Introduction to C⇤-Algebras
Thomas Sinclair
Model Theory of Operator Algebras UC Irvine
September 20, 2017
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 1 / 52 Hilbert Spaces
I H complex Hilbert space with ( , )innerproduct · · 1/2 I ⇠ := (⇠,⇠) (complete) norm k k I H⇤ continuous linear functionals : H C ! I (F. Reisz) H then ⌘ H ⇠ H( (⇠)=(⇠,⌘)) 2 ⇤ 9 2 8 2
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 2 / 52 Hilbert Spaces
I H complex Hilbert space with ( , )innerproduct · · 1/2 I ⇠ := (⇠,⇠) (complete) norm k k I H⇤ continuous linear functionals : H C ! I (F. Reisz) H then ⌘ H ⇠ H( (⇠)=(⇠,⌘)) 2 ⇤ 9 2 8 2
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 2 / 52 Hilbert Spaces
I H complex Hilbert space with ( , )innerproduct · · 1/2 I ⇠ := (⇠,⇠) (complete) norm k k I H⇤ continuous linear functionals : H C ! I (F. Reisz) H then ⌘ H ⇠ H( (⇠)=(⇠,⌘)) 2 ⇤ 9 2 8 2
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 2 / 52 Hilbert Spaces
I H complex Hilbert space with ( , )innerproduct · · 1/2 I ⇠ := (⇠,⇠) (complete) norm k k I H⇤ continuous linear functionals : H C ! I (F. Reisz) H then ⌘ H ⇠ H( (⇠)=(⇠,⌘)) 2 ⇤ 9 2 8 2
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 2 / 52 Bounded Operators
I (H, K)= T : H K : T linear, continuous B { ! } I operator norm T ⇠ T := sup k k < k k ⇠=0 ⇠ 1 6 k k I (H, K) complete B I T =sup (T ⇠,⌘) k k ⇠ = ⌘ =1 | | k k k k Proof. Wlog ⌘ T (H) 2
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 3 / 52 Bounded Operators
I (H, K)= T : H K : T linear, continuous B { ! } I operator norm T ⇠ T := sup k k < k k ⇠=0 ⇠ 1 6 k k I (H, K) complete B I T =sup (T ⇠,⌘) k k ⇠ = ⌘ =1 | | k k k k Proof. Wlog ⌘ T (H) 2
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 3 / 52 Bounded Operators
I (H, K)= T : H K : T linear, continuous B { ! } I operator norm T ⇠ T := sup k k < k k ⇠=0 ⇠ 1 6 k k I (H, K) complete B I T =sup (T ⇠,⌘) k k ⇠ = ⌘ =1 | | k k k k Proof. Wlog ⌘ T (H) 2
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 3 / 52 Bounded Operators
I (H, K)= T : H K : T linear, continuous B { ! } I operator norm T ⇠ T := sup k k < k k ⇠=0 ⇠ 1 6 k k I (H, K) complete B I T =sup (T ⇠,⌘) k k ⇠ = ⌘ =1 | | k k k k Proof. Wlog ⌘ T (H) 2
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 3 / 52 Bounded Operators
I (H):= (H, H) B B I ST S T k kk k·k k I (H) A(⇠,⌘) C : A(⇠,⌘) C ⇠ ⌘ , bilinear B ${ ! | | k k·k k } I Adjoint T (⇠,T ⌘) ⇤ $
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 4 / 52 Bounded Operators
I (H):= (H, H) B B I ST S T k kk k·k k I (H) A(⇠,⌘) C : A(⇠,⌘) C ⇠ ⌘ , bilinear B ${ ! | | k k·k k } I Adjoint T (⇠,T ⌘) ⇤ $
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 4 / 52 Bounded Operators
I (H):= (H, H) B B I ST S T k kk k·k k I (H) A(⇠,⌘) C : A(⇠,⌘) C ⇠ ⌘ , bilinear B ${ ! | | k k·k k } I Adjoint T (⇠,T ⌘) ⇤ $
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 4 / 52 Bounded Operators
I (H):= (H, H) B B I ST S T k kk k·k k I (H) A(⇠,⌘) C : A(⇠,⌘) C ⇠ ⌘ , bilinear B ${ ! | | k k·k k } I Adjoint T (⇠,T ⌘) ⇤ $
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 4 / 52 Adjoints
I (T ⇤)⇤ = T
I ( T )⇤ = ¯T ⇤,(ST )⇤ = T ⇤S⇤ I T = T k ⇤k k k 2 I T T = T k ⇤ k k k Proof.
2 sup (T ⇤T ⇠,⌘) =sup (T ⇠,T ⌘) =sup T ⇠ ⇠ = ⌘ =1 | | ⇠ = ⌘ =1 | | ⇠ =1 k k k k k k k k k k k k
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 5 / 52 Adjoints
I (T ⇤)⇤ = T
I ( T )⇤ = ¯T ⇤,(ST )⇤ = T ⇤S⇤ I T = T k ⇤k k k 2 I T T = T k ⇤ k k k Proof.
2 sup (T ⇤T ⇠,⌘) =sup (T ⇠,T ⌘) =sup T ⇠ ⇠ = ⌘ =1 | | ⇠ = ⌘ =1 | | ⇠ =1 k k k k k k k k k k k k
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 5 / 52 Adjoints
I (T ⇤)⇤ = T
I ( T )⇤ = ¯T ⇤,(ST )⇤ = T ⇤S⇤ I T = T k ⇤k k k 2 I T T = T k ⇤ k k k Proof.
2 sup (T ⇤T ⇠,⌘) =sup (T ⇠,T ⌘) =sup T ⇠ ⇠ = ⌘ =1 | | ⇠ = ⌘ =1 | | ⇠ =1 k k k k k k k k k k k k
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 5 / 52 Adjoints
I (T ⇤)⇤ = T
I ( T )⇤ = ¯T ⇤,(ST )⇤ = T ⇤S⇤ I T = T k ⇤k k k 2 I T T = T k ⇤ k k k Proof.
2 sup (T ⇤T ⇠,⌘) =sup (T ⇠,T ⌘) =sup T ⇠ ⇠ = ⌘ =1 | | ⇠ = ⌘ =1 | | ⇠ =1 k k k k k k k k k k k k
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 5 / 52 C⇤-Algebras
Definition A concrete C -algebra is a norm closed, adjoint closed, subalgebra of (H) ⇤ B for some Hilbert space H.
Definition
AC⇤-algebra is a: I complex Banach algebra - ab a b k kk k·k k I with involution - ( a)⇤ = ¯a⇤,(a⇤)⇤ = a
I which satisfies the C⇤-identity
2 a⇤a = a . k k k k
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 6 / 52 C⇤-Algebras
Definition A concrete C -algebra is a norm closed, adjoint closed, subalgebra of (H) ⇤ B for some Hilbert space H.
Definition
AC⇤-algebra is a: I complex Banach algebra - ab a b k kk k·k k I with involution - ( a)⇤ = ¯a⇤,(a⇤)⇤ = a
I which satisfies the C⇤-identity
2 a⇤a = a . k k k k
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 6 / 52 C⇤-Algebras
Definition A concrete C -algebra is a norm closed, adjoint closed, subalgebra of (H) ⇤ B for some Hilbert space H.
Definition
AC⇤-algebra is a: I complex Banach algebra - ab a b k kk k·k k I with involution - ( a)⇤ = ¯a⇤,(a⇤)⇤ = a
I which satisfies the C⇤-identity
2 a⇤a = a . k k k k
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 6 / 52 C⇤-Algebras
Theorem (Big Theorem) Every C -algebra is isometrically -isomorphic to a concrete C -algebra. ⇤ ⇤ ⇤ Done by Gelfand and Naimark in 1943 with extra axioms. It took almost two decades and the work of many others (Segal, Kaplansky, etc.) to reach this form.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 7 / 52 C⇤-Algebras
Theorem (Big Theorem) Every C -algebra is isometrically -isomorphic to a concrete C -algebra. ⇤ ⇤ ⇤ Done by Gelfand and Naimark in 1943 with extra axioms. It took almost two decades and the work of many others (Segal, Kaplansky, etc.) to reach this form.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 7 / 52 Theorem
Every finite dimensional C⇤-algebra is of this form.
Examples - Finite Dimension
Here are some examples to finite-dimensional C⇤-algebras: I C 2 I Mn(C)= (` ) B n I Mn1 (C) Mnk (C) ···
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 8 / 52 Theorem
Every finite dimensional C⇤-algebra is of this form.
Examples - Finite Dimension
Here are some examples to finite-dimensional C⇤-algebras: I C 2 I Mn(C)= (` ) B n I Mn1 (C) Mnk (C) ···
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 8 / 52 Theorem
Every finite dimensional C⇤-algebra is of this form.
Examples - Finite Dimension
Here are some examples to finite-dimensional C⇤-algebras: I C 2 I Mn(C)= (` ) B n I Mn1 (C) Mnk (C) ···
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 8 / 52 Examples - Finite Dimension
Here are some examples to finite-dimensional C⇤-algebras: I C 2 I Mn(C)= (` ) B n I Mn1 (C) Mnk (C) ··· Theorem
Every finite dimensional C⇤-algebra is of this form.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 8 / 52 Examples - Abelian
X is a set. For f ` (X ), M (`2(X )) by 2 1 f 2B M ⇠(x):=f (x)⇠(x), x X f 8 2 Theorem f M gives an isometric -embedding ` (X ) , (`2(X )) 7! f ⇤ 1 !B Proof.
I Mf¯ = Mf⇤ I Mfg = Mf Mg = Mg Mf I Mf =sup (Mf ⇠,⌘) =sup ( , f ) =sup f (x) k k ⇠ = ⌘ =1 | | =1 | | x | | k k2 k k2 k k1
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 9 / 52 Examples - Abelian
X is a set. For f ` (X ), M (`2(X )) by 2 1 f 2B M ⇠(x):=f (x)⇠(x), x X f 8 2 Theorem f M gives an isometric -embedding ` (X ) , (`2(X )) 7! f ⇤ 1 !B Proof.
I Mf¯ = Mf⇤ I Mfg = Mf Mg = Mg Mf I Mf =sup (Mf ⇠,⌘) =sup ( , f ) =sup f (x) k k ⇠ = ⌘ =1 | | =1 | | x | | k k2 k k2 k k1
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 9 / 52 Examples - Abelian
X is a set. For f ` (X ), M (`2(X )) by 2 1 f 2B M ⇠(x):=f (x)⇠(x), x X f 8 2 Theorem f M gives an isometric -embedding ` (X ) , (`2(X )) 7! f ⇤ 1 !B Proof.
I Mf¯ = Mf⇤ I Mfg = Mf Mg = Mg Mf I Mf =sup (Mf ⇠,⌘) =sup ( , f ) =sup f (x) k k ⇠ = ⌘ =1 | | =1 | | x | | k k2 k k2 k k1
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 9 / 52 Fact C(X ) is a closed -subalgebra of ` (Xˇ) (`2(Xˇ)) ⇤ 1 ⇢B
I “Xˇ”isX considered as a (discrete) set.
Examples - Abelian
X compact, Hausdor↵.
C(X ):= f : X C : f is continuous { ! }
Fact
C(X ) is an (abstract) C⇤-algebra under pointwise multiplication, pointwise conjugation, and “sup-norm”.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 10 / 52 I “Xˇ”isX considered as a (discrete) set.
Examples - Abelian
X compact, Hausdor↵.
C(X ):= f : X C : f is continuous { ! }
Fact
C(X ) is an (abstract) C⇤-algebra under pointwise multiplication, pointwise conjugation, and “sup-norm”.
Fact C(X ) is a closed -subalgebra of ` (Xˇ) (`2(Xˇ)) ⇤ 1 ⇢B
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 10 / 52 Examples - Abelian
X compact, Hausdor↵.
C(X ):= f : X C : f is continuous { ! }
Fact
C(X ) is an (abstract) C⇤-algebra under pointwise multiplication, pointwise conjugation, and “sup-norm”.
Fact C(X ) is a closed -subalgebra of ` (Xˇ) (`2(Xˇ)) ⇤ 1 ⇢B
I “Xˇ”isX considered as a (discrete) set.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 10 / 52 Examples - Abelian
X locally compact, Hausdor↵. C0(X )isallf : X C continuous s.t. ! ✏>0 K X compact( f (x) ✏ x K) 8 9 ⇢ | | ) 2
I C0(X )isan(abstract)C⇤-algebra under pointwise multiplication, pointwise conjugation, and “sup-norm”.
I C (X ) , C( X ), X = Stone–Cechˇ compactification. 0 ! I C0(X )isunitali↵X is compact.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 11 / 52 Examples - Abelian
X locally compact, Hausdor↵. C0(X )isallf : X C continuous s.t. ! ✏>0 K X compact( f (x) ✏ x K) 8 9 ⇢ | | ) 2
I C0(X )isan(abstract)C⇤-algebra under pointwise multiplication, pointwise conjugation, and “sup-norm”.
I C (X ) , C( X ), X = Stone–Cechˇ compactification. 0 ! I C0(X )isunitali↵X is compact.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 11 / 52 Examples - Abelian
X locally compact, Hausdor↵. C0(X )isallf : X C continuous s.t. ! ✏>0 K X compact( f (x) ✏ x K) 8 9 ⇢ | | ) 2
I C0(X )isan(abstract)C⇤-algebra under pointwise multiplication, pointwise conjugation, and “sup-norm”.
I C (X ) , C( X ), X = Stone–Cechˇ compactification. 0 ! I C0(X )isunitali↵X is compact.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 11 / 52 Examples - Abelian
X locally compact, Hausdor↵. C0(X )isallf : X C continuous s.t. ! ✏>0 K X compact( f (x) ✏ x K) 8 9 ⇢ | | ) 2
I C0(X )isan(abstract)C⇤-algebra under pointwise multiplication, pointwise conjugation, and “sup-norm”.
I C (X ) , C( X ), X = Stone–Cechˇ compactification. 0 ! I C0(X )isunitali↵X is compact.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 11 / 52 Abelian C⇤-Algebras
Theorem (Gelfand–Naimark, 1943) An abelian C -algebra is isometrically -isomorphic to C (X ) for some X ⇤ ⇤ 0 locally compact, Hausdor↵.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 12 / 52 More Examples
Definition An operator T (H)isfiniterankif 2B n T = ( ,⌘ )⇠ · i i Xi=1 The compact operators (H) (H) are the closure of the finite rank K ⇢B operators.
Fact (H)= (H) if dim H < . Otherwise (H) is a maximal proper closed K B 1 K ideal in (H). B
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 13 / 52 More Examples
Fact 2 (` )= 1 Mn(C) where K i=1 S A 0 Mn(C) A Mn+1(C). 3 7! 002 ✓ ◆ Example
The CAR algebra is i1=1 M2n (C)where
S A 0 M2n (C) A M2n+1 (C). 3 7! 0 A 2 ✓ ◆
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 14 / 52 Some Operator Theory
Definition An operator T (H)issaidtobe: 2B I self adjoint if T ⇤ = T . ⇠((T ⇠,⇠) R)) )8 2 I positive if self adjoint and ⇠((T ⇠,⇠) 0). 8 I normal if T ⇤T = TT ⇤.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 15 / 52 Some Operator Theory
Definition An operator T (H)issaidtobe: 2B I self adjoint if T ⇤ = T . ⇠((T ⇠,⇠) R)) )8 2 I positive if self adjoint and ⇠((T ⇠,⇠) 0). 8 I normal if T ⇤T = TT ⇤.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 15 / 52 Some Operator Theory
Definition An operator T (H)issaidtobe: 2B I self adjoint if T ⇤ = T . ⇠((T ⇠,⇠) R)) )8 2 I positive if self adjoint and ⇠((T ⇠,⇠) 0). 8 I normal if T ⇤T = TT ⇤.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 15 / 52 The Spectrum
Definition For T (H), the spectrum 2B (T ):= C : T I GL(H) . { 2 62 } Fact (T ) is nonempty compact.
Fact If (T ) then either: 2 I ⇠ = 0(T ⇠ = ⇠) 9 6 I ✏>0 ⇠( ⇠ =1 T ⇠ ⇠ <✏) 8 9 k k ^k k I (T I )(H) = H. 6
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 16 / 52 The Spectrum
Definition For T (H), the spectrum 2B (T ):= C : T I GL(H) . { 2 62 } Fact (T ) is nonempty compact.
Fact If (T ) then either: 2 I ⇠ = 0(T ⇠ = ⇠) 9 6 I ✏>0 ⇠( ⇠ =1 T ⇠ ⇠ <✏) 8 9 k k ^k k I (T I )(H) = H. 6
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 16 / 52 The Spectrum
Definition For T (H), the spectrum 2B (T ):= C : T I GL(H) . { 2 62 } Fact (T ) is nonempty compact.
Fact If (T ) then either: 2 I ⇠ = 0(T ⇠ = ⇠) 9 6 I ✏>0 ⇠( ⇠ =1 T ⇠ ⇠ <✏) 8 9 k k ^k k I (T I )(H) = H. 6
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 16 / 52 The Spectrum
Fact If T (H) is normal and (T ),thenthereexists(⇠ ), ⇠ =1s.t. 2B 2 n k nk T ⇠ ⇠ 0. k n nk! Proof. T normal T ⇠ ⇠ = T ⇠ ⇠¯ . )k k k ⇤ k
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 17 / 52 The Spectrum
Definition T (H)thespectralradius (T )ismax : (T ) . 2B | | {| | 2 } Fact > T then T I GL(H). | | k k 2 Proof. Use power series to invert T I .
Corollary (T ) T | | k k
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 18 / 52 Spectral Radius
Theorem If T (H), T 0,then T = (T ) 2B k k | | Proof. Wlog T = 1. Choose (⇠ ), ⇠ = 1, T ⇠ 1. Define k k n k nk k nk! (⇠,⌘)T := (T ⇠,⌘) positive semidefinite. We have 1 = limn(⇠n, T ⇠n)T and lim (⇠ ,⇠ ) , lim (T ⇠ , T ⇠ ) 1whence⇠ = T ⇠ + ⌘ where n n n T n n n T n n n T ⌘ 0(Cauchy-Schwarz).Butthen ⌘ 0 by optimality of norm k nk! k nk! estimate. Thus T ⇠ ⇠ 0and1 (T ). k n nk! 2
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 19 / 52 Spectral Mapping
p(z) complex -polynomial, ⇤ p(z)=(z a ) (z a )(¯z b ) (¯z b ). 1 ··· m 1 ··· n T (H), normal, 2B
p(T ):=(T a I ) (T a I )(T ⇤ b I ) (T ⇤ b I ) 1 ··· j 1 ··· k
Theorem (Spectral Mapping I) If T (H), T normal, then p( (T )) = (p(T )). 2B Proof. p(T )⇠ 0 either lim inf T ⇠ a ⇠ = 0 for some i or k nk! , k n i nk lim inf T ⇠ b ⇠ = 0 for some j. k ⇤ n j nk
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 20 / 52 Spectral Mapping
p (x) uniformly convergent on (T )tof , T = T . { n } ⇤ Theorem (Spectral Mapping II)
(f (T )⇠,⌘):=limn(pn(T )⇠,⌘) exists and f ( (T )) = (f (T )).
Proof. Uniformly over pairs ⇠,⌘ in unit ball (p (T )⇠,⌘) is Cauchy by spectral { n } mapping, so defines a (bounded) bilinear form, thus an operator f (T ). Then p (T ) f (T ) 0. k n k!
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 21 / 52 Corollary T (H),thenT 0 i↵ S(T = S S) i↵ S(T = S2 S = S ). 2B 9 ⇤ 9 ^ ⇤
Spectral Theorem
T (H), T = T , C (T ) smallest norm closed, real algebra containing 2B ⇤ R⇤ T and I . Theorem (Spectral Theorem I) There is an isometric isomorphism
C ( (T )) C ⇤(T ) R $ R
sending id (T ) to T .
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 22 / 52 Spectral Theorem
T (H), T = T , C (T ) smallest norm closed, real algebra containing 2B ⇤ R⇤ T and I . Theorem (Spectral Theorem I) There is an isometric isomorphism
C ( (T )) C ⇤(T ) R $ R
sending id (T ) to T .
Corollary T (H),thenT 0 i↵ S(T = S S) i↵ S(T = S2 S = S ). 2B 9 ⇤ 9 ^ ⇤
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 22 / 52 Spectral Theorem
T (H), T = T , C (T ) smallest norm closed, real algebra containing 2B ⇤ R⇤ T and I . Theorem (Spectral Theorem I) There is an isometric isomorphism
C ( (T )) C ⇤(T ) R $ R
sending id (T ) to T .
Corollary T (H),thenT 0 i↵ S(T = S S) i↵ S(T = S2 S = S ). 2B 9 ⇤ 9 ^ ⇤
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 22 / 52 Spectral Theorem
Corollary
Every element of a C⇤-algebra is a linear combination of four positive elements.
Corollary
In a unital C⇤-algebra the set of positive elements A+ is a complete cone
Definition x, y A , x y if y x 0. 2 + Fact If x A ,thenx x 1. 2 + k k·
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 23 / 52 Spectral Theorem
T (H), T normal, C (T )smallestC-algebra containing T , I . 2B ⇤ ⇤ Theorem (Spectral Theorem II) There is an isometric isomorphism
C( (T )) C ⇤(T ) $
sending id (T ) to T .
I How to bootstrap to abelian C⇤-algebras? I Need a complete set of “eigenvectors”.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 24 / 52 Spectral Theorem
T (H), T normal, C (T )smallestC-algebra containing T , I . 2B ⇤ ⇤ Theorem (Spectral Theorem II) There is an isometric isomorphism
C( (T )) C ⇤(T ) $
sending id (T ) to T .
I How to bootstrap to abelian C⇤-algebras? I Need a complete set of “eigenvectors”.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 24 / 52 I Let (A) be the set of states. (A) is nonempty, convex, weak* S S compact.
Fact A is positive i↵ (1) = . 2 ⇤ k k
States
A (H)unitalC-subalgebra ⇢B ⇤ Definition A is positive if a 0 (a) 0andastateif 0and (1) = 1. 2 ⇤ ) (x):=(x⇠,⇠), ⇠ = 1 is a state on (H). Also lim (x⇠ ,⇠ ), ⇠ = 1. k k B ! n n k nk
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 25 / 52 Fact A is positive i↵ (1) = . 2 ⇤ k k
States
A (H)unitalC-subalgebra ⇢B ⇤ Definition A is positive if a 0 (a) 0andastateif 0and (1) = 1. 2 ⇤ ) (x):=(x⇠,⇠), ⇠ = 1 is a state on (H). Also lim (x⇠ ,⇠ ), ⇠ = 1. k k B ! n n k nk I Let (A) be the set of states. (A) is nonempty, convex, weak* S S compact.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 25 / 52 States
A (H)unitalC-subalgebra ⇢B ⇤ Definition A is positive if a 0 (a) 0andastateif 0and (1) = 1. 2 ⇤ ) (x):=(x⇠,⇠), ⇠ = 1 is a state on (H). Also lim (x⇠ ,⇠ ), ⇠ = 1. k k B ! n n k nk I Let (A) be the set of states. (A) is nonempty, convex, weak* S S compact.
Fact A is positive i↵ (1) = . 2 ⇤ k k
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 25 / 52 States
Fact (Cauchy–Schwarz) A , 0,then 2 ⇤ 1/2 1/2 (y ⇤x) (x⇤x) (y ⇤y) . | | Corollary 0, (1) = 0,then 0. ⌘ Fact (State extension) Any state (A) extends to a state ( (H)). 2S 0 2S B Proof. Hahn–Banach
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 26 / 52 Pure States
Definition Astate (A)ispureif is an extreme in (A). Let (A) denote the 2S S P pure states.
Fact (Krein–Milman) The convex hull of (A) is (A). P S Fact a A, a =0,thereisapurestatesuchthat (a) =0 2 6 6 Proof. If H is a complex Hilbert space, then (T ⇠,⇠) = 0, ⇠ H T 0. 8 2 ) ⌘
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 27 / 52 States
Let A = `1(N). What is (A)? P I n(f ):=f (n).
Fact
In general, , X N, µ(X ):= ( X ) defines a finitely additive 2S ⇢ probability measure. Conversely for any such measure µ, f fdµ is a 7! state. R
I (`1(N)) µ(X ) 0, 1 , X N 2P $ 2{ } 8 ⇢ I (`1(N)) ultrafilters on N !! P ${ }
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 28 / 52 States
Let A = `1(N). What is (A)? P I n(f ):=f (n).
Fact
In general, , X N, µ(X ):= ( X ) defines a finitely additive 2S ⇢ probability measure. Conversely for any such measure µ, f fdµ is a 7! state. R
I (`1(N)) µ(X ) 0, 1 , X N 2P $ 2{ } 8 ⇢ I (`1(N)) ultrafilters on N !! P ${ }
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 28 / 52 States
Definition , A , is (a) (a)foralla 0. 2 ⇤ Fact 0 , (A),then = c for some c [0, 1]. 2P · 2 Proof. Wlog := (1) (0, 1). 2 1 1 = [ ]+(1 ) [(1 ) ( )]. · ·
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 29 / 52 Characters
Definition Acharacteris a unital -homomorphism : A C ⇤ !
I a 0, (a)= (x x)= (x) (x) 0, so (A). ⇤ 2S I (A). 2P I A = C(X ), X compact T , x X , (f ):=f (x)character. 2 8 2 x I Any character on C(X ) is of the form x . (Reisz representation theorem)
I May be no characters, e.g., Mn(C), n 2.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 30 / 52 Characters
Definition Acharacteris a unital -homomorphism : A C ⇤ !
I a 0, (a)= (x x)= (x) (x) 0, so (A). ⇤ 2S I (A). 2P I A = C(X ), X compact T , x X , (f ):=f (x)character. 2 8 2 x I Any character on C(X ) is of the form x . (Reisz representation theorem)
I May be no characters, e.g., Mn(C), n 2.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 30 / 52 Characters
Definition Acharacteris a unital -homomorphism : A C ⇤ !
I a 0, (a)= (x x)= (x) (x) 0, so (A). ⇤ 2S I (A). 2P I A = C(X ), X compact T , x X , (f ):=f (x)character. 2 8 2 x I Any character on C(X ) is of the form x . (Reisz representation theorem)
I May be no characters, e.g., Mn(C), n 2.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 30 / 52 Characters
Definition Acharacteris a unital -homomorphism : A C ⇤ !
I a 0, (a)= (x x)= (x) (x) 0, so (A). ⇤ 2S I (A). 2P I A = C(X ), X compact T , x X , (f ):=f (x)character. 2 8 2 x I Any character on C(X ) is of the form x . (Reisz representation theorem)
I May be no characters, e.g., Mn(C), n 2.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 30 / 52 Characters
Definition Acharacteris a unital -homomorphism : A C ⇤ !
I a 0, (a)= (x x)= (x) (x) 0, so (A). ⇤ 2S I (A). 2P I A = C(X ), X compact T , x X , (f ):=f (x)character. 2 8 2 x I Any character on C(X ) is of the form x . (Reisz representation theorem)
I May be no characters, e.g., Mn(C), n 2.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 30 / 52 Abelian C⇤-Algebras
Theorem
If A is a unital, abelian C⇤-algebra, then any pure state is a character.
Proof. (A), a, b 0, a , b 1. 2P k k k k 1/2 1/2 I x 0 (x):= (xb)= (b xb ) 0 ) b 1/2 1/2 1/2 1/2 I (x)= (xb)= (x bx ) b (x 1x ) (x) 0 b k k ) b I (A) = (x) 2P ) b · I (b)= (1) = (1) = b · I (ab)= (a) (b)
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 31 / 52 Abelian C⇤-Algebras
Theorem
If A is a unital, abelian C⇤-algebra, then any pure state is a character.
Proof. (A), a, b 0, a , b 1. 2P k k k k 1/2 1/2 I x 0 (x):= (xb)= (b xb ) 0 ) b 1/2 1/2 1/2 1/2 I (x)= (xb)= (x bx ) b (x 1x ) (x) 0 b k k ) b I (A) = (x) 2P ) b · I (b)= (1) = (1) = b · I (ab)= (a) (b)
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 31 / 52 Abelian C⇤-Algebras
Theorem
If A is a unital, abelian C⇤-algebra, then any pure state is a character.
Proof. (A), a, b 0, a , b 1. 2P k k k k 1/2 1/2 I x 0 (x):= (xb)= (b xb ) 0 ) b 1/2 1/2 1/2 1/2 I (x)= (xb)= (x bx ) b (x 1x ) (x) 0 b k k ) b I (A) = (x) 2P ) b · I (b)= (1) = (1) = b · I (ab)= (a) (b)
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 31 / 52 Abelian C⇤-Algebras
Theorem
If A is a unital, abelian C⇤-algebra, then any pure state is a character.
Proof. (A), a, b 0, a , b 1. 2P k k k k 1/2 1/2 I x 0 (x):= (xb)= (b xb ) 0 ) b 1/2 1/2 1/2 1/2 I (x)= (xb)= (x bx ) b (x 1x ) (x) 0 b k k ) b I (A) = (x) 2P ) b · I (b)= (1) = (1) = b · I (ab)= (a) (b)
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 31 / 52 Abelian C⇤-Algebras
Theorem
If A is a unital, abelian C⇤-algebra, then any pure state is a character.
Proof. (A), a, b 0, a , b 1. 2P k k k k 1/2 1/2 I x 0 (x):= (xb)= (b xb ) 0 ) b 1/2 1/2 1/2 1/2 I (x)= (xb)= (x bx ) b (x 1x ) (x) 0 b k k ) b I (A) = (x) 2P ) b · I (b)= (1) = (1) = b · I (ab)= (a) (b)
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 31 / 52 Gel’fand Spectrum
Definition
Let A be a unital, abelian C⇤-algebra. The Gel’fand spectrum is
⌃ := (A)= all characters of A A P { } ⌃ is a weak* closed subset of (A), whence is compact, Hausdor↵itself. A S Fact (Gel’fand–Mazur) ⌃ maximal proper closed ideals of A A ${ }
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 32 / 52 Gel’fand Transform
Definition :A C(⌃ ), (a)( ):= (a), Gel’fand transform. ! A Fact Easy to check that is a unital contractive -homomorphism. ⇤ Fact is an isometry.
Proof. A,abelian -algebra every a A is normal. Hence there is a state s.t. ⇤ ) 2 (a)= a . k k
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 33 / 52 Gel’fand Transform
Theorem is surjective.
Proof. Image of is a -subalgebra that separates points. The image is closed ⇤ since (a) determines the values of (a⇠,⇠), ⇠ H,whencethebilinear 8 2 form (a⇠,⌘). an isometry, thus (an) converges uniformly implies an converges uniformly and limn (an)= (limn an). Stone–Weierstrass finishes the job.
T. Sinclair (Purdue University) Intro to C⇤-algebras September 20, 2017 34 / 52 Gel’fand transform
Fact ( (a)( )= ) (a) 9 , 2 Proof.