Documentation

Atlas.DifferentialAnalysis.code.ConvolutionDensity

The cocompact filter on Euclidean space is countably generated: it has a countable basis given by complements of closed balls of integer radius.

theorem ConvolutionDensity.approxIdentity_tendstoUniformly {G : Type u_1} [SeminormedAddCommGroup G] [MeasurableSpace G] [BorelSpace G] [SecondCountableTopology G] {μ : MeasureTheory.Measure G} [μ.IsAddLeftInvariant] [MeasureTheory.SFinite μ] {ι : Type u_2} {l : Filter ι} {φ : ιG} (f : ZeroAtInftyContinuousMap G ) (hnφ : ∀ᶠ (i : ι) in l, ∀ (x : G), 0 φ i x) (hiφ : ∀ᶠ (i : ι) in l, (x : G), φ i x μ = 1) ( : Filter.Tendsto (fun (i : ι) => Function.support (φ i)) l (nhds 0).smallSets) (hmφ : ∀ᶠ (i : ι) in l, MeasureTheory.AEStronglyMeasurable (φ i) μ) :
TendstoUniformly (fun (i : ι) => MeasureTheory.convolution (φ i) (⇑f) (ContinuousLinearMap.lsmul ) μ) (⇑f) l

Approximate-identity convergence for C₀ functions: if φ i is a family of nonnegative, integral-one functions whose supports shrink to {0}, then the convolutions φ i ⋆ f converge uniformly to f for any f ∈ C₀(G, ℝ). This is a key ingredient for proving density results via mollification.

Density between smoothness classes of C₀ functions: any C^p zero-at-infinity function can be approximated in the C^p norm by C^k zero-at-infinity functions whenever p ≤ k.

theorem ConvolutionDensity.ckNorm_convolution_threshold {n : } (k : ) (v : EuclideanSpace (Fin n)) (hv_supp : HasCompactSupport v) (hv_smooth : ContDiff (↑k) v) (ε : ) ( : 0 < ε) :

Mollification threshold in the C^k norm: for any compactly-supported C^k function v and any ε > 0, there exists a radius r > 0 such that convolution with any normalised bump of outer radius at most r approximates v within ε in the C^k norm. The proof proceeds by induction on k, using uniform continuity of derivatives at the base step and commuting convolution with fderiv at the inductive step.

theorem ConvolutionDensity.approx_compactly_supported_Ck_by_schwartz {n : } (k : ) (v : EuclideanSpace (Fin n)) (hv_supp : HasCompactSupport v) (hv_smooth : ContDiff (↑k) v) (ε : ) ( : 0 < ε) :
∃ (φ : SchwartzMap (EuclideanSpace (Fin n)) ), (TestFunctions.ckNorm n k fun (x : EuclideanSpace (Fin n)) => v x - φ x) < ε

Any compactly-supported C^k function on Euclidean space can be approximated in the C^k norm by a Schwartz function. The approximant is constructed by convolving v with a sufficiently narrow normalised bump, which produces a compactly-supported smooth function — and hence a Schwartz function — within ε of v in the C^k norm.

CkNormBdd n k f is the recursive predicate stating that all derivatives of f : EuclideanSpace ℝ (Fin n) → ℂ up to order k are bounded; that is, f is k times differentiable and f together with all its partial derivatives up to order k have bounded norm ranges.

Instances For

    The difference of a C^k zero-at-infinity function and a compactly supported C^k function has all derivatives up to order k bounded.

    theorem ConvolutionDensity.ckNormBdd_of_smooth_sub_schwartz {n : } (k : ) (v : EuclideanSpace (Fin n)) (φ : SchwartzMap (EuclideanSpace (Fin n)) ) (hv_supp : HasCompactSupport v) (hv_smooth : ContDiff (↑k) v) :
    CkNormBdd n k fun (x : EuclideanSpace (Fin n)) => v x - φ x

    The difference of a compactly supported C^k function and a Schwartz function has all derivatives up to order k bounded.

    theorem ConvolutionDensity.ckNorm_add_le_base {n : } (f g : EuclideanSpace (Fin n)) (hf : BddAbove (Set.range fun (x : EuclideanSpace (Fin n)) => f x)) (hg : BddAbove (Set.range fun (x : EuclideanSpace (Fin n)) => g x)) :

    Base case of the triangle inequality for the C^k norm: at order zero, the supremum norm satisfies ‖f + g‖_∞ ≤ ‖f‖_∞ + ‖g‖_∞ whenever each is bounded above.

    theorem ConvolutionDensity.ckNorm_add_le {n : } (k : ) (f g : EuclideanSpace (Fin n)) (hf : CkNormBdd n k f) (hg : CkNormBdd n k g) :

    Triangle inequality for the C^k norm: for functions with bounded derivatives up to order k, the C^k norm of a sum is at most the sum of the C^k norms. Proved by induction on k using linearity of fderiv on sums of differentiable functions.

    Schwartz functions are dense in the space of C^k zero-at-infinity functions: any such u can be approximated in the C^k norm to within ε by a Schwartz function. The proof first approximates u by a compactly supported C^k function, then approximates that by a Schwartz function via mollification.

    Continuity of translation in : for u ∈ L²(ℝⁿ), the norm of u(· + t) - u(·) tends to 0 as t → 0. This is the standard "continuity in mean" property of L^p spaces.

    Two finite measures on Euclidean space are equal if they integrate every Schwartz function to the same value: Schwartz functions form a separating family for finite measures.

    Multiplication by a compactly-supported function on the left preserves compact support.

    theorem ConvolutionDensity.contDiff_mul_of_both {n : } (μ g : EuclideanSpace (Fin n)) (hμ_smooth : ContDiff (↑) μ) (hg_smooth : ContDiff (↑) g) :
    ContDiff fun (x : EuclideanSpace (Fin n)) => μ x * g x

    The pointwise product of two smooth functions on Euclidean space is smooth.

    General-target form: the compactly-supported Schwartz functions 𝓢(E, F) are dense in 𝓢(E, F). The proof uses bump-cutoff multiplication and the convergence of the corresponding Schwartz seminorms.

    noncomputable def SchwartzMap.cutoffMul {E : Type u_1} {F : Type u_2} [NormedAddCommGroup E] [NormedSpace E] [NormedAddCommGroup F] [NormedSpace F] (φ : E) (hφcs : HasCompactSupport φ) (hφcd : ContDiff φ) (f : SchwartzMap E F) :

    Multiplication of a Schwartz function f by a compactly-supported smooth real-valued function φ. The result is again Schwartz with compact support contained in the support of φ.

    Instances For
      @[simp]
      theorem SchwartzMap.cutoffMul_apply {E : Type u_1} {F : Type u_2} [NormedAddCommGroup E] [NormedSpace E] [NormedAddCommGroup F] [NormedSpace F] (φ : E) (hφcs : HasCompactSupport φ) (hφcd : ContDiff φ) (f : SchwartzMap E F) (x : E) :
      (cutoffMul φ hφcs hφcd f) x = φ x f x

      Pointwise evaluation of SchwartzMap.cutoffMul: at each x, the cut-off product equals φ x • f x.

      The Plancherel isometry: the Fourier transform on a real finite-dimensional inner product space V is a -linear isometric equivalence of L²(V; ℂ) with itself. This is the operator-theoretic incarnation of the Plancherel theorem.

      Instances For