DOI 10.1007/s10013-015-0146-yA Study of the Sequence of Norm of Derivatives or Primitives of Functions Depending on Their Beurling Spectrum Ha Huy Bang 1 · Vu Nhat Huy 2 Received: 22 Oct
Trang 1DOI 10.1007/s10013-015-0146-y
A Study of the Sequence of Norm of Derivatives
(or Primitives) of Functions Depending on Their
Beurling Spectrum
Ha Huy Bang 1 · Vu Nhat Huy 2
Received: 22 October 2014 / Accepted: 8 December 2014
© Vietnam Academy of Science and Technology (VAST) and Springer Science+Business Media Singapore 2015
Abstract In this paper, we characterize the behavior of the sequence of norm of derivatives
(or primitives) of functions by their Beurling spectrum in Banach spaces The Bernstein inequality for Banach spaces is also obtained
Keywords Banach Spaces· Beurling spectrum
Mathematics Subject Classification (2010) 26D10· 46E30
1 Introduction
The following result was proved in [9]: Let 1 ≤ p ≤ ∞ and f (m) ∈ L p ( R), m =
0, 1, 2, Then there always exists the following limit
lim
m→∞f (m)1/m
p
m→∞f (m)1/m
p = σ f = sup{|ξ| : ξ ∈ supp ˆ f },
where ˆ f is the Fourier transform of f
The result shows that the behavior of the sequencef (m)1/m
p is wholly characterized by
the spectrum of f , and it reminds us of the following well-known spectral radius formula in
Ha Huy Bang
hhbang@math.ac.vn
Vu Nhat Huy
nhat huy85@yahoo.com
1 Institute of Mathematics, Vietnamese Academy of Science and Technology,
18 Hoang Quoc Viet Street, Cau Giay, Hanoi, Vietnam
2 Department of Mathematics, College of Science, Vietnam National University,
334 Nguyen Trai Street, Thanh Xuan, Hanoi, Vietnam
Trang 2the Theory of Banach algebras: Let A be a unital Banach algebra and x ∈ A Then the limit
limm→∞x m1/malways exists and satisfies
lim
m→∞x m1/m = r A (x),
where r A (x) = sup{|λ| : λ ∈ σ A (x) } is the spectral radius of x, σ A (x) is the spectrum of x which is the set of all λ ∈ C such that x − λ1 is not invertible in A and 1 is the unit in A (see
[18,22]) That is why σ f is called the local spectral radius of the differential operator D,
and this result was studied and developed by many authors (see [1 12,15–20,26–29]) It is natural to ask what will happen when we replace derivatives by integrals To this question,
Tuan answered for p= 2 in [29] and we answered in [13] for 1≤ p ≤ ∞, and in [14] for
a more general case (but with the assumption that the spectrum must be compact) In this paper, we extend the results in [9,13,14,29] to Banach spaces, we also extend the results
in [17] to a general case and the Bernstein inequality is also obtained
2 Main Results
Let f ∈ L1( R) and ˆ f = Ff be its Fourier transform
ˆ
f (ξ )= √1
2π
+∞
−∞ e
−ixξ f (x)dx,
and ˇf = F−1fbe its inverse Fourier transform
ˇ
f (ξ )= √1
2π
+∞
−∞e
ixξ f (x)dx.
Let ( X, · X) denote a complex Banach space and let BC( R → X) denote the set of all X valued bounded continuous functions f : R → X For a given function f ∈ BC(R → X),
we definef ∞ = sup{f (t)X: t ∈ R} Then (BC(R → X), · ∞)itself is a Banach
space We define the derivative Df of f ∈ BC(R → X), as usual,
Df (s)= lim
δ→0
f (s + δ) − f (s)
Note that for every λ ∈ C \ iR the operator λ − D is invertible Hence, the spectrum of the differential operator is iR
Clearly, the inverse of λ − D is a bounded operator on (BC(R → X), · ∞) Moreover,
+∞
−∞ ϕ(t)(λ − D) −n f (t)dt =
+∞
−∞
(λ − D) −n ϕ(t)
f (t)dt
for any f ∈ BC(R → X) The convolution ϕ ∗ f of f with a Schwartz function is defined
by letting
ϕ ∗ f (s) =
+∞
−∞ ϕ(s − t)f (t)dt.
Proposition 1 (Young inequality for Banach spaces) Let f ∈ BC(R → X), ϕ ∈ S(R) Then ϕ ∗ f ∈ BC(R → X) and
ϕ ∗ f ∞≤ f ∞ϕ1.
The Beurling spectrum Spec(f ) of a function f ∈ BC(R → X) is defined by
Spec(f ) = {ξ ∈ R : ∀ > 0, ∃ϕ ∈ S(R) : supp ϕ ⊂ (ξ − , ξ + )), ϕ ∗ f ≡ 0}.
Trang 3The Beurling spectral radius ρ(f ) of f is defined by
ρ(f ) = sup{|ξ| : ξ ∈ Spec(f )}.
Note that Spec(f ) is always a closed subset ofR Moreover,
Proposition 2 We have the following properties of Beurling spectrum:
– ifϕ = 0 on Spec(f ) then ϕ ∗ f = 0;
– ifϕ = 1/√2π on Spec(f ) then ϕ ∗ f = f ;
– Spec((λ − D)−1f ) = Spec(f ) for every f ∈ BC(R → X) and λ ∈ C \ iR.
See [21] for more details Now, we defineD = −iD and the differential operator P (D)
is obtained from P (x) by substituting x → −iD (for P (x) =n
k=0a k x k then P ( D)f =
n
k=0a k D k f)
Theorem 1 Let P (x) be a polynomial and f ∈ BC(R → X), f ≡ 0 Assume that Spec(f )
is a compact set Then
lim
m→∞P m (D)f 1/m
∞ = sup
ξ ∈Spec(f ) |P (ξ)|. (1)
Proof Let us first show that
lim
m→∞P m ( D)f 1/m
∞ ≥ sup
ξ ∈Spec(f ) |P (ξ)|. (2)
Indeed, let σ ∈ Spec(f ) satisfy P (σ ) = 0 For a small enough positive number , we have
P (x) = 0 for all x ∈ B(σ, ) From the definition of Beurling spectrum, there exists a function ϕ ∈ C∞( R), supp ϕ ⊂ B(σ, ) such that ϕq∗f ≡ 0 We put
ϕ m = F−1(ϕ(x)/P m (x)).
Then ϕ mis well defined and it follows from
D k ϕ m = F−1(x kϕ m (x)) = F−1(x k ϕ(x)/P m (x)) ∀k ∈ Z+
that
P m ( D)ϕ m = F−1(P m (x)ϕ(x)/P m (x))=ϕ.q (3)
Since ϕ m ∗ (D k f ) = (D k ϕ m ) ∗ f for all k ∈ Z+, we have
ϕ m ∗ (P m ( D)f ) = (P m ( D)ϕ m ) ∗ f.
Thus, by (3), we get
0 <
q
ϕ ∗ f∞=ϕ m ∗ P m ( D)f∞.
Applying Young inequality for Banach spaces, we obtain
0 <
q
ϕ ∗ f
∞=ϕ m ∗ P m ( D)f
∞≤ P m ( D)f ∞ϕ m1
and then
lim
m→∞P m (D)f 1/m
∞ ≥ lim
m→∞ϕ m1/m
1
−1
According to [14] and ϕ(x) = 0 ∀x /∈ B(σ, ), we conclude that
lim
m→∞ϕ m1/m
1 ≤ sup
ξ ∈B(σ,) |1/P (ξ)|.
Trang 4Therefore, by (4), we obtain
lim
m→∞P m ( D)f 1/m
∞ ≥ inf
ξ ∈B(σ,) |P (ξ)|.
Letting → 0, we get
lim
m→∞P m ( D)f 1/m
Because (5) holds for any σ ∈ Spec(f ), P (σ ) = 0, we arrive at (2)
Finally, we show
lim
m→∞P m ( D)f 1/m
∞ ≤ sup
where K = Spec(f ) Indeed, for > 0, we can choose a function h ∈ C∞( R) satisfying the conditions h(ξ ) = 1/√2π for ξ ∈ K /2and h(ξ ) = 0 for ξ ∈ K Using Proposition 2,
we have f = ˇh ∗ f Clearly, D k f = D k ( ˇ h ∗ f ) = (D k ˇh) ∗ f for all k ∈ Z+ That gives
P m ( D)f = (P m ( D) ˇh) ∗ f.
Then, since P m ( D) ˇh = F−1(h(x)P m (x)), we get
P m ( D)f = F−1(h(x)P m (x)) ∗ f.
Applying Young inequality, we obtain
P m ( D)f ∞≤ f ∞F−1(h(ξ )P m (ξ ))1.
Hence,
lim
m→∞P m ( D)f 1/m
∞ ≤ limm→∞F (h(ξ)P m (ξ ))1/m
We have known in [11] that
lim
m→∞F (h(ξ)P m (ξ ))1/m
1 = limm→∞g m1/m
1 ≤ sup
ξ ∈K
|P (ξ)|. (8) Combining (7) and (8), we get
lim
m→∞P m (D)f 1/m
∞ ≤ sup
ξ ∈K
|P (ξ)|
and then (6) by letting → 0 Combing (2) and (6), we arrive at (1)
The proof is complete
We define I λ = (λ − D)−1, where λ ∈ C \ iR The integral operator P (I λ )is obtained
from P (x) by substituting x → I λ We have the following theorem:
Theorem 2 Let P (x) be a polynomial, f ∈ BC(R → X), f ≡ 0 and λ ∈ C \ iR Assume
that Spec(f ) is a compact set Then there always exists the following limit
lim
m→∞P m (I λ )f1/m
∞
and
lim
m→∞P m (I λ )f1/m
∞ = sup
ξ ∈Spec(f ) |P (1/(λ − iξ))|.
Proof Now we show that
lim
m→∞P m (I λ )f1/m
∞ ≥ sup
ξ ∈Spec(f ) |P (1/(λ − iξ))|. (9)
Trang 5Indeed, let σ be an arbitrary element in Spec(f ) satisfying P (1/(λ − iσ )) = 0 For a small enough positive number , we have P m ( 1/(λ − iξ)) = 0 for all ξ ∈ B(σ, ) From the definition of Beurling spectrum, there exists ϕ ∈ C∞
0 ( R), supp ϕ ⊂ B(σ, ) such that
q
ϕ ∗ f ≡ 0 Put
ϕ m = F−1 ϕ(ξ )/P m ( 1/(λ − iξ)).
Then, ϕ m is well defined, ϕ m∈S(R) and P m (I λ )ϕ m =ϕq Clearly, ϕ m ∗(I k
λ f ) = (I k
λ ϕ m ) ∗f for all k∈ Z+ That gives ϕ m ∗ (P m (I λ )f ) = (P m (I λ )ϕ m ) ∗ f and then ϕ m ∗ P m (I λ )f = q
ϕ ∗ f So, applying Young inequality, we obtain
0 <
q
ϕ ∗ f
∞=ϕ m ∗ P m (I λ )f
∞≤P m (I λ )f
∞ϕ m1. (10) Clearly,
sup
x∈R|(1 + x2)ϕ m (x)|
≤√1
2π
ξ ∈B(σ,)
D2 ϕ(ξ )/P m ( 1/(λ − iξ)) + ϕ(ξ )/P m ( 1/(λ − iξ)) dξ.
So
sup
x∈R
(1 + x2)ϕ m (x) ≤ C
1m2 sup
ξ ∈B(σ,)
1/P m+2( 1/(λ − iξ)) , (11) where
C1 := √2
2π
ξ ∈K
|(D2ϕ)(ξ )P2( 1/(λ − iξ))| + 2|(Dϕ)(ξ)D(P (1/(λ − iξ)))
×P (1/(λ − iξ))| + |ϕ(ξ)D2(P ( 1/(λ − iξ)))P (1/(λ − iξ))
+|ϕ(ξ)(D(P (1/(λ − iξ))))2| + |ϕ(ξ)P2( 1/(λ − iξ))|dξ.
Further, we have
R|ϕ m (x) |dx ≤ π sup
x∈R|(1 + x2)ϕ m (x) |. (12) Combining (11)–(12), we obtain
lim
m→∞ϕ m1/m
1 ≤ sup
ξ ∈B(σ,) |1/P (1/(λ − iξ))|. (13) Relations (10) and (13) imply
lim
m→∞P m (I λ )f1/m
∞ ≥ inf
ξ ∈B(σ,) |P (1/(λ − iξ))|.
Letting → 0, we get
lim
m→∞P m (I λ )f1/m
∞ ≥ |P (1/(λ − iσ ))|. (14) Because (14) holds for any σ ∈ Spec(f ), P (1/(λ − iσ )) = 0, we confirm (9)
Put K = Spec(f ) Now we claim that
lim
m→∞P m (I λ )f1/m
∞ ≤ sup
ξ ∈K |P (1/(λ − iξ))|. (15)
Indeed, for any > 0, there exists h ∈ C∞( R) such that h(x) = 1/√2π if x ∈ K /2 and
h(x) = 0 if x /∈ K Note that λ − iξ = 0 for all ξ ∈ K , so the following function is well defined:
H m = F−1(h(ξ )P m ( 1/(λ − iξ))).
Trang 6Using Proposition 2, we have f = ˇh∗f and observe from I k
λ f = (I k
λ ˇh) ∗ f ∀k ∈ Z+
that
P m (I λ )f = (P m (I λ ) ˇ h) ∗ f. (16)
From I λ k ˇh = F−1(h(ξ )/(λ − iξ) k ) , we conclude P m (I λ ) ˇ h = F−1(h(ξ )P m ( 1/(λ − iξ))) =
H m Therefore, by using (16), we have P m (I λ )f = H m ∗ f Hence, it follows from
Proposition 1 that
P m (I λ )f∞≤ f ∞H m1.
So
lim
m→∞P m (I λ )f1/m
∞ ≤ limm→∞H m1/m
Clearly,
√
2π sup
x∈R|(1 + x2)H m (x)|
≤
ξ ∈K
|D2(h(ξ )P m ( 1/(λ − iξ)))| + |h(ξ)P m ( 1/(λ − iξ))| dξ
≤
ξ ∈K
|D2h(ξ )P m ( 1/(λ −iξ))|+2m|Dh(ξ)D(P (1/(λ − iξ)))P m−1( 1/(λ − iξ))| +m(m − 1)|h(ξ)(D(P (1/(λ − iξ))))2P m−2( 1/(λ − iξ))|
+m|h(ξ)D2(P ( 1/(λ − iξ)))P m−1( 1/(λ − iξ))| + |h(ξ)P m ( 1/(λ − iξ))|dξ.
So
sup
x∈R|(1 + x2)H m (x) | ≤ C2m2 sup
ξ ∈K
|P m−2( 1/(λ − iξ))|, (18)
where C2does not depend on m Further, we have
R|H m (x) |dx ≤ π sup
x∈R|(1 + x2)H m (x) |. (19) Combining (18)–(19), we obtain
lim
m→∞H m1/m
1 ≤ sup
ξ ∈K
|P (1/(λ − iξ))|. (20)
We combine inequality (17) with inequality (20) to conclude that
lim
m→∞P m (I λ )f1/m
∞ ≤ sup
ξ ∈K
|P (1/(λ − iξ))|.
Letting → 0, we confirm (15) Combining (15) and (9), we have
lim
m→∞P m (I λ )f1/m
∞ = sup
ξ ∈Spec(f ) |P (1/(λ − iξ))|.
The proof is complete
Based on the results about Bernstein inequality in L p ( R)-spaces (see [23–25]), we obtain Bernstein inequality for Banach spaces in the following theorem:
Theorem 3 Let σ > 0, f ∈ BC(R → X) and Spec(f ) ⊂ [−σ, σ ] Then there exists a
constant C not depending on f, m, σ such that
D m f∞≤ Cσ m f ∞, m = 1, 2, (21)
Trang 7Proof Let us first prove (21) for the case σ = 1 Indeed, put K := [−1, 1] and the function
is defined as follows
=
C1e 1/(ξ2−1) if|ξ| < 1,
0 if|ξ| ≥ 1, where C1is chosen such that
R = 1 We define the sequence of functions φ m (ξ ) m≥1
via the formula
φ m (ξ ) = (1 K 3/(4m) 1/(4m) )(ξ ),
where
1/(4m) (ξ )
1/(4m) (ξ ) = 0 for all ξ ∈ [−1/(4m), 1/(4m)] andR 1/(4m) (ξ )dξ = 1 Hence, for
any m ≥ 1, we have φ m (ξ ) ∈ C∞
0 ( R), and φ m (ξ ) = 1 ∀ξ ∈ K 1/(2m) , φ m (ξ ) = 0 ∀ξ /∈
K 1/m So, it follows from Spec(f ) ⊂ K that f = (2π) −1/2 F (φ
m ) ∗ f Hence,
D m f∞= (2π) −1/2 F (φ m (ξ )ξ m ) ∗ f ∞.
Therefore, applying Young inequality (Proposition 1), we obtain the estimate
D m f∞= (2π) −1/2 F (φ m (ξ )ξ m ) ∗ f ∞≤ (2π) −1/2 f ∞F (φ m (ξ )ξ m )1. (22) Define
k m:= 1 + 1
m , g m (ξ ) = φ m
ξ
k m
, m (ξ ) = φ m (ξ ) − g m (ξ ).
Then
(F (g m (ξ )ξ m ))(x) = (k m ) m
F
φ m
ξ
k m
ξ
k m
m
(x)
= (k m ) m+1(F (φ
m (ξ )ξ m ))(k m x).
So
F (g m (ξ )ξ m )
1= (k m ) mF (φ m (ξ )ξ m )
1.
Then, it follows from (k m ) m = (1 + 1
m ) m ≥ 2 that
F (g m (ξ )ξ m )
1≥ 2F (φ m (ξ )ξ m )
1.
Therefore, since m (ξ ) = φ m (ξ ) − g m (ξ )we get
F ( m (ξ )ξ m )
1≥F (g m (ξ )ξ m )
1−F (φ m (ξ )ξ m )
1≥F (φ m (ξ )ξ m )
1. (23) From (22)–(23), we obtain
D m f∞≤ (2π) −1/2 f ∞F ( m (ξ )ξ m )1. (24)
Now, we show the existence of a constant C such that
F ( m (ξ )ξ m )1≤ (2π) 1/2 C
for all m ≥ 1 Indeed, we put C2 (j )1, j 1/(4m) (x)
(j )
1/(4m) (x) = (4m) j+1 (j ) ( 4mx) and then
(j )
1/(4m)1= (4m) j (j )1≤ C2( 4m) j ∀j ≤ 3.
Therefore,
φ (j )
m
∞=(1
K 3/(4m) 1/(4m) (j ) )
∞≤ (j )
1/(4m)
1≤ (4m) j C2 ∀j ≤ 3. (25)
Trang 8Note that φ m (ξ ) = 1 ∀ξ ∈ (−1 − (1/2m), 1 + (1/2m)), and φ m (ξ ) = 0 ∀ξ ∈ (−∞, −1 −
( 1/m)) ∪(1+(1/m), +∞) So, if |ξ| < 1 then |ξ/k m | < |ξ| < 1 and φ m (ξ ) = φ m (ξ /k m )=
1, i.e., m (ξ )= 0
Further, if|ξ| > 1+(3/m) then |ξ| > |ξ/k m | > 1+(1/m) and then φ m (ξ ) = φ m (ξ /k m )=
0, i.e., m (ξ )= 0
So we have
supp m ⊂ [1, 1 + (3/m)] ∪ [−1 − (3/m), −1]. (26)
Now, if ξ ∈ [1, 1 + (3/m)] ∪ [−1 − (3/m), −1] then
ξ − k ξ m = (k m − 1)ξ
k m
= mk ξ m ≤ m4. (27) From (25) and (27), we get the following estimate for ξ ∈ [1, 1+(3/m)]∪[−1−(3/m), −1]
| m (ξ ) | = |φ m (ξ ) − g m (ξ )| =
φ m (ξ ) − φ
ξ
k m
≤
ξ − ξ
k m
·φ
m
∞≤ 4
m (ξ ) = φ
m (ξ ) − g m(ξ ) = φ
m (ξ )−
φ m
ξ
k m
=
φ
m (ξ )− 1
k m
φ
m
ξ
k m
≤ φ
m (ξ ) − φ m
ξ
k m
+ 1− 1
k m
φ
m
ξ
k m
≤
ξ − ξ
k m
·φ
m
∞+
1 − 1
k m
·φ
m
∞
≤ 4
m ( 4m)
2C2+
1 −k1m 4mC2≤ 68mC2, (29) and
m (ξ ) = φ
m (ξ ) − gm (ξ ) = φ
m (ξ )−
φ m
ξ
k m
=
φ
m (ξ )− 1
k2
m
φ
m
ξ
k m
≤ φ
m (ξ ) − φ m
ξ
k2
m
+ 1− 1
k2
m
φ
m
ξ
k m
≤
ξ − ξ
k2
m
· φ
m∞+
1 − 1
k2
m
·φ
m
∞
≤ 8
m ( 4m)
3C2+
1 −k12
m
(4m)2C2≤ 528m2C2. (30)
Put H m (x) = (F ( m (ξ )ξ m ))(x) Then
H m (x)= √1
2π
Re
−ixξ
m (ξ )ξ m dξ.
Therefore, from (26), we obtain
sup
x∈R|H m (x)| ≤√1
2π
R
m (ξ )ξ m dξ = 1
√
2π
1≤|ξ|≤1+ 3
m
m (ξ )ξ m dξ and it follows from (25) that
sup
x∈R|H m (x)| ≤ 6
m√
2π ξsup∈R| m (ξ )|
1+ 3
m
m
≤ 96e3C2
m√
Trang 9We also have
x2H m (x) = 1
√
2π
Re −ixξ (
m (ξ )m(m − 1)ξ m−2+
m (ξ ) 2mξ m−1+
m (ξ )ξ m )dξ
≤ √1
2π
R
e −ixξ ( m (ξ )m(m − 1)ξ m−2+ m (ξ ) 2mξ m−1+ m ξ m)dξ dξ and then
sup
x∈R
x2H m (x)
≤ √1
2π
1≤|ξ|≤1+ 3
m
| m (ξ )m(m − 1)ξ m−2+ (ξ ) 2mξ m−1+ (ξ )ξ m |dξ
m√
2π
sup
ξ∈R|(ξ)|m(m − 1)
1+ 3
m
m−2
+ sup
ξ∈R|(ξ ) |2m
1+ 3
m
m−1
+ sup
ξ∈R|(ξ )|
1+ 3
m
m
.
So, from (28)–(30), we get
sup
x∈R
x2H m (x) ≤ 8
m√
2π 16C2m(m − 1)e3+ 68mC22me3+ 528m2C2e3
≤ 5280e3C2m
√
We have
H m1 =
|x|≤m |H m (x) |dx +
|x|≥m |H m (x) |dx
≤ sup
x∈R|H m (x)|
|x|≤m 1dx+ sup
x∈R|x2H m (x)|
|x|≥m
x12
dx
= 2m sup
x∈R|H m (x)| + 2
m xsup∈R|x2H m (x) |. (33) From (31)–(33), we obtain
H m1≤ 2m
96e3C2
m√
2π
+ 2
m
5280e3C2m
√
2π .
Therefore, from (24), we can choose a constant C such that
D m f∞≤ Cf ∞.
So, (21) has been proved with σ= 1
Next, we prove (21) for any σ > 0 Indeed, put
g(x) = fx
σ
.
Since Spec(f ) ⊂ [−σ, σ ], Spec(g) ⊂ [−1, 1] Therefore
Since g(x) = f ( x
σ ), we have
g∞= f ∞, D m g∞= σ −m D m f∞.
Trang 10Hence, it follows from (34) that
σ −m D m f∞≤ Cf ∞.
We conclude that
D m f∞≤ Cσ m f ∞.
The proof is complete
The minimum of C satisfying the above inequalities is called the Bernstein constant.
Theorem 4 Let f ∈ BC(R → X), σ ∈ R+ Assume that Spec(f ) ⊂ [−σ, σ ] Then
lim
m→∞σ −m D m f∞≤ C lim
λ→1 −f (·) − f (λ·)∞, (35)
where C is the Bernstein constant.
Proof Fix λ ∈ (0, 1) Put f λ (x) = f (λx) Then Spec(f λ ) = λSpec(f ) Then it follows from Spec(f ) ⊂ [−σ, σ ] that
Applying the Bernstein inequality for Beurling spectrum, we get
D m f λ∞≤ C(λσ ) m f λ∞ (37)
for all m ∈ N Using (36) and Spec(f ) ⊂ [−σ, σ ], we obtain Spec(f − f λ ) ⊂ [−σ, σ ].
Then it follows from the Bernstein inequality for Beurling spectrum that
D m (f − f λ )∞≤ Cσ m f − f λ∞. (38) Therefore, using (37) and (38), we have
σ −m D m f∞ ≤ σ −m D m (f − f λ )∞+ σ −m D m f λ∞
≤ Cf − f λ∞+ Cλ m f λ∞
= Cf − f λ∞+ Cλ m f ∞
for all m∈ N Hence,
lim
m→∞σ −m D m f∞≤ Cf − f λ∞.
Letting λ→ 1−, we get lim
m→∞σ −m D m f∞≤ C lim λ→1 −f (·)−f (λ·)∞ The proof
is complete
Corollary 1 Let f ∈ BC(R → X), σ ∈ R+ Assume that Spec(f ) ⊂ [−σ, σ ] Then the
sequence {σ −m D m f∞} is bounded Moreover, if lim λ→1 −f (·) − f (λ·)∞= 0 then
lim
m→∞σ
−m D m f∞= 0.
Acknowledgments This research is funded by Vietnam National Foundation for Science and Technol-ogy Development (NAFOSTED) under grant number 101.01-2011.32 The authors would like to thank the referees for the exact corrections.
References
1 Abreu, L.D.: Real Paley–Wiener theorems for the Koornwinder–Swarttouw q-Hankel transform J Math.
Anal Appl 334, 223–231 (2007)
...≡ 0}. Trang 3The Beurling spectral radius ρ(f ) of f is defined by
ρ(f... class="page_container" data-page="4">
Therefore, by (4), we obtain
lim
m→∞P m ( D)f 1/m
∞... chosen such that
R = We define the sequence of functions φ m (ξ ) m≥1
via the formula
φ m (ξ )