Minimizing Redundancy and Data Requirements of Machine Learning Potential: A Case Study in Interface Combustion.

The machine learning potential has emerged as a promising approach for addressing the accuracy-versus-efficiency dilemma in molecular modeling. Efficiently exploring chemical spaces with high accuracy presents a significant challenge, particularly for the interface reaction system. This study introduces a workflow aimed at achieving this goal by incorporating the classical SOAP descriptor and practical PCA strategy to minimize redundancy and data requirements, while successfully capturing the features of complex potential energy surfaces. Specifically, the study focuses on interface combustion behaviors within promising alloy-based solid propellants. A neural network potential model tailored for modeling AlLi-AP interface reactions under varying conditions is constructed, showcasing excellent predictive capabilities in energy prediction, force estimation, and bond energies. A series of large-scale MD simulations reveal that Li doping significantly influences the initial combustion stage, enhancing reactivity and reducing thermal conductivity. Mass transfer analysis also highlights the considerably higher diffusion coefficient of Li compared to Al, with the former being three times greater. Consequently, the overall combustion process is accelerated by approximately 10%. These breakthroughs pave the way for virtual screening and the rational design of advanced propellant formulations and microstructures incorporating alloy-formula propellants.

The thermal decomposition mechanism of RDX/AP composites: ab initio neural network MD simulations.

A neural network potential (NNP) is developed to investigate the decomposition mechanism of RDX, AP, and their composites. Utilizing an ab initio dataset, the NNP is evaluated in terms of atomic energy and forces, demonstrating strong agreement with ab initio calculations. Numerical stability tests across a range of timesteps reveal excellent stability compared to the state-of-the-art ReaxFF models. Then the thermal decomposition of pure RDX, AP, and RDX/AP composites is performed using NNP to explore the coupling effect between RDX and AP. The results highlight a dual interaction between RDX and AP, i.e., AP accelerates RDX decomposition, particularly at low temperatures, and RDX promotes AP decomposition. Analyzing RDX trajectories at the RDX/AP interface unveils a three-part decomposition mechanism involving N-N bond cleavage, H transfer with AP to form Cl-containing acid, and chain-breaking reactions generating small molecules such as N2, CO, and CO2. The presence of AP enhances H transfer reactions, contributing to its role in promoting RDX decomposition. This work studies the reaction kinetics of RDX/AP composites from the atomic point of view, and can be widely used in the establishment of reaction kinetics models of composite systems with energetic materials.

Ab Initio Driven Exploration on the Thermal Properties of Al-Li Alloy.

Al-Li alloys are feasible and promising additives in advanced energy and propellant systems due to the significantly enhanced heat release and increased specific impulse. The thermal properties of Al-Li alloys directly determine the manufacturing, storage safety, and ignition delay of propellants. In this study, a neural network potential (NNP) is developed to investigate the thermal behaviors of Al-Li alloys from an atomistic perspective. The novel NNP demonstrates an excellent predictive ability for energy, atomic force, mechanical behaviors, phonon vibrations, and dynamic evolutions. A series of NNP-based molecular dynamics simulations are performed to investigate the effect of Li doping on the thermal properties of Al-Li alloys. All calculated results for Al-Li alloys are consistent with experimental values for Al, ensuring their validity in predicting Al-Li interactions. The simulation results suggest that a minor increment in the Li content results in a slight change in the melting point, thermal expansion, and radical distribution functions. These three properties are associated with the lattice characteristics; nonetheless, it causes a substantial reduction in thermal conductivity, which is related to the physical properties of the elements. The lower thermal conductivity allows heat accumulation on the particle surface, thereby speeding up the surface premelt and ignition. This provides an alternative atomic explanation for the improved combustion performance of Al-Li alloys. These findings integrate insights from the field of alloy material science into crucial combustion applications, serving as an atomistic guide for developing manufacturing techniques.

Determining the mechanical and decomposition properties of high energetic materials (α-RDX, ß-HMX, and ε-CL-20) using a neural network potential.

Molecular simulations of high energetic materials (HEMs) are limited by efficiency and accuracy. Recently, neural network potential (NNP) models have achieved molecular simulations of millions of atoms while maintaining the accuracy of density functional theory (DFT) levels. Herein, an NNP model covering typical HEMs containing C, H, N, and O elements is developed. The mechanical and decomposition properties of 1,3,5-trinitroperhydro-1,3,5-triazine (RDX), hexahydro-1,3,5-trinitro-1,3,5-triazine (HMX), and 2,4,6,8,10,12-hexanitrohexaazaisowurtzitane (CL-20) are determined by employing the molecular dynamics (MD) simulations based on the NNP model. The calculated results show that the mechanical properties of α-RDX, ß-HMX, and ε-CL-20 agree with previous experiments and theoretical results, including cell parameters, equations of state, and elastic constants. In the thermal decomposition simulations, it is also found that the initial decomposition reactions of the three crystals are N-NO2 homolysis, corresponding radical intermediates formation, and NO2-induced reactions. This decomposition trajectory is mainly divided into two stages separating from the peak of NO2: pyrolysis and oxidation. Overall, the NNP model for C/H/N/O elements in this work is an alternative reactive force field for RDX, HMX, and CL-20 HEMs, and it opens up new potential for future kinetic study of nitramine explosives.

Predicting the enthalpy of formation of energetic molecules via conventional machine learning and GNN.

Machine learning (ML) provides a promising method for efficiently and accurately predicting molecular properties. Using ML models to predict the enthalpy of formation of energetic molecules helps in fast screening of potential high-energy molecules, thereby accelerating the design of energetic materials. A persistent challenge is to determine the optimal featurization methods for molecular representation and use an appropriate ML model. Thus, in our study, we evaluate various featurization methods (CDS, ECFP, SOAP, GNF) and ML models (RF, MLP, GCN, MPNN), dividing them into two groups: conventional ML models and GNN models, to predict the enthalpy of formation of potential high-energy molecules. Our results demonstrate that CDS and SOAP have advantages over the ECFP, while the GNFs in GCN and MPNN models perform better. Furthermore, the MPNN model performs best among all models with a root mean square error (RMSE) as low as 8.42 kcal mol-1, surpassing even the best performing CDS-MLP model in conventional ML models. Overall, this study provides a benchmark for ML in predicting enthalpy of formation and emphasizes the tremendous potential of GNN in property prediction.

Monitoring the melting behavior of boron nanoparticles using a neural network potential.

The melting behavior of metal additives is fundamental for various propulsion and energy-conversion applications. A neural network potential (NNP) is proposed to examine the size-dependent melting behaviors of boron nanoparticles. Our NNP model is proven to possess a desirable computational efficiency and retain ab initio accuracy, allowing investigation of the physicochemical properties of bulk boron crystals from an atomic perspective. In this work, a series of NNP-based molecular dynamics simulations were conducted and numerical evidence of the size-dependent melting behavior of boron nanoparticles with diameters from 3 to 6 nm was reported for the first time. Evolution of the intermolecular energy and the Lindemann index are used to monitor the melting process. A liquid layer forms on the particle surface and further expands with increased temperature. Once the liquid layer reaches the core region, the particle is completely molten. The reduced melting temperature of the boron nanoparticle decreases with its particle size following a linear relationship with reciprocal size, similar to other commonly used metals (Al and Mg). Additionally, boron nanoparticles are more sensitive to particle size than Al particles and less sensitive than Mg particles. These findings provide an atomistic perspective for developing manufacturing techniques and tailoring combustion performance in practical applications.

Revealing the thermal decomposition mechanism of RDX crystals by a neural network potential.

A neural network potential (NNP) is developed to investigate the complex reaction dynamics of 1,3,5-trinitro-1,3,5-triazine (RDX) thermal decomposition. Our NNP model is proven to possess good computational efficiency and retain the ab initio accuracy, which allows the investigation of the entire decomposition process of bulk RDX crystals from an atomic perspective. A series of molecular dynamics (MD) simulations are performed on the NNP to calculate the physical and chemical properties of the RDX crystal. The results show that the NNP can accurately describe the physical properties of RDX crystals, such as the cell parameters and the equation of state. The simulations of RDX thermal decomposition reveal that the NNP could capture the evolution of species at ab initio accuracy. The complex reaction network was established, and a reaction mechanism of RDX decomposition was provided. The N-N homolysis is the dominant channel, which cannot be observed in previous DFT studies of isolated RDX molecule. In addition, the H abstraction reaction by NO2 is found to be the critical pathway for NO and H2O formation, while the HONO elimination is relatively weak. The NNP gives an atomic insight into the complex reaction dynamics of RDX and can be extended to investigate the reaction mechanism of novel energetic materials.

Exploring Complex Reaction Networks Using Neural Network-Based Molecular Dynamics Simulation.

Ab initio molecular dynamics (AIMD) is an established method for revealing the reactive dynamics of complex systems. However, the high computational cost of AIMD restricts the explorable length and time scales. Here, we develop a fundamentally different approach using molecular dynamics simulations powered by a neural network potential to investigate complex reaction networks. This potential is trained via a workflow combining AIMD and interactive molecular dynamics in virtual reality to accelerate the sampling of rare reactive processes. A panoramic visualization of the complex reaction networks for decomposition of a novel high explosive (ICM-102) is achieved without any predefined reaction coordinates. The study leads to the discovery of new pathways that would be difficult to uncover if established methods were employed. These results highlight the power of neural network-based molecular dynamics simulations in exploring complex reaction mechanisms under extreme conditions at the ab initio level, pushing the limit of theoretical and computational chemistry toward the realism and fidelity of experiments.

Molecular Dynamics Study on the Condensation of PAH Molecules on Quasi Soot Surfaces.

In this paper, the condensation efficiency of polycyclic aromatic hydrocarbon (PAH) molecules up to coronene, from 500 to 2000 K, is calculated based on hundreds of collisions between a PAH molecule and the quasi soot surface, which is composed of stacked coronene molecules with periodic boundary conditions, using molecular dynamics simulations. The results show that the condensation efficiency increases with the PAH molecular mass but decreases as the temperature increases, following a Gaussian function. Meanwhile, when the presence of aliphatic chains on soot particle surfaces is considered, the condensation efficiency can be lowered by up to 40%, being affected more significantly at higher temperatures. A condensation efficiency model is thus proposed from the molecular trajectories. Finally, when this newly proposed PAH condensation efficiency model is adopted, better agreement with the experiments is achieved in predicting soot volume fractions of an ethylene/oxygen/nitrogen mixture in a tandem jet-stirred reactor and a plug-flow reactor.

On Modeling the Combustion of a Single Micron-Sized Aluminum Particle with the Effect of Oxide Cap.

The formation of oxide cap, which results from the condensation of gaseous aluminum oxide, makes a non-negligible impact on the combustion process of micron-sized aluminum particles, but its growth and effect are still unknown. Also, the transition of combustion modes during the combustion process, which affects the growth rate of the oxide cap, needs to be explored. Therefore, a detailed combustion model of a single micron-sized aluminum particle is developed to predict the transition of combustion modes and the effect of the oxide cap. This combustion model consists of a vapor-phase kinetic model and a particle model coupled by the Strang splitting algorithm. The predicted ignition delay and combustion times are compared with experimental data to validate the combustion model. Three combustion modes including vapor-phase, transitional, and surface combustions are considered in this combustion model. We find that the two modes coexist for particles between 100 and 200 µm when the ambient temperature and pressure are 2500 K and 1 atm, respectively. A higher ambient temperature extends the transition of combustion mode toward smaller sizes. An oxide cap model considering surface free energy is proposed to study the growth and effect of the oxide cap on the combustion process of micron-sized aluminum particles. We find that the formation of oxide cap limits the evaporation rate of aluminum directly due to the reduced active surface area. The oxide cap stabilizes the evolution of particle temperature and determines the burning time. The predicted burning time is reduced by a factor of 2 at least considering the growth of oxide cap.

Hydrogen abstraction/addition reactions in soot surface growth.

The hydrogen abstraction (HB) and addition reactions (HD) by H radicals are examined on a series of polycyclic aromatic hydrocarbon (PAH) monomers and models of quasi-surfaces using quasi-classical trajectory (QCT) method. QCT results reproduce the rate constants of HB reactions on PAH monomers from density functional theory (DFT) in the range of 1500-2700 K. The PAH size has a minor impact on the rates of HB reactions, especially at temperatures beyond 2100 K. In contrast, HD reactions have a clear size dependence, and a larger PAH yields a higher rate. It was also found that the preferred reaction pathway changes from HB to HD reactions at ∼1900 K. The rates of surface HB and HD reactions exceed those in the gas phase by nearly one factor of magnitude. Further analysis of the detailed trajectory of the QCT method reveals that about 50% of surface reactions can be attributed to the events of surface diffusion, which depends on the local energy transfer in gas-surface interactions. However, this phenomenon is not preferred in PAH monomers, as expected. Our finding here questions the treatment of the surface reactions of soot as the product of the first collision between the gaseous species and particle surface. The surface diffusion-induced reactions should be accounted for in the rates of the surface HB and HD reactions. The rate constants of HB and HD reactions on each reactive site (surface zig-zag, surface free-edge and pocket free-edge sites) were calculated by QCT method, and are recommended for the further development of surface chemistry models in soot formation.