Fifty years on, nobody can say why prime numbers and nuclei share a law
Most of this module has been about things that can be measured. Fill a matrix with noise and its eigenvalues form a semicircle: measurable. They repel with an exponent set by symmetry: measurable. The largest one fluctuates according to a skewed universal law: measurable. The zeros of the zeta function share the spacing statistics of random Hermitian eigenvalues: measurable, and measured to a precision that leaves no room for doubt about the fact.
None of that amounts to an explanation. There is no argument anywhere that starts from the distribution of prime numbers and arrives at the spectral statistics of random matrices. Fifty years after Dyson recognised Montgomery’s formula, the connection is exactly as well established and exactly as unexplained as it was that afternoon.
There is a shape a solution might take, and it has been sitting there since long before anyone noticed the statistics. It is usually called the Hilbert–Pólya idea, and it is less a conjecture than a wish with a precise form: find the right operator and the Riemann hypothesis falls out as a triviality.
If some natural self-adjoint operator had the imaginary parts of the zeta zeros as its spectrum, the Riemann hypothesis would follow immediately — the zeros would be real because eigenvalues of such an operator have to be. This is the Hilbert–Pólya idea. Nobody has produced the operator.
Two of these are theorems, one is the most famous open problem there is, and the link between them is a missing object nobody has been able to construct. The statistics in this module are the fourth line: the strongest evidence that the object exists, and no help at all in finding it.
Try it: Note which lines are marked established. The two theorems are the easy ones; the chain breaks precisely where an object would have to be produced. The statistics measured in this module are the evidence that it exists, and no guide at all to where.
There is a reason the agreement feels like a clue rather than a fluke. Random matrix statistics are not merely one distribution among many — they appear to be the signature of a particular kind of dynamics. Quantum systems whose classical behaviour is chaotic show them; systems whose classical behaviour is regular show Poisson statistics instead. The spacing distribution acts as a detector for what lies underneath.
Try it: Switch between the two worlds. The regular spectrum piles up at zero spacing — levels coincide freely. The chaotic one pushes them apart. The zeta zeros sit unambiguously on the second side, which is a strange thing to learn about a question in arithmetic.
Montgomery derived the pair correlation of the zeros under an assumption about prime pairs, and it came out matching the random matrix answer. But a derivation of the pair correlation is not a reason. There is no argument saying "the zeros must behave this way because the primes do that", and finding one would be a far deeper result than the numerical agreement.
It is worth laying out what this module has actually established, and under what standard. A theorem, a numerical agreement and a conjecture with heavy empirical support are three different things, and the interesting feature of this subject is how much of its most famous content sits in the weakest category.
Four of these are theorems. The two that matter most for the story of this module are in the bottom category, and one of them is not even a precise conjecture — it is a request for an explanation that nobody has been able to phrase, let alone answer.
The through-line of this module is that randomness is not the absence of structure. Fill a matrix with numbers chosen carelessly and you do not get chaos; you get a semicircle, a precise repulsion law, and a skewed distribution at the edge — none of which depend on what you put in, only on the symmetry of how you put it in. That is what universality means, and it is why a random matrix can stand in for a nucleus nobody can solve.
The same indifference to detail is what makes the zeta result so hard to dismiss and so hard to explain. If the spacing law only cared about symmetry, then something about the zeros has the symmetry of a chaotic quantum system without time-reversal invariance — and the prime numbers, of all things, are what produced it. There is a fair chance that the eventual explanation, whenever it arrives, will make the question look obvious in hindsight. For now the honest summary is that two subjects agree to many decimal places and nobody can say why. If you want the arithmetic side of it, the zeta function is where the zeros come from, and the comparison itself is worth running again once you know what you are looking at.