Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Boundary behavior of the iterates of a self-map of the unit disc
Manuel D. Contreras
Departamento de Matem ´atica Aplicada II and IMUS Universidad de Sevilla
First Workshop in Complex Analysis and Operator Theory M ´alaga 2016
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Introduction.
Let M be a complex manifold (the unit disc D, the complex plane C, the Riemann sphere C∞, a half plane, an annulus, ...) and φ : M → M a holomorphic function.
Since φ(M) ⊆ M, we can define the iterates φn= φn-times◦...◦ φ.
One of the goals ofComplex Dynamicsis to analyze the behavior of the sequence of iterates {φn}.
Many classical authors worked in Complex Dynamics: Poincar ´e, Julia, Fatou, Denjoy, Wolff, Carath ´eodory, ... In the unit disc, we could add to the above list the names of Valir ´on, Baker, Pommerenke, Cowen, ...
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Introduction.
Let M be a complex manifold (the unit disc D, the complex plane C, the Riemann sphere C∞, a half plane, an annulus, ...) and φ : M → M a holomorphic function.
Since φ(M) ⊆ M, we can define the iterates φn= φn-times◦...◦ φ.
One of the goals ofComplex Dynamicsis to analyze the behavior of the sequence of iterates {φn}.
Many classical authors worked in Complex Dynamics:
Poincar ´e, Julia, Fatou, Denjoy, Wolff, Carath ´eodory, ...
In the unit disc, we could add to the above list the names of Valir ´on, Baker, Pommerenke, Cowen, ...
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Introduction.
φ :M → M holomorphic.
Remark
Let g : M → N be a biholomorphism. Then the function ψ :=g ◦ φ ◦ g−1:N −→ N
is holomorphic and
ψn =
n-times
(g ◦ φ ◦ g−1) ◦ (g ◦ φ ◦ g−1) ◦ ... ◦ (g ◦ φ ◦ g−1)
=
n-times
g ◦ φ ◦ φ ◦ ... ◦ φ ◦ g−1=g ◦ φn◦ g−1.
Thus, the behavior of {φn} on M is similar to the behavior of {ψn} on N.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Introduction.
Theorem (Uniformization Theorem of Poincar ´e and Koebe) Everysimply connectedRiemann surface is biholomorphic either to the unit disc D, or to the complex plane C, or to the Riemann sphere C∞.
So, in the setting ofsimply connectedRiemann surfaces we can reduce to:
The Riemann sphere C∞: iteration of rational functions.
The complex plane C: iteration of entire functions.
The unit disc D.
Theorem (Montel)
Let F be a family of homomorphic functions on a domain Ω. If there are two points that are omitted by every f ∈ F , then F is a normal family.
If φ : D −→ D, then the family {φn:n ∈ N} is normal.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Introduction.
Theorem (Uniformization Theorem of Poincar ´e and Koebe) Everysimply connectedRiemann surface is biholomorphic either to the unit disc D, or to the complex plane C, or to the Riemann sphere C∞.
So, in the setting ofsimply connectedRiemann surfaces we can reduce to:
The Riemann sphere C∞: iteration of rational functions.
The complex plane C: iteration of entire functions.
The unit disc D.
Theorem (Montel)
Let F be a family of homomorphic functions on a domain Ω. If there are two points that are omitted by every f ∈ F , then F is a normal family.
If φ : D −→ D, then the family {φn:n ∈ N} is normal.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Introduction.
Theorem (Uniformization Theorem of Poincar ´e and Koebe) Everysimply connectedRiemann surface is biholomorphic either to the unit disc D, or to the complex plane C, or to the Riemann sphere C∞.
So, in the setting ofsimply connectedRiemann surfaces we can reduce to:
The Riemann sphere C∞: iteration of rational functions.
The complex plane C: iteration of entire functions.
The unit disc D.
Theorem (Montel)
Let F be a family of homomorphic functions on a domain Ω. If there are two points that are omitted by every f ∈ F , then F is a normal family.
If φ : D −→ D, then the family {φn:n ∈ N} is normal.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Introduction.
Iteration in the Riemann sphere and the in complex plane:
J. Milnor,Dynamics in One Complex Variable,2000.
L. Carleson and T.W. Gamelin,Complex Dynamics, 1993.
N. Fagella and X. Jarque,Iteraci ´on Compleja y Fractales, 2007.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
Examine theautomorphismsof the unit disc Aut(D), that is, Aut(D) := {T : D → D such that T is a biholomorphism}.
For a ∈ D, let Ta: D → C be the holomorphic map defined by Ta(z) := a − z
1 − az.
It is easy to see that Ta(D) = D and Ta−1(z) = Ta(z), that is, Ta
is an automorphism of D. Note also that Ta(a) = 0, Ta(0) = a.
Proposition
Let T ∈ Aut(D). Then there exists θ ∈ R and a ∈ D such that T (z) = eiθTa(z).
In particular, every automorphism of D extends as a homeomorphism from D to D.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
Examine theautomorphismsof the unit disc Aut(D), that is, Aut(D) := {T : D → D such that T is a biholomorphism}.
For a ∈ D, let Ta: D → C be the holomorphic map defined by Ta(z) := a − z
1 − az.
It is easy to see that Ta(D) = D and Ta−1(z) = Ta(z), that is, Ta
is an automorphism of D. Note also that Ta(a) = 0, Ta(0) = a.
Proposition
Let T ∈ Aut(D). Then there exists θ ∈ R and a ∈ D such that T (z) = eiθTa(z).
In particular, every automorphism of D extends as a homeomorphism from D to D.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
Lemma
Let T ∈ Aut(D) \ {idD}. Then T has at least one fixed point in D. Moreover, ifT has no fixed points in D, then it has two fixed points τ, σ ∈ ∂D, possibly τ = σ, such that T0(σ) ·T0(τ ) =1.
Proof.
T ∈ Aut(D) ⇒ T (z) = λ1−aza−z for a a ∈ D and λ ∈ C, |λ| = 1. Then
T (z) = z if and only if az2− (1 + λ)z + λa = 0.
If a = 0, since T 6= idD, then the unique fixed point of T in C is z = 0. If a 6= 0, the above equation has two solutions z1,z2∈ C which satisfy z1z2= λaa ∈ ∂D (in particular, |z1||z2| = 1) and az1+az2=1 + λ. Moreover,
T0(z1)T0(z2) = λ2 (1 − |a|2)2
((1 − az1)(1 − az2))2 =1.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
Lemma
Let T ∈ Aut(D) \ {idD}. Then T has at least one fixed point in D. Moreover, ifT has no fixed points in D, then it has two fixed points τ, σ ∈ ∂D, possibly τ = σ, such that T0(σ) ·T0(τ ) =1.
Proof.
T ∈ Aut(D) ⇒ T (z) = λ1−aza−z for a a ∈ D and λ ∈ C, |λ| = 1.
Then
T (z) = z if and only if az2− (1 + λ)z + λa = 0.
If a = 0, since T 6= idD, then the unique fixed point of T in C is z = 0.
If a 6= 0, the above equation has two solutions z1,z2∈ C which satisfy z1z2= λaa ∈ ∂D (in particular, |z1||z2| = 1) and az1+az2=1 + λ.
Moreover,
T0(z1)T0(z2) = λ2 (1 − |a|2)2
((1 − az1)(1 − az2))2 =1.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
Lemma
Let T ∈ Aut(D) \ {idD}. Then T has at least one fixed point in D. Moreover, ifT has no fixed points in D, then it has two fixed points τ, σ ∈ ∂D, possibly τ = σ, such that T0(σ) ·T0(τ ) =1.
Proof.
T ∈ Aut(D) ⇒ T (z) = λ1−aza−z for a a ∈ D and λ ∈ C, |λ| = 1.
Then
T (z) = z if and only if az2− (1 + λ)z + λa = 0.
If a = 0, since T 6= idD, then the unique fixed point of T in C is z = 0.
If a 6= 0, the above equation has two solutions z1,z2∈ C which satisfy z1z2= λaa ∈ ∂D (in particular, |z1||z2| = 1) and az1+az2=1 + λ.
Moreover,
T0(z1)T0(z2) = λ2 (1 − |a|2)2
((1 − az1)(1 − az2))2 =1.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
Definition
Let T ∈ Aut(D) \ {idD}. Then we say that
1 T isellipticif it has a fixed point in D,
2 T isparabolicif it has a unique fixed point in ∂D,
3 T ishyperbolicif it has two different fixed points in ∂D.
We investigate now the dynamics of an automorphism of D according to the previous classification.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
The elliptic case:
Assume T is an elliptic automorphism with a fixed point τ ∈ D.
The map Tτ◦ T ◦ Tτ is an automorphism that fixes the origin.
So, there exists λ ∈ ∂D s. t. (Tτ ◦ T ◦ Tτ)(z) = λz, for all z ∈ D. The automorphism T is then holomorphically conjugated to a rotation.
Thus, given z ∈ D,
T2(z) = Tτ(λTτ(Tτ(λTτ(z)))) = Tτ(λ2Tτ(z)), ... Tn(z) = Tτ(λTτ(Tτ(λn−1Tτ(z)))) = Tτ(λnTτ(z)). Remark: T0(τ ) = λ.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
The elliptic case:
Assume T is an elliptic automorphism with a fixed point τ ∈ D.
The map Tτ◦ T ◦ Tτ is an automorphism that fixes the origin.
So, there exists λ ∈ ∂D s. t. (Tτ ◦ T ◦ Tτ)(z) = λz, for all z ∈ D.
The automorphism T is then holomorphically conjugated to a rotation.
Thus, given z ∈ D,
T2(z) = Tτ(λTτ(Tτ(λTτ(z)))) = Tτ(λ2Tτ(z)), ... Tn(z) = Tτ(λTτ(Tτ(λn−1Tτ(z)))) = Tτ(λnTτ(z)). Remark: T0(τ ) = λ.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
The elliptic case:
Assume T is an elliptic automorphism with a fixed point τ ∈ D.
The map Tτ◦ T ◦ Tτ is an automorphism that fixes the origin.
So, there exists λ ∈ ∂D s. t. (Tτ ◦ T ◦ Tτ)(z) = λz, for all z ∈ D.
The automorphism T is then holomorphically conjugated to a rotation.
Thus, given z ∈ D,
T2(z) = Tτ(λTτ(Tτ(λTτ(z)))) = Tτ(λ2Tτ(z)), ...
Tn(z) = Tτ(λTτ(Tτ(λn−1Tτ(z)))) = Tτ(λnTτ(z)).
Remark: T0(τ ) = λ.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
The hyperbolic case:
Let τ, σ ∈ ∂D be its fixed points, τ 6= σ.
Since T0(τ )T0(σ) =1, we may assume that |T0(τ )| ≤1.
Consider C(z) = τ +zτ −z, C : D → H, we can conjugate T to an automorphism of the right half plane Φ = C ◦ T ◦ C−1that fixes the point ∞.
Thus Φ(w ) = aw + b, for all w ∈ C.
Since Φ(H) = H we conclude that: Re b = 0, Im a = 0, a > 0. Also, a 6= 1, for otherwise Φ would not have fixed points in C. Now, a direct computation shows that a = T01(τ ). So
T0(τ ) ∈ (0, 1) and T0(σ) = T01(τ ) ∈ (1, +∞).
Moreover, since Φn(w ) = anw +1−a1−anb for all w ∈ H, it follows that {Φn} converges to ∞ uniform on compacta of H. Hence, {Tn} = {C−1◦ Φn◦ C} converges uniformly on compacta to τ .
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
The hyperbolic case:
Let τ, σ ∈ ∂D be its fixed points, τ 6= σ.
Since T0(τ )T0(σ) =1, we may assume that |T0(τ )| ≤1.
Consider C(z) = τ +zτ −z, C : D → H, we can conjugate T to an automorphism of the right half plane Φ = C ◦ T ◦ C−1that fixes the point ∞.
Thus Φ(w ) = aw + b, for all w ∈ C.
Since Φ(H) = H we conclude that: Re b = 0, Im a = 0, a > 0.
Also, a 6= 1, for otherwise Φ would not have fixed points in C. Now, a direct computation shows that a = T01(τ ). So
T0(τ ) ∈ (0, 1) and T0(σ) = T01(τ ) ∈ (1, +∞).
Moreover, since Φn(w ) = anw +1−a1−anb for all w ∈ H, it follows that {Φn} converges to ∞ uniform on compacta of H. Hence, {Tn} = {C−1◦ Φn◦ C} converges uniformly on compacta to τ .
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
The hyperbolic case:
Let τ, σ ∈ ∂D be its fixed points, τ 6= σ.
Since T0(τ )T0(σ) =1, we may assume that |T0(τ )| ≤1.
Consider C(z) = τ +zτ −z, C : D → H, we can conjugate T to an automorphism of the right half plane Φ = C ◦ T ◦ C−1that fixes the point ∞.
Thus Φ(w ) = aw + b, for all w ∈ C.
Since Φ(H) = H we conclude that: Re b = 0, Im a = 0, a > 0.
Also, a 6= 1, for otherwise Φ would not have fixed points in C.
Now, a direct computation shows that a = T01(τ ). So T0(τ ) ∈ (0, 1) and T0(σ) = T01(τ ) ∈ (1, +∞).
Moreover, since Φn(w ) = anw +1−a1−anb for all w ∈ H, it follows that {Φn} converges to ∞ uniform on compacta of H. Hence, {Tn} = {C−1◦ Φn◦ C} converges uniformly on compacta to τ .
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
The hyperbolic case:
Let τ, σ ∈ ∂D be its fixed points, τ 6= σ.
Since T0(τ )T0(σ) =1, we may assume that |T0(τ )| ≤1.
Consider C(z) = τ +zτ −z, C : D → H, we can conjugate T to an automorphism of the right half plane Φ = C ◦ T ◦ C−1that fixes the point ∞.
Thus Φ(w ) = aw + b, for all w ∈ C.
Since Φ(H) = H we conclude that: Re b = 0, Im a = 0, a > 0.
Also, a 6= 1, for otherwise Φ would not have fixed points in C.
Now, a direct computation shows that a = T01(τ ). So T0(τ ) ∈ (0, 1) and T0(σ) = T01(τ ) ∈ (1, +∞).
Moreover, since Φn(w ) = anw +1−a1−anb for all w ∈ H, it follows that {Φn} converges to ∞ uniform on compacta of H. Hence, {Tn} = {C−1◦ Φn◦ C} converges uniformly on compacta to τ .
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
The hyperbolic case:
Let τ, σ ∈ ∂D be its fixed points, τ 6= σ.
Since T0(τ )T0(σ) =1, we may assume that |T0(τ )| ≤1.
Consider C(z) = τ +zτ −z, C : D → H, we can conjugate T to an automorphism of the right half plane Φ = C ◦ T ◦ C−1that fixes the point ∞.
Thus Φ(w ) = aw + b, for all w ∈ C.
Since Φ(H) = H we conclude that: Re b = 0, Im a = 0, a > 0.
Also, a 6= 1, for otherwise Φ would not have fixed points in C.
Now, a direct computation shows that a = T01(τ ). So T0(τ ) ∈ (0, 1) and T0(σ) = T01(τ ) ∈ (1, +∞).
Moreover, since Φn(w ) = anw +1−a1−anb for all w ∈ H, it follows that {Φn} converges to ∞ uniform on compacta of H. Hence, {Tn} = {C−1◦ Φn◦ C} converges uniformly on compacta to τ .
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
The parabolic case:
Assume that T is parabolic, with a unique fixed point τ ∈ ∂D.
Notice that T0(τ )2=1.
Arguing as in the hyperbolic case, the function
Φ(w ) = C ◦ T ◦ C−1(w ) = aw + b is a M ¨obius transformation that has only one fixed point in C∞and Φ(H) = H.
Hence, Φ(w ) = w + b, for all w ∈ H, with a = T01(τ ) >0 and Re b = 0. In particular, T0(τ ) =1.
Thus Φn(w ) = w + nb, which implies Φn(w ) → ∞ for all w ∈ H. Therefore, {Tn} converges to τ for all z ∈ D, and, even in this case, {Tn} converges uniformly on compacta to τ .
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
The parabolic case:
Assume that T is parabolic, with a unique fixed point τ ∈ ∂D.
Notice that T0(τ )2=1.
Arguing as in the hyperbolic case, the function
Φ(w ) = C ◦ T ◦ C−1(w ) = aw + b is a M ¨obius transformation that has only one fixed point in C∞and Φ(H) = H.
Hence, Φ(w ) = w + b, for all w ∈ H, with a = T01(τ ) >0 and Re b = 0. In particular, T0(τ ) =1.
Thus Φn(w ) = w + nb, which implies Φn(w ) → ∞ for all w ∈ H. Therefore, {Tn} converges to τ for all z ∈ D, and, even in this case, {Tn} converges uniformly on compacta to τ .
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
The parabolic case:
Assume that T is parabolic, with a unique fixed point τ ∈ ∂D.
Notice that T0(τ )2=1.
Arguing as in the hyperbolic case, the function
Φ(w ) = C ◦ T ◦ C−1(w ) = aw + b is a M ¨obius transformation that has only one fixed point in C∞and Φ(H) = H.
Hence, Φ(w ) = w + b, for all w ∈ H, with a = T01(τ ) >0 and Re b = 0. In particular, T0(τ ) =1.
Thus Φn(w ) = w + nb, which implies Φn(w ) → ∞ for all w ∈ H.
Therefore, {Tn} converges to τ for all z ∈ D, and, even in this case, {Tn} converges uniformly on compacta to τ .
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The easiest examples: automorphisms of the unit disc.
We have proved:
Conclusion I
Let T ∈ Aut(D) \ {idD}. If T is not elliptic, then there is τ ∈ ∂D such that {Tn} converges uniformly on compacta to the constant map D 3 z 7→ τ .
τ is the fixed point such that T0(τ ) ∈ (0, 1].
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Linear fractional maps.
Let φ : D → D be a linear fractional maps (LFM) given by φ(z) = az + b
cz + d with ad − bc 6= 0.
Then |c| < |d | and φ : D → D.
Theorem (Brouwer’s fixed-point theorem)
Every continuous function from a closed disk to itself has at least one fixed point.
Thus there is τ ∈ D such that φ(τ ) = τ .
We follow a similar argument to study the behavior of {φn}.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Linear fractional maps.
Let φ : D → D be a linear fractional maps (LFM) given by φ(z) = az+bcz+d with ad − bc 6= 0.
Assume that there isτ ∈ Dsuch thatφ(τ ) = τ andφ /∈ Aut(D).
Write again Tτ(z) = 1−τ zτ −z ∈ Aut(D).
The map ψ = Tτ◦ φ ◦ Tτ : D → D is a LFM that fixes the origin, let us say, ψ(z) = βz+γαz . We may assume that γ = 1.
Then
|ψ0(0)| = |α| < 1, ψn(z) = αnz
βwnz + 1, where wn= 1 − αn 1 − α. So {ψn} converges to 0 uniform on compacta of D. Hence, {φn} = {Tτ ◦ ψn◦ Tτ} converges uniformly on compacta to τ . Important: This is not satisfied if φ ∈ Aut(D).
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Linear fractional maps.
Let φ : D → D be a linear fractional maps (LFM) given by φ(z) = az+bcz+d with ad − bc 6= 0.
Assume that there isτ ∈ Dsuch thatφ(τ ) = τ andφ /∈ Aut(D).
Write again Tτ(z) = 1−τ zτ −z ∈ Aut(D).
The map ψ = Tτ◦ φ ◦ Tτ : D → D is a LFM that fixes the origin, let us say, ψ(z) = βz+γαz . We may assume that γ = 1.
Then
|ψ0(0)| = |α| < 1, ψn(z) = αnz
βwnz + 1, where wn= 1 − αn 1 − α.
So {ψn} converges to 0 uniform on compacta of D. Hence, {φn} = {Tτ ◦ ψn◦ Tτ} converges uniformly on compacta to τ . Important: This is not satisfied if φ ∈ Aut(D).
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Linear fractional maps.
Let φ : D → D be a linear fractional maps (LFM) given by φ(z) = az+bcz+d with ad − bc 6= 0.
Assume that there isτ ∈ Dsuch thatφ(τ ) = τ andφ /∈ Aut(D).
Write again Tτ(z) = 1−τ zτ −z ∈ Aut(D).
The map ψ = Tτ◦ φ ◦ Tτ : D → D is a LFM that fixes the origin, let us say, ψ(z) = βz+γαz . We may assume that γ = 1.
Then
|ψ0(0)| = |α| < 1, ψn(z) = αnz
βwnz + 1, where wn= 1 − αn 1 − α. So {ψn} converges to 0 uniform on compacta of D. Hence, {φn} = {Tτ ◦ ψn◦ Tτ} converges uniformly on compacta to τ . Important: This is not satisfied if φ ∈ Aut(D).
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Linear fractional maps.
Let φ : D → D be a linear fractional maps (LFM) given by φ(z) = az+bcz+d with ad − bc 6= 0.
Assume that φ hasno fixed point in D. Then τ ∈ ∂D.
Consider C(z) = τ +zτ −z, C : D → H, we can conjugate φ to an automorphism of the right half plane Φ = C ◦ φ ◦ C−1that fixes the point ∞.
Thus Φ(w ) = αw + β, for all w ∈ C.
, Re β ≥ 0, α > 0. Since Φ(H) ⊆ H we conclude that Re β = Re Φ(0) ≥ 0. Moreover
Re Φ(ic) = c Re (iα)+Re β = −c Im α+Re β ≥ 0, for all c ∈ R. So Im α = 0 and α ∈ R. On the other hand,
Re Φ(r ) = r α + Re β ≥ 0, for all r > 0. Thus α > 0.
If α = 1, then Φn(w ) = w + nβ, for all w ∈ H. If α 6= 1, then Φn(w ) = αnw +1−α1−αnβ for all w ∈ H.
In any case, it follows that {Φn} converges either to ∞ or to 0 uniformly on compacta of H.
Hence, {φn} converges uniformly on compacta of D to the constant map z 7→ σ where σ = τ or σ = C−1(β/(1 − α)). Remark: φ0(τ ) ≤1.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Linear fractional maps.
Let φ : D → D be a linear fractional maps (LFM) given by φ(z) = az+bcz+d with ad − bc 6= 0.
Assume that φ hasno fixed point in D. Then τ ∈ ∂D.
Consider C(z) = τ +zτ −z, C : D → H, we can conjugate φ to an automorphism of the right half plane Φ = C ◦ φ ◦ C−1that fixes the point ∞.
Thus Φ(w ) = αw + β, for all w ∈ C
.
, Re β ≥ 0, α > 0.
Since Φ(H) ⊆ H we conclude that Re β = Re Φ(0) ≥ 0.
Moreover
Re Φ(ic) = c Re (iα)+Re β = −c Im α+Re β ≥ 0, for all c ∈ R.
So Im α = 0 and α ∈ R. On the other hand,
Re Φ(r ) = r α + Re β ≥ 0, for all r > 0.
Thus α > 0.
If α = 1, then Φn(w ) = w + nβ, for all w ∈ H. If α 6= 1, then Φn(w ) = αnw +1−α1−αnβ for all w ∈ H.
In any case, it follows that {Φn} converges either to ∞ or to 0 uniformly on compacta of H.
Hence, {φn} converges uniformly on compacta of D to the constant map z 7→ σ where σ = τ or σ = C−1(β/(1 − α)). Remark: φ0(τ ) ≤1.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Linear fractional maps.
Let φ : D → D be a linear fractional maps (LFM) given by φ(z) = az+bcz+d with ad − bc 6= 0.
Assume that φ hasno fixed point in D. Then τ ∈ ∂D.
Consider C(z) = τ +zτ −z, C : D → H, we can conjugate φ to an automorphism of the right half plane Φ = C ◦ φ ◦ C−1that fixes the point ∞.
Thus Φ(w ) = αw + β, for all w ∈ C
.
, Re β ≥ 0, α > 0.
Since Φ(H) ⊆ H we conclude that Re β = Re Φ(0) ≥ 0. Moreover
Re Φ(ic) = c Re (iα)+Re β = −c Im α+Re β ≥ 0, for all c ∈ R. So Im α = 0 and α ∈ R. On the other hand,
Re Φ(r ) = r α + Re β ≥ 0, for all r > 0. Thus α > 0.
If α = 1, then Φn(w ) = w + nβ, for all w ∈ H.
If α 6= 1, then Φn(w ) = αnw +1−α1−αnβ for all w ∈ H.
In any case, it follows that {Φn} converges either to ∞ or to 0 uniformly on compacta of H.
Hence, {φn} converges uniformly on compacta of D to the constant map z 7→ σ where σ = τ or σ = C−1(β/(1 − α)). Remark: φ0(τ ) ≤1.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Linear fractional maps.
Let φ : D → D be a linear fractional maps (LFM) given by φ(z) = az+bcz+d with ad − bc 6= 0.
Assume that φ hasno fixed point in D. Then τ ∈ ∂D.
Consider C(z) = τ +zτ −z, C : D → H, we can conjugate φ to an automorphism of the right half plane Φ = C ◦ φ ◦ C−1that fixes the point ∞.
Thus Φ(w ) = αw + β, for all w ∈ C
.
, Re β ≥ 0, α > 0.
Since Φ(H) ⊆ H we conclude that Re β = Re Φ(0) ≥ 0. Moreover
Re Φ(ic) = c Re (iα)+Re β = −c Im α+Re β ≥ 0, for all c ∈ R. So Im α = 0 and α ∈ R. On the other hand,
Re Φ(r ) = r α + Re β ≥ 0, for all r > 0. Thus α > 0.
If α = 1, then Φn(w ) = w + nβ, for all w ∈ H.
If α 6= 1, then Φn(w ) = αnw +1−α1−αnβ for all w ∈ H.
In any case, it follows that {Φn} converges either to ∞ or to 0 uniformly on compacta of H.
Hence, {φn} converges uniformly on compacta of D to the constant map z 7→ σ where σ = τ or σ = C−1(β/(1 − α)).
Remark: φ0(τ ) ≤1.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Linear fractional maps.
We have proved:
Conclusion I
Let T ∈ Aut(D) \ {idD}. If T is not elliptic, then there is τ ∈ ∂D such that {Tn} converges uniformly on compacta to the constant map D 3 z 7→ τ .
Conclusion II
Let φ : D → D be a LFM different from an elliptic automorphism.
Then there is τ ∈ D such that {φn} converges uniformly on compacta to the constant map D 3 z 7→ τ .
If τ ∈ ∂D, then φ0(τ ) ∈ (0, 1].
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Other examples
Example
(i) φ : D → D given by φ(z) = z2.Then, φn(z) = z2n y φn(z) → 0 for all z ∈ D.
(ii) φ : D → D given by φ(z) = 1 − (1 − z)α,where 0 < α < 1. Then, φn(z) = 1 − (1 − z)αn and φn(z) → 0 for all z ∈ D. In both cases, 0 is a fixed point of φ.
Example
(iii) φ : D → D given by φ(z) = 1 −(1+(1−z)1−z2)1/2.Then φn(z) = 1 − (1+n(1−z)1−z 2
)1/2 and φn(z) → 1 for all z ∈ D. Notice that 1 is a “fixed point” of φ.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Other examples
Example
(i) φ : D → D given by φ(z) = z2.Then, φn(z) = z2n y φn(z) → 0 for all z ∈ D.
(ii) φ : D → D given by φ(z) = 1 − (1 − z)α,where 0 < α < 1.
Then, φn(z) = 1 − (1 − z)αn and φn(z) → 0 for all z ∈ D.
In both cases, 0 is a fixed point of φ. Example
(iii) φ : D → D given by φ(z) = 1 −(1+(1−z)1−z2)1/2.Then φn(z) = 1 − (1+n(1−z)1−z 2
)1/2 and φn(z) → 1 for all z ∈ D. Notice that 1 is a “fixed point” of φ.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Other examples
Example
(i) φ : D → D given by φ(z) = z2.Then, φn(z) = z2n y φn(z) → 0 for all z ∈ D.
(ii) φ : D → D given by φ(z) = 1 − (1 − z)α,where 0 < α < 1.
Then, φn(z) = 1 − (1 − z)αn and φn(z) → 0 for all z ∈ D.
In both cases, 0 is a fixed point of φ.
Example
(iii) φ : D → D given by φ(z) = 1 −(1+(1−z)1−z2)1/2.Then φn(z) = 1 − (1+n(1−z)1−z 2
)1/2 and φn(z) → 1 for all z ∈ D. Notice that 1 is a “fixed point” of φ.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Other examples
Example
(i) φ : D → D given by φ(z) = z2.Then, φn(z) = z2n y φn(z) → 0 for all z ∈ D.
(ii) φ : D → D given by φ(z) = 1 − (1 − z)α,where 0 < α < 1.
Then, φn(z) = 1 − (1 − z)αn and φn(z) → 0 for all z ∈ D.
In both cases, 0 is a fixed point of φ.
Example
(iii) φ : D → D given by φ(z) = 1 −(1+(1−z)1−z2)1/2.Then φn(z) = 1 − (1+n(1−z)1−z 2
)1/2 and φn(z) → 1 for all z ∈ D.
Notice that 1 is a “fixed point” of φ.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Other examples
Example
(i) φ : D → D given by φ(z) = z2.Then, φn(z) = z2n y φn(z) → 0 for all z ∈ D.
(ii) φ : D → D given by φ(z) = 1 − (1 − z)α,where 0 < α < 1.
Then, φn(z) = 1 − (1 − z)αn and φn(z) → 0 for all z ∈ D.
In both cases, 0 is a fixed point of φ.
Example
(iii) φ : D → D given by φ(z) = 1 −(1+(1−z)1−z2)1/2.Then φn(z) = 1 − (1+n(1−z)1−z 2
)1/2 and φn(z) → 1 for all z ∈ D.
Notice that 1 is a “fixed point” of φ.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Denjoy-Wolff Theorem
Theorem (Denjoy, Wolff, 1926)
Let φ : D → D be holomorphic, not an elliptic automorphism.
Then there exists a unique point τ ∈ D such that the sequence {φn} converges uniformly on compacta of the unit disc to the constant map D 3 z → τ .
Remark
τ ∈ D if and only if φ has a fixed point in D. In such case, φ(τ ) = τ .
Definition
The point τ is called theDenjoy-Wolff pointof φ.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Denjoy-Wolff Theorem
Theorem (Denjoy, Wolff, 1926)
Let φ : D → D be holomorphic, not an elliptic automorphism.
Then there exists a unique point τ ∈ D such that the sequence {φn} converges uniformly on compacta of the unit disc to the constant map D 3 z → τ .
Remark
τ ∈ D if and only if φ has a fixed point in D.
In such case, φ(τ ) = τ . Definition
The point τ is called theDenjoy-Wolff pointof φ.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Example in another domain: Newton’s method
In numerical analysis, Newton’s method is a method for finding successively better approximations to the roots (or zeroes) of a real-valued function.
The method starts with a function f defined over R and an initial guess x0for a root of the equation f (x ) = 0. If the function satisfies a certain assumptions about f0 and the initial guess is close enough, then a better approximation x1is given by
x1=x0− f (x0) f0(x0). The process is repeated as
xn+1=xn− f (xn) f0(xn). until a sufficiently accurate value is reached.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Example in another domain: Newton’s method
Let us see what happens with the roots of z2− 1 = 0. Applying Newton’s method we generate the sequence
zn+1 =zn− f (zn) f0(zn) = 1
2
zn+ 1
zn
, where f (z) = z2− 1.
Define φ(z) = 12 z + 1z .
To analyze the convergence of Newton’s method is equivalent to characterize the values z such that {φn(z)} converges. Problem:We have to find a domain Ω, Ω 6= C, s. t. φ : Ω → Ω. Since φ : H −→ H and φ(1) = 1, applying Denjoy-Wolff
Theorem to the simply connected domain H (biholomorphic to D) and the function φ that fixes a point in H, we obtain
if z ∈ H, then {φn(z)} converges to 1.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Example in another domain: Newton’s method
Let us see what happens with the roots of z2− 1 = 0. Applying Newton’s method we generate the sequence
zn+1 =zn− f (zn) f0(zn) = 1
2
zn+ 1
zn
, where f (z) = z2− 1.
Define φ(z) = 12 z + 1z .
To analyze the convergence of Newton’s method is equivalent to characterize the values z such that {φn(z)} converges.
Problem:We have to find a domain Ω, Ω 6= C, s. t. φ : Ω → Ω. Since φ : H −→ H and φ(1) = 1, applying Denjoy-Wolff
Theorem to the simply connected domain H (biholomorphic to D) and the function φ that fixes a point in H, we obtain
if z ∈ H, then {φn(z)} converges to 1.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Example in another domain: Newton’s method
Let us see what happens with the roots of z2− 1 = 0. Applying Newton’s method we generate the sequence
zn+1 =zn− f (zn) f0(zn) = 1
2
zn+ 1
zn
, where f (z) = z2− 1.
Define φ(z) = 12 z + 1z .
To analyze the convergence of Newton’s method is equivalent to characterize the values z such that {φn(z)} converges.
Problem:We have to find a domain Ω, Ω 6= C, s. t. φ : Ω → Ω.
Since φ : H −→ H and φ(1) = 1, applying Denjoy-Wolff
Theorem to the simply connected domain H (biholomorphic to D) and the function φ that fixes a point in H, we obtain
if z ∈ H, then {φn(z)} converges to 1.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Example in another domain: Newton’s method
Let us see what happens with the roots of z2− 1 = 0. Applying Newton’s method we generate the sequence
zn+1 =zn− f (zn) f0(zn) = 1
2
zn+ 1
zn
, where f (z) = z2− 1.
Define φ(z) = 12 z + 1z .
To analyze the convergence of Newton’s method is equivalent to characterize the values z such that {φn(z)} converges.
Problem:We have to find a domain Ω, Ω 6= C, s. t. φ : Ω → Ω.
Since φ : H −→ H and φ(1) = 1, applying Denjoy-Wolff
Theorem to the simply connected domain H (biholomorphic to D) and the function φ that fixes a point in H, we obtain
if z ∈ H, then {φn(z)} converges to 1.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Example in another domain: Newton’s method
Let us see what happens with the roots of z2− 1 = 0. Applying Newton’s method we generate the sequence
zn+1 =zn− f (zn) f0(zn) = 1
2
zn+ 1
zn
, where f (z) = z2− 1.
Define φ(z) = 12 z + 1z .
If Re z > 0, then {φn(z)} converges to 1.
If Re z < 0, then {φn(z)} converges to −1.
What happens with the sequence {φn(z)} if Re z = 0?
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Example in another domain: Newton’s method
Let us see what happens with the roots of z2− 1 = 0. Applying Newton’s method we generate the sequence
zn+1 =zn− f (zn) f0(zn) = 1
2
zn+ 1
zn
, where f (z) = z2− 1.
Define φ(z) = 12 z + 1z .
If Re z > 0, then {φn(z)} converges to 1.
If Re z < 0, then {φn(z)} converges to −1.
What happens with the sequence {φn(z)} if Re z = 0?
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Boundary behavior of the iterates.
What happens with the sequence {φn(ξ)}when ξ ∈ ∂D?
Does the sequence {φn(ξ)}make sense when ξ ∈ ∂D?
Theorem (Fatou)
Let φ : D → D be a holomorphic function. Then the redial limit φ(ξ) := lim
r →1φ(r ξ) ∈ D exists a.e. ξ ∈ ∂D. Moreover, the function
φ : ∂D −→ D
ξ 7−→ φ(ξ) is measurable.
φis calledinnerif φ(ξ) ∈ ∂D a.e. ξ ∈ ∂D.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Boundary behavior of the iterates.
What happens with the sequence {φn(ξ)}when ξ ∈ ∂D?
Does the sequence {φn(ξ)}make sense when ξ ∈ ∂D?
Theorem (Fatou)
Let φ : D → D be a holomorphic function. Then the redial limit φ(ξ) := lim
r →1φ(r ξ) ∈ D exists a.e. ξ ∈ ∂D. Moreover, the function
φ : ∂D −→ D
ξ 7−→ φ(ξ) is measurable.
φis calledinnerif φ(ξ) ∈ ∂D a.e. ξ ∈ ∂D.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Boundary behavior of the iterates.
So, there exists a set A of measure 0 such that φn(ξ) := lim
r →1φn(r ξ) exists for all ξ ∈ ∂D \ A and for all n ∈ N.
Therefore, we can reformulate our question:
Problem
Let φ : D → D be a holomorphic function with Denjoy-Wolff point τ ∈ D.
Is it satisfied that {φn(ξ)}converges to τ a.e. ξ ∈ ∂D?
NO.
Example: φ(z) = z2. Its Denjoy-Wolff point is 0. If ξ ∈ ∂D, then |φn(ξ)| =1 (φis inner).
So if ξ ∈ ∂D, then {φn(ξ)}does not converge to 0.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Boundary behavior of the iterates.
So, there exists a set A of measure 0 such that φn(ξ) := lim
r →1φn(r ξ) exists for all ξ ∈ ∂D \ A and for all n ∈ N.
Therefore, we can reformulate our question:
Problem
Let φ : D → D be a holomorphic function with Denjoy-Wolff point τ ∈ D.
Is it satisfied that {φn(ξ)}converges to τ a.e. ξ ∈ ∂D?
NO.
Example: φ(z) = z2. Its Denjoy-Wolff point is 0.
If ξ ∈ ∂D, then |φn(ξ)| =1 (φis inner).
So if ξ ∈ ∂D, then {φn(ξ)}does not converge to 0.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Boundary behavior of the iterates.
Theorem (Bourdon-Matache-Shapiro, Poggi-Corradini (2005)) Let φ : D → D be holomorphic with Denjoy-Wolff point τ ∈ D.
Then (φn(ξ))converges to τ a.e. ξ ∈ ∂D if and only if φ is not an inner function.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Hyperbolic distance.
Thehyperbolic distanceis defined by ρ(z, w ) := 1
2log 1 +
z−w 1−zw
1 −
z−w 1−zw
for all z, w ∈ D.
Theorem (Schwarz’s Lemma)
Let φ : D → D be holomorphic. Assume that φ(0) = 0. Then
|φ(z)| ≤ |z| for all z ∈ D and |φ0(0)| ≤ 1.
Theorem (Schwarz-Pick’s Lemma)
Let φ : D → D be holomorphic. Then, for all z, w ∈ D,
φ(z) − φ(w ) 1 − φ(z)φ(w )
≤
z − w 1 − zw
.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Hyperbolic distance.
Thehyperbolic distanceis defined by
ρ(z, w ) := 1 2log
1 +
z−w 1−zw
1 −
z−w 1−zw
for all z, w ∈ D.
Theorem (Schwarz-Pick’s Lemma)
Let φ : D → D be holomorphic. Then, for all z, w ∈ D,
φ(z) − φ(w ) 1 − φ(z)φ(w )
≤
z − w 1 − zw
.
Since the map x 7→ 12log1+x1−x is non-decreasing we obtain ρ(φ(z), φ(w )) ≤ ρ(z, w ) for all z, w ∈ D.
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
Hyperbolic distance.
Therefore Proposition
Let φ ∈ Hol(D, D) be holomorphic
ρ(φ(z), φ(w )) ≤ ρ(z, w ) for all z, w ∈ D.
Equality holds for some z 6= w if and only if φ ∈ Aut(D).
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The hyperbolic step.
Let φ : D → D be holomorphic, z0∈ D y zn= φn(z0):
ρ(zn,zn+1) = ρ(φ(zn−1), φ(zn)) ≤ ρ(zn−1,zn).
Thus, the sequence {ρ(zn,zn+1)}is decreasing. So, there exists limnρ(zn,zn+1) ∈ [0, +∞). Some examples: τ =Denjoy-Wolff point.
- If τ ∈ D and φ is not an elliptic automorphism, then limnρ(zn,zn+1) = ρ(τ, τ ) =0. - If φ ∈ Aut(D), then ρ(zn,zn+1) = ρ(zn−1,zn)for all n. So, if z06= τ then limnρ(zn,zn+1) >0.
- If φ is a LFM with τ ∈ ∂D and φ0(τ ) <1, then limnρ(zn,zn+1) >0. - If φ is a LFM with τ ∈ ∂D and φ0(τ ) =1, then
limnρ(zn,zn+1) >0 if and only if φ ∈ Aut(D).
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The hyperbolic step.
Let φ : D → D be holomorphic, z0∈ D y zn= φn(z0):
ρ(zn,zn+1) = ρ(φ(zn−1), φ(zn)) ≤ ρ(zn−1,zn).
Thus, the sequence {ρ(zn,zn+1)}is decreasing. So, there exists limnρ(zn,zn+1) ∈ [0, +∞).
Some examples: τ =Denjoy-Wolff point.
- If τ ∈ D and φ is not an elliptic automorphism, then limnρ(zn,zn+1) = ρ(τ, τ ) =0. - If φ ∈ Aut(D), then ρ(zn,zn+1) = ρ(zn−1,zn)for all n. So, if z06= τ then limnρ(zn,zn+1) >0.
- If φ is a LFM with τ ∈ ∂D and φ0(τ ) <1, then limnρ(zn,zn+1) >0. - If φ is a LFM with τ ∈ ∂D and φ0(τ ) =1, then
limnρ(zn,zn+1) >0 if and only if φ ∈ Aut(D).
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The hyperbolic step.
Let φ : D → D be holomorphic, z0∈ D y zn= φn(z0):
ρ(zn,zn+1) = ρ(φ(zn−1), φ(zn)) ≤ ρ(zn−1,zn).
Thus, the sequence {ρ(zn,zn+1)}is decreasing. So, there exists limnρ(zn,zn+1) ∈ [0, +∞).
Some examples: τ =Denjoy-Wolff point.
- If τ ∈ D and φ is not an elliptic automorphism, then limnρ(zn,zn+1) = ρ(τ, τ ) =0.
- If φ ∈ Aut(D), then ρ(zn,zn+1) = ρ(zn−1,zn)for all n.
So, if z06= τ then limnρ(zn,zn+1) >0.
- If φ is a LFM with τ ∈ ∂D and φ0(τ ) <1, then limnρ(zn,zn+1) >0. - If φ is a LFM with τ ∈ ∂D and φ0(τ ) =1, then
limnρ(zn,zn+1) >0 if and only if φ ∈ Aut(D).
Introduction Automorphisms LFM Denjoy-Wolff Theorem Boundary behavior
The hyperbolic step.
Let φ : D → D be holomorphic, z0∈ D y zn= φn(z0):
ρ(zn,zn+1) = ρ(φ(zn−1), φ(zn)) ≤ ρ(zn−1,zn).
Thus, the sequence {ρ(zn,zn+1)}is decreasing. So, there exists limnρ(zn,zn+1) ∈ [0, +∞).
Some examples: τ =Denjoy-Wolff point.
- If τ ∈ D and φ is not an elliptic automorphism, then limnρ(zn,zn+1) = ρ(τ, τ ) =0.
- If φ ∈ Aut(D), then ρ(zn,zn+1) = ρ(zn−1,zn)for all n.
So, if z06= τ then limnρ(zn,zn+1) >0.
- If φ is a LFM with τ ∈ ∂D and φ0(τ ) <1, then limnρ(zn,zn+1) >0.
- If φ is a LFM with τ ∈ ∂D and φ0(τ ) =1, then
limnρ(zn,zn+1) >0 if and only if φ ∈ Aut(D).