Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Predictive Data analytics”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 829 records · Page 46
All sourcesNASA NTRSDOE OSTI
DOE OSTI2021

Machine Learning Self-Diffusion Prediction for Lennard-Jones Fluids in Pores

Predicting the diffusion coefficient of fluids under nanoconfinement is important for many applications including the extraction of shale gas from kerogen and product turnover in porous catalysts. Due to the large number of important variables, including pore shape and size, fluid temperature and density, and the fluid–wall interaction strength, simulating diffusion coefficients using molecular dynamics (MD) in a systematic study could prove to be prohibitively expensive. Here, we use machine learning models trained on a subset of MD data to predict the self-diffusion coefficients of Lennard-Jones fluids in pores. Our MD data set contains 2280 simulations of ideal slit pore, cylindrical pore, and hexagonal pore geometries. We use the forward feature selection method to determine the most useful features (i.e., descriptors) for developing an artificial neutral network (ANN) model with an emphasis on easily acquired features. Our model shows good predictive ability with a coefficient of determination (i.e., R 2 ) of ~0.99 and a mean squared error of ~2.9 × 10 –5 . Finally, we propose an alteration to our feature set that will allow the ANN model to be applied to nonideal pore geometries.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Real-Time Alignment and Reorientation of Polymer Chains in Liquid Crystal Elastomers

Liquid crystal elastomers (LCEs) exhibit soft elasticity due to the alignment and reorientation of mesogens upon mechanical loading, which provides additional mechanisms to absorb and dissipate energy. This enhanced response makes LCEs potentially transformative materials for biomedical devices, tissue replacements, and protective equipment. However, there is a critical knowledge gap in understanding the highly rate-dependent dissipative behaviors of LCEs due to the lack of real-time characterization techniques that probe the microscale network structure and link it to the mechanical deformation of LCEs. In this work, we employ in situ optical measurements to evaluate the alignment and reorientation degree of mesogens in LCEs. The data are correlated to the quantitative physical analysis using polarized Fourier-transform infrared spectroscopy. The time scale of mesogen alignment is determined at different strain levels and loading rates. The mesogen reorientation kinetics is characterized to establish its relationship with the macroscale tensile strain, and compared to theoretical predictions. Overall, this work provides the first detailed study on the time-dependent evolution of mesogen alignment and reorientation in deformed LCEs. It also provides an effective and more accessible approach for other researchers to investigate the structural-property relationships of different types of polymers.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

Superradiance and Directional Exciton Migration in Metal–Organic Frameworks

Crystalline metal–organic frameworks (MOFs) are promising synthetic analogues of photosynthetic light-harvesting complexes (LHCs). The precise assembly of linkers (organic chromophores) around the topology-defined pores offers the evolution of unique photophysical behaviors that are reminiscence of LHCs. These include MOF excited states with photoabsorbed energy that is spatially dispersed over multiple linkers defining the molecular excitons. The multilinker molecular excitons display superradiance–a hallmark of coupled oscillators seen in LHCs–with radiative rate constant (k rad ) exceeding that of a single linker. Our theoretical model and experimental results on three zirconium MOFs, namely, PCN-222(Zn), NU-1000, and SIU-100, with similar topology but varying linkers suggest that the size of such molecular excitons depends on the electronic symmetry of the linker. This multilinker exciton model effectively predicts the energy transfer rate constant; corresponding single-step exciton hopping time, ranging from a few picoseconds in SIU-100 and NU-1000 to a few hundreds of picoseconds in PCN-222(Zn), matches well with the experimental data. The model also predicts the anisotropy of exciton displacement with preferential migration along the crystallographic c-axis. Overall, these findings establish various missing links defining the exciton size and dynamics in MOF-assembled linkers. Furthermore, the understandings will provide design principles, especially, positioning the catalysts or electrode relative to the linker orientation for low-density solar energy conversion systems.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2022

How Generalizable is a Machine-Learning Approach for Modeling Hub-Height Turbulence Intensity?

Hub-height turbulence intensity is essential for a variety of wind energy applications. However, simulating it is a challenging task. Simple analytical models have been proposed in the literature, but they all come with significant limitations. Even state-of-the-art numerical weather prediction models, such as the Weather Research and Forecasting model, currently struggle to predict hub-height turbulence intensity. Here, we propose a machine-learning-based approach to predict hub-height turbulence intensity from other hub-height and ground-level atmospheric measurements, using observations from the Perdigao field campaign and the Southern Great Plains atmospheric observatory. We consider a random forest regression model, which we validate first at the site used for training and then under a more robust round-robin approach, and compare its performance to a multivariate linear regression. The random forest successfully outperforms the linear regression in modeling hub-height turbulence intensity, with a normalized root-mean-square error as low as 0.014 when using 30-minute average data. In order to achieve such low root-mean-square error values, the knowledge of hub-height turbulence kinetic energy (which can instead be modeled in the Weather Research and Forecasting model) is needed. Interestingly, we find that the performance of the random forest generalizes well when considering a round-robin validation (i.e., when the algorithm is trained at one site such as Perdigao or Southern Great Plains) and then applied to model hub-height turbulence intensity at the other location.

17 WIND ENERGY↗
DOE OSTI2022

Deciphering Distinct Overpotential-Dependent Pathways for Electrochemical CO 2 Reduction Catalyzed by an Iron–Terpyridine Complex

[Fe(tpyPY2Me)] 2+ ([Fe] 2+ ) is a homogeneous electrocatalyst for converting CO 2 into CO featuring low overpotentials of <100 mV, near-unity selectivity, and high activity with turnover frequencies faster than 100 000 s -1 . To identify the origins of its exceptional performance and inform future catalyst design, we report a combined computational and experimental study that establishes two distinct mechanistic pathways for electrochemical CO 2 reduction catalyzed by [Fe] 2+ as a function of applied overpotential. Electrochemical data shows the formation of two catalytic regimes at low (η TOF/2 of 160 mV) and high (η TOF/2 of 590 mV) overpotential plateaus. We propose that at low overpotentials [Fe] 2+ undergoes a two-electron reduction, two-proton-transfer mechanism (electrochemical-electrochemical-chemical-chemical, EECC), where turnover occurs through the dicationic iron complex, [Fe] 2+ . Computational analysis supports the importance of the singlet ground-state electronic structure for CO 2 binding and that the rate-limiting step is the second protonation in this low-overpotential regime. When more negative potentials are applied, an additional electron-transfer event occurs through either a stepwise or proton-coupled electron-transfer (PCET) pathway, enabling catalytic turnover from the monocationic iron complex ([Fe] + ) via an electrochemical-chemical-electrochemical-chemical (ECEC) mechanism. Comparison of experimental kinetic data obtained from variable controlled potential electrolysis (CPE) experiments with direct product detection with calculated rates obtained from the energetic span model supports the PCET pathway as the most likely mechanism. Moreover, we build upon this mechanistic understanding to propose the design of an improved ligand framework that is predicted to stabilize the key transition states identified in our study and explore their electronic structures using an energy decomposition analysis. Finally, taken together, this work highlights the value of synergistic computational/experimental approaches to decipher mechanisms of new electrocatalysts and direct the rational design of improved platforms.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Improving Data and Prediction Quality of High-Throughput Perovskite Synthesis with Model Fusion

Combinatorial fusion analysis (CFA) is an approach for combining multiple scoring systems using the rank-score characteristic function and cognitive diversity measure. One example is to combine diverse machine learning models to achieve better prediction quality. In this work, we apply CFA to the synthesis of metal halide perovskites containing organic ammonium cations via inverse temperature crystallization. Using a data set generated by high-throughput experimentation, four individual models (support vector machines, random forests, weighted logistic classifier, and gradient boosted trees) were developed. We characterize each of these scoring systems and explore 66 possible combinations of the models. When measured by the precision on predicting crystal formation, the majority of the combination models improves the individual model results. The best combination models outperform the best individual models by 3.9 percentage points in precision. In addition to improving prediction quality, we demonstrate how the fusion models can be used to identify mislabeled input data and address issues of data quality. In particular, we identify example cases where all single models and all fusion models do not give the correct prediction. Experimental replication of these syntheses reveals that these compositions are sensitive to modest temperature variations across the different locations of the heating element that can hinder or enhance the crystallization process. In summary, we demonstrate that model fusion using CFA can not only identify a previously unconsidered influence on reaction outcome but also be used as a form of quality control for high-throughput experimentation.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Data-driven discovery of dynamics from time-resolved coherent scattering

Coherent X-ray scattering (CXS) techniques are capable of interrogating dynamics of nano- to mesoscale materials systems at time scales spanning several orders of magnitude. However, obtaining accurate theoretical descriptions of complex dynamics is often limited by one or more factors—the ability to visualize dynamics in real space, computational cost of high-fidelity simulations, and effectiveness of approximate or phenomenological models. In this work, we develop a data-driven framework to uncover mechanistic models of dynamics directly from time-resolved CXS measurements without solving the phase reconstruction problem for the entire time series of diffraction patterns. Our approach uses neural differential equations to parameterize unknown real-space dynamics and implements a computational scattering forward model to relate real-space predictions to reciprocal-space observations. This method is shown to recover the dynamics of several computational model systems under various simulated conditions of measurement resolution and noise. Moreover, the trained model enables estimation of long-term dynamics well beyond the maximum observation time, which can be used to inform and refine experimental parameters in practice. Finally, we demonstrate an experimental proof-of-concept by applying our framework to recover the probe trajectory from a ptychographic scan. Our proposed framework bridges the wide existing gap between approximate models and complex data.

36 MATERIALS SCIENCE↗
DOE OSTI2020

Adaptive model tuning studies for non-invasive diagnostics and feedback control of plasma wakefield acceleration at FACET-II

Beam-driven plasma wakefield acceleration (PWFA) achieves the same energy gain in a single meter, for which conventional accelerators require several kilometers, however much work is still required to match the beam quality of conventional accelerators. The PWFA processes to be studied at the FACET-II facility will utilize extremely short (down to a few fs) high energy (10 GeV), high charge (few nC) electron and eventually positron bunches. The PWFA process is extremely sensitive to the detailed longitudinal current profiles of these bunches and it would be of great benefit to have precise measurement and control of these profiles. We present an adaptive model tuning technique for FACET-II to adaptively tune online models based on real time accelerator and beam data to continuously provide a non-invasive diagnostic which predicts the longitudinal phase space (LPS) of extremely short and highly compressed electron beams, which otherwise require destructive X-band transverse deflecting cavity (XTCAV)-based measurements. Based on simulation studies, our method has the potential for: (1). The development of a non-invasive longitudinal phase space diagnostic by adaptively tuning models based on non-invasive measurements such as energy spread spectra which could be recorded at all of the bunch compressors in the facility. (2). Utilize these diagnostic to perform model-independent feedback-based to achieve desired longitudinal phase space distributions. (3). Utilize the model-independent feedback approach to maximize energy gain while minimizing emittance growth and energy gain variance of the PWFA process by tuning accelerator parameters, while monitoring what longitudinal phase space the algorithm has found, information that will be useful for further analytical and simulation studies.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2020

Evaluation of a New Chemical Mechanism for 2-Amino-2-methyl-1-propanol in a Reactive Environment from CSIRO Smog Chamber Experiments

Amines are considered as an emerging class of atmospheric pollutants that are of great importance to atmospheric chemistry and new particle formation. As a typical amine, 2-amino-2-methyl-1-propanol (AMP) is one of the proposed solvents for capturing CO 2 from flue gas streams in amine-based post-combustion CO 2 capture plants, and it is expected to result in AMP emission and secondary products formation in the atmosphere. However, the current knowledge of its atmospheric chemistry and kinetics is poorly understood, particularly in a reactive environment. As such, in this work we used the CSIRO smog chamber to study the photo-oxidation of AMP in the presence of VOCs–NOx surrogate mixtures over a range of initial amine concentrations. O 3 formation was significantly inhibited when AMP was added to the surrogate VOCs–NOx mixtures, implying that AMP could alter known atmospheric chemical reaction pathways and the prevailing reactivity. Simultaneously, a large amount of AMP-derived secondary aerosol was formed, with a considerably high aerosol mass yield (i.e. ratio of aerosol formed to reacted AMP) of 1.06±0.20. Based on updated knowledge of its kinetics, oxidation pathways and product yields, we have developed a new mechanism (designated as CSIAMP-19) and integrated into the Carbon Bond 6 (CB6) chemical mechanism, and evaluated it against available smog chamber data. Compared with the existing AMP mechanism (designated as CarterAMP-08), the modified CB6 with CSIAMP-19 mechanism improves prediction against AMP–VOCs–NOx experiments across a range of initial AMP concentrations, within ±10% model error for gross ozone production. Our results contribute to scientific understanding of AMP photochemistry and to the development of the chemical mechanism of other amines. The updated AMP chemical reactions scheme can be further embedded into the chemical transport model for regional modelling scenarios where AMP-related emissions are of concern.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2020

A cyclic electrochemical strategy to produce acetylene from CO 2 , CH 4 , or alternative carbon sources

This paper presents the first combined measurement of the double-differential muon neutrino and antineutrino charged-current cross sections with no pions in the final state on hydrocarbon at the off-axis near detector of the T2K experiment. The data analyzed in this work comprise 5.8 × 10$^{20}$ and 6.3 × 10$^{20}$ protons on target in neutrino and antineutrino mode respectively, at a beam energy peak of 0.6 GeV. Using the two measured cross sections, the sum, difference, and asymmetry were calculated with the aim of better understanding the nuclear effects involved in such interactions. The extracted measurements have been compared with the prediction from different Monte Carlo generators and theoretical models showing that the difference between the two cross sections have interesting sensitivity to nuclear effects.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Turbulence theories and statistical closure approaches

When discussing research in physics and in science more generally, it is common to ascribe equal importance to the three components of the scientific trinity: theoretical, experimental, and computational studies. This review will explore the future of modern turbulence theory by tracing its history, which began in earnest with Kolmogorov’s 1941 analysis of turbulence cascade and inertial range [A.N. Kolmogorov, Dokl. Akad. Nauk SSSR, 30, 299, (1941); 32, 19, (1941)]. The 80th Anniversary of Kolmogorov’s landmark study is a welcome opportunity to survey the achievements and evaluate the future of the theoretical approach of turbulence research. Over the years, turbulence theories have been critically important in laying the foundation of our understanding of the nature of turbulent flows. In particular, the Direct Interaction Approximation (DIA) [R.H. Kraichnan, J. Fluid Mech., 5, 497 (1959)] and its subsequent development, known as the statistical closure approach, can be identified as perhaps the most profound single advancement. The remarkable success of the statistical closure has furnished a platform to study such essential concepts as the energy transfer process and interacting scales, and the roles of the straining and sweeping motions. More recently, the quasi-Lagrangian formulation of V. L’vov & I. Procaccia and Kraichnan’s solvable passive scalar model provided powerful ways to explore another fundamental aspect of turbulent flows, the phenomena of intermittency, and the associated anomalous scaling exponents. In the meantime, the theory of fluid equilibria has been developed to describe the large-scale structures that can emerge from turbulent cascades of two-dimensional and geophysical flows at a later time. And yet, despite all these successes, analytical treatments suffer from mathematical complexities. As a result, the utility of theoretical approaches has been limited to relatively idealized flows. On the other hand, in recent decades, computational abilities and experimental facilities have reached an unprecedented scale. Looking beyond the horizon, the imminent deployment of exascale supercomputers will generate complete datasets of the entire flow field of key benchmark flows, allowing researchers to extract additional measurements concerning fully developed, complex turbulent flow fields far beyond those available from the statistical closure theories. Some other developments that could potentially influence the future course of turbulence theories include the advancement of machine learning, artificial intelligence, and data science; likely disruptions arising from the advent of quantum computation; and the increasingly prominent role of turbulence research in providing more accurate climate scientific data. Finally, turbulence theorists can leverage these developments by asking the right questions and developing advanced, sophisticated frameworks that will be able to predict and correlate vast amounts of data from the other two components of the trinity.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2020

CoSim: A Simulator for Co-Scheduling of Batch and On-Demand Jobs in HPC Datacenters

The increasing scale and complexity of scientific applications are rapidly transforming the ecosystem of tools, methods, and workflows adopted by the high-performance computing (HPC) community. Big data analytics and deep learning are gaining traction as essential components in this ecosystem in a variety of scenarios, such as, steering of experimental instruments , acceleration of high-fidelity simulations through surrogate computations, and guided ensemble searches. In this context, the batch job model traditionally adopted by the supercomputing in-frastructures needs to be complemented with support to schedule opportunistic on-demand analytics jobs, leading to the problem of efficient preemption of batch jobs with minimum loss of progress. In this paper, we design and implement a simulator, CoSim, that enables on-the-fly analysis of the trade-offs arising between delaying the start of opportunistic on-demand jobs, which leads to longer analytics latency, and loss of progress due to preemption of batch jobs, which is necessary to make room for on-demand jobs. To this end, we propose an algorithm based on dynamic programming with predictable performance and scalability that enables supercomputing infrastructure schedulers to analyze the aforementioned trade-off and take decisions in near real-time. Compared with other state-of-art approaches using traces of the Theta pre-Exascale machine, our approach is capable of finding the optimal solution, while achieving high performance and scalability.

batch job preemption↗
DOE OSTI2020

Improving Enzyme Optimum Temperature Prediction with Resampling Strategies and Ensemble Learning

Accurate prediction of the optimal catalytic temperature ( T opt ) of enzymes is vital in biotechnology, as enzymes with high T opt values are desired for enhanced reaction rates. Recently, a machine learning method (temperature optima for microorganisms and enzymes, TOME) for predicting T opt was developed. TOME was trained on a normally distributed data set with a median T opt of 37 °C and less than 5% of T opt values above 85 °C, limiting the method’s predictive capabilities for thermostable enzymes. Due to the distribution of the training data, the mean squared error on T opt values greater than 85 °C is nearly an order of magnitude higher than the error on values between 30 and 50 °C. Here, we apply ensemble learning and resampling strategies that tackle the data imbalance to significantly decrease the error on high T opt values (>85 °C) by 60% and increase the overall R 2 value from 0.527 to 0.632. The revised method, temperature optima for enzymes with resampling (TOMER), and the resampling strategies applied in this work are freely available to other researchers as Python packages on GitHub.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Catalytic Reduction of Esters over Zirconia-Supported Metal Catalysts

Esters are often produced as unwanted byproducts during the catalytic upgrading of ethanol to diesel fuel precursors through Guerbet coupling. Removal of esters from the product stream is important to prevent the loss of downstream catalyst activity from ester-derived carboxylic acids. In this work, we studied ester hydrogenolysis to the parent alcohols as a viable route for enhanced diesel fuel production. Specifically, we investigated the reduction of hexyl acetate in butanol over ZrO 2 -supported Ni, Co, Cu, Rh, Pd, and Pt catalysts, where Cu/ZrO 2 was the most selective catalyst for the hydrogenolysis of hexyl acetate into hexanol and ethanol. Thermodynamic analysis reveals that a 90% alcohol yield can be obtained at 200 °C, 30 bar, and a relatively high H 2 :hexyl acetate molar ratio of 480:1. Experimentally, an alcohol yield of 88% yield was obtained with a 10 wt % Cu/ZrO 2 catalyst at these conditions with a residence time of 5.4 h kg cat kmol gas –1 . Catalytic tests on the support revealed that ZrO 2 catalyzes the transesterification reaction between hexyl acetate and butanol. However, only the Cu sites can catalyze the hydrogenolysis of the esters into the final alcohols. We developed a kinetic model for our experimental results, which shows that the transesterification and hydrogenolysis reactions run at two different timescales, the former being 10 times faster than the latter. Data regression has been used to develop a model to predict the mole fraction distribution of ester hydrogenolysis products over a wide range of contact times. Cu/ZrO 2 loses half its catalytic activity after 80 h of time on stream. Modeling of deactivation data reveals that the ZrO 2 support conserves a residual activity due to external active sites, while active sites over the Cu surface deactivate at different rates. Furthermore, the catalytic conversion of esters into their parent alcohols is relevant to the production of surrogate liquid fuels since alcohols can be bimolecularly dehydrated to produce a blend of ethers with diesel fuel-like properties.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2024

Exploring CO 2 reduction and crossover in membrane electrode assemblies

Electrochemical CO 2 reduction (CO 2 R) using renewable electricity is a key pathway toward synthesizing fuels and chemicals. In this study, multi-physics modeling is used to interpret experimental data obtained for CO 2 R to CO using Ag catalysts in a membrane electrode assembly. The one-dimensional model is validated using measured CO 2 crossover and product formation rates. The kinetics of CO formation are described by Marcus–Hush–Chidsey kinetics, which enables accurate prediction of the experimental data by accounting for the reorganization of the solvent during CO 2 R. Further, the results show how the performance is dictated by competing phenomena including ion formation and transport, CO 2 solubility, and water management. The model shows that increasing the ion-exchange capacity of the membrane and surface area of the catalyst increases CO formation rates by >100 mA cm –2 without negatively impacting CO 2 utilization. Here we provide insights into how to manage the trade-off between productivity and CO 2 utilization in CO 2 electrolyzers.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Mock light-cones and theory friendly catalogues for the CANDELS survey

Here we present mock catalogues created to support the interpretation of the Cosmic Assembly Near-IR Deep Extragalactic Legacy Survey (CANDELS). We extract haloes along past light-cones from the Bolshoi Planck dissipationless N-body simulations and populate these haloes with galaxies using two different independently developed semi-analytic models of galaxy formation and the empirical model UNIVERSEMACHINE. Our mock catalogues have geometries that encompass the footprints of observations associated with the five CANDELS fields. In order to allow field-to-field variance to be explored, we have created eight realizations of each field. In this paper, we present comparisons with observable global galaxy properties, including counts in observed frame bands, luminosity functions, colour–magnitude distributions and colour–colour distributions. We additionally present comparisons with physical galaxy parameters derived from SED fitting for the CANDELS observations, such as stellar masses and star formation rates. We find relatively good agreement between the model predictions and CANDELS observations for luminosity and stellar mass functions. We find poorer agreement for colours and star formation rate distributions. All of the mock light-cones as well as curated ‘theory friendly’ versions of the observational CANDELS catalogues are made available through a web-based data hub.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2024

Optimizing Direct Air Capture Solvents to Minimize Energy Consumption of CO 2 Release in a Carbonate Electrolyzer

Addressing climate change by carbon management is critical to achieving the goal of net zero carbon emissions by 2050. In this work, we examined the electrochemically-driven recovery of CO 2 during alkaline solvent regeneration for solvent-based direct air capture. A mathematical model was developed by incorporating carbonate chemistry with water electrolysis to predict the energy consumption per unit of CO 2 released. The predicted results were consistent with the experimental data, in which the experimental work was achieved by characterizing alkalinity and carbon loading values of solvent collected from a flow carbonate electrolyzer. Through this study, we learned that minimizing the energy expended on CO 2 release can be achieved by using an anolyte with a lower alkalinity, increasing the electric charge input to the electrolyzer, and reducing the ohmic resistance of the electrolyzer. Furthermore, using a supporting electrolyte, e.g., Na 2 SO 4 in the present work, effectively compensates for the higher ohmic resistance from using an anolyte with a lower alkalinity.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Importance of Engineered and Learned Molecular Representations in Predicting Organic Reactivity, Selectivity, and Chemical Properties

Machine-readable chemical structure representations are foundational in all attempts to harness machine learning for the prediction of reactivities, selectivities, and chemical properties directly from molecular structure. The featurization of discrete chemical structures into a continuous vector space is a critical phase undertaken before model selection, and the development of new ways to quantitatively encode molecules is an active area of research. Here, we highlight the application and suitability of different representations, from expert-guided “engineered” descriptors to automatically “learned” features, in different prediction tasks relevant to organic and organometallic chemistry, where differing amounts of training data are available. These tasks include statistical models of stereo- and enantioselectivity, thermochemistry, and kinetics developed using experimental and quantum chemical data. The use of expert-guided molecular descriptors provides an opportunity to incorporate chemical knowledge, domain expertise, and physical constraints into statistical modeling. In applications to stereoselective organic and organometallic catalysis, where data sets may be relatively small and 3D-geometries and conformations play an important role, mechanistically informed features can be used successfully to obtain predictive statistical models that are also chemically interpretable. We provide an overview of several recent applications of this approach to obtain quantitative models for reactivity and selectivity, where topological descriptors, quantum mechanical calculations of electronic and steric properties, along with conformational ensembles, all feature as essential ingredients of the molecular representations used. Alternatively, more flexible, general-purpose molecular representations such as attributed molecular graphs can be used with machine learning approaches to learn the complex relationship between a structure and prediction target. This approach has the potential to out-perform more traditional representation methods such as “hand-crafted” molecular descriptors, particularly as data set sizes grow. One area where this is particularly relevant is in the use of large sets of quantum mechanical data to train quantitative structure–property relationships. A general approach toward curating useful data sets and training highly accurate graph neural network models is discussed in the context of organic bond dissociation enthalpies, where this strategy outperforms regression using precomputed descriptors. Finally, we describe how graph neural network predictions can be incorporated into mechanistically informed statistical models of chemical reactivity and selectivity. Once trained, this approach avoids the expensive computational overhead associated with quantum mechanical calculations, while maintaining chemical interpretability. We illustrate examples for which fast predictions of bond dissociation enthalpy and of the identities of radicals formed through cleavage of a molecule’s weakest bond are used in simple physical models of site-selectivity and reactivity.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗