Bourbaki

Corollary 2

Tag 02N8, permanent. Cite it as [Bourbaki, Tag 02N8].

Théories spectrales, chapter II, § 3. Sous-espaces invariants

Let Φ be a filter on G invariant under translation. Let f∈L1(G)f\in L^1(G) such that the Fourier transform of ff does not vanish and such that ∫Gf(x)dx=1\int_Gf(x)dx= 1. Let g∈L∞(G)g\in L^{\infty}(G). Suppose that f∗gf*g has a finite limit α\alpha following Φ.

a) For every function h∈L1(G)h\in L^1(G) such that ∫Gh(x)dx=1\int_Gh(x)dx= 1, the limit of h∗gh*g following Φ is equal to α\alpha;

b) Suppose in addition that gg is slowly oscillating following Φ. Then gg tends to α\alpha following Φ.

Replacing gg by g−αg-\alpha, we reduce to the case where α=0\alpha = 0. Let I be the set of functions h∈L1(G)h\in L^1(G) such that h∗gh*g tends to 0 following Φ. The set I is a vector subspace of L1(G)L^1(G) invariant under translation. It is a closed space. Indeed, let h∈I‾h\in \overline{I}. For every function h0∈L1(G)h_0\in L^1(G) and every x∈Gx\in G, we have

∣(h∗g)(x)∣⩽∣((h−h0)∗g)(x)∣+∣(h0∗g)(x)∣ |(h*g)(x)|\leqslant |((h-h_0)*g)(x)|+|(h_0*g)(x)|
⩽∥h−h0∥1∥g∥∞+∣(h0∗g)(x)∣ \leqslant \|h-h_0\|_1\|g\|_{\infty}+|(h_0*g)(x)|

For every ε>0\varepsilon >0, there exists h0∈Ih_0\in I such that ∥h−h0∥1∥g∥∞<ε\|h-h_0\|_1\|g\|_{\infty}< \varepsilon. Let M∈M\in Φ such that ∣(h0∗g)(x)∣<ε|(h_0*g)(x)|< \varepsilon for every x∈Mx\in M. We then have ∣(h∗g)(x)∣<2ε|(h*g)(x)|<2\varepsilon for every x∈Mx\in M, hence h∗gh*g converges to 0 following Φ. This shows that h∈Ih\in I.

The space I is therefore a closed ideal of L1(G)L^1(G). We have f∈If\in I by assumption, hence I=L1(G)I = L^1(G) by th. 1 since the Fourier transform of ff does not vanish. This implies a).

Let us place ourselves under the hypotheses of b). Let ε>0\varepsilon >0. Since gg is slowly oscillating according to Φ, there exists M∈ΦM\in \Phi and a compact neighbourhood V of ee such that

x∈Mx\in M and y∈V=⇒∣g(y−1x)−g(x)∣⩽εy\in V =\Rightarrow |g(y^{-1}x)-g(x)|\leqslant \varepsilon.

Let φ\varphi be the characteristic function of V and μ=∫φ(x)dx\mu=\int\varphi (x)dx. For every x∈Gx\in G, we have

1μ(φ∗g)(x)=1μ∫Vg(y−1x)dy=g(x)+1μ∫V(g(y−1x)−g(x))dy \frac{1}{\mu}(\varphi *g)(x) =\frac{1}{\mu}\int_Vg(y^{-1}x)dy=g(x) +\frac{1}{\mu}\int_V(g(y^{-1}x)-g(x))dy

Thus for every x∈Mx\in M, we have

∣1μ(φ∗g)(x)−g(x)∣⩽ε |\frac{1}{\mu}(\varphi *g)(x)-g(x)|\leqslant \varepsilon

Since, by a), the limit of φ∗g\varphi *g according to Φ is zero, we have lim supΦ∣g∣⩽ε_{\Phi}|g|\leqslant \varepsilon. Since ε\varepsilon is arbitrary, we conclude that the limit of gg according to Φ is zero, which proves b).

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-5.4, gpt-5-6-mini, gpt-5-mini, 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.