Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 4.1 — Gundy's decomposition f = X + Y + Z of an L¹-bounded martingale, with one stopping time

Proved
BurkholderDFI.Gundy.theorem_4_1

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

calderon-zygmundgundy-decompositionmartingalesp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1

Let (Ω,A,P)(\Omega,\mathcal A,P)(Ω,A,P) be a probability space, A0⊆A1⊆⋯\mathcal A_0\subseteq\mathcal A_1\subseteq\cdotsA0​⊆A1​⊆⋯ a filtration, and f=(f1,f2,… )f=(f_1,f_2,\dots)f=(f1​,f2​,…) an L1L^1L1-bounded martingale relative to A1,A2,…\mathcal A_1,\mathcal A_2,\dotsA1​,A2​,…, i.e. ∥f∥1=sup⁡n≥1E∣fn∣<∞\|f\|_1=\sup_{n\ge1}E|f_n|<\infty∥f∥1​=supn≥1​E∣fn​∣<∞. Let λ\lambdaλ be a positive real number. Then there are martingales X,Y,ZX,Y,ZX,Y,Z, all relative to the same σ\sigmaσ-fields A1,A2,…\mathcal A_1,\mathcal A_2,\dotsA1​,A2​,…, with difference sequences x,y,zx,y,zx,y,z (y1=Y1y_1=Y_1y1​=Y1​, yk=Yk−Yk−1y_k=Y_k-Y_{k-1}yk​=Yk​−Yk−1​, and so on), such that

  1. f=X+Y+Zf=X+Y+Zf=X+Y+Z, that is, fn=Xn+Yn+Znf_n=X_n+Y_n+Z_nfn​=Xn​+Yn​+Zn​ for every n≥1n\ge1n≥1;
∥X∥22≤2λ∥f∥1,\|X\|_2^2\le 2\lambda\|f\|_1,∥X∥22​≤2λ∥f∥1​,

where ∥X∥2=sup⁡n≥1(EXn2)1/2\|X\|_2=\sup_{n\ge1}(EX_n^2)^{1/2}∥X∥2​=supn≥1​(EXn2​)1/2; 3.

∥∑k=1∞∣yk∣∥1≤4∥f∥1;\Big\|\sum_{k=1}^\infty|y_k|\Big\|_1\le 4\|f\|_1;​k=1∑∞​∣yk​∣​1​≤4∥f∥1​;
P(Z∗>0)≤∥f∥1/λ,P(Z^*>0)\le\|f\|_1/\lambda,P(Z∗>0)≤∥f∥1​/λ,

where Z∗=sup⁡n≥1∣Zn∣Z^*=\sup_{n\ge1}|Z_n|Z∗=supn≥1​∣Zn​∣.

The decomposition splits an L1L^1L1-bounded martingale at height λ\lambdaλ into an L2L^2L2-bounded part, a part of bounded total variation in L1L^1L1, and a part that vanishes identically outside an event of probability at most ∥f∥1/λ\|f\|_1/\lambda∥f∥1​/λ. It is the martingale analogue of the Calderón–Zygmund decomposition and reduces weak-type (1,1)(1,1)(1,1) estimates for martingale operators to L2L^2L2 estimates.

Formalization Note The three martingales are Mathlib martingales for the same filtration as fff; asserting this is no weaker than the paper, since a martingale indexed from 111 extends to one indexed from 000 by X0:=E(X1∣A0)X_0:=E(X_1\mid\mathcal A_0)X0​:=E(X1​∣A0​), and the decomposition (1) is required only for n≥1n\ge1n≥1, the paper's indices (Lean's index 000 is otherwise free). Likewise the hypothesis that fff is a Mathlib martingale is no restriction. All norms are computed in [0,∞][0,\infty][0,∞] with the lower Lebesgue integral: ∥X∥22\|X\|_2^2∥X∥22​ is the square of the supremum over n≥1n\ge1n≥1 of ∥Xn∥2\|X_n\|_2∥Xn​∥2​ (not the norm of a limit), the left side of (3) is the integral of the series ∑k≥1∣yk∣\sum_{k\ge1}|y_k|∑k≥1​∣yk​∣ with y1=Y1y_1=Y_1y1​=Y1​, and ∥f∥1/λ\|f\|_1/\lambda∥f∥1​/λ is a quotient in [0,∞][0,\infty][0,∞], finite since ∥f∥1<∞\|f\|_1<\infty∥f∥1​<∞ and λ>0\lambda>0λ>0.

Preamble
import Mathlib
import Definitions.Def_BurkholderDFI_SquareFnLp_Martingale

open MeasureTheory ProbabilityTheory Filter Topology
open scoped ENNReal NNReal
Formal statement
namespace BurkholderDFI.Gundy

/-- Theorem 4.1, p. 23 (Gundy's decomposition, with one stopping time): an `L¹`-bounded martingale
`f` and `λ > 0` admit martingales `X, Y, Z` for the same filtration with `f = X + Y + Z`,
`‖X‖₂² ≤ 2λ‖f‖₁`, `‖Σ_{k ≥ 1} |y_k|‖₁ ≤ 4‖f‖₁` and `P(Z^* > 0) ≤ ‖f‖₁/λ`. -/
theorem theorem_4_1 {Ω : Type*} [mΩ : MeasurableSpace Ω] {P : Measure Ω} [IsProbabilityMeasure P]
    {ℱ : Filtration ℕ mΩ} {f : ℕ → Ω → ℝ} (hf : Martingale f ℱ P) (hL1 : BurkholderDFI.SquareFnLp.pNorm P 1 f < ⊤)
    (l : ℝ) (hl : 0 < l) :
    ∃ X Y Z : ℕ → Ω → ℝ, Martingale X ℱ P ∧ Martingale Y ℱ P ∧ Martingale Z ℱ P ∧
      (∀ n, 1 ≤ n → f n = X n + Y n + Z n) ∧
      BurkholderDFI.SquareFnLp.pNorm P 2 X ^ (2 : ℝ) ≤ ENNReal.ofReal (2 * l) * BurkholderDFI.SquareFnLp.pNorm P 1 f ∧
      ∫⁻ ω, ∑' k : ℕ, ENNReal.ofReal |BurkholderDFI.SquareFnLp.dseq Y (k + 1) ω| ∂P ≤ 4 * BurkholderDFI.SquareFnLp.pNorm P 1 f ∧
      P {ω | 0 < BurkholderDFI.SquareFnLp.maxFn Z ω} ≤ BurkholderDFI.SquareFnLp.pNorm P 1 f / ENNReal.ofReal l := by sorry

end BurkholderDFI.Gundy
Source
Burkholder, Distribution Function Inequalities for Martingales, Ann. Probability 1 (1973), Theorem 4.1, (4.1)–(4.4), p. 23
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 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