TheoremBase

Glue the couplings one at a time by amalgamated free products over the law mu into a consistent sequence of joint laws, realise these as a chain of tracial W*-probability spaces linked by marginal embeddings, and read off s and tkt^k in the inductive limit.

Proof

Each result cited is universally quantified over the data in its own statement. Fix dd, RR, μ\mu, (νk)k∈N(\nu_{k})_{k\in\mathbb{N}} and (γk)k∈N(\gamma_{k})_{k\in\mathbb{N}} as in the statement.

Notation. For a law λ\lambda and a polynomial pp in its variables, LpλL^{\lambda}_{p} is the left multiplication operator on the complex GNS space Hλ\mathcal{H}_{\lambda} given by Left and Right Multiplication Operators on the Complex GNS Space of a Noncommutative Law: Boundedness, Algebra Rules, Adjoints, Commutation, the Vacuum and the Conjugation §multiplication; Ωλ\Omega_{\lambda}, Mλ\mathcal{M}_{\lambda} and τλ\tau_{\lambda} are the vacuum vector, tracial algebra and trace, as in Noncommutative Laws, Couplings and the Wasserstein Distance: Standing Notation §gns. The letter II denotes an identity operator. For k∈Nk\in\mathbb{N}, with kdkd the product of kk and dd, let

βk=σ(x1,…,x(k+1)d):P(k+1)d→P(k+2)d,δk=σ(x1,…,xd,xkd+1,…,x(k+1)d):P2d→P(k+1)d,αk=σ(x1,…,xd):Pd→P(k+1)d\beta_{k}=\sigma_{(x_{1},\dots,x_{(k+1)d})}:\mathcal{P}_{(k+1)d}\to\mathcal{P}_{(k+2)d},\qquad\delta_{k}=\sigma_{(x_{1},\dots,x_{d},x_{kd+1},\dots,x_{(k+1)d})}:\mathcal{P}_{2d}\to\mathcal{P}_{(k+1)d},\qquad\alpha_{k}=\sigma_{(x_{1},\dots,x_{d})}:\mathcal{P}_{d}\to\mathcal{P}_{(k+1)d}

be the substitutions of the indicated tuples of variables. A map Γ\Gamma satisfies (Bk)(B_{k}) if Γ∈Σ(k+1)d,R\Gamma\in\Sigma_{(k+1)d,R} and Γ∘δk=γk\Gamma\circ\delta_{k}=\gamma_{k}. For l∈Nl\in\mathbb{N}, every variable xix_{i} of Pl\mathcal{P}_{l} lies in Pl,sa\mathcal{P}_{l,\mathrm{sa}} by Noncommutative Polynomials Form a Unital Complex Algebra with Involution: Linear Extension from Monomials, Products, Adjoints and Self-Adjoint Parts §self-adjoint and satisfies xi∗=xix_{i}^{*}=x_{i} by Noncommutative Polynomials Form a Unital Complex Algebra with Involution: Linear Extension from Monomials, Products, Adjoints and Self-Adjoint Parts §adjoint; hence, for every λ∈Σl\lambda\in\Sigma_{l} (which lies in some Σl,r\Sigma_{l,r} by Noncommutative Laws of Finitely Many Self-Adjoint Variables with a Norm Bound §law) and i∈[l]i\in[l], Left and Right Multiplication Operators on the Complex GNS Space of a Noncommutative Law: Boundedness, Algebra Rules, Adjoints, Commutation, the Vacuum and the Conjugation §adjoint gives (Lxiλ)∗=Lxiλ(L^{\lambda}_{x_{i}})^{*}=L^{\lambda}_{x_{i}} and that LxiλL^{\lambda}_{x_{i}} is self-adjoint.

Step 1 (the couplings and the first block). Let k∈Nk\in\mathbb{N}. Since μ,νk∈Σd,R\mu,\nu_{k}\in\Sigma_{d,R}, Couplings of Noncommutative Laws: the Norm Bound, the Cost Identity, the Tensor, Diagonal and Swapped Couplings, Weak-Star Closedness, and Displacement Interpolants §bound gives γk∈Σ2d,R\gamma_{k}\in\Sigma_{2d,R}, so γk∈Σ2d\gamma_{k}\in\Sigma_{2d} by Noncommutative Laws of Finitely Many Self-Adjoint Variables with a Norm Bound §law. By Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §composition, δk∘ι1=σc\delta_{k}\circ\iota^{1}=\sigma_{c} with cj=δk(xj)c_{j}=\delta_{k}(x_{j}) for j∈[d]j\in[d], and δk(xj)=xj\delta_{k}(x_{j})=x_{j} by Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §values (the first dd entries of the tuple of δk\delta_{k} are x1,…,xdx_{1},\dots,x_{d}); thus δk∘ι1=αk\delta_{k}\circ\iota^{1}=\alpha_{k}, where ι1=σ(x1,…,xd)\iota^{1}=\sigma_{(x_{1},\dots,x_{d})} by Couplings of Two Noncommutative Laws and Their Quadratic Cost §marginals. Consequently, if Γ\Gamma satisfies (Bk)(B_{k}), then by Couplings of Two Noncommutative Laws and Their Quadratic Cost §coupling (for γk∈Π(μ,νk)\gamma_{k}\in\Pi(\mu,\nu_{k}))

Γ∘αk=Γ∘δk∘ι1=γk∘ι1=μ.(1.1)\Gamma\circ\alpha_{k}=\Gamma\circ\delta_{k}\circ\iota^{1}=\gamma_{k}\circ\iota^{1}=\mu.\qquad(1.1)

Step 2 (one gluing step: the amalgamation data). Fix k∈Nk\in\mathbb{N} and a map Γ\Gamma satisfying (Bk)(B_{k}). Take the data of Two Noncommutative Laws with a Common Marginal: Standing Notation for Their Amalgamated Free Product §data with n=dn=d; with m1=(k+1)dm_{1}=(k+1)d, first law Γ\Gamma and the dd-tuple a1=(x1,…,xd)a^{1}=(x_{1},\dots,x_{d}) in P(k+1)d,sa\mathcal{P}_{(k+1)d,\mathrm{sa}}; and with m2=2dm_{2}=2d, second law γk+1\gamma_{k+1} and the dd-tuple a2=(x1,…,xd)a^{2}=(x_{1},\dots,x_{d}) in P2d,sa\mathcal{P}_{2d,\mathrm{sa}}. Here Γ∈Σ(k+1)d\Gamma\in\Sigma_{(k+1)d} by (Bk)(B_{k}) and Noncommutative Laws of Finitely Many Self-Adjoint Variables with a Norm Bound §law, and γk+1∈Σ2d\gamma_{k+1}\in\Sigma_{2d} by Step 1. The common marginal condition holds with marginal μ\mu: σa1=αk\sigma_{a^{1}}=\alpha_{k}, so Γ∘σa1=μ\Gamma\circ\sigma_{a^{1}}=\mu by (1.1); and σa2=ι1\sigma_{a^{2}}=\iota^{1}, so γk+1∘σa2=μ\gamma_{k+1}\circ\sigma_{a^{2}}=\mu by Couplings of Two Noncommutative Laws and Their Quadratic Cost §coupling. The two laws which that setting calls γ1,γ2\gamma_{1},\gamma_{2} are here Γ\Gamma and γk+1\gamma_{k+1} (our couplings γ1,γ2\gamma_{1},\gamma_{2} play no role in this step); its common marginal is our μ\mu. In the notation of Two Noncommutative Laws with a Common Marginal: Standing Notation for Their Amalgamated Free Product §algebras and Two Noncommutative Laws with a Common Marginal: Standing Notation for Their Amalgamated Free Product §embeddings: N=MμN=\mathcal{M}_{\mu}, A1=MΓA_{1}=\mathcal{M}_{\Gamma}, A2=Mγk+1A_{2}=\mathcal{M}_{\gamma_{k+1}}, and π1:N→A1\pi_{1}:N\to A_{1}, π2:N→A2\pi_{2}:N\to A_{2} are the embeddings of Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §embedding for (Γ,a1)(\Gamma,a^{1}) and (γk+1,a2)(\gamma_{k+1},a^{2}). Let H\mathcal{H}, Ω\Omega and φ\varphi be the amalgamated free product space, its vacuum vector and its vacuum state (The Amalgamated Free Product Space of Two Noncommutative Laws over a Common Marginal, Its Vacuum Vector and Vacuum State §space, The Amalgamated Free Product Space of Two Noncommutative Laws over a Common Marginal, Its Vacuum Vector and Vacuum State §vectors, The Amalgamated Free Product Space of Two Noncommutative Laws over a Common Marginal, Its Vacuum Vector and Vacuum State §state), so that φ(X)=⟨Ω,XΩ⟩H\varphi(X)=\langle\Omega,X\Omega\rangle_{\mathcal{H}} for X∈L(H)X\in\mathcal{L}(\mathcal{H}), and let Λ1,Λ2\Lambda_{1},\Lambda_{2} be the actions of The Amalgamated Free Product Space: Multilinearity of the Tuple Vectors and the Bounded Left Actions of the Two Tracial Algebras §actions; The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State applies to them. All these objects are determined by (k,Γ)(k,\Gamma) through the cited constructions; no choice is made.

For j∈[d]j\in[d], Lxjμ∈NL^{\mu}_{x_{j}}\in N 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, and Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §homomorphism with Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §values gives π1(Lxjμ)=Lσa1(xj)Γ=LxjΓ\pi_{1}(L^{\mu}_{x_{j}})=L^{\Gamma}_{\sigma_{a^{1}}(x_{j})}=L^{\Gamma}_{x_{j}} and π2(Lxjμ)=Lσa2(xj)γk+1=Lxjγk+1\pi_{2}(L^{\mu}_{x_{j}})=L^{\gamma_{k+1}}_{\sigma_{a^{2}}(x_{j})}=L^{\gamma_{k+1}}_{x_{j}}. Hence The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State §amalgamation yields

Λ1(LxjΓ)=Λ1(π1(Lxjμ))=Λ2(π2(Lxjμ))=Λ2(Lxjγk+1)(j∈[d]).(2.1)\Lambda_{1}(L^{\Gamma}_{x_{j}})=\Lambda_{1}(\pi_{1}(L^{\mu}_{x_{j}}))=\Lambda_{2}(\pi_{2}(L^{\mu}_{x_{j}}))=\Lambda_{2}(L^{\gamma_{k+1}}_{x_{j}})\qquad(j\in[d]).\qquad(2.1)

Step 3 (one gluing step: the glued law). Keep the data of Step 2 and put m=(k+2)dm=(k+2)d. Define f:[m]→{1,2}f:[m]\to\{1,2\} and ci∈Af(i)c_{i}\in A_{f(i)} by f(i)=1f(i)=1, ci=LxiΓc_{i}=L^{\Gamma}_{x_{i}} for i∈[(k+1)d]i\in[(k+1)d], and f((k+1)d+j)=2f((k+1)d+j)=2, c(k+1)d+j=Lxd+jγk+1c_{(k+1)d+j}=L^{\gamma_{k+1}}_{x_{d+j}} for j∈[d]j\in[d]; these operators lie in A1A_{1}, respectively A2A_{2}, 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. Let TT be the mm-tuple in L(H)\mathcal{L}(\mathcal{H}) with Ti=Λf(i)(ci)T_{i}=\Lambda_{f(i)}(c_{i}).

(a) Self-adjointness and norms. By Notation, ci∗=cic_{i}^{*}=c_{i}, so Ti∗=Λf(i)(ci∗)=TiT_{i}^{*}=\Lambda_{f(i)}(c_{i}^{*})=T_{i} by The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State §representation; by Adjoint of a Linear Map between Complex Inner Product Spaces §adjoint this means ⟨Tiξ,η⟩=⟨ξ,Tiη⟩\langle T_{i}\xi,\eta\rangle=\langle\xi,T_{i}\eta\rangle for all ξ,η∈H\xi,\eta\in\mathcal{H}, that is, TiT_{i} is self-adjoint in the sense of Self-Adjoint Operator. Since Γ∈Σ(k+1)d,R\Gamma\in\Sigma_{(k+1)d,R} and γk+1∈Σ2d,R\gamma_{k+1}\in\Sigma_{2d,R}, Left and Right Multiplication Operators on the Complex GNS Space of a Noncommutative Law: Boundedness, Algebra Rules, Adjoints, Commutation, the Vacuum and the Conjugation §multiplication (with (k+1)d(k+1)d, respectively 2d2d, and RR in place of its dd and rr) gives ∥ci∥op≤R\lVert c_{i}\rVert_{\mathrm{op}}\le R, so ∥Ti∥op≤∥ci∥op≤R\lVert T_{i}\rVert_{\mathrm{op}}\le\lVert c_{i}\rVert_{\mathrm{op}}\le R by The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State §representation, for every i∈[m]i\in[m].

(b) Traciality. We show φ(TuTv)=φ(TvTu)\varphi(T_{u}T_{v})=\varphi(T_{v}T_{u}) for all u,v∈Wmu,v\in W_{m}. If u=∅u=\varnothing, then Tu=IT_{u}=I by Evaluation of Noncommutative Polynomials at a Tuple of Bounded Operators §word-products and both sides equal φ(Tv)\varphi(T_{v}); likewise if v=∅v=\varnothing. For nonempty u,vu,v, TuTv=TuvT_{u}T_{v}=T_{uv} and TvTu=TvuT_{v}T_{u}=T_{vu} by Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §values, so it suffices to prove

φ(Tuv)=φ(Tvu)for all nonempty u,v∈Wm.(3.1)\varphi(T_{uv})=\varphi(T_{vu})\qquad\text{for all nonempty }u,v\in W_{m}.\qquad(3.1)

First, for a letter (i)(i) and a nonempty word vv of length ll with letters v1,…,vlv_{1},\dots,v_{l}, we have φ(T(i)v)=φ(Tv(i))\varphi(T_{(i)v})=\varphi(T_{v(i)}) (3.2). Indeed, by Evaluation of Noncommutative Polynomials at a Tuple of Bounded Operators §word-products and Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §values, T(i)v=TiTv1⋯TvlT_{(i)v}=T_{i}T_{v_{1}}\cdots T_{v_{l}} and Tv(i)=Tv1⋯TvlTiT_{v(i)}=T_{v_{1}}\cdots T_{v_{l}}T_{i}, products of the same l+1≥2l+1\ge2 factors Λf(⋅)(c⋅)\Lambda_{f(\cdot)}(c_{\cdot}) with the first moved to the end, so (3.2) is The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State §trace with r=l+1r=l+1. We prove (3.1) by induction on the length l′∈Nl'\in\mathbb{N} of uu, for all nonempty vv simultaneously. If l′=1l'=1, then uu is a letter by Basic Properties of Words: Associativity, Reversal, Finitely Many Factorisations, and Countability §last-letter and (3.1) is (3.2). Suppose (3.1) holds for all uu of length l′l', and let uu have length l′+1l'+1. By Basic Properties of Words: Associativity, Reversal, Finitely Many Factorisations, and Countability §last-letter, u=u′(i)u=u'(i) with u′u' of length l′l'; by Basic Properties of Words: Associativity, Reversal, Finitely Many Factorisations, and Countability §monoid, uv=u′((i)v)uv=u'((i)v), ((i)v)u′=(i)(vu′)((i)v)u'=(i)(vu') and (vu′)(i)=vu(vu')(i)=vu, the words (i)v(i)v and vu′vu' being nonempty. The induction hypothesis (for u′u' and (i)v(i)v) and (3.2) (for ii and vu′vu') give

φ(Tuv)=φ(Tu′((i)v))=φ(T(i)(vu′))=φ(T(vu′)(i))=φ(Tvu).\varphi(T_{uv})=\varphi(T_{u'((i)v)})=\varphi(T_{(i)(vu')})=\varphi(T_{(vu')(i)})=\varphi(T_{vu}).

(c) The law. By The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State §marginals, ∥Ω∥H=1\lVert\Omega\rVert_{\mathcal{H}}=1. By (a), (b) and The Law of a Tuple of Bounded Self-Adjoint Operators in a Tracial Vector State §law (with mm, H\mathcal{H}, TT in place of its nn, HH, TT), the map

Γ′:Pm→C,Γ′(p)=⟨Ω,p(T)Ω⟩H=φ(p(T)),\Gamma':\mathcal{P}_{m}\to\mathbb{C},\qquad\Gamma'(p)=\langle\Omega,p(T)\Omega\rangle_{\mathcal{H}}=\varphi(p(T)),

belongs to Σ(k+2)d,R\Sigma_{(k+2)d,R}. It is determined by (k,Γ)(k,\Gamma).

Step 4 (one gluing step: the marginals of Γ′\Gamma'). (a) Let p∈P(k+1)dp\in\mathcal{P}_{(k+1)d}. By Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §substitution (for the tuple (x1,…,x(k+1)d)(x_{1},\dots,x_{(k+1)d}) of βk\beta_{k} and the tuple TT) and Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §values, (βkp)(T)=p(T1,…,T(k+1)d)=p(Λ1(Lx1Γ),…,Λ1(Lx(k+1)dΓ))(\beta_{k}p)(T)=p(T_{1},\dots,T_{(k+1)d})=p(\Lambda_{1}(L^{\Gamma}_{x_{1}}),\dots,\Lambda_{1}(L^{\Gamma}_{x_{(k+1)d}})). Apply Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §transport with HΓ\mathcal{H}_{\Gamma}, H\mathcal{H}, (k+1)d(k+1)d, (Lx1Γ,…,Lx(k+1)dΓ)(L^{\Gamma}_{x_{1}},\dots,L^{\Gamma}_{x_{(k+1)d}}), A1A_{1}, Λ1\Lambda_{1} in place of its HH, KK, nn, TT, A\mathcal{A}, Φ\Phi: the set A1A_{1} contains II and these operators and is closed under sums, complex multiples and composition 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, and Λ1\Lambda_{1} is linear, multiplicative and unital by The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State §representation. With Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §gns (for Γ∈Σ(k+1)d\Gamma\in\Sigma_{(k+1)d}), which gives p(Lx1Γ,…,Lx(k+1)dΓ)=LpΓp(L^{\Gamma}_{x_{1}},\dots,L^{\Gamma}_{x_{(k+1)d}})=L^{\Gamma}_{p}, this yields (βkp)(T)=Λ1(LpΓ)(\beta_{k}p)(T)=\Lambda_{1}(L^{\Gamma}_{p}). By The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State §marginals 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, Γ′(βkp)=φ(Λ1(LpΓ))=τΓ(LpΓ)=Γ(p)\Gamma'(\beta_{k}p)=\varphi(\Lambda_{1}(L^{\Gamma}_{p}))=\tau_{\Gamma}(L^{\Gamma}_{p})=\Gamma(p). Thus Γ′∘βk=Γ\Gamma'\circ\beta_{k}=\Gamma.

(b) Let q∈P2dq\in\mathcal{P}_{2d}. The tuple of δk+1\delta_{k+1} is (x1,…,xd,x(k+1)d+1,…,x(k+2)d)(x_{1},\dots,x_{d},x_{(k+1)d+1},\dots,x_{(k+2)d}), so as in (a), (δk+1q)(T)=q(T1,…,Td,T(k+1)d+1,…,T(k+2)d)(\delta_{k+1}q)(T)=q(T_{1},\dots,T_{d},T_{(k+1)d+1},\dots,T_{(k+2)d}). By (2.1), Tj=Λ1(LxjΓ)=Λ2(Lxjγk+1)T_{j}=\Lambda_{1}(L^{\Gamma}_{x_{j}})=\Lambda_{2}(L^{\gamma_{k+1}}_{x_{j}}), and by definition T(k+1)d+j=Λ2(Lxd+jγk+1)T_{(k+1)d+j}=\Lambda_{2}(L^{\gamma_{k+1}}_{x_{d+j}}), for j∈[d]j\in[d]. So this 2d2d-tuple is (Λ2(Lx1γk+1),…,Λ2(Lx2dγk+1))(\Lambda_{2}(L^{\gamma_{k+1}}_{x_{1}}),\dots,\Lambda_{2}(L^{\gamma_{k+1}}_{x_{2d}})), and exactly as in (a), with Hγk+1\mathcal{H}_{\gamma_{k+1}}, 2d2d, A2A_{2}, Λ2\Lambda_{2}, γk+1\gamma_{k+1} in place of HΓ\mathcal{H}_{\Gamma}, (k+1)d(k+1)d, A1A_{1}, Λ1\Lambda_{1}, Γ\Gamma,

Γ′(δk+1q)=φ(Λ2(Lqγk+1))=τγk+1(Lqγk+1)=γk+1(q).\Gamma'(\delta_{k+1}q)=\varphi(\Lambda_{2}(L^{\gamma_{k+1}}_{q}))=\tau_{\gamma_{k+1}}(L^{\gamma_{k+1}}_{q})=\gamma_{k+1}(q).

Thus Γ′∘δk+1=γk+1\Gamma'\circ\delta_{k+1}=\gamma_{k+1}, and with Step 3(c), Γ′\Gamma' satisfies (Bk+1)(B_{k+1}).

Step 5 (the consistent laws, by recursion on k∈Nk\in\mathbb{N}). Let X=⋃n∈NΣnX=\bigcup_{n\in\mathbb{N}}\Sigma_{n} and define F:N×X→XF:\mathbb{N}\times X\to X by F(k,Γ)=Γ′F(k,\Gamma)=\Gamma', the law of Step 3(c) built from (k,Γ)(k,\Gamma), if Γ\Gamma satisfies (Bk)(B_{k}), and F(k,Γ)=ΓF(k,\Gamma)=\Gamma otherwise; this is a map into XX since Γ′∈Σ(k+2)d,R⊆Σ(k+2)d\Gamma'\in\Sigma_{(k+2)d,R}\subseteq\Sigma_{(k+2)d} (Noncommutative Laws of Finitely Many Self-Adjoint Variables with a Norm Bound §law) and Γ′\Gamma' is determined by (k,Γ)(k,\Gamma), so no choice is involved. As γ1∈Σ2d⊆X\gamma_{1}\in\Sigma_{2d}\subseteq X by Step 1, Definition of Sequences by Recursion on the Natural Numbers §recursion gives a sequence (Γk)k∈N(\Gamma_{k})_{k\in\mathbb{N}} in XX with Γ1=γ1\Gamma_{1}=\gamma_{1} and Γk+1=F(k,Γk)\Gamma_{k+1}=F(k,\Gamma_{k}) for every k∈Nk\in\mathbb{N}. By induction on k∈Nk\in\mathbb{N}, Γk\Gamma_{k} satisfies (Bk)(B_{k}): γ1∈Σ2d,R\gamma_{1}\in\Sigma_{2d,R} by Step 1, and δ1=σ(x1,…,x2d)\delta_{1}=\sigma_{(x_{1},\dots,x_{2d})} is the identity of P2d\mathcal{P}_{2d} by Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §identity, so Γ1∘δ1=γ1\Gamma_{1}\circ\delta_{1}=\gamma_{1}; and if Γk\Gamma_{k} satisfies (Bk)(B_{k}), then Γk+1\Gamma_{k+1} is the law Γ′\Gamma' built from (k,Γk)(k,\Gamma_{k}), which satisfies (Bk+1)(B_{k+1}) by Step 4(b). Consequently, for every k∈Nk\in\mathbb{N}, Step 4(a) applies to (k,Γk)(k,\Gamma_{k}), and

Γk∈Σ(k+1)d,R,Γk∘δk=γk,Γk+1∘βk=Γk.(5.1)\Gamma_{k}\in\Sigma_{(k+1)d,R},\qquad\Gamma_{k}\circ\delta_{k}=\gamma_{k},\qquad\Gamma_{k+1}\circ\beta_{k}=\Gamma_{k}.\qquad(5.1)

Step 6 (a chain of tracial W-probability spaces).* For k∈Nk\in\mathbb{N} write Hk=HΓkH_{k}=\mathcal{H}_{\Gamma_{k}}, Mk=MΓkM_{k}=\mathcal{M}_{\Gamma_{k}}, Ωk=ΩΓk\Omega_{k}=\Omega_{\Gamma_{k}} and Lp(k)=LpΓkL^{(k)}_{p}=L^{\Gamma_{k}}_{p} for p∈P(k+1)dp\in\mathcal{P}_{(k+1)d}. As Γk∈Σ(k+1)d\Gamma_{k}\in\Sigma_{(k+1)d}, The Tracial Algebra of a Noncommutative Law is a Tracial W*-Probability Space: the W*-Closure of the Left Multiplications §w-star shows that (Hk,Mk,Ωk)(H_{k},M_{k},\Omega_{k}) is a tracial W*-probability space with trace τΓk\tau_{\Gamma_{k}}. Let θk:Mk→Mk+1\theta_{k}:M_{k}\to M_{k+1} be the map π\pi of Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §embedding with (k+2)d(k+2)d, (k+1)d(k+1)d, Γk+1\Gamma_{k+1} and the tuple (x1,…,x(k+1)d)(x_{1},\dots,x_{(k+1)d}) in P(k+2)d,sa\mathcal{P}_{(k+2)d,\mathrm{sa}} in place of its mm, nn, γ\gamma and aa; then its σa\sigma_{a} is βk\beta_{k}, and its marginal law γ∘σa\gamma\circ\sigma_{a} is Γk+1∘βk=Γk\Gamma_{k+1}\circ\beta_{k}=\Gamma_{k} by (5.1), so θk\theta_{k} indeed maps MΓk=Mk\mathcal{M}_{\Gamma_{k}}=M_{k} into MΓk+1=Mk+1\mathcal{M}_{\Gamma_{k+1}}=M_{k+1}. By Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §homomorphism, θk\theta_{k} is linear, θk(I)=I\theta_{k}(I)=I, θk(ST)=θk(S)θk(T)\theta_{k}(ST)=\theta_{k}(S)\theta_{k}(T), θk(T∗)=θk(T)∗\theta_{k}(T^{*})=\theta_{k}(T)^{*} and τΓk+1(θk(T))=τΓk(T)\tau_{\Gamma_{k+1}}(\theta_{k}(T))=\tau_{\Gamma_{k}}(T) for S,T∈MkS,T\in M_{k}; so, the traces being τΓk+1\tau_{\Gamma_{k+1}} and τΓk\tau_{\Gamma_{k}}, θk\theta_{k} is a trace-preserving embedding of (Hk,Mk,Ωk)(H_{k},M_{k},\Omega_{k}) into (Hk+1,Mk+1,Ωk+1)(H_{k+1},M_{k+1},\Omega_{k+1}) in the sense of Trace-Preserving Embeddings of Tracial W*-Probability Spaces, Their Implementing Isometries and Conditional Expectations §embedding. The same clause and Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §values give

θk(Lxi(k))=Lβk(xi)(k+1)=Lxi(k+1)(i∈[(k+1)d]).(6.1)\theta_{k}(L^{(k)}_{x_{i}})=L^{(k+1)}_{\beta_{k}(x_{i})}=L^{(k+1)}_{x_{i}}\qquad(i\in[(k+1)d]).\qquad(6.1)

Step 7 (the inductive limit and the tuples). Apply Inductive Limits of Tracial W*-Probability Spaces along Trace-Preserving Embeddings with θk\theta_{k} in place of its πk\pi_{k}: this gives a tracial W*-probability space (H,M,Ω)(H,M,\Omega) and trace-preserving embeddings ρk\rho_{k} of (Hk,Mk,Ωk)(H_{k},M_{k},\Omega_{k}) into (H,M,Ω)(H,M,\Omega) with ρk+1(θk(S))=ρk(S)\rho_{k+1}(\theta_{k}(S))=\rho_{k}(S) for k∈Nk\in\mathbb{N} and S∈MkS\in M_{k} (Inductive Limits of Tracial W*-Probability Spaces along Trace-Preserving Embeddings §compatible). Let k∈Nk\in\mathbb{N} and i∈[(k+1)d]i\in[(k+1)d]; we show by induction on n∈Nn\in\mathbb{N} with n≥kn\ge k that

ρn(Lxi(n))=ρk(Lxi(k)).(7.1)\rho_{n}(L^{(n)}_{x_{i}})=\rho_{k}(L^{(k)}_{x_{i}}).\qquad(7.1)

For n=kn=k this is trivial. If it holds for nn, then i≤(k+1)d≤(n+1)di\le(k+1)d\le(n+1)d, so by (6.1) and compatibility ρn+1(Lxi(n+1))=ρn+1(θn(Lxi(n)))=ρn(Lxi(n))=ρk(Lxi(k))\rho_{n+1}(L^{(n+1)}_{x_{i}})=\rho_{n+1}(\theta_{n}(L^{(n)}_{x_{i}}))=\rho_{n}(L^{(n)}_{x_{i}})=\rho_{k}(L^{(k)}_{x_{i}}). Define, for j∈[d]j\in[d] and k∈Nk\in\mathbb{N},

sj=ρ1(Lxj(1)),tjk=ρk(Lxkd+j(k)),s_{j}=\rho_{1}(L^{(1)}_{x_{j}}),\qquad t^{k}_{j}=\rho_{k}(L^{(k)}_{x_{kd+j}}),

and s=(s1,…,sd)s=(s_{1},\dots,s_{d}), tk=(t1k,…,tdk)t^{k}=(t^{k}_{1},\dots,t^{k}_{d}); here kd+j∈[(k+1)d]kd+j\in[(k+1)d]. These are elements of MM. By Trace-Preserving Embeddings Preserve Operator Norms and Compose §norm and Left and Right Multiplication Operators on the Complex GNS Space of a Noncommutative Law: Boundedness, Algebra Rules, Adjoints, Commutation, the Vacuum and the Conjugation §multiplication (for Γk∈Σ(k+1)d,R\Gamma_{k}\in\Sigma_{(k+1)d,R}, by (5.1)), ∥sj∥op=∥Lxj(1)∥op≤R\lVert s_{j}\rVert_{\mathrm{op}}=\lVert L^{(1)}_{x_{j}}\rVert_{\mathrm{op}}\le R and ∥tjk∥op=∥Lxkd+j(k)∥op≤R\lVert t^{k}_{j}\rVert_{\mathrm{op}}=\lVert L^{(k)}_{x_{kd+j}}\rVert_{\mathrm{op}}\le R.

Step 8 (the laws). Fix k∈Nk\in\mathbb{N} and let uku^{k} be the 2d2d-tuple (Lx1(k),…,Lxd(k),Lxkd+1(k),…,Lx(k+1)d(k))(L^{(k)}_{x_{1}},\dots,L^{(k)}_{x_{d}},L^{(k)}_{x_{kd+1}},\dots,L^{(k)}_{x_{(k+1)d}}). Its entries lie in MkM_{k} 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 and are self-adjoint by Notation, so uku^{k} is a self-adjoint 2d2d-tuple in MkM_{k} (Self-Adjoint Tuples in a Tracial W*-Probability Space and Their Laws §tuple). By (7.1) (with 11, jj, kk in place of kk, ii, nn), sj=ρk(Lxj(k))s_{j}=\rho_{k}(L^{(k)}_{x_{j}}) for j∈[d]j\in[d]; hence (ρk(u1k),…,ρk(u2dk))=(s1,…,sd,t1k,…,tdk)(\rho_{k}(u^{k}_{1}),\dots,\rho_{k}(u^{k}_{2d}))=(s_{1},\dots,s_{d},t^{k}_{1},\dots,t^{k}_{d}). By Laws of Self-Adjoint Tuples in a Tracial W*-Probability Space: Moments, Affine Images, Couplings, Embeddings and L^2 Approximation §embedding (with 2d2d, (Hk,Mk,Ωk)(H_{k},M_{k},\Omega_{k}), (H,M,Ω)(H,M,\Omega), ρk\rho_{k}, uku^{k} in place of its dd, (H,M,Ω)(H,M,\Omega), (K,N,Ψ)(K,N,\Psi), π\pi, ss), this is a self-adjoint 2d2d-tuple in MM with law λuk\lambda_{u^{k}}. In particular every sjs_{j} and tjkt^{k}_{j} is a self-adjoint element of MM, so ss and tkt^{k} are self-adjoint dd-tuples in MM, and the 2d2d-tuple (s,tk)(s,t^{k}) of Laws of Self-Adjoint Tuples in a Tracial W*-Probability Space: Moments, Affine Images, Couplings, Embeddings and L^2 Approximation §coupling is this tuple; thus λ(s,tk)=λuk\lambda_{(s,t^{k})}=\lambda_{u^{k}}.

Let ee be the 2d2d-tuple (x1,…,xd,xkd+1,…,x(k+1)d)(x_{1},\dots,x_{d},x_{kd+1},\dots,x_{(k+1)d}) of δk\delta_{k} and Lx(k)=(Lx1(k),…,Lx(k+1)d(k))L^{(k)}_{x}=(L^{(k)}_{x_{1}},\dots,L^{(k)}_{x_{(k+1)d}}). By Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §values, e(Lx(k))=uke(L^{(k)}_{x})=u^{k}, so for p∈P2dp\in\mathcal{P}_{2d}, Self-Adjoint Tuples in a Tracial W*-Probability Space and Their Laws §law, Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §substitution, The Tracial Algebra of a Noncommutative Law is a Tracial W*-Probability Space: the W*-Closure of the Left Multiplications §law (for Γk∈Σ(k+1)d\Gamma_{k}\in\Sigma_{(k+1)d}) and (5.1) give

λuk(p)=⟨Ωk,p(uk)Ωk⟩=⟨Ωk,(δkp)(Lx(k))Ωk⟩=Γk(δkp)=γk(p).\lambda_{u^{k}}(p)=\langle\Omega_{k},p(u^{k})\Omega_{k}\rangle=\langle\Omega_{k},(\delta_{k}p)(L^{(k)}_{x})\Omega_{k}\rangle=\Gamma_{k}(\delta_{k}p)=\gamma_{k}(p).

Hence λ(s,tk)=γk\lambda_{(s,t^{k})}=\gamma_{k}. By Laws of Self-Adjoint Tuples in a Tracial W*-Probability Space: Moments, Affine Images, Couplings, Embeddings and L^2 Approximation §coupling, λ(s,tk)∈Π(λs,λtk)\lambda_{(s,t^{k})}\in\Pi(\lambda_{s},\lambda_{t^{k}}), so by Couplings of Two Noncommutative Laws and Their Quadratic Cost §coupling, applied to this coupling and to γk∈Π(μ,νk)\gamma_{k}\in\Pi(\mu,\nu_{k}),

λs=λ(s,tk)∘ι1=γk∘ι1=μ,λtk=λ(s,tk)∘ι2=γk∘ι2=νk.\lambda_{s}=\lambda_{(s,t^{k})}\circ\iota^{1}=\gamma_{k}\circ\iota^{1}=\mu,\qquad\lambda_{t^{k}}=\lambda_{(s,t^{k})}\circ\iota^{2}=\gamma_{k}\circ\iota^{2}=\nu_{k}.

Together with the norm bounds of Step 7, this proves the claim.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…