凝聚态物理
材料科学
热导率
热电效应
三元运算
无量纲量
合金
微晶
声子
热电材料
电阻率和电导率
热的
热传导
热力学
散射
格子(音乐)
非弹性中子散射
塞贝克系数
声子散射
电导率
介观物理学
非晶态金属
斯库特绿铁矿
微观结构
作者
Christophe Candolfi,Bartłomiej Wiendlocha,V. Ohorodniichuk,Petr Levinský,Sylvie Migot,Gwladys Steciuk,Ilayda Terzi,Arthur Wieder,A. Dauscher,Soufiane El Oualid,B. Lenoir
标识
DOI:10.1038/s41467-025-66578-y
摘要
Ternary and quaternary alloys based on Bi2Te3 remain to date the most efficient materials for thermoelectric cooling around room temperature. Among the various strategies developed over the last decades to further enhance their performance, nanostructuration induced by melt-spinning has proven to be effective, yielding higher values of the dimensionless thermoelectric figure of merit, attributed primarily to a significant decrease in the lattice thermal conductivity κ L . Here, we present measurements of the total thermal conductivity κ of two polycrystalline samples of the alloy Bi0.4Sb1.6Te3 prepared by conventional powder metallurgy and melt-spinning at low temperatures and under high magnetic fields of up to 14 T that challenge this view. By directly accessing κ L without relying on the Wiedemann-Franz law, we show that both samples exhibit a similar κ L ( T ) dependence below ∼70 K. The difference with the κ L ( T ) dependences predicted by this simple law is tied to anomalously low Lorenz number L , the values and temperature dependence of which are shaped by inelastic scattering processes that relax charge and energy flows in an inequivalent manner. Not captured by conventional transport models, the finite-temperature, downward deviations of L ( T ) are successfully reproduced by a theoretical model that includes inelastic electron-electron and electron-optical phonon scattering. These findings show that the melt-spinning-induced nanostructuration has a weaker influence on κ L than previously thought, and pave the way to a better understanding of the deviations from the Wiedemann-Franz law reported in a growing number of complex thermoelectric materials.
科研通智能强力驱动
Strongly Powered by AbleSci AI