Summary
The preceding lesson proved that every absolutely convergent improper integral converges, and announced that the converse is false. This lesson studies the integrals that converge without converging absolutely, called conditionally convergent. Its central example is : the integrals over the arcs of the graph alternate in sign and decrease, and they cancel enough for the integral to converge, although those of its absolute value grow like the sums of the reciprocals of the natural numbers, which are unbounded. Two tests decide that convergence without computing the integral: Dirichlet’s, for the product of a monotonic function that tends to and another one whose integrals are bounded, and Abel’s, for the product of a monotonic bounded function and another one with a convergent integral. Both are proved with the second mean value theorem for integrals and the Cauchy criterion, without requiring derivatives of either function. A change of variable, proved for improper integrals, brings within reach of those tests integrals such as the Fresnel integrals, , whose integrand does not even tend to . The value of , which is , is left for the following lesson.
Learning objectives
- Distinguish conditional convergence from absolute convergence, and prove that converges and that the integral of its absolute value diverges.
- Prove Dirichlet’s and Abel’s tests with the second mean value theorem for integrals, with the tail bound given by the former, and recognize which hypotheses cannot be omitted.
- Prove the change of variable in improper integrals and use it to bring integrals such as the Fresnel integrals within reach of those tests.
- Decide for which values of a parameter an oscillating integral converges, absolutely or conditionally, and compute one of them with the Wallis product.
Conditional convergence
The lesson on improper integrals and convergence tests proved, with the theorem that absolute convergence implies convergence, that if converges, converges too. Its comparison tests hold only for nonnegative functions, and for a function of variable sign the only tool was to compare its absolute value. Functions that oscillate with decreasing amplitude escape that method: the integral of over each arc between two consecutive zeros has the sign of the arc, and the absolute values of these integrals decrease like the reciprocals of the natural numbers. The conventions of that lesson are kept: is a real number or the symbol , abbreviates or , and everything proved at the right endpoint holds at the left one, by the lemma on the reflection of an improper integral.
Let be a function defined on and integrable on for every with . We say that converges conditionally if it converges and diverges. Likewise at the left endpoint and, splitting the integral at a point, on an open interval.
By the theorem on integrals of nonnegative functions, from the same lesson, converges or diverges to ; therefore a convergent integral converges absolutely or conditionally, and not in both ways. If , then , and no integral of a nonnegative function converges conditionally: conditional convergence requires the positive and negative parts of the integrand to cancel. In fact, if converges conditionally, the integrals of and of , integrable on each by the corollary on the maximum and minimum of two integrable functions, from the lesson on the properties of the definite integral, both diverge: their difference is and their sum is , and if one of them converged, the other would converge by linearity, and with them .
Loading the content…
Completed
Sign in to save your progress