High-contrast homogenization: Provider.Response.CutoffDerivatives #
Imported from the Apache-2.0 HighContrastHomogenization development at commit
7a13dbcd8d6609264a713373f5c69ceeac870472.
Quantitative adapted-coordinate derivatives #
Pulling the physical cutoff back by the grid matrix is the normalized reference
cutoff. The estimates below therefore control exactly the tensors denoted by
q ∇φ and q² ∇²φ in the response argument.
theorem
HCPolySupport.HighContrast.Response.adaptedPreYoungCutoff_gradientCoefficient_le_common
(d : ℕ)
[NeZero d]
:
32 * ↑d ^ 2 * smoothTransitionProfile.derivBound ≤ 1024 * ↑d ^ 4 * max 1 (max smoothTransitionProfile.derivBound smoothTransitionProfile.secondDerivBound) ^ 2
The common second-order coefficient also dominates the sharper gradient coefficient.
theorem
HCPolySupport.HighContrast.Response.adaptedPreYoungCutoff_pullback_fderiv_bound
{d : ℕ}
[NeZero d]
{q : Mat d}
(hq : Matrix.PosDef q)
(t : ℤ)
(x : Vec d)
:
The adapted-coordinate first derivative has the common dimension-only bound used for both derivative orders.
theorem
HCPolySupport.HighContrast.Response.adaptedPreYoungCutoff_pullback_iteratedFDeriv_two_bound
{d : ℕ}
[NeZero d]
{q : Mat d}
(hq : Matrix.PosDef q)
(t : ℤ)
(x : Vec d)
:
‖iteratedFDeriv ℝ 2 (fun (y : Vec d) => adaptedPreYoungCutoff q hq t (matVecMul q y)) x‖ ≤ 1024 * ↑d ^ 4 * max 1 (max smoothTransitionProfile.derivBound smoothTransitionProfile.secondDerivBound) ^ 2 * 3 ^ (-2 * t)
The adapted-coordinate Hessian has the same dimension-only coefficient as the first derivative.