Indian Institute of Science Bangalore
etd@IISc Electronic Theses and Dissertations at Indian Institute of ScienceNot a member yet
6204 research outputs found
Sort by
Self-Assembled Coordination Architectures for Fluorescence Modulation, Photocatalysis, and Light Harvesting
No full textSelf-assembly and noncovalent interactions play a pivotal role in the design of complex, and
intricate functional structures. Among several design approaches, metal-ligand coordinationdriven self-assembly has emerged as one of the most efficient methodology for constructing
complex 2D and 3D architectures. This approach is favored due to its relatively simple design
principle, predictable directionality, and high bond enthalpy. Over the years, a vast range of
topologically intricate structures was designed using this approach. However, reports on the
construction of complex 2D, and 3D architectures using highly symmetric and rigid square planner
Pd(II)/Pt(II)-based metal acceptor in combination with rigid polypyridyl donor building blocks
dominate the literature so far. In this context, imidazole-based donors and flexible donors remained
less explored than conventional pyridyl donors. It is envisioned that the use of imidazole-based
donors as well as flexible donors might provide interesting results in terms of the structure of the
final assembly. The rotational degree of freedom in these cases can offer different bite angles to
the rigid pyridyl donor and may affect the structure of the final assembly.
On the other hand, supramolecular coordination polymers (SCPs) that are made up of an ordered
arrangement of repeating monomeric units have gained significant attention as they offer high
surface area, ordered porosity, and better stability compared to discrete supramolecular
coordination complexes.
The objective of the thesis is to synthesize various functional supramolecular architectures (both
discrete and coordination polymers) using imidazole-based donors and flexible donors via the
metal-ligand coordination approach. And to explore these self-assembled architectures for
fluorescence modulation, visible-light-driven photocatalysis, and light harvesting
Stress-Assisted Pitting Corrosion Studies in Metallic Structures Using Advanced Modeling Methodologies
No full textPitting corrosion poses a significant threat to deepwater oil and gas production facilities, yet our current understanding of its impact on specific systems such as steel catenary riser (SCR), tensioned leg platforms (TLP), and deep draft caisson vessels (DDCV) is limited. Existing research primarily focuses on pitting corrosion behavior in controlled laboratory environments and standard numerical methodologies, which fail to capture the complex and harsh conditions in deepwater facilities. This research gap hampers our ability to fully comprehend the extent and consequences of pitting corrosion on these systems. Moreover, there is a lack of comprehensive modeling methodologies to assess the effectiveness of corrosion mitigation strategies designed explicitly for SCR, TLP, and DDCV systems. While corrosion mitigation measures are commonly employed in the oil and gas industry, their efficacy in protecting these systems from pitting corrosion remains largely unexplored. Addressing these research gaps is crucial to improve our understanding of pitting corrosion behavior in deepwater facilities and to develop targeted and efficient mitigation strategies that ensure these critical assets' integrity, safety, and long-term viability.
In the field of corrosion fatigue, several research gaps exist, including understanding the mechanisms of pit initiation, quantifying the influence of stress and temperature on pit growth, characterizing the pit-to-crack transition, investigating the interaction between mechanical and electrochemical factors, exploring the influence of multiple pits cluster formations, studying corrosion fatigue short crack growth, and developing comprehensive models for corrosion fatigue. These gaps hinder our understanding of corrosion fatigue mechanisms and limit the development of effective strategies for mitigating its damaging effects. By addressing these research gaps, we can enhance our understanding of corrosion fatigue, improve predictive models, and develop strategies to mitigate corrosion-induced damage.
This thesis focuses on the development of advanced computational methods to analyze and understand the mechanisms of localized pitting corrosion damage, pit-to-crack transition, and thermal diffusion in evolving discontinuities in metallic structures. The primary objective is to devise integrated multi-stage modeling methodologies that encompass various stages of corrosion-related processes, including pit initiation, pit growth, pit-to-crack transition, and stable crack growth. Furthermore, the research aims to investigate corrosion fatigue and stress-assisted pitting corrosion by considering relevant factors. This work also seeks to characterize the pit-to-crack transition and quantify the effects of stress, strain, and temperature on pit growth in corrosive environments. Additionally, the thesis aims to develop computationally efficient models for non-local thermal diffusion in evolving discontinuities, leveraging the power of machine learning algorithms.
To achieve these aims, the thesis will pursue specific objectives, including the development of an advanced model for corrosion fatigue, validation of the model through experimental data, investigation of the pit-to-crack transition, analysis of factors influencing the pit-to-crack transition, quantification of stress levels and pit growth rates, exploration of temperature's influence on pit growth, development of a predictive model incorporating temperature influence, validation of the temperature influence model, incorporation of corrosion fatigue short crack growth into the advanced computational model, and develop computationally efficient model for nonlocal thermal diffusion in evolving discontinuities.
By accomplishing these aims and objectives, this thesis will expand our knowledge of localized corrosion mechanisms, enhance the accuracy of numerical models, and contribute to the development of effective strategies for managing pitting corrosion and its associated challenges. The thesis is organized into four parts, each focusing on different aspects of the problem. These parts employ various computational approaches such as Probabilistic Cellular Automata (PCA), the eXtended Finite Element Method (XFEM), and hybrid Peridynamics-based Machine Learning (PD-ML) models to simulate damage mechanisms accurately. These computational techniques are employed in a multi-stage sequential coupling to simulate damage mechanisms using various combinations. The chapters within each part provide detailed explanations and discussions of the specific approaches and their application to the research problem.
The first part of the work introduces a novel approach to understanding localized corrosion and cracking damage mechanisms in pipeline steel subjected to fatigue loading. The methodology consists of a sequential coupling of PCA and the XFEM to simulate the initiation of pitting corrosion, pit-to-crack transition, and stable cracking. PCA describes the stochastic nature of localized corrosion damage, whereas XFEM is used to model arbitrary inhomogeneities, such as cracks, voids, and material interfaces. Level Set Methods (LSM) are implemented to track the crack propagation location accurately. Additionally, XFEM is enriched by additional degrees of freedom using the partition of unity concept, which enables efficient solutions without requiring mesh conformity to internal boundaries or re-meshing. A localized strain criterion initiates cracks from the pit surface when the pit reaches a critical strain value. The study demonstrates that the proposed coupling of PCA and XFEM provides a reliable approximation of experimental data and may be applied to predict the service life and integrity of pipelines subjected to similar loading conditions.
The second part of the work presents a hybrid formulation of a PD-ML model for thermal diffusion analysis in one-dimensional and two-dimensional problems with evolving discontinuities. Here, we use a multivariate linear regression approach to establish the relationship between the temperature values of material points, their neighboring points, and the applied external heat fluxes. The thermal modal analysis method uses the finite element method to generate training and testing data. An efficient numerical procedure is developed to couple the peridynamics and PD-ML models. The model is analyzed under multiple configurations of micro-thermal conductivity functions for one-dimensional thermal bar problems under both steady-state and transient loading conditions to identify the configuration that exhibits superior convergence behavior towards the local solution. The hybrid model effectively captures intricate discontinuities and boundaries while being computationally efficient, indicating its potential for thermal diffusion analysis in one- and two-dimensional problems with evolving discontinuities.
The third part of the work explores the potential of theory-based data science in the field of material science by introducing a hybrid formulation of a PD-ML model for pitting corrosion in one- and two-dimensional problems. Using a multivariate linear regression model, we establish relationships between the concentration value of material points, its family members' concentration, and the corresponding applied external mass fluxes within the linear regime for isotropic materials. Training and testing data are generated using the model analysis for mass diffusion using the finite element method for the machine learning model. An efficient numerical algorithm is developed to couple the peridynamics and PD-ML models. The model is analyzed under multiple configurations of micro-diffusivity functions for one-dimensional bar problems under transient loading conditions to identify the configuration that exhibits superior convergence behavior towards the local solution. Furthermore, an integral aspect of this study entailed the integration of a cutting-edge predictive model that effectively incorporates the intricate influence of temperature, followed by rigorous experimental and computational validation.
The last part of the work presents a novel approach to modeling localized corrosion and cracking damage mechanisms in pipeline steel subjected to mechanical loading. Here, we employ a sequential coupling of PCA and a PD-ML model to simulate the initiation of pitting corrosion and pit-to-crack transition. To maintain consistency, we are utilizing the PCA model that was previously developed in part one of this thesis. PD-ML is a hybrid approach that couples machine learning and ordinary state-based peridynamics models for fracture prediction of linear elastic 2D structures. We assume that a material point's displacements have a linear relationship with the displacement of its neighbors and its applied external forces. Weighted coefficients for the PDML model are obtained using multivariate linear regression analysis. The training and testing data are generated from structural modal analysis using the finite element method for a square plate. The study finds that coupling PCA and PD-ML provides a good approximation of experimental data of stress-assisted pitting corrosion and may be used to predict the integrity of pipelines subject to mechanical loading.
Overall, the thesis showcases novel approaches to modeling and simulating the behavior of materials and structures under various loading conditions, focusing on pitting corrosion, pit-to-crack transition, and thermal diffusion and aiming to contribute to developing improved numerical methods for their analysis. Combining the above works, we present a comprehensive approach for predicting the localized corrosion and cracking damage mechanisms in pipeline steels under fatigue and mechanical loading. We also provide thermal diffusion analysis with evolving discontinuities in one- and two-dimensional problems. The approaches combine PCA, the XFEM, and the hybrid PD-ML models to simulate the damage mechanisms accurately. The approaches presented in the thesis can potentially improve computational efficiency and the accuracy of predicting pipelines' service life and integrity subjected to similar loading conditions and provide insight into the thermal diffusion process in materials with evolving discontinuities. The thesis also strives to advance our understanding of localized corrosion damage mechanisms, improve numerical modeling techniques, and provide valuable tools for predicting and mitigating corrosion-induced damage in metallic structures. The findings of this research work have practical implications for various industries and contribute to ensuring the integrity and long-term viability of critical assets
True Random Number Generators based on Amplified Spontaneous Emission
No full textLow-cost, high-speed quantum random number generators (QRNGs) are imperative to develop in order to have a widespread application in areas ranging from cryptography and stochastic simulations to banking and internet. Amplified spontaneous emission (ASE) based QRNGs offer a great advantage in this pursuit. However, most ASE-based QRNGs use costlier and more complex components compared to their speed. In this work, we discuss two QRNG schemes where we alleviate this problem. We also develop a comprehensive theoretical study to understand then in detail. In our ASE-ASE beating based QRNG, random numbers are generated by performing balanced detection of two independent ASE signals. Our second QRNG scheme is based on mixing a laser and an ASE of similar power and measuring the beat noise using a balanced detector. Our theoretical analysis shows that introducing the laser improves the standard deviation of the raw data, which can result a higher minimum entropy in the later QRNG. This fact is corroborated by the experiments as well. Our first experimental implementation of ASE based QRNGs use a balanced detector of 1.6GHz bandwidth and clock in bit generation rates of 7.44Gbps (ASE-ASE) and 7.9898Gbps (ASE-Laser) using an 8-bit ADC. Then we reduce the bandwidth of the detector to 100MHz since most available ADCs with higher bit resolution have bandwidths of few hundreds of MHz. In this case also, we achieve half a Gbps speeds for our QRNGs using the aforementioned 8-bit ADC - 546.03Mbps (ASE-ASE), 507.94Mbps (ASE-Laser). Furthermore, we develop two multiplexed QRNG schemes to double the speed of the QRNGs. Our experimental implementations of them show that multiplexing can achieve 1Gbps speed with a very slow detector (bandwidth of 100MHz). All the random numbers are verified using industry-standard statistical tests - NIST and DIEHARD - and obtain satisfactory results. Moreover, they can be easily integrated on a chip for commercial use
Molecular pathways governing maturation and decay of precursor piRNAs in Caenorhabditis elegans
No full textPiwi-interacting RNAs (piRNAs) are an animal-specific class of germline-enriched small non-coding RNAs that play a crucial role in germline development and fertility in diverse organisms. In Caenorhabditis elegans (C. elegans), mature piRNAs are uniformly 21 nucleotides (nt) long, and begin with monophosphorylated uridine (therefore, also known as 21U-RNA). They are processed from 25-30 nt long capped RNA precursors, which are transcribed by RNA polymerase II and initiate precisely 2 nt upstream of the 5’-end of mature 21U-RNAs. Therefore, to generate 21 nt long mature piRNAs, the 5’ cap and first two nucleotides from the 5’-end must be removed, and extra nucleotides must be trimmed from the 3’-end. It has previously been shown that PARN-1, a conserved 3’-5’ exoribonuclease, contributes to the 3’-end processing of the precursors. However, the enzyme(s) responsible for processing the 5’-end of precursor piRNAs is still unknown.
Objective 1: Identification and characterization of factor(s) involved in the processing of precursor piRNAs in C. elegans.
Here, we report that the endoribonuclease activity of XRN-2 is involved in the 5’-end processing of a small set of piRNAs in C. elegans. Our findings also demonstrate that XRN-2 is capable of degrading precursor piRNAs that are not bound and protected by PRG-1, revealing a potential surveillance mechanism for ensuring quality control during precursor piRNA processing. Our research suggests that piRNAs originating from longer precursor molecules (≥ 60 nt) undergo an initial ribonucleolytic processing of their 3’-ends, which is mediated by ENDU-1, either directly or indirectly. Finally, we show that XRN-2 is not only important for the generation of mature piRNAs and piRNA-dependent 22G-RNAs, but through yet unknown pathways, it also affects piRNA-independent 22G-RNAs that shape transcriptome, as well as contribute to genomic integrity via regulation of transposable elements.
Objective 2: Elucidating a tailing-mediated surveillance pathway for the degradation of unprocessed intermediate precursor 21UR-1 in C. elegans.
Here, we report on a surveillance pathway mediated by 3’-tailing that targets unprocessed intermediate precursor 21UR-1 in the adult germline of C. elegans. The absence of proper 5’- and 3’-end processing events leads to the tailing of intermediate precursor 21UR-1, which acts as a signal for their degradation. CID-1 is identified as the mediator of 3’-tailing of precursor 21UR-1. Additionally, it has been shown that tailing affects the stability of PRG-1. Abrogation of the tailing event leads to a significant accumulation of unprocessed intermediate precursor 21UR-1 in C. elegans embryo, which would otherwise be degraded during the maternal to zygotic transition.
Overall, our study provides insight into precursor piRNA processing and quality control mechanisms in C. elegans.Council of Scientific and Industrial Research (CSIR
Deciphering the biochemical and biophysical properties of the Holliday junction resolvases RuvC and RuvX from Mycobacterium smegmatis
No full textHomologous recombination (HR) is a ubiquitous cellular process that occurs in all three domains of life as well as in DNA and RNA viruses. In eukaryotes, HR is critically important for homology-directed DNA repair (HDDR) in mitotic cells and synapsis of homologous chromosomes during prophase I of meiosis. Therefore, cells defective in HR exhibit multiple chromosomal aberrations, which raises the probability of genomic instability and cancer in eukaryotic organisms. In bacteria, similar to eukaryotes, HR is required for DNA repair and to facilitate horizontal gene transfer during transformation, transduction, and conjugation, while it plays a vital role in the evolution of viruses. A central intermediate formed during the late steps of HR and HDDR is a four-way joint DNA structure, also known as the Holliday junction (HJ), which is one among several different types of branched DNAs formed during HR and HDDR. A suite of studies have demonstrated that HJs adopt two global conformations: open and stacked. Whilst, the stacked form has two quasi-continuous helices and is induced by the binding of metal ions, the open form is a four-fold symmetric structure with a square, planar configuration, and is capable of branch migration between homologous duplexes, an essential step during DNA recombination and HDDR. The HJs are resolved by structure-specific endonucleases, known as junction resolvases, which are essential for faithful segregation of chromosomes during meiosis and double-strand break repair by HR in mitotic cells. Holliday junction resolvases (HJRs) have been identified in a wide variety of organisms, including animals, plants, fungi, and even microorganisms like bacteria and viruses based on their shared structural and functional characteristics, although they exhibit significant diversity and belong to different classes of structure-specific endonucleases.
Genetic, functional and structural analyses of Escherichia coli RuvC, the founding member of canonical HJRs, has provided key insights into the mechanism of processing branched DNA intermediates that arise during HR and HDDR. In eukaryotes, three conserved nucleases, such as Mus81–Mms4/MUS81–EME1 (budding yeast/human) and Yen1/GEN1 and Slx1–Slx4/SLX1–SLX4 have been found to have properties similar to those of E. coli RuvC. In contrast to the large body of information on E. coli HJRs, our knowledge about the HJRs in Gram-positive bacteria, and in particular of mycobacterial species is incompletely understood. Of note, genetic studies have revealed that the mycobacterial HR pathway differs from that of HR in E. coli. Thus, a full understanding of the biophysical and biochemical characteristics of HJRs in mycobacterial species may reveal important insights into interspecies and intergenus divergence and evolution of their domain structure and function.
A bioinformatics search on the whole genome sequence of Mycobacterium smegmatis revealed the presence of genes encoding two putative HJRs: ruvC and ruvX. Intuitively, this observation implies that ruvC and ruvX genes may play a crucial role in DNA recombination and repair. This finding raised intriguing questions: why does M. smegmatis encode teo putative HJRs; do they function independently or in a mutually exclusive manner? Thus, the overall purpose of our research is aimed at a fuller, more complete understanding of the functional aspects of M. smegmatis HJRs RuvC and RuvX. In the first half the thesis, we describe the molecular cloning, expression, purification and biochemical characterization of a canonical HJ resolvase RuvC from M. smegmatis (MsRuvC). The recombinant MsRuvC was purified to homogeneity using heparin affinity column and gel filtration chromatography under native conditions from the supernatant. The purified MsRuvC occurs as a homodimer in solution and shows binding preference to a wide variety of DNA structures, many of them mimic the intermediates generated either during recombination or replication. Importantly, we found that MsRuvC cleaves the mobile HJ (mHJ) at 5'-T↓C-3' around the crossover point, generating ligatable, nicked duplex products in Mg2+/Mn2+ dependent manner and resolution of three-way junctions occurred in the presence of Mn2+, but failed to act on immobile HJ, flaps, splayed arms, and replication fork structures. However, we cannot rule out the possibility that MsRuvC might cleave these substrates in the in vivo context. Our studies with various types of HJs revealed that MsRuvC-catalyzed incisions occur at the 3′ side of thymidine (5'-T↓C-3') positioned at or one base pair away from the branch point in HJ. We found that alanine substitution of Asp7 and Glu68 in the RNase-H domain of MsRuvC independently impaired the ability of MsRuvC to catalyse HJ resolution. However, their HJ binding activity remains unperturbed, indicating that binding by itself is not sufficient to ensure efficient HJ cleavage. Notably, MsRuvC was found to resolve the 3-way DNA junction as well, but failed to act on the Y-shaped substrate, flap and fork substrates, and duplex DNA. Overall, our data provide important insights into the substrate specificity of MsRuvC, mode of its substrate binding, as well as catalysis, advancing our knowledge on HJRs in mycobacteria. Viewed together, these findings indicate MsRuvC can be considered as a canonical HJ resolvase, similar to the prototype E. coli RuvC.
Many studies have provided strong genetic evidence that certain bacterial and eukaryotic organisms encode enzymes belonging to the RNaseH/retroviral integrase superfamily, known as YqgF/RuvX proteins. Since YqgF shares striking sequence and structural similarities with RuvC, including the presence of an RNase H-like motif, it was predicted to function as an alternative HJR. Herein, we report the cloning of M. smegmatis ruvX, heterologous expression, and biophysical and biochemical characterization of recombinant MsRuvX, and compare the results with M. tuberculosis RuvX (MtRuvX). Unlike dimeric MtRuvX, MsRuvX exists as a monomer in solution and exhibits high binding affinity for a suite of branched DNAs. Rather unexpectedly, we found that MsRuvX has no HJ resolution activity, while delivering a very robust cleavage activity on a suite of branched DNAs, generating DNA products of different lengths in reaction buffers containing either Mg2+ or Ca2+. However, MsRuvX mediated flap resolution was found to be ~2 fold higher in presence of Mg2+ compared to that of Ca2+. Furthermore, MsRuvX promotes robust cleavage of single-stranded DNA (ssDNA), but not double-stranded DNA, and that it resects ssDNA in a 5'-to-3' direction. Mutations of amino acid residues D25 and D142 in the MsRuvX RNase H-like domain rendered the resulting variants functionally inactive; however, their DNA binding activity was unaffected. These results suggest that MsRuvX might play an important role in the processing of multiple types of branched DNAs, other than HJs, arising from cellular processes such as produced during HR and HDDR, thereby furthering our understanding of HR and HDDR in M. smegmatis.
Overall, this study provides novel insights into the non-overlapping functions of RuvC and RuvX in the processing of a wide range of branched DNA structures in M. smegmatis. Importantly, this study has uncovered a strong division of labour between RuvC and RuvX in M. smegmatis, which may enhance the processing efficiency of different types of branched DNA structures. Thus, on the one hand, RuvC function might be limited to the resolution of HJ, on the other hand, RuvX may act in concert with, or independently from that of, RuvC in the processing of other types of branched DNAs that arise during the processes of HR and HDDR to maintain the integrity of M. smegmatis genome under different growth conditions
Design Principles of Phenotypic Robustness and Plasticity in Gene Regulatory Networks underlying Cancer Metastasis
No full textMetastasis – the process of cancer cells leaving the primary tumor and colonizing multiple organs – remains a major cause of cancer mortality. However, it is a highly inefficient process with only 0.01% of disseminated cells eventually succeeding in colonization. The reason for such high rates of attrition is inability of disseminated cancer cells to constantly adapt to various obstacles such as rapidly changing biochemical environments, anoikis and immune attack.
Successful metastasis requires disseminating cells to strike a balance between two contrasting dynamical features: phenotypic plasticity and phenotypic robustness. Phenotypic plasticity is the ability of cells to switch among different phenotypes reversibly, while robustness is their ability to retain a phenotype against intrinsic or extrinsic fluctuations they face. For example, successful colonization is often achieved by disseminating cells acquiring an epithelial phenotype upon reaching a secondary organ, and then maintaining it despite fluctuating microenvironments at metastatic site. While individual molecules driving robustness and plasticity have been reported, how cells maintain this delicate balance remains an open question. Here, we have investigated the dynamics and design principles of regulatory networks that can govern these properties, taking the Epithelial Mesenchymal Plasticity (EMP) – a key component of metastasis – as a case study.
First, to understand the emergent dynamics of EMP networks of varying sizes and densities, we employed two simulation formalisms: RACIPE (a continuous, parameter agnostic approach) and Boolean modeling (a discrete, parameter independent formalism). These networks enabled epithelial (E) and mesenchymal (M) phenotypes as predominant ones, alongside the less frequent hybrid E/M ones. Using these two formalisms, we found that phenotypic frequency distributions obtained from EMP networks are more robust to structural perturbations (mutations that change the nature of edge connecting two nodes) and dynamical perturbations (mutations that effect the strength of connection between two nodes), as compared to their randomized counterparts. Similarly, we observed that EMP networks have a higher tendency to allow for phenotypic plasticity than randomized networks, indicating that they may have evolved to facilitate both phenotypic robustness and plasticity.
Second, focusing on the design principles of EMP networks, we demonstrated that both phenotypic robustness and plasticity are supported by a larger number of positive feedback loops and a smaller number of negative feedback loops embedded in these networks. Importantly, we noted that these feedback loops intertwine to manifest themselves as two well-defined “teams” of nodes. These “teams” allow for the co-expression of specific genes that characterize the dominant E and M phenotypes. Also, the “teams” structure confers robustness to multiple structural and dynamical perturbations specifically for E and M phenotypes while allowing for hybrid E/M phenotypes to be more plastic relatively. Finally, we identify two network topology metrics – team strength and the fraction of positive feedback loops – that can explain the extent of plasticity and robustness emergent from EMP networks. These metrics allow us to identify the single-edge perturbations capable of significantly altering these networks' plasticity and/or robustness.
Together, our analysis elucidates that phenotypic plasticity and robustness in cancer are emergent properties of the topology of EMP regulatory networks and present a platform to isolate specific network perturbations to curtail these dynamical features, thereby potentially impacting metastasis.PMRF, DS
Temporal Point Processes for Forecasting Events in Higher-Order Networks
No full textReal-world systems consisting of interacting entities can be effectively represented as time-evolving networks or graphs, where the entities are depicted as nodes, and the interactions between them are represented as instantaneous edges. Modeling the evolution of these systems and forecasting interaction events are important for many fields, such as e-commerce, financial markets, neuroscience, etc. This is achieved using the Temporal Point Process (TPP) framework, a stochastic process that models these interactions as discrete events occurring in continuous time. The existing works on interaction forecasting are applicable only to pair-wise edges.
However, real-world interactions are much more complex than pair-wise interactions. It involves a group of entities interacting in a complex way rather than just two entities. This leads to the formation of time-evolving higher-order networks. There has not been much research to develop machine learning algorithms for event prediction in these types of networks. This thesis addresses this by providing solutions to the following problems: (i) How can we train models on the events observed in a higher-order network? (ii) Considering the number of possible events grows exponentially when problem setting changes from pair-wise to higher-order, can we forecast the next event in a scalable way? (iii) How to model the influence of interactions on nodes in a higher-order network, and how to improve its scalability? (iv) How can we incorporate relations and group structure within an interaction to an event forecasting model?
The first contribution of this thesis is a model for forecasting interactions among a group of entities as instantaneous hyperedge events in a network. In this model, we introduce a TPP on each hyperedge, with the conditional intensity parameterized by a hyperedge link predictor that uses node embeddings. We employ a dynamic node embedding strategy to account for the temporal evolution of entities with each interaction. Here, all the model parameters are learned by a mini-batch based negative log-likelihood calculation with negative sampling for incorporating unseen hyperedges. Furthermore, our model has been extended to accommodate bipartite interactions, where interactions occur between two distinct groups of entities of different types. To achieve this, we introduce a bipartite hyperedge link predictor and use separate node embedding modules for each node type.
Secondly, we propose a model to forecast directed higher-order interactions occurring between two distinct groups of entities. Unlike the previous approach that focuses on representation learning from higher-order network events, here we also introduce a strategy to forecast hyperedges in a scalable way. For that, we employ a multi-task framework for forecasting candidate hyperedges. This involves a TPP-based model to predict the time of events on each node, followed by pairwise neighborhood and hyperedge size prediction modules for generating candidate hyperedges. This will reduce the exponential search in forecasting future hyperedges in the previous models. Then a directed hyperedge link predictor is used to identify the true hyperedge from false ones. Further, we devise a dynamic node embedding architecture that processes samples in a batch using memory network and temporal graph attention modules. This improves the scalability of the model when dealing with datasets containing a large number of events.
In our final contribution, we extend the existing interaction forecasting approaches based on TPP to accommodate real-world interactions that involve internal group structures of different types. Here, each group is associated with a specific relation type, and to address this complexity, we introduce the concept of multi-relational recursive hyperedge formation events. In this framework, hyperedges can serve as nodes within other hyperedges, creating a hierarchical structure. Further, we also introduce a contrastive learning strategy to learn model parameters without using the intractable likelihood of TPP.
In conclusion, this thesis emphasizes and demonstrates the importance of employing higher-order temporal network models to forecast interactions in real-world systems accurately. To achieve this, we created deep neural network based approaches for dynamic node embedding to extract information from historical data and link predictors to model the occurrence of the edge formation event. Then devised different strategies to train these models and forecast events from them
Astrocytes regulate oligodendrocyte development and myelination in the mammalian brain
No full textOligodendrocytes (OLs), a type of glial cell, are the main myelinating cells of the mammalian central nervous system (CNS), enabling efficient saltatory mode of nerve conduction. On the other hand, astrocytes, another glial cell type, have diverse functions including neurotransmitter reuptake, maintenance of the blood-brain barrier, and providing trophic support. In the murine brain, there is a temporal sequence in the generation of major cell types with neurons differentiating first from the ventricular/sub-ventricular zone neural progenitor cells (NPCs), followed by astrocytes and oligodendrocytes.
Recent findings from our lab have shown that astrocyte-specific deletion of serum response factor (Srf) early during murine brain development leads to hypertrophic astrocytes exhibiting reactive-like phenotype throughout the brain (Jain et al., 2021). This reactive-like phenotype persists throughout adulthood. Interestingly, the SrfGFAPcKO brains also exhibit severe loss of myelin in different grey and white matter regions. The myelin deficits become evident around four weeks of age and worsen over time. Further investigations have shown that the myelin loss is not due to deficits in oligodendrocyte lineage cell proliferation, differentiation, or loss of oligodendrocyte progenitors or mature oligodendrocytes. Instead, the observations suggest that oligodendrocytes fail to mature into the myelinating phenotype.
Comprehensive evaluation of motor behavior provided compelling evidence of significant abnormalities in motor coordination and gait in these Srf mutant mice as evident from open field exploration, accelerating rotarod test and footprint analysis. These findings not only shed light on the specific behavioral deficits but also establish a potential connection between the observed hypomyelination in these mice and the impairments in their motor functions.
Transcriptomic analysis of RNA isolated from control and SrfGFAPcKO astrocytes revealed the downregulation of genes involved in lipid and cholesterol metabolism in SRF mutant astrocytes. Astrocytes are known to supply essential lipids and cholesterol to oligodendrocytes for myelin synthesis. Any disruptions in this astrocytic lipid and cholesterol supply can result in defects in myelination. Furthermore, a comparative analysis of the Srf mutant astrocyte transcriptome with that of astrocytes in a mouse model of Alexander disease has revealed shared genes between the two datasets. This further suggests that the myelin defects observed in the Srf mutant mice resemble those observed in the Alexander disease model.
These findings shed light on the intricate interplay between astrocytes and oligodendrocytes and highlights the importance of Srf and associated pathways in the regulation of oligodendrocyte development and myelination
Thermochemical conversion of biomass – single particle and packed bed: Experimental and Numerical Studies
No full textThe current work uses experimental and numerical techniques to analyze the thermochemical conversion of a single biomass particle exposed to various reactive environments – varying temperatures and O2, CO2 and H2O concentrations. The single-particle analysis is extended to packed beds, wherein a biomass bed is subjected to controlled reactive environments as occurring in practical gasification units: the evolution of temperature, conversion and gas composition through the bed is studied. The work ultimately results in a multi-scale packed bed numerical model, resolved to the individual particle within the bed and further resolved to pore-scale diffusion within the particle. The model provided with fuel properties and oxidizer conditions computes the product gas composition, gas yield and conversion rate.
Identifying char combustion as the limiting step in biomass combustion and recognizing that char combustion is a diffusion-limited process (~0; =1.97), the oxygen concentration in the feed is increased to curtail the diffusion limitation. Very interestingly, it is observed that at higher oxygen concentrations (> 40 %) and higher temperatures (>673 K), β surpasses the theoretical threshold of 2 and goes as high as 2.37. In parallel, it is observed that under cases wherein β>2, a luminous film envelope the particle. The higher flux of CO from the particle at high temperatures and oxygen concentration resists the diffusion of reactant to the particle surface, curtailing the conversion process, increasing the conversion time and thereby β>2. An analytical solution is formulated from the first principles, and it is noted that the conversion time, in addition to being a function of d0, also depends on the film diameter through the factor (1-(2/3)(d0/df)).
Gasification is also studied at a particle level by exposing a single biomass particle to CO2 and H2O environments. The threshold temperature beyond which practically significant reactions
occur is 737 0C and 850 0C for Char-H2O and Char-CO2. An increase in temperature and reactant concentration is found to enhance the gasification rate, albeit depending on the underlying conversion regime. The reactant flux is found to improve the conversion rate by increasing the transport of reactant to the surface. However, beyond a threshold flux, the conversion rate is controlled by an intra-particle gradient and is insensitive to flux. An increase in particle size leads to higher carbon mass and longer conversion time. The porosity and internal surface area decrease with an increase in density; hence, the gasification rate decreases.
The current work notes that the entire range of the conversion regime (from kinetic limit to diffusion limit) can be spanned by controlling the temperature and reactant concentration, irrespective of particle size or reaction. For the first time, the current work presents a conversion regime map, a contour plot to identify the nature of the conversion regime by readily measurable and controllable parameters – temperature and reactant concentration.
In packed bed gasification of char by steam, an intriguing observation pertains to conversion inhibition by H2. The H2 generated upstream of the bed inhibits the conversion process downstream. Char surface analysis indicates that along the bed length, the C-O and O-H bonds decrease, and the C-H bonds increase, confirming the hypothesis. To further characterize the system, H2 of the known fraction was introduced in the feed gas. It was found that the steam-char reaction is completely inhibited beyond a threshold concentration of H2 in the feed gas. For a given carbon surface with a finite number of active sites, H2 quickly saturates and bonds with the char surface curtailing the C+H2O reaction and, thereby, the conversion rate. The experimental findings and numerical analysis are used to arrive at threshold H2 concentration and bed height beyond which inhibition occurs for a given temperature, reactant concentration and residence time.
Having validated the numerical model over a range of experimental conditions discussed above, the model is used to compute the product gas composition and reaction propagation rate for packed bed gasification of char and biomass in air and oxy-steam mixtures. The model estimates agree well with in-house experimental results and experimental data from the literature
Tailoring Van – der – Waal materials for nonlinear nanophotonics
No full textVan der Waals or two-dimensional (2D) material are emerging nanomaterials which has shown tremendous potential in building next-generation nonlinear optical devices. Monolayer and multilayer 2D materials exhibit strong nonlinear optical response owing to their highly ordered crystalline structure, layer-tunable bandgap and symmetry, and polarization dependence, and which offer the possibility of incorporating electrical tunability due to their unique layer-dependent, electric and strain-tunable electrical and optical properties. The unique properties of these van der Waal materials have led to their proposed use in various applications such as wavelength conversion, saturable absorption, optical limiting, optical modulation, and parametric down-conversion. For instance, ultra-thin active photonic devices made from 2D materials could potentially provide significant advantages such as smaller device size, higher efficiency, and faster response time when compared to traditional photonic devices. Therefore, characterizing the strength of nonlinear response and boosting it further by augmenting it with a resonant photonic structure is of utmost importance. The thesis aims to characterize the nonlinear response of emerging 2D materials and will discuss optimally designed nanophotonic structures to enhance the nonlinear response.
In the first part of the work, we report strong second-harmonic generation (SHG) from the 2H polytype of multilayer Tin diselenide (SnSe2) for fundamental excitation close to its indirect band-edge in the absence of excitonic resonances. Rapid oscillations in signal strength for slight changes in flake thickness have been observed which are in good agreement with a nonlinear wave propagation model considering nonlinear polarization with alternating signs from each monolayer. We observed enhanced SHG at 1040 nm compared to 1550nm which is attributed to the enhanced nonlinear optical response for fundamental excitation close to the indirect band-edge. We also studied SHG from heterostructures of monolayer MoS2/multilayer SnSe2 and found the SHG signal and any interference effect in the overlap region to be dominated by the SnSe2 layer for the excitation wavelengths considered. The comparison of SHG from SnSe2 and MoS2 underscores that the choice of the 2D material for a particular nonlinear optical application is contextual and requires consideration of the wavelength range of interest and its optical properties at those wavelengths. This work further highlights the usefulness of near-band-edge enhancement of nonlinear processes in emerging 2D materials towards realizing useful nanophotonic devices.
Furthermore, we report up-conversion of 1550 nm incident light using third harmonic generation (THG) in multi-layered tin di-selenide (SnSe2) and study its thickness dependence by simultaneously acquiring spatially-resolved THG images in the forward and backward propagation direction. We find good agreement between the experimental measurements and a coupled-wave equation model we have developed when including the effect of Fabry-Perot interference between the SnSe2 layer and the surrounding medium. We extract the magnitude of the third-order electronic nonlinear optical susceptibility of SnSe2, for the first time to our knowledge, by comparing its nonlinear response with a glass substrate and find this to be ∼1500 times higher than that of glass. The large nonlinear optical susceptibility of multi-layer SnSe2 makes it a promising material for studying nonlinear optical effects.
The next part of this thesis focused on building nanophotonic structures for enhanced nonlinear optical response. Here in the first part, we report hybrid-genetic optimization (HGA) based design and experimental demonstration of second harmonic generation (SHG) enhancement from Fabry–Perot structures of single and double multilayer gallium selenide (GaSe) flakes with bottom silicon dioxide and index-matched polymethyl methacrylate spacer/encapsulation layers. The HGA technique utilized here speeds up the multilayer cavity design by 8.8 and 89 times for the single and double GaSe structures when compared to the full parameter-sweep, with measured SHG enhancement of 128- and 400-times, respectively, when compared to a reference sample composed of GaSe layer of optimized thickness on 300 nm silicon dioxide layer. SHG conversion efficiencies obtained from the HGA structures are 1–2 orders of magnitude higher than previous reports on 2D material integrated resonant metasurfaces or Bragg cavities.
In the final section of this thesis, we leverage the high refractive index exhibited by transition metal dichalcogenides (TMDCs) to create isolated Mie-resonant optical structures that facilitate strong nonlinear optical interaction. To achieve this, we performed scattering cross-sectional simulations to determine the appropriate dimensions for the MoS2 disk to enable higher-order anapole modes within our desired signal wavelength range of 1400-1700nm. We then fabricated the MoS2 disks by patterning a dry-transferred MoS2 flake on a 2.2-micron thick thermal SiO2 layer deposited on a silicon wafer. Our experiments show that the fabricated isolated MoS2 disks do exhibit higher-order anapole resonances, leading to enhanced second harmonic generation. Specifically, we observed a maximum experimental SHG enhancement of 96 times at 1470nm from the MoS2 disk compared to the un-patterned MoS2 flake region. We also performed detailed nonlinear wave propagation simulations, which were in good agreement with our experimental data. These results demonstrate that optical nanostructures based on layered materials, such as MoS2, have the potential to serve as efficient wavelength converters across widely separated wavelength bands