Pipeline-generated

How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

✓ 19 results · all verified · 10 also independently AI-judged

Every result on this page is machine-checked by a proof checker and read in full and owner-audited; the judge is an additional, independent cross-model AI review of the proofs. The 9 not AI-judged were verified by owner audit (typically over a confirmed judge false positive), not failures.

Fourier Restriction and the Stein–Tomas Theorem

1 · Prerequisites

2 · Summary

This page develops $L^{2}$ -density Fourier restriction theory for a compact curved hypersurface. The definition fixes the pointwise restriction operator $R_{0} f = \hat{f} ∣_{S}$ on Schwartz data — necessarily, since surface measure is carried by a Lebesgue-null set — and the adjoint extension operator $E g = (g σ)^{\lor}$ , together with its boundedness, uniform continuity and $L^{1}$ -- $L^{2}$ bound. A finite-measure pairing lemma supplies the two identities that drive everything: the $L^{2} (σ)$ pairing against Schwartz data and the convolution identity $(\hat{F} μ)^{\lor} = F * \overset{ˇ}{μ}$ . Restriction and extension estimates are then proved to be equivalent, with equal least constants and adjoint extensions.

The endpoint route is local. One-dimensional van der Corput estimates are proved by bounded primitives, and a separate multidimensional stationary-phase argument gives the decay of the spherical surface measure, $∣ \hat{σ} (ξ) ∣ ≲ (1 + ∣ ξ ∣)^{- (n - 1) / 2}$ , and, after the finite spherical graph partition and the localized curved-patch decay, the graph-patch slice family $U (t)$ obeys the dispersive bound $∥ U (t) g ∥_{\infty} ≲ ⟨ t ⟩^{- (n - 1) / 2} ∥ g ∥_{1}$ and the uniform Plancherel bound $∥ U (t) g ∥_{2} ≲ ∥ g ∥_{2}$ . Interpolating these at $p_{0} = 2 (n + 1) / (n + 3)$ produces the decay $⟨ t ⟩^{- β}$ with $β = (n - 1) / (n + 1)$ and exponent identity $1 / p_{0} - 1 / p_{0}^{'} = 2 / (n + 1) = 1 - β$ , so the one-dimensional Hardy–Littlewood–Sobolev theorem of order $2 / (n + 1)$ bounds the slice convolution and completes the $T T^{*}$ reduction. Knapp cap packets on dual tubes of dimensions $δ^{- 1} \times \dots \times δ^{- 1} \times δ^{- 2}$ give the matching necessary condition $p \leq p_{0}$ (equivalently $q \geq q_{0} = 2 (n + 1) / (n - 1)$ ), and the finite curved graph localization transfers the theorem to compact hypersurfaces with everywhere nonvanishing extrinsic Gaussian curvature.

Two recorded orientations close the page: the general restriction problem remains open for $n \geq 3$ , and the Strichartz comparison concerns paraboloid extension and is non-load-bearing in this run. Countable Choice is declared and propagated through the chart, partition, density, duality and extension interfaces used by the proofs.

3 · Logical flowchart

4 · Definitions, theorems and proofs

DefinitionDefinition: Literature-sourcedProof: Not applicablejudge pass (gpt-6.1-sol)Open item page →

Euclidean hypersurface normals, shape operators and curvature

Definition

For $n \geq 2$ , a smooth embedded hypersurface $S \subset R^{n}$ has the embedded-submanifold meaning of Embedded submanifolds and slice charts. If $X : U \subset R^{n - 1} \to S$ is a smooth local parametrization of rank $n - 1$ , define $T_{X (y)} S = im D X (y)$ with its Euclidean inner product. A smooth local unit normal is a smooth map $ν : V \to R^{n}$ on a relatively open subset $V \subseteq S$ with $∣ ν ∣ = 1$ and $ν ⊥ T S$ . Define the Euclidean shape operator by $S_{ν} v = - d ν_{p} (v)$ on $T_{p} S$ , and the extrinsic Gaussian (Gauss–Kronecker) curvature by $K_{ν} (p) = \det S_{ν} (p)$ . Here $d ν_{p} (D X (y) u) = D (ν \circ X) (y) u$ . Smooth functions and compact supports on $S$ use its subspace topology and these local parametrizations. Nonvanishing curvature means $K_{ν} \neq 0$ for either choice of local unit normal at each point. The graph and localization lemma Smooth Euclidean hypersurface graphs and compact localization ↗ proves that these definitions are independent of parametrization, that the derivative takes values in $T_{p} S$ , and that changing the unit normal only changes the sign of the shape operator. This is the Euclidean specialization of the usual Weingarten definition; the equivalence is proved there, without requiring the later Riemannian theory.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Smooth Euclidean hypersurface graphs and compact localization

Statement

For $n \geq 2$ , every smooth embedded hypersurface $S \subset R^{n}$ is locally, after a rigid motion, the graph $X (y) = (y, h (y))$ of a $C^{\infty}$ function. The tangent, normal, shape operator and curvature of Euclidean hypersurface normals, shape operators and curvature are well defined and agree with their usual Euclidean hypersurface meanings. Every continuous unit normal on $S$ is locally smooth. Every compact $K \subset S$ admits finitely many graph pieces $S_{j}$ and nonnegative smooth functions $χ_{j}$ on $S$ , compactly supported in $S_{j}$ , whose sum is one on a neighbourhood of $K$ . If $S$ itself is compact, that sum is one everywhere.

Facts & Assumptions

[F1]

The earlier inverse and implicit function theorems supply $C^{1}$ inverses and their derivative formulas. (The Euclidean inverse function theorem, The Euclidean implicit function theorem with derivative formula)

[F2]

The first-order total chain rule and sum/scalar rules hold. One-variable product and quotient rules apply on coordinate lines; continuous partials give total derivatives, whose columns are the partials. Smoothness is defined by all ordered iterated partials, mixed partials are symmetric at the stated regularity, and the cited determinant and spectral identities hold. (The chain rule for total derivatives: $D (g \circ f) (a) = D g (f (a)) \circ D f (a)$ , Sums and scalar multiples of totally differentiable maps are totally differentiable with the expected derivatives, Sums, scalar multiples, products and quotients: $(f + g)^{'} (c) = f^{'} (c) + g^{'} (c)$ , $(α f)^{'} (c) = α f^{'} (c)$ , $(f g)^{'} (c) = f^{'} (c) g (c) + f (c) g^{'} (c)$ , and $(f / g)^{'} (c) = (f^{'} (c) g (c) - f (c) g^{'} (c)) / g (c)^{2}$ when $g (c) \neq 0$ , If all partial derivatives exist on a neighbourhood and are continuous at a point, then the map is totally differentiable there with Jacobian derivative, A total derivative computes every directional derivative, and its matrix is the Jacobian, $C^{k}$ maps and multi-index derivative notation in Euclidean space, Continuous mixed partials of order $k$ are invariant under permutations, Laplace expansion computes the determinant along every row and every column over a commutative ring, For same-sized finite square matrices over a commutative ring, $\det (A B) = \det (A) \det (B)$ , Real spectral theorem: a self-adjoint endomorphism of a finite-dimensional real inner product space has an orthonormal eigenbasis)

[F3]

Embedded hypersurfaces have slice charts. (Embedded submanifolds and slice charts)

[F4]

Tangents, normals and curvature have the local Euclidean definitions. (Euclidean hypersurface normals, shape operators and curvature)

[F5]

Smooth cutoffs exist on Euclidean balls. (Explicit compactly supported smooth cutoffs)

[F6]

Euclidean compactness gives finite subcovers and extrema. (For a nonempty subset of $R^{n}$ with $n \geq 1$ , compactness, closedness and boundedness, pseudocompactness, and attainment of extrema by every continuous real-valued function are equivalent)

Proof

Given: A smooth embedded hypersurface $S$ and, for the localization assertion, a compact subset $K \subset S$ .

1.1F1F2algebra

Smooth inverse bootstrap. On coordinate lines the one-variable rules [F2] give $\partial_{j} (u v) = (\partial_{j} u) v + u \partial_{j} v$ and $\partial_{j} (1 / u) = - (\partial_{j} u) / u^{2}$ where $u \neq 0$ . Induction on derivative order proves that products and nonvanishing quotients of $C^{r}$ functions are $C^{r}$ . For compositions, [F2] gives $\partial_{j} (f \circ g) = \sum_{k} (\partial_{k} f \circ g) \partial_{j} g_{k}$ ; induction using the product rule and continuity proves $C^{r}$ closure under composition for each finite $r$ . The earlier $C^{1}$ inverse theorem [F1] gives a $C^{1}$ inverse $g$ to a smooth map $f$ with invertible derivative, with $D g = (D f \circ g)^{- 1}$ . For an invertible finite matrix $A$ , cofactor expansion gives $A^{- 1} = adj (A) / \det A$ : multiplying either side by $A$ gives the identity by Laplace expansion, including the off-diagonal expansions with two equal rows. Its entries are polynomial quotients with nonzero denominator, hence smooth. If $g$ is $C^{r}$ and $f$ smooth, the repeated product and chain rules make $(D f \circ g)^{- 1}$ $C^{r}$ , so the displayed derivative makes $g$ $C^{r + 1}$ . Induction from $r = 1$ proves $g$ smooth. The same bootstrap applies to the implicit theorem, whose solution is a component of the inverse of $(x, z) \mapsto (x, F (x, z))$ .

2.1F1F2F3step 1.1algebra

Graphs from slices. In a slice chart $θ$ near $p$ , let $F$ be its last coordinate. Then $S \cap V = F^{- 1} (0)$ and $D F$ has rank one, because $D θ$ is invertible by differentiating the chart inverse identities. Apply the real spectral theorem in [F2] to the self-adjoint rank-one orthogonal projection $v \mapsto (v \cdot u) u$ , where $u = \nabla F (p) / ∣ \nabla F (p) ∣$ . Its eigenvalue-one space is $R u$ , so, after reordering and changing one sign, its orthonormal eigenbasis has last vector $u$ ; in these rigid coordinates $\partial_{n} F (p) \neq 0$ . The implicit theorem and step 1.1 solve $F (y, z) = 0$ as $z = h (y)$ on a product neighbourhood, with $h$ smooth. Projection onto $y$ is the inverse of $X (y) = (y, h (y))$ ; its derivative has independent columns $(e_{j}, h_{j})$ , so $X$ is a smooth parametrization of the relatively open piece.

3.1F2F4step 1.1step 2.1algebra

Coordinate independence and normal derivatives. If two parametrizations overlap, their transition is smooth and has invertible derivative, so the chain rule makes their derivative images identical and makes $D (ν \circ X) (D X)^{- 1}$ independent of the parametrization. The positive square root is smooth because it is the inverse of $t \mapsto t^{2}$ on $(0, \infty)$ , to which step 1.1 applies. For a graph, $ν_{h} = (- \nabla h, 1) / \sqrt{1 + ∣ \nabla h ∣^{2}}$ is therefore smooth, unit, and perpendicular to every $(e_{j}, h_{j})$ . Since the normal space has dimension one, any continuous unit normal is $ϵ ν_{h}$ with $ϵ$ continuous and valued in ${1, - 1}$ , hence constant on a sufficiently small connected piece. It is therefore smooth. Differentiating $∣ ν ∣^{2} = 1$ gives $d ν (v) \cdot ν = 0$ , so $d ν (v) \in T_{p} S$ . Also differentiating $ν \cdot X_{k} = 0$ gives $- \partial_{j} ν \cdot X_{k} = ν \cdot X_{j k}$ , a symmetric expression by equality of mixed partials. Thus $S_{ν}$ is self-adjoint.

4.1F2F4step 3.1algebra

Equivalence with the Euclidean Weingarten operator. Cartesian differentiation of an ambient field along a curve is its componentwise derivative; extending a field from a graph by holding its graph coordinates fixed in the last ambient coordinate gives exactly that derivative along tangent vectors, independent of the extension because two extensions agree on every curve in $S$ . Orthogonal projection therefore gives the usual Euclidean second fundamental form $(D_{v} w)^{⊥} = (D_{v} w \cdot ν) ν$ . The differentiated orthogonality in step 3.1 gives $⟨ S_{ν} v, w ⟩ = ⟨ (D_{v} w)^{⊥}, ν ⟩$ and $S_{ν} = - (D_{v} ν)^{⊤} = - d ν (v)$ . This is precisely the shape operator for the Euclidean ambient connection. Its self-adjoint eigenvalues are the usual principal curvatures, and their product is its determinant. Replacing $ν$ by $- ν$ replaces $S_{ν}$ by $- S_{ν}$ , so $K_{- ν} = (- 1)^{n - 1} K_{ν}$ and nonvanishing is independent of local orientation. Rigid motions conjugate the shape operators by their orthogonal derivative and preserve the determinant.

5.1F5F6step 2.1algebra∎

Compact localization. For each point of $K$ , choose a graph neighbourhood and an ambient ball whose closed, slightly larger ball meets $S$ only inside that graph neighbourhood. An ambient smooth bump supported in the larger ball and equal to one on the smaller ball exists by [F5]. Compactness gives finitely many smaller balls covering $K$ , with bumps $β_{j}$ . Their restrictions have supports compact in their graph pieces: inside the larger closed ball the hypersurface is a relatively closed zero set of the defining function from step 2.1. Put $B = \sum_{j} β_{j}$ . It has positive minimum on $K$ . Choose a smooth real cutoff $η$ that is zero for $B \leq c / 2$ and one for $B \geq c$ , where $0 < c < \min_{K} B$ . Define $χ_{j} = η (B) β_{j} / B$ where $B > 0$ , and zero where $B = 0$ . These functions are smooth, nonnegative, supported compactly in their graph pieces, and sum to $η (B) = 1$ on a neighbourhood of $K$ . If $K = S$ , the sum is one on $S$ . If $K$ is empty, take the empty family.

DefinitionDefinition: Literature-sourcedProof: Not applicableOpen item page →

Fourier restriction and adjoint extension operators

Definition

Assume the Axiom of Countable Choice (The Axiom of Countable Choice ( ${A C}_{ω}$ )). Fix $n \geq 2$ and a compact embedded $C^{\infty}$ hypersurface $S \subseteq R^{n}$ . Its surface measure $σ$ is fixed as follows.

For $S = S^{n - 1}$ it is the polar surface measure of The polar surface set function on the unit sphere, which agrees with the chart surface measure by Agreement with the existing polar sphere measure.
For a general compact hypersurface it is the chart measure of Surface integration on compact C1 hypersurfaces, a finite Borel measure whose graph density is $\sqrt{1 + ∣ D h ∣^{2}}$ by Chart and partition independence of surface measure.

Restriction. For a Schwartz function $f \in S (R^{n})$ the transform $\hat{f}$ is again a Schwartz function, in particular an actual smooth function on $R^{n}$ (Fourier transform acts continuously on Schwartz space, Schwartz space and its seminorms), so $R_{0} f : = \hat{f} ∣_{S}$ is a pointwise-defined function on $S$ .

Extension. For $g \in L^{1} (σ; C)$ the set function $g σ : E \mapsto \int_{E} g d σ$ is a complex measure on the Borel sets of $S$ with total variation $∣ g ∣ σ$ (A complex L^1 density defines a complex measure whose total variation is |h| dmu, A complex measure is a finite-valued countably additive set function), and one writes $(g σ)^{\lor} (x) : = \hat{g σ} (- x)$ for the reflected transform of that finite measure, so that $E g : = (g σ)^{\lor}, E g (x) = \int_{S} e^{2 π i x \cdot ω} g (ω) d σ (ω) (x \in R^{n}) .$ The transform theorem Fourier transform of a finite complex Borel measure gives that $E g$ is a bounded uniformly continuous function and that $∣ E g (x) ∣ \leq ∣ g σ ∣ (R^{n}) = \int_{S} ∣ g ∣ d σ = ∥ g ∥_{L^{1} (σ)} (x \in R^{n}) .$ For $g \in L^{2} (σ; C)$ , finiteness of $σ$ gives $g \in L^{1} (σ; C)$ and the additional estimate $∣ E g (x) ∣ \leq σ (S)^{1 / 2} ∥ g ∥_{L^{2} (σ)} (x \in R^{n}),$ where the last inequality is the $p = p^{'} = 2$ case of the complex Holder/Cauchy–Schwarz inequality Complex Holder, Minkowski, and the quotient norm together with finiteness of the surface measure; here $L^{p} (σ; C)$ are the complex Lebesgue classes of Complex Lp classes and Euclidean test-function conventions. The assignment $g \mapsto E g$ is complex-linear, since $g \mapsto g σ$ is additive in the density and integration against a finite measure is additive. The symbol $R$ denotes the unique bounded extension of $R_{0}$ when one exists.

Restriction starts on Schwartz data by necessity: for $S = S^{n - 1}$ the measure $σ$ is carried by the Lebesgue-null set $S$ (The unit sphere is Lebesgue null), so no pointwise restriction is available for a general ambient $L^{p^{'}}$ class; the companion page's counterexample records the explicit failure of well-definedness on equivalence classes.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Fourier pairing for a finite measure and Schwartz data

Statement

Assume Countable Choice. Let $μ$ be a finite complex Borel measure on $R^{n}$ and let $F, G \in S (R^{n})$ . (i) $\int_{R^{n}} (G μ)^{\lor} (x) \overline{F (x)} d x = \int_{R^{n}} G (ω) \overline{\hat{F} (ω)} d μ (ω)$ , where $(G μ)^{\lor} (x) = \int e^{2 π i x \cdot ω} G (ω) d μ (ω)$ . (ii) Writing $\overset{ˇ}{μ} (x) : = \hat{μ} (- x) = \int e^{2 π i x \cdot ω} d μ (ω)$ , one has $(\hat{F} μ)^{\lor} = F * \overset{ˇ}{μ}$ as everywhere-defined bounded continuous functions. (iii) $∣ \overset{ˇ}{μ} (x) ∣ \leq ∣ μ ∣ (R^{n})$ for every $x$ , and $\overset{ˇ}{μ}$ is uniformly continuous.

Facts & Assumptions

Given: Countable Choice, a finite complex Borel measure $μ$ on $R^{n}$ with $M : = ∣ μ ∣ (R^{n}) < \infty$ , and Schwartz functions $F, G \in S (R^{n})$ .

[F1]

Every complex measure has finite total variation: $∣ ν ∣ (X) < \infty$ ; in particular $M < \infty$ . (Every complex measure has finite total variation)

[F2]

For a complex Borel measure $ν$ of finite total variation, $\hat{ν} (ξ) = \int e^{- 2 π i x \cdot ξ} d ν (x)$ is bounded and uniformly continuous with $\sup ∣ \hat{ν} ∣ \leq ∣ ν ∣ (R^{n})$ . (Fourier transform of a finite complex Borel measure)

[F3]

Fubini and Tonelli hold on sigma-finite products: Tonelli's identity for nonnegative product-measurable integrands, and the threefold equality $\int f d (μ \times ν) = \int \int f d ν d μ = \int \int f d μ d ν$ for $f \in L^{1} (μ \times ν)$ . (Fubini's theorem for L^1 functions on a sigma-finite product, Tonelli's theorem for nonnegative measurable functions on a sigma-finite product)

[F4]

Schwartz functions and their transforms are bounded and integrable: $F : S \to S$ is continuous, and $x^{α} \partial^{β} h \in L^{p}$ for all $p$ with norm bounded by finitely many Schwartz seminorms; in particular $F, \hat{F}, G \in L^{1} \cap L^{\infty}$ . (Fourier transform acts continuously on Schwartz space, Schwartz derivatives are integrable)

[F5]

Fourier transform laws for $f \in L^{1} (R^{n}; C)$ : with $τ_{a} f (x) = f (x - a)$ , $\hat{τ_{a} f} (ξ) = e^{- 2 π i a \cdot ξ} \hat{f} (ξ)$ , $\hat{f (- \cdot)} (ξ) = \hat{f} (- ξ)$ , and $\hat{\overline{f}} (ξ) = \overline{\hat{f} (- ξ)}$ , all at every frequency. (Translation, modulation, linear dilation and reflection laws)

[F6]

Convolution: $(f * g) (x) = \int f (x - y) g (y) d y$ whenever $y \mapsto f (x - y) g (y)$ is measurable and integrable. (Convolution of two functions on $R^{n}$ )

[F7]

Measurability toolkit: $B (R^{2 n}) = B (R^{n}) \otimes B (R^{n})$ ; composition of a Borel function with a Borel function is Borel; sums, products and scalar multiples of Borel functions are Borel; $\sin$ and $\cos$ are $1$ -Lipschitz, and $e^{2 π i x \cdot ω} = \cos (2 π x \cdot ω) + i \sin (2 π x \cdot ω)$ in the Cartesian form of the exponential. (The Borel product of R^m and R^n is the Borel sigma-algebra of R^{m+n}, Composition with a Borel measurable outer map preserves measurability, Arithmetic and lattice operations preserve measurability whenever they are defined, Sine and cosine are $1$ -Lipschitz on $R$ , $\exp (x + i y) = e^{x} (\cos y + i \sin y)$ , $∣ \exp (x + i y) ∣ = e^{x}$ , and $e^{i π} + 1 = 0$ )

[F8]

Integration against a signed or complex measure is the published $L^{1} (ν) = L^{1} (∣ ν ∣)$ integral and obeys $∣ \int f d ν ∣ \leq \int ∣ f ∣ d ∣ ν ∣$ . (Integration against a signed or complex measure, and the class L^1(nu) = L^1(|nu|), Integrals against signed or complex measures are bounded by total variation, A complex measure is a finite-valued countably additive set function)

Proof

technique · direct; the pairing identity is Fubini applied to the product $\mathbb R^n\times\mathbb R^n$ against $\lambda_n\times\mu$, the conjugation identity is the published transform law, and the convolution identity substitutes the translation law into the defining integral

1.1F1F3F4F7F8

Joint measurability and absolute integrability. The characters $e^{2 π i x \cdot ω}$ are Borel on $R^{2 n}$ by the Cartesian form, the $1$ -Lipschitz sine and cosine, and the closure rules of [F7]; the projections $(x, ω) \mapsto x$ , $(x, ω) \mapsto ω$ are Borel, so $G (ω)$ , $\overline{F (x)}$ and $\hat{F} (ω)$ pull back to Borel functions, and $p (x, ω) : = e^{2 π i x \cdot ω} G (ω) \overline{F (x)}$ , $q (z, ω) : = e^{2 π i z \cdot ω} F (x - z)$ for fixed $x$ are Borel by [F7]. For $p$ , using $∣ e^{2 π i x \cdot ω} ∣ = 1$ and $∣ G ∣ \leq ∥ G ∥_{\infty}$ , $∣ F ∣ \leq ∥ F ∥_{\infty}$ , with $∥ F ∥_{1} < \infty$ from [F4], $\int_{R^{n}} ⁣ ⁣ \int_{R^{n}} ∣ p ∣ d ∣ μ ∣ d x = ∥ F ∥_{1} ⁣ \int ∣ G ∣ d ∣ μ ∣ \leq ∥ F ∥_{1} ∥ G ∥_{\infty} M < \infty,$ and for $q$ , translation invariance gives $\int \int ∣ q ∣ d ∣ μ ∣ d z = ∥ F ∥_{1} M < \infty$ . Thus both kernels are integrable against $λ_{n} \times ∣ μ ∣$ . Fubini against the complex measure follows first for simple functions by linearity and then by approximation in this absolute-integral norm, using [F8]; this licenses the interchange below.

1.2F2

The clause (iii). By definition $\overset{ˇ}{μ} (x) = \hat{μ} (- x)$ , so $∣ \overset{ˇ}{μ} (x) ∣ = ∣ \hat{μ} (- x) ∣ \leq M$ for every $x$ by [F2], and $∣ \overset{ˇ}{μ} (x) - \overset{ˇ}{μ} (y) ∣ = ∣ \hat{μ} (- x) - \hat{μ} (- y) ∣$ tends to $0$ uniformly as $∣ x - y ∣ \to 0$ because $\hat{μ}$ is uniformly continuous [F2].

2.1F3F5step 1.1

The identity (i). By definition of $(G μ)^{\lor}$ and step 1.1, $\int (G μ)^{\lor} (x) \overline{F (x)} d x = \int \int p (x, ω) d μ (ω) d x$ , and Fubini [F3] rewrites this iterated integral as $\int [\int p (x, ω) d x] d μ (ω) = \int G (ω) [\int e^{2 π i x \cdot ω} \overline{F (x)} d x] d μ (ω)$ , because $G (ω)$ does not depend on $x$ . By the conjugation law of [F5] with $ξ = - ω$ , $\int e^{2 π i x \cdot ω} \overline{F (x)} d x = \hat{\overline{F}} (- ω) = \overline{\hat{F} (ω)}$ , so the last expression is $\int G (ω) \overline{\hat{F} (ω)} d μ (ω)$ . This proves (i).

2.2F3F5F6step 1.1

The convolution identity (ii). Fix $x \in R^{n}$ and put $F_{x} : = F (- \cdot)$ , so that $τ_{x} F_{x} (z) = F_{x} (z - x) = F (x - z)$ . By the translation and reflection laws of [F5], for every $ω$ , $\int F (x - z) e^{2 π i z \cdot ω} d z = \hat{τ_{x} F_{x}} (- ω) = e^{- 2 π i x \cdot (- ω)} \hat{F_{x}} (- ω) = e^{2 π i x \cdot ω} \hat{F} (ω) .$ Substituting this into the definition of $(\hat{F} μ)^{\lor}$ and applying Fubini [F3], which is licensed by step 1.1, gives $(\hat{F} μ)^{\lor} (x) = \int e^{2 π i x \cdot ω} \hat{F} (ω) d μ (ω) = \int \int q (z, ω) d z d μ (ω) = \int F (x - z) [\int e^{2 π i z \cdot ω} d μ (ω)] d z .$ The inner integral is $\overset{ˇ}{μ} (z)$ by its defining formula, so the last expression is $\int F (x - z) \overset{ˇ}{μ} (z) d z = (F * \overset{ˇ}{μ}) (x)$ by [F6]; this holds for every $x$ .

3.1F2F4F8step 2.2

Bounded continuity. By [F2], $∣ \overset{ˇ}{μ} (z) ∣ \leq M$ for every $z$ , so $∣ F * \overset{ˇ}{μ} (x) ∣ \leq ∥ F ∥_{1} M < \infty$ ; for the other side, $∣ (\hat{F} μ)^{\lor} (x) ∣ = ∣ \int e^{2 π i x \cdot ω} \hat{F} (ω) d μ (ω) ∣ \leq \int ∣ \hat{F} ∣ d ∣ μ ∣ \leq ∥ \hat{F} ∥_{\infty} M \leq ∥ F ∥_{1} M < \infty$ by the modulus bound of [F8] and [F4], and the equality of step 2.2 therefore holds between two bounded functions. For continuity of $F * \overset{ˇ}{μ}$ , let $ω_{\overset{ˇ}{μ}}$ be the modulus of uniform continuity of [F2]: for all $x, x^{'}$ , $∣ F * \overset{ˇ}{μ} (x) - F * \overset{ˇ}{μ} (x^{'}) ∣ \leq \int ∣ F (z) ∣ ∣ \overset{ˇ}{μ} (x - z) - \overset{ˇ}{μ} (x^{'} - z) ∣ d z \leq ∥ F ∥_{1} ω_{\overset{ˇ}{μ}} (∣ x - x^{'} ∣) ⟶ 0$ as $x^{'} \to x$ ; hence $F * \overset{ˇ}{μ}$ is continuous, and by step 2.2 so is $(\hat{F} μ)^{\lor}$ . Thus the identity of (ii) holds as everywhere-defined bounded continuous functions.

4.1step 2.1step 2.2step 3.1step 1.2∎

Conclusion. Step 2.1 proves (i); steps 2.2 and 3.1 prove (ii) as an identity of everywhere-defined bounded continuous functions; step 1.2 proves (iii). Countable Choice is spent only through the sigma-finite Fubini/Tonelli interfaces and the transform law of the cited suppliers, whose own hypotheses carry it.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

The unit sphere is Lebesgue null

Statement

Assume Countable Choice and let $n \geq 2$ . The unit sphere $S^{n - 1}$ satisfies $λ_{n} (S^{n - 1}) = 0$ , where $λ_{n}$ is $n$ -dimensional Lebesgue measure. Proof route: apply the polar-coordinate formula to the Borel indicator of $S^{n - 1}$ ; the inner integral vanishes unless $r = 1$ , so the double integral is zero because a single radius is a null set in $(0, \infty)$ .

Facts & Assumptions

Given: Countable Choice, $n \geq 2$ , the unit sphere $S^{n - 1} \subseteq R^{n}$ , the polar surface measure $σ$ of The polar surface set function on the unit sphere, and the Borel function $f = 1_{S^{n - 1}}$ .

[F1]

Polar coordinates: $σ$ is a finite Borel measure on $S^{n - 1}$ and for every Borel measurable $f : R^{n} \to [0, \infty]$ , $\int_{R^{n}} f d λ_{n} = \int_{0}^{\infty} \int_{S^{n - 1}} f (r ω) r^{n - 1} d σ (ω) d r$ . (Polar coordinates decompose Lebesgue measure into r^{n-1} dr d sigma, The polar surface set function on the unit sphere)

[F2]

The Euclidean norm is continuous, so $S^{n - 1} = ∣ \cdot ∣^{- 1} ({1})$ is a Borel subset of $R^{n}$ and $1_{S^{n - 1}}$ is a Borel, hence measurable, $[0, \infty]$ -valued function. (A continuous map has Borel preimages of Borel sets, The nonnegative Lebesgue integral)

[F3]

The integral of an indicator over a measurable set is the measure of that set: $\int_{E} 1_{A} d ν = ν (E \cap A)$ for measurable $E, A$ ; in particular the section integral of the indicator of a measurable set is a measure value. (Integral over a measurable subset, The nonnegative Lebesgue integral)

[F4]

A singleton ${r_{0}} \subseteq R$ is Lebesgue null, and every at most countable subset of $R^{n}$ is Lebesgue null. (Every at most countable subset of $R^{n}$ is Lebesgue null; in particular $λ_{1} (Q) = 0$ )

[F5]

A nonnegative measurable function has integral zero exactly when it vanishes almost everywhere. (A nonnegative measurable function has integral $0$ exactly when it vanishes almost everywhere)

[F6]

The iterated integral in [F1] is a Tonelli integral over the sigma-finite product $(0, \infty) \times S^{n - 1}$ ; in particular the inner integral is a measurable function of $r$ . (Tonelli's theorem for nonnegative measurable functions on a sigma-finite product)

Proof

technique · direct; evaluate the polar-coordinate formula at the indicator of the sphere and observe that the inner integral is supported on the single null radius $r=1$

1.1F1F3givenalgebra

The section integral. Fix $r > 0$ . For every $ω \in S^{n - 1}$ one has $∣ r ω ∣ = r$ , so $r ω \in S^{n - 1}$ if and only if $r = 1$ . Hence $1_{S^{n - 1}} (r ω) = 1_{{1}} (r)$ for every $ω$ , and by [F3] and [F1], $\int_{S^{n - 1}} 1_{S^{n - 1}} (r ω) d σ (ω) = \int_{S^{n - 1}} 1_{{1}} (r) d σ (ω) = 1_{{1}} (r) σ (S^{n - 1}) .$ The factor $σ (S^{n - 1})$ is finite by [F1].

2.1F1F2F6step 1.1

The polar integral. The function $f = 1_{S^{n - 1}}$ is Borel and nonnegative by [F2], so the polar-coordinate formula [F1] applies and, with the section computation of step 1.1, $λ_{n} (S^{n - 1}) = \int_{R^{n}} 1_{S^{n - 1}} d λ_{n} = \int_{0}^{\infty} [\int_{S^{n - 1}} 1_{S^{n - 1}} (r ω) d σ (ω)] r^{n - 1} d r = σ (S^{n - 1}) \int_{0}^{\infty} 1_{{1}} (r) r^{n - 1} d r .$

3.1F4F5step 2.1algebra

The radial integral vanishes. The function $r \mapsto 1_{{1}} (r) r^{n - 1}$ is nonnegative, measurable and vanishes for every $r \neq 1$ ; the singleton ${1}$ is Lebesgue null in $(0, \infty)$ by [F4], so the function vanishes almost everywhere. By [F5] its integral over $(0, \infty)$ is $0$ , and since $σ (S^{n - 1}) < \infty$ the right-hand side of step 2.1 is $σ (S^{n - 1}) \cdot 0 = 0$ . Therefore $λ_{n} (S^{n - 1}) = 0$ .

4.1F1F6step 1.1step 2.1step 3.1∎

Conclusion. Steps 1.1–3.1 evaluate the polar-coordinate formula at $1_{S^{n - 1}}$ and prove $λ_{n} (S^{n - 1}) = 0$ for every $n \geq 2$ ; Countable Choice is inherited exactly from the polar-coordinate, sigma-finite Tonelli, null-set and integral-interface suppliers.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Sphere graph charts, surface density, and a finite partition

Statement

Assume Countable Choice and let $n \geq 2$ . For each $i \in {1, \dots, n}$ and each sign $ε \in {\pm 1}$ the map $X_{i}^{ε} (y) = (y_{1}, \dots, y_{i - 1}, ε \sqrt{1 - ∣ y ∣^{2}}, y_{i}, \dots, y_{n - 1})$ , defined on the open unit ball $B \subseteq R^{n - 1}$ , is a $C^{\infty}$ graph chart onto the hemisphere ${ε x_{i} > 0}$ , and the $2 n$ hemispheres cover $S^{n - 1}$ . On such a chart the polar surface measure $σ$ has Lebesgue density $\sqrt{1 + ∣ \nabla h ∣^{2}} = (1 - ∣ y ∣^{2})^{- 1 / 2}$ , and the graphing function $h = ε \sqrt{1 - ∣ y ∣^{2}}$ satisfies $\det D^{2} h = (- ε)^{n - 1} (1 - ∣ y ∣^{2})^{- (n + 1) / 2} \neq 0$ on all of $B$ . There exist finitely many nonnegative $C^{\infty}$ functions $χ_{1}, \dots, χ_{m}$ on $S^{n - 1}$ with $\sum_{j} χ_{j} = 1$ , each compactly supported in the image of one of these charts.

Facts & Assumptions

Given: Countable Choice, $n \geq 2$ , the open unit ball $B \subseteq R^{n - 1}$ , and for each $i, ε$ the map $X_{i}^{ε}$ and graphing function $h_{i}^{ε} = ε \sqrt{1 - ∣ y ∣^{2}}$ .

[F1]

The polar surface set function $σ$ on $S^{n - 1}$ is the published Borel surface measure, and the chart surface measure on $S^{n - 1}$ equals $σ$ ; the chart measure of a compact hypersurface is computed by chart densities and is independent of the charts. (The polar surface set function on the unit sphere, Polar coordinates decompose Lebesgue measure into r^{n-1} dr d sigma, Agreement with the existing polar sphere measure, Surface integration on compact C1 hypersurfaces, Chart and partition independence of surface measure)

[F2]

In graph coordinates $X (y) = (y, h (y))$ the chart density is $\sqrt{1 + ∣ D h (y) ∣^{2}}$ ; more precisely the Gram determinant of the tangent columns is $\det D X^{T} D X = 1 + ∣ D h (y) ∣^{2}$ . (Chart and partition independence of surface measure)

[F3]

The one-variable sum, product and quotient derivative rules and the one-variable chain rule hold on open intervals. Applying them on coordinate lines computes partial derivatives; smoothness means continuity of all ordered iterated partials. The gradient and Jacobian conventions are the published ones. (Sums, scalar multiples, products and quotients: $(f + g)^{'} (c) = f^{'} (c) + g^{'} (c)$ , $(α f)^{'} (c) = α f^{'} (c)$ , $(f g)^{'} (c) = f^{'} (c) g (c) + f (c) g^{'} (c)$ , and $(f / g)^{'} (c) = (f^{'} (c) g (c) - f (c) g^{'} (c)) / g (c)^{2}$ when $g (c) \neq 0$ , The chain rule, in one line from Carathéodory: if $g$ is differentiable at $c$ and $f$ is differentiable at $g (c)$ , then $f \circ g$ is differentiable at $c$ with $(f \circ g)^{'} (c) = f^{'} (g (c)) g^{'} (c)$ , $C^{k}$ maps and multi-index derivative notation in Euclidean space, The Jacobian matrix of partial derivatives and the gradient in the scalar-valued case)

[F4]

Compactness: a subset of $R^{n}$ is compact if and only if it is closed and bounded; $S^{n - 1}$ is closed and bounded, hence compact. (For a nonempty subset of $R^{n}$ with $n \geq 1$ , compactness, closedness and boundedness, pseudocompactness, and attainment of extrema by every continuous real-valued function are equivalent)

[F5]

Every countable cover of a smooth manifold by coordinate balls admits a smooth partition of unity subordinate to that cover; subordinate means locally finite supports inside the corresponding cover members summing to one. (Smooth partitions subordinate to a countable coordinate cover, Smooth partitions of unity subordinate to an open cover)

[F6]

Embedded submanifolds and smooth manifolds: a subset that is locally the graph of a smooth function is an embedded submanifold, and compatible charts with smooth transitions define the smooth structure. (Embedded submanifolds and slice charts, Smooth manifolds and their smooth charts)

[F7]

Determinant expansion: the determinant is multilinear and alternating in the columns and in the rows, a matrix with two proportional columns has determinant zero, and transposition preserves the determinant. (The determinant is alternating and multilinear in the rows as well as in the columns, A square matrix with a zero column or two equal columns has determinant zero, For every square matrix over a commutative ring, $\det (A^{T}) = \det (A)$ )

Proof

technique · direct; identify each hemisphere with a graph, compute the density and the Hessian determinant, then subordinate a finite smooth partition to the hemisphere cover of the compact sphere

1.1F3givenalgebra

Calculus for the graph expressions. For $t, s > 0$ , $∣ \sqrt{t} - \sqrt{s} ∣ = ∣ t - s ∣ / (\sqrt{t} + \sqrt{s}) \leq ∣ t - s ∣ / \sqrt{s}$ , proving continuity at $s$ . Rationalizing the difference quotient gives $(\sqrt{\cdot})^{'} (s) = 1 / (2 \sqrt{s})$ . By the product and quotient rules in [F3], induction shows that every derivative of $\sqrt{t}$ is a constant times an integer power of $\sqrt{t}$ , hence exists and is continuous for $t > 0$ . Put $q (y) = 1 - \sum_{j} y_{j}^{2}$ . Applying the chain rule on each coordinate line gives $\partial_{j} \sqrt{q (y)} = - y_{j} / \sqrt{q (y)}$ . Repeated coordinate product and quotient rules show that every ordered partial of $\sqrt{q}$ and of its reciprocal is a finite sum of polynomial numerators divided by positive integer powers of $\sqrt{q}$ . All these expressions are continuous on $B$ , where $q > 0$ ; thus the graph functions and density are smooth in the sense of [F3].

1.2F7algebra

A rank-one determinant. For every $t \geq 0$ and $v \in R^{m}$ with $m \geq 1$ , $\det (I + t v v^{T}) = 1 + t ∣ v ∣^{2}$ . Indeed the $j$ -th column of $I + t v v^{T}$ is $e_{j} + t v_{j} v$ , so multilinearity in the columns [F7] expands the determinant over subsets $S$ of the column indices: the term for $S$ is $(\prod_{j \in S} t v_{j}) \det (c^{(S)})$ , where $c^{(S)}$ has column $v$ in the slots $S$ and $e_{j}$ elsewhere. If $∣ S ∣ \geq 2$ two columns are equal to $v$ , so the determinant vanishes [F7]; the term $S = \emptyset$ is $\det I = 1$ ; and for $S = {j}$ the matrix has determinant $v_{j}$ (expanding along the standard basis columns), giving $t v_{j}^{2}$ . Summing gives $1 + t \sum_{j} v_{j}^{2} = 1 + t ∣ v ∣^{2}$ .

1.3F1F2F3algebra

The density. Here $h (y) = ε \sqrt{1 - ∣ y ∣^{2}}$ , so by [F3] $\nabla h (y) = ε \cdot \frac{- y}{\sqrt{1 - ∣ y ∣^{2}}} = - ε (1 - ∣ y ∣^{2})^{- 1 / 2} y, ∣ \nabla h (y) ∣^{2} = \frac{∣ y ∣^{2}}{1 - ∣ y ∣^{2}},$ hence $\sqrt{1 + ∣ \nabla h (y) ∣^{2}} = (1 - ∣ y ∣^{2})^{- 1 / 2}$ on $B$ . By [F2] this is the chart density of the graph, and by [F1] the chart measure is the polar surface measure $σ$ ; the density is finite and strictly positive on $B$ because $1 - ∣ y ∣^{2} > 0$ there.

2.1F3step 1.1givenalgebra

The charts and the cover. Fix $i$ and $ε$ and put $π_{i} (x) = (x_{1}, \dots, \hat{x_{i}}, \dots, x_{n})$ for the coordinate projection. On the hemisphere $H_{i}^{ε} = {x \in S^{n - 1} : ε x_{i} > 0}$ the map $π_{i}$ is a left inverse of $X_{i}^{ε}$ : $π_{i} (X_{i}^{ε} (y)) = y$ ; conversely $X_{i}^{ε} (π_{i} (x)) = x$ for $x \in H_{i}^{ε}$ , because the removed coordinate is recovered by $x_{i} = ε \sqrt{1 - ∣ π_{i} (x) ∣^{2}}$ , which is exactly the defining equation of the sphere with the sign $ε$ . The coordinates of $X_{i}^{ε}$ are smooth by step 1.1, and its inverse is coordinate projection, since $1 - ∣ y ∣^{2} > 0$ on $B$ and the coordinates of a unit vector satisfy $∣ x_{i} ∣ \leq 1$ ; hence $X_{i}^{ε}$ is a $C^{\infty}$ chart of $S^{n - 1}$ onto $H_{i}^{ε}$ . Every $x \in S^{n - 1}$ satisfies $\sum_{k} x_{k}^{2} = 1$ , so some coordinate is nonzero; then $x \in H_{i}^{sgn (x_{i})}$ for that $i$ , and the $2 n$ hemispheres cover $S^{n - 1}$ .

2.2F3step 1.2step 1.3algebra

The Hessian determinant. Differentiating the gradient of step 1.3 with [F3] gives $\partial_{i} \partial_{j} h (y) = - ε [(1 - ∣ y ∣^{2})^{- 1 / 2} δ_{i j} + (1 - ∣ y ∣^{2})^{- 3 / 2} y_{i} y_{j}],$ that is $D^{2} h (y) = - ε a [I + a^{2} y y^{T}]$ with $a : = (1 - ∣ y ∣^{2})^{- 1 / 2} > 0$ . Taking determinants and using step 1.2 with $t = a^{2}$ , $v = y$ , $\det D^{2} h (y) = (- ε)^{n - 1} a^{n - 1} (1 + a^{2} ∣ y ∣^{2}) = (- ε)^{n - 1} a^{n - 1} \cdot \frac{1}{1 - ∣ y ∣^{2}} = (- ε)^{n - 1} (1 - ∣ y ∣^{2})^{- (n + 1) / 2},$ because $1 + a^{2} ∣ y ∣^{2} = 1 + ∣ y ∣^{2} / (1 - ∣ y ∣^{2}) = 1 / (1 - ∣ y ∣^{2})$ and $a^{n - 1} a^{2} = a^{n + 1}$ . The value is nonzero for every $y \in B$ since $1 - ∣ y ∣^{2} > 0$ .

3.1F3F4F6step 2.1

The smooth structure and compactness. Each $X_{i}^{ε}$ is a bijection of the open ball onto its image with smooth inverse and smooth transitions: on overlaps, the transition is $π_{i^{'}} \circ X_{i}^{ε}$ , a coordinate selection of a smooth graph map, smooth by [F3]. The ambient coordinate map that deletes $x_{i}$ and appends $x_{i} - h_{i}^{ε} (π_{i} (x))$ has smooth inverse obtained by reinserting $h_{i}^{ε} (y) + z$ in the $i$ th slot; near each point of the hemisphere it carries the sphere to $z = 0$ . These are slice charts in [F6], so the sphere is an embedded smooth manifold with the displayed graph charts. Being the zero set of the continuous function $x \mapsto ∣ x ∣^{2} - 1$ , the sphere is closed; it is bounded by $∣ x ∣ = 1$ , so it is compact by [F4].

4.1F5step 2.1step 3.1

The finite partition. The $2 n$ hemispheres $H_{i}^{ε}$ are coordinate balls of the smooth manifold $S^{n - 1}$ of step 3.1 and form a countable (indeed finite) open cover. By [F5] this cover admits a smooth partition of unity $(χ_{i}^{ε})_{i, ε}$ subordinate to it: the supports are locally finite, $supp χ_{i}^{ε} \subseteq H_{i}^{ε}$ , and $\sum_{i, ε} χ_{i}^{ε} = 1$ with every $χ_{i}^{ε} \geq 0$ . Since $S^{n - 1}$ is compact by step 3.1, every support, being closed in $S^{n - 1}$ , is compact; enumerating the $2 n$ pairs $(i, ε)$ as $1, \dots, m$ with $m = 2 n$ gives the asserted functions, each compactly supported inside the image $H_{i}^{ε}$ of the chart $X_{i}^{ε}$ .

5.1step 2.1step 3.1step 1.3step 2.2step 4.1∎

Conclusion. Step 2.1 gives the $2 n$ smooth graph charts and their hemisphere cover, step 1.3 computes the density $(1 - ∣ y ∣^{2})^{- 1 / 2}$ , step 2.2 computes $\det D^{2} h = (- ε)^{n - 1} (1 - ∣ y ∣^{2})^{- (n + 1) / 2} \neq 0$ , and step 4.1 produces the finite smooth partition with compactly supported pieces. Each clause holds for every $n \geq 2$ , with the case $n = 2$ covered by the same computation ( $m = n - 1 = 1$ and the determinant is $- ε (1 - ∣ y ∣^{2})^{- 3 / 2}$ ).

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Van der Corput oscillatory integral estimates in one dimension

Statement

Let $k \geq 1$ . (a) (First-derivative version) If $φ \in C^{1} (R)$ is real with $∣ φ^{'} ∣ \geq λ_{1} > 0$ and $φ^{'}$ monotone on a bounded interval $J$ , then $∣ \int_{J} e^{2 π i λ φ (x)} d x ∣ \leq 2 (π λ λ_{1})^{- 1}$ for every $λ > 0$ , uniformly in the length of $J$ ; if moreover $a \in C_{c}^{1} (R)$ is complex-valued and $∣ φ^{'} ∣ \geq λ_{1}$ with $φ^{'}$ monotone on $supp a$ , then $∣ \int e^{2 π i λ φ (x)} a (x) d x ∣ \leq (π λ λ_{1})^{- 1} ∥ a^{'} ∥_{L^{1}}$ for every $λ > 0$ . (b) (Higher-derivative version) If $a \in C_{c}^{1} (R)$ is complex-valued and $φ \in C^{k} (R)$ is real with $∣ φ^{(k)} ∣ \geq 1$ on $supp a$ for some $k \geq 2$ , then $∣ \int e^{2 π i λ φ (x)} a (x) d x ∣ \leq C λ^{- 1 / k}$ for $λ \geq 1$ , where $C$ depends only on $k$ , on $a$ and on finitely many derivative bounds for $φ$ near $supp a$ ; for the quadratic phase $φ (x) = \pm x^{2} / 2$ this gives the Fresnel bound $∣ \int e^{\pm π i λ x^{2}} a (x) d x ∣ \leq C_{0} (∥ a ∥_{\infty} + ∥ a^{'} ∥_{L^{1}}) λ^{- 1 / 2}$ with an absolute constant $C_{0}$ .

Facts & Assumptions

Given: $k \geq 1$ , a real phase $φ \in C^{k}$ , and, where an amplitude occurs, a complex $a \in C_{c}^{1} (R)$ .

[F1]

Riemann–Stieltjes integration by parts and the $C^{1}$ -integrator reduction: if $f$ has bounded variation and $α$ is continuous, $\int f d α$ exists; when $α = F$ is $C^{1}$ its derivative is continuous and $\int_{u}^{v} f d F = \int_{u}^{v} f F^{'}$ ; and $\int_{u}^{v} f d F + \int_{u}^{v} F d f = f (v) F (v) - f (u) F (u)$ . Integrals of step functions against $i d$ agree with ordinary integrals, and the Stieltjes integral is linear in the integrand. (Riemann–Stieltjes integration by parts, A bounded-variation integrand is Riemann–Stieltjes integrable against every continuous integrator, The total-variation bound for a Riemann–Stieltjes integral, A continuously differentiable integrator reduces Stieltjes integration to ordinary integration, The identity integrator recovers the Riemann integral, Linearity and interval additivity of the Riemann–Stieltjes integral)

[F2]

A continuous monotone real function has variation equal to the absolute difference of its endpoint values. If it has constant sign and modulus at most $M$ , that variation is at most $M$ . For a complex $C^{1}$ function $a$ , the fundamental theorem gives $∣ a (v) - a (u) ∣ \leq \int_{u}^{v} ∣ a^{'} ∣$ and hence $Var (a) \leq \int ∣ a^{'} ∣$ . Stieltjes identities apply componentwise. (Bounded variation and total variation on an interval, The total-variation bound for a Riemann–Stieltjes integral, Newton–Leibniz remains valid across finitely many exceptional interior points when the primitive is continuous)

[F3]

Smooth calculus: products, quotients with nonvanishing denominators, compositions and higher derivatives are computed by the algebra and chain rules; $φ^{'}$ monotone and nonzero makes $1 / φ^{'}$ monotone and nonzero, and $φ \in C^{k}$ gives continuity of $φ^{(k)}$ on compacta. (Sums, scalar multiples, products and quotients: $(f + g)^{'} (c) = f^{'} (c) + g^{'} (c)$ , $(α f)^{'} (c) = α f^{'} (c)$ , $(f g)^{'} (c) = f^{'} (c) g (c) + f (c) g^{'} (c)$ , and $(f / g)^{'} (c) = (f^{'} (c) g (c) - f (c) g^{'} (c)) / g (c)^{2}$ when $g (c) \neq 0$ , The chain rule for total derivatives: $D (g \circ f) (a) = D g (f (a)) \circ D f (a)$ , $C^{k}$ maps and multi-index derivative notation in Euclidean space)

Proof

technique · direct; prove uniform pure-phase estimates on intervals, then transfer them to amplitudes by integration against a bounded primitive on each component of the nonzero set

1.1F1F2F3givenalgebra

First derivative on an interval. Write $F = e^{2 π i λ φ}$ and $g = 1 / φ^{'}$ on $[u, v] \subset J$ . The continuous derivative has constant sign; $g$ is monotone of that sign and $∣ g ∣ \leq λ_{1}^{- 1}$ . Stieltjes integration by parts gives $\int_{u}^{v} F = (2 π i λ)^{- 1} (g (v) F (v) - g (u) F (u) - \int_{u}^{v} F d g)$ . Its modulus is at most $(2 π λ)^{- 1} (∣ g (v) ∣ + ∣ g (u) ∣ + ∣ g (v) - g (u) ∣) \leq (π λ λ_{1})^{- 1}$ . This stronger bound implies the displayed pure-phase estimate in (a), and also bounds the primitive on every subinterval. Endpoint inclusion does not change the integral.

2.1F1F2F3step 1.1

First-derivative amplitude bound. On each component $(u, v)$ of ${a \neq 0}$ , $a$ vanishes at the endpoints and the hypotheses on $supp a$ give the estimate of step 1.1 on every subinterval. Thus $P (x) = \int_{u}^{x} F$ has modulus at most $(π λ λ_{1})^{- 1}$ , and ordinary integration by parts gives $\int_{u}^{v} F a = - \int_{u}^{v} P a^{'}$ . Summing gives $∣ \int F a ∣ \leq (π λ λ_{1})^{- 1} ∥ a^{'} ∥_{1}$ . The components are canonically at most countable by assigning to each its first rational in a fixed enumeration; the sum is justified by $\int ∣ a ∣ < \infty$ and the sum of $\int_{u}^{v} ∣ a^{'} ∣$ being at most $∥ a^{'} ∥_{1}$ . No reciprocal of $φ^{'}$ is used in gaps outside the support.

2.2F1F3step 1.1algebra

Higher-derivative interval estimate. Suppose $∣ φ^{(k)} ∣ \geq γ > 0$ throughout a bounded interval $J$ , $k \geq 2$ . We prove that every subinterval has pure-phase integral bounded by $C_{k} (λ γ)^{- 1 / k}$ , independently of its length. The derivative $φ^{(k)}$ has constant sign, so $φ^{(k - 1)}$ is monotone. For $τ > 0$ , the set $∣ φ^{(k - 1)} ∣ \leq τ$ is an interval of length at most $2 τ / γ$ , by the fundamental theorem. Its complement has at most two intervals on which $∣ φ^{(k - 1)} ∣ \geq τ$ . For $k = 2$ , step 1.1 applies there because $φ^{'}$ is monotone. For $k > 2$ , use induction with lower bound $τ$ . The resulting bound is $2 τ / γ + 2 C_{k - 1} (λ τ)^{- 1 / (k - 1)}$ . Set $τ = γ^{(k - 1) / k} λ^{- 1 / k}$ to obtain the asserted bound. The same reasoning on any subinterval proves the primitive bound required below.

3.1F1F2F3step 2.1step 2.2

Higher-derivative amplitude bound. On every component $(u, v)$ of ${a \neq 0}$ , the hypothesis $∣ φ^{(k)} ∣ \geq 1$ holds throughout that interval. Step 2.2 with $γ = 1$ gives a primitive $P (x) = \int_{u}^{x} F$ bounded by $C_{k} λ^{- 1 / k}$ . Since $a (u) = a (v) = 0$ , integration by parts yields $∣ \int_{u}^{v} F a ∣ \leq C_{k} λ^{- 1 / k} \int_{u}^{v} ∣ a^{'} ∣$ . Sum over the canonically countable components as in step 2.1 to obtain $∣ \int F a ∣ \leq C_{k} λ^{- 1 / k} ∥ a^{'} ∥_{1}$ . This stronger estimate implies (b) with the stated constant dependence, even for disconnected support; no lower derivative bound in its gaps is assumed.

4.1F1F2step 2.2∎

Quadratic phase. For $φ = \pm x^{2} / 2$ , the second derivative has modulus one on every interval. Apply step 2.2 directly on one interval containing $supp a$ and integrate against its bounded primitive. This gives $∣ \int e^{\pm π i λ x^{2}} a ∣ \leq C_{0} λ^{- 1 / 2} ∥ a^{'} ∥_{1}$ , which implies the stated Fresnel estimate with $∥ a ∥_{\infty} + ∥ a^{'} ∥_{1}$ . No scale-dependent cutoff derivative enters.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Shape operator and Gauss-Kronecker curvature of a graph

Statement

Let $n \geq 2$ , let $U \subseteq R^{n - 1}$ be open, let $h \in C^{\infty} (U; R)$ , and let $S = {(y, h (y)) : y \in U} \subseteq R^{n}$ be the graph with the unit normal $ν (y) = (- \nabla h (y), 1) / \sqrt{1 + ∣ \nabla h (y) ∣^{2}}$ of positive last coordinate. At $p = (y, h (y))$ the shape operator $S_{ν}$ of $S$ with respect to $ν$ satisfies $\det S_{ν} = \det D^{2} h (y) / (1 + ∣ \nabla h (y) ∣^{2})^{(n + 1) / 2}$ . In particular the extrinsic Gaussian (Gauss-Kronecker) curvature of $S$ vanishes at $p$ if and only if $\det D^{2} h (y) = 0$ .

Facts & Assumptions

Given: $h \in C^{\infty} (U)$ , $X (y) = (y, h (y))$ , $b = \sqrt{1 + ∣ \nabla h ∣^{2}}$ , and $ν = (- \nabla h, 1) / b$ .

[F1]

The local Euclidean shape operator is $S_{ν} = - d ν$ , and it agrees with the usual hypersurface operator; curvature is its determinant. (Euclidean hypersurface normals, shape operators and curvature, Smooth Euclidean hypersurface graphs and compact localization)

[F2]

Determinants are multiplicative and invariant under change of basis. (For same-sized finite square matrices over a commutative ring, $\det (A B) = \det (A) \det (B)$ , The determinant of an endomorphism of a finite-dimensional vector space: its matrix determinant in an ordered basis in positive dimension, and $1$ on the zero space)

[F3]

Mixed second partials agree. (Continuous mixed partials of order $k$ are invariant under permutations)

[F4]

The graph Gram determinant and surface density are $1 + ∣ \nabla h ∣^{2}$ and its square root. (Chart and partition independence of surface measure)

Proof

technique · direct; differentiate orthogonality and compute the determinant in the graph frame

1.1givenF1F3algebra

The columns $X_{j} = (e_{j}, h_{j})$ are independent, and their Gram matrix is $G = I + \nabla h \nabla h^{T}$ . The vector $ν$ is unit and perpendicular to each column. Hence $G$ is positive definite and this is the positive-last-coordinate normal. Differentiating $ν \cdot X_{k} = 0$ gives $⟨ S_{ν} X_{j}, X_{k} ⟩ = - \partial_{j} ν \cdot X_{k} = ν \cdot X_{j k} = h_{j k} / b$ . In particular the normal component of $X_{j k} = h_{j k} e_{n}$ is $(h_{j k} / b) ν$ ; the vertical vector $e_{n}$ itself need not be normal.

2.1F1F2F4step 1.1algebra∎

If $A$ is the matrix of $S_{ν}$ in the frame $X_{j}$ , the pairing in step 1.1 says $G A = D^{2} h / b$ . Consequently $A = G^{- 1} D^{2} h / b$ , and multiplicativity gives $\det S_{ν} = \det D^{2} h / (b^{n - 1} \det G) = \det D^{2} h / (1 + ∣ \nabla h ∣^{2})^{(n + 1) / 2}$ . The denominator is positive, so curvature vanishes exactly when the Hessian determinant vanishes. The Euclidean equivalence in [F1] identifies this determinant with the promised extrinsic Gaussian curvature.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Restriction and extension estimates are dual

Statement

Assume Countable Choice and let $1 < p < \infty$ with conjugate exponent $p^{'}$ . For a compact hypersurface $S$ with surface measure $σ$ and the operators $R_{0}, E$ of Fourier restriction and adjoint extension operators the following are equivalent: (a) there is $C < \infty$ with $∥ \hat{f} ∥_{L^{2} (σ)} \leq C ∥ f ∥_{L^{p}}$ for all $f \in S (R^{n})$ , so that $R_{0}$ has a unique bounded extension $R : L^{p} (R^{n}) \to L^{2} (σ)$ ; (b) there is $C < \infty$ with $∥ E g ∥_{L^{p^{'}} (R^{n})} \leq C ∥ g ∥_{L^{2} (σ)}$ for all $g \in L^{2} (σ)$ . The least constants agree, and $E$ is the adjoint of $R$ under the $L^{p}$ – $L^{p'}$ and $L^{2} (σ)$ pairings: $\int E g \overline{f} d x = \int_{S} g \overline{R f} d σ$ .

Facts & Assumptions

Given: Countable Choice, $1 < p < \infty$ with conjugate $p^{'}$ , a compact hypersurface $S$ with surface measure $σ$ (finite), and the operators $R_{0} : S (R^{n}) \to L^{2} (σ)$ , $f \mapsto \hat{f} ∣_{S}$ , and $E : L^{1} (σ) \to C_{b} (R^{n})$ , $g \mapsto (g σ)^{\lor}$ , of Fourier restriction and adjoint extension operators.

[F1]

The operator $E$ is defined on $L^{1} (σ) \supseteq L^{2} (σ)$ by an everywhere-defined bounded uniformly continuous function, the surface measure is finite, and $R_{0}$ is pointwise defined on Schwartz data; $L^{p} (σ; C)$ and $L^{p} (R^{n}; C)$ are the complex Lebesgue classes. (Fourier restriction and adjoint extension operators)

[F2]

Pairing identity: for every finite complex Borel measure $μ$ and all $F, G \in S (R^{n})$ , $\int (G μ)^{\lor} \overline{F} d x = \int G \overline{\hat{F}} d μ$ . (Fourier pairing for a finite measure and Schwartz data)

[F4]

$C_{c}^{\infty} (R^{n}) \subset S$ is dense in $L^{p} (R^{n})$ for $1 \leq p < \infty$ ; smooth ball cutoffs exist. ( $C_{c}^{\infty} (R^{n})$ is dense in $L^{p} (R^{n})$ for $1 \leq p < \infty$ , Explicit compactly supported smooth cutoffs)

[F5]

Complex $L^{p}$ spaces are complete under Countable Choice; a bounded map from a dense subspace to a Banach space extends uniquely with the same norm. Cauchy–Schwarz and complex Hölder hold, including endpoints. (Complex Lp completeness and almost-everywhere subsequences, A bounded linear map from a dense normed subspace into a Banach space extends uniquely with the same norm, The complex $L^{2}$ pairing is well-defined and satisfies Cauchy–Schwarz, Complex Holder, Minkowski, and the quotient norm)

[F6]

Absolutely integrable product kernels admit Fubini, and nonnegative integrals obey monotone convergence. (Fubini's theorem for L^1 functions on a sigma-finite product, Monotone convergence for the integral)

Proof

technique · direct; prove the pairing by Fubini and obtain extension integrability by compactly supported norm tests

1.1F1F2F6

For $g \in L^{2} (σ) \subset L^{1} (σ)$ and $F \in S$ , the kernel $e^{2 π i x \cdot ω} g (ω) \overline{F (x)}$ has absolute integral $∥ g ∥_{1} ∥ F ∥_{1} < \infty$ . Fubini gives $\int E g \overline{F} d x = \int_{S} g \overline{\hat{F}} d σ$ . This is an absolutely convergent integral pairing, without asserting $E g \in L^{2} (R^{n})$ .

2.1F1F4F5F6step 1.1algebra

Assume (a), fix $g$ , and set $h = E g$ , $M = C ∥ g ∥_{2}$ . Step 1.1 and Cauchy–Schwarz give $∣ \int h \overline{F} ∣ \leq M ∥ F ∥_{p}$ on Schwartz functions. For a bounded ball $B$ , put $v = 1_{B} h ∣ h ∣^{p^{'} - 2}$ , assigning zero where $h = 0$ . Since $h$ is bounded, $v \in L^{p}$ and $A = \int_{B} ∣ h ∣^{p^{'}} < \infty$ . Approximate $v$ in $L^{p}$ by $C_{c}^{\infty}$ functions and multiply by a fixed smooth cutoff equal to one on $B$ , supported in a larger bounded ball. These approximants still converge to $v$ in $L^{p}$ , and their integrals against $h$ converge, because $h$ is bounded and their supports have uniformly finite measure. Passing to the limit gives $A \leq M A^{1 / p}$ . Thus $A^{1 / p^{'}} \leq M$ (also when $A = 0$ ). Letting $B$ increase to $R^{n}$ and using monotone convergence proves $h \in L^{p^{'}}$ and $∥ h ∥_{p^{'}} \leq M$ , establishing (b).

2.2F1F5step 1.1algebra

Assume (b). For Schwartz $F$ , step 1.1 and Hölder give $∣ \int_{S} g \overline{R_{0} F} d σ ∣ \leq C ∥ g ∥_{2} ∥ F ∥_{p}$ for every $g \in L^{2} (σ)$ . If $R_{0} F \neq 0$ , take $g = R_{0} F / ∥ R_{0} F ∥_{2}$ ; otherwise the desired estimate is immediate. Hence $∥ R_{0} F ∥_{2} \leq C ∥ F ∥_{p}$ , proving (a). No density assertion on surface functions is needed.

3.1F4F5step 1.1step 2.1step 2.2∎

By [F4] and [F5], (a) gives the unique extension $R : L^{p} \to L^{2} (σ)$ . For each $g$ , the pairing in step 1.1 extends by continuity in $f$ from Schwartz data to every $f \in L^{p}$ , since $E g \in L^{p^{'}}$ by step 2.1. This identifies $E$ as the adjoint under the displayed Banach dual pairings. Steps 2.1 and 2.2 preserve each admissible constant, so the least constants, and the operator norms, agree. Countable Choice is the hypothesis of the density and completeness suppliers.

LemmaStatement: Literature-sourcedProof: Literature-sourcedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Stationary phase with a compactly supported amplitude

Statement

Let $d \geq 1$ , $a \in C_{c}^{\infty} (R^{d})$ , $φ \in C^{\infty} (R^{d}; R)$ and $I (λ) = \int e^{2 π i λ φ (x)} a (x) d x$ . (a) If $\nabla φ \neq 0$ on a neighbourhood of $supp a$ , then $∣ I (λ) ∣ \leq C_{N} λ^{- N}$ for every integer $N \geq 0$ and $λ > 0$ . (b) If $φ$ has exactly one stationary point $x_{0}$ , in the interior of $supp a$ , with $D^{2} φ (x_{0})$ invertible, then $∣ I (λ) ∣ \leq C λ^{- d / 2}$ for $λ \geq 1$ , and more precisely $∣ \partial_{λ}^{k} (e^{- 2 π i λ φ (x_{0})} I (λ)) ∣ \leq C_{k} λ^{- d / 2 - k}$ for every $k \geq 0$ . The constants depend only on $d$ , the diameter of the amplitude support, the chosen localization radius, finitely many derivatives of $a$ and $φ$ , on a positive lower bound for $∣ \det D^{2} φ (x_{0}) ∣$ , and on a positive lower bound for $∣ \nabla φ ∣$ on the amplitude support off a small ball about $x_{0}$ .

Facts & Assumptions

Given: $d \geq 1$ , a real phase $φ \in C^{\infty} (R^{d}; R)$ , an amplitude $a \in C_{c}^{\infty} (R^{d}; C)$ , $I (λ) = \int e^{2 π i λ φ (x)} a (x) d x$ .

[F1]

Divergence and integration by parts: for a compactly supported smooth vector field $v$ one has $\int_{R^{d}} \nabla \cdot v d x = 0$ (integrate the last coordinate first and apply the one-dimensional fundamental theorem across the compact support, using Fubini); hence for smooth $h$ and any $j$ , $\int (\partial_{j} ψ) h d x = - \int ψ \partial_{j} h d x$ . (Fubini's theorem for L^1 functions on a sigma-finite product, Newton–Leibniz needs only continuity on $[a, b]$ , differentiability on $(a, b)$ , and a Riemann-integrable extension of the interior derivative, Newton–Leibniz remains valid across finitely many exceptional interior points when the primitive is continuous, Sums, scalar multiples, products and quotients: $(f + g)^{'} (c) = f^{'} (c) + g^{'} (c)$ , $(α f)^{'} (c) = α f^{'} (c)$ , $(f g)^{'} (c) = f^{'} (c) g (c) + f (c) g^{'} (c)$ , and $(f / g)^{'} (c) = (f^{'} (c) g (c) - f (c) g^{'} (c)) / g (c)^{2}$ when $g (c) \neq 0$ )

[F2]

Taylor with Lagrange remainder near the stationary point: since $\nabla φ (x_{0}) = 0$ , on a sufficiently small ball $B (x_{0}, η)$ one has $φ (x) = φ (x_{0}) + \frac{1}{2} ⟨ Q (x - x_{0}), x - x_{0} ⟩ + R (x)$ with $Q = D^{2} φ (x_{0})$ and $∣ R (x) ∣ \leq C_{3} ∣ x - x_{0} ∣^{3}$ , $∣ \nabla R (x) ∣ \leq C_{3} ∣ x - x_{0} ∣^{2}$ ; consequently $\nabla φ (x) = Q (x - x_{0}) + \nabla R (x)$ and, $Q$ being invertible with smallest singular value $μ > 0$ , $∣ \nabla φ (x) ∣ \geq \frac{μ}{2} ∣ x - x_{0} ∣$ for $∣ x - x_{0} ∣$ small. (Multivariable Taylor formula with a Lagrange remainder along a line segment, Second-order Taylor expansion $f (a + h) = f (a) + \nabla f (a) \cdot h + \frac{1}{2} h^{T} H_{f} (a) h + o (∥ h ∥^{2})$ , The multivariable Taylor polynomial in multi-index notation)

[F5]

Finite compact localization: for a compact set $K$ covered by finitely many open balls, first cover $K$ by finitely many smaller balls whose closures lie in members of the original cover. Choose smooth nonnegative bumps $β_{j}$ supported in those cover members and equal to one on the smaller balls. Their sum $B$ is positive near $K$ . For $0 < c < \min_{K} B$ , choose a smooth scalar cutoff $η$ zero when $B \leq c / 2$ and one when $B \geq c$ ; then $ρ_{j} = η (B) β_{j} / B$ where $B > 0$ , extended by zero, are smooth compactly supported functions subordinate to the original balls and sum to one near $K$ . This uses only finite choices and smooth Euclidean cutoffs. (Explicit compactly supported smooth cutoffs, For a nonempty subset of $R^{n}$ with $n \geq 1$ , compactness, closedness and boundedness, pseudocompactness, and attainment of extrema by every continuous real-valued function are equivalent, Sums, scalar multiples, products and quotients: $(f + g)^{'} (c) = f^{'} (c) + g^{'} (c)$ , $(α f)^{'} (c) = α f^{'} (c)$ , $(f g)^{'} (c) = f^{'} (c) g (c) + f (c) g^{'} (c)$ , and $(f / g)^{'} (c) = (f^{'} (c) g (c) - f (c) g^{'} (c)) / g (c)^{2}$ when $g (c) \neq 0$ , $C^{k}$ maps and multi-index derivative notation in Euclidean space)

[F6]

Differentiation under the integral sign: for compactly supported smooth integrands depending smoothly on a parameter, $\partial_{λ}^{k} \int e^{2 π i λ (φ (x) - φ (x_{0}))} a (x) d x = \int (2 π i (φ (x) - φ (x_{0})))^{k} e^{2 π i λ (φ (x) - φ (x_{0}))} a (x) d x$ for every $k \geq 0$ . (Differentiation under the integral sign, Dominated convergence)

[F7]

Volume and annulus integrals: the ball of radius $δ$ has volume $ω_{d} δ^{d}$ , and for $2 N > d$ the annulus $(δ, η)$ satisfies $\int_{δ \leq ∣ x ∣ \leq η} ∣ x ∣^{- 2 N} d x \leq C_{d, N} δ^{d - 2 N}$ for $0 < δ \leq η \leq 1$ . (The volume of a radius- $r$ closed $n$ -ball is $π^{n / 2} r^{n} / Γ (n / 2 + 1)$ , Fubini's theorem for L^1 functions on a sigma-finite product, The nonnegative Lebesgue integral)

Proof

technique · direct; localize with cutoffs, gain $\lambda^{-1}$ per integration by parts against the phase gradient, and balance the gain against the $|x|^{-2}$ loss and the annulus volume to obtain the exponent $d/2$

1.1F1F5givenalgebra

Non-stationary decay (a). Assume $∣ \nabla φ ∣ \geq κ > 0$ on a neighbourhood of $supp a$ . Cover $supp a$ by finitely many balls on each of which some partial derivative satisfies $∣ \partial_{j} φ ∣ \geq κ / \sqrt{d}$ ; such a cover exists because $∣ \nabla φ ∣$ is the Euclidean norm of the gradient. By [F5] choose a smooth partition of unity $\sum_{r} ρ_{r} = 1$ on a neighbourhood of $supp a$ subordinated to that cover and replace $a$ by $a ρ_{r}$ , reducing to the case where $∣ \partial_{j} φ ∣ \geq c > 0$ on $supp a$ for one index $j$ . Put $ψ : = 1 / \partial_{j} φ$ on a neighbourhood of $supp a$ and $F : = e^{2 π i λ φ}$ . Since $\partial_{j} F = 2 π i λ \partial_{j} φ F$ , [F1] gives $\int e^{2 π i λ φ} a d x = (2 π i λ)^{- 1} \int (\partial_{j} F) ψ a d x = - (2 π i λ)^{- 1} \int F \partial_{j} (ψ a) d x .$ Iterating this identity $N$ times (each step replaces the amplitude by $\partial_{j} (ψ \cdot)$ , a smooth compactly supported function) yields $∣ I (λ) ∣ \leq (2 π λ)^{- N} C_{N}$ , where $C_{N}$ is a finite supremum of derivatives of $a, ψ$ on the support; summing the finitely many patches gives (a).

2.1F1F2F5step 1.1

Localization at the stationary point (b). Fix $x_{0}$ and $Q$ as in [F2], and choose $η > 0$ so small that the expansion and the lower bound $∣ \nabla φ (x) ∣ \geq \frac{μ}{2} ∣ x - x_{0} ∣$ hold on the ball $B (x_{0}, η)$ and $B (x_{0}, η)$ is contained in the interior of $supp a$ where needed; choose a smooth cutoff $ρ$ with $ρ = 1$ on $B (x_{0}, η / 2)$ , $ρ = 0$ outside $B (x_{0}, η)$ , and put $a_{1} : = ρ a$ , $a_{2} : = (1 - ρ) a$ . On $supp a_{2}$ the gradient does not vanish and is bounded below by a positive number depending on $η$ , $a$ and $φ$ , so [F2] does not apply there but (a) of step 1.1 does: $∣ \int e^{2 π i λ φ} a_{2} ∣ \leq C_{N} λ^{- N}$ for every $N$ . Hence it suffices to estimate $I_{1} (λ) = \int e^{2 π i λ φ} a_{1}$ .

3.1F5F7step 2.1algebra

Smooth dyadic decomposition. Translate $x_{0}$ to zero. Choose a smooth radial cutoff $θ$ equal to one for $∣ x ∣ \leq 1$ and zero for $∣ x ∣ \geq 2$ . Put $r = λ^{- 1 / 2}$ . If $r$ is comparable to or larger than the fixed localization radius $η$ , the crude compact-support bound already gives the claimed estimate, after adjusting a constant on that bounded interval of $λ$ . Otherwise $a_{1} θ (x / r)$ is supported in $∣ x ∣ \leq 2 r$ and its integral is bounded by $C r^{d} = C λ^{- d / 2}$ . The remaining amplitude has the telescoping smooth decomposition $a_{1} \sum_{j \geq 0} [θ (x / (2^{j + 1} r)) - θ (x / (2^{j} r))]$ ; only finitely many summands meet its support. The $j$ th term is supported where $s_{j} \leq ∣ x ∣ \leq 4 s_{j}$ , $s_{j} = 2^{j} r$ , and its derivative of order $ℓ$ is bounded by $C_{ℓ} s_{j}^{- ℓ}$ , for $s_{j}$ below a fixed constant.

4.1F1F2F5F7step 3.1algebra

Smooth annulus estimates. On each such annulus put $V = \nabla φ / ∣ \nabla φ ∣^{2}$ . Taylor's estimate [F2] and the product and quotient rules give $∣ \partial^{α} V (x) ∣ \leq C_{α} s_{j}^{- 1 - ∣ α ∣}$ there: the numerator is $O (s_{j})$ , its higher derivatives are bounded, and the denominator is at least $c s_{j}^{2}$ ; differentiating the reciprocal and using the product rule gives the stated bounds by induction. For its smooth compactly supported amplitude $u_{j}$ , integration by parts over all of $R^{d}$ gives $\int e^{2 π i λ φ} u_{j} = - (2 π i λ)^{- 1} \int e^{2 π i λ φ} div (V u_{j})$ . After $N$ iterations the new amplitude is bounded by $C_{N} λ^{- N} s_{j}^{- 2 N}$ : each divergence consumes one derivative and one factor $V$ , and the derivative estimates of step 3.1 and of $V$ give this bound by the Leibniz rule. Its support has volume at most $C s_{j}^{d}$ , so its integral is at most $C_{N} λ^{- N} s_{j}^{d - 2 N}$ . Choose $2 N > d$ and sum the geometric series over $s_{j} = 2^{j} r$ ; it is bounded by $C λ^{- N} r^{d - 2 N} = C λ^{- d / 2}$ . Every integration uses smooth compact support away from zero, so there is no boundary term or singular vector field at the stationary point. Combining this with steps 2.1 and 3.1 proves the basic estimate.

5.1F2F6F7step 1.1step 3.1step 4.1algebra

Differentiated bounds. By [F6] the $k$ th derivative of the centered local integral has amplitude $u_{k} = (2 π i (φ - φ (0)))^{k} a_{1}$ . Taylor's formula gives $∣ \partial^{ℓ} u_{k} ∣ \leq C_{k, ℓ} ∣ x ∣^{\max (2 k - ℓ, 0)}$ near zero: a derivative of order $ℓ$ distributed among the $k$ factors reduces the total vanishing order by at most $ℓ$ . On the small ball its absolute integral is at most $C r^{d + 2 k}$ . On the smooth annulus the cutoff amplitude has derivative bounds $C_{k, ℓ} s_{j}^{2 k - ℓ}$ (also for $ℓ > 2 k$ , since $s_{j}$ is bounded above and $s_{j}^{2 k - ℓ}$ is then bounded below). The same $N$ integrations give the bound $C λ^{- N} s_{j}^{d + 2 k - 2 N}$ . Choosing $2 N > d + 2 k$ and summing yields $C λ^{- N} r^{d + 2 k - 2 N} = C λ^{- d / 2 - k}$ . On the nonstationary support, differentiation of the centered exponential multiplies its fixed amplitude by $(2 π i (φ - φ (0)))^{k}$ ; step 1.1 still gives arbitrarily fast decay. This proves every centered derivative estimate.

6.1step 1.1step 2.1step 3.1step 4.1step 5.1∎

Conclusion. Step 1.1 proves (a) for all $N$ ; steps 2.1–4.1 prove the $O (λ^{- d / 2})$ bound of (b) with the stated dependence on $d$ , finitely many derivatives of $a, φ$ , $μ$ and the lower bound for $∣ \nabla φ ∣$ off the small ball; step 5.1 proves the differentiated bounds. The argument uses finite-dimensional Taylor estimates, the stated Fubini and differentiation-under-the-integral interfaces, the fundamental theorem, and smooth compact-support integration by parts. All spatial partitions in [F5] use finitely many Euclidean bumps. No additional Choice principle is invoked beyond the stated hypotheses of these integration suppliers; this does not assert that every item in their transitive foundational closure has a choice-free proof.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Spherical cap and dual slab scales

Statement

Assume Countable Choice and let $n \geq 2$ . For $δ \in (0, 1]$ let $C_{δ} = {ω \in S^{n - 1} : 1 - ω \cdot e_{n} \leq δ^{2}}$ and, for a fixed $c > 0$ , $T_{δ} = {ξ \in R^{n} : ∣ ξ_{n} ∣ \leq c δ^{- 2}, ∣ ξ_{j} ∣ \leq c δ^{- 1} (j < n)}$ . Then $C_{δ}$ lies in the closed hemisphere ${ω_{n} \geq 0}$ and, in the graph chart $ω = (y, \sqrt{1 - ∣ y ∣^{2}})$ whose surface density is $(1 - ∣ y ∣^{2})^{- 1 / 2}$ , $σ (C_{δ}) = \int_{∣ y ∣^{2} \leq 2 δ^{2} - δ^{4}} (1 - ∣ y ∣^{2})^{- 1 / 2} d y$ ; consequently there are constants $0 < c_{n} \leq C_{n} < \infty$ with $c_{n} δ^{n - 1} \leq σ (C_{δ}) \leq C_{n} δ^{n - 1}$ , while $λ_{n} (T_{δ}) = (2 c)^{n} δ^{- (n + 1)}$ . Moreover $diam C_{δ} \leq 2 \sqrt{2} δ$ , and the orthogonal group preserves $σ$ , so the same scales hold for every cap ${ω : 1 - ω \cdot v \leq δ^{2}}$ with $v \in S^{n - 1}$ .

At $δ = 1$ , the integral formula is read on $∣ y ∣ < 1$ ; the omitted equator has surface measure zero, as proved below.

Facts & Assumptions

Given: Countable Choice, $n \geq 2$ , $δ \in (0, 1]$ , $c > 0$ , the cap $C_{δ}$ and the slab $T_{δ}$ .

[F1]

Sphere chart and density: in the graph chart $ω = (y, \sqrt{1 - ∣ y ∣^{2}})$ over $B \subseteq R^{n - 1}$ the polar surface measure has Lebesgue density $(1 - ∣ y ∣^{2})^{- 1 / 2}$ , and the chart measure equals $σ$ ; orthogonal transformations preserve $σ$ , and the chart measure is invariant under the linear change of variables used below. (Sphere graph charts, surface density, and a finite partition, Agreement with the existing polar sphere measure, A linear map $T$ of $R^{n}$ sends Lebesgue measurable sets to Lebesgue measurable sets, with $λ_{n} (T [E]) = ∣ \det T ∣ λ_{n} (E)$ when $T$ is invertible and $T [E]$ Lebesgue null when it is not, The Jacobian matrix of partial derivatives and the gradient in the scalar-valued case)

[F2]

Product structure: for a box $R = \prod_{i < n} (a_{i}, b_{i}]$ one has $λ_{n} (R) = \prod_{i} (b_{i} - a_{i})$ , and the $(n - 1)$ -dimensional ball of radius $r$ has volume $ω_{n - 1} r^{n - 1}$ ; iterated integrals over product domains are computed by Tonelli. (A box in $R^{n}$ with parameters $a_{i} \leq b_{i}$ is Lebesgue measurable of measure $\prod_{i < n} (b_{i} - a_{i})$ , whichever of its faces are included, The volume of a radius- $r$ closed $n$ -ball is $π^{n / 2} r^{n} / Γ (n / 2 + 1)$ , Tonelli's theorem for nonnegative measurable functions on a sigma-finite product)

[F3]

Polar coordinates identify $σ (S^{n - 1}) = n λ_{n} (B_{1}^{n})$ and give finiteness of $σ$ ; the sphere has unit radius. (Polar coordinates decompose Lebesgue measure into r^{n-1} dr d sigma, Agreement with the existing polar sphere measure)

Proof

technique · direct; rewrite the cap inequality in the graph chart, integrate the chart density with two-sided bounds, compute the slab as a box, and estimate the diameter from the cap condition

1.1F1F2givenalgebra

The cap lies in the closed hemisphere since $ω_{n} \geq 1 - δ^{2} \geq 0$ . For $δ < 1$ it lies in the open upper chart, and squaring $\sqrt{1 - ∣ y ∣^{2}} \geq 1 - δ^{2}$ gives $∣ y ∣^{2} \leq 2 δ^{2} - δ^{4}$ , yielding the stated integral with [F1]. At $δ = 1$ the cap is the closed upper hemisphere. Its equator has surface measure zero: cover the equator by the other hemisphere charts; there $ω_{n}$ is one parameter coordinate, and the equator is a coordinate hyperplane of Lebesgue measure zero by Tonelli (a singleton coordinate has zero length). The chart density is finite on its open domain, so integrating it over that null set gives zero. Thus the formula also holds at $δ = 1$ , integrating over $∣ y ∣ < 1$ ; boundary values may be assigned arbitrarily.

1.2F2algebra

The slab. $T_{δ}$ is the box $\prod_{j < n} [- c δ^{- 1}, c δ^{- 1}] \times [- c δ^{- 2}, c δ^{- 2}]$ , a product of $n - 1$ intervals of length $2 c δ^{- 1}$ and one of length $2 c δ^{- 2}$ . By [F2], $λ_{n} (T_{δ}) = (2 c δ^{- 1})^{n - 1} (2 c δ^{- 2}) = (2 c)^{n} δ^{- (n - 1) - 2} = (2 c)^{n} δ^{- (n + 1)}$ .

2.1F2F3step 1.1algebra

Two-sided bounds for the cap. Lower bound: the ball $∣ y ∣ \leq δ$ satisfies $∣ y ∣^{2} \leq δ^{2} \leq 2 δ^{2} - δ^{4}$ and on it the density is at least $1$ ; by [F2], $σ (C_{δ}) \geq ω_{n - 1} δ^{n - 1}$ . Upper bound: on the cap $1 - ∣ y ∣^{2} \geq (1 - δ^{2})^{2}$ , so the density is at most $(1 - δ^{2})^{- 1}$ ; the domain is contained in the ball of radius $\sqrt{2 δ^{2} - δ^{4}} \leq \sqrt{2} δ$ , so for $δ \leq 1 / 2$ the density factor is at most $4 / 3$ and $σ (C_{δ}) \leq (4 / 3) ω_{n - 1} 2^{(n - 1) / 2} δ^{n - 1}$ . For $δ \geq 1 / 2$ one has $C_{δ} \subseteq S^{n - 1}$ and $δ^{n - 1} \geq 2^{- (n - 1)}$ , so $σ (C_{δ}) \leq σ (S^{n - 1}) = n λ_{n} (B_{1}^{n}) \leq n λ_{n} (B_{1}^{n}) 2^{n - 1} δ^{n - 1}$ by [F3]. Thus $c_{n} δ^{n - 1} \leq σ (C_{δ}) \leq C_{n} δ^{n - 1}$ with $c_{n} : = ω_{n - 1}$ and $C_{n} : = \max {(4 / 3) ω_{n - 1} 2^{(n - 1) / 2}, n λ_{n} (B_{1}^{n}) 2^{n - 1}}$ .

2.2givenstep 1.1algebra

The diameter. Let $ω, ω^{'} \in C_{δ}$ and write $ω = (u, ω_{n})$ , $ω^{'} = (u^{'}, ω_{n}^{'})$ with $u, u^{'} \in R^{n - 1}$ . From step 1.1, $∣ u ∣^{2} \leq 2 δ^{2} - δ^{4}$ and $∣ u^{'} ∣^{2} \leq 2 δ^{2} - δ^{4}$ , while $ω_{n}, ω_{n}^{'} \geq 1 - δ^{2}$ . Hence $ω \cdot ω^{'} = u \cdot u^{'} + ω_{n} ω_{n}^{'} \geq - ∣ u ∣ ∣ u^{'} ∣ + (1 - δ^{2})^{2} \geq - (2 δ^{2} - δ^{4}) + (1 - δ^{2})^{2} = 1 - 4 δ^{2} + 2 δ^{4},$ so $∣ ω - ω^{'} ∣^{2} = 2 - 2 ω \cdot ω^{'} \leq 8 δ^{2} - 4 δ^{4} \leq 8 δ^{2}$ and $diam C_{δ} \leq 2 \sqrt{2} δ$ .

3.1F1step 2.1step 2.2

Rotational reduction. For $v \in S^{n - 1}$ choose an orthogonal map $R$ with $R e_{n} = v$ ; finite-dimensional orthogonal algebra supplies such an $R$ without choice. The cap ${ω : 1 - ω \cdot v \leq δ^{2}} = R [C_{δ}]$ is the image of $C_{δ}$ under an orthogonal transformation, which preserves $σ$ by [F1]; hence it has the same measure and the same diameter bound.

4.1step 1.1step 2.1step 1.2step 2.2step 3.1∎

Conclusion. Step 1.1 rewrites the cap in the graph chart, step 2.1 gives the two-sided scale $c_{n} δ^{n - 1} \leq σ (C_{δ}) \leq C_{n} δ^{n - 1}$ , step 1.2 computes $λ_{n} (T_{δ}) = (2 c)^{n} δ^{- (n + 1)}$ , step 2.2 gives $diam C_{δ} \leq 2 \sqrt{2} δ$ , and step 3.1 transfers the scales to arbitrary axis $v$ by orthogonal invariance.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Compact curved hypersurfaces admit a finite curved graph cover

Statement

Assume Countable Choice. Let $S \subseteq R^{n}$ ( $n \geq 2$ ) be a compact embedded $C^{\infty}$ hypersurface with a continuous unit normal field $ν$ and everywhere nonvanishing extrinsic Gaussian curvature $K = \det S_{ν} \neq 0$ . Then there exist finitely many open sets $U_{j} \subseteq R^{n - 1}$ , smooth $h_{j} : U_{j} \to R$ with $\det D^{2} h_{j} \neq 0$ on $U_{j}$ , embeddings $X_{j} (y) = (y, h_{j} (y))$ of $graph h_{j}$ onto relatively open pieces $S_{j} \subseteq S$ covering $S$ (after ambient rigid motions), and nonnegative $C^{\infty}$ functions $χ_{j}$ on $S$ with $\sum_{j} χ_{j} = 1$ and $supp χ_{j}$ compactly contained in $S_{j}$ .

Facts & Assumptions

Given: The compact embedded smooth hypersurface, continuous unit normal and nonzero curvature in the statement, with Countable Choice.

[F1]

Smooth graph charts, compact smooth localization, normal independence and the Euclidean curvature convention are established locally. (Smooth Euclidean hypersurface graphs and compact localization, Euclidean hypersurface normals, shape operators and curvature)

[F2]

The graph curvature is $\det D^{2} h / (1 + ∣ \nabla h ∣^{2})^{(n + 1) / 2}$ . (Shape operator and Gauss-Kronecker curvature of a graph)

[A1]

Countable Choice is assumed. (The Axiom of Countable Choice ( ${A C}_{ω}$ ))

Proof

technique · direct; apply the local graph and compact localization constructions

1.1givenF1F2

By [F1], every point has a smooth graph chart after a rigid motion. The given continuous normal is locally smooth and equals either the graph normal or its negative with constant sign on a connected smaller chart. Its shape operator therefore differs by that sign; nonvanishing curvature is unchanged. Formula [F2] implies $\det D^{2} h \neq 0$ throughout the smaller graph chart. This argument also works with local normals only, without a global orientation.

2.1F1A1step 1.1∎

Apply the compact localization part of [F1] with $K = S$ to the graph neighbourhoods of step 1.1. Its ambient-ball bumps give finitely many pieces covering $S$ and nonnegative smooth $χ_{j}$ with compact support inside their pieces and sum one. Their graph functions retain their nondegenerate Hessians on the whole chart. These are all the asserted data. The construction needs only finite choices; the assumed Countable Choice remains available to surface-measure consumers.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

TT-star reduces extension to convolution with the surface-measure transform

Statement

Assume Countable Choice. Let $μ$ be a finite positive Borel measure on $R^{n}$ , let $E_{μ} g = (g μ)^{\lor}$ for $g \in L^{1} (μ) \cap L^{2} (μ)$ , and let $E_{μ}^{*} F$ be the restriction of $\hat{F}$ to $supp μ$ for $F \in S (R^{n})$ . Then $E_{μ} E_{μ}^{*} F = F * \overset{ˇ}{μ}$ for every $F \in S (R^{n})$ , and for every $1 \leq p \leq \infty$ , $∥ \hat{F} ∥_{L^{2} (μ)}^{2} = ⟨ (\hat{F} μ)^{\lor}, F ⟩ = ⟨ F * \overset{ˇ}{μ}, F ⟩ \leq ∥ F ∥_{L^{p}} ∥ F * \overset{ˇ}{μ} ∥_{L^{p^{'}}}$ . In particular, for a compact hypersurface $S$ with surface measure $σ$ , a bound on $∥ F * \overset{ˇ}{σ} ∥_{L^{p^{'}}}$ bounds $∥ \hat{F} ∥_{L^{2} (σ)}$ .

The brackets here denote the absolutely convergent integral $⟨ H, F ⟩ = \int H \overline{F} d x$ , not a claim that $H \in L^{2} (R^{n})$ . Positivity of $μ$ is essential to the squared-norm identity.

Facts & Assumptions

Given: Countable Choice, a finite positive Borel measure $μ$ on $R^{n}$ with $μ (R^{n}) < \infty$ , $F \in S (R^{n})$ , and $1 \leq p \leq \infty$ with conjugate $p^{'}$ .

[F1]

For $g \in L^{1} (μ)$ the extension $E_{μ} g = (g μ)^{\lor}$ is the bounded uniformly continuous function $x \mapsto \int e^{2 π i x \cdot ω} g (ω) d μ (ω)$ , and $E_{μ}^{*} F = \hat{F} ∣_{supp μ}$ , so that $g = E_{μ}^{*} F$ is the restriction of the Schwartz transform; $\overset{ˇ}{μ} (x) = \hat{μ} (- x) = \int e^{2 π i x \cdot ω} d μ (ω)$ is bounded with $∣ \overset{ˇ}{μ} ∣ \leq ∣ μ ∣ (R^{n})$ . (Fourier restriction and adjoint extension operators, Fourier transform of a finite complex Borel measure, A complex L^1 density defines a complex measure whose total variation is |h| dmu)

[F2]

Fourier pairing and convolution identity: for all $F, G \in S (R^{n})$ , $\int (G μ)^{\lor} \overline{F} d x = \int G \overline{\hat{F}} d μ$ , and $(\hat{F} μ)^{\lor} = F * \overset{ˇ}{μ}$ holds as an identity of everywhere-defined bounded continuous functions. (Fourier pairing for a finite measure and Schwartz data, A complex measure is a finite-valued countably additive set function)

[F3]

Hölder and inner products: on a measure space, $\int ∣ f g ∣ d ν \leq ∥ f ∥_{r} ∥ g ∥_{r^{'}}$ for conjugate $r, r^{'}$ and complex measurable $f, g$ (endpoint cases included), and the complex $L^{2}$ pairing is $⟨ h, k ⟩ = \int h \overline{k}$ with $∣ ⟨ h, k ⟩ ∣ \leq ∥ h ∥_{2} ∥ k ∥_{2}$ ; the $L^{2}$ norm is $∥ h ∥_{2}^{2} = ⟨ h, h ⟩ = \int ∣ h ∣^{2}$ . (Complex Holder, Minkowski, and the quotient norm, Holder's inequality for integrals, including the endpoint cases, The complex $L^{2}$ pairing on equivalence classes, The complex $L^{2}$ pairing is well-defined and satisfies Cauchy–Schwarz, Conjugate exponents, including the endpoint conventions)

Proof

technique · direct; apply the pairing and convolution identities of the finite-measure lemma to $G=\widehat F$ and then bound the resulting pairing by Hölder

1.1F1F2

The $T T^{*}$ identity. Applying the convolution identity [F2] with the given $F$ gives $(\hat{F} μ)^{\lor} = F * \overset{ˇ}{μ}$ as functions on $R^{n}$ . Since $E_{μ}^{*} F = \hat{F} ∣_{supp μ}$ and $E_{μ} (\hat{F} ∣_{supp μ}) = (\hat{F} μ)^{\lor}$ , this reads $E_{μ} E_{μ}^{*} F = F * \overset{ˇ}{μ}$ , which is the first assertion.

2.1F1F2F3step 1.1

The squared norm identity. The $L^{2} (μ)$ norm is computed by [F3]: $∥ \hat{F} ∥_{L^{2} (μ)}^{2} = \int ∣ \hat{F} ∣^{2} d μ = \int \hat{F} \overline{\hat{F}} d μ$ . Applying the pairing identity [F2] with $G = \hat{F}$ (a Schwartz function) gives $\int (\hat{F} μ)^{\lor} \overline{F} d x = \int \hat{F} \overline{\hat{F}} d μ$ , that is $\int (\hat{F} μ)^{\lor} \overline{F} d x = ∥ \hat{F} ∥_{L^{2} (μ)}^{2}$ . By step 1.1 the left side equals $⟨ F * \overset{ˇ}{μ}, F ⟩$ , so the first two equalities of the statement hold.

3.1F3step 2.1

The Hölder bound. By [F3] for the $L^{p}$ – $L^{p^{'}}$ pairing with $h = F * \overset{ˇ}{μ}$ and $k = F$ , $∣ ⟨ F * \overset{ˇ}{μ}, F ⟩ ∣ = ∣ \int (F * \overset{ˇ}{μ}) \overline{F} d x ∣ \leq ∥ F * \overset{ˇ}{μ} ∥_{p^{'}} ∥ F ∥_{p}$ when $F * \overset{ˇ}{μ} \in L^{p^{'}}$ ; if that norm is infinite, the inequality is automatic. The pairing itself is absolutely convergent since $F \in L^{1}$ and $F * \overset{ˇ}{μ}$ is bounded. Combining with step 2.1, $∥ \hat{F} ∥_{L^{2} (μ)}^{2} = ⟨ F * \overset{ˇ}{μ}, F ⟩ \leq ∥ F ∥_{p} ∥ F * \overset{ˇ}{μ} ∥_{p^{'}}$ .

4.1F1F2F3step 2.1step 3.1

Specialization to a hypersurface. Let $S$ be a compact hypersurface with surface measure $σ$ ; then $σ$ is a finite Borel measure, $\overset{ˇ}{σ} = \hat{σ} (- \cdot)$ obeys $∣ \overset{ˇ}{σ} ∣ \leq σ (S)$ , and the identities of steps 1.1–3.1 hold with $μ = σ$ . Consequently, if there is $C < \infty$ with $∥ F * \overset{ˇ}{σ} ∥_{L^{p^{'}}} \leq C ∥ F ∥_{L^{p}}$ for every $F \in S (R^{n})$ , then $∥ \hat{F} ∥_{L^{2} (σ)}^{2} \leq C ∥ F ∥_{p}^{2}$ , that is $∥ \hat{F} ∥_{L^{2} (σ)} \leq C^{1 / 2} ∥ F ∥_{L^{p}}$ : a bound on the convolution with $\overset{ˇ}{σ}$ bounds the restriction estimate.

5.1step 1.1step 2.1step 3.1step 4.1∎

Conclusion. Step 1.1 proves $E_{μ} E_{μ}^{*} F = F * \overset{ˇ}{μ}$ ; steps 2.1 and 3.1 prove the chain $∥ \hat{F} ∥_{L^{2} (μ)}^{2} = ⟨ (\hat{F} μ)^{\lor}, F ⟩ = ⟨ F * \overset{ˇ}{μ}, F ⟩ \leq ∥ F ∥_{p} ∥ F * \overset{ˇ}{μ} ∥_{p^{'}}$ for every $1 \leq p \leq \infty$ ; step 4.1 records the hypersurface specialization. Countable Choice is inherited from the finite-measure pairing and transform suppliers.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Cap wave packets concentrate on the dual tube

Statement

Assume Countable Choice and let $n \geq 2$ . There is $c_{n} > 0$ such that for every $δ \in (0, 1]$ , every $v \in S^{n - 1}$ , every $η \in R^{n}$ and every rotation $R$ with $R e_{n} = v$ the following holds. With $C_{δ} (v) = {ω \in S^{n - 1} : 1 - ω \cdot v \leq δ^{2}}$ , the tube $T = η + {ξ : ∣ ξ \cdot v ∣ \leq c_{n} δ^{- 2}, ∣ ξ - (ξ \cdot v) v ∣ \leq c_{n} δ^{- 1}}$ and the data $g_{η} (ω) = e^{- 2 π i η \cdot ω} 1_{C_{δ} (v)} (ω)$ , one has $∣ E g_{η} (x) ∣ \geq \frac{1}{2} σ (C_{δ} (v))$ for every $x \in T$ . In particular the extension of cap data of angular radius $δ$ is essentially coherent on a dual tube of dimensions $δ^{- 1} \times \dots \times δ^{- 1} \times δ^{- 2}$ .

Facts & Assumptions

Given: Countable Choice, $n \geq 2$ , $δ \in (0, 1]$ , $v \in S^{n - 1}$ , $η \in R^{n}$ , and $ξ \in R^{n}$ with $∣ ξ \cdot v ∣ \leq c_{n} δ^{- 2}$ and $∣ ξ - (ξ \cdot v) v ∣ \leq c_{n} δ^{- 1}$ for a constant $c_{n} \in (0, 1 / 100]$ to be fixed below; write $x = η + ξ$ .

[F1]

Extension: for $g \in L^{1} (σ)$ one has $E g (x) = \int_{S^{n - 1}} e^{2 π i x \cdot ω} g (ω) d σ (ω)$ , with $L^{1} (σ)$ computed componentwise for complex functions, and for a real $φ$ one has $∣ E g (x) ∣ \geq Re (e^{- i φ} E g (x))$ . (Fourier restriction and adjoint extension operators, Complex Lp classes and Euclidean test-function conventions, The Lebesgue integral is linear on $L^{1} (μ)$ )

[F2]

Cap geometry: $ω \cdot v \geq 1 - δ^{2}$ on $C_{δ} (v)$ , the diameter satisfies $diam C_{δ} (v) \leq 2 \sqrt{2} δ$ , and $σ (C_{δ} (v)) > 0$ . (Spherical cap and dual slab scales)

[F3]

Unit-circle estimates: $\sin 0 = 0$ and $∣ e^{i θ} - 1 ∣ = 2 ∣ \sin (θ / 2) ∣ \leq ∣ θ ∣$ for every real $θ$ , because $∣ \sin u ∣ \leq ∣ u ∣$ and $e^{i θ} = \cos θ + i \sin θ$ ; consequently $1 - \cos θ = \frac{1}{2} ∣ e^{i θ} - 1 ∣^{2} \leq ∣ e^{i θ} - 1 ∣ \leq ∣ θ ∣$ , so $Re e^{i θ} = \cos θ \geq 1 - ∣ θ ∣$ . In particular $Re e^{i θ} \geq \frac{1}{2}$ whenever $∣ θ ∣ \leq \frac{1}{2}$ . ( $\exp (x + i y) = e^{x} (\cos y + i \sin y)$ , $∣ \exp (x + i y) ∣ = e^{x}$ , and $e^{i π} + 1 = 0$ , Sine and cosine are $1$ -Lipschitz on $R$ , The zero sets of sine and cosine and the least positive common period 2 pi)

Proof

technique · direct; evaluate the extension of the modulated cap data, expand the phase in the cap parameter, and use the elementary lower bound for the real part of a unit-modulus exponential

1.1F1given

The extension of the data. With $g_{η} = e^{- 2 π i η \cdot ω} 1_{C_{δ} (v)}$ and $x = η + ξ$ , [F1] gives $E g_{η} (x) = \int_{C_{δ} (v)} e^{2 π i x \cdot ω} e^{- 2 π i η \cdot ω} d σ (ω) = \int_{C_{δ} (v)} e^{2 π i ξ \cdot ω} d σ (ω)$ , since $x - η = ξ$ .

1.2F2F3givenalgebra

The phase on the cap. Let $ω \in C_{δ} (v)$ and write $ω = v + (ω - v)$ . Then $ξ \cdot ω = ξ \cdot v + ξ \cdot (ω - v)$ , and by [F2] and the tube inequalities $∣ ξ \cdot (ω - v) ∣ \leq ∣ (ξ \cdot v) (v \cdot (ω - v)) ∣ + ∣ (ξ - (ξ \cdot v) v) \cdot (ω - v) ∣ \leq c_{n} δ^{- 2} δ^{2} + 2 \sqrt{2} c_{n} δ^{- 1} δ \leq (1 + 2 \sqrt{2}) c_{n} .$ Choose $c_{n} : = 1 / 100$ , so $∣ ξ \cdot (ω - v) ∣ \leq (1 + 2 \sqrt{2}) / 100 < 1 / (4 π)$ and $∣ 2 π ξ \cdot (ω - v) ∣ \leq \frac{1}{2}$ . By the last clause of [F3], $Re e^{2 π i ξ \cdot (ω - v)} \geq \frac{1}{2}$ for every $ω \in C_{δ} (v)$ .

2.1F1step 1.1step 1.2

The lower bound. Multiplying step 1.1 by the unimodular factor $e^{- 2 π i ξ \cdot v}$ and applying [F1], $∣ E g_{η} (x) ∣ \geq Re (e^{- 2 π i ξ \cdot v} E g_{η} (x)) = Re \int_{C_{δ} (v)} e^{2 π i ξ \cdot (ω - v)} d σ (ω) = \int_{C_{δ} (v)} Re e^{2 π i ξ \cdot (ω - v)} d σ (ω) \geq \frac{1}{2} σ (C_{δ} (v)),$ the middle equality by componentwise integration of complex-valued functions [F1] and the last inequality by step 1.2 integrated against the positive measure $σ$ ; this holds for every $x = η + ξ \in T$ . The tube is a product of a tangential ball of radius $c_{n} δ^{- 1}$ and a normal interval of length $2 c_{n} δ^{- 2}$ . In any orthonormal tangential frame it contains the box with each tangential half-width $c_{n} δ^{- 1} / \sqrt{n - 1}$ and normal half-width $c_{n} δ^{- 2}$ ; this has the stated scales.

3.1step 1.2step 2.1∎

Conclusion. Steps 1.1–2.1 prove that for the explicit constant $c_{n} = 1 / 100$ the extension of the cap data $g_{η}$ is bounded below by $\frac{1}{2} σ (C_{δ} (v))$ on the whole dual tube $T$ , uniformly in $δ, v, η$ .

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Decay of a localized measure on a curved graph patch

Statement

Assume Countable Choice. Let $n \geq 2$ , let $U \subseteq R^{n - 1}$ be open, $h \in C^{\infty} (U)$ , $a \in C_{c}^{\infty} (U)$ , let $S = graph h$ carry the graph surface measure $σ$ , and put $d μ : = a (y) \sqrt{1 + ∣ \nabla h (y) ∣^{2}} d y$ (the localization of $σ$ by the pullback of $a$ ). If $\det D^{2} h \neq 0$ on $supp a$ , then $∣ \overset{ˇ}{μ} (x) ∣ = ∣ \int e^{2 π i (x^{'} \cdot y + x_{n} h (y))} a (y) \sqrt{1 + ∣ \nabla h (y) ∣^{2}} d y ∣ \leq C_{a} (1 + ∣ x ∣)^{- (n - 1) / 2}$ for all $x = (x^{'}, x_{n}) \in R^{n}$ , with $C_{a}$ depending on $a, h, n$ . More generally, if $S$ is a smooth hypersurface and $μ = φ σ$ for $φ \in C_{c}^{\infty} (R^{n})$ whose restriction to $S$ has compact support in $S$ , with the Gaussian curvature of $S$ nonvanishing on $S \cap supp φ$ , then the same decay holds.

Facts & Assumptions

Given: The graph, amplitude, nondegenerate Hessian on its compact support, and Countable Choice in the statement.

[F1]

Nonstationary phase gives arbitrary inverse powers of the parameter; near one nondegenerate stationary point stationary phase gives the power $- (n - 1) / 2$ , with constants controlled by finite derivative bounds, inverse Hessian bounds and the gradient away from the point. (Stationary phase with a compactly supported amplitude)

[F2]

Smooth inverse/implicit bootstrap, graph charts and compactly supported finite localization follow from earlier Euclidean calculus. (Smooth Euclidean hypersurface graphs and compact localization)

[F3]

Graph Hessian nondegeneracy is equivalent to nonvanishing extrinsic Gaussian curvature, independent of local normal orientation. (Shape operator and Gauss-Kronecker curvature of a graph, Euclidean hypersurface normals, shape operators and curvature)

[F4]

Smooth cutoffs, compactness and graph surface density are available. (Explicit compactly supported smooth cutoffs, For a nonempty subset of $R^{n}$ with $n \geq 1$ , compactness, closedness and boundedness, pseudocompactness, and attainment of extrema by every continuous real-valued function are equivalent, Chart and partition independence of surface measure)

[A1]

Countable Choice is assumed. (The Axiom of Countable Choice ( ${A C}_{ω}$ ))

Proof

technique · direct; use finitely many neighbourhoods of directions and retain stationary points in slightly enlarged spatial patches

1.1givenF3F4algebra

Put $d = n - 1$ , $b = a \sqrt{1 + ∣ \nabla h ∣^{2}}$ and $K = supp a$ . Choose a compact neighbourhood $K^{+}$ of $K$ inside $U$ on which $D^{2} h$ is invertible. This exists by continuity and a finite cover of $K$ . Every derivative of $h$ needed below is bounded there. For $x = ρ ω$ , $ρ \geq 1$ and $ω \in S^{n - 1}$ , the phase is $ψ_{ω} (y) = ω^{'} \cdot y + ω_{n} h (y)$ and the integral is $\int e^{2 π i ρ ψ_{ω}} b$ . Its gradient is $ω^{'} + ω_{n} \nabla h$ . At a zero in $K^{+}$ , $∣ ω_{n} ∣ = (1 + ∣ \nabla h ∣^{2})^{- 1 / 2}$ , so the Hessian $ω_{n} D^{2} h$ is invertible with uniformly bounded inverse.

2.1F2F4step 1.1

Fix a direction $ω_{0}$ . Its zeros in $K^{+}$ are isolated by the inverse theorem in [F2]. Only finitely many lie in a smaller compact neighbourhood of $K$ : otherwise compactness gives an accumulating zero in $K^{+}$ , contradicting local invertibility. Surround these finitely many zeros by disjoint small balls compactly contained in $K^{+}$ , on which $\nabla h$ is injective; choose smaller concentric balls around the zeros. Every remaining point of $K$ has nonzero phase gradient at $ω_{0}$ . A fixed finite smooth partition on a neighbourhood of $K$ therefore splits $b$ into amplitudes supported either in these zero balls or on a compact set where $∣ \nabla ψ_{ω_{0}} ∣ \geq c > 0$ .

3.1F1F2step 1.1step 2.1algebra

Shrink a neighbourhood $V$ of $ω_{0}$ in the direction sphere. On the nonstationary support, continuity keeps the gradient at least $c / 2$ . For each zero ball, the implicit theorem provides a smooth critical point $z (ω)$ remaining in its smaller ball for $ω \in V$ . Injectivity of $\nabla h$ and $ω_{n} \neq 0$ ensure it is the only critical point in the larger ball. By shrinking the ball and $V$ , Taylor's formula makes $∣ \nabla ψ_{ω} (y) ∣ \geq c_{0} ∣ y - z (ω) ∣$ near that point uniformly: subtract the gradient at $z$ and use uniform closeness of the Hessian to its invertible value at $(ω_{0}, z (ω_{0}))$ . On the compact remainder of the ball the gradient stays bounded below after further shrinking $V$ . All required derivatives and Hessian inverses are uniformly bounded.

4.1F1F4step 3.1algebra

Apply [F1] on these supports. The nonstationary amplitudes give $O (ρ^{- N})$ uniformly on $V$ . The zero-ball amplitudes give $O (ρ^{- d / 2})$ uniformly, even when the critical point lies outside the amplitude support: add a fixed smooth bump supported in that ball, equal to one on its smaller ball and multiplied by a constant larger than the amplitude bound, then subtract the same bump. Each of the two new amplitudes has the critical point in the interior of its support and uniformly bounded derivatives, so the stated stationary estimate applies to each; the local proof uses only the phase on that ball. Thus the original amplitude has the same bound by subtraction. This also handles critical points entering or leaving the original support.

5.1F4step 1.1step 4.1algebra

The neighbourhoods $V$ constructed for each direction cover the compact sphere, so finitely many suffice. Taking the maximum of their finite constants gives $∣ \overset{ˇ}{μ} (x) ∣ \leq C ∣ x ∣^{- d / 2}$ for $∣ x ∣ \geq 1$ . For every $x$ , the pointwise estimate $∣ \overset{ˇ}{μ} (x) ∣ \leq \int ∣ b (y) ∣ d y < \infty$ follows from unit modulus of the exponential and bounded compact support. Combining the two bounds yields $C_{a} (1 + ∣ x ∣)^{- d / 2}$ after increasing $C_{a}$ . No constancy of the number of critical points over the whole sphere is asserted or used.

6.1F2F3F4A1step 5.1∎

For the general clause, $K = {supp}_{S} (φ ∣_{S})$ is compact by the explicit hypothesis. Apply [F2] to $K$ , and [F3] to its graph charts; shrink the charts to retain nondegenerate Hessians on the compact supports of the localized weights. The graph amplitudes $a_{j} = (χ_{j} φ) \circ X_{j}$ are smooth and compactly supported in their parameter domains. The graph measure formula in [F4] writes $φ σ$ as their finite sum. A rigid motion rotates the frequency and contributes only a scalar exponential of modulus one, so step 5.1 applies without changing $∣ x ∣$ . Summing proves the asserted general decay.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Stationary-phase decay for spherical surface measure

Statement

Assume Countable Choice and let $n \geq 2$ . With $σ$ the polar surface measure on $S^{n - 1}$ , $∣ \hat{σ} (ξ) ∣ \leq C_{n} (1 + ∣ ξ ∣)^{- (n - 1) / 2}$ for every $ξ \in R^{n}$ , and $\overset{ˇ}{σ}$ obeys the same bound.

Facts & Assumptions

Given: Countable Choice, $n \geq 2$ , the polar surface measure $σ$ on $S^{n - 1}$ and its transform $\hat{σ} (ξ) = \int_{S^{n - 1}} e^{- 2 π i ξ \cdot ω} d σ (ω)$ .

[F1]

Sphere charts and partition: the $2 n$ hemispheres of the graph charts $X_{i}^{ε} (y) = (y_{1}, \dots, y_{i - 1}, ε \sqrt{1 - ∣ y ∣^{2}}, y_{i}, \dots, y_{n - 1})$ cover $S^{n - 1}$ , the chart measure is $σ$ , and there is a finite smooth partition of unity $(χ_{j})$ subordinate to the images of these charts, with compactly supported pieces; the density of each chart is $(1 - ∣ y ∣^{2})^{- 1 / 2}$ and the composition with a chart turns $\int ψ d σ$ into an integral of $ψ (X_{i}^{ε} (y))$ times that density over the unit ball. (Sphere graph charts, surface density, and a finite partition, Locally finite partitions of unity and subordination to an open cover)

[F2]

Orthogonal invariance: orthogonal transformations preserve $σ$ , so for $ξ = r ν$ with $r \geq 0$ , $ν \in S^{n - 1}$ and an orthogonal $R$ with $R ν = e_{n}$ , $\hat{σ} (ξ) = \int e^{- 2 π i r ω \cdot e_{n}} d σ (ω) = \hat{σ} (r e_{n})$ . (Agreement with the existing polar sphere measure, Polar coordinates decompose Lebesgue measure into r^{n-1} dr d sigma)

[F3]

Stationary phase: for $d \geq 1$ , a compactly supported smooth amplitude $b$ on $R^{d}$ and a real phase $ψ \in C^{\infty} (R^{d})$ : if $\nabla ψ$ does not vanish on a neighbourhood of $supp b$ , then $∣ \int e^{2 π i λ ψ} b ∣ \leq C_{N} λ^{- N}$ for all $N$ ; if $ψ$ has exactly one stationary point in $R^{d}$ , lying in the interior of $supp b$ with invertible Hessian, then $∣ \int e^{2 π i λ ψ} b ∣ \leq C λ^{- d / 2}$ for $λ \geq 1$ , with constants depending on finitely many derivatives of $b, ψ$ , on a lower bound for $∣ \det D^{2} ψ ∣$ at the point and on a lower bound for $∣ \nabla ψ ∣$ off a small ball about it. (Stationary phase with a compactly supported amplitude)

[F4]

The graphing functions $h (y) = \sqrt{1 - ∣ y ∣^{2}}$ are smooth on the unit ball. Differentiating $h^{2} = 1 - ∣ y ∣^{2}$ gives $\nabla h = - y / h$ , and differentiating again gives $D^{2} h (0) = - I$ . The nonpolar coordinate phase $y_{n - 1}$ has gradient $e_{n - 1}$ . (Sphere graph charts, surface density, and a finite partition, The Jacobian matrix of partial derivatives and the gradient in the scalar-valued case)

[F5]

Trivial bound: $∣ \hat{σ} (ξ) ∣ \leq σ (S^{n - 1}) = n λ_{n} (B_{1}^{n}) < \infty$ for every $ξ$ , and for $∣ ξ ∣ \leq 1$ one has $(1 + ∣ ξ ∣)^{- (n - 1) / 2} \geq 2^{- (n - 1) / 2}$ . (Polar coordinates decompose Lebesgue measure into r^{n-1} dr d sigma, Sphere graph charts, surface density, and a finite partition)

[F6]

One-variable product, quotient and chain rules compute coordinate partials, and smoothness means continuity of all ordered partials. The standard smooth step $s_{0}$ is smooth, takes values in $[0, 1]$ , is zero on $(- \infty, 0]$ and one on $[1, \infty)$ . Every continuous real function on an interval has the integral primitive, unique up to a constant. (Sums, scalar multiples, products and quotients: $(f + g)^{'} (c) = f^{'} (c) + g^{'} (c)$ , $(α f)^{'} (c) = α f^{'} (c)$ , $(f g)^{'} (c) = f^{'} (c) g (c) + f (c) g^{'} (c)$ , and $(f / g)^{'} (c) = (f^{'} (c) g (c) - f (c) g^{'} (c)) / g (c)^{2}$ when $g (c) \neq 0$ , The chain rule, in one line from Carathéodory: if $g$ is differentiable at $c$ and $f$ is differentiable at $g (c)$ , then $f \circ g$ is differentiable at $c$ with $(f \circ g)^{'} (c) = f^{'} (g (c)) g^{'} (c)$ , $C^{k}$ maps and multi-index derivative notation in Euclidean space, The standard smooth step function, Every continuous function on an interval has a primitive; two primitives differ by a constant; and $\int_{a}^{b} f = G (b) - G (a)$ for any primitive $G$ )

[F7]

For $0 < r < R$ there is a smooth bump equal to one on ${\overline{B}}_{r} (0)$ and with support inside $B_{R} (0)$ . (A smooth bump between concentric Euclidean balls)

Proof

technique · direct; rotate to the preferred axis, split the sphere with the finite graph partition, and apply the non-stationary and stationary alternatives of the stationary-phase lemma to each chart, with the polar caps contributing the exponent $(n-1)/2$

1.1F2F5given

Reduction to the axis. For $ξ \neq 0$ write $ξ = r ν$ with $r = ∣ ξ ∣$ and $ν \in S^{n - 1}$ , and choose an orthogonal map $R$ with $R ν = e_{n}$ . By the orthogonal invariance of [F2], $\hat{σ} (ξ) = \hat{σ} (r e_{n}) = \int_{S^{n - 1}} e^{- 2 π i r ω \cdot e_{n}} d σ (ω)$ ; the same identity with $ξ = 0$ is trivial and is covered by the bound of [F5] for $r \leq 1$ . It therefore suffices to estimate $J (r) : = \hat{σ} (r e_{n})$ for $r \geq 1$ and to add the trivial bound at small $r$ .

1.2F1F6given

Splitting with the chart partition. Let $(χ_{j})$ be the finite smooth partition of [F1] subordinate to the hemisphere charts, so that $J (r) = \sum_{j} J_{j} (r)$ with $J_{j} (r) = \int_{S^{n - 1}} e^{- 2 π i r ω \cdot e_{n}} χ_{j} (ω) d σ (ω)$ . Each $χ_{j}$ is supported in the image of one chart $X_{i}^{ε}$ , and by [F1] the chart formula writes $J_{j} (r)$ as an integral over the unit ball $B \subseteq R^{n - 1}$ of $e^{- 2 π i r Φ (y)} b_{j} (y) d y$ with $b_{j} : = (χ_{j} \circ X_{i}^{ε}) (1 - ∣ y ∣^{2})^{- 1 / 2}$ and the smooth phase $Φ (y) = y_{n - 1}$ if $i \neq n$ , or $Φ (y) = ε \sqrt{1 - ∣ y ∣^{2}}$ if $i = n$ (the $n$ -th coordinate of the chart being $ε \sqrt{1 - ∣ y ∣^{2}}$ ). The support of $χ_{j} \circ X_{i}^{ε}$ is compact inside $B$ , so $b_{j}$ extends by zero to $C_{c}^{\infty} (R^{n - 1})$ . Coordinate-line product rules in [F6] justify its smoothness.

2.1F3F4step 1.2

The non-stationary charts. If $i \neq n$ , then $Φ (y) = y_{n - 1}$ has $\nabla Φ = e_{n - 1} \neq 0$ everywhere, so the first alternative of [F3] applied to the global phase $- y_{n - 1}$ with $d = n - 1$ and $λ = r$ gives $∣ J_{j} (r) ∣ \leq C_{N} r^{- N}$ for every $N$ ; such patches contribute negligibly for every $N$ .

2.2F4F6step 1.2constructalgebra

A global polar phase. For a polar amplitude $b_{j}$ , choose $0 < r_{0} < r_{1} < 1$ with $supp b_{j} \subset B_{r_{0}} (0)$ . Put $κ (t) = 1 - s_{0} ((t - r_{0}^{2}) / (r_{1}^{2} - r_{0}^{2}))$ and define $a (t) = κ (t) (1 - t)^{- 1 / 2} + 1 - κ (t)$ for $t < r_{1}^{2}$ , and $a (t) = 1$ for $t \geq r_{1}^{2}$ . Rationalization gives $(\sqrt{u})^{'} = 1 / (2 \sqrt{u})$ for $u > 0$ ; repeated product and quotient rules show that $(1 - t)^{- 1 / 2}$ is smooth for $t < 1$ . Thus $a$ is a positive smooth function on $R$ : the first formula is smooth for $t < 1$ , and flatness of the smooth cutoff glues it to one at $r_{1}^{2}$ . Set $A (s) = 1 - \frac{1}{2} \int_{0}^{s} a (t) d t$ . By [F6], $A^{'} = - a / 2$ , so $A$ is smooth. On $s \leq r_{0}^{2}$ its derivative and value at zero agree with $\sqrt{1 - s}$ , hence $A (s) = \sqrt{1 - s}$ . Thus $\tilde{Φ} (y) = ε A (∣ y ∣^{2})$ agrees with the original phase near $supp b_{j}$ , is smooth on all of $R^{n - 1}$ , and satisfies $\nabla \tilde{Φ} (y) = - ε a (∣ y ∣^{2}) y$ . Its only stationary point is zero, with Hessian $- ε I$ .

3.1F3F7step 2.2algebra

Application at the pole. Choose a bump $β$ equal to one near zero and supported inside $B$ by [F7], and a real $M > \sup ∣ b_{j} ∣$ . Both $b_{j} + M β$ and $M β$ have zero in the interior of their supports, since $Re (b_{j} + M β) > 0$ near zero. Apply the stationary alternative [F3] to each amplitude with the global phase $- \tilde{Φ}$ from step 2.2 and $λ = r$ , then subtract their integrals. This gives $∣ J_{j} (r) ∣ \leq C_{j} r^{- (n - 1) / 2}$ without any assumption that zero belongs to the original amplitude support.

4.1F5step 2.1step 3.1algebra

Summation and the small-frequency bound. Summing the finitely many chart contributions of the non-stationary and polar-cap steps gives $∣ \hat{σ} (ξ) ∣ \leq C_{n}^{'} r^{- (n - 1) / 2}$ for $r = ∣ ξ ∣ \geq 1$ . For $∣ ξ ∣ \leq 1$ , [F5] gives $∣ \hat{σ} (ξ) ∣ \leq n λ_{n} (B_{1}^{n}) \leq n λ_{n} (B_{1}^{n}) 2^{(n - 1) / 2} (1 + ∣ ξ ∣)^{- (n - 1) / 2}$ . Taking $C_{n} : = 2^{(n - 1) / 2} \max {C_{n}^{'}, n λ_{n} (B_{1}^{n})}$ yields the asserted bound for every $ξ$ . Since $\overset{ˇ}{σ} (ξ) = \hat{σ} (- ξ)$ and $∣ - ξ ∣ = ∣ ξ ∣$ , the same bound holds for $\overset{ˇ}{σ}$ .

5.1step 1.1step 4.1∎

Conclusion. Step 4.1 proves $∣ \hat{σ} (ξ) ∣ \leq C_{n} (1 + ∣ ξ ∣)^{- (n - 1) / 2}$ for all $ξ$ , and the identity $\overset{ˇ}{σ} (ξ) = \hat{σ} (- ξ)$ transfers it to $\overset{ˇ}{σ}$ . Countable Choice is inherited from the sphere chart and partition suppliers.

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Graph-patch extension family: dispersive and L2 slice bounds

Statement

Assume the hypotheses and notation of Decay of a localized measure on a curved graph patch ( $U, h, a$ with $\det D^{2} h \neq 0$ on $supp a$ , and $μ$ the localized graph measure). Put $K (x^{'}, t) = \overset{ˇ}{μ} (x^{'}, t) = \int e^{2 π i (x^{'} \cdot η + t h (η))} a (η) \sqrt{1 + ∣ \nabla h (η) ∣^{2}} d η$ and $U (t) g (x^{'}) = \int K (x^{'} - y^{'}, t) g (y^{'}) d y^{'}$ for $g \in S (R^{n - 1})$ . Then (i) $∣ K (x^{'}, t) ∣ \leq C_{a} ⟨ t ⟩^{- (n - 1) / 2}$ and hence $∥ U (t) g ∥_{\infty} \leq C_{a} ⟨ t ⟩^{- (n - 1) / 2} ∥ g ∥_{1}$ ; (ii) the partial Fourier transform satisfies $\hat{K (\cdot, t)} (ξ^{'}) = e^{2 π i t h (ξ^{'})} a (ξ^{'}) \sqrt{1 + ∣ \nabla h (ξ^{'}) ∣^{2}}$ for $ξ^{'} \in U$ and zero outside, so $∥ U (t) g ∥_{2} \leq C_{a} ∥ g ∥_{2}$ uniformly in $t$ ; (iii) for every $1 \leq p \leq 2$ , $∥ U (t) g ∥_{p^{'}} \leq C_{a} ⟨ t ⟩^{- (n - 1) (1 / p - 1 / 2)} ∥ g ∥_{p}$ .

Facts & Assumptions

Given: Countable Choice, the data $U, h, a, μ$ of Decay of a localized measure on a curved graph patch with $\det D^{2} h \neq 0$ on $supp a$ , the kernel $K (x^{'}, t) = \overset{ˇ}{μ} (x^{'}, t)$ , and $g \in S (R^{n - 1})$ .

[F1]

Localized decay: $∣ \overset{ˇ}{μ} (x) ∣ \leq C_{a} (1 + ∣ x ∣)^{- (n - 1) / 2}$ for all $x = (x^{'}, x_{n}) \in R^{n}$ , with $C_{a}$ depending on $a, h, n$ . (Decay of a localized measure on a curved graph patch)

[F2]

Convolution and Young bound: $U (t) g (x^{'}) = \int K (x^{'} - y^{'}, t) g (y^{'}) d y^{'}$ converges absolutely when $K (\cdot, t)$ is bounded and $g \in L^{1}$ , with $∥ U (t) g ∥_{\infty} \leq ∥ K (\cdot, t) ∥_{\infty} ∥ g ∥_{1}$ ; the convolution conventions are the published ones, and $⟨ t ⟩ : = (1 + t^{2})^{1 / 2}$ satisfies $⟨ t ⟩ \leq 1 + ∣ t ∣ \leq 1 + ∣ (x^{'}, t) ∣$ . (Decay of a localized measure on a curved graph patch, Complex Lp classes and Euclidean test-function conventions, Holder's inequality for integrals, including the endpoint cases)

[F3]

Fourier conventions and Plancherel: for $h_{k} \in L^{1} (R^{m}) \cap L^{2} (R^{m})$ , ${\overset{ˇ}{h}}_{k} (x) = \int e^{2 π i x \cdot η} h_{k} (η) d η$ is bounded uniformly continuous, and $F_{2} {\overset{ˇ}{h}}_{k} = h_{k}$ in the Plancherel $L^{2}$ sense; the transform is an isometry on $L^{2}$ and $∥ H ∥_{2} = ∥ \hat{H} ∥_{2}$ . For $g \in S$ the product $F_{t} \hat{g}$ is Schwartz, where $F_{t} (η) : = e^{2 π i t h (η)} a (η) \sqrt{1 + ∣ \nabla h (η) ∣^{2}}$ , and $\overset{ˇ}{(F_{t} \hat{g})}$ is its everywhere-defined inverse transform. (Plancherel theorem, The L1 transform is bounded and uniformly continuous, Schwartz convolution and product laws, Schwartz derivatives are integrable, Schwartz space is dense in L2)

[F4]

Riesz–Thorin interpolation: a finite-simple-core operator on sigma-finite spaces with bounds $A$ from $L^{1}$ to $L^{\infty}$ and $B$ from $L^{2}$ to $L^{2}$ satisfies $∥ T f ∥_{p^{'}} \leq A^{2 / p - 1} B^{2 - 2 / p} ∥ f ∥_{p}$ for $1 < p < 2$ , and under countable choice it extends uniquely to the full $L^{p}$ spaces. (Interpolate L1 to Linfinity and L2 to L2 bounds, Conjugate exponents, including the endpoint conventions, A bounded linear map from a dense normed subspace into a Banach space extends uniquely with the same norm, Tonelli's theorem for nonnegative measurable functions on a sigma-finite product)

[F5]

Fubini licenses the multiplier interchange, Fourier inversion and Schwartz stability apply, and the smooth density proof supplies simultaneous $L^{1}$ and $L^{2}$ approximations by truncation and mollification. (Fubini's theorem for L^1 functions on a sigma-finite product, Fourier inversion on Schwartz space, Fourier transform acts continuously on Schwartz space, $C_{c}^{\infty} (R^{n})$ is dense in $L^{p} (R^{n})$ for $1 \leq p < \infty$ )

Proof

technique · direct; read the dispersive bound off the localized decay, identify the slice operator as a Fourier multiplier with a bounded compactly supported symbol, and interpolate the two bounds

1.1F1F2algebra

The dispersive bound. Since $K (x^{'}, t) = \overset{ˇ}{μ} (x^{'}, t)$ and $⟨ t ⟩ \leq 1 + ∣ (x^{'}, t) ∣$ , [F1] gives $∣ K (x^{'}, t) ∣ \leq C_{a} (1 + ∣ (x^{'}, t) ∣)^{- (n - 1) / 2} \leq C_{a} ⟨ t ⟩^{- (n - 1) / 2}$ . Substituting this bound into the convolution of [F2], $∣ U (t) g (x^{'}) ∣ \leq \int ∣ K (x^{'} - y^{'}, t) ∣ ∣ g (y^{'}) ∣ d y^{'} \leq C_{a} ⟨ t ⟩^{- (n - 1) / 2} ∥ g ∥_{1} .$ This proves (i) together with the absolute convergence of the defining integral.

2.1F3F5step 1.1algebra

The multiplier form. For $g \in S (R^{n - 1})$ insert the definition of $K$ into the convolution and apply Fubini (the absolute double integral is $∥ F_{t} ∥_{1} ∥ g ∥_{1} < \infty$ ): $U (t) g (x^{'}) = \int ⁣ ⁣ \int e^{2 π i ((x^{'} - y^{'}) \cdot η + t h (η))} a (η) \sqrt{1 + ∣ \nabla h (η) ∣^{2}} g (y^{'}) d η d y^{'} = \int e^{2 π i x^{'} \cdot η} F_{t} (η) \hat{g} (η) d η,$ where $F_{t} (η) = e^{2 π i t h (η)} a (η) \sqrt{1 + ∣ \nabla h (η) ∣^{2}}$ is supported on $supp a$ and $∣ \nabla h ∣$ is bounded there. Thus $U (t) g$ is the everywhere-defined inverse transform of the Schwartz function $F_{t} \hat{g}$ , so $\hat{U (t) g} = F_{t} \hat{g}$ ; equivalently, the kernel transform identity $\hat{K (\cdot, t)} = F_{t}$ holds pointwise in the Schwartz sense: $F_{t}$ , extended by zero outside $U$ , is smooth with compact support, so $K (\cdot, t) = {\overset{ˇ}{F}}_{t}$ is Schwartz for each fixed $t$ . By the Plancherel isometry [F3], $∥ U (t) g ∥_{2} = ∥ F_{t} \hat{g} ∥_{2} \leq ∥ F_{t} ∥_{\infty} ∥ \hat{g} ∥_{2} = ∥ F_{t} ∥_{\infty} ∥ g ∥_{2} \leq B_{a} ∥ g ∥_{2}$ with $B_{a} : = ∥ F_{0} ∥_{\infty} < \infty$ , uniformly in $t$ because $∣ e^{2 π i t h} ∣ = 1$ ; enlarge the constant $C_{a}$ from step 1.1 to be at least $B_{a}$ . This is (ii).

3.1F4F5step 1.1step 2.1algebra

Interpolation. For a finite simple $g$ of finite-measure support, $g \in L^{1} \cap L^{2}$ . Its convolution equals the Plancherel multiplier: choose smooth compactly supported approximants converging in both $L^{1}$ and $L^{2}$ (truncate the support to balls and mollify; the density proof applies in both norms). The convolution converges uniformly by the bounded kernel, while the multipliers converge in $L^{2}$ by Plancherel, so their limits agree almost everywhere. Thus $U (t)$ defines a compatible complex-linear operator on the finite simple core and satisfies the $L^{1} \to L^{\infty}$ bound $A (t) = C_{a} ⟨ t ⟩^{- (n - 1) / 2}$ by step 1.1 and the $L^{2} \to L^{2}$ bound $B = C_{a}$ by step 2.1. Applying the Riesz–Thorin corollary [F4] with $1 < p < 2$ gives $∥ U (t) g ∥_{p^{'}} \leq A (t)^{2 / p - 1} C_{a}^{2 - 2 / p} ∥ g ∥_{p} = C_{a}^{2 / p - 1 + 2 - 2 / p} ⟨ t ⟩^{- \frac{n - 1}{2} (2 / p - 1)} ∥ g ∥_{p} = C_{a} ⟨ t ⟩^{- (n - 1) (1 / p - 1 / 2)} ∥ g ∥_{p},$ because $2 / p - 1 = 2 (1 / p - 1 / 2)$ ; the endpoint cases $p = 1$ and $p = 2$ are the bounds of steps 1.1 and 2.1. The unique compatible bounded extensions to the full $L^{p}$ spaces exist by [F4]. This proves (iii).

4.1step 1.1step 2.1step 3.1∎

Conclusion. Step 1.1 gives the kernel decay and the $L^{1} \to L^{\infty}$ slice bound, step 2.1 identifies the slice operator as the Fourier multiplier by the bounded symbol $F_{t}$ and gives the uniform $L^{2}$ bound, and step 3.1 interpolates to the full range $1 \leq p \leq 2$ with the exponent $(n - 1) (1 / p - 1 / 2)$ . All constants depend only on $a, h, n$ and not on $t$ .

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Knapp necessary condition for spherical L2 restriction

Statement

Assume Countable Choice, let $n \geq 2$ , and take $p, q \in [1, \infty]$ . (a) If there is $C < \infty$ with $∥ \hat{f} ∥_{L^{2} (σ)} \leq C ∥ f ∥_{L^{p} (R^{n})}$ for all $f \in S (R^{n})$ , then $p \leq 2 (n + 1) / (n + 3)$ . Equivalently, if $E : L^{2} (S^{n - 1}) \to L^{q} (R^{n})$ satisfies $∥ E g ∥_{q} \leq C ∥ g ∥_{L^{2} (σ)}$ for all $g \in L^{2} (σ)$ , then $q \geq 2 (n + 1) / (n - 1)$ . (b) The necessity is exhibited by the cap data $g = 1_{C_{δ}}$ and the limit $δ ↓ 0$ .

Facts & Assumptions

Given: Countable Choice, $n \geq 2$ , $δ \in (0, 1]$ , the cap $C_{δ} = C_{δ} (e_{n}) = {ω \in S^{n - 1} : 1 - ω \cdot e_{n} \leq δ^{2}}$ , and the tube $T_{δ} = {ξ : ∣ ξ_{n} ∣ \leq c_{n} δ^{- 2}, ∣ ξ_{j} ∣ \leq c_{n} δ^{- 1} (j < n)}$ where $a_{n} > 0$ is a constant furnished by Cap wave packets concentrate on the dual tube and $c_{n} = a_{n} / \sqrt{n - 1}$ . For this box, $∣ (ξ_{1}, \dots, ξ_{n - 1}) ∣ \leq \sqrt{n - 1} c_{n} δ^{- 1} = a_{n} δ^{- 1}$ and $∣ ξ_{n} ∣ \leq c_{n} δ^{- 2} \leq a_{n} δ^{- 2}$ , so it lies in the cylindrical concentration tube.

[F1]

Duality: for $1 < p < \infty$ the restriction estimate $∥ \hat{f} ∥_{L^{2} (σ)} \leq C ∥ f ∥_{L^{p}}$ for all Schwartz $f$ is equivalent to the extension estimate $∥ E g ∥_{L^{p^{'}}} \leq C ∥ g ∥_{L^{2} (σ)}$ for all $g \in L^{2} (σ)$ , with the same least constant. (Restriction and extension estimates are dual, Conjugate exponents, including the endpoint conventions)

[F2]

Cap and tube scales: $c^{'} δ^{n - 1} \leq σ (C_{δ}) \leq C^{'} δ^{n - 1}$ and $λ_{n} (T_{δ}) = (2 c_{n})^{n} δ^{- (n + 1)}$ for $δ \in (0, 1]$ , with constants depending only on $n$ . (Spherical cap and dual slab scales)

[F3]

Cap concentration on the cylindrical tube of the supplier, and hence on the box specified in Given: for $g = 1_{C_{δ}}$ one has $∣ E g (x) ∣ \geq \frac{1}{2} σ (C_{δ})$ for every $x \in T_{δ}$ , hence $∥ E g ∥_{q}^{q} \geq (\frac{1}{2} σ (C_{δ}))^{q} λ_{n} (T_{δ})$ for every $1 \leq q < \infty$ ; and $∥ g ∥_{L^{2} (σ)} = σ (C_{δ})^{1 / 2}$ . (Cap wave packets concentrate on the dual tube, The nonnegative Lebesgue integral, Fourier pairing for a finite measure and Schwartz data, Explicit compactly supported smooth cutoffs, Monotone convergence for the integral, Differentiation under the integral sign)

[F4]

Positive-base real powers satisfy the exponent, product and iterated-power laws and are defined by $δ^{a} = \exp (a \log δ)$ . The logarithm is the inverse of the exponential, and $\exp t \to \infty$ as $t \to \infty$ , while $\exp t \to 0$ as $t \to - \infty$ . Thus if $δ^{a} \leq K δ^{b}$ for all $δ \in (0, 1]$ and $K > 0$ , then $a \geq b$ : otherwise put $δ = \exp (- k)$ for large $k$ to obtain $δ^{a - b} = \exp (k (b - a)) \to \infty$ , contradicting the bound. (The exponent, product, quotient, and iterated-power laws for positive real bases and real exponents, Real powers for positive bases, with the zero-base positive-exponent convention, The natural logarithm as the inverse of the exponential function, The exponential tends to $+ \infty$ at $+ \infty$ and to $0$ at $- \infty$ )

[F5]

The finite-measure pairing, smooth cutoffs, monotone convergence and compact-frequency differentiation are available. (Fourier pairing for a finite measure and Schwartz data, Explicit compactly supported smooth cutoffs, Monotone convergence for the integral, Differentiation under the integral sign)

Proof

technique · direct; evaluate the assumed extension bound on the cap data, insert the two exact scales, and let $\delta$ tend to $0$; the restriction form follows by duality

1.1F3algebra

If $q = \infty$ , the necessary inequality $q \geq 2 (n + 1) / (n - 1)$ is automatic. For $1 \leq q < \infty$ , assume $∥ E g ∥_{q} \leq C ∥ g ∥_{L^{2} (σ)}$ for all $g \in L^{2} (σ)$ and let $g = 1_{C_{δ}} \in L^{2} (σ)$ . By [F3], $∥ g ∥_{L^{2} (σ)} = σ (C_{δ})^{1 / 2}$ and $∥ E g ∥_{q}^{q} \geq (\frac{1}{2} σ (C_{δ}))^{q} λ_{n} (T_{δ})$ . Combining with the assumed bound, $\frac{1}{2} σ (C_{δ}) λ_{n} (T_{δ})^{1 / q} \leq ∥ E g ∥_{q} \leq C σ (C_{δ})^{1 / 2} .$

2.1F2step 1.1algebra

Inserting the scales. Dividing by $σ (C_{δ})^{1 / 2}$ and inserting [F2], $\frac{1}{2} c^{' 1 / 2} (2 c_{n})^{n / q} δ^{(n - 1) / 2 - (n + 1) / q} \leq \frac{1}{2} σ (C_{δ})^{1 / 2} λ_{n} (T_{δ})^{1 / q} \leq C .$ Thus there is a constant $K$ depending only on $n, q$ with $δ^{a} \leq K$ for all $δ \in (0, 1]$ , where $a : = (n - 1) / 2 - (n + 1) / q$ .

3.1F4step 2.1algebra

Forcing the exponent. Applying [F4] with $b = 0$ to the inequality $δ^{a} \leq K δ^{0}$ gives $a \geq 0$ , that is $(n - 1) / 2 \geq (n + 1) / q$ , hence $q \geq 2 (n + 1) / (n - 1)$ : no extension bound can hold for smaller $q$ .

4.1F1F2F3F5step 3.1algebra

Suppose the restriction estimate holds at exponent $p$ . For $1 < p < \infty$ , [F1] and step 3.1 imply $p^{'} \geq 2 (n + 1) / (n - 1)$ , hence $p \leq 2 (n + 1) / (n + 3)$ . At $p = 1$ the necessary inequality is automatic. At $p = \infty$ , apply the pairing identity [F5] with the finite positive measure $μ = 1_{C_{δ}} σ$ and a Schwartz cutoff equal to one near $S^{n - 1}$ as its frequency factor. Thus an $L^{\infty}$ restriction bound would imply $∣ \int E g \overline{F} ∣ \leq C ∥ g ∥_{2} ∥ F ∥_{\infty}$ on compactly supported smooth $F$ . For $g = 1_{C_{δ}}$ , $h = E g$ is smooth (differentiate its finite compact-frequency integral), and the tests $F = χ_{R} h / (∣ h ∣^{2} + ϵ^{2})^{1 / 2}$ have supremum at most one. Choose $0 \leq χ_{R} \leq 1$ equal to one on the radius- $R$ ball. The nonnegative pairing integrand therefore bounds the integral over that ball by $C ∥ g ∥_{2}$ . Let the balls increase to $R^{n}$ , then let $ϵ ↓ 0$ , by monotone convergence to obtain $∥ E g ∥_{1} \leq C ∥ g ∥_{2}$ , contradicted by the cap lower bound at $q = 1$ . Thus $p = \infty$ is impossible too.

4.2F2F3F4step 2.1step 3.1

The exhibited family. The data witnessing the necessity are exactly the functions $g_{δ} = 1_{C_{δ}}$ , $δ \in (0, 1]$ : for any $q < 2 (n + 1) / (n - 1)$ the quotient $∥ E g_{δ} ∥_{q} / ∥ g_{δ} ∥_{L^{2} (σ)} \geq \frac{1}{2} σ (C_{δ})^{1 / 2} λ_{n} (T_{δ})^{1 / q}$ is unbounded along $δ = \exp (- k) ↓ 0$ by [F4] and steps 1.1–3.1, so no finite constant works. This is clause (b).

5.1step 1.1step 3.1step 4.1step 4.2∎

Conclusion. Steps 1.1–4.1 prove that any $L^{2} \to L^{q}$ extension bound forces $q \geq 2 (n + 1) / (n - 1)$ , step 4.1 transfers this to the restriction form $p \leq 2 (n + 1) / (n + 3)$ through the duality lemma, and step 4.2 exhibits the cap family and the limit $δ ↓ 0$ .

LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-6.1-sol)Open item page →

Stein-Tomas TT-star bound from fractional integration

Statement

Assume Countable Choice. Let $n \geq 2$ and let $U, h, a, μ$ be as in Graph-patch extension family: dispersive and L2 slice bounds, with $\det D^{2} h \neq 0$ on $supp a$ . Set $p = 2 (n + 1) / (n + 3)$ , so that $p^{'} = 2 (n + 1) / (n - 1)$ and $1 / p - 1 / p^{'} = 2 / (n + 1)$ . Then $∥ f * \overset{ˇ}{μ} ∥_{L^{p^{'}} (R^{n})} \leq C_{a} ∥ f ∥_{L^{p} (R^{n})}$ for every $f \in S (R^{n})$ .

Facts & Assumptions

Given: Countable Choice, $n \geq 2$ , the graph-patch data $U, h, a$ with $\det D^{2} h \neq 0$ on $supp a$ , the localized measure $μ$ , its transform $\overset{ˇ}{μ}$ , the exponent $p = 2 (n + 1) / (n + 3)$ and $f \in S (R^{n})$ ; write $f_{s} (x^{'}) : = f (x^{'}, s)$ for the slice at height $s$ .

[F1]

Slice family: with $K (x^{'}, t) = \overset{ˇ}{μ} (x^{'}, t)$ and $U (t) g (x^{'}) = \int K (x^{'} - y^{'}, t) g (y^{'}) d y^{'}$ , one has for $1 \leq p \leq 2$ the bound $∥ U (t) g ∥_{L^{p^{'}} (R^{n - 1})} \leq C_{a} ⟨ t ⟩^{- (n - 1) (1 / p - 1 / 2)} ∥ g ∥_{L^{p} (R^{n - 1})}$ , uniformly in $t$ , and $⟨ z ⟩^{- (1 - α)} \leq ∣ z ∣^{- (1 - α)}$ for $z \neq 0$ for the order $α : = 2 / (n + 1)$ . (Graph-patch extension family: dispersive and L2 slice bounds, Conjugate exponents, including the endpoint conventions)

[F2]

Minkowski's integral inequality and Tonelli: for measurable $F$ with $\int_{Y} ∥ F (\cdot, y) ∥_{L^{r} (X)} d ν (y) < \infty$ one has $∥ \int_{Y} ∣ F (\cdot, y) ∣ d ν ∥_{L^{r} (X)} \leq \int_{Y} ∥ F (\cdot, y) ∥_{L^{r} (X)} d ν (y)$ for $1 \leq r < \infty$ ; iterated integrals over sigma-finite products agree for nonnegative integrands. (Minkowski's integral inequality, Tonelli's theorem for nonnegative measurable functions on a sigma-finite product)

[F3]

Hardy–Littlewood–Sobolev in one dimension: for $0 < α < 1$ and $1 < p < 1 / α$ with $1 / p^{'} = 1 / p - α$ , the unit-normalized Riesz potential $I_{α} g (t) = \int ∣ t - s ∣^{α - 1} g (s) d s$ satisfies $∥ I_{α} g ∥_{L^{p^{'}} (R)} \leq C_{α, p} ∥ g ∥_{L^{p} (R)}$ for $g \in L^{p} (R)$ ; the kernel convention is $K_{α} (z) = ∣ z ∣^{α - 1}$ for $z \neq 0$ and $K_{α} (0) = 0$ . (Hardy–Littlewood–Sobolev fractional integration inequality, Riesz potential of order alpha)

[F4]

Complex $L^{p}$ conventions: norms on complex classes, Tonelli for nonnegative measurable functions, and the modulus estimates used to pass from complex functions to their pointwise moduli. (Complex Lp classes and Euclidean test-function conventions, Complex Holder, Minkowski, and the quotient norm)

[F5]

Fubini applies to absolutely integrable complex product kernels, and Schwartz decay gives integrability of all slices and arbitrarily rapid decay of their norms in the remaining coordinate. (Fubini's theorem for L^1 functions on a sigma-finite product, Schwartz derivatives are integrable)

Proof

technique · direct; write the full convolution as a superposition of slice operators, apply Minkowski in the space variable, and recognize the resulting one-dimensional kernel as a Riesz potential of order $2/(n+1)$

1.1F1F5givenalgebra

The slice superposition. The bounded kernel and $f \in L^{1}$ make the full integral absolutely convergent; Fubini therefore licenses splitting the variable $y = (y^{'}, s) \in R^{n - 1} \times R$ and writing $x = (x^{'}, t)$ , the defining convolution and [F1] give, at every point, $f * \overset{ˇ}{μ} (x^{'}, t) = \int_{R} (\int_{R^{n - 1}} f (x^{'} - y^{'}, t - s) K (y^{'}, s) d y^{'}) d s = \int_{R} U (s) f_{t - s} (x^{'}) d s .$

2.1F1F2F4F5step 1.1

Minkowski in the slice variable. Fix $t$ . By [F2] applied in the space variable $x^{'}$ with $ν$ the Lebesgue measure in $s$ , $∥ \int_{R} U (s) f_{t - s} d s ∥_{L^{p^{'}} (R^{n - 1})} \leq \int_{R} ∥ U (s) f_{t - s} ∥_{L^{p^{'}} (R^{n - 1})} d s .$ For Schwartz $f$ , $g (τ) = ∥ f_{τ} ∥_{p}$ decays faster than any prescribed power (bound $∣ f (x', τ) ∣$ by $C_{N} (1 + ∣ x' ∣)^{- N} (1 + ∣ τ ∣)^{- N}$ ). Thus the slice estimate of [F1] bounds the right side by $C \int g < \infty$ , licensing [F2] for every $t$ ; this is the Minkowski inequality for the complex-valued measurable integrand of [F4].

3.1F1F2F4step 2.1algebra

The slice decay. For the exponent $p = 2 (n + 1) / (n + 3)$ one has $1 / p - 1 / 2 = 1 / (n + 1)$ , so [F1] gives $∥ U (s) f_{t - s} ∥_{L^{p^{'}}} \leq C_{a} ⟨ s ⟩^{- (n - 1) / (n + 1)} ∥ f_{t - s} ∥_{L^{p} (R^{n - 1})}$ . Substituting into step 2.1 and substituting $τ = t - s$ , $∥ \int_{R} U (s) f_{t - s} d s ∥_{L^{p^{'}} (R^{n - 1})} \leq C_{a} \int_{R} ⟨ t - τ ⟩^{- β} g (τ) d τ, g (τ) : = ∥ f_{τ} ∥_{L^{p} (R^{n - 1})}, β : = \frac{n - 1}{n + 1} .$ The function $g$ is measurable and nonnegative and, by Tonelli, $∥ g ∥_{L^{p} (R)}^{p} = \int_{R} \int_{R^{n - 1}} ∣ f (x^{'}, τ) ∣^{p} d x^{'} d τ = ∥ f ∥_{L^{p} (R^{n})}^{p}$ , so $g \in L^{p} (R)$ and $∥ g ∥_{p} = ∥ f ∥_{p}$ .

4.1F1F3step 3.1algebra

Fractional integration. Let $α : = 1 - β = 2 / (n + 1) \in (0, 1)$ and compare kernels: $⟨ t - τ ⟩^{- (1 - α)} \leq ∣ t - τ ∣^{- (1 - α)} = K_{α} (t - τ)$ for $τ \neq t$ by [F1]. Hence the function $h (t) : = ∥ (f * \overset{ˇ}{μ}) (\cdot, t) ∥_{L^{p^{'}} (R^{n - 1})}$ satisfies $h (t) \leq C_{a} 2^{1 - α} \int_{R} K_{α} (t - τ) g (τ) d τ = C_{a} 2^{1 - α} I_{α} g (t)$ for almost every $t$ ; the single diagonal point $τ = t$ has zero measure and does not affect the comparison. The HLS hypotheses hold: $0 < α < 1$ because $n \geq 2$ , and $1 < p < 1 / α = (n + 1) / 2$ because $p = 2 (n + 1) / (n + 3)$ and $n > 1$ . By [F3] applied in dimension one, $∥ \int_{R} U (s) f_{t - s} d s ∥_{L^{p^{'}} (R^{n})} = ∥ h ∥_{L^{p^{'}} (R)} \leq C_{a} 2^{1 - α} C_{α, p} ∥ g ∥_{L^{p} (R)} = C_{a}^{'} ∥ f ∥_{L^{p} (R^{n})},$ with $C_{a}^{'}$ depending on $a, h, n$ .

5.1step 1.1step 2.1step 3.1step 4.1∎

Conclusion. Steps 1.1–4.1 show that for the endpoint exponent $p = 2 (n + 1) / (n + 3)$ the convolution $f * \overset{ˇ}{μ}$ lies in $L^{p^{'}} (R^{n})$ with norm controlled by $∥ f ∥_{p}$ , the constants depending only on $a, h, n$ . The exponent identity $1 / p - 1 / p^{'} = 2 / (n + 1) = 1 - β$ used above is exactly the conjugacy of $p$ and $p^{'}$ .

TheoremStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Stein-Tomas spherical restriction theorem

Statement

Assume Countable Choice and let $n \geq 2$ . Let $σ$ be the polar surface measure on $S^{n - 1}$ and set $p_{0} = 2 (n + 1) / (n + 3)$ , so that $p_{0}^{'} = 2 (n + 1) / (n - 1)$ . Then (a) there is $C_{n} < \infty$ with $∥ \hat{f} ∥_{L^{2} (σ)} \leq C_{n} ∥ f ∥_{L^{p_{0}} (R^{n})}$ for all $f \in S (R^{n})$ ; (b) $R_{0}$ extends uniquely to a bounded linear $R : L^{p} (R^{n}) \to L^{2} (σ)$ for every $1 \leq p \leq p_{0}$ , and $E : L^{2} (σ) \to L^{p^{'}} (R^{n})$ is its adjoint under the $L^{p}$ – $L^{p^{'}}$ and $L^{2} (σ)$ pairings with the same norm; equivalently $E : L^{2} (σ) \to L^{q} (R^{n})$ is bounded for every $q \geq q_{0} = 2 (n + 1) / (n - 1)$ ; (c) the result is sharp: (a) fails for $p > p_{0}$ , and no $L^{2} \to L^{q}$ bound holds for $q < q_{0}$ .

Here and in the sharpness clause the exponents belong to $[1, \infty]$ .

Facts & Assumptions

Given: Countable Choice, $n \geq 2$ , the polar surface measure $σ$ on $S^{n - 1}$ , the exponents $p_{0} = 2 (n + 1) / (n + 3)$ , $p_{0}^{'} = 2 (n + 1) / (n - 1)$ , and the operators $R_{0}, E$ of Fourier restriction and adjoint extension operators.

[F1]

Sphere charts and partition: the polar measure is written as a finite sum of localizations $χ_{j} σ$ supported in the images of the graph charts $X_{j} (y) = (y, h_{j} (y))$ with $h_{j} = \pm \sqrt{1 - ∣ y ∣^{2}}$ and $\det D^{2} h_{j} \neq 0$ on the chart domain; the chart density is $(1 - ∣ y ∣^{2})^{- 1 / 2}$ . (Sphere graph charts, surface density, and a finite partition)

[F2]

Graph-patch endpoint bound: for each chart localization $μ_{j} = χ_{j} σ$ as a graph measure with $\det D^{2} h_{j} \neq 0$ one has $∥ f * {\overset{ˇ}{μ}}_{j} ∥_{p_{0}^{'}} \leq C_{j} ∥ f ∥_{p_{0}}$ for all Schwartz $f$ . (Stein-Tomas TT-star bound from fractional integration, Graph-patch extension family: dispersive and L2 slice bounds)

[F3]

TT*: $∥ \hat{f} ∥_{L^{2} (σ)}^{2} = ⟨ f * \overset{ˇ}{σ}, f ⟩ \leq ∥ f * \overset{ˇ}{σ} ∥_{p_{0}^{'}} ∥ f ∥_{p_{0}}$ ; the identity reduces the restriction bound to a convolution bound. (TT-star reduces extension to convolution with the surface-measure transform)

[F4]

Duality and extensions: for $1 < p < \infty$ , the restriction estimate at $p$ is equivalent to the extension estimate at $p^{'}$ with the same constant; a bounded linear map on a dense subspace of a normed space with Banach target has a unique bounded extension with the same norm; $S$ is dense in every $L^{p}$ , $1 \leq p < \infty$ . (Restriction and extension estimates are dual, A bounded linear map from a dense normed subspace into a Banach space extends uniquely with the same norm, $C_{c}^{\infty} (R^{n})$ is dense in $L^{p} (R^{n})$ for $1 \leq p < \infty$ )

[F5]

For a measurable $h \in L^{q_{0}} \cap L^{\infty}$ and $q_{0} \leq q < \infty$ , integration of $∣ h ∣^{q} \leq ∥ h ∥_{\infty}^{q - q_{0}} ∣ h ∣^{q_{0}}$ gives $∥ h ∥_{q} \leq ∥ h ∥_{q_{0}}^{q_{0} / q} ∥ h ∥_{\infty}^{1 - q_{0} / q}$ . At $q = \infty$ use the given supremum bound. (Complex Lp classes and Euclidean test-function conventions, Translation, modulation, linear dilation and reflection laws, A linear map $T$ of $R^{n}$ sends Lebesgue measurable sets to Lebesgue measurable sets, with $λ_{n} (T [E]) = ∣ \det T ∣ λ_{n} (E)$ when $T$ is invertible and $T [E]$ Lebesgue null when it is not)

[F6]

Sharpness: every restriction estimate forces $p \leq p_{0}$ , and every extension estimate forces $q \geq q_{0}$ . (Knapp necessary condition for spherical L2 restriction, Conjugate exponents, including the endpoint conventions)

[F7]

Orthogonal coordinate changes preserve Lebesgue norms and Schwartz space; translations of frequency surfaces modulate their inverse transforms. (Translation, modulation, linear dilation and reflection laws, A linear map $T$ of $R^{n}$ sends Lebesgue measurable sets to Lebesgue measurable sets, with $λ_{n} (T [E]) = ∣ \det T ∣ λ_{n} (E)$ when $T$ is invertible and $T [E]$ Lebesgue null when it is not)

Proof

technique · direct; localize the sphere into curved graph patches, sum the endpoint $TT^*$ bounds, dualize to the full range, and invoke the Knapp obstruction for sharpness

1.1F1F2F7algebra

Summing the graph-patch bounds. Let $σ = \sum_{j} μ_{j}$ with $μ_{j} = χ_{j} σ$ the finite chart localization of [F1]. For $f \in S (R^{n})$ , rotate each patch to graph coordinates; orthogonal changes preserve Lebesgue norms and Schwartz space, and chart translations only modulate the data by a unit character. Thus [F2] applies in ambient coordinates. Now $\overset{ˇ}{σ} = \sum_{j} {\overset{ˇ}{μ}}_{j}$ and hence $f * \overset{ˇ}{σ} = \sum_{j} f * {\overset{ˇ}{μ}}_{j}$ as everywhere-defined bounded continuous functions. By [F2] and the triangle inequality for $L^{p_{0}^{'}}$ , $∥ \sum_{j} f * {\overset{ˇ}{μ}}_{j} ∥_{p_{0}^{'}} \leq \sum_{j} ∥ f * {\overset{ˇ}{μ}}_{j} ∥_{p_{0}^{'}} \leq (\sum_{j} C_{j}) ∥ f ∥_{p_{0}} = : C_{n} ∥ f ∥_{p_{0}} .$ This is the convolution bound at the endpoint.

2.1F3step 1.1algebra

The restriction bound at $p_{0}$ . Applying the $T T^{*}$ identity [F3] to the sum of step 1.1, $∥ \hat{f} ∥_{L^{2} (σ)}^{2} = ⟨ f * \overset{ˇ}{σ}, f ⟩ \leq ∥ f * \overset{ˇ}{σ} ∥_{p_{0}^{'}} ∥ f ∥_{p_{0}} \leq C_{n} ∥ f ∥_{p_{0}}^{2},$ so $∥ \hat{f} ∥_{L^{2} (σ)} \leq C_{n}^{1 / 2} ∥ f ∥_{p_{0}}$ , which is (a).

3.1F4F5step 2.1algebra

By [F4], the endpoint restriction estimate gives $∥ E g ∥_{q_{0}} \leq C_{n}^{1 / 2} ∥ g ∥_{2}$ . Also $∥ E g ∥_{\infty} \leq σ (S^{n - 1})^{1 / 2} ∥ g ∥_{2}$ by the definition. Applying [F5] with $θ = q_{0} / q$ gives $∥ E g ∥_{q} \leq (C_{n}^{1 / 2})^{θ} σ (S^{n - 1})^{(1 - θ) / 2} ∥ g ∥_{2}$ for $q_{0} \leq q < \infty$ , and the supremum estimate gives $q = \infty$ . For $1 < p \leq p_{0}$ , duality gives the unique restriction extensions with the same norms. At $p = 1$ , $∥ R_{0} f ∥_{2} \leq σ (S^{n - 1})^{1 / 2} ∥ f ∥_{1}$ directly; density and completeness give its unique extension. The pairing $\int E g \overline{f} = \int g \overline{R f} d σ$ extends from Schwartz tests by Hölder, also for $p = 1$ . Conversely, norm testing in $L^{2} (σ)$ shows that each restriction norm is bounded by its extension norm; testing the extension against $L^{1}$ functions gives the reverse inequality at $p = 1$ . Thus the adjoint pairing and equality of norms hold throughout the asserted range.

4.1F6step 3.1

Sharpness. By [F6], the existence of a restriction estimate at exponent $p$ forces $p \leq p_{0}$ , and the existence of an $L^{2} (σ) \to L^{q}$ extension estimate forces $q \geq q_{0}$ ; the Knapp cap family $g = 1_{C_{δ}}$ with $δ ↓ 0$ exhibits both failures. This proves (c).

5.1step 1.1step 2.1step 3.1step 4.1∎

Conclusion. Steps 1.1–2.1 prove the endpoint restriction bound (a), step 3.1 gives the full range and the adjoint formulation (b), and step 4.1 records sharpness (c) from the Knapp necessary condition.

CorollaryStatement: Literature-sourcedProof: AI-adaptedprecheck passOpen item page →

Stein-Tomas for compact hypersurfaces with nonzero curvature

Statement

Assume Countable Choice. Let $S \subseteq R^{n}$ ( $n \geq 2$ ) be a compact embedded $C^{\infty}$ hypersurface with everywhere nonvanishing extrinsic Gaussian curvature and surface measure $σ$ . Set $p_{0} = 2 (n + 1) / (n + 3)$ and $q_{0} = 2 (n + 1) / (n - 1)$ . Then there is $C_{S} < \infty$ , depending on $S$ , such that $∥ \hat{f} ∥_{L^{2} (σ)} \leq C_{S} ∥ f ∥_{L^{p_{0}} (R^{n})}$ for all $f \in S (R^{n})$ ; $R_{0}$ extends uniquely to a bounded linear $R : L^{p} (R^{n}) \to L^{2} (σ)$ for every $1 \leq p \leq p_{0}$ ; and $E : L^{2} (σ) \to L^{q} (R^{n})$ is bounded for every $q \geq q_{0}$ .

Facts & Assumptions

Given: Countable Choice, $n \geq 2$ , a compact embedded $C^{\infty}$ hypersurface $S \subseteq R^{n}$ with everywhere nonvanishing extrinsic Gaussian curvature and surface measure $σ$ , the exponents $p_{0}, q_{0}$ , and the operators $R_{0}, E$ .

[F1]

Every smooth embedded Euclidean hypersurface admits smooth graph charts after rigid motions. Compact sets admit finite graph localization, with smooth compactly supported nonnegative weights summing to one. (Smooth Euclidean hypersurface graphs and compact localization)

[F2]

Euclidean shape operators and extrinsic Gaussian curvature have their usual meanings; local normal reversal preserves curvature nonvanishing. On a graph the curvature determinant equals the Hessian determinant divided by the positive graph factor. (Euclidean hypersurface normals, shape operators and curvature, Smooth Euclidean hypersurface graphs and compact localization, Shape operator and Gauss-Kronecker curvature of a graph)

[F3]

Finite cover and partition: the compact hypersurface is covered by finitely many relatively open graph pieces of [F1] with subordinate nonnegative smooth functions $χ_{j}$ , $\sum_{j} χ_{j} = 1$ , each $χ_{j}$ compactly supported in its piece; the localized measures $μ_{j} : = χ_{j} σ$ are graph-patch localizations of the form treated by the patch decay and slice estimates. (Compact curved hypersurfaces admit a finite curved graph cover, Decay of a localized measure on a curved graph patch, Chart and partition independence of surface measure)

[F4]

Endpoint patch bound: for every such localization $μ_{j}$ one has $∥ f * {\overset{ˇ}{μ}}_{j} ∥_{p_{0}^{'}} \leq C_{j} ∥ f ∥_{p_{0}}$ for all Schwartz $f$ , with $C_{j}$ depending on the patch data. (Stein-Tomas TT-star bound from fractional integration)

[F5]

TT*: $∥ \hat{f} ∥_{L^{2} (σ)}^{2} = ⟨ f * \overset{ˇ}{σ}, f ⟩ \leq ∥ f * \overset{ˇ}{σ} ∥_{p_{0}^{'}} ∥ f ∥_{p_{0}}$ . (TT-star reduces extension to convolution with the surface-measure transform)

[F6]

Duality, extension and interpolation: for $1 < p < \infty$ , restriction at $p$ is equivalent to extension at $p^{'}$ with the same constant and adjoint identification; bounded maps on dense subspaces have unique bounded extensions; and a map bounded $L^{2} \to L^{p_{0}^{'}}$ and $L^{2} \to L^{\infty}$ with constants $C$ and $σ (S)^{1 / 2}$ satisfies $∥ E g ∥_{q} \leq C^{θ} σ (S)^{(1 - θ) / 2} ∥ g ∥_{2}$ for $q \geq q_{0}$ with $1 / q = θ / p_{0}^{'}$ . (Restriction and extension estimates are dual, A bounded linear map from a dense normed subspace into a Banach space extends uniquely with the same norm, Interpolate L1 to Linfinity and L2 to L2 bounds, Riesz–Thorin estimate on the finite simple core, Conjugate exponents, including the endpoint conventions, Complex Lp classes and Euclidean test-function conventions)

[F7]

Euclidean smooth density, complex completeness and bounded extension apply at $p = 1$ . Rigid coordinate changes preserve the required norms, and translations of surface patches give modulation of the data. ( $C_{c}^{\infty} (R^{n})$ is dense in $L^{p} (R^{n})$ for $1 \leq p < \infty$ , Complex Lp completeness and almost-everywhere subsequences, Translation, modulation, linear dilation and reflection laws, A linear map $T$ of $R^{n}$ sends Lebesgue measurable sets to Lebesgue measurable sets, with $λ_{n} (T [E]) = ∣ \det T ∣ λ_{n} (E)$ when $T$ is invertible and $T [E]$ Lebesgue null when it is not)

Proof

technique · direct; cover the compact curved hypersurface by finitely many curved graph patches, sum the endpoint patch bounds, and conclude by $TT^*$, duality and interpolation

1.1F1F2F3

Finite graph cover. The local reduction of [F1] and the nondegeneracy of the curvature via [F2] produce, at every point, a curved graph chart; the argument of the finite-graph-cover lemma (compactness plus local construction) selects finitely many such charts covering $S$ . The compact localization construction of [F1] uses the local graph normals; their sign-independent nonvanishing curvature in [F2] supplies the curved charts without needing a global normal field. Write the resulting pieces as $S_{1}, \dots, S_{m}$ with graphing functions $h_{j}$ satisfying $\det D^{2} h_{j} \neq 0$ and localizations $μ_{j} = χ_{j} σ$ as in [F3].

2.1F4F7step 1.1algebra

Endpoint convolution bound. Each patch may be rotated to graph coordinates; its translation contributes a modulation of the data. These operations preserve all relevant norms, so [F4] applies in the original coordinates. Since $\overset{ˇ}{σ} = \sum_{j} {\overset{ˇ}{μ}}_{j}$ and $f * \overset{ˇ}{σ} = \sum_{j} f * {\overset{ˇ}{μ}}_{j}$ , the patch estimates [F4] and the triangle inequality give $∥ f * \overset{ˇ}{σ} ∥_{p_{0}^{'}} \leq (\sum_{j} C_{j}) ∥ f ∥_{p_{0}} = : C_{S} ∥ f ∥_{p_{0}}$ for every Schwartz $f$ .

3.1F5step 2.1algebra

The restriction bound. By the $T T^{*}$ identity [F5] and step 2.1, $∥ \hat{f} ∥_{L^{2} (σ)}^{2} = ⟨ f * \overset{ˇ}{σ}, f ⟩ \leq ∥ f * \overset{ˇ}{σ} ∥_{p_{0}^{'}} ∥ f ∥_{p_{0}} \leq C_{S} ∥ f ∥_{p_{0}}^{2}$ , hence $∥ \hat{f} ∥_{L^{2} (σ)} \leq C_{S}^{1 / 2} ∥ f ∥_{p_{0}}$ for every $f \in S (R^{n})$ .

4.1F6F7step 3.1algebra

The range and the extension. By [F6] the restriction bound of step 3.1 gives a unique bounded extension $R : L^{p_{0}} \to L^{2} (σ)$ and the dual extension estimate $∥ E g ∥_{p_{0}^{'}} \leq C_{S}^{1 / 2} ∥ g ∥_{L^{2} (σ)}$ ; moreover $∣ E g ∣ \leq ∥ g ∥_{L^{1} (σ)} \leq σ (S)^{1 / 2} ∥ g ∥_{L^{2} (σ)}$ , so $E : L^{2} (σ) \to L^{\infty}$ is bounded. For finite $q \geq q_{0}$ , integrating $∣ E g ∣^{q} \leq ∥ E g ∥_{\infty}^{q - q_{0}} ∣ E g ∣^{q_{0}}$ gives $E : L^{2} (σ) \to L^{q}$ for every $q \geq q_{0}$ , and dualizing as in [F6] extends $R_{0}$ uniquely to a bounded $R : L^{p} \to L^{2} (σ)$ for every $1 < p \leq p_{0}$ . At $p = 1$ , $∥ R_{0} f ∥_{2} \leq σ (S)^{1 / 2} ∥ f ∥_{1}$ directly, and smooth density with completeness gives the unique extension. Hölder extends the integral adjoint pairing to $L^{1}$ as well. Thus the full asserted range holds.

5.1step 1.1step 2.1step 3.1step 4.1∎

Conclusion. Steps 1.1–2.1 sum the endpoint patch estimates into a convolution bound for the full compact hypersurface, step 3.1 converts it into the endpoint restriction estimate, and step 4.1 gives the full range and the adjoint formulation. The constant depends on $S$ through the finitely many patch constants, the geometry of the cover, and $σ (S)$ .

RemarkRemark: Literature-sourcedProof: Not supplied‡ not proved hereOpen item page →

‡ Recorded, not proved here. This statement is included so the library can refer to it honestly, with a citation to the literature. It is not proved anywhere in this library: the track that would prove it has not been developed here yet.

The general Fourier restriction problem remains open

Remark

Recorded orientation, not proved here. For $f \in L^{\infty} (S^{n - 1})$ the restriction conjecture asks for $∥ \hat{f d σ} ∥_{L^{q} (R^{n})} ≲_{q} ∥ f ∥_{\infty}$ for all $q > 2 n / (n - 1)$ ; the constant density shows this range is best possible. The conjecture is known for $n = 2$ (Fefferman and Zygmund) and remains open for $n \geq 3$ . The Stein-Tomas theorem of this page settles only the $L^{2}$ -density line $q \geq 2 (n + 1) / (n - 1)$ , which lies strictly above the conjectured $L^{\infty}$ range; no claim here settles the general restriction problem, and the Knapp examples of the companion page constrain every such estimate.

The operators in the display are those of Fourier restriction and adjoint extension operators; the necessity of the $L^{2}$ -density threshold recorded here is the theorem Knapp necessary condition for spherical L2 restriction, and the companion page's counterexample exhibits the same cap family in the limit $δ ↓ 0$ .

RemarkRemark: Literature-sourcedProof: Not supplied‡ not proved hereOpen item page →

Restriction estimates and the missing Strichartz interface

Remark

Recorded orientation, not proved here. The Schrödinger initial-value problem $\partial_{t} u + i Δ u = h$ , $u (0) = f$ on $R^{1 + d}$ uses a paraboloid extension operator, as seen in Williams's formula (11.21); it is not the spherical extension theorem proved on this page. The two arguments share dispersive bounds, Plancherel, $T T^{*}$ and fractional integration. Williams, Definition 11.5 and Theorem 11.6, calls $(p, q)$ admissible when $2 / p + d / q = d / 2$ , $2 \leq p, q \leq \infty$ , excluding $(2, \infty)$ ; for admissible pairs with $p > 2$ and Schwartz data he proves $∥ u ∥_{L_{t}^{p} L_{x}^{q}} \leq C (∥ f ∥_{2} + ∥ h ∥_{L_{t}^{p^{'}} L_{x}^{q^{'}}})$ . This records that nonendpoint theorem with the same pair for solution and forcing. It does not assert endpoint Strichartz estimates or identify them with spherical restriction.

The comparison is anchored to Stein-Tomas spherical restriction theorem. No PDE page is commissioned in this run, and this sourced remark remains a non-load-bearing leaf; it supplies no proof to another item.

5 · Examples, counterexamples and false statements

None yet.