Source. Scott D. Hughes, Sums of distinct divisors of factorials,
arXiv:2609.10902v1, Remark 6, physical pp. 4–5 of the five-page PDF held by
Hughes (2026);
the library records it on
its result page.
Read in the canonical conversion beside the PDF and checked against the page
images.
Standing. Author-recorded reconstruction; not an independent review; it
changes no status and assigns no tier. The only imported input is
Chebyshev's bound π(x)≪x/logx.
Definitions
h(N) is defined on
the Theorem 1 page.
τ(N) is the number of positive divisors of N, vp the p-adic
valuation, and π(x) the number of primes up to x. Chebyshev's bound is
used in the form π(x)≤Cx/logx for x≥2 with an absolute C.
Statement
h(n!)≫(logn)2: there is an absolute constant c>0 with
h(n!)≥c(logn)2 for all sufficiently large n.
Proof
Write T=τ(n!) and k=h(n!), and let n≥4.
Counting. Every integer 1≤m≤n! is the sum of some set of at most
k distinct divisors of n!, and distinct m need distinct sets. Hence
n!≤i=0∑k(iT).
The binomial tail. If 1≤k≤T then ∑i≤k(iT)≤(eT/k)k:
for 0<x≤1,
i≤k∑(iT)≤x−ki=0∑T(iT)xi=x−k(1+x)T≤x−kexT,
and x=k/T gives (T/k)kek. Taking logarithms in the counting
inequality, for 1≤k≤T,
logn!≤klogkeT≤k(1+logT).
The divisor count.logT=∑p≤nlog(vp(n!)+1). The
primes p≤n are at most n in number and each has
vp(n!)≤n/(p−1)≤n, so they contribute O(nlogn). A prime
p>n has p2>n, hence vp(n!)=⌊n/p⌋. Such a prime
lies in (n/2r+1,n/2r] for exactly one integer r≥0 with
n/2r>n; there ⌊n/p⌋<2r+1, so
log(vp(n!)+1)≤(r+1)log2, and the number of primes in the range is at
most π(n/2r)≤Cn/(2rlog(n/2r))≤2Cn/(2rlogn), using
log(n/2r)>21logn. The primes above n therefore
contribute at most
logn2Cnlog2r≥0∑2rr+1≪lognn,
and altogether logT≪n/logn+nlogn≪n/logn.
Conclusion. Since
logn!≥∑n/2<j≤nlogj≥2nlog2n≫nlogn, the
case k≤T/2 (so k≤T) gives
nlogn≪logn!≤k(1+logT)≪klognn,
that is, k≫(logn)2. In the case k>T/2, every integer 1,…,n
divides n!, so T≥n and k>n/2≫(logn)2. In both cases
h(n!)≫(logn)2.
Qualifications
The source splits at k≤T/2; the binomial-tail bound holds for all
k≤T, so the split is only for convenience. The source also restates the
two questions of Erdős and Graham (pp. 37–38) on h(n!), which are
questions (b) and (c) of Problem 18; the remark
proves that any answer to (c) has exponent at least 2.