High-contrast homogenization: Provider.Response.WeakNormGeometric #
Imported from the Apache-2.0 HighContrastHomogenization development at commit
7a13dbcd8d6609264a713373f5c69ceeac870472.
The geometric sum of the scale weights #
Both scale sums of the weak-norm estimate are closed by the geometric sum over scales,
Σ_{k = -∞}^{t} 3^{(s - ρ/2)(k - t)} ≤ C/(2s - ρ),
used at the exponent s - ρ/2 ∈ (0, 1) on the bad event and at the exponent s
on the old scales, where the printed right side C/(2s-ρ) is 2/(s - ρ/2) for
C = 4. Reindexed by j = t - k, the sum is the ordinary geometric series in
the ratio 3^{-a}, and the bound comes from the convexity estimate
3^{-a} ≤ 1 - a/2 on 0 < a ≤ 1: since log 3 ≥ 1,
3^{-a} ≤ e^{-a} ≤ (1 + a)^{-1} ≤ 1 - a/2,
the last step because a(1-a) ≥ 0. The tail form is the same bound after the
common factor 3^{-aH} of the old scales k < t - H is taken out.
Nothing here is specific to the seminorm; the module is placed beside it because the weights it sums are the weights of the scale-average seminorm.
The geometric sum of the scale weights #
The geometric sum of the scale weights: Σ_{k ≤ t}3^{a(k-t)} ≤ 2/a for
0 < a ≤ 1, applied at a = s and at a = s - ρ/2, where the printed bound
C/(2s-ρ) is 4/(2s-ρ) = 2/(s-ρ/2).