Look at the primes up close and they seem to be scattered almost at random: 2, 3, 5, 7, 11, 13, 17, 19, 23, 29 — no formula spits them out, no obvious rhythm tells you where the next one will land. The primes are infinite, but up close their appearance looks chaotic.

Now zoom out. Instead of asking "where exactly is the next prime?", ask "roughly how many primes are there below a million? A billion? A trillion?" Something remarkable happens: the apparent chaos resolves into a smooth, precise, predictable rate at which primes thin out. Individual primes are unpredictable; their overall density obeys a law good enough to belong in a physics textbook. That law is one of the crown jewels of number theory.

Worked example: counting primes below 10 and below 100

Write \pi(x) for the number of primes up to and including x. Let's count by hand and watch what happens to the density — the fraction of numbers up to x that are prime.

Below 10: the primes are 2, 3, 5, 7, so \pi(10) = 4. That's 4 primes out of 10 numbers — a hefty 40% are prime.

Below 100, a careful sieve (or a look-up) gives \pi(100) = 25 primes. That's only 25% — the density has nearly halved. Push on to 1000: there are \pi(1000) = 168 primes, a density of just 16.8%.

The trend is unmistakable: as the numbers grow, primes become steadily rarer. The question this page answers is how fast — and the answer turns out to be exact enough to predict \pi(x) for numbers you could never hope to sieve by hand.

Counting the primes: \pi(x)

\pi(x) is a staircase: it jumps up by one at every prime and is flat in between. It looks jagged and irregular close up. Watch how its rising shape is shadowed almost exactly by a smooth curve as x grows:

The smooth curve: x / \ln x

Gauss, as a teenager poring over tables of primes for fun, noticed that \pi(x) is tracked closely by x / \ln x. Equivalently, the density of primes near a large number x is about 1/\ln x — so a number picked near x has roughly a 1 / \ln x chance of being prime. This is the celebrated Prime Number Theorem, which we'll state and prove properly later — for now, meet its shape:

Proving it took a century after Gauss's guess and required the deep machinery of complex analysis — we will return to it in the analytic stage.

Worked example: testing the estimate

Let's put x/\ln x to the test at x = 1000, where we already counted the true answer by hand.

The true count was \pi(1000) = 168. The estimate is in the right ballpark but noticeably low — off by about 14\%. That's expected: the theorem only promises the ratio \pi(x) / (x/\ln x) creeps towards 1 as x \to \infty, and at x = 1000 we are nowhere near infinity. Try x = 10^9: there \pi(x) = 50\,847\,534 while x/\ln x \approx 48\,254\,942 — a much closer 5\% gap. The bigger the number, the better the smooth law fits.

Gaps: growing on average, but not steadily

If primes near x occur with density about 1/\ln x, then the typical spacing — the gap between one prime and the next — should be about \ln x. Near x = 10, \ln 10 \approx 2.3, and indeed the gaps 3-2, 5-3, 7-5 are 1, 2, 2 — small, matching a small \ln x. Near x = 100, \ln 100 \approx 4.6, and the primes 89, 97, 101, 103, 107, 109, 113 have gaps 8, 4, 2, 4, 2, 4 — bigger on average, exactly as predicted.

Gaps can also grow enormously, on purpose. Take n = 5, so n! = 120. Then 120 + 2 = 122 is divisible by 2, 120 + 3 = 123 is divisible by 3, 120 + 4 = 124 is divisible by 4, and 120 + 5 = 125 is divisible by 5 — four composite numbers in a row, guaranteed, with no primality testing needed. Crank n up and the same trick — using n! + 2, n! + 3, \dots, n! + n — manufactures a prime-free run of any length you like. So gaps really can be made arbitrarily large; it's the typical gap, not the largest possible one, that grows like \ln x.

But look closely at the list of primes near 100: 101 and 103 are only 2 apart, and so are 107 and 109. Pairs of primes exactly 2 apart are called twin primes. Even as the average gap climbs relentlessly with \ln x, tight little gaps of 2 keep turning up, seemingly forever. The law governs the average; it says almost nothing about any one gap.

Order and chaos, side by side

Here is the honest, slightly unsettling truth: mathematicians understand the overall density of primes extremely well — the Prime Number Theorem pins it down with total confidence — yet nobody can predict where the next individual prime will fall, short of testing each candidate one by one. It's a bit like weather and climate: climate scientists can state with confidence how much warmer next July will be on average than this January, while no one can tell you whether it will rain in your town on a specific afternoon eight months from now. Zoomed out, primes obey a precise law. Zoomed in, they still spring surprises — that tension between order and chaos is exactly why number theorists find them endlessly fascinating.