Lemma. The scalar Gaussian integral [fdip-002S]

For real \(a\), consider the extended nonnegative Lebesgue integral \[I(a)=\int _{\mathbb R}e^{-a x^2/2}\,dx.\] For \(a>0\) it is finite and positive, and \[I(a)=\sqrt {2\pi /a},\qquad I(1)^2=2\pi .\] For \(a\leq 0\) it is infinite.

The calculation adapts Keith Conrad's improper-integral argument in [conradGaussianIntegral, Section 2, Second proof: Another change of variables, PDF pp. 1-2] to Lebesgue integration. The explicit tail bound and nonnegative Tonelli justify the adaptation; Tonelli is [tao2011measure, Theorem 1.7.15, pp. 200-201, preliminary version]. Mathlib's integrable_exp_neg_mul_sq and integral_gaussian support the real one-variable result with coefficient \(b=a/2>0\). The value theorem uses a totalized integral; its value at a nonintegrable parameter does not mean the nonnegative normalizer is finite.