Close Encounters
Walking home, past where the streetlamp gives up, the dark above her clarifies into pieces.
Most of the sky moves with the year. A few things do not.
She has known their names since she was twelve — small, dense, faraway, filed in the textbooks under old, which is the word we use for things we have stopped asking after.
Tonight she counts them, and they count back. The page she has been holding upside down turns in her hand.
The disk we live in is the weather. What hangs above it, what hangs below, was standing before there was a here to stand in.
And we, who thought we were the ones arriving, are the long swing of a pendulum through a room where the furniture has not moved in ten billion years.
The Furniture of the Galaxy
I want to tell you about a sky you already know.
If you have ever looked up on a dark summer night and seen the Milky Way stretched in a soft band across the sky, you have seen the disk we live inside. The disk is where most of what we recognize as the galaxy lives — the spiral arms, the new stars, the molecular clouds, the planets, us. It is the part of the galaxy that moves.
This is going to be about the part that doesn't.
What we forgot to look at
There are roughly 150 globular clusters orbiting the Milky Way. Each is a tight ball of a hundred thousand to a million stars, a few dozen light-years across. They are old — radiometrically dated to between twelve and thirteen billion years, which makes them older than the galactic disk itself. The disk formed around them. They are the senior structures of the galaxy.
They are not in the disk. They orbit the galactic center on highly inclined, often nearly perpendicular paths, plunging through the plane and back out into the halo on cycles of hundreds of millions of years. From our position inside the disk, they appear as small luminous beads scattered above and below us, strung on invisible threads.
Charles Messier catalogued the brightest of them in 1771. Omega Centauri and 47 Tucanae are visible to the naked eye. Every amateur astronomer has put an eye to a telescope and seen one resolve into countable stars. We have measured every one of their distances and motions to remarkable precision with the Gaia satellite. They are, in the strictest sense, the most catalogued and least studied objects in the sky.
Why we stopped asking
When the search for life elsewhere became a respectable line of inquiry in the late twentieth century, globular clusters were dismissed as candidates almost reflexively. Two arguments did the work.
First, they are metal-poor. The stars in globular clusters belong to Population II — the second generation in the history of the universe, formed before enough supernovae had cooked the heavy elements that rocky planets and life require. Without metals, the argument went, no rocky planets; without rocky planets, no biology.
Second, they are dense. The inner regions of a globular cluster can be a hundred thousand times more crowded with stars than our solar neighborhood. Stable planetary orbits, in such a crowd, were assumed to be impossible. Passing stars would disrupt them within tens of millions of years.
Both arguments are now being quietly revised.
Planets have been confirmed in globular clusters. The first, PSR B1620-26 b, was found in M4 in 2003. More candidates have followed. The metal-poor argument turns out to overstate the case: while metal-rich stars host planets at higher rates, planet formation can proceed at metallicities well below the solar value, and the outer regions of globular clusters — where stellar densities are much lower — appear capable of supporting planetary systems for billions of years. Recent dynamical studies suggest that the "no planets in globulars" rule applies mostly to the dense cores, not to the bulk of the cluster.
In other words: the case against globular clusters was built on assumptions that have weakened. And while we were debating those assumptions, a separate fact about them quietly sat unexamined.
They are exactly where they need to be
Our Sun orbits the galactic center once every 225 million years or so — a galactic year. Along that orbit, it does not move smoothly through a flat plane. It bobs vertically, oscillating up through the disk and back down again, with a period of about thirty to thirty-five million years. Over one galactic year, the Sun crosses the disk plane somewhere between six and eight times.
Each of those crossings brings the Sun closer to the orbital paths of globular clusters, which are themselves passing through the disk on their own plunging orbits. The geometry is not a coincidence in any spooky sense — it is just what happens when two populations of objects share a galactic center but live on different orbital surfaces. Sometimes the surfaces nearly touch.
If you reconstruct the Sun's orbit backward in time using modern data — which is now possible thanks to the European Space Agency's Gaia mission, which has measured the positions and motions of nearly two billion stars — you can ask, for each globular cluster: when was the closest approach? How close did the Sun and the cluster come, and how long ago?
A handful of places to look
It helps to put names to the question. Of the hundred-and-fifty globular clusters known, a few have stories worth telling whatever turns out to be true about them.
Omega Centauri. The largest globular cluster in the Galaxy — three and a half million stars, visible from anywhere south of about thirty degrees north latitude as a fuzzy "star" in the constellation Centaurus. It is almost certainly not a true globular cluster at all. Its multiple stellar populations, its unusual mass, its retrograde orbit, all point to it being the surviving core of a dwarf galaxy that the Milky Way ate some billions of years ago. Whatever passed near us when it passed near us was carrying the chemistry of an entire small galaxy.
M13, the Hercules Cluster. Twenty-five thousand light-years away, visible in binoculars between the keystone stars of Hercules. Edmund Halley first noticed it in 1714. In 1974, when humanity decided to send its first deliberate radio message to the stars, the Arecibo telescope was pointed at M13. The signal is still on its way; it will arrive in about twenty-three thousand more years. (ed: cleanup needed: Long before either thing happens, the cluster itself will have passed near us many times.)
M4. The closest globular cluster to Earth, only seven thousand light-years away. The first planet ever confirmed in a globular cluster was found here in 2003 — a Jupiter-sized world orbiting a pulsar and a white dwarf, the survivor of more cosmic violence than any planet we knew of before. M4 was the cluster that proved the dismissal wrong.
M92. Possibly the oldest object in the Milky Way. Stellar evolution models put its age at at least 13.2 billion years — formed when the universe was less than a tenth of its current age. If life requires time, M92 has had all the time there is.
47 Tucanae. The other great naked-eye southern globular. A million stars, twelve billion years old, only thirteen thousand light-years from us, second-brightest globular in the sky. It is one of the most metal-rich clusters known, and one of the most-studied — almost every planet-hunting survey ever pointed at a globular has been pointed at 47 Tuc.
These are specific places. They have addresses in the published catalogues. Their orbits are known. Their masses are known. If the hypothesis means anything, it means that one or more of these — or one of the hundred-and-forty-five others — was somewhere specific, at some specific time, when something specific happened on Earth.
The arithmetic of a tidal kick
Here is the only equation in this essay. It comes from a 1981 paper by Jack Hills.
When a massive body passes the Sun at impact parameter b with relative velocity v, the gravitational kick it imparts to a comet at distance a from the Sun — sitting at the edge of our Oort cloud, say — is approximately:
ΔV ≈ 2 G M a / (b² v)
That is the differential velocity change: how much faster or slower the comet is moving, relative to the Sun, after the encounter than before.
For numbers: Omega Centauri, at three and a half million solar masses, passing us at one kiloparsec at two hundred kilometers per second, with our outer Oort cloud at fifty thousand astronomical units, gives a kick of about one meter per second. Comets at the Oort cloud orbit the Sun at roughly a hundred meters per second. A one-percent perturbation to a fifty-thousand-AU orbit, applied over a few million years of slow encounter, sends some fraction of those comets inward. Inward, they pass the planets. Some are captured into the inner solar system. Some are evaporated by the Sun. Some hit the Earth.
This is not speculation. Every comet that has ever been examined contains organic chemistry. The Rosetta mission found glycine — an amino acid — at comet 67P. The Murchison meteorite contains dozens of amino acids in non-terrestrial ratios. Earth gains tons of cometary and interstellar dust every single day. We are not asking whether comets carry organics. We are asking whether the rate at which they fall is sometimes higher than usual, and what determines that rate.
You do not need to imagine that anyone is sending anything. You only need to imagine that ancient stellar structures have been quietly producing and shedding the chemistry of life for ten billion years, and that the Sun, on its long pendulum swing through the galaxy, periodically passes near enough for the tidal field of one to perturb the cometary halo of the other. When the flux rises, more organic material falls. When more organic material falls, biospheres respond.
How the question gets asked
In 2026, the question can be asked from a browser.
The tool that does it integrates the Sun's orbit and the orbits of the catalogued globular clusters backward in time inside a model of the Milky Way's gravitational potential. For each cluster, it finds the moment of closest approach over the last few hundred million years and computes the four numbers that matter: how close they came (in kiloparsecs), when it happened (in megayears ago), how fast they were moving relative to each other (in kilometers per second), and the tidal kick to a comet at the Oort cloud (in meters per second).
It then matches each closest approach against a list of events on Earth. The end-Permian extinction. The Cretaceous-Paleogene impact. The Paleocene-Eocene thermal maximum. The Carnian Pluvial Event when the dinosaurs first rose. The list is editable; you can put in whatever you want to test against.
The output is a sorted list. At the top: clusters that came closest, were most massive, and whose nearest approach falls within a few million years of a known event. At the bottom: clusters that passed far away long ago and have nothing to say.
The list is unlikely to be definitive on first reading. Backward orbital integration over two hundred million years is sensitive to the model of the galactic gravitational field; switching from one reasonable potential to another shifts cluster trajectories at the five-to-fifteen-percent level. The cluster phase-space measurements have their own uncertainties; a serious analysis requires sampling each cluster's six-dimensional state from its published error ellipsoid a thousand times. The clusters are not point masses. The cluster-to-comet-to-biology causal chain has its own large uncertainties. Even a strong perturbation does not deterministically produce an extinction event fifty million years later.
But the list will surface candidates. Specific cluster names, specific times, specific encounter parameters. That is what you want, before doing the harder work.
What is testable
Several things, in increasing order of difficulty:
One. The closest-approach time and distance for each known globular cluster, propagated back to the limit where the integration is still meaningful. It takes about a hundred milliseconds for the full catalogue.
Two. The robustness of those approach times to the choice of galactic potential. Re-run with three or four different published Milky Way models and look for clusters that agree across all of them. Agreement between independent models is far stronger evidence than any single result.
Three. A proper Monte Carlo over input uncertainties. Sample each cluster's phase space from its published Gaussian errors a thousand times. Report not single approach times but distributions. A candidate whose approach time is robust against both potential variation and input uncertainty is a candidate worth chasing.
Four. A correlation analysis against the geological record as a whole. Not just "does the K-Pg match," but: across the full record of biological pulses, is the rate of cometary infall (proxied by tidal kick) anomalously high during pulse intervals and anomalously low between? A signal in that direction would be hard to fake.
Five. Direct geochemistry. If the hypothesis is true, the cometary infall should leave isotopic signatures in sediments from the affected periods — anomalous abundances of certain noble gases, iridium and other siderophile spikes (already used for K-Pg), specific organic ratios. Some of this data already exists. Some of it doesn't.
None of this requires a new instrument. All of it requires somebody to do the work.
The old things have not stopped happening
We catalogued the globular clusters before we catalogued most of what we now know about the galaxy. Messier put them on a list because they got in the way when he was looking for comets. They have been there the whole time, the same names in the same atlases, with the same orbits, swinging through the disk like the slow pendulums of a deep clock.
If the correlations are real, the work of finding them will be unspectacular. A script. A list. A sort. The names of the structures responsible will already be known. They will have Messier numbers and NGC numbers. They will be in every observatory atlas printed in the last two hundred years.
We have stopped asking after old things.
The old things have not stopped happening.
The tool that runs the integration is at catpea.github.io/cluster-encounters. It runs in your browser. It takes less time to use than to read this essay.