Why vacuum makes this hard in both directions
On Earth almost all waste heat leaves a building by convection. In vacuum there is no convection, so the only path out is radiation, and radiation scales with the fourth power of absolute temperature. That sounds generous until you notice it also means a radiator running near room temperature is a very poor emitter.
The base has to reject about 46 kW at peak. A radiator at 20 °C with an emissivity of 0.92 radiating into a 3 K sky sheds about 385 W per square metre, before you subtract what it absorbs back from the ground and from the low sun. That is why the radiators on this design are large, vertical, and pointed at the sky rather than at the surface.
The other direction is worse. Whenever the sun drops behind the rim, and through the polar winter when the ridge itself sits at 50 to 70 K, every square metre of habitat skin is a heat leak. Survival heating through a dark spell costs power the array is not making.
The heat budget
The loop
No ammonia inside the pressure boundary, ever. The interface heat exchanger (HX) sits in the connector, outside the habitable volume, and is the only place the two fluids meet. This is the same rule the International Space Station follows, for the same reason.
Radiators
The radiators are mounted vertically on the flanks of each module. At the equator a horizontal radiator spends the day looking at ground near 120 °C and loses much of its output to it. Here the ground is cold, near 200 K on the sunlit ridge on a summer average and 220 to 300 K at the Diviner summer maximum, and far colder in shadow, so a vertical panel sees sky and cold ground and works well. The catch is the sun. It sits within about 1.5 degrees of the horizon and circles the whole sky once a lunation, so no fixed vertical face is edge on to it all month. Keeping the sun off the panels is an open design point, listed below.
Each module carries a pair sized for its own load, rather than a single central farm. That costs about 15 percent more area in total and buys the ability to lose any one module without losing the cooling for the rest.
| Element | Area | Rejects | Radiator temperature | Note |
|---|---|---|---|---|
| Core radiatorstwo wings | 34 m² | 13.2 kW | 20 °C | Sized for crew, avionics and galley |
| Greenhouse radiatorstwo wings | 32 m² | 12.6 kW | 22 °C | Matched to the diode load |
| Laboratory radiatorstwo wings | 14 m² | 5.4 kW | 18 °C | Includes the freezer reject |
| Airlock and cupolaone wing each | 10 m² | 3.7 kW | 20 °C | |
| ISRU plant radiatorstandalone | 28 m² | 13.1 kW | 95 °C | High grade heat, so a smaller panel does more |
| Total | 118 m² | 48.0 kW | 4 percent margin on the 46 kW peak |
The ISRU radiator, serving the in situ resource utilisation plant, is the interesting one. Because its loop runs at 95 °C rather than 20, it rejects nearly as much heat as the core does from a smaller panel. At 95 °C and the same emissivity a panel sheds about 960 W per square metre, so 28 square metres could carry roughly twice the 13.1 kW listed; the surplus is margin for dust and for sun on the face. Radiating at the highest temperature the process will tolerate is the cheapest thing you can do in a thermal design, and it is why the plant does not share the habitat loop.
Surviving the night
A frozen coolant line is unrecoverable in the field. Every external run is trace heated, the heaters are on the essential bus, and the phase change buffer holds enough energy to keep the lines liquid for eleven hours with no power at all. That eleven hours is the single number the night architecture is designed around.
Phase change buffer
A salt hydrate buffer sits between the loop and the radiators. It absorbs the peaks so the radiators can be sized for the average rather than the maximum, and it holds the base through short losses of power without the lines freezing.
The material is Glauber's salt, sodium sulfate decahydrate. It melts at 32.4 °C and the literature puts its latent heat at 239 to 254 kJ per kilogram; the buffer is sized here at 180 kJ per kilogram, below that, which leaves margin. Even so it holds far more than the same mass of water would as sensible heat over the same range.
Dust, and why it matters more here than anywhere else
A radiator works by having a high emissivity and a low solar absorptivity. Regolith dust has the opposite of both. A dusted radiator absorbs more sunlight and emits less of its own heat, so contamination attacks the system from both ends at once.
On Apollo 16 and 17 the rover batteries ran over their temperature limits because dust on their radiators could not be brushed off, and John Young said afterwards that he regretted the time spent trying. On a ten year outpost that is not a nuisance, it is a design driver. The radiators here are vertical, which sheds most settling dust, and carry the same electrodynamic clearing the solar array uses, which NASA demonstrated on a radiator surface on Blue Ghost in March 2025. The table below is a design estimate anchored to the clean figure, not a measurement; the Apollo record gives the direction and the seriousness, not these numbers.
| Condition | Absorptivity | Emissivity | Net rejection | Loss |
|---|---|---|---|---|
| Cleanas installed | 0.09 | 0.92 | 385 W/m² | none |
| Light dustone lunation | 0.14 | 0.88 | 353 W/m² | 8% |
| Moderatesix lunations, uncleared | 0.26 | 0.79 | 283 W/m² | 27% |
| Heavytwo years, uncleared | 0.41 | 0.68 | 203 W/m² | 47% |
Failure modes
| Failure | Effect | Time to impact | Response |
|---|---|---|---|
| Pump package failsone of two strings | Half the loop capacity | Hours | Second string carries the habitat. ISRU and greenhouse shed. |
| Ammonia leakexternal loop | Loss of external rejection | 2 h | Isolate the affected branch. Phase change buffer holds while the crew reconfigure. |
| Radiator puncturemicrometeorite | Local loss | Slow | Each wing isolates. Losing one wing costs about 12 percent of capacity. |
| Trace heater failurein a dark spell | Line freezes | 40 min | Redundant heaters on the essential bus. The line is drained if both fail. |
| Interface HX foulinggradual | Reduced transfer | Months | Scheduled flush. The reason the internal loop is water and not something exotic. |
What is not yet known
The sun circles the horizon once a lunation within about 1.5 degrees of it, so no fixed vertical radiator is edge on to it all month; how the panels are shaded or turned has not been decided, and it bears on the 118 square metre figure. The longest dark spell at the site has not been computed for the exact position. The 240 m resolution Lunar Orbiter Laser Altimeter (LOLA) model gives the best Shackleton rim site about 1.5 days and the 20 m model gives the best rim clusters 65 to 66 hours, so the 11 hour buffer and the 6.2 kW survival load have to be shown to cover a spell of days on the fission unit alone. The dust table is a design estimate; the Apollo record gives the direction, not these numbers. The salt hydrate buffer assumes its latent heat holds over hundreds of cycles, which has not been shown in a flight unit. The power conversion loss here, 4.4 kW, does not match the 1.1 kW the power document carries, and the two have to be reconciled.
Increment plan
Increment 1, habitat loop only
Water inside, ammonia outside, 90 square metres of radiator sized for the habitat alone, the four module rows in the table above. No ISRU load yet, so the peak is about 35 kW.
Increment 2, plant loop and buffer
The high grade ISRU loop at 95 degrees, its own radiator, and the phase change buffer that lets the whole system ride out a power interruption.
Increment 3, dust mitigation
Electrodynamic clearing on every radiator face. This is the step that stops the uncleared degradation the table above estimates.
Increment 4, capacity for construction
The sintering gantry adds 9 kW of intermittent load with a very high grade reject. It gets its own small panel rather than being plumbed into anything.
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.