Closing the Material-to-Device Efficiency Gap in Thermoelectric Generators with a Socket-Electrode Architecture

Thermoelectric generator (TEG) efficiency is often constrained not by the materials but by device-level losses: electrical contact resistance, parasitic heat leakage, and the difficult assembly of many small elements. We introduce a socket-electrode device architecture that addresses these losses together. Each cylindrical element is brazed into a machined cavity (a socket) in the electrode, forming a three-dimensional junction that also locates and supports it; the couples are sealed in an insulating encapsulant, and the module is built from only three standardized, self-registering electrode variants. We demonstrate the concept in all-skutterudite couples whose figure of merit (ZT) is comparable to, not better than, the state of the art, and isolate the effect of the architecture by comparing four device configurations built from the same materials and measured in the same rig. Progressively removing the electrical and thermal loss channels lowers the internal resistance of a unicouple by about one third at an unchanged open-circuit voltage, increasing its maximum output power by about 46%, from 219 to 321 mW, and its conversion efficiency from 7.7 to 11.3% at a hot-side temperature of 600 °C and a cold side of 140 °C; a six-leg module retains this performance. Against the efficiency calculated from the same transport properties, the material-to-device utilization factor U rises from ≈62 to ≈92%. Recovering performance already present in the material, rather than improving the material itself, is a scalable and largely material-independent route to higher-efficiency thermoelectric generation.