Let f be a vector function defined and continuous on a closed bounded interval I=[a, b] of R (where a<b) and taking values in a normed space E over R; let g be a continuous increasing real function on I. Suppose that f and g have right derivatives at all points of the relative complement in [a, b[ of a countable subset A of this interval (allowing gr′(x) to be infinite at some of the points x∈/A), and suppose that at each of these points we have
∥fd′(x)∥≤gr′(x).
Under these hypotheses one has
∥f(b)−f(a)∥≤g(b)−g(a).
The proof proceeds similarly to that of prop. 2. Let ε>0 be arbitrary, and (an) the sequence obtained by enumerating A in some order. Let J be the set of points y∈I such that, for all x such that a≤x≤y one has
∥f(x)−f(a)∥≤g(x)−g(a)+ε(x−a)+εan<x∑2n1;
we shall show that J=I. One sees immediately, as in prop. 2, that J is an interval with left-hand endpoint a; if c is its right-hand endpoint then c∈J; indeed, for all x<c one has (8), and a fortiori
∥f(x)−f(a)∥≤g(c)−g(a)+ε(c−a)+εan<c∑2n1
from which, letting x tend to c in this inequality, it follows from the continuity of f that c satisfies (8).
Let us show that we must have c=b. So suppose that c<b and that moreover c∈/A: then fd′(c) and gr′(c) exist and satisfy (6); suppose in the first place that gr′(c) (which is necessarily ≥0) is finite; then one can always write fd′(c)=ugr′(c), with ∥u∥≤1; since the function f(x)−ug(x) has zero right derivative at the point c there must exist a y such that c<y≤b and such that for c≤x≤y one has
∥f(x)−f(c)−u(g(x)−g(c))∥≤ε(x−c)
from which
∥f(x)−f(c)∥≤g(x)−g(c)+ε(x−c)
and, taking account of (8), in which x is replaced by c,
∥f(x)−f(a)∥≤g(x)−g(a)+ε(x−a)+εan<c∑2n1≤g(x)−g(a)+ε(x−a)+εan<x∑2n1.
Thus one has y∈J, which is a contradiction. Suppose next that c∈/A and that gr′(c)=+∞; then there is a y such that c<y≤b and such that for c≤x≤y one has on the one hand
∥f(x)−f(c)∥≤(∥fd′(c)∥+1)(x−c)
while on the other hand
g(x)−g(c)≥(∥fd′(c)∥+1)(x−c)
from which
∥f(x)−f(c)∥≤g(x)−g(c)
and one concludes as above. Finally, if one has c=ak, then there is a y such that c<y≤b, and such that for c<x≤y one has
∥f(x)−f(c)∥≤2kε
from which, taking account of (8), with x replaced by c,
∥f(x)−f(a)∥≤g(c)−g(a)+ε(c−a)+εan<x∑2n1≤g(x)−g(a)+ε(x−a)+εan<x∑2n1
which again entails a contradiction. The proof finishes as that of prop. 2.
Q.E.D.
Bourbaki puts the proof directly after the statement, so what is above is the statement and its proof together.
Nothing.