In my textbook, there is the following theorem:

For all $x>0$, we have $$\psi(x)=\sum_{\alpha=1}^\infty\theta(x^{1/\alpha})$$ and hence $$\psi(x)=\theta(x)+O(\sqrt x\log x).$$

Here $\theta(x)=\sum_{p\le x}\log p$ and $\psi(x)=\sum_{p^\alpha\le x}\log p$ are the Chebyshev functions. The first formula is proven with a little sum manipulations, but the second formula is not explained.

My guess is the author is writing $\sum_{\alpha=1}^\infty\theta(x^{1/\alpha})=\theta(x) +\sum_{\alpha=2}^\infty\theta(x^{1/\alpha})$, and the leading term of the remainder is $\theta(\sqrt x)$. Since this is introductory material we don't have the PNT, so we don't know $\theta(x)\sim x$, but a simple estimate gives $\theta(x)\le x\log x$ so the leading term is indeed $\sqrt x\log(\sqrt x)=\frac12\sqrt x\log x$. But it is not obvious to me that the other terms, of order $x^{1/3}\log x$, $x^{1/4}\log x$ etc. don't necessarily overwhelm the first term, if there are enough of them. Again, a naive bound on $\alpha$ gives $\alpha\le \log_2 x$, but this is much too weak if you just take $\log x$ copies of $x^{1/2}\log x$, and I don't think $\sum_\alpha x^{1/\alpha}$ converges.


We can do the following elementary estimate: \begin{align*} \psi(x)-\theta(x) & =\sum_{p^m\le x}\log(p)-\sum_{p\le x}\log(p)\\ & =\sum_{p\le \sqrt{x}}\log(p)\sum_{2\le m\le\frac{\log(x)}{\log(p)}}1 \\ & \le \sum_{p\le \sqrt{x}}\log(p)\Bigl[\frac{\log(x)}{\log(p)} \Bigr] \\ & \le \sqrt{x}\log(x). \end{align*}

  • $\begingroup$ Hm, I suppose this works, although since it goes from a completely different starting point I wonder what the author meant by "and hence". $\endgroup$ – Mario Carneiro Apr 8 '16 at 14:32

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.