Operator Algebras and Noncommutative Geometry
David E Evans
Cardiff University
LMS Graduate Meeting 15 November 2013
1 / 24 Overview
Gelfand Theorem
operator algebras
dimension
rotation algebra and orbifolds
Stone-von Neumann theorem
2 / 24 characters χx : A → C , f → f (x) Hom(A, C) Irred Rep (A, B(H)) Spectrum Spec(A)
recover topology of [−1, 1] from C[−1.1]
Algebra of continuous functions
A = C[−1, 1] = continuous maps f :[−1, 1] → C
||f || = sup x |f (x)|
3 / 24 Irred Rep (A, B(H)) Spectrum Spec(A)
recover topology of [−1, 1] from C[−1.1]
Algebra of continuous functions
A = C[−1, 1] = continuous maps f :[−1, 1] → C
||f || = sup x |f (x)|
characters χx : A → C , f → f (x) Hom(A, C)
3 / 24 recover topology of [−1, 1] from C[−1.1]
Algebra of continuous functions
A = C[−1, 1] = continuous maps f :[−1, 1] → C
||f || = sup x |f (x)|
characters χx : A → C , f → f (x) Hom(A, C) Irred Rep (A, B(H)) Spectrum Spec(A)
3 / 24 Algebra of continuous functions
A = C[−1, 1] = continuous maps f :[−1, 1] → C
||f || = sup x |f (x)|
characters χx : A → C , f → f (x) Hom(A, C) Irred Rep (A, B(H)) Spectrum Spec(A)
recover topology of [−1, 1] from C[−1.1]
3 / 24 2 Represent f ∈ C[−1, 1] on L (T) by multiplication operators
2 π(f ) ∈ B(L (T)) (π(f )g)(x) = f (x)g(x)
2 A = C[−1, 1] ⊂ B(L (T))
Geland-Naimark Theorem
If A is a Banach ∗-algebra where
kT ∗T k = kT k2, T ∈ A then A is a C ∗-algebra, — there exists an isometric ∗-homomorphism π : A → B(H) for a Hilbert space H.
4 / 24 Geland-Naimark Theorem
If A is a Banach ∗-algebra where
kT ∗T k = kT k2, T ∈ A then A is a C ∗-algebra, — there exists an isometric ∗-homomorphism π : A → B(H) for a Hilbert space H.
2 Represent f ∈ C[−1, 1] on L (T) by multiplication operators
2 π(f ) ∈ B(L (T)) (π(f )g)(x) = f (x)g(x)
2 A = C[−1, 1] ⊂ B(L (T))
4 / 24 H = H∗ ∈ B(H), C ∗(H, 1) ' C(σ(H)) Spec(A) = σ(H) f (H), f ∈ C(σ(H)) x → exp(itx), U(t) = exp(itH) U(t)U(s) = U(s + t)
Gelfand Theorem
If A is a commutative, unital Banach ∗-algebra with kf ∗f k = kf k2, f ∈ A then A ' C(Ω) for a compact Hausdorff Ω.
5 / 24 Gelfand Theorem
If A is a commutative, unital Banach ∗-algebra with kf ∗f k = kf k2, f ∈ A then A ' C(Ω) for a compact Hausdorff Ω.
H = H∗ ∈ B(H), C ∗(H, 1) ' C(σ(H)) Spec(A) = σ(H) f (H), f ∈ C(σ(H)) x → exp(itx), U(t) = exp(itH) U(t)U(s) = U(s + t)
5 / 24 A separable , Spec A = • =⇒ A 'K(H)
B = continuous maps from [−1, 1] → M2
Spec(B) = [−1, 1] f → f (x)
noncommutative algebra
points to matrices Mn representations are all multiples of x → x
Spectrum of Mn = singleton • Spectrum K(H) = •
6 / 24 B = continuous maps from [−1, 1] → M2
Spec(B) = [−1, 1] f → f (x)
noncommutative algebra
points to matrices Mn representations are all multiples of x → x
Spectrum of Mn = singleton • Spectrum K(H) = •
A separable , Spec A = • =⇒ A 'K(H)
6 / 24 noncommutative algebra
points to matrices Mn representations are all multiples of x → x
Spectrum of Mn = singleton • Spectrum K(H) = •
A separable , Spec A = • =⇒ A 'K(H)
B = continuous maps from [−1, 1] → M2
Spec(B) = [−1, 1] f → f (x)
6 / 24 Z2 action on B = C([−1, 1], M2) 0 1 by x → −x and U()U where U = 1 0 f → g(x) = Uf (−x)U 2 BZ2 ⊂ B : f (x) ∈ M2, x ∈ (0, 1] and f (0) = Uf (0)U i.e. f (0) ∈ C Spec(BZ2 ) =
∗ BZ2 = C ([−1, 1]/Z2)
orbifold
[−1, 1] under Z2 action x → −x
7 / 24 Z2 action on B = C([−1, 1], M2) 0 1 by x → −x and U()U where U = 1 0 f → g(x) = Uf (−x)U 2 BZ2 ⊂ B : f (x) ∈ M2, x ∈ (0, 1] and f (0) = Uf (0)U i.e. f (0) ∈ C Spec(BZ2 ) =
∗ BZ2 = C ([−1, 1]/Z2)
orbifold
[−1, 1] under Z2 action x → −x
7 / 24 Z2 action on B = C([−1, 1], M2) 0 1 by x → −x and U()U where U = 1 0 f → g(x) = Uf (−x)U 2 BZ2 ⊂ B : f (x) ∈ M2, x ∈ (0, 1] and f (0) = Uf (0)U i.e. f (0) ∈ C Spec(BZ2 ) =
∗ BZ2 = C ([−1, 1]/Z2)
orbifold
[−1, 1] under Z2 action x → −x
7 / 24 Z2 action on B = C([−1, 1], M2) 0 1 by x → −x and U()U where U = 1 0 f → g(x) = Uf (−x)U 2 BZ2 ⊂ B : f (x) ∈ M2, x ∈ (0, 1] and f (0) = Uf (0)U i.e. f (0) ∈ C Spec(BZ2 ) =
∗ BZ2 = C ([−1, 1]/Z2)
orbifold
[−1, 1] under Z2 action x → −x
7 / 24 0 1 by x → −x and U()U where U = 1 0 f → g(x) = Uf (−x)U 2 BZ2 ⊂ B : f (x) ∈ M2, x ∈ (0, 1] and f (0) = Uf (0)U i.e. f (0) ∈ C Spec(BZ2 ) =
∗ BZ2 = C ([−1, 1]/Z2)
orbifold
[−1, 1] under Z2 action x → −x
Z2 action on B = C([−1, 1], M2)
7 / 24 2 BZ2 ⊂ B : f (x) ∈ M2, x ∈ (0, 1] and f (0) = Uf (0)U i.e. f (0) ∈ C Spec(BZ2 ) =
∗ BZ2 = C ([−1, 1]/Z2)
orbifold
[−1, 1] under Z2 action x → −x
Z2 action on B = C([−1, 1], M2) 0 1 by x → −x and U()U where U = 1 0 f → g(x) = Uf (−x)U
7 / 24 Spec(BZ2 ) =
∗ BZ2 = C ([−1, 1]/Z2)
orbifold
[−1, 1] under Z2 action x → −x
Z2 action on B = C([−1, 1], M2) 0 1 by x → −x and U()U where U = 1 0 f → g(x) = Uf (−x)U 2 BZ2 ⊂ B : f (x) ∈ M2, x ∈ (0, 1] and f (0) = Uf (0)U i.e. f (0) ∈ C
7 / 24 orbifold
[−1, 1] under Z2 action x → −x
Z2 action on B = C([−1, 1], M2) 0 1 by x → −x and U()U where U = 1 0 f → g(x) = Uf (−x)U 2 BZ2 ⊂ B : f (x) ∈ M2, x ∈ (0, 1] and f (0) = Uf (0)U i.e. f (0) ∈ C Spec(BZ2 ) =
∗ BZ2 = C ([−1, 1]/Z2) 7 / 24 Iterating
C → M2 → M4 → M8 → · · · → M2n → M2n+1 → · · ·
Nn Nn 2 M2 ' M2n ' End( C )
Nn M2 ' M2n
the Fermion algebra
M2n → M2n+2n = M2n+1 ' M2n ⊗ M2 a 0 a → ' a ⊗ 1 0 a
8 / 24 the Fermion algebra
M2n → M2n+2n = M2n+1 ' M2n ⊗ M2 a 0 a → ' a ⊗ 1 0 a Iterating
C → M2 → M4 → M8 → · · · → M2n → M2n+1 → · · ·
Nn Nn 2 M2 ' M2n ' End( C )
Nn M2 ' M2n
8 / 24 C → M2 → M4 → M8 → · · · → M2n → M2n+1 → · · ·
X n Trace(a) = aii /2
a 0 a → (1) 0 a
e = e∗ = e2 : 1 2 3 dim(e) = trace(e) ∈ {0, 2n , 2n , 2n ..., 1} → N[1/2]
K0(⊗NM2) = Z[1/2]
Dimensions of Operator Algebras
Projections e = e2 = e∗ in A ⊂ B(H) correspond to decompositions H = eH ⊕ eH⊥
9 / 24 e = e∗ = e2 : 1 2 3 dim(e) = trace(e) ∈ {0, 2n , 2n , 2n ..., 1} → N[1/2]
K0(⊗NM2) = Z[1/2]
Dimensions of Operator Algebras
Projections e = e2 = e∗ in A ⊂ B(H) correspond to decompositions H = eH ⊕ eH⊥
C → M2 → M4 → M8 → · · · → M2n → M2n+1 → · · ·
X n Trace(a) = aii /2
a 0 a → (1) 0 a
9 / 24 K0(⊗NM2) = Z[1/2]
Dimensions of Operator Algebras
Projections e = e2 = e∗ in A ⊂ B(H) correspond to decompositions H = eH ⊕ eH⊥
C → M2 → M4 → M8 → · · · → M2n → M2n+1 → · · ·
X n Trace(a) = aii /2
a 0 a → (1) 0 a
e = e∗ = e2 : 1 2 3 dim(e) = trace(e) ∈ {0, 2n , 2n , 2n ..., 1} → N[1/2]
9 / 24 Dimensions of Operator Algebras
Projections e = e2 = e∗ in A ⊂ B(H) correspond to decompositions H = eH ⊕ eH⊥
C → M2 → M4 → M8 → · · · → M2n → M2n+1 → · · ·
X n Trace(a) = aii /2
a 0 a → (1) 0 a
e = e∗ = e2 : 1 2 3 dim(e) = trace(e) ∈ {0, 2n , 2n , 2n ..., 1} → N[1/2]
K0(⊗NM2) = Z[1/2]
9 / 24 discrete rotation matrices
0 1 0 0 0 V = ·· · 1 1 0 0 1 0 0 ω U = ·· · 0 0 ωn−1 ωn = 1
VUV ∗ = ωU
r Ver = er−1 Uer = ω er 10 / 24 −1 ∗ ∼ Ad V = V ( · )V leaves C (U) = C(T) invariant and acts by rotation through θ on the circle
if θ is irrational, Aθ is simple, unique 2 cf. C(T ) when θ = 0.
rotation algebra
2 multiplication and translation operators on L (T),
Uh(z) = zh(z)
Vh(z) = h(exp(2πiθ)z)
VU = exp(2πiθ)UV
∗ Aθ = C (U, V ) VUV −1 = exp(2πiθ)U;
11 / 24 if θ is irrational, Aθ is simple, unique 2 cf. C(T ) when θ = 0.
rotation algebra
2 multiplication and translation operators on L (T),
Uh(z) = zh(z)
Vh(z) = h(exp(2πiθ)z)
VU = exp(2πiθ)UV
∗ Aθ = C (U, V ) VUV −1 = exp(2πiθ)U; −1 ∗ ∼ Ad V = V ( · )V leaves C (U) = C(T) invariant and acts by rotation through θ on the circle
11 / 24 rotation algebra
2 multiplication and translation operators on L (T),
Uh(z) = zh(z)
Vh(z) = h(exp(2πiθ)z)
VU = exp(2πiθ)UV
∗ Aθ = C (U, V ) VUV −1 = exp(2πiθ)U; −1 ∗ ∼ Ad V = V ( · )V leaves C (U) = C(T) invariant and acts by rotation through θ on the circle
if θ is irrational, Aθ is simple, unique 2 cf. C(T ) when θ = 0.
11 / 24 trace on rotation algebra
2 T action α on Aθ by
2 (t1, t2) ∈ T
U → t1U, V → t2V
RR unique trace τ = αs,t ds dt, Z n τ(f (U)V ) = f (t) dt δn,0
K-group of dimensions generated by trace of projections
12 / 24 ∼ K0(Aθ) = Z + θZ (m, n) n + θm > 0 ∼ 2 2 K1(Aθ) = Z , If θ = 0, K1(C(T )) is generated by the coordinate unitary maps (z1, z2) → zi , i = 1, 2, i.e. U and V .
13 / 24 F pF (1 − F ) e = χ + [b,c] pF (1 − F ) 1 − F
−1 ∗ = V g + f + gV ∈ C (U, V ) ≡ Aθ ,
F f = χ + [b,c] 1 − F p g continuous function = F (1 − F )χ[a,b] Z τ(eθ) = f = θ
projections in rotation algebras
14 / 24 F pF (1 − F ) e = χ + [b,c] pF (1 − F ) 1 − F
−1 ∗ = V g + f + gV ∈ C (U, V ) ≡ Aθ ,
F f = χ + [b,c] 1 − F p g continuous function = F (1 − F )χ[a,b] Z τ(eθ) = f = θ
projections in rotation algebras
14 / 24 F pF (1 − F ) e = χ + [b,c] pF (1 − F ) 1 − F
−1 ∗ = V g + f + gV ∈ C (U, V ) ≡ Aθ ,
F f = χ + [b,c] 1 − F p g continuous function = F (1 − F )χ[a,b] Z τ(eθ) = f = θ
projections in rotation algebras
14 / 24 F pF (1 − F ) e = χ + [b,c] pF (1 − F ) 1 − F
−1 ∗ = V g + f + gV ∈ C (U, V ) ≡ Aθ ,
F f = χ + [b,c] 1 − F p g continuous function = F (1 − F )χ[a,b] Z τ(eθ) = f = θ
projections in rotation algebras
14 / 24 −1 ∗ = V g + f + gV ∈ C (U, V ) ≡ Aθ ,
F f = χ + [b,c] 1 − F p g continuous function = F (1 − F )χ[a,b] Z τ(eθ) = f = θ
projections in rotation algebras
F pF (1 − F ) e = χ + [b,c] pF (1 − F ) 1 − F
14 / 24 −1 ∗ = V g + f + gV ∈ C (U, V ) ≡ Aθ ,
p g continuous function = F (1 − F )χ[a,b] Z τ(eθ) = f = θ
projections in rotation algebras
F pF (1 − F ) e = χ + [b,c] pF (1 − F ) 1 − F
F f = χ + [b,c] 1 − F
14 / 24 projections in rotation algebras
F pF (1 − F ) e = χ + [b,c] pF (1 − F ) 1 − F
−1 ∗ = V g + f + gV ∈ C (U, V ) ≡ Aθ ,
F f = χ + [b,c] 1 − F p g continuous function = F (1 − F )χ[a,b] 14 / 24 Z τ(eθ) = f = θ projections in rotation algebras
F pF (1 − F ) e = χ + [b,c] pF (1 − F ) 1 − F
−1 ∗ = V g + f + gV ∈ C (U, V ) ≡ Aθ ,
Z τ(eθ) = f = θ
15 / 24 toroidal orbifold Z/2 action on T2
2 The flip (z1, z2) → (z1, z2) on T
180◦
2 The orbifold T /(Z/2) yields a sphere.
16 / 24 2 Toroidal orbifold, T /(Z/2)
2 C(T ) o (Z/2) = M2 valued functions on the sphere, restricted to 2 C at 4 points
toroidal orbifold
17 / 24 2 Toroidal orbifold, T /(Z/2)
2 C(T ) o (Z/2) = M2 valued functions on the sphere, restricted to 2 C at 4 points
toroidal orbifold
17 / 24 2 C(T ) o (Z/2) = M2 valued functions on the sphere, restricted to 2 C at 4 points
toroidal orbifold
2 Toroidal orbifold, T /(Z/2)
17 / 24 2 C(T ) o (Z/2) = M2 valued functions on the sphere, restricted to 2 C at 4 points
toroidal orbifold
2 Toroidal orbifold, T /(Z/2)
17 / 24 toroidal orbifold
2 Toroidal orbifold, T /(Z/2)
2 C(T ) o (Z/2) = M2 valued functions on the sphere, restricted to 2 C at 4 points 17 / 24 non-commutative toroidal orbifold
symmetry U, V → U−1, V −1 preserves relation VU = e2πiθUV .
Z/2 Aθ ⊂ Aθ ⊂ Aθ o (Z/2)
2 C(T/Z) = C = C(T /R)
2 C(T) o Z ∼Morita C(T ) o R
18 / 24 Aθ = lim→Mr (C(T)) ⊕ Ms (C(T))
Z2 Aθ = lim→Mr1 (C) ⊕ Mr2 (C) ⊕ · · ·
non-commutative toroidal orbifold
19 / 24 Z2 Aθ = lim→Mr1 (C) ⊕ Mr2 (C) ⊕ · · ·
non-commutative toroidal orbifold
Aθ = lim→Mr (C(T)) ⊕ Ms (C(T))
19 / 24 non-commutative toroidal orbifold
Aθ = lim→Mr (C(T)) ⊕ Ms (C(T))
Z2 Aθ = lim→Mr1 (C) ⊕ Mr2 (C) ⊕ · · ·
19 / 24 almost Mathieu operator
Z/2 ∗ −1 −1 Aθ = C (U + U , V + V ) is AF Hamiltonians
H = U + U−1 + λ(V + V −1)
Uh(z) = zh(z), Vh(z) = h(exp(2πiθ)z). almost Mathieu operator:
f → f (n + 1) + f (n − 1) + 2λ cos(2πnθ)f (n)
has Cantor spectrum for θ irrational, λ 6= 0.
20 / 24 PQ − QP = −i
irreducible Schr¨odingerrepresentation is
2 Q = x , P = −id/dx on L (R)
Stone-von Neumann Theorem
U(m) = Um, V (n) = V n U(m)V (n) = exp(2πimnθ)V (n)U(m) , m, n ∈ Z Continuous version – CCR Canonical Commutation Relations U(t)V (s) = exp(2πistθ)V (s)U(t) , s, t ∈ R U(t) = exp(itP), V (s) = exp(isQ)
21 / 24 Stone-von Neumann Theorem
U(m) = Um, V (n) = V n U(m)V (n) = exp(2πimnθ)V (n)U(m) , m, n ∈ Z Continuous version – CCR Canonical Commutation Relations U(t)V (s) = exp(2πistθ)V (s)U(t) , s, t ∈ R U(t) = exp(itP), V (s) = exp(isQ)
PQ − QP = −i
irreducible Schr¨odingerrepresentation is
2 Q = x , P = −id/dx on L (R)
21 / 24 Mackey-Rieffel picture:
Reps of CCR ∼ Reps of C0(R) o R ∼Morita C
∗ C0(G/H) o G ∼Morita C (H)
Stone-von Neumann theorem
There is an unique irreducible representation of the Canonical Commutation Relations.
22 / 24 ∗ C0(G/H) o G ∼Morita C (H)
Stone-von Neumann theorem
There is an unique irreducible representation of the Canonical Commutation Relations.
Mackey-Rieffel picture:
Reps of CCR ∼ Reps of C0(R) o R ∼Morita C
22 / 24 Stone-von Neumann theorem
There is an unique irreducible representation of the Canonical Commutation Relations.
Mackey-Rieffel picture:
Reps of CCR ∼ Reps of C0(R) o R ∼Morita C
∗ C0(G/H) o G ∼Morita C (H)
22 / 24 References
Akeman, C. and Weaver, N. (2004) Consistency of a counter example to Naimark’s problem. Proc. Nat. Acad. Sci. USA. 101, 7522-7525 Bratteli, O. and Kishimoto, A. (1992). Non-commutative spheres, III, Irrational rotations. Commun, in Mathematical Physics, 147, 605–624. Connes, A. (1994). Noncommutative geometry. Academic Press, New York. Elliott, G. A. and Evans, D. E. (1993). The structure of the irrational rotation C ∗-algebra. Annals of Mathematics, 138, 477–501. Evans, D.E. (1992). Some non-commutative orbifolds. Ideas and methods in mathematical analysis, stochastics, and applications. eds S.Albeverio, J.E. Fenstad, H. Holden, T. Lindstrom. Cambridge University Press. 344–364. Evans, D. E. and Kawahigashi, Y. (1998). Quantum Symmetries on Operator Algebras. Oxford University Press. Mackey, G.W. (1949). A theorem of Stone and von Neumann. Duke Math. J. 16, 313-326. Rieffel, M.A. (1972). On the uniqueness of the Heisenberg commutation relations. Duke Math. J. 39 745-752. Rieffel, M. A. (1981). C ∗-algebras associated with irrational rotations. Pacific Journal of Mathematics, 93, 415–429. Rosenberg, J. A. (2004). Selective History of the Stone-von Neumann Theorem. Operator algebras, quantization, and noncommutative geometry,Contemp. Math., 365, Amer. Math. Soc., Providence, RI, 123-158. Stone M. H. (1930). Linear transformations in Hilbert space, III: operational methods and group theory, Proc. Nat. Acad. Sci. U.S.A. 16 172-175. von Neumann, J. (1931). Die Eindeutigkeit der Schr¨odingerischenOperatoren. Math. Ann. 104 570-578. 23 / 24