Prove2Me
Navigate
DiscoverFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

Theorem 2.7 — amenability, an invariant mean, non-paradoxicality and the invariant extension property are equivalent

Proved
Garrido.isAmenable_tfae_four

by dbenbenn · Sep 23, 2026 · Mathlib 0df444a (Lean v4.33.1)

amenabilityfunctional-analysisgroup-theory

For a group GGG the following four are equivalent:

  1. GGG is amenable — there is a finitely additive left-invariant probability measure on P(G)\mathcal{P}(G)P(G);
  2. there is a left-invariant mean on ℓ∞(G)\ell^\infty(G)ℓ∞(G);
  3. GGG is not paradoxical;
  4. GGG has the invariant extension property.

This is the source's running list of equivalent definitions extended by the fourth clause; the first three are the same as in Theorem 1.15, which the source states separately.

Preamble
import Mathlib
import Definitions.Def_Garrido_Equidecomposability
import Definitions.Def_Garrido_Amenability
Formal statement
namespace Garrido

theorem isAmenable_tfae_four (G : Type*) [Group G] :
    [IsAmenable G,
      HasInvariantMean G,
      ¬ IsParadoxical G (Set.univ : Set G),
      HasInvariantExtensionProperty G].TFAE := by
  sorry

end Garrido
Source
A. Garrido, "An introduction to amenable groups", lecture notes, Oxford Advanced Class in Algebra, Michaelmas 2013 (PDF, Feb 2015), p. 7, Theorem 2.7; https://web.archive.org/web/20260805000803/https://www.math.uni-duesseldorf.de/~garrido/amenable.pdf
Read-back

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

Read-back

Setting. Let GGG be an arbitrary group (any size, including the trivial group; no topology, finiteness or countability assumption). The statement is also parametrised by a second universe level vvv, which enters only through condition (4) below; see the remark there. For each such GGG (and each choice of vvv), the statement asserts that the following four propositions are pairwise equivalent: for every two of them, each holds if and only if the other does.

Throughout, [0,∞][0,\infty][0,∞] denotes the extended non-negative reals, with a+∞=∞+a=∞a + \infty = \infty + a = \inftya+∞=∞+a=∞. For g∈Gg \in Gg∈G acting on a set XXX and S⊆XS \subseteq XS⊆X, write gS={ g⋅x:x∈S }gS = \{\, g\cdot x : x \in S \,\}gS={g⋅x:x∈S} (the image of SSS under x↦g⋅xx \mapsto g\cdot xx↦g⋅x). When X=GX = GX=G, the action is left multiplication, g⋅x=gxg\cdot x = gxg⋅x=gx, so gS={gx:x∈S}gS = \{gx : x \in S\}gS={gx:x∈S}.

A function m:P(X)→[0,∞]m : \mathcal P(X) \to [0,\infty]m:P(X)→[0,∞] defined on all subsets of XXX is called finitely additive here when

m(∅)=0andm(S∪T)=m(S)+m(T)  for all disjoint S,T⊆X.m(\varnothing) = 0 \quad\text{and}\quad m(S \cup T) = m(S) + m(T)\ \text{ for all disjoint } S, T \subseteq X.m(∅)=0andm(S∪T)=m(S)+m(T)  for all disjoint S,T⊆X.

It is called GGG-invariant when m(gS)=m(S)m(gS) = m(S)m(gS)=m(S) for every g∈Gg \in Gg∈G and every S⊆XS \subseteq XS⊆X.


(1) Amenability (measure form). There exists a function m:P(G)→[0,∞]m : \mathcal P(G) \to [0,\infty]m:P(G)→[0,∞], defined on all subsets of GGG, that is finitely additive, satisfies m(G)=1m(G) = 1m(G)=1, and is invariant under left translation:

m(gS)=m(S)for all g∈G, S⊆G, where gS={gx:x∈S}.m(gS) = m(S)\qquad\text{for all } g\in G,\ S \subseteq G,\ \text{where } gS = \{gx : x\in S\}.m(gS)=m(S)for all g∈G, S⊆G, where gS={gx:x∈S}.

(2) Invariant mean. Let ℓ∞(G)\ell^\infty(G)ℓ∞(G) be the real vector space of bounded functions f:G→Rf : G \to \mathbb Rf:G→R (bounded meaning sup⁡h∣f(h)∣<∞\sup_{h}|f(h)| < \inftysuph​∣f(h)∣<∞), with pointwise addition and scalar multiplication. For g∈Gg \in Gg∈G and f∈ℓ∞(G)f \in \ell^\infty(G)f∈ℓ∞(G) define the left shift

(g⋅f)(h)=f(g−1h)(h∈G).(g\cdot f)(h) = f(g^{-1}h)\qquad (h \in G).(g⋅f)(h)=f(g−1h)(h∈G).

Condition (2) says: there exists an R\mathbb RR-linear map M:ℓ∞(G)→RM : \ell^\infty(G) \to \mathbb RM:ℓ∞(G)→R (no continuity is required) such that

  • (positivity) if f(h)≥0f(h) \ge 0f(h)≥0 for all h∈Gh \in Gh∈G, then M(f)≥0M(f) \ge 0M(f)≥0;
  • (normalisation) if f(h)=1f(h) = 1f(h)=1 for all h∈Gh\in Gh∈G, then M(f)=1M(f) = 1M(f)=1;
  • (left invariance) M(g⋅f)=M(f)M(g\cdot f) = M(f)M(g⋅f)=M(f) for every g∈Gg\in Gg∈G and every f∈ℓ∞(G)f \in \ell^\infty(G)f∈ℓ∞(G).

(3) GGG is not paradoxical under its own left-multiplication action. This is the negation of the following statement:

There exist subsets A,B⊆GA, B \subseteq GA,B⊆G such that A≠GA \ne GA=G, B≠GB \ne GB=G, A∩B=∅A \cap B = \varnothingA∩B=∅, AAA is equidecomposable with GGG, and BBB is equidecomposable with GGG.

Here, for subsets P,Q⊆GP, Q \subseteq GP,Q⊆G, "PPP is equidecomposable with QQQ" (with respect to GGG acting on itself by left multiplication g⋅x=gxg\cdot x = gxg⋅x=gx) means: there exist a bijection φ:P→Q\varphi : P \to Qφ:P→Q and a finite set F⊆GF \subseteq GF⊆G such that for every a∈Pa \in Pa∈P there is some g∈Fg \in Fg∈F with φ(a)=ga\varphi(a) = gaφ(a)=ga. (Formally φ\varphiφ is recorded as a map G→GG\to GG→G together with a map G→GG \to GG→G that are mutually inverse between PPP and QQQ, map PPP into QQQ and QQQ into PPP; the values outside PPP, resp. QQQ, are irrelevant. The finite set FFF is only required to exist; the pieces are implicit, namely the sets {a∈P:φ(a)=ga}\{a \in P : \varphi(a) = ga\}{a∈P:φ(a)=ga}, which need not be disjoint for different ggg.) The conditions A⊆GA \subseteq GA⊆G, B⊆GB\subseteq GB⊆G are also stated but are automatic.

So (3) says: there are no two disjoint proper subsets A,BA, BA,B of GGG each of which is equidecomposable, in the above sense, with all of GGG.

(4) Invariant extension property. For every set XXX in the universe level max⁡(u,v)\max(u, v)max(u,v) (where uuu is the level of GGG), every action of GGG on XXX (a genuine group action: 1⋅x=x1\cdot x = x1⋅x=x and (gh)⋅x=g⋅(h⋅x)(gh)\cdot x = g\cdot(h\cdot x)(gh)⋅x=g⋅(h⋅x)), every family R\mathcal RR of subsets of XXX, and every two functions μ,ν:P(X)→[0,∞]\mu, \nu : \mathcal P(X) \to [0,\infty]μ,ν:P(X)→[0,∞] defined on all subsets of XXX: if

  • (a) R\mathcal RR is closed under the action: S∈R⇒gS∈RS \in \mathcal R \Rightarrow gS \in \mathcal RS∈R⇒gS∈R for all g∈Gg \in Gg∈G;
  • (b) μ\muμ is invariant on R\mathcal RR: μ(gS)=μ(S)\mu(gS) = \mu(S)μ(gS)=μ(S) for all g∈Gg\in Gg∈G, S∈RS \in \mathcal RS∈R;
  • (c) ν\nuν agrees with μ\muμ on R\mathcal RR: ν(S)=μ(S)\nu(S) = \mu(S)ν(S)=μ(S) for all S∈RS \in \mathcal RS∈R;
  • (d) ν\nuν is finitely additive on all of P(X)\mathcal P(X)P(X),

then there exists μˉ:P(X)→[0,∞]\bar\mu : \mathcal P(X) \to [0,\infty]μˉ​:P(X)→[0,∞] that is finitely additive on all of P(X)\mathcal P(X)P(X), agrees with μ\muμ on R\mathcal RR (i.e. μˉ(S)=μ(S)\bar\mu(S) = \mu(S)μˉ​(S)=μ(S) for S∈RS\in\mathcal RS∈R), and is GGG-invariant on all subsets: μˉ(gS)=μˉ(S)\bar\mu(gS) = \bar\mu(S)μˉ​(gS)=μˉ​(S) for all g∈Gg \in Gg∈G, S⊆XS\subseteq XS⊆X.

Remarks on (4):

  • Nothing is assumed of R\mathcal RR beyond (a): it need not contain ∅\varnothing∅ or XXX, be closed under unions, complements or intersections, and may be empty. When R=∅\mathcal R = \varnothingR=∅, hypotheses (a)–(c) are vacuous and the conclusion only asks for some finitely additive GGG-invariant μˉ\bar\muμˉ​ on P(X)\mathcal P(X)P(X) (the identically-zero function is one such). More generally, hypotheses (c)–(d) together require that the values of μ\muμ on R\mathcal RR be the values of some finitely additive function on P(X)\mathcal P(X)P(X); they are not a restriction on μ\muμ outside R\mathcal RR.
  • μ\muμ itself is not assumed additive; only its values on R\mathcal RR matter to the conclusion, and those values are forced by (c) to coincide with those of the finitely additive ν\nuν.
  • Values ∞\infty∞ are permitted for μ\muμ, ν\nuν and μˉ\bar\muμˉ​.
  • The quantification over XXX ranges over all types of universe level max⁡(u,v)\max(u,v)max(u,v) only, where vvv is a free universe parameter of the whole statement. The statement is therefore a separate assertion for each vvv: for each fixed vvv, conditions (1), (2), (3) and this level-vvv version of (4) are pairwise equivalent. It does not assert anything that quantifies over all universe levels at once. (Every level max⁡(u,v)\max(u,v)max(u,v) contains a copy of GGG itself, so XXX of the "size" of GGG are always included.)

Summary. For every group GGG (and each universe level vvv):

(1)  ⟺  (2)  ⟺  (3)  ⟺  (4),(1) \iff (2) \iff (3) \iff (4),(1)⟺(2)⟺(3)⟺(4),

where (1) is the existence of a left-invariant, finitely additive, [0,∞][0,\infty][0,∞]-valued set function on all subsets of GGG with total mass 111; (2) is the existence of a positive, normalised, left-shift-invariant real linear functional on bounded real functions on GGG; (3) is the non-existence of a paradoxical decomposition of GGG under left multiplication, as defined above; and (4) is the invariant extension property as stated above.

Human review
  • Endorsed by Shuze Chen · Sep 24, 2026

    Confirmed by the moderator at approval.

  • Endorsed by dbenbenn · Sep 24, 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