Automated reaction mechanism generation : data collaboration, Heteroatom implementation, and model validation

Abstract

Nearly two-thirds of the United States' transportation fuels are derived from non-renewable fossil fuels. This demand of fossil fuels requires the United States to import ~ 60% of its total fuel consumption. Relying so heavily on foreign oil is a threat to national security, not to mention that burning all of these fossil fuels produces increased levels of CO₂, a greenhouse gas that contributes to global warming. This is not a sustainable model. The United States government has recently passed legislation that requires greenhouse gas emissions to be reduced to 80% of the 2005 level by the year 2050. Furthermore, new legislation under the Energy Independence and Security Act (EISA) requires that 36 billion gallons of renewable fuel be blended into transportation fuel by 2022. Solving these types of problems will require the fuel industry to shift away from petroleum fuels to biomass-derived oxygenated hydrocarbon fuels. These fuels are generated through different biological pathways, using different "bugs." The question of which fuel molecules should we be burning, and thus, which bugs should we be engineering, arises. To answer that question, a detailed understanding of the fuel chemistry under a wide range of operating conditions, i.e. temperature, pressure, fuel equivalence ratio, and fuel percentage, must be known. Understanding any fuel chemistry fully requires significant collaboration: experimental datasets that span a range of temperatures, pressures, and equivalence ratios, high-level ab initio quantum chemistry calculations for single species and reactions, and a comprehensive reaction mechanism and reactor model that utilizes the theoretical calculations to make predictions. A shortcoming in any of these three fields limits the knowledge gained from the others. This thesis addresses the third field of the collaboration, namely constructing accurate reaction mechanisms for chemical systems. In this thesis, reaction mechanisms are constructed automatically using a software package Reaction Mechanism Generator (RMG) that has been developed in the Green Group over the last decade. The predictive capability of any mechanism depends on the parameters employed. For kinetic models, these parameters consist of species thermochemistry and reaction rate coefficients. Many parameters have been reported in the literature, and it would be beneficial if RMG would utilize these values instead of relying on estimation routines purely. To this end, the PrIMe Warehouse C/H/O chemistry has been validated and a means of incorporating said data in the RMG database has been implemented. Thus, all kinetic models built by RMG may utilize the community's reported thermochemical parameters.

(cont.) A kinetic model is evaluated by how accurately it can predict experimental data. In this thesis, it was shown that the RMG software, with the PrIMe Warehouse data collaboration, constructs validated kinetic models by using RMG to predict the pyrolysis and combustion chemistry of the four butanol isomers. The kinetic model has been validated against many unique datasets, including: pyrolysis experiments in a flow reactor, opposed-flow and doped methane diffusion flames, jet-stirred reactors, shock tube and rapid compression machine experiments, and low-pressure and atmospheric premixed laminar flames. The mechanism predicts the datasets remarkably well across all operating conditions, including: speciation data within a factor of three, ignition delays within a factor of two, and laminar burning velocities within 20% of the experimental measurements. This accurate, comprehensivelyvalidated kinetic model for the butanol isomers is valuable itself, and even more so as a demonstration of the state-of-the-art in predictive chemical kinetics. Although the butanol kinetic model was validated against many datasets, the model contained no nitrogen-containing species, and also had limited pathways for benzene formation. These limitations were due to the RMG software, as RMG was initially written with only carbon, hydrogen, and oxygen chemistry in mind. While this functionality has been sufficient in modeling the combustion of hydrocarbons, the ability to make predictions for other chemical systems, e.g. nitrogen, sulfur, and silicon compounds, with the same tools is desired. As part of this thesis, the hardcoded C/H/O functional groups were removed from the source code and database, enabling our RMG software to model heteroatom chemistry. These changes in the RMG software also allows for robust modeling of aromatic compounds. The future in the transportation sector is uncertain, particularly regarding which fuels our engines will run on. Understanding the elementary chemistry of combustion will be critical in efficiently screening all potential fuel alternatives. This thesis demonstrates one method of understanding fuel chemistry, through detailed reaction mechanisms constructed automatically using the RMG software. Specifically, a method for data collaboration between the RMG software and the PrIMe Warehouse has been established, which will facilitate collaboration between researchers working on combustion experiments, theory, and modeling. The RMG software's algorithm of mechanism construction has been validated by comparing the RMG-generated model predictions for the combustion of the butanol isomers against many unique datasets from the literature; many new species thermochemistry and reaction rate kinetics were calculated and this validation shows RMG to be a capable tool in constructing reaction mechanisms for combustion. Finally, the RMG source code and database have been updated, to allow for robust modeling of heteroatom and aromatic chemistry; these two features will be especially important for future modeling of combustion systems as they relate to the formation of harmful pollutants such as NOx and soot.