Date of Award
Doctor of Philosophy
Biochemistry and Cellular and Molecular Biology
Jeremy C. Smith
Hong Guo, Xiaolin Cheng, Tongye Shen
An understanding of the recalcitrance of plant biomass is important for efficient economic production of biofuel. Lignins are hydrophobic, branched polymers and form a residual barrier to effective hydrolysis of lignocellulosic biomass. Understanding lignin's structure, dynamics and its interaction and binding to cellulose will help with finding more efficient ways to reduce its contribution to the recalcitrance. Molecular dynamics (MD) using the GROMACS software is employed to study these properties in atomic detail. Studying complex, realistic models of pretreated plant cell walls, requires simulations significantly larger than was possible before. The most challenging part of such large simulations is the computation of the electrostatic interaction. As a solution, the reaction-field (RF) method has been shown to give accurate results for lignocellulose systems, as well as good computational efficiency on leadership class supercomputers. The particle-mesh Ewald method has been improved by implementing 2D decomposition and thread level parallelization for molecules not accurately modeled by RF. Other scaling limiting computational components, such as the load balancing and memory requirements, were identified and addressed to allow such large scale simulations for the first time. This work was done with the help of modern software engineering principles, including code-review, continuous integration, and integrated development environments. These methods were adapted to the special requirements for scientific codes. Multiple simulations of lignocellulose were performed. The simulation presented primarily, explains the temperature-dependent structure and dynamics of individual softwood lignin polymers in aqueous solution. With decreasing temperature, the lignins are found to transition from mobile, extended to glassy, compact states. The low-temperature collapse is thermodynamically driven by the increase of the translational entropy and density fluctuations of water molecules removed from the hydration shell.
Schulz, Roland, "Development of High Performance Molecular Dynamics with Application to Multimillion-Atom Biomass Simulations. " PhD diss., University of Tennessee, 2015.