VOL. 21 NO. 2 (1998) 397-402
FOURIER TRANSFORMS OF LIPSCHITZ FUNCTIONS ON THE HYPERBOLIC PLANE H =
M.S. YOUNIS Departmentof Mathematics
YarmoukUniversity Irbid,JORDAN
(ReceivedApril 24,
1996)
ABSTRACT. The purposeof the present work is to study the order of magnitude of theFourier transforms
f(A)
forlarse A
ofcomplex-valued
functions,f (z) sating
certainLipschitz
conditions in the non-Euclideanhyperbolic planeH .
KEY
WORDS AND PHRASES: Fouriertransforms, Lipschitz functions, absoluteconvergence of Fouriertransforms.1991AMSSUB/ECTCLASSIFICATION CODES: Primary 42B10.
1. INTRODUCTION
Therelation between smoothness conditions imposed on functions
f(z)
andthe behavior of its Fouriertransformsf
nearinfinity is well knowninthe literature.In fact, the Fourier transforms of Lipschitz functions defined on various domains have been extensivelystudiedover the lastdecades. Thepurposeofthe present researchis totracethe behaviorof the Fouriertransformsofcomplex-valuedfunctionssatisfyingLipschimconditions inthehyperbolic plane
H 9-.
This will pavethe ground forhandling the transforms ofLipschitz functions defined on other domainssuchasSL(2.R)
andSL(2.C)
inparticular.2. DEFINITIONS AND NOTATIONS
Our main reference on the Fourier analysis on
H
2 is the book by Helgason([2],
p. 29 infra);reference[7]willbe consulted (especially Chapter 10)aswell. Inthe beginningwewouldlike tomake clear an idea which will befrequentlyencoumered in thesequel. Let
D
betheopendiscIzl <
1intheplane
R 2.
Then ahorocycleinD
is a circletangentialtothe boundaryB
ODat acertainpointb. This horocycleisdenotedby.
Ifzisa point on,
the distanced(0, z)
from the origin0ofD
tozisdenoted by(z, b).
Letd(0, z)
r. Thenr
d(0, z)
log-iz----
1Thisgives
z[
tanhr.We
indicate here that tanhr is ofthe order ofr nearr 0and that it approaches1asrgoestoinfinity. Thesetwolimits will occurfrequentlyin duecourse. Inthis section weintroduce the basic definitionsnecessary fortheproofof themaintheorems. Other definitionswillbe givenwhenthey are needed. Thus westart withDEFINITION 2.1. Let
f(z)
be a complex-valued function defined on the unit discD
when endowedwiththe Pdemannianstructure. Thenf(z)
is said tobelongtothe Lipschitz class Lip(c,2)
ifII.f(z + h) .f(z)ll O(Ihl) (2)
398 M.S.YOUNIS
ash 0, 0
<
a<
1. WhereI[-[[2
isthe usualL
norm. The generalization ofthisdefinitionfor higher differencesAhlf(z
of orderk instep h takestheformll/khl](z)ll2 O(lh[ ()
0<
a_<
k+
1.(3)
DEFINITIOM2.2([2],p.3). If
.f(z)
is acomplex-valuedfunctiononD,
its Fouriertransformis definedby, b) =/ ()-’/l)I"b)d ()
for all
C,
bB
forwhichthis integralexists. The Parseval’s identityinthis case iswrittenas/o If(z)l 2dz
1JoB ]fl 2A
tanh dAdb. (5)3. MAIN
THEOREMS
Ourmainresultmaybe viewed as the non-Euclideananalogueofthefollowing theorem.
THEOREM3.1([5],Theorem85). Let
f(z)
belongtoL2(R).
Then the conditionsIll(=
/h) f(=)l12 o(h ) ()
as h--,O, 0
<
c,<
1and+ ]}12dz O(X-2(’) (7)
as
X
ooareequivalent. Themaintheoremofthis section isstated asTHEOREM3.2. Let
f(z)
beacomplex-valuedfunctionontheunit discD. Thenthe conditionsIIf(z + h) f(z)ll2 O(]hla) (8)
aslh
--,0,O<a<
landas ovareequivalent.
oo ilddb O("x-2-’) (9)
PROOF.
By
definitionoff (A, b)
itiseasilyseenthat the transform of(f(z + h) f (z))
isgiven(e
(-iA+l)(-h’b)1) f(A, b) (0)
by
observe that
h,b) (h,b) d(O,h)
rand hence r--
0 with[hi
whereIhl
tanhr. So, thefactore{h,b) e tendsto oneasrgoestozeroandthereforecanbesuppressedwithoutharm. Also, since E
R
+ inthe right side of the Parseval’s identity([7],p. 376), (10)could be simplified sothatthe transform ofIf(z
/h) .f(z)l
isgivenas12
sin]’1.
Thus2sin-- f
tanh dAdbO(tanhr)
2aO(r2a).
Observethat
tanh(-)
tendstobeaboundedconstantforlarge A. Thisyields(as
in theproof ofTheorem3.1)
Z-Z, .l,aeae o( -o) o(-o),
equivalently
Fromthisoneobtains
ITl2dAdb + +... +
(11)
(12)
Itisclear that the integral db Iwhenthe normalizedHaarmeasureisin action. The lastquantityis bounded as
, --
oo if1-a/-/ <
0, so that</3 <_
2, givingthe condition a>
0for/9
1.Thisshows that theFouriertransform
f(,, b)
off(z)
convergesabsolutely foranyc,greater than zero.This reflects the strength of the conclusion in Theorem 3.2 in contrast to those ofsimilartheorems provedfor Lipschitz functionsin
R, R
andT
forexample. Thisis mainlydue tothepresenceof the weightA
inthe inversionformula aswell as in one sideofParseval’s identity. Hadit notbeenforthis extrafactorinthe present situation, the conclusion of Theorem3.2would have been exactlythe same as that of Theorem3.1 asfar asthe order of magnitude and the absoluteconvergenceoff
areconcerned.Weshallencounterthe same situation whendealingwiththe spherical
Furier
transforms ofLipschitz functionsinH
2whichisthe subjectmatterof thenextsection.4. SPHERICALTRANSFORbISIN
H
2In this section we prove an analogue of Theorem 3.2 for the spherical Fourier transforms of Lipschitz functions. These are related to spherical functions which (by theirvery
nature)
are radial eigenfunctions of the Laplacianontheunit disc(see [2],
pp.38,39fordetails). Theyaredefinedbyx(z) f el-’x+l><z’b>db. (13)
Iff(z)
is aradialcomplex-valuedfunctioninD,
thenitsspherical-Fourier transformisgiven byJD (14)
with this inhand,we statethe following
THEOREM4.1. Let
f(z)
beradial inD. Thenthe conclusion of Theorem 3.2 holdswiththe integralas
hl "
0and theproofiscomplete.REMARK
3.3. Weindicateherethatifthe k-th differenceAhlf(z
isemployedinDefinition2.2of the Lipschitz condition, thenthiswould resultintheappearanceof the factor
[rAI
2kinplace of[rAI
in theestimatesgivenby(10), (11),butthis will notaffect thefinal conclusionof Theorem 3.2. This is due tothewayin which rand aretiedupin theirvariation.REMARK
3.4. Applying HOlder’s inequalityto(12) for/ff <
2weget asAtendstoinfinity,thisprovesthefirst part of the theorem.Ontheotherhand,given
(12),
then by following the reverse argumentasintheproofof Theorem 3.1 onecanarriveeasilyatthe estimateIIf(z + h)-/(z)ll O(r ) O(tanhr 9-0)
O(Ihl )
400 M.S. YOUNIS
beingreplacedwith
PROOF. Theproofwill carried briefly;wemainlypointoutthenecessarymodifications which will suitthe presentsituation. Oneessentialaspect here is thepresenceof theHarish-Chandra c-function bothintheinversionformulaaswellasintheParseval’s identity. Thus one has
c(a)=r r (a+), c
whereasforA6
R
t
TheParseval’s idemity readsin thiscase
If(z)ldz I]’1 IC(A)
dA. (15)Thetransform of
f (z + h)
isfo.f(z + h)_x(z)dz fo f(w) [f e(-iX+l){u’-h’b>db]
dwe(-iA+l)(h.b)
T()"
Tng
imoaoum e or
ofe(h,b) e rtendstorodttIC()l -= o()
forl$eA,
eprfof
eorem
3.2 could be appfi(moword forword)to@eld
ther
reslt ofe
theory.
Weretookat
s pot tt
thedeee
ofsim
bnthe 1o r
isnott
mdsg
ew
ofthe fte te
overthed B
ofe t
didoesnotplayysifit
rolepnen
to the ord ofmde
d inthe ablute nvergen of boe
FoYer
dthesphtrsfo ofons
inH 2. Ts
imeg cody suppress out
dgetothe coseof theproof. Thus asf ase
problofe
orderofmde
de
solute nvg e
nm e or
off(A, b)
df(A)
isost
the e4.2. Wewouldfike to
t
outt e prous
ysis appfiesost exay
tothe sphec sfo on theLobaschowsspacesociat the
complex goupSL(2.C)
(s [7], pp.40002). r
tridenthe11 beve
brief.In
ts
cethesphtrsfoisdetobeI () I(t) itei’dt
so that thetransformof
f (t + h)
isf (x) sinh(x h)eU’(-h)
dx.Ash 0
sinh(x h) O(sinhx)
and the transform off(t + h) f(t)
isequalto(e
-’vh1)]’Iv
asusualand thisleadstothefinal estimate
L ITl,2d, o(, -,)
TRANSFORMSOF LIPSCHITZ FUNCTIONS 401 orequivalently
as
,
goestoinfinity. The otherpart of the proofisvery clear bynow. Moreovertherangeof for whichf [f[d,
isboundedisbetter than that foundinthe case the spherical transforms forfunctions
on
H2.
Inthiscase HOlder’s inequalityfor/_<
2leadstowhich inturngives
2 <_<2 for the boundedness of the last quantityasv oo.
.
CONCLUDING REMARKSItwould beconvenient toend with afew comments-ofarather heuristicnature-whichmightcast somefightonthetreatmentof the problemonotherdomains. Ourfirsttargetin thisrespectis thegroup
SL(2.R)
denoted here byGfor brevity. Thus fora functionf(g)gEGasmall translation suitablefor the Lipschitzconditions inthiscaseisgiven byf(gh),h G, Ih
O. Thusf
belongstoLip(a,2)
onG
ifIlf(gh) f(g)ll2 O(Ihl)
0
<
a<
1,]hi
--,0. With this inhand,one canapplyParseval’s identity(see [4],
p. 346,[6],
Vol. 2, p.53)inordertoobtain the required estimates for theFourier(spherical) transformsf
correspondingto theprincipal and discrete seriesrepresentationsofG. The occurrenceof thetwointegralsandtheone summation onthe rightsideof the identity causesnoproblembecauseonthe onehand cothAtendsto onefor largeA,
thus thetwointegralsaretreated in the same manner. Onthe other hand the three parts aremajorizedby thesamequantityO(Ihl);
thisenables ustodealwitheach part onits own. Thiswould be easier; besidesit will notaffect the final conclusions.Secondly, we hintthat the previous analysis is amenable to treatment ofthe problem on other classicalgroupsandtheir alliedsymmetric andhomogeneousspaces(see [1],ChapterXfor examples of thosegroups). Itgoeswithoutsayingthat therewould besome modificationsinthemaincourseof the proof when handlingconcretesituations suchasSL(2,C) for example. Thirdly,onecould explore the validity of the presentlineofthoughtsonsemi-simpleLiegroupsingeneral. Acluetoinvestigatingthis problemlies in anestimate forthe Harish-Chandrac-functionnear infinityinthat case(see
[2],
p.450and [3], p.183).
These points willbetakenupin aforthcomingpaper.REFERENCES
[1] HELGASON, S., Differential Geometry,
LieGroups,
and SymmetricSpaces,
AcademicPress,
New York, 1978.[2] HELGASON, S.,
Groups
and GeometricAnalysis,AcademicPress,New York,1984.[3] GANGOLLI,R. andVARADARAJAN, V.S.,HarmonicAnalysts
of
SphericalFunctions on Real ReducttveGroups,
SpringerVerlag New York,1988.[4]
SUG1URA,M.,
UnitaryRepresentations andHarmonicAnalysis, Wiley,NewYork, 1975.[5] TITCHMARSH,E.C.,Theory
of
FourierIntegrals,Oxford UniversityPress,1948.[6]
WARNER,
G., HarmonicAnalysisonSemi-SimpleLieGroups,
Vol. 2, Springer Verlag, New York,1972.[7] WAWRZYNCZYK,