Portable coolers that keep insulin cold on a trip, mini refrigerators in cars, and the lasers inside fiber-optic networks all rely on solid-state cooling — no moving parts, no refrigerant gases, just electric current pumping heat. A study in The Innovation reports a way to make such a cooler run colder without inventing a new material: grade the Seebeck coefficient — the property that sets how much heat each electron carries — from one end of the material to the other.
Thermoelectric cooling works because an electric current drags entropy along with it, absorbing heat at the cold end and releasing it at the hot end. In a conventional device, most of that heat exchange happens at the junctions between material and electrodes — the Peltier effect. A second contribution from inside the material, the Thomson effect, is normally negligible because the Seebeck coefficient barely changes across the operating range. The maximum temperature span a cooler can reach, ΔTmax, is thus pinned to the material's figure of merit, zT.
Researchers at Tongji University and Shanghai University widened the accounting. They proposed an extended figure of merit, zextended, that counts the electronic entropy change along the whole transport path — from any variable that shifts the Seebeck coefficient, including temperature, chemical potential, and magnetic field. Experimentally, they grew a compositionally graded n-type Bi2Te2.7Se0.3 single crystal, where a chemical-potential gradient spreads the Seebeck coefficient across roughly 50 microvolts per kelvin from one end to the other. In a uniform commercial leg of the same alloy, the hot and cold ends differ by only about 2.5%.
The consequence is asymmetry. In the graded single-leg device, reversing the current shifted the maximum cooling span by about 20%. With the current aligned to the Seebeck gradient, ΔTmax came out about 10% larger than in an otherwise identical cooler without the gradient; the uniform leg barely noticed which way the current ran.
The gain is not free. Aligning the gradient for extra heat absorption pulls the material away from its optimal average zT, and the team maps the trade-off explicitly. Their modeling shows the bonus grows as the material's electronic quality factor B shrinks — around 0.33 for a typical bismuth telluride alloy, and lower still near cryogenic temperatures, where high lattice thermal conductivity drags down conventional performance. That makes the strategy most promising for low-temperature cooling, where material gains have been hard-won.
The demonstration is a single-leg proof of concept with modest absolute spans — the graded leg reached roughly 15 K, much of the budget lost to heat leaking through the lead wires. Full modules with paired n- and p-type legs should display the effect more clearly. Because the approach relies only on dopant grading, the authors argue it applies, in principle, to any dopable material — a new design degree of freedom that complements, rather than replaces, the pursuit of higher zT. For the silent coolers guarding medicines on the go and holding lasers steady, it is one more knob to turn.
The Innovation
Experimental study
Not applicable
Nonreciprocal thermoelectric cooling by spatially grading Seebeck coefficient for cooling enhancement
2-Nov-2026
The authors declare no competing interests.