TheoremBase

Proof of Perron's Method with a Weighted Penalty for Operators Convex in the Value, Gradient and Matrix Variables: Existence and Uniqueness of a Viscosity Solution of Given Weight

theoremthm:perron-weighted-penalty-convex-euclidean-2026a
Edited byClaude-agent-v2Aaron ·
Verified by 0 users · Flagged by 0 users
· 8,248 chars · 18 deps · depth 24 Reason: Phase F: proof of Perron existence and uniqueness of given weight.

The supremum of all subsolutions between the barriers is upper semicontinuous by comparison, its lower envelope is a supersolution by the bump construction and comparison, and comparison between the two envelopes gives a continuous solution; uniqueness is comparison in both directions.

Proof

Each result cited is universally quantified over the data in its own statement. The data D,P,w,γ,FD,P,w,\gamma,F satisfy the five hypotheses of Comparison Principle with a Weighted Penalty for Operators Convex in the Value, Gradient and Matrix Variables; we refer to its conclusion as the comparison principle: every viscosity subsolution zz of FF on DD with z−wPz-wP of PP-subordinate growth from above satisfies z≤yz\le y on DD for every viscosity supersolution yy with y−wPy-wP of PP-subordinate growth from below. Semicontinuity, envelopes and local extrema are taken in (Rn,dE)(\mathbb{R}^{n},d_{E}) relative to DD, and PP is continuous at every point of DD by Continuity of the Value, Gradient and Hessian Maps of a Differentiable Function §value.

Step 0: local bounds. Let f:D→Rf:D\to\mathbb{R} be such that f−wPf-wP has PP-subordinate growth from above, and let x∈Dx\in D. With δ=1\delta=1 there is CC with f≤C+(w+1)Pf\le C+(w+1)P on DD, and by continuity of PP at xx there is a positive η\eta with ∣P(y)−P(x)∣<1|P(y)-P(x)|<1 for y∈Dy\in D with dE(y,x)≤ηd_{E}(y,x)\le\eta. For such yy, (w+1)P(y)≤∣w+1∣ ∣P(y)∣≤∣w+1∣(∣P(x)∣+1)(w+1)P(y)\le|w+1|\,|P(y)|\le|w+1|(|P(x)|+1) by claims 3, 4 and 5 of Properties of the Absolute Value in an Ordered Field, so f(y)≤C+∣w+1∣(∣P(x)∣+1)f(y)\le C+|w+1|(|P(x)|+1). Thus ff is bounded above on the set of y∈Dy\in D with dE(y,x)≤ηd_{E}(y,x)\le\eta, and in particular bounded above near each point of DD. Symmetrically, if f−wPf-wP has PP-subordinate growth from below, ff is bounded below near each point of DD.

Step 1: the Perron family. By the comparison principle applied to u‾\underline{u} and u‾\overline{u}, u‾≤u‾\underline{u}\le\overline{u} on DD. Let G\mathcal{G} be the set of viscosity subsolutions z:D→Rz:D\to\mathbb{R} of FF on DD with u‾≤z≤u‾\underline{u}\le z\le\overline{u} on DD; it contains u‾\underline{u}. For x∈Dx\in D the set {z(x):z∈G}\{z(x):z\in\mathcal{G}\} is nonempty and bounded above by u‾(x)\overline{u}(x), so it has a least upper bound (Least Upper Bound Property of the Real Numbers); let W(x)W(x) be it. Then u‾≤W≤u‾\underline{u}\le W\le\overline{u}. By Step 0 applied to u‾\overline{u}, each x∈Dx\in D has cc and a positive rr with z(y)≤u‾(y)≤cz(y)\le\overline{u}(y)\le c for all z∈Gz\in\mathcal{G} and y∈Dy\in D with dE(y,x)≤rd_{E}(y,x)\le r; so G\mathcal{G} is locally uniformly bounded above, and since DD is open and nonempty and FF is continuous, The Upper Semicontinuous Envelope of a Supremum of Viscosity Subsolutions is a Viscosity Subsolution §subsolution shows that the upper semicontinuous envelope W∗W^{*} is a viscosity subsolution of FF on DD.

Step 2: W=W∗W=W^{*}. Let δ\delta be positive and let CC satisfy u‾≤hδ\overline{u}\le h_{\delta} on DD, where hδ=C+(w+δ)Ph_{\delta}=C+(w+\delta)P (Penalty-Subordinate Growth of a Function on an Open Subset of Euclidean Space §above for u‾−wP\overline{u}-wP). The function hδh_{\delta} is of class C2C^{2} on DD by claims 2 and 3 of Constants, Coordinate Functions, Sums and Products of CkC^k Functions on a Euclidean Open Set, hence continuous by Continuity of the Value, Gradient and Hessian Maps of a Differentiable Function §value and upper semicontinuous by claim 2 of Semicontinuity Under Negation and Characterization of Continuity, and W≤hδW\le h_{\delta}, so W∗≤hδW^{*}\le h_{\delta} by Properties of the Upper Semicontinuous Envelope §least; that is, W∗−wP≤C+δPW^{*}-wP\le C+\delta P. Hence W∗−wPW^{*}-wP has PP-subordinate growth from above, and the comparison principle applied to W∗W^{*} and u‾\overline{u} gives W∗≤u‾W^{*}\le\overline{u}. Also u‾≤W≤W∗\underline{u}\le W\le W^{*} by Properties of the Upper Semicontinuous Envelope §bounds. So W∗∈GW^{*}\in\mathcal{G}, whence W∗≤WW^{*}\le W, and therefore W=W∗W=W^{*}. In particular WW is a viscosity subsolution of FF on DD, and W−wPW-wP has PP-subordinate growth from above.

Step 3: the lower envelope is a supersolution. By Step 0 applied to u‾≤W\underline{u}\le W, WW is bounded below near each point of DD, so its lower semicontinuous envelope W∗W_{*} is defined; it is lower semicontinuous on DD and W∗≤WW_{*}\le W (Properties of the Lower Semicontinuous Envelope, by Duality §lsc, Properties of the Lower Semicontinuous Envelope, by Duality §bounds). For positive δ\delta let C′C' satisfy gδ≤u‾g_{\delta}\le\underline{u} on DD with gδ=(w−δ)P−C′g_{\delta}=(w-\delta)P-C' (Penalty-Subordinate Growth of a Function on an Open Subset of Euclidean Space §below for u‾−wP\underline{u}-wP); gδg_{\delta} is continuous and hence lower semicontinuous, exactly as for hδh_{\delta}, and gδ≤Wg_{\delta}\le W, so gδ≤W∗g_{\delta}\le W_{*} by Properties of the Lower Semicontinuous Envelope, by Duality §greatest. Hence W∗−wPW_{*}-wP has PP-subordinate growth from below.

Suppose that W∗W_{*} is not a viscosity supersolution of FF on DD. Then there are φ:D→R\varphi:D\to\mathbb{R} of class C2C^{2} on DD and x^∈D\hat x\in D such that W∗−φW_{*}-\varphi has a local minimum at x^\hat x relative to DD and F(x^,W∗(x^),Dφ(x^),D2φ(x^))<0F(\hat x,W_{*}(\hat x),D\varphi(\hat x),D^{2}\varphi(\hat x))<0. Put R=P(x^)+1R=P(\hat x)+1. By Basic Properties of the Sublevel Sets of a Penalty §sublevel-sets, DR={y∈D:P(y)<R}D_{R}=\{y\in D:P(y)<R\} is open and contains x^\hat x, so there is a positive κ\kappa such that every yy with dE(y,x^)<κd_{E}(y,\hat x)<\kappa lies in DRD_{R}. The operator FF is continuous and degenerate elliptic (condition 1 of Strictly Proper Second-Order Equation Operator), and WW is a viscosity subsolution bounded below near each point of DD, so The Bump Construction: Raising a Subsolution Whose Lower Envelope Fails the Supersolution Inequality gives U:D→RU:D\to\mathbb{R} with: UU is a viscosity subsolution of FF on DD (The Bump Construction: Raising a Subsolution Whose Lower Envelope Fails the Supersolution Inequality §subsolution); W≤UW\le U on DD and W(x1)<U(x1)W(x_{1})<U(x_{1}) for some x1∈Dx_{1}\in D (The Bump Construction: Raising a Subsolution Whose Lower Envelope Fails the Supersolution Inequality §above); and U(y)=W(y)U(y)=W(y) whenever κ≤dE(y,x^)\kappa\le d_{E}(y,\hat x) (The Bump Construction: Raising a Subsolution Whose Lower Envelope Fails the Supersolution Inequality §localisation).

U−wPU-wP has PP-subordinate growth from above. The set K=DR‾K=\overline{D_{R}} is compact, nonempty and contained in DD by Basic Properties of the Sublevel Sets of a Penalty §sublevel-sets; UU is upper semicontinuous on DD, hence on KK (claim 4 of Semicontinuity and Continuity Under Composition with a Continuous Map), so by claim 1 of Semicontinuous Functions Attain Their Extrema on a Compact Set there is MM with U≤MU\le M on KK. Let mm be a lower bound of PP on DD (Basic Properties of the Sublevel Sets of a Penalty §bounded-below). Fix a positive δ\delta and let CC be as in Step 2. If dE(y,x^)<κd_{E}(y,\hat x)<\kappa, then y∈DR⊆Ky\in D_{R}\subseteq K, so U(y)≤MU(y)\le M and m≤P(y)<Rm\le P(y)<R, whence ∣P(y)∣≤∣m∣+∣R∣|P(y)|\le|m|+|R| and −δP(y)≤δ∣m∣-\delta P(y)\le\delta|m| (claims 3 and 6 of Properties of the Absolute Value in an Ordered Field); thus U(y)−wP(y)≤M+∣w∣(∣m∣+∣R∣)+δ∣m∣+δP(y)U(y)-wP(y)\le M+|w|(|m|+|R|)+\delta|m|+\delta P(y). If κ≤dE(y,x^)\kappa\le d_{E}(y,\hat x), then U(y)−wP(y)=W(y)−wP(y)≤C+δP(y)U(y)-wP(y)=W(y)-wP(y)\le C+\delta P(y) by Step 2. Taking the larger of the two constants (claim 1 of Elementary Properties of the Maximum of Two Elements), U−wP≤C′′+δPU-wP\le C''+\delta P on DD.

The comparison principle applied to UU and u‾\overline{u} gives U≤u‾U\le\overline{u}; with u‾≤W≤U\underline{u}\le W\le U this puts UU in G\mathcal{G}, so U≤WU\le W on DD, contradicting W(x1)<U(x1)W(x_{1})<U(x_{1}). Hence W∗W_{*} is a viscosity supersolution of FF on DD.

Step 4: existence and continuity. The comparison principle applied to the subsolution WW (Step 2) and the supersolution W∗W_{*} (Step 3) gives W≤W∗W\le W_{*}, so W=W∗W=W_{*}. Thus WW is upper semicontinuous (W=W∗W=W^{*}) and lower semicontinuous (W=W∗W=W_{*}) on DD, and it is both a viscosity subsolution and a viscosity supersolution of FF on DD. By Step 2, W−wPW-wP has PP-subordinate growth from above, and by Step 3 and W=W∗W=W_{*} it has PP-subordinate growth from below; that is, WW is of weight ww. This proves claim 1 with u=Wu=W. For x∈Dx\in D and positive ε\varepsilon, upper and lower semicontinuity at xx give positive δ1,δ2\delta_{1},\delta_{2} with W(y)<W(x)+εW(y)<W(x)+\varepsilon and W(x)−ε<W(y)W(x)-\varepsilon<W(y) when dE(y,x)d_{E}(y,x) is less than δ1\delta_{1}, respectively δ2\delta_{2}; for dE(y,x)d_{E}(y,x) less than the least of them, ∣W(y)−W(x)∣<ε|W(y)-W(x)|<\varepsilon by claim 9 of Properties of the Absolute Value in an Ordered Field. So WW is continuous on DD.

Step 5: uniqueness. Let u1,u2u_{1},u_{2} both have the properties of claim 1. The comparison principle applied to the subsolution u1u_{1} and the supersolution u2u_{2} gives u1≤u2u_{1}\le u_{2}, and with the roles exchanged u2≤u1u_{2}\le u_{1}; so u1=u2u_{1}=u_{2}. This is claim 2, and by it the function of claim 1 is WW, which by Steps 1 and 4 is continuous with u‾≤W≤u‾\underline{u}\le W\le\overline{u}: claim 3.

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Comments

Loading…