Automated Mechanism Generation Research Articles

Species selection for automatic chemical kinetic mechanism generation

AbstractMany important chemical kinetic systems require detailed chemical kinetic models to resolve. These detailed kinetic models can involve thousands of species and hundreds of thousands of chemical reactions, making them difficult to construct by hand. Modern automatic mechanism generation algorithms can mostly be divided into two classes: rule and rate based. Rule‐based generators choose species based on user defined constraints on species and reaction classes. Rate‐based generators generate a much larger set of potentially important species and reactions and then choose which ones to add based on running simulations of species and reactions deemed important and calculating the flux to potentially important species. In principle, the latter is preferable, as it requires the user to make far fewer assumptions about what is important in the system. However, while the effectiveness of the rate‐based approach has been demonstrated in a wide variety of systems, it has also been demonstrated to have difficulty picking up important low‐flux chemistries. Here we present a discussion of the challenges associated with rate‐based mechanism generation and new algorithms that are able to efficiently mitigate these challenges improving species selection during mechanism generation in a set of case studies.

Predicting the autoignition behaviour of tailorable advanced biofuel blends using automatically generated mechanisms

Emerging processes such as biomass alcoholysis have the potential to provide tailorable, advanced biofuels to replace conventional fossil fuels. Knowledge of the engine-relevant behaviour for such fuels is evolving but is currently limited. Simulation tools may assist in the exploration of this behaviour but are reliant on the availability of robust, detailed kinetic mechanisms to produce accurate predictions. Automatic mechanism generation (AMG) techniques may be applied to facilitate the production of such mechanisms, as long as the utilised databases contain high-quality kinetic and thermochemical information of relevance to the functional groups of interest. To model the combustion characteristics of complex fuel blends, Reaction Mechanism Generator (RMG) is applied in this work to produce detailed ethyl (ethyl levulinate, diethyl ether, ethanol) and butyl (n‑butyl levulinate, di-n‑butyl ether, n-butanol) kinetic mechanisms. The predictive capabilities of these mechanisms are evaluated against a combination of literature data for individual components and new experimental measurements of ethyl and butyl blends. Stoichiometric blend ignition delay times are measured in a rapid compression machine at a compressed pressure of 20 bar and compressed temperatures of 645–960 K. The investigated blends are formulated to achieve a desired research octane number (blends ELV1 and BLV1) or to match physical property limits of diesel fuels (blends ELV2 and BLV2). Ethyl blend measurements show clear examples of complex low temperature oxidation behaviour, but this is not observable in the butyl cases, as the high boiling point of butyl levulinate necessitates the use of significant dilution in ignition delay experiments. The generated models show a high degree of accuracy when compared to measurements at thermodynamic conditions of relevance to engine technologies. The blending behaviour shown by the experimental measurements is also well predicted. This work highlights the importance of database additions/modifications based on low uncertainty, high quality, literature data when using AMG methods.

Automatic mechanism generation involving kinetics of surface reactions with bidentate adsorbates

RMG was expanded with multidentate functionalities, which enables the automated discovery of mechanisms for the complex non-oxidative dehydrogenation of ethane.

Accelerating Electrocatalyst Innovation: High-Throughput Automated Microkinetic Modeling

To reduce energy-related emissions, we must create chemicals from carbon dioxide with renewable energy, instead of from petroleum. Designing such a process needs computer models that can describe all the chemical reactions that are happening, including those driven by electrochemistry. There may be thousands of such reactions, so we must build a tool to find them automatically. Our reaction mechanism generator for electrocatalysis will create detailed kinetic models for many electrochemical processes, but for this initial project we are targeting the reduction of carbon dioxide to produce propanol, a useful hydrocarbon product. The kinetic models will be used to simulate an electrochemical reactor, then analyzed both technically and economically to determine which catalyst materials and reactor designs are best. This will allow new catalyst materials to be discovered using powerful computers, instead of using very time-consuming and expensive experiments.Building on the state-of-the-art open-source Reaction Mechanism Generator (RMG) software [1,2], we are creating the first automated mechanism generator for electrochemical reactions on a catalyst. Supplied with a feedstock, catalyst, and conditions, it will propose a detailed kinetic model comprising hundreds of intermediate species and reactions. Reactor simulations will predict product yields as a function of feed, concentrations, and potential, allowing conditions to be optimized and the trade-off between conversion and selectivity to be investigated. In turn this could inform a detailed technoeconomic analysis, enabling catalyst materials to be screened in silico on an economic basis. Meanwhile sensitivity analysis of the simulations will select estimated parameters for refinement with DFT calculations.We must first extend RMG to allow charged species, and implement proton coupled electron transfer (PCET) reactions. We will at first use the computational hydrogen electrode model to determine the Gibbs energy of a species (and hence reaction) as a function of chemical potential. Potential-dependent kinetics will be estimated using a charge transfer coefficient model. In preliminary work with a framework like this we have constructed a reaction mechanism for carbon dioxide reduction on Cu(111) with 37 species and 292 reactions (23 of them electrochemical) [3]. The model discovered many key electrochemical pathways to reduce CO2 to experimentally observed products such as methane, methanol, formic acid, ethylene, and ethanol. Acknowledgements The information, data, or work presented herein was funded in part by the Advanced Research Projects Agency-Energy (ARPA-E), U.S. Department of Energy, under Award Number DE-AR0001786. The authors thank Dr. David Farina Jr. for his significant contributions to the proof-of-concept.

Automatic mechanism generation for the combustion of advanced biofuels: A case study for diethyl ether

AbstractAdvanced biofuels have the potential to supplant significant fractions of conventional liquid fossil fuels. However, the range of potential compounds could be wide depending on selected feedstocks and production processes. Not enough is known about the engine relevant behavior of many of these fuels, particularly when used within complex blends. Simulation tools may help to explore the combustion behavior of such blends but rely on robust chemical mechanisms providing accurate predictions of performance targets over large regions of thermochemical space. Tools such as automatic mechanism generation (AMG) may facilitate the generation of suitable mechanisms. Such tools have been commonly applied for the generation of mechanisms describing the oxidation of non‐oxygenated, non‐aromatic hydrocarbons, but the emergence of biofuels adds new challenges due to the presence of functional groups containing oxygen. This study investigates the capabilities of the AMG tool Reaction Mechanism Generator for such a task, using diethyl ether (DEE) as a case study. A methodology for the generation of advanced biofuel mechanisms is proposed and the resultant mechanism is evaluated against literature sourced experimental measurements for ignition delay times, jet‐stirred reactor species concentrations, and flame speeds, over conditions covering φ = 0.5–2.0, P = 1–100 bar, and T = 298–1850 K. The results suggest that AMG tools are capable of rapidly producing accurate models for advanced biofuel components, although considerable upfront input was required. High‐quality fuel specific reaction rates and thermochemistry for oxygenated species were required, as well as a seed mechanism, a thermochemistry library, and an expansion of the reaction family database to include training data for oxygenated compounds. The final DEE mechanism contains 146 species and 4392 reactions and in general, provides more accurate or comparable predictions when compared to literature sourced mechanisms across the investigated target data. The generation of combustion mechanisms for other potential advanced biofuel components could easily capitalize on these database updates reducing the need for future user interventions.

Automated Generation of a Compact Chemical Kinetic Model for n-Pentane Combustion.

We have employed automated mechanism generation tools to construct a detailed chemical kinetic model for combustion of n-pentane, as a step toward the generation of compact kinetic models for larger alkanes. Pentane is of particular interest as a prototype for combustion of alkanes and as the smallest paraffin employed as a hybrid rocket fuel, albeit at cryogenic conditions. A reaction mechanism for pentane combustion thus provides a foundation for modeling combustion of extra-large alkanes (paraffins) that are of more practical interest as hybrid rocket fuels, for which manual construction becomes infeasible. Here, an n-pentane combustion kinetic model is developed using the open-source software package Reaction Mechanism Generator (RMG). The model was generated and tested across a range of temperatures (650 to 1350 K) and equivalence ratios (0.5, 1.0, 2.0) at pressures of 1 and 10 atm. Available thermodynamic and kinetic databases were incorporated wherever possible. Predictions using the mechanism were validated against the published laminar burning velocities (Su) and ignition delay times (IDT) of n-pentane. To improve the model performance, a comprehensive analysis, including reaction path and sensitivity analyses of n-pentane oxidation, was performed, leading us to modify the thermochemistry and rate parameters for a few key species and reactions. These were combined as a separate data set, an RMG library, that was then used during mechanism generation. The resulting model predicted IDTs as accurately as the best manually constructed mechanisms, while remaining much more compact. It predicted flame speeds to within 10% of published experimental results. The degree of success of automated mechanism generation for this case suggests that it can be extended to construct reliable and compact models for combustion of larger n-alkanes, particularly when using this mechanism as a seed submodel.

Automating the generation of detailed kinetic models for halocarbon combustion with the Reaction Mechanism Generator

Originally developed to predict the chemical kinetics of hydrocarbon combustion via automated generation of detailed reaction mechanisms, Reaction Mechanism Generator (RMG) contains extensive thermokinetic data for C,H,O chemisty, and has more recently been expanded to nitrogen and sulfur. In this work, we present the addition of halogen (fluorine, chlorine, and bromine) chemisty to RMG to enable automated generation of detailed kinetic models for halocarbon combustion. RMG’s existing reaction templates are updated to include halogens, and 11 new reactions families are created specific to halogen chemistry. Notably, kinetics for more than 1000 elementary reactions are calculated via ab inito methods and transition state theory, and these kinetic data are combined with kinetics from literature sources to train rate rule decision tree estimators. Additionally, halogen groups are added to RMG’s statistical mechanics database, enabling model generation with RMG’s pressure dependence module and automated computation of microcanonical rate constants for unimolecular networks. Halogen groups are also incorporated in RMG’s transport database to provide estimated parameters for the Lennard-Jones potential, important for transport-dependent simulations including laminar flame speeds. To demonstrate RMG’s capability for predicting halocarbon combustion, RMG is used to build a flame suppression model for 2-BTP (CH2=CBrCF3) in methane flames. The laminar flame speeds of RMG’s 2-BTP model show good agreement with a published model under a variety of reaction conditions. Automating the generation of detailed kinetic models for halocarbon combustion will facilitate the exploration of previously unexplored reaction pathways, thereby accelerating the development of greener refrigerants and suppressants, as well as advancing the field of automated mechanism generation.

Detailed Microkinetics for the Oxidation of Exhaust Gas Emissions through Automated Mechanism Generation

Emissions from vehicles contain a variety of pollutants that must be either oxidized or reduced efficiently in the catalytic converter. Improvements to the catalyst require knowledge of the microkinetics, but the complexity of the exhaust gas mixture makes it challenging to identify the reaction network. This complexity was tackled by using the “Reaction Mechanism Generator” (RMG) to automatically generate microkinetic models for the oxidation of combustion byproducts from stoichiometric gasoline direct injection engines on Pt(111). The possibilities and the limitations encountered during the generation procedure are discussed in detail. A combination of first-principles-based mechanism construction and top-down parameter refinement allows a description of experimental results obtained by kinetic testing of a Pt/Al2O3 monolith under stoichiometric conditions. The study can serve as a blueprint for the usage of RMG for other challenging heterogeneously catalyzed reactions.

Transfer Learning Approach to Multitarget Temperature-Dependent Reaction Rate Prediction.

Accurate prediction of temperature-dependent reaction rate constants of organic compounds is of great importance to both atmospheric chemistry and combustion science. Extensive work has been done on developing automated mechanism generation systems but the lack of quality reaction rate data remains a huge bottleneck in the application of highly detailed mechanisms. Machine learning prediction models have been recently adopted to alleviate the data gap in thermochemistry and have great potential to do the same for kinetic data with the recent release of quality reaction rate data compilations. The ultimate goal is to formulate easily accessible, general-purpose, temperature-dependent, and multitarget models for the prediction of reaction rates. To that end, we propose a model that depends on the well-known Morgan fingerprints as well as learned representations transferred from the QM9 data set. We propose the use of an Arrhenius-based loss where predictions of the three modified-Arrhenius parameters (A, n, and B = Ea/R) are given instead of the direct prediction of reaction rate constants. Our model is >35% more accurate compared to a baseline model of feed forward network (FFN) on Morgan fingerprints.

Estimation of mechanistic parameters in the gas-phase reactions of ozone with alkenes for use in automated mechanism construction

Abstract. Reaction with ozone is an important atmospheric removal process for alkenes. The ozonolysis reaction produces carbonyls and carbonyl oxides (Criegee intermediates, CI), which can rapidly decompose to yield a range of closed shell and radical products, including OH radicals. Consequently, it is essential to accurately represent the complex chemistry of Criegee intermediates in atmospheric models in order to fully understand the impact of alkene ozonolysis on atmospheric composition. A mechanism construction protocol is presented which is suitable for use in automatic mechanism generation. The protocol defines the critical parameters for describing the chemistry following the initial reaction, namely the primary carbonyl/CI yields from the primary ozonide fragmentation, the amount of stabilisation of the excited CI, the unimolecular decomposition pathways, rates and products of the CI, and the bimolecular rates and products of atmospherically important reactions of the stabilised CI (SCI). This analysis implicitly predicts the yield of OH from the alkene–ozone reaction. A comprehensive database of experimental OH, SCI and carbonyl yields has been collated using reported values in the literature and used to assess the reliability of the protocol. The protocol provides estimates of OH, SCI and carbonyl yields with root mean square errors of 0.13 and 0.12 and 0.14, respectively. Areas where new experimental and theoretical data would improve the protocol and its assessment are identified and discussed.

Extensive High-Accuracy Thermochemistry and Group Additivity Values for Halocarbon Combustion Modeling

Standard enthalpies, entropies, and heat capacities are calculated for 16,813 halocarbons using an automated high-fidelity thermochemistry workflow. This workflow generates conformers at density functional tight binding (DFTB) level, optimizes geometries, calculates harmonic frequencies, and performs 1D hindered rotor scans at DFT level, and computes electronic energies at G4 level. The computed enthalpies of formation for 400 molecules show good agreement with literature references, but the majority of the calculated species have no reference in the literature. Thus, this work presents the most accurate thermochemistry for many halocarbons to date. This new data set is used to train an extensive ensemble of group additivity values and hydrogen bond increment groups within the Reaction Mechanism Generator (RMG) framework. On average, the new group values estimate standard enthalpies for halogenated hydrocarbons within 3 kcal/mol of their G4 values. A significant contribution towards automated mechanism generation of halocarbon combustion, this research provides thermochemical data for thousands of novel halogenated species and presents a self-consistent set of halogen group additivity values.

A Study on Fundamental Combustion Properties of Oxymethylene Ether-2

Abstract Oxymethylene ethers (OMEn, n = 1–5) are a promising class of synthetic fuels that have the potential to be used as additives or substitutes to diesel in compression ignition engines. A comprehensive understanding of their combustion properties is required for their safe and efficient utilization. In this study, the results of a combined experimental and modeling work on oxidation of OME2 are reported: (i) Ignition delay time measurements of stoichiometric OME2/synthetic air mixtures diluted 1:5 with nitrogen using the shock tube method at pressures of 1, 4, and 16 bar, and (ii) laminar flame speeds of OME2/air mixtures using the cone angle method at atmospheric and elevated pressures of 3 and 6 bar. The experimental data sets obtained have been used for validation of three detailed reaction mechanisms of OME2 obtained from literature. The results of ignition delay time measurements showed that ignition of OME2 is characterized by pre-ignition activity at the low temperature side of the measurements regardless of the pressure. Regarding the performance of the different reaction mechanisms, the model from Cai et al. (2020, “Auto-Ignition of Oxymethylene Ethers (OMEn, n = 2–4) as Promising Synthetic e-Fuels From Renewable Electricity: Shock Tube Experiments and Automatic Mechanism Generation,” Fuel, 264, p. 116711) best predicted the temperature and pressure dependence of ignition delay times. For laminar flame speeds, the experimental data were well matched by the mechanism from Ren et al. (2019) at p = 1, 3, and 6 bar and for all equivalence ratios considered. From sensitivity analyses calculations, it was observed that chain reactions involving small radicals, i.e., H, O, OH, HO2, and CH3 control the oxidation of OME2. The comparison of the results of this work and our previous work (Ngugi et al. (2021)) on OME1 show that these two fuels have similar oxidation pathways. The results obtained in this work will contribute to a better understanding of the combustion of oxymethylene ethers, and thus, to the design and optimization of burners and engines as well.

Quantifying the Impact of Parametric Uncertainty on Automatic Mechanism Generation for CO2 Hydrogenation on Ni(111).

Automatic mechanism generation is used to determine mechanisms for the CO2 hydrogenation on Ni(111) in a two-stage process while considering the correlated uncertainty in DFT-based energetic parameters systematically. In a coarse stage, all the possible chemistry is explored with gas-phase products down to the ppb level, while a refined stage discovers the core methanation submechanism. Five thousand unique mechanisms were generated, which contain minor perturbations in all parameters. Global uncertainty assessment, global sensitivity analysis, and degree of rate control analysis are performed to study the effect of this parametric uncertainty on the microkinetic model predictions. Comparison of the model predictions with experimental data on a Ni/SiO2 catalyst find a feasible set of microkinetic mechanisms within the correlated uncertainty space that are in quantitative agreement with the measured data, without relying on explicit parameter optimization. Global uncertainty and sensitivity analyses provide tools to determine the pathways and key factors that control the methanation activity within the parameter space. Together, these methods reveal that the degree of rate control approach can be misleading if parametric uncertainty is not considered. The procedure of considering uncertainties in the automated mechanism generation is not unique to CO2 methanation and can be easily extended to other challenging heterogeneously catalyzed reactions.

Screening for New Pathways in Atmospheric Oxidation Chemistry with Automated Mechanism Generation.

In the Earth's atmosphere, reactive organic carbon undergoes oxidation via a highly complex, multigeneration process, with implications for air quality and climate. Decades of experimental and theoretical studies, primarily on the reactions of hydrocarbons, have led to a canonical understanding of how gas-phase oxidation of organic compounds takes place. Recent research has brought to light a number of examples where the presence of certain functional groups opens up reaction pathways for key radical intermediates, including alkyl radicals, alkoxy radicals, and peroxy radicals, that are substantially different from traditional oxidation mechanisms. These discoveries highlight the need for methods that systematically explore the chemistry of complex, functionalized molecules without being prohibitively expensive. In this work, automated reaction network generation is used as a screening tool for new pathways in atmospheric oxidation chemistry. The reaction mechanism generator (RMG) is used to generate reaction networks for the OH-initiated oxidation of 200 mono- and bifunctionally substituted n-pentanes. The resulting networks are then filtered to highlight the reactions of key radical intermediates that are fast enough to compete with traditional atmospheric removal processes as well as "uncanonical" processes which differ from traditionally accepted oxidation mechanisms. Several recently reported, uncanonical atmospheric mechanisms appear in the RMG dataset. These "proof of concept" results provide confidence in this approach as a tool in the search for overlooked atmospheric oxidation chemistry. Several previously unreported reaction types are also encountered in the dataset. The most potentially atmospherically important of these is a radical-carbonyl ring-closure reaction that produces a highly functionalized cyclic alkoxy radical. This pathway is proposed as a promising target for further study via experiments and more detailed theoretical calculations. The approach presented herein represents a new way to efficiently explore atmospheric chemical space and unearth overlooked reaction steps in atmospheric oxidation.

ChemDyME: Kinetically Steered, Automated Mechanism Generation through Combined Molecular Dynamics and Master Equation Calculations.

In many scientific fields, there is an interest in understanding the way in which chemical networks evolve. The chemical networks which researchers focus upon have become increasingly complex, and this has motivated the development of automated methods for exploring chemical reactivity or conformational change in a "black-box" manner, harnessing modern computing resources to automate mechanism discovery. In this work, we present a new approach to automated mechanism generation which couples molecular dynamics and statistical rate theory to automatically find kinetically important reactions and then solve the time evolution of the species in the evolving network. The key to this chemical network mapping through combined dynamics and ME simulation approach is the concept of "kinetic convergence", whereby the search for new reactions is constrained to those species which are kinetically favorable at the conditions of interest. We demonstrate the capability of the new approach for two systems, a well-studied combustion system and a multiple oxygen addition system relevant to atmospheric aerosol formation.

Learning Chemistry of Complex Reaction Systems via a Python First-Principles Reaction Rule Stencil (pReSt) Generator.

Complex reaction networks can be generated with automated network generators from initial reactants and reaction rules. Reaction rule specification is central to network generation. These reaction rules are, at present, user-defined based on (intuitive) expert knowledge of chemistry and are often transferred from gas-phase to surface processes. The catalyst active site geometry is usually left out but is often responsible for selectivity. We propose a first-principles-based reaction mechanism generation framework using density functional theory (DFT) data of published reaction mechanisms. The framework "learns the chemistry" from published mechanisms. It can generate reaction networks not studied before, "flag" reactions not seen before for further DFT convergence tests, and easily reconcile differences between catalysts and reactants that may introduce new pathways never seen before. As such, it can be a diagnostic tool for data (mechanism) quality assessment and novel pathway discovery to new molecules. A software, the Python Reaction Stencil (pReSt), was developed for this purpose to wrap around automatic mechanism generation software. Multiple catalytic chemistries are considered to show the efficacy of the proposed framework.

Automated Mechanism Generation Using Linear Scaling Relationships and Sensitivity Analyses Applied to Catalytic Partial Oxidation of Methane

Automated Mechanism Generation Using Linear Scaling Relationships and Sensitivity Analyses Applied to Catalytic Partial Oxidation of Methane

Automatic construction of transition states and on-the-fly accurate kinetic calculations for reaction classes in automated mechanism generators

A detailed combustion mechanism of hydrocarbon fuels usually contains a large number of reactions and need to have kinetic data assigned to them, where the rate constants are usually given by approximate estimation leading to inaccuracies. In this work, a method for construction transition state (TS) and an isodesmic reactions correction scheme for accurate and automatic calculation of rate constants have been designed. These methods are tested using the cyclization reactions of QOOH radicals and the concerted elimination reactions of RO2 radicals. The results show that the automatic TS construction method is effective to construct the initial guesses of the transition states, the isodesmic reactions corrected results are close to the CCSD(T) results and some of the rate constants are significantly different from the rate constants used in the automated mechanism generator, indicating that it is necessary and significant to perform on-the-fly accurate calculation of rate constants in an automated mechanism generator.

Auto-ignition of oxymethylene ethers (OMEn, n = 2–4) as promising synthetic e-fuels from renewable electricity: shock tube experiments and automatic mechanism generation

Carbon-neutral synthetic fuels can be produced from renewable electricity by the hydrogenation of carbon dioxide captured from air or exhaust gas. A promising class of these synthetic fuels are long-chain oxymethylene ethers (OMEs), which exhibit good auto-ignition characteristics for compression-ignition engine application. This study aims to investigate the auto-ignition of three oxymethylene ethers (OMEn, n = 2–4) numerically and experimentally. A shock tube is applied to measure ignition delay times over a range of initial conditions and the obtained results serve as the validation and optimization targets for a chemical mechanism of OME2-4 developed in this work. This model is derived first using an automatic reaction class-based mechanism generator. To ensure the chemical validity of the mechanism, the automatic generator applies reaction classes and rate rules consistently for OME2-4, which are adopted from a recently published OME1 mechanism. For improved model prediction accuracy of ignition delay times, the mechanism is then optimized automatically by calibrating these rate rules within their uncertainties using data for all OMEn fuels. It is shown that this highly automated model development process is able to provide accurate chemical mechanisms for large fuel components in a very efficient manner, if accurate prior kinetic knowledge exists for their short-chain counterparts.

Kinetics of the hydrogen abstraction alkane + O2 → alkyl + HO2 reaction class: an application of the reaction class transition state theory

In this work, the kinetics of reaction class of hydrogen abstraction from saturated hydrocarbons by O2 molecules has been studied. The high-pressure reaction rate constants were determined using reaction class transition state theory/structure-activity relationship (RC-TST/SAR) methodology, augmented by linear energy relationship (LER) and/or barrier height grouping (BHG) approximations for evaluation of the reaction barrier heights. The parameters needed have been derived from DFT calculations at M06-2X/aug-cc-pVTZ level for a training set of 23 reactions, involving hydrogen abstraction by O2 molecule at primary, secondary, and tertiary carbon sites. The reference reaction rate constant C2H6 + O2 → C2H5 + HO2 was obtained by extrapolation of the simplest reaction within the title family CH4 + O2 → CH3 + HO2. Kinetic parameters of the later one, calculated from canonical variational transition state theory (CVT), were taken from literature. The influence of low-frequency internal rotations has been investigated in details. The error analysis shows that the average systematic error of RC-TST/SAR-derived rate constants at low temperatures is within 25% compared to the explicit RC-TST results and diminishes at higher temperatures. This suggests that the proposed methodology can be effectively implemented in the automated mechanism generation codes to create the fuel combustion mechanisms.