Bourbaki

Example

Tag 088G, permanent. Cite it as [Bourbaki, Tag 088G].

Functions of a Real Variable, chapter II, § 2. INTEGRALS OVER NON-COMPACT INTERVALS

i) The integral ∫1+∞dt/t2\int_1^{+\infty} dt/t^2 is convergent and equal to 1, for

∫1xdtt2=1−1x. \int_1^x \frac{dt}{t^2} = 1 - \frac{1}{x}.

2) The integral ∫01dt/t\int_0^1 dt / \sqrt{t} is convergent and equal to 2, for

∫1xdtt=2(1−x)forx>0. \int_1^x \frac{dt}{\sqrt{t}} = 2(1 - \sqrt{x}) \quad \text{for} \quad x > 0.

3) Let (un)n≥1(\mathbf{u}_n)_{n \geq 1} be an infinite sequence of points of E, and let f\mathbf{f} be the step function defined on the interval [1,+∞[[1, +\infty[ by the conditions: f(x)=un\mathbf{f}(x) = \mathbf{u}_n for n≤x<n+1n \leq x < n+1. Then for the integral ∫1+∞f(t) dt\int_1^{+\infty} \mathbf{f}(t)\,dt to be convergent it is necessary and sufficient that the series with general term un\mathbf{u}_n be convergent in E; indeed, one has

∫1nf(t) dt=∑p=1n−1up, \int_1^n \mathbf{f}(t)\,dt = \sum_{p=1}^{n-1} \mathbf{u}_p,
so the condition is necessary; conversely, if the series with general term un\mathbf{u}_n converges in E, then lim⁡n→∞un=0\lim_{n \to \infty} \mathbf{u}_n = 0; now, if n≤x≤n+1n \leq x \leq n+1, one has ∫1nf(t) dt=∑p=1n−1up+un(x−n)\int_1^n \mathbf{f}(t)\,dt = \sum_{p=1}^{n-1} \mathbf{u}_p + \mathbf{u}_n(x-n). so this integral has the limit ∑n=1∞un\sum_{n=1}^{\infty} \mathbf{u}_n when xx tends to +∞+\infty.

It is immediate that if a piecewise regulated function f\mathbf{f} admits an integral over I then the formulae (4) to (9) of II, p. 59 remain valid. Similarly, formula (10) of II, p. 59 extends in the following manner: f\mathbf{f} and g\mathbf{g} are assumed to be primitives of the regulated functions f′\mathbf{f}' and g′\mathbf{g}' on ]a, b[]a,\ b[, and one denotes by [f.g]∣ab[\mathbf{f}.\mathbf{g}]|_a^b the limit (if it exists) of [f.g]∣xy[\mathbf{f}.\mathbf{g}]|_x^y as (x, y)(x,\ y) tends to (a, b)(a,\ b) (with a<x≤y<ba < x \leq y < b); then, if two of the (three) expressions [f.g]∣ab[\mathbf{f}.\mathbf{g}]|_a^b, ∫ab[f(t).g′(t)] dt\int_a^b [\mathbf{f}(t).\mathbf{g}'(t)]\,dt, and ∫ab[f′(t).g(t)] dt\int_a^b [\mathbf{f}'(t).\mathbf{g}(t)]\,dt have a meaning, then so has the third, and the formula (10) of II, p. 59 is valid.

Finally, let ff be a real function which is defined and continuous on I=]a, b[I = ]a,\ b[, and is the primitive of a regulated function f′f' on ]a, b[]a,\ b[; let on the other hand gg be a continuous vector function on an open interval J containing f(I)f(I); if the function g(f(x))f′(x)g(f(x))f'(x) admits an integral over I, and if ff tends to a limit (finite or not) at the points aa and bb, then gg admits an integral from f(a+)f(a+) to f(b−)f(b-), and one has the formula

∫abg(f(t))f′(t) dt=∫f(a+)f(b−)g(u) du. \int_a^b g(f(t))f'(t)\,dt = \int_{f(a+)}^{f(b-)} g(u)\,du.
Indeed, if (x, y)(x,\ y) tends to (a, b)(a,\ b), then (f(x), f(y))(f(x),\ f(y)) tends to (f(a+), f(b−))(f(a+),\ f(b-)) by hypothesis; it suffices to apply formula (12) of II, p. 60 between xx and yy, and to pass to the limit to obtain (1).

Given a regulated function f\mathbf{f} on a non-compact interval I⊂RI \subset \mathbf{R}, with endpoints aa and bb (a<ba < b), the condition for f\mathbf{f} to have an integral over I can be presented in the following manner. The compact intervals J⊂IJ \subset I form a directed set K(I)\mathscr{K}(I) with respect to the relation ⊂1\subset ^1, for if [α,β][ \alpha , \beta ] and [γ,δ][ \gamma , \delta ] are two compact intervals contained in II, and if one puts λ=min⁡(α,γ)\lambda = \min(\alpha, \gamma), μ=max⁡(β,δ)\mu = \max(\beta, \delta), then the interval [λ,μ][ \lambda , \mu ] is contained in II and contains the two intervals considered. For each compact interval J=[α,β]J = [ \alpha , \beta ] contained in II, let us put

∫Jf(t) dt=∫αβf(t) dt; \int_J f(t) \, dt = \int_\alpha^\beta f(t) \, dt;

for ff to admit an integral over II it is necessary and sufficient that the map J↦∫Jf(t) dtJ \mapsto \int_J f(t) \, dt have a limit with respect to the directed set K(I)\mathfrak{K}(I); this limit is then the integral ∫abf(t) dt\int_a^b f(t) \, dt, which we again denote by ∫If(t) dt\int_I f(t) \, dt.

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.