ebook img

A characterization of orthogonal vector fields over ${\bf W^*}$-algebras of type ${\bf I_2}$ PDF

0.18 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview A characterization of orthogonal vector fields over ${\bf W^*}$-algebras of type ${\bf I_2}$

A CHARACTERIZATION OF ORTHOGONAL VECTOR FIELDS OVER W∗-ALGEBRAS OF TYPE I2 3 1 G.D.LUGOVAYAANDA.N.SHERSTNEV* 0 2 Abstract. In the paper we give a characterization of a w∗-continuous n orthogonal vector field F over an W∗-algebra N of type I2 in terms of a reductions F onthecenter of N. As anapplicationitisobtained aproof J of the assertion that an arbitrary w∗-continuous orthogonal vector field 1 overaW∗-algebraoftypeI2 isstationary. 2 ] A Introduction O . In [1],[2] Masani studied the integration with respect to orthogonal vector h t measures defined on rings of the sets. These works obtained an extension in a numerouspublicationsincontextofthe noncommutativemeasuretheory. Some m settings of problems and approaches to their solutions may be found in [3], [ [4,§31]. One of intrinsic problem in this subject is the problem of characteri- 1 zation of Hilbert-valued linear mappings of W∗-algebra preserving the orthog- v onality property (so-called orthogonal vector fields). In some sense there is a 8 standard (well known) result for the commutative W∗-algebra (see for instance 6 [4, Theorem 31.19]). 7 4 In the paper we give a characterizationof a w∗-continuous orthogonalvector 1. field F overan W∗-algebraN of type I2 in terms of reductions F on the center 0 of N. As anapplication it is obtained a proofof the assertionthat anarbitrary 3 w∗-continuousorthogonalvectorfieldoveraW∗-algebraoftypeI isstationary. 2 1 : 1. Preliminaries v i X Let A be a W∗-algebra, and Apr, Asa, Aun denote the sets of orthogonal r projections, selfadjoint, unitaries elements in A, respectively. We denote by a rp(x)therangeprojectionofx∈A+. Itistheleastprojectionofallprojections p∈Apr such that px=x. In the paper we shall mainly deal with W∗-algebras of type I . It is known that any W∗-algebra of type I can be represented in 2 2 the form N = M⊗M where M is a commutative W∗-algebra and M is the 2 2 1991 Mathematics Subject Classification. Primary46L10,46L51. Key words and phrases. W∗-algebra,orthogonal vectorfield,stationaryvector field. 1 2 G.D.LUGOVAYAANDA.N.SHERSTNEV algebra of all 2×2 matrices over C. So the elements of N are matrices in the x x form of (x )= 11 12 where x ∈M. We denote by I and 1 the units ij x x ij 21 22 (cid:18) (cid:19) ofalgebrasN andMrespectively;{ε } denotethematrixunitofalgebra ij i,j=1,2 N in the form 1 0 0 1 0 0 0 0 ε = , ε = , ε = , ε = . 11 0 0 12 0 0 21 1 0 22 0 1 (cid:18) (cid:19) (cid:18) (cid:19) (cid:18) (cid:19) (cid:18) (cid:19) It is known [5, Prop. 1.18.1] that a commutative W∗-algebra M may be realized as algebra L∞(Ω,ν) of all essentially bounded locally ν-measurable functions on a localizable measure space (Ω,ν) (i. e. Ω is direct sum of fi- nite measure spaces, see [6]). In this case, the Banach space L1(Ω,ν) is the predual of L∞(Ω,ν). Now we shall identify M with L∞(Ω,ν). In this case the characteristic functions 1, if ω ∈π, π(ω)≡χ (ω)= π ⊂Ω. π 0, if ω 6∈π, (cid:26) correspondtoprojectionsπ ∈Mpr. (Weusethesameletterπtodesignatethree objects: a projection in Mpr, a ν-measurable set in Ω and the characteristic function of this set.) Accordingly,the W∗-algebraN =M⊗M is realizedasthe algebraof2×2- 2 matrices (x ), x ∈L∞(Ω,ν) with the predual ij ij N∗ =L1(Ω,ν)⊗(M2)∗ =L1(Ω,ν,M2) (1) where L1(Ω,ν,M ) is the Banach space of all M -valued Bochner ν-integrable 2 2 functions on Ω [5, Prop. 1.22.12]. Elements of the space M on the right-hand 2 side of (1) can be viewed as density matrices. In other words, elements ϕ of N∗ are the matrices ϕ ϕ ϕ= 11 12 , ϕ ∈L1(Ω,ν), (2) ϕ ϕ ij 21 22 (cid:18) (cid:19) and the action of ϕ on an element a = (a ), a ∈ L∞(Ω,ν), is given by the ij ij equality ϕ(a)= Tr[(ϕ )(a )]dν = (ϕ a +ϕ a +ϕ a +ϕ a )dν. ij ij 11 11 12 21 21 12 22 22 Z Z Ω Ω In this case ϕ(πε )= ϕ dν, π ∈Mpr, i,j =1,2,1 ij ji Z π 1Weusethenotations oftypeaε12≡(cid:18) 00 a0 (cid:19), a∈M. ORTHOGONAL VECTOR FIELDS OVER W∗-ALGEBRAS OF TYPE I2 3 i. e. ϕ = d ϕ((·)ε ) is the Radon-Nikodym derivative of the charge π → ij dν ji ϕ(πε ), π ∈Mpr, with respect to measure ν. ji Let H be a complex Hilbert space. A bounded linear map F :A→H is said to be an orthogonal vector field (OVF) if pq =0 (p,q ∈Apr)⇒hF(p),F(q)i=0. (3) If a linear map F with the property (3) is continuous in w∗-topology on A and weak topology on H, then the map F is referred to as w∗-continuous OVF. It should be noted that a w∗-continuous OVF F is bounded, so that it is an OVF [7, Corollary 3]. A w∗-continuous OVF F is said to be stationary [3], if there exist two functionals ϕ,ψ ∈A+ such that ∗ hF(x),F(y)i=ϕ(y∗x)+ψ(xy∗), x,y ∈A. (4) Given OVF F :A→H assign a positive linear functional ̺∈A∗, ̺(x)≡hF(x),F(1)i, x∈A (5) (̺∈A+ as soon as F is w∗-continuous). The following equalities hold ∗ kF(x)k2 =̺(x2), x∈Asa, (6) 1 RehF(x),F(y)i= ̺(xy+yx), x,y ∈Asa. (7) 2 Equality (6) follows from the spectral theorem with regard to (3), and the fol- lowing computation gives (7): 1 RehF(x),F(y)i= [hF(x+y),F(x+y)−F(x−y),F(x−y)i] 4 1 1 = [̺((x+y)2)−̺((x−y)2)]= ̺(xy+yx), x,y ∈Asa. 4 2 It is easily seen that for a stationary OVF the next equality is valid ̺=ϕ+ψ. (8) 2. Orthogonal vector fields over W∗-algebras of type I2 Let N = M⊗M be a W∗-algebra of type I , H be a Hilbert space over 2 2 C and F : N → H be a H-valued OVF. By the proof of Theorem 31.6 [4], the mappings F :M→H given by the equalities ij F (a)≡F(aε ), a∈M, i,j =1,2, (9) ij ij are OVFs over M. Moreover, 2 F(x)= F (x ), x=(x )∈N. ij ij ij i,j=1 X 4 G.D.LUGOVAYAANDA.N.SHERSTNEV SincethecenterofN isisomorphictoalgebraM,wecanconsidertheorthogonal vector fields F as reductions F on the center of N. ij Proposition 1. Let F :N →H be a OVF and F are defined in (9). Then ij (i) hF (a),F (b)i=hF (a),F (b)i=0, a,b∈M, 11 22 12 21 (ii) kF (a)k2+kF (a)k2 =kF (a)k2+kF (a)k2, a∈M, 12 21 11 22 (iii) hF (a),F (b)i=hF (b∗a),F (1)i, a,b∈M, i,j,k,l=1,2, ij kl ij kl (iv) hF (π),F (1)i=hF (1),F (π)i, hF (π),F (1)i=hF (1),F (π)i, 12 11 22 21 21 11 22 12 π ∈Mpr. Proof. Let us verify the second equality in (i). In view of the spectral theorem it suffices to prove the assertion in case a = σ, b = τ are projections in Mpr. Let us first assume that στ =0. We have 0 σ 1 σ ωσ 0 0 1 τ ωτ = ω , = ω , 0 0 4 ωσ σ τ 0 4 ωτ τ (cid:18) (cid:19) ω=±1,±i (cid:18) (cid:19) (cid:18) (cid:19) ω (cid:18) (cid:19) X X (10) where in right-hand sides of (10) there are linear combinations of projections from N. It now follows in view of (3) that στ =0⇒hF (σ),F (τ)i=0. 12 21 Now we suppose that σ =τ. By the equalities 0 σ 1 0 σ i 0 −i σ = + , 0 0 2 σ 0 2 i σ 0 (cid:18) (cid:19) (cid:18) (cid:19) (cid:18) (cid:19) 0 0 1 0 σ i 0 i σ = + σ 0 2 σ 0 2 −i σ 0 (cid:18) (cid:19) (cid:18) (cid:19) (cid:18) (cid:19) we have 0 σ 0 0 4hF (σ),F (σ)i=4 F ,F 12 21 0 0 σ 0 (cid:28) (cid:18) (cid:19) (cid:18) (cid:19)(cid:29) 2 0 σ 0 −i σ 0 i σ = F + F ,F σ 0 i σ 0 −i σ 0 (cid:13) (cid:18) (cid:19)(cid:13) (cid:28) (cid:18) (cid:19) (cid:18) (cid:19)(cid:29) (cid:13)(cid:13) 0 σ (cid:13)(cid:13) 0 i σ 0 −i σ 0 σ −i(cid:13) F (cid:13),F +i F ,F σ 0 −i σ 0 i σ 0 σ 0 (cid:28) (cid:18) (cid:19) (cid:18) (cid:19)(cid:29) (cid:28) (cid:18) (cid:19) (cid:18) (cid:19)(cid:29) 2 2 0 σ 0 −i σ = F − F σ 0 i σ 0 (cid:13) (cid:18) (cid:19)(cid:13) (cid:13) (cid:18) (cid:19)(cid:13) (cid:13)(cid:13) 0 σ(cid:13)(cid:13) (cid:13)(cid:13) 0 −i σ (cid:13)(cid:13) 0 −i σ 0 σ +i(cid:13) F (cid:13) ,F(cid:13) (cid:13)+ F ,F . σ 0 i σ 0 i σ 0 σ 0 (cid:26)(cid:28) (cid:18) (cid:19) (cid:18) (cid:19)(cid:29) (cid:28) (cid:18) (cid:19) (cid:18) (cid:19)(cid:29)(cid:27) ORTHOGONAL VECTOR FIELDS OVER W∗-ALGEBRAS OF TYPE I2 5 In addition (see (6),(7)), 2 2 0 −i σ 0 −i σ F = F ,F(I) i σ 0 i σ 0 (cid:13) (cid:18) (cid:19)(cid:13) * (cid:18) (cid:19) ! + (cid:13) (cid:13) (cid:13) (cid:13) σ 0 0 σ 2 (cid:13) (cid:13) = F ,F(I) = F , 0 σ σ 0 (cid:28) (cid:18) (cid:19) (cid:29) (cid:13) (cid:18) (cid:19)(cid:13) (cid:13) (cid:13) The expression in braces vanishes: (cid:13) (cid:13) (cid:13) (cid:13) 0 σ 0 −i σ {...}=2Re F ,F σ 0 i σ 0 (cid:28) (cid:18) (cid:19) (cid:18) (cid:19)(cid:29) 0 σ 0 −i σ 0 −i σ 0 σ =̺ + =0, σ 0 i σ 0 i σ 0 σ 0 (cid:20)(cid:18) (cid:19)(cid:18) (cid:19) (cid:18) (cid:19)(cid:18) (cid:19)(cid:21) and the second equality in (i) is established. The equality hF (a),F (b)i = 11 22 0 (a,b∈M) is obvious in view of the spectral theorem. Equality (ii) follows from computation (with regard to (i)) kF (a)k2+kF (a)k2 =kF (a)k2+kF (a∗)k2 =kF (a)+F (a∗)k2 12 21 12 21 12 21 2 0 a 0 a 0 a = F ,F = F ,F(I) a∗ 0 a∗ 0 a∗ 0 (cid:28) (cid:18) (cid:19) (cid:18) (cid:19)(cid:29) * (cid:18) (cid:19) ! + =hF (a∗a),F (1)i+hF (a∗a),F (1)i=kF (a)k2+kF (a)k2. 11 11 22 22 11 22 Weusedhere(forG=F )thefollowingpropertyoftheOVFovercommutative 21 W∗-algebra M (it follows easily from the spectral theorem): kG(a)k2 =kG(a∗)k2, a∈M. Let us establish (iii). We first note that representations (10) give στ =0 (σ,τ ∈Mpr)⇒hF (σ),F (τ)i=hF (σ),F (τ)i=0, i=1,2. (11) ii 12 ii 21 InviewofthespectraltheoremandboundednessoflinearmapsF itissufficient ij to establish (iii) for elements a,b in the form a= µ π , b= λ π (λ ,µ ∈C, π ∈Mpr, π π =0 (t6=s)). t t t t t t t t s t t X X (Here the sums are finite and π = 1.) Taking into account (i) and (11), we t t have P hF (a),F (b)i= µ λ hF (π ),F (π )i= µ λ hF (π ),F (1)i ij kl t s ij t kl s t t ij t kl t,s t X X =hF ( µ λ π ),F (1)i=hF (b∗a),F (1)i. ij t t t kl ij kl t X 6 G.D.LUGOVAYAANDA.N.SHERSTNEV (iv). We show first that hF (aπ)+F (a∗π),F (π)i=hF (π),F (aπ)+F (a∗π)i, a∈M, π ∈Mpr. 12 21 11 22 12 21 (12) 0 aπ Define f =F12(aπ)+F21(a∗π)=F a∗π 0 . By virtue of (iii) and (7) (cid:18) (cid:19) hf,F (π)i=hf,F(I)i−[hf,F (π)i+hF (π),fi]+hF (π),fi 11 22 22 22 0 aπ 0 0 =hf,F(I)i−2Re F a∗π 0 ,F 0 π +hF22(π),fi (cid:28) (cid:18) (cid:19) (cid:18) (cid:19)(cid:29) =̺(aπε +a∗πε )−̺((aπε +a∗πε )·πε +πε ·(aπε +a∗πε )) 12 21 12 21 22 22 12 21 +hF (π),fi=hF (π),fi, 22 22 and (12) is established. By setting a=1 and a=i1 in (12), we have hF (π)+F (π),F (π)i=hF (π),F (π)+F (π)i, (13) 12 21 11 22 12 21 ihF (π)−F (π),F (π)i=−ihF (π),F (π)−F (π)i. (14) 12 21 11 22 12 21 Multiplying both sides of (14) by i and taking into account (13), we have (iv). This proves the proposition. (cid:3) Proposition 2. Let F (i = 1,2) be OVFs satisfying conditions (iii) and (iv) ij of Prop. 1. Then equality (12) holds. Proof. Because the linear mappings F are bounded, it is sufficient to prove ij (12) for finite sums in the form a = λ π where λ ∈ C,π ∈ Mpr, π π = s s s s s t s 0 (s6=t), πs =1. In view of (iii) anPd (iv) we have s P hF (aπ)+F (a∗π),F (π)i= [λ hF (ππ ),F (1)i+λ hF (ππ ),F (1)i] 12 21 11 s 12 s 11 s 21 s 11 s X = [λ hF (1),F (ππ )i+λ hF (1),F (ππ )i] s 22 21 s s 22 12 s s X =hF (π),F ( λ ππ )i+hF (π),F ( λ ππ )i 22 21 s s 22 12 s s s s X X =hF (π),F (a∗π)+F (aπ)i. 22 21 12 The proposition follows. (cid:3) Theorem 3. Let N = M⊗M be a W∗-algebra of type I , F : M → H be 2 2 ij H-valued OVFs (w∗-continuous OVFs) with the properties (i)−(iv) of Prop. 1. Then the linear mapping F :N →H defined by the equalities F(aε )≡F (a), a∈M, i,j =1,2, ij ij is the OVF (respectively, w∗-continuous OVF). ORTHOGONAL VECTOR FIELDS OVER W∗-ALGEBRAS OF TYPE I2 7 Proof. Since F is w∗–continuous if and only if F are w∗-continuous, it is suf- ij ficient to consider the case when F be OVFs. We have only to verify the ij equality pq =0 ⇒ hF(p),F(q)i=0, p,q∈Npr. We use the canonical representation projections in N. Specifically, every pro- jection r ∈Npr can be expressed in the form [8, Lemma 1] r =π ⊕π +p(a,v,π), (15) 1 2 where π 0 π ⊕π ≡ 1 , π ∈Mpr, 1 2 0 π i 2 (cid:18) (cid:19) aπ vπ(a(1−a))1/2 p(a,v,π)≡ , π ∈Mpr,v ∈Mun. v∗π(a(1−a))1/2 (1−a)π (cid:18) (cid:19) In addition, π ≤1−π (i=1,2), a∈M, 0≤a≤1,rp a(1−a))=1. Now we i present p,q ∈Npr in the form of (15): p=τ ⊕τ +p(a,v,τ ), q =σ ⊕σ +p(b,w,σ ), 1 2 3 1 2 3 where τ ,σ ∈ Mpr, 0 ≤ a,b ≤ 1, v,w ∈ Mun. From pq = 0 it follows [8, i i Lemma 2] that τ σ =τ τ =σ σ =τ σ =τ σ =0, wπ =−vπ, bπ =(1−a)π, (16) i i i 3 i 3 3 i i 3 where i=1,2, π ≡σ τ . Denoting for brevity κ(x)≡(x(1−x))1/2 we have 3 3 F(p)=F (τ +aτ )+F (vκ(a)τ )+F (v∗κ(a)τ )+F (τ +(1−a)τ ), 11 1 3 12 3 21 3 22 2 3 F(q)=F (σ +bσ )+F (wκ(b)σ )+F (w∗κ(b)σ )+F (σ +(1−b)σ ). 11 1 3 12 3 21 3 22 2 3 Takingthe scalarproductof vectorsF(p) and F(q) with regardto (i), (iii), (iv) in Prop. 1 and equalities (16) we have hF(p),F(q)i=hF (aπ),F ((1−a)π)i−hF (aπ),F (vκ(a)π)i 11 11 11 12 −hF (aπ),F (v∗κ(a)π)i+hF (vκ(a)π),F ((1−a)π)i 11 21 12 11 −kF (vκ(a)π)k2+hF (vκ(a)π),F (aπ)i 12 12 22 +hF (v∗κ(a)π),F ((1−a)π)i−kF (v∗κ(a)π)k2 21 11 21 +hF (v∗κ(a)π),F (aπ),i−hF ((1−a)π),F (vκ(a)π)i 21 22 22 12 −hF ((1−a)π),F (v∗κ(a)π)i+hF ((1−a)π),F (aπ)i. 22 21 22 22 ByProp. 1(ii),thefirstandthelastsummandsontherighthand-sideofobtained equality are mutually annihilated with the fifth and the eights ones. Grouping 8 G.D.LUGOVAYAANDA.N.SHERSTNEV the remained summands and setting c =vaκ(a), d=v(1−a)κ(a) we have by (12) hF(p),F(q)i=[hF (vκ(a)π)+F (v∗κ(a)π),F (aπ)i 12 21 22 −hF (aπ),F (vκ(a)π)+F (v∗κ(a)π)i] 11 12 21 +[hF (vκ(a)π)+F (v∗κ(a)π),F ((1−a)π)i 12 21 11 −hF ((1−a)π),F (vκ(a)π)+F (v∗κ(a)π)i] 22 12 21 =[hF (cπ)+F (c∗π),F (π)i−hF (π),F (cπ)+F (c∗π)i] 12 21 22 11 12 21 +[hF (dπ)+F (d∗π),F (π)i−hF (π),F (dπ)+F (d∗π)i]=0. 12 21 11 22 12 21 The proof is complete. (cid:3) 3. The stationarity of OVFs over W∗-algebras of type I2 As an application of Theorem 3 we will show that every w∗-continuous OVF overanW∗-algebraoftype I is stationary. Let N =M⊗M be a W∗-algebra 2 2 oftypeI ,F :N →H beaw∗-continuousOVFandthefunctional̺begivenby 2 (5). Letusobservesomepropertiesofreductions(10). Withtheabovenotations of Section 1 we have ̺(πε )=hF(πε ),F(I)i=hF(πε ),F(ε )i=hF (π),F (1)i, (17) ii ii ii ii ii ii where i=1,2, π ∈Mpr. Therefore, d ρ = hF (·),F (1)i, i=1,2. (18) ii dν ii ii Similarly, d ρ = hF (·),F (1)+F (1)i, i,j =1,2 (i6=j). (19) ij dν ji 11 22 In addition there are defined functions r ∈L1(Ω,ν)+ such that ij d hF (π),F (1)i= r dν, r = hF (·),F (1)i, i,j =1,2. (20) ij ij ij ij dν ij ij Z π In this connection r =̺ (i=1,2), r +r =̺ +̺ a. e. (21) ii ii 12 21 11 22 (The second equality in (21) follows from Prop. 1(ii).) If F is stationary there exist (see (4)) ϕ,ψ ∈N+ such that ∗ hF(x),F(y)i=ϕ(y∗x)+ψ(xy∗), x,y ∈N. (22) ORTHOGONAL VECTOR FIELDS OVER W∗-ALGEBRAS OF TYPE I2 9 In this case (8) is satisfied. Let (ϕ ) be the matrix of ϕ (see (2)). We have ij ϕ dν =ϕ(πε )=ϕ(ε (πε )ε )=hF((πε )ε ),F(ε )i 21 12 11 12 22 12 22 11 Z π =hF(πε ),F(ε )i=hF (π),F (1)i, π ∈Mpr. 12 11 12 11 Thus, d d ϕ = hF (·),F (1)i, ϕ =ϕ = hF (·),F (1)i. 21 12 11 12 21 21 22 dν dν (The equality ϕ =ϕ follows from the positivity of ϕ.) Further, 12 21 (ϕ −ϕ )dν =ϕ(πε −πε )=ϕ(πε ε∗ )−ϕ(πε∗ ε ) 11 22 11 22 12 12 12 12 Z π =ϕ(πε ε∗ )+ψ(πε∗ ε )−̺(πε∗ ε ) 12 12 12 12 12 12 =hF(πε∗ ),F(πε∗ )i−̺(πε ) 12 12 22 =kF(πε )k2−̺(πε )=hF (π),F (1)i− ̺ dν 21 22 21 21 22 Z π = (r −̺ )dν. 21 22 Z π (The function r is defined by (20).) Putting φ≡ϕ we have 21 11 φ ϕ ̺ −φ ̺ −ϕ ϕ= 12 , ψ = 11 12 12 . (23) ϕ φ+̺ −r ̺ −ϕ r −φ 12 22 21 12 12 21 (cid:18) (cid:19) (cid:18) (cid:19) It should be noted that all elements of matrices ϕ and ψ, with the exception of φ, are defined as the Radon-Nikodym derivatives associated with the fields F . ij The function φ∈L1(Ω,ν)+ necessarily satisfies inequalities max{0,r (ω)−̺ (ω)}≤φ(ω)≤min{̺ (ω),r (ω)} a. e., (24) 21 22 11 21 |ϕ (ω)|≤[φ(ω)(φ(ω)+̺ (ω)−r (ω))]1/2 a. e., (25) 12 22 21 |̺ (ω)−ϕ (ω)|≤[(ρ (ω)−φ(ω))(r (ω)−φ(ω))]1/2 a. e. (26) 12 12 11 21 Theorem 4. Let N = M⊗M be a W∗-algebra of type I . Then every w∗- 2 2 continuous OVF F :N →H is stationary. The following lemma gives conditionally ahead of time proof of the theorem. Lemma 5. If the function φ ∈ L1(Ω,ν) satisfies inequalities (24) – (26), then the equality (22) holds when ϕ and ψ are defined by matrices (23). 10 G.D.LUGOVAYAANDA.N.SHERSTNEV Proof. As F is linear and w∗-continuous (and by Prop. 3(iii)) we must verify only that the following equalities hold hF(πε ),F(ε )i=ϕ(ε (πε ))+ψ((πε )ε ), π ∈Mpr, i,j,k,l∈{1,2}. ij kl lk ij ij lk They are easily verified directly. For example, in view of (17),(8),(9) and by Prop. 1(iv) we have hF(πε ),F(ε )i= ̺ dν = (ϕ +ψ )dν =ϕ(πε )+ψ(πε ) ii ii ii ii ii ii ii Z Z π π =ϕ(ε (πε ))+ψ((πε )ε ), i=1,2, ii ii ii ii hF(πε ),F(ε )i=hF (π),F (1)i=hF (1),F (π)i 11 12 11 12 11 12 =hF (π),F (1)i= ϕ dν =ϕ(πε ) 21 22 12 21 Z π =ϕ((πε )ε )+ψ(ε (πε )). 21 11 11 21 Similarly we can verify remaining equalities, and the lemma follows. (cid:3) Lemma 6. Any OVF over the factor of type I is stationary. 2 Proof. Let F : M →H be a OVF. Without loss of generality, we may assume 2 that kF(I)k=1. Let us denote for brevity F =F(ε ). By virtue of standard ij ij argumentsconnectedwithunitaryinvariance,itissufficienttoconsiderthecase when the matrix ̺=(̺ ) (here it is scalar) is diagonal: ij ̺ 0 ̺= 11 , ̺ +̺ =1, ̺ ≥0. 0 ̺ 11 22 ii 22 (cid:18) (cid:19) Also note that r =kF k2, r +r =1. ij ij 12 21 Case 1: the rank of matrix ̺ equals 1. Let̺ =1,̺ =0(the samearguments 11 22 can be used if ̺ =0,̺ =1). By the equality ̺ =0 it follows that {F } is 11 22 12 ij theorthogonalsystemofvectorsinH. Nowweseethatequality(22)issatisfied with r 0 r 0 ϕ= 21 , ψ = 12 0 0 0 0 (cid:18) (cid:19) (cid:18) (cid:19)

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.