By OpenAISourceGitHub

Manuscripts from openai/math

Fork
Clone
git clone https://openversions.com/git/index/openai-math.git
Cite
Version link
https://openversions.com/index/openai-math/commit/dfd8c46870b48511e697cbb43a3c8c91d375fa2f
Version identifier (commit hash)
dfd8c46870b48511e697cbb43a3c8c91d375fa2f

Fork to create your own copy with credit to the source. Clone to get a Git copy on your computer.

All papers

Claims

Findings

An explicit obstruction to nuclear norm-ultrapower embeddings

Let G=Z[1/2]3⋊(SL3(Z)×Z)G = \mathbb{Z}[1/2]^3 \rtimes (\mathrm{SL}_3(\mathbb{Z}) \times \mathbb{Z}) with (M,k)⋅v=2kMv(M,k) \cdot v = 2^k M v, and let A=C∗(G)A = C^*(G) be its full group C∗C^*-algebra. Then AA is separable and unital, and for every nonzero unital nuclear C∗C^*-algebra BB and every free ultrafilter ω\omega on N\mathbb{N}, there is no unital embedding A→BωA \to B^\omega. In particular, AA does not embed unitally into O2ω\mathcal{O}_2^\omega, giving a negative answer to Kirchberg's norm-ultrapower embedding problem.

The group Γ=Z3⋊SL3(Z)\Gamma = \mathbb{Z}^3 \rtimes \mathrm{SL}_3(\mathbb{Z}) is finitely generated and has property (T). The endomorphism θ(v,M)=(2v,M)\theta(v, M) = (2v, M) is injective with [Γ:θ(Γ)]=8[\Gamma : \theta(\Gamma)] = 8 and tγt−1=θ(γ)t\gamma t^{-1} = \theta(\gamma) for γ∈Γ\gamma \in \Gamma. Moreover, there is a nonzero finite-dimensional irreducible unitary representation π∗\pi_* of Γ\Gamma factoring through reduction modulo 2 that is nontrivial on Z3\mathbb{Z}^3, so π∗≄π∗∘θj\pi_* \not\simeq \pi_* \circ \theta^j for all j≥1j \ge 1.

For Γ,S,κ\Gamma, S, \kappa satisfying the Kazhdan spectral gap (2) and any unitary representation σ:Γ→U(K)\sigma : \Gamma \to \mathcal{U}(K) with no nonzero finite-dimensional subrepresentation, L((Us⊗σ(s))s∈S)≥b1L((U_s \otimes \sigma(s))_{s \in S}) \ge b1 with b=κ2/(4∣S∣)b = \kappa^2 / (4|S|) for any Hilbert space EE and any unitaries (Us)s∈S(U_s)_{s \in S} on EE without any group relations among the UsU_s.

For every nonzero unital nuclear C∗C^*-algebra BB and every unitary SS-tuple u\mathbf{u} in BB (with no group relations assumed), the set D(u)={[π]:d(u,π)<a}D(\mathbf{u}) = \{[\pi] : d(\mathbf{u}, \pi) < a\} of unitary equivalence classes of nonzero finite-dimensional irreducible unitary representations π\pi of Γ\Gamma is finite, where 0<a<min⁡{κ,b}0 < a < \min\{\kappa, b\}.

For every sequence (πn)(\pi_n) of nonzero finite-dimensional unitary representations of Γ\Gamma, with no bound on their dimensions, lim⁡ωd(xn,πn)∈{0}∪[κ,R]\lim_\omega d(\mathbf{x}_n, \pi_n) \in \{0\} \cup [\kappa, R]. In particular, if {n:d(xn,πn)<a}∈ω\{n : d(\mathbf{x}_n, \pi_n) < a\} \in \omega, then lim⁡ωd(xn,πn)=0\lim_\omega d(\mathbf{x}_n, \pi_n) = 0.

Under a unital embedding Φ:C∗(G)→Bω\Phi : C^*(G) \to B^\omega, we have lim⁡ωd(xn,π∗)=0\lim_\omega d(\mathbf{x}_n, \pi_*) = 0, and consequently {n:[π∗]∈Dn}∈ω\{n : [\pi_*] \in D_n\} \in \omega.

Under a unital embedding Φ:C∗(G)→Bω\Phi : C^*(G) \to B^\omega, the set {n:(∀[π]∈Dn)(∃[π′]∈Dn) π≺π′∘θ}\{n : (\forall [\pi] \in D_n)(\exists [\pi'] \in D_n)\, \pi \prec \pi' \circ \theta\} belongs to ω\omega.

Let θ:Γ→Γ\theta : \Gamma \to \Gamma be a homomorphism and DD a finite set of equivalence classes of nonzero finite-dimensional irreducible unitary representations of Γ\Gamma such that every [π]∈D[\pi] \in D has [π′]∈D[\pi'] \in D with π≺π′∘θ\pi \prec \pi' \circ \theta. Then for every [π]∈D[\pi] \in D there is an integer p≥1p \ge 1 such that π≃π∘θp\pi \simeq \pi \circ \theta^p.

Neither O2\mathcal{O}_2 nor any exact or nuclear unital C∗C^*-algebra is existentially closed among unital C∗C^*-algebra extensions; for every free ultrafilter ω\omega on N\mathbb{N}, there are a separable unital C∗C^*-algebra DD and α∈Aut(D)\alpha \in \mathrm{Aut}(D) such that DD embeds unitally into O2ω\mathcal{O}_2^\omega but D⋊αZD \rtimes_\alpha \mathbb{Z} does not; and some satisfiable finite condition p(x)p(x) in the unital C∗C^*-algebra language lacks good nuclear witnesses.

Methods

Spectral test for approximate representation matching d(z,π)=min⁡spec L((zs⊗πˉ(s))s∈S)d(\mathbf{z}, \pi) = \min\mathrm{spec}\, L((z_s \otimes \bar{\pi}(s))_{s \in S}) defined for arbitrary unitary SS-tuples z\mathbf{z} without requiring group relations.

Source excerpts

“Define d(z, π) = min spec L((zs ⊗ π¯(s))s∈S), (4) where the spectrum is taken in C ⊗ B(V¯π).” · Show in paper

Wassermann's property (T) quotient method combined with reduced density operators and the Powers–Størmer square-root trace inequality to prove finiteness of approximate finite-dimensional representation matches in a nuclear C∗C^*-algebra.

Source excerpts

“Fix a unit vector ξ ∈ E ⊗ K. Its reduced density operator T on K is characterized by Tr(TC) = ⟨(1 ⊗ C)ξ, ξ⟩ (C ∈ B(K)).” · Show in paper

“The following quotient argument uses Wassermann’s property (T) method [16]; see the exposition in [10, Section 2.1].” · Show in paper

Varying-dimension norm ultraproduct Cω=∏n(B⊗B(Vπn‾))/{(cn):lim⁡ω∥cn∥=0}C_\omega = \prod_n (B \otimes \mathcal{B}(\overline{V_{\pi_n}})) / \{(c_n) : \lim_\omega \|c_n\| = 0\} paired with finite-index induction and Frobenius reciprocity to force a predecessor chain among finite sets DnD_n.

Source excerpts

“Let Cn = B ⊗ B(V¯πn), and form the norm ultraproduct Cω = Qn Cn / {(cn) : limω ∥cn∥ = 0}.” · Show in paper

“We transfer these vectors from Λ to Γ by a fixed eight-coordinate induction.” · Show in paper