ITR: Large-scale, Grid-enabled Gaussian Orbital Implementation of Current Density and Spin Density Functional Theory for Ordered Systems

University Of Florida, Gainesville FL

Investigators

Abstract

This award is made in response to a proposal submitted to the Information Technology Research (ITR) Initiative. The objective of the research is to design, develop and implement, test, and apply major enhancements of the high-performance, all-electron, full-potential, relativistic gaussian-basis density functional theory (DFT) code for crystals, slabs, and periodic polymers called "GTOFF" (Gaussian Type Orbitals for Fitting Functions). The goals are (a) to include imposed magnetic fields as a major extension of materials-specific DFT prediction and interpretation of periodic systems; (b) make major speed-ups to handle much larger systems, e.g., nano-featured surfaces, ordered hard-soft interfaces; (c) introduce algorithms and data structures that will enable treatment of extremely large and/or complicated systems via grid technologies; (d) substantially improve the capability for revising and/or adding methods and capabilities to do more science. A major advantage of GTOFF is use of the same kinds of basis sets and procedures as the majority of molecular codes. It thus provides a seamless way to study chemically differentiated systems, from constituent atoms to molecule, cluster or nanoparticle to crystal or slab, including cluster-surface interactions and nano-featured surfaces. Successful, robust spin DFT algorithms in GTOFF will be reimplemented in C++ (from Fortran 77) to achieve modularity, manageability, and extendability now lacking. Those algorithms include generalized Ewald techniques to sum long-ranged contributions (rather than more common, sometimes less reliable multipole expansions), variational Coulomb fitting to eliminate 4-center integrals, fitting of exchange-correlation densities to fitted densities (eliminates extra sums over Brillouin Zone points), accuracy-conserving constraints on fittings, Douglas-Kroll-Hess relativity (including spin-orbit contributions). GTOFF now is serial. The redsign/reimplementation will exploit known parallelism opportunities as well as provide grid enablement. Current-density Functional Theory (C-DFT) will be built into the new design as a major extension to provide predictive treatment of materials at high magnetic field. The planned effort involves several major components: (a) redesign (not mere transcription) of the present GTOFF to take advantage of modern programming practices, construct a proper organization for parallelism (with hooks for MPI), include algorithms and data structures for grid-enabled computing; (b) design of gaussian orbital algorithms and techniques for current DFT; (c) re-implementation in C++ of the redesigned GTOFF for spin DFT; (d) extensive testing to assure reproducibility in serial mode, then in parallel; (e) implementation and testing of current DFT in the new GTOFF; (f) grid-enablement, including (but not limited to) interpolation among precalculated integral arrays, interpolation among approximate energy calculations, and energy gradient calculations direct from density fitting. The grant will provide support for a postdoctoral associate and graduate students who will will learn programming skills in addition to the fundamental chemistry and physics necessary to develop and apply the GTOFF codes. %%% This award is made in response to a proposal submitted to the Information Technology Research (ITR) Initiative. The objective of the research is to design, develop and implement, test, and apply major enhancements of the high-performance, all-electron, full-potential, relativistic gaussian-basis density functional theory (DFT) code for crystals, slabs, and periodic polymers called "GTOFF" (Gaussian Type Orbitals for Fitting Functions). The goals are (a) to include imposed magnetic fields as a major extension of materials-specific DFT prediction and interpretation of periodic systems; (b) make major speed-ups to handle much larger systems, e.g., nano-featured surfaces, ordered hard-soft interfaces; (c) introduce algorithms and data structures that will enable treatment of extremely large and/or complicated systems via grid technologies; (d) substantially improve the capability for revising and/or adding methods and capabilities to do more science. A major advantage of GTOFF is use of the same kinds of basis sets and procedures as the majority of molecular codes. It thus provides a seamless way to study chemically differentiated systems, from constituent atoms to molecule, cluster or nanoparticle to crystal or slab, including cluster-surface interactions and nano-featured surfaces. The grant will provide support for a postdoctoral associate and graduate students who will will learn programming skills in addition to the fundamental chemistry and physics necessary to develop and apply the GTOFF codes. ***