4
$\begingroup$

The book "The Riemann Hypothesis, A Resource for the Afficionado and Virtuoso Alike", Borwein, Choi, Rooney, Weirathmueller, Eds., states on its page six the following theorem (Theorem 1.2):

The Riemann hypothesis is equivalent to the statement that for every fixed $\epsilon > 0$,

$$\lim_{n \to \infty} \frac{\lambda(1) + \lambda(2)+ ...+\lambda(n)}{n^{\frac 12 + \epsilon}} = 0.$$

(Here, $\lambda$ is the Liouville function $\lambda: n \mapsto (-1)^{\omega(n)}$ where $\omega(n)$ is the number of, not necessarily distinct, prime factors of $n$.) The editors remark that this statement among others were considered by Landau in his doctoral thesis of 1899. There is translation by Coons of a dissertation of Landau, which is described as his doctoral dissertation, here:

https://arxiv.org/abs/0803.3787

But I haven't been able to discern a direct connection between this article and the theorem above. Am I missing something? Is there perhaps a proof of this theorem available somewhere else? I've looked a bit, and I haven't found one.

Reason for asking: I wonder, under the Riemann hypothesis, what happens to the exponent in the denominator if powers of $-1$ in the definition of the Liouville function are replaced by powers of some other root of unity $\exp(2 \pi i)/k, k >2$.

$\endgroup$
  • 1
    $\begingroup$ Work out why the Dirichlet series for the multiplicative function $\lambda(n)$ is $\zeta(2s)/\zeta(s)$ and see how those calculations change if you use the multiplicative function $\lambda_z(n)$ instead where on prime powers $\lambda_z(p^r) = z^r$ for a root of unity $z$ with order $k$. How does the identity $1/(1+x) = (1-x)/(1-x^2)$ change when the left side is $1/(1-zx)$? $\endgroup$ – KConrad May 11 '17 at 2:34
2
$\begingroup$

About your reason for asking. See https://projecteuclid.org/DPubS?service=UI&version=1.0&verb=Display&handle=euclid.nmj/1306851587 or the arxiv version https://arxiv.org/pdf/0906.1029.pdf where such things are discussed. In particular, the partial sums $\sum_{n\leq x}\exp(2\pi i\Omega(n)/k)$ are not $o(x^\alpha)$ for any $\alpha<1$.

$\endgroup$

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.