Pilipović, Stevan - Stanković, Bogoljub - Vindas, Jasson - Asymptotic Behavior of Generalized Functions PDF

Series on Analysis, Applications and Computation – Vol.
ISAAC
Asymptotic Behavior of
Generalized Functions
8266hc.9789814366847-tp.indd 1 9/5/11 3:00 PM

Series on Analysis, Applications and Computation
Series Editors: Heinrich G W Begehr (Freie Univ. Berlin, Germany)

Robert Pertsch Gilbert (Univ. Delaware, USA)
M. W. Wong (York Univ., Canada)
Advisory Board Members:
Mikhail S Agranovich (Moscow Inst. of Elec. & Math., Russia),
Ryuichi Ashino (Osaka Kyoiku Univ., Japan),
Alain Bourgeat (Univ. de Lyon, France),
Victor Burenkov (Cardiff Univ., UK),
Jinyuan Du (Wuhan Univ., China),
Antonio Fasano (Univ. di Firenez, Italy),
Massimo Lanza de Cristoforis (Univ. di Padova, Italy),
Bert-Wolfgang Schulze (Univ. Potsdam, Germany),
Masahiro Yamamoto (Univ. of Tokyo, Japan) &
Armand Wirgin (CNRS-Marseille, France)
Published
Vol. 1: Boundary Values and Convolution in Ultradistribution Spaces

by R D Carmichael, A Kami½ski & S Pilipoviƒ
Vol. 2: Complex Analysis

by M W Wong
Vol. 3: Topics in Mathematical Analysis

edited by P Ciatti, E Gonzalez, M L de Cristoforis & G P Leonardi
Vol. 4: Nonlinear Waves: An Introduction

by P Popivanov & A Slavova
Vol. 5: Asymptotic Behavior of Generalized Functions

by S Pilipović, B Stanković & J Vindas
He Yue - Asymptotic Behavior.pmd 1 9/15/2011, 11:32 AM

Series on Analysis, Applications and Computation – Vol. 5
ISAAC
Asymptotic Behavior of
Generalized Functions
º Stevan Pilipović
University of Novi Sad, Serbia
º Bogoljub Stanković
University of Novi Sad, Serbia
º Jasson Vindas
Ghent University, Belgium
World Scientific
NEW JERSEY • LONDON • SINGAPORE • BEIJING • SHANGHAI • HONG KONG • TA I P E I • CHENNAI
8266hc.9789814366847-tp.indd 2 9/5/11 3:00 PM

Published by
World Scientific Publishing Co. Pte. Ltd.
5 Toh Tuck Link, Singapore 596224
USA office: 27 Warren Street, Suite 401-402, Hackensack, NJ 07601
UK office: 57 Shelton Street, Covent Garden, London WC2H 9HE
British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library.
Series on Analysis, Applications and Computation — Vol. 5

ASYMPTOTIC BEHAVIOR OF GENERALIZED FUNCTIONS
Copyright © 2012 by World Scientific Publishing Co. Pte. Ltd.
All rights reserved. This book, or parts thereof, may not be reproduced in any form or by any means,
electronic or mechanical, including photocopying, recording or any information storage and retrieval
system now known or to be invented, without written permission from the Publisher.
For photocopying of material in this volume, please pay a copying fee through the Copyright
Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA. In this case permission to
photocopy is not required from the publisher.
ISBN-13 978-981-4366-84-7
ISBN-10 981-4366-84-6
Printed in Singapore.
He Yue - Asymptotic Behavior.pmd 2 9/15/2011, 11:32 AM

September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
Preface
There are several approaches to generalized asymptotics within spaces of

generalized functions. Probably the most developed approaches are that
of Vladimirov, Drozhinov and Zavyalov [192], and of Kanwal and Estrada
[56]. The first approach is followed by the authors of this book and extended
in the direction of the S-asymptotics. The second approach is related to
moment asymptotic expansions of generalized functions and the Cesàro
behavior, and we refer to papers of Estrada, Kanwal and Vindas given in the
references. We refer to [135] for the different definitions of the asymptotic
behavior of distributions.
Asymptotic analysis is a very old and wide branch of mathematics. In
general, the state of a system after certain period of time, or the behavior
of a system in the frame of some limit procedure described by a suitable
mathematical model, leads to the use and development of the methods of
asymptotic analysis. With the general development of various branches of
mathematical analysis and, especially, the theory of differential equations,
the asymptotic analysis has obtained new impulses that resulted in new
approaches and methods. In our analysis, such impulses came from the use
of slowly varying functions and the introduction of asymptotic behavior of
generalized functions.
In the first years of the twentieth century, the needs for a class of func-
tions with growth order between the constant functions and power func-
tions, in the scale of functions used in asymptotic analysis, appeared in
many papers. Already Landau [87] used a term similar to “regular varia-
tion”, but the theory of regularly varying and slowly varying functions (cf.
0.3) belongs to Karamata and his pupils. Karamata’s basic results were
published in [76] and elaborated in [77]. Although the theory of regularly
v
vi Preface
varying functions was invented for proving Tauberian type theorems for the
Laplace-Stieltjes transform, such functions have been successfully employed
in many other areas of mathematics: for Mercerian type theorems, in an-
alytic number theory, the theory of entire functions and in the analysis of
various classes of differential equations. Also, Karamata regularly varying
functions were recognized as an important tool in probability theory [60].
At the present time the class of regularly varying functions is an integral
part of asymptotic analysis. In the monographs [148], [64] and [9] one
can find collected together the theory and applications of regular varying
functions, the last one being certainly the most comprehensive treatise.
The class of regularly varying functions became more interesting and
more applicable after the introduction of the notions of asymptotic behav-
ior and expansion of generalized functions. We refer to [135] for a general
overview of the development of the theory of asymptotic behavior of gen-
eralized functions up to 1989.
The notion of asymptotic behavior of generalized functions have had
a very important role in quantum physics, [8], [10], [191], [192], where
rigorous proofs of results in the foundation of this theory were obtained
by the use of generalized asymptotic behavior. These papers motivated
mathematicians to develop further the methods of asymptotic analysis on
spaces of generalized functions.
The first definition of an asymptotic behavior for a generalized function
is attributed to Lighthill [91]. His definition can be applied only to the
so-called semi-regular distributions. Such a distribution T is equal to a
locally integrable function f in a neighborhood of the observed point (in
our case, infinity). The asymptotics of the function f at infinity gave the
asymptotics of the distribution T at infinity. The same idea was adopted
by Jones [72], Lavoine [88], Mangad [95], and Zemanian[203].
The next step forward was made by Silva in his axiomatic approach to
distribution theory [147] and later by Lavoine and Misra [89], [90]. Silva
introduced the growth order symbols small o and big O, in order to measure
the asymptotic behavior of a distribution. Let r ∈ C ∞ (a, ∞), a ∈ R and
let F ∈ C(a, ∞). If a distribution T satisfies
F (x)
T = r · Dm F in (a, ∞), m ∈ N0 , and lim = 0,
x→∞ xm
Preface vii
then T = o(r), x → ∞. If there exists a constant M such that

F (x)

xm ≤ M, x ∈ (b, ∞), b ≥ a ,
then T = O(r), x → ∞. Finally, if T = Dm F in (a, ∞) and

F (x) C
lim m
= ,
x→∞ x m!
then T behaves as the constant C in D′ (R).
Lavoine and Misra defined the notion of “equivalence at infinity”. The
distribution T is equivalent at infinity with Axα , A 6= 0, α ∈
/ −N, if there
exist m ∈ N0 , m + α > 0, and F ∈ C(R), such that T = Dm F on (a, ∞)
and
F (x) ∼ ACα,m xα+m , x → ∞,

1, m=0
Cα,m =
Γ(α + 1)/Γ(α + m + 1), m > 0 .
Later on, they replaced Axa by Axa log x in their definition of equivalence
at infinity.
A natural generalization came with the introduction of regularly varying
functions (cf. 0.3) as the basic comparison functions for the asymptotics
of generalized functions.
The notion of the quasi-asymptotic behavior (the quasi-asymptotics)
of tempered distributions appeared as a qualitatively new important step
in the asymptotic analysis of generalized functions. The first paper with
the definition of the quasi-asymptotic behavior was the one of Zavyalov
[200]. He introduced the quasi-asymptotics as automodel asymptotics “av-
tomodel’naya asimptotika”. First, he defined a class M+ of tempered dis-
tributions as follows: T ∈ M+ if and only if there exists a ∈ R such that
r
T ∈ C ∞ (−∞, a) and decreases faster than |x| , as x → −∞, for any pos-
itive r. For Zavyalov, T ∈ M+ has the quasi-limit (q − lim) equal to C
if
T (kx) → CH(x), k → ∞ in S ′ (R) ,
where H(x) is the Heaviside function. In short, q − limx→∞ T (x) = C. If

T ∈ M+ and q − lim T (α) (x) = C 6= 0, α ∈ R, then T admits a quasi-
asymptotic part of degree α at infinity. Zavyalov proved an equivalent
viii Preface
statement: T ∈ M+ admits a quasi-asymptotic part of degree α at infinity

if and only if
k −α T (kx) → g(x), x → ∞ in S ′ (R)
for some distribution g 6= 0. This second statement has been used thereafter
as the definition of the quasi-asymptotics.
Let us emphasize that the predominant role in the development of the
quasi-asymptotics belongs to the Russian school with Vladimirov and his
pupils and collaborators. Their results have given important impulses to
the study of asymptotic behavior in different spaces of generalized functions
and their applications. Especially, the robust work of Drozhzhinov and
Zavyalov [38], [39], [40], [42], in the last ten years has been an extremely
important contribution to the subject.
The notion of the shift asymptotics appeared for the first time in [146],
Chapter VII, the second remark after Theorem VI:
inf{k ∈ R; {T (x + h)/(1 + |h|2 )k/2 }h∈Rn is a bounded subset of D′ (Rn )}
was called the degree of growth of the distribution T at infinity.
The shift asymptotics, defined in [15], has been later called S-
asymptotics by the first two named authors of this book. It is said that
′
T ∈ S+ has shift asymptotics at infinity related to the regularly varying
function ρ(h) = hα L(h) if
T (x + h)/ρ(h) → g(x), h → ∞ in S ′ .
It was proved that if α > −1, then T has quasi-asymptotics related to ρ
at infinity. This result can give the impression that the S-asymptotics is
a particular case of the quasi-asymptotics; however, this is not really the
case. The relation between the S-asymptotics and the quasi-asymptotics
is described in this book, Chapter I 2.13. The first two authors of the
book adapted and made precise the S-asymptotic behavior of generalized
functions, and studied this type of asymptotic behavior for distributions
ultradistributions and hyperfunctions, giving structural characterizations
of comparison functions and the limit generalized function. It is worth to
mention that the quasi-asymptotics is a natural notion in the analysis of
tempered generalized functions, while the framework for the S-asymptotics
is the space of exponential distributions K1′ .
It should be mentioned that this book has a minor part in common
with the previous books [135] and [192], which are also devoted to the
Preface ix
asymptotic behavior of distributions. This was necessary in order to make

the exposition complete and more transparent. Conceptually, we first treat
the asymptotic behavior in a dual space Fg′ of a suitable, but rather general,
barrelled locally convex space of functions Fg . Many important spaces of
generalized functions are of this kind. The study is then specialized to a
space F ′ , a space of distribution, ultradistribution, or Fourier hyperfunction
type. In this way we develop a general theory of asymptotics in F ′ . We then
specify the space F ′ in order to analyze special properties of asymptotic
behaviors or expansions which are intrinsically connected with the nature
of each space under consideration. We illustrate the theory by examples
and point out some open problems which may serve as an encouragement
for further investigations.
The book consists of two parts. The first one deals with the basic prop-
erties and analysis of the S-asymptotics and quasi-asymptotics in the space
Fg′ and in various special spaces of generalized functions. The asymptotic
expansions and the Taylor expansion occupy also an important place in this
part of the book.
In the second part one can find several applications of the asymptotic
behavior of generalized functions. Abelian and Tauberian type theorems for
integral transforms of convolution and Mellin-type convolution are studied
for general kernels as well as for various special kernels such as the Laplace,
Stieltjes, Weierstrass and Poisson transforms. This part also contains the
asymptotic analysis of solutions to linear differential and partial differential
equations, and to equations with ultradifferential or local operators. In
addition, we give some applications of the quasi-asymptotic behavior of
distributions to the study of summability of Fourier series and integrals.
We point out that our exposition, based on the work of the first two
authors and their coauthors, has been extended with recent results of the
third author and his collaborators. In this way, this book contains novel
and strong results which give answers to some old problems in the area.
Those results are presented in Chapter I 2.10–2.12 and Chapter II 5, and
make the presentation up-to-date.
This page intentionally left blank

Contents
Preface v
I. Asymptotic Behavior of Generalized Functions 1
0 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . 1
1 S-asymptotics in Fg′ . . . . . . . . . . . . . . . . . . . . . 10
1.1 Definition . . . . . . . . . . . . . . . . . . . . . . . 10
1.2 Characterization of comparison functions and
limits . . . . . . . . . . . . . . . . . . . . . . . . . 11
′
1.3 Equivalent definitions of the S-asymptotics in F 16
1.4 Basic properties of the S-asymptotics . . . . . . . 18
1.5 S-asymptotic behavior of some special classes of
generalized functions . . . . . . . . . . . . . . . . 22
1.6 S-asymptotics and the asymptotics of a function . 31
1.7 Characterization of the support of T ∈ F0′ . . . . 34
1.8 Characterization of some generalized function spaces 40
1.9 Structural theorems for S-asymptotics in F ′ . . . 41
1.10 S-asymptotic expansions in Fg′ . . . . . . . . . . . 52
1.11 S-asymptotics in subspaces of distributions . . . . 66
1.12 Generalized S-asymptotics . . . . . . . . . . . . . 78
2 Quasi-asymptotics in F ′ . . . . . . . . . . . . . . . . . . . 82
xi
xii Contents
2.1 Definition of quasi-asymptotics at infinity over a

cone . . . . . . . . . . . . . . . . . . . . . . . . . . 82
2.2 Basic properties of quasi-asymptotics over a cone 85
2.3 Quasi-asymptotic behavior at infinity of some
generalized functions . . . . . . . . . . . . . . . . 96
2.4 Equivalent definitions of quasi-asymptotics at
infinity . . . . . . . . . . . . . . . . . . . . . . . . 101
2.5 Quasi-asymptotics as an extension of the classical
asymptotics . . . . . . . . . . . . . . . . . . . . . 104
2.6 Relations between quasi-asymptotics in D′ (R) and
S ′ (R) . . . . . . . . . . . . . . . . . . . . . . . . . 106
2.7 Quasi-asymptotics at ±∞ . . . . . . . . . . . . . 113
2.8 Quasi-asymptotics at the origin . . . . . . . . . . 117
2.9 Quasi-asymptotic expansions . . . . . . . . . . . 125
2.10 The structure of quasi-asymptotics. Up-to-date
results in one dimension . . . . . . . . . . . . . . 130
2.11 Quasi-asymptotic extension . . . . . . . . . . . . 150
2.12 Quasi-asymptotic boundedness . . . . . . . . . . . 160
2.13 Relation between the S-asymptotics and
quasi-asymptotics at ∞ . . . . . . . . . . . . . . . 166
II. Applications of the Asymptotic Behavior of Generalized

Functions 171
3 Asymptotic behavior of solutions to partial differential

equations . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
3.1 S-asymptotics of solutions . . . . . . . . . . . . . 171
3.2 Quasi-asymptotics of solutions . . . . . . . . . . . 177
3.3 S-asymptotics of solutions to equations with
ultra-differential or local operators . . . . . . . . . 185
4 Asymptotics and integral transforms . . . . . . . . . . . . 188
4.1 Abelian type theorems . . . . . . . . . . . . . . . 188
4.2 Tauberian type theorems . . . . . . . . . . . . . . 202
Contents xiii
5 Summability of Fourier series and integrals . . . . . . . . 237

5.1 The Cesàro behavior . . . . . . . . . . . . . . . . 238
5.2 Summability of the Fourier transform and
distributional point values . . . . . . . . . . . . . 248
Bibliography 283
Index 293
Chapter I
Asymptotic Behavior of Generalized

Functions
0 Preliminaries
0.1. We denote by R and N the sets of real and natural numbers;

N0 = N∪{0}. The following notation will be used. If x = (x1 , . . . , xn ), y =
(y1 , . . . , yn ) ∈ Rn , then x · y = x1 y1 + · · · + xn yn ; kxk2 = x21 + · · · +
x2n ; x ≥ 0 ⇐⇒ xi ≥ 0, i = 1, . . . , n; x → ∞ ⇐⇒ xi → ∞, i =
1, . . . , n; Rn+ = {x ∈ Rn ; x > 0}. If x = (x1 , . . . , xn ) ∈ Rn and
k = (k1 , . . . , kn ) ∈ Nn0 , then |k|= k1 + · · · + kn , k! = k1 ! . . . kn ! , xk =
xk11 . . . xknn ; Dk = ∂ k1 /∂xk11 . . . ∂ kn /∂xknn , f (k) = Dk f (f (0) = f );
z = (z1 , . . . , zn ) ∈ Cn ; D(a, r) denotes the polydisk {z ∈ Cn ; |zi − ai |<
ri , i = 1, . . . , n}, where |zj |2 = x2j + yj2 , zj = xj + iyj ∈ C. B(a, r) denotes
the open ball {x ∈ Rn , ka − xk< r} and H is the Heaviside function:
H(t) = 0, t ≤ 0; H(t) = 1, t > 0.
0.2. A cone with vertex at zero in Rn is a non-empty set Γ such that
x ∈ Γ and k > 0 imply kx ∈ Γ. The cone Γ is called solid if int Γ 6= ∅. The
conjugate cone (dual cone) Γ∗ to the cone Γ is the set {ξ ∈ Rn ; x · ξ ≥ 0
for each x ∈ Γ}. It is obvious that Γ∗ is also a cone which is convex and
closed. The cone Γ is called acute if Γ∗ is a solid cone.
0.3. A function ρ : (a, ∞) → R, a ∈ R+ , is called regularly varying at
infinity [76] if it is positive, measurable, and if there exists a real number
α such that for each x > 0
ρ(kx)
lim = xα . (0.1)
k→∞ ρ(k)
The number α is called index of regular variation. If α = 0, then ρ is called

slowly varying at infinity and for such a function the letter “L” will be used.
We then have that any regularly varying function can be written as ρ(x) =
1
2 Asymptotic Behavior of Generalized Functions
xα L(x), x > a, where L is slowly varying. It is known that the convergence

of (0.1) is uniform on every fixed compact interval [a′ , b′ ], a < a′ < b′ < ∞,
and that ρ is necessarily bounded (hence integrable) on it [9].
Let L be a slowly varying function at infinity. Then, for each ε > 0,
(i) there exist constants C1 , C2 > 0 and X > a such that
C1 x−ε ≤ L(x) ≤ C2 xε , x ≥ X ; (0.2)
(ii) lim xε L(x) = +∞, lim x−ε L(x) = 0.

x→∞ x→∞
We know ([9], p. 16) that if L2 (x) → ∞, x → ∞, and L1 , L2 are
slowly varying, then L1 ◦ L2 = L1 (L2 ) is slowly varying, as well. Hence, for
x > −∞,
L(x + h) L(log ut)
lim = lim = 1. (0.3)
h→∞ L(h) u→∞ L(log u)
The definition of a regular varying function at zero is similar. For the

definition of regularly generalized functions see [156] and [162].
0.4. The class of distributions fα , α ∈ R, belonging to S ′ + (see 0.5.1.)
is defined in the following way:

 α−1
 H(t)t /Γ(α), α > 0,
fα (t) =

 f (m) (t),
α+m α ≤ 0, α + m > 0,
where H is the Heaviside function, and the derivative f (m) is taken to be in

the distributional sense (see [192], Chapter I, §1). We therefore have f0 = δ,
the Dirac delta distribution; f−m = δ (m) , m ∈ N; and fp ∗ fq = fp+q . We
also use the notations tα α α
+ = Γ(α + 1)fα+1 (t) and t− = (−t)+ , α ∈ / −N.
′
Let g ∈ S+ . We denote g (−α) = fα ∗ g, α ∈ R (∗ the is convolution
symbol).
The sequence hδm im∈N ⊂ C ∞ (Rn ) is called a δ-sequence if:
a) Rsupp δm ⊂ [−αm , αm ], αm → 0, m → ∞; b) δm ≥ 0, m ∈ N;
c) δm (t)dt = 1, m ∈ N.
Rn
If ϕ ∈ D, then δm ∗ ϕ → ϕ, m → ∞ in D, hence {δm ∗ ϕ; m ∈ N} is a
bounded set in D.
0.5. We will repeat definitions and some basic properties of generalized
functions defined as elements of dual spaces.
0. Preliminaries 3
0.5.1. The Schwartz spaces of test functions and distributions on Rn are

denoted by D = D(Rn ) and D′ = D′ (Rn ), respectively (Rn will be omitted
wherever n is not fixed). The space D is a locally convex, barrelled, Montel
and complete space. If a filter with a countable basis is weakly convergent
in D′ , then it is convergent in D′ with the strong topology, as well (cf.
[146]).
S is the space of rapidly decreasing functions and its dual S ′ is the space
of tempered distributions [146]. For a closed cone Γ ⊂ Rn , S ′ Γ = {f ∈
S ′ ; supp f ⊂ Γ}. In the one-dimensional case S ′ + = {f ∈ S ′ (R); supp f ⊂
[0, ∞)}. Recall ([189]):
S(Γ) = {ϕ ∈ C ∞ (Γ); kϕkp < ∞, p ∈ N} ,
where
kϕkp = sup (1 + kxk2 )p/2 |ϕ(β) (x)| .

x∈Γ, |β|≤p
By Sp (Γ) is denoted the completion of the set S(Γ) with respect to the
T S ′
norm kϕkp . Note S(Γ) = Sp (Γ) and S ′ (Γ) = S p (Γ), where the
p∈N0 p∈N0
intersection and the union have topological meaning. A sequence hfn in∈N
in S ′ (Γ) converges to f ∈ S ′ (Γ) if and only if it belongs to some S ′ q (Γ) and
converges to f ∈ S ′ q (Γ) in the norm of S ′ q (Γ). The space S ′ Γ is isomorphic
to S ′ (Γ) if Γ is closed convex solid cone ([189], [192]).
E ′ the space of distributions with compact support; it is isomorphic to
the dual space of E = C ∞ (Rn ) (cf. [146]).
DLp , 1 ≤ p ≤ ∞, is the space of smooth functions with all derivatives
belonging to Lp ([146]), DLp ⊂ DLq if p < q.
. .
B is a subspace of B = DL∞ , defined as follows: ϕ ∈B if and only if
|ϕ(α) (x)|→ 0 as kxk→ ∞ for every α ∈ Nn0 .
1 1
D′ Lp , 1 < p ≤ ∞ is the dual space of DLq , 1 ≤ q < ∞, + = 1. D′ L1
p q
·
is the dual of B and D′ L∞ is denoted by B ′ .
O′ c is the space of distributions with fast descent:
O′ c = {T ∈ D′ ; (1 + kxk2 )m T ∈ B ′ , for every m ∈ N}.
Kp , p ≥ 1, is the spaces of functions ϕ ∈ C ∞ with the property:
νm (ϕ) = sup exp(mkxkp )|Dα ϕ(x)|< ∞, m = 1, 2, . . .

x∈Rn ,|α|≤m
The elements of Kp are called rapidly exponentially decreasing functions.

Then K′ p is the dual of Kp .
The convolution in D′ is defined as follows. Let hηm im∈N be a sequence
in D(R2n ) such that for every compact set K ⊂ R2n there exists m0 (K)
such that ηm (x) = 1, x ∈ K, m ≥ m0 (K) and
(β)
sup |ηm (x)|< Cβ , β ∈ Nn0 .
x∈R2n
The convolution of T, S ∈ D′ is defined by

hT ∗ S, ϕi = lim hT (x)S(y), ηm (x, y)ϕ(x + y)i, ϕ ∈ D ,
m→∞
if this limit exists for every hηm im (then, it does not depend on hηm im ).
By the Banach–Steinhaus theorem we know that T ∗ S ∈ D′ .
The spaces D′ (Γ) and S ′ (Γ) with the operation ∗ are associative and
commutative algebras. The convolution in this case is separately continu-
ous.
We refer also to [63] and [3] for the theory of distributions.
0.5.2 Ultradistribution spaces
We follow the notation and definitions from [79], [81] and [86]. By hMp ip
is denoted a sequence of positive numbers, M0 = M1 = 1, satisfying some
of the following conditions:
(M.1) Mp2 ≤ Mp−1 Mp+1 , p ∈ N;
(M.2) Mp /(Mq Mp−q ) ≤ AB p , 0 ≤ q ≤ p, p ∈ N;
(M.2)′ Mp+1 ≤ AB p Mp , p ∈ N0 , N0 = N ∪ {0};
P∞
(M.3) Mq−1 /Mq ≤ Ap Mp /Mp+1 , p ∈ N;
q=p+1
P
∞
(M.3)′ Mp−1 /Mp < ∞,
p=1
where A and B are constants independent of p.

In the sequel, we will always assume (M.1), (M.2)′ and (M.3)′ .
Let h be a positive number and let hhp ip be a real positive sequence
increasing to ∞. We denote
p
h , for the ultradifferentiable functions of class (Mp )
Hp = (0.4)
h1 . . . hp , for the ultradifferentiable functions of class {Mp }.
0. Preliminaries 5
Let K be a compact set in an open set Ω of Rn .

H Mp
We denote by EK p the space of smooth functions ϕ ∈ E such that
sup |ϕ(α) (x)|/H|α| M|α| < ∞ . (0.5)
x∈K,α∈Nn
0
H Mp H Mp
We set DKp = {ϕ ∈ EK p ; supp ϕ ⊂ K}, it is a Banach space with
norm
|ϕ(α) (x)|
qHp Mp (f ) = sup . (0.6)
x∈K,α∈Nn
0
H|α| M|α|
Then, the basic spaces are defined by
(Mp ) hp Mp {Mp } H Mp
DK = proj lim DK , DK = ind lim DKp
h→0 {Hp =h1 ...hp }
D∗ (Ω) ≡ DΩ
∗ ∗
= ind lim DK ,
K⊂⊂Ω
where ∗ denotes either (Mp ) or {Mp }.

The spaces with the upper index (Mp ) are the Beurling-type spaces of ul-
tradifferentiable functions and with the upper index {Mp } are the Roumieu-
type spaces of ultradifferentiable functions. Their strong duals are spaces
of Beurling and Roumieu-type ultradistributions, respectively.
∗ ∗
The space E ′ = E ′ (Ω) is the dual of
hp Mp
E (Mp ) = proj lim proj lim EK ;
K⊂⊂Ω h→0
H Mp
E {Mp } = proj lim ind lim EK p .
K⊂⊂Ω
Weighted ultradistribution spaces are defined by

(M ) (M ) Mα {M }
Mα
DL1 α (Rn ) = DL1 α = proj lim DL 1 ,h , DL1
α
= ind lim DL 1 ,h ,
h→∞ h→0
Mα
where DL 1 ,h , h > 0, is the Banach space of smooth functions ϕ on Rn with
finite norm
h|α|
kϕkL1 ,h = sup kϕ(α) kL1 .
α∈Nn
0
M|α|
D∗ = D∗ (Rn ) is dense in DL∗
1 , and the inclusion mapping is continuous.
∗
The strong dual of DL1 is denoted by B ′ .
∗
Spaces of tempered ultradistributions are defined as the strong duals of

the following testing function spaces:
S (Mα ) (Rn ) = S (Mα ) = proj lim ShMα ,
h→∞
S {Mα } (Rn ) ≡ S {Mα } = ind lim ShMα ,

h→0
where ShMα ,
h > 0, is the Banach space of smooth functions ϕ on R with
finite norm
h|α|+|β|
γh (ϕ) = sup k(1 + |x|2 )|α|/2 ϕ(β) kL∞ .
α,β∈Nn0
M |α| M |β|
The Fourier transform is an isomorphism of S ∗ onto S ∗ ; D∗ = D∗ (Rn ) is

dense in S ∗ , S ∗ is dense in DL
∗
1 and the inclusion mappings are continuous.
∗
The strong dual of S ∗ , S ′ , is the space of tempered ultradistributions
(of Beurling and Roumieu types). There holds D∗ ֒→ S ∗ ֒→ E ∗ , where ֒→
means that the left space is dense in the right one and that the inclusion
∗ ∗ ∗ ∗
mapping is continuous. Thus, E ′ ⊂ S ′ ⊂ D′ . We denote S ′ + = {f ∈
∗
S ′ (R); supp f ⊂ [0, ∞)}.
Let
∗
S[0,∞) = {ψ ∈ C ∞ [0, ∞); ψ = ϕ|[0,∞) for some ϕ ∈ S ∗ }
with the induced convergence structure from S ∗ ; its strong dual is in fact
∗
S ′+.
P
∞
An operator of the form: P (D) = aα Dα , aα ∈ C, α ∈ Nn0 , is
|α|=0
called ultradifferential operator of class (Mp ) (of class {Mp }) if there are
constants L > 0 and C > 0 (for every L > 0 there is C > 0) such that
|aα |≤ CL|α| /M|α| , |α|∈ N0 .
0.5.3. Fourier hyperfunctions. There are many equivalent definitions
of hyperfunctions, Laplace hyperfunctions and Fourier hyperfunctions (cf.
[71], [80], [144], [145], [204]), but we will use definitions and results collected
in [75]. Let I be a convex neighborhood of zero in Rn and let α be a non-
negative constant. A function F, holomorphic on Rn +iI, is said to decrease
exponentially with type (−α), α ≥ 0, if for every compact subset K ⊂⊂ I
and every ε > 0, there exists CK,ε > 0 such that
|F (z)|≤ CK,ε exp(−(α − ε)|Re z|), z ∈ Rn + iK . (0.7)
−α n n
The set of all such functions is denoted by Õ (D + iI), (Õ(D + iI) for
α = 0), where Dn denotes the directional compactification of Rn : Dn =
R n ∪ S∞n−1 n−1
(S∞ consists of all points at infinity in all direction). Space P∗
is defined by
P∗ = ind lim ind lim Õ−α (Dn + iI) . (0.8)
I∋0 α↓0
0. Preliminaries 7
The dual space of P∗ is the space Q(Dn ) of the Fourier hyperfunctions.

Q(Dn ) is a space of F S-type. We can give a representation of elements
from Q(Dn ). Let O be the sheaf of holomorphic functions. Denote
Uj = (Dn + iI) ∩ {Im zj 6= 0}, j = 1, . . . , n; (Dn + iI)#Dn = U1 ∩ · · · ∩ Un ,
(Dn + iI)#j Dn = U1 ∩ · · · ∩ Uj−1 ∩ Uj+1 ∩ · · · ∩ Un .

Then
n
X
Q(Dn ) = Õ((Dn + iI)#Dn )/ Õ((Dn + iI)#j Dn ) . (0.9)
j=1
Thus, f ∈ Q(Dn ) is defined as the class [F ], where F ∈ Õ((Dn + iI)#Dn ).

F is called a defining function of f and it is represented by 2n functions
Fσ , F = hFσ i, where Fσ ∈ Õ(Dn + iIσ ); Dn + iIσ = Dn + i(I ∩ Γσ ) is an
infinitesimal wedge of type Rn + iΓσ 0, Γσ are open σ-th orthants in Rn .
An f ∈ L1loc (Rn ) is called a function of infra-exponential type if for
every ε > 0, there exists Cε > 0 such that |f (x)|≤ Cε exp(ε|x|), x ∈ Rn .
Then by ℓf is denoted the hyperfunction defined by f.
We denote by Λ the set of n-th variations of {−1, 1}.
The boundary-value representation of f ∈ Q(Dn ) is:
X
f = [F ] := Fσ (x + iΓσ 0) . (0.10)
σ∈Λ
Fσ (x + iΓσ 0) denotes the element of the quotient space given in (0.9); it is

determined by Fσ .
The dual pairing between ϕ ∈ P∗ and g = [G] ∈ Q(Dn ) is given by
Z X Z
hg, ϕi = g(x)ϕ(x)dx = Gσ (w)ϕ(w)dw, w = u + iv,
Rn σ∈Λ Im w=v
σ
where vσ ∈ Iσ .
Similarly, Q−α (Dn ), α > 0, is defined using Õ−α instead of Õ (cf.
Definition 8.2.5 in [75]).
An infinite-order differential operator J(D) of the form
X p
J(D) = ba Da with lim |a| |ba |a! = 0, ba ∈ C ,
|a|→∞
|a|≥0
is called a local operator. J(D) maps continuously O(U ) into O(U ), U

being an open set in Cn , and also Q(Dn ) into Q(Dn ).
The Fourier transform on Q(Dn ) is defined by using the functions

χσ = χσ1 (z1 ) . . . χσn (zn ), where σk = ±1, k = 1, . . . , n, σ = (σ1 , . . . , σn )
and χ+ (t) = et /(1 + et ), χ− (t) = 1/(1 + et ), t ∈ R. Let
X X X
u(x) ∼ = Uσ (x + iΓσ 0) = (χσ̃ Uσ )(x + iΓσ 0) ,
σ∈Λ σ∈Λ σ̃∈Λ
n
where χσ̃ Uσ ∈ Q(D + iIσ ), σ̃ ∈ Λ and χσ̃ Uσ decreasing exponentially
along the real axis outside the closed σ̃-th orthant. The Fourier transform
of u is defined by
X X
F (u)(ξ) ∼
= F (χσ̃ Uσ )(ξ − iΓσ̃ 0)
σ∈Λ σ̃∈Λ
X X Z
= eizζ (χσ̃ Uσ )(z)dz, y σ ∈ Iσ , ζ = ξ + iη,
σ∈Λ σ̃∈Λ Imz=y σ
where F (χσ̃ Uσ ) ∈ Õ(Dn − iIσ̃ ) and F (χσ̃ Uσ )(z) = O(e−ω|x| ) for a suitable
ω > 0 along the real axis outside the closed σ-orthant. F is an automor-
phism of Q(Dn ).
0.6. Let E be a locally convex topological vector space. An absolute
convex closed absorbent subset of E is called a barrel. If every barrel is
a neighborhood of zero in E, then E is called barreled. Throughout this
book Fg stands for a locally convex barrelled complete Hausdorff space of
smooth functions (the subscript g stands for “general”), Fg ֒→ E = E(Rn ),
and Fg′ stands for the strong dual space of Fg ; observe E ′ ⊂ Fg′ . If T ∈ Fg′
and ϕ ∈ Fg , then hT, ϕi is the dual pairing between T and ϕ. We write F0
if all elements of Fg are compactly supported; F0′ denotes the dual space of
F0 ; observe that a notion of support for elements of F0′ can be defined in
the usual way. In Fg′ , a weakly bounded set is also a strongly bounded one
(Mackey–Banach–Steinhaus theorem). The spaces of distributions, ultra-
distributions, Fourier hyperfunctions, . . . , are of this kind. Furthermore, we
shall always use the notation F = Fg if A ֒→ F , where A = D, D∗ , or P∗ ;
in such case F ′ ⊂ A′ = D′ , D′∗ , or Q(Dn ), respectively, and we say that
F ′ is a distribution, ultradistribution, or Fourier hyperfunction space, re-
spectively. We set FΓ′ = {T ∈ F ′ ; supp T ⊂ Γ}.
We suppose that the following operations are well defined on Fg′ :
Differentiation: We assume that ∂/∂xi , Fg → Fg , are continuous oper-
ators. Let k ∈ Nn0 . Then,

∂k |k| ∂k
T (x), ϕ(x) = T (x), (−1) ϕ(x) , ϕ ∈ Fg .
∂xk11 . . . ∂xknn ∂xk11 . . . ∂xknn
0. Preliminaries 9
Change of variables: If T ∈ Fg′ and k > 0, then by definition

1 x
hT (kx), ϕ(x)i = T (x), n ϕ , ϕ ∈ Fg .
k k
If T ∈ Fg′ and h ∈ Rn ,
hT (x + h), ϕ(x)i = hT (x), ϕ(x − h)i, ϕ ∈ Fg .
Furthermore, it is always assumed that k → ϕ(·/k), R+ → Fg , and h →
ϕ(· + h), Rn → Fg , are continuous. Consequently, by the mean value
theorem, one readily verifies that both maps are indeed C ∞ .
Let ψ ∈ E, if ϕ 7→ ψϕ is a continuous mapping from Fg into Fg , then we
say that ψ is a multiplier of Fg′ . Then ψT is by definition hψT, ϕi = hT, ψϕi.
The set of multipliers of Fg′ is denoted by M(·) .
We shall say that θ ∈ E is a convolutor of Fg′ if the mapping ϕ 7→ θ̌ ∗
ϕ, Fg → Fg , is well defined and continuous, where θ̌(t) = θ(−t). We denote
by M(∗) the set of convolutors of Fg′ . If θ ∈ M(∗) , then for T ∈ Fg′ , (T ∗ θ)
is defined by hT ∗ θ, ψi = hT, θ̌ ∗ ϕi, ϕ ∈ Fg .
hδm im∈N , a sequence in M(∗) ∩ Fg , is called a δ-sequence in Fg if δm ≥
0, m ∈ N, and for every ϕ ∈ Fg , δm ∗ ϕ → ϕ in Fg , m → ∞.
We shall say that the convolution with compactly supported elements is
well defined in Fg′ if a notion of support makes sense in Fg′ and the following
definition applies: given S, T ∈ Fg′ , where supp S = K is compact in Rn ,
their convolution is defined by
hS ∗ T, ϕi = hS(x) × T (y), α(x)ϕ(x + y)i = hSx × Ty , α(x)ϕ(x + y)i ,
where the function α ∈ Fg has compact support, supp α = K̃, so that the
set K ⊂ int K̃ and α(x) = 1, x ∈ K. For F ′ it coincides with the usual
definition of convolution.
We assume that Fg′ contains regular elements f ∈ L1loc . They are iden-
tified with f itself:
Z
f : hf, ϕi = f (x)ϕ(x)dx, ϕ ∈ Fg ,
Rn
1
if f ϕ ∈ L for every ϕ ∈ Fg and if ϕm → 0 in F implies hf, ϕm i → 0.
There is also another situation in which we shall identify locally in-
tegrable functions with elements T ∈ F ′ . Let A = D, D∗ , or P∗ and
suppose A ֒→ F . We identify T with f ∈ L1loc if hT, ϕi = hf, ϕi,
for all ϕ ∈ A. In such a case we simply write T = f . For example,
2 2
T (x) = xex sin(ex ) ∈ S ′ (R) is defined in this way, and not as a regular
element of S ′ (R) in the sense described above.
1 S-asymptotics in Fg′
1.1 Definition
Definition 1.1. Let Γ be a cone with vertex at zero and let c be a positive
real-valued function defined on Γ. It is said that T ∈ Fg′ has S-asymptotic
behavior related to c with limit U if T (x + h)/c(h) converges weakly in Fg′
to U when h ∈ Γ, khk→ ∞, i.e. (w. lim = weak limit):
w. lim T (x + h)/c(h) = U in Fg′ (1.1)
h∈Γ,khk→∞
or
lim hT (x + h)/c(h), ϕ(x)i = hU, ϕi, ϕ ∈ Fg . (1.2)
h∈Γ,khk→∞
If (1.1) is satisfied, it is also said that T has S-asymptotics and we write in

s
short: T (x + h) ∼ c(h)U (x), h ∈ Γ .
Remarks. 1) If Γ is a convex cone we could use another limit in Γ.

Let h1 , h2 ∈ Γ. We say that h1 ≥ h2 if and only if h1 ∈ h2 + Γ; Γ is now
partially ordered.
For a real-valued function ρ defined on Γ, we write
lim ρ(h) = A ∈ R
h∈Γ, h→∞
if for any ε > 0 there exists h(ε) ∈ Γ such that ρ(h) ∈ (A − ε, A + ε) when
h ≥ h(ε), h ∈ Γ.
If Γ is a convex cone, then the S-asymptotics with respect to this limit
might be defined as:
lim hT (x + h)/c(h), ϕ(t)i = hu, ϕi, ϕ ∈ Fg . (1.3)
h∈Γ, h→∞
In case n = 1, the limits (1.2) and (1.3) coincide. We will mostly use
Definition 1.1 in this book, for S-asymptotics defined by (1.3) see also [135].
2) If Fg is a Montel space, then the strong and the weak topologies in
Fg′ are equivalent on a bounded set. If B is a filter with a countable basis
and if w. lim T (h) = U in Fg′ , then this limit exists in the sense of the
h∈B
strong topology. Hence, (1.1) is equivalent to
s. lim T (x + h)/c(h) = U in Fg′ .
h∈Γ,khk→∞
1. S-asymptotics in Fg′ 11
(Furthermore, from now on, we will omit the symbol s. for the strong
convergence).
For the first ideas of the S-asymptotics see [3] and [146]. The starting
point of the theory is [127].
1.2 Characterization of comparison functions and limits
Proposition 1.1. Let Γ be a convex cone. Suppose T ∈ Fg′ has the S-

s
asymptotics T (x + h) ∼ c(h)U (x), h ∈ Γ. If U 6= 0, then:
a) There exists a function d on Γ such that
lim c(h + h0 )/c(h) = d(h0 ) , for every h0 ∈ Γ . (1.4)

h∈Γ, khk→∞
b) The limit U satisfies the equation
U (· + h) = d(h)U, h ∈ Γ .
Proof. a) Since U 6= 0, there exists a ϕ̃ ∈ Fg such that hU, ϕ̃i 6= 0. For

this ϕ̃ and a fixed h0 ∈ Γ

c(h + h0 ) T (x + (h + h0 ))
lim , ϕ̃(x)
h∈Γ, khk→∞ c(h) c(h + h0 )
(1.5)
T ((x + h0 ) + h)
= lim , ϕ̃(x) .
h∈Γ, khk→∞ c(h)
Hence, for every h0 ∈ Γ

c(h + h0 ) hU (x + h0 ), ϕ̃(x)i
lim = = d(h0 ) .
h∈Γ, khk→∞ c(h) hU, ϕ̃i
b) Now we can take in (1.5) any function ϕ ∈ Fg instead of ϕ̃. Then we

have
d(h0 )hU, ϕi = hU (x + h0 ), ϕi, ϕ ∈ Fg
which proves b).

We now restrict the space of generalized functions to a space of distri-
bution, ultradistribution, or Fourier hyperfunction type. So, we have the
following explicit characterization of the comparison function and limit.
Proposition 1.2. Let Γ be a convex cone with int Γ 6= ∅ (int Γ is the in-
s
terior of Γ). Let T ∈ F ′ have S-asymptotics T (x + h) ∼ c(h)U (x), h ∈ Γ,
where U 6= 0 and c is a positive function defined on Rn . Then:
a) For every h0 ∈ Rn there exists
lim ˜ 0) .
c(h + h0 )/c(h) = d(h
h∈(h0 +Γ)∩Γ,khk→∞
˜ = exp(α · x), x ∈ Rn .
b) There exists α ∈ Rn such that d(x)
c) There exists C ∈ R such that U (x) = C exp(α · x).
Proof. a) Let a ∈ int Γ. Then there exists r > 0 such that B(a, r) ⊂ Γ.
Consequently, for every β > 0, B(βa, βr) ⊂ Γ, as well.
We shall prove that for every h0 ∈ Rn and every R > 0 the set (h0 +
Γ) ∩ Γ ∩ {x ∈ Rn ; kxk> R} is not empty. The first step is to prove that
(h0 + Γ) ∩ Γ is not empty.
Suppose that y ∈ B(a, r/2) ⊂ Γ. Then, for every β ≥ β0 > 2kh0 k/r > 0,
kβa − (h0 + βy)k≤ βka − yk+kh0k≤ βr ,
hence h0 + βy ∈ B(βa, βr) ⊂ Γ.
For a fixed R > 0, we can choose β such that kh0 + βyk> R. Then
h0 + βy is a common element for (h0 + Γ), Γ and {x ∈ Rn ; kxk> R}.
Now we can use the limit (1.5) when h ∈ (h0 + Γ) ∩ Γ, and in the same
way as in the proof of Proposition 1.1 a), we obtain
lim
c(h + h0 ) ˜ 0 ) = hU (t + h0 ), ϕ̃i .
= d(h
h∈(h0 +Γ)∩Γ, khk→∞ c(h) hU, ϕ̃i
From the existence of this limit, it follows:
1) d˜ extends d to the whole Rn ;
˜ = 1; d˜ ∈ C ∞ (Rn ) and d˜ satisfies
2) d(0)
˜ + h0 ) = d(h)d(h
d(h ˜ n
0 ), h, h0 ∈ R . (1.6)
n
We can take h0 = (0, . . . 0, ti , 0, . . . , ) ∈ R in (1.6); then the limit
˜ + h0 ) − d(h)
d(h ˜ ˜ 0 ) − d(0)
d(h ˜
lim ˜
= d(h) lim
ti →0 ti ti →0 ti
exists and gives

∂ ˜ ∂ ˜ ˜
d(h) = d(h) d(h), for i = 1, . . . , n .
∂hi ∂hi h=0
We introduce the function V given by

˜ (α·h) ∂ ˜
d(h) = e V (h), αi = d(h) , i = 1, . . . , n .
∂hi h=0
∂
Then V (h) = 0 for every i = 1, . . . , n. Consequently, V (h) = 1, h ∈ Rn .
∂hi
For the proof of c), we now have U (x + h) = exp(α · h)U (x). Differenti-
ating with respect to h and then setting h = 0, we obtain that U satisfies
∂
the differential equations U = αj , 1 ≤ j ≤ n. Proceeding as in the
∂xj
proof of b), we obtain U (x) = C exp(α · x), for some C ∈ R.
Remarks.
1. We only assumed that c is a positive function. But if we know that
s
there exist T ∈ F ′ g and U 6= 0 such that T (x + h) ∼ c(h)U (t), h ∈ Γ, then
we can find a function c̃ ∈ C ∞ and with the property
lim c̃(h)/c(h) = 1.
h∈Γ,khk→∞
This function c̃ can be defined as follows: c̃(h) = hT (x + h), ϕ̃(x)i/hU, ϕ̃i,

where ϕ̃ is chosen so that hU, ϕ̃i =
6 0. In this sense we can suppose, whenever
needed, that c ∈ C ∞ , and we do not loose generality.
Similarly, we have (see also Lemma 1.4 in 1.12) that
lim c(h)/c̃(h + x) = exp(−α · x) in E

h∈Γ′ ,khk→∞
if Γ is a convex cone, int Γ 6= ∅, Γ′ ⊂⊂ Γ (Γ′ is compact int Γ).

2. If in Proposition 1.2 we replace F ′ for the general space Fg′ , then
a) and b) still hold. On the other hand, c) will not be longer true, in
general. We will show this fact in Remark 5 below by constructing explicit
counterexamples.
3. In the one-dimensional case the cone Γ can be only R, R+ or R− .
In all these three cases int Γ 6= ∅. Consequently, d˜ from Proposition 1.2 has
˜ = exp(αx), where α ∈ R.
always the form d(x)
Let us write c(x) = L(ex ) exp(αx), x ∈ R. We will show that L is a
slowly varying function. Proposition 1.2 a) gives us the existence of the
limit
lim L(exp(h + h0 ))/L(exp(h)) = 1, h0 ∈ R .

h∈Γ,khk→∞
If Γ = R+ , then
lim L(xp)/L(x) = 1, p ∈ R+
x∈R+ ,x→∞
and this defines a slowly varying function (cf. 0.3). Thus if T ∈ Fg′ (R)
s
and T (x + h) ∼ c(h)U (x), in R+ , with U 6= 0, then it follows that c has
the form c(x) = exp(αx)L(exp(x)), x ≥ a > 0, where L is a slowly varying
function at infinity. Similarly, if Γ = R− , then L is slowly varying at the
origin, while if Γ = R, then L is slowly varying at both infinity and the
origin.
4. The explicit form of the function c given in 3. is not known in
the n-dimensional case, n ≥ 2. This problem is related to the extension
of the definition of a regularly varying function to the multi-dimensional
case ([135], [162]) and with certain q-admissible and q-strictly admissible
functions ([192]).
5. As mentioned before, c) in Proposition 1.2 does not have to hold for
a general space Fg′ . From the proof of Proposition 1.2, we can still obtain
the weaker conclusion U (x + h) = exp(α · h)U (x), h ∈ Rn , which in turn
implies the differential equations (∂/∂xj )U = αj . For distribution, ultra-
distribution, and Fourier hyperfunction spaces, these differential equations
imply that U must have the form c) of Proposition 1.2. However, the latter
fact is not true in general. We provide two related counterexamples below.
Let
A0 (R) = {ϕ ∈ C ∞ (R); lim ϕ(m) (x) exists and is finite, m ∈ N0 } ,
x→−∞
it is a Frechet space with seminorms:
βk (ϕ) = sup |ϕ(m) (x)| .

x∈(−∞,k],m≤k
Let us first observe that the definition of A0 (R) does not tell all the true
about its elements. Notice that if ϕ ∈ A0 (R), then for m = 1, 2, . . . , we
have limx→−∞ ϕ(m) (x) = 0, while limx→−∞ ϕ(x) may not be zero. Indeed,
the proof is easy, it is enough for m = 1, if limx→−∞ ϕ′ (x) = M , then
Z x
ϕ(x) = ϕ(0) + ϕ′ (t)dt ∼ ϕ(0) + M x, x → −∞ ,
0
but since ϕ has limit at −∞, then M = 0. So, we have

A0 (R) = {ϕ ∈ C ∞ (R); lim ϕ(x) exists and lim ϕ(m) (x) = 0, m ∈ N} .
x→−∞ x→−∞
Its dual space A′0 (R) contains a generalized function concentrated at −∞

which contradicts c) of Proposition 1.2. Define the Dirac delta concentrated
at −∞ by
hδ−∞ , ϕi := lim ϕ(x), ϕ ∈ A0 (R) .
x→−∞
Notice that the constant multiples of δ−∞ are the only elements of A′0 (R)
satisfying the differential equation U ′ = 0. This generalized function is
translation invariant, i.e., δ−∞ (· + h) = δ−∞ , for all h ∈ R; in particular,
it has the S-asymptotics
s
δ−∞ (x + h) ∼ δ−∞ (x), h ∈ Γ = R .
Therefore, we have found an example of a non-constant limit for S-
asymptotics related to the constant function c(h) = 1.
We can go beyond the previous example and give a counterexample
for the failure of conclusion c) of Proposition 1.2 with a general c and S-
asymptotics in Fg′ . Let c(h) = exp(αh)L(exp h), where L is slowly varying
at infinity and α ∈ R. We assume that c is C ∞ and, for all m ∈ N,
c(m) (h) ∼ αm c(h), h → ∞ (otherwise replace c by c̃ given in Remark 1).
Next, we define
ϕ(m) (x)
Ac (R) = {ϕ ∈ C ∞ (R); lim exists and is finite, m ∈ N0 } .
x→−∞ c(−x)
It is a Frechet space with seminorms:

|ϕ(m) (x)|
βk,c (ϕ) = sup .
x∈(−∞,k],m≤k c(−x)
Note that Ac (R) = c(−x) · A0 (R), consequently,

ϕ′ (x) ∼ −αCϕ c(−x), x → −∞ ,
where Cϕ = limx→−∞ ϕ(x)/c(−x). An inductive argument shows that
for all m, ϕ(m) (x) ∼ (−1)m αm Cϕ c(−x), x → −∞. Set now gc,−∞ =
(1/c(−x)) · δ−∞ ∈ A′c (R), a generalized function concentrated at infinity
and given by

ϕ(x) ϕ(x)
hgc,−∞ , ϕi = δ−∞ (x), = lim , ϕ ∈ Ac (R) .
c(−x) x→−∞ c(−x)
It is easy to show that the constant multiples of gc,−∞ are the only element
of A′c (R) satisfying the functional equation U (x + h) = exp(αh)U (x) (and
hence the differential equation U ′ = αU ); in particular,
s
gc,−∞ (x + h) ∼ eαh gc,−∞ (x), h ∈ Γ = R .
In addition, there is an infinite number of elements of A′c (R) having S-

asymptotics in the cone Γ = R+ related to c(h) with limit of the form
Cgc,−∞ (x), C ∈ R. In fact, consider δ (m) , the derivatives of the Dirac
delta concentrated at the origin. We have that
s
δ (m) (x + h) ∼ c(h)αm gc,−∞ (x), h ∈ Γ = R+ ,
since

(m)
δ (x + h), ϕ(x) ϕ(m) (−h)
lim = lim (−1)m = αm hgc,−∞ (x), ϕ(x)i .
h→∞ c(h) h→∞ c(h)
1.3 Equivalent definitions of the S-asymptotics in F ′
Theorem 1.1. Let T ∈ F ′ and let int Γ 6= ∅. The following assertions are
equivalent:
a) w. lim T (x + h)
= U (x) = M exp(αx) in F ′ , M 6= 0 . (1.7)
h∈Γ,khk→∞ c(h)
b) For a δ-sequence hδm im (cf. 0.6) there exists a sequence hMm im in

R, such that Mm → M 6= 0, m → ∞, and
(T ∗ δm )(x + h)
w. lim = Mm exp(αx), in F ′ , uniformly in m ∈ N .
(1.8)
c) For a δ-sequence hδm im , (cf. 0.6),
(T ∗ δm )(h)
lim = pm , m ∈ N , (1.9)
where pm 6= 0 for some m, and for every φ ∈ F ,

(T ∗ φ)(h)
sup < ∞. (1.10)
h∈Γ,khk≥0 c(h)
′
If F ′ = D′ or F ′ = D ∗
the following assertion is also equivalent to a).
d) (T (· + h)/c(h)) ∗ ϕ converges to (U ∗ ϕ)(h) in E(Rn ), for each ϕ ∈
D (ϕ ∈ D∗ ).
Proof. a) ⇒ b). Let hδm im be a δ-sequence. For any φ ∈ F , {δm ∗ φ;

m ∈ N} is a compact set in F . We have by the properties of the convolution
(cf. 0.6) and the Banach–Steinhaus theorem

(T ∗ δm )(x + h) h(T ∗ δm )(x), φ(x − h)i
lim , φ(x) = lim
h∈Γ,khk→∞ c(h) h∈Γ,khk→∞ c(h)

hT, (δ̌m ∗ φ)(· − h)i T (x + h)
= lim = lim , (δ̌m ∗ φ)(x)
= hM eα·x , (δ̌m ∗ φ)(x)i = hMm eα·x , φ(x)i , (1.11)

uniformly in m. Now (1.11) implies (1.8) and b).
b) ⇒ a). Let φ ∈ F and

(T ∗ δm )(x + h) (T ∗ (δ̌m ∗ φ))(h)
am,h = , φ(x) = , m ∈ N, h ∈ Γ .
c(h) c(h)
We have am,h → am , h ∈ Γ, khk→ ∞, uniformly for m ∈ N, where
am = Mm hexp(α · x), φ(x)i, m ∈ N,
am → a = M hexp(α · x), φ(x)i, m → ∞ .

Also am,h → ah , m → ∞, where

T (x + h)
ah = , φ(x) , h ∈ Γ .
c(h)
This implies ah → a, h ∈ Γ, khk→ ∞, what is in fact a).
a) ⇒ c). From (1.7) it follows that (T ∗ ϕ)(h)/c(h) converges for every
ϕ ∈ F , when h ∈ Γ, khk→ ∞. Hence, (T ∗ ϕ)(h)/c(h) is bounded, h ∈
Γ, khk≥ 0; (1.9) follows directly from (1.7).
c) ⇒ a). First, we shall prove that the set G = {δm (· + x), m ∈ N, x ∈
R } is dense in F . Suppose that T ∈ F ′ and that
n
hT, δm (· + x)i = 0, m ∈ N, x ∈ Rn .
It follows that (T ∗ δ̌m )(−x) = 0, m ∈ N, x ∈ Rn . Then, for any ϕ ∈
F , hT ∗ δm , ϕi = 0, m ∈ N, and consequently
hT, ϕi = lim hT, δm ∗ ϕi = lim hT ∗ δ̌m , ϕi = 0 .
m→∞ m→∞
This implies that T = 0 and hence, the set G is dense in D by the Hahn-
Banach theorem. Thus, by (1.10) and the Banach–Steinhaus theorem, c)
implies a).
a) ⇒ d). Note that (1.1) implies the strong convergence of T ( · + h)/c(h)
∗ ′
to U in D′ and D′ respectively. Since the convolution in D′ and D ∗ is
hypocontinuous, it follows the following equality in the sense of convergence
in E(Rn )

T ( · + h) T ( · + h)
lim ∗ϕ = lim ∗ϕ =U ∗ϕ
in both cases (ϕ ∈ D, respectively ϕ ∈ D∗ ).
The paper which can be consulted for related results is [128].
1.4 Basic properties of the S-asymptotics
Theorem 1.2. Let T ∈ Fg′ .

s s
a) If T (x + h) ∼ c(h)U (x), h ∈ Γ, then for every k ∈ Nn0 , T (k) (x + h) ∼
c(h)U (k) (x), h ∈ Γ.
b) Assume additionally that Fg is a Montel space. Let g ∈ M(·) (set
of multipliers of Fg′ (cf. 0.6)); let c, c1 be positive functions. If for every
ϕ ∈ Fg , (g(x + h)/c1 (h))ϕ(x) converges to G(x)ϕ(x) in Fg when h ∈
s s
Γ, khk→ ∞ and if T (x + h) ∼ c(h)U (x), h ∈ Γ, then g(x + h)T (x + h) ∼
c1 (h)c(h)G(x)U (x), h ∈ Γ.
s
c) If T ∈ F0′ and supp T is compact, then T (x + h) ∼ c(h) · 0, h ∈ Γ,
for every positive function c.
d) Suppose that Fg′ is a Montel spaces in which the convolution with
compactly supported elements is well defined and hypocontinuous (cf. 0.6).
s
Let S ∈ Fg′ , supp S being compact. If T (x + h) ∼ c(h)U (x), h ∈ Γ, then
s
(S ∗ T )(x + h) ∼ c(h)(S ∗ U )(x), h ∈ Γ.
′ s ′
e) Let T ∈ F ′ = D ∗ and T (x + h) ∼ c(h) · U (x), h ∈ Γ, in D ∗ . Assume
that (M.2) holds. Let P (D) be an ultradifferential operator of class ∗. Then
s ′
(P (D)T ) (x + h) ∼ c(h) · (P (D)U )(x), h ∈ Γ, in D ∗ .
f ) Let T ∈ F ′ = Q(Dn ). Let P (D) be a local operator and let T (x +

s s
h) ∼ c(h) · U (x), h ∈ Γ, khk→ ∞ in Q(Dn ), then (P (D)T )(x + h) ∼
c(h) · (P (D)U )(x), h ∈ Γ, khk→ ∞ in Q(Dn ), as well.
Proof. a) The assertion is a consequence of the definition of the deriva-

tive of a generalized function. Namely,

T (k) (x + h) T (x + h)
lim , ϕ(x) = lim , (−1)k ϕ(k) (x)
= (−1)k hU (x), ϕ(k) (x)i
= hU (k) , ϕi, ϕ ∈ Fg .
T (x + h)
b) Since Fg is a Montel space, we have that lim
= U (x)
in Fg′ with respect to strong topology. As (g(x + h)/c1 (h) − G(x))ϕ(x), h ∈
Γ, khk≥ 0, is a bounded set in Fg , it follows:
lim hg(x + h)T (x + h)/(c1 (h)c(h)), ϕ(x)i

h∈Γ,khk→∞

g(x + h)
= lim T (x + h)/c(h), − G(x) ϕ(x)
h∈Γ,khk→∞ c1 (h)
+ lim hT (x + h)/c(h), G(x)ϕ(x)i

h∈Γ,khk→∞

g(x + h)
= U (x), lim − G(x) ϕ(x) + hU (x), G(x)ϕ(x)i
= 0 + hU (x)G(x), ϕ(x)i = hU G, ϕi, ϕ ∈ Fg .
c) For each ϕ ∈ F0 there exists rϕ > 0 such that supp ϕ ⊂ B(0, rϕ ) =

{x ∈ Rn ; kxk< rϕ }. The support of T (x + h) is (supp T − h). Thus, by
our assumption, there exists βrϕ such that for all h ∈ Γ, khk> βrϕ the set
(supp T − h) ∩ B(0, rϕ ) is empty and consequently hT (x + h), ϕ(x)i = 0, h ∈
Γ, khk≥ βrϕ .
d) By definition of the convolution
h(S ∗ T )(x + h), ϕ(x)i = h(S ∗ T )(x), ϕ(x − h)i

= hSt × Ty , α(t)ϕ(t + y − h)i (1.12)
= hSt × Ty (· + h), α(t)ϕ(t + y)i .
Hence, (S ∗ T )(x + h) = (S ∗ T (· + h))(x).

Since in a Montel space the weak and the strong convergence are equiv-
alent, we have by (1.12)

T (· + h)
lim (S ∗T )(x+h)/c(h) = lim S∗ (x) = (S ∗U )(x) .
h∈Γ,khk→∞ h∈Γ,khk→∞ c(h)
For the proof of e) and f), we note that an ultradifferential operator

′ ′
P (D) maps continuously D ∗ into D ∗ and a local operator maps continu-
ously Q(Dn ) into Q(Dn ).
Remark. From assertion a) of Theorem 1.2, a natural question arises

for spaces F ′ : The limit U can be a constant generalized function, hence
U ′ = 0. Is there a positive function c̃ such that T ′ has S-asymptotics related
to this c̃, but with a limit different from zero?
In general the answer is negative as shown by the following example: Let
s
T be defined by x2 + sin(exp x2 ), x ∈ R. Then, T (x + h) ∼ h2 · 1, h ∈ R+ .
But T ′ (x) = 2x(1 + exp(x2 ) cos(exp x2 )). The same situation is obtained
with the distribution f (x) = x2 + x sin x.
We can now formulate an open problem: Suppose that T (x +

s
h) ∼ c(h)U (x), h ∈ Γ. If U (x) is a constant generalized function, then
the problem is to find some additional conditions on T which guarantee the
existence of a function c̃ such that T ′ has the S-asymptotics in Γ related to
c̃.
s
More generally, let S ∈ F ′ and T = (∂/∂xk )S. If T (x + h) ∼
c(h)U (x), h ∈ Γ. The question is what we can say about the S-asymptotics
of S.
Recall the well known result: Let h be a real-valued function which
has the first derivative h′ (x) 6= 0, x ≥ x0 and h(x) → ∞, x → ∞.
If a function F has its first derivative on (x0 , ∞) such that there exists
lim F ′ (x)/h′ (x) = A, then there exists lim F (x)/h(x) = A, as well.
x→∞ x→∞
We know that an opposite assertion does not hold, and this is at the
basis of the open problem quoted above.
We give a theorem to illustrate the relation between the S-asymptotics
of a distribution and the S-asymptotics of its primitive. We refer to [114]
for the proof.
Theorem 1.3. 1) Let f, g ∈ D′ (R) and for some m ∈ N, g (m) = f.

s s
a) If f (x + h) ∼ hν L(h) · 1, h ∈ R+ , where ν > −1, then g(x + h) ∼
ν+m
h L(h) · 1, h ∈ R+ .
s
b) If f (x + h) ∼ exp(αh)L(exp h) exp(αx), h ∈ R+ , α ∈ R, and
Rx
exp(αh)L(exp h)dh → ∞, when x → ∞, then
0
 h hm−1 
Z Z Zh1
s
g(x+h) ∼  . . . exp(αt)L(exp t)dtdh1 . . . dhm−1  exp(αx), h ∈ R+ .
0 0 0
R
2) Let φ0 ∈ D′ (R) such that φ0 (t)dt = 1. If

g (i) (x + h)
lim , φ0 (x) = αi hexp(αx), φ0 (x)i, i = 0, 1, . . . , m − 1
h→∞ exp(αh)L(exp h)
and
s
f (x + h) ∼ exp(αh)L(exp h)αm exp(αx), h ∈ R+ ,
then
s
g(x + h) ∼ exp(αh)L(exp h) exp(αx), h ∈ R+ .
3) Suppose that T ∈ D′ , Γ = {x ∈ Rn ; x = (0, . . . , xk , 0, . . . , 0)} and

s
T = (∂/∂xk )S. If T (x + h) ∼ c(h)U (x), h ∈ Γ and c(h) is locally integrable
in hk such that
Zhk
c1 (hk ) = c(v)dvk → ∞ as hk → ∞, h0k ≥ 0,
h0k
s
then S(x + h) ∼ c1 (h)U (x), h ∈ Γ.
4) Suppose that S ∈ D′ and that for an m ∈ {1, 2, . . . , n},
s
(Dtm S)(x + h) ∼ c(h) · U (x), h ∈ Γ .
R
Let V ∈ D′ , Dtm V = U and φ0 ∈ D(R), φ0 (τ )dτ = 1. Let
R
lim hS(x + h)/c(h), φ0 (xm )λm (x̃)i = hV, φ0 λm i ,

h∈Γ,khk→∞
where x̃ = (x1 , . . . , xm−1 , xm+1 , . . . , xn ) and

Z
λm (x̃) = ψ(x1 , . . . , xm , . . . , xn )dxm , ψ ∈ D .
R
s
Then S(x + h) ∼ c(h)V (x), h ∈ Γ.
The following theorem asserts that the S-asymptotics is a local property

if the elements of Fg are compactly supported.
Theorem 1.4. Let T1 , T2 ∈ F0′ . Let the open set Ω ⊂ Rn have the following
property: for every r > 0 there exists a βr > 0 such that the ball B(0, r) =
{x ∈ Rn ; kxk< r} is in {Ω − h; h ∈ Γ, khk≥ βr }. If T1 = T2 on Ω and
s s
T1 (x + h) ∼ c(h)U (x), h ∈ Γ, then T2 (x + h) ∼ c(h)U (x), h ∈ Γ, as well.
Proof. Let ϕ ∈ F0 with supp ϕ ⊂ B(0, r). We shall prove that

T1 (x + h) − T2 (x + h)
lim , ϕ(x) = 0 . (1.13)
The complement of the set supp(T1 (x + h) − T2 (x + h)) contains the set
{Ω − h, h ∈ Γ}. By our supposition the number βr is fixed in such a way
that the sets {Ω − h; h ∈ Γ, khk≥ βr } contain B(0, r) and consequently
supp ϕ. Since T1 has the S-asymptotics related to c and with the limit U,
(1.13) implies
hT2 (x + h), ϕ(x)i hT1 (x + h), ϕ(x)i
lim = lim = hU, ϕi .

0
Remark. The open set Ω, by its property, has to contain a set Γ ∩{x ∈
Rn ; kxk> R} where Γ0 is an open acute cone such that Γ ⊂ Γ0 and R is a
positive number.
1.5 S-asymptotic behavior of some special classes of

generalized functions
1.5.1 Examples with regular distributions

s
1. exp(a · (x + h)) ∼ exp(a · h) exp(a · x), h ∈ Rn .

p s 1
2
2. exp( (x + h) + (x + h)) ∼ exp h exp x + , h ∈ R+ .
2
n
3. Let w ∈ S n−1 = {x ∈ Rn ; kxk= 1}, p = (p1 , . . . , pn ) ∈ R+ and
P
Γ = {qw; q ∈ R+ }. Denote, J = {k ∈ {1, . . . , n}; wk 6= 0} and β = pi .
i∈J
Then
s
Y
(x + h)p ∼ q β wipi · 1, h ∈ Γ ,
i∈J
and
s
Y
(1 + x + h)−p ∼ q −β wi−pi · 1, h ∈ Γ ,
i∈J
((x + h)p = (x1 + h1 )p1 . . . (xn + hn )pn ) .
4. For a slowly varying function L(t), t ≥ α > 0 we have

s
L(t + h) ∼ L(h) · 1, h ∈ R+ .
Namely,
Zr
lim hL(t + h)/L(h), ϕ(t)i = lim ϕ(t)L(t + h)/L(h)dt
h→∞ h→∞
−r
r
Ze Z
dy
= lim ϕ(log y)L(log(yq))/L(log q) = ϕ(t)dt, ϕ ∈ D(R) .
q→∞ y
e−r R
We used above that L(log y) is also a slowly varying function (cf. 0.3)
and that L(log(uh))/L(log h) converges uniformly to 1 as h → ∞ if u stays
in compact intervals [α1 , α2 ], 0 < α1 < α2 < ∞.
5. Let g ∈ L1 (R). A distribution defined by g has the S-asymptotic behavior
related to c = 1 and with limit U = 0. To show this, let α ∈ R and
Zt
T (t) = g(x)dx, t ∈ R.
α
s R∞ s
Then, T (t + h) ∼ 1 · g(x)dx, h ∈ R+ . By Theorem 1.2 a), g(t + h) ∼
α
1 · 0, h ∈ R+ .
1.5.2 Examples with distributions in subspaces of D′

′
6. Let fα , α ∈ R, be given as in 0.4. It defines a distribution in D[0,∞) .
For α = −k, k = 0, 1, . . . , f−k = δ (k) . Since δ (k) is supported by {0},
it has the S-asymptotics zero related to every c > 0 (cf. Theorem 1.2 c)).
For α > 0,
s
fα (x + h) ∼ (1/Γ(α))hα−1 · 1, h ∈ R+ .
In the case α < 0, α 6= −1, −2, . . . , fα (x) = (1/Γ(α)) · Pf(H(x)xα−1 ),

where Pf is the finite part or “partie finie” (see [146], pp. 41–43).
Note, supp φ(x − h) = supp φ + h, φ ∈ D(R). Thus for φ ∈ D(R), we
can find h0 such that supp φ(x − h) ⊂ (a, ∞), a > 0, h ≥ h0 . Then
Z∞ α−1
α−1 1 x
lim hfα (x)/h , φ(x − h)i = lim φ(x − h)dx
h→∞ h→∞ Γ(α) h
a
Z α−1 Z
1 u+h 1
= lim · φ(u)du = φ(u)du .
h→∞ Γ(α) h Γ(α)
R R
s 1
Hence, fα (x + h) ∼ hα−1 · , h ∈ R+ , α < 0, α 6= −1, −2, . . .
Γ(α)
7. If T ∈ S ′ , then there exists a real number k0 such that T has S-
asymptotic behavior related to c(h)khkk0 , where c(h) tends to infinity as
|h|→ ∞, h ∈ Rn and with limit U = 0.
By Theorem VI, chapter VII in [146], there exists a number k0 such
that the set of distributions {T (· + h)/(1 + khk2 )k0 /2 ; h ∈ Rn } is bounded
in D′ . Hence, this set is weakly bounded and
D E (1 + khk2 )k0 /2 T (x + h)
k0
T (x + h)/(c(h)khk , ϕ(x) = , ϕ(x)
c(h)khkk0 (1 + khk2 )k0 /2
tends to zero as khk→ ∞.
8. Let T ∈ D′ (R) have the following property:
For a δ-sequence hδm im there is a sequence hpm im in R, such that
pm → p 6= 0, m → ∞, and
(T ∗ δm )(h)
lim = pm , m ∈ N, (∗)
h→∞ c(h)
where the limit is uniform for m ∈ N.
Then T has S-asymptotics related to c.
We will prove that b) in Theorem 1.1 is satisfied, which is equivalent to
a) in the same theorem. For every compact set K ⊂ R
(T ∗ δm )(x + h) (T ∗ δm )(x + h) c(x + h)
= → pm exp(αx), h ∈ Γ, khk→ ∞ ,
c(h) c(x + h) c(h)
c(x + h)
uniformly for x ∈ K, because of → exp(αx), khk→ ∞, uniformly
c(h)
for x ∈ K (cf. Remarks after Proposition 1.2). Then, T has the S-
asymptotics related to c.
Property (*) is not equivalent to the existence of the S-asymptotics.

The next example illustrates that this condition is not necessary.
Assume that S ∈ C(R) ∩ L1 (R) but not being bounded on R. This
function has S-asymptotics equal to zero related to c = 1 (see 5.). For T
s
we take 1 + S(x). Then, T (x + h) ∼ 1 · 1, h ∈ R+ and
lim [(1 + S) ∗ δm ](h) = lim h1 + S(x + h), δ̌m (x)i = h1, δ̌m (x)i = pm .
h→∞ h→∞
This limit is not uniform in m ∈ N because lim [(1 + S) ∗ δm ](h) does not
m→∞
necessarily exist.
9. Every distribution in D′ Lp , 1 ≤ p < ∞ has S-asymptotic behavior related

to c = 1, and Γ = Rn , with limit U = 0. Let us show this.
By Theorem XXV, Chapter VI in [146] it follows that (T ∗ ϕ) ∈ Lp (Rn )
∂
for every ϕ ∈ D. Every derivative of (T ∗ ϕ), (T ∗ ϕ)(h) = (T ∗
∂hk
∂
ϕ)(h), h ∈ Rn , is also in Lp (Rn ). Hence (T ∗ ϕ̌) ∈ DLp . We know
∂xk
that every element of DLp , 1 ≤ p < ∞ is bounded over Rn and tends to
zero when khk→ ∞ ([146], p. 199).
It has been examined in [153] how slowly the function (T ∗ϕ) ∈ DLp , ϕ ∈
D, tends to zero as khk→ ∞.
In fact, the question is:
Let p ≥ 1. Is it possible to find a positive function c such that c(x) →
0, kxk→ ∞ and |ϕ(x)/c(x)|≤ Cϕ , kxk≥ Rϕ , for every ϕ ∈ DLp ? Cϕ and
Rϕ are positive constants depending on ϕ.
The answer is negative (see [153]).
A similar question can be asked for T ∈ D′ Lp , 1 ≤ p < ∞, but related to
the S-asymptotics. Precisely, whether there exists c(h) > 0, c(h) → 0, khk→
∞, such that
|hT (x + h)/c(h), φ(x)i|≤ Mφ , khk> βφ φ ∈ D,
where βφ and Mφ are positive constants depending on φ.

The answer to this question is also negative (cf. [153]).
10. Let T ∈ K′ p . Then there exists k0 ∈ N0 such that T has S-asymptotic

behavior with limit U = 0 related to c(h) exp(k0 khkp ), where c(h) tends to
infinity as khk→ ∞.
First, we prove that there exists a positive integer k, such that the set
{T (· + h) exp(−kkhkp ), h ∈ Rn } is bounded in D′ .
We start by giving a bound for the seminorms νk (ϕ(· − h)), ϕ ∈ Kp :
νk [ϕ(· − h)] = sup exp(kkxkp )|Da ϕ(x − h)|

x∈Rn ,|a|≤k
= sup exp(kkx + hkp )|Da ϕ(x)|

x∈Rn ,|a|≤k
≤ exp(2p kkhkp ) sup exp(2p kkxkp )|Da ϕ(x)|

x∈Rn ,|a|≤2pk
≤ exp(2p kkhkp )ν2p k (ϕ).
By assumption, T is a continuous linear functional on Kp . Note, the

sequence of norms hνk ik is increasing. Thus, there exist ε > 0 and k0 ∈ N0
such that
|hT, ϕi|≤ 1 for ϕ ∈ Kp , νk0 (ϕ) ≤ ε .
This inequality holds for all k ≥ k0 . Hence
|hT, ϕi|≤ ε−1 νk (ϕ), k ≥ k0 for every ϕ ∈ Kp .
We know that D ⊂ Kp and that the inclusion is continuous. Let ϕ ∈ D,

Then
|hexp(−2p kkhkp )T (x + h), ϕ(x)i|= |hT (x), exp(−2p kkhkp )ϕ(x − h)i|
≤ ε−1 exp(−2p kkhkp )νk [ϕ(x − h)] ≤ ε−1 ν2p k (ϕ), k > k0 .
We can choose k0 ≥ 2p k. The set {exp(−k0 khkp )T (x + h); h ∈ Rn } is

bounded in D′ (and weakly bounded in D′ , as well).
Now, for every ϕ ∈ D :
lim hexp(−k0 khkp )T (x + h)/c(h), ϕ(x)i

khk→∞
1
= lim hexp(−k0 khkp )T (x + h), ϕ(x)i = 0.
khk→∞ c(h)
1.5.3 S-asymptotics of ultradistributions and Fourier hyperfunc-

tions — Comparisons with the S-asymptotics of distributions
11. If a distribution has S-asymptotics in D′ (see Definition 1.1), it has the

∗
same S-asymptotics in D′ , as well. But the opposite is not true. This is
illustrated by the following example:
∞
X
T (x) = 1 + δ (n) (x − n)/Mn , x ∈ R,
n=1
′ (Mp )
has S-asymptotics in D (R), but it does not have S-asymptotics in
D′ (R). We will show this. Let φ ∈ D(Mp ) (R). Then,
* ∞ + ∞
X X
δ (n) (x+h−n)/Mn, φ(x) = (−1)n φ(n) (n−h)/Mn → 0, h → ∞ .
n=1 n=1
This is a consequence of the property

sup|φ(n) (x)|/k n Mn → 0, n → ∞ for every k > 0 .
x∈R
Suppose that there exists a function c such that for every ϕ ∈ D(R)
∞
X
(−1)n ϕ(n) (n − h)/Mn c(h) ,
n=1
converges, as h → ∞. Taking h = n this implies that ϕ(n) (0)/Mn c(n)

converges to zero, as n → ∞, for every ϕ ∈ D(R). However, such a func-
tion c does not exist (Borel’s theorem). Namely, for every given sequence
Mn c(n), n ∈ N, there exists ϕ ∈ C0∞ (R) such that ϕ(n) (0)/Mn c(n) → ∞,
as n → ∞.
This example is the motivation for the following assertion.
∗
12. Let T ∈ D′ (R) be such that lim T (x + h)/c(h) exists in D′ (R).
h→∞
∗
Assume that for some s ∈ N and ω ∈ D (R) with the property
Z Z
ω(t)dt = 1, tj ω(t)dt = 0, j = 1, . . . , s,
R R
the following limit

T (t + h) ts t
lim , s+1 ω , p ∈ (0, 1],
h→∞ c(h) p p
exists uniformly in p. Then, lim T (x + h)/c(h) exists in D′ (R) as well.
h→∞
Proof. Let φ ∈ D(R), h ≥ h0 > 0. The function Fh (t) = hT (x + h +

t), φ(x)i, t ∈ R, is smooth and by the Taylor formula, we have:
d ts−1 ds−1
Fh (t) = Fh (0) + t Fh (0) + · · · + Fh (0)
dt (s − 1)! dts−1
Z1
ts (s)
+ (1 − p)s−1 Fh (pt)dp, s ≥ 1, 0 < p ≤ 1 .
(s − 1)!
0
This implies
hT (x + h), φ(x)i = hT (x + h + t), φ(x)i − hT (x + h), φ′ (x)i
(−1)s−1 ts−1
−···− hT (x + h), φ(s−1) (x)i
(s − 1)!
Z1
(−1)s ts
− (1 − p)s−1 hT (x + h + pt), φ(s) (x)idp,
(s − 1)!
0
h ≥ h0 , t ∈ R .
Multiplying both sides of the last equality by ω(t), integrating with respect
to t, and using Fubinni’s theorem, we obtain
(−1)s
hT (x + h), φ(x)i = hhT (x + h + t), ω(t)i, φ(x)i − · · · −
(s − 1)!
Z1
(1 − p)s−1 hhT (x + h + pt), ts ω(t)i, φ(s) (x)idp .
0
Set
G(x, h, p) = hT (x + h + pt), ts ω(t)i, x ∈ R, h ≥ h0 , p ∈ [0, 1] .
We have G(x, h, 0) = 0, x ∈ R, h ≥ h0 and
lim G(x, h, p)/c(h) = Ceax heapt , ts ω(t)i, x ∈ R, p ∈ (0, 1] ,
h→∞
for some a ∈ R (cf. Proposition 1.2 a)), where the limit is uniform in
p ∈ (0, 1] and x ∈ supp φ. The limit function is continuous in x ∈ R and
p ∈ [0, 1], because of heapt , ts ω(t)i → 0, as p → 0. Because of that, we
obtain
1
lim hhT (x + h + pt), ts ω(t)i, φ(s) (x)i
h→∞ c(h)

1
= lim hT (x + h + pt), ts ω(t)i, φ(s) (x)
h→∞ c(h)
= hhCeax+apt , ts ω(t)i, φ(s) (x)i = Cheax , φ(s) (x)iheapt , ts ω(t)i .

This implies

1 T (x + h + t)
lim hT (x + h), φ(x)i = lim , ω(t) , φ(x)
h→∞ c(h) h→∞ c(h)
Z1
(−1)s
−···− (1 − p)s−1 heapt , ts ω(t)idphCeax , φ(s) (x)i ,
(s − 1)!
0
which proves the assertion.

Such an assertion can be proved in the multi-dimensional case by ad-
justing the previous argument. But an open problem is to find necessary
and sufficient conditions for a distribution T, which has S-asymptotics in
∗
D′ , to have the S-asymptotics in D′ as well.
13. We shall construct an ultradistribution out of the space of Schwartz
distributions and having S-asymptotics.
Assume (M.2) holds. Let P (D) be an ultradifferential operator of class
∗ of infinite order (aα 6= 0 for infinitely α) (see 0.5.2). Then, P (D)δ is an
∗
element of D′ which is not a distribution and which has S-asymptotics in
∗
D′ equal to zero related to any c.
s
If T ∈ D′ and T (x + h) ∼ 1 · 1, h ∈ Γ in D′ , the ultradistribution
s
T + P (D)δ is not a distribution, but (T + P (D)δ)(x + h) ∼ 1 · 1, h ∈ Γ in
′∗
D :
lim (h(T (x + h), φ(x)i + hP (D)δ(x + h), φ(x)i)

h∈Γ,khk→∞
X
= h1, φi + lim hδ(x + h), (−1)m am φ(m) (x)i = h1, φi, φ ∈ D∗ .
h∈Γ,khk→∞
P α α α
14. Let P (D) be a local operator |α|≥0 b D , b 6= 0, |α|≥ 0 (see
0.5.3.). The Fourier hyperfunction f = 1 + P (D)δ has S-asymptotics
related to c = 1 in any cone Γ and with limit U = 1, but f is not a
distribution. For the S-asymptotics of f it is enough to prove that
lim hP (D)δ(x + h), ϕ(x)i = 0, ϕ ∈ P∗ .

h∈Γ,khk→∞
Since P (D) maps P∗ into P∗ ,
hP (D)δ(x + h), ϕ(x)i = hδ(x + h), P (−D)ϕ(x)i = ψ(h) ,

where ψ = P (−D)ϕ. By the property of the elements of P∗ (see 0.5.3), we

have lim ψ(h) = 0, for every cone Γ.
h∈Γ,khk→∞
A hyperfunction g supported by the origin is uniquely expressible as

g = P̃ (D)δ, where P̃ (D) is a local operator. In such a way, the above
proof implies that every Fourier hyperfunction with support at {0} has
S-asymptotics with limit equal zero.
P
Since P (D)δ = bα Dα δ is a distribution if and only if bα 6= 0 for a
|α|≥0
finite number of α, we note that 1 + P (D)δ is not a distribution, but it has
the S-asymptotics related to c = 1.
We can also find coefficients bα of a local operator P (D) such that
f = 1+P (D)δ is not defined by an ultradistribution belonging to the Gevrey
′ ′
class D(s) or D{s} , s > 1 (see [81]). For the sake of simplicity, we shall
consider the one-dimensional case. Choose P (D) such that the coefficients
of P (D) are: bn = (n! )−(1+cn ) , n ∈ N, where cn = (log log n)−1 . With
these coefficients, P (D) is a local operator. Namely,
p 1
lim n bn n! = lim (n! )− n log log n = 0 .
n→∞ n→∞
Also, any ultradistribution in the Gevrey class s > 1, supported by {0},

is of the form
X∞
J(D)δ = an Dn δ, necessarily with |an |≤ Ck n /(n! )s
n=0
for some constants k and C (Beurling’s type) or for any k > 0 with a con-
stant C (Roumieu’s type). But the coefficients bn = (n! )−(1+cn ) do not
satisfy these conditions, therefore P (D)δ cannot represent an ultradistri-
bution. Namely, since cn → 0 when n → ∞, for any s > 1, there exists n0
such that 1 + cn < s, n ≥ n0 . Thus,
(n! )−(1+cn ) > Ck n /(n! )s , n ≥ n0 , k > 0 .
Consequently, P (D)δ does not represent an ultradistribution of Gevrey
type.
On the other hand, if we suppose that g = P (D)δ is an ultradistribu-
tion with support {0} in the Gevrey class s > 1, then, we would have an
ultradifferential operator J1 (D) such that
∞
X
g = J1 (D)δ = en Dn δ, |en |≤ Ck n /(n! )s .
n=0
But in this case J1 (D) is also a local operator and J1 (D) 6= P (D). This
contradicts the fact that a hyperfunction with support at {0} is given by a
unique local operator.
1.6 S-asymptotics and the asymptotics of a function
We suppose in this subsection that elements of F are compactly supported

and, as usual, that the topology in F is stronger than the topology in E.
Recall, we use the notation F0 in this case.
Every locally integrable function f defines an element of F0′ (regular
generalized functions).
We shall compare the asymptotic behavior of a locally integrable func-
tion f and the S-asymptotic behavior of the generalized function generated
by it.
A function f has asymptotics at infinity if there exists a positive function
c such that lim f (x)/c(x) = A 6= 0, (in short f (x) ∼ Ac(x), x → ∞) .
x→∞
1. The following example points out that a continuous and L1 -integrable

function can have S-asymptotics as a distribution without having an ordi-
nary asymptotics. Suppose that g ∈ L1 (R) ∩ C(R) has the property that
g(n) = n, n ∈ N and that it is equal to zero outside suitable small intervals
Rt
In ∋ n, n ∈ N. Denote by f (t) = et g(x)dx, t ∈ R. It is easy to see that
0
Z∞
s h t
f (t + h) ∼ e · e g(x)dx, h ∈ R+ .
0
By Theorem 1.2 a) f ′ (t) has S-asymptotics related to eh and with the same
limit. But, in view of the properties of g, f ′ (t) = f (t) + et g(t) has not the
same asymptotics (in the ordinary sense). Moreover, g can be chosen so
that f ′ has no asymptotics at all.
2. The following example shows that a function f can have asymptotic
behavior without having S-asymptotics with limit U different from zero. An
example is x 7→ exp(x2 ), x ∈ R. Suppose that exp(x2 ) has S-asymptotics
related to a c(h) > 0, h ∈ R+ with a limit U different from zero. By
Proposition 1.2 c), U has the form U (x) = C exp(ax), C > 0. Then, for
every ϕ ∈ F0 such that ϕ > 0 we have

Z
1
lim exp[(x + h + h0 )2 ]ϕ(x)dx = eah0 hCeax , ϕ(x)i .
h→∞ c(h)
Therefore,
Z
1 2
eah0 hU, ϕi = exp(h20 ) lim e(x+h) e2h0 (x+h) ϕ(x)dx
h→∞ c(h)
≥ exp(h20 )hU, ϕi, for every h0 > 0 .

But this inequality is absurd. Consequently, exp(x2 ) cannot have such an
S-asymptotic behavior.
One can prove a more general assertion.
Proposition 1.3. Let f ∈ L1loc (R) ⊂ F0′ (R) have one of the four properties
for α > 1, β > 0, x ≥ x0 , h > 0, M > 0 and N > 0 :
a) f (x + h) ≥ M exp(βhα )f (x) ≥ 0,
a’) −f (x + h) ≥ −M exp(βhα )f (x) ≥ 0,
b) 0 ≤ f (x + h) ≤ N exp(−βhα )f (x),
b’) 0 ≤ −f (x + h) ≤ −N exp(−βhα )f (x).
Then f cannot have S-asymptotics with limit U 6= 0, but the function f can
have asymptotics.
For the proof see ([135], p. 89).

It is easy to show that for some classes of real functions f on R the
asymptotic behavior at infinity implies the S-asymptotics.
Proposition 1.4. a) Let c be a positive function and let T ∈ L1loc (Rn ).

Suppose that there exist locally integrable functions U (x) and V (x), x ∈ Rn ,
such that for every compact set K ⊂ Rn
|T (x + h)/c(h)|≤ V (x), x ∈ K, khk> rK ,
lim T (x + h)/c(h) = U (x), x ∈ K .
h∈Γ,khk→∞
s
Then, T (x + h) ∼ c(h)U (x), h ∈ Γ in F0′ .
b) Let T ∈ L1loc (R) have the ordinary asymptotic behavior
T (x) ∼ exp(αx)L(exp x), x → ∞, α ∈ R ,
where L is a slowly varying function. Then,
s
T (x + h) ∼ exp(αh)L(exp h) exp(αx), h ∈ R+ , in F0′ (R) .
Proof. a) For every ϕ ∈ F0

Z
T (x + h) T (x + h)
lim , ϕ(x) = lim ϕ(x)dx . (1.14)
Rn
Since supp ϕ ⊂ K ⊂ Rn and T has all the listed properties, Lebesgue’s
theorem implies the result.
b) It is enough to use in (1.14) that L(yt)/L(t) → 1, t → ∞ uniformly
in y, when y stays in compact interval contained in R+ .
A more general result is the following one ([135], pp. 89–90).
Proposition 1.5. Let Γ be a cone and let Ω ⊂ Rn be an open set such that
for every r > 0 there exists a βr such that B(0, r) ⊂ {Ω − h; h ∈ Γ, khk≥
βr }. Suppose that G ∈ L1loc (Ω) and it has the following properties: There
exist locally integrable functions U and V in Rn such that for every r > 0
we have
|G(x + h)/c(h)|≤ V (x), x ∈ B(0, r), h ∈ Γ, khk≥ βr ;
lim G(x + h)/c(h) = U (x), x ∈ B(0, r).

h∈Γ,khk→∞
If G0 ∈ F0′ coincides with G on Ω, then

s
G0 (x + h) ∼ c(h)U (x), h ∈ Γ .
Proof. By Theorem 1.4, it is enough to proof that

Z Z
G(x + h)
lim ϕ(x)dx = U (x)ϕ(x)dx;
Ω Rn
but as in the proof of Proposition 1.4 a), the exchange of the limit and the
integral sign is justified by our assumptions and Lebesgue’s theorem.
The following proposition gives a sufficient condition under which the
S-asymptotics of f ∈ L1loc (R), in D′ (R), implies the ordinary asymptotic
behavior of f.
Proposition 1.6. Let f ∈ L1loc (R), c(h) = hβ L(h), where β > −1

and L be a slowly varying function. If for some m ∈ N, xm f (x), x >
s
0, is monotonous and f (x + h) ∼ c(h) · 1, h ∈ R+ ,, in D′ (R), then
lim f (h)/c(h) = 1. If we suppose that L is monotonous, then we can omit
h→∞
the hypothesis β > −1.
For the proof see [115].

1.7 Characterization of the support of T ∈ F0′
We suppose in this subsection that the topology in F0 is defined in such

a way that a sequence {ϕn } in F0 converges if and only if there exists a
(k)
compact set K ⊂ Rn and ϕm converges to ϕ(k) uniformly on K for every
k ∈ Nn0 as m → ∞.
We already proved in Theorem 1.4 a relation between the support of a
distribution and its S-asymptotics. Now, we shall complete this result.
We need a property of the S-asymptotics given in the next lemma.
Lemma 1.1. Let Γ be a cone and let Γ̃ be a convex cone (it is partially
ordered). A necessary and sufficient condition that for every c(h) > 0,
h∈Γ
a) w. lim T (x + h)/c(h) = 0 in F ′ 0 ,
khk→∞,h∈Γ
b) w.lim T (x + h)/c(h) = 0 in F ′ 0 ,
h→∞,h∈Γ̃
is that for every φ ∈ F0 the following holds:

In case a): There exists β(φ) > 0 such that
hT (x + h), φ(x)i = 0, khk≥ β(φ), h ∈ Γ . (1.15)
In case b): There exists hφ ∈ Γ̃ such that
hT (x + h), φ(x)i = 0, h ≥ hφ , h ∈ Γ̃ . (1.16)
Proof. We have to prove only that the condition is necessary in both

cases. It is obvious that the condition is sufficient. Let us suppose the
opposite, that is, the condition is not necessary. Then, we could find a
sequence ham im in Γ such that in case a) kam k→ ∞ in Γ, and in case b)
am → ∞ in Γ̃, (m → ∞) and that
hT (x + am ), φ(x)i = αm 6= 0, m ∈ N .
Let find c such that c(h) = αm for h = am , m ∈ N. Clearly, for such

a c(h) the function h 7→ hT (x + h)/c(h), φ(x)i cannot converge to zero as
khk→ ∞, h ∈ Γ, or h → ∞, h ∈ Γ̃. This contradicts our supposition that
the S-asymptotics equals zero in both cases.
Theorem 1.5. Let Γ be a cone and let T ∈ F ′ 0 . A necessary and sufficient

condition that for every r > 0 there exists βr such that the sets
supp T ∩ B(h, r), h ∈ Γ, khk≥ βr are empty
s
is that T (x + h) ∼ c(h) · 0, h ∈ Γ for every positive function c on Γ.
Proof. Theorem 1.2 c) and Theorem 1.4 assert that the condition in
Theorem 1.5 is necessary. We have to prove only that this condition is also
sufficient.
Let us suppose that
w. lim T (x + h)/c(h) = 0 in F0′ .
khk→∞,h∈Γ
Let φ ∈ F0 . By Lemma 1.1 a), we know that there exists β0 (φ) =

inf β(φ), where β(φ) is such that (1.15) holds. We shall prove that the
set {β0 (φ); φ ∈ F0 , suppϕ ⊂ K} is bounded for every compact set K ⊂
Rn . Let us suppose the opposite. Then we could find a sequence hφk ik in
F0 , suppϕk ⊂ K, k ∈ N, and a sequence hhk ik in Γ, khk k→ ∞ such that

ak 6= 0, p = k
hT (t + hk ), φp (t)i = Ak,p =
0, p < k.
We give the construction of sequences {φk } and {hk }. Let φk ∈

F0 , suppφk
⊂ K, k ∈ N, be such that hβ0 (φk )ik is a strictly increasing sequence which
tends to infinity. Then, there exist hhk ik in Γ and εk > 0, k ∈ N, such that
β0 (φk−1 ) + εk ≤ khk k≤ β0 (φk ) − εk . Now, we shall construct the sequence
hψp (t)ip in F0 , suppψp ⊂ K, p ∈ N, for which we have

0, p 6= k
hT (t + hk ), ψp (t)i = (∗)
ak , p = k .
Put
ψp (t) = φp (t) − λp1 φ1 (t) − · · · − λpp−1 φp−1 (t), p > 1, t ∈ Rn .
The sequence hλpi ii will be determined in such a way that ψp (t) satisfies (*)
the sought property, p ∈ N.
It is easy to see that hT (t + hk ), ψk (t)i = ak and hT (t + hk ), ψp (t)i =
0, k > p. For a fixed p and k < p we can find λpi , i = 1, . . . , p − 1, such that
for k = 1, . . . , p − 1,
0 = hT (t + hk ), ψp (t)i = Ak,p − λp1 Ak,1 − · · · − λpp−1 Ak,p−1 .
Hence
λp1 Ak,1 + · · · + λpp−1 Ak,p−1 = Ak,p , k = 1, . . . , p − 1, p > 1 .
Since Ak,k 6= 0 for every k, the system always has a solution.
We introduce a sequence of numbers hbk ik ,
(i)
bk = sup{2k |ψk (t)|; i ≤ k}, k ∈ N .
Then, the function
X∞
ψ(t) = ψp (t)/bp , t ∈ Rn , is in F0 and suppψ ⊂ K .
p=1
Thus, we have
∞
X
hT (t + hk ), ψ(t)i = hT (t + hk ), ψp (t)/bp i = ak /bk .
p=1
Now, if we choose c(h) such that c(hk ) = ak /bk , k ∈ N, then

D T (t + h), ψ(t) E
does not converge to zero when khk→ ∞, h ∈ Γ.
c(h)
This proves that for every compact set K there exists a β0 (K) such that
hT (t + h), φ(t)i = 0, khk≥ β0 (K), h ∈ Γ, φ ∈ F0 , suppφ ⊂ K.
It follows that T (t + h) = 0 over B(0, r), khk≤ β(r), h ∈ Γ and T (t) = 0

over B(h, r), khk≥ β(r), h ∈ Γ.
Remark. The condition of Theorem 1.5 implies that the support of
T has the following property: The distance from supp T to a point h ∈
Γ, d(supp T, h), tends to infinity when khk→ ∞, h ∈ Γ.
The next proposition shows that if in Definition 1.1 we take the limit
(1.3) instead of the limit (1.2), then a more precise result is obtained.
Theorem 1.6. Let T ∈ F ′ 0 and Γ̃ be an acute, open and convex cone

(partially ordered, see Remark 1 after Definition 1.1). A necessary and
sufficient condition for
supp T ∈ CRn (a + Γ̃) for some a ∈ Γ̃ (1.17)
is that
lim T (x + h)/c(h) = 0 in F0′ for every c(h) (1.18)
h∈Γ̃,h→∞
(CRn A = Rn \ A).
Proof. If (1.17) holds, then for any ball B(0, 1), there exists a hr ∈ Γ̃
such that B(h, r) ⊂ (a + Γ̃) for h ≥ hr , h ∈ Γ̃. This implies (1.18), and by
Lemma 1.1, (1.17) follows.
Let us suppose now that (1.16) and consequently (1.18) hold, but for
any a ∈ Γ̃, supp T 6⊂ CRn (a + Γ̃). We fix such an a = a0 > 0. There exists
an a1 ∈ (a0 + Γ̃) ∩ supp T. Since supp T 6⊂ CRn (2a1 + Γ̃), there exists an
a2 ∈ (2a1 + Γ̃ ∩ supp T ). In such a way, we construct a sequence hak ik in Γ̃
such that ak ∈ (kak−1 + Γ̃) ∩ supp T and ak ≥ ka0 , k ∈ N.
Since ak ∈ supp T, k ∈ N, it follows that there exists a sequence hφk ik
in F0 such that supp φk ⊂ B(0, 1), k ∈ N, and
hT (x + ak ), φk (x)i 6= 0, k ∈ N.
We put now:
ck,i = hT (x + ai ), φk (x)i, k, i ∈ N;
bk = supp{2k |φ(j) (x)|; j ≤ k, x ∈ Rn }, k ∈ N.
We have to prove that there exists a sequence hck ik such that ck ≥

1, k ∈ N and
X
ck,i /(bk ck ) 6= 0, i ∈ N . (1.19)
k≥1
P
∞
First, we notice that ck,k 6= 0. If ck,1 /bk = 0, then we take c1 > 1,
k=1
and if this series is different from 0, then we put c1 = 1. Let i = 2. If
X
c1,2 /(b1 c1 ) + ck,2 /bk = 0 (6= 0), we take c2 > 1 (c2 = 1)
k≥2
such that
X
c1,1 /(c1 b1 ) + c2,1 /(c2 b2 ) + ck,1 /bk 6= 0 .
k≥3
Let i = 3,
∞
X
c1,3 /(b1 c1 ) + c2,3 /(b2 c2 ) + ck,3 /bk = 0 or 6= 0 .
k=3
Then we take c3 > 1 and c3 = 1 respectively such that

∞
X
c1,1 /(b1 c1 ) + c2,1 /(b2 c2 ) + c3,1 /(b3 c3 ) + ck,1 /bk 6= 0,
k=4
∞
X
c1,2 /(b1 c1 ) + c2,2 /(b2 c2 ) + c3,2 /(b3 c3 ) + ck,2 /bk 6= 0.
k=4
Continuing in this way, we construct a sequence hck ik for which (1.19)

holds.
Let us put
X
ψk (x) = φk (x)/(bk ck ), k ∈ N and ψ = ψk .
k≥i
From the properties of sequences hbk ik and hck ik , we can easily show
that
N
X
ψk → ψ in F0 when N → ∞.
k=1
Relation (1.18) implies

∞
X
hT (x + ai ), ψ(x)i = ck,1 /(bk ck ) 6= 0, i ∈ N.
k=1
We obtain that (1.18) does not hold for ψ. This completes the proof.
In Theorem 1.6 the support of T can be just CRn (a + Γ̃). The question
is: Is it possible to obtain a similar proposition for the S-asymptotics given
by Definition 1.1? This question is analyzed in the next examples. We take
F0 = D in which the S-asymptotics by the weak and strong convergence
are equivalent.
P
Examples. i) Let T (x, y) = m≥1 mδ(x−m, y). The given series converges
in D′ (R2 ). Since for a φ ∈ D(R2 ), supp φ ⊂ B(0, r), we have
n
X X
lim h mδ(x − m, y), φ(x, y)i = mφ(m, 0).
n→∞
m=1 1≤m≤r
It follows that T is a distribution on R2 (see Theorem XIII, Chapter 3

in [146]).
Let us remark that the support of T lies on the half line γ ≡ {(p, 0) ∈
R2 ; p > 0}. We can take for Γ the cone R2 + ≡ {(α, β) ∈ R2 ; α > 0, β >
0}. It is a convex, open and acute cone in R2 . We shall show that the limit
lim hT (u + h), φ(u)i
h∈Γ,khk→∞
does not exist. To do this, it is enough to take the limit over the half line
γ ′ ≡ {(0, α0 ) + γ}, α0 > 0, which belongs to Γ. If we choose φ such that
φ > 0 and φ(0, −α0 ) = 1, then for h = (p, α0 ) ∈ γ ′
X
hT (u + h), φ(u)i = mφ(m − p, −α0 ) ≥ p.
m≥1
Note, if khk→ ∞, then p → ∞, as well. Consequently, the answer to the
posed question is negative.
ii) The following example shows that if
lim T (x + h)/c(h) = 0 in D′
h∈γ̃,khk→∞
for every positive c(h), and every γ̃ = {pw, p > 0}, w ∈ Γ, then this does
not imply that
lim T (x + h)/c(h) = 0 in D′ .
h∈Γ,khk→∞
Let us remark that both limits on a γ̃, when h → ∞ or khk→ ∞, are equal.
P 1
Let T be given by T (x, y) = mδ(x − m, y − ) on R2 . The support
m≥1 m
of T lies on the curve {(x, 1/x); x > 0}. Let Γ = R2+ , w = (cos α, sin α), 0 <
α < π2 and γ̃ ≡ {pw; p > 0}. Then, for a φ ∈ D(R2 )
X
1
hT (x+h1 , y+h2 ), φ(x, y)i = mφ m−h1 , −h2 → 0, khk→ ∞, h ∈ γ̃.
m
m≥1
In order to show that
lim hT (x + h1 , y + h2 ), φ(x, y)i
h∈Γ,khk→∞
does not exist, take h to belong to γ = {(x, a); x > 0} for a fixed a > 0, as
we did in example i).
Remarks. As a consequence of Theorem 1.5 and Theorem 1.6, we have
some results concerning the convolution product (see [135], p. 102).
Let
G1 ≡ {f ∈ C ∞ ; supp f ⊂ CRn {h ∈ Γ; khk≥ βf } ,
G2 ≡ {f ∈ C ∞ ; supp f ⊂ CRn {h ∈ Γ̃; h ≥ hf }.
Corollary 1.1. For a fixed T ∈ D′ , the convolution T ∗ φ maps D into G1

if and only if the support of T has the property given in Theorem 1.5.
Corollary 1.2. For a fixed T ∈ D′ and for a convex, open and partially
ordered cone Γ̃ the convolution T ∗φ maps D into G2 if and only if supp T ⊂
CRn (a + Γ̃) for some a ∈ Γ̃.
1.8 Characterization of some generalized function spaces
Theorem 1.7. A necessary and sufficient condition for a distribution T to

belong to:
s
a) E ′ is that T (x + h) ∼ c(h) · 0, h ∈ Rn for every positive c.
b) Oc′ is that T has S-asymptotic behavior related to every c(h) =
khk−α , α ∈ R+ and with limit U = 0.
c) B ′ is that T has S-asymptotic behavior related to every positive
c, c(h) → ∞, khk→ ∞, and with limit U = 0.
Proof. a) It is a direct consequence of Theorem 1.5.

b) It is enough to apply Theorem IX Chapter VII in [146] which states:
A necessary and sufficient condition that a distribution T belongs to Oc′
is that for every ϕ ∈ D the function (T ∗ ϕ)(h) is continuous and of fast
descent at infinity. (see 0.5.1).
c) By Theorem XXV, Chapter VI in [146] a distribution T ∈ B ′ if and
only if T ∗ ϕ ∈ L∞ (Rn ) for every ϕ ∈ D. Suppose that T ∈ B ′ . Then for
every ϕ ∈ D and c(h) → ∞, khk→ ∞

T (x + h) (T ∗ ϕ̌)(h)
lim , ϕ(x) = lim = 0.
Suppose that (T ∗ ϕ̌)(h)/c(h) → 0, khk→ ∞, for every ϕ ∈ D and for

every c, c(h) → ∞ as khk→ ∞. We will show that (T ∗ ϕ̌)(h) ∈ L∞ (Rn )
for every ϕ ∈ D. Then, by the same theorem, it follows that T ∈ B ′ .
Let us assume the contrary, i.e., that (T ∗ ϕ̌o )(h) is not bounded for
a ϕo ∈ D. Then, we could find two sequences hhm im in Rn and hcm im
in R such that |cm |→ ∞ as m → ∞,p khm k≥ m and (T ∗ ϕ̌o )(hm ) = cm .
Now, for co (h) such that co (hm ) = |cm |, m ∈ N, the limit < T (x +
h)/co (h), ϕo (x) > would not exist, as khk→ ∞. This is in a contradiction
with our assumption that T has the S-asymptotics related to every c(h)
which tends to infinity as khk→ ∞.
Proposition 1.7. a) If for every rapidly decreasing function c, T has the

S-asymptotic behavior related to c−1 and with limit Uc (Uc = 0 is included),
then T ∈ S ′ .
b) If for every rapidly exponentially decreasing function c (for every

k > 0, c(h) exp(kkhk) → 0, khk→ ∞) a distribution T has the S-asymptotic
behavior related to c−1 with limit Uc (Uc = 0 is included), then T ∈ K′ 1 .
Proof. a) Let c be given. There exists β0 such that for every h, khk≥ β0 ,
and for every ϕ ∈ D
|< T (x + h) · c(h), ϕ(x) > |≤ |< Uc , ϕ > |+εϕ ≤ Mϕ + εϕ .
Therefore, the set {T (x + h) · c(h); h ≥ β0 } is weakly bounded and thus
bounded in D′ . By Theorem VI 4o , Chapter VII in [146] if {c(h)T (x +
h); khk> β0 } is bounded in D′ , for every c of fast descent, then T ∈ S ′ .
b) The proof is similar to that of a), if we use the following theorem
proved which will be showed below (cf. Theorem 1.15):
Let T ∈ D′ . If for every rapidly exponentially decreasing function r on
R the set {r(h)T (x + h); h ∈ Rn } is bounded in D′ , then T ∈ K′ 1 .
n

1.9 Structural theorems for S-asymptotics in F ′
In the analysis of the S-asymptotics and its applications, it is useful to

know analytical expression for generalized functions having S-asymptotics,
especially if it is given via continuous functions and their derivatives. The
next theorems are of this kind. These types of theorems are usually referred
as structural theorems.
Our first result concerns the one-dimensional case, a refinement will be
obtained in Theorem 1.9 below.
Theorem 1.8. Let α ∈ R and let L be a slowly varying function. Suppose

s
that f ∈ D′ (R). If f (x + h) ∼ c(h) · eαx , h ∈ R+ , then there is an m0 ∈ N
such that for every m ≥ m0 the following holds:
(1) Let α 6= 0 and c(h) = eαh L(eh ), h > 0. Then there are gm,i ∈
C(1, ∞), i = 0, 1, . . . , m such that
m
X (i)
f (x) = gm,i (x), x ∈ (1, ∞),
i=0
and
gm,i (x) ∼ Ci xm exp(αx)L(exp x), x → ∞, i = 0, . . . , m ,
where Ci are suitable constants.
(2) Let α = 0 and c(h) = hν L(h), h > 0. Then,

(m)
a) If ν > −1, then there is Fm ∈ C(1, ∞) such that f = Fm and
Fm (x) ∼ xm+ν L(x), x → ∞;
b) If ν ≤ −1, then there are fm,i ∈ C(1, ∞) and Am,i 6= 0, i =
0, 1, . . . , m, such that
fm,i (x) ∼ Am,i xm+ν−i L(x), i = 0, 1, . . . , m
and
m
X (m−i)
f (x) = fm,i (x), x ∈ (1, ∞) .
i=0
Proof. Case α 6= 0. By the remark 1. after Proposition 1.2, there exists

a function c̃ ∈ C ∞ (R) such that c(h)/c̃(x + h) → exp(−αx), h → ∞ in R
and c̃(i) (x + h)/c(h) → αi exp(αx), i ∈ N0 .
Since f (x + h)/c(h) → exp(αx) with strong convergence and the set
{c(h)φ(x)/c̃(x + h); h ≥ A} is bounded in D(R), we can then use Theorem
XI, Chapter III in [146] to obtain
f (x + h) f (x + h) c(h)
lim h , φ(x)i = lim h , φ(x)i
h→∞ c̃(x + h) h→∞ c(h) c̃(c + h)
= h1, φ(x)i, φ ∈ D(R) .

∞
Let θ ∈ C , θ(x) = 0 for x < 0 and θ(x) = 1 for x > 1. We have
θ(x + h)f (x + h)
→ 1 in D′ (R) as h → ∞ .
c̃(x + h)
Thus, {θ(x+h)f (x+h)/c̃(x+h); h > 0} is a bounded subset of D′ (R). This
implies that this set is bounded in S ′ (R) as well (cf. Theorem XXV, Chap-
ter VI in [146]). Since D(R) is dense in S(R), by the Banach–Steinhaus
theorem, we obtain
θ(x + h)f (x + h)
→ 1 in S ′ (R) as h → ∞, i.e.,
c̃(x + h)

(θf /c̃)(x + h)
lim , φ(x) = h1, φi, for every φ ∈ S(R),
h→∞ d(h)
where d(h) = 1, h > A. We shall now borrow two theorems about quasi-
asymptotics (see Definition 2.2) which will be shown later. Since the S-
asymptotics in S ′ + with ν > −1 implies the quasi-asymptotics of θf /c̃
(see Theorem 2.47), the structural theorem for the quasi-asymptotics (see
Theorem 2.2) implies that there is m0 ∈ N such that for every m > m0
there is Fm ∈ C(R) such that
(m)
(θf /c̃)(x) = Fm (x), x ∈ R,
and Fm (x) ∼ xm as x → ∞. Thus, we obtain
(m)
f (x) = c̃(x)Fm (x), x ∈ (1, ∞).
The Leibnitz formula implies
Xm
m
f (x) = (−1)i (c̃(i) (x)Fm (x))(m−i) , x ∈ (1, ∞).
i=0
i
Since
c̃(i) (x + h)
→ αi eαx , h → ∞, x ∈ R,
c(h)
we obtain
c̃(i) (h) ∼ αi c(h), h → ∞.
This implies the result when α 6= 0 and c(h) = eαh L(eh ), h > 0.
In case α = 0 and c(h) = hν L(h), h > 0, ν > −1, the S-asymptotics of
θf related to c(h) = hν L(h), h > 0, ν > −1, implies the quasi-asymptotics
of θf related to the same c(h) (see Theorem 2.47). The assertion follows
now from Theorem 2.2.
If ν ≤ −1, then we take k > 0 such that k + ν > −1. With θ as in the
preceding proof and by Theorem 1.2 b), we have
s
(1 + (x + h)2 )k/2 θ(x + h)f (x + h) ∼ hk+ν L(h) · 1, h ∈ R+ .
By the same arguments as in the preceding proof, we have that there is
m0 ∈ N such that for every m > m0 there is an Fm ∈ C(R), supp Fm ⊂
(0, ∞)
Fm (x) ∼ xν+k+m L(x), x → ∞
(1 + x2 )k/2 θ(x)f (x) = Fm

(m)
(x), x ∈ R.
Thus, for x ∈ (1, ∞),
Xm (i) (m−i)
m 1
f (x) = (−1)i Fm (x) .
i=0
i (1 + x2 )k/2
The result now follows from the fact

(i)
1
∼ Ci x−k−i , x → ∞ ,
(1 + x2 )k/2
where Ci 6= 0 are suitable constants, i = 0, . . . , m.
We can show a more general result which describes the precise structure
of a distribution having S-asymptotics.
Theorem 1.9. If T ∈ D′ has S-asymptotics related to the open cone Γ and

the continuous and positive function c(h), h ∈ Γ, then for the ball B(0, r)
there exist continuous functions Fi , |i|≤ m, such that Fi (x + h)/c(h) con-
verges uniformly for x ∈ B(0, r) when h ∈ Γ, khk→ ∞, and the restriction
P
of the distribution T on B(0, r)+Γ can be given in the form T = Di Fi .
|i|≤m
We need the following lemma for the proof of the theorem:

s
Lemma 1.2. Let T ∈ D′ , T (x + h) ∼ c(h) · U (x), h ∈ Γ. Then, for an
open ball B(0, r) and a relatively compact open neighborhood Ω of zero in
Rn , there exists an m ∈ N0 such that for every ϕ, ψ ∈ DΩ m
the function
(T ∗ϕ∗ψ)(x) is continuous for x ∈ B(0, r)+Γ. Moreover, the set of functions
{(Th ∗ϕ∗ψ)(x); h ∈ Γ} converges uniformly for x ∈ B(0, r) to (U ∗ϕ∗ψ)(x),
as h ∈ Γ, khk→ ∞; Th = T (x + h)/c(h).
Proof. Suppose that T has S-asymptotics related to c(h). Then, the set
{T (x + h)/c(h) ≡ Th ; h ∈ Γ} is weakly bounded in D′ and consequently,
bounded in D′ . A necessary and sufficient condition that a set B ′ ⊂ D′ is
bounded in D′ is: for every α ∈ D the set of functions {T ∗ α; T ∈ B ′ }
is bounded on every compact set K ⊂ Rn (see §7, Chapter VI in [146]).
Moreover {T ∗ α; T ∈ B ′ } defines a bounded set of regular distributions.
Let Cl(Ω) = K (Cl(Ω) is the closure of Ω); K is a compact set. For a
fixed α ∈ C∞0 , supp α ⊂ K, the linear mappings β 7→ (Th ∗ α) ∗ β, h ∈ Γ,
are continuous mappings of DK into E because of the separate continuity
of the convolution. Since the set {Th ∗ α; h ∈ Γ} is a bounded set in D′ , for
every ball B(0, r) the set of mappings β → {(Th ∗ α) ∗ β; h ∈ Γ} is the set
of equicontinuous mappings of DK into L∞ B , where B = B(0, r). Now there
exists an m ≥ N0 such that the linear mappings (α, β) → Th ∗ α ∗ β which
map DK × DK into L∞ m m
B can be extended to DΩ × DΩ in such a way that
m m
(α, β) 7→ Th ∗ α ∗ β, h ∈ Γ, are equicontinuous mappings of DΩ × DΩ into
∞
LB (see for example the proof of Theorem XXII, Chapter VI in [146]).
m
We proved that for every ϕ, ψ ∈ DΩ and every h ∈ Γ the functions
x 7→ (Th ∗ ϕ ∗ ψ)(x) are continuous in x ∈ B(0, r). From the relation
(Th ∗ ϕ ∗ ψ)(x) = (T ∗ ϕ ∗ ψ)(x + h)/c(h) and from the properties of c it
follows that y 7→ (T ∗ ϕ ∗ ψ)(y) is a continuous function for y ∈ B(0, r) + Γ
m
and ϕ, ψ ∈ DΩ .
It remains to prove that Th ∗ϕ∗φ converges to U ∗ϕ∗φ as khk→ ∞, h ∈
Γ, in L∞ m m
B for ϕ, ψ ∈ DΩ . We know that D is a dense subset of D , m ≥ 0.
m
We can construct a subset A of DK to be dense in DΩ . The set of functions
Th ∗α∗β converges in L∞ B for α, β ∈ A, when h ∈ Γ, khk→ ∞. Taking care of
m m
the equicontinuity of the mappings DΩ × DΩ into L∞
B , defined by Th ∗ ϕ∗ ψ,
we can use the Banach–Steinhaus theorem to prove that Th ∗ϕ∗ψ converges
in L∞B when h ∈ Γ, khk→ ∞.
Proof of the Theorem. We shall use (VI, 6; 23) from [146].
∆2k ∗ (γE ∗ γE ∗ T ) − 2∆k ∗ (γE ∗ ξ ∗ T ) + (ξ ∗ ξ ∗ T ) = T , (1.20)
where E is a solution of the iterated Laplace equation ∆k E = δ; γ, ξ ∈ DΩ .

We have only to choose the natural number k large enough so that E belongs
m
to DΩ . Now, it is possible to take F1 = γE ∗ γE ∗ T, F2 = γE ∗ ξ ∗ T and
F3 = ξ ∗ ξ ∗ T. All of these functions are of the form Fi = T ∗ ϕi ∗ ψ, ϕi , ψ, ∈
m
DΩ , i = 1, 2, 3.
The following holds:
Fi (x + h)/c(h) = (Fi (x)/c(h)) ∗ τ−h = (T ∗ ϕi ∗ ψi , )(x) ∗ τ−h /c(h)

= ((T (x) ∗ τ−h )/c(h) ∗ (ϕi ∗ ψi ) = (Th ∗ ϕi ∗ ψi )(x), x ∈ B(0, r), h ∈ Γ .
Hence, by Lemma 1.2 it follows that Fi (x + h)/c(h) converges uniformly

for x ∈ B(0, r) when h ∈ Γ, khk→ ∞.
Consequences of Theorem 1.9

a) If the functions Fi , |i|≤ m, have the property given in Theorem 1.9
and if Γ = Rn , then the regular distributions defined by the functions Fi /c
have the S-asymptotic behavior related to c1 (h) ≡ 1.
b) If T ∈ S ′ , then all functions Fi , |i|≤ m, are continuous for x ∈
B(0, r) + Γ, and of slow growth (Fi = (1 + r2 )q fi , |i|≤ m, q ∈ R, where
r = kxk and fi , |i|≤ m, are continuous and bounded functions).
c) The converse of Theorem 1.9 is also true. Therefore, it completely
characterizes those distributions having S-asymptotics.
Proof. a) Suppose that Γ = Rn . Then by the properties of Fi , |i|≤ m,

the functions Fi /c are continuous and Fi (x)/c(x) converge to the numbers
Ci as kxk→ ∞. Now, for |i|≤ m we have
lim hFi (x + h)/c(x + h), ϕ(x)i
khk→∞
Z
= lim (Fi (x + h)/c(x + h))ϕ(x)dx = hCi , ϕi, ϕ ∈ D.
khk→∞
Rn
b) If T ∈ S ′ , then there exists a q ∈ R such that the set of distributions

{T (x + h)/(1 + khk2 )q ; h ∈ Rn } = W is bounded in D′ (Theorem VI,
Chapter VII in [146]). We can now repeat the first part of the proof of
Lemma 1.2 but with c(h) = (1 + khk2 )q , h ∈ Γ = Rn . In this may we
p
obtain that there exists a p ∈ N0 such that for ϕ, ψ ∈ DΩ and x ∈ Rn the
function x 7→ (T ∗ϕ∗ψ)(x) is continuous and (T ∗ϕ∗ψ)(x)/(1+kxk2)q , x ∈ R
is bounded. It remains only to choose the number k in (1.20) large enough
max(m,p)
so that γE ∈ DΩ .
c) It follows directly from Theorem 1.2 a) and Proposition 1.5.
We can also characterize the structure of ultradistributions having S-
asymptotics.
′
Theorem 1.10. ([132]). Let T ∈ DΩ∗ . Suppose: (M.1), (M.2) and (M.3)
are satisfied by Mp ; Ω = Ω1 +Γ1 , where Ω1 ⊂ Rn is open set and Γ1 ⊂ Rn is
convex cone; Γ is a subcone of Γ1 . Then T has the S-asymptotics related to c
and Γ if and only if for a given open and relatively compact set A( A ⊂ Ω),
there exist an ultradifferential operator P (D) of class ∗ and continuous
functions f1 and f2 on A + Γ such that
lim fi (x + h)/c(h), i = 1, 2,
h∈Γ,khk→∞
exist uniformly in x ∈ A, i = 1, 2, and T = P (D)f1 + f2 on A + Γ.
Proof. One can easily prove that the condition is sufficient.

s
The condition is necessary. Suppose that T (x + h) ∼ c(h) · U (x), h ∈ Γ
in D′ ∗Ω . Denote by Th = τh T /c(h) and by (CB)A the space of continuous
and bounded functions on A. By Theorem 6.10 in [79], Fh : ϕ 7→ Th ∗ϕ, h ∈
∗
Γ, are continuous mappings: DB r+ε
→ (CB)A . Consequently, Fh are the
∗
continuous mappings: DK → (CB)A , where K = B r ⊂ Ω. The set {Th ; h ∈
∗
Γ, khk≥ γ}, γ > 0, is bounded in D′ Ω because of the S-asymptotics of T.
∗
Thus, for a fixed ϕ ∈ DK , the set {Fh (ϕ); h ∈ Γ, khk≥ γ} is bounded in
∗
(CB)A . Since DK is a barrelled space, by the Banach theorem it follows that
the family of functions {Fh } = {Fh ; h ∈ Γ, khk≥ γ} is equicontinuous.
Therefore, there exists Hp , p ∈ N of the form (0.1) and β > 0 such that
any function in {Fh } maps the neighborhood of zero

∗
Vβ = φ ∈ DK ; sup |φ(α) (x)|/H|α| M|α| < β (1.21)
x∈K,|α|∈N0
H Mp
into the unit ball B(0, 1) in (CB)A . Denote by D̃Kp the completion of
∗
DK under the norm qHp Mp (see (0.5)).
Using the extension of a function through its continuity (see [12]), we
H M
shall show that the family {Fh } can be extended on D̃Kp p , keeping the
uniform continuity; let us denote it by {F h }.
H M ∗
Let φ ∈ D̃Kp p and let hφj ij be a sequence in DK which converges to
φ being a Cauchy sequence in the norm qHp Mp . We shall prove that Th ∗ ψj
converges, as j → ∞, in (CB)A uniformly in h ∈ Γ, khk≥ γ. It is enough
to prove that hTh ∗ ψj ij is a Cauchy sequence in (CB)A .
Every neighborhood of zero W in (CB)A contains the ball B(0, δ) for
some δ > 0. The neighborhood of zero Vβδ , given by (1.21), satisfies Th ∗
Vβδ ⊂ W, when h ∈ Γ, khk≥ γ.
Let j0 ∈ N0 be such that ψi − ψj ∈ Vβδ if i, j ≥ j0 . Then,
Th ∗ ψi − Th ∗ ψj = Th ∗ (ψi − ψj ) ∈ W, h ∈ Γ, khk≥ γ .
This proves the existence of the family {F h } and that lim Th ∗ ψj =

j→∞
gh ∈ (CB)A , khk≥ γ. We shall prove that gh = Th ∗ φ. The sequence hψj ij
converges to φ also in E ′ ∗Br+ε , as well. So, Th ∗ ψj converges to Th ∗ φ as
∗
j → ∞, in D′ ω , where Ω = ω − Br+ε (see [79], p. 73). Thus, (Th ∗ φ)|A
must be gh .
H M
It remains to prove that for φ ∈ D̃Kp p , Th ∗ φ converges in (CB)A as
h ∈ Γ, khk→ ∞. Since {F h } is an equicontinuous family of functions, for
H M
every φ ∈ D̃Kp p the set {Th ∗ φ; h ∈ Γ, khk≥ γ} is bounded in (CB)A .
By the Banach–Steinhaus theorem Th ∗ φ converges in (CB)A .
Now we will give the analytic form of T. Let φ ∈ DQ such that Q is
a compact subset of the interior of K and qHp Mp (φ) < ∞. One can easily
H M
prove that φ ∈ D̃Kp p . Now, by Theorem 2.11 in [81], T = P (D)(φ ∗ T ) −
H M ∗
w ∗ T, where we can take that φ ∈ D̃Kp p and w ∈ DK . Let us denote φ ∗ T
by f1 and w ∗ T by f2 . The properties of these two functions follow by the

previous part of the proof.
We pass now to the case of Fourier hyperfunctions.
Theorem 1.11. Let f = [F ′ ] ∈ Q(Dn ), F ′ ∈ Õ((Dn +iI#Dn ). If f has the

S-asymptotics related to c and Γ, then there exist an elliptic local operator
J(D) and functions qs ∈ C ∞ (Rn ), s ∈ Λ, is the set of n-vectors with entry
{−1, 1}), of infra exponential type such that:
1. qs (z) ∈ Õ(Dn + iIs ), where Dn + iIs is an infinitesimal wedge of type
n
D + iΓs 0, s ∈ Λ.
P
2. f = J(D) qs (x).
s∈Λ
3. For every s ∈ Λ there exists ε0 > 0 such that qs (x + iεs)/c(x), s ∈ Λ

converge as kxk→ ∞, x ∈ Γ, for every fixed ε, 0 < ε < ε0 .
Proof. Let
X
f∼
= Fσ (x + iΓσ ), Fσ = sgnσFσ′
σ∈Λ
and
XX
F (f ) = [R] ∼
= F (χσ Fσ )(ξ − iΓσ̃ 0) .
σ∈Λ σ̃∈Λ
Then there exists a monotone increasing continuous positive valued function

ε(r), r ≥ 0, which satisfies ε(0) = 1, ε(r) → ∞, r → ∞ and such that
1
|R(ζ)|≤ Ck exp(|ζ|/ε(|ζ|), ≤ |Im ζj |≤ 1 ,
k
where j = 1, . . . , n and k ∈ N (cf. [74], p. 652).
By Lemma 1.2 in [73], we can choose an entire function J of infra
exponential type on Cn which satisfies the estimate:
|J(ζ)|≥ C exp(|ζ|/ε(|ζ|)), |Im ζ|≤ 1 .
Then J −2 (ζ) ∈ Õ(Dn + i{|µ|< 1}) and
|R(ζ)/J 2 (ζ)|≤ Ck exp(−|ζ|/ε(|ζ|)) ≤ Ck exp(−|ζ|γ ) ,

1
where ≤ |Im ζj |≤ 1, j = 1, . . . , n, and 0 < γ < 1.
k
Denote by g = F −1 (1/J 2 ). By Theorem 8.2.6 in [75], g ∈ Q−1 (Dn ). By

Proposition 8.4.3. in [75]
Z
f ∗ g = f (· − t)g(t)dt ∈ Q(Dn )
R
and
F (f ∗ g) = F (f )F (g) .
By the Corollaries of Proposition 2 in [165]
f = J0 (D)(g ∗ f ), J0 = J 2 . (1.22)
We can always assume that there exists α ∈ R+ such that I = (−α, α)n .
Then we denote by Iσ = I ∩ Γσ , σ ∈ Λ.
By Proposition 8.3.2 in [75] the following assertions are true:
Fσ χσ̃ ∈ Õ(x + iIσ ) and it decreases exponentially outside any cone
containing Γσ̃ as a proper subcone.
F (Fσ χσ̃ ) ∈ Õ(x − iIσ ) and it decreases exponentially outside any cone
containing Γσ̃ as a proper subcone.
F (Fσ χσ̃ )J −2 has the same cited properties as F (Fσ χσ̃ ).
F (Fσ χσ̃ )J −2 χs ∈ Õ(x − iIs ) and it decreases exponentially outside any
cone containing Γσ and Γs proper subcone.
F −1 (F (Fσ χσ̃ )J −2 χs ) ∈ Õ(x + i(Iσ ∪ Is )) and it decreases exponentially
outside any cone containing Γs as a proper subcone.
Consider now the Fourier hyperfunction f ∗ g given in (1.22):
f ∗ g = F −1 (F (f )F (g))
Z
∼ 1 XX
= eizσ ζσ F (χσ̃ Fσ )(ζσ̃ )/J 2 (ζσ̃ )dξ ,
(2π)n
σ∈Λ σ̃∈Λ Rn
where ησ̃ ∈ −Iσ̃ and zσ ∈ Rn + iIσ .

Let σ ∈ Λ be fixed. Then for σ̃ ∈ Λ and zσ ∈ Rn + iIσ , we have
Z
1
Sσ,σ̃ (zσ ) = eizσ ζσ̃ F (χσ̃ Fσ )(ζσ̃ )/J 2 (ζσ̃ )dξ ;
(2π)n
Rn
Z
1
|Sσ,σ̃ (zσ )| ≤ e−xησ̃ −yσ ξ F (χσ̃ Fσ )(ζσ̃ )/J 2 (ζσ̃ )dξ .
(2π)n
Rn
One can see that Sσ,σ̃ (zσ ), zσ ∈ Rn + Iσ are continuable to the real
axis. The corresponding functions x 7→ Sσ,σ̃ (x) are continuous and of infra
exponential type on Rn . By Lemma 8.4.7 in [75], for x ∈ Rn ,
Sσ,σ̃ (x) ∼
= Sσσ̃ (x + iΓσ 0), σ̃ ∈ Λ,
X X (1.23)
(f ∗ g)(x) = Sσ,σ̃ (x), x ∈ Rn .
σ∈Λ σ̃∈Λ
Functions Sσ,σ̃ can be written in the form

Z
1 X
Sσ,σ̃ (zσ ) = eizσ ζσ̃ F (χσ Fσ )(ζσ̃ )/J 2 (ζσ̃ )χs (ζσ̃ )dξ, zσ ∈ Rn +Iσ .
(2π)n
σ∈Λ Rn
Let s ∈ Λ. Then functions

Z
1
Sσ,σ̃,s (zσ ) = eizσ ζσ̃ F (χσ̃ Fσ )(ζσ̃ )/J 2 (ζσ̃ )χs (ζσ̃ )dξ ,
(2π)n
Rn
n
zσ ∈ R + iIσ , σ, σ̃, s ∈ Λ, are also continuable to the real axis. The corre-
sponding functions x → Sσ,σ̃,s (x) are continuous and of infra exponential
type on Rn . Moreover, on Rn ,
Sσ,σ̃,s (x) ∼
= Sσ,σ̃,s (x + iΓσ 0) ,
X (1.24)
Sσ,σ̃ (x) = Sσ,σ̃,s (x) .
s∈Λ
Let us analyze functions

Isε (ζ) = J −2 (ζ)e−εsζ χs (ζ), ζ ∈ (Rn + i{|η|< 1}) ,
where s ∈ Λ and ε > 0. These functions are elements of P∗ because of

−2
P
n Qn
|Is,ε (ζ)| = |J (ζ)|exp −ε si ζi i=1 |χsi (ζi )|
i=1
Qn
≤ |J −2 (ζ)| i=1 |χsi (ζi ) exp(−εsi ζi )
P
n
≤ C exp(−ε |ξi |), |η|< 1.
i=1
Therefore, Is,ε ∈ Õ−ε (Dn + i{|η|< 1}), s ∈ Λ. Since the Fourier transform
maps P∗ onto P∗ , there exists ψs,ε ∈ P∗ such that F (ψs,ε ) = Is,ε . By
Proposition 8.2.2 in [75],
ψs,ε ∈ Õ−1 (Dn + i{|y|< ε}), s ∈ Λ .
Denote by
X X
qs (x) = Sσ,σ̃,s (x)
σ∈Λ σ̃∈Λ
X X
∼
= F −1 (F (Fσ χσ̃ )J −2 χσ̃ )(x + i(Γσ ∪ Γs )0), x ∈ Rn . (1.25)
σ∈Λ σ̃∈Λ
We prove that functions qs , s ∈ Λ, have the properties cited in Theorem

1.11.
Property 1 follows from (1.24) and (1.25).
By (1.22) and (1.23) property 2 is satisfied.
It remains to prove property 3.
If f ∈ Q(Dn ) and ϕ ∈ P∗ , then f ∗ ϕ ∈ Õ(Dn + iI ′ ), where I ′ is
an interval containing zero. We shall use this fact and the properties of
functions Is,ε , already analyzed.
For a fixed s ∈ Λ there exists ε0 > 0 such that εs belongs to all in-
finitesimal wedges of the form Dn + i(Γσ ∪ Γs )0 which appear in (1.25). For
ε ∈ (0, ε0 ], we have
Z
P P 1
qs (x + iεs) = n
ei(x+iεs)ζσ̃ F (Fσ χσ̃ )(ζσ̃ )J −2 (ζσ̃ )χsi (ζσ̃ )dξ
σ∈Λ σ̃∈Λ (2π)
Rn
Z
P P 1
= n
eixζσ̃ F (Fσ χσ̃ )(ζσ̃ )F (ψs,ε )(ζσ̃ )dξ
σ∈Λ σ̃∈Λ (2π)
Rn
P P
= ((Fσ χσ̃ ) ∗ ψs,ε )(x)
σ∈Λ σ̃∈Λ

P
= Fσ ∗ ψs,ε (x) = (f ∗ ψs,ε )(x).
σ∈Λ
Now, for every fixed ε ∈ (0, ε0 ], and s ∈ Λ,
lim qs (x + iεs)/c(x) = lim (f ∗ ψs,ε )(x)/c(x)

x∈Γ,kxk→∞ x∈Γ,kxk→∞
= lim hf (t + x)/c(x), ψs,ε (t)i.

x∈Γ,kxk→∞
We cite some papers related to this problem: ([131], [132], [155],[151], [123]).
1.10 S-asymptotic expansions in Fg′
A sequence hψn in of positive real-valued functions ψn (t), t ∈ (t0 , ∞), t0 ≥ 0

(defined on (0, t0 ), t0 > 0) is said to be asymptotic if and only if ψn+1 (t) =
P
o(ψn (t)), t → ∞ (t → 0). The formal series un (t) is an asymptotic
n≥1
expansion of the function u related to the asymptotic sequence {ψp (t)} if
k
X
u(t) − un (t) = o(ψk (t)), t → ∞ (t → 0) (1.26)
n=1
for every k ∈ N. We write in this case:

∞
X
u(t) ∼ un (t)|{ψn (t)}, t → ∞ (t → 0). (1.27)
n=1
If un (t) = cn ψn (t), for every n ∈ N, where cn are complex numbers, then

expansion (1.27) is unique. Indeed, the numbers cn can be unambiguously
computed from (1.26). In this case, we omit from the notation {ψn (t)} in
(1.27). A series which is an asymptotic expansion of a function f can be
also convergent. However, the series is divergent in general; nevertheless,
several terms of it can give valuable information, and very often its good
approximation properties come actually from the fact that the series is
divergent. Sometimes, if we take more terms from the asymptotic series,
we obtain a worse approximation; consequently, the determination of the
optimal number of terms for a good approximation depends on a careful
analysis of the problem under consideration.
An asymptotic expansion does not determine only one function. The
following example illustrates this fact:
X∞ ∞
X
1 x−k 1 x−k
exp ∼ and exp + exp(−x2 ) ∼ , x → ∞.
x k! x k!
k=0 k=0
In many problems of applied mathematics one is led to the use of asymp-

totic series. (See [44], [10], [15], [14], [192], [69]). A clear exposition of the
theory and the use of asymptotic series of functions and distributions can
be found in ([50]–[56]).
We shall discuss in this section the S-asymptotic expansion of general-
ized functions belonging to Fg′ .
1.10.1 General definitions and assertions
In this section Γ will be a convex cone with the vertex at zero belonging to
Rd and Σ(Γ) the set of all real-valued and positive functions c(h), h ∈ Γ.
We shall consider the asymptotic expansion when khk→ ∞, h ∈ Γ.
Definition 1.2. A distribution T ∈ Fg′ has a S-asymptotic expansion re-

lated to the asymptotic sequence hcn (h)in ⊂ Σ(Γ), if for every ϕ ∈ F
∞
X
hT (t + h), ϕ(t)i ∼ hUn (t, h), ϕ(t)i|{cn (h)}, khk→ ∞, h ∈ Γ , (1.28)
n=1
where Un (t, h) ∈ Fg′ (with respect to t) for n ∈ N and h ∈ Γ. We write in

short:
X∞
s
T (t + h) ∼ Un (t, h)|{cn (h)}, khk→ ∞, h ∈ Γ . (1.29)
n=1
Remarks. 1) In the special case Un (t, h) = un (t)cn (h), un ∈ Fg′ , n ∈

N, we simply write
∞
X
s
T (t + h) ∼ un (t)cn (h), khk→ ∞, h ∈ Γ . (1.30)
n=1
In this case the given S-asymptotic expansion is unique.

2) Brychkov’s general definition is in S ′ (R) and slightly different from
ours ([15], [16] and [20]); his idea reformulated in Fg′ (R) gives the following
definition. Suppose that f ∈ Fg′ (R) and that the function exp(ixh), where
h is a real parameter, is a multiplier in Fg′ (R).
Definition 1.3. Suppose that f ∈ Fg′ (R). It is said that f (x)eixh has an
asymptotic expansion related to the asymptotic sequence hψn (h)in if for
every ϕ ∈ Fg (R)
∞
X
hf (x)eixh , ϕ(x)i ∼ hCn (x, h), ϕ(x)i|{ψn (h)}, h → ∞,
n=1
where Cn (x, h) ∈ Fg′ (R) (with respect to x), n ∈ N, h ≥ h0 . We write in

short:
X∞
f (x)eixh ∼ Cn (x, h)|{ψn (h)}, h → ∞ .
n=1
To obtain an equivalent definition of this asymptotic expansion,

Brychkov has supposed that Fg′ (R) = S ′ (R) and Fg (R) = S(R). Then, by
putting g = fˆ = F (f ), cn (·, t) = F (Cn (·, t)) and φ = F (ϕ), Definition 1.3
reduces to:
Definition 1.4. A distribution g ∈ S ′ (R) has an asymptotic expansion

related to the sequence hψn (h)in if for every φ ∈ S(R)
∞
X
hg(h − t), φ(t)i ∼ hcn (t, h), φ(t)i|{ψn (h)}, h → ∞,
n=1
where cn (·, h) ∈ S ′ (R), n ∈ N and h ≥ h0 .
Definition 1.4 is, of course, a particular case of Definition 1.2 if we use

T (t) = g(−t) and the cone Γ = R− .
In [15] and [20] authors studied asymptotic expansions of tempered dis-
tributions given by Definition 1.3. In [16] Brychkov extended Definition 1.4
to the n-dimensional case, but only on a ray {εy; ε > 0} for a fixed y ∈ Rn .
We study in this section the asymptotic expansion not only in S ′ (R)
and not only on a ray, but on a cone in Rn . The next remarks state some
motivations for such investigations.
Remarks. 1) A distribution in S ′ (R) can have an S-asymptotic expan-
sion in D′ (R) without having the same S-asymptotic expansion in S ′ (R).
Such an example is the regular distribution f defined by
f (t) = H(t) exp(1/(1 + t2 )) exp(−t), t ∈ R ,
where H is the Heaviside function.

It is easy to prove that for h ∈ R+
∞
X
s 1
f (t + h) ∼ (1 + (t + h)2)1−n exp(−t − h)|{e−hh2(1−n) }, h → ∞ ,
n=1
(n − 1)!
while
t 7→ Un (t, h) = (1 + (t + h)2 )1−n exp(−t − h), n ∈ N, h > 0
do not belong to S ′ (R).

2) The regular distribution g defined by
g(t) = exp(1/(1 + t2 )) exp(t), t ∈ R

belongs to D′ (R) but not to S ′ (R). It has the S-asymptotic expansion in

D′ (R) :
∞
X
s 1
g(t + h) ∼ (1 + (t + h)2 )1−n exp(t + h)|{eh h2(1−n) }, h → ∞ ,
n=1
(n − 1)!
where Γ = R+ .
3) We can distinguish two cases of S-asymptotic expansions. If in
Definition 1.2,
Un (t, h) = un (t)cn (h), n ∈ N ,
then the S-asymptotic expansion is called of second type. (See Remark

after Definition 1.2). If Un (t, h) = un (t + h), n ∈ N, then the S-asymptotic
expansion is of first type.
The following example illustrates the difference between these two types
of S-asymptotic expansions.
√
Let f (x) = x2 + x, x > 0 and f (x) = 0, x ≤ 0. A S-asymptotic
expansion of the first type for this distribution is
X∞
s 1/2
f (x + h) ∼ (x + h)2−n |{h2−n }, h → ∞ ,
n=1
n − 1
1/2
2−n
but the sequence un (x) = n−1 x cannot give a S-asymptotic expansion
of the second type for f.
S-asymptotic expansions have similar properties as those of S-
asymptotics.
Theorem 1.12. Let T ∈ Fg′ and
∞
X
s
T (t + h) ∼ Un (t, h)|{cn (h)}, khk→ ∞, h ∈ Γ.
n=1
Then:
∞
X
s
a) T (k) (t + h) ∼ Un(k) (t, h)|{cn (h)}, khk→ ∞, h ∈ Γ.
n=1
b) Let the open set Ω have the property: for every r > 0 there exists a
β0 such that the closed ball B(0, r) = {x ∈ Rn , kxk≤ r} is in {Ω − h, h ∈
Γ, khk≥ β0 }. If T, T1 ∈ F ′ 0 and T1 = T over Ω, then
∞
X
s
T1 (t + h) ∼ Un (t, h)|{cn (h)}, khk→ ∞, h ∈ Γ ,
n=1
as well.
c) Assume additionally that Fg is a Montel space and that in Fg the
convolution is well defined (see 0.6) and hypocontinuous. Let S ∈ Fg′ ,
supp S being compact. Then
∞
X
s
(S ∗ T )(t + h) ∼ (S ∗ Un )(t, h)|{cn (h)}, khk→ ∞, h ∈ Γ .
n=1
Proof. We prove only a). The proofs of b) and c) are the same as in the
proof of Theorem 1.2. We have

Pm
(k)
T (k) (t + h) − Un (t, h), ϕ(t)
n=1
lim
h∈Γ,khk→∞ cm (h)

P
m
T (t + h) − Un (t, h), (−1)|k| ϕ(k) (t)
n=1
= lim = 0.
h∈Γ,khk→∞ cm (h)
A relation between the asymptotic expansion of a locally integrable func-

tion f and its S-asymptotic expansion when seen as a regular generalized
function is provided in the following proposition (see [152]).
Proposition 1.8. Let f (t), Un (t, h) and Vn (t), t ∈ Rn , n ∈ N, h ∈ Γ,

be locally integrable functions such that for every compact set K ⊂ Rn the
following ordinary asymptotic expansion holds,
X∞
f (t + h) ∼ Un (t, h)|{cn (h)}, khk→ ∞, h ∈ Γ, t ∈ K
n=1
and for every k ∈ N

1
k
X
f (t + h) − Un (t, h) ≤ Vk (t), t ∈ K, h ∈ Γ, khk≥ r(k, K) .
ck (h) n=1

Then for f ∈ F0′ , we have
X∞
s
f (t + h) ∼ Un (t, h)|{cn (h)}, khk→ ∞, h ∈ Γ .
n=1
Proposition 1.9. Suppose that Γ has the nonempty interior. Let T ∈ F ′

and
∞
X
s
T (t + h) ∼ un (t)cn (h), khk→ ∞, h ∈ Γ .
n=1
If um 6= 0, m ∈ N, then um has the form
m
X
um (t) = Pkm (t) exp(ak · t), t ∈ Rn , m ∈ N ,
k=1
k
where a = (ak1 , . . . , akn )
∈ Rn and Pkm are polynomials with degrees less
than k at every ti , i = 1, . . . , n.
Proof. Definition 1.2 and the given asymptotics implies

lim T (t + h)/c1 (h) = u1 (t) 6= 0 in F ′ .
h∈Γ,khk→∞
Now Proposition 1.2 implies the explicit form of u1 .

The following limit gives u2 :
hT (t + h), ϕ(t)i − hu1 (t), ϕ(t)ic1 (h)
lim = hu2 , ϕi, ϕ ∈ F .
Note,
h(Dti − a1i )T (t + h), ϕ(t)i
lim = h(Dti − a1i )u2 (t), ϕ(t)i, ϕ ∈ F .
Two cases are possible.
a) If (Dti − a1i )u2 = 0, i = 1, . . . , n, then u2 (t) = C2 exp(a1 · t).
b) If (Dti − a1i )u2 6= 0 for some i, then by Proposition 1.2, (Dti −
a1i )u2 (t) = c exp(a2 · t) and u2 has the form
C2 exp(a1 · t) + P22 (t1 , . . . , tn ) exp(a2 · t) ,
where P22 is a polynomial of the degree less than 2 with respect to each
ti , i = 1, . . . , n.
In the same way, we prove the assertion for every um .
We will give an example of a function which has S-asymptotic expansion
of the first type but does not have the asymptotic expansion as a function:
Let ψ(t) = 1, t ∈ (n − 2−n , n + 2−n ), n ∈ N, and ψ(t) = 0 outside of
these intervals. Let
Z x
ψα (x) = eαx ψ(t)dt, x ∈ R, α ∈ R .
0
Rx
Since 0 ψ(t)dt → 2 as x → ∞, we have that ψα (x) ∼ 2eαx , x → ∞ but
ψα′ (x) does not have an ordinary asymptotic behavior (see Example 5. in
1.5.1).
Let hαj ij be a strictly decreasing sequence of positive numbers. Let θ
be a function, θ ∈ C ∞ , θ ≡ 1 for x > 1, θ ≡ 0 for x < 1/2 and f (x) =
P
∞
ψαi (x)θ(x − i), x ∈ R. We have
i=1
∞
X
f (x) ∼ ψαi (x)|{2 exp(αi x)}, x → ∞ .
i=1
This implies that f has the S-asymptotic expansion of the first type:
∞
X
s
f (x + h) ∼ ψαi (x + h)|{2eαi h }, h → ∞
i=1
and
∞
X
s
F (x + h) = f ′ (x + h) ∼ ψα′ i (x + h)|{2eαi h }, h → ∞
i=1
but F does not have the ordinary asymptotic expansion.

For S-asymptotic expansions see also ([108], [152]).
1.10.2 S-asymptotic Taylor expansion
Estrada and Kanwal have introduced in [53] and [54] the asymptotic Taylor
expansion of distributions (see also [56]). We generalize it to Fg′ .
Definition 1.5. Let f ∈ Fg′ and Dk f be the k-th partial derivative of f.

A formal series
∞
X Dk f (x)y k |k|
ε , where y ∈ Rn is fixed,
k!
|k|=0
is the asymptotic Taylor expansion of f, as ε → 0.
It means that for any test function φ ∈ Fg

N
X hDk f (x), φ(x)iy k |k|
hf (x + εy), φ(x)i − ε = O(εN +1 ), as ε → 0 .
k!
|k|=0
We write in short
∞
X ε|k|
f (x + εy) ∼ (Dk f (x)y k ) , ε → 0. (1.31)
k!
|k|=0
In fact, the asymptotic Taylor expansion is the S-asymptotic expansion
related to the asymptotic sequence {cn (h) = hn ; n ∈ N}, where h → 0,
with Un (x, h) = un (x)cn (h) and Γ = {εy; ε > 0}, where y ∈ Rn is fixed
(compare with Definition 1.2). We illustrate Definition 1.5 by two examples:
P∞ 1
1. δ(x + εy) ∼ (DN δ(x)y N )εN , as ε → 0,
|N|=0 N !
where εN = εN1 +···+Nn .
2. If α 6∈ Z, then the distribution xα ′
+ ∈ D (R) has the asymptotic Taylor
expansion
X∞
α
(x − ε)α
+ ∼ (−1)k xα−k k
+ ε , ε → 0,
k
k=0
This means that if φ ∈ D(R), then

Z∞ X∞ Z∞
α α k α−k
F.p. (x − ε) φ(x)dx ∼ (−1) F.p. x φ(x)dx εk , ε → 0 ,
k
0 k=0 0
where F.p. stands for the finite part (cf. [146] , [56]).
We now discuss the problem of the convergence of an asymptotic Taylor
expansion.
If a regular distribution is defined by a real analytic function f, then
the asymptotic Taylor expansion for f is a convergent power series in an
appropriate domain. But if f is only smooth, then the asymptotic Taylor
expansion for f does not necessarily have a domain of convergence. The
asymptotic Taylor expansion of δ distribution, given in Example 1, is not
convergent in D′ (R); we know that the series Σak δ (k) diverges unless ak =
0, kkk≥ n0 ∈ N.
One could attempt to solve this problem by trying to interpret the
asymptotic Taylor expansion for a distribution in a wider space F ′ g of
generalized functions asking the question: Find a necessary and sufficient
condition for the asymptotic Taylor expansion of a generalized function to
be convergent in Fg′ , i.e., so that it becomes its Taylor series.
In the following theorems we give the answer to this question if Fg′ is the
space of distributions (Fg′ = D′ ), the space of ultradistributions in which
∗
(M.1), (M.2) and (M.3) are satisfied (see 0.5.2) (Fg′ = D′ ) and the space
of Fourier hyperfunctions (Fg′ = Q).
∗
Theorem 1.13. ([163]). Let f ∈ D′ (f ∈ D′ ) and y = (y1 , . . . , yn ) yi 6=
0, i = 1, . . . , n. The asymptotic Taylor expansion for f , on the straight line
′
{εy; ε ∈ R}, is a convergent Taylor series in D′ (in D ∗ ) if and only if
there exists an r = (r1 , . . . , rn ), ri > 0, i = 1, . . . , n, such that f is a real
analytic function on Rn which can be extended as a holomorphic function
on {z ∈ Cn ; |Im zi |< ri , i = 1, . . . , n}.
Proof. Suppose that f ∈ D′ (f ∈ D′ ∗ ) and that w ∈ D (w ∈ D∗ ). Then
hf (x + εy), w(x)i = (f ∗ w̌)(εy), y ∈ Rn , ε ∈ R (1.32)
and
(f ∗ w̌)(εy + y0 ) = (f (t) ∗ w̌(t − y0 ))(εy), y, y0 ∈ Rn , (1.33)
where ∗ is the sign of convolution and w̌(t − y0 ) = w(−t + y0 ). This shows

that the expansion in a polydisc around y0 can be transferred to the ex-
pansion around 0.
In the sequel we shall use the following results proved in ([146], Chap.
VI, Theorem XXIV and [161], Theorem 1):
′
The generalized function f ∈ D′ (f ∈ D ∗ ) is a real analytic function if
and only if f ∗ w is real analytic for every w ∈ D (w ∈ D∗ ).
⇐ Suppose that there exists r = (r1 , . . . , rn ), ri > 0, i = 1, . . . , n such
that f is holomorphic in {x ∈ Cn ; |Im zi |< ri , i = 1, . . . , n}. Then its
Taylor series converges in D(a, r) = D(a1 , r1 ) × · · · × D(an , rn ) for every
a ∈ Rn . Therefore for every r′ = (r1′ , . . . , rn′ ), 0 < ri′ < ri , i = 1, . . . , n
and every compact set K ⊂ Rn , by the characterization of a real analytic
function,
|f (k) (x)|≤ C(r′ )−k k! , k ∈ Nn0 , x ∈ K .
Let us prove that

∞
X f (k) (x) ′′ k
(r ) (1.34)
k!
|k|=0
′
converges in D′ (in D∗ ) for any r′′ , 0 < ri′′ < ri′ , i = 1, . . . , n. It is enough
to prove that for every w ∈ D (w ∈ D∗ ) and every p ∈ N
* N +p +
X f (k) (x)
′′ k
lim (r ) , w(x)
N →∞ k!
|k|=N
N
X +p Z
= lim (f (k) (x)/(k! ))w(x)dx(r′′ )k = 0 .
N →∞
|k|=N Rn
Since for every w ∈ D (w ∈ D∗ )

Z Z

(f (x)/(k! ))w(x)dx ≤ C(r′ )−k |w(x)|dx, |k|≥ n0 (K) ,
(k)

Rn Rn
the series (1.34) converges for every r′′ , 0 < ri′′ < ri′ . Consequently, there
exists ε 6= 0 such that 0 < |εyi |< ri and that series (1.31) converges in D′
′
(in D ∗ ).
′
⇒ Suppose now that the series (1.31) converges in D′ (in D ∗ ) for a
fixed y ∈ Rn and a fixed ε > 0. Then for every w ∈ D (w ∈ D∗ )
∞
X hDk f (x), w(x)i
(yε)k
k!
|k|=0
is a convergent series in the polydisc D(0, r), where ri = yi ε, i = 1, . . . , n.

By (1.33), (f ∗ w)(z) is a holomorphic function on {z ∈ Cn ; |Im zi |< ri , i =
1, . . . , n}. By the cited theorem, f is a real analytic function. We now
continue the proof dividing it into two cases.
Case f ∈ D′ .
By (VI, 6; 22) in [146]
f = ∆k (gE ∗ f ) − (v ∗ f ) , (1.35)
where ∆ is the Laplacian, v ∈ D, and k is large enough so that for a fixed
m, gE ∈ DVm , where V is a relatively compact neighborhood of zero.
By the property of f ∗ w we just proved, it follows that for every w ∈ D
and for every y0 ∈ Rn there exists M > 0 such that
(r′′ )k |(f (k) ∗ ω)(y0 )|≤ M k! ,
where k ∈ Nn0 and 0 < r′′ < r. By Theorem XXII Chap. VI in [146], there
exists m ≥ 0 such that this inequality is also true if w = w̃ ∈ DVm , where V
is a relatively compact neighborhood of zero. Consequently, it follows that

f ∗ w̃ is a holomorphic function in B(y0 , r). By (1.35), f is also holomorphic
in B(y0 , r) for every y0 ∈ Rn .
′
Case f ∈ D ∗ . In the proof of this case we use the following Theorem
([81]):
Let the sequence Mp satisfy conditions (M1), (M2), and (M3). For
a given Hp and a compact neighborhood Q of zero in Rn there exist an
ultradifferential operator P (D) of class ∗ and functions ϕ ∈ C ∞ , and w ∈
∗
DQ such that
P (D)ϕ = δ + w
suppϕ ⊂ Q, sup |ϕ(k) (x)|/H|k| M|k| → 0, |k|→ ∞ .

x∈Q
Following the proof of Theorem 1 in [161], one can conclude that if

|(Dk f ∗ w)(y0 )|= |(f ∗ w)(k) (y0 )|≤ Ck! (r′ )−k ,
where w ∈ D∗ and k ∈ Nn0 , then there exists Hp such that the same
H M H M
inequality holds for w = w̃ ∈ D̃Kp p , where D̃Kp p is the completion of
∗
DK under the norm gHp Mp . Therefore f ∗ ϕ is a holomorphic function in
B(y0 , r).
By the cited theorem
f = P (D)(ϕ ∗ f ) + f ∗ w , (1.36)
where ϕ ∗ f and f ∗ w are holomorphic in a ball around any point of Rn .
So the proof will be finished if we prove that for a real analytic function θ
and an ultradifferential operator P (D) of (*)-class, P (D)θ is real analytic.
Therefore, we will prove first the next assertion.
Lemma 1.3. Let P (D) be an ultradifferential operator of (*)-class and θ

be real analytic in a neighborhood of x0 ∈ Rn . Then P (D)θ is real analytic
in a neighborhood of x0 .
Proof. We will prove the assertion in the case n = 1 and P (D) being of
{Mp }-class which is equivalent to P (D) = Σ ak Dk , where ak ∈ C, k ∈ N0
and there exist C > 0 and h > 0 such that
Chk
|ak |≤ , k ∈ N0 .
Mk
We have to prove that there exists H > 0 such that in a ball Br = {x; |x −
x0 |≤ r}
H α |(P (D)θ)(α) (x)|
sup < ∞.
α∈Nn
0 ,x∈Br
Mα
We have (x ∈ Br , α ∈ N0 )

1 α α X Hα X
∞ ∞
k α+k
H D ak D θ(x) = ak D θ(x)
Mα Mα
k=0 k=0
∞
!
CC1 H α X hk k+α
≤ C (k + α)! ,
Mα Mk 1
k=0
where we have used |Dk θ(x)|≤ C1k+1 k! , x ∈ Br which holds for some
C1 > 0. Since (k + α)! ≤ ek+α k! α! , we continue
∞
X ∞
H α hk C1k+α k! (C1 e)α α! H α X (C1 h)k k!
≤ CC1 eα+k ≤ C̃ sup .
Mα Mk α Mα Mk
k=0 k=0
The right side is bounded for every H > 0. This proves the lemma and
consequently the Theorem.
Theorem 1.14. ([164]) Let q ∈ Q, ξ = (ξ1 , . . . , ξn ), ξi > 0, i = 1, . . . , n

and ε > 0. The asymptotic Taylor expansion (1.31) for q on the straight
line {hξ; h ∈ R} is a convergent Taylor series in the topology of Q if and
only if there exists an r = (r1 , . . . , rn ), ri > 0, i = 1, . . . , n, such that q is
given by a real analytic function which can be extended as a holomorphic
function on {z ∈ Cn ; |Im zi |< ri , i = 1, . . . , n}.
Proof. In the first part of the proof, we suppose that series (1.31)
converges in Q for a fixed y = ξ = (ξ1 , . . . , ξn ), ξi > 0, i = 1, . . . , n, and a
fixed ε > 0. Since Q is an F S-space it is equivalent to suppose that series
(1.31) converges weakly in Q.
By Theorem 1.3 and Remark 1.4 in [73], there exists an elliptic local
operator J1 (D) and an infinitely differentiable function g rapidly decreasing
(|g(x)|≤ C exp(−αkxk), x ∈ Rn for some α > 0) such that
δ = J1 (D)g (δ is the delta distribution) .
Also (cf. Theorem in [74]) for every Fourier hyperfunction q ∈ Q, we

can find an elliptic local operator J2 (D) and an infinitely differentiable
function f of infra-exponential growth such that q = J2 (D)f. Then, by the

properties of the convolution:
q = (J2 (D)f ) ∗ (J1 (D)g) = J2 (D)J1 (D)((ℓf ) ∗ (ℓg)) , (1.37)
where J(D) = J2 (D)J1 (D) is also an elliptic operator. (ℓf denotes the
hyperfunction defined by the function f ).
From the proof of two mentioned theorems and by Theorem 8.2.6 in
[75] it follows that there exist two sets of functions {Fσ ; σ ∈ Λ} and
{Gσ̃ ; σ̃ ∈ Λ} such that:
a) Fσ ∈ Õ(Dn + iIσ ) and Gσ̃ ∈ Õ−α (Dn + iIσ̃ ), α > 0, σ, σ̃ ∈ Λ;
b) the functions Fσ and Gσ̃ can be extended to the real axis Rn as
infinitely differentiable functions Fσ (x) and Gσ̃ (x), respectively. Fσ (x) are
infra-exponential and Gσ̃ (x) are rapidly decreasing, σ, σ̃ ∈ Λ.
P P
c) f (x) = Fσ (x) and g(x) = Gσ̃ (x), x ∈ Rn ; g is analytic
σ∈Λ σ̃∈Λ
outside {0}.
By Carlemann’s theorem, we have
X X
f (x) := Fσ (x + iΓσ 0), g(x) := Gσ̃ (x + iΓσ̃ 0), x ∈ Rn
σ∈Λ σ̃∈Λ
(cf. [75] Lemma 8.4.7, and [145], §7).

Our first step is to prove that the Fourier hyperfunction (ℓf ) ∗ (ℓg) is
defined by the infra-exponential function
Z
f (x − u)g(u)du, x ∈ Rn .
Rn
By the definition of the convolution, we have

XX Z
(lf ) ∗ (lg)(u) := Fσ (z − w)Gσ̃ (w)du, w = u + iv, vσ̃ ∈ Iσ̃ ,
σ∈Λ σ̃∈ΛImw=v
σ̃
(1.38)
where z − w ∈ Rn + iΓσ 0. Hence z can move inside Rn + i(Γσ + Γσ̃ )0.
We can shift the integral path in the last integral to the real axis Rn . This
change of the path is justified by Cauchy’s integral formula and the growth
rate of the functions f and g at infinity. Then we have
XX Z
(ℓf ) ∗ (ℓg)(x) := Fσ (z − u)Gσ̃ (u)du .
σ σ̃ Rn
Every integral in the last sum is a function into Õ(Dn + i(Γσ + Γσ̃ )0),
respectively, and can be extended to the real axis as a slowly increasing
continuous function. Therefore by the same Carlemann’s theorem
!
XX Z
(ℓf ) ∗ (ℓg)(x) = ℓ Fσ (x − u)Gσ̃ (u)du
σ σ̃ Rn
Z !
=ℓ f (x − u)g(u)du . (1.39)
Rn
Hence, the first step is proved.
In the second step, we shall prove that (lf ) ∗ (lg)(x) is a real analytic
function in a neighborhood of zero, that is it can be extended on a complex
neighborhood of zero as a holomorphic function.
By definition of an infinitesimal wedge Uσ1 of type Dn + Γσ1 0 for every
proper subcone Γ′σ1 ⊂⊂ Γσ1 there exists w > 0 such that Uσ1 ⊃ Dn +i(Γ′σ1 ∩
{|y|< w}). Put Ǧ(u) = G(−u + iw/2). Then Ǧ ∈ Õ−α (Dn + i{|v|< w/2})
and Ǧ ∈ P∗ , as well.
Let us consider an addend in (1.38) and the corresponding integral in
(1.39). We have
Z
1 1
ℓ(Fσ ∗ Gσ̃ )(x) := Fσ z − u + ivσ̃ Gσ̃ (u + ivσ̃ )du, x ∈ Rn ,
2 2
Rn
where vσ̃ ∈ ∩ {|y|< 12 w}. Thus, for x ∈ Rn ,
Γ′σ̃
Z
1
(Fσ ∗ Gσ̃ ) x + ivσ̃ = Fσ (x − u)Ǧσ̃ (−u)du
2
R n
Z
= Fσ (x + u)Ǧσ̃ (u)du
Rn
Z
1 1
= Fσ x + u + ivσ̃ Ǧσ̃ u + ivσ̃ du .(1.40)
2 2
Rn
By (1.38)–(1.40), we have
X
1
((ℓf ) ∗ (ℓg)) x + ivσ̃ = hℓf (x + u), Ǧσ̃ (u)i, Ǧσ̃ ∈ P∗ .
2 σ̃
Let ε > 0 be fixed. The assumption of the theorem, implies that

hq(x + u), Gσ̃ (u)i, Gσ̃ ∈ P∗ , as a function in x ∈ Rn , can be extended
on B(0, εξ) ⊂ Cn as a holomorphic function. Then

hq(x + u), Gσ̃ (u)i = hJ2 (D)f (x + u), Gσ̃ (u)i
= J2 (D)hf (x + u), Gσ̃ (u)i, x ∈ Rn .
Since J2 (D) is an elliptic local operator, it follows that for every σ̃ ∈
Λ, hf (· + u), Gσ̃ (u)i is also real analytic. Consequently, ((ℓf ) ∗ (ℓg)) is
holomorphic in B(0, β) for β, βi > 0, i = 1, . . . , n.
Now it is easy to prove that ((ℓf ) ∗ (ℓg))(z) is holomorphic on Rn +
B(0, β).
In order to finish the first part of the proof, we have only to remark that
from (1.37) and from the property of an elliptic local operator it follows that
q is a real analytic function which can also be extended as a holomorphic
function on Rn + B(0, β).
For the second part of the proof suppose that q is given by a real analytic
function which can be extended on Rn + B(0, β), βi > 0, i = 1, . . . , n, as
a holomorphic function. Then for ε > 0 and ξ ∈ Rn ,
ℓq(εξ + x) := q(εξ + x + iIσ0 ), x ∈ Rn
and, for y0 ∈ Iσ0 , Z
hℓq(εξ + x), ϕ(x)i = q(εξ + x + iy0 )ϕ(x + iy0 )dx, ϕ ∈ P∗ .
Imz=y0
By the same arguments which we have used to transform (1.38) into

(1.39), we have
Z
hℓq(εξ + x), ϕ(x)i = q(εξ + x)ϕ(x)dx, ϕ ∈ P∗ .
Rn
Since q(z) is holomorphic on Rn + B(0, β), the same holds for hℓq(εξ +
x), ϕ(x)i.
For asymptotic Taylor expansions, see also [163], [1], [50], [56].
1.11 S-asymptotics in subspaces of distributions
We discuss in this section the following problem: Let A′ be a subspace of

D′ . If T ∈ A′ and if T has the S-asymptotics in D′ , is it true that T has the
S-asymptotics in A′ , as well? The answer is not simple. We shall analyze
two cases, A′ = S ′ and A′ = K′ 1 . First, we shall illustrate the problem by
an example.
The distribution H(t) exp(−t) (H is Heaviside’s function) belongs to

S (R). It has S-asymptotics in D′ (R) related to c(h) = exp(−h) with limit
′
U = exp(−t). We know that exp(−t) does not belong to S ′ (R) and D(R)
is dense in S(R). Hence H(t) exp(−t) cannot have the S-asymptotics in
S ′ (R).
To answer our question, we need to introduce some additional conditions
over c(h).
Although we defined the S-asymptotics of distributions in a general cone
Γ, we shall restrict ourselves to the cone Γ = Rn+ for the sake of simplicity.
Recall, for a, b ∈ Rn , a ≥ b means ai ≥ bi , i = 1, . . . , n.
The set of real-valued functions h → c(h), h ∈ Rn , defined on Rn ,
different from zero for h ∈ Rn+ , is denoted by Σ(Rn ). We assume, without
losing generality, that function c in Σ(Rn ) are positive and equal to 1 in
Rn+ \ (Rn+ + a), where a ∈ Rn+ depends on c and Rn+ + a = {x + a; x ∈ Rn+ }.
By Σe (Rn ) is denoted the subset of Σ(Rn ) such that c ∈ Σe (Rn ) if and
only if for some C > 0, d > 0, k > 0 and h0 = (h0,1 , . . . , h0,n ) ∈ Rn+
(E.1) c(h + r) ≤ Cc(h) exp(kkrk), h > h0 , r ∈ Rn .
(E.2) c(h) exp(kkhk) ≥ d, h > h0 .
By Σp (Rn ) is denoted another subset of Σ(Rn ) defined as follows,
c ∈ Σp (Rn ) if and only if for some C > 0, d > 0, k > 0 and h0 =
(h0,1 , . . . , h0,n ) ∈ Rn+ :
(P.1) c(h + r) ≤ Cc(h)(1 + krk)k , h > h0 , r ∈ Rn .
(P.2) c(h)(1 + khk)k ≥ d, h > h0 .
Obviously, Σp (Rn ) ⊂ Σe (Rn ).
We begin with some properties of the sets Σe (Rn ) and Σp (Rn ), we
restrict our attention to n = 1 in order to analyze explicit representations
for their elements.
The set Σp (R).
By Remark 3 after Proposition 1.2, a function c from Definition 1.1 has
the form c(h) = exp(αh)L(exp(h)), h > h0 > 0, where α ∈ R and L is a
slowly varying function, and the limit distribution U (x) = C exp(αx). For
the S-asymptotics in S ′ , α has to be zero: Then c(h) = L(exp h), h > h0 .
Proposition 1.10. Suppose that c ∈ Σ(R).

s
a) If there exists T ∈ S ′ (R) such that T (x + h) ∼ c(h)U (x), h ∈ R+
with U 6= 0, then c(h) = L(exp h) where L is a slowly varying function, and
so c satisfies (P.2).
b) If c = hν L(h), h > h1 and c(h) = 1, h ≤ h1 , where ν ∈ R, h1 > 0 and
L is a slowly varying and monotonous function for h > h1 , then c satisfies
(P.1).
c) If c is of the same form as in b), where L is only slowly varying, then
c satisfies (P.1) but for r ∈ R+ .
For the proof cf. [107].

The set Σe (R).
Proposition 1.11. Suppose that c ∈ Σ(R).

s
a) If there exists T ∈ D′ (R) such that T (x + h) ∼ c(h)U (x), U 6= 0, then
(E.2) holds for c.
b) If c(h) = exp(α|h|)hν L(h), h > h1 , c(h) = 1, h ≤ h1 , where α, ν ∈
R, h1 > 0 and L is a monotonous slowly varying function for h > h1 , then
(E.1) holds for c.
c) If c is of the same form as in b), where L is only a slowly varying
function, then c satisfies (E.1) but for r ∈ R+ .
For the proof cf. [107].

We prove now that if T ∈ K′ 1 (R) (T ∈ S ′ (R)) has the S-asymptotics
P P
in the cone R+ related to some c(h) ∈ e (R) (c(h) ∈ p (R)) with limit
U in the space D′ (R), then the limit
lim T (x + h)/c(h) = U (x)
h→∞
also exists in K′ 1 (R) (in S ′ (R)). If n > 1, we have to assume some addi-
tional conditions which imply the same assertion for the multidimensional
case.
First, we shall prove a theorem which extends Proposition 1.7 b) and
which will be used later.
Theorem 1.15. Let T ∈ D′ . If for every rapidly exponentially decreas-

ing function r (for every k > 0, r(x) exp(kkxk) → 0 as kxk→ ∞) the set
{r(h)T (x + h); h ∈ Rn } is bounded in D′ , then T ∈ K′ 1 .
Proof. Let K be an arbitrary compact set in Rn . For every φ ∈ DK ⊂ D,

h 7→ r(h)hT (x + h), φ(x)i, h ∈ Rn ,
is a bounded function. This implies that for some k1 = k1 (φ) > 0 and some
C = C(φ) > 0
|hT (x + h), φ(x)i|≤ C exp(k1 khk), h ∈ Rn . (1.41)
If h ∈ Rn is fixed, then
log+ |hT (x + h), φ(x)i | /(1 + khk), φ ∈ DK ,
a defines a continuous function on DK . It follows from (1.41) that {φ 7→
log+ |< T (x + h), φ(x) > |/(1 + khk); h ∈ Rn } is a bounded family of
continuous functions on DK ; from the classical theorem of Baire, it follows
the existence of some k > 0, which does not depend on φ ∈ DK , such that
the set {hT (x+ h) exp(−kkhk), φ(x)i; h ∈ Rn } is bounded for each φ ∈ DK ,
and hence, for each φ ∈ D. This implies (see Theorem XXII, Chapter VI
in [146]), that for a given open bounded set Ω ⊂ Rn , 0 ∈ Ω, there exists a
m
compact neighborhood of zero K and m ∈ N0 such that for every φ ∈ DK ,
{x 7→ (T (t + h) ∗ φ(t))(x)/exp(kkhk); h ∈ Rn }
is a bounded family of continuous bounded functions on Ω.
Since (T (t + h) ∗ φ(t))(x) = (T ∗ φ)(x + h), setting x = 0, we obtain that
h 7→ (T ∗ ψ)(h)/exp(kkhk), h ∈ Rn , is a bounded function on Rn for any
m
ψ ∈ DK . Now, by (VI, 6; 22) in [146], we obtain
T = ∆N (γE ∗ T ) − ψ ∗ T , (1.42)
where E is the fundamental solution of ∆N E = δ (∆ is the Laplacian),
γ ∈ DK , γ ≡ 1 in a neighborhood of 0 and ψ ∈ DK . If N is sufficiently
m
large, γE ∈ DK . Thus, γE ∗ T and ψ ∗ T are in K′ 1 , and this completes the
proof.
P
Theorem 1.16. Let T ∈ K′ 1 (R) (T ∈ S ′ (R)) and c ∈ e (R) (c ∈
P ′
p (R)). a) If the set {T (x + h)/c(h); h > a} is bounded in D (R), then
′ ′
this set is bounded in K 1 (R) (in S (R)) as well.
b) If there exists the limit
lim hT (x + h)/c(h), ϕ(x)i = hS, ϕi, ϕ ∈ D(R),
h→∞
then this limit exists for every ϕ ∈ K1 (R) (for every ϕ ∈ S(R)). In partic-
ular, S ∈ K′ 1 (R)(S ∈ S ′ (R)).
Proof. We prove the theorem for T ∈ K′ 1 (R), because for T ∈ S ′ (R) it

can be done in a similar way. (We have to replace exp(kk·k) by (1 + k·k)k ).
a) Using the last part of the proof of Theorem 1.15, we obtain that for
some m1 ∈ N0 and some compact neighborhood of zero, K1 , h 7→ (T ∗
m1
ψ)(h)/c(h), h > 0 is a bounded function for every ψ ∈ DK 1
(R). Since
′
T ∈ K 1 (R), it holds that for some k > 0, m2 ∈ N0 and some compact
neighborhood of zero K2 , h 7→ (T ∗ ψ)(h)/exp(k|h|), h ∈ R, is a bounded
m2
function for every ψ ∈ DK 2
(R). Thus, by taking N in (1.42) sufficiently
2 2
large (∆ = d /dx ) and K = K1 ∩ K2 , we obtain that for some m ∈ N0
m
X (i)
T = Fi ,
i=0
where Fi , i = 0, . . . , m, are continuous functions on R such that for some

M1 > 0, M2 > 0 and k > 0
sup{|Fi (x)/c(x)|; x > 0, i = 0, . . . , m} ≤ M1 , (1.43)
sup{|Fi (x)|exp(k|x|); x ∈ R, i = 0, . . . , m} ≤ M2 ,
Let φ ∈ K1 (R) and h > h0 be fixed. We put

Z∞
Ii (h, φ) ≡ Ii = (|φ(i) (x)kFi (x + h)|/c(h))dx
−∞
Z−h Z∞ !
= + (|φ(i) (x)kFi (x + h)|/c(h))dx
−∞ −h
= Ii (−∞, −h) + Ii (−h, ∞), i = 0, . . . , m, h > h0 .
If x ∈ (−∞, −h), then |x + h|= |x|−h and by (E.2), we obtain

Z−h
Ii (−∞, −h) ≤ d−1 |φ(i) (x)|exp(2k)|x|)|Fi (x + h)/exp(k|x + h|)|dx
−∞
Z∞
−1
≤ Md |φ(i) (x)|exp(k|x|)dx.
−∞
From the definition of the space K1 (R) it follows that the last integral
is finite.
Because of (E.1) and (1.43), we obtain that for some k1 > 0,

Z∞
Ii (−h, ∞) ≤ |φ(i) (x)kFi (x + h)/c(x + h)|exp(k1 |x|)dx
−h
Z∞
≤ M1 |φ(i) (x)|exp(k1 |x|)dx < ∞ .
−∞
Since
m
X
|hT (x + h)/c(h), φ(x)i|≤ Ii (h, φ),
i=0
from the preceding inequalities, we obtain that for some A > 0 which does
not depend on h > h0 and for φ ∈ K1 ,
|hT (x + h)/c(h), φ(x)i|≤ A, φ ∈ K1 , h > h0 .
Thus the proof of a) is complete.

b) If the limit given in b) exists, then for an a ∈ R, the set {T (x +
h)/c(h); h ≥ a} is bounded in K′ (R) (in S ′ (R)). Since D(R) is dense
in K1 (R) (in S(R)), it is now enough to use the assertion of a) and the
Banach–Steinhaus theorem.
If we assume more on a distribution T , then we can assume less on the
function c.
Theorem 1.17. Let T ∈ D′ (R) and supp T ⊂ [0, ∞). Suppose that c ∈
Σ(R) and satisfies (E.2) and (E.1) with r ∈ R+ (satisfies (P.2) and (P.1)
with r ∈ R+ ).
a) If the set {T (x + h)/c(h); h > a} is bounded in D′ (R), then this set
is bounded in K′ 1 (R) (in S ′ (R)), as well.
b) If there exists the limit
lim hT (x + h)/c(h), ϕ(x)i = hS, ϕi, ϕ ∈ D(R),

h→∞
then this limit exists for every ϕ ∈ K1 (R) (for every ϕ ∈ S(R)). In partic-
ular, S ∈ K′ 1 (R) (S ∈ S ′ (R)).

We give other versions of Theorem 1.17 because we want to empha-

size that the space S ′ (R) is “natural” for the distributions having the
S-asymptotics related to c(h) = hν L(h), h > 0, ν ∈ R, while the space
K′ 1 (R) is “natural” for those distributions having the S-asymptotics re-
lated to c(h) = exp(ah)L(exp(h)), h > 0, a ∈ R, where L is a slowly
varying function.
s
Theorem 1.18. Let T ∈ D′ (R), supp T ⊂ (−ω, ∞), ω > 0 and T (x + h) ∼
s
eah L(eh ) · Ceax , h ∈ R+ in D′ (R) (and T (x + h) ∼ hν L(h) · C, h ∈ R+
in D′ (R)), then T ∈ K′ 1 (R) (T ∈ S ′ (R)) and T has the S-asymptotics in
K′ 1 (R) (in S ′ (R)) related to the same c and with the same limit.
Proof. By Theorem 1.9 and Lemma 1.2 there exist functions Fi , i =

0, . . . , m, continuous on (−ω, ∞), such that
m
X
T = Di Fi on (−ω, ∞) ,
i=0
where
|Fi (x + h)/c(h)|≤ Mi , h ∈ R+ , x ∈ (−ω, ω), i = 0, . . . , m,
supp Fi ⊂ (−ω − δ, ∞), i = 0, . . . , m, δ > 0.
For a slowly varying function L there exists a slowly varying function

L∗ ∈ C(c,∞)
∞
(R), α > 0, such that L∗ (x)/L(x) → 1, x → ∞. We refer
to the Remark after Proposition 1.2 for the construction of L∗ . Therefore,
we can suppose, without any restriction, that L, which appeared in c, is
∞
in C(α,∞) (R), α > 0. Let us denote by Mi (h) = Fi (h)/c(h), h ∈ R; Mi
is a continuous and bounded function on R and Fi = cMi . This im-
plies that T ∈ K′ 1 . The next step is to prove that for h0 ∈ R the set
{T (x + h)/c(h); h ≥ h0 } is weakly bounded in K′ 1 which is equivalent to
the strong boundedness in K′ 1 .
We need the following inequality:
c(x + h)/c(h) = eαx L(ex eh )/L(eh ) ≤ Aeax+δ|x| , x, h ∈ R,
where δ > 0 (see [9], p. 25). Let ϕ ∈ K1 , then

m Z
X
|hT (x + h)/c(h), ϕ(x)i| ≤ |Fi (x + h)/c(h)kϕ(i) (x)|dx
i=0 R
m Z
X
≤ |Mi (x + h)c(x + h)/c(h)kϕ(i) (x)|dx
i=0 R
m Z
X
≤ eax+δ|x||ϕ(i) (x)|dx.
i=0 R
This proves that {T (x+h)/c(h); h ≥ h0 } is weakly bounded in K′ 1 . Since D

is dense in K′ 1 , the Banach–Steinhaus theorem implies the S-asymptotics
in K′ 1 related to c.
In order to extend Theorem 1.17 to the multidimensional case, we have
to introduce the following notation.
We denote by Λ the set of all n − th class variations of elements {−1, 1}.
If (a1 , . . . , an ) ∈ Λ, then we put
n
X
Γ(a1 , . . . , an ) = {h ∈ Rn ; sgn(ai · hi ) = n}.
i=1
(This means if ai = 1, (ai = −1), then hi > 0(hi < 0)). For example
Γ(1, . . . , 1) = Rn+ and Γ(−1, . . . , −1) = Rn− . Let
c(h) = c1 (h1 ) . . . cn (hn ), hi ∈ R, i = 1, . . . , n , (1.44)
where
ci ∈ Σe (R) (ci ∈ Σp (R)), i = 1, . . . , n.
Obviously, c ∈ Σe (Rn ) (c ∈ Σp (Rn )). Let (a1 , . . . , an ) ∈ Λ be given. We
denote by ji , i = 1, . . . , r the components of (a1 , . . . , an ) which are equal to
1, and those which are equal to −1 by si , i = 1, . . . , m (r + m = n).
Let h ∈ Γ(a1 , . . . , an ) and k > 0. We put
ck(a1 ,...,an ) (h) = cj (hj ) . . . cjr (hjr ) · exp(k(|hs1 |+ · · · + |hsm |))
Theorem 1.19. Let T ∈ K′ 1 (T ∈ S ′ ) and c be of the form (1.44).

a) If there exists k > 0 such that for every (a1 , . . . , an ) ∈ Λ
{T (x + h)/ck(a1 ,...,an ) (h); h ∈ Γ(a1 , . . . , an )}
is bounded in D′ , then {T (x + h)/c(h); h > 0} is bounded in K′ 1 (in S ′ ).
b) If
lim hT (x + h)/c(h), φ(x)i = hS(x), φ(x)i, φ ∈ D,
h→∞
and if for some k > 0 and every (a1 , . . . , an ) ∈ Λ\{(1, . . . , 1), (−1, . . . , −1)}
the sets
{T (x + h)/ck (a1 , . . . , an )(h); h ∈ Γ(a1 , . . . , an )},
are bounded in D′ , then T (x + h)/c(h) converges to S in K′ 1 (in S ′ ), as

h → ∞. In particular, S ∈ K1′ (S ∈ S ′ ).

Instead of analyzing the two special subspaces S and K1 of D we can
consider a general subspace A of D.
Let Γ be a convex cone. We denote Σq (Γ) a subset of Σ(Γ) such that
c ∈ Σq (Γ) if and only if there exist C > 0 and a positive locally integrable
function p such that
c(h + x) ≤ Cc(h)p(x), h, x ∈ Γ\B(0, r) . (1.45)
In the sequel, put G = {x ∈ Rn \(Γ ∪ B(0, r))}. We denote by A a barrelled

vector space of smooth functions such that D is dense in A with its topology
finer than the topology induced by A; A′ is the dual space of A, A′ ⊂ D′ .
We will suppose that the elements φ of A satisfy the following condition:
for every y ∈ B(0, r), p(x)φ(x + y) ∈ L1 .
Theorem 1.20. Suppose that T ∈ A′ and c ∈ Σq (Γ).

a) If the sets:
Q1 = {T (x + h)/c(h); h ∈ Γ}
Q2 = {T (x + k + h)/(c(h)p(k)); h ∈ Γ, k ∈ G}
are weakly bounded in D′ , then the set Q1 is weakly bounded in A′ as well.

b) If T has S-asymptotics in D′ related to c(h) with the limit U and if
the set Q2 is weakly bounded in D′ , then T has the S-asymptotics in A′
related to c(h) and with the limit U.
We remark that the well-known basic spaces as K1 , S, DLp (1 ≤ p < ∞)

·
and B satisfy our conditions assumed in Theorem 1.20 for the space A. The
space B = DL∞ is an example of one which does not satisfy them, due to
the fact that D is not dense in B.
If a distribution f has the S-asymptotic expansion in D′ , with adequate

additional conditions, it can have the S-asymptotic expansion in a subspace
of D′ . The following theorem gives such conditions.
Theorem 1.21. Suppose that f ∈ K′ 1 (R) (f ∈ S ′ (R)) and {ui } ⊂ K′ 1 (R)

({ui } ⊂ S ′ (R)). If
∞
X
s
f (x + h) ∼ ui (x + h)|{ci (h)}, h → ∞ in D′ (R)
i=1
and ci (h) = exp(αi h)Li (exp h) (resp. ci = hνi Li (h)), h > h0 ), where
Li , i ∈ N, are monotonous slowly varying functions, then
∞
X
s
f (x + h) ∼ ui (x + h)|{ci (h)}, h → ∞, in K′ 1 (R) (in S ′ (R)),
i=1
as well.
Proof. The proof is based on Proposition 1.11 and Theorem 1.18, we

leave the details to the reader.
The next two propositions give the relation between the asymptotic
behavior of functions and the S-asymptotics of distributions defined by
these functions.
Proposition 1.12. Let f ∈ L1loc (R) be such that f defines a regular distri-
bution in K′ 1 (R), i.e., f φ ∈ L1 (R) for every φ ∈ K1 (R). Further, assume
f (x) ∼ L(exp x) exp(αx) as x → ∞,
where L is a slowly varying monotonous function defined on (a, ∞). If
c(h) = L(exp h) exp(αh), h > A, and g(x) = exp(αx), x ∈ R, then
s
f (x + h) ∼ c(h)g(x), h ∈ R+ in K′ 1 .
Proof. Let φ ∈ D and supp φ ⊂ [a, b]. Since (L(λh)/L(h)) → 1, h → ∞,

uniformly on any compact interval contained in (0, ∞), we obtain
Z∞ Zb
f (x + h) f (x + h)
φ(x)dx =
exp(αh)L(exp h) exp(α(x + h))L(exp(x + h))
−∞ a
Zb
L(exp x · exp h)
· exp(αx) φ(x)dx → exp(αx)φ(x)dx, h → ∞ .
L(exp h)
a
The assertion follows now from Proposition 1.11 and Theorem 1.16.
Proposition 1.13. Suppose that the functions f and ui , i ∈ N, are locally

integrable and define regular distributions belonging to K′ 1 (R), i.e., f ϕ
and ui ϕ are in L1 (R) for every ϕ ∈ K1 (R). Let ci (h) = exp(αi h)Li (exp h),
where Li are monotonous slowly varying functions and αi ∈ R, i ∈ N. If
∞
X
f (x) ∼ Ui (x)| {ci (x)}, x → ∞,
i=1
then
∞
X
s
f (x + h) ∼ ui (x + h)|{ci (h)}, h → ∞ .
i=1
Proof. The proof is similar to that of Proposition 1.13.

′
We shall prove a structural theorem for a distribution T ∈ B having
the S-asymptotics in a cone Γ with the nonempty interior.
s
Theorem 1.22. Suppose T0 ∈ B ′ and T0 (x + h) ∼ 1 · U (x), h ∈ Γ in D′ ,
then
a) U = C.
P
2
b) T0 = ∆ik Fi , where Fi are continuous functions belonging to L∞ ;
i=0
c) For every 0 ≤ i ≤ 2 functions Fi (x + h) of part b) converge uniformly
to a constant when x belongs to a compact set K and h ∈ Γ, khk→ ∞.
d) T0 has the S-asymptotics in B ′ , related to c = 1 and with the limit
U = C in the cone Γ.
Proof. a) U has to be a constant because c = 1 (Proposition 1.2).

b) From the fact that T0 ∈ B ′ it follows that (T0 ∗ ζ̌) = hT0 (x+·), ζ(x)i ∈
∞
L for every ζ ∈ D (see [146], VI, §8) and the set of distributions Q =
{Th ≡ T0 (x + h); h ∈ Rn } is weakly bounded and bounded in D′ .
We will construct another bounded set of distributions. Denote S =
{ψ ∈ D; kψkL1 ≤ 1}. We have seen that for a fixed ζ ∈ D, (T0 ∗ ζ̌) ∈ L∞ .
Now, for every ψ ∈ S :
Z
|hT0 ∗ ψ̌, ζi|= |hT0 ∗ ζ̌, ψi|= | (T0 ∗ ζ̌)(t)ψ(t)dt| ≤ kT0 ∗ ζ̌kL∞ kψkL1 .
R
Hence, the set of regular distributions, defined by the set of continuous

functions {Uψ ≡ T0 ∗ ψ̌; ψ ∈ S} is weakly bounded and therefore bounded
in D′ .
A set W ′ ∈ D′ is bounded if and only if for every α ∈ D the set of
functions {T ∗ α; T ∈ W ′ } is bounded on every compact set M belonging
to Rn (see [146], VI, §7 ). Hence {T ∗ α; T ∈ W ′ } defines a bounded set of
regular distributions. In such a way {Th ∗ζ; Th ∈ Q} and {Uψ ∗ζ; Uψ ∈ H}
are bounded sets of regular distributions. Now, for these two sets we can
repeat twice a part of the proof of Theorem XXII, Chapter VI in [146].
We denote by Ω an open neighborhood of zero in Rn which is relatively
compact in Rn , namely, Ω̄ = K is a compact set. Then, by the mentioned
part of the proof, there exist m1 ≥ 0 and m2 ≥ 0, such that the mappings
(α, β) → Uψ ∗ (α ∗ β) or (α, β) → Th ∗ (α ∗ β), h ∈ Rn , are equicontinuous
m1 m1 m2 m2
and map DΩ × DΩ and DΩ × DΩ into L∞
B respectively; B is the ball
B(0, r), where r is a positive number. Hence, for every x ∈ B and h ∈ Rn
the function (Th ∗ α ∗ β)(x) = (T0 ∗ α ∗ β)(x + h) is continuous.
Let Z(0, ρ) be a ball in L∞ B . Then there exists a neighborhood V1 (m1 , ε1 ,
m1
K1 ) in DΩ , such that Uψ ∗(α∗β) ∈ Z(0, ρ) for α, β ∈ V1 (m1 , ε1 , K1 ), Uψ ∈
m2
H and a neighborhood V2 (m2 , ε2 , K2 ) ⊂ DΩ , such that Th ∗(α∗β) ∈ Z(0, ρ)
for α, β ∈ V2 (m2 , ε2 , K2 ), Th ∈ Q. Let K0 = K1 ∩ K2 , ε0 = min(ε1 , ε2 ) and
m = max(m1 , m2 ). We shall use (VI, 6; 23) in [146].
T0 = ∆2k ∗ (γE ∗ γE ∗ T0 ) − 2∆k ∗ (γE ∗ ξ ∗ T0 ) + (ξ ∗ ξ ∗ T0 ) ,
where E is a solution of the iterated Laplace equation; ∆k E = δ; γ, ξ ∈

DΩ , supp γ and supp ξ belonging to K0 = K1 ∩ K2 . We have only to
m
choose the number k large enough so that γE ∈ DΩ . Now, we can take:
F2 = γE ∗ γE ∗ T0 ; F1 = γE ∗ ξ ∗ T0 and F0 = ξ ∗ ξ ∗ T0 . All of these
functions are of the form: Fi = T0 ∗ αi ∗ βi ; αi , βi ∈ V (m, ε′0 , K0 ), ε′0 > 0.
We have to prove that Fi , i = 0, 1, 2, have the properties given in
Theorem 1.22. For αi , βi ⊂ V (m, ε′0 , K0 ) and ψ ∈ S
|h(T0 ∗ αi ∗ βi ), ψi|= |[T0 ∗ ψ̌) ∗ (α̌i ∗ β̌i )](0)|≤ ρ(ε′0 /ε0 )2 ≡ M .
Now let µ 6= 0 be any element in L1 . Then µ/kµkL1 ∈ S and |h(T0 ∗ (αi ∗

βi )), µi|≤ M kµkL1 which proves that T0 ∗ (αi ∗ βi ), i = 0, 1, 2, belong to
L∞ . Since Fi = T0 ∗ (αi ∗ βi ), αi , βi ∈ V (m, ε′0 , K0 ), Fi , i = 0, 1, 2, are
continuous and belong to L∞ .
c) We shall continue the investigations of the properties of Fi . By the

properties of the convolution, we have
m
Fi (x + h) = Fi ∗ τ−h = T0 ∗ (αi ∗ βi ) ∗ τ−h = Th ∗ (αi ∗ βi ), αi , βi ∈ DΩ ,
where τ−h is the translation operator.

We have proved that the mappings (α, β) → Th ∗ α ∗ β, Th ∈ Q, are
m m
equicontinuous and map DΩ ×DΩ into L∞ m
B . D is a dense subset of D , m ≥
m
0. We can construct a subset of DK , clΩ = K, which is dense in DΩ . Since
Th ∗ (ζ ∗ ψ) → C ∗ ζ ∗ ψ for ζ ∗ ψ ∈ DΩ × DΩ , then Th ∗ αi ∗ βi converges to
C ∗ αi ∗ βi , as well (see [146], VI, §7),when khk→ ∞, h ∈ Γ, i = 0, 1, 2.
d) It remains to prove the last part of Theorem 1.22. For µ ∈ DL1 and
T ∈ B, we have:
2 Z
X X2 Z
ik
|hT0 (x + h), µ(x)i|≤ |Fi (x + h)∆ µ(x)|dx ≤ Mi |∆ik µ(x)|dx,
i=0 Rn i=0 Rn
where Mi = sup|Fi (x)|, x ∈ Rn . Hence the set {T0 (x + h), h ∈ Rn }, is

weakly bounded in B ′ . Since D is dense in DL1 , by the Banach–Steinhaus
theorem the limit:
lim hT0 (x + h), µ(x)i, µ ∈ DL1 ,

h∈Γ,h→∞
exists, as well, and it is equal to hC, µi.
1.12 Generalized S-asymptotics
Definition 1.1 corresponds to some subsets of distributions of a limited

growth (see Theorem 1.16). This is the reason for introducing a general-
ization of the S-asymptotics ([119]).
Definition 1.6. Suppose that e is a function on Rn such that e−1 ∈ M(·)

(see 0.6). Then, it is said that T ∈ Fg′ has generalized S-asymptotics in
the cone Γ, related to e, if there exists
w. lim T (x + h)/e(x + h) = 1 in Fg′ .

khk→∞,h∈Γ
gs
We write in short T (x) ∼ e(x), kxk→ ∞, x ∈ Γ.
If we compare Definition 1.1 and Definition 1.6, we obtain.
Proposition 1.14. Let e be a function on Rn such that e−1 ∈ M(·) . Then,

T ∈ Fg′ has generalized S-asymptotics in Γ, related to e if and only if T /e
has the S-asymptotics in Γ related to c = 1 and with limit U = 1.
In such a way, dividing T ∈ Fg′ by e, we may apply our results on

the S-asymptotics to this new context. So, all the assertions on the S-
asymptotics can be transferred for the generalized S-asymptotics by simply
using Proposition 1.14. For this reason, we underline only some of the key
properties of generalized S-asymptotics.
The generalized S-asymptotics is a local property, as well (see Theorem
1.4). From Theorems 1.2,d) and 1.3,4) it follows:
gs
Proposition 1.15. a) Let S ∈ E ′ and T ∈ D′ . If T (x) ∼ e(x), kxk→
s
∞, x ∈ Γ, then (S ∗ (T /e))(c + h) ∼ 1 · (S ∗ 1)(x), h ∈ Γ.
gs
As consequence of T (x) ∼ e(x), kxk→ ∞, x ∈ Γ, it follows that for any
partial derivative Dxi
s
(Dxi (T /e))(x + h) ∼ 1 · 0, khk→ ∞, h ∈ Γ .
In order to compare the existence of the generalized S-asymptotics with

that of the S-asymptotics of a T ∈ F ′ , we need the following notation and
lemma.
Let Γ be a cone with the vertex at zero. We employ the notation pr Γ
for the intersection of Γ and the unit sphere in Rn .
Lemma 1.4. Let T ∈ F ′ Rand let c ∈ Σ(Γ) where Γ is a convex cone with
the nonempty interior, ( Γ 6= θ). Let Γ′ be a closed cone, Γ′ ⊂ int Γ.
s
Suppose that T (x + h) ∼ c(h) · U (x), h ∈ Γ with U 6= 0. Then, there exist
e ∈ C ∞ , positive on Rn , and a ∈ Rn such that
lim c(h)/e(x + h) = exp(−(a · x)) in E . (1.46)
h∈Γ′ ,khk→∞
Proof. We know (see Proposition 1.2) that U (x) = C exp(a · x), a ∈ Rn ,

C 6= 0. Let φ0 ∈ F such that hU, φ0 i 6= 0. We introduce functions
e1 , e2 , e3 , e4 and e in the following way: e1 is a smooth function
e1 (y) = hT (x + y), φ0 (x)i/hU, φ0 i, y ∈ Rn .
Since
lim hT (x + h)/c(h), φ0 (x)i = hU, φ0 i,
h∈Γ,khk→∞
there exists β0 > 0 such that e1 (y) > 0, y ∈ {z ∈ Γ′ , kzk≥ β0 }.

Functions e2 , e3 and e4 are defined as follows: e2 (x) = max{e1 (x), 0},
x ∈ Rn , e3 is the characteristic function of the set CRn {x ∈ Γ′ , kxk≥ β0 },
e4 = e2 + e3 . Thus e4 is positive and locally integrable.
Let w be a non-negative function belonging to C ∞ such that w(x) = 1
for x ∈ B(0, 1), w(x) = 0 for kxk> 2, and
Z
exp(−(a · x))w(x)dx = 1.
Rn
Now, we can construct the sought function e ∈ C ∞ : e = (e4 ∗ w)(x),

x ∈ Rn . The function e(x) is positive because of
Z Z
e(x) = e4 (x − t)w(t)dt ≥ e4 (x − t)dt > 0, x ∈ Rn .
Rn B(0,1)
It only remains to prove (1.45). The number m := d(pr Γ′ , CRn int Γ) is

positive because pr Γ′ is a compact set in Rn and CRn int Γ is a closed one.
Every ball B(b, m′ ), 0 < m′ < min (m, 1), b ∈ pr Γ′ is contained in int Γ
and for every λ > 0 the ball B(λb, λm′ ), 0 < m′ < min(m, 1), is contained
in int Γ, as well.
For every compact set K ⊂ Rn , K ⊂ B(0, ρ) and every β0 ∈ R+ there
exists β > 0 such that x + h ∈ {y ∈ Γ′ ; kyk≥ β0 } for x ∈ K and h ∈ {y ∈
Γ′ , {yk> β}. Since h ∈ Γ′ , h can be written as h = λb, b ∈ pr Γ′ , λ > 0.
Thus B(λb, λm′ ) ⊂ Γ for every λ > 0. If λ > ρ/m′ , then x + h ∈ B(λb, λm′ )
for x ∈ K ⊂ B(0, ρ) and h = λb because kx + h − λbk= kxk≤ ρ < λm′ .
Now,
kx + hk≥ |kxk−khk|≥ λ(1 − m′ ).
Then, we can take β = max{ρ/m′ , β0 /(1 − m′ )}.

For a compact set K0 ⊂ Rn the set K = K0 \B(0, 2) is also a compact
set and belongs to a ball B(0, ρ). Let us suppose that we have found β
which corresponds to K and β0 as above. Then, by the definition of e4 we
have e4 (x − t + h) = e1 (x − t + h) for x − t ∈ K; h ∈ Γ′ , khk≥ β. Hence

T (y + x − t + h)
lim e4 (x − t + h)/(c(h)) = lim , φ0 (y)
h∈Γ′ ,khk→∞ h∈Γ′ ,khk→∞ c(h)hU, φ0 i
= exp(a · (x − t))
and this limit is uniform for x − t ∈ K. Now, for x ∈ K0 the following limit

e(x + h) e4 (t + h)
lim = lim ∗ w(t) (x)
h∈Γ′ ,khk→∞ c(h) h∈Γ′ ,khk→∞ c(h)
= (exp(a · t) ∗ w(t))(x) = exp(a · x)
is uniform, as well.
R
Proposition 1.16. Let Γ be a convex cone with nonempty interior, Γ 6=
∅, and let Γ′ be a closed cone such that Γ′ ∩ ∂ B̄(0, 1) ⊂ int Γ. Let c ∈ Σ(Γ)
and T ∈ F ′ . If for U (x) = C exp(α · x), x ∈ R, C ∈ R, C 6= 0,
s
T (x + h) ∼ c(h) · U (x), h ∈ Γ, U 6= 0,
gs
then there exists e ∈ C ∞ such that e(x) 6= 0, x ∈ Rn , and T (x) ∼
Ce(x), kxk→ ∞, x ∈ Γ′ .
Proof. Let e be as in Lemma 1.4. It is enough to apply Proposition 1.2

and Theorem 1.2,b) to

c(h) T (x + h)
hT (x + h)/e(x + h), ϕ(x)i = , ϕ(x) , ϕ ∈ F .
e(x + h) c(h)
The next proposition also gives a relation between the S-asymptotics
and the generalized S-asymptotics.
Proposition 1.17. Suppose that T ∈ F ′ , Γ = {x ∈ Rn , x = (0, . . . , xk ,

gs s
0, . . . , 0)}, T /c = Dxk S and T (x) ∼ ac(x), kxk→ ∞, x ∈ Γ. Then S(x+h) ∼
hk · 1, h ∈ Γ.
Proof. By L’Hospital’s rule with Stolz’s improvement, we have (with

h = (0, . . . , hk , 0, . . . , 0) and S ∗ φ(h) = hS(x + h), φ(x)i)

(S ∗ φ̆(h)) T
lim hS(x + h)/hk , φ(x)i = lim = lim ∗ φ̆ (h)
hk →∞ hk →∞ hk hk →∞ c

T (x + h)
= lim , φ(x) = ha, φ(x)i.
hk →∞ c(x + h)
Finally, let us point out that if we take the limit
lim T (x + h)/(c(h)e(x + h)),
h∈Γ,khk→∞
in Definition 1.6., then nothing new is obtained (cf. Definition 1.6) for the
following three cases: the one-dimensional case, when Γ ⊂ Rn is a ray, and
when Γ has nonempty interior.
2 Quasi-asymptotics in F ′
2.1 Definition of quasi-asymptotics at infinity over a cone
Quasi-asymptotics has been originally defined and studied for tempered dis-
tributions. The motivation for such a choice can be found in Theorem 2.3
below. The first paper dealing with the analysis of the quasi-asymptotics
was written by Zavyalov [200]. Thereafter, many results concerning the
theory and applications of this notion have appeared. The main features of
the theory (published until the year 1986) have been collected in the mono-
graph [192]. We start with the most general definition of quasi-asymptotics
of tempered distributions ([192]).
Let Γ be a closed convex acute solid cone (cf. 0.2) in Rn and let
{Uk ; k ∈ I ⊂ R} be a family of linear nonsingular transforms of Rn which
leave the cone Γ invariant (automorphisms of Γ). We assume that Jk =
det Uk > 0 and that I has ∞ as a limit point. Furthermore, let ρ(k) be
a positive function defined on I. Denote by S ′ Γ the space of tempered
distributions with supports in the cone Γ. It is important to mention that
S ′ Γ is isomorphic to the space S ′ (Γ) (see 0.5).
Definition 2.1. Let T ∈ S ′ Γ . It is said that T has the quasi-asymptotics

in the cone Γ over the family {Uk ; k ∈ I} related to the positive measurable
function ρ(k) if there exists a tempered distribution g 6= 0 such that
1
lim T (Uk x) = g(x) in S ′ .
k∈I,k→∞ ρ(k)
We will discuss the quasi-asymptotics in a particular case, namely, if

T ∈ F ′ , Uk Γ = kΓ and I = (0, ∞), where Γ is a closed, convex and acute
cone in Rn (see [33]). For results related to Definition 2.1, see [192].
Definition 2.2. Let T ∈ FΓ′ . It is said that T has the quasi-asymptotics in

the cone Γ related to a positive measurable function ρ if there exists g 6= 0
such that
1
w.lim T (kx) = g(x) in F ′ .
k→∞ ρ(k)
q
We write in short T (kx) ∼ ρ(k)g(x), k → ∞ in F ′ ; we will omit the
space from the notation whenever it is clear from the context.
2. Quasi-asymptotics in F ′ 83
The definition itself forces ρ and g to have very specific forms. The
following proposition states precisely to which classes of functions and dis-
tributions ρ and g must belong.
q
Proposition 2.1. Let T ∈ FΓ′ . If T (kx) ∼ ρ(k)g(x), k → ∞, then:
a) ρ is a regularly varying function of the form ρ(k) = k α L(k), k ≥ k0 ,
(see 0.3) where α ∈ R and L is a slowly varying function.
b) g is a homogeneous element of FΓ′ with degree α, that is, g(kx) =
α
k g(x), k > 0.
Proof. We take φ ∈ FΓ such that hT, φi 6= 0. Let K ⊂ (0, ∞) be any

compact set. Then, by the assumption, we have

T (ktx)
lim , φ(x) = hg(x), φ(x)i (2.1)
k→∞ ρ(kt)
uniformly in t, provided that t belongs to the compact set K. On the other
hand, we have uniformly in t ∈ K

T (kx) φ(x/t) φ(x/t)
lim , = g(x), . (2.2)
k→∞ ρ(k) t t
Combining relations (2.1) and (2.2), we obtain for t > 0,

ρ(kt) hT (kx), φ(x/t)
t i hg(x), φ(x/t)
t i
lim = lim = = C(t)
k→∞ ρ(k) k→∞ hT (ktx), φ(x)i hg(x), φ(x)i
uniformly in t ∈ K. This is just the definition of a regularly varying function
(see 0.3). We thus have C(t) = tα , t > 0, for some α ∈ R; therefore, ρ has
the desired form.
In order to prove that g is homogeneous of degree α, we have to use
(2.1):

T (ktx)
hg(tx), ϕ(x)i = lim , ϕ(x)
k→∞ ρ(k)

ρ(kt) T (ktx)
= lim , ϕ(x)
k→∞ ρ(k) ρ(kt)
= htα g(x), ϕ(x)i, ϕ ∈ FΓ , t > 0 .
Remark. It is easy to prove that in case FΓ′ = SR ′

+
≡ S ′ + , then
g = Cfα+1 , where fα is the homogeneous tempered distribution of degree
α − 1 given in 0.4.
We emphasize that that all homogeneous distributions on the real line

are explicitly known. Recall ([56]), they are of the form
g(x) = C− xα α
− + C+ x+ if α 6∈ N ,
(2.3)
g(x) = γδ (k−1) (x) + β x−k if α = −k ∈ N
where C− , C+ γ, β ∈ C and the distribution x−k stands for the standard

regularization of the corresponding function [56, 63], i.e., x−1 = (log |x|)′ ,
−k x−k−1 = (x−k )′ . We refer to ([40, 56, 63]) for the explicit form of
multidimensional homogeneous distributions.
An more general analysis of homogeneous generalized functions can be
found in [85], but only in one dimension. We comment only part of it.
Let Wβ (I) be an abstract locally convex function space whose elements
are defined on I = (−∞, ∞) or on I = (0, ∞) and for which

·
φ(·) 7→ φ , y > 0,
y
is a continuous mapping from Wβ (I) to Wβ (I). If an element g ∈ Wβ′ (I),
where Wβ′ (I) is the dual space of Wβ (I), satisfies the equation
g(y·) = y α g(·), y > 0, α ∈ R ,
in the sense that hg(yx), φ(x)i = y α hg(x), φ(x)i for all φ ∈ Wβ (I), then g
is called a homogeneous generalized function from Wβ′ (I) of degree α.
Let us suppose that Wβ (I) is given by the subspace of C ∞ (I \ {0}) for
which all the seminorms
Z

pn,β (φ) = |x|β xn φ(n) (x)dx, n ∈ N0
I
are finite. In particular p0,β is a norm.

One can easily prove that for a given φ ∈ Wβ (I) the sequence hηn ·φin∈N ,
where ηn is an even positive function in D(I) such that ηn (x) = 0 for
x ∈ [0, 1/2n] ∪ [n + 1, ∞) and ηn (x) = 1 for x ∈ [1/n, n], n ∈ N, converges
to φ in Wβ (I) as n → ∞. This implies that the space D(I \ {0}) is dense
in Wβ (I). Thus, all homogeneous generalized functions of order α ∈ / −N in
Wβ (I) are of the form
A1 xα α α
+ + A2 x− resp. A1 x+ if I = R resp. I = (0, ∞) . (2.4)
Let Uβ (I), β ∈ R, be a subspace of Wβ (I) with a locally convex topol-

ogy, such that the inclusion mapping i : Uβ (I) → Wβ (I) is continuous, as
well as the following one
Uβ (I) → Uβ (I) : φ(·) 7→ φ(·/y), y ∈ (0, ∞) .
If f ∈ Wβ′ (I) and if it is homogeneous of order α, then f |Uβ (I) , the

restriction of f on Uβ (I), is of the form (2.4). A natural question is then
to characterize spaces for which all homogeneous generalized functions are
of the form (2.4). One can show that this is the case for Uβ′ (I). At a first
sight, the introduction of these spaces may seem to be artificial, but it is
naturally connected with the problem we are discussing.
Note that D(I \ {0}) need not be dense in Uβ (I). This is in fact the only
interesting case. Uβ (I) can contain D(I \ {0}) but not as a dense subspace.
Also, Uβ (I) may not contain D(I) at all. If Uβ (I) contains D(I \ {0}) then
clearly Uβ (I) is dense in Wβ (I). Even in that case, we could not use this fact
for the proof that any homogeneous element g on Uβ (I) can be extended
to Wβ (I), i.e., g ∈ Wβ′ (I), and thus it is of the form (2.4).
For Uβ (I), we assume: if φ ∈ Uβ (I) and y0 ∈ (0, ∞), then

φ(·/y) − φ(·/y0 ) d
lim = φ(·/y)
y→y0 y − y0 dy y=y0
in the sense of convergence in Uβ (I). We have the following proposition (for

the proof see [85]).
Proposition 2.2. A generalized function f ∈ Uβ′ (I) is homogeneous of

order α ∈ R if and only if for each φ ∈ Uβ (I)
hf (x), (xφ(x))′ + αφ(x)i = 0 ,
i.e., if and only if f fulfills the equation xf ′ = αf in Uβ′ (I).
A further analysis shows that Uβ′ (I) has the desired properties, for de-
tails see [85].
2.2 Basic properties of quasi-asymptotics over a cone
Recall, Γ∗ = {y; y · x ≥ 0, for every x ∈ Γ} is the conjugate cone to the cone

Γ (cf. 0.2). Denote by C = int Γ∗ . The characteristic function of a closed
convex solid acute cone Γ is denoted by θ1,Γ . The function

Z
KC (z) = eiz·t dt, z ∈ Rn + iC
Γ
is called the Cauchy-Szegö kernel of the tube domain Rn + iC. The closed
convex acute solid cone Γ is regular if KC (z) is a divisor of unity in the
Vladimirov algebra H(C) = L(S ′ (Γ)), where L denotes the Laplace trans-
form (cf. [189], §12]). In case n = 1, 2, 3 it is well-known (see Chapter 1,
§2.7 in [192]) that all (closed convex acute solid) cones are regular. For
regular cones the distribution θα,Γ is given by
Z
eiz·t θα,Γ (t)dt = KC
α
(z), z ∈ Rn + iC, α ∈ R.
Γ
This distribution has many interesting properties (see [33]):

1) supp θα,Γ ⊂ Γ and θα,Γ ∈ S ′ Γ ;
2) θα,Γ (kt) = k n(α−1) θα,Γ (t), t ∈ Γ, k > 0;
3) θα,Γ ∗ θβ,Γ = θα+β,Γ ;
4) For every m ≥ 0 there exists n0 such that θα,Γ ∈ C m (Rn ), α > n0 ;
5) {supp θα,Γ (e − t)} ∩ Γ ⊂ {ktk< R}, where R > 0 do not depend on α
and e ∈ pr Γ = Γ ∩ {kxk= 1}.
6) For any p ∈ N0 , there exists N0 > 0 such that for some A and q > 0
and for every n ≥ N0 , kθn,Γ (x − t)kp,Γ ≤ A(1 + kxk)q , x ∈ Γ (cf. 0.5 for
k kp,Γ ).
Operations of fractional derivatives (for α ≤ −1) and fractional integrals
(for α > −1) can be defined on S ′ Γ via
θα,Γ : T → θα,Γ ∗ T ,
the latter defines a continuous linear operator on SΓ′ . We will use the
notation T (α) for θ−α,Γ ∗ T. By the properties of the convolution, we have
T (−α) (t) = hT (τ ), θα,Γ (t − τ )i ,
if α is sufficiently large.
Definition 2.3. Suppose that f ∈ Lloc (Γ). It is said that f has an asymp-
totic behavior in the cone Γ related to the positive function ρ(k), k ∈ (0, ∞)
if there exists a function g 6= 0 such that
lim f (kx)/ρ(k) = g(x), x ∈ Γ , (2.5)
k→∞
|f (kx)/ρ(k)|≤ h(x), x ∈ Γ, k ≥ k0 ≥ 0 , (2.6)

1
and ϕ(x)h(x) ∈ L (Γ), for every ϕ ∈ F .
Proposition 2.3. If a locally integrable function f has the asymptotic be-

havior in the cone Γ related to ρ (Definition 2.3), and if f defines a regular
q
element of FΓ′ , then f (kx) ∼ ρ(k)g(x), k → ∞.
Proof. We have Z
lim hf (kx)/ρ(k), ϕ(x)i = lim f (kx)ϕ(x)/ρ(k)dx, ϕ ∈ F .
k→∞ k→∞
Γ
The assumption over f allows one to exchange the limit with the integral,
by the Lebesgue theorem.
t
Proposition 2.4. Let F have the property that {ϕ ; k ≥ k0 > 0}
k
is bounded in F for each ϕ ∈ F . Suppose that S, T ∈ F ′ , and ρ(k) =
k α L(k), α > −n, or α = −n and L(k) → ∞ as k → ∞.
a) If S has compact support, then
w.lim S(kx)/ρ(k) = 0, in F ′ .
k→∞
q
b) If T = S on Γ ∩ {kxk> R} for some R > 0, and S(kx) ∼
q
ρ(k)g(x), k → ∞, then T (kx) ∼ ρ(k)g(x), k → ∞, as well.
Proof. a) For every ϕ ∈ F

1 x
hS(kx)/ρ(k), ϕ(x)i = S(x), ϕ .
k n ρ(k) k
The set {ϕ(x/k); k ≥ k0 > 0} is a bounded set in F . Consequently, the set
{hf (x), ϕ(x/k)i; k ≥ k0 > 0} is bounded in R, too. Since k n ρ(k) → ∞, k →
∞, the assertion in a) follows.
b) By assumption, T − S has a compact support. Then, by a), for each
ϕ∈F
lim hT (kx)/ρ(k), ϕ(x)i = lim h(T − S)(kx)/ρ(k), ϕ(x)i
k→∞ k→∞
+ lim hS(kx)/ρ(k), ϕ(x)i

k→∞
= lim hS(kx)/ρ(k), ϕ(x)i

k→∞
which proves b).
Remark. Proposition 2.4 b) asserts that the quasi-asymptotics related

to ρ(k) = k α L(k), α > −n, is a local property. If α < −n, then the quasi-
asymptotics has no longer this property. The next example, in S ′ (R),
illustrates this fact: If f = δ (m) + (xβ+ ), m ∈ N0 , β ∈/ N , then f has
quasi-asymptotics related to ρ(k) = k q , where q = max(−m − 1, −β) (see
Examples 4 and 7 in 2.3). On the other hand, if α = −1, then this
property depends on L. See also Example 5 of 2.3 in relation to Proposition
2.4 b).
Proposition 2.5. Let m = (m1 , . . . , mn ) ∈ (N0 )n and T ∈ F ′ Γ . If

q
T (kx) ∼ ρ(k)g(x), k → ∞ and xm ∈ M(·) then
q
(kx)m T (kx) ∼ k |m| ρ(k)xm g(x), k → ∞ .
Proof. We have

(kx)m T (kx) T (kx) m
lim , ϕ(x) = lim , x ϕ(x)
k→∞ k |m| ρ(k) k→∞ ρ(k)
= hg(x), xm ϕ(x)i = hxm g(x), ϕ(x)i, ϕ ∈ F .
Remark. If g is a distribution with support {0}, then tm g(t) may be

equal zero.
One of the most useful and frequently used theorems that characterize
the quasi-asymptotic behavior is the following one.
Theorem 2.1. Let T ∈ S ′ Γ and Γ be a regular cone. Then, T has quasi-

asymptotics in Γ related to ρ if and only if there exists α ∈ R+ such that
T (−α) has the quasi-asymptotics in Γ related to k nα ρ(k).
Proof. By the mentioned properties of θα,Γ ,
1 k −n(α+1) (−α)
hT (−α) (kx), ϕ(x)i = hT (x), ϕ(x/k)i
k nα ρ(k) ρ(k)
k −n(α+1)
= hT (x), hθα,Γ (τ ), ϕ((x + τ )/k)ii
ρ(k)
1
= hT (kx), hθα,Γ (τ ), ϕ(x + τ )ii,
ρ(k)
where ϕ ∈ S(Γ). Since the cone Γ is regular, the mapping: ϕ(x) → hθα,Γ (t),
ϕ(x + t)i is an automorphism of the space S(Γ). In order to complete the
proof, it is enough to take limit in the last equality.
′
∗
The same theorem can be proved in the same way for S if (M.1),
(M.2)’ and (M.3)’ are satisfied.
We now give a structural theorem for quasi-asymptotics on cones.
Theorem 2.2. Suppose that Γ is a regular cone. Then T ∈ S ′ Γ has quasi-

asymptotics in Γ related to ρ if and only if there exists an integer N such
that T (−N ) is a continuous function and has the asymptotic behavior in Γ
related to k nN ρ(k) (see Definition 2.3).
Proof. The sufficiency follows from Theorem 2.1. Let us prove the
necessity.
Since T (kx)/ρ(k) converges to g in S ′ (in S ′ (Γ)), there exists p ∈ N0
such that it converges in S ′ p (Γ), as well (see 0.5.1). Also, by properties 4)
and 5) of θα,Γ there exists N0 ∈ N such that for every t ∈ Rn and N ≥ N0
the function θN,Γ (t − x) ∈ S ′ p (Γ). Then T (−N ) is continuous and
1 1
limk→∞ T (−N ) (kx) = limk→∞ nN hT (t), θN,Γ (kx − t)i
k nN ρ(k) k ρ(k)
= limk→∞ hT (kt)/ρ(k), θN,Γ(x − t)i
= hg(t), θN,Γ (x − t)i .
By property 6) of θα,Γ , there exist A and q > 0 such that

1
(−N )
(kx) ≤ CkθN,Γ (x − t)kp,Γ ≤ A(1 + kxk)q , x ∈ Γ .
k nN ρ(k) T (2.7)
In this way we proved that T (−N ) has the asymptotic behavior in Γ related
to k nN ρ(k) (cf. Definition 2.3) .
We shall quote a remark given in [192]:
“In many cases it is important to know exactly which primitive of a dis-
tribution f, having quasi-asymptotics, already has an asymptotics. There
is no simple or universal criterion. For instance, it would be natural to
suppose that the condition f ∈ S ′ p (Γ) guarantees the existence of such an
N (depending perhaps on p, on a family {Uk , k ∈ I} and on a function
ρ(k), k ∈ I) for which the primitive f (−N ) has the asymptotics.”
The following example shows that it is not the case, in general:

Let Γ = R+ and let {fn (ξ) = H(ξ) exp(iξ 1/n ); n = 1, 2, . . . } be a family
of functions in S ′ + . Each function fn has the quasi-asymptotics −in n! δ(ξ)
over the family of transforms {Uk ξ = kξ; k > 0} related to the function
(−N )
ρ(k) = 1/k, k > 0 (cf. Definition 2.1). For every n ∈ N the function fn
has the asymptotics over the family of transforms {Uk ξ = kξ, k > 0} with
respect to the function k N −1 , k > 0 if and only if N − 1 > ((n − 1)/n)N ;
that is, when N > n and, hence, N → ∞, as n → ∞. On the other hand,
all the functions fn , n = 1, 2, . . . are infinitely differentiable for ξ > 0 and
uniformly bounded (cf. [192]).
We prove now that the space of tempered distributions is naturally
related to the concept of quasi-asymptotic behavior at infinity.
Theorem 2.3. Let T ∈ D′ , supp T ⊂ Γ, where Γ is regular. If

T (kx)
lim , φ(x) = hg, φi, φ ∈ D (2.8)
k→∞ ρ(k)
for some regularly varying function ρ and g 6= 0, then T ∈ S ′ Γ , and it has
quasi-asymptotic behavior at infinity related to ρ.
Proof. By property 5) of θα,Γ there exists a sufficiently large R > 0

such that θm,Γ (e − x) = 0 for e ∈ pr Γ, x ∈ Γ and kxk> R. Let η ∈ D be
T (kx)
such that η(x) = 1 for kxk≤ R. Then, η(x) , k ∈ N, is convergent
ρ(k)
in S (Γ). Hence there exists a p ∈ N such that it converges in S ′ p (Γ) (see
′
0.5.1). By properties 4) and 5) of θα,Γ , we can find a sufficiently large

m ∈ N such that
1
lim hη(x)T (kx), θm,Γ (e − x)i = gm (e),
k→∞ ρ(k)
where θm,Γ (e−x) is in S ′ p (Γ). Now the left-hand side can be written without
η since θm,Γ (e − x) = 0 for |x|≥ R, and this implies that
1 1 1
lim T (−m) (ke) = lim hT (kx), θm,Γ (e − x)i = gm (e) (2.9)
k→∞ ρ(k) k mn k→∞ ρ(k)
and the last limit exists.

As in Theorem 2.2, it follows that T (−N ) ∈ L1loc and that the limit in
(2.5) for f = T (−N ) and with k N m ρ(k) instead ρ(k) exists. Also (2.6) is
satisfied.
Consequently, T (−N ) is a locally integrable function having the asymp-

totic behavior related to k nN ρ(k) in the cone Γ. Now, we can use Theorem
2.2 to complete the proof.
We now discuss some results in the context of ultradistributions.
Theorem 2.4. Suppose that Mp satisfies (M.1), (M.2) and (M.3). If f ∈

∗
E ′ (R) and supp f ⊂ [0, ∞), then there exists a p ∈ N0 such that
lim hk p+1 f (kx), ϕ(x)i = C, ϕ ∈ S[0,∞)
∗
,
k→∞
where C can be zero, too.
Proof. By Theorem 10.3 in [79], there exist an ultradifferential operator

P
∞
P (D) = an Dn of ∗ class and a compactly supported continuous function
n=0
G, supp G = K ⊃ suppf, such that
∞
X
f= an D n G .
n=0
Suppose that ap 6= 0 and an = 0, n < p. We have to analyze

X
∞
hk p+1 f (kx), ϕ(x)i = k p+1 an Dn G (kx), ϕ(x)
n=0
∞
X Z
x x
= kp (−1)n an k −n G(x), ϕ(n) = (−1)p ap G(x)ϕ(p) dx
n=p
k k
K
∞
X Z
(n) x
+ (−1)n an k −n+p G(x)ϕ ∗
dx, ϕ ∈ S[0,∞) .
n=p+1
k
K
Since
Z Z
x
G(x)ϕ(p) dx → G(x)dxϕ(p) (0), k → ∞ ,
k
K K
it remains to prove that

∞ Z Z
X X∞
Ln
n −n+p (n) x −n+p
(−1) a n k G(x)ϕ dx ≤ C k |G(x)| dx
k Mn
n=p+1 K n=p+1 K
∞
X Z
≤ C1 k −n+p |G(x)|dx → 0, k → ∞ ,
n=p+1 K
in both cases, (Mp ) and {Mp }.
Theorem 2.5. Suppose that (M.1), (M.2) and (M.3) are satisfied. Let
∗
T ∈ D′ (R), supp T ⊂ [0, ∞) and c(k) = k α L(k), k > 0, α > −1. If
s ∗
T (x + h) ∼ c(h)U, then T ∈ S ′ + and T has the quasi-asymptotics related
to c, as well.
Proof. Without loss of generality, we can suppose that c is continuous

in (0, ∞) (cf. Remarks after Proposition 1.2).
Denote by ω a function belonging to D∗ such that ω(x) = 1, x ∈
[0, h0 ], h0 > 0. Then, T = ωT + (1 − ω)T. The support of ωT is compact,
∗
therefore ωT ∈ E ′ . By Theorem 2.4

ω(kt)T (kt) ∗
lim , ϕ(t) = 0, ϕ ∈ S[0,∞) .
k→∞ c(k)
The S-asymptotics is a local property if α > −1 (Theorem 1.2, c)).

Therefore (1 − ω)T has the same S-asymptotics as T and supp(1 − ω)T ⊂
[h0 , ∞).
We shall use Theorem 1.10 with the assumptions that the set A of
that theorem satisfies A + R+ ⊂ [h0 , ∞), h0 > 0, (Γ = R+ ) and that
supp fi ⊂ (h0 − ε, ∞), h0 − ε > 0, i = 1, 2.
Denote by gi (h) = fi (h)/c(h), h ∈ (h0 − ǫ, ∞). Then fi (h) =
c(h)gi (h), h ≥ h0 − ǫ > 0, i = 1, 2, and lim gi (h) = Ci , i = 1, 2. Therefore,
h→∞
((1 − ω)T )(x) = P (D)c(x)g1 (x) + c(x)g2 (x), x > h0 > 0 .
It follows that (1 − ω)T belongs to S ′ ∗ . Consequently, T ∈ S ′ ∗ . Suppose

∗
now that φ ∈ S[0,∞) . Then

(1 − ω(kt))T (kt)
lim , φ(t)
k→∞ c(k)
∞ i
c(kt)g1 (kt) X (−1)i d
= lim , ai φ(t)
k→∞ c(k) i=1
ki dt

c(kt)
+ (a0 g1 (kt) + g2 (kt)), φ(t) . (2.10)
c(k)
By the properties of g1 and g2 , the last summand in (2.10) has a limit for
∗
every φ ∈ S[0,∞) . More precisely,
D E
c(kt)
lim (a0 g1 (kt) + g2 (kt)), φ(t) = (a0 C1 + C2 )xα , φ(x) .
k→∞ c(k)
n c(k·)g (k·) o
1 ∗
Since ; k ≥ 1 is a bounded set in S ′ [0,∞) and since
c(k)
X ∞ i i
−1 d
ai φ(x)
i=1
k dx
∗
tends to zero in S[0,∞) as k → ∞, the first summand in (2.10) tends to zero
as k → ∞.
From the existence of the S-asymptotics of T related to c, it follows that
a0 C1 + C2 6= 0.
∗
Thus, we have proved that T ∈ S ′ + and that T has the quasi-
asymptotics related to c.
∗
Let ϕ ∈ S[0,∞) and let fγ , γ ∈ R, be the function defined in 0.4.
∗
Then (fγ ∗ ϕ)(ξ) = ψ(ξ), ξ ≥ 0 belongs to S[0,∞) . One can prove that the
∗
mapping ϕ 7→ fγ ∗ ϕ is an automorphism of S[0,∞) . The method for proving
this property of fγ is the same as for the space S[0, ∞).
∗
Let f ∈ S ′ [0,∞) . Recall, we denote by f (−m) an element belonging to
′∗
S [0,∞) defined by
hf (−γ) (x), ϕ(x)i = hf (x), (fˇγ ∗ ϕ)(x)i, ϕ ∈ S[0,∞)
∗
.
∗
Proposition 2.6. Suppose that f ∈ S ′ [0,∞) and that γ is a real number.
Then, f has quasi-asymptotics related to ρ if and only if f (−γ) has the
quasi-asymptotics related to k γ ρ(k).
Proof. By the definition of f (−γ) ,

hf (−γ) , ϕi = hf (ξ), (fˇγ ∗ ϕ)(ξ)i
∗ ∗
and this exists for every ϕ ∈ S[0,∞) . Therefore, for every ϕ ∈ S[0,∞) and
k > 0, we have
D E
(−γ) 1 (−γ) ξ
f (kξ), ϕ(ξ) = f (ξ), ϕ
k k
γ
k ξ
= f (ξ), (fˇγ ∗ ϕ)
k k
γ ˇ
= k hf (kξ), (fγ ∗ ϕ)(ξ)i.
Since the mapping: ϕ → fˇγ ∗ ϕ is an automorphisms of S[0,∞)

∗
, this
completes the proof.
Theorem 2.6. Suppose that (M.1), (M.2) and (M.3) are satisfied and that
s
c(h) = hα L(h), h > h0 , α ≤ −1. If T ∈ D′ ∗ (R), T (x + h) ∼ c(h)U, h ∈
∗
R+ and supp T ⊂ [0, ∞), then T ∈ S ′ + .
Rx −1
b
If α = −1 and L(x) = t L(t)dt → ∞, as x → ∞, then T has
h1
b
quasi-asymptotics related to t−1 L(t). In the other cases, T has the quasi-
−p
asymptotics related to t , p ∈ N, but the limit may be zero.
Proof. Let w be the function used in Theorem 2.5. Then T = wT +

(1 − w)T. Since wT has a compact support, by Theorem 2.4, wT has quasi-
asymptotics related to c(k) = k −n for an n ∈ N.
The support of (1 − w)T belongs to [h0 , ∞), h0 > 0. Since the S-
asymptotics is a local property, (1 − w)T has the S-asymptotics related to
c, as well, and
∞
X
((1 − w)T )(t) = ai Di (tα L(t)E1 (t)) + (tα L(t)E2 (t)), t > h0 , (2.11)
i=0
where supp Ei ⊂ (h1 , ∞), 0 < h1 ≤ h0 and lim Ei (t) = Ci , i = 1, 2. We

t→∞
know that a0 C1 + C2 6= 0 because (1 − w)T has the S-asymptotics related
to hα L(h).
We can choose h0 and h1 in such a way that a0 E1 (t) + E2 (t) does not
change the sign when t ∈ (h1 , ∞). Using (2.11) and the convolution with
f1 (cf. 0.4) it follows
((1 − w)T )(−1) = f1 ∗ (tα L(t)(a0 E1 (t) + E2 (t)))
∞
X
+ ai Di (f1 ∗ (tα L(t)E1 (t))).
i=1
We have to analyze the function F = f1 ∗ (tα L(t)(a0 E1 (t) + E2 (t))).

This function is equal to zero in (0, h1 ) and
Zx
F (x) = tα L(t)(a0 E1 (t) + E2 (t))dt, x ≥ h1 .
h1
R∞
Case 1. Let α < −1 or α = −1 and x−1 L(x)dx < ∞. Then,
h1
Z∞
F (x) → tα L(t)(a0 E1 (t) + E2 (t))dt, as x → ∞ .
h1
R∞
Consequently, F has S-asymptotics related to 1. Since t−1 L(t)E1 (t)dt
h1
< ∞, ((1 − w)T )(−1) has S-asymptotics related to 1, too. By Theorem 2.5,
((1 − w)T )(−1) has quasi-asymptotics related to 1, and by Proposition 2.6,
(1 − wT ) has quasi-asymptotics related to h−1 .
R∞
Case 2. Let α = −1 and t−1 L(t) → ∞, as x → ∞. Then, F (x) =
h1
b
L(x), where Lb is also a slowly varying function (see Proposition 1.5.9 in
b
[9]); f has S-asymptotics related to L(x). We have the same situation with
R∞ −1
t L(t)E1 (t)dt = L b 1 (x). By Theorem 2.5, ((1 − w)T )(−1) has quasi-
h1
asymptotics related to L0 = max(L,b Lb 1 ) and by Proposition 2.6, (1 − w)T
−1
has quasi-asymptotics related to k L0 (k).
Taking care of the quasi-asymptotics of wT and (1 − w)T, we have
established the assertion of Theorem 2.6.
The following example illustrates different possibilities in case α < −1.
We use three functions

0, 0≤x≤1
f (x) =
x−1−ε , 1 < x, 1 < ε < 2,
and F (x) = H(x)H(1 − x), x ∈ R; H is the Heaviside function.
The quasi-asymptotics are given by:
Za
−1−ε
hf (kx), ϕ(x)i = k x−1−ε ϕ(x)dx
1/k
Za
1 1 1 1 1 ′ 1 1 1 1 1
= ϕ − ϕ − x1−ε ϕ′′ (x)dx, k > 0;
ε k k ε1−ε k k2 ε 1 − ε k 1+ε
1/k
Z1/k Z1
t
khF (kx), ϕ(x)i = k ϕ(x)dx = ϕ dt → ϕ(0), k → ∞;
k
0 0
∗
k 2 h(DF )(kx), ϕ(x)i = khF (kx), ϕ′ (x)i → ϕ′ (0), k → ∞, ϕ ∈ S ′ + .
Take now the distribution T = f +C1 F +C2 DF. For appropriated constants
C1 and C2 , T can have the quasi-asymptotics related to k −1 , k −2 , k −1−ε .
′
The following problem is discussed in [38]: Let f ∈ S+ and let {ϕk }k∈N
be a sequence in S. Assume that the following limit exists
1
lim hf (kt), ϕk (t)i = c ,
k→∞ ρ(k)
where ρ is a regularly varying function.

The question is to find conditions under which the limit
1
lim hf (kt), ϕ(t)i = cϕ
k→∞ ρ(k)
′
exists for all ϕ ∈ S, i.e., that f ∈ S+ has the quasi-asymptotics related to
ρ.
The authors proved that {ϕk (t)}k∈N cannot belong to S. But they con-
a,M a,M ′
structed a new space Sb,N,δ of test functions and the space (Sb,N,δ ) of dis-
tributions in which the asked question has a positive answer. (cf. Theorem
7 in [38]).
a,M
The relation between the spaces S+ and Sb,N,δ is given by the following
projective limits:
\ a,M a,∞
\
Sb,N,δ = Sb,∞ ≡ Sba , Sba = S+ .
M,M∈Z+ a,b∈R
The results of Theorem 7 in [38] can be applied in the analysis of

quasi-asymptotics of solutions to differential equations, and to Abelian and
Tauberian type theorems for integral transforms of distributions.
2.3 Quasi-asymptotic behavior at infinity of some

generalized functions
We examine first the case when a regular distribution T is defined by a

locally integrable function F on R and has support in an interval [a, ∞), a >
0. We write T = H(x − a)F, where H is the Heaviside function.
1. Let T = H(x − a)F (x), x ∈ R, for a > 0, where F is a locally
R∞
integrable function satisfying |F (x)|dx < ∞. Then T has the quasi-
a
asymptotics related to k −1 .
It follows at once from

Z∞ x
lim hkT (kx), φ(x)i = lim F (x)φ dx = hCδ, φi ,
k→∞ k→∞ k
a
R∞
where C = F (x)dx.
a
We obtain special cases if either F (x) ∼ xα L(x) as x → ∞ and α < −1
R∞
or F (x) ∼ L(x)/x, x → ∞, provided that |L(x)/x|dx < ∞.
a
2. Let a > 0 and T (x) = H(x − a)F (x), x ∈ R, where F is a locally
integrable function such that F (x) ∼ L(x)/x as x → ∞. If
Zx
L(t)
L∗ (x) := dt, x > a
t
a
diverges to infinity as x → ∞, then T has the quasi-asymptotics related to
L∗ (k)/k, k → ∞.
L∗ is also slowly varying at infinity (see 0.3 and [9]). Let
Zx
G(x) := (H ∗ T )(x) = F (t)dt, x ∈ R .
a
Since
G(x) F (x)
lim
∗
= lim =1
x→∞ L (x) x→∞ L(x)/x
d
and G = T, the Structural Theorem 2.2 implies the claim.
dx
3. Let a > 0 and T (x) = H(x − a)F (x), x ∈ R, where F is a locally
integrable function such that F (x) ∼ xα L(x) as x → ∞ for α > −1. Then
T has the quasi-asymptotics related to ρ(k) = k α L(k), k → ∞.
It is obvious, since G = (H ∗ T ) is a continuous function on R such that
G(x) ∼ xα+1 /(α + 1) L(x), as x → ∞.
Now we give the quasi-asymptotics of some distributions.
4. Denote by δ(x − a), a ≥ 0, the delta distribution with support in a.
q
Then δ (m) (kx − a) ∼ k −m−1 δ (m) (x), k → ∞.
Indeed, observe that
1 a φ(0) + O(1/k) φ(0)
hδ(kx − a), φ(x)i = φ = ∼ , k → ∞, φ ∈ S;
k k k k
q
therefore, δ(kx − a) ∼ k −1 δ(x), k → ∞. The result now follows by differ-
entiating m-times the last quasi-asymptotics (cf. Theorem 2.1).
5. For every S ∈ E ′ ∩ S ′ + there exists a natural number p such that S
has the quasi-asymptotics related to k −p , k → ∞.
For a given S ∈ E ′ ∩ S ′ + there exists m ∈ N0 and a continuous function
d
G on R with supp G ⊂ [0, ∞) such that S = Dm G (D = ). If supp S ⊂
dx
[0, a], a ≥ 0, then we have that G is equal to some polynomial of the order
≤ m − 1 on the interval (a, ∞). Thus for some 0 ≤ q ≤ m − 1 and some
C 6= 0
G(x) ∼ Cxq as x → ∞ .
This implies that G has the quasi-asymptotics related to k q , k → ∞. The
Structural Theorem 2.2 implies that S has the quasi-asymptotics related to
q
k q−m , k → ∞; in fact S(kx) ∼ Ck q−m δ (m−q−1) (x).
6. Let F be a locally integrable function on R \ {0} equal to zero outside
of some interval [0, a], a > 0, such that
F (x) ∼ xα L(x) as x → 0+ ,
where α ≤ −1 and L is a slowly varying function at zero.
This function can be identified with a distribution S defined by
Za
xm−1 (m−1)
hS, φi := F (x)(φ(x) − φ(0) − · · · − φ (0))dx (2.12)
(m − 1)!
0
if −(m + 1) < α ≤ −m, m ∈ N and φ ∈ S(R) (see [[138], p. 13]).
The distribution S defined by (2.12) has quasi-asymptotics related to
1
m+1
, k → ∞. Let us prove this. For φ ∈ S(R), we have
k
Za
m+1 m x xm−1
hk S(kx), φ(x)i= k F (x) φ −φ(0)−· · ·− m−1 φ(m−1) (0) dx
k k (m − 1)!
0
Za
1 x m (m) ξx
= km F (x) φ dx, 0 < ξx < x,
m! k k
0
hence
Za
m+1 (−1)m (m)
lim hk S(kx), φ(x)i = hδ , φi xm F (x)dx, as x → ∞ .
k→∞ m!
0
Let us remark that if α > −1, then the distribution

Ra S given by F is
q −1
regular and S(kx) ∼ Ck δ, k → ∞, with C = 0 F (x)dx.
7. Let L be slowly varying at zero and at infinity.
The distribution R(x) = (xα L(x))+ (see [138]) has the quasi-
asymptotics related to ρ(x) = xα L(x), x → ∞, if α 6∈ Z− = {−1, −2, . . . }
Rx
and related to ρ1 (x) = xα L∗ (x) if α ∈ Z− , where L∗ (x) = L(t)t−1 dt, x ≥
a
a.
8. Let f denote a measurable function on R with support in [0, ∞)
satisfying
f (x) ∼ xβ L1 (x) as x → 0+ , (2.13)
where L1 is a slowly varying function at zero and −m − 1 < β ≤ −m. We
suppose that f satisfies the following additional condition
xm f (x) is integrable on (a, ∞), a > 0 . (2.14)
We denote by f˜ the following distribution in S ′ (R) defined by f :

Z∞
xm−1 (m−1)
hf˜, φi = f (x)(φ(x) − φ(0) − · · · − φ (0))dx, φ ∈ S(R) ,
(m − 1)!
0
(2.15)
if −m − 1 < β < −m, and for some a > 0, and if β = −m,
Z∞
xm−2 (m−2)
hf˜, φi = f (x)(φ(x) − φ(0) − · · · − φ (0)
(m − 2)!
0
xm−1 (m−1)
− φ (0)H(a − x))dx, φ ∈ S(R) . (2.16)
(m − 1)!
The distribution f˜ from (2.15), respectively (2.16), has quasi-

asymptotics related to k −m−1 , resp. k −m , provided that both (2.13) and
(2.14) hold.
9. Let f satisfy (2.13) for −m − 1 < β < −m and
f (x) ∼ xγ L2 (x) as x → ∞ (γ < −m),
where L2 is slowly varying at infinity. Then f˜ defined by (2.15) (−m − 1 <
β < −m), respectively by (2.16) (β = −m), has quasi-asymptotics related
to k −m−1 and k −m respectively.
Observe that this quasi-asymptotics does not depend on the functions

L1 and L2 .
10. Let f satisfy (2.13) for −m − 1 < β < −m and
f (x) ∼ xν L2 (x) as x → ∞ (ν > 0) , (2.17)
where L2 is slowly varying at infinity.

Then f˜ defined by (2.15) has the quasi-asymptotics related to xν L2 (x).
We suppose in the next example, as usual, that Mp satisfies (M.1),
(M.2)’ and (M.3)’.
The quasi-asymptotics of ultradistributions is a natural extension of the
same notion for distributions. Suppose that f ∈ S ′ [0,∞) ; it defines an ul-
∗
tradistribution f ∈ S ′ [0,∞) . If it has the quasi-asymptotics as a distribution
related to ρ, then it has also the quasi-asymptotics as an ultradistribution
related to ρ, and with the same limit.
On the other hand, the next examples show that there exist elements
∗
of S ′ [0,∞) which are not in S ′ [0,∞) and have the quasi-asymptotics as ultra-
distributions.
We deal with the Beurling ultradistributions. In the case of Roumieu’s
ultradistributions the treatment is the same.
P∞
11. Let P (D) = an Dn be an ultradifferential operator of class (Mp )
n=0
and let a0 6= 0 and ai 6= 0 for infinitely many i. Then P (D)δ is an element
(Mp ) q
of S ′ [0,∞) which is not a distribution. We shall show that (P (D)δ) (kx) ∼
a0 k −1 δ(x).
By Definition 2.2, we have to consider the following limit:
lim kh(P (D)δ)(kx), ϕ(x)i

k→∞
∞
X
= lim hδ(y), (−1)n an k −n ϕ(n) (y/k)i
k→∞
n=0
∞
X (M )
= a0 ϕ(0) + lim (−1)n an k −n ϕ(n) (0), ϕ ∈ S[0,∞)
p
.
k→∞
n=1
It remains to show that the last limit equals zero. Since for every L > 0
sup Ln /Mn sup |ϕ(n) (x)|< C1 ,

n x∈[0,∞)
it follows
X∞ ∞
X
n −n (n) −n n (n)
(−1) an k ϕ (0)| ≤ C1 k L /Mn |ϕ (0)

n=1 n=1
≤ C2 k −1 → 0, k → ∞.
q (M )
Consequently, (P (D)δ)(kx) ∼ (a0 /k)δ(x) as k → ∞ in S ′ [0,∞)
p
.
12. There are also distributions which have no quasi-asymptotics as
distributions, but they have the quasi-asymptotics as ultradistributions. One
of such distributions is the following one
∞
X
f= δ (n) (x − en )/Mn .
n=0
q
Suppose that there exist ρ and g ∈ S ′ [0,∞) such that f (kx) ∼ ρ(k)g(x)
in S ′ [0,∞) . Then, we would have
X∞
1 (n) n
lim δ (kx − e )/Mn , ϕ(x) = hg, ϕi
k→∞ ρ(k)
n=0
for every ϕ ∈ S[0,∞) .
By Borel’s theorem, there exists ϕ0 ∈ S[0,∞) such that supp ϕ0 ⊂
(p)
[e , e], and the set {exp(−p2 − p)/(ρ(ep )Mp )|ϕ0 (1)|; p ∈ N} is not
−1
bounded in R. For this ϕ0 and for the subset {ep ; p ∈ N}, we have
∞
1 1 X (−1)n (n) n−p
hf (e p
x), ϕ (x)i = ϕ (e )
ρ(ep ) 0 ρ(ep ) Mn exp(np + p) 0
n=0
exp(−p2 − p) (p)
= |ϕ0 (1)|→ ∞, p → ∞.
ρ(ep )Mp
q
This is in contradiction to our assumption f (kx) ∼ ρ(k)g(x) in S ′ [0,∞) .
This distribution has the quasi-asymptotics as ultradistribution; the
proof is the same as for P (D)δ.
2.4 Equivalent definitions of quasi-asymptotics at infinity
Let T ∈ S ′ + . We denote by LT the Laplace transform of T : L(T )(z) =

hT (t), eizt i, z ∈ R + iR+ (cf. [190], Ch. II, Part 9, [189] and [192], Ch.
I, Part 2). We collect in the following theorem some equivalent ways to
define quasi-asymptotics over the cone R+ . The part c) is a prototype of a
Tauberian characterization for quasi-asymptotics; Tauberian theorems for

various integral transforms will be the main subject of 4.2.
Theorem 2.7. Let T ∈ S ′ + and ρ(k) = k α L(k), k ≥ k0 . The following

statements are equivalent:
T (k · )
a) lim = Cfα+1 , in S ′ (R), C =
6 0.
k→∞ ρ(k)
T (k · +b)
b) lim = Cfα+1 , in D′ (R), C 6= 0, b ∈ R.
k→∞ ρ(k)
y
c) A) lim+ LT (iy) = M 6= 0;
y→0 ρ(1/y)
B) there exist D1 > 0, m ∈ N0 , and r0 > 0 such that

r D1
iϕ
ρ(1/r) LT (re ) ≤ sinm ϕ , 0 < r < r0 , 0 < ϕ < π .
d) For every φ ∈ D(R),

(T ∗ φ)
lim (k · ) = Mφ fα+1 , in D′ (R), where Mφ 6= 0.
k→∞ ρ(k)
e) There exists φ0 ∈ D(R) with the property Lφ0 (0) 6= 0 such that
(T ∗ φ0 )(k · )
lim = Mφ0 fα+1 , in D′ (R), Mφ0 6= 0 .
k→∞ ρ(k)
f ) For a δ-sequence hδn in (cf. 0.4) there is a C 6= 0 such that

(T ∗ δn )(k · )
lim = Cfα+1 , in D′ (R), and uniformly for n ∈ N .
k→∞ ρ(k)
Proof. a) ⇒ b). We start with the relation

T (kx + b) T (kx) b
, φ(x) = ,φ x − , φ ∈ D(R) .
ρ(k) ρ(k) k
b
The set {φ( · − ); k ≥ 1} is bounded in D′ (R). By using the equivalence
k
of the weak and strong sequential convergence in D′ (R) and the fact that
φ( · −b/k) → φ, k → ∞, in D(R), we obtain

T (kx + b) T (kx) b
lim , φ(x) = lim ,φ x −
k→∞ ρ(k) k→∞ ρ(k) k
= hCfα+1 , φi, φ ∈ D(R).
By Theorem 2.3 the last limit holds in S ′ (R), as well.

b) ⇒ a) Take b = 0 and apply Theorem 2.3.
a) ⇒ c) and c) ⇒ a) is proved in [192].
c) ⇒ d) For a T ∈ S ′ + , we have T ∗ φ ∈ S ′ [a,∞) , a ∈ R. We shall show
that T ∗ φ satisfies c), it would imply that T ∗ φ satisfies a) and then T
would satisfy d). We have
y y
lim LT (yi)Lφ(iy) = lim LT (yi)Lφ(0) = Mφ0 .
y→0+ ρ(1/y) y→0+ ρ(1/y)
Moreover, there exist D2 > 0, m ∈ N0 and r0 > 0 such that

r
iϕ ϕi
ρ(1/r) LT (re )Lφ(re )

r D2
≤ LT (reiϕ ) max |Lφ(reϕi )|≤ .
ρ(1/r) 0<r≤r0 sinm ϕ
0≤ϕ<π
d) ⇒ e) It is obvious.
e) ⇒ a) The assumptions in e) imply a) and consequently c) for T ∗ φ0 .
Now, we have
y y
lim LT (iy)Lφ0 (iy) = lim+ LT (iy)Lφ0 (i0) = M 6= 0.
y→0+ ρ(1/y) y→0 ρ(1/y)
Taking care of the property that Lφ0 (0) 6= 0, we have

r D′
iϕ
ρ(1/r) LT (re ) ≤ sinm′ ϕ ,
where 0 < r ≤ r0′ , and 0 ≤ ϕ < π for appropriate r0′ , D′ and m′ . This gives
c) for T and consequently a).
a) ⇒ f ) Let φ ∈ D(R) and n ∈ N. Then

(T ∗ δn )(kx) 1 x
, φ(x) = (T ∗ δn )(x), φ
ρ(k) kρ(k) k

1 ·
= T (x), δ̌n ∗ φ (x)
kρ(k) k
Z
1 t
= T (kx), δn (−t)φ x − dt .
ρ(k) k
R
The set
(Z )
t
δn (−t)φ x − dt; n ∈ N, k ∈ (0, ∞)
k
R
is bounded in D(R) because of the properties of hδn in (cf. 0.4). Therefore,
we have
Z Z
T (kx) t
, δn (−t)φ x − dt → hCfα+1 , δn (−y)dy φi, k → ∞ ,
ρ(k) k
R R
uniformly for n ∈ N. Thus, for any ε > 0 there is a k0 (ε) such that

(T ∗ δn )(kx)
, φ(x) − Cf , φ
ρ(k)
α+1
Z
T (kx) t
≤ h , δn (−t)φ x − dt
ρ(k) k
R
Z
t
− hCfα+1 (x), δn (−t)φ x − dti
k
R
Z
t
+ hCfα+1 (x), δn (−t)φ x − dti − hCfα+1 , φi < ε, k > k0 (ε);
k
R
which proves a) ⇒ f ).
R∞
f ) ⇒ e) Since δn (x) = 1, n ∈ N, we have Lδn (0) 6= 0. This implies
−∞
that e) holds.
2.5 Quasi-asymptotics as an extension of the classical

asymptotics
We have seen in Proposition 2.3 that if a locally integrable function f

has asymptotic behavior related to a function ρ, then f˜ has the quasi-
asymptotics related to ρ. The following theorem also goes in this direction.
Theorem 2.8. Let Γ be a closed convex acute solid cone in Rn , f ∈ S ′ (Γ)∩

L1loc (Γ∩{ξ ∈ Rn ; |ξ|> R}) for some R > 0 and let ρ(k) = k α L(k), α > −n.
If, for any e ∈ pr Γ (pr Γ = {x ∈ Rn ; x ∈ Γ, kxk= 1}), the limits
1
lim f (|ξ|e) = g(e) 6= 0
|ξ|→∞ ρ(|ξ|)
exist, and for some g1 ∈ L1 (prΓ) the estimate

|f (ξ)| ξ
≤ g1 , |ξ|> R, ξ ∈ Γ,
ρ(|ξ|) |ξ|
is satisfied, then f has quasi-asymptotics related to ρ.
The proof is similar to that of Proposition 2.3 (see Theorem 2, Chapter

I, §3.3. in [192]).
Remark. The condition α > −n in Theorem 2.8 is essential. This fact
is illustrated by the following two examples.
Let f (x) = H(x − 1)x−1−ε , x ∈ R, ε > 0. Then f (x) ∼ x−1−ε , x →
∞, but by Example 1 in 2.3 the regular distribution f has the quasi-
asymptotics related to ρ(k) = k −1 for any ε > 0.
Let T (x) = H(x− 2)(x log2 x)−1 , x ∈ R. Then T (x) ∼ (x log2 x)−1 , x →
∞. But by the same Example 1 in 2.3, T has the quasi-asymptotics related
to ρ(k) = k −1 .
A more difficult question is to find conditions under which the quasi-
asymptotics implies the ordinary asymptotics of f. A partial answer is given
in the ensuing theorem.
Theorem 2.9. Let T ∈ S ′ + be equal to a locally integrable function f

in some interval [b, +∞), b > 0, with quasi-asymptotic behavior related to
ρ(k) = k α L(k), α > −1. If for some m ∈ N the function xm f (x), x ≥ b, is
monotonous, then f has the asymptotic behavior at infinity related to the
same regularly varying function ρ.
Proof. By Proposition 2.5, the function g(x) = xm f (x), x ≥ b, has

the quasi-asymptotic behavior at infinity related to ρ1 (k) = k α+m L(k).
The monotonicity of g implies that its distributional derivative Dg can be
written as Dg = B + µ, where B ∈ E ′ + and µ is a positive measure with
support in [b, ∞). By Theorem 2.1, Dg has the quasi-asymptotics related
to k α+m−1 L(k). From Proposition 2.4 a), we see that Dg and µ have the
same quasi-asymptotic behavior at infinity, related to k α+m−1 L(k), i.e.,
1
lim hµ(kx), φ(x)i = Chfα+m , φi, φ ∈ S(R) . (2.18)
k→∞ k α+m−1 L(k)
Choosing φε and ψε in D(R) with the properties
ε
φε (x) = 1, for |x|≤ 1 − and φε (x) = 0 for |x|≥ 1,
2
ε
ψε (x) = 1, for |x|≤ 1 and ψε (x) = 0 for |x|≥ 1 +
2
(0 < ε < 1), we obtain
1 1
hµ(kx), φε (x)i ≤ α+m−1 hµ(kx), H(1 − x)i
k α+m−1 L(k) k L(x)
1
≤ hµ(kx)ψε (x)i , (2.19)
k α+m−1 L(k)
since
φε (x) ≤ H(1 − x) ≤ ψǫ (x) for x > 0.
Using (2.18), we see that, both, the left and the right hand side of (2.19)
tend to numbers which do not differ more than ε. Thus the expression in
the middle of (2.18) tends to some limit independent of ε. Since (H ∗ µ) is
equal to g(x) on (b1 , ∞), b1 ≥ b and
1 1
hµ(kx), H(1 − x)i = (H ∗ µ)(k),
k α+m−1 L(k) k α+m L(k)
we conclude that g has ordinary asymptotic behavior at infinity related to
ρ1 , and this implies the statement.
Remark. It is easy to find a function which has quasi-asymptotics but
not an asymptotic behavior. An example is given by f (x) = H(x) sin x, x ∈
q
R; f (kx) ∼ k −1 δ(x), k → ∞.
In such a way, the quasi-asymptotics extends the notion of the asymp-
totic behavior of a locally integrable function.
2.6 Relations between quasi-asymptotics in D ′ (R)

and S ′ (R)
In this section we make a preliminary investigation of some questions raised

by Theorem 2.3:
As in Theorem 2.3, suppose that T ∈ D′ satisfies (2.8), with g ∈ D′ , but we
now remove the assumptions over the support of T (we allow the support to
be any subset of Rn ). It then is natural to ask: Does T ∈ S ′ ? Furthermore,
does the limit exist in S ′ ? Let us observe that Theorem 2.3 does not give
an answer to such a question because the cone Γ is acute and it cannot be
the whole Rn .
The following theorem and its corollary give a partial answer to this
question. We will postpone the complete answer (in one dimension) for
2.10.
Theorem 2.10. Let f ∈ D′ (R) and suppose that there exists, in the sense
of convergence in D′ (R), the limit
lim f (kx)/ρ(k) = g(x) 6= 0 , (2.20)
k→∞
where ρ(k), k ∈ (0, ∞), is a positive continuous function on (0, ∞). Then,
(i) ρ(x) = xv L(x), x ∈ (0, ∞), for some v ∈ R and some slowly varying
function L, and g is a homogeneous distribution with degree of homogeneity
v.
(ii) f ∈ S ′ (R).
(iii) If v > −1, then the limit (2.20) exists in the sense of convergence
in S ′ (R).
(iv) If v = −1 and 1/L(x), x ∈ (a, ∞), is bounded, then the limit (2.20)
exists in the sense of convergence in S ′ (R), as well.
It should be noticed that the proof is quite different from that of Theo-
rem 2.3 where we used the fact that D′ Γ and S ′ Γ are convolution algebras.
The absence of the convolution algebra structure makes the argument more
complex.
Proof. (i) Let ϕ ∈ D(R) such that hg, ϕi 6= 0. We have
lim hf (kmx)/ρ(k), ϕ(x)i = hg(mx), ϕ(x)i;
k→∞
lim (ρ(mk)/ρ(k))hf (kmx)/ρ(km), ϕ(x)i

k→∞
= hg(x), ϕ(x)i lim (ρ(mk)/ρ(k)), m > 0 .
k→∞
This implies that, for every m > 0, we have the existence of the limit
lim (ρ(mk)/ρ(k)) = d(m) .
k→∞
By [[148], p. 17], we obtain that for some v ∈ R and some slowly varying
function L, ρ(k) and d(k) are of the form
ρ(k) = k v L(k), d(k) = k v , k ∈ (a, ∞) .
Since hg(mx), ϕ(x)i = mv hg(x), ϕ(x)i, m > 0, we obtain that g is homoge-

neous of degree v.
(ii) The set {f (kx)/(k v L(k)); k > 0} is a bounded subset of D′ (R).
From Theorem XXII, Chapter VI in [146], it follows that for a given open
bounded neighborhood of zero Ω there exists a compact neighborhood of
m
zero K and a non-negative integer m such that for any ϕ ∈ DK (R)
Ω ∋ x 7→ ((f (kt)/(k v L(k))) ∗ ϕ(t))(x), k ∈ (0, ∞) , (2.21)
is a family of functions which are continuous and uniformly bounded on
Ω. Let Ω = (−2, 2) and K = [−ε, ε]. Since the weakly bounded family
m
(2.21) is strongly bounded in DK (R), we obtain that for every bounded set
m
A ⊂ D[−ε,ε] , the set of functions
{Ω ∋ x 7→ ((f (kt)/(k v L(k))) ∗ ϕ(t))(x); k > 0, ϕ ∈ A}
m
is a bounded family of continuous functions on Ω. Let ψ ∈ D[−ε,ε] and let
ϕk (x) = ψ(kx)/k m , x ∈ R, k ≥ 1 .
Since supp ϕk (x) ⊂ {x; |x|≤ ε/k} ⊂ [−ε, ε], we have that A = {ϕk (x); k ≥
m
1} is a bounded family in D[−ε,ε] and that
{((f (kt)/(k v L(k))) ∗ ϕr (t))(x); k > 0, r ≥ 1}
is a bounded family of continuous functions on Ω. Taking r = k, we obtain
that for some M > 0
| ((f (kt)/(k v L(k))) ∗ (ψ(kt)/k m ))(x) |≤ M, x ∈ (−2, 2), k ≥ 1 .
From
(f (kt) ∗ ψ(kt))(x) = hf (kt), ψ(k(x − t))i = k −1 hf (t), ψ(kx − t)i
= k −1 (f ∗ ψ)(kx),
we obtain that
| (f ∗ ψ)(kx)/(k v+m+1 L(k)) |< M for x ∈ (−2, 2), k ≥ 1 .
m
Taking x = 1 and x = −1 it follows that for any ψ ∈ D[−ε,ε] there exists
Mψ > 0 such that
| (f ∗ ψ)(x) |≤ Mψ (1 + |x|v+m+1 L(|x|)), x ∈ R .
By (VI, 6; 22) in [146], we obtain
d2s
f= (γE ∗ f ) − ψ ∗ f , (2.22)
dx2s
where E is the fundamental solution of d2s E/dx2s = δ, γ ∈ D[−ε,ε] , γ ≡ 1

in a neighborhood of zero and ψ ∈ D[−ε,ε] . If s is large enough, then γE ∈
m
D[−ε,ε] . Thus (2.22) implies that f ∈ S ′ (R).
(iii) We can rewrite (2.22) in the form f = d2s f1 /dx2s + f2 , where f1 is
a continuous and f2 a smooth function such that
sup{|f1 (x)|, |f2 (x)|} ≤ M (1 + |x|v+m+1 L(|x|)), x ∈ R ,
for some M > 0.

Since

v d2s
f (kx)/(k L(k)) = f1 (kx) + f2 (kx) /(k v L(k))
dx2s
d2s
= (f1 (kx))/(k v+2s L(k)) + f2 (kx)/(k v L(k)),
dx2s
we obtain, for s large enough that for any ϕ ∈ D(R) :
lim hf (kx)/(k v L(k)), ϕ(x)i

k→∞
= lim hf2 (kx)/(k v L(k)), ϕ(x)i = hg(x), ϕ(x)i . (2.23)

k→∞
Let us set
( (
f2 (x), x>0 f2 (x), x<0
f2+ (x) = , f2− (x) =
0, x≤0 0, x ≥ 0.
Clearly, for any ϕ ∈ D(0, ∞) (supp ϕ ⊂ (0, ∞)),
lim hf2 (kx)/(k v L(k)), ϕ(x)i = lim hf2+ (kx)/(k v L(k)), ϕ(x)i .
k→∞ k→∞
If ϕ ∈ D(0, ∞), then ψ(t) = ϕ(et )et , t ∈ R, is an element in D(R).

Moreover, the mapping ϕ 7→ ψ defined above is a bijection. Since for
ϕ ∈ D(0, ∞),
Z∞ Z∞
f2+ (kx)ϕ(x)dx = f2+ (ket )ϕ(et )et dt,
0 −∞
r
by putting k = e , r ∈ R, we obtain that for a function F1,2 (t) =
f2+ (et ), t ∈ R, there exists the limit
lim hF2,1 (t + r)/(evr L(er )), ψ(t)i

r→∞
for any ψ ∈ D(R). Using again Theorem XXII, Chapter VI in [146], we

obtain that for any open set Ω ∋ 0 there exists a compact neighborhood of
m
0 and a non-negative integer m such that for any ϕ ∈ DK (R),
vr r
x 7→ (F2,1 (t + r) ∗ ϕ(t))(x)/(e L(e )), r ≥ 0,
is a bounded family of continuous functions on Ω. Since (F2,1 (t + r) ∗
ϕ(t))(x) = (F2,1 (t) ∗ ϕ(t))(x + r), x ∈ R taking x = 0 and using (VI;
6; 22) in [146], we obtain
d2l
F2,1 = 2l H2,1 + G2,1 ,
dx
where H2,1 is a continuous function and G2,1 is a smooth function on R
such that
sup{|H2,1 (t)|, |G2,1 (t)|} < M evt L(et ), t > 0 . (2.24)
If t ∈ (−∞, 0), then et ∈ (0, 1) and since F2,1 (t) is bounded on (−∞, 0),
we obtain that H2,1 and G2,1 are bounded on (−∞, 0). Namely, both func-
tions are equal to the convolution of F2,1 with suitable functions with com-
pact supports.
From
d2l
f2+ (et ) = H2,1 (t) + G2,1 (t), t ∈ (−∞, ∞),
dx2l
we obtain
X dp
f2+ (x) = ap xp (H2,1 (log x)) + G2,1 (log x), x > 0,
p=1
dxp
where ap are suitable constants.
Set now
G2,1 (log x), x>0 H2,1 (log x), x>0
Ḡ2,1 (x) = H̄2,1 (x) =
0, x ≤ 0, 0, x ≤ 0.
P
2l
d p
The distributions f2+ ∈ D′ (−∞, ∞) and ap xp dx p H̄2,1 + Ḡ2,1 are equal
p=1
to each other. For every ϕ ∈ D(R), we have
Z∞
f2+ (kx)ϕ(x)dx = hf2+ (kx), ϕ(x)i
0
2l
X 1
= ap h(kx)p (H̄2,1 (kx))(p) , ϕ(x)i + hḠ2,1 (kx), ϕ(x)i
p=1
kp
2l
X
= ap (−1)p hH̄2,1 (kx), (xp ϕ(x))(p) i + hḠ2,1 (kx), ϕ(x)i .
p=1
The functions H2,1 and G2,1 are bounded on (−∞, 0). Thus, by (2.24),
we have that for some M > 0
1
sup{|Ḡ2,1 (kx)|, |H̄2,1 (kx)|} ≤ M (kx)v L(kx), x > , (2.25)
k
1
sup{|Ḡ2,1 (kx)|, |H̄2,1 (kx)|} ≤ M, 0 < x ≤
, (2.26)
k
Since v > −1, these inequalities imply that for any ϕ ∈ D(R),
Z∞
f2+ (kx)ϕ(x)dx/(k v L(k)) < ∞ . (2.27)
0
In a similar way as in the proof of Theorem 2.3, one can prove that
(2.25) holds for every ϕ ∈ S(R). If we put F2,2 (t) = f2− (−et ), t > 0, then,
by the same arguments as above, one can prove that for every ϕ ∈ S(R),
Z0
f2− (kx)ϕ(x)dx/(k v L(k)) < ∞ . (2.28)
−∞
By the Banach–Steinhaus Theorem, it follows from (2.23), (2.27) and

(2.28) that
lim f (kx)/k v L(k) = g(x)
k→∞
in the sense of convergence in S ′ (R).
(iv) The proof is the same as (iii). Namely, in this case the estimates
(2.25) and (2.26) imply the claim as well.
We can extend (ii) in Theorem 2.10 to the case v ∈
/ −N. First, we shall
recall two assertions (see Lemma 1 and Lemma 2, §7.4 in [192]).
Lemma 2.1. Let f ∈ S ′ (R) and let ρ be a regularly varying function of

the degree β 6= −1, −2, . . .
a) If there exists the limit
f (kt)/ρ(k) → g 6= 0, k → ∞, in S ′ (R) ,
where f = f+ + f− (f+ ∈ S ′ + and f− ∈ S ′ − ), then there exists N ∈ N0
such that for f+,N = tN f+ and f−,N = tN f− ,
lim f±,N (kx)/(k N ρ(k)) = C± fβ+N +1 (±t) ,
k→∞
where (C+ , C− ) 6= (0, 0).

b) If f ∈ S ′ + and if, for N ∈ N0 , fN = tN f has quasi-asymptotics at
∞ related to k N ρ(k), then:
10 If β > −1, then f has quasi-asymptotics at ∞ related to ρ(k), k > 0.
20 If β < −1, then there exist aj ∈ R, j = 0, 1, . . . , p, such that
p
X
g(t) = f (t) + aj δ (j) (t)
j=0
has the quasi-asymptotics related to ρ(k).
Corollary of Theorem 2.10. Let f ∈ D′ (R). Suppose the following

limit exists, in the sense of convergence in D′ (R),
lim f (kx)/(k v L(k)) = g(x) 6= 0,
k→∞
where v ∈ R \ {−1, −2, −3, . . . }. Then this limit exists in the sense of
convergence in S ′ (R) as well.
Proof. Theorem 2.10. implies that f ∈ S ′ (R).
Let n ∈ N be such that v + n > −1. Clearly, for f1 (x) = xn f (x), x ∈ R,
there holds
lim hf1 (kx)/(k v L(k)), ϕ(x)i = hxn g(x), ϕ(x)i, ϕ ∈ D(R) .
k→∞
There exist distributions f+ (x) ∈ S ′ + and f− ∈ S ′ − (supp f− ⊂

(−∞, 0]) such that f = f+ +f− . This decomposition of f implies the decom-
position of f1 : f1 (x) = xn f+ (x) + xn f− (x), x ∈ R, where xn f+ (x) ∈ S ′ +
and xn f− (x) ∈ S ′ − .
From Lemma 2.1 a), it follows that for some m ∈ N,
(tm+n f± (t))(kx)/(k m+n+v L(k)) → C± fv+m+n+1 (x) in S ′ (R) as k → ∞ ,
where (C+ , C− ) 6= (0, 0).
Now, Lemma 2.1 b) implies that for some constants aα , α = 0, 1, . . . , p,
and bβ , β = 0, 1, . . . , r,
p
X
(f+ (kt) + aα δ (α) (kt))/(k v L(k)) → C1 fv+1 (t) in S ′ (R) as k → ∞ ,
α=0
r
X
(f− (kt) + bβ δ (β) (kt))/(k v L(k)) → C2 fv+1 (−t) in S ′ (R) as k → ∞ ,
β=0
where (C1 , C2 ) 6= (0, 0).

Let us notice that in Lemma 2.1 b) it is assumed that the limit distri-
bution g is different from 0. But this assertion also holds for g = 0.
Thus for suitable constants C̃α , α = 0, 1, . . . , s, s = max{p, r}
s
X
(f (kt) + C̃α δ (α) (kt))/(k v L(k)) → C1 fv+1 (t) + C2 fv+1 (−t),
α=0
as k → ∞ in S ′ (R). The fact that v ∈ R \ N implies that

Xs
1
C̃α δ (α) (kt) → 0, k → ∞,
k v L(k) α=0
and this completes the proof.
2.7 Quasi-asymptotics at ±∞
In 2.1–2.5 the essential assumption was that the generalized functions were
supported by an acute cone Γ. In this section, we relax this restriction over
the support in the one-dimensional case. In fact, we already faced this sit-
uation in 2.6 and studied some problems occurring when the distributions
are not supported by an acute cone. We will focus in the case of distri-
bution spaces. Let us observe that some of the results below can be also
generalized to F ′ (Rn ).
Definition 2.4. It is said that f ∈ F ′ (R) has quasi-asymptotics at ±∞

related to some positive measurable function c(k), k ∈ (a, ∞), a > 0, if for
some g ∈ F ′ (R)
lim hf (kx)/c(k), φ(x)i = hg(x), φ(x)i, φ ∈ F (R).
k→∞
q q
We write in short: f (kx) ∼ c(k)g(x), k → ∞ in F ′ (R), or simply f ∼ g at
±∞ related to c(k).
The results of 2.7 give us already some important properties of quasi-

asymptotics at ±∞, they are stated in the following remark.
Remark. If g 6= 0 and F ′ (R) = D′ (R) in Definition 2.4, then, by
Theorem 2.10, f ∈ S ′ (R), c(x) = xv L(x), x ∈ (0, ∞), for some v ∈ R and
some slowly varying function L, and g is a homogeneous distribution with
degree of homogeneity v.
Several properties of the quasi-asymptotics at ±∞ are listed in the

following theorems.
q
Theorem 2.11. Let f ∈ D′ (R) and f ∼ g at ±∞ related to k ν L(k). Then:
q
(i) f (m) ∼ g (m) at ±∞ related to k ν−m L(k), k > a, m ∈ N;
q
(ii) if m ∈ N then xm f ∼ xm g at ±∞ related to k ν+m L(k);
(iii) if φ ∈ E(R) and c1 is a measurable positive function on some
interval (a, ∞), a > 0, such that
φ(kx)/c1 (k) → φ0 (x) in E(R), k → ∞, x ∈ R ,
q
then f φ ∼ gφ0 at ±∞ related to c1 (k)k ν L(k).
Proof. Properties (i) and (ii) follow easily from the definition. For (iii),
observe that if ϕ ∈ D(R), then ϕ(x)φ(kx)/c1 (k) → ϕ(x)φ0 (x) in D(R). By
the equivalence between weak and strong sequential convergence in D′ , we
have

f (kx)φ(kx) f (kx) φ(kx)
lim , ϕ(x) = lim , ϕ(x) = hgφ0 , φi .
k→∞ c1 (k)k ν L(k) k→∞ k ν L(k) c1 (k)
q
Theorem 2.12. Let f ∈ E ′ (R) and f ∼ g at ±∞ related to k ν L(k), g 6= 0.
Then L(k) = 1, k > a , ν ∈ −N, and g(x) = Cδ (−ν−1) (x), for some
constant C. Moreover, the limit in Definition 2.4 can be extended on S(R).
Proof. It is the same as that of Example 5 in 2.3.
Theorem 2.13. Let F be a locally integrable function on R and ν ∈ R, ν >

−1, such that
F (x)
lim = C± ,
x→±∞ |x|ν L(|x|)
q
where L is slowly varying at ∞. Then F ∼ g at ±∞ related to k ν L(k),
where
g(x) = C+ xν+ + C− xν− .
Proof. Let us put F+ (x) = H(x)F (x) and F− (x) = H(−x)F (x), x ∈ R,
we can now apply Example 3 in 2.3 to each F± (see also Theorem 2.8).
We have the following structural theorem, it will be extended in 2.10.
q
Theorem 2.14. Let f ∈ D′ (R) and f ∼ g at ±∞ related to k ν L(k), where
g 6= 0 and ν ∈ R\(−N). There are m ∈ N0 and a locally integrable function
F such that
F (x)
f = F (m) and lim = C± ,
x→±∞ |x|ν+m L(|x|)
where (C+ , C− ) 6= (0, 0).
Proof. Since f ∈ S ′ (R) (Theorem 2.10), let f = f+ +f− , where f+ ∈ S ′ +

and f− ∈ S ′ − (supp f− ⊂ (−∞, 0]). The Corollary of Theorem 2.10 implies
that for every φ ∈ S(R),

f (kx)
, φ(x) → hg(x), φ(x)i as k → ∞ .
k ν L(k)
As in the proof of the Corollary of Theorem 2.10, one may choose f±
satisfying the additional requirement

f± (kx)
, φ(x) → hC̃± fν+1 (±x), φ(x)i as k → ∞ .
k ν L(k)
By Theorem 2.2, there exist locally integrable functions F1 and F2 with

supp F1 ⊂ [0, ∞), supp F2 ⊂ (−∞, 0], and m ∈ N0 such that
(m) (m)
f+ (x) = F1 (x), f− (x) = F2 (x), x ∈ R ,
and
F1 (x) F2 (x)
lim = C+ , lim = C− .
x→∞ xν+m L(x) x→−∞ |x|ν+m L(|x|)
This completes the proof.
The proof of the next result can be found in [111].
R
Theorem 2.15. Let f ∈ S ′ (R) and φ0 ∈ D(R) such that φ0 (t)dt = 1.
q
Let f ′ ∼ g at ±∞ related to k v L(k), v ∈ R, g 6= 0, and

f (kx)
, φ0 (x) → hg0 (x), φ0 (x)i ,
k v+1 L(k)
q
where g0 ∈ S ′ (R) and g0′ = g. Then f ∼ g0 at ±∞ related to k v+1 L(k).
Recall that the Fourier transform of a tempered distribution f is denoted

by F (f ) or fˆ. We shall analyze in 2.8 the quasi-asymptotics of a distribution
at zero (cf. Definition 2.5). In the following theorem, which is very useful,
we already compare these two notions via the Fourier transform.
Theorem 2.16. Let f ∈ D′ (R) and v ∈ R \ (−N). If

q
f ∼ g at ± ∞ related to k v L(k) , (2.29)
with g 6= 0, then
fˆ(x/k)
lim = ĝ(x) in S ′ (R) , (2.30)
k→∞ (1/k)−v−1 L1 (1/k)
where L1 (·) = L(1/·) is slowly varying at the origin.

Conversely, if f ∈ S ′ (R) and (2.30) holds with v ∈ R, then (2.29) holds,
as well.
Proof. Let φ ∈ S(R). By the Corollary of Theorem 2.10, f ∈ S ′ (R) and

f (kx) b fˆ(x/k)
, φ(x) = , φ(x) , k > 0 .
k v L(k) (1/k)−v−1 L1 (1/k)
This implies the assertion.
q
Theorem 2.17. Let T ∈ E ′ (R) and T ∼ g1 at ±∞ related to k v , v ∈ −N,
q
g1 6= 0. Let f ∈ D′ (R) and f ∼ g at ±∞ related to k α L(k), α ∈ R \ (−N),
q
g 6= 0. Then T ∗ f ∼ g1 ∗ g at ±∞ related to k α+v+1 L(k).
Proof. Let φ ∈ S(R). Using the properties of the Fourier transform, we

have

(T ∗ f )(kx) b T̂ (x/k)fˆ(x/k)
, φ(x) = , φ(x)
k α+v+1 L(k) k α+v+2 L(k)

fˆ(x/k) T̂ (x/k)
= , φ(x) . (2.31)
(1/k)−α−1 L(1/k) (1/k)−v−1
Since T̂ is an entire function of polynomial growth when |x|→ ∞, it

must be of the form T̂ (x) = x−v−1 T1 (x), x ∈ R, where T1 is an entire
function of polynomial growth such that T1 (0) = C 6= 0. All the derivatives
of T̂ are of polynomial growth when |x|→ ∞. So, the same holds for T1 .
This implies that for any φ ∈ S(R)
1
v+1
(x/k)−v−1 T1 (x/k)φ(x) = x−v−1 T1 (x/k)φ(x) → x−v−1 T1 (0)φ(x) ,
k
k → ∞, in the sense of convergence in S(R). Let us note that gb1 (x) =

x−v−1 T1 (0), x ∈ R.
In the spaces S(R) and S ′ (R) the strong and weak sequential conver-
gence are equivalent. This implies that

fˆ(x/k) T̂ (x/k)
, φ(x)
(1/k)−α−1 L1 (1/k) (1/k)−v−1
→ hĝ(x), ĝ1 (x)φ(x)i = h(g1 ∗ g)(x), φ̂(x)i k → ∞ .
By (2.31), the claim follows from Theorem 2.16.
Theorem 2.18. Let f ∈ D′ (R) and {f (kx)/k α L(k); k > a}, α ∈ R\

q
(−N), be a bounded subset of D′ (R). Let T ∈ E ′ (R) and T ∼ g1 at ±∞
q
related to k −1 , g1 6= 0. If T ∗ f ∼ g2 at ±∞ related to k α L(k), g2 6= 0, then
q
f ∼ g at ±∞ related to k α L(k) and g1 ∗ g = g2 . (Note, g1 = Cδ).
Proof. The same arguments, as in the proof of Theorem 2.10 yield that
f ∈ S ′ (R) and that
{f (k·)/(k α L(k)); k > a}
is a bounded subset of S(R). With the same arguments as above, we have

(φ ∈ S(R))

fˆ(x/k) T̂ (x/k)
, φ(x) 1 − → 0 as k → ∞ .
(1/k)−α−1 L1 (1/k) T̂ (0)
2.8 Quasi-asymptotics at the origin
Definition 2.5. Let f ∈ F ′ (R) and c(x), x ∈ (0, a), a > 0, be a positive
measurable function. It is said that f has quasi-asymptotics at 0 in F ′ (R)
related to c(1/k) if there is g ∈ F ′ (R) such that

f (x/k)
lim , ϕ(x) = hg(x), ϕ(x)i, ϕ ∈ F (R).
k→∞ c(1/k)
q
We write in short: f ∼ g at 0 related to c(1/k) in F ′ (R).
Remark. The limit in Definition 2.5 may be formulated as

f (εx)
lim , φ(x) = hg(x), φ(x)i, φ ∈ F (R).
ε→+0 c(ε)
Therefore, we often denote also quasi-asymptotics at 0 by

q
f (εx) ∼ c(ε)g(x), ε → 0+ in F ′ (R),
q
or simply by f ∼ g at 0 related to c(ε).
If c = 1 then this Definition 2.5 is a slight generalization of the well-
known Lojasiewicz definition of a “value at 0” of a distribution (see [93]
and [105]), and it leads to a notion of jump behavior for distributions [173],
[176], [178], [181] and [57]. Note that in the Lojasiewicz definition ε → 0
from both sides.
We list some properties of the quasi-asymptotics at 0 for distributions,
we omit the proofs if they are similar to those of the corresponding prop-
erties of the quasi-asymptotics at infinity. For details see [112].
The next proposition shows that we may always assume c is regularly
varying at the origin (cf. 0.3). The proof is essentially the same as that of
(i) in Theorem 2.10.
Proposition 2.7. Let f ∈ D′ (R) and c satisfy the conditions of Definition

2.5. Assume further that g 6= 0. Then, for some real number v and some
slowly varying function L at 0+
c(x) = xv L(x), x ∈ (0, a) .
Moreover, g is homogeneous with the degree of homogeneity v.
Some obvious properties of the quasi-asymptotics at 0 in S ′ (R) are given

in the following proposition.
q
Proposition 2.8. Let f ∈ S ′ (R) and f ∼ g at 0 related to εv L(ε) in S ′ (R).
Then:
q
(i) f (m) ∼ g (m) at 0 related to εv−m L(ε) in S ′ (R), m ∈ N;
q
(ii) xm f (x) ∼ xm g(x) at 0 related to εv+m L(ε) in S ′ (R), m ∈ N.
The same assertions hold for the quasi-asymptotics at 0 in D′ (R).

We have seen (see remark after Proposition 2.4) that the quasi-
asymptotics at infinity is not in general a local property. The next propo-
sition asserts that the quasi-asymptotics at 0 is a local property.
q
Proposition 2.9. Let f ∈ D′ (R) and f ∼ g at 0 related to εv L(ε) in
D′ (R), and let f1 ∈ D′ (R) be such that f = f1 in some neighborhood of
q
zero. Then f1 ∼ g at 0 related to εv L(ε), as well.
Proof. The assertion follows from hf (x), ϕ(x/ε)i = hf1 (x), ϕ(x/ε)i
which holds for any ϕ ∈ D(R) if ε < ε0 (ϕ).
Remark. The same assertion holds for the quasi-asymptotics at 0 in
S ′ (R). This was proved in [34], Lemma 1.6. This claim also follows directly
if we combine Proposition 2.9 with Theorem 2.35 from 2.11.1.
Proposition 2.10. (Theorem 3, Chapter I, 3.3 in [192]) Let f ∈ S ′ (R) be

a locally integrable function in (−a, a), a > 0. Let c(ε) = εα L(ε), α > −1,
where as usual L is slowly varying at 0+ . If
lim f (x)/c(|x|) = C± ,
x→±0
then f has the quasi-asymptotics at zero in S ′ (R) related to c and

lim f (εx)/c(ε) = C+ xα α ′
+ + C− x− in S (R) .
ε→0+
Proof. According to the remark after Proposition 2.9, it is sufficient to

prove that for any ϕ ∈ S(R) the limits of the following expressions exist,
when ε → 0+ ,
Za Z0
f (εx)/c(ε)ϕ(x)dx and f (εx)/c(ε)ϕ(x)dx .
0 −a
Let us consider the first one:

Za Za
c(εx) f (εx)
lim f (εx)/c(ε)ϕ(x)dx = lim ϕ(x)dx .
k→∞ k→∞ c(ε) c(εx)
0 0
But since the term in the second integral is dominated by an L1 function

(c(εx)/c(ε) = O(εα−σ ), where α > σ − 1, [148]), it is possible to pass the
limit under the integral. The second expression can be considered in the
same way.
Proposition 2.10 provides a relation between the asymptotic behavior

at zero of a locally integrable function and the quasi-asymptotics at zero of
the distribution defined by it.
The next Proposition is a direct consequence of Theorem 2.15 and
Theorem 2.16.
q
Proposition 2.11. Let f ∈ S ′ (R) such that xf ∼ g atR 0 related to
εv+1 L(ε), v ∈ R \ (−N) in S ′ (R). Let ϕ0 ∈ D(R) such that ϕ0 (x)dx = 1
R
and

f (εx)
, ϕ0 (x) → hg0 (x), ϕ0 (x)i as ε → 0+
εv L(ε)
q
such that g0 ∈ S ′ (R) and xg0 (x) = g(x), x ∈ R. Then, f ∼ g0 at 0
related to εv L(ε) in S ′ (R) (g and g0 are homogeneous of order v + 1 and
v, respectively).
Proposition 2.10 and (i) of Proposition 2.8 directly yield
Proposition 2.12. Let f ∈ S ′ (R) and f = F (m) in some neighborhood of

0, where m ∈ N0 and F is a locally integrable function such that for some
v > −1 and some slowly varying function L,
F (x)
lim = C± .
x→±0 |x|v L(|x|)
q
Then f ∼ g at 0 related to εv L(ε) in S ′ (R), where g = (C+ xv+ + C− xv− )(m) .
We now discuss an structural theorem. A more complete result will be

the subject of 2.10.3–2.10.5.
Theorem 2.19. Let f ∈ D′ (R) have quasi-asymptotics at 0 in D′ (R) re-

lated to εv L(ε). If v > 0 or if v > −1 but L is bounded on some interval
(0, a), a > 0, then there exist a continuous function F , defined on (−1, 1),
an integer m, and constants C+ , C− , such that
F (x)
f = F (m) and lim = C± . (2.32)
x→±0 |x|v+m L(|x|)
Proof. The proof of this theorem is similar to the proof of the

Lojasiewicz structural theorem for a distribution having a value at 0, given
in [102], pp. 49–52. However, it is necessary to make several non-trivial

refinements in the quoted argument.
Let I = (−2, 2). Since f (εx)/(εv L(ε)) → g(x) in D′ (R), ε → 0+ , there
is a family of continuous functions Fε , ε ∈ (0, ε0 ], defined on I, an m ∈ R0
such that
f (εx)
Fε(m) (x) = , x ∈ I, ε ∈ (0, ε0 ]
εv L(ε)
and Fε (x) → g1 (x) uniformly on I when ε → 0+ , where
g1 (x) = (C+ xv+m
+ + C− xv+m
− )/m! .
With no loss of generality, we assume that ε0 ≥ 1. Let us put

F̃ε (x) = Fε (x)εv L(ε), x ∈ I, 0 < ε ≤ 1 .
(m)
From F̃1 = f it follows (ε−m F1 (εx))(m) = f (εx), x ∈ I, ε ∈ (0, 1].
(m)
So, because F̃ε (x) = f (εx), we obtain
(F̃1 (εx) − εm F̃ε (x))(m) = 0, x ∈ I, ε ∈ (0, 1] .
This implies that there is a polynomial which depends on ε such that
F̃ε (x) = ε−m (F̃1 (εx) + b0 (ε) + b1 (ε)εx + · · · + bm−1 (ε)(εx)m−1 ) ,
x ∈ I, ε ∈ (0, 1]. We have
F̃ε (x) 1
v
− g1 (x) = m+v [F̃1 (εx) − g1 (εx)L(ε|x|)
ε L(ε) ε L(ε)
+ b0 (ε) + · · · + bm−1 (ε)(εx)m−1 ]

L(ε|x|)
+ g1 (x) − 1 , x ∈ I, ε ∈ (0, 1].
L(ε)
We obtain that for any x ∈ I \ {0}
1
(G(εx) + b0 (ε)(εx) + · · · + bm−1 (ε)(εx)m−1 ) → 0
εm+v L(ε)
as ε → 0+ , where we put G(x) = F̃1 (x) − g1 (x)L(|x|).
Note that the last limit is not uniform in general because we have only
the following property of a slowly varying function:
L(ε|x|)
→ 1, ε → 0+ , uniformly for |x|∈ [a, b], b > a > 0 .
L(ε)
Let us fix m points x1 , . . . , xm ∈ I such that xi 6= 0, i = 1, . . . , m, and let

1
d = min{|xi |; i = 1, . . . , m}, J = {x; |x|> d} ∩ I .
2
Because of the quoted property of L, we have
1
(G(εx) + b0 (ε) + b1 (ε)εx + · · · + bm−1 (ε)(εx)m−1 ) → 0, ε → 0+ ,
εm+v L(ε)
uniformly on J. This implies that for some monotonously increasing func-
tion η(ε), ε > 0, η(ε) → 0 as ε → 0+ , there holds
|G(εx) + b0 (ε) + b1 (ε)εx + · · · + bm−1 (ε)εm−1 xm−1 |< εm+v L(ε)η(ε),
x ∈ J, ε ∈ (0, 1] . (2.33)
Let β > 0 and ε < β ≤ 2ε. If we put in (2.33) β instead of ε and xi ε/β
instead of x, i = 1, . . . , m), (note xi ε/β ∈ J, i = 1, . . . , m), we obtain
|G(εxi ) + b0 (β) + εb1 (β)xi + · · · + εm−1 bm−1 (β)xm−1
i |< β m+v L(β)η(β) .
From (2.33) and the last inequality it follows
|b0 (ε) − b0 (β) + ε(b1 (ε) − b1 (β))xi
+ · · · + εm−1 (bm−1 (ε) − bm−1 (β))xm−1
i |< 2β m+v L(β)η(β).
Now, in the same way as in [102], p. 51 one can prove
|bi (ε) − bi (β)|< 2i+1 Kη(β)L(β)β m+v−i ,
ε < β ≤ 2ε, i = 0, . . . , m − 1 , (2.34)
where K is a suitable constant. Let ε < β < 1/2. Take r ∈ N0 such that
β/2r+1 ≤ ε < β/2r . (2.34) implies
|bi (β/2j−1 ) − bi (β/2j )|< 2i+1 Kη(β/2j−1 )L(β/2j−1 )(β/2j−1 )m+v−i ,
(2.35)
i = 0, . . . , m − 1; j = 1, . . . , r, and
|bi (β/2r ) − bi (ε)|< 2i+1 Kη(β/2r )L(β/2r )(β/2r )m+v−i , (2.36)
i = 0, . . . , m − 1.
Let v > 0 and C = sup{tL(t); t ∈ (0, β)}. (2.35) and (2.36) imply
(0 < ε < β < 1/2)
X
r+1 m+v−i−1
|bi (ε) − bi (β)| ≤ 2i+1 Kη(β)C (β/2j−1 )
j=1
i+v+m+2
≤2 Kη(β)Cβ m+v−i−1 , i = 0, 1, . . . , m − 1 . (2.37)
Note that the assumption v > 0 is essential in the above inequality.

From (2.37) it follows bi = lim+ bi (ε) < ∞, i = 0, 1, . . . , m − 1, and
ε→0
|bi − bi (ε)|< Kη(ε)εm+v−i−1 , i = 0, . . . , m − 1,
where K is a suitable constant.

From (2.33) and the last inequality it follows
|G(εx) + b0 + εb1 x + · · · + εm−1 bm−1 xm−1 |< K1 εm+v L(ε)η(ε) , (2.38)
x ∈ J, 0 < ε ≤ 1 (with suitable K1 ).

Let us show that the function
F (x) = G(x) + b0 + b1 x + · · · + bm−1 xm−1 + g1 (x)L(|x|), x ∈ (−2, 2) ,
satisfies the conditions of the Theorem. Clearly, F (m) = f. Put in (2.38)

x = ±1. We obtain
|F (±ε) − g(±ε)L(ε)|< K1 εm+ν L(ε)η(ε), 0 < ε ≤ 1.
Let v > −1 and L(x) < C, x ∈ (0, a). Then from (2.35) and (2.36) it
follows (for 0 < ε < β, β < a, β < 1/2)
r+1
X
|bi (ε) − bi (β)| ≤ 2i+1 KCη(β) (β/2j−1 )m+v−i
j=1
r+1
X
i+1 m+v−i
≤2 KCη(β)β (1/2j−1 )m+v−i , i = 0, . . . , m − 1.
j=1
Now, the proof follows as in the previous case. The proof is complete.
The assertion of Theorem 2.19 also holds for v < 0. This situation is
analyzed in the following theorem. The proof of it has been given in [116].
We omit the proof because more general results will be given in 2.10.3.
Theorem 2.20. Let f ∈ S ′ (R) and let f have the quasi-asymptotic behav-
ior at 0 in S ′ (R) relate to αv L(α), where v ∈ (−∞, 0), v 6= −1, −2, . . . and
L is bounded on some interval (0, a), a > 0. Then there exist a continuous
function F defined on (−1, 1), an integer m and (C+ , C− ) 6= (0, 0) such
that (2.32) holds.
Remarks. 1. If v = 0 and L ≡ 1, then Theorem 2.19 generalizes the

well-known Lojasiewicz structural theorem ([93]) for a distribution which
has a value at 0.
q
2. Let f ∈ S ′ (R) and f ∼ g at 0 related to c(ε) in D′ (R). The question
is: Does the same hold in S ′ (R)? We shall prove in this section that for
c(ε) = εv L(ε), ε < a, the answer to the question is affirmative if v > 0 or
if 0 ≥ v > −1 and L is bounded in some interval (0, µ), µ > 0. Otherwise,
this question remained open for quite long time. A complete affirmative
answer has been recently obtained in [186], it will be presented below in
2.11 (Theorem 2.35).
3. Theorems 2.19 and 2.20 describes the structure of quasi-asymptotics
only under restrictions over ν and L. The complete structural theorems
will be discussed in 2.10.
Proposition 2.12 and Theorem 2.19 directly imply
Theorem 2.21. Let f ∈ S ′ (R) satisfy conditions of Theorem 2.19. Then

q
f ∼ g at 0 related to εv L(ε) in S ′ (R).
Let us denote by Z the space of Fourier transforms of elements from

D(R) supplied by the convergence structure transferred from D(R)(Z =
q
F (D)). Let f ∈ S ′ (R). As usual, we write f ∼ g at 0 related to εv L(ε) in
Z ′ if g ∈ Z ′ , and

f (εx)
lim , ϕ(x) = hg(x), ϕ(x)i, ϕ ∈ Z .
ε→0+ εv L(ε)
Using the Fourier transform and Theorem 2.10 (ii) one can easily obtain
that g ∈ S ′ (R).
For v < 0, we have the ensuing related result. above ([112]).
q
Theorem 2.22. Let f ∈ S ′ (R) and f ∼ g at 0 related to εv L(ε) in Z ′ ,
q
where v < 0, v 6∈ −N, and g 6= 0. Then f ∼ g at 0 related to εv L(ε) in
S ′ (R).
Proof. Apply Fourier transform, (iii) in Theorem 2.10, and then

Theorem 2.16.
2.9 Quasi-asymptotic expansions
Chapter III, §10 in [192] and §4 in [32] were devoted to the quasi-asymptotic
expansion of tempered distributions with support in [0, ∞). Two kind of
expansions were defined therein. We will recall these definitions but in F ′ + .
We also survey other definitions appearing in the literature.
Let {fα ; α ∈ R} be the family of tempered distributions belonging to
S ′ + which is defined in 0.4.
We denote by Σ∞ (by Σ0 ) the set of all positive slowly varying functions
at ∞ (at 0+ ).
Let α ∈ R and L ∈ Σ∞ (L ∈ Σ0 ). We introduce another family of
tempered distributions:
(
H(t)L(t)tα /Γ(α + 1), α > −1 ,
(fL )α+1 (t) = (2.39)
Dn (fL )α+n+1 (t), α ≤ −1, α + n > −1 ,
where n is the smallest natural number such that α + n > −1.
q
Obviously, (fL )α+1 ∼ fα+1 at ∞ (at 0+ ) related to k α L(k) (to
(1/k)α L(1/k)).
Definition 2.6. ([192]) A distribution g ∈ S ′ + is said to have a closed

quasi-asymptotic expansion of order α and of length ℓ, 0 ≤ ℓ ≤ ∞ if there
exist N ∈ N, αj ∈ R and cj ∈ C, j = 1, . . . , N, such that
X N
1
lim α−ℓ g(kt) − cj fαj +1 (kt) = 0, in S+ .
k→∞ k
j=1
Definition 2.7. ([192]) A distribution g ∈ S ′ + is said to have an open

quasi-asymptotic expansion of α and of length ℓ, 0 < ℓ ≤ ∞ if for every
ℓ1 < ℓ, g has closed quasi-asymptotic expansion of order α and of length
ℓ1 .
In [101, 117, 122] these two definitions were slightly altered and ex-
tended by using the family {(fL )α ; α ∈ R} instead of the family {fα ; α ∈
R} (cf. (2.39)). In the next definition we assume that {(fL )α ; α ∈ R} ⊂
′
F+ .
Definition 2.8. We say that an f ∈ F ′ + has a closed quasi-asymptotic

expansion at ∞ (at 0+ ) of order (α, L) ∈ R × Σ∞ (of order (α, L) ∈
R × Σ0 ) and of length ℓ, 0 ≤ ℓ < ∞, related to k α−ℓ L0 (k) (related to

(1/k)α+ℓ L0 (1/k)) if f has quasi-asymptotics at ∞ (at 0+ ) related to
k α L(k) ((1/k)α L(1/k)) and if there exist αi ∈ R, Li ∈ Σ∞ (Li ∈ Σ0 ), and
ci ∈ C, i = 1, . . . , N, N ∈ N, α1 ≥ α2 ≥ · · · ≥ αN (α1 ≤ α2 . . . ≤ αN ),
and such that f is of the form
N
X
f (t) = ci (fLi )αi+1 (t) + h(t) , (2.40)
i=1
where for every φ ∈ F (R)

h(kt) h(t/k)
lim , φ(t) = 0, lim , φ(t) = 0 .
k→∞ k α−ℓ L0 (k) k→∞ (1/k)α+ℓ L0 (1/k)
We write in short:
N
X
q.e.
f ∼ ci (fLi )αi +1 at ∞ (at 0+ ) of order (α, L).
i=1
Obviously, we shall assume that ci 6= 0 and that αN ≥ α−ℓ (αN ≤ α+ℓ).

We shall always assume that in representation (2.40), α1 > α2 > · · · >
αN (α1 < α2 < · · · < αN ), because, (fLj )β+1 + (fLk )β+1 ∼ (fLj +Lk )β+1 .
Observe (fLj )β1 +1 and (fLk )β2 +1 have the same quasi-asymptotics at
∞ (0+ ) if and only if β1 = β2 and Lj ∼ Lk . So, we have:
Proposition 2.13. Let f ∈ F ′ + satisfy the conditions of Definition 2.8

and assume that there are two representations of f, with the same length l,
N
X M
X
f (t) = ci (fLi )αi +1 + h(t), f (t) = c̃i (fLi )α̃i +1 + h̃(t)
i=1 i=1
for which all the assumptions given above hold. Then, M = N, α1 =

α̃1 , . . . , αN = α̃N , L1 ∼ L̃1 , . . . , LN ∼ L̃N .
′
Examples: All examples are in S+ (R).
P∞ H(x − 1)
1. We have that , x ∈ R, converges uniformly to H(x −
r=1 r! xr
1)e1/x but
q.e.
H(x − 1)e1/x ∼ H(x) + ((log x)+ )′ at ∞ of order (0, L ≡ 1)
related to k −1 log k and

X∞
1/x q.e. ′ 1 1
H(x − 1)e ∼ H(x) + ((log x)+ ) + −1 + δ(x) at ∞
r=2
r! r − 1
of order (0, L ≡ 1) related to k −1 .

q.e.
2. H(t − 1)/t3 ∼ (δ − δ ′ )/2 at ∞ of order (−1, L ≡ 1) related to
−3
k log k. Moreover, let n > 2; then for j ≤ n − 2
q.e. 1 1 (−1)j−1 δ (j−1)
H(t − 1)/tn ∼ δ− δ′ + · · · +
(n − 1) (n − 2)1! (n − 1)(j − 1)!
at ∞ of order (−1, L ≡ 1) related to k −j ;
q.e. 1 1 (−1)n−2 (n−2)
H(t − 1)/tn ∼ δ+ δ′ + · · · + δ
(n − 1) (n − 2)1! (n − 2)!
at ∞ of order (−1, L ≡ 1) related to k −n log k.
Following [32], we define the extended open quasi-asymptotic expansion.
Definition 2.9. An f ∈ F ′ + has open quasi-asymptotic expansion at

∞ (at 0+ ) of order (α, L) ∈ R × Σ∞ ((α, L) ∈ R × Σ0 ) and of the
length s, 0 < s ≤ ∞, if and only if for every ℓ, 0 ≤ ℓ < s, f has
closed quasi-asymptotic expansion of order (α, L) and of length ℓ, related
to k α−ℓ Lℓ (k) ((1/k)α+ℓ Lℓ (1/k)).
By the same arguments as in Proposition 2.13 one can prove the follo-
wing proposition:
Proposition 2.14. Let f ∈ F ′ + have open quasi-asymptotic expansion at

∞ (at 0+ ) of order (α, L) and of length s and let 0 ≤ ℓ1 < ℓ2 < s. Suppose
that
N
X
q.e.
f ∼ ai (fLi )αi +1 at ∞ ( at 0+ )
i=1
α−ℓ1
related to k Lℓ1 (k)((1/k)α+ℓ1 Lℓ1 (1/k)), and
M
X
q.e.
f ∼ bi (fLi )βi +1 at ∞ ( at 0+ )
i=1
α−ℓ2
related to k Lℓ2 (k)((1/k)α+ℓ2 Lℓ2 (1/k)).
Then, M ≥ N and ai = bi , αi = βi , Li ∼ L̃i , i = 1, . . . , N.
Let us note that if f has the closed quasi-asymptotic expansion at ∞

of order (α, L) related to k α−ℓ L(k), then for any s ≤ ℓ, f has the open
quasi-asymptotic expansion at ∞ of order (α, L) and of length s. A similar
conclusion holds for the point 0+ as well.
We will actually use the notation
N
X
q.e.
f ∼ ci (fLi )αi +1 |{k αi Li (k)} (εki Li (ε))
i=1
and instead of the “open or closed quasi-asymptotic expansion of order

(α, L) and length s”, we will just say that f has the quasi-asymptotic
expansion at ∞ or with respect to the given scale.
Let us redefine the notion of quasi-asymptotic expansion. We state
our definition at 0+ , but one can do the same for the quasi-asymptotic
expansion at ∞.
We denote by Λ the set N or a finite set of the form {1, 2, . . . , N }, N ∈
N. In the second case, we shall sometimes use the symbol ΛN . In the
following definition, for αi ∈ R and Li ∈ Σ0 , i ∈ Λ, we assume that if i, j ∈
Λ, i < j, then αi ≤ αj and if αi = αj , then Lj (ε)/Li (ε) → 0, ε → 0+ .
Definition 2.10. An f ∈ F ′ (R) has quasi-asymptotic expansion at 0+

related to {εαk Lk (ε); k ∈ Λ} if there are complex numbers Ak 6= 0, k ∈ Λ,
so that for any m ∈ Λ

Pm
f− Ak (fLk )αk +1 (εx)
k=1
w.lim = 0 in F ′ (R) .
ε→0+ εαm Lm (ε)
We write in short:
q.e. X
f ∼ Ak (fLk )αk +1 |{εαk Lk (ε)},
k∈Λ
q.e. P αk
or simply f ∼ k∈Λ Ak (fLk )αk +1 related to {ε Lk (ε); k ∈ Λ}. One can
easily prove that in this case
q.e. X
f′ ∼ Ak (fL′ k )αk +1 |{εαk Lk (ε)}.
k∈Λ
Proposition 2.15. Let f ∈ F ′ (R). If

q.e. X
f ∼ Ak (fLk )αk +1 |{εαk Lk (ε)}
k∈Λ
and
q.e. X
f ∼ Ãk (fL̃k )α̃k +1 |{εα̃k L̃k (ε)} ,
k∈Λ
then αk = α̃k , Ak = Ãk and Lk (ε) ∼ L̃k (ε), ε → 0+ , k ∈ Λ.
Proof. Since 1 ∈ Λ, by the properties of the quasi-asymptotics at

0+ , α1 = α̃1 , A1 = Ã1 and L1 (ε) ∼ L̃1 (ε), ε → 0+ . If
A1 (fL1 )α1 +1 + A2 (fL2 )α2 +1 + R2 = A1 (fL̃1 )α̃1 +1 + Ã2 (fL̃2 )α̃2 +1 + R̃2 ,
where
R2 (εx) R̃2 (εx)
w.lim α
= 0 and w.lim = 0 in F ′ (R) .
ε→0+ ε L2 (ε)
2 ε→0 εα̃2 L̃2 (ε)
+
Since α2 > α1 and α̃2 > α̃1 , the assumption α2 6= α̃2 gives a contradiction.
In the same way L2 (ε) ∼ L̃2 (ε), ε → 0+ and thus A2 = Ã2 . The rest follows
by induction.
′
We give a structural proposition in S+ .
Proposition 2.16. If f ∈ S ′ + ,
q.e. X
f ∼ Ak fαk +1 |{εαk Lk (ε)} ,
k∈Λ
then for each m ∈ Λ there is a pm ∈ N0 and a continuous function Fm with

supp Fm ⊂ [0, ∞) such that
Xm (pm )
Fm (x)
f = Fm + Ak fαk +pm +1 and lim αm +pm = 0.
x→0+ x Lm (x)
k=1
Proof. Observe that a version of Theorem 2.2 holds at 0+ . Now, let

s ≤ m. By Definition 2.10
* f (εx) − P Ak fα +1 (εx)
m
+
k
k=1
, ϕ(x) → 0, as ε → 0+ , ϕ ∈ S(R) ,
εαm Lm (ε)
and the claim follows.
Other definitions and results of the quasi-asymptotic expansions of dis-
tributions are given in [52], [53], [55], [102], [118], [194], and [205].
We end this section with an important example of quasi-asymptotic
expansion at ∞.
Example 3. The quasi-asymptotic expansion at ∞

∞
q X (−1) µj (j)
j
f (kx) ∼ j+1
δ (x) ,
j=0
j! k
is called the Estrada-Kanwal moment asymptotic expansion.
Estrada and Kanwal have extensively studied this expansion in several
distribution spaces as well as its numerous applications. For example, it
holds if f has compact support, and actually not just in the space D′ but
in E ′ . The

constants
µj are the moments of f , namely, they are determined
by µj = f (x), xj . Observe that the moment asymptotic expansion gives
a clear explanation of why the quasi-asymptotics at infinity is not a local
property. For theory and applications of this interesting and important
asymptotic expansion, we refer to [47], [52], [53], and [54].
2.10 The structure of quasi-asymptotics. Up-to-date

results in one dimension
In this section we present results related to quasi-asymptotics in D′ (R) and

S ′ (R). We give complete answers to some questions started in the previous
sections (cf. 2.6, 2.7 and 2.8). We shall follow the exposition from [171],
[172], [173] and [186].
Our first aim is to describe the structure of (one-dimensional) quasi-
asymptotics by means of complete structural theorems; this will be done in
2.10.3 and 2.10.5. The key tool for obtaining such results is the concepts
of asymptotically and associate asymptotically homogeneous functions; we
discuss these classes of functions in detail in 2.10.2 and 2.10.4.
We should employ a new notation for quasi-asymptotics at 0 and ±∞,
which is more convenient for the purposes of this section. In order to
emphasize the role of the slowly varying function, we will use the following
notation for quasi-asymptotics,
f (λx) = L(λ)g(λx) + o(λα L(λ)) in F ′ (R) , (2.41)
where the parameter λ is taken to either zero or infinity. We call α the
degree of the quasi-asymptotic behavior. Observe that g may be identically
zero, and all the results presented in this section are applicable to this
situation as well. Recall that if (2.41) is satisfied and F ′ = D′ , then g is
automatically homogeneous of degree α and it therefore has the form (2.3),
depending on whether α ∈ R \ (−N) or not.
Let gj ∈ F ′ (R) and let ρj be arbitrary measurable functions, j =

1, 2, . . . , n, we write
n
X
f (λx) = ρj (λ)gj (x) + o(ρn (λ)) in F ′ (R) ,
j=1
if
n
X
hf (λx), φ(x)i = ρj (λ) hgj (x), φ(x)i + o(ρn (λ)), φ ∈ F (R).
j=1
2.10.1 Remarks on slowly varying functions
In this section we collect some results about slowly varying functions to be

used in the future.
Let us assume that L is a slowly varying function at the origin (cf. 0.3).
Similar considerations are applicable for slowly varying functions at infinity.
Our first obvious observation is that only the behavior of L near 0 plays
a role in (2.41), and so we may impose to L any behavior we want in
intervals of the form [A, ∞). Moreover, if L̃ is any slowly varying function
which satisfies
L̃(x)
lim+ = 1,
x→0 L(x)
we may replace L by L̃ in any statement about quasi-asymptotics without
loosing generality.
One of the most basic (and most important) results in the theory of
slowly varying functions is the representation formula (see first two pages
of [148]). Furthermore, the representation formula completely characterizes
all the slowly varying functions; L is slowly varying at the origin if and only
if there exist measurable functions u and w defined on some interval (0, B],
u being bounded and having a finite limit at 0 and w being continuous on
[0, B] with w(0) = 0, such that
Z B !
w(t)
L(x) = exp u(x) + dt , x ∈ (0, B].
x t
This formula is important because it enables us to obtain some estimates

on L. Since we are looking for suitable modifications of L, our first remark
is that we can always assume that L is defined in the whole (0, ∞) and L is
everywhere positive. This is shown by extending u and w to (0, ∞) in any

way we want.
Given any fixed σ > 0, then, by modifying u and w, we can assume,
when it is convenient, that B = 1, u is bounded on all over (0, ∞) and
|w(x)| < σ, x ∈ (0, ∞). In particular, we obtain the estimate
L(λx)
M̃ min x−σ , xσ < < M max x−σ , xσ , ∀x, λ ∈ (0, ∞) , (2.42)
L(λ)
for some positive constants M and M̃ . This result is known as Potter’s

estimate [9], p. 25, and will be of vital importance in our investigations of
the structural properties for quasi-asymptotics. Under the assumption of
the last estimate we can use Lebesgue’s dominated convergence theorem in
Z ∞
L (λx)
− 1 φ(x)dx,
0 L(λ)
for φ ∈ S(R), to deduce that
L(λx)H(x) = L(λ)H(x) + o (L(λ)) , in S ′ (R) . (2.43)
The reader should keep in mind (2.42) and (2.43), since from now on they
will be implicitly used without any further reference, especially for differen-
tiating asymptotic expressions in the future sections. We finally comment
a well-known fact [9], [148]: As soon as L(ax) ∼ L(x) holds for each a > 0,
it automatically holds uniformly for a in compact subsets of (0, ∞) .
2.10.2 Asymptotically homogeneous functions
We study some properties of asymptotically homogeneous functions which

will be applied later to the structural study of quasi-asymptotics. Let us
proceed to define this class of functions.
Definition 2.11. A function b is said to be asymptotically homogeneous of

degree α at the origin (resp. at infinity) with respect to the slowly varying
function L, if it is measurable and defined in some interval (0, A) (resp. on
(A, ∞)), A > 0, and for each a > 0,
b(ax) = aα b(x) + o(L(x)), x → 0+ (resp. x → ∞) . (2.44)
Obviously, asymptotically homogeneous functions at the origin and at

infinity are connected by the change of variables x ↔ x−1 ; therefore, most
of the properties of the class of asymptotically homogeneous functions at

infinity can be obtained from those of the corresponding class at the origin.
Observe that no uniformity with respect to a is assumed in Definition
2.11; however, the definition itself forces (2.44) to hold uniformly for a
on compact subsets. Indeed, we will show this fact by using a classical
argument of J. Korevaar, T. van Aardenne Ehrenfest and N. G. de Bruijn
[83], [9], [148] and [186].
Lemma 2.2. Let b be asymptotically homogeneous of degree α with respect

to L. Then, the relation
b(ax) = aα b(x) + o(L(x)),
holds uniformly for a in compact subsets of (0, ∞).
Proof. We show the assertion at the origin, the case at infinity can
be obtained by the change of variables x ↔ x−1 . So assume that b is
asymptotically homogeneous function of degree α at the origin with respect
to L. We may assume that b is defined on (0, 1]. We rather work with the
functions c(x) = eαx b (e−x ) and s(x) = L (e−x ), hence c and s are defined
in [0, ∞). By using a linear transformation between an arbitrary compact
subinterval of [0, ∞) and [0, 1], it is enough to show that
c(h + x) − c(x) = o(eαx s(x)), x → ∞ , (2.45)
uniformly for h ∈ [0, 1]. Suppose that (2.45) is false. Then, there exist
∞
0 < ε < 1, a sequence hhm im=1 ∈ [0, 1]N and an increasing sequence of real
∞
numbers hxm im=1 , xm → ∞, m → ∞, such that
|c(hm + xm ) − c(xm )|≥ εeαxm s(xm ), m ∈ N . (2.46)
Define, for n ∈ N,

ε αxm
An = h ∈ [0, 2]; |c (h + xm ) − c(xm )| < e s (xm ) , m ≥ n ,
3

ε
Bn = h ∈ [0, 2]; |c(h+xm +hm )−c(hm +xm )|< eαxm s(xm +hm ), m ≥ n .
3
Note that
[ [
[0, 2] = An = Bn ,
n∈N n∈N
so we can select N such that µ(An ), µ(Bn ) > 32 (here µ(·) stands for
Lebesgue measure), for all n ≥ N . For each n ∈ N, put Cn = {hn } + Bn .
Then, we have µ(Cn ) > 32 , n ≥ N , and Cn , An ⊆ [0, 3]. It follows that

T T
An Cn 6= ∅, n > N . For each n ≥ N , select un ∈ An Cn . In particular,
we have un − hn ∈ Bn , and hence,
ε αxn
|c (un + xn ) − c (xn )| < e s (xn ) ,
3
ε
|c (un + xn ) − c (xn + hn )| < eαxn s (xn + hn )
3
which implies that for all n ≥ N,
ε αxn
|c (xn + hn ) − c (xn )| < e (s (xn ) + s (xn + hn )) .
3
Using that s(x + h) − s(x) = o(s(x)), x → ∞, uniformly for h on compact
subsets of (0, ∞), we have that for all n sufficiently large, s (xn + hn ) ≤
2s (xn ), which implies that for n big enough
|c (xn + hn ) − c (xn )| < εeαxn s (xn ) ,
in contradiction to (2.46), Therefore, (2.45) must hold uniformly for h ∈

[0, 1].
Corollary 2.1. If b is asymptotically homogeneous at the origin (resp. at

infinity) with respect to a slowly varying function, then b is locally bounded
in some interval of the form (0, B) (resp. (B, ∞)).
Proof. It follows directly from Lemma 2.2.

We now obtain the behavior of asymptotically homogeneous functions
when the degree is not a negative integer.
Theorem 2.23. Suppose that b is asymptotically homogeneous at the origin

(resp. at infinity) with respect to the slowly varying function L. Assume that
its degree is not a negative integer. Then
(i) If α > 0 (resp. α < 0 for the case at infinity), then
b(x) = o(L(x)), x → 0+ (resp. x → ∞) . (2.47)
(ii) If α < 0 (resp. α > 0), then there exists a constant γ such that
b(x) = γxα + o (L(x)) , x → 0+ (resp. x → ∞) . (2.48)

Proof. We show only the assertion at the origin, the case at infinity
follows again from a change of variables.
Let us first show i). Assume that α > 0. Let 0 < η be any arbitrary
number. We keep η < 2α − 1. Let x0 > 0 such that
x

b − 2−α b(x) ≤ ηL(x) and |L(2x) − L(x)| ≤ ηL(x), 0 < x < x0 ,
2
(2.49)

and M = sup |b(x)| /L(x); 12 x0 ≤ x ≤ x0 < ∞. Let x ∈ [x0 /2, x0 ]. We
obtain from (2.49),
n−1
b(x/2n ) X j
≤ 2−αn |b(x)| + η 2 −α(n−1−j) L(x/2 )
.
L(x/2n ) L(x/2n ) L(x/2n )
j=0
Thus, with t = x/2n , and t ∈ [x0 /2n+1 , x0 /2n ],

n−1
b(t) n X j+1
≤ 2−nα M L(2 t) + η 2 −jα L(2 t)
.
L(t) L(t) L(t)
j=0
By this and
L(2j+1 t)/L(2j t) ≤ (1 + η), j = 0, . . . , n − 1 ,

we have that if t ∈ 2−(n+1) x0 , 2−n x0 , then
n ∞ j
b(t) X
≤ M 1+η + η(1 + η)
1+η
L(t) 2α 2α
j=0
n
1+η 2α
=M + η(1 + η) .
2α 2α − 1 − η
Let us prove that for every ε > 0 there exists a positive σ such that
|b(t)/L(t)|< ε, t ∈ (0, σ). First, we have to take η, small enough, such that
2α ε
η(1 + η) α
<
2 −1−η 2
and n0 ∈ N such that
n
1+η ε
M < , n ≥ n0 .
2α 2
Then, it follows that |b(t)/L(t)|< ε, t ∈ (0, σ), if we take σ = x0 /2n0 . This
completes the first part of the proof.
We now show (ii). Assume that α < 0. We rather work with c(x) =
αx
e b(e−x ) and s(x) = L(e−x ). Then c satisfies
c(h + x) − c(x) = o (eαx s(x)) , x → ∞,
uniformly for h ∈ [0, 1]. Given ε > 0, we can find x0 > 0 such that for all
x > x0 and h ∈ [0, 1],
α
|c(x + h) − c(x)| ≤ εeαx s(x) and |s(h + x) − s(x)| ≤ e− 2 − 1 s(x).
So we have that
|c(h + n + x) − c(x)| ≤ |c(h + n + x) − c(n + x)| + |c(n + x) − c(x)|

n−1
X
≤ εeα(n+x) s(n + x) + |c(j + 1 + x) − c(j + x)|
j=0
n
X
≤ εeαx eαj s(j + x)
j=0
1
≤ εeαx s(x) α ,
1−e2
where the last estimate follows from s(x + j) ≤ s(x)e−αj/2 . Since s(x) =
o (e−αx ) as x → ∞, it shows that there exists γ ∈ R such that
lim c(x) = γ .
x→∞
Moreover, the estimate shows that
c(x) = γ + o (eαx s(x)) , x → ∞,
thus, changing the variables back, we have obtained,
b(x) = γxα + o (L(x)) , x → 0+ .
We remark that (2.47) and (2.48) trivially imply that b is asymptotically

homogeneous of degree α with respect to L.
Asymptotically homogeneous functions of degree zero have a more com-
plex asymptotic behavior. For example if L ≡ 1, any asymptotically homo-
geneous function is the logarithm of a slowly varying function. Instead of
attempting to find their behavior in the classical sense, we will study their
distributional behavior.
Lemma 2.3. Let b be asymptotically homogeneous of degree 0 at the origin

(respectively at infinity) with respect to the slowly varying function L. If
σ < 0 (resp. σ > 0) then,
b(x) = o (xσ ) , x → 0+ (resp. x → ∞).
−1
In particular, b(x) (L(x)) is integrable near the origin (resp. locally inte-
grable near ∞).
Proof. We know that L(x) = o (xσ ). Then for each a > 0, b(ax) =
b(x) + o (xσ ) and this implies that x−σ b(x) is asymptotically homogeneous
of degree −σ with respect to the constant function 1. From (i) of Theorem
2.23, it follows that b(x) = o (xσ ).
We now describe the behavior of asymptotically homogeneous functions
of degree zero at the origin. The next theorem will be very important in
the next section.
Theorem 2.24. Let b be asymptotically homogeneous of degree zero at the

origin with respect to the slowly varying function L. Suppose that b is
integrable on (0, A]. Then
b(εx)(H(x) − H(εx − A)) = b(ε)H(x) + o(L(ε)) as ε → 0+ in D′ (R) ,
(2.50)
where H is the Heaviside function.
Proof. Let φ ∈ D(R). Find B such that supp φ ⊆ [−B, B], then there
exists εφ < 1 such that
Z Aε Z B
hb(εx), φ(x)i = b(εx)φ(x)dx = b(εx)φ(x)dx, ε < εφ . (2.51)
0 0
Replacing φ(x) by Bφ(Bx) and εφ by Bεφ , we may assume that B = 1.
Our aim is to show that for some ε0 < 1,
b(εx) − b(ε)
, x ∈ (0, 1], ε < ε0 ,
L(ε)
is dominated by an integrable function in (0, 1] for the use of the Lebesgue
theorem. For this goal, we assume that L satisfies the following estimate,
L(εx) 1
≤ M x− 2 , x ∈ (0, 1], ε ∈ (0, εφ ) . (2.52)
L(ε)
By Lemma 2.2, there exists 0 < ε0 < εφ such that
|b(εx) − b(ε)| < L(ε), x ∈ [1/2, 2], ε < ε0 .

We keep ε < ε0 and x ∈ 2−n−1 , 2−n . Then
|b(εx) − b(ε)| ≤ |b(2εx) − b((2xε)/2)| + |b(2εx) − b(ε)|

Xn

≤ L(2εx) + |b(2εx) − b(ε)| ≤ L 2i εx + L(ε)
i=1
n
X
≤ (2i x)−1/2 L(ε) + L(ε),
i=1
where the last inequality follows from (2.52). Then, if ε < ε0 and x ≤ 1,
then

b(εx) − b(ε)
≤ M1 x− 12 + 1,
L(ε)
√
where M1 = M ( 2 + 1). Therefore we can apply Lebesgue’s dominated
convergence theorem to deduce (2.50).
We also have a similar result at infinity, this fact is stated in the next
theorem. Since its a corollary of Theorem 2.27 , we omit its proof and refer
the reader to 2.10.4.
Theorem 2.25. Let b be asymptotically homogeneous of degree zero at in-

finity with respect to the slowly varying function L. Suppose that b is locally
integrable on [A, ∞). Then
b(λx)H(λx − A) = b(λ)H(x) + o(L(λ)) as λ → ∞ in S ′ (R).
2.10.3 Relation between asymptotically homogeneous functions

and quasi-asymptotics
We introduced asymptotically homogeneous functions in order to study

the structure of the quasi-asymptotics for Schwartz distributions. We now
derive structural theorems for quasi-asymptotics in some cases from the
fundamental properties of this class of functions (Theorems 2.23, 2.24 and
2.25).
The technique to be employed here is based on the analysis of the
parametric coefficients resulting after performing several integrations of the
quasi-asymptotic behavior, these coefficients are naturally connected with
the class of asymptotically homogeneous functions. The technique of in-
tegration of distributional asymptotic relations goes back to the classical
work of Lojasiewicz [93, 116] (cf. Theorem 2.19 in 2.8). Later on, the
properties of the parametric coefficients were singled out and recognized
as asymptotically homogeneous functions in [45], [171], [172], [173], [176],
[186].
The next proposition provides the intrinsic link between quasi-asympto-
tics and asymptotically homogeneous functions.
Proposition 2.17. Let f ∈ D′ (R) have quasi-asymptotic behavior in

D′ (R)
f (λx) = L(λ)g(λx) + o (λα L(λ)) as λ → ∞ (resp. λ → 0+ ) , (2.53)
where L is a slowly varying function and g is a homogeneous distribution
of degree α ∈ R. Let n ∈ N. Suppose that g admits a primitive of order n,
(n)
that is, Gn ∈ D′ (R) and Gn = g, which is homogeneous of degree n + α.
Then, for any given Fn , an n-primitive of f in D′ (R), there exist functions
b0 , . . . , bn−1 , continuous on (0, ∞), such that
n−1
X xn−1−j
Fn (λx) = L(λ)Gn (λx)+ λα+n bj (λ) +o λα+n L(λ) (2.54)
j=0
(n − 1 − j)!
as λ → ∞ (resp. λ → 0+ ) in D′ (R), where each bj is asymptotically

homogeneous of degree −α − j − 1.
Proof. Recall that any φ ∈ D(R) is of the form

Z ∞
φ = Cφ φ0 + θ′ , where Cφ = φ(t)dt, θ ∈ D(R) , (2.55)
−∞
R∞
and φ0 ∈ D(R) is chosen so that −∞ φ0 (t)dt
= 1. The evaluations of
primitives F1 of f and G1 of g on φ are given by
hF1 , φi = Cφ hF1 , φ0 i − hf, θi and hG1 , φi = Cφ hG1 , φ0 i − hg, θi.
This implies

F1 (λx) F1 (λx) f (λx)
, φ(x) = Cφ , φ0 (x) − , θ(x) , (2.56)
λα+1 L(λ) λα+1 L(λ) λα L(λ)
and

G1 (λx) G1 (λx) g(λx)
, φ(x) = Cφ , φ0 (x) − , θ(x) . (2.57)
λα+1 L(λ) λα+1 L(λ) λα L(λ)
With c0 (λ) = h(F1 − G1 )(λx), φ0 (x)i, λ ∈ (0, ∞), from (2.53), it follows

F1 (λx) = L(λ)G1 (λx) + c0 (λ) + o λα+1 L(λ) in D′ (R) . (2.58)
So relation (2.54) follows by induction from (2.58) and (2.53).

We shall now concentrate in showing the property of the bj ’s. We set
(n−m) (n−m)
Fm = Fn and Gm = Gn , m ∈ {1, . . . , n}. By differentiating rela-
tion (2.54) (n − m)-times, it follows that
m−1
X xm−1−j
Fm (λx) = L(λ)Gm (λx) + λα+m bj (λ) + o λα+m L(λ)
j=0
(m − 1 − j)!
R∞ (2.59)
in D′ (R).
R∞ Choose φ ∈ D(R) such that −∞
φ(x)x j
dx = 0 for j = 1, . . . , m−
1, and −∞ φ(x)dx = 1. Then evaluating (2.59) at φ, we have that

(aλ)α+m bm−1 (aλ) + L(aλ) hGm (aλx), φ(x)i + o λα+m L(λ)
1D x E
= hFm (aλx), φ(x)i = Fm (λx), φ
a a
= λα+m bm−1 (λ) + L(λ) hGm (aλx), φ(x)i + o λα+m L(λ) ,
and so, with j = m − 1 ∈ {0, . . . , n − 1}, for each a > 0,
bj (aλ) = a−α−j−1 bj (λ) + o (L(λ)) .
With the aid of asymptotically homogeneous functions, we can now

obtain our first structural theorem. Observe that for the case at ±∞ we
recover Theorem 2.14, while at 0 we actually extend Theorem 2.19 and
Theorem 2.20 (cf. Remark 2 in 2.8).
Theorem 2.26. Let f ∈ D′ (R) have quasi-asymptotic behavior at ±∞

(resp. at the origin) in D′ (R),
(λx)α
− (λx)α
+
f (λx) = C− L(λ) + C+ L(λ) + o (λα L(λ)) . (2.60)
Γ(α + 1) Γ(α + 1)
If α ∈
/ {−1, −2, . . . } , then there exist a non-negative integer m > −α − 1
and an m-primitive F of f such that F is continuous (resp. continuous
near 0) and
Γ(α + m + 1)F (x)

lim = C± (resp. lim± ) . (2.61)
x→±∞ xm |x|α L (|x|) x→0
Conversely, if these conditions hold, then (by differentiation) (2.60) follows.

Proof. On combining Proposition 2.17 and Theorem 2.23, one obtains

that for each n ∈ N and Fn , an n-primitive of f , there exist constants
γ0 , . . . , γn−1 such that in the sense of convergence in D′ (R),
n−1
X (λx)j (−1)n L(λ)(λx)α+n
− L(λ)(λx)α+n
+
Fn (λx) = γj + C− + C+
j=0
j! Γ(α + n + 1) Γ(α + n + 1)

+ o λα+n L(λ) . (2.62)
It follows from the convergence D′ (R) that there is m ∈ N, sufficiently
large, such that any m-primitive of f is continuous and (2.62) holds (with
n = m) uniformly for x ∈ [−1, 1]. Pick a specific m-primitive of f , say Fm ,
then from (2.62) there is a polynomial p of degree at most m − 1 such that
(−1)m (λx)α+m
− (λx)α+m
+
Fm (λx) = p(λx) + C− L(λ) + C+ L(λ)
Γ(α + m + 1) Γ(α + m + 1)

+ o λα+m L(λ) ,
uniformly for x ∈ [−1, 1]. Then setting F = Fm − p, x = 1, −1 and
replacing λ by x, relation (2.61) follows at once. The converse follows by
differentiation and the properties of regularly varying functions.
We now start to analyze quasi-asymptotics of negative integral de-
q
grees. In this section we will only study the quasi-asymptotics f (λx) ∼
γL(λ)δ(λx). We postpone the general case for 2.10.5, after the introduc-
tion of associate asymptotically homogeneous function in 2.10.4.
Proposition 2.18. Let f ∈ D′ (R) have quasi-asymptotic behavior at ±∞

(at the origin) in D′ (R),

f (λx) = γL(λ)δ(λx) + o λ−1 L(λ) as λ → ∞ (resp. λ → 0+ ) . (2.63)
Then, there exist m ∈ N, a function b, being asymptotically homogeneous
of degree 0 with respect to L, and an (m + 1)-primitive F of f such that F
is continuous (resp. continuous near 0) and
xm xm
F (x) = γL(|x|) sgnx + b (|x|) + o (|x|m L(|x|)) . (2.64)
2m! m!
Conversely, if (2.64) holds, then (2.63) follows by differentiation.
Proof. The existence of m, b, and F satisfying (2.64) follows from the

weak convergence of (2.63), Proposition 2.17 and Theorem 2.23, as in the
proof of Theorem 2.26. The converse is shown by applying Theorem 2.25
(resp. Theorem 2.24) and differentiating (m + 1)-times.
2.10.4 Associate asymptotically homogeneous functions
We now introduce the main tool for the study of structural properties of
quasi-asymptotics of negative integral degree. Associate asymptotically
homogeneous functions are a generalization of asymptotically homogeneous
functions. Let us define this class of functions.
Definition 2.12. A function b is said to be associate asymptotically ho-

mogeneous of degree 0 at the origin (resp. at infinity) with respect to the
slowly varying function L, if it is measurable and defined in some interval
(A, ∞), A > 0, and there exists a constant β such that for each a > 0,
b(ax) = b(x) + βL(x) log a + o(L(x)) , x → 0+ (resp. x → ∞) . (2.65)
Remark. Associate asymptotically homogeneous functions are also

known as de Haan functions (cf. [9] and [11]).
We may use the same argument employed in the proof of Lemma 2.2 to
show uniform convergence.
Lemma 2.4. Relation (2.65) holds uniformly in compact subsets of (0, ∞).
We shall study the distributional asymptotic properties of this class of

functions in detail. We first roughly estimate the behavior of associate
asymptotically homogeneous functions of degree 0.
Lemma 2.5. Let b be associate asymptotically homogeneous of degree 0 at

the origin (resp. at infinity) with respect to L, then for each σ < 0 (resp.
σ > 0),
b(x) = o(xσ ) , x → 0+ (resp. x → ∞) . (2.66)
Hence, b is integrable near the origin (resp. locally integrable near infinity).
Proof. We know that L(x) = o(xσ ), for each σ > 0 [148]. Hence b(ax) =
b(x) + o(xσ ) and thus x−σ b(x) is asymptotically homogeneous of degree −σ
with respect to L ≡ 1, so (2.66) follows from Theorem 2.23.
The next two theorems will be crucial in 2.10.5. They generalize
Theorems 2.24 and 2.25. We only give the proof at infinity, the proof
at the origin is similar to that of Theorem 2.24.
Theorem 2.27. Let b be a locally integrable associate asymptotically homo-

geneous function of degree zero at infinity with respect to the slowly varying
function L. Suppose that b is defined on [A, ∞). Then
b(λx)H(λx−A) = b(λ)H(x)+L(λ)βH(x) log x+o(L(λ)) in S ′ (R) , (2.67)
λ → ∞, where H is the Heaviside function.
Proof. Let λ0 be any positive number. The function b can be de-

composed as b = b1 + b2 , where b1 ∈ L1 (R) has compact support and
b2 (x) = b(x)H(x − λ0 ) is associate asymptotically homogeneous function
of degree zero at infinity. Since b1 satisfies the moment asymptotic expan-
sion (cf. Example 3 in 2.9), it follows that b1 (λx) = O(λ−1 ) = o(L(λ)) as
λ → ∞ in S ′ (R). Therefore, we can always assume that A = λ0 , where λ0
is selected at our convenience.
Our aim is to show that there is some λ0 > 1 such that
b(λx) − b(λ) − βL(λ) log x
J(x, λ) := φ(x) H(λx − λ0 )
L(λ)
is dominated by an integrable function, whenever φ ∈ S(R), for the use of
the Lebesgue dominated convergence theorem. For this goal, we can always
assume that L is positive everywhere and satisfies the following estimate
(cf. 2.10.1)
L(λx) n 1 1o
≤ M max x− 4 , x 4 , x, λ ∈ (0, ∞) , (2.68)
L(λ)
for some positive constant M . Because of the uniformity of (2.65) on com-
pact sets, there exists a λ0 > 1 such that
|b(λx) − b(λ) − βL(λ) log x| < L(λ),
x ∈ [1, 2], λ0 < λ .

Let n be a positive integer. We keep λ0 < λ and x ∈ 2n , 2n+1 . Then
|b(λx) − b(λ) − βL(λ) log x| ≤ |b(λx) − b(λ)| + |β| L(λ) log x
≤ |β|L(λ) log x + |b(2(λx/2)) − b(λx/2) − βL(λx/2) log 2|
+ |β|L(λx/2) log 2 + |b(λx/2) − b(λ)|
≤ |β|L(λ) log x + (1 + |β| log 2) L(λx/2) + |b(λx/2) − b(λ)|
n
X
≤ (1 + |β|log 2) L(2−j λx) + |β|L(λ) log 2x + L(λ)
j=1
n
!
1
X j
≤ M x (1 + |β| log 2)
4 (1/2) + |β| log 2x + 1 L(λ) ,
4
j=1
where the last inequality follows from (2.68). So if λ0 < λ and 1 < x, then

≤ M1 x 14 ,
L(λ)
for some M1 > 0. Now if λ0 /λ < x < 1, we have that


L(λ)

L(λx) b(λ) − b(λx) − βL(λx) log x−1
≤ 1+ |β log x| +

L(λ) L(λ)

1 L(λx) b(λx(x−1 )) − b(λx) − βL(λx) log x−1
≤ 1 + M x− 4 |β log x| +
L(λ) L(λx)
1
1
≤ 1 + M x− 4 |β log x| + M M1 x− 2 .
Therefore J(x, λ) is dominated by an integrable function for λ > λ0 , so we

apply Lebesgue dominated convergence theorem to deduce that
Z ∞
lim J(x, λ)dx = 0 .
λ→∞ 0
Finally,
Z ∞ Z ∞
hb(λx)H(λx − λ0 ), φ(x)i − b(λ) φ(x)dx − βL(λ) log x φ(x)dx
0 0
Z ∞ Z ∞ Z ∞
= b(λx)φ(x)dx − b(λ) φ(x)dx − βL(λ) log x φ(x)dx
λ0 /λ 0 0
Z ∞
log λ b(λ)
= L(λ) J(x, λ)dx + L(λ)O +O
0 λ λ

b(λ)
= o(L(λ)) + L(λ)O = o(L(λ)) , λ → ∞ ,
λL(λ)
where in the last equality we have used Lemma 2.5 and the fact that slowly
varying functions are o(λσ ) for any σ > 0. This completes the proof of
(2.67).
Theorem 2.28. Let b be a locally integrable associate asymptotically ho-

mogeneous function of degree zero at the origin with respect to the slowly
varying function L. Suppose that b is defined on (0, A]. Then
b(εx)(H(x) − H(εx − A)) = b(ε)H(x) + L(ε)βH(x) log x + o(L(ε)) , (2.69)
as ε → 0+ in D′ (R).
Corollary 2.2. Let b be an associate asymptotically homogeneous function

of degree 0 with respect to L. Then, there exists an associate asymptotically
homogeneous function c ∈ C ∞ [0, ∞) such that b(x) = c(x) + o(L(x)).
Proof. We may assume that L ∈ C ∞ [0, ∞) [148], Section 1.4. Find

B such that b is locally bounded in [B, ∞), this
R ∞ can be done because
of Lemma
R∞ 2.4. Take φ ∈ D(R)
R∞ such that 0 φ(t)dt = 1 and set
c(x) = B/x b(xt)φ(t)dt − βL(x) 0 φ(t) log tdt, the corollary now follows
from Theorem 2.27 (resp. Theorem 2.28).
We may also use Corollary 2.2 to obtain a representation formula for
associate asymptotically homogeneous functions, this is the analog to [148],
Theorem 1.2 for slowly varying functions.
Theorem 2.29. The function b is associate asymptotically homogeneous

of degree 0 at ∞ satisfying (2.65) if and only if there is a positive number
A such that Z x
τ (t)
b(x) = η(x) + dt , x ≥ A , (2.70)
A t
where η is a locally bounded measurable function on [A, ∞) such that η(x) =
M + o(L(x)) as x → ∞, for some number M , and τ is a C ∞ -function such
that τ (x) ∼ βL(x) as x → ∞.
Proof. Assume first that b1 is C ∞ , defined on [0, ∞) and satisfies that

hypothesis of the theorem. We can find L1 ∼ L which is C ∞ and satisfies
xL′1 (x) = o(L(x)) as x → ∞ [148], p. 7. Let φ and c as in the proof of
Corollary 2.2 corresponding to b1 and L1 , additionally assume that suppφ ⊆
(0, ∞). From Theorem 2.27, we have that

′ b1 (λ) L(λ) H(x) L(λ)
b1 (λx) = δ(x) + β Pf +o as λ → ∞
λ λ x λ
in S ′ (R), where Pf(H(x)/x) = (H(x) log |x|)′ , since distributional asymp-
totics can be differentiated. Then, for x positive
Z ∞ Z ∞
′ ′ ′
xc (x) = x b1 (xt)tφ(t)dt − βxL1 (x) φ(t) log t dt
Z0 ∞ 0
=x b′1 (xt)tφ(t)dt + o(L(x))

0
Z ∞
= b1 (x) · 0 + βL(x) φ(t)dt + o(L(x))
0
= βL(x) + o(L(x)) as x → ∞ .
Set τ (x) = xc′ (x). Find A >R0 such that L is locally integrable on [A, ∞),
x
one has that b1 (x) = c(A) + A (τ (t)/t)dt + o(L(x)).
In the general case, let A be a number such that b and L are locally
bounded on [A, ∞) and let b1 the function from Corollary 2.2 such that
b(x) = b1 (x) + o(L(x)), then we can apply the previous argument
Rx to b1 to
find τ as before, so we obtain (2.70) with η(x) = b(x) − A (τ (t)/t) dt =
c(A) + o(L(x)).
A change of variables x ↔ x−1 in Theorem 2.29 implies the analog
result at 0.
Theorem 2.30. The function b is associate asymptotically homogeneous

of degree 0 at the origin satisfying (2.65) if and only if there is a positive
number A such that
Z A
τ (t)
b(x) = η(x) + dt , x ≥ A , (2.71)
x t
where η is a locally bounded measurable function on [A, ∞) such that η(x) =
M + o(L(x)) as x → 0+ , for some number M , and τ is a C ∞ -function such
that τ (x) ∼ βL(x) as x → 0+ .
A slightly different representation formula is given in [148] , but, except

for the smoothness of τ , both are equivalent.
2.10.5 Structural theorems for negative integral degrees. The

general case
This section is dedicated to the study of structural properties of quasi-

asymptotics with negative integral degree, solving the question posed in
Remark 3 of 2.8.
The next lemma reduces the analysis of negative integral degrees to the
case of degree −1.
Lemma 2.6. Let f ∈ D′ (R) and k be a positive integer. Then f has the
quasi-asymptotic behavior in D′ (R),

f (λx) = γλ−k L(λ) δ (k−1) (x) + (−1)k−1 (k − 1)! βL(λ)(λx)−k + o λ−k L(λ)
if and only if there exists a (k − 1)-primitive g of f satisfying

g(λx) = γλ−1 L(λ) δ(x) + βL(λ)(λx)−1 + o λ−1 L(λ) in D′ (R).
Proof. Apply Proposition 2.17.

We should introduce some notation that will be needed. In the following

for all n ∈ N we denote by ln the primitive of log |x| with the property that
ln (0) = 0 and ln′ = ln−1 . We have an explicit formula for them:
n
xn xn X 1
ln (x) = log |x| − , x ∈ R,
n! n! j=1 j
which can be easily verified by direct differentiation. They satisfy
(ax)n
ln (ax) = an ln (x) + log a, a > 0 . (2.72)
n!
Theorem 2.31. Let f ∈ D′ (R) have quasi-asymptotics at ±∞ (resp. at
the origin) of the form

f (λx) = γλ−1 L(λ)δ(x)+ βλ−1 L(λ)x−1 + o λ−1 L(λ) in D′ (R) . (2.73)
Then, there exist an associate asymptotically homogeneous function b
satisfying
b(ax) = b(x) + β log aL(x) + o(L(x), x → ∞ (resp. x → 0+ ) , (2.74)
an integer m, and a continuous (resp. continuous near 0) (m + 1)-primitive
F of f such that
m
xm xm xm X 1
F (x) = b (|x|) +γ L (|x|) sgnx − βL (|x|) + o (|x|m L (|x|))
m! 2n! m! j=1 j
(2.75)
as x → ±∞ (resp. x → 0), in the ordinary sense. Conversely, relation
(2.75) implies (2.73).
Proof. We will show the assertion only at infinity, the proof at the origin
is exactly the same. We shall study, as we have been doing, the coefficients
of the integration of (2.73). For each n ∈ N, choose an n primitive Fn of
f satisfying Fn′ = Fn−1 . We now proceed to integrate (2.73) once, so we
obtain
γ
F1 (λx) = b(λ) + L(λ)sgnx + βL(λ) log |x| + o(L(λ)) in D′ (R) . (2.76)
2
Now,
R∞ using the standard trick of evaluating at φ ∈ D(R) with the property
−∞
φ(x)dx = 1, one obtains that
Z ∞ Z ∞
γ
b(λa) + L(λa) sgnx φ(x)dx + βL(λa) log|x|φ(x)dx + o(L(λ))
2 −∞ −∞
1D x E
= hF1 (λax), φ(x)i = F1 (λx), φ
Z ∞a a Z ∞
γ
= b(λ) + L(λ) sgnxφ(x)dx + βL(λ) log|ax|φ(x)dx + o(L(λ)) ,
2 −∞ −∞
λ → ∞, for each a > 0. So, we see that b satisfies (2.74) for each a > 0.
Further integration of (2.76) gives,
n n
(λx)n X n xn−j (λx)
Fn+1 (λx) = b(λ) + λ bj (λ) + γL(λ)sgnx
n! j=1
(n − j)! 2n!
+ βL(λ)λn ln (x) + o (λn L(λ)) as λ → ∞ in D′ (R) .
As in the proof of Proposition 2.17, one shows that the bj ’s are asymptot-
ically homogeneous functions of degree −j with respect to L. Hence if we
apply Theorem 2.23 to the bj ’s, we obtain that
(λx)m (λx)m
Fm+1 (λx) = b(λ) + γL(λ) sgnx + βL(λ)λm lm (x) + o (λm L(λ))
m! 2m!
(2.77)
in the sense of convergence in D′ (R). Moreover, it follows from the defini-
tion of convergence in D′ (R) there exists m0 ∈ N such that for all m ≥ m0
the distribution Fm+1 is a continuous function and (2.77) holds uniformly
for x ∈ [−1, 1]. Relation (2.75) is shown by making x = ±1 in (2.77) and
then changing λ ↔ x.
Conversely, since only the behavior of b at infinity plays a roll in (2.75),
we may assume that b is locally integrable, so the converse is obtained after
application of Theorem 2.27 and then (m + 1) differentiations.
Theorem 2.31 is a structural theorem, but we shall give a version free
of b.
Theorem 2.32. Let f ∈ D′ (R). Then f has quasi-asymptotics at ±∞

(resp. at the origin) of the form (2.73) if and only if there exists an (m+1)-
primitive F of f , continuous (resp. continuous near 0), such that for each
a > 0,
m! (a−m F (ax) − (−1)m F (−x))
lim m
= γ + β log a resp. lim+ . (2.78)
x→∞ x L(x) x→0
Proof. The limit (2.78) follows from (2.75), (2.74) and (2.72) by direct
computation. For the converse, rewrite (2.78) as
xm
a−m F (ax) − (−1)m F (−x) = (γ + β log a) L(x) + o (xm L(x)) ,
m!
for each a > 0. Set
 
Xm
γ 1
b(x) = m! x−m F (x) −  − β  L(x) , x > 0.
2 j=1
j
By setting a = 1 in (2.78), one sees that for x < 0,

m
xm xm xm X 1 m
F (x) = b (|x|) + γL (|x|) sgnx− βL (|x|) + o (|x| L (|x|)) .
m! 2m! m! j=1 j
Since
xm
a−m F (ax) − F (x) = β log a L(x) + o (xm L(x)) ,
m!
it is clear that for each a > 0,
b(ax) = b(x) + β log aL(x) + o(L(x)) .
It is remarkable that, initially, no uniform condition on a is assumed

in (2.78). However, the proof of Theorem 2.32 forces this relation to hold
uniformly for a in compact subsets.
We are now ready to state the general structural theorem for negative
integral degrees which now follows directly from Lemma 2.6, Theorem 2.31
and Theorem 2.32.
Theorem 2.33. Let f ∈ D′ (R) and let k be a positive integer. Then f

has the quasi-asymptotic behavior in D′ (R) at ±∞ (resp. at the origin),

f (λx) = γλ−k L(λ) δ (k−1) (x) + (−1)k−1 (k − 1)! βL(λ)(λx)−k + o λ−k L(λ)
if and only if there exist m ∈ N, m ≥ k, an associate asymptotically
homogeneous function b of degree 0 at infinity (resp. at the origin) with
respect to L satisfying
b(ax) = b(x) + β log aL(x) + o(L(x)) , x → ∞ (resp. x → 0+ ),
for each a > 0, and an m-primitive F of f which is continuous (resp.
continuous near 0) and satisfies
xm−k xm−k
F (x) = b (|x|) + γL (|x|) sgnx
(m − k)! 2(m − k)!
m−k
xm−k X 1
− βL (|x|) + o(|x|m−k L(|x|))
(m − k)! j=1 j
as x → ±∞ (resp. x → 0), in the ordinary sense. The last property is

equivalent to

(m − k)! ak−m F (ax) − (−1)m−k F (−x)
lim m−k
= γ+β log a resp. lim+ ,
x→∞ x L(x) x→0
(2.79)
for each a > 0.
It should be noticed that in (2.79) is not absolutely necessary to assume

that the limit is of the form γ + β log a. Indeed, we have the following
corollary.
Theorem 2.34. Let f ∈ D′ (R). Then f has quasi-asymptotics at infinity

(resp. at the origin) of degree −k, k ∈ N, if and only if there exists a
continuous (resp. continuous near 0) m-primitive F of f , m ≥ k, such that
the following limit exists for each a > 0,

ak−m F (ax) − (−1)m−k F (−x)
lim = I(a) resp. lim .
x→∞ xm−k L(x) x→0+
Proof. We show that I(a) must be of the form I(1) + β log a, for some
constant β. We easily see that I is measurable and satisfies
I(ab) = I(a) + I(b) − I(1) ,
setting h(x) = eI(x)−I(1) , one has that h is positive, measurable and satisfies
h(ab) = h(a)h(b), from where it follows [148] that h(x) = xβ , for some β,
and so I has the desired form.
2.11 Quasi-asymptotic extension
We analyze some problems about the extensions of distributions to other

spaces together with their quasi-asymptotic properties, we name this prob-
lem quasi-asymptotic extension problem.
Let f ∈ F ′ have quasi-asymptotic behavior in F ′ , that is,
hf (λx), φ(x)i ∼ λα L(λ) hg(x), φ(x)i , ∀φ ∈ F . (2.80)
Suppose that f belongs also to another space U ′ such that F ⊂ U (not
necessarily densely contained). For various spaces, we investigate in this
section the possibility of extending (2.80) to U ′ , in the sense of obtaining
the asymptotic behavior of hf (λx), ϕ(x)i , for each ϕ ∈ U. Sometimes the
statement “f ∈ U ′ ” is also part of the problem. The results of the present
section were obtained in [171], [172], [173], [186].
2.11.1 Quasi-asymptotics at the origin in D′ (R) and S ′ (R)
In this subsection we conclude the discussion initiated in Remark 2 of 2.8

(cf. Theorem 2.21).
Theorem 2.35. Let f ∈ S ′ (R). If f has quasi-asymptotic behavior at 0 in

D′ (R), then f has the same quasi-asymptotic behavior at 0 in in the space
S ′ (R).
Proof. Let α be the degree of the quasi-asymptotic behavior. We shall

divide the proof into three cases:
α∈
/ {−1, −2, −3, . . . },
α = −1,
α = −2, −3, . . .
Suppose its degree is α ∈
/ {−1, −2, −3, . . . } and
(εx)α
− (εx)α
+
f (εx) = C− L(ε) + C+ L(ε) + o (εα L(ε))
Γ(α + 1) Γ(α + 1)
as ε → 0+ in D′ (R). Then, by using Theorem 2.26 and the fact f ∈ S ′ (R),
we conclude the existence of an integer m, a real number β such that
m > −α, β > m + α, and a continuous m-primitive F of f such that
|x|m+α
F (x) = L (|x|) ((−1)m C− H(−x) + C+ H(x))
Γ(m + α + 1)
m+α
+ o( |x| L (|x|) ), x → 0+ ,
and
F (x) = O( |x|β ), |x| → ∞ . (2.81)
We make the usual assumptions over L. Assume (cf. 2.10.1) that L is

positive, defined in (0, ∞) and there exists M1 > 0 such that
L(εx) n 1 1o
≤ M1 max x− 2 , x 2 , ε, x ∈ (0, ∞) . (2.82)
L(ε)
Let φ ∈ S(R), then we can decompose φ = φ1 + φ2 + φ3 , where supp φ1 ⊆
(−∞, 1], supp φ2 is compact and supp φ3 ⊆ [1, ∞). Observe that since
φ2 ∈ D(R) we have that

m α xα−
hf (εx), φ2 (x)i = (−1) C− ε L(ε) , φ2 (x)
Γ(α + 1)

α xα
+
+ C+ ε L(ε) , φ2 (x) + o (εα L(ε)) , ε → 0+ .
Γ(α + 1)
(2.83)
So, if we want to show (2.83) for φ, it is enough to show it for φ3 placed

instead of φ2 in the relation because by symmetry it would follow for φ1
and hence for φ. Set
F (x)
G(x) = α+m , x > 0.
x L(x)
Then
C+
lim G(x) = . (2.84)
x→0+ Γ(α + m + 1)
On combining (2.81), (2.82) and (2.84), we find a constant M2 > 0 such
that
1
|G(x)| < M2 (1 + xβ+ 2 −m−α ), x > 0 . (2.85)
Relation (2.85) together with (2.82) show that for ε ≤ 1,

G(εx) L(εx) xα+m φ(m) (x) ≤ 2M1 M2 xβ+1 φ(m) (x) H(x − 1) .
L(ε) 3 3
The right hand side of the last estimate belongs to L1 (R) and thus we can
use the Lebesgue dominated convergence theorem to obtain,
1
lim α hf (εx), φ3 (x)i
ε→0 ε L(ε)
+
Z ∞
L(εx) α+m (m)
= lim (−1)m G(εx) x φ3 (x)dx
ε→0+ 0 L(ε)
Z ∞
m C+ (m)
= (−1) xα+m φ3 (x)dx
Γ(α + m + 1) 0

xα+
= C+ , φ3 (x) .
Γ(α + 1)
This shows the result in the case α ∈ / {−1, −2, −3, . . . } .
We now consider the case α = −1. Assume that

f (εx) = γε−1 L(ε)δ(x) + βε−1 L(ε)x−1 + o ε−1 L(ε) as ε → 0+ in D′ (R).
As in the last case, it suffices to assume that φ ∈ S(R), supp φ ⊆ [1, ∞)
and show that Z ∞
ε φ(x)
lim hf (εx), φ(x)i = β dx.
ε→0+ L(ε) 1 x
We may proceed as in the previous case to apply the structural theorem,
but we rather reduce it to the previous situation. So, set g(x) = xf (x),
then
g(εx) = βL(ε) + o(L(ε)) as ε → 0+ in D′ (R) . (2.86)
But g ∈ S ′ (R), then since the degree of the quasi-asymptotics is 0, the first
case implies that (2.86) is valid in S ′ (R). Therefore
Z ∞
ε 1 φ(x) φ(x)
lim hf (ǫx), φ(x)i = lim g(εx), =β dx .
ε→0 L(ε)
+ ε→0 L(ε)
+ x 1 x
This shows the case α = −1.
It remains to show the theorem when α ∈ {−2, −3, . . . }. Suppose the
degree is −k, k ∈ {2, 3, . . . }. It is easy to see that any primitive of degree
(k − 1) of f has quasi-asymptotics of degree −1 at the origin with respect
to L (in fact this is the content of Proposition 2.17 when combined with
Theorem 2.23). The (k − 1)-primitives of f are in S ′ (R), so we can apply
the case α = −1 to them, and then, by differentiation, it follows that f has
quasi-asymptotics at the origin in S ′ (R).
This completes the proof of Theorem 2.35.
2.11.2 Quasi-asymptotic extension problem in D′ (0, ∞)
The purpose of this subsection is to study the following problem. Suppose

that a distribution f ∈ D′ (R) with support in [0, ∞) has quasi-asymptotics
of degree α in the space D′ (0, ∞), that is, for each φ ∈ D(0, ∞)
hf (λx), φ(x)i ∼ λα L(λ) hg(x), φ(x)i . (2.87)
′
What can we say about the quasi-asymptotic properties of f in D (R)?
We can apply the techniques from 2.10.3 and 2.10.5 to give a complete
answer to this question. The answer depends on α. We start with the
quasi-asymptotic behavior at infinity, the same arguments are applicable
to quasi-asymptotics at the origin.
Let us start with the case α > −1. It is not difficult to show that g
must be of the form Cxα + /Γ(α + 1), for some constant C. Next, Proposi-
tion 2.17 still holds replacing the space D′ (R) by D′ (0, ∞) (actually this
holds without the restriction α > −1). Hence, the same argument given
in Theorem 2.26 applies here, but this time we only require the uniform
convergence on [1/2, 2], and hence we can still conclude the existence of
the integer such that (2.61) holds with the limit taken only as x → ∞.
Actually, because α > −1, relation (2.61) holds for any m-primitive of f .
Let F be the m-primitive of f supported on the interval [0, ∞), then we
have that
Cxα+m L(x)
F (x) ∼ , x → ∞,
Γ(α + m + 1)
so we have that F (λx) = CL(λ)(λx)α+m+ /Γ(α + m + 1) + o(λα+m L(λ)) in

′
the space S (R), differentiating m-times, we obtain the following result.
Theorem 2.36. Let f ∈ D′ (R) be supported on [0, ∞). If f has quasi-

asymptotic behavior of degree α > −1 in D′ (0, ∞), then it is a tempered
distribution and has the same quasi-asymptotic behavior in the space S ′ (R).
Suppose now that α < −1 and α is not a negative integer. This case
differs from the last one essentially in one point, we cannot conclude (2.61)
for every m-primitive of f but only for some of them. In any case, denoting
again by F the m-primitive (we keep m > −α−1) of f supported on [0, ∞),
we have that there exists a polynomial of degree at most m − 1 such that
Cxα+m L(x)
F (x) − p(x) ∼ , x → ∞;
Γ(α + m + 1)
therefore,
m−1
X
CL(λ)(λx)α+m
F (λx) = +
+ aj (λx)j+ + o(λα+m L(λ)) as λ → ∞ ,
Γ(α + m + 1) j=0
in the space S ′ (R), for some constants a0 , . . . , am−1 . Thus, our arguments
immediately yield the next theorem.
Theorem 2.37. Let f ∈ D′ (R) be supported on [0, ∞). Suppose that

(λx)α
+
f (λx) = CL(λ) + o(λα L(λ)) as λ → ∞ in D′ (0, ∞).
Γ(α + 1)
If α < −1 and α is not a negative integer, then f is a tempered distribution.
Moreover, there exist constants a0 , a1 , . . . , an (n < −α − 1) such that
n
X
(λx)α
+ δ (j) (x)
f (λx) = CL(λ) + aj j+1 + o(λα L(λ)) as λ → ∞ in S ′ (R).
Γ(α + 1) j=0 λ
When α = −k, k being a positive integer, the distribution g in (2.87)

must have the form Cx−k ∈ D′ (0, ∞), for some constant C; these distri-
butions are homogeneous as elements of D′ (0, ∞), but they do not have
homogeneous extensions to D(R). The behavior of f (λx) as λ → ∞ in
S ′ (R) is described in the next theorem.
We denote by Pf(H(x)/xk ) the distribution
Z ∞
H(x) φ(x)
Pf , φ(x) = F.p. dx, φ ∈ D(R),
xk 0 xk
where F.p. stands for the Hadamard finite part of the divergent integral
[56].
Theorem 2.38. Let f ∈ D′ (R) be supported on [0, ∞). Suppose that

H(x) L(λ)
f (λx) = CL(λ) +o as λ → ∞ in D′ (0, ∞) ,
(λx)k λk
where k is a positive integer. Then f is a tempered distribution and there
exist an associate asymptotically homogeneous function b satisfying
(−1)k−1
b(ax) = b(x) + CL(x) log a + o(L(x)), x → ∞ , (2.88)
(k − 1)!
for each a > 0, and constants a0 , a1 , . . . , ak−2 such that
k−2
X δ (j) (x)
L(λ) H(x) b(λ) (k−1) L(λ)
f (λx) = C k Pf + k δ (x) + aj j+1 + o
λ xk λ j=0
λ λk
(2.89)
′
as λ → ∞ in S (R).
Proof. For each n ∈ N, let Fn denote the n-primitive of f with support

in [0, ∞). Set C1 = (−1)k−1 C/(k − 1)!. Adapting the arguments of 2.10.5
and reasoning as in the previous two cases, we obtain the existence of a
positive integer m > k such that Fm is continuous and
m−k
xm−k xm−k X 1
Fm (x) = b1 (x) −C1 L(x) +pm−1 (x)+o(xm−k L(x)) ,
(m − k)! (m − k)! j=1 j
x → ∞, where b1 is a locally integrable associate asymptotically homo-
geneous function satisfying (2.88) and pm−1 is a polynomial of degree at
most m − 1. Throwing away the irrelevant terms of the polynomial pm−1
and using Theorem 2.27, we obtain the following asymptotic expansion as
λ → ∞ in the space S ′ (R),
(λx)m−k
+
Fm (λx) = b1 (λ) + C1 λm−k L(λ)lm−k (x)H(x)
(m − k)!
k−2
X (λx)m−j−1
+
+ aj + o(λm−k L(λ)) .
j=0
(m − j − 1)!
Differentiating (m − k)-times this expansion, we have that
k−2
X (λx)k−j−1
+
Fk (λx) = b1 (λ)H(x) + C1 L(λ)H(x) log x + aj + o(L(λ)) .
j=0
(k − j − 1)!
(2.90)
The well known formulas [56], p. 68,

d H(x)
(H(x) log x) = Pf
dx x
and

d H(x) H(x) (−1)n δ (n) (x)
Pf = −nPf +
dx xn xn+1 n!
imply that
k−1
X1
dk−1 H(x) k−1 H(x) (k−1)
k−1
Pf = (−1) (k − 1)! Pf k
− δ (x) .
dx x x j=1
j
Hence, differentiating (2.90) k-times, one has (2.89) with
 
k k−1
X
(−1) C  1
b(x) = b1 (x) + L(x).
(k − 1)! j=1 j
The corresponding result at the origin is stated in the next theorem.
Theorem 2.39. Let f0 ∈ D′ (0, ∞). Let L be slowly varying at the origin
and α ∈ R. Suppose that
f0 (εx) = εα L(ε)g0 (x) + o (εα L(ε)) as ε → 0+ in D′ (0, ∞) , (2.91)
′
g0 ∈ D (0, ∞). Then f0 admits extensions to [0, ∞). Let f be any of such
extensions. Then f has the following asymptotic properties at the origin:
(i) If α ∈
/ −N, then there exist constants C, a0 , . . . , an such that
n
X
f (εx) = aj δ (j) (εx) + Cεα L(ε)xα α
+ + o(ε L(ε)) ,
j=0
+ ′
as ε → 0 in D (R).
(ii) If α = −k, k ∈ N, then there are constants C, a0 , . . . , an and
an associate asymptotically homogeneous function of degree 0 with
respect to L satisfying
(−1)k−1
b(ax) = b(x) + CL(x) log a + o(L(x)) , (2.92)
(k − 1)!
such that
Xn
H(x)
f (εx) = aj δ (j) (εx) + b(ε)δ (k−1) (εx) + Cε−k L(ε)Pf
j=0
xk

+ o ε−k L(ε) , (2.93)
as ε → 0+ in D′ (R).
2.11.3 Quasi-asymptotics at infinity and spaces Vβ′ (R)
Sometimes is very useful to have the right of evaluating (2.41) in more test
functions than in S(R), this section is dedicated to give some conditions
under the test function which guarantee one can do this for the quasi-
asymptotic behavior at ±∞. We shall now improve Theorem 2.10 and its
corollary (cf. 2.6).
We need the following definition.
Definition 2.13. Let φ ∈ E(R) and β ∈ R. We say that
φ(x) = O(|x|β ) strongly as |x| → ∞ , (2.94)
if for each m ∈ {0, 1, 2, . . . }
φ(m) (x) = O(|x|β−m ) as |x| → ∞ . (2.95)
The set of φ satisfying Definition 2.13 for a particular β forms the space
Vβ (R) which we topologize in the obvious way [56]. These spaces and their
dual spaces are very important in the theory of asymptotic expansions of
S
distributions [56]. In fact, if we set V(R) = Vβ (R) (the union having
a topological meaning), we have that V ′ (R) is the space of distributional
small distributions at infinity [47], [56], they satisfy the moment asymptotic
expansion at infinity (cf. Example 3 in 2.9). We point out that in [56] these
spaces are denoted by = Vβ′ = Kβ′ and V ′ = K′ .
The next theorem shows that if f has quasi-asymptotics at ±∞, then
the distributional evaluation of f at φ ∈ Vβ (R) makes sense under some
conditions on β, specifically, we show that f has extensions to some of the
spaces Vβ (R).
Theorem 2.40. Let f ∈ D′ (R) have quasi-asymptotic behavior of degree

α at ±∞ with respect to the slowly varying function L. If α + β < −1, then
f admits extensions to Vβ (R).
Proof. Let σ > 0 such that α + β + σ < −1, then from Theorem 2.26,
Theorem 2.33 and Lemma 2.5 we deduce that there exist m ∈ N and a
continuous m-primitive of f , say F , such that
m+α+σ
F (x) = O(|x| ) as |x| → ∞ . (2.96)
Notice that here we have used that L(x) = O(xσ ) as x → ∞ [148]. So it is

evident that the extension of f to Vβ (R) is given by
Z ∞
m
hf (x), φ(x)i = (−1) F (x)φ(m) (x)dx, φ ∈ Vβ (R) , (2.97)
−∞
which in view of (2.95) and (2.96) is well-defined and defines an element of

Vβ′ (R).
We now show that the quasi-asymptotic behavior remains valid in Vβ′ (R)
for one extension of f , with the assumption under β imposed in Theorem
2.40.
Theorem 2.41. Let f ∈ D′ (R) have quasi-asymptotic behavior at ∞ of

degree α with respect to a slowly varying function L, then f admits an ex-
tension to Vβ (R) which has the same quasi-asymptotics in Vβ′ (R), provided
that α + β < −1.
Proof. The proof is similar to that of Theorem 2.35 with some modifica-
tions in the estimates. We use one of the extensions of f found in Theorem
2.40; denote the extension by f˜. We shall divide the proof into two cases:
α∈ / {−1, −2, −3, . . . } and α ∈ {−1, −2, −3, . . . }.
Suppose its degree is α ∈
/ {−1, −2, −3, . . . } and
(λx)α
− (λx)α
+
f (λx) = C− L(λ) + C+ L(λ) + o (λα L(λ)) as λ → ∞ ,
Γ(α + 1) Γ(α + 1)
in D′ (R). Find σ > 0 such that α + β + σ < −1. Then from Theorem 2.26,
there are an m such that m + α > 0 and a continuous m-primitive F of f
such that
xm |x|
α
m+α
F (x) = L (|x|) (C− H(−x) + C+ H(x)) + o |x| L (|x|) ,
Γ(m + α + 1)
x → ∞. We recall that H denotes the Heaviside function. We make the
usual assumptions over L (cf. 2.10.1), assume that L is positive, defined
and continuous in (0, ∞) and there exists M1 > 0 such that
L(λx)
≤ M1 max xσ , x−σ , λ ≥ 1, x ∈ (0, ∞) . (2.98)
L(λ)
Let φ ∈ Vβ (R). As in the proof of Theorem 2.35, we may assume that
supp φ ⊆ [1, ∞), and the proof would be complete after we show
D E
xα
f˜(λx), φ(x) ∼ C+ λα L(λ) +
, φ(x) , (2.99)
Γ(α + 1)
as λ → ∞.
Set
F (x)
G(x) = for x ≥ 1 , (2.100)
xα+m L(x)
then
C+
lim G(x) = . (2.101)
x→∞ Γ(α + m + 1)
So, we can find a constant M2 > 0 such that
|G(x)| < M2 , globally. (2.102)
Relation (2.102) together with (2.98) show that for λ ≥ 1,

G(λx) L(λx) xα+m φ(m) (x) ≤ M1 M2 xα+m+σ φ(m) (x) H(x − 1) .
L(λ)
Since φ ∈ Vβ (R), the right hand side of the last estimate belongs to
L1 (R) and thus we can use the Lebesgue dominated convergence theorem
to obtain,
D E
f˜(λx), φ(x) Z ∞
m L(λx) α+m (m)
lim α
= lim (−1) G(λx) x φ (x)dx
λ→∞ λ L(λ) λ→∞ 0 L(λ)
Z ∞
C+
= (−1)m xα+m φ(m) (x)dx
Γ(α + m + 1) 0

xα+
= C+ , φ(x) .
Γ(α + 1)
This shows the result in the case α ∈
/ {−1, −2, −3, . . . } .
We now consider the case α = −k, k ∈ N. Assume that

f (λx) = γλ−k L(λ)δ (k−1) (x) + βλ−k L(λ)x−k + o λ−k L(λ) ,
as λ → ∞ in D′ (R). As in the last case, it suffices to assume that φ ∈
Vβ (R), supp φ ⊆ [1, ∞) and show that
Z ∞
λk D ˜ E φ(x)
lim f (λx), φ(x) = β dx .
λ→∞ L(λ) 1 xk
So, set g(x) = xk f (x), then
g(λx) = βL(λ) + o(L(λ)) as λ → ∞ in D′ (R). (2.103)
k
But φ ∈ Vβ (R) implies φ(x)/x ∈ Vβ−k (R) then since the degree of the
quasi-asymptotic behavior of g is 0, last case implies that (2.103) is valid
′
in Vβ−k (R) for a suitable extension g̃ because β − k < −1, therefore
Z ∞
λk D ˜ E 1 φ(x) φ(x)
lim f (λx), φ(x) = lim g̃(λx), k =β dx .
λ→∞ L(λ) λ→∞ L(λ) x 1 xk
This completes the proof of Theorem 2.41.

The importance Theorem 2.41 lies in the fact that we can relax the
growth restrictions over the test functions, this permits to apply quasi-
asymptotics to obtain ordinary asymptotics in many interesting situations,
for example for certain integral transforms or for solutions to partial differ-
ential equations. We discuss a simple example.
Example. Let f ∈ D′ (R) have quasi-asymptotic behavior at infinity
of degree α < 1,
f (λx) = λα L(λ)g(x) + o (λα L(λ)) as λ → ∞ in D′ (R) .
Consider the Poisson kernel,
1
P (t) = .
π (t2 + 1)
Clearly P ∈ V2 (R). By Theorem 2.41, f has an extension f˜ such that the
evaluation of f˜ at P is well defined and f˜ preserves the quasi-asymptotic
properties of f . Thus

1 x−t
U (z) = U (x + yi) = f (t), P
y y
is a solution of the boundary value problem
∂2U ∂2U
2
+ = 0, U (x + i0+ ) = f (x) (in D′ (R)) .
∂x ∂y 2
Using Theorem 2.41, we can find the asymptotic behavior of U at infinity
over cones. Indeed, let 0 < σ < π/2, then Theorem 2.41 implies that as
r→∞
U (reiϑ ) ∼ sinα (ϑ)Cϑ rα L(r), uniformly for ϑ ∈ [σ, π − σ] ,
where Cθ = g ∗ P (cot ϑ).
2.12 Quasi-asymptotic boundedness
This section is intended to study the structure of the distributional relation

f (λx) = O(ρ(λ)) ,
where here λ → ∞ or λ → 0+ and ρ is a regularly varying function. Our
approach to the problem follows the exposition from [172]. Distributions
satisfying this relation will be called quasi-asymptotically bounded distribu-
tions, we make this more precise in the following definition.
Definition 2.14. Let L be a slowly varying function at infinity (respec-

tively at the origin). We say that f ∈ D′ is quasi-asymptotically bounded
at infinity (at the origin) in D′ (R) with respect to λα L(λ), α ∈ R, if
hf (λx), φ(x)i = O (λα L(λ)) as λ → ∞ ∀φ ∈ D(R), (2.104)
(respectively λ → 0+ ). If (2.104) holds, it is also said that f is quasi-

asymptotically bounded of degree α at infinity (at the origin) with respect
to the slowly varying function L. We express (2.104) by
f (λx) = O(λα L(λ)) as λ → ∞ in D′ (R) , (2.105)
(respectively λ → 0+ ).
Note that in analogy to the quasi-asymptotic behavior of distributions,

we may talk about (2.105) in other spaces of distributions. In the case
at infinity, It will follow from our structural theorem that f ∈ S ′ (R) and
actually the relation (2.105) holds in S ′ (R). The case at the origin is related
to the problem of extension of distributions from R \ {0} to R. Indeed, if
f ∈ D′ (R \ {0}) and (2.104) holds for all φ ∈ D(R \ {0}), we will see later
that f admits an extension to R.
We now proceed to obtain the structure of quasi-asymptotically
bounded distributions. For this aim, the program established in 2.10 will
be followed. We will integrate the relation (2.105) and the coefficients of
this integration will satisfy the properties of the following definition.
Definition 2.15. A function b is said to be asymptotically homogeneously

bounded of degree α at infinity with respect to the slowly varying function
L if it is measurable and defined in some interval (A, ∞), A > 0, and for
each a > 0
b(ax) = aα b(x) + O(L(x)), x → ∞ . (2.106)
Similarly, one defines asymptotically homogeneously bounded functions

at the origin. Our first goal is to study the asymptotic properties of this
class of functions. Proceeding as in Lemma 2.2, or using the results of [148],
Section 2.4, one has that (2.106) must hold uniformly in compact subsets of
(0, ∞). Most of the proofs of the following results are the analog to those
for asymptotically homogeneous functions by replacing the o symbol by
the O symbol and making obvious modifications to the estimates, therefore
they will be omitted.
Proposition 2.19. Let b be asymptotically homogeneously bounded at in-

finity (at the origin) with respect to the slowly varying function L. If the
degree is negative (respectively positive), then b(x) = O(L(x)), as x → ∞
(x → 0+ ).
Proposition 2.20. Let b be asymptotically homogeneously bounded at in-

finity (at the origin) with respect to the slowly varying function L. If the
degree α is positive (respectively negative), then there exits a constant γ
such that b(x) = γxα + O(L(x)), as x → ∞ (x → 0+ ).
Note that for the case at infinity since L(x) = O(xσ ) as x → ∞, for
any σ > 0, then any asymptotically homogeneously bounded function of
degree 0 at infinity satisfies that b(x)/xσ is asymptotically homogeneously
bounded of degree −σ with respect to the trivial slowly varying function
L ≡ 1 and hence by Proposition 2.19 it satisfies b(x) = O(xσ ) as x → ∞,
hence for large argument it is a regular tempered distribution. Similarly,
any asymptotically homogeneously bounded function of degree 0 at the
origin satisfies b(x) = O(x−σ ) as x → 0+ , for any σ > 0, consequently it
is a distribution for small argument. The proof of the next proposition is
totally analogous to those of Theorems 2.24 and 2.27, and therefore it will
be omitted.
Proposition 2.21. Let b be asymptotically homogeneously bounded of de-

gree zero at the infinity (at the origin) with respect to the slowly vary-
ing function L. Suppose that b is locally integrable on [A, ∞) (respectively
(0, A]). Then
b(λx)H(λx − A) = b(λ)H(x) + O(L(λ)) as λ → ∞ in S ′ (R), (2.107)
(resp. b(λx)(H(x) − H(λx − A)) = b(λ)H(x) + O(L(λ)) as λ → 0+ in
D′ (R)).
Corollary 2.3. Let b be an asymptotically homogeneously bounded function

of degree 0 at infinity (at the origin) with respect to L. Then, there exists
c ∈ C ∞ [0, ∞), being asymptotically homogeneously bounded of degree 0,
such that b(x) = c(x) + O(L(x)) as x → ∞ (resp. as x → 0+ ).
Proof. We only show the assertion at infinity, the case at the origin is sim-
R ∞ bounded in [B, ∞). Take
ilar. Find B such that b is locally R∞φ ∈ D(R) sup-
ported in (0, ∞) such that 0 φ(t)dt = 1 and set c(x) = B/x b(xt)φ(t)dt,
the corollary now follows from Proposition 2.21.
The main connection between quasi-asymptotically bounded distribu-

tions and asymptotically homogeneously bounded functions is given in the
next proposition, again the proof will be omitted since it is analogous to
that of Proposition 2.17.
Proposition 2.22. Let f ∈ D′ (R) be quasi-asymptotically bounded of de-

gree α at infinity (at the origin) with respect to the slowly varying function
L. Let m ∈ N. Then, for any given Fm , an m-primitive of f in D′ (R),
there exist functions b0 , . . . , bm−1 , continuous on (0, ∞), such that
m−1
X xm−1−j
Fm (λx) = λα+m bj (λ) + O λα+m L(λ) in D′ (R),
j=0
(m − 1 − j)!
(2.108)
as λ → ∞ (respectively λ → 0+ ), where each bj is asymptotically homoge-
neously bounded of degree −α − j − 1 with respect to L.
Thus we obtain from Propositions 2.19–2.22 our first structural theorem.
Theorem 2.42. Let f ∈ D′ (R) and α ∈ / −N. Then f is quasi-

asymptotically bounded of degree α at infinity (at the origin) with respect to
the slowly varying function L if and only if there exist m ∈ N, m + α > 0,
and a continuous (continuous near 0) m-primitive F of f such that

m+α
F (x) = O |x| L (|x|) , (2.109)
as |x| → ∞ (respectively x → 0) in the ordinary sense. Moreover, in the
case at infinity, f belongs to S ′ (R) and is quasi-asymptotically bounded of
degree α with respect to L in S ′ (R).
Proof. We only discuss the case at infinity, the proof of the assertion at
the origin is similar to this case. It follows from Proposition 2.22, Proposi-
tion 2.19 and Proposition 2.20 that given m ∈ N and an m-primitive Fm ,
there is a polynomial pm−1 of degree at most m − 1 such that
Fm (λx) = pm−1 (λx) + O(λα+m L(λ)) as λ → ∞ in D′ (R), (2.110)
′
from the definition of boundedness in D (R) it follows that there is an
m > −α such that (2.110) holds uniformly for x ∈ [−1, 1]. We let F =
Fm − pm−1 , so by taking x = −1, x = 1 and replacing λ by x in (2.110) we
obtain (2.109). The converse follows by observing that (2.109) implies that
F (λx) = O(λα+m L(λ)) in S ′ (R) which gives the result after differentiating
m-times.
We now analyze the case of negative integral degree.
Theorem 2.43. Let f ∈ D′ (R) and let k be a positive integer. Then f is

quasi-asymptotically bounded of degree −k at infinity (at the origin) with
respect to L if and only if there exist m > k, an asymptotically homoge-
neously bounded function b of degree 0 at infinity (at the origin) with respect
to L and a continuous (continuous near 0) m-primitive F of f such that

m−k
F (x) = b (|x|) xm−k + O |x| L (|x|) , (2.111)
as |x| → ∞ (x → 0). Moreover (2.111) is equivalent to have

ak−m F (ax) − (−1)m−k F (−x) = O xm−k L(x) , (2.112)
as x → ∞ (x → 0+ ), for each a > 0. In the case at infinity, it follows that

f is quasi-asymptotically bounded of degree −k with respect to L in S ′ (R).
Proof. Again we only give the proof of the assertion at infinity, the
case at the origin is similar. If f (λx) = O(λ−k L(λ)) in D′ (R), then af-
ter k − 1 integrations Proposition 2.22 and Proposition 2.20 provide us of a
(k−1)-primitive of f which is quasi-asymptotically bounded of degree −1 at
infinity with respect to L, hence we may assume that k = 1. Next, Propo-
sition 2.22, Proposition 2.19 and the definition of boundedness in D′ (R)
give to us the existence of an m > 1, an asymptotically homogeneously
bounded function of degree −1 with respect to L and an m-primitive F of
f such that F (λx) is continuous for x ∈ [−1, 1] (hence F is continuous on
R) and F (λx) = λm−1 b(λ)xm−1 + O(λm−1 L(λ)) as λ → ∞ uniformly for
x ∈ [−1, 1]. By taking x = −1, x = 1 and replacing λ by x one gets (2.111).
Assume now (2.111), by using Corollary 2.3, we may assume that b is lo-
cally integrable on [0, ∞). This allows the application of Proposition 2.21
to deduce that F (λx) = λm−1 b(λ)xm−1 +O(λm−1 L(λ)) as λ → ∞ in S ′ (R)
and hence the converse follows by differentiating m-times. That (2.111) im-
plies (2.112) is a simple calculation; conversely, setting b(x) = xk−m F (x)
for x > 0, one obtains (2.111).
Remark I.1. Even if not assumed initially, the proof of Theorem 2.43
forces (2.112) to hold uniformly on compact subsets of (0, ∞).
We remark that the results from 2.11 are also true in the context of
quasi-asymptotic boundedness. Indeed, if one proceeds as in 2.11 but now
using the structural theorems of the present section, then one obtains the
proofs for the following theorems.
Theorem 2.44. Let f ∈ S ′ (R). If f is quasi-asymptotically bounded at

0, with respect to a slowly varying function L, in D′ (R), then f is quasi-
asymptotically bounded at 0 of the same degree with respect to L in the space
S ′ (R).
Theorem 2.45. Let f ∈ D′ (R) satisfy f (λx) = O(λα L(λ)) as λ → ∞ in

the space D′ (R). If α + β < −1, then f admits an extension to Vβ (R)
which is equally quasi-asymptotic bounded in the space Vβ′ (R).
Theorem 2.46. Let f0 ∈ D′ (0, ∞). Let L be slowly varying at the origin
and α ∈ R. Suppose that
f0 (εx) = O (εα L(ε)) as ε → 0+ in D′ (0, ∞). (2.113)
Then f0 admits extensions to [0, ∞). Let f be any of such extensions. Then
f has the following asymptotic properties at the origin:
(i) If α ∈
/ −N, then there exist constants a0 , . . . , an such that
n
X
f (εx) = aj δ (j) (εx) + O(εα L(ε)),
j=0
as ε → 0+ in D′ (R).
(ii) If α = −k, k ∈ N, then there are constants a0 , . . . , an and an
associate asymptotically homogeneously bounded function of degree
0 with respect to L,
b(ax) = b(x) + O(L(x)), (2.114)
such that
n
X
f (εx) = aj δ (j) (εx) + b(ε)δ (k−1) (εx) + O ε−k L(ε) , (2.115)
j=0
as ε → 0+ in D′ (R).
2.13 Relation between the S-asymptotics and

quasi-asymptotics at ∞
It is not easy to find conditions for a T ∈ F ′ which imply that T has both
the S-asymptotics and quasi-asymptotics. The S-asymptotics is a local
property (Theorem 1.4), whereas the quasi-asymptotics has in general a
global character (Proposition 2.4, Example 4 in 2.3, and Example 3 in
2.9). Also, we have to find a subspace of F ′ in which we can compare these
two definitions of the asymptotic behavior and choose the class of functions
which will measure the asymptotic behavior following these definitions.
We have seen that the space of tempered distributions is a “natural”
one for the quasi-asymptotics (Theorem 2.3, Theorem 2.10 and 2.11.2),
while for the S-asymptotics the space K′ 1 has this role (Theorem 1.17).
The following examples illustrate the problem of comparison of these two
types of asymptotic behavior in S ′ (R).
i) The regular distribution T (t) = H(t)eiat , t ∈ R, a 6= 0, has the
i
quasi-asymptotics δ in S ′ (R) related to c(k) = k −1 (cf. [32]):
a
Z∞ Z∞
ikat ikat 1 x
khH(kt)e , ϕ(t)i = k e ϕ(t)dt = ϕ d(eiax )
ia k
0 0
Z∞
−1 1 x i
= ϕ(0) − eiax ϕ′ dx → ϕ(0), k → ∞ .
ia k k a
0
But the distribution T has no S-asymptotics related to hα with a U 6= 0

for any α ∈ R. We start with
Z∞
ia(t+h) iah
hH(t + h)e , ϕ(t)i = e eiat ϕ(t)dt
−h
Z∞
iah
∼e eiat ϕ(t)dt, h ∈ R+ .
−∞
This distribution has the S-asymptotics, but related to the oscillatory func-
tion c(h) = eiah .
ii) The regular distribution T (t) = H(t) sin t, t ∈ R, has the quasi-
asymptotics related to c(k) = k −1 , but it has no S-asymptotics at all:
Z∞ u Z∞
1 u
khH(kt) sin kt, ϕ(t)i = sin uϕ du = ϕ(0) + ϕ′ cos u du .
k k k
0 0
For the S-asymptotics we have, h → ∞,

Z∞ Z∞
hH(t+h) sin(t+h), ϕ(t)i = cos h sin tϕ(t)dt+sin h cos tϕ(t)dt+o(1) .
−∞ −∞
√
iii) For the regular distribution T = H(t) sin t, t ∈ R, we cannot find
an α ∈ R and a distribution Uα 6= 0 such that
Z∞ √
lim k α sin kt ϕ(t)dt = hUα , ϕi, ϕ ∈ S(R) .
k→∞
0
Suppose on the contrary that such α and Uα do exist. We choose ϕ ∈ S+

such that, ϕ(t) = e−pt , t > 0, where Rep > 0. Then, we have
Z∞ √
α
lim k sin kte−pt dt = hUα (t), e−pt i .
k→∞
0
√ p
The value of the last integral is πk/ 4p3 exp(−k/4p) and hUα (y), e−pt i
is the Laplace transform of Uα . Thus the last relation says that the Laplace
transform of Uα equals zero for Rep > 0, hence Uα = 0.
If we change the basic space, the conclusion can be quite different. Sup-
pose that the basic space is K1 (R).
H(x) x
The function H(x) cosh x = (e + e−x ), x ∈ R defines an element
2
in K′ 1 and {exp(−px2 ); p > 0} is in K1 .
In order to find the quasi-asymptotics at infinity in K′ 1 of H(x)cosh(x),
x ∈ R, we use
Z∞ Z∞ √
2 1 √ dx π
cosh(kx)e−px dx = cosh(k x)e−px √ = √ exp(k 2 /4p) .
2 x 4p
0 0
r
This shows that there exists no function c(k) of the form eak k b L(k),
where a, b ∈ R, 0 ≤ r < 2, L(k) is a slowly varying function, such that
H(x) cosh x has the quasi-asymptotics related to c(k) with a limit U 6= 0.
For the S-asymptotics, we have

Z∞ Z∞ Z∞
−h 1 x 1 −2h
e cosh(x + h)ϕ(x)dx = e ϕ(x)dx + e ϕ(x)dx
2 2
−h −h −h
Z∞
1
→ ex ϕ(x)dx, k → ∞, ϕ ∈ K1 .
2
−∞
Therefore, H(x) cosh x, x ∈ R, has the S-asymptotics related to c(h) =

1
eh with the limit U = ex .
2
The common space in which we can compare the two types of asymp-
totic behavior is the space of tempered distributions and the common class
of functions for the comparison is the class of regularly varying functions
{hα L(h); α ∈ R}, where L is a slowly varying function (cf. 0.3).
The first result of this kind is the following one.
Proposition 2.23. ([32]). Let g ∈ S ′ + , α > −1, and let ϕ0 ∈ S(R) be

such that (F ϕ0 )(x) = 1 on a neighborhood of zero. If there exists
Z
α
lim hg(x + h)/h , ϕ0 (x)i = C ϕ0 (x)dx, C 6= 0 ,
h→∞
q
then g(kx) ∼ k α Cfα+1 (x), k → ∞, in S ′ (R). (For the function fα see
0.4).
Later on, this result has been improved (see Theorem 6, Chapter I, §3.3
in [192]) in such a way that, instead of hα , hα L(h) was used, where L is a
slowly varying function. In case α ≤ −1, the situation is quite different. In
[120], one can find precise results for α = −1, but for α < −1 the obtained
results must include the quasi-asymptotic behavior when the limit equals
zero, as well.
We quote some results of this kind proved in [159]. Denote L(x) b =
Rx L(t)
b
dt, x > a. If L(x) → ∞ and L is slowly varying, then L b is slowly
a t
varying, as well.
Theorem 2.47. Suppose that T ∈ S ′ + has S-asymptotics related to

hα L(h). Then T has also the quasi-asymptotics related to hβ L′ (h), where
β and L′ are determined as follows: If α > −1, then α = β, L′ = L; if
b
α = −1 and L(x) → ∞, x → ∞, then β = −1, L′ = L; b if α < −1 or
b
α = −1 and L(x) ′
< ∞, then β = −1, L = const., but the limit can be zero.
Finally, as an illustration of the case α < −1 in Theorem 2.47, we

can use the distribution T = C1 δ (r) + C2 (xα L(x))+ ; the quasi-asymptotics
depends not only on α but also on r.
We can compare the quasi-asymptotics with the S-asymptotics not only
for the elements belonging to S ′ + but also when they belong to D′ . In this
case, we shall use Definition 2.4 instead of Definition 2.2. We know that
the general form of the function c, related to which we can measure the
S-asymptotics, is c(h) = exp(αh)L(exp h), α ∈ R.
Theorem 2.48. If T ∈ D′ (R) has S-asymptotics related to c(h) = exp(αh)·

L(exp h) with α 6= 0 and with non-zero limit, then T cannot have quasi-
asymptotics.
At the end of this part dealing with the quasi-asymptotics, we shall

cite additionally the following papers: [8], [46], [50], [55], [113], [121], [124],
[159], [174], [199] and [201].

Chapter II
Applications of the Asymptotic

Behavior of Generalized
Functions
3 Asymptotic behavior of solutions to partial differential

equations
3.1 S-asymptotics of solutions
We refer to [152], [154], [69] and [160] for the S-asymptotics of solutions of
partial differential equations.
Let P (y), y ∈ Rn be a polynomial. By reg(1/P (y))we denote a solu-
tion, belonging to S ′ , of the equation P (y) · X = 1. It is well known that
Hörmander [70] proved that the last equation can always be solved in S ′ if
P 6≡ 0 (cf. [70], [190]).
Proposition 3.1. Let

m
X
P (D) = aα Dα , aα ∈ R, α ∈ Nn0 . (3.1)
|α|≥0
A necessary and sufficient condition that there exists a solution to the

equation
P (D)E = δ ,
such that
s
E(t + h) ∼ c(h)U (t), h ∈ Γ in S ′ (3.2)
is that there exists
1 1
lim exp(−it · h)reg = F [U ](t) in S ′ . (3.3)
h∈Γ,khk→∞ c(h) P (−it)
(F denotes the Fourier transform).
171
172 Applications of the Asymptotic Behavior of Generalized Functions
Proof. We know that E ∈ S ′ is a fundamental solution of the operator

P (D) if and only if F [E] is a solution to the equation P (−ix)F [E] = 1 (see
[190], p. 192]) where, F [E] = reg(1/P (−it)) = f ∈ S ′ .
We have E(· + h) = F −1 [exp(−it · h)f (t)] and
hE(x + h)/c(h), φ(t)i

1
= F −1 [exp(−it · h)f (t)](x), φ(x) . (3.4)
c(h)
Let us suppose now that the limit (3.2) exists in S ′ . Then by (3.4) there
exists the limit
1
lim F −1 [exp(−it · h)f (t)](x) in S ′ .
We know that
exp(−it · h)f (t) = F [F −1 [exp(−iy · h)f (y)](x)](t) .
The continuity of F implies (3.3).
Theorem 3.1. Let P (D) be of the form (3.1). A sufficient condition that
there exists a solution X to the equation
P (D)X = G, G ∈ E ′ , G 6= 0, (3.5)
s
such that X(t + h) ∼ c(h) (G ∗ U )(t), h ∈ Γ in S ′ is that there exists the
limit (3.3).
Proof. We proved that the limit (3.3) is necessary and sufficient for
s
E(t + h) ∼ c(h)U (t), h ∈ Γ in S ′ , where E is a solution to equation (3.1).
In order to find the S-asymptotics of X = E ∗ G, which is a solution to
(3.5), we have only to apply Theorem 1.2 b).
Remark. Theorem 3.1 also holds if G ∈ O′ C ⊂ S ′ . The convolution
E ∗ G exists and it is a separately continuous mapping of (O′ C , S ′ ) →
O′ C ∗S ′ ⊂ S ′ (see Chapter VII, 5 in [146]). Moreover, a solution to equation
(3.5) is of the form X = E ∗ G. This solution is unique in the class of
distributions for which the convolution with E exists (see Chapter III, §11
in [190]).
∂ ∂
Let P0 (i ) be the principal part of P (i ), where P is given by (3.1).
∂x ∂x
∂
If P0 (y) 6= 0 for any real y 6= 0, then P (i ) is called an elliptic operator.
∂x
3. Asymptotic behavior of solutions to partial differential equations 173
Assume that the coefficients of P are real numbers. If the elliptic oper-
ator is also homogeneous (i.e., if it coincides with its principal part), then
the fundamental solution G(x), x ∈ Rn , has the form [61]:

 A x rm−n , x ∈ Rn ,

 for n odd, or n even and m < n
 r
G(x) =

 x m−n

B r + C(x) log r, x ∈ Rn , for n even and m ≥ n,
r
(3.6)
n
where A and B are analytic functions on kyk= 1 and C(x), x ∈ R , is a
polynomial in x of degree m−n; m is the degree of the principal polynomial
Pn
P0 (y) (m is then necessarily an even number); r2 = kxk2 = x2i .
i=1
Our aim is to find the S-asymptotics of the fundamental solution G
given by (3.6).
x
First, we shall treat the function A( )rm−n , x ∈ Rn , m > n; A is
r
analytic on the sphere {y; kyk= 1}. Γ will be the ray {ρw; ρ > 0} for a
fixed w ∈ Rn , kwk= 1. In this case the S-asymptotics of G is given by
J ≡ lim hρn−m A((x + ρw)/kx + ρwk)kx + ρwkm−n , ϕ(x)i

ρ→∞
Z m−n
x x x
= lim A +w ϕ(x)dx, ϕ ∈ S .
ρ→∞ ρ ρ + w ρ + w
Rn
Since
m−n
x x x
A +w
ρ ρ + w ρ + w
" n
#(m−n)/2
X
2
≤M (|xi |+|wi |) , ρ > 1,
i=1
where M = sup{|A(y)|, kyk= 1}, we can use Lebesgue’s theorem which

gives
Z Z
m−n
J = A(w/kwk)kwk ϕ(x)dx = A(w) ϕ(x)dx, ϕ ∈ S .
Rn Rn
x
We prove that A rm−n , x ∈ Rn , m > n, has the S-asymptotics
r
related to ρm−n on the ray {ρw, ρ > 0}, kwk= 1 and with the limit U = A.
x The next step is to find the S-asymptotics of the function

A rm−n , x ∈ Rn , but in case when m < n, m > 0.
r
Suppose that in the expression for J the function ϕ belongs to D(Rn )
and supp ϕ ⊂ I n , I = (−a, a), a ∈ R+ . Then, for ρ ≥ a/{maxkwi k; i =
1, . . . , n}
Z
lim ρn−m A((x + ρw)/kx + ρwk)kx + ρwkm−n ϕ(x)dx
ρ→∞
Rn
Z
= A(w) ϕ(x)dx, ϕ ∈ D .
Rn
x
Hence, A rm−n , x ∈ Rn , m < n, has the S-asymptotics in D′ with the
r
limit A. Since D is dense in S, to prove that this function has the same
S-asymptotics in S ′ it suffices to prove that ρn−m J(ρ), ρ > 0, is bounded
(cf. the Banach–Steinhaus theorem), where
Z
x + ρw
J(ρ) = A kx + ρwkm−n ϕ(x)dx, ρ > 0, ϕ ∈ S .
kx + ρwk
Rn
Therefore, we shall split the integral J(ρ) into three parts and use the
following inequalities:
For kxk≤ ρ/2 and kxk≥ 3ρ/2, we have kx + ρwk≥| kxk−ρ |≥ ρ/2 and

Z Z
x + ρw
n−m m−n
ρ + A kx + ρwk ϕ(x)dx
kx + ρwk
kxk≤ ρ kxk≥ 3ρ
2 2
n−m Z
2
≤2 M |ϕ(x)|dx, ϕ ∈ S, 0 < m < n ,
ρ
Rn
where M = sup|A(w)|, kwk= 1. But

kxk≤ 3ρ

Z 2 x + ρw

A kx + ρwkm−n ϕ(x)dx
kx + ρwk
kxk≥ ρ
2
3ρ
kxk≤
Z 2
M
≤K ρ2 k
kx + ρwkm−n dx → 0, ρ → ∞ ,
(1 + 4 )
kxk≥ ρ
2
where k is any positive number.

x
Consequently, A rm−n , x ∈ Rn , m < n, has the S-asymptotics in
r
S ′ , as well. Let us note that we omitted the case m = n, because m is an
even number.
Now, we have to find the S-asymptotics of G for n even and n ≥ m.
After this, it is enough to treat only C(x) log r, x ∈ Rn .
P
Assume now that C(x) = αi xi , x ∈ Rn , m ≥ n, where αi ∈ R,
|i|≤m−n
i = (i1 , . . . , in ), xi = xi11 . . . xinn ; |i|= i1 + · · · + in . Then,
lim ρn−m (log ρ)−1 hC(x + ρw) logkx + ρwk, ϕ(x)i
ρ→∞
Z X
n−m −1
= lim ρ (log ρ) αi (x + ρw)i (log ρ
ρ→∞
Rn |i|≤m−n
X Z
x
+ logk + wk)ϕ(x)dx = αi wi ϕ(x)dx
ρ
|i|=m−n Rn
by the same reason as in the first case.

The results of this paragraph can be expressed in the following
proposition:
Proposition 3.2. If G ∈ S ′ is a fundamental solution to an elliptic ho-

mogeneous operator P of degree m (see (3.6)), then on the ray {ρw; ρ >
0}, kwk= 1, G has the S-asymptotics:
s
G(x + ρw) ∼ ρm−n · G(w), ρ → ∞, for n odd, or n even
and m < n;
s (3.7)
G(x + ρw) ∼ ρm−n log ρ · D(w), ρ → ∞, for n even
and m ≥ n,
where D(w) = lim G(ρw)/(ρm−n log ρ).
ρ→∞
Now, we can give the S-asymptotic behavior of the solutions to

∂
P i u(x) = f (x), f ∈ O′ C , (3.8)
∂x
where P is an elliptic homogeneous operator of degree m in n dimensions.
We recall that a solution to equation (3.8) is of the form u = G ∗ f (G
being the fundamental solution to the operator P given by (3.6) and belongs
to S ′ . Namely, since G belongs to S ′ and f to O′ C , it follows that G ∗ f
exists and belongs to S ′ [190]. Moreover, the mapping: (G, f ) → G ∗ f is
separately continuous.
Theorem 3.2. Let G be the fundamental solution to the elliptic differential

equation

∂
P i G(x) = δ(x)
∂x
and
D(w) = lim G(ρw)/(ρm−n log ρ);
ρ→∞
P is a homogeneous operator of degree m in n dimensions. Equation

(3.8) has a solution u = G ∗ f belonging to S ′ . The solution u has the
S-asymptotics on the ray {ρw; ρ > 0}, kwk= 1, ρ → ∞ :
s
u(x + ρw) ∼ ρm−n · (G(w) ∗ f ), for n odd, or n even
and m < n;
s
u(x + ρw) ∼ ρm−n log ρ · (D(w) ∗ f ), for n even
and m ≥ n.
Proof. Since G has the S-asymptotics given by (3.7), u = G ∗ f and the

assertion of the Theorem follows from (3.7).
The next proposition shows the use of the S-asymptotic expansion in
D′ .
Proposition 3.3. Suppose that E is a fundamental solution of the operator

X
P (D) = aα Dα , aα ∈ R, α ∈ Nn0
|α|≥0
such that
∞
X
s
E(t + h) ∼ Un (t, h) | {cn (h)}, khk→ ∞, h ∈ Γ .
n=1
Then, there exists a solution X to equation
P (D)X = G, G ∈ E ′ (3.9)
which has the S-asymptotic expansion
∞
X
s
X(t + h) ∼ (G ∗ Un (t, h)) | {cn (h)}, khk→ ∞, h ∈ Γ .
n=1
Proof. The well-known Malgrange-Ehrenpreis theorem (cf. for Example

[190], p. 212) asserts that there exists a fundamental solution to the operator
P (D) which belongs to D′ . The solution to equation (3.9) exists and can
be expressed by the formula X = E ∗ G. To find the S-asymptotics of X,
we have only to apply Theorem 1.2 b).
Remark. The solution X = E ∗ G is unique in D′ (cf. [190], p. 87).
3.2 Quasi-asymptotics of solutions
Chapter IV in [192] is devoted to asymptotic properties of solutions to

systems of convolution equations. We shall quote only two theorems.
Let Γ be a closed convex and regular cone, and Γ∗ be its dual cone. Let
{Uk ; k ∈ I} be a family of linear nonsingular transforms of Rn which leaves
the cone Γ invariant. Denote by UkT the transposed operator for Uk and
Vk = (UkT )−1 . Then, Vk Γ∗ = Γ∗ , k ∈ I. Let Jk = det Uk > 0. We denote by
L(f )(z), z ∈ C = int Γ∗ , the Laplace transform of f ∈ S ′ (Γ). (We refer for
the definition and the properties of the Laplace transform [192], Chapter I,
§2).
Theorem 3.3. ([192]) Suppose that K ∈ S ′ (Γ) has the quasi-asymptotics

in the cone Γ over the family {Uk ; k ∈ I} related to ρk with the limit K0
and that LK(z) has a bounded argument for z ∈ Rn + iC. If f ∈ S ′ (Γ)
has the quasi-asymptotics in Γ over {Uk ; k ∈ I} related to rk and with the
limit g0 , then the equation K ∗ u = f has a solution in S ′ (Γ) which has
the quasi-asymptotics in Γ over {Uk ; k ∈ I} related to (Jk ρk )−1 rk with the
limit u0 which satisfies K0 ∗ u0 = g0 .
Theorem 3.4. ([192]) Let P (−iz) be a polynomial which is different from

zero for z ∈ Rn + iC, C = int Γ∗ . Let Ω ⊂ C be a domain. If there exists
ρk such that
Jk ρk P (Vk y) → h(iy), k → ∞, k ∈ I, y ∈ Ω ,
then the fundamental solution E of the hyperbolic operator P (D) exists and
has the quasi-asymptotics in Γ over {Uk ; k ∈ I} related to ρk and with the
1
limit E0 , where E0 has the property that L(E0 )(z) = , z ∈ Rn + iC,
h(z)
and h(z) has a bounded argument Rn + iC.
Now, we will consider in the one-dimensional case applications of the

B-transform to equation
((xu)′′ + ru) ∗ h = g, r ≤ 0, h, g ∈ S ′ + (3.10)
(cf. [136]). First we give the definition of the B-transform.

We need an equivalent definition of S ′ + .
Let ln = e−x/2 Ln (x), x > 0, n ∈ N0 , be the Laguerre orthonormal
system in L2 (R+ ), where
X n
n (−x)m
Ln = , x > 0, n ∈ N0 ,
m=0
n−m m!
are the Laguerre polynomials, and ln are the eigenfunctions of the operator
R = ex/2 Dxe−x Dex/2 , for which R(ln ) = −nln , n ∈ N0 .
Then S+ , the space of functions φ ∈ C ∞ ([0, ∞)), |xk ϕ(ℓ) (x)|< ∞,
k, ℓ ∈ N0 can be defined also through the finite norms
Z∞
kφkk = ( |Rk φ(k)|2 dx)1/2 , k ∈ N0
0
and the following property:

(Rk φ, ln ) = (φ, Rk ln ), k, n ∈ N0 , (Rk+1 = R(Rk )) .
(cf. [43] and References in [136]).

P
∞
Then b-transform on S+ is defined as follows: If φ = an ln ∈ S+ ,
n=0
then
p √
b[φ](t) = φ(0) + 1/2hφ(τ ), t/τ J0′ ( tτ )i
∞
X ∞
X
= −2 (−1)n (2 ai + an )ln (t), t > 0.
n=0 i=n+1
J0 is the Bessel function:

∞
X 2j
iz 1
J0 (z) = .
j=0
2 (j! )2
The B-transform on S ′ + is defined by dualizing the b-transform on S+ :
hB[f ], φi = hf, b[φ], i, φ ∈ S+ .

P
∞
Thus, for f = b n ln ,
n=0
∞
X n−1
X
B[f ] = [2 (−1)m bm + (−1)n bn ]ln .
n=0 m=0
Recall that B[fα ] = 4α f−α , α ∈ R, where {fα ; α ∈ R} is given in 0.4.

We shall use the identity
B[f (εx)](t) = k 2 B[f (x)](kt), t > 0, ε = 1/k, k > 0.
Proposition 3.4. Let f ∈ S ′ + , and let L(ε), ε ∈ (0, ε0 ), be slowly varying

at 0+ . Then the following condition are equivalent:
1. f has the quasi-asymptotics at 0+ (at ∞) related to εσ L(ε) (to
σ
k L(1/k)).
2. Bf has the quasi-asymptotics at ∞ (at 0+ ) related to k −σ−2 L(1/k)
(to ε−σ−2 L(ε)).
Proof. Since B : S ′ + → S ′ + is an isomorphism (cf. [136]), we have to

prove only the part of this assertion which corresponds to 0+ .
1. ⇒ 2. Let φ ∈ S+ . Then, with ε = 1/k, k → ∞,

(Bf )(kt) B(f (x/k))(t) f (x/k)
, φ(t) = , φ(t) = , b[φ](x)
k −σ−2 L(1/k) k −σ−2+2 L(1/k) k −σ L(1/k)

f (εx)
= , b[φ](x) → hCfσ+1 (x), b[φ](x)i = hCB(fσ+1 ), φi (3.11)
εσ L(ε)
= hC̃f−σ−1 , φi, where C̃ = C4σ+1 .
2. ⇒ 1. For given φ ∈ S+ let φ = b[ψ], ψ ∈ S+ . From b[φ] = b[b[ψ]] = ψ

(cf. [136]) and (3.11), we have

f (x/k) (Bf )(kt)
, φ(t) = , b[b[ψ]](x) → hCf−σ−2+1 , ψi, k → ∞ ,
k σ+2 L(k) k σ L(k)
and this implies the assertion.
Now, we shall use the B-transform and the quasi-asymptotics for the
analysis of (3.10).
Since
B[xuxx + 2ux](t) = (−t/4)B[u](t), t > 0, (3.12)
it is equivalent to the equation
(−t/4 + r)ũ = g̃, where g̃ = B[g] ∈ S ′ + , and ũ = B[u]. (3.13)
Let us remark that the equation xp = q, q ∈ S ′ + has solutions in S ′ +

uniquely determined up to Cδ, C ∈ C, and that the equation
(x + r)p = q, r > 0, q ∈ S ′ + ,
has the unique solution in S ′ + .

We need the following assertion
Proposition 3.5. (i) Let r = 0 in (3.10) and let g have the quasi-
asymptotics at 0+ related to εσ L(ε), where σ 6= −2, −1, 0, 1, . . . Then:
1. If σ < −2, then the solution u has the quasi-asymptotics at 0+ related
to εσ+1 L(ε);
2. If σ > −2, then there exist numbers bj , j = 0, 1, . . . , p such that
Pp
u+ bj fj has the quasi-asymptotics related to εσ+1 L(ε);
j=0
(ii) Let r < 0 in (3.12) and let g have the quasi-asymptotics at 0+

related to εσ L(ε), σ ∈ R. Then the solution u to equation (3.12) has the
quasi-asymptotics at 0+ related to εσ+1 L(ε) if σ < −2. If σ > −2 and σ 6=
P
r
−1, 0, 1, . . . , then there exist aj ∈ C, j = 0, . . . , r such that u + aj (xj )+
j=0
has the quasi-asymptotics at 0+ with respect to εσ+1 L(ε).
Proof. (i) 1. Since g̃ has the quasi-asymptotics at ∞ related to

k −σ−2 L(1/k) and σ < −2, by the first part of Lemma 2.1, it follows
that ũ has the quasi-asymptotics at ∞ related to k −σ−3 L(1/k), because
−σ − 3 > −1. Proposition 3.4 implies that u has the quasi-asymptotics at
0+ related to εσ+1 L(ε).
2. If σ > −2 then there exist p ∈ N and numbers aj , j =
p
P
0, 1, . . . , p, such that ũ + aj δ (j) has the quasi-asymptotics at ∞ related
j=0
to k −σ−3 L(1/k). This follows from the second part of Lemma 2.1.
p
P
Thus u + bj xj−1 −j
+ /Γ(j), where bj = 4 aj , has the quasi-asymptotics
j=0
at 0+ related to εσ+1 L(ε) because B[δ (j) ] = 4−j fj , j ∈ N.
(ii) Let d > 0 and f ∈ S ′ + . We will prove that if

(kt + d)f (kt) (kt)f (kt)
α
→ C in S ′ (R) (k → ∞), then α → C in S ′ (R)
k L(k) k L(k)
(k → ∞). This follows from
f (kt)
→ 0 in S ′ (R), k → ∞.
k α L(k)
Let us prove this fact.
Our assumption implies that there exist s ∈ N and a continuous function
F, suppF ⊂ [0, ∞), such that
F (t)
(t + d)f (t) = F (s) (t) and → C as t → ∞ .
tα+s L(t)
Since
s (i) (s−i)
F (s) (t) X i 1
f (t) = = (−1) F (t)
t+d i=0
t+d
and
F (t) (s−i)
(t+d)1+i
(kt) → 0 in S ′ (R) as k → ∞, i = 0, . . . , s,
k α L(k)
f (kt)
we have → 0 in S ′ (R) as k → ∞.
k α L(k)
Thus (3.13) implies that (t/4−r)ũ and thus tũ has the quasi-asymptotics
k −σ−2 L(1/k). By Lemma 2.1, we have
ũ(kt)
→ C in S ′ (R) as k → ∞ if − σ − 3 > −1,
k −σ−3 L(1/k)
P
r
(ũ + aj δ j )(kt)
j=1
→ C in S ′ (R) as k → ∞ if − σ − 3 < −1.
k −σ−3 L(1/k)
Hence, by Proposition 3.4, we have that the solution u has the quasi-
Pr
asymptotics at 0+ with respect to εσ+1 L(ε) if σ < −2 or u + aj (xj )+
j=0
has the quasi-asymptotics at 0+ with respect to εσ+1 L(ε) if σ > −2 and
σ 6= −1, 0, 1, . . .
The main part of previous examinations will be used for the qualitative
analysis of equation (3.10). For example, the equation
j
X j
X
′′ (k)
ak ((xu) + ru) = g (with h = ak δ (k) )
k=0 k=0
is of this form.
Proposition 3.6. Assume that g, h ∈ S ′ + and that the Laplace trans-

form of h, (Lh)(x + iy), x ∈ R, y > 0, has a bounded argument. Let h
have the quasi-asymptotics at 0+ related to εσ1 L1 (ε) and g have the quasi-
asymptotics at 0+ related to εσ2 L2 (ε).
1. Let r = 0 in (3.10). If σ1 − σ2 > 1, then (3.10) has the solution u
with the quasi-asymptotics at 0+ related to εσ1 −σ2 L1 (ε).
If σ1 −σ2 < 1 and σ1 −σ2 6= 1, 0, −1, −2, . . . , then there are numbers bj ∈
Pp
C, j = 0, 1, . . . , p, such that (3.10) has the solution u so that u + bj fj
j=0
has the quasi-asymptotics at 0+ related to εσ2 −σ1 L2 (ε)/L1 (ε).
2. Let r < 0 in (3.10). Then (3.10) has the solution u with the
quasi-asymptotics at 0+ related to εσ2 −σ1 L2 (ε)/L1 (ε) if σ2 − σ1 < −2.
If σ2 − σ1 > −2 and σ2 − σ1 6= −1, 0, 1, . . . , then there exist aj ∈ C, j =
p
P
0, . . . , p such that u + aj (xj )+ has the quasi-asymptotics at 0+ related
j=0
to εσ2 −σ1 L2 (ε)/L1 (ε).
Proof. First, we prove: If the Laplace transform of h has a bounded

argument, then the same holds for B[h]. Namely, by
ti
b[eizτ ](t) = e− 4z , t > 0, Im z > 0 ([192], Chapter 1, §2.3),
we have

1
L(B[f ])(z) = h(B[f ](τ ), eiτ z i = hf (t), b[eizτ ]i = Lf − , Im z > 0 ,
4z
which implies the assertion.
By (3.13) and B[f ∗ g] = B[f ] ∗ B[g], (see [136]), (3.10) becomes
(−t/4 + r)ũ ∗ h̃ = g̃.
We shall prove only part 1 of Proposition 3.6. Let r = 0. We shall use
Theorem 1, Chapter IV, §1.1 in [192]. In the one-dimensional case this
theorem reads as follows.
“Let K ∈ S ′ + , has the quasi-asymptotics at ∞ related to k α L1 (k) with

the limit C1 fα+1 , C1 6= 0, and f ∈ S ′ + has the quasi-asymptotics at ∞
related to k β L2 (k) with the limit C2 fβ+1 , C2 6= 0. Let the Laplace transform
of K, LK(x+iy), has a bounded argument in R+iR+. Then the convolution
equation K ∗u = f has the solution u ∈ S ′ + which has the quasi-asymptotics
at ∞ with respect to k β−α−1 L2 (k)/L1 (k) with the limit (C2 /C1 )fβ−α ” (for
fβ−α see 0.4).
Now, we can continue with the proof of Proposition 3.6. Since the
Laplace transform of h has a bounded argument, it follows that there exists
s̃ ∈ S ′ + such that s̃ ∗ h̃ = g̃, and
s̃(kx)
→ const.fσ2 −σ1 , k → ∞ in S ′ + ,
k −σ2 −2−(−σ1 −2)−1 L2 (1/k)/L1 (1/k)
because g̃ has the quasi-asymptotics related to k −σ2 −2 L2 (1/k) and h̃ has
the quasi-asymptotics related to k σ1 −2 L1 (1/k).
Let u be a solution to
xu′′ + 2u′ = s ⇐⇒ (−t/4)ũ = s̃ .
As in the proof of Proposition 3.5, we have the following situations:
1. Since s̃ has the quasi-asymptotics at ∞ related to
k −(σ2 −σ1 −1)−2 L2 (1/k)/L1 (1/k) ,
if σ2 − σ1 < −1 it follows that ũ has the quasi-asymptotics at ∞ related to
k −(σ2 −σ1 −1)−3 L2 (1/k)/L1 (1/k)
and by Proposition 3.4, h has the quasi-asymptotics at 0+ related to
εσ2 −σ1 L2 (ε)/L1 (ε) .
2. If σ2 − σ1 > −1 and σ1 − σ2 6= 1, 0, −1, −2, . . . , then as in the proof

of Proposition 3.5, we conclude that there are numbers aj , j = 0, . . . , p,
p
P
such that ũ + aj δ j has the quasi-asymptotics at ∞ related to
j=0
k (σ2 −σ1 −1)−3 L2 (1/k)/L1(1/k)

which implies that
p
X
u+ bj xj−1 j
+ /Γ(j), (bj = 4 aj , j = 0, . . . , p),
j=0
has the quasi-asymptotics at 0+ related to εσ2 −σ1 L2 (ε)/L1 (ε).

As an application to hyperbolic problems, we shall quote the results of
P. Wagner [194] in which explicit formulas for the open quasi-asymptotic
expansion of the causal fundamental solution of a class of hyperbolic equa-
tions are given.
Let λ0 < λ1 < . . . be a sequence of real numbers and Tj be distributions
in S ′ (Γ), homogeneous of degree −λj . Recall,
∞
X
E(kx) ∼ k −λj Tj (x), k → ∞
j=0
means that for every N ∈ N0

N
X
λN −λj
lim k E(kx) − k Tj (x) = 0 in S ′ (Γ) .
k→∞
j=0
(cf. Definition 2.7).

A linear partial differential operator (with constant coefficients) P (∂) =
P
m
Pj (∂), Pj homogeneous of degree j, Pℓ 6≡ 0, Pm 6≡ 0, is called hy-
j=ℓ
perbolic with respect to the direction N ∈ Rn \ 0 if Pm (N ) 6= 0 and
P (−ix + τ N ) 6= 0 for all x ∈ Rn and τ > τ0 . In this case τ0 = 0.
P
m
Proposition 3.7. ([194]) Let P (∂) = Pj (∂) and Γ be as above. Denote
j=ℓ
the causal fundamental solutions (cf. [70]) of P (∂) and of Pℓ (∂)i by E and
El,i respectively. Then E has the following quasi-asymptotic expansion in
S ′ (Γ) for k → ∞ :
∞
X
E∼ k l−n−r Tr ,
r=0
where Tr consists of homogeneous distributions of degree l − n − r given by

X m−1
Y
|α|
Tr = (−1)|α| Pℓ+j (∂)αj El,1+|α| .
α j=1
α1 +2α2 +···+(m−ℓ)αm−1 =r
αi ≥0
Let Z(∂) be a system of N linear partial differential equations with

constant coefficients containing N unknown functions, i.e., Z(∂) is an N ×
N -matrix of operators. Assume that P (∂) := det Z(∂) has the property
that P (−iz) 6= 0 for z ∈ Rn + iC. The unique (again called causal) two-
sided fundamental solution EZ of Z(∂) having its support in Γ := C ∗ is
given by the formula EZ = Z adj (∂)E, wherein Z adj denotes the adjoint
matrix to Z and E is the causal fundamental solution of P (∂). Decompose
Z adj into its homogeneous components:
m1
X
adj adj
Z = Zsadj , where Zs,ij are homogeneous of degree s, 1 ≤ i, j ≤ N .
s=l1
P
m
Proposition 3.8. ([194]) Let Z(∂) = Zs (∂), where Zs,ij are homo-
s=0
geneous of degree s, and EZ has the property that det Z(−iz) 6= 0, z ∈
Rn + iC. Assume that det Z0 = det Z(0) 6= 0. Then the quasi-asymptotic
expansion of EZ (kx) in S ′ (Γ)N ×N for k → ∞ is given by the formula
∞
X X
EZ (kx) ∼ k −n−r (−1)j Z0−1 Zr1 (∂)Z0−1 . . . Z0−1 Zrj (∂)Z0−1 δ .
r=0 r1 +···+rj =r
j≥0,rj ≥0
3.3 S-asymptotics of solutions to equations with

ultra-differential or local operators
We refer to [132] for the material of this section.

We shall cite the following theorem which is a simple consequence
of Theorem 1.2e) in order to illustrate possible applications of the S-
asymptotics to
∗
P (D)X = G, G ∈ D′ , (3.14)
where P (D) is an ultradifferential operator of (∗)-class.
Theorem 3.5. A necessary condition that (3.14) has a solution χ0 , with

s ∗ s
the property χ0 ∼ c(h) · U, h ∈ Γ in D′ is that G ∼ c(h) · P (D)U, h ∈ Γ
∗
in D′ .
The proof will be given with the proof of Theorem 3.6.
∗ ∗
Theorem 3.6. Let Mp satisfy (M.1), (M.2) and (M.3) and G ∈ O′ C (O′ C
′
is the space of convolutors of S ∗ ).
∗
Suppose that there exists a solution E ∈ S ′ to equation P (D)Y = δ
such that:
∗
a) the set {E(x + h)/(1 + khk2 )q/2 ; h ∈ Rn }, q ∈ R is bounded in D′ ;
s ∗
b) E ∼ (1 + khk2 )q/2 · U, h ∈ Γ in D′ .
∗
Then, there exists a solution χ0 ∈ S ′ of the equation P (D)χ = G such
s ∗
that χ0 ∼ (1 + khk2 )q/2 · (U ∗ G), h ∈ Γ in D′ .
Proof. If we apply Theorem 1.10 to E with the property a), or if we

use Theorem 1 in [123], we obtain that E = Q(D)F1 + F2 , where Q(D)
is an ultradifferential operator of class ∗ and Fi , i = 1, 2, are continuous
functions such that
|Fi (x)/(1 + kxk2 )q/2 |≤ Mi , x ∈ Rn , i = 1, 2 .
Since for every φ ∈ S ∗ , Q(−D)φ ∈ S ∗ , it follows that {E(x + h)/(1 +

∗
khk2 )q/2 ; h ∈ Γ} is bounded in S ′ .
Since D∗ is dense in S ∗ , by the Banach–Steinhaus theorem
D E
E(x + h)
lim , φ(x) = U, φ , φ ∈ S∗. (3.15)
h∈Γ,khk→∞ (1 + khk2 )q/2
Clearly, χ0 = E ∗ G is a solution of equation P (D)χ = G. We have to

prove that
lim h(E ∗ G)(x + h)/(1 + khk2 )q/2 , φ(x)i = hU ∗ G, φi, φ ∈ D∗ .
h∈Γ,khk→∞
This follows from (3.15) and

h(E ∗ G)(x + h), ϕ(x)i = hE(x + h), (Ǧ ∗ ϕ)(x)i, Ǧ(x) = G(−x), x ∈ Rn ,
∗
because G ∈ O′ C is a convolutor on S ∗′ .
Let r0 > 0. Denote by Sr0 ,h , h > 0, the space of smooth functions φ on
Rn such that
(1 + kxk)r0 kφ(α) (x)kL∞
sup < ∞.
x∈Rn ,|α|∈N0 h|α| M|α|
Define
Sr(M
0
p)
= proj lim Sr0 ,h , Sr{M
0
p}
= ind lim Sr0 ,h .
h→0 h→∞
Clearly, D∗ is dense in Sr∗0 . We shall use the definition of the convolution
of two ultradistributions as given in [79]: hf ∗ g, ϕi = hg, fˇ ∗ ϕi, whenever
this defines a continuous linear functional in ϕ.
Theorem 3.7. Suppose:

∗
1. g ∈ S ′ r0 , r0 > 0, and
lim hg(x + h)/c(h), φi = hu, φi, φ ∈ Sr∗0 . (3.16)
h∈Γ,khk→∞
P
2. For the differential operator L(D) = aα Dα , aα ∈ R, there
|α|≤m
exists a fundamental solution f ∈ S ′ of the form
Xm
f (x) = (Fβ (x)/(1 + kxk)rβ )(β) , rβ ≥ r0 , Fβ ∈ L∞ , β = 0, . . . , m .
β=0
Then, the partial differential equation L(D)X = g has a solution X =

s ∗
f ∗ g and X ∼ c(h) · f ∗ u, h ∈ Γ in D′ .
Proof. One can easily prove that for F ∈ L∞ , φ ∈ D∗ and r ≥ r0 , the

functions:
x 7→ ((F (t)/(1 + ktk)r ) ∗ φ(β) (t))(x), β = 0, . . . , m,
belong to Sr∗0 . This, in accordance with (3.16), implies

(f ∗ g)(x + h)
lim , φ(x) = hu, fˇ ∗ φi = hu ∗ f, φi, φ ∈ D∗ .
Since f ∗ g = X is a solution of the equation L(D)X = g, the proof is
completed.
Similar theorems can be proved for hyperfunctions. We quote only the
following.
Proposition 3.9. Let P be a local operator. Then a necessary condition

that a solution to equation
P (D)x = f
has the S-asymptotics related to c and to the cone Γ with the limit u, is
s
that f (t + h) ∼ c(h)u(t), h ∈ Γ.
Proof. Since Q(Dn ) is a Montel space, from

w.lim x(t + h)/c(h) = u(t) in Q(Dn )
h→∞,h∈Γ
it follows that the same holds for the strong convergence. Since a local
operator maps continuously Q(Dn ) into Q(Dn ), it follows
s
P (D)x(t + h) ∼ c(h)P (D)u(t), h ∈ Γ ,
and thus
s
f (t + h) ∼ c(h)P (D)u(t), h ∈ Γ.
4 Asymptotics and integral transforms
Integral transforms of distributions have been elaborated in the last forty

years, or so and they appeared as one of powerful tools, especially in math-
ematical physics. The book by Zemanian [202] was the first systematic
monograph which gave different integral transforms of generalized func-
tions. Brychkov and Prudnikov [19] collected results on the most impor-
tant integral transforms of generalized functions. The results of Abelian
and Tauberian type have been elaborated only for some special integral
transforms of distributions.
We shall apply two defined asymptotics to integral transforms with gen-
eral kernels to obtain results which contain those proved for special kernels.
It turned out that the S-asymptotics is well appropriate to integral trans-
forms of convolution type and the quasi-asymptotics to integral transforms
of the Melline convolution type.
Integral transforms of generalized functions can be defined in various
ways. In the direct approach, one constructs a basic space A ⊂ D(R) of
smooth functions to which belongs the kernel of the integral transform,
K(s, ·), s ∈ S0 ⊂ C. Then, for T belonging to the dual space A′ , the
transform K is defined by the expression: K(T )(s) = hT (t), K(s, t)i, s ∈ S0 .
We shall prove the theorems of Abelian type for the integral transforms
defined in this way. The book by Estrada and Kanwal [56] gives another
approach to this matter.
We refer to [129], [130], [133], [134], [135], [158] and [192] for the material
on Abelian and Tauberian type theorems.
4.1 Abelian type theorems
4.1.1 Transforms with general kernels
Let L be a slowly varying function (see 0.3). One can introduce the function
L∗ ∈ C ∞ (R+ ) in the following way: Let w ∈ C ∞ , supp w ⊂ [−1, 1], w > 0
and
Z1 Z
w(t)dt = 1; L∗ (x) = w(x − t)L(exp t)dt, x > 0 .
−1 R
4. Asymptotics and integral transforms 189
Then, L∗ has the following properties:

1, k = 0
L∗(k) (x + h)/L(exp h) → h → ∞, x > 0 . (4.1)
0, k ∈ N
For every k ∈ N0 there are Ck > 0 and Ck∗ > 0 such that
|L∗(k) (x)|≤ Ck L(exp x) ≤ Ck∗ L∗ (x), x > 0 . (4.2)
By (4.2), we have that for every k ∈ N0 there exists C ∗ > 0 such that
(k)
1 ∗
≤ C , x > 0. (4.3)
L∗ (x) L∗ (x)
We shall use the following functions:

exp(ax)L(exp x), x ≥ 0
c(x) = , a > b, (4.4)
exp(bx), x<0

exp(ax)L∗ (x), x > 1
e(x) = , a > b, (4.5)
exp(bx), x<0
where e is extended on [0, 1] to be smooth and positive on R and

exp((a + δ)x), x > 1
eδ (x) = , a > b, δ > 0 . (4.6)
exp(bx), x<0
Function eδ is also extended on [0, 1] to be smooth and to satisfy the in-
(k)
equality eδ (x) ≥ e(k) (x), x ∈ R, k ∈ N0 .
We need the following two properties of the function e :
For every compact set B ⊂ R and every k ∈ N0 there is CB,k > 0 such
that
(k)
c(h)
(x) ≤ CB,k , h ∈ R, x ∈ B . (4.7)
e(x + h)
For every k ∈ N0 and δ > 0 there is a Ck > 0 such that for h > 0

(k) Ck exp(ax + δ|x|), x + h ≥ 0 ,
|(e(x + h)/c(h) − exp(ax)) |≤ (4.8)
Ck exp(bx), x+h < 0.
Note, (4.3) implies that for every k ∈ N0 there exists ck > 0 such that

 ck L(et )eat , t > 1,
(k)
|e (x)|≤ ck , 0 ≤ t ≤ 1, (4.9)

ck ebt , t < 0.
Thus (4.7) follows.
Let us prove (4.8). If x + h > 1, by Leibnitz’s formula, it is enough to

prove that for every j ∈ N0 there exists cj > 0 such that
(j)
e(· + h)
(x) ≤ cj exp(δ|x|) for any δ > 0. (4.10)
L(exp h) exp(a(· + h))
This easily follows from (4.9) and (4.3). Then it implies (4.8).
Let x + h ∈ [0, 1]. We shall prove (4.10) because it implies (4.8). Since
e(x + h)
, x + h ∈ [0, 1],
exp(a(x + h))
is bounded, (4.10) follows from the estimate
1 1
= ≤ eδ|x| , x + h = t ∈ [0, 1] for any δ > 0,
L(exp h) L(exp(x − t))
which is a consequence of (4.3).
If x + h < 0, then (4.8) follows from (4.9) and the assumption a > b.
′
We shall introduce the space DL 1 ,e (R)weighted distributions in a sim-
ilar way as it has been done by Ortner and Wagner [103] for the space
DL1 ,e (R) = {ψ ∈ C ∞ (R); eψ ∈ DL1 (R)}.
The function f : f (e, ψ) = eψ, which maps {e} × DL1 ,e (R) onto DL1 (R)
is continuous. The same holds for g : e−1 × (DL1 ,e )′ (R), onto B ′ (R), given
by g(e−1 , S) = e−1 S.
A topology in DL1 ,e (R) is defined by means of the isomorphism
DL1 ,e (R) → DL1 (R) : ψ → eψ, through seminorms
Z
(k)
kψkp = max |(eψ) (t)|dt, 0 ≤ k ≤ p , p ∈ N0 .
R
By the properties of e and eδ , we have DL1 ,eδ (R) ⊂ DL1 ,e (R). The topology
of DL1 ,eδ (R) is finer than the topology induced from DL1 ,e (R). The strong
dual space (DL1 ,e )′ (R) = {S ∈ D′ (R); e−1 S ∈ B ′ (R)} and (DL1 ,e )′ (R) ⊂
(DL1 ,eδ )′ (R), δ > 0 (see 0.5 for B ′ (R)).
Proposition 4.1. A necessary and sufficient condition that T belongs to

(DL1 ,e )′ (R) is the existence of continuous and bounded functions Fi , i =
0, . . . , p such that
Xp
T = Dj (Fj e) ,
j=0
where D is the derivative in the sense of distributions.

Proof. Suppose that T is given by the sum of derivatives of (Fj e), then
Xp Xp
−1 (k) j
T =e e e Dj−k Fj = eS .
k
k=0 j=k
−1 (k) ∞
Since e e ∈ C (R), 0 ≤ k ≤ p, and by (4.2) these functions are
bounded, then by the structural theorem for bounded distributions, it fol-
lows that S is a bounded distribution. Consequently, T ∈ (DL1 ,e )′ .
Suppose now that T ∈ (DL1 ,e )′ , then T is of the form
p
X p X
X j
j
T =e D Ej = aj,k Dk (e−1 e(j−k) eEj ), aj,k ∈ R,
j=0 j=0 k=0
where Ej , j = 0, . . . , p are continuous and bounded functions. Since

e(j−k) e−1 , j = 0, . . . , p; k = 0, . . . , j, are smooth and bounded functions,
T has the required form.
s
Proposition 4.2. Suppose that T ∈ D′ (R) and T (x+ h) ∼ c(h)·Aeax , h ∈
R+ , in D′ (R). If T ∈ (DL1 ,e )′ (R), then T0 = e−1 T ∈ B ′ (R) has the prop-
s
erty T0 (x + h) ∼ 1 · A, h > 0 in D′ (R) (and in B ′ (R) as well).
Proof. For h large enough and ϕ ∈ D(R)

T (x + h) c(h)
hT0 (x + h), ϕ(x)i = , ϕ(x)
c(h) e(x + h)

T (x + h) L(exp h)
= , ∗ exp(−ax)ϕ(x) .
c(h) L (x + h)
By relations (4.1) and (4.2) the function x 7→ ϕ(x) exp(−ax)L(exp h)/
L∗ (x + h) converges to x 7→ ϕ(x) exp(−ax) in D(R) when x → ∞. Now,
we can use Theorem XI Chapter III in [146] to obtain
lim hT0 (x + h), ϕ(x)i = hA, ϕi for every ϕ ∈ D(R) .
h→∞
The last is valid for ϕ ∈ DL1 (R), as well. This follows from Theorem 1.17.

Definition 4.1. Suppose that T ∈ (DL1 ,e )(R) and K(s + x) ∈ DL1 ,eδ (R)
for s ∈ S0 ⊂ C, and δ > 0, where S0 − R+ ⊂ S0 (i.e., for every s ∈ S0 and
x > 0, S0 − x ∈ S0 .). Then the K-transform of T is defined by
K(T )(s) = hT (x), K(s + x)i = hT (x)e−1 (x), e(x)K(s + x)i , (4.11)
for s ∈ S0 .

1. T ∈ (DL1 ,e )′ (R) and K(s + x) ∈ DL1 ,eδ (R), s ∈ S0 , S0 − R+ ⊂ S0 ,
for a δ > 0.
s
2. T (x + h) ∼ Ac(h)eax , h ∈ R+ in D′ (R), where c is given by (4.4).
Then for the K-transform of T , we have
lim K(T )(s − h)/c(h) = hA, K(s + x) exp(ax)i, s ∈ S0 . (4.12)

h→∞
Proof. Note K(s + x) exp(ax) ∈ DL1 (R), s ∈ S0 because of a > b. By

Proposition 4.2, if s ∈ S0 , then
lim hT (x + h)/e(x + h), K(s + x) exp(ax)i = hA, K(s + x) exp(ax)i .

h→∞
To prove relation (4.12), it is enough to show

T (x + h) e(x + h)
lim , − exp(ax) K(s + x) = 0, s ∈ S0 . (4.13)
h→∞ e(x + h) c(h)
For every h > 0 the distribution T (· + h)/e(· + h) is in B ′ (R) and thus

T /e can be given in the form (cf. Theorem 1.22)
2
X
T /e = D2ik Fi ,
i=0
where Fi , i = 0, 1, 2 are bounded and continuous functions. The left side

in (4.13) becomes,
2
X Z (2ik)
e(x + h)
lim (−1)2ik Fi (x + h) − exp(ax) K(s + x) dx .
h→∞
i=0
c(h)
R
(4.14)
If we prove that for every j ∈ N0 and s ∈ S0 the function
x 7→ |((e(x + h)/c(h) − exp(ax))K(s + x))(j) |, x ∈ R,
is bounded by a function integrable on R which does not depend on h, then

by the use Lebesgue’s theorem in (4.14) and by (4.1), it follows (4.13).
Let x > 0, s ∈ S0 and h > 0. By (4.8), we have for s ∈ S0
(i)
e(x + h)
− exp(ax) K (j)
(x + s) ≤ Ci |K (j) (x + s)|exp((a + δ)x) .
c(h)
Let x < 0 and h > 0. Then (4.8) gives

(i)
e(x + h)
− exp(ax) K (x + s)
(j)
c(h)
≤ Ci (exp(ax + δ|x|) + exp(bx))|K (j) (x + s)|
≤ 2Ci |K (j) (x + s)|exp(bx), s ∈ S0 .

If K(· + s) ∈ DL1 (R), s ∈ S0 , we can not apply Theorem 4.1 because
of supposition that a > b and a0 > 0. In this case, we have

1. T ∈ B ′ (R) and K(x + s) ∈ DL1 (R), s ∈ S0 ;
s
2. T (· + h) ∼ 1 · A, h ∈ R+ in D′ (R).
Then for the K-transform of T, K(T )(s) = hT (x), K(x + s))i, s ∈ S0 , we
have
lim K(T )(s − h) = hA, K(x + s)i .

h→∞
Proof. We have only to apply assertions a) and d) in Theorem 1.19, and

to use K(T )(s − h) = hT (x + h), K(x + s)i.
Here arises a natural question: If (1+x2 )k T (x) ∈ B ′ (R) for some k > 0,
how does the asymptotic behavior of K(T )(−h), h → ∞, depend on k? The
following example shows that this dependence is very limited.
The following formula is known
Z∞
(is + t)−r exp(−t)dt = (is)−r/2 exp(is/2)W−r/2,(1−r)/2(is)
0
∼ (is)−r , s → ∞, s ∈ R+ , r > 1 ,
where Wp,q is the Whittaker function. We have t 7→ H(t)(1 +

t2 )k exp(−t), t ∈ R defines a distribution which belongs to B ′ for any
k > 0, but the dependence on k in the asymptotic behavior of the K-
transform of H(t)e−tx , t ∈ R, with K(s + t) = (is + t)−r is limited by r (H
is the Heaviside function).
We shall give several theorems which correspond to Theorem 4.1 in

D (R+ ) or in S ′ + . By the use of the Mellin convolution type transforms
′
(see Zemanian [202]).

The function x → log x is a C ∞ -diffeomorphism of R+ onto R. Its
inverse is the function: y 7→ exp y. To an f ∈ D′ (R), we assign f (log x) ∈
D′ (R+ ) by the relation
hf (log x), ϕ(x)i = hf (y), ϕ(exp y) exp yi, ϕ ∈ D(R+ )
(see Chapter 1, Section 1 in [192]) or,
hf (log x), ψ(log x)/xi = hf (y), ψ(y)i, ψ ∈ D(R)
and
hf (log xu)/c(log u), ψ(log x)/xi = hf (y + h)/c(h), ψ(y)i, (4.15)
where u = exp h. The mapping: ψ(y) → ψ(log x)/x is an isomorphism
of the space D(R) onto D(R+ ). Note, D(R+ ) ⊂ S+ and the topology of
D(R+ ) is finer then the topology induced by S+ , D(R+ ) is not dense in
′
S+ . Thus, we have to consider the restrictions of f ∈ S+ onto D(R+ ). In
this sense S ′ + ⊂ D′ (R+ ) (for the space S+ and S ′ + see 0.5.1).
Definition 4.2. Ea+δ,b (R+ ) is the vector space of all smooth functions w
R∞
on (0, ∞)such that for each k ∈ N0 , αk (w) = qδ (x)xk−1 |w(k) (x)|dx < ∞,
0
where by qδ is denoted a positive and smooth function such that
−(a+δ)
x , 0 < x < e−1 , δ > 0
qδ (x) = −b , a > b.
x , x>1
Proposition 4.3. If w ∈ Ea+δ,b (R+ ), then w(exp(−x)) ∈ DL1 ,eδ (R).
Proof. For t > 1 and m ∈ N0

m
X
(eδ (t)w(exp(−t)))(m) = ak,m exp((a + δ − 1)t)w(k) (exp(−t)) ,
k=0
where ak,m are constants. Hence,

Z∞ m
X Ze−1
(m)
(eδ (t)w(exp(−t))) dt = ak,m qδ (x)xk−1 w(k) (x)dx .
1 k=0 0
The proof for t < 0 is similar.

Theorem 4.3. Suppose that:

1.{F (u·)/c(log u); u ∈ R+ } is a bounded set in D′ (R+ ), where c is
given by (4.4).
2. Let a > b and V0 ⊂ C be such that R+ V0 ⊂ V0 (R+ V0 = {xV ; x ∈
R+ , v ∈ V0 }). Let k(·/v) ∈ Ea+δ,b (R+ ) for every v ∈ V0 and for a δ > 0. If
for every ψ ∈ D(R+ )
lim hF (xu)/c(log u), ϕ(x)i = hAxa , ϕ(x)i, (4.16)
h→∞
then there exists a Mellin convolution type transform Mk (F ) (see Zemanian

[202])
Mk (F )(v) = (F ∗ k)(v) = hF (x), k(1/vx)/xi, v ∈ V0
M
= hF (x)/e(log x), e(log x)k(1/xv)/xi (4.17)

and
lim Mk (F )(vu)/c(− log u) = hA, xa−1 k(1/xv)i, v ∈ V0 .
u→0+
Proof. We shall prove first that the distribution F (exp ·) belongs to

(DL1 ,e )′ (R). We have

F (ux) F (exp(t + h))
, ψ(x) = , ψ(exp t) exp t , u > 0, h ∈ R, ψ ∈ D+ .
c(log u) c(h)
By the isomorphism D(R+ ) onto D(R) : ψ(x) → ψ(exp t) · exp t, and by as-
sumption 1. of Theorem 4.3 it follows that the set {F (exp(· + h))/c(h); h ∈
R} is bounded in D′ (R). To prove that F (exp ·) ∈ (DL1 ,e )′ (R) and
F (exp ·)e−1 ∈ B ′ (R), it is enough to show that for every ϕ ∈ D(R) we
have
((F (exp ·)/e) ∗ ϕ)(h) ∈ L∞ (R), h ∈ R. (4.18)
Let us consider relation:

((F (exp ·)/e) ∗ ϕ)(h) = hF (exp(x + h))/c(h), ϕ(−x)c(h)/e(x + h)i.
The set {ϕ(−·)c(h)/e(· + h); h ∈ R} is bounded in D(R) because of (4.7).
Now, (4.18) follows from the fact that {F (exp(· + h))/c(h); h ∈ R} is
a bounded set in D′ (R). Hence, the distribution T = F (exp ·) and the
function K(s+·) = k(exp(−s−·)) satisfy condition 1 of Theorem 4.1. From
(4.15) and (4.16) it follows that the distribution T = F (exp ·) satisfies also
condition 2 of Theorem 4.1.
By the change of variables: x → exp t, v → exp s, we have

F (x) F (exp t)
, e(log x)k(1/xv)/x = , e(t)k(exp(−t − s)) , v ∈ V0 .
e(log x) e(t)
(4.19)
Now, from the assertion of Theorem 4.1 and from (4.19) the assertion of
Theorem 4.3 follows.
The next theorem is a modification of Theorem 4.3 which is better
adapted to some special cases. In this theorem, we use the family of tem-
pered distributions {fα ; α ∈ R} (see 0.4).
Theorem 4.4. Suppose that F ∈ S ′ (R+ ) and suppF ⊂ [a, ∞), a > 0, and:
1. The set {F (u)/c(log u); u ∈ R+ } is bounded in S ′ (R+ ), where c is
given by (4.4)
2. Let a > b and V0 ⊂ C be such that R+ V0 ⊂ V0 . Let k(·/v) ∈
Ea+δ,b (R+ ) for every v ∈ V0 and for a δ > 0.
3. The function k̂, defined by k̂(1/t)/t = (−1)m Dtm (k(1/t)/t), t > 0 has
the property that k̂(·/v) ∈ Ea+m,b+m′ (R+ ), where m ≥ m′ ≥ 0; a+m−m′ >
b, a + m > −1; v ∈ V0 .
q
4. F ∼ c(log u)Afr+1 , u → ∞ in S ′ (R).
Then, there exists
Mk (F )(v) = hF (x), k(1/xv)/xi = v m h(fm ∗ F )(x), k̂(1/xv)/xi, v ∈ V0
and
limu→0+ Mk (F )(vu)/c(− log u) = v m hAfr+m+1 (x), k̂(1/xv)/xi
= hAfr+1 (x), k(1/xv)/xi, v ∈ V0 .
Proof. We denote by ĉ the following function:

(
exp(ax)L(exp x), x≥0
ĉ = ′
exp((b + m − m)x), x < 0 .
We have

(fm ∗ F )(ux) c(log u) F (ux)
, ϕ(x) = , hfm (t), ϕ(x + t)i , u > 0 .
um ĉ(log u) ĉ(log u) c(log u)
Since ϕ(x) → hfm (t), ϕ(x + t)i is an automorphism of S+ (see Chapter 1,

Section 3 in [192]), it follows that {(fm ∗ F )(u·)/(um ĉ(log u)); u ∈ R+ }
is a bounded set in S ′ (R+ ) and in D′ (R+ ), as well. Hence, conditions 1

and 2 of Theorem 4.3 are satisfied, where instead of F, k, a and b, we have
fm ∗ F, k̂, a + m and b + m′ , respectively. Now, from assumptions 1,2 and
3 in Theorem 4.4 it follows that for 0 < u < 1 and v ∈ V0
Mk (F )(uv) = hF (x), k(1/xuv)/x)i = (uv)m h(fm ∗ F )(x), k̂(1/xuv)/xi
= v m h(fm ∗ F )(t/u)/(u−m ), k̂(1/tv)/ti.
By assumption 4 in Theorem 4.4 and by Theorem 1, Chapter 1, Section 3
q
in [192], we have fm ∗ F ∼ k m c(log k) · Afr+m+1 , k → ∞ in S ′ (R). Hence,
Mk (F )(vu)
lim = v m hAfr+m+1 (x), k̂(1/xv)/xi
u→0+ c(− log u)
= hAfr+1 , k(1/xv)/xi, v ∈ V0 .
4.1.2 Special integral transforms
Now we apply results of the previous two sections to some special integral
transforms (cf. [158]).
Note that Zemanian [202] introduced the spaces Lc,d and Mc,d in order
to analyze the asymptotic behavior of integral transforms of distributions:
Lc,d = {f ∈ C ∞ (R); γn (f ) = sup|gc,d(t)f (n) (t)|< ∞, n ∈ N0 } ,
t∈R
where gc,d (t) = exp(ct) if t ≥ 0 and gc,d (t) = exp(dt) if t < 0; {γn } is a
multi-norm in Lc,d . It is easily seen that Lc,d ⊂ DL1 ,eδ if c > a + δ and
d < b, and the topology in Lc,d is finer than the topology induced by DL1 ,eδ .
This is a consequence of the relation
Z m
X
|(eδ f )(m) (t)|dt ≤ ak γk (f ),
R k=0
where ak , k = 0, 1, . . . , m are positive numbers. Hence, for strong duals,

we have D′ L1 ,eδ ⊂ L′c,d if c > a + δ and d < b. Therefore, Theorem 4.1 can
be applied to all integral transforms with the kernels in Lc,d .
The space Mc,d is isomorphic to Lc,d via the mapping: f (x) →
f (e−t )e−t = g(t), where x = exp(−t). It can be defined in the following
way. Let Gc,d be the function: Gc,d (x) = x−c if 0 < x ≤ 1; Gc,d (x) = x−d
if 1 < x < ∞. Then
Mc,d = {w ∈ C ∞ (R+ ); Hk (w) = sup |Gc,d (x)xk+1 w(k) (x)|< ∞, k ∈ N0 }.
0<x<∞
{Hk }k∈N0 is a multi norm in Mc,d .
Proposition 4.4. If w ∈ Mc,d and c > a + δ + 1, for a δ > 0, d < b + 1,

then w ∈ Ea+δ,b (R+ ), as well.
Proof. For 0 < x < e−1 ,

|xk−1 qδ (x)w(k) (x)|≤ x−(a−c+δ+2) |xk+1−c w(k) (x)|≤ C ′ x−(a−c+δ+2) .
For x > 1,
|xk−1 qδ (x)w(k) (x)|≤ x−(b−d+2) |xk+1−d w(k) (x)|≤ C ′′ x−(b−d+2) .
These two inequalities imply the assertion of Proposition 4.4.
From Proposition 4.4 it follows that Theorem 4.3 and Theorem 4.4 can
be applied to all integral transforms with the kernels in Mc,d . Such kernels
appeared in [24], and [104].
Weierstrass transform. (See [202]). The kernel x 7→ K(s + x) =
(4π)−1/2 · exp(−(x + s)2 /4), x ∈ R belongs to DL1 ,eδ (R) for every a, b ∈ R
and s ∈ C. We can apply Theorem 4.1 and have the same result as in [108].
Poisson transform. (See [189]. The kernel
x 7→ K(s + x) = y((x + s)2 + y 2 )−1 , x ∈ R ,
belongs to DL1 (R) for y > 0 and s ∈ C. We can apply Theorem 4.2.
Laplace transform. (See [192]) L[f ](z) = hf (x), eixz i, z ∈ R + iR+ .
The kernel of the Laplace transform is x 7→ exp(−vx), Rev > 0. Hence, for
t > 0, k(1/t)/t = exp(−t) and k̂(1/t)/t = exp(−t), as well. By the relation
X n
1 1 1
Dxn exp − = an,k x−k x−n−1 exp −
x x x
k=0
it follows that k(·/v) ∈ Ea,b (R+ ) for any a ∈ R, b > −1 and Rev > 0; also
k̂(·/v) ∈ Ea+m,b (R+ ) for any m ≥ 0; we can choose m in such a way that
a + m > b and a + m > −1. A consequence of Theorem 4.4 is now:
Theorem 4.5. Suppose that F ∈ S ′ + and b > −1. Let:

1. {F (·u)/c(log u); u ∈ R+ } be a bounded set in S ′ (R);
q
2. F ∼ c(log u) · Afα+1 , u → ∞ in S ′ (R).
Then
lim u/c(− log u)hF (x), exp(−xvu)i = Av −α−1 , Rev > 0 .
u→0+
Remark. Theorem 4.5 asserts the same as Theorem 1, Section 4.2, in

[192] in the part concerning the results of Abelian type.
Stieltjes transform. Lavoine and Misra [89] defined the Stieltjes
transform by introducing a space J ′ (r), r ∈ R \ (−N); it consists of all
distributions f ∈ S ′ + such that there exists k ∈ N0 and a locally integrable
function F, supp F ⊂ [0, ∞) such that f = Dk F (D is the derivative in the
sense of distributions) and
Z
|F (t)|(t + 1)−(r+n+1) dt < ∞ .
R+
The Stieltjes transform of f ∈ J′ (r) is defined by

Z∞
Sr [f ](s) = (r + 1)n F (t)(t + s)−(r+n+1) dt, s ∈ C \ R− ,
0
where (r + 1)n = (r + 1) . . . (r + n), n ≥ 1, (r + 1)0 = 1. Sr [f ] can be

written in the form:
Sr [f ](s) = h(fn ∗ f )(t), (r + 1)n (t + s)−(r+n+1) i, s ∈ C \ R− .
Or, if we replace s by 1/v, v ∈ C \ R−
v −(r+n+1) Sr [f ](1/v) = h(fn ∗ f )(t), (r + 1)n (tv + 1)−(r+n+1) i. (4.20)
For other definitions of the Stieltjes transform, we refer to [100], [110],

[135], [150] and [157]. In order to apply Theorem 4.4 to the Stieltjes trans-
form, we have to remark that k(1/t)/t = (t + 1)−r−1 , t > 0, and
k̂(1/t)/t = (−1)n Dtn (k(1/t)/t) = (r + 1)n (t + 1)−(r+n+1) , t > 0 .
Hence, k(x) = xr (x+1)−r−1 and k̂(x) = (r+1)n xr+n (x+1)−(r+n+1) , x > 0.
If we compare (4.17) and (4.20), we find that Mk (f )(v) = v −r Sr [f ](1/v).
Using the following relation
p
X
k (p) (x) = ap,i xr−p+i (x + 1)−(r+i+1) , x > 0 ,
i=1
where ap,i are real numbers, we can see that k(·/v) ∈ Ea,b if r > a, b > −1
and v ∈ C \ R− . In the same way, we have that k̂(·/v) ∈ Ea+m,b (R+ ). Now,
by Theorem 4.4, we have
Theorem 4.6. Suppose that f ∈ S ′ + . Moreover, suppose that a, b are real

numbers such that r > a > b > −1, r 6∈ −N. Let:
1. The set {f (·u)/c(log u); u ∈ R+ } be a bounded set in S ′ (R);

q
2. f ∼ c(log u) · Afα+1 , u → ∞ in S ′ (R).
Then, there exist
Sr [f ](s) = h(fn ∗ f )(x), (r + 1)n (x + s)−(r+n+1) i
and
lim Sr [f ](sh)/(h−r c(log h)) = AΓ(r − a)/Γ(r + 1)sa−r ,
h→∞
where s ∈ C \ R− .
This is the well-known result (see, for example, [167]).

We shall now apply asymptotics expansions at infinity to obtain Abelian
type theorems.
The Weierstrass transform with parameter t of an f ∈ K′ 1 is defined by
Wt [f ](s) = hf (x), k(s − x, t)i ,
where k(s, t) = (4πt)−1/2 exp(s−2 /(4t)), s ∈ C, t > 0.
Proposition 4.5. Let f ∈ K′ 1 (R), um ∈ K′ 1 (R) and let the slowly vary-
ing functions Lm be monotonous for sufficiently large arguments, m =
1, . . . , p < ∞ or m ∈ N. Denote by cm (h) = exp(am h)Lm (exp h), am ∈ R.
If f has the S-asymptotic expansion of the first type, related to the sequence
{cm (h)} (see Remark 3) after Definition 1.4):
∞
X
s
f (x + h) ∼ um (x)|{cm (h)}, h ∈ R+ , h → ∞ ,
m=1
then, for any s ∈ C in the sense of the ordinary asymptotics
m
X
Wt [f ](s + h) ∼ A(s + h)t,i |{ci (h)}, h → ∞,
i=1
where A(s)t,m = Wt [um ](s), m = 1, . . . , p or m ∈ N.
See [118] for the proof.

Now, we present the Abelian type results for the Stieltjes and Laplace
transforms of distributions at zero by using asymptotic expansion at zero.
Suppose that f ∈ S ′ + and that there exist m ∈ N and a locally
integrable function F, supp F ⊂ [0, ∞) such that f = Dm F and that
F (x)(x + s)−r−m−1 ∈ L1 , s ∈ R+ .
Proposition 4.6. Let hLk ik∈N , be a sequence of slowly varying functions

at zero, hαk ik∈N , be a strictly increasing sequence of real numbers, hAk ik∈N ,
be a sequence of real numbers and let r ∈ R\(−N), r > αk , k ∈ N. Assume
that f ∈ S ′ + has the quasi-asymptotic expansion at zero of the first kind:
q.e. X
f ∼ Ak fαk +1 |{εαk Lk (ε)}, ε → 0+ .
k∈N
Then, in the sense of ordinary asymptotic expansion, we have

∞
X εαk −1
Sr [f ](ε) ∼ Ak Γ(r − αk )|{εαm Lm (ε)}, ε → 0+ . (4.21)
Γ(r + 1)
k=1
Proof. Since for every m ∈ N,

m
X
(f (εx) − Ak fαk +1 (εx)/(εαm Lm (ε)) → 0, ε → 0+ , in S ′ (R) ,
k=1
Theorem 2.17 (at zero) implies that there exists pm ∈ N0 and a continuous
function F, supp F ∈ [0, ∞), such that f = Dpm F and
m
X
(F (εx) − Ak fαk +pm +1 (εx)/(εαm +pm Lm (ε)) → 0, ε → 0+ , in S ′ (R) .
k=1
First, assume that r − 1 ≤ αm < r. Then, as in ([[135], Ch. 4]), with

η ∈ C ∞ , η = 1, x > −δ, (δ > 0), η = 0, x < −2δ, we have
" m
#
1 X
Sr+1 f − Ak fαk +1 (εs)
εαm −1−r Lm (ε)
k=1
* m
! +
(r + 2)p X η(x)
= αm −r−1 m F− Ak fαk +pm +1 (x),
ε Lm (ε) (εs + x)r+pm +2
k=1
* F (εx) − P Ak fα +p +1 (εx)
m
+
k m
k=1 η(x)
= (r + 2)pm , ,
εαm +pm L m (ε) (s + x)r+pm +2
where h·, ·i is the dual pairing of S ′ pm +r+1 (R) and Spm +r+1 (R). Because
of αm > r − 1, it follows that η(x)/(s + x)r+pm +2 ∈ Spm +r+1 and, hence,
for s > 0,
m
X
1
(F (εs)− Ak fαk +pm +1 (εs)) →0 in S ′ pm +r+1 (R) as ε → 0+ .
εαm +pm Lm (ε)
k=1
This implies that for every s > 0,

m
X
1
S r+1 f − Ak f α +1 (εs) → 0, ε → 0+ .
εαm −r−1 Lm (ε) k
k=1
Since
Z∞ m
X
Pm
Sr f − k=1 Ak fαk +1 (ε) = (r + 1) Sr+1 f − Ak fαk +1 (u)du
ε k=1
Z∞ m
X
= ε(r + 1) Sr+1 f − Ak fαk +1 (εx)dx.
1 k=1
By Lebesgue’s theorem, we have (4.21) for αm ≥ r − 1. If αm < r −

P
m
1, then we consider the Sr transform of f − Ak fαk +1 by noting that
k=1
η(x)
∈ Sr+pm . Then for x > 0 it follows that
(s + x)r+pm +1
m
X
1
(F (εx) − Ak fαk +pm +1 (εx)) → 0, ε → 0+ in S ′ r+pm (R).
εαm +pm Lm (ε)
k=1
The proof is completed by using the equality

εαk −r Γ(r − αk )
Sr [fαk +1 ](ε) = , ε > 0, r > αk .
Γ(r + 1)
In a similar way, we can prove the assertion for the Laplace transform.
Let us remark that there is a lot of published papers dealing with
Abelian type theorems for the integral transforms of generalized functions.
We mention some of them: ([22], [23], [35], [65], [90], [96], [98], [125], [126],
[190], [203]).
4.2 Tauberian type theorems
Recall the celebrated Wiener–Tauberian theorem [198]: Suppose that f ∈

L∞ (R), k ∈ L1 (R) and F [k](y) 6= 0, for y ∈ R ( F [k] is the Fourier
transform of k). If
Z Z
lim f (y)k(x − y)dy = a k(y)dy, a ∈ R ,
x→∞
R R
then for every G ∈ L1 (R)

Z Z
lim f (y)G(x − y)dy = a G(y)dy.
x→∞
R R
This theorem has been much used in various branches of mathemat-

ics and so generalizations of it are, even now, important. Pitt’s form of
Wiener’s theorem [140], [141] describes the behavior of the function f as
x → ∞, with some additional conditions on f. Many generalizations of
these two basic results have been proved (c.f. [9]). For the Tauberian re-
mainder theorems, we refer to Ganelius [62]. Recently, some interesting
results on Tauberian type theorems have been published in [28].
The Wiener–Pitt type theorems can be proved in a quick and elegant
way by a Banach-algebra approach (see [142]), or by the method of gen-
eralized functions as in [60]. One can make use of the direct connection
between spectral synthesis and Tauberian theorems (see [7]).
Tauberian type results related to generalized asymptotic behavior have
been elaborated only for some special integral transforms of distributions
(see for example [135], [192]). It was natural to expect theorems of Wiener’s
or Pitt’s form. In [105] Peetre proved a Wiener–Tauberian theorem when
the kernel k belongs to a Banach space W, with a translation invariant norm
and with the properties that the space of rapidly decreasing functions is a
dense subspace of W and W ∗ W ′ ⊂ L1 .
We shall discuss the Wiener–Tauberian theorems in a more general
setting.
4.2.1 Convolution type transforms in spaces of distributions
We quote the next two theorems only with the ideas of the proof. For the
complete proof cf. [129]. Similar theorems are Theorems 4.9 and 4.10, but
in the space of ultradistributions, which are with the complete proof.
Theorem 4.7. Let f ∈ B ′ (R) and K ∈ DL1 (R) be such that F [K](ξ) 6= 0,
for ξ ∈ R. If
Z
lim (f ∗ K)(x) = A K(t)dt, A ∈ R ,
x→∞
R
then for every ψ ∈ DL1 (R),

Z
lim (f ∗ ψ)(x) = A ψ(t)dt.
x→∞
R
Proof. We follow the proof of the Wiener–Tauberian Theorem given in

([31], pp. 234–235). Let M be the subspace of DL1 (R) consisting of all
finite linear combinations of translations of K. First, we show that M is
dense in DL1 (R).
We recall that by Theorem XXV, Chapter VI in [146], f ∈ B ′ (R) if and
only if
m
X (i)
f= Fi , (4.22)
i=0
where Fi , i = 0, 1, . . . , m, are bounded continuous functions on R and the

derivatives are in the distributional sense.
Let ψ ∈ DL1 (R) and ε > 0. There is H ∈ M such that
kψ (k) − H (k) kL1 < ε, k = 0, . . . , m .
By Lebesgue’s theorem and by (4.22)
Z X
m
|(f ∗ ψ)(x) − a ψ(t)dt|≤ kFi kL∞ +a + 1 ε.
R i=0

We denote by η a smooth function which is equal to 1 in a neighborhood
of ∞, and to 0 in a neighborhood of −∞.
Theorem 4.8. For f ∈ D′ (R) and K ∈ C ∞ (R), we assume that

(i) the set {f (· + h)/(c); h ∈ R} is bounded in D′ (R),
(ii) there exists δ > 0 such that
η Ǩ exp((α + δ)·), (1 − η)Ǩ exp(β·) ∈ DL1 .
Then
(a) F [K exp(−α·)](ξ) = F [K](ξ − iα), for ξ ∈ R, exists,
(b) the convolution f ∗ K exists.
Moreover, if we assume that
(iii) F [K](ξ − iα) 6= 0, for ξ ∈ R,

Z
(f ∗ K)(x)
(iv) lim = A K(t) exp(−αt)dt, for A ∈ R,
x→∞ L(exp x) exp(αx)
R
then
(c) for every ψ ∈ C ∞ (R) for which η ψ̌ exp((α + δ)·), (1 − η)ψ̌ exp(β·) ∈
DL1 ,
Z
(f ∗ ψ)(x)
lim = A ψ(t) exp(−αt)dt .
R
Proof. We shall use the functions L∗ , c, e and eδ defined in 4.1 by (4.4),

(4.5) and (4.6), respectively in which we replace a by α and b by β.
First step: The proof of (a) and (b). From assumption (ii), we have
Ǩ(α·) ∈ DL1 (R) and this implies (a). To prove (b), we first prove that
f /e ∈ B ′ (R).
Then the proof of assertion (b) follows from the equality

f (x + h)
(f ∗ K)(h) = hf (x + h), Ǩ(x)i = , e(x + h)Ǩ(x) , h ∈ R ,
e(x + h)
if we prove that for every h ∈ R, e(· + h)Ǩ ∈ DL1 (R).
Second step: We prove that assumption (iv) (with (ii) and (i)) implies
that

f f (x + h)
∗ (K exp(−α·)) (h) = · Ǩ(x) exp(αx)
e e(x + h)
Z
→ A K(t) exp(−αt)dt, as h → ∞. (4.23)
R
Third step. We prove that (f ∗ ψ)(h), h ∈ R, exists and that (c) holds.
Proposition 4.7. Let f ∈ D′ (R) be such that the set

{f (· + h)/(L(eh )eαh ); h > 0}
is bounded in D′ (R) and let φ ∈ D(R) be such that the Fourier transform
of φ exp(−α·) is different from zero on R. If
Z
(f ∗ φ)(x)
lim = A φ(t) exp(−αt)dt, A ∈ R ,
R
then for every ψ ∈ D(R)

Z
(f ∗ ψ)(x)
lim =A ψ(t) exp(−αt)dt .
R
Proof. If in Theorem 4.8, we take K ∈ D(R), then assumption (ii) is

satisfied for every α, β, δ ∈ R, and assumptions (i), (iii) and (iv) of Theorem
4.8 imply the assertion in Proposition 4.7.
Remark. A direct consequence of Proposition 4.7 is: If
lim hf (x + h)/(eαh L(h)), φ(x)i
h→∞
exists for a φ such that F [e−αφ· ](x) 6= 0, x ∈ R and the set {f (x +

h)/(eαh L(h)); h > 0} is bounded in D′ (R), then f has the S-asymptotics
related to eαh L(eh ).
Proposition 4.8. Let f and k be from L1loc (R) and let α > β be such that:
(i) f /c ∈ L∞ (R), where c(x) = L(exp x) exp(αx), if x ≥ 0 and c(x) =
exp(βx) if x < 0;
(ii) ǩ exp((α + δ)·) ∈ L1 ((a, ∞)) and ǩ exp(β·) ∈ L1 ((−∞, b)) for some
a, b ∈ R, and some δ > 0, where ǩ(x) = k(−x).
Then (f ∗ k)(x), x ∈ R, exists and ǩ exp(α·) ∈ L1 (R).
Moreover, if we assume:
(iii) F [k exp(−α·)](y) 6= 0, for y ∈ R,
Z
(f ∗ k)(x)
(iv ) lim = A k(t) exp(−αt)dt, A ∈ R,
R
then
Z
(f ∗ ψ)(x)
lim =A ψ(t) exp(−αt)dt
R
for every ψ ∈ L1loc (R)

such that ψ̌ exp((α + δ)·) ∈ L1 ((a, ∞)) and
1
ψ̌ exp(β·) ∈ L ((−∞, b)).
In particular, if β + 1 > 0, then
Zx
A
f (log t)dt ∼ xα+1 L(x), x → ∞ .
α+1
0
Proof. Proposition 4.8 is not a direct consequence of Theorem 4.8, but

the proof of it is a version similar to the proof of Theorem 4.8 (cf. [129]).
Remark. Theorem 1.7.5 in [9] gives conditions on f under which
f (t) ∼ AL(exp t) exp(αt), t → ∞ .
Proposition 4.8 contains both forms, Wiener’s and Pitt’s, of a Tauberian

theorem (cf. [9], [29], [140]).
For simplicity, all results are given in one-dimensional case. They can
also be obtained for a many-dimensional case with the asymptotic behavior
in a cone.
Let us denote by W ′ the subspace of D′ (R) consisting of those T for
which {T (x + h); h ∈ R} is a bounded subset in D′ (R). This space is the
union of all the spaces introduced in [105]. Let θ be a smooth function on
R such that supp θ ⊂ [a, ∞) for some a ∈ R and θ(x) = 1 for x > b for
g.s
some b > a. Let f (x) ∼ c(x), x → ∞. Then by Theorem VI, Chapter VII
in [146], we have, θf /c ∈ S ′ ; θ(x + h)f (x + h)/c(x + h) → 1, h → ∞, in
the sense of convergence in S ′ (R).
By Theorem 1 in [105], we directly obtain
Proposition 4.9. Let f ∈ D′ (R), c ∈ C ∞ (R), c(x) 6= 0, x ∈ R, and

ϕ ∈ S(R) be such that the Fourier transform of ϕ, denoted by ϕ,
b is different
′
from 0 on R. If θf /c ∈ W and hf (x + h)/c(x + h), ϕ(x)i → A · B, h → ∞,
R
+∞
g.s
where A 6= 0, B = ϕ(x)dx = ϕ̂(0) 6= 0, then f (x) ∼ A · c(x), x → ∞.
−∞
We assume in the next proposition that c̃(x), x > 0, is constructed as

e(x) in Lemma 1.4.
Proposition 4.10. Let f be a non-negative distribution on R such that

f /c̃ ∈ W ′ . Let ϕ ∈ D(R) and ϕ be different from 0 on R and such that
lim hf (x + h)/c(h), ϕ(x)i = Ch1, ϕ(x)i, C 6= 0. (4.24)

h→∞
s
Then, f (x + h) ∼ c(h) · C, h ∈ R+ .
Proof. By Lemma 1.4, we have c(h)/c̃(x+h)−1 → 0, h → ∞, uniformly

on supp ϕ. Because of that, for a sequence of positive numbers hεn in∈N
which tends monotonically to 0 there is an increasing sequence hhn in∈N

such that
−εn ϕ(x) ≤ (c(h)/c̃(x + h) − 1)ϕ(x) ≤ εn ϕ(x), x ∈ K, h > hn .
This implies (for h > hn )

f (x + h) f (x + h) c(h)
−εn , ϕ(x) ≤ , − 1 ϕ(x)
c(h) c(h) c̃(x + h)

f (x + h)
≤ εn , ϕ(x) ,
c(h)
and because of (4.24)

f (x + h) c(h)
, − 1 ϕ(x) → 0 as h → ∞ .
c(h) c̃(x + h)
Whence
f (x + h) f (x + h)
lim , ϕ(x) = lim , ϕ(x)
h→∞ c̃(x + h) h→∞ c(h)

f (x + h) c(h)
+ , − 1 ϕ(x) = Ch1, ϕ(x)i . (4.25)
c(h) c̃(x + h)
g.s
Now Proposition 4.9 implies that for suitable A ∈ R, f (x) ∼ Ac̃(x), x →
∞. This implies that (4.25) holds for any test function from S(R) and the
assertion is proved.
Proposition 4.11. Let f be a non-negative distribution, f /c̃ ∈ W ′ , and

ϕ ∈ D(R) such that ϕ is different from 0 on R. Let
lim hf (x + h)/c(h), e−αx ϕ(x)i → Ch1, ϕ(x)i, α 6= 0 .
h→∞
s
Then f (x + h) ∼ c(h)Ceαx , h → ∞.
Proof. We apply the preceding proof on f (x)e−αx , x ∈ R.

Let c(h) = hβ L(h), h > 0, where β ∈ R and L(h), h > 0, is a slowly
varying function for which we assume to be monotonous and to belong to
C ∞ on (0, ∞). Proposition 1.6 with the Remark asserts. “If f ∈ L1loc and
for some m0 ∈ N0 and x0 ∈ R, f (x)xm0 is non-decreasing for x > x0 , then
s
the assumption f (x + h) ∼ c(h) · 1, h ∈ R+ , implies f (x) ∼ c(x), x → ∞”.
This assertion and Proposition 4.10 imply the following result.
Proposition 4.12. Let f ∈ L1loc (R) be such that for some m0 ∈ N and
x0 ∈ R the function f (x)xm0 is non-negative and non-decreasing for x >
x0 . If for some ϕ ∈ S(R) the Fourier transform is different from 0 on R,

and
Z∞ Z∞
1
f (x + h)ϕ(x)dx → C ϕ(x)dx, h → ∞ ,
hβ L(h)
x0 −∞
β
then f (x) ∼ Cx L(x), x → ∞.
Proof. We consider the function x → θ(x)f (x), where θ ∈ C ∞ (R), θ

is non-decreasing, θ(x) = 0 for x ≤ x0 , θ(x) = 1 for x ≥ x0 + 1, and apply
the previous assertions.
4.2.2 Convolution type transforms in other spaces of generalized

functions
First, we consider in the space of ultra distributions D∗ (R) (cf. 0.5) in

which (M.1), (M.2) and (M.3)’ are satisfied. We use the function c0 and its
regularization c given by
(
L(ex )eαx , x ≥ 0,
c0 (x) = α, β ∈ R, α > β . (4.26)
eβx , x < 0,
and
c(x) = (c0 ∗ ω)(x), x ∈ R, (4.27)
∗
R
where ω ∈ D , supp ω ⊂ [−1, 1], ω ≥ 0 and ω(t)dt = 1.
R
Let us denote by η a function from E ∗ with the properties
η(x) = 1, x > x0 > 0, η(x) = 0, x < −x0 . (4.28)
We shall use the following property of a slowly varying function (cf. [9]).
For every δ > 0 there is Cδ > 0 such that
δ δ δ δ
1 x y L(x) x y
min , ≤ ≤ Cδ max , , x > 0, y > 0 .
Cδ y x L(y) y x
(4.29)
∗
Theorem 4.9. Let f ∈ B ′ (R) and K ∈ DL∗
1 (R) such that F [K](ξ) 6=
0, ξ ∈ R. If
Z
lim (f ∗ K)(x) = a K(t)dt, a ∈ R ,
x→∞
R
∗
then for every φ ∈ DL 1 (R),
Z
lim (f ∗ ψ)(x) = a ψ(t)dt.
x→∞
R
Proof. First, we need the following version of Beurling’s theorem ([31])

for bounded ultradistributions.
∗
“Let f ∈ B ′ (R). A point ξ0 belongs to supp fˆ if and only if there is a
sequence of functions hϕn in∈N from S ∗ (R) such that
fn (x) = (f ∗ ϕn )(x), x ∈ R, n ∈ N ,
converges narrowly to f0 (x) = eixξ0 , x ∈ R, n → ∞”.

Recall from [31] that a sequence of continuous and bounded functions
fn on R converges narrowly to a continuous bounded function f0 if and
only if fn converges to f0 uniformly on bounded sets in R and
kfn kL∞ → kf0 kL∞ , n → ∞ .
The proof of this assertion is the same as for bounded distributions

since all the properties of Schwartz’s test functions which were used in
[31], pp. 230–231], have been proved in [79] and [123] for ultradifferentiable
functions. The same holds for the next assertion, based on the previous
one, which is analogous to Theorem on p. 232 in [31].
∗
“Let f ∈ B ′ (R) and K ∈ DL
∗
1 (R). If K ∗ f = 0 on R, then K̂(ξ) = 0
ˆ
for ξ ∈ supp f ”.
First, we shall prove that the set M which consists of finite linear com-
∗ ∗
binations of translations of K ∈ DL 1 (R) is dense in DL1 (R). By the
∗
property of dual pairing, M is dense in DL1 (R) if and only if for ev-
∗
ery S ∈ B ′ (R), S ∗ Ǩ = 0 ⇐⇒ S = 0. For, if M were not dense,
∗
we would have an S0 ∈ B ′ (R), S0 6= 0 such that S0 ∗ Ǩ = 0. Thus
F [Ǩ](ξ) = 0, ξ ∈ supp F [S0 ]. Since we assume that F [Ǩ](ξ) = F [K](−ξ)
∗
is never zero, we conclude that M is dense in DL 1 (R).
From that and previous statement, we obtain the proof of the quoted
Wiener theorem.
Note (cf. [123]), f ∈ B ′ ∗ (R) if and only if it is of the form
∞
X
f= Dα Fα , Fα ∈ L∞ , α ∈ N0 ,
α=0
∗
where D is the derivative in B ′ (R) and Fα , α ∈ N0 , are such that for some
h > 0 (in the (Mα )-case), respectively, for every h > 0 (in the {Mα }-case)
X∞
Mα
kFα kL∞ = Kh < ∞ . (4.30)
α=0
hα
∗ ∗
Let ψ ∈ DL 1 (R). Since M is dense in DL1 (R), then: In the (Mp ) case,
for every ε > 0 and every h > 0, there is Hh ∈ M such that

kHh − ψkL1 ,h < ε . (4.31)
In the {Mp } case, we have that for every ε > 0 there is h > 0 and Hh such
that (4.31) holds.
In the {Mp } case, the assumption of the theorem and Lebesgue’s theo-
rem give that for x > x0 (ε), where x0 (ε) is large enough, we have
Z Z
|(f ∗ ψ)(x) − a ψ(ξ)dt| ≤ |((ψ − H) ∗ f )(x) − a (ψ(t) − H(t))dt|
R R
Z
+ |(H ∗ f )(x) − a H(t)dt|
R
X Z
∞

≤ (ψ − H)(α) (t)Fα (x − t)dt|
i=0 R
Z Z
+a |ψ(t)−H(t)|dt + |(H ∗ f )(x)− a H(t)dt|
R R
∞
X
hα Mα
≤ sup k(ψ − H)(α) kL1 kFα kL∞ +aε
α Mα α=0
hα
Z
+ |(H ∗ f )(x) − a H(t)dt|≤ εKh + aε + ε .
R
The (Mp )-case can be proved similarly. The proof is completed.

∗
Theorem 4.10. Let f ∈ D′ (R) and K ∈ C ∞ (R). Assume:
∗
(i) f /c ∈ B ′ (R).
(ii) There exists δ > 0 such that η Ǩe(α+δ)· , (1 − η)Ǩeβ· ∈ DL
∗
1 (R).
(iii) F [K](ξ − iα) 6= 0, ξ ∈ R.

Z
(f ∗ K)(x)
(iv) lim =a K(t)e−αt dt, a ∈ R.
x→∞ L(xx )eαx
R
Then, for every ψ ∈ C ∞ (R) for which
η ψ̌e(α+δ)· , (1 − η)ψ̌eβ· ∈ DL
∗
1 (R), (4.32)
there holds
Z
(f ∗ ψ)(x)
lim =a ψ(t)e−αt dt .
x→∞ L(ex )eαx
R
Proof. We shall only prove the (Mp )-case because this proof can be
simply transferred to the {Mp }-case.
The proof is organized as follows. We shall prove in part I estima-
tions (4.33), (4.34) and (4.35) which will be used in part II to prove that
F (Ke−α· )(ξ), ξ ∈ Rn , and f ∗ K exist. In part III, we will prove the
assertion of Theorem 4.10.
Part I. Note, from the assumption that (1−η)Ǩeβ· and η Ǩe(α+δ)· belong
(M )
to DL1 p (R) and from (M.2) it follows that for every r > 0
m h
r
sup keβx((1 − η(x)Ǩ(x))(m) kL1 + (4.33)
Mm
i
+ke(α+δ)x (η(x)Ǩ(x))(m) kL1 , m ∈ N0 < ∞ .
Since e(α+δ)x ≤ eβx , for x < 0, we also have that for every r > 0
m
r
sup ke(α+δ)x ((1 − η(x))Ǩ(x))(m) kL1 (−∞,0) , m ∈ N < ∞ . (4.34)
Mm
We need the following estimate:
For every r > 0 there is C > 0 such that
k (k) αx+δ|x|
r c(x + h) αx

≤ Ce , x+h > 0,
sup − e (4.35)
k∈N0 Mk L(e )e h αh C βx , x+h < 0,
where we choose δ such that 0 < δ < α − β.
Let r > 0, k ∈ N0 and x + h > 1. By (4.27) and (4.29) we have (with
suitable constants)
(k)
rk c(x + h) αx

h αh
− e
Mk L(e )e
Z1
rk L(ex+h−t ) α(x−t) (k) |rα|k αx
≤ e |ω (t)|dt + e
Mk L(eh ) Mk
−1
≤ C1 eαx+δ|x| kωk[−1,1],r +C2 eαx ≤ Ceαx+δ|x| .
Similarly, for x + h < −1, we get that for a given r > 0 there is C > 0
such that
k (k)
r c0 (x + h) αx

≤ Ceβx .
sup h αh
−e
k∈N0 Mk L(e )e
Let u = x + h ∈ [−1, 1], r > 0 and k ∈ N0 . From (4.29), we have that

1
≤ Ceδ|x| , x ∈ R,
L(eu−x )
where C > 0 and δ > 0. Thus,
(k)
rk c(x + h) αx

h αh
− e
Mk L(e )e
1
≤ kωk[−1,1],r sup {c0 (x + h − t)} + C2 eαx
t∈[−1,1] L(eu−x )eα(u−x)
≤ C3 eαx+δ|x| + C2 eαx .
These inequalities and the assumption δ ∈ (0, α − β) imply (4.35).

Part II. Let ψ ∈ E (Mα ) (R) be such that ψ(x) = 1 on (−∞, −1) and
ψ(x) = 0 on [0, ∞). We have
eαx Ǩ(x) = eαx Ǩ(x)(1 − η(x))ψ(x)

+ eαx Ǩ(x)(1 − η(x))(1 − ψ(x)) + eαx Ǩ(x)η(x), x ∈ R .
(4.36)
(M )
Since the multiplication in DL1 α (R) is an inner operation, one can
easily prove that all the members on the right-hand side of (4.36) are
(M )
from DL1 α (R), and so the same holds for eα· Ǩ. This implies that
F (Ke )(ξ) = F (K)(ξ − iα), ξ ∈ Rn , exists.
−α·
Since

f (x + h)
(f ∗ K)(h) = , c(x + h)Ǩ(x) , h ∈ R,
c(x + h)
the existence of the convolution f ∗ K will be proved if we prove that for

every h ∈ R,
(M )
c(· + h)Ǩ ∈ DL1 α (R),
(Mp )
because by (i) f (· + h)/c(· + h) ∈ B ′ (R).
For a fixed h ⊂ R and a ψ as in (4.34), we have
c(x + h)Ǩ(x) = c(x + h)Ǩ(x)(1 − η(x))ψ(x)

+ c(x + h)Ǩ(x)(1 − η(x))(1 − ψ(x))
+ c(x + h)Ǩ(x)η(x), x ∈ R . (4.37)
(M )
By using (4.35), (4.33), (4.34) and that eα· Ǩ ∈ DL1 α (R), we prove that
(M )
c(· + h)Ǩ ∈ DL1 α (R). We only prove that c(· + h)Ǩ(1 − η)ψ is from
(M )
DL1 α (R) for every h ∈ R, because the proof that c(· + h)Ǩη belongs to
(M )
DL1 α (R) is similar. Then, one can easily see that c(·+h)Ǩ(1−η)(1−ψ) ∈
(M )
DL1 α (R).
Since

c(x + h)
αh h αx
c(x + h)Ǩ(x)(1 − η)(x))ψ(x) = e L(e ) αh −e Ǩ(x)
e L(eh )

+ eα(x+h) L(eh )Ǩ(x) (1 − η(x))ψ(x), x ∈ R,
we have to prove that

c(· + h) α· (M )
(1 − η)ψ αh h
−e Ǩ ∈ DL1 p (R) .
e L(e )
For every r > 0, x ∈ R, k ∈ N0 , by using (M.2) and (4.35), we have

k (j)
rk
X k c(x + h)

Ǩ(x)(1 − η(x))ψ(x))(k−j) αh h
− e αx

Kk j=0 j e L(e ) 1
L
k
X
AC k (2rH)k−j
≤ {k(Ǩ(x)(1 − η(x))ψ(x))(k−j) eβx kL1 }
2kj=0
j M k−j

(2rH)α
≤ C1 sup k(Ǩ(x)(1 − η(x))ψ(x))(α) eβx kL1 ,
α∈N0 Mα
where C and C1 are suitable constants.
To prove that the last supremum is bounded, we have to use the follow-
ing estimates
α
(4rH)α 1 X α
kǨ(x)(1 − η(x))(j) eβx ψ (α−j) kL1
Mα 2α j=0 j

(4rH 2 )j
≤ C2 sup kǨ(x)(1 − η(x))(j) eβx kL1
j∈N0 Mj

(4rH 2 )k−j (α−j)
× sup kψ kL∞ (−1,0) ≤ C3 ,
α∈N0 Mα−j
j≤α
where C2 and C3 are suitable constants. Thus, we have proved that the
convolution f ∗ K exists.
Part III. We are going to prove that the assumptions of the theorem
imply

f −α· f (x + h) αx
∗ Ke (h) = , Ǩ(x)e
c c(x + h)
Z
→ a K(t)e−αt dt, h → ∞ . (4.38)
R
It is enough to prove that

f (x + h) c(x + h) αx
, αh − e )Ǩ(x) → 0, as h → 0 .
c(x + h) e L(eh )
∗ P
∞
Since B ′ (R) ∋ f /c = Di Fi in order that (4.38) holds, we have to prove
i=0
that
∞
X Z (i)
i c(x + h) αx
Sh = (−1) Fi (x + h) −e Ǩ(x) dx → 0, h → ∞ .
i=0
L( eh )eαh
R
We have
N
X Z (i)
i c(x + h) αx
Sh = (−1) Fi (x + h) −e Ǩ(x) dx
i=0
L( eh )eαh
R
∞
X Z (i)
c(x + h)
+ (−1)i Fi (x + h) − e αx
Ǩ(x) dx
i=N +1
L( eh )eαh
R
= Sh,N + Sh,∞ .
Because the sum in Sh,N is finite, the proof that Sh,N → 0, h → ∞, is

the same as in the main assertion of [129]. By using (4.14), we obtain
∞
X (i)
1 Mi kFi kL∞ (2r)i
c(x + h) αx

Sh,∞ ≤ L(eh )eαh − e )Ǩ(x) 1
2i r i Mi L
i=N +1
∞
X (i)
C (2r)i
c(x + h) αx

≤ − e )Ǩ(x) .
2N +1 i=0
Mi L(eh )eαh
L1
So, if we prove that the last series is bounded with respect to h for
h ≥ h0 , then the proof that Sh,∞ → 0, h → ∞, simply follows. Put
c(· + h)
Im,h = k(( − eα· )Ǩ)(m) kL1 , m ∈ N0 , h ≥ h0 .
eαh L(eh )
We are going to prove that for every r > 0, there is a C > 0 such that
m
r
sup Im,h < C, h > h0 . (4.39)
m∈N0 Mm
This implies that the quoted series is bounded.
Let η(x), x0 > 0 be as in (4.28). We have
Z 0
−x (m)
c(x + h)
Im,h = − e αx
(1 − η(x))Ǩ(x) dx
eαh L(eh )
−∞
Zx0 (m)
c(x + h)
αx dx
+ eαh L(eh ) − e Ǩ(x)
−x0
Z∞ (m)
c(x + h)
αx dx = I1 + I2 + I3 .
+ eαh L(eh ) − e η(x)Ǩ(x)
x0
By the Leibnitz formula and by (4.35), (4.33), (4.34) and (4.29) there
are constants C1 and C which do not depend on m and p (but depend on
r) such that for δ ∈ (0, α − β)
Xm Z −h
m c(x + h)
I1 ≤ |( αh h)
− eαx )(p) e−βx keβx ((1 − η(x))Ǩ(x))(m−p) |dx
p=0
p e L(e
−∞
Z 0
−x (p)
c(x + h) αx−δx
αx
e−(αx−δx) (m−p)
+ eαh L(eh ) − e e ((1 − η(x))Ǩ(x)) dx
−h
Xm m−p
m Mp Mm−p r βx (m−p)
≤ C p m−p sup ke (1 − η(x))(Ǩ(x)) kL1
p=0
p r r m,p Mm−p
p≤m

Mp Mm−p rm−p
eαx−δx (1 − η(x))Ǩ(x)kL1
+ sup
rp rm−p m,p Mm−p (−∞,0)
p≤m
m
X
m Mm Mm
≤ C1 = C1 .
p=0
p rm (r/2)m
n (r/2)m o
This gives sup I1 < C1 .
m∈N0 Mm
In a similar way one can prove the corresponding estimates for I2 and
I3 , and the proof of (4.37) is completed.
Thus, we have proved (4.36).
If ψ ∈ E (Mα ) (R) satisfies the assumption given in (4.32), then ψeα· ∈
(M )
DL1 α (R), and we have
Z
f
∗ ψe−α· (h) → a ψ(x)e−αx dx, h → ∞.
c
R
As above, we can prove that (f ∗ ψ)(h), h ∈ R, exists.

To prove that
Z
f (x + h)
, ψ̌(x) → a ψ(x)e−αx dx, h → ∞ ,
L(eh )eαh
we have to prove that

f (x + h) c(x + h) αx
, − e ψ(x) → 0, h → ∞
c(x + h) eαh L(eh )
but this has already been done (with K instead of ψ), and the proof of
Theorem 4.10 is completed.
Remark. It is an open problem whether the assumption that the set
n f (· + h) o
∗ ∗
; h ∈ R is bounded in D′ (R) implies that f /c ∈ B ′ (R). Note
c(h)
that for distributions the corresponding assertion holds (see [146]).
∗ ∗
Corollary 4.1. Let f ∈ D′ (R) be such that f /c ∈ B ′ (R) and let φ ∈
D∗ (R) be such that F [φ](ξ − iα) 6= 0, ξ ∈ R. If
Z
(f ∗ φ)(x)
lim = a φ(t)e−αt dt, a ∈ R ,
x→∞ L(ex )eαx
then for every ψ ∈ D∗ (R)

Z
(f ∗ ψ)(x)
lim =a ψ(t)e−αt dt.
x→∞ L(ex )eαx
The proof of corollary simply follows from the given Theorem 4.10 be-
cause φ in the corollary satisfies conditions assumed for K and functions
from D∗ (R) satisfy condition (4.32) of Theorem 4.10.
The next space of generalized functions in which we consider convolution
transform will be the space of Fourier hyperfunctions Q(Dn )(cf.0.5).
We quote only the results, which are published in [112], because the idea
and the technique of the proofs are similar to those used for distributions
and ultradistributions.
Proposition 4.13. If f = [F ] ∈ Q(Dn ) and ϕ ∈ P∗ , then f ∗ ϕ ∈ Q(Dn +

iI n ) for an appropriate interval I n ⊂ Rn , I n ∋ 0 (i.e., f ∗ ϕ is a slowly
increasing real analytic function).
Proposition 4.14. Let ϕ ∈ p P∗ and F (ϕ) = ψ. Assume that there exists

δ > 0 such that ψ −1 exp(−δ ·2 + 1) ∈ P∗ . Let M be a subspace of P∗
consisting of all finite linear combinations of ϕ(· + x), x ∈ Rn . Then M is
dense in P∗ .
Theorem 4.11. Let ϕ ∈ P∗ , ψ = F (ϕ), f = [F ] ∈ Q(Dn ), F ∈ Õ((Dn +

iI)#Dn )(I = (−α, α)n , α > 0) and let c be a positive and measurable func-
tion on Rn . Assume:
c(x)
1. (i) lim = 1, t ∈ Rn .
x→∞ c(x + t)
(ii) For every ǫ > 0 there exist positive constants B and B1 such that
c(x)
Be−ǫ|t| ≤ ≤ B1 eε|t| , x, t ∈ Rn .
c(x + t)
√
z 2 +1
2. There exists ω > 0 such that ψ −1 (z)e−ω ∈ P∗ .
3. For every σ ∈ Λ, every compact set Kσ ⊂⊂ Iσ = I ∩ Γσ and for

every η > 0, there exists C > 0 such that

Fσ (x + h + iyσ )
≤ Ceη|x| , x ∈ Rn , h ∈ Rn+ , yσ ∈ Kσ .
c(h)
4. There exists A ∈ C such that lim (f ∗ ϕ)(x + h)/c(h) = hA, ϕi in

h→∞
Q(Dn ).
Then
lim (f ∗ λ)(x + h)/c(h) = hA, λi in Q(Dn ), for every λ ∈ P∗ . (4.40)
h→∞
Theorem 4.12. Let ϕ ∈ P∗ , ψ = F (ϕ) and f = [F ] ∈ Q, F ∈ Õ((Dn +

iI)#Dn ) (I = (−α, α)n , α > 0). Let c and ψ satisfy assumptions 1 and 2
of Theorem 4.11.
Assume also:
3. For every Kσ ⊂⊂ Iσ , σ ∈ Λ, there exists N > 0 such that for every
yσ ∈ K σ

Fσ (x + h + iyσ )
≤ N, x ∈ Rn , h ∈ Rn+ .
c(h)
4. There exists A ∈ C such that lim (f ∗ ϕ)(x)/c(x) = hA, ϕi in C.

x→∞
Then
lim (f ∗ λ)(x)/c(x) = hA, λi for every λ ∈ P∗ in C. (4.41)
x→∞
Comments on Theorems 4.11 and 4.12

If n = 1, then assumption 1 (ii) follows from assumption 1 (i). This
follows from Theorem 1.4.1 and Theorem 1.5.6 in [8] (with the change of
variables x = log u, u > 0). In fact, in this case we have that c(x) =
eαx L(ex ), x > x0 (cf. [135]).
The function
c(x) = (x21 + 1)p1 L1 (ex1 ) . . . (x2n + 1)pn Ln (exn ), x = (x1 , . . . , xn ) ∈ Rn ,
where Li , i = 1, . . . , n are slowly varying functions and p = (p1 , . . . , pn ) ∈
Rn , satisfies assumption 1 of Theorem 4.11 and 4.12.
√ √
2. Let ϕδ = F −1 (exp(−δ ·2 + 1)). Since exp(−δ ·2 + 1) ∈ P∗ , δ >
0, it follows that ϕδ ∈ P∗ . One can simply show that ψ = ϕδ satisfies
assumption 2 of Theorem 4.11 and 4.12.
An important function which satisfies assumption 2 in Theorem 4.11

and 4.12 is the Fourier transformation of the function K introduced by
Hörmander (cf. [70], Section 8.4):
Z
1
K(z) = eihz,ξi /I(ξ)dξ, z ∈ Ω = {z ∈ Cn ; |Im z|< 1},
(2π)n
Rn
R
where I(ξ) = e−ihω,ξi dω. Recall, I(ξ) = I0 (hξ, ξi1/2 ), ξ ∈ Rn , where
|ω|=1
Z1
2π (n−1)/2 n−1
I0 (ρ) = (1 − t2 ) 2 −1 e−tρ dt, ρ ∈ C
Γ((n − 1)/2)
−1
is an entire function which satisfies the estimate

|I0 (ρ)|≤ C(1 + |ρ|)−(n−1)/2 e|Reρ| , ρ ∈ C (4.42)
(cf. Lemma 8.4.9 in [70]).
We shall prove that K ∈ P∗ . By Lemma 8.4.10 in [70], K is analytic
e = {z ∈ Cn ; hz, zi 6∈
in every open connected set Ω satisfying Ω ⊂ Ω
(−∞, −1]}. One can simply prove that the strip Ω = {z ∈ Cn ; |yk |<
√ e
1/(2 n), k = 1, . . . , n} is a subset of Ω.
Let Γ be a closed cone such that if z ∈ Γ \ {0}, then |xk |> |yk |, k =
1, . . . , n. If z ∈ Γ \ {0}, then hz, zi 6∈ (−∞, 0]. By Lemma 8.4.10 in [70],
there exists c > 0 such that K(z) = O(e−c|z| ), z ∈ Γ, |z|→ ∞. Hence,
|K(z)|≤ CK,ε e−(c−ε)|Re z| , z ∈ Rn + iK ,
√
for every compact set K ⊂ {y ∈ Rn ; |yk |< 1/(2 n), k = 1, . . . , n} and
every ε > 0. Consequently K ∈ P∗ .
We denote by ψ the Fourier transform of K, i.e., ψ = F (K), and ω > 1.
Then 1/ψ(ζ) = I(ζ), ζ ∈ C is an entire function. Let ζ = ξ + iη, |η|< 1,
then
Z

|I(ζ)|= e−hω,ξ+iηi dxi = I(ξ) = I0 (ξ), ξ ∈ C .
|ω|=1
Now, by (4.40)
|I(ζ)|≤ |I(ξ)|≤ |I0 (|ζ|)|≤ C(1 + |ζ|)−(n−1)/2 e|ξ|+1 , ξ ∈ Rn , |η|< 1 .
p
This implies I(ζ) exp(−ω ζ 2 + 1) ∈ P∗ .
3. The S-asymptotics of Fourier hyperfunctions is defined in the usual
way:
Definition 4.3. Suppose that c is a positive function defined on Rn and

f ∈ Q(Dn ). It is said that f has the S-asymptotics related to c with the
limit u ∈ Q(Dn ) if
lim hf (t + x)/c(x), ϕ(t)i = hu, ϕi for every ϕ ∈ P∗ .
x→∞
Theorem 4.12 asserts that if f, ϕ̌ and c satisfy suppositions of Theorem

4.12 and if
* +
f (· + x) D E
lim , ϕ̌ = A, ϕ̌ ,
x→∞ c(x)
then f has the S-asymptotics related to c with the limit A.
~ ′ which is isomorphic to Q(Dn ).
Zharinov in [204] defined the space Φ
But in the same paper he constructed the space Λ ~ ′ (Σ) ⊂ Φ
~ ′ , where Σ is a
n
domain in R , 0 ∈ Σ. He defined the quasi-asymptotics in Λ ~ ′ (Σ).
Let Γ be a convex closed acute cone in Rn . We denote by Σ = intΓ∗ ,
where Γ∗ is the dual cone to Γ. We will follow definitions and results given
in [204] and [205].
Let A and B be bounded domains in Rn . Denote by sB (ξ) =
sup{−yξ; y ∈ B} and by Λ(A, B) the Banach space of functions holomor-
phic on Rn + iA and such that
kϕkA
−sB = sup{e
−sB (ξ)
|ϕ(ξ + iη)|; ζ ∈ Rn + iA} < ∞
with the topology given by the norm k·kA
−sB . It is easy to see that Λ(A, B) ⊂
′ ′ ′ ′
Λ(A , B ), when A ⊂ A and B ⊂ B . With the inclusion mapping ρAB,A′ B ′ :
Λ(A, B) → Λ(A′ , B ′ ), we can define
−
→ ←
−
Λ (Σ) = ind lim Λ(A, B); Λ (Σ) = proj lim Λ(B, A) .
A∋0,B⊂⊂Σ B⊂⊂Σ,0∈A
−
→ −
→
The space Λ (Σ) is a DFS space and its dual space Λ′ (Σ) is an FS space.
←
− −
→
But Λ (Σ) is an FS space. Zharinov proved that Φ ~ ′ Γ ⊂ Λ′ (Σ) ⊂ Φ
~ ′ , 0 ∈ Σ,
~ ′ ~ ′
where Φ Γ = {g ∈ Φ ; supp g ⊂ Γ}.
Now, we can quote the definition of the quasi-asymptotics.
Definition 4.4. ([205]) Suppose that g ∈ Λ ~ ′ (Σ) and that ρ is a positive

and continuous function on (0, ∞). If there exists
lim g(tζ)/ρ(t) = h(ζ) ~ ′ (Σ), h 6= 0,
in Λ
t→∞
then it is said that g has the quasi-asymptotics related to ρ.
~′ (Σ) is an FS space, the limit in Definition 4.4 is equivalent to

Since Λ
lim hg(tξ)/ρ(t), ϕ(ξ)i = hh, ϕi, h 6= 0

t→∞
~
for every ϕ ∈ Λ(Σ).
In a similar way, as for the quasi-asymptotics of distributions, one can
prove that ρ and h in Definition 4.4 have the following properties:
1) ρ has the form ρ(t) = tα L(t), α ∈ R and L is a slowly varying
function
2) h is homogeneous of degree α.
We will present an Abelian type theorem for the Laplace transform of
Fourier hyperfunctions (cf. [205]). But we have to define first the Laplace
transform of elements belonging to Λ~ ′ (Σ).
For a fixed z ∈ Rn + iB, where B is a bounded subset of Σ, ξ 7→ eizξ ∈
Λ(A, B) for every bounded set A and
keiz· kA
−sB = e
sA (x)
, z = x + iy .
~
Thus, for every fixed z ∈ Rn + iΣ, eiz· ∈ Λ(Σ) (cf [204]).
~ ′ (Σ), Lg, is defined by

Definition 4.5. The Laplace transform of g ∈ Λ
Lg(z) = hg(ξ), eizξ i, z ∈ Rn + iΣ .
In [205] it has been proved that the Laplace transform defines an isomor-
phism Λ ~′ (Σ) onto ←−
Λ (Σ). With this property and the mentioned properties
of the family of functions {eizξ ; z ∈ Rn + iΣ} it is easy to prove the
following Abelian type result.
Proposition 4.15. ([205]) Suppose that g, h ∈ Λ ~′ (Σ) and ρ(t) =

←
−
tα L(t), α ∈ R. If G = Lg and H = Lh, then G, H ∈ Λ (Σ). If
g(tξ)/ρ(t) → h(ξ), t → ∞, ~′ (Σ) ,

in Λ
then
←
−
G(z/t)/tn ρ(t) → H(z), t → ∞, in Λ (Σ).
In [205] one can find other properties of the quasi-asymptotics of the

Fourier hyperfunctions.
4.2.3 Integral transforms of Mellin convolution type
Now, we discuss Tauberian theorems for integral transforms which are of the
Mellin convolution type and whose kernels belong to suitable test function
spaces. The results are based on the Wiener–Tauberian Theorem 4.8 and
we apply then to the Laplace, Stieltjes, Weierstrass and Poisson transforms.
First, we recall some definitions and notions. By c and e, we denote the
same functions as in 4.1, defined by (4.4), (4.5) and by qδ as in Definition 4.2,
respectively, with the change of a by α and b by β. We introduce function
c∗ (x) = c(log x) and e∗ (x) = e(log x), x ∈ R+ . Denote by Eα,β (R+ ) the
space from Definition 4.2 and by D̃L1 (R+ ) the space of smooth functions
φ defined on R+ for which all the seminorms
Z
rk (φ) = |(Dx)k φ(x)|dx, k ∈ N0 ,
R+
d
are finite, where (Dx)φ(x) = (xφ(x)).
dx
Note that r0 is a norm. This sequence of seminorms defines the topo-
logical structure in D̃L1 (R+ ).
Recall that the classical Mellin transform of f ∈ L1loc (R+ ) is defined by
Z∞
M[f ](s) = f (x)xs−1 dx (4.43)
0
for those s ∈ C for which this integral exists (see [202]).

Suppose that k ∈ C ∞ (R+ ) and that the function: t 7→ (1/t)e∗ (t)k(x/t),
t ∈ R+ , belongs to D̃L1 (R+ ) for every x ∈ R+ . Let F ∈ D′ (R+ ) such that
F/e∗ ∈ D̃′ L1 (R+ ). Then, the Mellin convolution is defined by

∗ 1 ∗ x
(F ∗ k)(x) = F (t)/e (t), e (t)k
M t t

∗ 1 ∗ 1
= F (xt)/e (xt), e (xt)k , x ∈ R+ . (4.44)
t t
We refer to [202], Chapter 4, for this convolution. Note that several impor-
tant integral transforms of distributions are of this form.
Define
χ1 : C ∞ (R+ ) → C ∞ (R), φ(x) 7→ ey φ(ey ), x = ey , y ∈ R ,
1
χ2 : C ∞ (R) → C ∞ (R+ ), ψ(y) 7→ ψ(log x), y = log x, x ∈ R+ .
x
Clearly, they are inverse to each other and thus, they are bijections.
Proposition 4.16. The mapping χ1 is a topological isomorphism of D(R+ )

onto D(R), and of D̃L1 (R+ ) onto DL1 (R). Its inverse is χ2 .
Proof. We shall prove only that χ1 : D̃L1 (R+ ) → DL1 (R) is a topologi-
cal isomorphism. The following formulas can be simply proved:
k
X
(Dx)k φ(x) = ak,j xj φ(j) (x), x ∈ R+ , k ∈ N0 ,
j=0
p
X
e(p+1)y φ(p) (ey ) = bp,i (ey φ(ey ))(i) , y ∈ R, p ∈ N0 , φ ∈ C ∞ (R+ ) ,
i=0
where all the coefficients ak,j and bp,i are different from zero. This implies
Z∞ Xk Z∞
k
|(Dx) φ(x)|dx ≤ ck,j |(ey φ(ey ))(j) |dy, k ∈ N0 ,
0 j=0 −∞
and
Z∞ p
X Z∞
y y (p)
|(e φ(e )) |dy ≤ dp,i |(Dx)i φ(x)|dx, p ∈ N0 ,
−∞ i=0 0
where ck,j and dp,i are suitable positive constants. This implies that χ1 is
a topological isomorphism.
Proposition 4.13 and D(R) ֒→ DL1 (R) imply that D(R+ ) ֒→ D̃L1 (R+ ),
where ֒→ means that the left space is dense in the right one and that the
inclusion mapping is continuous. Clearly, S+ is a subspace of D̃L1 (R+ ),
and since it contains D(R+ ), it follows S+ ֒→ D̃L1 (R+ ). This implies that
D̃′ L1 (R+ ) ⊂ D′ (R+ ) and that all the elements of D̃′ L1 (R+ ) are tempered
distributions with the support contained in [0, ∞).
Proposition 4.17. (i) The mapping χ3 : φ(x) 7→ qδ (e−y )φ(e−y ), x =

e−y , y ∈ R, is a topological isomorphism of Eα+δ,β onto DL1 (R). (For
qδ cf. Definition 2).
(ii) If k ∈ Eα+δ,β (R+ ) and K(t) = k(et ), t ∈ R, then η Ǩe(α+δ) , (1 −
η)Ǩeβ· ∈ DL1 (R), where Ǩ denotes the function defined by Ǩ(x) = K(−x),
x ∈ R.
(iii) If k ∈ Eα+δ,β (R) and e is given by (4.5), then χ3 (k)e(· + h) ∈

DL1 (R) for every h ∈ R.
Proof. (i) One can prove by induction that for m ∈ N0 and φ ∈

Eα+δ,β (R+ ),
m
X
(qδ (e−t )φ(e−t ))(m) = ak,m e(α+δ−k)t φ(k) (e−t ), t > 1 ,
k=0
and
m
X
−t −t (m)
(qδ (e )φ(e )) = bk,m e(β−k)t φ(k) (e−t ), t < 0 .
k=0
This implies that there is a constant C > 0 such that

m
X
k(qδ (e−t )φ(e−t ))(m) kL1 ≤ C αi (φ),
i=0
where αi are seminorms in Eα+δ,β (R+ ) (see Definition 4.2).

Thus the mapping Eα+δ,β (R+ ) → DL1 (R) is continuous. Let us prove
the continuity of DL1 (R) → Eα+δ,β (R+ ).
Let ψ ∈ DL1 (R). Then, there exists φ ∈ Eα+δ,β (R+ ) such that
qδ (e−y )φ(e−y ) = ψ(y), y ∈ R. Indeed, we have to prove that φ(x) =
1
ψ(log x), x ∈ R+ , is in Eα+δ,β (R+ ). It is easy to prove that
qδ (x)
k
X
qδ (x)xk−1 φ(k) (x) = ci,k x−1 ψ (i) (log x) for x > 1/e.
i=0
This implies
Z∞ k
X Z∞
k−1 (k) dx
qδ (x)x |φ (x)|dx ≤ |ci,k | |ψ (i) (log x)|
i=0
x
1 1
k
X Z∞
≤ |ci,k | |ψ (i) (y)|dy .
t=0 0
In the same way, we can prove that

Z1 k
X
qδ (x)xk−1 |φ(k) (x)|dx < C kψ (i) kL1 .
0 i=0
This completes the proof of (i).

(ii) It follows by direct computation of DL1 -seminorms for given func-
tions.
(iii) Let i, j ∈ N0 , h ∈ R be fixed and x0 = max(1 − h, h). Then
x + h > 1 when x > x0 and x + h < 0 when x < −x0 .
Let K(t) = k(et ), t ∈ R. Using the estimate, for fixed h,
(
ci,j exp(αx + δx)|Ǩ (j) (x)|, x + h > 1
|e(i) (x + h)Ǩ (j) (x) ≤
ci,j exp(βx)|Ǩ (j) (x)|, x + h < 0,
proved in the first step of the proof of Theorem 4.8, it follows that e(· +
h)Ǩ ∈ DL1 (R).
Theorem 4.13. Let F ∈ D′ (R+ ), k ∈ C ∞ (R+ ) and α > β such that

n F (u·) o
(i)’ the set ; u ∈ R + is bounded in D′ (R+ ) and
c∗ (u)
(ii)’ the kernel k belongs to Eα+δ,β (R+ ) for some δ > 0.
Then,
a)’ M[x−α k(x)](iξ), ξ ∈ R, exists and
b)’ (F ∗ k), k ∈ R+ , exists, as well.
M
Moreover, if we assume that
(iii)’ M[x−α k(x)](iξ) 6= 0, ξ ∈ R, and
(iv)’ lim (F ∗ k)(x)/c∗ (x)
x→∞ M
Z
1 F (t) e∗ (t) x
= lim ∗
h ∗ , k( )i = A t−α−1 k(t)dt, A ∈ R ,
x→∞ c (x) e (t) t t
R+
then,
c)’ the Mellin convolution (F ∗ g)(x), x ∈ R+ , exists for every g ∈
M
Eα+δ,β (R+ ) and
(F ∗ g)(x) Z
M
→A t−α−1 g(t)dt, x → ∞ .
c∗ (x)
R+
Proof. We will show that the assumptions of theorem imply the as-
sumptions of Theorem 4.8. Then, we will prove the assertions using the
conclusions of Theorem 4.8.
Put K(t) = k(et ), t ∈ R. By Proposition 4.8 (ii), Assumption (ii)’
implies that
η Ǩe(α+δ)· , (1 − η)Ǩeβ· ∈ DL1 (R) .
Assumptions α > β and k ∈ Eα+δ,β (R+ ), imply that
Z∞
x−α k(x)x−iξ−1 dx
0
is finite for every ξ ∈ R. Thus, we proved a)’.
Let as prove the existence of F ∗ k.
M
Since k ∈ Eα+δ,β (R+ ), it follows that for every x ∈ R+ the function:
t 7→ (1/t)e∗ (t)k(x/t), t ∈ R+ belongs to D̃L1 (R+ ). Proposition 4.14 (iii)
implies e(· + h)Ǩ ∈ DL1 (R) for every h ∈ R.
Since
1
χ2 [e(· + h)Ǩ](t) = e(log t + log x)k(e− log t )
t

1 ∗ 1
= e (tx)k , t ∈ R+ , x = eh , h ∈ R ,
t t
it follows that for every x ∈ R+ , t 7→ (1/t)e∗ (tx)k(1/t), t ∈ R+ , belongs
to D̃L1 (R+ ).
Next, we will prove F/e∗ ∈ D̃′ L1 (R+ ). Proposition 4.16 implies that
1
φ 7→ hF (x), φ(log x)i, φ ∈ D(R),
x
defines a distribution f ∈ D′ (R). Assumption (i)’ and

f (t + h) F (xp) 1
, φ(t) = , φ(log x) , h = log p, p > 0 ,
c(h) c∗ (p) x
imply that for every φ ∈ D(R)

f (t + h)
, φ(t) ; h ∈ R
c(h)
is a bounded set of continuous functions on R. This implies that f /e ∈
B ′ (R). Let ϕ ∈ D(R+ ). We have

F (x) f (y) y y
, ϕ(x) = , e ϕ(e ) .
e∗ (x) e(y)
Proposition 4.13 implies that F/e∗ = (f /e) ◦ χ1 on DL1 (R+ ). Since D(R)
is dense in D̃L1 (R), we have F/e∗ = (f /e) ◦ χ1 on D̃L1 (R+ ). Hence, the
Mellin convolution F ∗ k exists and
M
(f ∗ K)(h) = (F ∗ k)(p), p = eh , h ∈ R. (4.45)

M
Thus, we have proved b)’. Note that assumption (iii)’ is equivalent to the
assumption F [K](ξ − iα) 6= 0, ξ ∈ R. By (4.45), we have that (iv)’ implies
(iv) in Theorem 4.8. Now, Theorem 4.8 c) and (4.45), with g instead of K,
imply assertion c)’.
Corollary of Theorem 4.13. If in Theorem 4.13, F ∈ S ′ + , A 6= 0
and α > β > −1, then the assumptions (i)’–(iv)’ imply that F has the
quasi-asymptotics related to c∗ with the limit AΓ(α + 1)fα+1 .
Proof. Formally, we have
(F ∗ g)(x)
M 1 1 x 1
= ∗ F (t), g = ∗ hF (tx), ψ(t)i,
c∗ (x) c (x) t t c (x)
1 1
where ψ(x) = g , x ∈ R+ .
x x
Let δ > 0. If we prove that for every ψ ∈ S+ there exists g ∈ Eα+δ,β (R+ )
such that x−1 g(x−1 ) = ψ(x), x ∈ R+ , the above formal calculation holds
(since F ∈ S ′ + ) and the assertion of the corollary follows from c)’ because
1
it implies that for every ψ ∈ S+ , lim ∗ hF (tx), ψ(t)i exists.
x→∞ c (x)
So, we have to prove that g(x) = (1/x)ψ(1/x), x ∈ R, belongs to

Eα+δ,β (R+ ). Since for k ∈ N and x ∈ R+ ,
k
X
(k) 1 1 (i) 1
g (x) = ai,k i ψ ; ai,k ∈ R, k ∈ N0 ,
xk+1 i=0
x x
where ai,k are suitable numbers, we have

k
X
k−1 (k) −2−i−α−δ (i)
1
|qδ (x)x g (x)|≤ |ai,k |x ψ < C, 0 < x < e−1 ,
i=0
x
and
k
X 1
k−1 (k) −β−i−2 (i)
|qδ (x)x g (x)|≤ |ai,k |x ψ ≤ Cx−β−2 , x > 1.
i=0
x
This proves that g ∈ Eα+δ,β (R+ ), if α > β > −1.

Let {fα ; α ∈ R} be the family of distributions given in 0.4 and F ∈ S ′ + .

Let F (−α) = fα ∗ F, α ∈ R. If ϕ ∈ S+ , and m ∈ N, then
u−m hF (−m) (uξ), ϕ(ξ))i = hF (uξ), hfm (t), ϕ(ξ + t)ii, (4.46)
where the mapping ϕ 7→ hfm (t), ϕ(· + t)i is an automorphism on S+ (cf.

§3 in [192]).
The following theorem is formulated for future use.
Theorem 4.14. Let f ∈ S ′ + , m ∈ N0 and α > β. Assume:

f (·u)
(i”) The set ; u ∈ R+ is bounded in S ′ + .
u−m c∗ (u)
(ii”) The kernel km belongs to Eα+δ,β (R) for some δ > 0.
Then,
a”) M[x−α km (x)](iξ), ξ ∈ R exists and
b”) (f (−m) ∗M km )(x), x ∈ R+ , exists, as well.
Moreover, if
(iii”) M[x−α km (x)](iξ) 6= 0, ξ ∈ R, and
R 1 1
(iv”) lim (f (−m) ∗M km )(x)/c∗ (x) = A(m) km ( )xα dx,
x→∞ R+ x x
where A(m) ∈ R, then:
c”) For every g ∈ Eα+δ,β , the Mellin convolution (f (−m) ∗M g)(x), x ∈
R+ , exists and
Z
(f (−m) ∗M g)(x) 1 1 α
d”) lim = A(m) g( )t dt.
x→∞ c∗ (x) t t
R+
q
In particular, if β > −1 and A(m) 6= 0, then F ∼ A(m)Γ(α + 1)fα+1−m
related to u−m c∗ (u).

f (−m) (u·)
Proof. If f satisfies (i”), then (4.46) implies that ; u ∈ R+
c∗ (u)
is a bounded set in D′ (R+ ) as well. Thus, f (−m) satisfies condition (i’) in
Theorem 4.8. Applying Theorem 4.8 to f (−m) and km , we obtain a”), b”),
c”) and d”) in Theorem 4.13. By the corollary of Theorem 4.13 it follows
that
q
f (−m) ∼ A(m)Γ(α + 1)fα+1 related to c∗ (u) .
Theorem 2.1 implies

q
f ∼ Am Γ(α + 1)fα+1−m related to u−m c∗ (u).
Theorem 4.7 and the change of variables t = ex , x ∈ R, imply the following

theorem.
Theorem 4.15. Let F, k ∈ L1loc (R+ ) and α > β, δ > 0, be such that
i) F/c∗ ∈ L∞ (R+ ),
ii) k(t)t−(α+δ+1) ∈ L1 ((0, 1)) and k(t)t−(β+1) ∈ L1 ((1, ∞)).
Then, there exist
Z∞
x dt
(F ∗ k)(x) = F (t)k , x ∈ R+ ,
M t t
0
−α
and M[x k(x)](iξ), ξ ∈ R. If
iii) M[x−α k(x)](iξ) 6= 0, ξ ∈ R, and
R∞
iv) lim (F ∗ k)(x)/c∗ (x) = A K(t)t−α−1 dt,
x→∞ M 0
then
Z∞
∗
lim (F ∗ ψ)(x)/c (x) = A ψ(t)t−α−1 dt, A ∈ R,
x→∞ M
0
for every ψ ∈ L1loc (R+ ) such that ψ(t)t−(α+δ+1) ∈ L1 ((0, 1)) and
ψ(t)t−β−1 ∈ L1 ((1, ∞)) hold. In particular, if β > −1, then
Zx
A α+1
F (t)dt ∼ x L(x), x → ∞.
α+1
0
Remark. The last relation and Theorem 1.7.5 in [9], imply that F (x) ∼
α
Ax L(x), x → ∞.
In the context of Theorem 4.17 see also [76] and Theorem 4.8.3 in [9].
4.2.4 Special integral transforms
We shall discuss some special cases of integral transforms.

As we mentioned in 4.1.2, Zemanian, introduced in [202] the spaces La,b

and Ma,b in order to define the Laplace and Mellin integral transforms of
generalized functions.
If the kernel Ǩ belongs to La,b , then it satisfies condition (ii) of Theorem
4.8 for a > α > β > b. This follows from
(η(x) exp((α + δ)x)Ǩ(x))(m)

X m
m
= (α + δ)m−j e−(a−α−δ)x eax Ǩ (i) (x), x > ω ,
i=0
i
and
((1 − η(x)) exp(βx)Ǩ(x))m

Xm
m m−j (β−b)x bx (i)
= β e e Ǩ (x), x < −ω ,
i=0
i
where 0 < δ < a − α, ω is a positive number large enough and η ∈

C ∞ , η(x) = 1 if x ∈ (1, ∞) and η(x) = 0 if x ∈ (−∞, −1).
In such a way Theorem 4.8 can be applied to all transforms of convolu-
tion type whose kernels belong to La,b with a > α > β > b, 0 < δ < a − α.
If ψ ∈ Ma,b and a > α + 1 > β + 1 > b, 0 < δ < a − α − 1, then
ψ ∈ Eα+δ,β (R+ ). This follows from
|xk−1 qδ (x)ψ (k) (x)| ≤ x−(α−a+δ+2) |xk+1−a ψ (k) (x)|

≤ C1 x−(α−a+δ+2) , 0 < x < e−1 ,
|xk−1 qδ (x)ψ (k) (x)| ≤ x−(β−b+2) |xk+1−b ψ (k) (x)|

≤ C2 x−(β−b+2) , x > e−1 .
Hence, Theorem 4.14 can be applied to transforms which are of the

Mellin convolution type with kernels in Ma,b , a > α + 1 > β + 1 > b, 0 <
δ < a − α − 1.
Laplace transform. The most precise Tauberian theorem for the
Laplace transform of distributions can be found in [192]. Tauberian theo-
rems concerning positive measures are also considered in [149].
Recall, the Laplace transform of an S ∈ S ′ + , is defined by
L[S](z) = hS(t), eizt i, z = x + iy ∈ R + iR+ .

Theorem 4.16. Let f ∈ S ′ + , m ∈ N0 , α > β > −1. Assume:

f (u·)
a) The set ; u ∈ R+ is bounded in S ′ + .
u−m c∗ (u)
Z∞
y
b) lim m ∗ 1 hf (t), e−yt i = A e−t tα dt, A ∈ R .
y→0+ y c ( )
y
0
q ∗
Then, f (kx) ∼ c (k)AΓ(α + 1)fα−m+1 (x), k → ∞.
Proof. If y > 0, then e−y ∈ S+ and

hf (t), e−yt i = y m hf (−m) (t), e−yt i, y > 0 .
Put km (t) = t−1 exp(−1/t), t ∈ R+ . Then km ∈ Eα+δ,β (R+ ) for any
α > β, β > −1 and δ > 0. Therefore,

1
(f (−m) ∗M km ) = y −m+1 hf (t), e−yt i
y
and
Z∞
−α
M(t km (t))(iξ) = e−t tα+iξ dt = Γ(α + 1 + iξ) 6= 0, ξ ∈ R, α + 1 > 0 .
0
Now, Theorem 4.16 follows from Theorem 4.14 (with k = km ).

Remark. In [192] Theorem 2, Section 7, the authors proved the as-
sertion of Theorem 4.16 under the assumptions b) and, instead of a), the
following one:
There are M > 0, σ > 0 and y0 > 0 such that
n−m+1
y
−yt σ
c∗ ( 1 ) hf (t), e i ≤ M n! n , 0 < y ≤ y0 , n ∈ N. (4.47)
y
Since a) implies (4.47), (4.48) and b) is equivalent to

q
f (kx) ∼ c∗ (k)AΓ(α + 1)fα−m+1 (x), k → ∞ ,
it follows that conditions a) and b) are equivalent to (4.47) and b).
Stieltjes transform. We shall use the same notation as in 4.1. Refor-
mulate (4.20)in another form
sr+m+1 D t −r−m−1 E
Sr [f ](s) = f (−m) (t), +1 , s ∈ R+ . (4.48)
(r + 1)m s
The next theorem follows from Theorem 4.14.
Theorem 4.17. Let f ∈ S ′ + , m ∈ N0 , r + m + 1 > α + 1 > β + 1 > 0

and δ > 0. Assume:

f (u·)
a) ; u ∈ R+
u−m c∗ (u)
is bounded in S ′ + . Then, Sr [f ](u), u ∈ R+ , exists. If moreover,
1
b) lim Sr [f ](u) = Cm , Cm ∈ R,
u→∞ u−(r+m) c∗ (u)
then for every g ∈ Eα+δ,β (R+ ),
(−m)
f (ξu) 1 1 Γ(r + 1) 1 1
lim ∗
, g = Cm f α+1 (ξ), g .
u→∞ c (u) ξ ξ Γ(r + m − α) ξ ξ
(4.49)
Hence,
q Γ(r + 1)
f (−m) (ux) ∼ c∗ (u)Cm fα+1 (x), u → ∞ (4.50)
Γ(r + m − α)
and
q
f (ux) ∼ u−m c∗ (u)Cm (Γ(r + 1)/Γ(r + m − α))fα+1−m (x), u → ∞ . (4.51)
Proof. Let F = f (−m) and

−(r+m+1)
1 1
km (t) = +1 , t ∈ R+ .
t t
Since r + m + 1 > α + 1 > β + 1 > 0 and

Xp −(r+m+i+1)
p−1 (p) −i−2 1
qδ (x)x km (x) = ai,p qδ (x)x +1 , x ∈ R+ ,
i=0
x
it follows that km ∈ Eα+δ,β (R+ ). Note that t−1 km (t−t ) = (t + 1)−(r+m+1) ,

t ∈ R+ . This and (4.48) imply.
(f (−m) ∗M km )(u) = hf (−m) (ξu), (ξ + 1)−(r+m+1) i

−(r+m+1)
1 (−m) t
= hf (t), +1 i
u u
ur+m
= Sr [f (−m) ](u), u ∈ R+ . (4.52)
(r + 1)m
We shall show that conditions of Theorem 4.14 hold for f and k = km . In

fact, (i”) is a) and (iii”) follows from
Γ(α + 1 + iξ)Γ(r + m − α − iξ)
M[x−α km (x)](iξ) = 6= 0, ξ ∈ R .
Γ(r + m + 1)
Condition b) and (4.52) imply (iv”) of Theorem 4.14
(f (−m) ∗M km )(u) ur+m
lim = lim Sr [f ](u)
u→∞ c∗ (u) u→∞ (r + 1)m c∗ (u)
Z∞ Z∞
Cm tα 1 1 α
= = Am dt = Am km x dx ,
(r + 1)m (t + 1)r+m+1 x k
0 0
Cm Γ(r + 1)
where Am = .
Γ(r + m − α)
Thus, Theorem 4.14 implies (4.49) but (4.50) follows from S+ ⊂
Eα+δ,β (R+ ). As in the proof of Theorem 4.14, we have that the quasi-
asymptotic behavior of f (−m) imply the appropriate quasi-asymptotic be-
havior of (f (−m) )(m) = f and this is given in (4.51).
Weierstrass transform. As earlier we will use the Kernel
1
k(s, t) = exp(−s2 /4t), s ∈ C, t > 0 .
(4πt)1/2
The Weierstrass transform, Wt (f ), t > 0, of f ∈ K′ 1 is defined by
Wt [f ](s) = hf (x), k(s − x, t)i = (f ∗ k(·, t))(s), s ∈ C, |args|< π/4 .
A Tauberian type theorem for this transform, in the case when c(h) ≡ 1,
is given in [108]. The next one is more general.
n f (· + h) o
Theorem 4.18. Let f ∈ D′ (R), α > β, and the set ; h ∈ R
c(h)
be bounded in D′ (R). Then, Wt (f )(s) exists for s ∈ C, |args|≤ π/4 and
t > 0. Moreover, if we assume
Z∞
1 A(t0 )
lim Wt0 [f ](h) = exp(−x2 /4t0 ) exp(−αx)dx
h→∞ c(h) (4πt0 )1/2
−∞
= A(t0 ) exp(α2 t0 ) = B(t0 ) ,

then
f (x + h)
lim , ψ(x) = B(t0 ) exp(−α2 t0 )heαx , ψ(x)i (4.53)
h→∞ c(h)
for every ψ ∈ C ∞ (R) such that η ψ̌ exp((α+δ)·) and (1−η)ψ̌ exp(β·) belong
to DL1 (R).
The proof follows by showing that the conditions of Theorem 4.8 are
satisfied. We omit the details.
Remark. Of A(t0 ) 6= 0, then (4.36) implies that f has the S-
asymptotics related to c with the limit B(t0 ) exp(−α2 t0 ) exp(αx) in D′ (R)
and in K′ 1 (R).
Poisson transform. The Poisson transform of a distribution f ∈
D′ (R) is defined by

y y
P[f ](x, y) = f (t) ∗ 2 (x) = f (t), , y > 0, x ∈ R .
t + y2 (x − t)2 + y 2
The equality
Z∞
y
e−iξt dt = e−|yξ| , y > 0, ξ ∈ R,
t2 + y 2
−∞
and Theorem 4.13 imply:
Theorem 4.19. Let f ∈ B ′ (R). If for some y > 0

Z∞
y
lim P[f ](x, y) = a dt, a ∈ R,
x→∞ t + y2
2
−∞
then
lim hf (t + x), ψ(t)i = ha, ψ(t)i
x→∞
for every ψ ∈ DL1 (R).
Remark. If a 6= 0, then the last equality means that f has S-

asymptotics in B ′ (R) related to c = 1.
4.2.5 Localization of Tauberian type theorems
In the Tauberian type theorems for generalized functions Vladimirov,

Drozhzhinov, and Zavyalov ([192]) have used results on functions holo-
morphic in some tubular domains over cones. In [36] it is proved that such
results can remain valid in the local variant, as well. Authors elaborated
a technique which can be applied reconstruct asymptotic properties of a
holomorphic function from the asymptotic behavior if its real part, a theo-
rem on the non-compensation of singularities of holomorphic functions and
a multidimensional theorem of Lindelöf type.
The basic idea lies in the following theorem.
Theorem 4.20. [36]. Assume that C1 and C2 are open, strictly convex
and acute cones in Rn with their vertices at the origin and such that
C1 ∩ C2 = C 6= 0, ch(C1 ∪ C2 ) = Rn .
Let f1 (z) and f2 (z) be functions holomorphic in local tubular domains over
the cones C1 and C2 , respectively. Consider the sum of these two functions
φ(z) = f1 (z) + f2 (z).
It is holomorphic in the local tubular domain over the cone C. Assume
that for some α > 0 there are constants M, N and r0 such that
M

φ(z) ≤ α for z ∈ T C , |z|< r0 ,
|z| ϕN
where ϕ is the angle between the vector z and the boundary of the tubular
domain T C . Then there are constants M1 , N1 such that
M1

f1 (z) ≤ α N1 , |z|< r0 ,
|z| ϕ1
where ϕ1 is the angle between the vector z and the boundary of the tubu-
lar domain over the cone C1 . A similar estimate is valid for f2 (z) with
corresponding M2 , N2 and ϕ2 .
In Theorem 4 of [36] is given a local variant of Theorem 4.16. First it

was defined the space Λ(Γ) of type K{Mp } with a weight Mp (t) (cf. [63],
V.2) which is complete, perfect and nuclear countably normed space.
Λ′ (Γ) is the corresponding space of continuous linear functionals and
Γ is a closed, convex acute, solid cone in Rn with vertex at the origin;
int Γ∗ = C. Let Cq,r = {y ∈ C; |y|2 < q∆C (y), |y|< r} and let H(T Cq,r ) be
the class of functions g(z),
|g(x + iy)|≤ M (1 + |x|)a /∆bCq,r (y) for z ∈ T Cq,r
for some a, b and M.

We recall also
Theorem 4.21. (cf. [36], Theorem 4). Assume that Lf ∈ H(T Cq,r ) (L
denote the Laplace transform) and let ρ(k) be a regularly varying function
of order α. Assume that the following conditions are fulfilled:
5. Summability of Fourier series and integrals 237
a) there is a domain Ω ⊂ C such that

1 iy
lim Lf = Lg(iy), y ∈ Ω ;
k→∞ k n ρ(k) k
b) there are y0 ∈ prC and constants M, N and r1 such that

N
ρ(1/|z|) z
Lf (z) ≤ M if z = x + iτ y0 , τ > 0, |z|< r1 .
|z|n y
Then the spectral function f (t) for Lf (z) has quasi-asymptotics relative to
ρ(k) in Λ′ (Γ). In other words
1
lim f (kt) = g(t) in Λ′ (Γ).
k→∞ ρ(k)
For the asymptotics and integral transforms one can also consult the
following papers: [117], [122] and [188].
5 Summability of Fourier series and integrals
In this part we discuss applications of the quasi-asymptotic behavior of

distributions to the study of summability of one-dimensional Fourier series
and integrals.
The study of the relationship between the local behavior of a periodic
function and the convergence or summability of its Fourier series is an old
and interesting problem. It has a long tradition [68], [206]. Since conver-
gence fails in many interesting cases, one is led to use summability methods
rather than ordinary convergence. In the case of periodic distributions, that
is, periodic elements of the space S ′ (R), the local problems have classically
focused around the notion of the value of a distribution at a point in the
sense of Lojasiewicz [93], [94] (cf. 5.2). A pioneer in this direction was
G. Walter [195], [196] who study Cesàro and Abel summability of Fourier
series under the presence of Lojasiewicz point values. Surprisingly, it is pos-
sible to characterize Lojasiewicz point values in terms of the summability
of the Fourier series; a complete characterization in terms of asymmetric
Cesàro limits of partial sums was obtained first by R. Estrada in [45]. The
situation with Fourier integrals is similar to that of series, even in the clas-
sical cases one needs to use summability methods [170]. Recent studies
[49], [57], [58], [172], [173], [175], [176], [178], [181] and [182] have shown
the deep connection between quasi-asymptotics and such problems, and it
is exactly the approach that we will follow in this part of the book.
The first section 5.1 is of preliminary character. We start by discussing
several summability procedures for series and integrals, and we then extend
them to summability for distributional evaluations. We follow closely the
expositions from [47], [56], [66] and [173].
The main section is 5.2 Let us state the problem to be considered. Fix
the constants in the Fourier transform,
Z ∞
φ̂(t) = e−itx φ(x)dx,
−∞
for φ ∈ S(R), so that the Fourier inversion formula is

Z ∞
1
φ(x) = eixt φ̂(t)dt. (5.1)
2π −∞
The first question that we should address is that of giving pointwise sense
to (5.1) for very general tempered distributions, that is, if a distribution has
a value at a point, we show that (5.1) holds pointwise for several summabil-
ity methods. The second important problem that we want to consider is to
find a summability method so that the pointwise Fourier inversion formula,
interpreted with such a summability procedure, becomes a full characteri-
zation of Lojasiewicz point values. For this goals, we follow the results from
[175], [176]; however, we present a simplification of the proofs based on the
results from 2.10. It is interesting to mention that the summability method
which provides the right characterization is nothing but a structural char-
acterization of a quasi-asymptotics, and it was actually the precedent to the
structural theorems from 2.10. In the remaining subsections, we discuss
other related problems. Finally, we should mention that we do not discuss
problems about summability order in our formulas; we refer to [181] for a
complete study in that direction.
5.1 The Cesàro behavior
It is the intension of this section to introduce two methods of summability

for distributional evaluations. We are only interested in the one-dimensional
case; for the multidimensional case we refer to [173] and [182].
We start by presenting a very brief introduction to summability of di-

vergent series and integrals. It will serve as a motivation to the study of
more general notions applicable to Schwartz distribution. There is a very
rich and extensive literature on this traditional subject; for instance, the
reader is referred to the classical and beautiful book of Hardy [66]. See also
[170], [206] for connections with Fourier series and integrals.
We will then discuss the Cesàro behavior of distributions and some basic
properties of this concept, it will be the base to define limits of distributions
and distributional evaluations in the Cesàro sense. We follow the approach
from [47] and [56], where we refer for a more complete account. In 5.1.2,
we shall confine ourselves with the definition for integral Cesàro orders and
comparison with respect to power functions. We point out that the Cesàro
behavior of distributions can also be defined for fractional orders [181], [183]
and [184], in addition, regularly varying functions may be included in the
theory [183].
5.1.1 Cesàro, Riesz, and Abel summability of series and integral
We shall discuss the summability methods by Abel, Cesàro and Riesz means
for series and integrals.
Let us start with Cesàro summability . In general we say that a numer-
P∞
ical series n=0 cn , possibly divergent, is summable to a complex number
γ in the average, or Cesàro, sense of order 1, if the averages of its partial
sums converge to γ, that is,
s0 + s1 + s2 + · · · + sn
lim = γ, (5.2)
n→∞ n+1
Pn
where sn = j=0 cj , in such a case one writes
∞
X
cn = γ (C, 1). (5.3)
n=0
It is elementary to check that if the series is convergent, then it is summable
by the (C, 1) method, but the converse is naturally false. For example,
P∞
one may take n=0 (−1)n , which is evidently divergent; but its average
P∞
converges to 1/2, hence n=0 (−1)n = 1/2 (C, 1).
The Cesàro method of summability is important in the analysis of sev-
eral series expansions of functions and generalized functions; in particular
for Fourier series. In fact, it is a famous result of Féjer that the Fourier se-
ries of a continuous function, although not necessarily convergent, is (C, 1)
summable to the value of the function at any point [206]. Furthermore, Kol-
mogorov proved [206] that there are functions in the class L1 [0, 2π] whose
Fourier series diverge everywhere; therefore, even in the case of classical
functions, it is imperative the use of summability methods for the point-
wise analysis of trigonometric series. In 5.2.1, we will generalize Féjer’s
classical result to include periodic distributions, for that we will use higher
order Cesàro means.
We can extend the (C, 1) to higher order average means. There are
several approaches, and all of them are equivalent. Perhaps the simplest,
but analytically inadequate, is that of Hölder means. We can define recur-
Pn Pn
sively the sequences, skn := ( j=0 sk−1
j )/(n + 1), with s0n := sn = j=0 cj .
Then, we call skn the Hölder means of order k of the series, and say that
P∞ k
n=0 cn = γ (H, k), if sn → γ as n → ∞. As we remarked before, Hölder
means present serious difficulties associated with their analytical manipu-
lation [66], we shall therefore avoid their use in the future.
Another approach to the extension of (5.3) is via higher order Cesàro
P∞
means. Given a series n=0 cn we define its Cesàro means of order β,
β > −1, by
n
Γ(β + 1) X β + j
Cnβ = cn−j , (5.4)
nβ j=0
β
then we say that the series is Cesàro summable of order β to γ, and write
P∞
c = γ (C, β), if Cnβ → γ as n → ∞. An interesting example
P∞ n
n=0
n α
is n=0 (−1) n , α > −1, which is (C, β) summable whenever β > α,
oscillates finitely when β = α, and oscillates infinitely for β < α; we refer
to [66] for a proof of this fact.
We shall also discuss the method of Marcel Riesz by typical means [66].
Actually, the Riesz method will be the most important for us in the sub-
∞
sequent sections. Let hλn in=0 be an increasing sequence of non-negative
numbers such that λn → ∞ as n → ∞. We say that a series is summable
by the Riesz means, with respect to hλn i, of order β ≥ 0 if
X β
λn
lim cn 1 − = γ; (5.5)
x→∞ x
0≤λn <x
and then we write

∞
X
cn = γ (R, hλn i , β). (5.6)
n=0
These three methods of summability can be compared. If β = mN, then

the (H, m) and the (C, m) methods are equivalent [66]. While if β ≥ 0
and λn = n, the (C, β) and the (R, hni , β) methods sum the same series
to the same value, and so they are also equivalent [66]. Here the use of
a continuous variable in (5.5) is absolutely necessary for the equivalence
[66]. The Riesz method has an advantage over the other two methods, it is
easily generalizable to integrals, even to distributions as we shall see later in
5.1.3. Therefore, we advise the reader that whenever we talk about Cesàro
summability, even if we write (C, β), the means should be thought as Riesz
means.
Let now f be a locally integrable function supported in [0, ∞). Let
β > 0. We write
lim f (x) = γ (C, β) , (5.7)
x→∞
if
Z β−1
β x t
lim f (t) 1 − dt = γ . (5.8)
x→∞ x 0 x
Note that (5.8) basically says that f (−β) (x), the β-primitive of f , is asymp-
totic to γxβ /Γ(β + 1) as x → ∞. The last approach will allows us to
consider Cesàro limits of distributions in the future. Suppose that f is a
function of local bounded variation, then its distributional derivative is a
Radon measure, a continuous linear functional over the space of continuous
functions with compact support, say f ′ = µ. Hence integration by parts in
(5.8) shows that it is equivalent to
Z x β
t
lim 1− dµ(t) = γ . (5.9)
x→∞ 0 x
The latter can be taken as Z the definition of the relation
∞
dµ(x) = γ (C, β) . (5.10)
0
Observe that (5.6) holds if and only (5.10) holds for the Radon measure
P
µ= ∞ n=0 cn δ( · − λn ).
We end this discussion by considering Abel summability of series [66].

P∞
For a series n=0 cn , we consider its Abel means, that is, the power series
P∞ n
P∞ n
n=0 cn r . We say that the series is Abel summable to γ, if n=0 cn r
is convergent for |r| < 1 and the power series approaches to the limit γ at
the boundary point r = 1, i.e.,
∞
X
lim cn r n = γ , (5.11)
r→1−
n=0
we write
∞
X
cn = γ (A). (5.12)
n=0
It will be more convenient for us to write r = e−y , so that the power

series becomes a Dirichlet series. So, we have a natural extension for mea-
R ∞ supported in [0, ∞) in terms of the Laplace transform. We say that
sures
0 dµ(x) is Abel summable to γ and write
Z ∞
dµ(x) = γ (A), (5.13)
0
R∞
if for any y > 0 the integral 0 e−yt dµ(t) exists as an improper integral,
and
Z ∞
lim e−yt dµ(t) = γ . (5.14)
y→0+ 0
P∞
When the Radon measure is given by n=0 cn δ(x − λn ), we write
∞
X
cn = γ (A, hλn i), (5.15)
n=0
P∞ −yλn
if (5.14) holds, that is, if the Dirichlet series n=0 cn e is convergent
for y > 0 and it tends to γ as y → 0+ .
We finally comment some inclusion between the Cesàro and Abel
method of summation, if (5.10) holds then (5.13) is satisfied, this is shown
below (Corollary 5.1). In the case of power series this fact is the well known
Abel’s theorem [66]. Naturally, the converse is not true. The reader may
wish to verify that the series whose coefficients are given by those of the
1 P∞ n
power series e 1−r = n=0 cn r is an explicit example of a series which
is (A) summable but not (C, β) summable [66], no matter what value of
β be taken. Furthermore, in [48], it is constructed a series which is Abel
summable with coefficients cn = O(nm ), but it is not (C, β) summable
for any β. The study of additional hypotheses to ensure the converse of
Abel’s theorem motivated the beginning of the Tauberian theory. For in-
stance, Littlewood Tauberian condition cn = O(1/n) together with Abel
summability imply the convergence of the series [66]. We will obtain a sim-
ple and quick proof of Littlewood’s theorem in Theorem 5.17, as a direct
consequence of the use of generalized asymptotic behavior.
5.1.2 The Cesàro behavior of distributions
Let us define the Cesàro behavior of distributions at infinity . It is studied

by using the order symbols O (xα ) and o (xα ) in the Cesàro sense.
Definition 5.1. Let f ∈ D′ (R), m ∈ N0 , and α ∈ R \ {−1, −2, −3, . . . }.

We say that f (x) = O (xα ) as x → ∞ in the Cesàro sense of order m (in
the (C, m) sense) and write
f (x) = O (xα ) (C, m) , x → ∞ , (5.16)
if each primitive F of order m, i.e., F (m) = f , is an ordinary function for
large arguments and satisfies the ordinary order relation

F (x) = p(x) + O xα+m , x → ∞ , (5.17)
for some suitable polynomial p of degree at most m − 1. Similarly for the
little o symbol. We say that f is asymptotic to Cxα as x → ∞ in the
Cesàro sense of order m and write
f (x) ∼ Cxα
+ (C, m), x → ∞ , (5.18)
if we have f (x) − Cxα α
+ = o(x ) (C, m), x → ∞.
Notice that if α > −1, then the polynomial p is irrelevant in (5.17). A

similar definition applies when x → −∞. One may also consider the case
when α = −1, −2, −3, . . . [56] , but we shall not do so. Obviously, if f
vanishes for large arguments, then f (x) = o(xα ) (C, m), for any m and α.
When we do not want to make reference to the order m in (5.16) or (5.18),
we simply write (C), meaning (C, m) for some m.
For α = 0, we obtain the notion of Cesàro limits at infinity.
Definition 5.2. Let f ∈ D′ (R) and m ∈ N0 . We say that f has a limit γ

at infinity in the Cesàro sense of order m (in the (C, m) sense) and write
lim f (x) = γ (C, m) ,
x→∞
if we have that f (x) = γ + o(1) (C, m), x → ∞.
We may also consider Cesàro limits as x → −∞. We will be mainly

concerned with limits and not asymptotics in the Cesàro; however, we
want discuss the close relation between Cesàro asymptotics and the quasi-
asymptotic behavior. For further properties, we refer to [56].
The next theorem shows that the Cesàro behavior, in the case α > −1,
is totally determined by the quasi-asymptotic properties of the distribution
on intervals being bounded at the left.
Proposition 5.1. Let f ∈ D′ (R), m ∈ N0 , and α > −1. Let f+ be any

distribution supported on an interval of the form [a, ∞), a ∈ R, coinciding
with f for large arguments, i.e., in some open interval with finite left end
point. Then, we have the next equivalences.
(i) The following two conditions are equivalent,
f (x) = O (xα ) (C), x → ∞, (5.19)
and f+ belongs to S ′ (R) and is quasi-asymptotically bounded of

degree α, i.e.,
f+ (λx) = O(λα ) as λ → ∞ in S ′ (R). (5.20)
(ii) The conditions,
f (x) ∼ Cxα
+ (C), x → ∞, (5.21)
and f+ ∈ S ′ (R) has the quasi-asymptotic behavior

q
f+ (λx) ∼ Cλα xα
+ as λ → ∞ in S ′ (R), (5.22)
are equivalent.
Proof. We can assume that f = f+ , and so the equivalence between

(5.19) and (5.20) is precisely the structural theorem for quasi-asymptotic
boundedness, see Section 2.12 On the other hand, the equivalence between
(5.21) and (5.22) is precisely the content of the structural theorem for quasi-
asymptotic behavior of degree α > −1 (cf. Theorem 2.2 in 2.2 and Theorem
2.26 in 2.10).
When α < −1, we do not exactly obtain a characterization in terms of
quasi-asymptotics because delta terms could appear in the expansion.
Proposition 5.2. Let f ∈ D′ (R), m ∈ N0 , and α < −1, −α ∈ / N. Let

f+ be any distribution supported on an interval of the form [a, ∞), a ∈ R,
coinciding with f for large arguments. Then, we have the next equivalences.
(i) The following two conditions are equivalent,

f (x) = O xα+ (C), x → ∞, (5.23)
and there exist n > −α constants a0 , . . . , an−1 , in general depend-

ing on f+ , such that f+ has the quasi-asymptotic expansion
n−1
X δ (j) (x)
f+ (λx) = aj + O(λα ) as λ → ∞ in S ′ (R). (5.24)
j=0
λj+1
(ii) The conditions,
f (x) ∼ Cxα
+ (C) , x → ∞, (5.25)
and the existence of constants n > −α constants a0 , . . . , an−1 , in

general depending on f+ , such that
n−1
X δ (j) (x)
f+ (λx) = Cλα xα
++ aj + o(λα ) as λ → ∞ in S ′ (R),
j=0
λj+1
(5.26)
are equivalent.
Proof. We can assume f = f+ . We only show (ii), the proof of (i) is

similar to this case and is left to the reader. Assume (5.25), then there
exist G1 , G2 , m > −α − 1, and m constants c0 , . . . , cm−1 such that f =
(m) (m)
G1 + G2 , G1 has compact support, G2 is a locally integral functions
with support on [0, ∞), and
m−1
X xj Γ(α + 1)
G2 (x) = cj +C xm+α + o(xm+α ), x → ∞ .
j=0
j! Γ(m + α + 1)
Since G1 has compact support, then G1 (λx) = O(λ−1 ), in S ′ (R), and so

G(m) (λx) = O(λ−m−1 ) = o(λα ); then, since it does not contribute for
(5.26), we can assume that G1 = 0. On the other hand, the ordinary
asymptotic expansion of G2 implies
m−1
X (λx)j+ Γ(α + 1)
G2 (λx) = cj +C (λx)m+α
+ + o(λm+α )
j=0
j! Γ(m + α + 1)
in S ′ (R). Differentiating m-times the above asymptotic formula, and dis-

carding the irrelevant constants, we obtain (5.26) with aj = cm−1−j . The
converse follows from the structural theorem for quasi-asymptotics (cf. The-
Pn−1
orem 2.2) applied to f+ − j=0 aj δ (j) .
5.1.3 Summability of distributional evaluations
We now study two methods of summability for distributional evaluations,

the two-sided Cesàro method, and Abel summability. Two more methods
will be introduced in 5.2 (Definitions 5.6 and 5.8).
We start with summability in the Cesàro sense. First we assume that
our distributions have support bounded at the left. Recall that H denotes
the Heaviside function, i.e., the characteristic function of (0, ∞).
Definition 5.3. Let f ∈ D′ (R) have support bounded at the left. Let
φ ∈ E(R) and m ∈ N0 . We say the evaluation hf (x), φ(x)i has a value γ
in the Cesàro sense of order m, and write
hf (x), φ(x)i = γ (C, m) (5.27)
if the first order primitive F = (φf )(−1) = (φf )∗H, the first order primitive
of φf with support bounded at the left, satisfies limx→∞ F (x) = γ (C, m).
Example
R∞ II.1. Let µ be a Radon measure with support on [0, ∞). Then
0 dµ(x) = γ (C, m) if and only if hµ(x), 1i = γ (C, m). In particular
∞
X
cn = γ (R, hλn i , m)
n=0
if and only if
* ∞
+
X
cn δ(x − λn ), 1 =γ (C, m).
n=0
If f has support bounded at the right then we say that hf (x), φ(x)i (C)
exists if and only if hf (−x), φ(−x)i = γ (C) exists and we define
hf (x), φ(x)i = γ (C).
The distributional evaluations with respect to compactly supported dis-
tribution can always be computed in the (C) sense, actually with order
m = 0.
Lemma 5.1. Let f ∈ E ′ (R) and φ ∈ E(R). Then hf (x), φ(x)i (C, 0) always
exists.
Proof. We can assume that φ ≡ 1. Consider f (−1) , it is obviously con-

stant for large arguments, we must show that it satisfies f (−1) = hf (x), 1i
(a constant distribution) on certain interval (a, ∞). Decompose f (−1) (x) =

g(x) + cH(x − a), where g has compact support and c and a are constants.
Then hf (x), 1i = hg ′ (x), 1i + hcH ′ (x − a), 1i = 0 + chδ(x − a), 1i = c, from
where the result follows.
We now define two-sided Cesàro evaluations.
Definition 5.4. Let f ∈ D′ (R), φ ∈ E(R), and m ∈ N0 . We say the

evaluation hf (x), φ(x)i exists in the Cesàro sense of order m if there is a
decomposition f = f− + f+ , suppf− ⊆ (−∞, 0] and suppf+ ⊆ [0, ∞), such
that both evaluations hf± (x), φ(x)i = γ± (C, m) exist. In this case we write
hf (x), φ(x)i = γ (C, m), (5.28)
where γ = γ− + γ+ .
We must check the consistence of Definition 5.4. Let f = f1 +f2 = g1 +g2

be two decompositions such that f2 and g2 have supports bounded at the
left, respectively, f1 and g1 have supports bounded at the right. Then
h = g1 −f1 = f2 −g2 has compact support. If both hfj (x), φ(x)i = γj (C, m)
exist, then, by Lemma 5.1, both hgj (x), φ(x)i = βj (C, m) exist, and we
have the two equalities β1 = γ1 +β and β2 = γ2 −β, where β = hfj (x), φ(x)i.
Hence the number γ = γ1 + γ2 = β1 + β2 is independent on the choice of
the decomposition.
Let us now define Abel summability for distributional evaluations.
Definition 5.5. Let f ∈ D′ (R) and φ ∈ E(R). We say the evaluation

hf (x), φ(x)i exists in the Abel sense if there is a decomposition f = f− +f+ ,
suppf− ⊆ (−∞, 0] and suppf+ ⊆ [0, ∞), such that both e∓yx φ(x)f± ∈
S ′ (R), for each y > 0, and

lim+ hφ(x)f− (x), eyx i + φ(x)f+ (x), e−yx = γ , (5.29)
y→0
in this case we write hf (x), φ(x)i = γ (A).
The notion of distributional evaluations in the Cesàro sense admits a

characterization in terms of the quasi-asymptotic behavior.
Proposition 5.3. Let f ∈ D′ (R) and φ ∈ E(R). Then hf (x), φ(x)i =

γ (C) if and only if there exist a decomposition f = f− + f+ , where
suppf− ⊆ (−∞, 0] and suppf+ ⊆ [0, ∞), and a constant β such that the
following quasi-asymptotic behaviors hold

q γ δ(x)
φ(λx)f+ (λx) ∼ +β as λ → ∞ in S ′ (R) (5.30)
2 λ
and
q γ δ(x)
φ(λx)f− (λx) ∼ −β as λ → ∞ in S ′ (R) . (5.31)
2 λ
In particular, we obtain that φf ∈ S ′ (R) and it has the quasi-asymptotic
behavior,
q δ(x)
φ(λx)f (λx) ∼ γ as λ → ∞ in S ′ (R) . (5.32)
λ
(−1)
Proof. We may assume that φ ≡ 1. Put f− equal to the primitive
of f− (−x) with support on [0, ∞). Because of the assumptions on the
(−1)
supports, note that (5.30) and (5.31) are equivalent to limλ→∞ f± (λx) =
((γ/2) ± β)H(x) in S ′ (R). By Proposition 5.1, the latter are equivalent to
(−1)
limλ→∞ f± (x) = (γ/2) ± β (C), which are equivalent to hf± (x), 1i =
(γ/2)±β (C). And so we obtain the equivalence with hf (x), φ(x)i = γ (C).

We can use Proposition 5.3 to obtain Abel’s theorem in the context of
distributional evaluations. The converse is false [48].
Corollary 5.1. Let f ∈ D′ (R) and φ ∈ E(R). Suppose that hf (x), φ(x)i =
γ (C), then hf (x), φ(x)i = γ (A).
Proof. Using Proposition 5.3, we obtain that, as λ → ∞,

x
x
φ(x)f− (x), e λ + φ(x)f+ (x), e− λ

= λ hφ (λx) f− (λx) , ex i + φ (λx) f+ (λx) , e−x

γ γ
= − β hδ(x), ex i + + β hδ(x), e−x i + o(1)
2 2
= γ + o(1) .
5.2 Summability of the Fourier transform and

distributional point values
We have arrived to the main section of 5. We characterize the value of

a tempered distribution at point in terms of the summability its Fourier
transform. Let us adopt the following convention to denote the value of a

distribution at a point. Let f ∈ D′ (R) have the value γ at the point x = x0
in the sense of Lojasiewicz [93, 94], that is, the following quasi-asymptotic
behavior is satisfied
f (x0 + εx) = γ + o(1) as ε → 0 in D′ (R). (5.33)
We will refer to Lojasiewicz point values as distributional point values, and

will use the following notation for the existence of the distributional point
value at x = x0 with value γ,
f (x0 ) = γ, distributionally. (5.34)
In order to characterize distributional point values. We will define an

intermediate method of summability, weaker than Cesàro summability but
stronger than Abel summability. We discuss applications of such a char-
acterization to the study of convergence of Fourier series and integrals of
distributions. Several Tauberian type results are also presented, in partic-
ular we obtain a quick simple distributional proof of the celebrated Little-
wood’s theorem [92, 177]. In the last part, we also formulate and solve the
Hardy-Littlewood problem for Cesàro summability [68, 181] in the frame-
work of distributions. The original sources for the results of this section are
[172, 173, 176, 177, 181]. Finally, we mention references [181, 184], where
more refined versions can be found, especially some results concerning to
the order of summability that we do not treat here.
5.2.1 Characterization of distributional point values of tempered

distributions
Let f ∈ S ′ (R) have distributional point γ at x0 . Then, the quasi-

asymptotics (5.33) actually holds in the space S ′ (R)(cf. Theorem 2.35
in 2.11.1). Therefore, we can take Fourier transform in (5.33) and obtain
the equivalent quasi-asymptotic expression
q δ(x)
eiλx0 x fˆ(λx) ∼ 2πγ as λ → ∞ in S ′ (R) . (5.35)
λ
Let us state this simple, but useful, observation as a lemma
Lemma 5.2. Let f ∈ S ′ (R). Then, f (x0 ) = γ, distributionally, if and

only if the Fourier transform satisfies the quasi-asymptotic behavior (5.35).
Therefore, on the Fourier side, distributional point values look like

(5.35). Since our ultimate goal is to characterize distributional point values
by certain type of summability of the Fourier transform, it is clear that
our summability method should provide a characterization of the quasi-
asymptotic behavior
q δ(x)
g(λx) ∼ γ as λ → ∞ in S ′ (R) . (5.36)
λ
A naive first approach to this problem might lead us to consider directly
Cesàro summability. However, Proposition 5.3 tell us that it is not going
to work: Cesàro summability is too strong to give a characterization. Let
us be more precise on this matter. Observe that if hg(x), 1i = γ (C), then
Proposition 5.3 implies (5.36). However the converse is not true.
Example II.2. Consider the regular distributions g(x) = (1/(x log |x|))
H(|x| − 3). Note that for any m ≥ 0,
Z x m Z m−1
1 t 1 m x t
1− dt = − + log(log t) 1 − dt
3 t log t x 3 log 3 x 3 x
∼ log(log x), x → ∞ .
Then, the evaluation hg(x), 1i does not exist in the Cesàro sense. However,
g(λx) = o(λ−1 ) as λ → ∞ in S ′ (R). In fact, if φ ∈ S(R), then
Z
1 ∞ φ(t) − φ(−t) 1
hg(λx), φ(x)i = dt = o , λ → ∞.
λ λ3 t log(λt) λ
Therefore, the Cesàro summability is not adequate for the characteriza-

tion of distributional point values. If we now think carefully in (5.36), it is
actually a quasi-asymptotic behavior of degree −1 with respect to the triv-
ial slowly varying function, L ≡ 1, and we have already characterized the
structure of such behaviors in 2.10.5. Let us state the structural theorem
for this particular case.
Proposition 5.4. Let g ∈ D′ (R), then,

q δ(x)
f (λx) ∼ γ as λ → ∞ in D′ (R) , (5.37)
λ
if and only if there exist m ∈ N and (m + 1)-primitive Gm+1 of g, i.e.,
G(m+1) = f , which is locally integrable for large positive and negative argu-
ments, such that following limit holds for each a > 0
lim (a−m Gm+1 (ax) − (−1)m Gm+1 (−x)) = γ . (5.38)
x→∞
Furthermore, the above relation is satisfied if and only if there exists an

asymptotically homogeneous function c of degree zero, i.e., c(ax) = c(x) +
o(1) as x → ∞ for each a > 0, such that
γsgnx m xm m
Gm+1 (x) = x + c(|x|) + o(|x| ), |x|→ ∞ , (5.39)
2m! m!
in the ordinary sense. In such a case g is a tempered distribution and (5.37)
holds in S ′ (R).
Proposition 5.4 implicitly suggests the method of summability: It should

involve (5.38). Let us reformulate (5.38). Set G = G(m) , then G is a first
order primitive of g, and in fact (5.38) can be rewritten as
lim (G(ax) − G(−x) = γ (C, m) .
x→∞
Hence, we have found the right summability method!
Definition 5.6. Let g ∈ D′ (R), φ ∈ E(R) and m ∈ N0 . We say that the

special value of hg (x) , φ (x)i exists in the Cesàro sense of order m (e.v.
Cesàro sense), and write
e.v. hg (x) , φ (x)i = γ (C, m) , (5.40)
′
if for some primitive G of φg, i.e., G = φg, and each a > 0, we have
lim (G(ax) − G(−x)) = γ (C, m) . (5.41)
x→∞
As a corollary we obtain.
Corollary 5.2. Let g ∈ D′ (R), φ ∈ E(R). Then

e.v. hg (x) , φ (x)i = γ (C) (5.42)
if and only if
δ(x)
q
φ(λx)g(λx) ∼ γ as λ → ∞ in S ′ (R). (5.43)
λ
In addition, we have that φg ∈ S ′ (R).
As expected, the Cesàro method is strictly stronger than the e.v Cesàro
summability (see also Example II.2).
Proposition 5.5. Let g ∈ D′ (R), φ ∈ E(R). Any evaluation summable

(C, m) is also summable in e.v.(C, m), that is, hg (x) , φ (x)i = γ (C, m),
implies e.v. hg (x) , φ (x)i = γ (C, m).
Proof. Let G be a first order primitive of φg. Decompose it as G =

G− + G+ , with supp G− ⊆ (−∞, 0] and supp G− ⊆ [0, ∞). Then, by
Proposition 5.3,
γ
lim ±G± (±x) = ± β (C, m),
x→∞ 2
for some β. Thus
lim (G(ax) − G(−x)) = lim (G+ (ax) − G− (−x)) = γ (C, m).
x→∞ x→∞
In summary, we succeeded characterizing distributional point values in

terms of the summability of the Fourier inversion formula.
Theorem 5.1. Let f ∈ S ′ (R). We have f (x0 ) = γ, distributionally, if and

only if there exists an m ∈ N0 such that
1 D E
e.v. fˆ (x) , eix0 x = γ (C, m). (5.44)
2π
Proof. Combine Lemma 5.2 with Corollary 5.2.
Let g = µ be a Radon measure. It convenient in this case to write

Z ∞
e.v. φ(x)dµ(x) = γ (C, m) (5.45)
−∞
for (5.40). When m = 0, we suppress (C, 0) from the notation, and simply
write
Z ∞
e.v. φ(x)dµ(x) = γ .
−∞
P∞
In particular, if µ = n=−∞ cn δ( · − n) and φ ≡ 1, we use the notation
∞
X
e.v. cn = γ (C, m), (5.46)
n=−∞
omitting again (C, 0) when m = 0.
Observe that if we use the family of summability kernels

m
m m x
φa (x) = (1 + x) (H(−x) − H(−1 − x)) + 1 − (H(x) − H(x − a)),
a
(5.47)
where H is the Heaviside function, then (5.45) holds if and only if
Z ∞
t
lim φma φ(t)dµ(t) = γ, for each a > 0 . (5.48)
x→∞ −∞ x
For series we obtain that (5.46) holds if and only if

∞
X
m n
lim φa cn = γ, for each a > 0 . (5.49)
x→∞
n=−∞
x
Let us now discuss some immediate consequences of Theorem 5.1.
Corollary 5.3. Let f ∈ S ′ (R) be such that fˆ = µ is a Radon measure.

Then, we have f (x0 ) = γ, distributionally, if and only if there exists an
m ∈ N0 such that
Z ∞
1
e.v. eix0 x dµ(x) = γ (C, m), (5.50)
2π −∞
or which amounts to the same,
Z ∞
1 ix0 t m t
lim e φa dµ(t) = γ, for each a > 0 . (5.51)
x→∞ 2π −∞ x
The next corollary is a result of R. Estrada [45], the characterization of

Fourier series having a distributional point value.
Corollary 5.4. Let f ∈ S ′ (R) be a 2π-periodic distribution having Fourier

series
X∞
f (x) = cn einx . (5.52)
n=−∞
Then, we have f (x0 ) = γ, distributionally, if and only if there exists an

m ∈ N0 such that
∞
X
e.v. cn einx0 = γ (C, m), (5.53)
n=−∞

X∞
m n
lim φa cn einx0 = γ, for each a > 0 . (5.54)
x→∞
n=−∞
x
P
Proof. We have that fˆ(x) = 2π ∞n=−∞ cn δ(x − n), the rest follows
from Corollary 5.3.
Let us state Corollary 5.3 when fˆ ∈ L1loc (R). A particular case is

obtained if f ∈ Lp (R) with 1 ≤ p ≤ 2, since fˆ ∈ Lq (R) with q = p/(p − 1)
[170].
Corollary 5.5. Let f ∈ S ′ (R) be such that fˆ ∈ L1loc (R). Then, we have
f (x0 ) = γ, distributionally, if and only if there exists an m ∈ N0 such that
Z ∞
1
e.v. eix0 x fˆ(x)dx = γ (C, m), (5.55)
2π −∞
Z ∞
1 t
lim φam
eix0 t fˆ(t)dt = γ, for each a > 0 . (5.56)
x→∞ 2π −∞ x
It is important to observe that the characterization of distributional

point values is given in terms of slightly asymmetric means and that the
corresponding result for symmetric means does not hold. The result for
separate integration over both the positive and negative parts of the spec-
trum does not hold either (we already discussed the latter in Example II.2).
Let us provide two further examples.
Example II.3. If f ∈ S ′ (R) and f (x0 ) = γ, distributionally, then by

taking a = 1 we obtain that the symmetric means converge to γ, in the
Cesàro sense, so that, in case fˆ(t)eix0 t is locally integrable,
Z x
1
lim fˆ(t)eix0 t dt = γ (C, m), (5.57)
x→∞ 2π −x
for some m. However, (5.57) does not imply the existence of the distribu-
tional value f (x0 ). A simple example is provided by f (x) = δ ′ (x) at x = 0,
since fˆ(t) = it, so that (5.57) exists and equals 0, but f (0) does not exist.
Example II.4. If f ∈ S ′ (R) and the two Cesàro limits

Z x
1
lim fˆ(t)eix0 t dt = γ+ (C, k), (5.58)
x→+∞ 2π 0
Z 0
1
lim fˆ(t)eix0 t dt = γ− (C, k), (5.59)
x→+∞ 2π −x
exist then the distributional value f (x0 ) exists and equals γ = γ+ + γ− .

However, the existence of the distributional point value f (x0 ) does not
imply the existence of both Cesàro limits. For instance, if
Z ∞
sin xt dt
f (x) = , (5.60)
0 t ln (t2 + a2 )
for some a > 1, then f is continuous and f (0) = 0, but fˆ(t) = −πit−1
(log(t2 + a2 ))−1 , and in that case both limits (5.58) and (5.59) give infinite
results, |γ+ |= |γ− |= ∞.
There is one case in which the distributional point values can be charac-
terized by Cesàro summability of the Fourier inversion formula, not needing
the asymmetric means, that is, when the distribution has support on a half-
ray.
Theorem 5.2. Let f ∈ S ′ (R) have support bounded at the left . We have
f (x0 ) = γ, distributionally, if and only if there exists an m ∈ N0 such that
1 Dˆ E
f (x) , eix0 x = γ (C, m) . (5.61)
2π
Proof. The converse follows from Proposition 5.5. Let now F be the
primitive of (1/2π)eix0 x fˆ with support bounded at the left. Then, by The-
orem 5.1, we have that
lim F (x) = lim (F (x) − F (−x)) = γ (C, m) .
x→∞ x→∞
Corollary 5.6. Let f ∈ S ′ (R) be such that fˆ = µ is a Radon measure

supported on [0, ∞). Then, we have f (x0 ) = γ, distributionally, if and only
if
Z ∞
1
eix0 x dµ(x) = γ (C) . (5.62)
2π 0
We also obtain a corresponding result for Riesz summability.
P∞
Corollary 5.7. Let f = n=0 cn eiλn x in S ′ (R), where λn ր ∞. Then,
we have f (x0 ) = γ, distributionally, if and only if
X∞
cn eiλn x0 = γ (R, hλn i) . (5.63)
n=0
These ideas can be applied to study some types of multiple series. It

is not our scope to investigate problems in several variables (which are
actually still open questions, see Open Problem 5.1 below), but the next
theorem shows that some problems in summability of multiple series can
be solved using this theory. The next result is an example of that.
Theorem 5.3. Let f ∈ S ′ (R) and be ρ be a real-valued function defined

on Rd which only takes non-negative values at points j ∈ Zd . Suppose that
X
f (x) = cj eiρ(j)x in S ′ (R) .
j∈Zd
∞
Enumerate the image ρ(Zd ) by an increasing sequence hλn in=0 Then,
f (x0 ) = γ, distributionally, if and only if there exists an m ∈ N such
that
 
∞
X X
 cj eiρ(j)x0  = γ (R, hλn i , m) , (5.64)
n=0 ρ(j)=λn
or equivalently,
X m
ρ(j)
lim cj eiρ(j)x0 1− =γ. (5.65)
λ→∞ λ
ρ(j)≤λ
Proof. It follows immediately from Corollary 5.7, since

 
∞
X X
f (x) =  cj eiρ(j)x0  eiλn x .
n=0 ρ(j)=λn
2
If in particular we take ρ(y) = |y| (here y ∈ Rd and | · | is the standard
euclidean norm) in Theorem 5.3, we obtain that f (x0 ) = γ, distributionally,
if and only if the multiple series is Bochner-Riesz summable by spherical
means [25].
We end this subsection by pointing out an important open problem.

While we have completely characterized the value of a one dimensional
distribution at point by the summability of its Fourier transform in the e.v.
Cesàro sense, the corresponding multidimensional problem is still an open
question.
Open Problem 5.1. For distributions in S ′ (Rd ), d > 1, find a charac-

terization of distributional point values in terms of the summability of the
multidimensional Fourier transform. The method of summability is part of
the open question.
5.2.2 Abel summability
We now analyze Abel summability of the Fourier inversion formula in the

presence of distributional point values.
Let us first state an interesting theorem, which we may be regarded as

a decomposition theorem for the quasi-asymptotic behavior (5.36).
Theorem 5.4. Let g ∈ S ′ (R). Then

δ(x) 1
g(λx) = γ +o as λ → ∞ in S ′ (R) (5.66)
λ λ
if and only if there exist a decomposition g = g− + g+ , where supp g− ⊆

(−∞, 0] and suppg+ ⊆ [0, ∞), and an asymptotically homogeneous function
c of degree zero, i.e., c(ax) = c(x) + o(1) as x → ∞ for each a > 0, such
that the following asymptotic relations hold

γ δ(x) 1
g+ (λx) = + c(λ) +o as λ → ∞ in S ′ (R) (5.67)
2 λ λ
and

γ δ(x) 1
g− (λx) = − c(λ) +o as λ → ∞ in S ′ (R) . (5.68)
2 λ λ
Proof. Proposition 5.4 implies the existence of an (m + 1)-primitive of

g, say G, such that
γsgnx m xm m
G(x) = x + c(|x|) + o(|x| ), |x| → ∞ . (5.69)
2m! m!
Set G± (x) = G(x)H(±x), where H is the Heaviside function. We have that
(cf. 2.10.2)
c(λx)H(x) = c(λ)H(x) + o(1) as λ → ∞ in S ′ (R),
which implies
γ (λx)m
±
G± (λx) = (±1)m+1 (λx)m m
± + (±1) c(λ) + o(λm ) as λ → ∞.
2m! m!
(m+1)
in S ′ (R). If we set g± = G± , differentiating (m + 1)-times the last two
asymptotic expressions we obtain (5.67) and (5.68). Conversely, setting
h± (x) = g± (x) ∓ (c(x)H(x))′ , an application of the structural theorem for
the quasi-asymptotic behavior of degree −1 with one-sided support to each
h± implies that there exists m such that (5.69) is satisfied, and hence (5.66)
follows.
Due to Corollary 5.2, Theorem 5.5 may also be stated in the following
equivalent form.
Theorem 5.5. Let g ∈ D′ (R) and φ ∈ E(R). Then e.v. hf (x), φ(x)i =
γ (C) if and only if there exist a decomposition g = g− + g+ , where
supp g− ⊆ (−∞, 0] and supp g+ ⊆ [0, ∞), and an asymptotically homoge-
neous function c of degree zero, i.e., c(ax) = c(x) + o(1) as x → ∞ for each
a > 0, such that the following asymptotic relations hold

γ δ(x) 1
φ(λx)g+ (λx) = + c(λ) +o as λ → ∞ in S ′ (R) (5.70)
2 λ λ
and

γ δ(x) 1
φ(λx)g− (λx) = − c(λ) +o as λ → ∞ in S ′ (R) . (5.71)
2 λ λ
We can now obtain the next Abelian result.
Proposition 5.6. Let g ∈ D′ (R) and φ ∈ E(R). Suppose that

e.v. hg(x), φ(x)i = γ (C) .
Then, hg(x), φ(x)i = γ (A). Moreover, Let g = g− + g+ be a decomposition
satisfying the support requirements of Theorem 5.5, then

lim g− (t), φ(t)eiz̄t + g+ (t), φ(t)eizt = γ , (5.72)
z→0
in any sector Im z ≥ M |Re z|, with M > 0.
Proof. We may assume that φ ≡ 1. We use (5.67) and (5.68). Write

z = (1/λ)(τ + i), so |τ | ≤ (1/M ), hence, as λ → ∞,

g− (t), eiz̄t + g+ (t), eizt
D E D E
= λ g− (−λt), e−(iτ +1)t + g+ (t), e(iτ −1)t
D E γ D E
γ
= − c(λ) δ(t), e(iτ +1)t + + c(λ) δ(t), e(iτ −1)t + o(1)
2 2
= γ + o(1) ,

with uniform convergence since e(iτ −1)t H(t) M|τ |≤1 is compact in S+ .
So, we obtain the Fourier inversion formula in the Abel sense.

Corollary 5.8. Let g ∈ S ′ (R). Suppose f (x0 ) = γ, distributionally. Then

the Fourier inversion formula holds in the Abel sense, i.e.,
1 Dˆ E
f (x), eix0 x = γ (A). (5.73)
2π
Moreover, let fˆ = fˆ− + fˆ+ , with suppfˆ− ⊆ (−∞, 0] and suppfˆ+ ⊆ [0, ∞),
then
1 D ˆ E D E
lim f− (t), eiz̄t + fˆ+ (t), eizt =γ, (5.74)
z→x0 2π
in any sector Im z ≥ M |Re z − x0 |, with M > 0.
In the case of Fourier series, we recover a result from [196].
Corollary 5.9. Let f ∈ S ′ (R) be a 2π-periodic distribution having Fourier

series
X∞
f (x) = cn einx . (5.75)
n=−∞
Suppose f (x0 ) = γ, distributionally. Then

∞
!
X
−inz̄ inz
lim c0 + c−n e + cn e =γ (5.76)
z→x0
n=1
in any sector Im z ≥ M |Re z − x0 |, with M > 0. In particular, if an =

c−n + cn and bn = cn − c−n , we obtain that
∞
a0 X
+ (an cos nx0 + bn sin nx0 ) = γ (A) . (5.77)
2 n=1
Proof. Relation (5.76) follows directly form Corollary 5.8. If we set

z = x0 + iy in (5.76), we obtain
∞
!
a0 X −ny
lim + (an cos nx0 + bn sin nx0 ) e =γ
y→0+ 2 n=1
which gives (5.77).
Let now f ∈ D′ (R) have f (x0 ) = γ, distributionally. We cannot longer

talk about Abel summability of the Fourier inversion formula, since the
Fourier transform is not available in D′ (R). Nevertheless, there is a substi-
tute of Abel summability, if we interpret it as the boundary limit at x = x0
of a harmonic representation. Recall that a harmonic function U , harmonic

on Im z > 0, is called a harmonic representation [13, 5] of f if
lim U (x + iy) = f (x), in D′ (R) . (5.78)
y→0+
It is very well known that any distribution admits a harmonic representation

[13].
Theorem 5.6. Let f ∈ D′ (R). Suppose that U is a harmonic representa-

tion of f in the upper half-plane Im z > 0. If f (x0 ) = γ, distributionally,
then
lim U (z) = γ , (5.79)
z→x0
in any sector Im z ≥ M |Re z − x0 |, with M > 0.
Proof. We first see that we may assume f ∈ S ′ (R). Indeed we can

decompose f = f1 + f2 where f2 is zero in a neighborhood of x0 and
f1 ∈ S ′ (R). Let U1 and U2 be harmonic representations of f1 and f2 ,
respectively; then U2 represents the zero distribution in a neighborhood of
x0 . Then by applying the reflection principle to the real and imaginary
parts of U2 ([5] , [169] ), we have that U admits a harmonic extension
to a (complex) neighborhood of x0 , and so it is real analytic, therefore,
U (z) − U1 (z) = U2 (z) = O (|z − x0 |) as z → x0 . Additionally, f1 (x0 ) = γ,
distributionally, thus, we can assume that f = f1 . The same argument with
the reflection principle shows that (5.79) is independent of the choice of U .
Therefore, we can assume that U is the Fourier-Laplace representation [13]
of f , that is, let fˆ = fˆ+ + fˆ− be a decomposition such that suppfˆ− ⊆
(−∞, 0] and suppfˆ+ ⊆ [0, ∞), we can assume that
1 D ˆ E D E
U (z) = f− (t), eiz̄t + fˆ+ (t), eizt , Im z > 0.
2π
But in this case, Corollary 5.8 yields (5.79).
Naturally, the converse of Theorem 5.6 is not true.
5.2.3 Convergence of Fourier Series
We now analyze sufficient conditions under which the existence of the dis-
tributional point value implies the convergence of the Fourier series at the
point. Note that, in particular, any result of this type gives a Tauberian
condition for Cesàro summability of series. The next theorem is our first
result in this direction. We denote by lp , 1 ≤ p < ∞, the set of those
∞ P∞ p
sequences hcn in=−∞ such that n=−∞ |cn | < ∞.
P∞ inx
Theorem 5.7. Let f (x) = n=−∞ cn e in S ′ (R). Suppose that hcn i ∈
lp , 1 ≤ p < ∞ and
∞
X
p 1
rN,p = |cn | = O , N → ∞. (5.80)
N p−1
|n|≥N
Then, f (x0 ) = γ, distributionally, implies

∞
X
e.v. cn einx0 = γ , (5.81)
n=−∞
or which amounts to the same

X
lim cn einx0 = γ , (5.82)
x→∞
−x≤n≤ax
for each a > 0.
Proof. If p = 1, it is trivial. Let us assume 1 < p < ∞, and let us find

1 1
q so that + = 1. If f (x0 ) = γ, we have
p q
∞
X
lim cn eix0 n ψ(εn) = γψ(0) ,
ε→0+
n=−∞
for each ψ ∈ S(R). Choose φ ∈ D(R) such that 0 ≤ φ ≤ 1, and φ(x) = 1

for x ∈ [−1, a]. Hence
∞
X X X
cn einx0 φ(nε) = cn eix0 n + cn eix0 n φ(nε)
n=−∞ a
− 1ε ≤n≤ a
ε ε <n
X
+ c−n e−ix0 n φ(−nε) + o(1) ,
1
ε <n
as ε → 0+ . Therefore,
 
X X X

lim sup cn einx0 − γ ≤ lim sup |cn | |φ(εn)| + |c−n | |φ(−εn)|.
ε→0+ 1 ε→0+ a 1
− ≤n≤ε
a
ε ε <n <n ε
But,
  p1   1q
X X  X 
p q
|cn | |φ(εn)| ≤ |cn | |φ(εn)| ,
a
a  a 
ε <n ε <n ε <n
By (5.80), we can find M > 0 such that

  1q
X 1
 X 
q
|cn | |φ(εn)| ≤ M a− q ε |φ(εn)| .
a
 a 
ε <n ε <n
Then,
X Z ∞ 1q
1 q
lim sup lim |cn | |φ(εn)| ≤ M a− q |φ(x)| dx .
ε→0+ a a
ε <n
′
Similarly, ∃M > 0 such that
X Z −1 1q
′ − 1q q
lim sup lim |c−n | |φ(−εn)| ≤ M a |φ(x)| dx .
ε→0+ −∞
1
ε <n
Now, we are free to choose φ such that the right sides of the last two
inequalities are both less than η/2. Therefore,

X

lim sup lim cn e inx0
− γ < η .
ε→0+
− 1 ≤n≤ a ε ε
Since this can be done for each η > 0, we conclude that

X
lim cn einx0 = γ,
x→∞
−x≤n≤ax
as required.
As an example of the use of Theorem 5.7, let us obtain a classical

Tauberian result of Hardy for Cesàro summability of series [66].
P
Corollary 5.10. Suppose that ∞ n=0 cn = γ (C). The Tauberian condi-
tion ncn = O(1) implies the convergence of the series to γ.
Proof. We associate to the sequence a Fourier series, f (x) =

P∞ inx
c
n=0 n e . The (C) summability to γ implies f (0) = γ, distributionally.
Now, Hardy’s Tauberian hypothesis obviously implies (5.80), so Theorem
5.7 gives the convergence.
We can generalize Theorem 5.7 to other norms.
P∞ inx
Theorem 5.8. Let f (x) = n=−∞ cn e in S ′ (R). Suppose that
X
p −pr 1
|cn | |n| =O (5.83)
N rp+p−1
|n|≥N
for some r and p with 1 < p < ∞. If f (x0 ) = γ, distributionally, then

∞
X
e.v. cn einx0 = γ, (5.84)
n=−∞
or which amounts to the same

X
lim cn einx0 = γ, (5.85)
x→∞
−x≤n≤ax
for each a > 0.
Proof. Use the inequality

  p1   1q
X X  X 
p q
|cn | |φ(εn)| ≤ |cn | n−rp nrq |φ(εn)|
a
a  a 
ε <n ε <n ε <n
and follow a similar argument as the one in the proof of Theorem 5.7.
If we take r = (1/p) − 1 in the last theorem, we obtain the following

Hardy-Littlewood Tauberian condition for (C) summability [67].
Corollary 5.11. If
∞
X
cn = γ (C, k), (5.86)
n=0
for some k ∈ N, then the Tauberian condition (p ≥ 1)

∞
X
np−1 |cn |p < ∞ (5.87)
n=0
P∞
implies that n=0 cn is convergent to γ.
Next, we would like to make some comments about the results we just
P
discussed. If hcn i ∈ lp for 1 ≤ p ≤ 2, then f (x) = cn einx belongs to
q p
L [0, 2π], but the converse is not true. Similarly, if f ∈ L [0, 2π], 1 ≤ p ≤ 2,
then hcn i, belongs to lq , but the converse is not necessarily true. Hence,
the results for hcn i ∈ lp with 1 ≤ p ≤ 2 are about functions. However,
for p > 2, these results are about distributions, in general. For example, as
follows from [206] , if hcn i ∈ lp \ l2 for some p > 2 then for almost all choices
P
of signs ρn = ± the distribution ∞ n=0 ρn cn e
inx
is not locally integrable; or
p 2
P ∞ inx
if hcn i ∈ l \ l is lacunary then n=0 cn e is never a regular distribution.
We conclude this subsection discussing a type of Tauberian result in

summability of Fourier series of distributions, where the conclusion is not
the convergence of the series but the (C,m) summability for a specific m.
As it has been mentioned before, any result of this type gives a result in the
theory of Cesàro summability of series. Let us suppose that f is a periodic
P∞ inx
distribution of period 2π, and f (x) = n=−∞ cn e . We want to find
sufficient conditions under which the existence of f (x0 ), distributionally,
implies that
X
lim cn einx0 = f (x0 ) (C, m), (5.88)
x→∞
−x≤n≤ax
for an specific positive integer m. A partial answer to this question is given

in Theorem 5.9.
P∞
Theorem 5.9. Let f ∈ S ′ (R) such that f (x) = n=−∞ cn einx . Suppose
that f (x0 ) = γ, distributionally. If for a fixed a
X
cn einx0 = O(1) (C, m), (5.89)
−x≤n≤ax
then
X
lim cn einx0 = γ (C, m + 1). (5.90)
x→∞
−x≤n≤ax
Proof. For x > 0, set

X
ga (x) = cn einx0 ,
−x≤n≤ax
and put ga (x) = 0 for x ≤ 0. Condition (5.89) means that there is an

m-primitive G of ga , such that suppG ⊆ [0, ∞) and
G(x) = O (xm ) , x → ∞ .
In addition, since f (x0 ) = γ, we have that

m m
q γλ x+
G(λx) ∼ as λ → ∞ in D′ (R),
m!
i.e., for each φ ∈ D(R)
Z ∞ Z
γλm ∞ m
G(λx)φ(x) dx = x φ(x) dx + o (λm ) .
0 m! 0
Pick φ ∈ D(R) such that φ(x) = 1 for x ∈ [−1, 1] and supp φ ⊆ [−1, 2].
Evaluating G at φ, we obtain
Z Z
1 λ 1 2λ x
G(x) dx + G(x)φ dx
λ 0 λ λ λ
Z
γλm γλm 2 m
= + x φ(x) dx + o(λm ), λ → ∞ ,
(m + 1)! m! 1
which implies

(m + 1)! Z λ

m+1 G(x) dx − γ
λ 0
Z 2 Z
(m + 1)! 2λ x
≤ o(1) + γ(m + 1) xm φ(x) dx + m+1
|G(x)| φ dx
1 λ λ λ
Z 2
= o(1) + {γ(m + 1) + (m + 1)! O(1)} xm φ(x) dx, λ → ∞ ,
1
R2
since we can choose φ such that 1 xm φ(x) dx is as small as we want, we
conclude that
Z
(m + 1)! λ
lim G(x) dx = γ ,
λ→∞ λm+1 0
and the result follows.
We obtain the following interesting corollary of Theorem 5.9, known as

convexity theorem [66].
Corollary 5.12. Let hcn in∈N0 be a sequence of complex numbers. Suppose

that
X∞
cn = γ (C, k) , (5.91)
n=0
for some k ∈ N. If the m-Cesàro means are bounded then
∞
X
cn = γ (C, m + 1) . (5.92)
n=0
5.2.4 Series with gaps
In this subsection we apply the ideas of the last subsection to series with
gaps. In particular, we shall find examples of continuous functions whose
distributional derivatives do not have distributional point values at any
point.
P
Theorem 5.10. Let f (x) = ∞ n=0 cn e
inx
, in S ′ (R). In addition, suppose
∞
that hcn in=0 is lacunary, in the sense of Hadamard, i.e., cn = 0 except for
a sequence nk ∈ N0 with nk+1 > αnk for some α > 1. Then f (x0 ) = γ,
distributionally, if and only if
X ∞
cn einx0 = γ . (5.93)
n=0
In particular, cnk = o(1), k → ∞.
Proof. Let φ ∈ D(R) such that 0 ≤ φ ≤ 1, φ(x) = 1 for x ∈ [0, 1] and

supp φ ⊆ [−1, α]. Set bn = cn einx0 , for each n ∈ N. We have that
X X
nk
M (λ) = bnk + bnk φ − γ = o(1), λ → ∞ .
λ
nk ≤λ λ<nk <αλ
Note that given λ > 0 there exists at most one kλ such that λ < kλ < αλ.
Therefore if λ = nm , we obtain
M (nm ) = o(1) , m → ∞ ,
which is the same as
m
X
bnk − γ = o(1), m → ∞ .
k=0
Moreover, with a little modification of the last argument, we obtain the

following result.
P∞ inx
Theorem 5.11. Let f (x) = n=−∞ cn e , in S ′ (R). Suppose that
hcn in∈Z is lacunary in both directions; then, f (x0 ) = γ, distributionally,
if and only if
X
lim cn einx0 = γ ,
x→∞
−x≤n≤ax
for each a > 0.
We obtain several interesting corollaries from the last two theorems.

The second part of the following corollary is a result of Kolmogorov.
Corollary 5.13. If f ∈ L1 [0, 2π] and hcn in∈Z lacunary, then the Fourier
series of f converges to f (x0 ) at every point where f (x0 ) exists distri-
butionally in the sense of Lojasiewicz. In particular, it converges almost
everywhere.
Proof. Indeed, the first part follows directly from Theorem 5.11, while
the second statement is true because f has distributional point values at
every point of the Lebesgue set of f .
P∞
Corollary 5.14. Let f (x) = n=−∞ cn einx , in S ′ (R). If hcn in∈Z is lacu-
nary, but cn 6= o(1), then the distributional value f (x0 ) does not exist at
any point x0 .
The next corollary allows us to find examples of continuous functions

whose distributional derivatives do not have point values anywhere.
Corollary 5.15. Let hcn in∈Z be a lacunary sequence such that cn 6= o(1)
but cn = O(1), |n| → ∞. Then
X∞
cn inx
g(x) = e , (5.94)
n=−∞
n
is continuous but g ′ (x) does not have distributional point values at any
point; in particular, g is nowhere differentiable.
That g ′ does not have point values in the sense of Lojasiewicz at any
point is stronger than the fact that g is nowhere differentiable. For example,
consider g(x) = x sin x−1 ; g ′ (0) does not exist in the usual sense, but g ′ has
the value 0 at x = 0 distributionally [93].
A good illustration of Corollary 5.14 is obtained when we consider the

two Weierstrass functions
∞
X
fα (x) = b−nα cos(bn x),
n=0
and
∞
X
gα (x) = b−nα sin(bn x),
n=0
where b > 1 is an integer and α is a positive number less or equal to 1.
Observe that fα and gα are continuous. Weierstrass showed that for α small
enough they are nowhere differentiable. The extension to 0 < α ≤ 1 was
first proved by Hardy. Using Corollary 5.15, we obtain a stronger result for
it, namely, fα′ and gα′ do not have distributional point values at any point.
5.2.5 Convergence of Fourier integrals
We now extend the results from 5.2.3 to Fourier integrals.
Theorem 5.12. Let f ∈ S ′ (R). Assume that fˆ ∈ Lp , 1 ≤ p < ∞, and

Z ∞
ˆ p 1
rp,x = f (t) dt = O , x → ∞. (5.95)
|t|≥x xp−1
Then, f (x0 ) = γ, distributionally, if and only if
Z ∞
1
e.v. fˆ(t)eix0 t dt = γ . (5.96)
2π −∞
Proof. We only consider the case 1 < p < ∞. Assume that f (x0 ) = γ,
distributionally. Fix a > 0. Taking Fourier transform in

x
f x0 + = γ + o(1), λ → ∞ , (5.97)
λ
we obtain

2πγδ(x) 1
eix0 λx fˆ(λx) = +o , λ → ∞, in S ′ (R) . (5.98)
λ λ
Set g(x) = eix0 x fˆ(x). Take φ ∈ D(R), such that φ(x) = 1 for x ∈ [−1, a]
1 1
and 0 ≤ φ ≤ 1. Take q such that + = 1. Thus, we have
p q
Z λa Z −λ Z ∞
t t
g(t) dt − 2πγ = − g(t)φ dt − g(t)φ dt + o(1),
−λ −∞ λ aλ λ
as λ → ∞. We show that
Z ∞
t
lim g(t)φ dt = 0 , (5.99)
λ→∞ aλ λ
and
Z −λ
t
lim g(t)φ dt = 0 . (5.100)
λ→∞ −∞ λ
We have
Z ∞ 1 Z ∞ 1q
t q
g(t)φ
dt ≤ O λ−q
λ |φ(t)| dt
λ
aλ a
Z ∞ q1
q
= O(1) |φ(t)| dt .
a
R ∞ q 1
Since a |φ(t)| dt q can be made arbitrarily small, we conclude (5.99).
Similarly, (5.100) follows.
Likewise, one can show.
Theorem 5.13. Let f ∈ S ′ (R). Suppose that fˆ is locally integrable and

Z
ˆ p −rp 1
f (t) |t| dt = O , x → ∞.
|t|≥x xpr+p−1
for some 1 < p < ∞ and r ∈ R. Then, f (x0 ) = γ, distributionally, if and

only if
Z ∞
1
e.v. fˆ(t)eix0 t dt = γ .
2π −∞
If we take r = (1/p) − 1 in Theorem 5.13, we obtain the next corollary.
Corollary 5.16. If f ∈ S ′ (R), fˆ is locally integrable, and

Z ∞ p
p−1 ˆ
|t| f (t) dt < ∞ , (5.101)
−∞
then, f (x0 ) = γ, distributionally, if and only if

Z ∞
1
e.v. fˆ(t)eix0 t dt = γ .
2π −∞
5.2.6 Tauberian theorem for distributional point values
We now turn our attention to Tauberian theorems for Abel summability.

Our aim is to obtain a Tauberian converse to the following result of Con-
stantinescu [26]:
Suppose that f ∈ D′ (R) is the boundary value of a function F analytic

in the upper half-plane, that is, f (x) = F (x + i0) ; if f (x0 ) = γ distribu-
tionally, then F (x0 + iy) → γ as y → 0+ .
Notice that the above result is a particular case of Theorem 5.6, which
we already remarked that can be viewed as Abel summability for non-
tempered distributions. Observe also that the converse is false, as we have
pointed out many times before [48].
Our Tauberian condition will be distributional boundedness at a point.
Definition 5.7. Let f ∈ D′ (R). We say that f is distributionally bounded

at x = x0 if it is quasi-asymptotic bounded of degree 0 at x0 with respect
to the trivial slowly varying function, that is,
f (x0 + εx) = O(1) as ε → 0+ in D′ (R) . (5.102)
Example II.5. The function f (x) = |x|i is bounded in the ordinary sense,
and thus it defines a unique regular distribution which is distributionally
bounded at x = x0 . It easy to see that f (0) does not exist distributionally.
In general the evaluation hf (εx) , φ (x)i does not tend to a limit as ε → 0
if φ ∈ D (R) .
Observe the Definition 5.7 is meaningful if f is just defined in a neigh-

borhood of x = x0 , since the quasi-asymptotics at 0 are local properties.
We shall introduce the equivalent approach of Campos-Ferreira to distri-

butional point values and distributional boundedness [21]. It is in somehow
connected with the structure of these two quasi-asymptotic concepts. Let
us introduce the operator µa which is defined on complex valued locally
integrable functions defined in R as
Z x
1
µa {f (t) ; x} = f (t) dt, x 6= a, (5.103)
x−a a
while the operator ∂a is the inverse of µa ,

′
∂a (g) = ((x − a) g (x)) , (5.104)
and it is defined on distributions. Suppose first that f0 = f is real. Then if
it is bounded near x = a, we can define
f0 (a) = lim sup f (x) , f0 (a) = lim inf f (x) . (5.105)
x→a x→a
Then f1 = µa (f ) will be likewise bounded near x = a and actually

f0 (a) ≤ f1 (a) ≤ f1 (a) ≤ f0 (a) (5.106)
and, in particular, if f (a) = f0 (a) exists, then f1 (a) also exists and f1 (a) =
f0 (a) . The next lemma is not difficult to show, we leave the verification as
an exercise to the reader (see also [21]).
Lemma 5.3. A distribution f ∈ D′ (R) is distributionally bounded at x =

x0 if and only if there exist n ∈ N and fn ∈ D′ (R) , continuous and
bounded in a pointed neighborhood (x0 − ε, x0 ) ∪ (x0 , x0 + ε) of x0 , such
that f = ∂xn0 fn .
If f0 is distributionally bounded at x = x0 , then there exists a unique

distributionally bounded distribution near x = x0 , f1 , with f0 = ∂x0 f1 .
Therefore, ∂x0 and µx0 are isomorphisms of the space of distributionally
bounded distributions near x = x0 . Given f0 we can form a sequence of
distributionally bounded distributions {fn }∞n=−∞ with fn = ∂x0 fn+1 for
each n ∈ Z.
We have an analogous result for distributional point values, again, we

leave the proof of the following lemma as an exercise for the reader (see
also [21]).
Lemma 5.4. A distribution f ∈ D′ (R) satisfies f (x0 ) = γ, distribution-

ally, if and only if there exist n ∈ N and fn ∈ D′ (R) , continuous near
x0 , such that f = ∂xn0 fn and fn (x0 ) = γ. We call n the order of the point
value.
Observe also that if f = ∂xn0 fn , and fn is bounded near x = x0 , then

f (x0 ) exists distributionally, and equals γ, if and only if fn (x0 ) = γ, dis-
tributionally.
We start with a Tauberian result for bounded analytic functions.
Theorem 5.14. Let F be analytic and bounded in a rectangular region of

the form (a, b) × (0, R) . Suppose f (x) = limy→0+ F (x + iy) in the space
D′ (a, b) . Let x0 ∈ (a, b) such that
lim F (x0 + iy) = γ . (5.107)
y→0+
Then
f (x0 ) = γ, distributionally. (5.108)
In fact, (5.108) is a point value of the first order, that is,
Z x
1
lim f (t) dt = γ . (5.109)
x→x0 x − x0 x
0
Proof. We shall first show that it is enough to prove the result if the
rectangular region is the upper half-plane H = {z ∈ C : Im z > 0} . Indeed,
let C be a smooth simple closed curve contained in (a, b) × [0, R) such that
C ∩ (a, b) = [x0 − η, x0 + η] , and which is symmetric with respect to the
line Re z = x0 . Let ϕ be a conformal bijection from H to the region
enclosed by C such that the image of the line Re z = x0 is contained in
Re z = x0 , so that, in particular, ϕ (x0 ) = x0 . Then (5.107) holds if and
only if F ◦ ϕ (x0 + iy) → γ as y → 0+ , while (5.108) and (5.109) hold if and
only if the corresponding equations hold for a distribution given locally as
f ◦ ϕ near x = x0 .
In the sequel we will use the weak* topology of L∞ . This is the topology
on L∞ determined by the family of seminorms f → pϕ (f ) = |hf, ϕi|, f ∈
L∞ , where ϕ ∈ L1 (R).
Therefore we shall assume that a = −∞, and b = R = ∞. In this case, f

belongs to the Hardy space H ∞ , the closed subspace of L∞ (R) consisting of
the boundary values of bounded analytic functions on H ([82]); moreover,
one easily verifies that H ∞ is a weak* closed subspace of L∞ , this fact
will be used below. Let fε (x) = f (x0 + εx) . Then the set
1
′ {fε : ∞
ε 6= 0}
is weak* bounded (as a subset of the dual space L (R) = L (R))
∞
and, consequently, a relatively weak* compact set. Suppose that {εn }n=0
is a sequence of non-zero numbers with εn → 0 such that the sequence
∞
{fεn }n=0 is weak* convergent to g ∈ L∞ (R) . It will be shown that g ≡ γ.
Since g ∈ H ∞ , we can write it as g (x) = G (x + i0) where G is a bounded
analytic function in H, then the weak* convergence of fεn to g implies

that F (x0 + εn z) converges to G (z) uniformly on compacts of H, and thus
G (iy) = γ for all y > 0. It follows that G ≡ γ, and so g ≡ γ. Since any
∞
sequence {fεn }n=0 with εn → 0 has a weak* convergent subsequence, and
since that subsequence converges to the constant function γ, we conclude
that fε → γ in the weak* topology of L∞ (R) .
That f (x0 ) = γ, distributionally, is now clear, because D (R) ⊂ L1 (R) .
On the other hand, (5.109) follows by taking x = x0 + ε and φ (t) =

χ[0,1] (t) , the characteristic
R ∞function of the unit interval, in the limit
limε→0 hfε (t) , φ (t)i = γ −∞ φ (t) dt, which in view of the previous argu-
ment holds now for φ ∈ L1 (R).
We can now prove our Tauberian theorem.
Theorem 5.15. Let F be analytic in a rectangular region of the form

(a, b) × (0, R) . Suppose f (x) = limy→0+ F (x + iy) in the space D′ (a, b) .
Let x0 ∈ (a, b) such that limy→0+ F (x0 + iy) = γ. If f is distributionally
bounded at x = x0 then f (x0 ) = γ, distributionally.
Proof. There exists n ∈ N and a function fn bounded in a neighborhood

of x0 such that f = ∂xn0 fn ; notice that f (x0 ) = γ, distributionally, if and
only if fn (x0 ) = γ, distributionally. But fn (x) = Fn (x + i0) in D′ (a, b),
where Fn is analytic in (a, b)×(0, R) ; here Fn is the only angularly bounded
solution of F (z) = ∂xn0 Fn (z) (derivatives with respect to z). Since fn is
bounded near x = x0 , Fn is also bounded in a rectangular region of the form
(a1 , b1 ) × (0, R1 ) , where x0 ∈ (a1 , b1 ) . Clearly limy→0+ Fn (x0 + iy) = γ, so
the Theorem 5.14 yields fn (x0 ) = γ, distributionally, as required.
Observe that in general the result (5.109) does not follow if f is not
bounded but just distributionally bounded near x0 .
The condition (5.107) may seem weaker than the angular convergence
of F (z) to γ as z → x0 , however, if F is angularly bounded, which is the
case if f is distributionally bounded at x = x0 , then angular convergence
and radial convergence are equivalent. In fact [27] both conditions are
equivalent to the existence of an arc κ : [0, 1] −→ (a, b) × [0, R) such that
κ ([0, 1)) ⊂ {z ∈ C : Im , z ≥ m |Re z − x0 |} for some m > 0 and such that
κ (1) = x0 , for which
lim F (κ (t)) = γ . (5.110)
t→1−
Therefore, we may use a conformal map to obtain the following general

form of the Theorem 5.15.
Theorem 5.16. Let C be a smooth part of the boundary ∂Ω of a region Ω of

the complex plane. Let F be analytic in Ω, and suppose that f ∈ D′ (C) is the
distributional boundary limit of F. Let ξ0 ∈ C and suppose that κ is an arc in
Ω that ends at ξ0 and that approaches C angularly. If limt→1− F (κ (t)) = γ
and f is distributionally bounded at ξ = ξ0 , then f (ξ0 ) = γ, distributionally.
Theorem 5.15 may also be used to obtain Littlewood type Tauberian

results for distributions. The first corollary is also contained in the gen-
eral theorem of Vladimirov, Drozhzhinov, and Zavyalov [192], but it is
convenient to state it here for future applications. We will use Corollary
5.17 to produce in 5.2.7 a simple proof of the celebrated Hardy-Littlewood
Tauberian theorem for the converse of Abel’s theorem.
Corollary 5.17. Let g be a tempered distribution supported on [0, ∞). Sup-

pose that

lim g(x), e−yx = γ . (5.111)
y→0+
Then, the Tauberian condition

1
g(λx) = O as λ → ∞ in D′ (R) (5.112)
λ
implies that g has the quasi-asymptotic behavior
q δ(x)
g(λx) ∼ γ as λ → ∞ in S ′ (R) . (5.113)
λ
Proof. Let f be such that fˆ

= g. Then (5.111) translates into F (iy) → γ
+ izx
as y → 0 , where F (z) = g(x), e (hence f (x) = F (x + i0)) and
(5.112) corresponds to the statement f distributionally bounded at x = 0,
by Theorem 5.15, we have that f (0) = γ, distributionally. Thus, Fourier
inverse transform yields (5.113).
Corollary 5.18. Let g be a tempered distribution supported on [0, ∞)

and φ ∈ E(R). Suppose that hg(x), φ(x)i = γ (A) . Then, the Tauberian
condition
1
φ(λx)g(λx) = O as λ → ∞ in D′ (R)
λ
implies that hg(x), φ(x)i = γ (C).
Proof. Corollary 5.17 gives that

q δ(x)
φ(λx)g(λx) ∼ γ as λ → ∞ in S ′ (R),
λ
which by Proposition 5.3 implies that hg(x), φ(x)i = γ (C) .
5.2.7 Application: Littlewood’s Tauberian theorem
We now discuss a non-trivial application of Theorem 5.15. Our application

is a proof of a famous Tauberian theorem of Hardy and Littlewood. This
distributional proof was originally found in [177]. In fact, we give a proof
for the version proposed by Littlewood but first proved by Ananda Rau [2].
We begin with a lemma whose proof can be tracked down to the proof of
the original first Tauber’s theorem ([66] , [168]).
Lemma 5.5. Let hbn i∞ n=0 be a sequence of complex numbers. Suppose that
∞
hλn in=0 is an increasing sequence of non-negative real numbers such that
λn → ∞ as n → ∞. If

λn − λn−1
bn = O , (5.114)
λn
then,
∞
X X
bn e−λn y − bn = O(1), as y → 0+ . (5.115)
n=0 λn < y1
Proof. Choose M such that |bn | ≤ M λ−1 n (λn − λn−1 ), for every n. Then,

X
∞ X X X
b e −λn y
− b ≤ |b | 1 − e −λn y
+ |bn | e−λn y
n n n
n=0 λn < 1 λn < 1 1
≤λn
y y y
X X
≤ My (λn − λn−1 ) + M y (λn − λn−1 ) e−λn y
λn < y1 1
y ≤λn
Z ∞
= O(1) + M y e−yt dt = O(1), y → 0+ ,
1
y
as required.
P
Recall (cf. 5.1.1) that a series ∞n=0 cn is (A, hλn i) summable to γ if
P∞ −λn y
n=0 cn e → γ as y → 0+ . When λn = n we obtain the notion of
Abel summability. Then we have the ensuing Hardy-Littlewood Tauberian
theorem.
∞
Theorem 5.17. Suppose that hλn in=0 is an increasing sequence of non-
negative real numbers such that λn → ∞, as n → ∞. If
∞
X
cn = γ (A, λn ) , (5.116)
n=0
P∞
and cn = O λ−1
n (λn − λn−1 ) , then n=0 cn = γ.
Proof. The plan of the proof is to associate to the series the tempered
P∞ q
distribution g(x) = n=0 cn δ(x − λn ) and show that g(λx) ∼ γδ(λx), as
λ → ∞ in S ′ (R), based on this conclusion, we will deduce the convergence
of the series.
Let us first verify that f defines a tempered distribution; indeed from

P
Lemma 5.5 and the assumption (5.116), we have that G(x) = λn <x cn
is a bounded function, hence g is a tempered distribution; furthermore,
G(λx) = O(1) in S ′ (R), and therefore by differentiating G(λx) with respect
to x, we obtain that λg(λx) is bounded in S ′ (R). Therefore, by Corollary
5.17, we obtain that

δ(x) 1
g(λx) = γ +o , in S ′ (R) . (5.117)
λ λ
So, we have that
∞
X
λn
lim cn φ = γ, for each φ ∈ D(R) . (5.118)
λ→∞
n=0
λ
To conclude the proof, we take in (5.118) suitable test functions. Let η > 0
and let us choose the test function φ ∈ D(R) such that 0 ≤ φ ≤ 1, φ(x) = 1
for x ∈ [0,R1], supp φ ⊆ [−1, 2], φ is decreasing on the interval (1, 2), and
2
such that 1 φ(x)dx < η. Then

XN X λn − λn−1 λn

lim sup cn − γ ≤ lim sup φ
N →∞ n=0 N →∞ λN λN
λN <λn ≤2λN
Z 2
≤ φ(x) dx < η .
1
∞
X
Since η was arbitrary, we conclude that cn = γ.
n=0
5.2.8 Solution to the Hardy-Littlewood (C) summability problem

for distributions
As an application of Theorem 5.1, we now formulate and solve the so-called

Hardy-Littlewood (C) summability problem in the context of tempered
distributions. This classical problem aims to characterize trigonometric
series, in sine-cosine form, which are (C) summable to some value at a
point x = x0 , that is,
∞
a0 X
+ (an cos nx0 + bn sin nx0 ) = γ (C, m) ,
2 n=1
for some γ and m ∈ N0 . One also imposes the restrictions an = O(nk ) and
bn = O(nk ), for some k; thus, the trigonometric series represents a tempered
distribution! The problem for trigonometric series was first formulated by
Hardy and Littlewood in [68]; a complete treatment with historical remarks
is found in [206] (see also [56] )
In order to formulate the problem for tempered distributions, we need

the following summability notion for distributional evaluations.
Definition 5.8. Let g ∈ D′ (R), φ ∈ E(R), and m ∈ N0 . We say that the

principal value evaluation p.v. hg(x), φ(x)i exists and is equal to γ in the
Cesàro sense of order m, and write
p.v. hg(x), φ(x)i = γ (C, m), (5.119)
if some first order primitive G of φg, i.e., G′ = φg, satisfies
lim (G(x) − G(−x)) = γ (C, m) . (5.120)

x→∞
Note that e.v. hg(x), φ(x)i = γ (C, m) implies p.v. hg(x), φ(x)i = γ (C, m),
as the reader can easily verify. On the other hand the converse is not
true; take for example p.v. hx, 1i = 0 (C, 0), but clearly the evaluation
e.v. hx, 1i (C) does not exist.
When g = µ is a Radon measure, we write

Z ∞
p.v. φ(x)dµ(x) = γ (C, m) (5.121)
−∞
for (5.119). Observe that (5.121) explicitly means that

Z x m
t
lim φ(t) 1 − dµ(t) = γ . (5.122)
x→∞ −x |x|
P∞
If µ = n=−∞ cn δ( · − n) and φ ≡ 1, then we write (5.121) as
∞
X
p.v. cn = γ (C, m), (5.123)
n=−∞
which is equivalent to have

∞
X
c0 + (cn + c−n ) = γ (C, m) . (5.124)
n=1
P∞ inx0
Example II.6. Consider the trigonometric series n=−∞ cn e then
∞
X
p.v. cn einx0 = γ (C, m)
n=−∞
if and only if
∞
a0 X
+ (an cos nx0 + bn sin nx0 ) = γ (C, m),
2 n=1
with an = cn + c−n and bn = cn − c−n .
We can now formulate our problem: we want to characterize tempered

distributions f such that
1 D E
p.v. fˆ(x), eix0 x = γ (C) . (5.125)
2π
We study some properties of the principal value evaluations in the (C)

sense. They admit a quasi-asymptotic characterization, but unlike e.v.
Cesàro evaluations, the existence of p.v. hg(x), φ(x)i = γ (C) does not imply
that φg ∈ S ′ (R). We first need the following lemmas.
Lemma 5.6. Let g ∈ D′ (R) be an even distribution. There exists h ∈

D′ (R) such that supp h ⊆ [0, ∞) and g(x) = h(x) + h(−x).
Proof. Decompose g = g− + g+ , where supp g− ⊆ (−∞, 0] and

supp g+ ⊆ [0, ∞). The parity of g implies that g+ (x) − g− (−x) is con-
centrated at the origin, and so there exist constants such that
n
X
g− (x) = g+ (−x) + aj δ (j) (x), (5.126)
j=0
Pn
Since, g(x)−g+ (−x)−g+ (x) = aj δ (j) (x) is even, it follows that aj = 0
j=0
Pk
whenever j is odd. So, n = 2k, and hence h = g+ + (1/2) j=0 a2j δ (2j)
satisfies the requirements.
Lemma 5.7. Let g ∈ D′ (R) be an even distribution. Then

δ(x) 1
g(λx) = γ +o as λ → ∞ in D′ (R) (5.127)
λ λ
if and only if any h ∈ S ′ (R) such that supp h ⊆ [0, ∞), and g(x) =
h(x) + h(−x), satisfies

γδ(x) 1
h(λx) = +o as λ → ∞ in S ′ (R). (5.128)
2λ λ
Proof. The converse is clear. On the other hand take h as in Lemma

5.6. Proposition 5.4 implies the existence of m such that
γx2m−1 x2m−1
h(−2m) (x) = + c(x) + o(x2m−1 )
2(2m − 1)! (2m − 1)!
and
γx2m−1 x2m−1
h(−2m) (x) = − c(|x|) + o(x2m−1 ),
2(2m − 1)! (2m − 1)!
x → ∞, but comparison between the last two expressions gives that c(x) =
o(1), and hence
γx2m−1
h(−2m) (x) ∼ , x → ∞,
2(2m − 1)!
which implies (5.128).
Proposition 5.7. Let g ∈ D′ (R) and φ ∈ E(R). Then,
p.v. hg(x), φ(x)i = γ (C) (5.129)
if and only if

δ(x) 1
φ(−λx)g(−λx) + φ(λx)g(λx) = 2γ +o as λ → ∞ in S ′ (R) ;
λ λ
(5.130)
if and only if for any decomposition g = g− + g+ , where supp g− ⊆ (−∞, 0]
and supp g+ ⊆ [0, ∞),

δ(x) 1
φ(−λx)g− (−λx) + φ(λx)g+ (λx) = γ +o as λ → ∞ in S ′ (R) .
λ λ
(5.131)
In particular, we obtain that φ(−x)g(−x) + φ(x)g(x) ∈ S ′ (R).
Proof. Assume that φ ≡ 1. We have that g(−x) + g(x) is an even

distribution, then, by Lemma 5.6, we can find h with supp h ⊆ [0, ∞)
such that g(−x) + g(x) = h(x) + h(−x). It is easy to see that (5.129) is
equivalent to limx→∞ h(−1) (x) = γ (C) which holds if and only if (5.128)
(with γ replaced by γ/2), and by Lemma 5.7, it is equivalent to (5.130).
The equivalence with (5.131) follows by taking h(x) = g− (−x) + g+ (x).
The right notion to characterize (5.125) is that of distributional sym-

metric point values.
Definition 5.9. Let f ∈ D′ (R). We say that f has a (distributional)

symmetric point value at x = x0 if the following quasi-asymptotic limit
holds
lim (f (x0 − εx) + f (x0 + εx)) = γ, in D′ (R). (5.132)

ε→0
In this case we write fsym (x0 ) = γ, distributionally.
Observe that if we define χfx0 (x) = f (x0 − x) + f (x0 + x), the symmetric
part of f about the point x = x0 , then (5.132) holds if and only if χfx0 (0) =
γ, distributionally. In addition, note that if χfx0 ∈ S ′ (R), then (5.132)
actually holds in the space S ′ (R) (cf. Theorem 2.35). We have set the
ground to solve our problem. The following theorem is the solution to the
Hardy-Littlewood (C) problem for tempered distributions.
Theorem 5.18. Let f ∈ S ′ (R). Then

1 D E
p.v. fˆ(x), eix0 x = γ (C) (5.133)
2π
if and only if fsym (x0 ) = γ, distributionally.
Proof. By definition fsym (x0 ) = γ, distributionally, if and only if,

f (x0 − εx) + f (x0 + εx) = γ + o(1) as ε → 0+ in S ′ (R),
which, by taking Fourier transform, is equivalent to

−iλx0 x ˆ iλx0 x ˆ δ(x) 1
e f (−λx) + e f (λx) = 2πγ +o as λ → ∞ in S ′ (R),
λ λ
and, by Proposition 5.7, the latter is equivalent to (5.133).
We immediately obtain, by Theorem 5.18 and Example II.6, the fol-

lowing result of Hardy and Littlewood. Naturally, the language in the
original statement differs from ours, at that time distribution theory and
quasi-asymptotics did not even exist!
Corollary 5.19. Let f ∈ S ′ (R) be a 2π periodic distribution having Fourier

series, in sine-cosine form,
∞
a0 X
+ (an cos nx + bn sin nx) .
2 n=1
Then,
∞
a0 X
+ (an cos nx0 + bn sin nx0 ) = γ (C)
2 n=1
if and only if fsym (x0 ) = γ, distributionally.
We end this section by showing two Abelian results.
Theorem 5.19. Let g ∈ D′ (R) and φ ∈ E(R). If

p.v. hg(x), φ(x)i = γ (C),
then,
hg(x), φ(x)i = γ (A).
Proof. Take g− and g+ as in Proposition 5.7, then, by (5.131), as λ → ∞,

x
x
φ(x)g− (x), e λ + φ(x)g+ (x), e− λ

= λ φ(−λx)g− (−λx) + φ(λx)g+ (λx), e−x

= γ δ(x), e−x + o(1)
= γ + o(1).
For symmetric point values, we get a radial version of Theorem 5.6.
Theorem 5.20. Let f ∈ D′ (R). Let U be a harmonic representation of f

on the upper half-plane Im z > 0. If fsym (x0 ) = γ, distributionally, then
lim U (x0 + iy) = γ . (5.134)
y→0+
Proof. As in the proof of Theorem 5.6, we may assume that f is

a tempered distribution. If fˆ = fˆ+ + fˆ− is a decomposition such that
suppfˆ− ⊆ (−∞, 0] and suppfˆ+ ⊆ [0, ∞), we can assume that
1 D ˆ E D E
U (z) = f− (t), eiz̄t + fˆ+ (t), eizt , Im z > 0.
2π
But in this case, Theorem 5.19 yields (5.134).
Bibliography
[1] M. Aguirre-Téllez, The expansion of δ (k−1) (m2 + P ), Integral Transform

Spec. Funct. 8 (1999), 139–148.
[2] K. Ananda Rau, On the converse of Abel’s theorem, J. London Math. Soc.
3 (1928), 349–355.
[3] P. Antosik, J. Mikusiński, R. Sikorski, Theory of Distributions, The
Sequential Approach, Polish Scientific Publishers, Warszawa, 1973.
[4] J. M. Ash, P. Erdős, L. A. Rubel, Very slowly varying functions, Aequa-
tiones Math. 10 (1974), 1–9.
[5] E. J. Beltrami, M.R. Wohlers , Distributions and the Boundary Values of
Analytic Functions, Academic Press, New York, 1966.
[6] J. J. Benedetto, Analytic representation of generalized functions, Math. Z.
97 (1967), 303–319.
[7] J. J. Benedetto, Spectral synthesis, Academic Press, New York, 1975.
[8] S. Berceanu, A. Gheorghe, On the asymptotics of distributions with support
in a cone, J. Math. Phys. 26 (1985), 2335–2341.
[9] N. H. Bingham, C. M. Goldie, J. L. Teugels, Regular Variation, Cambridge
Univ. Press, Cambridge, 1989.
[10] N. N. Bogolubov, V. S. Vladimirov, A. N. Tavkelidze, On automodel asymp-
totics in quantum field theory, Proc. Steklov Inst. Math. 135 (1978), 27–54.
[11] R. Bojanić, J. Karamata, On a class of functions of regular asymptotic
behavior, Math. Res. Centre, U.S. Army, Madison, Wis., Tech. Summary
Rep. No. 436 (1963).
[12] N. Bourbaki, Topologie Générale, Chapitres 1, 2, Hermann, Paris, 1951.
[13] H. Bremermann, Distributions, Complex Variables and Fourier Transforms,
Addison-Wesley, Reading, Massachusetts, 1965.
[14] N. G. de Bruijn, Asymptotic Methods in Analysis, Dover Publications, Inc.,
New York, 1981.
[15] Yu. A. Brychkov, Asymptotic Expansions of Generalized Functions I,
Theoret. Mat. Fiz. 5 (1970), 98–109 (In Russian).
[16] Yu. A. Brychkov, On Asymptotic Expansions of Generalized Functions,
Mat. Zametki 12 (1972), 131–138 (In Russian).
283
284 Bibliography
[17] Yu. A. Brychkov, Asymptotic expansions of generalized functions II,

Theoret. Mat. Fiz. 15 (1973), 375-381 (In Russian).
[18] Yu. A. Brychkov, Asymptotic expansions of generalized functions III,
Theoret. Mat. Fiz. 23 (1975), 191-198 (In Russian).
[19] Yu. A. Brychkov, A. P. Prudnikov, Integral Transforms of Generalized
Functions, Gordon and Breach Sci. Publ., New York, 1989.
[20] Yu. A. Brychkov, Yu. M. Shirokov, Asymptotic behavior of the Fourier
transforms, Teoret. Mat. Fiz. 4 (1970), 301–309 (In Russian).
[21] J. Campos Ferreira, Introduction to the Theory of Distributions, Longman,
Essex, 1997.
[22] R. D. Carmichael, Abelian Theorems for the Stieltjes Transform of func-
tions, Bull. Calcutta Math. Soc. 68 (1976), 49–52.
[23] R. D. Carmichael, E. O. Milton, Abelian theorems for the distributional
Stieltjes transform, J. Math. Anal. Appl. 72 (1979), 195–205.
[24] R. D. Carmichael, R. S. Pathak, Asymptotic analysis of the H-function
transform, Glasnik Math. 25 (1990), 103–127.
[25] K. Chandrasekharan, S. Minakshisundaram, Typical Means, Oxford Uni-
versity Press, 1952.
[26] F. Constantinescu, Boundary values of analytic functions, Comm. Math.
Phys. 7 (1968), 225–233.
[27] J. B. Conway, Functions of One Complex Variable II, Springer Verlag, New
York, 1995.
[28] J. Čižek, J. Jelinek, A Tauberian theorem for distributions, Comment.
Math. Univ. Carolinae 37 (1996), 479–488.
[29] H. Delange, Sur les théorèmes inverses des procédés de somation des séries
divergentes, I, II, Ann. Sci. Ecole Norm Sup. 67 (1950), 99–160, and 199–
242.
[30] G. Doetsch, Handbuch der Laplace Transformation, I, Birkhäuser, Basel,
1950.
[31] W. F. Donoghue, Distributions and Fourier transforms, Academic Press,
New York, 1969.
[32] Yu. N. Drozhzhinov, B. I. Zavyalov, Quasi-asymptotics of generalized func-
tions and Tauberian theorems in the complex domain, Mat. Sb. 102 (1977),
372–390 (In Russian). English transl. Math. USSR Sb. 36 (1977).
[33] Yu. N. Drozhzhinov, B. I. Zavyalov, Tauberian theorems for generalized
functions with supports in cones, Mat. Sb. 108 (1979), 78–90 (In Russian).
[34] Yu. N. Drozhzhinov, B. I. Zavyalov, Asymptotic properties of some classes
of generalized functions, Izv. Akad. Nauk. SSR, Ser. Mat. 49 (1985), 81–140
(In Russian).
[35] Yu. N. Drozhzhinov, B. I. Zavyalov, Multidimensional Abelian and Taube-
rian comparing theorems, Mat. Sb. 180 (1989), 1234–1258 (In Russian).
[36] Yu. N. Drozhzhinov, B. I. Zavyalov, Local Tauberian theorems in spaces of
distributions related to cones, and their applications, Izvestiya RAN. Ser.
Mat. 61:6 (1997), 59–102.
[37] Yu. N. Drozhzhinov, B. I. Zavyalov, A Wiener-type Tauberian theorem for
tempered distributions, Mat. Sb. 189 (1998), 91–130.
Bibliography 285
[38] Yu. N. Drozhzhinov, B. I. Zavyalov, Tauberian-type theorems for a general-

ized multiplicative convolution, Izv. Ross. Akad. Nauk Ser. Mat. 64 (2000),
37–94 (In Russian). Translation in Izv. Math. 64 (2000), 35-92.
[39] Y. N. Drozhzhinov, B. I. Zavyalov, Multidimensional Tauberian theorems
for Banach-space valued generalized functions, Sb. Math. 194 (2003), 1599–
1646.
[40] Yu. N. Drozhzhinov, B. I. Zavyalov, Asymptotically homogeneous general-
ized functions and boundary properties of functions holomorphic in tubular
cones, Izv. Math. 70 (2006), 1117–1164.
[41] Yu. N. Drozhzhinov, B. I. Zavyalov, Applications of Tauberian theorems
in some problems of mathematical physics, Teoret. Mat. Fiz. 157 (2008),
373–390 (In Russian). Translation in Theoret. and Math. Phys. 157 (2008),
1678–1693.
[42] Yu. N. Drozhzhinov, B. I. Zavyalov, Generalized functions asymptotically
homogeneous with respect to special transformation groups, Sb. Math. 200
(2009), 803-844.
[43] A. Duran, Laguerre Expansions of Tempered Distributions and Generalized
Functions, J. Math. Anal. Appl. 150 (1990), 166–180.
[44] A. Erdely, Asymptotic Expansions, Dover Publ., New York, 1956.
[45] R. Estrada, Characterization of the Fourier series of distributions having a
value at a point, Proc. Amer. Math. Soc. 124 (1996), 1205–1212.
[46] R. Estrada, Regularization of distributions, Internat. J. Math. & Math.
Sci. 21 (1998), 625–636.
[47] R. Estrada, The Cesàro behaviour of distributions, Proc. R. Soc. Lond. A
454 (1998), 2425–2443.
[48] R. Estrada, Boundary values of analytic functions without distributional
point values, Tamkang J. Math. 35 (2004), 53–60.
[49] R. Estrada, A Distributional Version of the Ferenc-Lukács Theorem, Sara-
jevo J. Math. 1 (2005), 75–92.
[50] R. Estrada, S. A. Fulling, Distributional asymptotic expansion of spectral
functions and of the associated Green kernels, Electronic J. Diff. Eqns. 6
(1999), 1–37.
[51] R. Estrada, R. P. Kanwal, Moment sequences for a certain class of distri-
butions, Compl. Variabl. 9 (1987), 31–39.
[52] R. Estrada, R. P. Kanwal, A distributional theory of asymptotic expan-
sions, Proc. Roy. Soc. London Ser. A 428 (1990), 399–430.
[53] R. Estrada, R. P. Kanwal, The asymptotic expansion of certain multidi-
mensional generalized functions, J. Math. Anal. Appl. 163 (1992), 264–283.
[54] R. Estrada, R. P. Kanwal, Taylor expansions for distributions, Math.
Methods Appl. Sci. 16 (1993), 297–304.
[55] R. Estrada. R. P. Kanwal, Asymptotic separation of variables, J. Math.
Anal. Appl. 178 (1993), 130–142.
[56] R. Estrada, R. P. Kanwal, A Distributional Approach to Asymptotics.
Theory and Applications, Second Edition, Birkhäuser, Basel, 2002.
[57] R. Estrada, J. Vindas, Determination of jumps of distributions by
differentiated means, Acta Math. Hungar. 124 (2009), 215–241.
286 Bibliography
[58] R. Estrada, J. Vindas, On the Point Behavior of Fourier Series and Conju-
gate Series, Z. Anal. Anwend. 29 (2010), 487–504.
[59] R. Estrada, J. Vindas, A General Integral, preprint, 2011.
[60] W. Feller, An Introduction to Probability Theory and its Applications, I,
Wiley, New York, 1968.
[61] A. Friedman, Generalized functions and partial differential equations,
Prentice-Hall, New Jercey, London, 1963.
[62] T. H. Ganelius, Tauberian remainder theorems, Lect. Not. Math. 232,
Springer, Berlin, 1971.
[63] I. M. Gelfand, G. E. Shilov, Generalized Functions 1, Academic Press, New
York, 1964.
[64] J. L. Geluk, L. De Haan, Regular variation Extensions and Tauberian
Theorems, CWI Tracts 40, Amsterdam, 1987.
[65] H. J. Glaske, D. Müller, Abelsche Sätze für die Laplace-Transformation von
Distribution, Z. Anal. Anw. 10 (1991), 232–238.
[66] G. H. Hardy, Divergent Series, Clarendon Press, Oxford, 1949.
[67] G. H. Hardy, J. E. Littlewood, Some theorems concerning Dirichlet’s series,
Messenger of Mathematics 43 (1914), 134–147.
[68] G. H. Hardy, J. E. Littlewood, Solution of the Cesàro summability problem
for power series and Fourier series, Math. Z. 19 (1923), 67–96.
[69] L. G. Hernández, R. Estrada, Solution of ordinary differential equations by
series of delta functions, J. Math. Anal. Appl. 191 (1995), 40–55.
[70] L. Hörmander, Linear partial differential operators I, II, Springer, Berlin,
1963.
[71] I. Imai, Applied Hyperfunction Theory, Kluwer, Dordrecht, 1992.
[72] D. S. Jones, Generalized Functions, McGraw-Hill, New York, Toronto,
Ont.-London, 1966.
[73] A. Kaneko, Representation of hyperfunctions by measures and some of its
applications, J. Fac. Sci. Univ. Tokyo, Sec. 1A, 19 (1972), 321–352.
[74] A. Kaneko, On the Structure of Fourier Hyperfunctions, Proc. Japan Acad.
48 (1972), 651–653.
[75] A. Kaneko, Introduction to Hyperfunctions, Kluwer, Dordrecht, 1988.
[76] J. Karamata, Sur un mode de croissance regulière des functions, Mathe-
matica (Cluj) 3 (1930), 33–48.
[77] J. Karamata, Sur un mode de croissance régulière. Théorèmes fondamen-
taux, Bull. Soc. Math. France 61 (1933), 55–62.
[78] T. Kawai, On the theory of Fourier hyperfuctions and its applications to
partial differential equations with constant coefficients, J. Fac. Sci. Univ.
Tokyo, 1A, 17 (1970), 467–517.
[79] H. Komatsu, Ultradistributions I, J. Fac. Sci. Univ. Tokyo Sect. 1A Mat.
20 (1973), 25–105.
[80] H. Komatsu, Laplace transform of hyperfunctions-A new fundation of the
Heaviside calculus, J. Fac. Sci. Univ. Tokyo IA 34 (1987), 805–820.
[81] H. Komatsu, Microlocal analysis in Gevrey classes and in complex domains,
Lect. Not. Math. 1726, 426–493, Springer-Verlag, Berlin, 1989.
[82] P. Koosis, Introduction to Hp spaces, London Mathematical Society
Bibliography 287
Lecture Note Series, 40. Cambridge University Press, Cambridge-New York,

1980.
[83] J. Korevaar, T. van Aardenne-Ehrenfest, N. G. de Bruijn, A note on slowly
oscillating functions, Nieuw. Arch. Wisk. 23 (1949), 77–88.
[84] J. Korevaar, Distribution proof of Wiener Tauberian theorem, Proc. Amer.
Math. Soc. 16 (1965), 353–355.
[85] D. Kovačević, S. Pilipović, On Homogeneous Generalized Functions, Publ.
Math. Debrecen 41 (1992), 227–287.
[86] D. Kovačević, S. Pilipović, Structural properties of the space of tempered
ultradistributions, in: Proc. Conf. Complex Analysis and Generalized Func-
tions (Varna, 1991), 169–184, 1993.
[87] E. Landau, Sur les valueurs moyennes de certains functions arithmétiques,
Bull. Acad. Roy. Belgique (1911), 443–472.
[88] J. Lavoine, Sur des théoèmes abéliens et taubériens de la transformation
de Laplace, Ann. Inst. Henri Poincare Nouv. S. 4 (1966), 49–65.
[89] J. Lavoine, O. P. Misra, Théorèmes Abéliens pour la transformation de
Stieltjes des distributions, C. R. Acad. Sci. Paris 279 (1974), 99–102.
[90] J. Lavoine, O. P. Misra, Abelian theorems for the distributional Stieltjes
transformation, Math. Proc. Cambridge Philos. Soc. 86 (1979), 287–293.
[91] M. J. Lighthill, Introduction to Fourier Analysis and Generalized Functions,
Cambridge Univ. Press, London, 1958.
[92] J. E. Littlewood, The converse of Abel’s theorem on power series, Proc.
London Math. Soc. 9 (1911), 434–448.
[93] S. Lojasiewicz, Sur la valeur et la limit d’une distribution dans un point,
Studia Math. 16 (1957), 1–36.
[94] S. Lojasiewicz, Sur la fixation des variables dans une distribution, Studia
Math. 17 (1958), 1–64.
[95] M. Mangad, Asymptotic expansions of Fourier transforms and discrete
polyharmonic Green’s functions, Pacific J. Math. 20 (1967), 85–98.
[96] V. Marić, M. Skendžić, A. Takači, On Stieltjes transform of distributions
behaving as regularly varying functions, Acta. Sci. Math. Szeged 50 (1986),
405–410.
[97] T. Matsuzawa, A Calculus Approach to Hyperfunctions, II, Transactions
of the American Mathematical Society 313 (1989), 619–654.
[98] O. P. Misra, Some Abelian theorems for distributional Stieltjes transform,
J. Math. Anal. Appl. 39 (1972), 590–599.
[99] M. Nedeljkov, S. Pilipović, Generalized Solutions to a Semilinear Hyper-
bolic System with a non-Lipshitz Nonlinearity, Monatsh. Math. 125 (1998),
255–261.
[100] D. Nikolić-Despotović, S. Pilipović, Tauberian theorem for the distribu-
tional Stieltjes transform, Internat. J. Math. Math. Sci. 9 (1986), 521–524.
[101] D. Nikolić-Despotović, S. Pilipović, The quasiasymptotic expansion of tem-
pered distributions and the Stieltjes transform, Univ. u Novom Sadu Zb.
Rad. Prirod.-Mat. Fak. Ser. Mat. 18 (1988), 31–44.
[102] D. Nikolić-Despotović, S. Pilipović, The quasiasymptotic expansion at the
origin. Abelian-type results for the Laplace and Stieltjes transforms, Math.
288 Bibliography
Nachr. 174 (1995), 231–238.

′
[103] N. Ortner, P. Wagner, Applications of weighted DL p -spaces to the convo-
lution of distributions, Bull. Polish Acad. Sci. Math. 37 (1989), 579–595.

[104] R. S. Pathak, S. K. Misra, Asymptotic behavior of a class of integral trans-
forms in complex domains, Proc. Edinburgh Math. Soc. 32 (1989), 231–247.
[105] J. Peetre, On the value of a distribution at a point, Portugal. Math. 27
(1968), 149–159.
[106] S. Pilipović, On the S-asymptotics of tempered distributions and K1′ distri-
butions, Part I, Univ. u Novom Sadu Zb. Rad. Prirod.-Mat. Fak. Ser. Mat.
15 (1985), 47–58.
[107] S. Pilipović, On the S-asymptotics of tempered distributions and K1′ dis-
tributions, Part II, Univ. u Novom Sadu Zb. Rad. Prirod.-Mat. Fak. Ser.
Mat. 15 (1985), 59–67.
[108] S. Pilipović, Asymptotic behavior of the distributional Weierstrass trans-
form, Appl. Anal. 25 (1987), 171–179.
[109] S. Pilipović, Remarks on supports of distributions, Glasnik Mat. 22 (1987),
375–380.
[110] S. Pilipović, On the quasiasymptotic behavior of the Stieltjes transform of
distributions, Publ. Inst. Math. Beograd 40 (1987), 143–152.
[111] S. Pilipović, Some properties of the quasi-asymptotics of Schwartz distri-
butions, Part I, Quasiasymptotic at ±∞, Publ. Inst. Math. Beograd 43
(1988), 125–130.
[112] S. Pilipović, Some properties of the quasiasymptotic of Schwartz distri-
butions, Part II-Qiasiasymptotic at 0, Publ. Inst. Math. Beograd 43 (57)
(1988), 131–135.
[113] S. Pilipović, On the quasiasymptotic of Schwartz distributions, Math.
Nachr. 137 (1988), 19–25.
[114] S. Pilipović, On the S-asymptotics of tempered distributions and K1′ -
distributions, Part III, Univ. u Novom Sadu, Zb. Rad. Prirod.-Mat. Fak.
Ser. Mat. 18 (1988), 179–189.
[115] S. Pilipović, On the S-asymptotics of tempered and K1′ distributions, Part
IV. S-asymptotic and the ordinary asymptotic, Univ. u Novom Sadu Zb.
Rad. Prirod.-Mat. Fak., Ser. Mat. 18 (1988), 191–195.
[116] S. Pilipović, On the Behavior of Distribution at the Origin, Math. Nachr.
141 (1989), 27–32.
′
[117] S. Pilipović, Quasiasymptotic expansion of distributions from S+ and the
asymptotic expansion of the distributional Stieltjes transform, Internat. J.
Math. Sci. 12 (1989) 673–684.
[118] S. Pilipović, Asymptotic expansion of Schwartz’s distributions, Publ. Inst.
Math. Beograd 45 (1989), 119–127.
[119] S. Pilipović, Generalized S-asymptotics, Tauberian Theorem, Publ. Inst.
Math. Beograd 48 (1990), 129–132.
[120] S. Pilipović, The translation asymptotic and the quasiasymptotic behavior
of distributions, Acta. Math. Hungarica 55 (1990), 239–243.
[121] S. Pilipović, On the behavior of distributions at infinity, Wiener-Tauberian
type results, Publ. Inst. Mat. Beograd, 48 (1990), 129–132.
Bibliography 289
[122] S. Pilipović, Quasiasymptotic expansion and the Laplace transform, Appl.

Anal. 35 (1991), 247–261.
[123] S. Pilipović, Characterization of bounded sets in spaces of ultradistribu-
tions, Proc. Amer. Math. Soc. 120 (1994), 1191–1206.
[124] S. Pilipović, Quasiasymptotics and S-asymptotics in S ′ and D′ , Publ. Inst.
Math. Beograd 58 (1995), 13–20.
[125] S. Pilipović, B. Stanković, Abelian theorems for the distributional Stieltjes
transform, Z. Anal. Anw. 6 (1987), 341–349.
[126] S. Pilipović, B. Stanković, Initial value Abelian theorems for the distribu-
tional Stieltjes transform, Studia Math. 86 (1987), 239–254.
[127] S. Pilipović, B. Stanković, S-asymptotics of a distribution, Pliska Studia
Math. Bulgar. 10 (1989), 147–156.
[128] S. Pilipović, B. Stanković, Structural theorems for the S-asymptotics and
quasi-asymptotics of distributions, Mathematica Panonica 1 (1993), 23–35.
[129] S. Pilipović, B. Stanković, Wiener-type Tauberian theorems for distribu-
tions, J. London Math. Soc. 47 (1993), 507–515.
[130] S. Pilipović, B. Stanković, Wiener-type Tauberian Theorems for ultra-
distributions, Rend. Sem. Math. Univ. Padova 92 (1994), 209–220.
[131] S. Pilipović, B. Stanković, Quasiasymptotics and S-asymptotics of ultra-
distributions, Bull. Acad. Serbe. Sci. 20 (1995), 61–74.
[132] S. Pilipović, B. Stanković, Properties of ultradistributions having the S-
asymptotics, Bull. Acad. Serbe Sci. 21 (1996), 47–59.
[133] S. Pilipović, B. Stanković, Tauberian Theorems for integral transforms of
distributions, Acta Math. Hungarica 74 (1997), 135–153.
[134] S. Pilipović, B. Stanković, Wiener Tauberian Theorems for Hyperfunctions,
Z. Anal. Anw. 21 (2002), 1127–1042.
[135] S. Pilipović, B. Stanković, A. Takači, Asymptotic Behavior and Stieltjes
Transformation of Distributions, Teubner, Leipzig BSB, 1990.
[136] S. Pilipović, M. Stojanović, B-transform and its quasi-asymptotics — Ap-
plication to the convolution equation (xuxx + 2ux + mu) ∗ g = h, m ≤ 0,
Publ. Math. Debrecen 44 (1994), 275–283.
[137] S. Pilipović, M. Stojanović, Generalized Solutions to Nonlinear Volterra
Integral Equations with non-Lipshitz Nonlinearity, Nonlinear Analysis,
TMA 37 (1999), 319–335.
[138] S. Pilipović, A. Takači, The quasiasymptotic behavior of some distributions,
Univ. u Novom Sadu, Zb. Rad. Prirod. Mat. Fak., Ser. Math. 15 (1985),
37–46.
[139] S. Pilipović, J. Vindas, Multidimensional Tauberian theorems for wavelet
and non-wavelet transforms, preprint.
[140] H. R. Pitt, Generalized Tauberian Theorems, Proc. London Math. Soc. 44
(1938), 243–288.
[141] H. R. Pitt, Tauberian Theorems, Oxford Univ. Press, Oxford, 1958.
[142] H. Reiter, Classical harmonic analysis and locally compact groups, Claren-
don Press, Oxford, 1968.
[143] K. Saneva, J. Vindas, Wavelet expansions and asymptotic behavior of
distributions, J. Math. Anal. Appl. 370 (2010), 543–554.
290 Bibliography
[144] M. Sato, On a generalization of the concept of functions, Proc. Japan Acad.

34 (1958), 126–130, and 604–608.
[145] M. Sato, Theory of hyperfunctions I, J. Fac. Sci. Univ. Tokyo IA 8 (1959),
139–193.
[146] L. Schwartz, Théorie des Distributions, Hermann, Paris, 1966.
[147] J. Sebastião e Silva, Axiomatic approach to the distribution theory, Inte-
grals and orders of distributions, Lisbonne, Centro de Calculo Cientifico,
1964.
[148] E. Seneta, Regularly Varying Functions, Springer, Berlin, 1976.
[149] B. Stanković, Theorems of Tauberian Type for Measures, Glasnik Mat. 20
(1985), 383–390.
[150] B. Stanković, Abelian and Tauberian theorems for the Stieltjes transform
of distributions, Usp. Mat. Nauk 40 (1985), 91–103. Russian Math. Surveys
40 (1985), 99–113.
[151] B. Stanković, Characterization of some subspaces of D′ by S-asymptotics,
Publ. Inst. Math. Beograd 41 (1987), 111–117.
[152] B. Stanković, S-asymptotic expansion of distributions, Internat. J. Math.
Math. Sci. 11 (1988), 449–456.
′
[153] B. Stanković, S-Asymptotics of Elements Belonging to DLp and DL p , Univ.
u Novom Sadu, Zb. Rad. Prir. Mat. Fak. Ser. Mat. 18 (1988), 169–175.
[154] B. Stanković, S-asymptotics of solutions of the elliptic partial differential
equation, Univ. u Novom Sadu, Zb. Rad. Prir. Mat. Fak. Ser. Mat. 19
(1989), 65–72.
[155] B. Stanković, A Structural Theorem for the Distributions Having S-
asymptotics, Publ. Inst. Math. Beograd 45 (1989) 129–132.
[156] B. Stanković, Regularly varying distributions, Publ. Inst. Math. Beograd
48 (1990), 119–128.
[157] B. Stanković, Abelian Theorem for the Stieltjes Transform of Distributions,
Int. J. Math. Sci. 13 (1990), 677–686.
[158] B. Stanković, Abelian Theorems for integral transforms of distributions,
Int. Trans. Spec. Func. 1 (1993), 61–73.
[159] B. Stanković, Relation between quasiasymptotics, equivalence at infinity
and S-asymptotics, Univ. u Novom Sadu Zb. Rad. Prir. Mat. Fak. Ser.
Mat. 21 (1993), 1–11.
[160] B. Stanković, Asymptotic almost periodic distributions and applications to
PDE, Trudy Steklov Math. Inst. Moscow 203 (1994), 449–460.
[161] B. Stanković, Analyltic ultradistributions, Proc. Amer. Math. Soc. 123
(1995), 3365–3369.
[162] B. Stanković, Regularly varying ultradistributions, Publ Inst. Math.
Beograd 58 (1995), 117–128.
[163] B. Stanković, Taylor Expansion for Generalized Functions, J. Math. Anal.
Appl. 203 (1996), 31–37.
[164] B. Stanković, Asymptotic Taylor expansion for Fourier hyperfunctions,
Proc. Royal Soc. London A 453 (1997), 913–918.
[165] B. Stanković, Convergence structures and S-asymptotic behaviour of
Fourier hyperfunctions, Publ. Inst. Math. (Beograd) 64 (78), (1998), 98–
106.
Bibliography 291
[166] H. Struwe, Semilinear Wave Equations, Bulletin Amer. Math. Soc. 26

(1992), 53–85.
[167] A. Takači, A note on the distributional Stieltjes transform, Math. Proc.
Camb. Philos. Soc. 94 (1983), 523–527.
[168] A. Tauber, Ein Satz aus der Theorie der unendlichen Reihen, Monatsh.
Math. Phys. 8 (1897), 273–277.
[169] E. C. Titchmarsh, The Theory of Functions, Second Edition, Oxford Uni-
versity Press, Oxford, 1939.
[170] E. C. Titchmarsh, Introduction to the theory of Fourier integrals, Second
Edition, Clarendon Press, Oxford, 1948.
[171] J. Vindas, Structural theorems for quasiasymptotics of distributions at
infinity, Publ. Inst. Math. (Beograd) (N.S.) 84(98) (2008), 159–174.
[172] J. Vindas, The structure of quasiasymptotics of Schwartz distributions, in:
Linear and non-linear theory of generalized functions and its applications,
297–314, Banach Center Publ. 88, Polish Acad. Sc. Inst. Math., Warsaw,
2010.
[173] J. Vindas, Local behavior of distributions and applications, Dissertation,
Louisiana State University, Baton Rouge, 2009.
[174] J. Vindas, Regularizations at the origin of distributions having prescribed
asymptotic properties, Integral Transforms Spec. Funct. (2011), in press.
[175] J. Vindas, R. Estrada, Distributionally regulated functions, Studia Math.
181 (2007), 211–236.
[176] J. Vindas, R. Estrada, Distributional Point Values and Convergence of
Fourier Series and Integrals, J. Fourier. Anal. Appl. 13 (2007), 551–576.
[177] J. Vindas, R. Estrada, A Tauberian theorem for distributional point values,
Arch. Math. (Basel) 91 (2008), 247–253.
[178] J. Vindas, R. Estrada, On the jump behavior of distributions and logarith-
mic averages, J. Math. Anal. Appl. 347 (2008), 597–606.
[179] J. Vindas, R. Estrada, Measures and the distributional φ-transform,
Integral Transforms Spec. Funct. 20 (2009), 325–332.
[180] J. Vindas, R. Estrada, A quick distributional way to the prime number
theorem, Indag. Math. (N.S.) 20 (2009), 159–165.
[181] J. Vindas, R. Estrada, On the Order of Summability of the Fourier Inversion
Formula, Anal. Theory Appl. 26 (2010), 13–42.
[182] J. Vindas, R. Estrada, On the support of tempered distributions, Proc.
Edinb. Math. Soc. 53 (2010), 255–270.
[183] J. Vindas, R. Estrada, On the Cesàro behavior and some asymptotic
notions for Schwartz distributions, preprint.
[184] J. Vindas, R. Estrada, On Tauber’s second Tauberian theorem, preprint.
[185] J. Vindas, R. Estrada, Local boundary behavior of harmonic and analytic
functions I: Abelian theorems, preprint.
[186] J. Vindas, S. Pilipović, Structural theorems for quasiasymptotics of distri-
butions at the origin, Math. Nachr. 282 (2009), 1584–1599.
[187] J. Vindas, S. Pilipović, D. Rakić, Tauberian theorems for the wavelet trans-
form, J. Fourier Anal. Appl. 17 (2011), 65–95.
[188] V. S. Vladimirov, Multidimensional Generalization of a Tauberian Theorem
292 Bibliography
of Hardy and Littlewood, Math. USSR, Izvestiya 10 (1976), 1032–1048.

[189] V. S. Vladimirov, Generalized Functions in Mathematical Physics, Mir,
Moscow, 1979.
[190] V. S. Vladimirov, Equations in Mathematical Physics, Nauka, Moscow,
1988 (In Russian).
[191] V. S. Vladimirov, B. I. Zavyalov, Tauberian theorems in quantum field
theory, Itogi Nauk. Tehn. 15 (1980), 95–130 (In Russian).
[192] V. S. Vladimirov, Yu. N. Drozhzhinov and B. I. Zavyalov, Tauberian
Theorems for Generalized Functions, Kluwer, Dordrecht, 1988.
[193] V. S. Vladimirov, Yu. N. Drozhzhinov, B. I. Zavyalov, Tauberian theo-
rems for generalized functions in the scale of regularly varying functions
and functionals, Publ. Inst. Math. (Beograd) 71 (85) (2002), 123–132. (In
Russian).
[194] P. Wagner, On the Quasi-asymptotic Expansion of the Causal Fundamental
Solution of Hyperbolic Operators and Systems, Z. Anal. Anw. 10 (1991),
159–167.
[195] G. Walter, Pointwise convergence of distribution expansions, Studia
Math. 26 (1966), 143–154.
[196] G. Walter, Fourier series and analytic representation of distributions, SIAM
Review 12 (1970), 272–276.
[197] G. Walter, Local boundary behavior of harmonic functions, Illinois J. Math.
16 (1972), 491–501.
[198] N. Wiener, Tauberian Theorems, Ann. of Math. 33 (1932), 1–100.
[199] R. Wong, J. P. McClure, Generalized Mellin convolution and their asymp-
totic expansions, Canad. J. Math. 36 (1984), 924–960.
[200] B. I. Zavyalov, Automodel asymptotics of electromagnetic form-factors and
the behavior of their Fourier transforms in the neighborhood of the light
cone, Teoret. Mat. Fiz. 17 (1973), 178–188 (In Russian).
[201] B. I. Zavyalov, Asymptotic properties of functions that are holomorphic
in tubular cones, Math. Sb. (N.S.) 136 (178) (1988), 97–114 (In Russian).
English transl. Math. USSR, Sb. 64 (1989), 97–113.
[202] A. H. Zemanian, Generalized integral transforms, Interscience, John Wiley,
New York, 1968.
[203] A. H. Zemanian, Some Abelian Theorems for the Distributional Hankel
and K-Transformation, SIAM J. Appl. Anal 14 (1966), 1255–1265.
[204] V. V. Zharinov, Laplace transform of Fourier hyperfunctions and other
related classes of analytical functionals, Teoret. Mat. Fiz. 33 (1977), 291–
309 (In Russian).
[205] V. V. Zharinov, On the quasiasymptotics of Fourier hyperfunctions, Teoret.
Mat. Fiz. 43 (1980), 32–38 (In Russian).
[206] A. Zygmund, Trigonometric Series, Vols. I & II, Second Edition, Cambridge
University Press, Cambridge, 1959.
Index
Abel summability, 241, 247, 256 Convolution type transforms, 209

Abel’s theorem, 242 convolutor of Fg′ , 9
Abelian type theorems, 188
acute cone, 1 decrease exponentially, 6
associate asymptotically defining function, 7
homogeneous function, 142 Differentiation, 8
asymptotic expansion, 52 distributional boundedness at a
asymptotic sequence, 52 point, 270
asymptotic Taylor expansion, 59 distributional evaluations in the Abel
asymptotically homogeneously sense, 247
bounded function, 161 distributional evaluations in the
asymptotically homogeneous Cesàro sense, 247
function, 132 distributional point values, 249
distributions, 2, 3
Beurling type spaces of
distributions with fast descent, 3
ultradifferentiable functions, 5
distributions with compact support, 3
dual cone, 1
Cauchy-Szegö kernel, 86
causal fundamental solutions, 184
elliptic homogeneous operator, 175
Cesàro behavior of distributions, 243
Cesàro evaluations, 247 evaluation in the Cesàro sense, 247
Cesàro means, 240 evaluations in the Abel sense, 247
Cesàro summability, 239
Change of variables, 9 Fourier hyperfunctions, 6
characteristic function of a closed Fourier integrals, 268
convex acute cone, 86 Fourier inversion formula, 252, 259
closed quasi-asymptotic expansion, Fourier series, 253, 259, 260
125 Fourier transform on Q(Dn ), 8
cone, 1 function of infra-exponential type, 7
conjugate cone (dual cone), 1 fundamental solution, 172
convolution of S, T ∈ Fg′ , 9
convolution of T, S ∈ D′ , 4 generalized S-asymptotics, 78
293
294 Index
Hölder means, 240 regularly varying function, 1

Heaviside function, 1 Riesz typical means, 240
homogeneous distributions, 84 Roumieu type spaces of
ultradifferentiable functions, 5
infinite-order differential operator
J(D), 8 S-asymptotic behavior, 10
S-asymptotic expansion, 53
Laguerre orthonormal system, 178 S-asymptotic expansion of first type,
Laguerre polynomials, 178 55
Laplace transform, 101, 198 S-asymptotic expansion of second
Littlewood’s Tauberian theorem, 275 type, 55
local operator, 7 S-asymptotic expansions in Fg′ , 52
Lojasiewicz point values, 249 S-asymptotic Taylor expansion, 58
S-asymptotics, 10
Malgrange-Ehrenpreis theorem, 177 Schwartz’s spaces, 3
Mellin convolution type transform, set of convolutors of Fg′ , 9
195 slowly varying function, 1
Mellin transform, 223 solid cone, 1
moment asymptotic expansion, 130 space of rapidly decreasing functions,
multiplier of Fg′ , 9 3
space of tempered distributions, 3
open quasi-asymptotic expansion, 125 Spaces of tempered
open quasi-asymptotic expansion at ultradistributions, 5
∞ (at 0+ ) of order (α, L), 127 Stieltjes transform, 199
structure of quasi-asymptotics, 130,
Poisson transform, 198 140, 149
polydisc, 1 symmetric point value, 280
Potter estimate, 132
principal value evaluation, 277 Tauberian theorems, 270
Tauberian type theorems, 202
quasi-asymptotic behavior at infinity, transforms of Mellin convolution
90 type, 223
quasi-asymptotic behavior at infinity
in F ′ , 82 ultradifferentiable functions, 5
quasi-asymptotic boundedness, 160 ultradifferential operator, 6
quasi-asymptotic expansion, 128 ultradistribution, 4
quasi-asymptotic extension, 150
quasi-asymptotics at 0, 117 Weierstrass functions, 267
quasi-asymptotics at±∞, 113 Weierstrass transform, 198
quasi-asymptotics over a cone Γ, 82 weighted distributions, 190
weighted ultradistribution spaces, 5
rapidly decreasing functions, 3 Wiener–Tauberian theorem, 202
rapidly exponentially decreasing
functions, 4
regular cone, 86
regular elements, 9

Pilipović, Stevan - Stanković, Bogoljub - Vindas, Jasson - Asymptotic Behavior of Generalized Functions PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Pilipović, Stevan - Stanković, Bogoljub - Vindas, Jasson - Asymptotic Behavior of Generalized Functions PDF

Uploaded by

Copyright:

Available Formats

Series on Analysis, Applications and Computation – Vol.

8266hc.9789814366847-tp.indd 1 9/5/11 3:00 PM

Series Editors: Heinrich G W Begehr (Freie Univ. Berlin, Germany)

Vol. 1: Boundary Values and Convolution in Ultradistribution Spaces

Vol. 2: Complex Analysis

Vol. 3: Topics in Mathematical Analysis

Vol. 4: Nonlinear Waves: An Introduction

Vol. 5: Asymptotic Behavior of Generalized Functions

He Yue - Asymptotic Behavior.pmd 1 9/15/2011, 11:32 AM

8266hc.9789814366847-tp.indd 2 9/5/11 3:00 PM

British Library Cataloguing-in-Publication Data

Series on Analysis, Applications and Computation — Vol. 5

He Yue - Asymptotic Behavior.pmd 2 9/15/2011, 11:32 AM

There are several approaches to generalized asymptotics within spaces of

then T = o(r), x → ∞. If there exists a constant M such that

then T = O(r), x → ∞. Finally, if T = Dm F in (a, ∞) and

T (kx) → CH(x), k → ∞ in S ′ (R) ,

where H(x) is the Heaviside function. In short, q − limx→∞ T (x) = C. If

statement: T ∈ M+ admits a quasi-asymptotic part of degree α at infinity

asymptotic behavior of distributions. This was necessary in order to make

This page intentionally left blank

I. Asymptotic Behavior of Generalized Functions 1

2.1 Definition of quasi-asymptotics at infinity over a

II. Applications of the Asymptotic Behavior of Generalized

3 Asymptotic behavior of solutions to partial differential

5 Summability of Fourier series and integrals . . . . . . . . 237

Asymptotic Behavior of Generalized

0.1. We denote by R and N the sets of real and natural numbers;

The number α is called index of regular variation. If α = 0, then ρ is called

2 Asymptotic Behavior of Generalized Functions

xα L(x), x > a, where L is slowly varying. It is known that the convergence

C1 x−ε ≤ L(x) ≤ C2 xε , x ≥ X ; (0.2)

(ii) lim xε L(x) = +∞, lim x−ε L(x) = 0.

The definition of a regular varying function at zero is similar. For the

where H is the Heaviside function, and the derivative f (m) is taken to be in

0.5.1. The Schwartz spaces of test functions and distributions on Rn are

S(Γ) = {ϕ ∈ C ∞ (Γ); kϕkp < ∞, p ∈ N} ,

kϕkp = sup (1 + kxk2 )p/2 |ϕ(β) (x)| .

νm (ϕ) = sup exp(mkxkp )|Dα ϕ(x)|< ∞, m = 1, 2, . . .

4 Asymptotic Behavior of Generalized Functions

The elements of Kp are called rapidly exponentially decreasing functions.

The convolution of T, S ∈ D′ is defined by

where A and B are constants independent of p.

Let K be a compact set in an open set Ω of Rn .

where ∗ denotes either (Mp ) or {Mp }.

Weighted ultradistribution spaces are defined by

Spaces of tempered ultradistributions are defined as the strong duals of

6 Asymptotic Behavior of Generalized Functions

S {Mα } (Rn ) ≡ S {Mα } = ind lim ShMα ,

The Fourier transform is an isomorphism of S ∗ onto S ∗ ; D∗ = D∗ (Rn ) is

The dual space of P∗ is the space Q(Dn ) of the Fourier hyperfunctions.

(Dn + iI)#j Dn = U1 ∩ · · · ∩ Uj−1 ∩ Uj+1 ∩ · · · ∩ Un .

Thus, f ∈ Q(Dn ) is defined as the class [F ], where F ∈ Õ((Dn + iI)#Dn ).

Fσ (x + iΓσ 0) denotes the element of the quotient space given in (0.9); it is

is called a local operator. J(D) maps continuously O(U ) into O(U ), U

8 Asymptotic Behavior of Generalized Functions

The Fourier transform on Q(Dn ) is defined by using the functions

Change of variables: If T ∈ Fg′ and k > 0, then by definition

10 Asymptotic Behavior of Generalized Functions

If (1.1) is satisfied, it is also said that T has S-asymptotics and we write in

Remarks. 1) If Γ is a convex cone we could use another limit in Γ.

1.2 Characterization of comparison functions and limits

Proposition 1.1. Let Γ be a convex cone. Suppose T ∈ Fg′ has the S-