(de Gruyter 2003
Complex geometry of generalized annuli
Chiara de Fabritiis
(Communicated by G. Gentili)
Abstract. We study the complex geometry of a class of domains inCn which generalize the annuli inC, i.e., which are quotients of the unit ballBnofCnfor the action of a group gen- erated by a hyperbolic element of AutBn. In particular, we prove that the degree of holomor- phic maps between two such domains is bounded by a constant which depends on the ‘‘radii’’
of the domains only and we give some results on the existence of complex geodesics for the Kobayashi distance in these domains.
Key words.Holomorphic maps, annuli, degree of a map, complex geodesics.
2000 Mathematics Subject Classification. Primary 32A10, 32A40; secondary 32H15, 32A30
1 Introduction
The aim of this paper is to study the complex geometry of a class of domains in Cn which are quotients of the unit ball BnHCn for the action of a group gen- erated by a hyperbolic element of AutBn (see Section 2 for definitions). Since the annuli in C are obtained as quotients of the unit disk D¼ fxAC:jxj<1g for the action of a group generated by a hyperbolic element of AutD, the domains we study can be seen as a generalization of annuli to several complex variables.
In Section 2 we give some definitions we need in the sequel of the paper and we recall some statements concerning this class of domains, which were introduced in [3]. In Section 3 we generalize to several complex variables a result which is due to Schi¤er in the one-dimensional case: the degree of a holomorphic map between two annuli is bounded by a constant which depends on the moduli of the annuli only.
In Section 4 we study the geometry of extremal mappings and complex geodesics for the Kobayashi distance in this class of domains. In particular we prove that there always exists an extremal mapping through two given points of a ‘‘generalized annu- lus’’ and we give several results on existence and non-existence of complex geodesics according to the ‘‘radius’’ of the domain and to other parameters which classify these domains.
2 Preliminaries and statements
We denote the unit ball for the Euclidean metric inCnbyBn; the following result is well known.
Theorem 2.1.Any holomorphic automorphismgofBncan be extended holomorphically to a neighborhood of the closure ofBn;ifghas no fixed points inBn,then its extension has either one or two fixed points inqBn.
From now on we shall denote by the same symbol a holomorphic automorphism of Bnand its extension to the closure ofBn.
Definition 2.2.LetgAAutBn: ifghas at least one fixed point inBn, thengis said to beelliptic; ifghas no fixed points inBnand has one fixed point inqBn, it is said to be parabolic; ifghas no fixed points inBn and has two fixed points inqBn, it is said to behyperbolic.
To generalize the construction of annuli to several complex variables, we will focus our attention to the action of hyperbolic elements on Bn. First of all, we recall a result which is due to de Fabritiis and Gentili (see [5]).
Proposition 2.3.Let gbe a hyperbolic element inAutBn;then there exist TAR and y2;. . .;ynARsuch thatgis conjugate to
g0:z7! z1coshTþsinhT
z1sinhTþcoshT; eiy2z2
z1sinhTþcoshT;. . .; eiynzn z1sinhTþcoshT
: ð2:1Þ
In the sequel it will be useful to consider the problem on the Siegel half-spaceHn ¼ fwACnjIw1>jw2j2þ þ jwnj2g which is biholomorphic to Bn via the Cayley transformC(see e.g. Rudin [7] or Abate [1]), so we also give the form of hyperbolic elements in AutHn.
Corollary 2.4. Let mAAutHn be hyperbolic; then there exist lARþnf1g and y2;. . .;ynARsuch thatmis conjugate to
Cg0C1 ¼m0:w7! ðl2w1;leiy2z2;. . .;leiynznÞ; ð2:2Þ wherel¼eT.
This result enables us to consider the quotients ofBn (resp.Hn) for the action of the groupGðMÞgenerated by a hyperbolic elementgAAutBn ðmAAutHnÞ. Since the quotients Hn=M1 andHn=M2 are biholomorphic i¤ M1 andM2 are conjugate in AutHn, then it is enough to consider the case of a group generated by an element of the form (2.2). As we are interested in the groupM generated bym0, rather than in the element m0 itself, we can always suppose that l>1, that is T>0. Let
ln:H1!R ð0;pÞbe a branch of the logarithm, set b¼1=lnl¼1=T and con- sider the holomorphic map pr:Hn!Cngiven by
prðwÞ ¼ ðeibplnw1;ebðlnlþiy2Þlnw1=2w2;. . .;ebðlnlþiynÞlnw1=2wnÞ: ð2:3Þ In [3] and [4] the following result is proved:
Theorem 2.5.Let m0 be given by(2.2)and M be the group generated bym0.Then the mappr:Hn!Cn given by(2.3)factors onHn=M giving a biholomorphism between
Hn=M and the bounded domain
Wðr;y2;. . .;ynÞ ¼
xACn
r<jx1j<1;Xn
j¼2
jxjj2jx1jyj=p<sin plnjx1j lnr
;
where r¼ep2=lnl¼ep2=T Að0;1Þ, y2;. . .;ynAR. In particular Hn=M is a Stein manifold which is biholomorphic to a bounded domain inCn.
As a consequence of the previous theorem, the domainsWðr;y2;. . .;ynÞcan be seen as a generalization to several complex variables of the annuli, which are the quotients of the unit diskD¼B1 for the action of the groups generated by hyperbolic auto- morphisms ofD. In factWðrÞ ¼ fxACjr<jxj<1gis an annulus inCand for this reason we shall often call these domains ‘‘generalized annuli’’.
To simplify notation, when no confusion can arise we only write W instead of Wðr;y2;. . .;ynÞ. Of course, via the Cayley transform we can also study the problem onBn: in this case the covering will be given byðBn!w WÞwherew¼prC.
Remark 2.6. Notice that the domainWðr;y2;. . .;ynÞretracts by deformation on the annulusWðrÞ, in particular it is doubly connected.
From now on, we shall denote byW1the domainWðr1;y2;. . .;yn1ÞHCn1 and byW2 the domainWðr2;Q2;. . .;Qn2ÞHCn2; in order to simplify the notation, the symbolH will stand for HolðW1;W2Þ ¼ ff :W1!W2jf holomorphicg. For all f AH, we will denote the degree of f bydðfÞ.
Definition 2.7.Let f;gAH. We say that f andgarehomotopicif there exists a con- tinuous mapF :½0;1 W1!W2 such that
(i) Fð0;Þ ¼ f,Fð1;Þ ¼g;
(ii) Fðt;ÞAHfor alltA½0;1.
Of course, the fact of being homotopic is an equivalence relation on H; we shall denote the homotopy class of f AHby½f.
At last, ifDis a domain inCn, we shall denote bykDð;ÞtheKobayashi distance on D and bykDð;Þthe Kobayashi metric on D(for a comprehensive reference on this topic see [1] or [6]).
3 Holomorphic maps between generalized annuli
In this section we generalize the estimate of the degree of a holomorphic map between two annuli which is due to Schi¤er (see [8]) in the one-dimensional case.
Theorem 3.1.If f AHolðWðr1Þ;Wðr2ÞÞ,thenjdðfÞjc½lnr2=lnr1.Moreover,if equal- ity holds there existsyARsuch that for allxAWðr1Þ
fðxÞ ¼ eiyxdðfÞ if dðfÞ>0;
r2eiyxdðfÞ if dðfÞ<0:
(
LetW1¼Wðr1;y2;. . .;yn1ÞHCn1 andW2¼Wðr1;Q2;. . .;Qn2ÞHCn2 be two general- ized annuli; since the fundamental group of bothW1 andW2 is isomorphic toZ, we can define the degree dðfÞAZof a holomorphic map f AH by choosing genera- tors aa~j of p1ðWjÞ represented by ajðtÞ ¼ ðpffiffiffiffirj
e2pit;0;. . .;0Þ for j¼1;2 and setting fðaa~1Þ ¼dðfÞaa~2.
Theorem 3.2.If f AH,thenjdðfÞjc½lnr2=lnr1.
Proof.LetSn ¼ fwACnjIw1Að0;pÞ;Iew1>jw2j2þ þ jwnj2g; it is easily verified that the mapE:SnCw! ðew1;w2;. . .;wnÞAHnis a biholomorphism.
Then we can consider the coveringsðSnj !qj WjÞfor j¼1;2, where the mapsq1and q2are given by
q1ðwÞ ¼ ðeib1pw1;eb1ðlnl1þiy2Þw1=2w2;. . .;eb1ðlnl1þiyn1Þw1=2wn1Þ; q2ðwÞ ¼ ðeib2pw1;eb2ðlnl2þiQ2Þw1=2w2;. . .;eb2ðlnl2þiQn2Þw1=2wn2Þ;
withbj ¼ lnrj=p2 and lnlj¼1=bj. The group of deck-transformations of the cov- eringsðSnj !qj WjÞis generated bynj ¼E1mjE:Snj !Snj given by
n1ðwÞ ¼ ðw1þ2 lnl1;l1eiy2w2;. . .;l1eiyn1wn1Þ;
n2ðwÞ ¼ ðw1þ2 lnl2;l2eiQ2w2;. . .;l2eiQn2wn2Þ;
respectively. Since the domainSn1is simply connected, there exists a continuous map ff~:Sn1 !Sn2 such that q2 ff~¼ f q1; the mapsq1 andq2 being local biholomor- phisms, we immediately obtain that ff~is holomorphic. Interpreting the degree of f via the isomorphism between the fundamental groups ofWj and the groups of deck- transformations of the coverings, we obtain the following equality
ff~n1 ¼n2dðfÞ ff~; ð3:1Þ the comparison between (3.1) and the contracting property of the Kobayashi dis- tance will yield the conclusion. Let us consider the points w0¼ ðip=2;0;. . .;0Þand
w1¼n1ðw0Þ ¼ ð2 lnl1þip=2;0;. . .;0ÞASn1; if kD denotes the Kobayashi distance onD, we have the following chain of inequalities
kSn
2ðff~ðw0Þ;ff~ðw1ÞÞckSn
1ðw0;w1Þ ¼kHn1ðEðw0Þ;Eðw1ÞÞ
¼kH1ði;l12iÞ ¼lnl1¼ p2=lnr1;
where the second equality is due to the fact thatEðw0Þ;Eðw1ÞAH1 f0gwhich is a holomorphic retract ofHn. Since
kSn
2ðff~ðw0Þ;ff~ðw1ÞÞ ¼kSn
2ðff~ðw0Þ;ff~ðn1ðw0ÞÞÞ ¼kSn
2ðff~ðw0Þ;n2dðfÞðff~ðw0ÞÞÞ;
considering the projection on the first component and denoting f1ðw0Þ by c, the above inequality and the contracting property of the Kobayashi distance yield
kS1ðc;cþ2dðfÞlnl2ÞckSn
2ðff~ðw0Þ;n2dðfÞðff~ðw1ÞÞÞcp2=lnr1:
Via the biholomorphismE:S1!H1, we can evaluatekS1ðc;cþ2dðfÞlnl2Þobtain- ing
kS1ðc;cþ2dðfÞlnl2Þ ¼kH1ðec;l22dðfÞecÞdp2jdðfÞj=lnr2;
asr2<1, this impliesjdðfÞjclnr2=lnr1. r This means that the ratio of the logarithms of ‘‘inner radii’’ (that is, the generaliza- tion of the ratio of the moduli) bounds the degree of holomorphic maps between gen- eralized annuli. The following proposition gives an even deeper interest to the above theorem, since it tells us that the degree is a complete homotopy invariant.
Proposition 3.3.Let f;gAH.Then dðfÞ ¼dðgÞif and only if½f ¼ ½g,that is if and only if f and g are homotopic.
Proof.The ‘‘if ’’ part is obvious. In order to prove the converse implication, we write f ¼ ðf1;. . .;fn2Þandg¼ ðg1;. . .;gn2Þ; using the retraction by deformation ofW2onto Wðr2Þ f0ggiven byrðt;xÞ ¼ ðx1;tx2;. . .;txnÞ, it is easily seen that we can limit our- selves to the casen2¼1.
Settingd¼dðfÞ ¼dðgÞ, we then have 1
2pi ð
ffiffiffir1
p S1
qf
qx1ðx1;0;. . .;0Þ=fðx1;0;. . .;0Þdx1
¼ 1 2pi
ð ffiffiffir1
p S1
qg
qx1ðx1;0;. . .;0Þ=gðx1;0;. . .;0Þdx1 ¼d:
Letj;c:W1!Cbe given by jðxÞ ¼xd1 fðxÞandcðxÞ ¼xd1 gðxÞ; it is easily veri- fied that
1 2pi
ð ffiffiffir1
p S1
qj
qx1ðx1;0;. . .;0Þ=jðx1;0;. . .;0Þdx1
¼ 1 2pi
ð ffiffiffir1
p S1
qc
qx1ðx1;0;. . .;0Þ=cðx1;0;. . .;0Þdx1¼0:
Consider the maps j;c:p1ðW1Þ !p1ðCÞ; the above equality implies that both these maps are trivial and therefore j;c can be lifted to continuous maps j~
j;cc~:W1!Csuch that expjj~¼jand expcc~¼c. Since exp:C!C is a local biholomorphism and both j and c are holomorphic, the two maps jj~ and cc~ are holomorphic. Then we have found holomorphic maps jj;~cc~:W1!C such that
fðxÞ ¼x1de~jjðxÞ andgðxÞ ¼x1deccðxÞ~ for allxAW1. Now set Hðt;xÞ ¼x1dejjðxÞþtð~ ccðxÞ~~ jjðxÞÞ;
obviouslyH is continuous, and the mapHðt;Þis holomorphic for alltA½0;1; more- overHð0;Þ ¼ f andHð1;Þ ¼g, so we are left to verify thatHð½0;1 W1ÞHWðr2Þ.
FixxAW1and consider the map
b:½0;1Ct7! jHðt;xÞj ¼ jx1dje<~jjðxÞþt<ðccðxÞ~~ jjðxÞÞAR;
the mapb is monotonic and we haver2<jfðxÞj ¼bð0Þ<1,r2 <jgðxÞj ¼bð1Þ<1.
Thus r2<jHðt;xÞj ¼bðtÞ<1 for all tA½0;1; so Hð½0;1 W1ÞHWðr2Þ and this concludes the proof of the assertion, since H is the required homotopy between the
maps f andg. r
As a consequence of the proof of Proposition 3.3 we obtain the following
Corollary 3.4. Let f AH, then there exists a holomorphic map u:W1!Csuch that f1ðxÞ ¼x1dðfÞeuðxÞfor allxAW1.
Moreover, gathering Theorem 3.2 and Proposition 3.3, we obtain that, if the ‘‘hole’’
inW1is smaller than the ‘‘hole’’ inW2, any holomorphic map fromW1toW2is homo- topic to a constant map.
Corollary 3.5.Let f AH;if r1<r2 then f is homotopic to a constant.
Now we turn to the study of the homotopy classes Hd ¼ ff AHjdðfÞ ¼dg. If jdj<lnr2=lnr1 the following remark and proposition ensure that the family Hd is very ‘‘ample’’.
Remark 3.6.For all f AH0, there exists a holomorphic map ff~:W1!Hn2such that pr2 ff~¼ f. Vice versa, for all holomorphic map ff~:W1!Hn2, the map pr2 ff~ belongs toH0.
Proof. Since Hn1 is simply connected, the existence of a continuous ff~:W1 !Hn2
such that pr2 ff~¼ f is equivalent to the triviality of the map f(that is todðfÞ ¼0).
As pr2 is a local biholomorphism, the assertion follows. r This simple remark classifies all elements of H0; moreover the boundedness of W1 ensures that there exists a huge family of holomorphic maps fromW1toHn2; in this sense we can say thatH0is ‘‘big’’.
Now let us consider the case when 0<jdj<lnr2=lnr1. First of all, ifn2d2 we set L¼maxfjQjj=pjj¼2;. . .;n2g. If 0<d <lnr2=lnr1, choose CAR such that r2 <
Cr1d <r1d <C; if 0<d <lnr2=lnr1, choose CAR such that r2<Cr2 <r2r1d <
Cr2r1d <1 and set
d¼ minfð1CÞ=2;ðCr1dr2Þ=2g if d >0;
minfðCr2r2Þ=2;ð1Cr1dr2Þ=2g if d <0;
K¼
min sin plnðð1þCÞ=2Þ lnr2
;sin plnððCr
d 1þr2Þ=2Þ lnr2
n o
if d >0;
min sin plnððCrln2rþr2Þ=2Þ
2
;sin plnðð1þCrlnr1dr2Þ=2Þ
2
n o
if d <0:
8>
<
>:
Then we can describe an ample set of the elements belonging toHd:
Proposition 3.7.Let dAZbe such that0<jdj<lnr2=lnr1and C;d;L;K be as above.
For anyyAR,any holomorphic functions:W1!Csuch thatkskycdand any holo- morphic map h:W1!Cn21such thatPn21
j¼1 jhjðxÞj2<r2LK for allxAW1,the map f :W1Cx7! ðCeiyx1dþsðxÞ;hðxÞÞAW2; if d>0;
ðCr2eiyx1dþsðxÞ;hðxÞÞAW2; if d<0;
(
belongs toHd.
Proof.We shall perform the proof in the cased >0, the cased <0 can be obtained from this by minor changes. First of all, let us prove that the map f belongs toH.
Since it is obvious that f is a holomorphic map, we are left to prove that it mapsW1 intoW2. LetxAW1; the choice ofCanddimplies that the first component of f sat- isfies the following inequality
r2<Cr1dþr2
2 <jf1ðxÞj<Cþ1
2 <1; ð3:2Þ
which yields
0<plnððCþ1Þ=2Þ
lnr2 <plnjf1ðxÞj
lnr2 <plnððCr1dþr2Þ=2Þ lnr2 <p:
The map sin:½0;p !Ris concave and therefore the choice ofKentails
Kcsin plnjf1ðxÞj lnr2
ð3:3Þ
for allxAW1. By (3.2) and (3.3) we then have Xn2
j¼2
jfjðxÞj2jf1ðxÞjQj=pcrL2 nX21
j¼1
jhjðxÞj2cK<sin plnjf1ðxÞj lnr2
and therefore f mapsW1toW2. At last, we prove thatdðfÞ ¼d, that is f AHd. Let H :½0;1 W1!Cn2 be given byHðt;xÞ ¼ ðCeiyx1dþtsðxÞ;thðxÞÞ. The mapHis of course continuous and by the above reasoning we haveHð½0;1 W1ÞHW2; more- overHð1;Þ ¼ f andHð0;xÞ ¼ ðCeiyx1d;0;. . .;0Þ. Since the degree ofHð0;Þis equal
tod, we are done. r
Now we consider the case when jdðfÞj ¼lnr2=lnr1 which of course can occur i¤
lnr2=lnr1AN. The following results describe the maps f AHd whenjdj ¼lnr2=lnr1: the restriction of f to Wðr1Þ f0g has to follow a prescribed pattern, while outside this annulus the behaviour of f can be quite ‘‘free’’.
Theorem 3.8. Let f AHd where jdj ¼lnr2=lnr1. Then there exist yAR and pAHolðW1;CÞsuch that pjWðr1Þf0g10,qxqp
j
Wðr1Þf0g10for all j¼2;. . .;n1and f1ðxÞ ¼ eiyx1depðxÞ if d>0;
eiyr2x1depðxÞ if d<0:
(
ð3:4Þ
Proof.Since we are interested in the behaviour of f1only andW2retracts by deforma- tion onto Wðr2Þ f0g, it is enough to consider the case n2¼1. Moreover we shall perform the proof only in the cased >0, the cased <0 being obtained from this by a few minor changes.
By Corollary 3.4 we can find a holomorphic mapu:W1!Csuch that
fðxÞ ¼x1deuðxÞ: ð3:5Þ Let us consider the map t:Wðr1ÞCx17! fðx1;0;. . .;0ÞAWðr2Þ; since Wðr1Þ is a deformation retract ofW1 the degree oftis equal tod. AsS1¼R ð0;pÞis simply connected andq2:S1!Wðr2Þis a local biholomorphism, there exists a holomorphic lifting ~ttof tq1, that is a holomorphic map ~tt:S1!S1 such that q2~tt¼tq1. Interpreting the degree oftvia the action of the group of deck-transformations yields
~ttn1 ¼n2d~tt, that is
~ttðzþ2 lnl1Þ ¼~ttðzÞ þ2dlnl2
for allzAS1. Takingz¼ip=2 and using the equalitydlnl2 ¼lnl1 we obtain
~tt ip
2 þ2 lnl1
¼~tt ip
2 þ2 lnl1: ð3:6Þ
Transferring the problem onH1, we obtain as a consequence of the Schwarz lemma that~ttis an automorphism ofS1of the form~tt:z7!zþxfor a suitablexAR.
In fact, let s¼Ett~E1:H1!H1 and set ~ttðip=2Þ ¼xþiy. Since sðiÞ ¼ expðxþiyÞandsðil21Þ ¼l12expðxþiyÞ, by the contracting property of the Kobaya- shi distance we obtain
kH1ðsðiÞ;sðil12ÞÞ ¼tanh1 l211 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi l14þ12l12cos 2y q
0 B@
1
CAckH1ði;il12Þ ¼lnl1:
Now the fact that tanh is increasing yield l121 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi l41þ12l12cos 2y
q cl211 l21þ1
and hence cos 2y¼ 1, that is y¼p=2. Moreover the equality at one point of the Kobayashi distance yields, by the Schwarz lemma, that s is an automorphism of H1; sincesðiÞ ¼iex andsðil12Þ ¼iexl12, we havesðwÞ ¼exwfor all wAH1, that is
~ttðzÞ ¼zþxfor a suitablexAR. Then there existsyARsuch thattðxÞ ¼eiyxd for allxAWðr1Þand we have proved that fðx1;0;. . .;0Þ ¼eiyx1d for allx1 AWðr1Þ. Com- paring this equality with (3.5) we infer that u is a holomorphic map which takes values iniyþ2piZonWðr1Þ f0gand hence, setting pðxÞ ¼uðxÞ uðpffiffiffiffir1
;0;. . .;0Þ, we obtain that pðx1;0;. . .;0Þ ¼0 for allx1AWðr1Þ.
In order to prove that qxqp
j
Wðr1Þf0g10 for all j¼2;. . .;n1, it is enough to con- sider the map Wðr1;yjÞCðx1;xjÞ 7! fðx1;0;. . .;0;xj;0;. . .;0ÞAWðr2Þ and hence we can limit ourselves to the casen1¼2.
Since for anyx1AWðr1Þand anyR<rjy12j=2pðsinðplnjx1j=lnr1ÞÞ1=2the setfx1g DR
is contained inW1, we can find holomorphic maps a1:Wðr1Þ !Canda2 :W1!C such that pðxÞ ¼a1ðx1Þx2þa2ðxÞx22. Our last assumption is therefore equivalent to a110. As f maps W1 in Wðr2Þ, using the form of f we have that for all xAW1 the following inequality holds:
lnr2dlnjx1j<<pðxÞ<dlnjx1j: ð3:7Þ Fix x10AWðr1Þ and for any 0<e<sinðplnjx01j=lnr1Þ denote by R the number rjy12j=2pðsinðplnjx01j=lnr1Þ eÞÞ1=2. Then for allx2ADR the pointx¼ ðx10;x2Þbelongs toW1. Now let us setr¼R=2; then we have
rja1ðx10Þjcr max
jx2jcrja1ðx10Þ þa2ðxÞx2j ¼r max
jx2j¼rja1ðx10Þ þa2ðxÞx2j ¼max
jx2j¼rjpðxÞj: The Borel–Carathe´odory theorem and (3.7) imply that
jxmax2j¼rjpðxÞjc R Rr max
jx2j¼r<pðxÞ ¼2 max
jx2j¼r<pðxÞc2dlnjx10j;
and therefore ja1ðx01Þjc4rjy1 2j=2pdlnjx10jðsinðplnjx10j=lnr1Þ eÞ1=2; the arbitrari- ness ofeyields that for allx01 AWðr1Þthe following inequality holds
ja1ðx10Þjc 4rjy1 2j=2pdlnjx10j ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi sinðplnjx10j=lnr1Þ
q : ð3:8Þ
It is easily seen that the right hand term of (3.8) goes to 0 whenjx01j !1, yielding limjx0
1j!1ja1ðx10Þj ¼0. The same reasoning can be performed for jx10j !rþ1; also in this case we obtain that limjx0
1j!rþ1ja1ðx10Þj ¼0. Then we can invoke the maximum
modulus principle and we are done. r
Now we consider the behaviour of the last n21 components of f on the annulus Wðr1Þ f0g.
Proposition 3.9. Let f AHd where jdj ¼lnr2=lnr1; then fjðx1;0;. . .;0Þ ¼0 for all x1AWðr1Þand all j¼2;. . .;n2.
Proof.First of all note that it is enough to prove the assertion forn1¼1 andn2¼2;
in fact for any jAf2;. . .;n2gconsider the map
g:Wðr1ÞCx1 7! ðf1ðx1;0;. . .;0Þ;fjðx1;0;. . .;0ÞÞAWðr2;QjÞ;
it is obvious that f andghave the same degree, thus we can limit ourselves to study the casen1¼1,n2¼2. Moreover, as in the previous theorem, we perform the proof ford >0 only, leaving the proof of the cased <0 to the reader.
Theorem 3.8 tells us thatjf1ðxÞj ¼ jxjd; since f mapsWðr1ÞintoWðr2;QjÞwe have that for allxAWðr1Þthe following inequality holds
jf2ðxÞj2jxjdQj=p¼ jf2ðxÞj2jf1ðxÞjQj=p<sinðplnjf1ðxÞj=lnr2Þ ¼sinðplnjxj=lnr1Þ; that is
jf2ðxÞj2<jxjdQj=psinðplnjxj=lnr1Þ ð3:9Þ for all xAWðr1Þ. Taking the limit of the right hand side both for jxj !1 and jxj !rþ1 we obtain that
lim
jxj!rþ1jf2ðxÞj2¼ lim
jxj!1jf2ðxÞj2¼0;
the maximum modulus principle yields the conclusion. r The big di¤erence inHd between the one-dimensional and the multi-dimensional case arises just whenjdj ¼lnr2=lnr1. In the one-dimensional caseHdis a one-dimensional topological space which is isomorphic to S1, while in the multidimensional case it contains an open set in an infinite-dimensional Fre´chet space. In fact, provided n1þn2d3, the following proposition gives a ‘‘large’’ family of maps belonging to
Hd whered ¼lnr2=lnr1 (and of course the same can be done, mutatis mutandis, for d ¼ lnr2=lnr1).
Since the domainWðr;y2;. . .;yn1Þis biholomorphic toWðr;y2þ2pl;. . .;yn1þ2plÞ for alllAZ(see [4] for a proof ), in order to simplify computations we can suppose thatyjc0 for all j¼2;. . .;n1andQjd0 for all j¼2;. . .;n2. Now set
K¼r2lnr2
p and K1¼lnð1þK=ðn11ÞÞ
d if n1d2:
For any yAR, sjk AHolðW1;CÞ for j;k¼2;. . .;n1 and hjk AHolðW1;CÞ for j¼ 2;. . .;n1,k¼2;. . .;n2, settðxÞ ¼Pn1
j;k¼2xjxksjkðxÞ.
Theorem 3.10.If Xn1
j;k¼2
jsjkðxÞjcK1 and nX1;n2
j;k¼2
jhjkðxÞj2c1þpðn11ÞK1
lnr1 p2K12
2 ln2r1; ð3:10Þ then the map
f :W1Cx7!
eiyx1dedtðxÞ;Xn1
j¼2
xjhj2ðxÞ;. . .;Xn1
j¼2
xjhjn2ðxÞ
AW2
belongs toHd where d ¼lnr2=lnr1.
Proof.By definition the map f is obviously holomorphic. Now we prove that f maps W1intoW2; after that a simple remark will show that the degree of f is equal todand therefore f belongs toHd. First of all, we give two estimates oftwhich will be useful in the sequel. By the definition oftwe havejtðxÞjcPn1
j;k¼2jxjxksjkðxÞj; sinceykc0 for k¼2;. . .;n1, then we obtain jxkj<1 for all xAW1 and for k¼2;. . .;n1, and therefore
jtðxÞjcXn1
j;k¼2
jsjkðxÞjcK1 ð3:11Þ
for allxAW1. Moreover for allxAW1the following bound ontalso holds jtðxÞjc Xn1
j;k¼2
jxjj jxkj Xn1
j;k¼2
jsjkðxÞj
cK1Xn1
j¼2
jxjj 2
cðn11ÞK1
Xn1
j¼2
jxjj2cðn11ÞK1
Xn1
j¼2
jxjj2jx1jyj=p
cðn11ÞK1sin plnjx1j lnr1
: ð3:12Þ
Let us notice that f1ðxÞcan be written aseiyx1dþeiyx1dqðxÞ, where
qðxÞ ¼edtðxÞ1¼X
l>0
ðdtðxÞÞl l! : Now we estimateqðxÞ: by (3.11) and (3.12) we obtain
jqðxÞjcX
l>0
jdtðxÞjl
l! cdjtðxÞj X
l>0
ðdK1Þl1 l!
!
cdðn11ÞK1
X
l>0
ðdK1Þl1 l!
!
sin plnjx1j lnr1
cðn11ÞðedK11Þsin plnjx1j lnr1
¼Ksin plnjx1j lnr1
:
As for anyxAW1 we havejx1jc1, the following inequalities yield jx1jdKsin plnjx1j
lnr1
cjf1ðxÞjcjx1jdþKsin plnjx1j lnr1
: ð3:13Þ
Now consider the functionsF;C:½r1;1 !Rgiven by FðtÞ ¼tdKsin plnt
lnr1
; CðtÞ ¼tdþKsin plnt lnr1
;
it is easily seen thatFðr1Þ ¼Cðr1Þ ¼r2, thatFð1Þ ¼Cð1Þ ¼1 and that both of them are increasing (in fact their derivatives on½r1;1are always positive due to the choice ofK). Then (3.13) implies thatr2 <jf1ðxÞj<1 for allxAW1.
In order to prove that f maps W1 to W2, we have to check the second condition, namely thatPn2
j¼2jfjðxÞj2jf1ðxÞjQj=p<sin plnjlnfr1ðxÞj
2
for allxAW1. The definition of f1
and the relationdlnr1¼lnr2entail sin plnjf1ðxÞj
lnr2
¼sin plnjx1dedtðxÞj lnr2
!
¼sin plnjx1j lnr1
cos p<tðxÞ lnr1
þcos plnjx1j lnr1
sin p<tðxÞ lnr1
dsin plnjx1j lnr1
1p2jtðxÞj2 2 ln2r1
!
þpjtðxÞj lnr1 ; by (3.11) and (3.12) we obtain that
sin plnjf1ðxÞj lnr2
dsin plnjx1j lnr1
1 p2K12
2 ln2r1þpðn11ÞK1
lnr1
: ð3:14Þ
SinceQk is non-negative fork¼2;. . .;n2, then for allxAW1we havejf1ðxÞjQk=pc1 fork¼2;. . .;n2, therefore the second inequality in (3.10) yields that for allxAW1 the following chain of inequalities holds
Xn2
k¼2
jfkðxÞj2jf1ðxÞjQk=pcXn2
k¼2
jfkðxÞj2 ¼Xn2
k¼2
Xn1
j¼2
xjhjkðxÞ
2
cXn2
k¼2
Xn1
j¼2
jxjj2Xn1
j¼2
jhjkðxÞj2
cXn1
j¼2
jxjj2jx1jyj=p Xn1;n2
j;k¼2
jhjkðxÞj2
<sin plnjx1j lnr1
1þpðn11ÞK1
lnr1
p2K12 2 ln2r1
; ð3:15Þ
together with (3.14) this ensures that fðW1Þis contained inW2. Now consider the map
H:½0;1 W1Cx7!
eiyx1dedttðxÞ;tXn1
j¼2
xjhj2ðxÞ;. . .;tXn1
j¼2
xjhjn2ðxÞ
ACn2;
since for anytA½0;1the inequalities contained in (3.10) are both satisfied, we obtain that H maps ½0;1 W1 intoW2. MoreoverHðt;Þis holomorphic for any tA½0;1, and Hð1;Þ ¼ f, while it is easily seen that Hð0;Þ has degree d and therefore
degðfÞ ¼d, which concludes the proof. r
Even if in the general case the bounds given by (3.10) can be non-optimal, there is at least one case in which they are optimal. IfW1¼Wðr1;0ÞandW2¼Wðr2;0Þthen the following corollary holds.
Corollary 3.11.For anyyARand hAHolðW1;DÞthe map f :W1Cx7! ðeiyx1d;x2hðxÞÞAW2
belongs to Hd where d¼lnr2=lnr1. Vice versa, for any map f in Hd of the form fðxÞ ¼ ðf1ðx1Þ;f2ðxÞÞ there exist yAR and hAHolðW1;DÞ such that the equality fðxÞ ¼ ðeiyx1d;x2hðxÞÞholds for allxAW1.
Proof.Since the su‰ciency of the condition can be obtained by direct computation as in the proof of the previous theorem, we are left to prove its necessity. By Theorem 3.8 there existsyARsuch that f1ðx1;0Þ ¼eiyx1d; as f1does not depend onx2we have that f1ðxÞ ¼eiyx1d for any xAW1. By Proposition 3.9 there exists a holomorphic
functionh:W1 !Csuch that f2ðxÞ ¼x2hðxÞ. Since fðW1ÞHW2 we obtain that for allxAW1
jx2hðxÞj2<sin plnjeiyx1dj lnr2
!
¼sin plnjx1j lnr1
:
Ifx0¼ ðx10;x02ÞAW1choosee>0 such thatjx02j2csinðplnjx10j=lnr1Þ eand setR¼ ðsinðplnjx01j=lnr1Þ eÞ1=2; then the following chain of inequalities holds
jhðx0Þj2c max
jx2jcRjhðx10;x2Þj2¼max
jx2j¼Rjhðx01;x2Þj2¼R2max
jx2j¼Rjf2ðx10;x2Þj2
¼maxjx2j¼Rjf2ðx10;x2Þj2
sinðplnjx10j=lnr1Þ ec sinðplnjx10j=lnr1Þ sinðplnjx10j=lnr1Þ e:
Lettingego to 0 we obtain thatjhðx0Þjc1 and then we are done. r 4 Complex geodesics for generalized annuli
In this section we prove some results on complex geodesics in generalized annuli.
Definition 4.1. Given x;zAD an extremal map j through x and z is a holomor- phic map j:D!D for which there exist t;sAD such that jðtÞ ¼x, jðsÞ ¼zand kDðx;zÞ ¼kDðt;sÞ. Acomplex geodesicfor the domainDis a holomorphic isometry j:D!Dwith respect to the Kobayashi distance ofDandD(that is, a holomorphic map which is extremal through any point of its image).
Analogous definitions can be given replacing the Kobayashi distance with the Kobayashi metric: in this case we speak of aninfinitesimal extremal mapand of an infinitesimal complex geodesic. Recall that for anyx;zAWand for anyz0Aw1ðxÞ, w0Aw1ðzÞwe have
kWðx;zÞ ¼inffkBnðz0;wwÞ~ : ~wwAw1ðzÞg ¼inffkBnðz0;g0jðw0ÞÞ: jAZg; ð4:1Þ this equality yields both the existence of extremal maps through any couple of points in generalized annuli and a characterization of complex geodesics which will be use- ful in order to solve some problems concerning existence of complex geodesics inW.
Remark 4.2.For anyx;zAWthere exists an extremal map throughxandz.
Proof. Consider the covering ðBn!w WÞ: since Bn is complete hyperbolic we can choose z;wABn such that wðzÞ ¼x, wðwÞ ¼z and kWðx;zÞ ¼kBnðz;wÞ. Let j~
j:D!Bn be a complex geodesic throughzandw(jj~does exist sinceBnis a strictly convex bounded domain inCn) and setj¼wjj. Setting~ t¼jj~1ðzÞands¼jj~1ðwÞ, it is easily seen thatjis an extremal map throughxandz. r
Proposition 4.3. Let jj~:D!Bn be a complex geodesic in Bn; then the holomorphic mapj¼wjj~:D!Wis a complex geodesic inWi¤
kBnðjjðtÞ;~ jjðsÞÞ ¼~ inffkBnðjjðtÞ;~ g0jðjjðsÞÞÞ~ : jAZg ð4:2Þ for any t;sAD.Vice versa,ifjis a complex geodesic inWthen any liftingjj~ofjtoBn is a complex geodesic inBn for which(4.2)holds for any t;sAD.
Proof.Ifjj~:D!Bnis a complex geodesic for which (4.2) holds for anyt;sAD, then (4.1) and (4.2) imply thatkDðt;sÞ ¼kWðwjjðtÞ;~ wjjðsÞÞ~ for allt;sADand therefore j¼wjj~is a complex geodesic inW.
Vice versa, ifj:D!Wis a complex geodesic, thenkDðt;sÞ ¼kWðjðtÞ;jðsÞÞholds for anyt;sAD. Fixs0AD, choosea0ABnsuch thatwða0Þ ¼jðs0Þand letjj~:D!Bn be the lifting ofj through a0, i.e. the unique holomorphic map fromDto Bn such that j¼wjj~andjjðs~ 0Þ ¼a0. Since the Kobayashi distance is contracted by holo- morphic maps, we then have
kDðt;sÞ ¼kWðjðtÞ;jðsÞÞ ¼kWðwjjðtÞ;~ wjjðsÞÞ~ ckBnðjjðtÞ;~ jjðsÞÞ~ ckDðt;sÞ and therefore equality holds at each t;sAD. Equation (4.1) implies that (4.2) holds
for anyt;sADand this concludes the proof. r
It is well known that there exist no complex geodesics in the annuli WðrÞ: this state- ment can be generalized to any couple of points belonging toWðrÞ f0gHW.
Proposition 4.4.For anyx1;z1AWðrÞwithx10z1there are no complex geodesics inW throughx¼ ðx1;0;. . .;0Þandz¼ ðz1;0;. . .;0Þ.For anyx1AWðrÞand v1ACthere are no infinitesimal complex geodesics in Wthroughx¼ ðx1;0;. . .;0Þwith tangent vector v¼ ðv1;0;. . .;0Þ.
Proof. We perform the proof in the case of complex geodesics, the case of infinites- imal complex geodesics is analogous and is left to the reader.
Suppose j is a complex geodesic through x and zand let jj~be a lifting of j to Bn; by Proposition 4.3 the mapjj~is a complex geodesic inBn. The form ofwimplies that w1ðWðrÞ f0gÞ ¼D f0g, and hence jjðDÞ~ intersects D f0g in two distinct pointsz¼ ðz1;0;. . .;0ÞAw1ðxÞandw¼ ðw1;0;. . .;0ÞAw1ðzÞ. Since the image of a complex geodesic inBn is an a‰ne subset of Bn, i.e. the intersection of Bn with an a‰ne line, we havejjðDÞ ¼~ D f0g, and thereforejðDÞ ¼wð~jjðDÞÞ ¼WðrÞ f0g. As WðrÞ f0gis a holomorphic retract ofW, we obtain that the map
j^
j:DCt7!j1ðtÞAWðrÞ
is a complex geodesic inWðrÞand this is a contradiction. r As we already noticed, complex geodesics in W are projections on W of complex
geodesics in Bn for which (4.2) holds for any t;sAD. To simplify computations, which are very long in general, we will focus our attention on complex geodesics in Wpassing through the pointP0¼ ðpffiffir
;0;. . .;0Þ, i.e. complex geodesics inBn passing through the origin. Up to holomorphic automorphisms of Dwe can therefore sup- pose thatjð0Þ ¼0; then, since complex geodesics inBnpassing through the origin of Bn are given by maps of the formDCt7!tpfor any pAqBn, we are led to investi- gate the following question, which is equivalent to the existence of a complex geode- sic inWpassing through the pointP0and with tangent vectordw0ðpÞatP0:
Does the equality
kDðs;tÞ ¼inffkBnðg0jðspÞ;tpÞ: jAZg ð4:3Þ hold for anyt;sAD?
Denote by h;i the standard Hermitian product in Cn and for any aABnnf0g definePa;Qa:Cn!CnandsaARby
PaðzÞ ¼hz;ai
ha;aia; QaðzÞ ¼zPaðzÞ; sa¼ ð1 kak2Þ1=2 and considerga:Bn!Cngiven by
gaðzÞ ¼aPaðzÞ saQaðzÞ 1hz;ai :
Thengais an involution in AutBn which mapsato the origin and
1 kgaðzÞk2¼ ð1 kak2Þð1 kzk2Þj1hz;aij2 ð4:4Þ holds for anyzABn (for a proof see [1] p. 152–153).
Let gtp be the involution defined above which mapstp to the origin; since tanh is increasing, by developing computations and by (4.4) we obtain that (4.3) is equiva- lent to
1 jsj2
j1tsj2 d 1 kg0jðspÞk2
j1htp;g0jðspÞij2¼ 1 kg0jðspÞk2 j1hg0jðspÞ;tpij2 for anyt;sADand any jAZ.
Setting cj¼coshðjTÞ, sj¼sinhðjTÞ, p0¼ ðp2;. . .;pnÞ, W ¼diag½eiy2;. . .;eiyn and developing computations, we obtain the following question which is again equi- valent to the existence of a complex geodesic inWpassing through the pointP0and with tangent vectordw0ðpÞatP0:
Does
j1tsj2cjcjþsjsp1hðcjsp1þsj;Wjðsp0ÞÞ;tpij2 ð4:5Þ hold for anyt;sADand any jAZ?
A first, very simple algebraic remark again stresses the fact thatp0cannot be equal to zero. In fact, if p0¼0, takings¼ p1AqDandt¼ s, by continuity (4.5) implies 2c2ðcjsjÞfor all jAZ, which is impossible sinceT >0.
Remark 4.5.Inequality (4.5) holds for anyt;sADand any jAZif and only if it holds for anyt;sAqDand any jAZ.
Proof. If j¼0, it is obvious that (4.5) is an equality for any t;sAD and there is nothing to prove. Now suppose that (4.5) holds for anyt;sAqDand any jAZ. First of all we prove thatcjþsjsp1hðcjsp1þsj;Wjðsp0ÞÞ;tpi 00 for allt;sADand for all j00. In fact, if cjþsjsp1hðcjsp1þsj;Wjðsp0ÞÞ;tpi¼0 for some t;sADand some jAZnf0g, then we obtain thathg0jðspÞ;tpi¼1, and thereforetp¼g0jðspÞand jtj ¼ kg0jðspÞk ¼1 which implies t;sAqD. Then the fact that (4.5) holds for any t;sAqDgives 1ts¼0, that is t¼s. Since j00, the unique fixed points of g0j are Ge1, and hence sp¼Ge1, that is p0¼0, which is a contradiction to the previous
remark. So, for any jAZnf0g, the holomorphic maps hj:DDCðt;sÞ 7! 1ts
cjþsjsp1hðcjsp1þsj;Wjðsp0ÞÞ;tpi AC
extend continuously to the boundary; ifjhjðt;sÞjc1 on the Shıˇlov boundary of the bidisk, thenjhjðt;sÞjc1 for anyt;sADand any jAZnf0g; this implies (4.5) for any t;sADand any jAZ. The other implication is trivial by continuity. r Then we are led to investigate on the following question: for which pAqBn does
j1tsj2cjcjþsjsp1hðcjsp1þsj;Wjðsp0ÞÞ;tpij2 ð4:6Þ hold for anyt;sAqDand any jAZ?
To simplify notation, we denote byqj the quantityhp0;Wjp0iand obtain j1tsj2cjcjtsðcjjp1j2þqjÞ þsjðsp1tp1Þj2;
settingz¼ts(which belongs toqDif bothtandsdo) we get j1zj2cjcjþsjsp1zðcjjp1j2þqjþsjsp1Þj2
for all z;sAqD and jAZ. A simple computation proves that the above inequality holds for anyz;sAqDand jAZif and only if
2þ2jðcjjp1j2þqjþsjsp1Þðcjþsjsp1Þ 1jcjcjþsjsp1j2þ jcjjp1j2þqjþsjsp1j2 for allsAqDand jAZ. SettingGj¼cjjp1j2þqj we get
2þ2jcjGj1þsjðcjþGjÞsp1þs2jðsp1Þ2jcjcjþsjsp1j2þ jGjþsjsp1j2