What transfers
The geometry, and on better terms.
Both Mercurian poles have permanently shadowed floors holding 100 billion to a trillion tons of ice with illuminated high ground nearby, and the value hierarchy holds: power and thermal management and launch on the high ground, water extraction and cold storage on the floor, a transit corridor between them.
The corridor transfers on better terms than it has on the Moon. Mercury’s obliquity of 2.04 arcminutes means no seasonal variation at all, so an alignment surveyed once is aligned forever. And thermal cycling is set by a 176-day solar day rather than the Moon’s 29.5, which is an order of magnitude fewer cycles over any design life — a real advantage for a rail, a pipeline or a power line.
The floors are 100 K, not 40
A cold trap with hot walls.
Lunar floors reach 25 to 40 kelvin, colder than Pluto’s surface, and that is the physical basis of the cold-compute argument at Shackleton. Mercury’s coldest floors are modeled below 100 kelvin — cold enough to hold ice for billions of years, and nowhere near 40 — because a crater floor in permanent shadow still sees its own sunlit walls, and at Mercury that surrounding terrain is receiving 6.7 times the flux.
The cold trap is warm by lunar standards for exactly the same reason the illuminated ground is valuable. Superconducting logic at 77 kelvin has no margin against a 100-kelvin floor where at Shackleton it has 37 kelvin of it. Water, cryogenic propellant storage and shielding all transfer intact; the cold-compute half of the Shackleton argument does not.
The same duty cycle, 10 times the storage
Outage length scales with the solar day.
The best-documented high-illumination site on Mercury receives sunlight about 82 percent of the time, at 89°S, 50.7°E. No true peak of eternal light has been confirmed at either pole. Shackleton’s rim reaches roughly 90 percent. 82 against 90 looks like a small difference and is not, because duty cycle is the wrong metric.
What sizes the storage is the longest outage, and outage length scales with the solar day.
| The Moon | Mercury | |
|---|---|---|
| Solar day | 29.5 d | 175.94 d |
| Illumination at the best site | ~90% | ~82% |
| Longest outage | ~3 d, usually shorter and more frequent | 32 d, one contiguous block |
| Storage for 100 MW through it | ~7 GWh | 76 GWh |
| Battery mass at 200 Wh/kg | ~35,000 t | 380,000 t |
Because the Sun tracks the horizon exactly once per solar day, terrain shadowing at Mercury tends to come in one contiguous block rather than in frequent short gaps. Storing 100 megawatts through 32 days is 380,000 tons of battery. That is not a solution; it is a disqualification, and it is what forces the ring.
The ring
Three to five sites, tied together.
Mercury’s polar power system is a ring of three to five collector sites distributed in azimuth within 20 kilometers of the pole, tied together by a power loop of 60 to 120 kilometers, sized so that shadowing at any one site is covered by the others. Each site is a heliostat field feeding a solar-thermal receiver — not photovoltaic, for the reasons on the sorting rule page.
That loop is the largest single structure in the Mercury baseline and it is built first, because nothing else runs without it. The Moon’s polar architecture is a point with a corridor. Mercury’s is a ring.
Redundancy is N+2 rather than N+1, because outages are month-long and simultaneous darkness at two adjacent sites is the design case rather than a contingency.
The ring assumes that ring-for-illumination and ring-for-survivability are the same geometry, and nobody has checked. Azimuthal distribution against terrain shadowing and dispersal against a counter-infrastructure weapon are different optimizations that happen to produce similar-looking layouts. Whether they coincide is unresolved.
Go north, probably
A magnetic argument against a survey artifact.
MESSENGER established that Mercury’s magnetic dipole is offset 484 ± 11 kilometers north of center, with a moment about one percent of Earth’s and a tilt under one degree. The consequences are large and one-sided.
| North | South | |
|---|---|---|
| Surface field at high latitude | 3.4× stronger | Baseline |
| Open-field polar cap area | Baseline | ~4× larger |
| Cusp proton precipitation | Baseline | ~4× higher |
| Best-documented illumination site | Not yet identified | 82% at 89°S, 50.7°E |
| MESSENGER coverage | Excellent | Worse |
The tension is a mapping gap rather than a physics gap. MESSENGER’s orbit was highly eccentric with periapsis in the north, so the absence of a documented northern site is partly an artifact of what was observed. BepiColombo settles it from 2027, and siting Mercury’s outpost is blocked on a measurement that will be made within two years.
The recommendation, stated conditionally: north, unless the northern illumination maps come back materially worse. The magnetic argument is permanent and physical; the illumination argument is a survey result about to improve.
Whatever the pole, the radiation design case is not the mean environment. It is the compressed-magnetosphere event, and the shelter is regolith over a shadowed floor, which the site already has.
The build sequence
Reconnaissance now, scale in the 2060s.
| Phase | Window | What happens | Status |
|---|---|---|---|
| Reconnaissance | 2026–2032 | BepiColombo delivers polar illumination maps, altimetry, magnetospheric characterization and surface chemistry. Its deliverables here are a siting decision and an ice-grade estimate | Funded and flying |
| The power ring | 2040s | Solar-electric tugs deliver three to five collector sites within 20 km of the chosen pole on a 60–120 km loop. Ice-grade determination at the floor. A regolith campaign on whether magnesium-bearing film feedstock can be made locally — a gate, not a task | Unfunded |
| The pilot line | 2050s | Film production at pilot scale. A mass driver test article at a few hundred meters. The first orbital collector deployed and allowed to sail itself into station. The first beam demonstration at Venus or Earth range, where apertures are under 110 m | Unfunded |
| Scale | 2060s onward | Mass driver at full length and 1.87 million tons a year. Beam service to Ceres, then to the Jovian system. The collector program as a 30-year build | Unfunded |
Three of the four technologies this needs are being built by somebody else for their own reasons — high-power solar-electric propulsion, off-Earth thin-film manufacture, and phased optical arrays with sub-wavelength metrology. The fourth gate is Mercury’s own and has no sponsor.
Four failure modes
Specific to this destination.
Heliostat attrition is the dominant maintenance load. Everything of value sits in permanent shadow and the only equipment in sunlight is the mirrors, taking 9,000 watts per square meter with a 2.3-fold annual swing plus micrometeoroid flux at a heliocentric distance where dust density is high. Budget continuous mirror replacement as a first-class line item rather than as maintenance.
Dust may be worse than the Moon’s. Mercury’s regolith is mature and fine, and its photoelectron sheath is driven by 6.7 times the ultraviolet flux, so charging and levitation should be more vigorous than the lunar case that has already caused so much trouble. Nobody has measured it.
The magnetospheric compression event is the radiation design case. Mercury’s field is one percent of Earth’s and has been observed pressed to the surface by extreme coronal mass ejections, with internal response timescales of one to two minutes. The protection is statistical and fails episodically.
Beam loss is a scheduling failure that is still a failure, and it is covered on the departure corridor page.
No rescue and no crew, which for once simplifies rather than complicates. With nobody on the surface, the no-rescue principle reduces to a spares-and-autonomy problem with an excellent supply cadence — Mercury’s spares queue is seven to eight months, the shortest outside cislunar space. Graceful degradation applies to the power ring above all.