Polygamma theory, the Li/Keiper constants and the Li criterion for the Riemann hypothesis (Q618664)

The Riemann \(\xi\) function is defined by \(\xi(s)=s(s-1)\pi^{-s/2}\Gamma(s/2)\zeta(s)\). Regarding to this functions we consider the sequence \(\{\lambda_n\}_{n\geq 1}\) with general term \[ \lambda_n=\frac{1}{(n-1)!}\frac{d^n}{ds^n}\left(s^{n-1}\ln\xi(s)\right)_{s=1}. \] It is known that the Riemann hypothesis is equivalent to the Li criterion asserting that all terms of the above sequence are nonnegative. The paper under review studies various properties of associated sets of certain constants \(c_n\) and \(d_n\) in eleven propositions, such that in some of them, the precise relation of these numbers to the Li/Keiper constants is obtained. Indeed, the sequences \(c_n\) and \(d_n\) are the coefficient of \(z^n\) in the Maclaurin expansions of the functions \[ \ln\left(\frac{z}{1-z}\zeta\left(\frac{1}{1-z}\right)\right) \qquad\text{ and }\qquad \ln\Gamma\left(\frac{1}{2(1-z)}\right), \] respectively. The author shows that the leading behavior \((\ln n)/2\) of \(\lambda_n/n\) is absent in \(c_n\). This result suggests the validity of the Riemann hypothesis. Moreover, the author presents the conjecture \(c_n\ll 1/n^{1/2-\varepsilon}\) for arbitrary \(\varepsilon>0\), and gives some numerical evidences suggesting that optimal order of \(|c_n|\) is very close to \(O(\ln n/\sqrt{n})\).