This image needs WebGL2, which this browser does not provide.
AI Mathematics · Number theory
Riemann’s Line, 1859 → 2026 Still · Ten thousand zeros
What you are looking at Every point here is a complex number s, with a real part σ across and an imaginary part t upward. The Riemann zeta function, ζ, turns each point into another complex number. The colour shows which way that number points, like a compass rose: one full turn of the colour wheel is one full turn around zero. The dark lines mark where its size doubles, the faint ones each quarter of a doubling, and within each band the brightness climbs as the size grows.
A zero is a point where ζ(s) = 0. There every colour meets and the rings of doubling shrink onto a single point. At the foot of the image there are two special places: s = 1, where ζ becomes infinite and the colours meet turning the other way, and s = −2, the first of the so-called trivial zeros on the real line.
The vertical band between real parts 0 and 1 is the critical strip. Every other zero lies inside it, and in this image the colours always converge on the line through its middle, real part ½. The image climbs to height t = 9,878.2, just past the 10,000th of them.
Curiosities Twelve places worth a closer look. Choose one to fly there; turn on Curiosities in the controls, or press c , to see their letters on the image.
The Riemann Hypothesis Bernhard Riemann conjectured in 1859 that every zero in the strip lies exactly on that middle line. The positions of these zeros control how irregularly the prime numbers are spread out, so the hypothesis is equivalent to the strongest possible statement about the error in counting primes.
The announcement On 20 November 2026, Levent Alpöge announced that, working with Claude, he had found a proof, extending to Dirichlet L-functions (his note ). He released it as a formalization in the Lean proof assistant, and wrote that a readable account would follow. The same day Sahibzada Allahyar posted an expository reconstruction of the argument.
Status As of 21 November 2026 the proof is unverified, and this page does not check it: an independent build on a public server whose operator is not named reported on 20 November that the whole project compiles with no failures and that its final theorems rest only on Lean’s three standard axioms (a fresh kernel replay was skipped; the audit ), but no named mathematician or peer review has confirmed it. What it shows is classical and independent of the claim. The zeros marked on request are Andrew Odlyzko’s computed values (tables ); the hypothesis has been verified numerically for every zero up to height 3 × 10¹² (Platt and Trudgian, 2021 ). This image does not test it beyond its own height.
How it is computed ζ was computed at every pixel centre in 64-bit floating point by Euler–Maclaurin summation, the same sum the Phase and Landscape views compute live, accurate to about 10⁻⁶ of |ζ| and of a radian in direction over the whole image. The full strip, real parts −2.5 to 3, is 352 by 632,205 pixels at 64 pixels per unit; the lower heights are also stored at 128 and 256 pixels per unit, and the foot, Lehmer’s pair and the top at 512. The tiles hold the size and direction of ζ, not colours, so your browser colours and draws the lines at any zoom; between stored points it interpolates ζ itself, which is smooth, rather than its size, so the rings round a zero stay round. Zoom in more than about three times past the stored sharpness and ζ is computed afresh on your graphics card, by the same sum the Phase view uses, down to 8,192 pixels per unit. Checks: all 10,000 tabulated zeros fall within 0.54 pixel of the darkest point of the computed grid on the middle line, and the same ζ, sampled along that line every 0.01 (as Hardy’s real-valued Z), changes sign exactly 10,000 times below the top. Nothing is drawn that was not computed; the zero marks are an overlay, off by default.