TheoremBase

The Periodised Kernel of a Periodic Convolution

Wrapping a compactly supported kernel around the lattice gives a kernel that is continuous and periodic in each variable separately, as smooth as the original, has mass over the cell no larger than the original, is jointly measurable on the cell squared, and represents the periodic convolution as an integral over the torus. Differentiating it in the second variable returns, up to sign, the periodised kernel of the derivative.

Statement

We work in the setting of The Flat Torus: Standing Notation, used here with a natural number nn satisfying 1≤n1\le n; the cell QQ, the lattice Zn\mathbb{Z}^{n}, the measure space (Q,BQ,λQ)(Q,\mathcal{B}_{Q},\lambda_{Q}), the integral over Tn\mathbb{T}^{n}, the classes Lr(Tn)\mathcal{L}^{r}(\mathbb{T}^{n}) for a real rr with 1≤r1\le r, the periodic classes CperC_{\mathrm{per}} and Cper∞C^{\infty}_{\mathrm{per}}, the restriction v∣Qv|_{Q}, and B(Rn)\mathcal{B}(\mathbb{R}^{n}), λn\lambda_{n} and the Euclidean norm ∥ ⋅ ∥\lVert\,\cdot\,\rVert are the ones fixed there. A map Rn→R\mathbb{R}^{n}\to\mathbb{R} is called continuous, and smooth, as in Periodic Convolution and Mollification on the Torus; ψ⋆u\psi\star u denotes the periodic convolution; and a sum indexed by a finite set is that of Sum over a Finite Index Set.

Throughout, RR is a real number with 0<R0<R and ψ:Rn→R\psi:\mathbb{R}^{n}\to\mathbb{R} is continuous with ψ(y)=0\psi(y)=0 for every y∈Rny\in\mathbb{R}^{n} satisfying R<∥y∥R<\lVert y\rVert. Then the following hold.

1. (The kernel is integrable) The map ψ\psi is compactly supported, bounded, and integrable with respect to λn\lambda_{n}. We write

K=∫Rn∣ψ∣ dλn,K=\int_{\mathbb{R}^{n}}|\psi|\,d\lambda_{n},

a real number with 0≤K0\le K.

2. (The periodised kernel) For all x,y∈Rnx,y\in\mathbb{R}^{n} the set S(x,y)={m∈Zn:ψ(x−y−m)≠0}S(x,y)=\{m\in\mathbb{Z}^{n}:\psi(x-y-m)\ne0\} is finite, and for every nonempty finite F⊆ZnF\subseteq\mathbb{Z}^{n} with S(x,y)⊆FS(x,y)\subseteq F — for instance F=S(x,y)∪{0}F=S(x,y)\cup\{0\} — the sum

∑m∈Fψ(x−y−m)\sum_{m\in F}\psi(x-y-m)

has one and the same value. Writing Ψ(x,y)\Psi(x,y) for that common value defines a map Ψ:Rn×Rn→R\Psi:\mathbb{R}^{n}\times\mathbb{R}^{n}\to\mathbb{R}, the periodised kernel of ψ\psi.

3. (Regularity in each variable) Let x∈Rnx\in\mathbb{R}^{n}. Then the map y↦Ψ(x,y)y\mapsto\Psi(x,y) belongs to CperC_{\mathrm{per}}, and belongs to Cper∞C^{\infty}_{\mathrm{per}} if ψ\psi is smooth. Symmetrically, for y∈Rny\in\mathbb{R}^{n} the map x↦Ψ(x,y)x\mapsto\Psi(x,y) belongs to CperC_{\mathrm{per}}, and to Cper∞C^{\infty}_{\mathrm{per}} if ψ\psi is smooth.

4. (Mass bound over the cell) Let x∈Rnx\in\mathbb{R}^{n} and let Ψx\Psi_{x} be the restriction of y↦Ψ(x,y)y\mapsto\Psi(x,y) to QQ. Then Ψx∈L1(Tn)\Psi_{x}\in\mathcal{L}^{1}(\mathbb{T}^{n}) and

∫Tn∣Ψx∣ dy≤K.\int_{\mathbb{T}^{n}}|\Psi_{x}|\,dy\le K .

Symmetrically, for y∈Rny\in\mathbb{R}^{n} the restriction Ψy\Psi^{y} of x↦Ψ(x,y)x\mapsto\Psi(x,y) to QQ lies in L1(Tn)\mathcal{L}^{1}(\mathbb{T}^{n}) and satisfies the same bound.

5. (Representation of the periodic convolution) Let u∈L1(Tn)u\in\mathcal{L}^{1}(\mathbb{T}^{n}) and let x∈Rnx\in\mathbb{R}^{n}. Then the pointwise product Ψx u\Psi_{x}\,u is integrable with respect to λQ\lambda_{Q} and

(ψ⋆u)(x)=∫TnΨx u dy.(\psi\star u)(x)=\int_{\mathbb{T}^{n}}\Psi_{x}\,u\,dy .

6. (Joint measurability over the cell) The restriction of Ψ\Psi to Q×QQ\times Q is measurable with respect to the product σ\sigma-algebra BQ⊗BQ\mathcal{B}_{Q}\otimes\mathcal{B}_{Q}.

7. (Differentiating the periodised kernel) Suppose ψ\psi is smooth, and let i∈[n]i\in[n]. Then ψ\psi is of class C1C^{1} on Rn\mathbb{R}^{n} by claim 2 of Euclidean Space Is Open in Itself, and the partial derivative ∂iψ\partial_{i}\psi is continuous and vanishes at every y∈Rny\in\mathbb{R}^{n} with R<∥y∥R<\lVert y\rVert, by Convolution of a Locally Integrable Function with a Compactly Supported Kernel (iv); so ∂iψ\partial_{i}\psi satisfies the standing hypotheses placed on ψ\psi above, and claim 2 applied to it yields its periodised kernel, which we write Ψ(i)\Psi^{(i)}. Let x∈Rnx\in\mathbb{R}^{n} and write Ψ(x,⋅)\Psi(x,\cdot) for the map y↦Ψ(x,y)y\mapsto\Psi(x,y), a member of Cper∞C^{\infty}_{\mathrm{per}} by claim 3. Then

∂iΨ(x,⋅)(y)=−Ψ(i)(x,y)for every y∈Rn.\partial_{i}\Psi(x,\cdot)(y)=-\Psi^{(i)}(x,y)\qquad\text{for every }y\in\mathbb{R}^{n}.

Proofs

Log in to submit a proof.

Loading...

Citations

Loading…

Dependencies

Loading…

Related

0 relations

Curated associations between results. These are editable and subjective — they do not replace the dependency graph, which is derived from the references in the text.

No relations recorded yet.

Comments

Log in to comment.

Loading…