Chapter 14: Tapping the Furnace

In 1904, in a valley in Tuscany where hot vapour had been rising out of the ground for as long as anyone had records, a man connected a small engine to the steam and lit five light bulbs.

The place is Larderello. The steam had been exploited for centuries before that for boric acid, and the surrounding landscape has been called an inferno for longer still — Dante is said to have taken some of his imagery from it. In 1913 the first commercial geothermal power station opened there.

It is still generating.

That is a century and a bit of continuous electricity from a hole in the ground in a volcanic district, and it is worth beginning with because it establishes the least controversial claim in this chapter: using the engine is not speculative. It is an old, working, unglamorous industry.

This is the book's constructive chapter. Everything before it has been about surviving the engine, understanding it, or crediting it with the world. Here it becomes something to use on purpose.

Register: empirical for existing technology and its performance; theoretical to speculative for the frontier technologies, each marked. The closing assessment is my own judgement and is labelled.

Where it actually stands

Let me start with the honest scale, because geothermal's advocates have a long history of overpromising and I do not intend to add to it.

Global installed geothermal electricity capacity is on the order of sixteen gigawatts. For comparison, the world has installed something like a thousand gigawatts of wind and rather more than that of solar. In global energy terms, geothermal is a rounding error.

But raw capacity is the wrong measure, and here is why. Solar produces when the sun shines and wind when the wind blows; their capacity factors — the fraction of nameplate capacity actually delivered over a year — are typically somewhere between fifteen and forty-five per cent. A geothermal plant runs at seventy-five to ninety. It produces at three in the morning in January. It is unaffected by weather. It is one of the very few low-carbon sources that is genuinely firm.

In a grid increasingly full of intermittent generation, firmness is the scarce commodity, and its value is rising rather than falling. That, and not raw scale, is geothermal's actual argument.

Now the cases where it is not marginal at all.

Kenya. The Olkaria complex, in the Rift Valley, generates something approaching a gigawatt, and geothermal supplies a very large share of Kenyan electricity — on some accounting close to half. This deserves more attention than it gets, because it is the clearest demonstration anywhere that a country can build a substantially decarbonized grid on volcanic heat. The East African Rift is the surface expression of a continent being pulled apart, which will eventually make a new ocean; in the meantime it boils water for Nairobi.

Iceland generates a large fraction of its electricity geothermally and, more impressively, heats the overwhelming majority of its buildings that way, through district heating networks that pipe hot water directly to homes. Reykjavík's air quality improved dramatically when the city stopped burning coal for heat, which is a public-health benefit rather than a climate one and is easy to overlook.

The Philippines, Indonesia, Turkey, New Zealand, Mexico, the United States all have substantial capacity, and in each case it is concentrated exactly where you would predict from Chapter 1 — arcs, rifts, hotspots.

Which is the constraint. Conventional geothermal needs three things in the same place: heat, water, and permeable rock to hold and move it. Heat alone is not enough; you need a reservoir. That combination occurs in volcanic and rift settings and largely not elsewhere, which is why the industry map is the tectonic map.

Two further honesty items. Geothermal fields can deplete — the Geysers field in California suffered significant pressure decline before it was partially rescued by injecting treated wastewater to recharge it. And geothermal is low-carbon rather than zero-carbon: the fluids carry dissolved carbon dioxide and hydrogen sulfide, emissions vary considerably between fields, and while the figures are far below any fossil alternative they are not nothing.

The supercritical frontier

Register: demonstrated in individual wells; not yet demonstrated as a commercial technology.

Water has a critical point at around three hundred and seventy-four degrees and two hundred and twenty atmospheres. Above it, the distinction between liquid and vapour disappears; there is a single supercritical phase with high enthalpy and low viscosity — a great deal of energy per unit mass, in a fluid that flows easily.

The prize is straightforward: a supercritical well could deliver something like five to ten times the power of a conventional one. Fewer wells, less surface footprint, better economics.

We know this because of the accident described in Chapter 3.

The Iceland Deep Drilling Project's first well, at Krafla in 2009, was aiming for supercritical conditions at four and a half kilometres and instead struck rhyolite magma at about two. Rather than abandon the hole, the project completed it as a producer — and it delivered superheated steam at around four hundred and fifty degrees, at the highest enthalpy ever obtained from a geothermal well, with an estimated electrical output several times that of a conventional well on the same field.

It also failed within a couple of years, and the reason is the frontier's actual problem. The fluids at those conditions are corrosive — acidic, chloride-bearing, and chemically aggressive toward steel at temperature. Valve and casing materials are the binding constraint, not the geology and not the drilling.

A second well, drilled on the Reykjanes peninsula and completed in 2017, reached beyond four and a half kilometres and encountered supercritical conditions without hitting magma, demonstrating that the drilling itself is achievable.

So the position is: the resource is real, the physics is favourable, one well proved the output, and the engineering problem is metallurgy and well control. That is a much better place to be than most energy frontiers, and it is still a long way from a power station.

Enhanced geothermal, and the earthquakes

Register: technically demonstrated, commercially unproven, and carrying a specific hazard that must not be minimized.

Everything above depends on finding a natural reservoir. The much larger prize is not needing one.

Heat is everywhere. Drill deep enough anywhere on Earth and the rock is hot. What is usually missing is permeability — the rock does not transmit fluid, so there is nothing to circulate and nothing to bring the heat up. Enhanced geothermal systems propose to manufacture the permeability: drill into hot dry rock, hydraulically stimulate it to open and connect fractures, and circulate water through the fracture network between injection and production wells.

If it worked at scale and at cost, it would decouple geothermal from volcanic settings entirely and turn a niche resource into a globally distributed one. The technical resource is enormous.

It has also, twice, caused serious harm, and I want to give that its full weight rather than a parenthesis.

In Basel in 2006, a deep geothermal project stimulating rock beneath the city induced an earthquake of around magnitude three and a half. There was no major structural damage but there were widespread damage claims, considerable public alarm, and the project was suspended and ultimately abandoned.

In Pohang, South Korea, on the fifteenth of November 2017, an earthquake of magnitude five point four struck near an EGS site. Eighty-two people were injured and the damage was extensive — it is the most damaging earthquake recorded in South Korea since instrumental observation began there in 1905, and the largest known induced earthquake at any enhanced geothermal site anywhere. A Korean government commission concluded in 2019 that it had been triggered by fluid injection at the project, and the project was terminated.

Pohang is the single most important fact about EGS. It is not a theoretical risk or a nuisance-tremor problem; it was a damaging earthquake in a city, attributed by an official inquiry to a fluid injection project. Any account of enhanced geothermal that does not put that near the front is selling something.

The mitigations are real but they are mitigations rather than solutions: siting away from critically stressed faults, characterizing the local stress field before stimulating, limiting injection pressures and volumes, and operating traffic-light protocols that suspend injection when seismicity crosses thresholds. Pohang's specific failure appears to have involved stimulating close to a previously unrecognized fault, which is precisely the kind of thing the inference problems of Chapters 3 and 4 make hard to rule out in advance.

There is, alongside this, genuine recent progress worth reporting. Techniques developed by the oil and gas industry — long horizontal wells with multiple stimulation stages — have been applied to geothermal by several companies, with demonstration projects producing power and larger installations following. The performance improvements have been rapid, and this is currently the most promising line in the field.

It is also, unavoidably, the same toolkit that carries the seismicity concern. The capability and the hazard arrived together.

What is in the rock

Alongside heat, the engine concentrates matter, and the economic consequence is enormous.

Chapter 1 described how partial melting distils incompatible elements upward. The last stages of that process — magmatic fluids exsolving from cooling intrusions, circulating through fractured rock, precipitating what they carry as they cool — are what makes ore.

Porphyry deposits, formed above cooling intrusions at volcanic arcs, supply the large majority of the world's copper and most of its molybdenum, along with substantial gold. Chile, Peru and Indonesia between them hold much of the global endowment.

Epithermal deposits, formed in shallow volcanic hydrothermal systems, are a principal source of gold and silver.

Volcanogenic massive sulfides are the fossilized remains of seafloor black smokers of the kind Chapter 6 opened with — the mineral deposit left behind when a vent field is buried and preserved. Cyprus was mined for copper in antiquity from exactly such deposits, and the metal is named after the island.

Rare earth elements are concentrated in carbonatites, a genuinely strange class of igneous rock composed dominantly of carbonate minerals rather than silicates. The world's rare earth supply comes overwhelmingly from a small number of carbonatite bodies. There is exactly one carbonatite volcano currently active, in Tanzania, and its lava erupts at a few hundred degrees — cool enough to appear black in daylight rather than glowing — and weathers to a pale crust within days.

Lithium is present in solution in geothermal brines, and there is active work on extracting it from the fluid that is already being pumped for power generation, most prominently at the Salton Sea in California. Co-production of energy and battery metals from the same well is an attractive proposition and is at demonstration stage rather than in routine operation.

The irony is worth naming plainly rather than leaving implicit. The mineral inputs to the energy transition — copper for electrification, lithium for storage, rare earths for magnets — are almost all concentrated by magmatic and hydrothermal processes. In order to stop burning the buried remains of an ancient biosphere, we are mining the products of the volcanic engine. There is a certain symmetry in it.

Putting the carbon back

Now the application I find most elegant, because it runs Chapter 7's thermostat deliberately and at speed.

Basalt is rich in calcium, magnesium and iron. Carbon dioxide dissolved in water forms a weak acid that attacks those minerals and precipitates stable carbonates. That is the silicate weathering reaction from Chapter 7 — the sink half of the thermostat — and in nature it takes hundreds of thousands of years.

At the Hellisheiði power station in Iceland, the CarbFix project dissolves carbon dioxide in water and injects the solution into basaltic rock at a few hundred metres depth.

The result, when it was first properly measured, was the surprise. Mineralization had been expected to take centuries or longer. Monitoring using isotopic tracers found that the great majority of the injected carbon had mineralized within about two years. The reaction, under the right conditions of dissolution and rock chemistry, runs several orders of magnitude faster than the natural process it imitates.

Once mineralized, the carbon is carbonate rock. It is not stored under a seal that might leak; it has become part of the rock, which is about as permanent as geological storage gets.

The approach has been replicated in the Columbia River basalts in Washington State with comparable results, and the Icelandic site is now paired with direct air capture plants, so that carbon pulled from the atmosphere is turned into stone on site.

The constraints are real and should be stated. It needs basalt — though basalt is abundant, in continental flood provinces and across essentially the entire ocean floor, so the resource is not the limitation. It needs water, in quantity, which the original method used a great deal of and which later work has sought to reduce, including versions using seawater. And it needs injection infrastructure at a scale that does not currently exist.

The honest scale check: current injection is measured in thousands of tonnes per year against global emissions measured in tens of billions. This is a demonstration, not a solution. But the mineralization rate finding is genuinely important, because it converts basalt carbon storage from a theoretical curiosity into an engineering problem about deployment.

The one intervention I would endorse

I have been sceptical throughout this book about grand interventions in volcanic systems. Chapter 12 dismissed the proposal to cool a caldera on arithmetic. Chapter 12 also declined to make a recommendation about deliberate climate intervention. This chapter has repeatedly qualified.

So let me be clear about the exception.

Chapter 3 established the discipline's founding constraint: you cannot instrument a magma body, because the instrument is destroyed. And it established the surprising empirical result that on three occasions drills have hit magma unintentionally, and on all three occasions the magma quenched against the borehole, sealed it, and did nothing else.

The Krafla Magma Testbed is a proposal to do it on purpose. Drill deliberately into the shallow rhyolitic magma body at Krafla — whose position is now known, because a drill found it — and build the world's first magma observatory: a set of instrumented boreholes at and near the magma-rock interface, designed from the outset around the certainty that sensors will be destroyed and must be replaceable.

That last point is the conceptual move that makes it possible. The constraint was never really that magma destroys instruments. It was that we designed as though it should not. Accept the loss, make the sensors cheap and the boreholes repeatable, and the constraint becomes an operating cost.

What it would yield is difficult to overstate for a discipline built entirely on indirect inference. Direct measurement of pressure and temperature in a magma body. Direct observation of how the interface between melt and country rock actually behaves. Ground truth against which every tomographic image, every deformation model, every one of the underdetermined inversions in Chapter 4 could finally be calibrated. And, incidentally, the best possible laboratory for supercritical geothermal, since the highest-enthalpy fluids sit exactly where the magma is.

The precedent for safety is three accidents, which is not a proof, and I said as much in Chapter 3. But the project is being designed by people who take that question seriously, at a site where it has already happened three times without incident, in a country with more experience of drilling into hot rock than anywhere else.

If the discipline gets one large intervention this century, this is the one worth having — because unlike every other proposal in this book, it does not try to control the engine. It tries to look at it.

What scales, and what does not

Let me close with an assessment, flagged as my own judgement rather than as reporting.

What scales now. Conventional geothermal in favourable settings, and it is under-built relative to its potential — the Kenyan example should be far more widely copied than it is. District heating from geothermal, which is unglamorous and enormously effective. Mineral co-production. And, separately from everything in this chapter, ground-source heat pumps, which have nothing to do with volcanism but which extract shallow heat anywhere and are among the most consequential decarbonization technologies available.

What might scale. Enhanced geothermal, where recent progress is real and where the constraint is now cost and seismicity management rather than feasibility. Supercritical, where the constraint is materials. Basalt carbon storage, where the science is settled and the constraint is deployment.

What does not scale. Intervening in eruptions beyond the scale of one lava flow and one harbour. Cooling a caldera. Any proposal that treats the engine as something to be managed rather than used.

And the honest summary: geothermal will not be a dominant global energy source, and anyone who tells you otherwise is selling. The geography does not permit it. What it can be is dominant in particular places, firm in a grid that badly needs firmness, and a source of heat rather than only electricity — which is where the largest unexploited opportunity actually sits, since heat is most of what humans use energy for and most of what is hardest to decarbonize.

That is a modest conclusion. It is also, I think, the true one, and this book has tried throughout to prefer the second.

There is one thing left to say, and it is not about us.

Every application in this chapter — every well, every plant, every tonne of carbon turned to stone — depends on the same thing: that there is heat down there, and that it is still coming up.

That will not always be so.

Draft notes — verification status

Standing convention. This chapter contains more numbers than any other and several are commercially or politically sensitive. Pohang in particular attributes a damaging earthquake to a named project on the basis of an official inquiry, and must be stated precisely.

Pending verification — high priority:

Pohang 2017: earthquake magnitude (M5.4/5.5 — sources differ), injuries (~135), damage extent, and the 2019 government investigation's conclusion attributing it to the EGS project. Verify the inquiry's exact finding and wording; this is the most consequential attribution in the chapter.
Basel 2006: induced magnitude (~M3.4), damage claims, project suspension and abandonment.
CarbFix mineralization result: the Matter et al. (2016) finding that >95% of injected CO₂ mineralized within roughly two years; the tracer methodology; injection depth and current annual injection tonnage.
IDDP-1: 2009 drilling, magma encountered at ~2.1 km against a 4.5 km target, steam temperature (~450 °C), estimated electrical output relative to conventional wells, and the cause and timing of the well's failure. Already partly covered in Chapter 3 notes — keep the two consistent.
IDDP-2 / Reykjanes 2017: depth reached (~4.6 km), temperature (~427 °C), and whether supercritical conditions were confirmed.
Krafla Magma Testbed: current project status, funding, and target drilling date. Time-sensitive — update before publication.

Pending verification — standard:

Global installed geothermal electricity capacity (~16 GW) and comparison figures for wind and solar. Update to most recent year available.
Geothermal capacity factors (75–90%) versus wind and solar.
Kenya: Olkaria capacity (~950 MW) and geothermal share of national electricity. The draft says "close to half on some accounting" — this figure is quoted variously and should be sourced with its basis stated.
Iceland: geothermal share of electricity and of building heat; the Reykjavík air quality improvement following the shift from coal.
Larderello: 1904 demonstration (Piero Ginori Conti, five bulbs), 1913 first commercial plant, continuous operation since. Also verify the Dante association, which is traditional and may be apocryphal — flag or cut if unsupported.
The Geysers field pressure decline and wastewater injection recharge.
Geothermal CO₂ and H₂S emissions ranges and their variability between fields.
Water's critical point (374 °C, ~22.1 MPa) and the enthalpy advantage of supercritical fluids (5–10×).
Recent EGS commercial progress (horizontal drilling and multi-stage stimulation; demonstration project outputs). Fast-moving — update before publication.
Porphyry deposits as the source of the majority of world copper and molybdenum; epithermal gold-silver; VMS deposits and the Cyprus etymology.
Carbonatites as the dominant rare earth source; Ol Doinyo Lengai as the only active carbonatite volcano; natrocarbonatite eruption temperature and the black-to-white weathering.
Salton Sea lithium-from-brine: resource estimates and current development stage.
Wallula, Washington basalt CO₂ injection results.
Basalt carbon storage capacity estimates, including oceanic basalt.
Water requirements per tonne CO₂ in the original CarbFix method and in later reduced-water and seawater variants.