https://etd.iisc.ac.in/handle/2005/25 2026-04-17T22:23:27Z https://etd.iisc.ac.in/handle/2005/6199 2D Piezotronics: Performance to Functionality Yarajena, Sai Saraswathi In the pursuit of interactive electronic devices, there is a need for smart materials which can serve multiple functionalities. 2D (two-dimensional) layered materials have gained attention in semiconductor technology because of their versatile electrical and optical properties. Furthermore, some materials exhibit piezoelectricity at 2D scale and can withstand enormous strain. These properties make them suitable as smart materials involving electromechanical signals. In the literature, materials which are semiconducting and piezoelectric are termed piezotronic (piezo+electronic) materials. Theoretical studies have indicated many materials as piezoelectric in 2D form. However, experimental tools to investigate the extent of piezoelectric coupling in 2D materials are limited, and their relevance for piezotronics has not been studied in detail. This dissertation presents some key aspects of 2D Piezotronics for improved performance and to achieve additional functionalities with heterojunctions. The work constitutes proposing a technique to estimate piezoelectric coupling coefficients, choice of flexible substrates for piezotronics, methods to reduce the charge screening effects, measurement strategies to extract the actual piezoelectric output from the bending measurements, and the study of heterojunctions for rectifying behaviour. In this work, Molybdenum disulfide (MoS2) is used as active piezoelectric material. In the initial part of the work, I propose a technique to estimate in-plane piezoelectric coupling quantitatively for 2D materials. The method involves a novel approach for in-plane field excitation in lateral Piezo force microscopy (PFM). Contact resonance gain of the tip-sample system is leveraged to measure the piezoelectric coupling coefficients in a few pm/V to sub pm/V range. However, I have shown that operating PFM at contact resonance can cause pseudo piezoelectric signals. Therefore, a detailed methodology for signal calibration and electrostatic background subtraction is developed in this work. The technique is verified by estimating the in-plane piezoelectric coupling coefficients (d11) for freely suspended MoS2 of one to five atomic layers. The technique presented is useful in estimating the piezoelectric coupling strengths in emerging 2D materials. Piezotronic devices are made on flexible substrates for practical applications. Fabrication on flexible substrates often poses great difficulties in handling them, depositing inorganic materials, and carrying out lithography processes. I propose the commercially available nano flex film as a prospective substrate for piezotronics. Carrying out fabrication on these substrates is as seamless as that on rigid substrates. Substrates such as PET, Nano flex and TPU can be used for low-temperature (<150 deg C) applications. Kapton is one of the flexible substrates that can handle higher temperatures(>200 deg C). However, they tend to twist when heated, making the fabrication difficult. I have proposed a gel-based bonding for the Kapton substrates wherein the debonding process is automatic. The method is helpful for the fabrication of 2D material devices on Kapton. Besides selecting the substrates, suitable base layers and passivation techniques are studied to reduce the charge screening effects and thus improve the performance of piezotronic devices. It is verified that open circuit voltages and strain gauge factors obtained for the current monolayer MoS2 device on SiO2 are three folds higher than those presented in the literature. A simple measurement setup which does not require probe needles or wire bonding is developed for the bending strain measurements. The open circuit voltage and short circuit current signals obtained from a single 2D material device are very small. The noise signals that originate from various triboelectric and electrostatic sources of the measurement setup can be of similar magnitude. Consequently, the electrical outputs from these devices during bending measurements are often misinterpreted. Thus, it is essential to analyse various noise sources in bending measurements. I then discuss ways to reduce the background noise and identify the valid piezoelectric output. Finally, I have studied some homogeneous and heterogeneous junctions of MoS2 to achieve good rectifying junction behaviour, which can add extra functionalities for piezotronics. The rectification ratio values as high as 5000 could be achieved at 1 V bias. Besides the rectifying ratio, I have observed that the heterojunctions of MoS2 and MoTe2 have superior piezoelectric behaviour compared to other 2D material junctions reported so far with open circuit voltages as high as ~1 V and peak power density of ~200 mW/m2 at 0.44% bending strain. Formation of the p-n and Schottky junction hybrid in MoS2-MoTe2 heterojunction could achieve high rectification ratios and open circuit voltages and is fascinating for further study. https://etd.iisc.ac.in/handle/2005/4668 3D Packaging for Integration of Heterogeneous Systems Nittala, Pavani Vamsi Krishna With several new applications getting developed around wearable technologies for Internet of Things (IoT), there has been a growing need for development of the miniaturized systems. Emerging applications in healthcare, structural monitoring, consumer accessories, etc are fuelling the need for these miniaturized hybrid systems. Such micro-nano systems will be enabled through the development of heterogeneous integration technologies that will allow co-packaging of several chips with different functionalities in a single vertical 3D stack. Therefore, the consumer electronics industry has initiated development of 3D integration of CMOS devices in vertical stacks which are electrically interconnected using thru-silicon-via (TSV) technology. This technology is however not suitable for stacks having a complex combination of GaN-HEMT’s, MEMS, microfluidics, optical devices and CMOS. Moreover, due to the cross-contamination issues, most of these devices are never accepted in the standard silicon CMOS foundries. To address these issues, we have developed innovative processing technologies that would allow 3D packaging by the post fab vertical stacking technique, suitable for the packaging industry. In the First Part of the thesis, we have developed processing technologies for the 3D stacking of the homogenous silicon systems. Using them, we have demonstrated a low temperature process to transfer MOS devices on ultra-thin silicon layers (1.5 μm) from a parent substrate to a foreign substrate or stack. In order to enable this transfer, we have analysed and resolved the associated stress issues. Furthermore, we demonstrate three-layer stacking of the ultra-thin silicon layers with functional MOSFET’s in each layer. We extensively characterize the changes in the device performance, which arise due to the transfer process. In the Second Part of the work, we have demonstrated an approach for stacking the III-nitride-on-Si HEMTs and Si-MOSFETs on to a copper substrate. The developed process flow offers a significant improvement in the device behaviour due to the transfer to a thermally conducting substrate like copper. The functional AlGaN/GaN epi-layer stack from the HEMT-on-silicon wafer is lifted-off and bonded to a copper substrate using novel Cu-In bond. Next, an ultra-thin silicon layer (~1.5 μm) with functional NMOS transistors fabricated in-house, on an SOI wafer are separated from the parent SOI wafer and then stacked over the GaN devices already bonded on the copper substrate, using cost-effective epoxy bonding approach. The devices are characterised to study the improvements in their performance. In the Third Part, we have demonstrated a 3D integration method for miniaturisation of hybrid systems. Using this 3D packaging technique, a fluorescence-sensing platform consisting of (i) a silicon photodetector, (ii) plastic optical filters, (iii) commercial LED and (iv) a glass micro-heater chip is demonstrated. We have resolved several fabrication challenges related to planarization, stacking and interconnection of these divergent chips. The above process flow developed in this work, can be scaled to stack a larger number of layers for achieving more complicated systems with enhanced functionality and applications. Finally, we have demonstrated interconnection methodologies using the nonconventional inkjet printing technique for via filling to enable identical die size stacking. https://etd.iisc.ac.in/handle/2005/4073 Accelerated and Accurate Alignment of Short Reads in High Throughput Next Generation Sequencing [NGS] Platforms Natarajan, Santhi The genome of an organism encompasses the unique set of genetic instructions for every individual in a species. The genome, in totality, guides the course of evolution, development, genetic and epigenetic growth factors of an individual. Genomics, the study of genome, presents an interdisciplinary landscape, with a multistage data analytics pipeline. Understanding the genome involves determining the order of the four constituent nucleotides or bases or genetic alphabets, namely adenine (A), cytosine (C), guanine (G) and thymine (T), within the genome’s DNA sequence, and the process is widely known as sequencing. Next Generation Sequencing (NGS) involves massively parallel sequencing of genetic data with high throughput. NGS offers an unparalleled interrogation of the genome, throwing deeper insight into the functional and structural investigation of genetic data. The deductions from such a study leave a huge impact across fields, including medical diagnostics, therapeutics and drug discovery, and as well form the basis for genomic medicine. Data processing with NGS happens over an elaborated multi-stage data analytics pipeline. During the primary data analysis, the sequencing process produces billions of short fragments, called short reads, of the target genome. This amounts to petabytes of unprocessed genomic raw data. Short read mapping (SRM) is the process of mapping these short reads to their respective positions in the target genome. Due to the sheer volume of data that needs to be handled, SRM serves as a major sequential bottleneck to the NGS data analytics pipeline in genomics, and presents profound technical and computing challenges. Classified as a complex big data engineering problem, SRM thus calls for innovative computational, scientific and statistical approaches towards big data analysis. A strict validation of various algorithms and softwares in an NGS pipeline is essential, to ensure reliable and accurate results. With growing volume of NGS big data, the SRM and subsequent analytic steps de-mand a High Performance Computing (HPC) environment for data storage and analyses. Existing solutions for accelerating SRM provide notable performance, while leveraging heuristics and incurring significant error rates. Given the impact of the results of SRM in subsequent diagnostics and therapeutics, such heuristics and error rates are not affordable. In this context, we need precise, affordable, reliable and actionable results from SRM, to support any application, with uncompromised accuracy and performance.In this work, we present a massively parallel and scalable archetype, for accurate alignment of short reads, at a fine-grained single nucleotide resolution. The significant contributions of this work are presented below: 1. We present a robust and efficient indexing scheme for the reference genome, which is devoid of heuristics. The scheme reports all possible regions of mapping for a short read, inclusive of repeat regions. The lookup scheme efficiently handles the redundancy in reads. Though this leaves the rest of the pipeline with more data for SRM as compared to the heuristic solutions, it provides the end user with reliable and actionable results. 2. We present an efficient parallel implementation of an accurate sequence alignment algorithm based on the Dynamic Programming (DP) method. Our alignment kernels can seamlessly perform the traceback process in hardware simultaneously with the forward scan, thus eliminating the computational and memory bottlenecks associated with such algorithms. These kernels thus report alignment in a minimum deterministic time, which forms the first level of acceleration for SRM. 3. We present AccuRA, a hardware accelerator targeting reconfigurable hardware platforms. The model scales well at multiple levels of granularity, which precisely aligns short reads, at a fine-grained single nucleotide resolution, and offers full coverage of the genome. 4. We present GMAccS, a GPGPU based solution, for the SRM accelerator. This employs a platform independent model, capable of targeting a heterogeneous set of GPU hardware. 5. We present a performance and scalability analysis model for both the archetypes. The results from the prototypes substantiate the scalability of these architectures at multiple levels of granularity. 6. We present the generalization of our solution across applications which exhibit similar data patterns in terms of volume, variety, rate of production and analysis, randomness and uncertainty involved in data, and use Approximate String Matching (ASM) as the fundamental operation for data analytics. 7. We present the various problems within the biological domain, in terms of complexity and quantity of data, to which our solution can be customized and scaled, at various levels of granularity. We have presented the results from various prototype models of both AccuRA and GMAccS. The AccuRA prototype, hosting eight kernel units on a single reconfigurable device, aligns short reads with an alignment performance of 20.48 Giga Cell Updates Per Second (GCUPs). AccuRA can be ported onto devices as diverse as SoCs, ASICs or reconfigurable platform based hardware coprocessors or accelerators. The scalability analysis proved to substantiate the parallel AccuRA architecture, making it a promising target to accelerate the SRM process in the NGS pipeline. The in-house supercomputing platform SahasraT, which is a Cray XC40 system, hosted the prototype for the GMAccS archetype. The GMAccS prototypes align with an optimal performance of 23.69 Million Maps Per Second (MMPS) to 528.69 MMPS, while scaling from a single GPU to 24 GPUs. The performance model for GMAccS, as well as the results from the prototypes, substantiates the scalability of the GMAccS archetype and the subsequent performance enhancement achieved by it. Both AccuRA and GMAccS accommodate the big data of genomics, with uncompromised accuracy, precision and performance, while aligning the smaller archeal, bacterial and fungal genomes, to the much larger mammalian human genomes. These models have successfully handled redundant reads and multiread alignments. The results from AccuRA and GMAccS are available in the Sequence Alignment/Map (SAM) format, making it compatible with the downstream applications in the NGS pipeline. With a basic parameterized SRM model, and the results from its various prototypes for small and large genome benchmarks, we have gained the confidence that this solution can serve the requirements of accurate and scalable alignment of NGS big data. We believe that our model can serve as a reliable candidate in the future of genomics, called the "genomic highway", where data belonging to multiple applications can be streamed in, and can be aligned real time, with minimal memory and storage requirements, on a generalized alignment engine. https://etd.iisc.ac.in/handle/2005/6560 Advanced Architectures for Cascaded Raman Fiber Lasers Deheri, Rashmita Fiber lasers have exhibited significant expansion in their diverse applications within the fields of communications, industrial operations, defence, and the medical sector. They require a rare earth doped element as their gain medium to absorb light and re-emit a coherent, intense laser beam. Although the fiber lasers are used for high-power amplifiers ranging from a few watts to kilowatts, the wavelength coverage is limited by the emission spectra of the rare-earth elements. The limitation of wavelength spanning in fiber lasers is overcome by cascaded Raman fiber lasers (CRFLs) are used. It utilizes stimulated Raman scattering (SRS) to produce multiple Raman Stokes orders, thereby generating wavelengths outside the emission spectrum band of rare-earth doped fiber lasers. The CRFL technology has been demonstrated to be suitable for developing high-power, scalable lasers with wavelength agility due to its versatile and compact fiber-based configuration. Despite all these advantages, the wavelength tunability of conventional CRFLs is constrained by the fixed wavelength of input/output highly reflective fiber Bragg gratings (FBGs). Hence, randomly distributed feedback (RDFB) on the CRFLs platform could improve the controllability of wavelength and enable the system to be more adaptable to different wavelengths. This configuration allows for efficient energy transfer and enables broad spectral coverage with high output power. However, there are some limitations to this, such as getting an efficiently desired wavelength spectrum, wavelength tunability reduced spectral purity, and line broadening of the output wavelengths. This thesis explores advanced architectures for CRFLs to overcome the tunability and spectral purity limitations. Previously, using RDFB Raman lasers, wavelength tunability was achieved using a tunable pump laser module. This module enhances the system complexity and increases the overall cost. Here, we proposed a system configuration to address these limitations by allowing the tunability of the output wavelengths using a fixed wavelength pump source. The proposed architecture also enables the tuning of the linewidth of the laser. Our proposed architecture incorporates a reflective Fourier spectral/pulse shaper as an advanced feedback mechanism. The shaper achieved a spectral resolution of 0.5 nm, and loss through the spectral shaper is less than 10 dB for a range of wavelengths from 1100 nm to 1250 nm. This configuration enables filtering out unwanted higher-order Raman Stokes lines, enhancing power in the desired output wavelength. The desired wavelength tunability and linewidth tunability are achieved by using the desired spatial mask patterns at the Fourier plane of the spectral shaper. We demonstrated a CRFL with wavelength tunability over three Raman Stokes orders with spectral purity of > 90% and in-band power of ~10 W. Further, the proposed architecture achieves linewidth tuning over an order of magnitude from ~0.5 nm to > 4 nm. Spectral purity, the power ratio in the desired output wavelength to the power in all other wavelengths, is a key performance measure for laser sources, and many applications require only a single wavelength with high spectral purity. The next part of the thesis proposes methods to enhance the spectral purity of Raman lasers. By analyzing the reasons for spectral purity degradation, we proposed a new architecture that achieves highly spectrally pure random distributed feedback CRFLs over six orders of Raman shifts. The proposed architecture used a narrow linewidth source as a seed with less intensity noise (-147 dBc/Hz from 9 kHz to 10 GHz). The seed is line-broadened by dual-phase modulation, both white noise source and sinusoid using phase modulators, and then amplified with Ytterbium amplifiers, and it is used for Raman conversion. At the output of the Raman fiber laser, we achieved up to 23 W power, tunable from the pump wavelength (1064 nm) all the way to 1480 nm. This approach yields high spectral purity, ~ 99%, over the entire range of Raman conversion. The last part of the thesis explores the measurement and analysis of relative intensity noise (RIN) in Raman fiber lasers. A Raman fiber laser's RIN refers to the fluctuations in the laser output power relative to the average power. Pump RIN influences RIN in CRFLs. We measured the RIN of the pump Raman Stokes orders for a narrow linewidth pumped Raman fiber laser and compared them with the conventional FBG-based pumped Raman fiber laser. The maximum RIN measured for a narrow linewidth pump is less, -128 dBc/Hz compared to a conventional fiber laser of -98 dBc/Hz. There is a 36 dB reduction in low-frequency RIN for phase modulated narrow linewidth pumped Raman Stokes compared to conventional pumped Raman fiber lasers. However, the high-frequency RIN (beyond a few GHz) is the same for both. In addition to the goal of reduction in intensity noise of Raman lasers, an added goal of the intensity noise studies is to investigate methods to create low linewidth Raman lasers with the goal of efficient harmonic conversion to the visible or mid-infrared regions. We investigated whether low-intensity noise Raman lasers have reduced linewidth due to lower self-phase modulation type line-broadening. However, this was not found to be the case. There are primarily two factors that can cause line-broadening; the first one is the inherently broad Raman gain spectrum, and the second is the nonlinear spectral broadening due to the intensity noise. In our case, though the low-frequency (< 1 GHz) RIN has been reduced, SPM effects would still be relevant with high-frequency RIN, which is the same as before. Further, linewidth broadening could arise from broadband spontaneous Raman scattering. In the future work section of the thesis, using the insights developed from these studies, we propose a new configuration which simultaneously uses the narrow linewidth feedback together with low intensity noise pumping to develop high-performance narrow linewidth CRFLs.