For every real x>0 one has
∫1x+1logΓ(t) dt=x(logx−1)+21log2π (Raabe’s integral).
First we establish formula (13) for x=0. Since logΓ(x)∼logx1 as x tends to 0, the integral ∫01logΓ(x) dx converges. Further, the function logΓ(x) decreases on ]0,1] (VII, p. 310); for every α>0 one thus has
n1k=1∑qlogΓ(nk)⩽∫0αlogΓ(x) dx,
q being the largest integer such that q/n⩽α. Since ∫0αlogΓ(x)dx tends to 0 with α and also n1∑k=q+1nlogΓ(nk) tends to ∫α1logΓ(x)dx as n tends to +∞ (II, p. 57, prop. 5) one has
∫01logΓ(x)dx=n→∞limn1k=1∑nlogΓ(nk).
But, by (12), the right-hand side of this formula is the limit of
2nn−1log2π−21nlogn,
whence
∫01logΓ(x)dx=21log2π. (14)
Next we remark that from the identity
logΓ(x+1)=logΓ(x)+logx
one deduces, on integrating, that for x>0
∫0xlogΓ(t+1)dt=∫0xlogΓ(t)dt+∫0xlogtdt.
But the integral on the left-hand side is also equal to ∫1x+1logΓ(t)dt. Thus, by (14),
∫1x+1logΓ(t)dt=∫0xlogtdt+21log2π=x(logx−1)+21log2π.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.
Nothing.