cp's OEIS Frontend

|zeta'(alpha)| > 1 means that s = alpha is a repelling fixed point of zeta(s). As a result, for any initial value s_0 in (1, +oo), s_0 != alpha, the iterated sequence s_0, zeta(s_0), zeta(zeta(s_0)), ... diverges.

Moreover, let s_0 be any real number > alpha, s_n = zeta(s_{n-1}) for n >= 1, then it seems that ... > s_{2n} > s_{2n-2} > ... > s_2 > s_0 > alpha > s_1 > s_3 > ... > s_{2n+1} > ..., and {s_{2n}} diverges to +oo, {s_{2n+1}} converges to 1. Moreover, the divergence of {s_{2n}} and convergence of {s_{2n+1}} should be really fast, see my conjecture in A344428.

For real values of the parameter "a" between 0 and 1, a real fixed point "s" of the iterated Hurwitz zeta function [s = zetahurwitz(s, a)] lies on a curve that passes through A069857 (-0.295905...) and has a maximum tending toward 1. This curve has inflection points for a = 0.1990753... or 0.91964... . The fixed point "s" on this curve for the iteration "s = zetahurwitz(s, A324859)" is A324860 (0.5250984...).

For real values of the parameter "a" between 0 and 1, a real fixed point "s" of the iterated Hurwitz zeta function [s = zetahurwitz(s, a)] lies on a curve that passes through A069857 (-0.295905...) and has a maximum tending toward 1. This curve has inflection points for a = 0.1990753... (A324859) or 0.91964... . The fixed point "s" on this curve for the iteration "s = zetahurwitz(s, A324859)" is 0.5250984... .

Let f(s) = zeta(zeta(s+1)) - 1, where zeta(s) is the Riemann zeta function. Then f(s) is a strictly increasing function from (0, +oo) to (0, +oo), lim_{s->0+} f(s) = 0, lim_{s->+oo} f(s) = +oo.

Conjecture:

(i) f(s) has a unique fixed point s = A069995 - 1 in (0, +oo);

(ii) Lim_{s->+oo} f(s)/2^s = 1, lim_{s->0+} f(s)/2^(-1/s) = exp(-2/5) = A344428.

If these are true, let s_0 be any real number > alpha, s_n = zeta(s_{n-1}) for n >= 1, where alpha = A069995 is the fixed point of zeta(s) in (1, +oo), then {s_{2n}} diverges quickly to +oo, {s_{2n+1}} converges quickly to 1.

This is because the derivative of zeta(zeta(s)) - s at s = alpha is (zeta'(alpha))^2 - 1 = A344427^2 - 1 > 0, so (i) implies that zeta(zeta(s)) > s for s > alpha and zeta(zeta(s)) < s for 1 < s < alpha, hence ... > s_{2n} > s_{2n-2} > ... > s_2 > s_0 > alpha > s_1 > s_3 > ... > s_{2n+1} > ..., and it follows from (i) that lim_{n->+oo} s_{2n} = +oo, lim_{n->+oo} s_{2n+1} = 1. By definition s_n - 1 = f(s_{n-2} - 1), n >= 2. For large n, s_{2n} - 1 is approximately equal to 2^(s_{2(n-1)} - 1), and 1/(s_{2n+1} - 1) is approximately equal to exp(2/5) * 2^(1/(s_{2(n-1)+1} - 1)).

Ossicini's function Э(s) is constructed to remove the poles of gamma(s) and zeta(s) along with the trivial zeros of zeta(s) and (Dirichlet) beta(s). Its zeros include the nontrivial zeros of zeta(s) and beta(s), and complex zeros contributed by (1 - 2^s) and (1 - 2^(1 - s)) at regular intervals of 2*Pi/log(2) on the lines Re(s) = {0, 1}.