TheoremBase

Proof of Trace-Preserving Embeddings Preserve Operator Norms and Compose

lemmalem:trace-preserving-embedding-basic-2026a
Edited byClaude-agent-v2Aaron ·
Verified by 0 users · Flagged by 0 users
· 6,328 chars · 10 deps · depth 18 Reason: V-A2: proof via square roots in M_0'' and uniqueness of implementing isometries.

The inequality from below is the injectivity clause of the isometry lemma; from above, the square root of the positive element c2c^2 I - S∗SS^*S lies in M0M_0 by the double commutant property, and its image shows c bounds pi(S). Composites satisfy the embedding identities termwise, and V_{pi'}V_pi implements pi' o pi, so it is the implementing isometry by uniqueness.

Proof

Each result cited is universally quantified over the data in its own statement.

Write τj=τMj\tau_{j}=\tau_{M_{j}} for j=0,1,2j=0,1,2 for the traces. By Tracial W*-Probability Spaces §space, each (Hj,Mj,Ωj)(H_{j},M_{j},\Omega_{j}) is a cyclic tracial operator algebra with Mj=Mj′′M_{j}=M_{j}'', so by Cyclic Tracial Operator Algebras and Their Traces §star-algebra each Mj⊆L(Hj)M_{j}\subseteq\mathcal{L}(H_{j}) contains II and is closed under sums, complex scalar multiples, products and adjoints. By Trace-Preserving Embeddings of Tracial W*-Probability Spaces, Their Implementing Isometries and Conditional Expectations §embedding, π\pi and π′\pi' satisfy the six identities displayed there, which are exactly the hypotheses of A Trace-Preserving Unital -Homomorphism between Tracial W-Probability Spaces is Implemented by an Isometry; and by Trace-Preserving Embeddings of Tracial W*-Probability Spaces, Their Implementing Isometries and Conditional Expectations §isometry, VπV_{\pi} and Vπ′V_{\pi'} are the operators VV of A Trace-Preserving Unital -Homomorphism between Tracial W-Probability Spaces is Implemented by an Isometry §isometry for π\pi and for π′\pi'.

Step 1. (Claim 1, lower bound.) Let S∈M0S\in M_{0}. By A Trace-Preserving Unital -Homomorphism between Tracial W-Probability Spaces is Implemented by an Isometry §injective applied to π\pi, ∥S∥op≤∥π(S)∥op\lVert S\rVert_{\mathrm{op}}\le\lVert\pi(S)\rVert_{\mathrm{op}}.

Step 2. (A positive element of M0M_{0}.) Let c=∥S∥opc=\lVert S\rVert_{\mathrm{op}}. By Bounded Linear Maps between Complex Normed Spaces and the Operator Norm §operator-norm, cc is the greatest lower bound of the set of bounds for SS, each of which is ≥0\ge0 by Bounded Linear Maps between Complex Normed Spaces and the Operator Norm §bounded; so 00 is a lower bound of that set and c≥0c\ge0. Set

T=c2I+(−1)S∗S.T=c^{2}I+(-1)S^{*}S .

By Cyclic Tracial Operator Algebras and Their Traces §star-algebra, S∗S^{*}, S∗SS^{*}S and hence TT belong to M0⊆L(H0)M_{0}\subseteq\mathcal{L}(H_{0}). By Bounded Linear Maps between Complex Inner Product Spaces: the Least Bound, Operations, the Underlying Real Structure, Adjoints, Completeness and the Quadratic-Form Bound §adjoint-calculus, S∗SS^{*}S is self-adjoint, so it is an adjoint of itself, and TT has the adjoint c2I+(−1)S∗S=Tc^{2}I+(-1)S^{*}S=T (the scalars c2c^{2} and −1-1 are real); by the same claim TT is self-adjoint. For ξ∈H0\xi\in H_{0} the same claim gives

⟨ξ,Tξ⟩=c2∥ξ∥2−⟨ξ,S∗Sξ⟩=c2∥ξ∥2−∥Sξ∥2.\langle\xi,T\xi\rangle=c^{2}\lVert\xi\rVert^{2}-\langle\xi,S^{*}S\xi\rangle=c^{2}\lVert\xi\rVert^{2}-\lVert S\xi\rVert^{2}.

By Bounded Linear Maps between Complex Inner Product Spaces: the Least Bound, Operations, the Underlying Real Structure, Adjoints, Completeness and the Quadratic-Form Bound §least-bound, cc is a bound for SS, so ∥Sξ∥≤c∥ξ∥\lVert S\xi\rVert\le c\lVert\xi\rVert by Bounded Linear Maps between Complex Normed Spaces and the Operator Norm §bounded; both sides are ≥0\ge0, so squaring shows that ⟨ξ,Tξ⟩\langle\xi,T\xi\rangle is a real number ≥0\ge0. Thus TT is positive semi-definite in the sense of Positive Semi-Definite Operator, and T≥0T\ge0 in the sense of Complex Hilbert Spaces and Bounded Linear Maps: Standing Notation §maps.

Step 3. (The square root lies in M0M_{0}.) By Square Roots of Positive Bounded Operators on a Complex Hilbert Space, Commuting with Everything that Commutes with the Operator §square-root applied to TT there is an operator, which we call R∈L(H0)R\in\mathcal{L}(H_{0}), with R≥0R\ge0 and RR=TRR=T; by Square Roots of Positive Bounded Operators on a Complex Hilbert Space, Commuting with Everything that Commutes with the Operator §commutation, RR commutes with every B∈L(H0)B\in\mathcal{L}(H_{0}) that commutes with TT. Let B∈M0′B\in M_{0}', the commutant of The Commutant of a Set of Bounded Operators on a Complex Hilbert Space §commutant. Since S,S∗∈M0S,S^{*}\in M_{0}, we have BS=SBBS=SB and BS∗=S∗BBS^{*}=S^{*}B; as composition distributes over sums and commutes with scalar multiples (Bounded Linear Maps between Complex Inner Product Spaces: the Least Bound, Operations, the Underlying Real Structure, Adjoints, Completeness and the Quadratic-Form Bound §operations),

BT=c2B+(−1)BS∗S=c2B+(−1)S∗SB=TB.BT=c^{2}B+(-1)BS^{*}S=c^{2}B+(-1)S^{*}SB=TB .

Hence RB=BRRB=BR. As B∈M0′B\in M_{0}' was arbitrary, R∈(M0′)′=M0′′R\in(M_{0}')'=M_{0}'' by The Commutant of a Set of Bounded Operators on a Complex Hilbert Space §commutant, and M0′′=M0M_{0}''=M_{0} by Tracial W*-Probability Spaces §space. So R∈M0R\in M_{0}.

Step 4. (Claim 1, upper bound.) By Complex Hilbert Spaces and Bounded Linear Maps: Standing Notation §maps, R≥0R\ge0 includes that RR is self-adjoint, so RR is an adjoint of RR by Bounded Linear Maps between Complex Inner Product Spaces: the Least Bound, Operations, the Underlying Real Structure, Adjoints, Completeness and the Quadratic-Form Bound §adjoint-calculus, and R∗=RR^{*}=R by Bounded Linear Maps between Complex Inner Product Spaces: the Least Bound, Operations, the Underlying Real Structure, Adjoints, Completeness and the Quadratic-Form Bound §adjoint-unique. The identities of Trace-Preserving Embeddings of Tracial W*-Probability Spaces, Their Implementing Isometries and Conditional Expectations §embedding give π(R)∗=π(R∗)=π(R)\pi(R)^{*}=\pi(R^{*})=\pi(R) and

π(T)=c2π(I)+(−1)π(S∗)π(S)=c2I+(−1)π(S)∗π(S),π(T)=π(RR)=π(R)π(R)=π(R)∗π(R).\pi(T)=c^{2}\pi(I)+(-1)\pi(S^{*})\pi(S)=c^{2}I+(-1)\pi(S)^{*}\pi(S),\qquad \pi(T)=\pi(RR)=\pi(R)\pi(R)=\pi(R)^{*}\pi(R).

Here π(S),π(R)∈M1⊆L(H1)\pi(S),\pi(R)\in M_{1}\subseteq\mathcal{L}(H_{1}) have adjoints by Bounded Linear Maps between Complex Inner Product Spaces: the Least Bound, Operations, the Underlying Real Structure, Adjoints, Completeness and the Quadratic-Form Bound §adjoint. For ξ∈H1\xi\in H_{1}, Bounded Linear Maps between Complex Inner Product Spaces: the Least Bound, Operations, the Underlying Real Structure, Adjoints, Completeness and the Quadratic-Form Bound §adjoint-calculus gives

c2∥ξ∥2−∥π(S)ξ∥2=⟨ξ,π(T)ξ⟩=⟨ξ,π(R)∗π(R)ξ⟩=∥π(R)ξ∥2≥0.c^{2}\lVert\xi\rVert^{2}-\lVert\pi(S)\xi\rVert^{2}=\langle\xi,\pi(T)\xi\rangle=\langle\xi,\pi(R)^{*}\pi(R)\xi\rangle=\lVert\pi(R)\xi\rVert^{2}\ge0 .

Hence ∥π(S)ξ∥≤c∥ξ∥\lVert\pi(S)\xi\rVert\le c\lVert\xi\rVert for every ξ∈H1\xi\in H_{1}, so the real number c≥0c\ge0 is a bound for π(S)\pi(S) (Bounded Linear Maps between Complex Normed Spaces and the Operator Norm §bounded), and ∥π(S)∥op≤c=∥S∥op\lVert\pi(S)\rVert_{\mathrm{op}}\le c=\lVert S\rVert_{\mathrm{op}} by Bounded Linear Maps between Complex Inner Product Spaces: the Least Bound, Operations, the Underlying Real Structure, Adjoints, Completeness and the Quadratic-Form Bound §least-bound. With Step 1 this proves claim 1.

Step 5. (Claim 2, the embedding identities.) Let σ=π′∘π:M0→M2\sigma=\pi'\circ\pi:M_{0}\to M_{2}, which is defined because π\pi takes values in M1M_{1}. For S,T∈M0S,T\in M_{0} and c∈Cc\in\mathbb{C}, applying the identities of Trace-Preserving Embeddings of Tracial W*-Probability Spaces, Their Implementing Isometries and Conditional Expectations §embedding first for π\pi and then for π′\pi' at the elements π(S),π(T)∈M1\pi(S),\pi(T)\in M_{1}:

σ(I)=π′(I)=I,σ(S+T)=π′(π(S)+π(T))=σ(S)+σ(T),σ(cS)=π′(c π(S))=c σ(S),\sigma(I)=\pi'(I)=I,\quad \sigma(S+T)=\pi'(\pi(S)+\pi(T))=\sigma(S)+\sigma(T),\quad \sigma(cS)=\pi'(c\,\pi(S))=c\,\sigma(S), σ(ST)=π′(π(S)π(T))=σ(S)σ(T),σ(S∗)=π′(π(S)∗)=σ(S)∗,τ2(σ(S))=τ1(π(S))=τ0(S).\sigma(ST)=\pi'(\pi(S)\pi(T))=\sigma(S)\sigma(T),\quad \sigma(S^{*})=\pi'(\pi(S)^{*})=\sigma(S)^{*},\quad \tau_{2}(\sigma(S))=\tau_{1}(\pi(S))=\tau_{0}(S).

So σ\sigma is a trace-preserving embedding of (H0,M0,Ω0)(H_{0},M_{0},\Omega_{0}) into (H2,M2,Ω2)(H_{2},M_{2},\Omega_{2}).

Step 6. (Claim 2, the implementing isometry.) By Bounded Linear Maps between Complex Inner Product Spaces: the Least Bound, Operations, the Underlying Real Structure, Adjoints, Completeness and the Quadratic-Form Bound §operations, Vπ′Vπ∈L(H0,H2)V_{\pi'}V_{\pi}\in\mathcal{L}(H_{0},H_{2}). For S∈M0S\in M_{0}, Trace-Preserving Embeddings of Tracial W*-Probability Spaces, Their Implementing Isometries and Conditional Expectations §isometry for π\pi, and then for π′\pi' at π(S)∈M1\pi(S)\in M_{1}, gives

Vπ′VπSΩ0=Vπ′π(S)Ω1=π′(π(S))Ω2=σ(S)Ω2.V_{\pi'}V_{\pi}S\Omega_{0}=V_{\pi'}\pi(S)\Omega_{1}=\pi'(\pi(S))\Omega_{2}=\sigma(S)\Omega_{2}.

By Trace-Preserving Embeddings of Tracial W*-Probability Spaces, Their Implementing Isometries and Conditional Expectations §isometry, VσV_{\sigma} is the unique element of L(H0,H2)\mathcal{L}(H_{0},H_{2}) with VσSΩ0=σ(S)Ω2V_{\sigma}S\Omega_{0}=\sigma(S)\Omega_{2} for every S∈M0S\in M_{0}, the uniqueness being that of A Trace-Preserving Unital -Homomorphism between Tracial W-Probability Spaces is Implemented by an Isometry §isometry applied to σ\sigma. Hence Vπ′∘π=Vπ′VπV_{\pi'\circ\pi}=V_{\pi'}V_{\pi}, which completes claim 2.

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Prerequisites

Loading...

Comments

Loading…