NASA traded thermophotovoltaics against Brayton for Mars power

What the work claims

The paper is a comparative architecture study, not a hardware result. Working from the master equipment list of an Advanced Concepts Office internal study of a 3.1 MWe nuclear electric propulsion element for a crewed, opposition-class Mars mission, the authors re-traded the heat engine: a baseline four-unit Brayton power system versus a thermophotovoltaic system in which 1,050 small channels lined with tungsten emitters radiate onto adjacent cells at 27 percent conversion efficiency. Their finding: electrical power divided by system dry mass for the TPV element came out roughly 12 percent lower than for the Brayton element, implying lower acceleration and worse performance, while the TPV power system itself was about 4 metric tons lighter and the radiator grew by about 22 percent in area and mass.¹

Weight this as what it is: a paper trade using modelled component performance, published alongside the authors' own caveat that a more detailed trade is planned. Nothing in it has been built, thermal-vacuum tested, or flown.

How the two engines work

A thermophotovoltaic cell is a photovoltaic cell tuned to infrared: a hot emitter glows, the cell absorbs the band of the spectrum its bandgap admits, and current flows. The physics constraint that drives everything here is that cell efficiency falls as cell temperature rises, while radiator mass falls as radiator temperature rises. In a reactor system the cell sits between those two curves, and the paper assumes a 2673 K tungsten emitter feeding cells held to 600 K, giving a modelled system efficiency of about 27 percent. That figure is not measured: it is the Nature-published model of a research cell that reached 40 percent only at a 2400 K emitter with the cell at 300 K, degraded analytically to the flight-relevant 600 K cell temperature.¹² The researchers behind that cell see a path to 56 percent at roughly 2500 K emitters, again under laboratory conditions far cooler than a flight cell would run.²

The Brayton side is the mature incumbent. The baseline vehicle carries four 788 kWe Brayton units fed by a common 10 MWt reactor through a molten lithium loop, generating 960 VAC that transformer-rectifier units convert to 650 VDC for the Hall thrusters and 120 VDC for the bus. Because a Brayton runs at a fixed shaft speed, a fast load drop (one thruster switching off) would overspeed the turbine, so a shunt regulator dumps the imbalance into a parasitic-load radiator running at 1000 to 1200 K.¹ The paper also rehearses the lineage: 1960s Brayton Rotating Units that hit 32 percent efficiency and accumulated about 50,000 hours of endurance testing, a Space Station Freedom design that completed ground tests but never flew, and the cancelled Jupiter Icy Moons Orbiter, whose 2 kWe Brayton power conversion unit was tested before the mission died.¹

The TPV vehicle replaces all of that with channel modules whose cell strings behave as current sources. An array regulation unit meets constant-voltage loads not by spinning anything faster but by switching strings between the loads and the parasitic radiator. The same physics then collects its toll: at 27 percent efficiency against the Brayton's higher figure, more reactor heat becomes waste heat, and the Stefan-Boltzmann scaling of the radiator (area proportional to waste heat divided by the fourth power of radiator temperature) pushes radiator area and mass up by about 22 percent.¹ The reactor, shielding, and PMAD changes roughly wash out the 4 metric ton power-system saving.

One historical footnote deserves mention because it explains why TPV keeps resurfacing: radioisotope TPV studies, including NASA contracts from 2003, indicated a system at 20 percent efficiency could reach 16 W/kg including radiator mass against under 4 W/kg for competing radioisotope options, and TPV cells have since been stress-tested under curium-244 for over 120 days with less than 1 percent per year degradation inferred in that low-power isotope environment.¹³ A 2500 K gray-body emitter, the paper notes, radiates an areal power density about 1,300 times the solar flux at 1 AU, which is why TPV advocates keep calling it compact.¹

The strongest case for taking TPV seriously

The steelman starts with the observation that the 12 percent deficit is narrow and rests on one cell-temperature assumption. The 27 percent system figure comes from cells that, as the paper itself concedes, were never designed to run at 600 K; cells engineered for that temperature would narrow the gap or erase it.¹ If the cell is the whole game, then the game is being played in a discipline, semiconductor photovoltaics, that improves on commercial timelines rather than aerospace ones.

Second, the qualitative column is genuinely loaded. TPV has almost no rotational inertia, which makes artificial-gravity spacecraft far easier to design; its cells are small, flexible, and modular, so scaling to a new power level is a re-stringing exercise rather than a new turbomachinery development; manufacture can start immediately on existing lines; and the same emitter-and-cell channels could plausibly wrap a nuclear thermal propulsion reactor, a synergy a Brayton loop cannot offer.¹

Third, the failure modes are boring in the good way. A Brayton at 36,000 rpm is a single point of mechanical failure with precision bearings in a reactor bay; a TPV channel is a diode. For a 25,000-hour Mars transit, the reliability arithmetic favors solid state even before mass enters it.

Where a skeptic should push

The single most load-bearing assumption is the 600 K cell temperature with a 27 percent efficiency attached. Everything measured about the impressive cells behind that number was measured at 300 K. The step from a 40 percent laboratory cell to a 27 percent reactor converter is a model, and the paper says so in its own results section.¹ If engineering for 600 K operation costs more efficiency than the model assumes, the 12 percent deficit widens quickly, because the radiator penalty compounds it.

Radiation is the second push, and it is where the paper is most honest. The curium-244 stress test that produced the sub-1-percent-per-year figure is a low-power isotope environment; the paper states plainly that TPV radiation tolerance is not well characterized for the environment near a fission reactor, quotes an anticipated dose of 10 to 6,000 rad per hour at the power conversion system depending on placement, and calls for neutron and gamma testing at shielded-reactor flux levels that has not been done.¹ Photovoltaics near an unshielded-megawatt-class neutron source is exactly the regime where pretty cells have historically misbehaved. The 533 kg shielding addition carries a 30 percent margin and comes from a simplified composite-shield estimate, with a three-dimensional analysis deferred to a companion report.¹

Finally, TRL honesty cuts both ways in this paper. The Brayton side has real hardware history: 50,000 hours on BRU units, a tested 2 kWe conversion unit. The TPV side has a modelled system and somebody else's room-temperature cells. Presenting this trade as "TPV nearly matches Brayton" would launder a simulation into a near-capability; the correct statement is that a favourable cell-temperature future exists in which it might.

What TPV heat engines mean for space power

The non-obvious implication is about where the engineering actually lives. In the Brayton vehicle, a dedicated carbon-carbon radiator at 1000 to 1200 K exists purely to eat electrical load transients and keep a shaft at speed; in the TPV vehicle, the entire throttling function is a switch matrix rerouting current-source strings to a parasitic radiator.¹ Either way, at megawatt-electric scale, power management is substantially a thermal design problem: the regulator is a radiator with electronics attached. Anyone planning high-power electric propulsion buses should read the PMAD and thermal subsystems as one coupled mass line, because that is where both heat engines spend their tonnes.

The genuine threat is unpriced degradation. If TPV cells near a 10 MWt reactor degrade at anything like typical fast-neutron damage rates for photovoltaics rather than the measured isotope-environment rate, the 12 percent paper deficit grows year over year over a 25,000-hour mission, and the modular-string advantage becomes a maintenance schedule nobody budgeted. Until the reactor-flux testing the paper requests exists, any TPV-based Mars vehicle mass estimate carries an open radiation term of unknown sign.¹

The opportunity is the one the paper lists and the field keeps underusing: because TPV cells are made on commercial semiconductor lines and the converter is a collection of identical channels, a space nuclear program could start qualifying converter hardware now, in parallel with reactor development, instead of after it. For a NEP architecture that must also integrate with nuclear thermal propulsion, a converter that wraps the reactor and shares its thermal infrastructure is a genuinely different path from the turbine-in-a-bay school.¹

The bottom line

Calibrated to what the source supports: I have high confidence that Brayton beats this particular paper TPV on power to dry mass for Mars-class NEP, because the result follows from the authors' own stated assumptions and efficiency model. I have moderate confidence the gap narrows if cells are purpose-built for 600 K operation; that is a development program, not a fact. And I have low confidence in any specific TPV-NEP system mass until reactor-flux radiation data exists. What would confirm the claim: TPV cells tested at neutron and gamma fluxes characteristic of a shielded fission reactor, holding efficiency at 600 K. What would break it: measured degradation above roughly 1 percent per year in that environment, or 600 K cells that fall well short of the modelled 27 percent system efficiency.