Radioisotope Thermoelectric Generator

How an RTG works

Strip an RTG down to its idea and it is almost insultingly simple: a hot lump of metal on one side, cold space on the other, and a wall of thermocouples in between converting the temperature difference into a trickle of electricity. There is nothing to spin, nothing to pump, nothing to ignite. The "battery" inside is a brick of plutonium-238 dioxide that pours out heat simply because it is radioactive, and it will keep doing so for a human lifetime whether you ask it to or not.

The heat comes from alpha decay. A plutonium-238 nucleus spits out an alpha particle (a helium nucleus) and becomes uranium-234. That alpha particle slams to a stop within microns inside the fuel pellet, and its kinetic energy becomes heat. One gram of Pu-238 releases about 0.54 watts continuously — a thumb-sized pellet glows a dull red on its own. A full General Purpose Heat Source (GPHS) module holds four PuO₂ pellets (~600 g of fuel) and produces ~250 W of heat; stack 8 of them and you have the ~2000 W thermal source of a Mars-rover RTG.

Turning that heat into volts is the job of the thermocouple, exploiting the Seebeck effect: hold a junction of two dissimilar conductors at different temperatures and a voltage appears across it. An RTG wires hundreds of these couples electrically in series and thermally in parallel, sandwiched between the hot fuel core and a finned radiator facing cold space.

         HOT side  (~1000 °C, plutonium-238 core)
   ┌───────────────────────────────────────────┐
   │  n-leg │ p-leg │ n-leg │ p-leg │ n-leg ... │   ← thermocouples
   │   ▲ e⁻ │ ▼ h⁺  │   ▲    │   ▼   │    ▲      │     in series
   └───────────────────────────────────────────┘
         COLD side (~300 °C, fin radiator → space)
                 │
                 └──►  DC out  (~28–30 V bus)

Electrons diffuse from hot to cold in the n-type legs, holes do the same in the p-type legs, and because the legs alternate p-n-p-n in series, all those tiny voltages add up to a usable DC bus. No part of this moves. The only "motion" is charge carriers drifting down a temperature gradient — which is exactly why an RTG can run unattended for 40+ years.

The governing physics

Two equations carry almost everything you need to size an RTG. The first is the Seebeck voltage of a single couple, and the second is the thermoelectric figure of merit that decides how good your materials are.

Seebeck voltage of one couple:
    V = (S_p − S_n) · (T_hot − T_cold) = S_pn · ΔT

  S      = Seebeck coefficient (V/K), material property
  ΔT     = hot-side minus cold-side temperature (K)

Dimensionless figure of merit (per material):
    ZT = (S² · σ / κ) · T

  σ      = electrical conductivity (S/m)   — want HIGH
  κ      = thermal conductivity (W/m·K)    — want LOW
  T      = absolute temperature (K)

Maximum thermoelectric efficiency:
    η_max = (ΔT / T_hot) · ( √(1+ZT) − 1 ) / ( √(1+ZT) + T_cold/T_hot )
            └── Carnot ──┘   └──── material penalty ────┘

The first bracket is the Carnot ceiling — you can never beat it. The second bracket is the material tax: it would equal 1 (perfect) only if ZT were infinite. Real space-grade thermoelectrics have ZT ≈ 0.6–1.0, which is why even a 700 K temperature drop yields only ~6–7% device efficiency. Notice the tension baked into ZT: you want high electrical conductivity but low thermal conductivity in the same material, and in most solids those two move together (the Wiedemann–Franz law). Decades of materials work — silicon-germanium for high temperature, lead telluride and skutterudites for mid-range — has been a hunt for the rare "phonon-glass, electron-crystal" material that conducts charge well but heat poorly.

Decay sets the time axis. The thermal power of the fuel follows the exponential decay law:

P(t) = P₀ · (1/2)^(t / t½) = P₀ · e^(−λt),   λ = ln2 / t½

For Pu-238, t½ = 87.7 yr  →  λ = 0.0079 /yr  →  ~0.79% thermal loss per year

Specific power of fresh PuO₂ fuel:  ≈ 0.39 W/g  (oxide form; 0.54 W/g for pure metal)

Worked example: sizing a Voyager-class RTG

Each Voyager carries three MHW-RTGs (Multi-Hundred-Watt). Let's reconstruct one from first principles.

Fuel:               ~4.5 kg PuO₂ spheres, ~2400 W thermal at launch (1977)
Conversion:         312 silicon-germanium (SiGe) thermocouples in series
Hot junction:       T_hot ≈ 1273 K (1000 °C)
Cold junction:      T_cold ≈ 573 K  (300 °C, radiator)
ΔT:                 700 K

Carnot ceiling:     ΔT / T_hot = 700 / 1273 = 55%
Material factor (ZT≈0.5): ≈ 0.12
Device efficiency:  0.55 × 0.12 ≈ 6.6%

Electrical output:  2400 W × 0.066 ≈ 158 W  per RTG at launch
Three RTGs:         ≈ 470 W  total — matches the spec sheet ✓

Now run the clock forward. The fuel alone loses 0.79%/yr, so after 48 years (to 2025) the thermal source has dropped to 2400 × (0.5)^(48/87.7) ≈ 1640 W — a 32% fall. But Voyager's measured electrical output fell from 470 W to ~230 W, a 51% drop. The extra loss is the thermocouples degrading: SiGe legs sublime and their bonds grow resistive over decades. That gap between fuel decay and measured output is the single most important practical fact about RTGs — the converter ages faster than the fuel.

That declining budget is why Voyager's operators have spent the 2020s switching off instrument heaters and science instruments one by one: every watt the RTGs lose must be subtracted from somewhere, and the spacecraft is now running on roughly the power of a dim refrigerator bulb across 24 billion kilometres.

Real flown systems

Generator	Mission(s)	Thermal / Electrical (BOL)	Thermoelectric material	Notes
SNAP-27	Apollo 12–17 ALSEP	1480 W / ~70 W	Lead telluride (PbTe)	Astronauts hand-loaded the fuel cask on the Moon
MHW-RTG	Voyager 1 & 2, LES-8/9	2400 W / ~158 W each	Silicon-germanium (SiGe)	Three per Voyager; still operating after 48 years
GPHS-RTG	Galileo, Ulysses, Cassini, New Horizons	~4400 W / ~290 W	Silicon-germanium (SiGe)	The workhorse; uses stacked GPHS modules
MMRTG	Curiosity, Perseverance, Dragonfly	~2000 W / ~110 W	PbTe / TAGS	Multi-Mission; runs in atmosphere or vacuum, ~14 yr design life
RHU (heater, not generator)	Many probes & rovers	1 W / 0 W	None — pure heat	~2.7 g pellet, keeps electronics from freezing

RTG vs other space power sources

	RTG (Pu-238)	Solar PV array	Fission reactor (Kilopower)	Chemical battery	Fuel cell
Works in deep space / no sun	Yes	No (inverse-square falloff)	Yes	Yes (short term)	Yes (short term)
Moving parts	None	Deploy/gimbal only	Pumps, turbine (Stirling)	None	Pumps/valves
Typical power class	100 W – 300 W	100 W – 30 kW	1 – 10 kW	Wh, not continuous	1 – 10 kW
Conversion efficiency	~6–7%	~30% (cell)	~25–35% (Stirling)	n/a (storage)	~50–60%
Operational lifetime	14 – 45+ years	10 – 20 years (degrades)	~15 years	Hours – days	Days – weeks (reactant-limited)
Power decay cause	Decay + couple aging (~3%/yr early)	Radiation damage, dust	Fuel burnup (slow)	Self-discharge	Reactant depletion
Throttleable?	No — fixed heat output	Yes (point away)	Yes (control rods)	Yes	Yes
Best home	Outer planets, Mars night, lunar night	Inner solar system, Earth orbit	Surface bases, high-power probes	Launch, landers, buffering	Crewed short missions

Design tradeoffs and failure modes

• Efficiency vs reliability. 6% efficiency looks terrible until you remember a Stirling converter at 25% efficiency adds a vibrating piston that must not fail for 14 years in a place no one can service it. NASA flew the Advanced Stirling Radioisotope Generator (ASRG) to TRL-6, then cancelled it in 2013 precisely because the moving-piston reliability risk outweighed the 4× efficiency gain. RTGs win by being boring.
• Thermocouple sublimation. At 1000 °C the SiGe legs slowly evaporate, thinning the conductor and raising resistance. Designers fight this with sealed cover gases (argon) and getters, but it remains the dominant electrical-degradation mechanism and the reason output falls faster than fuel decay.
• Fuel availability. Pu-238 is not a waste product; it must be deliberately bred by neutron-irradiating neptunium-237 in a reactor, then chemically separated. The US stopped production in 1988, ran down its stockpile, and only restarted at Oak Ridge in 2015 — now making roughly 400–500 g/yr against a goal of 1.5 kg/yr. A single GPHS-RTG holds 18 modules with roughly 7.8 kg of Pu-238 (about 10.9 kg as the dioxide). Fuel, not engineering, is the binding constraint on America's outer-planets program.
• Waste-heat management. Of ~4400 W thermal, ~4100 W is waste heat that must be radiated away — but that same waste heat is a feature, piped through the spacecraft to keep propellant and electronics from freezing. RTG missions design the thermal and electrical budgets together.
• Radiation to instruments. Pu-238 is a near-pure alpha emitter, but trace Pu-236 and spontaneous fission produce neutrons and gammas. Sensitive instruments (star trackers, particle detectors) are placed on a boom away from the RTG — the reason Voyager and Cassini have their generators on long arms.
• Launch-accident containment. The whole GPHS architecture exists to keep fuel contained if the rocket fails. Iridium cladding (ductile, high-melting, oxidation-resistant) around each pellet, carbon-carbon impact shells, and an aeroshell rated for reentry. This is also the source of most public controversy at every Pu-238 launch.

Why plutonium-238, specifically

The choice of fuel is the whole ballgame, and only a handful of isotopes even qualify. You want a long-ish half-life (so the mission outlives the fuel), high specific power (so the generator isn't enormous), and clean radiation (so you don't need a tonne of shielding). Those pull in opposite directions: high specific power means a short half-life, and a long half-life means low specific power. Pu-238 sits at the rare intersection.

Isotope	Half-life	Specific power	Radiation	Verdict for an RTG
Plutonium-238	87.7 yr	0.54 W/g	Alpha (clean)	The standard. Long life, modest shielding
Strontium-90	28.8 yr	0.46 W/g	Beta → bremsstrahlung	Cheap (fission waste) but needs heavy shield; used in Soviet terrestrial RTGs
Polonium-210	138 days	140 W/g	Alpha	Huge power density, but gone in months — short missions only
Curium-244	18.1 yr	2.8 W/g	Alpha + neutrons	High power but neutron-noisy; needs more shielding
Americium-241	432 yr	0.11 W/g	Alpha + gamma	ESA's pick — very long life, low power; bulkier generators

Europe, lacking a Pu-238 supply, has invested in americium-241 RTGs and heater units: its 432-year half-life and availability from civil plutonium stockpiles outweigh its low 0.11 W/g specific power and the gamma shielding it demands. It's a reminder that the "best" fuel is partly a supply-chain question, not just a physics one.

Common misconceptions and pitfalls

• "An RTG is a tiny nuclear reactor." No fission, no chain reaction, no criticality. It cannot melt down or run away. It is a heat-emitting fuel slug wired to a thermopile — closer to a glow-stick than to a power plant.
• "You can throttle it like a reactor." You can't. The heat output is fixed by physics and falls only as the fuel decays. Spacecraft regulate the electrical side with a shunt regulator that dumps excess power as heat, and store surplus in a battery for peak loads.
• "More efficient is always better." On a 14-year unattended mission, an extra moving part that might seize is a worse bet than 18% of wasted heat. The ASRG cancellation is the textbook case: higher efficiency lost to reliability.
• "Output drops only because the fuel decays." The thermocouples degrade faster than the fuel early in life. Half of Voyager's lost power is converter aging, not decay.
• "It's dangerous because it's radioactive." Pu-238 is an alpha emitter; a sheet of paper or the fuel's own cladding stops the radiation. The genuine hazard is inhalation of dispersed fuel — which is exactly what the GPHS containment is engineered to prevent in an accident. The flown ceramic form (PuO₂) is designed to fracture into coarse, non-respirable chunks rather than fine dust.
• "RTGs are obsolete." The opposite — Dragonfly (launching to Titan) and the proposed enhanced MMRTG show the architecture is still the only proven option for the outer solar system. The bottleneck is Pu-238 supply, not the technology.