QSPR studies of impact sensitivity of nitro energetic compounds using three-dimensional descriptors

Which molecular properties determine the impact sensitivity of an explosive? A machine learning quantitative investigation of nitroaromatic explosives

We decomposed density functional theory charge densities of 53 nitroaromatic molecules into atom-centered electric multipoles using the distributed multipole analysis that provides a detailed picture of the molecular electronic structure. Three electric multipoles, (the charge of the nitro groups), (the total dipole, i.e., polarization, of the nitro groups), (the total electron delocalization of the C ring atoms), and the number of explosophore groups (#NO₂) were selected as features for a comprehensive machine learning (ML) investigation. The target property was the impact sensitivity h₅₀ (cm) values quantified by drop-weight measurements, with a large h₅₀ (e.g., 150 cm) indicating that an explosive is insensitive and vice versa. After a preliminary screening of 42 ML algorithms, four were selected based on the lowest root mean square errors: Extra Trees, Random Forests, Gradient Boosting, and AdaBoost. Compared to experimental data, the predicted h₅₀ values of molecules having very different sensitivities for the four algorithms have differences in the range 19-28%. The most important properties for predicting h₅₀ are the electron delocalization in the ring atoms and the polarization of the nitro groups with averaged weights of 39% and 35%, followed by the charge (16%) and number (10%) of nitro groups. A significant result is how the contribution of these properties to h₅₀ depends on their actual sensitivities: for the most sensitive explosives (h₅₀ up to ∼50 cm), the four properties contribute to reducing h₅₀, and for intermediate ones (∼50 cm ≲ h₅₀ ≲ 100 cm) #NO₂ and contribute to increasing it and the other two properties to reducing it. For highly insensitive explosives (h₅₀ ≳ 200 cm), all four properties essentially contribute to increasing it. These results furnish a consistent molecular basis of the sensitivities of known explosives that also can be used for developing safer new ones.

Building Chemical Property Models for Energetic Materials from Small Datasets Using a Transfer Learning Approach

For many experimentally measured chemical properties that cannot be directly computed from first-principles, the existing physics-based models do not extrapolate well to out-of-sample molecules, and experimental datasets themselves are too small for traditional machine learning (ML) approaches. To overcome these limitations, we apply a transfer learning approach, whereby we simultaneously train a multi-target regression model on a small number of molecules with experimentally measured values and a large number of molecules with related computed properties. We demonstrate this methodology on predicting the experimentally measured impact sensitivity of energetic crystals, finding that both characteristics of the computed dataset and model architecture are important to prediction accuracy of the small experimental dataset. Our directed-message passing neural network (D-MPNN) ML model using transfer learning outperforms direct-ML and physics-based models on a diverse test set, and the new methods described here are widely applicable to modeling many other structure-property relationships.

Applying machine learning to balance performance and stability of high energy density materials

The long-standing performance-stability contradiction issue of high energy density materials (HEDMs) is of extremely complex and multi-parameter nature. Herein, machine learning was employed to handle 28 feature descriptors and 5 properties of detonation and stability of 153 HEDMs, wherein all 21,648 data used were obtained through high-throughput crystal-level quantum mechanics calculations on supercomputers. Among five models, namely, extreme gradient boosting regression tree (XGBoost), adaptive boosting, random forest, multi-layer perceptron, and kernel ridge regression, were respectively trained and evaluated by stratified sampling and 5-fold cross-validation method. Among them, XGBoost model produced the best scoring metrics in predicting the detonation velocity, detonation pressure, heat of explosion, decomposition temperature, and lattice energy of HEDMs, and XGBoost predictions agreed best with the 1,383 experimental data collected from massive literatures. Feature importance analysis was conducted to obtain data-driven insight into the causality of the performance-stability contradiction and delivered the optimal range of key features for more efficient rational design of advanced HEDMs.

Review of the molecular and crystal correlations on sensitivities of energetic materials

Highly efficient design on the levels of molecule and crystal, as well as formulation, is highly desired for accelerating the development of energetic materials (EMs). Sensitivity is one of the most important characteristics of EMs and should be compulsorily considered in the design. However, owing to multiple factors responsible for the sensitivity, it usually undergoes a low predictability. Thus, it becomes urgent to clarify which factors govern the sensitivity and what is the importance of these factors. The present article focuses upon the progress of the molecular and crystal correlations on the sensitivity, and the molecule-based numerical models for sensitivity prediction in the past decades. On the molecular level, composition, geometric structure, electronic structure, energy and reactivity can be correlated with the sensitivity; while the sensitivity can be also related with molecular packing pattern, intermolecular interaction, crystal morphology, crystal size and distribution, crystal surface/interface and crystal defect on the crystal level. And most of these factors, in particle on the crystal level, have been employed as variables in numerical models for predicting sensitivity of categorized EMs. Besides, we stress that more attention should be paid to the sensitivity correlations on the inherent structures of EMs, molecule and crystal packing, because they can be readily dealt by molecular simulations nowadays, facilitating to reveal the physical nature of sensitivity.

Impact sensitivity of aryl diazonium chlorides: Limitations of molecular and solid-state approach

The mechanism of the compression-induced decomposition of aryl diazonium chlorides is proposed on the basis of quantum-chemical calculations of both the isolated cations and crystalline materials. The electron transfer from the anion to the cation, followed by the crystal decomposition, is observed with the rise of pressure. Taking the known nature of the structural changes in cations undergone upon reduction, five structural, vibrational and electronic determinants of impact sensitivity are found. Thus, a correlation (R² = 0.79) between the experimentally known impact sensitivity of 40 different aryl diazonium cations and the developed empirical function Ω, which includes the above-mentioned parameters, is obtained. Meanwhile, an abnormal impact sensitivity of 4-nitrobenzenediazonium chloride (4 J) compared to the parent benzenediazonium chloride (3 J) is rationalized on the basis of first-principles calculations of the latter and its three nitro derivatives. Using our recently proposed solid-state criteria of impact sensitivity, another empirical function Ω was developed being able to predict impact sensitivity of these four salts with very good confidence (R² = 0.97).

Applying machine learning techniques to predict the properties of energetic materials

We present a proof of concept that machine learning techniques can be used to predict the properties of CNOHF energetic molecules from their molecular structures. We focus on a small but diverse dataset consisting of 109 molecular structures spread across ten compound classes. Up until now, candidate molecules for energetic materials have been screened using predictions from expensive quantum simulations and thermochemical codes. We present a comprehensive comparison of machine learning models and several molecular featurization methods - sum over bonds, custom descriptors, Coulomb matrices, Bag of Bonds, and fingerprints. The best featurization was sum over bonds (bond counting), and the best model was kernel ridge regression. Despite having a small data set, we obtain acceptable errors and Pearson correlations for the prediction of detonation pressure, detonation velocity, explosive energy, heat of formation, density, and other properties out of sample. By including another dataset with ≈300 additional molecules in our training we show how the error can be pushed lower, although the convergence with number of molecules is slow. Our work paves the way for future applications of machine learning in this domain, including automated lead generation and interpreting machine learning models to obtain novel chemical insights.

QSAR modeling in ecotoxicological risk assessment: application to the prediction of acute contact toxicity of pesticides on bees (Apis mellifera L.)

Despite their indisputable importance around the world, the pesticides can be dangerous for a range of species of ecological importance such as honeybees (Apis mellifera L.). Thus, a particular attention should be paid to their protection, not only for their ecological importance by contributing to the maintenance of wild plant diversity, but also for their economic value as honey producers and crop-pollinating agents. For all these reasons, the environmental protection requires the resort of risk assessment of pesticides. The goal of this work was therefore to develop a validated QSAR model to predict contact acute toxicity (LD₅₀) of 111 pesticides to bees because the QSAR models devoted to this species are very scarce. The analysis of the statistical parameters of this model and those published in the literature shows that our model is more efficient. The QSAR model was assessed according to the OECD principles for the validation of QSAR models. The calculated values for the internal and external validation statistic parameters (Q ² and [Formula: see text] are greater than 0.85. In addition to this validation, a mathematical equation derived from the ANN model was used to predict the LD₅₀ of 20 other pesticides. A good correlation between predicted and experimental values was found (R ² = 0.97 and RMSE = 0.14). As a result, this equation could be a means of predicting the toxicity of new pesticides.

Computational study of the structure and properties of bicyclo[3.1.1]heptane derivatives for new high-energy density compounds with low impact sensitivity

To design new high-energy density compounds (HEDCs), a series of new bicyclo[2.2.1]heptane derivatives containing an aza nitrogen atom and nitro substituent were designed and studied theoretically. The density, heat of sublimation and impact sensitivity were estimated by electrostatic potential analysis of the molecular surface. Based on the designed isodesmic reaction, and the reliable heat of formation (HOF) of the reference compounds, HOFs were calculated and compared at B3LYP/6-311G(d,p) and B3P86/6-311G(d,p), respectively. The detonation performances, bond dissociation energies (BDE) and impact sensitivity were calculated to evaluate the designed compounds. The calculated results show that the number of aza nitrogen atoms and NO₂ groups are two important factors for improving HOF, density and detonation properties. Thermal stability generally decreases with increasing nitro groups. And the N-NO₂ bond is the trigger bond for all designed compounds except B8, whose trigger bond is C-NO₂. Importantly, the BDE values are between 86.95 and 179.71 kJ mol^-1 and meet the requirement for HEDCs. Detonation velocity and detonation pressure were found to be 5.77-9.65 km s^-1 and 12.30-43.64 GPa, respectively. After comprehensive consideration of thermal stability, impact sensitivity and detonation properties, A7, A8, B8, C8, D7, E7, F7 and G6 may be considered as potential HEDCs. Especially, A8, B8, C8, and D7 have better detonation properties than the famous caged nitramine CL-20 (D = 9.40 km/s, P = 42.00GPa). Besides, all the designed potential HEDCs have reasonable impact sensitivity. Graphical abstract New high-energy density compounds (HEDCs) with low impact sensitivity (A8, B8, C8 and D7 have better detonation properties than CL-20).

A Quantitative Structure Activity Relationship for acute oral toxicity of pesticides on rats: Validation, domain of application and prediction

Quantitative Structure Activity Relationship (QSAR) models are expected to play an important role in the risk assessment of chemicals on humans and the environment. In this study, we developed a validated QSAR model to predict acute oral toxicity of 329 pesticides to rats because a few QSAR models have been devoted to predict the Lethal Dose 50 (LD50) of pesticides on rats. This QSAR model is based on 17 molecular descriptors, and is robust, externally predictive and characterized by a good applicability domain. The best results were obtained with a 17/9/1 Artificial Neural Network model trained with the Quasi Newton back propagation (BFGS) algorithm. The prediction accuracy for the external validation set was estimated by the Q(2)ext and the root mean square error (RMS) which are equal to 0.948 and 0.201, respectively. 98.6% of external validation set is correctly predicted and the present model proved to be superior to models previously published. Accordingly, the model developed in this study provides excellent predictions and can be used to predict the acute oral toxicity of pesticides, particularly for those that have not been tested as well as new pesticides.

Predicting impact sensitivities of nitro compounds on the basis of a semi-empirical rate constant

A physically motivated model is put forward to estimate impact sensitivity of nitro compounds on the basis of the relationship h(50) ∝ k(pr)(-4) between drop weight impact test data h(50) and rate constant k(pr) for the propagation of the decomposition. An approximate expression involving two adjustable parameters is introduced to estimate k(pr) from molecular structure. As a result, using only a hand-held calculator, ln(h(50)) values are estimated with a good reliability (R(2) ≃ 0.8) compared to previous schemes. These results support the underlying assumption that sensitivity primarily depends on the ability of reacting species to propagate the decomposition before the released energy dissipates away.

Toward a physically based quantitative modeling of impact sensitivities

Among the subsequent steps leading from impact to explosive decomposition in nitro compounds, the ability of early exothermic reactions to trigger the decomposition of neighboring molecules before the released energy has dissipated away is assumed to be critical. The rate of this process is roughly estimated using as inputs the energy content and the dissociation energy of the weakest X-NO2 bonds. While the sensitivity index thus obtained was previously shown to exhibit striking correlations with gap test pressures, its correlation with drop weight impact test data is poorer. Nevertheless, considering four different subsets of molecules depending on the environment of the most labile nitro groups, straightforward regressions against this sensitivity index yield a practical method to estimate impact sensitivity, whose combination of fair performance and generality is provided by no alternative approach, except purely empirical models based on extensive parametrization.

Development of validated QSPR models for impact sensitivity of nitroaliphatic compounds

The European regulation of chemicals named REACH implies the assessment of a large number of substances based on their hazardous properties. However, the complete characterization of physico-chemical, toxicological and eco-toxicological properties by experimental means is incompatible with the imposed calendar of REACH. Hence, there is a real need in evaluating the capabilities of alternative methods such as quantitative structure-property relationship (QSPR) models, notably for physico-chemical properties. In the present work, the molecular structures of 50 itroaliphatic compounds were correlated with their impact sensitivities (h(50%)) using such predictive models. More than 400 olecular descriptors (constitutional, topological, geometrical, quantum chemical) were calculated and linear and multi-linear regressions were performed to find accurate quantitative relationships with experimental impact sensitivities. Considering different sets of descriptors, four predictive models were obtained and two of them were selected for their predictive reliability. To our knowledge, these QSPR models for the impact sensitivity of nitroaliphatic compounds are the first ones being rigorously validated (both internally and externally) with defined applicability domains. They hence follow all OECD principles for regulatory acceptability of QSPRs, allowing possible application in REACH.