Interpretable Performance Models for Energetic Materials using Parsimonious Neural Networks.

Predictive models for the performance of explosives and propellants are important for their design, optimization, and safety. Thermochemical codes can predict some of these properties from fundamental quantities such as density and formation energies that can be obtained from first principles. Models that are simpler to evaluate are desirable for efficient, rapid screening of material screening. In addition, interpretable models can provide insight into the physics and chemistry of these materials that could be useful to direct new synthesis. Current state-of-the-art performance models are based on either the parametrization of physics-based expressions or data-driven approaches with minimal interpretability. We use parsimonious neural networks (PNNs) to discover interpretable models for the specific impulse of propellants and detonation velocity and pressure for explosives using data collected from the open literature. A combination of evolutionary optimization with custom neural networks explores and trains models with objective functions that balance accuracy and complexity. For all three properties of interest, we find interpretable models that are Pareto optimal in the accuracy and simplicity space.

Megahertz-rate background-oriented schlieren tomography in post-detonation blasts.

The understanding and predictive modeling of explosive blasts require advanced experimental diagnostics that can provide information on local state variables with high spatiotemporal resolution. Current datasets are predominantly based on idealized spherically symmetric explosive charges and point-probe measurements, although practical charges typically involve multidimensional spatial structures and complex shock-flow interactions. This work introduces megahertz-rate background-oriented schlieren tomography to resolve transient, three-dimensional density fields, as found in an explosive blast, without symmetry assumptions. A numerical evaluation is used to quantify the sources of error and optimize the reconstruction parameters for shock fields. Average errors are ∼3% in the synthetic environment, where the accuracy is limited by the deflection sensing algorithm. The approach was experimentally demonstrated on two different commercial blast charges (Mach ∼1.2 and ∼1.7) with both spherical and multi-shock structures. Overpressure measurements were conducted using shock-front tracking to provide a baseline for assessing the reconstructed densities. The experimental reconstructions of the primary blast fronts were within 9% of the expected peak values. The megahertz time resolution and quantitative reconstruction without symmetry assumptions were accomplished using a single high-speed camera and light source, enabling the visualization of multi-shock structures with a relatively simple arrangement. Future developments in illumination, imaging, and analysis to improve the accuracy in extreme environments are discussed.

Identification of Elusive Keto and Enol Intermediates in the Photolysis of 1,3,5-Trinitro-1,3,5-Triazinane.

Enols have emerged as critical reactive intermediates in combustion processes and in fundamental molecular mass growth processes in the interstellar medium, but the elementary reaction pathways to enols in extreme environments, such as during the decomposition of molecular energetic materials, are still elusive. Here, we report on the original identification of the enol and keto isomers of oxy-s-triazine, as well as its deoxygenated derivative 1,3,5-triazine, formed in the photodecomposition processes of 1,3,5-trinitro-1,3,5-triazinane (RDX)-a molecular energetic material. The identification was facilitated by exploiting isomer-selective tunable photoionization reflectron time-of-flight mass spectrometry (PI-ReTOF-MS) in conjunction with quantum chemical calculations. The present study reports the first experimental evidence of an enol intermediate in the dissociation domain of a nitramine-based energetic material. Our investigations suggest that the enols like 1,3,5-triazine-2-ol could be the source of hydroxyl radicals, and their inclusion in the theoretical models is important to understand the unprecedented chemistry of explosive materials.

Prediction of Energetic Material Properties from Electronic Structure Using 3D Convolutional Neural Networks.

We develop a convolutional neural network capable of directly parsing the 3D electronic structure of a molecule described by spatial point data for charge density and electrostatic potential represented as a 4D tensor. This method effectively bypasses the need to construct complex representations, or descriptors, of a molecule. This is beneficial because the accuracy of a machine learned model depends on the input representation. Ideally, input descriptors encode the essential physics and chemistry that influence the target property. Thousands of molecular descriptors have been proposed, and proper selection of features requires considerable domain expertise or exhaustive and careful statistical downselection. In contrast, deep learning networks are capable of learning rich data representations. This provides a compelling motivation to use deep learning networks to learn molecular structure-property relations from "raw" data. The convolutional neural network model is jointly trained on over 20,000 molecules that are potentially energetic materials (explosives) to predict dipole moment, total electronic energy, Chapman-Jouguet (C-J) detonation velocity, C-J pressure, C-J temperature, crystal density, HOMO-LUMO gap, and solid phase heat of formation. This work demonstrates the first use of complete 3D electronic structure for machine learning of molecular properties.

Investigating the Photochemical Decomposition of Solid 1,3,5-Trinitro-1,3,5-triazinane (RDX).

Energetic materials such as 1,3,5-trinitro-1,3,5-triazinane (RDX) are known to photodissociate when exposed to UV light. However, the fundamental photochemical process(es) that initiate the decomposition of RDX is (are) still debatable. In this study we investigate the photodissociation of solid-phase RDX at four distinct UV wavelengths (254 nm (4.88 eV), 236 nm (5.25 eV), 222 nm (5.58 eV), 206 nm (6.02 eV)) exploiting a surface science machine at 5 K. We also conducted dose-dependent studies at the highest and lowest photon energy of 206 nm (6.02 eV) and 254 nm (4.88 eV). The products were monitored online and in situ via infrared spectroscopy. During the temperature-programmed desorption phase, the subliming products were detected with a reflectron time-of-flight mass spectrometer coupled with soft-photoionization at 10.49 eV (PI-ReTOF-MS). Infrared spectroscopy revealed the formation of small molecules including nitrogen monoxide (NO), nitrogen monoxide dimer ([NO]2), dinitrogen trioxide (N2O3), carbon dioxide (CO2), carbon monoxide (CO), dinitrogen monoxide (N2O), water (H2O), and nitrite group (-ONO) while ReTOF-MS identified 32 cyclic and acyclic products. Among these, 11 products such as nitryl isocyanate (CN2O3), 5-nitro-1,3,5-triazinan-2-one (C3H6N4O3) and 1,5-dinitro-1,3,5-triazinan-2-one (C3H5N5O5) were detected for the first time in photodecomposition of RDX. Dose-dependent in combination with wavelength-dependent photolysis experiments aid to identify key primary and secondary products as well as distinguished pathways that are more preferred at lower and higher photon energies. Our experiments reveled that N-NO2 bond fission and nitro-nitrite isomerization are the initial steps in the UV photolysis of RDX. Reaction mechanisms are derived by comparing the experimental findings with previous electronic structure calculations to rationalize the origin of the observed products. The present study can assist in understanding the complex chemistry behind the photodissociation of electronically excited RDX molecule, thus bringing us closer to unraveling the decomposition mechanisms of nitramine-based explosives.

The Elusive Ketene (H2 CCO) Channel in the Infrared Multiphoton Dissociation of Solid 1,3,5-Trinitro-1,3,5-Triazinane (RDX).

Understanding of the fundamental mechanisms involved in the decomposition of 1,3,5-trinitro-1,3,5-triazinane (RDX) still represents a major challenge for the energetic materials and physical (organic) chemistry communities mainly because multiple competing dissociation channels are likely involved and previous detection methods of the products are not isomer selective. In this study we exploited a microsecond pulsed infrared laser to decompose thin RDX films at 5 K under mild conditions to limit the fragmentation channels. The subliming decomposition products during the temperature programed desorption phase are detected using isomer selective single photoionization time-of-flight mass spectrometry (PI-ReTOF-MS). This technique enables us to assign a product signal at m/z=42 to ketene (H2 CCO), but not to diazomethane (H2 CNN; 42 amu) as speculated previously. Electronic structure calculations support our experimental observations and unravel the decomposition mechanisms of RDX leading eventually to the elusive ketene (H2 CCO) via an exotic, four-membered ring intermediate. This study highlights the necessity to exploit isomer-selective detection schemes to probe the true decomposition products of nitramine-based energetic materials.

Probing the Reaction Mechanisms Involved in the Decomposition of Solid 1,3,5-Trinitro-1,3,5-triazinane by Energetic Electrons.

The decomposition mechanisms of 1,3,5-trinitro-1,3,5-triazinane (RDX) have been explored over the past decades, but as of now, a complete picture on these pathways has not yet emerged, as evident from the discrepancies in proposed reaction mechanisms and the critical lack of products and intermediates observed experimentally. This study exploited a surface science machine to investigate the decomposition of solid-phase RDX by energetic electrons at a temperature of 5 K. The products formed during irradiation were monitored online and in situ via infrared and UV-vis spectroscopy, and products subliming in the temperature programmed desorption phase were probed with a reflectron time-of-flight mass spectrometer coupled with soft photoionization at 10.49 eV (ReTOF-MS-PI). Infrared spectroscopy revealed the formation of water (H2O), carbon dioxide (CO2), dinitrogen oxide (N2O), nitrogen monoxide (NO), formaldehyde (H2CO), nitrous acid (HONO), and nitrogen dioxide (NO2). ReTOF-MS-PI identified 38 cyclic and acyclic products arranged into, for example, dinitro, mononitro, mononitroso, nitro-nitroso, and amines species. Among these molecules, 21 products such as N-methylnitrous amide (CH4N2O), 1,3,5-triazinane (C3H9N3), and N-(aminomethyl)methanediamine (C2H9N3) were detected for the first time in laboratory experiments; mechanisms based on the gas phase and condensed phase calculations were exploited to rationalize the formation of the observed products. The present studies reveal a rich, unprecedented chemistry in the condensed phase decomposition of RDX, which is significantly more complex than the unimolecular gas phase decomposition of RDX, thus leading us closer to an understanding of the decomposition chemistry of nitramine-based explosives.

Dynamic imaging of the temperature field within an energetic composite using phosphor thermography.

An improved understanding of energy localization ("hot spots") is needed to improve the safety and performance of explosives. We propose a technique to visualize and quantify the properties of a dynamic hot spot from within an energetic composite subjected to ultrasonic mechanical excitation. The composite is composed of an optically transparent binder and a countable number of octahydro 1,3,5,7-tetranitro-1,3,5,7-tetrazocine (HMX) crystals. The evolving temperature field is measured by observing the luminescence from embedded phosphor particles and subsequent application of the intensity ratio method. The spatial temperature precision is less than 2% of the measured absolute temperature in the temperature regime of interest (23°C-220°C). The temperature field is mapped from within an HMX-binder composite under periodic mechanical excitation.

Innovative scheme for high-repetition-rate imaging of CN radical.

We have employed, to the best of our knowledge, a novel excitation scheme to perform the first high-repetition-rate planar laser-induced fluorescence (PLIF) measurements of a CN radical in combustion. The third harmonic of a Nd:YVO4 laser at 355 nm due to its relatively large linewidth overlaps with several R branch transitions in a CN ground electronic state. Therefore, the 355 nm beam was employed to directly excite the CN transitions with good efficiency. The CN measurements were performed in premixed CH4-N2O flames with varying equivalence ratios. A detailed characterization of the high-speed CN PLIF imaging system is presented via its ability to capture statistical and dynamical information in these premixed flames. Single-shot CN PLIF images obtained over a HMX pellet undergoing self-supported deflagration are presented as an example of the imaging system being applied towards characterizing the flame structure of energetic materials.

Removing hydrochloric acid exhaust products from high performance solid rocket propellant using aluminum-lithium alloy.

Hydrochloric acid (HCl) pollution from perchlorate based propellants is well known for both launch site contamination, as well as the possible ozone layer depletion effects. Past efforts in developing environmentally cleaner solid propellants by scavenging the chlorine ion have focused on replacing a portion of the chorine-containing oxidant (i.e., ammonium perchlorate) with an alkali metal nitrate. The alkali metal (e.g., Li or Na) in the nitrate reacts with the chlorine ion to form an alkali metal chloride (i.e., a salt instead of HCl). While this technique can potentially reduce HCl formation, it also results in reduced ideal specific impulse (ISP). Here, we show using thermochemical calculations that using aluminum-lithium (Al-Li) alloy can reduce HCl formation by more than 95% (with lithium contents ≥15 mass%) and increase the ideal ISP by ∼7s compared to neat aluminum (using 80/20 mass% Al-Li alloy). Two solid propellants were formulated using 80/20 Al-Li alloy or neat aluminum as fuel additives. The halide scavenging effect of Al-Li propellants was verified using wet bomb combustion experiments (75.5±4.8% reduction in pH, ∝ [HCl], when compared to neat aluminum). Additionally, no measurable HCl evolution was detected using differential scanning calorimetry coupled with thermogravimetric analysis, mass spectrometry, and Fourier transform infrared absorption.

Using time-frequency analysis to determine time-resolved detonation velocity with microwave interferometry.

Two time-frequency analysis methods based on the short-time Fourier transform (STFT) and continuous wavelet transform (CWT) were used to determine time-resolved detonation velocities with microwave interferometry (MI). The results were directly compared to well-established analysis techniques consisting of a peak-picking routine as well as a phase unwrapping method (i.e., quadrature analysis). The comparison is conducted on experimental data consisting of transient detonation phenomena observed in triaminotrinitrobenzene and ammonium nitrate-urea explosives, representing high and low quality MI signals, respectively. Time-frequency analysis proved much more capable of extracting useful and highly resolved velocity information from low quality signals than the phase unwrapping and peak-picking methods. Additionally, control of the time-frequency methods is mainly constrained to a single parameter which allows for a highly unbiased analysis method to extract velocity information. In contrast, the phase unwrapping technique introduces user based variability while the peak-picking technique does not achieve a highly resolved velocity result. Both STFT and CWT methods are proposed as improved additions to the analysis methods applied to MI detonation experiments, and may be useful in similar applications.

Amine-boranes: green hypergolic fuels with consistently low ignition delays.

Complexation of amines with borane converts them to hypergols or decreases their ignition delays (IDs) multifold (with white fuming nitric acid as the oxidant). With consistently low IDs, amine-boranes represent a class of compounds that can be promising alternatives to toxic hydrazine and its derivatives as propellants. A structure-hypergolicity relationship study reveals the necessary features for the low ID.

High-repetition-rate three-dimensional OH imaging using scanned planar laser-induced fluorescence system for multiphase combustion.

Imaging dynamic multiphase combusting events is challenging. Conventional techniques can image only a single plane of an event, capturing limited details. Here, we report on a three-dimensional, time-resolved, OH planar laser-induced fluorescence (3D OH PLIF) technique that was developed to measure the relative OH concentration in multiphase combustion flow fields. To the best of our knowledge, this is the first time a 3D OH PLIF technique has been reported in the open literature. The technique involves rapidly scanning a laser sheet across a flow field of interest. The overall experimental system consists of a 5 kHz OH PLIF system, a high-speed detection system (image intensifier and CMOS camera), and a galvanometric scanning mirror. The scanning mirror was synchronized with a 500 Hz triangular sweep pattern generated using Labview. Images were acquired at 5 kHz corresponding to six images per mirror scan, and 1000 scans per second. The six images obtained in a scan were reconstructed into a volumetric representation. The resulting spatial resolution was 500×500×6 voxels mapped to a field of interest covering 30 mm×30 mm×8 mm. The novel 3D OH PLIF system was applied toward imaging droplet combustion of methanol gelled with hydroxypropyl cellulose (HPC) (3 wt. %, 6 wt. %), as well as solid propellant combustion, and impinging jet spray combustion. The resulting 3D dataset shows a comprehensive view of jetting events in gelled droplet combustion that was not observed with high-speed imaging or 2D OH PLIF. Although the scan is noninstantaneous, the temporal and spatial resolution was sufficient to view the dynamic events in the multiphase combustion flow fields of interest. The system is limited by the repetition rate of the pulsed laser and the step response time of the galvanometric mirror; however, the repetition rates are sufficient to resolve events in the order of 100 Hz. Future upgrade includes 40 kHz pulsed UV laser system, which can reduce the scan time to 125 µs, while keeping the high repetition rate of 1000 Hz.

Fate and toxicity of CuO nanospheres and nanorods used in Al/CuO nanothermites before and after combustion.

Although nanotechnology advancements should be fostered, the environmental health and safety (EHS) of nanoparticles used in technologies must be quantified simultaneously. However, most EHS studies assess the potential implications of the free nanoparticles which may not be directly applicable to the EHS of particles incorporated into in-use technologies. This investigation assessed the aquatic toxicological implications of copper oxide (CuO) nanospheres relative to CuO nanorods used in nanoenergetic applications to improve combustion. Particles were tested in both the as-received form and following combustion of a CuO/aluminum nanothermite. Results indicated nanospheres were more stable in water and slowly released ions, while higher surface area nanorods initially released more ions and were more toxic but generally less stable. After combustion, particles sintered into larger, micrometer-scale aggregates, which may lower toxicity potential to pelagic organisms due to deposition from water to sediment and reduced bioavailability after complexation with sediment organic matter. Whereas the larger nanothermite residues settled rapidly, implying lower persistence in water, their potential to release dissolved Cu was higher which led to greater toxicity to Ceriodaphnia dubia relative to parent CuO material (nanosphere or rod). This study illustrates the importance of considering the fate and toxicology of nanoparticles in context with their relevant in-use applications.

Tuning azolium azolate ionic liquids to promote surface interactions with titanium nanoparticles leading to increased passivation and colloidal stability.

The passivation and stability of suspensions of titanium nanoparticles in azolium azolate ionic liquids can be tuned by introducing metal specific binding sites in the azolate anion.

Hypergolic ionic liquids to mill, suspend, and ignite boron nanoparticles.

Boron nanoparticles prepared by milling in the presence of a hypergolic energetic ionic liquid (EIL) are suspendable in the EIL and the EIL retains hypergolicity leading to the ignition of the boron. This approach allows for incorporation of a variety of nanoscale additives to improve EIL properties, such as energetic density and heat of combustion, while providing stability and safe handling of the nanomaterials.

Experimental modeling of explosive blast-related traumatic brain injuries.

This study aims to characterize the interaction of explosive blast waves through simulated anatomical systems. We have developed physical models and a systematic approach for testing traumatic brain injury (TBI) mechanisms and occurrences. A simplified series of models consisting of spherical poly(methyl methacrylate) (PMMA) shells housing synthetic gelatins as brain simulants have been utilized. A series of experiments was conducted to compare the sensitivity of the system response to mechanical properties of the simulants under high strain-rate explosive blasts. Small explosive charges were directed at the models to produce a realistic blast wave in a scaled laboratory setting. Blast profiles were measured and analyzed to compare system response severity. High-speed shadowgraph imaging captured blast wave interaction with the head model while particle tracking captured internal response for displacement and strain correlation. The results suggest amplification of shock waves inside the head near material interfaces due to impedance mismatches. In addition, significant relative displacement was observed between the interacting materials suggesting large strain values of nearly 5%. Further quantitative results were obtained through shadowgraph imaging of the blasts confirming a separation of time scales between blast interaction and bulk movement. These results lead to a conclusion that primary blast effects may potentially contribute significantly to the occurrence of military associated TBI.

Kinetics of high temperature reaction in Ni-Al system: influence of mechanical activation.

High temperature (>1000 K) reaction kinetics in the stoichiometric (1:1 by molar ratio) Al-Ni system was investigated by using the, so-called, electrothermal analysis (ETA) method. ETA is the only technique that allows studying kinetics of a heterogeneous gasless reaction at temperatures above the melting points of the precursors. Special attention was focused on methodological aspects of the ETA method. Two different reaction systems were studied: (i) initial Al/Ni clad particles; (ii) the same powders but after 15 min of high energy ball milling. Analysis of the obtained results leads to the conclusion that such mechanical treatment decreases the apparent activation energies of the reaction in the Ni-Al system, from 47 +/- 7 kcal/mol for the initial powder to 25 +/- 3 kcal/mol after ball milling. Comparison of these data with those reported previously was also made.

Thermal explosion in Al-Ni system: influence of mechanical activation.

The influence of short-term (5-15 min) highly energetic ball milling on the ignition characteristics of a gasless heterogeneous Ni-Al reactive system has been investigated. By using Al-Ni clad particles (30-40 microm diameter Al spheres coated by a 3-3.5 microm layer of Ni, that corresponds to a 1:1 Ni/Al atomic ratio), it was shown that such mechanical treatment leads to a significant decrease in the self-ignition temperature of the system. For example, after 15 min of ball milling, the ignition temperature appears to be approximately 600 K, well below the eutectic (913 K) in the considered binary system, which is the ignition temperature for the initial clad particles. Thus, it was demonstrated that the thermal explosion process for mechanically treated reactive media can be solely defined by solid-state reactions. Additionally, thermal analysis measurements revealed that mechanical activation results in a substantial decrease in the effective activation energy (from 84 to 28 kcal/mol) of interaction between Al and Ni. This effect, that is, mechanical activation of chemical reaction, is connected to a substantial increase of contact area between reactive particles and fresh interphase boundaries formed in an inert atmosphere during ball milling. It is also important that by varying the time of mechanical activation one can precisely control the ignition temperature in high-density energetic heterogeneous systems.

Experimental study of blast-induced traumatic brain injury using a physical head model.

This study was conducted to quantify intracranial biomechanical responses and external blast overpressures using physical head model to understand the biomechanics of blast traumatic brain injury and to provide experimental data for computer simulation of blast-induced brain trauma. Ellipsoidal-shaped physical head models, made from 3-mm polycarbonate shell filled with Sylgard 527 silicon gel, were used. Six blast tests were conducted in frontal, side, and 45 degrees oblique orientations. External blast overpressures and internal pressures were quantified with ballistic pressure sensors. Blast overpressures, ranging from 129.5 kPa to 769.3 kPa, were generated using a rigid cannon and 1.3 to 3.0 grams of pentaerythritol tetranitrate (PETN) plastic sheet explosive (explosive yield of 13.24 kJ and TNT equivalent mass of 2.87 grams for 3 grams of material). The PETN plastic sheet explosive consisted of 63% PETN powder, 29% plasticizer, and 8% nitrocellulose with a density of 1.48 g/cm3 and detonation velocity of 6.8 km/s. Propagation and reflection of the shockwave was captured using a shadowgraph technique. Shockwave speeds ranging from 423.3 m/s to 680.3 m/s were recorded. The model demonstrated a two-stage response: a pressure dominant (overpressure) stage followed by kinematic dominant (blast wind) stage. Positive pressures in the brain simulant ranged from 75.1 kPa to 1095 kPa, and negative pressures ranged from -43.6 kPa to -646.0 kPa. High- and normal-speed videos did not reveal observable deformations in the brain simulant from the neutral density markers embedded in the midsagittal plane of the head model. Amplitudes of the internal positive and negative pressures were found to linearly correlate with external overpressure. Results from the current study suggested a pressure-dominant brain injury mechanism instead of strain injury mechanism under the blast severity of the current study. These quantitative results also served as the validation and calibration data for computer simulation models of blast brain injuries.