TheoremBase

The Lipschitz regularity theorem gives every uNu_N, and u, one N-independent Lipschitz constant L, so (uN)u_N) has the common modulus Lt in (H, dH)d_H). By the stability theorem the limsup of the uNu_N is a viscosity subsolution and the liminf a supersolution of F. Comparison for F (zero drift, modulus lgl_g t) puts them on either side of u on V; density of V and the Lipschitz bound extend this to H, so liminf = limsup = u: pointwise convergence. Closed V-balls are sequentially compact in (H, dH)d_H) by the compactness hypothesis and the closure lemma, giving uniform convergence there.

Proof

Each result cited below is universally quantified over the data in its own statement. Elementary facts about real numbers (arithmetic, the order, absolute values, and the fact that a real number ss with s≤εs\le\varepsilon for every real ε>0\varepsilon>0 satisfies s≤0s\le0) are used without comment; they are carried by The Real Numbers: Standing Notation and Background. Throughout, dH(x,y)=∣x−y∣Hd_{H}(x,y)=|x-y|_{H} is the distance of HH, as fixed in Hilbert Triples: Standing Notation and Background §triple, and we put

L=ℓgλ0+2C′+1.L=\frac{\ell_{g}}{\lambda_{0}}+2C'+1 .

Since ℓg≥0\ell_{g}\ge0, λ0>0\lambda_{0}>0 and C′≥0C'\ge0, we have L≥1>0L\ge1>0.

Step 1 (claim 1). Fix N∈NN\in\mathbb{N}. We apply Lipschitz Regularity of Bounded Continuous Viscosity Solutions of a Viscous Hamilton-Jacobi Equation with a Monotone Nonlinearity on a Hilbert Triple §lipschitz with the data λ0,Cg,ℓg\lambda_{0},C_{g},\ell_{g}, the function gNg_{N} in place of gg, the nonlinearity BNB_{N} in place of BB, the number ν\nu, the sequence fNf^{N} in place of ff, the form ΓN\Gamma_{N} in place of Γ\Gamma, the constant C′C', and the function uNu_{N} in place of uu. Its hypotheses hold: HH is not finite-dimensional by assumption; 0<λ00<\lambda_{0}, 0≤Cg0\le C_{g}, 0≤ℓg0\le\ell_{g} and 0≤ν0\le\nu; gN:V→Rg_{N}:V\to\mathbb{R} satisfies ∣gN(x)∣≤Cg|g_{N}(x)|\le C_{g} and ∣gN(x)−gN(y)∣≤ℓg∣x−y∣H|g_{N}(x)-g_{N}(y)|\le\ell_{g}|x-y|_{H} for x,y∈Vx,y\in V; BNB_{N} is a monotone nonlinearity for (H,V,A)(H,V,A); fNf^{N} is square-summable in VV; ΓN∈Sym(H)\Gamma_{N}\in\mathrm{Sym}(H) satisfies 0Sym⪯ΓN⪯IH0_{\mathrm{Sym}}\preceq\Gamma_{N}\preceq I_{H}; the function on D(A)×R×H×Sym(V)D(A)\times\mathbb{R}\times H\times\mathrm{Sym}(V) that the cited theorem builds from these data is, expression for expression, the operator FNF_{N} of the statement; and uN:H→Ru_{N}:H\to\mathbb{R} is continuous on HH, satisfies ∣uN(x)∣≤C′|u_{N}(x)|\le C' for every x∈Hx\in H, and is a viscosity solution of FNF_{N} on HH, all by hypothesis. The cited clause therefore gives

∣uN(x)−uN(y)∣≤L ∣x−y∣Hfor all x,y∈H.|u_{N}(x)-u_{N}(y)|\le L\,|x-y|_{H}\qquad\text{for all }x,y\in H .

The constant LL involves only ℓg,λ0,C′\ell_{g},\lambda_{0},C', so the bound is the same for every N∈NN\in\mathbb{N}. The same application with (B,f,Γ,g)(B,f,\Gamma,g) in place of (BN,fN,ΓN,gN)(B_{N},f^{N},\Gamma_{N},g_{N}) and uu in place of uNu_{N}, whose hypotheses hold for the same reasons, the operator built there being FF, gives ∣u(x)−u(y)∣≤L∣x−y∣H|u(x)-u(y)|\le L|x-y|_{H} for all x,y∈Hx,y\in H. This proves claim 1.

Step 2 (a common modulus and the limit functions). Let ω\omega be the function on the nonnegative reals with ω(t)=Lt\omega(t)=Lt. Since L≥0L\ge0, ω\omega is a modulus of continuity by Linear Moduli of Continuity §modulus (with c=Lc=L). By Step 1 and the hypothesis ∣uN(x)∣≤C′|u_{N}(x)|\le C',

∣uN(x)∣≤C′and∣uN(x)−uN(y)∣≤ω(dH(x,y))for all N∈N and x,y∈H.|u_{N}(x)|\le C'\quad\text{and}\quad|u_{N}(x)-u_{N}(y)|\le\omega\bigl(d_{H}(x,y)\bigr)\qquad\text{for all }N\in\mathbb{N}\text{ and }x,y\in H .

For x∈Hx\in H let uˉ(x)\bar{u}(x) and u‾(x)\underline{u}(x) be the limit superior and the limit inferior of the bounded sequence (uN(x))N∈N(u_{N}(x))_{N\in\mathbb{N}}. These are the functions uˉ,u‾\bar{u},\underline{u} of Limits of a Uniformly Bounded Sequence of Real Functions with a Common Modulus of Continuity: Bounds, Extraction and Uniform Convergence on Sequentially Compact Sets for the metric space (H,dH)(H,d_{H}), the constant C=C′C=C', the modulus ω\omega and the sequence (uN)N∈N(u_{N})_{N\in\mathbb{N}}, and they are also the functions uˉ,u‾\bar{u},\underline{u} of Stability of Viscosity Sub- and Supersolutions on a Hilbert Triple: Limits of Uniformly Bounded, Equicontinuous Sequences under Convergence of the Operators on Bounded Test Data for the same CC, ω\omega and (uN)N∈N(u_{N})_{N\in\mathbb{N}}. By Limits of a Uniformly Bounded Sequence of Real Functions with a Common Modulus of Continuity: Bounds, Extraction and Uniform Convergence on Sequentially Compact Sets §bounds, ∣uˉ(x)∣≤C′|\bar{u}(x)|\le C' and ∣u‾(x)∣≤C′|\underline{u}(x)|\le C' for every x∈Hx\in H, and by Limits of a Uniformly Bounded Sequence of Real Functions with a Common Modulus of Continuity: Bounds, Extraction and Uniform Convergence on Sequentially Compact Sets §modulus,

∣uˉ(x)−uˉ(y)∣≤L ∣x−y∣Hand∣u‾(x)−u‾(y)∣≤L ∣x−y∣Hfor all x,y∈H.|\bar{u}(x)-\bar{u}(y)|\le L\,|x-y|_{H}\quad\text{and}\quad|\underline{u}(x)-\underline{u}(y)|\le L\,|x-y|_{H}\qquad\text{for all }x,y\in H .

Moreover, for each x∈Hx\in H we have u‾(x)≤uˉ(x)\underline{u}(x)\le\bar{u}(x) by claim 1 of Basic Properties of the Limit Inferior and Limit Superior of a Bounded Real Sequence.

Step 3 (stability). As recorded in the statement, FF and every FNF_{N} are second-order equation operators on HH relative to (H,V,A)(H,V,A) by A Viscous Hamilton-Jacobi Operator with a Monotone Nonlinearity Satisfies the Second-Order Comparison Hypotheses §operator, and (FN)N∈N(F_{N})_{N\in\mathbb{N}} converges to FF on bounded test data by hypothesis; the hypothesis that every sequence in VV bounded in VV has a subsequence converging in HH is also assumed. For every N∈NN\in\mathbb{N}, uNu_{N} is a viscosity solution of FNF_{N} on HH, hence by Viscosity Subsolution, Supersolution and Solution of a Second-Order Equation on a Hilbert Triple §solution both a viscosity subsolution and a viscosity supersolution of FNF_{N} on HH. Together with the bounds of Step 2, this verifies the hypotheses of Stability of Viscosity Sub- and Supersolutions on a Hilbert Triple: Limits of Uniformly Bounded, Equicontinuous Sequences under Convergence of the Operators on Bounded Test Data. By Stability of Viscosity Sub- and Supersolutions on a Hilbert Triple: Limits of Uniformly Bounded, Equicontinuous Sequences under Convergence of the Operators on Bounded Test Data §subsolution, uˉ\bar{u} is a viscosity subsolution of FF on HH, and by Stability of Viscosity Sub- and Supersolutions on a Hilbert Triple: Limits of Uniformly Bounded, Equicontinuous Sequences under Convergence of the Operators on Bounded Test Data §supersolution, u‾\underline{u} is a viscosity supersolution of FF on HH. Likewise uu, a viscosity solution of FF on HH, is both a viscosity subsolution and a viscosity supersolution of FF on HH by Viscosity Subsolution, Supersolution and Solution of a Second-Order Equation on a Hilbert Triple §solution.

Step 4 (comparison on VV). The operator FF is the operator of A Viscous Hamilton-Jacobi Operator with a Monotone Nonlinearity Satisfies the Second-Order Comparison Hypotheses for the data λ0\lambda_{0}, CgC_{g}, ωg(t)=ℓgt\omega_{g}(t)=\ell_{g}t, gg, BB, ℓ=0\ell=0, the zero map L0:H→HL_{0}:H\to H (with L0(x)=0HL_{0}(x)=0_{H}) as the Lipschitz drift, Γ\Gamma, ν\nu and ff. Indeed: ωg\omega_{g} is a modulus of continuity by Linear Moduli of Continuity §modulus (with c=ℓg≥0c=\ell_{g}\ge0); for x,y∈Vx,y\in V we have ∣x−y∣H≤∣x−y∣V|x-y|_{H}\le|x-y|_{V} by Hilbert Triples: Standing Notation and Background §triple, so ∣g(x)−g(y)∣≤ℓg∣x−y∣H≤ℓg∣x−y∣V=ωg(∣x−y∣V)|g(x)-g(y)|\le\ell_{g}|x-y|_{H}\le\ell_{g}|x-y|_{V}=\omega_{g}(|x-y|_{V}), while ∣g(x)∣≤Cg|g(x)|\le C_{g}; the map L0L_{0} is Lipschitz with constant 00 from (H,dH)(H,d_{H}) to itself in the sense of Lipschitz Map Between Metric Spaces, since dH(L0(x),L0(x′))=dH(0H,0H)=0=0⋅dH(x,x′)d_{H}(L_{0}(x),L_{0}(x'))=d_{H}(0_{H},0_{H})=0=0\cdot d_{H}(x,x'); and since Ax+B(x)+L0(x)=Ax+B(x)Ax+B(x)+L_{0}(x)=Ax+B(x), the function defined there is exactly FF. Since HH is not finite-dimensional, the comparison clause A Viscous Hamilton-Jacobi Operator with a Monotone Nonlinearity Satisfies the Second-Order Comparison Hypotheses §comparison applies to FF with the constant C=C′C=C' in two ways.

First, with uˉ\bar{u} as the subsolution and uu as the supersolution: uˉ(x)≤C′\bar{u}(x)\le C' for every x∈Hx\in H by Step 2 and −C′≤u(x)-C'\le u(x) for every x∈Hx\in H by hypothesis, and uˉ\bar{u}, uu are a viscosity subsolution and a viscosity supersolution of FF on HH by Step 3. Hence uˉ(y)≤u(y)\bar{u}(y)\le u(y) for every y∈Vy\in V.

Second, with uu as the subsolution and u‾\underline{u} as the supersolution: u(x)≤C′u(x)\le C' by hypothesis and −C′≤u‾(x)-C'\le\underline{u}(x) by Step 2, for every x∈Hx\in H, and uu, u‾\underline{u} are a viscosity subsolution and a viscosity supersolution of FF on HH by Step 3. Hence u(y)≤u‾(y)u(y)\le\underline{u}(y) for every y∈Vy\in V.

Step 5 (extension to HH by density). Fix x∈Hx\in H and a real ε>0\varepsilon>0. The set VV is dense in HH by Hilbert Triples: Standing Notation and Background §triple, that is, its closure in (H,dH)(H,d_{H}) is HH, so xx lies in that closure; by the equivalence of claims 1 and 3 of Characterization of the Closure in a Metric Space by Open Balls, applied in the metric space (H,dH)(H,d_{H}) with the subset VV and the radius ε/(2L)>0\varepsilon/(2L)>0, there is y∈Vy\in V with ∣x−y∣H<ε/(2L)|x-y|_{H}<\varepsilon/(2L). Using Step 1 for uu, Step 2 for uˉ\bar{u} and u‾\underline{u}, and Step 4 at the point y∈Vy\in V,

uˉ(x)−u(x)=(uˉ(x)−uˉ(y))+(uˉ(y)−u(y))+(u(y)−u(x))≤L∣x−y∣H+0+L∣x−y∣H<ε,\bar{u}(x)-u(x)=\bigl(\bar{u}(x)-\bar{u}(y)\bigr)+\bigl(\bar{u}(y)-u(y)\bigr)+\bigl(u(y)-u(x)\bigr)\le L|x-y|_{H}+0+L|x-y|_{H}<\varepsilon, u(x)−u‾(x)=(u(x)−u(y))+(u(y)−u‾(y))+(u‾(y)−u‾(x))≤L∣x−y∣H+0+L∣x−y∣H<ε.u(x)-\underline{u}(x)=\bigl(u(x)-u(y)\bigr)+\bigl(u(y)-\underline{u}(y)\bigr)+\bigl(\underline{u}(y)-\underline{u}(x)\bigr)\le L|x-y|_{H}+0+L|x-y|_{H}<\varepsilon .

As ε>0\varepsilon>0 was arbitrary, uˉ(x)≤u(x)\bar{u}(x)\le u(x) and u(x)≤u‾(x)u(x)\le\underline{u}(x). Combined with u‾(x)≤uˉ(x)\underline{u}(x)\le\bar{u}(x) from Step 2 this gives u‾(x)≤uˉ(x)≤u(x)≤u‾(x)\underline{u}(x)\le\bar{u}(x)\le u(x)\le\underline{u}(x), so

u‾(x)=uˉ(x)=u(x)for every x∈H.\underline{u}(x)=\bar{u}(x)=u(x)\qquad\text{for every }x\in H .

Step 6 (claim 2). Fix x∈Hx\in H. The sequence (uN(x))N∈N(u_{N}(x))_{N\in\mathbb{N}} is a bounded sequence of real numbers, since ∣uN(x)∣≤C′|u_{N}(x)|\le C', and by Step 5 its limit inferior u‾(x)\underline{u}(x) and its limit superior uˉ(x)\bar{u}(x) both equal u(x)u(x). By claim 5 of Basic Properties of the Limit Inferior and Limit Superior of a Bounded Real Sequence, with a=u(x)a=u(x), the sequence (uN(x))N∈N(u_{N}(x))_{N\in\mathbb{N}} converges to u(x)u(x). This proves claim 2.

Step 7 (claim 3). Fix real numbers R>0R>0 and ε>0\varepsilon>0, and let KR={x∈V:∣x∣V≤R}⊆HK_{R}=\{x\in V:|x|_{V}\le R\}\subseteq H. We show that KRK_{R} is sequentially compact in (H,dH)(H,d_{H}) in the sense of Sequentially Compact Subset of a Metric Space. Let (xm)m∈N(x_{m})_{m\in\mathbb{N}} be a sequence in HH with xm∈KRx_{m}\in K_{R} for every mm. It is a sequence in VV with dV(xm,0H)=∣xm∣V≤Rd_{V}(x_{m},0_{H})=|x_{m}|_{V}\le R for every mm, hence bounded in VV; by the compactness hypothesis of the statement there are a strictly increasing sequence (nk)k∈N(n_{k})_{k\in\mathbb{N}} in N\mathbb{N} and a point x∈Hx\in H such that the subsequence (xnk)k∈N(x_{n_{k}})_{k\in\mathbb{N}} converges to xx in HH. This subsequence is a sequence in VV with ∣xnk∣V≤R|x_{n_{k}}|_{V}\le R for every kk, and (V,dV)(V,d_{V}) is separable by Hilbert Triples: Standing Notation and Background §separable; so Weak Compactness and Closedness of Bounded Subsets of the Small Space of a Hilbert Triple §closure, applied with the constant RR to the sequence (xnk)k∈N(x_{n_{k}})_{k\in\mathbb{N}}, gives x∈Vx\in V and ∣x∣V≤R|x|_{V}\le R, that is x∈KRx\in K_{R}. Thus KRK_{R} is sequentially compact in (H,dH)(H,d_{H}).

By Step 5, uˉ(x)=u‾(x)\bar{u}(x)=\underline{u}(x) for every x∈KRx\in K_{R}. With the data of Step 2, Limits of a Uniformly Bounded Sequence of Real Functions with a Common Modulus of Continuity: Bounds, Extraction and Uniform Convergence on Sequentially Compact Sets §uniform, applied to K=KRK=K_{R} and to ε\varepsilon, gives N0∈NN_{0}\in\mathbb{N} such that ∣uN(x)−uˉ(x)∣≤ε|u_{N}(x)-\bar{u}(x)|\le\varepsilon for every N∈NN\in\mathbb{N} with N≥N0N\ge N_{0} and every x∈KRx\in K_{R}. Since uˉ(x)=u(x)\bar{u}(x)=u(x) by Step 5, this reads ∣uN(x)−u(x)∣≤ε|u_{N}(x)-u(x)|\le\varepsilon for every N≥N0N\ge N_{0} and every x∈Vx\in V with ∣x∣V≤R|x|_{V}\le R. This proves claim 3.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…