27
$\begingroup$

A weaker version of the Riemann hypothesis is the claim that if $\zeta(s) = 0$ then $Re(s) \leq 1 - h$ for some constant $h> 0$. What would the consequences be of a result of this type?

$\endgroup$
1
  • $\begingroup$ Comments are not for extended discussion; this conversation has been moved to chat. $\endgroup$ – Ben Webster Jun 13 at 0:29
16
$\begingroup$

If this were proven, it would be a huge breakthrough in number theory. The most direct improvement would of course be a power savings in the error term of $\lvert \pi (x) - \mathrm{Li} (x) \rvert$, but there are many more applications for such things. For example, this would imply that $\zeta \left( \sigma + i t \right) = \mathcal{O} \left( \lvert t \rvert^{\varepsilon} \right)$ for $\sigma \geq 1 - 2 h$, which is a significant improvement on Heath-Brown's bound $\zeta \left( \sigma + i t \right) = \mathcal{O} \left( \lvert t \rvert^{\frac{1}{2} (1 - \sigma)^{\frac{3}{2}} + \varepsilon} \right)$, which (if I recall correctly) is the currently best known bound for values close to $\sigma = 1$. Bounds like this are very useful in all types of applications, where integrals containing the zeta function appear.

Of course, this is just one of many and varied consequences.

$\endgroup$
1
  • $\begingroup$ Comments are not for extended discussion; this conversation has been moved to chat. $\endgroup$ – Ben Webster Jun 13 at 0:29
14
$\begingroup$

Many number-theoretic functions have error bounds contingent on the supremum of the real parts of zeroes of the Riemann zeta function. Call this $\Theta$. If we write $f(x)=\Omega_{\pm}(g(x))$ for the notion that \begin{align*} \limsup_{x\to\infty}\frac{f(x)}{g(x)}&>0,\text{ and}\\ \liminf_{x\to\infty}\frac{f(x)}{g(x)}&<0, \end{align*} then (e.g. in Montgomery and Vaughan, Section 15.1), for any $\epsilon>0$, \begin{align*} \psi(x)-x&=\Omega_{\pm}(x^{\Theta-\epsilon}),\\ \pi(x)-\operatorname{li}(x)&=\Omega_{\pm}(x^{\Theta-\epsilon}),\\ M(x)&=\Omega_{\pm}(x^{\Theta-\epsilon}). \end{align*} Here, $\psi(x)=\log\operatorname{lcm}(1,2,\dots,n)$ is the Chebyshev Psi function, $\pi(x)$ is the prime counting function, $\operatorname{li}(x)$ is the logarithmic integral, and $M(x)$ is the Mertens function. In other words, the maximal error of these summatory functions from the standard "simple" approximations is (somewhat) precisely determined by how far the zeroes of $\zeta$ may stray from the line $\Re s=1/2$. (In fact, sharper results involving $\Omega_{\pm}(x^\Theta)$ directly can be shown if there is a zero $\rho$ with $\Re\rho=\Theta$). So, knowing that $\Theta<1$ gives quite strong information about the distribution of prime numbers, and as $\Theta\to1/2$ such information approximates the best possible bounds, at least as far as functions of the form $x^\alpha$ are concerned.

$\endgroup$
1
  • 6
    $\begingroup$ The $\Omega_\pm$ bounds give you lower bounds on the oscillation of the error terms, not upper bounds. Your answer would make more sense to me if you were citing the $O(x^{\Theta+\epsilon})$ bounds. $\endgroup$ – Wojowu Jun 13 at 1:04

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.