Equilibrium States are determined by their unstable conditionals. - - PowerPoint PPT Presentation

equilibrium states are determined by their unstable
SMART_READER_LITE
LIVE PREVIEW

Equilibrium States are determined by their unstable conditionals. - - PowerPoint PPT Presentation

Equilibrium States are determined by their unstable conditionals. Pablo D. Carrasco (joint w/ Federico Rodriguez-Hertz) ICMC-USP visiting ICTP July of 2016 ICTP P. Carrasco, ICMC-USP visiting ICTP Equilibrium States are determined by their


slide-1
SLIDE 1

Equilibrium States are determined by their unstable conditionals.

Pablo D. Carrasco (joint w/ Federico Rodriguez-Hertz)

ICMC-USP visiting ICTP

July of 2016 ICTP

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 1/20

slide-2
SLIDE 2

General Thermodynamic Formalism.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 2/20

slide-3
SLIDE 3

General Thermodynamic Formalism.

M compact metric space. f ∈ Homeo(M).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 2/20

slide-4
SLIDE 4

General Thermodynamic Formalism.

M compact metric space. f ∈ Homeo(M). Topological pressure:

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 2/20

slide-5
SLIDE 5

General Thermodynamic Formalism.

M compact metric space. f ∈ Homeo(M). Topological pressure: Pf (M) := set of f -invariant Borel probability measures on M.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 2/20

slide-6
SLIDE 6

General Thermodynamic Formalism.

M compact metric space. f ∈ Homeo(M). Topological pressure: Pf (M) := set of f -invariant Borel probability measures on M. If ϕ : M → R is continuous (potential), define

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 2/20

slide-7
SLIDE 7

General Thermodynamic Formalism.

M compact metric space. f ∈ Homeo(M). Topological pressure: Pf (M) := set of f -invariant Borel probability measures on M. If ϕ : M → R is continuous (potential), define Ptop(ϕ) = sup

ν∈Pf (M)

{hν(f ) +

  • ϕdν}.
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 2/20

slide-8
SLIDE 8

Equilibrium States.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 3/20

slide-9
SLIDE 9

Equilibrium States. Definition

µ ∈ Pf (M) is an equilibrium state for the potential ϕ if Ptop(ϕ) = hµ(f ) +

  • ϕdµ (i.e. the supremum is attained).
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 3/20

slide-10
SLIDE 10

Equilibrium States. Definition

µ ∈ Pf (M) is an equilibrium state for the potential ϕ if Ptop(ϕ) = hµ(f ) +

  • ϕdµ (i.e. the supremum is attained).

We are interested in:

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 3/20

slide-11
SLIDE 11

Equilibrium States. Definition

µ ∈ Pf (M) is an equilibrium state for the potential ϕ if Ptop(ϕ) = hµ(f ) +

  • ϕdµ (i.e. the supremum is attained).

We are interested in: Existence of equilibrium states.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 3/20

slide-12
SLIDE 12

Equilibrium States. Definition

µ ∈ Pf (M) is an equilibrium state for the potential ϕ if Ptop(ϕ) = hµ(f ) +

  • ϕdµ (i.e. the supremum is attained).

We are interested in: Existence of equilibrium states. General methods (Bowen).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 3/20

slide-13
SLIDE 13

Equilibrium States. Definition

µ ∈ Pf (M) is an equilibrium state for the potential ϕ if Ptop(ϕ) = hµ(f ) +

  • ϕdµ (i.e. the supremum is attained).

We are interested in: Existence of equilibrium states. General methods (Bowen). Uniqueness of equilibrium states

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 3/20

slide-14
SLIDE 14

Equilibrium States. Definition

µ ∈ Pf (M) is an equilibrium state for the potential ϕ if Ptop(ϕ) = hµ(f ) +

  • ϕdµ (i.e. the supremum is attained).

We are interested in: Existence of equilibrium states. General methods (Bowen). Uniqueness of equilibrium states General (but more restrictive) methods

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 3/20

slide-15
SLIDE 15

Equilibrium States. Definition

µ ∈ Pf (M) is an equilibrium state for the potential ϕ if Ptop(ϕ) = hµ(f ) +

  • ϕdµ (i.e. the supremum is attained).

We are interested in: Existence of equilibrium states. General methods (Bowen). Uniqueness of equilibrium states General (but more restrictive) methods (Bowen, or more recently, Climenhaga-Thompson).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 3/20

slide-16
SLIDE 16

Equilibrium States. Definition

µ ∈ Pf (M) is an equilibrium state for the potential ϕ if Ptop(ϕ) = hµ(f ) +

  • ϕdµ (i.e. the supremum is attained).

We are interested in: Existence of equilibrium states. General methods (Bowen). Uniqueness of equilibrium states General (but more restrictive) methods (Bowen, or more recently, Climenhaga-Thompson). Properties/Description of equilibrium states.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 3/20

slide-17
SLIDE 17

Equilibrium States. Definition

µ ∈ Pf (M) is an equilibrium state for the potential ϕ if Ptop(ϕ) = hµ(f ) +

  • ϕdµ (i.e. the supremum is attained).

We are interested in: Existence of equilibrium states. General methods (Bowen). Uniqueness of equilibrium states General (but more restrictive) methods (Bowen, or more recently, Climenhaga-Thompson). Properties/Description of equilibrium states. (Mixing. Bernoulli. Decay of correlations.)

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 3/20

slide-18
SLIDE 18

A very important theorem in smooth ergodic theory.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 4/20

slide-19
SLIDE 19

A very important theorem in smooth ergodic theory.

Smooth Ergodic Theory: M is a manifold and f ∈ Diff r(M).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 4/20

slide-20
SLIDE 20

A very important theorem in smooth ergodic theory.

Smooth Ergodic Theory: M is a manifold and f ∈ Diff r(M).

Theorem

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 4/20

slide-21
SLIDE 21

A very important theorem in smooth ergodic theory.

Smooth Ergodic Theory: M is a manifold and f ∈ Diff r(M).

Theorem

Λ ⊂ M hyperbolic attractor for f (in particular, f |Λ is transitive).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 4/20

slide-22
SLIDE 22

A very important theorem in smooth ergodic theory.

Smooth Ergodic Theory: M is a manifold and f ∈ Diff r(M).

Theorem

Λ ⊂ M hyperbolic attractor for f (in particular, f |Λ is transitive). Then for every H¨

  • lder potential ϕ : Λ → R there exists a unique

equilibrium state µϕ for f |Λ.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 4/20

slide-23
SLIDE 23

A very important theorem in smooth ergodic theory.

Smooth Ergodic Theory: M is a manifold and f ∈ Diff r(M).

Theorem

Λ ⊂ M hyperbolic attractor for f (in particular, f |Λ is transitive). Then for every H¨

  • lder potential ϕ : Λ → R there exists a unique

equilibrium state µϕ for f |Λ. Furthermore, is f |Λ is topologically mixing then the system (f , µϕ) is metrically isomorphic to a Bernoulli shift.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 4/20

slide-24
SLIDE 24

A very important theorem in smooth ergodic theory.

Smooth Ergodic Theory: M is a manifold and f ∈ Diff r(M).

Theorem

Λ ⊂ M hyperbolic attractor for f (in particular, f |Λ is transitive). Then for every H¨

  • lder potential ϕ : Λ → R there exists a unique

equilibrium state µϕ for f |Λ. Furthermore, is f |Λ is topologically mixing then the system (f , µϕ) is metrically isomorphic to a Bernoulli shift. Due to: Sinai-Ruelle-Bowen,

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 4/20

slide-25
SLIDE 25

A very important theorem in smooth ergodic theory.

Smooth Ergodic Theory: M is a manifold and f ∈ Diff r(M).

Theorem

Λ ⊂ M hyperbolic attractor for f (in particular, f |Λ is transitive). Then for every H¨

  • lder potential ϕ : Λ → R there exists a unique

equilibrium state µϕ for f |Λ. Furthermore, is f |Λ is topologically mixing then the system (f , µϕ) is metrically isomorphic to a Bernoulli shift. Due to: Sinai-Ruelle-Bowen, Ornstein and Weiss.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 4/20

slide-26
SLIDE 26

Examples:

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 5/20

slide-27
SLIDE 27

Examples:

ϕ ≡ 0 ⇒ µ is the entropy maximizing measure.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 5/20

slide-28
SLIDE 28

Examples:

ϕ ≡ 0 ⇒ µ is the entropy maximizing measure. ϕ(x) = − log |det df |E u

x | ⇒ µ is the SRB measure (need to assume

f is C 1+θ)

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 5/20

slide-29
SLIDE 29

Main idea and Generalizations.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 6/20

slide-30
SLIDE 30

Main idea and Generalizations.

Main idea of the proof:

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 6/20

slide-31
SLIDE 31

Main idea and Generalizations.

Main idea of the proof: Using Markov partitions reduce to a subshift of finite type and establish the theorem there. Then push everything back to the manifold using the semi-conjugacy.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 6/20

slide-32
SLIDE 32

Main idea and Generalizations.

Main idea of the proof: Using Markov partitions reduce to a subshift of finite type and establish the theorem there. Then push everything back to the manifold using the semi-conjugacy. Several generalizations of the above theorem exists.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 6/20

slide-33
SLIDE 33

Main idea and Generalizations.

Main idea of the proof: Using Markov partitions reduce to a subshift of finite type and establish the theorem there. Then push everything back to the manifold using the semi-conjugacy. Several generalizations of the above theorem exists. Highlight: Anosov flows.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 6/20

slide-34
SLIDE 34

Main idea and Generalizations.

Main idea of the proof: Using Markov partitions reduce to a subshift of finite type and establish the theorem there. Then push everything back to the manifold using the semi-conjugacy. Several generalizations of the above theorem exists. Highlight: Anosov flows. Some hyperbolicity is usually required, either for f of for the potential (see for example Y. Lima’s course next week).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 6/20

slide-35
SLIDE 35

Main idea and Generalizations.

Main idea of the proof: Using Markov partitions reduce to a subshift of finite type and establish the theorem there. Then push everything back to the manifold using the semi-conjugacy. Several generalizations of the above theorem exists. Highlight: Anosov flows. Some hyperbolicity is usually required, either for f of for the potential (see for example Y. Lima’s course next week). We discuss now one important example where the available methods fail.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 6/20

slide-36
SLIDE 36

Diagonal actions (on locally homogeneous spaces).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 7/20

slide-37
SLIDE 37

Diagonal actions (on locally homogeneous spaces).

In the course of Mohammadi it’s been discussed diagonal actions on quotients of SL(n, R).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 7/20

slide-38
SLIDE 38

Diagonal actions (on locally homogeneous spaces).

In the course of Mohammadi it’s been discussed diagonal actions on quotients of SL(n, R). G = Sl(3, R) .

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 7/20

slide-39
SLIDE 39

Diagonal actions (on locally homogeneous spaces).

In the course of Mohammadi it’s been discussed diagonal actions on quotients of SL(n, R). G = Sl(3, R) . A = {diag(ea, eb, ec) : a + b + c = 0} < G

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 7/20

slide-40
SLIDE 40

Diagonal actions (on locally homogeneous spaces).

In the course of Mohammadi it’s been discussed diagonal actions on quotients of SL(n, R). G = Sl(3, R) . A = {diag(ea, eb, ec) : a + b + c = 0} < G Γ < G co-compact torsion free lattice.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 7/20

slide-41
SLIDE 41

Diagonal actions (on locally homogeneous spaces).

In the course of Mohammadi it’s been discussed diagonal actions on quotients of SL(n, R). G = Sl(3, R) . A = {diag(ea, eb, ec) : a + b + c = 0} < G Γ < G co-compact torsion free lattice. α : A M = Sl(3, R)/Γ

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 7/20

slide-42
SLIDE 42

Diagonal actions (on locally homogeneous spaces).

In the course of Mohammadi it’s been discussed diagonal actions on quotients of SL(n, R). G = Sl(3, R) . A = {diag(ea, eb, ec) : a + b + c = 0} < G Γ < G co-compact torsion free lattice. α : A M = Sl(3, R)/Γ is an Anosov action

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 7/20

slide-43
SLIDE 43

Diagonal actions (on locally homogeneous spaces).

In the course of Mohammadi it’s been discussed diagonal actions on quotients of SL(n, R). G = Sl(3, R) . A = {diag(ea, eb, ec) : a + b + c = 0} < G Γ < G co-compact torsion free lattice. α : A M = Sl(3, R)/Γ is an Anosov action, meaning there exists an element f = α(g) having an invariant splitting TM = E s ⊕ E c ⊕ E u

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 7/20

slide-44
SLIDE 44

Diagonal actions (on locally homogeneous spaces).

In the course of Mohammadi it’s been discussed diagonal actions on quotients of SL(n, R). G = Sl(3, R) . A = {diag(ea, eb, ec) : a + b + c = 0} < G Γ < G co-compact torsion free lattice. α : A M = Sl(3, R)/Γ is an Anosov action, meaning there exists an element f = α(g) having an invariant splitting TM = E s ⊕ E c ⊕ E u such that (for some metric) df |E u expansion, df |E s contraction and E c is tangent to the orbits of the action.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 7/20

slide-45
SLIDE 45

Center isometries.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 8/20

slide-46
SLIDE 46

Center isometries.

The metric in M can be chosen so that f acts as an isometry on the orbit foliation.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 8/20

slide-47
SLIDE 47

Center isometries.

The metric in M can be chosen so that f acts as an isometry on the orbit

  • foliation. Hence, f is a center isometry (the definition is nearly evident).
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 8/20

slide-48
SLIDE 48

Center isometries.

The metric in M can be chosen so that f acts as an isometry on the orbit

  • foliation. Hence, f is a center isometry (the definition is nearly evident).

Remark: no hyperbolicity whatsoever along the center.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 8/20

slide-49
SLIDE 49

Center isometries.

The metric in M can be chosen so that f acts as an isometry on the orbit

  • foliation. Hence, f is a center isometry (the definition is nearly evident).

Remark: no hyperbolicity whatsoever along the center. For a center isometry all bundles E s, E u, E c, E cs = E c ⊕ E s, E cu = E c ⊕ E u are integrable to f -invariant foliations W ∗.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 8/20

slide-50
SLIDE 50

Main Theorems.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 9/20

slide-51
SLIDE 51

Main Theorems.

f : M → M center isometry of class C 2 s.t

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 9/20

slide-52
SLIDE 52

Main Theorems.

f : M → M center isometry of class C 2 s.t W s, W u are minimal.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 9/20

slide-53
SLIDE 53

Main Theorems.

f : M → M center isometry of class C 2 s.t W s, W u are minimal. ϕ : M → R H¨

  • lder potential.
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 9/20

slide-54
SLIDE 54

Main Theorems.

f : M → M center isometry of class C 2 s.t W s, W u are minimal. ϕ : M → R H¨

  • lder potential.

Theorem A [P.C., F. Rodriguez-Hertz]

There exist µϕ ∈ Pf (M) and families of measures µu = {µu

x}x∈M, µs = {µs x}x∈M, µcu = {µcu x }x∈M, µcs = {µcs x }x∈M

satisfying the following.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 9/20

slide-55
SLIDE 55

Main Theorems.

f : M → M center isometry of class C 2 s.t W s, W u are minimal. ϕ : M → R H¨

  • lder potential.

Theorem A [P.C., F. Rodriguez-Hertz]

There exist µϕ ∈ Pf (M) and families of measures µu = {µu

x}x∈M, µs = {µs x}x∈M, µcu = {µcu x }x∈M, µcs = {µcs x }x∈M

satisfying the following.

  • 1. The probability µϕ is an equilibrium state for the potential ϕ.
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 9/20

slide-56
SLIDE 56

Main Theorems.

f : M → M center isometry of class C 2 s.t W s, W u are minimal. ϕ : M → R H¨

  • lder potential.

Theorem A [P.C., F. Rodriguez-Hertz]

There exist µϕ ∈ Pf (M) and families of measures µu = {µu

x}x∈M, µs = {µs x}x∈M, µcu = {µcu x }x∈M, µcs = {µcs x }x∈M

satisfying the following.

  • 1. The probability µϕ is an equilibrium state for the potential ϕ.
  • 2. For every x ∈ M the measure µσ, σ ∈ {u, s, cu, cs} is a Radon

measure on W σ(x) which is positive on relatively open sets, and y ∈ W σ(x) implies µσ

x = µσ y .

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 9/20

slide-57
SLIDE 57

Cont.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 10/20

slide-58
SLIDE 58

Cont.

  • 3. If ξ is a measurable partition that refines the partition by unstable

(stable) leaves then the conditionals (µϕ)ξ

x of µϕ are equivalent to

µu

x (resp. µs x) for µϕ − a.e.(x).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 10/20

slide-59
SLIDE 59

Cont.

  • 3. If ξ is a measurable partition that refines the partition by unstable

(stable) leaves then the conditionals (µϕ)ξ

x of µϕ are equivalent to

µu

x (resp. µs x) for µϕ − a.e.(x).

  • 4. For every ǫ > 0 sufficiently small, for every x ∈ M the measure

µϕ|D(x; ǫ) has product structure with respect to the pair µu

x, µcs x ,

i.e. its equivalent to µu

x × µcs x .

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 10/20

slide-60
SLIDE 60

Cont.

  • 3. If ξ is a measurable partition that refines the partition by unstable

(stable) leaves then the conditionals (µϕ)ξ

x of µϕ are equivalent to

µu

x (resp. µs x) for µϕ − a.e.(x).

  • 4. For every ǫ > 0 sufficiently small, for every x ∈ M the measure

µϕ|D(x; ǫ) has product structure with respect to the pair µu

x, µcs x ,

i.e. its equivalent to µu

x × µcs x .

  • 5. Given ǫ > 0 there exist a(ǫ), b(ǫ) > 0 such that if

U(x, ǫ, n) = {y ∈ W u(x, ǫ) : d(f jx, f jy) < ǫ, j = 0, . . . , n − 1} then a(ǫ) ≤ µu

x(U(x, ǫ, n)))

eSnϕ(x)−nPtop(ϕ) ≤ b(ǫ).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 10/20

slide-61
SLIDE 61

Bernoulli property and uniqueness.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 11/20

slide-62
SLIDE 62

Bernoulli property and uniqueness. Theorem B [P.C., F. Rodriguez-Hertz]

Under certain technical conditions the system (f , µϕ) is metrically isomorphic to a Bernoulli shift.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 11/20

slide-63
SLIDE 63

Bernoulli property and uniqueness. Theorem B [P.C., F. Rodriguez-Hertz]

Under certain technical conditions the system (f , µϕ) is metrically isomorphic to a Bernoulli shift.

Theorem C [P.C., F. Rodriguez-Hertz]

If either dimE s, dimE u = 1, or

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 11/20

slide-64
SLIDE 64

Bernoulli property and uniqueness. Theorem B [P.C., F. Rodriguez-Hertz]

Under certain technical conditions the system (f , µϕ) is metrically isomorphic to a Bernoulli shift.

Theorem C [P.C., F. Rodriguez-Hertz]

If either dimE s, dimE u = 1, or f is an ergodic automorphisms of TN (no repeated eigenvalues),

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 11/20

slide-65
SLIDE 65

Bernoulli property and uniqueness. Theorem B [P.C., F. Rodriguez-Hertz]

Under certain technical conditions the system (f , µϕ) is metrically isomorphic to a Bernoulli shift.

Theorem C [P.C., F. Rodriguez-Hertz]

If either dimE s, dimE u = 1, or f is an ergodic automorphisms of TN (no repeated eigenvalues), the equilibrium state µϕ is unique.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 11/20

slide-66
SLIDE 66

Bernoulli property and uniqueness. Theorem B [P.C., F. Rodriguez-Hertz]

Under certain technical conditions the system (f , µϕ) is metrically isomorphic to a Bernoulli shift.

Theorem C [P.C., F. Rodriguez-Hertz]

If either dimE s, dimE u = 1, or f is an ergodic automorphisms of TN (no repeated eigenvalues), the equilibrium state µϕ is unique. Work in progress: Uniqueness also holds in the homogeneous examples (Weyl Chambers’ flow).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 11/20

slide-67
SLIDE 67

SRB measures.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 12/20

slide-68
SLIDE 68

SRB measures.

In the general setting, an SRB measure is a invariant measure whose unstable conditionals are absolutely continuous with respect to Lebesgue.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 12/20

slide-69
SLIDE 69

SRB measures.

In the general setting, an SRB measure is a invariant measure whose unstable conditionals are absolutely continuous with respect to Lebesgue. We keep working with center isometries.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 12/20

slide-70
SLIDE 70

SRB measures.

In the general setting, an SRB measure is a invariant measure whose unstable conditionals are absolutely continuous with respect to Lebesgue. We keep working with center isometries.

  • SRB measures exist (Sinai-Pesin).
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 12/20

slide-71
SLIDE 71

SRB measures.

In the general setting, an SRB measure is a invariant measure whose unstable conditionals are absolutely continuous with respect to Lebesgue. We keep working with center isometries.

  • SRB measures exist (Sinai-Pesin).
  • Ledrappier-Young: µ is an SRB if and only if

hµ(f ) =

  • log Ju(x)dµ(x)

Ju(x) = det(df |E u

x )

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 12/20

slide-72
SLIDE 72

SRB measures.

In the general setting, an SRB measure is a invariant measure whose unstable conditionals are absolutely continuous with respect to Lebesgue. We keep working with center isometries.

  • SRB measures exist (Sinai-Pesin).
  • Ledrappier-Young: µ is an SRB if and only if

hµ(f ) =

  • log Ju(x)dµ(x)

Ju(x) = det(df |E u

x )

Implicit: Unstable manifolds coincide with Pesin’s unstable manifolds.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 12/20

slide-73
SLIDE 73
  • Eq. states are determined by their unstable conditionals.
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 13/20

slide-74
SLIDE 74
  • Eq. states are determined by their unstable conditionals.

Theorem D [P.C., F. Rodriguez-Hertz]

If µ is an equilibrium state for ϕ then µ has conditionals along unstables equivalent to µu.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 13/20

slide-75
SLIDE 75
  • Eq. states are determined by their unstable conditionals.

Theorem D [P.C., F. Rodriguez-Hertz]

If µ is an equilibrium state for ϕ then µ has conditionals along unstables equivalent to µu. Similarly it has conditionals along stables equivalent to µs.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 13/20

slide-76
SLIDE 76
  • Eq. states are determined by their unstable conditionals.

Theorem D [P.C., F. Rodriguez-Hertz]

If µ is an equilibrium state for ϕ then µ has conditionals along unstables equivalent to µu. Similarly it has conditionals along stables equivalent to µs. Conversely if µ ∈ Pf (M) has unstable conditionals absolutely continuous wrt µu, then µ is an equilibrium state for ϕ.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 13/20

slide-77
SLIDE 77
  • Eq. states are determined by their unstable conditionals.

Theorem D [P.C., F. Rodriguez-Hertz]

If µ is an equilibrium state for ϕ then µ has conditionals along unstables equivalent to µu. Similarly it has conditionals along stables equivalent to µs. Conversely if µ ∈ Pf (M) has unstable conditionals absolutely continuous wrt µu, then µ is an equilibrium state for ϕ.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 13/20

slide-78
SLIDE 78
  • Eq. states are determined by their unstable conditionals.

Theorem D [P.C., F. Rodriguez-Hertz]

If µ is an equilibrium state for ϕ then µ has conditionals along unstables equivalent to µu. Similarly it has conditionals along stables equivalent to µs. Conversely if µ ∈ Pf (M) has unstable conditionals absolutely continuous wrt µu, then µ is an equilibrium state for ϕ. The families µu, µs provide the reference measures to which one can compare.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 13/20

slide-79
SLIDE 79

Sketch of the proof: Thm D.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 14/20

slide-80
SLIDE 80

Sketch of the proof: Thm D. Definition

We call a measurable partition ξ a SPLY partition if

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 14/20

slide-81
SLIDE 81

Sketch of the proof: Thm D. Definition

We call a measurable partition ξ a SPLY partition if f ξ < ξ. ξ subordinated to Wu µ-a.e. x the atom ξ(x) contains a neighbourhood of x inside W u(x).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 14/20

slide-82
SLIDE 82

Sketch of the proof: Thm D. Definition

We call a measurable partition ξ a SPLY partition if f ξ < ξ. ξ subordinated to Wu µ-a.e. x the atom ξ(x) contains a neighbourhood of x inside W u(x). SPLY partitions exist: (Sinai, Pesin - Ledrappier, Strelcyn).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 14/20

slide-83
SLIDE 83

Sketch of the proof: Thm D. Definition

We call a measurable partition ξ a SPLY partition if f ξ < ξ. ξ subordinated to Wu µ-a.e. x the atom ξ(x) contains a neighbourhood of x inside W u(x). SPLY partitions exist: (Sinai, Pesin - Ledrappier, Strelcyn). We fix one SPLY partition and take m ∈ Pf (M) equilibrium state for ϕ;

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 14/20

slide-84
SLIDE 84

Sketch of the proof: Thm D. Definition

We call a measurable partition ξ a SPLY partition if f ξ < ξ. ξ subordinated to Wu µ-a.e. x the atom ξ(x) contains a neighbourhood of x inside W u(x). SPLY partitions exist: (Sinai, Pesin - Ledrappier, Strelcyn). We fix one SPLY partition and take m ∈ Pf (M) equilibrium state for ϕ; denote mx = unstable conditional of m on ξ(x).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 14/20

slide-85
SLIDE 85

’Change of variables’ for µu

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 15/20

slide-86
SLIDE 86

’Change of variables’ for µu Theorem

For every x ∈ M

  • 1. µσ

fx = ePtop(ϕ)−ϕf∗µσ x

σ ∈ {u, cu}.

  • 2. µσ

fx = eϕ−Ptop(ϕ)f∗µσ x

σ ∈ {s, cs}.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 15/20

slide-87
SLIDE 87
  • Cont. proof
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 16/20

slide-88
SLIDE 88
  • Cont. proof

If mx << µu

x m − a.e.(x) we can write

dmx = ρdµu

x

where ρ is measurable on M.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 16/20

slide-89
SLIDE 89
  • Cont. proof

If mx << µu

x m − a.e.(x) we can write

dmx = ρdµu

x

where ρ is measurable on M. One can then show x → ρ(x) ρ(f −1x)eP−ϕ(f −1x) is constant on the atoms of ξ.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 16/20

slide-90
SLIDE 90
  • Cont. proof

If mx << µu

x m − a.e.(x) we can write

dmx = ρdµu

x

where ρ is measurable on M. One can then show x → ρ(x) ρ(f −1x)eP−ϕ(f −1x) is constant on the atoms of ξ. From here one deduces that y ∈ ξ(x) ⇒ ρ(y) ρ(x) =

  • k=1

eϕ◦f −k(y) eϕ◦f −k(x) = ∆x(y).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 16/20

slide-91
SLIDE 91
  • Cont. proof
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 17/20

slide-92
SLIDE 92
  • Cont. proof

⇒ ρ (provided is defined) should have the form ρ(y) = ∆x(y) L(x) , y ∈ ξ(x) with L(x) =

  • ξ(x) ∆x(y)dµu

x(y).

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 17/20

slide-93
SLIDE 93
  • Cont. proof

⇒ ρ (provided is defined) should have the form ρ(y) = ∆x(y) L(x) , y ∈ ξ(x) with L(x) =

  • ξ(x) ∆x(y)dµu

x(y).

Define a measure ν by requiring ν = m on Bξ and such its conditionals

  • n ξ are given by dνx = ρdµx.
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 17/20

slide-94
SLIDE 94
  • Cont. proof

⇒ ρ (provided is defined) should have the form ρ(y) = ∆x(y) L(x) , y ∈ ξ(x) with L(x) =

  • ξ(x) ∆x(y)dµu

x(y).

Define a measure ν by requiring ν = m on Bξ and such its conditionals

  • n ξ are given by dνx = ρdµx. We want to prove m = ν.
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 17/20

slide-95
SLIDE 95
  • Cont. proof

⇒ ρ (provided is defined) should have the form ρ(y) = ∆x(y) L(x) , y ∈ ξ(x) with L(x) =

  • ξ(x) ∆x(y)dµu

x(y).

Define a measure ν by requiring ν = m on Bξ and such its conditionals

  • n ξ are given by dνx = ρdµx. We want to prove m = ν.

It is enough to show m = ν on every Bf −nξ, n ≥ 0.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 17/20

slide-96
SLIDE 96
  • Cont. proof

⇒ ρ (provided is defined) should have the form ρ(y) = ∆x(y) L(x) , y ∈ ξ(x) with L(x) =

  • ξ(x) ∆x(y)dµu

x(y).

Define a measure ν by requiring ν = m on Bξ and such its conditionals

  • n ξ are given by dνx = ρdµx. We want to prove m = ν.

It is enough to show m = ν on every Bf −nξ, n ≥ 0. An induction argument shows that after proving m = ν on Bf −1ξ we are done.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 17/20

slide-97
SLIDE 97
  • Cont. proof

q(x) = νx(f −1ξ(x)) ⇒

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 18/20

slide-98
SLIDE 98
  • Cont. proof

q(x) = νx(f −1ξ(x)) ⇒ q(x) = 1 L(x)

  • f −1(ξ(fx))

∆x(f −1fy) eP−ϕ(y) eP−ϕ(f −1fy) dµu

x

= 1 L(x)

  • ξ(fx)

∆x(f −1z)eϕ(f −1z)−Pdµu

fx(z) = L(fx)

L(x) eϕ(x)−P ≤ 1

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 18/20

slide-99
SLIDE 99
  • Cont. proof

q(x) = νx(f −1ξ(x)) ⇒ q(x) = 1 L(x)

  • f −1(ξ(fx))

∆x(f −1fy) eP−ϕ(y) eP−ϕ(f −1fy) dµu

x

= 1 L(x)

  • ξ(fx)

∆x(f −1z)eϕ(f −1z)−Pdµu

fx(z) = L(fx)

L(x) eϕ(x)−P ≤ 1 L(fx) L(x) ≤ eϕ(x)−P ∈ L1(M, m)

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 18/20

slide-100
SLIDE 100
  • Cont. proof

q(x) = νx(f −1ξ(x)) ⇒ q(x) = 1 L(x)

  • f −1(ξ(fx))

∆x(f −1fy) eP−ϕ(y) eP−ϕ(f −1fy) dµu

x

= 1 L(x)

  • ξ(fx)

∆x(f −1z)eϕ(f −1z)−Pdµu

fx(z) = L(fx)

L(x) eϕ(x)−P ≤ 1 L(fx) L(x) ≤ eϕ(x)−P ∈ L1(M, m)

  • log L ◦ f

L dm = 0 ⇒

  • − log q(x)dm(x) = P −
  • ϕdm.
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 18/20

slide-101
SLIDE 101

End of the proof.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 19/20

slide-102
SLIDE 102

End of the proof.

⇒ P −

  • ϕdm =
  • − log νx(f −1ξ(x))dm(x).

Since ξ is a SPLY and m equilibrium state, Ledrappier, Young tell us that ⇒ P −

  • ϕdm = hm(f ) = H(f −1ξ|ξ) =
  • − log mx(f −1ξ(x))dm(x)
  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 19/20

slide-103
SLIDE 103

End of the proof.

⇒ P −

  • ϕdm =
  • − log νx(f −1ξ(x))dm(x).

Since ξ is a SPLY and m equilibrium state, Ledrappier, Young tell us that ⇒ P −

  • ϕdm = hm(f ) = H(f −1ξ|ξ) =
  • − log mx(f −1ξ(x))dm(x)

  • − log νx(f −1ξ(x))

mx(f −1ξ(x))dm(x) = 0

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 19/20

slide-104
SLIDE 104

End of the proof.

⇒ P −

  • ϕdm =
  • − log νx(f −1ξ(x))dm(x).

Since ξ is a SPLY and m equilibrium state, Ledrappier, Young tell us that ⇒ P −

  • ϕdm = hm(f ) = H(f −1ξ|ξ) =
  • − log mx(f −1ξ(x))dm(x)

  • − log νx(f −1ξ(x))

mx(f −1ξ(x))dm(x) = 0 Using strict convexity of the logarithm function plus working (carefully) with the partitions f −1ξ(x))|ξ(x) we deduce dν dm

  • f −1ξ

(x) = 1 m − a.e.x ∈ M.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 19/20

slide-105
SLIDE 105

End of the proof.

⇒ P −

  • ϕdm =
  • − log νx(f −1ξ(x))dm(x).

Since ξ is a SPLY and m equilibrium state, Ledrappier, Young tell us that ⇒ P −

  • ϕdm = hm(f ) = H(f −1ξ|ξ) =
  • − log mx(f −1ξ(x))dm(x)

  • − log νx(f −1ξ(x))

mx(f −1ξ(x))dm(x) = 0 Using strict convexity of the logarithm function plus working (carefully) with the partitions f −1ξ(x))|ξ(x) we deduce dν dm

  • f −1ξ

(x) = 1 m − a.e.x ∈ M. and afterwards, mx = νx on ξ(x) for ALL atoms, as we wanted to show.

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 19/20

slide-106
SLIDE 106

Grazie!!!

  • P. Carrasco, ICMC-USP visiting ICTP

Equilibrium States are determined by their unstable conditionals. 20/20