OctoInformBeta

OpenAI preprint, September 2026 · §6. From masked energies to the maximal transform

Proposition 6.5: Spatial (3/2)(3/2)-moment bounds over finite families of positive scales for continuous compactly supported or bounded box-supported inputs

Moment bounds for hard-truncated cyclic kernel operators.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

For complex-valued functions FF and GG on ℝ2\mathbb{R}^2, let B*(F,G)B_*(F, G) denote the maximal triangular Hilbert transform over finite hard-truncation annuli {t∈ℝ:ε<|t|<R}\{t \in \mathbb{R} : \varepsilon < |t| < R\} with 0<ε≤R0 < \varepsilon \le R, and let BB and CC be the determinant-one cyclic coordinate pullbacks of FF and GG with (x,y)=(−a,−b)(x, y) = (-a, -b). The question is how to bound the spatial integral of the maximal (3/2)(3/2)-moment of the cyclic hard-truncated integrals of BB and CC over any finite family of positive scale pairs (rk,Rk)(r_k, R_k) in terms of the L3L^3 norms of BB and CC, when BB and CC are continuous and either bounded on a box or compactly supported.

Definitions

A function f:ℝ2→Yf : \mathbb{R}^2 \to Y into a set YY with a zero element has box support M∈ℝM \in \mathbb{R} if |z1|≤M|z_1| \le M and |z2|≤M|z_2| \le M whenever f(z1,z2)≠0f(z_1, z_2) \ne 0.✦

Given B,C:ℝ2→ℂB, C : \mathbb{R}^2 \to \mathbb{C} and a kernel K:ℝ→ℝK : \mathbb{R} \to \mathbb{R}, the cyclic kernel operator TK(B,C):ℝ2→ℂT_K(B, C) : \mathbb{R}^2 \to \mathbb{C} is the function defined at z=(z1,z2)∈ℝ2z = (z_1, z_2) \in \mathbb{R}^2 by (1): TK(B,C)(z1,z2)=∫ℝB(z2,c)C(c,z1)K(z1+z2+c)dc.T_K(B, C)(z_1, z_2) = \int_{\mathbb{R}} B(z_2, c)\, C(c, z_1)\, K(z_1 + z_2 + c)\, dc.(1)✦

The hard-truncated Hilbert kernel Hr,R:ℝ→ℝH_{r,R} : \mathbb{R} \to \mathbb{R} at scale parameters r,R∈ℝr, R \in \mathbb{R} is the function defined for t∈ℝt \in \mathbb{R} by (2): Hr,R(t)=𝟏r<|t|−𝟏R<|t|t,H_{r,R}(t) = \frac{\mathbf{1}_{r < |t|} - \mathbf{1}_{R < |t|}}{t},(2)✦ with Hr,R(0)=0H_{r,R}(0) = 0.

For a complex number w∈ℂw \in \mathbb{C}, the (3/2)(3/2)-moment |w|3/2|w|^{3/2} is the real number defined by |w|3/2=|w||w||w|^{3/2} = |w|\sqrt{|w|}.✦ For a family f:κ→ℂf : \kappa \to \mathbb{C} indexed by a finite nonempty set κ\kappa, the finite maximal moment ℳκ(f)∈ℝ\mathcal{M}_\kappa(f) \in \mathbb{R} is the real number defined by (3): ℳκ(f)=maxk∈κ|f(k)|3/2.\mathcal{M}_\kappa(f) = \max_{k \in \kappa} |f(k)|^{3/2}.(3)✦

With variance σ∈ℝ\sigma \in \mathbb{R}, the Gaussian density gσ:ℝ→ℝg_\sigma : \mathbb{R} \to \mathbb{R} is the function defined at t∈ℝt \in \mathbb{R} by (4): gσ(t)=(2πσ)−1/2exp⁡(−t2/(2σ)).g_\sigma(t) = (2\pi\sigma)^{-1/2} \exp\bigl(-t^2/(2\sigma)\bigr).(4)✦ The heat kernel constant is the real number cheat=−23g3(0)c_{\mathrm{heat}} = -\frac{2}{3} g_3(0)✦, and the smooth kernel profile ψ:ℝ→ℝ\psi : \mathbb{R} \to \mathbb{R} and kernel error profile E:ℝ→ℝE : \mathbb{R} \to \mathbb{R} are the functions defined for v∈ℝv \in \mathbb{R} by (5): ψ(v)=cheat−2(v29−13)g3(v),E(v)=ψ(v)−𝟏|v|>1cheat.\psi(v) = c_{\mathrm{heat}} - 2\left(\frac{v^2}{9} - \frac{1}{3}\right) g_3(v), \qquad E(v) = \psi(v) - \mathbf{1}_{|v| > 1}\, c_{\mathrm{heat}}.(5) The error envelope constant Kenv>0K_{\mathrm{env}} > 0 is a chosen positive real constant satisfying |E(v)|≤Kenv|v|g6(v)|E(v)| \le K_{\mathrm{env}} |v| g_6(v) for all v∈ℝv \in \mathbb{R}.✦

The mesh mass constant cmesh∈ℝc_{\mathrm{mesh}} \in \mathbb{R}, jump scalar constant cjump∈ℝc_{\mathrm{jump}} \in \mathbb{R}, and strict positivity constant cstrict∈ℝc_{\mathrm{strict}} \in \mathbb{R} are the real numbers defined by (6): cmesh=2e1/2,cjump=2+482,cstrict=12min⁡(1−352,1110−122).c_{\mathrm{mesh}} = 2e^{1/2}, \qquad c_{\mathrm{jump}} = 2 + 48\sqrt{2}, \qquad c_{\mathrm{strict}} = \frac{1}{2}\min\left(1 - \frac{3}{5}\sqrt{2},\, \frac{11}{10} - \frac{1}{2}\sqrt{2}\right).(6) In terms of cmeshc_{\mathrm{mesh}}, cjumpc_{\mathrm{jump}}, cstrictc_{\mathrm{strict}}, and KenvK_{\mathrm{env}}, the masked constant cmask∈ℝc_{\mathrm{mask}} \in \mathbb{R}, selected scale constant csel∈ℝc_{\mathrm{sel}} \in \mathbb{R}, and error pairing constant cerr∈ℝc_{\mathrm{err}} \in \mathbb{R} are the real numbers defined by (7): cmask=cmesh(22+3cjump),csel=440cmaskcstrict,cerr=Kenvcmeshcjump.c_{\mathrm{mask}} = c_{\mathrm{mesh}}\bigl(2\sqrt{2} + 3 c_{\mathrm{jump}}\bigr), \qquad c_{\mathrm{sel}} = \frac{440\,c_{\mathrm{mask}}}{c_{\mathrm{strict}}}, \qquad c_{\mathrm{err}} = K_{\mathrm{env}}\, c_{\mathrm{mesh}}\, c_{\mathrm{jump}}.(7) The hard pairing constant chard∈ℝc_{\mathrm{hard}} \in \mathbb{R} and the moment bound constant Cmom∈ℝC_{\mathrm{mom}} \in \mathbb{R} are the real numbers defined by (8): chard=csel+2cerr|cheat|,Cmom=4chard(chard+1).c_{\mathrm{hard}} = \frac{c_{\mathrm{sel}} + 2 c_{\mathrm{err}}}{|c_{\mathrm{heat}}|}, \qquad C_{\mathrm{mom}} = 4\, c_{\mathrm{hard}}\bigl(c_{\mathrm{hard}} + 1\bigr).(8)

The Claim

Theorem 1. Let B,C:ℝ2→ℂB, C : \mathbb{R}^2 \to \mathbb{C} be continuous functions, let κ\kappa be a finite nonempty index set, and let rk,Rk∈ℝr_k, R_k \in \mathbb{R} satisfy 0<rk≤Rk0 < r_k \le R_k for all k∈κk \in \kappa. If BB and CC are bounded and both have box support MM for some M∈ℝM \in \mathbb{R}, then (9) holds: ∫ℝ2ℳκ((THrk,Rk(B,C)(z))k∈κ)dz≤Cmom(∫ℝ2|B(z)|3dz+∫ℝ2|C(z)|3dz).\int_{\mathbb{R}^2} \mathcal{M}_\kappa\Bigl(\bigl(T_{H_{r_k, R_k}}(B, C)(z)\bigr)_{k \in \kappa}\Bigr)\, dz \le C_{\mathrm{mom}} \left(\int_{\mathbb{R}^2} |B(z)|^3\, dz + \int_{\mathbb{R}^2} |C(z)|^3\, dz\right).(9)✦ If instead BB and CC are compactly supported, then the Lebesgue integral satisfies (10): ∫ℝ2ℳκ((THrk,Rk(B,C)(z))k∈κ)dz≤Cmom(∫ℝ2|B(z)|3dz+∫ℝ2|C(z)|3dz).\int_{\mathbb{R}^2} \mathcal{M}_\kappa\Bigl(\bigl(T_{H_{r_k, R_k}}(B, C)(z)\bigr)_{k \in \kappa}\Bigr)\, dz \le C_{\mathrm{mom}} \left(\int_{\mathbb{R}^2} |B(z)|^3\, dz + \int_{\mathbb{R}^2} |C(z)|^3\, dz\right).(10)✦ Under the same compact-support hypothesis, writing a+=max⁡(a,0)∈[0,∞]a_+ = \max(a, 0) \in [0, \infty], the unsigned Lebesgue integral in [0,∞][0, \infty] satisfies (11): ∫ℝ2(ℳκ((THrk,Rk(B,C)(z))k∈κ))+dz≤(Cmom(∫ℝ2|B(z)|3dz+∫ℝ2|C(z)|3dz))+.\int_{\mathbb{R}^2} \left(\mathcal{M}_\kappa\Bigl(\bigl(T_{H_{r_k, R_k}}(B, C)(z)\bigr)_{k \in \kappa}\Bigr)\right)_+\, dz \le \left(C_{\mathrm{mom}} \left(\int_{\mathbb{R}^2} |B(z)|^3\, dz + \int_{\mathbb{R}^2} |C(z)|^3\, dz\right)\right)_+.(11)✦

In the paper, the result reads: for all complex-valued F,G∈Cc∞(ℝ2)F, G \in C_c^\infty(\mathbb{R}^2), ∥B*(F,G)∥L3/2(ℝ2)≤C0∥F∥L3(ℝ2)∥G∥L3(ℝ2)\|B_*(F, G)\|_{L^{3/2}(\mathbb{R}^2)} \le C_0 \|F\|_{L^3(\mathbb{R}^2)} \|G\|_{L^3(\mathbb{R}^2)} with an absolute constant C0C_0, where the supremum defining B*B_* is over both finite hard-truncation endpoints inside the spatial norm, and BB and CC stand for the cyclic coordinate pullbacks of FF and GG. The bounds (9), (10), and (11) work directly in the cyclic coordinates BB and CC under continuity alone, replace the full supremum with an arbitrary finite scale family κ\kappa, and take an additive (3/2)(3/2)-power form with explicit constant CmomC_{\mathrm{mom}} in place of the L3L^3 product.

What Is Checked

The claims (9), (10), and (11) are stated over continuous complex-valued functions B,CB, C on ℝ2\mathbb{R}^2 that are either bounded with box support MM or compactly supported, finite nonempty families κ\kappa of positive scale pairs (rk,Rk)(r_k, R_k) with rk≤Rkr_k \le R_k, the cyclic kernel operator, the hard-truncated Hilbert kernel Hrk,RkH_{r_k, R_k}, the finite maximal (3/2)(3/2)-moment ℳκ\mathcal{M}_\kappa, the moment bound constant CmomC_{\mathrm{mom}} and its defining Gaussian density, kernel profiles, and pairing constants, under both the Lebesgue integral and the unsigned Lebesgue integral in [0,∞][0, \infty]. It also holds that there exists a positive real constant KK such that for every real number vv, the absolute value of the kernel error profile at vv is at most K|v|g6(v)K |v| g_6(v).✦