro,rt=otïp=,llI'rr<,tÍ(*)ll.¡0,'¡
to the Ditzian-Totik modulusof
smoothness
l3f.
Tireonnu 4, Lei u=u(n) =o(nt), and 0<þ<2' i¡ f eCi1,il
anciai U, h)
cp.tt = o(hþ),
then ll P:f -,f
ll"ro,, r =o(n-þ''
)'In order to prove theorems, we need some lemmas
first'
i2 3
On Some Properties
10r
thenf is convex on[0,11.
Proof'
suppose thatf is
not convex._Then, using the parabora technique[1,p.
124], there exist xoe(a,ó),ô>0
ande@)="ri +d**r,".0
such that"f(xò=e(xo)
andf(xo+t)<eeo+/) for allt,
ltl<ô. wecanexrend ftoa
fturction x e
[0,l]. F eC[0,1)
By Lemma 4.1.16,p,so that F(x)=f ilg4l (x),
x e [xo _ô,xo+ð], F(x)<ee),
(e)
so we obtain
pno
(u', r)
=*
Prt
(r,
x) = xfx(l-x) I
L , +x{x+c)j
Pn" (1, x) = 1,
and
n[Pf
(F,x)
-F(xo)]t
n[p," (e, xo) _eeù]
But Lemma
I
shows that=c.xo(t-,ïo)
11'?g<0
n[p,"
U, x)
_ -f(xù]
< c. xo(l _ xo).
t,* no +o(l),
which contradicts the assumption
(9),
I + ctIn order to prove Theorem
2le
need some other lemmas:Lewre
3. For.f
eCIO,1] we have llp,ïf
llrto,,l s l]./ll.ro,,i .Proof. Since
.i wn,t(x,u)
=I
[6, p. l lg4], we havek =0
I P," (.f
,")
I=
å
w n. t ( x, a) I "f Ø t n) I <ll.f
llqo,,t .5 On Solne Properties
103
v/here G(v) = (p 2 (y)
f
,,(v).
By Lemma 5 we obrain(16) llpi Gr(f
,.,x),x)llc¡et,.t_Atnt<,r(F#,ìll.,^,.,.^,,
llM(G,x)llcro, rland (17) Since
we have (18)
ll M (G,x) llcro,,t slle'
Í,,
llrrc,,t.P," ((u
- ,)' .g-, e),
x)= n(l+o)'
1: noll
p:
( R, (f
,.,x),
xllc¡ e r u. t _,r r,1 sffi ,r, f
,, llr¡0,t:,
using (16) and (17). By Taylor's formula:
u
I
@) =f
(x) +î,(x) (u_,) +
J @_ v)f
,,(v) dv.Then
P,ï
(f
(.), x) =f
(x) +f
,(x)p:
((. _ x), x) + p," (Rr(-f ,., x), x).Hence, by (18), we obtain (15).
4. PROOFS
ProofofTheorem
I.If
lP:
U,x)-.f(x)l< M.'P'..(*),
x e[0,t], n=r,2,3,.,.
n
is satisfìed, rye pur g(u) =
f
(u) + Mtf
/2
and obtain' n[P,"(g,'x)-g(x)lr r'x(l-x)
(n-l)htn
2 l+a
Then, by Lemma 2, g is convex on [0,
l]
and, therefore, g(x + h) _2g(x) + g(x _ h)>0.Hence
f (x+h)-2f (x)+/(x-h)>-Mh2.
Analogously,forg(u)=_f (u)+
Mu212Zoltan Finta 4
102
(14) So
llP:
f
ll.¡0,,1 < ll"/ ll.ro;'r'Lnwa
4.For
some constantC>0 andall f aCl},ll'
(10) lP: (f ,x)- lQ)l<Ca'(f ,qç¡tJi),
x e[0,1]'Proof.LeT.
f
eC2¡0,1.1. By Taylor's formula:(1,) t(+)
=t G).(+- ò,' ur.(:- ò'lir
"u,. r(+-i]'
where
s0)
=s,o)-) 0
as y)
0.This shows that ths last term in(l l)
does notexceed ll,f " ll.ro,
¡(k
I n-
Ð2 12. Then(r2) llp. Í-
lll"¡0,,1<ll./"ll.ro,,r4*
#, ¡
ec2¡0,t1.Using Lemma 3, we obtain the boundedness of the operator
Pf
and, by (12) andTheorem 5.3 12,p. 2181, we obtain (10)'
Remark.
For
c¿=0 the inequality (10) reduces to the known inequalityof
Popoviciu [5] conesponding to the Bernstein polynomial (2):(13) lB,(f ,x)- f (x)l<Ca'(f ,q(r) l^ñ),
x e [0,1]' Theorem 4 requires the following lemmasLEÀ/ß44 5.
For
Rr(f ,u, x)= ! f"-v)
,f " (v) dv we have,T
lu-
xllRr(f
,u,x)l< +
Q- (x) 92(v) |
f "(v)ldv
J
for
x, zr e [0, 1].Proof. Lemma 5 is a particular case of Lemma
9'6'I
13, p. 1aOl'LEMMA 6.
For
A >0
a fixed number andf
a dffirentiable function on 10,l)
with
f '
locally absolutely continuous in f0,ll
we have(ls)
llP.l -îllrrn,,.,-n,,t ffi
ffs'"f
"11"10.,7'Proof. We consider the maximal function of Hardy-Littlewood
M(G,x)=supl lr'l t'
IC1u;Arl,
u lx-u "
I(20)
llo's','llr¡o.t¡ 32nKz,eç
,n-1)cto.tt.By Kz,r(f
't')cro.¡1 we denote the K-ñrnctionar of the pair of spaces
c[0,1]
anda coresponding weighted Sobolev space with weight frrnction g2 given by
Kr.r(-f
,t2)qo,,r =ir"tr{llf
-gllc¡0,r1+tt llq,
g,,llc1o,r1ig,e
A.c.,o"}, whereg'
e A.C.,o" means that g is differentiable
andg,
is absolutely continuous m everyÍ
= "f- g,
closed*g,,
intervar we estimate[a,b]c-[',1].
using Lemma 3 and Lemma 6 and writing(21)
and
llP:
(l - s,) - (Í-&
)llc¡0, tt< C ll.f-
B,llc¡o,ry(22)
llp,is, - Inllc¡an,t_,u,.1s#ft.ller8,,ll.¡0.,;
.
By Theorem 7.2.1 and, Theorem 7.3.1 [3,
p.79
and,p, g4, respectively], \,].e can chooseg' to be
{",
theu*st tJãi-¡å
degreeór;";;
approximario' in C[0,1]. We have(23) ilPi Prt;t-
PtJ;tlic¡',1¡ sMltpi \.t;t-
prJ;tllr¡n,,,,_n,n1,using Theorem g.4.g [3,
p.
10g] translatcd from [_1,l]
to [0, 1] andn to [\Ç].
But, in view of Theorem2.l.l [3, p. I 1], there
exists
M>
'such
(24) M-'rtr(.f ,o-,,r)"to,,l I Kr,r(f ,n,l)cro,rt<
Ma2r(-f ,n,,,r).,r0,,r.thatThen. by (23), (24)^ (.22¡ vn¿120). we obtain
(2s) llP"
Pr^t,t- PrJ;tLlc¡0,,11" ñft'2n.Kr.*(f ,r-')"ro,,r r
<
M' 'ff '
'l{'f 'n-"')
',.,,t'On the other hand, by (21),( l9) and (24), we have
(26)
llp: (f - \r;t)-U -{ø¡)ll.¡0.,1<
<2CK,.e(f
,n-')rto,,l s
M,, .rrrÇ,r_"r).t0,,1.
7 On Some Properties
t05 and
104 Zoltrán Finta
we have f (x+h)-ZlG)+ Í(x-h)<
Mh2, so l\'ol(Ðl< Mhz.
Thereforea'(.f
,h) < lrlht,
h> o.Conversely,
let f
eC[g,1] with o'
(-f , h) < lrlh'.
h>0'
Thenf
belongsto the Lipschitz space
Lip(2,C[0,1]),
which implies,in view,of
Theorem 9'3 12,p.53f, that there exists/'
on [0, 1.1 and/' is
absolutely continuous withlf "(x)l< M
a.e. .r e[0,1].Let
x e[0, 1] be a fixed point. For the linear functionl(y)
=f
(x) +f
'(x)'
'
(, - y),
by Taylor's formula, we have l-f0) - t(y)l<
M (y-
x¡2 I2'
Since P,"preserues linear frurctions, we obtain
lP: (f,x)- f(x)l=lP: (Í -t,x)l<
Y
p,: ((.- x)z;x)= M. t(l_-t)
+!Y
<-2 2n
1+ GBy the hypothesis
n'a(n) (
1, forn=!,7,...
we obtainlP: (f,x)- f(x)l< M q2(x), xe[o,i],
n
Proof
of
Theorem 2. Lemma 4 and ol2(f,h)<
Mhþ' 0<P<2,
imply theinequality
lP: (f
, x)- I Q)l<
M(ç'
(x) I n)þtz ' Proof of Theorem 3 . We havePf(l,x¡=1, Pi@,x)=x
and
[_ . I lxfl_x)+r("+o)-].
'ì
(u",x)
=I
r.üL---.,, j
Then
Pi
(@-
x)2,x)
=x(l - Ð
'+:+=
n(t +ü)
q2 1x¡ 'crl 'fot
n=1,2,...So,
our statement is a direct consequence of Theorem 1 [7' p. 165]' Proof offneorei
+. We can chooseg, eCf0,1]
such that <p2 g'n'e C[0,l]
and g', is locally absolutely continuous in [0,
l],
which satisfies(19)
ll"f-B,llcr.l,'l <2Kz,q(f ,r-t)rro.,t
6
n =
lr2,
MEAN- VALUE FOR'MULAE FOR INTEGRALS INVOLVING
GENERALIZED ORTHOGONAL POLYNOMIALS
REVUE D'ANALYSE NUMERIQUE ET DE THÉORIE DE L'APPROXIMATION Tome XXVII, No 1,199g, pp. f07_ll5
LATIRA GORI, D. D. STANCU
l. It
is known thatif
we have a seassociated with a nonnegative measure
d
ul polynomiaß(p,), if
we consider a f,inction -f el,,,(o,,
tlrn'jlti:JffT
value formula, of N, Cioranescu
p],
for(1) [
btra p,G)da(_r)
=f
(')(E)
I ", p,(x) do(x),
a <E <b..
In this paper we-give several.*t.nriolr.
of.it to some crassesof nonclassical orthogonal polynomiars, incruding
th;;;;r-"rrhogonal
porynoiiur, correspond_ing to a measure of the form oldä,
*rrir.
given real
rrtt¡t. ,"ro*,
cÙ ls a nonnegative polynomial having, t'
is remarkable the speciar case of s-orthogonal polynomialsp^,",
foïwhichI ,Ì::'
(x) da(.r) =ñh,
when we obtain the exrensionØ [
brø p];i' u)dcr(x)
=f (2G)
I
*^ o:,,,-,1.r¡ dcrlx¡,n"r"il||,1lces to
formula(1)
when s =0
(the caseof
ordinary ofihogonar2'
we
start from-a
"methodof
parameters,,. (see [20])for
constructing affiii,r"J,:ï:T:i:rrer
quadratu'"roilîu
uy u,i,,g,nu;tî oî.u,,,*ed
nodesn., o,Ll;.jffj};l;llå:
innnitery many poinrs or increase andrhat
da(x)AMS Subjecr Classification; ZStZ+, SZCZS, 4]ASS,65D32.
Consequently, from (25) and (26),we obtain
llP,ï
f -./ llcro,,,t(*, # * r,,). r'*{r, r-''' ),p.,t.
tJsingthehypotheses
a=o(n-t) and of (f
,'n-''')cto,rì =O(hþ)' 0<B< 2'
we obtain llP:f - f
llrw,tt= O(n-þt2 ). Thus the theolem i5 proved completely'REFERENCES
l. R. A. Devore, The Approximation of Continuous Functions by Positive Linear operators, Lecture Notes in Mathematics, 293, Springer Yerlag,7972'
2. R. A. DeVore and G. G. Lorentz, Constructive Approxímation, Springer Verlag, Berlin-Heidel- berg-New York-London, 1 993'
3, Z. Ditziw.tand V. Totik, Moduli of Smoothenes.s, Springer Verlag, Berlin-Heidelberg-New York- London, 1987.
4, G. G. Loreutz, Approximation of Functíons, Athena Series' Holt' Rinehart and Winston' New York' 1966'
'^-)-'nti¡ø ztoo rnn¡tinns rn lathematicacluj"l0 5.T.Popovicìu,Srtl'approximaliondesfonclionsconvexesd'ordresuptirieur'lt
(1935), 49- s4.
6.D. D. Starrcu, Approxintation of functions by a netv class of linear polynomíal operators,Rev' Routnaine Math. Pures Appl. XIII, 8 (1968), 1173-1194'
7. V. Totik, LIníþrn approximation by posítive oPerotofs on infnile intervals' Aural' Math' 10 (1984), 163-182.
Received May 10, 1997 Depaitmenl of Mathematics
" Bab eS-Bolyai " Universi IY
3400 Cluj-Napoca Rl\mania
Zoltán Finta 8
106