Each result cited below is universally quantified over the data appearing in its own statement, and is applied here to the data named at the point of use. We write ∣ ⋅ ∣ H |\cdot|_{H} ∣ ⋅ ∣ H and d H d_{H} d H for the norm and distance of H H H , so that d H ( x , y ) = ∣ x − y ∣ H d_{H}(x,y)=|x-y|_{H} d H ( x , y ) = ∣ x − y ∣ H by Real Inner Product Space §distance , and we use the triangle inequality of The Norm Metric of a Real Inner Product Space: Triangle Inequalities, Limits and Continuity §triangle and the symmetry of d H d_{H} d H from the metric axioms without further mention. Every member of S \mathcal{S} S , being a viscosity subsolution of F F F on U U U , is bounded above near each point of U U U , so its δ \delta δ -envelopes are defined for every real δ > 0 \delta>0 δ > 0 .
Claim 1. Let x ∈ U x\in U x ∈ U and let c ∈ R c\in\mathbb{R} c ∈ R and the positive r ∈ R r\in\mathbb{R} r ∈ R be as supplied for x x x by the hypothesis The Pointwise Supremum of a Locally Uniformly Bounded Family of Viscosity Subsolutions on a Hilbert Triple is a Viscosity Subsolution §locally-bounded . For y ∈ U y\in U y ∈ U with d H ( y , x ) ≤ r d_{H}(y,x)\le r d H ( y , x ) ≤ r , the number c c c is an upper bound of { v ( y ) : v ∈ S } \{v(y):v\in\mathcal{S}\} { v ( y ) : v ∈ S } , whose least upper bound is u ( y ) u(y) u ( y ) ; hence u ( y ) ≤ c u(y)\le c u ( y ) ≤ c . Thus c c c belongs to the set A u ( x ) A_{u}(x) A u ( x ) of Upper and Lower Semicontinuous Envelopes of a Real-Valued Function , and as x x x was arbitrary, u u u is bounded above near each point of U U U by Upper and Lower Semicontinuous Envelopes of a Real-Valued Function §near-bounds . For v ∈ S v\in\mathcal{S} v ∈ S and x ∈ U x\in U x ∈ U , v ( x ) v(x) v ( x ) belongs to { w ( x ) : w ∈ S } \{w(x):w\in\mathcal{S}\} { w ( x ) : w ∈ S } and u ( x ) u(x) u ( x ) is an upper bound of that set, so v ( x ) ≤ u ( x ) v(x)\le u(x) v ( x ) ≤ u ( x ) by Upper Bound and Least Upper Bound . The last assertion is then claim 5 of Basic Properties of the δ \delta δ -Envelopes: Semicontinuity, Duality, Closed Superlevel Sets, Bounds and Monotonicity , applied to the pair v v v , u u u with δ ′ = δ \delta'=\delta δ ′ = δ .
Claim 2. Let δ > 0 \delta>0 δ > 0 , let φ ∈ C 2 ( U ) \varphi\in C^{2}(U) φ ∈ C 2 ( U ) , let x ^ ∈ V ∩ U \hat{x}\in V\cap U x ^ ∈ V ∩ U be a point at which the function V ∩ U → R V\cap U\to\mathbb{R} V ∩ U → R with value u δ − ( x ) − φ ( x ) u^{-}_{\delta}(x)-\varphi(x) u δ − ( x ) − φ ( x ) at x x x has a local maximum relative to V ∩ U V\cap U V ∩ U , and let ε > 0 \varepsilon>0 ε > 0 . We must produce witnesses as in Viscosity Subsolution, Supersolution and Solution of a Second-Order Equation on a Hilbert Triple §subsolution .
Step 1: a closed ball on which the maximum is global. Since U U U is open in H H H there is a positive ρ 1 \rho_{1} ρ 1 with B d H ( x ^ , ρ 1 ) ⊆ U B_{d_{H}}(\hat{x},\rho_{1})\subseteq U B d H ( x ^ , ρ 1 ) ⊆ U , by Open Subset of a Metric Space , and by Local Maximum of a Function Relative to a Subset of a Metric Space there is a positive δ 1 \delta_{1} δ 1 such that every x ∈ V ∩ U x\in V\cap U x ∈ V ∩ U with d H ( x ^ , x ) < δ 1 d_{H}(\hat{x},x)<\delta_{1} d H ( x ^ , x ) < δ 1 satisfies u δ − ( x ) − φ ( x ) ≤ u δ − ( x ^ ) − φ ( x ^ ) u^{-}_{\delta}(x)-\varphi(x)\le u^{-}_{\delta}(\hat{x})-\varphi(\hat{x}) u δ − ( x ) − φ ( x ) ≤ u δ − ( x ^ ) − φ ( x ^ ) . Put r 0 = 1 2 min { ρ 1 , δ 1 } r_{0}=\tfrac{1}{2}\min\{\rho_{1},\delta_{1}\} r 0 = 2 1 min { ρ 1 , δ 1 } , positive by claim 2 of Elementary Properties of the Minimum of Two Elements and claim 8 of Elementary Order Arithmetic in an Ordered Field , and satisfying r 0 < ρ 1 r_{0}<\rho_{1} r 0 < ρ 1 and r 0 < δ 1 r_{0}<\delta_{1} r 0 < δ 1 by claim 1 of the former and claim 2 of the latter. Let
K = B ˉ d H ( x ^ , r 0 ) = { x ′ ∈ H : d H ( x ^ , x ′ ) ≤ r 0 } , A = V ∩ K , m = u δ − ( x ^ ) − φ ( x ^ ) . K=\bar{B}_{d_{H}}(\hat{x},r_{0})=\{x'\in H:d_{H}(\hat{x},x')\le r_{0}\},\qquad A=V\cap K,\qquad m=u^{-}_{\delta}(\hat{x})-\varphi(\hat{x}). K = B ˉ d H ( x ^ , r 0 ) = { x ′ ∈ H : d H ( x ^ , x ′ ) ≤ r 0 } , A = V ∩ K , m = u δ − ( x ^ ) − φ ( x ^ ) .
Then K ⊆ U K\subseteq U K ⊆ U , because d H ( x ^ , x ′ ) ≤ r 0 < ρ 1 d_{H}(\hat{x},x')\le r_{0}<\rho_{1} d H ( x ^ , x ′ ) ≤ r 0 < ρ 1 puts x ′ x' x ′ in B d H ( x ^ , ρ 1 ) B_{d_{H}}(\hat{x},\rho_{1}) B d H ( x ^ , ρ 1 ) ; K K K is closed in H H H by claim 3 of Elementary Properties of the Closed Ball in a Metric Space ; x ^ ∈ K \hat{x}\in K x ^ ∈ K by claim 1 of that lemma, so x ^ ∈ A \hat{x}\in A x ^ ∈ A and A A A is nonempty; and
u δ − ( x ) − φ ( x ) ≤ m for every x ∈ A , u^{-}_{\delta}(x)-\varphi(x)\le m\qquad\text{for every }x\in A , u δ − ( x ) − φ ( x ) ≤ m for every x ∈ A ,
since A ⊆ V ∩ U A\subseteq V\cap U A ⊆ V ∩ U and d H ( x ^ , x ) ≤ r 0 < δ 1 d_{H}(\hat{x},x)\le r_{0}<\delta_{1} d H ( x ^ , x ) ≤ r 0 < δ 1 for x ∈ A x\in A x ∈ A .
Step 2: a radius on which φ \varphi φ and its derivatives are nearly constant. The function φ \varphi φ is continuous on U U U by claim 2 of Basic Properties of Differentiability on an Open Subset of a Real Inner Product Space , and its gradient map and Hessian map are continuous on U U U , as maps into ( H , d H ) (H,d_{H}) ( H , d H ) and into ( S y m ( H ) , d S y m ) (\mathrm{Sym}(H),d_{\mathrm{Sym}}) ( Sym ( H ) , d Sym ) , by The Classes C 1 C^1 C 1 and C 2 C^2 C 2 on an Open Subset of a Real Inner Product Space §c1 and The Classes C 1 C^1 C 1 and C 2 C^2 C 2 on an Open Subset of a Real Inner Product Space §c2 . Applying Continuous Map Between Metric Spaces three times at x ^ \hat{x} x ^ , with the positive numbers ε 16 \tfrac{\varepsilon}{16} 16 ε , ε 8 \tfrac{\varepsilon}{8} 8 ε and ε 8 \tfrac{\varepsilon}{8} 8 ε respectively, and taking for σ \sigma σ half the minimum of the three resulting radii together with r 0 r_{0} r 0 and ε 2 \tfrac{\varepsilon}{2} 2 ε , we obtain a positive σ \sigma σ with σ ≤ r 0 \sigma\le r_{0} σ ≤ r 0 and σ ≤ ε 2 \sigma\le\tfrac{\varepsilon}{2} σ ≤ 2 ε such that every z ∈ U z\in U z ∈ U with d H ( x ^ , z ) ≤ σ d_{H}(\hat{x},z)\le\sigma d H ( x ^ , z ) ≤ σ satisfies
∣ φ ( z ) − φ ( x ^ ) ∣ < ε 16 , ∣ D φ ( z ) − D φ ( x ^ ) ∣ H < ε 8 , ∥ D 2 φ ( z ) − D 2 φ ( x ^ ) ∥ < ε 8 . |\varphi(z)-\varphi(\hat{x})|<\tfrac{\varepsilon}{16},\qquad |D\varphi(z)-D\varphi(\hat{x})|_{H}<\tfrac{\varepsilon}{8},\qquad \lVert D^{2}\varphi(z)-D^{2}\varphi(\hat{x})\rVert<\tfrac{\varepsilon}{8}. ∣ φ ( z ) − φ ( x ^ ) ∣ < 16 ε , ∣ D φ ( z ) − D φ ( x ^ ) ∣ H < 8 ε , ∥ D 2 φ ( z ) − D 2 φ ( x ^ )∥ < 8 ε .
Step 3: the parameters. Put
μ = ε 16 , λ = min { 1 8 , σ 8 } , β = min { μ λ 2 4 , ε 32 } , ε ′ = min { ε 8 , σ 4 } , \mu=\tfrac{\varepsilon}{16},\qquad \lambda=\min\bigl\{\tfrac{1}{8},\tfrac{\sigma}{8}\bigr\},\qquad \beta=\min\bigl\{\tfrac{\mu\lambda^{2}}{4},\tfrac{\varepsilon}{32}\bigr\},\qquad \varepsilon'=\min\bigl\{\tfrac{\varepsilon}{8},\tfrac{\sigma}{4}\bigr\}, μ = 16 ε , λ = min { 8 1 , 8 σ } , β = min { 4 μ λ 2 , 32 ε } , ε ′ = min { 8 ε , 4 σ } ,
all positive by claim 2 of Elementary Properties of the Minimum of Two Elements . By claim 1 of that lemma and claim 5 of Elementary Arithmetic in an Ordered Field we record, for later use,
16 μ λ = ε λ ≤ ε 8 , 2 μ λ 2 = ε 8 λ 2 ≤ ε 8 , 4 λ ≤ σ 2 , 3 β ≤ μ λ 2 , 3 β ≤ ε 8 , 2 μ = ε 8 , 16\mu\lambda=\varepsilon\lambda\le\tfrac{\varepsilon}{8},\qquad 2\mu\lambda^{2}=\tfrac{\varepsilon}{8}\lambda^{2}\le\tfrac{\varepsilon}{8},\qquad 4\lambda\le\tfrac{\sigma}{2},\qquad 3\beta\le\mu\lambda^{2},\qquad 3\beta\le\tfrac{\varepsilon}{8},\qquad 2\mu=\tfrac{\varepsilon}{8}, 16 μ λ = ε λ ≤ 8 ε , 2 μ λ 2 = 8 ε λ 2 ≤ 8 ε , 4 λ ≤ 2 σ , 3 β ≤ μ λ 2 , 3 β ≤ 8 ε , 2 μ = 8 ε ,
using λ ≤ 1 8 ≤ 1 \lambda\le\tfrac{1}{8}\le 1 λ ≤ 8 1 ≤ 1 for the first two. Finally, by continuity of φ \varphi φ at x ^ \hat{x} x ^ again, choose a positive ρ 2 ≤ σ 4 \rho_{2}\le\tfrac{\sigma}{4} ρ 2 ≤ 4 σ such that every z ∈ U z\in U z ∈ U with d H ( x ^ , z ) ≤ ρ 2 d_{H}(\hat{x},z)\le\rho_{2} d H ( x ^ , z ) ≤ ρ 2 satisfies ∣ φ ( z ) − φ ( x ^ ) ∣ < β |\varphi(z)-\varphi(\hat{x})|<\beta ∣ φ ( z ) − φ ( x ^ ) ∣ < β .
Step 4: an almost optimal member of the family. The function V ∩ U → R V\cap U\to\mathbb{R} V ∩ U → R with value u ( x ) − δ h ( x ) u(x)-\delta h(x) u ( x ) − δ h ( x ) at x x x is bounded above near each point of V ∩ U V\cap U V ∩ U by The Penalty Function h = 1 2 ∣ ⋅ ∣ V 2 h=\tfrac12|\cdot|_V^2 h = 2 1 ∣ ⋅ ∣ V 2 of a Hilbert Triple: Expansion Identities, Closed Sublevel Sets, Density and Local Bounds §near-bounds , and u δ − u^{-}_{\delta} u δ − is its upper semicontinuous envelope by The δ \delta δ -Envelopes u δ − u^-_\delta u δ − and u δ + u^+_\delta u δ + of a Function on an Open Subset of a Hilbert Triple §minus . By claim 5 of Properties of the Upper Semicontinuous Envelope , applied at x ^ \hat{x} x ^ with the positive number min { ρ 2 , β } \min\{\rho_{2},\beta\} min { ρ 2 , β } , there is z ∈ V ∩ U z\in V\cap U z ∈ V ∩ U with
d H ( z , x ^ ) ≤ ρ 2 and ∣ ( u ( z ) − δ h ( z ) ) − u δ − ( x ^ ) ∣ < β . d_{H}(z,\hat{x})\le\rho_{2}\qquad\text{and}\qquad \bigl|\bigl(u(z)-\delta h(z)\bigr)-u^{-}_{\delta}(\hat{x})\bigr|<\beta . d H ( z , x ^ ) ≤ ρ 2 and ( u ( z ) − δ h ( z ) ) − u δ − ( x ^ ) < β .
Since u ( z ) u(z) u ( z ) is the least upper bound of { v ( z ) : v ∈ S } \{v(z):v\in\mathcal{S}\} { v ( z ) : v ∈ S } , claim 3 of Approximation Property of the Supremum and the Infimum in R \mathbb{R} R , applied with the positive number β \beta β , gives v ∈ S v\in\mathcal{S} v ∈ S with u ( z ) − β < v ( z ) u(z)-\beta<v(z) u ( z ) − β < v ( z ) . By claim 1 of Basic Properties of the δ \delta δ -Envelopes: Semicontinuity, Duality, Closed Superlevel Sets, Bounds and Monotonicity , v ( z ) − δ h ( z ) ≤ v δ − ( z ) v(z)-\delta h(z)\le v^{-}_{\delta}(z) v ( z ) − δ h ( z ) ≤ v δ − ( z ) , and by claim 9 of Properties of the Absolute Value in an Ordered Field , u δ − ( x ^ ) − β < u ( z ) − δ h ( z ) u^{-}_{\delta}(\hat{x})-\beta<u(z)-\delta h(z) u δ − ( x ^ ) − β < u ( z ) − δ h ( z ) ; hence
u δ − ( x ^ ) − 2 β < v δ − ( z ) . u^{-}_{\delta}(\hat{x})-2\beta<v^{-}_{\delta}(z). u δ − ( x ^ ) − 2 β < v δ − ( z ) .
Also d H ( z , x ^ ) ≤ ρ 2 ≤ σ 4 ≤ r 0 d_{H}(z,\hat{x})\le\rho_{2}\le\tfrac{\sigma}{4}\le r_{0} d H ( z , x ^ ) ≤ ρ 2 ≤ 4 σ ≤ r 0 , so z ∈ A z\in A z ∈ A , and ∣ φ ( z ) − φ ( x ^ ) ∣ < β |\varphi(z)-\varphi(\hat{x})|<\beta ∣ φ ( z ) − φ ( x ^ ) ∣ < β by the choice of ρ 2 \rho_{2} ρ 2 .
Step 5: the perturbed maximum. Let Φ : A → R \Phi:A\to\mathbb{R} Φ : A → R be given by Φ ( x ) = v δ − ( x ) − φ ( x ) \Phi(x)=v^{-}_{\delta}(x)-\varphi(x) Φ ( x ) = v δ − ( x ) − φ ( x ) ; this is defined because A ⊆ V ∩ U A\subseteq V\cap U A ⊆ V ∩ U . By claim 1 of the present proposition and Step 1, Φ ( x ) ≤ u δ − ( x ) − φ ( x ) ≤ m \Phi(x)\le u^{-}_{\delta}(x)-\varphi(x)\le m Φ ( x ) ≤ u δ − ( x ) − φ ( x ) ≤ m for every x ∈ A x\in A x ∈ A , so Φ \Phi Φ is bounded above and sup x ∈ A Φ ( x ) ≤ m \sup_{x\in A}\Phi(x)\le m sup x ∈ A Φ ( x ) ≤ m . By claim 3 of The δ \delta δ -Envelopes on an Open Subset, under Penalisation of a Continuous Function, and on a Closed Subset , applied to v v v with the closed set K K K and the continuous function φ \varphi φ , Φ \Phi Φ has closed superlevel sets in H H H . Moreover, by Step 4,
Φ ( z ) = v δ − ( z ) − φ ( z ) > ( u δ − ( x ^ ) − 2 β ) − ( φ ( x ^ ) + β ) = m − 3 β , \Phi(z)=v^{-}_{\delta}(z)-\varphi(z)>\bigl(u^{-}_{\delta}(\hat{x})-2\beta\bigr)-\bigl(\varphi(\hat{x})+\beta\bigr)=m-3\beta , Φ ( z ) = v δ − ( z ) − φ ( z ) > ( u δ − ( x ^ ) − 2 β ) − ( φ ( x ^ ) + β ) = m − 3 β ,
so that sup x ∈ A Φ ( x ) ≤ m < Φ ( z ) + 3 β ≤ Φ ( z ) + μ λ 2 \sup_{x\in A}\Phi(x)\le m<\Phi(z)+3\beta\le\Phi(z)+\mu\lambda^{2} sup x ∈ A Φ ( x ) ≤ m < Φ ( z ) + 3 β ≤ Φ ( z ) + μ λ 2 .
Applying A Perturbed Maximum Principle of Borwein-Preiss Type in a Real Hilbert Space to the real Hilbert space H H H , the nonempty set A A A , the function Φ \Phi Φ , the positive numbers μ \mu μ and λ \lambda λ and the point x 0 = z x_{0}=z x 0 = z , we obtain y ˉ ∈ H \bar{y}\in H y ˉ ∈ H and x ˉ ∈ A \bar{x}\in A x ˉ ∈ A such that, with Ψ ( x ) = Φ ( x ) − μ ∣ x − y ˉ ∣ H 2 \Psi(x)=\Phi(x)-\mu\,|x-\bar{y}|_{H}^{2} Ψ ( x ) = Φ ( x ) − μ ∣ x − y ˉ ∣ H 2 ,
∣ x ˉ − z ∣ H ≤ 4 λ , ∣ x ˉ − y ˉ ∣ H ≤ 8 λ , sup x ∈ A Φ ( x ) ≤ Φ ( x ˉ ) + 2 μ λ 2 , |\bar{x}-z|_{H}\le 4\lambda,\qquad |\bar{x}-\bar{y}|_{H}\le 8\lambda,\qquad \sup_{x\in A}\Phi(x)\le\Phi(\bar{x})+2\mu\lambda^{2}, ∣ x ˉ − z ∣ H ≤ 4 λ , ∣ x ˉ − y ˉ ∣ H ≤ 8 λ , x ∈ A sup Φ ( x ) ≤ Φ ( x ˉ ) + 2 μ λ 2 ,
and Ψ \Psi Ψ attains a sequentially strict maximum on A A A at x ˉ \bar{x} x ˉ . In particular Ψ ( x ) ≤ Ψ ( x ˉ ) \Psi(x)\le\Psi(\bar{x}) Ψ ( x ) ≤ Ψ ( x ˉ ) for every x ∈ A x\in A x ∈ A , by Sequentially Strict Maxima and Minima on a Subset of a Metric Space §maximum . Two consequences are recorded. First,
d H ( x ^ , x ˉ ) ≤ d H ( x ^ , z ) + d H ( z , x ˉ ) ≤ σ 4 + 4 λ ≤ σ 4 + σ 2 = 3 σ 4 ≤ σ , d_{H}(\hat{x},\bar{x})\le d_{H}(\hat{x},z)+d_{H}(z,\bar{x})\le\tfrac{\sigma}{4}+4\lambda\le\tfrac{\sigma}{4}+\tfrac{\sigma}{2}=\tfrac{3\sigma}{4}\le\sigma , d H ( x ^ , x ˉ ) ≤ d H ( x ^ , z ) + d H ( z , x ˉ ) ≤ 4 σ + 4 λ ≤ 4 σ + 2 σ = 4 3 σ ≤ σ ,
so the three estimates of Step 2 apply at x ˉ \bar{x} x ˉ . Secondly, Φ ( x ˉ ) ≥ Φ ( z ) − 2 μ λ 2 > m − 3 β − 2 μ λ 2 \Phi(\bar{x})\ge\Phi(z)-2\mu\lambda^{2}>m-3\beta-2\mu\lambda^{2} Φ ( x ˉ ) ≥ Φ ( z ) − 2 μ λ 2 > m − 3 β − 2 μ λ 2 , whence
v δ − ( x ˉ ) = Φ ( x ˉ ) + φ ( x ˉ ) > u δ − ( x ^ ) + ( φ ( x ˉ ) − φ ( x ^ ) ) − 3 β − 2 μ λ 2 > u δ − ( x ^ ) − ε 16 − ε 8 − ε 8 , v^{-}_{\delta}(\bar{x})=\Phi(\bar{x})+\varphi(\bar{x})>u^{-}_{\delta}(\hat{x})+\bigl(\varphi(\bar{x})-\varphi(\hat{x})\bigr)-3\beta-2\mu\lambda^{2}>u^{-}_{\delta}(\hat{x})-\tfrac{\varepsilon}{16}-\tfrac{\varepsilon}{8}-\tfrac{\varepsilon}{8}, v δ − ( x ˉ ) = Φ ( x ˉ ) + φ ( x ˉ ) > u δ − ( x ^ ) + ( φ ( x ˉ ) − φ ( x ^ ) ) − 3 β − 2 μ λ 2 > u δ − ( x ^ ) − 16 ε − 8 ε − 8 ε ,
using φ ( x ˉ ) − φ ( x ^ ) > − ε 16 \varphi(\bar{x})-\varphi(\hat{x})>-\tfrac{\varepsilon}{16} φ ( x ˉ ) − φ ( x ^ ) > − 16 ε from Step 2 and claim 3 of Properties of the Absolute Value in an Ordered Field . In the other direction x ˉ ∈ A \bar{x}\in A x ˉ ∈ A , so v δ − ( x ˉ ) ≤ u δ − ( x ˉ ) ≤ m + φ ( x ˉ ) = u δ − ( x ^ ) + ( φ ( x ˉ ) − φ ( x ^ ) ) < u δ − ( x ^ ) + ε 16 v^{-}_{\delta}(\bar{x})\le u^{-}_{\delta}(\bar{x})\le m+\varphi(\bar{x})=u^{-}_{\delta}(\hat{x})+\bigl(\varphi(\bar{x})-\varphi(\hat{x})\bigr)<u^{-}_{\delta}(\hat{x})+\tfrac{\varepsilon}{16} v δ − ( x ˉ ) ≤ u δ − ( x ˉ ) ≤ m + φ ( x ˉ ) = u δ − ( x ^ ) + ( φ ( x ˉ ) − φ ( x ^ ) ) < u δ − ( x ^ ) + 16 ε . Writing Δ = v δ − ( x ˉ ) − u δ − ( x ^ ) \Delta=v^{-}_{\delta}(\bar{x})-u^{-}_{\delta}(\hat{x}) Δ = v δ − ( x ˉ ) − u δ − ( x ^ ) , we have shown
− 5 ε 16 < Δ < ε 16 . -\tfrac{5\varepsilon}{16}<\Delta<\tfrac{\varepsilon}{16}. − 16 5 ε < Δ < 16 ε .
Step 6: testing v v v . Let q 0 : H → R q_{0}:H\to\mathbb{R} q 0 : H → R be given by q 0 ( x ) = 2 μ 2 ∣ x − y ˉ ∣ H 2 q_{0}(x)=\tfrac{2\mu}{2}|x-\bar{y}|_{H}^{2} q 0 ( x ) = 2 2 μ ∣ x − y ˉ ∣ H 2 . By claim 3 of Affine and Quadratic Functions on a Real Hilbert Space are of Class C 2 C^2 C 2 , applied with α = 2 μ \alpha=2\mu α = 2 μ and y 0 = y ˉ y_{0}=\bar{y} y 0 = y ˉ , and by claim 4 of that lemma, the restriction of q 0 q_{0} q 0 to U U U belongs to C 2 ( U ) C^{2}(U) C 2 ( U ) with D q 0 ( x ) = 2 μ ( x − y ˉ ) Dq_{0}(x)=2\mu\,(x-\bar{y}) D q 0 ( x ) = 2 μ ( x − y ˉ ) and D 2 q 0 ( x ) = 2 μ I H D^{2}q_{0}(x)=2\mu I_{H} D 2 q 0 ( x ) = 2 μ I H . Hence, by claim 2 of Constants, Sums, Scalar Multiples and Differences of Differentiable Functions on an Open Subset of a Real Inner Product Space , the function ψ = φ + q 0 \psi=\varphi+q_{0} ψ = φ + q 0 on U U U belongs to C 2 ( U ) C^{2}(U) C 2 ( U ) , with
D ψ ( x ) = D φ ( x ) + 2 μ ( x − y ˉ ) , D 2 ψ ( x ) = D 2 φ ( x ) + 2 μ I H ( x ∈ U ) . D\psi(x)=D\varphi(x)+2\mu\,(x-\bar{y}),\qquad D^{2}\psi(x)=D^{2}\varphi(x)+2\mu I_{H}\qquad (x\in U). D ψ ( x ) = D φ ( x ) + 2 μ ( x − y ˉ ) , D 2 ψ ( x ) = D 2 φ ( x ) + 2 μ I H ( x ∈ U ) .
For x ∈ A x\in A x ∈ A we have Ψ ( x ) = v δ − ( x ) − ψ ( x ) \Psi(x)=v^{-}_{\delta}(x)-\psi(x) Ψ ( x ) = v δ − ( x ) − ψ ( x ) . We claim that the function V ∩ U → R V\cap U\to\mathbb{R} V ∩ U → R with value v δ − ( x ) − ψ ( x ) v^{-}_{\delta}(x)-\psi(x) v δ − ( x ) − ψ ( x ) at x x x has a local maximum at x ˉ \bar{x} x ˉ relative to V ∩ U V\cap U V ∩ U . Indeed, let x ∈ V ∩ U x\in V\cap U x ∈ V ∩ U satisfy d H ( x ˉ , x ) < r 0 4 d_{H}(\bar{x},x)<\tfrac{r_{0}}{4} d H ( x ˉ , x ) < 4 r 0 . By Step 5 and σ ≤ r 0 \sigma\le r_{0} σ ≤ r 0 we have d H ( x ^ , x ˉ ) ≤ 3 σ 4 ≤ 3 r 0 4 d_{H}(\hat{x},\bar{x})\le\tfrac{3\sigma}{4}\le\tfrac{3r_{0}}{4} d H ( x ^ , x ˉ ) ≤ 4 3 σ ≤ 4 3 r 0 , so
d H ( x ^ , x ) ≤ d H ( x ^ , x ˉ ) + d H ( x ˉ , x ) < 3 r 0 4 + r 0 4 = r 0 ; d_{H}(\hat{x},x)\le d_{H}(\hat{x},\bar{x})+d_{H}(\bar{x},x)<\tfrac{3r_{0}}{4}+\tfrac{r_{0}}{4}=r_{0}; d H ( x ^ , x ) ≤ d H ( x ^ , x ˉ ) + d H ( x ˉ , x ) < 4 3 r 0 + 4 r 0 = r 0 ;
hence x ∈ K x\in K x ∈ K and so x ∈ A x\in A x ∈ A , and therefore v δ − ( x ) − ψ ( x ) = Ψ ( x ) ≤ Ψ ( x ˉ ) = v δ − ( x ˉ ) − ψ ( x ˉ ) v^{-}_{\delta}(x)-\psi(x)=\Psi(x)\le\Psi(\bar{x})=v^{-}_{\delta}(\bar{x})-\psi(\bar{x}) v δ − ( x ) − ψ ( x ) = Ψ ( x ) ≤ Ψ ( x ˉ ) = v δ − ( x ˉ ) − ψ ( x ˉ ) . Thus r 0 4 \tfrac{r_{0}}{4} 4 r 0 is a witnessing radius in Local Maximum of a Function Relative to a Subset of a Metric Space .
Since v v v is a viscosity subsolution of F F F on U U U , Viscosity Subsolution, Supersolution and Solution of a Second-Order Equation on a Hilbert Triple §subsolution , applied with δ \delta δ , the test function ψ \psi ψ , the point x ˉ \bar{x} x ˉ and the tolerance ε ′ \varepsilon' ε ′ , yields y ∈ W y\in W y ∈ W , s ∈ R s\in\mathbb{R} s ∈ R , q ∈ H q\in H q ∈ H and Y ∈ S y m ( H ) Y\in\mathrm{Sym}(H) Y ∈ Sym ( H ) with
∣ y − x ˉ ∣ H < ε ′ , ∣ v δ − ( y ) − v δ − ( x ˉ ) ∣ < ε ′ , ∣ s − v δ − ( x ˉ ) ∣ < ε ′ , |y-\bar{x}|_{H}<\varepsilon',\quad |v^{-}_{\delta}(y)-v^{-}_{\delta}(\bar{x})|<\varepsilon',\quad |s-v^{-}_{\delta}(\bar{x})|<\varepsilon', ∣ y − x ˉ ∣ H < ε ′ , ∣ v δ − ( y ) − v δ − ( x ˉ ) ∣ < ε ′ , ∣ s − v δ − ( x ˉ ) ∣ < ε ′ ,
∣ q − D ψ ( x ˉ ) ∣ H < ε ′ , ∥ Y − D 2 ψ ( x ˉ ) ∥ < ε ′ , F δ − ( y , s , q , Y ) ≤ ε ′ . |q-D\psi(\bar{x})|_{H}<\varepsilon',\quad \lVert Y-D^{2}\psi(\bar{x})\rVert<\varepsilon',\quad F^{-}_{\delta}(y,s,q,Y)\le\varepsilon' . ∣ q − D ψ ( x ˉ ) ∣ H < ε ′ , ∥ Y − D 2 ψ ( x ˉ )∥ < ε ′ , F δ − ( y , s , q , Y ) ≤ ε ′ .
Step 7: these are witnesses for u u u . We verify the six conditions of Viscosity Subsolution, Supersolution and Solution of a Second-Order Equation on a Hilbert Triple §subsolution for u u u , δ \delta δ , φ \varphi φ , x ^ \hat{x} x ^ and ε \varepsilon ε , with the same y , s , q , Y y,s,q,Y y , s , q , Y . Note first that
d H ( x ^ , y ) ≤ d H ( x ^ , x ˉ ) + d H ( x ˉ , y ) ≤ 3 σ 4 + ε ′ ≤ 3 σ 4 + σ 4 = σ , d_{H}(\hat{x},y)\le d_{H}(\hat{x},\bar{x})+d_{H}(\bar{x},y)\le\tfrac{3\sigma}{4}+\varepsilon'\le\tfrac{3\sigma}{4}+\tfrac{\sigma}{4}=\sigma , d H ( x ^ , y ) ≤ d H ( x ^ , x ˉ ) + d H ( x ˉ , y ) ≤ 4 3 σ + ε ′ ≤ 4 3 σ + 4 σ = σ ,
so the estimates of Step 2 apply at y y y as well; moreover y ∈ W ⊆ V ∩ U y\in W\subseteq V\cap U y ∈ W ⊆ V ∩ U and d H ( x ^ , y ) ≤ σ ≤ r 0 d_{H}(\hat{x},y)\le\sigma\le r_{0} d H ( x ^ , y ) ≤ σ ≤ r 0 , so y ∈ A y\in A y ∈ A .
(i) ∣ y − x ^ ∣ H = d H ( x ^ , y ) ≤ σ ≤ ε 2 < ε |y-\hat{x}|_{H}=d_{H}(\hat{x},y)\le\sigma\le\tfrac{\varepsilon}{2}<\varepsilon ∣ y − x ^ ∣ H = d H ( x ^ , y ) ≤ σ ≤ 2 ε < ε by claim 8 of Elementary Order Arithmetic in an Ordered Field .
(ii) Since y ∈ A y\in A y ∈ A , Step 1 gives u δ − ( y ) − u δ − ( x ^ ) ≤ φ ( y ) − φ ( x ^ ) < ε 16 u^{-}_{\delta}(y)-u^{-}_{\delta}(\hat{x})\le\varphi(y)-\varphi(\hat{x})<\tfrac{\varepsilon}{16} u δ − ( y ) − u δ − ( x ^ ) ≤ φ ( y ) − φ ( x ^ ) < 16 ε . In the other direction, claim 1 of the present proposition and Step 6 give
u δ − ( y ) ≥ v δ − ( y ) > v δ − ( x ˉ ) − ε ′ = u δ − ( x ^ ) + Δ − ε ′ > u δ − ( x ^ ) − 5 ε 16 − ε 8 . u^{-}_{\delta}(y)\ge v^{-}_{\delta}(y)>v^{-}_{\delta}(\bar{x})-\varepsilon'=u^{-}_{\delta}(\hat{x})+\Delta-\varepsilon'>u^{-}_{\delta}(\hat{x})-\tfrac{5\varepsilon}{16}-\tfrac{\varepsilon}{8}. u δ − ( y ) ≥ v δ − ( y ) > v δ − ( x ˉ ) − ε ′ = u δ − ( x ^ ) + Δ − ε ′ > u δ − ( x ^ ) − 16 5 ε − 8 ε .
Since 5 ε 16 + ε 8 = 7 ε 16 < ε \tfrac{5\varepsilon}{16}+\tfrac{\varepsilon}{8}=\tfrac{7\varepsilon}{16}<\varepsilon 16 5 ε + 8 ε = 16 7 ε < ε and ε 16 < ε \tfrac{\varepsilon}{16}<\varepsilon 16 ε < ε , claim 9 of Properties of the Absolute Value in an Ordered Field gives ∣ u δ − ( y ) − u δ − ( x ^ ) ∣ < ε |u^{-}_{\delta}(y)-u^{-}_{\delta}(\hat{x})|<\varepsilon ∣ u δ − ( y ) − u δ − ( x ^ ) ∣ < ε .
(iii) ∣ s − u δ − ( x ^ ) ∣ ≤ ∣ s − v δ − ( x ˉ ) ∣ + ∣ Δ ∣ < ε ′ + 5 ε 16 ≤ ε 8 + 5 ε 16 = 7 ε 16 < ε |s-u^{-}_{\delta}(\hat{x})|\le|s-v^{-}_{\delta}(\bar{x})|+|\Delta|<\varepsilon'+\tfrac{5\varepsilon}{16}\le\tfrac{\varepsilon}{8}+\tfrac{5\varepsilon}{16}=\tfrac{7\varepsilon}{16}<\varepsilon ∣ s − u δ − ( x ^ ) ∣ ≤ ∣ s − v δ − ( x ˉ ) ∣ + ∣Δ∣ < ε ′ + 16 5 ε ≤ 8 ε + 16 5 ε = 16 7 ε < ε , by claim 5 of Properties of the Absolute Value in an Ordered Field and the bounds on Δ \Delta Δ from Step 5.
(iv) By the triangle inequality, Step 6 and Step 2,
∣ q − D φ ( x ^ ) ∣ H ≤ ∣ q − D ψ ( x ˉ ) ∣ H + 2 μ ∣ x ˉ − y ˉ ∣ H + ∣ D φ ( x ˉ ) − D φ ( x ^ ) ∣ H < ε ′ + 16 μ λ + ε 8 ≤ ε 8 + ε 8 + ε 8 < ε , |q-D\varphi(\hat{x})|_{H}\le|q-D\psi(\bar{x})|_{H}+2\mu\,|\bar{x}-\bar{y}|_{H}+|D\varphi(\bar{x})-D\varphi(\hat{x})|_{H}<\varepsilon'+16\mu\lambda+\tfrac{\varepsilon}{8}\le\tfrac{\varepsilon}{8}+\tfrac{\varepsilon}{8}+\tfrac{\varepsilon}{8}<\varepsilon , ∣ q − D φ ( x ^ ) ∣ H ≤ ∣ q − D ψ ( x ˉ ) ∣ H + 2 μ ∣ x ˉ − y ˉ ∣ H + ∣ D φ ( x ˉ ) − D φ ( x ^ ) ∣ H < ε ′ + 16 μ λ + 8 ε ≤ 8 ε + 8 ε + 8 ε < ε ,
using D ψ ( x ˉ ) − D φ ( x ˉ ) = 2 μ ( x ˉ − y ˉ ) D\psi(\bar{x})-D\varphi(\bar{x})=2\mu(\bar{x}-\bar{y}) D ψ ( x ˉ ) − D φ ( x ˉ ) = 2 μ ( x ˉ − y ˉ ) and ∣ x ˉ − y ˉ ∣ H ≤ 8 λ |\bar{x}-\bar{y}|_{H}\le 8\lambda ∣ x ˉ − y ˉ ∣ H ≤ 8 λ .
(v) Likewise, by claim 3 of Elementary Properties of Bounded Symmetric Bilinear Forms: Norm, Quadratic Form, Order and Continuity ,
∥ Y − D 2 φ ( x ^ ) ∥ ≤ ∥ Y − D 2 ψ ( x ˉ ) ∥ + ∥ 2 μ I H ∥ + ∥ D 2 φ ( x ˉ ) − D 2 φ ( x ^ ) ∥ < ε ′ + 2 μ + ε 8 ≤ 3 ε 8 < ε , \lVert Y-D^{2}\varphi(\hat{x})\rVert\le\lVert Y-D^{2}\psi(\bar{x})\rVert+\lVert 2\mu I_{H}\rVert+\lVert D^{2}\varphi(\bar{x})-D^{2}\varphi(\hat{x})\rVert<\varepsilon'+2\mu+\tfrac{\varepsilon}{8}\le\tfrac{3\varepsilon}{8}<\varepsilon , ∥ Y − D 2 φ ( x ^ )∥ ≤ ∥ Y − D 2 ψ ( x ˉ )∥ + ∥ 2 μ I H ∥ + ∥ D 2 φ ( x ˉ ) − D 2 φ ( x ^ )∥ < ε ′ + 2 μ + 8 ε ≤ 8 3 ε < ε ,
where ∥ 2 μ I H ∥ ≤ 2 μ \lVert 2\mu I_{H}\rVert\le 2\mu ∥ 2 μ I H ∥ ≤ 2 μ because ∣ 2 μ ⟨ x ′ , y ′ ⟩ H ∣ ≤ 2 μ ∣ x ′ ∣ H ∣ y ′ ∣ H |2\mu\langle x',y'\rangle_{H}|\le 2\mu|x'|_{H}|y'|_{H} ∣2 μ ⟨ x ′ , y ′ ⟩ H ∣ ≤ 2 μ ∣ x ′ ∣ H ∣ y ′ ∣ H for all x ′ , y ′ ∈ H x',y'\in H x ′ , y ′ ∈ H by The Cauchy-Schwarz Inequality in a Real Inner Product Space and claim 4 of Properties of the Absolute Value in an Ordered Field , so that claim 2 of Elementary Properties of Bounded Symmetric Bilinear Forms: Norm, Quadratic Form, Order and Continuity applies.
(vi) F δ − ( y , s , q , Y ) ≤ ε ′ ≤ ε F^{-}_{\delta}(y,s,q,Y)\le\varepsilon'\le\varepsilon F δ − ( y , s , q , Y ) ≤ ε ′ ≤ ε .
As δ \delta δ , φ \varphi φ , x ^ \hat{x} x ^ and ε \varepsilon ε were arbitrary, and u u u is bounded above near each point of U U U by claim 1, the function u u u is a viscosity subsolution of F F F on U U U .