Notes 1 2018-19 Infinitely Many Primes
Notes 1 2018-19 Infinitely Many Primes
v1 [3 lectures]
One method to show there are infinitely many primes is to start with a
sequence {an }n≥1 of real numbers such
P that an 6= 0 if n is prime and an = 0
if n is not prime, i.e. composite. If n≤N an → ∞ as N → ∞ then there
must be infinitely many primes. Thus we are led to estimating sums and
a fundamental idea is to replace the sum by an integral. One way this is
achieved is by applying the important idea that, for an integrable function
f, Z b
glbf (t) (b − a) ≤ f (t) dt ≤ lubf (t) (b − a) . (1)
[a,b] a [a,b]
Proof For the lower bound in (3) sum the upper inequality in (2) over
n = 1, ..., N. For the upper bound in (3) sum the lower inequality in (2) over
n = 1, ..., N −1 and then change the variable of summation from n to n+1,
though still calling it n.
1
Proof This follows immediately from Lemma 1.1 with f (t) = 1/t.
For our proof of the infinitude of primes we will require the following
lemma. Recall that the Taylor Series for − log (1 − x) is
x2 x3 x4
x+ + + + ...
2 3 4
for |x| < 1. Thus we might expect x to be a good first approximation to
− log (1 − x). The following result quantifies how good an approximation.
As seen in (1)
Z b
glbf (t) (b − a) ≤ f (t) dt ≤ lubf (t) (b − a) .
[a,b] a [a,b]
for N ≥ 2.
2
P
Letting N → ∞ proves that the series p 1/p diverges which it can only
do if it contains a infinite number of terms, i.e. we have an infinitude of
primes.
Proof Let
N = {n ∈ N : p|n ⇒ p ≤ N } .
Note that 1 ∈ N since there are no primes which divide 1 and so the condition
is trivially satisfied.
Another way of writing N is to list all the primes up to N as p1 < p2 <
... < pr ≤ N . By the prime factorisation of integers every n ∈ N can
be written as pa11 pa22 ...par r for some ai ≥ 0 ∀ 1 ≤ i ≤ r while, conversely, every
product pa11 pa22 ...par r with ai ≥ 0 is an element of N . Hence
N = {pa11 pa22 ...par r : ai ≥ 0 ∀ 1 ≤ i ≤ r} .
Next by unique factorisation into prime there is no n ∈ N which is
represented by two different products pa11 pa22 ...par r . Thus we get the first
equality in
X1 X 1
= a1 a2
n∈N
n p p ...par r
a ≥0 1 ≤ i ≤ r 1 2
i
r
Y 1 1 1
= 1 + + 2 + 3 + ... (4)
i=1
pi pi pi
To see this last equality multiply the product out, taking from each
bracket a term of the form 1/pai i for some ai ≥ 0. Multiplying them together
gives a term of the form
r
Y 1 1 1
ai = Q r ai = ,
i=1
p i i=1 p i n
with n ∈ N .
The result (4) can be written more succinctly as
X1 Y 1 1 1
= 1 + + 2 + 3 + ... . (5)
n∈N
n p≤N p p p
3
Thus −1
X1 Y 1
= 1− . (6)
n∈N
n p≤N p
n ≤ N =⇒ n ∈ N or, equivalently, N ⊇ {1 ≤ n ≤ N } .
Hence X1 X1
≤ . (7)
n≤N
n n∈N n
Apply Lemma 1.3 with x = 1/p which is ≤ 1/2 since primes satisfy p ≥ 2.
Then
(1/p)2
1 1 1
0 < − log 1 − − < = .
p p 1 − 1/p p (p − 1)
So
X
1
1
X 1 X 1
0< − log 1 − − < < ,
p≤N
p p p≤N
p (p − 1) 2≤n≤N n (n − 1)
replacing the sum over primes by a larger sum over all integers. Note how
this latter sum starts at n = 2 and not 1, because the smallest prime is 2.
4
Continue using partial fractions
X 1 X 1 1
= −
2≤n≤N
n (n − 1) 2≤n≤N
n − 1 n
1 1 1 1 1 1
= − + − + − + ...
1 2 2 3 3 4
1 1 1 1
... + − + −
N −2 N −1 N −1 N
1
= 1− , by cancellation,
N
< 1. (10)
We say that the sum has “telescoped down”, the second term in a bracket
cancelling the first term in the next. Thus
X
1
1
0< − log 1 − − < 1.
p≤N
p p
By (8) we get
X1
log log (N + 1) < + 1,
p≤N
p
which rearranges to
X1
> log log (N + 1) − 1,
p≤N
p
as required.
5
Definition 1.5 The Riemann zeta-function is defined by
∞
X 1
ζ(s) =
n=1
ns
Example 1.6
∞
X 1
ζ(2) =
n=1
n2
is known to converge.
and this result will be used throughout the course with no further comment.
For s ∈ R, i.e. a real variable we have
Theorem 1.7 For s = σ real, the series defining ζ(σ) diverges for all σ ≤ 1
and converges for all σ > 1 with
1 1
≤ ζ(σ) ≤ + 1.
σ−1 σ−1
6
Proof If σ ≤ 0 then the terms of the series defining ζ(σ) satisfy 1/nσ > 1, in
particular they do not tend to 0 as n → ∞ and so the series cannot converge,
i.e. it diverges.
Assume σ > 0. Lemma 1.1 with f (u) = 1/uσ gives
N +1 N N
du X 1 du
Z Z
≤ ≤1+ ,
1 uσ n=1
nσ 1 uσ
for σ 6= 1.
P
First Case 0 < σ ≤ 1. If σ = 1 then, since by CorollaryP1.2, n≤N 1/N >
log (N + 1) which tends to ∞ as N → ∞, we deduce that ∞ n=1 1/n diverges.
using N > 0. Then the sequence of partial sums n≤N 1/nσ is an increasing
P
sequence (positive terms 1/nσ are added as N increases) bounded above.
Thus, by a first year analysis result, the sequence of partial sums converges,
which is the definition that the series ζ(σ) converges. Also
1
ζ(σ) ≤ 1 − . (12)
1−σ
Next, 1 − σ < 0 so (N + 1)1−σ → 0 as N → ∞. So, in the limit, the first
inequality in (11) gives
1 1
ζ(σ) ≥ − = . (13)
1−σ σ−1
Combine (12) and (13) to give the displayed conclusion.
Note that for complex s = σ + it Theorem 1.7 says that ζ(s) converges
absolutely for Re s > 1 because, for such s,
7
∞ ∞
X 1 X 1
= = ζ(σ) .
ns nσ
n=1 n=1
N
Y
lim (1 + an )
N →∞
n=1
8
is convergent.
Proofs of both results can be found in Appendix C of The Prime Number
Theorem by G.J.O. Jameson, Pub. London Mathematical Society, Student
Text 13, 2003. It can also be found in the Appendix on my web site.
For s ∈ C, i.e. a complex variable, we have
Solution Assume Re s > 1. Use again the fact that an infinite series over
primes of positive terms is less than the series over all integers, i.e.
X 1 X 1 ∞
− =
X 1
ps σ
≤ = ζ(σ) .
p p
p n=1
nσ
converges.
For real s > 1 this result is due to L. Euler, 1737, and for this reason:
Definition 1.10 Infinite products over primes are known as Euler Prod-
ucts.
9
The product seen in Example 1.9 is often referred to as the Euler Product
because of
Proof Let N > 1 and N as seen in the proof of Theorem 1.4. The exponent
of s makes no differences to the arguments seen in the proof of Theorem 1.4
and so, by the unique factorization of integers, we have
Y −1 X
1 1
1− s = s
.
p≤N
p n∈N
n
Consider
−1 X
Y ∞
1 1 X 1 X 1
ζ(s) − 1− s = − = .
p n=1 ns ns ns
p≤N n∈N n∈N
/
Next we will take the modulus into the series, allowable since the resulting
series is convergent as the subsequent argument will show. Thus
X 1 X 1 X 1
≤ = .
s
n ns nσ
n∈N
/ n∈N
/ n∈N
/
n ∞
X Z du du 1
Z
≤ = = .
n≥N +1 n−1 uσ N u σ (σ − 1) N σ−1
Therefore −1
Y 1 1
ζ(s) − 1− s ≤ .
p (σ − 1) N σ−1
p≤N
Let N → ∞ when the bound here tends to 0 since σ > 1. This means we
have again shown that the infinite product converges but now we know the
limit is ζ(s).
10
Corollary 1.12 For Re s > 1
ζ(s) 6= 0.
Proof From Example 1.9 we see that the Euler product for ζ(s) converges
which means by definition of convergence for infinite products that the Euler
product is non-zero. Thus by Theorem 1.11 we have ζ(s) 6= 0 for Re s > 1.
An advert for a substantial part of this course is that we will later prove
ζ(s) 6= 0 for Re s ≥ 1. This may not look much of an improvement but the
fact that ζ(s) 6= 0 for Re s = 1 is equivalent to the Prime Number Theorem,
another major result of this course.
as seen in the previous proof for complex s in place of the real σ. Take
logarithms,
! # −1 X
X 1 Y 1
1
log σ
≤ log 1− σ = − log 1 − σ
1≤n≤N
n p≤N
p p≤N
p
X 1 X
1
1
= σ
+ − log 1 − σ − σ . (14)
p≤N
p p≤N
p p
11
By Lemma 1.3 with x = 1/pσ we have
X
1
1
X
1
0 ≤ − log 1 − σ − σ ≤ σ σ
p≤N
p p p≤N
p (p − 1)
N N
X 1 X 1
≤ ≤ since σ > 1
n=2
nσ (nσ − 1) n=2 n (n − 1)
≤ 1,
for all N ≥ 1, as seen in an earlier proof. Hence the sum over primes
converges and
X
1
1
0≤ − log 1 − σ − σ ≤ 1. (15)
p
p p
Let N → ∞ in (14) . On the left hand side the continuity of log gives
! # ! #
X 1 X 1
lim log σ
= log lim σ
= log ζ(σ)
N →∞
1≤n≤N
n N →∞
1≤n≤N
n
That is, the sum over primes converges, and the lower bound in the Theorem
follows from (15) and Theorem 1.7.
In later Chapters we will sharpen and generalise all these methods and
results.
12