Tea at the IAS
Museum cafe
Tea served 15:00
Institute for Advanced Study, Princeton
Nobody took a photo at tea. This is the formula they talked about.
- Place
- Institute for Advanced Study, teatime
- Date
- Early April 1972
- People
- Hugh Montgomery, Freeman Dyson
- Object
- 1 − (sin πu / πu)²
Table 7 · reserved
Montgomery & Dyson
In early April 1972 Hugh Montgomery stopped by the Institute to show Atle Selberg a new result about how the zeros of zeta space out. The zeros push each other apart.
At tea, the number theorist Daman Chowla wanted to introduce him to Freeman Dyson. Montgomery said he didn’t need to meet Dyson, and it ended with Chowla “dragging me across the room.”
Montgomery mentioned his formula. Dyson said something like: that’s the pair correlation of eigenvalues of random matrices in the Gaussian Unitary Ensemble. Physicists use those matrices for the energy levels of heavy nuclei.
“I’d never seen a random matrix,” Montgomery said later. “I’ve hardly seen one since.” They had one conversation in their lives, about five minutes long. I love that this whole room comes out of five minutes at tea.
Dyson wrote to Selberg afterwards with the references. He had found the same formula a decade earlier, coming from physics. “They were coming from completely different directions and you get the same answer,” he said. “At the moment, it is just a miracle.”
In 1989 Andrew Odlyzko computed eight million zeros near zero number 1020 and found the same match. Nobody knows why it holds. The same spacing shows up in the buses of Cuernavaca.
There's more in the Run Data Run Deep Dive.
Zeros get closer together as you go up the line, so each gap is rescaled to make the average gap 1. Bars are the zeros. Yellow is the random-matrix curve. For gaps that’s the Wigner surmise, a close stand-in for the exact curve Michel Gaudin worked out in 1961. For all pairs it’s Montgomery’s formula. The dashed curve and pale bars are random points with the same average gap. Zeros almost never sit close together. Random points often do.