Mountain pass theorem
DefinitionCritical point#✓fully proved
Let \(E\) be a real Banach space and \(f : E \to \mathbb{R}\). A point \(x \in E\) is a critical point of \(f\) when the Fréchet derivative of \(f\) at \(x\) vanishes, \(f'(x) = 0\). Here \(f'(x)\) is taken to be \(0\) when \(f\) is not differentiable at \(x\), so such points are critical as well.
def IsCriticalPoint (f : E → ℝ) (x : E) : Prop :=
fderiv ℝ f x = 0
Used by 1
DefinitionPalais–Smale condition#✓fully proved
Let \(E\) be a real Banach space. A function \(f : E \to \mathbb{R}\) satisfies the Palais–Smale condition when every sequence \((u_k)\) in \(E\) such that \((f(u_k))\) is bounded and the Fréchet derivatives \(f'(u_k)\) tend to \(0\) in operator norm has a subsequence converging in \(E\). Here \(f'(x)\) is taken to be \(0\) at points where \(f\) is not differentiable.
def PalaisSmale (f : E → ℝ) : Prop :=
∀ u : ℕ → E, (∃ M : ℝ, ∀ k, |f (u k)| ≤ M) →
Filter.Tendsto (fun k => fderiv ℝ f (u k)) Filter.atTop (nhds 0) →
∃ (x : E) (φ : ℕ → ℕ), StrictMono φ ∧
Filter.Tendsto (u ∘ φ) Filter.atTop (nhds x)
Used by 1
DefinitionMountain range#✓fully proved
Let \(E\) be a real Banach space, \(f : E \to \mathbb{R}\), \(a, b \in E\) and \(\varepsilon, r \in \mathbb{R}\). The points \(a\) and \(b\) are separated by a mountain range of \(f\) at height \(\varepsilon\) and radius \(r\) when \(f(a) = 0\), \(\varepsilon > 0\), \(r > 0\), \(f(y) \ge \varepsilon\) for every \(y \in E\) with \(\|y - a\| = r\), \(\|b - a\| > r\) and \(f(b) \le 0\).
def MountainRange (f : E → ℝ) (a b : E) (ε r : ℝ) : Prop :=
f a = 0 ∧ 0 < ε ∧ 0 < r ∧
(∀ y ∈ Metric.sphere a r, ε ≤ f y) ∧ r < ‖b - a‖ ∧ f b ≤ 0
Used by 1
DefinitionMountain pass level#✓fully proved
Let \(E\) be a real Banach space, \(f : E \to \mathbb{R}\) and \(a, b \in E\). The mountain pass level of \(f\) between \(a\) and \(b\) is
the infimum taken over all continuous paths \(\gamma : [0, 1] \to E\) with \(\gamma(0) = a\) and \(\gamma(1) = b\). Suprema and infima are taken in \(\mathbb{R}\), where a family that is unbounded on the relevant side has supremum or infimum \(0\) by convention.
noncomputable def mountainPassLevel (f : E → ℝ) (a b : E) : ℝ :=
⨅ γ : Path a b, ⨆ t : I, f (γ t)
Used by 1
TheoremMountain pass theorem#●ready to prove
Let \(E\) be a real Banach space and let \(f : E \to \mathbb{R}\) be continuously differentiable and satisfy the Palais–Smale condition. Let \(a, b \in E\) and let \(\varepsilon, r\) be real numbers such that \(a\) and \(b\) are separated by a mountain range of \(f\) at height \(\varepsilon\) and radius \(r\): \(f(a) = 0\), \(\varepsilon > 0\), \(r > 0\), \(f(y) \ge \varepsilon\) for every \(y\) with \(\|y - a\| = r\), \(\|b - a\| > r\) and \(f(b) \le 0\). Then the mountain pass level \(c\) of \(f\) between \(a\) and \(b\) satisfies \(c \ge \varepsilon\), and \(f\) has a critical point \(x\) with \(f(x) = c\).
public theorem_wanted mountain_pass (f : E → ℝ) (_hf : ContDiff ℝ 1 f) (_hps : PalaisSmale f)
{a b : E} {ε r : ℝ} (_hmr : MountainRange f a b ε r) :
∃ x : E, IsCriticalPoint f x ∧
f x = mountainPassLevel f a b ∧ ε ≤ mountainPassLevel f a b
def PalaisSmale (f : E → ℝ) : Prop :=
∀ u : ℕ → E, (∃ M : ℝ, ∀ k, |f (u k)| ≤ M) →
Filter.Tendsto (fun k => fderiv ℝ f (u k)) Filter.atTop (nhds 0) →
∃ (x : E) (φ : ℕ → ℕ), StrictMono φ ∧
Filter.Tendsto (u ∘ φ) Filter.atTop (nhds x)
WantedExt/Analysis/MountainPassWanted.lean:73
def MountainRange (f : E → ℝ) (a b : E) (ε r : ℝ) : Prop :=
f a = 0 ∧ 0 < ε ∧ 0 < r ∧
(∀ y ∈ Metric.sphere a r, ε ≤ f y) ∧ r < ‖b - a‖ ∧ f b ≤ 0
WantedExt/Analysis/MountainPassWanted.lean:82
def IsCriticalPoint (f : E → ℝ) (x : E) : Prop :=
fderiv ℝ f x = 0
WantedExt/Analysis/MountainPassWanted.lean:68
noncomputable def mountainPassLevel (f : E → ℝ) (a b : E) : ℝ :=
⨅ γ : Path a b, ⨆ t : I, f (γ t)
WantedExt/Analysis/MountainPassWanted.lean:88
Sources
- A. Ambrosetti and P.H. Rabinowitz, 'Dual variational methods in critical point theory and applications', J. Funct. Anal. 14 (1973) 349–381. Listed as §119 in O. Knill, Some Fundamental Theorems in Mathematics (https://people.math.harvard.edu/~knill/graphgeometry/papers/fundamental.pdf). Problem
mountain_passof the lean-eval benchmark.
Uses 4
Loading dependency explorer…
The dependency explorer needs JavaScript and its graph data.