Bourbaki

Theorem 1

Tag 0872, permanent. Cite it as [Bourbaki, Tag 0872].

Functions of a Real Variable, chapter II, § 1. PRIMITIVES AND INTEGRALS

Let AA be a set filtered by a filter F\mathfrak{F}, and (fα)α∈A(f_\alpha)_{\alpha \in A} a family of vector functions with values in a complete normed space EE over R\mathbf{R}, defined on an interval I⊂RI \subset \mathbf{R}: for each α∈A\alpha \in A let gαg_\alpha be a primitive of fαf_\alpha. We suppose that: 1. with respect to the filter F\mathfrak{F} the functions fαf_\alpha converge uniformly on every compact subset of II to a function ff; 2. there is a point a∈Ia \in I such that, with respect to the filter F\mathfrak{F}, the family (gα(a))(g_\alpha(a)) has a limit in EE. Under these hypotheses the functions gαg_\alpha converge uniformly (with respect to F\mathfrak{F}) on every compact subset of II to a primitive gg of ff.

By the remark at the beginning of this subsection we can restrict ourselves to the case where II is a compact interval.

First let us show that the gαg_\alpha converge uniformly on II to a continuous function gg. By hypothesis, for every ε>0\varepsilon > 0 there is a set M∈FM \in \mathfrak{F} such that, for any two indices α,β\alpha, \beta belonging to MM, one has ∥fα(x)−fβ(x)∥≤ε\|f_\alpha(x) - f_\beta(x)\| \leq \varepsilon for each x∈Ix \in I; in consequence one has (I, p. 15, th. 2)

∥gα(x)−gβ(x)−(gα(a)−gβ(a))∥≤ε∣x−a∣≤εl \|g_\alpha(x) - g_\beta(x) - (g_\alpha(a) - g_\beta(a))\| \leq \varepsilon |x - a| \leq \varepsilon l

where ll denotes the length of II; since by hypothesis gα(a)g_\alpha(a) approaches a limit with respect to F\mathfrak{F}, it follows from the Cauchy criterion that the gαg_\alpha converge uniformly on II. It remains to show that the limit gg of the gαg_\alpha is a primitive of ff.

For each integer n>0n > 0 let αn\alpha_n be an index such that ∥f(x)−fαn(x)∥≤1/n\|f(x) - f_{\alpha_n}(x)\| \leq 1/n on II; it is clear that the sequence (fαn)(f_{\alpha_n}) converges uniformly to ff and that the sequence (gαn)(g_{\alpha_n}) converges uniformly to gg on II. Let HnH_n be the countable subset of II where fαnf_{\alpha_n} is not the derivative of gαng_{\alpha_n}, and let HH be the union of the HnH_n, which is thus a countable subset of I; we shall see that at every point x∈Ix \in I not belonging to H the function g has a derivative equal to f(x)f(x). Indeed, one sees as above that for every m≥nm \geq n and every y∈Iy \in I one has

∥gαm(y)−gαm(x)−(gαn(y)−gαn(x))∥≤2n∣y−x∣. \| g_{\alpha_m}(y) - g_{\alpha_m}(x) - (g_{\alpha_n}(y) - g_{\alpha_n}(x)) \| \leq \frac{2}{n} |y - x|.

Letting mm increase indefinitely one also has

∥g(y)−g(x)−(gαn(y)−gαn(x))∥≤2n∣y−x∣ \| g(y) - g(x) - (g_{\alpha_n}(y) - g_{\alpha_n}(x)) \| \leq \frac{2}{n} |y - x|

for every y∈Iy \in I; now, there exists an h>0h > 0 such that, for ∣y−x∣≤h|y - x| \leq h and y∈Iy \in I, one has ∥gα(y)−gαn(x)−fαn(x)(y−x)∥≤∣y−x∣/n\| g_\alpha(y) - g_{\alpha_n}(x) - f_{\alpha_n}(x)(y - x) \| \leq |y - x| / n; since, on the other hand, we have ∥f(x)−fαn(x)∥≤1/n\| f(x) - f_{\alpha_n}(x) \| \leq 1 / n, we finally obtain

∥g(y)−g(x)−f(x)(y−x)∥≤4n∣y−x∣ \| g(y) - g(x) - f(x)(y - x) \| \leq \frac{4}{n} |y - x|

for y∈Iy \in I and ∣y−x∣≤h|y - x| \leq h, which completes the proof.

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

Cited by

Cites

Transcribed from the printed volume. Not translated.