23.11.2012 Views

Discrete Holomorphic Local Dynamical Systems

Discrete Holomorphic Local Dynamical Systems

Discrete Holomorphic Local Dynamical Systems

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

196 Tien-Cuong Dinh and Nessim Sibony<br />

We will use the following version of a lemma due to Briend-Duval [BD2]. Their<br />

proof uses the theory of moduli of annuli.<br />

Lemma 1.55. Let g : Δr → P k be a holomorphic map from a disc of center 0 and<br />

of radius r in C. Assume that area(g(Δr)) counted with multiplicity, is smaller than<br />

1/2. Then for any ε > 0 there is a constant λ > 0 independent of g,r such that the<br />

diameter of g(Δ λ r) is smaller than ε � area(g(Δr)).<br />

Proof. Observe that the lemma is an easy consequence of the Cauchy formula if<br />

g has values in a compact set of C k ⊂ P k . In order to reduce the problem to this<br />

case, it is enough to prove that given an ε0 > 0, there is a constant λ0 > 0such<br />

that diam(g(Δ λ0r)) ≤ ε0. Forε0 small enough, we can apply the above case to g<br />

restricted to Δ λ0r.<br />

By hypothesis, the graphs Γg of g in Δr × P k have bounded area. So, according to<br />

Bishop’s theorem [BS], these graphs form a relatively compact family of analytic<br />

sets, that is, the limits of these graphs in the Hausdorff sense, are analytic sets.<br />

Since area(g(Δr)) is bounded by 1/2, the limits have no compact components.<br />

So, they are also graphs and the family of the maps g is compact. We deduce that<br />

diam(g(Δ λ0r)) ≤ ε0 for λ0 small enough. ⊓⊔<br />

Sketch of the proof of Proposition 1.51. The last assertion in the proposition is<br />

deduced from the first one and Proposition 1.46 applied to a generic point in B. We<br />

obtain that �μ ′ − μ�≤2 √ ν for every ν strictly larger than ν(R,a) which implies<br />

the result.<br />

For the first assertion, the idea is to construct inverse branches for many discs<br />

passing through a and then to apply Theorem 1.54 in order to construct inverse<br />

branches on balls. We can assume that ν is smaller than 1. Choose constants δ > 0,<br />

ε > 0 small enough and then a constant κ > 0 large enough; all independent of n.<br />

Fix now the integer n. Recall that �( f n )∗(ωFS)� = d (k−1)n . By Lemmas 1.52 and<br />

1.53, there is a family F ′ ⊂ F and a constant r > 0 such that L (F ′ ) > 1 − δ<br />

and for any Δ in F ′ , the mass of R ∧ [Δ κ 2 r ] is smaller than ν and the mass of<br />

( f n )∗(ωFS) ∧ [Δκr] is smaller than Ad (k−1)n with A > 0.<br />

Claim. For each Δ in F ′ , f n admits at least (1 − 2ν)d kn inverse branches<br />

gi : Δ κ 2 r → Vi with area(Vi) ≤ Aν −1 d −n . The inverse branches gi can be extended<br />

to a neighbourhood of Δ κ 2 r .<br />

Assuming the claim, we complete the proof of the proposition. Let a1,...,al be<br />

the points in f −n (a), with l ≤ d kn ,andF ′ s ⊂ F ′ the family of Δ’s such that one of<br />

the previous inverse branches gi : Δ κ 2 r → Vi passes through as, thatis,Vi contains<br />

as. The above claim implies that ∑L (F ′ s) ≥ (1 − δ)(1 −2ν)d kn . There are at most<br />

d kn terms in this sum. We only consider the family S of the indices s such that<br />

L (F ′ s) ≥ 1 − 3 √ ν.SinceL (F ′ s) ≤ 1foreverys, wehave<br />

#S +(d kn − #S )(1 − 3 √ ν) ≥ ∑L (F ′ s ) ≥ (1 − δ)(1 − 2ν)dkn .<br />

Therefore, since δ is small, we have #S ≥ (1 − √ ν)d kn . For any index s ∈ S<br />

and for Δ in F ′ s , by Lemma 1.55, the corresponding inverse branch on Δκr, which

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!