Earth's violent birth

Foundations

Three explanations, two of them wrong

Before Apollo, there were three respectable ideas about the Moon. It was captured, having formed elsewhere and wandered too close. It condensed alongside Earth from the same swirl of material, a sister planet. Or it was flung off a rapidly spinning young Earth, leaving a scar that some suggested was the Pacific basin.

All three had problems, and the rocks brought back between 1969 and 1972 killed two of them outright.

Capture fails because the Earth-Moon system has far too much angular momentum, and because capturing a body that large requires a coincidence bordering on the miraculous. Co-formation fails on composition: the Moon has almost no iron core, about two percent of its mass against Earth’s thirty, and two bodies forming side by side from the same material should not differ that much. Fission fails because the Earth would have had to spin impossibly fast.

What the Apollo rocks did show was stranger than any of the three. The Moon is made of material that is chemically depleted in exactly the things that boil away easily, and isotopically almost identical to Earth’s mantle.

The impact

The explanation now accepted was proposed in 1975 and became convincing when computers grew powerful enough to test it. Near the end of Earth’s assembly, when it had reached about ninety percent of its present mass, it was struck by another growing planet roughly the size of Mars.

The impact was not head-on but glancing, at something like ten kilometres per second. It was, by human standards, unimaginable: enough energy to melt the entire outer Earth and vaporise a substantial part of the impactor. The impactor’s iron core sank and merged with Earth’s. A disc of vaporised and molten rock, drawn mostly from the outer layers of both bodies, was thrown into orbit.

That disc, poor in iron because the iron had already sunk, cooled and gathered itself into the Moon. Simulations suggest it took a matter of decades to a century. The impactor has been given a name, Theia, after the mother of the Moon goddess in Greek myth.

The model explains what the others could not: the angular momentum, the small lunar core, the depletion of volatile elements that were lost as vapour, and the timing.

The isotopic problem

There is one thing it explains less well, and it is worth saying so.

Every group of meteorites has its own isotopic fingerprint, particularly in oxygen, which varies with where in the solar system a body formed. Earth and the Moon have essentially the same fingerprint, in oxygen and in titanium and in several other elements. They are isotopic twins.

The difficulty is that in most simulations, the majority of the material that ends up in orbit comes from the impactor, not from Earth. If Theia formed somewhere else, it should have had a different fingerprint, and the Moon should show it. It does not.

There are ways out. Theia might have formed at almost exactly the same distance from the Sun and so have had the same composition. The impact might have been more violent than the standard model, vaporising and thoroughly mixing both bodies. Or the disc might have exchanged material with Earth for long enough to homogenise. Each proposal fixes the isotopes and strains something else. The giant impact is not in doubt; its details are being argued over now.

An ocean of magma

Whatever the details, the aftermath is clear. Earth’s outer several hundred kilometres were molten. There was no crust, no ocean and no air worth the name, only a glowing surface under a thick atmosphere of rock vapour and steam, radiating heat into space.

That state did not last. A magma ocean loses heat quickly once the rock vapour condenses, and models give it a few million years at most to solidify. As it froze, denser minerals sank and lighter ones floated, which is the beginning of the layering in the next lesson.

The Hadean, revised

The first eon of Earth’s history is called the Hadean, after the Greek underworld, and the name reflects what geologists assumed: a hellish world of magma and bombardment for half a billion years. Almost no rock survives from it, so there was little to argue with.

Then, in the Jack Hills of Western Australia, geologists found grains of zircon. Zircon is nearly indestructible, and when it crystallises it traps uranium and rejects lead, which makes it an excellent clock. Some of these grains are 4.4 billion years old, older than any rock on Earth, eroded out of vanished rocks and recycled into younger sandstone.

Their chemistry is the surprise. The oxygen isotopes in them are consistent with crystals that formed from rock which had interacted with liquid water at low temperature. If that reading holds, then within about 150 million years of its formation the Earth had cooled enough to have liquid water at its surface, and possibly continental crust. Not hellish at all, and not for very long.

The Moon since

The Moon has been moving away ever since. Tides raised on Earth by the Moon are dragged ahead by Earth’s rotation, and the resulting pull nudges the Moon outward while slowing Earth’s spin. Reflectors left on the surface by the Apollo missions let its distance be measured to millimetres: it recedes about 3.8 centimetres a year.

That rate cannot simply be run backwards, because it depends on the shape of the ocean basins, which change. But the direction is certain, and it is confirmed independently by layered tidal sediments that record shorter days in the deep past. The Moon was once far closer, the tides were far larger, and the day was a matter of hours.

One likely consequence matters for everything that follows: a large Moon steadies Earth’s axial tilt. Mars, without one, wobbles chaotically over tens of degrees. A stable tilt means stable seasons over geological time, which is a reasonable thing to have while life is getting established.

Deeper

The angular momentum constraint

The Earth-Moon system’s angular momentum, about 3.5 × 10³⁴ kilogram metres squared per second, is high for a planet-satellite pair and corresponds to an Earth day of about five hours if the Moon’s orbital momentum were returned to Earth’s spin. Capture scenarios cannot supply this without a third body to absorb energy. In the canonical giant-impact simulations, a glancing collision by an impactor of about 0.1 Earth masses at close to the mutual escape velocity delivers the observed angular momentum naturally, which was the model’s original selling point and is also what makes the isotopic problem hard: a low-energy glancing impact leaves the impactor’s material largely unmixed.

Depletion of volatiles

Lunar samples are depleted relative to Earth in elements that condense at low temperature, including potassium, sodium and zinc, and are extremely dry compared with Earth’s mantle, though not anhydrous: Apollo samples reanalysed since 2008 contain measurable water in volcanic glasses and apatite. The depletion pattern follows condensation temperature, consistent with material that passed through a vapour phase and lost what did not recondense. The presence of any water at all constrains how hot and how long-lived the disc was.

Dating the Moon

Lunar formation ages come from several systems that disagree in detail. Hafnium-tungsten chronometry of lunar samples, and lead-lead ages of ferroan anorthosites from the lunar crust, have given ages between about 4.36 and 4.51 billion years. Barboni and colleagues dated zircons from Apollo 14 by uranium-lead and hafnium and obtained 4.51 billion years, implying the impact occurred within about 60 million years of solar system formation. The spread among methods reflects genuine difficulty: different systems close at different times during the magma ocean’s crystallisation, so they may be dating different events.

Magma ocean crystallisation

Elkins-Tanton’s review sets out the sequence. Once the rock-vapour atmosphere condenses, a magma ocean radiates efficiently and can solidify in as little as a few million years, though a thick steam atmosphere slows this considerably. Crystallisation proceeds from the bottom upward as the adiabat intersects the liquidus, with dense minerals settling and buoyant plagioclase floating, producing a chemically stratified mantle and, on the Moon, an anorthositic crust that is the lunar highlands. The resulting layering may be gravitationally unstable and overturn, which has consequences for the mantle’s later composition.

The late veneer

Highly siderophile elements, those that partition strongly into iron, should have been almost entirely stripped from Earth’s mantle during core formation. The mantle contains them at roughly a thousand times the expected level, in approximately chondritic proportions. The standard explanation is a late veneer: about 0.5 percent of Earth’s mass added as chondritic material after core formation ceased. This is where the gold accessible in Earth’s crust comes from, and it links to lesson 3.2: the gold was made in neutron-star mergers, incorporated into asteroids, and delivered here after the core had closed.

Frontier

Making an isotopic twin

Several classes of model attempt to reconcile the isotopes with the dynamics: a fast-spinning proto-Earth struck by a smaller impactor, with the excess angular momentum later removed by a solar resonance; a much more energetic impact producing a supercritical fluid structure in which Earth and disc material fully mix; and a series of smaller impacts rather than one. Each reproduces some observations and struggles with others, and the community has not converged. This is one of the clearest cases in planetary science of a well-established conclusion with unresolved mechanics.

How cool was the early Earth

The Jack Hills zircons are single grains, and inferring a global surface environment from a few thousand crystals requires care. Alternative readings attribute their oxygen isotope signatures to processes other than surface water, and the case for continental crust at 4.4 billion years is contested. Atom-probe tomography has confirmed that the ages are real and not artefacts of lead mobility, which strengthens the chronology without settling the environmental interpretation.

Transcript and shot list

Target runtime about 7 minutes. Built as killing the wrong answers with evidence, installing the right one, then spending real time on the thing it still cannot explain. Shot types: ANIM is rendered from code with Manim, AI is a generated shot from Higgsfield with the shared style reference, REAL is agency imagery with credit, CAM is the narrator.

[00:00] S01 · AI · A full Moon low over a horizon, seen from Earth, unusually large in frame.

Before 1969 there were three respectable ideas about where this came from. It was captured, having formed somewhere else and wandered too close. It formed alongside Earth from the same material, a sister world. Or it was flung off a spinning young Earth, leaving a scar some people said was the Pacific. Then we went and got some rocks. And two of those three ideas died immediately.

[00:31] S02 · ANIM · Cosmic calendar; zoom into the first hours of 2 September.

This is the first hour or two of the 2nd of September on the cosmic calendar. Four and a half billion years ago.

[00:42] S03 · ANIM · Three failed models with their fatal problems: capture and the angular momentum budget; co-formation and the Moon’s tiny core; fission and an impossible spin rate.

Capture fails because the Earth-Moon system has far too much angular momentum, and because catching something that big requires a near-miracle. Co-formation fails on composition: the Moon has almost no iron core, about two percent of its mass against Earth’s thirty. Two bodies forming side by side from the same stuff shouldn’t differ like that. And fission fails because Earth would have had to spin impossibly fast. What the Apollo rocks actually showed was stranger than all three.

[01:18] S04 · ANIM · A Mars-sized body strikes the 90-percent-grown Earth at a glancing angle; the impactor’s core sinks and merges; a disc of vaporised rock is thrown into orbit; it gathers into the Moon within a century.

Here’s what happened. Near the end of Earth’s assembly, when it was about ninety percent grown, it was hit by another growing planet roughly the size of Mars. Not head-on. A glancing blow, at something like ten kilometres a second. Enough energy to melt the entire outer Earth and vaporise a good part of the impactor. The impactor’s iron core sank and merged with ours. And a disc of vaporised and molten rock, drawn mostly from the outer layers of both bodies, was thrown into orbit. Poor in iron, because the iron had already sunk. That disc cooled and gathered itself into the Moon, in a matter of decades. We’ve given the impactor a name: Theia, after the mother of the Moon goddess.

[02:14] S05 · ANIM · A checklist ticking off: angular momentum, small lunar core, missing volatile elements, timing.

And it explains what the others couldn’t. The angular momentum. The tiny core. The fact that the Moon is depleted in exactly the elements that boil away easily, because they were lost as vapour. And the timing.

[02:31] S06 · ANIM · Oxygen isotope fingerprints of meteorite groups scattered on a plot; Earth and Moon sit on top of each other; then a simulation shows most lunar material coming from the impactor, and a red flag.

There’s one thing it explains less well, and it’s worth saying so out loud. Every group of meteorites has its own isotopic fingerprint, because it depends on where in the solar system a body formed. Earth and the Moon have essentially the same fingerprint. In oxygen, in titanium, in several other elements. They’re isotopic twins. The problem: in most simulations, most of the material that ends up in orbit comes from the impactor. If Theia formed somewhere else, the Moon should carry its fingerprint. It doesn’t.

[03:11] S07 · ANIM · Three proposed fixes, each with a tick and a cross: same-orbit Theia; a far more violent, fully mixing impact; prolonged disc-Earth exchange.

There are ways out. Theia might have formed at almost exactly the same distance from the Sun. The impact might have been far more violent, vaporising and completely mixing both bodies. Or the disc and the Earth might have kept exchanging material long enough to even out. Each fix solves the isotopes and strains something else. The giant impact isn’t in doubt. Its details are being argued over right now.

[03:43] S08 · ANIM · The molten Earth: several hundred kilometres of magma ocean, a rock-vapour and steam atmosphere, radiating heat; a clock of a few million years; denser minerals sink, lighter ones float.

Whatever the details, the aftermath is clear. Earth’s outer several hundred kilometres were molten. No crust, no ocean, no air worth the name. A glowing surface under an atmosphere of rock vapour and steam. It didn’t last: a few million years at most before it froze. And as it froze, denser minerals sank and lighter ones floated, which is where the next lesson begins.

[04:13] S09 · REAL · A Jack Hills zircon grain under cathodoluminescence, showing growth zoning (research or museum image; confirm credit line).

Geologists called this first eon the Hadean, after the underworld, because they assumed half a billion years of magma and bombardment. Almost no rock survives from it, so there was little to argue with. Then, in the Jack Hills of Western Australia, they found these. Zircon crystals. Nearly indestructible, and when they form they trap uranium and reject lead, which makes them excellent clocks. Some are four point four billion years old. Older than any rock on Earth: eroded out of vanished rocks and recycled into younger sandstone.

[04:54] S10 · ANIM · Oxygen isotopes in the zircon consistent with low-temperature water interaction; a timeline showing liquid water within ~150 million years of formation; the word “Hadean” with a question mark.

Their chemistry is the surprise. The oxygen isotopes are consistent with crystals that grew from rock which had interacted with liquid water at low temperature. If that reading holds, then within about a hundred and fifty million years of forming, Earth had cooled enough for liquid water at its surface. Not hellish at all. And not for very long. That interpretation is still contested, and I’ll mark it as such.

[05:26] S11 · ANIM · Laser ranging to an Apollo reflector; 3.8 cm/yr; the Moon rewinds closer; huge tides; a short day; tidal sediment layers recording shorter days.

The Moon has been leaving ever since. Tides raised on Earth are dragged ahead by our rotation, and the pull nudges the Moon outward while slowing our spin. Reflectors left by Apollo let us measure the distance to millimetres: three point eight centimetres a year. You can’t just run that backwards, because it depends on the shape of the ocean basins, and those change. But the direction is certain, and layered tidal sediments record shorter days in the deep past. The Moon was once far closer, the tides enormous, and the day a matter of hours.

[06:10] S12 · ANIM · Earth’s axial tilt held steady by the Moon; Mars’s tilt wandering chaotically over tens of degrees.

And one consequence matters for everything after this. A large Moon steadies Earth’s tilt. Mars, without one, wobbles chaotically over tens of degrees. Stable tilt means stable seasons over geological time, which is a useful thing to have while life is getting started.

[06:30] S13 · ANIM · Labels: ESTABLISHED for the giant impact and tidal recession; ACTIVE RESEARCH for the isotopic twin problem, the Moon’s exact age and the cool early Earth.

The scorecard. That a giant impact made the Moon, and that it’s been receding ever since: established. Why Earth and Moon are isotopic twins when the simulations say they shouldn’t be, exactly when the impact happened, and how quickly the surface cooled: active research.

[06:50] S14 · AI into ANIM · The Moon again, then rewinding to fill half the sky; the calendar returns.

So the next clear night, look up. That thing is a piece of the Earth, thrown into the sky by the largest collision this planet has ever had, and it’s been drifting away ever since. Meanwhile, down here, a molten planet is about to sort itself into layers. Next lesson: core, mantle and crust.

Shot list

Two AI shots, eleven code-rendered shots, one research image. Estimated Higgsfield use: 3 to 6 generations including retries.