Bourbaki

Proposition 1

Tag 08IX, permanent. Cite it as [Bourbaki, Tag 08IX].

Functions of a Real Variable, chapter VII, § 1. THE GAMMA FUNCTION IN THE REAL DOMAIN

There exists one and only one convex function gg defined on ]0, +∞[ that satisfies the equation (3) and takes the value 0 for x=1x = 1.

First we show that if there is a function gg satisfying the conditions stated then it is well-determined on the interval ]0, 1], and consequently on the interval ]0, +∞[. Indeed, for every integer n>1n > 1 the gradient of the line joining the point (n,g(n))(n, g(n)) to the point (x,g(x))(x, g(x)) is an increasing function of xx, since gg is convex (I, p. 27, prop. 5); one thus must have, for 0<x≤10 < x \leq 1,

g(n−1)−g(n)(n−1)−n≤g(n+x)−g(n)(n+x)−n≤g(n+1)−g(n)(n+1)−n \frac{g(n-1) - g(n)}{(n-1) - n} \leq \frac{g(n+x) - g(n)}{(n+x) - n} \leq \frac{g(n+1) - g(n)}{(n+1) - n}

that is, by (3),

xlog⁡(n−1)≤g(x+n)−g(n)≤xlog⁡n. x \log(n-1) \leq g(x+n) - g(n) \leq x \log n.

Now, by (3),

g(x+n)−g(n)=g(x)+log⁡x+∑k=1n−1(log⁡(x+k)−log⁡k). g(x+n) - g(n) = g(x) + \log x + \sum_{k=1}^{n-1} (\log(x+k) - \log k).

Moreover, one can write log⁡n=∑k=2nlog⁡kk−1\log n = \sum_{k=2}^n \log \frac{k}{k-1} so the inequality (4) can be written

x∑k=2n−1log⁡kk−1≤g(x)+log⁡x+∑k=2n(log⁡(x+k−1)−log⁡(k−1)) x \sum_{k=2}^{n-1} \log \frac{k}{k-1} \leq g(x) + \log x + \sum_{k=2}^n (\log(x+k-1) - \log(k-1))
≤x∑k=2nlog⁡kk−1. \leq x \sum_{k=2}^n \log \frac{k}{k-1}.

Let us put, for every n≥2n \geq 2,

un(x)=xlog⁡nn−1−log⁡(x+n−1)+log⁡(n−1) u_n(x) = x \log \frac{n}{n-1} - \log(x+n-1) + \log(n-1)

(5)

and

gn(x)=−log⁡x+∑k=2nuk(x). g_n(x) = -\log x + \sum_{k=2}^n u_k(x).

For 0<x≤10 < x \leq 1 one then has

gn(x)−xlog⁡nn−1≤g(x)≤gn(x). g_n(x) - x \log \frac{n}{n-1} \leq g(x) \leq g_n(x).
(6)

Since log⁡nn−1\log \frac{n}{n-1} tends to 0 as nn tends to +∞+\infty one deduces from (6) that if a solution gg exists then it is necessarily equal, on ]0,1]]0, 1], to the limit of the gn(x)g_n(x).

Now one deduces immediately from (5) that for xx fixed and >0> 0 one has

un(x)=−xlog⁡(1−1n)−log⁡(1+x−1n)+log⁡(1−1n)∼x(x−1)2n2 u_n(x) = -x \log \left( 1 - \frac{1}{n} \right) - \log \left( 1 + \frac{x-1}{n} \right) + \log \left( 1 - \frac{1}{n} \right) \sim \frac{x(x-1)}{2n^2}
as nn tends to +∞+\infty, which proves that the series with general term un(x)u_n(x) converges for every x>0x > 0. Each of the functions un(x)u_n(x) being convex on ]0,+∞[]0, +\infty[, as is −log⁡x-\log x, the function g(x)=−log⁡x+∑n=2∞un(x)g(x) = -\log x + \sum_{n=2}^{\infty} u_n(x) is convex on this interval (I, p. 27, prop. 2 and prop. 4); finally, one has un(1)=0u_n(1) = 0, whence g(1)=0g(1) = 0, and
un(x+1)=un+1(x)+x(log⁡nn−1−log⁡n+1n) u_n(x+1) = u_{n+1}(x) + x \left( \log \frac{n}{n-1} - \log \frac{n+1}{n} \right)
whence
g(x+1)=−log⁡(x+1)+xlog⁡2+∑n=3∞un(x)=log⁡x+g(x); g(x+1) = -\log(x+1) + x \log 2 + \sum_{n=3}^{\infty} u_n(x) = \log x + g(x);
in other words, gg satisfies equation (3) of VII, p. 306.

Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.

Cited by

Nothing.

Cites

Transcribed from the printed volume. Not translated.