More accurate computer simulations for bigger molecular systems
Over the course of the last three decades, it has become increasingly common to interpret various macroscopic chemical phenomena and reaction mechanisms in terms of specific inter- and intramolecular interactions. Today, this is the case not only within the classical fields of physics and chemistry, but also within such modern areas of natural science as molecular biology and nanotechnology. Thus, quantum chemistry, or the application of quantum mechanics to molecular systems and phenomena, has become an integral tool for all chemical, biological and material sciences. Besides contributing qualitative information on molecules and their different interactions, modern quantum chemistry may also provide a deeper understanding of molecular processes, which cannot be solely derived from such experimental work as elusive chemical reaction intermediates. ‘The majority of experimental results are supported by modern computational work, and theory is now used more than ever as an apparatus to lead and guide future experimental work within the medicinal industry and material sciences,’ says Poul Jorgensen, a researcher with the EU-funded LCC (Coupled Cluster Calculations on Large Molecular Systems) project. ‘As a result, accurate computer simulations on increasingly larger molecular systems are desired, not only by academia, but also by various industrial research labs.’ However, according to Jorgensen, the problem is that computation efforts grow dramatically with the size of the molecular system whenever an increasingly higher accuracy is requested. ‘In order to circumvent this computational problem, so-called local correlation methods have been devised, which describe the fundamental repulsive interactions between individual electrons in a spatially local manner instead of the typical delocalised, canonical manner,’ he says. Improving the LSDalton code Jorgensen was part of the research team that developed the quantum chemistry code called LSDalton – a massively parallel and linear-scaling programme used for the accurate determination of energies and other molecular properties for large molecular systems. Now, through the LCC project, Jorgensen and his team have further developed the LSDalton code. ‘The ultimate goal of this project was to obtain cluster methods that scale linearly with system size and where the calculations are massively parallel, such that calculations for small and large molecular systems require the same computational wall time,’ says Jorgensen. The key to accomplishing this goal was to express the coupled cluster wave function in a basis of local Hartree-Fock (HF) orbitals. ‘We successfully demonstrated how such a local HF basis may be obtained and described how linear-scaling, massively parallel coupled cluster energies can be obtained,’ explains Jorgensen. ‘At different levels of coupled cluster theory we performed efficient massively parallel calculations for the energy and the molecular gradient, and in the future the same technology will be applied for even higher-level coupled cluster methods to yield not only the energy and gradient of a large molecule, but also other molecular properties such as excitation energies and transition moments, nuclear shielding, polarisabilities and electronic and vibrational circular dichroism.’ A new era for quantum calculations The developments achieved in the LCC project will open a new era of accurate quantum calculations on large molecular systems such as nanoparticles and proteins. ‘The improved performance has the potential to benefit all areas of molecular science and engineering by enabling increases in both the maximum molecular system size that can be simulated and the overall accuracy achievable,’ explains Jorgensen. The developments are particularly interesting in the context of supercomputers, where the wall time to solution is the most important measure and the LSDalton program may be efficiently utilised. This has been recognised at Oak Ridge National Laboratory (ORNL) in the US, where one of the largest supercomputers, TITAN, is located and where the world’s largest supercomputer SUMMIT will be operational in about a year. Here, LSDalton was successfully utilised to perform extremely large correlational calculations on TITAN. ‘The LSDalton programme is now being further developed at ORNL and will soon be ready to explore new challenging applications on SUMMIT on for example carbon nanotubes, graphene and the preferred crystal form of organic molecules,’ says Jorgensen.
Keywords
LCC, European Union EU, LSDalton, supercomputers, quantum chemistry
