Startseite Approximation of entire transcendental functions of several complex variables in some Banach spaces for slow growth
Artikel Öffentlich zugänglich

Approximation of entire transcendental functions of several complex variables in some Banach spaces for slow growth

  • Devendra Kumar EMAIL logo
Veröffentlicht/Copyright: 19. Mai 2017

Abstract

In the present paper, the coefficients characterizations of generalized type Tm(f;α,α) of entire transcendental functions f of several complex variables m (m2) for slow growth have been obtained in terms of the sequence of best polynomial approximations of f in the Hardy Banach spaces Hq(Um) and in the Banach spaces Bm(p,q,λ). The presented work is the extension and refinement of the corresponding assertions made by Vakarchuk and Zhir [22, 20, 21, 23, 24, 25], Gol’dberg [4] and Sheremeta [17, 18] to the multidimensional case.

MSC 2010: 30E10; 41A15

1 Introduction

By we denote a space of complex numbers z=x+iy, and

m={𝐳=(z1,,zm):zj,j=1,,m}

is the m-dimensional complex space. Let f be an entire function of m complex variables 𝐳=(z1,,zm)m and let {DR}m, R>0, be a family of complete m-circular domains depending on the parameter R and such that 𝐳DR if and only if

𝐳R=(z1R,,zmR)D,D=D1.

Suppose that Mf,D(R)=max{|f(𝐳)|:𝐳DR}. Gol’dberg [4] introduced the order and type of growth of f(𝐳) as (see also [2])

(1.1)ρD=lim supRloglogMf,D(R)logR,

and, for 0<ρD<,

(1.2)T=Dlim supRlogMf,D(R)RρD.

The quantities ρD and TD are called the D-order and D-type of the entire function f, respectively.

We now write the expansion of the entire function f in the Taylor series

f(𝐳)=|𝐤|=0c𝐤(f)𝐳𝐤,

where 𝐤=(k1,km)+m, |𝐤|=k1++km, 𝐳𝐤=z1k1zmkm and c𝐤(f)=ck1,,km(f) are the Taylor coefficients of f. Gol’dberg [4] established the following relationship between ρ and T with the moduli of the coefficients |c𝐤(f)|:

(1.3)ρ=lim sup|𝐤||𝐤|log|𝐤|-log|c𝐤(f)|

and

(1.4)T=lim sup|𝐤||𝐤|eρ|c𝐤(f)|-ρ|𝐤|.

Sheremeta [17, 18] generalized Gol’dberg’s results with the help of general functions as follows.

Let L denote the class of functions h satisfying the following conditions:

  1. h(x) is defined on [a,) and is positive, strictly increasing and differentiable, and tends to as x.

  2. For any function φ(x) such that φ(x) as x, we have

    limxh{(1+1φ(x))x}h(x)=1.

Let Δ denote the class of functions h satisfying condition (i) and

limxh(cx)h(x)=1for every c>0,

that is, h(x) is slowly increasing.

Sheremeta [17] introduced the following notions of generalized order and generalized type of growth for entire functions of several complex variables:

(1.5)ρm(f;α,β)=lim supRα(logMf,D(R))β(logR)

for α(x)Δ and β(x)L. Further, for α(x)L, β-1(x)L and γ(x)L,

(1.6)Tm(f;α,β)=lim supRα(logMf,D(R))β((γ(R))ρm).

Setting α(x)=logx and β(x)=x in (1.5) and α(x)=β(x)=γ(x)=x in (1.6), we obtain the Gol’dberg definitions of order and type of entire functions (1.1) and (1.2), respectively.

Sheremeta also established the relationship between the generalized order of growth (1.5) and (1.6) of an entire function f and its Taylor coefficients. We now summarize these results in the form of the following theorems.

Theorem A.

Let f(z)=|k|=0ck(f)zk be an entire function of m complex variables with generalized order of growth ρm(f;α,β), where αΔ and βL. If for any c(0,), the function F(x,c)=β-1(cα(x)), where β-1 is a function inverse to β, satisfies the condition

dF(x,c)d(logx)=O(1)as x,

then

(1.7)ρm(f;α,β)=lim sup|𝐤|α(|𝐤|)β(-|𝐤|-1log|c𝐤(f)|).

Theorem B.

Let f(z)=|k|=0ck(f)zk be an entire function of m complex variables with generalized order of growth 1<ρm(f;α,β)<. Then f(z) is of generalized type Tm(f;α,β), where α(x)L, β(x)L and γ(x)L. Set F(x,Tm,ρm)=γ-1{[β-1(Tmα(x))]1ρm}. Suppose that for all Tm,0<Tm<, F satisfies:

  1. If γ(x)Δ and α(x)Δ, then dF(x,Tm,ρm)dlogx=O(1) as x.

  2. If γ(x)L-Δ or α(x)L-Δ, then limxd(logF(x,Tm,ρm))d(logx)=1ρm.

Then we have

(1.8)Tm(f;α,β)=lim sup|𝐤|α(|𝐤|ρm)β{[γ(e1ρm|c𝐤(f)|-1|𝐤|)]}ρm.

Remark 1.1.

For the functions α,β and γ defined above, the relations (1.7) and (1.8) yield the Gol’dberg results (1.3) and (1.4), respectively.

It has been noticed that the above characterizations of generalized order and generalized type of f(𝐳) in terms of Taylor coefficients do not hold when α=β=γ, i.e., entire functions of slow growth. To overcome this difficulty, Kapoor and Nautiyal [9] introduced a new class of functions Ω for m=1, defined as follows. A function ϕ(x) belongs to Ω if ϕ(x) satisfies (i) and the following condition:

  1. There exists a function δ(x)Δ and xo,K1 and K2 such that for all x>xo ,

    0<K1d(ϕ(x))d(δ(logx))K2<.

Further, a function ϕ(x) belongs to Ω¯ if ϕ(x) satisfies (i) and

limxd(ϕ(x))d(logx)=K,0<K<.

Kapoor and Nautiyal [9, p. 66] showed that Ω,Ω¯Δ and ΩΩ¯=ϕ. Let α(x)Ω or Ω¯. Then, following [9, p. 66], for an entire function f(𝐳), we define the generalized order and generalized type as

ρρm(f;α,α)=lim supRα(logMf,D(R))α(logR)

and

TTm(f;α,α)=lim supRα(logMf,D(R))(α(logR))ρ,

respectively, where α(x) either belongs to Ω or Ω¯.

Recently Vakarchuk and Zhir [24] studied the best polynomial approximations of entire transcendental functions of several complex variables in Banach spaces. Let Um={𝐳m:|zj|<1,j=1,,m} be a unit polydisk in m and let Γm={𝐳m:|zj|=1,j=1,,m} be its skeleton. Also, let m be an m-dimensional real space. Further, by

Tm={𝐱=(x1,,xm)m:0xj2π,j=1,,m}

and

Πm={𝐫=(r1,,rm)m:0rj<1,j=1,,m},

we denote m-dimensional cubes in the space m. Let A(Um) be the set of all analytic functions in the set Um. For any function fA(Um), we get

Mq(f,r)={1(2π)mTm|f(𝐫ei𝐭)|qd𝐭}1q,0<q<,

where f(𝐫ei𝐭)=f(r1eit1,,rmeitm), d𝐭=dt1dtm and

M(f,r)=max{|f(𝐫ei𝐭)|:𝐭Tm},𝐫Πm.

Let Hq(Um), 0<q, denote the Hardy space of functions f(𝐳)A(Um) satisfying the condition

fHq=sup{Mq(f,r):𝐫Πm}<,

and let Hq(Um) denote the Bergman space of functions f(𝐳)A(Um) satisfying the condition

fHq={1(2π)mTm|f(ei𝐭)|qd𝐭}1q,q>0.

For q=, let fH=fH=sup{|f(𝐳)|:𝐳Um}. Then Hq and Hq are Banach spaces for q1. Following [24, p. 1794], we say that a function fA(Um) belongs to the space Bm(p,q,λ) if for 0<p<q,

(1.9)fp,q,λ={Πm(1-𝐫)λ(1p-1q)-1Mqλ(f,𝐫)𝑑𝐫}1λ<,0<λ<,

and

fp,q,=sup{(1-𝐫)(1p-1q)Mq(f,𝐫):𝐫Πm}<,λ=.

It is known (see [6, 7]) that Bm(p,q,λ) is a Banach space for p>0 and q,λ1, otherwise it is a Frechet space.

Let Pn denote a subspace of algebraic polynomials of m complex variables of the form

Pn={|𝐤|=0c𝐤𝐳𝐤:c𝐤},

where n+. Let X be one of the Banach spaces of analytic functions of m complex variables listed earlier. Let En(f,X) denote the value of the best polynomial approximation of the function fX by elements of the subspace Pn, i.e.,

(1.10)En(f,X)=inf{f-pnX:pnPn}.

Several authors established the relationship between the order and type of an entire function and the rate of its best polynomial approximation in different domains, see [1, 15, 3, 8, 16]. Kumar [10, 11] investigated the growth and approximation of entire function solutions of the Helmholtz equation. Vakarchuk and Zhir [22, 20, 21, 23] studied some problems of approximation of entire transcendental functions in some Banach spaces. It has been noticed that the study of growth of entire transcendental functions in terms of En(f,X) in several complex variables has not been done extensively (see [12, 13, 14, 19, 24, 25]) as in a single complex variable. Vakarchuk and Zhir [24] obtained the necessary and sufficient condition for fX to be an entire transcendental function of the generalized order of growth ρm(f;α,β) in terms of En(f,X), where X is one of the Banach spaces of functions analytic in Um. It is significant to mention here that these results do not hold for α=β. So, in the present paper we have tried to refine this scale. To the best of our knowledge, the generalized type Tm(f;α,α) has not been characterized in terms of En(f,X) in m-complex variables so far. The results of Vakarchuk and Zhir [24] can be improved by our results.

2 Main results

Theorem 2.1.

Let α(x)Ω¯ and let f(z)=|k|=0ck(f)zk be an entire function of m complex variables of generalized order ρm(f;α,α), 1<ρm(f;α,α)<. Then f(z) is of generalized type Tm(f;α,α) if and only if

(2.1)TTm(f;α,α)=lim sup|𝐤|α(|𝐤|ρ){α[ρρ-1log|c𝐤(f)|-1|𝐤|]}(ρ-1),ρρm(f;α,α),

provided that dF(x;T,ρ)/dlogx=O(1) as x for all T, 0<T<.

Proof.

By the definition of T, we have

TTm(f;α,α)=lim supRα(logMf,D(R))(α(logR))ρ.

Suppose T<. Then, for every ε>0, there exists R(ε) such that

α(logMf,D(R))(α(logR))ρ<T+ε=T¯for all R>R(ε)

or

(2.2)logMf,D(R)<(α-1{T¯[α(logR)]ρ}).

The minimum value of the right-hand side of (2.2) is attained at R=R(|𝐤|) and satisfies the equation

|𝐤|=ρlogRF[logR;T¯,1ρ].

Then

logR=α-1[(1T¯α(|𝐤|ρ))1(ρ-1)]=F[|𝐤|ρ;1T¯,ρ-1].

By the Cauchy inequality, we have

|c𝐤(f)|R-𝐤Mf,D(R)R-|𝐤|Mf,D(R)exp{-|𝐤|logR+(α-1{T¯[α(logR)]ρ})}

or

|c𝐤(f)|exp{-|𝐤|F+|𝐤|ρF}

or

ρρ-1log|c𝐤(f)|-1|𝐤|α-1[(1T¯α(|𝐤|ρ))1(ρ-1)]

or

T¯α(|𝐤|ρ){α[ρρ-1log|c𝐤(f)|-1|𝐤|]}(ρ-1).

Proceeding to limits, we obtain

(2.3)Tlim sup|𝐤|α(|𝐤|ρ){α[ρρ-1log|c𝐤(f)|-1|𝐤|]}(ρ-1).

For T=, the inequality (2.3) obviously holds.

In order to prove the reverse inequality in (2.3), we suppose that

lim sup|𝐤|α(|𝐤|ρ){α[ρρ-1log|c𝐤(f)|-1|𝐤|]}(ρ-1)=T1.

Let T1<. Then, for every ε>0 and for all k1N1(ε),k2N2(ε),,kmNm(ε) or |𝐤|mN(ε), taking N=min(N1,N2,,Nm), we have

α(|𝐤|ρ){α[ρρ-1log|c𝐤(f)|-1|𝐤|]}(ρ-1)T1+ε=T¯1.

or

|c𝐤(f)|1exp{(ρ-1)|𝐤|ρF[|𝐤|ρ;1T1¯,ρ-1]}.

The inequality

(2.4)(|c𝐤(f)|R|𝐤|)1|𝐤|Re-(ρ-1ρ)F[|𝐤|ρ;1T1¯,ρ-1]12

is satisfied for some |𝐤|=|𝐤|(R). Then

(2.5)|𝐤|=|𝐤|(R)+1|c𝐤(f)|R|𝐤||𝐤|=|𝐤|(R)+112|𝐤|1.

From inequality (2.4), we have

2Rexp{(ρ-1ρ)F[|𝐤|ρ;1T1¯,ρ-1]}.

Taking |𝐤|(R)=E[ρα1{T¯(α(logR+log2))ρ-1}] and letting φ(x)=Rxexp{-(ρ-1ρ)xF[xρ;1T1¯,ρ-1]}, we get

(2.6)φ(x)φ(x)=logR-(ρ-1ρ)F[xρ;1T1¯,ρ-1]-dF[xρ;1T1¯,ρ-1]dlogx=0as x.

By the assumption of the theorem, for finite T1(0<T1<), dF[x;T1¯,ρ-1]dlogx is bounded. So, there exists A>0 such that for x>x1, we have

(2.7)|dF[xρ;1T1¯,ρ-1]dlogx|A.

We can take A>log2. Inequalities (2.4) and (2.5) hold for |𝐤||𝐤|(R)+1. We let k0 designate the number max(N(ε),E[x1]+1). For R>R1(k0), we have φ(k0)φ(k0)>0. From (2.7) and (2.6) it follows that φ(|𝐤|R+1)φ(|𝐤|R+1)<0. Therefore, if for R>R1(k0), we let x*(R) designate the point where φ(x*(R))=maxk0x|𝐤|R+1φ(x), then k0x|𝐤|R+1 and x*(R)=ρα-1{T1¯(α(logR-a(R)))ρ-1}, where

-A<a(R)=|dF[xρ;1T1¯,ρ-1]dlogx|x=x*(R)A.

Further,

maxmk0<|𝐤|<|𝐤|R+1(|c𝐤(f)|R|𝐤|)maxmk0<|𝐤|<|𝐤|R+1φ(x)
=Rρα-1{T1¯(α(logR-a(R)))ρ-1}eρα-1{T1¯(α(logR-a(R)))ρ-1}(logR-a(R))
=exp{a(R)ρα-1{T1¯(α(logR-a(R)))ρ-1}}
exp{Aρα-1{T1¯(α(logR+A))ρ-1}}.

It is clear that (for R>R1(k0))

Mf,D(R)|𝐤|=0|c𝐤(f)|R|𝐤|
=|𝐤|=0k0|c𝐤(f)|R|𝐤|+|𝐤|=k0+1|𝐤|R+1|c𝐤(f)|R|𝐤|+|𝐤|=|𝐤|R+1|c𝐤(f)|R|𝐤|
O(Rk0)+(|𝐤|R+1)maxmk0<|𝐤|<|𝐤|R+1(|c𝐤(f)|R|𝐤|)+1

or

Mf,D(R)(1+o(1))exp{(Aρ+o(1))α-1[T1¯(α(logR+A))ρ-1]}

or

α(logMf,D(R))T1¯[α(logR+A)]ρ-1T1¯[α(logR+A)]ρ.

This gives

α[(Aρ+o(1))-1logMf,D(R)][α(logR+A)]ρT1¯=T1+ε.

Since α(x)Ω¯Δ, we then have

(2.8)lim supRα[logMf,D(R)][α(logR+A)]ρT1.

Using the definition of Tm(f;α,α) and combining inequalities (2.8) and (2.3), we get the required result. ∎

Theorem 2.2.

Let α(x)Ω¯. Then a necessary and sufficient condition for an entire transcendental function f(z)Bm(p,q,λ) to be of generalized type Tm(f;α,α) having finite generalized order ρm(f;α,α),1<ρm(f;α,α)< is

(2.9)TTm(f;α,α)=lim supnα(nρ){α[ρρ-1log(|En(f,Bm(p,q,λ))|-1n)]}(ρ-1).

Proof.

We prove the result in two steps. First, we consider the case where q=2, i.e., the space Bm(p,2,λ), where 0<p<2 and λ1. Let f(𝐳)Bm(p,q,λ) be of generalized type T with generalized order ρ. Then, from Theorem 2.1, we have

(2.10)lim sup|𝐤|α(|𝐤|ρ){α[ρρ-1log|c𝐤(f)|-1|𝐤|]}(ρ-1)=T.

For a given ε>0, there exists a natural number no=no (ε>0) such that the inequality

α(|𝐤|ρ){α[ρρ-1log|c𝐤(f)|-1|𝐤|]}(ρ-1)T+ε=T¯

is true for all |𝐤|>no. Thus, we have

(2.11)|c𝐤(f)|1exp{(ρ-1)|𝐤|ρF[|𝐤|ρ;1T1¯,ρ-1]}.

Denote

{κBm(p,2,λ)*(n+1)}-1=max|𝐤|=n+1{κBm(p,2,λ)*(𝐤)}-1.

Let Tn(f,𝐳)=|𝐤|=0nc𝐤(f)𝐳𝐤 be the nth partial sum of the Taylor series of the function f. Following [24, p. 1805], we obtain

En(f;Bm(p,2,λ)){κBm(p,2,λ)*(n+1)}-1λ{|𝐤|=n+1|c𝐤(f)|2}12.

By using (2.11), we have

(2.12)En(f;Bm(p,2,λ)){κBm(p,2,λ)*(n+1)}-1λexp{(ρ-1)n+1ρF[n+1ρ;1T¯,ρ-1]}{|𝐤|=n+1Ω𝐤2(α,α)}12,

where

Ωk(α,α)exp{n+1ρ(ρ-1)[α-1{(α(n+1ρ)T+ε)1ρ-1}]}exp{|𝐤|ρ(ρ-1)[α-1{(α(|𝐤|ρ)T+ε)1ρ-1}]}.

Set

Ω^(α,α)exp{-1ρ(ρ-1)[α-1{(α(1ρ)T+ε)1ρ-1}]}.

Clearly, Ω^(α,α)<1. Since α(x) is increasing and |𝐤|>n+1, we get

(2.13)Ω𝐤(α,α)exp{(n+1)-|𝐤|ρ(ρ-1)[α-1{(α(n+1ρ)T+ε)1ρ-1}]}Ω^|𝐤|-(n+1)(α,α).

From (2.12) and (2.13), we get

(2.14)En(f;Bm(p,2,λ)){κBm(p,2,λ)*(n+1)}-1λ(1-Ω^2(α,α))m2[exp{n+1ρ(ρ-1)[α-1{(α(n+1ρ)T+ε)1ρ-1}]}].

For n>no,n, this yields

(2.15)T+εα(n+1ρ){α(ρ(1+1n)(ρ-1){log(|En(f;Bm(p,2,λ))|-1n)+log({κBm(p,2,λ)*(n+1)}-1λ(1-Ω^2(α,α))m2)})}(ρ-1).

Passing to the limit as n on the right-hand side of (2.15) and taking into account the fact that the number ε>0 is chosen arbitrarily, we get

(2.16)Tlim supnα(nρ){α(ρ(ρ-1){log(|En(f;Bm(p,2,λ))|-1n)})}(ρ-1).

In order to prove the reverse inequality, by [24, p. 1804], we have

|c𝐤(f)|j=1mBkj,p,2,λf-Tn(f)p,2,λ=En(f;Bm(p,2,λ)).

For sufficiently large n, we get

α(nρ){α(ρ(ρ-1){log(|En(f;Bm(p,2,λ))|-1n)})}(ρ-1)α(nρ){α(ρ(ρ-1){(1+1n)log(|En(f;Bm(p,2,λ))|-1n+1)})}(ρ-1)
α((1-1|𝐤|)|𝐤|ρ)[α{ρρ-1(1+1|𝐤|-1)(log(|c𝐤(f)|-1|𝐤|)+log(κBm(p,2,λ)*(𝐤))1|𝐤|)}]ρ-1,

where

κBm(p,2,λ)*(𝐤)=(j=1mBkj,p,2,λ)-1

is a constant which depends on the space Bm(p,2,λ) and 𝐤, and does not depend on f.

Now proceeding to the limit as n (|𝐤|) and taking into account equality (2.1), we get

(2.17)lim supnα(nρ){α(ρ(ρ-1){log(|En(f;Bm(p,2,λ))|-1n)})}(ρ-1)T.

Combining (2.16) and (2.17), we get the required result.

Now, in the second step, we consider the spaces Bm(p,q,λ) for 0<p<q, q2, and q,λ1. Note that (see [5, p. 103]) for pp1, qq1 and λλ1, if at least one of the inequalities is strict, then the strict inclusion Bm(p,q,λ)Bm(p1,q1,λ1) holds, and for any function fBm(p,q,λ), the following relation is true:

(2.18)fp,q,λ1Cp1,q1,λ1;p,q,λfp,q,λ,

where Cp1,q1,λ1;p,q,λ is a positive constant that depends only on the indicated subscripts and is independent of f. In view of the definition of the best polynomial approximation, the inequality (2.18) yields

En(f;Bm(p1,q1,λ1))Cp,q,λ;p1,q1,λ1En(f;Bm(p,q,λ)),

where fBm(p,q,λ), and Cp,q,λ;p1,q1,λ1 is a constant independent of f and n.

For the general case Bm(p,q,λ),q2, we prove the necessity of condition (2.9). Let fBm(p,q,λ) be an entire transcendental function having finite generalized order ρm(f;α,α), whose generalized type is defined by (2.9). We set

{κBm(p,q,λ)*(n+1)}-1=max|𝐤|=n+1{κBm(p,q,λ)*(𝐤)}-1

and write (see [10, p. 1807])

(2.19)En(f;Bm(p,q,λ)){κBm(p,q,λ)*(n+1)}-1λ{|𝐤|=n+1|c𝐤(f)|}.

Now, using the relation (2.11) and (2.19), and the same reasoning as in deducing relations (2.12)–(2.14) for n>no, n, we obtain

(2.20)En(f;Bm(p,q,λ)){κBm(p,q,λ)*(n+1)}-1λ(1-Ω^2(α,α))m2[exp{n+1ρ(ρ-1)[α-1{(α(n+1ρ)T+ε)1ρ-1}]}].

For n>no, (2.20) gives

T+εα(n+1ρ){α(ρ(1+1n)(ρ-1){log(|En(f;Bm(p,q,λ))|-1n)+log({κBm(p,q,λ)*(n+1)}-1λ(1-Ω^2(α,α))m2)})1n}(ρ-1).

Since Ω^(α,α)<1, and αΩ^, proceeding to limits, we get

Tlim supnα(nρ){α(ρ(ρ-1){log(|En(f;Bm(p,q,λ))|-1n)})}(ρ-1).

For the reverse inequality, let 0<p<q<2 and λ,q1. By virtue of relation (2.9), with p1=p, q1=2 and λ1=λ, and relation (2.8) already established for q=2, we get

lim supnα(nρ){α(ρ(ρ-1){log(|En(f;Bm(p,q,λ))|-1n)})}(ρ-1)lim supnα(nρ){α(ρ(ρ-1){log(|En(f;Bm(p,2,λ))|-1n)})}(ρ-1)=T.

Now let 0<p2<q. Since

M2(f;𝐫)Mq(f;𝐫),

where 𝐫Πm, from [24, p. 1808], we get

(2.21)En(f;Bm(p,q,λ))|c𝐤(f)|{κBm(p,q,λ)*(𝐤)}-1λ,

where |𝐤|=n+1. Then, for sufficiently large n, we have

α(nρ){α(ρ(ρ-1){log(|En(f;Bm(p,q,λ))|-1n)})}(ρ-1)α(|𝐤|ρ){α(ρ(ρ-1){log(|c𝐤(f)|-1|𝐤|+log{κBm(p,q,λ)*}1|𝐤|λ)})}(ρ-1).

Using (2.1) and proceeding to limits, we get

lim supnα(nρ){α(ρ(ρ-1){log(|En(f;Bm(p,q,λ))|-1n)})}(ρ-1)lim sup|𝐤|α(|𝐤|ρ){α(ρ(ρ-1){log(|c𝐤(f)|-1|𝐤|)})}(ρ-1)=T.

Now, using the definition of the best polynomial approximation (1.10) in relation (2.18), we get the following inequality:

(2.22)En(f;Bm(p1,q1,λ1)Cp,q,λ;p1,q1,λ1En(f;Bm(p,q,λ)).

Let 2p<q. Setting in (2.22) q1=q, λ1=λ and p1(0,2), where p1 is a fixed number, we get

(2.23)En(f;Bm(p,q,λ))En(f;Bm(p1,q,λ))Cp,q,λ;p1,q,λ.

Substituting p1=p in (2.21), we obtain the following relation:

(2.24)En(f;Bm(p1,q,λ))|c𝐤(f)|{kBm(p,q,λ)*(𝐤)}-1λ.

Now combining (2.23) and (2.24), we get

(2.25)En(f;Bm(p,q,λ))|c𝐤(f)|Cp,q,λ;p1,q,λ{kBm(p,q,λ)*(𝐤)}-1λ,

where |𝐤|=n+1, and Cp,q,λ;p1,q,λ is a constant independent of n and f. Using (2.25) and applying the same analogy as in the previous case 0<p2<q, for sufficiently large n or |𝐤|, we have

α(nρ){α(ρ(ρ-1){log(|En(f;Bm(p,q,λ))|-1n)})}(ρ-1)
α(|𝐤|ρ){α(ρ(ρ-1){log(|c𝐤(f)|-1|𝐤|+log(Cp,q,λ;p1,q,λ)1|𝐤|+log{κBm(p1,q,λ)*}1|𝐤|λ)})}(ρ-1).

By applying limits and using (2.1), we get

lim supnα(nρ){α(ρ(ρ-1){log(|En(f;Bm(p,q,λ))|-1n)})}(ρ-1)T.

Hence, the proof is completed. ∎

Theorem 2.3.

Let α(x)Ω¯ and f(z)=|k|=0ck(f)zk be an entire function of m complex variables having generalized order ρm(f;α,α), 1<ρm(f;α,α)<. Then f(z)Hq(Um) is of generalized type ξm(f;α,α) if and only if

ξmξm(f;α,α)=lim sup|𝐤|α(|𝐤|ρ){α[ρρ-1log(|En(f,Hq(Um))|-1n)]}(ρ-1).

Proof.

Let f(𝐳)=|𝐤|=0c𝐤(f)𝐳𝐤 be an entire transcendental function of m complex variables having finite generalized order ρm(f;α,α) and generalized type Tm(f;α,α). Since

lim|𝐤||c𝐤(f)|-1|𝐤|=0

and f(𝐳)Bm(p,q,λ), where 0<p<q and q,λ1, in view of relation (1.9), we get

(2.26)En(f;Bm(q2,q,q))En(f;Hq(Um)),1q<.

In the case of the Hardy space H(Um), we have

(2.27)En(f;Bm(p,,))En(f;H(Um)),1p<.

By using (2.26), we can write

ξm=lim supnα(nρ){α[ρρ-1log(|En(f,Hq(Um))|-1n)]}(ρ-1)
lim supnα(nρ){α[ρρ-1log(|En(f,Bm(q2,q,q))|-1n)]}(ρ-1)
(2.28)Tm(f;α,α),1q<.

Using estimate (2.27), we can prove inequality (2.28) in the case q=. For the reverse inequality

(2.29)ξmT,

we use inequality (2.10), which is true for any |𝐤|>no, and estimate from above the generalized type Tm of f having finite generalized order ρm(f;α,α) as follows. We have

En(f;Hq(Um))f-Tn(f)Hq
|𝐤|=n+1|c𝐤(f)|
1exp{n+1ρ(ρ-1)[α-1{(α(n+1ρ)T+ε)1ρ-1}]}|𝐤|=n+1Ω^|𝐤|-n-1(α,α)
1(1-Ω^(α,α))[exp{n+1ρ(ρ-1)[α-1{(α(n+1ρ)T+ε)1ρ-1}]}]

or

1En(f;Hq(Um))(1-Ω^(α,α))[exp{n+1ρ(ρ-1)[α-1{(α(n+1ρ)T+ε)1ρ-1}]}].

This yields

T+εlim supnα(n+1ρ){α[ρρ-1log(|En(f,Hq(Um))|-1n+1)+log((1-Ω^(α,α))-1n+1)]}(ρ-1).

Since Ω^(α,α)<1, by using the properties of α and by passing to the limit as n, we obtain inequality (2.29). Thus, finally, we get ξm=T. Hence, the proof is completed. ∎

Remark 2.4.

We can find an analog of Theorem 2.3 for Bergman spaces by using (1.9) for 1q< and for q=, from Theorem 2.2.

Acknowledgements

The author is very much thankful to learned reviewers for giving fruitful comments that improved the paper.

References

[1] M. Z. Dveirin, On the relationship between the rate of polynomial approximation of an entire function and its order and type, Ukrainian Math. J. 67 (2015), no. 10, 1498–1514. 10.1007/s11253-016-1168-7Suche in Google Scholar

[2] B. A. Fuks, Introduction to Theory of Analytic Functions of many Complex Variables (in Russian), Fizmatgiz, Moscow, 1962. Suche in Google Scholar

[3] A. Giroux, Approximation of entire functions over bounded domains, J. Approx. Theory 28 (1980), no. 1, 45–53. 10.1016/0021-9045(80)90103-3Suche in Google Scholar

[4] A. A. Gol’dberg, Elementary remarks on the formulas for the determination of the order and type of entire functions of many complex variables, Dokl. Akad. Nauk Arm. SSSR 29 (1959), no. 4, 145–151. Suche in Google Scholar

[5] M. I. Gvaradze, On one class of spaces of analytic functions (in Russian), Candidate Degree Thesis (Physics and Mathematics), Tbilisi, 1975. Suche in Google Scholar

[6] M. I. Gvaradge, On one class of spaces of analytic functions, Mat. Zametki 21 (1977), no. 2, 141–150. Suche in Google Scholar

[7] M. I. Gvaradze, Factors of one class of analytic functions defined on a polydisk, Tr. Tbilis. Mat. Inst. 66 (1988), 15–21. Suche in Google Scholar

[8] I. I. Ibragimov and N. I. Shikhaliev, On the best polynomial approximation of analytic functions in a space of analytic functions, Dokl. Akad. Nauk SSSR 227 (1976), no. 2, 280–283. Suche in Google Scholar

[9] G. P. Kapoor and A. Nautiyal, Polynomial approximation of an entire function of slow growth, J. Approx. Theory 32 (1981), 64–75. 10.1016/0021-9045(81)90022-8Suche in Google Scholar

[10] D. Kumar, Approximation of entire function solutions of Helmholtz equation having slow growth, J. Appl. Anal. 18 (2012), 179–196. 10.1515/jaa-2012-0012Suche in Google Scholar

[11] D. Kumar, Generalized growth and approximation of entire function solution of Helmholtz equation in Banach spaces, Ann. Univ. Ferrara Sez. VII Sci. Mat. 62 (2016), no. 1, 83–97. 10.1007/s11565-015-0233-8Suche in Google Scholar

[12] D. Kumar, On the growth and approximation of transcendental entire functions on algebraic varieties, Int. J. Anal. Appl. 12 (2016), no. 1, 22–29. Suche in Google Scholar

[13] D. Kumar, On the growth and best polynomial approximation of entire functions in m(m2) in some Banach spaces, J. Anal. (2017), 10.1007/s41478-017-0033-x. 10.1007/s41478-017-0033-xSuche in Google Scholar

[14] V. M. Muradov, On the relationship between the best polynomial approximation of analytic functions of many complex variables and the Faber coefficients, Special Problems of the Theory of Functions, Academy of Sciences of Aserbaidschan, Baku (1986), 195–212. Suche in Google Scholar

[15] G. Ramesh and G. S. Srivastava, Approximation of entire functions of slow growth, General Math. 14 (2006), no. 2, 59–76. Suche in Google Scholar

[16] S. M. Shah, Polynomial approximation of an entire function and generalized orders, J. Approx. Theory 19 (1977), no. 4, 315–324. 10.1016/0021-9045(77)90095-8Suche in Google Scholar

[17] M. N. Sheremeta, On the relationship between the growth of the maximum modulus of an entire function and the moduli of coefficients of its power expansions, Izv. Vyssh. Uchebn. Zaved. Ser. Mat. (1967), no. 2, 100–108. Suche in Google Scholar

[18] M. N. Sheremeta, On the relationship between the growth of entire or analytic functions of zero order in a disk and the coefficients of their power expansions, Izv. Vyssh. Uchebn. Zaved. Ser. Mat. (1968), no. 6, 115–121. Suche in Google Scholar

[19] S. B. Vakarchuk, On the best approximation by generalized polynomials in one space of analytic functions of two complex variables, Izv. Vyssh. Uchebn. Zaved. Ser. Mat. (1991), no. 7, 14–25. Suche in Google Scholar

[20] S. B. Vakarchuk, On the best polynomial approximation of entire transcendental functions in Banach spaces. I, Ukraïn. Mat. Zh. 47 (1994), no. 9, 1123–1133. Suche in Google Scholar

[21] S. B. Vakarchuk, On the best polynomial approximation of entire transcendental functions in Banach spaces. II, Ukraïn. Mat. Zh. 47 (1994), no. 10, 1318–1322. 10.1007/BF01066088Suche in Google Scholar

[22] S. B. Vakarchuk, On the best polynomial approximation of functions analytic in the unit circle in Banach spaces, Mat. Zametki 55 (1994), no. 4, 6–14. Suche in Google Scholar

[23] S. B. Vakarchuk and S. I. Zhir, On some problems of polynomial approximation of entire transcendental functions, Ukrainian Math. J. 54 (2002), no. 9, 1393–1401. 10.1023/A:1023407416027Suche in Google Scholar

[24] S. B. Vakarchuk and S. I. Zhir, On the best polynomial approximations of entire transcendental functions of many complex variables in some Banach spaces, Ukrainian Math. J. 66 (2015), no. 12, 1763–1811. 10.1007/s11253-015-1052-xSuche in Google Scholar

[25] S. I. Zhir and S. B. Vakarchuk, Some problems of the best polynomial approximation of entire transcendental functions of one and many complex variables, Abstract of the International Conference “Approximation Theory of Functions and Its Applications” on the 70th Birthday of O.İ. Stepanets, Ukrainian National Academy of Science, Kyiv (2012), 44–45. Suche in Google Scholar

Received: 2016-4-23
Revised: 2017-3-24
Accepted: 2017-5-1
Published Online: 2017-5-19
Published in Print: 2017-6-1

© 2017 Walter de Gruyter GmbH, Berlin/Boston

Heruntergeladen am 21.11.2025 von https://www.degruyterbrill.com/document/doi/10.1515/jaa-2017-0002/html
Button zum nach oben scrollen