, ,

Following up the compact metric spaces given by Connes in connection with his theory of quantum Riemannian geometry defined by Dirac operators [3] , Rieffel defined the notion of a compact quantum metric space

(A, L_{A})

in [17] as an order unit space

A

equipped with a Lip-norm

L_{A}

, which is a generalization of the usual Lipschitz seminorm on functions which one associates to an ordinary metric. Many interesting examples of compact quantum metric space have been constructed [14, 16, 11, 10] . Motivated by the type of convergence of spaces that has recently begun to play a central role in string theory, Rieffel introduces the quantum Gromov-Hausdorff distance for the compact quantum metric spaces as a quantum analogue of Gromov-Hausdorff distance, and shows that the basic theorems of the classical theory have natural quantum analogues.

In [20] and [21] , we formulated matrix Lipschitz seminorms on matrix order unit spaces. This operator space version of Lipschitz seminorm has many nice properties which are similar to those for ordinary metric spaces. These data may then be thought of as some `noncommutative metric spaces'. So it is natural to ask, as does Rieffel in [17] , if it is possible to develop a corresponding operator space version of quantum Gromov-Hausdorff distance. This is the aim of the present article.

In contrast to the matricial quantum Gromov-Hausdorff distance in [8] and operator Gromov-Hausdorff distance in [9] , our quantized Gromov-Hausdorff distance operates entirely at the “matrix” level. Not only the matrix state spaces but also the matrix Lipschitz seminorms and the complete isometries are brought into our picture. This should be important in the background of operator systems.

The paper has eight sections. Section 2 contains preliminaries, mainly to fix some terminology and notation. In Section 3 we define quantized metric space and develop an operator “quotient”. Section 4 defines our quantized Gromov-Hausdorff distance, and we prove that it satisfies the triangle inequality. Section 5 deals with the operator Gromov-Hausdorff distance zero. We establish that it implies a complete isometry. Section 6 treats the completeness theorem of the complete isometry classes of quantized metric spaces. In Section 7 we show that a quantized metric space with 1-exact underlying matrix order unit space is a limit of matrix algebras with respect to quantized Gromov-Hausdorff distance. It is established in Section 8 that matrix algebras converge naturally to the sphere for quantized Gromov-Hausdorff distance.

2 Preliminaries

All vector spaces are assumed to be complex throughout this paper. Given a vector space

V

, we let

M_{m, n} (V)

denote the matrix space of all

m

n

matrices

v = [v_{i j}]

with

v_{i j} \in V

, and we set

M_{n} (V) = M_{n, n} (V)

. If

V = C

, we write

M_{m, n} = M_{m, n} (C)

and

M_{n} = M_{n, n} (C)

, which means that we may identify

M_{m, n} (V)

with the tensor product

M_{m, n} \otimes V

. We identify

M_{m, n}

with the normed space

ℬ (C^{n}, C^{m})

We use the standard matrix multiplication and *-operation for compatible scalar matrices, and

1_{n}

for the identity matrix in

M_{n}

There are two natural operations on the matrix spaces. For

v \in M_{m, n} (V)

and

w \in M_{p, q} (V)

, the direct sum

v \oplus w \in M_{m + p, n + q} (V)

is defined by letting

v \oplus w = [\begin{matrix} v & 0 \\ 0 & w \end{matrix}],

and if we are given

α \in M_{m, p}

v \in M_{p, q} (V)

and

β \in M_{q, n}

, the matrix product

α v β \in M_{m, n} (V)

is defined by

α v β = [\sum_{k, l} α_{i k} v_{k l} β_{l j}] .

A *-vector space

V

is a complex vector space together with a conjugate linear mapping

v \mapsto v^{*}

such that

v^{* *} = v

. A complex vector space

V

is said to be matrix ordered if:

(1) $V$ is a *-vector space;
(2) each $M_{n} (V)$ , $n \in N$ , is partially ordered;
(3) $γ^{*} M_{n} (V)^{+} γ \subseteq M_{m} (V)^{+}$ if $γ = [γ_{i j}]$ is any $n \times m$ matrix of complex numbers.

A matrix order unit space

(V, 1)

is a matrix ordered space

V

together with a distinguished order unit

1

satisfying the following conditions:

(1) $V^{+}$ is a proper cone with the order unit $1$ ;
(2) each of the cones $M_{n} (V)^{+}$ is Archimedean.

Each matrix order unit space

(V, 1)

may be provided with the norm

∥ v ∥ = inf {t \in R : [\begin{matrix} t 1 & v \\ v^{*} & t 1 \end{matrix}] \geq 0} .

As in [17] , we will not assume that

V

is complete for the norm.

V

and

W

are *-vector spaces and

φ : V \mapsto W

is a linear mapping, we have a linear mapping

φ^{*} : V \mapsto W

defined by

φ^{*} (v) = φ (v^{*})^{*}

Given vector spaces

V

and

W

and a linear mapping

φ : V \mapsto W

and

n \in N

, we have a corresponding

φ_{n} : M_{n} (V) \mapsto M_{n} (W)

defined by

φ_{n} ([v_{i j}]) = [φ (v_{i j})] .

V

and

W

are vector spaces in duality, then they determine the matrix pairing

≪ \cdot, \cdot ≫ : M_{n} (V) \times M_{m} (W) \mapsto M_{n m},

where

≪ [v_{i j}], [w_{k l}] ≫ = [< v_{i j}, w_{k l} >]

for

[v_{i j}] \in M_{n} (V)

and

[w_{k l}] \in M_{m} (W)

A graded set

S = (S_{n})

is a sequence of sets

S_{n} (n \in N)

. If

V

is a locally convex topological vector space, then the canonical topology on

M_{n} (V) (n \in N)

is that determined by the natural linear isomorphism

M_{n} (V) \sim = V^{n^{2}}

, that is, the product topology. A graded set

S = (S_{n})

with

S_{n} \subseteq M_{n} (V)

is closed or compact if that is the case for each set

S_{n}

in the product topology in

M_{n} (V)

. Given a vector space

V

, we say that a graded set

B = (B_{n})

with

B_{n} \subseteq M_{n} (V)

is absolutely matrix convex if for all

m, n \in N

(1) $B_{m} \oplus B_{n} \subseteq B_{m + n}$ ;
(2) $α B_{m} β \subseteq B_{n}$ for any contractions $α \in M_{n, m}$ and $β \in M_{m, n}$ .

A matrix convex set in

V

is a graded set

K = (K_{n})

of subsets

K_{n} \subseteq M_{n} (V)

such that

\sum_{i = 1}^{k} γ_{i}^{*} v_{i} γ_{i} \in K_{n}

for all

v_{i} \in K_{n_{i}}

and

γ_{i} \in M_{n_{i}, n}

for

i = 1, 2, \dots, k

satisfying

\sum_{i = 1}^{k} γ_{i}^{*} γ_{i} = 1_{n}

Let

V

and

W

be vector spaces in duality, and let

S = (S_{n})

be a graded set with

S_{n} \subseteq M_{n} (V)

. The absolute operator polar

S^{⊚} = (S_{n}^{⊚})

with

S_{n}^{⊚} \subseteq M_{n} (W)

, is defined by

S_{n}^{⊚} = {w \in M_{n} (W) : ∥ ≪ v, w ≫ ∥ \leq 1 f o r a l l v \in S_{r}, r \in N}

. The matrix polar

S^{π} = (S_{n}^{π})

with

S_{n}^{π} \subseteq M_{n} (W)

, is defined by

S_{n}^{π} = {w \in M_{n} (W) : R e ≪ v, w ≫ \leq 1_{r \times n} f o r a l l v \in S_{r}, r \in N}

. Given a subset

S \subseteq V

, the absolute polar of

S

is defined by

S^{\circ} = {w \in W : | < v, w > | \leq 1 for all v \in S}

A gauge on a vector space

V

is a function

g : V \mapsto [0, + \infty]

such that

(1) $g (v + w) \leq g (v) + g (w)$ ;
(2) $g (α v) \leq | α | g (v)$ ,

for all

v, w \in V

and

α \in C

. We say that a gauge

g

is a seminorm on

V

g (v) < + \infty

for all

v \in V

. Given an arbitrary vector space

V

, a matrix gauge

G = (g_{n})

V

is a sequence of gauges

g_{n} : M_{n} (V) \mapsto [0, + \infty]

such that

(1) $g_{m + n} (v \oplus w) = max {g_{m} (v), g_{n} (w)}$ ;
(2) $g_{n} (α v β) \leq ∥ α ∥ g_{m} (v) ∥ β ∥$ ,

for any

v \in M_{m} (V)

w \in M_{n} (V)

α \in M_{n, m}

and

β \in M_{m, n}

. A matrix gauge

G = (g_{n})

is a matrix seminorm on

V

if for any

n \in N, g_{n} (v) < + \infty

for all

v \in M_{n} (V)

If each

g_{n}

is a norm on

M_{n} (V)

, we say that

G

is a matrix norm. An operator space is a vector space together with a matrix norm on it. For a matrix order unit space

(V, 1)

, it is an operator space with the matrix norm determined by the matrix order on it.

3 Quantized metric space

First we recall the following definitions given in [20, 21] :

Definition 3.1. Given a matrix order unit space

(V, 1)

, a matrix Lipschitz seminorm

ℒ

(V, 1)

is a sequence of seminorms

L_{n} : M_{n} (V) \mapsto [0, + \infty)

such that

(1) the null space of each $L_{n}$ is $M_{n} (C 1)$ ;
(2) $L_{m + n} (v \oplus w) = max {L_{m} (v), L_{n} (w)}$ ;
(3) $L_{n} (α v β) \leq ∥ α ∥ L_{m} (v) ∥ β ∥$ ;
(4) $L_{m} (v^{*}) = L_{m} (v)$ ,

for any

v \in M_{m} (V)

w \in M_{n} (V)

α \in M_{n, m}

and

β \in M_{m, n}

Let

(V, 1)

be a matrix order unit space. The matrix state space of

(V, 1)

is the collection

C S (V) = (C S_{n} (V))

of matrix states

C S_{n} (V) = {φ : φ i s a u n i t a l c o m p l e t e l y p o s i t i v e l i n e a r m a p p i n g f r o m V i n t o M_{n}} .

ℒ = (L_{n})

is a matrix Lipschitz seminorm on

(V, 1)

, we have a collection

D_{ℒ} = (D_{L_{n}})

of metrics on

C S (V)

given by

D_{L_{n}} (φ, ψ) = sup {∥ ≪ φ, a ≫ - ≪ ψ, a ≫ ∥ : a \in M_{r} (V), L_{r} (a) \leq 1, r \in N},

for

φ, ψ \in C S_{n} (V)

(notice that it may take value

+ \infty

). And in turn we obtain a sequence

ℒ_{D_{ℒ}} = (L_{D_{L_{n}}})

of gauges on

(V, 1)

L_{D_{L_{n}}} (a) = sup {\frac{∥ ≪ φ, a ≫ - ≪ ψ, a ≫ ∥}{D_{L_{r}} (φ, ψ)} : φ, ψ \in C S_{r} (V), φ \neq ψ, r \in N},

for all

a \in M_{n} (V)

Definition 3.2. Let

(V, 1)

be a matrix order unit space. By a matrix Lip-norm on

(V, 1)

we mean a matrix Lipschitz seminorm

ℒ = (L_{n})

(V, 1)

such that the

D_{ℒ}

-topology on

C S (V)

agrees with the BW-topology.

We are now prepared to make:

Definition 3.3. By a quantized metric space we mean a pair

(V, ℒ)

consisting of a matrix order unit space

(V, 1)

with a matrix Lip-norm

ℒ

defined on it.

Example 3.4. Let

(X, ρ)

be an ordinary compact metric space, let

A

denote the set of Lipschitz functions on

X

, and let

L_{ρ}

denote the Lipschitz seminorm on

A

. Then

A \subseteq C (X)

, and for

f, g \in A

and

α \in C

, we have

L_{ρ} (f^{*}) = L_{ρ} (f), L_{ρ} (α f) = | α | L_{ρ} (f), L_{ρ} (f + g) \leq L_{ρ} (f) + L_{ρ} (g) .

Thus

A

is a self-adjoint linear subspace of

C (X)

which contains constant functions, and so

A

is a matrix order unit space by Theorem 4.4 in [2] .

Since

L_{ρ}

is lower semicontinuous,

K = {f \in A : L_{ρ} (f) \leq 1}

is an absolutely convex normed-closed (and hence is weakly closed) set in

A

K

determines a graded set

K_{n} = {\begin{matrix} K, & i f n = 1, \\ {0}, & i f n > 1 . \end{matrix}

The minimal envelope

\hat{K}

K

is the matrix bipolar

K^{⊚ ⊚}

K

\hat{K}

is an absolutely matrix convex weakly closed graded set. We let

\hat{ℒ} = ({\hat{L}}_{n})

be the corresponding matrix gauge of

\hat{K}

. Since

{\hat{L}}_{1} = L_{ρ}

is a Lipschitz seminorm,

\hat{ℒ}

is a matrix Lipschitz seminorm.

ρ_{L_{ρ}} = ρ

implies that

\hat{ℒ}

is also a matrix Lip-norm(see Theorem 1.9 in [14] and Proposition 7.5 in [21] ). Therefore,

(A, \hat{ℒ})

is a quantized metric space. It is called the minimal quantized metric space of

(X, ρ)

. The maximal envelope

\overset{ˇ}{K}

K

is the matrix polar

(K^{\circ})^{⊚}

K^{\circ} = (K_{n}^{\circ})

, where

K_{n}^{\circ} = {\begin{matrix} K^{\circ}, & i f n = 1, \\ {0}, & i f n > 1 . \end{matrix}

Similarly,

\overset{ˇ}{K}

is an absolutely matrix convex weakly closed graded set, and the corresponding matrix gauge

\overset{ˇ}{ℒ}

\overset{ˇ}{K}

makes

A

into a quantized metric space.

(A, \overset{ˇ}{ℒ})

is called the maximal quantized metric space of

(X, ρ)

Moreover, if

C = (C_{n})

is an absolutely matrix convex weakly closed graded set with

C_{1} = K

, then

\hat{K} \subseteq C \subseteq \overset{ˇ}{K},

and the corresponding matrix gauge

ℒ = (L_{n})

satisfies

{\overset{ˇ}{L}}_{n} \leq L_{n} \leq {\hat{L}}_{n}, n \in N,

(see page 181 in [4] ). So

(A, ℒ)

is a quantized metric space. It is called a quantized metric space of

(X, ρ)

Example 3.5. Let

(A, L)

be a compact quantum metric space, that is, an order unit space

(A, e)

equipped with a seminorm

L

, called Lip-norm, on

A

such that

L (a) = 0

if and only if

a \in R e

, and the topology on the state space

S (A)

A

from the metric

ρ_{L} (μ, ν) = sup {| μ (a) - ν (a) | : L (a) \leq 1}

is the

w^{*}

-topology (see Definition 2.2 in [17] ). So

(S (A), ρ_{L})

is an ordinary compact metric space. Let

A

denote the set of Lipschitz functions on

S (A)

. By Example 3.4 , there exists a quantized metric space structure

(A, ℒ_{1})

(S (A), ρ_{L})

, where

ℒ_{1} = (L_{1, n})

. From Lemma 3.2 in [15] ,

A \subseteq A

and

L_{1, 1} (a) \leq L (a)

for

a \in A

. Let

∥ \cdot ∥ = (∥ \cdot ∥_{n})

be the matrix norm determined by the matrix order on

(A, 1)

. By the basic representation theorem of Kadison[7] , we also have that

∥ a ∥ = ∥ a ∥_{1}

for

a \in A

. If

L

is lower semicontinuous, the embedding of

A

into

A

is isometric, that is,

∥ a ∥ = ∥ a ∥_{1}

and

L (a) = L_{1, 1} (a)

for all

a \in A

, according to Theorem 4.1 in [15] .

Set

V = A \cap (A + i A) .

We denote the restriction of

ℒ_{1}

V

ℒ = (L_{n})

. Then

V

is a self-adjoint linear subspace of

A

and contains the order unit of

A

. So

V

is a matrix order unit space. Because the

D_{ℒ_{1}}

-topology on

C S (A)

agrees with the BW-topology, the image of

L_{1, 1}^{1} = {a \in A : L_{1, 1} (a) \leq 1}

\tilde{A} = A / C 1

is totally bounded for

∥ \cdot ∥_{1}^{\sim}

by Theorem 5.3 in [20] . Since

L_{1}^{1} \subseteq L_{1, 1}^{1}

, the image of

L_{1}^{1}

\tilde{V} \subseteq \tilde{A}

is totally bounded for

∥ \cdot ∥_{1}^{\sim}

, and so, by Theorem 5.3 in [20] , the

D_{ℒ}

-topology on

C S (V)

is the BW-topology. Therefore,

(V, ℒ)

is a quantized metric space, and the embedding of

(A, L)

into

(V, ℒ)

is an isometry if

L

is lower semicontinuous.

Let

(V, 1)

and

(W, 1)

be matrix order unit spaces, and let

φ : V \mapsto W

be a unital completely positive linear mapping. Then we have the dual mapping

φ^{'} : W^{*} \mapsto V^{*}

determined by

φ^{'} (f) (v) = f (φ (v))

. Let

{\tilde{V}}^{*}

denote the dual space of

\tilde{V} = V / (C 1)

{\tilde{V}}^{*}

is just the subspace of

V^{*}

consisting of those

f \in V^{*}

such that

f (1) = 0

. For any

v \in M_{n} (V)

and

g \in M_{m} (W^{*})

, we have

≪ g, φ_{n} (v) ≫ = [g_{k l} (φ (v_{i j}))] = [(φ^{'} (g_{k l})) (v_{i j})] = ≪ (φ^{'})_{m} (g), v ≫ .

∥ φ ∥_{c b} = 1

(φ^{'})_{m} (M_{m} ({\tilde{W}}^{*})) \subseteq M_{m} ({\tilde{V}}^{*})

and

(φ^{'})_{m} (C S_{m} (W)) \subseteq C S_{m} (V)

Moreover,

φ^{'}

w^{*}

-continuous. Let

φ_{n}^{c} = (φ^{'})_{n} |_{C S_{n} (W)}

for

n \in N

. Then for

v \in M_{n} (V)

f_{i} \in C S_{n_{i}} (W)

and

γ_{i} \in M_{n_{i}, m}

satisfying

\sum_{i = 1}^{k} γ_{i}^{*} γ_{i} = 1_{m}

, we have

\begin{matrix} ≪ φ_{m}^{c} (\sum_{i = 1}^{k} γ_{i}^{*} f_{i} γ_{i}), v ≫ & = & ≪ \sum_{i = 1}^{k} γ_{i}^{*} f_{i} γ_{i}, φ_{n} (v) ≫ \end{matrix}

\begin{matrix} = & \sum_{i = 1}^{k} (γ_{i} \otimes 1_{n})^{*} ≪ f_{i}, φ_{n} (v) ≫ (γ_{i} \otimes 1_{n}) \end{matrix}

\begin{matrix} = & \sum_{i = 1}^{k} (γ_{i} \otimes 1_{n})^{*} ≪ φ_{n_{i}}^{c} (f_{i}), v ≫ (γ_{i} \otimes 1_{n}) \end{matrix}

\begin{matrix} = & ≪ \sum_{i = 1}^{k} γ_{i}^{*} φ_{n_{i}}^{c} (f_{i}) γ_{i}, v ≫ . \end{matrix}

φ^{c} = (φ_{n}^{c})

is a BW-continuous matrix affine mapping of

C S (W)

into

C S (V)

. In particular,

φ^{c} (C S (W)) = (φ_{n}^{c} (C S_{n} (W)))

is a closed matrix convex subset of

C S (V)

Clearly

φ_{n}^{c}

is injective if

φ

is surjective.

Let

ℒ

be a matrix Lipschitz seminorm on

V

. On

V^{*}

, we define the matrix gauge

ℒ^{'} = (L_{n}^{'})

L_{n}^{'} (f) = sup {∥ ≪ f, a ≫ ∥ : a \in L_{r}^{1}, r \in N}, f \in M_{n} (V^{*}) .

Then

L_{n}^{'} (φ - ψ) = D_{L_{n}} (φ, ψ)

for

φ, ψ \in C S_{n} (V)

by Lemma 4.3 in [21] .

Proposition 3.6. Let

(V, 1)

and

(W, 1)

be matrix order unit spaces, and let

φ : V \mapsto W

be a unital completely positive linear mapping which is surjective. Let

ℒ

be a matrix Lipschitz seminorm on

V

, and let

ℒ_{W} = (L_{W, n})

be a sequence of the corresponding quotient seminorms on

W

, defined by

L_{W, n} (b) = inf {L_{n} (a) : φ_{n} (a) = b}, b \in M_{n} (W) .

Then

(1) $ℒ_{W}$ is a matrix Lipschitz seminorm on $W$ ;
(2) $φ^{'}$ is a complete isometry for the matrix norms $ℒ_{W}^{'}$ and $ℒ^{'}$ on ${\tilde{W}}^{*}$ and ${\tilde{V}}^{*}$ ;
(3) $φ^{c}$ is a complete isometry for the corresponding matrix metrics $D_{ℒ_{W}}$ and $D_{ℒ}$ ;
(4) If $ℒ$ is a matrix Lip-norm, then so is $ℒ_{W}$ .

Proof. (1) For

b_{1} \in M_{m} (W), b_{2} \in M_{n} (W)

, we have

\begin{matrix} L_{W, m + n} (b_{1} \oplus b_{2}) \end{matrix}

\begin{matrix} = & inf {L_{m + n} (a) : φ_{m + n} (a) = b_{1} \oplus b_{2}} \end{matrix}

\begin{matrix} \leq & inf {L_{m + n} (a_{1} \oplus a_{2}) : φ_{m + n} (a_{1} \oplus a_{2}) = b_{1} \oplus b_{2}} \end{matrix}

\begin{matrix} = & inf {max {L_{m} (a_{1}), L_{n} (a_{2})} : φ_{m} (a_{1}) = b_{1}, φ_{n} (a_{2}) = b_{2}} \end{matrix}

\begin{matrix} = & max {inf {L_{m} (a_{1}) : φ_{m} (a_{1}) = b_{1}}, inf {L_{n} (a_{2}) : φ_{n} (a_{2}) = b_{2}}} \end{matrix}

\begin{matrix} = & max {L_{W, m} (b_{1}), L_{W, n} (b_{2})} . \end{matrix}

α \in M_{m, n}

β \in M_{n, m}

and

b \in M_{n} (W)

, we have

\begin{matrix} L_{W, m} (α b β) & = & inf {L_{m} (a) : φ_{m} (a) = α b β} \end{matrix}

\begin{matrix} \leq & inf {L_{m} (α a β) : φ_{n} (a) = b} \end{matrix}

\begin{matrix} \leq & ∥ α ∥ ∥ β ∥ inf {L_{n} (a) : φ_{n} (a) = b} \end{matrix}

\begin{matrix} = & ∥ α ∥ ∥ β ∥ L_{W, n} (b), \end{matrix}

and

\begin{matrix} L_{W, n} (b^{*}) & = & inf {L_{n} (a) : φ_{n} (a) = b^{*}} \end{matrix}

\begin{matrix} = & inf {L_{n} (a) : φ_{n} (a^{*}) = b} \end{matrix}

\begin{matrix} = & inf {L_{n} (a^{*}) : φ_{n} (a) = b} \end{matrix}

\begin{matrix} = & inf {L_{n} (a) : φ_{n} (a) = b} \end{matrix}

\begin{matrix} = & L_{W, n} (b) . \end{matrix}

Given

[λ_{i j}] \in M_{n}

. We have

\begin{matrix} L_{W, n} ([λ_{i j} 1]) & = & inf {L_{n} (a) : φ_{n} (a) = [λ_{i j} 1]} \end{matrix}

\begin{matrix} \leq & L_{n} ([λ_{i j} 1]) = 0, \end{matrix}

and so

L_{W, n} ([λ_{i j} 1]) = 0

. If

b = [b_{i j}] \in M_{n} (W)

with

L_{W, n} (b) = 0

, then

L_{W, 1} (b_{i j}) = 0, i, j = 1, 2, \dots, n .

Letting

b_{i j} = c_{i j} + i d_{i j} (i, j = 1, 2, \dots, n)

, where

c_{i j}^{*} = c_{i j}, d_{i j}^{*} = d_{i j}

, we get

L_{W, 1} (c_{i j}) = 0, L_{W, 1} (d_{i j}) = 0, i, j = 1, 2, \dots, n .

Since

L_{W, 1} (c_{i j}) = inf {L_{1} (a) : φ_{1} (a) = c_{i j}}

and

φ

is positive, we have

L_{W, 1} (c_{i j}) = inf {L_{1} (a) : φ_{1} (a) = c_{i j}, a = a^{*}} .

Now by Proposition 3.1 in [17] , there exists an

α_{i j} \in R

such that

c_{i j} = α_{i j} 1

Similarly, there exists a

β_{i j} \in R

such that

d_{i j} = β_{i j} 1

. Therefore,

b \in M_{n} (C 1)

Thus

ℒ_{W}

is a matrix Lipschitz seminorm on

W

(2) Let

f \in M_{m} ({\tilde{W}}^{*})

. For any

a \in M_{n} (V)

we clearly have

L_{W, n} (φ_{n} (a)) \leq L_{n} (a)

, and so if

L_{n} (a) \leq 1

we have

∥ ≪ (φ^{'})_{m} (f), a ≫ ∥ = ∥ ≪ f, φ_{n} (a) ≫ ∥ \leq L_{W, m}^{'} (f) .

Consequently,

L_{m}^{'} ((φ^{'})_{m} (f)) \leq L_{W, m}^{'} (f)

. But let

δ > 0

be given, and let

b \in M_{n} (W)

with

L_{W, m} (b) \leq 1

. Then there is an

a \in M_{n} (V)

such that

φ_{n} (a) = b

and

L_{n} (a) \leq 1 + δ

. Thus,

L_{n} (a / (1 + δ)) \leq 1

. Consequently,

\begin{matrix} L_{m}^{'} ((φ^{'})_{m} (f)) & \geq & ∥ ≪ (φ^{'})_{m} (f), a / (1 + δ) ≫ ∥ \end{matrix}

\begin{matrix} = & ∥ ≪ f, φ_{n} (a) ≫ ∥ / (1 + δ) \end{matrix}

\begin{matrix} = & ∥ ≪ f, b ≫ ∥ / (1 + δ) . \end{matrix}

Taking the supremum over

b \in M_{n} (W)

with

L_{W, n} (b) \leq 1

, we see that

L_{m}^{'} ((φ^{'})_{m} (f)) \geq L_{W, m}^{'} (f) / (1 + δ) .

Since

δ

is arbitrary, we obtain that

L_{m}^{'} ((φ^{'})_{m} (f)) \geq L_{W, m}^{'} (f)

. Thus

φ^{'}

is a complete isometry.

(3) By (2), we have

\begin{matrix} D_{L_{W, n}} (φ, ψ) & = & L_{W, n}^{'} (φ - ψ) \end{matrix}

\begin{matrix} = & L_{n}^{'} ((φ^{'})_{n} (φ - ψ)) \end{matrix}

\begin{matrix} = & D_{L_{n}} (φ_{n}^{c} (φ), φ_{n}^{c} (ψ)), \end{matrix}

where

φ, ψ \in C S_{n} (W)

, that is,

φ^{c}

is a complete isometry for the corresponding matrix metrics

D_{ℒ_{W}}

and

D_{ℒ}

. (4) Suppose that

ℒ

is a matrix Lip-norm. Since

φ^{'}

w^{*}

-continuous,

φ

is surjective, and

C S_{n} (V)

is BW-compact,

(φ^{'})_{n}

is a homeomorphism of

C S_{n} (W)

onto

(φ^{'})_{n} (C S_{n} (W)) \subseteq C S_{n} (V)

. Because

D_{L_{n}}

gives the BW-topology on

C S_{n} (V)

{D_{L_{n}} |}_{(φ^{'})_{n} (C S_{n} (W))}

gives the relative topology of

(φ^{'})_{n} (C S_{n} (W))

. According to (3),

D_{L_{W, n}}

gives the BW-topology on

C S_{n} (W)

. Therefore,

ℒ_{W}

is a matrix Lip-norm. □

Notation 3.7. Under the conditions of Propositions 3.6 we will say that

ℒ

induces

ℒ_{W}

via

φ

For a matrix convex set

K

in a locally convex vector space, let

A (K)

be the set of all matrix affine mappings from

K

C

(see §6 in [21] ). On the other hand, we have

Proposition 3.8. Let

(V, 1)

be a matrix order unit space, and let

K = (K_{n})

be a compact matrix convex subset of

C S (V)

. View the elements of

V

as matrix affine mapping from

C S (V)

C

(Proposition 6.1 in [21] ), and let

W

consist of their restrictions to

K

, with

φ

the restriction mapping of

V

onto

W

. Then

(W, φ (1))

is a matrix order unit space, and

K = φ^{c} (C S (W))

Proof. Clearly, with the natural matrix order structure on $W$ and the order unit $φ (1)$ , $(W, φ (1))$ is a matrix order unit space.
For $φ \in K_{n} \subseteq C S_{n} (V)$ , we define the mapping $ψ : W \mapsto M_{n}$ by $ψ (φ (v)) = φ (v)$ .
Then $ψ \in C S_{n} (W)$ and $(φ_{n}^{c} (ψ)) (v) = ψ (φ (v)) = φ (v)$ for $v \in V$ . Thus $K_{n} \subseteq φ_{n}^{c} (C S_{n} (W))$ .
Suppose that $φ_{0} \in C S_{n} (V)$ and $φ_{0} / \in K_{n}$ . By Theorem 1.6 in [19] , there is a $v = [v_{i j}] \in M_{n} (V)$ and a self-adjoint $α = [α_{i j}] \in M_{n}$ such that $R e ≪ φ, v ≫ \leq α \otimes 1_{r},$ for all $r \in N$ , $φ \in K_{r}$ , and $R e ≪ φ_{0}, v ≫ ≰ α \otimes 1_{n} .$ So we obtain $φ_{n} (R e [α_{i j} 1 - v_{i j}]) \geq 0$ for all $r \in N$ and $φ \in K_{r}$ . Thus $φ_{n} (R e [α_{i j} 1 - v_{i j}]) \geq 0$ in $W$ . If $φ_{0} = φ_{n}^{c} (ψ_{0})$ for some $ψ_{0} \in C S_{n} (W)$ , we would then have that $R e ≪ φ_{0}, v ≫ = R e ≪ ψ_{0}, φ_{n} (v) ≫ = α \otimes 1_{n} - R e ≪ ψ_{0}, φ_{n} ([α_{i j} 1 - v_{i j}]) ≫ = α \otimes 1_{n} - ≪ ψ_{0}, φ_{n} (R e [α_{i j} 1 - v_{i j}]) ≫ \leq α \otimes 1_{n}$ . Thus, $φ_{0} / \in φ^{c} (C S_{n} (W))$ . Therefore, $K = φ^{c} (C S (W))$ . □

Notation 3.9. We will call the matrix order unit space

(W, φ (1))

in the Proposition 3.8 the quotient of

(V, 1)

with respect to

K

, and will identify

C S (W)

with

K

. When

(V, ℒ)

is a quantized metric space,

(W, ℒ_{W})

is a quantized metric space by Proposition 3.6 .

(W, ℒ_{W})

is called the quotient space of

(V, ℒ)

with respect to

K

and

φ

Proposition 3.10. Let

(V_{1}, 1)

(V_{2}, 1)

and

(V_{3}, 1)

be matrix order unit spaces. Suppose that

φ : V_{1} \mapsto V_{2}

and

ψ : V_{2} \mapsto V_{3}

are unital completely positive linear mappings which are surjective. Denote

φ = ψ \circ φ

. If

ℒ

is a matrix Lipschitz seminorm on

V_{1}

ℒ_{V_{2}}

and

ℒ_{V_{3}}

are the induced matrix Lipschitz seminorms of

ℒ

via

φ

and

φ

, respectively, and

ℒ_{V_{23}}

is the induced matrix Lipschitz seminorm of

ℒ_{V_{2}}

via

ψ

, then

ℒ_{V_{23}} = ℒ_{V_{3}}

Proof. This follows by exactly the same argument used for quantum Gromov-Hausdorff distance in [17] . □

4 Quantized Gromov-Hausdorff distance

As in the situation of compact quantum metric spaces, we need a corresponding notion of bridge for estimating distance between quantized metric spaces.

Let

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

be two quantized metric spaces with the matrix norms

∥ \cdot ∥_{1} = (∥ \cdot ∥_{1, n})

and

∥ \cdot ∥_{2} = (∥ \cdot ∥_{2, n})

determined by their matrix orders on

(V_{1}, 1)

and

(V_{2}, 1)

, respectively. We form the direct sum

V_{1} \oplus V_{2}

of operator spaces (see §2.6 in [13] ).

(V_{1} \oplus V_{2}, (1, 1))

becomes a matrix order unit space.

Definition 4.1. Let

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

be quantized metric spaces.

A matrix bridge between

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

is a matrix seminorm

N

V_{1} \oplus V_{2}

such that

(1) $N$ is matrix continuous for the matrix norm $∥ \cdot ∥$ on $V_{1} \oplus V_{2}$ , that is, each $N_{n}$ is continuous for $∥ \cdot ∥_{n}$ on $M_{n} (V_{1} \oplus V_{2})$ .
(2) $N_{n} ((a, b)^{*}) = N_{n} (a, b)$ for $a \in M_{n} (V_{1})$ and $b \in M_{n} (V_{2})$ and $n \in N$ .
(3) $N_{1} (1, 1) = 0$ but $N_{1} (1, 0) \neq 0$ .
(4) For any $n \in N$ , $a \in M_{n} (V_{1})$ and $ε > 0$ , there is a $b \in M_{n} (V_{2})$ such that $max {L_{2, n} (b), N_{n} (a, b)} \leq L_{1, n} (a) + ε,$ and similarly for $V_{1}$ and $V_{2}$ interchanged.

Example 4.2. Suppose

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

are quantized metric spaces. Choose

φ_{1} \in C S_{1} (V_{1})

and

ψ_{1} \in C S_{1} (V_{2})

. For

n \in N

, we define

N_{n} : M_{n} (V_{1} \oplus V_{2}) \mapsto [0, + \infty)

N_{n} (a, b) = ∥ ≪ φ_{1}, a ≫ - ≪ ψ_{1}, b ≫ ∥ .

Then

N = (N_{n})

is a matrix seminorm on

V_{1} \oplus V_{2}

, and satisfies the conditions (1), (2) and (3) of Definition 4.1 . For any

a \in M_{n} (V_{1})

and

ε > 0

, choose

b = [φ_{1} (a_{i j}) 1] \in M_{n} (V_{2})

. Then we have

max {L_{2, n} (b), N_{n} (a, b)} = 0 \leq L_{1, n} (a) + ε,

and similarly if we are given

b \in M_{n} (V_{2})

. So

N

is a matrix bridge between

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

Proposition 4.3. Let

N

be a matrix bridge between quantized metric spaces

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

. Define

ℒ = (L_{n})

V_{1} \oplus V_{2}

L_{n} (a, b) = max {L_{1, n} (a), L_{2, n} (b), N_{n} (a, b)}, a \in M_{n} (V_{1}), b \in M_{n} (V_{2}), n \in N .

Let

π_{1}

and

π_{2}

be the projections from

V_{1} \oplus V_{2}

onto

V_{1}

and

V_{2}

, respectively, which are unital completely positive linear surjective mappings. Then

ℒ

is a matrix Lip-norm on

(V_{1} \oplus V_{2}, (1, 1))

, and it induces

V_{1}

and

V_{2}

via

π_{1}

and

π_{2}

, respectively. If

ℒ_{1}

and

ℒ_{2}

are lower semicontinuous, then so is

ℒ

Proof. For

a_{i} \in M_{n} (V_{i})

and

b_{i} \in M_{m} (V_{i}), i = 1, 2

, we have

\begin{matrix} L_{n + m} (a_{1} \oplus b_{1}, a_{2} \oplus b_{2}) \end{matrix}

\begin{matrix} = & max {L_{1, n + m} (a_{1} \oplus b_{1}), L_{2, n + m} (a_{2} \oplus b_{2}), N_{n + m} (a_{1} \oplus b_{1}, a_{2} \oplus b_{2})} \end{matrix}

\begin{matrix} = & max {max {L_{1, n} (a_{1}), L_{1, m} (b_{1})}, max {L_{2, n} (a_{2}), L_{2, m} (b_{2})}, \end{matrix}

\begin{matrix} max {N_{n} (a_{1}, a_{2}), N_{m} (b_{1}, b_{2})}} \end{matrix}

\begin{matrix} = & max {max {L_{1, n} (a_{1}), L_{2, n} (a_{2}), N_{n} (a_{1}, a_{2})}, \end{matrix}

\begin{matrix} max {L_{1, m} (b_{1}), L_{2, m} (b_{2}), N_{m} (b_{1}, b_{2})}} \end{matrix}

\begin{matrix} = & max {L_{n} (a_{1}, a_{2}), L_{m} (b_{1}, b_{2})}, \end{matrix}

and

\begin{matrix} L_{n} ((a_{1}, a_{2})^{*}) & = & L_{n} (a_{1}^{*}, a_{2}^{*}) \end{matrix}

\begin{matrix} = & max {L_{1, n} (a_{1}^{*}), L_{2, n} (a_{2}^{*}), N_{n} (a_{1}^{*}, a_{2}^{*})} \end{matrix}

\begin{matrix} = & max {L_{1, n} (a_{1}), L_{2, n} (a_{2}), N_{n} (a_{1}, a_{2})} \end{matrix}

\begin{matrix} = & L_{n} (a_{1}, a_{2}), \end{matrix}

and for

α \in M_{m, n}

and

β \in M_{n, m}

, we have

\begin{matrix} L_{m} (α (a_{1}, a_{2}) β) \end{matrix}

\begin{matrix} = & L_{m} (α a_{1} β, α a_{2} β) \end{matrix}

\begin{matrix} = & max {L_{1, m} (α a_{1} β), L_{2, m} (α a_{2} β), N_{m} (α a_{1} β, α a_{2} β)} \end{matrix}

\begin{matrix} \leq & max {∥ α ∥ L_{1, n} (a_{1}) ∥ β ∥, ∥ α ∥ L_{2, n} (a_{2}) ∥ β ∥, ∥ α ∥ N_{n} (a_{1}, a_{2}) ∥ β ∥} \end{matrix}

\begin{matrix} = & ∥ α ∥ L_{n} (a_{1}, a_{2}) ∥ β ∥ . \end{matrix}

Thus

ℒ

is a matrix seminorm. Since

L_{1} (a_{s t}) \leq L_{n} ([a_{i j}]) \leq \sum_{i, j = 1}^{n} L_{1} (a_{i j})

for

s, t = 1, 2, \dots, n

and

[a_{i j}] \in M_{n} (V_{1} \oplus V_{2})

L_{n} ([λ_{i j} (1, 1)]) = 0

for

[λ_{i j}] \in M_{n}

L_{n} ([(a_{i j}, b_{i j})]) = 0

, then

L_{1, n} ([a_{i j}]) = L_{2, n} ([b_{i j}]) = 0

, and hence

a_{i j} = λ_{i j} 1

and

b_{i j} = μ_{i j} 1

i, j = 1, 2, \dots, n

, where

λ_{i j}, μ_{i j} \in C

. From

N_{n} ([a_{i j}, b_{i j}]) = 0

and

N_{1} (a_{s t}, b_{s t}) \leq N_{n} ([a_{i j}, b_{i j}])

for

s, t = 1, 2, \dots, n

, we have

N_{1} (λ_{i j} 1, μ_{i j} 1) = 0, i, j = 1, 2, \dots, n,

and so for

i, j = 1, 2, \dots, n

\begin{matrix} | λ_{i j} - μ_{i j} | N_{1} (1, 0) & = & N_{1} ((λ_{i j} - μ_{i j}) 1, 0) \end{matrix}

\begin{matrix} = & N_{1} ((λ_{i j} 1, μ_{i j} 1) - (μ_{i j} 1, μ_{i j} 1)) \end{matrix}

\begin{matrix} \leq & N_{1} (λ_{i j} 1, μ_{i j} 1) + N_{1} (μ_{i j} 1, μ_{i j} 1) \end{matrix}

\begin{matrix} = & 0 . \end{matrix}

Thus

[(a_{i j}, b_{i j})] = [(λ_{i j} 1, λ_{i j} 1)] = [λ_{i j} (1, 1)]

. So

ℒ

is a matrix Lipschitz seminorm.

Similar to the same argument used in Theorem 5.2 of [17] , we have that

ℒ

induces

ℒ_{1}

and

ℒ_{2}

via

π_{1}

and

π_{2}

, respectively. By Proposition 3.1 in [20] , Proposition 7.5 in [21] and Theorem 5.2 in [17] (see also §2 in [17] ), the

D_{ℒ}

-topology on

C S (V_{1} \oplus V_{2})

agrees with the BW-topology. Therefore,

ℒ

is a matrix Lip-norm on

(V_{1} \oplus V_{2}, (1, 1))

Suppose that

ℒ_{1}

and

ℒ_{2}

are lower semicontinuous. Clearly,

ℒ

is lower semicontinuous since

N

is matrix continuous. □

Notation 4.4. We will denote by

ℳ (ℒ_{1}, ℒ_{2})

the set of matrix Lip-norms on

V_{1} \oplus V_{2}

which induce both

ℒ_{1}

and

ℒ_{2}

via

π_{1}

and

π_{2}

, respectively.

By Proposition 4.3 and Example 4.2 ,

ℳ (ℒ_{1}, ℒ_{2}) \neq \emptyset

. From Proposition 3.6 , we can view

C S (V_{1})

and

C S (V_{2})

as closed matrix convex subsets of

C S (V_{1} \oplus V_{2})

Now we introduce our notion of distance for quantized metric spaces.

Definition 4.5. Let

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

be quantized metric spaces. We define the quantized Gromov-Hausdorff distance

d i s t_{N C} (V_{1}, V_{2})

between them by

d i s t_{N C} (V_{1}, V_{2}) = {inf}_{ℒ = (L_{n}) \in ℳ (ℒ_{1}, ℒ_{2})} {sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{1}), C S_{n} (V_{2}))},

where

d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{1}), C S_{n} (V_{2}))

is the Hausdorff distance between

C S_{n} (V_{1})

and

C S_{n} (V_{2})

for

D_{L_{n}}

Given a quantized metric space

(V, ℒ)

, we define its diameter

d i a m (V, ℒ)

to be the diameter of

C S_{1} (V)

with respect to

D_{L_{1}}

. The following proposition indicates that the quantized Gromov-Hausdorff distance is always finite.

Proposition 4.6. Let

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

be quantized metric spaces.

Then

d i s t_{N C} (V_{1}, V_{2}) \leq 2 (d i a m (V_{1}, ℒ_{1}) + d i a m (V_{2}, ℒ_{2})) .

Proof. Choosing arbitrarily

α > 0

φ_{0} \in C S_{1} (V_{1}), ψ_{0} \in C S_{1} (V_{2})

, we set

N_{n} (a, b) = α^{- 1} ∥ ≪ φ_{0}, a ≫ - ≪ ψ_{0}, b ≫ ∥, a \in M_{n} (V_{1}), b \in M_{n} (V_{2}), n \in N .

As Example 4.2 ,

N = (N_{n})

is a matrix bridge between

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

. By Proposition 4.3 ,

ℒ = (L_{n})

, where

L_{n} (a, b) = max {L_{1, n} (a), L_{2, n} (b), N_{n} (a, b)}, a \in M_{n} (V_{1}), b \in M_{n} (V_{2}), n \in N,

is a matrix Lip-norm in

ℳ (ℒ_{1}, ℒ_{2})

. Then for

φ \in C S_{n} (V_{1}), ψ \in C S_{n} (V_{2})

, and

(a, b) \in M_{n} (V_{1} \oplus V_{2})

with

L_{n} (a, b) \leq 1

, we have

\begin{matrix} ∥ ≪ φ, a ≫ - ≪ ψ, b ≫ ∥ \end{matrix}

\begin{matrix} \leq & ∥ ≪ φ, a ≫ - ≪ {\underset{︸}{φ_{0} \oplus \dots \oplus φ_{0}}}_{n}, a ≫ ∥ \end{matrix}

\begin{matrix} + ∥ ≪ {\underset{︸}{φ_{0} \oplus \dots \oplus φ_{0}}}_{n}, a ≫ - ≪ {\underset{︸}{ψ_{0} \oplus \dots \oplus ψ_{0}}}_{n}, b ≫ ∥ \end{matrix}

\begin{matrix} + ∥ ≪ {\underset{︸}{ψ_{0} \oplus \dots \oplus ψ_{0}}}_{n}, b ≫ - ≪ ψ, b ≫ ∥ \end{matrix}

\begin{matrix} \leq & \sum_{i, j} ∥ ≪ φ_{i j} - δ_{i j} φ_{0}, a ≫ ∥ + α + \sum_{i, j} ∥ ≪ ψ_{i j} - δ_{i j} ψ_{0}, b ≫ ∥ \end{matrix}

n = 1

, we get

∥ ≪ φ, a ≫ - ≪ ψ, b ≫ ∥ \leq d i a m (V_{1}, ℒ_{1}) + α + d i a m (V_{2}, ℒ_{2}),

by Proposition 3.1 in [20] . If

n > 1

, similar to the proof of Proposition 4.2 in [20] , there are

φ_{i j}^{(k)} \in C S_{1} (V_{1}), k = 1, 2, 3, 4

, such that

φ_{i j} - δ_{i j} φ_{0} = φ_{i j}^{(1)} - φ_{i j}^{(2)} + i (φ_{i j}^{(3)} - φ_{i j}^{(4)}) .

Since

L_{1, n} (a) \leq L_{n} (a, b) \leq 1

, we obtain

\begin{matrix} \sum_{i, j} ∥ ≪ φ_{i j} - δ_{i j} φ_{0}, a ≫ ∥ \end{matrix}

\begin{matrix} \leq & \sum_{i, j} (∥ ≪ φ_{i j}^{(1)}, a ≫ - ≪ φ_{i j}^{(2)}, a ≫ ∥ + ∥ ≪ φ_{i j}^{(3)}, a ≫ - ≪ φ_{i j}^{(4)}, a ≫ ∥) \end{matrix}

\begin{matrix} \leq & \sum_{i, j} (D_{L_{1, 1}} (φ_{i j}^{(1)}, φ_{i j}^{(2)}) + D_{L_{1, 1}} (φ_{i j}^{(3)}, φ_{i j}^{(4)})) \end{matrix}

\begin{matrix} \leq & 2 n^{2} d i a m (V_{1}, ℒ_{1}) . \end{matrix}

Applying the same argument, we have

\sum_{i, j} ∥ ≪ ψ_{i j} - δ_{i j} ψ_{0}, b ≫ ∥ \leq 2 n^{2} d i a m (V_{2}, ℒ_{2}) .

Hence

∥ ≪ φ, a ≫ - ≪ ψ, b ≫ ∥ \leq 2 n^{2} (d i a m (V_{1}, ℒ_{1}) + α + d i a m (V_{2}, ℒ_{2})) .

The arbitrariness of

α

implies that

d i s t_{N C} (V_{1}, V_{2}) \leq 2 (d i a m (V_{1}) + d i a m (V_{2}))

by Proposition 3.1 in [20] . □

It is clear that the quantized Gromov-Hausdorff distance is symmetric in

V_{1}

and

V_{2}

. We come to prove that it satisfies the triangle inequality.

Theorem 4.7. If

(V_{1}, ℒ_{1})

(V_{2}, ℒ_{2})

and

(V_{3}, ℒ_{3})

be quantized metric spaces, then

d i s t_{N C} (V_{1}, V_{3}) \leq d i s t_{N C} (V_{1}, V_{2}) + d i s t_{N C} (V_{2}, V_{3}) .

Proof. Given

ε > 0

. Then there are

ℒ_{12} \in ℳ (ℒ_{1}, ℒ_{2})

and

ℒ_{23} \in ℳ (ℒ_{2}, ℒ_{3})

such that

{sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{12, n}}} (C S_{n} (V_{1}), C S_{n} (V_{2}))} \leq d i s t_{N C} (V_{1}, V_{2}) + ε

and

{sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{23, n}}} (C S_{n} (V_{2}), C S_{n} (V_{3}))} \leq d i s t_{N C} (V_{2}, V_{3}) + ε .

We define

ℒ = (L_{n})

V_{1} \oplus V_{2} \oplus V_{3}

L_{n} (a_{1}, a_{2}, a_{3}) = max {L_{12, n} (a_{1}, a_{2}), L_{23, n} (a_{2}, a_{3})} .

Then for

a_{i} \in M_{n} (V_{i})

and

b_{i} \in M_{m} (V_{i}), i = 1, 2, 3

, we have

\begin{matrix} L_{n + m} (a_{1} \oplus b_{1}, a_{2} \oplus b_{2}, a_{3} \oplus b_{3}) \end{matrix}

\begin{matrix} = & max {L_{12, n + m} (a_{1} \oplus b_{1}, a_{2} \oplus b_{2}), L_{23, n + m} (a_{2} \oplus b_{2}, a_{3} \oplus b_{3})} \end{matrix}

\begin{matrix} = & max {max {L_{12, n} (a_{1}, a_{2}), L_{12, m} (b_{1}, b_{2})}, max {L_{23, n} (a_{2}, a_{3}), L_{23, m} (b_{2}, b_{3})}} \end{matrix}

\begin{matrix} = & max {max {L_{12, n} (a_{1}, a_{2}), L_{23, n} (a_{2}, a_{3})}, max {L_{12, m} (b_{1}, b_{2}), L_{23, m} (b_{2}, b_{3})}} \end{matrix}

\begin{matrix} = & max {L_{n} (a_{1}, a_{2}, a_{3}), L_{m} (b_{1}, b_{2}, b_{3})}, \end{matrix}

and

\begin{matrix} L_{n} ((a_{1}, a_{2}, a_{3})^{*}) & = & L_{n} (a_{1}^{*}, a_{2}^{*}, a_{3}^{*}) \end{matrix}

\begin{matrix} = & max {L_{12, n} (a_{1}^{*}, a_{2}^{*}), L_{23, n} (a_{2}^{*}, a_{3}^{*})} \end{matrix}

\begin{matrix} = & max {L_{12, n} (a_{1}, a_{2}), L_{23, n} (a_{2}, a_{3})} \end{matrix}

\begin{matrix} = & L_{n} (a_{1}, a_{2}, a_{3}), \end{matrix}

and for

α \in M_{m, n}

and

β \in M_{n, m}

, we have

\begin{matrix} L_{m} (α (a_{1}, a_{2}, a_{3}) β) & = & L_{m} (α a_{1} β, α a_{2} β, α a_{3} β) \end{matrix}

\begin{matrix} = & max {L_{12, m} (α a_{1} β, α a_{2} β), L_{23, m} (α a_{2} β, α a_{3} β)} \end{matrix}

\begin{matrix} \leq & max {∥ α ∥ L_{12, n} (a_{1}, a_{2}) ∥ β ∥, ∥ α ∥ L_{23, n} (a_{2}, a_{3}) ∥ β ∥} \end{matrix}

\begin{matrix} = & ∥ α ∥ L_{n} (a_{1}, a_{2}, a_{3}) ∥ β ∥ . \end{matrix}

L_{n} (a_{1}, a_{2}, a_{3}) = 0

if and only if

L_{12, n} (a_{1}, a_{2}) = 0

and

L_{23, n} (a_{2}, a_{3}) = 0

, and this is equivalent to that

(a_{1}, a_{2}, a_{3}) \in M_{n} (C (1, 1, 1))

. Therefore,

ℒ

is a matrix Lipschitz seminorm.

Similar to the same argument used in Lemma 4.4 of [17] , we have that

ℒ

induces

ℒ_{12}

ℒ_{23}

ℒ_{1}

ℒ_{2}

and

ℒ_{3}

for the evident quotient mappings by Proposition 3.10 .

By Proposition 3.1 in [20] , Proposition 7.5 in [21] and Lemma 4.4 in [17] (see also §2 in [17] ), the

D_{ℒ}

-topology on

C S (V_{1} \oplus V_{2} \oplus V_{3})

agrees with the BW-topology. So

ℒ

is a matrix Lip-norm on

(V_{1} \oplus V_{2} \oplus V_{3}, (1, 1, 1))

By Proposition 3.6 , we have

{sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{1}), C S_{n} (V_{2}))} \leq d i s t_{N C} (V_{1}, V_{2}) + ε,

{sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{2}), C S_{n} (V_{3}))} \leq d i s t_{N C} (V_{2}, V_{3}) + ε

and

d i s t_{N C} (V_{1}, V_{3}) \leq {sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{1}), C S_{n} (V_{3}))} .

\begin{matrix} d i s t_{N C} (V_{1}, V_{3}) \end{matrix}

\begin{matrix} \leq & {sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{1}), C S_{n} (V_{3}))} \end{matrix}

\begin{matrix} \leq & {sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{1}), C S_{n} (V_{2})) \end{matrix}

\begin{matrix} + n^{- 2} d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{2}), C S_{n} (V_{3}))} \end{matrix}

\begin{matrix} \leq & {sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{1}), C S_{n} (V_{2}))} \end{matrix}

\begin{matrix} + {sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{2}), C S_{n} (V_{3}))} \end{matrix}

\begin{matrix} \leq & d i s t_{N C} (C S_{n} (V_{1}), C S_{n} (V_{2})) + d i s t_{N C} (C S_{n} (V_{2}), C S_{n} (V_{3})) + 2 ε . \end{matrix}

Since

ε

is arbitrary, we obtain

d i s t_{N C} (V_{1}, V_{3}) \leq d i s t_{N C} (V_{1}, V_{2}) + d i s t_{N C} (V_{2}, V_{3}) .

□

Proposition 4.8. Let

(V, ℒ)

be a quantized metric space, and let

K^{(1)} = (K_{n}^{(1)})

and

K^{(2)} = (K_{n}^{(2)})

be compact matrix convex subsets of

C S (V)

(V_{j}, ℒ_{j})

is the quotient space of

(V, ℒ)

with respect to

K^{(j)}

and

φ^{(j)}

j = 1, 2

, then we have

d i s t_{N C} (V_{1}, V_{2}) \leq {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{L_{k}}} (K_{k}^{(1)}, K_{k}^{(2)})} .

Proof. Let

p_{1}

and

p_{2}

be the projections from

V \oplus V

onto the first space

V

and the second space

V

, respectively. Denote

G_{n}^{(j)} = (p_{1}^{c})_{n} (K_{n}^{(j)}), H_{n}^{(j)} = (p_{2}^{c})_{n} (K_{n}^{(j)}), j = 1, 2, n \in N,

and set

G^{(j)} = (G_{n}^{(j)})

ℋ^{(j)} = (H_{n}^{(j)})

j = 1, 2

, and

K = (K_{n}) = \bar{m c o} (G^{(1)} \cup ℋ^{(2)})

, the BW-closed matrix convex hull of the graded set

(G_{n}^{(1)} \cup H_{n}^{(2)})

. Let

(W, φ (1 \oplus 1))

be the quotient of

(V \oplus V, 1 \oplus 1)

with respect to

K

. Then

K = φ^{c} (C S (W))

by Proposition 3.8 .

For

(a, b) \in K e r φ

, we have

φ (a, b) = 0

, that is,

≪ (a, b), φ ≫ = 0_{n}

for

φ \in K_{n}

This is equivalent to

≪ (a, b), φ ≫ = 0_{n}

for

φ \in G_{n}^{(1)} \cup H_{n}^{(2)}

n \in N

since

(a, b) \in A (K)

. And this holds if and only if

≪ a, φ_{1} ≫ = 0_{n}

and

≪ b, φ_{2} ≫ = 0_{n}

for

φ_{1} \in G_{n}^{(1)}

and

φ_{2} \in H_{n}^{(2)}

n \in N

, that is, if and only if

a \in K e r φ^{(1)}

and

b \in K e r φ^{(2)}

K e r φ = K e r φ^{(1)} \oplus K e r φ^{(2)}

. And thus there is a complete order isomorphism

Ψ

from

W

onto

V_{1} \oplus V_{2}

Given

ε > 0

. We define a matrix seminorm

N = (N_{n})

V \oplus V

N_{n} (a, b) = ε^{- 1} ∥ a - b ∥_{n}, a, b \in M_{n} (V) .

Then

N

is a matrix bridge between

(V, ℒ)

and

(V, ℒ)

, and

Q = (Q_{n}) \in ℳ (ℒ, ℒ)

by Proposition 4.3 , where

Q_{n} (a, b) = max {L_{n} (a), L_{n} (b), N_{n} (a, b)}, a, b \in M_{n} (V), n \in N .

Thus

Q

is a matrix Lip-norm on

(V \oplus V, (1, 1))

. Let

P = (P_{n})

and

(W, P)

be the quotient space of

(V \oplus V, Q)

with respect to

K

and

φ

. Then

P \in ℳ (ℒ_{1}, ℒ_{2})

by Propositon 3.10 .

Since

D_{P_{k}} (φ_{1}, φ_{2}) = D_{Q_{k}} (φ_{k}^{c} (φ_{1}), φ_{k}^{c} (φ_{2}))

for

φ_{1}, φ_{2} \in C S_{k} (W)

, we have that

d i s t_{H}^{D_{P_{k}}} (C S_{k} (V_{1}), C S_{k} (V_{2})) = d i s t_{H}^{D_{Q_{k}}} (G_{k}^{(1)}, H_{k}^{(2)})

. For

ψ \in K_{k}^{(2)}

, we have

\begin{matrix} D_{Q_{k}} ((p_{1}^{c})_{k} (ψ), (p_{2}^{c})_{k} (ψ)) \end{matrix}

\begin{matrix} = & sup {∥ ≪ (p_{1}^{c})_{k} (ψ), (a, b) ≫ - ≪ (p_{2}^{c})_{k} (ψ), (a, b) ≫ ∥ : Q_{r} (a, b) \leq 1, r \in N} \end{matrix}

\begin{matrix} \leq & sup {∥ ≪ ψ, a ≫ - ≪ ψ, b ≫ ∥ : N_{r} (a, b) \leq 1, r \in N} \end{matrix}

\begin{matrix} \leq & ε, \end{matrix}

that is,

d i s t_{H}^{D_{Q_{k}}} (G_{k}^{(2)}, H_{k}^{(2)}) \leq ε

. Because

Q \in ℳ (ℒ, ℒ)

, we get that

d i s t_{H}^{D_{Q_{k}}} (G_{k}^{(1)}, G_{k}^{(2)}) = d i s t_{H}^{D_{L_{k}}} (K_{k}^{(1)}, K_{k}^{(2)}) .

\begin{matrix} d i s t_{N C} (V_{1}, V_{2}) & \leq & {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{P_{k}}} (C S_{k} (V_{1}), C S_{k} (V_{2}))} \end{matrix}

\begin{matrix} = & {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{Q_{k}}} (G_{k}^{(1)}, H_{k}^{(2)})} \end{matrix}

\begin{matrix} = & {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{Q_{k}}} (G_{k}^{(1)}, G_{k}^{(2)})} \end{matrix}

\begin{matrix} + {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{Q_{k}}} (G_{k}^{(2)}, H_{k}^{(2)})} \end{matrix}

\begin{matrix} \leq & {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D Q_{P_{k}}} (G_{k}^{(1)}, G_{k}^{(2)}) + k^{- 2} ε} \end{matrix}

\begin{matrix} \leq & {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{Q_{k}}} (G_{k}^{(1)}, G_{k}^{(2)})} + ε \end{matrix}

\begin{matrix} = & {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{L_{k}}} (K_{k}^{(1)}, K_{k}^{(2)})} + ε \end{matrix}

Since

ε

is arbitrary, we obtain the desired inequality. □

Let

(A, L_{A})

and

(B, L_{B})

be compact quantum metric spaces. The quantum Gromov-Hausdorff distance between them is defined by

d i s t_{q} (A, B) = inf d i s t_{H}^{ρ_{L}} (S (A), S (B)),

where the infimum is taken over all Lip-norms

L

A \oplus B

which induce

L_{A}

and

L_{B}

(see Definition 4.2 in [17] ).

Proposition 4.9. Let

(A_{j}, L_{j})

for

j = 1, 2

be compact quantum metric spaces, and let

(V_{j}, ℒ_{j})

be an associated quantized metric space of

(A_{j}, L_{j})

(see Example 3.5 ). Then

d i s t_{q} (A_{1}, A_{2}) \leq d i s t_{N C} (V_{1}, V_{2}) .

Proof. Suppose

Q \in ℳ (ℒ_{1}, ℒ_{2})

. Then

Q_{V_{j}, 1} = L_{j, 1}

for

j = 1, 2

and

L_{j, 1} (a) = L_{j}^{s} (a)

for

a \in A_{j}

, where

L_{j}^{s} = L_{ρ_{L_{j}}}

(see Example 3.5 ). So for

a \in A_{1}

, we have

\begin{matrix} L_{1}^{s} (a) & = & L_{1, 1} (a) = Q_{V_{j}, 1} (a) \end{matrix}

\begin{matrix} = & inf {Q_{1} (a_{1}, b_{1}) : π_{1} (a_{1}, b_{1}) = a, (a_{1}, b_{1}) = V_{1} \oplus V_{2}} \end{matrix}

\begin{matrix} = & inf {Q_{1} (a, b_{1}) : b_{1} \in V_{2}} \end{matrix}

\begin{matrix} \leq & inf {Q_{1} (a, b) : b \in A_{2}} \end{matrix}

\begin{matrix} = & inf {R (a, b) : b \in A_{2}} \end{matrix}

\begin{matrix} = & R_{A_{1}} (a), \end{matrix}

where

π_{1}

is the projection from

V_{1} \oplus V_{2}

onto

V_{1}

and

R

is the restriction of

Q_{1}

A_{1} \oplus A_{2}

. Denote

c = inf {Q_{1} (a, b) : b \in V_{2}}

. Let

ε > 0

be given. Then there is a

y \in V_{2}

such that

Q_{1} (a, y) \leq c + ε

. Setting

x = \frac{1}{2} (y + y^{*})

, we have that

x \in A_{2}

and

\begin{matrix} R (a, x) & = & Q_{1} (a, x) \end{matrix}

\begin{matrix} = & Q_{1} (a, \frac{1}{2} (y + y^{*})) \end{matrix}

\begin{matrix} \leq & \frac{1}{2} Q_{1} ((a, y) + (a, y)^{*}) \end{matrix}

\begin{matrix} \leq & \frac{1}{2} (Q_{1} (a, y) + Q_{1} ((a, y)^{*})) \end{matrix}

\begin{matrix} = & Q_{1} (a, y) \end{matrix}

\begin{matrix} \leq & c + ε . \end{matrix}

Thus

L_{1}^{s} (a) = R_{A_{1}} (a)

for

a \in A_{1}

. Similarly, we have that

L_{2}^{s} (b) = R_{A_{2}} (b)

for

b \in A_{2}

. So

R \in ℳ (L_{1}^{s}, L_{2}^{s})

For

φ \in C S_{1} (V_{1})

and

ψ \in C S_{1} (V_{2})

, let

φ_{1} = φ |_{A_{1}}

and

ψ_{1} = ψ |_{A_{2}}

. Then

φ_{1} \in S (A_{1})

and

ψ_{1} \in S (A_{2})

. Since

Q_{1} ((a, b)^{*}) = Q_{1} (a, b)

, we obtain

\begin{matrix} D_{Q_{1}} (φ, ψ) & = & sup {| φ (c) - ψ (d) | : Q_{1} (c, d) \leq 1, (c, d) \in V_{1} \oplus V_{2}} \end{matrix}

\begin{matrix} = & sup {| φ (c) - ψ (d) | : Q_{1} (c, d) \leq 1, (c, d) = (c, d)^{*} \in V_{1} \oplus V_{2}} \end{matrix}

\begin{matrix} = & sup {| φ_{1} (c) - ψ_{1} (d) | : R (c, d) \leq 1, (c, d) \in A_{1} \oplus A_{2}} \end{matrix}

\begin{matrix} = & ρ_{R} (φ_{1}, ψ_{1}), \end{matrix}

(see §2 in [17] ). So

d i s t_{H}^{ρ_{R}} (S (A_{1}), S (A_{2})) = d i s t_{H}^{D_{Q_{1}}} (C S_{1} (V_{1}), C S_{1} (V_{2})) .

Therefore, by Theorem 4.3 and Proposition 7.1 in [17] , we have

\begin{matrix} d i s t_{q} (A_{1}, A_{2}) & \leq & d i s t_{q} ((A_{1}, L_{1}), (A_{1}, L_{1}^{s})) + d i s t_{q} ((A_{1}, L_{1}^{s}), (A_{2}, L_{2}^{s})) \end{matrix}

\begin{matrix} + d i s t_{q} ((A_{2}, L_{2}^{s}), (A_{2}, L_{2})) \end{matrix}

\begin{matrix} = & d i s t_{q} ((A_{1}, L_{1}^{s}), (A_{2}, L_{2}^{s})) \end{matrix}

\begin{matrix} \leq & d i s t_{H}^{ρ_{R}} (S (A_{1}), S (A_{2})) \end{matrix}

\begin{matrix} = & d i s t_{H}^{D_{Q_{1}}} (C S_{1} (V_{1}), C S_{1} (V_{2})) \end{matrix}

\begin{matrix} \leq & {sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{Q_{n}}} (C S_{n} (V_{1}), C S_{n} (V_{2}))} . \end{matrix}

Consequently,

d i s t_{q} (A_{1}, A_{2}) \leq d i s t_{N C} (V_{1}, V_{2})

. □

5 Distance zero

In this section, we show that

d i s t_{N C} (V_{1}, V_{2}) = 0

is equivalent to the existence of a complete isometry between them in the following sense.

(V, ℒ)

is a quantized metric space, then

ℒ_{D_{ℒ}}

is the largest lower semicontinuous matrix Lip-norm smaller than

ℒ

by Corollary 4.5 in [21] . From Proposition 7.1 in [21] ,

ℒ_{D_{ℒ}}

extends uniquely to a closed matrix Lip-norm

ℒ^{c}

on the subspace

V^{c} = {a \in \bar{V} : L_{1}^{c} (a) < + \infty}

, where

\bar{V}

is the completion of

V

for its matrix norm.

Definition 5.1. Let

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

be quantized metric spaces. By a complete isometry from

(V_{1}, ℒ_{1})

onto

(V_{2}, ℒ_{2})

we mean a unital complete order isomorphism

Φ

from

V_{1}^{c}

onto

V_{2}^{c}

such that

ℒ_{1}^{c} = ℒ_{2}^{c} \circ Φ

, that is,

L_{1, n}^{c} = L_{2, n}^{c} \circ Φ_{n}

for all

n \in N

Lemma 5.2. Let

(V, ℒ)

be a quantized metric space. Then

d i s t_{N C} (V, V^{c}) = 0, d i s t_{N C} ((V, ℒ), (V, ℒ_{D_{ℒ}})) = 0 .

Proof. Let

ε > 0

be given, and define

N_{n} (a_{1}, a_{2}) = ε^{- 1} ∥ a_{1} - a_{2} ∥_{n},

for

a_{1} \in M_{n} (V)

a_{2} \in M_{n} (V^{c})

and

n \in N

. Clearly

N = (N_{n})

is a matrix continuous matrix seminorm on

V \oplus V^{c}

, and

N_{1} (1, 1) = 0

and

N_{1} (1, 0) = ε^{- 1} \neq 0

For

n \in N

and

a_{1} \in M_{n} (V)

and

δ > 0

, setting

a_{2} = a_{1} \in M_{n} (V^{c})

, we have

max {L_{n}^{c} (a_{2}), N_{n} (a_{1}, a_{2})} = L_{n}^{c} (a_{2}) = L_{D_{L_{n}}} (a_{2}) \leq L_{n} (a_{2}) < L_{n} (a_{2}) + δ

by Proposition 3.6, Proposition 7.1 and Proposition 3.4 in [21] . Given

n \in N

and

a_{2} \in M_{n} (V^{c})

and

δ > 0

. By Lemma 7.3 in [21] , there is a sequence

{a_{1}^{(k)}}

of elements in

M_{n} (V)

such that

L_{n} (a_{1}^{(k)}) \leq L_{n}^{c} (a_{2})

and

{a_{1}^{(k)}}

converges to

a_{2}

in norm. Consequently, we can find an

a_{1}^{(k_{0})}

such that

ε^{- 1} {‖ a_{1}^{(k_{0})} - a_{2} ‖}_{n} \leq L_{n}^{c} (a_{2}) + δ

max {L_{n} (a_{1}^{(k_{0})}), N_{n} (a_{1}^{(k_{0})}, a_{2})} \leq L_{n}^{c} (a_{2}) + δ

. Thus

N

is a matrix bridge between

(V, ℒ)

and

(V^{c}, ℒ^{c})

Define

L_{n} (a_{1}, a_{2}) = max {L_{n} (a_{1}), L_{n}^{c} (a_{2}), N_{n} (a_{1}, a_{2})}

for

a_{1} \in M_{n} (V)

a_{2} \in M_{n} (V^{c})

and

n \in N

. By Proposition 4.3 ,

ℒ = (L_{n}) \in ℳ (V, V^{c})

. For

n \in N

and

φ \in C S_{n} (V^{c})

, we have that

ψ = φ |_{V} \in C S_{n} (V)

, and hence

\begin{matrix} D_{L_{n}} (ψ, φ) \end{matrix}

\begin{matrix} = & sup {∥ ≪ (π_{1})_{n}^{c} (ψ), (a_{1}, a_{2}) ≫ - ≪ (π_{2})_{n}^{c} (φ), (a_{1}, a_{2}) ≫ ∥ : \end{matrix}

\begin{matrix} L_{r} (a_{1}, a_{2}) \leq 1, (a_{1}, a_{2}) \in M_{r} (V \oplus V^{c}), r \in N} \end{matrix}

\begin{matrix} = & sup {∥ ≪ ψ, a_{1} ≫ - ≪ φ, a_{2} ≫ ∥ : \end{matrix}

\begin{matrix} L_{r} (a_{1}, a_{2}) \leq 1, (a_{1}, a_{2}) \in M_{r} (V \oplus V^{c}), r \in N} \end{matrix}

\begin{matrix} = & sup {∥ ≪ φ, a_{1} - a_{2} ≫ ∥ : \end{matrix}

\begin{matrix} L_{r} (a_{1}, a_{2}) \leq 1, (a_{1}, a_{2}) \in M_{r} (V \oplus V^{c}), r \in N} \end{matrix}

\begin{matrix} \leq & sup {∥ a_{1} - a_{2} ∥_{r} : L_{r} (a_{1}, a_{2}) \leq 1, (a_{1}, a_{2}) \in M_{r} (V \oplus V^{c}), r \in N} \end{matrix}

\begin{matrix} \leq & ε, \end{matrix}

where

π_{1}

and

π_{2}

are the projections from

V \oplus V^{c}

onto

V

and

V^{c}

, respectively. For

n \in N

and

φ \in C S_{n} (V)

, there is a

ψ \in C S_{n} (V^{c})

such that

ψ |_{V} = φ

by Arveson's extension theorem. So

\begin{matrix} D_{L_{n}} (φ, ψ) \end{matrix}

\begin{matrix} = & sup {∥ ≪ (π_{1})_{n}^{c} (φ), (a_{1}, a_{2}) ≫ - ≪ (π_{2})_{n}^{c} (ψ), (a_{1}, a_{2}) ≫ ∥ : \end{matrix}

\begin{matrix} L_{r} (a_{1}, a_{2}) \leq 1, (a_{1}, a_{2}) \in M_{r} (V \oplus V^{c}), r \in N} \end{matrix}

\begin{matrix} = & sup {∥ ≪ φ, a_{1} ≫ - ≪ ψ, a_{2} ≫ ∥ : \end{matrix}

\begin{matrix} L_{r} (a_{1}, a_{2}) \leq 1, (a_{1}, a_{2}) \in M_{r} (V \oplus V^{c}), r \in N} \end{matrix}

\begin{matrix} = & sup {∥ ≪ ψ, a_{1} - a_{2} ≫ ∥ : \end{matrix}

\begin{matrix} L_{r} (a_{1}, a_{2}) \leq 1, (a_{1}, a_{2}) \in M_{r} (V \oplus V^{c}), r \in N} \end{matrix}

\begin{matrix} \leq & sup {∥ a_{1} - a_{2} ∥_{r} : L_{r} (a_{1}, a_{2}) \leq 1, (a_{1}, a_{2}) \in M_{r} (V \oplus V^{c}), r \in N} \end{matrix}

\begin{matrix} \leq & ε . \end{matrix}

Thus

d i s t_{H}^{D_{L_{n}}} (C S_{n} (V), C S_{n} (V^{c})) \leq ε

for

n \in N

, and so

{sup}_{n \in N} {d i s t_{H}^{D_{L_{n}}} (C S_{n} (V), C S_{n} (V^{c}))} \leq ε .

Therefore,

d i s t_{N C} (V, V^{c}) \leq ε

. By the arbitrariness of

ε

, we obtain

d i s t_{N C} (V, V^{c}) = 0 .

By Proposition 3.4 in [21] and the proof of Theorem 4.4 in [21] , we can prove that

d i s t_{N C} ((V, ℒ), (V, ℒ_{D_{ℒ}})) = 0

similarly. □

Theorem 5.3. Suppose

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

are quantized metric spaces.

If there exists a complete isometry

Φ

from

(V_{1}, ℒ_{1})

onto

(V_{2}, ℒ_{2})

, then

d i s t_{N C} (V_{1}, V_{2}) = 0 .

Proof. For

ε > 0

, we define

N_{n} (a_{1}, a_{2}) = ε^{- 1} ∥ Φ_{n} (a_{1}) - a_{2} ∥_{n},

for

a_{1} \in M_{n} (V_{1}^{c})

a_{2} \in M_{n} (V_{2}^{c})

and

n \in N

. Clearly

N = (N_{n})

is a matrix seminorm on

V_{1}^{c} \oplus V_{2}^{c}

and

N_{n} ((a_{1}, a_{2})^{*}) = N_{n} (a_{1}, a_{2})

for

a_{1} \in M_{n} (V_{1}^{c})

a_{2} \in M_{n} (V_{2}^{c})

and

n \in N

. And we have that

N_{1} (1, 1) = ε^{- 1} ∥ Φ (1) - 1 ∥_{1} = 0

and

N_{1} (1, 0) = ε^{- 1} ∥ Φ (1) - 0 ∥_{1} = ε^{- 1}

. If

{a_{1}^{(k)}} \subseteq M_{n} (V_{1}^{c})

and

{a_{2}^{(k)}} \subseteq M_{n} (V_{2}^{c})

with

{lim}_{k \to \infty} a_{1}^{(k)} = a_{1} \in M_{n} (V_{1}^{c})

and

{lim}_{k \to \infty} a_{2}^{(k)} = a_{2} \in M_{n} (V_{2}^{c})

, we have that

{lim}_{k \to \infty} N_{n} (a_{1}^{(k)}, a_{2}^{(k)}) = {lim}_{k \to \infty} ε^{- 1} ∥ Φ_{n} (a_{1}^{(k)}) - a_{2}^{(k)} ∥_{n} = ε^{- 1} ∥ Φ_{n} (a_{1}) - a_{2} ∥_{n} = N_{n} (a_{1}, a_{2})

since

Φ

is completely bounded(see Proposition 3.5 in [12] ).

Given

a_{1} \in M_{n} (V_{1}^{c})

and

δ > 0

. Taking

a_{2} = Φ_{n} (a_{1})

, we have

\begin{matrix} max {L_{2, n}^{c} (a_{2}), N_{n} (a_{1}, a_{2})} & = & max {L_{2, n}^{c} (Φ_{n} (a_{1})), ε^{- 1} ∥ Φ_{n} (a_{1}) - a_{2} ∥} \end{matrix}

\begin{matrix} = & L_{1, n}^{c} (a_{1}) < L_{1, n}^{c} (a_{1}) + δ . \end{matrix}

While if

a_{2} \in M_{n} (V_{2}^{c})

and

δ > 0

, we can take

a_{1} \in M_{n} (V_{1}^{c})

such that

Φ_{n} (a_{1}) = a_{2}

, and hence we have

\begin{matrix} max {L_{1, n}^{c} (a_{1}), N_{n} (a_{1}, a_{2})} & = & max {L_{2, n}^{c} (Φ_{n} (a_{1})), ε^{- 1} ∥ Φ_{n} (a_{1}) - a_{2} ∥} \end{matrix}

\begin{matrix} = & L_{2, n}^{c} (a_{2}) < L_{2, n}^{c} (a_{2}) + δ . \end{matrix}

Therefore,

N

is a matrix bridge between

(V_{1}^{c}, ℒ_{1}^{c})

and

(V_{2}^{c}, ℒ_{2}^{c})

. Define

L_{n} (a_{1}, a_{2}) = max {L_{1, n}^{c} (a_{1}), L_{2, n}^{c} (a_{2}), N_{n} (a_{1}, a_{2})},

for

a_{1} \in M_{n} (V_{1}^{c})

a_{2} \in M_{n} (V_{2}^{c})

and

n \in N

. By Proposition 4.3 ,

ℒ = (L_{n}) \in ℳ (ℒ_{1}^{c}, ℒ_{2}^{c})

. For

n \in N

and

φ \in C S_{n} (V_{2}^{c})

, we have that

φ \circ Φ \in C S_{n} (V_{1}^{c})

, and so

\begin{matrix} D_{L_{n}} (φ \circ Φ, φ) \end{matrix}

\begin{matrix} = & sup {∥ ≪ φ \circ Φ, a_{1} ≫ - ≪ φ, a_{2} ≫ ∥ : \end{matrix}

\begin{matrix} L_{r} (a_{1}, a_{2}) \leq 1, (a_{1}, a_{2}) \in M_{r} (V_{1}^{c} \oplus V_{2}^{c}), r \in N} \end{matrix}

\begin{matrix} = & sup {∥ ≪ φ, Φ_{r} (a_{1}) - a_{2} ≫ ∥ : \end{matrix}

\begin{matrix} L_{r} (a_{1}, a_{2}) \leq 1, (a_{1}, a_{2}) \in M_{r} (V_{1}^{c} \oplus V_{2}^{c}), r \in N} \end{matrix}

\begin{matrix} \leq & sup {∥ Φ_{r} (a_{1}) - a_{2} ∥_{r} : L_{r} (a_{1}, a_{2}) \leq 1, (a_{1}, a_{2}) \in M_{r} (V_{1}^{c} \oplus V_{2}^{c}), r \in N} \end{matrix}

\begin{matrix} \leq & ε, \end{matrix}

Similarly, for

n \in N

and

ψ \in C S_{n} (V_{1}^{c})

, we have that

D_{L_{n}} ((ψ, ψ \circ Φ^{- 1}) \leq ε

. Thus we obtain that

d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{1}^{c}), C S_{n} (V_{2}^{c})) \leq ε

for

n \in N

, and so

{sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{n}}} (C S_{n} (V_{1}^{c}), C S_{n} (V_{2}^{c}))} \leq ε .

Therefore,

d i s t_{N C} (V_{1}^{c}, V_{2}^{c}) \leq ε

. Since

ε

is arbitrary, we conclude

d i s t_{N C} (V_{1}^{c}, V_{2}^{c}) = 0 .

Now, by Theorem 4.7 and Lemma 5.2 we have

d i s t_{N C} (V_{1}, V_{2}) \leq d i s t_{N C} (V_{1}, V_{1}^{c}) + d i s t_{N C} (V_{1}^{c}, V_{2}^{c}) + d i s t_{N C} (V_{2}^{c}, V_{2}) = 0 .

d i s t_{N C} (V_{1}, V_{2}) = 0

. □

Given a quantized metric space

(V, ℒ)

. From Proposition 6.1 in [21] and the proof of Proposition 3.5 in [19] , the mapping

Ψ : V \mapsto A (C S (V))

, defined by

Ψ (a) (φ) = φ (a)

for

φ \in C S_{r} (V)

, is a unital complete order isomorphism from

V

into

A (C S (V))

, and

Ψ

can be extended to a unital complete order isomorphism

\bar{Ψ}

from the completion

\bar{V}

V

onto

A (C S (V))

. Define

L_{D_{L_{n}}} (F^{(n)}) = sup {\frac{∥ F_{r}^{(n)} (φ) - F_{r}^{(n)} (ψ) ∥}{D_{L_{r}} (φ, ψ)} : φ \neq ψ, φ, ψ \in C S_{r} (V), r \in N},

where

F^{(n)} \in A (C S (V), M_{n})

. Then

ℒ_{D_{ℒ}} = (L_{D_{L_{n}}})

is a matrix gauge on

A (C S (V))

Denote

K_{n} = {F^{(n)} \in A (C S (V), M_{n}) : L_{D_{L_{n}}} (F^{(n)}) < + \infty} .

Let

L_{n}^{1} = {a \in V : L_{n} (a) \leq 1}

and

{\bar{L}}_{n}^{1}

be the norm closure of

L_{n}^{1}

\bar{V}

. Denote

ℒ^{1} = (L_{n}^{1})

and

{\bar{ℒ}}^{1} = ({\bar{L}}_{n}^{1})

. The matrix gauge

\bar{ℒ} = ({\bar{L}}_{n})

(\bar{V}, 1)

determined by

{\bar{ℒ}}^{1}

is called the closure of

ℒ

ℒ

is closed if

ℒ = \bar{ℒ}

on the subspace where

\bar{ℒ}

is finite (see Definition 7.2 in [21] ).

Lemma 5.4. If

ℒ

is closed, then

Ψ_{n} (M_{n} (V)) = K_{n}

for

n \in N

Proof. Denote

M_{n}^{1} = {Ψ_{n} (a) : a \in M_{n} (V), L_{n} (a) \leq 1}, n \in N,

L_{D_{L_{n}}}^{1} = {F^{(n)} \in A (C S (V), M_{n}) : L_{D_{L_{n}}} (F^{(n)}) \leq 1}, n \in N,

and set

ℳ^{1} = (M_{n}^{1})

. Define

L_{n}^{'} (f) = sup {∥ ≪ f, Ψ_{r} (a) ≫ ∥ : L_{r} (a) \leq 1, a \in M_{r} (V), r \in N},

for

f \in M_{n} ((A (C S (V)) / (C I))^{*})

, where

I

is the order unit of

A (C S (V))

. Here we view

M_{n} ((A (C S (V)) / (C I))^{*})

as the subspace of

M_{n} ((A (C S (V)))^{*})

consisting of those

f \in M_{n} ((A (C S (V)))^{*})

with

f (a) = 0_{n}

for

a \in C I

. Clearly,

ℒ^{'} = (L_{n}^{'})

is a matrix gauge on

(A (C S (V)) / (C I))^{*}

and

L_{n}^{'} (f^{*}) = L_{n}^{'} (f)

for all

f \in M_{n} ((A (C S (V)) / (C I))^{*})

, The generalized bipolar theorem says that

(ℳ^{1})^{⊚ ⊚}

is the smallest weakly closed absolutely matrix convex set containing

ℳ^{1}

(see Proposition 4.1 in [4] ). Since

ℒ = (L_{n})

is a matrix gauge and

Ψ

is a unital complete order isomorphism,

ℳ^{1}

is absolutely matrix convex. The closeness of

ℒ

implies that

ℳ^{1}

is normed-closed by Lemma 7.4 in [21] . So

ℳ^{1}

is weakly closed. Thus

(ℳ^{1})^{⊚ ⊚} = ℳ^{1} .

For

n \in N

, we have

\begin{matrix} (M_{n}^{1})^{⊚} & = & {Ψ_{n} (a) : a \in M_{n} (V), L_{n} (a) \leq 1}^{⊚} \end{matrix}

\begin{matrix} = & {f \in M_{n} ((A (C S (V)))^{*}) : ∥ ≪ f, Ψ_{r} (a) ≫ ∥ \leq 1 \end{matrix}

\begin{matrix} f o r a l l a \in M_{r} (V), L_{r} (a) \leq 1, r \in N} \end{matrix}

\begin{matrix} = & {f \in M_{n} ((A (C S (V)) / (C I))^{*}) : ∥ ≪ f, Ψ_{r} (a) ≫ ∥ \leq 1 \end{matrix}

\begin{matrix} f o r a l l a \in M_{r} (V), L_{r} (a) \leq 1, r \in N} \end{matrix}

\begin{matrix} = & {f \in M_{n} ((A (C S (V)) / (C I))^{*}) : L_{n}^{'} (f) \leq 1} \end{matrix}

and

\begin{matrix} (M_{n}^{1})^{⊚ ⊚} \end{matrix}

\begin{matrix} = & {f \in M_{n} ((A (C S (V)) / (C I))^{*}) : L_{n}^{'} (f) \leq 1}^{⊚} \end{matrix}

\begin{matrix} = & {F^{(n)} \in M_{n} (A (C S (V))) : ∥ ≪ f, F^{(n)} ≫ ∥ \leq 1 \end{matrix}

\begin{matrix} f o r a l l f \in M_{r} ((A (C S (V)) / (C I))^{*}), L_{r}^{'} (f) \leq 1, r \in N} . \end{matrix}

F^{(n)} \in (M_{n}^{1})^{⊚ ⊚}

if and only if

∥ ≪ f, F^{(n)} ≫ ∥ \leq L_{r}^{'} (f)

for all

f \in M_{r} ((A (C S (V)) / (C I))^{*})

and

r \in N

Suppose that

∥ ≪ f, F^{(n)} ≫ ∥ \leq L_{r}^{'} (f)

for all

f = f^{*} \in M_{r} ((A (C S (V)) / (C I))^{*})

and

r \in N

. Then for

g \in M_{r} ((A (C S (V)) / (C I))^{*})

, we have

\begin{matrix} ∥ ≪ g, F^{(n)} ≫ ∥ & = & ‖ [\begin{matrix} 1 & 0 \end{matrix}] [\begin{matrix} 0 & ≪ g, F^{(n)} ≫ \\ ≪ g^{*}, F^{(n)} ≫ & 0 \end{matrix}] [\begin{matrix} 0 \\ 1 \end{matrix}] ‖ \end{matrix}

\begin{matrix} \leq & ‖ ≪ [\begin{matrix} 0 & g \\ g^{*} & 0 \end{matrix}], F^{(n)} ≫ ‖ \end{matrix}

\begin{matrix} \leq & L_{2 r}^{'} ([\begin{matrix} 0 & g \\ g^{*} & 0 \end{matrix}]) \end{matrix}

\begin{matrix} \leq & L_{2 r}^{'} ([\begin{matrix} g & 0 \\ 0 & g^{*} \end{matrix}] [\begin{matrix} 0 & 1 \\ 1 & 0 \end{matrix}]) \end{matrix}

\begin{matrix} \leq & L_{2 r}^{'} ([\begin{matrix} g & 0 \\ 0 & g^{*} \end{matrix}]) \end{matrix}

\begin{matrix} = & L_{r}^{'} (g), \end{matrix}

Thus

F^{(n)} \in (M_{n}^{1})^{⊚ ⊚}

exactly if

∥ ≪ f, F^{(n)} ≫ ∥ \leq L_{r}^{'} (f)

for all

f = f^{*} \in M_{r} ((A (C S (V)) / (C I))^{*})

and

r \in N

. According to Lemma 4.1 in [21] ,

F^{(n)} \in (M_{n}^{1})^{⊚ ⊚}

exactly if

∥ ≪ φ, F^{(n)} ≫ - ≪ ψ, F^{(n)} ≫ ∥ \leq L_{r}^{'} (φ - ψ)

for all

φ, ψ \in C S_{r} (A (C S (V)))

and

r \in N

. So

F^{(n)} \in (M_{n}^{1})^{⊚ ⊚}

exactly if

\begin{matrix} ∥ ≪ φ \circ Ψ^{- 1}, F^{(n)} ≫ - ≪ ψ \circ Ψ^{- 1}, F^{(n)} ≫ ∥ \end{matrix}

\begin{matrix} \leq & L_{r}^{'} (φ \circ Ψ^{- 1} - ψ \circ Ψ^{- 1}) \end{matrix}

\begin{matrix} = & sup {∥ ≪ (φ - ψ) \circ Ψ^{- 1}, Ψ_{k} (a) ≫ ∥ : L_{k} (a) \leq 1, a \in M_{k} (V), k \in N} \end{matrix}

\begin{matrix} = & sup {∥ ≪ φ - ψ, a ≫ ∥ : L_{k} (a) \leq 1, a \in M_{k} (V), k \in N} \end{matrix}

\begin{matrix} = & D_{L_{r}} (φ, ψ) . \end{matrix}

for all

φ, ψ \in C S_{r} (A (C S (V)))

and

r \in N

. Because

∥ ≪ φ \circ Ψ^{- 1}, F^{(n)} ≫ - ≪ ψ \circ Ψ^{- 1}, F^{(n)} ≫ ∥ = ∥ F_{r}^{(n)} (φ) - F_{r}^{(n)} (ψ) ∥

F^{(n)} \in (M_{n}^{1})^{⊚ ⊚}

if and only if

∥ F_{r}^{(n)} (φ) - F_{r}^{(n)} (ψ) ∥ \leq D_{L_{r}} (φ, ψ)

for all

φ, ψ \in C S_{r} (A (C S (V)))

and

r \in N

. And this says exactly that

F^{(n)} \in L_{D_{L_{n}}}^{1}

. Therefore,

M_{n}^{1} = L_{D_{L_{n}}}^{1}

for

n \in N

. So

Ψ_{n} (M_{n} (V)) = K_{n}

for

n \in N

. □

Lemma 5.5. Let

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

be quantized metric spaces such that

ℒ_{1}

and

ℒ_{2}

are closed. For every matrix affine mapping

α = (α_{n})

from

C S (V_{1})

onto

C S (V_{2})

which is completely isometric for

D_{ℒ_{1}}

and

D_{ℒ_{2}}

, there is a unital complete order isomorphism

Ψ

from

V_{1}

onto

V_{2}

such that

ℒ_{1} = ℒ_{2} \circ Ψ

Proof. Define $Φ : A (C S (V_{2})) \mapsto A (C S (V_{1}))$ by ${(Φ (F^{(1)}))}_{r} (φ) = F_{r}^{(1)} (α_{r} (φ))$ for $F^{(1)} \in A (C S (V_{2}))$ and $φ \in C S_{r} (V_{1})$ . Since $α$ is isometric and matrix affine, $Φ$ is well-defined. Clearly, $Φ$ is unital and surjective. On the level of matrices, we have ${(Φ_{n} (F^{(n)}))}_{r} (φ) = F_{r}^{(n)} (α_{r} (φ))$ for $F^{(n)} \in M_{n} (A (C S (V_{2})))$ and $φ \in C S_{r} (V_{1})$ . Since $F^{(n)} \geq 0$ in $M_{n} (A (C S (V_{2})))$ if and only if $F_{r}^{(n)} (φ) \geq 0$ for all $r \in N$ and $φ \in C S_{r} (V_{2})$ , $Φ$ is a unital complete order isomorphism from $A (C S (V_{2}))$ onto $A (C S (V_{1}))$ . Since $ℒ_{1}$ and $ℒ_{2}$ are closed, $Φ$ is a unital complete order isomorphism from $V_{2}$ onto $V_{1}$ by Lemma 5.4 . That $α$ is completely isometric for $D_{ℒ_{1}}$ and $D_{ℒ_{2}}$ implies that $L_{D_{L_{1, n}}} (Φ_{n} (a_{2})) = L_{D_{L_{2, n}}} (a_{2})$ for all $a_{2} \in M_{n} (V_{2})$ and $n \in N$ . Because $ℒ_{1}$ and $ℒ_{2}$ are closed, they are lower semicontinuous, so that $ℒ_{D_{ℒ_{1}}} = ℒ_{1}$ on $V_{1}$ by Theorem 4.4 in [21] , and similarly for $V_{2}$ . Thus $Φ^{- 1}$ is a unital complete order isomorphism from $V_{1}$ onto $V_{2}$ such that $ℒ_{1} = ℒ_{2} \circ Ψ$ . □

Theorem 5.6. Suppose

(V_{1}, ℒ_{1})

and

(V_{2}, ℒ_{2})

are quantized metric spaces.

d i s t_{N C} (V_{1}, V_{2}) = 0,

then there exists a complete isometry

Φ

from

(V_{1}, ℒ_{1})

onto

(V_{2}, ℒ_{2})

Proof. Since

d i s t_{N C} (V_{1}, V_{2}) = 0

, we have

d i s t_{N C} (V_{1}^{c}, V_{2}^{c}) = 0

by Lemma 5.2 and Theorem 4.7 . From that

d i s t_{N C} (V_{1}^{c}, V_{2}^{c}) = 0

, there is a sequence

{ℒ^{(k)}} \subseteq ℳ (V_{1}^{c} \oplus V_{2}^{c})

of matrix Lip-norms such that

{sup}_{n \in N} {n^{- 2} d i s t_{H}^{D_{L_{n}^{(k)}}} (C S_{n} (V_{1}^{c}), C S_{n} (V_{2}^{c}))} < \frac{1}{k} .

So for each

n \in N

, we have

n^{- 2} d i s t_{H}^{D_{L_{n}^{(k)}}} (C S_{n} (V_{1}^{c}), C S_{n} (V_{2}^{c})) < \frac{1}{k} .

And for

φ, ψ \in C S_{n} (V_{i}^{c})

i = 1, 2

, by Proposition 3.6 we have

D_{L_{n}^{(k)}} ((π_{i})_{n}^{c} (φ), (π_{i})_{n}^{c} (ψ)) = D_{L_{i, n}^{c}} (φ, ψ),

where

π_{i}, i = 1, 2

, is the projection from

V_{1}^{c} \oplus V_{2}^{c}

onto

V_{i}^{c}

. Therefore, for each

n \in N

we get

d i s t_{G H} (C S_{n} (V_{1}^{c}), C S_{n} (V_{2}^{c})) = 0,

where

d i s t_{G H} (C S_{n} (V_{1}^{c}), C S_{n} (V_{2}^{c}))

is the Gromov-Hausdorff distance(see Definition 3.4 in [6] ) between

(C S_{n} (V_{1}^{c}), D_{L_{1, n}^{c}})

and

(C S_{n} (V_{2}^{c}), D_{L_{2, n}^{c}})

. As in the proofs of Theorem 7.6 and Theorem 7.7 in [17] , there is a subsequence

{D_{L_{1}^{(k_{j_{1}})}}}

which converges uniformly on the disjoint union

C S_{1} (V_{1}^{c}) ⊔ C S_{1} (V_{2}^{c})

to a semi-metric

σ_{1}

and

σ_{1}

determines an isometry

α_{1}

from

C S_{1} (V_{1}^{c})

onto

C S_{1} (V_{2}^{c})

by the condition that

σ_{1} (φ, α_{1} (φ)) = 0

. Similarly, there is a subsequence

{D_{L_{2}^{(k_{j_{1}, j_{2}})}}}

{D_{L_{2}^{(k_{j_{1}})}}}

which converges uniformly on

C S_{2} (V_{1}^{c}) ⊔ C S_{2} (V_{2}^{c})

to a semi-metric

σ_{2}

and

σ_{2}

determines an isometry

α_{2}

from

C S_{2} (V_{1}^{c})

onto

C S_{2} (V_{2}^{c})

by the condition that

σ_{2} (φ, α_{2} (φ)) = 0

. In general, once

{D_{L_{2}^{(k_{j_{1}})}}}, {D_{L_{2}^{(k_{j_{1}, j_{2}})}}}, \dots, {D_{L_{2}^{(k_{j_{1}, j_{2}, \dots, j_{n}})}}}

have been choosen, there is a subsequence

{D_{L_{2}^{(k_{j_{1}, j_{2}, \dots, j_{n}, j_{n + 1}})}}}

{D_{L_{2}^{(k_{j_{1}, j_{2}, \dots, j_{n}})}}}

which converges uniformly on

C S_{n + 1} (V_{1}^{c}) ⊔ C S_{n + 1} (V_{2}^{c})

to a semi-metric

σ_{n + 1}

and

σ_{n + 1}

determines an isometry

α_{n + 1}

from

C S_{n + 1} (V_{1}^{c})

onto

C S_{n + 1} (V_{2}^{c})

by the condition that

σ_{n + 1} (φ, α_{n + 1} (φ)) = 0

Given

φ_{i} \in C S_{n_{i}} (V_{1}^{c})

and

γ_{i} \in M_{n_{i}, n}

i = 1, 2, \dots, m

, satisfying

\sum_{i = 1}^{m} γ_{i}^{*} γ_{i} = 1_{n}

, and

ε > 0

. Let

s = max {n, n_{1}, \dots, n_{m}}

. We can find

K \in N

such that if

k_{j_{1}, \dots, j_{s}} > K

then

‖ σ_{l} - D_{L_{l}^{(k_{j_{1}, \dots, j_{s}})}} ‖ < \frac{ε}{2}

for

l = n, n_{1}, \dots, n_{m}

. Now for

k_{j_{1}, \dots, j_{s}} > K

we have

\begin{matrix} σ_{n} (\sum_{i = 1}^{m} γ_{i}^{*} φ_{i} γ_{i}, \sum_{i = 1}^{m} γ_{i}^{*} α_{n_{i}} (φ_{i}) γ_{i}) \end{matrix}

\begin{matrix} \leq & D_{L_{n}^{(k_{j_{1}, \dots, j_{s}})}} (\sum_{i = 1}^{m} γ_{i}^{*} φ_{i} γ_{i}, \sum_{i = 1}^{m} γ_{i}^{*} α_{n_{i}} (φ_{i}) γ_{i}) + \frac{ε}{2} \end{matrix}

\begin{matrix} \leq & D_{L_{n_{1} + \dots + n_{m}}^{(k_{j_{1}, \dots, j_{s}})}} (φ_{1} \oplus \dots \oplus φ_{m}, α_{n_{1}} (φ_{1}) \oplus \dots \oplus α_{n_{m}} (φ_{m})) + \frac{ε}{2} \end{matrix}

\begin{matrix} = & max {D_{L_{n_{1}}^{(k_{j_{1}, \dots, j_{s}})}} (φ_{1}, α_{n_{1}} (φ_{1})), \dots, D_{L_{n_{m}^{(k_{j_{1}, \dots, j_{s}})}}} (φ_{m}, α_{n_{m}} (φ_{m}))} + \frac{ε}{2} \end{matrix}

\begin{matrix} < & max {σ_{n_{1}} (φ_{1}, α_{n_{1}} (φ_{1})) + \frac{ε}{2}, \dots, σ_{n_{m}} (φ_{m}, α_{n_{m}} (φ_{m})) + \frac{ε}{2}} + \frac{ε}{2} \end{matrix}

\begin{matrix} = & ε . \end{matrix}

Since

ε

is arbitrary, we have

σ_{n} (\sum_{i = 1}^{m} γ_{i}^{*} φ_{i} γ_{i}, \sum_{i = 1}^{m} γ_{i}^{*} α_{n_{i}} (φ_{i}) γ_{i}) = 0

. But

σ_{n} (\sum_{i = 1}^{m} γ_{i}^{*} φ_{i} γ_{i}, α_{n} (\sum_{i = 1}^{m} γ_{i}^{*} φ_{i} γ_{i})) = 0 .

By Lemma 7.4 in [17] , we obtain

α_{n} (\sum_{i = 1}^{m} γ_{i}^{*} φ_{i} γ_{i}) = \sum_{i = 1}^{m} γ_{i}^{*} α_{n_{i}} (φ_{i}) γ_{i} .

α = (α_{n})

is matrix affine.

Now, by Lemma 5.5 we conclude that there exists a unital complete order isomorphism

Φ

from

V_{1}^{c}

onto

V_{2}^{c}

such that

ℒ_{1}^{c} = ℒ_{2}^{c} \circ Φ

, that is,

Φ

is a complete isometry from

(V_{1}, ℒ_{1})

onto

(V_{2}, ℒ_{2})

. □

6 Completeness

For the metric space of complete isometry classes of quantized metric spaces with the quantized Gromov-Hausdorff distance, we show in this section that it is complete.

Let

{(V_{i}, 1)}

be a sequence of matrix order unit space. We will denote by

\oplus_{i \in N} V_{i}

the operator space direct sum that is formed of all sequences

{a_{i}}

with

a_{i} \in V_{i}

and

{sup}_{i \in N} ∥ a_{i} ∥ < + \infty

, and by

\oplus_{i = 1}^{n} V_{i}

the operator space direct sum of

V_{1}, V_{2}, \dots, V_{n}

(see §2.6 in [13] ). Then

(\oplus_{i \in N} V_{i}, {1})

and

(\oplus_{i = 1}^{n} V_{i}, ({\underset{︸}{1, \dots, 1}}_{n}))

are matrix order unit spaces.

Suppose we have a sequence

{(V_{i}, ℒ_{i})}

of quantized metric spaces. Suppose further that we have a sequence

{ℳ_{i}}

of matrix Lip-norms with

ℳ_{i} \in ℳ (ℒ_{i}, ℒ_{i + 1})

Define

Q = (Q_{k})

\prod_{i \in N} V_{i}

, the full product, by

Q_{k} ({a_{i}}) = {sup}_{i \in N} {M_{i, k} (a_{i}, a_{i + 1})}, {a_{i}} \in M_{k} (\prod_{i \in N} V_{i}),

and set

ℰ_{1} = {{a_{i}} \in \oplus_{i \in N} V_{i} : Q_{1} ({a_{i}}) < + \infty} .

It is easy to check that

ℰ_{1}

is a self-adjoint subspace of

\oplus_{i \in N} V_{i}

containing

{1}

, and so is a matix order unit space, and that

Q

is a matrix Lipschitz seminorm on

ℰ_{1}

For the evident identifications, we have

C S (V_{i}) \subseteq C S (\oplus_{j = 1}^{i} V_{j}) \subseteq C S (\oplus_{j = 1}^{n} V_{j}) \subseteq C S (ℰ_{1}), 0 \leq i \leq n .

Given a family of graded sets

S_{i} = (S_{i, n}), i \in I

. We denote by

\cup_{i \in I} S_{i}

the graded set

(\cup_{i \in I} S_{i, n})

. If

S = (S_{n})

is a graded set in a vector space, we denote by

m c o (S)

the matrix convex hull of

S

. Let

Z = (Z_{n}) = m c o (\cup_{i \in N} C S (V_{i}))

U = \cup_{n \in N} C S (\oplus_{j = 1}^{n} V_{j})

and

W_{n} = m c o (\cup_{i = 1}^{n} C S (V_{i}))

Proposition 6.1.

Z

and

U

are BW-dense in

C S (ℰ_{1})

. In particular,

W_{n}

is BW-dense in

C S (\oplus_{j = 1}^{n} V_{j})

for

n \in N

Proof. Since the matrix polar

\begin{matrix} Z_{n}^{π} & = & {a \in M_{n} (ℰ_{1}) : R e ≪ a,, φ ≫ \leq 1_{r \times n} for all φ \in Z_{r}, r \in N} \end{matrix}

\begin{matrix} = & {a \in M_{n} (ℰ_{1}) : φ_{n} (R e (a) \leq 1_{r \times n} for all φ \in Z_{r}, r \in N} \end{matrix}

\begin{matrix} = & {a \in M_{n} (ℰ_{1}) : φ_{n} (1_{n} - R e (a)) \geq 0 for all φ \in Z_{r}, r \in N} \end{matrix}

\begin{matrix} = & {a \in M_{n} (ℰ_{1}) : 1_{n} - R e (a) \geq 0} \end{matrix}

\begin{matrix} = & {a \in M_{n} (ℰ_{1}) : ∥ a_{+} ∥ \leq 1}, \end{matrix}

we have

\begin{matrix} Z_{n}^{π π} & = & {f \in M_{n} (ℰ_{1}^{*}) : R e ≪ f, a ≫ \leq 1_{r \times n} \end{matrix}

\begin{matrix} when 1_{r} - R e (a) \geq 0, a \in M_{r} (\oplus_{i \in N} V_{i}), r \in N} . \end{matrix}

For

f \in Z_{n}^{π π}, λ \in R

and

a = a^{*} \in M_{r} (ℰ_{1})

, we have that

1_{r} - R e (i λ a) \geq 0

, and so

R e ≪ i λ a, f ≫ \leq 1_{r \times n}

. Thus

0 = R e ≪ i a, f ≫ = R e (i ≪ a, f ≫) = - I m ≪ a, f ≫,

that is,

I m ≪ a, f ≫ = 0

. If

a \in M_{r} (ℰ_{1}), a \geq 0

and

λ \leq 0

, then

1_{r} - R e (λ a) \geq 0

and so

≪ λ a, f ≫ = R e ≪ λ a, f ≫ \leq 1_{r \times n}

. Thus

≪ a, f ≫ \geq 0

. Clearly,

≪ 1_{r}, f ≫ \leq 1_{r \times n}

. Therefore,

Z_{n}^{π π} = {f \in M_{n} (ℰ_{1}^{*}) : f is completely positive and f (1) \leq 1_{n}} .

By the bipolar theorem in matrix convexity (see Corollary 5.5 in [5] ),

m c o (Z \cup {0})

is BW-dense in

Z^{π π} = (Z_{n}^{π π})

. Evaluting the completely positive mappings at

1

, we see that

Z

is BW-dense in

C S (ℰ_{1})

. Because

Z \subseteq U \subseteq C S (ℰ_{1})

U

is BW-dense in

C S (ℰ_{1})

. □

Define

P_{n} = (P_{n, k})

\oplus_{i = 1}^{n} V_{i}

P_{n, k} (a_{1}, \dots, a_{n}) = max {M_{i, k} (a_{i}, a_{i + 1}) : 1 \leq i \leq n - 1},

for

(a_{1}, \dots, a_{n}) \in M_{k} (\oplus_{i = 1}^{n} V_{i})

. Similar to the proof of Proposition 4.3 , we have

Proposition 6.2.

P_{n}

is a matrix Lip-norm on

\oplus_{i = 1}^{n} V_{i}

, and induces

ℒ_{j}, 1 \leq j \leq n

and

ℳ_{i}

and

P_{i}

1 \leq i \leq n - 1

, via the evident projections.

For

b \in M_{i} (\oplus_{j = 1}^{n} V_{j})

and

ε > 0

, set

b_{n} = b

. Since

P_{n + 1}

induces

P_{n}

, we can find

b_{n + 1} \in M_{i} (\oplus_{j = 1}^{n + 1} V_{j})

such that

(π_{n})_{i} (b_{n + 1}) = b_{n}

and

P_{n + 1, i} (b_{n + 1}) < P_{n, i} (b_{n}) + \frac{ε}{2^{n}}

, where

π_{n}

is the evident projection from

\oplus_{j = 1}^{n + 1} V_{j}

onto

\oplus_{j = 1}^{n} V_{j}

. Similarly, we can find

b_{n + 2} \in M_{i} (\oplus_{j = 1}^{n + 2} V_{j})

such that

(π_{n + 1})_{i} (b_{n + 2}) = b_{n + 1}

and

P_{n + 2, i} (b_{n + 2}) < P_{n + 1, i} (b_{n + 1}) + \frac{ε}{2^{n + 1}}

. Continuing in this way, for

t \geq n

we get

b_{t + 1} \in M_{i} (\oplus_{j = 1}^{t + 1} V_{j})

such that

(π_{t})_{i} (b_{t + 1}) = b_{t}

and

P_{t + 1, i} (b_{t + 1}) < P_{t, i} (b_{t}) + \frac{ε}{2^{t}}

. We let

c = {c_{j}}

be the unique element of

M_{i} (\prod_{j \in N} V_{j})

such that

(π_{t})_{i} (c) = b_{t}

for

t \geq n

. Then

Q_{i} (c) \leq P_{n, i} (b) + ε

. So,

P_{n, i} (b) = Q_{\oplus_{j = 1}^{n} V_{j}, i} (b)

. Set

d_{k} = (c_{1}, \dots, c_{k}), k \in N

Since

P_{n}

induces

ℳ_{i} (1 \leq i \leq n - 1)

, via the evidence projections,

d i s t_{H}^{D_{P_{n, i}}} (C S_{i} (V_{k}), C S_{i} (V_{k + 1})) = d i s t_{H}^{D_{M_{k, i}}} (C S_{i} (V_{k}), C S_{i} (V_{k + 1})) .

For

m, n \in N

with

m < n

and

φ_{n} \in C S_{i} (V_{n})

, we can find

φ_{n - 1} \in C S_{i} (V_{n - 1})

with

D_{P_{n, i}} (φ_{n - 1}, φ_{n}) \leq d i s t_{H}^{D_{M_{n - 1, i}}} (C S_{i} (V_{n - 1}), C S_{i} (V_{n})) .

Similarly, we can find

φ_{n - 2} \in C S_{i} (V_{n - 2})

with

D_{P_{n, i}} (φ_{n - 2}, φ_{n - 1}) \leq d i s t_{H}^{D_{M_{n - 2, i}}} (C S_{i} (V_{n - 2}), C S_{i} (V_{n - 1})) .

Inductively, we can find

φ_{m}, \dots, φ_{n - 1}

with

φ_{k} \in C S_{i} (V_{k})

and

D_{P_{n, i}} (φ_{k}, φ_{k + 1}) \leq d i s t_{H}^{D_{M_{k, i}}} (C S_{i} (V_{k}), C S_{i} (V_{k + 1})),

for

m \leq k \leq n - 1

. Consequently,

D_{P_{n, i}} (φ_{m}, φ_{n}) \leq \sum_{j = m}^{n - 1} d i s t_{H}^{D_{M_{j, i}}} (C S_{i} (V_{j}), C S_{i} (V_{j + 1})), 2 \leq k \leq n - 1 .

Similarly, for

φ_{m} \in C S_{i} (V_{m})

we can find a

φ_{n} \in C S_{i} (V_{n})

such that the inequality above holds. Thus by Proposition 6.1, we have

d i s t_{H}^{D_{P_{n, i}}} (C S_{i} (\oplus_{j = 1}^{m} V_{j}), C S_{i} (V_{n})) \leq \sum_{j = m}^{n - 1} d i s t_{H}^{D_{M_{j, i}}} (C S_{i} (V_{j}), C S_{i} (V_{j + 1})) .

Proposition 6.3. For

m < n

, we have

d i s t_{H}^{D_{P_{n, i}}} (C S_{i} (\oplus_{j = 1}^{m} V_{j}), C S_{i} (\oplus_{j = 1}^{n} V_{j})) \leq \sum_{k = 1}^{i} \sum_{j = m}^{n - 1} d i s t_{H}^{D_{M_{j, k}}} (C S_{k} (V_{j}), C S_{k} (V_{j + 1})) .

Proof. For

φ \in C S_{i} (\oplus_{j = 1}^{m} V_{j})

, we can find a

ψ \in C S_{i} (V_{n}) \subseteq C S_{i} (\oplus_{j = 1}^{n} V_{j})

such that

D_{P_{n, i}} (φ, ψ) \leq \sum_{k = 1}^{i} \sum_{j = m}^{n - 1} d i s t_{H}^{D_{M_{j, k}}} (C S_{k} (V_{j}), C S_{k} (V_{j + 1}))

from the discussion before the proposition. Suppose

φ \in C S_{i} (\oplus_{j = 1}^{n} V_{j})

and

ε > 0

. For each

i \in N

\cup_{j = 1}^{n} C S_{i} (V_{j})

is a BW-closed subset of

C S_{i} (\oplus_{j = 1}^{n} V_{j})

, and

γ^{*} (\cup_{j = 1}^{n} C S_{i} (V_{j})) γ \subseteq \cup_{j = 1}^{n} C S_{k} (V_{j})

for all isometries

γ \in M_{i, k}

. From Proposition 6.1 , the BW-closure

\bar{m c o} (\cup_{j = 1}^{n} C S_{i} (V_{j}))

m c o (\cup_{j = 1}^{n} C S_{i} (V_{j}))

C S (\oplus_{j = 1}^{n} V_{j})

, and so by Theorem 4.6 and Theorem 4.3 in [19] , there exist

φ_{k} \in C S_{l_{k}} (V_{j_{k}})

and

γ_{k} \in M_{l_{k}, i}

for

k = 1, 2, \dots, s, 1 \leq j_{k} \leq n, 1 \leq l_{k} \leq i

satisfying

\sum_{k = 1}^{s} γ_{k}^{*} γ_{k} = 1_{i}

such that

D_{P_{n, i}} (φ, \sum_{k = 1}^{s} γ_{k}^{*} φ_{k} γ_{k}) < ε .

For each

φ_{k}

, we can find

ψ_{k} \in C S_{l_{k}} (\oplus_{j = 1}^{m} V_{j})

so that

D_{P_{n, l_{k}}} (φ_{k}, ψ_{k}) \leq \sum_{j = m}^{n - 1} d i s t_{H}^{D_{M_{j, l_{k}}}} (C S_{l_{k}} (V_{j}), C S_{l_{k}} (V_{j + 1})) .

Thus

\begin{matrix} D_{P_{n, i}} (φ, \sum_{k = 1}^{s} γ_{k}^{*} ψ_{k} γ) \end{matrix}

\begin{matrix} \leq & D_{P_{n, i}} (φ, \sum_{k = 1}^{s} γ_{k}^{*} φ_{k} γ) + D_{P_{n, i}} (\sum_{k = 1}^{s} γ_{k}^{*} φ_{k} γ, \sum_{k = 1}^{s} γ_{k}^{*} ψ_{k} γ) \end{matrix}

\begin{matrix} < & ε + D_{P_{n, i}} (\sum_{k = 1}^{s} γ_{k}^{*} φ_{k} γ, \sum_{k = 1}^{s} γ_{k}^{*} ψ_{k} γ) \end{matrix}

\begin{matrix} \leq & ε + D_{P_{n, \sum_{k = 1}^{s} l_{k}}} (φ_{1} \oplus \dots \oplus φ_{s}, ψ_{1} \oplus \dots \oplus ψ_{s}) \end{matrix}

\begin{matrix} = & ε + max {D_{P_{n, l_{1}}} (φ_{1}, ψ_{1}), \dots, D_{P_{n, l_{s}}} (φ_{s}, ψ_{s})} \end{matrix}

\begin{matrix} \leq & ε + \sum_{k = 1}^{i} \sum_{j = m}^{n - 1} d i s t_{H}^{D_{M_{j, k}}} (C S_{k} (V_{j}), C S_{k} (V_{j + 1})), \end{matrix}

because

D_{P_{n}} = (D_{P_{n, k}})

is a matrix metric (see Example 5.2 in [21] ). Since

ε

is arbitrary, we obtain the desired inequality. □

Now for

φ, ψ \in C S_{i} (\oplus_{j = 1}^{n} V_{j})

, there are

φ_{1}, ψ_{1} \in C S_{i} (V_{1})

with

D_{P_{n, i}} (φ_{1}, φ) \leq \sum_{j = 1}^{n - 1} d i s t_{H}^{D_{M_{j, i}}} (C S_{i} (V_{j}), C S_{i} (V_{j + 1})),

and

D_{P_{n, i}} (ψ_{1}, ψ) \leq \sum_{j = 1}^{n - 1} d i s t_{H}^{D_{M_{j, i}}} (C S_{i} (V_{j}), C S_{i} (V_{j + 1})) .

\begin{matrix} D_{P_{n, i}} (φ, ψ) & \leq & D_{P_{n, i}} (φ, φ_{1}) + D_{P_{n, i}} (φ_{1}, ψ_{1}) + D_{P_{n, i}} (ψ_{1}, ψ) \end{matrix}

\begin{matrix} \leq & d i a m (C S_{i} (V_{1}), D_{L_{1, i}}) + 2 \sum_{j = 1}^{n - 1} d i s t_{H}^{D_{M_{j, i}}} (C S_{i} (V_{j}), C S_{i} (V_{j + 1})) \end{matrix}

\begin{matrix} \overset{△}{=} & h_{i}, \end{matrix}

where

d i a m (C S_{i} (V_{1}), D_{L_{1, i}})

is the diameter of

C S_{i} (V_{1})

with respect to

D_{L_{1, i}}

By Proposition 5.2 and Proposition 3.8 in [20] , we have

∥ {\tilde{d}}_{n} ∥_{i}^{\sim} \leq h_{i} P_{n, i} (d_{n}) \leq h_{i} Q_{i} (c) < h_{i} Q_{i} (c) + ε,

where

ε > 0

. So there is an

α_{n, i} = [α_{s t}^{(n, i)}] \in M_{i}

such that

{‖ d_{n} - [α_{s t}^{(n, i)} (1, \dots, 1)] ‖}_{i} \leq h_{i} Q_{i} (c) + ε, n \in N .

Set

G_{n, i} = {β_{n, i} = [β_{s t}^{(n, i)}] \in M_{i} : {‖ d_{n} - [β_{s t}^{(n, i)} (1, \dots, 1)] ‖}_{i} \leq h_{i} Q_{i} (c) + ε} .

Then

G_{n, i}

is a non-empty closed bounded subset of

M_{i}

. Clearly,

G_{n + 1, i} \subseteq G_{n, i}

So there exists a

β_{0} \in \cap_{n = 1}^{\infty} G_{n, i}

. We have

∥ d_{n} ∥_{i} \leq ∥ β_{0} ∥ + h_{i} Q_{i} (c) + ε, n \in N .

Thus

c \in M_{i} (\oplus_{j \in N} V_{j})

, and we obtain

Proposition 6.4. For

n \in N

Q

induces

P_{n}

via the evident projection.

Theorem 6.5. The metric space

ℜ

of complete isometry classes of quantized metric spaces, with the metric

d i s t_{N C}

, is complete.

Proof. Let

{(V_{n}, ℒ_{n})}

be a sequence in

ℜ

which is Cauchy with respect to the quantized Gromov-Hausdorff distance

d i s t_{N C}

. To show that

{(V_{n}, ℒ_{n})}

converges it suffices to show that a subsequence converges. Since

{(V_{n}, ℒ_{n})}

is Cauchy, we can choose a subsequence, still denoted by

{(V_{n}, ℒ_{n})}

, such that

d i s t_{N C} (V_{n}, V_{n + 1}) < \frac{1}{2^{n}},

for all

n \in N

. By definition, there exist

ℳ_{n} = (M_{n, k}) \in ℳ (ℒ_{n}, ℒ_{n + 1})

with

{sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{M_{n, k}}} (C S_{k} (V_{n}), C S_{k} (V_{n + 1}))} < \frac{1}{2^{n}},

for all

n \in N

. It follows that

\sum_{n = 1}^{\infty} {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{M_{n, k}}} (C S_{k} (V_{n}), C S_{k} (V_{n + 1}))} < + \infty .

Let

ε > 0

be given. Then there is an

m \in N

such that

\sum_{n = m}^{\infty} {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{M_{n, k}}} (C S_{k} (V_{n}), C S_{k} (V_{n + 1}))} < ε .

By Proposition 6.2 , Proposition 6.3 and Proposition 6.4 , we have

\begin{matrix} d i s t_{H}^{D_{Q_{i}}} (C S_{i} (\oplus_{j = 1}^{m} V_{j}), C S_{i} (\oplus_{j = 1}^{n} V_{j})) \end{matrix}

\begin{matrix} \leq & \sum_{k = 1}^{i} \sum_{j = m}^{n - 1} d i s t_{H}^{D_{M_{j, k}}} (C S_{k} (V_{j}), C S_{k} (V_{j + 1})) \end{matrix}

\begin{matrix} \leq & \sum_{k = 1}^{i} k^{2} \sum_{j = m}^{n - 1} {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{M_{j, k}}} (C S_{k} (V_{j}), C S_{k} (V_{j + 1}))} \end{matrix}

\begin{matrix} < & (\sum_{k = 1}^{i} k^{2}) ε, \end{matrix}

for

n > m

. This says that

C S_{i} (\oplus_{j = 1}^{m} V_{j})

(\sum_{k = 1}^{i} k^{2}) ε

-dense for

D_{Q_{i}}

Z_{i}

. But

C S_{i} (\oplus_{j = 1}^{m} V_{j})

is BW-compact for the topology from

D_{Q_{i}} = D_{P_{m, i}}

by Proposition 6.2 . Thus

C S_{i} (\oplus_{j = 1}^{m} V_{j})

is totally bounded for

D_{Q_{i}}

, and so

Z_{i}

is totally bounded for

D_{Q_{i}}

. Let

\hat{Z} = ({\hat{Z}}_{n})

be the completion of

Z

for

D_{Q}

. We let

D_{Q}

denote also the extension of

D_{Q}

\hat{Z}

. Then

\hat{Z}

is a compact matrix convex set. For

{a_{i}} \in M_{n} (ℰ_{1})

, we have

≪ \sum_{j = 1}^{m} γ_{j}^{*} φ_{j} γ_{j}, {a_{i}} ≫ = \sum_{j = 1}^{m} (γ_{j} \otimes 1_{n})^{*} ≪ φ_{j}, {a_{i}} ≫ (γ_{j} \otimes 1_{n}),

and

\begin{matrix} ‖ ≪ \sum_{j = 1}^{m} γ_{j}^{*} φ_{j} γ_{j}, {a_{i}} ≫ - ≪ \sum_{k = 1}^{p} λ_{k}^{*} ψ_{k} λ_{k}, {a_{i}} ≫ ‖ \end{matrix}

\begin{matrix} \leq & L_{D_{Q_{n}}} ({a_{i}}) D_{Q_{r}} (\sum_{j = 1}^{m} γ_{j}^{*} φ_{j} γ_{j}, \sum_{k = 1}^{p} λ_{k}^{*} ψ_{k} λ_{k}) \end{matrix}

\begin{matrix} \leq & Q_{n} ({a_{i}}) D_{Q_{r}} (\sum_{j = 1}^{m} γ_{j}^{*} φ_{j} γ_{j}, \sum_{k = 1}^{p} λ_{k}^{*} ψ_{k} λ_{k}), \end{matrix}

where

φ_{j} \in C S_{n_{j}} (V_{q_{j}})

ψ_{k} \in C S_{m_{k}} (V_{l_{k}})

, and

γ_{j} \in M_{n_{j}, r}

, and

λ_{k} \in M_{m_{k}, r}

satisfying

\sum_{j = 1}^{m} γ_{j}^{*} γ_{j} = 1_{r}

and

\sum_{k = 1}^{p} λ_{k}^{*} λ_{k} = 1_{r}

. So the map

Φ : ℰ_{1} \mapsto A (Z)

, given by

(Φ ({a_{i}})) (\sum_{j = 1}^{m} γ_{j}^{*} φ_{j} γ_{j}) = \sum_{j = 1}^{m} γ_{j}^{*} φ_{j} (a_{q_{j}}) γ_{j},

for

{a_{i}} \in ℰ_{1}

φ_{j} \in C S_{n_{j}} (V_{q_{j}})

and

γ_{j} \in M_{n_{j}, r}

satisfying

\sum_{j = 1}^{m} γ_{j}^{*} γ_{j} = 1_{r}

, is well-defined and

Φ ({a_{i}})

can be extended to an element

\hat{Φ ({a_{i}})} \in A (\hat{Z})

. Moreover if

{a_{i}} \geq 0

ℰ_{1}

then

\hat{Φ ({a_{i}})} \geq 0

A (\hat{Z})

and

\hat{Φ ({1})} = I

. Thus

ℰ_{1}

can be regarded as a matrix order unit subspace of

A (\hat{Z})

Define the map

Ψ_{r} : {\hat{Z}}_{r} \mapsto C S_{r} (ℰ_{1}), r \in N

, by

Ψ_{r} (z) ({a_{i}}) = \hat{Ψ ({a_{i}})} (z),

for

z \in {\hat{Z}}_{r}

and

{a_{i}} \in ℰ_{1}

. Clearly,

Ψ

is continuous. For

z = \sum_{j = 1}^{m} γ_{j}^{*} φ_{j} γ_{j} \in Z_{r}

with

φ_{j} \in C S_{n_{j}} (V_{q_{j}})

γ_{j} \in M_{n_{j}, r}

satisfying

\sum_{j = 1}^{m} γ_{j}^{*} γ_{j} = 1_{r}

, we have

Ψ_{r} (z) ({a_{i}}) = \hat{Ψ ({a_{i}})} (z) = Ψ ({a_{i}}) (z) = z ({a_{i}}),

that is,

Ψ_{r} (z) = z

. Since

Z_{r}

is dense in

C S_{r} (ℰ_{1})

and

{\hat{Z}}_{r}

is compact, we obtain that

Ψ_{r} ({\hat{Z}}_{r}) = C S_{r} (ℰ_{1})

z_{1}, z_{2} \in {\hat{Z}}_{r}

with

z_{1} \neq z_{2}

and

k = D_{Q_{r}} (z_{1}, z_{2})

, we can find

y_{1}, y_{2} \in Z_{r}

such that

D_{Q_{r}} (z_{i}, y_{i}) < \frac{k}{4}, i = 1, 2

. Thus

D_{Q_{r}} (y_{1}, y_{2}) > \frac{k}{2}

. So we can find

{w_{i}} \in M_{r} (ℰ_{1})

with

Q_{r} ({w_{i}}) \leq 1

and

∥ ≪ {w_{i}}, y_{1} ≫ - ≪ {w_{i}}, y_{2} ≫ ∥ > \frac{k}{2}

. But

L_{D_{Q_{r}}} (\hat{Φ_{r} ({w_{i}})} \leq 1

so that

∥ ≪ \hat{Φ_{r} ({w_{i}})}, z_{i} ≫ - ≪ \hat{Φ_{r} ({w_{i}})}, y_{i} ≫ ∥ < \frac{k}{4}, i = 1, 2

. Thus

∥ ≪ \hat{Φ_{r} ({w_{i}})}, z_{1} ≫ - ≪ \hat{Φ_{r} ({w_{i}})}, z_{2} ≫ ∥ > 0

. Therefore,

Ψ_{r}

is injective. So

Ψ_{r}

is a homeomorphism of

{\hat{Z}}_{r}

onto

C S_{r} (ℰ_{1})

for

r \in N

. From this we see that the

D_{Q}

-topology on

C S (ℰ_{1})

agrees with the BW-topology. Hence

Q

is a matrix Lip-norm on

ℰ_{1}

By Proposition 6.2 and Proposition 6.4 , we obtain

\sum_{n = 1}^{\infty} {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{Q_{k}}} (C S_{k} (V_{n}), C S_{k} (V_{n + 1}))} < + \infty,

which indicate that, for

k \in N

{C S_{k} (V_{n})}

is a Cauchy sequence for

d i s t_{H}^{D_{Q_{k}}}

, and has a limit

K_{k} \subseteq C S_{k} (ℰ_{1})

. Clearly

K = (K_{k})

is a compact matrix convex set.

Because

ℰ_{1}

is completely order isomorphic to a dense subspce of

A (C S (ℰ_{1}))

(Proposition 6.1(1) in [21] ), we can view

ℰ_{1}

as a dence subspace of

A (C S (ℰ_{1}))

. Let

φ

be the map which restricts the elements of

A (C S (ℰ_{1}))

K

and

V = φ (ℰ_{1})

. Then

(V, Q_{V})

is a quantized metric space.

Given

ε > 0

. Then there is an

N \in N

such that

\sum_{j = n}^{\infty} {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{Q_{k}}} (C S_{k} (V_{j}), C S_{k} (V_{j + 1}))} < ε, n \geq N .

For

k, p \in N

, we have

\begin{matrix} k^{- 2} d i s t_{H}^{D_{Q_{k}}} (C S_{k} (V_{n}), C S_{k} (V_{n + p})) \end{matrix}

\begin{matrix} \leq & \sum_{j = n}^{n + p - 1} k^{- 2} d i s t_{H}^{D_{Q_{k}}} (C S_{k} (V_{j}), C S_{k} (V_{j + 1})) \end{matrix}

\begin{matrix} \leq & \sum_{j = n}^{\infty} {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{Q_{k}}} (C S_{k} (V_{j}), C S_{k} (V_{j + 1}))} \end{matrix}

\begin{matrix} < & ε, \end{matrix}

for

n \geq N

. Letting

p \to + \infty

, we obtain

k^{- 2} d i s t_{H}^{D_{Q_{k}}} (C S_{k} (V_{n}), K_{k}) \leq ε,

for

k \in N

, and so

{sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{Q_{k}}} (C S_{k} (V_{n}), K_{k})} \leq ε

. By Proposition 4.8 , for

n \geq N

we have

d i s t_{N C} (V_{n}, V) \leq {sup}_{k \in N} {k^{- 2} d i s t_{H}^{D_{Q_{k}}} (C S_{k} (V_{n}), K_{k})} \leq ε .

Therefore,

{lim}_{n \to \infty} d i s t_{N C} (V_{n}, V) = 0

. □

7 Matrix approximability

In this section, we establish a matrix approximability theorem for

1

-exact matrix order unit spaces.

Lemma 7.1. Let

(V, ℒ)

be a quantized metric space and let

x = [x_{s t}] \in M_{k} (V)

x_{s t} = x_{s t}^{(1)} + i x_{s t}^{(2)}

with

(x_{s t}^{(p)})^{*} = x_{s t}^{(p)}

for

p = 1, 2, s, t = 1, 2, \dots, k

Suppose

λ_{s t}^{(p)} \in σ (x_{s t}^{(p)})

for

p = 1, 2, s, t = 1, 2, \dots, k

. Then

∥ x - [(λ_{s t}^{(1)} + i λ_{s t}^{(2)}) 1] ∥_{k} \leq 2 k^{2} L_{k} (x) d i a m (V, ℒ)

Proof. By Proposition 2.11 in [8] , we have

\begin{matrix} {‖ x - [(λ_{s t}^{(1)} + i λ_{s t}^{(2)}) 1] ‖}_{k} & \leq & {‖ [x_{s t}^{(1)} - λ_{s t}^{(1)} 1] ‖}_{k} + {‖ [x_{s t}^{(2)} - λ_{s t}^{(2)} 1] ‖}_{k} \end{matrix}

\begin{matrix} \leq & \sum_{p = 1}^{2} \sum_{s, t = 1}^{k} {‖ x_{s t}^{(p)} - λ_{s t}^{(p)} 1 ‖}_{1} \end{matrix}

\begin{matrix} \leq & \sum_{p = 1}^{2} \sum_{s, t = 1}^{k} L_{1} (x_{s t}^{(p)}) d i a m (V, ℒ) \end{matrix}

\begin{matrix} \leq & \sum_{s, t = 1}^{k} 2 L_{k} (x) d i a m (V, ℒ) \end{matrix}

\begin{matrix} = & 2 k^{2} L_{k} (x) d i a m (V, ℒ) . \end{matrix}

□

An operator space

X

is said to be

1

-exact if for every finite-dimensional subspace

ℰ \subseteq X

and

λ > 1

there is an isomorphism

α

from

ℰ

onto a subspace of a matrix algebra such that

∥ α ∥_{c b} ∥ α^{- 1} ∥_{c b} \leq λ

. A matrix order unit space

(V, 1)

is said to be

1

-exact if it is

1

-exact as an operator space.

Theorem 7.2. Let

(V, ℒ)

be a quantized metric space. If

V

1

-exact, then for every

ε > 0

, there is a quantized metric space

(M_{n_{λ_{ε}}}, N)

such that

d i s t_{N C} (V, M_{n_{λ_{ε}}}) < ε .

Proof. Since

V

1

-exact, by Lemma 5.1 in [9] there is a unital complete order embedding

ι : V \mapsto ℬ (ℋ)

and a net

V - - \to φ_{λ} M_{n_{λ}} - - \to ψ_{λ} ℬ (ℋ)

of unital completely positive mappings through matrix algebras such that

ψ_{λ} \circ φ_{λ}

converges pointwise to

ι

. Given

ε > 0

. By Lemma 7.2, we have

L_{1}^{1} = B_{1}^{2 d i a m (V, ℒ)} + C 1,

where

B_{1}^{2 d i a m (V, ℒ)} = {a \in V : L_{1} (a) \leq 1, ∥ a ∥_{1} \leq 2 d i a m (V, ℒ)}

. From Proposition 7.5 in [21] ,

B_{1}^{2 d i a m (V, ℒ)}

is totally bounded for

∥ \cdot ∥_{1}

. So there is a

λ_{ε}

such that

∥ (ψ_{λ_{ε}} \circ φ_{λ_{ε}}) (x) - x ∥ < \frac{ε}{5}, x \in L_{1}^{1} .

Denote

W = φ_{λ_{ε}} (V)

and

Q_{k} (y) = inf {L_{k} (x) : (φ_{λ_{ε}})_{k} (x) = y}

for

y \in M_{k} (W)

and

k \in N

We define

N_{k} (x, y) = \frac{5}{ε} ∥ (φ_{λ_{ε}})_{k} (x) - y ∥_{k}, (x, y) \in M_{k} (V \oplus W), k \in N .

It is clear that

N = (N_{k})

is a matrix seminorm on

V \oplus W

and satisfies the conditions (1), (2) and (3) of Definition 4.1 . For

x \in M_{k} (V)

and

δ > 0

, we can choose

y = (φ_{λ_{ε}})_{k} (x)

. Then

max {Q_{k} (y), N_{k} (x, y)} = Q_{k} (y) \leq L_{k} (x) \leq L_{k} (x) + δ .

For

y \in M_{k} (W)

and

δ > 0

, we can take

x \in M_{k} (V)

such that

y = (φ_{λ_{ε}})_{k} (x)

and

L_{k} (x) \leq Q_{k} (y) + δ

. Then

max {L_{k} (x), N_{k} (x, y)} = L_{k} (x) \leq Q_{k} (y) + δ .

N

is a matrix bridge between

(V, ℒ)

and

(W, Q)

. Define

P_{k} (x, y) = max {L_{k} (x), Q_{k} (y), N_{k} (x, y)}, (x, y) \in M_{k} (V \oplus W), k \in N .

Then

P = (P_{k}) \in ℳ (ℒ, Q)

by Proposition 4.3 . If

f \in C S_{k} (W)

, we have

f \circ φ_{λ_{ε}} \in C S_{k} (V)

and

\begin{matrix} D_{P_{k}} (f, f \circ φ_{λ_{ε}}) \end{matrix}

\begin{matrix} = & sup {∥ ≪ f, y ≫ - ≪ f \circ φ_{λ_{ε}}, x ≫ ∥ : P_{k} (x, y) \leq 1, (x, y) \in M_{k} (V \oplus W)} \end{matrix}

\begin{matrix} = & sup {∥ ≪ f, y - (φ_{λ_{ε}})_{k} (x) ≫ ∥ : P_{k} (x, y) \leq 1, (x, y) \in M_{k} (V \oplus W)} \end{matrix}

\begin{matrix} \leq & sup {∥ y - (φ_{λ_{ε}})_{k} (x) ∥_{k} : P_{k} (x, y) \leq 1, (x, y) \in M_{k} (V \oplus W)} \end{matrix}

\begin{matrix} \leq & \frac{ε}{5} \end{matrix}

On the other hand, if

g \in C S_{k} (V)

g

can be extended to a

\bar{g} \in C S_{k} (ℬ (ℋ))

by Arveson's extension theorem. We have

\bar{g} \circ ψ_{λ_{ε}} \in C S_{k} (W)

and

\begin{matrix} D_{P_{k}} (g, \bar{g} \circ ψ_{λ_{ε}}) \end{matrix}

\begin{matrix} = & sup {∥ ≪ g, x ≫ - ≪ \bar{g} \circ ψ_{λ_{ε}}, y ≫ ∥ : P_{k} (x, y) \leq 1, (x, y) \in M_{k} (V \oplus W)} \end{matrix}

\begin{matrix} = & sup {∥ ≪ \bar{g}, x - (ψ_{λ_{ε}})_{k} (y) ≫ ∥ : P_{k} (x, y) \leq 1, (x, y) \in M_{k} (V \oplus W)} \end{matrix}

\begin{matrix} \leq & sup {∥ ≪ \bar{g}, x - (ψ_{λ_{ε}} \circ φ_{λ_{ε}})_{k} (x) ≫ ∥ + ∥ ≪ \bar{g} \circ ψ_{λ_{ε}}, (φ_{λ_{ε}})_{k} (x) - y ≫ ∥ : \end{matrix}

\begin{matrix} P_{k} (x, y) \leq 1, (x, y) \in M_{k} (V \oplus W)} \end{matrix}

\begin{matrix} \leq & sup {∥ x - (ψ_{λ_{ε}} \circ φ_{λ_{ε}})_{k} (x) ∥_{k} + ∥ (φ_{λ_{ε}})_{k} (x) - y ∥_{k} : \end{matrix}

\begin{matrix} P_{k} (x, y) \leq 1, (x, y) \in M_{k} (V \oplus W)} \end{matrix}

\begin{matrix} \leq & sup {\sum_{i, j = 1}^{k} ∥ x_{i j} - (ψ_{λ_{ε}} \circ φ_{λ_{ε}}) (x_{i j}) ∥_{1} : P_{k} (x, y) \leq 1, \end{matrix}

\begin{matrix} (x, y) \in M_{k} (V \oplus W)} + \frac{ε}{5} \end{matrix}

\begin{matrix} \leq & (k^{2} + 1) \frac{ε}{5} \end{matrix}

So we obtain that

d i s t_{N C} (V, W) < \frac{ε}{2}

Since

W \subseteq M_{n_{λ_{ε}}}

is finite-dimensional,

K = Q_{1}^{1}

is a normed-closed (and hence weakly closed) absolutely convex set in

M_{n_{λ_{ε}}}

, and

Q^{1} = (Q_{k}^{1})

is a normed-closed (and hence weakly closed) absolutely matrix convex set in

M_{n_{λ_{ε}}}

. Then for the corresponding matrix seminorm

\overset{ˇ}{ℛ} = ({\overset{ˇ}{R}}_{k})

of the maximal envelope

\overset{ˇ}{K}

K

M_{n_{λ_{ε}}}

(see Example 3.4 ), we have

{\overset{ˇ}{R}}_{1}^{1} = Q_{1}^{1}, {{\overset{ˇ}{R}}_{k} |}_{M_{k} (W)} \leq {Q_{k} |}_{M_{k} (W)}, k \in N,

(see page 181 in [4] ). It is clear that

\overset{ˇ}{ℛ}

is a matrix Lipschitz seminorm. Since the image of

Q_{1}^{1} = {\overset{ˇ}{R}}_{1}^{1}

W / (C 1)

is totally bounded for

∥ \cdot ∥^{\sim}

and

W \subseteq M_{n_{λ_{ε}}}

, the image of

{\overset{ˇ}{R}}_{1}^{1}

M_{n_{λ_{ε}}} / (C 1)

is totally bounded for

∥ \cdot ∥^{\sim}

. By Theorem 5.3 in [20] ,

D_{ℛ}

-topology on

C S (M_{n_{λ_{ε}}})

agrees with the BW-topology. So

\overset{ˇ}{ℛ}

is a matrix Lip-norm on

(M_{n_{λ_{ε}}}, 1)

. By Lemma 3.2.3 in [1] , there is a (real linear) projection

T

from

(M_{n_{λ_{ε}}})_{s a}

onto

(W)_{s a}

with

∥ T ∥ \leq n_{λ_{ε}}

. We define

S : M_{n_{λ_{ε}}} \mapsto W

S (a + i b) = T (a) + i T (b)

for

a, b \in (M_{n_{λ_{ε}}})_{s a}

. Then

S

is a bounded linear mapping with

∥ S ∥ \leq 2 n_{λ_{ε}}

. Define

N_{k} (x) = max {Q_{k} (S_{k} (x)), {\overset{ˇ}{R}}_{k} (x), \frac{4}{ε} ∥ x - S_{k} (x) ∥_{k}}, x \in M_{k} (M_{n_{λ_{ε}}}), k \in N .

It is clear that

N = (N_{k})

is a matrix Lip-norm on

M_{n_{λ_{ε}}}

since

{\overset{ˇ}{R}}_{k} \leq N_{k}

for all

k \in N

and

\overset{ˇ}{ℛ}

is a matrix Lip-norm. And for

y \in M_{k} (W)

, we have

\begin{matrix} N_{k} (y) & = & max {Q_{k} (S_{k} (y)), {\overset{ˇ}{R}}_{k} (y), \frac{4}{ε} ∥ y - S_{k} (y) ∥_{k}} \end{matrix}

\begin{matrix} = & max {Q_{k} (y), {\overset{ˇ}{R}}_{k} (y)} \end{matrix}

\begin{matrix} = & Q_{k} (y) . \end{matrix}

Define

X_{k} (x, y) = \frac{4}{ε} ∥ y - S_{k} (x) ∥_{k}, (x, y) \in M_{k} (M_{n_{λ_{ε}}} \oplus W), k \in N .

It is clear that

N = (N_{k})

is a matrix seminorm on

M_{n_{λ_{ε}}} \oplus W

and satisfies the conditions (1), (2) and (3) of Definition 4.1 . For

x \in M_{k} (M_{n_{λ_{ε}}})

and

δ > 0

, we choose

y = S_{k} (x)

. Then we have

max {Q_{k} (y), X_{k} (x, y)} = Q_{k} (S_{k} (x)) \leq N_{k} (x) \leq N_{k} (x) + δ .

For

y \in M_{k} (W)

and

δ > 0

, we choose

x = y

. Then we have

max {N_{k} (x), X_{k} (x, y)} = N_{k} (y) = Q_{k} (y) \leq Q_{k} (x) + δ .

X = (X_{k})

is a matrix bridge between

(M_{k} (M_{n_{λ_{ε}}}), N)

and

(W, Q)

. Define

Y_{k} (x, y) = max {N_{k} (x), Q_{k} (y), X_{k} (x, y)}, (x, y) \in M_{k} (M_{n_{λ_{ε}}} \oplus W), k \in N .

Then

Y = (Y_{k}) \in ℳ (N, Q)

by Proposition 4.3 . For

φ \in C S_{k} (M_{n_{λ_{ε}}})

ψ = φ |_{W} \in C S_{k} (W)

and

\begin{matrix} D_{Y_{k}} (φ, ψ) \end{matrix}

\begin{matrix} = & sup {∥ ≪ φ, x ≫ - ≪ ψ, y ≫ ∥ : (x, y) \in M_{k} (M_{n_{λ_{ε}}} \oplus W), Y_{k} (x, y) \leq 1} \end{matrix}

\begin{matrix} \leq & sup {∥ ≪ φ, x ≫ - ≪ φ, S_{k} (x) ≫ + ≪ φ, S_{k} (x) ≫ - ≪ φ, y ≫ ∥ : \end{matrix}

\begin{matrix} (x, y) \in M_{k} (M_{n_{λ_{ε}}} \oplus W), Y_{k} (x, y) \leq 1} \end{matrix}

\begin{matrix} \leq & sup {∥ x - S_{k} (x) ∥_{k} + ∥ S_{k} (x) - y ∥_{k} : (x, y) \in M_{k} (M_{n_{λ_{ε}}} \oplus W), Y_{k} (x, y) \leq 1} \end{matrix}

\begin{matrix} \leq & \frac{ε}{2} . \end{matrix}

For

ψ \in C S_{k} (W)

ψ

can be extended to a

φ \in C S_{k} (M_{n_{λ_{ε}}})

by Arveson's extension theorem. We have

\begin{matrix} D_{Y_{k}} (φ, ψ) \end{matrix}

\begin{matrix} = & sup {∥ ≪ φ, x ≫ - ≪ ψ, y ≫ ∥ : (x, y) \in M_{k} (M_{n_{λ_{ε}}} \oplus W), Y_{k} (x, y) \leq 1} \end{matrix}

\begin{matrix} \leq & sup {∥ ≪ φ, x ≫ - ≪ φ, S_{k} (x) ≫ + ≪ φ, S_{k} (x) ≫ - ≪ φ, y ≫ ∥ : \end{matrix}

\begin{matrix} (x, y) \in M_{k} (M_{n_{λ_{ε}}} \oplus W), Y_{k} (x, y) \leq 1} \end{matrix}

\begin{matrix} \leq & sup {∥ x - S_{k} (x) ∥ + ∥ S_{k} (x) - y ∥ : (x, y) \in M_{k} (M_{n_{λ_{ε}}} \oplus W), Y_{k} (x, y) \leq 1} \end{matrix}

\begin{matrix} \leq & \frac{ε}{2} . \end{matrix}

d i s t_{N C} (W, M_{n_{λ_{ε}}}) < \frac{ε}{2}

. Therefore,

d i s t_{N C} (V, M_{n_{λ_{ε}}}) \leq d i s t_{N C} (V, W) + d i s t_{N C} (W, M_{n_{λ_{ε}}}) < ε .

□

8 Sphere as the limit of matrix algebras

Let

G

be a connected compact semisimple Lie group with a continuous length function

l

G

, which satisfies the additional condition

l (x y x^{- 1}) = l (y)

for all

x, y \in G

. Fix an irreducible unitary representation

(U, ℋ)

G

. Then

(U, ℋ)

have a highest weight vector

ξ

, of norm

1

, unique up to a scalar multiple. Let

P

be the rank-one projection for

ξ

. Denote by

H

the stability subgroup for

P

under

α

. For any

n \in N

, we form the

n^{t h}

inner tensor power

(U^{\otimes n}, ℋ^{\otimes n})

(U, ℋ)

Let

(U^{(n)}, ℋ^{(n)})

denote the subrepresentation generated by

ξ^{(n)} = ξ^{\otimes n}

. Then

(U^{(n)}, ℋ^{(n)})

is irreducible with

ξ^{(n)}

as highest weight vector. We let

ℬ^{(n)} = ℬ (ℋ^{(n)})

The action of

G

ℬ^{(n)}

by conjugation by

U^{(n)}

is denoted by

α^{(n)}

. We let

λ

denote the action of

G

G / H

, and so on

A = C (G / H)

, by left-translation. We denote the corresponding Lip-norm for

α^{(n)}

and

l

ℬ^{(n)}

L^{(n)}

, that is,

L^{(n)} (T) = sup {\frac{∥ α_{x}^{(n)} (T) - T ∥}{l (x)} : x \neq e, x \in G}, T \in ℬ^{(n)},

and we denote the Lip-norm for

λ

and

l

A

L

, that is,

L (f) = sup {\frac{∥ λ_{x} (f) - f ∥_{\infty}}{l (x)} : x \neq e, x \in G}, f \in A,

here we view

C (G / H)

as a subalgebra of

C (G)

. By Theorem 3.2 in [18] , the quantum metric spaces

(ℬ^{(n)}, L^{(n)})

converge to

(A, L)

for quantum Gromov-Hausdorff distance as

n

goes to

\infty

. In this section, a more general statement is established.

Let

∥ \cdot ∥_{\infty} = (∥ \cdot ∥_{\infty, k})

be the matrix norm on

A

. Set

ℒ^{(n)} = (L_{k}^{(n)})

, where

L_{k}^{(n)} (T) = sup {\frac{∥ [α_{x}^{(n)} (T_{i j}) - T_{i j}] ∥}{l (x)} : x \neq e, x \in G}, T = [T_{i j}] \in M_{k} (ℬ^{(n)}), k \in N,

and

ℒ = (L_{k})

, where

L_{k} (f) = sup {\frac{∥ [λ_{x} (f_{i j}) - f_{i j}] ∥_{\infty, k}}{l (x)} : x \neq e, x \in G}, f = [f_{i j}] \in M_{k} (A), k \in N .

Then

(ℬ^{(n)}, ℒ^{(n)})

and

(A, ℒ)

are quantized metric spaces for all

n \in N

by Example 6.5 in [21] . As in [18] , we will not restrict

ℒ

to the Lipschitz functions. Let

P^{(n)}

denote the rank-one projection for

ξ^{(n)}

. We denote the corresponding Berezin symbol mapping from

ℬ^{(n)}

A

σ^{(n)}

. Then

σ^{(n)}

is unital, positive, norm-nonincreasing and

α^{(n)}

λ

-equivariant (see page 73 in [18] ). For

k \in N

and

T = [T_{i j}] \in M_{k} (ℬ^{(n)})

, define

σ_{T}^{(n)} (x) = [σ_{T_{i j}}^{(n)} (x)], x \in G .

For

ε > 0

, define

N_{k} (f, T) = ε^{- 1} ∥ f - σ_{T}^{(n)} ∥_{\infty, k}, f \in M_{k} (A), T \in M_{k} (ℬ^{(n)}),

and denote

N = (N_{k})

Lemma 8.1. For any

T \in M_{k} (ℬ^{(n)})

, we have

L_{k} (σ_{T}^{(n)}) < L_{k}^{(n)} (T) + ε .

Proof. Since

σ^{(n)}

is a unital positive mapping from

ℬ^{(n)}

A

σ^{(n)}

is unital completely positive and hence

∥ σ^{(n)} ∥_{c b} = 1

by Theorem 3.8 and Proposition 3.5 in [12] . So we have

\begin{matrix} L_{k} (σ_{T}^{(n)}) \end{matrix}

\begin{matrix} = & sup {\frac{{‖ [λ_{x} (σ_{T_{i j}}^{(n)}) - σ_{T_{i j}}^{(n)}] ‖}_{\infty, k}}{l (x)} : x \neq e, x \in G} \end{matrix}

\begin{matrix} = & sup {\frac{{‖ [σ_{(α_{x}^{(n)} (T_{i j}) - T_{i j})}^{(n)}] ‖}_{\infty, k}}{l (x)} : x \neq e, x \in G} \end{matrix}

\begin{matrix} \leq & sup {\frac{∥ [α_{x}^{(n)} (T_{i j}) - T_{i j}] ∥}{l (x)} : x \neq e, x \in G} \end{matrix}

\begin{matrix} = & L_{k}^{(n)} (T) \end{matrix}

\begin{matrix} < & L_{k}^{(n)} (T) + ε, \end{matrix}

by the

α^{(n)}

λ

-equivariation of

σ^{(n)}

. □

Put on

A

the inner product from

L^{2} (G / H)

, while on

ℬ^{(n)}

its Hilbert-Schmidt inner product. Then the mapping

σ^{(n)}

from

ℬ^{(n)}

A

has an adjoint operator

{\overset{˘}{σ}}^{(n)}

from

A

ℬ^{(n)}

. For any

T \in ℬ^{(n)}

, a function

f \in A

such that

{\overset{˘}{σ}}_{f}^{(n)} = T

is called a Berezin contravariant symbol for

T

. Moreover,

{\overset{˘}{σ}}^{(n)}

is unital, positive, norm-nonincreasing, and

λ

α^{(n)}

-equivariant (see page 75 in [18] ). From Theorem 3.10 and Proposition 3.5 in [12] ,

{\overset{˘}{σ}}^{(n)}

is unital completely positive and

∥ {\overset{˘}{σ}}^{(n)} ∥_{c b} = 1

So by the same argument as in the proof of Lemma 8.1 , we obtain:

Lemma 8.2. For any

f = [f_{i j}] \in M_{k} (A)

, we have

L_{k}^{(n)} ({\overset{˘}{σ}}_{f}^{(n)}) < L_{k} (f) + ε,

where

{\overset{˘}{σ}}_{f}^{(n)} = [{\overset{˘}{σ}}_{f_{i j}}^{(n)}] \in M_{k} (ℬ^{(n)})

Denote

D_{L_{k}} (φ, ψ) = sup {∥ ≪ f, φ ≫ - ≪ f, ψ ≫ ∥ : L_{r} (f) \leq 1, f \in M_{r} (A), r \in N},

for

φ, ψ \in C S_{k} (A)

k \in N

, and

h_{P^{(n)}} (x) = d^{(n)} τ^{(n)} (P^{(n)} α_{x}^{(n)} (P^{(n)})), x \in G / H,

where

τ^{(n)}

denotes the usual (un-normalized) trace on

ℬ^{(n)}

and

d^{(n)}

is the dimension of

ℋ^{(n)}

. Set

γ^{(n)} = \int_{G / H} D_{L_{1}} (\hat{e}, \hat{y}) h_{P^{(n)}} (y) d y,

where every

y \in G / H

is naturally identified with an element

\hat{y}

C S_{1} (A)

. Then:

Lemma 8.3. For

f \in M_{k} (A)

, we have

∥ f - σ^{(n)} ({\overset{˘}{σ}}_{f}^{(n)}) ∥_{\infty, k} \leq γ^{(n)} L_{k} (f) .

Proof. Suppose

f = [f_{i j}]

. Then for any

x \in G / H

, we have

\begin{matrix} ∥ f (x) - (σ^{(n)} ({\overset{˘}{σ}}_{f}^{(n)})) (x) ∥ \end{matrix}

\begin{matrix} = & ‖ [\int_{G / H} (f_{i j} (x) - f_{i j} (y)) h_{P^{(n)}} (y^{- 1} x) d y] ‖ \end{matrix}

\begin{matrix} = & ‖ \int_{G / H} [f_{i j} (x) - f_{i j} (y)] h_{P^{(n)}} (y^{- 1} x) d y ‖ \end{matrix}

\begin{matrix} \leq & \int_{G / H} ∥ [f_{i j} (x) - f_{i j} (y)] ∥ h_{P^{(n)}} (y^{- 1} x) d y \end{matrix}

\begin{matrix} \leq & L_{k} (f) \int_{G / H} D_{L_{1}} (\hat{x}, \hat{y}) h_{P^{(n)}} (y^{- 1} x) d y \end{matrix}

\begin{matrix} = & L_{k} (f) \int_{G / H} D_{L_{1}} (\hat{e}, \hat{y}) h_{P^{(n)}} (y) d y \end{matrix}

\begin{matrix} \leq & γ^{(n)} L_{k} (f), \end{matrix}

by the formula (2.2) in [18] . So

∥ f - σ^{(n)} ({\overset{˘}{σ}}_{f}^{(n)}) ∥_{\infty, k} = max {∥ f (x) - (σ^{(n)} ({\overset{˘}{σ}}_{f}^{(n)})) (x) ∥ : x \in G / H} \leq γ^{(n)} L_{k} (f) .

□

Since the sequence

{γ^{(n)}}

converges to

0

n \to \infty

(see page 80 in [18] ), there is an

N_{1} \in N

such that

γ^{(n)} < \frac{ε}{2}

for

n > N_{1}^{(n)}

. So we obtain:

Proposition 8.4. For

n > N_{1}

N

is a matrix bridge between

(ℬ^{(n)}, ℒ^{(n)})

and

(A, ℒ)

, and hence

Q = (Q_{k}) \in ℳ (ℒ^{(n)}, ℒ)

, where

Q_{k} (f, T) = max {L_{k}^{(n)} (T), L_{k} (f), N_{k} (f, T)}, (f, T) \in M_{r} (ℬ^{(n)} \oplus A) .

From Theorem 6.1 in [18] , we have:

Lemma 8.5. There is an

N_{2} \in N

such that

‖ T - {\overset{˘}{σ}}^{(n)} (σ_{T}^{(n)}) ‖ < \frac{ε}{2} L_{1}^{(n)} (T),

for all

T \in ℬ^{(n)}

and

n > N_{2}

Theorem 8.6. With notation as above, the quantized metric spaces

(ℬ^{(n)}, ℒ^{(n)})

converge to

(A, ℒ)

for quantized Gromov-Hausdorff distance as

n

goes to

\infty

Proof. Given

ε > 0

. Choose

N = max {N_{1}, N_{2}}

. Then for

n > N

, we have that

Q \in ℳ (ℒ^{(n)}, ℒ)

by Proposition 8.4 . Given

φ \in C S_{k} (A)

. we have

φ \circ σ^{(n)} \in C S_{k} (ℬ^{(n)})

, and

\begin{matrix} D_{L_{k}} (φ, φ \circ σ^{(n)}) \end{matrix}

\begin{matrix} = & sup {∥ ≪ φ, f ≫ - ≪ φ \circ σ^{(n)}, T ≫ ∥ : L_{r} (f, T) \leq 1, \end{matrix}

\begin{matrix} (f, T) \in M_{r} (A \oplus ℬ^{(n)}), r \in N} \end{matrix}

\begin{matrix} = & sup {∥ ≪ φ, f - σ_{T}^{(n)} ≫ ∥ : L_{r} (f, T) \leq 1, (f, T) \in M_{r} (A \oplus ℬ^{(n)}), r \in N} \end{matrix}

\begin{matrix} \leq & sup {∥ f - σ_{T}^{(n)} ∥_{\infty, r} : L_{r} (f, T) \leq 1, (f, T) \in M_{r} (A \oplus ℬ^{(n)}), r \in N} \end{matrix}

\begin{matrix} \leq & ε . \end{matrix}

On the other hand, if

ψ \in C S_{k} (ℬ^{(n)})

, then

ψ \circ {\overset{˘}{σ}}^{(n)} \in C S_{k} (A)

, and

\begin{matrix} D_{L_{k}} (ψ \circ {\overset{˘}{σ}}^{(n)}, ψ) \end{matrix}

\begin{matrix} = & sup {∥ ≪ ψ \circ {\overset{˘}{σ}}^{(n)}, f ≫ - ≪ ψ, T ≫ ∥ : L_{k} (f, T) \leq 1, \end{matrix}

\begin{matrix} (f, T) \in M_{k} (A \oplus ℬ^{(n)})} \end{matrix}

\begin{matrix} = & sup {∥ ≪ ψ, {\overset{˘}{σ}}_{f}^{(n)} - T ≫ ∥ : L_{k} (f, T) \leq 1, (f, T) \in M_{k} (A \oplus ℬ^{(n)})} \end{matrix}

\begin{matrix} \leq & sup {∥ {\overset{˘}{σ}}_{f}^{(n)} - T ∥ : L_{k} (f, T) \leq 1, (f, T) \in M_{k} (A \oplus ℬ^{(n)})} \end{matrix}

\begin{matrix} \leq & sup {∥ {\overset{˘}{σ}}_{f}^{(n)} - {\overset{˘}{σ}}^{(n)} (σ_{T}^{(n)}) ∥ + ∥ {\overset{˘}{σ}}^{(n)} (σ_{T}^{(n)}) - T ∥ : L_{k} (f, T) \leq 1, \end{matrix}

\begin{matrix} (f, T) \in M_{k} (A \oplus ℬ^{(n)})} \end{matrix}

\begin{matrix} \leq & ∥ f - σ_{T}^{(n)} ∥_{\infty, k} + sup {∥ {\overset{˘}{σ}}^{(n)} (σ_{T}^{(n)}) - T ∥ : L_{k} (f, T) \leq 1, \end{matrix}

\begin{matrix} (f, T) \in M_{k} (A \oplus ℬ^{(n)})} \end{matrix}

\begin{matrix} \leq & \frac{ε}{2} + \frac{1}{2} k^{2} ε \end{matrix}

\begin{matrix} \leq & k^{2} ε, \end{matrix}

by Lemma 8.5 . Therefore, for

n > N

, we have

d i s t_{N C} (ℬ^{(n)}, A) \leq ε,

that is,

{lim}_{n \to \infty} d i s t_{N C} (ℬ^{(n)}, A) = 0

. □

Acknowledgements

I am grateful to Marc Rieffel for valuable discussions. I also thank Hanfeng Li for helpful comments. This research was partially supported by Shanghai Priority Academic Discipline, China Scholarship Council, and National Natural Science Foundation of China. References

B. Blackadar, E. Kirchberg, Generalized inductive limits of finite-dimensional $C^{*}$ -algebras, Math.Ann. 307 (1997), 343–380.
M. -D. Choi, E. G. Effros, Injectivity and operator spaces, J. Funct. Anal., 24 (1977), 156–209.
A. Connes, Compact metric spaces, Fredholm modules, and hyperfiniteness, Ergodic Theory Dynam. Systems,9(2) (1989), 207–220.
E. G. Effros and C. Webster, Operator analogues of locally convex spaces, Operator algebras and applications(Samos, 1996), 163–207, NATO Adv. Sci. Inst. Ser. C Math. Phys. Sci., 495, Kluwer Acad. Publ., Dordrecht, 1997.
E. G. Effros, S. Winkler, Matrix convexity: operator analogues of the Bipolar and Hahn-Banach theorems, J. Funct. Anal., 144 (1997), 117–152.
M. Gromov, Metric structures for Riemannian and non-Riemannian spaces, Birkhäuser Boston Inc., Boston, MA, 1999.
R. V. Kadison, A representation theory for commutative topological algebra. Mem. Amer. Math. Soc., 7, 1951.
D. Kerr, Matricial quantum Gromov-Hausdorff distance. J. Funct. Anal., 205 (2003), 132–167.
D. Kerr, H. Li, On Gromov-Hausdorff convergence for operator metric spaces, arXiv:math.OA/0411157 v2, 2004.
H. Li, $θ$ -deformations as compact quantum metric spaces, to appear in Comm. Math. Phys., arXiv:math.OA/0311500.
N. Ozawa, M. A. Rieffel, Hyperbolic group $C^{*}$ -algebras and free-product $C^{*}$ -algebras as compact quantum metric spaces, to appear in Canad. J. Math., arXiv:math.OA/0302310.
V. I. Paulsen, Completely bounded maps and dilations, Pitman Research Notes in Mathematics Series, 146. Longman Scientific & Technical, Harlow; John Wiley & Sons, Inc., New York, 1986.
G. Pisier, Introduction to operator space theory, London Mathematical Society Lecture Note Series, 294. Cambridge University Press, Cambridge, 2003.
M. A. Rieffel, Metrics on states from actions of compact groups, Doc. Math., 3 (1998), 215–229.
M. A. Rieffel, Metrics on state spaces, Doc. Math., 4 (1999), 559–600.
M. A. Rieffel, Group $C^{*}$ -algebras as compact quantum metric spaces, Doc. Math., 7 (2002), 605–651.
M. A. Rieffel, Gromov-Hausdorff distance for quantum metric spaces, Mem. Amer. Math. Soc. 168 (2004), 1–65.
M. A. Rieffel, Matrix algebras converge to the sphere for quantum Gromov-Hausdorff distance, Mem. Amer. Math. Soc. 168 (2004), 67–91.
C. Webster, S. Winkler, The Krein-Milman theorem in operator convexity, Trans. Amer. Math. Soc., 351 (1999), 307–322.
W. Wu, Non-commutative metric topology on matrix state spaces, to appear in Proc. Amer. Math. Soc., arXiv:math.OA/0410587.
W. Wu, Non-commutative metrics on matrix state spaces, arXiv:math.OA/0411475.

Department of Mathematics, East China Normal University, Shanghai 200062, P.R. China E-mail address : wwu@math.ecnu.edu.cn Current address : Department of Mathematics, University of California, Berkeley, CA 94720-3840 E-mail address : wwu@math.berkeley.edu

wei wu