Professional Documents
Culture Documents
ISAAC
Asymptotic Behavior of
Generalized Functions
Published
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
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.
Preface
v
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
Preface vii
viii Preface
Preface ix
Contents
Preface v
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
xii Contents
Contents xiii
Bibliography 283
Index 293
This page intentionally left blank
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
Chapter I
0 Preliminaries
1
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
0. Preliminaries 3
where
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:
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
0. Preliminaries 5
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
H Mp
E {Mp } = proj lim ind lim EK p .
K⊂⊂Ω
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 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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
0. Preliminaries 7
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
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
0. Preliminaries 9
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 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→∞
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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].
U (· + h) = d(h)U, h ∈ Γ .
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
1. S-asymptotics in Fg′ 13
∂
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→∞
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→−∞
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
1. S-asymptotics in Fg′ 15
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 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 .
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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)
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)
(T ∗ δm )(x + h)
w. lim = Mm exp(αx), in F ′ , uniformly in m ∈ N .
h∈Γ,khk→∞ c(h)
(1.8)
c) For a δ-sequence hδm im , (cf. 0.6),
(T ∗ δm )(h)
lim = pm , m ∈ N , (1.9)
h∈Γ,khk→∞ 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∗ ).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
1. S-asymptotics in Fg′ 17
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→∞
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
h∈Γ,khk→∞ c(h) h∈Γ,khk→∞ c(h)
in both cases (ϕ ∈ D, respectively ϕ ∈ D∗ ).
The paper which can be consulted for related results is [128].
1. S-asymptotics in Fg′ 19
= hU (k) , ϕi, ϕ ∈ Fg .
T (x + h)
b) Since Fg is a Montel space, we have that lim
= U (x)
h∈Γ,khk→∞ c(h)
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:
g(x + h)
= U (x), lim − G(x) ϕ(x) + hU (x), G(x)ϕ(x)i
h∈Γ,khk→∞ c1 (h)
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)
1. S-asymptotics in Fg′ 21
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+ .
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.
1. S-asymptotics in Fg′ 23
and
s
Y
(1 + x + h)−p ∼ q −β wi−pi · 1, h ∈ Γ ,
i∈J
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. S-asymptotics in Fg′ 25
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.
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 :
|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 .
1
= lim hexp(−k0 khkp )T (x + h), ϕ(x)i = 0.
khk→∞ c(h)
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
1. S-asymptotics in Fg′ 27
′ (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
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
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)
1. S-asymptotics in Fg′ 29
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
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
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
1. S-asymptotics in Fg′ 31
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.
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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.
1. S-asymptotics in Fg′ 33
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 ;
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
hT (x + am ), φ(x)i = αm 6= 0, m ∈ N .
1. S-asymptotics in Fg′ 35
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
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
(CRn A = Rn \ A).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
1. S-asymptotics in Fg′ 37
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;
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
∞
X
c1,2 /(b1 c1 ) + c2,2 /(b2 c2 ) + c3,2 /(b3 c3 ) + ck,2 /bk 6= 0.
k=4
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
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
1. S-asymptotics in Fg′ 39
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 1.2. 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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
1. S-asymptotics in Fg′ 41
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. S-asymptotics in Fg′ 43
(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 → ∞
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]).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
1. S-asymptotics in Fg′ 45
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→ ∞.
1. S-asymptotics in Fg′ 47
∗
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 .
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
1. S-asymptotics in Fg′ 49
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
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
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 .
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
1. S-asymptotics in Fg′ 51
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
P P
= ((Fσ χσ̃ ) ∗ ψs,ε )(x)
σ∈Λ σ̃∈Λ
P
= Fσ ∗ ψs,ε (x) = (f ∗ ψs,ε )(x).
σ∈Λ
We cite some papers related to this problem: ([131], [132], [155],[151], [123]).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
k
X
u(t) − un (t) = o(ψk (t)), t → ∞ (t → 0) (1.26)
n=1
1. S-asymptotics in Fg′ 53
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 1.3. 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
while
1. S-asymptotics in Fg′ 55
∞
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,
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.
∞
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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)
1. S-asymptotics in Fg′ 57
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
Estrada and Kanwal have introduced in [53] and [54] the asymptotic Taylor
expansion of distributions (see also [56]). We generalize it to Fg′ .
1. S-asymptotics in Fg′ 59
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
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
∗
(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}.
and
1. S-asymptotics in Fg′ 61
′
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
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
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
1. S-asymptotics in Fg′ 63
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.
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) .
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
1. S-asymptotics in Fg′ 65
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
1. S-asymptotics in Fg′ 67
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 .
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+ .
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.
1. S-asymptotics in Fg′ 69
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)).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
sup{|Fi (x)|exp(k|x|); x ∈ R, i = 0, . . . , m} ≤ M2 ,
Z−h Z∞ !
= + (|φ(i) (x)kFi (x + h)|/c(h))dx
−∞ −h
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
1. S-asymptotics in Fg′ 71
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 ,
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
then this limit exists for every ϕ ∈ K1 (R) (for every ϕ ∈ S(R)). In partic-
ular, S ∈ K′ 1 (R) (S ∈ S ′ (R)).
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.
where
1. S-asymptotics in Fg′ 73
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 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 |))
and if for some k > 0 and every (a1 , . . . , an ) ∈ Λ\{(1, . . . , 1), (−1, . . . , −1)}
the sets
Q1 = {T (x + h)/c(h); h ∈ Γ}
Q2 = {T (x + k + h)/(c(h)p(k)); h ∈ Γ, k ∈ G}
1. S-asymptotics in Fg′ 75
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.
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 .
Zb
L(exp x · exp h)
· exp(αx) φ(x)dx → exp(αx)φ(x)dx, h → ∞ .
L(exp h)
a
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
The assertion follows now from Proposition 1.11 and Theorem 1.16.
then
∞
X
s
f (x + h) ∼ ui (x + h)|{ci (h)}, h → ∞ .
i=1
1. S-asymptotics in Fg′ 77
gs
We write in short T (x) ∼ e(x), kxk→ ∞, x ∈ Γ.
If we compare Definition 1.1 and Definition 1.6, we obtain.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
1. S-asymptotics in Fg′ 79
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→∞
1. S-asymptotics in Fg′ 81
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′ .
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2 Quasi-asymptotics in F ′
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).
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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.
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
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
A1 xα α α
+ + A2 x− resp. A1 x+ if I = R resp. I = (0, ∞) . (2.4)
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 85
A further analysis shows that Uβ′ (I) has the desired properties, for de-
tails see [85].
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→∞
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 87
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. 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 .
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)
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 89
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.
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
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.”
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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)
2. Quasi-asymptotics in F ′ 91
Since
Z Z
x
G(x)ϕ(p) dx → G(x)dxϕ(p) (0), k → ∞ ,
k
K K
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.
2. Quasi-asymptotics in F ′ 93
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).
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.
2. Quasi-asymptotics in F ′ 95
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
∗
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)
2. Quasi-asymptotics in F ′ 97
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 99
xm−1 (m−1)
− φ (0)H(a − x))dx, φ ∈ S(R) . (2.16)
(m − 1)!
It remains to show that the last limit equals zero. Since for every L > 0
2. Quasi-asymptotics in F ′ 101
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.
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)
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)
2. Quasi-asymptotics in F ′ 103
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)
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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. Quasi-asymptotics in F ′ 105
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
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. Quasi-asymptotics in F ′ 107
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→∞
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, ∞) .
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 109
Let us set
( (
f2 (x), x>0 f2 (x), x<0
f2+ (x) = , f2− (x) =
0, x≤0 0, x ≥ 0.
lim hf2 (kx)/(k v L(k)), ϕ(x)i = lim hf2+ (kx)/(k v L(k)), ϕ(x)i .
k→∞ k→∞
r
by putting k = e , r ∈ R, we obtain that for a function F1,2 (t) =
f2+ (et ), t ∈ R, there exists the limit
2. Quasi-asymptotics in F ′ 111
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)
−∞
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→∞
r
X
(f− (kt) + bβ δ (β) (kt))/(k v L(k)) → C2 fv+1 (−t) in S ′ (R) as k → ∞ ,
β=0
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 113
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 ).
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).
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).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 115
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|)
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).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 117
Proof. The same arguments, as in the proof of Theorem 2.10 yield that
f ∈ S ′ (R) and that
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).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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.
2. Quasi-asymptotics in F ′ 119
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.
2. Quasi-asymptotics in F ′ 121
2. Quasi-asymptotics in F ′ 123
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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).
2. Quasi-asymptotics in F ′ 125
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)).
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+ .
We write in short:
N
X
q.e.
f ∼ ci (fLi )αi +1 at ∞ (at 0+ ) of order (α, L).
i=1
′
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)
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 127
By the same arguments as in Proposition 2.13 one can prove the follo-
wing proposition:
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
2. Quasi-asymptotics in F ′ 129
and
q.e. X
f ∼ Ãk (fL̃k )α̃k +1 |{εα̃k L̃k (ε)} ,
k∈Λ
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
2. Quasi-asymptotics in F ′ 131
if
n
X
hf (λx), φ(x)i = ρj (λ) hgj (x), φ(x)i + o(ρn (λ)), φ ∈ F (R).
j=1
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. Quasi-asymptotics in F ′ 133
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 .
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
(ii) If α < 0 (resp. α > 0), then there exists a constant γ such that
2. Quasi-asymptotics in F ′ 135
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
By this and
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
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
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→∞
2. Quasi-asymptotics in F ′ 137
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.
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 .
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
We keep ε < ε0 and x ∈ 2−n−1 , 2−n . Then
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.
2. Quasi-asymptotics in F ′ 139
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.
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) ,
(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
2. Quasi-asymptotics in F ′ 141
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.
Lemma 2.4. Relation (2.65) holds uniformly in compact subsets of (0, ∞).
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 143
j=1
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
where the last inequality follows from (2.68). So if λ0 < λ and 1 < x, then
b(λx) − b(λ) − βL(λ) log x
≤ M1 x 14 ,
L(λ)
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).
as ε → 0+ in D′ (R).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 145
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.
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).
2. Quasi-asymptotics in F ′ 147
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.
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 149
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. Quasi-asymptotics in F ′ 151
|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
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)
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 153
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.
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.
2. Quasi-asymptotics in F ′ 155
where F.p. stands for the Hadamard finite part of the divergent integral
[56].
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).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 157
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.
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).
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)
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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)
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 159
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 161
(respectively λ → 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.
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 163
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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, ∞).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 165
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 2.46. Let f0 ∈ D′ (0, ∞). Let L be slowly varying at the origin
and α ∈ R. Suppose that
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,
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).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
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 .
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
2. Quasi-asymptotics in F ′ 167
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
Z∞
1
→ ex ϕ(x)dx, k → ∞, ϕ ∈ K1 .
2
−∞
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.
2. Quasi-asymptotics in F ′ 169
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.
Chapter II
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]).
171
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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 ′ .
h∈Γ,khk→∞ c(h)
We know that
exp(−it · h)f (t) = F [F −1 [exp(−iy · h)f (y)](x)](t) .
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
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
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
3ρ
kxk≤
Z 2
M
≤K ρ2 k
kx + ρwkm−n dx → 0, ρ → ∞ ,
(1 + 4 )
kxk≥ ρ
2
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
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.
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
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
(x + r)p = q, r > 0, q ∈ S ′ + ,
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
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, . . .
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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.
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
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
∗ ∗
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 ∗ ).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
∗
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′ .
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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)).
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
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 .
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
∼ (is)−r , s → ∞, s ∈ R+ , r > 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
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 .
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
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:
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
where s ∈ C \ R− .
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
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
* 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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
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]).
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
Third step. We prove that (f ∗ ψ)(h), h ∈ R, exists and that (c) holds.
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,
x→∞ L(exp x) exp(αx)
R
then
Z
(f ∗ ψ)(x)
lim =A ψ(t) exp(−αt)dt
x→∞ L(exp x) exp(αx)
R
s
Then, f (x + h) ∼ c(h) · C, h ∈ R+ .
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 >
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
∗
then for every φ ∈ DL 1 (R),
Z
lim (f ∗ ψ)(x) = a ψ(t)dt.
x→∞
R
fn (x) = (f ∗ ϕn )(x), x ∈ R, n ∈ N ,
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
∗
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,
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
Z1
rk L(ex+h−t ) α(x−t) (k) |rα|k αx
≤ e |ω (t)|dt + e
Mk L(eh ) Mk
−1
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
≤ C3 eαx+δ|x| + C2 eαx .
Since
f (x + h)
(f ∗ K)(h) = , c(x + h)Ǩ(x) , h ∈ R,
c(x + h)
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
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,∞ .
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
∞
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
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]).
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
∗ ∗
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
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.
(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∗ .
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
Z1
2π (n−1)/2 n−1
I0 (ρ) = (1 − t2 ) 2 −1 e−tρ dt, ρ ∈ C
Γ((n − 1)/2)
−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:
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
−
→ −
→
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.
~
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]).
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.
then
←
−
G(z/t)/tn ρ(t) → H(z), t → ∞, in Λ (Σ).
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
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.
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, ∞).
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
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
proved in the first step of the proof of Theorem 4.8, it follows that e(· +
h)Ǩ ∈ DL1 (R).
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+
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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)
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
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)
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
u−m hF (−m) (uξ), ϕ(ξ))i = hF (uξ), hfm (t), ϕ(ξ + t)ii, (4.46)
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].
and
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)
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
−∞
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
−∞
then
lim hf (t + x), ψ(t)i = ha, ψ(t)i
x→∞
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 .
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:
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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].
[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,
−∞
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.
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)
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
we write
∞
X
cn = γ (A). (5.12)
n=0
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.
f (x) ∼ Cxα
+ (C), x → ∞, (5.21)
are equivalent.
f (x) ∼ Cxα
+ (C) , x → ∞, (5.25)
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.
where γ = γ− + γ+ .
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).
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
As a corollary we obtain.
As expected, the Cesàro method is strictly stronger than the e.v Cesàro
summability (see also Example II.2).
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=−∞
P
Proof. We have that fˆ(x) = 2π ∞n=−∞ cn δ(x − n), the rest follows
from Corollary 5.3.
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 −∞
or which amounts to the same,
Z ∞
1 t
lim φam
eix0 t fˆ(t)dt = γ, for each a > 0 . (5.56)
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.
Z 0
1
lim fˆ(t)eix0 t dt = γ− (C, k), (5.59)
x→+∞ 2π −x
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→∞
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
∞
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
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].
and
γ δ(x) 1
g− (λx) = − c(λ) +o as λ → ∞ in S ′ (R) . (5.68)
2 λ λ
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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 now analyze sufficient conditions under which the existence of the dis-
tributional point value implies the convergence of the Fourier series at the
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
But,
p1 1q
X X X
p q
|cn | |φ(εn)| ≤ |cn | |φ(εn)| ,
a
a a
ε <n 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
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
and follow a similar argument as the one in the proof of Theorem 5.7.
Corollary 5.11. If
∞
X
cn = γ (C, k), (5.86)
n=0
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.
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
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 → ∞.
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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 .
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].
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.
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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.
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].
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) .
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).
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−
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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.
∞
X
Since η was arbitrary, we conclude that cn = γ.
n=0
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] )
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
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
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).
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
Then,
∞
a0 X
+ (an cos nx0 + bn sin nx0 ) = γ (C)
2 n=1
Bibliography
283
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
284 Bibliography
Bibliography 285
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
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
Bibliography 287
288 Bibliography
Bibliography 289
290 Bibliography
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.
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
Bibliography 291
292 Bibliography
Index
293
September 1, 2011 10:53 World Scientific Book - 9in x 6in maj16
294 Index