surya.narada.systems

R7: the closing argument for non-photovoltaic solar

Sweep executed 18 August 2026. Verdict: closed. On measured, system-level numbers, no non-photovoltaic solar technology beats photovoltaics on specific power, cost, or readiness. Only unbuilt study projections compete, and they require exactly the components nobody is building for space. The question is retired from future trade studies; this page is the answer they should cite.

The question

Does anything that turns sunlight into electricity without a p-n junction — heat engines, thermionics, thermophotovoltaics — deserve a line in space-power trade studies? The space-unique variants were already closed on the kill table. What remained was the general case, with current numbers, written once.

The numbers

TechnologyBest measuredSystem specific powerStatus
Solar-dynamic Stirling/Brayton29% cycle, 17% orbital end-to-end (2 kWe ground test, 800 h, 372 simulated LEO orbits)3.4 W/kg engineered (Freedom 25 kWe module, 7,440 kg); 10–13 W/kg study-level; 120–700 W/kg projected, never builtNo flight program since 1991; never flown
Solar thermionics6–13% converters (best ~16% at 2,200 K); JPL solar units 7–11% at ~1,900 K~14 W/kg — one study, 1978; concentrator alone 57% of massSolar program discontinued in the 1970s as uncompetitive with PV
Solar TPVCell physics: 41.1% at 2,400 °C (2022), 44% at 1,435 °C (2024). Complete systems: 0.8–3.2% solar-to-electric, measuredNo orbital system estimate exists. Space TPV heritage is nuclear-heated: 100 We class, 6–8 W/kg realisticCells advancing fast; systems stagnant; commercial work is terrestrial storage
PV baseline30.7–34% AM0 production III-V cells50 W/kg rigid; 100–200 W/kg flexible flown (UltraFlex 175 W/kg); thin-film cells >1 kW/kg at cell level, TRL 3–4Shipping, improving, and getting cheaper (Si entrants target $20/W)

Solar-dynamic: the engine works, the system doesn't

established The conversion is real: cycle efficiencies of 25–35% are documented across Stirling, Brayton, and Rankine studies, and NASA's 2 kWe closed-Brayton ground demonstration measured 29% cycle efficiency and 17% orbital end-to-end efficiency over 800 hours and 372 simulated LEO orbits. The 1968–78 Brayton Rotating Unit ran over 38,000 hours without degradation. Nobody is disputing the physics.

The mass is the kill. The only fully engineered system point — Space Station Freedom's 25 kWe module — closed at 7,440 kg:3.4 W/kg, with a two-axis concentrator held to ±0.1°, thermal storage alone half the system mass, and a radiator rejecting ~4 W for every watt delivered. Studies without full engineering reach 10–13 W/kg with storage, 20–26 W/kg day-only, ~40 W/kg with assumed component improvements; Mason's 120–700 W/kg SSP projections require inflatable concentrators at 0.2 kg/m² and 2,000 K ceramic turbines — components that do not exist. Meanwhile the PV side of that 1991 trade shipped, and has since improved array-level specific power by roughly an order of magnitude.

Thermionics: abandoned once, correctly

established JPL's solar-thermionic converters ran 7–11% at ~1,900 K for up to 11,000 hours — and the program was discontinued in the 1970s, explicitly because it was uncompetitive with emerging photovoltaics. Nothing since has changed the physics: space charge still halves efficiency, cesium neutralization still costs 30–50% of it, emitters still want >1,500 K, and the best modern vacuum-gap devices sit at ~10% and ~1 W/cm² against a theoretical 30%+/100 W/cm² that no prototype has approached. The only system-level study ever done — NASA, 1978, 50 kWe — closed at ~14 W/kg with the concentrator at 57% of the mass budget. That was against 1978 photovoltaics.

Solar TPV: the live one, and why it still loses

evidence-backed TPV cell physics is genuinely moving: 41.1 ± 1% at a 2,400 °C emitter (Nature, 2022), 44% at 1,435 °C (Joule, 2024). This is the strongest non-PV thread, and it still closes, because a cell is not a system. The two measured complete solar-TPV systems stand at 0.8% and 3.2% solar-to-electric — against 30%+ direct PV at the same sunlight. Every space application of TPV to date is nuclear-heated (radioisotope or reactor), 100 We-class, with realistic specific power of 6–8 W/kg, limited by the cold-side radiator that the low cell rejection temperature forces. No orbital solar-TPV system-level W/kg estimate exists in the literature; building one from concentrator, emitter, and radiator mass models points the wrong way on every term. The active commercial TPV work (thermal-battery storage) is terrestrial and does not move the space case.

The incumbent baseline

evidence-backed Production III-V cells run 30.7–34% AM0 (Spectrolab XTJ Prime, Rocket Lab/SolAero IMM-α in volume at 32%, IMM-β averaging 33.3% BOL). Arrays ship at ~50 W/kg rigid and 100–200 W/kg flexible — UltraFlex flew at 175 W/kg. Cost: cells run hundreds of $/W ($400–500 per 4×8 cm cell, NASA Spinoff 2016; Airbus Ventures 2022 concurs); the last rigorous public array-level dataset (GSFC 2002) puts delivered arrays at ~$1,650/W. Silicon entrants (Solestial, mPower) are attacking exactly that cost with a $20/W target, and thin-film/perovskite cells exceed 1 kW/kg at cell level, TRL 3–4 — which is what this program's R1 and R2 are for. PV is a moving target; non-PV stood still for three decades and never caught the 1991 version.

Verdict

Closed. No non-photovoltaic solar approach beats PV on any measured system-level axis; the only competitive figures are unbuilt projections depending on components no one is developing for space. Future trade studies should cite this page instead of re-opening the question.

What would change the conclusion

  1. Flight-proven ultra-light concentrators (<0.5 kg/m²), developed for any reason — solar sailing, thermal propulsion. Concentrator mass is the term that killed every thermal scheme.
  2. A 2,400 °C-class emitter plus spectral-control stack with multi-year space durability — moves solar TPV systems toward their cell physics.
  3. Mission profiles where PV+battery mass explodes (deep space, eclipse-heavy orbits) — re-run the trade with storage charged to the PV side.
  4. Anyone flies a non-PV solar demonstrator above 50 W/kg system-level.

Annual check against these four triggers. Otherwise the direction stays closed.

Perimeter and provenance

Sweep of 18 August 2026, four parallel literature sweeps, twenty sources. Findings admitted to the Surya epistemic graph as eventev-000000000006 with assessmentsurya-assessment-r7-closing-2026-08, judgmentestablished-with-gap. Known gaps, flagged there: no current merchant III-V $/W primary source (the 2002 GSFC dataset is the last rigorous public one; a 2026 market-report figure of $307/W exists but its methodology is opaque — treated as indicative, not load-bearing); no modern system-level W/kg for thermionics or orbital solar TPV. Neither gap can reverse the conclusion: both would require order-of-magnitude reversals in technologies that have been static or abandoned for decades, while PV kept moving. The earlier kill-table figure of 12–25 W/kg for solar-dynamic is refined here: 3.4 W/kg engineered, 10–13 W/kg study-level.

Sources