Coda: The Deep-Time Horizon

Every argument in this book has rested on a quantity that is decreasing.

The heat inside this planet comes from two sources, and neither is being replenished. There is the primordial heat — the energy of accretion and of core formation, released four and a half billion years ago and leaking out ever since, an inheritance being spent. And there is the radiogenic heat, from the decay of uranium, thorium and potassium threaded through the rock.

Radioactive decay is exponential. Every isotope has a half-life, and every half-life that passes leaves less of it. Uranium-235 has a half-life of about seven hundred million years, so there is only a small fraction of the original left. Potassium-40 is at around one and a quarter billion years, and is likewise well depleted. Uranium-238 and thorium-232 are longer-lived and are what remains of the fuel.

Put those together and the early Earth generated several times the radiogenic heat that it does now. It ran hotter, convected faster, and was more volcanically productive by a wide margin. What we are living through — the eruptions in this book, the plate motions, the entire tectonic system that Part III argued built the world — is the late, cool, comparatively sedate phase of a process that has been winding down since it started.

The engine has a fuel gauge, and the needle only moves one way.

This closing section is about what happens as it empties, and about a discovery I did not expect when I began writing, which is that the thing which ends the living world is not the engine failing.

Register: the physics of the decay is empirical. The projected sequence is theoretical, model-dependent, and carries error bars measured in hundreds of millions of years. The final pages are openly interpretive and I will say when I get there.

What winding down looks like

The order of events, with the caveat that the timescales are approximate to a degree that would be embarrassing anywhere else in this book.

As the mantle cools, it stiffens. Rock viscosity is strongly temperature-dependent, so a cooler mantle convects more sluggishly, and the plates — which are, as Chapter 1 established, simply the cold top of that convecting system — move more slowly.

There is a complication here that is more interesting than the simple thermal story, and it involves water. Subduction appears to depend on it: water weakens rock, lowers melting points, and lubricates the interface along which one plate slides beneath another. Chapter 7 described the deep water cycle, with ocean water carried into the mantle by subducting slabs and returned by arc volcanism. If the return is not quite balanced — if the mantle is a net sink — then the oceans are slowly draining into the interior, and estimates for how long the surface stays wet on that account are measured in the region of a billion years rather than several billion.

A drier Earth is a harder Earth to subduct. So plate tectonics may not end because the planet has run out of heat. It may end because the planet has run out of water to lubricate the machinery — which would stop the return path, and therefore the thermostat, and therefore everything Chapter 7 built its case on.

After that: arc volcanism ends with subduction. Hotspot volcanism persists longer, since it does not require plate boundaries, but declines with the heat. The core continues to solidify inward; at some point the convection driving the dynamo becomes insufficient and the magnetic field fails.

And then Earth is a stagnant-lid planet with a dead core. It is Mars, at greater size and with a longer runway.

Estimates for that endpoint span billions of years and I am not going to pretend to a precision the models do not have.

The thing that ends it first

But here is what I did not expect, and it reorganizes the whole ending.

The biosphere does not last long enough to see any of that.

The Sun is brightening. It has been doing so throughout Earth's history — this is the faint young Sun problem from Chapter 7, run forward instead of backward — at a rate of roughly one per cent every hundred million years or so. That is slow on any human reckoning and relentless on a geological one.

The carbonate–silicate thermostat has been compensating for it, and this is the mechanism I called the load-bearing argument of the book. As the planet warms, silicate weathering accelerates, carbon dioxide is drawn out of the atmosphere faster than volcanism replaces it, and the greenhouse effect weakens by exactly enough to offset the additional sunlight. It has done this for four billion years. It is why there has been liquid water throughout.

Now notice what the mechanism requires in order to keep working.

It compensates for a brighter Sun by lowering atmospheric carbon dioxide. That is its only move. And carbon dioxide cannot go below zero — in fact it cannot usefully go below a great deal more than zero, because plants need it.

Photosynthesis has a floor. The most common photosynthetic pathway becomes unable to operate below a carbon dioxide concentration in the region of a hundred and fifty parts per million. A second pathway, used by a minority of plants, can manage down to something nearer ten. Below that there is no carbon fixation, and below carbon fixation there is no food chain.

The projections — made repeatedly since the early 1990s, with refinements but no reversal of the conclusion — put the crossing of that threshold at roughly a billion years from now, once the minority of plants using the more carbon-efficient pathway are counted; without them it is sooner. More recent modelling of the associated collapse in atmospheric oxygen arrives at a comparable horizon, and some recent work extends the outer bound further still.

So the sequence is this. The Sun brightens. The thermostat responds, correctly and as designed, by drawing carbon dioxide down. It keeps drawing it down. And eventually it draws it down past the point at which plants can live, at which point photosynthesis stops, and then oxygen — which is entirely biological in origin, as Chapter 7 noted — declines, and the ozone layer with it, and complex life ends.

Not from a catastrophe. Not from an impact or an eruption or a nearby star. From the successful continued operation of the mechanism that made the whole thing possible.

The thermostat does not fail. The thermostat works, and working is what kills us.

The engine loses

And now the part that makes this the ending of this book rather than a general essay about deep time.

Throughout that final decline, the volcanoes are still erupting.

The engine is still there, still hot, still outgassing carbon dioxide into the atmosphere at broadly the rate it always has. It is doing precisely what Chapter 7 credited it with: supplying the source term, holding up its half of the loop, keeping carbon in circulation.

It is simply not enough. The sink — weathering, accelerating as the Sun brightens — pulls harder than the source can push. Volcanic outgassing does not stop. It gets outrun.

I find that a genuinely strange thing to sit with. Chapter 11 described the great extinctions as occasions when the source term overwhelmed the sink; the engine ran too fast and the weathering could not keep up, and most of what was alive died of it. The far future is the same relationship with the sign reversed. The sink overwhelms the source. The engine cannot outgas fast enough to keep the air breathable against a star that will not stop getting brighter.

Both ends of the story are about the same imbalance between two rates. Only the direction differs.

And the final irony, which I want to state carefully because it is the book's thesis arriving at its conclusion: the engine will still be running when the biosphere ends. Earth will be a geologically active planet with a dead surface — volcanoes erupting into an atmosphere nothing breathes, plate boundaries grinding under an empty sky, the whole apparatus that Part III argued built the living world continuing to operate over its remains.

The engine does not merely sustain the living world. Its inability to keep up is what will end it.

Mars, completed

Chapter 8 used the dead worlds as a control experiment. It is worth returning to Mars with everything now assembled, because it shows the endpoint and because it shows it in a specific and instructive way.

Mars died differently from how Earth will. It was small, so it lost its heat early; it never established a return path, so its carbon drained one way; and its atmosphere thinned until the surface was uninhabitable. It failed from the inside, from insufficiency.

Earth is scheduled to fail from the outside, from a star that will not stop, while its interior is still working.

But the result converges, and what Mars offers is a preview of the condition rather than the route.

There is one property of a dead planet that I find unexpectedly affecting, and it is this. On Earth, the record of the past is continuously destroyed. Erosion wears the mountains down, sediment buries what is left, subduction takes the ocean floor back into the mantle and melts it, and volcanism resurfaces whatever remains. The oldest ocean floor is a couple of hundred million years old; almost everything Earth has ever done to itself has been erased by the same processes that keep it alive.

Mars has none of that. Its surface is billions of years old in places. Its river valleys are still there, dry, with their channels intact. Its volcanoes still stand at full height because nothing has worn them down and no plate has carried them away.

A dead planet becomes its own museum. A living one is amnesiac. The price of being a world that renews itself is that it cannot remember, and the reason Mars can show us its youth is that it has had no youth since.

That is what the far side of the horizon looks like. Everything preserved, and nobody to read it.

Where we are

Register: interpretive from here.

Let me put us in the sequence, because I think the position is worth knowing and it is not widely held in mind.

The Earth is about four and a half billion years old. The habitable window has perhaps a billion years left, give or take a great deal. So we are somewhere around four-fifths of the way through the period during which this planet can support life at all.

Complex, multicellular life is younger — it has been around for something over half a billion years, and on these projections it has perhaps another half-billion or so ahead of it. Which places us, very roughly, near the middle of the era of complex life on Earth. Not at the beginning of something. Not at the end. Somewhere around halfway through the only interval in this planet's history when there has been anything here capable of noticing.

A billion years is not a deadline in any sense that should trouble anyone. It is four times longer than there have been animals. Nothing in this coda is a call to action and I would be embarrassed to pretend otherwise; the problems worth acting on are all several orders of magnitude closer, and Chapter 12 was careful to say that the mechanism described here is of no help with any of them.

What it is, I think, is a piece of orientation. And I want to say something about why orientation of this kind might matter, because I named the idea in the introduction and owe it a return.

The habit of thinking in deep time is unusual and fairly new. Almost nothing in ordinary life trains for it, and there is a reasonable case that a society which cannot hold hundreds of thousands of years in mind will keep making a particular class of error — will keep mistaking the recent for the normal, and the stable-looking for the permanent, and the slow for the absent. Learning that the ground under a city is a thin cooled skin over a body that is still losing heat is a specific instance of a general discipline, and I do not think the discipline is merely decorative.

What a firing planet is worth

So here is what I have come to, at the end of fourteen chapters of trying to be careful.

We tend to treat habitability as a property of a location — a planet is in the habitable zone or it is not, has water or does not, is the right size or is not. A checklist of attributes, possessed or lacked.

That is not what habitability is. Habitability is a process, actively maintained, consuming energy, and it can stop.

This planet is warm and wet and breathable because a mechanism has been running continuously for four billion years: heat escaping from an interior, carrying material as well as energy, building crust at ridges and consuming it at trenches, exhaling gas and drawing it back down, holding a chemical balance in the air against a star that has been getting brighter the whole time. It is not a state the Earth is in. It is a thing the Earth is doing, and it will not always be doing it.

That reframing has been the whole argument. The eruption is not an interruption of the ordinary condition. The eruption is the ordinary condition, made briefly visible.

Which brings me back to a cornfield in Michoacán, on an afternoon in February 1943, where a farmer stood in his own field and watched a mountain begin.

I said in the introduction that this was not a metaphor, and it was not. What Dionisio Pulido saw was the mechanism that made the ground he was standing on, and the air he was breathing, and the continent under both — running, as it always is, and visible, just that once, to a man with a plough.

We do not live on a stable world that occasionally erupts.

We live inside a window that is being held open by heat.

The eruption is the sound of it being held.

Draft notes — verification status

Standing convention. This coda's projections are model-dependent and the timescales vary between studies. The central claim — that the biosphere ends via the carbon cycle while the engine is still running — is the book's thesis in its final form and must be sourced properly.

Pending verification — high priority:

• The photosynthesis threshold. C3 pathway failure at ~150 ppm CO₂ and C4 at ~10 ppm. Verify both figures and the framing.
• The timescale to that threshold. Caldeira & Kasting (1992) is the standard reference and gives a range; later work has refined it. The draft says "five hundred million to a billion years" — check this against current estimates, which vary considerably depending on assumptions about weathering and continental configuration.
• Ozaki & Reinhard (2021) on atmospheric deoxygenation timing (~1 Gyr) — confirm the finding and its stated horizon.
• Solar luminosity increase rate (~1% per 110 Myr). Already flagged as inconsistent between Chapters 7 and 9 — reconcile across all three and this coda.
• The ocean-into-mantle argument. That subduction returns less water than it consumes, that surface water may be substantially depleted on ~1 Gyr timescales, and that this could stall plate tectonics before thermal decline does. This is a real position in the literature but it is one position among several — verify and characterize the uncertainty honestly, as the draft attempts to.

Pending verification — standard:

• Half-lives: U-235 (~704 Myr), U-238 (~4.47 Gyr), Th-232 (~14.05 Gyr), K-40 (~1.25 Gyr).
• Radiogenic heat production at 4.5 Ga relative to present ("several times"). Chapter 1 makes a similar claim — keep consistent.
• Total surface heat flow (~46–47 TW) — already flagged in Chapter 1 notes; the draft here avoids the figure, which may be the right choice.
• Timescale for geodynamo shutdown and inner core growth — very uncertain; the draft deliberately declines to give a number.
• Age of oldest oceanic crust (~180–200 Ma) versus oldest Martian surfaces. Chapter 7 makes the oceanic crust claim — keep consistent.
• Age of complex/multicellular life (~540 Ma for the Cambrian radiation; earlier for multicellularity generally — be precise about which claim is being made).
• Red giant phase timing (~5 Gyr) — not stated in the draft; add only if needed.
• The Bjornerud deep-time-as-civic-capacity argument — already flagged in the Introduction notes; verify once and use consistently in both places.
• Paricutín and Dionisio Pulido — already flagged in the Introduction notes as needing primary-source checking. The coda's closing bookend depends on the same facts; if the Pulido attribution does not survive verification, both passages change together.