Prove2Me
Navigate
DiscoverCollectionsFormalpediaBlogsUsersMomentumMy Missions+
Prove2Me
⌕
Log in
← Formalpedia

The Bochner–Riesz conjecture implies the restriction conjecture

Open
TaoBR.restrictionConjecture_of_bochnerRieszConjecture

by dbenbenn · Oct 29, 2026 · Mathlib 0df444a (Lean v4.33.1)

bochner-rieszfourier-restrictionharmonic-analysiskakeya

For every dimension n≥2n \ge 2n≥2: if the Bochner–Riesz conjecture holds in Rn\mathbf R^nRn (BochnerRieszConjecture n: BR(p,ε)BR(p, \varepsilon)BR(p,ε) for all 1≤p≤∞1 \le p \le \infty1≤p≤∞ and ε>0\varepsilon > 0ε>0), then the restriction conjecture holds in Rn\mathbf R^nRn (RestrictionConjecture n: R(p,0)R(p, 0)R(p,0) for all 1≤p<2n/(n+1)1 \le p < 2n/(n+1)1≤p<2n/(n+1)).

Tao, p. 2: “In this note we present a formal proof of the implication in the reverse direction, that the Bochner-Riesz conjecture implies the restriction conjecture, as an immediate consequence of the following two results.” The title of the paper states the same implication.

Preamble
import Mathlib
import Definitions.Def_TaoBR
Formal statement
namespace TaoBR

theorem restrictionConjecture_of_bochnerRieszConjecture (n : ℕ) (hn : 2 ≤ n) :
    BochnerRieszConjecture n → RestrictionConjecture n := by
  sorry

end TaoBR
Source
Tao, T., The Bochner–Riesz conjecture implies the restriction conjecture, Duke Math. J. 96 (1999) 363–375, https://doi.org/10.1215/S0012-7094-99-09610-2 (the author's preprint, https://www.math.ucla.edu/~tao/preprints/restriction.dvi, whose page numbers are used), p. 2, the implication Bochner–Riesz ⟹ restriction (the title)
Read-back

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

Statement. Let nnn be a natural number with n≥2n \ge 2n≥2. If the Bochner–Riesz estimate (H) below holds in dimension nnn, then the restriction estimate (C) below holds in dimension nnn.

This is a one-way implication, (H) ⇒\Rightarrow⇒ (C), for each fixed n≥2n \ge 2n≥2. There are no other variables or assumptions: the only inputs are the natural number nnn and the hypothesis n≥2n \ge 2n≥2.


Setting and conventions

  • Space. Rn\mathbb{R}^nRn carries the Euclidean inner product ⟨x,ξ⟩=∑i=1nxiξi\langle x, \xi\rangle = \sum_{i=1}^n x_i \xi_i⟨x,ξ⟩=∑i=1n​xi​ξi​ and the Euclidean norm ∣x∣=(∑ixi2)1/2|x| = \bigl(\sum_i x_i^2\bigr)^{1/2}∣x∣=(∑i​xi2​)1/2. Its measure dxdxdx is Lebesgue measure, so the unit cube has volume 111.
  • Test functions. S(Rn)\mathcal{S}(\mathbb{R}^n)S(Rn) is the Schwartz class of complex-valued functions fff that are C∞C^\inftyC∞ and satisfy sup⁡x∣x∣k ∥Djf(x)∥<∞\sup_x |x|^k\,\|D^j f(x)\| < \inftysupx​∣x∣k∥Djf(x)∥<∞ for all k,j≥0k, j \ge 0k,j≥0.
  • Fourier transform. For f:Rn→Cf : \mathbb{R}^n \to \mathbb{C}f:Rn→C,
f^(ξ)=∫Rne−2πi⟨x,ξ⟩ f(x) dx,\widehat{f}(\xi) = \int_{\mathbb{R}^n} e^{-2\pi i \langle x, \xi\rangle}\, f(x)\, dx ,f​(ξ)=∫Rn​e−2πi⟨x,ξ⟩f(x)dx,

and the inverse transform of g:Rn→Cg : \mathbb{R}^n \to \mathbb{C}g:Rn→C is

gˇ(x)=∫Rne2πi⟨ξ,x⟩ g(ξ) dξ.\check{g}(x) = \int_{\mathbb{R}^n} e^{2\pi i \langle \xi, x\rangle}\, g(\xi)\, d\xi .gˇ​(x)=∫Rn​e2πi⟨ξ,x⟩g(ξ)dξ.

Both are Lebesgue (Bochner) integrals, and by convention an integral whose integrand is not absolutely integrable is assigned the value 000. For Schwartz fff the integral defining f^(ξ)\widehat f(\xi)f​(ξ) converges absolutely, so this convention does not affect f^\widehat ff​.

  • LpL^pLp norms. For a measure μ\muμ and an exponent p∈[1,∞]p \in [1,\infty]p∈[1,∞], ∥g∥Lp(μ)\|g\|_{L^p(\mu)}∥g∥Lp(μ)​ is the extended nonnegative number (∫∣g∣p dμ)1/p∈[0,∞]\bigl(\int |g|^p\, d\mu\bigr)^{1/p} \in [0,\infty](∫∣g∣pdμ)1/p∈[0,∞] when p<∞p < \inftyp<∞, and the μ\muμ-essential supremum of ∣g∣|g|∣g∣ when p=∞p = \inftyp=∞. Every inequality below compares elements of [0,∞][0,\infty][0,∞], and the constants CCC are finite nonnegative reals.

The hypothesis (H): Bochner–Riesz estimates in dimension nnn

For a real number δ\deltaδ, define the symbol

mδ(ξ)=(1−∣ξ∣2)+δ={(1−∣ξ∣2)δif ∣ξ∣<1,0if ∣ξ∣≥1,m_\delta(\xi) = (1 - |\xi|^2)_+^{\delta} = \begin{cases} (1 - |\xi|^2)^{\delta} & \text{if } |\xi| < 1,\\ 0 & \text{if } |\xi| \ge 1,\end{cases}mδ​(ξ)=(1−∣ξ∣2)+δ​={(1−∣ξ∣2)δ0​if ∣ξ∣<1,if ∣ξ∣≥1,​

where the power is the real power of a positive number. Define the Bochner–Riesz operator of order δ\deltaδ on functions f:Rn→Cf : \mathbb{R}^n \to \mathbb{C}f:Rn→C by

Sδf(x)=∫Rne2πi⟨ξ,x⟩ mδ(ξ) f^(ξ) dξ,S^\delta f(x) = \int_{\mathbb{R}^n} e^{2\pi i \langle \xi, x\rangle}\, m_\delta(\xi)\, \widehat f(\xi)\, d\xi ,Sδf(x)=∫Rn​e2πi⟨ξ,x⟩mδ​(ξ)f​(ξ)dξ,

that is, the inverse Fourier transform of mδf^m_\delta \widehat fmδ​f​ (with the convention above that a non-integrable integrand gives 000).

For p∈[1,∞]p \in [1, \infty]p∈[1,∞], define the critical index

β(p)=max⁡(n ∣1p−12∣−12, 0),\beta(p) = \max\Bigl( n\,\Bigl|\tfrac1p - \tfrac12\Bigr| - \tfrac12,\ 0 \Bigr),β(p)=max(n​p1​−21​​−21​, 0),

with 1∞=0\tfrac1\infty = 0∞1​=0. So β(p)≥0\beta(p) \ge 0β(p)≥0 always, and β(∞)=max⁡(n−12,0)=n−12\beta(\infty) = \max\bigl(\tfrac{n-1}{2}, 0\bigr) = \tfrac{n-1}{2}β(∞)=max(2n−1​,0)=2n−1​.

(H) asserts: for every exponent ppp with 1≤p≤∞1 \le p \le \infty1≤p≤∞ (the endpoint p=∞p = \inftyp=∞ included) and every real ε>0\varepsilon > 0ε>0, there is a finite constant C≥0C \ge 0C≥0 such that

∥Sβ(p)+εf∥Lp(Rn)  ≤  C ∥f∥Lp(Rn)for every f∈S(Rn).\bigl\| S^{\beta(p) + \varepsilon} f \bigr\|_{L^p(\mathbb{R}^n)} \;\le\; C\, \|f\|_{L^p(\mathbb{R}^n)} \qquad \text{for every } f \in \mathcal{S}(\mathbb{R}^n).​Sβ(p)+εf​Lp(Rn)​≤C∥f∥Lp(Rn)​for every f∈S(Rn).

The constant CCC may depend on nnn, ppp and ε\varepsilonε, but not on fff. The estimate is required only for Schwartz functions, not for all of LpL^pLp.

Because ε>0\varepsilon > 0ε>0 and β(p)≥0\beta(p) \ge 0β(p)≥0, the order δ=β(p)+ε\delta = \beta(p) + \varepsilonδ=β(p)+ε is strictly positive. So mδm_\deltamδ​ takes values in [0,1][0,1][0,1] and vanishes outside the open unit ball, and mδf^m_\delta \widehat fmδ​f​ is a bounded measurable function supported in the closed unit ball. Its inverse-transform integral therefore converges absolutely, and the zero-value convention plays no role in (H).


The conclusion (C): restriction estimates for the unit sphere in dimension nnn

Let Sn−1={ω∈Rn:∣ω∣=1}S^{n-1} = \{\omega \in \mathbb{R}^n : |\omega| = 1\}Sn−1={ω∈Rn:∣ω∣=1} be the unit sphere, regarded as a measure space in its own right. Its measurable sets are the traces on Sn−1S^{n-1}Sn−1 of Borel sets of Rn\mathbb{R}^nRn. Its measure σ\sigmaσ is the cone measure

σ(A)=n⋅∣{ tω:0<t<1, ω∈A }∣(A⊆Sn−1 measurable),\sigma(A) = n \cdot \bigl|\{\, t\omega : 0 < t < 1,\ \omega \in A \,\}\bigr| \qquad (A \subseteq S^{n-1} \text{ measurable}),σ(A)=n⋅​{tω:0<t<1, ω∈A}​(A⊆Sn−1 measurable),

where ∣⋅∣|\cdot|∣⋅∣ is Lebesgue measure on Rn\mathbb{R}^nRn: nnn times the volume of the open cone over AAA inside the unit ball, with the origin removed. Its total mass is σ(Sn−1)=n⋅∣B(0,1)∣\sigma(S^{n-1}) = n \cdot |B(0,1)|σ(Sn−1)=n⋅∣B(0,1)∣.

For f:Rn→Cf : \mathbb{R}^n \to \mathbb{C}f:Rn→C, write f^∣Sn−1\widehat f|_{S^{n-1}}f​∣Sn−1​ for the function ω↦f^(ω)\omega \mapsto \widehat f(\omega)ω↦f​(ω) on the sphere. Write B‾R={x:∣x∣≤R}\overline{B}_R = \{x : |x| \le R\}BR​={x:∣x∣≤R} for the closed ball of radius RRR about the origin, and supp⁡f\operatorname{supp} fsuppf for the closure of {x:f(x)≠0}\{x : f(x) \ne 0\}{x:f(x)=0}.

(C) asserts: for every exponent ppp with

1≤p<2nn+11 \le p < \frac{2n}{n+1}1≤p<n+12n​

(real division, so the bound is the real number 2n/(n+1)2n/(n+1)2n/(n+1); this forces ppp to be finite, and for n≥2n \ge 2n≥2 the range is nonempty because 2n/(n+1)≥4/32n/(n+1) \ge 4/32n/(n+1)≥4/3), there is a finite constant C≥0C \ge 0C≥0 such that for every real R≥1R \ge 1R≥1 and every f∈S(Rn)f \in \mathcal{S}(\mathbb{R}^n)f∈S(Rn) with supp⁡f⊆B‾R\operatorname{supp} f \subseteq \overline{B}_Rsuppf⊆BR​,

∥f^∣Sn−1∥Lp(Sn−1, σ)  ≤  C⋅R0⋅∥f∥Lp(B‾R, dx).\bigl\| \widehat f|_{S^{n-1}} \bigr\|_{L^p(S^{n-1},\, \sigma)} \;\le\; C \cdot R^{0} \cdot \|f\|_{L^p(\overline{B}_R,\, dx)} .​f​∣Sn−1​​Lp(Sn−1,σ)​≤C⋅R0⋅∥f∥Lp(BR​,dx)​.

Remarks on reading (C) literally:

  • The factor R0R^0R0 is the real power R0=1R^0 = 1R0=1, so the bound is just C ∥f∥Lp(B‾R)C\,\|f\|_{L^p(\overline B_R)}C∥f∥Lp(BR​)​. The constant CCC depends on nnn and ppp only. It is chosen before RRR and fff, so it is uniform in the radius R≥1R \ge 1R≥1 and in fff.
  • The same exponent ppp appears on both sides: an LpL^pLp norm of fff on the ball controls an Lp(σ)L^p(\sigma)Lp(σ) norm of f^\widehat ff​ on the sphere.
  • The right-hand norm is taken over the closed ball B‾R\overline B_RBR​ with Lebesgue measure. Since fff vanishes outside B‾R\overline B_RBR​, it equals ∥f∥Lp(Rn)\|f\|_{L^p(\mathbb{R}^n)}∥f∥Lp(Rn)​.
  • The functions considered are exactly the Schwartz functions with compact support in B‾R\overline B_RBR​, that is, smooth compactly supported functions. This class is nonempty (it contains 000 and smooth bump functions), so (C) is not vacuous.
Human review
  • Endorsed by Shuze Chen · Oct 29, 2026

    Confirmed by the moderator at approval.

  • Endorsed by dbenbenn · Oct 29, 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 paper
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me