Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Claim. Put T(n)=ϕ(n)+1T(n)=\phi(n)+1, and for a prime pp let BpB_p be the set of positive integers nn that reach pp after zero or more iterations of TT. For every odd prime pp, Bp={p,2p}B_p=\{p,2p\} if and only if the only positive integer solutions of ϕ(n)=p−1\phi(n)=p-1 are n=pn=p and n=2pn=2p. For every δ>0\delta>0 there is CδC_\delta such that, for every x≥3x\ge3, at most Cδx/(log⁡x)2−δC_\delta x/(\log x)^{2-\delta} primes p≤xp\le x have Bp≠{p,2p}B_p\ne\{p,2p\}. Thus, for almost every prime pp, exactly two positive integers reach pp, and the set of nn that reach pp has density zero.

Covers. The second and third questions of Problem 409 at almost every prime, that is, at every prime outside an exceptional set of at most Cδx/(log⁡x)2−δC_\delta x/(\log x)^{2-\delta} primes up to xx: only finitely many nn reach such a prime, and the density of the nn that reach it is zero. The union of these two-element basins has natural density zero among the positive integers, so almost every integer reaches an exceptional prime; the claim gives no information about the basin of any exceptional prime. Whether some prime is reached by infinitely many nn, the density of each exceptional basin, and the first question, how many iterations are needed, remain open.

Credit. The counting input is Theorem 4.1 of Banks, Ford, Luca, Pappalardi and Shparlinski, Values of the Euler Function in Various Sequences, Monatsh. Math. 146 (2005), 1–19, which bounds the primes p≤xp\le x for which ϕ(n)=p−1\phi(n)=p-1 has a solution other than pp and 2p2p; their introduction records the almost-all-primes consequence for these preimages. Plasma AI supplies the identification of the whole basin, using that 2p2p is not a value of TT. An argument posted in the problem's forum thread on 22 July 2026 by the account ankit, recorded on its own claim page, gives the basin {p,2p}\{p,2p\} for all but O(x)O(\sqrt x) primes p≤xp\le x with p≡3(mod4)p\equiv3\pmod4, ending with the same step.

Standing. Tier 0: the deduction is author-recorded by Plasma AI, with no independent review or Lean proof. The 2005 theorem is a published input: its definition, statement and proof were read in the authors' manuscript, and the totient-counting and sieve results on which it rests were not independently audited. Catalog acceptance is pending.

Depends on. L104 (tier 0), with its external input Banks, Ford, Luca, Pappalardi and Shparlinski's Theorem 4.1.