Where Does Earth’s Water Come From?
We don’t yet know for sure, but we have some ideas.
1. The Two Mysteries of Earth’s Water
Earth looks blue from space, but we are not a water world.
Several icy moons (Europa, Ganymede, Titan, Callisto, and others) have a far better claim to the moniker. These moons are predicted to contain much more water than Earth by volume, and to have a far larger fraction of their mass as water. All the water on Earth — oceans, ice caps, groundwater, clouds, every river and lake, the glass of water on your nightstand, all of it — masses about ~1021 kg. The planet is 6×1024 kg. So water is about ~0.02% of the whole. In other words, we are a rock with condensation on it. If Earth were a basketball, and you collected all that condensation, all the water would form a bead about the size of a marble.
The inner solar system, the zone where Earth, Venus, Mars, and Mercury formed about 4.5 billion years ago, was hellishly hot during planetary formation. All water molecules should have been evaporated and pushed beyond the snow line, the distance from the young Sun beyond which water ice could condense as a solid. For our sun, that snow line was 2.5 to 3 AU (Astronomical Unit, defined as the average distance between the Earth and the Sun), which is well beyond Mars.
In other words, the rocks that built Earth coalesced from dust that must have been too hot to hold water ice. Moons of Jupiter and Saturn having vast subsurface oceans of liquid water isn’t surprising. The real surprise runs in the opposite direction: Earth shouldn’t have any water at all. But it does. This is the first mystery.
The second mystery: Earth has an extraordinarily specific amount of water, a kind of Goldilocks quantity.
If the Earth had less water, there would be no oceans, no plate tectonics1, no silicate weathering cycle to regulate CO2; no CO2 regulation would mean no planetary thermostat, and potentially no complex life. Mars was probably like this: it may have had surface water early on, but not enough to sustain a geodynamic cycle.
But if the Earth had more water, it would be an ocean world with no exposed continents. Exposed continents are required for carbon-silicate and nutrient cycling.2 Carbon-silicate cycling, like the silicate weathering cycle, is necessary for regulating CO2. And the absence of nutrient cycling would potentially mean no complex life. Scientific modeling suggests that just ~10X more water, i.e., if Earth was 0.2% water instead of 0.02% water, we would have no exposed land at all.
Earth sits in a narrow band. Enough water to run plate tectonics, which requires hydrated minerals sinking back into the mantle. Enough to sustain a surface biosphere. But not so much that you drown all surfaces and stop the carbon-silicate cycle — that million-year synchronized dance stabilizing long-term climate by sequestering carbon. Whether this is a coincidence, or the anthropic selection principle at play (we can only observe it because we’re here), or a natural outcome of planetary formation, is a part of the mystery. This mystery is connected to the deeper question of how common or rare Life in the Universe is.
Because we shouldn’t have water, but we have exactly the right amount, scientists are forced to play detective to find out who delivered it—or if it was here all along. Think of it as an inverse of a missing-person case, a cosmic mystery of found-water. It is a cosmic mystery because of its implications: If we can explain how a planet born in a hellishly hot zone ended up with a life-sustaining Goldilocks’ coating of water, we may be able to say how many other life-sustaining blue planets might be possible in the Universe.
2. The Exogenous Orthodoxy
Earth is a middle-aged planet, approximately 4.5 billion years old. It’s mostly settled in now, barring a few hot flashes of volcanism and some tectonic instabilities now and then. But our planet had an intense and explosive early childhood. So much so that geologists have named this period the Hadean Eon, after Hades, the Greek god of the underworld. This eon encompasses the first 500 million years of our planet’s existence.
Earth inherited its violent early childhood from its parental solar nebula, a churning primordial soup of incredibly hot, highly pressurized gases that collapsed into the protoplanetary disk about 4.6 billion years ago.
From this disk, our solar system eventually emerged. The Hadean Earth began with a 3,000°C magma ocean that solidified in 5 million years – quick by the standards of geological time. The Hadean atmosphere was marked by extreme volcanism, and there was a constant bombardment of and collisions with other bodies as they tried to find their footing in the protoplanetary churn. Scientists have found evidence of water in rocks from this era, dating back 4.4 billion years. Where did this water come from?
The “late veneer” hypothesis posits that between about 4.45 and 4.35 billion years ago, even before the Moon formed, Earth was struck by a group of meteorites called chondrites that were rich in volatiles, including water.
Along with water, these chondrites also contained highly “siderophile” elements. “Siderophile” just means “soluble in molten iron.” These siderophile elements, brought to Earth by chondrites, became part of Earth’s mantle but not its core. As per the late-veneer hypothesis, these chondrites hit Earth after the Earth’s core had formed, but while Earth’s mantle was still setting. The elements that made up these meteorites thinly coated the Earth, making a ‘veneer.’
What does the evidence say about this hypothesis?
Since we are talking about mysteries, let’s consider the forensic tools available to us.
The first tool we will consider is the noble gases, named for their chemical inertness. They are the ultimate control group. They are chemically inert, so they just sit there, preserving the properties they had when the solar system formed. Their current whereabouts can be very informative about the deep past.
Light noble gases (He, Ne) are found deep in Earth’s inner mantle, bearing signatures of the solar nebula. On the other hand, heavy noble gases (Kr, Xe) are found in the shallower mantle and in the atmosphere. These show signatures of chondrites, indicating that heavy noble gases were delivered exogenously via chondrites. So far, so good for the late veneer: the heavier gases come from beyond the snow line and are concentrated near the surface.3
The second forensic tool we have is the D/H isotopic ratio, which is hydrogen’s fingerprint.
Every element comes in slightly different versions called isotopes: the same number of protons (which define the element), but a different number of neutrons (which changes its mass). Most of the time, these versions behave identically in chemistry. But their slight mass differences mean that the isotopes get sorted by physical processes. These processes can be evaporation, condensation, nuclear reactions, radiation chemistry, and others.
Different environments in the early solar system processed isotopes differently, leaving distinct fingerprints. By measuring these ‘fingerprints’ in Earth’s water, in meteorites, in oceans, and in the deep mantle, scientists can try to match them to their source in the Solar system, like detectives matching crime-scene evidence to a suspect.
Hydrogen, the simplest atom, has two stable isotopes: ordinary hydrogen (1H, one proton, no neutrons) and deuterium (2H or D, one proton, one neutron). Deuterium is rare — roughly 1 atom for every 6,400 hydrogen atoms in Earth’s ocean — but its relative abundance varies dramatically across the solar system, because different environments process these isotopes differently.
Cold, radiation-rich environments in the outer solar system concentrate deuterium. The hot inner disk depletes it. This makes the deuterium-to-hydrogen ratio (D/H) a fingerprint: if you measure the D/H of Earth’s water and the D/H of a suspected source, you can check whether the fingerprints match.
When we measure the D/H ratios of Earth’s oceans, it closely matches the wet, carbonaceous (i.e., carbon-containing) chondrites from the outer solar system.
So we now have two pieces of evidence supporting the late veneer: the heavy noble gas abundances in Earth’s outer mantle and the D/H ratio of Earth’s oceans.
But how did these water-rich siderophile-containing chondritic meteorites from beyond the snow line bombard Earth so soon after the solar system had started to form? A possible explanation involves Jupiter and something called the Grand Track Hypothesis: About 2 to 5 million years after the Solar System’s birth, Jupiter took a 1-million-year brisk walk inward, reaching around 1.5 AU before turning back out again. This migration may have scattered water‑rich planetesimals from beyond the snow line into the inner Solar System.
All of this sounds good for the theory of exogenous delivery. Almost too good.
3. The Data Rebellion
Here’s the thing: no tool is perfect.
For example, take D/H ratios. D/H ratio can be altered by several processes. Light H escapes to space faster than heavy D, a process called atmospheric escape. There’s also something called ‘isotopic fractionation’ that happens during the formation of the planetary core: Deuterium, being heavier and soluble in molten iron at high pressures, sinks into the core. As a result, the present-day ocean D/H may not faithfully record the original source. We have to take this into account while considering the balance of evidence.
Here’s another thing: different forensic tools may point in different directions.
Consider Highly Siderophile Elements, or HSEs. Although there are many HSEs, we will limit ourselves to three here: Platinum (Pt), Gold (Au), and Ruthenium (Ru).
“Siderophile” literally means “iron-loving.” And the HSEs represent an abundance anomaly that supports late-veneer. These elements are so iron-loving that core formation should have stripped them from the mantle almost completely. The sinking, molten iron should have ruthlessly stripped the mantle of virtually all its platinum and gold, dragging it down into the inaccessible core. Yet the mantle contains them at 100–1,000 times the concentration they should have if the core stripped them. Enough that the world’s economy ran on Gold.
The presence of HSEs in Earth’s mantle makes the late veneer hypothesis almost elegant. The late-stage bombardment of wet, carbonaceous chondrites after solid core formation is all we need to explain the presence of gold and platinum that should have been absent from the mantle. It delivers the missing HSEs and fills the oceans.
However, while late veneer appears to resolve the abundance anomaly for Gold and Platinum, it offers no explanation for the isotopic anomaly in Ruthenium.
Ruthenium is one of those HSEs delivered by the late-veneer, and it carries its own isotopic fingerprint. When geochemists finally measured the Ruthenium isotopes in Earth’s mantle, the data contradicted the elegant model. The Ruthenium does not match wet, carbonaceous chondrites from the outer solar nebula. But it perfectly matches dry, enstatite-like rocks from the inner solar system. These enstatite rocks were formed in the inner disk closer to the sun, too much on the wrong side of the snow-line to contain any water.
We are forced into a mathematically necessary concession: The Late Veneer was dry. It successfully delivered platinum and gold to the upper mantle via bombardment by bone-dry enstatite meteorites, but it brought absolutely no water.
The mystery deepens.
4. The Triple Oxygen Isotopes
Let’s now understand what I think is the most impressive forensic tool available to us based on isotopic ratios: the triple oxygen isotopes.
Oxygen has three isotopes,16O, 17O, and 18O, with 8, 9, and 10 neutrons, respectively (16O is the most abundant, ‘normal’ isotope of Oxygen, containing 8 protons and 8 neutrons). As with any other mixture of isotopes, normal processes fractionate them in predictable mass-dependent ways: 18O, containing two extra neutrons, is ~12% heavier than 16O; while the 17O isotope, containing one extra neutron, is ~6% heavier than 16O.
The notable feature here is the linear relationship: the mass difference between 18O and 16O is twice that between 17O and 16O. Therefore, any standard chemical process that alters the ratio of 18O/16O in a system will mathematically alter the 17O/16O ratio by about half as much.
If you plot the 17O/16O ratio (y-axis) against the 18O/16O ratio (x-axis) for any standard geological sample on Earth, they all fall on a single, predictable line with a slope of roughly 0.52. Normal planetary chemistry can shift a sample's isotopic signature up or down along this line, depending on the temperature and phases involved, but it cannot move a sample off this line. This is mass-dependent fractionation.
Here’s a question: if an extreme heating event caused a liquid body to rapidly evaporate, selectively losing its highly mobile, lighter isotopes to space, how would you expect the isotopic signature of the remaining liquid to shift on that mass-dependent plot? Answer: The lighter isotopes vaporize first, enriching the residue in the heavier ones. You will move up on the mass-dependent line.
Now, this is where things get interesting in cosmochemistry. Because when cosmochemists analyze certain early solar system materials, specifically samples from primitive meteorites, they find isotopic signatures that completely ignore the mass rules. Instead of following the ~0.52 slope, these components often plot along a slope of roughly 1.0. This indicates that processes in the early solar nebula (likely photochemical reactions or the mixing of stardust from different supernovas) were adding or removing 17O and 18O in equal proportions, regardless of their mass difference. This is mass-independent fractionation.
This deviation from the mass-dependent straight line due to mass-independent fractionation is quantified as Δ17O. If Δ17O = 0, then the material shares Earth’s starting isotopic baseline. If Δ17O ≠ 0, then the material originated from a completely different isotopic reservoir in the solar nebula. The scientific literature calls this deviation a “genetic tag” that cannot be erased by ordinary chemistry. This makes Δ17O the most robust forensic tool in cosmochemistry. It is harder to fake than D/H.
Suppose a primitive asteroid forms from solar nebula dust that has a distinct Δ17O of +2.0, i.e., the vertical offset from the terrestrial water line. Millions of years later, radioactive decay causes the entire asteroid to melt, undergo complex igneous chemistry, and re-crystallize into various new rock types. When we measure and plot the 17O/16O ratio (y-axis) against the 18O/16O ratio (x-axis) of these new rock types, what will we see in comparison to the terrestrial mass-dependent line?
Remember, thermodynamic processes such as melting or crystallization merely shuffle isotopes along a mass-dependent slope of ~0.52. The process shifts the coordinate positions but preserves the vertical offset. What we will see when we plot the ratios is a line parallel to the terrestrial oxygen isotope line, but offset by 2.0. The process shifts the coordinate positions but preserves the vertical offset. The genetic tag survives the planetary blender.
Each meteorite group sits on its own oxygen isotope line. Let's consider the oldest meteorites, the chondrites. Chondrites are the primitive meteorites that represent the raw, unblended solid material of the early solar system. Chondrites are divided into different groups based on their composition. When we plot the oxygen isotopes of different chondrite groups, they do not share a single baseline. Each distinct group sits on its own parallel mass-dependent line. They each possess a unique Δ17O signature, demonstrating they accreted in isolated, poorly mixed reservoirs within the solar nebula.
Earth and chondrites containing the mineral enstatite sit together, lying exactly on the terrestrial line (Δ17O = 0). Most carbonaceous chondrites, on the other hand, sit at a different table in the isotope-line cafeteria. They plot on entirely different lines, indicating that they are isotopically foreign to Earth’s reservoir. However, there’s one exception: The CI carbonaceous chondrites overlap with Earth on the isotope line. These are highly water-rich but surprisingly share the terrestrial Δ17O = 0 baseline.
What does this mean for the exogenous and endogenous delivery theories?
5. The Exogenous Revival?
The isotopic mapping of triple oxygen imposes a strict boundary condition for any model of planetary accretion.
Consider the following three empirical facts:
The bone-dry enstatite chondrites plot exactly on the terrestrial line (Δ17O = 0), meaning they have an ancient kinship with Earth.
Most carbonaceous chondrites plot on entirely different lines, meaning they are isotopically foreign to Earth’s reservoir.
The highly water-rich CI Carbonaceous Chondrites share the terrestrial Δ17O = 0 baseline.
Q: What is the only way these three empirical facts fit with the fourth empirical fact that Earth has water?
A: The bulk of the planet accretes from the dry enstatite reservoir, and the hydrosphere is delivered by CI Carbonaceous chondrites. The isotopic ledger balances perfectly, explaining the empirical facts.
Isotopically, my dear waterworld.
But here’s a temporal problem with this model: Planetary accretion is violently hot. If CI material is mixed in during the main construction phase of the planet, then all that highly volatile water will evaporate on contact with Earth, which at this point is a 3,000°C ball of radiating magma. The only way this explains water on Earth is if our planet builds its bulk mass hot and dry from the enstatite reservoir, lets the crust solidify, and then a subsequent bombardment of CI material coats the cooled surface with water.
As for D/H ratios, even when we account for the kinetic fractionation over time that lets H preferentially escape the atmosphere, the empirical data hold up: the D/H ratio measured in the hydrated minerals of CI chondrites is a remarkably tight match to Earth’s bulk oceans.
But we have to look at the entire geochemical package of the C1 meteorites, not just oxygen and hydrogen. If CI meteorites delivered water on Earth, they did not hold back the rest of the package. These meteorites contain a specific, quantifiable ratio of volatiles. We must have gotten all of it, the entire elemental payload.
This is where the elemental spreadsheet shows an imbalance, even though the isotopic spreadsheet appears balanced.
Elementally, my dear waterworld.
If CI meteorites slammed enough water into early Earth to fill the oceans, then they must have also delivered their massive carbon and nitrogen payloads. BUT Earth’s ‘silicate’ layer — the crust, mantle, and oceans combined — is profoundly depleted in carbon and nitrogen relative to water when compared to CI chondrites. If we had gotten a proportional amount of carbon from these meteorites, there would have been a supercritical CO₂ atmosphere on Earth, making Venus look like a mild vacation spot in comparison. The most we can say about exogenous delivery is that we may have received some water from outside, but definitely not all of it.
There’s another possibility: maybe the CI chondrites delivered water after the Earth cooled, but this leads to the carbon anomaly. However, if CI chondrites slammed into a hot Earth, then that will explain the carbon anomaly: the sinking molten iron swallowed the carbon. And it also hints at the possibility for water: the silicate mantle retained the water, eventually outgassing it to form the hydrosphere. But if the water was present before Earth cooled, where was it hiding while the planet was a churning, 3,000°C ball of radiating magma? And if there was a way for the water to safely hide out from the heat until the planet cooled, do we even need to consider exogenous delivery at all?
Thanks to Jannik Reigl, Andrew Miller, Elizabeth Van Nostrand, Hiya Jain, and Tina Marsh Dalton for helpful and extensive comments. And thanks, as always, to Mike Riggs for editing and feedback on multiple drafts of this essay from the beginning to the end of the writing process.
Water lubricates subduction zones and lowers the melting point of the mantle wedge.
No exposed continent means no continental weathering, which means no phosphorus delivery to the ocean, hence no nutrient cycling.
The composition of asteroids beyond the snow line differs from that of asteroids that originate within the snow line. Chondrites escape the violence of the formation.
Thank you for reading! And your kind words!!
That’s amazing. I knew bits and pieces of it, but thank you so much for the deep dive