Generalizable chemical mechanism extraction from molecular dynamics simulations
- Hydrocarbon pyrolysis is a process that is thought to occur in the interior of icy giant planets. This reaction would transform the methane contained in the interior of these planets into diamond through compression to pressure of several tens of gigapascals and heating to temperatures of several thousands of kelvins. Even if it is commonly admitted that this reaction occurs, the conditions under which it takes place are still debated. Many experiments and theoretical calculations gave disagreeing predictions; some experiments led to the observation of diamond from hydrocarbon pyrolysis in pressure as low as 10 GPa and temperatures between 2,000 and 3,000 K, whereas other experiments and theoretical predictions claimed that diamond cannot be produced by hydrocarbon pyrolysis below 100 GPa. Several explanations have been proposed to explain these discrepancies; however, there is no consensus yet and one of the main reasons is that the precise mechanism of hydrocarbon pyrolysis is not precisely known. A crude mechanism is that methane separates into dihydrogen and carbon-rich molecules, such as large hydrocarbons or solids. Nevertheless, this simplified mechanism leaves many topics unstudied: first, the kinetic of the reaction remains unknown. This limitation implies that researchers cannot guarantee that their experiments or simulations were run for a sufficient time to reach equilibrium. Another poorly understood topic is the role of pressure and temperature; it is generally assumed that diamond is more likely to be synthesized when pressure and temperature increases. But the exact conditions in which diamond is produced and the role of pressure and temperature has not been studied yet. Finally, the role of the hydrogen content has not been investigated; Le Chatelier's principle implies that the less hydrogen in the system, the larger the hydrocarbons will be, but it is unknown which hydrogen content induces diamond production. Therefore, there is a need for a detailed mechanism of hydrocarbon pyrolysis. Discovering the hydrocarbon pyrolysis mechanism through experiments is impractical because of the extreme conditions undergone and the short time scales involved. Recently, the rise of reactive force fields, such as ReaxFF, for Molecular Dynamics (MD) simulations allowed researchers to extract mechanisms from these MD simulations. Several methods have been developed for this extraction; however, the state-of-the-art methods display some limitations making them unsuitable for the study of hydrocarbon pyrolysis. First, these methods are only appropriate for the study of small molecules; since large molecules or solid particles appear in hydrocarbon pyrolysis, a framework that can both describe short and long molecules is needed. Second, the method cannot rely on theories such as Transition State Theory that were developed for gases. Hydrocarbon pyrolysis involves liquid and solid phases making these theories inappropriate. Eventually, the mechanisms are usually fitted to some specific conditions of temperature and pressure, but no extrapolation to new conditions have been performed. These extrapolations are especially important in low-reactivity conditions (low temperature and pressure) where MD simulations cannot reach equilibrium in reasonable computational time. In this dissertation, I am presenting the developments to mechanism extraction from MD simulations that made this method suitable for hydrocarbon pyrolysis. Mechanism extraction was performed on training MD simulations of hydrocarbon pyrolysis. From these simulations, the possible reactions that can occur in this system were extracted and their reaction rates estimated. From the list of reactions in the mechanism and their associated reaction rates, the evolution of hydrocarbon pyrolysis in different conditions was predicted by running Kinetic Monte Carlo (KMC) simulations that served here as an ordinary differential equation solver. This general framework was improved during my PhD. As a first contribution, reactions were described at the atomic scale, considering only the local structure around the reactive site and ignoring the rest of the molecules. This description differs from common mechanism where the reactions are described with the full molecules. This new description allowed me to describe both short and long molecules, condense the number of reactions to describe the system, predict the appearance of molecules unobserved in the initial MD simulations, transfer to the pyrolysis of new hydrocarbons, and extrapolate MD simulations to longer timescales. My second contribution was to perform temperature extrapolation on the extracted mechanism. This extrapolation enabled me to make accurate predictions of the evolution of different hydrocarbon pyrolysis in temperatures that were outside of the temperature range of the MD simulations that were used to extract the mechanism. I was able to make accurate predictions several hundred kelvins outside of this temperature range, run simulations on the microsecond-scale that would be too long to run for an MD simulation, and give insights to some experimental results of hydrocarbon pyrolysis. My final contribution involved the development of a model that predicts the size distribution of hydrocarbon at equilibrium in many different conditions from a simplified mechanism and random graph theory. This model can give quickly and accurately the equilibrium conditions that would otherwise need an MD simulation or a KMC simulation to be predicted.
|Type of resource
|electronic resource; remote; computer; online resource
|1 online resource.
|Dufour, Vincent Jean Louis
|Blanchet, Jose H
|Blanchet, Jose H
|Wang, Hai, 1962-
|Degree committee member
|Wang, Hai, 1962-
|Stanford University, Department of Materials Science and Engineering
|Statement of responsibility
|Submitted to the Department of Materials Science and Engineering.
|Thesis Ph.D. Stanford University 2022.
- © 2022 by Vincent Jean Louis Dufour
Also listed in
Loading usage metrics...