Each result cited is universally quantified over the data in its own statement.
We work in the setting of The Intrinsic Calculus on the Wasserstein Space: Standing Notation ; elementary order and arithmetic of real numbers is carried by The Real Numbers: Standing Notation and Background §background , which we adopt for that purpose. By The Discounted Hamilton-Jacobi Equation with Common Noise and a Penalty Drift on the Wasserstein Space §operator , F F F is a second-order equation operator over D Σ \mathcal{D}_{\Sigma} D Σ with
F ( ν , r , q , Y ) = λ 0 r − 1 2 t r ( Γ ⊤ Γ Y ) + θ 2 ∥ q ∥ ν 2 + ⟨ Σ ( ν ) , q ⟩ ν − g ( ν ) , F(\nu,r,q,Y)=\lambda_{0}\,r-\frac{1}{2}\,\mathrm{tr}\bigl(\Gamma^{\top}\Gamma Y\bigr)+\frac{\theta}{2}\,\lVert q\rVert_{\nu}^{2}+\langle\Sigma(\nu),q\rangle_{\nu}-g(\nu), F ( ν , r , q , Y ) = λ 0 r − 2 1 tr ( Γ ⊤ Γ Y ) + 2 θ ∥ q ∥ ν 2 + ⟨ Σ ( ν ) , q ⟩ ν − g ( ν ) ,
and its δ \delta δ -shifts are given by The Bundle of Vector Fields over a Set of Measures, Second-Order Equation Operators on the Wasserstein Space, and Their Delta-Shifts §shifted .
Claim 1. Let δ , R , ν , r , Y , q , q ′ \delta,R,\nu,r,Y,q,q' δ , R , ν , r , Y , q , q ′ be as in the claim, and write σ = Σ ( ν ) \sigma=\Sigma(\nu) σ = Σ ( ν ) , which lies in T ν ⊆ L 2 ( ν ; R d ) T_{\nu}\subseteq L^{2}(\nu;\mathbb{R}^{d}) T ν ⊆ L 2 ( ν ; R d ) by Penalty Pairs on the Wasserstein Space: the Penalty, Its Score, and Their Domains §pair . By Square-Integrable Vector Fields Against a Probability Measure on Euclidean Space, and Test Functions: Standing Notation §l2mu , L 2 ( ν ; R d ) L^{2}(\nu;\mathbb{R}^{d}) L 2 ( ν ; R d ) is a real Hilbert space, in particular a real inner product space with inner product ⟨ ⋅ , ⋅ ⟩ ν \langle\cdot,\cdot\rangle_{\nu} ⟨ ⋅ , ⋅ ⟩ ν and norm ∥ ⋅ ∥ ν \lVert\cdot\rVert_{\nu} ∥ ⋅ ∥ ν , the norm satisfying ∥ x ∥ ν 2 = ⟨ x , x ⟩ ν \lVert x\rVert_{\nu}^{2}=\langle x,x\rangle_{\nu} ∥ x ∥ ν 2 = ⟨ x , x ⟩ ν by Real Inner Product Space §norm .
Step 1 (an identity for squared norms). For w , w ′ ∈ L 2 ( ν ; R d ) w,w'\in L^{2}(\nu;\mathbb{R}^{d}) w , w ′ ∈ L 2 ( ν ; R d ) , Elementary Identities in a Real Inner Product Space §bilinear and symmetry of the inner product (condition (a) of Real Inner Product Space §inner-product ) give
⟨ w − w ′ , w + w ′ ⟩ ν = ⟨ w , w ⟩ ν + ⟨ w , w ′ ⟩ ν − ⟨ w ′ , w ⟩ ν − ⟨ w ′ , w ′ ⟩ ν = ∥ w ∥ ν 2 − ∥ w ′ ∥ ν 2 . \langle w-w',w+w'\rangle_{\nu}=\langle w,w\rangle_{\nu}+\langle w,w'\rangle_{\nu}-\langle w',w\rangle_{\nu}-\langle w',w'\rangle_{\nu}=\lVert w\rVert_{\nu}^{2}-\lVert w'\rVert_{\nu}^{2}. ⟨ w − w ′ , w + w ′ ⟩ ν = ⟨ w , w ⟩ ν + ⟨ w , w ′ ⟩ ν − ⟨ w ′ , w ⟩ ν − ⟨ w ′ , w ′ ⟩ ν = ∥ w ∥ ν 2 − ∥ w ′ ∥ ν 2 .
Step 2 (the difference for F δ − F^{-}_{\delta} F δ − ). Put w = q + δ σ w=q+\delta\sigma w = q + δ σ and w ′ = q ′ + δ σ w'=q'+\delta\sigma w ′ = q ′ + δ σ , so that w − w ′ = q − q ′ w-w'=q-q' w − w ′ = q − q ′ and w + w ′ = q + q ′ + 2 δ σ w+w'=q+q'+2\delta\sigma w + w ′ = q + q ′ + 2 δ σ in the vector space L 2 ( ν ; R d ) L^{2}(\nu;\mathbb{R}^{d}) L 2 ( ν ; R d ) . By The Bundle of Vector Fields over a Set of Measures, Second-Order Equation Operators on the Wasserstein Space, and Their Delta-Shifts §shifted , F δ − ( ν , r , q , Y ) F^{-}_{\delta}(\nu,r,q,Y) F δ − ( ν , r , q , Y ) and F δ − ( ν , r , q ′ , Y ) F^{-}_{\delta}(\nu,r,q',Y) F δ − ( ν , r , q ′ , Y ) are the values of F F F at ( ν , r + δ E ( ν ) , w , Y + δ H E ( ν ) ) (\nu,r+\delta\mathcal{E}(\nu),w,Y+\delta H_{\mathcal{E}}(\nu)) ( ν , r + δ E ( ν ) , w , Y + δ H E ( ν )) and at ( ν , r + δ E ( ν ) , w ′ , Y + δ H E ( ν ) ) (\nu,r+\delta\mathcal{E}(\nu),w',Y+\delta H_{\mathcal{E}}(\nu)) ( ν , r + δ E ( ν ) , w ′ , Y + δ H E ( ν )) . The terms λ 0 ( r + δ E ( ν ) ) \lambda_{0}(r+\delta\mathcal{E}(\nu)) λ 0 ( r + δ E ( ν )) , − 1 2 t r ( Γ ⊤ Γ ( Y + δ H E ( ν ) ) ) -\tfrac12\mathrm{tr}(\Gamma^{\top}\Gamma(Y+\delta H_{\mathcal{E}}(\nu))) − 2 1 tr ( Γ ⊤ Γ ( Y + δ H E ( ν ))) and − g ( ν ) -g(\nu) − g ( ν ) are the same in both values and cancel in the difference, so by Step 1 and Elementary Identities in a Real Inner Product Space §bilinear ,
F δ − ( ν , r , q , Y ) − F δ − ( ν , r , q ′ , Y ) = θ 2 ( ∥ w ∥ ν 2 − ∥ w ′ ∥ ν 2 ) + ⟨ σ , w − w ′ ⟩ ν = θ 2 ⟨ q − q ′ , q + q ′ + 2 δ σ ⟩ ν + ⟨ σ , q − q ′ ⟩ ν . F^{-}_{\delta}(\nu,r,q,Y)-F^{-}_{\delta}(\nu,r,q',Y)=\frac{\theta}{2}\bigl(\lVert w\rVert_{\nu}^{2}-\lVert w'\rVert_{\nu}^{2}\bigr)+\langle\sigma,w-w'\rangle_{\nu}=\frac{\theta}{2}\,\langle q-q',\,q+q'+2\delta\sigma\rangle_{\nu}+\langle\sigma,q-q'\rangle_{\nu}. F δ − ( ν , r , q , Y ) − F δ − ( ν , r , q ′ , Y ) = 2 θ ( ∥ w ∥ ν 2 − ∥ w ′ ∥ ν 2 ) + ⟨ σ , w − w ′ ⟩ ν = 2 θ ⟨ q − q ′ , q + q ′ + 2 δ σ ⟩ ν + ⟨ σ , q − q ′ ⟩ ν .
Step 3 (the estimate). By the triangle inequality The Norm Metric of a Real Inner Product Space: Triangle Inequalities, Limits and Continuity §triangle and homogeneity Elementary Identities in a Real Inner Product Space §homogeneity , using δ > 0 \delta>0 δ > 0 ,
∥ q + q ′ + 2 δ σ ∥ ν ≤ ∥ q ∥ ν + ∥ q ′ ∥ ν + 2 δ ∥ σ ∥ ν ≤ 2 R + 2 δ R = 2 ( 1 + δ ) R . \lVert q+q'+2\delta\sigma\rVert_{\nu}\le\lVert q\rVert_{\nu}+\lVert q'\rVert_{\nu}+2\delta\lVert\sigma\rVert_{\nu}\le2R+2\delta R=2(1+\delta)R . ∥ q + q ′ + 2 δ σ ∥ ν ≤ ∥ q ∥ ν + ∥ q ′ ∥ ν + 2 δ ∥ σ ∥ ν ≤ 2 R + 2 δ R = 2 ( 1 + δ ) R .
Taking absolute values in Step 2, with the triangle inequality for the absolute value and the Cauchy-Schwarz inequality in L 2 ( ν ; R d ) L^{2}(\nu;\mathbb{R}^{d}) L 2 ( ν ; R d ) applied to each inner product, and using θ > 0 \theta>0 θ > 0 and ∥ σ ∥ ν ≤ R \lVert\sigma\rVert_{\nu}\le R ∥ σ ∥ ν ≤ R ,
∣ F δ − ( ν , r , q , Y ) − F δ − ( ν , r , q ′ , Y ) ∣ ≤ θ 2 ∥ q − q ′ ∥ ν ⋅ 2 ( 1 + δ ) R + R ∥ q − q ′ ∥ ν = ( θ ( 1 + δ ) + 1 ) R ∥ q − q ′ ∥ ν . \bigl|F^{-}_{\delta}(\nu,r,q,Y)-F^{-}_{\delta}(\nu,r,q',Y)\bigr|\le\frac{\theta}{2}\,\lVert q-q'\rVert_{\nu}\cdot2(1+\delta)R+R\,\lVert q-q'\rVert_{\nu}=\bigl(\theta(1+\delta)+1\bigr)R\,\lVert q-q'\rVert_{\nu}. F δ − ( ν , r , q , Y ) − F δ − ( ν , r , q ′ , Y ) ≤ 2 θ ∥ q − q ′ ∥ ν ⋅ 2 ( 1 + δ ) R + R ∥ q − q ′ ∥ ν = ( θ ( 1 + δ ) + 1 ) R ∥ q − q ′ ∥ ν .
Since 0 < δ < 1 0<\delta<1 0 < δ < 1 and θ > 0 \theta>0 θ > 0 , we have θ ( 1 + δ ) + 1 ≤ 2 θ + 1 \theta(1+\delta)+1\le2\theta+1 θ ( 1 + δ ) + 1 ≤ 2 θ + 1 , and as R ∥ q − q ′ ∥ ν ≥ 0 R\lVert q-q'\rVert_{\nu}\ge0 R ∥ q − q ′ ∥ ν ≥ 0 the right-hand side is at most ( 2 θ + 1 ) R ∥ q − q ′ ∥ ν (2\theta+1)R\lVert q-q'\rVert_{\nu} ( 2 θ + 1 ) R ∥ q − q ′ ∥ ν . This is the first inequality.
Step 4 (the difference for F δ + F^{+}_{\delta} F δ + ). Put w = q − δ σ w=q-\delta\sigma w = q − δ σ and w ′ = q ′ − δ σ w'=q'-\delta\sigma w ′ = q ′ − δ σ , so that w − w ′ = q − q ′ w-w'=q-q' w − w ′ = q − q ′ and w + w ′ = q + q ′ − 2 δ σ w+w'=q+q'-2\delta\sigma w + w ′ = q + q ′ − 2 δ σ . By The Bundle of Vector Fields over a Set of Measures, Second-Order Equation Operators on the Wasserstein Space, and Their Delta-Shifts §shifted the two values are those of F F F at ( ν , r − δ E ( ν ) , w , Y − δ H E ( ν ) ) (\nu,r-\delta\mathcal{E}(\nu),w,Y-\delta H_{\mathcal{E}}(\nu)) ( ν , r − δ E ( ν ) , w , Y − δ H E ( ν )) and ( ν , r − δ E ( ν ) , w ′ , Y − δ H E ( ν ) ) (\nu,r-\delta\mathcal{E}(\nu),w',Y-\delta H_{\mathcal{E}}(\nu)) ( ν , r − δ E ( ν ) , w ′ , Y − δ H E ( ν )) ; exactly as in Step 2,
F δ + ( ν , r , q , Y ) − F δ + ( ν , r , q ′ , Y ) = θ 2 ⟨ q − q ′ , q + q ′ − 2 δ σ ⟩ ν + ⟨ σ , q − q ′ ⟩ ν , F^{+}_{\delta}(\nu,r,q,Y)-F^{+}_{\delta}(\nu,r,q',Y)=\frac{\theta}{2}\,\langle q-q',\,q+q'-2\delta\sigma\rangle_{\nu}+\langle\sigma,q-q'\rangle_{\nu}, F δ + ( ν , r , q , Y ) − F δ + ( ν , r , q ′ , Y ) = 2 θ ⟨ q − q ′ , q + q ′ − 2 δ σ ⟩ ν + ⟨ σ , q − q ′ ⟩ ν ,
and The Norm Metric of a Real Inner Product Space: Triangle Inequalities, Limits and Continuity §triangle with Elementary Identities in a Real Inner Product Space §homogeneity gives ∥ q + q ′ − 2 δ σ ∥ ν ≤ 2 ( 1 + δ ) R \lVert q+q'-2\delta\sigma\rVert_{\nu}\le2(1+\delta)R ∥ q + q ′ − 2 δ σ ∥ ν ≤ 2 ( 1 + δ ) R . The estimate of Step 3, with Cauchy-Schwarz , gives the second inequality.
Claim 2. We verify Equation Operators on the Wasserstein Space with Momentum-Continuous Shifts §momentum , whose hypotheses on F F F hold since F F F is a second-order equation operator over D Σ \mathcal{D}_{\Sigma} D Σ . Let δ , R , η ∈ R \delta,R,\eta\in\mathbb{R} δ , R , η ∈ R with 0 < δ < 1 0<\delta<1 0 < δ < 1 , 0 < R 0<R 0 < R and 0 < η 0<\eta 0 < η be given; then ( 2 θ + 1 ) R (2\theta+1)R ( 2 θ + 1 ) R is positive, and we choose
ρ = η ( 2 θ + 1 ) R , \rho=\frac{\eta}{(2\theta+1)R}, ρ = ( 2 θ + 1 ) R η ,
which is positive. Let ν ∈ D Σ \nu\in\mathcal{D}_{\Sigma} ν ∈ D Σ with ∥ Σ ( ν ) ∥ ν ≤ R \lVert\Sigma(\nu)\rVert_{\nu}\le R ∥ Σ ( ν ) ∥ ν ≤ R , r ∈ R r\in\mathbb{R} r ∈ R with ∣ r ∣ ≤ R |r|\le R ∣ r ∣ ≤ R , q , q ′ ∈ L 2 ( ν ; R d ) q,q'\in L^{2}(\nu;\mathbb{R}^{d}) q , q ′ ∈ L 2 ( ν ; R d ) with ∥ q ∥ ν ≤ R \lVert q\rVert_{\nu}\le R ∥ q ∥ ν ≤ R , ∥ q ′ ∥ ν ≤ R \lVert q'\rVert_{\nu}\le R ∥ q ′ ∥ ν ≤ R and ∥ q − q ′ ∥ ν < ρ \lVert q-q'\rVert_{\nu}<\rho ∥ q − q ′ ∥ ν < ρ , and Y ∈ S ( d ) Y\in\mathcal{S}(d) Y ∈ S ( d ) with ∥ Y ∥ ≤ R \lVert Y\rVert\le R ∥ Y ∥ ≤ R . Claim 1 applies to these data (the bounds on r r r and Y Y Y are not needed) and gives
∣ F δ ∓ ( ν , r , q , Y ) − F δ ∓ ( ν , r , q ′ , Y ) ∣ ≤ ( 2 θ + 1 ) R ∥ q − q ′ ∥ ν < ( 2 θ + 1 ) R ρ = η \bigl|F^{\mp}_{\delta}(\nu,r,q,Y)-F^{\mp}_{\delta}(\nu,r,q',Y)\bigr|\le(2\theta+1)R\,\lVert q-q'\rVert_{\nu}<(2\theta+1)R\,\rho=\eta F δ ∓ ( ν , r , q , Y ) − F δ ∓ ( ν , r , q ′ , Y ) ≤ ( 2 θ + 1 ) R ∥ q − q ′ ∥ ν < ( 2 θ + 1 ) R ρ = η
for both shifts, the strict inequality because ( 2 θ + 1 ) R (2\theta+1)R ( 2 θ + 1 ) R is positive and ∥ q − q ′ ∥ ν < ρ \lVert q-q'\rVert_{\nu}<\rho ∥ q − q ′ ∥ ν < ρ . Hence F F F has momentum-continuous shifts relative to the pair. ■ \blacksquare ■