Joseph C. Fogarty

Parallel reactive molecular dynamics: Numerical methods and algorithmic techniques

Molecular dynamics modeling has provided a powerful tool for simulating and understanding diverse systems - ranging from materials processes to biophysical phenomena. Parallel formulations of these methods have been shown to be among the most scalable scientific computing applications. Many instances of this class of methods rely on a static bond structure for molecules, rendering them infeasible for reactive systems. Recent work on reactive force fields has resulted in the development of ReaxFF, a novel bond order potential that bridges quantum-scale and classical MD approaches by explicitly modeling bond activity (reactions) and charge equilibration. These aspects of ReaxFF pose significant challenges from a computational standpoint, both in sequential and parallel contexts. Evolving bond structure requires efficient dynamic data structures. Minimizing electrostatic energy through charge equilibration requires the solution of a large sparse linear system with a shielded electrostatic kernel at each sub-femtosecond long time-step. In this context, reaching spatio-temporal scales of tens of nanometers and nanoseconds, where phenomena of interest can be observed, poses significant challenges. In this paper, we present the design and implementation details of the Purdue Reactive Molecular Dynamics code, PuReMD. PuReMD has been demonstrated to be highly efficient (in terms of processor performance) and scalable. It extends current spatio-temporal simulation capability for reactive atomistic systems by over an order of magnitude. It incorporates efficient dynamic data structures, algorithmic optimizations, and effective solvers to deliver low per-time-step simulation time, with a small memory footprint. PuReMD is comprehensively validated for performance and accuracy on up to 3375 cores on a commodity cluster (Hera at LLNL-OCF). Potential performance bottlenecks to scalability beyond our experiments have also been analyzed. PuReMD is available over the public domain and has been used to model diverse systems, ranging from strain relaxation in Si-Ge nanobars, water-silica surface interaction, and oxidative stress in lipid bilayers (bio-membranes).

Atomically detailed lipid bilayer models for the interpretation of small angle neutron and X-ray scattering data.

We present a new atom density profile (ADP) model and a statistical approach for extracting structural characteristics of lipid bilayers from X-ray and neutron scattering data. Models for five lipids with varying head and tail chemical composition in the fluid phase, 1,2-dioleoyl-sn-glycero-3-phosphatidylcholine (DOPC), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine (POPC), 1,2-dipalmitoyl-sn-glycero-3-phosphatidylcholine (DPPC), 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylserine (POPS), and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylglycerol (POPG), are optimized using a simplex based method to simultaneously reproduce both neutron and X-ray scattering data. Structural properties are determined using statistical analysis of multiple optimal model structures. The method and models presented make minimal assumptions regarding the atomic configuration, while taking into account the underlying physical properties of the system. The more general model and statistical approach yield data with well defined uncertainties, indicating the precision in determining density profiles, atomic locations, and bilayer structural characteristics. Resulting bilayer structures include regions exhibiting large conformational variation. Due to the increased detail in the model, the results demonstrate the possibility of a distinct hydration layer within the interfacial (backbone) region.

Automated optimization of water-water interaction parameters for a coarse-grained model.

We have developed an automated parameter optimization software framework (ParOpt) that implements the Nelder–Mead simplex algorithm and applied it to a coarse-grained polarizable water model. The model employs a tabulated, modified Morse potential with decoupled short- and long-range interactions incorporating four water molecules per interaction site. Polarizability is introduced by the addition of a harmonic angle term defined among three charged points within each bead. The target function for parameter optimization was based on the experimental density, surface tension, electric field permittivity, and diffusion coefficient. The model was validated by comparison of statistical quantities with experimental observation. We found very good performance of the optimization procedure and good agreement of the model with experiment.

Push-based system for molecular simulation data analysis

Many scientific fields generate, and require manipulation of big data. Known scientific data analysis systems, as well as traditional DBMSs, follow a pull-based architectural design, where the executed queries mandate the data needed. This design, while suitable for traditional transaction-based workloads where number of queries retrieve small parts of data located at various places of the database, is ill-fitted for applications involving complex analysis on most of the data. Such design involves redundant and random I/O, considerably affecting the data throughput in the system. In this paper, we design and implement a push-based type system that allows high-throughput data analysis in the process of scientific discovery. Our design improves throughput in two ways: i) it uses a sequential scan-based I/O framework that loads the data into the main memory, and then ii) the system pushes the loaded data to a number of pre-programmed queries. By this way the system lowers the unnecessary I/O overhead imposed by the randomized, index-based scan and that of a multiple data reads if each query were to be fed separately. Considering the amount of data and the number of executed queries, we believe our system provides substantial improvement over the current data analyzing systems. The efficiency of the proposed system is backed by the results of extensive experiments using real MS data. The running times of our system are compared to those of the GROMACS system. The comparison shows the advantage and the potential of using such push-based system for data system analysis.

FFopt: An Automated Molecular Dynamics Force Field Parameter Optimization

An empirical Molecular Dynamics (MD) force-fields usefulness is limited by the accuracy of the parameter set. FFOpt is a general approach for optimization of these parameters. The method explores a vector field in parameter space, searching for minima. The field is determined by the error in the comparison of MD observables with external data. These data can consist of experimentaly determined quantities as well as the results of ab initio quantum calculations. Different geometries are used to generate comparable MD data. The method has been applied to Purdue Reactive Molecular Dynamics (PuReMD). A training set of quantum mechanical data is used to tune a force-field for the simulation of water systems. We present the results of the parameter fitting procedure, as well as comparison of PuReMD data with experimental and quantum data.