Data Science in Chemical Engineering: Applications to Molecular Science.

Chemical engineering is being rapidly transformed by the tools of data science. On the horizon, artificial intelligence (AI) applications will impact a huge swath of our work, ranging from the discovery and design of new molecules to operations and manufacturing and many areas in between. Early adoption of data science, machine learning, and early examples of AI in chemical engineering has been rich with examples of molecular data science-the application tools for molecular discovery and property optimization at the atomic scale. We summarize key advances in this nascent subfield while introducing molecular data science for a broad chemical engineering readership. We introduce the field through the concept of a molecular data science life cycle and discuss relevant aspects of five distinct phases of this process: creation of curated data sets, molecular representations, data-driven property prediction, generation of new molecules, and feasibility and synthesizability considerations.

Continuous Molecular Representations of Ionic Liquids.

Designing new ionic liquids (ILs) is of crucial importance for various industrial applications. However, this always leads to a daunting challenge, as the number of possible combinations of cation and anion are very high and it is impossible to experimentally propose and screen a wide pool of potential candidates. However, recent applications of machine learning (ML) models have greatly improved the overall chemical discovery pipeline. In this study, we compare different generative methods for producing ionic liquids. In this comparison, we show the following: (1) when training data is scarce, a transfer learning approach can be applied to variational autoencoders (VAEs) to generate molecular structures of the target molecule type; (2) in a VAE-like structure, separate latent spaces for the cationic and anionic moieties can result in meaningful representations for their combinative, macroscopic properties; (3) interpolating between ILs with desired properties can result in a new IL with attributes similar to the two structural end points.

Unveiling Carbon Ring Structure Formation Mechanisms in Polyacrylonitrile-Derived Carbon Fibers.

As the demand for electric vehicles (EVs) and autonomous vehicles (AVs) rapidly grows, lower-cost, lighter, and stronger carbon fibers (CFs) are urgently needed to respond to consumers' call for greater EV traveling range and stronger safety structures for AVs. Converting polymeric precursors to CFs requires a complex set of thermochemical processes; a systematic understanding of each parameter in fiber conversion is still, to a large extent, lacking. Here, we demonstrate the effect of carbonization temperature on carbon ring structure formation by combining atomistic/microscale simulations and experimental validation. Experimental testing, as predicted by simulations, exhibited that the strength and ductility of PAN CFs decreased, whereas the Young's modulus increased with increasing carbonization temperature. Our simulations unveiled that high carbonization temperature accelerated the kinetics of graphitic phase nucleation and growth, leading to the decrease in strength and ductility but increase in modulus. The methodology presented herein using combined atomistic/microscale simulations and experimental validation lays a firm foundation for further innovation in CF manufacturing and low-cost alternative precursor development.

Atomistic Scale Analysis of the Carbonization Process for C/H/O/N-Based Polymers with the ReaxFF Reactive Force Field.

During the carbonization process of raw polymer precursors, graphitic structures can evolve. The presence of these graphitic structures affects mechanical properties of the carbonized carbon fibers. To gain a better understanding of the chemistry behind the evolution of these structures, we performed atomistic-scale simulations using the ReaxFF reactive force field. Three different polymers were considered as a precursor: idealized ladder PAN (polyacrylonitrile), a proposed oxidized PAN, and poly( p-phenylene-2,6-benzobisoxazole). We determined the underlying molecular details of polymer conversion into a carbon fiber structure. Because these are C/H/O/N-based polymers, we first developed an improved force field for C/H/O/N chemistry based on the density functional theory data with a particular focus on N2 formation kinetics and its interactions with polymer-associated radicals formed during the carbonization process. Then, using this improved force field, we performed atomistic-scale simulations of the initial stage of the carbonization process for the considered polymers. On the basis of our simulation data, the molecular pathways for the formation of low-molecular-weight gas species and all-carbon ring formation were determined. We also examined the possible alignment of the developed all-carbon 6-membered ring clusters, which is crucial for the further graphitic structure evolution.

Accelerated ReaxFF Simulations for Describing the Reactive Cross-Linking of Polymers.

Various methods have been developed to perform atomistic-scale simulations for the cross-linking of polymers. Most of these methods involve connecting the reactive sites of the monomers, but these typically do not capture the entire reaction process from the reactants to final products through transition states. Experimental time scales for cross-linking reactions in polymers range from minutes to hours, which are time scales that are inaccessible to atomistic-scale simulations. Because simulating reactions on realistic time scales is computationally expensive, in this investigation, an accelerated simulation method was developed within the ReaxFF reactive force field framework. In this method, the reactants are tracked until they reach a nonreactive configuration that provides a good starting point for a reactive event. Subsequently, the reactants are provided with a sufficient amount of energy-equivalent or slightly larger than their lowest-energy reaction barrier-to overcome the barrier for the cross-linking process and form desired products. This allows simulation of cross-linking at realistic, low temperatures, which helps to mimic chemical reactions and avoids unwanted high-temperature side reactions and still allows us to reject high-barrier events. It should be noted that not all accelerated events are successful as high local strain can lead to reaction rejections. The validity of the ReaxFF force field was tested for three different types of transition state, possibly for polymerization of epoxides, and good agreement with quantum mechanical methods was observed. The accelerated method was further implemented to study the cross-linking of diglycidyl ether of bisphenol F (bis F) and diethyltoluenediamine (DETDA), and a reasonably high percentage (82%) of cross-linking was obtained. The simulated cross-linked polymer was then tested for density, glass transition temperature, and modulus and found to be in good agreement with experiments. Results indicate that this newly developed accelerated simulation method in ReaxFF can be a useful tool to perform atomistic-scale simulations on polymerization processes that have a relatively high reaction barrier at a realistic, low temperature.

Development of a Charge-Implicit ReaxFF Potential for Hydrocarbon Systems.

Molecular dynamics (MD) simulations continue to make important contributions to understanding chemical and physical processes. Concomitant with the growth of MD simulations is the need to have interaction potentials that both represent the chemistry of the system and are computationally efficient. We propose a modification to the ReaxFF potential for carbon and hydrogen that eliminates the time-consuming charge equilibration, eliminates the acknowledged flaws of the electronegativity equalization method, includes an expanded training set for condensed phases, has a repulsive wall for simulations of energetic particle bombardment, and is compatible with the LAMMPS code. This charge-implicit ReaxFF potential is five times faster than the conventional ReaxFF potential for a simulation of keV particle bombardment with a sample size of over 800 000 atoms.

Modified Random Sequential Adsorption Model for Understanding Kinetics of Proteins Adsorption at a Liquid-Solid Interface.

In this Article, we experimentally measure the adsorption kinetics of human serum albumin (HSA) on a hydrophobic hexadecanethiolated gold surface. We use micromachined quartz crystal resonators with fundamental frequency of 83 MHz to accomplish these measurements in real time. In this work, we focus on two key results: (i) asymptotic behavior of the sensor responses upon HSA adsorption and (ii) the jamming limit of adsorbed layer formed by both single-injection and multi-injection experiments with the same value of final concentration. We develop a new interface-depletion modified random sequential adsorption (RSA) model to elucidate the adsorption kinetics and the transport properties of the protein molecules. Analysis of the experimentally measured data shows that the results can be explained on the basis of the exponentially depleting interfacial layer RSA model. To better understand the origin of the formation of the interfacial depletion region where the supply of protein molecules is dramatically reduced, we performed a series of molecular dynamics (MD) simulations using the ReaxFF method. These simulations predict that the resulting adsorption of the protein molecules on the thiolated surface results in a specific orientation at the interface and the diffusion constant of the protein molecules in this layer is significantly reduced. This interplay between the surface adsorption rate and the reduced diffusion coefficient leads to the depletion of the protein molecules in the interfacial layer where the concentration of the protein molecules is much less than the bulk concentration and explains the observed slowdown of the HSA adsorption characteristics on a hydrophobic surface.

ReaxFF based molecular dynamics simulations of ignition front propagation in hydrocarbon/oxygen mixtures under high temperature and pressure conditions.

In this paper, we present the first atomistic-scale based method for calculating ignition front propagation speed and hypothesize that this quantity is related to laminar flame speed. This method is based on atomistic-level molecular dynamics (MD) simulations with the ReaxFF reactive force field. Results reported in this study are for supercritical (P = 55 MPa and Tu = 1800 K) combustion of hydrocarbons as elevated pressure and temperature are required to accelerate the dynamics for reactive MD simulations. These simulations are performed for different types of hydrocarbons, including alkyne, alkane, and aromatic, and are able to successfully reproduce the experimental trend of reactivity of these hydrocarbons. Moreover, our results indicate that the ignition front propagation speed under supercritical conditions has a strong dependence on equivalence ratio, similar to experimentally measured flame speeds at lower temperatures and pressures which supports our hypothesis that ignition front speed is a related quantity to laminar flame speed. In addition, comparisons between results obtained from ReaxFF simulation and continuum simulations performed under similar conditions show good qualitative, and reasonable quantitative agreement. This demonstrates that ReaxFF based MD-simulations are a promising tool to study flame speed/ignition front speed in supercritical hydrocarbon combustion.

Extension of the ReaxFF Combustion Force Field toward Syngas Combustion and Initial Oxidation Kinetics.

A detailed insight of key reactive events related to oxidation and pyrolysis of hydrocarbon fuels further enhances our understanding of combustion chemistry. Though comprehensive kinetic models are available for smaller hydrocarbons (typically C3 or lower), developing and validating reaction mechanisms for larger hydrocarbons is a daunting task, due to the complexity of their reaction networks. The ReaxFF method provides an attractive computational method to obtain reaction kinetics for complex fuel and fuel mixtures, providing an accuracy approaching ab-initio-based methods but with a significantly lower computational expense. The development of the first ReaxFF combustion force field by Chenoweth et al. (CHO-2008 parameter set) in 2008 has opened new avenues for researchers to investigate combustion chemistry from the atomistic level. In this article, we seek to address two issues with the CHO-2008 ReaxFF description. While the CHO-2008 description has achieved significant popularity for studying large hydrocarbon combustion, it fails to accurately describe the chemistry of small hydrocarbon oxidation, especially conversion of CO2 from CO, which is highly relevant to syngas combustion. Additionally, the CHO-2008 description was obtained faster than expected H abstraction by O2 from hydrocarbons, thus underestimating the oxidation initiation temperature. In this study, we seek to systemically improve the CHO-2008 description and validate it for these cases. Additionally, our aim was to retain the accuracy of the 2008 description for larger hydrocarbons and provide similar quality results. Thus, we expanded the ReaxFF CHO-2008 DFT-based training set by including reactions and transition state structures relevant to the syngas and oxidation initiation pathways and retrained the parameters. To validate the quality of our force field, we performed high-temperature NVT-MD simulations to study oxidation and pyrolysis of four different hydrocarbon fuels, namely, syngas, methane, JP-10, and n-butylbenzene. Results obtained from syngas and methane oxidation simulation indicated that our redeveloped parameters (named as the CHO-2016 parameter set) has significantly improved the C1 chemistry predicted by ReaxFF and has solved the low-temperature oxidation initiation problem. Moreover, Arrhenius parameters of JP-10 decomposition and initiation mechanism pathways of n-butylbenzene pyrolysis obtained using the CHO-2016 parameter set are also in good agreement with both experimental and CHO-2008 simulation results. This demonstrated the transferability of the CHO-2016 description for a wide range of hydrocarbon chemistry.