11

I was thinking on the twin prime conjecture, that there are an infinite number of twin primes... I came up with a proof. I have to think that it is incomplete or wrong, because many great minds have thought on this previously. I can't see the issue, so I thought I would raise it in a broader forum. What is wrong with this proof?

1) A number n is prime iff n mod p is non zero for every prime number 1 < p < n

This is easy to prove from the definition of a prime number and mod. It just says that n is not equally divisible by another prime, making it prime.

Let $p_n$ = the nth prime $p_0 = 1$, $p_1 = 2$ ...

Consider $N_n = \Pi_0^n p_n$.

It is easy to see that $N_n \mod p_j = 0 $ for all primes $0 < j < n$

$(N_n + 1) \mod p_j = 1 $ for all primes $0 < j < n$, and therefore, from #1 must be prime

$(N_n - 1) \mod p_j = (p_j - 1) $ for primes $0 < j < n$, and therefore, from #1 must be prime as well

The set $p_n$ has infinite members (as shown by Euclid) , so there are infinite $N_n$. Therefore there are an infinite set of twin prime numbers $(N_n-1,N_n+1)$ .

Eric Stucky
  • 13,018
Edward
  • 221
  • 2
  • 6

3 Answers3

40

This repeats a common misconception about Euclid's proof. Your argument does not show that either $N_n+1$ or $N_n-1$ is prime, but rather that these numbers must be divisible by a prime greater than $p_n$. Indeed, $N_4=210$ has $N_4+1$ prime but $N_4-1=209$ is divisible by $11$.

Kevin Carlson
  • 54,515
  • 5
    (An aside: the question of whether $1$ should be a prime or not is a sort of red herring, insofar as it doesn't really matter: other definitions and assertions would simply have to be modified accordingly. Indeed, until late in the 19th century, $1$ was usually considered prime, apparently. The point is that we understand that specific aspect, and it's not really the issue at all...) – paul garrett Dec 07 '14 at 23:19
  • I had heard that 1 can't be a prime because it breaks prime factorization... – k_g Dec 08 '14 at 04:39
  • 4
    @k_g: that's not why it can't be a prime, that's why it's convenient for it not to be a prime. Unique prime factorisation says that every integer has a prime factorisation that is unique up to the order of factors. If we called 1 prime then we'd have to say unique up to the order of factors and the number of times 1 occurs, which would be annoying. Then again, with 1 not prime we have a definition/theorem that $p$ is prime iff p is not zero or a unit and $\forall a \forall b (p|ab \Rightarrow p|a \lor p|b)$. We choose the lesser inconvenience, but "can't" is overstating the problem :-) – Steve Jessop Dec 08 '14 at 10:57
  • 3
    Anyway, there are statements true of primes, statements true of units, and statements true of both primes and units. It's largely a matter of convenient short names that we don't call those three categories, "primes except units", "units", and "primes". Or for that matter "foos", "bars" and "bazes". "1 is not prime" is simply the assertion, "it's more convenient to define this handy short name to exclude units than includes them" – Steve Jessop Dec 08 '14 at 11:03
  • I see, so basically it's just a matter of semantics. – k_g Dec 08 '14 at 21:15
  • @k_g : Semantics yes. But useful semantics. Units are the numbers that have multiplicative inverses. That alone makes they worthy of their own classification. And, as Steve Jessop pointed out, they mess up unique factorization if you call them prime. Like $6 = 2 \cdot 3 = 1 \cdot 2 \cdot 3 = (-1)^2 \cdot 2 \cdot 3$. It gets even worse when you consider complex numbers where $1 = (-1)^2 = i(-i)$. – Steven Alexis Gregory May 30 '21 at 17:05
12

$$N_4-1=p_0p_1p_2p_3p_4-1=1\cdot2\cdot3\cdot5\cdot7-1=210-1=209=11\cdot19$$ $$N_6+1=p_0p_1p_2p_3p_4p_5p_6+1=1\cdot2\cdot3\cdot5\cdot7\cdot11\cdot13+1=30030+1=30031=59\cdot509$$

bof
  • 82,298
4

$N_7+1 = 510511 = 19 ⋅ 97 ⋅ 277$, $N_7-1 = 510509 = 61 ⋅ 8369$ is the first example where both numbers are not primes. I would suggest that not only is $(N_k-1, N_k+1)$ not always a twin prime pair, but that this would actually be quite rare.

According to http://primes.utm.edu/top20/page.php?id=5 $N_k+1$, $N_k-1$ have been tested for k ≤ 100,000 with very few primes found, and with no twin primes found beyond the pair (2309, 2311).

gnasher729
  • 10,611
  • 20
  • 38