Alphabeta Math
LemmaStatement: AI-adaptedProof: AI-adaptedPipeline-generatedjudge pass (gpt-6-luna)
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.

Inward mollification and smooth finite-rank approximation on Lipschitz domains

Statement

Assume Countable Choice. Let D⊆Rd be a bounded Lipschitz open set, with d≥1, and let m∈N0. Let X be either Wm,p(D;R) for 1≤p<∞, including W0,p(D)=Lp(D), or Cm(Dˉ;R). In the latter space derivatives in D extend continuously to Dˉ, and the norm is max⁡∣α∣≤msup⁡Dˉ∣Dαu∣. For every norm-compact K⊆X and every ε>0, there is a bounded finite-rank linear map F:X→X such that sup⁡u∈K∥Fu−u∥X<ε and whose range is contained in the restrictions to D of functions in Cc∞(Rd), with their continuous boundary values in the Cm case. In particular, such smooth restrictions are norm-dense in X.

Moreover, there are finitely many ambient smooth cutoffs ψj, vectors vj, a unit-mass ρ∈Cc∞(Rd), and t0>0 such that for 0<t<t0 the kernel kt(x,y)=∑jψj(x)ρt(x+tvj−y), with ρt(z)=t−dρ(z/t), defines bounded linear smoothing maps Stu(x)=∫Dkt(x,y)u(y) dy. They satisfy sup⁡0<t<t0∥St∥X→X<∞ and sup⁡u∈K∥Stu−u∥X→0 for every norm-compact K⊂X. Each kernel is ambient smooth with fixed compact support in its first variable.

Facts & Assumptions

Given: Countable Choice, D,d,m,X,K,ε as in the statement. Lipschitz means that every boundary point has a rigid graph chart in which D is the strict epigraph of a Lipschitz function.

[F1]

The Sobolev norm is the finite ℓp sum of the Lp norms of the weak derivatives through order m, and W0,p=Lp; this is a well-defined norm under Countable Choice (Integer-order Sobolev spaces and their norms, The Sobolev norm descends to equivalence classes).

[F2]

A finite open cover of a compact Euclidean set admits nonnegative ambient smooth functions compactly supported in the respective cover members and summing to one on a neighborhood of that set (Finite ambient partitions near compact sets).

[F3]

Interior mollification commutes with every weak derivative through order m wherever the closed mollifier-support ball is compactly contained in the domain (Interior mollification commutes with weak derivatives).

[F4]

Under Countable Choice, translations are isometries in Lp(Rd) and are norm-continuous for p<∞; convolution with a compact smooth kernel is smooth and differentiates by differentiating the kernel (Complex translation, convolution, approximate identities, and mollification).

[F5]

Minkowski's integral inequality holds for all 1≤p<∞; its p=1 case is Tonelli's identity (Minkowski's integral inequality).

[F6]

Classical derivatives of a Cm function are weak derivatives, and the weak Leibniz rule holds for a smooth multiplier with bounded derivatives (Classical derivatives agree with weak derivatives, Weak Leibniz rule with a smooth factor).

[F8]

Bounded measurable Euclidean sets have finite measure, and Hölder, including p=1, gives ∥u∥L1(D)≤∣D∣1−1/p∥u∥Lp(D) for nonempty open D (Lebesgue measure is sigma-finite, and every metrically bounded subset of Rn has finite outer measure, Holder's inequality for integrals, including the endpoint cases).

[F9]

Nested balls admit a nonnegative compactly supported smooth bump equal to one on the inner ball (A smooth bump between concentric Euclidean balls).

Choice use: Countable Choice is used in the Sobolev, Lebesgue-measure, translation, convolution and weak-derivative interfaces [F1], [F3]–[F6], [F8]. The geometric cover and all later nets and cell selections are finite selections. No stronger choice axiom is used.

Proof

1.1givenF7F8constructalgebra

If D=∅, X is the zero space; take F=0, the empty cutoff family, kt=0, St=0 and t0=1. If K=∅ the compact approximation clause is vacuous and F=0 suffices; the smoothing construction below depends only on D and X, and can be obtained by carrying it out with K={0}. Henceforth both sets are nonempty. In particular 0<∣D∣<∞, because D contains a ball and is bounded. Fix one boundary chart Q, with Lipschitz graph function φ of constant L, and an ambient neighborhood U of its boundary point with Uˉ⊂Q compact. In chart coordinates set v=(L+2)ed. For x∈Dˉ∩Uˉ, ∣z∣≤1, and sufficiently small t>0, uniformly in x,z, the point y=x+tv−tz stays in Q because t(∣v∣+1)<dist⁡(Uˉ,Qc). Moreover yd−φ(y′)≥xd−φ(x′)+t(L+2)−t∣zd∣−Lt∣z′∣≥t>0. Thus B‾(x+tv,t)⋐D. Rotate v back to the original coordinates.

2.1F2F7step 1.1construct

The compact boundary admits finitely many such Uj, 1≤j≤J, covering it. The compact remainder Dˉ∖⋃j=1JUj lies in D; cover it by finitely many balls with closures in D, and let U0 be their union, omitting it if the remainder is empty. Then Uˉ0⋐D, so v0=0 has the same ball property for sufficiently small t. Apply [F2] to this finite open cover of Dˉ, obtaining ψj∈Cc∞(Uj) with ∑jψj=1 near Dˉ. The minimum of the finitely many positive bounds on t gives t0>0 such that B‾(x+tvj,t)⋐D for every x∈Dˉ∩supp⁡ψj and 0<t<t0. Taking the family of all admissible chart neighborhoods before extracting a finite subcover avoids selecting a chart independently at every boundary point.

3.1F4F8F9step 2.1construct

By [F9] choose a nonnegative smooth bump compactly supported in B1(0) and positive on B‾1/2(0), and divide it by its positive integral to obtain ρ∈Cc∞(B1(0)) with ∫ρ=1. Put ρt(w)=t−dρ(w/t). Every u∈X belongs continuously to L1(D): use [F8] for Sobolev spaces, and ∥u∥1≤∣D∣∥u∥Cm for Cm. For its zero extension E0u define Hj,tu(x)=(ρt∗E0u)(x+tvj) and Stu=∑jψjHj,tu. By [F4] these are well-defined linear constructions independent of representatives and Stu∈Cc∞(Rd).

4.1F1F3step 2.1step 3.1

First take X=Wm,p(D). For every ∣γ∣≤m and x∈Dˉ∩supp⁡ψj, [F3] and step 2.1 give DγHj,tu(x)=∫ρ(z)(E0Dγu)(x+t(vj−z)) dz. The translated center has distance greater than t from Dc: the entire closed ball of radius t is a compact subset of D. The identity consequently holds on a neighborhood of each indicated x. Equivalently the defining weak-derivative identity can be tested against y↦ρt(x+tvj−y), whose support is compact in D; its two derivative signs cancel. Only interior values are used, with no assertion that derivatives commute with zero extension across ∂D.

4.2F4F5step 3.1construct

For f∈Lp(Rd) define Tj,tf(x)=∫ρ(z)f(x+t(vj−z)) dz and ωf(r)=sup⁡∣h∣≤r∥f(⋅+h)−f∥p. Translation invariance and [F5] give ∥Tj,tf∥p≤∥f∥p and ∥Tj,tf−f∥p≤ωf((∣vj∣+1)t). By [F4], ωf(r)→0 as r↓0. The integral estimates also apply when p=1, where [F5] follows directly from Tonelli.

4.3F7step 3.1construct

Fix 0<t<t0. Write kt(x,y)=∑jψj(x)ρt(x+tvj−y), so Stu(x)=∫Dkt(x,y)u(y) dy. Fix a closed box Q containing Dˉ. The finitely many Dxαkt, ∣α∣≤m, are smooth and uniformly continuous on the compact set Dˉ×Q by [F7]. For any η>0, partition Q into finitely many measurable cells Qℓ of small enough diameter, assigning shared and outer faces consistently, and choose yℓ∈Qℓ. Their diameters can be chosen so that ∣Dxαkt(x,y)−Dxαkt(x,yℓ)∣<η for x∈Dˉ, y∈Qℓ and ∣α∣≤m. The point yℓ need not lie in D.

5.1F1F6step 2.1step 4.1step 4.2algebra

Set fγ=E0Dγu. The finite Leibniz rule [F6], step 4.1 and ∑jψj=1 near Dˉ give, for ∣α∣≤m, the a.e. identity on D Dα(Stu−u)=∑j∑β≤α(αβ)(Dβψj)(Tj,tfα−β−Dα−βu). Every derivative Dβψj vanishes off supp⁡ψj, so step 4.1 supplies exactly the identities used in this formula. All these fixed multiplier derivatives are bounded. Step 4.2 and the finite norm sum [F1] therefore imply ∥Stu−u∥Wm,p→0. The same Leibniz formula without subtracting u, and ∥Tj,tfγ∥p≤∥fγ∥p≤∥u∥Wm,p, give sup⁡0<t<t0∥St∥X→X≤MX<∞. For example take the ℓp norm over α of ∑j,β≤α(αβ)∥Dβψj∥∞. This argument includes m=0 and p=1.

6.1F6F7step 2.1step 4.1step 5.1algebra

Now take X=Cm(Dˉ). By [F6] each bounded classical derivative is its weak derivative, so u∈Wm,1(D) and step 4.1 applies. Each Dγu is uniformly continuous on Dˉ by [F7]. For x∈Dˉ∩supp⁡ψj, every sampled y=x+t(vj−z) lies in D and ∣y−x∣≤(∣vj∣+1)t. Positivity and mass one of ρ thus give sup⁡x∈Dˉ∩supp⁡ψj∣DγHj,tu(x)−Dγu(x)∣≤sup⁡x,y∈Dˉ∣x−y∣≤(∣vj∣+1)t∣Dγu(y)−Dγu(x)∣⟶0. The formula of step 5.1, first on D and then by continuity on Dˉ, gives ∥Stu−u∥Cm→0. Bounding each averaged derivative by ∥Dγu∥∞ and summing the same fixed multiplier constants gives sup⁡0<t<t0∥St∥X→X≤MX<∞ also in this case.

7.1F1F8step 3.1step 5.1step 6.1step 4.3constructalgebra

Define Ft,ηu(x)=∑ℓkt(x,yℓ)∫D∩Qℓu(y) dy. Each coefficient is a bounded linear functional by step 3.1, and there are finitely many coefficients. Each range function kt(⋅,yℓ) belongs to Cc∞(Rd), so the operator has finite rank with the required smooth range. Comparison with the integral for St gives max⁡∣α∣≤msup⁡Dˉ∣Dα(Ft,η−St)u∣≤η∥u∥L1(D). For X=Cm(Dˉ), this yields ∥Ft,η−St∥X→X≤∣D∣η. For X=Wm,p(D) let Am be the finite number of multi-indices of order at most m. Taking the derivative Lp norms and then applying [F8] gives ∥Ft,η−St∥X→X≤Am1/p∣D∣1/pη∣D∣1−1/p=Am1/p∣D∣η. This includes p=1, and it shows each Ft,η is bounded because St is bounded.

7.2step 5.1step 6.1givenchoosealgebra

By steps 5.1 and 6.1 the operators St:X→X converge pointwise to the identity and satisfy ∥St∥≤MX. Choose δ<ε/[4(MX+1)]. Compactness gives a finite δ-net u1,…,uN in K. Pointwise convergence permits one t∈(0,t0) such that max⁡i∥Stui−ui∥<ε/4. For any u∈K, choose one net point within δ; then ∥Stu−u∥≤(MX+1)δ+max⁡i∥Stui−ui∥<ε/2. Compactness also makes R=sup⁡u∈K∥u∥ finite: a finite cover by unit balls bounds all these norms by one plus the largest center norm.

8.1step 7.1step 7.2choosealgebra∎

Choose η>0 small enough that the relevant bound in step 7.1 is below ε/[2(R+1)]. For F=Ft,η, sup⁡u∈K∥Fu−u∥≤∥F−St∥R+sup⁡u∈K∥Stu−u∥<ε. Its range is a finite span of ambient compactly supported smooth functions by step 7.1. Applying this conclusion to the singleton compact set K={u} proves smooth-restriction density for every u∈X. All net, partition and representative-point choices here are finite; Countable Choice enters only the analytical interfaces stated above.

Depends on

Used by

Dependency tree · two levels

111 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources