You are on page 1of 5

Elem. Math.

76 (2021) 28 – 32 
c Swiss Mathematical Society, 2021
0013-6018/21/010028-5
DOI 10.4171/EM/425 Elemente der Mathematik

Infinitude of primes: Euclid’s proof


using angles between lattice vectors

Dario T. de Castro

Dario T. de Castro received his PhD in Quantum Optics from Universidade Federal
do Rio de Janeiro in 2004. Since then, he held several teaching positions at private
universities before joining the group at Instituto Federal do Rio de Janeiro – campus
Nilópolis, where he works as a teacher and researcher since 2008.

1 Introduction
Euclid’s geometric and constructive proof for his theorem on the infinitude of primes [3]
has been widely considered as an elegant example of cleverness in mathematical demon-
stration. Today, there exist several different proofs of this result [1, 4, 6]; each of these
proofs contributes to our understanding of this result in a different manner. In this paper,
we present a version of Euclid’s proof using a geometric argument based on vectors in k
dimensions and the angles between them. Subsequently, we enunciate and solve a purely
geometric problem, which is similar to proving Euclid’s theorem. Finally, based on the
same geometric context, we propose two additional proofs of the infinitude of primes.
One of these proofs depends on a theorem proved by Gersonides in 1342, which states
that 8 and 9 are the only consecutive powers of 2 and 3, with exponents greater than
one. This theorem is a special case of the Catalan–Mihăilescu theorem, proved by Preda
Mihăilescu in 2002, which states that 8 and 9 are the only consecutive integers that are
perfect powers. Although k-dimensional geometry is a complex subject, throughout this

Jeder neue Beweis eines klassischen Theorems lässt dieses in einem neuen Licht er-
scheinen. Zu Euklids Satz über das Nichtabbrechen der Primzahlfolge existiert eine
Vielzahl von Beweisen. Im vorliegenden Artikel folgt der Autor zwar Euklids ur-
sprünglichen Argumenten, stellt aber einen neuen, geometrischen Bezug zu endlichdi-
mensionalen Gittervektoren und deren Zwischenwinkeln her. Durch diese Sichtweise
ergeben sich gleich noch zwei weitere neue Beweise von Euklids Theorem, von denen
einer auf dem Satz von Gersonides beruht, einem Spezialfall des Catalan–Mihăilescu
Theorems.
.
Infinitude of primes: Euclid’s proof using angles between lattice vectors 29

study, we only consider pairs of vectors and the angles between them. Specifically, we ver-
ify whether pairs of consecutive vectors in a given sequence are orthogonal; this task can
be easily performed by obtaining the scalar product of the vectors. The next section will
be devoted to the proof of the infinity of primes using finite dimensional lattice vectors
known as exponent vectors.

2 The proof using lattice vectors


We shall now enunciate Euclid’s theorem and proceed to prove it.
Theorem 1 (Infinitude of Primes – IP) There are infinite prime numbers.
Proof #1 of IP. By contradiction, suppose that there are only k prime numbers with k ≥ 2.
We propose a bijective mapping of positive integers n – which can presumably be prime
factorized in the form n = 2α1 · 3α2 · 5α3 · · · · · pkαk – into points (or vectors) of an integer
k-dimensional lattice, restricted to the region where all coordinates are non-negative, i.e.,
Zk+ (or the first k-hyperoctant). A k-hyperoctant generalizes the idea of a quadrant (or
octant) for k dimensions. In the aforementioned mapping, the number n defined above is
mapped onto a vector P(n) = (α1 , α2 , α3 , . . . , αk ) in Zk+ . This map is sometimes referred
to as the exponent vector of n [5]. Because every two consecutive integers are co-prime, it
follows that pairs of consecutive numbers in this representation must always yield pairs of
orthogonal vectors in Zk+ . Additionally, considering the bijective nature of the mapping, all
integer points in this hyperoctant must be the image of some positive integer. Thus, based
on the above assumptions, we can state that, for every k, when the vector corresponding to
a given integer n has an orientation significantly close to that of a line r passing through
the origin and through the point with all coordinates equal to one, it will be impossible for
the image of its successor to exhibit orthogonality and still be in Zk+ . In fact, we leave it
to the reader to verify that the angles θk between the line r and principal axes are always
the largest possible angles starting from r and ending in the first k-hyperoctant. Moreover,
they are always smaller than π/2, as can be deduced from the following expression:
   
−1 (1, 0, 0, . . . , 0, 0) · (1, 1, 1, . . . , 1, 1) −1 1
θk = cos √ = cos √ , (1)
k k
where we choose the basis vector (1, 0, 0, . . . , 0, 0) as an example, without a loss of gen-
erality. Thus, because line r contains points of the integer lattice, it follows that there must
be more than k prime numbers, thus proving the theorem. 
Remark 1. Recall that the primorial of a prime, pk , denoted by pk #, represents the product
of the first k prime numbers. In the proof above, the point in Zk+ having all coordinates
equal to one plays the same role as the primorial pk # in Euclid’s demonstration. However,
the geometric context used in this study seems to make it clearer which other integer
numbers are also effective at triggering the need for new primes in addition to pk #.

3 A purely geometric theorem


Considering the possible pedagogical application of this study, we believe it would be
useful to present the above theorem as a purely geometric problem; this is defined as
follows:
30 D.T. de Castro

Theorem 2. Let k be a positive integer and let {v n }∞ n=1 be an infinite sequence of vectors
in k dimensions, satisfying the following conditions:
1. all vectors in the sequence are non-negative integer vectors, i.e., their Cartesian
coordinates are all non-negative integers (v i ∈ Zk+ , ∀ i ∈ N∗ ).
2. each vector is orthogonal to its successor (v i ⊥ v i+1 , ∀ i ∈ N∗ ).
3. to each point in this lattice (Zk+ ), there exists only one associated vector in the se-
quence with the same coordinates (Q ∈ Zk+ ⇔ ∃! i ∈ N∗ | v i = Q).
Then, it follows that the construction of this infinite sequence of vectors is impossible.
Proof. By contradiction, suppose there is such a sequence as described above, which im-
plies the existence of a bijection between the index i ∈ N∗ of the vectors v i in the sequence
and the points of Zk+ ; therefore, for some positive integer value i 0 of the index, the vec-
tor v i0 corresponds with the specific lattice point P = (1, 1, 1, . . . , 1, 1). However, the
maximum angular variation θk from the initial vector v i0 = P that still corresponds to
orientations of vectors in Zk+ is given by the following:
 
1
θk = cos−1 √ < π/2. (2)
k
We conclude that the next vector, v i0 +1 , of this sequence cannot exist inside Zk+ . Conse-
quently, the infinite sequence is not constructible, which completes the proof. 

4 Two additional proofs of Euclid’s theorem


Based on the geometric context described above, we present two additional proofs of Eu-
clid’s theorem. In the first proof, we shall use as a lemma a theorem proved by the French
astronomer Gersonides in 1342, which states that 8 and 9 are the only consecutive powers
of 2 and 3 with exponents greater than one. Here, assuming that there are only k primes
and using the same exponent vector mapping, we noted that the orthogonality requirement
and the spatial restriction imposed on the exponent vector compelled us to explore another
promising path. We enunciate and prove this lemma, as shown below.
Lemma. The equation 3 x − 2 y = 1 does not have solutions for integers x, y > 3.
Proof. Because y > 2, it follows that 3x ≡ 1 mod 4. This implies that x is even. Let
x = 2m with m ≥ 2. The equation in the lemma thereafter reads:
2 y = (3m − 1)(3m + 1) ⇒ (3m − 1) = 2u , (3m + 1) = 2v
(3)
⇒ 2 = 2v − 2u ⇒ v = 2, u = 1 and m = 1,
which proves the lemma. 
Furthermore, we consider positive integers n i defined as:
pk #
ni = , with 1 ≤ i ≤ k, (4)
pi
where each n i corresponds to an exponent vector of k dimensions with all components
equal to one, except the one with index i , which is zero. Clearly, for such vectors, there
Infinitude of primes: Euclid’s proof using angles between lattice vectors 31

exists only one orthogonal direction that still contains points in Zk+ . Effectively, the expo-
nent vector associated with n i +1 must only have the i -th component as nonzero, such that
the scalar product between it and its predecessor vanishes. As an example, for k = 4, n 1
will be mapped onto (0, 1, 1, 1), such that the only suitable form of its consecutive would
be (a, 0, 0, 0), for some natural number a. In the next proof, we intend to show that, for
every k > 3, either n 1 + 1 or n 2 + 1 will not be allowed to have its exponent vector in
Zk+ , implying that k is not sufficiently large (as a spatial dimension) to accommodate the
image of all positive integers consistently.
Proof #2 of IP. We prove by contradiction. Suppose there are only k prime numbers:
2, 3, 5, . . . , pk , with k ≥ 3. Let n 1 and n 2 be two positive integers given by
pk #
ni = , with i = 1, 2. (5)
pi
For each n i we must have
α
n i + 1 = pi i , (6)
with αi > 1, in general. Let us assume that n 1 and n 2 have successors of this form. Thus,
recalling that k ≥ 3 (meaning that pk # ≥ 30), we have
pk #
n1 + 1 = + 1 = 2α1 , with α1 > 3
2 (7)
pk #
n2 + 1 = + 1 = 3α2 , with α2 > 2.
3
This implies that
3α2 +1 − 2α1 +1 = 3 − 2 = 1, (8)
which contradicts the lemma. 
Remark 2. Although the above-mentioned lemma is a special case of the Catalan–Mihăi-
lescu theorem, this result could not be used in this work because the infinitude of primes is
assumed in the proof itself. In contrast, the lemma (Gersonides’ theorem) does not consider
the infinitude of primes, and thus, it can be used here.
A simpler version of the proof presented above can be obtained by considering the prede-
cessor and successor of n 1 as follows.
Proof #3 of IP. We prove by contradiction. Assume that there are only k prime numbers,
with k ≥ 3. Let n 1 be defined as follows:
pk # pk #
n1 = = ; (9)
p1 2
we note that n 1 ≥ 15. As consecutive integer numbers are co-prime, both the predecessor
and successor of n 1 must be powers of 2, such that
2d+1 − 2d = 2d = 2 ⇒ d = 1, (10)
which leads to an absurdity, because 2 cannot be (n 1 − 1), completing the proof. 
Remark 3. Demonstrations very similar to the above one are possible considering the
predecessor and successor of n i , with i = 2, 3, . . . , k.
32 D.T. de Castro

Remark 4. Although proofs #2 and #3 were inspired by the geometry of exponent vec-
tors, only purely arithmetic properties were used in them. These proofs are variations of
Euclid’s argument that use special and simpler to prove cases of two theorems from Cos-
grave [2] which state that both n i + 1 and n i − 1 cannot be powers of pi , for k ≥ 4 and
1 ≤ i ≤ k.

5 Conclusion
In this paper, we presented a new geometric version of Euclid’s proof of the infinitude of
primes based on vectors and the angles between them. The concept of exponent vector was
used to map every positive integer hypothetically into vectors in Zk+ . Hitherto, neither the
finite dimension hypothesis for the exponent vectors nor their geometric properties have
been used to demonstrate the infinity of the primes. We believe that this geometric proof
is easy to understand for students and can provide them with an instructive mathematical
experience because determining if two vectors in k dimensions are orthogonal is a task that
can be easily achieved using the scalar product. In addition, our study proposes a purely
geometric formulation of Euclid’s theorem with no connection to number theory concepts.
Finally, inspired by the same geometric concept, two additional proofs of Euclid’s theorem
were presented. The first of these proofs required Gersonides’ theorem, which is a special
case of the Catalan–Mihăilescu theorem.
Acknowledgement. The author would like to express his gratitude to professors C.G. Mo-
reira, R. Melo, and G. Guedes for their valuable observations and suggestions. Special
thanks are also due to the anonymous referee for his fruitful suggestions which have im-
proved this paper.

References
[1] Aigner, M., Ziegler, G. (2018). Proofs from THE BOOK, 6th ed. Berlin, New York: Springer-Verlag, pp.
3–8.
[2] Cosgrave, J.B. (1989). A remark on Euclid’s proof of the infinitude of primes. Amer. Math. Monthly. 96:
339–341.
[3] Heath, T.L. (1956). The Thirteen Books of the Elements, Vol. 2, Books III-IX, 2nd ed. New York, USA:
Dover, pp. 412–413
[4] Mes̆trović, R. (2018). Euclid’s Theorem on the Infinitude of Primes: A Historical Survey of its Proofs (300
B.C.–2017). Preprint arXiv:1202.3670v3.
[5] Pomerance, C. (1996). A tale of two sieves. Notices Amer. Math. Soc. 43: 1473–1485.
[6] Ribenboim, P. (1996). The New Book of Prime Number Records. New York, USA: Springer-Verlag, pp.
3–17.

Dario T. de Castro
Instituto Federal do Rio de Janeiro
Campus Nilópolis
Rua Cel. Délio Menezes Tavares 1045
Nilópolis – Rio de Janeiro
RJ – Brazil – CEP 26530-060
e-mail: dario.neto@ifrj.edu.br

You might also like