Machine Learning and Informatics Based Elucidation of Reaction Pathways for Upcycling Model Polyolefin to Aromatics.

Catalytic upcycling of plastics results in a complex network of potentially thousands of reactions and intermediates. Manual analysis of such a network using ab initio methods to identify plausible reaction pathways and rate-controlling steps is intractable. Here, we combine informatics-based reaction network generation and machine learning based thermochemistry calculation to identify plausible (nonelementary step) pathways involved in dehydroaromatization of a model polyolefin, n-decane, to form aromatic products. All 78 aromatic molecules found involve a sequence comprising dehydrogenation, ß-scission, and cyclization steps (in slightly different order). The plausible flux-carrying pathway depends on the family of reactions that is rate-controlling while the thermodynamic bottleneck is the first dehydrogenation step of n-decane. The adopted workflow is system agnostic and can be applied to understand the overall thermochemistry of other upcycling systems.

Combining Computational Modeling with Reaction Kinetics Experiments for Elucidating the In Situ Nature of the Active Site in Catalysis.

Microkinetic modeling based on density functional theory (DFT) derived energetics is important for addressing fundamental questions in catalysis. The quantitative fidelity of microkinetic models (MKMs), however, is often insufficient to conclusively infer the mechanistic details of a specific catalytic system. This can be attributed to a number of factors such as an incorrect model of the active site for which DFT calculations are performed, deficiencies in the hypothesized reaction mechanism, inadequate consideration of the surface environment under reaction conditions, and intrinsic errors in the DFT exchange-correlation functional. Despite these limitations, we aim at developing a rigorous understanding of the reaction mechanism and of the nature of the active site for heterogeneous catalytic chemistries under reaction conditions. By achieving parity between experimental and modeling outcomes through robust parameter estimation and by ensuring coverage-consistency between DFT calculations and MKM predictions, it is possible to systematically refine the mechanistic model and, thereby, our understanding of the catalytic active site in situ.Our general approach consists of developing ab initio informed MKM for a given active site and then re-estimating the energies of the transition and intermediate states so that the model predictions match quantities measured in reaction kinetics experiments. If (i) model-experiment parity is high, (ii) the adjustments to the DFT-derived energetics for a given model of the active site are rationalized within the errors of standard DFT exchange-correlation functionals, and (iii) the resultant MKM predicts surface coverages that are consistent with those assumed in the DFT calculations used to initialize the MKM, we conclude that we have correctly identified the active site and the reaction mechanism. If one or more of these requirements are not met, we iteratively refine our model by updating our hypothesis for the structure of the active site and/or by incorporating coverage effects, until we obtain a high-fidelity coverage-self-consistent MKM whose final kinetic and thermodynamic parameters are within error of the values derived from DFT.Using the catalytic reaction of formic acid (FA, HCOOH) decomposition over transition-metal catalysts as an example, here we provide an account of how we applied this algorithm to study this chemistry on powder Au/SiC and Pt/C catalysts. For the case of Au catalysts, on which the FA decomposition occurred exclusively through the dehydrogenation reaction (HCOOH → CO2+H2), our approach was used to iteratively refine the model starting from the (111) facet until we found that specific ensembles of Au atoms present in sub-nanometer clusters can describe the active site for this catalysis. For the case of Pt catalysts, wherein both dehydrogenation (HCOOH → CO2 + H2) and dehydration (HCOOH → CO + H2O) reactions were active, our approach identified that a partially CO*-covered (111) surface serves as the active site and that CO*-assisted steps contributed substantially to the overall FA decomposition activity. Finally, we suggest that once the active site and the mechanism are conclusively identified, the model can subsequently serve as a high-quality basis for designing specific goal-oriented experiments and improved catalysts.

Sustainable Aviation Fuel Molecules from (Hemi)Cellulose: Computational Insights into Synthesis Routes, Fuel Properties, and Process Chemistry Metrics.

Production of sustainable aviation fuels (SAFs) can significantly reduce the aviation industry's carbon footprint. Current pathways that produce SAFs in significant volumes from ethanol and fatty acids can be costly, have a relatively high carbon intensity (CI), and impose sustainability challenges. There is a need for a diversified approach to reduce costs and utilize more sustainable feedstocks effectively. Here, we map out catalytic synthesis routes to convert furanics derived from the (hemi)cellulosic biomass to alkanes and cycloalkanes using automated network generation with RING and semiempirical thermochemistry calculations. We find >100 energy-dense C8-C16 alkane and cycloalkane SAF candidates over 300 synthesis routes; the top three are 2-methyl heptane, ethyl cyclohexane, and propyl cyclohexane, although these are relatively short. The shortest, least endothermic process chemistry involves C-C coupling, oxygen removal, and hydrogen addition, with dehydracyclization of the heterocyclic oxygens in the furan ring being the most endothermic step. The global warming potential due to hydrogen use and byproduct CO2 is typically 0.7-1 kg CO2/kg SAF product; the least CO2 emitting routes entail making larger molecules with fewer ketonization, hydrogenation, and hydrodeoxygenation steps. The large number of SAF candidates highlights the rich potential of furanics as a source of SAF molecules. However, the structural dissimilarity between reactants and target products precludes pathways with fewer than six synthetic steps, thus necessitating intensified processes, integrating multiple reaction steps in multifunctional catalytic reactors.

Actively Learned Optimal Sustainable Operation of Plasma-Catalyzed Methane Bireforming on La0.7Ce0.3NiO3 Perovskite Catalyst.

Plasma-catalytic bireforming of methane was studied and actively optimized using a La0.7Ce0.3NiO3 perovskite catalyst via experimentation in tandem with response surface modeling. Plasma power, inlet flow rate, temperature, CO2/CH4 ratio, and steam concentration were tuned to maximize a variety of process- and sustainability-based metrics. Analysis of the optimal conditions (with respect to different metrics) with and without the catalyst reveals that dry reforming is driven largely via noncatalytic reactions, while steam reforming and water gas shift reactions require the catalyst. The experimental outcome demonstrated that under optimum reaction conditions using the La0.7Ce0.3NiO3 catalyst it is possible to minimize global warming potential (GWP), in terms of inferred CO2 footprint normalized to hydrogen throughput, resulting in maximizing hydrogen yield through steam reforming (and water gas shift reactions) at an SEI of ≈12 eV/molecule. Furthermore, the highest CH4 conversion reached was 87% while the catalyst showed good activity stability in DBD plasma experiments.The actively learned iterative optimization procedure developed in this work allows for a direct juxtaposition of thermal (heat needed to make steam and heat the plasma reactor) and electrical (power requirement for plasma generation) carbon footprints in a highly nonlinear multivariate process. Furthermore, the corresponding GWP was calculated using a conventional electricity mix, wind electricity, and solar electricity, allowing a direct sustainability assessment in catalyst-assisted plasma conversion of carbonaceous feedstock to H2 and CO.

Revealing the Nature of Active Oxygen Species and Reaction Mechanism of Ethylene Epoxidation by Supported Ag/α-Al2O3 Catalysts.

The oxygen species on Ag catalysts and reaction mechanisms for ethylene epoxidation and ethylene combustion continue to be debated in the literature despite decades of investigation. Fundamental details of ethylene oxidation by supported Ag/α-Al2O3 catalysts were revealed with the application of high-angle annular dark-field-scanning transmission electron microscopy-energy-dispersive X-ray spectroscopy (HAADF-STEM-EDS), in situ techniques (Raman, UV-vis, X-ray diffraction (XRD), HS-LEIS), chemical probes (C2H4-TPSR and C2H4 + O2-TPSR), and steady-state ethylene oxidation and SSITKA (16O2 → 18O2 switch) studies. The Ag nanoparticles are found to carry a considerable amount of oxygen after the reaction. Density functional theory (DFT) calculations indicate the oxidative reconstructed p(4 × 4)-O-Ag(111) surface is stable relative to metallic Ag(111) under the relevant reaction environment. Multiple configurations of reactive oxygen species are present, and their relevant concentrations depend on treatment conditions. Selective ethylene oxidation to EO proceeds with surface Ag4-O2* species (dioxygen species occupying an oxygen site on a p(4 × 4)-O-Ag(111) surface) only present after strong oxidation of Ag. These experimental findings are strongly supported by the associated DFT calculations. Ethylene epoxidation proceeds via a Langmuir-Hinshelwood mechanism, and ethylene combustion proceeds via combined Langmuir-Hinshelwood (predominant) and Mars-van Krevelen (minor) mechanisms.

Elucidating Local Structure and Positional Effect of Dopants in Colloidal Transition Metal Dichalcogenide Nanosheets for Catalytic Hydrogenolysis.

Tailoring nanoscale catalysts to targeted applications is a vital component in reducing the carbon footprint of industrial processes; however, understanding and controlling the nanostructure influence on catalysts is challenging. Molybdenum disulfide (MoS2), a transition metal dichalcogenide (TMD) material, is a popular example of a nonplatinum-group-metal catalyst with tunable nanoscale properties. Doping with transition metal atoms, such as cobalt, is one method of enhancing its catalytic properties. However, the location and influence of dopant atoms on catalyst behavior are poorly understood. To investigate this knowledge gap, we studied the influence of Co dopants in MoS2 nanosheets on catalytic hydrodesulfurization (HDS) through a well-controlled, ligand-directed, tunable colloidal doping approach. X-ray absorption spectroscopy and density functional theory calculations revealed the nonmonotonous relationship between dopant concentration, location, and activity in HDS. Catalyst activity peaked at 21% Co:Mo as Co saturates the edge sites and begins basal plane doping. While Co prefers to dope the edges over basal sites, basal Co atoms are demonstrably more catalytically active than edge Co. These findings provide insight into the hydrogenolysis behavior of doped TMDs and can be extended to other TMD materials.

Data-driven discovery of sparse dynamical model of cardiovascular system for model predictive control.

Cardiovascular diseases remain the leading cause of death globally. In recent years, vagal nerve stimulation (VNS) has shown promising results in the treatment of a number of cardiovascular diseases. In this approach, mild electrical pulses are sent to the brain via the vagus nerve. This open-loop neurostimulation, however, leads to various side effects due to physiological and inter-patient variability and therefore a closed-loop delivery strategy of electrical pulses that accounts for this variability is desired. In this context, we envision data-driven sparse dynamical model parameterized by patient-specific data as appropriate for use in closed loop controller design. In this work, we build a dynamical model for mean arterial pressure and heart rate using the method sparse identification of nonlinear dynamics (SINDy). As a proxy for real datasets or measurements from a patient, we simulate a mechanistic model from the literature and then discover a data-driven model for predicting mean arterial pressure and heart rate in response to neural stimulus. This discovered model is then used to design a controller to be implemented in closed-loop via model predictive control. We observe that this data-driven model is interpretable, consistent with experiments, provides insights on the sensitivity of different stimulation locations and simplifies the formulation of the optimal control problem. Noting the set-point tracking performance of this closed-loop model-based controller that uses this discovered model, we conclude that the model is adequate in capturing the dynamics of a highly nonlinear cardiovascular system for the purpose of optimal predictive controller design.

Site-dependent reactivity of MoS2 nanoparticles in hydrodesulfurization of thiophene.

The catalytically active site for the removal of S from organosulfur compounds in catalytic hydrodesulfurization has been attributed to a generic site at an S-vacancy on the edge of MoS2 particles. However, steric constraints in adsorption and variations in S-coordination means that not all S-vacancy sites should be considered equally active. Here, we use a combination of atom-resolved scanning probe microscopy and density functional theory to reveal how the generation of S-vacancies within MoS2 nanoparticles and the subsequent adsorption of thiophene (C4H4S) depends strongly on the location on the edge of MoS2. Thiophene adsorbs directly at open corner vacancy sites, however, we find that its adsorption at S-vacancy sites away from the MoS2 particle corners leads to an activated and concerted displacement of neighboring edge S. This mechanism allows the reactant to self-generate a double CUS site that reduces steric effects in more constrained sites along the edge.

Computational Investigation of Correlations in Adsorbate Entropy for Pure-Silica Zeolite Adsorbents.

Vast numbers of unstudied hypothetical porous frameworks continue to spark interest in optimizing adsorption and catalytic processes. Evaluating the use of such materials depends on the accessibility of thermodynamic metrics such as the free energy, which, in turn, depend on the satisfactory estimation or calculation of the adsorption entropy, which often remains elusive. Previous works using simulations and experimental data have demonstrated relationships between the entropy and system descriptors, allowing for sensible predictions based on more-easily obtained physical parameters. However, the resultant conclusions were either based on experimental data for industrially relevant alkanes or lacked a significant sample size. In this paper, we evaluate correlations between gas-phase and adsorbed-phase entropies for a larger and more chemically diverse set of adsorbate molecules by using force fields and statistical mechanical expressions to calculate those entropies. In total, we perform calculations for 37 molecules across 10 chemical categories available in the TraPPE force field set, as adsorbed in five siliceous zeolites. Our results show that linear correlations between the gas- and adsorbed-phase entropies persist for the larger and diverse set of adsorbate molecules studied here, proving a broader applicability and justifying the use of simple correlations for many adsorbates and, presumably, adsorbent materials.

Predicting Adsorption Energies Using Multifidelity Data.

In this paper, we show that the binding energy of small adsorbates on transition-metal surfaces can be modeled to a high level of fidelity with data from multiple sources using multitask Gaussian processes (MT-GPs). This allows us to take advantage of the relatively abundant "low fidelity" data (such as from density functional theory computations) and small amounts of "high fidelity" computational (e.g., using the random phase approximation) or experimental data. We report two case studies here-one using purely computational datasets and the other using a combination of experimental and computational datasets-to explore the performance of MT-GPs. In both cases, the performance of MT-GPs is significantly better than single-task models built on a single data source. We posit that this method can be used to learn improved models from fused datasets, thereby maximizing model accuracy under tight computational and experimental budget.

Designing Molecular Building Blocks for the Self-assembly of Complex Porous Networks.

The creation of molecular or colloidal building blocks which can self-assemble into complex, ordered porous structures has been long sought-after, and so are the guiding principles behind this creation. The pursuit of this goal has led to the creation of novel classes of materials like metal organic frameworks (MOFs) and covalent organic frameworks (COFs). In theory, a tremendous number of structures can be formed by these materials due to the variety of geometries available to their building blocks. However, most realized crystal structures tend to be simple or homoporous and typically assemble from building blocks with high degrees of symmetry. Building blocks with low degrees of symmetry suitable for assembly into the more complex structures tend to assemble into polymorphous or disordered structures instead. In this work, we use Monte Carlo simulations of patchy vertex-like building blocks to show how the addition of chemical specificity via orthogonally reacting functional sites can allow vertex-like building blocks with even asymmetric geometries to self-assemble into ordered crystallites of various complex structures. In addition to demonstrating the utility of such a strategy in creating ordered, heteroporous structures, we also demonstrate that it can be used as a means for tuning specific features of the crystal structure, accomplishing such aims as the control of relative pore sizes. We also discuss heuristics for properly designing molecules so that they can assemble into target structures.