ebook img

The Abel, Fourier and Radon Transforms on Symmetric Spaces. PDF

21 Pages·2006·0.33 MB·English
by  
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview The Abel, Fourier and Radon Transforms on Symmetric Spaces.

Indag.Mathem.,N.S.,16(3–4),531–551 December19,2005 TheAbel,FourierandRadontransformsonsymmetricspaces bySigurdurHelgason 77MassachusettsAvenue,Cambridge,MA02139,USA DedicatedtoGerritvanDijkontheoccasionofhis65thbirthday CommunicatedbyProf.J.J.DuistermaatatthemeetingofJune20,2005 1. INTRODUCTION In this paper we prove some recent results on the three transforms in the title and discusstheirrelationshipstoolderresults.Thespaceswedealwitharesymmetric spaces X = G/K of the noncompact type, G being a connected noncompact semisimpleLiegroupwithfinitecenterandK amaximalcompactsubgroup. For the two natural Radon transforms on X we prove a new inversion formula andasharpeningofanoldsupporttheorem;fortheAbeltransformweprovesome new identities with some applications and for the Fourier transform a result for integrable functions which has a strong analog of the Riemann–Lebesgue lemma. TheselatterresultsarefromacollaborationwithRawat,SenguptaandSitaram. Notation. Following Schwartz we use the notation D(X) for C∞(X), E(X) for c C∞(X)andS(Rn)forthespaceofrapidlydecreasingfunctionsonRn. 2. DIFFERENT RADON TRANSFORMS ON THE SYMMETRIC SPACE X Radon’s paper [40] suggested the general problem of determining a function on a manifold on the basis of its integrals over certain submanifolds. A natural case of E-mail:[email protected](S.Helgason). 531 thisproblemisthe inversion oftheX-raytransformonaRiemannian manifold.It isthetransformf →fˆdefinedbythearc-lengthintegral (cid:1) (2.1) fˆ(γ)= f(x)dm(x), γ f beingan“arbitrary”functionontheRiemannianmanifoldXandγ anycomplete geodesicinX. In general this injectivity problem seems to be unresolved. For a Cartan sym- metric space X (cid:3)=Sn the injectivity, however, holds. For a symmetric X of the noncompacttheinjectivityholdsinthestrongerformofthe Supporttheorem [26]. Iffˆ(γ)=0forallgeodesicsγ disjointfromaballB⊂X thenf(x)=0forx∈/B. Thislastresultrequiresstrongerdecayassumptionat∞thantheinjectivityresult does. Here we shall prove an explicit inversion formula for the X-ray transform for rankX>1.SeeSection5forthecontactwithRouvière’sdifferentsolution. Funk [11] and Radon [40] inverted this transform for the sphere S2 and R2. Denotingthesetofgeodesicsby(cid:2)wehavethecosetspacerepresentations S2=O(3)/O(2), (cid:2)=O(3)/O(2)Z2, R2=M(2)/O(2), (cid:2)=M(2)/M(1)Z , 2 M(n)denotingtheisometrygroupofRn. Thissuggeststhefollowinggeneralization.LetX=G/K and(cid:2)=G/H becoset spaces of the same locally compact group G, K and H being closed subgroups. Here it will be convenient to assume all these groups as well as L=K∩H to be unimodular. We do not assume the elements ξ ∈(cid:2) to be subsets of X but instead useChern’sconceptofincidence: x=gK isincidenttoξ =γH ifgK∩γH (cid:3)=∅assubsetsofG.Givenx∈X,ξ ∈(cid:2)define xˇ={ξ ∈(cid:2): x,ξ incident}, ξˆ ={x∈X: x,ξ incident}. Theseareorbitsofcertainsubgroupsof Gandhavenaturalmeasuresdm,dµ(up tofactors)andwedefinetheabstractRadontransformf →fˆanditsdual ϕ→ϕˇ by (cid:1) (cid:1) (2.2) fˆ(ξ)= f(x)dm(x), ϕˇ(x)= ϕ(ξ)dµ(ξ). ξˆ xˇ 532 Thenormalizationsofdmanddµareunifiedbytakingx =eK,ξ =eH and 0 0 (cid:1) (cid:1) (2.3) fˆ(γH)= f(γh·x )dh , ϕˇ(gK)= ϕ(gk·ξ )dk 0 L 0 L H/L K/L theinvariantmeasuresdh ,dk beingfixedbyHaarmeasuresofH,K andL. L L Mainproblems. (i) Injectivityoff →fˆ,ϕ→ϕˇ. (ii) Inversionformulas. (iii) Rangeandkernelquestionforthesetransforms. (iv) Applicationselsewhere. An easy general result relevant to problem (iii) is the following. For a suitable normalizationofthemeasuresdx=dg ,dξ =dγ wehave K H (cid:1) (cid:1) (2.4) f(x)ϕˇ(x)dx= fˆ(ξ)ϕ(ξ)dξ, X (cid:2) aresultwhichsuggeststheextensionof(2.3)todistributions. 3. d-PLANES IN Rn Here we consider the space X =Rn and (cid:2)=G(d,n) the set of d-dimensional planesinRn.ThesearebothhomogeneousunderthegroupG=M(n).Fixx ∈Rn, 0 ξ ∈G(d,n)atdistanced(x ,ξ )=p.Thenwehave 0 0 0 (3.1) X=Rn=M(n)/K , (cid:2)=G(d,n)=M(n)/H , p p whereK andH ,respectively,arethestabilitygroupsofx andξ .Sincevarious p p 0 0 p will be considered the transforms (2.2) will be denoted by fˆ and ϕˇ . Since p p the action of M(n) on X and (cid:2) is quite rich it turns out that for the coset space representation(3.1) z∈Xisincidenttoη∈(cid:2) ⇔ d(z,η)=p. Thusthetransformfˆ andϕˇ canbewritten p p (cid:1) (cid:1) (3.2) fˆ (ξ)= f(x)dm(x), ϕˇ (x)= ϕ(ξ)dµ(ξ). p p d(x,ξ)=p d(x,ξ)=p 533 ˆ ˆ Inparticular, f istheusuald-planetransformf,butinordertoinvertitweneed 0 ϕˇ forvariablep.Oneofseveralversionsoftheinversionformulaisthefollowing p (see[25,27]): (cid:2)(cid:3) (cid:4) (cid:1)∞ (cid:7) (3.3) f(x)=c(d) d d p(cid:5)p2−r2(cid:6)d/2−1(cid:5)fˆ(cid:6)∨(x)dp d(r2) p r r=0 ˆ ∨ withc(d)andconstant.Notethat(f) (x)istheaverageoftheintegralsoff over p alld-planesatdistancepfromx. Ford=1thisformulareducesto (cid:1)∞ (cid:5)(cid:5) (cid:6) (cid:6) (3.4) f(x)=−1 d fˆ ∨(x) dp, π dp p p 0 which for n=2 coincides with Radon’s original formula. Radon’s proof is very elegant and is based on an exhaustion of the exterior |x|>r by lines. As far as I know this proof has not been extended to higher dimensions n. Formula (3.4) for n>2iscrucialfortheinversionof(2.1)giveninTheorem5.1. 4. d-DIMENSIONAL TOTALLY GEODESIC SUBMANIFOLDS IN HYPERBOLIC SPACE Hn Asimilarmethodworkshereandtheanalogof(3.3)istheformula (cid:2) (cid:7) (cid:3) (cid:4) (cid:1)∞ (4.1) f(x)=C(d) d d (cid:5)t2−r2(cid:6)d/2−1td(cid:5)fˆ(cid:6)∨ (x)dt , d(r2) s(t) r r=1 whereC(d)isaconstantands(p)=cosh−1(p)(see[25,27]).Otherversionsofthe inversionexist(e.g.,[28]and[6]).Ford=1thisreducesto (cid:1)∞ (cid:5)(cid:5) (cid:6) (cid:6) (4.2) f(x)=−1 d fˆ ∨(x) dp π dp p sinhp 0 aformulawhichforn=2isstatedwithoutproofinRadon[40]. 5. X-RAY INVERSION ON THE SYMMETRIC SPACE X=G/K In communication from 2003, Rouvière proved an extension of formula (4.2) to symmetricspacesX ofrank(cid:7)=1.Inspiredbyhismethods,Iprovedtheinversion formula (5.2) for the X-ray transform for X of rank (cid:7)>1. Then Rouvière [43] extendedhisformulatoXofarbitraryrank(cid:7).Actuallyhehasseveralsuchformulas buttheyarealldifferentfromtheformula(5.2)below. Fix a flat totally geodesic submanifold E of X with dimE = (cid:7) > 1 ((cid:7)therankofX)passingthroughtheorigino=eK ofX.Letp>0andS=S (o) p be the sphere in E with radius p and center o. The geodesics γ in E tangent to 534 S are permuted transitively by the orthogonal group O(E). Let du and dk denote thenormalizedHaarmeasuresonU andK.Thespacesk·E ask runsthroughK constitute all flat totally geodesic subspaces of X through o of dimension (cid:7). Thus theimagesk·γ (k∈K,γ tangenttoS)constitutetheset(cid:8) ofallgeodesicsγ inX p lyingin someflat (cid:7)-dimensionaltotally geodesic submanifoldof X through o and d(o,γ)=p.Theset(cid:8) hasanaturalmeasureω givenbythefunctional p p (cid:8) (cid:9) (cid:1) (cid:1) (cid:5) (cid:6) (5.1) ω :ϕ→ ϕ k(u·γ) du dk. p K O(E) Theorem5.1. TheX-raytransform(2.1)onasymmetricspaceX=G/K ofrank (cid:7)>1isinvertedbytheformula (cid:8) (cid:9) (cid:1)∞ (cid:1) (5.2) f(o)=−1 d fˆ(γ)dω (γ) dp, f ∈D(X). p π dp p 0 (cid:8)p Since(cid:8) anddω areK-invarianttheformulaholdsateachpointx byreplacing p p f byf ◦g whereg∈Gissuchthatg·o=x. Proof. First assume f to be K-invariant and consider the restriction f|E. Fix an orthonormal frame H ,H ∈E , the tangent space to E at o, consider the one 0 0 parameter subgroups exptH , exptH and the geodesic γ (t)=exptH ·o. Then 0 0 0 the geodesic γ(t)=exppH ·γ (t) lies in E and is tangent to S (o). Because of o p (3.4)wehave (cid:1)∞ (cid:5) (cid:6) (5.3) f(o)=−1 d fˆ E(o)dp, π dp p p 0 wherethesuperscriptE standsforthedualtransformongeodesicsinthespaceE. Thus (cid:1) (cid:1) (cid:5) (cid:6) (cid:5) (cid:6) (cid:5) (cid:6) (5.4) fˆ E(o)= fˆ (γ)dν(γ)= fˆ (u·γ)du, p γ⊂E O(E) d(o,γ)=p whereν standsfortheaverageoverthesetofgeodesicstangenttoS (o). p Forf ∈D(X)arbitraryweuse(5.2)onthefunction (cid:1) f(cid:11)(x)= f(k·x)dk. K Takingintoaccountthedefinition(5.1)theinversionformula(5.2)followsimme- diately. (cid:1) 535 Remark. Note that the measure ω in (5.1) is a kind of convolution of the Haar p measuresdk and du.Howeveritisnotastrictconvolutionsincetheproduct ku is notdefined. 6. THE HOROCYCLE TRANSFORM IN X=G/K Consider the usual Iwasawa decomposition of G, G=NAK where N and A are nilpotent and abelian, respectively. A horocycle is by definition [12] an orbit in X ofaconjugategNg−1 ofN.ThegroupGpermutesthehorocyclestransitivelyand thespace(cid:2)ofhorocyclescanbewritten(cid:2)=G/MN whereM isthecentralizerof AinK.Inthedoublefibration G/M (cid:1)(cid:1) (cid:2) (cid:1) (cid:2)(cid:2) X=G/K G/MN =(cid:2) itturnsoutthatx=gK isincidenttoξ =γMN ifandonlyifx∈ξ.Thetransforms (2.2)become (cid:1) (cid:1) (6.1) fˆ(γMN)= f(γn·o)dn, ϕˇ(gK)= ϕ(gk·ξ )dk, 0 N K whereξ =N ·o.Whilethemap ϕ→ϕˇ hasabigkernel,thehorocycletransform o f →fˆisinjective(see[17]or[13]).Thefollowingresultfrom[21]isconsiderably stronger. Theorem6.1 (Supporttheorem). LetB beaclosedballinX.Then fˆ(ξ)=0 forξ ∩B=∅ implies f(x)=0 forx∈/B. Hereonerequiresstrongerdecayconditionsonf thanfortheinjectivity.Adif- ferentproofwasgivenin[14].WehavealsothefollowinginversionandPlancherel formula for the Radon transform [18,19]. The pseudodifferential operator (cid:12) and thedifferentialoperator (cid:1) belowareconstructedbymeansoftheHarish-Chandra c-function, and w denotes the order of the Weyl group. For G complex a result similarto(6.2)appearsin[12]. Theorem 6.2. For f ∈ D(X) or sufficiently rapidly decreasing we have the inversionformula (cid:5) (cid:6) (6.2) f = 1 (cid:12)(cid:12)¯fˆ ∨. w 536 IfallCartansubgroupsofGareconjugatetheformulahastheimprovedversion (cid:5)(cid:5) (cid:6) (cid:6) f = 1(cid:1) fˆ ∨ . w ForGarbitrary (cid:1) (cid:1) (cid:10) (cid:10) (cid:10) (cid:10) (6.3) w (cid:10)f(x)(cid:10)2dx= (cid:10)(cid:12)fˆ(cid:10)2(ξ)dξ, X (cid:2) withasuitablenormalizationoftheinvariantmeasuresdx anddξ. The range question (iii) for f → fˆ is more complicated. Consider first the hyperbolic plane H2 in the Poincaré unit disk model D. Here the horocycles are the circles in the disk tangential to the boundary {eiθ: θ ∈R}. Let ξ denote the t,θ horocyclethrougheiθ withdistancet (withsign)fromtheorigin.Thenwehavethe followingresultfrom[23]. Theorem6.3. TherangeD(D)(cid:11)consistsofthefunctionsψ∈D((cid:2)) (cid:12) ψ(ξ )= ψ (t)einθ t,θ n n where (cid:3) (cid:4) (cid:3) (cid:4) d d (6.4) ψ (t)=e−t −1 ··· −2|n|+1 ϕ (t) n n dt dt whereϕ ∈D(D)iseven. n This implies a relationship between ψ(ξt,θ) and ψ(ξ−t,θ). More specifically, if f(cid:11)(t)=f(−t),(cid:15) =etψ then∗denotingconvolutiononR n n (cid:15)(cid:11) =S ∗(cid:15) n n n wherethedistributionS onRhasFouriertransform n Sˆ = (iλ+1)···(iλ+2|n|−1), λ∈R. n (iλ−1)···(iλ−2|n|+1) Thisrelationshipbetweenψ(ξ−t,θ)andψ(ξt,θ)impliesthatin(6.3)f →(cid:12)fˆdoes notmapL2(X)ontoL2((cid:2)). For the generalization of (6.4) to X=G/K we need some additional notation. Let Kˆ betheunitarydualof K and d(δ) thedegreeofa δ∈Kˆ.Given δ actingon V let δ (cid:13) (cid:14) VM = v∈V : δ(m)v=vform∈M δ δ 537 andput(cid:7)(δ)=dimVM.Let δ (cid:13) (cid:14) Kˆ = δ∈Kˆ: (cid:7)(δ)>0 . M Inthefollowingtheoremfrom[25]theexpansion(6.5)isageneralizationof(6.4). Putρ(H)= 1Trace(adH|n). 2 Theorem6.4. TherangeD(X)(cid:11)consistsofthefunctionsψ∈D((cid:2)) (cid:12) (cid:5) (cid:6) (6.5) ψ(ka·ξ )= d(δ)Tr δ(k)(cid:15) (a) 0 δ δ∈KˆM (Tr = Trace) where (cid:15) is a function on A with values in Hom(V ,VM), i.e., δ δ δ (cid:15) ∈D(A,Hom(V ,VM)),givenby δ δ δ (cid:5) (cid:6) (6.6) (cid:15) (a)=e−ρ(loga)Qδ(D) (cid:19) (a) , a∈A, δ a δ where (cid:5) (cid:5) (cid:6)(cid:6) (6.7) (cid:19) ∈D A,Hom V ,VM δ δ δ is W-invariant and Qδ(D) is a certain (cid:7)(δ)×(cid:7)(δ) matrix of constant coefficient differentialoperatorsonA. From this result we can derive the following (unpublished) refinement of the + support theorem above. Let A be the Weyl chamber corresponding to the choice ofthegroupN. Theorem6.5. Supposef ∈D(X)satisfies fˆ(ka·ξ )=0 fork∈K, a∈A+, |loga|>R. 0 Then fˆ(ka·ξ )=0 fork∈K, |loga|>R, a∈A 0 sobyTheorem6.1 f(x)=0 ford(0,x)>R. Proof. LetQ (D)bethematrixofcofactorsofQδ(D)sothat c (6.8) Q (D)Qδ(D)=detQδ(D)I. c Then(6.6)implies (6.9) Q (D)(eρ(cid:15) )=detQδ(D)(cid:19) . c δ δ 538 Nowitisknown[34],[25,pp.267,348]thatdetQδ(D)isaproductoflinearfactors δ(H )+cwhereH ∈aand∂(H )thecorrespondingdirectionalderivative. i i i Supposethefunctionψ=fˆsatisfies ψ(ka·ξ )=0 fork∈K, a∈A+, |loga|>R. 0 Since (cid:1) (cid:15) (a)= ψ(ka·ξ )δ(k−1)dk δ 0 K wededucefrom(6.8)and(6.9)that detQδ(D)(cid:19) (a)=0 fora∈A+, |loga|>R. δ + Consider this equation on a ray in A starting at e. Because of the mentioned factorization of detQδ(D) we deduce that on this ray (cid:19) satisfies an ordinary δ differential equation on the interval (R,∞). Having compact support we deduce that(cid:19) (a)=0fora∈A+,|loga|>R.ByitsWeylgroupinvarianceitvanishesfor δ alla∈A,|loga|>R whichby(6.5)provesthetheorem. (cid:1) ConsiderthecaserankX=1.LetB (o)beaballinXwithradiusR andcenter R 0.FixaunitvectorH intheLiealgebraofAsuchthatexpH ∈A+.Puta =exptH. (cid:15) t TheinteriorofthehorocyclekNa ·oistheunion kNa ·o.Ahorocycleξ is t τ>t τ saidtobeexternaltoB (o)ifitsinteriorisdisjointfromB (o);ξ issaidtoenclose R R B (o)ifitsinteriorcontainsB (o). R R Corollary6.6. LetXhaverankoneandB (o)asabove.Letf ∈D(X).Thenthe R followingareequivalent: (i) fˆ(ξ)=0wheneverξ isexternaltoB (o). R (ii) fˆ(ξ)=0wheneverξ enclosesB (o). R (iii) f ≡0outsideB (o). R ForhyperbolicspacethisisclearfromTheorem6.3andwasprovedinadifferent waybyLaxandPhillips[35]. Problem(iii)forthedualtransformϕ→ϕˇ hasasatisfactoryanswer(see[25,IV §§2 and 4]). The kernel can be described in the spirit of Theorem 6.5 and for the rangeonehasthesurjectivity (6.10) E((cid:2))∨=E(X). 539 7. THE ABEL TRANSFORM LetD (X)denotethespaceofK-invariantfunctionsinD(X).TheAbeltransform K f →Af isdefinedby (cid:1) (7.1) (Af)(a)=eρ(loga) f(an·o)dn, a∈A, f ∈D (X). K N Except for the factor eρ it is the restriction of the Radon transform to K-invariant functions (7.2) Af =eρfˆ. Someofitspropertiesarebestanalyzedbymeansofthesphericalfunctions (cid:1) (7.3) ϕ (g)= e(iλ−ρ)(H(gk))dk, g∈G, λ∈a∗, λ c K whereH(g)∈aisdeterminedbyg∈kexpH(g)N anda∗ isthecomplexdualofa. c Thesphericaltransform (cid:1) (7.4) f˜(λ)= f(x)ϕ−λ(x)dx, f ∈DK(X) X (whereϕ (g·o)=ϕ (g))isahomomorphismrelativetoconvolution×onX: λ λ (7.5) (f ×f )∼(λ)=f˜(λ)f˜(λ). 1 2 1 2 Asprovedin[15],AintertwinesthesphericaltransformandtheEuclideanFourier transformF →F∗ onAso (cid:1) (cid:1) (7.6) (Af)(a)e−iλ(loga)da= ϕ−λ(x)f(x)dx, (Af)∗=f˜. A X ThusAf isW-invariantandby(7.5) (7.7) A(f ×f )=Af ∗Af , 1 2 1 2 where∗isconvolutiononA.LetD(X)denotethealgebraofG-invariantdifferential operators on X and (cid:8):D(X) → D (A) the isomorphism onto the W-invariant W constantcoefficientdifferentialoperatorsonA. TheAbeltransformisasimultaneoustransmutationoperatorbetweenD(X)and D (A),i.e., W (7.8) ADf =(cid:8)(D)Af, D∈D(X), f ∈D (X) K as shown in [29] which for example can be used to prove that each D has a fundamental solution. By the Paley–Wiener theorem for (7.3) one has that 540

Description:
Indag. Mathem., N.S., 16 (3–4), 531–551. December 19, 2005. The Abel, Fourier and Radon transforms on symmetric spaces by Sigurdur Helgason.
See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.