Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 1.15 — Kolmogorov–Riesz compactness criterion in Lᵖ(ℝⁿ)

Proved
HunterPDE.Compactness.kolmogorov_riesz

by mikedeng1 · Sep 28, 2026 · Mathlib 0df444a (Lean v4.33.1)

compactnessfunctional-analysislp-spacesp2o-batch-books5p2o-gran-per-chapterp2o-plan-bookp2o-v1pde

Let 1≤p<∞1 \le p < \infty1≤p<∞. A subset F\mathcal{F}F of Lp(Rn)L^p(\mathbb{R}^n)Lp(Rn) is precompact (its closure is compact) if and only if:

  1. there exists MMM such that ∥f∥Lp≤M\|f\|_{L^p} \le M∥f∥Lp​≤M for all f∈Ff \in \mathcal{F}f∈F;
  2. for every ε>0\varepsilon > 0ε>0 there exists RRR such that (∫∣x∣>R∣f(x)∣p dx)1/p<ε\big(\int_{|x|>R} |f(x)|^p \, dx\big)^{1/p} < \varepsilon(∫∣x∣>R​∣f(x)∣pdx)1/p<ε for all f∈Ff \in \mathcal{F}f∈F;
  3. for every ε>0\varepsilon > 0ε>0 there exists δ>0\delta > 0δ>0 such that if ∣h∣<δ|h| < \delta∣h∣<δ,
(∫Rn∣f(x+h)−f(x)∣p dx)1/p<εfor all f∈F.\Big(\int_{\mathbb{R}^n} |f(x+h) - f(x)|^p \, dx\Big)^{1/p} < \varepsilon \qquad \text{for all } f \in \mathcal{F}.(∫Rn​∣f(x+h)−f(x)∣pdx)1/p<εfor all f∈F.

This is the LpL^pLp analogue of the Arzelà–Ascoli theorem: bounded, tight and LpL^pLp-equicontinuous. The Rellich–Kondrachov theorem is proved by verifying these three conditions. The notes cite it without proof.

Formalization Note. F\mathcal{F}F is a set in Mathlib's Lp ℝ p volume on EuclideanSpace ℝ (Fin n), with p : ℝ≥0∞, 1 ≤ p and p ≠ ∞. Both LpL^pLp integrals are written with eLpNorm, which for finite p≥1p \ge 1p≥1 equals (∫∣⋅∣p)1/p(\int |\cdot|^p)^{1/p}(∫∣⋅∣p)1/p, and ε\varepsilonε is compared through ENNReal.ofReal.

Preamble
import Mathlib

open MeasureTheory
open scoped ENNReal
Formal statement
namespace HunterPDE.Compactness

/-- Theorem 1.15 of Hunter, *Notes on PDEs* (revised 6/18/2014), pp. 6–7 (Kolmogorov–Riesz,
Riesz–Tamarkin, Fréchet–Kolmogorov): for `1 ≤ p < ∞`, a subset `F` of `Lᵖ(ℝⁿ)` is precompact
(its closure is compact) if and only if
(1) `F` is bounded: `‖f‖_{Lᵖ} ≤ M` for all `f ∈ F`;
(2) `F` is tight: for every `ε > 0` there is `R` with `(∫_{|x|>R} |f|ᵖ dx)^{1/p} < ε` for all `f ∈ F`;
(3) `F` is `Lᵖ`-equicontinuous: for every `ε > 0` there is `δ > 0` such that `|h| < δ` implies
`(∫_{ℝⁿ} |f(x + h) − f(x)|ᵖ dx)^{1/p} < ε` for all `f ∈ F`.
The `Lᵖ` norms over `{|x| > R}` and of the translation difference are written with `eLpNorm`,
which for `1 ≤ p < ∞` is `(∫ |·|ᵖ)^{1/p}`. -/
theorem kolmogorov_riesz {n : ℕ} (p : ℝ≥0∞) [Fact (1 ≤ p)] (hp : p ≠ ∞)
    (F : Set (Lp ℝ p (volume : Measure (EuclideanSpace ℝ (Fin n))))) :
    IsCompact (closure F) ↔
      ((∃ M : ℝ, ∀ f ∈ F, ‖f‖ ≤ M) ∧
        (∀ ε : ℝ, 0 < ε → ∃ R : ℝ, ∀ f ∈ F,
          eLpNorm (f : EuclideanSpace ℝ (Fin n) → ℝ) p
            (volume.restrict {x : EuclideanSpace ℝ (Fin n) | R < ‖x‖}) < ENNReal.ofReal ε) ∧
        (∀ ε : ℝ, 0 < ε → ∃ δ : ℝ, 0 < δ ∧ ∀ h : EuclideanSpace ℝ (Fin n), ‖h‖ < δ → ∀ f ∈ F,
          eLpNorm (fun x => (f : EuclideanSpace ℝ (Fin n) → ℝ) (x + h) -
            (f : EuclideanSpace ℝ (Fin n) → ℝ) x) p volume < ENNReal.ofReal ε)) := by sorry

end HunterPDE.Compactness
Source
Hunter, Notes on Partial Differential Equations (revised 6/18/2014), pp. 6–7, Theorem 1.15
Read-back

What the Lean code literally says, in plain math · claude-opus-5-5

Setting. Let n∈Nn \in \mathbb{N}n∈N and let λ\lambdaλ be Lebesgue measure on Rn\mathbb{R}^nRn (Euclidean norm ∣⋅∣|\cdot|∣⋅∣). Let p∈[1,∞)p \in [1,\infty)p∈[1,∞): the hypotheses say 1≤p1 \le p1≤p and p≠∞p \ne \inftyp=∞. Let FFF be any subset of Lp(Rn,λ;R)L^p(\mathbb{R}^n,\lambda;\mathbb{R})Lp(Rn,λ;R), the Banach space of a.e.-equivalence classes with norm ∥f∥p=(∫∣f∣p dλ)1/p\|f\|_p = (\int |f|^p\,d\lambda)^{1/p}∥f∥p​=(∫∣f∣pdλ)1/p. For f∈Ff \in Ff∈F, "f(x)f(x)f(x)" means the value of a fixed representative. The quantities below do not depend on which representative is used.

Claim. The closure of FFF in LpL^pLp is compact if and only if all three of the following hold:

  1. Boundedness. There is a real MMM with ∥f∥p≤M\|f\|_p \le M∥f∥p​≤M for every f∈Ff \in Ff∈F.
  2. Tightness. For every real ε>0\varepsilon > 0ε>0 there is a real RRR, with no sign restriction, such that for every f∈Ff \in Ff∈F
(∫{x:∣x∣>R}∣f(x)∣p dx)1/p<ε.\Bigl(\int_{\{x : |x| > R\}} |f(x)|^p\,dx\Bigr)^{1/p} < \varepsilon .(∫{x:∣x∣>R}​∣f(x)∣pdx)1/p<ε.
  1. Equicontinuity. For every real ε>0\varepsilon > 0ε>0 there is a real δ>0\delta > 0δ>0 such that for every h∈Rnh \in \mathbb{R}^nh∈Rn with ∣h∣<δ|h| < \delta∣h∣<δ and every f∈Ff \in Ff∈F
(∫Rn∣f(x+h)−f(x)∣p dx)1/p<ε.\Bigl(\int_{\mathbb{R}^n} |f(x+h) - f(x)|^p\,dx\Bigr)^{1/p} < \varepsilon .(∫Rn​∣f(x+h)−f(x)∣pdx)1/p<ε.

These integrals are taken in [0,∞][0,\infty][0,∞], so the strict inequalities also assert finiteness.

Degenerate cases.

  • FFF empty. The closure is empty and compact, and all three conditions hold vacuously, so both sides are true.
  • n=0n = 0n=0. R0\mathbb{R}^0R0 is a single point, Lebesgue measure on it is the unit point mass, and LpL^pLp is a copy of R\mathbb{R}R with ∥f∥p=∣f(0)∣\|f\|_p = |f(0)|∥f∥p​=∣f(0)∣. Tightness holds automatically: take R=0R = 0R=0, and the region {∣x∣>0}\{|x| > 0\}{∣x∣>0} is empty. Equicontinuity holds automatically: the only hhh is 000, so the difference is 000. The statement then reduces to "the closure of F⊆RF \subseteq \mathbb{R}F⊆R is compact iff FFF is bounded".
  • R<0R < 0R<0. If RRR is negative, the region {∣x∣>R}\{|x| > R\}{∣x∣>R} is all of Rn\mathbb{R}^nRn.
Human review
  • Endorsed by Shuze Chen · Oct 2, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 2, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me