the space of uniform limit power functions

CHUANYI ZHANG AND WEIGUO LIU

Received 10 October 2005; Revised 16 June 2006; Accepted 5 July 2006

To answer a question proposed by Mari in 1996, we proposeᐁᏸᏼα(R⁺), the space of uniform limit power functions. We show thatᐁᏸᏼα(R⁺) has properties similar to that ofᏭᏼ(R⁺). We also proposed three other limit power function spaces.

Copyright © 2006 Hindawi Publishing Corporation. All rights reserved.

1. Introduction

In literature of Fourier transforms and Wavelet transforms, the basic space is L2(R).

From the point of view of signal analysis, a signal f ∈L2(R) can only be transient (or

“wavelets”). During recent years in some application areas, it has become more common to motivate a theory via persistent rather than transient signals (e.g., [16,28,30,42]).

To work on persistent signals, people have to seek a space different fromL2(R). One important example of such spaces isᏭᏼ(R), the space of almost periodic functions. Peo- ple have developed a profound theory and applications forᏭᏼ(R) (e.g., see [4,5,7–

15,18,19,24,26,30,32,37,38]).

As in [30], a function f is called limit power if the limit

Tlim→∞

1 2T

_T

−T

f(t)²dt (1.1)

exists. Denote byH2the set of all such functions.

It is well known thatᏭᏼ(R)⊂H2and so is the Besicovitch spaceB2[5], the comple- tion ofᏭᏼ(R) inH2. In fact, many useful persistent signals are inH2, for example, the bounded power signals studied in Wiener’s generalized harmonic analysis [36]. However, H2 is not a linear set. An example in [29] shows thatH2 is not closed under addition.

The lack of closedness under addition caused some difficulties in Robust control (e.g., see [27]).

As [29] pointed out that except for some subsets ofH2which are already known to be vector spaces (e.g.,L2(R),{f ∈L_∞(R) : lim_|t|→∞f(t) exists},Ꮽᏼ(R)), it is not clear whether a “nice” (e.g., Hilbert) large vector space could be defined.

Hindawi Publishing Corporation

International Journal of Mathematics and Mathematical Sciences Volume 2006, Article ID 17042, Pages1–14

DOI 10.1155/IJMMS/2006/17042

Let us recall when the spaces mentioned above were invented. The latest one isB2

invented by Besicovitch in 1926 (see [5]); a year earlier isᏭᏼ(R⁺) invented by Bohr [8–

10];L2(R) was invented even earlier. We remark that the function set studied by Wiener mentioned above is not closed under addition either. Some generalizations ofᏭᏼ(R), for example, the functions studied in [1,2,17,21,32–34,38], are vector spaces. They are larger thanᏭᏼ(R) inᏯ(R). However, they are the same withᏭᏼ(R) inH2. We remark that thoughH2is not linear, there have been Banach spaces containingH2, for example, the spaceB² proposed in [11] (in [28] for the discrete setting). However, [11,28] use lim instead of lim in (1.1) to construct the spaces. In many cases, lim is needed too.

The background of [29] and related problems being pointed out by some authors (e.g., [27,28,30] and references therein) show real needs for new, larger, nice spaces inH2.

The purpose of the paper is to propose such spaces. One will see that the new spaces are so natural that they come from what we call generalized trigonometric polynomials in the same way asᏭᏼ(R) andB2come from trigonometric polynomials. One will also see that they are so huge that to compareᏭᏼ(R) andB2with them is the same as to compare one point withR⁺.

The layout of the paper is as follows. In the next section, we show the existence of a larger orthonormal basis. In Section 3, we develop a theory of uniform limit power functions in a way parallel to that ofᏭᏼ(R) (e.g., [12,13]). InSection 4, we discuss the limit power type functions.

2. Orthonormal basis

It is well known that{e^iλt}is a complete orthonormal basis inB2[5]. In this section, we consider the set

e^iλtα, (2.1)

whereλ∈Rand 0< α <∞.

Whenα >1, in radar and sonar terminology, the functione^iλtαrepresents a chirp sig- nal because it is reasonably well defined but steadily rising frequency. By analyzingf(t)= sin(πt²), [25, Chapter 2] points out the fact that a chirp has a well-defined instantaneous frequency and ordinary Fourier analysis hides the fact. By using Windowed Fourier trans- form, the signal is reasonably well localized both in time and in frequency. In particular, whenα=2, the functione^it2, being an underlying kernel (e.g., in oscillatory integrals, optics, etc.), has important applications; we refer the reader to [3,6,20,22,23,31,35]

for details.

Whenα <1, the functione^iλtαbehaves conversely.

As{e^iλt}, the set{e^iλtα}is also orthonormal. We show this in the next two theorems.

Theorem 2.1. Forα≥β≥0 withα≥1 andλ,μ∈Rwithλ=0, the following limit

Tlim→∞

1 T

_T+a

a e^{i(λtα+μtβ)}dt=

⎧⎨

⎩

1, α=β,μ= −λ,

0, otherwise (2.2)

exists uniformly with respect toa∈R⁺.

Proof. The conclusion for the case ofα=β,μ= −λis obvious, so we only consider the other cases. In these cases, if we can finda0>0 such that

1 T

_T+a

a e^{i(λtα+μtβ)}dt−→0 (T−→ ∞) (2.3) uniformly with respect toa∈[a0,∞), then we have

1 T

_T+a

a e^{i(λtα+μtβ)}dt−→0 (T−→ ∞) (2.4) uniformly with respect toa∈R⁺. In fact, fora∈[0,a0], one has

1

T _T+a

a e^{i(λtα+μtβ)}dt≤ 1 T

^a0

a + _T+a0

a0

− _T+a0

T+a

e^{i(λtα+μtβ)}dt

≤2a0

T + 1 T

_T+a0

a0

e^{i(λtα+μtβ)}dt−→0 (T−→ ∞).

(2.5)

Note thatα,β,λ, andμare fixed, we can chosea0>1 such that|λα+μβa^β−α| ≥ε0>0 for some0>0 and alla∈[a0,∞). In fact, ifβ=αthenλ+μ=0 and for alla∈[1,∞) one has|λα+μβa^β−α| = |α(λ+μ)| =ε0>0; ifβ < α, thent^β−α→0 ast→ ∞, and therefore there existsa0>1 such that for alla∈[a0,∞) one has|λα+μβa^β−α| ≥ |λα/2| =ε0>0.

For suchaandt≥a,λαt^α−1+μβt^β−1=t^α−1[λα+μβa^β−α]=0 and _T+a

a e^{i(λtα+μtβ)}dt

= _T+a

a

e^{i(λtα+μtβ)}iλαt^α−1+μβt^β−1 iλαt^α−1+μβt^β−1 dt=

_T+a

a

e^{i(λtα+μtβ)}diλt^α+μt^β iλαt^α−1+μβt^β−1

= e^{i(λtα+μtβ)} iλαt^α−1+μβt^β−1

^T+a

a −1

i _T+a

a e^{i(λtα+μtβ)}d 1

(λαt^α−1+μβt^β−1)⁼I1+I2. (2.6)

So

|I1| ≤ 1

λα(T+a)^α−1+μβ(T+a)^β−1+ 1

λαa^α−1+μβa^β−1

= 1

(T+a)^α−1λα+μβ(T+a)^β−α+ 1

a^α−1λα+μβa^β−α≤M1,

(2.7)

whereM1is a constant which is independent ofTanda∈[a0,∞).

To estimateI2, we have I2≤

_T+a

a

λα(α−1)t^α−2+μβ(β−1)t^β−2 λαt^α−1+μβt^β−12

dt= _T+a

a

λα(α−1) +μβ(β−1)t^β−α t^2−αλαt^α−1+μβt^β−12

dt

≤ _T+a

a

λα(α−1)+μβ(β−1)

t^αλα+μβt^β−α2 dt≤M2

_T+a

a

1 t^αdt

=

⎧⎪

⎪⎪

⎪⎨

⎪⎪

⎪⎪

⎩

M2 ln T a+ 1

≤M2ln(T+ 1), α=1, M2 1

a α−1

− 1 (T+a)

α−1

, α >1,

(2.8) whereM2is a constant which is independent ofTanda∈[a0,∞).

It follows from (2.6)–(2.8) that

Tlim→∞

1 T

_T+a

a e^{i(λtα+μtβ)}=0 (2.9)

uniformly with respect toa∈[a0,∞), and therefore with respect toa∈R⁺, the proof is

complete.

Corollary 2.2. Forα≥1 andλ=0, the limit

Tlim→∞

1 T

_T+a

a e^iλtαdt=0 (2.10)

exists uniformly with respect toa∈R⁺.

Proof. Putμ=0 inTheorem 2.1to get the conclusion.

Theorem 2.3. Let 1> α≥β≥0 withα >0 andλ,μ∈Rwithλ=0. Then

Tlim→∞

1 T

_T

0 e^{i(λtα+μtβ)}dt=

⎧⎨

⎩

1, α=β,μ= −λ,

0, otherwise. (2.11)

Proof. First, we consider the caseα=βandw=λ+μ=0, _T

1 e^iwtαdt= _T

1

e^iwtαiwαt^α−1

iwαt^α−1 dt= e^iwtα iwαt^α−1

^T

1 − _T

1

e^iwtα

iwαt^α(1−α)dt. (2.12)

So

1 T

_T

0 e^iwtαdt= 1 T

1 0+

_T

1 e^iwtαdt

−→0 (2.13)

asT→ ∞.

Ifα > β, letT > a >1 be so large that|λα|>|μβa^β−α|. As in the proof ofTheorem 2.1, one has

_T

a e^{i(λtα+μtβ)}dt=I3+I4, I3≤ 1

λαT^α−1+μβT^β−1+ 1

λαa^α−1+μβa^β−1

≤ T^1−α

λα+μβT^β−α+ a^1−α λα+μβa^β−α

≤T^1−α

1

λα+μβT^β−α+ 1 λα+μβa^β−α

≤T^1−α

1

|λα| −μβa^β−α+ 1

|λα| −μβa^β−α

≤M3T^1−α.

(2.14)

To estimateI4we have the following:

I4≤ _T

a

λα(α−1)t^α−2+μβ(β−1)t^β−2 λαt^α−1+μβt^β−12

dt

= _T

a

λα(α−1)t^α−2+μβ(β−1)t^β−2 t^2(α−1)λα+μβt^β−α2

dt

= _T

a

λα(α−1)t^−α+μβ(β−1)t^β−2α λα+μβt^β−α2

dt

≤ _T

a

λα(α−1)t^−α+μβ(β−1)t^β−2α |λα| −μa^β−α2 dt

≤M4t^1−αT_a +M5t^1+β−2αT_a,

(2.15)

whereM4andM5are constants which are independent ofT.

It follows that 1

T _T

0 e^{i(λtα+μtβ)}dt= 1 T

_a

0 + _T

a e^{i(λtα+μtβ)}dt

≤ 1

T _a

0 e^{i(λtα+μtβ)}dt+1 T

_T

a e^{i(λtα+μtβ)}dt−→0

(2.16)

asT→ ∞. The proof is complete.

It follows from Theorems2.1and2.3that

Tlim→∞

1 T

_T

0

e^iλtα_·e^−iμtβdt₌

⎧⎨

⎩

1, α=β,μ=λ,

0, otherwise. (2.17)

That is, the set{e^iλtα}constitutes an orthonormal basis.

Remark 2.4. Since the domain of the functione^iλtα in general isR⁺, we considerR⁺only in the paper. For special numbers ofα, for example,αsare positive integers, the domain will beR. In this case all the results will hold forR.

3. Uniform limit power functions We call the functions

n k=1

a_ke^iλktα (3.1)

α-trigonometric polynomial, whereak∈Candλk∈R. AsᏭᏼ(R), we have the following definition.

Definition 3.1. Letα >0 be fixed. A function f onR⁺is called uniform limit power if for each>0 there exists anα-trigonometric polynomialPsuch that

f −P=supf(t)−P(t):t∈R⁺

<. (3.2)

Denote byᐁᏸᏼα(R⁺) the set of all such functions.

One sees that whenα=1,ᐁᏸᏼα(R⁺)=Ꮽᏼ(R⁺). Also one sees thatᐁᏸᏼα(R⁺) is the completion ofα-trigonometric polynomial inC(R⁺). Since the set ofα-trigonometric polynomials are closed under addition, multiplication, and conjugation, so is the com- pletionᐁᏸᏼα(R⁺). Thus we have shown the following statement:ᐁᏸᏼα(R⁺) is aC^∗- subalgebra ofC(R⁺) containing the constant functions.

Next, we discuss the Fourier expansion of f ∈ᐁᏸᏼα(R⁺). First of all, we show that the mean exists.

Theorem 3.2. If f ∈ᐁᏸᏼα(R⁺), then

Tlim→∞

1 T

_T

0 f(t)dt (3.3)

exists. In the case ofα≥1,

Tlim→∞

1 T

_T+a

a f(t)dt (3.4)

exists uniformly with respect toa∈R⁺.

Proof. We first show the theorem in the case thatf is anα-trigonometric polynomial. Let f(t)=P(t)=c0+

n k=1

cke^iλktα. (3.5)

Then by Theorems2.1and2.3,

Tlim→∞

1 T

_T

0 P(t)dt=c0. (3.6)

If f is an arbitrary function inᐁᏸᏼα(R⁺) then for >0 there exists an α-trigo- nometric polynomialP such that (3.2) holds. Since limT→∞(1/T)₀^TP(t)dt exists, we can find a numberT0such that whenT1,T2> T0,

1 T1

_T1

0 P(t)dt− 1 T2

_T2

0 P(t)dt<. (3.7) It follows from (3.2) and the last inequality above that whenT1,T2> T0,

1 T1

_T1

0 f(t)dt− 1 T2

_T2

0 f(t)dt≤ 1 T1

_T1

0

f(t)−P(t)dt

+1 T1

_T1

0 P(t)dt− 1 T2

_T2

0 P(t)dt + 1

T2

_T2

0

f(t)−P(t)dt <3.

(3.8)

Similarly, one shows the existence of the second limit. The proof is complete.

We call the limit inTheorem 3.2the mean off and denote it byM(f).

Forλ∈Rand f ∈ᐁᏸᏼα(R⁺) since the function f e^−iλtα is inᐁᏸᏼα(R⁺), the mean exists for the function. We write

a(λ)=Mf e^−iλtα. (3.9)

As the proof forᏭᏼ(R⁺) (see [12,13,18,26,37,38]), for a function f ∈ᐁᏸᏼα(R⁺) the frequency set

Freq(f)=

λ∈R:a(λ)=0 (3.10)

is countable (or finite). Let Freq(f)= {λk}and Ak=a(λk). Thus f has an associated Fourier series

f(t)∼^∞

k=1

Ake^iλktα, (3.11)

and Parseval’s equality holds:

∞ k=1

aλk²=M|f|²

. (3.12)

The unique theorem for almost periodic function is well known. That is, distinct al- most periodic functions have distinct Fourier series. We point out that this is also true for ᐁᏸᏼα(R⁺). To show this we need to set up some correspondence betweenᐁᏸᏼα(R⁺) andᏭᏼ(R⁺). For theα-trigonometric polynomialP inDefinition 3.1, lets=t^α. Then Pbecomes trigonometric polynomial ofs. That is,

P(s)= n k=1

ake^iλks. (3.13)

For f ∈ᐁᏸᏼα(R⁺) define the function

f(s)= fs^1/α. (3.14)

Thus, (3.2) becomes

f(s)−P(s)<

s∈R⁺

. (3.15)

So, f∈Ꮽᏼ(R⁺). Conversely, let h∈Ꮽᏼ(R⁺). By the approximation theorem of Ꮽᏼ(R⁺) for>0, there exists a trigonometric polynomialⁿ_k=1a_ke^iλkssuch that

h(s)− n k=1

a_ke^iλks<

s∈R⁺

. (3.16)

Lets=t^α(t∈R⁺) and leth(t)=h(t^α). It follows that h(t)−

n k=1

ake^iλktα<

t∈R⁺

. (3.17)

Therefore, we have h∈ᐁᏸᏼα(R⁺). Thus (3.14) is the correspondence between ᐁᏸᏼα(R⁺) andᏭᏼ(R⁺).

Note the translate property of almost periodic function, that is, for>0 there exists l >0 with the property that any intervalI⊂R⁺of lengthlhas a numberτ∈Isuch that

f(s+τ)−f(s)<

s∈R⁺

. (3.18)

By the correspondence (3.14), we have in fact already shown the following theorem.

Theorem 3.3. Let f ∈C(R⁺). Then the following statements are equivalent:

(1) f ∈ᐁᏸᏼα(R⁺);

(2) for>0 there existsl >0 with the property that any intervalI⊂R⁺of lengthlhas a numberτ∈Isuch that

f(t+τ)^1/α−ft^1/α<. (3.19) Furthermore, if f ∈ᐁᏸᏼα(R⁺) then so is|f(·)|.

Now, we make use of the unique theorem forfto get the same conclusion for f.

Lemma 3.4. Let f ∈ᐁᏸᏼα(R⁺) be nonnegative and f(t0)>0 for some t0∈R⁺. Then M(f)>0.

Proof. Let fbe the function in (3.14). So f(s)≥0 and f(s0)>0, whereso=t0^α. Since f∈Ꮽᏼ(R⁺), one has

Tlim→∞

1 T

_T+a

a

f(s)ds=M(f)>0 (3.20) uniformly with respect toa∈R⁺.

To show the theorem we need to discuss two cases.

(1)α≥1:

1 T

_T

1 f(t)dt= 1 T

_T^α

1 fs^1/αs^1/α−1 α ds=1

α 1 T

_T^α

1

f(s) s^1−1/αds

≥1 α

1 T

_T^α

1

f(s)

T^α1−1/αds= 1 α

1 T^α

_T^α

1

f(s)ds.

(3.21)

It follows from (3.20) that

Mf(t)≥1

αM f(s)>0. (3.22)

(2) 0< α <1: in this case, 1

T _T

0 f(t)dt= 1 T

_T^α

0 fs^1/αs^1/α−1 α ds=1

α 1 T

_T^α

0

f(s)s^1/α−1ds

≥1 α

1 T

_T^α

T^α/2

f(s) T^α 2

1/α−1

ds= 1 α

1 T

T^1−α 2^1/α−1

_T^α

T^α/2

f(s)ds

= 1 α2^1/α

1 T^α/2

_T^α

T^α/2

f(s)ds.

(3.23)

So, in this case we also have

Mf(t)≥ 1

α2^1/αM f(s)>0. (3.24)

The proof is complete.

By the lemma above we are able to show the following unique theorem.

Theorem 3.5. Distinct uniform limit power functions have distinct Fourier series.

Proof. Suppose that the distinct functions f1,f2∈ᐁᏸᏼα(R⁺) have the same Fourier se- ries. Then f1−f2will have Fourier series of all term zero. By Parseval’s equalityM(|f1− f2|²)=0. However, byLemma 3.4we getM(|f1−f2|²)>0. This is a contraction. The

proof is complete.

Forf(t)∈ᐁᏸᏼα(R⁺) since f(s)∈Ꮽᏼ(R⁺), the functionfhas an associated Fourier series

f(s)∼^∞

k=1

a_ke^iμks, (3.25)

whereak=M(f(s)e^−iμks). It is well known (e.g., see [12,13,38]) that f can be approxi- mated uniformly onRby the Bocher-Fej´er trigonometric polynomials

σ_m(s)=

n(m)

k=1

r_m,ka_ke^iμks, (3.26)

where the rational numbers 0≤r_m,k≤1 and limm→∞r_m,k=1. Thus, replacingsin (3.26) byt^αwe have

^n(m)

k=1

rm,kake^iμktα−ft^α= ^n(m)

k=1

rm,kake^iμktα−f(t)−→0, (3.27) uniformly onR⁺.

The following remark tells us an important conclusion.

Remark 3.6. In the section all the results are achieved under the assumption of fixed α. We may release the restriction onα. Let{α1,α2,. . .,αn} be nonnegative sequence. A generalized trigonometric polynomial is a function of the form

n k=1

ake^iλktαk, (3.28)

whereak∈Candλk∈R, 1≤k≤n. If inDefinition 3.1Pis a generalized trigonometric polynomial, then the function f is also called uniform limit power andᐁᏸᏼ(R⁺) is de- noted the set of all such functions. It is not difficult to show thatᐁᏸᏼ(R⁺) is a Banach space and (3.10)–(3.12) are valid. For the question if an f ∈ᐁᏸᏼ(R⁺) can be approx- imated by the Bochner-Fejer polynomials, as well as how to construct the polynomials, we refer the reader to [40,41] for details. Also

ᐁᏸᏼR⁺

=spanᐁᏸᏼα R⁺

: 0< α <∞

⊂CR⁺

∩H2, (3.29) where the closure is taken inC(R⁺). One can see how hugeᐁᏸᏼ(R⁺) is by comparing withᏭᏼ(R⁺).

Remark 3.7. As chirps, some existing results enable us to analyze and reconstruct f ∈ ᐁᏸᏼ(R⁺). For example, by [30, Theorem 2.1] a windowed Fourier transform off exists, by Theorem 2.2 of the same paper the transform satisfies some Parseval’s relation, and by Theorem 2.4 of that paper again a generalized frame exists.

One more remark is needed to end the section.

Remark 3.8. (1) The conclusion in the paragraph beforeRemark 2.4is mentioned in [39, Section 2] without proof. Here we not only prove it in details, but we also distinguish the limits between the caseα≥1 and the caseα <1 in Theorems2.1and2.3, respectively. (2) Also, the results inSection 3are presented in [39, Section 2] in an abstract-like form. To convince the reader the correctness of these results, we present and prove them in details here.

4. Limit power type functions

In this section, we will define and investigate three types of limit power function which are corresponding to the three types of well-known almost periodic functions (e.g., see [1,2,17,21,31,33,34,38]).

Let

C0

R⁺

=

ϕ∈CR⁺ : lim

t→∞ϕ(t)=0, ᏼᏭᏼ0

R⁺

=

ϕ∈CR⁺

:M|ϕ|

=0.

(4.1)

Definition 4.1. Let f ∈C(R⁺). A function f is called asymptotic limit power if f(t)=g(t) +ϕ(t) t∈R⁺

, (4.2)

whereg∈ᐁᏸᏼα(R⁺) andϕ∈C0(R⁺). Denote byᏭᏸᏼα(R⁺) all such functions.

By (3.14), we have

f(s)=g(s) +ϕ(s). (4.3)

It is easy to check thatϕ(t) is inC0(R⁺) if and only ifϕ(s) is in C0(R⁺). Sinceg(s)∈ Ꮽᏼ(R⁺), one gets that f(s)∈ᏭᏭᏼ(R⁺), the space of asymptotically almost periodic functions. Therefore, (3.14) is also a correspondence between ᏭᏭᏼ(R⁺) and Ꮽᏸᏼα(R⁺). Note the characterization of ᏭᏭᏼ(R⁺) (e.g., see [38, Theorem 1.2.11]), we have the following corresponding characterization.

Theorem 4.2. Let f ∈C(R⁺). Then the following statements are equivalent:

(1) f ∈Ꮽᏸᏼα(R⁺);

(2) the set{f[(t+x)^1/α] :x∈R⁺}is relatively compact inC(R⁺);

(3) for any>0 there exists a bounded closed intervalC=[0,a] andl >0 such that any intervalI⊂R⁺of lengthlhas a numberτ∈Iwith the property

f(t+τ)^1/α−f(t)^1/α<

t,t+τ∈R⁺\C. (4.4) If we only require the set inTheorem 4.2(2) to be weakly compact, then we get the following concept.

Definition 4.3. An f ∈C(R⁺) is called weak limit power if the set inTheorem 4.2(2) is weakly compact inC(R⁺). Denote byᐃᏸᏼα(R⁺) of all such functions.

To get the decomposition of a function f ∈ᐃᏸᏼα(R⁺), we introduce the following set:

ᐃᏸᏼ0(b)R⁺=

ϕ∈ᐃᏸᏼα R⁺

: 0∈

ϕ(t+x)^1/α:x∈R⁺

, (4.5)

where the closure is taken under weak topology inC(R⁺). The following result is a corre- spondence inᐃᏸᏼα(R⁺) to that inᐃᏭᏼ(R⁺).

Theorem 4.4. Let f ∈C(R⁺). Then the following statements are equivalent:

(1) f ∈ᐃᏸᏼα(R⁺);

(2) f =g+ϕ, whereg∈ᐁᏸᏼα(R⁺) andϕ∈ᐃᏸᏼ0(R⁺).

Finally we give the following concept.

Definition 4.5. An f ∈C(R⁺) is called pseudolimit power if f has the form f =g+ϕ, whereg∈ᐁᏸᏼα(R⁺) andϕ∈ᏼᏭᏼ0(R⁺). Denote byᏼᏸᏼα(R⁺) all such functions.

Let f ∈ᏼᏸᏼα(R⁺). Since f(t)e^−iλtα=g(t)e^−iλtα+ϕ(t)e^−iλtαand

Tlim→∞1/T _T

0 f(t)e^−iλtαdt=Mge^−iλtα (4.6) for allλ∈R,Theorem 3.5implies thatᏼᏸᏼα(R⁺) is a direct sum ofᐁᏸᏼα(R⁺) and ᏼᏭᏼ0(R⁺). Since the rangesR f andRfare the same (soRgandRg,RϕandRϕ) and Rf ⊃Rg[38, Lemma 1.5.2], we haveR f ⊃Rg. By this, we can show the following theo- rem.

Theorem 4.6. ᏼᏸᏼα(R⁺) is a Banach space.

Proof. Let {fn} ⊂ᏼᏸᏼα(R⁺) be Cauchy. Since R fn⊃Rgn, {gn} is Cauchy too. Note ᐁᏸᏼα(R⁺) is closed inC(_R⁺), there existsg_∈ᐁᏸᏼα(R⁺) such thatg_n−g _→0 as n→ ∞. Since{f_n−g_n}is also Cauchy andᏼᏭᏼ0(R⁺) is closed inC(R⁺), there exists ϕ∈ᏼᏭᏼ0(R⁺) such thatϕn−ϕ →0 asn→ ∞. Let f =g+ϕ. Then f ∈ᏼᏸᏼα(R⁺)

andfn−f →0 asn→ ∞. The proof is complete.

Since

ᏭᏼR⁺

⊂ᏭᏭᏼR⁺

⊂ᐃᏭᏼR⁺

⊂ᏼᏭᏼ(R⁺), (4.7)

one has the following inclusion relationship:

ᐁᏸᏼα

R⁺

⊂Ꮽᏸᏼα

R⁺

⊂ᐃᏸᏼα

R⁺

⊂ᏼᏸᏼα

R⁺

. (4.8)

Acknowledgments

The authors would like to thank the referees for the valuable comments. The research is supported by NSF of China (no.10671046).