Treasures @ UT Dallas
Not a member yet
7697 research outputs found
Sort by
Properties of Blade-coated Transparent Conducting Oxide Sol-gel Precursor Films on Plastic Substrates
No full textThe transparent conducting electrode (TCE) industry is dominated by indium-tin oxide (ITO)
films deposited on glass for rigid applications and flexible polymers such as polyethylene
terephthalate (PET) for flexible applications. These applications can range from liquid crystal
displays to light emitting diodes to perovskite solar cells, most of which require electrodes with
the maximum possible optical transmittance and the minimum possible sheet resistance. ITO
films are generally deposited on their substrates via chemical vapor deposition or sputtering, both
of which are low-throughput processes, on the order of 1 m/min web speeds or batch-processed
at slower rates.
This research presents a novel solution-deposition method for an alternative chemistry to TCE
films. Solution deposition is advantageous over chemical vapor deposition and sputtering
because it can be undertaken in an open air environment without much energy applied to coating.
The project focuses on doctor blade coating as an intermediate step between current batch-
processing methods such as spin coating and future roll-to-roll (R2R) compatible methods such
as slot-die coating, in order to understand the dynamics of meniscus coating.
The process entails synthesis, coating, and post-processing of indium-zinc oxide (IZO) sol-gel
precursor ink on PET substrates. The benefit of this chemistry is its flexibility compared to ITO,
its capacity for room-temperature solution deposition, and its decreased reliance on indium. The
compositional ratio of In:Zn is 7:3, whereas that of ITO is 9:1 In:Sn.
0.2 M IZO alone is not sufficient to support the desired electrical conductivity, however, so a
randomly oriented solution-deposited silver nanowire (AgNW) mesh layer is added to the IZO
layer on PET. This results in the architecture PET/AgNW/IZO that has shown promising results
in terms of transparency and conductivity.
It is shown in this research that the PET/AgNW/IZO architecture is capable of performing within
the range of figures of merit (FoMs) of commercially available PET/ITO and is processable at
approximately ten times the speed. In addition to a deep exploration of the mechanisms at play
during the blade coating process of each of these layers, this report includes a technoeconomic
analysis comparing the two architectures and the costs of each in order to establish that blade
coated PET/AgNW/IZO is less expensive to manufacture than PET/ITO is to purchase. This has
implications for the TCE industry as throughput demands grow higher and higher as our society
demands more high performance optoelectronics, and R2R solution processing is a suitable
alternative
Towards AI and Hardware Synergy
No full textThis research explores the symbiotic relationship between Artificial Intelligence (AI) and
hardware, with a specific emphasis on the intersection of AI and Hardware. In this dissertation,
we consider two research thrusts: (i) AI for Hardware and (ii) Hardware for AI.
In recent years, there has been a widespread adoption of custom hardware-based AI solutions
to solve a plethora of real-world problems. For instance, researchers have proposed the
incorporation of AI in numerous mission-critical applications, especially in high-assurance
environments. To this end, in the first research thrust, we focus on developing AI techniques
catered to hardware. Specifically, we propose novel low-latency and high-fidelity AI
workloads to ensure the reliability of automotive hardware. On the other hand, the second
research thrust is associated specialized hardware for AI. Despite the ubiquitous use of AI
solutions in various real-world applications, such as facial recognition and autonomous vehicles,
their deployment on hardware renders inefficiency, especially in resource-constrained
platforms. Therefore, the second research thrust aims to facilitate efficient deployment of
AI workloads on dedicated hardware platforms. To this end, we have formulated two main
problems in this research, which are explained in detail below.
The first aspect of this dissertation focuses on ensuring the Functional Safety (FuSa) of automotive
systems. With increasing prevalence of safety-critical applications in the automotive
domain, it is imperative to ensure the FuSa of circuits and components within the associated systems, which are predominantly Analog and Mixed-Signal (AMS) circuits. However,
existing AI-based AMS FuSa violation detection solutions are limited by predefined feature
inputs and lack a rationale for determining signals to be monitored for anomaly detection.
To address these challenges, we propose a novel unsupervised Machine Learning (ML)-based
framework for early anomaly detection in automotive AMS circuits. Our approach comprises
the injection of anomalies into automotive AMS circuits in order to generate a comprehensive
anomaly model, novel centroid selection and time-series methodologies for expedited
high-fidelity anomaly detection. The proposed anomaly detection framework furnishes up
to 100% detection accuracy and reduced the associated latency by 5× (compared to the
non-time series approach). Subsequently, we augment our existing solution via novel feature
and signal selection techniques, as well as an Explainable AI (XAI) framework for enhanced
user interpretability and transparency. We achieve up to 7.2% improvement in detection
accuracy and 2.3× reduction in detection latency over our prior approach. Following this,
we perform anomaly abstraction to study the impact of anomalies across multiple abstraction
levels in automotive systems, wherein we achieve high-fidelity anomaly detection (up to
100% detection accuracy) in both component-level and block-level implementations. Moreover,
since we aim to deploy our AI solution on-chip for in-field applications, it is imperative
to enable efficient resource utilization for real-world applicability. This necessitates real-time
and low-power AI workload deployment, which we describe in our second research thrust.
The second aspect of this dissertation pertains to Application-Specific Integrated Circuits
(ASICs), such as Deep Neural Network (DNN) hardware accelerators, wherein we optimize
the energy efficiency of DNN inference. The proliferation of DNNs in recent years has led to
their widespread application in addressing a myriad of real-world challenges. However, due to
the significant computational and power requirements, specialized hardware platforms, such
as DNN accelerators, have been developed. Despite these advancements, DNN inference
execution is associated with energy bottlenecks in these resource-constrained accelerators.
To address these issues, we first propose a novel low-power hardware-based memory compression
solution catered to commercial DNN accelerators. Our approach, which optimizes
the memory subsystem of deep learning systems, involves hardware-based post-quantization
weight trimming, followed by dictionary-based compression, and subsequent decompression
by a low-power hardware engine during inference in the accelerator. Our technique furnishes
up to 28571× reduction in memory footprint, while incurring negligible area and power
decompression overheads of around 0.02% and 0.002%, respectively. Following this, we propose
a novel sensor compression solution designed to optimize the energy efficiency of DNN
sensor subsystems. The proposed approach employs a two-step approach involving subsampling
followed by supersampling of sensor images through interpolation. Furthermore, we
develop a fault injection framework to assess the fault resilience of DNNs (with compressed
sensor inputs) to bit-flip faults in the DNN accelerator memory. Our solution furnishes up to
62.1% energy savings with marginal performance degradation of 0.83%. Furthermore, from
our results, we can infer that DNNs accelerators witness up to 21.56% loss in classification
accuracy for compressed sensor inputs, rendering them highly vulnerable to bit-flip faults
manifested in their memory blocks. Therefore, by optimizing both the memory and sensor
subsystems, we seek to enhance the overall efficiency and performance of of deep learning
systems, particularly in resource-constrained environments.
In conclusion, this dissertation proposes pioneering approaches to achieve synergy between AI
and hardware, with the objective of improving the performance, safety, and power efficiency
of systems situated at the confluence of AI and hardware. This research describes novel
approaches to address the challenges pertaining to automotive AMS functional safety and
low-power DNN implementation in resource-constrained IoT edge devices, offering valuable
contributions to the AI and hardware research communities, and can foster the development
of more robust and efficient systems
Advanced Log-of-power Extremum Seeking Control for Wind Power Maximization
No full textThe emergence of wind power generation as one of the most economic and viable forms
of renewable energy depends on the reduction of the Levelized Cost of Energy (LCOE).
Advanced control strategies designed to increase wind power capture and decrease the structural loads could play a vital role in reducing the LCOE. This dissertation research aims
to investigate a real-time model-free control strategy, the Logarithm of Power Extremum
Seeking Control (LP-ESC), to increase the energy capture by optimizing wind turbine control parameters such as generator torque gain, tip-speed ratio set point, blade pitch angle,
and nacelle yaw angle. LP-ESC can optimally tune these control parameters in off-design
conditions without detailed knowledge of the underlying dynamics or the environment. The
basic ESC algorithm suffers from slow and inconsistent convergence under changing wind
speeds in below-rated conditions. Although the inconsistent convergence of the basic ESC
algorithm can be handled by changing the performance index from the power signal to its
logarithm, i.e., by using the LP-ESC, the algorithm can still show a slow convergence in
some applications. In the current work, we propose a Log-of-Power Proportional-Integral
ESC (LP-PIESC) to accelerate the convergence of the LP-ESC algorithm.
The first part of the research involves design and simulation of a real-time weighted multiobjective Log-of-Power ESC (LP-ESC) scheme to maximize the wind turbine power capture
with load reduction. The next part is focused on the design and simulation study of the LPESC and the LP-PIESC (Log-of-Power PIESC) on an NREL 5MW reference wind turbine
model in NREL’s OpenFAST code. Torque gain and blade pitch angle are used as the tuning
parameters. Results indicate that the LP-PIESC is much faster in finding the unknown
optimum, leading to a more practical algorithm for real applications. To demonstrate the
performance of the LP-PIESC under blade degradation and contamination, the algorithm is
used to re-tune the optimal tip speed ratio (TSR) for a turbine using the recently proposed
NREL’s Reference Open-Source Controller (ROSCO). It is shown that the LP-PIESC can
determine the new optimal TSRs when there are changes in aerodynamic parameters due to
blade degradation and contamination.
This dissertation also describes the results from wind tunnel experiments performed to maximize wind plant total power output using wake steering via closed-loop yaw angle control.
The experimental wind plant consists of nine turbines arranged in two different layouts,
both are two-dimensional arrays and differ in the positioning of the individual turbines.
Both algorithms are implemented to maximize wind plant power: LP-ESC and LP-PIESC.
These algorithms command the yaw angles of the turbines in the upstream row. The results
demonstrate that the algorithms can find the optimal yaw angles that maximize total power
output. The LP-PIESC reached the optimal yaw angles much faster than the LP-ESC. The
sensitivity of the LP-PIESC to variations in free stream wind speed and initial yaw angles is
studied to demonstrate robustness to variations in wind speed and unknown yaw misalignment. Finally, the LP-PIESC algorithm is also tested experimentally on a complex terrain
consisting of eight scaled wind turbines. The complex terrain was modeled by scaling a real
wind farm to fit in the wind tunnel. The LP-PIESC converged almost 30 times faster than
the LP-ESC. Results show that the LP-PIESC could find the unknown optimal yaw angles
even in the presence of gusty winds
‘Send More Butter’: Finding Meaning in Civil War Food References
No full textFood in the American Civil War meant more than nutrition. It served as a means of
communication, status elevator, social lubricant, and bridge between home and front, and even
across battle lines. This work examines how food, cooking, and references to food can be
interpreted to tell us more about how the war operated on different levels.
Approached thematically, the study looks at express boxes, mess bonding, cooking, social
hierarchy of cooks, the blockade, and trade across lines. Central to the argument is that food
references in Civil War letters acted as a subtle communications tool that give insight into how
soldiers felt and responded to the historic events around them. Essentially, it seeks to decode the
language of food in Civil War letters.
In addition to the letter diagnostics, the study takes a food-centric look at Sherman’s actions in
Georgia and the Carolinas in 1864-65, with an eye toward how his seizure and destruction of the
resources can be interpreted, and why he felt so confident in his success. Another intervention
involves the express boxes and how they connected the home front and the war front. By
examining tax data, it becomes clear that many more boxes were sent to the front than previously
estimated, which changes how we should approach these gifts and civilian contributions to the
war effort. Food is also used as a lens into the blockade, women’s resistance, and the formation
of bonds between soldiers. Cooking is examined for its ability to change the social status of meal
preparers, both white and Black, free and enslaved. Cooking changed attitudes and lives during
the war, even as it is suspected to have ended others.
Food is more than calories and comfort, it is also a means of communication, identity,
commerce, and social tie. Through this perspective, the Civil War takes on fresh nuances
Aesthetics and Politics of Women’s Authorship: Examining American and British Literary Texts by Women (1855-1930)
No full textWhat did it mean for women writers in the late nineteenth-century to identify and assert
themselves as literary artists and to assert that status as an identity independent of their status as
wives and mothers? This Dissertation revisits and re-examines a selection of literary texts by
British and American women writers, published between 1855 and 1930, towards an exploration
of different stylistic, aesthetic and narrative elements through which they emerge as models of
early feminist artistic vision and authorship. The texts under study foreground their critique of
social norms and engage with the cultural ideologies around women’s authorship through
experimental narrative elements, and, in that sense, they also represent early instantiations of a
reconciliation between aesthetics and politics of women’s literary writings. The project
juxtaposes literary works whose shared political commitments emerge through their narrative
shaping. I begin by considering Harriet Jacobs’ Incidents in the Life of a Slave Girl (1861) and
Christina Rossetti’s “Goblin Market” (1862) as expressions of women’s narrative autonomy and
agency, and end with readings of Kate Chopin’s The Awakening (1899), and Virginia Woolf’s To
the Lighthouse (1927) proposing that these narratives make a case for artistic ambition of women
authors and present their aesthetic choices as capable of generating political alternatives. The
literary output spanned by the starting and finishing points is crucial to analyzing the opposition
between aesthetics and politics of literature that exerts a major influence on how we understand
the value of literary studies within the academia
Framing Contemporary Women’s Hysteria: the Archive, the Mediated Self, and the Recontextualization of Women’s Madness
No full textThis creative dissertation seeks to expose the rudimentary issues of male-posed photographic documents of “truth” about the female hysterical body. The hysterical photographs I refer to were taken in the 1870s at the Salpêtrière Hospital in Paris. These photographs depict women posed by male photographers, and the poses reveal sexuality within the hysterical pose. My work juxtaposes contemporary documentations of the female body against these historical documents as a means of clarifying what constitutes a hysterical female body and how women continue to be called hysterical.
It is evident to me that the images produced in the French sanitarium have been replicated in advertising and in the portrayal of sexualized females in movies and television, as well as in social media. If we look at paintings which pre-date the hysterical photographs we can also see similarity of the pose. The conclusions that I have drawn from comparing imagery produced across time is that there is ample evidence to suggest a formal continuity of posing. Throughout history and under the oppression of patriarchal social structures, the pose has been controlled by
men. Therefore, I argue, for centuries women have been socially engineered to accept the logic of this male perspective, which, I further argue, has influenced women’s beliefs about how to look, behave, and think about themselves and their places in the world. The images I produced for my exhibition and the ideas contained within my dissertation chapters explore—through visual analysis—whether the persistence of the aesthetic formalities of the past that I see in current photographic compositions constitute a continued oppressive cultural framework and a reinforcement of the status quo regarding women’s position in society
Ferroelectric HfxZr1-xO2 for Next Generation Non-volatile Memory Applications and Its Reliability
No full textTo keep up with the increasing memory demands, developing memories with higher densities,
speed and energy efficiency is necessary at different levels of the memory hierarchy.
Ferroelectric materials have been considered alternative memory components; however, the
conventional perovskite-based ferroelectric materials pose several challenges due to CMOS
integration, high thermal budget, and scaling to sub 70 nm thicknesses. In this regard, the
discovery of ferroelectricity in doped HfO2 thin films was revolutionary as HfO2 is already
employed in front end CMOS as a high-k dielectric material for scaled thicknesses (<10 nm) .
Additionally, doping HfO2 films with ZrO2, i.e., Hf0.5Zr0.5O2 (HZO) showed stable ferroelectric
phase crystallization at back-end of line compatible temperatures (<450 °C).
This dissertation addresses some critical issues on the stress-induced crystallization of the
ferroelectric phase in HZO films, the reliability properties of metal-ferroelectric-metal (MFM)
structures, scaling ferroelectric HZO films on silicon substrates, and their reliability. First, the
driving forces for the crystallization of pure ferroelectric phase in HZO thin films were addressed
and the role of the TiN top electrode in phase crystallization at low process temperatures (400
°C) is studied. Then, the reliability of 10 nm thick HZO films was studied, and the various
ferroelectric device reliability properties and mechanisms were evaluated for metal-ferroelectricmetal structures. Finally, the ferroelectric HZO films were integrated directly on silicon for FeFET applications and the effect of ferroelectric device reliability based on scaling HZO films on
silicon structures was studied
Empirical Analyses of Emerging Technologies in Digital Marketing
No full textThe emergence of new technologies often reshapes the digital marketing landscape, prompting this
dissertation to delve into their roles across multiple marketing contexts. Specifically, this
dissertation explores the implications of blockchain and machine learning technologies in online
marketplaces, targeting algorithms on social media platforms, and digital ad content design
strategies. In the first chapter, blockchain takes center stage as an emerging technology poised to
transition the digital landscape from Web 2 to Web 3. The study focuses on the role of provenance
in bidding behavior in a blockchain marketplace by analyzing how the reputation of previous
owners of digital items in the form of NFTs (Non-Fungible Tokens) affects item valuation. The
products studied are a hybrid between consumption goods and collectibles. So, provenance is
expected to manifest in use value and collectible value (defined as the celebrity value and liquidity
value) of past owners. The empirical analysis uses a trading dataset of an online card game on
Opensea, a blockchain marketplace for NFTs. The results show that in such markets where
information about items’ ownership history is available, there is a transference of value from past
owners to the current trade.
In the second chapter, attention shifts to the use of machine learning algorithms for targeted ad
delivery and its potential consequence of leading to algorithmic discrimination on social media
platforms. Specifically, the study addresses algorithmic bias in ad delivery on social media
platforms and highlights the role of ad design in mitigating this problem. To achieve this, we train
a machine learning-based counterfactual generation model on a comprehensive dataset related to
social issues and political ads on Facebook and Instagram. The primary objective of this model is
to provide personalized and practical recommendations for modifying ad designs, aiming for an
equitable ad delivery outcome for neutrally targeted ads. Following the study, we conducted
multiple field tests on Facebook and Instagram to demonstrate the validity and practicality of the
counterfactual ad design generated by the proposed model. In the third chapter, the primary focus
is empirically examining the effectiveness of digital political fundraising ads distributed through
social media platforms, especially on Facebook and Instagram. We delve into the significance of
employing negative (versus non-negative) ad tone and comparative (versus non-comparative)
messaging. Our findings reveal that the efficacy of negative political fundraising ads depends on
the comparative messaging integrated into their content. Moreover, since these ads are distributed
on social media, they are subject to algorithmic targeting, so our ad elasticity estimates may be
biased. We demonstrate the robustness of our results to this algorithmic user selection by
incorporating Machine Learning-based adjustments to account for Meta's ad delivery algorithm
Far Infrared Detection and Photonic Components in Complimentary Metal-oxide Silicon (CMOS) Technologies
No full textConsumer applications of Infrared (IR) and Far Infrared (FIR) imaging are emerging. These
systems require fabrication of detectors on a membrane, heterogeneous integration with read-out
IC, and specialized packaging, which all increase cost. To lower the cost of Far Infrared imaging
circuits, this dissertation investigated feasibility of realizing the circuits using CMOS.
Implementing an electronic circuit becomes increasingly challenging in the Far infrared region of
the spectrum since the state-of the-art active solid-state devices no longer have gain.
First, this work proposed a metal-n+
silicon junction that is fabricated in a 130-nm CMOS process
without process modifications. Its cut-off frequency (fT) is expected to be greater than 5 THz and
flicker noise corner frequency is less than 10 kHz at a bias current of 100 nA. This is the highest
fT for any junctions fabricated in silicon technologies to date. The parasitic effects of metal
interconnect on the fT of junctions and diodes are explored and techniques to overcome the
challenges are demonstrated by increasing the fT of Poly-Gate-Separated Schottky Barrier Diodes
(PGS SBD) from ~2 THz to 2.7 THz.
Second, challenges for characterization of on-wafer devices with an fT of over 1 THz are
investigated and a de-embedding method (Sub-array-SHORT) for improving the measurement
reliability is proposed. The proposed technique overcomes the limitations of vector network
analyses of structures with a large ratio of imaginary and real parts (|X|/R) of their impedance.
Measurements and comparisons are performed to verify the effectiveness of the proposed method.
The %-variations of measured fT and series resistance over the measurement frequency range (50
to 55 GHz) and over the samples are reduced by ~50% compared to the measurements using a
conventional OPEN structure. Proper scaling of the estimated resistance and capacitance with the
number of device cells indicates that the proposed de-embedding technique provides estimations
of resistance, capacitance, and fT with improved accuracy.
For the first time, 20-THz electronic detection using foundry CMOS technologies is demonstrated.
The effects of thermal and electronic detections are separated noting that the output from thermal
and electronic detections can have opposite signs. A 20-THz PGS SBD detector using a 110-nm
CMOS process without process modifications is integrated with an on-chip Lock-In Amplifier.
Detectors with larger aperture sizes for FIR detection are implemented. It occupies an area of 82
µm2
. The peak optical responsivity of 29.8 V/W and shot noise limited NEP of 2.2 nW/√Hz are
achieved. Additionally, a 20-THz detector using the proposed unsilicided metal-n
+
junction is
fabricated using a standard 130-nm CMOS technology once again without any process
modifications. The pixel area is 8.6 µm2
. This is significantly smaller than other detectors reported
in the literature which makes it a good potential candidate for large detector arrays. The
measurements show that its output is higher at low bias conditions at which noise is smaller
compared to the PGS SBD detectors. The performance is better than other previously reported state-of-the-art and commercially available FIR detectors. Finally, the frequency responses of the
antennas are verified by measurements using an FTIR.
To characterize the detectors, a computer controlled low-cost wideband imaging setup is
constructed. A blackbody light source is utilized along with a wideband pyroelectric thermal
detector that is combined with an automatic 3-axis stage
High-throughput Optimizing Halide Perovskite Solar Cells Processed by Photo-irradiation
No full textPerovskite solar cells (PSCs) have leapfrogged many photovoltaics in the past decade due to
their excellent optoelectronic properties and solution processability. With over 25% power
conversion efficiency (PCE), PSCs have reached a laboratory performance suitable for
commercial applications. Successful commercializing PSCs will, however, be contingent upon
the development of a low-cost film processing method that is suitable for large-scale production
and rigorous study of the materials’ properties processed by these unconventional processes.
Despite much research has focused on scalable deposition technologies of halide perovskite, the
post-deposition process, a critical step of converting the precursor into the photo-active phase
and determining the film quality is often overlooked. This is because in laboratories, the postdeposition process is easily done on a hotplate, which takes 10 min at 100 °C. However, in a
perovskite solar panel manufacturing line, conventional thermal annealing would lead to two
problems: 1. The lengthy annealing time limits the throughput. 2. As the solar panel area
increases, the oven size will become impractically large, and the energy loss would increase
tremendously due to the equilibrium heating of the devices as well as the surrounding
environment.
Herein, we address these challenges by using photo-irradiance, rather than heat conduction or
convection for post-deposition process. By using flash lamp annealing, often called intense
pulsed light sintering or photonic curing, the conversion can be reduced from 10 min to 20 ms
because of the higher annealing temperature based on the Arrhenius Law. More than that, the
intense pulsed light is a selective heating process, only the perovskite precursor films absorb
most of the radiant energy, reducing the energy loss. The goal is threefold. First step is to
develop a high-throughput photo-irradiance method that delivers decent PSCs, which will be
covered in Chapter 2. Second step is to study the crystallization mechanism during the photoirradiance and to make the PCE comparable to PSCs made by the conventional thermal
annealing, which will be mainly discussed in Chapter 3 and 4. Third is to establish a machine
learning optimization process for a high-dimensional process such as a photo-irradiance process
to quickly transfer the knowledge from rigid PSCs fabrication to help make flexible PSCs, which
will be focused in Chapter 5.
After all, we have demonstrated that photonic curing is a viable tool for high-throughput
fabrication of PSCs, which is a crucial advancement toward PSC commercialization.
Additionally, with the assistance of machine learning, optimizing photonic curing on perovskite
as well as many other materials is becoming much easier, streamlining the development of this
technology on many fields