Network theory for inhomogeneous thermoelectrics

TL;DR

This paper uses a network model based on Onsager-de Groot-Callen theory to analyze the Harman method for measuring thermoelectric properties, revealing systematic overestimations of key coefficients in mixed-material samples.

Contribution

It introduces a network simulation approach to evaluate the accuracy of the Harman method, highlighting its overestimation issues in inhomogeneous thermoelectric materials.

Findings

Harman method overestimates Seebeck coefficient in mixed materials.

Figure of merit is overestimated when thermal coupling is weak.

Overestimation exceeds 100% near metal percolation in metal-semiconductor mixtures.

Abstract

The Onsager-de Groot-Callen transport theory, implemented as a network model, is used to simulate the transient Harman method, which is widely used experimentally to determine all thermoelectric transport coefficients in a single measurement setup. It is shown that this method systematically overestimates the Seebeck coefficient for samples composed of two different materials. As a consequence, the figure of merit is also overestimated, if the thermal coupling of the measurement setup to the environment is weak. For a mixture of metal and semiconductor particles near metal percolation the figure of merit obtained by the Harman method is more than 100 \% too large. For a correct interpretation of the experimental data, information on composition and microstructure of the sample are indispensable.