THE ONE-CLASS EXPONENT over the exact 64-term ladder A048670 gamma(n) = log g(p_n#) / log p_n (the local exponent in x) shape predictions: x^2 (Iwaniec) ; x log^2 x (conjecture) ; x log x log_3 x / log_2 x (FGKMT lower bound) n p_n g run gamma g/(p logp) g/(p log2p) gamma_shape 1 2 2 1 1.0000 1.443 2.081 1.0000 2 3 4 3 1.2619 1.214 1.105 1.1712 3 5 6 5 1.1133 0.746 0.463 1.5914 4 7 10 9 1.1833 0.734 0.377 1.6842 5 11 14 13 1.1006 0.531 0.221 1.7295 6 13 22 21 1.2051 0.660 0.257 1.7345 7 17 26 25 1.1500 0.540 0.191 1.7351 8 19 34 33 1.1976 0.608 0.206 1.7335 9 23 40 39 1.1765 0.555 0.177 1.7289 10 29 46 45 1.1370 0.471 0.140 1.7211 15 47 100 99 1.1961 0.553 0.144 1.7003 20 71 174 173 1.2103 0.575 0.135 1.6803 25 97 258 257 1.2138 0.581 0.127 1.6648 30 113 330 329 1.2267 0.618 0.131 1.6572 35 149 432 431 1.2127 0.579 0.116 1.6436 40 173 538 537 1.2202 0.603 0.117 1.6363 45 197 642 641 1.2236 0.617 0.117 1.6301 50 229 762 761 1.2213 0.612 0.113 1.6230 55 257 876 875 1.2210 0.614 0.111 1.6176 60 281 1002 1001 1.2255 0.632 0.112 1.6135 64 311 1110 1109 1.2217 0.622 0.108 1.6089 gamma at n=13 (p=41) : 1.1590 gamma at n=64 (p=311) : 1.2217 Iwaniec's theorem allows exponent 2 exactly (g << (omega log omega)^2), and for n = x# that is x^2 (1+o(1)): at p=311, x^2 = 96721 vs g = 1110, a factor 87.1. FGKMT's lower bound gives about 570 at p=311: g exceeds it by a factor 1.95. So at the largest computed level the theorem is ~87x the truth and the construction is within a factor ~1.9: THAT interval is the open problem. Required saving for the target: g = o(x^2) needs a factor psi(x) -> infinity. Iwaniec : (omega log omega)^2 -> x^2 (1+o(1)) NOT o(x^2) single-log saving : omega^2 log omega -> x^2 / log x = o(x^2) Jacobsthal conj. : omega^2 -> x^2 / log^2 x = o(x^2)