Classification of Continuous Fields of C* Algebras
Classification of Continuous Fields of C* Algebras
Prahlad Vaidyanathan (Joint work with M. Dadarlat)
Department of Mathematics IISER Bhopal
Sept 11, 2014 2 Continuous Fields of C* Algebras
3 E-theory for Continuous fields
4 Classification Results
Outline
1 Elliott’s Classification Program 3 E-theory for Continuous fields
4 Classification Results
Outline
1 Elliott’s Classification Program
2 Continuous Fields of C* Algebras 4 Classification Results
Outline
1 Elliott’s Classification Program
2 Continuous Fields of C* Algebras
3 E-theory for Continuous fields Outline
1 Elliott’s Classification Program
2 Continuous Fields of C* Algebras
3 E-theory for Continuous fields
4 Classification Results A finite dimensional C* algebra is of the form
Mn1 (C) ⊕ Mn2 (C) ⊕ ... ⊕ Mnk (C)
A nuclear C* algebra is one that can be approximated “nicely" by finite dimensional C* algebras.
Elliott’s Classification Program C* algebras
A C* algebra is An algebra of bounded linear operators on a Hilbert space that is closed in the norm topology and closed under taking adjoints. A complex Banach algebra with an involution map a 7→ a∗ such that ka∗ak = kak2 A nuclear C* algebra is one that can be approximated “nicely" by finite dimensional C* algebras.
Elliott’s Classification Program C* algebras
A C* algebra is An algebra of bounded linear operators on a Hilbert space that is closed in the norm topology and closed under taking adjoints. A complex Banach algebra with an involution map a 7→ a∗ such that ka∗ak = kak2 A finite dimensional C* algebra is of the form
Mn1 (C) ⊕ Mn2 (C) ⊕ ... ⊕ Mnk (C) Elliott’s Classification Program C* algebras
A C* algebra is An algebra of bounded linear operators on a Hilbert space that is closed in the norm topology and closed under taking adjoints. A complex Banach algebra with an involution map a 7→ a∗ such that ka∗ak = kak2 A finite dimensional C* algebra is of the form
Mn1 (C) ⊕ Mn2 (C) ⊕ ... ⊕ Mnk (C)
A nuclear C* algebra is one that can be approximated “nicely" by finite dimensional C* algebras. The Elliott Program To classify nuclear C* algebras by their K-theory
Elliott’s Classification Program K-theory
To each C* algebra A, one can associate two groups
K0(A) and K1(A) Elliott’s Classification Program K-theory
To each C* algebra A, one can associate two groups
K0(A) and K1(A)
The Elliott Program To classify nuclear C* algebras by their K-theory A class of simple algebras for which the Elliott program has been successful is the class of Kirchberg algebras (Purely infinite, simple, unital, nuclear algebras) Fact Every separable nuclear C* algebra is K-theoretically equivalent to a Kirchberg algebra.
Elliott’s Classification Program Kirchberg Algebras
A C* algebra is called simple if it has no non-trivial closed two-sided ideals. Fact Every separable nuclear C* algebra is K-theoretically equivalent to a Kirchberg algebra.
Elliott’s Classification Program Kirchberg Algebras
A C* algebra is called simple if it has no non-trivial closed two-sided ideals.
A class of simple algebras for which the Elliott program has been successful is the class of Kirchberg algebras (Purely infinite, simple, unital, nuclear algebras) Elliott’s Classification Program Kirchberg Algebras
A C* algebra is called simple if it has no non-trivial closed two-sided ideals.
A class of simple algebras for which the Elliott program has been successful is the class of Kirchberg algebras (Purely infinite, simple, unital, nuclear algebras) Fact Every separable nuclear C* algebra is K-theoretically equivalent to a Kirchberg algebra. By the fact mentioned above, we may assume that all simple quotients of our C* algebras are Kirchberg algebras.
This Project
Goal To classify non-simple C* algebras using K-theory K-theory of all quotients Maps between these K-theory groups This Project
Goal To classify non-simple C* algebras using K-theory K-theory of all quotients Maps between these K-theory groups By the fact mentioned above, we may assume that all simple quotients of our C* algebras are Kirchberg algebras. Θ is called the structure morphism. We usually drop Θ and write
f · a := Θ(f )(a)
Continuous Fields of C* Algebras C(X)-algebras
Let X be a compact metric space. A structure of a C(X)-algebra on a separable C* algebra A is a unital *-homomorphism
Θ: C(X) → ZM(A)
from C(X) to the center of the multiplier algebra of A. In other words, A has the structure of a non-degenerate C(X)-module. Continuous Fields of C* Algebras C(X)-algebras
Let X be a compact metric space. A structure of a C(X)-algebra on a separable C* algebra A is a unital *-homomorphism
Θ: C(X) → ZM(A)
from C(X) to the center of the multiplier algebra of A. In other words, A has the structure of a non-degenerate C(X)-module.
Θ is called the structure morphism. We usually drop Θ and write
f · a := Θ(f )(a) Write πx : A → A(x) for the natural quotient map. For a ∈ A, write
a(x) := πx (a)
A C(X)-algebra A is called a continuous field of C* algebras over X if the function x 7→ ka(x)k is continuous on X for each a ∈ A
Continuous Fields of C* Algebras Definition and Examples Continuous Fields of C* Algebras
The C(X)-module structure allows us to localize at points of X.
For x ∈ X, the quotient
A A(x) := C0(X \{x}) · A is called the fiber of A at x. Continuous Fields of C* Algebras Definition and Examples Continuous Fields of C* Algebras
The C(X)-module structure allows us to localize at points of X.
For x ∈ X, the quotient
A A(x) := C0(X \{x}) · A is called the fiber of A at x. Write πx : A → A(x) for the natural quotient map. For a ∈ A, write
a(x) := πx (a)
A C(X)-algebra A is called a continuous field of C* algebras over X if the function x 7→ ka(x)k is continuous on X for each a ∈ A If X = Prim(A) Hausdorff, the Dauns-Hoffman Theorem says that ∼ C(X) = ZM(A)
and A(x) is simple for all x ∈ X. Thus,
A unital separable C* algebra with a Hausdorff primitive spectrum is a continuous field over its spectrum.
Continuous Fields of C* Algebras Definition and Examples The Primitive Spectrum
If A is a C* algebra A primitive ideal is the kernel of an irreducible representation.
The collection of primitive ideals, Prim(A), is a compact T0-space Thus,
A unital separable C* algebra with a Hausdorff primitive spectrum is a continuous field over its spectrum.
Continuous Fields of C* Algebras Definition and Examples The Primitive Spectrum
If A is a C* algebra A primitive ideal is the kernel of an irreducible representation.
The collection of primitive ideals, Prim(A), is a compact T0-space
If X = Prim(A) Hausdorff, the Dauns-Hoffman Theorem says that ∼ C(X) = ZM(A)
and A(x) is simple for all x ∈ X. Continuous Fields of C* Algebras Definition and Examples The Primitive Spectrum
If A is a C* algebra A primitive ideal is the kernel of an irreducible representation.
The collection of primitive ideals, Prim(A), is a compact T0-space
If X = Prim(A) Hausdorff, the Dauns-Hoffman Theorem says that ∼ C(X) = ZM(A)
and A(x) is simple for all x ∈ X. Thus,
A unital separable C* algebra with a Hausdorff primitive spectrum is a continuous field over its spectrum. If A is a continuous field of C* algebras such that, for each x ∈ X, there is a neighbourhood V of x and a C* algebra D such that ∼ C0(V ) · A = C0(V , D)
then A is said to be locally trivial. Locally trivial fields are rare when the fibers are infinite dimensional C* algebras.
Continuous Fields of C* Algebras Definition and Examples Examples of Continuous Fields
If D is any fixed C* algebra, then
A = C(X, D) = {f : X → D such that f is continuous }
is called the trivial field with fiber D Continuous Fields of C* Algebras Definition and Examples Examples of Continuous Fields
If D is any fixed C* algebra, then
A = C(X, D) = {f : X → D such that f is continuous }
is called the trivial field with fiber D If A is a continuous field of C* algebras such that, for each x ∈ X, there is a neighbourhood V of x and a C* algebra D such that ∼ C0(V ) · A = C0(V , D)
then A is said to be locally trivial. Locally trivial fields are rare when the fibers are infinite dimensional C* algebras. Continuous Fields of C* Algebras Definition and Examples A field with one singularity
Let D, H be two C* algebras, and γ : D → H an injective *-homomorphism. For x ∈ [0, 1] fixed, define
Bx := {(f , g) ∈ C([0, x], H) ⊕ C([x, 1], D): f (r) = γ(g(x))}
We represent this pictorially by
H •0 •x O γ D •x •1
Then, Bx is a continuous field over [0, 1] which is locally trivial at all points of [0, 1] except x. Continuous Fields of C* Algebras Definition and Examples Elementary fields over [0, 1]
An elementary continuous field over [0, 1] is a field which is locally trivial at all but finitely many points. We depict them pictorially by
H1 H2 H3 Hk •0 •x1 •x2 •x3 •x4 ... •xn •1 O O O O O γ1 γ2 γ3 γ4 γn D1 D2 Dk−1 •x1 •x2 •x3 •x4 ... •xn Question Given two continuous fields A and B over the same space X, how do we know if there is a C(X)-linear *-isomorphism
∼ ϕ : A −→ B
Continuous Fields of C* Algebras The Classification Problem
Let A and B be a continuous fields over a space X. A morphism of continuous fields is a *-homomorphism
ϕ : A → B
such that ϕ(f · a) = f · ϕ(a) ∀f ∈ C(X), a ∈ A Continuous Fields of C* Algebras The Classification Problem
Let A and B be a continuous fields over a space X. A morphism of continuous fields is a *-homomorphism
ϕ : A → B
such that ϕ(f · a) = f · ϕ(a) ∀f ∈ C(X), a ∈ A
Question Given two continuous fields A and B over the same space X, how do we know if there is a C(X)-linear *-isomorphism
∼ ϕ : A −→ B E-theory for Continuous fields The idea behind E-theory
The collection HomX (A, B) of C(X)-linear *-homomorphisms is not a group, but can be replaced by a more flexible invariant, the E-theory group.
EX (A, B) For each a ∈ A, t 7→ kϕt (a)k is a bounded continuous function from T to B ∗ ∗ ϕt (a + λb) − ϕt (a) − λϕt (b) → 0
ϕt (ab) − ϕt (a)ϕt (b) → 0
ϕt (f · a) − f · ϕt (a) → 0
E-theory for Continuous fields Asymptotic morphisms
Let A and B be two C(X)-algebras. An asymptotic morphism is a family ϕt : A → B, t ∈ T := [0, ∞)
such that {ϕt } behaves like a genuine C(X)-linear *-homomorphism as t → ∞. E-theory for Continuous fields Asymptotic morphisms
Let A and B be two C(X)-algebras. An asymptotic morphism is a family ϕt : A → B, t ∈ T := [0, ∞)
such that {ϕt } behaves like a genuine C(X)-linear *-homomorphism as t → ∞.
For each a ∈ A, t 7→ kϕt (a)k is a bounded continuous function from T to B ∗ ∗ ϕt (a + λb) − ϕt (a) − λϕt (b) → 0
ϕt (ab) − ϕt (a)ϕt (b) → 0
ϕt (f · a) − f · ϕt (a) → 0 E-theory for Continuous fields Homotopy of Asymptotic morphisms
A homotopy of asymptotic morphisms ϕt , ψt : A → B is a C(X)-linear asymptotic morphism
Φt : A → C([0, 1], B)
such that Φt (a)(0) = ϕt (a), and Φt (a)(1) = ψt (a)
We denote the homotopy class of an asymptotic morphism ϕt : A → B by [ϕt ] and we write [A, B]X for the collection of homotopy classes of asymptotic morphisms from A to B. And by passing to suspensions (SB = C0((0, 1), B)); if ϕt , ψt : A → SB, we can compose them by ( ϕt (a)(2s): 0 ≤ s ≤ 1/2 ϕt · ψt : A → SB, ϕt · ψt (a)(s) = ψt (2s − 1): 1/2 ≤ s ≤ 1
The two additions coincide when we consider maps
A 7→ S(B ⊗ K)
where K denotes the compact operators.
E-theory for Continuous fields Addition of asymptotic morphisms
We may add two asymptotic morphisms ϕt , ψt : A → B by Passing to matrices ϕt (a) 0 ϕt ⊕ ψt : A → M2(B), a 7→ 0 ψt (a) The two additions coincide when we consider maps
A 7→ S(B ⊗ K)
where K denotes the compact operators.
E-theory for Continuous fields Addition of asymptotic morphisms
We may add two asymptotic morphisms ϕt , ψt : A → B by Passing to matrices ϕt (a) 0 ϕt ⊕ ψt : A → M2(B), a 7→ 0 ψt (a)
And by passing to suspensions (SB = C0((0, 1), B)); if ϕt , ψt : A → SB, we can compose them by ( ϕt (a)(2s): 0 ≤ s ≤ 1/2 ϕt · ψt : A → SB, ϕt · ψt (a)(s) = ψt (2s − 1): 1/2 ≤ s ≤ 1 E-theory for Continuous fields Addition of asymptotic morphisms
We may add two asymptotic morphisms ϕt , ψt : A → B by Passing to matrices ϕt (a) 0 ϕt ⊕ ψt : A → M2(B), a 7→ 0 ψt (a)
And by passing to suspensions (SB = C0((0, 1), B)); if ϕt , ψt : A → SB, we can compose them by ( ϕt (a)(2s): 0 ≤ s ≤ 1/2 ϕt · ψt : A → SB, ϕt · ψt (a)(s) = ψt (2s − 1): 1/2 ≤ s ≤ 1
The two additions coincide when we consider maps
A 7→ S(B ⊗ K)
where K denotes the compact operators. Functoriality Given a morphism ϕ : B → C, there is an induced group homomorphism ϕ∗ : EX (A, B) → EX (A, C)
Similarly, EX (·, B) is a contravariant functor.
E-theory for Continuous fields The E-theory group
For two separable C(X)-algebras A and B,
EX (A, B) = [SA, SB ⊗ K]X
Then, EX (A, B) is a group under the addition defined above. Moreover, we have a map
HomX (A, B) → EX (A, B), ϕ 7→ [ϕ] E-theory for Continuous fields The E-theory group
For two separable C(X)-algebras A and B,
EX (A, B) = [SA, SB ⊗ K]X
Then, EX (A, B) is a group under the addition defined above. Moreover, we have a map
HomX (A, B) → EX (A, B), ϕ 7→ [ϕ]
Functoriality Given a morphism ϕ : B → C, there is an induced group homomorphism ϕ∗ : EX (A, B) → EX (A, C)
Similarly, EX (·, B) is a contravariant functor. E-theory for Continuous fields The composition product
The composition map
HomX (A, B) × HomX (B, C) → HomX (A, C)
extends to E-theory. Composition product
EX (A, B) × EX (B, C) → EX (A, C) (α, β) 7→ β ◦ α Theorem (Kirchberg (2000)) Let A and B be separable continuous fields over X each of whose fibers are stable Kirchberg algebras. If α ∈ EX (A, B) is invertible, then ∃ an isomorphism ϕ : A → B such that
α = [ϕ]
E-theory for Continuous fields Kirchberg’s theorem
An element α ∈ EX (A, B) is called invertible if ∃β ∈ EX (B, A) such that
β ◦ α = [idA], and α ◦ β = [idB]
If ϕ : A → B is an isomorphism, then α := [ϕ] is invertible. E-theory for Continuous fields Kirchberg’s theorem
An element α ∈ EX (A, B) is called invertible if ∃β ∈ EX (B, A) such that
β ◦ α = [idA], and α ◦ β = [idB]
If ϕ : A → B is an isomorphism, then α := [ϕ] is invertible. Theorem (Kirchberg (2000)) Let A and B be separable continuous fields over X each of whose fibers are stable Kirchberg algebras. If α ∈ EX (A, B) is invertible, then ∃ an isomorphism ϕ : A → B such that
α = [ϕ] E-theory for Continuous fields Towards Classification
Question
Can we effectively compute EX (A, B)?
Can we determine when EX (A, B) has an invertible element? The goal of this project was to answer these two questions in the case where X = [0, 1] Classification Results An approximation theorem
Recall that an elementary continuous field over [0, 1] is one that is locally trivial at all but finitely many points. Theorem (Dadarlat-Elliott (2007)) Let A be a continuous field over [0, 1] each of whose fibers are inductive limits of unital semi-projective Kirchberg algebras, then A can be represented as an inductive limit ∼ A = lim An n→∞
where each An is an elementary continuous field for each n ∈ N Classification Results Fibered morphisms between Elementary fields
Suppose A, B are two elementary fields over [0, 1]. We can construct a morphism of continuous fields in the following way :
H •0 •x O γ D •x •1
ϕ1 ϕ2 •0 •x H0 O µ
•x •1 D0
0 0 where ϕ1 : H → H and ϕ2 : D → D are such that ϕ1 ◦ γ = µ ◦ ϕ2 Classification Results
In general, a fibered morphism would look like
E1 E2 E3 Ek •0 •x1 •x2 •x3 •x4 ... •xn •1 O O O O O γ1 γ2 γ3 γ4 γn D1 D2 Dk−1 •x1 •x2 •x3 •x4 ... •xn
...
ϕ1 ψ1 ϕ2 ψ2 ϕ3 ψk−1 ϕk •0 •x •x •x •x ... •x •1 E0 1 2 E0 3 4 E0 n E0 1 O O 2 O O 3 O k µ1 µ2 µ3 µ4 µn 0 0 D0 D1 D2 k−1 •x1 •x2 •x3 •x4 ... •xn with ϕ1 ◦ γ1 = µ1 ◦ ψ1, etc. We show that, under certain conditions on the fibers, all morphisms look like this. Classification Results E-theory for Elementary Fields over [0, 1]
Let F denote the class of separable C* algebras D satisfying the UCT, and such that K0(D) is free of finite rank, and K1(D) = 0. Let C(F) denote the class of elementary continuous fields over [0, 1] whose fibers are in F Theorem (Dadarlat - V (2014))
If A, B ∈ C(F), then, for every α ∈ EX (A, B), there exists a fibered morphism ϕ : A → B such that
α = [ϕ] If I ⊂ J, then we get an induced quotient map A(J) → A(I), and hence an induced map K0(A(J)) → K0(A(I))
This collection {K0(A(I)) : I ⊂ [0, 1] closed subinterval}, together with these maps, has the structure of a sheaf on the family of closed subintervals of [0, 1], denoted by K0(A)
Classification Results Definition
Let A be a C[0, 1]-algebra in F, then for each closed sub-interval I ⊂ [0, 1], we consider the K0-group
A K0(A(I)) := K0 C0(X \ I) · A These groups capture the local structure of A via K-theory. This collection {K0(A(I)) : I ⊂ [0, 1] closed subinterval}, together with these maps, has the structure of a sheaf on the family of closed subintervals of [0, 1], denoted by K0(A)
Classification Results Definition
Let A be a C[0, 1]-algebra in F, then for each closed sub-interval I ⊂ [0, 1], we consider the K0-group
A K0(A(I)) := K0 C0(X \ I) · A These groups capture the local structure of A via K-theory.
If I ⊂ J, then we get an induced quotient map A(J) → A(I), and hence an induced map K0(A(J)) → K0(A(I)) Classification Results Definition
Let A be a C[0, 1]-algebra in F, then for each closed sub-interval I ⊂ [0, 1], we consider the K0-group
A K0(A(I)) := K0 C0(X \ I) · A These groups capture the local structure of A via K-theory.
If I ⊂ J, then we get an induced quotient map A(J) → A(I), and hence an induced map K0(A(J)) → K0(A(I))
This collection {K0(A(I)) : I ⊂ [0, 1] closed subinterval}, together with these maps, has the structure of a sheaf on the family of closed subintervals of [0, 1], denoted by K0(A) If I ⊂ J, then the following diagram commutes
ϕI K0(A(I)) −−−−→ K0(B(I)) y y ϕJ K0(A(J)) −−−−→ K0(B(J))
Classification Results E-theory and K-theory
If ϕ : A → B is a morphism of C[0, 1]-algebras, then for each closed interval I ⊂ [0, 1], we get a map
ϕI : A(I) → B(I)
and hence an induced map ϕI : K0(A(I)) → K0(B(I)). Classification Results E-theory and K-theory
If ϕ : A → B is a morphism of C[0, 1]-algebras, then for each closed interval I ⊂ [0, 1], we get a map
ϕI : A(I) → B(I)
and hence an induced map ϕI : K0(A(I)) → K0(B(I)).
If I ⊂ J, then the following diagram commutes
ϕI K0(A(I)) −−−−→ K0(B(I)) y y ϕJ K0(A(J)) −−−−→ K0(B(J)) Classification Results E-theory and K-theory
Thus, we get a map
HomX (A, B) → Hom(K0(A), K0(B))
which further extends to a natural homomorphism
Γ: EX (A, B) → Hom(K0(A), K0(B)) In particular, ∼ ∼ A = B ⇔ K0(A) = K0(B)
Classification Results A Classification Theorem
Theorem (Dadarlat-V (2014)) Let A and B be separable continuous fields over [0, 1] whose fibers are Kirchberg algebras satisfying the Universal Coefficient theorem with torsion-free K0 groups and zero K1 groups, then
Γ: EX (A, B) → Hom(K0(A), K0(B))
is an isomorphism. Classification Results A Classification Theorem
Theorem (Dadarlat-V (2014)) Let A and B be separable continuous fields over [0, 1] whose fibers are Kirchberg algebras satisfying the Universal Coefficient theorem with torsion-free K0 groups and zero K1 groups, then
Γ: EX (A, B) → Hom(K0(A), K0(B))
is an isomorphism.
In particular, ∼ ∼ A = B ⇔ K0(A) = K0(B) Use the approximation result to write
A = lim An, and B = lim Bk
where An, Bk are elementary fields. Use a continuity property of K0(·), together with Kirchberg’s theorem and Elliott’s intertwining argument.
Classification Results The idea behind the proof
First prove that, when A and B are both elementary fields, the map
Γ: EX (A, B) → Hom(K0(A), K0(B))
is an isomorphism. Use a continuity property of K0(·), together with Kirchberg’s theorem and Elliott’s intertwining argument.
Classification Results The idea behind the proof
First prove that, when A and B are both elementary fields, the map
Γ: EX (A, B) → Hom(K0(A), K0(B))
is an isomorphism. Use the approximation result to write
A = lim An, and B = lim Bk
where An, Bk are elementary fields. Classification Results The idea behind the proof
First prove that, when A and B are both elementary fields, the map
Γ: EX (A, B) → Hom(K0(A), K0(B))
is an isomorphism. Use the approximation result to write
A = lim An, and B = lim Bk
where An, Bk are elementary fields. Use a continuity property of K0(·), together with Kirchberg’s theorem and Elliott’s intertwining argument. Thank you for listening!