Return to Article Details Baskakov-Kantorovich operators reproducing affine functions: inverse results

Baskakov-Kantorovich operators reproducing affine functions: inverse results

Jorge Bustamante

received: June 18, 2022; accepted: June 22, 2022; published online: August 25, 2022.

In a previous paper the author presented a Kantorovich modification of Baskakov operators which reproduce affine functions and he provided an upper estimate for the rate of convergence in polynomial weighted spaces. In this paper, for the same family of operators, a strong inverse inequality is given for the case of approximation in norm.

MSC. 41A35, 41A36, 41A17, 41A37.

Keywords. Baskakov-Kantorovich operators, polynomial weighted spaces, strong inverse results.

Benemérita Universidad Autónoma de Puebla, Faculty of Physic and Mathematics, Avenida San Claudio y 18 Sur, Colonia San Manuel, Ciudad Universitaria, C.P. 72570, Puebla, Mexico, e-mail: jbusta@fcfm.buap.mx.

1 Introduction

Let C[0,) be the family of all real continuous functions on the semiaxis. Denote ek(t)=tk, k0.

Throughout the paper, we fix mN, m2, and set

ϱ(x)=1(1+x)mandφ(x)=x(1+x).

Moreover

Cϱ[0,)={fC[0,):fϱ<},

where fϱ=supx0|ϱ(x)f(x)|.

For a real λ>1, f:[0,)R and x0, the Baskakov operator is defined by

Vλ(f,x)=k=0f(kλ)vλ,k(x),vλ,k(x)=(λ+k1k)xk(1+x)λ+k,

whenever the series converges absolutely.

For some functions fC[0,), a family of Kantorovich-Baskakov type operators reproducing affine functions was introduced in [ 1 ] by setting

Mλ(f,x)=λk=0Qλ,k(f)vλ,k(x),Qλ,k(f)=Iλ,kf(akt)dt,

where

Iλ,k=[kλ,k+1λ]andak=2k2k+1,kN0.

Approximation properties of the operators Mλ in some weighted spaces were presented in [ 1 ] . The following notations are needed: for 0β1, set

Cϱ,β[0,)={hC[0,):h(0)=0,hϱ,β=φ2βhϱ<}

and

K(f,t)ϱ,β=inf{fgϱ,β+tφ2gϱ,β:gD(ϱ,β)},
1

where

D(ϱ,β)={gCϱ,β[0,):g,gACloc:φ2gϱ,β<}.

The following result was proved in [ 1 ] .

Theorem 1

If β[0,1] and m2, then there exists a constant C such that, for all λ>2(1+m) and every fCϱ,β[0,),

Mλ(f)fϱ,βCK(f,1λ)ϱ,β.

In this paper we present a strong inverse result related with theorem 1.

The work is organized as follows. In section 2 we present notations that will be used throughout the paper, as well as some identities related with the operators Mλ and their derivatives. Section 3 is devoted to prove inequalities related with the moments of the operators Mλ. In section 4 we collect several inequalities related with the weight ϱ. In section 5 we include some Bernstein type inequalities. A Voronovskaya type theorem is given in section 6. Finally the main result is proved in section 7.

In what follows C and Ci (iN) will denote absolute constants. They may be different on each occurrence. We remark that our arguments allow to obtain bounds for the constants, but not the best.

2 Notations and identities

We will use the notations

Mλ(f,x)=ddxMλ(f,x),Mλ(f,x)=d2dx2Mλ(f,x)

and

Mλ(f,x)=d3dx3Mλ(f,x).

For λ>1, kN0 and x0, we define

Rλ,k(x)=(kλx)2(1+2x)λ(kλx)φ2(x)λ,Iλ,1(x)=k=1vλ,k(x)ϱ(kλ)φ2β(kn)(kλx)2,Iλ,2(x)=1+2xλk=1|kλx|vλ,k(x)ϱ(kλ)φ2β(kn),

and

Iλ,3(x)=φ2(x)λk=1vλ,k(x)ϱ(kλ)φ2β(kλ).
5

For p,q0, set

Aλ,p,q(x)=k=1vλ,k(x)ϱp(kλ)φq(kλ).
6

For fC[0,) and k,jN, we use the notation

Jλ,k,j(f)=Iλ,kk/λak+j(t+j/λ)f(u)((ak+j(t+jλ)u)dudt.

Proposition 1

( [ 1 , Prop. 2.4 ] ) For each fCϱ[0,), λ>1 and x>0, one has

Mλ(f,x)=λ2k=0(Qλ,k+1(f)Qλ,k(f))vλ+1,k(x)=λ2φ2(x)k=0Qλ,k(f)(kλx)vλ,k(x).

If f(0)=0, the term corresponding to k=0 should be omitted.

Remark 1

The proof of proposition 1 (see [ 1 ] ) is a consequence of the identities (vλ,1=0)

vλ,k(x)=kλxx(1+x)vλ,k(x)=λ(vλ+1,k1(x)vλ+1,k(x))).
7

Here need the analogous of proposition 1 for the second and the third derivatives.

Proposition 2

For each λ>1, fCϱ[0,), and x>0, one has

Mλ(f,x)=λ2(λ+1)k=0(Qλ,k+2(f)2Qλ,k+1(f)+Qλ,k(f))vλ+2,k(x)=λ3φ4(x)k=0Qλ,k(f)Rλ,k(x)vλ,k(x),

where Rλ,k was defined in 2.

Moreover, if gCϱ2(0,) and hCϱ3(0,), then

Mλ(g,x)=λ2(λ+1)k=0(j=02(2j)(1)jJλ,k,j(g))vλ+2,k(x)
10

and

φ2(x)Mλ(g,x)λ2(λ+1)(λ+2)=k=0(j=02(2j)(1)jJλ,k,j(g))(kλ+2x)vλ+2,k(x).

If f(0)=0 or g(0)=0 the term corresponding to k=0 should be omitted.

Proof â–¼
It follows from proposition 1 and 7 that
Mλ(f,x)=λ2k=0(Qλ,k+1(f)Qλ,k(f))vλ+1,k(x)=λ2(λ+1)k=0(Qλ,k+1(f)Qλ,k(f))(vλ+2,k1(x)vλ+2,k(x)))=λ2(λ+1)k=0(Qλ,k+2(f)2Qλ,k+1(f)+Qλ,k(f))vλ+2,k(x).

On the other hand,

φ4(x)λ2Mλ(f,x)=φ2(x)λ(φ2(x)λMλ(f,x))1+2xλφ2(x)λMλ(f,x)=φ2(x)λ(λk=0Qλ,k(f)(kλx)vλ,k(x)λk=0Qλ,k(f)vλ,k(x))(1+2x)k=0Qλ,k(f)(kλx)vλ,k(x)=λk=0Qλ,k(f)((kλx)2(1+2x)λ(kλx)φ2(x)λ)vλ,k(x).

Notice that, if

Tλ(t,k)=(ak+2(t+2λ)kλ)2(ak+1(t+1λ)kλ)+(aktkλ),

then

2kλk+1λTλ(t,k)dt==ak+2((k+3λ)2(k+2λ)2)2ak+1((k+2λ)2(k+1λ)2)+ak((k+1λ)2(kλ)2)=1λ2(ak+2(2k+5)2ak+1(2k+3)+ak(2k+1))=1λ2(2(k+2)4(k+1)+2k)=0.

Hence, using the representation

g(y)=g(kλ)+g(kλ)(ykλ)+k/λyg(u)(yu)du,

one has

Qλ,k+2(g)2Qλ,k+1(g)+Qλ,k(g)==k/λ(k+1)/λ(g(ak+2(t+2/λ))2g(ak+1(t+1λ))+g(akt))dt=k/λ(k+1)/λ(k/λak+2(t+2/λ)g(u)(ak+2(t+2λ)u)2k/λak+1(t+2/λ)g(u)(ak+1(t+1λ)u)+k/λaktg(u)(aktu))du=j=02(2j)(1)jJλ,k,j(g).

On the other hand

φ2(x)Mλ(g,x)==φ2(x)λ2(λ+1)k=0(j=02(2j)(1)jJλ,k,j(g))vλ+2,k(x)=φ2(x)λ2(λ+1)k=0(j=02(2j)(1)jJn,k,j(g))k(λ+2)xx(1+x)vλ+2,k(x)=λ2(λ+1)(λ+2)k=0(j=02(2j)(1)jJn,k,j(g))(kλ+2x)vλ+2,k(x).

3 Estimates for the moments

Here, for qN, we should consider the absolute moments

Vλ,q(x)=Vλ(|e1x|q,x)andMλ,q(x)=Mλ(|e1x|q,x).

Proposition 3

For each fixed qN, there exists a constant Cq such that, if λ>1, x>0 and λx23, then

Vλ,q(x)Cqφq(x)λq2.

Proof â–¼
It was proved in [ 3 , Prop. 3.2 ] that, if q=2j, jN, the assertion holds whenever λx1. But the proof can be modified to include the case λx23. Of course, with a different constant.

If q=2j1, jN, then

Vλ(e1x2j1,x)(Vλ((e1x)2(j1),x)Vλ((e1x)2j,x))1/2C(φ2(j1)(x)λj1φ2j(x)λj)1/2=Cφ2j1(x)λj1/2.

We need an extension of proposition 3 to the case of the operators Mλ, but only for 1q6.

Proposition 4

For each fixed qN, 1q6, there exists a constant Cq such that, if λ>1, x>0 and λx23, then

Mλ,q(x)Cqφq(x)λq2.

Proof â–¼
It was proved in [ 1 , Cor. 2.3 ] that, for each λ>1 and x0, one has
Mλ,2(x)1312φ2(x)λandMλ,1(x)1312φ(x)λ.

Moreover, if x12(λ+1), then Mλ,4(x)16φ4(x)λ2. Since 23λ12(λ+1), the inequality holds under the conditions assumed above.

Notice that

Mλ,3(x)Mλ,2(x)Mλ,4(x)42φ3(x)λ3/2.

We should present a proof for Mλ,6(x). For kN,

0λIλ,k(aktx)6dt=λ7ak((akkλx+akλ)7(akkλx)7)=λ7aki=06(7i)(akλ)7i(akkλx)i=17i=06(7i)(akλ)6i(akkλx)i=17i=06(7i)(akλ)6i(kλxk(2k+1)λ)i17i=06(7i)2i(1λ)6i(|kλx|i+(k(2k+1)λ)i)17λ6i=06(7i)+17i=06(7i)2i(1λ)6i|kλx|i.

Therefore

λk=1Qλ,k((e1x)6)vλ,k(x)277λ6+i=16(7i)2iλ6iVλ(e1xi,x)277λ6+i=16(7i)2iCiλ6iφi(x)λi/2=277λ6+i=16(7i)2iCiφi(x)λ3+(6i)/233x3237λ3+i=16(7i)2i3(6i)/2Ciφi(x)x(6i)/22(6i)/2λ3φ6(x)λ3+C1i=16(7i)φi(x)φ6i(x)λ3C2φ6(x)λ3.

Moreover, taking into account proposition 3, one has

0λIλ,0(a0tx)6dtvλ,0=x6vλ,0Vλ((e1x)6,x)C6φ6(x)λ3.

This yields the inequality for m=6.

Finally the proof in the case m=5 is obtained by using Hölder inequality.

4 Preparatory computations

Proposition 5

Suppose that m2, γ[0,2), fC[0,), f(0)=0, and φ2γfϱ<. If λ>1 and kN, then

Qλ,k(f)∣≤2mλφ2γfϱϱ(kλ)φ2γ(kλ).

Proof â–¼
It was proved in [ 1 , Prop. 3.4 ] that, if γ[0,2), mγ, λ>1 and k>0, then
k/λ(k+1)/λdtϱ(akt)φ2γ(akt)2mλ1ϱ(kλ)φ2γ(kλ).

Proposition 6

Assume mN, p,q0, mpq0 and set s=2(1+2mpq). There exists a constant C(p,q) such that, if λs and x>0, then

Aλ,p,q(x)C(p,q)(1+x)mpφq(x),

where Aλ,p,q(x) is defined in 6.

Proof â–¼
It is known that (see [ 3 , Prop. 3.8 ] ), if a>1, qR, and r=max{2a,2|c|,|c|1+a}, then
k=1(λk)a(1+kλ)cvλ,k(x)C(1+x)cxa,λ2(1+r).

We apply this result with a=q2 and c=mpq2. Notice that q2mpq. Hence, if λ2(1+2mpq),

k=1vλ,k(x)ϱp(kλ)φq(kλ)=k=1(λk)q2(1+kλ)mpq2vλ,k(x)C(1+x)mpq2xq2=C(1+x)mpφq(x).

Proposition 7

Assume β[0,1] and m2. There exists a constant C such that, if k>0, j{0,1,2}, λ>2, and gD(ϱ,β), then

Jλ,k,j(g)∣≤Cϱ(kλ+2)φ2+2β(kλ+2)φ2gϱ,βλ3.

Proof â–¼
It is sufficient to consider the case φ2gϱ,β=1.

If we set g(x)=ϱ(x)φ2+2β(x), then

|Jλ,k,j(g)|Iλ,kk/λak+j(t+jλ)((ak+j(t+jλ)u)g(u)dudt.

(A) Let us first consider the case j{1,2}. Since

k(1+2(k+j))2k(k+j)+2j(k+j)=2(k+j)2,

if tkλ, then

kλ(2(k+j)1+2(k+j)(kλ+jλ)ak+j(t+jλ).

Moreover, if uak+j(t+j/λ) and t(k+1)/λ, then

1ϱ(u)=(1+u)m(1+(ak+j(k+1+jλ))m(1+(2k+2λ))m(1+(4kλ))m4m(1+kλ)m4m2mϱ(kλ+2).

On the other hand, since φ2+2β(x) increases, if kλu, then

1φ2+2β(u)1φ2+2β(kλ)1φ4β(kλ+2).

Therefore,

|Jλ,k,j(g)|23mg(kλ+2)Iλ,kk/λak+j(t+j/λ)(ak+j(t+jλ)u)dudt=23m2g(kλ+2)Iλ,k(ak+j(t+jλ)kλ)2dt23m2g(kλ+2)k/λ(k+1)/λ((t+jλ)kλ)2dt23m2g(kλ+2)k/λ(k+1)/λ(1+jλ)2dt923m2g(kλ+2)1λ3.

(B) Now assume j=0. First notice that, if kN, then

23ak<1.
11

In this case, taking into account 11, for tk/λ,

φ2(akt)=akt(1+akt)23t(1+2t3)49φ(kλ)49φ(kλ+2).

On the other hand, since

1+kλ2+kλ+2λ+2λ2(1+kλ+2),

one has

1ϱ(kλ)2mϱ(kλ+2).
12

Moreover

(k+1)ak=2k(k+1)2k+1k(2k+1)2k+1=k.

Therefore

Iλ,k|k/λaktaktug(u)du|dt==k/λk/(λak)aktk/λuaktg(u)dudt+k/(λak)(k+1)/λk/λaktaktug(u)dudtC1g(k/(λ+2))(k/nk/(λak)aktk/λ(uakt)dudt+k/(nak)(k+1)/λk/λakt(aktu)dudt)=C12g(kλ+2)(k/λk/(λak)(kλakt)2dt+k/(λak)(k+1)/λ(aktkλ)2dt)=C16akg(kλ+2)(k3λ3(1ak)3+(ak(k+1)λkλ)3)C14g(kλ+2)(k3λ31(2k+1)3+(k+1λkλ)3)C2g(kλ+2)1λ3,

where we use 12.â–¡

Proposition 8

If λ2, β[0,1), there exists a constant C such that, if x>0, k{1,2}, and λx<2/3, then

φ2β(x)vλ,k(x)λk/λ(k+1)/λ|xakt(akts)2ds(ϱφ3+2β(s)|dtC(k+1)2λ2.

Proof â–¼
Since x<2/(3λ)ak/λakk/λ  and 0<x<1,
φ2β(x)vλ,k(x)λk/λ(k+1)/λxakt(akts)2ds(ϱφ3+2β(s)dtxkφ2(x)1+xλk/λ(k+1)/λ(1+t)m(tx)2xtdss1/2dt=2xkφ2(x)1+xλk/λ(k+1)/λ(1+t)m(tx)2(tx)dtC1xk1(1+k+1λ)mλk/λ(k+1)/λ(tx)2dtC2(k+1λ)2.

Proposition 9

If m2 and β[0,1), there exists a constant C such that for i{1,2,3}, λ2(1+2m), and λx2/3,

Iλ,i(x)C(1+x)mφ2β(x)φ2(x)λ.

where we use the notations in 35.

Proof â–¼
We will use the notation in 6. Notice that 4β<42m, thus we can apply proposition 6 with λ2(1+2m)2(1+2m4β).

Taking into account Hölder’s inequality and using proposition 6 and proposition 3, we have

I1(λ,x)=k=1vλ,k(x)ϱ(kλ)φ2β(kn)(kλx)2Aλ,2,4β(x)(k=1vλ,k(x)(kλx)4)1/2C1(1+x)mφ2β(x)φ2(x)λ.

On the other hand

I2(λ,x)=1+2xλk=1|kλx|vλ,k(x)ϱ(kλ)φ2β(kn)1+2xλAλ,2,4β(x)(k=1vλ,k(x)(knx)2)1/2C2(1+x)mφ2β(x)(1+x)λx(1+x)λ=C2(1+x)mφ2β(x)x(1+x)λ(1+1x)1λC2(1+x)mφ2β(x)φ2(x)λ1+λλ2C2(1+x)mφ2β(x)φ2(x)λ.

Finally

λφ2(x)Iλ,3(x)=k=1vλ,k(x)ϱ(kλ)φ2β(kλ)=Aλ,1,2β(x)C3(1+x)mφ2β(x).

5 Bernstein type inequalities

Theorem 2

Suppose β[0,1) and m2. There exists a constant C such that, if λ2(1+2m) and fCϱ,β[0,), then

φ2Mλ(f)ϱ,βCλfϱ,β.

Proof â–¼
Taking into account the notations 35, it follows from 9, proposition 5, and proposition 9 that
ϱ(x)φ2+2β(x)|Mλ(f,x)|=ϱ(x)λ3φ2(1β)(x)|k=0Qλ,k(f)Rλ,k(x)vλ,k(x)|ϱ(x)λ2fϱ,βφ2(1β)(x)(I1(λ,x)+I2(λ,x)+I3(λ,x))Cλfϱ,β.

Theorem 3

Suppose β[0,1) and m2. There exist a constant Λ1 such that, if λ2(1+2m) and gD(ϱ,β), then

φ3Mλ(g)ϱ,βΛ1λφ2gϱ,β.

Proof â–¼
Set Z(f)=φ2gϱ,β. From ?? we obtain
(ϱφ3+2β)(x)|Mλ(g,x)|C1Z(f)(ϱφ1+2β)(x)λ4k=1(j=02(2j)Jn,k,j|kλ+2x|vλ+2,k(x)C2Z(f)(ϱφ1+2β)(x)λk=1|kλ+2x|vλ+2,k(x)ϱ(kλ+2)φ2+2β(kλ+2)C2Z(f)(ϱφ1+2β)(x)λAλ+2,2,2(1+β)(x)Vλ+2((e1x)2,x)C3Z(f)(ϱφ1+2β)(x)(1+x)mφ2+2β(x)λφ(x)λ+2C4Z(f)λ.

6 A Voronovskaya type theorem

We need a result given in Theorem 5.1 of [ 1 ] .

Theorem 4

If β[0,1] and m2, there exists a constant Λ2 such that, for all λ>2(1+m) and every fCϱ,β[0,), one has

Mλ(f)ϱ,βΛ2fϱ,β.

Let

Cϱ,β3[0,)={gCϱ,β[0,):g,g,gACloc,φ3gϱ,β<}.

Theorem 5

Suppose β[0,1) and m3 (or m=2 and β[0,1/2]). There exists a constant Λ3 such that, if gCϱ,β3[0,) and λ2(1+2m), then

Mλ(g)gg2Mλ,2ϱ,βΛ3λ3/2φ3gϱ.β.

Proof â–¼
Let us denote
Rλ(g,x)=Mλ(g,x)g(x)g(x)2Mλ((tx)2,x).

By Taylor’s expansion

g(t)=g(x)+g(x)(xt)+12g(x)(tx)2+12xtg(s)(ts)2ds,

one has

φ2β(x)ϱ(x)Rλ(g,x)∣==φ2β(x)ϱ(x)|λ2k=0vλ,k(x)(k/λ(k+1)/λxakt(akts)2g(s)ds)dt|φ2β(x)ϱ(x)φ3gϱ,βλ2k=0vλ,k(x)k/λ(k+1)/λ|xakt(akts)2ds(ϱφ3+2β(s)|dt=12φ3gϱ,βFn(x),

where

Fn(x)=φ2β(x)ϱ(x)λk=0vλ,k(x)k/λ(k+1)/λ|xakt(akts)2ds(ϱφ3+2β(s)|dt.

Case 1. Assume λx<2/3.

First we estimate the terms corresponding to k=0,1,2.

Since a0=0, one has

(ϱφ2β)(x)(1+x)λλ01/λ0xs2ds(ϱφ3+2β(s)dtϱ(x)xβ0xs(12β)/2(1+s)mdsϱ(x)xβ(1+x)m0xs(12β)/2ds=2xβx3/2β(32β)2(32β)1λ3/2.

On the other hand, it follows from proposition 8 that

(ϱφ2β)(x)λk=12vλ,k(x)k/λ(k+1)/λ|xakt(akts)2ds(ϱφ3+2β(s)|dtC1(22+32)λ2.

Now we consider the tail of the series. It is known that (see [ 3 ] )

kλvλ,k(x)=xvλ+1,k1(x).

In particular

kλk1λ+1k2λ+2vλ,k(x)=x3vλ+3,k3(x).

From this we obtain

(ϱφ2β)(x)λk=3vλ,k(x)k/λ(k+1)/λ|xakt(akts)2ds(ϱφ3+2β(s)|dt==(ϱφ2β)(x)λk=3vλ,k(x)k/λ(k+1)/λxakt(akts)2(1+s)mdsφ3+2β(s)dtλϱ(x)φ3(x)k=3vλ,k(x)k/λ(k+1)/λxt(ts)2(1+s)mdsdtλϱ(x)φ3(x)k=1(1+k+1λ)mvλ,k(x)k/λ(k+1)/λt3dtC2ϱ(x)φ3(x)k=3(1+kλ)mk3λ3vλ,k(x)C3ϱ(x)φ3(x)k=3vλ,k(x)(1+kλ+3)mkλk1λ+1k2λ+2C4ϱ(x)x3/2k=3vλ+3,k3(x)(1+k3λ+3)mC5x3/2C6λ3/2,

where we use proposition 6, with q=0 and p=1.

Case 2. Assume λx2/3 and set c=m3/2β.

From Proposition 3.3 of [ 3 ] we know that

|xakt(aktu)2du(ϱφ3+2β(u)|=|xakt(aktu)2(1+u)cduu(3+2β)/2||aktx|3(3(1+2β)/2)x(3+2β)/2((1+x)c+(1+t)c)

If λx2/3, since 3/2+β<5/2<3, one has

Fn(x)=(ϱφ2β)(x)λk=0vλ,k(x)k/λ(k+1)/λ|xakt(aktu)2du(ϱφ3+2β(u)|dtC1φ2β(x)ϱ(x)λx(3+2β)/2k=0vλ,k(x)k/λ(k+1)/λaktx3((1+x)c+(1+t)c)dt=C1φ2β(x)ϱ(x)(1+x)cx(3+2β)/2Mλ,3(x)+C1φ2β(x)ϱ(x)x(3+2β)/2λk=0vλ,k(x)k/λ(k+1)/λ|aktx|3(1+t)cdt=C1φ3(x)Mλ,3(x)+C1(1+x)βx3/2(1+x)mλk=0vλ,k(x)k/λ(k+1)/λ|aktx|3(1+t)cdt.

It follows from proposition 4 that

1φ3(x)Mλ(tx3,x)C2λ3/2.

For the other terms we first estimate the case k=0. Notice that, for t(0,1λ),

(1+t)c(1+1λ)c(1+3x2)c2c(1+x)c.

Here the condition m3 was used. Therefore

(1+x)βx3/2(1+x)mλvλ,0(x)01/λx3(1+t)cdt2c(1+x)βx3x3/2(1+x)mvλ,0(x)(1+x)c=2cx3x3/2(1+x)3/2vλ,0(x)2cφ3(x)Vλ(tx3,x)C3λ3/2.

On the other hand, since

k/λ(k+1)/λaktx3(1+t)cdt(k/λ(k+1)/λ(aktx)6dt)1/2(k/λ(k+1)/λ(1+t)2cdt)1/2,

from theorem 4 and proposition 4 we obtain

(1+x)βx3/2(1+x)mλk=1vλ,k(x)k/λ(k+1)/λaktx3(1+t)cdt(1+x)βx3/2(1+x)mMλ((1+t)2c,x)Mλ((e1x)6,x)C4(1+x)βx3/2(1+x)m(1+x)m3/2βφ3(x)λ3/2=C4λ3/2.

This completes the proof.â–¡

Remark 2

We do not know if theorem 5 holds m=2 and β(12,1).

7 Inverse result

Theorem 6

Suppose β[0,1) and m3 (or m=2 and β[0,1/2]). There exist positive constants κ and Λ4 such that, if fCϱ[0,) and λ2(1+2m), then

1λφ2(Mλ2(f))ϱ,βΛ3(Mλ(f)fϱ,β+Mκλ(f)fϱ,β),

where Mλ2(f)=Mλ(Mλ(f)).

Proof â–¼
Let Λ1 and Λ3 be constants such that the inequalities in theorem 3 and theorem 5 hold for λ2(1+2m). Set
μ=16(Λ1Λ3)2λ.

If λ2(1+2m) and g=Mλ2(f), from ?? we know that

12μφ2(Mλ2)fϱ,β12Mμ,2(Mλ2)fϱ,β||Mμ(Mλ2f)Mλ2f||ϱ,β+||Mμ(Mλ2f)Mλ2fMμ,22(Mλ2f)||ϱ,βMμ(Mλ2fMλf)+Mμ(Mλff)+Mμff+fMλf+MλfMλ2fϱ,β+Λ3μ3/2φ3(Mλ2f)ϱ,βC1(Mμffϱ,β+Mλffϱ,β)+Λ1Λ3λμ3/2φ2Mλ(f)ϱ,βC1(Mμffϱ,β+Mλffϱ,β)+Λ1Λ3λμ3/2φ2Mλ(fMλ(f))ϱ,β+Λ1Λ3λμ3/2φ2Mλ(Mλ(f))ϱ,βC1(Mμffϱ,β+Mλffϱ,β)+CΛ1Λ3λ3/2μ3/2fMλ(f)ϱ,β+14μφ2Mλ(Mλ(f))ϱ,β.

Therefore

14μφ2(Mλ2)fϱ,βC2(Mμffϱ,β+Mλffϱ,β),

and it is sufficient to prove the result, because

1λ||φ2(Mλ2(f))||ϱ,β=26(Λ1Λ2)24μ||φ2(Mλ2(f))||ϱ,β.

Theorem 7

Suppose β[0,1), m3 (or m=2 and β[0,12]), and κ is given as in theorem 6. There exists a constant C such that, if fCϱ[0,) and λ2(1+2m), then

Kβ(f,1λ)ϱC(Mλ(f)fϱ,β+Mκλ(f)fϱ,β),

where Kβ(f,t)ϱ is defined as in 1.

Proof â–¼
Fix fCϱ[0,) and denote g=Mλ2(f)=Mλ(Mλ(f)). From theorem 4 we know that gCϱ,b[0,).

From the definition of the K-functional Kβ(f,t)ϱ, theorem 4 and theorem 6 we know that

Kβ(f,1λ)ϱfMλ2(f)ϱ,β+1λφ2(Mλ2(f))ϱ,βfMλ(f)ϱ,β+Mλ(fMλ(f))ϱ,β+1λφ2(Mλ2(f))ϱ,β(1+Λ2)fMλ(f)ϱ,β+1λφ2(Mλ2(f))ϱ,βC(Mλ(f)fϱ,β+Mκ,λ(f)fϱ,β).

Remark 3

In the case β=0 the K-functional can be replaced by a weighted modulus of smoothness as in [ 4 ] . For the classical Baskakov operators Vn related results were given [ 6 ] and [ 5 ] . â–¡