Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

Rigidity of Holomorphic Mappings and a New Schwarz Lemma at the Boundary

Author(s): Daniel M. Burns and Steven G. Krantz


Source: Journal of the American Mathematical Society, Vol. 7, No. 3 (Jul., 1994), pp. 661-676
Published by: American Mathematical Society
Stable URL: http://www.jstor.org/stable/2152787 .
Accessed: 12/06/2014 23:33

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Mathematical Society is collaborating with JSTOR to digitize, preserve and extend access to Journal
of the American Mathematical Society.

http://www.jstor.org

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
JOURNALOF THE
AMERICANMATHEMATICALSOCIETY
Volume7, Number3, July1994

RIGIDITY OF HOLOMORPHIC MAPPINGS


AND NEW SCHWARZ LEMMA AT THE BOUNDARY
A

DANIEL M. BURNS AND STEVEN G. KRANTZ

1. INTRODUCTION

The uniquenessportionof theclassicalSchwarzlemmahas playedan inter-


estinghistoricalrole. Usuallyfirstencountered of thecon-
in theclassification
formalself-maps ofthedisc,itarisesin moregeneralconsiderations ofautomor-
phismgroups,in theconstruction of theCaratheodory and Kobayashi/Royden
metrics, and in a varietyofcontextsin analysison manifolds.Usefulreferences
forthismaterialare [AHL, KR1, KOB, GKl, GK2, YAU].
Generalizations ofthediscSchwarzlemmato multiply-connected domainsin
onecomplexvariablearegenerated naturallyandeasilyusingtheuniformization
theorem.The generalization to severalcomplexvariablesrequiresinsightsof
a differentnature.For example,theso-calledCaratheodory-Cartan-Kaup-Wu
theorem[WU] saysin partthatif Q is a boundeddomainin C' and if '1 is
a holomorphic self-mapof Q thatfixesa point P E Q, thentheholomorphic
Jacobiandeterminant det JacF(P) has modulusless thanor equal to 1, and
equals 1 if and onlyif '1 is a biholomorphism of Q. An interesting and
importantfeatureof thisresultis thatit has a globalhypothesis(that Q be
mappedto itself)and a local hypothesis (theconditionon thebehaviorof the
mappingat P). The conclusionis thena strongglobalone. We will see this
paradigmrepeatedin theworkthatfollows.
The purposeof thispaperis to seekversionsof thelast statedresultwhen
thepoint P lies in theboundaryof Q. (A primitive versionof sucha result
appearsin (KR2].) Thisproblem,whileofconsiderable intrinsic
interest,is also
relatedto a varietyof otherworkin the literature.We now indicatesome of
theseconnections.
Our interestin this problemarose originallyfroma questionof Warren
Wogen: Does thereexist a holomorphicself-mapping1Dof the ball B in
C2 (i) whichtakesthe point 1 = (1, 0) E dB to itself(in a suitablesense)
and (ii) forwhichtheset a (1(B)) has "highorderof contact",as a set,with
the boundaryOB of the targetball? Here part of the problemis to finda
suitableinterpretation forthe phrase"highorderof contact". The mappings
ReceivedbytheeditorsMay 3, 1993.
1991Mathematics Primary32H15, 32H20,32H99.
SubjectClassification.
Key wordsandphrases.Holomorphic Schwarzlemma.
mapping,boundaryrigidity,
in partbygrant#DMS9004149fromtheNationalScienceFoun-
authorwas supported
The first
dation. The secondauthorwas supportedin partby grants#DMS8800523and #DMS9101104
fromtheNationalScienceFoundation.
i 1994American MathematicalSociety
0894-0347/94$1.00+ $.25 perpage

661

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
662 D. M. BURNS AND S. G. KRANTZ

1D8(Z1I Z2) = ((1 - e)Z1 + I, (1 - e)Z2) e > 0, satisfy(i) and (ii) withthe
boundaryof theimageball havingorderof contact1 withtheboundaryof the
targetball (thetangentplanesagreebutthesecondfundamental formsdo not).
The mapping(Do is theidentityand theorderof contactis infinite.Wogen's
questionmaybe interpreted as askingwhether thereare mapswhichare inter-
mediateto the (D, e > 0, and id = (o. The originalinterestin constructing
suchmapswas in finding counterexamples to certainassertionsaboutcomposi-
tionoperators(see [WOG]).
As ourunderstanding oftheproblemdeveloped,we also sawthatitrelatesto a
resultofH. Alexander[ALE]: If UnB and U'nB areboundaryneighborhoods
in theball and if (D is a biholomorphic mappingof theseneighborhoods which
extendsC2 to the boundary,then (D mustbe the restriction to U n B of a
biholomorphism of the entireball. S. Pincuk[PIN] generalizedthisresultto
boundedstrictly pseudoconvexdomainswithreal analyticboundariesand W.
Rudin [RU1] reducedthehypothesis of C2 to theboundaryto an assumption
whichis even weakerthan continuity at a point. One interpretation of the
mainresultof thepresentpaperis thatthehypothesis of directcoincidenceof
boundaryneighborhoods in theresultsofAlexander, Pincuk,and Rudinmaybe
weakenedto highorderofcontactofimageboundaryand targetboundary.(See
also [GK3] forothermoregeneralversionsof theAlexanderphenomenon-in
particular, thatpapercharacterizes notjust thebiholomorphic mappingsof the
ball but mappingswhichare "approximately" biholomorphic mappingsof the
ball.)
The organization of thispaperis as follows:Section2 presentsa boundary
uniquenessresulton the disc. The relationship of thisresultto the questions
justdiscussedis notimmediately apparent, later.Section3 shows
butis clarified
how to extendtheresultof Section2 to theball in Cn. Section4 showshow
to use variantsof theFornmssImbeddingTheoremand theLemperttheoryof
extremaldiscs forthe Kobayashimetricto derivea resultforstrictly pseudo-
convexdomains.Section5 discussesa generalization oftheSchwarzuniqueness
theoremsoftheprevioussections.Section6 discussesgeometric interpretations
of themainresultsand returns to theoriginalquestionof orderof contact.The
questionof Wogenis thenrecalledand answered.Section7 considersonlyvery
briefly analogousproblemson weaklypseudoconvexdomains.

2. A RESULT ON THE DISC ABOUT BOUNDARY DERIVATIVES

Let D be theunitdisc centeredat 0 E C.


Theorem2.1. Let :D - D be a holomorphic fromthedisc to itself
function
suchthat
_
0(0) = I + (C - 1) + 0 ((C 1)4)
as I. Then =-(4-- on thedisc.
Remark1. Resultssuchas thisone appearin theliterature
of conformalmap-
pings(see, forinstance,[VEL]) withtheadditionalhypothesisthat 0 be univa-
lent(and oftenthefunctionis assumedto be quitesmooth-evenanalytic-in
a neighborhood of 1 ). The theorempresentedherehas no suchhypothesis;so

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
RIGIDITY OF HOLOMORPHIC MAPPINGS 663

faras we knowit is new. Evenmoresurprising


is thattheexponent4 is sharp:
simplegeometric arguments showthatthefunction

+(4)=4 llog -1)3

satisfiesthe hypotheses of the theoremwith 4 replacedby 3. Note also in


theproofthat 0((z - 1)4) can be replacedby o((z - 1)3) . A similarremark
appliesto Sections3, 4, and 5 below.
X. Huang,in thepapers[HU1, HU2, HU3], has exploredadditionalcondi-
tionsunderwhichtheexponentin theerrortermmaybe decreased.
Proofof Theorem2.1. Considertheholomorphic
function

_(4 l+q$b(4)

Then g mapsthedisc D to therighthalfplane. BytheHerglotzrepresentation


(thisis just an applicationof the Banach-Alaoglutheorem;see [AHL]), there
mustbe a positivemeasure,u on theinterval[0, 2Xr)and an imaginary
constant
& suchthat

(*) g(4) = + du(6) + W.


2|je1io
We use thehypothesis
on 0 to analyzethestructure
of g and hencethatof
,u. To wit,
I(4= C + o(C _ 1)4 =- + o(g _ 1)2.
Fromthisand equation (*) we easilyconcludethatthemeasureIu has the
form,u-= 0 + v, wheredo5is ( 2r times)theDirac massat theoriginand v is
anotherpositivemeasureon [0, 27). In fact,a nicewayto verify
thepositivity
of v is to use theequation
l+g + 1 {2neio +
(**) 1 g +? -_12
) =2 i_ d(go+ v)(0) + F

to derivea Fourier-Stieltjes + v and thento applytheHerglotz


expansionof do5
criterion[KAT, p. 38].
We maysimplify equation (**) to

o(4 - 1) eio +4dv(O) + W.

Pass to thereal partof thelast equation,thuseliminating


theconstantW.
Since v is a positivemeasure,we thussee thattherealpartof theintegralon
therightof thislastequationrepresentsa positiveharmonicfunctionh on the
disc thatsatisfies
2
h(g)= O(C- -1 )2
In particular,h takes a minimumat the point 4 = 1 and is O(jz - 11)
thereas well. This contradicts
Hopf'slemma[KR1] unless h 0. But h 0

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
664 D. M. BURNS AND S. G. KRANTZ

impliesthat v 0_ , hencethat

g(C) = +
Thereforeq(4) and thetheoremis proved. 0
Remark2. Noticethattheformulation
of Theorem2.1 is a boundaryunique-
nesstheoremin theveinoftheuniquenesspartof theclassicalSchwarzlemma.
questionsabout "orderof contact".
However,the resultsidestepsthe onrginal
In factthe RiemannMappingTheoremguaranteesthat,withany reasonable
definitionof thephrase,thereare univalentholomorphicmaps of the disc to
high(even infinite
the disc so thatthe imageboundaryhas arbitrarily order)
contactwiththe targetboundary. Thus Theorem 2.1 addressessome more
rigidstructuralphenomenon.
We deferour consideration questionto ?6.
of theoriginalgeometric
3. A RESULT FOR THE BALL

Theorem3.1. Let B c C' be theunitball. Let (1: B -+ B be a holomorphic


oftheballtoitself
mapping suchthat
D(Z) = I + (Z - 1) + O(z -_l14)
as z -+ 1. (Here 1 denotesthedistinguished point 1 = (1 , 0
boundary ... , 0)
oftheball.) Then+(z) _ z on theball.
Proof.There is no usefulHerglotzrepresentation
on the ball (however,see
[AIZ] forrelatedideas): this is a deep factwhichcannotbe circumvented.
Thus we presenta newargument thatreducestheball case to thedisc case. For
simplicity
we restrictattentionto dimensiontwo.
For each point a E B let Ya be thecomplexlinejoining a and 1. Let da
be thecomplexdiscgivenby YanB . Now forfixeda considertheholomorphic
function

Definealso the mappingOa : B-)B, whichis an automorphism of the ball


mappingdo onto da and fixing1. (That such maps existfollowsfromele-
mentarygeometric orsee [RU2] foran explicitformula.)Finally
considerations,
define
7cl: B B,
7r1 (Z1, Z2) (z1
(ZI 0)
and
1:do)D,
1: (Zl 0)i Z1 -
Thenthefunction
Ha -.DD

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
RIGIDITY OF HOLOMORPHIC MAPPINGS 665

is well defined.Moreover,it is straightforward


to checkthat H satisfiesthe
hypotheses of Theorem 2.1 (since 1Dagreeswiththe identityto highorder,
the composition (q$a)1 o (D 0? ,a is the identityto high order). It follows that
Ha(4)-C.
Now set
Ga = (0,,) ? (D ? ?, ? ty(C) =_(g,l(C) ga(:))
The statement
that Ha(C) meansthat ga(4) _ But

IgI(4C)2 + Ig2(C)12 <1


g2() 2
for 4 E D. LettingI4C-- 1 yieldsthenthat I0ga
(C) O. Thus ga _ 0. It now
followsthattheimageof Ga alreadylies in do. As a result,it mustbe that C)
preservesda . Butthiscan onlyholdforeverychoiceof a if D is theidentity
mapping.
The proofis now complete. 0

4. GENERALIZATION OF THE RESULT TO STRONGLY CONVEX


AND TO STRONGLY PSEUDOCONVEX DOMAINS

It is clear thatthe methodsof the last sectionwill not applyto strongly


pseudoconvexdomains.Indeed moststrongly pseudoconvexdomainshave no
automorphisms excepttheidentity (see [GKI]), so we are missinga majortool.
However,an inspectionoftheproofofTheorem3.1 showsthattheprincipal
geometricconstruct is of a familyof mappingsof the disc into the ball such
thattheimageanalyticdiscsareholomorphic retractsoftheentiredomain(this
is wheretheglobal natureof the result,alludedto in theintroduction, comes
in). Thanksto workof Lempert[LEM1], we knowa largefamilyof domains
forwhichsuchspecialanalyticdiscsexist:
Proposition 4.1 (Lempert).Let Q c C'( be a smoothly
bounded,strictly
convex
(in therealsense)domainwith ek-boundary(k > 6). Let Q E Q and P E (Q .
Thenthereexistuniqueholomorphic mappings
)
VPQ Q: D vp Q: Q D

whichare ?k-4 up to theboundary,


and suchthat:
(1) (0P Q(O) = Q, PpQ(l) = P;
(2) VpIQ o pp Q(4) = C, for every 4 E D;
(3) if (: D -Q is holomorphic, and if y1pQ o () C=g for C E D, then

Note that(2) impliesthat fp Q is extremalfortheinfinitesimal


Kobayashi
metric on Q in the direction := V
qpQ(0) E TQ(Q), while VIp,Q is the ex-
tremalforthe"dual" infinitesimal metricat 4.
Caratheodory
Proof.The statements
of thepropositionfollowfrom[LEM1, Theoreme3] and
[LEM2]. 0
GivenProposition4.1, theproofof Theorem3.1 givesthefollowing
result.

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
666 D. M. BURNS AND S. G. KRANTZ

Theorem4.2. Let Q bea smoothly


bounded, strictly
convex
domainin Cn with
g6 boundary, 9 If (: Q2 Q isa holomorphic
andletP E i2 befixed. mapping
suchthat
= Z + _ p14),
?(DZ) ?(jZ

then4?(z) _ z.
Proof.Indeed,themaps p, Q above replace Oa o V in theproofof Theorem
3.1, and the YpI Q play the role of 1o o?((a) 1 there. The uniqueness
argumentat the end of the proofof Theorem3.1 is hereachievedby (3) of
Proposition4.1. 0
Unfortunately, Lempert'sresultsfailfora generalstrictly
pseudoconvexdo-
main (see [SIB]), but theyremaintruein a neighborhood of P, in directions
approximately C-tangent to Oil at P. More precisely,
we have thefollowing.
4.3. Let Q be a smoothly
Proposition bounded,strongly
pseudoconvexdomain
in C' withW6 boundary, existsQ E Q,
andlet P E i2 befixed.Thenthere
9
V c Q of Q, andholomorphic
an openneighborhood mappings
fPpQI D41, VIpQI: LI )D
as inProposition
4.1 above,
foreveryQ' E V. Furthermore,givenanyneighbor-
hoodU c C' of P, onecanassumethatbothQ andtheimages(0p Q,(D) lie
in QnU.
disc (p Q existsforsome Q E U n Q is
Proof.Given U, thatthestationary
shownin [LEMI, p. 468]. The existenceof the open set V 3 Q, and of the
corresponding gpPQ,, Q' e V, followsfrom[LEMI, Proposition10, p. 446].
To getthedual functionsVIpQI , we needthefollowing
preparatory
geometrical
lemma.
Lemma4.4. Let Q and P be as in Proposition
4.3 above.Thenthereexista
Q' of C, a bounded,
neighborhood convexdomainE" c En, and a
strictly
map F :
holomorphic *' nC suchthat:
W of F(P), F is biholomorphic,
(1) fora suitableopenneighborhood
F :F 1(W) ,W;
(2) F (Q) C U' ;
(3) F(aQ) n W =n" n W.
Admitting thelemmaforthemoment,we returnto theproofof Proposition
4.3. We now choose the Q and V morecarefully, namelyso that V c U n
F 1 (W) n Q, where W, F are as in thelemma.ThenthemappingsF o p1 Q,
discs forthe strictly
are stationary convexdomain Q" C Cn. Thus thereexist
dual mappings
Q
Y1F(P),F(Q')
D

and themappingsV,p Qt= (p) F(Q)0


, F willhaveall thedesiredproperties
in
thestatement
of Proposition4.3, as followsdirectly
fromProposition4.1. o
UsingProposition4.3, theproofof Theorem4.2 extendsdirectly
to give:

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
RIGIDITY OF HOLOMORPHIC MAPPINGS 667

Theorem4.5. Let Q be a connected, smoothlybounded,strongly pseudoconvex


domain. Let P E 9El and let 'D: Q -- a be a holomorphic map such that
?(z) = z +O(jz - Pj4) . Then?(z) - z.
Proof. Find Q, V as in Proposition4.3 above. The proofas in Theorems3.1
and 4.2 showsthat F is the identityon each disc (p QQ,(D); in particular,
D(Q') = Q' forall Q' E V. By analyticcontinuation,F(z) -z. 0
We finishthissectionwiththeproofof Lemma4.4. We startwithFomrss's
embeddingtheorem[FOR], whichgives a neighborhoodQ' of LI, a holo-
N
morphicembeddingG: n' ?- (forsome large N), a smoothlybounded
strictlyconvexset Qi" c CN such that aO01 intersectsG(n') transversally,
and G(Q) = Q0 n G(Q'). Let T be thetangentspace to G(Q') at G(P); T
N
is isomorphicto C?, say E: T Cn linearly.Let Il1- CN- T be the
--

Hermitianorthogonalprojectiononto T (withrespectto the usual euclidean


metricon CN), so that II is C-linear.We claimthatthecomposition
F G:Q G ?N n T E) ?n
willbe the desiredmap, if we shrinkQ' about Q as necessary.To see this,
noticefirst thatproperty (1) of Lemma4.4 followsfromthestrictconvexity of
Q'i and the factthat G is an embedding.(We maybe requiredto shrinkQ'
here.) It is an easy exercisein the implicitfunctiontheoremto see that,in a
neighborhood of P, F is a local diffeomorphism convex
of an9i ontoa strictly
hypersurface in Cn. Since E o II is linear,thecompactset E o rI(f2g):='I' is
convexand F(Q) c S". It is again a simpleexercisein theimplicitfunction
theoremto see thatthe boundaryof l"' is smoothand strictly convexnear
F(P). Thus thereis a closed euclideanball B whichis tangentto a 2l" at
F(P) and whichcontains 27l.TranslateB a distanceE > 0 in thedirection
of theinwardeuclideannormalto a KY' at F(P), to a newball called BE. For
e sufficientlysmall,let He denotethe(real)convexhullof B8uF(f) in Cn, so
thattheboundaryof H. coincideswith F(lQ2) in a neighborhood of F(P).
The domain H, is convex,but not smoothand not strictlyconvex. We can
enlargeHe slightly to a smoothly bounded,strictlyconvexset n" c C" whose
boundarystillagreeswith F(Ol) in a neighborhood of F(P). That thislast
is possibleis clearfroma picture,butis elementary and tediousto writedown
explicitly,so we omitthisverification. The Q" just constructed satisfiesall the
claimsof Proposition4.4. o
We close thissectionwiththe remarkthatLemma 4.4 is an attemptto re-
embed Q2 in Cn as a strictly convexdomain,whichis certainlyimpossible.
The pointhereis thatF is globallydefined, and does re-embedQl as a strictly
convexdomain,at leastin a neighborhood of P.
5. ANOTHER DIRECT EXTENSION

For P #Q it is interesting
to considerthe settingof Theorem4.2 withthe
that
modifiedhypothesis
mD(z)
n
Q l fz7 - P) + Olz - PI)4

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
668 D. M. BURNS AND S. G. KRANTZ

here , is somelocal holomorphic transformation whichtakesthetangent plane


to ad at P to thetangentplane to ad at Q. Of coursewe cannothope to
concludeunderthesecircumstances that (D is the identity.To see whatthe
correctconclusionmightbe, considertwospecialcases:
(1) If Q is theball,then(D maybe composedwitha unitaryrotationtaking
Q to P. This reducesthesituationto theresultof ?3. We concludethat (D is
a rotation.
(2) If Q is strongly pseudoconvexand thereis a biholomorphism T of Q
whose extensionto the boundarytakes Q to P then z o (D satisfiesthe hy-
pothesesof Theorem4.5. We concludethat (D is a biholomorphism of Q.
A genericstrongly pseudoconvexQ has no nontrivial biholomorphisms (see
[GK1]); so thesecondcase abovecanbe considereddescriptive butnotprescrip-
tive. Moreover,foranypositiveintegerm it is a simplematterto construct a
strongly pseudoconvexdomain Q withboundarypoints P, Q such that Ad
near Q is (aftera rigidmotion)a 2m-order perturbation of ad near P-thus,
it can be arrangedthata rigidmotionof Q maps Q into Q, maps P to Q,
and agreeswiththe identityto order 2m, yetthe mappingis notbiholomor-
phic. In summary, a generalization
of ourresultsin thedirectionwe havebeen
discussingmusttakea morerestrictive form.We have thefollowing:
Theorem5.1. Let Q c C' be a smoothly bounded,stronglypseudoconvex
do-
mainwithCW boundary.Suppose P, Q E Ad and (D is a local biholomorphic
equivalenceof ad sendingP to Q. If F : Q Q2 is a holomorphicmapsuch
that
F(z) = (D?(z) + ?(lz-p14)
near P, then(D continuesanalytically
to a biholomorphism
of Q and F(z)
(D(z) on Q.
Proof.We firstnotethatthe localizedargument in theproofof Theorem4.5
I
can be appliedto (D o F in a neighborhood of P to showthat F _ (D near
P in K2. We wouldliketo use analyticcontinuation now,butmustdistinguish
betweentwocases.
(1) aO is notspherical,
i.e.,is nowherelocallyCR-equivalent to aB n. In this
case,thetheoremofKruzhilin-Vitushkin [VIT] saysthatthelocal holomorphic
equivalence(D can be analytically continuedalongarbitrary continuouscurves
in aO, as a local biholomorphic equivalencesendingaO to itself.By analytic
continuation, then,we concludethat F extendsCo to AO, and has nonvan-
ishingJacobiandeterminant. The mappingF is therefore a biholomorphism,
since it is a self-mapof Q.
(2) AO is spherical,i.e., locallyCR-equivalentto aOn at some (and hence
every,by analyticcontinuation) point. Thenby a resultof [BU], theuniversal
cover Q is biholomorphic to &n. The universalcovering Y' of a sufficiently
small open neighborhoodQ' of Q is also realizableas an open set Bn C
C c (Cn,and we let P E aBn be an inverseimageof P forthisextended
coveringmap. The mappingF liftsto F : B' _ By, and whatwe haveproved
above nowshowsthatin a neighborhood of P, F givesa local biholomorphic
equivalenceof B'n to itself.But it is knownfrom[CHM] thatsuch a local
equivalenceis therestriction of a globalautomorphism c1 of Bn . By analytic

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
RIGIDITY OF HOLOMORPHIC MAPPINGS 669

F _D on B' , and eD inducestheclaimedanalyticcontinuation


continuation,
of (D to Q. O
Remark3. In conclusion, we notethatthesetwocasesbehavedifferently
withre-
spectto analyticcontinuation ofboundaryequivalences,namely,thetheoremof
Kruzhilin-Vitushkin is falseif Ad is spherical.We do notknow,incidentally,
whetherin generaltheanalyticcontinuation givenby theKruzhilin-Vitushkin
theoremis singlevalued.
6. GEOMETRIC INTERPRETATIONS
AND RETURN TO THE QUESTION OF ORDER OF CONTACT

An eleganttheoremof Ian Graham[GRA] calculatestheasymptotic bound-


arybehaviorof the infinitesimal Caratheodoryand Kobayashimetricsneara
strongly pseudoconvexboundarypoint. The workof Fefferman in [FEF] cal-
culatestheasymptotic boundarybehaviorof theinfinitesimal Bergmanmetric
neara strongly pseudoconvexboundarypoint.In bothpaperstheproofsmake
it clearthatiftwostrongly pseudoconvex domainshavelocal defining functions
whichagreeto fourthorderat a strongly pseudoconvex boundarypoint P then
theCaratheodory (resp.Kobayashi,resp.Bergman)metrichas local boundary
behaviorat P whichis the same forbothdomains. Thus we see thatthehy-
pothesesof Theorem4.5 are preciselywhatare neededto guaranteethatthe
Caratheodory (resp.Kobayashi,resp.Bergman)metricnearthe distinguished
boundarypoint P has asymptotic boundarybehaviorwhichis identicalto that
forthe pullbackmetric(underthemapping(D) near P. To whatextentcan
thisreasoningbe reversed?That is, if (D is a self-mapof Q, takinga bound-
arypoint P to itselfand suchthat cD preservesasymptotically someinvariant
metricnear P, thenmust (D be an automorphism of Q?
Interestingly,theproblemis moresubtlethanthisdiscussionsuggests.For,
the fourthordercontactthatwe have been discussingis enoughto guarantee
identicallocal behaviorof themetrics(forinstance,it is enoughto pickup the
firsttermoftheasymptotic expansionfortheBergmankernelat P), butit does
not sufficeto capturetheglobalinformation thatis necessaryforthetruthof
the theoremswe have been considering.Thus, at thistimewe do not have a
purelygeometric methodforprovingthetheoremsof Sections3 and 4.
Now letus return to ouroriginalquestionconcerning "geometricorderofcon-
tact"oftheimageofa mappingwithitstargetboundary.To givea morenatural
geometrical notionof contacthereforourproblemwe consideran anisotropic
notionappropriateto the strongpseudoconvexity of our domains. Let Q as
beforebe a smoothly boundedstrongly pseudoconvexdomain,and fixP E aO.
Let p be a defining functionfor AO near P. Normalizethecoordinatesys-
tem at P so that z = 0 at P, the z -directionis the complexnormaldi-
rection,and z' (z2, ... Z,) are thecomplextangential
, directions.Define
2 .41
5(z) = (Iz112+ IZ I4)1/4.We say thata functionWN= WN(z) is of weight
N at P if IWN(Z)I< C5(z)N near P, forsome C > 0. Similarly, in anal-
ogywithLandau's notation,we will denoteby wN(z) a functionsuch that
I
limz 0 IWN(Z)/JN(z) = 0. These notionsare independent of the coordinate
normalizations.

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
670 D. M. BURNS AND S. G. KRANTZ

Now let (D be a holomorphic mapwhichis Wk on Q near P E AQ, (D(Q) c


Q', 4>(P) = Q E adQ', and let p' be a defining
functionfor adQ' at Q.
Definition
6.1. The set dI(D8) is said to have geometric
contactwith adQ' of
weightN at 4D(P) = Q if

(*) p o (D = h - p +wN

whereh is a Wk- positivefunctionon Q near P and WN is as above.


Remark4. In thisdefinition,k mustbe at least N forthedefinitionto make
intrinsicsense. See below for some remarkson more efficient
measuresof
here.
differentiability
Now letus relatethisgeometricnotionofcontactwiththeanalyticosculation
discussedearlier.
Proposition6.2. Let B' be theunitball in Cn, n > 2, and (D a holomorphic
mapping,(D: Bn Bn'. Suppose (D is F6 to theboundarynear P E adn and
Q = ??(P) E adBn. If (D(dB'n) has geometriccontactof weight6 witha3n at
Q, thenthereis a globalbiholomorphism T: Bn B'nsuchthat
D(z) = T(z) + o(IZ - p13)

near P.
Remark5. The RiemannMappingTheoremshowsthattheproposition is false
when n = 1. Concerningthe sharpnessof the proposition,and of Theorem
3.1, see Example6.3 below.
ProofofProposition by composing(D withau-
6.2. Withoutloss of generality,
tomorphismsof By, we can assume that P = Q = (1, 0, ... , 0). Conjugating
witha Cayleytransform,we can also replaceBnwithWn= {(Z, w) E Cn-I xC
Imw > Iz 2}, and P with0. The correspondence can be givenexplicitlyby

w =-iz+ X Z=
i z l j= ,. ,n-l
The coordinates(z, w) at 0 are,suitablyrenumbered, normalizedas in Def-
inition6.1, and the proofof the propositionis an examinationof our hy-
pothesis(*) with N = 6 in termshomogeneouswithrespectto the dilations
os
T
(z, w) -*(dz, 5~~~~~~~2
T,5f: W), > 0.
Again,composingwithautomorphisms of g/n we can assumewithoutloss
that (D has an expansionnear 0:
of generality
(D(z, w) = (z + A(z, w), w + B(z, w))
where
N
A(z, w) = Ea,,(z, w) + w(N+ 1),
v=2
N
B(z, w) = Ebv,(z, w) + w(N+ 1) .

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
RIGIDITY OF HOLOMORPHIC MAPPINGS 671

Here, a., bV are homogeneousof degreev forthe T -dilations.Also W(k),


as distinctfrom Wk, denotesan errortermin an obvioussense. Similarly,h
can be written
N
h = 1 +Eh, +w(N+ 1) .
v=2

We willuse N < 6 in theseexpansions.


Let av be the firstnonvanishing
termin A; then (*) showsthat bV+1 is
thefirstnonvanishing
termof B. Thus

(**) zP(z, w) = (z + aVO+ h.w.t.,w + bV0+1


+ h.w.t.)

where"h.w.t."is "higherweightterms"and vo > 2. The partof equation


(*) homogeneousof weightvo+ 1 (where vo < N), and with P = Q = 0,
p = pI = Imw - lz 2, isjust

(*)VO+l hv 1(Imw - IZ 2) = -2 Re(avo, z) + bv0+l

where,for v, v' E Cn-l, (V, V') = En-l VfV


Now theoperation

(au, bV0+1) - 2Re(av, z) + Imbv+1

evaluatedat Imw- = z12, is the operator"L" on formalpowerseriesthat


was definedin formula(2.6) of [CHM]. Equation (*)V +1 says preciselythat
L(a , bV+1) = 0. We shallconsidervo = 2, 3 directly;thecases vo= 4, 5,
whichwe shallalso need,willthenfollowfromLemma2.1 of [CHM].
The case vo= 2. In thiscase we have

h= (a, z)+(a,z), some a E C


a2= Q(Z) +,Bw , Q quadraticin z, lE n,
b3 = C(z) + w(z, ga), C cubic, y EC.

Then equation (*)3 becomes

(at Z)W _ (0t, Z)W 2 _


_d 2)IZ2 (at, Z)W _ o, )
2i 2i ) ( )l + , )z)+ 2i 2i )

= -((Q(z), z) + (Q(z), z)) - (w(fl, z) + w(f, z))

z_ - C(z) +(w(z, Y) 2w(z, w)


nlik m 2i o e2i 2i

Comparinglikemonomialson eitherside (let us label thesewiththenotation

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
672 D. M. BURNS AND S. G. KRANTZ

(p, q) ), we findthat

(3,0 ) 0 2i) or C(z) = 0;

(2,0 ) w(z, ) = W(z, Y) or a= y;


21 2i
(3, 1) -(z, a)(z, z) = -(Q(z), z) or Q(z) = (z, a)z;
(1, 1) (a, z)w (z ,)W = -w(z, a) - w(f, z);
21 21
hence a = -2i,B, some ,BE Cn.
Before proceeding,we recall that the automorphismsof Wn with 0 E aWn a
fixedpointand normalizedas in (**) have theform

(z, w)!- ( z +w
1-2i(z,/))+cw
________

1-2i(Z,,B)+ C

where ,B E Cn is arbitrary
and Imc = -Ifl12. The calculationsperformed
of ,n we
abovetellus that,aftercomposingwithone of theseautomorphisms 5

mayassumethat (**) holdswith vo > 3.


The case vo= 3. Now we have
h2= q(z) + aw + a-w+ q(z), q quadratic, a E n;
a3 = C(z) + wL(z) , C cubic, L linear;
b4= Q(z) +wS(z) + yw2, Q quartic, S quadratic, j E C.
Thus (*)4 becomes

Im(qw + aw2 2Req Iz12 - Im(qw) - 2e(aw lz) 2 Ima

Im(Q + wS + yw2) - 2Re[(C(z), z) + w(L(z), z)] .


=

Comparingliketermsnowyields
(4,0) Q 0;
(3, 0) S q
(2, 0) a= Y;
(3, 1) C(z) = q(z) z;
(1, 1) Ima = 0;
(2, 1) -aw(z, Z) -q2i = -w(L(z), z)

hence q O0 and L(z) = az.


Note that we have now that q = S = 0 and C = z * q = 0. The upshot
of our calculationis thatthemapping(D agreesin form(up to ordervo) with
an automorphism of Wn thatfixes0. Composing(D withtheinverseof said
automorphism (as we are obviouslyfreeto do), we have that (D satisfies(**)
=
with vo 4.
The cases vo= 4, 5 . ApplyingLemma2.1 of [CHM, p. 233], we findthat
(a4 b5) = (O,0 ) and (a5, b6) = 0.

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
RIGIDITY OF HOLOMORPHIC MAPPINGS 673

Thus we have provedthatthegeometric


orderof contact6 impliesthat
ID(z, w) = (z, w) + (termsof homogeneity
at leastsix) .
This is almostthe desiredresult.In termsof the usual notionof order,this
resultsaysthatthereis a constanta E C suchthat,withoutloss of generality,
3
(Dz, W) = (Z + aW W) + 4(1Z - pi,)

We can now use a globalargumentrelatedto Theorem3.1 above. Consider


the mappingof the upperhalf plane := {Im w > 0} to itselfgivenby
w -w + B(O, w) . By theabove we knowthat B(O, w) is O(1w14) at w = 0
and so bytheconjugatedformof theboundarySchwarzlemma,B(O, w) _ O.
Now considerthemap of D into Zn givenby
w-4(0, w) = (A(O, w), w + B(O, w)) = (A(O, w), w) .
By theuniquenessargumentat theend of theproofof Theorem3.1, it follows
that A(O, w) _ 0. Hence, the coefficient
a above is 0, whichprovesthe
proposition. 0
Remark6. Of course we could assume above that 1(0D11) had contactof weight
N > 7, and the propositionwould followfrompurelylocal calculations.It
wouldcertainly notbe sharpthen.
In theaboveproposition, we can weakenconsiderably thenotionof differen-
tiabilityon 1Dand h as follows:Say that D is admissiblydifferentiable at P
ofweightN, ifin coordinates normalizedat P = 0 as above,thereexistsan ap-
proximation of 5(z, w) = (A(z, w), B(z, w)) whereA _ EN 14a,,(z, w) =
WN(Z, w) and B - Ev=o b,,(z, w) = WN(Z, w). Here theapproximations are
to holdon admissibleapproachregionsat 0 in thesenseof Koranyi.Notethat
the a.,, b, are assumedto be T3-homogeneous as in theproofabove. One can
similarly definenontangential of orderN, usingnontangential
differentiability
approachregions.The corresponding conceptsfortherealfunctionh we leave
to thereader.Proposition6.2 followsif D is admissibly ofweight
differentiable
6 and nontangentially of order3 at P. Presumably
differentiable one does not
need all derivativesin z of (D up to order6 forthe validityof Proposition
6.2. We have chosento workwiththe notionof contactexpressedin terms
of weightbecauseof its naturalitywithrespectto CR-geometry, as in [CHM].
Example6.3 belowshowsthat,withthismeasureof contact,Theorem3.1 and
Proposition6.2 are sharp.
Example6.3. We will constructself-mapsof the Siegel domain W2 to itself
whichdescribewithsome precisionhow sharptheabove resultsare. We start
fromthe RiemannMappingTheoremin C, and take V to be a smoothly
bounded,simplyconnectedopenset c X (theupperhalfplanein C) suchthat
vnaA' in C containsa neighborhood of 0 in therealaxis. Let (o : -* V be '

a Riemannmapping(i.e.,biholomorphism). Since ip is one-to-one,(o'(w) $40,


for all w E X . Since ' is simply connected, we can choose a branch of
o'(w) globallyon X. By thePick-Schwarzlemma,appliedto themap q,

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
674 D. M. BURNS AND S. G. KRANTZ

we knowthat

< 1
(#)M___________'()_
IImo(w)12 ImwI -

By (#), themappingof Cx*cC2 to C2 givenby

F: (z, w)-)( ?0'(w)z, (0(w))


we maychoose (0 in sucha way
sends 2 to itself.Withoutloss of generality,
that
~P() =w +Aw3 +O(IwI4),
9v(w)=w+sw +0W14)

whereA is real, A > 0. Then,in a neighborhoodof 0, we have

F(z, w) = (Z + 3AW2z, W + ASW3)+ ?(I(Z' W)14)-

Here we have chosenthebranchof ?o'(w) whichis 1 at w = 0. For this


map, F(aO2) has contactof weight5 butnotweight6; thecase vo= 5 above
breaksdown here. This example(extendedto fn) also showsthatTheorem
past
3.1 is sharpforall n. Note thatthemap F actuallyextendsanalytically
0.
to attemptto carryout the proofof Proposition6.2 for
It is enlightening
n = 1 to see thatthe coefficientsof (D are not determinedby the contact
equation,no matterwhatorderof contactis assumed.Note also thatthehigher
dimensionalexampleabove has contactof F(aO2) withOW2 alongtheentire
curve (O, u), u = Rew E R , near u = 0. The contactweightis sharpat all
pointsof thiscurve.

7. DISCUSSION OF THE WEAKLY PSEUDOCONVEX CASE AND CLOSING REMARKS


The analysiswe have been discussingdoes not carryover to the case of
weaklypseudoconvex thewealth
domains.A simpleexampleservesto illustrate
of questionsand problemsavailable:Considerthedomains

im = {(Z1 , Z2): IZ12 + IZ21 < 1},


for m a positiveinteger.What conditionsnear 1 = (1, 0) on a holomor-
phic mapping(D1:Qm ' Qm will force Q to be the identitymapping?The
keyobservationis that Qm coverstheball B via the mapping /m/(ZimZ2) =
(Z1, (z2)m). For any a = (a, a2) E B witha2 4 0?,theanalyticdisc da (see
?3) liftsto m extremaldiscs daj j = 1, ... m, in Qm. Then theanalysis
of ?3 maybe carriedoutwiththeseextremaldiscs-providedthatthemapping
so thatafterit is pushed
restricted
(D has Taylorjet at P whichis sufficiently
downwiththe functionv/moffthebranchlocus,it satisfiestheconditionsof
Theorem3.1. Note in particularthatthe correctconditionon the Taylorjet
willbe nonisotropic and willdependon m. X. Huang[HU 1, HU2, HU3] has
exploredthesemattersin some detail.

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
RIGIDITY OF HOLOMORPHIC MAPPINGS 675

An argumentsimilarto the one just outlinedmaybe carriedout near any


weaklypseudoconvexboundarypointthatlocally"covers"a strongly pseudo-
class,and has recently
convexboundarypoint.This is in facta ratherrestrictive
been describedby Barlettaand Bedford[BABE] (however,see also [GAV]).

ACKNOWLEDGMENT

We thankJohnBland and RobertE. Greeneforhelpfuldiscussionsabout


thegeneralnatureof thisproblemand Al Baernsteinfora usefulconversation
theproofof Theorem2.1. We also thankJ.Moserand N. Sibonyfor
regarding
theirtimelyinterest
and suggestions.

REFERENCES

Conformal
[AHL] L. Ahlfors, McGraw-Hill,
invariants, New York,1973.
[ALE] H. Alexander,Holomorphicmappings from theball Math.Ann.209 (1974),
and polydisc,
249-256.
(AIZ] L. A. Aizenberg,Multidimensionalanaloguesof the Carlemanformulawithintegration
overboundary setsof maximaldimension, Akad. Nauk. SSSR Sibirsk.Otdel. Inst.Fiz.,
Krasnoyarsk,1985,pp. 12-22,272. (Russian)
[BABE] E. Barlettaand E. Bedford,Existenceofpropermappings fromdomainsin C2, Indiana
Univ.Math.J.39 (1990), 315-338.
(BU] D. Bums,A multi-valued Hartogstheoremand developingmaps,preprint.
[BSW] D. Bums, S. Shnider,and R. Wells,On deformations of strictly domains,
pseudoconvex
Invent.Math.46 (1978), 237-253.
in complexmanifolds,
[CHM] S. S. Chernand J.Moser,Real hypersurfaces ActaMath.133 (1975),
219-271.
(FEF] C. Fefferman, TheBergmankerneland biholomorphic mappings domains,
ofpseudoconvex
Invent.Math.26 (1974), 1-65.
[FOR] J. Fomrnss, pseudoconvex
Strictly domainsin convexdomains,Amer.J. Math.98 (1976),
529-569.
[GAV] E. Gavosto,Thesis,Washington 1990.
University,
[GRA] I. Graham,Boundary oftheCaratheodory
behavior and Kobayashimetricsonstronglypseu-
doconvex domains in Cn with smooth boundary,Trans. Amer. Math. Soc. 207 (1975),
219-240.
[GKl] R. E. Greeneand S. G. Krantz,Stability of theBergmankerneland curvature
properties
of boundeddomains,RecentProgressin SeveralComplexVariables,Princeton
properties
Univ.Press,Princeton,NJ,1982.
(GK2] -, Deformationofcomplexstructures, forthe a equation,and stability
estimates ofthe
Bergmankernel, Adv.Math.43 (1982), 1-86.
[GK3] -, Methodsforstudying theautomorphism groupsof weaklypseudoconvex domains,
Proc. Intemat.Conf.on ComplexGeometry(Cetraro,Italy),Mediterranean Press,Cal-
abria,1991.
[HUI] X. Huang,Some applicationsofBell's theorem domains,PacificJ.
to weaklypseudoconvex
Math.(to appear).
[HU2] , A boundary problemofholomorphic
rigidity mappingson weaklypseudoconvex do-
mains, preprint.
[HU3] -, Preservation ofextremalmappings
principle IllinoisJ.Math.(to
and itsapplications,
appear).
An introduction
[KAT] Y. Katznelson, analysis,Dover,New York,1976.
to harmonic

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions
676 D. M. BURNS AND S. G. KRANTZ

[KOB] S. Kobayashi,Hyperbolic manifolds and holomorphic mappings,MarcelDekker,NewYork,


1970.
[KRI] S. G. Krantz,Functiontheory ofseveralcomplexvariables, Wiley,New York,1982.
[KR2] -, A newcompactness principle in complexanalysis,Univ.Autonomade Madrid,1987.
[KOR] J. Kohn and H. Rossi,On theextension ofholomorphic functions
fromtheboundary ofa
complexmanifold, Ann.of Math.(2) 81 (1965), 451-472.
[KRV] N. KruzhilinandA. Vitushkin, Extensionoflocalmappings ofpseudoconvexsurfaces,
Dokl.
Akad.Nauk.SSSR 270 (1983), 271-274.
[LEMI] L. Lempert,La metrique Kobayashiet les representationdes domainssurla boule,Bull.
Soc. Math.France109 (1981), 427-474.
[LEM2] , Holomorphic retractsand intrinsicmetricsin convexdomains,Anal.Math.8 (1982),
257-261.
[LEM3] -, Intrinsicdistancesand holomorphic ComplexAnalysisand Applications,
retracts,
Sofia,1984.
[PIN] S. Pincuk,On theanalyticcontinuation ofholomorphic mappings,Mat.USSR-Sb.27 (1975),
416-435.
[RUI] W. Rudin,Holomorphic mapsthatextendto automorphisms ofa ball,Proc.Amer.Math.
Soc. 81 (1981), 429-432.
[RU2] -, Functiontheory Berlinand New York,1980.
oftheunitball in Cn , Springer-Verlag,
[SIB] N. Sibony,Unpublishednotes.
[VEL] J.Velling,Thesis,Stanford, 1985.
[VIT] A. G. Vitushkin, Real analytichypersurfacesofcomplexmanifolds, UspekhiMat.Nauk.40
(1985), 3-31.
[WOG] W. Wogen,Composition operators actingon spacesofholomorphicfunctionson domainsin
Cn, OperatorTheory:OperatorAlgebras Part2 (Durham,New Hamp-
and Applications,
shire,1988),Proc.Sympos.PureMath.,vol. 51,Part2, Amer.Math.Soc.,Providence, RI,
1990,pp. 361-366.
[WU] H. H. Wu,Normalfamilies ofholomorphic mappings, ActaMath.119 (1967), 193-2333.
[YAU] S. T. Yau, A generalized Schwarz lemma forKahler manifolds,Amer.J.Math.100 (1978),
197-204.

ABSTRACT. A rigiditytheoremforholomorphic mappings,in thenatureof the


uniquenessstatement of the classicalone-variableSchwarzlemma,is proved
at theboundaryof a strongly pseudoconvex domain.The resultreducesto an
and apparently
interesting, new,resultevenin one complexdimension.The
theoremhas a varietyof geometric and analyticinterpretations.

DEPARTMENT OF MATHEMATICS, UNIVERSITY OF MICHIGAN, ANN ARBOR, MICHIGAN 48109


DEPARTMENT OF MATHEMATICS, Box 1146, WASHINGTON UNIVERSITY IN ST. LOUIS, ST. LOUIS,
MISSOURI 63130

This content downloaded from 91.229.229.96 on Thu, 12 Jun 2014 23:33:03 PM


All use subject to JSTOR Terms and Conditions

You might also like