Documentation

LeanPool.HighContrastHomogenization.Provider.Quenched.UnitRangeCoarseBlockConcentration

High-contrast homogenization: Provider.Quenched.UnitRangeCoarseBlockConcentration #

Imported from the Apache-2.0 HighContrastHomogenization development at commit 7a13dbcd8d6609264a713373f5c69ceeac870472.

Concentration for the coarse block readouts #

The observables the endgame averages are readouts of the coarse block response of the standard aligned cubes of e.coarse.ellipticity. Two properties are needed before the concentration-for-sums estimate applies to them: each readout must be measurable for the local sigma-field of its own cube, and it must be bounded and centred.

Locality holds for every entry of the coarse block of a bounded open cell of positive volume, so in particular on the standard aligned cubes. Boundedness does not: on the high-contrast coefficient class the coarse block is not uniformly bounded, its size being controlled only through the reference block and the source scale. The estimate is therefore applied to the truncated readout, exactly as the printed derivation of the concentration condition does: a cutoff at a level B is applied first, the truncated readout is centred at its own mean and normalized by its own range, and the resulting family satisfies the hypotheses of the concentration estimate with dimensional constants.

The truncation level is a parameter of the statements below; nothing here fixes it, and no dimensional constant depends on it.

The truncated, centred, normalized readout #

Concentration for truncated local readouts #

The coarse block readout #

Every entry of the coarse block response of a standard aligned cube is measurable for the local sigma-field of that cube.