Each result cited below is universally quantified over the data in its own statement. We work in the setting of Noncommutative Laws, Couplings and the Wasserstein Distance: Standing Notation , with μ , ν ∈ Σ d \mu,\nu\in\Sigma_{d} μ , ν ∈ Σ d as in the statement, and use the following elementary facts throughout. Every variable satisfies x i ∗ = x i x_{i}^{*}=x_{i} x i ∗ = x i by Noncommutative Polynomials Form a Unital Complex Algebra with Involution: Linear Extension from Monomials, Products, Adjoints and Self-Adjoint Parts §adjoint , so the variables lie in the self-adjoint part by Noncommutative Polynomials Form a Unital Complex Algebra with Involution: Linear Extension from Monomials, Products, Adjoints and Self-Adjoint Parts §self-adjoint . The inner product is linear in its second and conjugate-linear in its first argument by Complex Hilbert Spaces and Bounded Linear Maps: Standing Notation §spaces ; for a bounded linear map T T T with adjoint T ∗ T^{*} T ∗ we have ⟨ T ∗ ξ , ξ ′ ⟩ = ⟨ ξ , T ξ ′ ⟩ \langle T^{*}\xi,\xi'\rangle=\langle\xi,T\xi'\rangle ⟨ T ∗ ξ , ξ ′ ⟩ = ⟨ ξ , T ξ ′ ⟩ , hence also ⟨ T ξ ′ , ξ ⟩ = ⟨ ξ ′ , T ∗ ξ ⟩ \langle T\xi',\xi\rangle=\langle\xi',T^{*}\xi\rangle ⟨ T ξ ′ , ξ ⟩ = ⟨ ξ ′ , T ∗ ξ ⟩ by conjugate symmetry. For k ∈ N k\in\mathbb{N} k ∈ N , a real r > 0 r>0 r > 0 , λ ∈ Σ k , r \lambda\in\Sigma_{k,r} λ ∈ Σ k , r and p , q ∈ P k p,q\in\mathcal{P}_{k} p , q ∈ P k , Left and Right Multiplication Operators on the Complex GNS Space of a Noncommutative Law: Boundedness, Algebra Rules, Adjoints, Commutation, the Vacuum and the Conjugation §vacuum gives
L p Ω λ = p ^ λ , ⟨ p ^ λ , q ^ λ ⟩ = λ ( p ∗ q ) , ⟨ Ω λ , L p Ω λ ⟩ = λ ( p ) ; L_{p}\Omega_{\lambda}=\widehat{p}^{\,\lambda},\qquad\langle\widehat{p}^{\,\lambda},\widehat{q}^{\,\lambda}\rangle=\lambda(p^{*}q),\qquad\langle\Omega_{\lambda},L_{p}\Omega_{\lambda}\rangle=\lambda(p); L p Ω λ = p λ , ⟨ p λ , q λ ⟩ = λ ( p ∗ q ) , ⟨ Ω λ , L p Ω λ ⟩ = λ ( p ) ;
for T ∈ M λ T\in\mathcal{M}_{\lambda} T ∈ M λ , τ λ ( T ) = ⟨ Ω λ , T Ω λ ⟩ \tau_{\lambda}(T)=\langle\Omega_{\lambda},T\Omega_{\lambda}\rangle τ λ ( T ) = ⟨ Ω λ , T Ω λ ⟩ by The Tracial Algebra of a Noncommutative Law and Its Trace §trace ; τ λ \tau_{\lambda} τ λ is linear with τ λ ( L p ) = λ ( p ) \tau_{\lambda}(L_{p})=\lambda(p) τ λ ( L p ) = λ ( p ) 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 ; and M λ \mathcal{M}_{\lambda} M λ contains I I I and every L p L_{p} L p and is closed under sums, complex multiples, composition and adjoints 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 . For a tuple T T T of bounded operators, evaluation p ↦ p ( T ) p\mapsto p(T) p ↦ p ( T ) is linear by Evaluation of Noncommutative Polynomials at a Tuple of Bounded Operators §evaluation , multiplicative by Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §homomorphism , and 1 ( T ) = I 1(T)=I 1 ( T ) = I , x i ( T ) = T i x_{i}(T)=T_{i} x i ( T ) = T i by Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §values .
Step 0 (Norm bounds, the marginal isometry and the barycentric vectors). By Noncommutative Laws of Finitely Many Self-Adjoint Variables with a Norm Bound §law there are reals R 1 , R 2 > 0 R_{1},R_{2}>0 R 1 , R 2 > 0 with μ ∈ Σ d , R 1 \mu\in\Sigma_{d,R_{1}} μ ∈ Σ d , R 1 and ν ∈ Σ d , R 2 \nu\in\Sigma_{d,R_{2}} ν ∈ Σ d , R 2 ; with R = max { R 1 , R 2 } R=\max\{R_{1},R_{2}\} R = max { R 1 , R 2 } , Basic Properties of Noncommutative Laws: Adjoints, the Self-Adjoint Pairing, Cauchy-Schwarz, Monotonicity in the Bound, and Laws of Constant Tuples §monotone gives μ , ν ∈ Σ d , R \mu,\nu\in\Sigma_{d,R} μ , ν ∈ Σ d , R . We fix this R R R .
Fix γ ∈ Π ( μ , ν ) \gamma\in\Pi(\mu,\nu) γ ∈ Π ( μ , ν ) and abbreviate z j = z j γ z_{j}=z^{\gamma}_{j} z j = z j γ , z = ( z 1 , … , z d ) z=(z_{1},\dots,z_{d}) z = ( z 1 , … , z d ) , ζ = ζ γ \zeta=\zeta_{\gamma} ζ = ζ γ , β = β γ \beta=\beta_{\gamma} β = β γ and E = E γ E=E_{\gamma} E = E γ ; the operators L p L_{p} L p act on H γ \mathcal{H}_{\gamma} H γ for p ∈ P 2 d p\in\mathcal{P}_{2d} p ∈ P 2 d and on H μ \mathcal{H}_{\mu} H μ for p ∈ P d p\in\mathcal{P}_{d} p ∈ P d , as in the statement. By Couplings of Noncommutative Laws: the Norm Bound, the Cost Identity, the Tensor, Diagonal and Swapped Couplings, Weak-Star Closedness, and Displacement Interpolants §bound , γ ∈ Σ 2 d , R \gamma\in\Sigma_{2d,R} γ ∈ Σ 2 d , R , so γ ∈ Σ 2 d \gamma\in\Sigma_{2d} γ ∈ Σ 2 d by Noncommutative Laws of Finitely Many Self-Adjoint Variables with a Norm Bound §law . Let a = ( x 1 , … , x d ) a=(x_{1},\dots,x_{d}) a = ( x 1 , … , x d ) and a ′ = ( x d + 1 , … , x 2 d ) a'=(x_{d+1},\dots,x_{2d}) a ′ = ( x d + 1 , … , x 2 d ) , d d d -tuples in P 2 d , s a \mathcal{P}_{2d,\mathrm{sa}} P 2 d , sa . By Couplings of Two Noncommutative Laws and Their Quadratic Cost §marginals , σ a = ι 1 \sigma_{a}=\iota^{1} σ a = ι 1 and σ a ′ = ι 2 \sigma_{a'}=\iota^{2} σ a ′ = ι 2 , so Couplings of Two Noncommutative Laws and Their Quadratic Cost §coupling gives γ ∘ σ a = μ \gamma\circ\sigma_{a}=\mu γ ∘ σ a = μ and γ ∘ σ a ′ = ν \gamma\circ\sigma_{a'}=\nu γ ∘ σ a ′ = ν ; and by Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §values , σ a ( x j ) = x j \sigma_{a}(x_{j})=x_{j} σ a ( x j ) = x j and σ a ′ ( x j ) = x d + j \sigma_{a'}(x_{j})=x_{d+j} σ a ′ ( x j ) = x d + j for j ∈ [ d ] j\in[d] j ∈ [ d ] .
Let V ∈ L ( H μ , H γ ) V\in\mathcal{L}(\mathcal{H}_{\mu},\mathcal{H}_{\gamma}) V ∈ L ( H μ , H γ ) and π : M μ → M γ \pi:\mathcal{M}_{\mu}\to\mathcal{M}_{\gamma} π : M μ → M γ be the isometry and the embedding of Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation for γ \gamma γ and a a a . By Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §isometry , V ∗ V = I V^{*}V=I V ∗ V = I , V Ω μ = Ω γ V\Omega_{\mu}=\Omega_{\gamma} V Ω μ = Ω γ and V p ^ μ = ι 1 ( p ) ^ γ V\widehat{p}^{\,\mu}=\widehat{\iota^{1}(p)}^{\,\gamma} V p μ = ι 1 ( p ) γ for p ∈ P d p\in\mathcal{P}_{d} p ∈ P d ; by Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §embedding , π ( T ) ∈ M γ \pi(T)\in\mathcal{M}_{\gamma} π ( T ) ∈ M γ , π ( T ) Ω γ = V T Ω μ \pi(T)\Omega_{\gamma}=VT\Omega_{\mu} π ( T ) Ω γ = V T Ω μ and ∥ π ( T ) ∥ o p ≤ ∥ T ∥ o p \lVert\pi(T)\rVert_{\mathrm{op}}\le\lVert T\rVert_{\mathrm{op}} ∥ π ( T ) ∥ op ≤ ∥ T ∥ op for T ∈ M μ T\in\mathcal{M}_{\mu} T ∈ M μ ; by Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §homomorphism , π \pi π is linear, π ( I ) = I \pi(I)=I π ( I ) = I , π ( S T ) = π ( S ) π ( T ) \pi(ST)=\pi(S)\pi(T) π ( ST ) = π ( S ) π ( T ) , π ( T ∗ ) = π ( T ) ∗ \pi(T^{*})=\pi(T)^{*} π ( T ∗ ) = π ( T ) ∗ and τ γ ( π ( T ) ) = τ μ ( T ) \tau_{\gamma}(\pi(T))=\tau_{\mu}(T) τ γ ( π ( T )) = τ μ ( T ) ; and by Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §expectation , E ( b ) = V ∗ b V E(b)=V^{*}bV E ( b ) = V ∗ bV and ∥ E ( b ) ∥ o p ≤ ∥ b ∥ o p \lVert E(b)\rVert_{\mathrm{op}}\le\lVert b\rVert_{\mathrm{op}} ∥ E ( b ) ∥ op ≤ ∥ b ∥ op for b ∈ M γ b\in\mathcal{M}_{\gamma} b ∈ M γ . For j ∈ [ d ] j\in[d] j ∈ [ d ] put y j = x d + j ^ γ = L x d + j Ω γ ∈ H γ y_{j}=\widehat{x_{d+j}}^{\,\gamma}=L_{x_{d+j}}\Omega_{\gamma}\in\mathcal{H}_{\gamma} y j = x d + j γ = L x d + j Ω γ ∈ H γ . Then
z j Ω μ = V ∗ L x d + j V Ω μ = V ∗ L x d + j Ω γ = V ∗ y j . ( 0.1 ) z_{j}\Omega_{\mu}=V^{*}L_{x_{d+j}}V\Omega_{\mu}=V^{*}L_{x_{d+j}}\Omega_{\gamma}=V^{*}y_{j}.\qquad(0.1) z j Ω μ = V ∗ L x d + j V Ω μ = V ∗ L x d + j Ω γ = V ∗ y j . ( 0.1 )
Moreover, for every y ∈ H γ y\in\mathcal{H}_{\gamma} y ∈ H γ we have ∥ V ∗ y ∥ ≤ ∥ y ∥ \lVert V^{*}y\rVert\le\lVert y\rVert ∥ V ∗ y ∥ ≤ ∥ y ∥ : indeed ∥ V V ∗ y ∥ 2 = ⟨ V ∗ y , V ∗ V V ∗ y ⟩ = ∥ V ∗ y ∥ 2 \lVert VV^{*}y\rVert^{2}=\langle V^{*}y,V^{*}VV^{*}y\rangle=\lVert V^{*}y\rVert^{2} ∥ V V ∗ y ∥ 2 = ⟨ V ∗ y , V ∗ V V ∗ y ⟩ = ∥ V ∗ y ∥ 2 because V ∗ V = I V^{*}V=I V ∗ V = I , and by Cauchy-Schwarz Inequality in a Complex Inner Product Space
∥ V ∗ y ∥ 2 = ⟨ V ∗ y , V ∗ y ⟩ = ∣ ⟨ V V ∗ y , y ⟩ ∣ ≤ ∥ V V ∗ y ∥ ∥ y ∥ = ∥ V ∗ y ∥ ∥ y ∥ , ( 0.2 ) \lVert V^{*}y\rVert^{2}=\langle V^{*}y,V^{*}y\rangle=\bigl|\langle VV^{*}y,y\rangle\bigr|\le\lVert VV^{*}y\rVert\,\lVert y\rVert=\lVert V^{*}y\rVert\,\lVert y\rVert,\qquad(0.2) ∥ V ∗ y ∥ 2 = ⟨ V ∗ y , V ∗ y ⟩ = ⟨ V V ∗ y , y ⟩ ≤ ∥ V V ∗ y ∥ ∥ y ∥ = ∥ V ∗ y ∥ ∥ y ∥ , ( 0.2 )
so ∥ V ∗ y ∥ ≤ ∥ y ∥ \lVert V^{*}y\rVert\le\lVert y\rVert ∥ V ∗ y ∥ ≤ ∥ y ∥ (trivially if V ∗ y = 0 V^{*}y=0 V ∗ y = 0 , and after division by ∥ V ∗ y ∥ \lVert V^{*}y\rVert ∥ V ∗ y ∥ otherwise).
Step 1 (Claim 1, coupling). Let S > 0 S>0 S > 0 be real with ν ∈ Σ d , S \nu\in\Sigma_{d,S} ν ∈ Σ d , S . Let π ′ : M ν → M γ \pi':\mathcal{M}_{\nu}\to\mathcal{M}_{\gamma} π ′ : M ν → M γ be the embedding of Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation for γ \gamma γ and a ′ a' a ′ (that lemma's μ \mu μ being our ν = γ ∘ σ a ′ \nu=\gamma\circ\sigma_{a'} ν = γ ∘ σ a ′ ), and write L p ν L^{\nu}_{p} L p ν for the multiplication operators on H ν \mathcal{H}_{\nu} H ν . By Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §homomorphism , π ′ ( L x j ν ) = L σ a ′ ( x j ) = L x d + j \pi'(L^{\nu}_{x_{j}})=L_{\sigma_{a'}(x_{j})}=L_{x_{d+j}} π ′ ( L x j ν ) = L σ a ′ ( x j ) = L x d + j ; by Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §embedding , π ′ \pi' π ′ does not increase operator norms; and 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 for ν ∈ Σ d , S \nu\in\Sigma_{d,S} ν ∈ Σ d , S , ∥ L x j ν ∥ o p ≤ S \lVert L^{\nu}_{x_{j}}\rVert_{\mathrm{op}}\le S ∥ L x j ν ∥ op ≤ S . With the contraction property of E E E from Step 0,
∥ z j ∥ o p = ∥ E ( L x d + j ) ∥ o p ≤ ∥ L x d + j ∥ o p = ∥ π ′ ( L x j ν ) ∥ o p ≤ ∥ L x j ν ∥ o p ≤ S ( j ∈ [ d ] ) . ( 1.1 ) \lVert z_{j}\rVert_{\mathrm{op}}=\lVert E(L_{x_{d+j}})\rVert_{\mathrm{op}}\le\lVert L_{x_{d+j}}\rVert_{\mathrm{op}}=\lVert\pi'(L^{\nu}_{x_{j}})\rVert_{\mathrm{op}}\le\lVert L^{\nu}_{x_{j}}\rVert_{\mathrm{op}}\le S\qquad(j\in[d]).\qquad(1.1) ∥ z j ∥ op = ∥ E ( L x d + j ) ∥ op ≤ ∥ L x d + j ∥ op = ∥ π ′ ( L x j ν ) ∥ op ≤ ∥ L x j ν ∥ op ≤ S ( j ∈ [ d ]) . ( 1.1 )
By the statement, z z z is a self-adjoint d d d -tuple in M μ \mathcal{M}_{\mu} M μ , and ( H μ , M μ , Ω μ ) (\mathcal{H}_{\mu},\mathcal{M}_{\mu},\Omega_{\mu}) ( H μ , M μ , Ω μ ) is a tracial W*-probability space by The Tracial Algebra of a Noncommutative Law is a Tracial W*-Probability Space: the W*-Closure of the Left Multiplications §w-star . Hence Laws of Self-Adjoint Tuples in a Tracial W*-Probability Space: Moments, Affine Images, Couplings, Embeddings and L^2 Approximation §law and (1.1) give ζ = l a w ( z ) ∈ Σ d , S \zeta=\mathrm{law}(z)\in\Sigma_{d,S} ζ = law ( z ) ∈ Σ d , S . This is the last sentence of claim 1; with S = R S=R S = R it gives ζ ∈ Σ d , R \zeta\in\Sigma_{d,R} ζ ∈ Σ d , R , and so ζ ∈ Σ d \zeta\in\Sigma_{d} ζ ∈ Σ d by Noncommutative Laws of Finitely Many Self-Adjoint Variables with a Norm Bound §law .
Let L x = ( L x 1 , … , L x d ) L_{x}=(L_{x_{1}},\dots,L_{x_{d}}) L x = ( L x 1 , … , L x d ) on H μ \mathcal{H}_{\mu} H μ , a self-adjoint d d d -tuple in M μ \mathcal{M}_{\mu} M μ (statement). By The Tracial Algebra of a Noncommutative Law is a Tracial W*-Probability Space: the W*-Closure of the Left Multiplications §law , ⟨ Ω μ , p ( L x ) Ω μ ⟩ = μ ( p ) \langle\Omega_{\mu},p(L_{x})\Omega_{\mu}\rangle=\mu(p) ⟨ Ω μ , p ( L x ) Ω μ ⟩ = μ ( p ) for p ∈ P d p\in\mathcal{P}_{d} p ∈ P d , so by Self-Adjoint Tuples in a Tracial W*-Probability Space and Their Laws §law the law of L x L_{x} L x is μ \mu μ . By Laws of Self-Adjoint Tuples in a Tracial W*-Probability Space: Moments, Affine Images, Couplings, Embeddings and L^2 Approximation §coupling , applied to the tuples L x L_{x} L x and z z z , the law β \beta β of s = ( L x 1 , … , L x d , z 1 , … , z d ) s=(L_{x_{1}},\dots,L_{x_{d}},z_{1},\dots,z_{d}) s = ( L x 1 , … , L x d , z 1 , … , z d ) belongs to Π ( μ , ζ ) \Pi(\mu,\zeta) Π ( μ , ζ ) . This proves claim 1.
Step 2 (Claim 2, orthogonal splitting). Let j ∈ [ d ] j\in[d] j ∈ [ d ] . Since x j ∗ = x j x_{j}^{*}=x_{j} x j ∗ = x j and x j ^ γ = ι 1 ( x j ) ^ γ = V x j ^ μ \widehat{x_{j}}^{\,\gamma}=\widehat{\iota^{1}(x_{j})}^{\,\gamma}=V\widehat{x_{j}}^{\,\mu} x j γ = ι 1 ( x j ) γ = V x j μ (Step 0), the vacuum identities for γ ∈ Σ 2 d , R \gamma\in\Sigma_{2d,R} γ ∈ Σ 2 d , R and μ ∈ Σ d , R \mu\in\Sigma_{d,R} μ ∈ Σ d , R and (0.1) give
γ ( x j x d + j ) = ⟨ x j ^ γ , x d + j ^ γ ⟩ = ⟨ V x j ^ μ , y j ⟩ = ⟨ x j ^ μ , V ∗ y j ⟩ = ⟨ L x j Ω μ , z j Ω μ ⟩ . \gamma(x_{j}x_{d+j})=\langle\widehat{x_{j}}^{\,\gamma},\widehat{x_{d+j}}^{\,\gamma}\rangle=\langle V\widehat{x_{j}}^{\,\mu},y_{j}\rangle=\langle\widehat{x_{j}}^{\,\mu},V^{*}y_{j}\rangle=\langle L_{x_{j}}\Omega_{\mu},z_{j}\Omega_{\mu}\rangle. γ ( x j x d + j ) = ⟨ x j γ , x d + j γ ⟩ = ⟨ V x j μ , y j ⟩ = ⟨ x j μ , V ∗ y j ⟩ = ⟨ L x j Ω μ , z j Ω μ ⟩ .
By Laws of Self-Adjoint Tuples in a Tracial W*-Probability Space: Moments, Affine Images, Couplings, Embeddings and L^2 Approximation §moments for the 2 d 2d 2 d -tuple s s s of Step 1 (with indices j j j and d + j d+j d + j ), β ( x j x d + j ) = ⟨ L x j Ω μ , z j Ω μ ⟩ \beta(x_{j}x_{d+j})=\langle L_{x_{j}}\Omega_{\mu},z_{j}\Omega_{\mu}\rangle β ( x j x d + j ) = ⟨ L x j Ω μ , z j Ω μ ⟩ . Hence
γ ( x j x d + j ) = β ( x j x d + j ) ( j ∈ [ d ] ) . ( 2.1 ) \gamma(x_{j}x_{d+j})=\beta(x_{j}x_{d+j})\qquad(j\in[d]).\qquad(2.1) γ ( x j x d + j ) = β ( x j x d + j ) ( j ∈ [ d ]) . ( 2.1 )
Next, ι 2 ( x j x j ) = x d + j x d + j \iota^{2}(x_{j}x_{j})=x_{d+j}x_{d+j} ι 2 ( x j x j ) = x d + j x d + j by Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §homomorphism and Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §values , so with γ ∘ ι 2 = ν \gamma\circ\iota^{2}=\nu γ ∘ ι 2 = ν and the vacuum identities, ν ( x j 2 ) = γ ( x d + j x d + j ) = ∥ y j ∥ 2 \nu(x_{j}^{2})=\gamma(x_{d+j}x_{d+j})=\lVert y_{j}\rVert^{2} ν ( x j 2 ) = γ ( x d + j x d + j ) = ∥ y j ∥ 2 . By Laws of Self-Adjoint Tuples in a Tracial W*-Probability Space: Moments, Affine Images, Couplings, Embeddings and L^2 Approximation §moments for the tuple z z z and (0.1), ζ ( x j 2 ) = ∥ z j Ω μ ∥ 2 = ∥ V ∗ y j ∥ 2 \zeta(x_{j}^{2})=\lVert z_{j}\Omega_{\mu}\rVert^{2}=\lVert V^{*}y_{j}\rVert^{2} ζ ( x j 2 ) = ∥ z j Ω μ ∥ 2 = ∥ V ∗ y j ∥ 2 . Summing over j j j , with M M M as in Noncommutative Laws, Couplings and the Wasserstein Distance: Standing Notation §laws ,
M ( ν ) = ∑ j = 1 d ∥ y j ∥ 2 , M ( ζ ) = ∑ j = 1 d ∥ V ∗ y j ∥ 2 , ( 2.2 ) M(\nu)=\sum_{j=1}^{d}\lVert y_{j}\rVert^{2},\qquad M(\zeta)=\sum_{j=1}^{d}\lVert V^{*}y_{j}\rVert^{2},\qquad(2.2) M ( ν ) = j = 1 ∑ d ∥ y j ∥ 2 , M ( ζ ) = j = 1 ∑ d ∥ V ∗ y j ∥ 2 , ( 2.2 )
and (0.2) gives M ( ζ ) ≤ M ( ν ) M(\zeta)\le M(\nu) M ( ζ ) ≤ M ( ν ) . Finally, Couplings of Noncommutative Laws: the Norm Bound, the Cost Identity, the Tensor, Diagonal and Swapped Couplings, Weak-Star Closedness, and Displacement Interpolants §cost , applied to γ ∈ Π ( μ , ν ) \gamma\in\Pi(\mu,\nu) γ ∈ Π ( μ , ν ) and to β ∈ Π ( μ , ζ ) \beta\in\Pi(\mu,\zeta) β ∈ Π ( μ , ζ ) (with μ , ζ ∈ Σ d \mu,\zeta\in\Sigma_{d} μ , ζ ∈ Σ d by Step 1), gives
I ( γ ) = M ( μ ) + M ( ν ) − 2 ∑ j = 1 d γ ( x j x d + j ) , I ( β ) = M ( μ ) + M ( ζ ) − 2 ∑ j = 1 d β ( x j x d + j ) , I(\gamma)=M(\mu)+M(\nu)-2\sum_{j=1}^{d}\gamma(x_{j}x_{d+j}),\qquad I(\beta)=M(\mu)+M(\zeta)-2\sum_{j=1}^{d}\beta(x_{j}x_{d+j}), I ( γ ) = M ( μ ) + M ( ν ) − 2 j = 1 ∑ d γ ( x j x d + j ) , I ( β ) = M ( μ ) + M ( ζ ) − 2 j = 1 ∑ d β ( x j x d + j ) ,
and subtracting, by (2.1), I ( γ ) = I ( β ) + M ( ν ) − M ( ζ ) I(\gamma)=I(\beta)+M(\nu)-M(\zeta) I ( γ ) = I ( β ) + M ( ν ) − M ( ζ ) . This proves claim 2.
Step 3 (A second coupling of ζ \zeta ζ and ν \nu ν ). For j ∈ [ d ] j\in[d] j ∈ [ d ] let g j = π ( z j ) g_{j}=\pi(z_{j}) g j = π ( z j ) , and let s ′ = ( g 1 , … , g d , L x d + 1 , … , L x 2 d ) s'=(g_{1},\dots,g_{d},L_{x_{d+1}},\dots,L_{x_{2d}}) s ′ = ( g 1 , … , g d , L x d + 1 , … , L x 2 d ) , a 2 d 2d 2 d -tuple of operators on H γ \mathcal{H}_{\gamma} H γ . By Step 0, g j ∈ M γ g_{j}\in\mathcal{M}_{\gamma} g j ∈ M γ , g j ∗ = π ( z j ∗ ) = π ( z j ) = g j g_{j}^{*}=\pi(z_{j}^{*})=\pi(z_{j})=g_{j} g j ∗ = π ( z j ∗ ) = π ( z j ) = g j and ∥ g j ∥ o p ≤ ∥ z j ∥ o p ≤ R \lVert g_{j}\rVert_{\mathrm{op}}\le\lVert z_{j}\rVert_{\mathrm{op}}\le R ∥ g j ∥ op ≤ ∥ z j ∥ op ≤ R by (1.1) with S = R S=R S = R . The operators L x d + j L_{x_{d+j}} L x d + j are self-adjoint elements of M γ \mathcal{M}_{\gamma} M γ (statement) with ∥ L x d + j ∥ o p ≤ R \lVert L_{x_{d+j}}\rVert_{\mathrm{op}}\le R ∥ L x d + j ∥ op ≤ R by (1.1). So s ′ s' s ′ is a self-adjoint 2 d 2d 2 d -tuple in the tracial W*-probability space ( H γ , M γ , Ω γ ) (\mathcal{H}_{\gamma},\mathcal{M}_{\gamma},\Omega_{\gamma}) ( H γ , M γ , Ω γ ) of The Tracial Algebra of a Noncommutative Law is a Tracial W*-Probability Space: the W*-Closure of the Left Multiplications §w-star ; let η \eta η be its law , η ( p ) = ⟨ Ω γ , p ( s ′ ) Ω γ ⟩ \eta(p)=\langle\Omega_{\gamma},p(s')\Omega_{\gamma}\rangle η ( p ) = ⟨ Ω γ , p ( s ′ ) Ω γ ⟩ . By Laws of Self-Adjoint Tuples in a Tracial W*-Probability Space: Moments, Affine Images, Couplings, Embeddings and L^2 Approximation §law , η ∈ Σ 2 d , R \eta\in\Sigma_{2d,R} η ∈ Σ 2 d , R .
Let g = ( g 1 , … , g d ) g=(g_{1},\dots,g_{d}) g = ( g 1 , … , g d ) and h = ( L x d + 1 , … , L x 2 d ) h=(L_{x_{d+1}},\dots,L_{x_{2d}}) h = ( L x d + 1 , … , L x 2 d ) . For q ∈ P d q\in\mathcal{P}_{d} q ∈ P d , Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §transport , applied with H μ \mathcal{H}_{\mu} H μ , H γ \mathcal{H}_{\gamma} H γ , d d d , z z z , M μ \mathcal{M}_{\mu} M μ , π \pi π in place of its H H H , K K K , n n n , T T T , A \mathcal{A} A , Φ \Phi Φ (the set M μ \mathcal{M}_{\mu} M μ contains I I I and every z j z_{j} z j and is closed under sums, complex multiples and composition, and π \pi π is linear, multiplicative and unital, by Step 0), gives
q ( z ) ∈ M μ and q ( g ) = π ( q ( z ) ) . ( 3.1 ) q(z)\in\mathcal{M}_{\mu}\qquad\text{and}\qquad q(g)=\pi\bigl(q(z)\bigr).\qquad(3.1) q ( z ) ∈ M μ and q ( g ) = π ( q ( z ) ) . ( 3.1 )
Hence ⟨ Ω γ , q ( g ) Ω γ ⟩ = τ γ ( π ( q ( z ) ) ) = τ μ ( q ( z ) ) = ⟨ Ω μ , q ( z ) Ω μ ⟩ = ζ ( q ) \langle\Omega_{\gamma},q(g)\Omega_{\gamma}\rangle=\tau_{\gamma}(\pi(q(z)))=\tau_{\mu}(q(z))=\langle\Omega_{\mu},q(z)\Omega_{\mu}\rangle=\zeta(q) ⟨ Ω γ , q ( g ) Ω γ ⟩ = τ γ ( π ( q ( z ))) = τ μ ( q ( z )) = ⟨ Ω μ , q ( z ) Ω μ ⟩ = ζ ( q ) , so the law of g g g is ζ \zeta ζ . For the second block, write L x γ = ( L x 1 , … , L x 2 d ) L^{\gamma}_{x}=(L_{x_{1}},\dots,L_{x_{2d}}) L x γ = ( L x 1 , … , L x 2 d ) ; then a ′ ( L x γ ) = h a'(L^{\gamma}_{x})=h a ′ ( L x γ ) = h by Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §values , so Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §substitution and Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §gns (for γ ∈ Σ 2 d \gamma\in\Sigma_{2d} γ ∈ Σ 2 d ) give q ( h ) = ( ι 2 ( q ) ) ( L x γ ) = L ι 2 ( q ) q(h)=(\iota^{2}(q))(L^{\gamma}_{x})=L_{\iota^{2}(q)} q ( h ) = ( ι 2 ( q )) ( L x γ ) = L ι 2 ( q ) , and ⟨ Ω γ , q ( h ) Ω γ ⟩ = γ ( ι 2 ( q ) ) = ν ( q ) \langle\Omega_{\gamma},q(h)\Omega_{\gamma}\rangle=\gamma(\iota^{2}(q))=\nu(q) ⟨ Ω γ , q ( h ) Ω γ ⟩ = γ ( ι 2 ( q )) = ν ( q ) . So the law of h h h is ν \nu ν , and Laws of Self-Adjoint Tuples in a Tracial W*-Probability Space: Moments, Affine Images, Couplings, Embeddings and L^2 Approximation §coupling , applied to the tuples g g g and h h h , gives η ∈ Π ( ζ , ν ) \eta\in\Pi(\zeta,\nu) η ∈ Π ( ζ , ν ) .
Step 4 (A martingale identity for η \eta η ). We show
η ( ι 1 ( q ) x d + j ) = η ( ι 1 ( q ) x j ) ( q ∈ P d , j ∈ [ d ] ) . ( 4.1 ) \eta\bigl(\iota^{1}(q)\,x_{d+j}\bigr)=\eta\bigl(\iota^{1}(q)\,x_{j}\bigr)\qquad(q\in\mathcal{P}_{d},\ j\in[d]).\qquad(4.1) η ( ι 1 ( q ) x d + j ) = η ( ι 1 ( q ) x j ) ( q ∈ P d , j ∈ [ d ]) . ( 4.1 )
Let k ∈ [ 2 d ] k\in[2d] k ∈ [ 2 d ] . By the homomorphism and value rules of evaluation, ( ι 1 ( q ) x k ) ( s ′ ) = ( ι 1 ( q ) ) ( s ′ ) s k ′ (\iota^{1}(q)x_{k})(s')=(\iota^{1}(q))(s')\,s'_{k} ( ι 1 ( q ) x k ) ( s ′ ) = ( ι 1 ( q )) ( s ′ ) s k ′ ; and a ( s ′ ) = g a(s')=g a ( s ′ ) = g by Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §values , so Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §substitution and (3.1) give ( ι 1 ( q ) ) ( s ′ ) = q ( g ) = π ( q ( z ) ) (\iota^{1}(q))(s')=q(g)=\pi(q(z)) ( ι 1 ( q )) ( s ′ ) = q ( g ) = π ( q ( z )) . With π ( q ( z ) ) ∗ = π ( q ( z ) ∗ ) \pi(q(z))^{*}=\pi(q(z)^{*}) π ( q ( z ) ) ∗ = π ( q ( z ) ∗ ) and π ( q ( z ) ∗ ) Ω γ = V q ( z ) ∗ Ω μ \pi(q(z)^{*})\Omega_{\gamma}=Vq(z)^{*}\Omega_{\mu} π ( q ( z ) ∗ ) Ω γ = V q ( z ) ∗ Ω μ (Step 0; q ( z ) ∗ ∈ M μ q(z)^{*}\in\mathcal{M}_{\mu} q ( z ) ∗ ∈ M μ ), we get
η ( ι 1 ( q ) x k ) = ⟨ Ω γ , π ( q ( z ) ) s k ′ Ω γ ⟩ = ⟨ V q ( z ) ∗ Ω μ , s k ′ Ω γ ⟩ . \eta\bigl(\iota^{1}(q)x_{k}\bigr)=\langle\Omega_{\gamma},\pi(q(z))\,s'_{k}\Omega_{\gamma}\rangle=\langle Vq(z)^{*}\Omega_{\mu},s'_{k}\Omega_{\gamma}\rangle. η ( ι 1 ( q ) x k ) = ⟨ Ω γ , π ( q ( z )) s k ′ Ω γ ⟩ = ⟨ V q ( z ) ∗ Ω μ , s k ′ Ω γ ⟩ .
For k = d + j k=d+j k = d + j , s k ′ Ω γ = y j s'_{k}\Omega_{\gamma}=y_{j} s k ′ Ω γ = y j and, by (0.1), ⟨ V q ( z ) ∗ Ω μ , y j ⟩ = ⟨ q ( z ) ∗ Ω μ , V ∗ y j ⟩ = ⟨ q ( z ) ∗ Ω μ , z j Ω μ ⟩ \langle Vq(z)^{*}\Omega_{\mu},y_{j}\rangle=\langle q(z)^{*}\Omega_{\mu},V^{*}y_{j}\rangle=\langle q(z)^{*}\Omega_{\mu},z_{j}\Omega_{\mu}\rangle ⟨ V q ( z ) ∗ Ω μ , y j ⟩ = ⟨ q ( z ) ∗ Ω μ , V ∗ y j ⟩ = ⟨ q ( z ) ∗ Ω μ , z j Ω μ ⟩ . For k = j k=j k = j , s j ′ Ω γ = π ( z j ) Ω γ = V z j Ω μ s'_{j}\Omega_{\gamma}=\pi(z_{j})\Omega_{\gamma}=Vz_{j}\Omega_{\mu} s j ′ Ω γ = π ( z j ) Ω γ = V z j Ω μ and ⟨ V q ( z ) ∗ Ω μ , V z j Ω μ ⟩ = ⟨ q ( z ) ∗ Ω μ , V ∗ V z j Ω μ ⟩ = ⟨ q ( z ) ∗ Ω μ , z j Ω μ ⟩ \langle Vq(z)^{*}\Omega_{\mu},Vz_{j}\Omega_{\mu}\rangle=\langle q(z)^{*}\Omega_{\mu},V^{*}Vz_{j}\Omega_{\mu}\rangle=\langle q(z)^{*}\Omega_{\mu},z_{j}\Omega_{\mu}\rangle ⟨ V q ( z ) ∗ Ω μ , V z j Ω μ ⟩ = ⟨ q ( z ) ∗ Ω μ , V ∗ V z j Ω μ ⟩ = ⟨ q ( z ) ∗ Ω μ , z j Ω μ ⟩ . The two values agree, proving (4.1).
Step 5 (Gluing freely over ζ \zeta ζ : the data). By Step 1, ζ , μ ∈ Σ d , R \zeta,\mu\in\Sigma_{d,R} ζ , μ ∈ Σ d , R , so The Noncommutative Wasserstein Distance: Existence of Optimal Couplings, Symmetry, Separation, a Moment Bound, Weak-Star Lower Semicontinuity, and Displacement Interpolation §attained gives a coupling α ∈ Π ( ζ , μ ) \alpha\in\Pi(\zeta,\mu) α ∈ Π ( ζ , μ ) that is optimal, i.e. by The Noncommutative Quadratic Wasserstein Distance and Optimal Couplings §optimal and The Noncommutative Wasserstein Distance: Existence of Optimal Couplings, Symmetry, Separation, a Moment Bound, Weak-Star Lower Semicontinuity, and Displacement Interpolation §symmetry ,
I ( α ) = W 2 ( ζ , μ ) 2 = W 2 ( μ , ζ ) 2 . ( 5.1 ) I(\alpha)=W_{2}(\zeta,\mu)^{2}=W_{2}(\mu,\zeta)^{2}.\qquad(5.1) I ( α ) = W 2 ( ζ , μ ) 2 = W 2 ( μ , ζ ) 2 . ( 5.1 )
By Couplings of Noncommutative Laws: the Norm Bound, the Cost Identity, the Tensor, Diagonal and Swapped Couplings, Weak-Star Closedness, and Displacement Interpolants §bound , α ∈ Σ 2 d , R \alpha\in\Sigma_{2d,R} α ∈ Σ 2 d , R ; also η ∈ Σ 2 d , R \eta\in\Sigma_{2d,R} η ∈ Σ 2 d , R (Step 3). Put m 1 = m 2 = 2 d m_{1}=m_{2}=2d m 1 = m 2 = 2 d , n = d n=d n = d , γ 1 = α \gamma_{1}=\alpha γ 1 = α , γ 2 = η \gamma_{2}=\eta γ 2 = η and a 1 = a 2 = a a^{1}=a^{2}=a a 1 = a 2 = a . Since σ a = ι 1 \sigma_{a}=\iota^{1} σ a = ι 1 , Couplings of Two Noncommutative Laws and Their Quadratic Cost §coupling gives α ∘ σ a = ζ = η ∘ σ a \alpha\circ\sigma_{a}=\zeta=\eta\circ\sigma_{a} α ∘ σ a = ζ = η ∘ σ a , so the data of Two Noncommutative Laws with a Common Marginal: Standing Notation for Their Amalgamated Free Product §data hold. The letters clash: the common marginal, which that setting (and every result stated in it) calls μ \mu μ , is here ζ \zeta ζ ; our μ \mu μ plays no role in the amalgamation. With this renaming we use the notation of its algebras and its embeddings and expectations : N = M ζ N=\mathcal{M}_{\zeta} N = M ζ on H ζ \mathcal{H}_{\zeta} H ζ , A 1 = M α A_{1}=\mathcal{M}_{\alpha} A 1 = M α on H α \mathcal{H}_{\alpha} H α , A 2 = M η A_{2}=\mathcal{M}_{\eta} A 2 = M η on H η \mathcal{H}_{\eta} H η , π ε : N → A ε \pi_{\varepsilon}:N\to A_{\varepsilon} π ε : N → A ε and E ε : A ε → N E_{\varepsilon}:A_{\varepsilon}\to N E ε : A ε → N ; let V ε V_{\varepsilon} V ε be the isometry of Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §isometry for γ ε \gamma_{\varepsilon} γ ε and a a a . Let H \mathcal{H} H , Ω \Omega Ω , φ \varphi φ and Λ ε \Lambda_{\varepsilon} Λ ε be the amalgamated free product space, its vacuum vector, its vacuum state and the actions, as in The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State ; by The Amalgamated Free Product Space of Two Noncommutative Laws over a Common Marginal, Its Vacuum Vector and Vacuum State §state , φ ( X ) = ⟨ Ω , X Ω ⟩ H \varphi(X)=\langle\Omega,X\Omega\rangle_{\mathcal{H}} φ ( X ) = ⟨ Ω , X Ω ⟩ H , so φ \varphi φ is linear. Write L q 1 L^{1}_{q} L q 1 , L q 2 L^{2}_{q} L q 2 for the multiplication operators on H α \mathcal{H}_{\alpha} H α , H η \mathcal{H}_{\eta} H η (q ∈ P 2 d q\in\mathcal{P}_{2d} q ∈ P 2 d ), and L q ζ L^{\zeta}_{q} L q ζ for those on H ζ \mathcal{H}_{\zeta} H ζ (q ∈ P d q\in\mathcal{P}_{d} q ∈ P d ). They lie in A 1 A_{1} A 1 , A 2 A_{2} A 2 , N N N respectively (preliminaries); 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 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 §adjoint for α , η ∈ Σ 2 d , R \alpha,\eta\in\Sigma_{2d,R} α , η ∈ Σ 2 d , R , ∥ L x i ε ∥ o p ≤ R \lVert L^{\varepsilon}_{x_{i}}\rVert_{\mathrm{op}}\le R ∥ L x i ε ∥ op ≤ R and ( L x i ε ) ∗ = L x i ε (L^{\varepsilon}_{x_{i}})^{*}=L^{\varepsilon}_{x_{i}} ( L x i ε ) ∗ = L x i ε for i ∈ [ 2 d ] i\in[2d] i ∈ [ 2 d ] ; and by Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §homomorphism and σ a ( x j ) = x j \sigma_{a}(x_{j})=x_{j} σ a ( x j ) = x j ,
π ε ( L x j ζ ) = L x j ε ( ε ∈ { 1 , 2 } , j ∈ [ d ] ) . ( 5.2 ) \pi_{\varepsilon}(L^{\zeta}_{x_{j}})=L^{\varepsilon}_{x_{j}}\qquad(\varepsilon\in\{1,2\},\ j\in[d]).\qquad(5.2) π ε ( L x j ζ ) = L x j ε ( ε ∈ { 1 , 2 } , j ∈ [ d ]) . ( 5.2 )
We use repeatedly, for ε ∈ { 1 , 2 } \varepsilon\in\{1,2\} ε ∈ { 1 , 2 } , b ∈ A ε b\in A_{\varepsilon} b ∈ A ε and X , Y ∈ N X,Y\in N X , Y ∈ N : π ε \pi_{\varepsilon} π ε is linear and multiplicative with π ε ( I ) = I \pi_{\varepsilon}(I)=I π ε ( I ) = I and τ γ ε ( π ε ( X ) ) = τ ζ ( X ) \tau_{\gamma_{\varepsilon}}(\pi_{\varepsilon}(X))=\tau_{\zeta}(X) τ γ ε ( π ε ( X )) = τ ζ ( X ) (Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §homomorphism ); E ε ( π ε ( X ) b π ε ( Y ) ) = X E ε ( b ) Y E_{\varepsilon}(\pi_{\varepsilon}(X)\,b\,\pi_{\varepsilon}(Y))=X\,E_{\varepsilon}(b)\,Y E ε ( π ε ( X ) b π ε ( Y )) = X E ε ( b ) Y , τ ζ ( E ε ( b ) ) = τ γ ε ( b ) \tau_{\zeta}(E_{\varepsilon}(b))=\tau_{\gamma_{\varepsilon}}(b) τ ζ ( E ε ( b )) = τ γ ε ( b ) and ⟨ ξ , E ε ( b ) ξ ′ ⟩ = ⟨ V ε ξ , b V ε ξ ′ ⟩ \langle\xi,E_{\varepsilon}(b)\xi'\rangle=\langle V_{\varepsilon}\xi,bV_{\varepsilon}\xi'\rangle ⟨ ξ , E ε ( b ) ξ ′ ⟩ = ⟨ V ε ξ , b V ε ξ ′ ⟩ for ξ , ξ ′ ∈ H ζ \xi,\xi'\in\mathcal{H}_{\zeta} ξ , ξ ′ ∈ H ζ (Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §expectation ); Λ ε \Lambda_{\varepsilon} Λ ε is linear and multiplicative with Λ ε ( b ∗ ) = Λ ε ( b ) ∗ \Lambda_{\varepsilon}(b^{*})=\Lambda_{\varepsilon}(b)^{*} Λ ε ( b ∗ ) = Λ ε ( b ) ∗ and ∥ Λ ε ( b ) ∥ o p ≤ ∥ b ∥ o p \lVert\Lambda_{\varepsilon}(b)\rVert_{\mathrm{op}}\le\lVert b\rVert_{\mathrm{op}} ∥ Λ ε ( b ) ∥ op ≤ ∥ b ∥ op (The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State §representation ); Λ 1 ( π 1 ( X ) ) = Λ 2 ( π 2 ( X ) ) \Lambda_{1}(\pi_{1}(X))=\Lambda_{2}(\pi_{2}(X)) Λ 1 ( π 1 ( X )) = Λ 2 ( π 2 ( X )) (The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State §amalgamation ); and φ ( Λ ε ( b ) ) = τ γ ε ( b ) \varphi(\Lambda_{\varepsilon}(b))=\tau_{\gamma_{\varepsilon}}(b) φ ( Λ ε ( b )) = τ γ ε ( b ) , ∥ Ω ∥ H = 1 \lVert\Omega\rVert_{\mathcal{H}}=1 ∥ Ω ∥ H = 1 (The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State §marginals ).
Step 6 (The expectation of the ν \nu ν -variables of η \eta η ). We show E 2 ( L x d + j 2 ) = L x j ζ E_{2}(L^{2}_{x_{d+j}})=L^{\zeta}_{x_{j}} E 2 ( L x d + j 2 ) = L x j ζ for j ∈ [ d ] j\in[d] j ∈ [ d ] . Let q ∈ P d q\in\mathcal{P}_{d} q ∈ P d . By Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §isometry , V 2 q ^ ζ = ι 1 ( q ) ^ η V_{2}\widehat{q}^{\,\zeta}=\widehat{\iota^{1}(q)}^{\,\eta} V 2 q ζ = ι 1 ( q ) η and V 2 Ω ζ = Ω η V_{2}\Omega_{\zeta}=\Omega_{\eta} V 2 Ω ζ = Ω η ; by Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §adjoint , Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §homomorphism and Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §values , ι 1 ( q ) ∗ = ι 1 ( q ∗ ) \iota^{1}(q)^{*}=\iota^{1}(q^{*}) ι 1 ( q ) ∗ = ι 1 ( q ∗ ) and ι 1 ( q ∗ ) x j = ι 1 ( q ∗ x j ) \iota^{1}(q^{*})x_{j}=\iota^{1}(q^{*}x_{j}) ι 1 ( q ∗ ) x j = ι 1 ( q ∗ x j ) . With the inner-product rule for E 2 E_{2} E 2 of Step 5, the vacuum identities for η ∈ Σ 2 d , R \eta\in\Sigma_{2d,R} η ∈ Σ 2 d , R and ζ ∈ Σ d , R \zeta\in\Sigma_{d,R} ζ ∈ Σ d , R , (4.1) for q ∗ q^{*} q ∗ , and η ∘ ι 1 = ζ \eta\circ\iota^{1}=\zeta η ∘ ι 1 = ζ (Step 3),
⟨ q ^ ζ , E 2 ( L x d + j 2 ) Ω ζ ⟩ = ⟨ ι 1 ( q ) ^ η , x d + j ^ η ⟩ = η ( ι 1 ( q ∗ ) x d + j ) = η ( ι 1 ( q ∗ ) x j ) = ζ ( q ∗ x j ) = ⟨ q ^ ζ , L x j ζ Ω ζ ⟩ . \langle\widehat{q}^{\,\zeta},E_{2}(L^{2}_{x_{d+j}})\Omega_{\zeta}\rangle=\langle\widehat{\iota^{1}(q)}^{\,\eta},\widehat{x_{d+j}}^{\,\eta}\rangle=\eta\bigl(\iota^{1}(q^{*})x_{d+j}\bigr)=\eta\bigl(\iota^{1}(q^{*})x_{j}\bigr)=\zeta(q^{*}x_{j})=\langle\widehat{q}^{\,\zeta},L^{\zeta}_{x_{j}}\Omega_{\zeta}\rangle. ⟨ q ζ , E 2 ( L x d + j 2 ) Ω ζ ⟩ = ⟨ ι 1 ( q ) η , x d + j η ⟩ = η ( ι 1 ( q ∗ ) x d + j ) = η ( ι 1 ( q ∗ ) x j ) = ζ ( q ∗ x j ) = ⟨ q ζ , L x j ζ Ω ζ ⟩ .
So D = E 2 ( L x d + j 2 ) − L x j ζ ∈ N D=E_{2}(L^{2}_{x_{d+j}})-L^{\zeta}_{x_{j}}\in N D = E 2 ( L x d + j 2 ) − L x j ζ ∈ N satisfies ⟨ q ^ ζ , D Ω ζ ⟩ = 0 \langle\widehat{q}^{\,\zeta},D\Omega_{\zeta}\rangle=0 ⟨ q ζ , D Ω ζ ⟩ = 0 for every q ∈ P d q\in\mathcal{P}_{d} q ∈ P d . The classes are the image of P d \mathcal{P}_{d} P d under the canonical map of the completion H ζ \mathcal{H}_{\zeta} H ζ (The Complex GNS Space of a Tracial State on Noncommutative Polynomials §gns and The Complex GNS Space of a Tracial State on Noncommutative Polynomials §classes ), which is dense by The Complex Hilbert Completion is a Complex Hilbert Space Containing a Dense Isometric Image, and Bounded Complex-Linear Maps Extend to It §dense ; choosing classes q k ^ ζ \widehat{q_{k}}^{\,\zeta} q k ζ converging to D Ω ζ D\Omega_{\zeta} D Ω ζ , continuity of the inner product gives ∥ D Ω ζ ∥ 2 = lim k ⟨ q k ^ ζ , D Ω ζ ⟩ = 0 \lVert D\Omega_{\zeta}\rVert^{2}=\lim_{k}\langle\widehat{q_{k}}^{\,\zeta},D\Omega_{\zeta}\rangle=0 ∥ D Ω ζ ∥ 2 = lim k ⟨ q k ζ , D Ω ζ ⟩ = 0 . Thus D Ω ζ = 0 D\Omega_{\zeta}=0 D Ω ζ = 0 , and D = 0 D=0 D = 0 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 §vacuum .
Step 7 (Mixed moments across the free product). We show φ ( Λ 1 ( b ) Λ 2 ( c ) ) = τ ζ ( E 1 ( b ) E 2 ( c ) ) \varphi(\Lambda_{1}(b)\Lambda_{2}(c))=\tau_{\zeta}(E_{1}(b)E_{2}(c)) φ ( Λ 1 ( b ) Λ 2 ( c )) = τ ζ ( E 1 ( b ) E 2 ( c )) for b ∈ A 1 b\in A_{1} b ∈ A 1 and c ∈ A 2 c\in A_{2} c ∈ A 2 . Let X = E 1 ( b ) X=E_{1}(b) X = E 1 ( b ) , Y = E 2 ( c ) Y=E_{2}(c) Y = E 2 ( c ) , and let b ∘ = b − π 1 ( X ) b^{\circ}=b-\pi_{1}(X) b ∘ = b − π 1 ( X ) and c ∘ = c − π 2 ( Y ) c^{\circ}=c-\pi_{2}(Y) c ∘ = c − π 2 ( Y ) be the centred parts , so b ∘ ∈ A 1 ∘ b^{\circ}\in A_{1}^{\circ} b ∘ ∈ A 1 ∘ , c ∘ ∈ A 2 ∘ c^{\circ}\in A_{2}^{\circ} c ∘ ∈ A 2 ∘ , i.e. E 1 ( b ∘ ) = 0 E_{1}(b^{\circ})=0 E 1 ( b ∘ ) = 0 and E 2 ( c ∘ ) = 0 E_{2}(c^{\circ})=0 E 2 ( c ∘ ) = 0 . As Λ 1 , Λ 2 , φ \Lambda_{1},\Lambda_{2},\varphi Λ 1 , Λ 2 , φ are linear and composition is bilinear, φ ( Λ 1 ( b ) Λ 2 ( c ) ) \varphi(\Lambda_{1}(b)\Lambda_{2}(c)) φ ( Λ 1 ( b ) Λ 2 ( c )) is the sum of the four terms below, evaluated with the rules of Step 5. First,
φ ( Λ 1 ( π 1 ( X ) ) Λ 2 ( π 2 ( Y ) ) ) = φ ( Λ 2 ( π 2 ( X Y ) ) ) = τ η ( π 2 ( X Y ) ) = τ ζ ( X Y ) . \varphi\bigl(\Lambda_{1}(\pi_{1}(X))\Lambda_{2}(\pi_{2}(Y))\bigr)=\varphi\bigl(\Lambda_{2}(\pi_{2}(XY))\bigr)=\tau_{\eta}(\pi_{2}(XY))=\tau_{\zeta}(XY). φ ( Λ 1 ( π 1 ( X )) Λ 2 ( π 2 ( Y )) ) = φ ( Λ 2 ( π 2 ( X Y )) ) = τ η ( π 2 ( X Y )) = τ ζ ( X Y ) .
Second,
φ ( Λ 1 ( π 1 ( X ) ) Λ 2 ( c ∘ ) ) = φ ( Λ 2 ( π 2 ( X ) c ∘ ) ) = τ η ( π 2 ( X ) c ∘ π 2 ( I ) ) = τ ζ ( X E 2 ( c ∘ ) ) = 0. \varphi\bigl(\Lambda_{1}(\pi_{1}(X))\Lambda_{2}(c^{\circ})\bigr)=\varphi\bigl(\Lambda_{2}(\pi_{2}(X)c^{\circ})\bigr)=\tau_{\eta}(\pi_{2}(X)c^{\circ}\pi_{2}(I))=\tau_{\zeta}\bigl(X\,E_{2}(c^{\circ})\bigr)=0. φ ( Λ 1 ( π 1 ( X )) Λ 2 ( c ∘ ) ) = φ ( Λ 2 ( π 2 ( X ) c ∘ ) ) = τ η ( π 2 ( X ) c ∘ π 2 ( I )) = τ ζ ( X E 2 ( c ∘ ) ) = 0.
Third,
φ ( Λ 1 ( b ∘ ) Λ 2 ( π 2 ( Y ) ) ) = φ ( Λ 1 ( b ∘ π 1 ( Y ) ) ) = τ α ( π 1 ( I ) b ∘ π 1 ( Y ) ) = τ ζ ( E 1 ( b ∘ ) Y ) = 0. \varphi\bigl(\Lambda_{1}(b^{\circ})\Lambda_{2}(\pi_{2}(Y))\bigr)=\varphi\bigl(\Lambda_{1}(b^{\circ}\pi_{1}(Y))\bigr)=\tau_{\alpha}(\pi_{1}(I)b^{\circ}\pi_{1}(Y))=\tau_{\zeta}\bigl(E_{1}(b^{\circ})\,Y\bigr)=0. φ ( Λ 1 ( b ∘ ) Λ 2 ( π 2 ( Y )) ) = φ ( Λ 1 ( b ∘ π 1 ( Y )) ) = τ α ( π 1 ( I ) b ∘ π 1 ( Y )) = τ ζ ( E 1 ( b ∘ ) Y ) = 0.
Fourth, ( ( 1 , b ∘ ) , ( 2 , c ∘ ) ) ((1,b^{\circ}),(2,c^{\circ})) (( 1 , b ∘ ) , ( 2 , c ∘ )) is an alternating tuple of type ( 1 , 2 ) (1,2) ( 1 , 2 ) , so by The Amalgamated Free Product of Two Noncommutative Laws over a Common Marginal: Representations, Amalgamation, Freeness and the Tracial Vacuum State §freeness , Λ 1 ( b ∘ ) Λ 2 ( c ∘ ) Ω = Ξ ( b ∘ , c ∘ ) \Lambda_{1}(b^{\circ})\Lambda_{2}(c^{\circ})\Omega=\Xi(b^{\circ},c^{\circ}) Λ 1 ( b ∘ ) Λ 2 ( c ∘ ) Ω = Ξ ( b ∘ , c ∘ ) and φ ( Λ 1 ( b ∘ ) Λ 2 ( c ∘ ) ) = ⟨ Ω , Ξ ( b ∘ , c ∘ ) ⟩ H = 0 \varphi(\Lambda_{1}(b^{\circ})\Lambda_{2}(c^{\circ}))=\langle\Omega,\Xi(b^{\circ},c^{\circ})\rangle_{\mathcal{H}}=0 φ ( Λ 1 ( b ∘ ) Λ 2 ( c ∘ )) = ⟨ Ω , Ξ ( b ∘ , c ∘ ) ⟩ H = 0 . Summing, φ ( Λ 1 ( b ) Λ 2 ( c ) ) = τ ζ ( X Y ) \varphi(\Lambda_{1}(b)\Lambda_{2}(c))=\tau_{\zeta}(XY) φ ( Λ 1 ( b ) Λ 2 ( c )) = τ ζ ( X Y ) .
Step 8 (The glued coupling ψ \psi ψ ). Define f : [ 2 d ] → { 1 , 2 } f:[2d]\to\{1,2\} f : [ 2 d ] → { 1 , 2 } and c i ∈ A f ( i ) c_{i}\in A_{f(i)} c i ∈ A f ( i ) by f ( i ) = 1 f(i)=1 f ( i ) = 1 , c i = L x d + i 1 c_{i}=L^{1}_{x_{d+i}} c i = L x d + i 1 and f ( d + i ) = 2 f(d+i)=2 f ( d + i ) = 2 , c d + i = L x d + i 2 c_{d+i}=L^{2}_{x_{d+i}} c d + i = L x d + i 2 for i ∈ [ d ] i\in[d] i ∈ [ d ] , and let T = ( T 1 , … , T 2 d ) T=(T_{1},\dots,T_{2d}) T = ( T 1 , … , T 2 d ) with T i = Λ f ( i ) ( c i ) ∈ L ( H ) T_{i}=\Lambda_{f(i)}(c_{i})\in\mathcal{L}(\mathcal{H}) T i = Λ f ( i ) ( c i ) ∈ L ( H ) . By Step 5, T i ∗ = Λ f ( i ) ( c i ∗ ) = T i T_{i}^{*}=\Lambda_{f(i)}(c_{i}^{*})=T_{i} T i ∗ = Λ f ( i ) ( c i ∗ ) = T i and ∥ T i ∥ o p ≤ ∥ c i ∥ o p ≤ R \lVert T_{i}\rVert_{\mathrm{op}}\le\lVert c_{i}\rVert_{\mathrm{op}}\le R ∥ T i ∥ op ≤ ∥ c i ∥ op ≤ R , and ∥ Ω ∥ H = 1 \lVert\Omega\rVert_{\mathcal{H}}=1 ∥ Ω ∥ H = 1 ; so The Law of a Tuple of Bounded Self-Adjoint Operators in a Tracial Vector State applies with 2 d 2d 2 d , H \mathcal{H} H , T T T in place of its n n n , H H H , T T T . Let ψ = λ T \psi=\lambda_{T} ψ = λ T , i.e. ψ ( p ) = ⟨ Ω , p ( T ) Ω ⟩ = φ ( p ( T ) ) \psi(p)=\langle\Omega,p(T)\Omega\rangle=\varphi(p(T)) ψ ( p ) = ⟨ Ω , p ( T ) Ω ⟩ = φ ( p ( T )) for p ∈ P 2 d p\in\mathcal{P}_{2d} p ∈ P 2 d .
To apply The Law of a Tuple of Bounded Self-Adjoint Operators in a Tracial Vector State §law we show φ ( T u T v ) = φ ( T v T u ) \varphi(T_{u}T_{v})=\varphi(T_{v}T_{u}) φ ( T u T v ) = φ ( T v T u ) for all words u , v ∈ W 2 d u,v\in W_{2d} u , v ∈ W 2 d . If u = ∅ u=\varnothing u = ∅ , then T u = I T_{u}=I T u = I by Evaluation of Noncommutative Polynomials at a Tuple of Bounded Operators §word-products and both sides equal φ ( T v ) \varphi(T_{v}) φ ( T v ) ; likewise if v = ∅ v=\varnothing v = ∅ . For nonempty u , v u,v u , v , T u T v = T u v T_{u}T_{v}=T_{uv} T u T v = T uv and T v T u = T v u T_{v}T_{u}=T_{vu} T 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 show φ ( T u v ) = φ ( T v u ) \varphi(T_{uv})=\varphi(T_{vu}) φ ( T uv ) = φ ( T vu ) for all nonempty u , v u,v u , v . First, for a letter ( i ) (i) ( i ) and a nonempty word w w w of length l l l with letters w 1 , … , w l w_{1},\dots,w_{l} w 1 , … , w l , we have φ ( T ( i ) w ) = φ ( T w ( i ) ) \varphi(T_{(i)w})=\varphi(T_{w(i)}) φ ( T ( i ) w ) = φ ( T w ( i ) ) : indeed T ( i ) w = T i T w T_{(i)w}=T_{i}T_{w} T ( i ) w = T i T w , T w ( i ) = T w T i T_{w(i)}=T_{w}T_{i} T w ( i ) = T w T i and T w = T w 1 ⋯ T w l T_{w}=T_{w_{1}}\cdots T_{w_{l}} T w = T w 1 ⋯ T w l by Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §values and Evaluation of Noncommutative Polynomials at a Tuple of Bounded Operators §word-products , so by associativity of composition T ( i ) w T_{(i)w} T ( i ) w is the product Λ f ( i ) ( c i ) Λ f ( w 1 ) ( c w 1 ) ⋯ Λ f ( w l ) ( c w l ) \Lambda_{f(i)}(c_{i})\Lambda_{f(w_{1})}(c_{w_{1}})\cdots\Lambda_{f(w_{l})}(c_{w_{l}}) Λ f ( i ) ( c i ) Λ f ( w 1 ) ( c w 1 ) ⋯ Λ f ( w l ) ( c w l ) of l + 1 ≥ 2 l+1\ge2 l + 1 ≥ 2 factors and T w ( i ) T_{w(i)} T w ( i ) is the product of the same factors with the first one moved to the end, and the claim 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 + 1 r=l+1 r = l + 1 . Now induct on the length k k k of u u u , for all nonempty v v v at once. If k = 1 k=1 k = 1 , u u u is a letter by Basic Properties of Words: Associativity, Reversal, Finitely Many Factorisations, and Countability §last-letter and the claim is the rotation just shown. If the claim holds for words of length k k k and u u u has length k + 1 k+1 k + 1 , then u = u ′ ( i ) u=u'(i) u = u ′ ( i ) with u ′ u' u ′ of length k k k by Basic Properties of Words: Associativity, Reversal, Finitely Many Factorisations, and Countability §last-letter ; by Basic Properties of Words: Associativity, Reversal, Finitely Many Factorisations, and Countability §monoid , u v = u ′ ( ( i ) v ) uv=u'((i)v) uv = u ′ (( i ) v ) , ( ( i ) v ) u ′ = ( i ) ( v u ′ ) ((i)v)u'=(i)(vu') (( i ) v ) u ′ = ( i ) ( v u ′ ) with ( i ) v (i)v ( i ) v and v u ′ vu' v u ′ nonempty, and ( v u ′ ) ( i ) = v ( u ′ ( i ) ) = v u (vu')(i)=v(u'(i))=vu ( v u ′ ) ( i ) = v ( u ′ ( i )) = vu ; so the induction hypothesis (for u ′ u' u ′ and ( i ) v (i)v ( i ) v ) and the rotation (for i i i and v u ′ vu' v u ′ ) give φ ( T u v ) = φ ( T ( i ) ( v u ′ ) ) = φ ( T ( v u ′ ) ( i ) ) = φ ( T v u ) \varphi(T_{uv})=\varphi(T_{(i)(vu')})=\varphi(T_{(vu')(i)})=\varphi(T_{vu}) φ ( T uv ) = φ ( T ( i ) ( v u ′ ) ) = φ ( T ( v u ′ ) ( i ) ) = φ ( T vu ) . Hence ψ ∈ Σ 2 d , R \psi\in\Sigma_{2d,R} ψ ∈ Σ 2 d , R by The Law of a Tuple of Bounded Self-Adjoint Operators in a Tracial Vector State §law ; in particular ψ \psi ψ is a tracial state by Noncommutative Laws of Finitely Many Self-Adjoint Variables with a Norm Bound §norm-bound .
Marginals of ψ \psi ψ . Let p ∈ P d p\in\mathcal{P}_{d} p ∈ P d . As a ( T ) = ( T 1 , … , T d ) a(T)=(T_{1},\dots,T_{d}) a ( T ) = ( T 1 , … , T d ) by Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §values , Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §substitution gives ( ι 1 ( p ) ) ( T ) = p ( T 1 , … , T d ) (\iota^{1}(p))(T)=p(T_{1},\dots,T_{d}) ( ι 1 ( p )) ( T ) = p ( T 1 , … , T d ) . Apply Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §transport with H α \mathcal{H}_{\alpha} H α , H \mathcal{H} H , d d d , ( L x d + 1 1 , … , L x 2 d 1 ) (L^{1}_{x_{d+1}},\dots,L^{1}_{x_{2d}}) ( L x d + 1 1 , … , L x 2 d 1 ) , A 1 A_{1} A 1 , Λ 1 \Lambda_{1} Λ 1 in place of its H H H , K K K , n n n , T T T , A \mathcal{A} A , Φ \Phi Φ (the set A 1 A_{1} A 1 has the required closure properties by the preliminaries, and Λ 1 \Lambda_{1} Λ 1 is linear, multiplicative and unital by Step 5): p ( T 1 , … , T d ) = Λ 1 ( p ( L x d + 1 1 , … , L x 2 d 1 ) ) p(T_{1},\dots,T_{d})=\Lambda_{1}\bigl(p(L^{1}_{x_{d+1}},\dots,L^{1}_{x_{2d}})\bigr) p ( T 1 , … , T d ) = Λ 1 ( p ( L x d + 1 1 , … , L x 2 d 1 ) ) . With L x 1 = ( L x 1 1 , … , L x 2 d 1 ) L^{1}_{x}=(L^{1}_{x_{1}},\dots,L^{1}_{x_{2d}}) L x 1 = ( L x 1 1 , … , L x 2 d 1 ) we have a ′ ( L x 1 ) = ( L x d + 1 1 , … , L x 2 d 1 ) a'(L^{1}_{x})=(L^{1}_{x_{d+1}},\dots,L^{1}_{x_{2d}}) a ′ ( L x 1 ) = ( L x d + 1 1 , … , L x 2 d 1 ) , so Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §substitution and Evaluation of Noncommutative Polynomials at Bounded Operators: a Unital Homomorphism Compatible with Adjoints, Substitution, Representations and the GNS Multiplication Operators §gns (for α ∈ Σ 2 d \alpha\in\Sigma_{2d} α ∈ Σ 2 d ) give p ( L x d + 1 1 , … , L x 2 d 1 ) = ( ι 2 ( p ) ) ( L x 1 ) = L ι 2 ( p ) 1 p(L^{1}_{x_{d+1}},\dots,L^{1}_{x_{2d}})=(\iota^{2}(p))(L^{1}_{x})=L^{1}_{\iota^{2}(p)} p ( L x d + 1 1 , … , L x 2 d 1 ) = ( ι 2 ( p )) ( L x 1 ) = L ι 2 ( p ) 1 . Hence, as α ∈ Π ( ζ , μ ) \alpha\in\Pi(\zeta,\mu) α ∈ Π ( ζ , μ ) ,
ψ ( ι 1 ( p ) ) = φ ( Λ 1 ( L ι 2 ( p ) 1 ) ) = τ α ( L ι 2 ( p ) 1 ) = α ( ι 2 ( p ) ) = μ ( p ) . \psi(\iota^{1}(p))=\varphi\bigl(\Lambda_{1}(L^{1}_{\iota^{2}(p)})\bigr)=\tau_{\alpha}(L^{1}_{\iota^{2}(p)})=\alpha(\iota^{2}(p))=\mu(p). ψ ( ι 1 ( p )) = φ ( Λ 1 ( L ι 2 ( p ) 1 ) ) = τ α ( L ι 2 ( p ) 1 ) = α ( ι 2 ( p )) = μ ( p ) .
In the same way, with a ′ ( T ) = ( T d + 1 , … , T 2 d ) = ( Λ 2 ( L x d + 1 2 ) , … , Λ 2 ( L x 2 d 2 ) ) a'(T)=(T_{d+1},\dots,T_{2d})=(\Lambda_{2}(L^{2}_{x_{d+1}}),\dots,\Lambda_{2}(L^{2}_{x_{2d}})) a ′ ( T ) = ( T d + 1 , … , T 2 d ) = ( Λ 2 ( L x d + 1 2 ) , … , Λ 2 ( L x 2 d 2 )) and H η \mathcal{H}_{\eta} H η , A 2 A_{2} A 2 , Λ 2 \Lambda_{2} Λ 2 , η \eta η in place of H α \mathcal{H}_{\alpha} H α , A 1 A_{1} A 1 , Λ 1 \Lambda_{1} Λ 1 , α \alpha α , we get ψ ( ι 2 ( p ) ) = τ η ( L ι 2 ( p ) 2 ) = η ( ι 2 ( p ) ) = ν ( p ) \psi(\iota^{2}(p))=\tau_{\eta}(L^{2}_{\iota^{2}(p)})=\eta(\iota^{2}(p))=\nu(p) ψ ( ι 2 ( p )) = τ η ( L ι 2 ( p ) 2 ) = η ( ι 2 ( p )) = ν ( p ) , as η ∈ Π ( ζ , ν ) \eta\in\Pi(\zeta,\nu) η ∈ Π ( ζ , ν ) . By Couplings of Two Noncommutative Laws and Their Quadratic Cost §coupling , ψ ∈ Π ( μ , ν ) \psi\in\Pi(\mu,\nu) ψ ∈ Π ( μ , ν ) .
Step 9 (Claim 3, upper bound). Let j ∈ [ d ] j\in[d] j ∈ [ d ] . By the homomorphism and value rules of evaluation, Step 7, Step 6, the expectation and trace rules of Step 5 (with π 1 ( I ) = I \pi_{1}(I)=I π 1 ( I ) = I ) and (5.2), then Left and Right Multiplication Operators on the Complex GNS Space of a Noncommutative Law: Boundedness, Algebra Rules, Adjoints, Commutation, the Vacuum and the Conjugation §algebra , the identity τ α ( L p 1 ) = α ( p ) \tau_{\alpha}(L^{1}_{p})=\alpha(p) τ α ( L p 1 ) = α ( p ) of the preliminaries, and condition (c) of Noncommutative Laws of Finitely Many Self-Adjoint Variables with a Norm Bound §tracial-state for α \alpha α ,
ψ ( x j x d + j ) = φ ( T j T d + j ) = τ ζ ( E 1 ( L x d + j 1 ) L x j ζ ) = τ ζ ( E 1 ( L x d + j 1 π 1 ( L x j ζ ) ) ) = τ α ( L x d + j 1 L x j 1 ) = α ( x d + j x j ) = α ( x j x d + j ) . \psi(x_{j}x_{d+j})=\varphi(T_{j}T_{d+j})=\tau_{\zeta}\bigl(E_{1}(L^{1}_{x_{d+j}})\,L^{\zeta}_{x_{j}}\bigr)=\tau_{\zeta}\bigl(E_{1}(L^{1}_{x_{d+j}}\pi_{1}(L^{\zeta}_{x_{j}}))\bigr)=\tau_{\alpha}(L^{1}_{x_{d+j}}L^{1}_{x_{j}})=\alpha(x_{d+j}x_{j})=\alpha(x_{j}x_{d+j}). ψ ( x j x d + j ) = φ ( T j T d + j ) = τ ζ ( E 1 ( L x d + j 1 ) L x j ζ ) = τ ζ ( E 1 ( L x d + j 1 π 1 ( L x j ζ )) ) = τ α ( L x d + j 1 L x j 1 ) = α ( x d + j x j ) = α ( x j x d + j ) .
By Couplings of Noncommutative Laws: the Norm Bound, the Cost Identity, the Tensor, Diagonal and Swapped Couplings, Weak-Star Closedness, and Displacement Interpolants §cost , applied to ψ ∈ Π ( μ , ν ) \psi\in\Pi(\mu,\nu) ψ ∈ Π ( μ , ν ) and to α ∈ Π ( ζ , μ ) \alpha\in\Pi(\zeta,\mu) α ∈ Π ( ζ , μ ) (with ζ , μ ∈ Σ d \zeta,\mu\in\Sigma_{d} ζ , μ ∈ Σ d ),
I ( ψ ) = M ( μ ) + M ( ν ) − 2 ∑ j = 1 d α ( x j x d + j ) , I ( α ) = M ( ζ ) + M ( μ ) − 2 ∑ j = 1 d α ( x j x d + j ) , I(\psi)=M(\mu)+M(\nu)-2\sum_{j=1}^{d}\alpha(x_{j}x_{d+j}),\qquad I(\alpha)=M(\zeta)+M(\mu)-2\sum_{j=1}^{d}\alpha(x_{j}x_{d+j}), I ( ψ ) = M ( μ ) + M ( ν ) − 2 j = 1 ∑ d α ( x j x d + j ) , I ( α ) = M ( ζ ) + M ( μ ) − 2 j = 1 ∑ d α ( x j x d + j ) ,
so I ( ψ ) = I ( α ) + M ( ν ) − M ( ζ ) I(\psi)=I(\alpha)+M(\nu)-M(\zeta) I ( ψ ) = I ( α ) + M ( ν ) − M ( ζ ) . By The Noncommutative Quadratic Wasserstein Distance and Optimal Couplings §distance , W 2 ( μ , ν ) 2 W_{2}(\mu,\nu)^{2} W 2 ( μ , ν ) 2 is the infimum of the costs of couplings of μ \mu μ and ν \nu ν , so W 2 ( μ , ν ) 2 ≤ I ( ψ ) W_{2}(\mu,\nu)^{2}\le I(\psi) W 2 ( μ , ν ) 2 ≤ I ( ψ ) , and with (5.1)
W 2 ( μ , ν ) 2 ≤ W 2 ( μ , ζ ) 2 + M ( ν ) − M ( ζ ) . ( 9.1 ) W_{2}(\mu,\nu)^{2}\le W_{2}(\mu,\zeta)^{2}+M(\nu)-M(\zeta).\qquad(9.1) W 2 ( μ , ν ) 2 ≤ W 2 ( μ , ζ ) 2 + M ( ν ) − M ( ζ ) . ( 9.1 )
As γ ∈ Π ( μ , ν ) \gamma\in\Pi(\mu,\nu) γ ∈ Π ( μ , ν ) was arbitrary, this proves claim 3.
Step 10 (Claim 4, optimal couplings). Let γ \gamma γ be optimal, so I ( γ ) = W 2 ( μ , ν ) 2 I(\gamma)=W_{2}(\mu,\nu)^{2} I ( γ ) = W 2 ( μ , ν ) 2 by The Noncommutative Quadratic Wasserstein Distance and Optimal Couplings §optimal . By Laws of Self-Adjoint Tuples in a Tracial W*-Probability Space: Moments, Affine Images, Couplings, Embeddings and L^2 Approximation §coupling for the tuples L x L_{x} L x and z z z of Step 1 (whose laws are μ \mu μ and ζ \zeta ζ ), I ( β ) ≥ W 2 ( μ , ζ ) 2 I(\beta)\ge W_{2}(\mu,\zeta)^{2} I ( β ) ≥ W 2 ( μ , ζ ) 2 . With claim 2 and (9.1),
W 2 ( μ , ν ) 2 = I ( γ ) = I ( β ) + M ( ν ) − M ( ζ ) ≥ W 2 ( μ , ζ ) 2 + M ( ν ) − M ( ζ ) ≥ W 2 ( μ , ν ) 2 . W_{2}(\mu,\nu)^{2}=I(\gamma)=I(\beta)+M(\nu)-M(\zeta)\ge W_{2}(\mu,\zeta)^{2}+M(\nu)-M(\zeta)\ge W_{2}(\mu,\nu)^{2}. W 2 ( μ , ν ) 2 = I ( γ ) = I ( β ) + M ( ν ) − M ( ζ ) ≥ W 2 ( μ , ζ ) 2 + M ( ν ) − M ( ζ ) ≥ W 2 ( μ , ν ) 2 .
Hence equality holds throughout: W 2 ( μ , ν ) 2 = W 2 ( μ , ζ ) 2 + M ( ν ) − M ( ζ ) W_{2}(\mu,\nu)^{2}=W_{2}(\mu,\zeta)^{2}+M(\nu)-M(\zeta) W 2 ( μ , ν ) 2 = W 2 ( μ , ζ ) 2 + M ( ν ) − M ( ζ ) and I ( β ) = W 2 ( μ , ζ ) 2 I(\beta)=W_{2}(\mu,\zeta)^{2} I ( β ) = W 2 ( μ , ζ ) 2 , so β ∈ Π ( μ , ζ ) \beta\in\Pi(\mu,\zeta) β ∈ Π ( μ , ζ ) is optimal by The Noncommutative Quadratic Wasserstein Distance and Optimal Couplings §optimal . This proves claim 4.
Step 11 (Claim 5, variational formula). By claim 3, W 2 ( μ , ν ) 2 W_{2}(\mu,\nu)^{2} W 2 ( μ , ν ) 2 is a lower bound of the set { W 2 ( μ , ζ γ ) 2 + M ( ν ) − M ( ζ γ ) : γ ∈ Π ( μ , ν ) } \{W_{2}(\mu,\zeta_{\gamma})^{2}+M(\nu)-M(\zeta_{\gamma}):\ \gamma\in\Pi(\mu,\nu)\} { W 2 ( μ , ζ γ ) 2 + M ( ν ) − M ( ζ γ ) : γ ∈ Π ( μ , ν )} , and by claim 4 it is the value of this expression at every optimal γ \gamma γ . Since μ , ν ∈ Σ d , R \mu,\nu\in\Sigma_{d,R} μ , ν ∈ Σ d , R , The Noncommutative Wasserstein Distance: Existence of Optimal Couplings, Symmetry, Separation, a Moment Bound, Weak-Star Lower Semicontinuity, and Displacement Interpolation §attained provides an optimal γ ∈ Π ( μ , ν ) \gamma\in\Pi(\mu,\nu) γ ∈ Π ( μ , ν ) , so W 2 ( μ , ν ) 2 W_{2}(\mu,\nu)^{2} W 2 ( μ , ν ) 2 belongs to the set and is its least element. This proves claim 5.