Documentation

LeanPool.Superdiffusion.Algsuperdiff.Section5.Support.SeriesEnergy

SeriesEnergy #

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

The energy estimate on a cube, and the Dirichlet problem for a perturbed coefficient field #

The C^{0,1/2} contraction of the Section 5.1 iteration says nothing about the H¹ size of its iterates, and a Hölder bound alone cannot identify a limit as a weak solution. What identifies it is the energy estimate

  lam ‖∇u‖_{L²(y+□_n)} ≤ d ‖g‖_{L²(y+□_n)}

for -∇·a∇u = ∇·g with zero boundary values: the antisymmetric part of a drops out when the equation is tested with the solution itself, so only the lower ellipticity constant survives. The dimensional factor d is the price of the supremum-norm carrier, paid once in abs_setIntegral_vecDot_le.

This module collects that estimate together with the algebra a perturbed coefficient field needs. If b = a + kap pointwise and both fields are elliptic on the cube, then a solution for a at forcing G is a solution for b at forcing G - kap ∇u; the perturbation kap is automatically bounded entrywise by the sum of the two upper ellipticity constants, and its action on an L² field is again L² because it is the difference of the actions of b and of a. No separate regularity or boundedness hypothesis on kap is required.

Main results #

References #

1. Elementary L² bookkeeping #

The energy integral of a field is nonnegative.

The L² size of a field from its supremum size on a set of finite volume.

2. Matrix algebra on the ambient carrier #

The action of a matrix field is additive in the matrix.

The action of a matrix field is subtractive in the matrix.

theorem Algsuperdiff.Section5.Support.sqrt_setIntegral_sq_matVecMul_le {d : ℕ} {U : Set (HCPolySupport.Vec d)} (hU : MeasurableSet U) {k : HCPolySupport.Vec d → HCPolySupport.Mat d} {delta : ℝ} (hdelta : 0 ≤ delta) (hk : ∀ x ∈ U, ∀ (i j : Fin d), |k x i j| ≤ delta) {F : HCPolySupport.Vec d → HCPolySupport.Vec d} (hF : HCPolySupport.MemVectorL2 U F) (hkF : HCPolySupport.MemVectorL2 U fun (x : HCPolySupport.Vec d) => HCPolySupport.matVecMul (k x) (F x)) :
√(∫ (x : HCPolySupport.Vec d) in U, ‖HCPolySupport.matVecMul (k x) (F x)‖ ^ 2) ≤ ↑d * delta * √(∫ (x : HCPolySupport.Vec d) in U, ‖F x‖ ^ 2)

The L² size of the action of an entrywise bounded matrix field.

3. The energy estimate #

The energy estimate for the localized Dirichlet problem. Testing -∇·a∇u = ∇·g with the solution itself and using only the lower ellipticity constant gives

  lam ‖∇u‖_{L²(y+□_n)} ≤ d ‖g‖_{L²(y+□_n)} .

The antisymmetric part of a contributes nothing: the hypothesis used is the one-sided bound lam |ξ|² ≤ ξ·aξ, which a skew addition to a leaves untouched.

4. Additivity of the localized Dirichlet problem #

The sum of two solutions solves the problem for the sum of the forcing fields.

5. The perturbed coefficient field #

theorem Algsuperdiff.Section5.Support.abs_perturbation_entry_le {d : ℕ} {U : Set (HCPolySupport.Vec d)} {a b : HCPolySupport.CoeffField d} {lamA LamA lamB LamB : ℝ} (hEllA : HCPolySupport.IsEllipticFieldOn lamA LamA U a) (hEllB : HCPolySupport.IsEllipticFieldOn lamB LamB U b) {kap : HCPolySupport.Vec d → HCPolySupport.Mat d} (hab : ∀ (x : HCPolySupport.Vec d), b x = a x + kap x) {x : HCPolySupport.Vec d} (hx : x ∈ U) (i j : Fin d) :
|kap x i j| ≤ LamA + LamB

The perturbation between two fields elliptic on the cube is bounded entrywise by the sum of their upper ellipticity constants.

theorem Algsuperdiff.Section5.Support.perturbation_bound_nonneg {d : ℕ} {U : Set (HCPolySupport.Vec d)} {a b : HCPolySupport.CoeffField d} {lamA LamA lamB LamB : ℝ} (hx : U.Nonempty) (hEllA : HCPolySupport.IsEllipticFieldOn lamA LamA U a) (hEllB : HCPolySupport.IsEllipticFieldOn lamB LamB U b) :
0 ≤ LamA + LamB

The sum of the two upper ellipticity constants is nonnegative.

The action of the perturbation on an L² field is again L²: it is the difference of the actions of the two elliptic fields.

theorem Algsuperdiff.Section5.Support.isDirichletSolutionAt_coeff_add {d : ℕ} {y : HCPolySupport.Vec d} {n : ℤ} {a b : HCPolySupport.CoeffField d} {lamA LamA lamB LamB : ℝ} (hEllA : HCPolySupport.IsEllipticFieldOn lamA LamA (cubeSetAt y n) a) (hEllB : HCPolySupport.IsEllipticFieldOn lamB LamB (cubeSetAt y n) b) {kap : HCPolySupport.Vec d → HCPolySupport.Mat d} (hab : ∀ (x : HCPolySupport.Vec d), b x = a x + kap x) {G : HCPolySupport.Vec d → HCPolySupport.Vec d} (hG : HCPolySupport.MemVectorL2 (cubeSetAt y n) G) {u : HCPolySupport.H1Function (cubeSetAt y n)} (hu : IsDirichletSolutionAt a y n u G) :
IsDirichletSolutionAt b y n u fun (x : HCPolySupport.Vec d) => G x - HCPolySupport.matVecMul (kap x) (u.grad x)

A solution for a is a solution for b = a + kap at the forcing field corrected by the action of the perturbation on its own gradient.