Since more than ten years the BigDFT project consortium has developed a Density Functional Theory code based on Daubechies wavelets. From its earliest conception the code has been employed in High Performance Computing environment. It makes use of multilevel shared/distributed parallelisation paradigms, effectively working on homogeneous and hybrid architectures accelerated with Graphic Processing Units. Recently, the BigDFT code has been extended with a Linear Scaling algorithm, that also exhibit excellent behaviour on large scale computing architectures. Such a scheme is developed on top of the previously existing operations based on Daubechies wavelets. The properties of Daubechies wavelets formalism enable the calculation of a wide variety of systems of many thousand atoms in effective and accurate way, giving also the possiblity of implementing approaches based on Fragments and Constrained Density Functional Theory calculations. In this presentation we will briefly illustrate the advantages of the Daubechies wavelets formalism in the context of Large Scale, High Performance, electronic structure calculations. We will then focus on the scientific challenges that similar software packages enable to address, and we will highlight the importance of identifying different solutions for the developments and the setup of unprecedented scientific simulations in this field, which will enable to tackle unprecedented problems in disciplines at the interface between Physics, Chemistry, and Biology.

Linear scaling density functional theory in Daubechies wavelets basis: Towards paradigm shifts in large-scale electronic structure calculations

254th National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.