Each result cited is applied with the data of its own statement. Write W = W 2 ( μ , ν ) W=W_{2}(\mu,\nu) W = W 2 ( μ , ν ) , and for k , l ∈ [ 2 d ] k,l\in[2d] k , l ∈ [ 2 d ] let δ k l \delta_{kl} δ k l be 1 1 1 if k = l k=l k = l and 0 0 0 otherwise. Throughout, j j j ranges over [ d ] [d] [ d ] and l l l over [ 2 d ] [2d] [ 2 d ] .
Step 1 (The data and their composites). By Affine Data and Affine Substitutions of Noncommutative Polynomials §tuple , the tuple of a datum ( A , c ) (A,c) ( A , c ) from 2 d 2d 2 d to 2 d 2d 2 d variables has components c i 1 + ∑ l = 1 2 d A i l x l c_{i}1+\sum_{l=1}^{2d}A_{il}x_{l} c i 1 + ∑ l = 1 2 d A i l x l ; since 1 , x 1 , … , x 2 d 1,x_{1},\dots,x_{2d} 1 , x 1 , … , x 2 d are distinct monomials, these components determine the numbers c i c_{i} c i and A i l A_{il} A i l , so a datum is determined by its tuple. Reading off the coefficients of the given tuples, T = ( A , 0 ) T=(A,0) T = ( A , 0 ) and T ′ = ( A ′ , 0 ) T'=(A',0) T ′ = ( A ′ , 0 ) with
A j l = δ j l , A j l ′ = δ d + j , l , A d + j , l = A d + j , l ′ = α δ j l − α δ d + j , l . A_{jl}=\delta_{jl},\qquad A'_{jl}=\delta_{d+j,l},\qquad A_{d+j,l}=A'_{d+j,l}=\alpha\delta_{jl}-\alpha\delta_{d+j,l}. A j l = δ j l , A j l ′ = δ d + j , l , A d + j , l = A d + j , l ′ = α δ j l − α δ d + j , l .
In particular, with h j = x j − x d + j h_{j}=x_{j}-x_{d+j} h j = x j − x d + j as in Marginal Isometries, Bounded Plans and Displacement Pairings for Noncommutative Laws , a d + j T = a d + j T ′ = α x j − α x d + j = α h j a^{T}_{d+j}=a^{T'}_{d+j}=\alpha x_{j}-\alpha x_{d+j}=\alpha h_{j} a d + j T = a d + j T ′ = α x j − α x d + j = α h j .
As γ ∈ Π ( μ , ν ) \gamma\in\Pi(\mu,\nu) γ ∈ Π ( μ , ν ) , γ \gamma γ is a tracial state on P 2 d \mathcal{P}_{2d} P 2 d with γ ∘ ι 1 = μ \gamma\circ\iota^{1}=\mu γ ∘ ι 1 = μ and γ ∘ ι 2 = ν \gamma\circ\iota^{2}=\nu γ ∘ ι 2 = ν by Couplings of Two Noncommutative Laws and Their Quadratic Cost §coupling , By Noncommutative Laws of Finitely Many Self-Adjoint Variables with a Norm Bound §law there are real 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 R R the larger of R 1 R_{1} R 1 and R 2 R_{2} 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 , and then γ ∈ Π ( μ , ν ) ⊆ Σ 2 d , R ⊆ Σ 2 d \gamma\in\Pi(\mu,\nu)\subseteq\Sigma_{2d,R}\subseteq\Sigma_{2d} γ ∈ Π ( μ , ν ) ⊆ Σ 2 d , R ⊆ Σ 2 d by Couplings of Noncommutative Laws: the Norm Bound, the Cost Identity, the Tensor, Diagonal and Swapped Couplings, Weak-Star Closedness, and Displacement Interpolants §bound and Noncommutative Laws of Finitely Many Self-Adjoint Variables with a Norm Bound §law . By Affine Substitutions of Noncommutative Laws: Self-Adjointness, Composition, Moment Formulas and Positivity, and the Coordinate Data §self-adjoint , the tuples a T a^{T} a T and a T ′ a^{T'} a T ′ are 2 d 2d 2 d -tuples in P 2 d , s a \mathcal{P}_{2d,\mathrm{sa}} P 2 d , sa , and π , π ′ ∈ Σ 2 d \pi,\pi'\in\Sigma_{2d} π , π ′ ∈ Σ 2 d . Let S = ( B , 0 ) S=(B,0) S = ( B , 0 ) and S ′ = ( B ′ , 0 ) S'=(B',0) S ′ = ( B ′ , 0 ) be the affine data from 2 d 2d 2 d to 2 d 2d 2 d variables with
B j l = δ j l , B d + j , l = δ j l − α − 1 δ d + j , l , B j l ′ = δ j l + α − 1 δ d + j , l , B d + j , l ′ = δ j l . B_{jl}=\delta_{jl},\qquad B_{d+j,l}=\delta_{jl}-\alpha^{-1}\delta_{d+j,l},\qquad B'_{jl}=\delta_{jl}+\alpha^{-1}\delta_{d+j,l},\qquad B'_{d+j,l}=\delta_{jl}. B j l = δ j l , B d + j , l = δ j l − α − 1 δ d + j , l , B j l ′ = δ j l + α − 1 δ d + j , l , B d + j , l ′ = δ j l .
By Affine Data and Affine Substitutions of Noncommutative Polynomials §composite , all composites below have zero shift, and their matrices are computed from ( P 1 A ) j l = A j l = δ j l = P j l 1 (P^{1}A)_{jl}=A_{jl}=\delta_{jl}=P^{1}_{jl} ( P 1 A ) j l = A j l = δ j l = P j l 1 , ( P 1 A ′ ) j l = A j l ′ = δ d + j , l = P j l 2 (P^{1}A')_{jl}=A'_{jl}=\delta_{d+j,l}=P^{2}_{jl} ( P 1 A ′ ) j l = A j l ′ = δ d + j , l = P j l 2 , where P 1 , P 2 P^{1},P^{2} P 1 , P 2 are the matrices of the marginal data , and
( B A ) j l = A j l = δ j l , ( B A ) d + j , l = A j l − α − 1 A d + j , l = δ d + j , l , ( B ′ A ′ ) j l = A j l ′ + α − 1 A d + j , l ′ = δ j l , ( B ′ A ′ ) d + j , l = A j l ′ = δ d + j , l . (BA)_{jl}=A_{jl}=\delta_{jl},\quad(BA)_{d+j,l}=A_{jl}-\alpha^{-1}A_{d+j,l}=\delta_{d+j,l},\quad(B'A')_{jl}=A'_{jl}+\alpha^{-1}A'_{d+j,l}=\delta_{jl},\quad(B'A')_{d+j,l}=A'_{jl}=\delta_{d+j,l}. ( B A ) j l = A j l = δ j l , ( B A ) d + j , l = A j l − α − 1 A d + j , l = δ d + j , l , ( B ′ A ′ ) j l = A j l ′ + α − 1 A d + j , l ′ = δ j l , ( B ′ A ′ ) d + j , l = A j l ′ = δ d + j , l .
Hence, with the identity datum of Affine Data and Affine Substitutions of Noncommutative Polynomials §identity ,
p r 1 ∘ T = p r 1 , p r 1 ∘ T ′ = p r 2 , S ∘ T = S ′ ∘ T ′ = i d 2 d . (1) \mathrm{pr}^{1}\circ T=\mathrm{pr}^{1},\qquad\mathrm{pr}^{1}\circ T'=\mathrm{pr}^{2},\qquad S\circ T=S'\circ T'=\mathrm{id}_{2d}.\tag{1} pr 1 ∘ T = pr 1 , pr 1 ∘ T ′ = pr 2 , S ∘ T = S ′ ∘ T ′ = id 2 d . ( 1 )
Since σ p r 1 = ι 1 \sigma_{\mathrm{pr}^{1}}=\iota^{1} σ pr 1 = ι 1 and σ p r 2 = ι 2 \sigma_{\mathrm{pr}^{2}}=\iota^{2} σ pr 2 = ι 2 by Affine Substitutions of Noncommutative Laws: Self-Adjointness, Composition, Moment Formulas and Positivity, and the Coordinate Data §coordinate , (1) and Affine Substitutions of Noncommutative Laws: Self-Adjointness, Composition, Moment Formulas and Positivity, and the Coordinate Data §composition give
σ T ∘ ι 1 = σ p r 1 ∘ T = ι 1 , σ T ′ ∘ ι 1 = σ p r 1 ∘ T ′ = ι 2 , π ∘ σ S = γ ∘ σ S ∘ T = γ , π ′ ∘ σ S ′ = γ ∘ σ S ′ ∘ T ′ = γ . (2) \sigma_{T}\circ\iota^{1}=\sigma_{\mathrm{pr}^{1}\circ T}=\iota^{1},\qquad\sigma_{T'}\circ\iota^{1}=\sigma_{\mathrm{pr}^{1}\circ T'}=\iota^{2},\qquad\pi\circ\sigma_{S}=\gamma\circ\sigma_{S\circ T}=\gamma,\qquad\pi'\circ\sigma_{S'}=\gamma\circ\sigma_{S'\circ T'}=\gamma.\tag{2} σ T ∘ ι 1 = σ pr 1 ∘ T = ι 1 , σ T ′ ∘ ι 1 = σ pr 1 ∘ T ′ = ι 2 , π ∘ σ S = γ ∘ σ S ∘ T = γ , π ′ ∘ σ S ′ = γ ∘ σ S ′ ∘ T ′ = γ . ( 2 )
Step 2 (Clause 1). By (2), π ∘ ι 1 = γ ∘ ( σ T ∘ ι 1 ) = γ ∘ ι 1 = μ \pi\circ\iota^{1}=\gamma\circ(\sigma_{T}\circ\iota^{1})=\gamma\circ\iota^{1}=\mu π ∘ ι 1 = γ ∘ ( σ T ∘ ι 1 ) = γ ∘ ι 1 = μ and π ′ ∘ ι 1 = γ ∘ ι 2 = ν \pi'\circ\iota^{1}=\gamma\circ\iota^{2}=\nu π ′ ∘ ι 1 = γ ∘ ι 2 = ν ; as π , π ′ ∈ Σ 2 d \pi,\pi'\in\Sigma_{2d} π , π ′ ∈ Σ 2 d , π \pi π and π ′ \pi' π ′ are bounded plans at μ \mu μ and at ν \nu ν by Marginal Isometries, Bounded Plans and Displacement Pairings for Noncommutative Laws §plans . For the momentum, the moment formula of Affine Substitutions of Noncommutative Laws: Self-Adjointness, Composition, Moment Formulas and Positivity, and the Coordinate Data §moments for the tracial state γ \gamma γ , the datum T T T (zero shift) and i = k = d + j i=k=d+j i = k = d + j , together with the symmetry m j , d + j ( γ ) = m d + j , j ( γ ) \mathrm{m}_{j,d+j}(\gamma)=\mathrm{m}_{d+j,j}(\gamma) m j , d + j ( γ ) = m d + j , j ( γ ) of the same clause, gives
π ( x d + j x d + j ) = ∑ l = 1 2 d ∑ l ′ = 1 2 d A d + j , l A d + j , l ′ m l l ′ ( γ ) = α 2 ( m j j ( γ ) + m d + j , d + j ( γ ) − 2 m j , d + j ( γ ) ) , \pi(x_{d+j}x_{d+j})=\sum_{l=1}^{2d}\sum_{l'=1}^{2d}A_{d+j,l}A_{d+j,l'}\,\mathrm{m}_{ll'}(\gamma)=\alpha^{2}\bigl(\mathrm{m}_{jj}(\gamma)+\mathrm{m}_{d+j,d+j}(\gamma)-2\,\mathrm{m}_{j,d+j}(\gamma)\bigr), π ( x d + j x d + j ) = l = 1 ∑ 2 d l ′ = 1 ∑ 2 d A d + j , l A d + j , l ′ m l l ′ ( γ ) = α 2 ( m jj ( γ ) + m d + j , d + j ( γ ) − 2 m j , d + j ( γ ) ) ,
and the same value for π ′ ( x d + j x d + j ) \pi'(x_{d+j}x_{d+j}) π ′ ( x d + j x d + j ) , since the rows d + j d+j d + j of A A A and A ′ A' A ′ agree. The same formula for the data p r 1 , p r 2 \mathrm{pr}^{1},\mathrm{pr}^{2} pr 1 , pr 2 (zero shift), whose substitutions are ι 1 , ι 2 \iota^{1},\iota^{2} ι 1 , ι 2 , gives m j j ( γ ∘ ι 1 ) = m j j ( γ ) \mathrm{m}_{jj}(\gamma\circ\iota^{1})=\mathrm{m}_{jj}(\gamma) m jj ( γ ∘ ι 1 ) = m jj ( γ ) and m j j ( γ ∘ ι 2 ) = m d + j , d + j ( γ ) \mathrm{m}_{jj}(\gamma\circ\iota^{2})=\mathrm{m}_{d+j,d+j}(\gamma) m jj ( γ ∘ ι 2 ) = m d + j , d + j ( γ ) . Summing over j j j and using Affine Substitutions of Noncommutative Laws: Self-Adjointness, Composition, Moment Formulas and Positivity, and the Coordinate Data §cost-identity , Couplings of Two Noncommutative Laws and Their Quadratic Cost §cost and the optimality of γ \gamma γ (The Noncommutative Quadratic Wasserstein Distance and Optimal Couplings §optimal ),
∑ j = 1 d π ( x d + j x d + j ) = α 2 ( M ( γ ∘ ι 1 ) + M ( γ ∘ ι 2 ) − 2 ∑ j = 1 d m j , d + j ( γ ) ) = α 2 γ ( Δ d ) = α 2 I ( γ ) = α 2 W 2 , \sum_{j=1}^{d}\pi(x_{d+j}x_{d+j})=\alpha^{2}\Bigl(M(\gamma\circ\iota^{1})+M(\gamma\circ\iota^{2})-2\sum_{j=1}^{d}\mathrm{m}_{j,d+j}(\gamma)\Bigr)=\alpha^{2}\gamma(\Delta_{d})=\alpha^{2}I(\gamma)=\alpha^{2}W^{2}, j = 1 ∑ d π ( x d + j x d + j ) = α 2 ( M ( γ ∘ ι 1 ) + M ( γ ∘ ι 2 ) − 2 j = 1 ∑ d m j , d + j ( γ ) ) = α 2 γ ( Δ d ) = α 2 I ( γ ) = α 2 W 2 ,
and likewise for π ′ \pi' π ′ . Since α W ≥ 0 \alpha W\ge0 α W ≥ 0 has square α 2 W 2 \alpha^{2}W^{2} α 2 W 2 , Existence and Uniqueness of the Nonnegative Square Root gives ∣ π ∣ m o m = ∣ π ′ ∣ m o m = α W |\pi|_{\mathrm{mom}}=|\pi'|_{\mathrm{mom}}=\alpha W ∣ π ∣ mom = ∣ π ′ ∣ mom = α W . Finally, by Calculus of Square-Integrable Noncommutative Laws: Agreement on Bounded Laws, Lipschitz Estimates, Functoriality of Push-Forwards, Moment Formulas, Positivity, the Cost and the Diagonal Coupling §bounded for the datum p r 1 \mathrm{pr}^{1} pr 1 ,
p r # 1 κ 2 d ( π ) = κ d ( π ∘ ι 1 ) = κ d ( μ ) , p r # 1 κ 2 d ( π ′ ) = κ d ( π ′ ∘ ι 1 ) = κ d ( ν ) , \mathrm{pr}^{1}_{\#}\kappa_{2d}(\pi)=\kappa_{d}(\pi\circ\iota^{1})=\kappa_{d}(\mu),\qquad\mathrm{pr}^{1}_{\#}\kappa_{2d}(\pi')=\kappa_{d}(\pi'\circ\iota^{1})=\kappa_{d}(\nu), pr # 1 κ 2 d ( π ) = κ d ( π ∘ ι 1 ) = κ d ( μ ) , pr # 1 κ 2 d ( π ′ ) = κ d ( π ′ ∘ ι 1 ) = κ d ( ν ) ,
so κ 2 d ( π ) ∈ Σ 2 d 2 \kappa_{2d}(\pi)\in\Sigma^{2}_{2d} κ 2 d ( π ) ∈ Σ 2 d 2 is a plan at κ d ( μ ) \kappa_{d}(\mu) κ d ( μ ) and κ 2 d ( π ′ ) \kappa_{2d}(\pi') κ 2 d ( π ′ ) one at κ d ( ν ) \kappa_{d}(\nu) κ d ( ν ) by Plans at a Square-Integrable Noncommutative Law §plan .
Step 3 (Tuples of law κ 2 d ( γ ) \kappa_{2d}(\gamma) κ 2 d ( γ ) ). Let ( H , M , Ω ) (H,M,\Omega) ( H , M , Ω ) be a tracial W*-probability space and X , Y X,Y X , Y L 2 L^{2} L 2 d d d -tuples with l a w ( X , Y ) = κ 2 d ( γ ) \mathrm{law}(X,Y)=\kappa_{2d}(\gamma) law ( X , Y ) = κ 2 d ( γ ) . We claim l a w ( X ) = κ d ( μ ) \mathrm{law}(X)=\kappa_{d}(\mu) law ( X ) = κ d ( μ ) , l a w ( Y ) = κ d ( ν ) \mathrm{law}(Y)=\kappa_{d}(\nu) law ( Y ) = κ d ( ν ) and ∥ X − Y ∥ 2 = W \lVert X-Y\rVert_{2}=W ∥ X − Y ∥ 2 = W . By Calculus of Laws of Square-Integrable Tuples: Bounded Tuples, the Lipschitz Bound, Moments, Affine Push-Forwards, Couplings and Embeddings §coupling , l a w ( X , Y ) ∈ Π 2 ( l a w ( X ) , l a w ( Y ) ) \mathrm{law}(X,Y)\in\Pi^{2}(\mathrm{law}(X),\mathrm{law}(Y)) law ( X , Y ) ∈ Π 2 ( law ( X ) , law ( Y )) , so by Square-Integrable Noncommutative Laws: the Wasserstein Completion of the Laws, Affine Push-Forwards, Moments, Couplings and Cost §couplings and Calculus of Square-Integrable Noncommutative Laws: Agreement on Bounded Laws, Lipschitz Estimates, Functoriality of Push-Forwards, Moment Formulas, Positivity, the Cost and the Diagonal Coupling §bounded ,
l a w ( X ) = p r # 1 κ 2 d ( γ ) = κ d ( γ ∘ ι 1 ) = κ d ( μ ) , l a w ( Y ) = p r # 2 κ 2 d ( γ ) = κ d ( γ ∘ ι 2 ) = κ d ( ν ) . \mathrm{law}(X)=\mathrm{pr}^{1}_{\#}\kappa_{2d}(\gamma)=\kappa_{d}(\gamma\circ\iota^{1})=\kappa_{d}(\mu),\qquad\mathrm{law}(Y)=\mathrm{pr}^{2}_{\#}\kappa_{2d}(\gamma)=\kappa_{d}(\gamma\circ\iota^{2})=\kappa_{d}(\nu). law ( X ) = pr # 1 κ 2 d ( γ ) = κ d ( γ ∘ ι 1 ) = κ d ( μ ) , law ( Y ) = pr # 2 κ 2 d ( γ ) = κ d ( γ ∘ ι 2 ) = κ d ( ν ) .
By the same clause of Calculus of Laws of Square-Integrable Tuples: Bounded Tuples, the Lipschitz Bound, Moments, Affine Push-Forwards, Couplings and Embeddings §coupling , the second part of Calculus of Square-Integrable Noncommutative Laws: Agreement on Bounded Laws, Lipschitz Estimates, Functoriality of Push-Forwards, Moment Formulas, Positivity, the Cost and the Diagonal Coupling §bounded (as γ ∈ Π ( μ , ν ) \gamma\in\Pi(\mu,\nu) γ ∈ Π ( μ , ν ) ) and optimality, ∥ X − Y ∥ 2 2 = I ( κ 2 d ( γ ) ) = I ( γ ) = W 2 \lVert X-Y\rVert_{2}^{2}=\mathcal{I}(\kappa_{2d}(\gamma))=I(\gamma)=W^{2} ∥ X − Y ∥ 2 2 = I ( κ 2 d ( γ )) = I ( γ ) = W 2 ; both ∥ X − Y ∥ 2 \lVert X-Y\rVert_{2} ∥ X − Y ∥ 2 and W W W are nonnegative, so they are equal. Moreover W ^ 2 ( κ d ( μ ) , κ d ( ν ) ) = W \widehat{W}_{2}(\kappa_{d}(\mu),\kappa_{d}(\nu))=W W 2 ( κ d ( μ ) , κ d ( ν )) = W by The Metric Completion is a Complete Metric Space with a Dense Isometric Copy of the Space, and Maps Preserving Cauchy Sequences Extend to It §isometry , and W ^ 2 ( λ , λ ) = 0 \widehat{W}_{2}(\lambda,\lambda)=0 W 2 ( λ , λ ) = 0 for all λ \lambda λ by The Metric Completion is a Complete Metric Space with a Dense Isometric Copy of the Space, and Maps Preserving Cauchy Sequences Extend to It §metric ; hence
φ ( κ d ( μ ) ) = α 2 W 2 , ψ ( κ d ( ν ) ) = − α 2 W 2 . (3) \varphi(\kappa_{d}(\mu))=\tfrac{\alpha}{2}W^{2},\qquad\psi(\kappa_{d}(\nu))=-\tfrac{\alpha}{2}W^{2}.\tag{3} φ ( κ d ( μ )) = 2 α W 2 , ψ ( κ d ( ν )) = − 2 α W 2 . ( 3 )
We also use that, by Sums, Real Multiples and the Pairing of Square-Integrable Tuples in a Tracial W*-Probability Space §pairing , ⟨ ⋅ , ⋅ ⟩ 2 \langle\cdot,\cdot\rangle_{2} ⟨ ⋅ , ⋅ ⟩ 2 is the inner product of H d H^{d} H d , real on L 2 L^{2} L 2 d d d -tuples, and ∥ ⋅ ∥ 2 \lVert\cdot\rVert_{2} ∥ ⋅ ∥ 2 its norm, while sums, real multiples and differences are computed componentwise (Sums, Real Multiples and the Pairing of Square-Integrable Tuples in a Tracial W*-Probability Space §operations , Square-Integrable Tuples in a Tracial W*-Probability Space: Their Norm, Affine Images, Pairs, Embedded Images and Laws §tuples ); so for L 2 L^{2} L 2 d d d -tuples U , V U,V U , V and real t t t , conjugate symmetry and the reality of the pairing give
∥ U + t V ∥ 2 2 = ∥ U ∥ 2 2 + 2 t ⟨ U , V ⟩ 2 + t 2 ∥ V ∥ 2 2 , ⟨ U , V ⟩ 2 = ⟨ V , U ⟩ 2 . (4) \lVert U+tV\rVert_{2}^{2}=\lVert U\rVert_{2}^{2}+2t\langle U,V\rangle_{2}+t^{2}\lVert V\rVert_{2}^{2},\qquad\langle U,V\rangle_{2}=\langle V,U\rangle_{2}.\tag{4} ∥ U + t V ∥ 2 2 = ∥ U ∥ 2 2 + 2 t ⟨ U , V ⟩ 2 + t 2 ∥ V ∥ 2 2 , ⟨ U , V ⟩ 2 = ⟨ V , U ⟩ 2 . ( 4 )
Step 4 (Clause 2). Let η > 0 \eta>0 η > 0 and put r = 2 η / ( α + β ) > 0 r=2\eta/(\alpha+\beta)>0 r = 2 η / ( α + β ) > 0 . Let ( H , M , Ω ) (H,M,\Omega) ( H , M , Ω ) be a tracial W*-probability space and X , P , X ′ X,P,X' X , P , X ′ L 2 L^{2} L 2 d d d -tuples with l a w ( X , P ) = κ 2 d ( π ) \mathrm{law}(X,P)=\kappa_{2d}(\pi) law ( X , P ) = κ 2 d ( π ) and ∥ X ′ − X ∥ 2 < r \lVert X'-X\rVert_{2}<r ∥ X ′ − X ∥ 2 < r . Let Y = X − α − 1 P Y=X-\alpha^{-1}P Y = X − α − 1 P . By Square-Integrable Tuples in a Tracial W*-Probability Space: Their Norm, Affine Images, Pairs, Embedded Images and Laws §operations , S ( X , P ) S(X,P) S ( X , P ) has components X j X_{j} X j and X j − α − 1 P j X_{j}-\alpha^{-1}P_{j} X j − α − 1 P j in the positions j j j and d + j d+j d + j , so S ( X , P ) = ( X , Y ) S(X,P)=(X,Y) S ( X , P ) = ( X , Y ) . By Calculus of Laws of Square-Integrable Tuples: Bounded Tuples, the Lipschitz Bound, Moments, Affine Push-Forwards, Couplings and Embeddings §push-forward , Calculus of Square-Integrable Noncommutative Laws: Agreement on Bounded Laws, Lipschitz Estimates, Functoriality of Push-Forwards, Moment Formulas, Positivity, the Cost and the Diagonal Coupling §bounded and (2),
l a w ( X , Y ) = S # κ 2 d ( π ) = κ 2 d ( π ∘ σ S ) = κ 2 d ( γ ) . \mathrm{law}(X,Y)=S_{\#}\kappa_{2d}(\pi)=\kappa_{2d}(\pi\circ\sigma_{S})=\kappa_{2d}(\gamma). law ( X , Y ) = S # κ 2 d ( π ) = κ 2 d ( π ∘ σ S ) = κ 2 d ( γ ) .
By Step 3, l a w ( X ) = κ d ( μ ) \mathrm{law}(X)=\kappa_{d}(\mu) law ( X ) = κ d ( μ ) , l a w ( Y ) = κ d ( ν ) \mathrm{law}(Y)=\kappa_{d}(\nu) law ( Y ) = κ d ( ν ) and ∥ X − Y ∥ 2 = W \lVert X-Y\rVert_{2}=W ∥ X − Y ∥ 2 = W ; also X − Y = α − 1 P X-Y=\alpha^{-1}P X − Y = α − 1 P componentwise. By Calculus of Laws of Square-Integrable Tuples: Bounded Tuples, the Lipschitz Bound, Moments, Affine Push-Forwards, Couplings and Embeddings §lipschitz , W ^ 2 ( l a w ( X ′ ) , κ d ( ν ) ) ≤ ∥ X ′ − Y ∥ 2 \widehat{W}_{2}(\mathrm{law}(X'),\kappa_{d}(\nu))\le\lVert X'-Y\rVert_{2} W 2 ( law ( X ′ ) , κ d ( ν )) ≤ ∥ X ′ − Y ∥ 2 and W ^ 2 ( l a w ( X ′ ) , κ d ( μ ) ) ≤ ∥ X ′ − X ∥ 2 \widehat{W}_{2}(\mathrm{law}(X'),\kappa_{d}(\mu))\le\lVert X'-X\rVert_{2} W 2 ( law ( X ′ ) , κ d ( μ )) ≤ ∥ X ′ − X ∥ 2 , and both sides are nonnegative. Since X − Y = α − 1 P X-Y=\alpha^{-1}P X − Y = α − 1 P , one has X ′ − Y = ( X ′ − X ) + ( X − Y ) = ( X ′ − X ) + α − 1 P X'-Y=(X'-X)+(X-Y)=(X'-X)+\alpha^{-1}P X ′ − Y = ( X ′ − X ) + ( X − Y ) = ( X ′ − X ) + α − 1 P and ∥ α − 1 P ∥ 2 2 = ∥ X − Y ∥ 2 2 = W 2 \lVert\alpha^{-1}P\rVert_{2}^{2}=\lVert X-Y\rVert_{2}^{2}=W^{2} ∥ α − 1 P ∥ 2 2 = ∥ X − Y ∥ 2 2 = W 2 ; so (4) with U = X ′ − X U=X'-X U = X ′ − X , V = P V=P V = P and t = α − 1 t=\alpha^{-1} t = α − 1 , together with the symmetry in (4), gives ∥ X ′ − Y ∥ 2 2 = ∥ X ′ − X ∥ 2 2 + 2 α − 1 ⟨ P , X ′ − X ⟩ 2 + α − 2 ∥ P ∥ 2 2 = ∥ X ′ − X ∥ 2 2 + 2 α − 1 ⟨ P , X ′ − X ⟩ 2 + W 2 \lVert X'-Y\rVert_{2}^{2}=\lVert X'-X\rVert_{2}^{2}+2\alpha^{-1}\langle P,X'-X\rangle_{2}+\alpha^{-2}\lVert P\rVert_{2}^{2}=\lVert X'-X\rVert_{2}^{2}+2\alpha^{-1}\langle P,X'-X\rangle_{2}+W^{2} ∥ X ′ − Y ∥ 2 2 = ∥ X ′ − X ∥ 2 2 + 2 α − 1 ⟨ P , X ′ − X ⟩ 2 + α − 2 ∥ P ∥ 2 2 = ∥ X ′ − X ∥ 2 2 + 2 α − 1 ⟨ P , X ′ − X ⟩ 2 + W 2 , where α − 2 ∥ P ∥ 2 2 = ∥ α − 1 P ∥ 2 2 \alpha^{-2}\lVert P\rVert_{2}^{2}=\lVert\alpha^{-1}P\rVert_{2}^{2} α − 2 ∥ P ∥ 2 2 = ∥ α − 1 P ∥ 2 2 by the real homogeneity of the pairing (the case U = 0 U=0 U = 0 of (4)). By Plan Jets and Hamiltonians on Square-Integrable Noncommutative Laws: Standing Notation §lifts and Lifts of Functions on Square-Integrable Noncommutative Laws to Square-Integrable Tuples §lift , φ M ( X ′ ) = φ ( l a w ( X ′ ) ) \varphi_{M}(X')=\varphi(\mathrm{law}(X')) φ M ( X ′ ) = φ ( law ( X ′ )) , so
φ M ( X ′ ) ≤ α 2 ∥ X ′ − Y ∥ 2 2 + β 2 ∥ X ′ − X ∥ 2 2 = α 2 W 2 + ⟨ P , X ′ − X ⟩ 2 + α + β 2 ∥ X ′ − X ∥ 2 2 . \varphi_{M}(X')\le\tfrac{\alpha}{2}\lVert X'-Y\rVert_{2}^{2}+\tfrac{\beta}{2}\lVert X'-X\rVert_{2}^{2}=\tfrac{\alpha}{2}W^{2}+\langle P,X'-X\rangle_{2}+\tfrac{\alpha+\beta}{2}\lVert X'-X\rVert_{2}^{2}. φ M ( X ′ ) ≤ 2 α ∥ X ′ − Y ∥ 2 2 + 2 β ∥ X ′ − X ∥ 2 2 = 2 α W 2 + ⟨ P , X ′ − X ⟩ 2 + 2 α + β ∥ X ′ − X ∥ 2 2 .
As 0 ≤ ∥ X ′ − X ∥ 2 < r 0\le\lVert X'-X\rVert_{2}<r 0 ≤ ∥ X ′ − X ∥ 2 < r , α + β 2 ∥ X ′ − X ∥ 2 2 ≤ α + β 2 r ∥ X ′ − X ∥ 2 = η ∥ X ′ − X ∥ 2 \tfrac{\alpha+\beta}{2}\lVert X'-X\rVert_{2}^{2}\le\tfrac{\alpha+\beta}{2}r\lVert X'-X\rVert_{2}=\eta\lVert X'-X\rVert_{2} 2 α + β ∥ X ′ − X ∥ 2 2 ≤ 2 α + β r ∥ X ′ − X ∥ 2 = η ∥ X ′ − X ∥ 2 . With (3),
φ M ( X ′ ) ≤ φ ( κ d ( μ ) ) + ⟨ P , X ′ − X ⟩ 2 + ( 0 + η ) ∥ X ′ − X ∥ 2 . \varphi_{M}(X')\le\varphi(\kappa_{d}(\mu))+\langle P,X'-X\rangle_{2}+(0+\eta)\lVert X'-X\rVert_{2}. φ M ( X ′ ) ≤ φ ( κ d ( μ )) + ⟨ P , X ′ − X ⟩ 2 + ( 0 + η ) ∥ X ′ − X ∥ 2 .
So the plan κ 2 d ( π ) \kappa_{2d}(\pi) κ 2 d ( π ) at κ d ( μ ) \kappa_{d}(\mu) κ d ( μ ) (Step 2) is a plan superdifferential of φ \varphi φ at κ d ( μ ) \kappa_{d}(\mu) κ d ( μ ) with slack 0 0 0 by Plan Superdifferentials, Plan Subdifferentials and Plan Jets with Slack on Square-Integrable Noncommutative Laws §super , and κ 2 d ( π ) ∈ J + φ ( κ d ( μ ) ) \kappa_{2d}(\pi)\in J^{+}\varphi(\kappa_{d}(\mu)) κ 2 d ( π ) ∈ J + φ ( κ d ( μ )) by Plan Superdifferentials, Plan Subdifferentials and Plan Jets with Slack on Square-Integrable Noncommutative Laws §superjet .
Step 5 (Clause 3). Let η > 0 \eta>0 η > 0 and r = 2 η / ( α + β ) r=2\eta/(\alpha+\beta) r = 2 η / ( α + β ) . Let ( H , M , Ω ) (H,M,\Omega) ( H , M , Ω ) be a tracial W*-probability space and Y , P , Y ′ Y,P,Y' Y , P , Y ′ L 2 L^{2} L 2 d d d -tuples with l a w ( Y , P ) = κ 2 d ( π ′ ) \mathrm{law}(Y,P)=\kappa_{2d}(\pi') law ( Y , P ) = κ 2 d ( π ′ ) and ∥ Y ′ − Y ∥ 2 < r \lVert Y'-Y\rVert_{2}<r ∥ Y ′ − Y ∥ 2 < r . Let X = Y + α − 1 P X=Y+\alpha^{-1}P X = Y + α − 1 P . By Square-Integrable Tuples in a Tracial W*-Probability Space: Their Norm, Affine Images, Pairs, Embedded Images and Laws §operations , S ′ ( Y , P ) S'(Y,P) S ′ ( Y , P ) has components Y j + α − 1 P j Y_{j}+\alpha^{-1}P_{j} Y j + α − 1 P j and Y j Y_{j} Y j in the positions j j j and d + j d+j d + j , so S ′ ( Y , P ) = ( X , Y ) S'(Y,P)=(X,Y) S ′ ( Y , P ) = ( X , Y ) , and as in Step 4, by (2), l a w ( X , Y ) = S # ′ κ 2 d ( π ′ ) = κ 2 d ( π ′ ∘ σ S ′ ) = κ 2 d ( γ ) \mathrm{law}(X,Y)=S'_{\#}\kappa_{2d}(\pi')=\kappa_{2d}(\pi'\circ\sigma_{S'})=\kappa_{2d}(\gamma) law ( X , Y ) = S # ′ κ 2 d ( π ′ ) = κ 2 d ( π ′ ∘ σ S ′ ) = κ 2 d ( γ ) . By Step 3, l a w ( X ) = κ d ( μ ) \mathrm{law}(X)=\kappa_{d}(\mu) law ( X ) = κ d ( μ ) , l a w ( Y ) = κ d ( ν ) \mathrm{law}(Y)=\kappa_{d}(\nu) law ( Y ) = κ d ( ν ) and ∥ X − Y ∥ 2 = W \lVert X-Y\rVert_{2}=W ∥ X − Y ∥ 2 = W , and X − Y = α − 1 P X-Y=\alpha^{-1}P X − Y = α − 1 P componentwise. Since X − Y ′ = ( X − Y ) − ( Y ′ − Y ) X-Y'=(X-Y)-(Y'-Y) X − Y ′ = ( X − Y ) − ( Y ′ − Y ) , (4) gives ∥ X − Y ′ ∥ 2 2 = W 2 − 2 α − 1 ⟨ P , Y ′ − Y ⟩ 2 + ∥ Y ′ − Y ∥ 2 2 \lVert X-Y'\rVert_{2}^{2}=W^{2}-2\alpha^{-1}\langle P,Y'-Y\rangle_{2}+\lVert Y'-Y\rVert_{2}^{2} ∥ X − Y ′ ∥ 2 2 = W 2 − 2 α − 1 ⟨ P , Y ′ − Y ⟩ 2 + ∥ Y ′ − Y ∥ 2 2 . By Calculus of Laws of Square-Integrable Tuples: Bounded Tuples, the Lipschitz Bound, Moments, Affine Push-Forwards, Couplings and Embeddings §lipschitz , W ^ 2 ( κ d ( μ ) , l a w ( Y ′ ) ) ≤ ∥ X − Y ′ ∥ 2 \widehat{W}_{2}(\kappa_{d}(\mu),\mathrm{law}(Y'))\le\lVert X-Y'\rVert_{2} W 2 ( κ d ( μ ) , law ( Y ′ )) ≤ ∥ X − Y ′ ∥ 2 and W ^ 2 ( l a w ( Y ′ ) , κ d ( ν ) ) ≤ ∥ Y ′ − Y ∥ 2 \widehat{W}_{2}(\mathrm{law}(Y'),\kappa_{d}(\nu))\le\lVert Y'-Y\rVert_{2} W 2 ( law ( Y ′ ) , κ d ( ν )) ≤ ∥ Y ′ − Y ∥ 2 . With the lift ψ M ( Y ′ ) = ψ ( l a w ( Y ′ ) ) \psi_{M}(Y')=\psi(\mathrm{law}(Y')) ψ M ( Y ′ ) = ψ ( law ( Y ′ )) of Plan Jets and Hamiltonians on Square-Integrable Noncommutative Laws: Standing Notation §lifts and (3),
ψ M ( Y ′ ) ≥ − α 2 ∥ X − Y ′ ∥ 2 2 − β 2 ∥ Y ′ − Y ∥ 2 2 = ψ ( κ d ( ν ) ) + ⟨ P , Y ′ − Y ⟩ 2 − α + β 2 ∥ Y ′ − Y ∥ 2 2 ≥ ψ ( κ d ( ν ) ) + ⟨ P , Y ′ − Y ⟩ 2 − ( 0 + η ) ∥ Y ′ − Y ∥ 2 , \psi_{M}(Y')\ge-\tfrac{\alpha}{2}\lVert X-Y'\rVert_{2}^{2}-\tfrac{\beta}{2}\lVert Y'-Y\rVert_{2}^{2}=\psi(\kappa_{d}(\nu))+\langle P,Y'-Y\rangle_{2}-\tfrac{\alpha+\beta}{2}\lVert Y'-Y\rVert_{2}^{2}\ge\psi(\kappa_{d}(\nu))+\langle P,Y'-Y\rangle_{2}-(0+\eta)\lVert Y'-Y\rVert_{2}, ψ M ( Y ′ ) ≥ − 2 α ∥ X − Y ′ ∥ 2 2 − 2 β ∥ Y ′ − Y ∥ 2 2 = ψ ( κ d ( ν )) + ⟨ P , Y ′ − Y ⟩ 2 − 2 α + β ∥ Y ′ − Y ∥ 2 2 ≥ ψ ( κ d ( ν )) + ⟨ P , Y ′ − Y ⟩ 2 − ( 0 + η ) ∥ Y ′ − Y ∥ 2 ,
the last step as in Step 4. Since κ 2 d ( π ′ ) \kappa_{2d}(\pi') κ 2 d ( π ′ ) is a plan at κ d ( ν ) \kappa_{d}(\nu) κ d ( ν ) (Step 2), Plan Superdifferentials, Plan Subdifferentials and Plan Jets with Slack on Square-Integrable Noncommutative Laws §sub and Plan Superdifferentials, Plan Subdifferentials and Plan Jets with Slack on Square-Integrable Noncommutative Laws §subjet give κ 2 d ( π ′ ) ∈ J − ψ ( κ d ( ν ) ) \kappa_{2d}(\pi')\in J^{-}\psi(\kappa_{d}(\nu)) κ 2 d ( π ′ ) ∈ J − ψ ( κ d ( ν )) .
Step 6 (Clause 4). Apply Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §isometry to γ ∈ Σ 2 d \gamma\in\Sigma_{2d} γ ∈ Σ 2 d and the 2 d 2d 2 d -tuple a T a^{T} a T in P 2 d , s a \mathcal{P}_{2d,\mathrm{sa}} P 2 d , sa (Step 1), whose marginal law is γ ∘ σ T = π \gamma\circ\sigma_{T}=\pi γ ∘ σ T = π : there is U ∈ L ( H π , H γ ) U\in\mathcal{L}(\mathcal{H}_{\pi},\mathcal{H}_{\gamma}) U ∈ L ( H π , H γ ) with U p ^ π = σ T ( p ) ^ γ U\widehat{p}^{\,\pi}=\widehat{\sigma_{T}(p)}^{\,\gamma} U p π = σ T ( p ) γ for all p ∈ P 2 d p\in\mathcal{P}_{2d} p ∈ P 2 d and U ∗ U = I U^{*}U=I U ∗ U = I , so that ⟨ U ξ , U ξ ′ ⟩ = ⟨ ξ , U ∗ U ξ ′ ⟩ = ⟨ ξ , ξ ′ ⟩ \langle U\xi,U\xi'\rangle=\langle\xi,U^{*}U\xi'\rangle=\langle\xi,\xi'\rangle ⟨ U ξ , U ξ ′ ⟩ = ⟨ ξ , U ∗ U ξ ′ ⟩ = ⟨ ξ , ξ ′ ⟩ for ξ , ξ ′ ∈ H π \xi,\xi'\in\mathcal{H}_{\pi} ξ , ξ ′ ∈ H π . Since π ∘ ι 1 = μ \pi\circ\iota^{1}=\mu π ∘ ι 1 = μ , Marginal Isometries, Bounded Plans and Displacement Pairings for Noncommutative Laws §isometries gives V π 1 ∈ L ( H μ , H π ) V^{1}_{\pi}\in\mathcal{L}(\mathcal{H}_{\mu},\mathcal{H}_{\pi}) V π 1 ∈ L ( H μ , H π ) with V π 1 q ^ μ = ι 1 ( q ) ^ π V^{1}_{\pi}\widehat{q}^{\,\mu}=\widehat{\iota^{1}(q)}^{\,\pi} V π 1 q μ = ι 1 ( q ) π , so U V π 1 ∈ L ( H μ , H γ ) UV^{1}_{\pi}\in\mathcal{L}(\mathcal{H}_{\mu},\mathcal{H}_{\gamma}) U V π 1 ∈ L ( H μ , H γ ) and, by (2), U V π 1 q ^ μ = σ T ( ι 1 ( q ) ) ^ γ = ι 1 ( q ) ^ γ UV^{1}_{\pi}\widehat{q}^{\,\mu}=\widehat{\sigma_{T}(\iota^{1}(q))}^{\,\gamma}=\widehat{\iota^{1}(q)}^{\,\gamma} U V π 1 q μ = σ T ( ι 1 ( q )) γ = ι 1 ( q ) γ for every q ∈ P d q\in\mathcal{P}_{d} q ∈ P d . As γ ∘ ι 1 = μ \gamma\circ\iota^{1}=\mu γ ∘ ι 1 = μ , V γ 1 V^{1}_{\gamma} V γ 1 is by Marginal Isometries, Bounded Plans and Displacement Pairings for Noncommutative Laws §isometries the operator 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 \gamma γ and ( x 1 , … , x d ) (x_{1},\dots,x_{d}) ( x 1 , … , x d ) , the only element of L ( H μ , H γ ) \mathcal{L}(\mathcal{H}_{\mu},\mathcal{H}_{\gamma}) L ( H μ , H γ ) with q ^ μ ↦ ι 1 ( q ) ^ γ \widehat{q}^{\,\mu}\mapsto\widehat{\iota^{1}(q)}^{\,\gamma} q μ ↦ ι 1 ( q ) γ for all q q q ; hence U V π 1 = V γ 1 UV^{1}_{\pi}=V^{1}_{\gamma} U V π 1 = V γ 1 . By Substitution is the Unique Unital Homomorphism with Prescribed Values on the Variables: Monomials, Products, Adjoints and Composition §values , σ T ( x d + j ) = a d + j T = α h j \sigma_{T}(x_{d+j})=a^{T}_{d+j}=\alpha h_{j} σ T ( x d + j ) = a d + j T = α h j , 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 §vacuum 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 §algebra for γ ∈ Σ 2 d , R \gamma\in\Sigma_{2d,R} γ ∈ Σ 2 d , R , α h j ^ = L α h j Ω γ = α L h j Ω γ = α h j ^ \widehat{\alpha h_{j}}=L_{\alpha h_{j}}\Omega_{\gamma}=\alpha L_{h_{j}}\Omega_{\gamma}=\alpha\widehat{h_{j}} α h j = L α h j Ω γ = α L h j Ω γ = α h j . Hence U x d + j ^ π = α h j ^ γ U\widehat{x_{d+j}}^{\,\pi}=\alpha\widehat{h_{j}}^{\,\gamma} U x d + j π = α h j γ , and for ζ ∈ H μ d \zeta\in\mathcal{H}_{\mu}^{d} ζ ∈ H μ d , by linearity of the inner product in its second argument,
⟨ V π 1 ζ j , x d + j ^ ⟩ H π = ⟨ U V π 1 ζ j , U x d + j ^ ⟩ H γ = α ⟨ V γ 1 ζ j , h j ^ ⟩ H γ . \bigl\langle V^{1}_{\pi}\zeta_{j},\widehat{x_{d+j}}\bigr\rangle_{\mathcal{H}_{\pi}}=\bigl\langle UV^{1}_{\pi}\zeta_{j},U\widehat{x_{d+j}}\bigr\rangle_{\mathcal{H}_{\gamma}}=\alpha\bigl\langle V^{1}_{\gamma}\zeta_{j},\widehat{h_{j}}\bigr\rangle_{\mathcal{H}_{\gamma}}. ⟨ V π 1 ζ j , x d + j ⟩ H π = ⟨ U V π 1 ζ j , U x d + j ⟩ H γ = α ⟨ V γ 1 ζ j , h j ⟩ H γ .
Taking real parts (α \alpha α is real) and summing over j j j , Marginal Isometries, Bounded Plans and Displacement Pairings for Noncommutative Laws §plan-pairing and Marginal Isometries, Bounded Plans and Displacement Pairings for Noncommutative Laws §coupling-pairing give J ( ζ , π ) = α J γ 1 ( ζ ) \mathcal{J}(\zeta,\pi)=\alpha\mathcal{J}^{1}_{\gamma}(\zeta) J ( ζ , π ) = α J γ 1 ( ζ ) . For π ′ \pi' π ′ the same argument with a T ′ a^{T'} a T ′ gives U ′ ∈ L ( H π ′ , H γ ) U'\in\mathcal{L}(\mathcal{H}_{\pi'},\mathcal{H}_{\gamma}) U ′ ∈ L ( H π ′ , H γ ) with U ′ ∗ U ′ = I U'^{*}U'=I U ′ ∗ U ′ = I ; now π ′ ∘ ι 1 = ν = γ ∘ ι 2 \pi'\circ\iota^{1}=\nu=\gamma\circ\iota^{2} π ′ ∘ ι 1 = ν = γ ∘ ι 2 and, by (2), U ′ V π ′ 1 q ^ ν = σ T ′ ( ι 1 ( q ) ) ^ γ = ι 2 ( q ) ^ γ U'V^{1}_{\pi'}\widehat{q}^{\,\nu}=\widehat{\sigma_{T'}(\iota^{1}(q))}^{\,\gamma}=\widehat{\iota^{2}(q)}^{\,\gamma} U ′ V π ′ 1 q ν = σ T ′ ( ι 1 ( q )) γ = ι 2 ( q ) γ , so the uniqueness in Marginals of a Noncommutative Law: the Isometry of GNS Spaces, the Trace-Preserving Embedding of Tracial Algebras and the Conditional Expectation §isometry for γ \gamma γ and ( x d + 1 , … , x 2 d ) (x_{d+1},\dots,x_{2d}) ( x d + 1 , … , x 2 d ) gives U ′ V π ′ 1 = V γ 2 U'V^{1}_{\pi'}=V^{2}_{\gamma} U ′ V π ′ 1 = V γ 2 ; and U ′ x d + j ^ π ′ = α h j ^ γ U'\widehat{x_{d+j}}^{\,\pi'}=\alpha\widehat{h_{j}}^{\,\gamma} U ′ x d + j π ′ = α h j γ since a d + j T ′ = α h j a^{T'}_{d+j}=\alpha h_{j} a d + j T ′ = α h j . Hence J ( ζ ′ , π ′ ) = α J γ 2 ( ζ ′ ) \mathcal{J}(\zeta',\pi')=\alpha\mathcal{J}^{2}_{\gamma}(\zeta') J ( ζ ′ , π ′ ) = α J γ 2 ( ζ ′ ) for ζ ′ ∈ H ν d \zeta'\in\mathcal{H}_{\nu}^{d} ζ ′ ∈ H ν d .
Step 7 (Clause 5). Let A γ = { L p : p ∈ P 2 d } \mathcal{A}_{\gamma}=\{L_{p}:\ p\in\mathcal{P}_{2d}\} A γ = { L p : p ∈ P 2 d } as in The Tracial Algebra of a Noncommutative Law is a Tracial W*-Probability Space: the W*-Closure of the Left Multiplications . By The Tracial Algebra of a Noncommutative Law is a Tracial W*-Probability Space: the W*-Closure of the Left Multiplications §w-star for γ ∈ Σ 2 d \gamma\in\Sigma_{2d} γ ∈ Σ 2 d , ( K , N , Ψ ) = ( H γ , M γ , Ω γ ) (K,N,\Psi)=(\mathcal{H}_{\gamma},\mathcal{M}_{\gamma},\Omega_{\gamma}) ( K , N , Ψ ) = ( H γ , M γ , Ω γ ) is a tracial W*-probability space with M γ = A γ ′ ′ \mathcal{M}_{\gamma}=\mathcal{A}_{\gamma}'' M γ = A γ ′′ . Let s = ( L x 1 , … , L x 2 d ) s=(L_{x_{1}},\dots,L_{x_{2d}}) s = ( L x 1 , … , L x 2 d ) . Each L x i L_{x_{i}} L x i lies in A γ ⊆ A γ ′ ′ = M γ \mathcal{A}_{\gamma}\subseteq\mathcal{A}_{\gamma}''=\mathcal{M}_{\gamma} A γ ⊆ A γ ′′ = M γ by Basic Properties of Commutants: Unital Algebras Closed under Weak Limits, Order Reversal, the Triple Commutant, and Conjugation §order , and is self-adjoint 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 §adjoint , since 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 s s s is a self-adjoint 2 d 2d 2 d -tuple in N N N (Self-Adjoint Tuples in a Tracial W*-Probability Space and Their Laws §tuple ), and its law is λ s = γ \lambda_{s}=\gamma λ s = γ by Self-Adjoint Tuples in a Tracial W*-Probability Space and Their Laws §law and The Tracial Algebra of a Noncommutative Law is a Tracial W*-Probability Space: the W*-Closure of the Left Multiplications §law . By Calculus of Laws of Square-Integrable Tuples: Bounded Tuples, the Lipschitz Bound, Moments, Affine Push-Forwards, Couplings and Embeddings §bounded , s Ψ s\Psi s Ψ is an L 2 L^{2} L 2 2 d 2d 2 d -tuple of ( K , N , Ψ ) (K,N,\Psi) ( K , N , Ψ ) with l a w ( s Ψ ) = κ 2 d ( γ ) \mathrm{law}(s\Psi)=\kappa_{2d}(\gamma) law ( s Ψ ) = κ 2 d ( γ ) . Let X = ( L x 1 Ψ , … , L x d Ψ ) X=(L_{x_{1}}\Psi,\dots,L_{x_{d}}\Psi) X = ( L x 1 Ψ , … , L x d Ψ ) and Y = ( L x d + 1 Ψ , … , L x 2 d Ψ ) Y=(L_{x_{d+1}}\Psi,\dots,L_{x_{2d}}\Psi) Y = ( L x d + 1 Ψ , … , L x 2 d Ψ ) , L 2 L^{2} L 2 d d d -tuples of ( K , N , Ψ ) (K,N,\Psi) ( K , N , Ψ ) with ( X , Y ) = s Ψ (X,Y)=s\Psi ( X , Y ) = s Ψ . By Step 3, ∥ X − Y ∥ 2 = W \lVert X-Y\rVert_{2}=W ∥ X − Y ∥ 2 = W , and l a w ( X ) = κ d ( μ ) \mathrm{law}(X)=\kappa_{d}(\mu) law ( X ) = κ d ( μ ) , l a w ( Y ) = κ d ( ν ) \mathrm{law}(Y)=\kappa_{d}(\nu) law ( Y ) = κ d ( ν ) ; so by Calculus of Laws of Square-Integrable Tuples: Bounded Tuples, the Lipschitz Bound, Moments, Affine Push-Forwards, Couplings and Embeddings §moments and Calculus of Square-Integrable Noncommutative Laws: Agreement on Bounded Laws, Lipschitz Estimates, Functoriality of Push-Forwards, Moment Formulas, Positivity, the Cost and the Diagonal Coupling §bounded , ∥ X ∥ 2 2 = M ^ ( κ d ( μ ) ) = M ( μ ) \lVert X\rVert_{2}^{2}=\widehat{M}(\kappa_{d}(\mu))=M(\mu) ∥ X ∥ 2 2 = M ( κ d ( μ )) = M ( μ ) and ∥ Y ∥ 2 2 = M ^ ( κ d ( ν ) ) = M ( ν ) \lVert Y\rVert_{2}^{2}=\widehat{M}(\kappa_{d}(\nu))=M(\nu) ∥ Y ∥ 2 2 = M ( κ d ( ν )) = M ( ν ) , where M ( μ ) = ∑ j = 1 d μ ( x j x j ) M(\mu)=\sum_{j=1}^{d}\mu(x_{j}x_{j}) M ( μ ) = ∑ j = 1 d μ ( x j x j ) is the second moment of Noncommutative Laws, Couplings and the Wasserstein Distance: Standing Notation §laws , as in the statement. By Square-Integrable Tuples in a Tracial W*-Probability Space: Their Norm, Affine Images, Pairs, Embedded Images and Laws §operations , T ( X , Y ) T(X,Y) T ( X , Y ) has components X j X_{j} X j and α X j − α Y j \alpha X_{j}-\alpha Y_{j} α X j − α Y j in the positions j j j and d + j d+j d + j , so T ( X , Y ) = ( X , α ( X − Y ) ) T(X,Y)=(X,\alpha(X-Y)) T ( X , Y ) = ( X , α ( X − Y )) , and likewise T ′ ( X , Y ) = ( Y , α ( X − Y ) ) T'(X,Y)=(Y,\alpha(X-Y)) T ′ ( X , Y ) = ( Y , α ( X − Y )) . By Calculus of Laws of Square-Integrable Tuples: Bounded Tuples, the Lipschitz Bound, Moments, Affine Push-Forwards, Couplings and Embeddings §push-forward and Calculus of Square-Integrable Noncommutative Laws: Agreement on Bounded Laws, Lipschitz Estimates, Functoriality of Push-Forwards, Moment Formulas, Positivity, the Cost and the Diagonal Coupling §bounded ,
l a w ( X , α ( X − Y ) ) = T # κ 2 d ( γ ) = κ 2 d ( γ ∘ σ T ) = κ 2 d ( π ) , l a w ( Y , α ( X − Y ) ) = T # ′ κ 2 d ( γ ) = κ 2 d ( π ′ ) . \mathrm{law}(X,\alpha(X-Y))=T_{\#}\kappa_{2d}(\gamma)=\kappa_{2d}(\gamma\circ\sigma_{T})=\kappa_{2d}(\pi),\qquad\mathrm{law}(Y,\alpha(X-Y))=T'_{\#}\kappa_{2d}(\gamma)=\kappa_{2d}(\pi'). law ( X , α ( X − Y )) = T # κ 2 d ( γ ) = κ 2 d ( γ ∘ σ T ) = κ 2 d ( π ) , law ( Y , α ( X − Y )) = T # ′ κ 2 d ( γ ) = κ 2 d ( π ′ ) .
This proves clause 5.