Documentation

TauCeti.Analysis.InnerProductSpace.Laplacian.DriftMaximumPrinciple

The maximum principle for Δ + b·∇ (a first-order drift term) #

TauCeti.Analysis.InnerProductSpace.Laplacian.WeakMaximumPrinciple proves the maximum principle for the bare Laplacian Δ (subharmonic functions), and TauCeti.Analysis.InnerProductSpace.Laplacian.ZerothOrderMaximumPrinciple adds a zeroth-order term c. This file supplies the missing first-order (transport/advection) term: the maximum principle for the second-order elliptic operator

L u = Δ u + ⟪b, ∇u⟫

with a bounded drift field b : E → E. In Lean the directional derivative ⟪b x, ∇u x⟫ is spelled fderiv ℝ u x (b x), so the operator value at x is Δ u x + fderiv ℝ u x (b x).

The two ingredients, both classical:

The exponential-barrier argument follows the weak maximum principle proof in Gilbarg--Trudinger, Elliptic Partial Differential Equations of Second Order, Chapter 3.

Main declarations #

theorem TauCeti.contDiff_exp_inner {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] (α : ℝ) (u : E) :
ContDiff ℝ ↑⊤ fun (y : E) => Real.exp (α * inner ℝ u y)

The exponential barrier y ↦ exp (α ⟪u, y⟫) is smooth, being the exponential of a continuous linear form.

@[simp]
theorem TauCeti.fderiv_exp_inner_apply {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] (α : ℝ) (u x v : E) :
(fderiv ℝ (fun (y : E) => Real.exp (α * inner ℝ u y)) x) v = Real.exp (α * inner ℝ u x) * (α * inner ℝ u v)

The directional derivative of the exponential barrier y ↦ exp (α ⟪u, y⟫).

@[simp]
theorem TauCeti.laplacian_exp_inner {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] (α : ℝ) (u x : E) :
Laplacian.laplacian (fun (y : E) => Real.exp (α * inner ℝ u y)) x = α ^ 2 * ‖u‖ ^ 2 * Real.exp (α * inner ℝ u x)

The Laplacian of the exponential barrier. For a fixed vector u, the function y ↦ exp (α ⟪u, y⟫) has Laplacian α² ‖u‖² exp (α ⟪u, y⟫), because it is an exponential of a linear form: the second directional derivative along an orthonormal basis vector eᵢ contributes α² ⟪u, eᵢ⟫², and these sum to α² ‖u‖². This is the barrier for the weak maximum principle with drift.

Strict interior obstruction. If 0 < Δ f x + fderiv ℝ f x v at a point where f is C², then f has no local maximum at x. The direction v is arbitrary: at a local maximum the derivative fderiv ℝ f x vanishes, so the first-order term contributes nothing and only the Laplacian's nonpositivity survives.

Strict interior obstruction, minimum form. If Δ f x + fderiv ℝ f x v < 0 at a point where f is C², then f has no local minimum at x.

theorem TauCeti.exists_mem_frontier_isMaxOn_of_laplacian_add_fderiv_pos {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] {K : Set E} (hK : IsCompact K) (hne : K.Nonempty) {f : E → ℝ} {b : E → E} (hcont : ContinuousOn f K) (hcd : ∀ ⦃x : E⦄, x ∈ interior K → ContDiffAt ℝ 2 f x) (hpos : ∀ ⦃x : E⦄, x ∈ interior K → 0 < Laplacian.laplacian f x + (fderiv ℝ f x) (b x)) :
∃ x ∈ frontier K, IsMaxOn f K x

Strict boundary maximum principle for Δ + b·∇. Let K be compact and nonempty. If f is continuous on K, is C² on interior K, and satisfies 0 < Δ f x + fderiv ℝ f x (b x) throughout interior K, then some maximum point of f on K lies on frontier K. No hypothesis on the drift field b is needed.

theorem TauCeti.exists_mem_frontier_isMinOn_of_laplacian_add_fderiv_neg {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] {K : Set E} (hK : IsCompact K) (hne : K.Nonempty) {f : E → ℝ} {b : E → E} (hcont : ContinuousOn f K) (hcd : ∀ ⦃x : E⦄, x ∈ interior K → ContDiffAt ℝ 2 f x) (hneg : ∀ ⦃x : E⦄, x ∈ interior K → Laplacian.laplacian f x + (fderiv ℝ f x) (b x) < 0) :
∃ x ∈ frontier K, IsMinOn f K x

Strict boundary minimum principle for Δ + b·∇. The dual of exists_mem_frontier_isMaxOn_of_laplacian_add_fderiv_pos.

theorem TauCeti.laplacian_add_fderiv_exp_inner_pos_of_norm_le {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] {u b : E} {β : ℝ} (hu : ‖u‖ = 1) (hb : ‖b‖ ≤ β) (x : E) :
0 < Laplacian.laplacian (fun (y : E) => Real.exp ((β + 1) * inner ℝ u y)) x + (fderiv ℝ (fun (y : E) => Real.exp ((β + 1) * inner ℝ u y)) x) b

The normalized exponential barrier for a drift bounded by β is a strict subsolution of Δ + b·∇.

theorem TauCeti.laplacian_add_fderiv_add_const_smul {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] (f w : E → ℝ) (b : E) (ε : ℝ) (x : E) (hf : ContDiffAt ℝ 2 f x) (hw : ContDiffAt ℝ 2 w x) :
Laplacian.laplacian (fun (y : E) => f y + ε • w y) x + (fderiv ℝ (fun (y : E) => f y + ε • w y) x) b = Laplacian.laplacian f x + (fderiv ℝ f x) b + ε * (Laplacian.laplacian w x + (fderiv ℝ w x) b)

The operator Δ + b·∇ is linear under addition of a constant multiple.

theorem TauCeti.le_of_laplacian_add_fderiv_nonneg_le_frontier {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [Nontrivial E] {K : Set E} (hK : IsCompact K) {f : E → ℝ} {b : E → E} {β m : ℝ} (hcont : ContinuousOn f K) (hcd : ∀ ⦃x : E⦄, x ∈ interior K → ContDiffAt ℝ 2 f x) (hb : ∀ ⦃x : E⦄, x ∈ interior K → ‖b x‖ ≤ β) (hlap : ∀ ⦃x : E⦄, x ∈ interior K → 0 ≤ Laplacian.laplacian f x + (fderiv ℝ f x) (b x)) (hbdry : ∀ ⦃x : E⦄, x ∈ frontier K → f x ≤ m) ⦃x : E⦄ :
x ∈ K → f x ≤ m

Weak maximum principle for Δ + b·∇ with bounded drift.

Let K be compact. If f is continuous on K, is C² on interior K, the drift field is bounded there (‖b x‖ ≤ β), and 0 ≤ Δ f x + fderiv ℝ f x (b x) (a subsolution of Δ + b·∇), then any bound m that f respects on frontier K bounds f on all of K.

theorem TauCeti.ge_of_laplacian_add_fderiv_nonpos_ge_frontier {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [Nontrivial E] {K : Set E} (hK : IsCompact K) {f : E → ℝ} {b : E → E} {β m : ℝ} (hcont : ContinuousOn f K) (hcd : ∀ ⦃x : E⦄, x ∈ interior K → ContDiffAt ℝ 2 f x) (hb : ∀ ⦃x : E⦄, x ∈ interior K → ‖b x‖ ≤ β) (hlap : ∀ ⦃x : E⦄, x ∈ interior K → Laplacian.laplacian f x + (fderiv ℝ f x) (b x) ≤ 0) (hbdry : ∀ ⦃x : E⦄, x ∈ frontier K → m ≤ f x) ⦃x : E⦄ :
x ∈ K → m ≤ f x

Weak minimum principle for Δ + b·∇ with bounded drift. The dual of le_of_laplacian_add_fderiv_nonneg_le_frontier for supersolutions (Δ f x + fderiv ℝ f x (b x) ≤ 0).

theorem TauCeti.le_of_laplacian_add_fderiv_le_laplacian_add_fderiv_of_le_frontier {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [Nontrivial E] {K : Set E} (hK : IsCompact K) {f g : E → ℝ} {b : E → E} {β : ℝ} (hfcont : ContinuousOn f K) (hgcont : ContinuousOn g K) (hfcd : ∀ ⦃x : E⦄, x ∈ interior K → ContDiffAt ℝ 2 f x) (hgcd : ∀ ⦃x : E⦄, x ∈ interior K → ContDiffAt ℝ 2 g x) (hb : ∀ ⦃x : E⦄, x ∈ interior K → ‖b x‖ ≤ β) (hL : ∀ ⦃x : E⦄, x ∈ interior K → Laplacian.laplacian g x + (fderiv ℝ g x) (b x) ≤ Laplacian.laplacian f x + (fderiv ℝ f x) (b x)) (hbdry : ∀ ⦃x : E⦄, x ∈ frontier K → f x ≤ g x) ⦃x : E⦄ :
x ∈ K → f x ≤ g x

Comparison principle for Δ + b·∇. Two functions acted on by the same bounded drift are ordered on a compact set if their operator values and frontier values are ordered.

theorem TauCeti.eqOn_of_laplacian_add_fderiv_eq_of_eqOn_frontier {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [Nontrivial E] {K : Set E} (hK : IsCompact K) {f g : E → ℝ} {b : E → E} {β : ℝ} (hfcont : ContinuousOn f K) (hgcont : ContinuousOn g K) (hfcd : ∀ ⦃x : E⦄, x ∈ interior K → ContDiffAt ℝ 2 f x) (hgcd : ∀ ⦃x : E⦄, x ∈ interior K → ContDiffAt ℝ 2 g x) (hb : ∀ ⦃x : E⦄, x ∈ interior K → ‖b x‖ ≤ β) (hL : ∀ ⦃x : E⦄, x ∈ interior K → Laplacian.laplacian f x + (fderiv ℝ f x) (b x) = Laplacian.laplacian g x + (fderiv ℝ g x) (b x)) (hbdry : Set.EqOn f g (frontier K)) :
Set.EqOn f g K

Uniqueness principle for Δ + b·∇. Functions with equal operator values for the same bounded drift and equal frontier data agree throughout the compact set.

theorem TauCeti.exists_mem_frontier_isMaxOn_of_laplacian_add_fderiv_nonneg {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [Nontrivial E] {K : Set E} (hK : IsCompact K) (hne : K.Nonempty) {f : E → ℝ} {b : E → E} {β : ℝ} (hcont : ContinuousOn f K) (hcd : ∀ ⦃x : E⦄, x ∈ interior K → ContDiffAt ℝ 2 f x) (hb : ∀ ⦃x : E⦄, x ∈ interior K → ‖b x‖ ≤ β) (hlap : ∀ ⦃x : E⦄, x ∈ interior K → 0 ≤ Laplacian.laplacian f x + (fderiv ℝ f x) (b x)) :
∃ x ∈ frontier K, IsMaxOn f K x

The ∃-form of the weak maximum principle for Δ + b·∇: a subsolution with bounded drift on a nonempty compact set attains a maximum on the frontier.

theorem TauCeti.exists_mem_frontier_isMinOn_of_laplacian_add_fderiv_nonpos {E : Type u_1} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E] [Nontrivial E] {K : Set E} (hK : IsCompact K) (hne : K.Nonempty) {f : E → ℝ} {b : E → E} {β : ℝ} (hcont : ContinuousOn f K) (hcd : ∀ ⦃x : E⦄, x ∈ interior K → ContDiffAt ℝ 2 f x) (hb : ∀ ⦃x : E⦄, x ∈ interior K → ‖b x‖ ≤ β) (hlap : ∀ ⦃x : E⦄, x ∈ interior K → Laplacian.laplacian f x + (fderiv ℝ f x) (b x) ≤ 0) :
∃ x ∈ frontier K, IsMinOn f K x

The ∃-form of the weak minimum principle for Δ + b·∇: a supersolution with bounded drift on a nonempty compact set attains a minimum on the frontier.