Thermophotovoltaics (TPV) — Heat to Electricity by Light, the Inverse of Cooling

Scope note. TPV is not a cooling technology — it converts heat into electricity, the inverse of moving heat out of a space. It sits in this KB because it runs on the same physics the cooling articles depend on (black-body and selective-surface radiation), and because its thermoradiative variant — a diode that makes power by radiating to a cold sink — is a direct cousin of night-sky radiative cooling. Read it as an adjacent energy-conversion concept, not a chiller. The article rests on one encyclopedia source, so treat the numbers as a well-cited overview rather than primary data.

How it works

A TPV system is a heat source, a hot emitter (often a metal or ceramic block, or the storage mass itself), a photovoltaic cell, and a waste-heat rejection path. The emitter radiates a thermal spectrum; the PV cell — an ordinary p–n junction, but tuned for that emitter’s spectrum rather than for sunlight — absorbs above-bandgap photons and drives current. Sub-bandgap photons do nothing; above-bandgap photons dump their excess energy as heat, so a higher bandgap gives more voltage from fewer photons and electrical power peaks at an intermediate bandgap.

The Sun’s AM1.5 spectrum (~1,000 W/m² at an apparent 5,780 K) is best captured near 1.4 eV — close to silicon’s 1.1 eV, which is why solar PV is cheap. Thermal emitters run far cooler (industrial burners 900–1,300 °C; a welding torch ~3,400 K) and, by Wien’s law, emit a broader spectrum peaking deep in the infrared: the source states that most emitter materials cap near 1,800 °C with a spectral peak near 1,600 nm (~0.75 eV), and ~0.5 eV at 1,200 °C. (Wien’s law puts the 1,600 nm peak at ~1,800 K, not °C — the source’s own Carnot example uses 1,800 K — so read the pairing as approximate.) Silicon’s 1.1 eV and GaAs’s 1.4 eV bandgaps are therefore too high for realistic emitters, and TPV nearly always uses narrow-bandgap, multi-junction cells to reach double-digit efficiency.

The key surface-engineering moves are the same ones the cooling side uses in reverse: selective emitters (radiate strongly only in the band the cell can convert) and spectral filters / back-mirrors that reflect sub-bandgap photons back into the emitter to be recycled rather than lost as heat.

Efficiency — and why the system number is so low

The theoretical ceiling is Carnot: η = 1 − T_cell/T_emit. With a room-temperature cell and an ~1,800 K emitter that’s ~83%. Reality is far lower and the gap between device and system efficiency is the important story:

Level	Efficiency	Note
Theoretical (Carnot)	~83%	T_cell ≈ 300 K, T_emit ≈ 1,800 K
Lab device (heat→electric)	~35% (2021, 1,773 K emitter) → 41% (MIT/NREL 2022) → 44% (May 2024)	2022: multi-layer III-V cell, gold reflector, tungsten emitter to 2,400 °C; 2024: SiC heat-storage block at 1,435 °C as the emitter, cell capturing 20–30% of photons plus air and gold-reflector layers
System (fuel→electric)	~5%	described as a “world record”; killed by heat-transfer losses, low output voltage, active cooling

Losses stack across four stages — emitter (non-ideal black-body, deep-IR phonon resonances), filter (imperfect reflection back to the emitter), converter (non-radiative recombination, Ohmic, Fresnel reflection, unabsorbed light), and geometry (incomplete converter coverage around the emitter). Keeping the cell cool is the recurring engineering fight: PV efficiency drops as the cell heats — the same “reject the waste heat” problem the cooling articles care about, here as a parasitic rather than the goal.

Materials

Emitters: polycrystalline SiC (the most common burner emitter — cheap, stable to ~1,700 °C, long-wavelength emission recycled by filters; also the 2024-record storage/emitter), tungsten (refractory selective emitter, emissivity 0.45–0.47 in the visible/near-IR vs 0.1–0.2 in the IR, but oxidizes readily), rare-earth oxides (Yb₂O₃ peaking at 1.29 eV for Si cells — ~13% demonstrated; Er₂O₃ at 0.827 eV for GaSb/InGaAs; selective only above ~1,100 °C), and photonic crystals (engineered emission bands; Sandia predicted 34% with tungsten photonic crystals, but manufacturing is not commercially feasible).
PV cells: GaSb (0.72 eV, invented 1989, the basis of most modern TPV; ~20% at a 1,000 °C black-body spectrum against a 52% radiative limit), InGaAs (0.74 eV lattice-matched to InP, ~15%; bandgap-engineered 0.4–1.4 eV with strain-graded layers, and lattice-matchable to germanium substrates), InGaAsSb (0.5–0.6 eV, internal quantum efficiency approaching 90–95% by MBE, but phase separation and cost limit commercialization), InPAsSb (0.3–0.55 eV, understudied, spectral response to 4.3 μm), and IV-VI PbSnSe/PbSrSe quantum wells on silicon (0.3–0.6 eV, Auger recombination

10× lower than comparable III-Vs). Germanium (0.66 eV) performs poorly; silicon would need >2,000 K emitters and no efficient Si TPV has been built.

History in brief

Henry Kolm built an elementary TPV system at MIT in 1956; Pierre Aigrain’s 1960–61 MIT lectures (he is credited as the inventor) catalyzed sustained R&D. Efficiencies approached 30% in the 1980s, Western Washington University ran the “Viking 29” TPV hybrid car in 1997, and the 2022 (41%) and 2024 (44%) device records above mark the modern push toward storage and waste-heat applications.

Where it touches this KB

TPV is adjacent, but three of its threads run straight into cooling-KB articles:

Thermoradiative / “negative-emission” cells put the photodiode on the hot side of a heat engine and generate power by radiating to a cold sink — the exact physics of night-sky radiative cooling, run for electricity instead of cold. Hybrid designs have been proposed that harvest from both hot and cold diodes.
Thermal-storage discharge. A proposed TPV storage scheme heats a carbon/phase-change block off-peak, keeps the surrounding cells “off” (photons reflected back by mirrors and insulation) until demand, then connects them to the load — making TPV a candidate discharge stage for a heat battery. Grid-scale variants cite a graphite storage medium with molten-tin heat transfer near 2,000 °C, with companies claiming costs below lithium-ion.
Waste-heat recovery. TPV is pitched as an auxiliary converter on otherwise-lost high-temperature heat (steam turbines, even solar cells) — the power-side mirror of driving an adsorption chiller off waste heat.

Real deployments remain niche: RTG replacements for spacecraft (Emcore/Creare/ Oak Ridge/NASA Glenn demonstrated 15–20%, University of Houston 30% — 3–4× thermocouples; heat rejection in vacuum is the hard problem, and a 1,350 K tungsten emitter with tandem filters and a cooled 0.6 eV InGaAs cell reached 20%); portable and CHP generators (JX Crystals’ SiC-emitter/GaSb units: a 25,000 BTU/hr gas stove yielding 100 W at 1.4%, a 2001 Army propane charger at 230 W and 2.5%, and a proposed 25 kW-thermal boiling-coolant CHP unit at an estimated 6–12% and ~0.08 €/kWh — never commercialized because the market was too small); and the storage/grid work above. Recreational-vehicle generators have been proposed but the required emitter temperatures make them unlikely.

Sources

Thermophotovoltaic energy conversion (Wikipedia) — mechanism, Carnot/device/system efficiencies, loss stages, emitter and cell materials, history, applications, and the Planck/Wien framing; the sole source, flagged as encyclopedia-grade