Article

ReacNetGenerator: an automatic reaction network generator for reactive molecular dynamic simulations

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

Reactive molecular dynamics (MD) simulation makes it possible to study the reaction mechanisms of complex reaction systems at the atomic level. However, the analysis of the MD trajectories which contain thousands of species and reaction pathways has become a major obstacle to the application of reactive MD simulation in large-scale systems. Here, we report the development and application of the Reaction Network Generator (ReacNetGenerator) method. It can automatically extract the reaction network from the reaction trajectory without any predefined reaction coordinates and elementary reaction steps. Molecular species can be automatically identified from the cartesian coordinates of atoms and the hidden Markov model is used to filter the trajectory noises which makes the analysis process easier and more accurate. The ReacNetGenerator has been successfully used to analyze the reactive MD trajectories of the combustion of methane and 4-component surrogate fuel for rocket propellant 3 (RP-3), and it has great advantages in efficiency and accuracy compared to traditional manual analysis.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... Reaction network for hydrogen combustion.The most important reason for reactive MD simulations of chemical reaction systems is to determine the reaction mechanism. To this end, we analyzed the trajectory with the help of ReacNetGenerator40.Fig. 3 shows ...
Preprint
Full-text available
Understanding and prediction of the chemical reactions are fundamental demanding in the study of many complex chemical systems. Reactive molecular dynamics (MD) simulation has been widely used for this purpose as it can offer atomic details and can help us better interpret chemical reaction mechanisms. In this study, two reference datasets were constructed and corresponding neural network (NN) potentials were trained based on them. For given large-scale reaction systems, the NN potentials can predict the potential energy and atomic forces of DFT precision, while it is orders of magnitude faster than the conventional DFT calculation. With these two models, reactive MD simulations were performed to explore the combustion mechanisms of hydrogen and methane. Benefit from the high efficiency of the NN model, nanosecond MD trajectories for large-scale systems containing hundreds of atoms were produced and detailed combustion mechanism was obtained. Through further development, the algorithms in this study can be used to explore and discovery reaction mechanisms of many complex reaction systems, such as combustion, synthesis, and heterogeneous catalysis without any predefined reaction coordinates and elementary reaction steps.
... It may be due to the high temperature that caus tion to take place. Then, the ·CH3O radicals release a ·H to form formaldehyd The trajectory was further analyzed by the ReacNetGenerator software [43], which can construct reaction networks automatically from atomic coordinates. The main reaction network detected from the trajectory is shown in Figure 6. ...
Article
Full-text available
We develop a fragment-based ab initio molecular dynamics (FB-AIMD) method for efficient dynamics simulation of the combustion process. 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 FB-AIMD calculated energies and atomic forces of several different systems with those obtained by standard full system quantum calculations. The computational cost of the FB-AIMD method scales linearly with the size of the system, and the calculation is easily parallelizable. The method is applied to methane combustion as a benchmark. Detailed reaction network of methane reaction is analyzed, and important reaction species are tracked in real time. The current result of methane simulation is in excellent agreement with known experimental findings and with prior theoretical studies.
... Interactive frameworks for exploring chemical reaction space have been developed, for example Molpher software [56]. Bespoke automated computational approaches for generating CRNRs can be constructed modularly, for example by integrating graph grammar operations such as those used in MØD [41], reactive molecular dynamics tools such as ReacNetGenerator [57] and Python (programming language frameworks) such as Reaction Mechanism Generator (RMG) [58,59], CGRtools (Condensed Graph of Reaction) [60], and Rule Input Network Generator (RING) for generating CRNRs from complex reactive systems [61], among others. Other tools potentially useful for constructing such pipelines include pReSt [62] to discover novel chemistries in automated CRNRs, and CERENA (ChEmical REaction Network Analyzer) [63] to model stochastic chemical kinetics in chemical reaction networks. ...
Article
Full-text available
Prebiotic chemistry often involves the study of complex systems of chemical reactions that form large networks with a large number of diverse species. Such complex systems may have given rise to emergent phenomena that ultimately led to the origin of life on Earth. The environmental conditions and processes involved in this emergence may not be fully recapitulable, making it difficult for experimentalists to study prebiotic systems in laboratory simulations. Computational chemistry offers efficient ways to study such chemical systems and identify the ones most likely to display complex properties associated with life. Here, we review tools and techniques for modelling prebiotic chemical reaction networks and outline possible ways to identify self-replicating features that are central to many origin-of-life models.
... The analysis of MD trajectories, which contain thousands of species and reaction pathways, has become a major obstacle to the application of reactive MD simulation in large-scale systems. In the current study, the ReacNetGenerator method 45 ...
Preprint
Full-text available
div> Reactive molecular dynamics (MD) simulation is a powerful tool to study the reaction mechanism of complex chemical systems. Central to the method is the potential energy surface (PES) that can describe the breaking and formation of chemical bonds. The development of PES of both accurate and efficent has attracted significant effort in the past two decades. Recently developed Deep Potential (DP) model has the promise to bring ab initio accuracy to large-scale reactive MD simulations. However, for complex chemical reaction processes like pyrolysis, it remains challenging to generate reliable DP models with an optimal training dataset. In this work, a dataset construction scheme for such a purpose was established. The employment of a concurrent learning algorithm allows us to maximize the exploration of the chemical space while minimize the redundancy of the dataset. This greatly reduces the cost of computational resources required by ab initio calculations. Based on this method, we constructed a dataset for the pyrolysis of n-dodecane, which contains 35,496 structures. The reactive MD simulation with the DP model trained based on this dataset revealed the pyrolysis mechanism of n-dodecane in detail, and the simulation results are in good agreement with the experimental measurements. In addition, this dataset shows excellent transferability to different long-chain alkanes. These results demonstrate the advantages of the proposed method for constructing training datasets for similar systems. </div
... Atomic-level simulations have contributed much to the understanding of the mechanisms of pyrolysis and combustion of hydrocarbons [15][16][17][18][19][20][21][22][23][24][25]. Reaction mechanisms and reaction rates can be extracted [15][16][17] from atomistic simulations such as MD simulations and compared directly to experimental data. ...
Preprint
The high computational cost of evaluating atomic interactions recently motivated the development of computationally inexpensive kinetic models, which can be parametrized from MD simulations and accelerate the prediction of the chemical evolution by up to four order of magnitude. Such kinetic models utilize molecular descriptions of reactions and have been constrained to only reproduce molecules previously observed in MD simulations. Therefore, these descriptions fail to predict the reactivity of unobserved molecules, for example in the case of large molecules or solids. Here we propose a new approach for the extraction of reaction mechanisms and reaction rates from MD simulations, namely the use of atomic-level features. Using the complex chemical network of hydrocarbon pyrolysis as example, it is demonstrated that kinetic models built using atomic features are able to explore chemical reaction pathways never observed in the MD simulations used to parametrize them, a critical feature to describe rare events. Atomic-level features are shown to construct reaction mechanisms and estimate reaction rates of unknown molecular species from elementary atomic events. Through comparisons of the model ability to extrapolate to longer simulation timescales and different chemical compositions than the ones used for parameterization, it is demonstrated that kinetic models employing atomic features retain the same level of accuracy and transferability as the use of features based on molecular species, while being more compact and parametrized with less data. We also find that atomic features can better describe the formation of large molecules enabling the simultaneous description of small molecules and condensed phases.
... 0.1 fs timestep was employed for all simulations up to 1 ns and trajectory was written every 10 fs for post-analysis. The evolution of radicals and reaction networks was analyzed with the efficient tool, called ReacNetGenerator [47]. ...
Article
Full-text available
Gasification with supercritical water is an efficient process that can be used for the valorization of biomass. Lignin is the second most abundant biopolymer in biomass and its conversion is fundamental for future energy and value-added chemicals. In this paper, the supercritical water gasification process of lignin by employing reactive force field molecular dynamics simulations (ReaxFF MD) was investigated. Guaiacyl glycerol-β-guaiacyl ether (GGE) was considered as a lignin model to evaluate the reaction mechanism and identify the components at different temperatures from 1000 K to 5000 K. The obtained results revealed that the reactions and breaking of the lignin model started at 2000 K. At the primary stage of the reaction at 2000 K the β-O-4 bond tends to break into several compounds, forming mainly guaiacol and 1,3-benzodioxole. In particular, 1,3-benzodioxole undergoes dissociation and forms cyclopentene-based ketones. Afterward, dealkylation reaction occurred through hydroxyl radicals of water to form methanol, formaldehyde and methane. Above 2500 K, H2, CO and CO2 are predominantly formed in which water molecules contributed hydrogen and oxygen for their formation. Understanding the detailed reactive mechanism of lignin’s gasification is important for efficient energy conversion of biomass.
Preprint
Full-text available
Combustion is a kind of important reaction that affects people's daily lives and the development of aerospace. Exploring the reaction mechanism contributes to the understanding of combustion and the more efficient use of fuels. Ab initio quantum mechanical (QM) calculation is precise but limited by its computational time for large-scale systems. In order to carry out reactive molecular dynamics (MD) simulation for combustion accurately and quickly, we develop the MFCC-combustion method in this study, which calculates the interaction between atoms using QM method at the level of MN15/6-31G(d). Each molecule in systems is treated as a fragment, and when the distance between any two atoms in different molecules is greater than 3.5 Å, a new fragment involved two molecules is produced in order to consider the two-body interaction. The deviations of MFCC-combustion from full system calculations are within a few kcal/mol, and the result clearly shows that the calculated energies of the different systems using MFCC-combustion are close to converging after the distance thresholds are larger than 3.5 Å for the two-body QM interactions. The methane combustion was studied with the MFCC-combustion method to explore the combustion mechanism of the methane-oxygen system.
Article
To understand the influence of volatile removal ratio from the system on pyrolysis characteristic of cotton, a simulation model was proposed by performing both non-isothermal and isothermal simulations using ReaxFF molecular dynamics. Thermogravimetric-Fourier transform infrared (TG-FTIR) was used to study the overall evolution behaviors of volatile from cotton pyrolysis as a reference to simulation results. In non-isothermal simulations, volatile was remained inside the simulation unit during the entire process, while in isothermal simulations a portion of volatile was removed from the simulation unit before starting next simulation at a higher temperature. Results showed that although the non-isothermal simulation method can predict a TG curve similar to the one obtained from the experiment and can reveal the initial evolution behaviors of the volatile, but the atomic contents of chars were unpractical. In isothermal simulations, 100, 80, 60 and 30 % volatile removal ratios (VRR) were employed to resemble the different extent of secondary reactions that volatiles undergo in practical pyrolysis. It was found that contents of aldehydes and C–H were mostly influenced by VRR; and simulations with higher VRR produced chars with lower O and H contents, closing to the practical chars. Simulations with 80, 100 % of VRR had a higher probability to match the reality.
Article
Reactive molecular dynamics (MD) simulation is a powerful tool to study the reaction mechanism of complex chemical systems. Central to the method is the potential energy surface (PES) that can describe the breaking and formation of chemical bonds. The development of both accurate and efficient PES has attracted significant effort in the past 2 decades. A recently developed deep potential (DP) model has the promise to bring ab initio accuracy to large-scale reactive MD simulations. However, for complex chemical reaction processes like pyrolysis, it remains challenging to generate reliable DP models with an optimal training data set. In this work, a data set construction scheme for such a purpose was established. The employment of a concurrent learning algorithm allows us to maximize the exploration of the chemical space while minimizing the redundancy of the data set. This greatly reduces the cost of computational resources required for ab initio calculations. Based on this method, we constructed a data set for the pyrolysis of n-dodecane, which contains 35 496 structures. The reactive MD simulation with the DP model trained based on this data set revealed the pyrolysis mechanism of n-dodecane in detail, and the simulation results are in good agreement with the experimental measurements. In addition, this data set shows excellent transferability to different long-chain alkanes. These results demonstrate the advantages of the proposed method for constructing training data sets for similar systems.
Article
This work reports the results of a perspective workshop held in summer 2021 discussing the current status and future needs for multiscale modeling in reaction engineering. This research topic is one of the most challenging and likewise most interdisciplinary in the chemical engineering community, today. Although it is progressing fast in terms of methods development, it is only slowly applied by most reaction engineers. Therefore, this perspective is aimed to promote this field and facilitate research and a common understanding. It involves the following areas: (1) reactors and cells with surface changes focusing on Density Functional Theory and Monte-Carlo simulations; (2) hierarchically-based microkinetic analysis of heterogeneous catalytic processes including structure sensitivity, microkinetic mechanism development, and parameter estimation; (3) coupling first-principles kinetic models and CFD simulations of catalytic reactors covering chemistry acceleration strategies and surrogate models; and finally (4) catalyst-reactor-plant systems with details on linking CFD with plant simulations, respectively. It therefore highlights recent achievements, challenges, and future needs for fueling this urgent research topic in reaction engineering.
Article
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. In this work, ab initio based neural network potential (NNP) energy surfaces for both β-CL-20 and CL-20/TNT co-crystals were constructed. To accurately simulate the thermal decomposition processes of these two crystal systems, reactive molecular dynamics simulations based on the NNPs were performed. Many important intermediate species and their associated reaction paths during the decomposition had been identified in the simulations and the direct results on detonation temperatures of both systems were provided. The simulations also showed clearly that 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 significant at lower temperatures. Specifically, the addition of TNT molecules in the CL-20/TNT co-crystal introduces intermolecular hydrogen bonds between CL-20 and TNT molecules in the system, thereby increasing the thermal stability of the co-crystal. The current reactive molecular dynamics simulation is performed based on the NNP which helps in accelerating the speed of ab initio molecular dynamics (AIMD) simulation by more than 3 orders of magnitude while preserving the accuracy of density functional theory (DFT) calculations. This enabled us to perform longer-time simulations at more realistic temperatures that traditional AIMD methods cannot achieve. With the advantage of the NNP in its powerful fitting ability and transferability, the NNP-based MD simulation can be widely applied to energetic material systems.
Article
Novel ferrocene functionalized graphene with different molecular structures were designed, fabricated and characterized via SEM, EDS, FTIR, XPS and RAMAN methods. SEM results show the two-dimensional structure of the as-prepared catalysts, and the active metal Fe is uniformly distributed on the surface of graphene. The FTIR, XPS and RAMAN results confirmed the successful preparation of ferrocene functionalized graphene. The catalytic effects of the as-synthesized catalysts for the thermal decomposition of energetic TKX-50 were monitored by DSC, and the corresponding kinetic parameters were calculated using multi kinetic methods including traditional and nonlinear models. The results showed that the as-prepared ferrocene functionalized graphene can effectively promote the thermal decomposition of TKX-50 with the reduced decomposition peak temperatures and activation energies. In addition, the effects of ferrocene functionalized graphene for TKX-50 decomposition are reflected in both high and low temperature stages, and the effect on the high temperature stage is more significant. The outstanding catalytic activity of ferrocene functionalized graphene is related not only to the good dispersion of active Fe, but also to the enhanced interaction of small molecule products on two-dimensional graphene. Among the ferrocene functionalized graphene studied, G-792-Fe and G-902-Fe exhibit better catalytic effects on the thermal decomposition of TKX-50, which can be used as candidate catalysts for TKX-50-based solid propellants.
Article
We explored the mechanism of ethylene combustion by combining density functional tight-binding based nanoreactor molecular dynamic method (DFTB-NMD) and a hidden Markov model (HMM) based reaction network generator approach. The...
Article
The pyrolysis and combustion mechanism of the hydrocarbon fuel has important scientific and practical significance. However, it is difficult to detect the whole intermediates and products using traditional methods, which brings trouble to the analysis of the reaction process. In this paper, the microscopic reaction mechanism and the main products of n-eicosane (C20H42) were simulated based on the reactive force field molecular dynamics (ReaxFF-MD). The effects of temperature (2000–3500 K) and oxygen on the initial decomposition, the distribution of main products, and the reactive pathways of C20H42 fuel were studied to determine its reaction mechanism. The initial decomposition of C20H42 was mainly initiated by small alkyl radicals in pyrolysis, and by the oxygen-containing radicals in combustion. The participation of oxygen had a greater effect on accelerating the decomposition reaction. The reactions involving oxygen of C20H42 initial decomposition accounted for 87.5% of the total reactions at 2000 K. Moreover, the detailed distribution and formation pathways of the main products of H2, C2H4, CH4, H2O, CO, and CO2 were depicted to construct the overall reaction mechanism of C20H42. •H radical formed from the composition of C2H4 was exactly consistent with the •H radical consumed by the generation of CH4 and H2 in the pyrolysis stage. The feasibility of the simulation method was verified by the result of thermal analysis. The results are helpful for further research on the reaction mechanism of hydrocarbon fuels.
Article
Full-text available
Ethylcyclohexane (ECH) is a model compound for cycloalkanes with long alkyl side-chains. A preliminary investigation on ECH (Wang et al., Proc. Combust. Inst., 35, 2015, 367–375) revealed that an accurate ECH kinetic model with detailed fuel consumption mechanism and aromatic growth pathways, as well as additional ECH pyrolysis and oxidation data with detailed species concentration covering a wide pressure and temperature range are required to understand the ECH combustion kinetics. In this work, the flow reactor pyrolysis of ECH at various pressures (30, 150 and 760 Torr) was studied using synchrotron vacuum ultraviolet (VUV) photoionization mass spectrometry (PIMS) and gas chromatography (GC). The mole fraction profiles of numerous major and minor species were evaluated, and good agreement was observed between the PIMS and GC data sets. Furthermore, a fuel-rich burner-stabilized laminar premixed ECH/O2/Ar flame at 30 Torr was studied using synchrotron VUV PIMS. A detailed kinetic model for ECH high temperature pyrolysis and oxidation was developed and validated against the pyrolysis and flame data performed in this work. Further validation of the kinetic model is presented against literature data including species concentrations in jet-stirred reactor oxidation, ignition delay times in a shock tube, and laminar flame speeds at various pressures and equivalence ratios. The model well predicts the consumption of ECH, the growth of aromatics, and the global combustion properties. Reaction flux and sensitivity analysis were utilized to elucidate chemical kinetic features of ECH combustion under various reaction conditions.
Article
Full-text available
We provide a methodology for deducing quantitative reaction models from reactive molecular dynamics simulations by identifying, quantifying, and evaluating elementary reactions of classical trajectories. Simulations of the inception stage of methane oxidation are used to demonstrate our methodology. The agreement of pathways and rates with available literature data reveals the potential of reactive molecular dynamics studies for developing quantitative reaction models.
Article
Full-text available
In order to develop efficient suppression techniques and isolation materials for gas explosion, it is greatly vital to understand the intrinsic mechanism of gas explosion at atomic level. The methane explosion process was systematically explored by reactive force field molecular dynamics (MD) simulation and density function theory (DFT) calculations. The results show that the initiation step of the gas explosion process is mainly induced by center dot OH free radical, and the chain transfer process is primarily carried by the conversion from center dot HO2 to center dot OH radical. The number of free radicals, such as center dot CH3, center dot HO2 and center dot OH, greatly affect the reaction velocity during the gas explosion process. These results not only identify the critical reaction intermediates and main radical carriers of methane explosion process, but also they will provide significant theoretical guide for development of the novel and efficient gas explosion suppression methods and isolation materials.
Article
Full-text available
Laminar flame speeds of n-decane/air mixtures were determined experimentally over an extensive range of equivalence ratios at 500 K and at atmospheric pressure. The effect of N2 dilution on the laminar flame speed was also studied at these same conditions. The experiments employed the stagnation jet–wall flame configuration with the flow velocity field determined by particle image velocimetry. Reference laminar flame speeds were obtained using linear extrapolation from low to zero stretch rate. The determined flame speeds are significantly different from those predicted using existing published kinetic models, including a model validated previously against high-temperature data from flow reactor, jet-stirred reactor, shock tube ignition delay, and burner-stabilized flame experiments. A significant update of this model is described that continues to predict the earlier validation experiments as well as the newly acquired laminar flame speed data and other recently published shock-tube ignition delay measurements.
Conference Paper
Full-text available
NetworkX is a Python language package for exploration and analysis of networks and network algorithms. The core package provides data structures for representing many types of networks, or graphs, including simple graphs, directed graphs, and graphs with parallel edges and self loops. The nodes in NetworkX graphs can be any (hashable) Python object and edges can contain arbitrary data; this flexibility mades NetworkX ideal for representing networks found in many different scientific fields. In addition to the basic data structures many graph algorithms are implemented for calculating network properties and structure measures: shortest paths, betweenness centrality, clustering, and degree distribution and many more. NetworkX can read and write various graph formats for eash exchange with existing data, and provides generators for many classic graphs and popular graph models, such as the Erdoes-Renyi, Small World, and Barabasi-Albert models, are included. The ease-of-use and flexibility of the Python programming language together with connection to the SciPy tools make NetworkX a powerful tool for scientific computations. We discuss some of our recent work studying synchronization of coupled oscillators to demonstrate how NetworkX enables research in the field of computational networks.
Article
Full-text available
A detailed kinetic model is proposed for the combustion of normal alkanes up to n-dodecane above 850 K. The model was validated against experimental data, including fuel pyrolysis in plug flow and jet-stirred reactors, laminar flame speeds, and ignition delay times behind reflected shock waves, with n-dodecane being the emphasis. Analysis of the computational results reveal that for a wide range of combustion conditions, the kinetics of fuel cracking to form smaller molecular fragments is fast and may be decoupled from the oxidation kinetics of the fragments. Subsequently, a simplified model containing a minimal set of 4 species and 20 reaction steps was developed to predict the fuel pyrolysis rate and product distribution. Combined with the base C1-C4 model, the simplified model predicts fuel pyrolysis rate and product distribution, laminar flame speeds, and ignition delays as close as the detailed reaction model.
Article
Full-text available
ABSTRACT: A frequent problem in computational modeling is the interconversion of chemical structures between different formats. While standard interchange formats exist (for example, Chemical Markup Language) and de facto standards have arisen (for example, SMILES format), the need to interconvert formats is a continuing problem due to the multitude of different application areas for chemistry data, differences in the data stored by different formats (0D versus 3D, for example), and competition between software along with a lack of vendor-neutral formats. We discuss, for the first time, Open Babel, an open-source chemical toolbox that speaks the many languages of chemical data. Open Babel version 2.3 interconverts over 110 formats. The need to represent such a wide variety of chemical and molecular data requires a library that implements a wide range of cheminformatics algorithms, from partial charge assignment and aromaticity detection, to bond order perception and canonicalization. We detail the implementation of Open Babel, describe key advances in the 2.3 release, and outline a variety of uses both in terms of software products and scientific research, including applications far beyond simple format interconversion. Open Babel presents a solution to the proliferation of multiple chemical file formats. In addition, it provides a variety of useful utilities from conformer searching and 2D depiction, to filtering, batch conversion, and substructure and similarity searching. For developers, it can be used as a programming library to handle chemical data in areas such as organic chemistry, drug design, materials science, and computational chemistry. It is freely available under an open-source license from http://openbabel.org.
Article
Full-text available
We extend the reactive force field ReaxFF to describe the high energy nitramine RDX and use it with molecular dynamics (MD) to study its shock-induced chemistry. We studied shock propagation via nonequilibrium MD simulations at various collision velocities. We find that for high impact velocities (>6 km/s) the RDX molecules decompose and react to form a variety of small molecules in very short time scales (<3 ps). These products are consistent with those found experimentally at longer times. For lower velocities only NO2 is formed, also in agreement with experiments.
Article
Full-text available
We use the recently developed reactive force field ReaxFF with molecular dynamics to study thermal induced chemistry in RDX [cyclic-[CH(2)N(NO(2))](3)] at various temperatures and densities. We find that the time evolution of the potential energy can be described reasonably well with a single exponential function from which we obtain an overall characteristic time of decomposition that increases with decreasing density and shows an Arrhenius temperature dependence. These characteristic timescales are in reasonable quantitative agreement with experimental measurements in a similar energetic material, HMX [cyclic-[CH(2)N(NO(2))](4)]. Our simulations show that the equilibrium population of CO and CO(2) (as well as their time evolution) depend strongly of density: at low density almost all carbon atoms form CO molecules; as the density increases larger aggregates of carbon appear leading to a C deficient gas phase and the appearance of CO(2) molecules. The equilibrium populations of N(2) and H(2)O are more insensitive with respect to density and form in the early stages of the decomposition process with similar timescales.
Article
ReaxFF is a computationally efficient force field to simulate complex reactive dynamics in extended molecular models with diverse chemistries, if reliable force-field parameters are available for the chemistry of interest. If not, they must be optimized by minimizing the error ReaxFF makes on a relevant training set. Because this optimization is far from trivial, many methods, in particular genetic algorithms (GAs), have been developed to search for the global optimum in parameter space. Recently, two alternative parameter calibration techniques were proposed, i.e. Monte-Carlo Force Field optimizer (MCFF) and Covariance Matrix Adaptation Evolutionary Strategy (CMA-ES). In this work, CMA-ES, MCFF and a GA method (OGOLEM) are systematically compared using three training sets from the literature. By repeating optimizations with different random seeds and initial parameter guesses, it is shown that a single optimization run with any of these methods should not be trusted blindly: non-reproducible, poor or premature convergence are common deficiencies. GA shows the smallest risk of getting trapped into a local minimum, whereas CMA-ES is capable of reaching the lowest errors for two third of the cases, albeit not systematically. For each method, we provide reasonable default settings and our analysis offers useful guidelines for their usage in future work. An important side effect impairing the parameter optimization is numerical noise. A detailed analysis reveals that it can be reduced, e.g. by using exclusively unambiguous geometry optimizations in the training set. Even without this noise, many distinct near-optimal parameter vectors can be found, which opens new avenues for improving the training set and detecting overfitting artifacts.
Article
With the development of theoretical and computational chemistry, as well as high-performance computing, molecular simulation can now be used not only as a tool to explain the experimental results but also as a means for discovery or prediction. Quantum chemical nanoreactor is such a method which can automatically explore the chemical process based only on the basic mechanics without prior knowledge of the reactions. Here, we present a new method which combines the semi-empirical quantum mechanical method density functional tight-binding (DFTB) with the nanoreactor molecular dynamic (NMD) method and we simulated the reaction process of graphene synthesis via detonation at different oxygen/acetylene mole ratio. The formation of graphene is initiated by the breaking of acetylene (C2H2) molecules by collision into pieces such as H atoms, ethynyl (HC≡C•) and vinylidene (H2C=C:) radicals. It is followed by the formation of long straight carbon chains coupled with few branched carbon chains, which then turned into the 2-D framework made of carbon rings. Trace oxygen could modulate the size of the rings during graphene formation, and promote the formation of regular graphene with fused six-membered rings as we see, but the addition of high oxygen content makes more C-containing species oxidized to small oxide molecules instead of polymerization. The calculation speed of the DFTB nanoreactor is greatly improved compared to ab initio nanoreactor, which makes it a valuable option to simulate chemical process of large sizes and long timescales, and to help us uncover the "unknown unknowns".
Book
This book differs from its out of print 1984 predecessorl primarily by lacking theoretical chapters on combustion modeling and elementary reaction rate coeffi­ cients. While noteworthy advances in these subjects have been made since 1984, it was decided to mention theory in this book only where appropriate in chap­ ters describing combustion chemistry itself. Otherwise, space limitation would have forced us to discuss only new developments in theoretical areas, thereby abandoning our goal of keeping this book readable by newcomers to the field of combustion modeling. Contemporary modeling and rate coefficient theory as applied to combustion deserve a book of their own. A second omission is a chapter devoted to reviewing the elementary reactions that contribute to the combustion chemistry of hydrogen, carbon monoxide, and hydrocarbon or alternate fuels. Readers looking for guidance to the current knowledge we have in this area will find a broad outline and extensive references to the review and archival literature in Chapter 1, where the essential features of combustion chemistry modeling are surveyed.
Article
This paper presents long time ReaxFF MD simulations of fuel-rich combustion for up to 10 ns to explore the initial mechanism of soot nanoparticle formation. A 24-component rocket propellant 1 (RP-1) model based on the major components of RP-1 fuel was employed. Simulations were performed by GPU-accelerated code GMD-Reax and reactions therein was revealed with the aid of VARxMD. Simulated evolution of physical and chemical properties of the largest molecule exhibits the overall structural transitions of three stages for incipient ring formation, nucleation and graphitization from fuel molecules to the formation of a single soot nanoparticle. The incipient ring formation takes place in stage 1 by large ring generation from activated aliphatic polyyne-like chains, ring number increase from internal bridging between carbon atoms of large rings, and consequent formation of PAH-like molecules with aliphatic side chains. Nucleation of a nanoparticle in stage 2 is the result of coalescence of PAH-like molecules, accompanied by ring closure reactions occurred at side chains of the PAH-like molecules, and following formation of internal bridged bonds. Graphitization of the nanoparticle in stage 3 is dominated by the transformation from C5/C7 rings to C6 rings with C3 rings as intermediates. This work demonstrates that ReaxFF MD simulation might be a promising approach for qualitatively characterizing morphological evolution and the underlying chemical complexity of soot formation using multi-component fuel models.
Article
ε-2,4,6,8,10,12-hexanitro-2,4,6,8,10,12-hexaazaisowurtzitane (CL-20) is currently the most powerful explosive commercially available. Nevertheless, the early decay events of shocked ε-CL-20 still remain unclear. We perform quantum based self-consistent charge density-functional tight-binding molecular dynamics simulations, in combination with the multiscale shock simulation technique, to reveal the events with four specified shock velocities (Us) of 8 to 11 km/s. We find that the temperature and pressure increases, and the volume reduction are enhanced with increasing shock strength. The ring opening is observed to trigger molecular decay at all four shock conditions; while, the sufficient NO2 fission is observed at Us=8 and 9 km/s, and strongly inhibited at Us=10 and 11 km/s. Moreover, the evolution of main chemical species, such as active intermediates, stable products and clusters, is strongly dependent on the shock strength. NO2 and H are dominant in the primary intermediates, responsible for weak and strong shock, respectively; CO2 and N2, as well as water, are the main stable products with a population gradation determined by the shock strength; and the bigger clusters with longer durations are found to be caused by the stronger shock, and their fast dissociation mainly undergoes through the ring opening. Besides, it is found that ε-CL-20 possesses weak anisotropy in the above-specified Us range. This work will richen the knowledge of shocked energetic materials, in particular, the important energetic materials.
Article
The high temperature oxidative mechanism of a new four-component RP-3 surrogate fuel model was investigated using the ReaxFF MD method. The evolution of the fuel molecules, oxygen, C2H4, and ·CH3, and the underlying reactions, were obtained by systematic analysis of the simulation trajectories with the aid of VARxMD, a unique tool for ReaxFF MD reaction analysis developed by the authors' group. The simulated consumption of fuel and oxygen, as well as the amount of ethylene and methyl radicals, in RP-3 oxidation are of the same magnitude in the ReaxFF MD simulations as that predicted by CHEMKIN under the same temperature and initial pressure conditions. Based on the chemical structures of all the species and the full set of reactions obtained, the detailed mechanisms observed in the simulations broadly agree with the previous literature. The first reactions of the fuel molecules can be categorized into H-abstraction and internal scission, with the latter dominating under various temperature conditions. Observation and statistical analysis of the oxygen reactions reveal that small species of C1-C3 are involved in a relatively large proportion, which may allow the simplification of the reaction mechanism. A reaction network for RP-3 oxidation at high temperature is obtained through the analysis of the reaction mechanisms. This work demonstrates that the ReaxFF MD method, combined with the unique reaction analysis capability of VARxMD, provides useful insights into the mechanism of fuel combustion and should aid the construction of combustion mechanism libraries.
Article
We describe a flexible and broadly applicable energy refinement method, "nebterpolation," for identifying and characterizing the reaction events in a molecular dynamics (MD) simulation. The new method is applicable to ab initio simulations with hundreds of atoms containing complex and multi-molecular reaction events. A key aspect of nebterpolation is smoothing of the reactive MD trajectory in internal coordinates to initiate the search for the reaction path on the potential energy surface. We apply nebterpolation to analyze the reaction events in an ab initio nanoreactor simulation that discovers new molecules and mechanisms, including a C-C coupling pathway for glycolaldehyde synthesis. We find that the new method, which incorporates information from the MD trajectory that connects reactants with products, produces a dramatically distinct set of minimum energy paths compared to existing approaches that start from information for the reaction endpoints alone. The energy refinement method described here represents a key component of an emerging simulation paradigm where molecular dynamics simulations are applied to discover the possible reaction mechanisms.
Article
We synthesized the alanine-based ionic liquid [C2mim][Ala] (1-ethyl-3-methylimidazolium alanine) by using the neutralization method and characterized it. Using a solution-reaction isoperibol calorimeter, we determined the molar enthalpies of the solution (ΔsolHm) at various molalities in water from (288.15 ± 0.01) to (308.15 ± 0.01) K in intervals of 5 K. Using Archer’s method, we obtained the standard molar enthalpy of solution for [C2mim][Ala] (ΔsolθHm) and calculated its apparent relative molar enthalpy (ΦL). Using Glasser’s theory of lattice energy, we obtained the lattice energy, UPOT = 566 kJ mol–1, the hydration enthalpy of the cation and anion, (ΔH+ + ΔH–) = –620 kJ·mol–1, and the hydration enthalpy of an anion, ΔH–([Ala]–) = –387 kJ·mol–1 at 298.15 K. Finally, we obtained the heat capacity of aqueous [C2mim][Ala] (Cp(sol)) and its apparent molar heat capacity (ΦCp) at various specific molalities.
Article
Aromatic hydrocarbon fuels, such as toluene, are important components in real jet fuels. In this work, reactive molecular dynamics (MD) simulations employing the ReaxFF reactive force field have been performed to study the high-temperature oxidation mechanisms of toluene at different temperatures and densities with equivalence ratios ranging from 0.5 to 2.0. From the ReaxFF MD simulations, we have found that the initiation consumption of toluene is mainly through three ways, (1) the hydrogen abstraction reactions by oxygen molecules or other small radicals to form the benzyl radical, (2) the cleavage of the C–H bond to form benzyl and hydrogen radicals, and (3) the cleavage of the C–C bond to form phenyl and methyl radicals. These basic reaction mechanisms are in good agreement with available chemical kinetic models. The temperatures and densities have composite effects on toluene oxidation; concerning the effect of the equivalence ratio, the oxidation reaction rate is found to decrease with the increasing of equivalence ratio. The analysis of the initiation reaction of toluene shows that the hydrogen abstraction reaction dominates the initial reaction stage at low equivalence ratio (0.5–1.0), while the contribution from the pyrolysis reaction increases significantly as the equivalence ratio increases to 2.0. The apparent activation energies, Ea, for combustion of toluene extracted from ReaxFF MD simulations are consistent with experimental results.
Article
In this study, the first GPU-enabled ReaxFF MD program with significantly improved performance, surpassing CPU implementations, was employed to explore the initial chemical mechanisms and product distributions in pyrolysis of Liulin coal, a bituminous coal from Shanxi, PRC. The largest coal model ever used in simulation via ReaxFF MD, the Liulin coal molecular model consisting of 28 351 atoms was constructed based on a combination of experiments and classical coal models. The ReaxFF MD simulations at temperatures of 1000–2600 K were performed for 250 ps to investigate the temperature effects on the product profile and the initial chemical reactions of the Liulin coal model pyrolysis. The generation rates of C14–C40 compounds and gas tend to equilibrate within 150–250 ps, indicating that the simulation should allow most of the thermal decomposition reactions complete and the simulated product profiles are reasonable for understanding the chemical reactions of the Liulin coal pyrolysis. The product (gas, tar, and char) evolution tendencies with time and temperature observed in the simulations are fairly in agreement with the experimental tendency reported in the literature. In particular, the evolution trends of three representative products (naphthalene, methyl-naphthalene and dimethyl-naphthalene) with temperature are very consistent with Py-GC/MS experiments. The detailed chemical reactions of the pyrolysis simulation have been generated using VARMD (Visualization and Analysis of Reactive Molecular Dynamics), which was newly created to examine the complexity of the chemical reaction network in ReaxFF MD simulation. The generation and consumption of HO· and H3C· radicals with time and temperature are reasonable and consistent both with the evolution of H2O and CH4, and with the detailed chemical reactions obtained as well. The amount of six-membered ring structures was observed to decrease with time and temperature, because of their conversion into 5-membered rings or 7–9-membered rings or even-larger-membered ring structures that will further open and decompose into small fragments. This work demonstrates a new methodology for investigating coal pyrolysis mechanism by combining GPU-enabled high-performance computing with cheminformatics analysis in ReaxFF MD.
Article
To make practical the molecular dynamics simulation of large scale reactive chemical systems (1000s of atoms), we developed ReaxFF, a force field for reactive systems. ReaxFF uses a general relationship between bond distance and bond order on one hand and between bond order and bond energy on the other hand that leads to proper dissociation of bonds to separated atoms. Other valence terms present in the force field (angle and torsion) are defined in terms of the same bond orders so that all these terms go to zero smoothly as bonds break. In addition, ReaxFF has Coulomb and Morse (van der Waals) potentials to describe nonbond interactions between all atoms (no exclusions). These nonbond interactions are shielded at short range so that the Coulomb and van der Waals interactions become constant as R-ij --> 0. We report here the ReaxFF for hydrocarbons. The parameters were derived from quantum chemical calculations on bond dissociation and reactions of small molecules plus heat of formation and geometry data for a number of stable hydrocarbon compounds. We find that the ReaxFF provides a good description of these data. Generally, the results are of an accuracy similar or better than PM3, while ReaxFF is about 100 times faster. In turn, the PM3 is about 100 times faster than the QC calculations. Thus, with ReaxFF we hope to be able to study complex reactions in hydrocarbons.
Article
Thermal cracking of n-decane and n-decane in the presence of several fuel additives are studied in order to improve the rate of thermal cracking by using reactive molecular dynamics (MD) simulations employing the ReaxFF reactive force field. From MD simulations, we find the initiation mechanisms of pyrolysis of n-decane are mainly through two pathways: (1) the cleavage of a C-C bond to form smaller hydrocarbon radicals, and (2) the dehydrogenation reaction to form an H radical and the corresponding decyl radical. Another pathway is the H-abstraction reactions by small radicals including H, CH(3), and C(2)H(5). The basic reaction mechanisms are in good agreement with existing chemical kinetic models of thermal decomposition of n-decane. Quantum mechanical calculations of reaction enthalpies demonstrate that the H-abstraction channel is easier compared with the direct C-C or C-H bond-breaking in n-decane. The thermal cracking of n-decane with several additives is further investigated. ReaxFF MD simulations lead to reasonable Arrhenius parameters compared with experimental results based on first-order kinetic analysis. The different chemical structures of the fuel additives greatly affect the apparent activation energy and pre-exponential factors. The presence of diethyl ether (DEE), methyl tert-butyl ether (MTBE), 1-nitropropane (NP), 3,6,9-triethyl-3,6,9-trimethyl-1,2,4,5,7,8-hexaoxonane (TEMPO), triethylamine (TEA), and diacetonediperodixe (DADP) exhibit remarkable promoting effect on the thermal cracking rates, compared with that of pure n-decane, in the following order: NP > TEMPO > DADP > DEE (∼MTBE) > TEA, which coincides with experimental results. These results demonstrate that reactive MD simulations can be used to screen for fuel additives and provide useful information for more comprehensive chemical kinetic model studies at the molecular level.
Article
The value of depth-first search or “backtracking” as a technique for solving problems is illustrated by two examples. An improved version of an algorithm for finding the strongly connected components of a directed graph and at algorithm for finding the biconnected components of an undirect graph are presented. The space and time requirements of both algorithms are bounded by $k_1 V + k_2 E + k_3 $ for some constants $k_1 ,k_2 $, and $k_3 $, where V is the number of vertices and E is the number of edges of the graph being examined.
Article
We present a density-functional-based scheme for determining the necessary parameters of common nonorthogonal tight-binding (TB) models within the framework of the linear-combination-of-atomic-orbitals formalism using the local-density approximation (LDA). By only considering two-center integrals the Hamiltonian and overlap matrix elements are calculated out of suitable input densities and potentials rather than fitted to experimental data. We can derive analytical functions for the C-C, C-H, and H-H Hamiltonian and overlap matrix elements. The usual short-range repulsive potential appearing in most TB models is fitted to self-consistent calculations performed within the LDA. The calculation of forces is easy and allows an application of the method to molecular-dynamics simulations. Despite its extreme simplicity, the method is transferable to complex carbon and hydrocarbon systems. The determination of equilibrium geometries, total energies, and vibrational modes of carbon clusters, hydrocarbon molecules, and solid-state modifications of carbon yield results showing an overall good agreement with more sophisticated methods.
Article
To investigate the failure of the poly(dimethylsiloxane) polymer (PDMS) at high temperatures and pressures and in the presence of various additives, we have expanded the ReaxFF reactive force field to describe carbon-silicon systems. From molecular dynamics (MD) simulations using ReaxFF we find initial thermal decomposition products of PDMS to be CH(3) radical and the associated polymer radical, indicating that decomposition and subsequent cross-linking of the polymer is initiated by Si-C bond cleavage, in agreement with experimental observations. Secondary reactions involving these CH(3) radicals lead primarily to formation of methane. We studied temperature and pressure dependence of PDMS decomposition by following the rate of production of methane in the ReaxFF MD simulations. We tracked the temperature dependency of the methane production to extract Arrhenius parameters for the failure modes of PDMS. Furthermore, we found that at increased pressures the rate of PDMS decomposition drops considerably, leading to the formation of fewer CH(3) radicals and methane molecules. Finally, we studied the influence of various additives on PDMS stability. We found that the addition of water or a SiO(2) slab has no direct effect on the short-term stability of PDMS, but addition of reactive species such as ozone leads to significantly lower PDMS decomposition temperature. The addition of nitrogen monoxide does not significantly alter the degradation temperature but does retard the initial production of methane and C(2) hydrocarbons until the nitrogen monoxide is depleted. These results, and their good agreement with available experimental data, demonstrate that ReaxFF provides a useful computational tool for studying the chemical stability of polymers.
Article
To study the initial chemical events related to the detonation of triacetonetriperoxide (TATP), we have performed a series of molecular dynamics (MD) simulations. In these simulations we used the ReaxFF reactive force field, which we have extended to reproduce the quantum mechanics (QM)-derived relative energies of the reactants, products, intermediates, and transition states related to the TATP unimolecular decomposition. We find excellent agreement between the QM-predicted reaction products and those observed from 100 independent ReaxFF unimolecular MD cookoff simulations. Furthermore, the primary reaction products and average initiation temperature observed in these 100 independent unimolecular cookoff simulations match closely with those observed from a TATP condensed-phase cookoff simulation, indicating that unimolecular decomposition dominates the thermal initiation of the TATP condensed phase. Our simulations demonstrate that thermal initiation of condensed-phase TATP is entropy-driven (rather than enthalpy-driven), since the initial reaction (which mainly leads to the formation of acetone, O(2), and several unstable C(3)H(6)O(2) isomers) is almost energy-neutral. The O(2) generated in the initiation steps is subsequently utilized in exothermic secondary reactions, leading finally to formation of water and a wide range of small hydrocarbons, acids, aldehydes, ketones, ethers, and alcohols.
Article
To investigate the initial chemical events associated with high-temperature gas-phase oxidation of hydrocarbons, we have expanded the ReaxFF reactive force field training set to include additional transition states and chemical reactivity of systems relevant to these reactions and optimized the force field parameters against a quantum mechanics (QM)-based training set. To validate the ReaxFF potential obtained after parameter optimization, we performed a range of NVT-MD simulations on various hydrocarbon/O2 systems. From simulations on methane/O2, o-xylene/O2, propene/O2, and benzene/O2 mixtures, we found that ReaxFF obtains the correct reactivity trend (propene > o-xylene > methane > benzene), following the trend in the C-H bond strength in these hydrocarbons. We also tracked in detail the reactions during a complete oxidation of isolated methane, propene, and o-xylene to a CO/CO2/H2O mixture and found that the pathways predicted by ReaxFF are in agreement with chemical intuition and our QM results. We observed that the predominant initiation reaction for oxidation of methane, propene, and o-xylene under fuel lean conditions involved hydrogen abstraction of the methyl hydrogen by molecular oxygen forming hydroperoxyl and hydrocarbon radical species. While under fuel rich conditions with a mixture of these hydrocarbons, we observed different chemistry compared with the oxidation of isolated hydrocarbons including a change in the type of initiation reactions, which involved both decomposition of the hydrocarbon or attack by other radicals in the system. Since ReaxFF is capable of simulating complicated reaction pathways without any preconditioning, we believe that atomistic modeling with ReaxFF provides a useful method for determining the initial events of oxidation of hydrocarbons under extreme conditions and can enhance existing combustion models.
Article
The Viterbi algorithm (VA) is a recursive optimal solution to the problem of estimating the state sequence of a discrete-time finite-state Markov process observed in memoryless noise. Many problems in areas such as digital communications can be cast in this form. This paper gives a tutorial exposition of the algorithm and of how it is implemented and analyzed. Applications to date are reviewed. Increasing use of the algorithm in a widening variety of areas is foreseen.
Article
This tutorial provides an overview of the basic theory of hidden Markov models (HMMs) as originated by L.E. Baum and T. Petrie (1966) and gives practical details on methods of implementation of the theory along with a description of selected applications of the theory to distinct problems in speech recognition. Results from a number of original sources are combined to provide a single source of acquiring the background required to pursue further this area of research. The author first reviews the theory of discrete Markov chains and shows how the concept of hidden states, where the observation is a probabilistic function of the state, can be used effectively. The theory is illustrated with two simple examples, namely coin-tossing, and the classic balls-in-urns system. Three fundamental problems of HMMs are noted and several practical techniques for solving these problems are given. The various types of HMMs that have been studied, including ergodic as well as left-right models, are described
Article
Three parallel algorithms for classical molecular dynamics are presented. The first assigns each processor a fixed subset of atoms; the second assigns each a fixed subset of inter--atomic forces to compute; the third assigns each a fixed spatial region. The algorithms are suitable for molecular dynamics models which can be difficult to parallelize efficiently --- those with short--range forces where the neighbors of each atom change rapidly. They can be implemented on any distributed--memory parallel machine which allows for message--passing of data between independently executing processors. The algorithms are tested on a standard Lennard--Jones benchmark problem for system sizes ranging from 500 to 100,000,000 atoms on several parallel supercomputers --- the nCUBE 2, Intel iPSC/860 and Paragon, and Cray T3D. Comparing the results to the fastest reported vectorized Cray Y--MP and C90 algorithm shows that the current generation of parallel machines is competitive with conventi...
  • M E Tuckerman
M. E. Tuckerman, J. Phys.: Condens. Matter, 2002, 14, 1297-1355.
  • L.-P Wang
  • R T Mcgibbon
  • V S Pande
  • T J Martinez
L.-P. Wang, R. T. McGibbon, V. S. Pande and T. J. Martinez, J. Chem. Theory Comput., 2016, 12, 638-649.
  • A C Van Duin
  • S Dasgupta
  • F Lorant
  • W A Goddard
A. C. Van Duin, S. Dasgupta, F. Lorant and W. A. Goddard, J. Phys. Chem. A, 2001, 105, 9396-9409.
  • A Strachan
  • E Kober
A. Strachan and E. Kober, J. Chem. Phys., 2005, 122, 054502.
  • G Shchygol
  • A Yakovlev
  • T Trnka
  • A C Van Duin
  • T Verstraelen
G. Shchygol, A. Yakovlev, T. Trnka, A. C. van Duin and T. Verstraelen, J. Chem. Theory Soc., 2005, 127, 7192-7202.
  • M Zheng
  • X X Li
  • J Liu
  • Z Wang
  • X M Gong
  • L Guo
  • W L Song
M. Zheng, X. X. Li, J. Liu, Z. Wang, X. M. Gong, L. Guo and W. L. Song, Energy & Fuels, 2014, 28, 522-534.
  • Y Z Wu
  • H Sun
  • L Wu
  • J D Deetz
Y. Z. Wu, H. Sun, L. Wu and J. D. Deetz, J. Comput. Chem., 2019, 40, 1586-1592.
  • D Zheng
  • W.-M Yu
  • B.-J Zhong
D. Zheng, W.-M. Yu and B.-J. Zhong, Acta Physico-Chimica Sinica, 2015, 31, 636-642.
  • Z He
  • X.-B Li
  • L.-M Liu
  • W Zhu
Z. He, X.-B. Li, L.-M. Liu and W. Zhu, Fuel, 2014, 124, 85-90.
  • X.-L Liu
  • X.-X Li
  • S Han
  • X.-J Qiao
  • B.-J Zhong
  • L Guo
X.-L. Liu, X.-X. Li, S. Han, X.-J. Qiao, B.-J. Zhong and L. Guo, Acta Physico-Chimica Sinica, 2016, 32, 1424-1433.
  • K Chenoweth
  • A C Van Duin
  • W A Goddard
K. Chenoweth, A. C. Van Duin and W. A. Goddard, J. Phys. Chem. A, 2008, 112, 1040-1053.
  • S Russell
  • P Norvig
S. Russell and P. Norvig, Artificial Intelligence A Modern Approach, Pearson Education Asia, Hong Kong, 2011.
  • N M O'boyle
  • M Banck
  • C A James
  • C Morley
  • T Vandermeersch
  • G R Hutchison
N. M. O'Boyle, M. Banck, C. A. James, C. Morley, T. Vandermeersch and G. R. Hutchison, J. Cheminform., 2011, 3, 33.
  • L P Cordella
  • P Foggia
  • C Sansone
  • M Vento
L. P. Cordella, P. Foggia, C. Sansone and M. Vento, ITPAM, 2004, 26, 1367-1372.
  • Q.-D Wang
  • X.-X Hua
  • X.-M Cheng
  • J.-Q Li
  • X.-Y. Li
Q.-D. Wang, X.-X. Hua, X.-M. Cheng, J.-Q. Li and X.-Y. Li, J. Phys. Chem. A, 2012, 116, 3794-3801.
  • X.-M Cheng
  • Q.-D Wang
  • J.-Q Li
  • J.-B Wang
  • X.-Y. Li
X.-M. Cheng, Q.-D. Wang, J.-Q. Li, J.-B. Wang and X.-Y. Li, J. Phys. Chem. A, 2012, 116, 9811-9818.
  • T Lei
  • W Guo
  • Q Liu
  • H Jiao
  • D B Cao
  • B Teng
  • Y W Li
  • X Liu
  • X D Wen
T. Lei, W. Guo, Q. Liu, H. Jiao, D. B. Cao, B. Teng, Y. W. Li, X. Liu and X. D. Wen, J. Chem. Theory Comput., 2019, 15, 3654-3665.
  • N M O'boyle
  • M Banck
  • C A James
  • C Morley
  • T Vandermeersch
  • G R Hutchison
N. M. O'Boyle, M. Banck, C. A. James, C. Morley, T. Vandermeersch and G. R. Hutchison, J. Cheminf., 2011, 3, 33.