Neural-network representations of chemical kinetics

(Georgia Institute of Technology, 2023-12-12) Sabenca Gusmao, Gabriel

High-fidelity microkinetic models (MKMs) have provided a framework for understanding and mathematically representing the elementary processes underpinning catalytic reactions in terms of differential equations. Computational chemistry methods, such as density functional theory, have enabled the estimation of the thermochemistry of elementary reactions on different catalytic materials from first-principles. MKMs constructed around computational chemistry methods have proven useful in determining trends in catalytic activity across different materials. Nevertheless, there are still challenges to overcome: (i) the inclusion of lateral interactions and solvation effects in models leads to over-parameterization, making the mean-field approximation useless and approximating MKMs to kinetic Monte-Carlo; (ii) uncertainties in the structure of the active site and the detailed mechanism, and (iii) non-standardization in the reported thermochemistry models. In this thesis, we introduce a general and unifying algebraic framework that uses singular value decomposition to assess the connectivity of complex reaction networks. Such a framework addresses the standardization issue by leveraging the use of thermochemical data from multiple sources, allowing them to be re-referenced or combined in extended reaction mechanisms. It also generalizes the construction of descriptor-based models by providing a means to quantify the explained variance by each descriptor. With this general algebraic representation of MKM thermochemistry, we set our focus on creating methods to bridge information from transient experiments to high-fidelity MKMs. Physics-informed neural networks (PINNs) have proven to be a suitable mathematical scaffold for solving inverse ordinary (ODE) and partial differential equations (PDE). In this work, we devise an application of the PINNs formulation designed to address inverse kinetics problems, which we call Kinetics-Informed Neural Networks (KINNs). It consists of soft-constrained multi-objective optimization problems that include a hyperparameter that controls the variance between adhering to physical laws and interpolating observed data. We further bridge the statistical formulation of the error probability density in inverse PINNs to frame it in terms of maximum-likelihood estimators (MLE), which allows explicit error propagation from interpolation to the physical model space through Taylor expansion, thereby eliminating the need for hyperparameter tuning. We explore its application to high-dimensional coupled ODEs constrained by differential algebraic equations that are common in transient chemical and biological kinetics. Furthermore, we show that singular-value decomposition (SVD) of the ODE coupling matrices (reaction stoichiometry matrices) provides reduced uncorrelated subspaces in which PINNs solutions can be represented and over which residuals can be projected. Finally, SVD bases serve as preconditioners for the inversion of covariance matrices in this hyperparameter-free robust application of MLE to KINNs, in robust-KINNs (rKINNs). Our exploration extends to applying rKINNs to high-fidelity MKMs constructed from an amalgamation of literature-reported DFT energies relying on the developed unified algebraic thermochemistry framework. We embed domain knowledge into the singular computational perturbation (CSP) approach to satisfy MKM constitutional constraints and generate synthetic data by solving the stiff forward differential equations associated with transient reactor models of laboratory-scale. In this endeavor, we probe the limits of realistic chemical dynamics recoverability, using the ab-initio predicted timescales of the RWGS MKM as a case study.
Modeling and Simulation of Industrial Membrane Processes Using Complex Mixtures for Integration in Process Simulation Environments

(Georgia Institute of Technology, 2023-12-12) Weber, Dylan Jacob

The goal of this work is to enable design, optimization, and control of membrane-based separation processes that encounter complex industrial streams of up to thousands of components. These mixture components can have boundless concentrations and interactions between them. Presently, tools for such processes are non-existent. For chemical engineers, after synthesizing the chemical of interest, half of the job is separating it. Traditional separations rely on energy intensive heat and specialty chemicals which generate pollutants and contribute to climate change. Membrane-based separations alleviate these effects by using electrical energy which can be based on renewable resources. This thesis achieves this goal by asserting the following objectives: (i) develop improved numerical methods for local membrane transport of complex mixtures, (ii) extend models for predicting complex mixture sorption and diffusion, (iii) develop a software package for membrane process simulation to use within process flowsheet simulation environments, and (iv) present preliminary process design and control strategies for transport of complex mixtures through ion-exchange membrane modules (to shift towards electrochemical membrane-based separations for nutrient recovery from ubiquitous waste streams). The numerical methods, models, and software package presented has been, and will continue to be utilized by researchers and engineers to design, optimize, and control membrane-based processes as a green alternative for separations in the oil, bio-refinery, paper making, and water treatment industries.
On the importance of robust and accurate mixture adsorption data for industrial separations

(Georgia Institute of Technology, 2023-12-11) Bingel, Lukas Willi

Large-scale implementation of adsorption processes in industrial separations has enormous potential for reducing both energy costs and environmental footprint. To ease this transition, there is a need for reliable mixture adsorption data. As these data are difficult to obtain experimentally, the use of mixture prediction theories is commonplace. Here, the Ideal Adsorbed Solution Theory (IAST) has been established as the benchmark method. However, IAST relies on accurate, robust isotherm inputs and is derived based on ideality assumptions. Here, a new isotherm model is derived to accurately fit isotherm types I, III, and V with special emphasis on type V due to its appearance in many emerging adsorption systems. Then, meta-analyses of published single-component isotherm data for alkanes and alcohols are conducted to assess reproducibility and derive consensus isotherms as more robust input parameters for process modeling and simulations. To challenge the ideality assumptions, the impact of different metal-organic framework (MOF) motifs and mixed-linker configurations on the accuracy of IAST is investigated by comparing its predictions to mixture experiments. Traditional prediction theories reach their limits for novel MOFs with intrinsic flexibility due to a linker rotation-introduced gate-opening behavior. Thus, a dynamic adsorbent with record inverse selectivity for propane/propylene separation under real industrial conditions is investigated and a novel process integration is proposed. For the flexible ZIF-7, it is presented how a post-synthetic ligand exchange approach can be utilized to improve the capability for biogas separation. These results help bring adsorption-based separation systems one step closer to being reliable, energy-efficient alternatives to existing separation processes.
Understanding Complex Mass Transfer In Chemical Separations By Computational Modeling

(Georgia Institute of Technology, 2023-12-08) Cai, Xuqing

This dissertation explores crucial aspects of adsorption equilibrium and diffusivity in porous materials like silica, zeolites, and MOFs. The goal is to enhance comprehensive understanding of adsorption process through computational simulations, considering defects, and providing insights applicable to chemical separations. Beginning with the BISON-20 dataset, the study emphasizes the importance of precise binary gas mixture data. Then, it investigates defects in UTSA-280 and Zn(tbip), revealing DFT simulations fall short for UTSA-280 without considering defects. Energy barriers for molecular diffusion through defects align with experimental data. The dissertation also introduces defect engineering in Zn(tbip). The final segment addresses challenges in implementing adsorption-based DAC technology for global decarbonization. The framework developed evaluates meteorological variables on DAC efficiency, highlighting the need to adapt to real-world conditions. Overall, this research offers valuable insights into adsorption and diffusion processes, with practical applications in chemical separations and DAC technology.
When it Fails Right: A Learning-Based Investigation of the Effect of Failure and Career Tenure on Creativity

(Georgia Institute of Technology, 2023-12-05) Zhu-Ireland, Jiani

Drawing from experiential learning theories and research on exploration-exploitation, I develop a learning-based perspective of how people revise their knowledge structures after experiencing failure, which affects their subsequent creativity. I theorize that individuals with different pre-existing knowledge structures, reflected by the length of their career tenure, learn differently following failure. Specifically, individuals later in their careers are more likely to learn through revising and refining existing knowledge after experiencing failure because of the high complexity and low flexibility of their knowledge structures. Conversely, individuals earlier in their careers are more likely to learn through pursuing new knowledge after experiencing failure due to the low complexity and high flexibility of their knowledge structures. Furthermore, I develop a two-phase longitudinal model to examine how failure and the two types of learning influence initial and long-term creativity and how career tenure influences these processes. I test my hypotheses in a setting suited to the research question: the science-fiction novel field. Results show that after experiencing a failure, authors with longer (vs. shorter) career tenure are more likely to engage in exploitation (vs. exploration), even though exploitation harms their creativity both immediately and in the long run. This dissertation contributes to the literature on creativity, career tenure, and learning from failure.
Computational Modeling of Energetic and Diffusion Properties in Metal-Organic Frameworks

(Georgia Institute of Technology, 2023-11-14) Ibikunle, Ifayoyinsola

Metal organic frameworks (MOFs) are versatile materials with potential for applications like gas storage and separation, catalysis, energy storage and conversion, and photoluminescence and sensing. Here, we consider MOFs suitable for photoluminescence and energy storage (i.e., battery) applications. For photoluminescence, we focus on MOFs containing Rare-Earth (RE) metals due to their spectroscopic properties such as sharp and characteristic emission bands, strong resistance to photobleaching and long luminescence lifetimes. There is particular interest in developing heterometallic MOFs that use multiple RE atoms in close spatial proximity. For these materials, it is important to understand the details of metal ordering and siting. We use Density Functional Theory (DFT) to study metal ordering in Nd-Yb heterometallic MOFs, including a new MOF structure synthesized by our collaborators. We also performed additional calculations to determine if the insights on electronic structures from these Nd-Yb MOFs can be extended to other RE metals. Lastly, we sought MOFs suitable for Li-ion battery applications by data mining from a set of approximately 170,000 MOF materials. Our screening process is facilitated by pore size and chemistry and yields 131 MOFs that show potential promise for effective transport of Li ions. We quantify the diffusion properties of these materials by performing Molecular Dynamics simulations using classical force fields. The fundamental and applied knowledge gained from this work will aid in the rational design of functional materials for several emerging technologies.
Understanding and controlling water-organic co-transport in carbon molecular sieve membranes

(Georgia Institute of Technology, 2023-08-14) Yoon, Young Hee

Addressing water scarcity and pollution is important for global sustainability and environmental conservation. Industrial wastewater management plays a critical role in mitigating water pollution and ensuring safe water supplies. Membrane separation has emerged as a cost-efficient and highly effective approach for separating water from dissolved organic solvent contaminants. This dissertation focused on exploring the potential of a scalable rigid microporous material as a membrane for achieving desired separation performance in water and organic solvent separation. Carbon molecular sieve (CMS) is investigated as an advanced membrane material known for its impressive separation performance in gas and organic solvent separation. This study focused on p-xylene as an organic contaminant, which is a common aromatic organic solvent encountered in industrial wastewater. Overall, the study sought to understand a structure-transport relationship of water-organic solvent mixture in CMS membranes to optimize the utilization of CMS membranes for the water-organic separation. The research examined the structure of amorphous CMS membranes derived from the PIM-1 (polymer of intrinsic microporosity-1) precursor. Experimental findings provided valuable insights into the microporous structures of CMS. Then, the transport studies on water and p-xylene vapors in CMS membrane derived from polyvinylidene fluoride (PVDF) was performed using a sorption-diffusion (SD) model. The investigation extends further to PIM-1 (polymer of intrinsic microporosity-1) derived CMS membranes fabricated under various conditions to achieve different micropores structures and surface chemistries. We were able to confirm that the separation performance of water and p-xylene can be controlled via adjusting the properties CMS membranes. The study further investigated the transport of water and the dilute concentration of p-xylene in the liquid phase, in PIM-1 derived CMS hollow fiber membrane using the SD model. Overall, this research may provide valuable insights for the engineering and optimization of CMS membranes for water-organic separation challenges by understanding the relationship between membrane structure, transport regime, and separation modality. What we learned here may be useful in understanding and developing CMS membranes as effective tools in addressing the complex separation of water and organic solvents in industrial wastewater treatment.
Analyzing Sorption Exclusion Effects for the CO2/CH4 Gas Pair Using Matrimid® CMS Dense Films

(Georgia Institute of Technology, 2023-07-31) Vessel, Taylor C.

This thesis uses the so-called dual-mode sorption model to analyze Matrimid® polyimide-derived carbon molecular (CMS) thin films. This model provides a useful framework to analyze and understand sorption in complex CMS morphology. The model is also helpful to connect morphology to the pyrolysis process used to create it. The dual mode model includes coexisting continuous and dispersed Langmuir terms. The model parameters related to these environments are reported and discussed in this work. CMS based membranes have been applied towards many different gas-pair separations. CMS thin films have appealing separation performance for key gas pairs; the carbon dioxide/methane pair, which is practically important, is the focus of this study. Previous measurements on CMS film sorption for this pair have been done. Surprisingly, applying the dual-mode sorption model to such thin films created from Matrimid® over a range of pyrolysis temperatures has not been done previously and is considered here. This thesis characterizes sorption properties of such CMS thin films for the CO2./CH4 pair and adds insights relevant to their separation. Suggestions for the next steps to extend this study are also provided.
Structure-Property Relationships for Mixed Ion- and Electron-Conducting Polymer Systems

(Georgia Institute of Technology, 2023-07-30) Durbin, Marlow Monnig

Mixed ion- and electron-conducting polymers are an emerging class of functional materials combining mechanical softness and pliability with electrochemical functionality. The possibility of building devices from “plastic electronics” has spurred great interest in the fields of electrochemical energy storage, bio-integrated electronics, and analog computing. In this thesis, a “polymer science” approach is taken in combination with prior synthetic efforts to improve fundamental understanding of how the chemical and physical structure of mixed conducting polymers results in their diverse electrochemical, spectroscopic, and mechanical properties. We begin with a family of poly (3,4-propylenedioxythiophene) (“ProDOT”) polymers and copolymers, functionalized with aliphatic and/or polar oligo(ether) side chains of varying length and linearity for redox activity in organic electrolytes. Based on the redox and swelling behavior of three P(ProDOT)s, we illustrate that passive (bias-free) swelling of such P(ProDOT) films is principally driven by surface polarity, while side-chain free volume dictates the materials’ active swelling under applied electrochemical potentials by directing ion “flow” (sorption/desorption). We next study the effects of electrochemical cycling on ProDOT solid-state ordering as a function of side chain linearity. Electrochemical doping is found to enhance structural coherence in the lamellar/side-chain direction in P(ProDOT) polymers cycled in organic electrolytes, suggesting that the combined effects of electrochemical oxidation and counter-ion insertion substantially re-order these polymers. We go on to explore the phase behavior of multiple binary semiconductor:ion conductor blend systems with redox activity in aqueous and/or organic electrolytes, showing that careful manipulation of solution processing parameters may result in blend systems with altered redox activity compared to their neat components. With the design rules developed here, one can readily envision innovations in mixed ion-/electron-conducting polymer systems that bring the field closer to engineering “materials on demand” for varied applications.
Dewatering of Cellulose Nanomaterials using Ultrasound

(Georgia Institute of Technology, 2023-07-30) Ringania, Udita

This dissertation aims to develop a sustainable dewatering technology that can dramatically lower the energy consumption in pulp and paper industries. Using ultrasound technology, the study offers an energy-efficient and sustainable technique for dewatering Cellulose Nanomaterials (CNMs), a highly valuable cellulosic materials used in various industries. Given their unique properties and renewable nature, CNMs offer potential for achieving a sustainable economy. However, the high energy requirements associated with their dewatering and drying acts as a bottleneck in their widescale application. The research is segmented into three key objectives: First, the dissertation aims to design an energy-efficient, scalable, low-cost platform for dewatering Cellulose Nanofibers (CNFs) and examines the effects of system parameters on dewatering efficiency. Second, the research investigates how the fines percentage in cellulose nanofibrils affects the ultrasonic dewatering process. Lastly, it seeks to automate the image analysis process for highly branched fibrils, enabling faster and easier characterisation. Overall, the goal is to provide a sustainable solution to dewatering challenges in pulp and paper industries, reducing both energy consumption and cost, while also addressing key issues such as agglomeration. Guide to dissertation: In Chapter 1, I outline the dissertation’s motivation and research aims. I familiarize the reader with CNMs, explaining their conventional drying and dewatering techniques, and shedding light on the challenges intrinsic to these methods. In Chapter 2, I introduce an innovative ultrasonic dewatering process for CNFs which is validation through proof-of-concept experiments using a static dewatering setup. I also present a design for a continuous dewatering platform and evaluates its efficiency by considering parameters such as the number and configuration of transducers, the CNF flow rate on the transducers, and mesh pore size. I assess the redispersibility of dewatered samples and benchmarks it against CNFs dewatered using other methods. I conclude the chapter by providing an energy estimation and comparison. In Chapter 3, I further examine the influence of material parameters — specifically, the percentage of fines in the CNFs sample — on ultrasonic dewatering efficiency. I explore possible water removal mechanism, guided by observed dewatering curves for varying fines percentages. Finally, I evaluate the dependence of suspension stability of the redispersed samples on fines percentage, final dewatered CNFs weight, and redispersion time. My experiences with CNFs in previous chapters highlighted the challenges associated with characterizing these highly branched fibrils, which span a broad dimension range. In Chapter 4, I discuss an automated image analysis method, developed to ease the characterization of these highly branched fibrils through application of machine learning algorithms. Finally, in Chapter 5, I conclude the dissertation, offering recommendations for sustainable implementation of the technology. Here, I also describe few other methods that can be followed up for future work.