2
$\begingroup$

Let $p_1=2<p_2<p_3<\dots$ are consecutive prime numbers.

Let $$S_n=\sum_{k_1=0}^{\infty}\sum_{k_2=0}^{\infty}\dots \sum_{k_n=0}^{\infty}\frac{1}{p_1^{k_1}p_2^{k_2}\dots p_n^{k_n}}$$

It is easy to check that $S_1=2,S_3=3$ and in general $S_n=\frac{1}{1-\frac{1}{p_1}}\frac{1}{1-\frac{1}{p_2}}\dots \frac{1}{1-\frac{1}{p_n}}$

Then if we can show that as $n\to\infty$, $S_n$ diverges to infinity we can say $\sum_{n=1}^{\infty} \frac{1}{n}$ also diverges to infinity.

If we can show that $S_n\thicksim f(n)$, where $f(n)$ increases to infinity as $n\to\infty$, then we are done.

What is this $f(n)$?

Is there any other way to show $\sum_{k=1}^{\infty}\frac{1}{k}$ diverges using $S_n$?


But I have no idea about the following problem:

$$T_n=\sum_{k_1\neq k_2\neq \dots \neq k_n}^{\infty}\frac{1}{p_1^{k_1}p_2^{k_2}\dots p_n^{k_n}}$$

Here $k_1\neq k_2\neq \dots\neq k_n$ means they are distinct($n!$ values).

How should we proceed to in this case?

$\endgroup$
  • $\begingroup$ You don't really need very precise estimates, but if you want you can show $S_n \sim e^\gamma \log(n)$, see e.g. math.stackexchange.com/questions/1341369/… $\endgroup$ – Winther Jul 23 '17 at 14:26
  • $\begingroup$ @Winther thanks! But you said You don't really need very precise estimates , how to show diverges using $S_n$ otherwise? $\endgroup$ – MAN-MADE Jul 23 '17 at 14:33
  • 1
    $\begingroup$ The title seems to be a tad misleading. $\endgroup$ – Simply Beautiful Art Jul 23 '17 at 18:01
  • $\begingroup$ @SimplyBeautifulArt can you please suggest one, thanks. $\endgroup$ – MAN-MADE Jul 23 '17 at 18:12
  • 1
    $\begingroup$ @MANMAID How did you arrive at the calculation relating $T(n) = S(n)$? You didn't define the notation $k_1\ne k_2 \ne\cdots \ne k_n$, but it would be natural to think it means "$k_1,k_2,\ldots,k_n$ are distinct". However, the calculation suggests that you really mean "$k_1,k_2,\ldots,k_n$ are not all identical", which is a very different criterion. Of the index sets $(k_1,k_2,k_3) \in \{ (1,1,2), (1,2,1), (3,2,1) ,(1,2,3) \}$, which of these four is included in your summation? $\endgroup$ – Erick Wong Jul 23 '17 at 21:34
2
$\begingroup$

A hint: since $$S_n=\frac{p_1}{p_1-1}\frac{p_2}{p_2-1}...\frac{p_n}{p_n-1}$$

From Mertens' 3rd theorem $$0<(1-\varepsilon)e^{-\gamma}<\ln{p_n} \cdot \prod_{k=1}^{n}\frac{p_k-1}{p_k} < (1+\varepsilon)e^{-\gamma}$$ or $$\frac{\ln{p_n}}{(1-\varepsilon)e^{-\gamma}}>\prod_{k=1}^{n}\frac{p_k}{p_k-1} > \frac{\ln{p_n}}{(1+\varepsilon)e^{-\gamma}}$$ Altogether $$C_1\cdot \ln{p_n} > S_n>C_2\cdot \ln{p_n}$$

$\endgroup$
  • $\begingroup$ Thanks a lot! I did not know Mertens' 3rd theorem. $\endgroup$ – MAN-MADE Jul 23 '17 at 16:05
  • $\begingroup$ Again can you say if $T_n$ diverges or not... yes or no at least! $\endgroup$ – MAN-MADE Jul 23 '17 at 16:05
  • $\begingroup$ I need to think about it, I only recalled Mertens' theorem because I used it a few times in past, like math.stackexchange.com/questions/1726488/… $\endgroup$ – rtybase Jul 23 '17 at 16:17
  • $\begingroup$ @MANMAID Otherwise try bounding $T_n$ below by a function of $S_n=\prod_{p \le n} \frac{1}{1-1/p}$, that $\lim_{n \to \infty} S_n$ diverges implies your series diverges $\endgroup$ – reuns Jul 24 '17 at 2:17

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.