Belt Power & the Fissile Leash

Solar still works

Solar Power at Ceres: 13% of Earth’s Flux, and Three Missions That Ran on It

The standard claim is that solar fails past Mars. The flight record says otherwise.

Sunlight falls off as the inverse square of distance. At Ceres’s mean 2.77 AU the flux is 1,361 ÷ 2.77² = 177 W/m², and Ceres’s eccentric orbit swings that between 153 and 210 W/m² over its 4.6-year year — a 37% peak-to-trough change any array has to be sized for at the worst point.

That is a large reduction and not a disqualifying one. Dawn carried arrays rated about 10 kW at Earth and ran its ion engines and instruments on roughly 1.3 kW at Ceres from 2015 to 2018. Psyche launched with arrays of about 21 kW at Earth and is flying solar-electric to a body near 3 AU now. Juno runs on solar at Jupiter, 5.2 AU, where the flux is 50 W/m² — its three wings produce around 14 kW at Earth and about 500 W in Jovian orbit.

Per square meter, the comparison with the ground is closer than intuition suggests. A sun-pointing array at Ceres receives 177 W/m² continuously, about 1,550 kWh/m² a year, with no night, no weather, no atmosphere and no dust. A fixed-tilt panel in Arizona — among the best sites on Earth — receives roughly 2,400 kWh/m² a year. The cold also helps: photovoltaic efficiency rises as cells get colder, and an array at belt temperature gains a large fraction of what an Arizona panel loses running at 60°C. Per unit of area, belt solar is in the same range as terrestrial utility solar. Observed

One refinement cuts the other way. Cells operated at low intensity and low temperature — the LILT regime — can suffer defect-driven losses that offset part of the cold-temperature gain, and those losses vary by cell type. The net effect is roughly a wash against the aphelion penalty, which is why the average figure is defensible and the aphelion figure is the one that sizes the hardware.

Where solar loses

The Penalty Is Array Mass, and It Scales Straight Up With Power

Area is not the problem out here. Kilograms per kilowatt are.

A modern deployable array at 1 AU — the roll-out designs flying on the ISS and on commercial satellites — runs around 6.6 kg per kilowatt of output. Move it to Ceres and every kilowatt needs 7.7 times the area, so the specific mass multiplies accordingly:

6.6 kg/kWe × 7.7 = ~51 kg/kWe at mean distance
6.6 kg/kWe × 8.9 = ~59 kg/kWe at aphelion, which is where arrays must be sized

That figure is array-only — no structure for a large field, no power management, no storage. It also scales linearly: 1 MWe of belt solar is on the order of 60 tonnes of array, 10 MWe is 600 tonnes, and nothing about building bigger makes each kilowatt cheaper in mass. At prospecting scale that is perfectly acceptable. At industrial scale it is the reason the belt goes nuclear. Modeled

The crossover

Solar Wins Below 100–300 kWe, Nuclear Wins Above It

The line is set by how reactors scale, not by how far the belt is from the Sun.

A reactor does not scale linearly. Shielding, controls and the core itself carry large fixed masses, so small reactors are heavy per kilowatt and large ones are light. Solar is flat at roughly 51 to 59 kg/kWe in the belt. The crossover is simply the power level where the reactor curve falls through the solar line.

Reactor figures are system-level and span flown hardware, program targets and studies; the high-power rows are Modeled. The Kilopower range reflects the 2018 KRUSTY ground test at 1 kWe and design extrapolation to 10 kWe.

That splits the belt’s timeline cleanly. Prospecting, survey and sample return run on solar — they need kilowatts, and every flown belt mission proves it. Extraction, refining and manufacturing run on nuclear — a single tonne-a-day electrolysis plant already needs 263 kW continuous, and a working operation needs megawatts. The exploration phase does not have to wait for space reactors to mature, which buys it roughly a decade of schedule.

Heat is the real gate

Megawatt Reactors in Space Are a Radiator Problem

Every watt of electricity comes with three watts of waste heat, and vacuum only lets it leave by radiating.

Converting reactor heat to electricity runs at roughly 20 to 30% efficiency, so a 20 MWe plant rejects 50 to 80 MW of waste heat. On Earth that goes into a river or a cooling tower. In vacuum the only route is thermal radiation, governed by the Stefan-Boltzmann law:

q = ε σ T⁴
at 500 K: ~3.2 kW/m² per side  ·  at 800 K: ~21 kW/m² per side

Because rejection goes as the fourth power of temperature, running the radiator at 800 K instead of 500 K shrinks it by a factor of about 6.6. Radiator area, and therefore radiator mass, sets the specific mass of every high-power system — which is why the belt’s fast transport ship, analyzed on Belt Corridors & Transfer Windows, is gated by rejecting heat at 800 K and not by the reactor core. Materials, working fluids and pumps that last years at that temperature in vacuum are the unretired development item.

The same arithmetic runs through the surface plant. A 20 MWe industrial site on Ceres needs hundreds to thousands of square meters of radiator depending on its operating temperature, and the belt’s cold background helps only a little — radiated power depends on the radiator’s own temperature, and 160 K surroundings barely change the net. Modeled

No uranium out here

The Belt Has No Uranium Ore, and Geochemistry Guarantees It

Five to six orders of magnitude below terrestrial ore, and no process on the belt concentrates it.

Uranium ore on Earth runs from about 1,000 parts per million at the low end of economic grade to 20,000 ppm and beyond in the richest Canadian deposits. Those concentrations are the product of billions of years of crustal recycling, granite formation and groundwater chemistry — uranium is soluble when oxidized and precipitates where conditions change, so water moving through rock gathers it into deposits.

None of that happened in the belt. Chondritic meteorites, the closest thing to average belt material, carry uranium at roughly 10 parts per billion. That is 100,000 to 2 million times below workable terrestrial ore. Thorium, the other fission fuel, sits near 40 ppb for the same reason.

Vesta is the one hope worth checking and it fails. Differentiation carries uranium into the crust with the other incompatible elements, and Vesta’s basaltic crust is indeed enriched — by a factor of a few. A factor of a few against a gap of 100,000 does not close anything. Ceres’s brines move dissolved salts to the surface, and there is no evidence that they concentrate uranium to anything near ore grade either.

Every gram of fissile material in the belt is an import from Earth. That is a statement about geochemistry, not politics or technology, and no future development changes it. Observed

The leash binds growth

Reactor Fuel Can Be Stockpiled, So the Leash Limits Growth, Not Existence

Thirty years of fuel for a 20-megawatt economy fits in one Starship.

How much fuel does a belt industrial economy actually need? Take a 20 MWe operation running at 25% conversion efficiency, so 80 MW of reactor heat, for 30 years:

80 MW × 30 yr × 365 d = 876,000 MW·d
at 30,000–90,000 MW·d per tonne of fuel burnup
= ~10 to 30 tonnes of high-assay low-enriched uranium

High-assay low-enriched uranium, or HALEU, is uranium enriched to between 5 and 20% U-235 — above ordinary power-reactor fuel and below weapons-usable material, and the fuel nearly every compact reactor design now specifies. Ten to thirty tonnes of it is a single Starship payload. U-235 has a half-life of 704 million years, so the fuel does not spoil sitting in a depot.

That changes what the dependency means. A belt operation that brings its fuel with it can run for its planned life without another delivery; Earth cannot switch it off. What Earth controls is expansion — every new reactor needs new fuel, and every new fuel shipment requires an enrichment plant on Earth, an export license and a launch. The fissile leash binds growth, not existence. And anything running below the 100 to 300 kWe solar crossover sits outside the leash entirely.

The enrichment side is concentrated. For most of the past two decades Russia’s Tenex was the only commercial HALEU supplier; the United States banned Russian uranium imports in 2024 with waivers running into 2027, and Centrus at Piketon, Ohio produced the first domestic HALEU in late 2023 at demonstration scale. China’s CNNC runs a fully domestic enrichment chain. Whoever controls HALEU export controls how fast anyone else’s belt presence can grow, and at present that is a short list. Modeled

Beamed power

Beaming Power to the Belt Means Building the Solar Farm Twice

The receiver has to be as large as an array that would collect the same power from sunlight directly.

The proposal recurs: collect sunlight in the inner system where it is strong, convert it to a laser or microwave beam, and send it out to the belt. Diffraction sets how tightly a beam can be focused, and for a transmitter of diameter D_t and a receiver of diameter D_r separated by distance L:

D_t × D_r ≥ 1.22 λ L

At a 1-micron infrared wavelength over 1 AU, the product works out to about 183,000 m² — two apertures each around 430 meters across. Microwaves at centimeter wavelengths make it roughly ten thousand times worse.

The receiving end is the problem. A 430-meter receiver in the belt has an area of about 145,000 m². Cover that same area with solar cells and at 177 W/m² it collects around 26 MW of sunlight with no transmitter, no beam and no pointing link at all. A beamed-power system builds a large collector in the inner system and then a second one of comparable size in the belt; the belt half alone already does the job. The idea fails on that economics, and the diffraction limit is only the reason both halves have to be so big. Modeled

Who is building reactors

Space Reactors: One Flown by the US in 1965, Dozens by the Soviet Union

The history is thin, the current programs are lunar, and nothing on anyone’s manifest reaches megawatt class.

The United States has flown exactly one reactor: SNAP-10A in 1965, producing about 500 W for 43 days. The Soviet Union flew more than 30 — the BUK reactors on its RORSAT ocean-surveillance satellites and two TOPAZ units in 1987 — which remains the largest operational record anyone has. NASA’s Jupiter Icy Moons Orbiter, a 200 kWe nuclear-electric mission under Project Prometheus, was cancelled in 2005. The KRUSTY test in Nevada in 2018 ran a 1 kWe Kilopower prototype at full power and is the most recent American reactor demonstrated for space.

Current programs aim at the Moon. NASA’s Fission Surface Power effort targeted 40 kWe and was redirected in 2025 toward a reactor of around 100 kWe for the lunar surface by about 2030. Russia and China announced in March 2024 a joint plan to place a nuclear power station on the Moon between 2033 and 2035 in support of the International Lunar Research Station, and Chinese institutes have reported design work on megawatt-class space reactors. Both are surface-power programs at or below the belt’s crossover.

The belt needs two things neither program delivers: a propulsion reactor at 10 to 20 MWe that makes 300-day crewed transits possible, and industrial surface reactors at 1 MWe and up. Neither is funded anywhere. Every belt mission currently in flight or in development — Psyche, Tianwen-2 on its way to 311P/PANSTARRS, AstroForge, Karman+ — runs on sunlight, and for what they are doing, sunlight is enough. Observed

Record

Provenance and Data Sources