Ab Initio Modeling of Optoelectronic Properties of Chalcopyrites and of their Point Defects

Abstract

Ab initio calculations have been instrumental in the understanding of important structural and chemical properties of Chalcopyrites, such as non stoichiometry, self compensation and stability. The benefit of such methods has been specially important as chalcopyrites are a very complex class of semiconductors, in which the interpretation of experiments is far from straightforward. The role of point defects in the electronic properties of chalcopyrites has been recognized long ago and challenging results have been presented using DFT calculations. Unfortunately, whereas DFT is well suited for ground state properties (structure, energy of formation, lattice vibrations), it is much less so when it comes to computing properties linked to excited states of the system (band gap, ionisation energies of defects, optical properties,..). These shortcomings have been addressed in different ways: either using semi-empirical electron-electron repulsion (LDA+U), or so called hybrid approaches (functionals combining from Hartree-Fock and correlation from other sources) or even computationally intensive methods based on a Green’s function description of the many body effects (GW approximation). GW calculations are often performed as “one-shot” corrections to self-consistent Kohn-Sham local density approximation calculations. This standard approach has been successful in many applications on solids. However, it has been shown that it fails for many compounds with d electrons, where only a self-consistent (SC) GW scheme allows to recover a good description of the quasi-particle energies. The objective of the present work is to obtain missing values on fundamental optoelectronic properties of chalcopyrite alloys and of their main defects to feed device level modelling. The calculations are based on, and go beyond, density functional theory within the two frameworks of all electron and pseudopotential-planewave calculations. We compare standard DFT results to hybrid approaches where a fixed amount of exact exchange has been introduced, to LDA+U and to SC-GW calculations within different approximations. The differences between these approaches and their reliabilities will be discussed. Only GW using self-consistency was able to give band gaps within 10% error. We calculate effective masses and show the contribution of the GW corrections to the band dispersion, as well as their importance for the different types of electronic states that are present in these complex compounds. Finally, we discuss the influence of many-body corrections on the defect levels. In this work we will also present results on how the different computing schemes affect intrinsic and extrinsic point defect formation energies.