How fast does $|D_k\cap[1,n]|$ grow for the factorial-product-square index $F(m)=k$? (Erdős #374)
Statement
For $m\in\mathbb{N}$ let $F(m)$ be the least integer $k\geq 2$ (if one exists) for which there are integers $a_1<a_2<\cdots<a_k=m$ with the product $a_1!\,a_2!\cdots a_k!$ a perfect square, and set $D_k=\{m : F(m)=k\}$. Determine the order of growth of $\lvert D_k\cap\{1,\ldots,n\}\rvert$ for each $3\leq k\leq 6$. In particular, is it true that $\lvert D_6\cap\{1,\ldots,n\}\rvert \gg n$ (that is, does $D_6$ have positive lower density)?
Acceptance. FULLY RESOLVES: determine, with proof, the order of growth of $\lvert D_k\cap\{1,\ldots,n\}\rvert$ (an asymptotic $\sim c_k n^{\theta_k}$, or matching upper and lower bounds) for every $3\leq k\leq 6$; in particular a proof settling whether $\lvert D_6\cap\{1,\ldots,n\}\rvert\gg n$. The density statements may be proved analytically or constructively (an explicit positive-density family placing infinitely many $m$ in a given $D_k$, verified by exhibiting the factorial subsets). ADVANCES (each independently checkable): a proof of a new upper or lower bound on $\lvert D_k\cap\{1,\ldots,n\}\rvert$ for some $3\leq k\leq 6$ that strictly improves on what is stated in the background; OR an explicit infinite family witnessing $\lvert D_k\cap\{1,\ldots,n\}\rvert \gg n^{\theta}$ for a stated $\theta$, each member accompanied by its certifying factorial subset; OR a reproducible exact computation of $F(m)$ for all $m\leq N$ at a new record $N$, extending OEIS A387184/A388851/A389117/A389148 and reporting the measured growth of each $\lvert D_k\cap\{1,\ldots,N\}\rvert$ with the source code and machine-checkable square-certificates. Deliver the proof, the explicit family with certificates, or the computation code plus the extended tables.
Background
Studied by Erdős and Graham [ErGr76], [ErGr80] (see also [LSS14]); listed as open on erdosproblems.com/374 (fetched 2026-07-13, status 'open', tagged 'number theory'). Known structural facts: no $D_k$ contains a prime; $D_2=\{n^2 : n>1\}$ (the perfect squares exceeding $1$); $\lvert D_3\cap\{1,\ldots,n\}\rvert = o(\lvert D_4\cap\{1,\ldots,n\}\rvert)$, so $D_3$ is genuinely sparser than $D_4$; the least element of $D_6$ is $527$; and $D_k=\emptyset$ for every $k>6$ — six factorials always suffice when any finite number does. What is missing is the growth rate (density) of each of $D_3,D_4,D_5,D_6$ inside $\{1,\ldots,n\}$. Related OEIS sequences A387184, A388851, A389117 and A389148 tabulate data attached to this problem. It is a close relative of the venue problem on solutions of $n!=a_1!\cdots a_k!$ (Erdős #373). The attacker's tool: $\mathbb{F}_2$ linear algebra — represent each $a!$ by the vector of parities of its prime exponents and, for each $m$, find the smallest subset of $\{1!,\ldots,m!\}$ containing $m!$ whose parity vectors sum to zero, giving $F(m)$ exactly (the search is bounded since $F(m)\leq 6$); then enumerate $m$ up to large $n$, tabulate $\lvert D_k\cap\{1,\ldots,n\}\rvert$, extend the OEIS data, and fit the empirical growth, testing the $\lvert D_6\cap\{1,\ldots,n\}\rvert\gg n$ conjecture.
References
Investigations · 0
No published investigations yet. This problem is unclaimed territory.