I'm trying to show that then $\Sigma a_n$ converges iff $\Sigma 2^n a_{2^n}$ converges if $a_n$ is any monotone sequence.

I am assuming I have to prove the cases:

  • $a_n$ is positive and increasing
  • $a_n$ is negative and increasing
  • $a_n$ is positive and decreasing
  • $a_n$ is negative and decreasing

Here is what I have worked out from some of the comments below:

Assume $a_n$ is an increasing sequence. If $a_n$ is positive, then $\lim a_n \neq 0$ and we have divergence.

I don't really understand why this is. Is it because $a_n$ is unbounded and so it diverges to $\infty$? I don't know a rigorous way to prove this. I'm assuming I have to make use of the theorem that monotonic $s_n$ converges iff it is bounded.

If $a_n$ is negative, then we have that $\lim a_n = 0$

Again, I don't know how to rigorously prove this. How do I find a bound for $a_n$?

Let $s_n = a_1 + a_2 + ... + a_n$ and $t_n = a_1 + 2 a_2 + ... + 2^k a_{2^k}$. For $n<2^k$, we have $s_n \geq a_1 + (a_2 + a_3) + ... _ (a_{2^k}+...+a_{2^{k+1}-1} \geq a_1 + a_2 + ... + 2^{k}a_{2k} = t_k $

And so $s_n \geq t_k$.

Similar proof for $2s_{n} \leq t_k$

Someone below has already proved nonnegative decreasing case. I'm assuming that:

If $a_n$ was decreasing and negative, then $\lim a_n \neq 0$ and we have divergence.

Again, I'm not sure how to prove this.


Consider the series $\sum a_n $ where $(a_n)$ is decreasing and positive. Also, consider the series $\sum 2^n a_{2^n}$. Consider their partial sums $(s_n), (t_n)$

Note if $n < 2^n$, then $s_n = a_1 + a_2 + ... + a_n \leq a_1 + (a_2+a_3) + (a_4 + a_5 + a_6 + a_7) + ... + (a_{2^n} + ... + a_{2^{n+1}-1}) \leq a_1 + 2a_2 + ... + 2^n a_{2n} = t_n \implies \boxed{ s_n \leq t_n }$

If $2^n < n$, then $s_n \geq a_1 + a_2 + (a_3 + a_4) + ... + (a_{2^{n-1}+1} + ... + a_{2^n} ) \geq \frac{1}{2} a_1 + a_2 + 2 a_4 + ... + 2^{n-1}a_n = \frac{1}{2} t_n \implies \boxed{ 2s_n \geq t_n }$

Since series of nonnegative terms converges iff its sequence of partial sums are bounded, then we are done.

  • $\begingroup$ Do we have the consider when $a_n$ is increasing and positive? I think someone below explained why negative terms will result in a divergence. Sorry if this seems like a dumb question. $\endgroup$
    – Nikitau
    Nov 9 '16 at 1:59

Plenty of information at the Wikipedia page on this very theorem:


  • $\begingroup$ I appreciate the link, but it states that the test holds true for nonnegative, decreasing sequences. I want to somehow show that it's true for any monotone sequence, but I don't see how this could be true given the conditions for the proof of Cauchy Condensation Test. $\endgroup$
    – Nikitau
    Nov 10 '16 at 20:34
  • $\begingroup$ If the sequence is monotone then $a_n \to 0$ and $a_{2^n} \to 0$ are equivalent. So the only cases that potentially converge are negatives of each other (positive decreasing, and negative increasing, both converging to 0). $\endgroup$
    – zyx
    Nov 11 '16 at 4:22
  • $\begingroup$ Could you explain why the positive increasing would not result in convergence? It might be a silly question, but I'm not sure how to prove it. $\endgroup$
    – Nikitau
    Nov 11 '16 at 5:35
  • $\begingroup$ The terms of the sum have to go to 0 in order for the sum to converge. $\endgroup$
    – zyx
    Nov 11 '16 at 5:41

This is called the Cauchy Condensation Test. Note the convergence or divergence of $\sum_na_n$ is unaffected by changing $a_n$ for any finite set of $n.$ So if $a_n<0$ for only finitely many n, we may substitute $0$ for $a_n$ whenever $a_n<0$ without affecting convergence or divergence. (Similarly if $a_n>0$ for finitely many $n.$)

Also it is sufficient that $(a_n)_n$ is monotonic for all but finitely many n. ... If $(a_n)_n$ is decreasing (for all but finitely many $n$) but has negative values for infinitely many $n$ then for some $r>0$ we have: $a_n<-r$ for all but finitely many $n$ so $\sum_na_n$ diverges, as we do not even have $a_n\to 0.$ ... I see that a proof of the Test has appeared as an answer just now.

Examples of application:

(1). Let $a_n=n^{-K}$ for each $n.$ Then $\sum_na_n$ converges iff $K>1.$

(2).$\sum_{n>1}\frac {1}{n(\log n)^K}$ converges iff $K>1.$

  • $\begingroup$ Thanks for the explanation! This might be a dumb question, but do we have the consider the case where the sequence is positive and increasing? $\endgroup$
    – Nikitau
    Nov 9 '16 at 2:00
  • $\begingroup$ If $(a_n)_n$ is positive and increasing then $\sum_{j=1}^na_j\geq na_1, $ which goes to $\infty$ as $n\to \infty,$ because $a_1>0.$ But the Cauchy Condensation Test works for all monotone sequences. $\endgroup$ Nov 9 '16 at 6:14
  • $\begingroup$ Thanks, but I still think I'm a little confused. Indeed, I am trying to prove that the Cauchy Condensation Test works for any monotone sequences. So in my proof, do I need to include the cases where we have a positive increasing, negative increasing, positive decreasing and negative decreasing sequence? What if we have a sequence with both positive and negative terms? $\endgroup$
    – Nikitau
    Nov 9 '16 at 23:31

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.