Accelerating The Computation Of Electrostatic Interactions In Dissipative Particle Dynamic Simulations

The computation of electrostatic interactions has always been the bottleneck of efficiency in molecular dynamic simulations. It accounts for the vast majority of running time in common simulations. In many simulations with little influence by electrostatic interactions, electrostatic computation is omitted for the consideration of efficiency, which affects the result accuracy. However, the simulation processes are time-consuming and tough jobs for simulations with electrostatic interactions. Therefore, in past decades, many researchers have been working on looking for efficient ways of electrostatic computation. In 1921, Ewald proposed a way of splitting electrostatic energy into three parts, respectively in real space, reciprocal space and self space, which has become a standard algorithm architecture for efficiently evaluating electrostatic energy. Afterwards, the PPPM method based on gridding algorithm by Hockney and Eastwood and PME method by Darden were presented as modified results of Ewald summation. In 2006, F. Hedman and A. Laaksonen proposed a new way of computing electrostatic interactions, namely ENUF, which first executes electrostatic computation in reciprocal space using non-uniform fast Fourier transform(NFFT) and lower executing time effectively. The tasks of this dissertation are as followings:(1) The emphasis in this dissertation is how to significantly improve the computing efficiency of electrostatic interaction based on ENUF method. First of all, we adopt a neighboring algorithm Near Distance to reduce interacting cells in neighbor-list when computing real-space electrostatic interactions, which effectively lowers computing time in real-space part. Furthermore, the whole electrostatic computations are implemented by GPU parallel mechanism based on NVIDIA CUDA technique. Especially, we develop a package of parallel non-uniform fast Fourier translation based on a new gridding algorithm and optimize it to the best performance according to computing characteristics in reciprocal-space part, which makes the efficiency improve dramatically in reciprocal space. This upgraded algorithm for computing electrostatic interactions is called CU-ENUF because of its technique's features based on ENUF and CUDA. Finally, a series of benchmarks are conducted to test the accuracy and performance of CU-ENUF. The results reflect that the efficiency increases by several orders of magnitude compared to traditional ENUF when setting the same decree of accuracy.(2) Madelung constant is non-dimensional and related to ion crystal binding energy. Its value depends only on lattice structure and is unique for one kind of crystal. In this dissertation, we calculate electrostatic lattice energies of Na Cl and Ca F2, through which their Madelung constants are obtained respectively. Compared with standard Madelung reference list, we validate the results calculated by CU-ENUF.(3) Incorporating CU-ENUF into DPD simulations, this dissertation studies the typical conformational distributions of polyelectrolyte with different charge ratio under condition of the same salt concentration. In the study, beads on the single chain of polyelectrolyte are positively charged in proportion. Addition of the same amount of counter ions in solution assures electric neutrality in whole system. Additionally, salt ions with a certain concentration are added into solution with the same amount of positive and negative ions. We measure degree of collapse or expansion of polyelectrolyte by calculating gyration radius of the single chain for different cases of charge proportion. The results reveal that the comformation of polyelectrolyte changes from complete collapse to full expansion while increasing proportion of charged beads on the chain. In the case of fully charged beads on the single chain, when increasing salt concentration in the system gradually, the chain conformation begins to shrink from full expansion because of effect of electrostatic screening by salt ions.(4) Due to its unique properties in many aspects, polyzwitterions(PZs) have been the focus of much scientific interest. Muthukumar et al. [Morozova, S.; Hu, G.; Emrick, T.; Muthukumar, M. ACS Macro Lett. 2016, 5, 118-122] have studied solution properties of poly(2-methacryloyloxyethyl phosphorylcholine)(PMPC) and n-butyl-substituted choline phosphate polymers(PMBP) using dynamic and static light scattering methods. Their results indicated that PMPC and PMBP exhibit dramatically different solution properties. In this dissertation, by using coarse-grained molecular dynamics simulations, we found that PMPC showed a strong solubility and a so-called antipolyelectrolyte effect(APE) in water. In contrast, obvious aggregation but no APE was found in PMBP solutions. The underlying mechanisms for the different solution behaviors of PMPC and PMBP were investigated in detail. Our results indicate that the presence of butyl groups in PMBP enhances both electrostatic interactions and hydrophobicity of PMBP molecules in the system, and both factors are found contributing to the formations of aggregates in PMBP system. Further researches reveal that hydrophobicity arising from butyl group plays more important role than electrostatic interactions in inducing the PMBP aggregation. In addition, strong hydrophobicity in PMBP is found being responsible for the absence of APE. These results are expected to contribute to a better understanding and a better design of the solution properties of polyzwitterions.