1.

Introduction
Let A be a finite dimensional C∗ -algebra and B a unital C∗ -algebra. Then there exists a
unital C∗ -algebra MA,B together with a unital ∗-homomorphism ΦA,B : B → A ⊗ MA,B such
that for any unital C∗ -algebra C and any unital ∗-homomorphism Ψ : B → A ⊗ C there exists a
unique unital ∗-homomorphism Λ : MA,B → C such that Ψ = (id ⊗ Λ) ◦ ΦA,B . In fact the same
kind of universality of (MA,B , ΦA,B ) holds also in the category of C∗ -algebras with morphisms
à la Woronowicz.
Free products. Given two unital C∗ -algebras A and B the universal unital free product
(i.e. the universal free product with amalgamation over the scalars) will be denoted by A * B
C1
and the canonical (injective) unital ∗-homomorphism A → A * B and B → A * B will be
C1 C1
denoted by ıA and ıB respectively.

2. What is known about MA,B

• MA,B is generated by the set {(ω ⊗ id)ΦA,B (b) | b ∈ B, ω ∈ A∗ }.
• MA1 ⊕A2 ,B ∼
= MA1 ,B * MA2 ,B and with this identification
C1



ΦA1 ⊕A2 ,B (b) = (1 ⊗ ıA1 ) ΦA1 ,B (b) + (2 ⊗ ıA2 ) ΦA2 ,B (b) , b ∈ B,
where for i = 1, 2 the maps i : Ai → A1 ⊕ A2 are the (non-unital) inclusions of direct
summands.
• In particular taking A = Cn we have MCn ,B ∼ = B * · · · * B (n copies) and denoting by
C1 C1
ıB,i the inclusion of the i-th copy of B in B * · · · * B we have
C1 C1
n
X
ΦCn ,B (b) = ei ⊗ ıB,i (b), b ∈ B,
i=1
where {e1 , . . . , en } is the standard basis of Cn .
• MMatn ,C(T) is isomorphic to the Brown algebra Unnc , while MMatn ,C2 is Brown’s non-
commutative Grassmannian Gnc n .
• The Gelfand spectrum of MA,B is the space of unital ∗-homomorphisms B → A with the
natural topology, i.e. the weakest topology making the evaluation maps continuous.
• For any unital ∗-homomorphisms ρ : B2 → B1 and π : A1 → A2 there exists a unique
Mρ,π : MA1 ,B1 → MA2 ,B2 such that
(id ⊗ Mρ,π ) ◦ ΦA1 ,B1 = (ρ ⊗ id) ◦ ΦA2 ,B2 ◦ π.
Furthermore, if
ρ : A2 −→ A1 , ρ0 : A3 −→ A2 ,
π : B1 −→ B2 , π 0 : B2 −→ B3
are unital ∗-homomorphisms then
Mρ◦ρ0 ,π0 ◦π = Mρ0 ,π0 ◦ Mρ,π .
• We have



π is surjective =⇒ Mid,π is surjective ,



π is injective =⇒ Mid,π is injective ,



ρ is surjective =⇒ Mρ,id is injective ,
1
2



ρ is injective =⇒ Mρ,id is surjective .
Consequently MA,B is not nuclear in either of the following cases
– 2 < dim B < +∞ and dim A > 1,
– B has at least two characters and dim B ∈ {3, 5, 6, 7, . . . },
– B has at least three characters and dim B > 1
(in each case MA,B surjects onto a non-nuclear algebra).
• If B is separable and has the lifting property then MA,B has the lifting property for
any A.
• If B is residually finite dimensional then MA,B is residually finite dimensional for any
A.
• The C∗ -algebra MA,A carries a natural structure of a quantum semigroup with unit
(i.e. it admits a comultiplication and co-unit) which is not a quantum group unless
dim A = 1.1 For any quantum group G (in particular for G = Qut(A, ψ) for some state
ψ on A) and any action of G on A the corresponding Λ : MA,A → C(G) is a morphism
of quantum semigroups (it intertwines the respective comultiplications).

3. K-theory
The C∗ -algebra Matn ⊗MMatn ,B is isomorphic to Matn * B in such a way that the universal
C1
homomorphism ΦMatn ,B : B → Matn ⊗MMatn ,B coincides with ıB : B → Matn * B. Conse-
C1
quently
Ki (MMatn ,B ) ∼
= Ki (Matn ⊗MMatn ,B ) ∼
= Ki (Matn * B).
C1
By highly non-trivial results of Fima and Germain there is a six-term exact sequence
αMatn ∗ ⊕αB∗ ıMatn ∗ +ıB∗
K0 (C) / K0 (Matn ) ⊕ K0 (B) / K0 (Matn B)
O *
C1


K1 (Matn * B) o ıMatn ∗ +ıB∗
K1 (Matn ) ⊕ K1 (B) o αMatn ∗ ⊕αB∗
K1 (C)
C1

where αMatn and αB are the unital inclusions of C into Matn and B respectively. In other
words
αMatn ∗ ⊕αB∗ ıMatn ∗ +ıB∗
ZO / Z ⊕ K0 (B) / K0 (Matn B)
*
C1


K1 (Matn * B) o ıB∗ K1 (B) o 0
C1

αMatn ⊕αB∗
Since the mapping Z −−−−−∗−−−−→ Z ⊕ K0 (B) is injective, the left vertical map must be
zero and hence
K1 (MMatn ,B ) ∼
= K1 (Matn * B) ∼
= K1 (B).
C1

Furthermore the quotient of Z ⊕ K0 (B) by the image of αMatn ∗ ⊕ αB∗ is isomorphic to the
second direct summand, so
∼ K0 (Matn B) ∼
K0 (MMatn ,B ) = = K0 (B).
*
C1

1The comultiplication ∆ : M
A,A → MA,A ⊗ MA,A is determined by (id ⊗ ∆) ◦ ΦA,A = (ΦA,A ⊗ id) ◦ ΦA,A .
 3

Remark 3.1. There is a canonical surjective map ΛA,B : MA,B → B defined by

b ∈ B.2


(id ⊗ ΛA,B ) ΦA,B (b) = 1 ⊗ b,
In the case A = Matn the map id ⊗ ΛMatn ,B : Matn ⊗MMatn ,B → Matn ⊗B induces the same
map on K-theory as ΛMatn ,B , while
B 3 b 7−→ 1 ⊗ b ∈ Matn ⊗B
clearly induces identity on Ki (B). Thus ΛMatn ,B ∗ ◦ ΦMatn ,B ∗ = id on K-theory.
The K-groups of MA,B for general A should in principle be determined from repeated use
of
MA1 ⊕A2 ,B = MA1 ,B * MA2 ,B
C1
and the six-term exact sequence.

4. Operations on B
4.1. Free product. The C∗ -algebra MA,B1 * B2 i isomorphic to MA,B1 * MA,B2 and the uni-
C1 C1
versal morphism ΦA,B1 * B2 arises from the pair of morphisms
C1

(id ⊗ ıMA,B1 ) ◦ ΦA,B1 : B1 −→ A ⊗ MA,B1 * MA,B2

C1

(id ⊗ ıMA,B2 ) ◦ ΦA,B2 : B2 −→ A ⊗ MA,B1 * MA,B2 .

C1

Indeed, any morphism Λ : B1 * B2 → A⊗C arises from two unital ∗-homomorphisms Ψi : Bi →

C1
A ⊗ C (i = 1, 2) and they are of the form
Ψi = (id ⊗ Λi ) ◦ ΦA,Bi
for unique Λi : MA,Bi → A ⊗ C. The pair (Λ1 , Λ2 ) gives rise to a unique Λ : MA,B1 * MA,B2 →
C1
A ⊗ C such that
Λ ◦ ıMA,Bi = Λi , i = 1, 2.
It is clear that with ΦA,B1 * B2 defined above we have
C1

(id ⊗ Λ) ◦ ΦA,B1 * B2 = Ψ.
C1

4.2. Inductive limit. Fix a finite-dimensional C∗ -algebra A. Let I be a directed set and
((Bi )i∈I , (πi,j : Bi → Bj )i,j∈I, i4j ) be an inductive system of unital C∗ -algebras (and unital
∗-homomorphisms), Set B = lim Bi and let βi : Bi → B be the natural map into the inductive
−→
limit. We note that
βj ◦ πi,j = βi , i, j ∈ I, i 4 j.
Clearly ((MA,Bi )i∈I , (Mid,πi,j )i,j∈I, i4j ) is an inductive system and we set M = lim MA,Bi
−→
with canonical maps µi : MA,Bi → M which satisfy
µj ◦ Mid,πi,j = µi , i, j ∈ I, i 4 j.
Next we note that the family ((id ⊗ µi ) ◦ ΦA,Bi )i∈I has the property that



(id ⊗ µj ) ◦ ΦA,Bj ◦ πi,j = id ⊗ (µj ◦ Mid,πi,j ) ◦ ΦA,Bi = (id ⊗ µi ) ◦ ΦA,Bi
2One easily sees that M ∼
C,B = B and with this identification ΦC,B maps b ∈ B to 1 ⊗ b ∈ C ⊗ B. We have
ΛA,B = MαA ,id : MA,B → MC,B .
4

whenever i 4 j, so that there exists a unique Φ : B → A ⊗ M with the property that

Φ ◦ βi = (id ⊗ µi ) ◦ ΦA,Bi , i ∈ I.
Now let Ψ : B → A⊗C be an arbitrary unital ∗-homomorphism and for each i set Ψi = Ψ◦βi .
Then automatically
Ψj ◦ πi,j = Ψ ◦ βj ◦ πi,j = Ψ ◦ βi = Ψi , i, j ∈ I, i 4 j.
Next we note that for each i there is a unique Λi : MA,Bi → C such that
Ψi = (id ⊗ Λi ) ◦ ΦA,Bi
and whenever i 4 j


id ⊗ (Λj ◦ Mid,πi,j ) ◦ ΦA,Bi = (id ⊗ Λj ) ◦ ΦA,Bj ◦ πi,j = Ψj ◦ πi,j = Ψi = (id ⊗ Λi ) ◦ ΦA,Bi .
It follows from this and the fact that MA,Bi is generated by
(ω ⊗ id)ΦA,Bi (b) b ∈ Bi , ω ∈ A∗


that
Λj ◦ Mid,πi,j = Λi , i, j ∈ I, i 4 j.
Thus there exists Λ : M → A ⊗ C with the property that
Λ ◦ µi = Λi , i ∈ I.
Furthermore


(id ⊗ Λ) ◦ Φ ◦ βi = id ⊗ (Λj ◦ µi )) ◦ ΦA,Bi = (id ⊗ Λi ) ◦ ΦA,Bi = Ψi ,
i.e. (id ⊗ Λ) ◦ Φ coincides with Ψ on the range of each βi . It follows that
(id ⊗ Λ) ◦ Φ = Ψ
which also guarantees the uniqueness of Λ. This way we proved that (M, Φ) has the universal
property of (MA,B , ΦA,B ). In particular
MA,lim Bi = lim MA,Bi .
−→ −→