Let $a_1,a_2,a_3,\ldots$ be a decreasing sequence of positive numbers. Show that

(a) if $a_1+a_2+\ldots$ converges then $\lim_{n\rightarrow\infty} n a_n=0$

(b) $a_1+a_2+\ldots$ converges if and only if $a_1+2 a_2+4 a_4 +\ldots $ converges.


If $\sum a_i$ converges then for any $\epsilon>0$ there is natural number $N_1$ such that if $n>N_1$ then $$2n \cdot a_{2n} \le\sum_{i=n}^{2n} a_i <\epsilon$$ We cam deal in the same way with the odd terms and for given $\epsilon>0$ find $N_2$ such that $$(2n+1) \cdot a_{2n+1} \le\sum_{i=n+1}^{2n+1} a_i <\epsilon$$ So for every $\epsilon>0$ there is $N=\max\{N_1,N_2\}$ such that whenever $n>N$ then $na_n <\epsilon$.

Is this the correct way of proving that fascinating fact?


If the second series converges then since $a_1,a_2,\ldots$ is decreasing sequence of nonegative numbers, from comparison test we know that the first series converges too.

For the converse I will show that partial sums of the second series are bounded. $$\begin{align*} a_1+\frac12\sum_{i=1}^N2^ia_{2^i}&=a_1+a_2+2a_4+4a_8+\dots+2^{N-1}a_{2^N}\\ &\leq a_1+ a_2+a_3+a_4+a_5+a_6+a_7+a_8+\dots+a_{2^{N-1}+1}+\dots+a_{2^N-1}+a_{2^N}\\ &\leq \sum_{i=1}^\infty a_i<\infty \end{align*}$$

  • 3
    $\begingroup$ Note that condition two is called Cauchy condensation. $\endgroup$ – user296602 Dec 22 '15 at 8:36
  • $\begingroup$ Please do not use pictures for critical portions of your post. Pictures cannot be searched and are inaccessible to those using screen readers. Please edit your question accordingly. $\endgroup$ – gebruiker Dec 22 '15 at 10:19
  • $\begingroup$ @gebruiker Edited. $\endgroup$ – luka5z Dec 22 '15 at 10:29
  • 2
    $\begingroup$ @luka5z You've taken away the fun out of it by solving it yourself, haha... $\endgroup$ – user98186 Dec 22 '15 at 15:42

If you're interested, the Cauchy condensation test is actually a special case of the Schlömilch test. The proof is actually relatively straightforward and very simply stated (with some missing details) here: http://arxiv.org/pdf/1011.4697.pdf.

The generalization covers many of the same series that the integral test would otherwise cover, so its utility is partially diminished.

| cite | improve this answer | |
  • $\begingroup$ Nice, Ill check this out. $\endgroup$ – luka5z Dec 22 '15 at 18:14
  • 1
    $\begingroup$ Yes! It's very interesting. The relevant part is just that s: N->N is monotonically increasing, and that it doesn't grow "too fast," since it is acting as a "weight" and would not always converge if the function grew to quickly. $\endgroup$ – Andres Mejia Dec 22 '15 at 18:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.