Browsing by Author "Xin, Hongliang"

Now showing 1 - 20 of 36
  • Thumbnail Image
    Accelerating Catalyst Discovery via Ab Initio Machine Learning
    Li, Zheng (Virginia Tech, 2019-12-03)
    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.
  • Thumbnail Image
    Accelerating Catalytic Materials Discovery for Sustainable Nitrogen Transformations by Interpretable Machine Learning
    Pillai, Hemanth Somarajan (Virginia Tech, 2023-01-12)
    Computational chemistry and machine learning approaches are combined to understand the mechanisms, derive activity trends, and ultimately to search for active electrocatalysts for the electrochemical oxidation of ammonia (AOR) and nitrate reduction (NO3RR). Both re- actions play vital roles within the nitrogen cycle and have important applications within tackling current environmental issues. Mechanisms are studied through the use of density functional theory (DFT) for AOR and NO3RR, subsequently a descriptor based approach is used to understand activity trends on a wide range of electrocatalysts. For AOR inter- pretable machine learning is used in conjunction with active learning to screen for active and stable ternary electrocatalysts. We find Pt3RuCo, Pt3RuNi and Pt3RuFe show great activity, and are further validated via experimental results. By leveraging the advantages of the interpretible machine learning model we elucidate the underlying electronic factors for the stronger *N binding which leads to the observed improved activity. For NO3RR an interpretible machine learning model is used to understand ways to bypass the stringent limitations put on the electrocatalytic activity due to the *N vs *NO3 scaling relations. It is found that the *N binding energy can be tuned while leaving the *NO3 binding energy unaffected by ensuring that the subsurface atom interacts strongly with the *N. Based on this analysis we suggest the B2 CuPd as a potential active electrocatalyst for this reaction, which is further validated by experiments
  • Thumbnail Image
    Algorithm-derived feature representations for explainable AI in catalysis
    Omidvar, Noushin; Xin, Hongliang (Elsevier, 2021-12-01)
    Machine learning (ML) has emerged as a critical tool in catalysis, attributed to its capability of finding complex patterns in high dimensional and heterogeneous data. A recently published article in Chem Catalysis (Esterhuizen et al.) used unsupervised ML for uncovering electronic and geometric descriptors of the surface reactivity of metal alloys and oxides.
  • Thumbnail Image
    Ambient ammonia synthesis via palladium-catalyzed electrohydrogenation of dinitrogen at low overpotential
    Wang, Jun; Yu, Liang; Hu, Lin; Chen, Gang; Xin, Hongliang; Feng, Xiaofeng (Springer Nature, 2018-05-15)
    Electrochemical reduction of N2 to NH3 provides an alternative to the Haber−Bosch process for sustainable, distributed production of NH3 when powered by renewable electricity. However, the development of such process has been impeded by the lack of efficient electrocatalysts for N2 reduction. Here we report efficient electroreduction of N2 to NH3 on palladium nanoparticles in phosphate buffer solution under ambient conditions, which exhibits high activity and selectivity with an NH3 yield rate of ~4.5 μg mg−1Pd h−1 and a Faradaic efficiency of 8.2% at 0.1 V vs. the reversible hydrogen electrode (corresponding to a low overpotential of 56 mV), outperforming other catalysts including gold and platinum. Density functional theory calculations suggest that the unique activity of palladium originates from its balanced hydrogen evolution activity and the Grotthuss-like hydride transfer mechanism on α-palladium hydride that lowers the free energy barrier of N2 hydrogenation to *N2H, the rate-limiting step for NH3 electrosynthesis.
  • Thumbnail Image
    Assessing Structure–Property Relationships of Crystal Materials using Deep Learning
    Li, Zheng (Virginia Tech, 2020-08-05)
    In recent years, deep learning technologies have received huge attention and interest in the field of high-performance material design. This is primarily because deep learning algorithms in nature have huge advantages over the conventional machine learning models in processing massive amounts of unstructured data with high performance. Besides, deep learning models are capable of recognizing the hidden patterns among unstructured data in an automatic fashion without relying on excessive human domain knowledge. Nevertheless, constructing a robust deep learning model for assessing materials' structure-property relationships remains a non-trivial task due to highly flexible model architecture and the challenge of selecting appropriate material representation methods. In this regard, we develop advanced deep-learning models and implement them for predicting the quantum-chemical calculated properties (i.e., formation energy) for an enormous number of crystal systems. Chapter 1 briefly introduces some fundamental theory of deep learning models (i.e., CNN, GNN) and advanced analysis methods (i.e., saliency map). In Chapter 2, the convolutional neural network (CNN) model is established to find the correlation between the physically intuitive partial electronic density of state (PDOS) and the formation energies of crystals. Importantly, advanced machine learning analysis methods (i.e., salience mapping analysis) are utilized to shed light on underlying physical factors governing the energy properties. In Chapter 3, we introduce the methodology of implementing the cutting-edge graph neural networks (GNN) models for learning an enormous number of crystal structures for the desired properties.
  • Thumbnail Image
    Bayesian learning of chemisorption for bridging the complexity of electronic descriptors
    Wang, Siwen; Pillai, Hemanth Somarajan; Xin, Hongliang (Springer Nature, 2020)
    Building upon the d-band reactivity theory in surface chemistry and catalysis, we develop a Bayesian learning approach to probing chemisorption processes at atomically tailored metal sites. With representative species, e.g., *O and *OH, Bayesian models trained with ab initio adsorption properties of transition metals predict site reactivity at a diverse range of intermetallics and near-surface alloys while naturally providing uncertainty quantification from posterior sampling. More importantly, this conceptual framework sheds light on the orbitalwise nature of chemical bonding at adsorption sites with d-states characteristics ranging from bulk-like semi-elliptic bands to free-atom-like discrete energy levels, bridging the complexity of electronic descriptors for the prediction of novel catalytic materials.
  • Thumbnail Image
    Bayesian-optimization-assisted discovery of stereoselective aluminum complexes for ring-opening polymerization of racemic lactide
    Wang, Xiaoqian; Huang, Yang; Xie, Xiaoyu; Liu, Yan; Huo, Ziyu; Lin, Maverick; Xin, Hongliang; Tong, Rong (Nature Research, 2023-06-20)
    Stereoselective ring-opening polymerization catalysts are used to produce degradable stereoregular poly(lactic acids) with thermal and mechanical properties that are superior to those of atactic polymers. However, the process of discovering highly stereoselective catalysts is still largely empirical.We aim to develop an integrated computational and experimental framework for efficient, predictive catalyst selection and optimization. As a proof of principle, we have developed a Bayesian optimization workflow on a subset of literature results for stereoselective lactide ring-opening polymerization, and using the algorithm, we identify multiple new Al complexes that catalyze either isoselective or heteroselective polymerization. In addition, feature attribution analysis uncovers mechanistically meaningful ligand descriptors, such as percent buried volume (%Vbur) and the highest occupied molecular orbital energy (Eₕₒₘₒ), that can access quantitative and predictivemodels for catalyst development.
  • Thumbnail Image
    Breaking adsorption-energy scaling limitations of electrocatalytic nitrate reduction on intermetallic CuPd nanocubes by machine-learned insights
    Gao, Qiang; Pillai, Hemanth Somarajan; Huang, Yang; Liu, Shikai; Mu, Qingmin; Han, Xue; Yan, Zihao; Zhou, Hua; He, Qian; Xin, Hongliang; Zhu, Huiyuan (Nature Portfolio, 2022-04-29)
    Machine learning is a powerful tool for screening electrocatalytic materials. Here, the authors reported a seamless integration of machine-learned physical insights with the controlled synthesis of structurally ordered intermetallic nanocrystals and well-defined catalytic sites for efficient nitrate reduction to ammonia. The electrochemical nitrate reduction reaction (NO3RR) to ammonia is an essential step toward restoring the globally disrupted nitrogen cycle. In search of highly efficient electrocatalysts, tailoring catalytic sites with ligand and strain effects in random alloys is a common approach but remains limited due to the ubiquitous energy-scaling relations. With interpretable machine learning, we unravel a mechanism of breaking adsorption-energy scaling relations through the site-specific Pauli repulsion interactions of the metal d-states with adsorbate frontier orbitals. The non-scaling behavior can be realized on (100)-type sites of ordered B2 intermetallics, in which the orbital overlap between the hollow *N and subsurface metal atoms is significant while the bridge-bidentate *NO3 is not directly affected. Among those intermetallics predicted, we synthesize monodisperse ordered B2 CuPd nanocubes that demonstrate high performance for NO3RR to ammonia with a Faradaic efficiency of 92.5% at -0.5 V-RHE and a yield rate of 6.25 mol h(-1) g(-1) at -0.6 V-RHE. This study provides machine-learned design rules besides the d-band center metrics, paving the path toward data-driven discovery of catalytic materials beyond linear scaling limitations.
  • Thumbnail Image
    Computational and Data-Driven Design of Perturbed Metal Sites for Catalytic Transformations
    Huang, Yang (Virginia Tech, 2024-05-23)
    We integrate theoretical, computational and data-driven approaches for the sake of understanding, design and discovery of metal based catalysts. Firstly, we develop theoretical frameworks for predicting electronic descriptors of transition and noble metal alloys, including a physics model of d-band center, and a tight-binding theory of d-band moments to systematically elucidate the distinct electronic structures of novel catalysts. Within this framework, the hybridization of semi-empirical theories with graph neural network and attribution analysis enables accurate prediction equipped with mechanistic insights. In addition, novel physics effect controlling surface reactivity beyond conventional understanding is uncovered. Secondly, we develop a computational and data-driven framework to model high entropy alloy (HEA) catalysis, incorporating thermodynamic descriptor-based phase stability evaluation, surface segregation modeling by deep learning potential-driven molecular simulation and activity prediction through machine learning-embedded electrokinetic model. With this framework, we successfully elucidate the experimentally observed improved activity of PtPdCuNiCo HEA in oxygen reduction reaction. Thirdly, a Bayesian optimization framework is employed to optimize racemic lactide polymerization by searching for stereoselective aluminum (Al) -complex catalysts. We identified multiple new Al-complex molecules that catalyzed either isoselective or heteroselective polymerization. In addition, feature attribution analysis uncovered mechanistically meaningful ligand descriptors that can access quantitative and predictive models for catalyst development.
  • Thumbnail Image
    Electrochemical Carbon Dioxide Reduction for Renewable Carbonaceous Fuels and Chemicals
    Han, Xue (Virginia Tech, 2023-03-15)
    Electrochemical CO2 reduction reaction (ECO2RR) powered by renewable electricity possesses the potential to store intermittent energy in chemical bonds while producing sustainable chemicals and fuels. Unfortunately, it is hard to achieve low overpotential, high selectivity, and activity simultaneously of ECO2RR. Developing efficient electrocatalysts is the most promising strategy to enhance electrocatalytic activity in CO2 reduction. Herein, we designed novel Bi-Cu2S heterostructures by a one-pot wet-chemistry method. The epitaxial growth of Cu2S on Bi results in abundant interfacial sites and these heterostructured nanocrystals demonstrated high electrocatalytic performance of ECO2RR with high current density, largely reduced overpotential, near-unity FE for formate production (Chapter 2). Meanwhile, we see a lot of opportunities for catalysis in a confined space due to their tunable microenvironment and active sites on the surface, leading to a broad spectrum of electrochemical conversion schemes. Herein, we reveal fundamental concepts of confined catalysis by summarizing recent experimental investigations. We mainly focus on carbon nanotubes (CNTs) encapsulated metal-based materials and summarize their applications in emerging electrochemical reactions, including ECO2RR and more (Chapter 3). Although we were able to obtain high activity and selectivity toward C1 products, it is more attractive to go beyond C1 chemicals to produce C2 products due to their high industrial value. Herein, we designed Ag-modified Cu alloy catalysts that can create a CO-rich local environment for enhancing C-C coupling on Cu for C2 formation. Moreover, Ag corporate in Cu can chemically improve the structural stability of Cu lattice. (Chapter 4) Nevertheless, advanced electrocatalytic platforms cannot be developed without a fundamental understanding of binding configurations of the surface-adsorbed intermediates and adsorbate-adsorbate interaction on the local environment in electrochemical CO2 reduction. In this case, we make discussions of recent developments of machine learning based models of adsorbate-adsorbate interactions, including the oversimplified linear analytic relationships, the cluster expansion models parameterized by machine learning algorithms, and the highly nonlinear deep learning models. We also discuss the challenges of the field, particularly overcoming the limitations of pure data driven models with the integration of computational theory and machine learning of lateral interactions for catalyst materials design. (Chapter 5).
  • Thumbnail Image
    Experimental Adsorption and Reaction Studies on Transition Metal Oxides Compared to DFT Simulations
    Chen, Han (Virginia Tech, 2021-06-11)
    A temperature-programmed desorption (TPD) study of CO and NH₃ adsorption on MnO(100) with complimentary density functional theory (DFT) simulations was conducted. TPD reveals a primary CO desorption signal at 130 K from MnO(100) in the low coverage limit giving an adsorption energy of -35.6 ±2.1 kJ/mol on terrace sites. PBE+U gives a more reasonable structural result than PBE, and the adsorption energy obtained by PBE+U and DFT-D3 Becke-Johnson gives excellent agreement with the experimentally obtained ΔEads for adsorption at Mn²⁺ terrace sites. The analysis of NH₃-TPD traces revealed that adsorption energy on MnO(100) is coverage-dependent. At the low-coverage limit, the adsorption energy on terraces is -58.7±1.0 kJ/mol. A doser results in the formation of a transient NH₃ multilayers that appears in TPD at around 110K. For a terrace site, PBE+U predicts a more realistic surface adsorbate geometry than PBE does, with PBE+U with Tkatchenko-Scheffler method with iterative Hirshfeld partitioning (TSHP) provides the best prediction. DFT simulations of the dehydrogenation elementary step of the ethyl and methyl fragments on α-Cr2O₃(101̅2) were also conducted to complement previous TPD studies of these subjects. On the nearly-stoichiometric surface of α-Cr₂O₃(101̅2), CD₃₋ undergoes dehydrogenation to produce CD₂=CD₂ and CD₄. Previous TPD traces suggest that the α-hydrogen (α-H) elimination of methyl groups on α-Cr₂O₃(101̅2) is the rate-limiting step, and has an activation barrier of 135±2 kJ/mol. DFT simulations showed that PBE gives reasonable prediction of the adsorption sites for CH3- fragments in accordance with XPS spectra, while PBE+U did not. Both PBE and PBE+U failed to predict the correct adsorption sites for CH₂=. When the simulation is set in accordance with the experimentally observed adsorption sites for the carbon species, PBE gives very accurate prediction on the reaction barrier when an adjacent I adatom is present, while PBE+U failed spectacularly. When the simulation is set in accordance with the DFT-predicted adsorption sites, PBE is still able to accurately predict the reaction barrier (<1% to 8.7% error) while PBE+U is less accurate. DFT is also used to complement the previous study of the β-H elimination an ethyl group on the α-Cr₂O₃(101̅2) surface. The DFT simulation shows that absent surface Cl adatoms, PBE predicts an activation barrier of 92.6 kJ/mol, underpredicting the experimental activation barrier by 28.7%, while PBE+U predicts a barrier of 27.0 kJ/mol, under-predicting the experimental barrier by 79.2%. The addition of chlorine on the adjacent cation improved the prediction on barrier by PBE+U marginally, while worsened the prediction by PBE marginally. Grant information: Financial support provided by the U.S. Department of Energy through grant DE-FG02 97ER14751.
  • Thumbnail Image
    Exploring Strategies to Break Adsorption-Energy Scaling Relations in Catalytic CO Oxidation
    Wang, Jiamin (Virginia Tech, 2020-01-21)
    An atomistic control of chemical bonds formation and cleavage holds the key to making molecular transformations more energy efficient and product selective. However, inherent scaling relations among binding strengths of adsorbates on various catalytic materials often give rise to volcano-shaped relationships between the catalytic activity and the affinity of critical intermediates to the surface. The optimal catalysts should bind the reactants 'just right', i.e., neither too strong nor too weak, which is the Sabatier's principle. It is extremely useful for searching promising catalysts, but also imposes serious constraints on design flexibility. Therefore, how to circumvent scaling constraints is crucial for advancing catalytic science. It has been shown that hot electrons can selectively activate the chemical bonds that are not responsive to phonon excitation, thus providing a rational approach beyond scaling limitation. Another emerging yet effective way to break the scaling constraint is single atom catalysis. Strong interactions of supported single atoms with supports dramatically affect the electronic structure of active sites, which reroutes mechanistic pathways of surface reactions. In my PhD research, we use CO oxidation reaction on metal-based active sites as a benchmark system to tailor mechanistic pathways through those two strategies 1) ultra-fast laser induced nonadiabatic surface chemistry and 2) oxide-supported single metal catalysis, with the aim to go beyond the Sabatier activity volcano in metal catalysis.
  • Thumbnail Image
    Infusing theory into deep learning for interpretable reactivity prediction
    Wang, Shih-Han; Pillai, Hemanth Somarajan; Wang, Siwen; Achenie, Luke E. K.; Xin, Hongliang (Nature Research, 2021)
    Despite recent advances of data acquisition and algorithms development, machine learning (ML) faces tremendous challenges to being adopted in practical catalyst design, largely due to its limited generalizability and poor explainability. Herein, we develop a theory-infused neural network (TinNet) approach that integrates deep learning algorithms with the wellestablished d-band theory of chemisorption for reactivity prediction of transition-metal surfaces. With simple adsorbates (e.g., *OH, *O, and *N) at active site ensembles as representative descriptor species, we demonstrate that the TinNet is on par with purely data-driven ML methods in prediction performance while being inherently interpretable. Incorporation of scientific knowledge of physical interactions into learning from data sheds further light on the nature of chemical bonding and opens up new avenues for ML discovery of novel motifs with desired catalytic properties.
  • Thumbnail Image
    Integrated Process Modeling and Data Analytics for Optimizing Polyolefin Manufacturing
    Sharma, Niket (Virginia Tech, 2021-11-19)
    Polyolefins are one of the most widely used commodity polymers with applications in films, packaging and automotive industry. The modeling of polymerization processes producing polyolefins, including high-density polyethylene (HDPE), polypropylene (PP), and linear low-density polyethylene (LLDPE) using Ziegler-Natta catalysts with multiple active sites, is a complex and challenging task. In our study, we integrate process modeling and data analytics for improving and optimizing polyolefin manufacturing processes. Most of the current literature on polyolefin modeling does not consider all of the commercially important production targets when quantifying the relevant polymerization reactions and their kinetic parameters based on measurable plant data. We develop an effective methodology to estimate kinetic parameters that have the most significant impacts on specific production targets, and to develop the kinetics using all commercially important production targets validated over industrial polyolefin processes. We showcase the utility of dynamic models for efficient grade transition in polyolefin processes. We also use the dynamic models for inferential control of polymer processes. Thus, we showcase the methodology for making first-principle polyolefin process models which are scientifically consistent, but tend to be less accurate due to many modeling assumptions in a complex system. Data analytics and machine learning (ML) have been applied in the chemical process industry for accurate predictions for data-based soft sensors and process monitoring/control. Specifically, for polymer processes, they are very useful since the polymer quality measurements like polymer melt index, molecular weight etc. are usually less frequent compared to the continuous process variable measurements. We showcase the use of predictive machine learning models like neural networks for predicting polymer quality indicators and demonstrate the utility of causal models like partial least squares to study the causal effect of the process parameters on the polymer quality variables. ML models produce accurate results can over-fit the data and also produce scientifically inconsistent results beyond the operating data range. Thus, it is growingly important to develop hybrid models combining data-based ML models and first-principle models. We present a broad perspective of hybrid process modeling and optimization combining the scientific knowledge and data analytics in bioprocessing and chemical engineering with a science-guided machine learning (SGML) approach and not just the direct combinations of first-principle and ML models. We present a detailed review of scientific literature relating to the hybrid SGML approach, and propose a systematic classification of hybrid SGML models according to their methodology and objective. We identify the themes and methodologies which have not been explored much in chemical engineering applications, like the use of scientific knowledge to help improve the ML model architecture and learning process for more scientifically consistent solutions. We apply these hybrid SGML techniques to industrial polyolefin processes such as inverse modeling, science guided loss and many others which have not been applied previously to such polymer applications.
  • Thumbnail Image
    Interpretable Machine Learning of Chemical Bonding at Solid Surfaces
    Omidvar, Noushin; Pillai, Hemanth Somarajan; Wang, Shih-Han; Mou, Tianyou; Wang, Siwen; Athawale, Andy; Achenie, Luke E. K.; Xin, Hongliang (American Chemical Society, 2021-11-25)
    Understanding the nature of chemical bonding and its variation in strength across physically tunable factors is important for the development of novel catalytic materials. One way to speed up this process is to employ machine learning (ML) algorithms with online data repositories curated from high-throughput experiments or quantum-chemical simulations. Despite the reasonable predictive performance of ML models for predicting reactivity properties of solid surfaces, the ever-growing complexity of modern algorithms, e.g., deep learning, makes them black boxes with little to no explanation. In this Perspective, we discuss recent advances of interpretable ML for opening up these black boxes from the standpoints of feature engineering, algorithm development, and post hoc analysis. We underline the pivotal role of interpretability as the foundation of next-generation ML algorithms and emerging AI platforms for driving discoveries across scientific disciplines.
  • Thumbnail Image
    Investigating Cathode–Electrolyte Interfacial Degradation Mechanism to Enhance the Performance of Rechargeable Aqueous Batteries
    Zhang, Yuxin (Virginia Tech, 2023-12-04)
    The invention of Li-ion batteries (LIBs) marks a new era of energy storage and allows for the large-scale industrialization of electric vehicles. However, the flammable organic electrolyte in LIBs raises significant safety concerns and has resulted in numerous fires and explosion accidents. In the pursuit of more reliable and stable battery solutions, interests in aqueous batteries composed of high-energy cathodes and water-based electrolytes are surging. Limited by the narrow electrochemical stability window (ESW) of water, conventional aqueous batteries only achieve inferior energy densities. Current development mainly focuses on manipulating the properties of aqueous electrolytes through introducing excessive salts or secondary solvents, which enables an unprecedentedly broad ESW and more selections of electrode materials while also resulting in some compromises. On the other hand, the interaction between electrodes and aqueous electrolytes and associated electrode failure mechanism, as the key factors that govern cell performance, are of vital importance yet not fully understood. Owing to the high-temperature calcination synthesis, most electrode materials are intrinsically moisture-free and sensitive to the water-rich environment. Therefore, compared to the degradation behaviors in conventional LIBs, such as cracking and structure collapse, the electrode may suffer more severe damage during cycling and lead to rapid capacity decay. Herein, we adopted multi-scale characterization techniques to identify the failure modes at cathode–electrolyte interface and provide strategies for improving the cell capacity and life during prolonged cycling. In Chapter 1, we first provide a background introduction of conventional non-aqueous and aqueous batteries. We then show the current development of modern aqueous batteries through electrolyte modification and their merits and drawbacks. Finally, we present typical electrode failure mechanism in non-aqueous electrolytes and discuss how water can further impact the degradation behaviors. In Chapter 2, we prepare three types of aqueous electrolytes and systematically evaluate the electrochemical performance of LiNixMnyCo1-x-yO2, LiMn2O4 and LiFePO4 in the aqueous electrolytes. Combing surface- and bulk-sensitive techniques, we identify the roles played by surface exfoliation, structure degradation, transition metal dissolution and interface formation in terms of the capacity decay in different cathode materials. We also provide fundamental insights into the materials selection and electrolyte design in the aqueous batteries. In Chapter 3, we select LiMn2O4 as the material platform to study the transition metal dissolution behavior. Relying on the spatially resolved X-ray fluorescence microscopy, we discover a voltage-dependent Mn dissolution/redeposition (D/R) process during electrochemical cycling, which is confirmed to be related to the Jahn–Teller distortion and surface reconstruction at different voltages. Inspired by the findings, we propose an approach to stabilize the material performance through coating sulfonated tetrafluoroethylene (i.e., Nafion) on the particle, which can regulate the proton diffusion and Mn dissolution behavior. Our study discovers the dynamic Mn D/R process and highlights the impact of coating strategy in the performance of aqueous batteries. In Chapter 4, we investigate the diffusion layer formed by transition metals at the electrode–electrolyte interface. With the help of customized cells and XFM technique, we successfully track the spatiotemporal evolution of the diffusion layer during soaking and electrochemical cycling. The thickness of diffusion layer is determined to be at micron level, which can be readily diminished when gas is generated on the electrode surface. Our approach can be further expanded to study the phase transformation and particle agglomeration at the interfacial region and provide insights into the reactive complexes. In Chapter 5, we reveal the correlation between the electrolytic water decomposition and ion intercalation behaviors in aqueous batteries. In the Na-deficient system, we discover that overcharging in the formation process can introduce more cyclable Na ions into the full cell and allows for a boosted performance from 58 mAh/g to 124 mAh/g. The mechanism can be attributed to the water oxidation on the cathode and Na-ion intercalation on the anode when the charging voltage exceeds the normal oxidation potential of cathode. We emphasize the importance of unique formation process in terms of the cell performance and cycle life of aqueous batteries. In Chapter 6, we summarize the results of our work and propose perspectives of future research directions.
  • Thumbnail Image
    Machine Learning and Multivariate Statistics for Optimizing Bioprocessing and Polyolefin Manufacturing
    Agarwal, Aman (Virginia Tech, 2022-01-07)
    Chemical engineers have routinely used computational tools for modeling, optimizing, and debottlenecking chemical processes. Because of the advances in computational science over the past decade, multivariate statistics and machine learning have become an integral part of the computerization of chemical processes. In this research, we look into using multivariate statistics, machine learning tools, and their combinations through a series of case studies including a case with a successful industrial deployment of machine learning models for fermentation. We use both commercially-available software tools, Aspen ProMV and Python, to demonstrate the feasibility of the computational tools. This work demonstrates a novel application of ensemble-based machine learning methods in bioprocessing, particularly for the prediction of different fermenter types in a fermentation process (to allow for successful data integration) and the prediction of the onset of foaming. We apply two ensemble frameworks, Extreme Gradient Boosting (XGBoost) and Random Forest (RF), to build classification and regression models. Excessive foaming can interfere with the mixing of reactants and lead to problems, such as decreasing effective reactor volume, microbial contamination, product loss, and increased reaction time. Physical modeling of foaming is an arduous process as it requires estimation of foam height, which is dynamic in nature and varies for different processes. In addition to foaming prediction, we extend our work to control and prevent foaming by allowing data-driven ad hoc addition of antifoam using exhaust differential pressure as an indicator of foaming. We use large-scale real fermentation data for six different types of sporulating microorganisms to predict foaming over multiple strains of microorganisms and build exploratory time-series driven antifoam profiles for four different fermenter types. In order to successfully predict the antifoam addition from the large-scale multivariate dataset (about half a million instances for 163 batches), we use TPOT (Tree-based Pipeline Optimization Tool), an automated genetic programming algorithm, to find the best pipeline from 600 other pipelines. Our antifoam profiles are able to decrease hourly volume retention by over 53% for a specific fermenter. A decrease in hourly volume retention leads to an increase in fermentation product yield. We also study two different cases associated with the manufacturing of polyolefins, particularly LDPE (low-density polyethylene) and HDPE (high-density polyethylene). Through these cases, we showcase the usage of machine learning and multivariate statistical tools to improve process understanding and enhance the predictive capability for process optimization. By using indirect measurements such as temperature profiles, we demonstrate the viability of such measures in the prediction of polyolefin quality parameters, anomaly detection, and statistical monitoring and control of the chemical processes associated with a LDPE plant. We use dimensionality reduction, visualization tools, and regression analysis to achieve our goals. Using advanced analytical tools and a combination of algorithms such as PCA (Principal Component Analysis), PLS (Partial Least Squares), Random Forest, etc., we identify predictive models that can be used to create inferential schemes. Soft-sensors are widely used for on-line monitoring and real-time prediction of process variables. In one of our cases, we use advanced machine learning algorithms to predict the polymer melt index, which is crucial in determining the product quality of polymers. We use real industrial data from one of the leading chemical engineering companies in the Asia-Pacific region to build a predictive model for a HDPE plant. Lastly, we show an end-to-end workflow for deep learning on both industrial and simulated polyolefin datasets. Thus, using these five cases, we explore the usage of advanced machine learning and multivariate statistical techniques in the optimization of chemical and biochemical processes. The recent advances in computational hardware allow engineers to design such data-driven models, which enhances their capacity to effectively and efficiently monitor and control a process. We showcase that even non-expert chemical engineers can implement such machine learning algorithms with ease using open-source or commercially available software tools.
  • Thumbnail Image
    Machine Learning for Millimeter Wave Wireless Systems: Network Design and Optimization
    Zhang, Qianqian (Virginia Tech, 2021-06-16)
    Next-generation cellular systems will rely on millimeter wave (mmWave) bands to meet the increasing demand for wireless connectivity from end user equipment. Given large available bandwidth and small-sized antenna elements, mmWave frequencies can support high communication rates and facilitate the use of multiple-input-multiple-output (MIMO) techniques to increase the wireless capacity. However, the small wavelength of mmWave yields severe path loss and high channel uncertainty. Meanwhile, using a large number of antenna elements requires a high energy consumption and heavy communication overhead for MIMO transmissions and channel measurement. To facilitate efficient mmWave communications, in this dissertation, the challenges of energy efficiency and communication overhead are addressed. First, the use of unmanned aerial vehicle (UAV), intelligent signal reflector, and device-to-device (D2D) communications are investigated to improve the reliability and energy efficiency of mmWave communications in face of blockage. Next, to reduce the communication overhead, new channel modeling and user localization approaches are developed to facilitate MIMO channel estimation by providing prior knowledge of mmWave links. Using advance mathematical tools from machine learning (ML), game theory, and communication theory, this dissertation develops a suite of novel frameworks using which mmWave communication networks can be reliably deployed and operated in wireless cellular systems, UAV networks, and wearable device networks. For UAV-based wireless communications, a learning framework is developed to predict the cellular data traffic during congestion events, and a new framework for the on-demand deployment of UAVs is proposed to offload the excessive traffic from the ground base stations (BSs) to the UAVs. The results show that the proposed approach enables a dynamical and optimal deployment of UAVs that alleviates the cellular traffic congestion. Subsequently, a novel energy-efficient framework is developed to reflect mmWave signals from a BS towards mobile users using a UAV-carried intelligent reflector (IR). To optimize the location and reflection coefficient of the UAV-carried IR, a deep reinforcement learning (RL) approach is proposed to maximize the downlink transmission capacity. The results show that the RL-based approach significantly improves the downlink line-of-sight probability and increases the achievable data rate. Moreover, the channel estimation challenge for MIMO communications is addressed using a distributional RL approach, while optimizing an IR-aided downlink multi-user communication. The results show that the proposed method captures the statistic feature of MIMO channels, and significantly increases the downlink sum-rate. Moreover, in order to capture the characteristics of air-to-ground channels, a data-driven approach is developed, based on a distributed framework of generative adversarial networks, so that each UAV collects and shares mmWave channel state information (CSI) for cooperative channel modeling. The results show that the proposed algorithm enables an accurate channel modeling for mmWave MIMO communications over a large temporal-spatial domain. Furthermore, the CSI pattern is analyzed via semi-supervised ML tools to localize the wireless devices in the mmWave networks. Finally, to support D2D communications, a novel framework for mmWave multi-hop transmissions is investigated to improve the performance of the high-rate low-latency transmissions between wearable devices. In a nutshell, this dissertation provides analytical foundations on the ML-based performance optimization of mmWave communication systems, and the anticipated results provide rigorous guidelines for effective deployment of mmWave frequency bands into next-generation wireless systems (e.g., 6G).
  • Thumbnail Image
    Measuring the Functionality of Amazon Alexa and Google Home Applications
    Wang, Jiamin (Virginia Tech, 2020-01)
    Voice Personal Assistant (VPA) is a software agent, which can interpret the user's voice commands and respond with appropriate information or action. The users can operate the VPA by voice to complete multiple tasks, such as read the message, order coffee, send an email, check the news, and so on. Although this new technique brings in interesting and useful features, they also pose new privacy and security risks. The current researches have focused on proof-of-concept attacks by pointing out the potential ways of launching the attacks, e.g., craft hidden voice commands to trigger malicious actions without noticing the user, fool the VPA to invoke the wrong applications. However, lacking a comprehensive understanding of the functionality of the skills and its commands prevents us from analyzing the potential threats of these attacks systematically. In this project, we developed convolutional neural networks with active learning and keyword-based approach to investigate the commands according to their capability (information retrieval or action injection) and sensitivity (sensitive or nonsensitive). Through these two levels of analysis, we will provide a complete view of VPA skills, and their susceptibility to the existing attacks.
  • Thumbnail Image
    Modeling of Thermal Transport Properties in Metallic and Oxide Fuels
    Chen, Weiming (Virginia Tech, 2021-08-26)
    Thermal conductivity is a critical fuel performance property not only for current UO2 oxide fuel based light water reactors but also important for next-generation fast reactors that use U-Zr based metallic fuels. In this work, the thermal transport properties of both UO2 based oxide fuels and U-Zr based metallic fuels have been studied. At first, molecular dynamics (MD) simulations were conducted to study the effect of dispersed Xe fission gas atoms on the UO2 thermal conductivity. Numerous studies have demonstrated that xenon (Xe) fission gas plays a major role on fuel thermal conductivity degradation. Even a very low Xe concentration can cause significant thermal conductivity reduction. In this work, the effect of dispersed Xe gas atoms on UO2 thermal conductivity were studied using three different interatomic potentials. It is found that although these potentials result in significant discrepancies in the absolute thermal conductivity values, their normalized values are very similar at a wide range of temperatures and Xe concentrations. By integrating this unified effect into the experimentally measured thermal conductivities, a new analytical model is developed to predict the realistic thermal conductivities of UO2 at different dispersed Xe concentrations and temperatures. Using this new model, the critical Xe concentration that offsets the grain boundary Kapitza resistance effect on the thermal conductivity in a high burnup structure is studied. Next, the mechanisms on how Xe gas bubbles affect the UO2 thermal conductivity have been studied using MD. At a fixed total porosity, the effective thermal conductivity of the bubble-containing UO2 increases with Xe cluster size, then reaches a nearly saturated value at a cluster radius of 0.6 nm, demonstrating that dispersed Xe atoms result in a lower thermal conductivity than clustering them into bubbles. In comparison with empty voids of the same size, Xe-filled bubbles lead to a lower thermal conductivity when the number ratio of Xe atoms to uranium vacancies (Xe:VU ratio) in bubbles is high. Detailed atomic-level analysis shows that the pressure-induced distortion of atoms at bubble surface causes additional phonon scattering and thus further reduces the thermal conductivity. For metallic fuels, temperature gradient and irradiation induced constituent redistribution in U-Zr based fuels cause the variation in fuel composition and the formation of different phases that have different physical properties such as thermal conductivity. In this work, a semi-empirical model is developed to predict the thermal conductivities of U-Zr alloys for the complete composition range and a wide range of temperatures. The model considers the effects of (a) scattering by defects, (b) electron-phonon scattering, and (c) electron-electron scattering. The electronic thermal resistivity models for the two pure components are empirically determined by fitting to the experimental data. A new mixing rule is proposed to predict the average thermal conductivity in U-Zr alloys based on their nominal composition. The thermal conductivity predictions by the new model show good agreement with many available experimental data. In comparison with previous models, the new model has further improvement, in particular for high-U alloys that are relevant to reactor fuel compositions and at the low-temperature regime for the high-Zr alloys. The average thermal conductivity model for the binary U-Zr fuel is also coupled with finite element-based mesoscale modeling technique to calculate the effective thermal conductivities of the U-Zr heterogeneous microstructures. For a U-10wt.%Zr (U-10Zr) fuel at temperatures below the ɑ phase transition temperature, the dominant microstructures are lamellar δ-UZr2 and ɑ-U. Using the mesoscale modeling, the phase boundary thermal resistance R (Kapitza resistance) between δ-UZr2 and ɑ-U has been determined at different temperatures, which shows a T-3 dependence in the temperature range between 300K and 800K. Besides, the Kapitza resistance exhibits a strong dependence on the aspect ratio of the δ-UZr2 phase in the alloying system. An analytical model is therefore developed to correlate the temperature effect and the aspect ratio effect on the Kapitza resistance. Combining the mesoscale modeling with the newly developed Kapitza resistance model, the effective thermal conductivities of many arbitrary δ-UZr2 + ɑ-U heterogeneous systems can be estimated.