| The objective of this research was to develop and apply new efficient multibody dynamics algorithms to improve the computational efficiency of coarse-grained molecular dynamics simulations. As a part of this research, a suite of four new, efficient algorithms were developed, tested and validated. This suite includes algorithms for (a) dynamics of multi-flexible-body systems, (b) dynamics of constrained multibody systems, (c) sensitivity analysis of multibody systems and (d) impulse-momentum formulation for multibody systems with discontinuous system definitions. These algorithms are all processor and time optimal, providing linear and logarithmic complexity in serial and parallel implementations respectively. The key advantages of these algorithms are (i) superior computational efficiency for dynamics analysis, (ii) robust handling of singular configurations and excellent constraint satisfaction, (iii) efficient adaptive changes in system definition without restarting a simulation, and (iv) concurrent sensitivity analysis eliminating backward integration, numerical perturbations and data storage issues.; The foundational aspects of a multi-scale (or multi-resolution) method based on multibody dynamics has been developed in this research for coarse-grained molecular dynamics simulations. The algorithms have been used for long time molecular dynamics simulations of bulk materials such as water, polymer chains and alkanes, and biomolecular simulations of Rhodoposin, Calmodulin, Ribosomal and RuBisCo proteins. These systems span several thousand degrees of freedom and involve several orders (10-15 -- 10-9 seconds) of temporal scales. On an average, the simulations were carried out with an order of magnitude increase in the integration time step. This resulted in significant computational savings, with simulations which took several days, being completed in several hours.; A C++ object oriented research software POEMS has been developed and houses implementations of the efficient algorithms. This software has been coupled with the molecular dynamics package LAMMPS from Sandia National Laboratories and is distributed freely with an open source license in the public domain. |
