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
Deep learning methods for light fluence compensation in two-dimensional and three-dimensional photoacoustic imaging
No full textPhotoacoustic imaging (PAI) employed the special properties of light or photons to obtain
detailed images of organs, tissues, cells, and even molecules. The method allowed for a
non-invasive or minimally invasive examination within the body. The PAI used nearinfrared
light (600 nm - 900 nm) as the scan media, which had the additional benefit of
being a non-ionizing imaging modality. The PAI can be integrated with other imaging
modalities, such as MRI or X-ray, to provide better information for complex diseases or
researchers working on complex experiments. Photoacoustic imaging has already been
widely used in pre-clinical research to image small animals. Although PAI is a multi-scale
modality, it is challenging to use for clinical research and interventional applications due
to the non-linear distribution of optical fluence.
Quantitative Photoacoustic Imaging (QPAI) has remained problematic due to the
influence of non-linear optical fluence distribution, which influences photoacoustic image
representation. Non-linear optical fluence correction in PA imaging was highly ill-posed,
leading to inaccurate recovery of optical absorption maps. Note that the traditional
optical fluence correction method needs precise estimation of optical fluence map. Many
different light transport models exist for estimating the optical fluence map when the
optical properties, i.e., optical absorption and optical scattering are known. However,
in reality the optical properties are unknown in advance, therefore fluence estimation
becomes difficult during PA imaging. Moreover, optical light illumination at the target
medium under the study is not uniform over the wavelength, the target medium introduce
spectral distortion between the measured PA spectrum and the true target spectrum.
Consequently, for true QPAI, the optical fluence must be simultaneously estimated and
compensated. This requires not only an appropriate fluence model, but also an effective
method to estimate the fluence distribution at each wavelength from PA measurements.
Based on prior knowledge of the target medium’s optical properties, many different
methods have been proposed for fluence compensation for a simple and homogeneous
medium. Unfortunately, none translate into clinical usage. To translate to clinical usage,
more complex and heterogeneous media need to be studied. And also the generated PA
signal may also change dynamically based on the background tissue properties. Hence,
consider complex, foreground and background non-homogeneity of the medium for accurate
recovery of optical absorption coefficient. None of the research groups adapted all
the above factors simultaneously for fluence compensation. This thesis study developed
a deep learning-based optical fluence correction approach to solving this limitation.
The main objective of this thesis was to investigate the non-linear distribution of optical
fluence effect in 2D and 3D medium and compensate this effect by using deep learning
(DL) models. This thesis explains the recovery of the optical absorption maps using
deep learning approaches by correcting the fluence effect. In this thesis, different deep
learning models were compared and investigated to enable optical absorption coefficient
recovery at a particular wavelength in a non-homogeneous foreground and background
medium. Data-driven models were trained with two-dimensional (2D) Blood vessel and
three-dimensional (3D) numerical breast phantom with highly heterogeneous/realistic
structures to correct for the non-linear optical fluence distribution. The trained deep
learning models like U-Net, FD U-Net, Y-Net, FD Y-Net, Deep ResUnet, and GAN
were tested to evaluate the performance of optical absorption coefficient recovery with
in-silico and in-vivo dataset. The results indicated that DL-based deconvolution improves
the reconstructed PAI in terms of PSNR and SSIM. Further, it was observed
that DL models can indeed highlight deep-seated structures with higher contrast due
to fluence compensation. Importantly, the DL models were found to be about 17 times
faster than solving diffusion equation for fluence correction and also able to compensate
for nonlinear optical fluence distribution more effectively and improve the photoacoustic
image quality
Improved Understanding of Standing Waves in Single Layer Coils and Elegant Methods to Estimate Transformer Winding Parameters
No full textAnalyzing the effect of impulse voltages (like lightning, switching) on transformer winding has occupied centerstage in core electrical engineering research for over a century. These investigations gather great significance and relevance as it eventually governs the design of insulation in the winding. Notwithstanding the colossal contribution this domain has witnessed from stalwarts in the past century, a closer scrutiny surprisingly reveals that there still exists tiny grey areas that demands attention. Pursuing this line of thought, the first part of this thesis aims to clearly describe what this grey area is and resolving it would provide a deeper insight about fundamental understanding of surge response in transformer windings – with special emphasis on its standing wave phenomenon. Following this, in the latter part, elegant procedures are stitched together to determine a few electrical parameters of the transformer winding equivalent circuit that have the potential to help in assessing mechanical status of windings. Objectives of the thesis are -
1. Formulate an analytical method to determine the exact shape of standing waves for all modes in a uniform single layer coil as a solution of its governing partial differential equation
2. Estimate series capacitance of a uniform transformer winding from its measured driving point impedance
3. Determine effective air-core inductance of an iron-core uniform winding as a function of its axial length from measured driving point impedance
First part of the thesis revisits a century-old classical theory of standing waves on uniform single layer coils. Accurate information about natural frequencies and shapes of the corresponding standing waves are essential for gaining a deeper understanding of the response of coils to impulse excitations. Analytical studies on coils have largely been based on the assumption that standing waves are sinusoids in both space and time. However, this contradicts the results from numerical circuit analysis and practical measurements. So, this thesis attempts to bridge this discrepancy by revisiting the classical standing wave phenomena in coils. It not only assesses the reason for the aforementioned inconsistency, but also makes a contribution by analytically deriving the exact mode shape of standing waves for both neutral open/short conditions. For this, the coil is modelled as a distributed network of elemental inductances and capacitances, while an exponential function describes the spatial variation of mutual inductance between turns. Initially, an elegant derivation of the governing partial differential equation (in terms of voltage as the variable instead of flux) for surge distribution is presented and to the best of our knowledge, for the first time, an analytical solution for the same has been found by the variable-separable method to find the complete solution (sum of time and spatial terms). Hyperbolic terms in the spatial part of the solution have always been neglected but are included here, thus, yielding the exact mode shapes. For verification, both voltage and current standing waves computed from the analytical solution were plotted and compared with PSPICE simulation results on a 100-section ladder network representing a uniform single-layer coil. Then, practical measurements were made on a tailor-made large-sized single layer coil with a length of 2.02 m, diameter of ~1 m and having 640 turns. It turns out that even in such simple single layer coils, the shape of standing waves of all modes deviates considerably from being sinusoidal. It was further observed that this deviation depends on spatial variation of mutual inductance, capacitive coupling, and order of the standing waves.
In the second part, an elegant method for determining the series capacitance (Cs) and air-core equivalent inductance of a uniform winding as a function of its axial length (termed as M0x in this thesis) of a uniform transformer winding, from its measured DPI magnitude, is discussed. Knowledge about the series capacitance of the winding is essential, which along with shunt capacitance, determines the initial impulse voltage distribution when a surge impinges on the winding. Unlike previously published approaches, the proposed method does not involve any cumbersome and time-consuming curve-fitting or running of optimization/search algorithms. Neither does it require winding geometry data. The proposed procedure for finding series capacitance relies on a property that is observable in the driving point impedance (DPI) function of a lossless winding with an open neutral condition, viz., the ratio of the product of squares of open circuit natural frequencies to the product of squares of short circuit natural frequencies bears a particular relation to the coefficients of the DPI function. A simple procedure involving a deft manipulation and combination of a few well-known properties that correlate the roots of a polynomial to its coefficients are then utilized for determining series capacitance.
Knowledge about equivalent air-core inductance distribution as a function of its axial length (i.e., M0x) is useful for localizing a minor/incipient mechanical fault in the winding. A physically realizable empirical relationship to estimate M0x is initially proposed. The corresponding constants of the empirical relationship are then calculated from the measured DPI. The proposed method requires three DPI measurements: one with neutral-end open and the other with neutral-end shorted. The third DPI is measured with a known external lumped capacitance connected between the neutral and ground. This method requires only the first few dominant natural frequencies observable in the first two of the DPIs.
Feasibility of proposed methods for estimating Cs and M0x was initially verified by simulation on an N-section ladder network and then by experiments on small-sized continuous-disk and interleaved-disk windings, and finally on a large-sized 33 kV, 3.5 MVA continuous-disk winding. Salient features of the proposed methods are – they are simple, elegant and involve minimum post-processing after measuring the DPI. Given its inherent simplicity and their relevance, the author is hopeful that industry will come forward to implement these procedures on an existing FRA measuring instruments – thus opening a new dimension to FRA measurements
Algorithms for Online Learning in Structured Environments
No full textOnline learning deals with the study of making decisions sequentially using information gathered along the way. Typical goals of an online learning agent can be to maximize the reward gained during learning or to identify the best possible action to take with the maximum (expected) reward. We study this problem in the setting where the environment has some inbuilt structure. This structure can be exploited by the learning agent while making decisions to accelerate the process of learning from data.
We study a number of such problems in this dissertation. We begin with regret minimization for multi-user online recommendation where the expected user-item reward matrix has low rank (much smaller than the number of users or items). We address the cold-start problem in recommendation systems where the agent initially has no information about the reward matrix and only gathers information gradually by interactions (i.e., recommending items) with arriving users. We use results from low-rank matrix estimation to design an efficient online algorithm to exploit the low rank of the underlying reward matrix. We analyze this algorithm and show that it enjoys better regret than algorithms that do not take the low-rank structure into account.
We then study the problem of pure exploration (or best arm identification) in linear bandits. In this problem, each time the learner chooses an action vector (arm), she receives a noisy realization of a reward whose mean is linearly dependent on an unknown vector and the chosen action. The aim here is to identify the arm which yields the maximum reward (in expectation) as quickly as possible. We are specifically interested in the situation where the ambient dimension of the unknown parameter vector is very small as compared to the number of arms. We show that by using this inherent problem structure, one can design provably optimal and efficient algorithms to identify the best arm quickly. We study how exploiting the intrinsic geometry of the problem leads to the design of statistically and computationally efficient algorithms for the best arm identification problem.
We finally formulate and study the problem of improper reinforcement learning, where for a given (unknown) Markov Decision Process (MDP), we are given a bag of (pre-designed) controllers or policies for the MDP. We study how the agent can leverage (combine) these pre-trained base controllers (instead of just the rudimentary actions of the MDP) to accelerate learning. This can be useful in tuning across ensembles of controllers, learning in mismatched or simulated environments, etc., to obtain a good controller for a given target environment with relatively few trials. This differs from the usual reinforcement learning setup where the learner observes the current state of the environment and chooses an action to play. In contrast, the improper learner chooses a given base controller and plays whichever action is recommended by the chosen controller. This indirect selection of actions via the base controllers helps to inherit desirable properties (e.g., interpretability, principled design, safety, etc) into the learned policy
Transcriptional regulation of a microRNA encoding gene MIR319C during leaf development in Arabidopsis thaliana
No full textThe evolutionarily conserved microRNA miR319 and its target transcription factors encoded by five CIN-TCPs (TCP2, 3, 4, 10 & 24) regulate leaf morphogenesis in Arabidopsis by triggering the division to differentiation switch of the leaf cells. In a young leaf, the expression of the miR319 encoding gene MIR319C is restricted at the basal region coinciding with the cell proliferation zone, whereas the CIN-TCP transcripts are detected in the more distal region where differentiation is initiated. How the complementary expression patterns of MIR319C and CIN-TCPs are established in leaf primordia is unknown. Moreover, the factors that activate and maintain MIR319C expression in the leaf primordia are yet to be uncovered.
Here, a detailed spatiotemporal analysis of the predominantly expressed TCP4 and MIR319C genes suggested the possibility of CIN-TCP mediated downregulation of MIR319C promoter activity in the leaf primordia. Loss of multiple CIN-TCPs resulted in the distal extension of the MIR319C expression domain, whereas ectopic TCP4 activity restricted the MIR319C domain more proximally. TCP4 was enriched at the MIR319C promoter, and increased TCP4 activity enhanced the deposition of H3K27me3 repressive marks on the MIR319C. Additionally, transgenic lines carrying mutations in TCP binding sites on MIR319C promoter exhibited miR319 overexpression phenotypes. Together with the previous knowledge that miR319 degrades CIN-TCP transcripts, our study suggests the existence of a double-negative feedback loop involving the miR319-CIN-TCP module in regulating leaf morphogenesis in Arabidopsis.
To uncover the activators of MIR319C in leaf primordia, we screened a leaf-specific Arabidopsis transcription factor (TF) library using a yeast one-hybrid assay to isolate proteins that bind to the 2.7 kb promoter of MIR319C. The screen yielded 57 positives including the six NAM/ATAF1/ATAF2/CUC (NAC) domain-containing TFs with DNA-binding preferences similar to that of the CUC sub-group of NAC TFs, i.e., CUC1, 2 & 3. In addition to the ability of the CUC proteins to bind to the MIR319C promoter region in yeast, the expression domain of CUC2 overlaps with that of MIR319C in early leaf primordia, suggesting a role for CUCs in the activation of MIR319C during leaf development. Loss of CUC2 activity significantly reduced the MIR319C expression domain, whereas increased CUC2 level led to a distal expansion of MIR319C expression. Elevated CUC2 level partly rescued the TCP4-mediated suppression of MIR319C expression suggesting that CUC2 and TCP4 interact to establish the domain of MIR319C expression in leaf primordia. Thus, we have identified CUCs as the activators of MIR319C in the leaf primordia.
In conclusion, we propose a model where the CUC proteins initially activate MIR319C throughout early leaf primordia. As development progresses, the CIN-TCP genes are expressed towards the distal end of the primordia by the action of yet unidentified factors, and the onset of CIN-TCP activity results in the downregulation of MIR319C transcription in the distal primordia, possibly by recruiting chromatin modifiers. Strong CUC activity at the base sustains MIR319C expression in the proximal region, where CIN-TCP transcripts are degraded by mature miR319. Thus, our study provides evidence that a CUC-MIR319C-CIN-TCP module patterns a uniformly growing leaf primordium into the proximal and the distal growth domains, where the cells in the basal region continue to divide and grow, whereas cells in the distal region stop dividing and start differentiating
Understanding the role of mtHsp70 in regulating mitochondrial homeostasis: revealing its significance in Congenital Sideroblastic Anemia progression
No full textMitochondria are ubiquitous organelles placed at the nexus of several metabolic and signaling pathways essential for cell survival. Therefore, maintaining a healthy and functional organelle becomes paramount for the cell. The complex structural organization and the bi-genomic nature of the mitochondria pose a significant challenge in maintaining their homeostasis. In addition to the proteins encoded by the mitochondrial DNA (mtDNA), the mito-proteome primarily consists of nuclear-encoded proteins synthesized in the cytosol and subsequently translocated to the mitochondria. Thus, the biogenesis and functioning of the mitochondria are dependent on the efficient transport, folding, and localization of the nuclear-encoded proteins. Any disruptions in this chain of events can be detrimental to the mitochondria and, thereby, to the cell. As a result, several quality-control mechanisms have evolved that operate at multiple levels to abate any mitochondrial damage due to internal and external cellular stress.
Within the mitochondria, the import, folding, targeting, and degradation of proteins are regulated by the molecular chaperones. Among these, the mitochondrial Hsp70 (mtHsp70) is a crucial mediator of protein quality control. In conjunction with multiple co-chaperones, mtHsp70 performs two critical functions: the vectorial import of the nascent polypeptides into the mitochondria and their subsequent folding within the matrix. At the organellar level, a quality check is monitored by the segregation and degradation of superfluous or dysfunctional mitochondria via the process of mitophagy. This is achieved by the concerted action of AuTophaGy (ATG) related proteins and the dynamics of the mitochondrial network. Interestingly, studies reveal that increased mitophagy mitigates the effects of mtHsp70 mutations identified in patients with Parkinson’s disease, thus, suggesting an overlap between the quality control pathways. However, the details and implications of this interaction remain unexplored. Thus, in the current study, we have employed an array of genetic and biochemical techniques in the yeast model system to understand the overlap between the quality checkpoints and, further, to delineate the involvement of mtHsp70-mediated quality control in the progression of Congenital Sideroblastic Anemia (CSA).
We have explored how mtHsp70-mediated quality control engages and responds to the abrogation of mitophagy. Utilizing an unbiased genetic screen, we have identified mtHsp70 mutants that exhibit compromised growth without the mitophagy receptor, Atg32. This is accompanied by an alteration in the mitochondrial physiology, general autophagy, lipid homeostasis, and redox balance overall, resulting in a reduction in cellular lifespan. Our findings highlight the role of mtHsp70 in maintaining mitochondrial integrity under stress conditions and underscore the need for an overlap between the quality control pathways.
Further, we have investigated the role of mtHsp70 in the onset and progression of Congenital Sideroblastic Anemia (CSA), a hereditary blood disorder characterized by the accumulation of iron-laden mitochondria. Preliminary analyses of analogous mutations in the yeast mtHsp70 reveal perturbations in the mitochondrial network and functionality. Further, we observe mutations in mtHsp70 impair its import and chaperone activity resulting in a loss of function that manifests as the disease phenotypes observed in Congenital Sideroblastic Anemia.
The current study provides insights into the interaction between the various mitochondrial quality checkpoints and highlights the relevance of protein quality control in the context of Congenital Sideroblastic Anemia progression.Indian Institute of Scienc
Study on optical and electrical transport properties of twisted bilayer transition metal dichalcogenides
No full textVan der Waals (vdW) heterostructures, where dissimilar atomically thin vdW crystals are vertically assembled, have initiated a new paradigm to create flexible multifunctional devices. Despite the weak nature of vdW interactions, unusually strong interlayer coupling and hybridization in these heterostructures lead to novel physical phenomena ranging from interfacial stress fields to modification of electronic band structure. In twisted van der Waals heterostructures (vdWHs), the angular mismatch between two similar lattices generates a large-scale interference pattern, known as the moiré pattern, which strongly impacts the electronic band structure of the superlattice. The moiré patterns in vdWHs create a periodic potential for electrons and excitons to yield many interesting phenomena such as Hofstadter butterfly spectrum, moiré excitons, tunable Mott insulator phases, unconventional superconductivity. In this thesis, we study the effects of moiré patterns on twisted TMDC bilayers by using Raman and PL measurements and try to probe the modified electronic properties in moiré superlattice through transport measurements.
The relative rotation between the adjacent layers or the twist angle between them plays a crucial role in changing the electronic band structure of the superlattice. The first part of the thesis attempts to create such twisted TMDC bilayers with highly accurate twist angle. The assembly of multi-layers of precisely twisted two-dimensional layered materials requires knowledge of the atomic structure at the edge of the flake. Here, we demonstrate a simple and elegant transfer protocol using only optical microscope as an edge identifier tool, using which controlled transfer of twisted homobilayer and heterobilayer transition metal dichalcogenides is performed with close to 100 % yield. The fabricated twisted van der Waals heterostructures have been characterized by SHG, Raman spectroscopy, and photoluminescence spectroscopy, confirming the desired twist angle within 0.50 accuracy. The presented method is reliable, and quick, and prevents the use of invasive tools, which is desirable for reproducible device functionalities.
Next, we study the phonon renormalization in twisted bilayer MoS2, which adds insight into the moiré physics. The interlayer coupling in these heterostructures is sensitive to twist angles (θ) and is key to controllably tuning several exotic properties. We demonstrate a systematic evolution of the interlayer coupling strength with twist angle in bilayer MoS2 using a combination of Raman spectroscopy and classical simulations. At zero doping, we show a monotonic increment of the separation between the A1g and E2g mode frequencies as θ decreases from 100 to 10, which saturates to that for a bilayer at small twist angles. Furthermore, we use doping-dependent Raman spectroscopy to reveal the θ-dependent softening and broadening of the A1g mode, whereas the E2g mode remains unaffected. Using first principles-based simulations, we demonstrate large (weak) electron-phonon coupling for the A1g (E2g) mode, explaining the observed trends. Our study provides a non-destructive way to characterize the twist angle and the interlayer coupling and establishes the manipulation of phonons in twisted bilayer MoS2 (twistnonics).
Besides the closely aligned moiré lattice, intermediate misorientation (twist angles > 150) bilayers also offer a unique opportunity to tune excitonic behavior within these concurrent physical mechanisms but are seldom studied. To explore the light-matter interaction at an intermediate angle, we measure many-body excitonic complexes in monolayer (ML), natural bilayer (BL), and twisted bilayer (tBL) WSe2. Neutral biexciton (XX) is observed in tBL for the first time while being undetected in non-encapsulated ML and BL, demonstrating the unique effects of disorder screening in twisted bilayers. The XX, as well as charged biexciton (XX-), are robust to thermal dissociation and are controllable by electrostatic doping. Vanishing of momentum indirect interlayer excitons with increasing electron doping is demonstrated in tBL, resulting from the near-alignment of Q'-K and K-K valleys. Intermediate misorientation samples offer a high degree of control of excitonic complexes while offering possibilities for studying exciton-phonon coupling, band alignment, and screening.
Finally, we investigate the electrical transport in Gr/tWSe2 heterostructure, using graphene as a sensing layer to probe the electronic effects of the underlying twisted TMDC structure on monolayer graphene. Unlike graphene, TMDC materials show a massive contact resistance. We tried to solve this issue by using different work function materials to reduce the Schottky barrier across the metal-semiconductor junction. However, getting an ohmic contact between the metal-semiconductor junction is a difficult technological challenge. We resolve this issue by using graphene as a sensing layer in monolayer graphene/twisted bilayer WSe2-based heterostructures. We observe the ferroelectricity in the sample, which can be understood by the presence of moiré ferroelectric domains in twisted TMDC lattice. We find that the polarization switching can be controlled through the vertical electric field. We also find a huge nonlocal signal in graphene at a zero magnetic field that can't be explained via classical contribution. Both the nonlocal and local resistance can be controlled through the electric field. We further explore the magnetotransport properties of the system and find that the magnetoresistance of the sample increases with an in-plane magnetic field
Deploying an Orthogonal Turn for Isolation of Rare Cell Populations from Vascular Environments
No full textCancer cells are shed from metastatic primary tumors, bearing the potential for blood-borne metastasis to distant vital organs. Metastasis is predominantly responsible for cancer-related deaths. These circulating tumor cells (also known as CTCs) are highly invasive and can be present in the vasculature as single cells or as clusters. CTCs have been proposed as an important biomarker to assess the aggressiveness of cancer, the effectiveness of the treatment, and disease progression. Although most of the cells have epithelial receptors on their surface, such as Epithelial Cell Adhesion Molecules (EpCAMs), the molecular diversity on the surfaces of such cells is still not completely characterized. The circulating tumor cells (or CTCs) tend to be larger in size and higher in density with respect to the rest of the blood cells and in their density. CTCs are extraordinarily rare, i.e., one among a billion blood cells which make their isolation difficult.
Existing technologies carry out CTC separation by either inducing external forces (active separation) or using intrinsic hydrodynamic forces (passive separation). In active separation, the external forces have to be larger than the flow-induced forces, which results in a limited throughput. Moreover, these techniques often involve biomarkers and labelling agents (usually EpCAM antibodies), which not only puts a question on the viability of the captured cells but also might fail to work for CTCs that do not express such markers. The passive separation method is carried out by simply controlling the hydrodynamic properties of the flow. Of these, inertial microfluidics separation has high throughput, whereby large sample volumes can be processed in a short time. High throughput vortex trapping and CTC separation has been described by Di Carlo et al. in their vortex-chip technology. However, the operation of their device requires drastically high flow velocities (particle velocity ̴ 4 m/s), which are prone to damage the cells and affect their viability.
In this dissertation, an inertial microfluidic vortex chip incorporating an orthogonal turn is investigated for the isolation and separation of CTCs. These chips function at significantly lower (38% of previously reported) flow velocities. Fluid flowing through the chip is constrained to exit the trapping chamber at right angles to that of its entry. Such a flow configuration leads to the formation of a vortex in the chamber and above a critical flow velocity, larger particles are trapped in the vortex, whereas smaller particles get ejected with the flow: we call this phenomenon the turn-effect.
I explain how different forces contribute to the turn-effect in the orthogonal design by acting on cells, and pushing them into specific vortices in a size- and velocity-dependent fashion. Furthermore, we have characterized the critical velocities for trapping particles of different sizes on chips with distinct entry-exit configurations. Optimal architectures for stable vortex trapping at low flow velocities are identified using polystyrene beads and blood cells.
Subsequently, I demonstrated selective trapping of human breast cancer cells mixed with whole blood at low concentrations. An isolation protocol to separate the trapped particles was developed and optimized on a scaled-up device that uses serialization and parallelization. After isolating spiked circulating cancer cells from diluted blood, we were also able to culture them.
In summary, a label-free inertial microfluidic vortex trapping setup incorporating an orthogonal turn was developed and optimized for the size-based gentle separation of CTCs which are larger and rarer than other blood cells. Some further design modifications were also suggested in the latter part of the work to increase the efficiency of enrichment
Employing Arynes in Unique Multicomponent Reactions and a Strain-Release Ene Reaction
No full textArynes are highly reactive electrophilic intermediates, extensively used for synthesizing various benzo-fused heterocycles and 1,2-disubstituted benzenes. Among the numerous reactions of arynes, multicomponent coupling (MCC) has been widely used to synthesize versatile and structurally complex 1,2-disubstituted benzenes. In a conventional aryne MCC, the nucleophile having no acidic hydrogen adds to the highly electrophilic aryne generating the aryl anion intermediate, which is subsequently intercepted by the electrophile to furnish 1,2-disubstituted arenes. In this context, Mr. Subrata Bhattacharjee has demonstrated the synthetic potential of KI as the nucleophilic iodide source in aryne three-component coupling using aldehydes as the third component. This mild and transition-metal-free coupling reaction allowed the straightforward synthesis of 2-iodobenzyl alcohols in moderate to good yields with good functional group compatibility. Mr. Subrata could also use KBr and KCl as the nucleophilic trigger and N-methylisatin and CO2 as the electrophilic third components in this aryne multicomponent coupling (MCC).
In the vast realm of transition-metal-free aryne MCCs, only aprotic nucleophiles are employed as triggers, and the use of protic nucleophiles is not demonstrated. This is because, in many cases, the use of protic nucleophiles leads to the protonation of the generated aryl anion intermediates leading to the arylation. In this line, Mr. Bhattacharjee has employed thiophenols as a protic nucleophilic trigger in the transition-metal-free and Grignard reagent-free three-component coupling involving arynes. Employing aldehydes as the third component, the reaction allowed the mild and broad scope synthesis of 2-arylthio benzyl alcohol derivatives in good yields. Moreover, Mr. Subrata has extended the reaction to selenophenol as the nucleophilic trigger, and activated ketones as the third component.
Mr. Bhattacharjee has also synthesised biologically important S-aryl dithiocarbamates by the aryne three component coupling involving CS2 and aliphatic amines. This transition-metal-free and mild reaction is scalable and operates with good functional group compatibility. Contrary to the known aryne MCCs involving amines and CO2, this reaction proceeds via the initial addition of amines to CS2 followed by the trapping with arynes to furnish the desired products. Mr. Bhattacharjee has presented detailed experiments and DFT studies to get insight into the mode of addition and product formation. Moreover, using 3-triflyloxybenzyne, a four-component coupling with the incorporation of THF was also presented.
The aryne ene reaction, although well-known, has not enjoyed widespread use in synthesis, primarily because of poor selectivity and low yields due to the competing [4+2], [2+2] pathways. However, the emergence of contemporary methods to generate arynes under mild conditions has led to significant advancements in the aryne ene reaction. Over the past two decades, a significant number of efficient and selective processes have been described, employing a wide range of substrates, including alkenes, alkynes, allenes etc. In this context, Mr. Bhattacharjee have demonstrated a unique strain-release ene reaction bicyclo[1.1.0]butanes (BCBs) employing arynes by utilizing the substantial π-character of the strained central C–C bond of the BCBs. The ene reaction proceeds under mild conditions and results in the formation of substituted cyclobutene derivatives in excellent yields
Self-Supervised Domain Adaptation Frameworks for Computer Vision Tasks
No full textThere is a strong incentive to build intelligent machines that can understand and adapt to changes in the visual world without human supervision. While humans and animals learn to perceive the world on their own, almost all state-of-the-art vision systems heavily rely on external supervision from millions of manually annotated training examples. Gathering such large-scale manual annotations for structured vision tasks, such as monocular depth estimation, scene segmentation, human pose estimation, faces several practical limitations. Usually, the annotations are gathered in two broad ways; 1) via specialized instruments (sensors) or laboratory setups, 2) via manual annotations. Both processes have several drawbacks. While human annotations are expensive, scarce, or error-prone; instrument-based annotations are often noisy or limited to specific laboratory environments. Such limitations not only stand as a major bottleneck in our efforts to gather unambiguous ground-truth but also limit the diversity in the collected labeled dataset. This motivates us to develop innovative ways to utilize synthetic environments to create labeled synthetic datasets with noise-free unambiguous ground-truths. However, the performance of models trained on such synthetic data markedly degrades when tested on real-world samples due to input distribution shift (a.k.a. domain shift). Unsupervised domain adaptation (DA) seeks learning techniques that can minimize the domain discrepancy between a labeled source and an unlabeled target. However, it mostly remains unexplored for challenging structured prediction based vision tasks.
Motivated by the above observations, my research focuses on addressing the following key aspects: (1) Developing algorithms that support improved transferability to domain and task shifts, (2) Leveraging inter-entity or cross-modal relationships to develop self-supervised objectives, and (3) Instilling natural priors to constrain the model output within the realm of natural distributions.
First, we present AdaDepth - an unsupervised domain adaptation (DA) strategy for the pixel-wise regression task of monocular depth estimation. Mode collapse is a common phenomenon observed during adversarial training in the absence of paired supervision. Without access to target depth-maps, we address this challenge using a novel content congruent regularization technique. In a follow-up work, we introduced UM-Adapt, a unified framework to address two distinct objectives in a multi-task adaptation framework, i.e., a) achieving balanced performance across all tasks and b) performing domain adaptation in an unsupervised setting. This is realized using two novel regularization strategies; Contour-based content regularization and exploitation of inter-task coherency using a novel cross-task distillation module. Moving forward, we identified certain key issues in existing domain adaptation algorithms that hinder their practical deployability to a large extent. Existing approaches demand the coexistence of source and target data, which is highly impractical in scenarios where data-sharing is restricted due to proprietary or privacy concerns. To address this, we propose a new setting termed as Source-Free DA and tailored learning protocols for the dense prediction task of semantic segmentation and image classification in both with and without category shift scenarios.
Further, we investigate the problem of Self-supervised Domain Adaptation for the challenging monocular 3D human pose estimation task. The key differentiating factor in our approach is the idea of infusing model-based structural prior as a means to constrain the pose estimation predictions within the realm of natural pose and shape distributions. Towards self-supervised learning, our contribution lies in the effective use of new inter-entity relationships to discern the co-salient foreground appearance and thereby the corresponding pose from just a pair of images having diverse backgrounds. Unlike self-supervised solutions that aim for better generalization, self-adaptive solutions aim for target-specific adaptation, i.e., adaptation to deployment-specific environmental attributes. To this end, we propose a self-adaptive method to align the latent space of human pose from unpaired image-to-latent and the pose-to-latent, by enforcing well-formed non-local latent space rules available for unpaired image (or video) and pose (or motion) domains. This idea of non-local relation distillation against the broadly employed general contrastive learning techniques shows significant improvements in the self-adaptation performance. Further, in a recent work, we propose a novel way to effectively utilize uncertainty estimation for out-of-distribution (OOD) detection, and thus enabling inference-time self-adaptation. The ability to discern OOD samples allows a model to assess when to perform re-adaptation while deployed in a continually changing environment. Such solutions are in high demand for enabling effective real-world deployment across various industries, from virtual and augmented reality to gaming and health-care applications
Charge Density Wave-driven Carrier Transport in Layered Heterostructures
No full textMetal-based electronics remain one of the longstanding goals of researchers to achieve ultra-fast and radiation-hard electronic circuits. Generally, metals are primarily used as passive conductors in modern electronics and do not play an active role. Nanoscale materials with distinctive size-dependent properties provide opportunities to achieve new device functionalities. Ta-based di-chalcogenides, particularly 1T-TaS2 and 2H-TaSe2, which form layered structures and exhibit charge density waves (CDW), are promising in this context. CDW is a macroscopic state shown by materials with reduced dimensions, for example, one-dimensional and layered two-dimensional crystals. It results from the modulation in the electronic charge arising due to a periodic modulation in the crystal lattice.
1T-TaS2 exhibits one of the strongest known CDW characteristics enabling temperature-dependent distinct resistivity phases. The nearly commensurate (NC) to the incommensurate (IC) CDW phase transition that usually occurs at 353 K and can be driven electrically at room temperature is of high practical interest. However, resistivity switching during this phase transition is weak (< 2) and cannot be modulated by an external gate voltage – limiting its widespread usage. Using a back-gated 1T-TaS2/2H-MoS2 heterojunction, we show resistivity switching up to 17.3, which is ~14.5-fold higher than standalone TaS2. We demonstrate a low barrier electrical contact between a TaS2 source and a MoS2 channel, promising “all-2D” flexible electronics. Additionally, we show that the usual resistivity switching in TaS2 due to different phase transitions is accompanied by a surprisingly strong modulation in the Schottky barrier height (SBH) at the TaS2/MoS2 interface – providing an additional knob to control the degree of the phase-transition-driven resistivity switching by an external gate voltage. In particular, the commensurate (C) to triclinic (T) CDW phase transition increases the SBH owing to a collapse of the Mott gap in TaS2. The change in SBH allows us to estimate an electrical Mott gap opening of ~71 ± 7 meV in the C phase of TaS2. The results show a promising pathway to externally control and amplify the CDW induced resistivity switching.
Further, we achieve gate- and light-controlled negative differential resistance (NDR) characteristics in an asymmetric 1T-TaS2/2H-MoS2 T-junction by exploiting the electrically driven CDW phase transition of TaS2. The device operation is purely governed by majority charge carriers, making it distinct from typical tunneling-based NDR devices, thus avoiding the bottleneck of weak tunneling efficiency in van der Waals heterojunctions. Consequently, we achieve a peak current density over 10^5 nA μm^(-2), which is about two orders of magnitude higher than that obtained in typical layered material-based NDR implementations. An external gate voltage and photo-gating can effectively tune the peak current density. The device characteristics show a peak-to-valley current ratio (PVCR) of 1.06 at 290 K, increasing to 1.59 at 180 K. To exploit the low thermal conductivity of 1T-TaS2 and 2H-TaSe2 in a local heater structure, we insert 2H-TaSe2 in between TaS2 and MoS2 layers, thereby forming a triple-layered 1T-TaS2/2H-TaSe2/2H-MoS2 T-junction. TaSe2 acts as a buffer layer preventing the CDW-induced SBH modulation at TaS2/MoS2 interface. This will allow efficient thermionic switching of carriers resulting from sharp temperature rise in the junction due to electrically driven TaS2 phase transitions. Interestingly, the device can toggle between the current increment and NDR characteristics by simply changing the biasing conditions. At TaS2 biasing, the heterostructure device shows a current increment by a factor of 3 at 300 K, which gets enhanced up to ~10^3 at 77 K, beneficial for various switching circuits and sensing applications. However, under TaSe2 biasing, the device exhibits NDR characteristics with a PVCR of 1.04 and 1.10 at 300 K and 77 K, respectively. The external back-gate voltage can effectively tune the current enhancement factor and NDR. The devices mentioned above are robust against ambiance-induced degradation, and the characteristics repeat in multiple measurements over more than six months.
Conventional metals, in general, do not exhibit strong photoluminescence. However, we found that 2H-TaSe2 exhibits a surprisingly strong optical absorption and photoluminescence resulting from inter-band transitions. We use this perfect combination of electrical and optical properties in several optoelectronic applications. We show a seven-fold enhancement in the photoluminescence intensity of otherwise weakly luminescent multi-layer MoS2 through non-radiative resonant energy transfer from TaSe2 transition dipoles. Using a combination of scanning photocurrent and time-resolved photoluminescence measurements, we also show that the hot electrons generated by light absorption in TaSe2 have a relatively long lifetime, unlike conventional metals, making TaSe2 an excellent hot-electron injector. Finally, we show a vertical TaSe2/MoS2/graphene photodetector demonstrating a responsivity greater than 10 AW^(-1) at 0.1 MHz - one of the fastest reported photodetectors using MoS2.
The findings will boost device applications that exploit CDW phase transitions, such as ultra-broadband photodetection, negative differential conductance, thermal sensors, fast oscillator, and threshold switching in neuromorphic chips. These functionalities will enable the implementation of active metal-based circuits