08 — Deep Time

Two Billion Years

the same number, read back out of rock and out of starlight

An atomic clock can watch α for as long as someone keeps it running — years, a few decades. To ask whether the number was different long before anyone was measuring it, physicists needed a record older than any instrument. They found two: one buried in the ground in Gabon, one arriving from the far side of the universe.

α_then = α_now  —  to within about one part in ten million

In 1972, a routine check at a French uranium-enrichment plant turned up ore from the Oklo mine in Gabon carrying less uranium-235 than uranium is supposed to carry anywhere on Earth. The missing atoms had already been split. Nearly two billion years ago, groundwater seeping through an unusually rich seam slowed neutrons enough to sustain a chain reaction; the heat boiled the water away, the reaction stopped, the rock cooled, the water returned. It ran like that, off and on, for a few hundred thousand years, at sixteen separate sites. A nuclear reactor, with no one there to build it.

In 1976, Alexander Shlyakhter, a physicist in Leningrad, saw that the reactor's leftovers were also a measurement. Samarium-149 swallows slow neutrons through a resonance at an extraordinarily low energy — so finely balanced that even a slight change in α would have moved it, and changed how much samarium the reactor burned. The samarium still in the rock matches what today's physics predicts. Revisiting the data in 1996, Thibault Damour and Freeman Dyson put it at 95% confidence: whatever α was when Oklo ran, it was the same as now to about one part in ten million.

The second record is older. Light from distant quasars passes through gas in galaxies along the way, and the atoms in that gas absorb it at wavelengths set by α. In 2011, a team led by John Webb combined spectra from telescopes in Hawaii and Chile and reported something no one expected: α seemed slightly smaller in one direction on the sky and slightly larger in the opposite one — a dipole, at 4.2 standard deviations. If real, the constant would not be constant across space.

Then the instruments themselves were examined. In 2015, Jonathan Whitmore and Michael Murphy — Murphy had been one of the 2011 paper's authors — showed that the spectrographs' wavelength scales were distorted in ways that could mimic much of the signal. In 2022, Murphy's team pointed ESPRESSO, a spectrograph calibrated against a laser frequency comb, at a single bright quasar, through a galaxy whose light left it more than eight billion years ago, and found no change in α to about one part per million. In 2026, radio observations with ALMA of molecules toward two lensed quasars set limits tight enough that their authors count them among the evidence ruling the dipole out.

None of this proves α has never changed — only that everywhere it has been checked, in a reactor in African rock and in light that set out before the Sun existed, it has not changed enough to see. A constant that holds still is not a constant explained. The record says the number has been the same for a very long time. It says nothing about why it is this number. What it does show, along the way, is a field working as it should: a startling result published, and one of its own authors building the tools that took it apart.