Puzzle for rational maps Pascale Rsch Institut of Mathematics of - - PowerPoint PPT Presentation

puzzle for rational maps
SMART_READER_LITE
LIVE PREVIEW

Puzzle for rational maps Pascale Rsch Institut of Mathematics of - - PowerPoint PPT Presentation

Puzzle for rational maps Pascale Rsch Institut of Mathematics of Toulouse 2019 Rsch P. (IMT) TCD2019 2019 1 / 72 Overview A puzzle associated to a map f : X X is a collection P of puzzle pieces satisfying certain properties.


slide-1
SLIDE 1

Puzzle for rational maps

Pascale Rœsch Institut of Mathematics of Toulouse

2019

Rœsch P. (IMT) TCD2019 2019 1 / 72

slide-2
SLIDE 2

Overview

A puzzle associated to a map f : X → X is a collection P of puzzle pieces satisfying certain properties.

Rœsch P. (IMT) TCD2019 2019 2 / 72

slide-3
SLIDE 3

Overview

A puzzle associated to a map f : X → X is a collection P of puzzle pieces satisfying certain properties.

Rœsch P. (IMT) TCD2019 2019 2 / 72

slide-4
SLIDE 4

P is a collection of jigsaw puzzles of any level : P = {P0, .....Pn, ....};

Rœsch P. (IMT) TCD2019 2019 3 / 72

slide-5
SLIDE 5

P is a collection of jigsaw puzzles of any level : P = {P0, .....Pn, ....}; Each jigsaw puzzles Pn is a collection of puzzle pieces defining a partition of X: X =

  • P∈Pn

P and P ∩ Q = ∅ for P = Q, P, Q ∈ Pn.

Rœsch P. (IMT) TCD2019 2019 3 / 72

slide-6
SLIDE 6

P is a collection of jigsaw puzzles of any level : P = {P0, .....Pn, ....}; Each jigsaw puzzles Pn is a collection of puzzle pieces defining a partition of X: X =

  • P∈Pn

P and P ∩ Q = ∅ for P = Q, P, Q ∈ Pn. P ∈ Pn = ⇒ ∃Q ∈ Pn−1, P ⊂ Q Q is unique. The idea is to get a more complicate jigsaw puzzle at each next level (with more puzzle pieces)

Rœsch P. (IMT) TCD2019 2019 3 / 72

slide-7
SLIDE 7

P is a collection of jigsaw puzzles of any level : P = {P0, .....Pn, ....}; Each jigsaw puzzles Pn is a collection of puzzle pieces defining a partition of X: X =

  • P∈Pn

P and P ∩ Q = ∅ for P = Q, P, Q ∈ Pn. P ∈ Pn = ⇒ ∃Q ∈ Pn−1, P ⊂ Q Q is unique. The idea is to get a more complicate jigsaw puzzle at each next level (with more puzzle pieces) the map f acts on the puzzle : P ∈ Pn = ⇒ f (P) ∈ Pn−1

Rœsch P. (IMT) TCD2019 2019 3 / 72

slide-8
SLIDE 8

The goal of such jigsaw puzzle is to give to any point of the space X a precise address (depending on the level n) which is compatible with the dynamics.

Rœsch P. (IMT) TCD2019 2019 4 / 72

slide-9
SLIDE 9

The goal of such jigsaw puzzle is to give to any point of the space X a precise address (depending on the level n) which is compatible with the dynamics. Any point x is included in some P with P ∈ Pn for each level n ∈ N. If x ∈ ∂P then P is not unique.

Rœsch P. (IMT) TCD2019 2019 4 / 72

slide-10
SLIDE 10

The goal of such jigsaw puzzle is to give to any point of the space X a precise address (depending on the level n) which is compatible with the dynamics. Any point x is included in some P with P ∈ Pn for each level n ∈ N. If x ∈ ∂P then P is not unique. On some subset of X, one can define Pn(x) as the unique puzzle piece of level containing x P0(x) ⊃ P1(x) ⊃ ... ⊃ Pn(x) ⊃ Pn+1(x) ⊃ .... ∋ x is the sequence of decreasing puzzle pieces containing x

Rœsch P. (IMT) TCD2019 2019 4 / 72

slide-11
SLIDE 11

The goal of such jigsaw puzzle is to give to any point of the space X a precise address (depending on the level n) which is compatible with the dynamics. Any point x is included in some P with P ∈ Pn for each level n ∈ N. If x ∈ ∂P then P is not unique. On some subset of X, one can define Pn(x) as the unique puzzle piece of level containing x P0(x) ⊃ P1(x) ⊃ ... ⊃ Pn(x) ⊃ Pn+1(x) ⊃ .... ∋ x is the sequence of decreasing puzzle pieces containing x P0(f (x)) = f (P1(x)) ⊃ P1(f (x)) = f (P2(x)) ⊃ ... ⊃ Pn−1(f (x)) = f (Pn(x)) is the sequence of decreasing puzzle pieces containing f (x)

Rœsch P. (IMT) TCD2019 2019 4 / 72

slide-12
SLIDE 12

If the diameter of the sequence of puzzle pieces shrinks to 0 P0(x) ⊃ P1(x) ⊃ ... ⊃ Pn(x) ⊃ Pn+1(x) ⊃ .... ∋ x then it gives a precise location of the point x.

Rœsch P. (IMT) TCD2019 2019 5 / 72

slide-13
SLIDE 13

Example

The tent map is defined on [0, 1] by T(x) =

  • 2x

if 0 ≤ x ≤ 1/2 2(1 − x) if 1/2 ≤ x ≤ 1

Rœsch P. (IMT) TCD2019 2019 6 / 72

slide-14
SLIDE 14

Example

The tent map is defined on [0, 1] by T(x) =

  • 2x

if 0 ≤ x ≤ 1/2 2(1 − x) if 1/2 ≤ x ≤ 1 P0 = {]0, 1[}, P1 = {]0, 1/2[, ]1/2, 1[}, P2 = {]0, 1/4[, ]1/4, 1/2[, ]1/2, 3/4[, ]3/4, 1[}... A basic method of studying its dynamics is to find a symbolic representation: an encoding of the points by sequences of symbols such that the map T becomes the shift map.

Rœsch P. (IMT) TCD2019 2019 6 / 72

slide-15
SLIDE 15

Puzzles are the analogous of a Markov partition for hyperbolic systems. Sinaï and Bowen used Markov partition to describe uniformly hyperbolic systems.

Rœsch P. (IMT) TCD2019 2019 7 / 72

slide-16
SLIDE 16

A way to define a puzzle is by cutting the space with a graph Γ

Rœsch P. (IMT) TCD2019 2019 8 / 72

slide-17
SLIDE 17

A way to define a puzzle is by cutting the space with a graph Γ Puzzle pieces are obtained by pull back by the dynamics of an starting partition

Rœsch P. (IMT) TCD2019 2019 8 / 72

slide-18
SLIDE 18

This way of understanding the dynamics using coding was already used in the study of hyperbolic systems. The presence of dilatation, contraction but also bending does not allow to build a general theory. It will be more a collection of examples.

Rœsch P. (IMT) TCD2019 2019 9 / 72

slide-19
SLIDE 19

Branner-Hubbard puzzle, cubic polynomials

Rœsch P. (IMT) TCD2019 2019 10 / 72

slide-20
SLIDE 20

Branner-Hubbard puzzle, cubic polynomials

Rœsch P. (IMT) TCD2019 2019 10 / 72

slide-21
SLIDE 21

Branner-Hubbard puzzle, cubic polynomials

Rœsch P. (IMT) TCD2019 2019 10 / 72

slide-22
SLIDE 22

Branner-Hubbard puzzle, cubic polynomials

Theorem (B-H)

For a cubic polynomial f with one critical point in K(f ), the Julia set K(f ) is a Cantor set if and only if the critical components of K(f ) are aperiodic.

Rœsch P. (IMT) TCD2019 2019 10 / 72

slide-23
SLIDE 23

Branner-Hubbard Puzzle

Let f be monic of degree 3. f is conjugated to z → z3 near ∞.

Rœsch P. (IMT) TCD2019 2019 11 / 72

slide-24
SLIDE 24

Branner-Hubbard Puzzle

Let f be monic of degree 3. f is conjugated to z → z3 near ∞. Let Γ0 be an equipotential containing the critical value

Rœsch P. (IMT) TCD2019 2019 11 / 72

slide-25
SLIDE 25

Branner-Hubbard Puzzle

Let f be monic of degree 3. f is conjugated to z → z3 near ∞. Let Γ0 be an equipotential containing the critical value Let Γ1 = f −1(Γ0) figure height curve

Rœsch P. (IMT) TCD2019 2019 11 / 72

slide-26
SLIDE 26

Branner-Hubbard Puzzle

Let f be monic of degree 3. f is conjugated to z → z3 near ∞. Let Γ0 be an equipotential containing the critical value Let Γ1 = f −1(Γ0) figure height curve ...let Γn+1 = f −1(Γn) for n ≥ 0

Rœsch P. (IMT) TCD2019 2019 11 / 72

slide-27
SLIDE 27

Branner-Hubbard Puzzle

Let f be monic of degree 3. f is conjugated to z → z3 near ∞. Let Γ0 be an equipotential containing the critical value Let Γ1 = f −1(Γ0) figure height curve ...let Γn+1 = f −1(Γn) for n ≥ 0 A piece of puzzle of level n is any bounded connected component of C \ Γn

Rœsch P. (IMT) TCD2019 2019 11 / 72

slide-28
SLIDE 28

Branner-Hubbard Puzzle

Let f be monic of degree 3. f is conjugated to z → z3 near ∞. Let Γ0 be an equipotential containing the critical value Let Γ1 = f −1(Γ0) figure height curve ...let Γn+1 = f −1(Γn) for n ≥ 0 A piece of puzzle of level n is any bounded connected component of C \ Γn

Puzzle pieces are topological disks .

Rœsch P. (IMT) TCD2019 2019 11 / 72

slide-29
SLIDE 29

Branner-Hubbard Puzzle

Let f be monic of degree 3. f is conjugated to z → z3 near ∞. Let Γ0 be an equipotential containing the critical value Let Γ1 = f −1(Γ0) figure height curve ...let Γn+1 = f −1(Γn) for n ≥ 0 A piece of puzzle of level n is any bounded connected component of C \ Γn

Puzzle pieces are topological disks .

For x ∈ K(f ), let Pn(x) be the puzzle piece containing x f (Pn+1(x)) = Pn(f (x))

Rœsch P. (IMT) TCD2019 2019 11 / 72

slide-30
SLIDE 30

Branner-Hubbard Puzzle

Let f be monic of degree 3. f is conjugated to z → z3 near ∞. Let Γ0 be an equipotential containing the critical value Let Γ1 = f −1(Γ0) figure height curve ...let Γn+1 = f −1(Γn) for n ≥ 0 A piece of puzzle of level n is any bounded connected component of C \ Γn

Puzzle pieces are topological disks .

For x ∈ K(f ), let Pn(x) be the puzzle piece containing x f (Pn+1(x)) = Pn(f (x)) f : Pn+1(x) → Pn(f (x)) is a covering of degree at most 2

Rœsch P. (IMT) TCD2019 2019 11 / 72

slide-31
SLIDE 31

Branner-Hubbard Puzzle

Every point x ∈ K(f ) defines a "nest" of puzzle pieces x ∈ Pn+1(x) ⊂ Pn(x) ⊂ · · · ⊂ P1(x) ⊂ P0(x)

Rœsch P. (IMT) TCD2019 2019 12 / 72

slide-32
SLIDE 32

Branner-Hubbard Puzzle

Every point x ∈ K(f ) defines a "nest" of puzzle pieces x ∈ Pn+1(x) ⊂ Pn(x) ⊂ · · · ⊂ P1(x) ⊂ P0(x) Kx the connected component of K(f ) containing x satisfies Kx =

  • n≥0

Pn(x)

Rœsch P. (IMT) TCD2019 2019 12 / 72

slide-33
SLIDE 33

Branner-Hubbard Puzzle

Every point x ∈ K(f ) defines a "nest" of puzzle pieces x ∈ Pn+1(x) ⊂ Pn(x) ⊂ · · · ⊂ P1(x) ⊂ P0(x) Kx the connected component of K(f ) containing x satisfies Kx =

  • n≥0

Pn(x)

Remark

K(f ) is a Cantor set ⇐ ⇒ diam(Pn(x)) → 0 for every x.

Rœsch P. (IMT) TCD2019 2019 12 / 72

slide-34
SLIDE 34

Branner-Hubbard Puzzle

Every point x ∈ K(f ) defines a "nest" of puzzle pieces x ∈ Pn+1(x) ⊂ Pn(x) ⊂ · · · ⊂ P1(x) ⊂ P0(x) Kx the connected component of K(f ) containing x satisfies Kx =

  • n≥0

Pn(x)

Remark

K(f ) is a Cantor set ⇐ ⇒ diam(Pn(x)) → 0 for every x. Kx is k-periodic ⇐ ⇒ the nest is k-periodic : f k(Pn+k(x)) = Pn(x) for n ≥ n0.

Rœsch P. (IMT) TCD2019 2019 12 / 72

slide-35
SLIDE 35

Branner-Hubbard Tableaux

The dynamics can be read on the diagonal of the tableaux P0(x) P1(x) P2(x) P3(x) . . . . . . Pn(x) Pn+1(x) . . .

Rœsch P. (IMT) TCD2019 2019 13 / 72

slide-36
SLIDE 36

Branner-Hubbard Tableaux

The dynamics can be read on the diagonal of the tableaux P0(x) P1(x) P2(x) P3(x) . . . . . . Pn(x) Pn+1(x) . . .

fր fր fր

. . . . . .

fր fր

Rœsch P. (IMT) TCD2019 2019 13 / 72

slide-37
SLIDE 37

Branner-Hubbard Tableaux

The dynamics can be read on the diagonal of the tableaux P0(x) P1(x) P2(x) P3(x) . . . . . . Pn(x) Pn+1(x) . . .

fր fր fր

. . . . . .

fր fր

P0(f (x)) P1(f (x)) P2(f (x)) . . . . . . Pn−1(f (x)) Pn(f (x)) Pn+1(f (x)) . . .

Rœsch P. (IMT) TCD2019 2019 13 / 72

slide-38
SLIDE 38

Branner-Hubbard Tableaux

The dynamics can be read on the diagonal of the tableaux P0(x) P1(x) P2(x) P3(x) . . . . . . Pn(x) Pn+1(x) . . .

fր fր fր

. . . . . .

fր fր

P0(f (x)) P1(f (x)) P2(f (x)) . . . . . . Pn−1(f (x)) Pn(f (x)) Pn+1(f (x)) . . .

fր fր

. . . . . .

fր fր fր

Rœsch P. (IMT) TCD2019 2019 13 / 72

slide-39
SLIDE 39

Branner-Hubbard Tableaux

The dynamics can be read on the diagonal of the tableaux P0(x) P1(x) P2(x) P3(x) . . . . . . Pn(x) Pn+1(x) . . .

fր fր fր

. . . . . .

fր fր

P0(f (x)) P1(f (x)) P2(f (x)) . . . . . . Pn−1(f (x)) Pn(f (x)) Pn+1(f (x)) . . .

fր fր

. . . . . .

fր fր fր

P0(f 2(x)) P1(f 2(x)) . . . . . . Pn−2(f 2(x)) Pn−1(f 2(x)) Pn(f 2(x)) . . . . . .

Rœsch P. (IMT) TCD2019 2019 13 / 72

slide-40
SLIDE 40

Some Analysis

To prove that Kx = Pn(x) is reduced to {x} one needs to understand this combinatorics and the following analysis.

1 The modulus of an annulus A estimates its "size", it is a conformal

invariant and mod (DR \ D1) =

1 2π log(R) ;

Rœsch P. (IMT) TCD2019 2019 14 / 72

slide-41
SLIDE 41

Some Analysis

To prove that Kx = Pn(x) is reduced to {x} one needs to understand this combinatorics and the following analysis.

1 The modulus of an annulus A estimates its "size", it is a conformal

invariant and mod (DR \ D1) =

1 2π log(R) ;

2 If an annulus D \ K has infinite modulus then K is one point ; Rœsch P. (IMT) TCD2019 2019 14 / 72

slide-42
SLIDE 42

Some Analysis

To prove that Kx = Pn(x) is reduced to {x} one needs to understand this combinatorics and the following analysis.

1 The modulus of an annulus A estimates its "size", it is a conformal

invariant and mod (DR \ D1) =

1 2π log(R) ;

2 If an annulus D \ K has infinite modulus then K is one point ; Rœsch P. (IMT) TCD2019 2019 14 / 72

slide-43
SLIDE 43

1 Consider the annuli An(x) = Pn(x) \ Pn+1(x) which are disjoint,

essential in P0(x) \ Kx ;

Rœsch P. (IMT) TCD2019 2019 15 / 72

slide-44
SLIDE 44

1 Consider the annuli An(x) = Pn(x) \ Pn+1(x) which are disjoint,

essential in P0(x) \ Kx ;

2 Grötzsch inequality : mod(P0(x) \ Kx) ≥ mod(An(x)) ; Rœsch P. (IMT) TCD2019 2019 15 / 72

slide-45
SLIDE 45

1 Consider the annuli An(x) = Pn(x) \ Pn+1(x) which are disjoint,

essential in P0(x) \ Kx ;

2 Grötzsch inequality : mod(P0(x) \ Kx) ≥ mod(An(x)) ; 3 it is enough to prove that mod(An(x)) = ∞. Rœsch P. (IMT) TCD2019 2019 15 / 72

slide-46
SLIDE 46

Generally f (An+1(x)) = An(f (x)) for An(x) = Pn(x) \ Pn+1(x) but It is critical if Pn+1(x) contains the critical point and mod (An(x)) = 1

2 mod (An−1(f (x))).

  • Rœsch P. (IMT)

TCD2019 2019 16 / 72

slide-47
SLIDE 47

Generally f (An+1(x)) = An(f (x)) for An(x) = Pn(x) \ Pn+1(x) but It is critical if Pn+1(x) contains the critical point and mod (An(x)) = 1

2 mod (An−1(f (x))).

  • Rœsch P. (IMT)

TCD2019 2019 16 / 72

slide-48
SLIDE 48

Generally f (An+1(x)) = An(f (x)) for An(x) = Pn(x) \ Pn+1(x) but It is critical if Pn+1(x) contains the critical point and mod (An(x)) = 1

2 mod (An−1(f (x))).

  • It is semi-critical if An(x) contains the critical point and

mod (An(x)) ≥ 1

2 mod (An−1(f (x))).

  • Rœsch P. (IMT)

TCD2019 2019 16 / 72

slide-49
SLIDE 49

Generally f (An+1(x)) = An(f (x)) for An(x) = Pn(x) \ Pn+1(x) but It is critical if Pn+1(x) contains the critical point and mod (An(x)) = 1

2 mod (An−1(f (x))).

  • It is semi-critical if An(x) contains the critical point and

mod (An(x)) ≥ 1

2 mod (An−1(f (x))).

  • Rœsch P. (IMT)

TCD2019 2019 16 / 72

slide-50
SLIDE 50

Generally f (An+1(x)) = An(f (x)) for An(x) = Pn(x) \ Pn+1(x) but It is critical if Pn+1(x) contains the critical point and mod (An(x)) = 1

2 mod (An−1(f (x))).

  • It is semi-critical if An(x) contains the critical point and

mod (An(x)) ≥ 1

2 mod (An−1(f (x))).

  • It is non critical if Pn(x) contains no critical point and

mod (An(x)) = mod (An−1(f (x))).

Rœsch P. (IMT) TCD2019 2019 16 / 72

slide-51
SLIDE 51

Two important properties : Pn+1(x) ⊂ Pn(x) Pn(x) is a topological disk So that there is a non degenerate annulus Pn(x) \ Pn+1(x). Remark : If Kx is l-periodic then f l : Pn+l(x) → Pn(f l(x)) = Pn(x) is a covering of degree at most 2. If the degree is 2 then f l : Pn+l(x) → Pn(f l(x)) = Pn(x) is a polynomial like map of degree 2 so conjugate to some z2 + c if the degree is 1 then Kx = {x} is periodic.

Rœsch P. (IMT) TCD2019 2019 17 / 72

slide-52
SLIDE 52

Theorem (McMullen)

For a cubic polynomial f with Cantor Julia set, the Lebesque measure of J(f ) is zero.

Rœsch P. (IMT) TCD2019 2019 18 / 72

slide-53
SLIDE 53

If the puzzle pieces / graph does not separate the Julia set then we just get Kx = K(f ) So the graph used has to cut the Julia set in two pieces at least. Cut it properly i.e. in one point

Rœsch P. (IMT) TCD2019 2019 19 / 72

slide-54
SLIDE 54

Let f : U → V with U ⊂ V Define a puzzle for the map f by a finite connected graph Γ ⊂ U satisfying f (Γ) ∩ U ⊂ U the forward orbits of critical points are disjoint from Γ The puzzle pieces of level n are the connected components of f −n(U \ Γ) intersecting K(f ) = f −n(U).

Rœsch P. (IMT) TCD2019 2019 20 / 72

slide-55
SLIDE 55

Yoccoz puzzles for quadratic polynomials

Rœsch P. (IMT) TCD2019 2019 21 / 72

slide-56
SLIDE 56

Yoccoz puzzles for quadratic polynomials

Rœsch P. (IMT) TCD2019 2019 21 / 72

slide-57
SLIDE 57

Yoccoz puzzles for quadratic polynomials

Rœsch P. (IMT) TCD2019 2019 21 / 72

slide-58
SLIDE 58

Yoccoz puzzles for quadratic polynomials

Rœsch P. (IMT) TCD2019 2019 21 / 72

slide-59
SLIDE 59

Yoccoz puzzles for quadratic polynomials

Yoccoz Theorem : The map is renormalizable or the impression of puzzle pieces is one point

Rœsch P. (IMT) TCD2019 2019 21 / 72

slide-60
SLIDE 60

Siegel disks

Carsten Petersen constructed a puzzle piece for Siegel disk working on the Blaschke model. Petersen, Petersen-Zakeri : Most Siegel Julia sets are locally connected

Rœsch P. (IMT) TCD2019 2019 22 / 72

slide-61
SLIDE 61

Higher degree polynomials

Those constructions have been generalized for many cubic polynomials and in higher degree for polynomials . The goal is to prove that

  • n∈N Pn(x) = {x}
  • r the map is renormalizable,

then find another puzzle for the renormalized map. There are several critical points and the degree is no more 2. New tools : develop the combinatorics by constructing particular nest called KSS-nest use analytic tools like Kahn-Lyubich covering Lemma.

Rœsch P. (IMT) TCD2019 2019 23 / 72

slide-62
SLIDE 62

Results that can be proved using puzzles

  • ne can get that some components of the Julia set are points, or

copies of Julia sets by getting renormalization domains (B-H)

Rœsch P. (IMT) TCD2019 2019 24 / 72

slide-63
SLIDE 63

Results that can be proved using puzzles

  • ne can get that some components of the Julia set are points, or

copies of Julia sets by getting renormalization domains (B-H)

  • ne can get local connectivity of a set X, where X is a Julia set, the

boundary of a Fatou component or parts of M.

Rœsch P. (IMT) TCD2019 2019 24 / 72

slide-64
SLIDE 64

Results that can be proved using puzzles

  • ne can get that some components of the Julia set are points, or

copies of Julia sets by getting renormalization domains (B-H)

  • ne can get local connectivity of a set X, where X is a Julia set, the

boundary of a Fatou component or parts of M. Xn(x) = Pn(x) ∩ X is a basis of connected neighbourhoods.

Rœsch P. (IMT) TCD2019 2019 24 / 72

slide-65
SLIDE 65

Results that can be proved using puzzles

  • ne can get that some components of the Julia set are points, or

copies of Julia sets by getting renormalization domains (B-H)

  • ne can get local connectivity of a set X, where X is a Julia set, the

boundary of a Fatou component or parts of M. Xn(x) = Pn(x) ∩ X is a basis of connected neighbourhoods.

  • ne can get measure 0 of the Julia set or parts of it based on

McMullen inequality area(Pn+1) ≤

area(Pn) 1+4π mod (An).

Rœsch P. (IMT) TCD2019 2019 24 / 72

slide-66
SLIDE 66

Results that can be proved using puzzles

  • ne can get that some components of the Julia set are points, or

copies of Julia sets by getting renormalization domains (B-H)

  • ne can get local connectivity of a set X, where X is a Julia set, the

boundary of a Fatou component or parts of M. Xn(x) = Pn(x) ∩ X is a basis of connected neighbourhoods.

  • ne can get measure 0 of the Julia set or parts of it based on

McMullen inequality area(Pn+1) ≤

area(Pn) 1+4π mod (An).

  • ne can get Rigidity: similar puzzles leads to combinatorially conjugacy

that can be promoted QC or conformal using analytic tools .

Rœsch P. (IMT) TCD2019 2019 24 / 72

slide-67
SLIDE 67

Results that can be proved using puzzles

  • ne can get that some components of the Julia set are points, or

copies of Julia sets by getting renormalization domains (B-H)

  • ne can get local connectivity of a set X, where X is a Julia set, the

boundary of a Fatou component or parts of M. Xn(x) = Pn(x) ∩ X is a basis of connected neighbourhoods.

  • ne can get measure 0 of the Julia set or parts of it based on

McMullen inequality area(Pn+1) ≤

area(Pn) 1+4π mod (An).

  • ne can get Rigidity: similar puzzles leads to combinatorially conjugacy

that can be promoted QC or conformal using analytic tools .

  • ne can get convergence of an access like external ray, since puzzle

pieces can be used like prime-ends.

Rœsch P. (IMT) TCD2019 2019 24 / 72

slide-68
SLIDE 68

Results that can be proved using puzzles

  • ne can get that some components of the Julia set are points, or

copies of Julia sets by getting renormalization domains (B-H)

  • ne can get local connectivity of a set X, where X is a Julia set, the

boundary of a Fatou component or parts of M. Xn(x) = Pn(x) ∩ X is a basis of connected neighbourhoods.

  • ne can get measure 0 of the Julia set or parts of it based on

McMullen inequality area(Pn+1) ≤

area(Pn) 1+4π mod (An).

  • ne can get Rigidity: similar puzzles leads to combinatorially conjugacy

that can be promoted QC or conformal using analytic tools .

  • ne can get convergence of an access like external ray, since puzzle

pieces can be used like prime-ends.

  • ne can get a description of a rational map as a mating using the

conjugacy given by puzzles.

Rœsch P. (IMT) TCD2019 2019 24 / 72

slide-69
SLIDE 69

Results that can be proved using puzzles

  • ne can get that some components of the Julia set are points, or

copies of Julia sets by getting renormalization domains (B-H)

  • ne can get local connectivity of a set X, where X is a Julia set, the

boundary of a Fatou component or parts of M. Xn(x) = Pn(x) ∩ X is a basis of connected neighbourhoods.

  • ne can get measure 0 of the Julia set or parts of it based on

McMullen inequality area(Pn+1) ≤

area(Pn) 1+4π mod (An).

  • ne can get Rigidity: similar puzzles leads to combinatorially conjugacy

that can be promoted QC or conformal using analytic tools .

  • ne can get convergence of an access like external ray, since puzzle

pieces can be used like prime-ends.

  • ne can get a description of a rational map as a mating using the

conjugacy given by puzzles.

  • ne can get model in parameter space via puzzles in parameter spaces

Rœsch P. (IMT) TCD2019 2019 24 / 72

slide-70
SLIDE 70

Rœsch P. (IMT) TCD2019 2019 25 / 72

slide-71
SLIDE 71

Rational maps

For rational maps there is no equipotential and rays cutting the Julia set like for polynomials Julia set of a rational map is more complicate

Rœsch P. (IMT) TCD2019 2019 26 / 72

slide-72
SLIDE 72

First example : cubic Newton map

. The Newton’s method NP of a polynomial P is defined by NP(z) = z − P(z) P′(z). The roots of P are super-attracting fixed points of NP.

Rœsch P. (IMT) TCD2019 2019 27 / 72

slide-73
SLIDE 73

Rœsch P. (IMT) TCD2019 2019 28 / 72

slide-74
SLIDE 74

The Julia set of a rational map is defined as the unique minimal compact subset of the Riemann sphere C totally invariant ( by N and N−1) containing at least 3 points.

Rœsch P. (IMT) TCD2019 2019 28 / 72

slide-75
SLIDE 75

To cut the Julia set in small pieces we need to construct the equivalent to external ray.

Rœsch P. (IMT) TCD2019 2019 29 / 72

slide-76
SLIDE 76

To cut the Julia set in small pieces we need to construct the equivalent to external ray. There are 3 basins corresponding to the 3 roots of P, ∞ is a common point, landing of fixed internal rays in the basins.

Rœsch P. (IMT) TCD2019 2019 29 / 72

slide-77
SLIDE 77

Except in the symmetric case, only two basins intersect and there is a last angle of intersection

Rœsch P. (IMT) TCD2019 2019 30 / 72

slide-78
SLIDE 78

Except in the symmetric case, only two basins intersect and there is a last angle of intersection There is a Cantor set of angles Θ defining the intersection.

Rœsch P. (IMT) TCD2019 2019 30 / 72

slide-79
SLIDE 79

Construction of articulated rays by iterated pull back It is a curve γ such that f k(γ) = γ ∪ R1(t) ∪ R2(−t) with t ∈ Θ. It consists in infinitely many internal rays alternating from basin 1 et 2.

Rœsch P. (IMT) TCD2019 2019 31 / 72

slide-80
SLIDE 80

Using the following two graphs,

Rœsch P. (IMT) TCD2019 2019 32 / 72

slide-81
SLIDE 81

Using the following two graphs,

Theorem (R)

The intersection of the puzzle piece is either a point or the homeomorphic image of the filled Julia set of a quadratic polynomial.

Rœsch P. (IMT) TCD2019 2019 32 / 72

slide-82
SLIDE 82

Theorem (R)

In most cases the Julia set is locally connected.

Rœsch P. (IMT) TCD2019 2019 33 / 72

slide-83
SLIDE 83

Theorem (R)

In most cases the Julia set is locally connected.

Theorem (R)

In particular J(N) ⊃ h(J(P)) where J(P) is a non locally connected Julia set of quadratic polynomials P and J(N) is locally connected.

Rœsch P. (IMT) TCD2019 2019 33 / 72

slide-84
SLIDE 84

We use this puzzle structure to prove Tan Lei’s conjecture

Theorem (Aspenberg, R)

There exists a subset RC of renormalizable cubic polynomials, a subset RN

  • f renormalizable cubic Newton maps and a map M : RC → RN which is
  • nto and such that M(f ) is the mating of f with the polynomial

f∞(z) = z(z2 + 3

2).

One can understand the dynamics of N through the dynamics of the

  • polynomials. But there is no external rays any more.

Rœsch P. (IMT) TCD2019 2019 34 / 72

slide-85
SLIDE 85

Rœsch P. (IMT) TCD2019 2019 35 / 72

slide-86
SLIDE 86

Rœsch P. (IMT) TCD2019 2019 35 / 72

slide-87
SLIDE 87

Rœsch P. (IMT) TCD2019 2019 35 / 72

slide-88
SLIDE 88

Rœsch P. (IMT) TCD2019 2019 35 / 72

slide-89
SLIDE 89

Sketch of the mating

Rœsch P. (IMT) TCD2019 2019 35 / 72

slide-90
SLIDE 90

Understand rational map via the two polynomials

Rœsch P. (IMT) TCD2019 2019 36 / 72

slide-91
SLIDE 91

Understand rational map via the two polynomials

Rœsch P. (IMT) TCD2019 2019 36 / 72

slide-92
SLIDE 92

Understand rational map via the two polynomials

Rœsch P. (IMT) TCD2019 2019 36 / 72

slide-93
SLIDE 93

Understand rational map via the two polynomials

Rœsch P. (IMT) TCD2019 2019 36 / 72

slide-94
SLIDE 94

Understand rational map via the two polynomials

Rœsch P. (IMT) TCD2019 2019 36 / 72

slide-95
SLIDE 95

Definition

Two polynomials f1 and f2 are said mateable, if there exist a rational map R and two semi-conjugacies φj : Kj → ˆ C conformal on the interior of Kj, such that φ1(K1) ∪ φ2(K2) = ˆ C and ∀(z, w) ∈ Ki × Kj, φi(z) = φj(w) ⇐ ⇒ z ∼r w. The relation ∼r is generated by : the landing point of R1(t) is equivalent to the landing point of R2(−t).

Rœsch P. (IMT) TCD2019 2019 37 / 72

slide-96
SLIDE 96

Rœsch P. (IMT) TCD2019 2019 38 / 72

slide-97
SLIDE 97

Rœsch P. (IMT) TCD2019 2019 38 / 72

slide-98
SLIDE 98

Rœsch P. (IMT) TCD2019 2019 38 / 72

slide-99
SLIDE 99

Theorem (Aspenberg, R)

There exists a subset RC of renormalizable cubic polynomials, a subset RN

  • f renormalizable cubic Newton maps and a map M : RC → RN which is
  • nto and such that M(f ) is the mating of f with the polynomial

f∞(z) = z(z2 + 3

2).

Idea of the proof : we construct the semi conjugacy by sending the puzzle pieces of the abstract mating to the puzzle pieces for the Newton map.

Rœsch P. (IMT) TCD2019 2019 39 / 72

slide-100
SLIDE 100

Rœsch P. (IMT) TCD2019 2019 40 / 72

slide-101
SLIDE 101

Rœsch P. (IMT) TCD2019 2019 40 / 72

slide-102
SLIDE 102

Rœsch P. (IMT) TCD2019 2019 40 / 72

slide-103
SLIDE 103

To find the cubic Newton map, one has to investigate the space of cubic Newton map.

Rœsch P. (IMT) TCD2019 2019 41 / 72

slide-104
SLIDE 104

To find the cubic Newton map, one has to investigate the space of cubic Newton map. It is a one parameter slice with symmetries. More precisely any Newton map is conjugate to one of the form Nλ(z) = 2z3 − (λ2 − 1

4)

3z2 − (λ2 + 3

4) with λ ∈ C \ {±3

2, 0}

Rœsch P. (IMT) TCD2019 2019 41 / 72

slide-105
SLIDE 105

To find the cubic Newton map, one has to investigate the space of cubic Newton map. It is a one parameter slice with symmetries. More precisely any Newton map is conjugate to one of the form Nλ(z) = 2z3 − (λ2 − 1

4)

3z2 − (λ2 + 3

4) with λ ∈ C \ {±3

2, 0} The graphs exist and define puzzles in some precise regions of the parameter plane called para-puzzle pieces.

Rœsch P. (IMT) TCD2019 2019 41 / 72

slide-106
SLIDE 106

To define them one has to transfer to the parameter plane the articulated rays and all the pre-images.

Rœsch P. (IMT) TCD2019 2019 42 / 72

slide-107
SLIDE 107

Theorem (Wang, R, Yin)

(Advances 2017) Any ray in any hyperbolic component lands. The boundary of any hyperbolic component is a Jordan curve. Rigidty : Two Newton maps with the same combinatorics are conformally conjugated. It generalizes the proof done with para-puzzle pieces of the following

Theorem (R)

The boundary of the principal hyperbolic components are Jordan curves.

Rœsch P. (IMT) TCD2019 2019 43 / 72

slide-108
SLIDE 108

Sketch of the proof in the case of the principal hyperbolic component: Assume λ1 and λ2 are two accumulation points of an irrational ray so that Rλi(t) lands at the free critical point of Nλi. Then the Newton maps Nλ1 and Nλ2 share the combinatorial dynamics with respect to the puzzles constructed with the same angles. There is a topological conjugacy ψ between Nλ1 and Nλ2, which is holomorphic in the Fatou set of Nλ1. Using control on the distortion, on the shape and the decreasing of puzzle pieces we get that the conjugacy is a quasi-conformal map. The Lebesgue measure of J(Nλi) is zero ( Lyubich, Shishikura arguments on rational like maps with an admissible puzzle) The conjugacy is a Möbius transformation More recent progress in the dynamical plane...

Rœsch P. (IMT) TCD2019 2019 44 / 72

slide-109
SLIDE 109

Theorem (Wang, Yin, Zeng)

Let fp be the Newton map for any non-trivial polynomial P. Then the boundary of any immediate root basin B is locally connected. This is proved by generalizing the work for cubic Newton maps. Namely the puzzles by applying KSS nest and Kahn Lyubich covering Lemma.

Rœsch P. (IMT) TCD2019 2019 45 / 72

slide-110
SLIDE 110

Theorem (Wang, Yin, Zeng)

Let fp be the Newton map for any non-trivial polynomial P. Then the boundary of any immediate root basin B is locally connected. This is proved by generalizing the work for cubic Newton maps. Namely the puzzles by applying KSS nest and Kahn Lyubich covering Lemma.

Theorem (R., Yin, Zeng)

(arxiv. 11/2018) Non-renormalizable Newton maps are rigid. More precisely, we prove that the topological conjugacy is equivalent to quasiconformal conjugacy in this case.

Theorem ( Drach, Lodge, Schleicher, Sowinski)

There exists an invariant graph for higher degree Newton maps that gives a Fatou-Shihikura injection.

Theorem (Drach,Schleicher)

Rigidity for non renormalisable Newton maps or in the same "way".

Rœsch P. (IMT) TCD2019 2019 45 / 72

slide-111
SLIDE 111

McMullen maps

We consider the maps fλ : z → zn + λ zn .

Rœsch P. (IMT) TCD2019 2019 46 / 72

slide-112
SLIDE 112

McMullen maps

We consider the maps fλ : z → zn + λ zn . For small λ, the map fλ is a "perturbation" of zn whose Julia set is the unit circle.

Rœsch P. (IMT) TCD2019 2019 46 / 72

slide-113
SLIDE 113

McMullen maps

We consider the maps fλ : z → zn + λ zn . For small λ, the map fλ is a "perturbation" of zn whose Julia set is the unit circle. McMullen showed that the Julia set of fλ is a Cantor set of simple closed curves provided n = 1, 2 and λ is small. We restrict to n ≥ 3.

Rœsch P. (IMT) TCD2019 2019 46 / 72

slide-114
SLIDE 114

There exist also maps which are renormalizable and contain copies of polynomial Julia sets.

Rœsch P. (IMT) TCD2019 2019 47 / 72

slide-115
SLIDE 115

In the parameter plane appear : the unbounded component which is the Cantor set region the neighborhood of 0 where J(fλ) is a Cantor set of circles the other "holes" where the Julia set is a Sierpinsky carpet. n = 3

Rœsch P. (IMT) TCD2019 2019 48 / 72

slide-116
SLIDE 116

In the parameter plane appear : the unbounded component which is the Cantor set region the neighborhood of 0 where J(fλ) is a Cantor set of circles the other "holes" where the Julia set is a Sierpinsky carpet. n = 3 H∞ : the set of λ so that the critical points converge to ∞.

Rœsch P. (IMT) TCD2019 2019 48 / 72

slide-117
SLIDE 117

Rœsch P. (IMT) TCD2019 2019 49 / 72

slide-118
SLIDE 118

H0 H2 H0 is the unbounded component H2 is the component contaning 0

Rœsch P. (IMT) TCD2019 2019 49 / 72

slide-119
SLIDE 119

H0 H2 H0 is the unbounded component H2 is the component contaning 0 Precisely,

Rœsch P. (IMT) TCD2019 2019 49 / 72

slide-120
SLIDE 120

H0 H2 H0 is the unbounded component H2 is the component contaning 0 Precisely,

Theorem (Devaney-Look-Uminsky; Devaney-Russell)

If λ ∈ H0, then J(fλ) is a Cantor set ; If λ ∈ H2 \ {0}, then J(fλ) is homeomorphic to the product of a Cantor set and a circle ; If λ ∈ H∞ \ (H0 ∪ H2) , then J(fλ) is a Sierpinsky carpet ; If λ / ∈ H∞ then J(fλ) is connected.

Rœsch P. (IMT) TCD2019 2019 49 / 72

slide-121
SLIDE 121

H0 H2

Theorem (Devaney)

The boundary of H2 is a Jordan curve.

Rœsch P. (IMT) TCD2019 2019 50 / 72

slide-122
SLIDE 122

H0 H2

Theorem (Devaney)

The boundary of H2 is a Jordan curve.

Conjecture (Devaney)

The boundary of any connected component of H∞ is a Jordan curve.

Rœsch P. (IMT) TCD2019 2019 50 / 72

slide-123
SLIDE 123

Theorem (Qiu, Rœsch, Wang, Yin )

Let H be any connected component of H∞. Then H is a Jordan domain.

Rœsch P. (IMT) TCD2019 2019 51 / 72

slide-124
SLIDE 124

Theorem (Qiu, Rœsch, Wang, Yin )

Let H be any connected component of H∞. Then H is a Jordan domain. Moreover

Proposition (Qiu, Rœsch, Wang, Yin )

The parametrization extends to the boundary as a function ν(θ). If θ is periodic then the dynamical ray lands at a parabolic point. If θ is not periodic then the dynamical ray lands at the critical value.

Rœsch P. (IMT) TCD2019 2019 51 / 72

slide-125
SLIDE 125

Theorem (Qiu, Rœsch, Wang, Yin )

Let H be any connected component of H∞. Then H is a Jordan domain. Moreover

Proposition (Qiu, Rœsch, Wang, Yin )

The parametrization extends to the boundary as a function ν(θ). If θ is periodic then the dynamical ray lands at a parabolic point. If θ is not periodic then the dynamical ray lands at the critical value. A parameter λ is a cusp if fλ has a parabolic cycle.

Rœsch P. (IMT) TCD2019 2019 51 / 72

slide-126
SLIDE 126

Theorem (Qiu, Rœsch, Wang, Yin )

Let H be any connected component of H∞. Then H is a Jordan domain. Moreover

Proposition (Qiu, Rœsch, Wang, Yin )

The parametrization extends to the boundary as a function ν(θ). If θ is periodic then the dynamical ray lands at a parabolic point. If θ is not periodic then the dynamical ray lands at the critical value. A parameter λ is a cusp if fλ has a parabolic cycle.

Corollary

The cusps are dense in the boundary of H0.

Rœsch P. (IMT) TCD2019 2019 51 / 72

slide-127
SLIDE 127

Some symmetries :

fλ(z) = fλ(z) and fλ(ωz) = ωfλω−2(z) where ω = e

2iπ n−1 .

We will always restrict to the fundamental domain : F = {λ ∈ C∗ | 0 ≤ argλ < 2π n − 1} F n = 4

Rœsch P. (IMT) TCD2019 2019 52 / 72

slide-128
SLIDE 128

Some dynamics The maps fλ(z) = zn + λ/zn are the composition of two simple maps z → z + λ z and zn.

Rœsch P. (IMT) TCD2019 2019 53 / 72

slide-129
SLIDE 129

Some dynamics The maps fλ(z) = zn + λ/zn are the composition of two simple maps z → z + λ z and zn. The map z → z + λ z is just conjugated to z → z + 1 z .

Rœsch P. (IMT) TCD2019 2019 53 / 72

slide-130
SLIDE 130

z + 1/z .

Rœsch P. (IMT) TCD2019 2019 54 / 72

slide-131
SLIDE 131

The critical set of the map fλ(z) = zn + λ/zn is Crit = {0, ∞} ∪ Cλ where Cλ = {c | c2n = λ} = {c0e

ikπ n | k ∈ [0, .., 2n − 1]}

n = 4 .

Rœsch P. (IMT) TCD2019 2019 55 / 72

slide-132
SLIDE 132

In each sector the map is one to one onto C \ ±v0[1, +∞]. fλ On can pull back any sector except the ones containing ±v0.

Rœsch P. (IMT) TCD2019 2019 56 / 72

slide-133
SLIDE 133

S1 \ (Θ0 ∪ Θn) = 1

2n, 1 2

1

2 + 1 2n, 1

  • τ(θ) = nθ mod 1.

Rœsch P. (IMT) TCD2019 2019 57 / 72

slide-134
SLIDE 134

S1 \ (Θ0 ∪ Θn) = 1

2n, 1 2

1

2 + 1 2n, 1

  • τ(θ) = nθ mod 1.

θ has itinerary (s0, · · · , sk, · · · ) if τ k(θ) ∈ Θsk

Rœsch P. (IMT) TCD2019 2019 57 / 72

slide-135
SLIDE 135

S1 \ (Θ0 ∪ Θn) = 1

2n, 1 2

1

2 + 1 2n, 1

  • τ(θ) = nθ mod 1.

θ has itinerary (s0, · · · , sk, · · · ) if τ k(θ) ∈ Θsk Θ =

  • θ | τ k(θ) ∈ S1 \ (Θ0 ∪ Θn)

∀k ≥ 0

  • .

Rœsch P. (IMT) TCD2019 2019 57 / 72

slide-136
SLIDE 136

Pulling back to the sectors without critical values − → fλ n = 4

Rœsch P. (IMT) TCD2019 2019 58 / 72

slide-137
SLIDE 137

Pulling back to the sectors without critical values − → fλ n = 4 The intersection of a decreasing sequence of sectors shrinks to a curve in some cases.

Rœsch P. (IMT) TCD2019 2019 58 / 72

slide-138
SLIDE 138

Theorem (Devaney, Qiu-Wang-Yin)

For any λ in the interior of the fundamental domain F and for any θ ∈ Θ with itinerary (s0, s1, · · · , ) the set Ωθ

λ :=

  • k≥0

f −k

λ

(Sλ

sk ∪ Sλ −sk)

is a Jordan curve intersecting the Julia set under a Cantor set. "cut rays" Ω1

λ = Ω1/2 λ

n = 3.

Rœsch P. (IMT) TCD2019 2019 59 / 72

slide-139
SLIDE 139

Theorem (Devaney, Qiu-Wang-Yin)

For any λ in the interior of the fundamental domain F and for any θ ∈ Θ with itinerary (s0, s1, · · · , ) the set Ωθ

λ :=

  • k≥0

f −k

λ

(Sλ

sk ∪ Sλ −sk)

is a Jordan curve intersecting the Julia set under a Cantor set. "cut rays" Ω1

λ = Ω1/2 λ

n = 3. There is a similar construction for λ ∈ R.

Rœsch P. (IMT) TCD2019 2019 59 / 72

slide-140
SLIDE 140

Rœsch P. (IMT) TCD2019 2019 60 / 72

slide-141
SLIDE 141

Rœsch P. (IMT) TCD2019 2019 61 / 72

slide-142
SLIDE 142

Notations : Bλ is the immediate basin of ∞, φλ : Bλ → C \ D the Böttcher map : φλ(fλ(z)) = (φλ(z))n, φ′

λ(∞) = 1,

Rλ(θ) is the ray of angle θ in Bλ : Rλ(θ) := φ−1

λ ((1, +∞)e2iπt).

Rœsch P. (IMT) TCD2019 2019 62 / 72

slide-143
SLIDE 143

Notations : Bλ is the immediate basin of ∞, φλ : Bλ → C \ D the Böttcher map : φλ(fλ(z)) = (φλ(z))n, φ′

λ(∞) = 1,

Rλ(θ) is the ray of angle θ in Bλ : Rλ(θ) := φ−1

λ ((1, +∞)e2iπt).

Properties of "cut rays" : Ωθ

λ = −Ωθ λ = Ωθ+1/2 λ

; fλ : Ωθ

λ → Ωτ(θ) λ

is two to one ; Ωθ

λ ∩ Bλ = Rλ(θ) ∪ Rλ(θ + 1/2) ∪ {∞} ;

Ωθ

λ ∩ (C \ J(fλ)) ⊂ k≥0 f −k λ

(Bλ).

Rœsch P. (IMT) TCD2019 2019 62 / 72

slide-144
SLIDE 144

The "cut rays" are used in order to construct a puzzle.

Rœsch P. (IMT) TCD2019 2019 63 / 72

slide-145
SLIDE 145

The "cut rays" are used in order to construct a puzzle.

Theorem ( Qiu-Wang-Yin)

If J(fλ) is not a Cantor set, then the boundary of Bλ is a Jordan curve. The precise result is on decreasing of puzzle pieces , it is used in order to get the rigidity in the parameter plane.

Rœsch P. (IMT) TCD2019 2019 63 / 72

slide-146
SLIDE 146

Parameter plane.

We restrict to H0.

Rœsch P. (IMT) TCD2019 2019 64 / 72

slide-147
SLIDE 147

Parameter plane.

We restrict to H0. To prove that ∂H0 is a Jordan curve we will prove that the impression of any ray is reduced to a single point.

Rœsch P. (IMT) TCD2019 2019 64 / 72

slide-148
SLIDE 148

Parameter plane.

We restrict to H0. To prove that ∂H0 is a Jordan curve we will prove that the impression of any ray is reduced to a single point. Let Φ0 : H0 → C \ D be a parametrization given by " the position of critical value in Bλ" : Φ0(λ) = (φλ(vλ))2.

Rœsch P. (IMT) TCD2019 2019 64 / 72

slide-149
SLIDE 149

Parameter plane.

We restrict to H0. To prove that ∂H0 is a Jordan curve we will prove that the impression of any ray is reduced to a single point. Let Φ0 : H0 → C \ D be a parametrization given by " the position of critical value in Bλ" : Φ0(λ) = (φλ(vλ))2. A ray R(t) is Φ−1

0 (]1, ∞]e2iπt).

Rœsch P. (IMT) TCD2019 2019 64 / 72

slide-150
SLIDE 150

The impression of a ray is the intersection : Imp(t) :=

  • k≥1

Φ−1

0 ({]1, 1 + 1/k[e2iπθ | |θ − t| < 1/k})

Rœsch P. (IMT) TCD2019 2019 65 / 72

slide-151
SLIDE 151

Idea of the proof :

1 We prove that the impression is a finite set ; since it is a connected

subset of ∂H0, it will be one point.

Rœsch P. (IMT) TCD2019 2019 66 / 72

slide-152
SLIDE 152

Idea of the proof :

1 We prove that the impression is a finite set ; since it is a connected

subset of ∂H0, it will be one point.

2 To get 1) we prove that parameters in the impression of a ray have

special properties :

Rœsch P. (IMT) TCD2019 2019 66 / 72

slide-153
SLIDE 153

Idea of the proof :

1 We prove that the impression is a finite set ; since it is a connected

subset of ∂H0, it will be one point.

2 To get 1) we prove that parameters in the impression of a ray have

special properties :

◮ either "the dynamical ray" lands at a critical value , ◮ or "the dynamical ray" lands at a parabolic cycle (finite set). Rœsch P. (IMT) TCD2019 2019 66 / 72

slide-154
SLIDE 154

Idea of the proof :

1 We prove that the impression is a finite set ; since it is a connected

subset of ∂H0, it will be one point.

2 To get 1) we prove that parameters in the impression of a ray have

special properties :

◮ either "the dynamical ray" lands at a critical value , ◮ or "the dynamical ray" lands at a parabolic cycle (finite set). 3 For two parameters in the same impression, which are not cusps, we

construct a conjugacy between the maps and use

◮ Thurston’s theorem in the post-critically finite case ; ◮ that the homeomorphism is quasi-conformal, conformal on the Fatou

set and that the Julia set is of measure zero.

Rœsch P. (IMT) TCD2019 2019 66 / 72

slide-155
SLIDE 155

Special properties at ∂H0 :

Proposition

For λ ∈ ∂H0 the boundary ∂Bλ either contains the critical set Cλ or a parabolic point.

Rœsch P. (IMT) TCD2019 2019 67 / 72

slide-156
SLIDE 156

Special properties at ∂H0 :

Proposition

For λ ∈ ∂H0 the boundary ∂Bλ either contains the critical set Cλ or a parabolic point. Assume that ∂Bλ ∩ Cλ = ∅ and that ∂Bλ contains no parabolic point.

Rœsch P. (IMT) TCD2019 2019 67 / 72

slide-157
SLIDE 157

Special properties at ∂H0 :

Proposition

For λ ∈ ∂H0 the boundary ∂Bλ either contains the critical set Cλ or a parabolic point. Assume that ∂Bλ ∩ Cλ = ∅ and that ∂Bλ contains no parabolic point. Then the map fλ : ∂Bλ → Bλ is expanding.

Rœsch P. (IMT) TCD2019 2019 67 / 72

slide-158
SLIDE 158

Special properties at ∂H0 :

Proposition

For λ ∈ ∂H0 the boundary ∂Bλ either contains the critical set Cλ or a parabolic point. Assume that ∂Bλ ∩ Cλ = ∅ and that ∂Bλ contains no parabolic point. Then the map fλ : ∂Bλ → Bλ is expanding. Therefore there exist Uλ, Vλ disks with Bλ ⊂ Vλ ⊂ Vλ ⊂ Uλ such that fλ : Vλ → Uλ is a polynomial-like map with one critical point.

Rœsch P. (IMT) TCD2019 2019 67 / 72

slide-159
SLIDE 159

Special properties at ∂H0 :

Proposition

For λ ∈ ∂H0 the boundary ∂Bλ either contains the critical set Cλ or a parabolic point. Assume that ∂Bλ ∩ Cλ = ∅ and that ∂Bλ contains no parabolic point. Then the map fλ : ∂Bλ → Bλ is expanding. Therefore there exist Uλ, Vλ disks with Bλ ⊂ Vλ ⊂ Vλ ⊂ Uλ such that fλ : Vλ → Uλ is a polynomial-like map with one critical point. For nearby λ, one should have a polynomial-like map with the same degree. In H0, J(fλ) is a Cantor set. Contradiction

Rœsch P. (IMT) TCD2019 2019 67 / 72

slide-160
SLIDE 160

Special properties in Imp(t) :

Proposition

For λ ∈ Imp(t), the dynamical ray Rλ(t/2) or Rλ(t/2 + 1/2) lands at a parabolic cycle, or at one of the critical values.

Rœsch P. (IMT) TCD2019 2019 68 / 72

slide-161
SLIDE 161

Special properties in Imp(t) :

Proposition

For λ ∈ Imp(t), the dynamical ray Rλ(t/2) or Rλ(t/2 + 1/2) lands at a parabolic cycle, or at one of the critical values. There is a critical value or a parabolic point on ∂Bλ : pλ. Denote by Rλ(t′) the ray landing at pλ. Assume t′ = t

2 and t′ = t+1 2 .

Rœsch P. (IMT) TCD2019 2019 68 / 72

slide-162
SLIDE 162

Special properties in Imp(t) :

Proposition

For λ ∈ Imp(t), the dynamical ray Rλ(t/2) or Rλ(t/2 + 1/2) lands at a parabolic cycle, or at one of the critical values. There is a critical value or a parabolic point on ∂Bλ : pλ. Denote by Rλ(t′) the ray landing at pλ. Assume t′ = t

2 and t′ = t+1 2 .

Rλ(t′) ∪ {pλ}, Rλ( t

2) and Rλ( t+1 2 ) are separated by cut rays Ωλα and Ωλβ.

Rλ(t′) Ωα

λ

V +

λ •

They move holomorphically Rλ( t

2)

Ωβ

λ λ( t+1 2 )

Rœsch P. (IMT) TCD2019 2019 68 / 72

slide-163
SLIDE 163

Special properties in Imp(t) :

Proposition

For λ ∈ Imp(t), the dynamical ray Rλ(t/2) or Rλ(t/2 + 1/2) lands at a parabolic cycle, or at one of the critical values. There is a critical value or a parabolic point on ∂Bλ : pλ. Denote by Rλ(t′) the ray landing at pλ. Assume t′ = t

2 and t′ = t+1 2 .

Rλ(t′) ∪ {pλ}, Rλ( t

2) and Rλ( t+1 2 ) are separated by cut rays Ωλα and Ωλβ.

Rλ(t′) Ωα

λ

V +

λ •

They move holomorphically Rλ( t

2)

Ωβ

λ λ( t+1 2 )

so stay in different components Contradiction. .

Rœsch P. (IMT) TCD2019 2019 68 / 72

slide-164
SLIDE 164

Lemma

For 0 ≤ t < 1/(n − 1) there is a finite number of cusps in Imp(t).

Rœsch P. (IMT) TCD2019 2019 69 / 72

slide-165
SLIDE 165

Lemma

For 0 ≤ t < 1/(n − 1) there is a finite number of cusps in Imp(t). Proof. The dynamical ray Rλ(t/2) lands at a parabolic point, then there exists k ≥ 1 such that τ k(t/2) = t/2 mod 1 k depends only on t. λ satisfies : ∃ x | f k

λ (x) = x,

(f k

λ )′(x) = 1

This is a finite set.

Rœsch P. (IMT) TCD2019 2019 69 / 72

slide-166
SLIDE 166

Lemma

For 0 ≤ t < 1/(n − 1) there is a finite number of cusps in Imp(t). Proof. The dynamical ray Rλ(t/2) lands at a parabolic point, then there exists k ≥ 1 such that τ k(t/2) = t/2 mod 1 k depends only on t. λ satisfies : ∃ x | f k

λ (x) = x,

(f k

λ )′(x) = 1

This is a finite set.

Lemma

If 0 ≤ t < 1/(n − 1) and λ1, λ2 ∈ Imp(t) are not cusps, then Rλ1(t/2) lands at vλ1 and Rλ2(t/2) lands at vλ2. By continuity.

Rœsch P. (IMT) TCD2019 2019 69 / 72

slide-167
SLIDE 167

Assume that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ Q.

Rœsch P. (IMT) TCD2019 2019 70 / 72

slide-168
SLIDE 168

Assume that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ Q. the Böttcher maps give a conjugacy φ on the basin of ∞ ;

Rœsch P. (IMT) TCD2019 2019 70 / 72

slide-169
SLIDE 169

Assume that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ Q. the Böttcher maps give a conjugacy φ on the basin of ∞ ; φ extends to the closure ;

Rœsch P. (IMT) TCD2019 2019 70 / 72

slide-170
SLIDE 170

Assume that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ Q. the Böttcher maps give a conjugacy φ on the basin of ∞ ; φ extends to the closure ; the maps are post-critically finite since Rλi(t/2) lands at the critical value ;

Rœsch P. (IMT) TCD2019 2019 70 / 72

slide-171
SLIDE 171

Assume that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ Q. the Böttcher maps give a conjugacy φ on the basin of ∞ ; φ extends to the closure ; the maps are post-critically finite since Rλi(t/2) lands at the critical value ; φ sends the postcritical set of fλ1 to the one of fλ2 ;

Rœsch P. (IMT) TCD2019 2019 70 / 72

slide-172
SLIDE 172

Assume that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ Q. the Böttcher maps give a conjugacy φ on the basin of ∞ ; φ extends to the closure ; the maps are post-critically finite since Rλi(t/2) lands at the critical value ; φ sends the postcritical set of fλ1 to the one of fλ2 ; φ extends to a homeomorphism of C ;

Rœsch P. (IMT) TCD2019 2019 70 / 72

slide-173
SLIDE 173

Assume that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ Q. the Böttcher maps give a conjugacy φ on the basin of ∞ ; φ extends to the closure ; the maps are post-critically finite since Rλi(t/2) lands at the critical value ; φ sends the postcritical set of fλ1 to the one of fλ2 ; φ extends to a homeomorphism of C ; its lifts ψ gives a combinatorial conjugacy in Thurston’s sense ;

Rœsch P. (IMT) TCD2019 2019 70 / 72

slide-174
SLIDE 174

Assume that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ Q. the Böttcher maps give a conjugacy φ on the basin of ∞ ; φ extends to the closure ; the maps are post-critically finite since Rλi(t/2) lands at the critical value ; φ sends the postcritical set of fλ1 to the one of fλ2 ; φ extends to a homeomorphism of C ; its lifts ψ gives a combinatorial conjugacy in Thurston’s sense ; by Thurston’s theorem fλ1 and fλ2 are Möbius conjugate.

Rœsch P. (IMT) TCD2019 2019 70 / 72

slide-175
SLIDE 175

Assume now that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ R \ Q.

Rœsch P. (IMT) TCD2019 2019 71 / 72

slide-176
SLIDE 176

Assume now that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ R \ Q. The cut ray Ω1

λi is a quasi-arc ;

Rœsch P. (IMT) TCD2019 2019 71 / 72

slide-177
SLIDE 177

Assume now that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ R \ Q. The cut ray Ω1

λi is a quasi-arc ;

there exists a quasi-conformal homeomorphism ψ0 such that ψ0(Ω1

λ1) = Ω1 λ2,

(ψ0)|BλR

1 = (φ−1

λ2 ◦ φλ1)|BλR

1 ; Rœsch P. (IMT) TCD2019 2019 71 / 72

slide-178
SLIDE 178

Assume now that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ R \ Q. The cut ray Ω1

λi is a quasi-arc ;

there exists a quasi-conformal homeomorphism ψ0 such that ψ0(Ω1

λ1) = Ω1 λ2,

(ψ0)|BλR

1 = (φ−1

λ2 ◦ φλ1)|BλR

1 ;

there exists a family (ψn) of quasi-conformal homeomorphisms such that ψn+1 lifts ψn ;

Rœsch P. (IMT) TCD2019 2019 71 / 72

slide-179
SLIDE 179

Assume now that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ R \ Q. The cut ray Ω1

λi is a quasi-arc ;

there exists a quasi-conformal homeomorphism ψ0 such that ψ0(Ω1

λ1) = Ω1 λ2,

(ψ0)|BλR

1 = (φ−1

λ2 ◦ φλ1)|BλR

1 ;

there exists a family (ψn) of quasi-conformal homeomorphisms such that ψn+1 lifts ψn ; the dilatation of (ψn) is uniformly bounded ;

Rœsch P. (IMT) TCD2019 2019 71 / 72

slide-180
SLIDE 180

Assume now that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ R \ Q. The cut ray Ω1

λi is a quasi-arc ;

there exists a quasi-conformal homeomorphism ψ0 such that ψ0(Ω1

λ1) = Ω1 λ2,

(ψ0)|BλR

1 = (φ−1

λ2 ◦ φλ1)|BλR

1 ;

there exists a family (ψn) of quasi-conformal homeomorphisms such that ψn+1 lifts ψn ; the dilatation of (ψn) is uniformly bounded ; the sequence ψn converges to a quasi-conformal homeomorphism ψ ;

Rœsch P. (IMT) TCD2019 2019 71 / 72

slide-181
SLIDE 181

Assume now that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ R \ Q. The cut ray Ω1

λi is a quasi-arc ;

there exists a quasi-conformal homeomorphism ψ0 such that ψ0(Ω1

λ1) = Ω1 λ2,

(ψ0)|BλR

1 = (φ−1

λ2 ◦ φλ1)|BλR

1 ;

there exists a family (ψn) of quasi-conformal homeomorphisms such that ψn+1 lifts ψn ; the dilatation of (ψn) is uniformly bounded ; the sequence ψn converges to a quasi-conformal homeomorphism ψ ; ψ conjugates fλ1 and fλ2 on the Fatou set, by continuity on C ;

Rœsch P. (IMT) TCD2019 2019 71 / 72

slide-182
SLIDE 182

Assume now that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ R \ Q. The cut ray Ω1

λi is a quasi-arc ;

there exists a quasi-conformal homeomorphism ψ0 such that ψ0(Ω1

λ1) = Ω1 λ2,

(ψ0)|BλR

1 = (φ−1

λ2 ◦ φλ1)|BλR

1 ;

there exists a family (ψn) of quasi-conformal homeomorphisms such that ψn+1 lifts ψn ; the dilatation of (ψn) is uniformly bounded ; the sequence ψn converges to a quasi-conformal homeomorphism ψ ; ψ conjugates fλ1 and fλ2 on the Fatou set, by continuity on C ; ψ is conformal on the Fatou set ;

Rœsch P. (IMT) TCD2019 2019 71 / 72

slide-183
SLIDE 183

Assume now that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ R \ Q. The cut ray Ω1

λi is a quasi-arc ;

there exists a quasi-conformal homeomorphism ψ0 such that ψ0(Ω1

λ1) = Ω1 λ2,

(ψ0)|BλR

1 = (φ−1

λ2 ◦ φλ1)|BλR

1 ;

there exists a family (ψn) of quasi-conformal homeomorphisms such that ψn+1 lifts ψn ; the dilatation of (ψn) is uniformly bounded ; the sequence ψn converges to a quasi-conformal homeomorphism ψ ; ψ conjugates fλ1 and fλ2 on the Fatou set, by continuity on C ; ψ is conformal on the Fatou set ; the Julia set has measure zero ;

Rœsch P. (IMT) TCD2019 2019 71 / 72

slide-184
SLIDE 184

Assume now that λ1, λ2 ∈ Imp(t) are not cusps and t ∈ R \ Q. The cut ray Ω1

λi is a quasi-arc ;

there exists a quasi-conformal homeomorphism ψ0 such that ψ0(Ω1

λ1) = Ω1 λ2,

(ψ0)|BλR

1 = (φ−1

λ2 ◦ φλ1)|BλR

1 ;

there exists a family (ψn) of quasi-conformal homeomorphisms such that ψn+1 lifts ψn ; the dilatation of (ψn) is uniformly bounded ; the sequence ψn converges to a quasi-conformal homeomorphism ψ ; ψ conjugates fλ1 and fλ2 on the Fatou set, by continuity on C ; ψ is conformal on the Fatou set ; the Julia set has measure zero ; then ψ is a Möbius conjugacy between fλ1 and fλ2.

Rœsch P. (IMT) TCD2019 2019 71 / 72

slide-185
SLIDE 185

Rigidity for non-recurrent exponential maps Let fc(z) = ez + c. Let Γ be a closed forward invariant graph formed by finitely many periodic rays together with their landing points.

Rœsch P. (IMT) TCD2019 2019 72 / 72

slide-186
SLIDE 186

Rigidity for non-recurrent exponential maps Let fc(z) = ez + c. Let Γ be a closed forward invariant graph formed by finitely many periodic rays together with their landing points. For each n, the the connected components of C\ where Γn = ∪n

j=0f −j(Γ)

are puzzle pieces of level n. A parameter c is combinatorially non-recurrent if there is a suitable Γ which separates the singular value from the postsingular set.

Theorem[Benini]

Let c; c0 be non-escaping parameters, and fc be combinatorially non-

  • recurrent. If fc0 is combinatorially equivalent to fc, then c0 = c.

Rœsch P. (IMT) TCD2019 2019 72 / 72