Infinitude of Primes
Via Harmonic Series

It's well known that the harmonic series diverges. We may thin it somewhat by removing all terms with denominators divisible by 2. What is left over is the series of the reciprocals of the odd whole numbers:

(1) Σi = 1 1/(2i - 1).

This series is divergent:

Σi = 1 1/(2i - 1) > Σi = 1 1/2i = ½Σi = 1 1/i,

which should be understood as a manipulation of partial sums of the two series. The partial sums of the reciprocals of the odd numbers are greater than the corresponding sums of the harmonic series divided by 2. Since the latter diverge, so do the former.

Next we exclude from (1) the terms with the denominators that are multiples of 3. The remaining terms are the reciprocals of the integers in one of the forms: either 6n + 1 or 6n + 5. The series of reciprocals taken separately diverge and so does the whole remaining series.

We may continue this process indefinitely removing eventually the reciprocals of all the composite numbers and leaving only those of the primes. The sieve of Eratosthenes of sorts.

I'll use the classic argument from Hardy and Wright to show that the remaining series, i.e. the series of the reciprocals of the primes is still divergent. From here it will follow that the number of primes cannot be finite. (This prooof is also discussed at length in Ingenuity in Mathematics, and a different proof can be found in, say Proofs from THE BOOK.)

Assume all the primes are enumerated according to their magnitude: 2, 3, 5, ..., pk , ..., and consider the first k. For a given x, define N(x) as the number of terms in the set

{nx: n is not divisible by any prime beyond pk}.

In words: N(x) is the number of n's not exceeding x with only prime divisors p satisfying ppk. Any n from that set can be represented in the form

n = (n1)2 m,

with a square-free m:

m = 2a13a25a3· ... ·pkak,

where every a is either 0 or 1. There are just 2k possible combinations of the k exponents and so not more than 2k possible values for m.

On the other hand,


and so there are not more than x different values for n1 implying that

N(x) ≤ 2kx.

If the number of primes is finite and their total is k then N(x) = x for every x and so

x ≤ 2kx


x ≤ 22k

which is false for x > 22k.

We now use the same approach to prove the following


Let 2, 3, 5, ..., pk be an enumeration of the primes. Then the series

Σ 1/pk = 1/2 + 1/3 + 1/5 + ... + 1/pk + ...

is divergent.


If the series convergent then its partial sums tend to a limit so that the remainders

rn = Σk > n 1/pk = 1/pn +1 + 1/pn +2 + ...

become arbitrary small. In particular, there is n such that rn < 1/2:

1/pn +1 + 1/pn +2 + ... < 1/2.

The number of nx which are divisible by p is at most x/p. Hence, x - N(x), i.e., the number of nx which are divisible by at least one of pn +1, pn +2, ... does not exceed

x/pn +1 + x/pn +2 + ... < x/2.

From the foregoing discussion we conclude that

x/2 ≤ N(x) ≤ 2nx,

or, again, a contradictory

x ≤ 22(n + 1).

We see that the assumption that the series of the reciprocals of the primes is convergent leads to a contradiction. Therefore, the series diverges.


  1. M. Aigner, G. Ziegler, Proofs from THE BOOK, Springer, 2000
  2. G. H. Hardy and E. M. Wright, An Introduction to the Theory of Numbers, Fifth Edition, Oxford Science Publications, 1996.
  3. R. Honsberger, Ingenuity in Mathematics, MAA, New Math Library, 1970

|Contact| |Front page| |Contents| |Algebra| |Up| |Store|

Copyright © 1996-2017 Alexander Bogomolny


Search by google: