for every f∈A(G). For f∈A(G), one has F(f)∈L1(G,ν)∩L2(G,ν).
Let f∈A(G). Let Ωf denote the open set in G formed by the χ∈G such that F(f)(χ)=0. Let φ be the characteristic function of G−Ωf. Then for every g∈A(G), one has
∫GF(g)d(φ⋅μf)=∫GφF(f)dμg=0
in view of formula (18).
By corollary 1 of II, p. 212, the image A(G) of A(G) under the Fourier transform is dense in C0(G). One then deduces from the preceding formula that φ⋅μf=0, hence that μf is concentrated on Ωf (INT, IV, §4, no7, def. 4). Let νf be the measure on Ωf with density F(f)−1 with respect to μf∣Ωf.
The sets Ωf, for f∈A(G), form an open covering of G (cor. 2). For all f and g in A(G), formula (18) shows that νf∣(Ωf∩Ωg)=νg∣(Ωf∩Ωg). Consequently, there exists a unique measure ν on G such that one has νf=ν∣Ωf for every f∈A(G) (INT, III, §2, No.o1, prop. 1).
If f∈A(G), the measures μf and F(f)⋅ν are concentrated on Ωf, and their restrictions to Ωf are equal to F(f)⋅νf; these measures are therefore equal.
Since μf is a bounded measure, the Fourier transform F(f) belongs to the space L1(G,ν). Since moreover F(f) belongs to C0(G), one has also F(f)∈L2(G,ν).
Formula (17) now takes the form, for φ∈ Stell(G) and f∈A(G): (20) (φ∗f)(e)=∫GF(φ)F(f)dν.
In particular, for f and g in A(G), one has
∫GF(f)F(g)dν=(f∗g)(e)=∫Gf(x)g(x−1)dx(21)
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.