Modeling of interface roughness in thermoelectric composite materials

We use a network model to calculate the influence of the mesoscopic interface structure on the thermoelectric properties of superlattice structures consisting of alternating layers of materials A and B. The thermoelectric figure of merit of such a composite material depends on the layer thickness, if interface resistances are accounted for, and can be increased by proper interface design. In general, interface roughness reduces the figure of merit, again compared to the case of ideal interfaces. However, the strength of this reduction depends strongly on the type of interface roughness. Smooth atomic surface diffusion leading to alloying of materials A and B causes the largest reduction of the figure of merit. Consequently, in real structures, it is important not only to minimize interface roughness, but also to control the type of roughness. Although the microscopic effects of interfaces are only empirically accounted for, using a network model can yield useful information about the dependence of the macroscopic transport coefficients on the mesoscopic disorder in structured thermoelectric materials.