5
$\begingroup$

I have been trying to solve the following problem

Assume $f : \mathbb{R} \to (0, \infty)$ is a function satisfying $\lim_{x \to 0} \frac{f(x)}{x} = 0$. Prove that there exists a sequence $(x_n) \subset \mathbb{R}$ such that the series $\sum_{n = 1}^\infty x_n$ diverges, but $\sum_{n = 1}^\infty f(x_n)$ converges.

While the claim seems pretty obvious, I feel I lack the necessary intuition to prove it.

$\endgroup$
6
  • $\begingroup$ To prove the existence of such a series, you should exhibit an example. Do you know about $p$-series $\sum n^{-p}$? $\endgroup$ Commented Jan 16 at 21:29
  • 1
    $\begingroup$ @SammyBlack yes. I tried sequences like $\frac{1}{n}$ because the harmonic series would diverge, but $\sum_{n = 1}^\infty f(1 / n)$ not necessarily, but then a function like $f(x) = x \ln(1 + \lvert x \rvert)$ is a counterexample. The same would happen if we decreased the exponent of $p$-series further. I've also tried to prove it by contradiction, but nothing much has come out of that. $\endgroup$
    – dawmd
    Commented Jan 16 at 21:38
  • 1
    $\begingroup$ not obvious. Where did you get the problem? $\endgroup$
    – Will Jagy
    Commented Jan 16 at 22:14
  • $\begingroup$ @WillJagy it appeared in one of the tests at my uni last year. I'm practising for this year's one $\endgroup$
    – dawmd
    Commented Jan 16 at 22:18
  • $\begingroup$ Does this answer your question? Functions such that $\sum \frac{1}{x_n}$ diverges $\Longrightarrow \sum \frac{1}{x_nf(x_n)}$ diverge $\endgroup$ Commented Jan 16 at 22:24

1 Answer 1

1
$\begingroup$

I don't know how to prove by contradiction, but here's a proof:

For every $k$, there exists $N_k$ so that whenever $0\leq x\leq \frac{1}{N_k}$ we have $f(x) < \frac{1}{k} x$. By enlarging $N_k$ if necessary we may assume $N_{k+1}>2N_k$.

For any sequence $x_n$, let $a_k$ denote the number of $n$'s where $x_n\in (\frac{1}{2N_k},\frac{1}{N_k})$. Then

$$\sum_n f(x_n) \leq \sum_k\frac{a_k}{N_k\cdot k}$$ because $a_k$ of the $x$'s are in $(\frac{1}{2N_k},\frac{1}{N_k})$ and so are at most $\frac{1}{N_k}$, but $f(x) \leq \frac{x}{k}$.

While for similar reasons $$\sum_n x_n \geq \sum_{k} \frac{a_k}{2N_k}$$

Now we just need to choose $a_k$ properly. So we will just choose $a_k = \lfloor N_k/\sqrt{k}\rfloor$, these bracket denotes the largest integer smaller than $\frac{N_k}{\sqrt{k}}$, its a technicality because $a_k$ needs to be an integer. Namely, from any interval between $(\frac{1}{2N_k},\frac{1}{N_k})$ we will arbitrarily choose $a_k$ elements. This guarantees that

$$\sum_n x_n \geq \sum_k (\frac{1}{2\sqrt k}-\frac{1}{N_k})$$ Note that $\sum_k \frac{1}{2\sqrt{k}}$ diverges, on the other hand, since $N_{k+1}>2N_k$ we have $N_k>2^{k-1}$ which is huge, so $\frac{1}{N_k}$ is too small to change the divergence.

On the other hand $$\sum_n f(x_n) \leq \sum_k \frac{1}{k^{1.5}}$$ which conveges.

$\endgroup$
0

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .