Cyclotomic primes (Q6979227)

For a positive integer \(m\), let \(\phi_m=\Phi_m(2)\), where \(\Phi_m(X)\in {\mathbb Z}[X]\) is the \(m\)th cyclotomic polynomial. The author asks himself when is \(\phi_m\) or a large algebraic factor of it composite. The integer \(\phi_m\) has an algebraic factor for certain positive integers \(m\). Namely, let \(p\) be an odd prime and put \(\ell(p)\) for the order of \(2\) modulo \(p\). It turns out that \(p\| \Phi_m(2)\) exactly when \(m=p^j\ell(p)\) for some positive integer \(j\). In this case \(p\) is uniquely determined by \(m\) as its largest prime factor. When this happens one calls \(p\) the intrinsic prime factor of \(\phi_m\). Put \N\[\N\psi_m:=\phi_m/\delta_m, \N\]\Nwhere \(\delta_m=1\) unless \(p\) is the intrinsic prime factor of \(\phi_m\) in which case one puts \(\delta_m=p\). There is an additional algebraic reason for \(\phi_m\) not to be prime when \(m\equiv 4\pmod 8\) which comes from the factorization \N\[\N\phi_{8k+4}=\gcd(\phi_{8k+4}, 2^{2k+1}+2^{k+1}+1)\gcd(\phi_{8k+4},2^{2k+1}-2^{k+1}+1):=\phi_{8k+4}^+\phi_{8k+4}^{-}. \N\]\NThis was observed by Schinzel who proved that the above factorization is nontrivial for \(k\ge 3\). This is known under the name of Aurifeuillan factorization. Putting \(\psi_{8k+4}=\psi^{+}_{8k+4}\psi^{-}_{8k+4}\) to account for the intrinsic prime in case it exists, the author considers the sets \N\[\NC_1=\{\psi_m: m\not\equiv 4\pmod 8\},\qquad C_2=\{\psi_m: m\equiv 4\pmod 8\}, \N\]\Nand studies how many elements of \(C_1\) are composite and how many elements of \(C_2\) are not a product of \(2\) primes as \(m\) varies. He shows that \(C_1\) contains more that \(x^{3/5}\) numbers which are not prime for \(m\le x\) and \(x\) large and \(C_2\) contains more than \(x^{3/5}\) numbers which are not product of two primes when \(m\le x\). Under the \(abc\)-conjecture the author shows a bit more namely that \(C_1\) contains at least \(x^{3/5}\) numbers which are divisible by at least \(2\) \textit{distinct} primes for \(m\le x\) and \(C_2\) contains at least \(x^{3/5}\) numbers which are divisible by at least \(3\) \textit{distinct} primes once \(x\) is sufficiently large. The proof uses sieve methods (the Bombieri-Vinogradov theorem) and results about primes in arithmetic progressions with large prime modulis due to \textit{R. C. Baker} and \textit{G. Harman} [Acta Arith. 83, No. 4, 331--361 (1998; Zbl 0994.11033)]. The author presents a quick elementary proof of the weaker version of the above statements in which the exponent \(3/5\) is replaced by the smaller exponent \(1/2\). The last section contains some statistics and surmises which are likely to inspire future research.