Subsystem 05 of 12 · ISRU
Lunar Habitat ACT, the Autonomous Construction Testbed·ISRU·Concept definition, revision A·August 2026·See it in the tour

Forty three percent of the ground is oxygen

Lunar regolith is mostly metal oxides. Melt it, pass a current, and the oxygen comes free along with an iron silicon alloy. That single fact is what turns a camp into a base.

43%oxygen by mass
1.8 kgoxygen per day
14 kWplant draw
Contents
  1. What the feedstock actually is
  2. Molten regolith electrolysis
  3. The energy account
  4. What comes out besides oxygen
  5. The plant
  6. Failure modes
  7. What is not yet known
  8. Increment plan
  9. Sources and further reading
Section 01

What the feedstock actually is

In situ resource utilisation (ISRU) means making what the base needs from what is already under it. Apollo, Luna and Chang'e have returned samples from eleven sites, the last from the far side in June 2024, and the composition is well characterised. Mare regolith is basaltic, highland regolith is anorthositic, and both are dominated by oxides. There is no processing step needed to concentrate the resource, because the resource is the bulk material.

OxideMare, by massHighland, by massOxygen fractionNote
SiO₂silica45.4%45.0%53%The bulk of the melt
Al₂O₃alumina14.9%24.6%47%Higher in the highlands
FeOiron oxide14.1%5.7%22%The alloy comes mostly from here
CaOlime11.8%15.8%29%
MgOmagnesia9.2%7.5%40%
TiO₂titania3.9%0.6%40%Mare only, and useful
Oxygen, total44.6%44.3%Call it 43 percent after volatiles
Why this matters more than water ice

Polar water ice gets the attention. The Lunar Crater Observation and Sensing Satellite (LCROSS) measured 5.6 plus or minus 2.9 percent water by mass in the ejecta it threw up from Cabeus, the shadowed floors run from about 25 K at the coldest to below 110 K, and the first attempt to drill for ice on the surface, PRIME 1 (the Polar Resources Ice Mining Experiment) on IM 2 in March 2025, ran its TRIDENT drill (The Regolith and Ice Drill for Exploring New Terrain) through its full range of motion but never reached the ground because the lander was on its side. The grade at any particular spot is unknown. Oxygen in regolith is everywhere, at a known concentration, in material a rover can scoop anywhere on the site. For an outpost that needs oxygen every single day, availability beats grade.

Section 02

Molten regolith electrolysis

Regolith is heated above 1,600 °C until it is a conductive melt, and a current is passed through it. Oxygen evolves at the anode as a gas and metal collects at the cathode. There is no reagent, no consumable, and nothing to ship: the only inputs are soil and electricity.

The competing process, hydrogen reduction of ilmenite, runs at a gentler 900 °C but only attacks the iron titanium oxide fraction, which is a few percent of the soil, and less in the highlands. A 2023 system study that sized whole plants for the three leading processes found hydrogen reduction competitive only where the feed holds 7.5 percent ilmenite or more, which the anorthositic rim does not. It needs hydrogen shipped from Earth and it wastes most of the material it handles. Molten regolith electrolysis (MRE) takes everything.

The third route, carbothermal reduction, heats a spot of regolith with methane present and pulls the oxygen out as carbon monoxide, then as water. NASA's Carbothermal Reduction Demonstration ran a reactor built by Sierra Space in the Dirty Thermal Vacuum Chamber at Johnson in April 2023 and extracted oxygen from simulant, the first time that had been done in vacuum, and Sierra Space ran a standalone automated version in the same chamber in 2024. It is the more mature process today. MRE is chosen here because it needs no carbon loop, because its metal output feeds construction, and because Blue Origin's Blue Alchemist reported silicon, metals and oxygen from simulant by MRE in February 2023, passed its critical design review under a 35 million dollar NASA Tipping Point award in September 2025, and is due to demonstrate autonomous operation in a simulated lunar environment in 2026.

Excavate and screen

A rover scrapes the top 100 mm at the excavation face and screens out anything above 5 mm. The plant takes about 14 kg a day, so one shift a week at a modest 14 kg an hour keeps the hopper ahead of it.

Feed and preheat

Screened fines go to a hopper and down a screw conveyor. Waste heat from the cell walls preheats the charge, which is worth about 18 percent of the total energy.

Melt and hold

The cell holds above 1,600 °C. Getting there is the expensive part; holding is cheap, which is why the plant runs continuously rather than in batches.

Electrolyse

Direct current through the melt. Oxygen at the anode, molten metal at the cathode. The anode material is the hard engineering problem: it has to survive oxygen at 1,600 degrees.

Tap and cast

Oxygen is cooled, dried and compressed to storage. The metal is tapped and cast into ingots, or poured directly into paver moulds for the berms.

Return the slag

What is left is a glassy slag that goes back out as fill or aggregate. Nothing is stockpiled as waste, because on a site with no waste disposal there is no such thing as waste.

Section 03

The energy account

Energy per kilogram of oxygen produced Melting the charge 8.4 kWh/kg sensible plus latent heat Electrolysis, theoretical 4.9 kWh/kg the thermodynamic floor Electrolysis, overpotential 3.1 kWh/kg real cells are not ideal Excavation and handling 0.7 kWh/kg Gas cooling and compression 1.4 kWh/kg Losses and standby 2.2 kWh/kg holding the cell hot overnight 20.7 kWh per kilogram is the process account. Schreiner's reactor model gives about 120 kWh per kilogram at 14 kW, most of it wall loss.
The per item figures are estimates for a well insulated cell. The theoretical minimum for splitting these oxides is a few kilowatt hours per kilogram and everything above it is the cost of getting the material hot and keeping it there. The sourced system figure is Schreiner's parametric model: a 400 kg, 14 kW molten regolith electrolysis system producing about 1,000 kg of oxygen a year from highland regolith, roughly 120 kWh per kilogram, because most of the power of a small cell leaves through its walls. That is why the plant here draws 14 kW continuously rather than in bursts, and why preheating with waste heat and running without interruption both matter so much.
Section 04

What comes out besides oxygen

OutputRateCompositionWhere it goes
Oxygenthe product1.80 kg/day99.2% O₂ after dryingLife support makeup, and extravehicular activity (EVA) recharge
Iron silicon alloycathode metal2.40 kg/dayFe with Si, Al, TiPrinter feedstock, brackets, repair stock
Glassy slagresidue9.20 kg/dayCa and Mg silicatesCast into pavers for the berms
Fines below 20 µmscreened out0.60 kg/dayas excavatedReturned to the face, never stockpiled
Feed14.0 kg/dayOne rover shift covers a week

The alloy is the underrated output. At 2.4 kg a day it is about 880 kg a year of structural metal that did not have to be launched, and it arrives in exactly the place the construction gantry needs it.

Section 05

The plant

1.0 tlanded mass
14 kWpeak draw
3electrolysis cells
1 t/yrrated at 14 kW, Schreiner model

The plant is rated at about 1,000 kg of oxygen a year at 14 kW, the figure Schreiner's model gives for a 400 kg reactor, and budgeted at 1.8 kg a day, about two thirds of that, which covers the 1.65 kg a day life support gap with margin for anode rebuilds. Three cells rather than one, for the same reason everything else here is plural: a cell with a failed anode can be taken offline and rebuilt while the other two carry the load at reduced rate. Anode life is the least certain number in this whole document.

Honest statement of maturity

Molten regolith electrolysis has been demonstrated at laboratory scale with simulant, most publicly by Blue Origin's Blue Alchemist. It has never run on the Moon, never run for years, and the anode problem is not solved. This is the highest technical risk item in the architecture, and the reason increment two carries a full oxygen store as backup rather than relying on the plant from day one.

Section 06

Failure modes

FailureEffectBufferResponse
Anode erosionexpected, gradualCell efficiency fallsTwo other cellsScheduled rebuild. Consumable anodes are carried as spares.
Feed blockagefines bridging in the hopperCell starvesHoursVibrator on the hopper wall, then manual clearing on EVA.
Melt freezepower loss at the cellCell is scrapMinutesCells hold on the essential bus. A frozen cell is drilled out, which is a multi day job.
Oxygen line contaminationcarbon carryoverProduct off spec96 h O₂ storeInline analyser diverts to vent until it is back in spec.
Excavator downthe roverNo feed6 days of screened stockScreened fines are kept ahead, deliberately, for exactly this.
Section 07

What is not yet known

Molten regolith electrolysis has run only on simulant, so anode life, the behaviour of the rim's highland feedstock in the melt, and the split of the energy account between electrolysis and wall loss are all estimates. The 14 kW and the 1,000 kg a year come from a parametric model, not a built machine, and the yield of 1.8 kg of oxygen from 14 kg of feed has not been shown on real regolith. Anode life is the least certain number in this whole document, and it is the reason increment two carries a full oxygen store as backup. Carbothermal reduction is further along, with vacuum tests in 2023 and 2024, and is the fallback if the anode problem does not close. Whether the ice next door is worth mining is unknown until a drill reaches it; PRIME 1 did not.

Section 08

Increment plan

Increment 1, no plant

Oxygen is shipped and stored. The site is surveyed and the excavation face is chosen on composition, not convenience.

Increment 2, one cell

A single cell at partial duty, proving the process and the anode life on the real feedstock rather than on simulant. Oxygen store still covers the crew.

Increment 3, three cells

Full 1.8 kg a day, which closes the life support gap entirely. This is the point the base stops importing oxygen.

Increment 4, metal into construction

The alloy tap is plumbed to the printer feedstock hopper, and the slag goes into paver moulds instead of back to the face.

Sources

Sources and further reading

Every figure in this document traces to one of these. Agency documents and peer reviewed work first, reporting only where it is the primary record.

Heiken, Vaniman and French, Lunar Sourcebook, 1991, hosted by the Lunar and Planetary Institute https://www.lpi.usra.edu/publications/books/lunar_sourcebook/Source for the oxide table
Schreiner et al., A parametric sizing model for molten regolith electrolysis reactors to produce oxygen on the Moon, Advances in Space Research, 2016 https://doi.org/10.1016/j.asr.2016.01.006A 400 kg, 14 kW system for 1,000 kg of oxygen a year from highland regolith; plant sizing and energy
NASA, NASA successfully extracts oxygen from lunar soil simulant, April 2023 nasa.govCarbothermal Reduction Demonstration with a Sierra Space reactor, first oxygen from simulant in vacuum
Sierra Space, carbothermal oxygen production reactor thermal vacuum test at Johnson, 2024 sierraspace.comStandalone automated carbothermal extraction in the lunar environment chamber
Blue Origin, Blue Alchemist technology powers our lunar future, February 2023 blueorigin.comMolten regolith electrolysis on simulant at scale; NASA Tipping Point award, July 2023, for a demonstration by 2026
NASA, NASA's lunar drill technology passes tests on the Moon, 2025 nasa.govPRIME 1 and TRIDENT on IM 2: full range of motion shown, no sample because the lander was on its side
Colaprete et al., Detection of water in the LCROSS ejecta plume, Science, 2010 https://doi.org/10.1126/science.11869865.6 plus or minus 2.9 percent water by mass at the Cabeus impact site
Williams et al., Seasonal polar temperatures on the Moon, JGR Planets, 2019 https://doi.org/10.1029/2019JE006028Diviner: cold traps below 110 K, and about 25 K at the coldest
State Council of China, Chang'e 6 returns first samples from the far side of the Moon, June 2024 https://english.www.gov.cn/news/202406/25/content_WS667a644dc6d0868f4e8e8864.htmlThe eleventh sample return site
ISECG, In Situ Resource Utilization Gap Assessment Report, April 2021 globalspaceexploration.orgAgency assessment of oxygen from regolith: complementary work at demonstration scale, and the gaps still open
Blue Origin, Blue Alchemist hits major milestone toward permanent and sustainable lunar infrastructure, September 2025 blueorigin.comCritical design review complete; autonomous demonstration in a simulated lunar environment due in 2026
Guerrero Gonzalez and Zabel, System analysis of an ISRU production plant: extraction of metals and oxygen from lunar regolith, Acta Astronautica 203, 2023 https://www.sciencedirect.com/science/article/abs/pii/S0094576522006579Hydrogen reduction, molten regolith electrolysis and molten salt electrolysis sized as whole plants; hydrogen reduction competitive only above 7.5 percent ilmenite
Schlüter and Cowley, Review of techniques for in situ oxygen extraction on the Moon, Planetary and Space Science, 2020 https://www.sciencedirect.com/science/article/abs/pii/S0032063319301825The process survey behind section 02
Modeling energy requirements for oxygen production on the Moon, PNAS, 2025 https://www.pnas.org/doi/10.1073/pnas.230614612224.3 kWh per kilogram of liquid oxygen for an end to end ilmenite reduction chain, the independent check on the energy account