Daniel R. Johnston
Abstract:We demonstrate the impact of a generic zero-free region and zero-density estimate on the error term in the prime number theorem. Consequently, we are able to improve upon previous work of Pintz and provide an essentially optimal error term for some choices of the zero-free region. As an example, we show that if there are no zeros $\rho=\beta+it$ of $\zeta(s)$ for $$\beta>1-\frac{1}{c(\log t)^{2/3}(\log\log t)^{1/3}},$$ then $$\frac{|\psi(x)-x|}{x}\ll\exp(-\omega(x))\frac{(\log x)^9}{(\log\log x)^3},$$ where $\psi(x)$ is the Chebyshev prime-counting function, and $$\omega(x)=\left(\frac{5^6}{2^2\cdot 3^4\cdot c^3}\right)^{1/5}\frac{(\log x)^{3/5}}{(\log\log x)^{1/5}}.$$ This improves upon the best known error term for the prime number theorem, previously given by $$\frac{|\psi(x)-x|}{x}\ll_{\varepsilon}\exp(-(1-\varepsilon)\omega(x))$$ for any $\varepsilon>0$.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to optimize the error term in the prime number theorem. Specifically, the author studied the influence of zero - point density estimates and zero - free regions on the error term in the prime number theorem and tried to provide the optimal error term under certain selected zero - free regions.
### Problem Background
The prime number theorem describes the distribution of prime numbers, which involves several important functions:
- \(\pi(x)\): the number of prime numbers less than or equal to \(x\).
- \(\theta(x)\): \(\sum_{p \leq x} \log p\), where \(p\) is a prime number.
- \(\psi(x)\): \(\sum_{p^k \leq x} \log p\), where \(p\) is a prime number and \(k \geq 1\).
The prime number theorem shows that:
\[
\pi(x) \sim \text{li}(x) := \int_2^x \frac{dt}{\log t}, \quad \theta(x) \sim x, \quad \psi(x) \sim x
\]
Therefore, it is usually necessary to make a strong boundary estimate of the (scaled) error term:
\[
\Delta_1(x) = \left| \frac{\pi(x) - \text{li}(x)}{x / \log x} \right|, \quad \Delta_2(x) = \left| \frac{\theta(x) - x}{x} \right|, \quad \Delta_3(x) = \left| \frac{\psi(x) - x}{x} \right|
\]
### Main Problem
Assuming the Riemann hypothesis holds, Schoenfeld [21] gave a relatively good error bound:
\[
\Delta_i(x) < \frac{\log^2 x}{8\pi \sqrt{x}}, \quad x \geq 2657
\]
However, without assuming the Riemann hypothesis, the error bound estimate will be much weaker. In order to optimize the estimation of these error bounds, it is necessary to know as much as possible about the zero - point distribution of the Riemann zeta function, which is usually achieved through zero - free regions and zero - point density estimates.
### Main Contributions of the Paper
This paper explored how to derive almost optimal error terms from zero - free regions by improving Pintz's method and using more general zero - point density estimates. Specifically, the author proved the following result:
#### Theorem 2.1
Let \(\eta(t)\) be a decreasing function with a continuous derivative \(\eta'(t)\) for \(t \geq 0\), such that \(0 < \eta(t) \leq \frac{1}{2}\), and \(\zeta(s)\) has no zero at \(\beta > 1 - \eta(|t|)\). Assume that \(N(\sigma, t)\) satisfies the boundary:
\[
N(\sigma, t) \ll t^A (1 - \sigma)^B (\log t)^C
\]
for some constants \(A>0, B \geq 1, C \geq 0\), uniformly in the range \(\sigma \in [\sigma_0, 1]\), where \(\sigma_0 \in (1/2, 1)\). In addition, assume that for a sufficiently large \(x\), the function \(f_x(t)=\eta(t) \log x+\log t\) is minimized at the unique value \(t = t_0\). If \(\omega(x)=f_x(t_0)\) satisfies:
\[
\omega(x) \geq 2 \log \log x \quad \text{and} \quad \omega(x)=o(\log x)
\]
then:
\[
\Delta_i(x) \ll \exp(-\omega(x)) \exp\left( \frac{2A \omega}{}\right)