Infinitude of Primes
Infinitude of Primes
ALDEN MATHIEU
1. Introduction
The question of how many primes exist dates back to at least ancient Greece,
when Euclid proved the infinitude of primes (circa 300 BCE). Later mathematicians
improved the efficiency of identifying primes and provided alternative proofs for the
infinitude of primes. We consider 6 such proofs here, demonstrating the variety of
approaches.
We follow Ronan O’Gorman’s presentation of the material from [1], including a
proof of Lagrange’s Theorem.
2. Infinitude of Primes
In the proofs below, we denote the set of prime numbers P = {2, 3, 5, ...}. We
also use two key facts:
2.1. First Fact. The set of natural numbers N = {1, 2, 3, ...} is unbounded, because
there is a natural ordering on N and we can always add 1 to the ”largest” n ∈ N to
generate an even larger one.
2.2. Second Fact. Prime factorization is unique. That is, for all natural numbers
n 6= ±1, we can uniquely represent n as the product of primes: n = pk11 pk22 · · · pkr r , pi ∈
P, ki ∈ Z and ki 6= 0, ∀i.
The below proofs consist of repackaging these two facts increasingly cleverly,
observed O’Gorman.
4. Goldbach (1742)
Goldbach follows a similar approach, while making more specific demands on
the form of the prime.
n
4.1. Fermat numbers. We define a Fermat number to be of the form Fn = 22 +1
for n = 0, 1, 2, . . . Goldbach noted that this form means all Fermat numbers are
relatively prime. We prove this below.
The following proof was omitted from O’Gorman’s presentation for time but we
include it here for completeness.
n−1
Y
Theorem 4.2. We can generate Fermat numbers using the recursion Fk = Fn − 2
k=0
(n ≥ 1).
It follows immediately that since the only common divisor any of them share is
q = 1, there are infinitely many primes.
5. Mersenne Primes
Similar to the two previous proofs, we consider prime ”Mersenne numbers,”
named for the 17th-century friar Marin Mersenne who studied them.
We first state and prove Lagrange’s Theorem, which will be used in the proof
regarding Mersenne primes.
Transitive
a ∼ b, b ∼ c =⇒ ba−1 ∈ U
=⇒ (ba−1 )(cb−1 ) ∈ U
=⇒ ea−1 c ∈ U
=⇒ a−1 c ∈ U
=⇒ ca−1 ∈ U ∴ a ∼ c
Thus it is indeed a valid equivalence relation and hence it forms a partition.
The equivalence class of a, [a], is the right coset U a = {xa : x ∈ U }. Since
the order of Ua equals the order of U, we see that G decomposes into equivalence
classes of size |U | =⇒ |U | divides |G|.
6. Euler (1737)
This proof, involving the manipulation of infinite series, is O’Gorman’s least
favorite. While [1] relies on a proof using logarithms, we follow O’Gorman’s pre-
sentation in using harmonic series.
Let P = {p1 , p2 , p3 , · · · } be the set of primes in increasing order. Let π(n) be
the function that couns the number of primes less than or equal to n ∈ R : π(n) :=
#{p ∈ P, p ≤ n}. Let Hn be the harmonic series: Hn = 1 + 12 + 13 + · · · + a1 , which
we know diverges by first year analysis.
Theorem 6.1. Hn diverges, thus π(n) diverges, and thus the number of primes is
infinite.
1
P
Proof. Let σn = m∈N m where m has only prime divisors ≤ n. Note that Hn < σn
(because Hn is contained in it) though we do not know if σn converges.
But because both series have exclusively positive terms, by first year analysis,
we know that if they are convergent, then they are absolutely convergent.
We rearrange σn , using our Second Fact:
4 ALDEN MATHIEU
π(n)
YX 1
σn ≤ ( i)
p
k=1 i≥0
Since the sum is positive and geometric, it converges; thus we have an absolutely
convergent series and equality:
π(n)
YX 1
σn = ( i)
p
k=1 i≥0
π(n)
Y 1
and note that pk ≥ k + 1.
k=1
1 − p1k
Because the right hand side is a telescoping series equal to π(n) + 1, we have
Hn < π(n) + 1, and since Hn diverges, π(n) + 1 must also diverge. Hence, the
quantity of primes is infinite.
7. Fürstenberg (1955)
This proof, written while Fürstenberg was an undergraduate in New York, relies
on topology and proof by contradiction.
Theorem 7.1. P cannot be finite.
Proof. We define sets Na,b = {a+nb : n ∈ Z} where a, b ∈ Z and b > 0. We define a
topology on Z(τ ) such that A ∈ τ if and only if ∀a ∈ A, ∃b ∈ N such that Na,b ∈ A.
O’Gorman omitted the proof that τ forms a topology but it is easily verified:
• ∅ is open by definition, and the whole space Z is N1,0 and hence also open.
• Any union of open sets is open: let ∪i=1 Ai be a union of open sets and let
x ∈ ∪i=1 Ai . Thus for any aj such that Naj ,x belongs to the open set Aj ,
Naj ,x also belongs to the union.
• Any finite intersection of open sets is open: let Ai , Aj be open and let x
be in their intersection. Then ∃ai such that Nai ,x ∈ Ai and ∃aj such that
Naj ,x ∈ Aj . We pick a = lcm(ai , aj ), which means Na,x = Ai ∩ Aj .
Hence, this defines a valid topology τ on Z.
We observe that any A 6= ∅ in the topology τ is infinite, because by the First
Fact, N is infinite.
Na,b = Z\ ∪b−1 b−1
i=1 Na+i,b where the union ∪i=1 Na+i,b is an open set. Hence, Na,b
is the complement of an open set and therefore closed.
SIX PROOFS OF THE INFINITUDE OF PRIMES 5
By the Second Fact, for all integers n ± 1, n ∈ N0,p for some prime p ∈ P. We
consider the set {-1, 1}. By the above, this clearly equals Z\ ∪p∈P N0,p .
If P were finite, this union would be a finite union of closed sets and hence
closed. As the complement of a closed set, {−1, 1} would be open and thus would
be infinite, a contradiction.
We conclude P must be infinite.
8. Erdös (1938)
Similiar to Euler’s proof, this is O’Gorman’s favorite. We prove the infinitude
of primes, and get that the sum of reciprocals of primes diverge for free.
1
P
Theorem 8.1. p∈P p diverges and hence P is infinite.
Proof. We assume p∈P p1 converges. If so, then there exists some k ∈ N such that
P
1 1
P
i≥k pi < 2 . Then we can extend this to arbitrary N ∈ N:
XN N
(1) <
pi 2
i≥k
This is useful becasue it tells us something about ”big primes” and we can show
that there isn’t enough ”small primes.”
We define ”big primes” Bn := {m < n : pi | m for some i ≥ k} and ”small
primes” Sn := {m < n : pi | m for some i < k}. Therefore, all numbers are
divisible by some Bn , Sn .
We can form a partition since #Bn + #Sn = n. Note that the floor function
b pNi c counts the multiples of pi ∈ Z. This places a bound, using (1) :
X N N
(2) Bn ≤ b c<
pi 2
i≥k
where am is square-free and equals 1 copy of each prime divisor raised to an odd
power.
Thus by our Second Fact, m is the product of distinct ”small primes.” Due to
#Sn , there are 2k−1 possibilities for am . √ √ √
Returning to (3), b2m ≤ m < n, so bm ≤ m < √
n =⇒ #Sn √= √ 2k−1 n. We
√
choose n large, such that n > 2k =⇒ 2k−1 < 2n =⇒ #Sn < n2 n = n2 . But
by (2), #Sn + #Bn < n, a contradiction. Hence, P must be infinite.
References
[1] Martin Aigner and Günter M. Ziegler. Proofs from The Book, chapter 1, pages viii+308.
Springer-Verlag, Berlin, fifth edition, 2014. Including illustrations by Karl H. Hofmann.