Theoretical Methods For Reaction Mechanism In Complex Reaction Systems

Combustion of fuels and thermal decomposition of energetic materials involve complex reaction processes.For combustion,studying the chemical reaction mechanisms in detail can facilitate the design of cleaner and more efficient fuels,which can reduce the emission of pollutants during the combustion process.While,exploring the thermal decomposition reaction mechanisms of energetic materials is helpful for studying the properties of energetic materials and provides theoretical support for the design of novel energetic materials.Reactive molecular dynamics（MD）simulation makes it possible to study the reaction mechanism of complex reaction systems at the atomic level.However,the analysis of simulation trajectories containing thousands of species and reaction paths has become a major obstacle to the application of reactive molecular dynamics simulations in large-scale systems.In addition,the existing reactive molecular dynamics simulation methods still need to be improved in terms of accuracy and efficiency.This article focused on how to obtain the reaction mechanism of complex reaction systems with high accuracy and efficiency.Firstly,ReacNetGenerator（Reaction Network Generator）method is developed,which can automatically extract reaction networks from reaction trajectories without any predefined reaction coordinates and elementary reaction steps.Species can be automatically identified from the Cartesian coordinates of the atoms,and Hidden Markov Models（HMM）are used to filter the“noise signal”of species in the trajectories.HMM filter makes the analysis process easier and more accurate.The ReacNetGenerator method has been successfully used to analyze the reactive MD simulation trajectories of combustion of methane and 4-component surrogate fuel for rocket propellant 3（RP-3）.It has great advantages in terms of efficiency and accuracy compared to traditional manual analysis.At present,ab initio molecular dynamics simulation（AIMD）undoubtedly has high accuracy in reactive molecular dynamics simulation.However,due to the expensive computational cost,the scale of the simulation system is very limited,and the simulation time is also limited to the order of picoseconds.The fragment-based ab initio molecular dynamics（FB-AIMD）method for efficient dynamics simulation of the combustion process is developed.In this method,the intermolecular interactions are treated by a fragment-based many-body expansion in which three-or higher-body interactions are neglected,while two-body interactions are computed if the distance between the two fragments is smaller than a cutoff value.The accuracy of the method was verified by comparing energies and atomic forces calculated by FB-AIMD with those calculated by the full system quantum method.The computational cost of the FB-AIMD method scales linearly with the size of the system,and the calculation is easily parallelizable.The method was applied to explore the reaction mechanism of methane combustion.The detailed reaction network was obtained by analyzing the simulation trajectory and some important intermediates were tracked in real time.The current result of simulation for methane combustion is in excellent agreement with experimental findings and prior theoretical studies.However,the computational cost of this method is still relatively high for complex reaction systems with a large scale to perform longer MD simulations.In recent years,the rapid development of machine learning has made it possible to develop a more efficient and accurate method to carry out reactive molecular dynamics simulation.Using molecular dynamics simulations based on neural network potential energy surfaces（NNPs）to explore chemical reaction mechanisms for complex reaction systems is a powerful research tool.CL-20（2,4,6,8,10,12-hexanitro-2,4,6,8,10,12-hexaazaisowurtzitane,also known as HNIW）is one of the most powerful energetic materials.However,its high sensitivity to environmental stimuli greatly reduces its safety and severely limits its application.CL-20/TNT co-crystal has a great improvement in stability.To obtain the molecular mechanism behind the improvement of the stability of CL-20/TNT,the reactive molecular dynamics simulations based on the NNPs were performed to simulate the thermal decomposition processes ofβ-CL-20 and CL-20/TNT co-crystal systems.The performance of the NNPs highly depends on the quality of the reference dataset.To ensure the completeness and low redundancy of the reference dataset,the“concurrent learning”algorithm was employed to construct the preparation process of the reference dataset.The NNPs of the two systems were trained based on their references datasets,respectively.Molecular dynamics simulations were performed at different temperatures based on NNPs.Through the analysis of the simulation trajectories,the detonation temperatures of the two systems were compared.It was found that the detonation temperature of the CL-20/TNT co-crystal system was significantly higher than that ofβ-CL-20,and the thermal stability of CL-20/TNT was improved.Moreover,the simulations also showed clearly that the 2,4,6-trinitrotoluene（TNT）molecules in the co-crystal act as a buffer to slow down the chain reactions triggered by nitrogen dioxide and this effect is more obvious at lower temperatures.The intermolecular hydrogen bonds between CL-20 and TNT molecules hinder the decomposition of CL-20,thus improving the thermal stability of the CL-20/TNT co-crystal.Compared with CL-20,TEX also has a cage-like structure,retains the high density of the crystal structure and low sensitivity.It is an insensitive energetic material with excellent performance.In previous work,it was found that during the thermal decomposition of TEX,clusters with a molecular weight larger than that of TEX were generated,which may have an impact on the stability of energetic materials.In order to explore the formation paths of clusters during the thermal decomposition of TEX,the reactive molecular dynamics simulation based on neural network potential was used in this work.Through the analysis of the simulation trajectories,it was found that there are not only products with relatively small molecular weights,but also species with larger molecular weights than TEX,which were defined as clusters,such as C₈H₈N₃O₄,C₉H₉N₄O₇,C₁₀H₁₀N₄O₅,C₁₂H₁₂N₅O₁₀,etc.These clusters were also detected in experiments using in situ atmospheric pressure photoionization mass spectrometry.By exploring the evolution of the number of species with the simulation time at different temperatures,it was found that when the temperature increases,the clusters（C>=7）would be generated earlier,but they would break down quickly.For the decomposition pathways of TEX,the dissociation of the N-NO₂ bond was the main initial decomposition reaction.After the TEX molecule lost the nitro group,the remaining structure became more unstable.The C-C bond or C-O bond would be broken by subsequent reactions,destroying the cage-like structure of TEX.Fragments produced during thermal decomposition would form clusters through re-polymerization.The reaction potential energy surfaces of some polymerization reactions were calculated by the DFT method.The energy barriers for these reactions were low enough,indicating that these reactions could easily occur during the thermal decomposition of TEX.The formation of clusters would prevent other intermediates from colliding and reacting with each other,which acts as a buffer and improves the stability of TEX.With the powerful fitting ability of the neural network,the potential energy surfaces of the thermal decomposition of different energetic materials were constructed,and the detailed reaction mechanism during the detonation process was explored by NNP-based reactive molecular dynamics simulation.The accuracy of NNPs is as high as DFT calculations,while their computational efficiency has a lead of 3 orders of magnitude.Through the continuous accumulation of training datasets,the reaction mechanism of a series of energetic materials with similar structures could be deeply explored,and the required cost of DFT computation will gradually decrease.The reactive molecular dynamics simulation based on neural network potential energy surface is a very powerful method for exploring the reaction mechanism of complex reaction systems.