In recent decades, machine learning techniques have received an explosion of interest in the domain of high-throughput materials discovery, which is largely attributed to the fastgrowing development of quantum-chemical methods and learning algorithms. Nevertheless, machine learning for catalysis is still at its initial stage due to our insufficient knowledge of the structure-property relationships. In this regard, we demonstrate a holistic machine-learning framework as surrogate models for the expensive density functional theory to facilitate the discovery of high-performance catalysts. The framework, which integrates the descriptor-based kinetic analysis, material fingerprinting and machine learning algorithms, can rapidly explore a broad range of materials space with enormous compositional and configurational degrees of freedom prior to the expensive quantum-chemical calculations and/or experimental testing. Importantly, advanced machine learning approaches (e.g., global sensitivity analysis, principal component analysis, and exploratory analysis) can be utilized to shed light on the underlying physical factors governing the catalytic activity on a diverse type of catalytic materials with different applications. Chapter 1 introduces some basic concepts and knowledge relating to the computational catalyst design. Chapter 2 and Chapter 3 demonstrate the methodology to construct the machine-learning models for bimetallic catalysts. In Chapter 4, the multi-functionality of the machine-learning models is illustrated to understand the metalloporphyrin's underlying structure-property relationships. In Chapter 5, an uncertainty-guided machine learning strategy is introduced to tackle the challenge of data deficiency for perovskite electrode materials design in the electrochemical water splitting cell. / Doctor of Philosophy / Machine learning and deep learning techniques have revolutionized a range of industries in recent years and have huge potential to improve every aspect of our daily lives. Essentially, machine-learning provides algorithms the ability to automatically discover the hidden patterns of data without being explicitly programmed. Because of this, machine learning models have gained huge successes in applications such as website recommendation systems, online fraud detection, robotic technologies, image recognition, etc. Nevertheless, implementing machine-learning techniques in the field of catalyst design remains difficult due to 2 primary challenges. The first challenge is our insufficient knowledge about the structure-property relationships for diverse material systems. Typically, developing a physically intuitive material feature method requests in-depth expert knowledge about the underlying physics of the material system and it is always an active field. The second challenge is the lack of training data in academic research. In many cases, collecting a sufficient amount of training data is not always feasible due to the limitation of computational/experimental resources. Subsequently, the machine learning model optimized with small data tends to be over-fitted and could provide biased predictions with huge uncertainties. To address the above-mentioned challenges, this thesis focus on the development of robust feature methods and strategies for a variety of catalyst systems using the density functional theory (DFT) calculations. Through the case studies in the chapters, we show that the bulk electronic structure characteristics are successful features for capturing the adsorption properties of metal alloys and metal oxides. While molecular graphs are robust features for the molecular property, e.g., energy gap, of metal-organics compounds. Besides, we demonstrate that the adaptive machine learning workflow is an effective strategy to tackle the data deficiency issue in search of perovskite catalysts for the oxygen evolution reaction.

No description available.

Carbon dioxide (CO2), as a greenhouse gas, has shown to achieve the highest level in history, causes the global warming issue, leading to a 1.2 ℃ increase of the global average temperature. The consumption of fossil fuels is one of the main reasons that cause CO2 emission. Current industrial production of chemicals accounts for 29% of total fossil fuels consumption, which can be the feedstock or raw materials for carbon source, or act as the fuel to generate heat and power. CO2 conversion technologies, e.g., thermo-catalytic reaction and electrochemical reduction, have drawn researchers' attention, since they have the potential to resolve the feedstock and fuel consumption sectors of chemical production at the same time. CO2 conversion technologies use CO2 as the direct carbon source of chemicals and store the intermittent renewable energies as the energy source, which can ultimately achieve a net-zero CO2 emission and produce value-added chemical products. However, there are challenges for a practical application of CO2 conversion technologies. For instance, electrochemical CO2 reduction reaction (ECO2RR) suffers from the low activity and selectivity, while thermocatalytic CO2 conversion, or the CO2 hydrogenation reaction, usually requires harsh reaction conditions and has a low selectivity. Nonetheless, the improvement of developing new promising catalysts remains limited, due to the lack of insights of the reactions. The complex reaction networks and kinetics lead to an elusive reaction mechanism, and various effects, e.g., solvation, potential, structure, and coverage, hinder our fundamental understanding of catalytic processes. Herein, we report the efforts that we have been put in to gain insights of reaction mechanism of CO2 reduction reactions. Bi has shown to reduce CO2 to formic acid (HCOOH), while we have found that, by constituting a Bi-Cu2S heterostructure catalyst, a better catalytic performance was achieved, due to the structural effect of the interface (Chapter 2). However, it is shown that the CO2 electrochemical reduction mechanism on Bi has changed when switching the electrolyte from water to aprotic media, e.g., ionic liquids, and CO was obtained as the main product instead of HCOOH, showing a shift of reaction pathway due to the electrolyte effect (Chapter 3). However, the fundamental understanding of reaction mechanism requires not only the reaction pathways, but the reaction kinetics under reaction conditions, where the lateral or adsorbate-adsorbate interactions play an important role. In this case, we summarized recent advances of applications of machine learning (ML) algorithms for adsorbate-adsorbate interaction model developments to deal with the realistic reaction kinetics (Chapter 4). The lattice based Kinetic Monte Carlo (KMC) has shown promising performances for considering the lateral interactions of surface reactions. We report the mechanistic and KMC kinetic study of CO2 hydrogenation on Cesium promoted Au(111) surface, to gain insights of alkali metal promoting effects under reaction conditions (Chapter 5). To expand the scope, the integration of CO2 reduction with the C-N bond formation provides a promising strategy to produce more value-added product such as urea. Recent studies show that urea can be produced by reducing CO2 and nitrate (NO3-) from wastewater, which mitigate both global warming and nitrate pollution issue. However, the reaction mechanism remains elusive due to the complicated reaction network. Therefore, we employed the first-principles molecular dynamics to reveal the reaction mechanism of C-N coupling and the effect of different reaction conditions including applied potential and electrolyte (Chapter 6). Although recent advances in the computational catalysis field have significantly push forward the understanding of the chemistry nature of heterogeneous catalysis, the gap between theory and experiment remains far beyond bridged due to the complexity nature of the problem in a wide range of time and length scales, hinders the development and discovery of active catalytic materials. Recent advances of narrowing and bridging the complexity gap between theory and experiment with machine learning have been summarized to emphasize the importance of utilizing machine learning for rational catalyst design (Chapter 7). / Doctor of Philosophy / Global warming issue is a rising topic in recent years which has severe impacts on environments. One of the main reasons is the increase level of greenhouse gases that prevent the release of heat that captured from the sun. Carbon dioxide (CO2) is achieving the highest level in history due to the human activities including the consumption of fossil fuels. Therefore, CO2 conversion technologies are needed to tackle reduce the CO2 level in the atmosphere and the emission of CO2 in industries. CO2 conversion technologies, e.g., thermo-catalytic reaction and electrochemical reduction, have drawn researchers' attention, since they have the potential to resolve the feedstock and fuel consumption sectors of chemical production at the same time. However, the complexity of the CO2 conversion processes hinders the development of new technologies. Since the nature of these technologies are heterogeneous catalytic reactions, all reactions are happening at the interface between catalysts and reactants/products, which calls for the understanding of interfacial mechanisms of CO2 reduction reactions. For this type of high degree of freedom problem where many phases including solid-solid, solid-liquid, and solid-gas phases exist, multiscale simulations turn out to be a proper approach since the wide time and length scale that can be covered. Herein, we employed different multiscale modeling methods to tackle various CO2 reduction problems. For electrochemical reduction of CO2, we designed a novel Bi-Cu2S hetero-structured catalyst, which has abundant interfacial sites between Bi and Cu2S, demonstrating the improved catalytic performance of ECO2RR toward formate production. At the same time, it has been found that in non-aqueous solution, the reaction pathway has been switched, where CO is obtained as the final product instead of formate. This effect has been investigated using constant potential calculation method to probe the reaction under reaction condition. For thermo-catalytic reactions, we studied the CO2 hydrogenation on Cesium promoted Au(111) surface using quantum mechanics and kinetic Monte Carlo (KMC) calculations, to gain insights of alkali metal promoting effects under reaction conditions. To expand the scope, the integration of CO2 electroreduction with C-N coupling is a promising strategy for global warming and pollution control, which utilizes the nitrate (NO3-) from wastewater and CO2 to produce high value-added product such as urea. The fundamental investigation of reaction mechanism of C-N coupling has been studied using first principles molecular dynamics.

A major contribution to global warming is CO2 emitted from the combustion of fossil fuels. Electrochemical processes can help to mitigate the elevated CO2 emissions through either the conversion of CO2 into value-added chemicals or the replacement of fossil fuels with clean fuels such as hydrogen produced from water oxidation. The present dissertation focuses on the mechanistic aspects of electrochemical processes. Electrochemical water oxidation is hindered by the low efficiency of oxygen evolution reaction (OER) at the anode whereas electrochemical reduction of CO2 (ERCO2) is hampered by high overpotentials and poor product selectivity. In this dissertation, we studied the catalytic activity of transition metal-based catalysts, including FeNi spinels, metal-oxide/copper, and d metal cyclam complexes, for both OER and ERCO2 using the density functional theory (DFT) computational approach.We report a combined effort of fabricating FeNi oxide catalysts and identifying the active component of the catalyst for OER. Our collaborators at the University of California, Santa Cruze fabricated a series of FeNi spinels-based materials including Ni(OH)Fe2O4(Cl), Ni(OH)Fe2O4, Fe(OH)Fe2O4(Cl), Fe(OH)Fe2O4, Ni(OH)O(Cl), Ni(OH)O and some show exceptional activity for OER. Combined experimental characterization and computational mechanistic study based on the computational hydrogen electrode (CHE) model revealed that Ni(OH)Fe2O4(Cl) is the active ensemble for exceptional OER performance. We also investigated CO2 reduction to C1 products at the metal-oxide/copper interfaces ((MO)4/Cu(100), M = Fe, Co and Ni) based on the CHE model. The effect of tuning metal-oxide/copper interfaces on product selectivity and limiting potential was clearly demonstrated. This study showed that the catalyst/electrode interface and solvent can be regulated to optimize product selectivity and lower the limiting potential for ERCO2. Applied potential affects the stability of species on the surface of the electrode. The proton-coupled electron transfer (PCET) equilibrium assumed in the CHE model does not capture the change in free energy under the influence of the applied potential. In contrast, the constant electrode potential (CEP) model captures changes in free energy due to applied potential, we applied the CEP model to ERCO2 and OER on (MO)4/Cu(100) and compared the results with those from the CHE model. The results demonstrate that the CHE and the CEP models predict different limiting potentials and product selectivity for ERCO2, but they predict similar limiting potentials for OER. The results demonstrate the importance of accounting for the applied potential effect in the study of more complex multi-step electrochemical processes. We also studied transition metal-based homogeneous catalysts for ERCO2. We examined the performance of transition metal(M) - cyclam(L) complexes as molecular catalysts for the reduction of CO2 to HCOO- and CO, focusing on the effect of changing the metal ions in cyclam on product selectivity (either HCOO- or CO), limiting potential and competitive hydrogen evolution reaction. Our results show that among the complexes, [LNi]2+ and [LPd]2+ can catalyze CO2 reduction to CO, and [LMo]2+ and [LW]3+ can reduce CO2 to HCOO-. Notably, [LMo]2+, [LW]3+, [LW]2+ and [LCo]2+ have a limiting potential less negative than -1.6 V and are based on earth-abundant elements, making them attractive for practical application. In summary, the dissertation demonstrates high-performance catalysts can be designed from earth-abundant transition metals for electrochemical processes that would alleviate the high CO2 level in the environment. On the other hand, completely reversing the increasing trend of CO2 level in the atmosphere requires a collective human effort.

Made available in DSpace on 2017-07-10T15:14:40Z (GMT). No. of bitstreams: 1 Glaucio_J_Gomes.pdf: 2760994 bytes, checksum: ed1586b3de70d1981cd1f4c6b68eb593 (MD5) Previous issue date: 2016-04-07 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior / The application of zeolites in catalytic processes is of great interest in several chemical reactions that involves transformation of biomass into higher value-added products. In a similar sense, conversion of fatty acids into esters catalyzed by Hzeolites have sprouted a widely-varied academic and industrial attention due to it providing a sustainable route inside the Green Chemistry for the production of chemicals and energy, of interest in biodiesel synthesis. Thermogravimetric analysis (TGA-IR), FTIR and FTIR-ATR spectroscopy, and theoretical studies [DFT method at M06-2X/6-31G(D) level] as support of experimental data are employed to investigate the adsorption of methanol and acetic acid as reaction models of fatty acids on the surface of H-Beta and H-ZSM-5 zeolites. The results obtained by TGA-IR for adsorption of acetic acid and methanol in both zeolites show different processes of mass loss that correlated with the adsorption mode (physical and chemical adsorption). The spectra obtained by FTIR-ATR show that acetic acid is molecularly adsorbed on H-Beta through the carbonyl group on the Brønsted site and, by the hydroxyl group on the Lewis site, the experimental evidence is supported by the theoretical vibrational analysis. The acetic acid adsorption by C=O on the Brønsted site is more stable than other proposals of adsorption found in the literature. For HZSM- 5 zeolite, FTIR-ATR spectrum reported changes in the structural composition of the material, therefore it was not possible to compare theoretical and experimental results. Moreover, different models of adsorption/co-adsorption of acetic acid and methanol on the surface of H-zeolites have been studied by theoretical calculations in order to provide information that could be helpful in the interpretation of the first step of the mechanism of acetic acid esterification with methanol on H-Beta and HZSM- 5 zeolites. / A aplicação de zeólitas em processos catalíticos é de grande interesse em inúmeras reações químicas para transformação de biomassa em produtos de maior valor agregado. Neste sentido a conversão de ácidos graxos em ésteres catalisadas por H-zeólitas recentemente ganha grande atenção na academia e na indústria uma vez que fornece uma rota sustentável dentro da Química Verde destinada à produção de produtos químicos e energia, de interesse na produção de biodiesel. Assim, a reação de esterificação por catálise heterogênea é complexa envolvendo inúmeros passos reacionais na superfície que podem ocorrer simultaneamente sendo de interesse também compreender os passos elementares envolvidos nesta reação. Nesta dissertação buscou-se compreender o processo de adsorção de metanol e ácido acético como modelos de reação para ácidos graxos sobre a superfície da zeólita H-Beta e H-ZSM-5 por termogravimetria (TGA-IV), espectroscopia (FTIR e FTIR-ATR) e a aplicação de estudos teóricos [cálculos DFT a nível M06-2X/6- 31G(D)] como suporte aos dados experimentais. Os resultados encontrados para ambas as zeólitas por TGA-IV permitiram estudar a adsorção de ácido acético e metanol, sendo identificados diferentes processos de perda de massa correlacionados com a forma de adsorção (adsorção física e química). Os espectros obtidos por FTIR-ATR nos indicaram que o ácido acético se encontra molecularmente adsorvido sobre H-Beta, a partir do grupo carbonila com os sítios de Brønsted e o grupo hidroxila com o sítio de Lewis, é comprovada pelas frequências vibracionais do modelo teórico. A adsorção de ácido acético por C=O sobre o sítio de Brønsted é mais estável que outras propostas encontradas na literatura. Para a zeólita H-ZSM-5 os espectros de FTIR-ATR informaram alterações na composição estrutural do material o qual não foi possível realizar comparações com estudo teórico. Por outro lado os cálculos teóricos puderam prever diferentes modelos de adsorção/co-adsorção de ácido acético e metanol sobre a superfície de H-zeólitas a fim de proporcionar informações que podem auxiliar a interpretação do primeiro passo da reação de esterificação ácido acético e metanol em H-Beta e H-ZSM-5.