Alphabeta Math
LemmaStatement: Literature-sourcedProof: AI-adaptedPipeline-generatedprecheck passverified 2026-10-22 (gpt-6.1-sol)
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.

Wavefront sets are invariant under smooth coordinate changes

Statement

Assume Countable Choice. For a distribution u on a smooth manifold, the chartwise rapid-decay definition of its wavefront set is independent of the chart: if y=ψ(x), its covector coordinates satisfy η=(Dψ(x))−Tξ. The resulting WF⁡(u) is closed and conic in T∗M∖0. For distributions u,v and a smooth diffeomorphism ψ, one has WF⁡(u+v)⊆WF⁡(u)∪WF⁡(v) and WF⁡(ψ∗u)=ψ∗WF⁡(u). Finally, WF⁡(u)=∅ exactly when u is represented by a smooth scalar function in every chart.

Facts & Assumptions

Given: Countable Choice, a smooth manifold M, and distributions u,v∈D′(M).

[A1]

Assume only the Axiom of Countable Choice ACω The Axiom of Countable Choice (ACω), used for the Schwartz Fourier inversion supplier in step 4.1.

[F1]

A manifold distribution is a continuous bilinear-functional dual to compactly supported smooth densities; its chart representatives transform by the density Jacobian (Scalar distributions on smooth manifolds).

[F2]

In each chart, the local wavefront set is the complement of pairs where a localized Fourier transform decays faster than every power in an open cone; the global set is the union of chartwise cotangent images (Local Fourier definition of a wavefront set, Wavefront and positive-frequency Hadamard data).

[F3]

On each compact support, a distribution has a finite derivative-order bound, with its own order and constant (Local finite order characterization of distributions).

[F4]

A compactly supported distribution has a smooth polynomially bounded localized Fourier transform with the cutoff-pairing formula in the stated 2π convention (Fourier transform of a compactly supported distribution is a smooth polynomially bounded multiplier).

[F5]

For a Schwartz test, the 2π Fourier convention is f^(ξ)=∫f(x)e−2πix⋅ξ dx, with inverse f(x)=∫f^(ξ)e2πix⋅ξ dξ under ACω (Fourier inversion on Schwartz space).

[F6]

A Schwartz-parameter family with integrable seminorm majorants may be integrated through a tempered-distribution pairing under ACω (Schwartz parameter pairing and integral interchange).

[F7]

Cotangent coordinates transform by the inverse transpose Jacobian (Cotangent coordinate changes use the inverse transpose Jacobian).

[F8]

A compact subset of a coordinate domain admits smooth compactly supported cutoffs equal to one near it (Test function cutoffs and euclidean localization).

[F9]

Absolutely integrable Fourier convolution integrands may be reordered by Fubini under ACω (Fubini's theorem for L^1 functions on a sigma-finite product).

Proof

Given: [A1] and [F1]–[F9].

1.1A1F1F2F3F4algebra

In a chart, localize u by a cutoff χ and write χu^(ξ)=⟨u,χ(x)e−2πix⋅ξ∣dx∣⟩. The finite-order bound [F3] gives polynomial growth in ξ by differentiating the test at most its fixed order; parameter derivatives multiply it by powers of x, which remain bounded on the fixed compact support. Difference quotients converge in each fixed-support test seminorm by the Taylor remainder bound, so pairing shows that χu^ is smooth. This is also the localized Fourier transform in [F4].

2.1A1F1F3F4F5F6F7F8step 1.1construct

Suppose (p,ξ0) is regular for u in the x chart, witnessed by a cutoff χx and an open cone Γx. Let y=ψ(x) be a smooth coordinate change, restrict to a compact patch about p, and put η0=(Dψ(p))−Tξ0. Shrink the patch and choose nested cones Γ0⋐Γ1⋐Γx around ξ0 and a target cone V around η0 so that Dψ(x)TV⊂Γ0 throughout the patch. Choose a target cutoff χy equal to one near p and supported where χx=1, using [F8]. The target localized Fourier value is the source distribution paired with aη(x)=χy(ψ(x))∣det⁡Dψ(x)∣e−2πiψ(x)⋅η; because χx=1 on its support this is also the pairing with χxu. Fourier inversion [F5], passed through this compactly supported distribution by [F6], expresses the value as ∫χxu^(ξ)aη^(−ξ) dξ. The phase in aη^(−ξ) is 2π(x⋅ξ−ψ(x)⋅η), with gradient 2π(ξ−Dψ(x)Tη). In the near-stationary region ∣ξ−Dψ(x)Tη∣≤c(∣ξ∣+∣η∣), for sufficiently small c>0, cone separation gives ξ∈Γx and ∣ξ∣≍∣η∣; there χxu^(ξ) decreases faster than every power, while ∣aη^(−ξ)∣≤∥aη∥1 is uniformly bounded, and the region has only polynomial volume in ∣η∣. On the complementary region the phase gradient is bounded below by c(∣ξ∣+∣η∣); repeated integration by parts gives arbitrary powers of joint decay for aη^(−ξ). The polynomial bound [F3] for χxu^ then makes this part rapidly decreasing after integration. This proves regularity in the target chart.

3.1F2F7step 2.1

Applying step 2.1 also to ψ−1 proves equivalence of regularity in both charts, and [F7] gives exactly η=(Dψ)−Tξ. Regularity is open because the same cutoff and a smaller cone work for nearby base points; positive rescaling of a covector does not change rapid decay. Hence its complement is closed and conic.

3.2A1F2F3F4F5F6F8F9step 2.1algebra

Suppose χ1 is a regularity witness for u on an open cone Γ, and let ρ∈Cc∞ have support where χ1=1. If Γ′ is an open cone whose closure of directions on the unit sphere is contained in Γ, then ρu^ is rapidly decreasing on Γ′. Indeed, ρu=ρ(χ1u) and Fourier inversion gives ρu^=ρ^∗χ1u^; passing the Schwartz integral through the compactly supported distribution uses [F6]. Repeated integration by parts makes ρ^ rapidly decreasing, and [F3] bounds χ1u^ polynomially. Choose an intermediate cone with closure of directions between those of Γ′ and Γ. In the convolution, for frequencies η in this intermediate cone with ∣η∣≥∣ξ∣/2, the second factor is rapidly decreasing at frequency comparable to ∣ξ∣. If ∣η∣<∣ξ∣/2, then ∣ξ−η∣>∣ξ∣/2; for the remaining frequencies, η lies outside the intermediate cone and ∣η∣≥∣ξ∣/2, so angular separation gives ∣ξ−η∣≥c(∣ξ∣+∣η∣) for some c>0. In these latter regions the first factor is rapidly decreasing, while [F3] bounds the other factor polynomially. These estimates give the arbitrary-power bound for ∣ξ∣≥1 on Γ′, and [F4] gives boundedness on bounded sets. The convolution is absolutely integrable by the same estimates, and [F9] justifies its Fourier-integral identity. Applying this cutoff transfer to a common ρ for witnesses for u and v, and intersecting their smaller cones, gives rapid decay for ρ(u+v)^=ρu^+ρv^. This proves WF⁡(u+v)⊆WF⁡(u)∪WF⁡(v). For a diffeomorphism, step 2.1 gives the cotangent image inclusion for pullback; applying the same argument to ψ−1 gives the reverse inclusion.

4.1A1F2F5F6F8step 3.2algebra∎

If u is smooth, integration by parts in a compact chart cutoff gives rapid decay in every cone. Conversely, suppose WF⁡(u)=∅ and fix a point p in a chart. In dimension zero every distribution is locally a scalar function, so assume the chart dimension is at least one. Form the set of all open cones Γ′ for which there exist a regularity witness (χ,Γ) at p with the closure of the directions of Γ′ on the unit sphere contained in Γ. This set covers the unit covector sphere: every direction has a witness by emptiness of the local wavefront set, and its open cone contains a smaller cone about that direction. Compactness gives a finite subcover Γ1′,…,Γm′. For each member of this finite family, take a corresponding witness χj,Γj; finite choice suffices. Their witness cutoffs all equal one on a neighborhood of p. By [F8], choose one cutoff ρ equal to one near p and supported in the intersection of those neighborhoods. Step 3.2 shows that the single localization ρu has rapidly decreasing Fourier transform on each Γj′. Since these finitely many cones cover the unit sphere, ρu^ is rapidly decreasing in every direction: for each N, take the maximum of the finitely many cone bounds outside a bounded ball. Therefore every differentiated inverse Fourier integral is absolutely convergent and defines a smooth function. Fourier inversion [F5] and the pairing interchange [F6], applied after a cutoff equal to one near supp⁡(ρu), identify this function with the compactly supported distribution ρu. Hence u is smooth near p; as p was arbitrary, it is smooth in every chart.

Depends on

Dependency tree · 2 levels

Sources