2
$\begingroup$

I am reading the book of the prime number theorem by G.J.O.Jameson on p.192

If the Riemann Hypothesis were true, the contour of integration in the proof of Theorem 5.1.1 could be shifted almost to the line $ \operatorname{Re}(s) = \frac{1}{2} $. A modified version of the argument would then suggest that
$ \psi(x) - x = O(x^c) \quad \text{for every } c > \frac{1}{2}. $
where $ \psi(x) = \sum_{n \leq x} \Lambda(n) $ is the Chebyshev function

theorem 5.1.1 Consider a sequence $ a(n) $ and an associated function $ f(s) $ satisfying the following conditions:

  • (FT1) The Dirichlet series

    $\sum_{n=1}^{\infty} \frac{a(n)}{n^s} $ converges absolutely to $ f(s) $ for $ \operatorname{Re}(s) > 1 $.

  • (FT2) The function $f(s) $ has the form
    $ f(s) = \frac{\alpha}{s-1} + \alpha_0 + (s-1) h(s), $ where $ h(s) $ is differentiable at $s = 1 $.

  • (FT3a) There exist constants $ c_0, M_0 $ such that $ f(s) $ is differentiable in the region $ |t| \geq 2, x \geq 1 - \frac{c_0}{\log |t|} $, and satisfies the bound
    $ |f(x + i t)| \leq M_0 (\log |t|)^3. $

  • (FT3b) The function $ f(s) $ is differentiable at all points $x + i t $ (except s = 1 ) for $ |t| \leq 2 $ and $ x \geq 1 - \frac{c_0}{\log 2} $.

    then the summatory function
    $ A(x) = \sum_{n \leq x} a(n) $ satisfies the integral bound
    $ \left| \int_x^\infty \frac{A(y) - \alpha y}{y^2} \, dy \right| \leq K \exp(-c (\log x)^{1/2}) $ for sufficiently large x , where $ K, c $ are constants.

My question is:
How $\psi(x)-x$ is a bound of the form $ O(x^{1/2 + \epsilon}) $ directly from Theorem 5.1.1 under RH?

$\endgroup$
1
  • $\begingroup$ in the proof of the theorem there's a vertical integral ("$J_1$") to the left of $\mathfrak R\mathfrak e(s)=1$. what's stopping you move it more to the left is that $1/\zeta (s)$ may not be holomorphic, because of zeros. RH says no zeros up to $1/2$, so you can move the integral to $1/2$. $\endgroup$ Commented Mar 18 at 23:02

2 Answers 2

9
$\begingroup$

Koch (1901) proved that the Riemann hypothesis implies $$\psi(x)=x+O(x^{1/2}\log^2 x).$$ This is stronger than the relation you ask about. See Theorem 13.1 and its proof in Montgomery-Vaughan: Multiplicative number theory I.

$\endgroup$
3
$\begingroup$

There is no way to obtain $\psi(x)-x = O(x^{1/2+\varepsilon})$ from Theorem 5.1.1. The conclusion is consistent with the hypothetical formula $A(y) = \alpha y + y^{1-\delta}$ for any $\delta>0$.

One would need to establish a different version of Theorem 5.1.1, with the Riemann hypothesis-like assumption built in to the theorem, to be able to get error terms this strong. (Or else use other known techniques, of course.)

$\endgroup$

You must log in to answer this question.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.