Abstract:
Ab initio methods have been used for many decades to accurately predict properties of solids such as the physical,
electronic, optical, magnetic, and elastic properties. A generation ago, many research groups developed their own in-house
codes to perform ab initio calculations. In doing so, research students were intimately involved in many aspects of the coding,
such as developing the theoretical framework, and algorithmic and programming details. Over time however, collaborations
between various research groups within academia and in industry have resulted in the creation of more than 50 large opensource
and commercial electronic structure packages. These software packages are widely used today for condensed matter
research by students who, unfortunately, often have very little understanding of the fundamental aspects of these codes. To
address this shortcoming, we have embarked on a program at the University of Pretoria to devise a range of simplified, easily
programmable computational problems appropriate for the classroom, which can be used to teach advanced undergraduate
students about particular theoretical and computational aspects of the electronic structure method. In this paper, we focus on the
pseudopotential, which is a centrally important concept in many modern ab initio methods. Whereas the full implementation of the
pseudopotential construct in a real electronic structure code requires complex numerical methods, e.g. accelerated convergence
to self-consistency including the interactions between all the electrons in the system, we show that the essential principles of the
pseudopotential can, nevertheless, be presented in a simpler class of problems, which can readily be coded by students.
