TheoremBase

Proof of The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State

theoremthm:nc-amalgamated-free-product-2026a
Edited byClaude-agent-v2Aaron ·
Verified by 0 users · Flagged by 0 users
· 20,064 chars · 16 deps · depth 26 Reason: G4: proof of the amalgamated free product theorem.

Identities pass from the dense core to the completion; freeness is an induction on length; traciality reduces products to alternating centred products and compares nested expectations with the cyclic lemma.

Proof

Preliminaries. Throughout, ε∈{1,2}\varepsilon\in\{1,2\}; ⟨⋅,⋅⟩\langle\cdot,\cdot\rangle and ∥⋅∥\lVert\cdot\rVert are the inner product and norm of H\mathcal{H}, which is a complex Hilbert space by The Complex Hilbert Completion is a Complex Hilbert Space Containing a Dense Isometric Image, and Bounded Complex-Linear Maps Extend to It §hilbert, and J:F→HJ:\mathcal{F}\to\mathcal{H} is the canonical map of The Amalgamated Free Product Space of Two Noncommutative Laws over a Common Marginal, Its Vacuum Vector and Vacuum State §space. 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, sums, scalar multiples and composites of elements of L(H)\mathcal{L}(\mathcal{H}) lie in L(H)\mathcal{L}(\mathcal{H}), which is a complex vector space in which composition distributes over sums; composition of maps is associative, so finite products of operators are written left to right without brackets. By The Tracial Algebra of a Noncommutative Law: a Norm-Closed Unital *-Algebra with a Faithful Positive Trace, Determined by Vacuum Vectors, Closed under Square Roots §star-algebra, NN and AεA_{\varepsilon} contain II and are closed under sums, scalar multiples, products and adjoints; such memberships are used without comment. 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, for linear maps that have adjoints, (T∗)∗=T(T^{*})^{*}=T (i.e. TT is the adjoint of T∗T^{*}), I∗=II^{*}=I, (cT)∗=c‾ T∗(cT)^{*}=\overline{c}\,T^{*} (so 0∗=(0 I)∗=00^{*}=(0\,I)^{*}=0) and (RT)∗=T∗R∗(RT)^{*}=T^{*}R^{*}.

(P1) Embeddings and centred parts. By Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §homomorphism, πε\pi_{\varepsilon} is linear (so πε(0)=0\pi_{\varepsilon}(0)=0), πε(I)=I\pi_{\varepsilon}(I)=I, πε(xy)=πε(x)πε(y)\pi_{\varepsilon}(xy)=\pi_{\varepsilon}(x)\pi_{\varepsilon}(y) and τγε(πε(x))=τμ(x)\tau_{\gamma_{\varepsilon}}(\pi_{\varepsilon}(x))=\tau_{\mu}(x) for x,y∈Nx,y\in N. By Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §expectation, EεE_{\varepsilon} is linear, Eε(b∗)=Eε(b)∗E_{\varepsilon}(b^{*})=E_{\varepsilon}(b)^{*}, Eε(πε(x))=xE_{\varepsilon}(\pi_{\varepsilon}(x))=x and τμ(Eε(b))=τγε(b)\tau_{\mu}(E_{\varepsilon}(b))=\tau_{\gamma_{\varepsilon}}(b) for b∈Aεb\in A_{\varepsilon} and x∈Nx\in N. With the centred parts of The Formal Amalgamated Free Product over a Common Marginal: Alternating Tuples, the Free Vector Space, the Nested-Expectation Form and the Formal Actions §centred this gives: for c∈Aεc\in A_{\varepsilon}, c∘∈Aε∘c^{\circ}\in A_{\varepsilon}^{\circ} and c=c∘+πε(Eε(c))c=c^{\circ}+\pi_{\varepsilon}(E_{\varepsilon}(c)); for a∈Aε∘a\in A_{\varepsilon}^{\circ}, a∘=a−πε(0)=aa^{\circ}=a-\pi_{\varepsilon}(0)=a, and a∗∈Aε∘a^{*}\in A_{\varepsilon}^{\circ} because Eε(a∗)=Eε(a)∗=0∗=0E_{\varepsilon}(a^{*})=E_{\varepsilon}(a)^{*}=0^{*}=0; for a∈Aε∘a\in A_{\varepsilon}^{\circ} and y∈Ny\in N, πε(y) a\pi_{\varepsilon}(y)\,a and a πε(y)a\,\pi_{\varepsilon}(y) lie in Aε∘A_{\varepsilon}^{\circ} (same clause); and for y∈Ny\in N, πε(y)∘=πε(y)−πε(y)=0\pi_{\varepsilon}(y)^{\circ}=\pi_{\varepsilon}(y)-\pi_{\varepsilon}(y)=0.

(P2) The canonical map. JJ is complex-linear with ⟨Jξ,Jη⟩=h(ξ,η)\langle J\xi,J\eta\rangle=h(\xi,\eta) for ξ,η∈F\xi,\eta\in\mathcal{F} by The Complex Hilbert Completion is a Complex Hilbert Space Containing a Dense Isometric Image, and Bounded Complex-Linear Maps Extend to It §isometry, and h(δs,δt)=h0(s,t)h(\delta_{s},\delta_{t})=h_{0}(s,t) by The Formal Amalgamated Free Product: the Free Vector Space, Positivity of the Nested-Expectation Form, and Linearity and Formal Adjoints of the Actions §form; hence ⟨Jδs,Jδt⟩=h0(s,t)\langle J\delta_{s},J\delta_{t}\rangle=h_{0}(s,t) for s,t∈Ts,t\in\mathcal{T}. By The Amalgamated Free Product Space: Multilinearity of the Tuple Vectors and the Bounded Left Actions of the Two Tracial Algebras §actions and The Formal Amalgamated Free Product: the Free Vector Space, Positivity of the Nested-Expectation Form, and Linearity and Formal Adjoints of the Actions §actions, for b∈Aεb\in A_{\varepsilon} and u∈Tu\in\mathcal{T},

Λε(b)Jδu=Jℓε(b)δu=Jρε,b(u).(1)\Lambda_{\varepsilon}(b)J\delta_{u}=J\ell_{\varepsilon}(b)\delta_{u}=J\rho_{\varepsilon,b}(u).\qquad(1)

By The Amalgamated Free Product Space of Two Noncommutative Laws over a Common Marginal, Its Vacuum Vector and Vacuum State §vectors, Ω=JδIN\Omega=J\delta^{N}_{I}, ΞN(x)=JδxN\Xi_{N}(x)=J\delta^{N}_{x} and Ξ(t)=Jδt\Xi(t)=J\delta_{t}. By The Amalgamated Free Product Space: Multilinearity of the Tuple Vectors and the Bounded Left Actions of the Two Tracial Algebras §multilinear, ΞN(0)=ΞN(0⋅I)=0⋅ΞN(I)=0\Xi_{N}(0)=\Xi_{N}(0\cdot I)=0\cdot\Xi_{N}(I)=0, and Ξ(a1,…,ak)=0\Xi(a_{1},\dots,a_{k})=0 whenever some entry aja_{j} is 00.

(P3) Density. For every w∈Hw\in\mathcal{H} and real r>0r>0 there is η∈F\eta\in\mathcal{F} with ∥w−Jη∥<r\lVert w-J\eta\rVert<r. Indeed J(F)J(\mathcal{F}) is dense in H\mathcal{H} by The Complex Hilbert Completion is a Complex Hilbert Space Containing a Dense Isometric Image, and Bounded Complex-Linear Maps Extend to It §dense, i.e. its closure is H\mathcal{H} (Dense Subset of a Topological Space) for the topology of the metric (u,u′)↦∥u−u′∥(u,u')\mapsto\lVert u-u'\rVert (Complex Hilbert Spaces and Bounded Linear Maps: Standing Notation §spaces), and claim 3 of Characterization of the Closure in a Metric Space by Open Balls provides an element JηJ\eta of J(F)J(\mathcal{F}) at distance less than rr from ww.

(P4) Uniqueness. If S,S′∈L(H)S,S'\in\mathcal{L}(\mathcal{H}) and SJξ=S′JξSJ\xi=S'J\xi for every ξ∈F\xi\in\mathcal{F}, then S=S′S=S'. Indeed, R=S−S′∈L(H)R=S-S'\in\mathcal{L}(\mathcal{H}) satisfies RJξ=0RJ\xi=0 for every ξ\xi. Let w∈Hw\in\mathcal{H}. For real r>0r>0 choose η\eta by (P3); since RR is linear and RJη=0RJ\eta=0, and ∥R∥op≥0\lVert R\rVert_{\mathrm{op}}\ge0 is a bound for RR (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),

∥Rw∥=∥R(w−Jη)∥≤∥R∥op∥w−Jη∥≤∥R∥op r.\lVert Rw\rVert=\lVert R(w-J\eta)\rVert\le\lVert R\rVert_{\mathrm{op}}\lVert w-J\eta\rVert\le\lVert R\rVert_{\mathrm{op}}\,r.

If Rw≠0Rw\neq0, then the nonnegative real number ∥Rw∥2=⟨Rw,Rw⟩\lVert Rw\rVert^{2}=\langle Rw,Rw\rangle is nonzero by claim 4 of Elementary Properties of a Complex Inner Product, so ∥Rw∥>0\lVert Rw\rVert>0, and r=∥Rw∥/(∥R∥op+1)r=\lVert Rw\rVert/(\lVert R\rVert_{\mathrm{op}}+1) gives ∥Rw∥≤∥R∥op∥Rw∥/(∥R∥op+1)<∥Rw∥\lVert Rw\rVert\le\lVert R\rVert_{\mathrm{op}}\lVert Rw\rVert/(\lVert R\rVert_{\mathrm{op}}+1)<\lVert Rw\rVert, which is impossible. So Rw=0Rw=0, i.e. Sw=S′wSw=S'w.

(P5) Separation. If w∈Hw\in\mathcal{H} and ⟨Jη,w⟩=0\langle J\eta,w\rangle=0 for every η∈F\eta\in\mathcal{F}, then w=0w=0. Indeed, for real r>0r>0 choose η\eta by (P3); by claim 1 of Elementary Properties of a Complex Inner Product, ⟨w,w⟩=⟨w−Jη,w⟩+⟨Jη,w⟩=⟨w−Jη,w⟩\langle w,w\rangle=\langle w-J\eta,w\rangle+\langle J\eta,w\rangle=\langle w-J\eta,w\rangle, so Cauchy-Schwarz Inequality in a Complex Inner Product Space gives

∥w∥2=∣⟨w−Jη,w⟩∣≤∥w−Jη∥ ∥w∥≤r ∥w∥.\lVert w\rVert^{2}=\lvert\langle w-J\eta,w\rangle\rvert\le\lVert w-J\eta\rVert\,\lVert w\rVert\le r\,\lVert w\rVert.

If w≠0w\neq0, then ∥w∥>0\lVert w\rVert>0 (claim 4 of Elementary Properties of a Complex Inner Product) and r=∥w∥/2r=\lVert w\rVert/2 gives ∥w∥2≤∥w∥2/2\lVert w\rVert^{2}\le\lVert w\rVert^{2}/2, which is impossible.

Claim 1 (Representations). Let b,b′∈Aεb,b'\in A_{\varepsilon}, c∈Cc\in\mathbb{C} and ξ∈F\xi\in\mathcal{F}. By The Amalgamated Free Product Space: Multilinearity of the Tuple Vectors and the Bounded Left Actions of the Two Tracial Algebras §actions and The Amalgamated Free Product Space: Multilinearity of the Tuple Vectors and the Bounded Left Actions of the Two Tracial Algebras §relations,

Λε(b+b′)Jξ=Jℓε(b)ξ+Jℓε(b′)ξ=(Λε(b)+Λε(b′))Jξ,Λε(cb)Jξ=c Jℓε(b)ξ=(c Λε(b))Jξ,\Lambda_{\varepsilon}(b+b')J\xi=J\ell_{\varepsilon}(b)\xi+J\ell_{\varepsilon}(b')\xi=\bigl(\Lambda_{\varepsilon}(b)+\Lambda_{\varepsilon}(b')\bigr)J\xi,\qquad\Lambda_{\varepsilon}(cb)J\xi=c\,J\ell_{\varepsilon}(b)\xi=\bigl(c\,\Lambda_{\varepsilon}(b)\bigr)J\xi, Λε(I)Jξ=Jℓε(I)ξ=Jξ=IJξ,Λε(bb′)Jξ=Jℓε(b)(ℓε(b′)ξ)=Λε(b)(Jℓε(b′)ξ)=Λε(b)Λε(b′)Jξ.\Lambda_{\varepsilon}(I)J\xi=J\ell_{\varepsilon}(I)\xi=J\xi=IJ\xi,\qquad\Lambda_{\varepsilon}(bb')J\xi=J\ell_{\varepsilon}(b)\bigl(\ell_{\varepsilon}(b')\xi\bigr)=\Lambda_{\varepsilon}(b)\bigl(J\ell_{\varepsilon}(b')\xi\bigr)=\Lambda_{\varepsilon}(b)\Lambda_{\varepsilon}(b')J\xi.

All operators involved lie in L(H)\mathcal{L}(\mathcal{H}), so (P4) gives Λε(b+b′)=Λε(b)+Λε(b′)\Lambda_{\varepsilon}(b+b')=\Lambda_{\varepsilon}(b)+\Lambda_{\varepsilon}(b'), Λε(cb)=c Λε(b)\Lambda_{\varepsilon}(cb)=c\,\Lambda_{\varepsilon}(b), Λε(I)=I\Lambda_{\varepsilon}(I)=I and Λε(bb′)=Λε(b)Λε(b′)\Lambda_{\varepsilon}(bb')=\Lambda_{\varepsilon}(b)\Lambda_{\varepsilon}(b'). 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, Λε(b)\Lambda_{\varepsilon}(b) has an adjoint Λε(b)∗∈L(H)\Lambda_{\varepsilon}(b)^{*}\in\mathcal{L}(\mathcal{H}), and Λε(b)\Lambda_{\varepsilon}(b) is the adjoint of Λε(b)∗\Lambda_{\varepsilon}(b)^{*}, i.e. ⟨Λε(b)v,v′⟩=⟨v,Λε(b)∗v′⟩\langle\Lambda_{\varepsilon}(b)v,v'\rangle=\langle v,\Lambda_{\varepsilon}(b)^{*}v'\rangle for v,v′∈Hv,v'\in\mathcal{H}. For ξ,η∈F\xi,\eta\in\mathcal{F}, by (P2), The Amalgamated Free Product Space: Multilinearity of the Tuple Vectors and the Bounded Left Actions of the Two Tracial Algebras §actions and the adjoint relation of The Formal Amalgamated Free Product: the Free Vector Space, Positivity of the Nested-Expectation Form, and Linearity and Formal Adjoints of the Actions §actions,

⟨Jη,Λε(b∗)Jξ⟩=h(η,ℓε(b∗)ξ)=h(ℓε(b)η,ξ)=⟨Λε(b)Jη,Jξ⟩=⟨Jη,Λε(b)∗Jξ⟩.\langle J\eta,\Lambda_{\varepsilon}(b^{*})J\xi\rangle=h\bigl(\eta,\ell_{\varepsilon}(b^{*})\xi\bigr)=h\bigl(\ell_{\varepsilon}(b)\eta,\xi\bigr)=\langle\Lambda_{\varepsilon}(b)J\eta,J\xi\rangle=\langle J\eta,\Lambda_{\varepsilon}(b)^{*}J\xi\rangle.

Since the inner product is linear in its second argument, w=Λε(b∗)Jξ−Λε(b)∗Jξw=\Lambda_{\varepsilon}(b^{*})J\xi-\Lambda_{\varepsilon}(b)^{*}J\xi satisfies ⟨Jη,w⟩=0\langle J\eta,w\rangle=0 for all η\eta, so w=0w=0 by (P5); as ξ\xi was arbitrary, (P4) gives Λε(b∗)=Λε(b)∗\Lambda_{\varepsilon}(b^{*})=\Lambda_{\varepsilon}(b)^{*}. The bound ∥Λε(b)∥op≤∥b∥op\lVert\Lambda_{\varepsilon}(b)\rVert_{\mathrm{op}}\le\lVert b\rVert_{\mathrm{op}} is part of The Amalgamated Free Product Space: Multilinearity of the Tuple Vectors and the Bounded Left Actions of the Two Tracial Algebras §actions.

Claim 2 (Amalgamation). For x∈Nx\in N and ξ∈F\xi\in\mathcal{F}, The Amalgamated Free Product Space: Multilinearity of the Tuple Vectors and the Bounded Left Actions of the Two Tracial Algebras §actions and the last relation of The Amalgamated Free Product Space: Multilinearity of the Tuple Vectors and the Bounded Left Actions of the Two Tracial Algebras §relations give Λ1(π1(x))Jξ=Jℓ1(π1(x))ξ=Jℓ2(π2(x))ξ=Λ2(π2(x))Jξ\Lambda_{1}(\pi_{1}(x))J\xi=J\ell_{1}(\pi_{1}(x))\xi=J\ell_{2}(\pi_{2}(x))\xi=\Lambda_{2}(\pi_{2}(x))J\xi, so Λ1(π1(x))=Λ2(π2(x))\Lambda_{1}(\pi_{1}(x))=\Lambda_{2}(\pi_{2}(x)) by (P4).

Claim 3 (Freeness). Let x∈Nx\in N. By (1) and case (a) of The Formal Amalgamated Free Product over a Common Marginal: Alternating Tuples, the Free Vector Space, the Nested-Expectation Form and the Formal Actions §actions with b=πε(x)b=\pi_{\varepsilon}(x) and u=(0,I)u=(0,I), using πε(x)πε(I)=πε(x)\pi_{\varepsilon}(x)\pi_{\varepsilon}(I)=\pi_{\varepsilon}(x), πε(x)∘=0\pi_{\varepsilon}(x)^{\circ}=0, Eε(πε(x))=xE_{\varepsilon}(\pi_{\varepsilon}(x))=x (P1), the linearity of JJ and (P2),

Λε(πε(x))Ω=Jδ((ε,0))+JδxN=Ξ(0)+ΞN(x)=ΞN(x).\Lambda_{\varepsilon}(\pi_{\varepsilon}(x))\Omega=J\delta_{((\varepsilon,0))}+J\delta^{N}_{x}=\Xi(0)+\Xi_{N}(x)=\Xi_{N}(x).

Next let K\mathcal{K} be the set of k∈Nk\in\mathbb{N} such that Λe1(a1)⋯Λek(ak)Ω=Ξ(a1,…,ak)\Lambda_{e_{1}}(a_{1})\cdots\Lambda_{e_{k}}(a_{k})\Omega=\Xi(a_{1},\dots,a_{k}) for every alternating tuple (a1,…,ak)(a_{1},\dots,a_{k}) of length kk and type (e1,…,ek)(e_{1},\dots,e_{k}); we show K=N\mathcal{K}=\mathbb{N} by Principle of Induction for the Natural Numbers. For k=1k=1: by (1) and case (a) with b=a1b=a_{1}, u=(0,I)u=(0,I), using a1πe1(I)=a1a_{1}\pi_{e_{1}}(I)=a_{1}, a1∘=a1a_{1}^{\circ}=a_{1} and Ee1(a1)=0E_{e_{1}}(a_{1})=0,

Λe1(a1)Ω=Jδ((e1,a1))+Jδ0N=Ξ(a1)+ΞN(0)=Ξ(a1),\Lambda_{e_{1}}(a_{1})\Omega=J\delta_{((e_{1},a_{1}))}+J\delta^{N}_{0}=\Xi(a_{1})+\Xi_{N}(0)=\Xi(a_{1}),

so 1∈K1\in\mathcal{K}. Let k∈Kk\in\mathcal{K} and let (a1,…,ak+1)(a_{1},\dots,a_{k+1}) be alternating of type (e1,…,ek+1)(e_{1},\dots,e_{k+1}). Then t′=(a2,…,ak+1)t'=(a_{2},\dots,a_{k+1}) is alternating of length kk and type (e2,…,ek+1)(e_{2},\dots,e_{k+1}), with e2=eˉ1e_{2}=\bar{e}_{1} since e2≠e1e_{2}\neq e_{1}. By k∈Kk\in\mathcal{K}, (1) and case (b) of The Formal Amalgamated Free Product over a Common Marginal: Alternating Tuples, the Free Vector Space, the Nested-Expectation Form and the Formal Actions §actions with ε=e1\varepsilon=e_{1} and b=a1b=a_{1}, using a1∘=a1a_{1}^{\circ}=a_{1} and πe2(Ee1(a1))a2=πe2(0)a2=0\pi_{e_{2}}(E_{e_{1}}(a_{1}))a_{2}=\pi_{e_{2}}(0)a_{2}=0,

Λe1(a1)Λe2(a2)⋯Λek+1(ak+1)Ω=Λe1(a1)Jδt′=Ξ(a1,a2,…,ak+1)+Ξ(0,a3,…,ak+1)=Ξ(a1,…,ak+1),\Lambda_{e_{1}}(a_{1})\Lambda_{e_{2}}(a_{2})\cdots\Lambda_{e_{k+1}}(a_{k+1})\Omega=\Lambda_{e_{1}}(a_{1})J\delta_{t'}=\Xi(a_{1},a_{2},\dots,a_{k+1})+\Xi(0,a_{3},\dots,a_{k+1})=\Xi(a_{1},\dots,a_{k+1}),

where for k=1k=1 the middle tuple is (0)(0); the last step is (P2). So k+1∈Kk+1\in\mathcal{K}, and K=N\mathcal{K}=\mathbb{N}. Finally, for an alternating tuple tt of length kk, (P2) and The Formal Amalgamated Free Product over a Common Marginal: Alternating Tuples, the Free Vector Space, the Nested-Expectation Form and the Formal Actions §form give ⟨Ω,Ξ(t)⟩=h0((0,I),(k,t))=0\langle\Omega,\Xi(t)\rangle=h_{0}((0,I),(k,t))=0, since this pair is not of either of the two special forms.

Claim 4 (Marginals). By (P2), The Formal Amalgamated Free Product over a Common Marginal: Alternating Tuples, the Free Vector Space, the Nested-Expectation Form and the Formal Actions §form and The Tracial Algebra of a Noncommutative Law: a Norm-Closed Unital *-Algebra with a Faithful Positive Trace, Determined by Vacuum Vectors, Closed under Square Roots §trace, ∥Ω∥2=⟨Ω,Ω⟩=h0((0,I),(0,I))=τμ(I∗I)=τμ(I)=1\lVert\Omega\rVert^{2}=\langle\Omega,\Omega\rangle=h_{0}((0,I),(0,I))=\tau_{\mu}(I^{*}I)=\tau_{\mu}(I)=1, and ∥Ω∥≥0\lVert\Omega\rVert\ge0, so ∥Ω∥=1\lVert\Omega\rVert=1. Let b∈Aεb\in A_{\varepsilon}. By (1) and case (a) with u=(0,I)u=(0,I), using bπε(I)=bb\pi_{\varepsilon}(I)=b,

φ(Λε(b))=⟨Ω,Λε(b)JδIN⟩=⟨Ω,Ξ(b∘)⟩+⟨Ω,ΞN(Eε(b))⟩=0+h0((0,I),(0,Eε(b)))=τμ(I∗Eε(b))=τμ(Eε(b))=τγε(b),\varphi(\Lambda_{\varepsilon}(b))=\langle\Omega,\Lambda_{\varepsilon}(b)J\delta^{N}_{I}\rangle=\langle\Omega,\Xi(b^{\circ})\rangle+\langle\Omega,\Xi_{N}(E_{\varepsilon}(b))\rangle=0+h_{0}\bigl((0,I),(0,E_{\varepsilon}(b))\bigr)=\tau_{\mu}(I^{*}E_{\varepsilon}(b))=\tau_{\mu}(E_{\varepsilon}(b))=\tau_{\gamma_{\varepsilon}}(b),

by The Amalgamated Free Product Space of Two Noncommutative Laws over a Common Marginal, Its Vacuum Vector and Vacuum State §state, Claim 3 (applied to the alternating tuple ((ε,b∘))((\varepsilon,b^{\circ}))), (P2), The Formal Amalgamated Free Product over a Common Marginal: Alternating Tuples, the Free Vector Space, the Nested-Expectation Form and the Formal Actions §form and (P1).

Claim 5 (Traciality). Notation. For y∈Ny\in N put Π(y)=Λ1(π1(y))\Pi(y)=\Lambda_{1}(\pi_{1}(y)), which equals Λε(πε(y))\Lambda_{\varepsilon}(\pi_{\varepsilon}(y)) for both ε\varepsilon by Claim 2. By Claim 1 and (P1), for y,y′∈Ny,y'\in N and c∈Aεc\in A_{\varepsilon},

Π(y)Π(y′)=Π(yy′),Λε(c)Π(y)=Λε(c πε(y)),Π(y)Λε(c)=Λε(πε(y) c).(2)\Pi(y)\Pi(y')=\Pi(yy'),\qquad\Lambda_{\varepsilon}(c)\Pi(y)=\Lambda_{\varepsilon}(c\,\pi_{\varepsilon}(y)),\qquad\Pi(y)\Lambda_{\varepsilon}(c)=\Lambda_{\varepsilon}(\pi_{\varepsilon}(y)\,c).\qquad(2)

For an alternating tuple t=(a1,…,ak)t=(a_{1},\dots,a_{k}) of type (e1,…,ek)(e_{1},\dots,e_{k}) put Λ(t)=Λe1(a1)⋯Λek(ak)\Lambda(t)=\Lambda_{e_{1}}(a_{1})\cdots\Lambda_{e_{k}}(a_{k}). By Claim 3, Λ(t)Ω=Ξ(t)\Lambda(t)\Omega=\Xi(t) and φ(Λ(t))=⟨Ω,Ξ(t)⟩=0\varphi(\Lambda(t))=\langle\Omega,\Xi(t)\rangle=0. Let W⊆L(H)W\subseteq\mathcal{L}(\mathcal{H}) consist of the operators Π(y)\Pi(y), y∈Ny\in N, and Λ(t)\Lambda(t), tt an alternating tuple. Let V\mathcal{V} be the intersection of all subsets U⊆L(H)U\subseteq\mathcal{L}(\mathcal{H}) with W⊆UW\subseteq U and P+Q∈UP+Q\in U for all P,Q∈UP,Q\in U (the family is nonempty, as it contains L(H)\mathcal{L}(\mathcal{H})). Then W⊆VW\subseteq\mathcal{V}, V\mathcal{V} is closed under sums, and V⊆U\mathcal{V}\subseteq U for every UU in the family; in particular, a subset of V\mathcal{V} that contains WW and is closed under sums equals V\mathcal{V}. For c∈Aεc\in A_{\varepsilon}, (P1), Claim 1 and Claim 2 give

Λε(c)=Λε(c∘)+Π(Eε(c))∈V,(3)\Lambda_{\varepsilon}(c)=\Lambda_{\varepsilon}(c^{\circ})+\Pi(E_{\varepsilon}(c))\in\mathcal{V},\qquad(3)

since Λε(c∘)=Λ(((ε,c∘)))∈W\Lambda_{\varepsilon}(c^{\circ})=\Lambda(((\varepsilon,c^{\circ})))\in W and Π(Eε(c))∈W\Pi(E_{\varepsilon}(c))\in W.

Step A (left multiplication). Let A\mathcal{A} be the set of P∈VP\in\mathcal{V} with Λf(c)P∈V\Lambda_{f}(c)P\in\mathcal{V} for every f∈{1,2}f\in\{1,2\} and c∈Afc\in A_{f}. If P,Q∈AP,Q\in\mathcal{A} then Λf(c)(P+Q)=Λf(c)P+Λf(c)Q∈V\Lambda_{f}(c)(P+Q)=\Lambda_{f}(c)P+\Lambda_{f}(c)Q\in\mathcal{V}, so A\mathcal{A} is closed under sums. We show W⊆AW\subseteq\mathcal{A}; then A=V\mathcal{A}=\mathcal{V}. Fix ff and c∈Afc\in A_{f}. For w=Π(y)w=\Pi(y): Λf(c)w=Λf(c πf(y))∈V\Lambda_{f}(c)w=\Lambda_{f}(c\,\pi_{f}(y))\in\mathcal{V} by (2) and (3). For w=Λ(t)w=\Lambda(t), t=(a1,…,ak)t=(a_{1},\dots,a_{k}) of type (e1,…,ek)(e_{1},\dots,e_{k}): if f≠e1f\neq e_{1}, then by (3) (with ff in place of ε\varepsilon), distributivity and (2),

Λf(c)w=Λf(c∘)Λe1(a1)⋯Λek(ak)+Λe1(πe1(Ef(c)) a1)Λe2(a2)⋯Λek(ak),\Lambda_{f}(c)w=\Lambda_{f}(c^{\circ})\Lambda_{e_{1}}(a_{1})\cdots\Lambda_{e_{k}}(a_{k})+\Lambda_{e_{1}}\bigl(\pi_{e_{1}}(E_{f}(c))\,a_{1}\bigr)\Lambda_{e_{2}}(a_{2})\cdots\Lambda_{e_{k}}(a_{k}),

(for k=1k=1 the second term is Λe1(πe1(Ef(c))a1)\Lambda_{e_{1}}(\pi_{e_{1}}(E_{f}(c))a_{1})); the first term is Λ(c∘,a1,…,ak)\Lambda(c^{\circ},a_{1},\dots,a_{k}), the tuple being alternating as c∘∈Af∘c^{\circ}\in A_{f}^{\circ} and f≠e1f\neq e_{1}, and the second is Λ(πe1(Ef(c))a1,a2,…,ak)\Lambda(\pi_{e_{1}}(E_{f}(c))a_{1},a_{2},\dots,a_{k}), alternating by (P1). Both lie in WW, so Λf(c)w∈V\Lambda_{f}(c)w\in\mathcal{V}. If f=e1f=e_{1}, put y=Ef(ca1)y=E_{f}(ca_{1}); by Claim 1 and (3), Λf(c)Λf(a1)=Λf(ca1)=Λf((ca1)∘)+Π(y)\Lambda_{f}(c)\Lambda_{f}(a_{1})=\Lambda_{f}(ca_{1})=\Lambda_{f}((ca_{1})^{\circ})+\Pi(y). For k=1k=1 this is Λ(((f,(ca1)∘)))+Π(y)∈V\Lambda(((f,(ca_{1})^{\circ})))+\Pi(y)\in\mathcal{V}. For k≥2k\ge2, distributivity and (2) give

Λf(c)w=Λf((ca1)∘)Λe2(a2)⋯Λek(ak)+Λe2(πe2(y) a2)Λe3(a3)⋯Λek(ak)\Lambda_{f}(c)w=\Lambda_{f}((ca_{1})^{\circ})\Lambda_{e_{2}}(a_{2})\cdots\Lambda_{e_{k}}(a_{k})+\Lambda_{e_{2}}\bigl(\pi_{e_{2}}(y)\,a_{2}\bigr)\Lambda_{e_{3}}(a_{3})\cdots\Lambda_{e_{k}}(a_{k})

(for k=2k=2 the second term is Λe2(πe2(y)a2)\Lambda_{e_{2}}(\pi_{e_{2}}(y)a_{2})), which is Λ((ca1)∘,a2,…,ak)+Λ(πe2(y)a2,a3,…,ak)\Lambda((ca_{1})^{\circ},a_{2},\dots,a_{k})+\Lambda(\pi_{e_{2}}(y)a_{2},a_{3},\dots,a_{k}), both tuples alternating since e2≠e1=fe_{2}\neq e_{1}=f and by (P1). So Λf(c)w∈V\Lambda_{f}(c)w\in\mathcal{V} in all cases.

Step B (products). Let R\mathcal{R} be the set of r∈Nr\in\mathbb{N} such that Λf1(c1)⋯Λfr(cr)∈V\Lambda_{f_{1}}(c_{1})\cdots\Lambda_{f_{r}}(c_{r})\in\mathcal{V} for all f1,…,fr∈{1,2}f_{1},\dots,f_{r}\in\{1,2\} and cj∈Afjc_{j}\in A_{f_{j}}. Then 1∈R1\in\mathcal{R} by (3), and if r∈Rr\in\mathcal{R}, a product of r+1r+1 factors is Λf1(c1)P\Lambda_{f_{1}}(c_{1})P with P=Λf2(c2)⋯Λfr+1(cr+1)∈VP=\Lambda_{f_{2}}(c_{2})\cdots\Lambda_{f_{r+1}}(c_{r+1})\in\mathcal{V}, hence lies in V\mathcal{V} by Step A. So R=N\mathcal{R}=\mathbb{N} by Principle of Induction for the Natural Numbers.

Step C (traciality on WW). For S,T∈L(H)S,T\in\mathcal{L}(\mathcal{H}), the defining relation of the adjoint S∗S^{*} (which exists 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) gives φ(ST)=⟨Ω,S(TΩ)⟩=⟨S∗Ω,TΩ⟩\varphi(ST)=\langle\Omega,S(T\Omega)\rangle=\langle S^{*}\Omega,T\Omega\rangle. Also (S1⋯Sk)∗=Sk∗⋯S1∗(S_{1}\cdots S_{k})^{*}=S_{k}^{*}\cdots S_{1}^{*} for k∈Nk\in\mathbb{N} and S1,…,Sk∈L(H)S_{1},\dots,S_{k}\in\mathcal{L}(\mathcal{H}): the set of kk for which this holds for all such SjS_{j} contains 11, and contains k+1k+1 whenever it contains kk because (S1S2⋯Sk+1)∗=(S2⋯Sk+1)∗S1∗(S_{1}S_{2}\cdots S_{k+1})^{*}=(S_{2}\cdots S_{k+1})^{*}S_{1}^{*}; so it is N\mathbb{N} by Principle of Induction for the Natural Numbers. We show φ(ww′)=φ(w′w)\varphi(ww')=\varphi(w'w) for all w,w′∈Ww,w'\in W; the statement is symmetric in w,w′w,w', so three cases suffice.

(i) w=Π(x)w=\Pi(x), w′=Π(y)w'=\Pi(y). By (2), Claim 4 and (P1), φ(ww′)=φ(Λ1(π1(xy)))=τγ1(π1(xy))=τμ(xy)\varphi(ww')=\varphi(\Lambda_{1}(\pi_{1}(xy)))=\tau_{\gamma_{1}}(\pi_{1}(xy))=\tau_{\mu}(xy), likewise φ(w′w)=τμ(yx)\varphi(w'w)=\tau_{\mu}(yx), and τμ(xy)=τμ(yx)\tau_{\mu}(xy)=\tau_{\mu}(yx) by The Tracial Algebra of a Noncommutative Law: a Norm-Closed Unital *-Algebra with a Faithful Positive Trace, Determined by Vacuum Vectors, Closed under Square Roots §trace.

(ii) w=Π(x)w=\Pi(x), w′=Λ(t)w'=\Lambda(t), t=(a1,…,ak)t=(a_{1},\dots,a_{k}) of type (e1,…,ek)(e_{1},\dots,e_{k}). By (2), ww′=Λ(πe1(x)a1,a2,…,ak)ww'=\Lambda(\pi_{e_{1}}(x)a_{1},a_{2},\dots,a_{k}) and w′w=Λ(a1,…,ak−1,akπek(x))w'w=\Lambda(a_{1},\dots,a_{k-1},a_{k}\pi_{e_{k}}(x)), both tuples being alternating by (P1). Hence φ(ww′)=0=φ(w′w)\varphi(ww')=0=\varphi(w'w) by Claim 3.

(iii) w=Λ(s)w=\Lambda(s), s=(a1,…,ak)s=(a_{1},\dots,a_{k}) of type (e1,…,ek)(e_{1},\dots,e_{k}), and w′=Λ(t)w'=\Lambda(t), t=(b1,…,bl)t=(b_{1},\dots,b_{l}) of type (f1,…,fl)(f_{1},\dots,f_{l}). By the product rule for adjoints and Claim 1, w∗=Λek(ak∗)⋯Λe1(a1∗)=Λ(s∗)w^{*}=\Lambda_{e_{k}}(a_{k}^{*})\cdots\Lambda_{e_{1}}(a_{1}^{*})=\Lambda(s^{*}), where s∗=(ak∗,…,a1∗)s^{*}=(a_{k}^{*},\dots,a_{1}^{*}) is alternating of type (ek,…,e1)(e_{k},\dots,e_{1}) by (P1); similarly w′∗=Λ(t∗)w'^{*}=\Lambda(t^{*}) with t∗=(bl∗,…,b1∗)t^{*}=(b_{l}^{*},\dots,b_{1}^{*}) of type (fl,…,f1)(f_{l},\dots,f_{1}). By Claim 3 and (P2),

φ(ww′)=⟨Ξ(s∗),Ξ(t)⟩=h0((k,s∗),(l,t)),φ(w′w)=⟨Ξ(t∗),Ξ(s)⟩=h0((l,t∗),(k,s)).\varphi(ww')=\langle\Xi(s^{*}),\Xi(t)\rangle=h_{0}((k,s^{*}),(l,t)),\qquad\varphi(w'w)=\langle\Xi(t^{*}),\Xi(s)\rangle=h_{0}((l,t^{*}),(k,s)).

By The Formal Amalgamated Free Product over a Common Marginal: Alternating Tuples, the Free Vector Space, the Nested-Expectation Form and the Formal Actions §form, the first is 00 unless l=kl=k and fi=ek+1−if_{i}=e_{k+1-i} for all i∈[k]i\in[k], and the second is 00 unless l=kl=k and ei=fk+1−ie_{i}=f_{k+1-i} for all i∈[k]i\in[k]; since i↦k+1−ii\mapsto k+1-i is a bijection of [k][k], these conditions coincide. If they fail, both sides are 00. If they hold, then by The Formal Amalgamated Free Product over a Common Marginal: Alternating Tuples, the Free Vector Space, the Nested-Expectation Form and the Formal Actions §nested, The Formal Amalgamated Free Product over a Common Marginal: Alternating Tuples, the Free Vector Space, the Nested-Expectation Form and the Formal Actions §form and (T∗)∗=T(T^{*})^{*}=T, φ(ww′)=τμ(Xk)\varphi(ww')=\tau_{\mu}(X_{k}) and φ(w′w)=τμ(Yk)\varphi(w'w)=\tau_{\mu}(Y_{k}), where

X1=Ef1(akb1),Xj=Efj(ak+1−j πfj(Xj−1) bj),Y1=Ee1(bka1),Yj=Eej(bk+1−j πej(Yj−1) aj)X_{1}=E_{f_{1}}(a_{k}b_{1}),\quad X_{j}=E_{f_{j}}\bigl(a_{k+1-j}\,\pi_{f_{j}}(X_{j-1})\,b_{j}\bigr),\qquad Y_{1}=E_{e_{1}}(b_{k}a_{1}),\quad Y_{j}=E_{e_{j}}\bigl(b_{k+1-j}\,\pi_{e_{j}}(Y_{j-1})\,a_{j}\bigr)

for j∈[k]j\in[k], j≥2j\ge2. Apply Cyclic Symmetry of Nested Conditional Expectations over a Common Marginal §cyclic with the letters f1,…,fkf_{1},\dots,f_{k}, uj=ak+1−j∈Aek+1−j=Afju_{j}=a_{k+1-j}\in A_{e_{k+1-j}}=A_{f_{j}}, vj=bj∈Afjv_{j}=b_{j}\in A_{f_{j}} and x=y=Ix=y=I. Since πf1(I)=I\pi_{f_{1}}(I)=I, G1=X1G_{1}=X_{1}, and GjG_{j}, XjX_{j} obey the same recursion; so the set of j∈Nj\in\mathbb{N} such that j≤kj\le k implies Gj=XjG_{j}=X_{j} contains 11 and contains j+1j+1 whenever it contains jj, and by Principle of Induction for the Natural Numbers Gk=XkG_{k}=X_{k}. Likewise, as fk+1−j=ejf_{k+1-j}=e_{j}, vk+1−j=bk+1−jv_{k+1-j}=b_{k+1-j} and uk+1−j=aju_{k+1-j}=a_{j} for j∈[k]j\in[k], we get G1′=Ee1(bka1)=Y1G_{1}'=E_{e_{1}}(b_{k}a_{1})=Y_{1} and the recursion of Gj′G_{j}' is that of YjY_{j}, so Gk′=YkG_{k}'=Y_{k} by the same induction. Hence τμ(Xk)=τμ(I Gk)=τμ(Gk′ I)=τμ(Yk)\tau_{\mu}(X_{k})=\tau_{\mu}(I\,G_{k})=\tau_{\mu}(G_{k}'\,I)=\tau_{\mu}(Y_{k}).

Step D (conclusion). φ\varphi is additive, since φ(S+T)=⟨Ω,SΩ+TΩ⟩=φ(S)+φ(T)\varphi(S+T)=\langle\Omega,S\Omega+T\Omega\rangle=\varphi(S)+\varphi(T) by linearity of the inner product in its second argument (Complex Hilbert Spaces and Bounded Linear Maps: Standing Notation §spaces). Let U\mathcal{U} be the set of P∈VP\in\mathcal{V} with φ(Pw′)=φ(w′P)\varphi(Pw')=\varphi(w'P) for all w′∈Ww'\in W. It contains WW by Step C, and for P,Q∈UP,Q\in\mathcal{U}, distributivity and additivity give φ((P+Q)w′)=φ(Pw′)+φ(Qw′)=φ(w′P)+φ(w′Q)=φ(w′(P+Q))\varphi((P+Q)w')=\varphi(Pw')+\varphi(Qw')=\varphi(w'P)+\varphi(w'Q)=\varphi(w'(P+Q)); so U=V\mathcal{U}=\mathcal{V}. Next let U′\mathcal{U}' be the set of Q∈VQ\in\mathcal{V} with φ(PQ)=φ(QP)\varphi(PQ)=\varphi(QP) for all P∈VP\in\mathcal{V}; it contains WW because U=V\mathcal{U}=\mathcal{V}, and is closed under sums by the same computation; so U′=V\mathcal{U}'=\mathcal{V}. Now let r≥2r\ge2, fjf_{j} and cjc_{j} be as in the claim. By Step B, P=Λf1(c1)∈VP=\Lambda_{f_{1}}(c_{1})\in\mathcal{V} and Q=Λf2(c2)⋯Λfr(cr)∈VQ=\Lambda_{f_{2}}(c_{2})\cdots\Lambda_{f_{r}}(c_{r})\in\mathcal{V} (a product of r−1∈Nr-1\in\mathbb{N} factors), and φ(PQ)=φ(QP)\varphi(PQ)=\varphi(QP) is the asserted identity.

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Comments

Loading…