Source: published paper, printed pp. 413–414
(PDF pp. 3–4), Theorem 3.
Statement
For every integer k≥2, with p1=2,
pk≥k(logk+loglogk−1).(1)
This is the inequality actually displayed in Theorem 3. The title and
abstract say “greater than”; the compilation does not silently replace
the displayed ≥ by a strict all-k conclusion. The proof below has
strict margins when pk≥1011. The smaller range is imported at
its stated weak scope. Equation numbers on this page are its own; the
print's (1) and (2) on p. 413 are Robin's estimate for θ(pk) and
Schoenfeld's final-range estimate.
Full proof relative to the exact external estimates
The cited Robin finite-range result proves (1)
for 3≤pk≤1011. This includes every k≥2 whose prime lies
in that range. No fresh enumeration of that range is asserted.
For the remaining ranges put
t=logk,q=logpk,a=2.1454,c=0.0077629.
They have k≥6, so all required inequalities from
Lemma 1 apply. Since pk≥k+1, we have t<q.
The bound pk≤klogpk also gives
t≥q−logq.(2)
First range: 1011≤pk≤e500.
We have t<q≤500. If t≤20, Lemma 1 would give
pk≤et(t+logt)≤e20(20+log20)<1011,
where monotonicity is valid for t≥log6>1, and the last inequality
is the certified endpoint bound. Thus 20<t<500.
Schoenfeld's first-range estimate and pk≤kq give
pk≥θ(pk)−cqpk≥θ(pk)−ck.
Insert Robin's lower estimate for θ(pk) to obtain
pk≥k(t+logt−1+tlogt−a−c).
The complete calculus bound g(t)>c on [20,500] proves (1), with
strict inequality in this range.
Intermediate range: e500<pk<e1800.
Since q−logq is increasing for q>1, equation (2) and
log500<7 give t>493; also t<q<1800.
By Theorem 2 and θ(x)≤ψ(x),
θ(pk)−pk<εpk,ε=0.905⋅10−7.
Combining this with Robin's estimate and
pk≤k(t+logt) gives
pk>k(t+logt−1+tlogt−a−ε(t+logt)).
The correction term is positive because
t(t+logt)logt−a>1.6⋅10−6>ε.
Thus (1) again holds strictly.
Final range: pk≥e1800.
The external Schoenfeld estimate, with η4=16570000, yields
∣θ(pk)−pk∣≤η4q4pk≤η4q3k≤18002η4tk,
using q≥1800 and q>t>0. Equation (2) gives
t≥1800−log1800>1792, hence logt>7.49.
Robin's estimate now implies
pk≥k(t+logt−1+tlogt−a−η4/18002).
The numerator of the last fraction exceeds 7.49−7.26>0 by the
certified rational constant inequality. This proves strict inequality
in the final range and completes the weak all-k statement (1).
Every essential same-paper step is supplied at the linked pages. The
analytic explicit formula, finite zero verification, Robin estimates,
Schoenfeld estimates and Lemma 1 are the exact external boundaries.
Use in the covering-system termination proof
The BBMST termination induction
needs only (1). In its notation
λi=logi+loglogi−3, the strict increment
i(λi−λi−1)>1 combines with the weak prime bound to give
pi−1≥i(λi+2)−1>i(λi−1+2).
Thus no stronger external inequality is needed for that strict
inductive step. This identifies the interface; the canonical BBMST
proof is not duplicated or modified here.
Bears on
- Problem 2: the weak
bound (1) is the external prime input of BBMST's Theorem 6.1, on which
their bound 616000 for the minimum modulus of a distinct cover rests.
- Problem 7: through the
same Theorem 6.1, (1) is an input to the 2021 square-free obstruction,
which covers only the square-free special case.