In only twenty five years, first principles quantum mechanical simulation techniques based on density functional theory (DFT) have progressed from modelling two atoms of silicon to modelling thousands of atoms of any species. Using such calculations, we can now make reliable predictions about the structure and properties of many atomistic systems.

There have been numerous applications of such first principles calculations to study surfaces, point defects such as vacancies and impurities, extended defects such as grain boundaries and dislocations, catalysis and many, many other applications. In recent years, these techniques have been extended to predict a wider range of physical and chemical properties and to predict theoretical values for experimentally measured spectra such as optical, nuclear magnetic resonance (NMR) and electron paramagnetic resonance (EPR) and many others. However, the computational cost of these calculations places significant restrictions on our ability to sample complex phase spaces and on the timescales that are accessible to such simulations.

In this talk, I shall describe some of the successes of DFT. I shall point out some of the limitations of the methodology and outline developments that should partially address some of these limitations. However, irrespective of these successes and future developments, it seems clear that truly complex problems, such as ab initio prediction of protein structure, will probably always lie beyond the scope of purely quantum mechanical calculations. In contrast, I believe that a judicious combination of experimental and computational methods should, within a few years, allow us to predict, for instance, protein function. ​