Odlyzko's tables
Computers have checked about 12.4 trillion zeros, and every one is on the line. That’s why nearly everybody believed the hypothesis. It never counted as a proof: there are infinitely many zeros, and 12.4 trillion of them is a rounding error.
- Maker
- Andrew Odlyzko
- Contents
- The first 100 zeros to over 1,000 digits, the first 100,000, the first 2,001,052, and blocks near the 1012th, 1021st and 1022nd zero
- Held
- Online, University of Minnesota
Odlyzko didn’t climb from the ground. He jumped far up the line and checked blocks of zeros up there, which is how the spacing in the tea room got tested at heights near 10²⁰.
Everyone else started at the bottom. Scroll down to climb with them.
There's more in the Run Data Run Deep Dive.
The climb, 1859 to 2021
- c. 18593Bernhard Riemann, by handNever published. Siegel found the working in his papers in 1932.
- 190315J. P. Gram, by hand, Euler-Maclaurin summationNoticed there's usually one zero between each pair of Gram points. That's Gram's law.
- 191479R. J. Backlund, by handFound a better way to check that every zero up to a height is on the line.
- 1925138J. I. Hutchinson, by handFound the first place Gram's law fails, at Gram point 126. The zeros stayed on the line anyway.
- 1935195E. C. Titchmarsh, the Riemann-Siegel formulaUsed Riemann's own method, rediscovered in his papers. Much faster than the old way.
- 19361,041E. C. Titchmarsh and L. J. Comrie, by handThe last count done by hand.
- 19531,104Alan Turing, the Manchester Mark 1 (run in 1950)The first digital computer on the job, and the counting check everyone still uses.
- 195625,000D. H. Lehmer, a computerFound two zeros at 7005.063 and 7005.101, so close that Z only just changes sign between them.
- 1966250,000R. S. Lehman, a computerA quarter of a million.
- 19683,500,000Rosser, Yohe and Schoenfeld, a computerStated Rosser's rule, a patch for the places Gram's law breaks.
- 197740,000,000R. P. Brent, a computerForty million.
- 197981,000,001R. P. Brent, a computerCounts this big get quoted to the last zero.
- 1982200,000,001Brent, van de Lune, te Riele and Winter, a computerTwo hundred million and one.
- 1983300,000,001van de Lune and te Riele, a computerThree hundred million and one.
- 19861,500,000,001van de Lune, te Riele and Winter, a computerOne and a half billion, with graphs of Z wherever it behaved oddly.
- 200110 billionJ. van de Lune, a computer, unpublishedTen billion.
- 2004about 900 billionSebastian Wedeniwski, ZetaGrid, volunteers' computersSpread across volunteers' spare machines.
- 200410 trillionXavier Gourdon and Patrick Demichel, the Odlyzko-Schönhage algorithmAlso checked two billion zeros around heights 10¹³, 10¹⁴ and on up to 10²⁴.
- 202112.4 trillionDave Platt and Tim Trudgian, a rigorous computer runEvery zero up to height 3 × 10¹², about 1.2363 × 10¹³ of them. The highest solid rung.
All of them
Claimed, still being checkedLevent Alpöge, working with Claude, posted a four-page note claiming a proof of the hypothesis. It’s backed by a computer proof of about 17.4 million lines. Checking more zeros never gets you here. A proof covers every zero at once.
As of 21 November, two separately written checking programs accept the whole proof. The last checks are still running.
Where the solid rungs stop
Platt and Trudgian checked every zero up to height 3 × 10¹². Here’s Z(t) just above that. Press the button and you’ll find zeros on the line up there too. It can’t tell you about the next one.
Checking isn’t proving
Peter Sarnak, writing for the Clay Institute in 2004, passed on Odlyzko’s warning: the first 10¹³ zeros sitting on the line “should not by itself be taken as very convincing evidence for RH”. The strange behaviour that would give away a zero off the line is expected to show up only at heights far beyond anything a computer can reach.