Thermoelectric Effect Explained: Seebeck and Peltier
One hot end and one cold end is a battery, worth a few hundred microvolts per kelvin in doped semiconductors. The ZT problem and 49 years of Voyager power.
Table of Contents
A Battery With No Moving Parts
Voyager 1 is more than 25 billion kilometres from Earth and still transmitting. Nothing aboard it spins, pumps, burns or wears out. Its electricity comes from three bricks of plutonium-238 dioxide wrapped in silicon-germanium, and the only thing happening inside is that one end is hot and the other is cold. That has been the entire mechanism since 5 September 1977.
A slab of the right material with a temperature difference across it is a battery — not as a metaphor, but as an open-circuit voltage you can put a meter on. Thomas Johann Seebeck found it in 1821, when he noticed that a loop of two different metals with its junctions held at different temperatures deflected a nearby compass needle. He thought heat was making magnetism. He was making a current.
The quantity that matters is the Seebeck coefficient S, the voltage produced per kelvin of temperature difference. In ordinary metals it is a few microvolts per kelvin, which is why nobody ever built a power station out of copper and iron. In heavily doped semiconductors it is a couple of hundred microvolts per kelvin, and wiring several hundred junctions in series turns that into something usable. The mechanism is diffusion: carriers at the hot end are more energetic and drift towards the cold end faster than cold ones come back, piling up charge until the field they create stops the flow.
One Effect, Running in Two Directions
Thirteen years later, Jean Charles Athanase Peltier ran the experiment backwards. In 1834 he drove a current through a junction between two different conductors and found that one side cooled while the other heated. That is not resistive heating, which warms everything regardless of direction; reverse the current and the cold side becomes the hot one. Something was pumping heat.
William Thomson, later Lord Kelvin, showed in 1851 that Seebeck and Peltier had found one phenomenon rather than two. Thermodynamics gives the first Kelvin relation, Π = S·T, where Π is the Peltier coefficient — the heat carried per unit charge across a junction — and T is the absolute temperature. Measure how many volts per kelvin a material generates and you have already measured how many joules per coulomb it will pump. Thomson also predicted a third effect nobody had looked for: a current flowing along a wire that already carries a temperature gradient absorbs or releases heat continuously along its length, with coefficient μ = T·(dS/dT). All three follow from one fact, that charge carriers in a solid transport entropy as well as charge, and the relations between them were an early victory for thermodynamic reasoning over experiment-by-experiment cataloguing.
The Figure of Merit That Asks for the Impossible
Everything useful about a thermoelectric material is compressed into one dimensionless number:
ZT = S²σT/κ
A good material needs three things at once: a large Seebeck coefficient S, high electrical conductivity σ so the current it drives is not thrown away as internal resistance, and low thermal conductivity κ so the temperature difference does not simply leak away through the slab. Those demands are in direct conflict. Metals and doped semiconductors conduct electricity well because they hold a sea of mobile electrons, and those same electrons carry heat — the Wiedemann-Franz law ties electronic thermal conductivity to electrical conductivity through a nearly universal constant, so pushing σ up drags κ up with it. Raising S usually means reducing the carrier concentration, which pulls σ down.
The one route that works is attacking the part of κ that electrons do not carry. Heat also travels through a solid as lattice vibrations, or phonons, and that channel can be sabotaged almost independently of the electrons: heavy atoms rattling in cage-like voids, nanoscale grain boundaries, deliberately disordered alloys. The slogan is phonon glass, electron crystal — make the material look crystalline to charge and glassy to vibrations. Nearly every advance of the last thirty years is a variation on that trick.
Real Materials, and Records That Keep Moving
Bismuth telluride, Bi₂Te₃, is the workhorse. Its Seebeck coefficient is around 200 µV/K per leg and its ZT sits close to 1 near room temperature, and it is what almost every module you can buy is made of. Lead telluride and its relatives take over between roughly 500 K and 800 K. Silicon-germanium, with ZT between about 0.5 and 0.8 in the forms actually flown, is the only family that survives above 1,000 K. Filled skutterudites such as Yb₀.₂Co₄Sb₁₂ reach ZT near 1.0 at 823 K, with the best multiply-filled compositions approaching 1.5.
Laboratory records are higher and they move. Sodium-doped polycrystalline tin selenide has been reported at ZT ≈ 3.1 at 783 K, above the ZT ≈ 2.6 measured in SnSe single crystals; silver antimony telluride reached 2.5 at 673 K in a 2024 paper, and nanostructured PbTe-SrTe has been pushed to 2.5. Read each as a peak value at one temperature in a small sample, often hard to reproduce. What you can actually buy is a module, and even good bismuth telluride modules built for waste-heat recovery land around 8% efficiency. The distance between a record ZT and a shipping device is the real story of the field, and it is the same gap that keeps ambient energy harvesting in the milliwatt range.
Forty-Nine Years on a Slab of Silicon-Germanium
Each Voyager carries three multi-hundred-watt radioisotope thermoelectric generators. Each one holds 24 pressed spheres of plutonium-238 dioxide making about 2,400 watts of heat, surrounded by 312 silicon-germanium couples with their hot junctions initially at 1,273 K and their cold junctions at 573 K. Each generator delivered about 157 watts of electricity, so 470 watts for the spacecraft at launch. That is 6.5% conversion, against a Carnot ceiling of 55%.
The output is now around 220 watts, falling by roughly four watts a year. Plutonium-238’s 87.7-year half-life accounts for only part of it: after 49 years the heat source should still be delivering about 68% of its original power, while the electricity is down to 47%. The missing third is the converter itself, slowly degrading as dopants migrate and sublimation thins the legs. Nothing moves, but things still age.
Perseverance runs on one newer unit with 768 lead-telluride and TAGS couples, turning about 2,000 watts of heat into roughly 110 watts of electricity. On Mars the inefficiency is partly a feature: the waste heat is plumbed around the rover to keep its electronics alive through nights at −90 °C, and no dust storm or polar winter cuts the output.
The Same Device, Driven Backwards
Wire a module to a power supply instead of a load and you have a Peltier cooler. A typical one is 127 couples of bismuth telluride sandwiched between two ceramic plates, the size of a large postage stamp, and it will pull its cold face about 70 K below its hot face if you give it enough current and somewhere to dump the heat.
The catch is the coefficient of performance, the heat moved per watt of electricity supplied. Peltier modules manage something between 0.3 and 0.8 in practice, where a vapour-compression fridge achieves 2 to 4. Thermoelectric cooling is three to ten times more expensive to run. It wins anyway wherever the alternative cannot fit or cannot be trusted: hotel minibars, camping coolers, cooled car seats, small dehumidifiers, and more seriously the temperature stabilisation of laser diodes, infrared detectors and astronomical camera sensors, where a vibrating compressor would ruin the measurement.
Where It Actually Wins
Thermoelectrics lose almost every contest decided on efficiency. A gas turbine beats them, a heat engine with pistons beats them, a compressor fridge beats them. They win every contest decided on the number of things that can break, which is why they end up in the places nobody can visit: deep-space probes, remote sensors, pacemakers in the 1970s, and the cold finger of an instrument that must not shake. The same logic drives interest in piezoelectric harvesting, where the figures of merit are equally unflattering and equally beside the point.
Voyager 1 will go quiet sometime in the 2030s, not because anything failed but because a slab with a hot end and a cold end finally runs out of temperature difference. Forty-nine years is a long time for 6.5%.
Frequently Asked Questions
What is the thermoelectric effect?
The thermoelectric effect is the direct conversion between a temperature difference and an electrical voltage inside a solid, working in either direction. Heat one end of a suitable conductor while holding the other end cold and a voltage appears across it, with no moving parts, no working fluid and nothing to wear out. That is the Seebeck effect, discovered in 1821. Drive a current through the same material instead and it pumps heat from one face to the other, cooling one side and warming the other. That is the Peltier effect, found in 1834. The size of the voltage is set by the Seebeck coefficient S, measured in microvolts per kelvin: a few µV/K in ordinary metals, and roughly 200 µV/K in good doped semiconductors such as bismuth telluride. Because the conversion is entirely solid-state it is extremely reliable but inefficient. Practical generators turn only about 5 to 8 per cent of the heat passing through them into electricity.
What is the difference between the Seebeck effect and the Peltier effect?
They are the same physics running in opposite directions. In the Seebeck effect a temperature difference across a material produces a voltage, so heat in gives electricity out. In the Peltier effect a current forced through a junction carries heat across it, so electricity in gives a temperature difference out: one face cools, the other warms, and reversing the current swaps which is which. William Thomson showed in 1851 that the two coefficients are not independent. The first Kelvin relation is Π = S·T, where Π is the Peltier coefficient in joules per coulomb, S is the Seebeck coefficient in volts per kelvin and T is the absolute temperature, so measuring one determines the other. Thomson added a third effect of his own: a current flowing along a wire that already carries a temperature gradient absorbs or releases heat along its length, with coefficient μ = T·(dS/dT). A Peltier cooling module and a thermoelectric generator are often the same hardware, wired differently.
What does ZT mean for a thermoelectric material?
ZT is the dimensionless figure of merit, ZT = S²σT/κ, where S is the Seebeck coefficient, σ the electrical conductivity, T the absolute temperature and κ the total thermal conductivity. It sets how close a material can come to the Carnot limit. Efficiency rises with ZT but saturates slowly, so even ZT = 2 recovers only about a quarter of the Carnot ceiling. The difficulty is that the three ingredients fight each other. High electrical conductivity needs plenty of mobile electrons, and those same electrons conduct heat, which raises κ. Raising S usually means fewer carriers, which lowers σ. The one clean win is suppressing the lattice-vibration part of κ, which is why modern materials use heavy rattling atoms, nanoscale grain boundaries and deliberately disordered alloys. Commercial bismuth telluride sits near ZT = 1 at room temperature. Laboratory peaks above 2.5 have been reported and the records keep moving, but those are single-temperature values on small samples.
How does a radioisotope thermoelectric generator power a spacecraft?
A radioisotope thermoelectric generator, or RTG, wraps a block of decaying plutonium-238 dioxide in thermoelectric couples and radiates the leftover heat to space, so the couples sit between a hot source and a cold sink and produce a voltage through the Seebeck effect. Each of the three generators on a Voyager spacecraft holds 24 pressed plutonium oxide spheres making about 2,400 watts of heat, and 312 silicon-germanium couples working between a hot junction near 1,273 K and a cold junction near 573 K, delivering roughly 157 watts of electricity. That is about 6.5 per cent conversion against a Carnot ceiling of 55 per cent. The three together supplied about 470 watts in 1977 and now deliver around 220 watts, fading by roughly four watts a year. Plutonium-238's 87.7-year half-life explains only part of that decline; the rest is gradual degradation of the couples. Perseverance uses one newer unit with 768 lead-telluride couples for about 110 watts.