DL_POLY 25^th Anniversary

It was my pleasure to organise and celebrate the 25^th anniversary of the DL_POLY project as a special event at the Royal Society’s Chicheley Hall in November 2017. This special issue of Molecular Simulation is a humble outcome of that special event and a follow-up from the two previous reviews of the project in the journal in 2006 (Vol. 32(12–13), pp. 933–993) and in 2002 (Vol. 28 (5), pp. 385–471). The issue contains a collection of articles that showcase the diversity of modelling and simulation research assisted and inspired by the DL_POLY project and the people behind it. This diversity includes and acknowledges the hidden and often unrecognised efforts of ‘software creators’ or ‘hack-a-demics’ in methodology creation and application development by means of software engineering and continuous program improvement. These efforts are an integral part of the project and very much at the heart of the DL_POLY’s home community, CCP5. CCP5 is the UK’s Collaborative Computational Project for the computer simulation of condensed phase materials at length scales spanning from the atomistic to the mesoscopic and it has been a long-standing driving force in advancing research in this area; DL_POLY was originally developed to meet demand from CCP5 for a general purpose package.

This series of articles begins with a biologically oriented paper investigating the pressure reversal mechanisms of general anaesthetics on membranes by Chau. It provides arguments based on molecular dynamics simulations as to why the natural hypothesis responsible for this effect – the aggregation of halothane molecules within DMPC and POPC bilayers – cannot be the key mechanism.

A discussion of the range of challenges – theoretical, numerical and software – when addressing the multiscale problem for mesoscale bio-simulations is provided by Melchionna et al. The article addresses the efficiency versus realism arguments for optimal combination when coupling molecular dynamics and lattice Boltzmann techniques.

Kaiser et al. discuss the use of neural network potentials and how to use them to represent the potential energy function of a system. It is followed by a description of the interface between the eanet library (of Artith and Urban) and the DL_POLY program. Its use is showcased by results for liquid water, with a detailed comparison of the results with those from a classical simulation.

The interplay between simulation and experiment to guide the improvement of force-field interactions for the single point charge water model is discussed by English et al. The authors describe the systematic Design-of-Experiments, factorial-design approaches they took to minimise the differences in diffusivity and radial distribution functions between simulation and experiment.

DL_MONTE, following in the footsteps of DL_POLY, is an ambitious project that has delivered a general purpose, atomistic Monte Carlo program. Underwood et al.’s long article describes the CCP5 Flagship Project and its outcomes in detail. The methods within DL_MONTE and the package’s versatile functionality are illustrated by two studies: free energy profiling of a lipid translocation through a lipid bilayer, and examination of the thermodynamic stability of two plastic crystal phases of water at high pressure.

Q. Parker et al. employ molecular dynamics to study comprehensively three different ionic liquids with two different force-fields to represent the interactions in these systems. They calculate a range of properties at a number of temperatures and compare and contrast the results and their sensitivity with experiment.

Dove et al. investigate the validity of the conventional Lorentz–Bertholet mixing rules for the Lennard–Jones potentials on mixed water and methane systems. Ab initio and classical molecular dynamics simulations of methane hydrate clathrates are compared and contrasted to determine the crucial interactions that these mixing-rules fail to address.

S.C. Parker et al. describe a computational workflow for modelling the sorption and transport of molecular hydrogen in cellulose frameworks. The study utilises a number of DL packages – DL_POLY, DL_MONTE and DL_FIELD in a concerted manner to provide insight into the behaviour of amorphous cellulose.

Seaton gives a perspective on the history and capability of another excellent simulation package, DL_MESO, that uses a technology similar in spirit to DL_POLY to address the mesoscopic scale via dissipative particle dynamics (DPD). The article gives information on the current performance of DL_MESO_DPD and demonstrates its versatility for a selection of applications.

Elena et al. take a detailed technical look into the complexity of dynamic load balancing via Task-Based Parallelism. They use an OpenMP strategy for two-body force calculations within DL_POLY_4 to demonstrate the value of this approach and discuss the challenges of extending it beyond two-body.

The following article by Guest et al. provides a comprehensive performance analysis of the DL_POLY project over the last two decades. It provides details of all project codes to date – DL_POLY_Classic and DL_POLY 3 & 4 – to generate a broad overview of performance across well over 100 HPC systems with discussions of optimum choices of hardware components for best parallel performance.

Seaton et al. discuss the need for inhomogeneous thermostats in high energy molecular dynamics simulations and implementation of the two-temperature model in DL_POLY_4. The paper provides high-level technical detail on the Domain Decomposition implementation of the method as well as a couple of example applications to materials of interest.

Sokhan et al. showcase a novel methodological approach of mapping classical molecular dynamics onto DPD. Their methods provide a new molecular foundation linking the atomistic and DPD levels of representation and a practical route to extract the DPD parameters as direct statistical averages over the full atomistic system.

Rosebottom et al. demonstrate the power of the integrated use of DL_POLY, DL_FIELD and DL_ANALYSER in a workflow to explore and correlate synthonic interactions in organic solute-solvent systems. The article investigates the solution chemistry of saturated benzoic acid in hexane solutions using a multi-scale and multi-technique approach to explore the transition from solution to crystal at the molecular level.

Archer et al. carry out very large molecular dynamics simulations to investigate the effects of radiation damage on crystalline matrices via multiple cascades in a range of gadolinium pyrochlore solid solutions Gd₂(Zr_xTi_1-x)₂O₇. The damage in these systems is analysed by calculating Steinhardt local order parameters, establishing the key factors and mechanisms that lead to improved healing of the damage and increase resistance to amorphisation. The authors highlight the importance of ion mobility, associated with the weaker Zr–O bonds, in the healing process.

The last article, by Bill Smith - the originator of the DL_POLY project and many others, provides a new, neat algorithm for rotational motion in molecular dynamics simulations. The algorithm requires neither quaternions nor Euler angles and works by updating the local Cartesian axes of the rotating body, the components of which also provide the rotation matrix. The algorithm’s algebraic simplicity and numerical implementation provide computationally cheaper rotational motion integration with well-bounded errors.

Not only do the articles presented in this issue demonstrate the versatility of computational research made possible by the DL_POLY program but they also showcase the impact the project has generated in methodology development, computer science and software engineering. The project's success, thanks to the perseverance needed to ensure its continuity over a long period, has inspired further software initiatives at Daresbury Laboratory (the DL_Software ecosystem) and elsewhere. These also include the development of analysis tools and workflows for processing simulation results from DL_Software. On behalf of all the DL_POLY project authors, I would like to thank all the contributors to this special issue for their excellent articles and for promoting the use of DL_POLY and the whole ecosystem of packages, known collectively as DL_Software, which includes DL_POLY, DL_MESO, DL_MONTE, DL_FIELD, DL_ANALYSER and ChemShell.