Abstract
<jats:p> Thermoelectric power is an efficient and flexible technology for improving energy efficiency, but widespread applications require cost-effective and sustainable materials with a high figure of merit <jats:italic toggle="yes">zT</jats:italic> . Alloying is a common strategy for optimising <jats:italic toggle="yes">zT</jats:italic> , but predicting and understanding its impact on the material properties is made challenging by the structural disorder. In this work, we develop a practical and rigorous approach to predicting the <jats:italic toggle="yes">zT</jats:italic> of alloys by combining electrical-transport and lattice thermal-conductivity calculations, with approximate models for the electron and phonon scattering, and apply it to study the Sn(S <jats:sub>0.2</jats:sub> Se <jats:sub>0.8</jats:sub> ) alloy. We compare predictions obtained by averaging over the full set of unique configurations in a 32-atom supercell with those obtained from special quasi-random structure (SQS) structures with supercell sizes of up to 256 atoms. We find that a 128-atom SQS model achieves a good balance balance between computational efficiency and realistic predictions of the physical properties. The calculations indicate that Sn(S <jats:sub>0.2</jats:sub> Se <jats:sub>0.8</jats:sub> ) shows comparable performance to the SnSe endpoint but with a modest enhancement to the <jats:italic toggle="yes">zT</jats:italic> due to a reduction in the lattice thermal conductivity, which has a particularly significant impact on the low-temperature performance. </jats:p>