Let $(a_n)$ be decreasing sequence of positive real numbers such that the square summabiliity assumption $$\sum_{n} a_n^2 < +\infty. \tag{SSA}$$ holds. It is known that in this case $na_n^2\to 0$ (see, e.g., https://www.encyclopediaofmath.org/index.php/Series) and hence $$\sqrt{n}a_n\to 0.$$ In turn, this implies (using Cauchy-Schwarz) that $a_n(a_1+\cdots+a_n)\leq a_n\sqrt{n}(a_1^2+\cdots+a_n^2)\to 0$ and hence that $$a_n(a_1+\cdots+a_n)\to 0. \tag{*} $$

My question is: What happens if we drop the assumption that $(a_n)$ be decreasing? Is (*) still true just assuming the square summability assumption (SSA)? Any reference or counterexample would be much appreciated.


1 Answer 1


Pick $a_n = \frac{1}{\log n}, n=2^{2^k}, k \ge 1, a_n =\frac{1}{n}$ otherwise, Clearly $\sum a_n^2 < \infty$ but for every $n=2^{2^k}, a_1+a_2+...a_n \ge a_1+a_3+...a_{n-1} > \frac{1}{2} \log (n-1)$ so $a_n(a_1+a_2+...a_n) > \frac{1}{2}$ for those infinitely many $n's$, so the answer to the question is negative

  • $\begingroup$ Awesome, many thanks! I think you can replace $\log(n-1)$ even by $\log(n)$. $\endgroup$
    – max_zorn
    Sep 24, 2019 at 15:36
  • $\begingroup$ Sure as actually the modified term is bigger than the corresponding harmonic one but I had the idea using odd terms in mind when writing and then got lazy to edit as it doesn't matter for the problem itself. $\endgroup$
    – Conrad
    Sep 24, 2019 at 17:16

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.