MathOverflow is a question and answer site for professional mathematicians. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Here's the setup: let $N(m)$, $m = 1,2, \cdots$ be a sequence of nonnegative integers for which

$(*)~~~~~\sum_{m = 1}^{\infty} \frac{N(m)}{m} < \infty$

Now let $C>0$ be any constant. My question is, does there exist a constant $K < \infty$ such that for all $n \geq 1$,

$\left(1+C \sum_{m = n+1}^{\infty} \frac{N(m)}{m}\right)^{\sum_{m = 1}^n N(m)} < K$

I already know the bound $\sum_{m = 1}^n N(m) < n$ for $n$ sufficiently large (uses Abel summation), but for some reason I can't seem to make the above work. For instance, one can check that

$\lim\limits_{n \rightarrow \infty} (1 + \frac{1}{\log n})^n = \infty$

but I'm not sure if this is a death knell for what I want to show. My confusion stems from the idea that the condition $(*)$ seems to suggest that "most" of the $N(m)$ are $0$, hence $\sum_{m = 1}^n N(m)$ ought to be much, much smaller than $n$.

Can anybody think of a counterexample?

share|cite|improve this question
Where does the question arise? – Yemon Choi Sep 27 '12 at 23:20
I am very sleep-deprived right now but one idea that comes to mind is to fix an infinite subset S of the natural numbers and take $N(\cdot)$ to be the indicator function/characteristic function of S. If you take S to be the "nearest integer perturbation" of the set $\{ k^{3/2} : k \in {\bf N}\}$ then some back of envelope calculations suggest $\sum_{m=1}^n N(m)$ is approx $n^{2/3}$ while $\sum_{m=n+1}^\infty N(m)/m$ is $O(n^{-1/3})$. Which looks like it would give a counter-example... however, this is all being done through a mental fog & I may have made a stupid error – Yemon Choi Sep 27 '12 at 23:31
up vote 1 down vote accepted

Counterexample: Define $N(m)$ by $$N(m):=\begin{cases} \lfloor n!/2\rfloor &\text{if } m=(n+2)!-1\\\\ \lceil n!/2\rceil &\text{if } m=(n+2)!\\\\ 0 &\text{otherwise}\\\\ \end{cases}$$ and note that $\sum\limits_{m=1}^\infty \frac{N(m)}{m}<\sum\limits_{n=1}^\infty \frac{1}{n^2}=\frac{\pi^2}{6}$, yet $$\begin{align} \left(1+\sum\limits_{m=(n+2)!}^{\infty} \frac{N(m)}{m}\right)^{\sum\limits_{m=1}^{(n+2)!-1} N(m)} &\ge \left(1+\frac{1}{2(n+1)(n+2)}\right)^{n!/2}\\\\ &\ge 1+\frac{n!/2}{2(n+1)(n+2)}\to \infty\\\\ \end{align}$$ thus clearly no upper bound can exist.

share|cite|improve this answer
Nice - looks similar to the construction I outlined. – Yemon Choi Sep 28 '12 at 4:52
Yes, I think basically any function which has sparse, sharp peaks which come close (in the sense of $1/n^{1+\epsilon}$) to their index will provide a counterexample. – Alex Becker Sep 28 '12 at 5:12
Yep, that's a counterexample. Oh well; this came up when I was trying to find a bound on the total variation of a sequence of signed measures. I think this is true if $\sum_{m = 1}^n N(m) \leq C \sqrt{n}$. Then the tail of the sum looks like $1/\sqrt{n}$ by using Abel summation. Do you think any weaker bound (e.g. $\sum_{m = 1}^n N(m) = O(n^{1-\epsilon}$ for $\epsilon < 1/2$) would work? – A Blumenthal Sep 28 '12 at 21:38
Sorry, Yemon Choi's earlier reply answers my question in the negative. – A Blumenthal Sep 28 '12 at 21:39

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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