betaFeedback

WorksMathematics

Cartesian products of groups as quantum metric spaces

Abstract

We provide a new class of finitely generated groups whose word-length functions give rise to compact quantum metric spaces. Kaad's recent theory of matrix compact quantum metric spaces is instrumental in obtaining this new class, and along the way we also show that any length function on an amenable group which gives rise to a compact quantum metric space in fact gives rise to a matrix compact quantum metric space, and that the corresponding result is false for the free group on two generators.

Introduction

Over the past decade, substantial progress has been made in developing Rieffel’s theory of compact quantum metric spaces, which provides a far-reaching non-commutative generalisation of classical compact metric spaces, with strong ties to Connes’ non-commutative geometry and theoretical physics. Recent advances in the field [1, 13, 19, 6], suggest that basing the theory on operator systems, rather than order unit spaces or full C∗C^*-algebras, provides a good blend of flexibility and rigidity, and with this in mind it is also very natural to seek a matricial version of the theory. A first approach to a matricial theory was developed by Kaad in [8], and the resulting theory of matrix compact quantum metric spaces has, among other things, the advantage that it admits a natural tensor product construction, compatible with the external product of spectral triples. The present paper continues the investigation of matrix compact quantum metric spaces, with particular emphasis on the examples arising from groups. These examples fit into the more general connection between compact quantum metric spaces and non-commutative (differential) geometry, which we now describe in more detail. Firstly, a compact quantum metric space consists of an operator system X\mathcal{X} together with a seminorm L:X→[0,∞)L:\mathcal{X}\to[0,\infty), such that the associated Monge-Kantorovič metric

mk⁡L(φ,ψ):=sup⁡{∣φ(x)−ψ(x)∣∣x∈X,L(x)≤1},φ,ψ∈S(X),\operatorname{mk}_{L}(\varphi,\psi):=\sup\{|\varphi(x)-\psi(x)|\mid x\in\mathcal{X},L(x)\leq1\},\qquad\varphi,\psi\in S(\mathcal{X}),

induces the weak* topology on the state space S(X)S(\mathcal{X}). The seminorm LL may be viewed as a generalisation of the example arising from a classical compact metric space (M,d)(M,d), where X\mathcal{X} is the algebra of Lipschitz continuous functions and LL the seminorm that assigns the (smallest) Lipschitz constant to the function in question. In the same vein, a matrix compact quantum metric space consists of an operator system X\mathcal{X} together with a family of seminorms Ln:Mn(X)→[0,∞)L_n:M_n(\mathcal{X})\to[0,\infty) satisfying a natural set of axioms; see § 2 for details.

On the other hand, the fundamental object in Connes’ non-commutative geometry [4] is the notion of a spectral triple, and from a spectral triple (A,H,D)(\mathcal{A},H,D) one obtains a natural seminorm LD:A→[0,∞)L_D:\mathcal{A}\to[0,\infty) given by LD(a):=∥[D,a]∥L_D(a):=\lVert[D,a]\rVert. In many interesting examples, but not always, this gives rise to a compact quantum metric space. From a discrete group GG equipped with a proper length function ℓ:G→[0,∞)\ell:G\to[0,\infty), one obtains a natural spectral triple (C[G],ℓ2(G),Dℓ)(C[G],\ell^2(G),D_\ell), and even in this situation it has proven very difficult to determine when one obtains a compact quantum metric structure from the associated seminorm LDℓL_{D_\ell}. On the positive side, this has been verified for hyperbolic groups and groups of polynomial growth in [14] and [5], respectively, and only very recently the first counterexample for a word-length function was found by Klisse [10]. The aim of the present paper is two-fold: First, we answer, negatively, a question posed in [8], asking if any compact quantum metric space is in fact automatically a matrix compact quantum metric space. Kaad conjectured that this is not the case, and we confirm this conjecture by showing that the free group on two generators provides a natural counterexample.

Theorem A (see Theorem 3.6). For the free group F2F_2 on two generators, aa and bb, the word-length function arising from the generating set S:={a,a−1,b,b−1}S := \{a,a^{-1},b,b^{-1}\} gives rise to a compact quantum metric space, which is not a matrix compact quantum metric space.

Contrasting this, we provide an affirmative answer in the presence of amenability.

Theorem B (see Theorem 3.1). If GG is an amenable group equipped with a proper length function yielding a compact quantum metric space, then it is automatically a matrix compact quantum metric space.

Secondly, exploiting the tensor product construction for matrix compact quantum metric spaces, we provide a new class of groups for which one does obtain a compact quantum metric structure.

Theorem C (see Theorem 4.10). Let GG and KK be finitely generated groups equipped with word-length functions ℓG\ell_G and ℓK\ell_K. If (C[G],LDℓG)(\mathbb{C}[G],L_{D_{\ell_G}}) is a compact quantum metric space and KK is either finite, abelian or the infinite dihedral group, then (C[G×K],LDℓ)(\mathbb{C}[G \times K],L_{D_\ell}) is a compact quantum metric space for the length function ℓ(g,k)=ℓG(g)+ℓK(k)\ell(g,k)=\ell_G(g)+\ell_K(k).

Theorem C does indeed provide genuinely new examples, in that one may, for instance, take GG to be any non-elementary hyperbolic group and KK to be the integers, then G×KG \times K is neither hyperbolic (as it contains a copy of Z2\mathbb{Z}^2) nor of polynomial growth (as GG must have exponential growth), and is thus not covered by the existing results in [14] and [5].

Notation. Throughout the article, all C∗C^*-algebras are unital, and all spectral triples are faithfully and unitally represented. Moreover, inner products are assumed linear in the first variable. Algebraic tensor products are denoted by ⊙\odot, while ⊗min⁡\otimes_{\min} will denote the minimal tensor product of C∗C^*-algebras, while ⊗\otimes will be used for tensor products in the category of Hilbert spaces. A unital completely positive map between operator systems will be referred to as a UCP map.

Acknowledgments. The present paper grew out of a series of interesting discussions with Jens Kaad and Nadia S. Larsen in the fall of 2025, and it is a pleasure to thank them both for sharing their insights. We also wish to acknowledge that the fundamental idea of geodesic cancellation, which plays a pivotal role in the proof of Theorem C, was originally discovered by Jens in the case of the integers. That Theorem 3.1 and Theorem 4.2 should be true, was also suggested by Jens. The first named author would also like to thank the University of Southern Denmark for its hospitality during a visit in November 2025, during which some of this work was conducted. The first named author was funded by the Research Council of Norway projects 324944 and 359602. The second named author was funded by the Independent Research Fund Denmark through grant no. 1026-00371B and by the European Commission through the MSCA Staff Exchanges grant no. 101086394.

AI declaration. The present paper consists of a mixture of human-generated results and AI-assisted proofs. As indicated above, the fundamental idea of geodesic cancellation leading to Theorem C was discovered and utilised by the authors (without AI assistance but through conversations with Jens Kaad and Nadia S. Larsen) in the case where KK is Z\mathbb{Z} or Z2\mathbb{Z}^2, but the full class of groups covered in Theorem C was then obtained via ChatGPT through a carefully designed research prompt. Similarly, Theorem B was obtained without any AI assistance, whereas the counterexample in Theorem A was obtained through a dialogue with ChatGPT, to explore if Theorem B could be extended to also cover the class of hyperbolic groups (which is then shown to be impossible by Theorem A).

Preliminaries

The present section introduces the necessary background concerning compact quantum metric spaces and their matrix analogues. In the interest of brevity, we will start by introducing the matrix version.

Matrix compact quantum metric spaces

Following [8], we choose the operator system approach to (matrix) compact quantum metric spaces. By an operator system, we mean a unital, ∗*-invariant subspace X\mathcal{X} of a unital C∗C^*-algebra AA. Note that we do not require it to be norm closed. As defined, the ambient C∗C^*-algebra is part of the definition of the operator system X\mathcal{X}, but we remark that the theory also admits an abstract description; see [15] for details.

We will tacitly identify the scalars inside an operator system X\mathcal{X} as C⋅1X\mathbb{C} \cdot1_{\mathcal{X}}. A Lipschitz seminorm on an operator system X\mathcal{X}, is a ∗*-invariant seminorm L:Dom⁡(L)→[0,∞)L:\operatorname{Dom}(L) \to[0,\infty), defined on a dense operator-subsystem Dom⁡(L)⊆X\operatorname{Dom}(L) \subseteq\mathcal{X}, which vanishes on the unit. As is common in the field, we will sometimes regard such a seminorm as an extended seminorm on the whole operator system by setting it to ∞\infty outside its domain. To discuss matrix compact quantum metric spaces, we will adopt the terminology developed in [8]. A Lipschitz operator seminorm on an operator system X\mathcal{X} is a family L=(Ln)n≥1\mathbb{L}=(L_n)_{n\geq1} of ∗*-invariant seminorms on Mn(X)M_n(\mathcal{X}), which vanish on Mn(C)M_n(\mathbb{C}), such that

Ln+m(x⊕y)≤max⁡{Ln(x),Lm(y)}andLn(αxβ)≤∥α∥⋅Ln(x)⋅∥β∥,L_{n+m}(x\oplus y)\leq\max\{L_n(x),L_m(y)\}\qquad\text{and}\qquad L_n(\alpha x\beta)\leq\|\alpha\|\cdot L_n(x)\cdot\|\beta\|,

for all x∈Mn(X)x\in M_n(\mathcal{X}), y∈Mm(X)y\in M_m(\mathcal{X}) and scalar matrices α,β∈Mn(C)\alpha,\beta\in M_n(\mathbb{C}).

Notation 2.1. For an operator system X\mathcal{X} and an n∈Nn\in\mathbb{N}, we will denote by [⋅]:Mn(X)→Mn(X)/Mn(C)[\cdot]:M_n(\mathcal{X})\to M_n(\mathcal{X})/M_n(\mathbb{C}) the natural quotient map. Moreover, for a linear map T:X→YT:\mathcal{X}\to\mathcal{Y} between operator systems, we denote by Tn:Mn(X)→Mn(Y)T_n:M_n(\mathcal{X})\to M_n(\mathcal{Y}) its entrywise matrix amplification.

Definition 2.1 ([8]). A pair (X,L)(\mathcal{X},\mathbb{L}) consisting of an operator system X\mathcal{X} and a Lipschitz operator seminorm L=(Ln)n∈N\mathbb{L}=(L_n)_{n\in\mathbb{N}} is said to be a matrix compact quantum metric space (MCQMS) if the following conditions hold:

  1. There exists an R>0R>0 such that

∥[x]∥≤R⋅Ln(x),for all n≥1, x∈Mn(X).\|[x]\|\leq R\cdot L_n(x),\qquad\text{for all }n\geq1,\ x\in M_n(\mathcal{X}).
  1. For every ε>0\varepsilon>0 there exist an operator system Yε\mathcal{Y}_{\varepsilon}, a unital complete isometry ιε:X→Yε\iota_{\varepsilon}:\mathcal{X}\to\mathcal{Y}_{\varepsilon}, and a finite-rank unital completely positive map Φε:X→Yε\Phi_{\varepsilon}:\mathcal{X}\to\mathcal{Y}_{\varepsilon} satisfying

∥(ιε)n(x)−(Φε)n(x)∥≤ε⋅Ln(x)for all n≥1, x∈Mn(X).\|(\iota_{\varepsilon})_n(x)-(\Phi_{\varepsilon})_n(x)\|\leq\varepsilon\cdot L_n(x)\qquad\text{for all }n\geq1,\ x\in M_n(\mathcal{X}).

Note in particular that RR and the approximating maps are independent of the matrix size nn. For a single Lipschitz seminorm LL, (X,L)(\mathcal{X},L) is called a compact quantum metric space (CQMS) if these conditions hold at n=1n=1, with L1=LL_1=L, and where ιε\iota_{\varepsilon} is only required to be a unital linear isometry and Φε\Phi_{\varepsilon} a bounded, unital positive finite-rank linear map. In this case, LL is called a Lip-norm.

Note first that any MCQMS is a CQMS by restricting the matrix approximations to the n=1n=1 case, see also [8 Proposition 5.4]. Furthermore, the notion of CQMS from Definition 2.1 agrees with Rieffel’s usual definition by [8 Theorem 3.1]. In particular, a Lipschitz seminorm L:X→[0,∞)L:\mathcal{X}\to[0,\infty) is a Lip-norm, if and only if the Monge-Kantorovič metric

mk⁡L(ϕ,ψ):=sup⁡{∣ϕ(x)−ψ(x)∣∣x∈X,L(x)≤1},ϕ,ψ∈S(X),\operatorname{mk}_{L}(\phi,\psi):=\sup\{|\phi(x)-\psi(x)|\mid x\in\mathcal{X},L(x)\leq1\},\qquad\phi,\psi\in S(\mathcal{X}),

induces the weak∗^* topology on the state space S(X)S(\mathcal{X}). Given a classical compact metric space (M,d)(M,d), one obtains [8 Proposition 5.7] an MCQMS by setting Dom⁡(Ln):={f:M→Mn(C)∣f Lipschitz continuous}\operatorname{Dom}(L_n):=\{f:M\to M_n(\mathbb{C})\mid f\text{ Lipschitz continuous}\} and

Ln(f):=sup⁡{∥f(p)−f(q)∥d(p,q)∣p,q∈M,p≠q}.L_n(f):=\sup\left\{\frac{\|f(p)-f(q)\|}{d(p,q)}\mathrel{\Big|}p,q\in M,p\ne q\right\}.

Here we tacitly identify C(M,Mn(C))C(M,M_n(\mathbb{C})) with Mn(C(M))M_n(C(M)).

We end this section, by recalling the following well known useful criterion.

Proposition 2.2. Fix μ∈S(X)\mu\in S(\mathcal{X}), and put

EL,μ={x∈X∣L(x)≤1,μ(x)=0}.E_{L,\mu}=\{x\in\mathcal{X}\mid L(x)\leq1,\mu(x)=0\}.

Then (X,L)(\mathcal{X},L) is a CQMS if and only if EL,μE_{L,\mu} is totally bounded in norm. In this case, ker⁡L=C\ker L=\mathbb{C}, the radius Rμ:=sup⁡x∈EL,μ∥x∥R_{\mu}:=\sup_{x\in E_{L,\mu}}\|x\| is finite, and

∥x−μ(x)1∥≤Rμ⋅L(x)\|x-\mu(x)1\|\leq R_{\mu}\cdot L(x)

for all x∈Xx\in\mathcal{X}. Moreover, for every s,t≥0s,t\geq0, the set {x∈X∣L(x)≤s,∣μ(x)∣≤t}\{x\in\mathcal{X}\mid L(x)\leq s,|\mu(x)|\leq t\} is totally bounded.

Proof. The statement on total boundedness of EL,μE_{L,\mu} follows from [14 Proposition 1.3]. It is well known that if (X,L)(\mathcal{X},L) is a CQMS, then ker⁡L=C\ker L = \mathbb{C}; see e.g. [9 Lemma 2.2] for a proof. The statement ∥x−μ(x)1∥≤RμL(x)\lVert x-\mu(x)1\rVert\le R_{\mu}L(x) follows from noting that x−μ(x)1∈L(x)⋅EL,μx-\mu(x)1\in L(x)\cdot E_{L,\mu}. Total boundedness of {x∈X∣L(x)≤s,∣μ(x)∣≤t}\{x\in\mathcal{X}\mid L(x)\le s,\lvert\mu(x)\rvert\le t\} now follows by noting that it is contained in the totally bounded set sEL,μ+{c1∣∣c∣≤t}sE_{L,\mu}+\{c1\mid\lvert c\rvert\le t\}.

Many interesting non-commutative examples of (M)CQMS arise from spectral triples [4], and these are the primary objects of study in the present paper. In fact, we will be focusing (almost) exclusively on the spectral triples that arise from length functions on discrete groups, for which the non-commutative metric geometry is already very complex [14, 5, 10].

Spectral triples and length functions on groups

To set the notation, we recall that a unital spectral triple (A,H,D)(\mathcal{A},H,D) consists of a unital ∗*-algebra faithfully represented on the Hilbert space HH together with a self-adjoint, possibly unbounded operator DD on HH with compact resolvent, such that A\mathcal{A} preserves Dom⁡D\operatorname{Dom}D and each commutator [D,a][D,a] is bounded. Given a spectral triple (A,H,D)(\mathcal{A},H,D), we denote the C∗C^*-closure of A\mathcal{A} by AA. Note that one obtains a semi-norm LDL_D on A\mathcal{A} by setting

LD(a)=∥[D,a]∥,L_D(a)=\lVert[D,a]\rVert,

which in many interesting cases, but not always [10, 11], gives rise to a CQMS. A spectral triple (A,H,D)(\mathcal{A},H,D) is called a spectral metric space if its commutator seminorm is a Lip-norm, i.e., if (A,LD)(\mathcal{A},L_D) is a CQMS. Amplifying to matrices, we obtain a seminorm LD,n:Mn(A)→[0,∞)L_{D,n}:M_n(\mathcal{A})\to[0,\infty) by setting

LD,n(x):=∥[1⊗D,x]∥L_{D,n}(x):=\lVert[1\otimes D,x]\rVert

and (A,H,D)(\mathcal{A},H,D) is called a matrix spectral metric space if (A,(LD,n)n≥1)(\mathcal{A},(L_{D,n})_{n\ge1}) is an MCQMS; see [8]. We will need to amplify unbounded operators as well. If D:Dom⁡(D)→HD:\operatorname{Dom}(D)\to H is an unbounded operator on a Hilbert space HH and KK is some auxiliary Hilbert space then we obtain an unbounded operator D⊙1:Dom⁡(D)⊙K→H⊗KD\odot1:\operatorname{Dom}(D)\odot K\to H\otimes K defined by (D⊙1)(ξ⊗η)=Dξ⊗η(D\odot1)(\xi\otimes\eta)=D\xi\otimes\eta. If DD is furthermore selfadjoint, then D⊙1D\odot1 is essentially selfadjoint and we denote its selfadjoint closure by D⊗1D\otimes1. For two spectral triples (A,HA,DA),(B,HB,DB)(\mathcal{A},H_A,D_A),(\mathcal{B},H_B,D_B), put

L(1)(z)=∥[DA⊙1,z]∥,L(2)(z)=∥[1⊙DB,z]∥,L^{(1)}(z)=\lVert[D_A\odot1,z]\rVert,\qquad L^{(2)}(z)=\lVert[1\odot D_B,z]\rVert,

for z∈A⊙Bz\in\mathcal{A}\odot\mathcal{B}. Note that these are indeed well defined, since for z=∑k=1nak⊗bk∈A⊙Bz=\sum_{k=1}^{n}a_k\otimes b_k\in\mathcal{A}\odot\mathcal{B} preserves Dom⁡(D)⊙HB\operatorname{Dom}(D)\odot H_B, and on this subspace we have

[DA⊙1,z]=∑k=1n[DA,ak]⊗bk,[D_A\odot1,z]=\sum_{k=1}^{n}[D_A,a_k]\otimes b_k,

which is bounded (and similarly for L(2)L^{(2)}).

The classical commutative example, upon which the definition of a spectral triple is modelled, arises by choosing A\mathcal{A} to be the algebra of smooth functions on a compact, connected spin manifold and DD to be the associated Dirac operator. An interesting source of non-commutative examples is obtained by considering length functions on discrete groups, as was already realised by Connes in his seminal paper [3].

Definition 2.3. By a length function function on a discrete group GG we will mean a function ℓ:G→[0,∞)\ell:G\to[0,\infty) with

  1. ℓ(g)=0\ell(g)=0 if and only if g=eg=e,

  2. ℓ(g−1)=ℓ(g)\ell(g^{-1})=\ell(g) for all g∈Gg\in G,

  3. ℓ(gh)≤ℓ(g)+ℓ(h)\ell(gh)\le\ell(g)+\ell(h) for all g,h∈Gg,h\in G.

We say ℓ\ell is proper if the balls Bℓ(t):={g∈G∣ℓ(g)≤t}B_{\ell}(t):=\{g\in G\mid\ell(g)\le t\} are finite for all t≥0t\ge0.

Assume that ℓ\ell is proper. We represent C[G]⊆C∗(G)\mathbb{C}[G]\subseteq C^*(G) on ℓ2(G)\ell^2(G) through the left regular representation λgδh=δgh\lambda_g\delta_h=\delta_{gh} and denote by (δg)g∈G(\delta_g)_{g\in G} the standard orthonormal basis in ℓ2(G)\ell^2(G). The relation Dℓ:span⁡{δg∣g∈G}→ℓ2(G)D_{\ell}:\operatorname{span}\{\delta_g\mid g\in G\}\to\ell^2(G) defined by Dℓδg:=ℓ(g)⋅δgD_{\ell}\delta_g:=\ell(g)\cdot\delta_g is essentially selfadjoint, and its closure DℓD_{\ell} has compact resolvent by properness of ℓ\ell. Each λg\lambda_g preserves the domain Dom⁡Dℓ={ξ∈ℓ2(G)∣∑gℓ(g)2∣ξ(g)∣2<∞}\operatorname{Dom}D_{\ell}=\{\xi\in\ell^2(G)\mid\sum_g\ell(g)^2|\xi(g)|^2<\infty\}, and

[Dℓ,λg]δh=(ℓ(gh)−ℓ(h))δgh,(1)[D_{\ell},\lambda_g]\delta_h=(\ell(gh)-\ell(h))\delta_{gh}, \tag*{(1)}

and hence ∥[Dℓ,λg]∥≤ℓ(g)\lVert[D_{\ell},\lambda_g]\rVert\leq\ell(g). Thus, (C[G],ℓ2(G),Dℓ)(\mathbb{C}[G],\ell^2(G),D_{\ell}) is a unital spectral triple. Set

Lℓ,n(x):=∥[1n⊗Dℓ,x]∥,andLℓ:=Lℓ,1.L_{\ell,n}(x):=\lVert[1_n\otimes D_{\ell},x]\rVert,\qquad\text{and}\qquad L_{\ell}:=L_{\ell,1}.

We record that it is known that (C[G],ℓ2(G),Dℓ)(\mathbb{C}[G],\ell^2(G),D_{\ell}) is a spectral metric space if ℓ\ell has bounded doubling [5], or if GG is word-hyperbolic and ℓ\ell is the word length function on GG [14] — or more generally a length function whose induced metric is hyperbolic and weakly geodesic [2]. A recent example by Klisse [10] shows that a word length on (Z/2)≀F2(\mathbb{Z}/2)\wr F_2 does not yield a spectral metric space.

For discrete groups GG and KK with length functions ℓG\ell_G and ℓK\ell_K, respectively, we obtain a natural length function on G×KG\times K by setting

ℓ(g,k):=ℓG(g)+ℓK(k).\ell(g,k):=\ell_G(g)+\ell_K(k).

Note that the Dirac operator DℓD_{\ell} agrees with DℓG⊗1+1⊗DℓKD_{\ell_G}\otimes1+1\otimes D_{\ell_K} on the dense subspace cc(G)⊙cc(K)⊆ℓ2(G)⊗ℓ2(K)c_c(G)\odot c_c(K)\subseteq\ell^2(G)\otimes\ell^2(K). Controlling ∥[DℓG⊗1,z]∥\lVert[D_{\ell_G}\otimes1,z]\rVert and ∥[1⊗DℓK,z]∥\lVert[1\otimes D_{\ell_K},z]\rVert in terms of ∥[Dℓ,z]∥\lVert[D_{\ell},z]\rVert will turn out to be a crucial step in our strategy for exhibiting a CQMS structure on the product group G×KG\times K using ℓ=ℓG+ℓK\ell=\ell_G+\ell_K. We will employ a geodesic cancellation argument, see § 4.2.

Lastly, note that if ℓG\ell_G and ℓK\ell_K are word length functions arising from symmetric generating sets SGS_G and SKS_K for GG and KK, respectively, then the length function on G×KG\times K defined above is also a word length function corresponding to the generating set {(s,e),(e,t)∣s∈SG,t∈SK}\{(s,e),(e,t)\mid s\in S_G,t\in S_K\}.

Groups and matrix compact quantum metric spaces

In this section we investigate the question of whether any CQMS is in fact an MCQMS. This question already appears in [8], in which Kaad conjectures that this is not generally the case. We first show that in the context of length functions on amenable groups, one does indeed automatically obtain an MCQMS from a CQMS, and then proceed to show that the free group on two generators provides a counter example.

Theorem 3.1. Let GG be an amenable, countable discrete group and ℓ\ell a proper length function on GG. If (C[G],Lℓ)(\mathbb{C}[G],L_{\ell}) is a CQMS, then (C[G],(Lℓ,n)n≥1)(\mathbb{C}[G],(L_{\ell,n})_{n\geq1}) is an MCQMS.

The proof is a modification of the argument for [8 Theorem 5.4]. Write A=Cr∗(G)A=C_r^*(G), H=ℓ2(G)H=\ell^2(G), and D=DℓD=D_{\ell}, and let Δ:C[G]→C[G]⊙C[G]\Delta:\mathbb{C}[G]\to\mathbb{C}[G]\odot\mathbb{C}[G] be the coproduct determined by Δ(λg)=λg⊗λg\Delta(\lambda_g)=\lambda_g\otimes\lambda_g. We shall use the isometric ∗*-isomorphism

ιp:Mp(B(H))⊗min⁡B(H)⟶Mp(B(H)⊗min⁡B(H)),ιp((aij)ij⊗b)=(aij⊗b)ij.\iota_p:M_p(B(H))\otimes_{\min}B(H)\longrightarrow M_p(B(H)\otimes_{\min}B(H)),\qquad\iota_p((a_{ij})_{ij}\otimes b)=(a_{ij}\otimes b)_{ij}.

That ιp\iota_p is indeed an isometry follows by applying the isomorphism Mp(E)≅Mp(C)⊗min⁡EM_p(E)\cong M_p(\mathbb{C})\otimes_{\min}E and associativity of the minimal tensor product. We use the same notation for its restrictions to the relevant subalgebras. We also use the isometric map

κp:Mp(B(H))⟶Mp(B(H)⊗min⁡B(H)),κp((aij)ij)=(1⊗aij)ij.\kappa_p:M_p(B(H))\longrightarrow M_p(B(H)\otimes_{\min}B(H)),\qquad\kappa_p((a_{ij})_{ij})=(1\otimes a_{ij})_{ij}.

Indeed, κp\kappa_p is the matrix amplification of the injective unital ∗*-homomorphism T↦1⊗TT\mapsto1\otimes T and hence isometric. For ω∈A∗\omega\in A^* and z∈Mp(C[G]⊙C[G])z\in M_p(\mathbb{C}[G]\odot\mathbb{C}[G]), these identifications give the identity

(id⁡A⊗ω)p(z)=(id⁡Mp(A)⊗ω)ιp−1(z).(2)(\operatorname{id}_A\otimes\omega)_p(z)=(\operatorname{id}_{M_p(A)}\otimes\omega)\iota_p^{-1}(z). \tag*{(2)}

We note the following lemma.

Lemma 3.2. For every p≥1p\geq1, every μ∈Mp(A)∗\mu\in M_p(A)^*, and every x∈Mp(C[G])x\in M_p(\mathbb{C}[G]), we have

Lℓ((μ⊗id⁡)ιp−1Δp(x))≤∥μ∥⋅Lℓ,p(x).L_{\ell}((\mu\otimes\operatorname{id})\iota_p^{-1}\Delta_p(x))\leq\lVert\mu\rVert\cdot L_{\ell,p}(x).

Proof. Consider the Kac–Takesaki unitary W∈B(H⊗H)W\in B(H\otimes H) given by

W(δs⊗δt)=δt−1s⊗δt.W(\delta_s\otimes\delta_t)=\delta_{t^{-1}s}\otimes\delta_t.

It satisfies W∗(1⊗λg)W=λg⊗λgW^{*}(1\otimes\lambda_g)W=\lambda_g\otimes\lambda_g. Consequently, Δ\Delta extends to an injective unital ∗*-homomorphism A→A⊗min⁡AA\to A\otimes_{\min}A, given by a↦W∗(1⊗a)Wa\mapsto W^{*}(1\otimes a)W. Both WW and W∗W^{*} preserve the dense subspace C:=cc(G)⊙cc(G)⊆ℓ2(G)⊗ℓ2(G)\mathcal{C}:=c_c(G)\odot c_c(G)\subseteq\ell^2(G)\otimes\ell^2(G) and commute thereon with 1⊗D1\otimes D. For a∈C[G]a\in\mathbb{C}[G], it follows that

[1⊗D,Δ(a)]=W∗(1⊗[D,a])W.(3)[1\otimes D,\Delta(a)]=W^{*}(1\otimes[D,a])W. \tag*{(3)}

The identity holds on C\mathcal{C}, and hence for the bounded extensions. Denote by eije_{ij} the standard matrix units in Mp(C)M_p(\mathbb{C}) and write λgeij\lambda_g e_{ij} for the matrix in Mp(C[G])M_p(\mathbb{C}[G]) whose (i,j)(i,j)-entry is λg\lambda_g and whose other entries are zero. For μ∈Mp(A)∗\mu\in M_p(A)^*, we get

(μ⊗id⁡)ιp−1Δp(λgeij)=μ(λgeij)λg.(\mu\otimes\operatorname{id})\iota_p^{-1}\Delta_p(\lambda_g e_{ij})=\mu(\lambda_g e_{ij})\lambda_g.

We therefore have

[D,(μ⊗id⁡)ιp−1Δp(λgeij)]=μ(λgeij)[D,λg]=(μ⊗id⁡)((λgeij)⊗[D,λg])=(μ⊗id⁡)ιp−1([1⊗D,Δ(λg)]eij)\begin{aligned} [D,(\mu\otimes\operatorname{id})\iota_p^{-1}\Delta_p(\lambda_g e_{ij})] &=\mu(\lambda_g e_{ij})[D,\lambda_g]\\ &=(\mu\otimes\operatorname{id})((\lambda_g e_{ij})\otimes[D,\lambda_g])\\ &=(\mu\otimes\operatorname{id})\iota_p^{-1}([1\otimes D,\Delta(\lambda_g)]e_{ij}) \end{aligned}

Since these matrix units with group entries span Mp(C[G])M_p(\mathbb{C}[G]) we therefore obtain

[D,(μ⊗id⁡)ιp−1Δp(x)]=(μ⊗id⁡)ιp−1(([1⊗D,Δ(xij)])ij)(4)[D,(\mu\otimes\operatorname{id})\iota_p^{-1}\Delta_p(x)]=(\mu\otimes\operatorname{id})\iota_p^{-1}(([1\otimes D,\Delta(x_{ij})])_{ij}) \tag*{(4)}

for every x∈Mp(C[G])x\in M_p(\mathbb{C}[G]). Since any bounded functional is completely bounded with cb-norm equal to its norm [15 Proposition 3.8], we have ∥μ⊗id⁡∥≤∥μ∥\|\mu\otimes\operatorname{id}\|\leq\|\mu\| on Mp(A)⊗min⁡B(H)M_p(A)\otimes_{\min}B(H). Using this bound together with the fact that ιp−1\iota_p^{-1} and κp\kappa_p are isometries and WW is unitary, we now get

Lℓ((μ⊗id⁡)ιp−1Δp(x))=∥[D,(μ⊗id⁡)ιp−1Δp(x)]∥=∥(μ⊗id⁡)ιp−1(([1⊗D,Δ(xij)])ij)∥(by (3.3))≤∥μ∥⋅∥([1⊗D,Δ(xij)])ij∥=∥μ∥⋅∥(W∗(1⊗[D,xij])W)ij∥(by (3.2))=∥μ∥⋅∥((1⊗[D,xij])ij∥=∥μ∥⋅∥κp([1⊗D,x])∥=∥μ∥⋅Lℓ,p(x).\begin{aligned} L_{\ell}((\mu\otimes\operatorname{id})\iota_p^{-1}\Delta_p(x)) &=\|[D,(\mu\otimes\operatorname{id})\iota_p^{-1}\Delta_p(x)]\|\\ &=\|(\mu\otimes\operatorname{id})\iota_p^{-1}(([1\otimes D,\Delta(x_{ij})])_{ij})\| &&\text{(by (3.3))}\\ &\leq\|\mu\|\cdot\|([1\otimes D,\Delta(x_{ij})])_{ij}\|\\ &=\|\mu\|\cdot\|(W^{*}(1\otimes[D,x_{ij}])W)_{ij}\| &&\text{(by (3.2))}\\ &=\|\mu\|\cdot\|((1\otimes[D,x_{ij}])_{ij}\|\\ &=\|\mu\|\cdot\|\kappa_p([1\otimes D,x])\|\\ &=\|\mu\|\cdot L_{\ell,p}(x). \end{aligned}

Proof of Theorem 3.1. Choose a Følner sequence (Fn)n≥0(F_n)_{n\geq0} for GG, with F0={e}F_0=\{e\}. Define unit vectors and their associated vector states by

ξn=∣Fn∣−1/2∑g∈Fnδgandχn(a)=⟨aξn,ξn⟩for a∈A.\xi_n=|F_n|^{-1/2}\sum_{g\in F_n}\delta_g \qquad\text{and}\qquad \chi_n(a)=\langle a\xi_n,\xi_n\rangle\quad\text{for }a\in A.

The Følner property gives χn(λg)=∣Fn∩gFn∣/∣Fn∣→1\chi_n(\lambda_g)=|F_n\cap gF_n|/|F_n|\to1 for every g∈Gg\in G. Thus the trivial representation ϵ(∑gagλg):=∑gag\epsilon(\sum_g a_g\lambda_g):=\sum_g a_g is the pointwise limit of χn\chi_n on C[G]\mathbb{C}[G] and, by amenability, ϵ\epsilon extends to a state on AA. By density, we therefore have lim⁡nχn=ϵ\lim_n\chi_n=\epsilon in the weak-∗* topology. The CQMS hypothesis implies that δn:=mk⁡Lℓ(ϵ,χn)→0\delta_n:=\operatorname{mk}_{L_\ell}(\epsilon,\chi_n)\to0 and by definition of mk⁡Lℓ\operatorname{mk}_{L_\ell} we have

∣(ϵ−χn)(b)∣≤δn⋅Lℓ(b)for b∈C[G].(5)|(\epsilon-\chi_n)(b)|\leq\delta_n\cdot L_\ell(b)\qquad\text{for }b\in\mathbb{C}[G]. \tag*{(5)}

Consider now the maps βn=(id⁡⊗χn)Δ:C[G]→C[G]\beta_n=(\operatorname{id}\otimes\chi_n)\Delta:\mathbb{C}[G]\to\mathbb{C}[G]. They are UCP, and they have finite rank because βn(λg)=χn(λg)λg\beta_n(\lambda_g)=\chi_n(\lambda_g)\lambda_g and χn(λg)=0\chi_n(\lambda_g)=0 if g∉FnFn−1g\notin F_nF_n^{-1}. We now show that these maps give matricial approximations. Fix p≥1p\geq1, x∈Mp(C[G])x\in M_p(\mathbb{C}[G]), and μ∈Mp(A)∗\mu\in M_p(A)^* with ∥μ∥≤1\|\mu\|\leq1. The identity (id⁡⊗ϵ)Δ=id⁡(\operatorname{id}\otimes\epsilon)\Delta=\operatorname{id} gives

∣μ(x−(βn)p(x))∣=∣μ((id⁡A⊗(ϵ−χn))pΔp(x))∣=∣μ((id⁡Mp(A)⊗(ϵ−χn))ιp−1Δp(x))∣=∣(ϵ−χn)((μ⊗id⁡)ιp−1Δp(x))∣≤δn⋅Lℓ((μ⊗id⁡)ιp−1Δp(x))≤δn⋅Lℓ,p(x).\begin{aligned} |\mu(x-(\beta_n)_p(x))| &=|\mu((\operatorname{id}_A\otimes(\epsilon-\chi_n))_p\Delta_p(x))|\\ &=|\mu((\operatorname{id}_{M_p(A)}\otimes(\epsilon-\chi_n))\iota_p^{-1}\Delta_p(x))|\\ &=|(\epsilon-\chi_n)((\mu\otimes\operatorname{id})\iota_p^{-1}\Delta_p(x))|\\ &\leq\delta_n\cdot L_\ell((\mu\otimes\operatorname{id})\iota_p^{-1}\Delta_p(x))\\ &\leq\delta_n\cdot L_{\ell,p}(x). \end{aligned}

The second equality is (2), the third interchanges slices on different tensor factors and the inequalities are (5) and Lemma 3.2, respectively. Since

∥x−(βn)p(x)∥=sup⁡μ∈Mp(A)∗∥μ∥≤1∣μ(x−(βn)p(x))∣,\lVert x-(\beta_n)_p(x)\rVert=\sup_{\substack{\mu\in M_p(A)^*\\\lVert\mu\rVert\leq1}}\left|\mu\left(x-(\beta_n)_p(x)\right)\right|,

we obtain

∥x−(βn)p(x)∥≤δnLℓ,p(x),(6)\lVert x-(\beta_n)_p(x)\rVert\leq\delta_nL_{\ell,p}(x), \tag*{(6)}

for all n≥0n\geq0 and all p≥1p\geq1. For a given ε>0\varepsilon>0, choose nn with δn≤ε\delta_n\leq\varepsilon. Then (id⁡,βn)(\operatorname{id},\beta_n) is a matricial ε\varepsilon-approximation in C[G]\mathbb{C}[G], with the same maps for every matrix size.

Finally, F0={e}F_0=\{e\} gives χ0=τ\chi_0=\tau and β0(a)=τ(a)1\beta_0(a)=\tau(a)1 where τ\tau denotes the canonical trace on AA. Applying (6) with n=0n=0 yields

∥[x]∥≤∥x−(β0)p(x)∥≤mk⁡Lℓ(ε,τ)⋅Lℓ,p(x),\lVert[x]\rVert\leq\lVert x-(\beta_0)_p(x)\rVert\leq\operatorname{mk}_{L_\ell}(\varepsilon,\tau)\cdot L_{\ell,p}(x),

and R:=mk⁡Lℓ(ε,τ)R:=\operatorname{mk}_{L_\ell}(\varepsilon,\tau) is finite because mk⁡Lℓ\operatorname{mk}_{L_\ell} metrizes the weak-* topology on the compact state space. This proves uniformly finite matrix diameter and completes the proof. □

Note in particular that the examples considered in [5] have bounded doubling and are therefore amenable. As noted therein, any finitely generated group with polynomial growth has bounded doubling with respect to any word length. We have the following immediate corollary.

Corollary 3.3. If GG is a finitely generated discrete group and ℓ\ell is a proper length function with bounded doubling, then (C[G],(Lℓ,n)n≥1)(\mathbb{C}[G],(L_{\ell,n})_{n\geq1}) is an MCQMS.

In addition to groups of polynomial growth, it is also known that word-length functions on hyperbolic groups yield compact quantum metric spaces [14]. Most hyperbolic groups are non-amenable, and as we shall now see the situation is quite different in this case. Indeed, we proceed to show that F2F_2 with its standard word length function does not yield an MCQMS, despite it being hyperbolic and thus yielding a CQMS by the results in [14]. Before proceeding to the proof of this result, we recall the rank-one case of the projection-commutator identity in [18 Example 6.2].

Lemma 3.4. Let B=B∗∈MN(C)B=B^*\in M_N(\mathbb{C}), and suppose that v∈CNv\in\mathbb{C}^N is a unit vector. Then the projection p=vv∗p=vv^* satisfies

∥[B,p]∥2=∥Bv∥2−∣v∗Bv∣2.\lVert[B,p]\rVert^2=\lVert Bv\rVert^2-\lvert v^*Bv\rvert^2.

In particular, if v=N−1/2(1,…,1)tv=N^{-1/2}(1,\ldots,1)^t and B=diag⁡(b1,…,bN)B=\operatorname{diag}(b_1,\ldots,b_N) with bj∈Rb_j\in\mathbb{R}, then

∥[B,p]∥2=1N∑j=1Nbj2−(1N∑j=1Nbj)2.\lVert[B,p]\rVert^2=\frac{1}{N}\sum_{j=1}^{N}b_j^2-\left(\frac{1}{N}\sum_{j=1}^{N}b_j\right)^2.

Proof. By [18 Example 6.2] we have ∥[P,A]∥=max⁡{∥PA(1−P)∥,∥(1−P)AP∥}\lVert[P,A]\rVert=\max\{\lVert PA(1-P)\rVert,\lVert(1-P)AP\rVert\} for a projection PP and a general element AA in a C∗C^*-algebra. For P=p=vv∗P=p=vv^* and A=B=B∗A=B=B^* the two terms coincide, and we get

∥[B,p]∥2=sup⁡∥w∥=1∥(1−p)Bpw∥2=sup⁡∥w∥=1∥(1−p)B⟨w,v⟩v∥2=∥(1−p)Bv∥2=∥Bv∥2−∥pBv∥2=∥Bv∥2−∥⟨v,Bv⟩v∥2=∥Bv∥2−∣v∗Bv∣2.\begin{aligned} \lVert[B,p]\rVert^2 &=\sup_{\lVert w\rVert=1}\lVert(1-p)Bpw\rVert^2 =\sup_{\lVert w\rVert=1}\lVert(1-p)B\langle w,v\rangle v\rVert^2 =\lVert(1-p)Bv\rVert^2\\ &=\lVert Bv\rVert^2-\lVert pBv\rVert^2 =\lVert Bv\rVert^2-\lVert\langle v,Bv\rangle v\rVert^2 =\lVert Bv\rVert^2-\lvert v^*Bv\rvert^2. \end{aligned}

□

Remark 3.5. Considering X={1,…,N}X=\{1,\ldots,N\} as a probability space with respect to the uniform probability measure μ\mu, then if b:X→Rb:X\to\mathbb{R} is some random variable, we have that

Var⁡(b):=E(b2)−E(b)2=1N∑j=1Nbj2−(1N∑j=1Nbj)2.\operatorname{Var}(b):=\mathbb{E}(b^2)-\mathbb{E}(b)^2=\frac{1}{N}\sum_{j=1}^{N}b_j^2-\left(\frac{1}{N}\sum_{j=1}^{N}b_j\right)^2.

The righthand side of the last formula in Lemma 3.4 may therefore be interpreted as the variance of a random variable, a point of view that will be important in the proof of the following result.

Theorem 3.6. Let G=F2G=\mathbb{F}_2 be the free group on two generators, which we denote aa and bb, and let ℓ\ell be the word length for the symmetric generating set {a,a−1,b,b−1}\{a,a^{-1},b,b^{-1}\}. Then (C[F2],Lℓ)(\mathbb{C}[\mathbb{F}_2],L_\ell) is a CQMS, but (C[F2],(Lℓ,n)n≥1)(\mathbb{C}[\mathbb{F}_2],(L_{\ell,n})_{n\geq1}) is not an MCQMS. More precisely, for every operator system YY, every linear complete isometry ι:C[F2]→Y\iota:\mathbb{C}[\mathbb{F}_2]\to Y, and every finite-rank linear map Φ:C[F2]→Y\Phi:\mathbb{C}[\mathbb{F}_2]\to Y bounded with respect to the reduced C∗C^*-norm, we have

sup⁡n≥1 sup⁡X∈Mn(C[F2])Lℓ,n(X)≤1∥ιn(X)−Φn(X)∥≥13.(7)\sup_{n\geq1}\ \sup_{\substack{X\in M_n(\mathbb{C}[\mathbb{F}_2])\\L_{\ell,n}(X)\leq1}}\|\iota_n(X)-\Phi_n(X)\|\geq\frac{1}{\sqrt{3}}. \tag*{(7)}

Proof. The claim that (C[F2],Lℓ)(\mathbb{C}[\mathbb{F}_2],L_\ell) is a CQMS follows from [14 Proposition 4.3]. Fix now an integer R≥1R\geq1, and set

SR={g∈G∣ℓ(g)=R}.S_R=\{g\in G\mid\ell(g)=R\}.

Since there are four choices for the first letter and three for each subsequent letter we have NR:=∣SR∣=4⋅3R−1N_R:=|S_R|=4\cdot3^{R-1}. Then define

PR=1NR[λgh−1]g,h∈SR∈MNR(C[F2]).P_R=\frac{1}{N_R}[\lambda_{gh^{-1}}]_{g,h\in S_R}\in M_{N_R}(\mathbb{C}[\mathbb{F}_2]).

Now note that the map

VR:ℓ2(G)→CNR⊗ℓ2(G),VRξ=NR−1/2(λgξ)g∈SRV_R:\ell^2(G)\to\mathbb{C}^{N_R}\otimes\ell^2(G),\qquad V_R\xi=N_R^{-1/2}(\lambda_g\xi)_{g\in S_R}

is an isometry with PR=VRVR∗P_R=V_RV_R^*; hence PRP_R is a nonzero projection. We show that Lℓ,NR(PR)≤3L_{\ell,N_R}(P_R)\leq\sqrt{3}. To this end, define

WR:=diag⁡g∈SR(λg)∈MNR(C[F2]),vR:=NR−1/2(1,…,1)t∈CNRandpR:=vRvR∗∈MNR(C).W_R:=\operatorname{diag}_{g\in S_R}(\lambda_g)\in M_{N_R}(\mathbb{C}[\mathbb{F}_2]),\qquad v_R:=N_R^{-1/2}(1,\ldots,1)^t\in\mathbb{C}^{N_R}\quad\text{and}\quad p_R:=v_Rv_R^*\in M_{N_R}(\mathbb{C}).

Then WR∗PRWR=[1NR]g,h∈SR=pR⊗1W_R^*P_RW_R=\left[\frac{1}{N_R}\right]_{g,h\in S_R}=p_R\otimes1. Note that WRW_R and WR∗W_R^* preserve the domain of INR⊗DI_{N_R}\otimes D. On the summand CNR⊗Cδx\mathbb{C}^{N_R}\otimes\mathbb{C}\delta_x the operator WR∗(INR⊗D)WRW_R^*(I_{N_R}\otimes D)W_R acts as Bx:=diag⁡g∈SRℓ(gx)B_x:=\operatorname{diag}_{g\in S_R}\ell(gx). Using that WRW_R is unitary, Lemma 3.4 then gives

Lℓ,NR(PR)2=∥[INR⊗Dℓ,PR]∥2=∥[WR∗(INR⊗Dℓ)WR,pR⊗1]∥2=sup⁡x∈F2∥[Bx,pR]∥2(8)\begin{aligned} L_{\ell,N_R}(P_R)^2&=\|[I_{N_R}\otimes D_\ell,P_R]\|^2\\ &=\|[W_R^*(I_{N_R}\otimes D_\ell)W_R,p_R\otimes1]\|^2\\ &=\sup_{x\in F_2}\|[B_x,p_R]\|^2 \tag*{(8)} \end{aligned}
=sup⁡x∈F2Var⁡(ℓ(−x)),(9)=\sup_{x\in F_2}\operatorname{Var}(\ell(-x)), \tag*{(9)}

where in the last equality we consider the function SR∋g↦ℓ(gx)∈RS_R\ni g\mapsto\ell(gx)\in\mathbb{R} as a random variable on SRS_R endowed with the uniform probability measure; see e.g. Remark 3.5.

For x∈F2x\in F_2 and g∈SRg\in S_R, denote by C(g,x)C(g,x) the number of cancellation pairs in the reduced product gxgx. Then

ℓ(gx)=R+ℓ(x)−2C(g,x).\ell(gx)=R+\ell(x)-2C(g,x).

We now fix an x∈F2x\in F_2 and consider the random variable C:SR→RC:S_R\to\mathbb{R} given by C(g):=C(g,x)C(g):=C(g,x). If the reduced word for xx is t1⋯tmt_1\cdots t_m, the event C≥kC\geq k, for 1≤k≤min⁡(R,m)1\leq k\leq\min(R,m), prescribes the suffix tk−1⋯t1−1t_k^{-1}\cdots t_1^{-1} to gg. There are 3R−k3^{R-k} such words, so

P(C≥k)=3R−k4⋅3R−1=14⋅3k−1.\mathbb{P}(C\geq k)=\frac{3^{R-k}}{4\cdot3^{R-1}}=\frac{1}{4\cdot3^{k-1}}.

For larger kk the probability is zero. We now note that C2=∑k=1∞(2k−1)1C≥kC^2=\sum_{k=1}^{\infty}(2k-1)1_{C\geq k}; indeed, if C(g)=jC(g)=j then it holds that j2=∑k=1j(2k−1)j^2=\sum_{k=1}^{j}(2k-1). We thus obtain

E[C2]=E[∑k=1∞(2k−1)1C≥k]=∑k=1∞(2k−1)P(C≥k)≤14∑k=1∞(2k−1)3−(k−1)=34.\mathbb{E}[C^2]=\mathbb{E}\left[\sum_{k=1}^{\infty}(2k-1)1_{C\geq k}\right]=\sum_{k=1}^{\infty}(2k-1)\mathbb{P}(C\geq k)\leq\frac{1}{4}\sum_{k=1}^{\infty}(2k-1)3^{-(k-1)}=\frac{3}{4}.

Now, from ℓ(gx)=R+ℓ(x)−2C(g)\ell(gx)=R+\ell(x)-2C(g) we obtain ℓ(−x)−E(ℓ(−x))=−2(C−E(C))\ell(-x)-\mathbb{E}(\ell(-x))=-2(C-\mathbb{E}(C)). Thus

Var⁡(ℓ(−x))=E[ℓ(−x)−E[ℓ(−x)]]2=4E[(C−E[C])2]=4Var⁡(C)=4(E[C2]−(E[C])2)≤4E[C2]≤3.\operatorname{Var}(\ell(-x))=\mathbb{E}\left[\ell(-x)-\mathbb{E}[\ell(-x)]\right]^2=4\mathbb{E}\left[(C-\mathbb{E}[C])^2\right] =4\operatorname{Var}(C)=4\left(\mathbb{E}[C^2]-(\mathbb{E}[C])^2\right)\leq4\mathbb{E}[C^2]\leq3.

The assertion Lℓ,NR(PR)≤3L_{\ell,N_R}(P_R)\leq\sqrt{3} now follows from (3.8).

Consider now a bounded finite rank map Φ:C[F2]→Y\Phi:\mathbb{C}[F_2]\to Y for some operator system YY. We now aim to show that

lim⁡R→∞∥ΦNR(PR)∥=0.\lim_{R\to\infty}\left\|\Phi_{N_R}(P_R)\right\|=0.

An application of [7 Lemma 1.4] shows that

∥∑g∈SRcgλg∥≤(R+1)(∑g∈SR∣cg∣2)1/2.(10)\left\|\sum_{g\in S_R}c_g\lambda_g\right\|\leq(R+1)\left(\sum_{g\in S_R}|c_g|^2\right)^{1/2}. \tag*{(10)}

for any family (cg)g∈SR(c_g)_{g\in S_R} of complex numbers. For a unit vector α∈CNR\alpha\in\mathbb{C}^{N_R}, set

Uα:=∑g∈SRαgλg.U_\alpha:=\sum_{g\in S_R}\alpha_g\lambda_g.

Then for any f∈A∗f\in A^* and any two unit vectors α,β∈CNR\alpha,\beta\in\mathbb{C}^{N_R}, we have

⟨fNR(PR)β,α⟩=α∗fNR(PR)β=1NR∑g,hαg‾βhf(λgh−1)=1NRf(UαUβ∗).\left\langle f_{N_R}(P_R)\beta,\alpha\right\rangle=\alpha^*f_{N_R}(P_R)\beta=\frac{1}{N_R}\sum_{g,h}\overline{\alpha_g}\beta_hf(\lambda_{gh^{-1}})=\frac{1}{N_R}f(U_\alpha U_\beta^*).

Applying (3.9) to UαU_\alpha and Uβ∗U_\beta^* we obtain

∥fNR(PR)∥=sup⁡∥α∥=∥β∥=1∣⟨fNR(PR)β,α⟩∣=sup⁡∥α∥=∥β∥=11NR∣f(UαUβ∗)∣≤1NRsup⁡∥α∥=∥β∥=1∥f∥⋅∥Uα∥⋅∥Uβ∥≤∥f∥⋅(R+1)2NR.(11)\begin{aligned} \left\|f_{N_R}(P_R)\right\|&=\sup_{\|\alpha\|=\|\beta\|=1}\left|\left\langle f_{N_R}(P_R)\beta,\alpha\right\rangle\right| =\sup_{\|\alpha\|=\|\beta\|=1}\frac{1}{N_R}\left|f(U_\alpha U_\beta^*)\right|\\ &\leq\frac{1}{N_R}\sup_{\|\alpha\|=\|\beta\|=1}\|f\|\cdot\|U_\alpha\|\cdot\|U_\beta\| \leq\|f\|\cdot\frac{(R+1)^2}{N_R}. \tag*{(11)} \end{aligned}

Now choose a basis y1,…,ysy_1,\ldots,y_s of Ran⁡Φ\operatorname{Ran}\Phi. We may then write Φ(a)=∑j=1sfj(a)yj\Phi(a)=\sum_{j=1}^s f_j(a)y_j for some fj∈A∗f_j\in A^*. Then (3.10) yields

∥ΦNR(PR)∥≤∑j=1s∥(fj)NR(PR)∥∥yj∥≤(∑j=1s∥fj∥∥yj∥)(R+1)24⋅3R−1⟶0.\left\|\Phi_{N_R}(P_R)\right\|\leq\sum_{j=1}^s\left\|(f_j)_{N_R}(P_R)\right\|\|y_j\| \leq\left(\sum_{j=1}^s\|f_j\|\|y_j\|\right)\frac{(R+1)^2}{4\cdot3^{R-1}}\longrightarrow0.

Finally, set XR=PR/3X_R=P_R/\sqrt{3}. Then Lℓ,NR(XR)≤1L_{\ell,N_R}(X_R)\leq1, and ι\iota being a complete isometry implies

∥ιNR(XR)−ΦNR(XR)∥≥13(1−∥ΦNR(PR)∥).\left\|\iota_{N_R}(X_R)-\Phi_{N_R}(X_R)\right\|\geq\frac{1}{\sqrt{3}}\left(1-\left\|\Phi_{N_R}(P_R)\right\|\right).

Letting R→∞R\to\infty proves (3.6). In particular, the finite rank UCP maps needed to establish an MCQMS cannot give an approximation with error constant smaller than 1/31/\sqrt{3}. □

Cartesian products

Our next aim is to show how one may use the theory of matrix compact quantum metric spaces to expand on the class of groups with length functions that yield compact quantum metric spaces. As mentioned already, the only structural results in this direction cover the classes of hyperbolic groups [14] and of groups of polynomial growth [5]. We first develop some general machinery that can be used to verify that a certain length function on a cartesian product of two groups yields a CQMS, and then proceed to verify that the relevant assumptions are met in a number of instances.

A tensor product criterion

We first recall some standard properties of the minimal tensor product. Let AA, BB, and CC be unital C∗C^*-algebras. For every w∈C⊗min⁡Bw \in C \otimes_{\min} B, we have

∥w∥=sup⁡{∥(id⁡C⊗ψ)(w)∥∣n≥1, ψ∈UCP⁡(B,Mn(C))}(12)\lVert w\rVert= \sup\left\{\lVert(\operatorname{id}_{C} \otimes\psi)(w)\rVert\mid n \ge1,\ \psi\in\operatorname{UCP}(B,M_{n}(\mathbb{C}))\right\} \tag*{(12)}

which may be seen by representing BB faithfully on a Hilbert space and considering the UCP maps arising from compressions to finite-dimensional subspaces.

By injectivity of the minimal tensor product, a unital complete isometry j ⁣:A→Cj\colon A \to C induces a complete isometry j⊗id⁡B ⁣:A⊗min⁡B→C⊗min⁡Bj \otimes\operatorname{id}_{B}\colon A \otimes_{\min} B \to C \otimes_{\min} B, see [12 Theorem 4.6]. Finally, every bounded linear functional f∈C∗f \in C^* satisfies ∥f∥cb=∥f∥\lVert f\rVert_{\mathrm{cb}} = \lVert f\rVert by [15 Proposition 3.8]. The completely bounded mapping property of the minimal tensor product (see e.g. [16]) therefore gives

∥(f⊗id⁡B)(w)∥≤∥f∥∥w∥,w∈C⊗min⁡B.\lVert(f \otimes\operatorname{id}_{B})(w)\rVert\le\lVert f\rVert\lVert w\rVert,\qquad w \in C \otimes_{\min} B.

Proposition 4.1. Let (X,LX)(\mathcal{X},L_{\mathcal{X}}) and (Y,LY)(\mathcal{Y},L_{\mathcal{Y}}) be CQMSs and let MM be a Lipschitz seminorm on X⊙Y\mathcal{X} \odot\mathcal{Y}. Suppose that there exist cX,cY>0c_{\mathcal{X}},c_{\mathcal{Y}}>0, such that

LX((id⁡⊗g)z)≤cX∥g∥M(z)(g∈Y∗),L_{\mathcal{X}}((\operatorname{id}\otimes g)z) \le c_{\mathcal{X}}\lVert g\rVert M(z)\qquad(g \in\mathcal{Y}^*),
LY((f⊗id⁡)z)≤cY∥f∥M(z)(f∈X∗).L_{\mathcal{Y}}((f\otimes\operatorname{id})z) \le c_{\mathcal{Y}}\lVert f\rVert M(z)\qquad(f \in\mathcal{X}^*).

Suppose, moreover, that for every ε>0\varepsilon>0 there exist a unital C∗C^*-algebra CεC_{\varepsilon}, a unital complete isometry ιε ⁣:X→Cε\iota_{\varepsilon}\colon\mathcal{X}\to C_{\varepsilon}, and a bounded linear map Φε ⁣:X→Cε\Phi_{\varepsilon}\colon\mathcal{X}\to C_{\varepsilon} of finite rank satisfying

∥((ιε−Φε)⊗id⁡)z∥≤εM(z),z∈X⊙Y.(13)\lVert((\iota_{\varepsilon}-\Phi_{\varepsilon})\otimes\operatorname{id})z\rVert\le\varepsilon M(z),\qquad z\in\mathcal{X}\odot\mathcal{Y}. \tag*{(13)}

Then (X⊙Y,M)(\mathcal{X}\odot\mathcal{Y},M) is a CQMS.

Here we consider X⊙Y\mathcal{X}\odot\mathcal{Y} with the operator system structure induced by the embedding into the minimal tensor product of their enveloping C∗C^*-algebras.

Proof. Choose states μ∈S(X)\mu\in S(\mathcal{X}) and ν∈S(Y)\nu\in S(\mathcal{Y}). By Proposition 2.2 it suffices to show that

EM,μ⊗ν={z∈X⊙Y∣M(z)≤1, (μ⊗ν)(z)=0}E_{M,\mu\otimes\nu}=\{z\in\mathcal{X}\odot\mathcal{Y}\mid M(z)\le1,\ (\mu\otimes\nu)(z)=0\}

is totally bounded in norm. Fix ε>0\varepsilon>0 and the maps ιε,Φε ⁣:X→Cε\iota_{\varepsilon},\Phi_{\varepsilon}\colon\mathcal{X}\to C_{\varepsilon} ensured by the assumptions. Choose a basis e1,…,ere_{1},\ldots,e_{r} of Ran⁡Φε\operatorname{Ran}\Phi_{\varepsilon}. We may then write Φε(x)=∑i=1rfi(x)ei\Phi_{\varepsilon}(x)=\sum_{i=1}^{r}f_{i}(x)e_{i} with fi∈X∗f_{i}\in\mathcal{X}^*. Defining bi(z)=(fi⊗id⁡)zb_{i}(z)=(f_{i}\otimes\operatorname{id})z, we then have

(Φε⊗id⁡)z=∑iei⊗bi(z),andLY(bi(z))≤cY∥fi∥,(\Phi_{\varepsilon}\otimes\operatorname{id})z=\sum_{i}e_{i}\otimes b_{i}(z),\qquad\text{and}\qquad L_{\mathcal{Y}}(b_{i}(z))\le c_{\mathcal{Y}}\lVert f_{i}\rVert,

for z∈EM,μ⊗νz\in E_{M,\mu\otimes\nu}. For z∈EM,μ⊗νz\in E_{M,\mu\otimes\nu}, we now define az:=(id⁡⊗ν)za_{z}:=(\operatorname{id}\otimes\nu)z. Then LX(az)≤cXL_{\mathcal{X}}(a_{z})\le c_{\mathcal{X}} and μ(az)=0\mu(a_{z})=0. Setting

ELX,μ:={x∈X∣LX(x)≤1, μ(x)=0}E_{L_{\mathcal{X}},\mu}:=\{x\in\mathcal{X}\mid L_{\mathcal{X}}(x)\le1,\ \mu(x)=0\}

and RX:=sup⁡x∈ELX,μ∥x∥R_{\mathcal{X}}:=\sup_{x\in E_{L_{\mathcal{X}},\mu}}\lVert x\rVert, we have that az/cX∈ELX,μa_{z}/c_{\mathcal{X}}\in E_{L_{\mathcal{X}},\mu} and

∣ν(bi(z))∣=∣fi(az)∣≤cXRX∥fi∥.|\nu(b_{i}(z))|=|f_{i}(a_{z})|\le c_{\mathcal{X}}R_{\mathcal{X}}\lVert f_{i}\rVert.

Since

{bi(z)∣z∈EM,μ⊗ν}⊆cY∥fi∥ELY,ν+{λ1Y∣∣λ∣≤cXRX∥fi∥},\{b_{i}(z)\mid z\in E_{M,\mu\otimes\nu}\}\subseteq c_{\mathcal{Y}}\lVert f_{i}\rVert E_{L_{\mathcal{Y}},\nu}+\{\lambda1_{\mathcal{Y}}\mid|\lambda|\le c_{\mathcal{X}}R_{\mathcal{X}}\lVert f_{i}\rVert\},

we deduce that the former set is totally bounded for i=1,…,ri=1,\ldots,r. Hence, (Φε⊗id⁡)(EM,μ⊗ν)(\Phi_{\varepsilon}\otimes\operatorname{id})(E_{M,\mu\otimes\nu}) is totally bounded as well.

We may therefore choose z1,…,zN∈EM,μ⊗νz_{1},\ldots,z_{N}\in E_{M,\mu\otimes\nu} such that (Φε⊗id⁡)z1,…,(Φε⊗id⁡)zN(\Phi_{\varepsilon}\otimes\operatorname{id})z_{1},\ldots,(\Phi_{\varepsilon}\otimes\operatorname{id})z_{N} form an ε\varepsilon-net in (Φε⊗id⁡)(EM,μ⊗ν)(\Phi_{\varepsilon}\otimes\operatorname{id})(E_{M,\mu\otimes\nu}). Since ιε⊗id⁡\iota_{\varepsilon}\otimes\operatorname{id} is isometric, for each z∈EM,μ⊗νz\in E_{M,\mu\otimes\nu} and a suitable i∈{1,…,N}i\in\{1,\ldots,N\} we have

∥z−zi∥=∥(ιε⊗id⁡)z−(ιε⊗id⁡)zi∥≤∥(ιε⊗id⁡)z−(Φε⊗id⁡)z∥+∥(Φε⊗id⁡)z−(Φε⊗id⁡)zi∥+∥(Φε⊗id⁡)zi−(ιε⊗id⁡)zi∥<3ε\begin{aligned} \lVert z-z_{i}\rVert=\lVert(\iota_{\varepsilon}\otimes\operatorname{id})z-(\iota_{\varepsilon}\otimes\operatorname{id})z_{i}\rVert\\ &\le\lVert(\iota_{\varepsilon}\otimes\operatorname{id})z-(\Phi_{\varepsilon}\otimes\operatorname{id})z\rVert+\lVert(\Phi_{\varepsilon}\otimes\operatorname{id})z-(\Phi_{\varepsilon}\otimes\operatorname{id})z_{i}\rVert+\lVert(\Phi_{\varepsilon}\otimes\operatorname{id})z_{i}-(\iota_{\varepsilon}\otimes\operatorname{id})z_{i}\rVert\\ &<3\varepsilon \end{aligned}

by using (4.2). It follows that EM,μ⊗νE_{M,\mu\otimes\nu} is totally bounded.

We now specialise to spectral triples, which are the main case of interest in the present paper. For the statement, recall that a spectral triple (A,HA,DA)(\mathcal{A},H_{\mathcal{A}},D_{\mathcal{A}}) is called a (matrix) spectral metric space if the commutator seminorm yields a (M)CQMS; see § 2.2 for details

Theorem 4.2. Suppose (A,HA,DA)(\mathcal{A}, H_A, D_A) is a matrix spectral metric space and (B,HB,DB)(\mathcal{B}, H_B, D_B) is a spectral metric space. For z=∑kak⊗bk∈A⊙Bz = \sum_k a_k \otimes b_k \in\mathcal{A} \odot\mathcal{B}, set

P(z)=∑k[DA,ak]⊗bkandQ(z)=∑kak⊗[DB,bk].P(z) = \sum_k [D_A,a_k]\otimes b_k \qquad\text{and}\qquad Q(z) = \sum_k a_k \otimes[D_B,b_k].

If MM is a Lipschitz seminorm on A⊙B\mathcal{A}\odot\mathcal{B} for which there exist cA,cB>0c_A,c_B>0 such that

∥P(z)∥≤cAM(z)and∥Q(z)∥≤cBM(z),(14)\lVert P(z)\rVert\le c_A M(z) \qquad\text{and}\qquad\lVert Q(z)\rVert\le c_B M(z), \tag*{(14)}

for all z∈A⊙Bz\in\mathcal{A}\odot\mathcal{B}, then (A⊙B,M)(\mathcal{A}\odot\mathcal{B},M) is a CQMS. The analogous assertion holds with the two factors interchanged.

Proof. We verify the hypotheses of Proposition 4.1. Denote by A\mathcal{A} and B\mathcal{B} the respective C∗\mathrm{C}^*-completions, and put LDA,n(x)=∥[1n⊗DA,x]∥L_{D_A,n}(x)=\lVert[1_n\otimes D_A,x]\rVert. Let δ>0\delta>0 be given, and use that (A,HA,DA)(\mathcal{A},H_A,D_A) is a matrix spectral metric space to obtain a C∗\mathrm{C}^*-algebra C\mathcal{C} along with maps ι,Φ:A→C\iota,\Phi:\mathcal{A}\to\mathcal{C} such that

∥ιn(x)−Φn(x)∥≤δ⋅LDA,n(x)(15)\lVert\iota_n(x)-\Phi_n(x)\rVert\le\delta\cdot L_{D_A,n}(x) \tag*{(15)}

for all n≥1n\ge1 and all x∈Mn(A)x\in M_n(\mathcal{A}). Here ι\iota is a complete isometry and Φ\Phi is UCP with finite-dimensional range; in particular both maps extend to A\mathcal{A}.

Fix now an n∈Nn\in\mathbb{N}, a UCP map ψ:B→Mn(C)\psi:\mathcal{B}\to M_n(\mathbb{C}) as well as a z∈A⊙Bz\in\mathcal{A}\odot\mathcal{B}, and set xψ=(id⁡⊗ψ)zx_\psi=(\operatorname{id}\otimes\psi)z. Under the canonical matrix identification A⊙Mn(C)≅Mn(A)≅Mn(C)⊙A\mathcal{A}\odot M_n(\mathbb{C})\cong M_n(\mathcal{A})\cong M_n(\mathbb{C})\odot\mathcal{A} we then have

[1n⊗DA,xψ]=(id⁡B(HA)⊗ψ)P(z).(16)[1_n\otimes D_A,x_\psi]=(\operatorname{id}_{B(H_A)}\otimes\psi)P(z). \tag*{(16)}

Setting W=((ι−Φ)⊗id⁡B)(z)∈C⊗min⁡BW=((\iota-\Phi)\otimes\operatorname{id}_{\mathcal{B}})(z)\in\mathcal{C}\otimes_{\min}\mathcal{B}, we now obtain

∥(id⁡C⊗ψ)(W)∥=∥ιn(xψ)−Φn(xψ)∥≤δ⋅LDA,n(xψ)≤δ⋅∥P(z)∥,\lVert(\operatorname{id}_{\mathcal{C}}\otimes\psi)(W)\rVert =\lVert\iota_n(x_\psi)-\Phi_n(x_\psi)\rVert \le\delta\cdot L_{D_A,n}(x_\psi) \le\delta\cdot\lVert P(z)\rVert,

where the first inequality follows from (15) and the last inequality from (16). Since this holds for all n∈Nn\in\mathbb{N} and all ψ∈UCP⁡(B,Mn(C))\psi\in\operatorname{UCP}(\mathcal{B},M_n(\mathbb{C})), (12) and (14) imply

∥((ι−Φ)⊗id⁡)z∥≤δ⋅∥P(z)∥≤δcA⋅M(z).\lVert((\iota-\Phi)\otimes\operatorname{id})z\rVert \le\delta\cdot\lVert P(z)\rVert \le\delta c_A\cdot M(z).

Given ε>0\varepsilon>0, choosing δ=ε/cA\delta=\varepsilon/c_A we see that (13) is satisfied. Moreover,

LDA((id⁡⊗g)z)≤∥g∥∥P(z)∥andLDB((f⊗id⁡)z)≤∥f∥∥Q(z)∥,L_{D_A}((\operatorname{id}\otimes g)z)\le\lVert g\rVert\lVert P(z)\rVert \qquad\text{and}\qquad L_{D_B}((f\otimes\operatorname{id})z)\le\lVert f\rVert\lVert Q(z)\rVert,

and so Proposition 4.1 applies and shows that (A⊙B,M)(\mathcal{A}\odot\mathcal{B},M) is a CQMS. It is clear that the argument is symmetric in the choice of matrix factor. □

Cancellation in one factor

We now return to the setting of groups and length functions to provide a criterion ensuring that the assumptions in Theorem 4.2 are satisfied. Suppose therefore GG and KK are countable discrete groups equipped with proper length functions ℓG\ell_G and ℓK\ell_K, and consider the length function ℓG×K:=ℓG+ℓK\ell_{G\times K}:=\ell_G+\ell_K on G×KG\times K. Suppressing the natural isomorphism ℓ2(G×K)≅ℓ2(G)⊗ℓ2(K)\ell^2(G\times K)\cong\ell^2(G)\otimes\ell^2(K), we then have

DℓG×K=DℓG⊗1+1⊗DℓK,D_{\ell_{G\times K}}=D_{\ell_G}\otimes1+1\otimes D_{\ell_K},

as unbounded operators on, say, cc(G)⊙cc(K)c_c(G)\odot c_c(K). For notational convenience, we set

D1:=DℓG⊗1,D2:=1⊗DℓK,andD:=DℓG×K.D_1:=D_{\ell_G}\otimes1,\qquad D_2:=1\otimes D_{\ell_K},\qquad\text{and}\qquad D:=D_{\ell_{G\times K}}.

For x∈Mp(C[G×K])x\in M_p(\mathbb{C}[G\times K]), we furthermore set

T1(x):=[1p⊗D1,x],T2(x):=[1p⊗D2,x]andT(x):=T1(x)+T2(x),T^1(x):=[1_p\otimes D_1,x],\qquad T^2(x):=[1_p\otimes D_2,x]\qquad\text{and}\qquad T(x):=T^1(x)+T^2(x),

which are all extendable to bounded operators. We make no notational difference between these densely defined operators and their bounded extensions. The strategy in the sequel will be to average certain sequences of unitary conjugates of T(x)T(x) which fix T1(x)T^1(x) and make T2(x)T^2(x) converge strongly to zero. This is formalised by means of the following definition.

Definition 4.3. A finite cancellation datum for (K,ℓK)(K,\ell_K) consists of a finite nonempty set II, numbers wi≥0w_i \ge0 with ∑iwi=1\sum_i w_i = 1, and sequences (gi(n))n⊆K(g_i(n))_n \subseteq K indexed by i∈Ii \in I, such that, for every fixed m,s∈Km,s \in K,

∑i∈Iwi(ℓK(msgi(n))−ℓK(sgi(n)))⟶n→∞0.(17)\sum_{i\in I} w_i\bigl(\ell_K(msg_i(n))-\ell_K(sg_i(n))\bigr) \longrightarrow_{n\to\infty} 0. \tag*{(17)}

It is eventually exact if this expression is zero for all sufficiently large nn, with the threshold allowed to depend on mm and ss.

Lemma 4.4. If KK admits a finite cancellation datum then for every p≥1p \ge1 and every x∈Mp(C[G×K])x \in M_p(\mathbb{C}[G \times K]), it holds that

∥T1(x)∥≤∥T(x)∥and∥T2(x)∥≤2∥T(x)∥.\lVert T^1(x)\rVert\le\lVert T(x)\rVert \qquad\text{and}\qquad \lVert T^2(x)\rVert\le2\lVert T(x)\rVert.

For the proof, we will need the right translations as well. For t∈Kt \in K, we therefore let ρtδs=δst\rho_t\delta_s=\delta_{st} be the right translation, which is a unitary operator. It commutes with every left translation λm\lambda_m, and just like for left translation, every ρt\rho_t preserves the domain of DℓKD_{\ell_K}.

Proof. Write x=∑m∈Fxm⊗λmx=\sum_{m\in F}x_m\otimes\lambda_m with xm∈Mp(C[G])x_m\in M_p(\mathbb{C}[G]) for some finite set F⊆KF\subseteq K. To simplify notation, we will denote 1p⊗1ℓ2(G)⊗ρt1_p\otimes1_{\ell^2(G)}\otimes\rho_t by VtV_t in the following. Since VtV_t operator commutes with xx and D1D_1, it will fix T1(x)T^1(x) when conjugating, that is Vt∗T1(x)Vt=T1(x)V_t^*T^1(x)V_t=T^1(x). For η∈Cp⊗ℓ2(G)\eta\in\mathbb{C}^p\otimes\ell^2(G) and s∈Ks\in K, we have

Vt∗T2(x)Vt(η⊗δs)=∑m∈F(ℓK(mst)−ℓK(st))xmη⊗δms.V_t^*T^2(x)V_t(\eta\otimes\delta_s) =\sum_{m\in F}\bigl(\ell_K(mst)-\ell_K(st)\bigr)x_m\eta\otimes\delta_{ms}.

Notably, the rightmost tensor-leg coordinate, msms, does not depend on tt. For each n∈Nn\in\mathbb{N}, we now define a contraction on B(Cp⊗ℓ2(G)⊗ℓ2(K))\mathcal{B}(\mathbb{C}^p\otimes\ell^2(G)\otimes\ell^2(K)) by

Ψn(S)=∑iwiVgi(n)∗SVgi(n).\Psi_n(S)=\sum_i w_iV_{g_i(n)}^*SV_{g_i(n)}.

For ξ=∑s∈Eξs⊗δs∈Cp⊗ℓ2(G)⊗ℓ2(K)\xi=\sum_{s\in E}\xi_s\otimes\delta_s\in\mathbb{C}^p\otimes\ell^2(G)\otimes\ell^2(K) with EE finite, we have

∥Ψn(T2(x))ξ∥≤∑s∈E,m∈F∥xm∥⋅∥ξs∥⋅∣∑iwi(ℓK(msgi(n))−ℓK(sgi(n)))∣⟶0.\left\lVert\Psi_n(T^2(x))\xi\right\rVert \le \sum_{s\in E,m\in F}\lVert x_m\rVert\cdot\lVert\xi_s\rVert\cdot \left|\sum_iw_i\bigl(\ell_K(msg_i(n))-\ell_K(sg_i(n))\bigr)\right| \longrightarrow0.

Such vectors are dense, and ∥Ψn(T2(x))∥≤∥T2(x)∥\lVert\Psi_n(T^2(x))\rVert\le\lVert T^2(x)\rVert, so we have strong convergence everywhere. Consequently, Ψn(T2(x))→nT1(x)\Psi_n(T^2(x))\to_n T^1(x) strongly. Contractivity then gives ∥T1(x)ξ∥≤∥T(x)∥∥ξ∥\lVert T^1(x)\xi\rVert\le\lVert T(x)\rVert\lVert\xi\rVert for every ξ\xi, which proves the first inequality in the lemma statement. The second then follows by writing T2(x)=T(x)−T1(x)T^2(x)=T(x)-T^1(x). □

Corollary 4.5. If (C[G],ℓ2(G),DℓG)(\mathbb{C}[G],\ell^2(G),D_{\ell_G}) is a spectral metric space and if (C[K],ℓ2(K),DℓK)(\mathbb{C}[K],\ell^2(K),D_{\ell_K}) is a matrix spectral metric space and if KK admits a finite cancellation datum, then (C[G×K],ℓ2(G×K),DℓG×K)(\mathbb{C}[G\times K],\ell^2(G\times K),D_{\ell_{G\times K}}) is a spectral metric space.

Proof. By Lemma 4.4, the assumptions in Theorem 4.2 are satisfied and the result follows. □

To illustrate cancellation, we include the following easy lemma.

Lemma 4.6. For any length function on a finite group KK, one obtains an exact cancellation datum by setting I=KI=K, wk=1∣K∣w_k=\frac{1}{|K|} and gk(n)=kg_k(n)=k for all n∈Nn\in\mathbb{N}.

Proof. For fixed m,sm,s the maps t↦mstt\mapsto mst and t↦stt\mapsto st both permute KK, and so the sums of their lengths are equal. □

For abelian and infinite dihedral groups, we construct geodesic rays along which the length defects eventually stabilize. Averaging over the two directions and a finite set of residues then gives cancellation. The construction is related to Rieffel’s horofunction methods [17]. For an equally weighted family it suffices to prove, for each fixed hh,

An(h):=∑i(ℓK(hgi(n))−ℓK(gi(n)))=0eventually.(18)A_n(h):=\sum_i\bigl(\ell_K(hg_i(n))-\ell_K(g_i(n))\bigr)=0\quad\text{eventually}. \tag*{(18)}

Indeed, the sum in (4.6) is (An(ms)−An(s))/∣I∣(A_n(ms)-A_n(s))/|I|.

We now proceed to show that certain natural groups admit a finite cancellation datum.

Finitely generated abelian groups

We first show that finitely generated abelian groups have a finite cancelation datum.

Proposition 4.7. Every finite symmetric word length on a finitely generated abelian group KK has an eventually exact cancellation datum. If KK is infinite, the sequences can be chosen as samples of geodesic rays.

Proof. Since Lemma 4.6 covers the case for finite KK, we assume KK is infinite in the sequel. We will write KK additively and denote its rank by dd and its torsion subgroup by Tor⁡(K)\operatorname{Tor}(K). Suppose SS is the given generating set, and consider the quotient map

π:K→K/Tor⁡(K)≅Zd⊆Rd\pi: K \to K/\operatorname{Tor}(K) \cong\mathbb{Z}^{d} \subseteq\mathbb{R}^{d}

for some d≥1d \ge1. The set P:=conv⁡(π(S))P := \operatorname{conv}(\pi(S)) is then a polytope of dimension dd, which is centrally symmetric and it contains 00 in its interior. Recall that a facet of PP is a face of dimension d−1d-1. Choose a facet FF together with a supporting linear functional σ:Rd→R\sigma: \mathbb{R}^{d} \to\mathbb{R}; that is σ(p)≤1\sigma(p) \le1 for x∈Px \in P and F=P∩{ξ∈Rd∣σ(ξ)=1}F = P \cap\{\xi\in\mathbb{R}^{d} \mid\sigma(\xi) = 1\}; see e.g. [Zie95] for background on polytopes. Since 00 is in the interior of PP, the maximum value for σ\sigma is positive. To ease notation we will also write σ(w)=σ(π(w))\sigma(w) = \sigma(\pi(w)) for w∈Kw \in K. Now define

V:={v∈S∣π(v)∈F},q:=∣V∣,z:=∑v∈VvandΛ:=⟨V⟩≤K.V := \{v \in S \mid\pi(v) \in F\}, \qquad q := |V|, \qquad z := \sum_{v \in V} v \qquad\text{and} \qquad\Lambda:= \langle V\rangle\le K.

Note that if w=s1+⋯+skw = s_1 + \cdots+ s_k with si∈Ss_i \in S, then σ(w)=∑iσ(si)≤k\sigma(w) = \sum_i \sigma(s_i) \le k. Taking the minimum over all representing words, we obtain ℓK(w)≥σ(w)\ell_K(w) \ge\sigma(w). Since zz is defined as the sum of qq letters from VV, nznz can be represented by nqnq such letters. Since σ(nz)=nq\sigma(nz) = nq, we deduce ℓK(nz)=nq\ell_K(nz) = nq for all n≥0n \ge0. As S=−SS = -S, we deduce ℓK(−w)=ℓK(w)\ell_K(-w) = \ell_K(w) for every w∈Kw \in K. If n<0n < 0, this means ℓK(nz)=ℓK((−n)z)=(−n)q\ell_K(nz) = \ell_K((-n)z) = (-n)q, and so

ℓK(nz)=∣n∣qfor all n∈Z.(19)\ell_K(nz) = |n|q \qquad\text{for all } n \in\mathbb{Z}. \tag*{(19)}

Let γ:N0→K\gamma: \mathbb{N}_0 \to K be the path obtained by successively adding the letters of this repeated listing of elements of VV, starting at 00; i.e. writing V={s1,…,sq}V = \{s_1, \ldots, s_q\} the path is defined as:

γ(0)=0,γ(1)=s1,γ(2)=s1+s2,…,γ(q)=z,γ(q+1)=z+s1,…\gamma(0) = 0,\qquad\gamma(1) = s_1,\qquad\gamma(2) = s_1 + s_2,\ldots,\qquad\gamma(q) = z,\qquad\gamma(q+1) = z + s_1,\ldots

If r<tr < t, the subword from γ(r)\gamma(r) to γ(t)\gamma(t) has t−rt-r letters, each of which takes the value 11 under σ\sigma. Hence

t−r=σ(γ(t)−γ(r))≤ℓK(γ(t)−γ(r))≤t−r.t-r = \sigma(\gamma(t)-\gamma(r)) \le\ell_K(\gamma(t)-\gamma(r)) \le t-r.

It follows that γ\gamma is geodesic, and γ(nq)=nz\gamma(nq) = nz.

We proceed to show that K/ΛK/\Lambda is finite. Note that F=conv⁡(π(V))F = \operatorname{conv}(\pi(V)) and that it has dimension d−1d-1, and hence π(V)\pi(V) contains dd affinely independent points u1,…,udu_1,\ldots,u_d. By affine independence, a relation ∑j=1dtjuj=0\sum_{j=1}^{d} t_j u_j = 0 with real coefficients satisfying ∑j=1dtj=0\sum_{j=1}^{d} t_j = 0 forces every tjt_j to vanish. In fact, these points are also linearly independent. We may see this by applying σ\sigma to any relation ∑j=1dtjuj=0\sum_{j=1}^{d} t_j u_j = 0 and noting that since σ(uj)=1\sigma(u_j) = 1, this gives ∑j=1dtj=0\sum_{j=1}^{d} t_j = 0. Affine independence then forces t1=⋯=td=0t_1 = \cdots= t_d = 0. Consequently, π(Λ)\pi(\Lambda) contains dd linearly independent integer vectors, w1,…,wdw_1,\ldots,w_d. The matrix AA whose columns consist of w1,…,wdw_1,\ldots,w_d is therefore invertible and its inverse is given by A−1=(det⁡(A))−1Adj⁡(A)A^{-1} = (\det(A))^{-1}\operatorname{Adj}(A), where the adjugate matrix Adj⁡(A)\operatorname{Adj}(A) has integer entries. We therefore have that

det⁡(A)⋅Zd=A⋅Adj⁡(A)⋅Zd⊆AZd⊆π(Λ),\det(A)\cdot\mathbb{Z}^{d} = A\cdot\operatorname{Adj}(A)\cdot\mathbb{Z}^{d} \subseteq A\mathbb{Z}^{d} \subseteq\pi(\Lambda),

and it follows that π(Λ)\pi(\Lambda) has finite index in Zd\mathbb{Z}^{d}. The surjection K/Λ→Zd/π(Λ)K/\Lambda\to\mathbb{Z}^{d}/\pi(\Lambda) has kernel (Λ+Tor⁡(K))/Λ(\Lambda+\operatorname{Tor}(K))/\Lambda which, by the second isomorphism theorem, is isomorphic to Tor⁡(K)/Λ∩Tor⁡(K)\operatorname{Tor}(K)/\Lambda\cap\operatorname{Tor}(K), and thus finite. Finiteness of K/ΛK/\Lambda now follows from noting that both kernel and range are finite.

For fixed w∈Kw \in K, consider the integer-valued sequence

bn(w)=ℓK(nz+w)−nq.b_n(w) = \ell_K(nz+w)-nq.

Then ℓK((n+1)z+w)≤ℓK(nz+w)+q\ell_K((n+1)z+w) \le\ell_K(nz+w)+q implies bn+1(w)≤bn(w)b_{n+1}(w) \le b_n(w), so the sequence is non-increasing. Using (19) we also have nq=ℓK(nz)≤ℓK(nz+w)+ℓK(w)nq = \ell_K(nz) \le\ell_K(nz+w)+\ell_K(w), and so we deduce −ℓK(w)≤bn(w)≤ℓK(w)-\ell_K(w) \le b_n(w) \le\ell_K(w), forcing the sequence to eventually stabilize at a value, which we denote by b(w)b(w). We will need to understand how bb changes under translations with elements from Λ\Lambda. For v∈Vv \in V, we have ℓK(v)=1\ell_K(v)=1. Moreover, as z−v=∑u∈V∖{v}uz-v=\sum_{u\in V\setminus\{v\}}u is a sum of q−1q-1 generators, we have ℓK(z−v)≤q−1\ell_K(z-v)\leq q-1. Writing (n+1)z+w=(nz+w+v)+(z−v)(n+1)z+w=(nz+w+v)+(z-v), we use the triangle inequality to calculate

ℓK(nz+w+v)≤ℓK(nz+w)+1,\ell_K(nz+w+v)\leq\ell_K(nz+w)+1,
ℓK((n+1)z+w)≤ℓK(nz+w+v)+q−1.\ell_K((n+1)z+w)\leq\ell_K(nz+w+v)+q-1.

Thus

bn+1(w)+1≤bn(w+v)≤bn(w)+1.b_{n+1}(w)+1\leq b_n(w+v)\leq b_n(w)+1.

As bnb_n eventually stabilizes to bb, we see that for sufficiently large nn, both the outer terms equal b(w)+1b(w)+1, and so b(w+v)=b(w)+1b(w+v)=b(w)+1. We also note that this yields b(w)−1=b(w−v+v)−1=b(w−v)b(w)-1=b(w-v+v)-1=b(w-v). Every a∈Λa\in\Lambda can be written as a=∑v∈Vkvva=\sum_{v\in V}k_vv for kv∈Zk_v\in\mathbb{Z}. By iterating these identities we get

b(w+a)=b(w)+∑v∈Vkv=b(w)+∑v∈Vkvσ(v)=b(w)+σ(∑v∈Vkvv)=b(w)+σ(a).b(w+a)=b(w)+\sum_{v\in V}k_v=b(w)+\sum_{v\in V}k_v\sigma(v)=b(w)+\sigma\left(\sum_{v\in V}k_vv\right)=b(w)+\sigma(a).

It follows that b(w+a)−σ(w+a)=b(w)−σ(w)b(w+a)-\sigma(w+a)=b(w)-\sigma(w), and so f([w]):=b(w)−σ(w)f([w]):=b(w)-\sigma(w) is a well-defined function on K/ΛK/\Lambda. Thus, for all sufficiently large nn (the threshold depending on ww) we have

ℓK(nz+w)=nq+σ(w)+f([w]).(20)\ell_K(nz+w)=nq+\sigma(w)+f([w]). \tag*{(20)}

With the analysis above at our disposal, we now obtain the claimed cancellation datum as follows: Choose as set of representatives, RR, for K/ΛK/\Lambda, put M=∣R∣M=|R|, and define 2M2M sequences as follows:

gr+(n):=γ(nq)+r=nz+r,for r∈R,g_r^+(n):=\gamma(nq)+r=nz+r,\qquad\text{for }r\in R,
gr−(n):=−γ(nq)−r=−nz−r,for r∈R.g_r^-(n):=-\gamma(nq)-r=-nz-r,\qquad\text{for }r\in R.

Fix h∈Kh\in K, and choose n0∈Nn_0\in\mathbb{N} such that (4.9) holds for all n≥n0n\geq n_0 and all ww in the finite set {r,r±h∣r∈R}\{r,r\pm h\mid r\in R\}. Then

∑r∈R(ℓK(h+nz+r)−ℓK(nz+r))=∑r∈R(σ(h)+f([h+r])−f([r]))=Mσ(h),\sum_{r\in R}\left(\ell_K(h+nz+r)-\ell_K(nz+r)\right) =\sum_{r\in R}\left(\sigma(h)+f([h+r])-f([r])\right)=M\sigma(h),
∑r∈R(ℓK(h−nz−r)−ℓK(−nz−r))=∑r∈R(−σ(h)+f([r−h])−f([r]))=−Mσ(h).\sum_{r\in R}\left(\ell_K(h-nz-r)-\ell_K(-nz-r)\right) =\sum_{r\in R}\left(-\sigma(h)+f([r-h])-f([r])\right)=-M\sigma(h).

For the second line, symmetry gives ℓK(h−nz−r)=ℓK(nz+r−h)\ell_K(h-nz-r)=\ell_K(nz+r-h) and ℓK(−nz−r)=ℓK(nz+r)\ell_K(-nz-r)=\ell_K(nz+r). In both lines the ff-terms sum to zero because [r]↦[r±h][r]\mapsto[r\pm h] permutes the finite quotient K/ΛK/\Lambda. The two sums therefore cancel, giving An(h)=0A_n(h)=0, where AnA_n is given as in (18). The weighted sum in (17) is then

An(m+s)−An(s)2M=0.\frac{A_n(m+s)-A_n(s)}{2M}=0.

Thus, the weights 1/(2M)1/(2M) give the required eventually exact datum. □

The infinite dihedral group

Next we show that also the infinite dihedral group D∞D_\infty admits a finite cancellation datum.

Proposition 4.8. Let KK be the infinite dihedral group

K=⟨a,b∣b2=e, bab=a−1⟩,K=\langle a,b\mid b^2=e,\ bab=a^{-1}\rangle,

and let ℓK\ell_K be the word length for any finite symmetric generating set SS. Then there is an eventually exact cancellation datum whose sequences are samples of geodesic rays.

Proof. Write elements of KK as ajbεa^jb^\varepsilon for j∈Zj\in\mathbb{Z} and ε∈{0,1}\varepsilon\in\{0,1\}, and note that

(ajbδ)(akbε)=aj+(−1)δkbδ+ε,(a^jb^\delta)(a^kb^\varepsilon)=a^{j+(-1)^\delta k}b^{\delta+\varepsilon},

where the exponent of bb is understood to be modulo 22. Now define

T={t∈Z:at∈S}andP=max⁡({∣t∣:t∈T}∪{0}),T=\{t\in\mathbb{Z}:a^t\in S\} \qquad\text{and}\qquad P=\max(\{|t|:t\in T\}\cup\{0\}),
Q={r∈Z:arb∈S}andΔ=max⁡Q−min⁡Q.Q=\{r\in\mathbb{Z}:a^rb\in S\} \qquad\text{and}\qquad \Delta=\max Q-\min Q.

Note that T=−TT=-T. Also, QQ is nonempty since KK is non-abelian, and ν:=max⁡{P,Δ/2}>0\nu:=\max\{P,\Delta/2\}>0 because otherwise SS would generate a group of order at most 2. Now choose positive integers q,Lq,L by

(q,L)={(P,1),if P≥Δ/2,(Δ,2),if P<Δ/2.(q,L)= \begin{cases} (P,1), & \text{if } P\geq\Delta/2,\\ (\Delta,2), & \text{if } P<\Delta/2. \end{cases}

Then q=νLq=\nu L. Note that if (q,L)=(P,1)(q,L)=(P,1), then aq=aP∈Sa^q=a^P\in S as T=−TT=-T. If (q,L)=(Δ,2)(q,L)=(\Delta,2), then (amax⁡Qb)(amin⁡Qb)=aΔ=aq(a^{\max Q}b)(a^{\min Q}b)=a^\Delta=a^q and (amax⁡Qb)(amin⁡Qb)(a^{\max Q}b)(a^{\min Q}b) has length 2. Thus, in either case, there is a word of length LL representing aqa^q. We first prove

ℓK(anq)=∣n∣L,for all n∈Z.\ell_K(a^{nq})=|n|L,\qquad\text{for all }n\in\mathbb{Z}.

To this end, put c:=(max⁡Q+min⁡Q)/2c:=(\max Q+\min Q)/2 and define X:K→RX:K\to\mathbb{R} by X(akbε):=k−cεX(a^kb^\varepsilon):=k-c\varepsilon. We now wish to analyse how much the value of XX changes upon multiplying the input by a generator. We first note that

X(akbεat)−X(akbε)=(−1)εt,X(a^kb^\varepsilon a^t)-X(a^kb^\varepsilon)=(-1)^\varepsilon t,
X(akbεarb)−X(akbε)=(−1)ε(r−c).X(a^kb^\varepsilon a^rb)-X(a^kb^\varepsilon)=(-1)^\varepsilon(r-c).

If at∈Sa^t\in S then ∣t∣≤P≤ν|t|\leq P\leq\nu and if arb∈Sa^rb\in S then ∣r−c∣≤Δ/2≤ν|r-c|\leq\Delta/2\leq\nu. So, upon multiplying a given element akbε∈Ka^kb^\varepsilon\in K by a generator from the right, the absolute change in the value of XX is at most ν\nu. Also, X(e)=0X(e)=0, so telescoping along a word of length kk representing hh gives ∣X(h)∣≤kν|X(h)|\leq k\nu. Thus ℓK(h)≥∣X(h)∣/ν\ell_K(h)\geq|X(h)|/\nu. If h=anqh=a^{nq} for some n∈Zn\in\mathbb{Z}, this reads ℓK(anq)≥∣n∣q/ν=∣n∣L\ell_K(a^{nq})\geq|n|q/\nu=|n|L, which proves the lower bound. But, as we saw above aqa^q can be represented by a word of length LL so anqa^{nq} can be expressed as a product of ∣n∣L|n|L generators, which provides the upper bound.

For σ∈{±1}\sigma\in\{\pm1\}, 0≤r<q0\leq r<q and ε∈{0,1}\varepsilon\in\{0,1\}, set

gσ,r,ε(n):=aσnq+rbε.g_{\sigma,r,\varepsilon}(n):=a^{\sigma nq+r}b^\varepsilon.

We note that

gσ,r,ε(n)−1gσ,r,ε(k)=a(−1)εσ(k−n)q.g_{\sigma,r,\varepsilon}(n)^{-1}g_{\sigma,r,\varepsilon}(k)=a^{(-1)^\varepsilon\sigma(k-n)q}.

The length of this element is ∣n−k∣L|n-k|L by what we noted above. In other words, the distance between gσ,r,ε(n)g_{\sigma,r,\varepsilon}(n) and gσ,r,ε(k)g_{\sigma,r,\varepsilon}(k) in the induced left-invariant word metric is ∣n−k∣L|n-k|L. Repeating a shortest word for a(−1)εσqa^{(-1)^\varepsilon\sigma q} from the initial point arbεa^rb^\varepsilon interpolates these samples to a geodesic ray. Indeed, the interpolating path has length ∣k−n∣L|k-n|L between any two samples. A shorter path between two intermediate vertices, together with the remaining portions of the interpolating path, would contradict the distance formula for two enclosing samples. For fixed (σ,r,ε)(\sigma,r,\varepsilon), the integers

fn=ℓK(aσnq+rbε)−nLf_n=\ell_K(a^{\sigma nq+r}b^\varepsilon)-nL

are non-increasing, since left multiplication by aσqa^{\sigma q} increases the length by at most LL. Using subadditivity and the reverse triangle inequality we obtain

nL−ℓK(arbε)≤ℓK(aσnq+rbε)≤nL+ℓK(arbε).nL-\ell_K(a^rb^\varepsilon)\leq\ell_K(a^{\sigma nq+r}b^\varepsilon)\leq nL+\ell_K(a^rb^\varepsilon).

It follows that fnf_n is bounded and integer-valued, and must therefore stabilize. We may thus find an N0N_0 and constants Cσ,r,ε∈ZC_{\sigma,r,\varepsilon}\in\mathbb{Z} such that

ℓK(gσ,r,ε(n))=nL+Cσ,r,εfor all n≥N0,(21)\ell_K(g_{\sigma,r,\varepsilon}(n))=nL+C_{\sigma,r,\varepsilon}\qquad\text{for all }n\geq N_0, \tag*{(21)}

for all 4q4q indices (σ,r,ε)(\sigma,r,\varepsilon).

With this, we may now finally verify cancellation. Fix h=ajbδh=a^jb^\delta. For every r∈{0,…,q−1}r\in\{0,\ldots,q-1\}, there exist unique tr∈Zt_r\in\mathbb{Z} and r′∈{0,…,q−1}r'\in\{0,\ldots,q-1\} such that

j+(−1)δr=qtr+r′.j+(-1)^\delta r=q t_r+r'.

We now define σ′:=(−1)δσ\sigma':=(-1)^\delta\sigma and ε′:=δ+ε(mod2)\varepsilon':=\delta+\varepsilon\pmod2. Then

hgσ,r,ε(n)=gσ′,r′,ε′(n+σ′tr).hg_{\sigma,r,\varepsilon}(n)=g_{\sigma',r',\varepsilon'}(n+\sigma't_r).

Note that trt_r is independent of σ\sigma and ϵ\epsilon. For n≥N0+max⁡r∣tr∣n \geq N_0+\max_r |t_r|, both nn and n+σ′trn+\sigma't_r are at least N0N_0, and so (4.10) yields

ℓK(hgσ,r,ϵ(n))−ℓK(gσ,r,ϵ(n))=σ′trL+Cσ′,r′,ϵ′−Cσ,r,ϵ.\ell_K(hg_{\sigma,r,\epsilon}(n))-\ell_K(g_{\sigma,r,\epsilon}(n))=\sigma't_rL+C_{\sigma',r',\epsilon'}-C_{\sigma,r,\epsilon}.

The map of indices (σ,r,ϵ)(\sigma,r,\epsilon) to (σ′,r′,ϵ′)(\sigma',r',\epsilon') is a permutation: from its image one recovers σ=(−1)δσ′\sigma=(-1)^\delta\sigma', ϵ=ϵ′−δ\epsilon=\epsilon'-\delta modulo two, and the unique r∈{0,…,q−1}r\in\{0,\ldots,q-1\} congruent to (−1)δ(r′−j)(-1)^\delta(r'-j) modulo qq. Consequently the sums of the two families of constants agree. For fixed r,ϵr,\epsilon, summing over σ∈{1,−1}\sigma\in\{1,-1\} cancels the carry terms because trt_r is fixed and σ′\sigma' takes both signs. Thus

∑σ,r,ϵ(ℓK(hgσ,r,ϵ(n))−ℓK(gσ,r,ϵ(n)))=0(22)\sum_{\sigma,r,\epsilon}\left(\ell_K(hg_{\sigma,r,\epsilon}(n))-\ell_K(g_{\sigma,r,\epsilon}(n))\right)=0 \tag*{(22)}

for every sufficiently large nn. This is An(h)=0A_n(h)=0, so the weights 1/(4q)1/(4q) give (4.6).

Products with hyperbolic groups

In this section we show how the analysis carried out above leads to new examples of length functions on groups which give rise to compact quantum metric spaces. We first summarize the upshot of the previous section as follows:

Proposition 4.9. Let GG be any group with a proper length function ℓG\ell_G. Let KK be either

  • a finite group with any length function, or

  • a finitely generated abelian group with any finite symmetric word length, or

  • the infinite dihedral group with any finite symmetric word length.

Then the bounds of Lemma 4.4 hold at every matrix level, with constants 1 and 2 independent of the group, generating set, and matrix size.

Proof. This is immediate by Lemma 4.4, Lemma 4.6, Proposition 4.7 and Proposition 4.8. □

Theorem 4.10. Let GG be a finitely generated group equipped with a word-length function ℓG\ell_G for which (C[G],LℓG)(\mathbb{C}[G],L_{\ell_G}) is a CQMS, and let K1,…,KmK_1,\ldots,K_m be a family or groups, each of which is either finite, finitely generated abelian or isomorphic to D∞D_\infty. Equipping each KjK_j with a word length function ℓKj\ell_{K_j} and

P:=G×K1×⋯×Km with ℓ(g,k1,…,km)=ℓG(g)+∑j=1mℓKj(kj),P:=G\times K_1\times\cdots\times K_m\ \text{with}\ \ell(g,k_1,\ldots,k_m)=\ell_G(g)+\sum_{j=1}^{m}\ell_{K_j}(k_j),

the pair (C[P],Lℓ)(\mathbb{C}[P],L_\ell) is a CQMS.

Proof. Each KjK_j is amenable and yields a CQMS by [5]. Theorem 3.1 therefore makes every KjK_j an MCQMS. For m=1m=1, the result therefore follows by Corollary 4.5. The result now follows by induction on mm. □

Corollary 4.11. If HH is word-hyperbolic and K1,…,KmK_1,\ldots,K_m are as in Theorem 4.10, then H×K1×⋯×KmH\times K_1\times\cdots\times K_m, with length function coming from the union of the chosen generating sets for the factors, yields a CQMS.

Remark 4.12. Corollary 4.11 covers groups of the form G=H×F×D∞k×ZdG=H\times F\times D_\infty^k\times\mathbb{Z}^d for HH word hyperbolic and FF finite, and with the length function as described in the result. In particular it covers the case d=0d=0, FF a finite Coxeter group, and HH a non-elementary hyperbolic Coxeter group, and k≥1k\geq1. As such, we have the first examples of word length functions on non-amenable, non-hyperbolic Coxeter groups with CQMS structures for these word-lengths. Indeed, these groups contain HH and a copy of Z2\mathbb{Z}^2, and are therefore neither amenable nor word-hyperbolic. They are thus in particular not covered by the results of [5] or [14].

References

Are Austad, Department of Mathematics, University of Oslo, P.O.Box 1053 Blindern, 0316 Oslo, Norway

Email address: [email protected]

David Kyed, Department of Mathematics and Computer Science, University of Southern Denmark, Campusvej 55, DK-5230 Odense M, Denmark

Email address: [email protected]

References

  1. [1][AKK22] Konrad Aguilar, Jens Kaad, and David Kyed. The Podleś spheres converge to the sphere. Comm. Math. Phys., 392(3):1029–1061, 2022.DOI
  2. [2][Aus26] Are Austad. Compact quantum metric spaces from weakly geodesic length functions on hyperbolic groups, 2026.arxiv.org/abs/2608.28303
  3. [3][Con89] Alain Connes. Compact metric spaces, Fredholm modules, and hyperfiniteness. Ergodic Theory Dynam. Syst., 9(2):207–220, 1989.
  4. [4][Con94] Alain Connes. Noncommutative Geometry. Academic Press, San Diego, 1994.
  5. [5]Michael Christ and Marc A. Rieffel. Nilpotent group C*-algebras as compact quantum metric spaces. Canad. Math. Bull., 60(1):77–94, 2017.
  6. [6]Alain Connes and Walter D. van Suijlekom. Spectral truncations in noncommutative geometry and operator systems. Commun. Math. Phys., 383(3):2021–2067, 2021.DOI
  7. [7]Uffe Haagerup. An example of a nonnuclear C*-algebra, which has the metric approximation property. Invent. Math., 50(3):279–293, 1978/79.
  8. [8]Jens Kaad. External products of spectral metric spaces. Pure Appl. Funct. Anal., 9(5):1175–1211, 2024.arxiv.org/abs/2304.03979
  9. [9]Jens Kaad and David Kyed. Dynamics of compact quantum metric spaces. Ergodic Theory and Dynamical Systems, 41(7):2069–2109, 2021.DOI
  10. [10]Mario Klisse. Word-length spectral triples of (Z/2Z) ≀ F_d are not metric, 2026.arxiv.org/abs/2608.12080
  11. [11]David Kyed and Ryszard Nest. Finiteness of metrics on state spaces. Bull. Lond. Math. Soc., 56(1):288–295, 2024.DOI
  12. [12]Ali Kavruk, Vern I. Paulsen, Ivan G. Todorov, and Mark Tomforde. Tensor products of operator systems. J. Funct. Anal., 261(2):267–299, 2011.arxiv.org/abs/0910.2210
  13. [13]Malte Leimbach. Convergence of Peter–Weyl truncations of compact quantum groups. Journal of Noncommutative Geometry, 2025. Published online first.arxiv.org/abs/2409.16698
  14. [14]Narutaka Ozawa and Marc A. Rieffel. Hyperbolic group C*-algebras and free-product C*-algebras as compact quantum metric spaces. Canad. J. Math., 57(5):1056–1079, 2005.
  15. [15]Vern Paulsen. Completely bounded maps and operator algebras, volume 78 of Cambridge Studies in Advanced Mathematics. Cambridge University Press, Cambridge, 2002.DOI
  16. [16]Gilles Pisier. Introduction to Operator Space Theory, volume 294 of London Mathematical Society Lecture Note Series. Cambridge University Press, 2003.DOI
  17. [17]Marc A Rieffel. Group C*-algebras as compact quantum metric spaces. Documenta Mathematica, 7:605–651, 2002.DOI
  18. [18]Marc A. Rieffel. Standard deviation is a strongly Leibniz seminorm. New York J. Math., 20:35–56, 2014.
  19. [19]Walter D. van Suijlekom. Gromov-Hausdorff convergence of state spaces for spectral truncations. J. Geom. Phys., 162:11, 2021. Id/No 104075.DOI

Paper details

vlnc:W_dwvs3ebw
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
  • Mathematics

ChatGPT

View statement
Looking for related papers…
Contents