N properly proximal relative toLΓ, we show thatN either has a properly proximal direct summand or is amenable relative toLΓinsideL(ZoΓ). In this step, the notion of normal bidual developed in [DKEP22, Section 2] is extensively used. Lastly, using a technique from [Ioa15], we conclude in Section 3.4 that if N⊂L(ZoΓ)is amenable relative toLΓ, thenN must be rigid with respect to the s-malleable deformation {αt}associated withL(ZoΓ). Altogether, we obtain that ifN⊂L(ZoΓ)has no amenable or properly proximal direct summand, thenNmust beαt-rigid.
3.2.4.1 Boundary pieces
GivenM a finite von Neumann algebra, anM-boundary pieceXis a hereditary C∗-subalgebra ofB(L2M) such thatM(X)∩M⊂MandM(X)∩JMJ⊂JMJare weakly dense, whereM(X)is the multiplier ofX. To avoid pathological examples, we will always assume thatX6={0}and it follows thatK(L2M)⊂Xfor any M-boundary pieceX.
Let M be a finite von Neumann algebra andXan M-boundary piece. Denote byKLX(M)⊂B(L2M) thek · k∞,2-closure of the norm closed left idealB(L2M)X, wherekTk∞,2=supa∈(M)
1kTakˆ 2 for anyT ∈ B(L2M), and setKX(M) = (KLX(M))∗∩KLX(M)to be the hereditary C∗-subalgebra generated byKLX(M).
The multiplier algebra ofKX(M)contains bothM andJMJand we denote byK∞,1X (M)thek · k∞,1-closure ofKX(M), wherekTk∞,1=supa,b∈(M)
1hTa,ˆ biˆ forT ∈B(L2M), andK∞,1X (M)coincides withXk·k∞,1. And denote bySX(M)the following operator system that containsM,
SX(M) ={T ∈B(L2M)|[T,x]∈K∞,1X (M),for anyx∈JMJ}.
WhenX=K(L2M), we omitXin the above notations for simplicity.
LetN⊂Mbe a von Neuman subalgebra. We sayN⊂Mis properly proximal relative toXif there does not exist anyN-central stateϕ onSX(M)such thatϕ|M is normal. And we sayM is properly proximal if M⊂Mproperly proximal relative toK(L2M). By [DKEP22, Theorem 6.2], a groupΓis properly proximal in the sense of [BIP21] if and only ifLΓis properly proximal.
One particular type of boundary pieces arise from subalgebras. LetN⊂Mbe a von Neumann subalgebra and we may associate withNanM-boundary pieceXN, which is the hereditary C∗-subalgebra ofB(L2M) generated byxJyJeNforx,y∈M, whereeN∈B(L2M)is the orthogonal projection fromL2MontoL2N.
Remark 3.2.2. LetΓ be a group that is not properly proximal, thenLΓ has no properly proximal direct summand. Indeed, supposez∈Z(LΓ)in a nonzero central projection such thatzLΓis not properly proximal, i.e., there exists azLΓ-central stateϕ:S(zLΓ)→Cthat is normal onzLΓ. We may consider theΓ-equivariant embeddingi:S(Γ)→S(LΓ)andE:=Ad(z):S(LΓ)→S(zLΓ)[DKEP22, Section 6]. Thenϕ◦E◦i:S(Γ)→ Cis then aΓ-invariant state, showing Γis not properly proximal. A similar argument shows that if Γis nonamenable, thenLΓhas no amenable direct summand.
Recall that a groupΓis biexact relative to a subgroupΛ<Γif the left action ofΓonSΛ(Γ) ={f ∈`∞Γ| f−Rtf ∈c0(Γ,{Λ})}is topologically amenable, wherec0(Γ,{Λ})is functions onΓthat converge to 0 when t∈Γescapes subsets ofΓthat are small relative toΛ(See [BO08, Chapter 15] for the precise definition).
We remark that this is equivalent toΓy SI(Γ)is amenable. Indeed, since we may embed`∞Γ,→I∗∗ in a
Γ-equivariant way, we haveΓyI∗∗⊕(SI(Γ)/I)∗∗=SI(Γ)∗∗is amenable, and it follows thatΓy SI(Γ)is an amenable action [BEW19, Proposition 2.7].
The following is an easy adaptation of [DKEP22, Theorem 7.1]. For completeness, we include the proof.
Proposition 3.2.3. Suppose a groupΓis biexact relative to a subgroupΛ<Γ. Then for every von Neumann subalgebra N⊂LΓ, either the inclusion N ⊂LΓ is properly proximal relative to XLΛ or else N has an amenable direct summand.
Proof. Suppose the inclusion N⊂LΓ is not properly proximal relative to XLΛ, and let φ :SXLΛ(LΓ)→ hp(LΓ)p,eN pibe ap(LΓ)p-bimodular u.c.p. map, where p∈Z(N)is a non-zero central projection.
If we consider theΓ-equivariant diagonal embedding`∞Γ⊂B(`2Γ), we see thatc0(Γ,{Λ})is mapped toXLΛ. Restricting to SΛ(Γ)then gives a Γ-equivariant embedding into SXLΛ(LΓ). We therefore obtain a ∗-homomorphism SΛ(Γ)of Γ→B(`2Γ) whose image is contained in SXLΛ(LΓ). Composing this ∗- homomorphism with the u.c.p. mapφ then gives a u.c.p. map ˜φ :SΛ(Γ)ofΓ→ hp(LΓ)p,eN pisuch that φ(t) =˜ putpfor allt∈Γ.
SinceΓis biexact relative toΛ, the actionΓy SΛ(Γ)is topologically amenable. Hence,SΛ(Γ)ofΓ= SΛ(Γ)orΓis a nuclear C∗-algebra. We setϕ(·):=τ(p)1 hφ˜(·)p,ˆ piˆ and note that forx∈Cr∗Γwe haveϕ(x) =
1
τ(p)τ(pxp). SinceC∗rΓis weakly dense inLΓan argument similar to Proposition 3.1 in [BC15] then gives a representationπϕ:SΛ(Γ)orΓ→B(Hϕ), a state ˜ϕ ∈B(Hϕ)∗ withϕ=ϕ˜◦πϕ, and a projectionq∈ πϕ(SΛ(Γ)orΓ)00with ˜ϕ(q) =1 such that there is a normal unital∗-homomorphismi:LΓ,→qπϕ(SΛ(Γ)or
Γ)00q.
Since SΛ(Γ)orΓ is nuclear, we have thatπϕ(SΛ(Γ)orΓ)00 is injective, and so there is a u.c.p. map i˜:B(`2Γ)→qπϕ(SΛ(Γ)orΓ)00qthat extendsi. Notice thatψ:=ϕ˜◦i˜is then anN p-central state onB(`2Γ) andψ(x) =τ(p)1 τ(pxp)forx∈LΓ. Therefore,N pis amenable.
3.2.4.2 A bidual characterization
Next we collect some basics of the normal bidual from [DKEP22, Section 2].
Given a finite von Neumann algebra M and a C∗-subalgebraA⊂B(L2M) such thatM andJMJ are contained in the multiplier algebraM(A), we recall thatAM]M(resp.AJMJ]JMJ)denotes the space ofϕ∈A∗ such that for each T ∈A the mapM×M 3(a,b)7→ϕ(aT b)(resp. JMJ×JMJ3(a,b)7→ϕ(aT b)) is separately normal in each variable. When there is no confusion about the von Neumann algebra that we are referring to, we will denoteAM]MbyA]andAM]M∩AJMJ]JMJbyA]J.
We may view (A]J)∗as a von Neumann algebra as follows. Denote by pnor∈M(A)∗∗ the supremum of support projections of states inM(A)∗ that restrict to normal states onM andJMJ, so thatM andJMJ
may be viewed as unital von Neumann subalgebras ofpnorM(A)∗∗pnor, which is canonically identified with (M(A)]J)∗. LetqA∈P(M(A)∗∗)be the central projection such thatqA(M(A)∗∗) =A∗∗ and we may then identify(A]J)∗withpnorqAM(A)∗∗pnor=pnorA∗∗pnor. Furthermore, ifB⊂Ais another C∗-subalgebra with M,JMJ⊂M(B), we may identify(B]J)∗withqBpnorA∗∗pnorqB, which is a non-unital subalgebra of(A]J)∗.
If we denote byι:M(A)→M(A)∗∗ the canonical embedding, we may then viewM(A)as an operator subsystem of(M(A)]J)∗through the isometric u.c.p. mapιnor:M(A)3T →pnorι(T)pnor∈(M(A)]J)∗, and the its restriction toAgives a natural embedding ofA⊂(A]J)∗as an operator system.
It is worth noting that ιnorandι are different in a few ways. On one hand, ι:M(A)→M(A)∗∗ is a
∗-homomorphism whileιnor:M(A)→(M(A)]J)∗is a u.c.p. map; on the other hand,ιnorgives rise to normal faithful representations when restricted toMandJMJ, butι|M andι|JMJare not normal in general.
The following is a bidual characterization of properly proximal.
Lemma 3.2.4. [DKEP22, Lemma 8.5] Let M be a separable tracial von Neumann algebra with an M- boundary pieceX. Then M is properly proximal relative to Xif and only if there is no M-central stateϕ on
eSX(M):=n T∈
B(L2M)]J∗
|[T,a]∈
KX(M)]J∗
for alla∈JMJo
such thatϕ|Mis normal.
WhenX=K(L2M), we will abbreviateXfor simplicity. It is worth noting thateSX(M)is a von Neumann algebra which containsMas a von Neumann subalgebra, whileSX(M)is only an operator system. Following the above discussion, we note that ˜SX(M)may be identified with
eSX(M) ={T ∈pnorB(L2M)∗∗pnor|[T,a]∈qX M(KX(M))∗∗
qX,for anya∈JMJ},
whereqX is the identity of (KX(M)]J)∗⊂(M(KX(M))]J)∗. If we setqK=qK(L2M) to be the identity of (K(L2M)]J)∗⊂(B(L2M)]J)∗, then using the above description of ˜SX(M), we haveq⊥XeSX(M)q⊥
X⊂q⊥KeS(M), as qXcommutes withMandJMJ.
Lemma 3.2.5. Let M be a separable tracial von Neumann algebra. Suppose M has no properly proximal direct summand, then there exists an M-central stateϕoneS(M)such thatϕ|Mis faithful and normal.
Proof. First we show that there exists anM-central stateϕ oneS(M)such thatϕ|Z(M) is faithful. Consider a pair (ϕ,p), where ϕ ∈eS(M)∗ is anM-central state such that ϕ|M is normal and p∈Z(M) is support projection ofϕ. And we may order such pairs by the order onZ(M), i.e.,(ϕ1,p1)≤(ϕ2,p2)ifp1≤p2. If{(ϕi,pi)}i∈I is a chain, then we may find a subsequence pi(n) such that limnpi(n)=∨i∈Ipi, andϕ0(·) =
∑n≥12−nϕi(n)(pi(n)·pi(n))then is anM-central state oneS(M) such that ϕ0|M is normal and ∨i∈Ipi is the support ofϕ0. Suppose(ϕ,p)is a maximal element andq=p⊥>0. Denote byEq: Ad(q):B(L2M)→ B(L2(qM))and one checks that(Eq)∗mapsB(L2(qM))]JtoB(L2M)]J. Therefore dualizingEqyeilds a u.c.p.
map ˜Eq:(B(L2M)]J)∗→(B(L2(qM))]J)∗, and(E˜q)|eS(M):eS(M)→eS(qM). SinceqMis not properly proximal, there exists a stateψ∈eS(qM)∗that isqM-central andψ|qMis normal. Setϕ0(T) =ϕ(pT p) +ψ(E˜q(qT q)), which is anM-central state on eS(M) that is normal on M with support strictly larger than p, which is a contradiction.
Now supposeϕis such a state withϕ|Z(M)faithful, andϕ(p) =0 for somep∈P(M), then we may write the central supportz(p) =∑∞i=1viv∗i, wherevi∈Mare partial isometries such thatv∗ivi≤p. Sinceϕis normal and tracial onM, we haveϕ(z(p)) =∑∞i=1ϕ(viv∗i)≤∑∞i=1ϕ(p) =0, which shows thatp≤z(p) =0.