The number-theoretical spin chain and the Riemann zeroes (Q1290537)

In an interesting paper that appeared some years ago [\textit{A. Knauf}, Commun. Math. Phys. 153, 77-115 (1993; Zbl 0776.11047)], the author interpreted the quotient \(Z(s):= \zeta(s-1)/ \zeta(s)\) of the Riemann zeta-function at two different points, with \(\text{Re} (s)>2\), as the partition function of an infinite spin chain at inverse temperature \(s\). The function \(Z(s)\) is given, in that half plane, by the Dirichlet series \(Z(s)= \sum_{n\in \mathbb{N}} \varphi(n) n^{-s}\), \(\varphi(n)\) being the Euler totient function. The function \(Z(s)\) was also shown there to be the thermodynamic limit of a certain series of partition functions \(Z_k(s)\) of spin chains with \(k\) spins, with an energy function \(H_k\) that can be easily defined inductively. This so-called number theoretical spin chain was later shown to have the properties of typical systems considered in statistical mechanics, in particular, to possess exactly one phase transition, at \(s=2\) (precise references are given in the paper). From the point of view of number theory, its most important property seems to be its ferromagnetic behavior, e.g., the fact that its Fourier coefficients (the interaction coefficients in the statistical mechanics terminology) are all positive (the exceptional negative coefficient not affecting the Gibbs measure of the spin chain for inverse temperature \(s\geq 0\)). This ferromagnetic property is of interest in the context of the Riemann hypothesis. It is an empirical observation that the Riemann zeta-function can be well approximated in its critical strip by using the number theoretical spin chain, and a precise proof of this would imply the Riemann hypothesis. In the paper, this question is related to one concerning the spectral radii of Markov chains, which on its turn leads to the investigation of the Ramanujan character of certain graphs. The general idea is to explain the pseudorandom features of certain number-theoretical functions, by considering them as observables of a spin chain of statistical mechanics. Finally, in an appendix of the paper the free energy of the number theoretical spin chain is related to the solutions of the Lewis equation \(\psi(z)= \psi(z+1)+ z^{-2s} \psi(1+1/z)\). Also worth mentioning is a resemblance (emphasized by the author) of the number theoretical spin chain with an approach by A. Connes, in which he interpretes the Riemann zeta-function directly (instead of \(Z(s)\)) as the partition function of a statistical mechanical system at inverse temperature \(s\).