Homepage Solution manuals Ivan Niven An Introduction to the Theory of Numbers Exercise 1.3.53* ($\varepsilon^2$ of Prime Number Theorem.)

Exercise 1.3.53* ($\varepsilon^2$ of Prime Number Theorem.)

Let π ( x ) denote the number of primes not exceeding x . Show that

p x 1 p = π ( x ) x + 2 x π ( u ) u 2 du .

Using Theorem 1.19, deduce that

limsup x π ( x ) x log ( x ) 1 .

Answers

Proof. a) u π ( u ) u is piecewise continuous on [ 2 , x ] , so is Riemann integrable on this interval.

Write p n the n -th prime number : p 1 = 2 , p 2 = 3 , p 3 = 5 , . Let x 2 a real number. Define n by

p n x < p n + 1 .

Then

2 x π ( u ) u 2 d u = i = 1 n 1 p i p i + 1 π ( u ) u 2 d u + p n x π ( u ) u 2 d u . (1)

For u [ p i , p i + 1 [ , π ( u ) is a constant, and equal to i , because the primes not exceeding u are p 1 , p 2 , , p i :

u [ p i , p i + 1 [ , π ( u ) = i .

Therefore

p i p i + 1 π ( u ) u 2 d u = i p i p i + 1 1 u 2 d u = i ( 1 p i 1 p i + 1 ) ,

and similarly

p n x π ( u ) u 2 d u = n ( 1 p n 1 x ) .

Replacing these values in (1), we obtain

2 x π ( u ) u 2 d u = i = 1 n 1 i ( 1 p i 1 p i + 1 ) + n ( 1 p n 1 x ) .

Using π ( x ) = n , this gives

π ( x ) x + 2 x π ( u ) u 2 d u = i = 1 n 1 i ( 1 p i 1 p i + 1 ) + n p n = n p n + i = 1 n 1 i p i i = 1 n 1 i p i + 1 = n p n + i = 1 n 1 i p i i = 2 n i 1 p i = i = 1 n i p i i = 2 n i p i + i = 2 n 1 p i = 1 p 1 + i = 2 n 1 p i = i = 1 n 1 p i

Since p 1 , , p n are the prime numbers p x , we conclude

p x 1 p = π ( x ) x + 2 x π ( u ) u 2 d u . (2)

b) The limit superior of a function f 0 , defined on a neighborhood of + , is defined by

limsup x f ( x ) = lim x ( sup u x f ( u ) ) .

Since x sup u x f ( u ) is a decreasing function, this limit exists in + { + } .

Assume for contradiction that

ρ = limsup x π ( x ) x log ( x ) < 1 .

Then, taking 𝜀 = 1 ρ 2 > 0 , there exists x 0 2 such that

x x 0 | sup u x π ( u ) u log ( u ) ρ | < 𝜀 = 1 ρ 2 ,

so

x x 0 sup u x π ( u ) u log ( u ) 1 + ρ 2 .

Writing k = 1 + ρ 2 , then k < 1 , and for all u x 0 ,

π ( u ) u log ( u ) k ,

that is

u x 0 π ( u ) k u log ( u ) , k < 1 . (3)

By Theorem 1.9.1,

p x 1 p > log ( log ( x ) ) 1 .

Then, by the equality (2), for every x x 0 ,

log ( log ( x ) ) 1 π ( x ) x + 2 x π ( u ) u 2 d u = π ( x ) x + 2 x 0 π ( u ) u 2 d u + x 0 x π ( u ) u 2 d u .

Writing c the constant c = 2 x 0 π ( u ) u 2 d u , this gives, using π ( x ) x ,

log ( log ( x ) ) 1 + c + π ( x ) x + x 0 x π ( u ) u 2 d u 2 + c + x 0 x π ( u ) u 2 d u

The inequality (3) gives

log ( log ( x ) ) 2 + c + k x 0 x 1 u log ( u ) d u = 2 + c + k [ log ( log ( u ) ) ] x 0 x = C + k log ( log ( x ) ) ,

where C = 2 + c k log ( log ( x 0 ) ) is a constant.

Dividing by log ( log ( x ) ) , we obtain, for all x x 0 ,

1 C log ( log ( x ) ) + k ,

therefore

1 lim x + ( C log ( log ( x ) ) + k ) = k .

This shows k 1 : this is a constradiction, since k < 1 , so

limsup x π ( x ) ( x log ( x ) ) 1 .

User profile picture
2024-07-21 16:39
Comments