Now showing items 1-20 of 243

    • Development and Fabrication of Low Voltage (600 V) to High Voltage (15 kV) 4H-Silicon Carbide (SiC) Power Devices

      Yun, Nick (Nung Jun) (2022-12)
      The research primarily focuses on the development and fabrication of 4H-Silicon Carbide (SiC) power devices. As of today, power devices play a substantial role in high power applications such as fast-charging stations for electric vehicles, inverters for solar power, and energy storage equipment, to name a few. To minimize power loss during the operation, one of the key elements is to develop an energy-efficient power device. Although silicon (Si)-based power devices are currently being used in various high power applications, Si reached its physical limit in power loss reduction. In this aspect, wide-bandgap material, especially 4H-Silicon Carbide (SiC), became an excellent candidate to replace Si to fabricate power semiconductor devices that enable further minimization of power dissipation beyond Si. To advance the present and future low voltage (600 V) and high voltage (15 kV) power applications, the development of both low voltage and high voltage power devices are imperative. The most unique feature of a power device is the ability to withstand high voltages (> 600 V) with a voltage supporting layer, called the “drift region”. The breakdown voltage of the power device depends on the thickness and doping concentration of the drift region, as most of the voltage is supported by the depletion region formed within the drift layer. The optimization of the drift region must be performed to meet the breakdown voltage requirements based on the application while minimizing the on-state voltage drop to reduce power dissipation. When compared to the Si counterparts, SiC allows for the design of a thin, heavily doped drift region to support a specified voltage due to its superior material properties. Additionally, leakage currents generated during the off-state mode are also significantly suppressed due to two orders of magnitude lower intrinsic carrier density than that of Si. These merits of SiC become more substantial when building high voltage power devices (>3.3 kV) where resistance in the drift region dominates the overall on-resistance of the device. The details of optimizing device structures, fabrication details, and electrical characterizations of 600 V to 15 kV 4H-SiC power devices are discussed in this dissertation. The fundamental of the power device including the design of the drift layer and edge termination techniques for the power device will be discussed. To improve the low voltage application (i.e. electrical vehicles and photovoltaic converters), 600 V-rated lateral and vertical MOSFETs were developed and fabricated. From this work, the world's first high current (10 A) and high voltage (600 V) SiC lateral MOSFET was demonstrated. The fabricated lateral MOSFET was compared with the state-of-the-art vertical power MOSFET to identify the performance gaps to further enhance the electrical performances of the lateral MOSFETs. 600 V vertical MOSFETs and JBSFET (Junction-Barrier-Schottky (JBS) diode integrated MOSFET) were also developed to reduce the power loss in the system by replacing the Si-IGBTs (insulated-gate-bipolar-transistor) in the circuitry. The utilization of unipolar devices (i.e. MOSFET) is often more favorable than the bipolar devices (i.e. IGBT) due to faster switching speed and lower switching loss. On the other hand, the development of high voltage (> 6.5kV) devices are essential for high power applications such as power grids, military vehicles, to name a few. The fabrication and application of single-chip, high voltage devices are advantageous in terms of replacing many series-connected devices used to withstand high voltage in power circuits. However, research on ≥ 6.5kV-rated 4H-SiC power devices are very limited. With this motivation, 6.5 kV to 15 kV SiC JBS diodes, MOSFETs, and JBSFETs were designed and fabricated. From this study, we identified that device optimization for high voltage (> 6.5 kV) devices are different from the low voltage (< 1700V) devices due low background doping concentration of high voltage devices. Critical design considerations for fabricating 6.5 kV to 15 kV devices will be discussed. Both static and dynamic characteristics were also evaluated and compared, respectively.
    • Forecast of COVID-19 New Cases Using an MLP Regressor

      Alarcon, Hector (2022-12)
      Over the past year the COVID-19 pandemic has overwhelmed healthcare systems and government institutions worldwide, creating the need for accurate prediction of confirmed cases. Current research focuses on prediction methodologies aiming at mitigating economical anxiety and aiding in detection, preemption and forecasting of the pandemic. Modeling using artificial neural networks (ANN) is an important component of this effort and this research aims to contribute with a methodology to forecast occurrence of new cases of COVID-19 1, 2 and 7 days ahead by using historical case data. These approaches were compared and contrasted with some published results in terms of network architecture and neural network of use. Three regressors were developed and showcased in this document with their accuracy evaluated using the root mean squared error (RMSE) from 1179 to 2806. The quality evaluation supports the conclusion that these regressors compete with those of current architectures.
    • Studying the Segregation Induced Resist Component Contribution to EUV Stochastic Failures

      Azhari, Nabihah (2022-12)
      Extreme Ultraviolet (EUV) technology is necessary for chip manufacturing technology for finer circuits into many components for building faster and more energy-efficient chips. The EUV process utilizes a plasma light source that emits 13.5 nanometers in wavelength to create higher resolution chip circuit designs to transfer an aerial image at smaller dimensions for advanced process nodes at lower exposure doses. Chemically amplified resists have been recently commercialized as it benefits higher resolution chip circuit designs. However, one of the issues concerning EUV technology is that it can suffer from different types of stochastic defects due to photon shot noise, random inhomogeneities, and non-random inhomogeneities. This study investigates the potential non-random stochastic effects that exist in the multicomponent resist. In the multicomponent resist, self-segregation occurs, creating an inhomogeneous distribution leading to failures in the resist. We approached this problem by looking at previous models of the phase diagram to understand the system and energetic favorability of segregation. Throughout our experiments we explore balancing the ratios of solvent, polymer, and PAG and hoping to define the line where we reach the 2-phase region indicating we have reached segregation. First, we observed phase segregated regions using AFM through a spin coating method and a drop coat method. Then we approached the issue by analyzing the bulk liquid. Although we were unable to find the exact parameters where we cross the 2-phase region, through several formulations we have narrowed down the region that segregation occurs.
    • The Photovoltaic Properties of Carbon Nanotube Network p-n Diodes

      Oyibo, Gideon (2022-11)
      Single Walled Carbon nanotubes (SWCNTs) are quasi one dimensional rolled up sheets of graphene with amazing optical and electronic properties. Depending on their diameter and roll up angle, SWCNTs come in varying chiralities with multiple bandgaps giving them exceptional properties that make them attractive for photovoltaic applications. One of such properties is the absorption of light across the broad solar spectrum, a highly desirable property in semiconducting solar cell absorbers. In this dissertation, we will be exploring our attempt to fabricate a fundamental device that enables us harness the full sunlight potential of semiconducting SWCNT (s-SWCNT) networks and have a better understanding of its photovoltaic properties. To fabricate this fundamental device, we look to nature for inspiration on solar energy conversion. We use the process of photosynthesis as a model for building our solar energy conversion device. Nature, through centuries of evolution, has perfected the harvesting of light for energy conversion through the process of photosynthesis by employing two main mechanisms carried out by distinct proteins: excitation energy transfer, where light harvesting complexes capture light from multiple regions of the solar spectrum and funnel photoexcitations to a reaction center, and charge separation, where the photoexcitations become free charges in the reaction center. As we will see in this dissertation, SWCNTs have similar properties to that of photosynthetic systems, one of which is the varying chiralities of SWCNTs with different diameters, analogous to the distinct proteins in photosynthetic systems absorbing light at different wavelengths. We fabricate p-n diodes on various networks of s-SWCNTs, we study the intrinsic electronic and optical properties of nearly monochiral and polychiral s-SWCNT networks and form a fundamental understanding of the best s-SWCNT films required to make more ideal diodes. We examine the current-voltage characteristics of these diodes in the dark and find correlations between the key figure of merits, including the diode leakage current and the ideality factor, to different s-SWCNT networks. We also examine their optical properties by measuring wavelength-dependent photocurrent spectroscopy to gain insights into the dynamics of excitons in a network of s-SWCNTs. We achieve ideal diodes, for the first time in a homogenous network of s-SWCNTs. We discuss the limitations of using ideal diodes in the measurement of the electronic bandgap of s-SWCNT networks and then use non – ideal diodes to measure the electronic bandgaps of the s-SWCNT networks for the first time. After a more in-depth understanding of the dark diode characteristics of the s-SWCNT networks, we progress to fabricating a fundamental solar energy conversion device, modelled after photosynthesis. We fabricate photovoltaic diodes mimicking photosynthetic systems. Using different s-SWCNT chiralities, we create an energy funnel in our diodes by layering different s-SWCNT networks according to their bandgaps. The photo excitations in the larger bandgap s-SWCNTs are funneled down to the smallest bandgap s-SWCNT, allowing us to increase the spectral response of our diodes. We show that the photocurrent generation in our energy funnel is more efficient than in diodes formed using single chirality s-SWCNT networks. Finally, we show that our device architecture increases the photocurrent without increasing the highly undesirable dark leakage current. Using the analogy to photosynthetic systems, we use the smallest bandgap s-SWCNT network to create the diode (Reaction Center). The larger bandgap s-SWCNT networks act as light harvesters. We demonstrate an increase in short circuit current and the open circuit voltage as we add these nanotubes sequentially. We use this device to implement the mechanisms of exciton energy transfer in our p-n diodes and study its properties as it applies to s-SWCNT networks. We see some new and exciting physics which we will cover in this dissertation.
    • Regulation of PTP1B Activity and Insulin Resistance by Cellular Cholesterol

      Sagabala, Reddy Sudheer (2022-08)
      Protein Tyrosine phosphatase 1B (PTP1B), is an endoplasmic resident protein and a well-known negative regulator of the insulin receptor, dephosphorylating Tyr-1162, and Tyr-1163, two residues located in the activation loop of the insulin receptor. Mice lacking the PTPN1 gene encoding for PTP1B exhibit increased insulin sensitivity and improved glucose tolerance. Apart from its role in insulin signaling, mice lacking PTP1B show resistance to weight gain on a highfat diet, increased basal metabolic rate, and decreased cholesterol levels. In addition, PTP1B was previously identified in a proteome-wide mapping of cholesterol-interacting proteins in mammalian cells. However, the relationship between PTP1B and cholesterol is still unclear. To better understand the role of cholesterol on PTP1B function and on insulin signaling, we first used an in silico approach to predict cholesterol-binding sites in the 3D structure of the phosphatase and confirmed the binding sites through fluorescence binding studies and mass fingerprinting. We confirmed that the association between PTP1B and cholesterol occurred in both in vitro and in mammalian cells. In an attempt to understand whether cholesterol affects the ability of PTP1B to dephosphorylate substrates, we performed activity assays in various conditions. We observed that cholesterol could reduce and reactivate the reversibly oxidized form of PTP1B in vitro. Treatment of mammalian cells with cholesterol confirmed that excess cholesterol kept PTP1B reduced, and decreased Insulin Receptor phosphorylation and downstream signaling. In vivo results obtained by exposing mice to a high cholesterol diet support a role in the cholesterol-mediated reduction of PTP1B and decreased insulin sensitivity in the liver. We have established an electron tunneling path between the allosteric site and the catalytic cysteine residue and used a redox-sensitive fluorophore to measure electron tunneling in vitro. Hence, our results demonstrate for the first time that cholesterol binds to PTP1B at an allosteric site and reduces the phosphatase to regulate its activity and insulin signaling. Based on these results we propose a novel role for cholesterol in activating enzymes and in the context of insulin resistance.
    • Resist and Process Pattern Variations in Advanced Node Semiconductor Device Fabrication

      Chih-Fang Liu, Eric (2022-06)
      Pattern variations can cause challenges in device scaling. Since the last few decades, the semiconductor industry has successfully utilized the device scaling technique by reducing the transistor area to meet the requirements needed for optimum device performance and fabrication cost during each generation of development. The main challenges in the development of this technique are imaging resolution and pattern variations. Extreme ultraviolet (EUV) lithography and the multiple-patterning method can be used to push the imaging resolution to sub-30 nm. This thesis investigates the mechanism of pattern variations and proposes methods for pattern improvement. The thesis begins by investigating the origin of pattern variations in an EUV–chemically amplified photoresist system. The experimental results show that the chemical composition and inhomogeneity of the material contribute to pattern variations in EUV lithography. A difference in the localized-material-removal rate indicates the contribution of stochastics chemical kinetics in the photoresist during the development process. The study then investigates the effects of the plasma etching process on the pattern variations. The plasma etching process can alter the pattern variations by modifying the etching behavior and the etching selectivity. The thesis also discusses the system-level or integrated process-induced pattern variations. The method proposed herein involves surface modification and tone inversion technique and reduces the line edge roughness by 26% on a 20-nm pitch line pattern. Using a multicolor line-cut process, the thesis experimentally demonstrated the control of the edge-placement error from system-level pattern variations.
    • Microfluidic Imaging Windows for Study of the Tumor Microenvironment

      Head, Tristen (2022-08)
      Despite decades of research and billions of dollars in funding, cancer has maintained its epidemiological prominence as the second leading cause of death in the US for nearly 90 years. Currently, the clinical trial success rates for oncologic drugs is ~3%, and approved drugs often have a modest impact on overall survival. This is due in part to the tumor microenvironment (TME) which promotes cancer development and mitigates therapeutic response. Study of this biological system, however, is limited by conventional in vitro and in vivo techniques, which compromise either physiological relevance or experimental control. To better understand the role of the TME, we have utilized microfabrication techniques to develop the microfluidic imaging window (MFIW), an implantable platform for the observation and manipulation of in vivo TMEs. This technology provides unique opportunities for assessing the pharmacologic effects of therapeutics within intact, living tissue. Among the applications explored, a novel photolithographic technique, termed post exposure lamination, was developed to integrate tapered SU-8 micro-nozzle structures and enhance fluid conduction into porous matrices. Using these features, it was found that micro-nozzles improved axial penetration of fluorescent dextran into agarose tissue mimics and reduced the radial dispersion of Trypan Blue dye. Applications of localized reagent delivery for enhanced assay control were also investigated using small molecule nuclear stains and cell-based reporter systems. Here, significant cell staining occurred rapidly using small volumes of reagent (100 nL), substrate delivery for enzymatic processing was detected using a bioluminescent readout, and induction of cell gene expression was used to upregulate the production of fluorescent protein. Collectively, these capabilities showcase applications of the MFIW for enhanced monitoring and modulation of the TME that are well suited for translation into in vivo animal studies.
    • Therapeutic Targeting of Oncogenic Gain-of-Function Mutant p53 by Proteasome Inhibition

      Oduah, Eziafa I. (2022-07)
      Non-small cell lung cancer (NSCLC) is a molecularly complex and heterogenous disease. Recent advances in genomic profiling have changed the therapeutic landscape of NSCLC to incorporate targeted and immunotherapeutic approaches. Despite these advances, lung cancer remains the leading cause of cancer mortality in the United States and worldwide. This is partly because these novel treatments are not applicable to all patients and are often associated with primary or secondary resistance. This highlights the need for continued search for new therapeutic agents and strategies for NSCLC patients. However, the drug discovery and development pipeline is protracted and inherently expensive for new drugs. The projected timeline from identification of a new drug candidate from preclinical research to clinical trials and approval is estimated at about 12-15 years with an average cost of $1.3 billion [1]. Moreover, the failure rate for new drugs during the clinical development stage is high, reaching up to 96% by some estimates [2] and is partly due to adverse risk profiles of candidate molecules. Given the ongoing need for continued drug development in lung cancer, repurposing previously approved drugs for new indications when possible is advantageous. Such strategies decrease the cost and timeframe of drug development and pose a lower safety risk to patients since the toxicity profiles of the repurposed drugs are already well established. Drug repurposing has had success in cancer therapy. Some of these include the repositioning of thalidomide for use in multiple myeloma and the repurposing of rituximab from lymphoma to incorporate its use in rheumatoid arthritis [3]. Interestingly, the observations that led to many drug repurposing efforts were serendipitous by nature. However, recently more systematic approaches to repurposing drugs are being employed and include retrospective clinical analysis, genetic associations and pathway matching, binding assays to identify relevant target interactions, and large-scale in vitro drug screens with paired genomic data [3]. In this thesis compilation, I first and foremost lay the groundwork for repurposing proteasome inhibitors for therapeutic targeting of gain-of-function (GOF) oncogenic mutant p53 using lung cancer as a model disease. This has a potential for generalizability across cancers that bear GOF p53 mutations since alterations in TP53 are central to carcinogenesis and prevalent across tumor types. As the ‘guardian of the genome’, p53 maintains the genome integrity by inducing DNA damage repair or forcing aberrant cells into apoptosis or senescence. Failure of this function results in propagation of abnormal cells and the progression from normal to precancerous and malignant cells. Moreover, gain-of-function (GOF) activities of mutated TP53 related the acquisition of novel oncogenic properties are well described in the literature and are related to excess accumulation of the mutant protein. This work describes the mechanism of paradoxical destabilization of GOF p53 by proteasome inhibition in lung cancer and identifies ‘hyperactive’ proteasome genes in mutant p53 as targetable vulnerabilities in this subset of NSCLC. Since proteasome inhibitors are FDA approved drugs and prior drug candidates targeting p53 have not had success in clinical development, the final goal is to repurpose proteasome inhibitors to target GOF p53 mutant NSCLC.
    • Stock Price Prediction Using Sentiment Analysis and LSTM

      Carter, Caymen (2022-05)
      This work presents multiple Long Short-Term Memory neural networks used in con- junction with sentiment analysis to predict stock prices over time. Multiple datasets and input features are used on a LSTM model to decipher which features produce the best output predictions and if there is correlation to the sentiment of posts and the rising of a stock. This project uses embedding based sentiment analysis on a dataset collected from Kaggle which includes over one million posts made on the subreddit r/wallstreetbets. This subreddit recently came under fire by the media with the shorting of Gamestop in the stock market. It was theorized that this subreddit was working as a collective to drive up the price of multiple stock, therefore hurting large corporations such as hedge funds that had large short positions on multiple stocks.
    • Application of Resistive Random Access Memory (RRAM) For Non-Von Neumann Computing

      Rafiq, Sarah (2022-05)
      The movement of data between physically separated memory and processing units in conventional computing systems (the so-called von Neumann architecture) incurs significant costs in energy and latency. This is known as the von Neumann bottleneck. With the advent of the Internet of Things (IoT) and edge computing, computing systems are also becoming significantly power limited. In this work, hafnium oxide resistive random access memory (ReRAM) integrated with 65nm CMOS technology on a 300 mm wafer platform was assessed to carry out two novel non-von Neumann computing applications that processes data within memory and avoid excessive data movement. These computing applications are based on regulating the flow of sneak path currents in memory arrays to perform computation, called flow-based computing, and detecting degree of association (correlation) between binary processes in an unsupervised manner using the ReRAM non-volatile accumulative behavior, termed as temporal correlation detection. Electrical characterization of hafnium oxide ReRAM arrays was conducted for multi-level resistance states for flow-based computing, which was then investigated for two functions, approximate edge detection and XOR Boolean logic, through both experiments and simulation. The effect of device non-idealities was also evaluated. A trade-off between the flow-based output resistance ratio and the variability of flow-based outputs was found for different patterned binary resistance Roff/Ron ratios. For the second non-von Neumann application, the feasibility of ReRAM as a non-volatile candidate device was investigated with an empirical ReRAM model through simulation. Experimental ReRAM analog incremental switching data, from both SET and RESET regimes, was also evaluated on the modified temporal correlation detection algorithm, where the RESET regime resulted in better performance. The ReRAM based implementation yielded 36,000-53,000 vi times lower energy consumption than similar implementation with phase change memory for 25 binary processes, and a speed-up of computation time by 1,600-2,100 times than that of a CPU-based implementation using 1xPOWER8 CPU. 1xPOWER8 CPU is a CPU available on the IBM* Power* System S822LC system, the POWER8 system series, where the CPU was run for 1 thread. In summary, hafnium oxide ReRAM based on 65nm CMOS technology has been evaluated for two non-von Neumann computing applications, and the effect of device non-idealities has also been assessed. These ReRAM in-memory computing applications show the promising potential of ReRAM in overcoming the von-Neumann bottleneck.
    • Generating Facial Character: A Systematic Method Accumulating Expressive Histories

      Jofre, Ana (MIT Press, 2022-04)
      The author presents a method to simulate facial character development by accumulating an expressive history onto a face. The model analytically combines facial features from Paul Ekman’s seven universal facial expressions using a simple Markov chain algorithm. The output is a series of 3D digital faces created in Blender with Python. The results show that systematically imprinting features from emotional expressions onto a neutral face transforms it into one with distinct character. This method could be applied to creative works that depend on character creation, ranging from figurative sculpture to game design, and allows the creator to incorporate chance into the creative process. The author demonstrates the method’s application to sculpture with ceramic casts of generated faces.
    • Crowdsourcing Image Extraction and Annotation: Software Development and Case Study

      Jofre, Ana; Berardi, Vincent; Brennan, Kathleen P.J.; Cornejo, Aisha; Bennett, Carl; Harlan, John (Digital Humanities Quarterly, 2020-03)
      We describe the development of web-based software that facilitates large-scale, crowdsourced image extraction and annotation within image-heavy corpora that are of interest to the digital humanities. An application of this software is then detailed and evaluated through a case study where it was deployed within Amazon Mechanical Turk to extract and annotate faces from the archives of Time magazine. Annotation labels included categories such as age, gender, and race that were subsequently used to train machine learning models. The systemization of our crowdsourced data collection and worker quality verification procedures are detailed within this case study. We outline a data verification methodology that used validation images and required only two annotations per image to produce high-fidelity data that has comparable results to methods using five annotations per image. Finally, we provide instructions for customizing our software to meet the needs for other studies, with the goal of offering this resource to researchers undertaking the analysis of objects within other image-heavy archives.
    • Faces extracted from Time Magazine 1923-2014

      Jofre, Ana; Berardi, Vincent; Bennett, Carl; Reale, Michael; Cole, Josh (Journal of Cultural Analytics, 2020-03-16)
      We present metadata of labeled faces extracted from a Time magazine archive that contains 3,389 issues ranging from 1923 to 2012. The data we are publishing consists of three subsets: Dataset 1) the gender labels and image characteristics for each of the 327,322 faces that were automatically-extracted from the entire Time archive, Dataset 2) a subset of 8,789 faces from a sample of 100 issues that were labeled by Amazon Mechanical Turk (AMT) workers according to ten dimensions (including gender) and used as training data to produce Dataset 1, and Dataset 3) the raw data collected from the AMT workers before being processed to produce Dataset 2.
    • What’s in a Face? Gender Representation of Faces in Time, 1940s-1990s

      Jofre, Ana; Cole, Josh; Berardi, Vincent; Bennett, Carl; Reale, Michael (Journal of Cultural Analytics, 2020-03-16)
      We extracted 327,322 faces from an archive of Time magazine containing 3,389 issues dating from 1923 to 2014, classified the gender of each extracted face, and discovered that the proportion of female faces contained within this archive varied in interesting ways over time. The proportion of female faces first peaked in the mid-to-late 1940s. This was followed by a dip lasting from the mid-1950s to the early 1960s. The 1970s saw another peak followed by a dip over the course of the 1980s. Finally, we see a slow and steady rise in the proportion of female faces from the early 1990s onwards. In this paper, we seek to make sense of these variations through an interdisciplinary framework drawing on psychology, visual studies (in particular, photography theory), and history. Through a close reading of our Time archive from the 1940s through the 1990s, we conclude that the visual representation of women in Time magazine correlates with attitudes toward women in both the historical context of the era and the textual content of the magazine.
    • Face Recognition and Emotion Identification

      Andriamanalimanana, Bruno R.; Thesis Advisor; Spetka, Scott; Thesis Committee; Chiang, Chen-Fu; Thesis Committee; Lagwankar, Akshara Avadhut (2021-05)
      While face recognition has been around in one form or another since the 1960s, recent technological developments have led to a wide proliferation of this technology. This technology is no longer seen as something out of science fiction movies like Minority Report. With the release of the iPhone X, millions of people now literally have face recognition technology in the palms of their hands, protecting their data and personal information. While mobile phone access control might be the most recognizable way face recognition is being used, it is being employed for a wide range of use cases including preventing crime, protecting events and making air travel more convenient. This project focuses on various advanced Python libraries to improve the face recognition accuracy such as OpenCV, Sklearn, face_recognition. The project understands the data and model, train it for further usage. The real time videos are considered for evaluating the results. Further the project glances the emotion recognition algorithms using CV2, Seaborn. The areas of the human faces are highlighted according to different emotions. The large data sets (fer2013, Olivetti faces) are used for training and testing the data sets. PCA, leave one out cross validation, grid search CV, machine learning pipelines, CNN models are used to estimate and increase the accuracy. The project is executed in Anaconda environment Jupyter Notebook. As the data sets are huge Google Collaboratory is used for execution.
    • sxRNA Switches: Hypothesis Through Automated Design Via a Genetic Algorithm Approach

      Tenenbaum, Scott; Chair; Melendez, J.A.; Cady, Nathaniel; Fasullo, Michael; Begley, Thomas; Doyle, Francis J., II (2021-12)
      The following document is meant to represent an overview of my work on structurally interacting RNA (sxRNA), which has already resulted in three publications with another two in preparation. Where appropriate, some text and data from these publications have been reproduced here. Ribonucleic Acid (RNA) is one of the fundamental macromolecules present in living systems. It can be found in all cells as varying length polymer chains composed of four primary bases (adenine, cytosine, guanine, uracil) capable of numerous modifications. Though generally characterized as an information carrier, RNA is a versatile molecule that exhibits both intra and inter-strand base pairing to form complex structures. Similar to protein, the particular shape of an RNA structure in combination with some degree of sequence specificity, can dictate its function (RNA binding protein recognition sites, ribozyme activity, aptameric affinity, etc.). Structurally interacting RNA (sxRNA) is a molecular switch technology that exploits predictable intermolecular RNA base pairing to form an otherwise absent functional structure in one RNA strand when it interacts with a specific, targeted second strand. Originally proposed as a potential regulatory mechanism in natural systems, we used characteristics of predicted pairings in that context to engineer purely synthetic sxRNA switches that have been successfully tested. There are many non-coding RNAs associated with pathological conditions, the ability to use these as triggers for sxRNA opens the door to potential applications ranging from diagnostics to therapeutics. Furthermore, other prospective triggers (including those synthetically designed) may allow use of the technology as a molecular tool for a variety of purposes including as an alternative to antibiotic selection in cell line development. The typical trigger sequences targeted by sxRNA switches are at least 20 bases in length. Combinatorial options with regard to structure positioning and base composition produce an enormous number of potential sxRNA sequences for any given target. Exhaustively examining these for feasible candidates (i.e., analyzing predicted interactions with unintended targets) is computationally impossible with current systems. Evolutionary computing is a subfield of artificial intelligence (AI) that has been inspired by biology. Genetic algorithms are a type of evolutionary algorithm and apply operators (such as recombination and mutation) to find candidate solutions to an optimization problem. The presented dissertation will describe the original sxRNA research as well as the development and testing of a genetic algorithm that automates the production of new sxRNA switch candidates. This algorithm takes into consideration factors that were previously impossible to account for in manual designs.
    • Multicyclic Loss for Multidomain Image-to-Image Translation

      Reale, Michael, Chair; Confer, Amos; Andriamanalimanana, Bruno; Schneider, Ethan H.
      GANs developed to Translate an Image’s style between different domains often only care about the initial translation, and not the ability to further translate upon an image This can cause issues where, if one would want to generate upon an image and then further on, change that image even more that person may come into issues. This creates a ”gap” between the base images and the generated images, and in this paper a Multicyclic Loss is presented, where the Neural Network also trains on further translations to images that were already translated. iv
    • Development of High-Performance Hafnium Oxide based Non-Volatile Memory Devices on 300mm Wafer Platform for Data Storage and Neuromorphic Applications

      Diebold, Alaine (Committee member); Ventrice, Carl A. Jr. (Committee member); Lloyd, James (Committee member); Kurinec, Santosh (External committee member); Cady, Nathaniel (Dissertation Committee Chair); Hazra, Jubin (2021-08)
      Fundamental limitations associated with scaling and modern von Neumann computing architectures illustrates the need for emerging memory solutions in the semiconductor industry. One such promising non-volatile memory (NVM) solution is resistive random access memory (RRAM), which is seen as a potential candidate that can meet the performance needs of DRAM and the density of NAND Flash in terms of scalability, reliability and switching performance. However, reliable operation of RRAM devices requires further development to remedy device- to-device and cycle-to-cycle uniformity variation, increase the conductance window, and to improve retention, yield and endurance properties. This research work primarily focuses on improving RRAM performance metrics through optimization of processing conditions and programming algorithms for CMOS-integrated nanoscale HfO2 RRAM devices on a full scale 300mm wafer platform. It was observed that tuning of ALD parameters during RRAM switching layer HfO2 deposition had a significant impact on device switching performance. An excellent memory window of >30 with switching yield ~90%, along with low cycle-to-cycle (σ <0.5) and cell-to-cell variability (σ <0.4) were achieved for tested 1 Transistor 1 RRAM (1T1R) cells across full 300mm wafers. The devices demonstrated excellent endurance (>1010 switching cycles) and data retention performance at elevated temperature (105 s at 373K). The fabricated RRAM cells were also optimized for multi-level-cell switching behavior and ~10 distinct resistance levels were obtained through a combined current- and voltage-control based programming approach. An incremental pulse write technique combined with read verification algorithm enabled accurate resistance states programming within a large resistance window along with linear and symmetric potentiation-depression characteristics yielding superior analog synaptic functionality of fabricated RRAM devices. In addition to RRAM devices, hafnium zirconium oxide (HZO) based nanoscale ferroelectric tunnel junction (FTJ) devices were successfully implemented on a 300 mm wafer platform. Current measurement, as a function of voltage for both up and down polarization states, yielded a tunneling electroresistance (TER) ratio of ~5 and switching endurance up to 106 cycles in TiN/ Al2O3/ Hf0.5Zr0.5O2/ TiN FTJ devices distributed across full 300 mm wafer. Investigation of current transport mechanisms showed that the conduction in these FTJ devices is dominated by direct tunneling (DT) at low electric field and by Fowler-Nordheim (F-N) tunneling at high electric field. The realization of CMOS-compatible nanoscale RRAM and FTJ devices on 300mm wafers demonstrates the promising potential of these devices in large scale high-yield NVM manufacturing for high performance embedded memory and mass data storage applications.
    • Text Detection from an Image

      Andriamanalimanana, Bruno R.; Thesis Advisor; Novillo, Jorge; Thesis Committee; Spetka, Scott; Thesis Committee; Goda, Piyush Jain (2020-12)
      Recently, a variety of real-world applications have triggered a huge demand for techniques that can extract textual information from images and videos. Therefore, image text detection and recognition have become active research topics in computer vision. The current trend in object detection and localization is to learn predictions with high capacity deep neural networks trained on a very large amount of annotated data and using a high amount of processing power. In this project, I have built an approach for text detection using the object detection technique. Our approach is to deal with the text as objects. We use an object detection method, YOLO (You Only Look Once), to detect the text in the images. We frame object detection as a regression problem to spatially separated bounding boxes and associated class probabilities. YOLO, a single neural network, that predicts bounding boxes and class probabilities directly from full images in one evaluation. Since the whole detection pipeline is a single network, it can be optimized end-to-end directly on detection performance. The MobileNet pre-trained deep learning model architecture was used and modified in different ways to find the best performing model. The goal is to achieve high accuracy in text spotting. Experiments on standard datasets ICDAR 2015 demonstrate that the proposed algorithm significantly outperforms methods in terms of both accuracy and efficiency.

      Rogers, Jack (2021-05)
      Insulators play an important role in the architecture and resulting performance of semiconductor devices manufactured today. Materials such as HfO2 and Al2O3 are utilized as gate oxides and spacers to control leakage current and enable bottom-up self-aligned patterning of device features. Understanding the electrostatic barrier that forms at the metal-oxide-semiconductor (MOS) interface is crucial in the development of field effect transistors and other devices, especially as the scaling of device features continues to shrink into the nanoscale. Characterization of the barrier height using current-voltage (IV) and capacitance-voltage (CV) techniques provides only a spatially averaged view of the interface, and is incapable of accounting for local nonuniformity which arises at nanoscale dimensions. Additionally, common lithographic strategies for patterning small feature oxides are limited by printing misalignments such as edge placement error (EPE), and in order to achieve smaller pitch sizes lithography steps must be repeated multiple times which adds time and cost to the process. The feasibility of uniform, cost-effective insulator films at the 5 nm technology node and beyond relies on the development of new deposition strategies. In this thesis, hafnium oxide grown using atomic layer deposition (ALD) is examined with ballistic electron emission microscopy (BEEM). Localized nonuniformities in the barrier height are found to exist for two identically prepared samples which reveal three distinct electrostatic barriers at the buried Au/HfO2/SiO2/Si-p interface, including a novel barrier found at 0.45 eV due to ultrathin HfO2. The results uncover changes in electrostatic behavior of the film which are otherwise impossible to detect using spatially averaged techniques. These variations in barrier height are visualized in a novel way that produces spatial maps showing transitions between high energy and lower energy barriers across a few nanometers. The resolution of this mapping technique is determined by comparing the measured barrier heights of Au/Si(001) and Au/Si(111) interfaces. Momentum conservation and electron scattering result in slightly different barrier heights for both interfaces that depends on metal thickness. The Rayleigh criterion is applied to the barrier height distributions as a function of metal thickness, resulting in a 10 meV resolution. Both aluminum oxide and hafnium oxide are also selectively grown on patterned metal / low-k silicon wafers using ALD. Self-assembled monolayer (SAM) materials such as octodecanethiol (ODT) and dodecanethiol (DDT) -which are functionalized to metal -are first deposited on the copper lines in order to block high-k film deposi¬tion on metal. Both HfO2 and Al2O3 are shown to selectively cover the low-k lines for linespace pitches greater than 100 nm and 5 mM concentration of SAM, and better selectivity is achieved for smaller pitches using lower SAM concentrations. Selectivity is measured qualitatively and quantitatively using x-ray photoemission spectroscopy and confirmed with transmission electron microscopy.