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.

Dynamics in Several Complex variables 241<br />

where the points in f −1 (z) are counted with multiplicity. Since f is a ramified<br />

covering, Λ(ϕ) is continuous when ϕ is continuous. If ν is a probability measure<br />

on V, define the measure f ∗ (ν) by<br />

〈 f ∗ (ν),ϕ〉 := 〈ν, f∗(ϕ)〉.<br />

This is a positive measure of mass dt supported on f −1 (supp(ν)). Observe that the<br />

operator ν ↦→ d −1<br />

t f ∗ (ν) is continuous on positive measures, see Exercise A.11.<br />

Theorem 2.11. Let f : U → V be a polynomial-like map as above. Let ν be a prob-<br />

ability measure supported on V which is defined by an L 1 form. Then d −n<br />

t ( f n ) ∗ (ν)<br />

converge to a probability measure μ which does not depend on ν. Forϕ p.s.h. on a<br />

neighbourhood of the filled Julia set K , we have 〈d −n<br />

t ( f n ) ∗ (ν),ϕ〉→〈μ,ϕ〉. The<br />

measure μ is supported on the boundary of K and is totally invariant: d −1<br />

t f ∗ (μ)=<br />

f∗(μ)=μ. Moreover, if Λ is the Perron-Frobenius operator associated to f and ϕ<br />

is a p.s.h. function on a neighbourhood of K ,thenΛn (ϕ) converge to 〈μ,ϕ〉.<br />

Note that in general 〈μ,ϕ〉 may be −∞. If〈μ,ϕ〉 = −∞, the above convergence<br />

means that Λ n (ϕ) tend locally uniformly to −∞; otherwise, the convergence is in<br />

L p<br />

loc for 1 ≤ p < +∞, see Appendix A.2. The above result still holds for measures ν<br />

which have no mass on pluripolar sets. The proof in that case is more delicate. We<br />

have the following lemma.<br />

Lemma 2.12. If ϕ is p.s.h. on a neighbourhood of K ,thenΛ n (ϕ) converge to a<br />

constant cϕ in R ∪{−∞}.<br />

Proof. Observe that Λ n (ϕ) is defined on V for n large enough. It is not difficult<br />

to check that these functions are p.s.h. Indeed, when ϕ is a continu-<br />

ous p.s.h. function, Λ n (ϕ) is a continuous function, see Exercise A.11, and<br />

dd c Λ n (ϕ)=d −n<br />

t ( f n )∗(dd c ϕ) ≥ 0. So, Λ n (ϕ) is p.s.h. The general case is obtained<br />

using an approximation of ϕ by a decreasing sequence of smooth p.s.h. functions.<br />

Consider ψ the upper semi-continuous regularization of limsupΛ n (ϕ). Wededuce<br />

from Proposition A.20 that ψ is a p.s.h. function. We first prove that ψ is constant.<br />

Assume not. By maximum principle, there is a constant δ such that sup U ψ <<br />

δ < sup V ψ. By Hartogs’ lemma A.20,forn large enough, we have Λ n (ϕ) < δ on U.<br />

Since the fibers of f are contained in U, we deduce from the definition of Λ that<br />

sup<br />

V<br />

Λ n+1 (ϕ)=sup<br />

V<br />

Λ(Λ n (ϕ)) ≤ supΛ<br />

U<br />

n (ϕ) < δ.<br />

This implies that ψ ≤ δ which contradicts the choice of δ. Soψ is constant.<br />

Denote by cϕ this constant. If cϕ = −∞, it is clear that Λ n (ϕ) converge to −∞<br />

uniformly on compact sets. Assume that cϕ is finite and Λ ni(ϕ) does not converge to<br />

cϕ for some sequence (ni). By Hartogs’ lemma, we have Λ ni(ϕ) ≤ cϕ − ε for some<br />

constant ε > 0andfori large enough. We deduce as above that Λ n (ϕ) ≤ cϕ − ε for<br />

n ≥ ni. This contradicts the definition of cϕ. ⊓⊔

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

Saved successfully!

Ooh no, something went wrong!