The 1.9 MWe Mars power plant that is really a radiator packaging problem

What the work claims

This is a trade study and concept development paper, and everything in it is analysis, not hardware. Under the Mars Transportation Assessment Study, the GRC COMPASS team found that a crewed Mars campaign using a hybrid nuclear electric and chemical propulsion architecture needs a 1.9 MWe-class nuclear electric power system (Figure 1), and the authors' reactor, shielding, conversion, heat rejection, and PMAD trades converged on this reference: a modified terrestrial microreactor derivative burning high-assay low-enriched uranium (HA-LEU, 19.75% enrichment), lithium-cooled, coupled to four 500 kWe supercritical CO2 Brayton units, rejecting heat through pumped-NaK composite radiators, and distributing power at 960 Vac.¹

The headline figures: total system mass of about 25,000 kg, which is about 13 kg/kWe at the full system boundary (reactor, shield, conversion, heat rejection, and PMAD included), and a 3500 kg (14%) mass reduction available if highly enriched uranium (93%) were permitted instead.¹ The paper's own parametric sweep (Figure 3) shows the ceiling was set by packaging: with roughly 2500 m2 the maximum radiator that stows inside the 8.4 m Space Launch System fairing, a 1200 K helium-xenon system tops out at 1.6 MWe, the selected 1200 K supercritical CO2 case at 1.9 MWe, and a 1500 K helium-xenon case at 2.9 MWe, the last rejected as too risky.¹

How it works

The reactor side is a DOE/Oak Ridge design space. Two concepts were evaluated: a fast-spectrum SP-100 derivative with uranium-nitride pin fuel and pumped lithium cooling, and a derivative of the Transformational Challenge Reactor using UN particle fuel in silicon-carbide elements with interspersed yttrium-hydride moderator, lithium-cooled in the preferred configuration.¹ Both were sized at 10 MWt, 1200 K coolant outlet, and two years of full-power life. The mass ordering is blunt: the HEU SP-100 is the lightest reactor at about 2400 kg including fuel, vessel, reflector, instrumentation, and lithium loop; the LEU version of the same reactor is "prohibitively heavy"; the HEU TCR derivative matches the SP-100's reactor mass but its larger diameter forces a 70% heavier shield; and the HA-LEU TCR derivative weighs about 4800 kg, roughly twice the HEU reactor, with the heaviest shield of all.¹ Yet the LEU TCR was selected as reference because its full system penalty, reactor plus shield, is only 3500 kg over the HEU SP-100, a difference the mission design absorbs.¹ At 10 MWt against 1.9 MWe output the reactor carries about 40% thermal margin, implying roughly 19% end-to-end thermal-to-electric efficiency, our arithmetic from the paper's stated figures.

The shield is a dedicated ORNL optimization driven by a 50 rem-per-year crew dose limit. Starting from a conical lithium-hydride/tungsten shield with a 26-degree half-angle that held PMAD electronics to 25 krad and 10^11 n/cm2 at 50 m after two years, ORNL compared four geometries; the masses run 13,800 kg for the full-thickness cone, 4800 kg for a hybrid beryllium/boron-carbide compound, 3500 kg for a lithium-hydride/tungsten compound, and 2800 kg for a compound cruciform that keeps full thickness only along the vehicle centerline and four 26-degree radiator-wing lobes.¹ The cruciform was selected; its radiation maps give 100 Mrad absorbed gamma dose at the Brayton converters, 25 krad at the PMAD electronics, and 3 mrem/hr at the forward crew-habitat face, about 100 rem over two years with credit taken for intervening structure, propellant, and equipment.¹ With shield included, the reference HA-LEU reactor and crew-rated shield together mass about 7600 kg, against about 4100 kg for the HEU equivalent.¹

Conversion and heat rejection are where the study's real thesis sits. The conversion trade favors supercritical CO2 Brayton over helium-xenon because the SCO2 cycle yields about 20% more power for the same radiator area, the decisive currency under the fairing constraint; the four 500 kWe converters total about 2100 kg, but each represents a scale-up of roughly fifty-to-one over the largest relevant NASA hardware, the 10 kWe Brayton Rotating Unit, with the 2 kWe mini-BRU, the 36 kWe Space Station Freedom solar-dynamic converter, and the 100 kWe Prometheus/JIMO unit as the intervening rungs.¹ Legacy HeXe Brayton hardware has about 50,000 hours of BRU testing behind it; SCO2 development has been mostly terrestrial, so the paper assigns the space adaptation work to the SCO2 path explicitly.¹ The heat rejection subsystem rejects about 4 MWt at 375 to 550 K through four pumped-NaK loops and 68 polymer-matrix composite panels of about 4 m by 5 m, roughly 100 kg each, for a subsystem total near 9500 kg; the paper notes the whole 2500 m2 array is about double the International Space Station's roughly 1200 m2 of radiator.¹

The PMAD architecture (Figure 9) is a 2.5 kHz three-phase 960 Vac bus run 50 m from the Brayton alternators to the power electronics, converted locally to 650 Vdc for the Hall-thruster direct-drive units and 120 Vdc for the spacecraft bus.¹ Because Brayton units need constant alternator loading, a pulse-width-modulated parasitic load radiator on each converter can dump the entire 500 kWe output at 550 C, letting the plant run at full power with zero external load; the four load radiators are about 30 m2 each, and each PMAD channel's electronics sit behind a dedicated 20 m2 cold plate rejecting about 15 kWt, roughly 3% of channel power, at 100 C.¹ PMAD including cabling, electronics, and thermal management totals about 5800 kg; the spacecraft spins the Braytons up electrically from batteries and solar arrays through a start inverter.¹

Steelman: the best case for this reference concept

The study's deepest contribution is identifying the true design driver. Everyone debates reactor and conversion technology for nuclear electric propulsion; this paper demonstrates that with near-term launch packaging, the radiator stowage volume is what caps system power, which is why a modest conversion-efficiency edge (the SCO2 cycle's 20%) translates directly into 300 kWe of delivered power (Figure 3).¹ That reframing is exactly the kind of insight a trade study exists to produce, and it is falsifiable: any launch vehicle with a larger fairing or any radiator that stows tighter moves the whole curve.

The LEU decision is equally well argued. Enrichment policy is a geopolitical given, not an engineering variable, and the paper quantifies the cost of that constraint honestly: about 3500 kg, or 14% of system mass, and a mission that still closes.¹ The fault-tolerance architecture deserves note too: four independent Brayton channels each with a dedicated radiator quadrant and the ability to produce partial power on surviving units, plus the ability to motor a failed unit, is a credible answer to the single-point-reactor objection.¹ And the mass figure itself, about 13 kg/kWe at the system level, is competitive with any published high-power space nuclear concept of comparable completeness.

Where a skeptic should push

The most load-bearing assumption is that supercritical CO2 Brayton technology adapts to space at the 500 kWe unit scale. The paper is candid that SCO2 development "has been mostly limited to terrestrial applications," which means the selected architecture's core advantage rests on a technology whose space-relevant-environment test record is thin; the largest space-flown-relevant Brayton experience cited is the 10 kWe BRU with about 50,000 hours of testing, two orders of magnitude below the reference unit.¹ The 1500 K helium-xenon alternative that would deliver 2.9 MWe was rejected for requiring fuel forms and refractory alloys beyond SP-100 experience, so the reference sits deliberately below the performance physics allows.¹ Every number above, 25,000 kg, 13 kg/kWe, the shield masses, is a paper study output resting on terrestrial cost and mass analogies, not a measurement from assembled hardware.

Second, the crew-dose arithmetic leans on credit-taking. The 3 mrem/hr at the habitat face incorporates attenuation by the Brayton engines, boom, PMAD equipment, xenon tanks, and structure, and the two-year integral of about 100 rem sits at double the 50 rem-per-year design requirement that drove shield mass.¹ That is a modeling choice that should be stress-tested against a conservative no-credit case before it anchors a crewed mission. Third, the mission wrapper is uncrewed-spiral and opposition-class with a 30-day Mars stay; the 1.9 MWe figure is specific to that COMPASS mission set (2035-2042 opportunities, LEO aggregation, 760-day round trip), not a universal NEP requirement.¹

What a 25-tonne Mars power plant means for space power

Non-obvious implication: at megawatt scale, packaging beats physics. The paper's own Figure 3 sweep shows delivered power set by how many square meters of radiator fit through the fairing, which means fairing geometry, radiator stowage efficiency, and deployment mechanics are power-system technologies in exactly the same sense the reactor is.¹ Programs that fund reactor and conversion research while treating the radiator as a downstream detail have the driver hierarchy inverted; the 20%-for-free power from SCO2 only exists because it buys back packaging volume.¹

Genuine threat: the reference architecture concentrates risk in one unproven element. The 1.9 MWe figure is unreachable without the SCO2 cycle's radiator-area advantage, so a slip in space-adapting supercritical CO2 convertors does not degrade the concept, it removes the concept, and the fallback is the 1.6 MWe HeXe case or the 1500 K materials program the authors judged too risky to baseline.¹ For mission planners, that is a single-point schedule dependency sitting inside a nominally four-channel fault-tolerant system.

Opportunity: the PMAD pattern generalizes. A medium-voltage, kilohertz-frequency AC bus with local DC conversion, per-channel cold plates sized to reject their own electronics heat, and parasitic load radiators that let generators run at constant full power under arbitrary load is a template for any high-power electric propulsion stage, and nearly all of it is buildable today from terrestrial medium-voltage drive technology.¹ A ground demonstrator of the Figure 9 bus, minus the reactor, would retire more near-term risk per dollar than reactor work, because the bus is where the kilowatts-per-kilogram and the fault tolerance both live.

The bottom line

Established by this study: a complete, internally consistent reference design for a 1.9 MWe crewed-Mars NEP power system at about 25,000 kg system mass exists, with credible subsystem trades and a clearly identified binding constraint (radiator stowage) that makes the result explainable rather than asserted.¹ Not established: any of it as hardware; the SCO2 convertors, the 10 MWt HA-LEU reactor, and the cruciform shield are all pre-hardware, and the crew-dose result is model output with generous structural credit. Confidence is moderate that the mass figure is in the right neighborhood, because its inputs trace to heritage component databases, and low that the reference configuration survives contact with a development program. What would confirm it is a space-relevant-environment test of a 100 kWe-class SCO2 Brayton loop; what would break it is that test failing, because the fallback paths both pay in either power or materials risk.