Bourbaki

Proposition 21

Tag 031L, permanent. Cite it as [Bourbaki, Tag 031L].

Théories spectrales, chapter IV, § 3. Distributions et distributions tempérées

Let N be the euclidean norm on Rn\mathbf{R}^n. Let kk be an integer ⩾0\geqslant 0. The Sobolev space Hk(Rn)H^k(\mathbf{R}^n) is the space of f∈S′(Rn)f\in \mathscr{S}'(\mathbf{R}^n) such that (1+Nk)F(f)(1 + N^k)\mathscr{F}(f) belongs to L2(Rn)L^2(\mathbf{R}^n).

We proceed by induction on kk. When k=0k= 0, the result is a consequence of the Plancherel theorem (II, p. 215, th. 1). Suppose that k=1k= 1. For f∈S′(Rn)f\in \mathscr{S}'(\mathbf{R}^n), one has (1+N)Ff∈L2(Rn)(1 + N)\mathscr{F}f\in L^2(\mathbf{R}^n) if and only if f∈L2(Rn)f\in L^2(\mathbf{R}^n) and NFf∈L2(Rn)N\mathscr{F}f\in L^2(\mathbf{R}^n). Moreover, NF∈L2(Rn)N\mathscr{F}\in L^2(\mathbf{R}^n) if and only if, for every α∈Nn\alpha \in \mathbf{N}^n such that ∣α∣=1|\alpha |= 1, one has XαFf∈L2(Rn)X^{\alpha}\mathscr{F}f\in L^2(\mathbf{R}^n). Since one has F(∂αf)=2iπXαFf\mathscr{F}(\partial^{\alpha}f) = 2i\pi X^{\alpha}\mathscr{F}f, this condition means that F(∂αf)∈L2(Rn)\mathscr{F}(\partial^{\alpha}f)\in L^2(\mathbf{R}^n) for every α\alpha with ∣α∣=1|\alpha |= 1, that is, ∂αf∈L2(Rn)\partial^{\alpha}f\in L^2(\mathbf{R}^n) for every α\alpha with ∣α∣=1|\alpha |= 1. It follows that the assertion is true for k=1k= 1.

Suppose now that k⩾k\geqslant 2 and that the assertion concerning Hℓ(Rn)H^{\ell}(\mathbf{R}^n) is valid for every positive integer ℓ⩽k−1\ell \leqslant k-1. Let f∈S′(Rn)f\in \mathscr{S}'(\mathbf{R}^n). By definition, one has f∈Hk(Rn)f\in H^k(\mathbf{R}^n) if and only if f∈L2(Rn)f\in L^2(\mathbf{R}^n) and, for every β∈Nn\beta \in \mathbf{N}^n such that ∣β∣⩽1|\beta |\leqslant 1, the distribution ∂βf\partial^{\beta}f belongs to Hk−1(Rn)H^{k-1}(\mathbf{R}^n). This is equivalent, by the induction hypothesis, to f∈L2(Rn)f\in L^2(\mathbf{R}^n) and (1+Nk−1)F(∂βf)∈L2(Rn)(1 + N^{k-1})\mathscr{F}(\partial^{\beta}f)\in L^2(\mathbf{R}^n) for every β∈Nn\beta \in \mathbf{N}^n such that ∣β∣⩽1|\beta |\leqslant 1. Since F(∂βf)=(2iπX)∣β∣F(f)\mathscr{F}(\partial^{\beta}f) = (2i\pi X)^{|\beta|}\mathscr{F}(f), the condition f∈Hk(Rn)f\in H^k(\mathbf{R}^n) is equivalent to saying that Ff∈L2(Rn)\mathscr{F}f\in L^2(\mathbf{R}^n) and (1+Nk−1)XβFf∈L2(Rn)(1 + N^{k-1})X^{\beta}\mathscr{F}f\in L^2(\mathbf{R}^n) for β∈Nn\beta \in \mathbf{N}^n such that ∣β∣⩽1|\beta |\leqslant 1.

The inequalities

1+Nk⩽1+Nk−1∑β∣β∈∣N⩽1n∣Xβ∣⩽1+n1/2Nk⩽(1+n1/2)(1+Nk) 1 + N^k\leqslant 1 + N^{k-1}\sum_{\beta|\beta\in|\mathbf{N}\leqslant 1^n}|X^{\beta}|\leqslant 1 +n^{1/2}N^k\leqslant (1 +n^{1/2})(1 + N^k)

then imply that f∈Hk(Rn)f\in H^k(\mathbf{R}^n) if and only if (1+Nk)F∈L2(Rn)(1+N^k)\mathscr{F}\in L^2(\mathbf{R}^n).

Bourbaki sets the proof directly after the statement, and the corpus does not separate them, so what is above is the statement and its proof together.

Cited by

Nothing.

Cites

Nothing.

Machine translation of the English, by gpt-5-mini, gpt-5.4, against glossary version 34. Not checked by a person. That model is a cut down version of the one the rest of this language was written by.