The Twin Prime Conjecture

59 and 61 are both prime. So are 101 and 103. So is a pair with hundreds of thousands of digits. Average gaps between primes grow. These two-step neighbors do not get the memo.

A prime is a whole number bigger than 1 that you cannot factor into smaller whole numbers. Two, three, five, seven, eleven. Euclid already knew there are infinitely many of them: if you had only a finite list, multiply them all, add one, and you force a new prime into existence. That argument says primes go on forever. Whether they keep showing up as roommates is a separate claim.

The roommate pattern is stubborn. Sometimes two primes differ by exactly two: 3 and 5, 5 and 7, 11 and 13, 17 and 19, 29 and 31. Pairs like that are called twin primes. As numbers grow, primes get rarer on average. The lonely-prime intuition says close pairs should eventually stop. They just get harder to find.

In 1849 Alphonse de Polignac asked a wider question: does every even gap happen infinitely often? A gap of two is the most stubborn special case. That special case is the twin prime conjecture: there are infinitely many primes p such that p + 2 is also prime.

The twentieth century made the twins look both real and sparse. In 1919 Viggo Brun proved that the sum of the reciprocals of twin primes converges. Ordinary primes are numerous enough that their reciprocal sum diverges, like the harmonic series. Twins are thinner than that. G. H. Hardy and John Littlewood later guessed a precise density: twins should keep appearing, just more and more slowly. Huge pairs have been found by machine. A proof that they never stop did not arrive.

In 2013 Yitang Zhang proved that there is some even number at most 70 million that appears infinitely often as a gap between primes. For the first time, consecutive primes were known to huddle inside a fixed window, over and over, forever. A collaborative “Polymath” project, and independent work by James Maynard, drove that window down. The best unconditional bound now is 246: infinitely often, two primes differ by at most 246.

That is astonishing progress. The conjecture still asks for a gap of 2, and the proved window is 246. Sieve methods, the main tools for counting primes in patterns, run into a parity obstruction: they struggle to tell numbers with an even number of prime factors from numbers with an odd number. Until that wall moves, or a different idea appears, twins remain a conjecture with a shrinking neighborhood around them.