We proceed by induction on n. The theorem holds for n=1. For arbitrary n one can, by the induction hypothesis, apply it to the derivative f′ of f: for any ε>0 there is an h>0 such that, if one puts
We apply the mean value theorem (I, p. 15, th. 2) on the interval with endpoints a,x (with ∣x−a∣≤h) to the vector function g and to the real increasing function equal to ε∣y−a∣n/n if x>a, and to −ε∣y−a∣n/n if x<a; it follows that ∥g(x)∥≤ε∣x−a∣n/n, which proves the theorem.
where u(x) approaches 0 as x approaches a while remaining in I; this formula is called Taylor’s formula of ordern at the point a, and the right-hand side of (8) is called the Taylor expansion of ordern of the function f at the point a. The last term rn(x)=u(x)(x−a)n/n! is called the remainder in the Taylor formula of order n.
When f has a derivative of ordern+1 on I, one can estimate ∥rn(x)∥ in terms of this n+1th derivative, on all of I, and not just on an unspecified neighbourhood of a:
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.