Source. Thomas F. Bloom, On a density conjecture about unit fractions,
arXiv:2112.03726v2 (12 October 2023). Printed and PDF page numbers agree.
Use R(A), Aq, QA and R(A;q) as defined in
Lemma 6;
QA consists of exact prime powers, and ω(n) counts distinct
prime divisors. Unqualified sums over q are sums over prime powers.
Statement (Lemma 4, p. 13). Let 0<ϵ<1/2, and let N be
sufficiently large in terms of ϵ. Suppose that A is a finite set
of integers satisfying
R(A)≥(logN)−ϵ/2
and
(1−ϵ)loglogN≤ω(n)≤2loglogN(n∈A).
Then
q∈QA∑q1≥(1−2ϵ)e−1loglogN.
Rewritten proof. Set
σ=q∈QA∑q1,I=[(1−ϵ)loglogN,2loglogN].
Every n∈A is the product of its distinct exact prime-power components,
and their number is ω(n)∈I. Summing over all possible collections
of components, and then enlarging to all ordered choices, gives
R(A)≤t∈I∑t!σt≤t∈I∑(teσ)t,
where the last inequality uses t!≥(t/e)t and t ranges over the
integers in I.
If σ≥(1−ϵ)loglogN, the claimed weaker bound is immediate.
Otherwise σ<t throughout I. The function (eσ/t)t is then
decreasing in t, so there are at most 2loglogN terms and
(logN)−ϵ/2≤R(A)≤2loglogN((1−ϵ)e−1loglogNσ)(1−ϵ)loglogN.
Taking the ((1−ϵ)loglogN)-th root yields
σ≥(1−ϵ)e−1loglogNe−ϵ/(2(1−ϵ))(2loglogN)−1/((1−ϵ)loglogN).
For 0<ϵ<1/2,
e−ϵ/(2(1−ϵ))≥1−ϵ. Choose N so large that
(2loglogN)2/loglogN≤1+ϵ2.
Since 1/(1−ϵ)≤2, the final factor in the preceding lower bound is
at least (1+ϵ2)−1. Consequently
σ≥1+ϵ2(1−ϵ)2e−1loglogN≥(1−2ϵ)e−1loglogN,
as required.
Dependencies
The elementary bound t!≥(t/e)t and prime-power factorization;
no earlier numbered lemma is used.
Bears on