Bourbaki

Proposition 6

Tag 089A, permanent. Cite it as [Bourbaki, Tag 089A].

Functions of a Real Variable, chapter II, § 3. DERIVATIVES AND INTEGRALS OF FUNCTIONS DEPENDING ON A PARAMETER

Supposing that the hypotheses of prop. 5 are satisfied, let a(α)a(\alpha), b(α)b(\alpha) be two functions defined on VV, with values in II; if the derivatives a′(α0), b′(α0)a'(\alpha_0),\ b'(\alpha_0) exist and are finite then the function g(α)=∫a(α)b(α)f(t,α) dtg(\alpha) = \int_{a(\alpha)}^{b(\alpha)} f(t, \alpha)\, dt admits at α0\alpha_0 a derivative given by the formula

g′(α0)=∫a(α0)b(α0)fα′(t,α0) dt+b′(α0)f(b(α0),α0)−a′(α0)f(a(α0),α0). g'(\alpha_0) = \int_{a(\alpha_0)}^{b(\alpha_0)} f'_\alpha(t, \alpha_0)\, dt + b'(\alpha_0)f(b(\alpha_0), \alpha_0) - a'(\alpha_0)f(a(\alpha_0), \alpha_0).

Indeed, for all α∈V\alpha \in V distinct from α0\alpha_0 one can write

g(α)−g(α0)α−α0=∫a(α0)b(α0)f(t,α)−f(t,α0)α−α0 dt+1α−α0∫b(α0)b(α)f(t,α) dt \frac{g(\alpha) - g(\alpha_0)}{\alpha - \alpha_0} = \int_{a(\alpha_0)}^{b(\alpha_0)} \frac{f(t, \alpha) - f(t, \alpha_0)}{\alpha - \alpha_0}\, dt + \frac{1}{\alpha - \alpha_0} \int_{b(\alpha_0)}^{b(\alpha)} f(t, \alpha)\, dt
−1α−α0∫a(α0)a(α)f(t,α) dt. - \frac{1}{\alpha - \alpha_0} \int_{a(\alpha_0)}^{a(\alpha)} f(t, \alpha)\, dt.

By prop. 5 of II, p. 74, the first integral on the right-hand side tends to ∫a(α0)b(α0)fα′(t,α0) dt\int_{a(\alpha_0)}^{b(\alpha_0)} f'_\alpha(t, \alpha_0)\, dt as α\alpha tends to α0\alpha_0. In the second integral we replace f(t,α)f(t, \alpha) by f(b(α0),α0)f(b(\alpha_0), \alpha_0) and show that the difference tends to 0. We put M=max⁡(∥f(b(α0),α0)∥,∣b′(α0)∣+1)M = \max (\|f(b(\alpha_0), \alpha_0)\|, |b'(\alpha_0)| + 1); the function b(α)b(\alpha) being continuous at the point α0\alpha_0 and the function ff continuous at the point (b(α0),α0)(b(\alpha_0), \alpha_0), for every ε\varepsilon such that 0<ε<10 < \varepsilon < 1 there exists an r>0r > 0 such that the relation ∣α−α0∣≤r|\alpha - \alpha_0| \leq r entails that ∥f(t,α)−f(b(α0),α0)∥≤ε\|f(t, \alpha) - f(b(\alpha_0), \alpha_0)\| \leq \varepsilon for all tt belonging to the interval with endpoints b(α0)b(\alpha_0) and b(α)b(\alpha); thus one may also suppose that the relation ∣α−α0∣≤r|\alpha - \alpha_0| \leq r entails ∣b(α)−b(α0)α−α0−b′(α0)∣≤ε\left| \frac{b(\alpha) - b(\alpha_0)}{\alpha - \alpha_0} - b'(\alpha_0) \right| \leq \varepsilon.

By the mean value formula (II, p 62, formula (17)) one thus has

∣1α−α0∫b(α0)b(α)f(t,α) dt−b(α)−b(α0)α−α0f(b(α0),α0)∣≤∣b(α)−b(α0)α−α0∣ε \left| \frac{1}{\alpha - \alpha_0} \int_{b(\alpha_0)}^{b(\alpha)} f(t, \alpha)\, dt - \frac{b(\alpha) - b(\alpha_0)}{\alpha - \alpha_0} f(b(\alpha_0), \alpha_0) \right| \leq \left| \frac{b(\alpha) - b(\alpha_0)}{\alpha - \alpha_0} \right| \varepsilon

and consequently

∣1α−α0∫b(α0)b(α)f(t,α) dt−b′(α0)f(b(α0),α0)∣≤2Mε \left| \frac{1}{\alpha - \alpha_0} \int_{b(\alpha_0)}^{b(\alpha)} f(t, \alpha)\, dt - b'(\alpha_0) f(b(\alpha_0), \alpha_0) \right| \leq 2M \varepsilon

which shows that 1α−α0∫b(α0)b(α)f(t,α) dt\frac{1}{\alpha - \alpha_0} \int_{b(\alpha_0)}^{b(\alpha)} f(t, \alpha)\, dt tends to b′(α0)f(b(α0),α0)b'(\alpha_0) f(b(\alpha_0), \alpha_0). In the same way one shows that 1α−α0∫a(α0)a(α)f(t,α) dt\frac{1}{\alpha - \alpha_0} \int_{a(\alpha_0)}^{a(\alpha)} f(t, \alpha)\, dt tends to a′(α0)f(a(α0),α0)a'(\alpha_0) f(a(\alpha_0), \alpha_0).

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.