The combustion performance of double-base propellants (DBPs) is crucial for ensuring their safe and efficient application in various propulsion systems. However, accurately predicting the combustion behavior, particularly the impact of catalytic effects under varying temperature and pressure conditions, remains a significant challenge due to the complexity of the involved microscopic mechanisms. This study introduces a novel approach to address these challenges by high-precision large-scale molecular dynamics (MD) simulations. First, a reactive neural network potential (NNP) is developed by integrating deep learning techniques with density functional theory (DFT) calculations, providing high-precision force fields for the combustion process of DBPs. This NNP model is capable of accurately predicting the energy and forces involved in reactions, overcoming the limitations of traditional methods in capturing combustion mechanisms at the microscopic level. Second, large-scale MD simulations, based on the machine learning potential, are conducted to model the combustion process of DBPs under extreme conditions, particularly focusing on the dynamic behavior of the flame front. The simulation framework demonstrates both accuracy and efficiency, offering a novel computational approach for predicting propellant combustion performance. Finally, the study reveals the catalytic effects on the combustion process by systematically investigating how catalysts regulate the reaction rate under various temperature and pressure conditions. The results show that catalysts significantly influence the thermal decomposition pathways and reaction kinetics, providing new theoretical insights into the catalytic reaction mechanisms of propellant combustion. This research offers a comprehensive and efficient framework for simulating and understanding the combustion of DBPs, contributing to the design and optimization of propellants.

RP-3 is the most widely used aviation fuel in China, and its consumption has increased significantly in recent years, being responsible for the release of considerable amounts of greenhouse gases and soot particles. Therefore, it is imperative to gain fundamental understanding of its burning behavior and design methodologies to achieve cleaner and more efficient combustion. This study presents an investigation into the capabilities of a tailored strategy for the kinetic modeling of the gas phase of RP-3 (HyChem) and its coupling with soot particle dynamics. A stagnation and coflow flame are used for simulating test cases using the CoFlame code, and compared to available experimental data about soot and flow characteristics. In burner stabilized stagnation flames, good agreement is observed between experiments and predictions of temperature profiles. For coflow flames, the model predicts soot volume fraction and flame height with good accuracy as compared to experimental data of an RP-3 flame, although the soot peak is slightly shifted towards the flame edge. Although the kinetic model was designed for RP-3, it also captures overall trends of soot production in a Jet A-1 flame. This approach demonstrates a reliable predictive capability for aviation fuel combustion and provides a basis for further refinement in the modeling of these complex fuels, with potential applications for cleaner combustion strategies.

1,1-Diamino-2,2-dinitroethylene (FOX-7) is a high-energy, low-sensitivity explosive, yet its decomposition pathway remains critical for safe application. In this study, the thermal decomposition of FOX-7 was investigated through a combination of thermogravimetric (TG) measurements and chemical reaction neural network (CRNN) modelling. Five sets of the experimental TG measurements were first selected to evaluate the inherent uncertainties. In particular, the two-stage decomposition characteristics and the solid residue were discussed in detail. Two CRNN models. i.e., the 5-2 model (five species and two reactions) and 5-4 model (five species and four reactions) were developed, with both accurately predicting initial decomposition activation energies. The 5-4 model elucidates detailed reaction pathways, including C & horbar;NO2, C & boxH;C, and C & horbar;H bond cleavages, alongside product interactions, aligning with prior theoretical studies. The overall reaction mechanism and the associated energy barriers for bond dissociation are consistent with previous theoretical studies. Our findings highlight the capability of the CRNN model to decode complex decomposition kinetics, including multi-stage reactions and residue formation. This approach offers a promising framework for modelling other energetic materials.

The combustion kinetics of ammonium perchlorate (AP), a key oxidizer in solid propellants, remain debated due to the complexity of its solid-phase decomposition. In this study, the reaction kinetics of AP is resolved by decoupling the complex combustion process into solid-phase pyrolysis and gas-phase oxidation sub-processes, proposed as EM-HyChem approach. By this approach, the key pyrolysis products and reaction mechanisms are identified through molecular dynamics simulations. The chemical reaction neural network model is then employed to extract the rate parameters in pyrolysis model from thermogravimetric experiments. Subsequently, the pyrolysis model is coupled with an oxidation model for the pyrolysis products to build a kinetic model for AP. The kinetic model is used to simulate AP laminar flame via a one-dimensional method. Predicted burning rates, surface temperatures, and species profiles show good agreement with results from other experimental measurements and models. Sensitivity analysis of kinetic parameters provides insights into the factors contributing to the N-shaped curve of the AP burning rate.

Iron powder is regarded as a highly promising zero-carbon energy carrier, with combustion as its primary mode of energy release. However, iron dust flames exhibit poor stability, prompting the common practice of co-firing with hydrocarbon fuels to ensure stable combustion. This approach still yields carbon emissions. In pursuit of a fully zero-carbon iron-fuel cycle, the present work firstly investigates the combustion characteristics of single iron particles under ammonia co-firing conditions. Two distinct combustion behaviors, including micro-explosion and fragment release, are observed. The fragments are inferred to be the nanoparticle cloud based on 30k fps high-speed shadowgraphy. Under ammonia as the carrier gas, the micro-explosion probability of iron particles exceeds that observed with methane or nitrogen, significantly at oxygen mole fractions of 10.9%-20.4%. This phenomenon likely arises from iron nitride decomposition at the liquid iron (L1)-liquid iron oxide (L2) interface. Furthermore, the micro-explosion probability in ammonia/iron combustion decreases with increasing oxygen concentration. The micro-explosion delay time (MDT) is defined to quantify the effect of particle size on liquid-phase combustion under ammonia co-firing conditions. Further experimental results show that at higher oxygen concentrations, MDT is nearly proportional to the inverse of oxygen mass fraction, suggesting that particle oxidation is limited by external oxygen diffusion. However, at YO2 = 12.5%, MDT deviates from the linear correlation. In the low oxygen concentration cases, iron nitride may react with absorbed oxygen and impede the internal transport of oxygen, thereby constraining the oxidation rate of iron and delaying the formation of a complete core-shell structure. Overall, ammonia/iron co-firing technology shows great promise for regulating micro-explosions and represents a crucial step toward realizing a genuinely zero-carbon iron-fuel cycle. Novelty and Significance Statement The fundamental combustion characteristics of iron particles under ammonia co-firing conditions were first investigated in this work. The micro-explosion probability of iron particles in a hot ammonia environment is significantly high and decreases with increasing oxygen concentration in the bulk gas. The effects of particle size and ambient oxygen concentration on the iron particles combustion time under ammonia co-firing conditions were quantitatively analyzed. The potential mechanisms underlying the influence of ammonia on the micro-explosion of iron particles were discussed. The ammonia/iron co-firing technology offers a novel approach for achieving a truly zero-carbon iron-fuel cycle.

Magnesium hydride (MgH2) is a promising energetic additive for solid propellants, yet the atomic-scale coupling between its rapid dehydrogenation and subsequent oxidation remains poorly understood. In this study, we developed a high-fidelity neural network potential (NNP) with ab initio accuracy to reveal the distinct combustion mechanisms of MgH2 versus pure Mg in oxygen environments. Molecular dynamics simulations uncover a unique “Coupled Dehydrogenation-Oxidation” (CDO) mechanism governing MgH2 combustion. Unlike the diffusion-limited shrinking-core oxidation of Mg, MgH2 acts as a “self-activating” fuel. We identify a critical “dehydrogenation-induced shattering” process, where the internal pressure from H2 release mechanically ruptures the MgO passivation shell, creating interconnected channels. This structural damage induces a dominant “channeling effect”, significantly enhancing the inward diffusion of O2 and enabling volumetric oxidation. Furthermore, the interfacial formation of H2O generates lattice defects that facilitate oxygen transport, creating a reaction-transport feedback loop that accelerates global kinetics. These findings provide a rigorous theoretical basis for the superior ignition sensitivity of MgH2 and offer atomistic guidance for designing high-performance energetic formulations.

Nitroguanidine (NQ) plays a central role in aerospace and industrial fields, and a deep understanding of its dynamic behavior and reaction mechanisms is crucial for accurately predicting its combustion and explosive properties. To this end, this work proposes a novel kinetic modelling approach that combines a chemical reaction neural network (CRNN) with thermogravimetric (TG) experiments to conduct an in-depth study of NQ reaction kinetics. The results demonstrate that the kinetic model constructed via the CRNN accurately fits the experimental data, revealing the main reaction pathways of NQ and extracting kinetic parameters. Two simplified models have been developed: a single-step reaction model and a multistep reaction model. The single-step model, regarded as the global reaction of NQ, effectively predicts the pyrolysis process of NQ and the formation of its solid products, with activation energy values that are consistent with the experimental results. The multistep reaction model successfully reproduces the TG curve and offers a detailed depiction of the NQ reaction mechanism, covering the initial decomposition pathway, intermediate products, and interconversion reactions among gases. Compared with other data-driven modelling techniques, the CRNN modelling approach incorporates constraints from both experimental and numerical results, making the derived kinetic models more physically reasonable.

Understanding the microscopic origin of impact sensitivity (IS) in energetic materials (EMs) requires a physically meaningful descriptor that links molecular-scale dynamic response to macroscopic behavior. In this work, molecular dynamics simulations based on a Deep potential (DeepMD) were employed to investigate the impact response of α-RDX nanocrystals under practical drop-weight-like loading conditions. An explicit atomic impactor was introduced to capture heterogeneous mechanical responses, including stress concentration, energy localization, and compression-shear coupled deformation. The simulations reveal that impact-induced reaction initiation proceeds through a sequence of impact energy deposition, mechanical compression, hotspot formation, and rapid decomposition. These processes collectively reflect the intrinsic resistance of material to impact-induced failure at the molecular scale. The decomposition fraction is used as an observable to identify the onset of irreversible reactions, from which the critical impact velocity (vc) is defined. vc serves as a molecular-level descriptor of impact sensitivity that quantifies the material resistance to impact-induced failure, providing a physically interpretable measure of impact resistance. Comparative simulations on eight energetic crystals (IS = 3.5-120 J) reproduce the experimental impact sensitivity ranking with a good correlation (R2 = 0.91) between the descriptor vc2 and IS. These results establish a direct link between atomistic failure processes and macroscopic impact sensitivity, providing a descriptor-based framework for the quantitative prediction and virtual screening of EMs with improved safety-performance balance.

Accurate atomistic simulation of polymer-bonded explosive (PBX) initiation demands a force field that simultaneously captures mechanical deformation, thermal response, and chemical bond scission. Classical force fields reproduce equilibrium physical properties but cannot describe bond breaking, and ReaxFF contains no parameterization for F23-series fluoropolymers, leaving binder degradation and explosive decomposition inaccessible to molecular simulations. Here, we report a reactive machine learning potential (MLP) for three fluoropolymer binders widely used in PBXs: F2311, F2313, and F2314. Trained on density functional theory (DFT) datasets via the deep potential framework with iterative active learning, the MLP achieves ab initio accuracy at a fraction of the computational cost. Validation spans both physical and chemical regimes: radial distribution functions, equilibrium densities, glass transition temperature (Tg), and elastic constants agree quantitatively with experiment and ab initio molecular dynamics benchmarks. A microscopic analysis of backbone torsional transition events further reveals a positive correlation between dihedral activation energy and Tg, providing atomic-scale mechanistic insight into their thermal-mechanical behavior. Critically, the MLP faithfully reproduces DFT atomic forces throughout bond scission under uniaxial tension with bond dissociation energies deviating only 0.14-0.15 eV from DFT reference values. This work provides a foundation for future atomistic studies of PBX initiation, which fills a critical gap in reactive force field coverage for F23-series fluoropolymer binders and establishes a validated framework spanning physical and reactive property regimes.

The catalytic promotion effect of 4,4'-bipyridine 1,1'-dioxide (BpyNO) on the thermal decomposition of ammonium perchlorate (AP) is investigated using combined thermal analysis and neural network potential (NNP)-based molecular dynamics simulations. A small addition of BpyNO (5 wt%) reduces the main decomposition peak of AP by approximately 100 K and increases the total heat release by 2.8-fold (1338 vs. 479 J g-1). The apparent activation energy is significantly lowered from 150.5 to 109.9 kJ mol-1, indicating an accelerated decomposition process. NNP simulations reveal a distinct interfacial decomposition mechanism in the AP/BpyNO system, in which oxygen transfer from ClO4 to the organic framework dominates the early-stage reactions, in contrast to the proton-transfer-dominated pathway in pure AP. The catalytic interface promotes rapid oxygen migration from ClO4, hydrogen abstraction, and early disruption of the AP crystal lattice. These synergistic effects result in enhanced reaction kinetics and a fundamentally different decomposition pathway, consistent with comparative simulations against structurally related bipyridine analogues. The findings provide atomic-level insight into organocatalytic regulation of oxidizer decomposition and offer a mechanistic foundation for designing safer and more efficient composite energetic materials.

Polyvinylidene fluoride (PVDF) is widely used in energetic materials and other fields. However, no detailed kinetic mechanism is available to describe the complete decomposition of PVDF in the gas phase. In this work, CH3CF2CH2CF2CH2CHF2 was selected as the PVDF model to build the complete decomposition mechanism. The potential energy surface for unimolecular and bimolecular reactions of PVDF decomposition was calculated via advanced quantum chemistry methods (DLPNO-CCSD(T)/cc-pVTZ//B3LYP-D3/6-311++G(d,p)). The temperature and pressure dependences of the rate constants of key reactions at 300-3000 K and high-pressure limits were calculated via the RRKM master equation method. The thermochemical properties of related species at the DLPNO-CCSD(T)/CBS level were calculated via the atomization method. Our calculations showed that the stepby-step decomposition of PVDF is dominated by HF elimination reactions, generating large amounts of HF and hydrofluoroolefins. In addition, the radical-driven (H, OH, and O) decomposition of PVDF produces small molecules such as H2, H2O, OH and fluoroalkyl groups. It is also found that further decomposition after H abstraction from the PVDF side chain is dominated by the beta-C-C scission reactions, and the product structure is consistent with existing experimental studies. The simulation results show that the 1,2-HF elimination reactions control the initial and secondary decomposition of PVDF. The migration reactions and HF elimination reactions jointly promote the further conversion of PVDF into species with polyene sequences.

Triethyl phosphite (TEPI) is an organophosphorus compound of interest as a flame inhibitor and additive. Yet, its high-temperature decomposition pathways remain poorly characterized. This study integrates ab initio quantum chemical calculations, ReaxFF molecular dynamics (MD) simulations, and numerical kinetic modeling to investigate TEPI's thermal degradation. Molecular geometries and thermochemical parameters were obtained using M06-2X/6-311++G(d,p) and composite methods (G3B3, CBS-QB3, G3), supported by MP2/CCSD-based complete basis set (CBS) extrapolations. The C-O bond was identified as the weakest link (similar to 60 kcal/mol), consistent across electronic structure methods, and machine learning-based bond dissociation energy predictions (ALFABET). Despite variations in absolute bond dissociation energies among ReaxFF parameterizations, all methods confirmed C-O cleavage as the dominant initiation step. Reactive MD showed that TEPI is stable below 1500 K, with decomposition initiated by homolytic C-O scission yielding ethyl and an oxygen-centered radical (adjacent to the phosphorus atom). At 5000 K, over 240 intermediates were identified, including C2H5, CH3, C2H4, CH2O, PO, PO2, and HOPO. A kinetic mechanism constructed from quantum chemistry and ReaxFF-MD-derived pathways was implemented in Chemkin to simulate ignition delay times (IDTs) for TEPI/oxidizer mixtures at equivalence ratios of 0.5, 1.0, and 2.0, pressures of 1 and 10 bar, and temperatures between 800 and 2000 K. The results show that IDTs shorten with increasing pressure and equivalence ratio, while decomposition is slowed under fuel-lean conditions due to radical scavenging by phosphorus-containing fragments. Flux and sensitivity analyses highlight initial C-O bond cleavage and subsequent PO/HOPO radical recombination as the dominant kinetic pathways. This work provides new mechanistic insights into TEPI decomposition as a transferable modeling framework for trivalent phosphorus-based compounds.

Al-Li alloys represent a promising class of additives for advanced energy and propellant systems, owing to their unique micro-explosion combustion mode and enhanced performance. This study employs molecular dynamics simulations to illuminate the micro-explosion behavior of Al-Li alloys from an atomistic perspective. Key processes in Al-Li alloy combustion upon laser heating, such as melting behavior, heat transfer, mass diffusion, and product flows, are investigated in the case of alloy-AP interface. The numerical results reveal that the reduced ignition delay of Al-Li alloys, as observed in experimental studies, is primarily attributed to Li doping, which promotes moderate interface combustion before the onset of melting. Upon melting, Li atoms diffuse toward the AP oxidizer and eventually eject into the gas-phase environment and trigger micro-explosion combustion. Notably, only Li atoms located on the substrate surface actively contribute to triggering the micro-explosion, while those in the interior of substrate remain confined. Furthermore, the formation of LiCl provides an efficient reaction pathway to neutralize Cl atoms, drastically reducing HCl emissions and enabling environmentally friendly combustion. The Al-Li alloys undergo complete combustion, resulting in superior propulsion capabilities, and exceptional overall performance compared to the case with neat Al. These pivotal numerical findings offer groundbreaking insights into the micro-explosion dynamics of Al-Li alloys, laying a robust atomistic foundation for the advanced structural design and optimization of high-performance solid propellants.

Ethylene leakage in enclosed petrochemical facilities poses serious safety risks, yet monitoring is often limited by sparse sensor deployment. Discrete measurements cannot characterise the global concentration distribution or its short-term evolution in geometrically complex indoor spaces. This study proposes a deep learning-based monitoring method that reconstructs and predicts two-dimensional concentration fields from sparse sensor data. Leakage monitoring is formulated as a spatiotemporal field inference problem, and a dual-module framework is developed, comprising a reconstruction module and a one-step prediction module. Both modules employ a customised LeakU-Net architecture that combines U-Net with depthwise separable convolutions and multi-scale feature extraction to efficiently capture dispersion features. High-fidelity Computational Fluid Dynamics (CFD) simulations are used to generate training and testing datasets, and the CFD data are validated through laboratory experiments. For safe validation while preserving similar dispersion characteristics to ethylene, nitrogen is used as the surrogate release gas (with an identical molecular weight to ethylene), and oxygen concentration is measured as a proxy indicator of dispersion. The proposed method achieves reconstruction MSE below 4.0 × 10–4 and prediction MSE below 1.6 × 10–5 across the evaluated scenarios. The integrated reconstruction–prediction pipeline runs in approximately 0.3 s, enabling near-real-time field estimation and short-term forecasting. The method provides an efficient and scalable solution for monitoring ethylene-related leakage scenarios under sparse sensing conditions.

Natural gas explosions in confined spaces can cause severe structural damage, yet the quantitative relationship between overpressure loading and structural damage remains insufficiently understood. This work investigates the coupling between overpressure and window damage through five full scale confined natural gas explosion experiments with varying leakage positions and equivalence ratios. Internal overpressure histories were measured near a venting window, and characteristic parameters were extracted for the stages before and after vent opening. After each test, the post-explosion scene was reconstructed using a 3D reconstruction framework, and window fragments were identified for quantitative analysis. A volume-weighted average fragment displacement was introduced to characterize the spatial dispersion of damage. The results show that internal overpressure exhibits a two stage evolution. Increasing equivalence ratio accelerates pressure accumulation and intensifies the post opening pressure response, while leakage position has a limited influence before vent opening but significantly affects the post opening stage. Window damage evolves from localized failure to extensive fragmentation as explosion intensity increases, accompanied by an increase in fragment displacement and a reduction in the volume proportion of the largest fragment. Image processing results and 3D reconstruction results show consistent overall trends, but the latter provide more reliable quantitative characterization by reducing misidentification and preserving fragment geometry and spatial position. A quantitative relationship is established between overpressure loading and structural damage, with the vent opening parameters, especially the opening impulse, showing the strongest relationship with fragment dispersion. These findings provide a quantitative basis for explosion damage assessment in confined spaces.

Developing advanced microstructured energetic composites provides an effective strategy for regulating combustion behavior and improving the energy utilization efficiency of aluminized solid propellants. Through tailored microstructural assembly, this advanced interface engineering not only mitigates aluminum agglomeration to boost the combustion efficiency of metallic fuels, but also effectively desensitizes high-energy explosives. In this study, large-scale molecular dynamics simulations driven by a high-fidelity neural network potential were employed to investigate the combustion dynamics of two distinct microunit architectures: Al-core/RDX-shell (Al@RDX) and RDX-core/Al-shell (RDX@Al). Particular emphasis was placed on clarifying how interfacial topology influences local decomposition behavior, heat release, and pressure evolution at the nanoscale. Under condensed-phase conditions, Al@RDX exhibits rapid initial heat release due to fast RDX decomposition. However, decomposition fragments can diffuse away from the Al interface, limiting sustained interfacial reactions. In contrast, the confined RDX@Al structure initially suppresses decomposition, but gradually enhances interfacial reactions through the accumulation of reactive fragments near the surrounding Al shell, leading to stronger interfacial coupling and accelerated Al consumption. Vacuum simulations further show that RDX@Al maintains stronger interfacial reactivity under pressure-release conditions, indicating lower sensitivity to pressure fluctuations These findings establish a qualitative relationship between interfacial topology and localized combustion behavior, providing mechanistic guidance for the design of high performance aluminized composite propellants.

Gradient microstructures offer a promising route to improve dynamic strength, yet their impact mechanisms remain unclear. In this work, oxygen free copper (C10100) cylinders were explosively hardened using a low detonation velocity ammonium nitrate/fuel oil (ANFO) charge, and the charge thickness was tuned to create controllable radial gradients. Optical microscopy (OM) and electron backscatter diffraction (EBSD) quantify millimeter scale refinement in fine grain fraction and near surface grain size reduction from 8.45 to 3.07 mu m, together with higher geometrically necessary dislocation density and low-angle grain boundary fraction, and a lower twin-boundary fraction. Radial nanoindentation hardness profiles map the hardness gradient and an area weighted hardness indicator captures the effective contribution of the strengthened shell. Taylor cylinder impact tests at 236 m/s show a monotonic rise in residual length and an estimated dynamic flow stress increase of similar to 20% for the strongest condition. Post impact EBSD on recovered specimens indicates that severe loading drives both untreated and hardened states toward similarly saturated substructures, implying that preexisting gradients mainly regulate early stress and strain partitioning. Molecular dynamics modeling of Taylor impact isolates grain size gradients, twins, and preexisting dislocations, and uses waveform characteristics (peak compressive magnitude, wavefront width, blunting index) and defect-localization entropy to link microstructure to stress-wave evolution. Grain gradients broaden the compressive front, twins promote more spatially distributed planar defects and can suppress gradient driven broadening, while dislocations sharpen and amplify peaks with strong blunting reduction. These results establish microstructure informed guidelines for impact resistant metallic materials.

Property-conditioned molecular generation enables precise targeting of enthalpy of formation, producing CHON energetic molecules with properties close to specified EoF targets.