This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

Homotopy type of the unitary group of the uniform Roe algebra on n\mathbb{Z}^{n}

Tsuyoshi Kato Department of Mathematics, Kyoto University, Kyoto, 606-8502, Japan [email protected] Daisuke Kishimoto Department of Mathematics, Kyoto University, Kyoto, 606-8502, Japan [email protected]  and  Mitsunobu Tsutaya Faculty of Mathematics, Kyushu University, Fukuoka 819-0395, Japan [email protected]
Abstract.

We study the homotopy type of the space of the unitary group U1(Cu(|n|))\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)) of the uniform Roe algebra Cu(|n|)C^{\ast}_{u}(|\mathbb{Z}^{n}|) of n\mathbb{Z}^{n}. We show that the stabilizing map U1(Cu(|n|))U(Cu(|n|))\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\to\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)) is a homotopy equivalence. Moreover, when n=1,2n=1,2, we determine the homotopy type of U1(Cu(|n|))\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)), which is the product of the unitary group U1(C(|n|))\operatorname{U}_{1}(C^{\ast}(|\mathbb{Z}^{n}|)) (having the homotopy type of U()\operatorname{U}_{\infty}(\mathbb{C}) or ×BU()\mathbb{Z}\times B\operatorname{U}_{\infty}(\mathbb{C}) depending on the parity of nn) of the Roe algebra C(|n|)C^{\ast}(|\mathbb{Z}^{n}|) and rational Eilenberg–MacLane spaces.

Key words and phrases:
uniform Roe algebra, Roe algebra, unitary group, homotopy type, operator KK-theory
2010 Mathematics Subject Classification:
55Q52 (Primary), 46L80 (Secondary)
Kato was supported by JSPS KAKENHI 17K18725 and 17H06461. Kishimoto was supported by JSPS KAKENHI 17K05248 and 19K03473. Tsutaya was supported by JSPS KAKENHI 19K14535

1. Introduction

For a CC^{\ast}-algebra AA, let GLd(A)\operatorname{GL}_{d}(A) and Ud(A)\operatorname{U}_{d}(A) denote the space of the invertible and unitary matrices with entries in AA, respectively. It it well-known that they always have the same homotopy type. We will often refer only to Ud(A)\operatorname{U}_{d}(A) but most statements are valid for GLd(A)\operatorname{GL}_{d}(A) as well. There have been a lot of works on the homotopy theory of Ud(A)\operatorname{U}_{d}(A) and some of them have important applications. For finite-dimensional case, the complex-valued unitary matrices Ud()\operatorname{U}_{d}(\mathbb{C}) is just the usual unitary group acting linearly on d\mathbb{C}^{d}. For inifinite-dimensional case, Kuiper [Kui65] proved that the space of all unitary operators on an inifinite-dimensional Hilbert space is contractible. This result is basic in the Atiyah–Singer index theory. This kind of contractibility result has been extended to Ud(A)\operatorname{U}_{d}(A) of some other algebras AA while AA is all the bounded operators on a infinite dimensional Hilbert space in the original result. Of course it is not always the case for Ud(A)\operatorname{U}_{d}(A) of other infinite-dimensional CC^{\ast}-algebras AA. In general, it is hard to determine the homotopy type of Ud(A)\operatorname{U}_{d}(A).

Let us use the notation

GL(A)=limdGLd(A)andU(A)=limdUd(A).\operatorname{GL}_{\infty}(A)=\lim_{d\to\infty}\operatorname{GL}_{d}(A)\quad\text{and}\quad\operatorname{U}_{\infty}(A)=\lim_{d\to\infty}\operatorname{U}_{d}(A).

It is well-known that U(A)\operatorname{U}_{\infty}(A) has the same homotopy type as U1(A𝒦)\operatorname{U}_{1}(A\otimes\mathcal{K}) where 𝒦\mathcal{K} is the space of compact operators. The KK-theory Ki(A)K_{i}(A) (i=0,1i=0,1) is a basic homotopy invariant of AA, which is characterized as

K0(A)=π1(U(A))andK1(A)=π0(U(A)).K_{0}(A)=\pi_{1}(\operatorname{U}_{\infty}(A))\quad\text{and}\quad K_{1}(A)=\pi_{0}(\operatorname{U}_{\infty}(A)).

Since Ud(A)\operatorname{U}_{d}(A) is not necessarily homotopy equivalent to U(A)\operatorname{U}_{\infty}(A), Ki(A)K_{i}(A) is not a so strong invariant in general. But sometimes the natural map Ud(A)U(A)\operatorname{U}_{d}(A)\to\operatorname{U}_{\infty}(A), which we will call the stabilizing map, becomes a homotopy equivalence. Study on such stability seems to trace back to the work of Bass [Bas64]. There have been a various works on this kind of stability. Rieffel introduced the topological stable rank in [Rie83] and applied it to show the stability of the non-commutative torus in [Rie87], which is a key tool in the present work. It is difficult in general to determine how stable a given CC^{\ast}-algebra is.

In the present paper, we study the stability of the uniform Roe algebra Cu(|n|)C^{\ast}_{u}(|\mathbb{Z}^{n}|) on n\mathbb{Z}^{n} and investigate its homotopy type. The uniform Roe algebra Cu(X)C^{\ast}_{u}(X) of a metric space XX introduced by Roe in [Roe88] to establish an index theory on open manifolds, where the index lives in the KK-theory K(Cu(X))K_{\ast}(C^{\ast}_{u}(X)). The algebra Cu(X)C^{\ast}_{u}(X) itself is also important since it encodes a kind of “large scale geometry” of XX. Studying the homotopy type of Ud(Cu(X))\operatorname{U}_{d}(C^{\ast}_{u}(X)) will provide more insights from a homotopy theoretic viewpoint, which cannot be obtained only from its KK-theory. But there are only a few works on the homotopy type of Ud(Cu(X))\operatorname{U}_{d}(C^{\ast}_{u}(X)) yet. For example, Manuilov and Troitsky [MT21] studied some condition for Ud(Cu(X))\operatorname{U}_{d}(C^{\ast}_{u}(X)) being contractible. In the present work, we observe the other extreme, that is, Ud(Cu(|n|))\operatorname{U}_{d}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)) has a highly nontrivial homotopy type.

We give some comment on the relation with our previous work [KKT] on the space 𝒰\mathcal{U} of finite propagation unitary operators on \mathbb{Z}. Note that U1(Cu(||))\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}|)) can be viewed as a kind of completion of 𝒰\mathcal{U}. We determined the homotopy type of 𝒰\mathcal{U} there. But it is not clear whether 𝒰\mathcal{U} has the same homotopy type as U1(Cu(||))\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}|)). Actually, they turn out to have the same homotopy type (Theorem 1.2). Also, the method there does not seem to be extended to n\mathbb{Z}^{n} when n2n\geq 2. We employ rather operator algebraic technique in the present paper. Our method here reduces the problem to determine the homotopy type of Ud(Cu(|n|))\operatorname{U}_{d}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)) to the one to show the surjectivity of the homomorphism on KK-theory K(Cu(|n|))K(C(|n|))K_{\ast}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\to K_{\ast}(C^{\ast}(|\mathbb{Z}^{n}|)) induced from the inclusion (Proposition 7.4), where C(|n|)C^{\ast}(|\mathbb{Z}^{n}|) denotes the Roe algebra of n\mathbb{Z}^{n}.

For stability, we show the following theorem in Section 3.

Theorem 1.1.

For any integer n1n\geq 1, the stabilizing maps

GL1(Cu(|n|))GL(Cu(|n|)),\displaystyle\operatorname{GL}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\to\operatorname{GL}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)), U1(Cu(|n|))U(Cu(|n|)),\displaystyle\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\to\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)),
GL1(C(|n|))GL(C(|n|)),\displaystyle\operatorname{GL}_{1}(C^{\ast}(|\mathbb{Z}^{n}|))\to\operatorname{GL}_{\infty}(C^{\ast}(|\mathbb{Z}^{n}|)), U1(C(|n|))U(C(|n|))\displaystyle\operatorname{U}_{1}(C^{\ast}(|\mathbb{Z}^{n}|))\to\operatorname{U}_{\infty}(C^{\ast}(|\mathbb{Z}^{n}|))

between the spaces of invertible and unitary elements are homotopy equivalences. This implies that these maps induce the following isomorphisms on homotopy groups for all i0i\geq 0:

πi(GL1(Cu(|n|)))πi(U1(Cu(|n|))){K1(Cu(|n|))i is even,K0(Cu(|n|))i is odd,\displaystyle\pi_{i}(\operatorname{GL}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)))\cong\pi_{i}(\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)))\cong\begin{cases}K_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))&\text{$i$ is even,}\\ K_{0}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))&\text{$i$ is odd,}\end{cases}
πi(GL1(C(|n|)))πi(U1(C(|n|))){K1(C(|n|))i is even,K0(C(|n|))i is odd.\displaystyle\pi_{i}(\operatorname{GL}_{1}(C^{\ast}(|\mathbb{Z}^{n}|)))\cong\pi_{i}(\operatorname{U}_{1}(C^{\ast}(|\mathbb{Z}^{n}|)))\cong\begin{cases}K_{1}(C^{\ast}(|\mathbb{Z}^{n}|))&\text{$i$ is even,}\\ K_{0}(C^{\ast}(|\mathbb{Z}^{n}|))&\text{$i$ is odd.}\end{cases}

Let K(V,i)K(V,i) denote the Eilenberg–MacLane space of type (V,i)(V,i) and BU()B\operatorname{U}_{\infty}(\mathbb{C}) denote the classifying space of the unitary group U()\operatorname{U}_{\infty}(\mathbb{C}). Also, for based spaces XiX_{i} (i=1,2,i=1,2,\ldots), define

i1Xi=limk(X1×X2××Xk).\prod_{i\geq 1}^{\circ}X_{i}=\lim_{k\to\infty}(X_{1}\times X_{2}\times\cdots\times X_{k}).

For the homotopy type of U1(Cu(|n|))\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)), we show the following results when n=1,2n=1,2.

Theorem 1.2.

There exist homotopy equivalences of infinite loop spaces

GL1(Cu(||))U1(Cu(||))×BU()×i1K((,)S,2i1).\operatorname{GL}_{1}(C^{\ast}_{u}(|\mathbb{Z}|))\simeq\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}|))\simeq\mathbb{Z}\times B\operatorname{U}_{\infty}(\mathbb{C})\times\prod_{i\geq 1}^{\circ}K(\ell^{\infty}(\mathbb{Z},\mathbb{Z})_{S},2i-1).
Theorem 1.3.

There exist homotopy equivalences of infinite loop spaces

GL1(Cu(|2|))U1(Cu(|2|))V1×U()×i1(K(V0,2i1)×K(V1,2i)),\operatorname{GL}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\simeq\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\simeq V_{1}\times\operatorname{U}_{\infty}(\mathbb{C})\times\prod_{i\geq 1}^{\circ}(K(V_{0},2i-1)\times K(V_{1},2i)),

where V0,V1V_{0},V_{1} are the rational vector spaces given by

V0=ker[K0(Cu(|2|))K0(C(|2|))],V1=K1(Cu(|2|))V_{0}=\ker[K_{0}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\to K_{0}(C^{\ast}(|\mathbb{Z}^{2}|))],\quad V_{1}=K_{1}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))

and the product factor V1V_{1} is a discrete space.

More detailed descriptions of the vector spaces V0V_{0} and V1V_{1} appear in the proof of Lemma 7.7.

We will see the existence of a homotopy section of the inclusion U1(Cu(||))U1(BSW)\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}|))\to\operatorname{U}_{1}(B^{\mathrm{SW}}) in Section 6, where U1(BSW)\operatorname{U}_{1}(B^{\mathrm{SW}}) is the Segal–Wilson restricted unitary group [SW85] having the homotopy type of ×BU()\mathbb{Z}\times B\operatorname{U}_{\infty}(\mathbb{C}). This implies Theorem 1.2. We also show in Section 7 that, for any integer n1n\geq 1, the inclusion U1(Cu(|n|))U1(C(|n|))\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\to\operatorname{U}_{1}(C^{\ast}(|\mathbb{Z}^{n}|)) admits a homotopy section if and only if the homomorphism K(Cu(|n|))K(C(|n|))K_{\ast}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\to K_{\ast}(C^{\ast}(|\mathbb{Z}^{n}|)) is surjective. Since we can see it is surjective when n=1,2n=1,2, Theorems 1.2 again and 1.3 follows. If one could show the surjectivity for n3n\geq 3, then a similar homotopy decomposition will immediately follow.

This paper is organized as follows. We fix our notation in Section 2. In Section 3, we recall Rieffel’s results on stability and show Theorem 1.1. In Section 4, we recall the Bott periodicity realized as a \ast-homomorphism. In Section 5, we recall the Segal–Wilson restricted unitary group and show its stability. In Section 6, we show Theorem 1.2 using the Segal–Wilson restricted unitary group. In Section 7, we discuss the homotopy type of Ud(Cu(|n|))\operatorname{U}_{d}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)) for general n1n\geq 1 and show Theorems 1.2 again and 1.3.

2. Notation

The CC^{\ast}-algebra of bounded operators on a Hilbert space VV is denoted by (V)\mathcal{B}(V) and the subalgebra of compact operators by 𝒦(V)\mathcal{K}(V). We write the operator norm of T(V)T\in\mathcal{B}(V) as T\|T\|. The Hilbert space of square summable sequences indexed by a discrete group Γ\Gamma will be written as

2(Γ)={(vg)ggΓ|vg|2<}.\ell^{2}(\Gamma)=\{(v_{g})_{g}\mid\sum_{g\in\Gamma}|v_{g}|^{2}<\infty\}.

We also consider the tensor product Hilbert space 2(Γ)\ell^{2}(\Gamma)\otimes\mathcal{H} with an infinite dimensional separable Hilbert space \mathcal{H}.

A bounded operator T(2(Γ))T\in\mathcal{B}(\ell^{2}(\Gamma)) can be expressed in the matrix form as

T=(Tg,h)g,h,Tg,h.T=(T_{g,h})_{g,h},\quad T_{g,h}\in\mathbb{C}.

For T(2(Γ))T\in\mathcal{B}(\ell^{2}(\Gamma)\otimes\mathcal{H}), we also have a similar expression T=(Tg,h)g,hT=(T_{g,h})_{g,h} with Tg,h()T_{g,h}\in\mathcal{B}(\mathcal{H}).

Definition 2.1.

Let Γ\Gamma be a finitely generated group and dd denote the word metric with respect to some finite set of generators. We say that a bounded operator T(2(Γ))T\in\mathcal{B}(\ell^{2}(\Gamma)) has finite propagation if

prop(T)=sup{d(g,h)Tg,h0}\operatorname{prop}(T)=\sup\{d(g,h)\mid T_{g,h}\neq 0\}

is finite. We define finite propagation for T=(Tg,h)g,h(2(Γ))T=(T_{g,h})_{g,h}\in\mathcal{B}(\ell^{2}(\Gamma)\otimes\mathcal{H}) similarly.

Example 2.2.

The shift Sx(2(Γ))S_{x}\in\mathcal{B}(\ell^{2}(\Gamma)) by xΓx\in\Gamma is defined by

Sx=((Sx)g,h)g,h,Sg,h={1g1h=x,0otherwise.S_{x}=((S_{x})_{g,h})_{g,h},\quad S_{g,h}=\begin{cases}1&g^{-1}h=x,\\ 0&\text{otherwise.}\end{cases}

The operator SxS_{x} is a unitary operator with prop(Sx)=d(x,1)\operatorname{prop}(S_{x})=d(x,1).

It is easy to see that the definition of having finite propagation is independent of the choice of generators while the value of prop(T)\operatorname{prop}(T) depends on the word metric. Since we have

prop(ST)prop(S)+prop(T),prop(T)=prop(T),prop(1)=0\operatorname{prop}(ST)\leq\operatorname{prop}(S)+\operatorname{prop}(T),\quad\operatorname{prop}(T^{\ast})=\operatorname{prop}(T),\quad\operatorname{prop}(1)=0

for any finite propagation operators S,T(2(Γ))S,T\in\mathcal{B}(\ell^{2}(\Gamma)), the subset of finite propagation operators becomes a unital \ast-subalgebra of (2(Γ))\mathcal{B}(\ell^{2}(\Gamma)). Similar properties hold for finite propagation operators S,T(2(Γ))S,T\in\mathcal{B}(\ell^{2}(\Gamma)\otimes\mathcal{H}) such that the components Tg,hT_{g,h} and Sg,hS_{g,h} are compact operators.

Definition 2.3.

The uniform Roe algebra Cu(|Γ|)C^{\ast}_{u}(|\Gamma|) of Γ\Gamma is the norm closure of the algebra of finite propagation operators in (2(Γ))\mathcal{B}(\ell^{2}(\Gamma)).

Definition 2.4.

The Roe algebra C(|Γ|)C^{\ast}(|\Gamma|) of Γ\Gamma is the norm closure of the algebra of finite propagation operators T(2(Γ))T\in\mathcal{B}(\ell^{2}(\Gamma)\otimes\mathcal{H}) such that each component Tg,hT_{g,h} is a compact operator.

Remark 2.5.

We follow the usual notation C(|Γ|)C^{\ast}(|\Gamma|) for the Roe algebra of Γ\Gamma to distinguish it from the group CC^{\ast}-algebra of Γ\Gamma though we do not consider the latter here.

We will consider the uniform Roe algebra Cu(|Γ|)C^{\ast}_{u}(|\Gamma|) is a subalgebra of the Roe algebra C(|Γ|)C^{\ast}(|\Gamma|) with respect to some inclusion \mathbb{C}\subset\mathcal{H}.

We use the symbol (Γ,)\ell^{\infty}(\Gamma,\mathbb{C}) to express the Banach algebra of \mathbb{C}-valued bounded sequences indexed by Γ\Gamma rather than the simpler symbol (Γ)\ell^{\infty}(\Gamma) since we also consider the abelian group of \mathbb{Z}-valued bounded sequences (Γ,)\ell^{\infty}(\Gamma,\mathbb{Z}).

The group Γ\Gamma acts on the algebras (Γ,)\ell^{\infty}(\Gamma,\mathbb{C}) and (Γ,𝒦())\ell^{\infty}(\Gamma,\mathcal{K}(\mathcal{H})) by right translation. The action by xΓx\in\Gamma is compatible with the conjugation by SxS_{x} through the diagonal inclusion (Γ,)Cu(|Γ|)\ell^{\infty}(\Gamma,\mathbb{C})\to C^{\ast}_{u}(|\Gamma|) or (Γ,𝒦())C(|Γ|)\ell^{\infty}(\Gamma,\mathcal{K}(\mathcal{H}))\to C^{\ast}(|\Gamma|) given by

(tg)g(Tg,h)g,h,Tg,h={tgg=h,0otherwise.(t_{g})_{g}\mapsto(T_{g,h})_{g,h},\quad T_{g,h}=\begin{cases}t_{g}&g=h,\\ 0&\text{otherwise.}\end{cases}

Moreover, this inclusion extends to the well-known isomorphisms

(Γ,)ΓCu(|Γ|),(Γ,𝒦())ΓC(|Γ|)\ell^{\infty}(\Gamma,\mathbb{C})\rtimes\Gamma\cong C^{\ast}_{u}(|\Gamma|),\quad\ell^{\infty}(\Gamma,\mathcal{K}(\mathcal{H}))\rtimes\Gamma\cong C^{\ast}(|\Gamma|)

from the reduced crossed products of CC^{\ast}-algebras. For example, see [Roe03, Theorem 4.28].

The d×dd\times d-matrix algebra Md(A)M_{d}(A) of a CC^{\ast}-algebra AA is again a CC^{\ast}-algebra. The spaces of invertible elements and unitary elements in Md(A)M_{d}(A) will be denoted as GLd(A)\operatorname{GL}_{d}(A) and Ud(A)\operatorname{U}_{d}(A). The stabilizing maps are given as

GL1(A)GL(A)=limdGLd(A),U1(A)U(A)=limdUd(A),\operatorname{GL}_{1}(A)\to\operatorname{GL}_{\infty}(A)=\lim_{d\to\infty}\operatorname{GL}_{d}(A),\quad\operatorname{U}_{1}(A)\to\operatorname{U}_{\infty}(A)=\lim_{d\to\infty}\operatorname{U}_{d}(A),

where the inductive limits are taken along the inclusions GLd(A)GLd+1(A)\operatorname{GL}_{d}(A)\subset\operatorname{GL}_{d+1}(A) and Ud(A)Ud+1(A)\operatorname{U}_{d}(A)\subset\operatorname{U}_{d+1}(A). The inductive limit spaces GL(A)\operatorname{GL}_{\infty}(A) and U(A)\operatorname{U}_{\infty}(A) are well-known to be homotopy equivalent to the spaces GL1(A𝒦())\operatorname{GL}_{1}(A\otimes\mathcal{K}(\mathcal{H})) and U1(A𝒦())\operatorname{U}_{1}(A\otimes\mathcal{K}(\mathcal{H})).

3. Stability

The aim of this section is to prove Theorem 1.1. Once the assumption of the following result by Rieffel [Rie87] is verified, the theorem will immediately follow.

Theorem 3.1 (Rieffel).

Let AA be a unital CC^{\ast}-algebra. If AA is tsr-boundedly divisible, then the stabilizing maps

GL1(A)GL(A)andU1(A)U(A)\operatorname{GL}_{1}(A)\to\operatorname{GL}_{\infty}(A)\quad\text{and}\quad\operatorname{U}_{1}(A)\to\operatorname{U}_{\infty}(A)

are homotopy equivalences.

Remark 3.2.

The original statement of Theorem 4.13 in [Rie87] is involved only with homotopy groups. But what is actually proved there is slightly stronger as above.

For the definitions of the topological stable rank tsr(A)1\operatorname{tsr}(A)\in\mathbb{Z}_{\geq 1}, see [Rie83]. A CC^{\ast}-algebra AA is said to be tsr-boundedly divisible [Rie87] if there is a constant KK such that for any integer mm, there exists an integer dmd\geq m such that AA is isomorphic to Md(B)M_{d}(B) for some CC^{\ast}-algebra BB with tsr(B)K\operatorname{tsr}(B)\leq K. To verify the assumption, we need the following two lemmas.

Lemma 3.3.

The topological stable ranks of Cu(|n|)C^{\ast}_{u}(|\mathbb{Z}^{n}|) and C(|n|)C^{\ast}(|\mathbb{Z}^{n}|) are estimated as

tsr(Cu(|n|))n+1andtsr(C(|n|))n+1.\operatorname{tsr}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\leq n+1\quad\text{and}\quad\operatorname{tsr}(C^{\ast}(|\mathbb{Z}^{n}|))\leq n+1.
Remark 3.4.

We will see that both Cu(|n|)C^{\ast}_{u}(|\mathbb{Z}^{n}|) and C(|n|)C^{\ast}(|\mathbb{Z}^{n}|) are tsr-boundedly divisible using this lemma. Thus we will actually obtain the estimates tsr(Cu(|n|))2\operatorname{tsr}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\leq 2 and tsr(C(|n|))2\operatorname{tsr}(C^{\ast}(|\mathbb{Z}^{n}|))\leq 2 by [Rie87, Proposition 4.6].

Proof.

Let A=A=\mathbb{C} or 𝒦()\mathcal{K}(\mathcal{H}). Since the invertible elements in (n,)\ell^{\infty}(\mathbb{Z}^{n},\mathbb{C}) and (n,𝒦())\mathbb{C}\oplus\ell^{\infty}(\mathbb{Z}^{n},\mathcal{K}(\mathcal{H})) are dense, we have

tsr((n,A))=1\operatorname{tsr}(\ell^{\infty}(\mathbb{Z}^{n},A))=1

by [Rie83, Proposition 3.1]. Considering the restricted action of mn\mathbb{Z}^{m}\subset\mathbb{Z}^{n} on the first mm factors of n\mathbb{Z}^{n}, we obtain the isomorphism

(n,A)m+1((n,A)m).\ell^{\infty}(\mathbb{Z}^{n},A)\rtimes\mathbb{Z}^{m+1}\cong(\ell^{\infty}(\mathbb{Z}^{n},A)\rtimes\mathbb{Z}^{m})\rtimes\mathbb{Z}.

Thus, by [Rie83, Theorem 7.1], we get the desired estimates on tsr(Cu(||))\operatorname{tsr}(C^{\ast}_{u}(|\mathbb{Z}|)) and tsr(C(|n|))\operatorname{tsr}(C^{\ast}(|\mathbb{Z}^{n}|)). ∎

Lemma 3.5.

For any integer d1d\geq 1, there exist isomorphisms

ϕ:Cu(|n|)Md(Cu(|n|))andϕ:C(|n|)Md(C(|n|)).\phi\colon C^{\ast}_{u}(|\mathbb{Z}^{n}|)\cong M_{d}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\quad\text{and}\quad\phi\colon C^{\ast}(|\mathbb{Z}^{n}|)\cong M_{d}(C^{\ast}(|\mathbb{Z}^{n}|)).
Proof.

Let V=V=\mathbb{C} or \mathcal{H}. According to the decomposition

2(n)V=(i1,,in)nV(i1,,in),V(i1,,in)V,\ell^{2}(\mathbb{Z}^{n})\otimes V=\bigoplus_{(i_{1},\ldots,i_{n})\in\mathbb{Z}^{n}}V_{(i_{1},\ldots,i_{n})},\quad V_{(i_{1},\ldots,i_{n})}\cong V,

we have the matrix expression for T(2(n)V)T\in\mathcal{B}(\ell^{2}(\mathbb{Z}^{n})\otimes V)

T=(T(i1,,in)(j1,,jn))(i1,,in)(j1,,jn),T(i1,,in)(j1,,jn):V(j1,,jn)V(i1,,in).T=(T_{(i_{1},\ldots,i_{n})(j_{1},\ldots,j_{n})})_{(i_{1},\ldots,i_{n})(j_{1},\ldots,j_{n})},\quad T_{(i_{1},\ldots,i_{n})(j_{1},\ldots,j_{n})}\colon V_{(j_{1},\ldots,j_{n})}\to V_{(i_{1},\ldots,i_{n})}.

Consider the map ϕ:(2(n)V)Md((2(n)V))\phi\colon\mathcal{B}(\ell^{2}(\mathbb{Z}^{n})\otimes V)\to M_{d}(\mathcal{B}(\ell^{2}(\mathbb{Z}^{n})\otimes V)) given by

ϕ(T)(i1,,in)(j1,,jn)=(T(di1,,in)(dj1,,jn)T(di1,,in)(dj1+d1,,jn)T(di1+d1,,in)(dj1,,jn)T(di1+d1,,in)(dj1+d1,,jn))Md((V)).\phi(T)_{(i_{1},\ldots,i_{n})(j_{1},\ldots,j_{n})}=\begin{pmatrix}T_{(di_{1},\ldots,i_{n})(dj_{1},\ldots,j_{n})}&\cdots&T_{(di_{1},\ldots,i_{n})(dj_{1}+d-1,\ldots,j_{n})}\\ \vdots&\ddots&\vdots\\ T_{(di_{1}+d-1,\ldots,i_{n})(dj_{1},\ldots,j_{n})}&\cdots&T_{(di_{1}+d-1,\ldots,i_{n})(dj_{1}+d-1,\ldots,j_{n})}\end{pmatrix}\in M_{d}(\mathcal{B}(V)).

The restrictions to Cu(|n|)(2(n))C^{\ast}_{u}(|\mathbb{Z}^{n}|)\subset\mathcal{B}(\ell^{2}(\mathbb{Z}^{n})) and C(|n|)(2(n))C^{\ast}(|\mathbb{Z}^{n}|)\subset\mathcal{B}(\ell^{2}(\mathbb{Z}^{n})\otimes\mathcal{H}) are desired isomorphisms. ∎

Remark 3.6.

When n=1n=1, the map ϕ\phi is just taking the block matrix of which each block is a d×dd\times d-matrix.

Proof of Theorem 1.1.

By Lemmas 3.3 and 3.5, we can apply Theorem 3.1 to Cu(|n|)C^{\ast}_{u}(|\mathbb{Z}^{n}|) and C(|n|)C^{\ast}(|\mathbb{Z}^{n}|). This completes the proof of the theorem. ∎

4. Bott periodicity

Let us recall the Bott periodicity of CC^{\ast}-algebras here. Let AA be a CC^{\ast}-algebra, which might be non-unital. The direct sum A\mathbb{C}\oplus A is considered as the unitization with unit (1,0)A(1,0)\in\mathbb{C}\oplus A. Define the unitary group Ud(A)\operatorname{U}^{\prime}_{d}(A) by

Ud(A)={UUn(A)U(Id,0)Md(A)}.\operatorname{U}^{\prime}_{d}(A)=\{U\in\operatorname{U}_{n}(\mathbb{C}\oplus A)\mid U-(I_{d},0)\in M_{d}(A)\}.

If AA is already unital, we have a canonical isomorphism Ud(A)Ud(A)\operatorname{U}^{\prime}_{d}(A)\cong\operatorname{U}_{d}(A). So we use the same symbol Ud(A)\operatorname{U}_{d}(A) for Ud(A)\operatorname{U}^{\prime}_{d}(A) even if AA is not unital.

Consider the following space of continuous functions:

C0(m,A)={T:mAT is continuous and lim|z|T(z)=0}.C_{0}(\mathbb{R}^{m},A)=\{T\colon\mathbb{R}^{m}\to A\mid T\text{ is continuous and }\lim_{|z|\to\infty}T(z)=0\}.

This is a CC^{\ast}-algebra without unit. Notice that C0(m,A)C_{0}(\mathbb{R}^{m},A) is isomorphic to the space ΩmA\Omega^{m}A of based maps from the mm-sphere SmS^{m} to AA where the basepoint Sm\ast\in S^{m} is mapped to 0A0\in A.

Set the element

pB(z)=11+|z|2(|z|2zz¯1)M2(C0(2,))(z2),p_{B}(z)=\frac{1}{1+|z|^{2}}\begin{pmatrix}|z|^{2}&z\\ \bar{z}&1\end{pmatrix}\in M_{2}(\mathbb{C}\oplus C_{0}(\mathbb{R}^{2},\mathbb{C}))\quad(z\in\mathbb{R}^{2}),

where we identify 2\mathbb{R}^{2}\cong\mathbb{C} in the matrix entries. The Bott map β:AM2(AC0(2,A))\beta\colon A\to M_{2}(A\oplus C_{0}(\mathbb{R}^{2},A)) is a \ast-homomorphism defined by

β(a)=pB(a00a)=(a00a)pB.\beta(a)=p_{B}\begin{pmatrix}a&0\\ 0&a\end{pmatrix}=\begin{pmatrix}a&0\\ 0&a\end{pmatrix}p_{B}.

Then we have the commutative square of unital CC^{\ast}-algebras

A\textstyle{\mathbb{C}\oplus A\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϵ\scriptstyle{\epsilon}β\scriptstyle{\beta}\textstyle{\mathbb{C}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}η\scriptstyle{\eta}M2(AC0(2,A))\textstyle{M_{2}(\mathbb{C}\oplus A\oplus C_{0}(\mathbb{R}^{2},A))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}ϵ\scriptstyle{\epsilon}M2(A)\textstyle{M_{2}(\mathbb{C}\oplus A)}

where ϵ:A\epsilon\colon\mathbb{C}\oplus A\to\mathbb{C} and ϵ:M2(AC0(2,A))M2(A)\epsilon\colon M_{2}(\mathbb{C}\oplus A\oplus C_{0}(\mathbb{R}^{2},A))\to M_{2}(\mathbb{C}\oplus A) are the projections and η:M2(A)\eta\colon\mathbb{C}\to M_{2}(\mathbb{C}\oplus A) is the unit map. This square induces the \ast-homomorphism between the kernels of ϵ\epsilon:

β:AM2(C0(2,A)).\beta\colon A\to M_{2}(C_{0}(\mathbb{R}^{2},A)).

We call this β\beta the Bott map as well. It is natural in the following sense: if f:ABf\colon A\to B is a \ast-homomorphism between CC^{\ast}-algebras, then the following square commutes:

A\textstyle{A\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}f\scriptstyle{f}β\scriptstyle{\beta}B\textstyle{B\ignorespaces\ignorespaces\ignorespaces\ignorespaces}β\scriptstyle{\beta}M2(C0(2,A))\textstyle{M_{2}(C_{0}(\mathbb{R}^{2},A))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}f\scriptstyle{f_{\ast}}M2(C0(2,B))\textstyle{M_{2}(C_{0}(\mathbb{R}^{2},B))}
Proposition 4.1.

The Bott map β:AM2(C0(2,A))\beta\colon A\to M_{2}(C_{0}(\mathbb{R}^{2},A)) induces an isomorphism on KK-theory.

Remark 4.2.

This can be seen as a formulation of the Bott periodicity. If you wish to deduce this proposition from the results appearing in [Bla86], it follows from the observation 9.2.10 on the generator of K0(C0(2,))K_{0}(C_{0}(\mathbb{R}^{2},\mathbb{C})) and the Künneth theorem for tensor products (Theorem 23.1.3).

The Bott periodicity provides the natural homotopy equivalence

U(A)𝛽U(M2(C0(2,A)))Ω2U(A),\operatorname{U}_{\infty}(A)\xrightarrow{\beta}\operatorname{U}_{\infty}(M_{2}(C_{0}(\mathbb{R}^{2},A)))\simeq\Omega^{2}\operatorname{U}_{\infty}(A),

which is a group homomorphism. Thus we obtain the following proposition on infinite loop structure.

Proposition 4.3.

The unitary group U(A)\operatorname{U}_{\infty}(A) of a CC^{\ast}-algebra AA is equipped with a canonical infinite loop space structure such that the map U(A)U(B)\operatorname{U}_{\infty}(A)\to\operatorname{U}_{\infty}(B) induced from a \ast-homomorphism ABA\to B is an infinite loop map. Moreover, the underlying loop structure of U(A)\operatorname{U}_{\infty}(A) coincides with the group structure of U(A)\operatorname{U}_{\infty}(A).

Remark 4.4.

The last sentence in the proposition means that there exists a homotopy equivalence BU(A)ΩU(A)B\operatorname{U}_{\infty}(A)\simeq\Omega\operatorname{U}_{\infty}(A) from the classifying space BU(A)B\operatorname{U}_{\infty}(A) of the topological group U(A)\operatorname{U}_{\infty}(A).

5. Segal–Wilson restricted unitary group

To study the homotopy type of U1(Cu(||))\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}|)), we will relate it with other spaces. One is the Segal–Wilson restricted unitary group U1(BSW)\operatorname{U}_{1}(B^{\mathrm{SW}}) and the other is the unitary group of the Roe algebra U1(C(||))\operatorname{U}_{1}(C^{\ast}(|\mathbb{Z}|)). We recall the former in this section.

We have another matrix expression for T(2())T\in\mathcal{B}(\ell^{2}(\mathbb{Z})) as

T=(TT+T+T++),T=\begin{pmatrix}T_{--}&T_{-+}\\ T_{+-}&T_{++}\end{pmatrix},

where

T:2(<0)2(<0),T+:2(0)2(<0),\displaystyle T_{--}\colon\ell^{2}(\mathbb{Z}_{<0})\to\ell^{2}(\mathbb{Z}_{<0}),\quad T_{-+}\colon\ell^{2}(\mathbb{Z}_{\geq 0})\to\ell^{2}(\mathbb{Z}_{<0}),
T+:2(<0)2(0),T++:2(0)2(0).\displaystyle T_{+-}\colon\ell^{2}(\mathbb{Z}_{<0})\to\ell^{2}(\mathbb{Z}_{\geq 0}),\quad T_{++}\colon\ell^{2}(\mathbb{Z}_{\geq 0})\to\ell^{2}(\mathbb{Z}_{\geq 0}).
Definition 5.1.

We define the CC^{\ast}-algebra BSWB^{\mathrm{SW}} by

BSW:={T()T+,T+ are compact}.\displaystyle B^{\mathrm{SW}}:=\{T\in\mathcal{B}(\mathcal{H})\mid\text{$T_{-+},T_{+-}$ are compact}\}.

The symbol “SW” stands for Segal–Wilson. The unitary group U1(BSW)\operatorname{U}_{1}(B^{\mathrm{SW}}) is called the restricted unitary group in the work of Segal and Wilson [SW85]. They used it as a model of the infinite Grassmannian.

Lemma 5.2 (Segal–Wilson).

The space U1(BSW)\operatorname{U}_{1}(B^{\mathrm{SW}}) has the homotopy type of ×BU()\mathbb{Z}\times B\operatorname{U}_{\infty}(\mathbb{C}). Moreover, the map

π0(U1(BSW)),[U]ind(U++),\pi_{0}(\operatorname{U}_{1}(B^{\mathrm{SW}}))\to\mathbb{Z},\quad[U]\mapsto\operatorname{ind}(U_{++}),

is bijective, where ind(U++)\operatorname{ind}(U_{++}) denotes the Fredholm index of the Fredholm operator U++U_{++}.

Let S=S+1BSWS=S_{+1}\in B^{\mathrm{SW}} the shift operator as in Example 2.2. We have indSn=n\operatorname{ind}S^{n}=n.

The goal of this section is to see the following.

Proposition 5.3.

The stabilizing maps

GL1(BSW)GL(BSW)andU1(BSW)U(BSW)\operatorname{GL}_{1}(B^{\mathrm{SW}})\to\operatorname{GL}_{\infty}(B^{\mathrm{SW}})\quad\text{and}\quad\operatorname{U}_{1}(B^{\mathrm{SW}})\to\operatorname{U}_{\infty}(B^{\mathrm{SW}})

are homotopy equivalences.

To show this, we do not use a kind of stability as in Section 3.

Lemma 5.4.

For any integer d1d\geq 1, the inclusion

U1(BSW)Ud(BSW)\operatorname{U}_{1}(B^{\mathrm{SW}})\to\operatorname{U}_{d}(B^{\mathrm{SW}})

induces an isomorphism on π0\pi_{0}.

Proof.

Consider the composite of the inclusion and the isomorphism ϕ:BSWMd(BSW)\phi\colon B^{\mathrm{SW}}\to M_{d}(B^{\mathrm{SW}}) similar to the one in the proof of Lemma 3.5:

U1(BSW)Ud(BSW)ϕ1U1(BSW).\operatorname{U}_{1}(B^{\mathrm{SW}})\to\operatorname{U}_{d}(B^{\mathrm{SW}})\xrightarrow{\phi^{-1}}\operatorname{U}_{1}(B^{\mathrm{SW}}).

It is easy to see that the image of the shift SU1(BSW)S\in\operatorname{U}_{1}(B^{\mathrm{SW}}) under this composite again has index 11. This implies the lemma. ∎

Lemma 5.5.

The KK-theory of BSWB^{\mathrm{SW}} is computed as

Ki(BSW){0i=0,i=1,K_{i}(B^{\mathrm{SW}})\cong\begin{cases}0&i=0,\\ \mathbb{Z}&i=1,\end{cases}

where K1(BSW)K_{1}(B^{\mathrm{SW}}) is generated by the shift SU1(BSW)S\in\operatorname{U}_{1}(B^{\mathrm{SW}}).

Proof.

This follows from the isomorphisms

K0(BSW)limdπ1(Ud(BSW))andK1(BSW)limdπ0(Ud(BSW))K_{0}(B^{\mathrm{SW}})\cong\lim_{d\to\infty}\pi_{1}(\operatorname{U}_{d}(B^{\mathrm{SW}}))\quad\text{and}\quad K_{1}(B^{\mathrm{SW}})\cong\lim_{d\to\infty}\pi_{0}(\operatorname{U}_{d}(B^{\mathrm{SW}}))

and Lemmas 5.2 and 5.4. ∎

Lemma 5.6.

For any i0i\geq 0, there exists an integer m1m\geq 1 such that the iterated Bott map

βi:Ud(BSW)Ud(M2i(C0(2i,BSW)))\beta^{i}\colon\operatorname{U}_{d}(B^{\mathrm{SW}})\to\operatorname{U}_{d}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}})))

induces an isomorphism on π0\pi_{0} if dmd\geq m.

Proof.

From the isomorphisms

π0(Ud(M2i(C0(2i,BSW)))π2i(U2id(BSW))π2i(U1(BSW))\pi_{0}(\operatorname{U}_{d}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}})))\cong\pi_{2i}(\operatorname{U}_{2^{i}d}(B^{\mathrm{SW}}))\cong\pi_{2i}(\operatorname{U}_{1}(B^{\mathrm{SW}}))\cong\mathbb{Z}

and

K1(M2i(C0(2i,BSW)))limdπ0(Ud(M2i(C0(2i,BSW)))K1(BSW),K_{1}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}})))\cong\lim_{d\to\infty}\pi_{0}(\operatorname{U}_{d}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}})))\cong K_{1}(B^{\mathrm{SW}})\cong\mathbb{Z},

we can find an integer m1m\geq 1 such that the stabilizing map

π0(Ud(M2i(C0(2i,BSW)))K1(C0(M2i(2i,BSW)))\pi_{0}(\operatorname{U}_{d}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}})))\to K_{1}(C_{0}(M_{2^{i}}(\mathbb{R}^{2i},B^{\mathrm{SW}})))

is an isomorphism if dmd\geq m. Consider the commutative diagram

π0(Ud(BSW))\textstyle{\pi_{0}(\operatorname{U}_{d}(B^{\mathrm{SW}}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}\scriptstyle{\cong}(βi)\scriptstyle{(\beta^{i})_{\ast}}K1(BSW)\textstyle{K_{1}(B^{\mathrm{SW}})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}(βi)\scriptstyle{(\beta^{i})_{\ast}}π0(Ud(M2i(C0(2i,BSW)))\textstyle{\pi_{0}(\operatorname{U}_{d}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}})))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}\scriptstyle{\cong}K1(M2i(C0(2i,BSW))\textstyle{K_{1}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}}))}

where the top arrow is an isomorphism by Lemma 5.4 and the right Bott map βi\beta^{i} is an isomorphism by Proposition 4.1. Then the lemma follows. ∎

Proof of Proposition 5.3.

Take an integer i0i\geq 0. We can find an integer m1m\geq 1 as in Proposition 5.6 and

π2i(Ud(BSW))π2i(U(BSW))\pi_{2i}(\operatorname{U}_{d}(B^{\mathrm{SW}}))\to\pi_{2i}(\operatorname{U}_{\infty}(B^{\mathrm{SW}}))\cong\mathbb{Z}

is an isomorphism if dmd\geq m. Consider the following commutative diagram:

U1(C0(2i,BSW))\textstyle{\operatorname{U}_{1}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}\scriptstyle{\cong}U1(M2i(C0(2i,BSW)))\textstyle{\operatorname{U}_{1}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}})))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}U1(BSW)\textstyle{\operatorname{U}_{1}(B^{\mathrm{SW}})\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}βi\scriptstyle{\beta^{i}}isom. on π0\pi_{0}Ud(C0(2i,BSW))\textstyle{\operatorname{U}_{d}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}\scriptstyle{\cong}Ud(M2i(C0(2i,BSW)))\textstyle{\operatorname{U}_{d}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}})))}Ud(BSW)\textstyle{\operatorname{U}_{d}(B^{\mathrm{SW}})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}βi\scriptstyle{\beta^{i}}

where the left horizontal arrows are the isomorphisms similar to the one in Lemma 3.5 and the vertical arrows are the inclusions. Since the composite

U1(BSW)Ud(BSW)βiUd(M2i(C0(2i,BSW)))\operatorname{U}_{1}(B^{\mathrm{SW}})\to\operatorname{U}_{d}(B^{\mathrm{SW}})\xrightarrow{\beta^{i}}\operatorname{U}_{d}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}})))

induces an isomorphism on π0\pi_{0}, the middle vertical arrow

U1(M2i(C0(2i,BSW)))Ud(M2i(C0(2i,BSW)))\operatorname{U}_{1}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}})))\to\operatorname{U}_{d}(M_{2^{i}}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}})))

induces a surjection on π0\pi_{0}. But it is indeed an isomorphism as their π0\pi_{0} are isomorphic to \mathbb{Z}. Then the map

U1(C0(2i,BSW))Ud(C0(2n,BSW))\operatorname{U}_{1}(C_{0}(\mathbb{R}^{2i},B^{\mathrm{SW}}))\to\operatorname{U}_{d}(C_{0}(\mathbb{R}^{2n},B^{\mathrm{SW}}))

induces an isomorphism on π0\pi_{0}. This implies that the map

U1(BSW)Ud(BSW)\operatorname{U}_{1}(B^{\mathrm{SW}})\to\operatorname{U}_{d}(B^{\mathrm{SW}})

induces an isomorphism on π2i\pi_{2i}. Thus the map

U1(BSW)U(BSW)\operatorname{U}_{1}(B^{\mathrm{SW}})\to\operatorname{U}_{\infty}(B^{\mathrm{SW}})

induces an isomorphism on π2i\pi_{2i}. This completes the proof. ∎

6. Homotopy type of U1(Cu(||))\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}|))

The goal of this section is to prove Theorem 1.2. The components T+T_{-+} and T+T_{+-} of a finite propagation operator T(H)T\in\mathcal{B}(H) are finite rank operators. This implies the inclusion

Cu(||)BSW.C^{\ast}_{u}(|\mathbb{Z}|)\subset B^{\mathrm{SW}}.

This map is a key to the proof of Theorem 1.2.

We begin with computing the KK-theory.

Proposition 6.1.

The following isomorphism holds:

K(Cu(||)){(,)Si=0,i=1,K_{\ast}(C^{\ast}_{u}(|\mathbb{Z}|))\cong\begin{cases}\ell^{\infty}(\mathbb{Z},\mathbb{Z})_{S}&i=0,\\ \mathbb{Z}&i=1,\end{cases}

where

(,)S=(,)/{aSaa(,)}\ell^{\infty}(\mathbb{Z},\mathbb{Z})_{S}=\ell^{\infty}(\mathbb{Z},\mathbb{Z})/\{a-Sa\mid a\in\ell^{\infty}(\mathbb{Z},\mathbb{Z})\}

is the coinvariant by the shift S:(,)(,)S\colon\ell^{\infty}(\mathbb{Z},\mathbb{Z})\to\ell^{\infty}(\mathbb{Z},\mathbb{Z}).

Proof.

Applying the Pimsner–Voiculescu exact sequence [PV80] to the crossed product

Cu(||)(,),C^{\ast}_{u}(|\mathbb{Z}|)\cong\ell^{\infty}(\mathbb{Z},\mathbb{C})\rtimes\mathbb{Z},

we get the six-term cyclic exact sequence:

K0((,))\textstyle{K_{0}(\ell^{\infty}(\mathbb{Z},\mathbb{C}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}1S\scriptstyle{1-S}K0((,))\textstyle{K_{0}(\ell^{\infty}(\mathbb{Z},\mathbb{C}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}K0(Cu(||))\textstyle{K_{0}(C^{\ast}_{u}(|\mathbb{Z}|))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}K1(Cu(||))\textstyle{K_{1}(C^{\ast}_{u}(|\mathbb{Z}|))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}K1((,))\textstyle{K_{1}(\ell^{\infty}(\mathbb{Z},\mathbb{C}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}K1((,))\textstyle{K_{1}(\ell^{\infty}(\mathbb{Z},\mathbb{C}))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}1S\scriptstyle{1-S}

As is well-known, we have

Ki((,)){(,)i=0,0i=1,K_{i}(\ell^{\infty}(\mathbb{Z},\mathbb{C}))\cong\begin{cases}\ell^{\infty}(\mathbb{Z},\mathbb{Z})&i=0,\\ 0&i=1,\end{cases}

where the induced homomorphism S:(,)(,)S\colon\ell^{\infty}(\mathbb{Z},\mathbb{Z})\to\ell^{\infty}(\mathbb{Z},\mathbb{Z}) is the shift as well. Thus we can compute K(Cu(||))K_{\ast}(C^{\ast}_{u}(|\mathbb{Z}|)) by the previous exact sequence. ∎

We saw the homotopy stabilities as in Theorem 1.1 and Proposition 5.3. Then it is sufficient to investigate the inclusion U(Cu(||))U(BSW)\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}|))\to\operatorname{U}_{\infty}(B^{\mathrm{SW}}).

Lemma 6.2.

The inclusion U(Cu(||))U(BSW)\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}|))\to\operatorname{U}_{\infty}(B^{\mathrm{SW}}) induces isomorphisms on π2i\pi_{2i} for i0i\geq 0.

Proof.

By Lemma 5.5, K1(BSW)K_{1}(B^{\mathrm{SW}}) is isomorphic to \mathbb{Z} and generated by the shift SBSWS\in B^{\mathrm{SW}}. Since SCu(||)S\in C^{\ast}_{u}(|\mathbb{Z}|) and K1(Cu(||))K_{1}(C^{\ast}_{u}(|\mathbb{Z}|))\cong\mathbb{Z}, the map K1(Cu(||))K1(BSW)K_{1}(C^{\ast}_{u}(|\mathbb{Z}|))\to K_{1}(B^{\mathrm{SW}}) is an isomorphism. Thus the map π2i(U(Cu(||)))π2i(U(BSW))\pi_{2i}(\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}|)))\to\pi_{2i}(\operatorname{U}_{\infty}(B^{\mathrm{SW}})) is also an isomorphism. ∎

Let F1F_{1} be the homotopy fiber of the inclusion U(Cu(||))U(BSW)\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}|))\to\operatorname{U}_{\infty}(B^{\mathrm{SW}}).

Proposition 6.3.

The space F1F_{1} has the homotopy type of the product of Eilenberg–MacLane spaces

i1K((,)S,2i1).\prod_{i\geq 1}^{\circ}K(\ell^{\infty}(\mathbb{Z},\mathbb{Z})_{S},2i-1).

where (,)S\ell^{\infty}(\mathbb{Z},\mathbb{Z})_{S} is a rational vector space of uncountable dimension.

Proof.

Observing the homotopy exact sequence

πi(F1)πi(U(Cu(||)))πi(U(BSW))πi1(F1),\cdots\to\pi_{i}(F_{1})\to\pi_{i}(\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}|)))\to\pi_{i}(\operatorname{U}_{\infty}(B^{\mathrm{SW}}))\to\pi_{i-1}(F_{1})\to\cdots,

we can see that the homotopy fiber inclusion F1U(Cu(||))F_{1}\to\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}|)) induces an isomorpshim on π2i1\pi_{2i-1} and π2i(F1)=0\pi_{2i}(F_{1})=0 by Lemma 6.2 and the fact that π2i1(U1(BSW))=0\pi_{2i-1}(\operatorname{U}_{1}(B^{\mathrm{SW}}))=0. By Proposition 6.1, we have π2i1(F1)(,)S\pi_{2i-1}(F_{1})\cong\ell^{\infty}(\mathbb{Z},\mathbb{Z})_{S}. The abelian group (,)S\ell^{\infty}(\mathbb{Z},\mathbb{Z})_{S} is a rational vector space of uncountable dimension as seen in [KKT, Section 5]. By [KKT, Lemma 5.4], F1F_{1} has the homotopy type of the product of Eilenberg–MacLane spaces as above. ∎

The following easy lemma is useful to study the homotopy type of the unitary group of a CC^{\ast}-algebra. Let prMd()p_{r}\in M_{d}(\mathbb{C}) denote the projection of rank rr.

Lemma 6.4.

Let AA be a CC^{\ast}-algebra, where we do not require the existence of unit. For any element uK0(A)u\in K_{0}(A), there exists a (non-unital in general) \ast-homomorphism f:Md(A)f\colon\mathbb{C}\to M_{d}(A) such that f[p1]K0(Md(A))K0(A)f_{\ast}[p_{1}]\in K_{0}(M_{d}(A))\cong K_{0}(A) equals to uu.

Proof.

We can find a projection pMd(A)p\in M_{d}(\mathbb{C}\oplus A) and r0r\geq 0 such that u=[p][pr]u=[p]-[p_{r}] in K0(A)K_{0}(A). Define a \ast-homomorphism f:Md(A)f\colon\mathbb{C}\to M_{d}(A) by f(1)=pf(1)=p. This is the desired map. ∎

Proposition 6.5.

The inclusion U(Cu(||))U(BSW)\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}|))\to\operatorname{U}_{\infty}(B^{\mathrm{SW}}) admits a homotopy section, which is an infinite loop map.

Proof.

Consider the inclusion of based loop spaces U(C0(,Cu(||)))U(C0(,BSW))\operatorname{U}_{\infty}(C_{0}(\mathbb{R},C^{\ast}_{u}(|\mathbb{Z}|)))\to\operatorname{U}_{\infty}(C_{0}(\mathbb{R},B^{\mathrm{SW}})). By Proposition 4.3, Lemma 6.4 and K0(C0(,Cu(||)))K_{0}(C_{0}(\mathbb{R},C^{\ast}_{u}(|\mathbb{Z}|)))\cong\mathbb{Z}, there exists an infinite loop map f:U()U(C0(,Cu(||)))f\colon\operatorname{U}_{\infty}(\mathbb{C})\to\operatorname{U}_{\infty}(C_{0}(\mathbb{R},C^{\ast}_{u}(|\mathbb{Z}|))) which induces an isomorphism on π2i1\pi_{2i-1} for any i1i\geq 1. It follows from this and Lemma 6.2 that the composite

U()𝑓U(C0(,Cu(||)))U(C0(,BSW))\operatorname{U}_{\infty}(\mathbb{C})\xrightarrow{f}\operatorname{U}_{\infty}(C_{0}(\mathbb{R},C^{\ast}_{u}(|\mathbb{Z}|)))\to\operatorname{U}_{\infty}(C_{0}(\mathbb{R},B^{\mathrm{SW}}))

is a homotopy equivalence. Then the inclusion of based loop spaces U(C0(,Cu(||)))U(C0(,BSW))\operatorname{U}_{\infty}(C_{0}(\mathbb{R},C^{\ast}_{u}(|\mathbb{Z}|)))\to\operatorname{U}_{\infty}(C_{0}(\mathbb{R},B^{\mathrm{SW}})) admits a homotopy section. This implies that the inclusion of the double loop space U(C0(2,Cu(||)))U(C0(2,BSW))\operatorname{U}_{\infty}(C_{0}(\mathbb{R}^{2},C^{\ast}_{u}(|\mathbb{Z}|)))\to\operatorname{U}_{\infty}(C_{0}(\mathbb{R}^{2},B^{\mathrm{SW}})) also admits a homotopy section. Thus the inclusion U(Cu(||))U(BSW)\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}|))\to\operatorname{U}_{\infty}(B^{\mathrm{SW}}) admits a homotopy section by Bott periodicity, which is again an infinite loop map. ∎

Proof of Theorem 1.2.

By Proposition 6.5, we have a homotopy equivalence

U(Cu(||))U(BSW)×F1\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}|))\simeq\operatorname{U}_{\infty}(B^{\mathrm{SW}})\times F_{1}

as infinite loop spaces. The homotopy types of the spaces U(BSW)\operatorname{U}_{\infty}(B^{\mathrm{SW}}) and F1F_{1} are determined in Lemma 5.2 and Proposition 6.3, respectively. Together with the homotopy stability in Theorem 1.1, this completes the proof of the theorem. ∎

7. Generalization

In this section, we study the relation between the homotopy type of U1(Cu(|n|))\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|)) and the inclusion U1(Cu(|n|))U1(C(|n|))\operatorname{U}_{1}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\subset\operatorname{U}_{1}(C^{\ast}(|\mathbb{Z}^{n}|)) for general n2n\geq 2. In view of Theorem 1.2, we propose the following question.

Question 7.1.

Does the inclusion Ud(Cu(|Γ|))Ud(C(|Γ|))\operatorname{U}_{d}(C^{\ast}_{u}(|\Gamma|))\to\operatorname{U}_{d}(C^{\ast}(|\Gamma|)) admits a homotopy section? Are the homotopy groups of its homotopy fiber are rational vector spaces?

Let us see the case when Γ=n\Gamma=\mathbb{Z}^{n} in view of this question.

Lemma 7.2.

The KK-theory of the Roe algebra C(|n|)C^{\ast}(|\mathbb{Z}^{n}|) is computed as

Ki(C(|n|)){in mod 2,0in mod 2.K_{i}(C^{\ast}(|\mathbb{Z}^{n}|))\cong\begin{cases}\mathbb{Z}&\text{$i\equiv n$ mod $2$,}\\ 0&\text{$i\not\equiv n$ mod $2$.}\end{cases}
Proof.

Let

Am=(n,𝒦())mA_{m}=\ell^{\infty}(\mathbb{Z}^{n},\mathcal{K}(\mathcal{H}))\rtimes\mathbb{Z}^{m}

with respect to the action of m\mathbb{Z}^{m} (mnm\leq n) on the first mm factors of n\mathbb{Z}^{n}. Let SjS_{j} denote the shift on the jj-th factor. Then by the Pimsner–Voiculescu exact sequence

K0(Am1)\textstyle{K_{0}(A_{m-1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}1Sm\scriptstyle{1-S_{m}}K0(Am1)\textstyle{K_{0}(A_{m-1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}K0(Am)\textstyle{K_{0}(A_{m})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}K1(Am)\textstyle{K_{1}(A_{m})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}K1(Am1)\textstyle{K_{1}(A_{m-1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}K1(Am1)\textstyle{K_{1}(A_{m-1})\ignorespaces\ignorespaces\ignorespaces\ignorespaces}1Sm\scriptstyle{1-S_{m}}

for Am=Am1SmA_{m}=A_{m-1}\rtimes_{S_{m}}\mathbb{Z}, we obtain the short exact sequence

(1) 0Ki(Am1)Sm\displaystyle 0\to K_{i}(A_{m-1})_{S_{m}}\to Ki(Am)K1i(Am1)Sm0\displaystyle K_{i}(A_{m})\to K_{1-i}(A_{m-1})^{S_{m}}\to 0

for i=0,1i=0,1, where Ki(Am1)SmK_{i}(A_{m-1})_{S_{m}} and Ki(Am1)SmK_{i}(A_{m-1})^{S_{m}} denote the coinvariant and the invariant by SmS_{m}, respectively. Since A0=(n,𝒦())A_{0}=\ell^{\infty}(\mathbb{Z}^{n},\mathcal{K}(\mathcal{H})) and we have the well-known isomorphism

Ki((n,𝒦)){ni=0,0i=1,K_{i}(\ell^{\infty}(\mathbb{Z}^{n},\mathcal{K}))\cong\begin{cases}\mathbb{Z}^{\mathbb{Z}^{n}}&i=0,\\ 0&i=1,\end{cases}

where n\mathbb{Z}^{\mathbb{Z}^{n}} is the group of all \mathbb{Z}-valued sequences over n\mathbb{Z}^{n}, we obtain

Ki(Am){nmim mod 2,0im mod 2,K_{i}(A_{m})\cong\begin{cases}\mathbb{Z}^{\mathbb{Z}^{n-m}}&\text{$i\equiv m$ mod $2$,}\\ 0&\text{$i\not\equiv m$ mod $2$,}\end{cases}

by induction on mm. The lemma is just the case when m=nm=n. ∎

Together with the previous lemma, the homotopy type of U(C(|n|))\operatorname{U}_{\infty}(C^{\ast}(|\mathbb{Z}^{n}|)) is determined by the following lemma.

Lemma 7.3.

Let AA be a CC^{\ast}-algebra, where we do not require the existence of unit. Consider the following two conditions on KK-theory:

(i)Ki(A){i=0,0i=1,(ii)Ki(A){0i=0,i=1.\text{(i)}\quad K_{i}(A)\cong\begin{cases}\mathbb{Z}&i=0,\\ 0&i=1,\end{cases}\qquad\qquad\text{(ii)}\quad K_{i}(A)\cong\begin{cases}0&i=0,\\ \mathbb{Z}&i=1.\end{cases}

If (i) holds, then U(A)\operatorname{U}_{\infty}(A) has the homotopy type of U()\operatorname{U}_{\infty}(\mathbb{C}) as an infinite loop space. If (ii) holds, then U(A)\operatorname{U}_{\infty}(A) has the homotopy type of ×BU()\mathbb{Z}\times B\operatorname{U}_{\infty}(\mathbb{C}) as an infinite loop space.

Proof.

Suppose the condition (i). By Lemma 6.4, there exists a homotopy equivalence f:U()U(A)f\colon\operatorname{U}_{\infty}(\mathbb{C})\to\operatorname{U}_{\infty}(A), which is an infinite loop map. When the condition (ii) holds, apply the result for the condition (i) to the algebra C0(,A)C_{0}(\mathbb{R},A). This implies that U(C0(,A))\operatorname{U}_{\infty}(C_{0}(\mathbb{R},A)) is homotopy equivalent to U()\operatorname{U}_{\infty}(\mathbb{C}). By the Bott periodicity, U(A)\operatorname{U}_{\infty}(A) is homotopy equivalent to ΩU()×BU()\Omega\operatorname{U}_{\infty}(\mathbb{C})\simeq\mathbb{Z}\times B\operatorname{U}_{\infty}(\mathbb{C}). ∎

Proposition 7.4.

The inclusion U(Cu(|n|))U(C(|n|))\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\to\operatorname{U}_{\infty}(C^{\ast}(|\mathbb{Z}^{n}|)) admits a homotopy section as an infinite loop map if and only if the homomorphism K(Cu(|n|))K(C(|n|))K_{\ast}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\to K_{\ast}(C^{\ast}(|\mathbb{Z}^{n}|)) is surjective.

Proof.

The only if part is obvious. For the if part, when nn is odd, this follows from Lemma 7.2 and the same argument as in the proof of Proposition 6.5. When nn is even, apply the same argument to the map on the based loop spaces U(C0(,Cu(|n|)))U(C0(,C(|n|)))\operatorname{U}_{\infty}(C_{0}(\mathbb{R},C^{\ast}_{u}(|\mathbb{Z}^{n}|)))\to\operatorname{U}_{\infty}(C_{0}(\mathbb{R},C^{\ast}(|\mathbb{Z}^{n}|))). Then the proposition follows from the existence of the homotopy section of the map on the double loop spaces U(C0(2,Cu(|n|)))U(C0(2,C(|n|)))\operatorname{U}_{\infty}(C_{0}(\mathbb{R}^{2},C^{\ast}_{u}(|\mathbb{Z}^{n}|)))\to\operatorname{U}_{\infty}(C_{0}(\mathbb{R}^{2},C^{\ast}(|\mathbb{Z}^{n}|))) and the Bott periodicity. ∎

Now all we have to do is to see that the homomorphism K(Cu(|n|))K(C(|n|))K_{\ast}(C^{\ast}_{u}(|\mathbb{Z}^{n}|))\to K_{\ast}(C^{\ast}(|\mathbb{Z}^{n}|)) is surjective. Let

Bm=(n,)mB_{m}=\ell^{\infty}(\mathbb{Z}^{n},\mathbb{C})\rtimes\mathbb{Z}^{m}

with respect to the action m\mathbb{Z}^{m} (mnm\leq n) on the first mm factors of n\mathbb{Z}^{n} and SjS_{j} denote the shift on the jj-th factor. We obtain the short exact sequences similar to (1)

(2) 0Ki(Bm1)Sm\displaystyle 0\to K_{i}(B_{m-1})_{S_{m}}\to Ki(Bm)K1i(Bm1)Sm0\displaystyle K_{i}(B_{m})\to K_{1-i}(B_{m-1})^{S_{m}}\to 0

for i=0,1i=0,1. For n=1,2n=1,2, we can see the surjectivity as follows.

Lemma 7.5.

The homomorphism K1(Cu(||))K1(C(||))K_{1}(C^{\ast}_{u}(|\mathbb{Z}|))\to K_{1}(C^{\ast}(|\mathbb{Z}|)) is an isomorphism.

Proof.

Consider the commutative square

K1(Cu(||))\textstyle{K_{1}(C^{\ast}_{u}(|\mathbb{Z}|))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}\scriptstyle{\cong}(,)S\textstyle{\ell^{\infty}(\mathbb{Z},\mathbb{Z})^{S}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}\scriptstyle{\cong}K1(C(||))\textstyle{K_{1}(C^{\ast}(|\mathbb{Z}|))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}\scriptstyle{\cong}()S\textstyle{(\mathbb{Z}^{\mathbb{Z}})^{S}}

obtained from the exact sequences (1) and (2). Thus the lemma follows. ∎

Lemma 7.6.

The homomorphism K0(Cu(|2|))K0(C(|2|))K_{0}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\to K_{0}(C^{\ast}(|\mathbb{Z}^{2}|)) is surjective.

Proof.

When n=2n=2, we can compute K(B1)K_{\ast}(B_{1}) by the exact sequence (2) as follows:

Ki(B1){(2,)S1i=0,(2,)S1i=1.K_{i}(B_{1})\cong\begin{cases}\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})_{S_{1}}&i=0,\\ \ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})^{S_{1}}&i=1.\end{cases}

Again by the exact sequences (1) and (2) for m=2m=2, we have the commutative diagram

0\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}(2,)S1S2\textstyle{\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})_{S_{1}S_{2}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}K0(Cu(|2|))\textstyle{K_{0}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}(2,)S1S2\textstyle{\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})^{S_{1}S_{2}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces\ignorespaces}\scriptstyle{\cong}0\textstyle{0}0\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}0\textstyle{0\ignorespaces\ignorespaces\ignorespaces\ignorespaces}K0(C(|2|))\textstyle{K_{0}(C^{\ast}(|\mathbb{Z}^{2}|))\ignorespaces\ignorespaces\ignorespaces\ignorespaces}\scriptstyle{\cong}(2)S1S2\textstyle{(\mathbb{Z}^{\mathbb{Z}^{2}})^{S_{1}S_{2}}\ignorespaces\ignorespaces\ignorespaces\ignorespaces}0\textstyle{0}

Thus the homomorphism K0(Cu(|2|))K0(C(|2|))K_{0}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\to K_{0}(C^{\ast}(|\mathbb{Z}^{2}|)) is surjective by the right square. ∎

To determine the homotopy type of Cu(|2|)C^{\ast}_{u}(|\mathbb{Z}^{2}|), we also need its KK-theory.

Lemma 7.7.

The KK-theory K1(Cu(|2|))K_{1}(C^{\ast}_{u}(|\mathbb{Z}^{2}|)) and the kernel of the homomorphism K0(Cu(|2|))K0(C(|2|))K_{0}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\to K_{0}(C^{\ast}(|\mathbb{Z}^{2}|)) are rational vector spaces of uncountable dimension.

Proof.

As seen in the proof of Lemma 7.6, the latter group is isomorphic to (2,)S1S2\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})_{S_{1}S_{2}}. The coinvariant (2,)S1\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})_{S_{1}} can be seen to be a rational vector space of uncountable dimension by the same argument as in [KKT, Section 5]. Then, since S2:(2,)S1(2,)S1S_{2}\colon\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})_{S_{1}}\to\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})_{S_{1}} is a linear map on a rational vector space, the coinvariant (2,)S1S2\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})_{S_{1}S_{2}} is a rational vector space of uncountable dimension. For K1(Cu(|2|))K_{1}(C^{\ast}_{u}(|\mathbb{Z}^{2}|)), we obtain the exact sequence

0((2,)S1)S2K1(Cu(|2|))((2,)S1)S200\to(\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})^{S_{1}})_{S_{2}}\to K_{1}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\to(\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})_{S_{1}})^{S_{2}}\to 0

from (2). Since ((2,)S1)S2(,)S(\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})^{S_{1}})_{S_{2}}\cong\ell^{\infty}(\mathbb{Z},\mathbb{Z})_{S} and (2,)S1\ell^{\infty}(\mathbb{Z}^{2},\mathbb{Z})_{S_{1}} are rational vector spaces, K1(Cu(|2|))K_{1}(C^{\ast}_{u}(|\mathbb{Z}^{2}|)) is also a rational vector space of uncountable dimension. ∎

Proof of Theorem 1.3.

By Proposition 7.4 and Lemma 7.6, the inclusion U(Cu(|2|))U(C(|2|))\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\to\operatorname{U}_{\infty}(C^{\ast}(|\mathbb{Z}^{2}|)) admits a homotopy section as an infinite loop map. Let F2F_{2} be the homotopy fiber of the inclusion. Then we have a homotopy equivalence

U(Cu(|2|))U(C(|2|))×F2\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\simeq\operatorname{U}_{\infty}(C^{\ast}(|\mathbb{Z}^{2}|))\times F_{2}

as infinite loop spaces. By Lemmas 7.2 and 7.3, U(C(|2|))\operatorname{U}_{\infty}(C^{\ast}(|\mathbb{Z}^{2}|)) is homotopy equivalent to U()\operatorname{U}_{\infty}(\mathbb{C}) as an infinite loop space. By the naturality of the Bott maps

β:U(Cu(|2|))U(C0(2,Cu(|2|)))andβ:U(C(|2|))U(C0(2,C(|2|))),\beta\colon\operatorname{U}_{\infty}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\xrightarrow{\simeq}\operatorname{U}_{\infty}(C_{0}(\mathbb{R}^{2},C^{\ast}_{u}(|\mathbb{Z}^{2}|)))\quad\text{and}\quad\beta\colon\operatorname{U}_{\infty}(C^{\ast}(|\mathbb{Z}^{2}|))\xrightarrow{\simeq}\operatorname{U}_{\infty}(C_{0}(\mathbb{R}^{2},C^{\ast}(|\mathbb{Z}^{2}|))),

we have the homotopy equivalence

β~:F2Ω2F2\tilde{\beta}\colon F_{2}\xrightarrow{\simeq}\Omega^{2}F_{2}

as well. The homotopy group of F2F_{2} can be computed by Lemma 7.6:

πi(F2){V1i is even,V0i is odd,\pi_{i}(F_{2})\cong\begin{cases}V_{1}&\text{$i$ is even,}\\ V_{0}&\text{$i$ is odd,}\end{cases}

where

V0=ker[K0(Cu(|2|))K0(C(|2|))],V1=K1(Cu(|2|))V_{0}=\ker[K_{0}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))\to K_{0}(C^{\ast}(|\mathbb{Z}^{2}|))],\quad V_{1}=K_{1}(C^{\ast}_{u}(|\mathbb{Z}^{2}|))

are rational vector spaces by Lemma 7.7. Again as in the proof of [KKT, Lemma 5.4], we can find maps

i1K(V0,2i1)F2andi1K(V1,2i1)ΩF2\prod_{i\geq 1}^{\circ}K(V_{0},2i-1)\to F_{2}\quad\text{and}\quad\prod_{i\geq 1}^{\circ}K(V_{1},2i-1)\to\Omega F_{2}

inducing isomorphisms on the odd degree homotopy groups. Then, using the homotopy equivalence β~\tilde{\beta}, we obtain the homotopy equivalence

i1(K(V0,2i1)×K(V1,2i))F2.\prod_{i\geq 1}^{\circ}(K(V_{0},2i-1)\times K(V_{1},2i))\to F_{2}.

This completes the proof of the theorem. ∎

Moreover, Lemma 7.5 provides another proof of Theorem 1.2 in a similar manner.

References

  • [Bas64] H. Bass. KK-theory and stable algebra. Inst. Hautes Études Sci. Publ. Math., (22):5–60, 1964.
  • [Bla86] Bruce Blackadar. KK-theory for operator algebras, volume 5 of Mathematical Sciences Research Institute Publications. Springer-Verlag, New York, 1986.
  • [KKT] T. Kato, D. Kishimoto, and M. Tsutaya. Homotopy type of the space of finite propagation unitary operators on \mathbb{Z}. preprint, arxiv: 2007.06787.
  • [Kui65] Nicolaas H. Kuiper. The homotopy type of the unitary group of Hilbert space. Topology, 3:19–30, 1965.
  • [MT21] Vladimir Manuilov and Evgenij Troitsky. On Kuiper type theorems for uniform Roe algebras. Linear Algebra Appl., 608:387–398, 2021.
  • [PV80] M. Pimsner and D. Voiculescu. Exact sequences for KK-groups and Ext-groups of certain cross-product CC^{\ast}-algebras. J. Operator Theory, 4(1):93–118, 1980.
  • [Rie83] Marc A. Rieffel. Dimension and stable rank in the KK-theory of CC^{\ast}-algebras. Proc. London Math. Soc. (3), 46(2):301–333, 1983.
  • [Rie87] Marc A. Rieffel. The homotopy groups of the unitary groups of noncommutative tori. J. Operator Theory, 17(2):237–254, 1987.
  • [Roe88] John Roe. An index theorem on open manifolds. I, II. J. Differential Geom., 27(1):87–113, 115–136, 1988.
  • [Roe03] John Roe. Lectures on coarse geometry, volume 31 of University Lecture Series. American Mathematical Society, Providence, RI, 2003.
  • [SW85] Graeme Segal and George Wilson. Loop groups and equations of KdV type. Inst. Hautes Études Sci. Publ. Math., (61):5–65, 1985.