A new massively parallel version of CRYSTAL for large systems on high performance computing architectures

Abstract

Fully ab initio treatment of complex solid systems needs computational software which is able to efficiently take advantage of the growing power of high performance computing (HPC) architectures. Recent improvements in CRYSTAL, a periodic ab initio code that uses a Gaussian basis set, allows treatment of very large unit cells for crystalline systems on HPC architectures with high parallel efficiency in terms of running time and memory requirements. The latter is a crucial point, due to the trend toward architectures relying on a very high number of cores with associated relatively low memory availability. An exhaustive performance analysis shows that density functional calculations, based on a hybrid functional, of low‐symmetry systems containing up to 100,000 atomic orbitals and 8000 atoms are feasible on the most advanced HPC architectures available to European researchers today, using thousands of processors. © 2012 Wiley Periodicals, Inc.

Silica Surface Features and Their Role in the Adsorption of Biomolecules: Computational Modeling and Experiments

“…The same model has been adopted in a number of studies devoted to study adsorption of simple molecules to silica [318][319][320] as well as to study the adsorption of glycine 309,[321][322][323] and alanine. 324,325 This approach has obvious limitations and it is preferable to rely on amorphous silica models, which is now feasible with the development of highly parallelized computer programs 326 on the modern high performance computing facilities.…”

Thermodynamics and phonon dispersion of pyrope and grossular silicate garnets from ab initio simulations

“…By virtue of recent advances, both in the hardware and in the software, (Bush et al 2011;Orlando et al 2012;Dovesi et al 2014a) ab initio investigation of thermodynamic properties of complex materials has become a feasible task. We show here thermodynamic data obtained for pyrope and Grossular end member garnets with the CRYSTAL14 program ) by including phonon dispersion in the direct (frozen-phonon) approach.…”

Electron density analysis of large (molecular and periodic) systems: A parallel implementation

“…The present implementation takes advantage of the several significant improvements recently made in Crystal in terms of increased parallel and massive-parallel scalability, reduced use of memory per node and increased exploitation of symmetry at all steps of the calculation, which recently allowed the program to be run in parallel mode over 32'000 CPUs and to study systems containing up to about 14'000 atoms and 200'000 basis functions. [17][18][19] Specific features of the current implementation are: i) possibility of studying systems of any dimensionality within the same formal and numerical framework (from 0D molecules, to 1D polymers, nanotubes, helices and nano-rods, to 2D slabs and 3D crystals); ii) efficient use of several DFT functionals, belonging to four rungs of the well-known "Jacob's ladder" 20 (local density approximation, LDA, generalized gradient approximation, GGA, global or range-separated hybrids and meta-GGA); iii) full exploitation of any residual symmetry; iv) parallelization of all algorithms related to the evaluation of ρ(r), its gradient and Laplacian, of the X-ray structure factors, of Bader's topological analysis (as generalized to periodic systems by C. Gatti's Topond package, 21,22 which has recently been merged into the Crystal program), of directional Compton profiles, of the electrostatic potential and its derivatives, of the electronic band structure and density-of-states. The Crystal program adopts an atom-centered basis set of Gaussian-type functions (GTF); all density matrix-based algorithms have been parallelized on the number of orbital shell-shell pairs, which guarantees a good load balance among processors and thus a satisfactory speedup for most systems.…”