Goldbach's comet: for every even number up to 100,000 a light dot marks the number of ways it is a sum of two primes, so that the dots form bands fanning out from the lower left on a dark ground
Unit 08 · All masterclasses · 75 min

The Great Problems of Number Theory

Goldbach, twin primes, the Riemann hypothesis, and Fermat's Last Theorem: what is known and what is open
Image: Own rendering

The closing unit of the Foundation masterclass presents the statements that nobody has proved and the one that was proved in 1995 after 358 years. It explains why a verification to four billion billion cases settles nothing, states Goldbach's conjecture, the twin prime conjecture, the prime number theorem, and the Riemann hypothesis in a form that needs no complex numbers, proves Euler's product formula and the divergence of the sum of the reciprocals of the primes from unique factorization, and tells the story of Fermat's Last Theorem from the margin note to Wiles. It ends with Landau's problems, the Collatz problem, number theory in daily use, and the roads that lead on from here.

Preview: the beginning of the unit

The seven units before this one proved everything they stated. This unit is about the statements that nobody has proved. Some of them were asked in letters of the eighteenth century and can be explained to a child; one was settled in 1995 after 358 years, with tools that did not exist for most of that time; one carries a prize of a million dollars and is the most famous open problem in mathematics. The unit says what each problem asks, how far computers have checked it, what has been proved around it, and why checking never settles a question about all integers. It also proves two results of Euler, the product formula for the zeta function and the divergence of the sum of the reciprocals of the primes, and it ends with a map of where the subject goes from here.

1 Conjectures and theorems

Unit 1 drew the line between a conjecture, a statement someone believes, and a theorem, a statement someone has proved, and it gave the warning example \(n^{2} + n + 41\), which is prime for \(n = 0, 1, \ldots, 39\) and composite for \(n = 40\). The great problems of this unit sit on the conjecture side of that line, and most of them have been tested far beyond anything a hand computation could reach. Goldbach’s conjecture has been checked for every even number up to \(4 \times 10^{18}\), which is four billion billion. A reader meeting that figure for the first time may feel that the matter is settled, and the first job of the unit is to say why it is not.

A statement about all even numbers is a statement about infinitely many cases, of which a check covers finitely many. That is the logical point, and on its own it may sound like a mere formality, so it helps to see that patterns in the integers really do break late. In 1919 George Pólya (1887-1985) conjectured that, for every \(x \ge 2\), at least half of the integers from \(1\) to \(x\) have an odd number of prime factors counted with multiplicity (so that \(12 = 2 \cdot 2 \cdot 3\) counts three), which is what the first few hundred cases suggest. The conjecture was disproved in 1958, and the smallest counterexample, found in 1980, is \(x = 906{,}150{,}257\): the pattern holds for nine hundred million consecutive cases and then fails. Section 5 meets a pattern that has been checked for every \(x\) up to \(10^{19}\) and is nevertheless known to fail somewhere below \(1.4 \times 10^{316}\). So a verification to \(4 \times 10^{18}\) is evidence of the kind that makes mathematicians confident, and it is not a proof. What it does prove is the finite statement itself, and several proofs in this unit use such a computation as one half of an argument whose other half covers all large numbers.

The full unit is part of the program

Want to see a complete unit first? Unit 3, Prime Numbers and Unique Factorization, is free to read.