The Chebyshev functions are defined as $\psi(x) = \sum_{p^m \leq x} \log n$ and $\theta(x) = \sum_{p\leq x} \log p$, where $p$ is a prime, $m\geq 1$ is an integer and $n=p^m$ in $\psi(x)$. It is known that there exist positive constants $c_{1}$ and $c_2$ such that $c_{1}x < \psi(x) < c_{2} x$ and $ \frac{1}{2}c_{1}x < \theta(x) < c_{2} x$. By these bounds we find that $ \dfrac{ \psi(x) - x}{\psi(x) - \theta(x)} < \dfrac{(c_{2} -1)x}{(c_2 - c_1)x} = \dfrac{c_{2} -1 }{c_2 - c_1} <\infty $.

Hence invoking the well known result $\mid \psi(x) - \theta(x) \mid = O(x^{\frac{1}{2}}\log x)$, it then follows that $\mid \psi(x) - x \mid = O(x^{\frac{1}{2}}\log x)$ ?

EDIT: It is also known that $\theta (x)$ tends to $x$ as $x$ tends to $\infty$. Surely, an impication of this is $\dfrac{ \psi(x) - x}{\psi(x) - \theta(x)} < \infty$ since $\psi(x) \neq \theta(x)$.

  • $\begingroup$ Note that $c_1 - c_2 < 0$. You don't have a lower bound $\psi(x) - \vartheta(x) > \gamma\cdot x$. $\endgroup$ – Daniel Fischer Feb 6 '16 at 20:33
  • $\begingroup$ I think you're referring to the typo i just edited. $\endgroup$ – User1 Feb 6 '16 at 20:37
  • $\begingroup$ In that case, what's your argument for $$\frac{\psi(x) - x}{\psi(x) - \vartheta(x)} < \frac{(c_2-1)x}{(c_2-c_1)x}\,?$$ $\endgroup$ – Daniel Fischer Feb 6 '16 at 20:39
  • $\begingroup$ The $x's$ on the RHS cancel and we remain with a constant. $\endgroup$ – User1 Feb 6 '16 at 20:45
  • $\begingroup$ You do that after the step I'm asking about. One thing after the other. $\endgroup$ – Daniel Fischer Feb 6 '16 at 20:46

The inequality \[\frac{\psi(x) - x}{\psi(x) - \vartheta(x)} < \frac{(c_2 - 1)x}{(c_2 - c_1)x}\] is false, as Daniel Fischer mentions. The mistake is in the denominator; you know that $c_1 x < \psi(x) < c_2 x$ and $\frac{1}{2} c_1 x < \vartheta(x) < c_2 x$, but this does not imply that \[\frac{1}{\psi(x) - \vartheta(x)} < \frac{1}{(c_2 - c_1)x},\] or equivalently that \[\psi(x) - \vartheta(x) > (c_2 - c_1)x.\] In fact, \[\psi(x) - \vartheta(x) = \sum_{\substack{p^m \leq x \\ m \geq 2}} \log p,\] and from this it is easy to show that \[\frac{1}{2} c_1 \sqrt{x} < \psi(x) - \vartheta(x) < c_3 \sqrt{x} \log x,\] so the inequality is far from true.


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.