The proof is based on inequalities for the function V(x)=T(x)−T(x2)−T(x3)−T(x6)
where
T(x)=∑1≤n≤xlogn
If ψ(x)=∑n≤xΛ(n)
with Λ the von Mangoldt function, then T(x)=∞∑n=1ψ(xn)
and thus
V(x)=ψ(x)+ψ(x5)−2ψ(x6)+ψ(x7)+ψ(x11)−2ψ(x12)+⋯
The coefficients of ψ(xn) in this expression have period 6. In contrast to a similar equation in Chebyshev's Mémoire, the signs in (2) are not alternating, this makes it less straightforward to use (2) to obtain bounds on ψ(x).
For all real x≥1, one has xlogx−x−logx+1≤T(x)≤xlogx−x+logx+1
From (1) and (3) one immediately deduces that for all real x≥6
V(x)≤Bx+4logx−2−2log6
and
Bx−4logx−2+2log6≤V(x)
where
B=12log2+13log3+13log3.
Calculation of upper bound on ψ(x)
Because n→ψ(xn) is decreasing, it follows that ψ(x7)+ψ(x11)−2ψ(x12)≥0
and
ψ(x13)+ψ(x17)−2ψ(x18)≥0
and so on. Therefore from (2) one obtains
V(x)≥ψ(x)+ψ(x5)−2ψ(x6)
Because ψ(x5)−ψ(x6)≥0, it follows also that
V(x)≥ψ(x)−ψ(x6)
One can obtain an upper bound on ψ(x) with a standard calculation: combining (4) and (6) and using −2−2log6<−5, one has for all real x≥6 that
ψ(x)−ψ(x6)≤Bx+4logx−5
Take m∈N such that 6≤x6m<62. Applying (7) several times gives
ψ(x6)−ψ(x62)≤Bx6+4logx6−5ψ(x62)−ψ(x63)≤Bx62+4logx62−5⋯ψ(x6m)−ψ(x6m+1)≤Bx6m+4logx6m−5
Adding the inequalities gives
ψ(x)≤ψ(x6m+1)+65B+4(m+1)logx−(m+1)5
The inequality 6m+1≤x gives m+1≤logxlog6. Also, 1≤x6m+1<6, thus ψ(x6m+1)≤3. Therefore, for all x≥6 one has
ψ(x)≤65Bx+4log2xlog6
One can verify in Mathematica that this inequality is also valid for x≥1
Calculation of lower bound on ψ(x)
One has −2ψ(x6)+ψ(x7)+ψ(x11)≤0
and
−2ψ(x12)+ψ(x13)+ψ(x17)≤0
and so on. Therefore from (2) one obtains
V(x)≤ψ(x)+ψ(x5)
Combining this inequality with (5) and (8) then gives for x≥5
Bx−4logx−2+2log6−65Bx5−4log2x/5log6≤ψ(x)
I simplify this expression slightly by using −2+2log6>0, then
1925Bx−4logx−4log2x/5log6≤ψ(x)
Proof of Bertrand's postulate
Because 6/5B19/25B<2, the upper and lower bound are sufficiently close and one can now deduce Bertrand's postulate in a standard way. I do not do so here, because the post is already quite long. Details can for example be found in Murty's book.
Other posts about Bertrand's postulate
No comments:
Post a Comment