49. Let T=T1, where Ta is defined in problem II.13.48. Show that
T=dN−2,
where N∼N(0,1) is a standard Gaussian random variable. Show also that the Laplace transform is given by
Ee−2λ2T=Ee−2N2λ2=e−λ,λ⩾0
and the Fourier transform is given by
EeitT=EeN2it=exp{−∣t∣1/2(1−i∣t∣t)},t∈R
Thus, the random variable N−2 can be considered as a constructively defined random variable with a stable distribution (with parameters α=1/2,β=0,θ=−1,d=1).
Remark. The results of II.13.48 and II.13.49 imply Lévy's formula
∫0∞2πt3ae−2ta2e−λtdt=e−a2λ
This one wants a proof. Work it on paper, read the official solution, then mark
yourself honestly — the ladder only means something if the record is true.
Official solution
Solution. Reasoning as in the solution of II.13.48, we find that P(T1⩽t)=2P(Bt⩾1)=2P(tN⩾1)=P(tN2⩾1)=P(N−2⩽t). Hence,
where the last equality follows from the fact that the Boolean transformation preserves the Lebesgue measure on R (see problem II.6.103).
Consider the function of a complex variable f(z)=Ee−2z2T, defined in the domain D={z∈C:Rez2⩾0}, where, in particular, e−z2T/2⩽1. It is now sufficient to show that the function f is holomorphic inside this domain and continuous on the boundary. Indeed, in this case, by the uniqueness theorem, f(z) necessarily equals e−z in the domain D, since f(z)=e−z for z∈R. The continuity of the function f in D follows from the Lebesgue dominated convergence theorem. We will prove that f is differentiable with respect to z inside D. Since