Let $$A = \lim_{x\to \infty} \sum_{k=2}^\infty \sum_{p^k\le x} \frac{\ln(p)}{p^k}$$

I want to show that $\displaystyle A - \sum_{k=2}^\infty \sum_{p^k\le x} \frac{\ln(p)}{p^k} \in \mathcal{O}(\frac{1}{\sqrt{x}})$

Generally, I want to know how fast $\displaystyle \sum_{k=2}^\infty \sum_{p^k\le x} \frac{\ln(p)}{p^k}$ converges.

Some results:

It is possible to express $A$ as $\displaystyle \sum_{p \in \mathbb{P}} \frac{\ln(p)}{p(p-1)}$ and I was able to show that $\displaystyle A - \sum_{p \le x} \frac{\ln(p)}{p(p-1)} \in \mathcal{O}(\frac{1}{x})$ using Abel summation. I wasn't able to use the same technique on the double sum above. I noticed, that there is a connection to Chebyshev's functions by $\displaystyle \psi(x) - \vartheta(x) = \sum_{k=2}^\infty \sum_{p^k \le x} \ln(p)$

Also $\displaystyle A = -\sum_{k=2}^\infty P'(k)$ where $P(s)$ denotes the Prime-Zeta function, but I don't think that this is useful for the stated problem.

I tried to approximate $A$ with the sum over all natural numbers instead of primes, and approximate that with an Integral.

I tried some other minor things but nothing seems to lead anywhere. If you got some idea, any help is appreciated.


I'm using the Prime Number Theorem below. Let $\kappa(p,x):=\max\{1,\lfloor\ln x/\ln p\rfloor\}$; then $$R(x):=\sum_{k=2}^\infty\sum_{p^k>x}\frac{\ln p}{p^k}=\sum_{p\in\mathbb{P}}\sum_{k>\kappa(p,x)}\frac{\ln p}{p^k}=\sum_{p\in\mathbb{P}}\frac{\ln p}{p^{\kappa(p,x)}(p-1)}.$$ Note that $\kappa(p,x)=1$ if and only if $p^2>x$; otherwise $\kappa(p,x)>\frac{\ln x}{\ln p}-1$ and $p^{\kappa(p,x)}>\frac{x}{p}$.

Hence $R(x)=S(x)+T(x)$, where $$S(x)=\sum_{p^2\leqslant x}\frac{\ln p}{p^{\kappa(p,x)}(p-1)}\leqslant\frac1x\sum_{p^2\leqslant x}\frac{p\ln p}{p-1}\in\mathcal{O}(x^{-1/2})$$ because $\frac{p\ln p}{p-1}\leqslant 2\ln p\leqslant \ln x$ and $\#\{p : p^2\leqslant x\}\in\mathcal{O}\left(\frac{x^{1/2}}{\ln x}\right)$ by PNT, and $$T(x)=\sum_{p^2>x}\frac{\ln p}{p^{\kappa(p,x)}(p-1)}=\sum_{p^2>x}\frac{\ln p}{p(p-1)}\in\mathcal{O}(x^{-1/2})$$ is what you've shown already (moreover, this can't be enhanced).

  • 1
    $\begingroup$ +1 and a big thank you for writing $\in \mathcal{O}(\cdot)$ instead of $= \mathcal{O}(\cdot)$ $\endgroup$ – user3733558 Mar 24 at 9:38
  • 1
    $\begingroup$ @user3733558 people write $= O(\cdot)$ to mean $\in O(\cdot)$, cause it's much easier that way (e.g., you can write $f = g+O(h)$ to mean $f-g \in O(h)$). $\endgroup$ – mathworker21 Mar 24 at 20:24

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.