Documentation

LeanPool.Superdiffusion.Algsuperdiff.Section4.Provider.ExcessDecay.OffGridErrorFluxCorrected

OffGridErrorFluxCorrected #

Ported from scottnarmstrong/Superdiffusion at 7f20476385d4af418f110303085ad0740c9eba46. Module imports, support namespace and local option directives adapted for Lean Pool.

The shallow re-parameterization, and the development carrier #

The lower layer (fluxCorrectedNormalizedBlockResponseRepresentative) already separates the correction cube Q from the response cube R, so the fix is a shallow re-parameterization.

Two definitions

together with the reduction

fluxCorrectedErrorFunctionalAtRoot M L k (originCube d k) (originCube d k) s
    (isotropicComparatorMatrix (Annealed.sigmaBar M k))
  = fluxCorrectedErrorFunctional M L k s

which holds by rfl: the generalized object is a conservative extension of the proved one, matching it exactly at the matched index.

Main results #

References #

1. Pointwise ellipticity of the flux-corrected field on a cube #

Pointwise ellipticity of the flux-corrected field on a triadic cube.

The proved witness is an almost everywhere statement; the countable subadditivity consumes CoarseGraining's IsEllipticFieldOn, which is pointwise on the set. Both clauses are available from the proved ingredients, on the half-open realization.

2. F1': the shell root separated from the correction and comparator #

The F1' descendant maximum. The shells are taken over the descendants of an arbitrary root cube, while the flux correction (Q, at scale m) and the comparator a0 stay at the parent's index. Compare fluxCorrectedMaxDescendantNormalizedBlockResponseRepresentative, which forces root = Q.

Equations
  • One or more equations did not get rendered due to their size.
Instances For

    The F1' error functional. 𝓔_{s,∞,2} at the cube root, taken with the flux correction and the comparator of the parent index.

    Equations
    • One or more equations did not get rendered due to their size.
    Instances For

      3. The representative identification, at every root #

      4. The composed estimate in the development carriers #

      The composed off-grid stability estimate in the development carriers.

      On one probability-one event, for every real translate w, every child cube P and every lattice cube K with w + □_P ⊆ □_K:

      𝓔_{t,∞,2}(w + □_P; ã_{L,m}, σ_m)
          ≤ (36dt/((t−u)(1−2u)))^{1/2} · 3^{u(K.scale − P.scale)}
              · 𝓔_{u,∞,2}(□_K; ã_{L,m}, σ_m) ,
      

      Both printed depths (K.scale − P.scale = 1, 2) are instances.