HAL Id: hal-02099190
https://hal.univ-antilles.fr/hal-02099190
Preprint submitted on 14 Apr 2019
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
FUNCTIONAL CALCULUS IN THE ALGEBRA OF GENERALIZED HYPERFUNCTIONS ON THE
CIRCLE AND APPLICATIONS
V. Valmorin
To cite this version:
V. Valmorin. FUNCTIONAL CALCULUS IN THE ALGEBRA OF GENERALIZED HYPERFUNC-
TIONS ON THE CIRCLE AND APPLICATIONS. 2019. �hal-02099190�
GENERALIZED HYPERFUNCTIONS ON THE CIRCLE AND APPLICATIONS
V. VALMORIN
Abstract. This paper deals with a functional calculus in the al- gebra H pTq of generalized hyperfunctions on the circle. This is done introducing an inductive family of complete ultrametric sub- algebras. Power series expansions of classical functions such as the exponential, logarithm or power ones are considered. As an appli- cation, a nonlinear Cauchy problem involving fractional powers of generalized hyperfunctions is studied.
11. Introduction
This paper aims to provide the algebra H pTq of generalized periodic hy- perfunctions with a functional calculus based on elementary functions but with high nonlinearities. This becomes essential when dealing with nonlinear differential or functional equations. The algebra H pTq was introduced in [18] and its ultrametric topology in [17]. Earlier a first version was given in [16] involving real 2π-periodic smooth functions.
Later on, using the framework of sequence spaces, see [5, 6, 7], the au- thor and his collaborators have given a general topological description of various algebras of generalized functions including H pTq . This des- crition involves projective and inductive limits of locally convex spaces.
It is well-known that contrary to projective limits inductive limits have a bad inheritance of completeness. Moreover it has never been proved that H pTq was a complete space or not. Then to overcome such a situ- ation, we introduce an inductive family p H
rpTqq
r¡1of complete ultra- metric differential algebras in such a way that H pTq ind lim
rÑ1H
rpTq in a set theoritical sense. Therefore it is shown that the induced induc- tive limit topology on H pTq is finer that its original one. Recall that the initial ultrametric topology of H pTq is given by ω p f, g q ν p f g q where ν is the so-called indicator introduced in [17]. We point out that ν p λ q 1 for all nonzero complex number λ. It follows that p H pTq , ω q is
1
2010 Mathematics Subject Classification : 30H50, 46F15, 46F30, 54E50.
Keywords: Generalized periodic hyperfunctions, Complete algebras, Functional cal- culus, Cauchy problem.
1
not a classical topological algebra over the field C of complex numbers since the multiplication by a nonzero complex number is not contin- uous. Nevertheless ν induces a complete ultrametric structure on the associated algebra C of generalized complex numbers over which H pTq is a classical topological algebra but it should be noticed that C is not a field nor a domain. In the same way the topology of each algebra H
rpTq is defined by an indicator ν
r. Endowed with the ultrametric ω
rsuch that ω p f, g q ν
rp f g q , H
rpTq is a complete algebras.
For the basic theory of Colombeau generalized functions, we refer to [3, 4, 9, 10, 13, 14]. Topological results on generalized functions can be found in [7, 13]. For the theory of periodic hyperfunctions we refer to [1, 2, 11, 12]. We notice that a product of hyperfunctions on the circle is defined in [8] in a more classical setting. This is done using conditions on Fourier coefficients. In the setting of Colombeau algebras, the first work on product of hyperfunctions has been done in [15].
The paper is organized as follows. Section 2 presents some preliminaries on the algebra H pTq which are useful for the sequel. References for this section are mainly [12, 17, 18]. In Section 3 we define and study the algebras H
rpTq , r ¡ 1. They are proved to be complete and the same is done for the algebra C of generalized numbers endowed with the ultrametric ω. In Section 4 we give necessary of sufficient conditions for the existence of log p h q , exp p h q or h
s, s P R where h P H pTq . Section 4 is concerned with the resolution of a nonlinear Cauchy problem in H pTq where the introduced functional calculus is used.
2. Preliminaries
2.1. The algebra of generalized hyperfunctions on the circle.
For this section we refer mainly to [12, 17, 18]. For r ¡ 1 let C
rt z P C , 1 { r | z | r u and } f }
rsup
zPCr
| f p z q|
for every bounded continuous function f defined in C
r. We denote by O
rthe Banach space of bounded holomorphic functions in C
ren- dowed with the norm } }
r. Then, the topological space of real analytic functions on the unit circle T is
A pTq ind lim
rÑ1O
r.
If X pTq is the set of sequences of functions p f
nq
nwith f
nP A pTq , we denote by X
epTq the subset of X pTq whose elements p f
nq
nsatisfy:
D a ¡ 0, D η P N , D r ¡ 1, f
nP O
r, } f
n}
r¤ a
n, n ¡ η.
We denote by N
epTq the subset of X
epTq constituted of elements p f
nq
nsatisfying:
@ b P p 0, 1 q , D η P N , D r ¡ 1, f
nP O
r, } f
n}
r¤ b
n, n ¡ η.
Clearly X
epTq is an algebra for usual termwise operations and N
epTq is an ideal of X
epTq .
Proposition 2.1. [18, Proposition 3.1] If p f
nq
nP X pTq , then:
(i) p f
nq
nP X
epTq if and only if
D a ¡ 0, D η P N , D r ¡ 1, | p f
np k q| ¤ a
nr
|k|, n ¡ η, k P Z . (ii) p f
nq
nP N
epTq if and only if
@ b P p 0, 1 q , D η P N , D r ¡ 1, | p f
np k q| ¤ b
nr
|k|, n ¡ η, k P Z . The algebra of generalized hyperfunctions on T is the factor algebra
H pTq X
epTq{ N
epTq
The class of p f
nq
nin H pTq will be denoted by cl p f
nq .
Embedding of B pTq and A pTq in H pTq . The space B pTq of periodic hyperfunctions is the topological dual of A pTq . For n P N we set
ϕ
np z q ¸
|k|¤n
z
k.
Then we have ϕ
nϕ
nϕ
nand lim
nÑ8ϕ
nδ in B pTq where δ is the periodic Dirac distribution. If H P B pTq , then p H ϕ
nqp z q
°
|k|¤n
H p p k q z
kand lim
nÑ8H ϕ
nH in B pTq . Moreover, the maps i : B pTq Ñ X
epTq defined by i p H q p H ϕ
nq
nand i
0: A pTq Ñ X
epTq defined by i
0p f q p f
nq
nwith f
nf , satisfy the following:
(i) i and i
0are linear embeddings;
(ii) i
0is a morphism of algebras.
We denote by B
θbe the differential operator defined for f P O
r, by B
θf iz df
dz where z P C
r. It follows that for every k P Z ,
pB z
θf qp k q ik f ˆ p k q .
Henceforth, H pTq is endowed with two structures of differential algebra defined by
df dz cl
df
ndz
and B
θf cl pB
θf
nq
where f P H pTq and p f
nq
nis any representative of f. Passing to the quotient spaces we get a linear embedding ¯i and an injective morphism of algebras ¯i
0such that ¯i |
ApTq¯i
0. For any H P B pTq one has
¯i p dH dz q d
dz ¯i p H q
and ¯i pB
θH q B
θ¯i p H q .
2.2. The algebra of generalized numbers of exponential type.
Let C
ebe the algebra of complex valued sequences p z
nq
n¥1such that:
D a ¡ 0, D η P N
, @ n P E
η, | z
n| ¤ a
n.
Elements of C
eare said to be of exponential growth. In the same way, we define I
eas the set of elements p z
nq
nP C
efor which
@ b P p 0, 1 q , D η P N
, @ n P E
η, | z
n| ¤ b
n.
The elements of I
eare said to be of exponential decrease. It may be seen that C
eis a subalgebra of C and that I
eis an ideal of C
e.
Definition 2.1. The algebra of complex generalized numbers of expo- nential type, is the quotient algebra C C
e{ I
e.
The complex number z is identified with a generalized number cl p z
nq where z
nz for all n. We denote by ˜ T the subalgebra of C constituted of elements z with a representative in T
N.
Definition 2.2. [18, Definition 3.3] Let f P H pTq and z P T ˜ . The value f p z q of f at z is the generalized number f p z q cl p f
np z
nqq where f cl p f
nq and z cl p z
nq with p z
nq
nP T
N.
2.2.1. Fourier coefficients of a generalized hyperfunction.
Definition 2.3. The Fourier coefficent of rank k P Z of the generalized hyperfunction f is the generalized number
f ˆ p k q cl 1
2iπ
»
|z|1
f
np z q z
k1dz
where p f
nq
nis an arbitrary representative of f.
The Fourier coefficients do not depend on the chosen representative and we have the following:
Proposition 2.2. [18, Proposition 3.8 ] If f P H pTq , then:
(i) There exists F P H pTq such that B
θF f if and only if f ˆ p 0 q 0.
(ii) There exists F P H pTq such that dF
dz f if and only if f ˆ p 1 q 0.
2.3. Invertibility. We denote by C
the subset of invertible elements in C . It follows from [18, Threorem 3.9], that z P C
if and only if z admits a representative p z
nq
nsuch that
D b P p 0, 1 q , D η P N
, @ n ¡ η, | z
n| ¥ b
n.
Let H
pTq denote the subset of invertible elements of H pTq . From [18, Theorem 3.10], we know that f P H
pTq if and only if it admits a representative p f
nq
nfor which there is r ¡ 1 such that f
nP O
rand:
D b P p 0, 1 q , D η P N
, @ n ¡ η, inf
zPCr
| f
np z q| ¥ b
n.
This means that the generalized number cl p inf
zPCr| f
np z q|q is invertible.
Moreover this condition does not depend on the chosen representative.
2.4. The topological structure of H pTq .
Definition 2.4. [17, Definition 3.1] The indicator of f P H pTq is:
ν p f q lim
rÑ1
lim sup
nÑ 8
} f
n}
1r{n(1) where p f
nq
nis an arbitrary representative of f .
It is shown (c.f. [17, Proposition 3.6] that ν p f q is also given by ν p f q lim
rÑ1
#
lim sup
nÑ 8
sup
kPZ
p r
|k|| p f
np k q|q
1{n+
. (2)
Then we have:
Proposition 2.3. [17, Proposition 3.1] Let f, g P H pTq and λ P C
. Then the following holds.
(i) ν p f q ¥ 0 and ν p f q 0 iff f 0;
(ii) ν p λf q ν p f q ; (iii) ν p f g q ¤ ν p f q ν p g q ;
(iv) ν p f g q ¤ sup p ν p f q , ν p g qq ; (v) | ν p f q ν p g q| ¤ ν p f g q ;
(vi) ν p f
1q ¥ p ν p f qq
1if f P H
pTq . Setting
ω p f, g q ν p f g q , f, g P H pTq ,
we define a translation invariant ultrametric distance on H pTq . More- over addition and multiplication are continuous mappings from H pTq
2to H pTq where H pTq
2is endowed with the ultrametric distance D de- fined by
D rp f, g q , p u, v qs sup p ω p f, u q , ω p g, v qq .
The inverse fonction is a continuous operator of H
pTq (see [17, Propo- sition 3.4 and Corollary 3.2]). We end this section by the following result.
Proposition 2.4. [17, Corollary 3.5] The following holds:
(i) If f P ¯i p B pTqq and f 0, then ν p f q 1.
(ii) The mapping ν is surjective from H pTq to R . 3. Completeness of basic subalgebras
3.1. Completeness of the ultrametric space C. The subalgebra C of H pTq is endowed with the restriction of ν and then with the restriction of the metric ω.
Theorem 3.1. The ultrametric space p C , ω q is complete. Then it is a closed subspace of H pTq .
Proof. Let p λ
mq
mbe a Cauchy sequence in C ; we denote by p λ
m,nq
na representative of λ
m. Then we have:
@ ε ¡ 0, D m
0P N
, @ p, q P N
, p ¡ q ¥ m
0, lim inf
nÑ 8
| λ
p,nλ
q,n|
1{n¤ ε { 2.
Hence, for each p p, q q as above there exists η ¡ 0 such that | λ
p,nλ
q,n|
1{n¤ ε. It follows that we can define two sequences p m
kq and p η
kq of positive integers both strictly increasing and such that:
@ k P N
, @ n P N
, n ¥ η
k, | λ
mk 1,nλ
mk,n| ¤ 1
2
kn. (3) We define the sequence p µ
mq
min C by
µ
k,nλ
mk,nif n ¥ η
kand µ
k,n0 if n η
k.
Since the sequence p η
kq is increasing, we have µ
k 1,n0 if n η
k. Then it follows that
@ k P N
, @ n P N
, | µ
k 1,nµ
k,n| ¤ 1
2
kn. (4)
Hence, we have
¸
8 k1| µ
k 1,nµ
k,n| ¤ ¸
8k1
1 2
n k1
2
n1 .
It follows that for each n P N
, the sequence p µ
k,nq
kconverges to ζ
nwhere
ζ
nµ
1,n¸
8 k1µ
k 1,nµ
k,n.
This shows that p ζ
nq is a moderate element, and then we set ζ cl p ζ
nq . Using (4), we have for every p P N
:
| µ
k p,nµ
k,n| ¤
p
¸
1 j0| µ
k j 1,nµ
k j,n| ¤
p
¸
1 j01 2
n k j¤ 1
2
n k11 2
n1 . Letting p Ñ 8 , we get that
| ζ
nµ
k,n| ¤ 1
2
n k11 2
n1 , from which it follows that
lim sup
nÑ 8
| ζ
nµ
k,n|
1{n¤ 1
2
k. This means that ν p µ
kζ q ¤
12kshowing that p µ
kq
kconverges to ζ in p C, ω q . But since µ
k,nλ
mk,nfor n ¥ η
k, it follows that µ
kλ
mkwhich implies that p λ
mq
mconverges to ζ and concludes the proof. l 3.2. The ultrametric algebras H
rpTq . For every r ¡ 1 we set X
erpTq tp f
nq
nP X
epTq , D η P N , @ n ¡ η, f
nP O
r, lim sup
nÑ 8
} f
n}
1{nr8u and we define
H
rpTq t f P H pTq , Dp f
nq
nP X
erpTq , cl p f
nq f u . Therefore, if R r 0, 8q , we get a well defined mapping
ν
r: H
rpTq Ñ R by setting
ν
rp f q inf t lim sup
nÑ 8
} f
n}
1r{n, p f
nq
nP X
erpTq , cl p f
nq f u . (5) Then, ν
rsatisfies to the following.
Proposition 3.2. Let f, g P H
rpTq and λ P C
. Then we have:
(i) ν
rp λ q ν p λ q ; (ii) ν
rp λf q ν
rp f q ; (iii) ν p f q ¤ ν
rp f q ;
(iv) ν
rp f q 0 if and only if f 0;
(v) ν
rp f g q ¤ ν
rp f q ν
rp g q ;
(vi) ν
rp f g q ¤ max p ν
rp f q , ν
rp g qq .
Proof. Assume that cl p λ
nq and cl p µ
nq are two representatives of λ.
Then, we have p λ
nµ
nq
nP N
eand consequently for every b P p 0, 1 q there is η P N such that | λ
nµ
n| b
nfor n ¡ η. Therefore
| λ
n|
1{n¤ p| µ
n| b
nq
1{n¤ | µ
n|
1{nb
and then lim sup
nÑ 8| λ
n|
1{n¤ lim sup
nÑ 8| µ
n|
1{n. It follows that lim sup
nÑ 8| λ
n|
1{nlim sup
nÑ 8| µ
n|
1{nwhich shows that
ν
rp λ q lim sup
nÑ 8
| λ
n|
1{nν p λ q
and proves (i). The proof of (ii) can be done following those of [17, Proposition 3.1], (see Proposition 2.3). To prove (iii), let α ¡ ν
rp f q . Then, there exists a representative p f
nq
nof f in X
erpTq such that lim sup
nÑ 8} f
n}
1{rnα. Since } f
n}
1{ρn¤ } f
n}
1{rnfor ρ r, it follows that ν p f q lim
ρÑ1p lim sup
nÑ 8} f
n}
1ρ{nq α. Thus, ν p f q ¤ ν
rp f q . We see that (iv) follows from (iii). Now take β ¡ ν p g q and choose a representative p g
nq
nof g such that lim sup
nÑ 8} g
n}
1r{nβ. Since lim sup
nÑ 8} f
ng
n}
1r{n¤ lim sup
nÑ 8} f
n}
1r{nlim sup
nÑ 8} g
n}
1r{n, it follows that ν
rp f g q ¤ αβ proving (v). Using the above notation, there exists η P N such that } f
n}
rα
nand } g
n}
rβ
nfor n ¡ η. It follows that
} f
ng
n}
1r{n¤ p α
nβ
nq
1{n. Assuming tha α ¥ β we get
p α
nβ
nq
1{nα
1 β
α
n1{nÑ α as n Ñ 8 which proves (vi). The proof of the proposition is then complete.
l
Clearly H
rpTq is a subalgebra of H pTq and H
rpTq H
spTq if r ¥ s ¡ 1 since ν
r¥ ν
s. Moreover we have H pTq Y
r¡1H
rpTq . We introduce the ultrametric distances ω
ron H
rpTq and D
ron H
rpTq
2as follows:
ω
rp f, g q ν
rp f g q and D
rpp f, u q , p g, v qq max p ω
rp f, g q , ω
rp u, v qq . It is easily seen that addition and multiplication are continuous maps from H
rpTq
2to H
rpTq , and the inverse map is a continuous operator on H
rpTq
the group of invertible elements in H
rpTq . Moreover, if r ¥ s ¡ 1 the embeddings u
s,r: H
rpTq Ñ H
rpTq and u
r: H
rpTq Ñ H pTq are continuous. It follows that
H pTq ind lim
rÑ1H
rpTq ,
can be endowed with the inductive limit topology of the spaces H
rpTq
which will be denoted by T . Then we have:
Proposition 3.3. The inductive limit topology defined by the ultramet- ric spaces H
rpTq on H pTq is finer that the one induced by ν.
Proof. Let V be an open set in H pTq for the topology defined by ν and take f P V . Then, there exists an open ball centered at f such that B p f, α q V . If r ¡ 1 is such that f P H
rpTq , the corresponding open ball B
rp f, α q for the topology induced by ν
rsatisfies B
rp f, α q B p f, α q since ν ¤ ν
r. It follows that B
rp f, α q V X H
rpTq which proves that V X H
rpTq is an open set in H
rfor the topology induced by ν
r. Hence V is an open set for the topology T , which concludes the proof. l For any bounded function g on T , we set
} g }
8,Tsup
zPT
| g p z q| . Then, the following holds:
Proposition 3.4. Let f P H pTq . If p f
nq
nand p g
nq
nare two represen- tatives of f , then
lim sup
nÑ 8
} f
n}
18{n,Tlim sup
nÑ 8
} g
n}
18{n,T.
Proof. Since p f
ng
nq
nP N
epTq , then for every b P p 0, 1 q there are r ¡ 1 and η P N such that f
n, g
nP O
rand } f
ng
n}
rb
nif n ¡ η.
Thus we have: @ b P p 0, 1 q , D r ¡ 1, D η P N , @ n ¡ η, } f
ng
n}
8,Tb
n, n ¡ η.
It folows that } f
n}
8,T¤ } g
n}
8,Tb
nfor n ¡ η and then lim sup
nÑ 8
} f
n}
18{n,T¤ max p lim sup
nÑ 8
} g
n}
18{n,T, b q .
- If lim sup
nÑ 8} g
n}
18{n,T0, then lim sup
nÑ 8} f
n}
18{n,T¤ b for every b P p 0, 1 q which implies that lim sup
nÑ 8} f
n}
1{8,nT0.
- If lim sup
nÑ 8} g
n}
18,T{n¡ 0, taking b lim sup
nÑ 8} f
n}
18,T{ngives lim sup
nÑ 8} f
n}
18{n,T¤ lim sup
nÑ 8} g
n}
18{n,T.
We have proved that in any case we have lim sup
nÑ 8
} f
n}
18{n,T¤ lim sup
nÑ 8
} g
n}
18{n,T.
The converse inequality can be shown to be true in the same way. l This allows us to define
ν
1p f q lim sup
nÑ 8
} f
n}
18,T{n(6)
where p f
nq
nis any representative of f . It is easy to see that properties
(i), (iii) and (vi) of Proposition 3.2 are satisfied for r 1 and ν
1¤ ν.
Theorem 3.5. For every r ¡ 1 and for every f P H
rpTq we have:
(i) ν
rp f q ¤ max p ν
rp f
1q , ν
1p f qq ; (ii) ν
1p f
1q ¤ a
ν
1p f q ν p f q .
Proof. For z P C
rset z
1z {| z | . If p f
nq
nis a representative of f , we have
f
np z q
»
rz1,zs
f
n1p ξ q dξ f
np z
1q and then
| f
np z q| ¤ | z z
1|} f
n1}
r} f
n}
8,T.
Since | z z
1| ¤ max p r 1, 1 1 { r q r 1, it follows that
| f
np z q| ¤ p r 1 q} f
n1}
r} f
n}
8,T. Finally we obtain
lim sup
nÑ 8
} f
n}
1r{n¤ max p lim sup
nÑ 8
} f
n1}
1r{n, lim sup
nÑ 8
} f
n}
18{n,Tq from which (i) follows.
Now let a P T and choose s ¡ 0 such D p a, s q C
rwhere D p a, s q t z P C , | z a | s u . Recall that the remainder after the term of degree m in the Taylor expansion of f
nabout a is
R
n,mp z q p z a q
m 12iπ
»
Γs
f
np ξ q dξ p ξ z qp ξ a q
m 1where Γ
st ξ P C , | ξ a | s u . It follows that if | z a | ¤ ρ s, then
| R
n,mp z q| ¤ s s ρ
ρ s
m 1
} f
n}
r.
Thus, if | z a | ρ and z P T , writting f
np z q f
np a q p z a q f
n1p a q R
n,1p z q and using the above inequality with m 1 gives
} f
n1}
8,T¤ 2 } f
n}
8,Tρ
ρ
s p s ρ q } f
n}
r. (7) Set ρ ts with t P p 0, 1 q . Therefore (7) becomes
} f
n1}
8,T¤ 1 s
2 } f
n}
8,Tt
t
1 t } f
n}
r. (8)
Let α 2 } f
n}
8,Tand β } f
n}
r. We let ϕ denote the function ϕ p t q α
t
βt 1 t where t P p 0, 1 q . A simple calculation gives
ϕ
1p t q p β α q t
22αt α
t
2p 1 t q
2.
For β α 0, the value of the reduced discriminant of the polynomials p β α q t
22αt α being equal to ?
αβ, we find that it has two roots t
0and t
1given by
t
0α ? αβ
β α and t
1α ? αβ β α . If β ¡ α, we find that
t
00 and t
1? α
? α ? β , If β α, we find that
t
0¡ 1 and t
1? α
? α ? β . If α β, ϕ
1p t q vanishes for t 1
2 and ϕ p 1
2 q 3α.
Therefore, in any case ϕ p t q reaches its minimum at t
? α
? α ? β in p 0, 1 q and we find that
ϕ
?
? α
α ?
β
α 2 a αβ.
This equality is also true when β α. Finally we obtain } f
n1}
8,T¤ 2
s p 2 } f
n}
8,Tb
2 } f
n}
8,T} f
n}
rq . It follows that
ν
1p f
1q ¤ max p ν
1p f q , a
ν
1p f q c
lim sup
nÑ 8
} f
n}
1r{nq .
Making r Ñ 1 and using ν p f q lim
rÑ1p lim sup
nÑ 8} f
n}
1r{nq gives (ii) and concludes the proof. l
Using Theorem 3.5, (ii) we get straightforwardly:
Corollary 3.6. Let f P H pTq . If ν
1p f q 0, then for every m P N
we have ν
1p f
pmqq 0.
3.3. Continuity of the differential operators d { dz and B
θ. To establish the continuity of these differential operators we state and prove the following.
Theorem 3.7. Let f P H
rpTq for some r ¡ 1. The following holds:
(i) ν
ρpB
θf q ν
ρp f
1q ¤ ν
rp f q , @ ρ P p 1, r q ; (ii) ν pB
θf q ν p f
1q ¤ ν p f q ;
(iii) If f ˆ p 0 q 0, then ν pB
θf q ν p f
1q ν p f q .
Proof. Let p f
nq
ndenote a representative of f in X
erpTq and let z P C
ρwith ρ P p 1, r q . We have pB
θf qp z q izf
1p z q with
1ρ¤ | z | ¤ ρ, and then 1
ρ } f
n1}
ρ¤ }B
θf
n}
ρ¤ ρ } f
n1}
ρwhich gives
lim sup
nÑ 8
}B
θf
n}
1ρ{nlim sup
nÑ 8
} f
n1}
1ρ{n. It follows that ν
ρpB
θf q ν
ρp f
1q and ν pB
θf q ν p f
1q .
Let ρ P p 1, r q and take r
1such that ρ r
1r. Hence, for all z P C
ρwe have
f
np z q 1 2iπ
»
|ξ|r1
f
np ξ q dξ ξ z 1
2iπ
»
|ξ|1{r1
f
np ξ q dξ ξ z and then
f
n1p z q 1 2iπ
»
|ξ|r1
f
np ξ q dξ p ξ z q
21
2iπ
»
|ξ|1{r1
f
np ξ q dξ p ξ z q
2. It follows that
| f
n1p z q| ¤ r
1} f
n}
r1p r
1ρ q
21 r1
} f
n}
r1p
1ρ r11q
2. Simple calculation gives
| f
n1p z q| ¤ r
1r
1ρ
2p r
1ρ q
2} f
n}
r1and then
} f
n1}
ρ¤ r
1r
1ρ
2p r
1ρ q
2} f
n}
r1. Using } f
n}
r1¤ } f
n}
rand letting r
1Ñ r yields
} f
n1}
ρ¤ r rρ
2p r ρ q
2} f
n}
r.
It follows that ν
ρpB
θf q ν
ρp f
1q ¤ ν
rp f q and ν pB
θf q ν p f
1q ¤ ν p f q which proves (i) and (ii).
Since pB {
θf
nqp k q ik f p
np k q for all k P Z , it follows from (2) that ν p f
1q lim
ρÑ1
#
lim sup
nÑ 8
sup
kPZ
p ρ
|k|| k || p f
np k q|q
1{n+
.
Hence, if ˆ f p 0 q 0, we can choose p f
nq
nsuch that f p
np 0 q 0 for every n and we will have
sup
kPZ
p ρ
|k|| k || p f
np k q|q ¥ sup
kPZ
p ρ
|k|| p f
np k q|q .
This leads to ν pB
θf q ¥ ν p f q and then ν pB
θf q ν p f q , proving (iii). l Thus, the following corollary is a straightforward consequence of The- orem 3.7.
Corollary 3.8. The differential operators d { dz and B
θare continuous in each of the following cases:
(i) from H pTq to H pTq ; (ii) from H
rpTq to H pTq ;
(iii) from H
rpTq to H
spTq with 1 s r.
Consequently H pTq is a topological differential algebra.
3.4. Completeness of the topological algebras H
rpTq .
Theorem 3.9. The ultrametric algebra p H
rpTq , ω
rq is a complete one.
Proof. Let p F
mq
mbe a Cauchy sequence in H
rpTq . It follows from the definition of ν
rthat there exist m
1, m
2P N
with m
2¡ m
1and two representatives p F
mr11s,nq
nand p F
mr12s,nq
nof F
m1and F
m2respectively such that:
lim sup
nÑ 8
} F
mr12s,nF
mr11s,n}
1r{n1
2
1. (9)
Then, we set
F
m1,nF
mr11s,nand F
m2,nF
mr12s,n. (10) In the same way we get m
3P N
with m
3¡ m
2and two representatives p F
mr2s2,nq
nand p F
mr2s3,nq
nof F
m2and F
m3respectively such that:
lim sup
nÑ 8
} F
mr2s3,n
F
mr2s2,n
}
1r{n1 2
2. Then, for each n P N
, we set
F
m3,nF
mr23s,nF
mr22s,nF
m2,n.
Hence, by induction, we get a subsequence p F
mkq
kalong with repre- sentatives p F
mrksk 1,n
q
nand p F
mrksk,n
q
nof F
mrksk 1
and F
mrksk
respectively such that for every k P N
,
lim sup
nÑ 8
} F
mrksk 1,n
F
mrksk,n
}
1r{n1
2
k. (11)
Then, for every p k, n q P N
N
we set F
mk 1,nF
mrksk 1,n
F
mrksk,n
F
mk,n. (12)
It follows that
F
mj 1,nF
mj,nF
mrjsj 1,n
F
mrjsj,n
for 1 ¤ j ¤ k, and summing up we find that for every k ¥ 2:
F
mk 1,nF
mrksk 1,n
¸
k j2p F
mrj1sj,nF
mrjsj,nq . (13) Since p F
mrj1sj,n
q
nand p F
mrjsj,n
q
nare both representatives of F
mj,n, it follows that °
kj2
r F
mrjj,n1sF
mrjjs,ns
n
P N
epTq and then p F
mk 1,nq
nis a repre- sentative of F
mk 1. Using (12), we get F
mk 1F
mkF
mrksk 1,n
F
mrksk,n
and then using (11) we find lim sup
nÑ 8
} F
mk 1,nF
mk,n}
1r{n1
2
k. (14)
Then, there exists a sequence p η
kq
kof positive integers which is strictly increasing and such that
@p k, n q P N
N
, n ¥ η
k, } F
mk 1,nF
mk,n}
r¤ 1
2
k n. (15) For each k P N
, we define the sequence of functions p G
k,nq
nas follows:
G
k,nF
mk,nif n ¥ η
kand G
k,n0 otherwise.
It follows that p G
k,nq
nis a moderate sequence, and if G
krp G
k,nqs , then G
kF
mk. We also have:
@p k, n q P N
N
, } G
k 1,nG
k,n} ¤ 1
2
n k. Using successively the above inequality, we get for every p P N
:
} G
k p,nG
k,n}
r¤ } G
k p,nG
k p1,n}
r} G
k 1,nG
k,n}
r¤
21nk p121nk
¤
21nk1 2n
p11 } G
k p,nG
k,n}
r¤
21nk1 12n1
.
It follows that for each n P N
, the sequence p G
k,nq
kis a Cauchy se- quence in O
rand then it converges to an element g
nin O
r. Letting p Ñ 8 in the above inequality gives
} g
nG
k,n}
r¤ 1
2
n k11
2
n1 . (16)
This shows that p g
nq is a moderate element; in fact we have:
} g
n}
r¤ } G
k,n}
r1 2
k1 n.
Then we set g rp g
nqs . Using (16), we have for every p P N
: } g
nG
k,n}
1r{n¤
1 2
k11 2
n1
1{nwhich gives
ν
rp g G
kq ¤ lim sup
nÑ 8
} g
nG
k,n}
1r{n¤ 1
2
kand proves that
kÑ 8
lim ν
rp g G
kq 0.
Hence, p F
mkq
kconverges to g in H pTq , and since p F
mq
mis a Cauchy sequence, it converges to g which concludes the proof. l
4. Functional calculus and applications
All the results stated in this section for the algebra H pTq are also true for the subalgebras H
rpTq and C.
4.1. Exponential, logarithm and power functions.
4.1.1. The exponential of a generalized hyperfunction. Let u P H pTq and let p u
nq be a representative of u such that u
nP O
rfor some r ¡ 1.
If z P C
r, then | exp p u
np z qq| exp p <u
np z qq and consequently } exp p u
nq}
rexp p sup
zPCr
<u
np z qq .
It follows that p u
nq satisfies } exp p u
nq}
r¤ a
nfor some positive constant a if and only if sup
zPCr<u
np z q ¤ n ln a.
Definition 4.1. A generalized hyperfunction u is said to be real sublin- ear if it admits a representative p u
nq
nsuch that u
nP O
rfor some r ¡ 1 and sup
zPCr<u
np z q ¤ λn for a real constant λ and n large enough.
We have the following:
Proposition 4.1. For a generalized hyperfunction u, the condition to be real sublinear does not depend on the chosen representative.
Proof. Let p u
nq
nand p v
nq
nbe two representatives of u where p u
nq
nis real sublinear; we set
α
nsup
zPCr
<u
np z q and β
nsup
zPCr
<u
np z q . It follows that
| e
βne
αn| |} e
vn}
r} e
un}
r| ¤ } e
vne
vn}
rand then using | e
z1 | ¤ | z | e
|z|, we get
| e
βne
αn| ¤ } e
unp e
vnun1 q}
r¤ } e
un}
r} e
vnun1 }
r¤ e
αne
}vnun}r} v
nu
n}
r.
Since p v
nu
nq
nis negligible, for every ε ¡ 0 there exists η
1P N such that e
}vnun}r} v
nu
n}
r¤ ε if n ¡ η
1. It follows that e
βn¤ p 1 ε q e
αnfor n ¡ η
1. Hence, if α
n¤ λn for n ¡ η ¡ η
1, then we have β
n¤ r λ ln p 1 ε qs n for n ¡ η which proves the proposition l
We notice that if u is bounded, i.e. } u
n}
r¤ α for some α ¡ 0 for n large enough, then it is real sublinear. Clearly, if u is real sublinear then λu is also real sublinear if λ is a nonnegative real number. It is easily seen that if u, v P H pTq , then
exp p u v q exp u exp v.
Moreover, since sup
zPCrp <u
np z qq inf
zPCr<u
np z q , it follows that p u q is real sublinear if and only if inf
zPCr<u
np z q ¥ µn for some µ P R when n is large enough. Thus u and p u q are both real sublinear if and only if there are λ, µ P R such that
µn ¤ inf
zPCr
<u
np z q ¤ sup
zPCr
p <u
np z qq ¤ λn.
Under this condition exp p u q and exp p u q are invertible with r exp p u qs
1exp p u q .
4.1.2. The exponential of u for ν p u q 1.
Theorem 4.2. If u P H pTq is such that ν p u q 1, then exp p u q is well defined in H pTq and is given by
exp p u q ¸
8k0
u
kk! .
Proof. Let u P H pTq satisfy ν p u q 1 and choose any representative p u
nq
nof u. Then we have:
ν p u q lim
rÑ1
p lim sup
nÑ8
} u
n}
1r{nq 1.
Hence, for every α such that ν p u q α 1, there exists ρ ¡ 1 such that
ν
ρp u q lim sup
nÑ8