2007
Ana Paula
Bran o Nolas o
Propriedades de Regularidade de Operadores de
Wiener-Hopf-Hankel
Regularity Properties of Wiener-Hopf-Hankel
2007
Ana Paula
Bran o Nolas o
Propriedades de Regularidade de Operadores de
Wiener-Hopf-Hankel
Regularity Properties of Wiener-Hopf-Hankel
Operators
teseapresentada àUniversidade de Aveiro para umprimento dosrequisitos
ne essários à obtenção do grau de Doutor em Matemáti a, realizada sob
a orientação ientí a do Doutor Luís Filipe Pinheiro de Castro, Professor
Asso iado omAgregaçãodoDepartamentodeMatemáti adaUniversidade
de Aveiro
Apoionan eiro da FCT, Bolsade
presidente Reitora da Universidade deAveiro
Doutor Vasile Stai u
ProfessorCatedráti odaUniversidadedeAveiro
Doutor Semyon Borisovi hYakubovi h
ProfessorAsso iado omAgregaçãodaFa uldadedeCiên iasdaUniversidadedoPorto
Doutor Luís Filipe Pinheiro deCastro
ProfessorAsso iado omAgregaçãodaUniversidadedeAveiro(Orientador)
Doutor Juan CarlosSan hez Rodriguez
ProfessorAuxiliardaFa uldadedeCiên iaseTe nologiadaUniversidadedoAlgarve
Doutor Frank-Olme EwalSpe k
disponibilidadee in entivoprestados aolongodestes últimos quatroanos.
À Unidade de Investigação "Matemáti a e Apli ações" e aoDepartamento
de Matemáti a da Universidade de Aveiro por todo o apoio e fa ilidades
on edidas para a realização deste trabalho e a todos os meus olegas e
amigos pelo ambiente de trabalho propor ionado e pelo apoio que sempre
manifestaram.
À Fundação para a Ciên iaeaTe nologiapelo apoio nan eiro on edido.
invertibilidade, fa torização, função quase periódi a, função semi-quase
periódi a,funçãoquaseperiódi a portroços
resumo Nesta tese estudamos as propriedades de regularidade de operadores de
Wiener-Hopf-Hankel om símbolos de Fourier perten entes às álgebrasdas
funções quase periódi as, das funções semi-quase periódi as e das funções
quase periódi as portroçose onsideramosestes operadoresa a tuarentre
espaçosdeLebesgue
L
p
(para
1
< p <
∞
). Porpropriedadesderegularidade entende-se invertibilidade lateral e bilateral, propriedade de Fredholm esolubilidade normal.
Propomos uma teoria de fa torização para operadores de
Wiener-Hopf--Hankel om símbolos de Fourier quase periódi os, e a a tuar entre
espaços de Lebesgue
L
2
, introduzindo uma fa torização para os símbolos
de Fourier quase periódi os de tal modo que as propriedades dos fa tores
irãopermitir orrespondentesfa torizaçõesdosoperadores. Um ritériopara
a propriedade de semi-Fredholm e para a invertibilidadelateral e bilateralé
assim obtidoemtermos de determinadosíndi esdas fa torizações.
Baseado na relação delta após extensão, estabele emos um teorema do
tipo de Sarason para operadores de Wiener-Hopf-Hankel om símbolos de
Fouriersemi-quaseperiódi os,aa tuarentreespaçosdeLebesgue
L
2
. Uma
generalizaçãodoteoremadotipodeSarasonétambémobtida onsiderando
agoraosoperadoresaa tuarentreespaçosdeLebesgue
L
p
. Paraoperadores
deWiener-Hopf-Hankel omsímbolosdeFourierquaseperiódi osportroços,
a a tuar entre espaços de Lebesgue
L
2
bilateral destes operadores em termos dos valores médios e das médias
geométri as dosrepresentantesquaseperiódi osnoinnitodossímbolosde
Fourier,assim omodasdes ontinuidadesdedeterminadasfunçõesauxiliares
(no aso das funções quase periódi as por troços). Para ada aso, é
apresentada umafórmulapara o índi ede Fredholm.
Finalmente, de volta aos operadores de Wiener-Hopf-Hankel om símbolos
de Fouriernasubálgebradasfunçõesquaseperiódi as
AP W
,aa tuarentre espaços de LebesgueL
2
, onsideramos o aso mais geralde operadores de
Wiener-Hopf-Hankel om símbolos matri iais de Fourier
AP W
. Para estes operadores, obtemos um ritério para a invertibilidade e a propriedade desemi-Fredholmbaseadonahipótesede umespe í o onjunto de Hausdor
pie ewise almostperiodi fun tion
abstra t In this thesis we study the regularity properties of Wiener-Hopf-Hankel
operatorswithFouriersymbolsbelongingtothealgebrasofalmostperiodi ,
semi-almostperiodi andpie ewisealmostperiodi fun tionsandwe onsider
these operators a ting between
L
p
Lebesgue spa es (for
1
< p <
∞
). By regularity properties one means one-sided and both-sided invertibility,Fredholm property and normal solvability.
We propose a fa torization theory for Wiener-Hopf-Hankel operators with
almostperiodi Fouriersymbols,anda tingbetween
L
2
Lebesguespa es,by
introdu inga fa torization on ept forthe almostperiodi Fouriersymbols
su h that the properties of the fa tors will allow orresponding operator
fa torizations. A riterionforthesemi-Fredholmproperty and forone-sided
and both-sidedinvertibilityis thereforeobtainedupon ertainindi es ofthe
fa torizations.
Based on the delta relation after extension, we establish a Sarason's
type theorem for Wiener-Hopf-Hankel operators with semi-almost periodi
Fourier symbols and a ting between
L
2
Lebesgue spa es. We also derive
a generalization of the Sarason's type theorem, the so- alled
Dudu hava--Saginashvili's type theorem, when we onsider the same kind of operators
a tingnowbetween
L
p
Lebesguespa es. ForWiener-Hopf-Hankeloperators
withpie ewisealmostperiodi Fouriersymbols,a tingbetween
L
2
Lebesgue
one-sided invertibility of these operators, based on the mean motions and
geometri meanvaluesof the almostperiodi representativesof theFourier
symbolsatminusandplusinnity,aswellasonthedis ontinuitiesof ertain
auxiliaryfun tions (inthe ase of pie ewise almostperiodi fun tions). For
ea h ase, formulaeforthe Fredholmindex of theoperators areprovided.
Finally, we return to Wiener-Hopf-Hankel operators with Fourier symbols
in the subalgebra of almost periodi fun tions
AP W
, a ting betweenL
2
Lebesgue spa es, and we onsider the more general ase of
Wiener-Hopf-Hankeloperatorswithmatrix
AP W
Fouriersymbols. Fortheseoperatorswe a hieveaninvertibilityandsemi-Fredholm riterionbasedontheassumptionList of Symbols v
Introdu tion ix
1 Convolution Type Operators 1
1.1 Some denitions,notations, and histori al notes . . . 2
1.2 Regularity properties . . . 8
1.3 Relationsbetween onvolution typeoperators . . . 11
1.3.1 Relationsbetween bounded linearoperators . . . 11
1.3.2 Fa torizationof Wiener-Hopf plus Hankeloperators . . . 13
1.3.3 Relationbetween Wiener-Hopf-Hankeloperators and Wiener-Hopf operators . . . 18
1.3.4 Relationbetween Wiener-Hopf-Hankeloperators and Toeplitz-Hankeloperators . . . 23
1.4 Ne essary onditions forthesemi-Fredholmproperty . . . 27
2 Fourier Symbols 33 2.1 Almost periodi fun tions . . . 34
2.2 Semi-almostperiodi fun tions. . . 41
2.3 Pie ewise almost periodi fun tions . . . 45
3 Wiener-Hopf plus Hankel Operators with Almost Periodi Symbols 53 3.1
AP
fa torizations . . . 563.3 Invertibility of Wiener-Hopf plus Hankel operators with
AP
symbols. . . . 734 Wiener-Hopf-Hankel Operators with Semi-Almost Periodi Symbols 77 4.1 A Sarason's typetheorem . . . 78
4.1.1 Motivation. . . 78
4.1.2 First approa h . . . 79
4.1.3 A stronger version of the Sarason's type theorem . . . 83
4.1.4 Anotherlookattheinvertibilityandsemi-Fredholm riteriaof Wiener-Hopf-Hankel operatorswith
AP
symbols . . . 924.2 A Dudu hava-Saginashvili's typetheorem . . . 94
4.2.1 Someparti ular lasses of almost and semi-almost periodi fun tions 95 4.2.2 The Dudu hava-Saginashvili's theorem . . . 96
4.2.3 Mainresult . . . 100
4.3 Fredholmindex formula . . . 103
4.3.1 FredholmindexformulaforWiener-Hopf-Hankeloperatorswith
SAP
Fouriersymbols . . . 1044.3.2 FredholmindexformulaforWiener-Hopf-Hankeloperatorswith
SAP
p
Fouriersymbols . . . 1195 Wiener-Hopf-HankelOperatorswithPie ewiseAlmostPeriodi Symbols127 5.1 A semi-Fredholmand invertibility riterion . . . 128
5.2 Fredholmindex formula . . . 132
5.2.1 Aformulaforthesumoftheindi esofFredholmWiener-Hopfplus/minus Hankel operators . . . 132
5.2.2 Examplesand invertibility . . . 148
6 Matrix Wiener-Hopf-Hankel Operators with Good Hausdor Sets 157 6.1 Preliminaries . . . 157
6.4 Anexample . . . 164
Con lusion 167
A
n
olumns of length
n
with entries inA
, 158A
n
×n
n × n
matri eswith entries inA
, 158AP
lass of almostperiodi fun tions,34AP
p
losure ofAP W
inM
p
(R)
,95
AP
−
smallest losed subalgebraof
L
∞
(R)
that ontainsallthe fun tions
e
λ
withλ ≤ 0
, 41AP
+
smallest losed subalgebraof
L
∞
(R)
that ontainsallthe fun tions
e
λ
withλ ≥ 0
, 41AP W
set of all almost periodi fun tions whi h an be written inthe form of an absolutely onvergent series, 40AP W
−
set of allfun tions
ψ ∈ AP W
su h thatΩ(ψ) ⊂ (−∞, 0]
, 41AP W
+
set of allfun tionsψ ∈ AP W
su h thatΩ(ψ) ⊂ [0, +∞)
,41A
p
(z
1
, z
2
) =
z
1
+ (z
2
− z
1
) σ
p
(µ), µ ∈ R
, 97
arg φ
ontinuous argument ofφ
, 39C
+
omplex upperhalf planeℑm z > 0
,15C
−
omplex lower half planeℑm z < 0
, 15C( ˙
R)
set of all(bounded) ontinuous ( omplex-valued) fun tions on˙
R
, 41C(R)
set of all (bounded) ontinuous ( omplex-valued) fun tions onR
with a possiblejump at∞
, 42C
0
( ˙
R)
set ofallfun tionsinC( ˙
R)
su hthatthelimitsat−∞
andat+∞
areequal to zero, 42C
p
( ˙
R)
losure ofC( ˙
R)
inM
p
(R)
, 95C
p
(R)
losure ofC(R)
inM
p
(R)
, 95Coker
T
okernel of the operatorT
, 8D
+
= {z ∈ C : |z| < 1}
, 24D
−
= {z ∈ C : |z| ≥ 1} ∪ {∞}
, 24d(T )
dimension of the okernel of the operatorT
,8d
(φ)
geometri mean value ofφ ∈ GAP
, 40d
l
(φ)
left geometri mean value ofφ ∈ GPAP
, 47sup
F
Fourier transformation,3F
−1
inverse of the Fouriertransformation
F
,3φ
l
almost periodi representative ofφ
at−∞
, 45, 47φ
r
almost periodi representative ofφ
at+∞
, 45, 47GB
group of allinvertible elements of aBana h algebraB
,17H
φ
Hankeloperator, 7H(Θ)
Hausdor set ornumeri al range of the omplex matrixΘ ∈ C
n
×n
, 159
H
∞
(C
±
)
set of allbounded and analyti fun tionsinC
±
, 15H
∞
(D
±
)
set of allbounded and analyti fun tionsinD
±
, 24H
p
(C
±
)
set of all fun tionsφ
whi h are analyti inC
±
and su h thatsup
±y>0
R
R
|φ(x + iy)|
p
dy < ∞
, 15
H
p
(D
+
)
set of all fun tionsφ
whi h are analyti inD
+
and su h thatsup
r
∈(0,1)
R
2π
0
|φ(re
iθ
)|
p
dθ < ∞
, 24H
p
(D
−
)
set of all fun tionsφ(z)
(z ∈ D
−
) for whi hφ(
1
z
)
is a fun tion inH
p
(D
+
)
, 24H
∞
±
(R)
set of all fun tions inL
∞
(R)
that are non-tangential limitsof elements in
H
∞
(C
±
)
, 15H
±
p
(R)
set of all fun tions inL
p
(R)
that are non-tangential limits of elements in
H
p
(C
±
)
,15H
±
p
(T)
set of all fun tions inL
p
(T)
that are non-tangential limits of elements in
H
p
(D
±
)
,24Im
T
image of the operator T, 8ℑm(z)
imaginary part of the omplex numberz
IndT
Fredholm index of the operatorT
,9 indφ
Cau hy index ofφ
, 103, 105, 108, 135inf
inmumI
Z
identity operator onthe Bana h spa eZ
, 7J
ree tion operator onR
, 5, 158J
T
ree tion operator onT
, 25 KerT
kernel of the operator T, 8κ(φ)
mean motion ofφ ∈ GAP
,39κ
l
(φ)
left mean motionofφ ∈ GPAP
, 47κ
r
(φ)
right mean motionofφ ∈ GPAP
, 47ℓ
0
zero extension operator fromL
p
(R
+
)
intoL
p
(R)
,2
ℓ
2
spa e of all innitesequen es
{ξ
i
}
∞
i=1
su h that∞
X
i=1
|ξ
i
|
2
<
∞
, 6ℓ
e
even extension operator from
L
p
(R
+
)
intoL
p
(R)
,7
ℓ
o
odd extensionoperatorfrom
L
p
(R
+
)
intoL
p
(R)
, 7
L(X, Y )
linear spa e of allbounded linear operators from the Bana h spa eX
into the Bana hspa eY
L
∞
(Ω)
L
p
(Ω)
Lebesgue spa e ofp
th
-power-integrablefun tionsonΩ
, 2L
p
+
(R)
subspa e ofL
p
(R)
formed by all the fun tions supported in the losure of
R
+
, 2L
p
−
(R)
subspa e ofL
p
(R)
formed by all the fun tions supported in the losure of
R
−
, 2L
p
even
(T)
set of all fun tions
φ ∈ L
p
(T)
su h that
φ(t) = φ(t
−1
)
, 58
L
p
JT
(T)
=
ImP
JT
|
L
p
(T)
, 58M(φ)
Bohr meanvalue ofφ
(or mean value ofφ
),36M
p
(R)
set of all Fouriermultiplierson
L
p
(R)
, 6
n(T )
dimension of the kernel of the operatorT
,8N
0
= {0, 1, 2, 3, . . .}
,144Ω
open set ofR
,2Ω(φ)
Bohr-Fourierspe trum ofφ
, 38P
set of all trigonometri polynomialsonT
, 30P
orthogonal proje tion ofL
2
(R)
ontoH
2
+
(R)
, 24P
T
orthogonal proje tion ofL
2
(T)
ontoH
2
+
(T)
, 24P
+
anoni al proje tor fromL
p
(R)
onto
L
p
+
(R)
, 2P
−
anoni al proje tor fromL
p
(R)
onto
L
p
−
(R)
, 2P
JT
=
I+JT
2
, 58PAP
algebraof pie ewise almost periodi fun tionsonR
,47P C
algebraof pie ewise ontinuous fun tions on˙
R
, 46P C
0
lassof allpie ewise ontinuous fun tionsforwhi hboth limitsat−∞
and at+∞
are equal to zero, 46P C(T)
algebraof pie ewise ontinuous fun tions onT
, 85ψ
#
ψ
#
: ˙
R × [0, 1] → C
,ψ
#
(x, µ) = (1 − µ) ψ(x − 0) + µ ψ(x + 0)
, 96ψ
p
ψ
p
: ˙
R × R → C
,ψ
p
(x, µ) = (1 − σ
p
(µ)) ψ(x − 0) + σ
p
(µ) ψ(x + 0)
, 98R
set of real numbersR
+
positive half-line, 2R
−
negative half-line, 2˙
R
= R ∪ {∞}
, 41R
= R ∪ {±∞}
, 42R(φ)
essential range ofφ
, 28ℜe(z)
real part of the omplex numberz
r
+
restri tion operator fromL
p
(R)
into
L
p
(R
+
)
, 2S
Cau hy singularintegral operator onL
2
(R)
, 24
S
T
Cau hy singularintegral operator onL
2
(T)
, 24
S(R)
S hwartz spa eof all rapidly de reasing fun tionsonR
, 3SAP
algebraof the semi-almostperiodi fun tions,42SAP
p
smallest losed subalgebra ofM
p
(R)
that ontains
AP
p
andC
p
(R)
, 96sp T
spe trumof the operatorT ∈ L(X)
, 28sp
ess
T
essential spe trum ofthe operatorT ∈ L(X)
, 28σ
p
σ
p
(µ) =
1
2
+
1
2
coth
h
π
i
p
+ µ
i
(µ ∈ R)
,97T
unit ir le,3T
+
= {t ∈ T : ℑm t > 0}
, 88(T +H)
ν
Toeplitz plus Hankel operator, 25(T −H)
ν
Toeplitz minusHankel operator, 25W
Wiener algebra, 4W
φ
Wiener-Hopf operator, 7(W +H)
φ
Wiener-Hopf plus Hankel operator,7(W −H)
φ
Wiener-Hopf minus Hankel operator,7wind
φ
winding number ofφ
, 104, 105, 108, 135, 137, 138k.k
A
norm in the linear spa eA
f ∗ g
onvolution off
andg
, 3e
ϕ
a tion of the ree tion operatoron the fun tionϕ
, 5∼
equivalen e relation,11∗
∼
equivalen e after extensionrelation,11⊕
dire t sum∂A
boundaryof the setA
[c
1
, c
2
]
line segment in the omplex plane between and in luding the endpointsc
1
, c
2
∈ C
,78Wiener-Hopf-Hankel operators (as well as their dis rete analogues based on Toeplitz
and Hankel operators) are well-known to play an important role in several applied areas.
E.g.,this isthe asein ertainwave dira tionproblems,digitalsignalpro essing,dis rete
inverses attering, andlinear predi tion. For on rete examplesofadetailedenrollmentof
thoseoperatorsinthese(andothers)appli ationswereferto[21,23, 32,49,52, 53,68,74℄.
In view of the needs of the appli ations, itis naturalto expe t a orresponding higher
interestinmathemati alfundamentalresear hforthosekindofoperators. Infa t,inre ent
yearsseveralauthorshave ontributedtothemathemati alunderstandingof
Wiener-Hopf--Hankeloperators(andtheirdis reteanalogues)underdierenttypesofassumptions( f.[6,
7,8,20,22,31,41,42,52,63,64,69℄). Asa onsequen e,thetheoryofWiener-Hopf-Hankel
operatorsisnowadayswelldevelopedforsome lassesofFouriersymbols(likeinthe aseof
ontinuousorpie ewise ontinuoussymbols). Inparti ular,the invertibilityand Fredholm
properties of su h kind of operators with pie ewise ontinuous Fourier symbols are now
well known (and of great importan efor the appli ations [23, 49, 52, 74℄). However, this
is not the ase for almostperiodi orsemi-almost periodi Fourier symbols whi h are also
importantinthe appli ations inview of their appearan e due to, e.g., (i)parti ular nite
boundaries in the geometry of physi al problems [21℄, or (ii) the needs of ompositions
withshift operatorswhi hintrodu ealmostperiodi and semi-almostperiodi elementsin
the Fourier symbols of those operators[19, 44℄.
Having this in mind, in this thesis we study the regularity properties of
periodi fun tions and a ting between
L
p
Lebesgue spa es (
1 < p < ∞
). In the asewhere both Wiener-Hopf and Hankel operators have the same Fourier symbol, we obtain
a Wiener-Hopf plus Hankel operator, and in the ase where Wiener-Hopf and Hankel
operatorshave symmetri Fourier symbols,we get a Wiener-Hopfminus Hankel operator.
To study the regularity properties of Wiener-Hopf-Hankel operators, we use ertain
relations between operators that allow us derive the regularity properties of the
Wiener--Hopf-Hankeloperatorsfromthe regularitypropertiesofothersoperatorsforwhi hresults
on erning toregularity properties are known.
This thesis is organized as follows. In the rst hapter, after givingthe formal
deni-tion of the operators under study - the Wiener-Hopf-Hankel operators - several relations
between some lasses of onvolution type operators are investigated. The rst relation is
a multipli ativerelation between Wiener-Hopf plus Hankel operators, whi h is a
general-izationofthe well-known resultfor theWiener-Hopf operators. Usingthe samereasoning,
an analogue multipli ative relation between Wiener-Hopf minus Hankel operators is also
possible to derive. After this, we present a relation between Wiener-Hopf plus Hankel
operators and Wiener-Hopf operators (based on Wiener-Hopf minus Hankel and paired
operators). Sin e this relationis in fa ta
∆
relationafter extensionbetween theWiener--Hopf plus Hankel operatorand the Wiener-Hopfoperator,and therefore there istransfer
of the regularity propertiesfrom the Wiener-Hopf operatorto the Wiener-Hopf plus
Han-kel operator, this relation turns out to be a de isive result in the study of the regularity
properties of the Wiener-Hopf-Hankel operators. The last relations under investigation
in this hapter are two relations of equivalen e between Wiener-Hopf plus/minus Hankel
operators and Toeplitz plus/minus Hankel operators, this due to the equivalen e relation
existent between Wiener-Hopf and Toeplitz operators. Finally, we use these equivalen e
relationsbetween Wiener-Hopf-Hankel and Toeplitz-Hankel operators toderive ne essary
onditions for the semi-Fredholmproperty of Wiener-Hopf-Hankel operators.
Wiener-algebra of almost periodi fun tions, the algebra of semi-almost periodi fun tions, and
the algebra of pie ewise almost periodi fun tions. Some properties of these algebras are
presented aswell.
The main purpose of Chapter 3 is to establish an invertibility and semi-Fredholm
ri-terion for Wiener-Hopf plus Hankel operators with almost periodi Fourier symbols, and
a tingbetween
L
2
Lebesgue spa es. Westart by onsideringthe ase of Wiener-Hopfplus
HankeloperatorswithFouriersymbolsinthesubalgebraofalmostperiodi fun tions
AP W
and then we onsider the more general ase of Wiener-Hopf plus Hankel operators with
almost periodi Fourier symbols. To obtain the invertibility and semi-Fredholm riterion
weintrodu eafa torization on ept forthe almostperiodi Fouriersymbolssu h thatthe
properties of the fa tors will allow orresponding operator fa torizations. Using then the
multipli ativerelationpresented inChapter 1,a riterionfor the semi-Fredholmproperty
and for the one-sided and both-sided invertibility is therefore obtained upon ertain
in-di es of the fa torizations. Under su h onditions, the one-sided and two-sided inverses
of the operators are alsoobtained. Moreover, the introdu ed fa torizations alsoallowthe
expositionofdependen ies between theinvertibilityof Wiener-HopfandWiener-Hopfplus
HankeloperatorswiththesameFouriersymbol. AlltheseresultsholdtrueforWiener-Hopf
minus Hankel operators sin e we use an analogue multipli ative relationfor Wiener-Hopf
minusHankeloperators of that one used for Wiener-Hopf plus Hankel operators.
InChapter4,Wiener-Hopf-Hankeloperatorswithsemi-almostperiodi Fouriersymbols
are info us. Consideringinrstpla e theseoperatorsa tingbetween
L
2
Lebesgue spa es,
and motivated by the Sarason's Theorem, we obtain a generalization of the invertibility
and semi-Fredholm riteria a hieved inChapter 3 for Wiener-Hopf-Hankel operatorswith
almost periodi symbols - the so- alled Sarason's type theorem. This result is attained
by using the
∆
relation after extension and also the main idea of the approa h of R.V. Dudu hava and A. I. Saginashvili [29℄. After settled the Sarason's type theorem, a
reformulation of the invertibility and semi-Fredholm riterionfor Wiener-Hopf plus
tors with almost periodi Fourier symbols without the assumption on the orresponding
fa torization of the Fourier symbol of the operator, being this also possible due to the
Sarason'stypetheorem. Inase ondpart ofthis hapter,we onsiderWiener-Hopf-Hankel
operators with semi-almost periodi Fourier symbols a ting between
L
p
Lebesgue spa es
(
1 < p < ∞
). Having by motivation the Dudu hava-Saginashvili's theorem, we a hieve ageneralizationof the Sarason's type theorem, the so- alledDudu hava-Saginashvili's type
theorem. The fundamental key to rea h this result is the
∆
relation after extensionbe-tweentheWiener-HopfplusHankeloperatorandtheWiener-Hopfoperator. Alltheresults
mentioned above meana hara terization of the Fredholmproperty, and one-sided
invert-ibility of the Wiener-Hopf-Hankel operators, based on the mean motions and geometri
mean values of the almost periodi representatives of the Fourier symbols at minus and
plus innity. We end up this hapter with a se tion devoted to the Fredholm index of
the Wiener-Hopf-Hankeloperatorsunder studyin this hapter. Inthe rst ase wherethe
operatorsare onsidered a tingbetween
L
2
Lebesgue spa es,a Fredholmindex formula is
derived in terms of the winding number of the Fourier symbols of the operators, while in
the se ond ase where we onsider the operators a ting between
L
p
Lebesgue spa es, the
Fredholm index formulais given in terms of the winding number of ontinuous fun tions
onstru tedfromtheFouriersymbolsoftheoperators. Additionally,inthe aseof
Wiener--Hopf-Hankeloperatorsa tingbetween
L
2
Lebesguespa es, onditions fortheinvertibility
of these operatorsare alsoobtained.
Chapter 5 presents a riterion for the semi-Fredholm property and for the one-sided
invertibilityof Wiener-Hopf-Hankeloperatorswith pie ewisealmost periodi Fourier
sym-bols,a tingbetween
L
2
Lebesguespa es. This riterionisalsoobtainedupontheuseofthe
∆
relation after extension and, in virtue of the nature of pie ewise almost periodi fun -tions, it is based on the mean motionsand geometri mean values of the almost periodirepresentatives of the Fourier symbols as well as on the dis ontinuities of ertain
Hankeloperators. Sin einthis hapterwearedealingwithoperatorshavingdis ontinuous
symbols,theFredholmindexformulaisalsointerpretedupondierent asesofsymmetries
of the dis ontinuities of the Fourier symbols. In ordertoexemplify the simpli ationthat
o urs in the Fredholm index formula due to the symmetries of the dis ontinuities of the
Fourier symbol, several examples are presented. Finally, the (both-sided) invertibility of
the operators instudy isalsodis ussed.
In the last hapter, motivated by a result due to R. G. Babadzhanyan and V. S.
Ra-binovi h that settles the (one-sided and both-sided) invertibility and the semi-Fredholm
property of Wiener-Hopf operators with matrix
AP W
Fourier symbols having Hausdorsets bounded away from zero, we derive a generalization of the invertibility and
semi--Fredholm riteria obtained in Chapter 3. Here, we onsider Wiener-Hopf-Hankel
opera-tors a ting in
L
2
Lebesgue spa es with matrix APW Fourier symbols having a parti ular
Hausdor set bounded away from zero and obtain for these operators aninvertibility and
semi-Fredholm riteria. Similarlytothe resultobtained inTheorem3.2.1, theinvertibility
and semi-Fredholm riteria here presented is given in terms of the mean motion, in this
ase,dependingonaparti ularHausdorsetboundedawayfromzero(insteadofthemean
motion of the Fouriersymbol of the operatoras inTheorem 3.2.1).
As far as the author knows, the results presented in this thesis are new. Most of the
material is published or a epted for publi ation in journals or onferen e pro eedings
[58, 59, 60, 61℄, and the material whi h is not published or a epted for publi ation is
submitted for publi ationin journals [17, 57℄. Auxiliary results of other authors in luded
Convolution Type Operators
The study of relations between dierent lasses of operators is an important subje t
in Operator Theory sin e, in parti ular, itallows the transfer of properties from one lass
of operators to the other ones. Due to this transfer of properties and in order to a hieve
the invertibilityand semi-Fredholm riteriapresented in Chapters 3,4, 5and 6,the main
purpose of this hapter isthe study of some relationsbetween some lassesof onvolution
type operators. In rst pla e, we present a multipli ative relation between Wiener-Hopf
plus Hankel operators (whi h in fa tis a orresponding result for Wiener-Hopf plus
Han-kel operators of the well-known result for the Wiener-Hopf operators). In se ond pla e,
it is exhibited a relation between Wiener-Hopf plus Hankel operators and Wiener-Hopf
operators (based on Wiener-Hopf minus Hankel and paired operators). This relation will
perform a signi ant role in the obtainment of several results be ause it allows to study
the regularity propertiesof Wiener-Hopf operators, and then transferthe regularity
prop-erties to Wiener-Hopf-Hankel operators. Finally, in what on erns to relations between
onvolution type operators, sin e Wiener-Hopf and Toeplitz operators in Hilbert spa es
areequivalentoperators,itisalsopresented awaytorelateWiener-Hopf-Hankeloperators
with Toeplitz-Hankel operators. Here and in what follows, we will simply all
Wiener--Hopf-HankeloperatorstobothWiener-Hopfplus Hankel,and Wiener-HopfminusHankel
Toeplitz plus Hankel, and Toeplitz minus Hankel operators. Using equivalen e relations
between Wiener-Hopf-Hankel and Toeplitz-Hankel operators, we end up this hapter by
deriving ne essary onditions for the semi-Fredholm property of Wiener-Hopf-Hankel
op-erators.
Before presenting the results mentioned above, we will rst introdu e the formal
de-nitions of the operators under study as well as some of the notation that will be used
throughout this thesis.
1.1 Some denitions, notations, and histori al notes
Let
Ω
be an open set ofR
. For1 ≤ p < ∞
,L
p
(Ω)
represents the Bana h spa e of
Lebesgue measurable omplex-valued fun tions
ϕ
onΩ
su h that|ϕ|
p
is integrable. This
spa e isendowed with the norm
kϕk
L
p
(Ω)
:=
Z
Ω
|ϕ(η)|
p
dη
1
p
.
(1.1.1) Furthermore,letL
p
+
(R)
denotethesubspa eofL
p
(R)
formedbyallthefun tionssupported
in the losure of
R
+
:= (0, +∞)
, andL
p
−
(R)
represent the subspa e ofL
p
(R)
formed by
all the fun tions supported in the losure of
R
−
:= (−∞, 0)
. We will use the anoni al proje torsP
+
andP
−
that mapL
p
(R)
ontoL
p
+
(R)
, andL
p
(R)
ontoL
p
−
(R)
, respe tively. Consideringr
+
being the restri tion operator fromL
p
(R)
into
L
p
(R
+
)
andℓ
0
the zero extension operator fromL
p
(R
+
)
intoL
p
(R)
, wehaveP
+
= ℓ
0
r
+
.
(1.1.2) ConsideralsoL
∞
(Ω)
astheBana hspa eofLebesguemeasurableandessentiallybounded
( omplex-valued)fun tions on
Ω
. The norm inL
∞
(Ω)
isgiven by
kφk
L
∞
(Ω)
:=
esssup |φ(η)|,
(1.1.3)where
For the ase of Lebesgue spa es dened on the unit ir le
T := {z ∈ C : |z| = 1}
, letL
p
(T) (1 ≤ p < ∞)
stand for the Bana h spa e of Lebesgue measurable omplex-valued
fun tions
ϕ
onT
su h that|ϕ|
p
is integrable. The norm in
L
p
(T)
is given bykϕk
L
p
(T)
:=
1
2π
Z
2π
0
|ϕ(e
iθ
)|
p
dθ
1
p
.
(1.1.5) Moreover, letL
∞
(T)
denote the Bana h spa e of Lebesgue measurable and essentially
bounded ( omplex-valued) fun tions dened on
T
.L
∞
(T)
is endowed with the essential
supremum norm
kφk
L
∞
(T)
:=
esssup |φ(η)|,
(1.1.6)where, inthis ase,
ess
sup |φ(η)| := inf{α : |{η ∈ T : |φ(η)| > α}| = 0}.
(1.1.7)Let
F
denotetheFouriertransformation dened inthe S hwartz spa eS(R)
ofrapidlyde reasing fun tions on
R
byFϕ(ξ) :=
Z
R
e
iξη
ϕ(η) dη,
ξ ∈ R ,
(1.1.8) and letF
−1
denote the inverse of the Fourier transformation
F
, also dened onS(R)
,given by
F
−1
ψ(η) :=
1
2π
Z
R
e
−iξη
ψ(ξ) dξ,
η ∈ R .
(1.1.9)A onvolution operator is dened by
Cϕ(ξ) := (K ∗ ϕ)(ξ) =
Z
R
K(ξ − η)ϕ(η) dη,
ξ ∈ R,
(1.1.10)
where
K
is alled the onvolution kernel ofC
. Convolution operators may be onsidereda ting in several fun tions spa es as well as their onvolution kernels that may also be
onsidered in dierent fun tions spa es. For instan e, we may onsider the onvolution
operator
C
a tinginL
1
(R)
with
K ∈ L
1
(R)
. The onvolutionoperator
C
anbewritteninthe form
with
φ
C
= FK
. From this, we see whyφ
C
is alled the Fourier symbol of the onvolution operatorC
. If in (1.1.10) instead ofR
we onsider the open setΩ ⊂ R
, then we obtain aonvolution type operator
Aϕ(ξ) := (K ∗ ϕ)(ξ) =
Z
Ω
K(ξ − η)ϕ(η) dη,
ξ ∈ Ω.
(1.1.12)
The probably best known onvolution type operators are the Wiener-Hopf operators
a ting between Lebesgue spa es on the half-line. We re all that the name Wiener-Hopf
operatorsisduetotheinitialworkofN.WienerandE.Hopfin1931[76℄whereareasoning
to solve integral equations whose kernels depend only on the dieren e of the arguments
was provided:
cf (x) +
Z
+∞
0
k(x − y)f(y)dy = g(x), x ∈ R
+
,
(1.1.13)i.e.theso- alledintegralWiener-Hopfequations. Here
c ∈ C, k ∈ L
1
(R)
and
f, g ∈ L
2
(R
+
)
, wherec
andk
are xed,g
is given andf
is the unknown element.From those Wiener-Hopfequations arise the ( lassi al)Wiener-Hopf operators dened
by
W
φ
f (x) = cf (x) +
Z
+∞
0
k(x − y)f(y)dy , x ∈ R
+
,
(1.1.14)where
φ = c + Fk
belongs tothe Wiener algebra. The Wiener algebrais dened byW := {φ : φ = c + Fk, c ∈ C, k ∈ L
1
(R)}
(1.1.15)and itis a Bana halgebra when endowed with the norm
kc + Fkk
W
:= |c| + kkk
L
1
(R)
(1.1.16)and the usual multipli ation operation. The Wiener algebra is a subalgebra of
L
∞
(R)
.
Having in mind the onvolution operation, the denition of the lassi al Wiener-Hopf
operators gives rise tothe following representation of the Wiener-Hopf operators
W
φ
:= r
+
F
−1
φ · F : L
2
+
(R) → L
2
(R
+
) .
(1.1.17) Making use of the anoni al proje tor onL
2
+
(R)
, the Wiener-Hopf operators onL
2
+
(R)
may alsobe writtenin the form
where
A
is the translation invariant operatorF
−1
φ · F
. Lookingnow to the stru ture of
the operators in (1.1.17) and (1.1.18), we re ognize that possibilities other than only the
Wiener algebra an be onsidered for the so- alled Fourier symbols
φ
of the Wiener-Hopfoperators. Namely,we may onsider to hoose
φ
among theL
∞
(R)
elements.
Withinthe ontextof (1.1.13)and(1.1.14),theHankelintegral operatorshavetheform
Hf (x) =
Z
+∞
0
k(x + y)f (y)dy ,
x ∈ R
+
(1.1.19) for somek ∈ L
1
(R)
. In this ase, itis well-known that
H
, asan operatordened betweenL
2
spa es, is a ompa t operator. However, as seen above, itis also possible to provide a
rigorousmeaning tothe expression (1.1.19) when the kernel
k
isa temperate distributionwhose Fourier transform belongs to
L
∞
(R)
. In su h ase, Hankel operators admit the
representation
H
φ
:= r
+
F
−1
φ · FJ : L
2
+
(R) → L
2
(R
+
) .
(1.1.20) Here and inwhat follows,J
isthe ree tion operator onR
given by the ruleJϕ(x) = e
ϕ(x) = ϕ(−x), x ∈ R .
(1.1.21)Likein (1.1.18),we may write Hankel operatorson
L
2
+
(R)
asP
+
AJ = ℓ
0
r
+
F
−1
φ · FJ : L
2
+
(R) → L
2
+
(R) ,
(1.1.22) beingA = F
−1
φ · F
. We would like to mention that the dis rete analogue of
H
has itsroots in the year of 1861 with the Ph.D. thesis of H. Hankel [39℄. There the study of
nite matri es with entries depending only on the sum of the oordinates was proposed.
Determinantsof innite omplexmatri eswith entries dened by
a
jk
= a
j+k
(forj, k ≥ 0
, and wherea = {a
j
}
j
≥0
is a sequen e of omplex numbers) were also studied. For these (innite) Hankel matri es, one of the rst main results was obtained by L. Krone ker in1881 [47℄ when hara terizing the Hankel matri es of nite rank as the ones that have
orresponding power series,
a(z) =
∞
X
j=0
a
j
z
j
, whi h are rational fun tions. In 1906, D. Hilbert proved that the operator(indu ed by the famous Hilbert matrix),H : ℓ
2
→ ℓ
2
, {b
j
}
j
≥0
7→
(
∞
X
k=0
b
k
j + k + 1
)
j
≥0
,
(1.1.23)is bounded on the spa e
ℓ
2
of all innite sequen es
{ξ
i
}
∞
i=1
su h that∞
X
i=1
|ξ
i
|
2
< ∞
. This resultmaybeviewed astheoriginof(dis rete)Hankeloperators,asnaturalobje tsarisingfromHankelmatri es. Lateron,in1957,Z.Neharipresenteda hara terizationofbounded
Hankel operators on
ℓ
2
[55℄. Due tothe importan eof su h hara terization, we may say
that itmarks the beginningof the ontemporary periodof the study of Hankel operators.
After a brief note on the origins of Wiener-Hopf and Hankel operators and before
presentthedenitionofWiener-HopfplusHankeloperatorsandWiener-HopfminusHankel
operators,let usre all the denition of Fourier multiplieron
L
p
(R)
.
A fun tion
φ ∈ L
∞
(R)
is alledaFouriermultiplier on
L
p
(R)
if the operatorF
−1
φ · F
, a tingonL
2
(R) ∩ L
p
(R)
, extends by ontinuity toa bounded operator on
L
p
(R)
. The set
of all Fouriermultiplierson
L
p
(R)
is denoted by
M
p
(R)
and it is a Bana h algebrawhen
endowed with the norm
kφk
M
p
(R)
:= kF
−1
φ · Fk
L(L
p
(R))
(1.1.24)andpointwise multipli ation. Further,the set ofallFouriermultiplierson
L
2
(R)
oin ides withL
∞
(R)
, i.e.,M
2
(R) = L
∞
(R).
(1.1.25)We are now in position to present the main obje ts of this thesis - the Wiener-Hopf
plus Hankel operators and the Wiener-Hopf minus Hankel operators. As the names
sug-gest,theseoperatorsresultfrom ombinationsbetweenWiener-HopfandHankeloperators.
Combinationsofsu hkindappearforthersttimein1979inthe lassi alworkofS.Power
[62℄, wherea study of the spe traand essentialspe tra of Hankel operatorswas presented
by investigating the
C
∗
algebra generated by Toeplitz and Hankel operators (in the two
ases ofpie ewise ontinuous symbols and almostperiodi symbols). Althoughinthe
ma-jorpartofthisthesiswewill onsider operatorsa tingbetween
L
2
Lebesguespa es,wewill
now dene Wiener-Hopf plus Hankel operators and Wiener-Hopf minusHankel operators
a tingbetween
L
p
For
φ ∈ M
p
(R)
, let
W
φ
andH
φ
be the Wiener-Hopf and Hankel operators dened byW
φ
:= r
+
F
−1
φ · F : L
p
+
(R) → L
p
(R
+
)
(1.1.26)H
φ
:= r
+
F
−1
φ · FJ : L
p
+
(R) → L
p
(R
+
),
(1.1.27)respe tively. The Wiener-Hopf plus Hankel operator is given by
(W +H)
φ
:= W
φ
+ H
φ
: L
p
+
(R) → L
p
(R
+
),
(1.1.28)and the Wiener-Hopf minus Hankeloperatoris dened by
(W −H)
φ
:= W
φ
− H
φ
: L
p
+
(R) → L
p
(R
+
).
(1.1.29) ForaBana hspa eZ
,letI
Z
representtheidentityoperatoronZ
. A ordingto(1.1.26), (1.1.27)and (1.1.28), we have(W +H)
φ
= r
+
(F
−1
φ · F + F
−1
φ · FJ) = r
+
F
−1
φ · F(I
L
p
+
(R)
+ J).
(1.1.30) Furthermore, sin eI
L
p
+
(R)
+ J = ℓ
e
r
+
,
(1.1.31) whereℓ
e
denotes the even extension operator from
L
p
(R
+
)
intoL
p
(R)
, we may write the
Wiener-Hopf plus Hankel operator as
(W +H)
φ
= r
+
F
−1
φ · Fℓ
e
r
+
.
(1.1.32)Combiningnow (1.1.26),(1.1.27) and (1.1.29), we have
(W −H)
φ
= r
+
(F
−1
φ · F − F
−1
φ · FJ) = r
+
F
−1
φ · F(I
L
p
+
(R)
− J).
(1.1.33) Denoting byℓ
o
the odd extension operator from
L
p
(R
+
)
intoL
p
(R)
,it follows that
I
L
p
+
(R)
− J = ℓ
o
r
+
.
(1.1.34)In this way, we may write the Wiener-Hopf minusHankeloperator as
From(1.1.32)and (1.1.35),weobservethatWiener-Hopf-Hankeloperatorsmaybewritten
as onvolutiontypeoperatorswith symmetry. Noti e that onvolution type operators with
symmetry (CTOS) are operatorsof the form
T = r
+
Al
c
: L
p
(R
+
) → L
p
(R
+
) ,
(1.1.36)where
A = F
−1
φ · F : L
p
(R) → L
p
(R)
is atranslation invariantoperator and supposed to
bebounded invertible,and
ℓ
c
denotes even or oddextension asa ontinuous operatorfrom
L
p
(R
+
)
intoL
p
(R)
. CTOS may also be onsidered in Bessel potential spa es. Operators
of this type were rst studied in[51℄and then in[22℄.
1.2 Regularity properties
Let
X
andY
be two Bana hspa es and onsiderT ∈ L(X, Y )
. The kernel KerT
andthe image Im
T
of the operatorT
are dened byKer
T := {x ∈ X : T x = 0},
ImT := {T x : x ∈ X}.
(1.2.37)Ker
T
and ImT
are linearsubspa es ofX
andY
, respe tively. Moreover, KerT
isa losedsubspa e. IfIm
T
is alsoa losedsubspa e, the operatorT
issaid tobenormally solvable.In this ase, the okernel of T isdened as
Coker
T := Y /
ImT.
(1.2.38)Fora normallysolvable operator
T
, the de ien y numbersofT
are dened asn(T ) := dim
KerT,
d(T ) := dim
CokerT.
(1.2.39)Con erning the kernel of
T
, it holds that KerT
satises (at least) one of the followingonditions:
(1)
n(T ) = 0
;(3) Ker
T
is omplementable inX
;(4) orKer
T
is losed inX
.A similar hara terization an be donefor the image of
T
, thatis, ImT
satises(at least)one of the following onditions:
(i)
d(T ) = 0
;(ii)
d(T ) < ∞
;(iii) Im
T
is omplementableinY
;(iv) orIm
T
is losed inY
.Combiningthese properties on erning to the kernel of
T
with the properties referring tothe image of
T
, we obtain several lasses that are alled regularity lasses [16, 18, 73℄. Inwhat follows,we willdene some of these regularity lasses.
The operator
T
is said to be a Fredholm operator if it is normally solvable andn(T )
and
d(T )
are nite. In this ase, the Fredholm index ofT
is dened byInd
T := n(T ) − d(T ).
(1.2.40)The operator
T
is said to be a semi-Fredholm operator if it is normally solvable and atleast one of the de ien y numbers
n(T )
andd(T )
is nite. A semi-Fredholm operatoris said to be
n
-normal ifn(T )
is nite, andd
-normal ifd(T )
is nite. In the ase whereonly one of the de ien y numbers is nite, the operator
T
is said to be a properlysemi--Fredholm operator. In this ase, the operator
T
is said to be properlyn
-normal ifn(T )
is nite and
d(T )
is innite, and properlyd
-normal ifd(T )
is nite andn(T )
is innite.Forsemi-Fredholmoperators,the indexformula(1.2.40)isalsowell-dened. In parti ular,
forproperlysemi-Fredholmoperators,we have Ind
T = −∞
ifT
isproperlyn
-normalandInd
T = +∞
ifT
is properlyd
-normal.We point out that in German and Russian literature, (semi-)Fredholm operators are
whowasthersttodis overthatsingularintegraloperatorswithnonvanishing ontinuous
symbols are normally solvable, and have nite kernel and okernel dimensions.
Still on erning to semi-Fredholm operators, we present next a property about the
index of these operators whi h willbe used later on.
Theorem 1.2.1. ( f. [35, 6.7℄) If
T ∈ L(X, Y )
is a Fredholm operator (resp. properlyn
-normal, properlyd
-normal), thenthere existsa numberδ > 0
su hthat, forall operatorsM ∈ L(X, Y )
withkMk < δ
,A + M ∈ L(X, Y )
is a Fredholm operator (resp. properlyn
-normal, properlyd
-normal)and Ind(A + M) =
IndA
.Weend up this se tion by introdu ingthe denition of reexive generalized
invertibil-ity. As we will see in a moment, it is possible to hara terize the regularity lass of all
operators
T
su hthat KerT
is a omplementable subspa e ofX
and ImT
is a losed andomplementable subspa e of
Y
in terms of the reexive generalizedinvertibilityofT
.Let
T : X → Y
be abounded linear operatora ting between Bana h spa es.T
is saidtobe reexive generalized invertible if thereexists abounded linear operator
T
−
: Y → X
su h that
T T
−
T = T
andT
−
T T
−
= T
−
.
(1.2.41)
Inthis ase,theoperator
T
−
isreferredtoasthereexivegeneralizedinverse (or
pseudoin-verse)of
T
. From the denition ofreexive generalizedinvertibleoperator,it follows thata linear bounded one-sided or two-sided invertible operator is also a reexive generalized
invertibleoperatorandoneofitsreexivegeneralizedinversesisthe one-sidedortwo-sided
inverse, respe tively. Ingeneral,reexive generalized inverses are not unique. However, in
the ase where
T
is invertible, the reexive generalized inverses are unique and oin idewith the inverse of
T
.Finally,and about the hara terization (mentionedabove)of the regularity lass of all
operators
T
su hthat KerT
is a omplementable subspa e ofX
and ImT
is a losed andomplementable subspa e of
Y
, wehavethe following:T
isreexivegeneralized invertibleif and only if Ker
T
is a omplementable subspa e ofX
and ImT
is a losedandgeneralized invertible.
1.3 Relations between onvolution type operators
1.3.1 Relations between bounded linear operators
In what follows, onsider
T : X
1
→ X
2
andS : Y
1
→ Y
2
two bounded linear operators a ting between Bana h spa es.The operators
T
andS
are said to be equivalent, and we will denotethis byT ∼ S
, ifthere are two boundedly invertible linear operators,
E : Y
2
→ X
2
andF : X
1
→ Y
1
, su h thatT = E S F.
(1.3.42)It dire tly follows from (1.3.42) that if two operators are equivalent, then they belong to
thesameregularity lass. Namely,oneofthese operatorsisinvertible,one-sidedinvertible,
Fredholm, (properly)
n
-normal, (properly)d
-normal or normally solvable, if and only ifthe other operator enjoys the same property.
Anoperatorrelationthatgeneralizestheoperatorequivalen erelationistheequivalen e
after extension relation. The operators
T
andS
are said tobe equivalent after extension,and we will denote this by
T
∗
∼ S
, if there exist two Bana h spa esW
andZ
su h thatT ⊕ I
W
andS ⊕ I
Z
are equivalent operators, i.e.,
T
0
0 I
W
= E
S
0
0 I
Z
F,
(1.3.43)for invertible bounded linearoperators
E : Y
2
× Z → X
2
× W
andF : X
1
× W → Y
1
× Z
. As we an easily see, the operator equivalen e relation orresponds to the ase where theextension spa es
W
andZ
are hosen to be the trivial spa e (in the equivalen e afterextension relation). Like inthe equivalen e ase, two equivalentafter extension operators
belong to the same regularity lass.
Another known relation between bounded linear operators is the matri ial oupling.
linear operators
T
T
2
T
1
T
0
: X
1
× Y
2
→ X
2
× Y
1
,
(1.3.44)
S
0
S
1
S
2
S
: X
2
× Y
1
→ X
1
× Y
2
(1.3.45) su h that
T
T
2
T
1
T
0
−1
=
S
0
S
1
S
2
S
.
(1.3.46)In [3℄, H. Bart, I. Gohberg, and M. A. Kaashoek proved that matri ial oupling implies
equivalen e after extension. Later on, in [4℄, H. Bart and V. E. Tsekanovshii proved that
the onverse is alsotrue, and therefore,matri ial ouplingand equivalen eafter extension
amount tothe same. That is tosay that twooperators a ting between Bana h spa es are
equivalent after extensionif and only if they are matri ially oupled.
Con erning additional operators relations, we nd the
∆
relation and the∆
relationafterextension thatwereintrodu edbyL.P.CastroandF.-O.Spe k in[15,18℄. Inregard
to the
∆
relation, that an be viewed as a generalization of the equivalen e relation, wesaythat
T
is∆
related withS
ifthereisabounded linearoperatora tingbetween Bana hspa es
T
∆
: X
1∆
→ X
2∆
and two invertible bounded linear operatorsE : Y
2
→ X
2
× X
2∆
andF : X
1
× X
1∆
→ Y
1
su h that
T
0
0 T
∆
= ESF.
(1.3.47)As we an see here, in general the extension is not made with the identity operator, like
in the equivalen e after extension relation ( f. (1.3.43)), but with a third new operator
T
∆
. Therefore, due to the presen e of three operators,T
,T
∆
andS
, the relation (1.3.47) is alled∆
relation.In its turn, the
∆
relation after extension appears a generalizationof the∆
relation.Thus,
T
issaid tobe∆
related afterextension withS
ifthereisaboundedlinearoperatorlinear operators
E : Y
2
× Z → X
2
× X
2∆
andF : X
1
× X
1∆
→ Y
1
× Z
su h that
T
0
0 T
∆
= E
S
0
0 I
Z
F.
(1.3.48)From (1.3.47)and (1.3.48)it follows that if we have
T
being∆
relatedwithS
orT
being∆
related after extension withS
, then the transfer of regularity properties an only be guaranteed in one dire tion, that is, from operatorS
to operatorT
, as stated in [18,Theorem 2.1℄. It is lear that this restri tion o urs only here and in ontrast to what
happenswiththe transferofregularitypropertiesbetween twoequivalent(after extension)
operators,where the transfer an bedone in both dire tions.
1.3.2 Fa torization of Wiener-Hopf plus Hankel operators
Inordertoobtaininvertibilityand semi-Fredholm riteriaforWiener-HopfplusHankel
operators with almost periodi Fourier symbols, and a ting between
L
2
Lebesgue spa es,
we need a fa torization theory for these operators. In this sense, in what follows, we
will onsider Wiener-Hopf, Hankel and Wiener-Hopf-Hankel operators a ting between
L
2
Lebesgue spa es. Wepointout that similar resultshold true for operatorsa ting between
L
p
Lebesgue spa es.
We will start by re alling two well-known relations between Wiener-Hopf and Hankel
operators that arise fromthe Wiener-Hopf and Hankeloperator theory.
Proposition 1.3.1. ([12, Proposition 2.10℄,[75℄) Let
φ, ϕ ∈ L
∞
(R)
. ThenW
φϕ
= W
φ
ℓ
0
W
ϕ
+ H
φ
ℓ
0
H
ϕ
e,
(1.3.49)H
φϕ
= W
φ
ℓ
0
H
ϕ
+ H
φ
ℓ
0
W
ϕ
e.
(1.3.50)Proof. Takingintoa ount that
I
L
2
(R)
= P
+
+ P
−
= P
+
+ P
−
JJP
−
(1.3.51)and
itholds
I
L
2
(R)
= P
+
+ JP
+
P
+
J = P
+
+ JP
+
J.
(1.3.53) Therefore, it followsW
φϕ
= r
+
F
−1
(φϕ) · F
= r
+
F
−1
φ · FF
−1
ϕ · F
= r
+
F
−1
φ · F(P
+
+ JP
+
J)F
−1
ϕ · F
= r
+
F
−1
φ · FP
+
F
−1
ϕ · F + r
+
F
−1
φ · FJP
+
JF
−1
ϕ · F
= r
+
F
−1
φ · F ℓ
0
r
+
F
−1
ϕ · F + r
+
F
−1
φ · FJ ℓ
0
r
+
F
−1
ϕ · FJ
e
= W
φ
ℓ
0
W
ϕ
+ H
φ
ℓ
0
H
ϕ
e (1.3.54) andH
φϕ
= r
+
F
−1
(φϕ) · FJ
= r
+
F
−1
φ · FF
−1
ϕ · FJ
= r
+
F
−1
φ · F(P
+
+ JP
+
J)F
−1
ϕ · FJ
= r
+
F
−1
φ · FP
+
F
−1
ϕ · FJ + r
+
F
−1
φ · FJP
+
JF
−1
ϕ · FJ
= r
+
F
−1
φ · F ℓ
0
r
+
F
−1
ϕ · FJ + r
+
F
−1
φ · FJ ℓ
0
r
+
F
−1
ϕ · F
e
= W
φ
ℓ
0
H
ϕ
+ H
φ
ℓ
0
W
ϕ
e.
(1.3.55)HavinginmindProposition1.3.1andthinkingonthe lassofWiener-HopfplusHankel
operators, it is natural to expe t a orresponding formula for this lass of operators. In
fa t,forWiener-HopfplusHankeloperatorsalsoholdsananalogueof formula(1.3.49)like
itis stated inthe following proposition.
Proposition 1.3.2. Let
φ, ϕ ∈ L
∞
(R)
. Then
(W +H)
φϕ
= (W +H)
φ
ℓ
0
(W +H)
ϕ
+ H
φ
ℓ
0
(W +H)
ϕ
e−ϕ
.
Proof. Adding (1.3.49)and (1.3.50),we obtain
(W +H)
φϕ
= W
φ
ℓ
0
(W +H)
ϕ
+ H
φ
ℓ
0
(W +H)
ϕ
e.
(1.3.57)
Addingand subtra ting
H
φ
ℓ
0
(W +H)
ϕ
onthe right-handsideof the lastidentity,itholds(W +H)
φϕ
= (W +H)
φ
ℓ
0
(W +H)
ϕ
+ H
φ
ℓ
0
(W +H)
ϕ
e−ϕ
.
(1.3.58) LetC
+
:= {z ∈ C : ℑm z > 0}
andC
−
:= {z ∈ C : ℑm z < 0}
. Asusual,letH
∞
(C
±
)
denote the set of all bounded and analyti fun tions in
C
±
. Fatou's Theorem asserts that fun tions inH
∞
(C
±
)
have non-tangential limits onR = ∂C
±
almost everywhere. In this sense, letH
∞
±
(R)
be the set of all fun tions inL
∞
(R)
that are non-tangential
limitsof elementsin
H
∞
(C
±
)
.H
∞
+
(R)
andH
∞
−
(R)
are losed subalgebras ofL
∞
(R)
. For
0 < p < ∞
,H
p
(C
±
)
denote the set of all fun tionsφ
whi h are analyti inC
±
and su h thatsup
±y>0
Z
R
|φ(x + iy)|
p
dy < ∞ .
(1.3.59)Like in the ase of
H
∞
(C
±
)
, by Fatou's theorem, it also holds that fun tions inH
p
(C
±
)
have non-tangential limits almost everywhere on
R
. The set of all these non-tangentialfun tions isdenoted by
H
p
±
(R)
. For1 < p < ∞
,H
p
±
(R)
is a losed subspa e ofL
p
(R)
. It is well-known that: ifφ ∈ H
∞
+
(R)
, thenH
eφ
= 0;
(1.3.60) ifφ ∈ H
∞
−
(R)
, thenH
φ
= 0.
(1.3.61)Thesetwosimplefa tsareveryimportantinthetheoryofWiener-Hopfoperators. Namely,
it is possible to fa torize the Wiener-Hopf operator if its Fourier symbol admits a
fa tor-ization wherethe left fa tor belongsto