Integrating reaction enumeration and machine learning for microkinetic modeling of bond-exchange reactions

Fast and accurate prospective predictions of regioselectivity can significantly reduce the time and resources spent on unproductive transformations in the pharmaceutical industry. Density functional theory (DFT) reaction modeling through transition state theory (TST) and machine learning (ML) methods has been widely used to predict reaction outcomes such as selectivity. However, TST reaction modeling and ML methods are either time-consuming or data-dependent. Herein, we introduce a prototype seamlessly bridging ML and TST modeling by triggering resource-intensive but much less domain-sensitive DFT calculations only on less confident ML predictions. The proposed workflow was trained and tested on both the Pfizer internal dataset and the USPTO public dataset to predict regioselectivity for SNAr reactions. Our method is accurate and fast, which achieves 96.3 and 94.7% accuracy in predicting the correct major product on Pfizer and USPTO datasets, respectively, in a fraction of conventional TST computing time.

CO oxidation is a very useful reference reaction in catalysis by nanoparticles (NPs). Two reaction models—an association mechanism (AM) and a carbonate-mediated mechanism (CMM)—have been suggested for CO oxidation catalyzed by small NPs. It is still unclear, however, when and why these mechanisms preferentially operate. With the Ag13 crystalline NPs and the Ag12Pd1 core−shell NP, using a density functional theory calculation and a microkinetic reaction model, we found that the different reaction mechanisms can operate with different reaction intermediates accompanied by a balance in the adsorption energy of reactants. Variations in the adsorption energy of adsorbates can result from the interchange of two reaction mechanisms, even in a single NP. An AM operates when both reactants interact strongly with a NP, whereas the contribution of the CMM in CO oxidation increases when a CO molecule interacts weakly or not at all with a NP.

Accurate modeling of catalytic reactions on undercoordinated sites requires accounting for the structural and ensemble-specific nature of the active sites. This study examines how common microkinetic modeling (MKM) assumptions affect predicted kinetics and mechanisms on the stepped Pt(211) facet for the ethane dehydrogenation (EDH) and the ethane hydrogenolysis (EH). Six (211) MKMs were developed, differing in (i) the number of active sites represented, (ii) adsorbate site occupancy treatment, and (iii) inclusion of cross-facet interactions. These models are benchmarked against a particle-based microkinetic model (PB-MKM), which best represents step-edge behavior. MKM assumptions caused deviations in turnover frequencies exceeding ten orders of magnitude and led to contrasting mechanistic and selectivity predictions. Multi-site MKMs overestimate activity by inflating free site availability, single-site models underestimate activity, and uniform occupancy models overpredict coverage of multi-dentate intermediates, leading to reaction-specific artifacts. Overall, the Combined Site Edge Model (CSEM), a single-site MKM accounting for site occupancy and cross-facet interactions, most closely approximates PB-MKM predictions. All models predict similar kinetics when surfaces are clean or primarily occupied by monodentate species. This work provides practical guidance for selecting MKM frameworks for undercoordinated catalytic surfaces and highlights the critical role of modeling assumptions in catalytic predictions.

The first principles modeling of electrochemical reactions has proven useful for the development of efficient, durable and low cost solid oxide full cells (SOFCs). In this account we focus on recent advances in modeling of structural, electronic and catalytic properties of the SOFC anodes based on density functional theory (DFT) first principle calculations. As a starting point, we highlight that the adequate analysis of cell electrochemistry generally requires modeling of chemical reactions at the metal/oxide interface rather than on individual metal or oxide surfaces. The atomic models of Ni/YSZ and Ni/CeO2 interfaces, required for DFT simulations of reactions on SOFC anodes are discussed next, together with the analysis of the electronic structure of these interfaces. Then we proceed to DFT-based findings on charge transfer mechanisms during redox reactions on these two anodes. We provide a comparison of the electronic properties of Ni/YSZ and Ni/CeO2 interfaces and present an interpretation of their different chemical performances. Subsequently we discuss the computed energy pathways of fuel oxidation mechanisms, obtained by various groups to date. We also discuss the results of DFT studies combined with microkinetic modeling as well as the results of kinetic Monte Carlo simulations. In conclusion we summarize the key findings of DFT modeling of metal/oxide interfaces to date and highlight possible directions in the future modeling of SOFC anodes.

A review of multiscale modeling of metal-catalyzed reactions: Mechanism development for complexity and emergent behavior

Molecular-level descriptions of surface chemistry in kinetic models using density functional theory

Work continued on the development of a microkinetic model of Fischer-Tropsch synthesis (FTS) on supported and unsupported Fe catalysts. The following aspects of the FT mechanism on unsupported iron catalysts were investigated on during this third year: (1) the collection of rate data in a Berty CSTR reactor based on sequential design of experiments; (2) CO adsorption and CO-TPD for obtaining the heat of adsorption of CO on polycrystalline iron; and (3) isothermal hydrogenation (IH) after Fischer Tropsch reaction to identify and quantify surface carbonaceous species. Rates of C{sub 2+} formation on unsupported iron catalysts at 220 C and 20 atm correlated well to a Langmuir-Hinshelwood type expression, derived assuming carbon hydrogenation to CH and OH recombination to water to be rate-determining steps. From desorption of molecularly adsorbed CO at different temperatures the heat of adsorption of CO on polycrystalline iron was determined to be 100 kJ/mol. Amounts and types of carbonaceous species formed after FT reaction for 5-10 minutes at 150, 175, 200 and 285 C vary significantly with temperature. Mr. Brian Critchfield completed his M.S. thesis work on a statistically designed study of the kinetics of FTS on 20% Fe/alumina. Preparation of a paper describing this work is in progress. Results of these studies were reported at the Annual Meeting of the Western States Catalysis and at the San Francisco AIChE meeting. In the coming period, studies will focus on quantitative determination of the rates of kinetically-relevant elementary steps on unsupported Fe catalysts with/without K and Pt promoters by SSITKA method. This study will help us to (1) understand effects of promoter and support on elementary kinetic parameters and (2) build a microkinetics model for FTS on iron. Calculations using periodic, self-consistent Density Functional Theory (DFT) methods were performed on models of defected Fe surfaces, most significantly the stepped Fe(211) surface. Binding Energies (BE's), preferred adsorption sites and geometries of all the FTS relevant stable species and intermediates were evaluated. Each elementary step of our reaction model was fully characterized with respect to its thermochemistry and comparisons between the stepped Fe(211) facet and the most-stable Fe(110) facet were established. In most cases the BE's on Fe(211) reflected the trends observed earlier on Fe(110), yet there were significant variations imposed on the underlying trends. Vibrational frequencies were evaluated for the preferred adsorption configurations of each species with the aim of evaluating the entropy-changes and preexponential factors for each elementary step. Kinetic studies were performed for the early steps of FTS (up to CH{sub 4} formation) and CO dissociation. This involved evaluation of the Minimum Energy Pathway (MEP) and activation energy barrier for the steps involved. We concluded that Fe(211) would allow for far more facile CO dissociation in comparison to other Fe catalysts studied so far, but the other FTS steps studied remained mostly unchanged.

We have developed a flexible, general-purpose microkinetic modeling code, Micki, to analyze complex, heterogeneously catalyzed chemical reactions based upon first-principles calculations. This Python-based code is modular and object oriented, framing the development of microkinetic models in familiar chemical terms. We also present novel approaches, incorporated into Micki, to describe diffusion limited reactions, multidentate bindings, thermodynamically consistent lateral interactions, and Brønsted-Evans-Polanyi estimates of changes in barrier heights. Micki has built-in modules for subsequent analysis of microkinetic models, including degree of rate control and rate order. As a demonstration of the power and flexibility of the code, we build a microkinetic model for the water-gas shift reaction and compare to previously published experimental results and microkinetic models, showing that Micki can quantitatively reproduce experimental turnover frequencies with minimal empirical optimization.

An improved microkinetic model for the water gas shift reaction on copper

This chapter provides a detailed description of how to build the so-called mean-field microkinetic models, proposed and popularized by Dumesic and coworkers, using inputs from density functional theory (DFT) and various types of experimental data. In particular, using illustrative examples, the chapter deals with (1) formulating thermodynamically consistent ordinary differential equation (ODE) based microkinetic models, (2) modern approaches to solving these ODEs, (3) estimation of parameters through optimization and Bayesian inference, and (4) local and global degree of rate control measures. We end with an illustrative example of building a microkinetic model of the water gas shift reaction on Cu(111).

The objective of this work is to elucidate controlling mechanisms in NOx reduction, develop reduced‐order reaction models, and analyze the reactor performance using the reduced‐order reaction model for the NO–CO reaction. We start with the microkinetic model on platinum, which describes the mechanism of catalytic reduction of NO by CO. The formation of the main product N2O and the competitive formation of the side product N2 are accounted for in the microkinetic model. Sensitivity and reaction path analysis have been carried out to determine the rate‐limiting steps as well as the most abundant reactive intermediates in the system. Owing to the differences between system performance at high and low temperatures, the model has been analyzed in detail in these temperature regimes. Two closed‐form expressions, corresponding to the two global reactions involved, have been derived. The characteristic features of the microkinetic model such as the sharp increase in NO conversion and the selectivity to N2O are captured well by the reduced model. The reduced‐order model has been extended to the rhodium catalyst as well. © 2012 Wiley Periodicals, Inc. Int J Chem Kinet 44: 577–585, 2012

Kinetic modeling of electro-catalytic reactions based on micro-kinetic approaches allows at obtaining information on reaction mechanisms and determine how some parameters, such as rate coefficients, are affected by the conditions under which the catalyst operates.1 Unlike a kinetic model, which only considers the overall reaction rate, micro-kinetic modeling provides access to fundamental information of elementary reaction steps. In this sense, we present a micro-kinetic model linked to a proposed reaction mechanism for the formic acid electro-oxidation reaction (FAEOR) on platinum. The model was tested by numerical simulations under voltammetric and oscillatory regimes.2 We formulated the micro-kinetic model using the following workflow stages: i) Gathering information from literature, including spectroscopy and computational chemistry studies; ii) Mechanism proposal and model description that consists of a set of ordinary differential equations involving charge and mass balances on the electrocatalytic surface; iii) Numerical resolution of the model using a set of test parameters; iv) Comparison with different sets of experiments including cyclic voltammetry and potential oscillations under the galvanostatic regime. The initial electrokinetic parameters, associated with the rate coefficients, were adjusted and re-evaluated by an iterative procedure to improve the description of the experimental observations.The type of electrochemical experiments performed in the micro-kinetic model is fundamental to the success of validation. Under oscillatory conditions, galvanostatic experiments are an excellent complement to other electrodynamic techniques because the rate of several processes that are happening in the electrode-solution interface can be associated with oscillatory features such as frequency, amplitude, and oscillation format.3 Thus, the self-organized mode offers a sensitive tool to evaluate the proposed reaction mechanism. In this work, we studied the oscillatory characteristics in the electrochemical response of the FAEOR to determine the rate coefficients as a function of the electrode potential. The consistency of the numerical solution of our model (Figure 1), namely frequency and amplitude of the potential oscillations, with the experimental response makes our proposal a plausible possibility for the FAEOR, providing evidence to the clarification of this controversial process through a micro-kinetic approach. References (1) Campbell, C. T. Micro- and Macro-Kinetics: Their Relationship in Heterogeneous Catalysis. Top. Catal. 1994, 1 (3–4), 353–366. https://doi.org/10.1007/BF01492288.(2) Calderón-Cárdenas, A.; Hartl, F. W.; Gallas, J. A. C.; Varela, H. Modeling the Triple-Path Electro-Oxidation of Formic Acid on Platinum: Cyclic Voltammetry and Oscillations. Catal. Today 2021, 359, 90–98. https://doi.org/10.1016/j.cattod.2019.04.054.(3) Machado, E. G.; Varela, H. Kinetic Instabilities in Electrocatalysis. Encyclopedia of Interfacial Chemistry; Elsevier, 2018; pp 701–718. https://doi.org/10.1016/B978-0-12-409547-2.13369-4. Figure 1

This project builds on the hypothesis that the hydrotreating processes for the removal of oxygen and sulfur are fundamentally similar at the atomic-scale and existing knowledge from the treatment of petroleum derived feedstock can be leveraged for the design of novel catalysts for the upgrade of bio-oil. We tested this hypothesis by comparing computed potential energy diagrams for hydrodesulfurization (HDS) of thiophene over MoS2 with hydrodeoxygenation (HDO) of furan over MoO3 and concluded that certain aspects, such as catalyst promotion with transition metals, are valid strategies for both reactions. On the other hand, we also noticed significant differences in the mechanism for hydrogen (H2) activation, which requires sites with metallic character. While MoS2 is known to have metallic edge states that can catalyze H2 dissociation, this elementary step is prohibitively slow on defect-free oxides. Only in the presence of vacancies or by creating metal/oxide interfaces can efficient H2 activation sites during HDO be formed. The need for bifunctional catalyst when it comes to efficient and selective HDO or dehydrogenation reactions was further corroborated in joint experimental and theoretical studies of the Guerbet reaction for the coupling of biomass derived oxygenates over PdCu alloys, nitrate reduction over In-promoted Pd nanoparticles, and ethylene dehydroaromatization over Ga-exchanged ZSM-5 zeolites. All of these catalytic systems have in common that catalytic sites with distinct functional requirements are needed to create a working catalyst. Detailed computational studies were carried out for HDO of m-cresol and phenol on Ru-modified TiO2 surfaces, which allowed us to attribute catalytic activity to the metal/oxide interface. A surprising finding was that heterolytic cleavage of the H-H bond across the Ru/TiO2 interface was critically important, despite lower barriers for homolytic H2 activation on Ru metal. The explanation lies in the high barriers for hydrogen spillover from Ru to TiO2, which becomes unnecessary in the heterolytic activation pathway. Moreover, we also reported that proton transfer steps between metal and oxide sites are mediated by weakly adsorbed surface water. During attempts to develop and validate a kinetic Monte Carlo (kMC) model for HDO reactions at the Ru/TiO2 interface, it became clear that lateral interactions are paramount to describe realistic surface chemistry and without these interactions, the reduction and hydroxylation behavior from our simulations was inconsistent with reported experiments. To assess the importance of lateral interactions in popular computational catalyst design strategies relying on the identification of reactivity descriptors, which can be used along with Brønsted–Evans–Polanyi (BEP) and scaling relations as input to a microkinetic model (MKM) to make predictions for activity or selectivity trends, we compared predicted trends with those obtained from descriptor-based kMC models. We critically evaluated the benefits of kMC over MKM in terms of trend predictions and computational cost when using only a small set of input parameters. After confirming that in the absence of lateral interactions the kMC and MKM approaches yield identical trends and mechanistic information, we observed substantial differences between the two kinetic models when lateral interactions were introduced. The mean-field implementation applies coverage corrections directly to the descriptors, causing an artificial overprediction of the activity of strongly binding metals. In contrast, the cluster expansion in kMC implementation can differentiate among the highly active metals but it is very sensitive to the set of included interaction parameters. Considering that computational screening relies on a minimal set of descriptors, for which MKM makes reasonable trend predictions at a ca. three orders of magnitude lower computational cost than kMC, we concluded that the MKM approach does provide an overall better entry point for computational catalyst design. Overall, this project has led to 11 peer-reviewed publications, and their scientific is impact is well illustrated by their combined 702 citations.

A comprehensive theoretical study of a Au15Cu15 cluster on MgO(100) supports and its catalytic activity for CO oxidation has been performed based on the density functional theory and microkinetic modeling. Molecular adsorption and different reaction paths based on the Langmuir–Hinshelwood (LH) and Eley–Rideal (ER) mechanisms have been explored by tuning the location of vacancies in MgO(100). The charge states of the Au15Cu15 cluster are negative on all supports, defect-free, O-vacancy (F-center), and Mg-vacancy (V-center), and the effect is significantly amplified on the F-center. In each case, the O2 molecule can be effectively activated upon adsorption and dissociated to 2 × O atoms easily, and the reaction modeling takes into account also the reaction paths with adsorbed O atoms. Overall, CO oxidation has lower reaction barriers on the cluster on the F-center. The microkinetic modeling analysis reveals that CO oxidation is very sensitive to the CO partial pressure, as the relatively strong CO binding l...

Based on the free text data of virtual health community, this paper aims to construct a detection model of adverse drug reaction events by using machine learning. Firstly, this paper reviews the research status related to the theme, builds the overall framework of the detection model of adverse drug reaction events, and expounds the basic elements and logical relationship of each layer model. DIKW system is adopted as the model to design the logical framework. Then, considering the dependence of output markers, the extraction model of adverse drug reactions based on BiLSTM and CRF is established. By analyzing the structural data of FAERS, we propose to classify the characteristics according to the differences of drug molecular structure, and construct a multicore function pool. The simulation structure shows that the model can help doctors reduce the risk of potential adverse reactions when prescribing, and assist the adverse drug reaction monitoring institutions to detect the potential adverse drug reactions in advance.