Remann Hypothesis
A mathematical field guide

The Riemann Hypothesis

Prime numbers have uneven gaps. The Riemann hypothesis links their pattern to a simple picture. Certain points on a graph all lie on one straight line.

A green wireframe surface shows the magnitude of zeta rising into ridges and falling into valleys over the complex plane.
Zeta landscape
|ζ(s)|Magnitude of the zeta function

Zeta landscape

A green wireframe surface shows the magnitude of zeta rising into ridges and falling into valleys over the complex plane.

Each point on the floor is a complex number s. The surface rises with the size of ζ(s). A zero is a place where the height falls to zero.

−0.25 ≤ Re(s) ≤ 1.75 · 11 ≤ Im(s) ≤ 32
Proof update Lean and Mathlib confirm the Claude-assisted proof. Read the update

What it is

A prime is a whole number above 1 that only 1 and itself divide evenly. The first few are 2, 3, 5, 7, and 11. Their gaps change.

In 1859, Riemann studied primes using a function called zeta. A zero is an input that makes the function give 0.

Set aside the zeros at −2, −4, −6, and so on. Riemann’s claim is that every other zero sits on the vertical line halfway between 0 and 1.

Why it matters

Count the primes up to 100 and you find 25. A formula estimates about 30. For this small example, we can see that the estimate misses by about 5.

Now imagine counting up to a number with a hundred digits. Listing the primes one by one would take too long. An estimate is much easier to get.

The line puts a limit on how far the estimate can miss. It tells us how much to trust a prime count even for numbers that huge.

Exact count25
Estimate≈ 30
x = 100

The essentials

Three ideas connect the primes to a single vertical line.

Prime numbers

A prime is a whole number greater than 1 with exactly two positive divisors. The first five are 2, 3, 5, 7, and 11.

Read moreShow less · Prime numbers

Every whole number above 1 has a unique prime factorization, apart from order. For example, 84 = 2 × 2 × 3 × 7.

Primes thin out as numbers grow. There are 25 up to 100 and 168 up to 1,000. The gaps between them vary.

Source · NIST

The zeta function

Zeta connects addition with multiplication. A sum over all positive whole numbers equals a product with one factor for each prime.

Read moreShow less · The zeta function

At s = 2, the sum begins 1 + 1/4 + 1/9 + 1/16. The prime factors in the product generate every term in that sum exactly once.

The sum and product converge when the real part of s exceeds 1. Mathematicians extend zeta beyond that region to study its zeros.

Source · NIST

The critical line

A zero is an input where zeta gives 0. The Riemann hypothesis says every nontrivial zero has real part exactly one half.

Read moreShow less · The critical line

A complex number has two coordinates. The real part runs across the page. The imaginary part runs upward. Real part ½ marks a vertical line.

Zeta has trivial zeros at −2, −4, −6, and so on. Its first nontrivial zero above the real axis lies near ½ + 14.134725i.

Source · Clay

The visual lab

Change a number and watch the mathematics unfold.

Count primes

Move the slider to count primes up to x. Each prime lifts the green staircase by one. The dashed curve gives a smooth estimate.

NIST Digital Library of Mathematical Functions
Prime count25
Estimate li(x)30.13
Count − estimate-5.13
010203040 0255075100 x
Exact countEstimate

Locate zeros

Select one of the first ten zeros above the real axis. Its height changes, while its horizontal coordinate stays at one half.

Andrew Odlyzko’s zero tables
Zero number1
Real part0.5
Height14.134725
01020304050 -0.500.511.5 ½ + 14.13i Im(s)Re(s)
Zeta zerosSelected zero

Primes in everyday life

Primes help your phone check a software update. The hypothesis explains how plentiful those primes are.

Counting primes with confidence

The function li(x) estimates how many primes lie below x. The zeros of zeta control the size of the error in that estimate.

The critical line bounds the error by a constant times √x log(x) for large x. This gives a guarantee even far beyond the reach of a direct count.

Primes in digital signatures

When an Android phone installs a system update, it checks the publisher’s signature. RSA is one way to create that signature.

RSA uses large prime factors to create keys. A digital signature made with a private key lets others check who signed a message and whether it changed.

A prime-count estimate says how plentiful primes are. Recovering the specific secret factors in an RSA key is a different problem.

Primes in repeating sequences

Consider numbers ending in 1, such as 11, 31, and 41. Dirichlet L-functions describe how primes spread among such repeating sequences.

The generalized hypothesis gives sharper error bounds for counting primes in these sequences.

Proof watch

A dated record of the proof, the people, and the independent checks.

News checked
  1. Independent checks match five theorems

    Andrew Sutherland’s MIT record reports that two independent tools matched all five formal statements to their targets.

    The checks include the generalized Riemann hypothesis for Dirichlet L-functions. These functions help count primes in arithmetic progressions.

    MIT verification record
  2. Lean and Mathlib certify the proof

    The two communities confirmed the classical Riemann hypothesis. Independent checkers accepted the proof and verified that it proves Mathlib’s statement.

    Joint verification statement
  3. Levent Alpöge releases the Lean proof

    Alpöge, a mathematician at Anthropic, developed the argument with Claude. He released about 17 million lines of Lean and a four-page announcement.

    Alpöge describes guiding the strategy and credits Claude with crucial late ideas. He plans to explain and simplify the argument.

    Alpöge’s announcement (PDF)
  4. A separate draft by Alessandro Masullo

    Ravi Vakil shared Alessandro Maria Masullo’s human-written draft on the generalized hypothesis. Brian Conrad had given feedback for months.

    Vakil later reported that readers found problems in the draft. Their resolution remains open.

    Vakil’s announcement and update
  5. An earlier result from Anthropic

    Anthropic reported a 67.2% lower bound for the share of zeros on the line. Levent Alpöge and Ralph Furman examined Claude’s argument.

    Anthropic research report

How Lean checks a proof

Lean is a programming language and a proof assistant. It lets researchers express both a mathematical statement and the reasoning that establishes it.

A first example

example : (2 : Nat) + 3 = 5 := by
  rfl

This line asks Lean to prove that 2 + 3 equals 5. The word rfl works because Lean can calculate both sides and see that they agree.

Explore the Lean language

Checking the Riemann proof

Its kernel checks each inference against a small set of rules. Mathlib supplies a shared library of definitions, including the Riemann hypothesis.

In the published project, this command lists the axioms on which the final theorem depends.

import Main

#print axioms riemannHypothesis
MIT verification record