Semiconductor dopability is inherently limited by intrinsic defect chemistry. In many thermoelectric materials, narrow band gaps due to strong spin-orbit interactions make accurate atomic level predictions of intrinsic defect chemistry and self-doping computationally challenging. Here we use different levels of theory to model point defects in PbTe, and compare and contrast the results against each other and a large body of experimental data. We find that to accurately reproduce the intrinsic defect chemistry and known self-doping behavior of PbTe, it is essential to (a) go beyond the semi-local GGA approximation to density functional theory, (b) include spin-orbit coupling, and (c) utilize many-body GW theory to describe the positions of individual band edges. The hybrid HSE functional with spin-orbit coupling included, in combination with the band edge shifts from G 0 W 0 is the only approach that accurately captures both the intrinsic conductivity type of PbTe as function of synthesis conditions as well as the measured charge carrier concentrations, without the need for experimental inputs. Our results reaffirm the critical role of the position of individual band edges in defect calculations, and demonstrate that dopability can be accurately predicted in such challenging narrow band gap materials.
INTRODUCTIONThe dopability of semiconductor materials plays a decisive role in device performance. Dopability refers to the carrier concentration limits achievable in a semiconductor material. These limits are set by the compensating intrinsic (or native) defects and the solubility of extrinsic dopants. The computational prediction of dopability has a multi-decade history in microelectronic and optoelectronic materials (e.g. III-V compounds, 1 transparent conducting oxides 2,3 ). Successful computational prediction of dopability has been enabled by the accurate description of native defect chemistry, their formation energies and the absolute position of band edges in these materials.