SI2-CHE: Development and Deployment of Chemical Software for Advanced Potential Energy Surfaces

Molecular dynamics simulations provide a powerful tool to study a wide range of chemical and biochemical systems. The simulations provide a movie of the atoms diffusing over time, from which important dynamic and thermodynamic properties may be calculated. The reliability of these simulations is, however, limited by the accuracy of the model used to describe how the atoms in the system interact with each other. Conventional force fields, in which fixed charges are used, represent a major factor limiting the successful application of computer simulations to a variety of grand challenge problems in computational chemistry, biochemistry and materials science. Polarizable empirical force fields, which offer a clear and systematic improvement by allowing atom-centred charges to change depending on their environment, have been recently developed by most major research groups in force field development to increase accuracy. These advanced potential energy surfaces are important for the future of grand challenge applications such as the design of environmentally friendly materials, chemical reactions and reactivity critical for chemical synthesis, and biological complexity such as protein-drug interactions.

However, there are obstacles to using advanced potential energy surfaces for these grand challenge chemistry problems: the computational cost of the models, limited dissemination to a broad range of community codes, and lagging quality software implementations on HPC architectures and newer GPU and multicore hardware. To address these issues, we have organized a UK and US consortium that represents a broad cross section of the computational chemistry software community involved in chemical and biochemical applications, force field development, electronic structure methods, molecular dynamics algorithms, and software engineering with computer science experts. In this project, state-of-the-art polarizable potential energy functions will be consistently implemented and tested in a number of the most widely used simulation codes. The latest software development practices will be used to ensure that the freely-available developed software meets the highest standards of robustness, maintainability, and usability. New methodologies to improve the computational performance of these models will also be implemented. An important aspect of our work is to combine these latest force field models with quantum mechanical methods, allowing the accurate modelling of chemical reactivity and excited states. Our international collaboration between US and UK universities and HPC centres will ensure that the investment made in molecular simulation software is successfully deployed on current and emergent hardware and will also realize a long term payoff in community availability and sustainability. This project will lead to a step-change in the use of advanced potential energy surfaces by delivering consistent and sustainable implementations of the latest science on a diverse range of readily available and widely utilised software platforms.

The software platforms within which the developments will be carried out in this proposal, TINKER, Amber, DL_POLY, ONETEP, and Q-Chem, comprise a diverse and popular suite of chemical software programs used for the simulation of molecular properties in gas or condensed phases. The close collaboration of experts in force field and electronic structure codes proposed here will allow us to develop important new methods and to communicate effectively to the wider chemical, biological and materials research communities. We aim to bring about a step change in molecular simulations both in the accuracy and the time scales that can be achieved and hence enable the end users to perform simulations at a whole new level of realism.

Academic research groups will benefit by being able to access the latest technology in potential energy surface calculations, consistently implemented in a range of distinctive classical and quantum mechanical computational chemistry codes. Not only will this allow these groups to perform a wider range of more accurate calculations than before, but the free availability of the software developed here will facilitate wider engagement by the academic community in terms of implementing our approaches in yet more codes, together with enabling further method development and optimisation.

The industrial community will also benefit through being able to access these new methods. Accurate potential energy functions allow intermolecular interactions to be described more completely, a crucial aspect leading to improvements in rational material and molecular design. The opportunities arising from advanced potential energy functions will therefore have a clear commercial benefit, improving our economic competitiveness by delivering new and more efficient products (such as greener and more efficient batteries, satisfying our energy needs in an environmentally friendly way), and delivering products that will enhance our quality of life through improved health and longevity (new drugs, for example). An important aspect of our proposal is the implementation and testing of accurate potential energy methods in a diverse range of widely-used software programs using the latest software design and development protocols. This will ensure the appropriate level of trust and confidence needed for software to be used in a commercial environment.

Policy makers will benefit through the more accurate calculations of molecular properties that the developed software will allow. This is particularly important given the current drive to reduce animal testing, of which the European Union REACH regulations are perhaps the clearest example. Our developed software therefore has the potential to impact public policy and legislation.

Another class of beneficiaries of this project will be the manufacturers of cutting edge HPC platforms. One of the main aims of our work is to harness the power of the latest HPC technologies to tackle chemical grant challenges. Through our access to top of the range national supercomputers and their manufactures we will assist them to make capability HPC more accessible to chemical problems. Our consortium is large enough to leverage the HPC manufacturers to view chemical applications as important user-cases when designing the next generation supercomputers.

Finally, the training of scientists working within our groups will make a significant impact. All will receive rigorous training in computational science, numerical methods and parallel programming, within a collaborative environment characterised by an ethos that embraces best practice in software development. This grant will therefore help to fill the skills gap recently identified by several international reports by delivering highly trained individuals with combined skills in computational science and software development.