Deep Carbon Cycle — Earth's Billion-Year Climate Thermostat
A diamond pulled from a kimberlite pipe in Juina, Brazil contains a fleck of ice. Not surface ice — ice-VII, a phase of water stable only above 2 gigapascals. The diamond crystallized 700 kilometers below the surface, and the water trapped inside it last saw the sky maybe 500 million years ago, as seawater on a plate that subsequently sank into the mantle. Earth has a slow plumbing system for carbon, and the diamonds are its leak tests.
The deep carbon cycle is the planetary subsystem that links seafloor chemistry to mantle mineralogy on million-to-billion-year timescales. Earth's mantle holds roughly 10⁸ gigatons of carbon — about 100,000× the atmosphere (800 GtC) and 2,500× the ocean (38,000 GtC). The cycle is small in annual flux and enormous in cumulative effect: it is the reason Earth has not frozen solid or cooked itself in CO₂ over four billion years.
How carbon goes down
Three carriers ride subducting oceanic plates into the mantle:
| Carrier | Flux (MtC/yr) | What it is |
|---|---|---|
| Carbonate sediments | ~15–60 | CaCO₃ from marine shells settled on the seafloor |
| Altered oceanic crust | ~18–61 | Basalt veined with carbonates by hydrothermal seawater |
| Organic carbon | ~10–25 | Buried biomass, isotopically light (δ¹³C −20 to −30‰) |
Total descending: ~40–120 MtC/yr. Total returning via arc volcanoes (Cascades, Andes, Sunda): ~35–65 MtC/yr. The mantle is a net sink on human timescales and a regulated reservoir on geological ones.
What pressure does to carbon
Carbonate is not stable everywhere in the mantle. Below ~250 km, the mantle contains trace metallic iron, and the chemistry flips reducing. Carbonate melt meeting Fe-bearing peridotite produces iron carbides (Fe₃C, Fe₇C₃), which reduce further into graphite and ultimately diamond. A 2025 paper in Scientific Reports mapped the redox conditions: CaCO₃ + Fe + SiO₂ → CaSiO₃ + FeO + Fe₃C at 25–53 GPa and 1,500–2,000 K, corresponding to 700–1,300 km depth. This is the deepest carbon cycling ever documented in natural samples.
The Fe-carbide intermediate also explains two long-standing puzzles. Kinetic isotope fractionation concentrates ¹²C into the carbide, leaving some superdeep diamonds at δ¹³C −40‰ — too light for any mantle source, fingerprinting organic carbon all the way down. And carbides act as nitrogen sinks, which is why these diamonds are anomalously nitrogen-poor.
Superdeep diamonds as messengers
Most diamonds form at 150–200 km, in the cold lithospheric keels under cratons. Superdeep diamonds form at 250–800+ km and arrive intact only because kimberlite eruptions punch them upward through the mantle in hours — too fast for their mineral inclusions to re-equilibrate.
Those inclusions are the data: CaSiO₃-perovskite (stable only below the 660 km discontinuity), ringwoodite, ferropericlase, ice-VII. A 2020 Nature paper confirmed several Juina diamonds carried inclusions only stable in the lower mantle. Whatever you think of diamonds as jewelry, they are the only direct samples we have of rock from below 660 kilometers.
The thermostat
Strip the cycle and Earth's surface CO₂ has two failure modes:
- No volcanic return → silicate weathering drains atmospheric CO₂ → temperatures drop → glaciers expand → Snowball Earth within tens of millions of years.
- No subduction sink → outgassing accumulates → runaway greenhouse → something like Venus (96% CO₂, 92 bar surface pressure).
The cycle gives Earth a negative feedback loop with a multi-million-year time constant. Cold periods slow weathering; warm periods accelerate it. A 2018 Science Advances analysis found ~26-million-year oscillations in CO₂ traceable to seafloor spreading rate — large enough to plausibly contribute to the disputed ~26 Myr periodicity in marine extinctions.
Why Mars froze and Venus cooked
The cycle requires mobile-lid plate tectonics — Earth's particular mode of convection. Mars switched to stagnant-lid early (possibly always was), lost its dynamo around 4 Ga, and lacked any mechanism to buffer atmospheric CO₂ against changing solar luminosity. Liquid water on Mars may have lasted only 100–500 million years. Venus runs an "episodic-squishy lid" regime — catastrophic resurfacing without sustained subduction. The atmosphere is what outgassing without a sink looks like at steady state. See concept planetary tectonics for the six-regime taxonomy.
What's contested
Flux numbers carry factor-of-three uncertainty even in 2025. Different teams using different methods (volatile inclusion analysis, slab thermal modeling, volcanic emission inventories) disagree on how much carbon actually reaches the lower mantle versus being recycled in the upper few hundred kilometers. The organic-carbon fraction of the subducting flux is especially contested — estimates range from 10% to 30% — and matters disproportionately because organic carbon is what gives superdeep diamonds their isotopic fingerprint of buried life.
Whether the deep biosphere (microbial communities 2–5 km down, feeding on serpentinization-derived H₂) is a meaningful node in the carbon cycle or a passive passenger is also unresolved. Some estimates put it at >50% of Earth's biomass; if true, biology participates in the deep cycle, not just on the surface.
Cross-realm bridge
Carbon's range across the wiki is absurd. The same element makes concept graphene (single-atom sheet, room temperature, lab synthesis) and superdeep diamond (cubic lattice, 25 GPa, mantle synthesis over geological time). Living things eat one form and exhale another; subduction takes the exhaust and presses it for 500 million years until it becomes the hardest natural material we know. The carbon in a wedding ring may be reincarnated phytoplankton.
The astrobiology stake is sharper. If mobile-lid tectonics is rare — and the current planetary census suggests it might be — then long-term climate stability via deep carbon cycling becomes a concept fermi paradox filter. Europa and Enceladus may have liquid water, serpentinization, and the chemistry to seed life. They do not have subduction. Any biosphere there exists without a planetary thermostat, which probably bounds how complex it can get and how long it can persist.
An open question
If the diamonds in Juina are crystallized marine life from the Cambrian, what's down there now from the Anthropocene — and when, in 400 million years, does it come back up?
Key sources
- Pearson, D. G. et al. (2014, Nature) — ringwoodite inclusion in superdeep diamond, first direct sample of mantle transition zone water.
- Walter, M. J. et al. (2011, Science) — Juina superdeep diamonds and subducted oceanic crust as their source.
- Smith, E. M. et al. (2016, Science) — CLIPPIR diamonds and metallic iron in the deep mantle.
- Scientific Reports (2025) — redox-driven superdeep diamond formation at 25–53 GPa; to verify exact author/title.
- Science Advances (2018) — 26 Myr oceanic crust carbon oscillations; to verify Müller & Dutkiewicz or similar.
- Deep Carbon Observatory synthesis volume (2019, Cambridge University Press) — load-bearing reference for flux estimates and the field's consensus state at the end of the DCO decade.
Further reading
- Symphony in C by Robert Hazen — carbon across realms, written by the geochemist who ran the Deep Carbon Observatory. The closest thing to a single-author overview of this whole topic.
- Deep Carbon Observatory final reports (deepcarbon.net archive) — ten years of coordinated research, freely accessible.
- The Story of Earth by Robert Hazen — wider planetary evolution context for why the carbon cycle even exists.
- Marie Edmonds' lectures on volcanic degassing (Cambridge, available on YouTube) — the volcanic-return half of the cycle explained by someone who measures it.
See Also
- concept planetary tectonics — the six tectonic regimes; mobile lid is the cycle's enabling condition.
- concept great oxygenation event — photosynthesis built the organic-carbon pool that eventually subducts and crystallizes as light-δ¹³C diamonds.
- concept extremophiles — the deep biosphere lives on serpentinization chemistry, which is also the cycle's H₂ producer.
- concept fermi paradox — long-term planetary thermostats may be a Rare Earth criterion; this is the candidate mechanism.
- concept graphene — carbon's other extreme, 2D and atmospheric-pressure, the same element in a wholly different lattice.
- concept deep ocean — hadal hydrothermal systems are where altered oceanic crust starts its trip down.
- concept permafrost methane — the fast carbon "bomb"; the deep cycle is its billion-year counterweight.