Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Operator Spaces Alonso Delf´ın Functional Analysis Seminar University of Oregon November 19, 2020 1/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Outline 1 Preliminaries. 2 Operator Spaces 3 Column and Row Hilbert space. 4 Generalization of Hilbert Modules 2/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Operator Algebras. We denote by L(E, F) the bounded linear operators between Banach spaces E and F. This is a Banach space (algebra if E = F) with norm kak := sup ka(ξ)kF kξkE=1 If H is a Hilbert space, any a 2 L(H) := L(H, H) has an adjoint a∗ 2 L(H) characterized by ha(ξ), ηi = hξ, a∗(η)i. If H is a Hilbert space, C∗-algebra A is a norm closed selfadjoint subalgebra of L(H). If (X, µ) is a measure space and p 2 [1, ), an Lp-operator algebra A is a norm closed subalgebra of L(Lp(X, µ)). 1 3/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Definition and Examples Definition Let H be a Hilbert space. An operator space E is a closed subspace of L(H). Example Any C∗-algebra A is an operator space. 4/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Definition and Examples Definition Let H be a Hilbert space. An operator space E is a closed subspace of L(H). Example Let H1 and H2 be Hilbert spaces. Then L(H1, H2) is regarded as an operator space by identifying L(H1, H2) in L(H1 ⊕ H2) by 0 0 L(H , H ) 3 a 7! 2 L(H ⊕ H ) 1 2 a 0 1 2 5/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Definition and Examples Definition Let H be a Hilbert space. An operator space E is a closed subspace of L(H). Example Any Banach space E is an operator space. Indeed, it's well known that BE∗ is a compact space when equipped with the weak-∗ topology and E is identified with a subspace of C(BE∗ ) via the isometric mapping ξ 7! ξb, where ξb(') = '(ξ) ∗ for any ' 2 BE∗ . Since C(BE∗ ) is a C -algebra, it follows that E is an operator space. 6/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Matrix Norms Let E be an operator space and n 2 Z>0. Then, Mn(E), the space of n × n matrices with entries in E, is a subspace of Mn(L(H)). Thus, Mn(E) has a natural norm k · kn, which comes from the n n 2 identification of Mn(L(H)) as L(H ), where H is the ` direct sum of H with it self. More precisely, if ξ := (ξ j,k) 2 Mn(E) 8 = 9 0 211 2 <> n n => k(ξ j,k)kn := sup @ ∑ ∑ ξ j,khk A : h := (hk) 2 BHn :> j=1 k=1 ;> 7/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Completely bounded maps. The main difference between the category of Banach spaces and that of operator spaces is in the morphisms. Let E and F be operator spaces and u : E ! F a linear map. For each n 2 Z>0, u induces a linear map un : Mn(E) ! Mn(F) in the obvious way un((ξ j,k)) := (u(ξ j,k)). Further, we set kunk := supfkun(ξ)kn : ξ 2 Mn(E), kξkn = 1g. We say that u is completely bounded (c.b.) if kukcb := sup kunk < . n2Z>0 1 8/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules We denote by CB(E, F) ⊂ L(E, F) to the set of all c.b maps from E to F. Definition Let E, F be operator spaces and u 2 CB(E, F). 1 If kukcb ≤ 1 we say u is completely contractive. 2 If each un is an isometry, we say u is a complete isometry. 3 We say E and F are completely isomorphic if u is an isomorphism with u−1 2 CB(F, E). 4 We say E and F are completely isometrically isomorphic if u is a complete isomorphism that's also a complete isometry. 9/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Proposition Let E, F and G be operator spaces and u 2 CB(E, F), v 2 CB(F, G), then vu 2 CB(E, G) and kvukcb ≤ kvkcbkukcb. Proposition Let E, F be operator spaces and u 2 CB(E, F) a rank one operator. That is, u(ξ) = '(ξ)η for ' 2 E∗ and η 2 F. Then, kukcb = kuk. 10/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Ruan's Axioms Proposition Let E ⊂ L(H) be an operator space and k · kn the norms on Mn(E) defined above. For ξ := (ξ j,k) 2 Mn(E) we have (R1) If α := (α j,k), β := (β j,k) 2 Mn := Mn(C) kαξβkn ≤ kαkkξknkβk. (R2) If η := (η j,k) 2 Mm(E), then ξ 0 = m´axfkξk , kηk g η n m 0 n+m 11/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Ruan's Theorem Theorem (Ruan 1987) Suppose that E is a vector space, and that for each n 2 Z>0 we are given a norm k · kn on Mn(E) satisfying conditions (R1) and (R2) above. Then E is completely isometrically isomorphic to a subspace of L(H), for some Hilbert space H. 12/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Proof. Let U be the collection of all completely contractive maps u : E ! Mn for some n 2 Z≥0. For each u 2 U, we define nu 2 Z≥0 to be so that Mnu is the co-domain of u. Then, sup M M := Mnu u2U is a C∗-algebra and hence an operator space. We define v : E ! M by v(ξ) = (u(ξ))u2U . One checks v is a complete contraction. Furthermore, if ξ 2 Mn(E), by Hahn-Banach there is a complete contraction u : E ! Mn(E) such that kun(ξ)k = kξnk. Now consider the projection pu : M ! Mnu and notice that kvn(ξ)k = k(v(ξ j,k))k ≥ k(pu(u(ξ j,k)))k = kun(ξ)k = kξnk. Thus, vn is an isometry and therefore v is a complete isometry. 13/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Column space Let H be any Hilbert space. There are several (completely isometrically isomorphic) ways of giving H a canonical operator space structure which we call the column Hilbert space and denote by Hc. Informally, one should think of Hc as a \column in L(H)". We now give 3 equivalent descriptions of Hc for a general Hilbert space H. 14/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Definition 1 Identify H with L(C, H) by regarding each h 2 H as a map th : C !H defined by th(λ) := λh. Notice that the operator ∗ ∗ th : H! C is such that th( f ) = h f , hi and therefore ∗ (tht f )(1) = h f , hi c Using this we notice that the induced norm in Mn(H ) takes a nice form: 1=2 1=2 n ! n ! ∗ k(th )k = t th = hhi k, hi,ji j,k ∑ hi,j i,k ∑ , i=1 j,k i=1 j,k where we've used the C∗-identity. 15/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Definition 2 Fix a unit vector f 2 H. Look at the rank one operators c θh, f (y) := hy, f ih and identify H with fθh, f : h 2 Hg ⊂ L(H). This gives an operator structure which is independent (up to complete isometry) of the unit vector f chosen to begin with. Further, such structure coincides with the above one: 1=2 1=2 1=2 n ! n ! n ! k(θ ) k = θ θ = hh , h iθ = hh , h i h j,k, f j,k ∑ f ,hh,j h j,k, f ∑ i,k i,j f , f ∑ i,k i,j i=1 j,k i=1 j,k i=1 j,k where we've used again the C∗-identity and that f has norm 1. 16/23 Alonso Delf´ın Operator Spaces Preliminaries. Operator Spaces Generalization of Hilbert Column and Row Hilbert space. Modules Definition 3 Fix an orthonormal basis for H and regard elements in L(H) as infinite matrices with respect to this basis. Then let Hc consist of all the matrices in L(H) that are zero except on the first column. That way, if H = `2 we can describe the column space as c 2 H = spanfδ j,1 : j 2 Z>0g ⊂ L(` ) which is of course isometric to `2.