Sorry, I've asked a similar question before, and the current question seems to have been answered there by Cam McLeman. The answer he gives is $PF(n)\approx ln(ln(n)).$ To me, this means that $E(N)$ should also be $ln(ln(N))$.