The Development And Application Of Parallel High Performance Dissipative Particle Dynamics Code

The self-assembly of block copolymers in dilute solution into complex structures, such as vesicles, giant micelles and multicompartment micelles, had received scientists'attentions, mainly due to their wide applications, for example, in the drug delivery and nanotechnology. A lot of factors, for example, the structure of the multiblock copolymer (including the chemical constitution, the relative lengths of the individual block, and the chain architecture) and the property of the solution (such as the concentration, the temperature, PH, and the specific solvent) can influence the morphology and structure of the self-assembly of block copolymers, which must be explored in a wide parameter space, taking a huge number of experimental effort. Therefore, molecular simulation technique, as a time and cost efficient tool, can not only complement experimental works with a wide parameter space, but also give a preview of phenomena prior to experiments. Dissipative particle dynamics (DPD) method is a mesoscopic simulation technique and very suitable to study the complex structures from block copolymers in dilute solution and dynamical behavior. However, these complex structures such as multicompartment micelles are formed in the dilute solution, and there are many solvent molecules that should be considered to build the suitable model. Thus, very large simulation system is requested and these computations couldn't be finised by only one computer. Therefore, DPD algorithm should be parallelized to study the systems such as complex structures self-assembled from block copolymers in dilute solution."A workman must first sharpen his tools if he is to do his work well". There are few applications of parallel dissipative particle dynamics until now. So the development of parallel dissipative particle dynamics is very important. In this dissertation, first realization is to develop parallel dissipative particle dynamics simulation method, try to improve the parallel performance, and build the source code to provide cross-platform support. The realization of the parallel dissipative particle dynamics is not only the key prerequisite to finish this dissertation, but also the important feature of this dissertation, especially could provide good basis of other investigations in the future.In this dissertation, we develop parallel dissipative particle dynamics code using a spatial domain decomposition method with the aid of standard Message Passing Interface (MPI). We also investigate the application of parallel dissipative particle dynamics on the spinodal decomposition, multicompartment micelles self-assembled from block copolymers in dilute solution, and multicomparment micelles formed from the mixture of diblock copolymers and homopolymer in dilute solution. The main contents include:(1) Development of parallel high performance dissipative particle dynamics. We develop parallel dissipative particle dynamics codes in two and three dimensions using a spatial domain decomposition method with the aid of the standard MPI. After massive tests, we confirm that the numerical results obtained by using parallel high performance DPD algorithm are consistent with those from serial DPD code. Thus, the parallel high performance DPD algorithm is valid and accurate. The parallel high performance DPD code satisfies large size requirement. For example, the system of 192×103 DPD beads can be computed using parallel high performance dissipative particle dynamics with 8 CPU. But, only 24×103 DPD beads can be performed using serial dissipative particle dynamics with 1 CPU. Thus, the computation ability of parallel code is seven times larger than that of serial code. Parallel high performance dissipative particle dynamics also has high efficiency, for example, Amdal speedup ratio (κs ) is 43.79 by using 8 CPU to simulate 98.304×103. This means that the work of 43 days using serial code can be performed in only 1 day. The parallel high performance dissipative particle dynamics code also has good portability. It has been compiled in the condition of different hardwares including SGI, IBM, Dawning servers, and individual computers, and has been tested suscessfully in the enviroments of different compilers. So it has the characteristic of cross-platform. The parallel high performance dissipative particle dynamics has been the important part of software DPD-JLU of State Key Laboratory of Theoretical Computational Chemistry in Jilin University and one of the most important investigation methods.(2) To study the finite size effects of the system, domain growth processes of spinodal decomposition with different quenching depth in two and three dimensional binary immiscible fluids are investigated by using parallel high performance dissipative particle dynamics. In two dimensions, the dynamic scaling exponent 1/2 for coalescence and 2/3 for inertial regimes in the shallow quench and strong finite size effects in the cases of deep quenching were obtained. In three dimensions, it was used that the diffusive regime with exponent n=1/3 in the shallow quench and the inertial hydrodynamic regime with n=2/3 for different quenches. The viscous effects are not clearly reflected, showing n=1/2 in both shallow and deep quenches in this time period, due to the soft nature of interaction potential adopted in dissipative particle dynamics.(3) Parallel high performance dissipative particle dynamics simulation is performed to study the sensitive influence of the molecular architecture and/or segment sequence on the morphology diversity of the multicompartment micelles. The morphologies of multicompartment micelles formed from ABC triblock copolymers with various molecular architectures, such as the linear, the pentalinear, the cyclic, the star-like, the tetra-arm, and theπ-shape are investigated, and different morphologies of the multicompartment micelles, for example, worm-like,"hamburger", sheet-like with pores,"sweet potato"with alternating layers, sheet-like with cylinder-inclusion, and three-dimensional network are observed in this work. The density profiles and the radial distribution functions are calculated to characterize the structures of the multicompartment micelles. The preparation of complex multicompartment micelles in experiments can be fulfilled by simply changing the segment sequence and molecular architecture.(4) Parallel high performance dissipative particle dynamics (DPD) method is applied to study the self-assembly of diblock copolymer Poly(Ethyl Ethylene)-block-Poly(Ethylene Oxide) (PEE-b-PEO) and homopolymer Poly(Propylene Oxide) (PPO) in aqueous solution. We adopt a systematic coarse-graining strategy that several segments are coarse-grained into a single bead, and the interaction parameters between beads are estimated from the Flory-Hugginsχparameters. Our results in silico are in agreement with those in the experiments not only the micelle morphologies but also on the micelle sizes. We then adopt the approach of mixing PEE-b-PEO with PPO in water and study the conditions of forming multicompartment micelles by changing the polymer chain lengths and the volume fraction ratios between PPO and PEE-b-PEO. The core-shell-corona micelle is observed in the condition of short chain length of PEE-b-PEO, whereas the micelle with two spherical compartments is formed in the condition of long chain length of PEE-b-PEO. The results in this work primarily complement the experiments with the possibilities on the preparation of multicompartment micelles using a more simple and economical approach.