II1 -factor representations of Higman-Thompson’s groups Classification of Characters Kostya Medynets United States Naval Academy Annapolis, Maryland Characters of infinite groups Fix an infinite group G . Definition A character of G is a function f : G → C such that (normalization) f (1) = 1 (class function) f (a) = f (bab −1 ) for all a, b ∈ G . (positivity) For any family {g1 , . . . , gn } ⊂ G , the matrix M = {f (gi−1 gj )}ni,j=1 is positive-definite, i.e., for any {c1 , . . . , cn } ⊂ C n n X X i =1 j=1 c̄i cj f (gi−1 gj ) ≥ 0. Where do the characters come from? Consider a unitary representation π : G → B(H) in a Hilbert space H. Denote by Wπ the von Neumann algebra generated by π(G ), i.e., Wπ = (π(G ))′′ . Suppose that Wπ is of finite type, of type In , n < ∞, or II1 . Let τ be the tracial state on Wπ . Set f (g ) = τ (π(g )). Then f is a character. To show P notice that τ (Q) = (Qξ, ξ). P that f is positive-definite, Then ni,j=1 c̄i cj f (gi−1 gj ) = || ci π(gi )ξ||2 for any c1 , . . . , cn . Do we have any other characters? Theorem (Gelfand-Naimark-Segal Construction) If f is a character of an infinite group G , then there is a finite type unitary representation π such that f (g ) = τ (π(g )), where τ is the trace of von Neumann algebra Wπ . Characters form a Choquet simplex. (Decomposition) Extreme characters correspond to factor representations. This means that the algebra generated by the group π(G ) is a factor. So the question is to describe all extreme characters for a group in question. For finite groups, extreme characters are precisely the irreducible characters. Characters and Ergodic Theory Vershik noticed that there is yet another way to produce characters. Suppose that a group G acts on a measure space (X , µ). For g ∈ G , set Fix(g ) = {x ∈ X : g (x) = x}, the set of fixed points. The function f (g ) = µ(Fix(g )) is a character. The function f is central. f (hgh−1 ) = µ(Fix(hgh−1 )) = µ(h(Fix(g ))) = µ(Fix(g )) = f (g ). Normalization. f (1) = µ(X ) = 1. The function f is positive-definite. Characters and Ergodic Theory To show that f is positive-definite, we will use the Feldman-Moore groupoid construction. Let R be the orbit equivalence relation generated by the action of G on (X , µ). R Define the measure µ̄ on R as µ(A) = X card(Ax )dµ(x), where Ax = {(x, y ) ∈ A}. Define the unitary representation π : G → U(L2 (R, µ̄)) by π(g )f (x, y ) = f (g −1 x, y ). Denote by ∆ the diagonal in X × X . Then f (g ) = µ(Fix(g )) = (π(g )1∆ , 1∆ ). Thompson’s group Thompson’s group F is generated by the functions 1.0 1.0 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0.2 0.4 0.6 0.8 1.0 0.2 0.4 0.6 0.8 1.0 The group F consists of all piecewise linear homeomorphisms from the closed interval [0, 1] to itself that are differentiable except at finitely many dyadic rational numbers and such that on the intervals of differentiability the derivatives are powers of 2. The elements of F preserve orientation. The Commutator Subgroup Denote by F ′ the commutator subgroup of F . The group F ′ consists of all homeomorphisms from F that are trivial in neighborhoods of 0 and 1. F ′ is simple. The action of F ′ on X = (0, 1) has no probability invariant measure. Theorem (Dudko-M.) The group F ′ has no non-trivial characters. Every group always admits at least two characters (1) the regular character δ(g ) = 0 if g 6= e and δ(e) = 0 and (2) the identity character ρ(g ) = 1 for every g . Corollaries Theorem (Dudko-M.) The only non-trivial II1 -factor representation of F ′ is the left regular representation. Theorem (Dudko-M.) If a countable group G has no non-trivial characters, then any ergodic action of G on a probability measure space is essentially free, i.e. µ(Fix(g )) = 0 for every g 6= e. Invariant Random Subgroups Definition Every group G acts on the space SUBG of its subgroups by conjugation. The space of subgroups is a closed subset in {0, 1}G . A probability invariant measure µ for (SUBG , G ) is called an invariant random subgroup. Abert, Glasner, and Viraǵ (2012) showed that for any invariant random subgroup λ there is a measure preserving action of G on (X , µ) such that F : X → SUBG given by F (x) = StabG (x) = {g ∈ G : g (x) = x} pushes µ onto λ. Corollary The commutator subgroup F ′ of the Thompson group has no non-trivial invariant random subgroups. Note that G always has measures supported by 1{e} and 1G . Amenability and Invariant Random Subgroups Open Question. Is F ′ amenable? A discrete group G is amenable if it admits an invariant finitely additive probability measure on G . Question. Is there a subgroup Γ < F ′ such that the orbit-closure of Γ, under F ′ , in SUBG does not contain the identity subgroup {e} and the entire group {F ′ }? If it were true, the group F ′ would be non-amenable. Indeed, suppose that F ′ is amenable. Then there exists an invariant random subgroup µ supported by the orbit-closure of Γ. This is impossible as F ′ has no non-trivial invariant random subgroups. Skudlarek (1971). There are locally finite groups with no non-trivial characters. Characters of Thompson’s groups The abelianization of F , F /F ′ is isomorphic to Z × Z. Theorem (Dudko-M.) If χ : F → C is an irreducible character, then either χ is a regular character or χ(g ) = ρ([g ]), where ρ : F /F ′ → T is a group homomorphism, where [g ] is the image of g in the abelianization F /F ′ . Proof. Classification of Characters for F ′ . Simplicity. Note that the group F ′ is simple. If χ is a character such that |χ(g )| = 1 for some g ∈ G , g 6= e, then χ ≡ 1. Using the GNS-construction, construct a unitary representation (π, H, ξ) such that χ(g ) = (π(g )ξ, ξ) with ||ξ|| = 1. Using the Cauchy-Schwarz inequality, we obtain 1 = |χ(g )| = |(π(g )ξ, ξ)| ≤ 1. Thus, π(g )ξ = cξ. Since ξ is a separating, we get that π(g ) = cId. Find h ∈ F ′ such that s = hgh−1 g −1 6= e. Then π(s) = Id. Since the group F ′ is simple, it is generated by elements of the form {qsq −1 }. It follows that π ≡ 1. Proof continued. If χ is irreducible, the representation is a factor, then χ is multiplicative in the sense that for “almost” any g , h ∈ F ′ with disjoint supports, χ(g · h) = χ(g ) · χ(h). The action of the group F ′ on X = (0, 1) is contractible in the sense that for any g ∈ F ′ and any open set U there is a conjugate of g supported by U. So for any open set U and any ε > 0, there is an element h ∈ F ′ supported by U with |χ(h)| < ε. Take |χ(g )| < 1, find conjugates {gi } of g with disjoint supports and supported by U. Then |χ(g1 · · · gn )| < ε. If χ(g ) = µ(Fix(g )). Then supp(gn ) → {x0 }, Fix(gn ) → X , µ(Fix(gn )) → 1, but χ(gn ) → 0, a contradiction.