If $(a_n)_{n \ge 1}$ is a non-negative sequence s.t., $$\sum\limits_{n = c_k}^\infty \frac{a_n}{\log^{(k)} n} < \infty, \, \forall k \ge 1 \overset{?}{\implies} \sum\limits_{n \ge 1} a_n < \infty$$ where, $\log^{(k)} = \underbrace{\log \circ \log \circ \cdots \circ \log}_{k \text{ times }}$ and $c_k$ is large say $2^{2^{ \cdots ^{2}}}$ ($k$ power tower of $2$).

Perhaps it is more pertinent to ask some sort of characterization of a subset $S$ of sequences in $c_0$ , s.t., if $\displaystyle \sum\limits_{n \ge 1} a_nu_n < \infty$ for all $(u_n) \in S$ implies $(a_n) \in \ell^1$.

A softer related question is as follows, if $(a_n)_{n \ge 1}$ is a non-negative sequence s.t., $$\sum\limits_{n=1}^\infty \frac{a_n}{n^s} < \infty, \, \forall s > 0 \overset{?}{\implies} \sum\limits_{n \ge 1} a_n < \infty$$

i.e, if the abscissa of convergence of the Dirichlet series $\displaystyle F(s) = \sum\limits_{n=1}^\infty \frac{a_n}{n^s}$ is $\sigma_c = 0$, then does it imply $\displaystyle \sum\limits_{n \ge 1} a_n < \infty$?

Apparently Landau's theorem guarantees that if $\sigma_c \ge 0$ then $\displaystyle \limsup\limits_{x \to \infty} \frac{\log |A(x)|}{\log x} = \sigma_c$ where, $\displaystyle A(x) = \sum\limits_{n < x} a_n$ is the summatory function.

Also asked on m.se.

  • 2
    $\begingroup$ I believe it is not difficult to find counterexamples to your questions. $\endgroup$ – Piotr Hajlasz Feb 13 at 17:17
  • $\begingroup$ @PiotrHajlasz Indeed I can guess the collection of sequences $(\log^{(k)} n)_{n \ge c_k}$ is definitely inadequate to force $(a_n)$ to be in $\ell^1$ (I'd naively guess we'd need the set $S$ to have atleast Baire class $2$). But I wish to see how'd we think about cooking up a counterexample in this case. $\endgroup$ – r9m Feb 13 at 17:25

Let $l_k:=\ln^{(k)}$. Let $(n_k)$ be any strictly increasing sequence of natural numbers such that $l_k(n_k)>k^2$. Let $a_n:=1$ if $n=n_k$ for some $k$ and let $a_n:=0$ otherwise. Then $\sum_n a_n=\infty$, but $\sum_n a_n/l_j(n)=\sum_k 1/l_j(n_k)<\infty$ for all $j$, since $l_j(n_k)\ge l_k(n_k)>k^2$ for $k\ge j$.

So, the implication in your main question is false in general. As for the "softer related question", just let $a_n=1/n$ to disprove that implication as well.


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.