BoundaryCoveringTrace #
Ported from scottnarmstrong/Superdiffusion at 7f20476385d4af418f110303085ad0740c9eba46. Module imports, support namespace and local option directives adapted for Lean Pool.
The boundary datum on the covering cube #
The last hypothesis CoarseGraining's boundary coarse Caccioppoli inequality
asks of its datum, once the covering cube of BoundaryCoveringGeometry is in
place, is the localized zero trace through the Dirichlet patch:
LocalizedZeroTraceFunctionOn (c + □_k) (openCubeAtScale x (k-1)) (u - v) .
Here the ambient domain is □_m but the Caccioppoli domain is the smaller
covering cube, so the H¹₀ witness has to descend. It does, and for exactly
the reason the covering was designed: a cutoff supported in the patch
multiplies the H¹₀(□_m) approximants into functions supported in patch ∩ □_m, and that set is inside the covering cube.
Main results #
h10RestrictOfApproxSupport— anH¹₀(Ω)function whose approximants are all supported in an openW ⊆ Ωis anH¹₀(W)function, with the same value, gradient and approximating sequence.localizedZeroTraceFunctionOn_of_memH10_of_inter_subset— the descent: a global zero trace onΩgives the localized zero trace onWthrough every windowVwithV ∩ Ω ⊆ W.localizedZeroTraceFunctionOn_wellPlacedCube— the instance at the covering cube, with the geometric hypothesis discharged byBoundaryCoveringGeometry.
What is not done here #
This is stated in the untranslated frame, with the covering cube as a
translated set. Feeding it to CoarseGraining's theorem, which is stated at a
TriadicCube, requires the A4 translation transport of the datum (sample
translation); that step belongs to the development's TranslationTransport*
lane and is not taken here. Nothing in this module asserts the Caccioppoli
estimate.
References #
- ABK26,
l.coarse.grained.Caccioppoli.RHS; boundary application. - CoarseGraining,
Sobolev/H1/LocalizedZeroTrace.lean,Sobolev/H1/Algebra/H10Function.lean,Sobolev/H1/BasicLemmas.lean.
1. The H¹₀ descent #
H¹₀ descends along the support of its approximants.
If every smooth compactly supported approximant of u ∈ H¹₀(Ω) is supported in
the open subset W ⊆ Ω, then u is an H¹₀(W) function: the same value, the
same weak gradient, the same approximating sequence. Only the two L²
convergences have to be re-read, and they are inherited because eLpNorm is
monotone in the measure.
Equations
- One or more equations did not get rendered due to their size.
Instances For
2. The localized zero trace on a subdomain #
The boundary datum descends to the covering set.
A function with a global zero trace on Ω has the localized zero trace on
any open W ⊆ Ω through any window V with V ∩ Ω ⊆ W.
3. The instance at the well-placed covering cube #
The Caccioppoli boundary datum on the covering cube.
For k ≤ m, a function with a global zero trace on □_m has the localized
zero trace on the well-placed covering cube c + □_k through the Dirichlet
patch openCubeAtScale x (k-1) — the exact hypothesis of CoarseGraining's
BoundaryForcedCaccioppoliDatum at an outer cube of scale k.
The geometric input is BoundaryCoveringGeometry: the cube stays in □_m, and
it covers the truncated patch (x+□_{k-1}) ∩ □_m.