Computing Sciences masthead Berkeley Lab Computing Sciences Berkeley Lab logo
Share/Bookmark

Exascale for Energy


Renewable Fuels: Bioenergy Conversion

Alternative and renewable fuels derived from biomass offer the potential to reduce our dependence on imported oil, support national economic growth, and mitigate global climate change. However, technological breakthroughs are needed to overcome key barriers to the development and commercialization of these fuels. These barriers include the high cost of pretreatment processes, enzymes, and microbial biocatalysts for biochemical conversion processes.

The lignocellulose in biomass is highly recalcitrant to most of the physical, chemical, and biochemical treatments currently used to liberate sugars. The cell walls of lignocellulose contain highly ordered, water-excluding microfibrils of crystalline cellulose that pose a significant barrier to enzymatic hydrolysis. The cellulose microfibrils themselves are laminated with hemicellulose, pectin, and lignin polymers. This complex matrix of heteropolymers is the main reason why plant biomass has resisted low-cost chemical and enzymatic treatments.

Cellulose fibril
Figure 22. A simulated cellulose fibril showing (a) the cross-section and (b) a side perspective. The fibril consists of 18 origin chains (blue) and 18 center chains (green). The axes of the unit cell are also indicated. Source: T. Splettstoesser, ORNL

Cellulases can be used to hydrolyze the polysaccharides in the plant cell wall to fermentable monosaccharides, but the large quantities of expensive cellulases that are currently needed make the process cost-prohibitive. Despite Herculean attempts, the specific activity of cellulases has not been improved after more than three decades of research. A better understanding of the structure, function, and relationships governing the activity of soluble enzymes on insoluble polymeric substrates is essential to break this bottleneck. Computation uniquely provides a multiscale framework of understanding to guide and interpret experimentation on complex biological systems.

The priority research directions for biomass conversion identified in the SC/EE workshop include understanding lignocellulosic biomass depolymerization and hydrolysis, and chemical energy extraction from heterogeneous biomass. At Oak Ridge National Laboratory’s BioEnergy Science Center (BESC), a new method for molecular dynamics simulation of lignocellulosic biomass is currently being tested.

BESC researchers have developed a strategy for a fast all-atom molecular dynamics simulation of multimillion-atom biological systems on massively parallel supercomputers, using models of cellulose and lignocellulosic biomass in an aqueous solution. Their approach involves using the reaction field (RF) method for the computation of long-range electrostatic interactions, which permits efficient scaling on many thousands of cores. Although the range of applicability of the RF method for biomolecular systems remains to be demonstrated, for the benchmark systems the use of the RF produces molecular dipole moments, Kirkwood G factors, other structural properties, and mean-square fluctuations in excellent agreement with those obtained with the commonly used Particle Mesh Ewald (PME) method.

Due to its complexity, lignocellulose poses significant challenges to molecular dynamics simulation. Among these are the characteristic length scales (Å–μm) and time scales (ns–μs and beyond) of events pertinent to the recalcitrance of biomass to hydrolysis into sugars. To access these length and time scales, standard molecular dynamics protocols must be modified to scale up to massively parallel machines.

The BESC researchers simulated a system of lignocellulosic biomass containing 52 lignin molecules each with 61 monomers, a cellulose fibril of 36 chains with 80 monomers per chain (Figure 22), and 1,037,585 water molecules, totaling 3,316,463 atoms.

Their studies showed that the properties derived using the PME method are well reproduced using the computationally less demanding reaction field method. Scaling benchmarks showed that the use of RF drastically improves the parallel efficiency of the algorithm relative to PME, yielding ~30 nanoseconds of simulated reaction time per computing day, running at 16.9 teraflops on 12,288 cores of ORNL’s “Jaguar” Cray XT5 system. Consequently, microsecond time scale molecular dynamics simulations of multimillion-atom biomolecular systems now appear to be within reach.

Despite this important advance, some critical biological phenomena, such as ligand binding, require the simulation of relatively long time scales (up to 1000 seconds). For this type of application, exascale computing will be required.

 


<< Previous page