Let φ∈L1(G)∩L2(G). Then f=FG(φ)∘η belongs to L2(G) and FG(f)=φ in L2(G).
The function f is continuous and bounded on G since φ∈L1(G). For every function g∈L1(G)∩L2(G), one has
∫Gg(x)f(x)dx=∫Gg(x)(∫G⟨χ,x⟩φ(χ)dx(χ))dx
=∫GFG(g)(χ)φ(χ)dx(χ)(28)
applying the Lebesgue-Fubini theorem (INT, V, §8, no4, th. 1, a)) to the function (x,χ)↦g(x)φ(χ)⟨χ,x⟩ which is integrable on G×G with respect to the product measure dx⊗dx. It follows that
∣∫Gg(x)f(x)dx∣⩽∥FG(g)∥2∥φ∥2=∥g∥2∥φ∥2
by Plancherel's formula. The linear form g↦∫Gfg is therefore continuous on L1(G)∩L2(G), and since L1(G)∩L2(G) is dense in the Hilbert space L2(G), it follows that f belongs to L2(G).
Applying then Theorem 1 of II, p. 215, one obtains on the other hand
∫Gg(x)f(x)dx=∫GFG(g)(χ)FG(f)(χ)dx(χ)
=∫GFG(g)(χ)FG(f)(χ)dx(χ)
for every g∈L2(G). Comparing with (28), one concludes that φ=FG(f) in L2(G), since A(G) is contained in L1(G)∩L2(G) and A(G) is dense in L2(G) (Lemma 5 of II, p. 215).
Proposition 12 (Fourier inversion formula)
Let f∈L2(G) be such that FG(f)∈L1(G). Then one has f = FG(FG(f))∘η in L2(G). In other words, for almost every x∈G, one has
f(x)=∫G⟨x,x⟩FG(f)(x)dx
The function φ=FG(f) belongs to L1(G)∩L2(G), and the desired formula is obtained by applying the lemma.
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.