betaFeedback

WorksMathematics

Solid factors associated to the linear action of SL2(Z)SL_2(\mathbb{Z}) on Zn\mathbb{Z}^n

Abstract

Let n≥2n\geq 2 be an integer. Set Γn=SL2(Z)Γ_n=SL_2(\mathbb{Z}) (respectively PSL2(Z)PSL_2(\mathbb{Z})) for nn even (respectively for nn odd). Then ΓnΓ_n acts on Zn\mathbb{Z}^n through the restriction of the linear action of SL2(R)SL_2(\mathbb{R}) on Rn\mathbb{R}^n. We prove that that the von Neumann algebra L(Zn⋊Γn)L(\mathbb{Z}^n\rtimes Γ_n) is a solid factor. This generalizes the case n=2n=2 proved by N. Ozawa.

Introduction

Let n≥2n \ge2 be an integer and let Pn−1(Z)P_{n-1}(\mathbb{Z}) be the space of homogeneous polynomials of degree n−1n-1 in two variables X,YX,Y with integer coefficients: we have a standard action ρn−1\rho_{n-1} of SL2(Z)\mathrm{SL}_2(\mathbb{Z}) on Pn−1(Z)P_{n-1}(\mathbb{Z}) defined by

(ρn−1((abcd)))(P)(X,Y)=P(aX+cY,bX+dY)\left(\rho_{n-1}\left(\begin{pmatrix} a & b \\ c & d \end{pmatrix}\right)\right)(P)(X,Y)=P(aX+cY,bX+dY)

for P∈Pn−1(Z)P \in P_{n-1}(\mathbb{Z}) and (abcd)∈SL2(Z)\begin{pmatrix} a & b \\ c & d \end{pmatrix} \in\mathrm{SL}_2(\mathbb{Z}). Identifying the abelian group Pn−1(Z)P_{n-1}(\mathbb{Z}) with Zn\mathbb{Z}^n by means of the canonical basis Xn−1,Xn−2Y,…,XYn−2,Yn−1X^{n-1},X^{n-2}Y,\ldots,XY^{n-2},Y^{n-1}, we get an SL2(Z)\mathrm{SL}_2(\mathbb{Z})-action on Pn−1(Z)=ZnP_{n-1}(\mathbb{Z})=\mathbb{Z}^n. If nn is odd, the representation ρn−1\rho_{n-1} factors through PSL2(Z)\mathrm{PSL}_2(\mathbb{Z}). Setting Γn=SL2(Z)\Gamma_n=\mathrm{SL}_2(\mathbb{Z}) for nn even and Γn=PSL2(Z)\Gamma_n=\mathrm{PSL}_2(\mathbb{Z}) for nn odd, the associated group von Neumann algebra L(Zn⋊Γn)L(\mathbb{Z}^n\rtimes\Gamma_n) is a type II1\mathrm{II}_1 factor, as it was proved in [4].

The present note consists in a generalization of the main result of the short note [7] where the author proves that the group von Neumann algebra L(Z2×SL2(Z))L(\mathbb{Z}^2\times\mathrm{SL}_2(\mathbb{Z})) is solid, i.e. the relative commutant of every diffuse von Neumann subalgebra of L(Z2×SL2(Z))L(\mathbb{Z}^2\times\mathrm{SL}_2(\mathbb{Z})) is injective.

Solid von Neumann algebras were introduced and studied first in the seminal article [5] as a sufficient condition which ensures that a solid type II1\mathrm{II}_1 factor is necessarily non-Γ\Gamma and prime unless it is injective. Thus, using ideas and techniques of [5, 6, 7], our main result is the following theorem.

Theorem 1.1. Let n≥2n \ge2 be an integer. Then the type II1\mathrm{II}_1 factor L(Zn⋊Γn)L(\mathbb{Z}^n\rtimes\Gamma_n) is solid.

As will be explained in § 2, the above theorem is a consequence of the following slightly more general result. We refer to § 2 and to [2] for precise definitions.

Theorem 1.2. Let Γ\Gamma and Λ\Lambda be countable groups. Assume that:

  1. Γ\Gamma is a bi-exact group that acts on Λ\Lambda;

  2. Λ\Lambda is an amenable group;

  3. Γ\Gamma acts amenably on a metric space (X,d)(X,d) and there is a Γ\Gamma-equivariant map φ:Λ\{e}→X\varphi:\Lambda\backslash\{e\}\to X such that

lim⁡t→∞d(φ(sts′),φ(t))=0\lim_{t\to\infty}d(\varphi(sts'),\varphi(t))=0

for all s,s′∈Λs,s'\in\Lambda.

Then the von Neumann algebra L(Λ⋊Γ)L(\Lambda\rtimes\Gamma) is solid.

Acknowledgements. I am very grateful to A. Valette for his help in Lemma 2.4.

Proofs of Theorems 1.1 and 1.2

Let GG be a discrete, countable group. If GG acts on some set SS and if ff is any function defined on SS then we denote by g.fg.f the function s↦f(g−1.s),g∈Gs\mapsto f(g^{-1}.s),g\in G.

Recall from [2 Theorem 5.1] that GG is exact if it admits an amenable action on some compact topological space, i.e. if there is a compact space XX on which GG acts by homeomorphisms so that there exists a sequence of continuous maps mi:X→Prob⁡(G)m_i:X\to\operatorname{Prob}(G) such that, for every g∈Gg\in G

lim⁡i→∞(sup⁡x∈X∥g.mix−migx∥1)=0.\lim_{i\to\infty}\left(\sup_{x\in X}\left\|g.m_i^x-m_i^{gx}\right\|_1\right)=0.

Notations (1) Let A,BA,B be C∗C^*-algebras. We denote by A⊙BA\mathbin{\odot}B the algebraic tensor product of AA and BB, and by A⊗BA\otimes B their minimal tensor product.

(2) If HH is a Hilbert space, K(H)\mathbb{K}(H) denotes the ideal of all compact operators in the algebra of all linear, bounded operators B(H)\mathbb{B}(H).

(3) Let GG be a group. We denote by λ\lambda (resp. ρ\rho) its left (resp. right) regular representation on ℓ2(G)\ell^2(G), and by Cλ∗(G)C^*_\lambda(G) and Cρ∗(G)C^*_\rho(G) the C∗C^*-algebras generated by λ(G)\lambda(G) and ρ(G)\rho(G) respectively. We write λG\lambda_G and ρG\rho_G when the group must be precised. Finally, we denote by L(G)L(G) the associated von Neumann algebra of GG, which is the σ\sigma-weak closure of the reduced C∗C^*-algebra Cλ∗(G)C^*_\lambda(G).

Definition 2.1. [2 Definition 15.1.2] The countable group GG is bi-exact (or has property SS) if it is exact and there is a map μ:G→Prob⁡(G)\mu:G\to\operatorname{Prob}(G) such that, for all g,h∈Gg,h\in G,

lim⁡x→∞∥μ(gxh)−g.μ(x)∥1=0.\lim_{x\to\infty}\|\mu(gxh)-g.\mu(x)\|_1=0.

We state now the following criterion of bi-exactness for exact groups. It is taken from [2 Lemma 15.1.4], and the last assertion is [5 Theorem 6].

Theorem 2.2. Let GG be a countable exact group. Then the following conditions are equivalent:

  1. GG is bi-exact;

  2. there exists a unital completely positive (ucp) map

θ:Cλ∗(G)⊗Cρ∗(G)→B(ℓ2(G))\theta:C^*_\lambda(G)\otimes C^*_\rho(G)\to\mathbb{B}(\ell^2(G))

such that θ(a⊗b)−ab∈K(ℓ2(G))\theta(a\otimes b)-ab\in\mathbb{K}(\ell^2(G)) for all a∈Cλ∗(G),b∈Cρ∗(G)a\in C^*_\lambda(G),b\in C^*_\rho(G).

If it is the case, then L(G)L(G) is a solid von Neumann algebra: for every diffuse von Neumann subalgebra N⊂L(G)N \subset L(G), its relative commutant N′∩L(G)N' \cap L(G) is injective.

Condition (2) above was first established for the free group on two generators by Akemann and Ostrand in [1], later by Skandalis [8] for discrete subgroups of connected simple Lie groups of rank one, and by Higson and Guentner for hyperbolic groups in [3]. In particular, SL2(Z)\mathrm{SL}_{2}(\mathbb{Z}) and PSL2(Z)\mathrm{PSL}_{2}(\mathbb{Z}) are bi-exact.

As we will prove, the semi-direct products Zn⋊Γn\mathbb{Z}^{n} \rtimes\Gamma_{n} defined in § 1 are bi-exact because they possess an extra property (introduced in [7] for the case n=2n=2) that implies their bi-exactness.

From now on, let Γ\Gamma be a countable, bi-exact group which acts on a countable, amenable group Λ\Lambda via an action denoted by α\alpha. We assume moreover that the pair (Γ,Λ)(\Gamma,\Lambda) satisfies the following additional condition: there is a compact metric space (X,d)(X,d) on which Γ\Gamma acts amenably and there is a Γ\Gamma-equivariant map φ:Λ\{ϵ}→X\varphi:\Lambda\backslash\{\epsilon\}\to X such that, for all s,s′∈Λs,s'\in\Lambda,

lim⁡t→∞d(φ(sts′),φ(t))=0.\lim_{t\to\infty} d(\varphi(sts'),\varphi(t))=0.

Then we say that (Γ,Λ)(\Gamma,\Lambda) is bi-invariant at infinity relative to XX.

The following lemma is taken from the proof of Theorem 3 in [7]; we reproduce its proof for the reader’s convenience.

Lemma 2.3. Let Γ,Λ\Gamma,\Lambda and XX be as above, and assume that (Γ,Λ)(\Gamma,\Lambda) is bi-invariant at infinity relative to XX. Then there exists a map ζ:Λ→Prob⁡(Γ)\zeta:\Lambda\to\operatorname{Prob}(\Gamma) such that, for all g∈Γ,s,s′∈Λg\in\Gamma,s,s'\in\Lambda,

lim⁡y→∞(∥g⋅ζ(y)−ζ(αg(y))∥1+∥ζ(sys′)−ζ(y)∥1)=0.\lim_{y\to\infty}\left(\lVert g\cdot\zeta(y)-\zeta(\alpha_g(y))\rVert_{1}+\lVert\zeta(sys')-\zeta(y)\rVert_{1}\right)=0.

Proof. Since the action Γ↷X\Gamma\curvearrowright X is amenable, choose a sequence of continuous maps μn:X→Prob⁡(Γ)\mu_n:X\to\operatorname{Prob}(\Gamma) for n≥1n\geq1 such that

lim⁡n→∞(sup⁡x∈X∥g⋅μnx−μng⋅x∥1)=0(g∈Γ).\lim_{n\to\infty}\left(\sup_{x\in X}\lVert g\cdot\mu_n^x-\mu_n^{g\cdot x}\rVert_{1}\right)=0\qquad(g\in\Gamma).

Then define the map ζn:Λ\{ϵ}→Prob⁡(Γ)\zeta_n:\Lambda\backslash\{\epsilon\}\to\operatorname{Prob}(\Gamma) by ζn(s)=μnφ(s)\zeta_n(s)=\mu_n^{\varphi(s)}, s∈Λ\{ϵ}s\in\Lambda\backslash\{\epsilon\}.

Taking a subsequence if necessary, we assume furthermore that

lim⁡t→∞(∥ζn(αg(t))−g⋅ζn(t)∥1+∥ζn(sts′)−ζn(t)∥1)<1n\lim_{t\to\infty}\left(\lVert\zeta_n(\alpha_g(t))-g\cdot\zeta_n(t)\rVert_{1}+\lVert\zeta_n(sts')-\zeta_n(t)\rVert_{1}\right)<\frac{1}{n}

for all n,g∈Γ,s,s′∈Λn,g\in\Gamma,s,s'\in\Lambda. Indeed, for g∈Γ,s,s′∈Λg\in\Gamma,s,s'\in\Lambda one has:

∥ζn(αg(t))−g⋅ζn(t)∥1+∥ζn(sts′)−ζn(t)∥1=∥μng⋅φ(t)−g⋅μnφ(t)∥1+∥μnφ(sts′)−μnφ(t)∥1→0.\lVert\zeta_n(\alpha_g(t))-g\cdot\zeta_n(t)\rVert_{1}+\lVert\zeta_n(sts')-\zeta_n(t)\rVert_{1} = \lVert\mu_n^{g\cdot\varphi(t)}-g\cdot\mu_n^{\varphi(t)}\rVert_{1}+\lVert\mu_n^{\varphi(sts')}-\mu_n^{\varphi(t)}\rVert_{1}\to0.

For n≥1,g∈Γ,s,s′∈Λn\geq1,g\in\Gamma,s,s'\in\Lambda, let us define the finite set

Dn(g;s,s′):={t∈Λ\{ϵ}:∥ζn(αg(t))−g⋅ζn(t)∥1+∥ζn(sts′)−ζn(t)∥1≥1/n}.D_n(g;s,s'):=\{t\in\Lambda\backslash\{\epsilon\}:\lVert\zeta_n(\alpha_g(t))-g\cdot\zeta_n(t)\rVert_{1}+\lVert\zeta_n(sts')-\zeta_n(t)\rVert_{1}\geq1/n\}.

Choose sequences of finite, symmetric sets {ϵ}=E0⊂E1⊂E2⊂⋯⊂Γ\{\epsilon\}=E_0\subset E_1\subset E_2\subset\cdots\subset\Gamma, ⋃nEn=Γ\bigcup_n E_n=\Gamma, and {ϵ}=F0⊂F1⊂F2⊂⋯⊂Λ\{\epsilon\}=F_0\subset F_1\subset F_2\subset\cdots\subset\Lambda, ⋃nFn=Λ\bigcup_n F_n=\Lambda. Define inductively finite sets {ϵ}=Ω0⊂Ω1⊂Ω2⊂⋯⊂Λ\{\epsilon\}=\Omega_0\subset\Omega_1\subset\Omega_2\subset\cdots\subset\Lambda by

Ωn=⋃g∈En,s,s′∈Fnt∈Ωn−1(Dn(g;s,s′)∪{αg(t),sts′})\Omega_n=\bigcup_{\substack{g\in E_n,s,s'\in F_n\\ t\in\Omega_{n-1}}}\left(D_n(g;s,s')\cup\{\alpha_g(t),st s'\}\right)

for each n≥1n \ge1. For y∈Λy \in\Lambda, set l(y)=min⁡{n:y∈Ωn}l(y)=\min\{n:y\in\Omega_n\}. Then finally, define ζ:Λ→Prob⁡(Γ)\zeta:\Lambda\to\operatorname{Prob}(\Gamma) by

ζ(y)=1l(y)∑n=0l(y)−1ζn(y)\zeta(y)=\frac{1}{l(y)}\sum_{n=0}^{l(y)-1}\zeta_n(y)

if y≠ey\ne e, and by ζ(y)=δe\zeta(y)=\delta_e if y=ey=e.

Fix g∈Γg\in\Gamma, s,s′∈Λs,s'\in\Lambda. Let kk be large enough so that g∈Ekg\in E_k, s,s′∈Fks,s'\in F_k. If y∈Λy\in\Lambda is such that l(y)>kl(y)>k, then y∈Ωl(y)y\in\Omega_{l(y)}, but y∉Ωny\notin\Omega_n for n≤kn\le k. One has αg(y),sys′∈Ωl(y)+1\alpha_g(y),sys'\in\Omega_{l(y)+1} hence ∣l(αg(y))−l(y)∣,∣l(sys′)−l(y)∣≤1|l(\alpha_g(y))-l(y)|,|l(sys')-l(y)|\le1 and thus

∥ζn(αg(y))−g⋅ζn(y)∥1+∥ζn(sys′)−ζn(y)∥1<1/n\|\zeta_n(\alpha_g(y))-g\cdot\zeta_n(y)\|_1+\|\zeta_n(sys')-\zeta_n(y)\|_1<1/n

for k≤n<l(y)k\le n<l(y). This proves that

lim⁡l(y)→∞(∥g⋅ζ(y)−ζ(αg(y))∥1+∥ζ(sys′)−ζ(y)∥1)=0\lim_{l(y)\to\infty}\bigl(\|g\cdot\zeta(y)-\zeta(\alpha_g(y))\|_1+\|\zeta(sys')-\zeta(y)\|_1\bigr)=0

for all g∈Γg\in\Gamma, s,s′∈Λs,s'\in\Lambda.

The following lemma was suggested by A. Valette.

Lemma 2.4. For n≥2n\ge2, let Pn−1(R)\mathbb{P}^{n-1}(\mathbb{R}) be the real projective space of dimension n−1n-1 equipped with the metric

d([v],[w])=arccos⁡(∣⟨v∣w⟩∣∥v∥∥w∥)(v,w∈Rn∖{0}).d([v],[w])=\arccos\left(\frac{|\langle v|w\rangle|}{\|v\|\|w\|}\right)\quad(v,w\in\mathbb{R}^n\setminus\{0\}).

(For v∈Rn∖{0}v\in\mathbb{R}^n\setminus\{0\}, [v][v] denotes the image of vv by the canonical projection from Rn∖{0}\mathbb{R}^n\setminus\{0\} onto Pn−1(R)\mathbb{P}^{n-1}(\mathbb{R}).) Then the map φ:Zn∖{0}→Pn−1(R)\varphi:\mathbb{Z}^n\setminus\{0\}\to\mathbb{P}^{n-1}(\mathbb{R}) defined by φ(v)=[v]\varphi(v)=[v] is Zn\mathbb{Z}^n-equivariant, the action of Γn\Gamma_n on Pn−1(R)\mathbb{P}^{n-1}(\mathbb{R}) is amenable and one has for every a∈Zna\in\mathbb{Z}^n

lim⁡∥v∥→∞d(φ(v+a),φ(v))=0.\lim_{\|v\|\to\infty}d(\varphi(v+a),\varphi(v))=0.

In particular, the pair (Γn,Zn)(\Gamma_n,\mathbb{Z}^n) is bi-invariant at infinity relative to Pn−1(R)\mathbb{P}^{n-1}(\mathbb{R}).

Proof. It is well-known that the linear action of SL⁡n(R)\operatorname{SL}_n(\mathbb{R}) on Rn\mathbb{R}^n induces an action on Pn−1(R)\mathbb{P}^{n-1}(\mathbb{R}), and it was observed in the proof of Theorem 1.1 in [4] that ρn−1(SL⁡2(Z))\rho_{n-1}(\operatorname{SL}_2(\mathbb{Z})) is a subgroup of SL⁡n(Z)\operatorname{SL}_n(\mathbb{Z}). Thus, the restriction φ\varphi to Zn∖{0}\mathbb{Z}^n\setminus\{0\} of the canonical map v↦[v]v\mapsto[v] to Pn−1(R)\mathbb{P}^{n-1}(\mathbb{R}) is Γn\Gamma_n-equivariant. By [4 Proposition 3.5], the action of Γn\Gamma_n on Pn−1(R)\mathbb{P}^{n-1}(\mathbb{R}) is amenable, and one obviously has

lim⁡∥v∥→∞d([v+a],[v])=0\lim_{\|v\|\to\infty}d([v+a],[v])=0

for every a∈Zna\in\mathbb{Z}^n by definition of dd. This proves that the pair (Γn,Zn)(\Gamma_n,\mathbb{Z}^n) is bi-invariant at infinity.

The previous lemma implies that Theorem 1.1 is a special case of Theorem 1.2. Thus we are left to prove the latter.

Proof of Theorem 1.2. We come back to the general situation where Γ\Gamma is bi-exact, that it acts on the amenable group Λ\Lambda and that there is a map ζ:Λ→Prob⁡(Γ)\zeta:\Lambda\to\operatorname{Prob}(\Gamma) such that, for all g∈Γg\in\Gamma, s,s′∈Λs,s'\in\Lambda,

lim⁡y→∞(∥g⋅ζ(y)−ζ(αg(y))∥1+∥ζ(sys′)−ζ(y)∥1)=0.\lim_{y\to\infty}\bigl(\|g\cdot\zeta(y)-\zeta(\alpha_g(y))\|_1+\|\zeta(sys')-\zeta(y)\|_1\bigr)=0.

We observe that, as exactness of groups is preserved under semi-direct products, the group G:=Λ⋊ΓG:=\Lambda\rtimes\Gamma is exact.

In order to prove Theorem 1.2, by Theorem 2.2, it suffices to show the existence of a ucp map θ ⁣:Cλ∗(G)⊗Cρ∗(G)→B(ℓ2(G))\theta\colon C_{\lambda}^{*}(G)\otimes C_{\rho}^{*}(G)\to\mathbb{B}(\ell^{2}(G)) such that θ(a⊗b)=ab∈K(ℓ2(G))\theta(a\otimes b)=ab\in\mathbb{K}(\ell^{2}(G)) for all a∈Cλ∗(G),b∈Cρ∗(G)a\in C_{\lambda}^{*}(G), b\in C_{\rho}^{*}(G). We adapt Section 4 of [6] to our situation.

Set ξ=ζ1/2\xi=\zeta^{1/2} and define V ⁣:ℓ2(Λ)→ℓ2(Λ)⊗ℓ2(Γ)=ℓ2(G)V\colon\ell^{2}(\Lambda)\to\ell^{2}(\Lambda)\otimes\ell^{2}(\Gamma)=\ell^{2}(G) by Vδy=δy⊗ξ(y)V\delta_{y}=\delta_{y}\otimes\xi(y) for every y∈Λy\in\Lambda, where (δy)y∈Λ(\delta_{y})_{y\in\Lambda} denotes the natural orthonormal basis of ℓ2(Λ)\ell^{2}(\Lambda). It is easy to see that VV is an isometry.

Let also u ⁣:Γ→U(ℓ2(Λ))u\colon\Gamma\to U(\ell^{2}(\Lambda)) be the unitary representation associated to the action Γ↷αΛ\Gamma\overset{\alpha}{\curvearrowright}\Lambda:

(u(g)η)(s)=η(αg−1(s))(g∈Γ,η∈ℓ2(Λ),s∈Λ).(u(g)\eta)(s)=\eta(\alpha_{g^{-1}}(s))\qquad(g\in\Gamma,\eta\in\ell^{2}(\Lambda),s\in\Lambda).

It is straightforward to check that, for (s,g),(t,h)∈G(s,g),(t,h)\in G, one has

((λΛ(s)⊗1)(u⊗λΓ)(g))((λΛ(t)⊗1)(u⊗λΓ)(h))=(λΛ(sαg(t))⊗1)(u⊗λΓ)(gh)\bigl((\lambda_{\Lambda}(s)\otimes1)(u\otimes\lambda_{\Gamma})(g)\bigr)\bigl((\lambda_{\Lambda}(t)\otimes1)(u\otimes\lambda_{\Gamma})(h)\bigr)=\bigl(\lambda_{\Lambda}(s\alpha_{g}(t))\otimes1\bigr)(u\otimes\lambda_{\Gamma})(gh)

and that

⟨(λΛ(s)⊗1)(u⊗λΓ)(g)δe⊗δe∣δe⊗δe⟩=τ(λG(s,g))\langle(\lambda_{\Lambda}(s)\otimes1)(u\otimes\lambda_{\Gamma})(g)\delta_{e}\otimes\delta_{e}\mid\delta_{e}\otimes\delta_{e}\rangle=\tau(\lambda_{G}(s,g))

hence that the map ϕ ⁣:L(Λ⋊Γ)→M:=((λ(Λ)⊗1)∪(u⊗λ)(Γ))′′\phi\colon L(\Lambda\rtimes\Gamma)\to M:=\left((\lambda(\Lambda)\otimes1)\cup(u\otimes\lambda)(\Gamma)\right)^{\prime\prime} defined by ϕ(λG(s,g))=(λΛ(s)⊗1)(u⊗λΓ)(g)\phi(\lambda_{G}(s,g))=(\lambda_{\Lambda}(s)\otimes1)(u\otimes\lambda_{\Gamma})(g) extends to a spatial ∗*-isomorphism of L(Λ⋊Γ)L(\Lambda\rtimes\Gamma) onto MM.

It identifies the reduced C∗C^{*}-algebra Cλ∗(G)C_{\lambda}^{*}(G) with the norm-closure of

B:=C∗((λ(Λ)⊗1)∪(u⊗λ)(Γ)).B:=C^{*}\bigl((\lambda(\Lambda)\otimes1)\cup(u\otimes\lambda)(\Gamma)\bigr).

Moreover, δe⊗δe\delta_{e}\otimes\delta_{e} is cyclic and separating for MM, and, denoting by JJ the involution on ℓ2(Λ)⊗ℓ2(Γ)\ell^{2}(\Lambda)\otimes\ell^{2}(\Gamma) defined by J(x(δe⊗δe))=x∗(δe⊗δe)J(x(\delta_{e}\otimes\delta_{e}))=x^{*}(\delta_{e}\otimes\delta_{e}), x∈Mx\in M, it is easy to see that

J(u⊗λ)(g)J=(u⊗ρ)(g)andJλ(s)⊗1J=ρ(s)⊗1(g∈Γ,s∈Λ).J(u\otimes\lambda)(g)J=(u\otimes\rho)(g)\quad\text{and}\quad J\lambda(s)\otimes1J=\rho(s)\otimes1\quad(g\in\Gamma,s\in\Lambda).

Thus B≅Cλ∗(G)B\cong C_{\lambda}^{*}(G) and C:=JBJ≅Cρ∗(G)C:=JBJ\cong C_{\rho}^{*}(G) satisfy the following properties: BB is locally reflexive by [2 Theorem 9.3.1] since GG is an exact group, and B⊂MB\subset M and C⊂M′C\subset M' are σ\sigma-weakly dense ∗*-subalgebras of MM and of its commutant M′M' respectively.

Let θ ⁣:B(ℓ2(Λ)⊗ℓ2(Γ))→B(ℓ2(Λ))\theta\colon\mathbb{B}(\ell^{2}(\Lambda)\otimes\ell^{2}(\Gamma))\to\mathbb{B}(\ell^{2}(\Lambda)) be defined by

θ(T)=V∗TV(T∈B(ℓ2(Λ)⊗ℓ2(Γ))).\theta(T)=V^{*}TV\quad\left(T\in\mathbb{B}(\ell^{2}(\Lambda)\otimes\ell^{2}(\Gamma))\right).

Lemma 2.5. With hypotheses and notation above, the following equalities hold:

(a) θ((u⊗λ)(g))δy=⟨g.ξ(y)∣ξ(αg(y))⟩δαg(y)(g∈Γ,y∈Λ)\theta((u\otimes\lambda)(g))\delta_{y}=\langle g.\xi(y)\mid\xi(\alpha_{g}(y))\rangle\delta_{\alpha_{g}(y)}\quad(g\in\Gamma,y\in\Lambda);

(b) θ(λ(s)⊗1)δy=⟨ξ(y)∣ξ(sy)⟩δsy(s,y∈Λ)\theta(\lambda(s)\otimes1)\delta_{y}=\langle\xi(y)\mid\xi(sy)\rangle\delta_{sy}\quad(s,y\in\Lambda);

(c) θ(ρ(s)⊗1)δy=⟨ξ(y)∣ξ(ys−1)⟩δys−1(s,y∈Λ)\theta(\rho(s)\otimes1)\delta_{y}=\langle\xi(y)\mid\xi(ys^{-1})\rangle\delta_{ys^{-1}}\quad(s,y\in\Lambda).

Moreover,

θ((u⊗λ)(g))−u(g),θ(λ(s)⊗1)−λ(s),θ(ρ(s)⊗1)−ρ(s)∈K(ℓ2(Λ))\theta((u\otimes\lambda)(g))-u(g),\quad\theta(\lambda(s)\otimes1)-\lambda(s),\quad\theta(\rho(s)\otimes1)-\rho(s)\in\mathbb{K}(\ell^{2}(\Lambda))

for all g∈Γ,s∈Λg\in\Gamma,s\in\Lambda.

Proof. (a) For g∈Γ,y,t∈Λg\in\Gamma,y,t\in\Lambda, one has

⟨θ((u⊗λ)(g))δy∣δt⟩=⟨V∗(u⊗λ)(g)VVδy∣δt⟩=⟨δαg(y)⊗g.ξ(y)∣δt⊗ξ(t)⟩=⟨δαg(y)∣δt⟩⟨g.ξ(y)∣ξ(t)⟩=⟨⟨g.ξ(y)∣ξ(αg(y))⟩δαg(y)∣δt⟩.\begin{aligned} \langle\theta((u\otimes\lambda)(g))\delta_{y}\mid\delta_{t}\rangle=\langle V^{*}(u\otimes\lambda)(g)VV\delta_{y}\mid\delta_{t}\rangle=\langle\delta_{\alpha_{g}(y)}\otimes g.\xi(y)\mid\delta_{t}\otimes\xi(t)\rangle\\ &=\langle\delta_{\alpha_{g}(y)}\mid\delta_{t}\rangle\langle g.\xi(y)\mid\xi(t)\rangle=\langle\langle g.\xi(y)\mid\xi(\alpha_{g}(y))\rangle\delta_{\alpha_{g}(y)}\mid\delta_{t}\rangle. \end{aligned}

Verifications of equalities (b) and (c) are similar and left to the reader. Finally, for g∈Γg\in\Gamma and y∈Λy\in\Lambda, one gets from (a)

(θ((u⊗λ)(g))−u(g))δy=u(g)(⟨g.ξ(y)∣ξ(αg(y))⟩−1)δy\bigl(\theta((u\otimes\lambda)(g))-u(g)\bigr)\delta_{y}=u(g)\bigl(\langle g.\xi(y)\mid\xi(\alpha_{g}(y))\rangle-1\bigr)\delta_{y}

and the function y↦⟨g.ξ(y)∣ξ(αg(y))⟩−1y \mapsto\langle g.\xi(y)|\xi(\alpha_g(y))\rangle-1 belongs to c0(Λ)c_0(\Lambda) because

∣⟨g.ξ(y)∣ξ(αg(y))⟩−1∣=∣⟨g.ξ(y)∣ξ(αg(y))−g.ξ(y)⟩∣≤∥g.ξ(y)∥2∥ξ(αg(y))−g.ξ(y)∥2≤∥ξ(αg(y))−g.ξ(y)∥11/2→0\begin{aligned} \left|\langle g.\xi(y)|\xi(\alpha_g(y))\rangle-1\right|=\left|\langle g.\xi(y)|\xi(\alpha_g(y))-g.\xi(y)\rangle\right| \\ &\leq\|g.\xi(y)\|_2\|\xi(\alpha_g(y))-g.\xi(y)\|_2 \\ &\leq\|\xi(\alpha_g(y))-g.\xi(y)\|_1^{1/2}\to0 \end{aligned}

as y→∞y\to\infty. Thus θ((u⊗λ)(g))−u(g)∈K(ℓ2(Λ))\theta((u\otimes\lambda)(g))-u(g)\in\mathcal{K}(\ell^2(\Lambda)). The last two claims are proved similarly.

Following Section 4 of [6], let K1K_1 be the hereditary C∗^*-subalgebra of B(ℓ2(Λ)⊗ℓ2(Γ))\mathbb{B}(\ell^2(\Lambda)\otimes\ell^2(\Gamma)) generated by K(ℓ2(Λ))⊗ℓ∞(Γ)\mathbb{K}(\ell^2(\Lambda))\otimes\ell^\infty(\Gamma). Then BB and CC are in the multiplier of K1K_1 and hence K:=K1∩C∗(B,C)K:=K_1\cap C^*(B,C) is an ideal in C∗(B,C)C^*(B,C).

Then Lemma 2.5 shows that [6 Proposition 4.3] applies with A=Cλ∗(Λ)A=C^*_\lambda(\Lambda) so that the homomorphism μ:B⊙C→C∗(B,C)/K\mu:B\mathbin{\odot}C\to C^*(B,C)/K defined by

μ(∑k=1mak⊗xk)=∑k=1makxk+K\mu\left(\sum_{k=1}^{m}a_k\otimes x_k\right)=\sum_{k=1}^{m}a_kx_k+K

is continuous with respect to the min-norm.

Next, set I1=B(ℓ2(Λ))⊗K(ℓ2(Γ))I_1=\mathbb{B}(\ell^2(\Lambda))\otimes\mathbb{K}(\ell^2(\Gamma)) and I:=I1∩C∗(B,C)I:=I_1\cap C^*(B,C). Then, as in the proof of [6 Corollary 4.5], the homomorphism Φ:B⊙C→C∗(B,C)/(I∩K)\Phi:B\mathbin{\odot}C\to C^*(B,C)/(I\cap K) defined by

Φ(∑k=1mak⊗xk)=∑k=1makxk+(I∩K)\Phi\left(\sum_{k=1}^{m}a_k\otimes x_k\right)=\sum_{k=1}^{m}a_kx_k+(I\cap K)

is continuous with respect to the min-norm and has a ucp lifting since BB is locally reflexive. As I1∩K1=K(ℓ2(Λ)⊗ℓ2(Γ))I_1\cap K_1=\mathbb{K}(\ell^2(\Lambda)\otimes\ell^2(\Gamma)), the group G=Λ⋊ΓG=\Lambda\rtimes\Gamma is bi-exact by Theorem 2.2, and L(G)L(G) is a solid von Neumann algebra.

This ends the proof of Theorem 1.2.

Université de Neuchâtel, Institut de Mathématiques, E.-Argand 11, 2000 Neuchâtel, Switzerland

Email address: [email protected]

References

  1. [1]C.A. Akemann and P.A. Ostrand. On a tensor product C*-algebra associated with the free group on two generators. J. Math. Soc. Japan, 27:589–599, 1975.DOI
  2. [2]N. Brown and N. Ozawa. C*-algebras and finite-dimensional approximations. Grad. Stud. Math. 88, Amer. Math. Soc., Providence, RI, 2008.DOI
  3. [3]N. Higson and E. Guentner. Group C*-algebras and K-theory, in Noncommutative Geometry, volume 1831, pages 137–251. Lecture Notes in Math., Springer, Berlin, 2004.DOI
  4. [4]P. Jolissaint and V. Valette. The linear SL2(Z)-action on Tⁿ: ergodic and von Neumann algebraic aspects. J. Operator Theory, 53:3–20, 2005.
  5. [5]N. Ozawa. Solid von Neumann algebras. Acta Math., 192:111–117, 2004.DOI
  6. [6]N. Ozawa. A Kurosh type theorem for II₁ factors. Int. Math. Res. Not., pages 1–21, 2006. doi.org/10.1155/IMRN/2006/97560.
  7. [7]N. Ozawa. An example of a solid von Neumann algebra. Hokkaido Math. Journal, 38:557–561, 2009.arxiv.org/abs/0804.0288
  8. [8]G. Skandalis. Une notion de nucléarité en K-théorie (d’après J. Cuntz). K-Theory, 1:549–573, 1988.DOI

Paper details

vlnc:W_e6rbugy9
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 1

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents