Prove2Me
Navigate
DiscoverFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

Groups of piecewise projective homeomorphisms of the projective line

Definition
Monod_PiecewiseProjective

by dbenbenn · Sep 30, 2026 · Mathlib 0df444a (Lean v4.33.1)

amenabilitygroup-theorypiecewise-projective

Monod (2013), pp. 1–3: the groups G(A)G(A)G(A) and H(A)H(A)H(A) of piecewise projective homeomorphisms, and three amenability-type notions used in the paper.

The projective line. p. 1: “Consider the natural action of the group PSL2(R)\mathrm{PSL}_2(\mathbf{R})PSL2​(R) on the projective line P1=P1(R)\mathbf{P}^1 = \mathbf{P}^1(\mathbf{R})P1=P1(R). We endow P1\mathbf{P}^1P1 with its R\mathbf{R}R-topology making it a topological circle.” P1\mathbf{P}^1P1 is OnePoint ℝ, the real line with a point ∞\infty∞ added (the one-point compactification, a circle). slToGL A sends SL2(A)\mathrm{SL}_2(A)SL2​(A), for a subring AAA of R\mathbf{R}R, into GL2(R)\mathrm{GL}_2(\mathbf{R})GL2​(R), and mob g x is the Möbius action x↦(ax+b)/(cx+d)x \mapsto (ax + b)/(cx + d)x↦(ax+b)/(cx+d) of g=(abcd)g = \begin{pmatrix} a & b \\ c & d\end{pmatrix}g=(ac​bd​) on P1\mathbf{P}^1P1 (Mathlib's action of GL2\mathrm{GL}_2GL2​ on OnePoint). Since −1-1−1 acts trivially this is the action of PSL2(A)\mathrm{PSL}_2(A)PSL2​(A), which the source names without defining.

Hyperbolic elements and PAP_APA​. p. 1: “Given a subring A<RA < \mathbf{R}A<R, we denote by PA⊆P1P_A \subseteq \mathbf{P}^1PA​⊆P1 the collection of all fixed points of all hyperbolic elements of PSL2(A)\mathrm{PSL}_2(A)PSL2​(A).” The source does not define “hyperbolic” on pp. 1–3; here g∈SL2(A)g \in \mathrm{SL}_2(A)g∈SL2​(A) is hyperbolic when its trace has absolute value greater than 222. P A is PAP_APA​, the set of points of P1\mathbf{P}^1P1 fixed by some hyperbolic g∈SL2(A)g \in \mathrm{SL}_2(A)g∈SL2​(A). For A=RA = \mathbf{R}A=R this is all of P1\mathbf{P}^1P1.

Piecewise projective maps. The source has no separate definition of this notion; it is part of the sentences defining GGG and G(A)G(A)G(A) below. IsPiecewiseProjOn A E f: there is a finite set B⊆EB \subseteq EB⊆E such that near every point outside BBB the homeomorphism fff agrees with mob g for some g∈SL2(A)g \in \mathrm{SL}_2(A)g∈SL2​(A). IsPiecewiseProj A f takes E=PAE = P_AE=PA​: piecewise in PSL2(A)\mathrm{PSL}_2(A)PSL2​(A) with breakpoints in PAP_APA​.

The groups.

  • p. 1: “We denote by GGG the group of all homeomorphisms of P1\mathbf{P}^1P1 which are piecewise in PSL2(R)\mathrm{PSL}_2(\mathbf{R})PSL2​(R), each piece being an interval of P1\mathbf{P}^1P1, with finitely many pieces.” Monod's GGG is Gpp, the group generated by the homeomorphisms piecewise in PSL2(R)\mathrm{PSL}_2(\mathbf{R})PSL2​(R) with breakpoints anywhere.
  • p. 1: “We let H<GH < GH<G be the subgroup fixing the point ∞∈P1\infty \in \mathbf{P}^1∞∈P1 corresponding to the first basis vector of R2\mathbf{R}^2R2.” fixInf is the stabilizer of ∞\infty∞, and Monod's HHH is Hpp, the stabilizer of ∞\infty∞ in Gpp.
  • p. 1: “We define G(A)G(A)G(A) to be the subgroup of GGG given by all elements that are piecewise in PSL2(A)\mathrm{PSL}_2(A)PSL2​(A) with all interval endpoints in PAP_APA​.” G A is G(A)G(A)G(A), the subgroup of Gpp generated by its elements that are IsPiecewiseProj A.
  • p. 1: “We write H(A)=G(A)∩HH(A) = G(A) \cap HH(A)=G(A)∩H, which is the stabilizer of ∞\infty∞ in G(A)G(A)G(A).” H A =G(A)∩= G(A) \cap=G(A)∩ fixInf; H(Z)H(\mathbf{Z})H(Z) is H ⊥.
  • p. 2: “The relation is as follows: if we modify the definition of H(Z)H(\mathbf{Z})H(Z) by requiring that the breakpoints be rational, then all its elements are automatically C1C^1C1 and the resulting group is conjugated to FFF. The corresponding relation holds between G(Z)G(\mathbf{Z})G(Z) and Thompson's group TTT.” The source defines the variants only through these two sentences (for G(Z)G(\mathbf{Z})G(Z) the modification is implicit in the second). ratPoints is Q∪{∞}\mathbf{Q} \cup \{\infty\}Q∪{∞}, and GRat, HRat are the variants of G(Z)G(\mathbf{Z})G(Z) and H(Z)H(\mathbf{Z})H(Z) with breakpoints in ratPoints. Both are defined as generated subgroups; Monod.mem_GRat_iff and Monod.mem_HRat_iff show that they consist exactly of the homeomorphisms that are piecewise in PSL2(Z)\mathrm{PSL}_2(\mathbf{Z})PSL2​(Z) with breakpoints in ratPoints (for HRat, those that also fix ∞\infty∞), the groups the two sentences describe.

Stabilizers and amenability notions.

  • fixSubgroup J E is the pointwise stabilizer of E⊆P1E \subseteq \mathbf{P}^1E⊆P1 in JJJ; the source only names it (Proposition 7, p. 1).
  • p. 3: “A subgroup KKK of a group JJJ is called co-amenable if there is an JJJ-invariant mean on J/KJ/KJ/K.” IsCoamenable K: J/KJ / KJ/K carries a JJJ-invariant finitely additive probability measure on all its subsets.
  • p. 3: “Recall that a group JJJ is inner amenable if there is a conjugacy-invariant mean on J\{e}J\backslash\{e\}J\{e}.” IsInnerAmenable J: JJJ carries a finitely additive probability measure on all subsets, invariant under conjugation and giving {1}\{1\}{1} measure 000 (equivalently, such a mean on J∖{e}J \setminus \{e\}J∖{e}).

Amenable measured equivalence relations. p. 2: “We recall that a measurable equivalence relation with countable classes is amenable if there is an a.e. defined measurable assignment of a mean on the orbit of each point in such a way that the means of two equivalent points coincide.” The operator form is used here (see the Formalization Note). For a measure μ\muμ on XXX and a relation R⊆X×XR \subseteq X \times XR⊆X×X, IsAmenableRel μ R says RRR has a left invariant mean (Connes–Feldman–Weiss, Ergodic Theory Dynam. Systems 1 (1981), Def. 5–6, p. 437; Schmidt, CBMS 76, Def. 1.4): a map P from bounded measurable functions on RRR to functions on XXX that is, up to μ\muμ-null sets, linear, positive, sends 111 to 111, has measurable values, identifies functions that differ on a null set of the left counting measure of RRR (RelNull: the set of first coordinates where they differ is μ\muμ-null), and is invariant under every partial transformation φ\varphiφ of RRR (PartialTransformation: a measurable isomorphism between measurable subsets of XXX with graph in RRR): P(fφ)=(Pf)φP(f^\varphi) = (Pf)^\varphiP(fφ)=(Pf)φ with fφ(y,x)=f(φ−1y,x)f^\varphi(y, x) = f(\varphi^{-1} y, x)fφ(y,x)=f(φ−1y,x) and Fφ=F∘φ−1F^\varphi = F \circ \varphi^{-1}Fφ=F∘φ−1 on the image of φ\varphiφ, both 000 elsewhere (shiftRel, shiftBase, CFW p. 436). volP1 is the Lebesgue measure class on P1\mathbf{P}^1P1, which the source leaves implicit: Lebesgue measure on R⊆P1\mathbf{R} \subseteq \mathbf{P}^1R⊆P1, ∞\infty∞ null, for the Borel σ-algebra.

Formalization Note. The finitely many interval pieces in the definition of GGG (p. 1, quoted above) are stated locally off a finite set; since a Möbius transformation is determined by its values on any open set, fff is a single Möbius map on each arc between breakpoints. Means are finitely additive probability measures on all subsets, as in the published Garrido.IsAmenable. Products of homeomorphisms are composition, (fg)(x)=f(g(x))(fg)(x) = f(g(x))(fg)(x)=f(g(x)). Monod's definition of an amenable relation (p. 2, quoted above) is a pointwise family of means on the orbits, invariant along the relation; that pointwise family is what CFW (p. 437) give as motivation for their definition, an operator PPP with (Pf)(x)(Pf)(x)(Pf)(x) the mean of fff on the class of xxx. Their official definition, used here, is the operator form, as in Schmidt, whom Monod cites for the next sentence. Under the operator form a measurable action of an amenable group produces an amenable relation in ZFC; for the pointwise family this is known only assuming CH (Kechris, The Theory of Countable Borel Equivalence Relations, 2025, Remark 9.23).

Definition code
import Definitions.Def_Garrido_Amenability
import Mathlib

/-!
# Groups of piecewise projective homeomorphisms

N. Monod, *Groups of piecewise projective homeomorphisms*, Proc. Natl. Acad. Sci. USA **110**
(2013) 4524–4527, pp. 1–3.

The projective line `P¹ = OnePoint ℝ` with the Möbius action of `SL(2, A)` for a subring `A` of
`ℝ`; the fixed points `P_A` of hyperbolic elements; the groups `G(A)` and `H(A) = G(A) ∩ H` of
piecewise projective homeomorphisms; the pointwise stabilizer of a set; co-amenable subgroups and
inner amenable groups (p. 3).
-/

namespace Monod

open OnePoint Filter Topology
open scoped Pointwise

/-- `SL(2, A)` acting on `ℝ²` as a subgroup of `GL(2, ℝ)`. -/
noncomputable def slToGL (A : Subring ℝ) : Matrix.SpecialLinearGroup (Fin 2) A →* GL (Fin 2) ℝ :=
  Matrix.SpecialLinearGroup.toGL.comp (Matrix.SpecialLinearGroup.map A.subtype)

/-- The Möbius action of `g ∈ SL(2, A)` on the projective line `P¹ = OnePoint ℝ`. -/
noncomputable def mob {A : Subring ℝ} (g : Matrix.SpecialLinearGroup (Fin 2) A) (x : OnePoint ℝ) :
    OnePoint ℝ :=
  slToGL A g • x

/-- `g ∈ SL(2, A)` is **hyperbolic**: its trace has absolute value greater than `2`. -/
def IsHyperbolic {A : Subring ℝ} (g : Matrix.SpecialLinearGroup (Fin 2) A) : Prop :=
  2 < |((Matrix.trace (g : Matrix (Fin 2) (Fin 2) A) : A) : ℝ)|

/-- `P_A` (p. 1): the fixed points in `P¹` of hyperbolic elements of `SL(2, A)`. -/
def P (A : Subring ℝ) : Set (OnePoint ℝ) :=
  {p | ∃ g : Matrix.SpecialLinearGroup (Fin 2) A, IsHyperbolic g ∧ mob g p = p}

/-- `f` is **piecewise in `PSL(2, A)` with breakpoints in `E`**: for some finite set `B ⊆ E`,
near every point outside `B` the map `f` agrees with the Möbius action of an element of
`SL(2, A)`. -/
def IsPiecewiseProjOn (A : Subring ℝ) (E : Set (OnePoint ℝ)) (f : OnePoint ℝ ≃ₜ OnePoint ℝ) :
    Prop :=
  ∃ B : Finset (OnePoint ℝ), (↑B : Set (OnePoint ℝ)) ⊆ E ∧
    ∀ x ∉ B, ∃ g : Matrix.SpecialLinearGroup (Fin 2) A, ∀ᶠ y in 𝓝 x, f y = mob g y

/-- The homeomorphisms of `P¹` that are piecewise in `PSL(2, A)` with all breakpoints in `P_A`
(p. 1). -/
def IsPiecewiseProj (A : Subring ℝ) (f : OnePoint ℝ ≃ₜ OnePoint ℝ) : Prop :=
  IsPiecewiseProjOn A (P A) f

/-- Monod's `G` (p. 1): the group generated by the homeomorphisms of `P¹` that are piecewise in
`PSL₂(ℝ)` with finitely many pieces, the breakpoints anywhere. -/
def Gpp : Subgroup (OnePoint ℝ ≃ₜ OnePoint ℝ) :=
  Subgroup.closure {f | IsPiecewiseProjOn ⊤ Set.univ f}

/-- `G(A)` (p. 1): the subgroup of `G` generated by its elements that are piecewise in `PSL(2, A)`
with all breakpoints in `P_A`. -/
def G (A : Subring ℝ) : Subgroup (OnePoint ℝ ≃ₜ OnePoint ℝ) :=
  Subgroup.closure {f | f ∈ Gpp ∧ IsPiecewiseProj A f}

/-- The homeomorphisms of `P¹` fixing `∞`. -/
def fixInf : Subgroup (OnePoint ℝ ≃ₜ OnePoint ℝ) where
  carrier := {f | f OnePoint.infty = OnePoint.infty}
  one_mem' := rfl
  mul_mem' {f g} hf hg := by
    show f (g OnePoint.infty) = OnePoint.infty
    rw [hg, hf]
  inv_mem' {f} hf := by
    show f.symm OnePoint.infty = OnePoint.infty
    have h := f.symm_apply_apply OnePoint.infty
    rwa [hf] at h

/-- `H(A)` (p. 1): the stabilizer of `∞` in `G(A)`. -/
def H (A : Subring ℝ) : Subgroup (OnePoint ℝ ≃ₜ OnePoint ℝ) :=
  G A ⊓ fixInf

/-- Monod's `H` (p. 1): the stabilizer of `∞` in `G`. -/
def Hpp : Subgroup (OnePoint ℝ ≃ₜ OnePoint ℝ) :=
  Gpp ⊓ fixInf

/-- The points of `P¹` that are rational or `∞`. -/
def ratPoints : Set (OnePoint ℝ) := {x | x = OnePoint.infty ∨ ∃ q : ℚ, x = ((q : ℝ) : OnePoint ℝ)}

/-- `G(ℤ)` with rational breakpoints (p. 2): the subgroup of `G` generated by its elements that are
piecewise in `PSL(2, ℤ)` with all breakpoints in `ℚ ∪ {∞}`. -/
def GRat : Subgroup (OnePoint ℝ ≃ₜ OnePoint ℝ) :=
  Subgroup.closure {f | f ∈ Gpp ∧ IsPiecewiseProjOn ⊥ ratPoints f}

/-- `H(ℤ)` with rational breakpoints (p. 2): the elements of `GRat` fixing `∞`. -/
def HRat : Subgroup (OnePoint ℝ ≃ₜ OnePoint ℝ) :=
  GRat ⊓ fixInf

/-- The pointwise stabilizer in `J` of a set `E ⊆ P¹`, for `J` a group of homeomorphisms of
`P¹`. -/
def fixSubgroup (J : Subgroup (OnePoint ℝ ≃ₜ OnePoint ℝ)) (E : Set (OnePoint ℝ)) : Subgroup J where
  carrier := {h | ∀ x ∈ E, (h : OnePoint ℝ ≃ₜ OnePoint ℝ) x = x}
  one_mem' _ _ := rfl
  mul_mem' {a b} ha hb x hx := by
    show (a : OnePoint ℝ ≃ₜ OnePoint ℝ) ((b : OnePoint ℝ ≃ₜ OnePoint ℝ) x) = x
    rw [hb x hx, ha x hx]
  inv_mem' {a} ha x hx := by
    show (a : OnePoint ℝ ≃ₜ OnePoint ℝ).symm x = x
    conv_lhs => rw [← ha x hx]
    exact (a : OnePoint ℝ ≃ₜ OnePoint ℝ).symm_apply_apply x

/-- `K ≤ J` is **co-amenable** (p. 3): there is a `J`-invariant mean on `J ⧸ K`, here a finitely
additive probability measure on all subsets of `J ⧸ K`. -/
def IsCoamenable {J : Type*} [Group J] (K : Subgroup J) : Prop :=
  ∃ m : Set (J ⧸ K) → ENNReal,
    Garrido.IsFinitelyAdditiveMeasure m ∧ m Set.univ = 1 ∧ Garrido.IsInvariant J m

/-- `J` is **inner amenable** (p. 3): there is a conjugation-invariant mean on `J ∖ {e}`, here a
finitely additive probability measure on all subsets of `J`, giving `{e}` measure `0`, invariant
under conjugation. -/
def IsInnerAmenable (J : Type*) [Group J] : Prop :=
  ∃ m : Set J → ENNReal,
    Garrido.IsFinitelyAdditiveMeasure m ∧ m Set.univ = 1 ∧ m {1} = 0 ∧
    ∀ (g : J) (s : Set J), m ((fun x => g * x * g⁻¹) '' s) = m s

section MeasuredRelations

open MeasureTheory

/-! ### Amenable measured equivalence relations (Connes–Feldman–Weiss [17], Def. 5–6, p. 437;
Schmidt [19], Def. 1.4). Monod, p. 2: "a measurable equivalence relation with countable classes is
amenable if there is an a.e. defined measurable assignment of a mean on the orbit of each point in
such a way that the means of two equivalent points coincide." -/

section Relations

variable {X : Type*} [MeasurableSpace X]

/-- A set `S ⊆ X × X` is null for the left counting measure of `R` (CFW p. 436) when the set of
first coordinates of `S ∩ R` is `μ`-null. -/
def RelNull (μ : Measure X) (R : Set (X × X)) (S : Set (X × X)) : Prop :=
  μ (Prod.fst '' (S ∩ R)) = 0

/-- A bounded measurable function on the relation `R`, a representative of an element of
`L^∞(R, m)`. -/
def IsBddMeasOn (R : Set (X × X)) (f : X × X → ℝ) : Prop :=
  Measurable f ∧ ∃ C : ℝ, ∀ p ∈ R, |f p| ≤ C

/-- A partial transformation of `R` (CFW p. 436): a measurable isomorphism between measurable
subsets of `X` whose graph lies in `R`. -/
structure PartialTransformation (R : Set (X × X)) where
  dom : Set X
  cod : Set X
  measurableSet_dom : MeasurableSet dom
  measurableSet_cod : MeasurableSet cod
  e : dom ≃ᵐ cod
  graph_subset : ∀ a : dom, ((a : X), (e a : X)) ∈ R

namespace PartialTransformation

variable {R : Set (X × X)} (φ : PartialTransformation R)

/-- CFW p. 436: `f^φ (y, x) = f (φ⁻¹ y, x)` for `y` in the image of `φ`, and `0` otherwise. -/
noncomputable def shiftRel (f : X × X → ℝ) : X × X → ℝ := fun p =>
  open Classical in if h : p.1 ∈ φ.cod then f ((φ.e.symm ⟨p.1, h⟩ : X), p.2) else 0

/-- CFW p. 436: `F^φ = F ∘ φ⁻¹` on the image of `φ`, and `0` elsewhere. -/
noncomputable def shiftBase (F : X → ℝ) : X → ℝ := fun y =>
  open Classical in if h : y ∈ φ.cod then F (φ.e.symm ⟨y, h⟩ : X) else 0

end PartialTransformation

/-- A left invariant mean on `R` (CFW Def. 5, p. 437): a positive map `P` with `P(1) = 1` from
`L^∞(R, m)` to `L^∞(X, μ)` with `P(f^φ) = (P f)^φ` for every partial transformation `φ` of `R`.
It is given on bounded measurable representatives and respects equality off a null set of the
left counting measure. -/
structure IsLeftInvariantMean (μ : Measure X) (R : Set (X × X)) (P : (X × X → ℝ) → X → ℝ) :
    Prop where
  aemeasurable : ∀ f, IsBddMeasOn R f → AEMeasurable (P f) μ
  congr : ∀ f g, IsBddMeasOn R f → IsBddMeasOn R g → RelNull μ R {p | f p ≠ g p} →
    P f =ᵐ[μ] P g
  add : ∀ f g, IsBddMeasOn R f → IsBddMeasOn R g → P (f + g) =ᵐ[μ] P f + P g
  smul : ∀ (c : ℝ) f, IsBddMeasOn R f → P (c • f) =ᵐ[μ] c • P f
  nonneg : ∀ f, IsBddMeasOn R f → (∀ p ∈ R, 0 ≤ f p) → ∀ᵐ x ∂μ, 0 ≤ P f x
  one : P 1 =ᵐ[μ] 1
  invariant : ∀ (φ : PartialTransformation R) f, IsBddMeasOn R f →
    P (φ.shiftRel f) =ᵐ[μ] φ.shiftBase (P f)

/-- CFW Def. 6, p. 437: `R` is amenable if it possesses a left invariant mean. -/
def IsAmenableRel (μ : Measure X) (R : Set (X × X)) : Prop :=
  ∃ P, IsLeftInvariantMean μ R P

end Relations

/-- The Borel σ-algebra on `𝐏¹ = OnePoint ℝ`. -/
instance measurableSpaceP1 : MeasurableSpace (OnePoint ℝ) := borel _

instance : BorelSpace (OnePoint ℝ) := ⟨rfl⟩

/-- The Lebesgue measure class on `𝐏¹`: Lebesgue measure on `ℝ ⊆ 𝐏¹`, with `∞` null. -/
noncomputable def volP1 : Measure (OnePoint ℝ) :=
  Measure.map ((↑) : ℝ → OnePoint ℝ) volume

end MeasuredRelations

end Monod
Source
Monod, N., Groups of piecewise projective homeomorphisms, Proc. Natl. Acad. Sci. USA 110 (2013) 4524–4527, https://doi.org/10.1073/pnas.1218426110 (arXiv:1209.5229v2, whose page numbers are used), p. 1–3, the construction of G, H, P_A, G(A), H(A) (p. 1), the rational-breakpoint variants (p. 2), co-amenable subgroups and inner amenability (p. 3)
Read-back

What the Lean code literally says, in plain math · claude-opus-5-5

Read-back: definitions for piecewise projective homeomorphisms of the projective line, coamenability, inner amenability and amenable measured relations

This item is a bundle of definitions. They share the following vocabulary.

The projective line. Throughout, P1\mathbb{P}^1P1 denotes R∪{∞}\mathbb{R}\cup\{\infty\}R∪{∞}, the one-point compactification of R\mathbb{R}R with its usual topology: a set U⊆P1U\subseteq\mathbb{P}^1U⊆P1 is open exactly when U∩RU\cap\mathbb{R}U∩R is open in R\mathbb{R}R and, if ∞∈U\infty\in U∞∈U, the complement R∖U\mathbb{R}\setminus UR∖U is compact. So the neighbourhoods of a real point xxx are the sets containing an open interval around xxx, and the neighbourhoods of ∞\infty∞ are the sets containing {∞}∪(R∖K)\{\infty\}\cup(\mathbb{R}\setminus K){∞}∪(R∖K) for some compact K⊆RK\subseteq\mathbb{R}K⊆R.

The homeomorphism group. Homeo(P1)\mathrm{Homeo}(\mathbb{P}^1)Homeo(P1) denotes the group of all homeomorphisms P1→P1\mathbb{P}^1\to\mathbb{P}^1P1→P1, with product (fg)(x)=f(g(x))(f g)(x)=f(g(x))(fg)(x)=f(g(x)) (composition, ggg applied first), identity the identity map, and inverse the inverse map. "The subgroup generated by a set SSS" means the smallest subgroup of Homeo(P1)\mathrm{Homeo}(\mathbb{P}^1)Homeo(P1) containing SSS. H1∩H2H_1\cap H_2H1​∩H2​ is the intersection of subgroups.

Subrings. Throughout, AAA is an arbitrary subring of R\mathbb{R}R (containing 111). Two particular subrings appear: the whole of R\mathbb{R}R, and the smallest subring, which is Z\mathbb{Z}Z (the set of real numbers of the form n⋅1n\cdot 1n⋅1, n∈Zn\in\mathbb{Z}n∈Z). SL2(A)\mathrm{SL}_2(A)SL2​(A) is the group of 2×22\times 22×2 matrices with entries in AAA and determinant 111.


1. The embedding SL2(A)→GL2(R)\mathrm{SL}_2(A)\to\mathrm{GL}_2(\mathbb{R})SL2​(A)→GL2​(R)

For a subring A⊆RA\subseteq\mathbb{R}A⊆R, this is the group homomorphism SL2(A)→GL2(R)\mathrm{SL}_2(A)\to\mathrm{GL}_2(\mathbb{R})SL2​(A)→GL2​(R) that sends a matrix with entries in AAA to the same matrix regarded as a real matrix (entries included into R\mathbb{R}R). It is injective.

2. The Möbius action of SL2(A)\mathrm{SL}_2(A)SL2​(A) on P1\mathbb{P}^1P1

For a subring AAA (left implicit, determined by ggg), a matrix

g=(abcd)∈SL2(A)g=\begin{pmatrix}a&b\\ c&d\end{pmatrix}\in\mathrm{SL}_2(A)g=(ac​bd​)∈SL2​(A)

and a point x∈P1x\in\mathbb{P}^1x∈P1, the point g⋅x∈P1g\cdot x\in\mathbb{P}^1g⋅x∈P1 is obtained by regarding ggg as a real matrix and applying the standard action of GL2(R)\mathrm{GL}_2(\mathbb{R})GL2​(R) on P1\mathbb{P}^1P1 (via the identification of P1\mathbb{P}^1P1 with the lines in R2\mathbb{R}^2R2: a real ttt corresponds to the line through (t,1)(t,1)(t,1) and ∞\infty∞ to the line through (1,0)(1,0)(1,0); ggg acts on column vectors). Explicitly, for real ttt,

g⋅t={∞if ct+d=0,at+bct+dotherwise,g⋅∞={∞if c=0,a/cotherwise.g\cdot t=\begin{cases}\infty & \text{if } ct+d=0,\\[2pt] \dfrac{at+b}{ct+d} & \text{otherwise,}\end{cases} \qquad g\cdot\infty=\begin{cases}\infty & \text{if } c=0,\\[2pt] a/c & \text{otherwise.}\end{cases}g⋅t=⎩⎨⎧​∞ct+dat+b​​if ct+d=0,otherwise,​g⋅∞={∞a/c​if c=0,otherwise.​

In particular ggg and −g-g−g act identically.

3. Hyperbolic elements

A matrix g=(abcd)∈SL2(A)g=\begin{pmatrix}a&b\\ c&d\end{pmatrix}\in\mathrm{SL}_2(A)g=(ac​bd​)∈SL2​(A) is called hyperbolic when its trace, computed in AAA and regarded as a real number, has absolute value strictly greater than 222:

∣a+d∣>2.|a+d|>2 .∣a+d∣>2.

4. The set PAP_APA​ of hyperbolic fixed points

For a subring AAA,

PA={ p∈P1:there is a hyperbolic g∈SL2(A) with g⋅p=p }.P_A=\{\,p\in\mathbb{P}^1 : \text{there is a hyperbolic } g\in\mathrm{SL}_2(A)\text{ with } g\cdot p=p\,\}.PA​={p∈P1:there is a hyperbolic g∈SL2​(A) with g⋅p=p}.

The point ∞\infty∞ is allowed.

5. Piecewise projective on a set of breakpoints

Given a subring AAA, a set E⊆P1E\subseteq\mathbb{P}^1E⊆P1 and a homeomorphism fff of P1\mathbb{P}^1P1, fff is piecewise AAA-projective with breakpoints in EEE when there exists a finite set B⊆P1B\subseteq\mathbb{P}^1B⊆P1 with B⊆EB\subseteq EB⊆E such that for every point x∈P1∖Bx\in\mathbb{P}^1\setminus Bx∈P1∖B there exists a matrix g∈SL2(A)g\in\mathrm{SL}_2(A)g∈SL2​(A) (which may depend on xxx) and a neighbourhood UUU of xxx in P1\mathbb{P}^1P1 with

f(y)=g⋅yfor all y∈U.f(y)=g\cdot y\quad\text{for all } y\in U.f(y)=g⋅yfor all y∈U.

Nothing is required at the points of BBB themselves. BBB may be empty (then fff is locally Möbius at every point). The condition allows x=∞x=\inftyx=∞, in which case UUU is a neighbourhood of ∞\infty∞ as described above.

6. Piecewise projective (breakpoints in PAP_APA​)

A homeomorphism fff of P1\mathbb{P}^1P1 is piecewise AAA-projective when it is piecewise AAA-projective with breakpoints in PAP_APA​ (definition 5 with E=PAE=P_AE=PA​ from definition 4): some finite set B⊆PAB\subseteq P_AB⊆PA​ exists such that every x∉Bx\notin Bx∈/B has a neighbourhood on which fff agrees with the Möbius action of a single element of SL2(A)\mathrm{SL}_2(A)SL2​(A).

7. The group GppG_{pp}Gpp​

GppG_{pp}Gpp​ is the subgroup of Homeo(P1)\mathrm{Homeo}(\mathbb{P}^1)Homeo(P1) generated by all homeomorphisms fff that are piecewise R\mathbb{R}R-projective with breakpoints anywhere, i.e. definition 5 with A=RA=\mathbb{R}A=R and E=P1E=\mathbb{P}^1E=P1: there is a finite set B⊆P1B\subseteq\mathbb{P}^1B⊆P1 such that every x∉Bx\notin Bx∈/B has a neighbourhood on which fff coincides with y↦g⋅yy\mapsto g\cdot yy↦g⋅y for some g∈SL2(R)g\in\mathrm{SL}_2(\mathbb{R})g∈SL2​(R).

8. The group G(A)G(A)G(A)

For a subring AAA, G(A)G(A)G(A) is the subgroup of Homeo(P1)\mathrm{Homeo}(\mathbb{P}^1)Homeo(P1) generated by the set of homeomorphisms fff such that both

  • f∈Gppf\in G_{pp}f∈Gpp​, and
  • fff is piecewise AAA-projective in the sense of definition 6 (finitely many breakpoints, all in PAP_APA​; locally given by elements of SL2(A)\mathrm{SL}_2(A)SL2​(A) elsewhere).

(Any fff satisfying the second condition also satisfies the first, since an element of SL2(A)\mathrm{SL}_2(A)SL2​(A) is also an element of SL2(R)\mathrm{SL}_2(\mathbb{R})SL2​(R) acting by the same map, and PA⊆P1P_A\subseteq\mathbb{P}^1PA​⊆P1; so the generating set is exactly the set of piecewise AAA-projective homeomorphisms.)

9. The stabiliser of ∞\infty∞

Fix(∞)\mathrm{Fix}(\infty)Fix(∞) is the subgroup of Homeo(P1)\mathrm{Homeo}(\mathbb{P}^1)Homeo(P1) consisting of all homeomorphisms fff with f(∞)=∞f(\infty)=\inftyf(∞)=∞.

10. The group H(A)H(A)H(A)

For a subring AAA,

H(A)=G(A)∩Fix(∞),H(A)=G(A)\cap\mathrm{Fix}(\infty),H(A)=G(A)∩Fix(∞),

the elements of G(A)G(A)G(A) (definition 8) that fix ∞\infty∞.

11. The group HppH_{pp}Hpp​

Hpp=Gpp∩Fix(∞),H_{pp}=G_{pp}\cap\mathrm{Fix}(\infty),Hpp​=Gpp​∩Fix(∞),

the elements of GppG_{pp}Gpp​ (definition 7) that fix ∞\infty∞.

12. The rational points

P1(Q)={∞}∪{ q:q∈Q }⊆P1,\mathbb{P}^1(\mathbb{Q})=\{\infty\}\cup\{\,q : q\in\mathbb{Q}\,\}\subseteq\mathbb{P}^1,P1(Q)={∞}∪{q:q∈Q}⊆P1,

the point at infinity together with all rational numbers (viewed as real numbers, viewed as points of P1\mathbb{P}^1P1).

13. The group GQG_{\mathbb{Q}}GQ​

GQG_{\mathbb{Q}}GQ​ is the subgroup of Homeo(P1)\mathrm{Homeo}(\mathbb{P}^1)Homeo(P1) generated by the set of homeomorphisms fff such that both

  • f∈Gppf\in G_{pp}f∈Gpp​, and
  • fff is piecewise Z\mathbb{Z}Z-projective with breakpoints in P1(Q)\mathbb{P}^1(\mathbb{Q})P1(Q) (definition 5 with A=ZA=\mathbb{Z}A=Z, the smallest subring of R\mathbb{R}R, and E=P1(Q)E=\mathbb{P}^1(\mathbb{Q})E=P1(Q)): there is a finite set BBB of points of P1(Q)\mathbb{P}^1(\mathbb{Q})P1(Q) such that every x∉Bx\notin Bx∈/B has a neighbourhood on which fff coincides with y↦g⋅yy\mapsto g\cdot yy↦g⋅y for some g∈SL2(Z)g\in\mathrm{SL}_2(\mathbb{Z})g∈SL2​(Z).

Note that the breakpoint set here is P1(Q)\mathbb{P}^1(\mathbb{Q})P1(Q), not the set PZP_{\mathbb{Z}}PZ​ of definition 4. (As in definition 8, the second condition implies the first.)

14. The group HQH_{\mathbb{Q}}HQ​

HQ=GQ∩Fix(∞),H_{\mathbb{Q}}=G_{\mathbb{Q}}\cap\mathrm{Fix}(\infty),HQ​=GQ​∩Fix(∞),

the elements of GQG_{\mathbb{Q}}GQ​ that fix ∞\infty∞.

15. Pointwise fixator of a set inside a subgroup

For a subgroup J≤Homeo(P1)J\le\mathrm{Homeo}(\mathbb{P}^1)J≤Homeo(P1) and a set E⊆P1E\subseteq\mathbb{P}^1E⊆P1, this is the subgroup of JJJ (a subgroup of the group JJJ itself, not of Homeo(P1)\mathrm{Homeo}(\mathbb{P}^1)Homeo(P1))

JE={ h∈J:h(x)=x for every x∈E }.J_E=\{\,h\in J : h(x)=x\text{ for every } x\in E\,\}.JE​={h∈J:h(x)=x for every x∈E}.

If EEE is empty this is all of JJJ.

16. Coamenable subgroup

Let JJJ be any group and KKK a subgroup of JJJ. Write J/KJ/KJ/K for the set of left cosets aKaKaK (a∈Ja\in Ja∈J), on which JJJ acts by left multiplication, g⋅(aK)=(ga)Kg\cdot(aK)=(ga)Kg⋅(aK)=(ga)K; for a set S⊆J/KS\subseteq J/KS⊆J/K put g⋅S={g⋅s:s∈S}g\cdot S=\{g\cdot s : s\in S\}g⋅S={g⋅s:s∈S}. KKK is coamenable in JJJ when there exists a function

m:{all subsets of J/K}→[0,∞]m:\{\text{all subsets of } J/K\}\to[0,\infty]m:{all subsets of J/K}→[0,∞]

such that

  1. m(∅)=0m(\varnothing)=0m(∅)=0, and m(S∪T)=m(S)+m(T)m(S\cup T)=m(S)+m(T)m(S∪T)=m(S)+m(T) whenever S,T⊆J/KS,T\subseteq J/KS,T⊆J/K are disjoint (finite additivity, on arbitrary subsets);
  2. m(J/K)=1m(J/K)=1m(J/K)=1;
  3. m(g⋅S)=m(S)m(g\cdot S)=m(S)m(g⋅S)=m(S) for every g∈Jg\in Jg∈J and every S⊆J/KS\subseteq J/KS⊆J/K.

(From 1 and 2, mmm is monotone and takes values in [0,1][0,1][0,1].) No hypothesis on JJJ or KKK beyond being a group and a subgroup; in particular J/KJ/KJ/K may be a single point, in which case the condition holds.

17. Inner amenable group

A group JJJ is inner amenable when there exists a function

m:{all subsets of J}→[0,∞]m:\{\text{all subsets of } J\}\to[0,\infty]m:{all subsets of J}→[0,∞]

such that

  1. m(∅)=0m(\varnothing)=0m(∅)=0 and m(S∪T)=m(S)+m(T)m(S\cup T)=m(S)+m(T)m(S∪T)=m(S)+m(T) for all disjoint S,T⊆JS,T\subseteq JS,T⊆J;
  2. m(J)=1m(J)=1m(J)=1;
  3. m({1})=0m(\{1\})=0m({1})=0, where 111 is the identity of JJJ;
  4. m(gSg−1)=m(S)m(gSg^{-1})=m(S)m(gSg−1)=m(S) for every g∈Jg\in Jg∈J and every S⊆JS\subseteq JS⊆J, where gSg−1={gxg−1:x∈S}gSg^{-1}=\{gxg^{-1}:x\in S\}gSg−1={gxg−1:x∈S}.

If JJJ is the trivial group, conditions 2 and 3 contradict each other, so the trivial group is not inner amenable under this definition.


The remaining definitions concern a set XXX equipped with a σ\sigmaσ-algebra (a measurable space), in an arbitrary universe; X×XX\times XX×X carries the product σ\sigmaσ-algebra and R\mathbb{R}R its Borel σ\sigmaσ-algebra. A measure μ\muμ on XXX is a countably additive measure on this σ\sigmaσ-algebra; μ(S)\mu(S)μ(S) for an arbitrary (possibly non-measurable) set S⊆XS\subseteq XS⊆X means its outer measure, inf⁡{μ(T):T⊇S measurable}\inf\{\mu(T): T\supseteq S \text{ measurable}\}inf{μ(T):T⊇S measurable}. "μ\muμ-a.e." means outside a μ\muμ-null set. R⊆X×XR\subseteq X\times XR⊆X×X is an arbitrary subset (not assumed measurable, nor an equivalence relation). π1:X×X→X\pi_1:X\times X\to Xπ1​:X×X→X is the first-coordinate projection.

18. RRR-null sets

For a measure μ\muμ on XXX, a set R⊆X×XR\subseteq X\times XR⊆X×X and a set S⊆X×XS\subseteq X\times XS⊆X×X, SSS is RRR-null (for μ\muμ) when

μ(π1(S∩R))=0,\mu\bigl(\pi_1(S\cap R)\bigr)=0,μ(π1​(S∩R))=0,

i.e. the set of first coordinates of points of RRR lying in SSS has μ\muμ-(outer) measure zero.

19. Bounded measurable on RRR

For R⊆X×XR\subseteq X\times XR⊆X×X and f:X×X→Rf:X\times X\to\mathbb{R}f:X×X→R, fff is bounded measurable on RRR when

  • fff is measurable on all of X×XX\times XX×X (product σ\sigmaσ-algebra to Borel sets), and
  • there is a real constant CCC with ∣f(p)∣≤C|f(p)|\le C∣f(p)∣≤C for every p∈Rp\in Rp∈R.

Boundedness is required only on RRR; measurability is required everywhere.

20. Partial transformations of RRR

For R⊆X×XR\subseteq X\times XR⊆X×X, a partial transformation of RRR consists of

  • two measurable sets D,C′⊆XD,C'\subseteq XD,C′⊆X (the domain and codomain; either may be empty),
  • a bijection φ:D→C′\varphi:D\to C'φ:D→C′ that is measurable with measurable inverse, where DDD and C′C'C′ carry the subspace σ\sigmaσ-algebras (a measurable isomorphism),

subject to the condition that its graph lies in RRR: (x,φ(x))∈R(x,\varphi(x))\in R(x,φ(x))∈R for every x∈Dx\in Dx∈D.

Associated to a partial transformation φ:D→C′\varphi:D\to C'φ:D→C′ are two shift operations.

  • For f:X×X→Rf:X\times X\to\mathbb{R}f:X×X→R, the function φ∗f:X×X→R\varphi_*f:X\times X\to\mathbb{R}φ∗​f:X×X→R is
φ∗f(y,z)={f(φ−1(y), z)if y∈C′,0otherwise.\varphi_*f(y,z)=\begin{cases} f\bigl(\varphi^{-1}(y),\,z\bigr) & \text{if } y\in C',\\ 0 & \text{otherwise.}\end{cases}φ∗​f(y,z)={f(φ−1(y),z)0​if y∈C′,otherwise.​

(It is defined for every pair (y,z)(y,z)(y,z), whether or not (φ−1(y),z)(\varphi^{-1}(y),z)(φ−1(y),z) or (y,z)(y,z)(y,z) lies in RRR.)

  • For F:X→RF:X\to\mathbb{R}F:X→R, the function φ∗F:X→R\varphi_*F:X\to\mathbb{R}φ∗​F:X→R is
φ∗F(y)={F(φ−1(y))if y∈C′,0otherwise.\varphi_*F(y)=\begin{cases} F\bigl(\varphi^{-1}(y)\bigr) & \text{if } y\in C',\\ 0 & \text{otherwise.}\end{cases}φ∗​F(y)={F(φ−1(y))0​if y∈C′,otherwise.​

21. Left-invariant mean on RRR

Let μ\muμ be a measure on XXX, R⊆X×XR\subseteq X\times XR⊆X×X, and let P\mathcal{P}P be any map sending every function f:X×X→Rf:X\times X\to\mathbb{R}f:X×X→R (not only the bounded measurable ones) to a function Pf:X→R\mathcal{P}f:X\to\mathbb{R}Pf:X→R. P\mathcal{P}P is a left-invariant mean on RRR (for μ\muμ) when all of the following hold. In each item, f,gf,gf,g range over functions that are bounded measurable on RRR (definition 19), except where said otherwise; all equalities and inequalities between functions on XXX are required only μ\muμ-almost everywhere.

  1. Measurability. For every bounded measurable fff on RRR, Pf\mathcal{P}fPf is μ\muμ-almost-everywhere measurable (it agrees μ\muμ-a.e. with a measurable function X→RX\to\mathbb{R}X→R).
  2. Null-set invariance. If f,gf,gf,g are bounded measurable on RRR and the set {p∈X×X:f(p)≠g(p)}\{p\in X\times X: f(p)\neq g(p)\}{p∈X×X:f(p)=g(p)} is RRR-null (definition 18) — that is, μ({x:∃z, (x,z)∈R, f(x,z)≠g(x,z)})=0\mu\bigl(\{x : \exists z,\ (x,z)\in R,\ f(x,z)\neq g(x,z)\}\bigr)=0μ({x:∃z, (x,z)∈R, f(x,z)=g(x,z)})=0 — then Pf=Pg\mathcal{P}f=\mathcal{P}gPf=Pg μ\muμ-a.e.
  3. Additivity. P(f+g)=Pf+Pg\mathcal{P}(f+g)=\mathcal{P}f+\mathcal{P}gP(f+g)=Pf+Pg μ\muμ-a.e.
  4. Homogeneity. For every real ccc, P(cf)=c Pf\mathcal{P}(cf)=c\,\mathcal{P}fP(cf)=cPf μ\muμ-a.e.
  5. Positivity. If fff is bounded measurable on RRR and f(p)≥0f(p)\ge 0f(p)≥0 for every p∈Rp\in Rp∈R (no condition off RRR), then Pf≥0\mathcal{P}f\ge 0Pf≥0 μ\muμ-a.e.
  6. Normalisation. P1=1\mathcal{P}\mathbf{1}=\mathbf{1}P1=1 μ\muμ-a.e., where 1\mathbf{1}1 denotes the constant function 111 (on X×XX\times XX×X, respectively on XXX).
  7. Invariance. For every partial transformation φ\varphiφ of RRR (definition 20) and every fff bounded measurable on RRR,
P(φ∗f)=φ∗(Pf)μ-a.e.,\mathcal{P}(\varphi_*f)=\varphi_*(\mathcal{P}f)\quad\mu\text{-a.e.},P(φ∗​f)=φ∗​(Pf)μ-a.e.,

with both shifts as in definition 20. (No assumption is made that φ∗f\varphi_*fφ∗​f is itself bounded measurable on RRR.)

The "a.e." set may depend on fff, ggg, ccc and φ\varphiφ. On functions that are not bounded measurable on RRR, P\mathcal{P}P is constrained only through items 6 and 7 as stated. If μ\muμ is the zero measure, every "a.e." condition holds automatically.

22. Amenable measured relation

For a measure μ\muμ on XXX and R⊆X×XR\subseteq X\times XR⊆X×X, RRR is amenable (for μ\muμ) when there exists a left-invariant mean on RRR for μ\muμ in the sense of definition 21, i.e. some map P\mathcal{P}P from functions X×X→RX\times X\to\mathbb{R}X×X→R to functions X→RX\to\mathbb{R}X→R satisfying items 1–7.


23. The σ\sigmaσ-algebra on P1\mathbb{P}^1P1

P1\mathbb{P}^1P1 is given the Borel σ\sigmaσ-algebra of its topology (the σ\sigmaσ-algebra generated by the open sets described at the top). This is declared as the standard σ\sigmaσ-algebra on P1\mathbb{P}^1P1 from this point on (globally, not only inside this item), and P1\mathbb{P}^1P1 is recorded as a Borel space with it.

24. The measure on P1\mathbb{P}^1P1

The measure λP1\lambda_{\mathbb{P}^1}λP1​ on P1\mathbb{P}^1P1 (with the Borel σ\sigmaσ-algebra of definition 23) is the pushforward of Lebesgue measure λ\lambdaλ on R\mathbb{R}R under the inclusion ι:R→P1\iota:\mathbb{R}\to\mathbb{P}^1ι:R→P1:

λP1(S)=λ(ι−1(S))=λ(S∩R)for Borel S⊆P1.\lambda_{\mathbb{P}^1}(S)=\lambda\bigl(\iota^{-1}(S)\bigr)=\lambda(S\cap\mathbb{R})\qquad\text{for Borel } S\subseteq\mathbb{P}^1 .λP1​(S)=λ(ι−1(S))=λ(S∩R)for Borel S⊆P1.

(The inclusion is continuous, hence measurable, so the pushforward is the genuine one.) In particular λP1({∞})=0\lambda_{\mathbb{P}^1}(\{\infty\})=0λP1​({∞})=0 and λP1(P1)=∞\lambda_{\mathbb{P}^1}(\mathbb{P}^1)=\inftyλP1​(P1)=∞: it is an infinite measure, not a probability measure.

Human review
  • Endorsed by Shuze Chen · Sep 30, 2026

    Confirmed by the moderator at approval.

  • Endorsed by dbenbenn · Sep 30, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me