High-contrast homogenization: Provider.Window.BurnSplitStandardToAdapted #
Imported from the Apache-2.0 HighContrastHomogenization development at commit
7a13dbcd8d6609264a713373f5c69ceeac870472.
The burn-split standard-to-adapted bridge #
adapted_scaled_of_standard transfers simultaneous standard-cell bounds to an
adapted cell and pays boundaryConst Cd g n for the transfer, because it applies
the eccentricity-carrying Whitney row estimate to every row β including the top
one, where the sharp estimate is the partition property.
The printed conversion (Lemma 2.13 of HC, whose conclusion is (2.124)) does
not pay that. It splits the Whitney rows at the burn depth D: on the band
u β€ D the hypothesis envelope is flat, so those rows are covered by the
total row weight alone, which is at most one; only the
rows below the band meet the geometric estimate, and there the envelope's growth
3 ^ (g (u - D)) is beaten by the volume decay 3 ^ (-u), leaving 3 ^ (-D).
The geometric constant therefore multiplies 3 ^ (-D) rather than the whole
sum, and is absorbed once the depth exceeds logβ (Cd π’ ΞΆ_g / Ξ΄) β HC (2.123),
and its stated principle that the aspect ratio enters through a scale
restriction and nowhere else.
Everything outside the row accounting is adapted_scaled_of_standard's proof
unchanged.
The burn-split Whitney bridge. Standard-cell bounds whose envelope is
truncated D scales above the adapted cell's own imply the adapted-cell bound
with constant 1 + Cd * π’ * ΞΆ_g * 3 ^ (-D): the eccentricity rides on the burn
depth instead of multiplying.