Abstracts

Thin films with a stoichiometry of CeFeO_x  and CeMnO_x were conformally deposited on high-surface-area
γ-Al₂O₃  by Atomic Layer Deposition (ALD). 1,2 X-ray diffraction (XRD) patterns, High-Resolution Transmission
Electron Microscopy (HR-TEM) images demonstrated that 2 nm-thick films exhibited a perovskite structure after
reduction at 1073 K but converted to a fluorite phase upon oxidation at 1073 K. The transition between the fluorite
and perovskite structures was reversible for at least five oxidation and reduction cycles. This demonstrates the
superior stability of thin-film CeFeO₃ over bulk CeFeO₃ , which will irreversibly phase separate into CeO₂ and Fe₂O₃
upon 1073-K calcination. It is noted that the formation of perovskite CeMnO₃ film is intriguing since no bulk
CeMnO₃ has ever been successfully synthesized yet. In addition, both the CeFeO_x  and CeMnO_x thin films were
found to have completely different thermodynamic properties compared to either their bulk counterparts or the
combination of their constituent monoxides, suggesting that they may be useful in various catalytic applications.

Chemical manufacturing processes can present significant dangers, and therefore, it is vital to incorporate safety and reliability measures during their design. To reduce the probability of catastrophic accidents, which can have grave consequences on human life and the environment, comprehensive instrumentation such as control systems, alarms, and automated Safety Instrumented Systems (SIS) are regularly utilized in chemical processes. Common reliability assessment methods such as failure mode and effect analysis (FMEA), fault-tree analysis (FTA), reliability-centered maintenance (RCM), root-cause analysis (RCA), and the like, have proven to be effective in identifying and handling postulated abnormal events that have occurred previously or are more likely to occur, based on process historian data. However, it is difficult to predict and counteract the impact of infrequent and unforeseeable un-postulated abnormal events in real-time, which, when not considered during process design, can lead to the most serious consequences. Hence, existing reliability/safety systems, alone, might prove to be insufficient in monitoring and alerting the operator for un-postulated abnormal events. Additionally, most fault diagnosis and risk assessment techniques for rare events in literature do not adequately address the “decision science” aspect of risk analysis, i.e., given the quantified probabilities of the rare events, what are the strategies and response actions that can be implemented in real-time to minimize the probabilities?
Previously, we developed an advisory system for analyzing and monitoring process reliability, consisting of novel, multivariate alarm systems with appropriate response actions introduced using process modeling and path sampling for un-postulated abnormal events. Its purpose is to augment and support existing reliability systems, suggesting actions when unanticipated reliability/quality events are approached. Our analyses were demonstrated initially on an exothermic CSTR process and led to promising alarm thresholds and reliability response actions. To quantify the probabilities and develop the alarm systems, we proposed parametric regression models using ordinary least squares (OLS) – such simple parametric models with inherent assumptions may not hold true for different values of process parameters.
Herein, we extend our analyses to a more complex case study, i.e., a polystyrene CSTR exhibiting a free radical polymerization (FRP) mechanism. Polymerization reactors are known to exhibit complex, nonlinear behavior, e.g., output multiplicity, input multiplicity, and isolas — often, it is desirable to operate at the intermediate unstable region, with potential abnormal transitions to multiple undesirable operating regions. This is a control, safety, and reliability problem, and hence, a good application for our analyses. Additionally, we develop non-parametric tree based ensemble machine learning models to quantify the probabilities and develop the alarm systems – such models do not have any inherent underlying assumptions regarding the mathematical functional form, and are capable of “learning” the model parameters from the data simulated during path-sampling. Hence, such models are applicable to wide ranges of the process parameters and are used reliably in real-time to estimate the probabilities of abnormal shifts to the undesirable operating regions.

Hydrodeoxygenation (HDO) of m-cresol was studied over WO_x-Pt/SBA-15 using several
light alkanes as H₂ carriers. While dehydrogenation of n-hexane at 773 K over a Pt/SBA-15 with
9-nm pores was unstable at pressures below 30 bar, the reaction was stable for more than 5 h at
pressures greater than 40 bar. The transition from stable to unstable operation depended on the
support pore size and structure, occurring at lower pressures on Pt/SBA-15 with 6-nm pores and
much higher pressures on conventional Pt/SiO₂. WO_x-decorated Pt/SBA-15 was active for both
dehydrogenation of n-hexane and HDO of m-cresol, and both reactions could be carried out in a
stable manner at high pressures. HDO of m-cresol could also be carried out using H2 from n-
pentane and 3-methylpentane over WO_x-Pt/SBA-15 at high pressures. Possible explanations for
the high-pressure stability are given.

Unmanned aerial vehicles (UAVs) or drones have gained significant attention in recent
years due to their applications in various fields, including surveillance, aerial
photography, environmental monitoring, and precision agriculture. The need for drones
with improved capabilities has led to the development of smaller and lighter drone
platforms that can perform tasks with greater precision and efficiency. As a result, there
is a growing demand for micro-drones that can navigate through confined spaces and
operate in highly sensitive environments. A metric directly influencing micro-drone
operational efficiency is flight duration. Flight duration depends on the specific energy
density of on-board power sources. Lithium-ion batteries, LIBs, have been the preferred
power source for drones, due to both LIB availability and scalability ranging from several
grams to several kilograms. Yet, the hermetic packaging mandates of LIBs can
significantly reduce available energy density in micro-drones. Aluminum-air batteries
(AABs) show higher theoretical energy density than LIBs, and may be particularly
applicable for this particular application. Employing 3D-printing techniques, we
developed lightweight battery packaging solutions for AABs. Anode structures and
electrolyte compositions were studied and engineered for high-discharge-rate AABs.
Utilizing a micro-drone, we conducted flight duration and thrust characterization
analyses to evaluate the performance of the AAB. The results of this work suggest that
the AAB is a promising approach for extending the flight duration of micro-drones.

Each year, the US produces enough mining waste to fill 38 million Olympic size
swimming pools. This waste could present a scalable opportunity to store carbon and unlock
domestic critical mineral recovery. The imperative for carbon removal on a gigatonne scale, as
emphasized by the Intergovernmental Panel on Climate Change, accentuates the urgency to
address rising CO₂ emissions. However, technical/engineered approaches to carbon removal have
failed to gain traction, largely due to high economic barriers to deployment. Early techno-
economic analyses suggest that carbon mineralization has the greatest potential to become
economically viable through consideration of mineral feedstocks with valorizable co-products.
This research examines metal extraction via pH swing on magnesium and calcium rich
silicates composed of heterogeneous crystal structures and integrates bench scale
experimentation with rigorous techno-economic analysis (TEA) to evaluate the potential to scale
on metrics of embodied carbon, energy consumption, environmental harms, and co-benefits for
ecosystems and communities. This acidic leaching process can extract up to 80% of carbon-
storing alkalinity, as well as valuable metals and critical minerals. These product streams may be
valorized through carbonation with captured CO₂ to yield carbonate products for use in other
industries, the concrete industry, which is responsible for 8% of the world’s annual emissions or
separated and sold into other scalable markets like iron and steel manufacturing. The in-house
TEA shows costs of $383-$1,010/tCO₂ for a process designed to intake 1Mt of magnesium-rich
silicate mine tailings annually to store 0.13Mt of CO₂ and produce 0.32Mt of mixed metal
hydroxide precipitant. The carbon balance in the TEA also demonstrates the process emits 0.01-
0.04Mt CO_2eq through energy usage and embodied carbon, giving an overall carbon efficiency of
70-90%. At commercial scale, this process has the potential to link multiple decarbonized
supply chains which are critical for the energy transition. Further research will investigate the
level of compatibility between the carbonates and concrete aggregate along with the fundamental
kinetics for each process step to elucidate the major levers for optimizing low-cost carbon
storage.

Biofouling, the undesired formation of biofilms, can contaminate a wide variety of surfaces
that operate in aqueous environments, such as medical device implants and water filtration membranes.
One way to combat this issue is to increase a surface’s hydration to prevent proteins from adsorbing onto
the surface, which is the first step in biofilm formation. While homogeneous surface modifications, such
as polar and/or zwitterionic surface coatings employ electrostatics to enhance surface hydration, more
heterogeneous surface modifications derived from proteins themselves could also provide enhanced
surface hydration, since proteins surfaces have evolved to resist non-specific aggregation and fouling by
other proteins in the crowded cellular environment. By identifying and characterizing the most
hydrophilic regions, we aim to uncover the chemical patterns responsible for protein-protein selectivity
and provide a basis for the creation of non-fouling surfaces. Using specialized molecular dynamics
simulations, we have characterized the atomic-level hydrophilicity of a protein surface directly based on
water affinity. Using this characterization and the DBSCAN clustering algorithm, we have identified
protein surface patches of similar water affinity for various proteins, ranging from highly hydrophobic to
highly hydrophilic. Our results indicate that patches of similar water affinity contain both polar and
nonpolar atoms as well as amino acids of varying hydrophilicity. We have also found that proteins
surface patches of similar partial charge do not have similar water affinity, indicating the importance of
other factors, such as chemical patterning and surface topology on protein surface hydrophilicity.

Complement opsonization, which is one of the most primitive pathways, occurs when
complement proteins mark a pathogen for phagocytosis by immune cells. The binding of
complement proteins to a pathogens’ surface enhances immune cells’ ability to
recognize and engulf the pathogen.
In the context of nanoparticles, such as naturally occurring exosomes or otherwise, the
principles of complement opsonization can be leveraged to improve the targeted
delivery of nanoparticles. By engineering nanoparticles with surfaces that can evade
complement proteins, these particles can mimic pathogens and gain an advantage over
opsonization. This can help enhance their stability and circulation time in the
bloodstream.
We show that Complement deposition onto surfaces is intricately fine-tuned to the
nanoscale spacing of surface molecules, we uncover a critical threshold spacing
between complement sites above which the surface is completely coated by
complement proteins. Through a dynamical systems model as well as an agent-based
model and by performing systems size scaling, we established the complement coating
of surface proceeds through a percolation phenomenon.
Our results suggest that the complement phenomenon is activated differently in different
systems (virus, nanoparticles, exosomes) and the threshold is context-specific knowing
which could help predict the immunogenicity of natural and engineered nanoparticles in
a physiological microenvironment including the tumor microenvironment.

The transcription factor Myc is known to modulate a multitude of genes and cellular processes. Myc is
often deregulated in human cancers and is commonly known as an “undruggable” molecule. Factors such
as Myc’s structure and primary nuclear localization make it difficult to directly target Myc in cancer
treatments. In this work, to identify indirect strategies to target Myc, we study the molecular mechanisms
behind 1) regulation and 2) function of Myc, and uncover the role of Myc in mechanosensitivity. In the first
part of this work, we studied Myc regulation by extracellular growth and mechanical cues transduced
through intracellular signaling pathways. Here, we built an ODE-based kinetic model consisting of MAPK,
Rho/ROCK, and PI3K/Akt pathways effected by EGF (growth) and integrin (mechanical) receptors. Our
modeling results show that in normal cellular phenotypes, Myc is primarily regulated by EGF receptors.
However, in cancerous phenotype, Myc is regulated by both EGF and integrin receptors. We, therefore,
conclude that growth and mechanical signals play a synergistic role in driving and sustaining cancers by
regulating Myc. Hence, while designing treatment strategies for Myc-driven solid cancers, mechanical
cues from remodeled tumor microenvironment must be considered as an essential factor. Further, we
systematically perturbed signaling interactions in the model to identify key processes governing Myc in
cells. From this analysis, we identified intermediate nodes critical for Myc deregulation, which can be used
as indirect targets to treat Myc-driven solid cancers. In the second part of this work, we performed
bioinformatics analysis on Myc ChIP-seq datasets to study Myc-DNA binding patterns. Through motif
discovery analysis, we identified a new DNA binding sequence of Myc which we propose is a low-affinity
binding site for Myc. From additional bioinformatics analysis, we show that Myc functions cooperatively
with other proteins to occupy this low-affinity site and modulate distinct cellular processes related to
mechanotransduction and mechanosensitivity. The DNA-binding patterns and protein/protein interactions
identified from this work will aid in the development of cancer treatment strategies targeting Myc. In
conclusion, we gain a mechanistic understanding of Myc activity by studying Myc regulation and function
in cells. We then uncovered a novel role of Myc in mechanotransduction in the tumor microenvironment
that could couple to several cancer hallmarks such as proliferation, survival/dormancy, and
migration/metastasis. We also identify new interactions and mechanisms critical to Myc activity, which can
be used as indirect ways to target Myc in cancer therapies.

Physics informed neural networks (PINNs) have grown from a toy machine learning concept to very
powerful and applicable in a variety of modern problems. As PINNs continue to grow, they will inevitably
be utilized more in the fields of medicine and biology, where not only are the domains of interest
incredibly complex but information on them incomplete. We study PINN performance in the said area
via the lens of aortic dissections (AD) informed by MRI scanning. Specifically, we consider the prediction
accuracy of PINNs as a function of 4D flow MRI in both spatial and temporal resolutions, and consider
PINN prediction of the gradient-based parameter, wall shear stress. Three AD aneurysms are analyzed,
those with large, medium and small mouths. These mouths lack any boundary conditions making this an
ill-posed problem with standard computational techniques. We utilize PINN aided by 2D MRI data to
learn the hemodynamics of the domain. We conclude that full MRI resolution may not be required,
saving on scanning cost, and in the case of AD aneurysms, larger mouths lead to more accurate results
due to the larger order of magnitude in the velocities which are easier for PINNs to learn.

It is shown that both epigenetic states and cell signaling pathways have significant effects on cell
reprogramming efficiency. However, previous techniques relied on fixed cells across cell population,
lacking temporal resolution and failing to account for cell-to-cell variability. The kinetics of how the
activation of the signaling pathway or epigenetic modifications affect downstream gene expression
dynamics remain unclear. We first utilized CRISPR-mediated live imaging to investigate how LIF signaling
pathway affects the downstream gene expression dynamics in mouse embryonic stem cells (mESCs). To
visualize endogenous transcriptional activity of key embryonic genes Sox2 and Klf4 in living cells, we
used CRISPR to tag endogenous gene loci with MS2 stem loops.
We first modulated the LIF pathway activity by adding 10μM JAK inhibitor I or by removing LIF from the
cell culture media. Cells were imaged every 24 hours after each perturbation to characterize how fast
the genes could respond to the cell signaling changes. Our results showed that the proportion of cells
showing the active Sox2 signal decreased after 48hr of JAKi addition and 72hr of LIF removal. This time
discrepancy is likely attributed to the slow degradation of LIF and high binding efficiency of JAK inhibitor.
We also found that total mRNA output of Sox2 was significantly reduced after 48hr of LIF removal and
24hr of JAKi addition. The decrease in total mRNA output following LIF removal and JAKi addition was
predominantly caused by a decrease in amplitude and duration of active transcription. This indicates
that Sox2 expression in mESCs becomes more stochastic following the perturbations, resulting in a more
rapid switch between the gene’s ON and OFF state. Interestingly, we also observed transcriptional
memory in mESCs, wherein active cells (showing Sox2 signal before mitosis) are more likely to exhibit
transcription after mitosis. Our work provides a quantitative and single-cell analysis on the impact of
signaling pathway modulation in transcription and in subsequent cell reprogramming.

Utilizing biomass as a renewable and clean energy source received great attention in the past
few decades. Metal or metal oxides are common catalysts to promote biomass upgrading
reactions, but their local structure of the active sites and detailed reaction mechanisms are
generally not well understood yet. Atomic Layer Deposition (ALD) was adopted for near atomic
level control of the surface structure and thereby the active site [1]. It provides insight into the
structure of the active sites on metal or metal oxide catalysts that are active for several of the
reactions.
Dehydra-decyclization of cyclic ethers to conjugated dienes could be a key step in biomass
upgrading, since they are used as monomers for the synthesis of range of polymers. While
Brønsted acid catalysts are used for this reaction, we have recently demonstrated that solid
Lewis acids, such as ZrO₂ , also exhibit a high selectivity for this reaction. However, the active
sites of ZrO₂ are not well understood. Sodium poisoning study reveals different sites are
required for decyclization and dehydration. The reaction highly depends on the local structure
of the catalysts based on different activity observed for single crystal, ultra-thin film and bulk
ZrO₂ . Combining with DFT calculation, we conclude that decyclization reaction requires a Lewis
acid-base pair while dehydration only requires Lewis acid sites.
Given its unique catalytic properties, tunning and utilizing ZrO₂ in various system recently
receives great attention. Zeolites are used as shape selective catalysts due to their microporous
structure. Their Lewis sites requires the substitution of heteroatoms into the native zeolite
structure, resulting in electron transfer from the heteroatoms to Si for an electron-deficient
Lewis acid site [2]. We successfully generate analogous sites with similar catalytic properties in
other types of mesoporous and microporous silica-based materials using ALD as a simpler and
less time-consuming synthesis technique. The as-synthesized materials show high catalytic
activity and selectivity towards transfer hydrogenation of hydroxymethylfurfural as well as
Diels-Alder cycloaddition-dehydration of dimethylfuran to p-xylene. Combining with
characterization of FTIR and TPD, depositing ZrO₂ on amorphous silica by ALD leads to the same
type of Lewis acid sites as those in zeolites, resulting from metal heteroatoms bonded by silanol
groups in a tetrahedral coordination.

The fracture of end-linked polymer networks and gels strongly affects the
performance of these versatile and widely used materials, and a molecular-level
understanding of the fracture energy is important to the design of new materials.
The Lake-Thomas theory serves as a framework to understand and quantify the
energy dissipation due to the chain scission in these materials based on an idealized
picture of fracture in networks. Recent extensions of the Lake-Thomas theory have
incorporated the effect of topological defects, such as loop defects, and in some
examples enabled accurate prediction of the fracture energy. In this talk, I will
describe how we use coarse-grained molecular dynamics simulations and network
analysis techniques to provide a molecular view of the energy dissipated during
chain scission in polymer networks. In addition to the energy of the broken strand,
we also consider the amount of energy released by the networks connected to the
broken chain. Network analysis techniques would be used to further understand
how the inhomogeneous nature of network structure affects energy and stress
transmission in polymer networks. Network analysis also provides a surprisingly
effective approach to identifying potential failure locations in our model. Our
results can be used to further refine the description of the processes at play during
the failure of polymer networks.

Highly loaded polymer nanocomposite films (PNCFs) have improved physical properties due to
the synergistic combination of organic polymers with inorganic nanomaterials. The properties of
a PNCF could theoretically be further augmented by incorporating a polymer blend, which may
impart the desired traits of each polymer. However, blend-PNCFs suffer from similar compatibility
issues that plague current polymer blends, as most polymers are thermodynamically immiscible.
Recent work has shown that the confined geometry of highly-loaded PNCFs strongly influence
the dynamics and morphology of these materials. What is relatively unexplored, however, is how
the confinement of polymer chains within the interstitial pores of a dense nanoparticle packing can
affect the thermodynamics of polymers blends. This experimental work studies the role of
confinement and polymer-nanoparticle interactions on the phase behavior of Polystyrene and
Polymethyl methacrylate, a pair of immiscible polymers. The phase behavior is characterized by
optical microscopy, scanning electron microscopy, and neutron scattering. Promising initial
findings point to confinement causing full-scale suppression of macroscopic phase separation in
favor of a pore-scale segregation regime. By understanding the role of confinement and polymer-
nanoparticle interactions in these systems, a myriad of novel applications could be unlocked for
blend-PNCFs with finely tuned mechanical, electrical, and transport properties.

Geothermal energy production provides both low-carbon electricity and heat. To date, geothermal
resources are primarily commissioned for electricity generation, but are the least efficient of all
thermal power plants, with the global average efficiency of 12%. Binary geothermal plants are of
particular interest because their closed-loop systems further reduce the carbon dioxide (CO₂e)
emissions released during operation, however, these plants tend to have an even lower average
efficiency, closer to ~4% with respect to electricity generation. To improve the resource efficiency
of binary geothermal plants, recent studies have shown that these plants may be able to provide
low-carbon thermal energy to support technologies such as direct air capture (DAC).
DAC is one of the most energy-intensive methods of achieving CO₂ removal to date, and coupling
this to low-carbon energy is one way to maximize its carbon removal efficiency . The energy
distribution required to operate DAC is typically 80 – 94% thermal energy and 6 – 20% electrical
energy. In the case of solvent- and mineralization-based DAC, the thermal energy must be
delivered between 800 – 900^oC (high-grade), while for sorbent-based DAC, this thermal energy
can be delivered between 80 – 120^oC (low-grade). The low-grade heat required by sorbent-based
DAC makes it a promising candidate for geothermal energy and waste heat. This work illustrates
ways in which the thermal energy from binary geothermal plants can be utilized to power the
regeneration step for sorbent-based DAC. Binary geothermal plants have been previously
optimized for electricity generation, but rather, this study focuses on optimizing the integration of
geothermal energy with DAC for net carbon abated. This optimization is considered by evaluating
the integration of geothermal energy and DAC under different geothermal brine production
scenarios that vary the thermal integration of the DAC plant to determine which results in the most
net carbon abated. On a CO_2eq basis, using geothermal energy to power the regeneration cycle for
DAC leads to greater CO₂ abated than if it was used to displace grid electricity, by more than a
factor of 3. Furthermore, there may be processes that allow for a portion of the geothermal brine
to be dedicated to DAC, while the other portion is dedicated to generating electricity for local
communities. This analysis will be expanded to consider the advantages and tradeoffs of direct-
use of geothermal brine to heat the DAC contactors and those of vacuum steam generation to
determine which of these regeneration configurations is best suited for low-enthalpy geothermal
resources in the United States.

Polymer upcycling has gained major attention as an economically feasible way to tackle global plastic
pollution. Several strategies to upcycle polyolefins involve adding various functional groups to the
polymer molecules to make high-value specialty polymers. A sustainable option for such polymeric
reactions involves using catalysis to perform the reactions at lower temperature and pressure
conditions. Heterogeneous catalysts typically use highly nanoporous support materials to maximize the
catalytically active surface area. While these nanoporous support materials perform extremely well for
reactions involving small molecules, there can be significant transport limitations in case of polymeric
reactions. Factors such as polymer confinement, wettability of polymer on the solid surface, pore
geometry and interdiffusion of reactant and product polymers within the pores can significantly affect
the efficacy of the catalyst. An effective way to quantify polymer – solid interactions (a.k.a. wettability)
is by measuring polymer – solid contact angles. We have developed a technique to precisely measure
the polymer – solid contact angle using atomic force microscopy (AFM). We measured the contact
angles between metal oxides (SiO₂ , TiO₂ and CaCO₃ ) commonly used as catalytic supports and two
polymers – polystyrene (PS) and polyethylene (PE). We found that PS has stronger interactions with TiO₂
and CaCO₃ as compared to SiO₂ while the opposite is true for PE. This has implications for preferentially
upcycling a particular polymer without having to physically separate it from a mixture of polymers
present in the waste streams. The technique can be further developed to measure the 3-phase contact
angle between two polymers and a metal oxide which will tell us if the reactant and product polymers
will interdiffuse within the nanopores.

Life is dependent on water: most self-assembly and binding processes of biomolecules take place
in water. Water-mediated interactions are an essential driving force behind these processes,
which is largely affected by the hydrophobicities of the binding surfaces. In this talk, I will
describe molecular dynamics simulation approaches and results that help us understand the
interactions of surfaces with different hydrophobicities. I will show that interfacial water is
essential to tell the binding pathway involving hydrophobic surfaces. Furthermore, I will show
how the bulk concentration of ions affects the strength of interaction between a hydrophobic
surface and a charged surface. Our results can be used to illuminate some design principles for
the self-assembly of functional materials in water.

Knowing the local viscoelastic mechanical profile of physical vapor deposited (PVD) glass film is
imperative for its nanoscale application in industry, and coarse-grained simulation method is used to
simulate PVD film of N, N-bis(3-methylphenyl)-N, N’-diphenylbenzidine (TPD) molecules with varying
thickness. In the bulk, the PVD glass shows higher elastic modulus and lower loss modulus compared to
the liquid quench glass (LQG), agreed with previous report of enhanced mechanical stability.
Interestingly, a region with exceptional high elastic modulus near the substrate of PVD film has shown
gradient of strong correlation among loss modulus, molecule orientation and out-of-plane mobility, and
could be attributed to the surface-substrate coupling effect due to the nature of PVD process, which has
not been researched thoroughly before. Quantitively defined modulus difference shows strong correlation
to the potential energy difference between PVD and LQG, providing more understanding to thermal and
mechanical properties correlation. Also, underlining mechanism may also help explain the extraordinary
stability of ultrathin film.

Traumatic brain injury (TBI) is one of the leading causes of death and disability for the most
active people in the population (ages < 45 years of age). Current approaches for treating TBI
include using systemic delivery of N-methyl D-aspartate (NMDA) receptor antagonists, calcium
channel blockers, and free radical scavengers. In this work, we combine these therapeutics with
microcapsule delivery systems which are characterized by their small size (diameter of 1–1000
μm) and core–shell formulation to shorten the time between occurrence of the injury and the
administration of the drug which is a critical factor and can affect the efficacy of the drug.
Mechano-sensitive microparticles (MSMPs) are an injectable drug delivery vehicle made
through microfluidics that collapse under sufficient hydrostatic pressure, triggering the release of
a drug payload within the MSMP. Specifically, MSMPs are capsules consisting of polymeric
shell, namely poly(lactic-co-glycolic acid) (PLGA), enclosing an aqueous core containing
microbubbles. Due to these material properties, these MSMPs offer an advantage compared to
conventional microcapsules, which are deformable under uniaxial tensile forces but less
responsive to hydrostatic pressure because of their incompressible aqueous core. The core region
of the proposed MSMP incorporates both a compressible gas bubbles and aqueous phase,
allowing drug solubility within the aqueous phase and pressure sensitivity through the gas
bubbles. The drugs encapsulated within MSMPs are shielded by the polymeric shell and remain
inactive unless there is sufficient hydrostatic pressure to trigger the collapse of the capsule. The
pressure sensitivity depends on the thickness to diameter ratio (t/D) and the volume of gas within
the capsule.

Temperature-dependent shape transformations of hexadecane droplets in C₁₈TAB aqueous
solution as well as shape transformations observed in other combinations of alkanes and
surfactants of similar carbon chain lengths have been well-studied whereas the influence of
particle adsorption in such systems has not yet been fully explored. In this study, Janus particles
that have high desorption energies are incorporated into the emulsion system. When C₁₈TAB is
introduced to densely-covered hexadecane droplets, however, a significant portion of particles
desorb and the resulting droplets display shape transformations upon cooling, which are similar
to what we have observed in a particle-free system. To probe the mechanism of particle
desorption, the surface charge of particles is determined and the interactions of particle-stabilized
emulsion droplets with anionic surfactants, sodium hexadecyl sulfate and sodium octadecyl
sulfate, are also studied to understand the role of electrostatic interactions. We will take
advantage of microwell arrays that trap droplets due to density difference to investigate the
behavior of individual droplets when they come into contact with different surfactants and vary
temperature during the exposure to study shape transformations in the presence of Janus
particles.

Physics-Informed Neural Networks (PINNs) have emerged as a robust framework for solving
partial differential equations by incorporating both physical equations and experimental data.
In search of computational efficiency, training PINNs using half precision rather than
conventional single or double precision has gained substantial interest, given the inherent
benefits of reduced computational time and memory consumed. In this study, we delve into the
idea of employing pure half precision training and address its limitations. To surmount these
limitations, we explore the integration of mixed precision techniques within the Tensorflow and
PyTorch frameworks. Our investigation showcases the remarkable potential of mixed precision
training to not only substantially decrease training times and memory demands but also to
uphold model accuracy. Through empirical evaluations, we emphasize the practical efficacy of
mixed precision methods and reinforce our observations with a comprehensive theoretical
analysis. This work not only contributes insights into the realm of PINNs but also extends
research into DeepOnet problems and how they can be run with high computational efficiency.

Stimuli-responsive polyelectrolyte brushes in large-scale membranes are particularly attractive
for improving adsorption-based separation technologies. My work investigated the principles
governing gold NP (Au NP) adsorption onto a poly(2-vinylpyridine) (P2VP) brush system. The
system consisted of carboxylic acid end-functionalized P2VP brushes with molecular weights of
10 and 53 kg/mol grafted-to a poly(glycidyl methacrylate) priming layer. The two surfaces
studied were silicon wafers and quartz crystal microbalance with dissipation (QCM-D) sensors.
Here, increasing the molecular weight increased brush height while reducing the grafting density.
Prior to adsorption studies, I utilized AFM to characterize the brush out-of-plane surface
structure to ensure a uniform and flat brush surface and ellipsometry to measure the film
thickness and swelling in situ under different pH conditions. I monitored the adsorption of 10
and 20-nm citrate-functionalized Au NP solutions with the pH conditions of 4 and 6 and
molarities of 200 and 1200 pM onto these brushes on the two substrates. I measured the kinetics
of Au NP adsorption with QCM-D and visually studied the adsorption packing with SEM. With
these combined techniques, I determined that while grafting density influences adsorption
behavior, the ratio of brush height to Au NP diameter is the most important parameter for
controlling adsorption behavior.

Cancer treatment regimens include chemotherapy, radiotherapy, surgery, and targeted
therapies like immunotherapy. Oncologists use clinical data at different scales (multi omics,
quantitative single cell sequencing, tissue level tumor histology data) to tailor treatments based
on patient-specific data. This patient-specific treatment selection can be significantly enhanced
through a robust physics-driven computational framework that integrates multi omics data
across scales. We hypothesize that AI-enabled agent-based modeling (ABM) plays a pivotal role
in leveraging and utilizing patient multi omics data and serve as an effective conduit to integrate
cancer systems models that encode signaling at the cellular scale with digital twin models that
predict behavior of the cellular response in a tumor microenvironment customized to patient
cohorts.
We built a Prostate Cancer Agent Based Model incorporating a multiscale systems biology
model (cellular model integrating androgen receptor signaling, mitogenic signaling, and DNA
damage response subject to androgen deprivation therapy) already established in the lab, to
simulate growth of a heterogeneous tumor population post radical prostatectomy. The cellular
model is coupled to the ABM using cell proliferation, migration, and adhesion propensities that
are local to microenvironment signals such as mitogenic factors, drug concentrations, and
cellular state (drug sensitive vs resistant cells). Extensive local, parametric, and global sensitivity
analysis on aggressive tumor cell adhesion-motility revealed conditions of clustering and
micrometastatic formations in heterogeneous cell populations.
In ongoing work, we are developing new coupling strategies between cellular systems models
and ABM enabled via ML-based algorithms to better account for intra-tumor heterogeneity.
Specifically, we are incorporating heterogeneity using machine learning to inform cancer cell
growth rates based on local EGF and testosterone concentrations. Our results demonstrate the
potential of ABMs in improving our understanding of complex tumor behaviors. The ABM work
can be further extended to capture heterogeneous response to therapies and development of
resistance.

The use of polymer membranes for separating gas mixtures provides an energetically attractive
alternative to distillation in processes such as O₂/N₂ air separation, natural gas upgrading, or the
capture of anthropogenic CO₂ emissions. The development of novel materials for these processes
can be accelerated by using simulation to predict the performance of systems which may be
synthesized experimentally. In this work we examine several candidate systems such as a
polysulfone, polyimide, and the ladder polymer PIM-1 using a fully atomistic simulation. We
utilize the Enhanced Monte Carlo package to construct simulations from a simplified molecular-
input line-entry system (SMILES) string and perform simulations in LAMMPS using a polymer-
consistent force field (PCFF). The gas permeance and separation factor for key pairs are
calculated from the solution-diffusion model using Monte Carlo methods to predict Henry’s law
solubilities and molecular dynamics to calculate the diffusivities of gases in the system and
compared with experimental results from literature. Finally, we describe how a neural network
model developed by collaborators allows for the prediction of new polymers to be studied
computationally and eventually synthesized for use in separations processes.

Drug carriers play a critical role in pharmaceutical industries. Existing commercial drug
carriers are typically chemically synthesized, causing them polydisperse in size, not
biocompatible and making it difficult to functionalize for biological applications. These
limitations have prompted an extensive effort to create drug carriers from other molecules. A
powerful alternative to chemical synthesis is the expression of recombinant proteins through
molecular biology. These proteins have the exact amino acid sequence as dictated and
monodisperse in weight. They would allow the direct incorporation of specific motifs that
mediate biological activities and permit morphology control through sequence design. In this
study, we have chosen a naturally occurring plant protein oleosin, which are surfactant-like and
can stabilize oil bodies, as a starting candidate, which has been demonstrated to self-assemble
into a variety of nanostructures. The excellent hydrophilicity of recombinantly derived resilin
offers unique opportunities in creating amphiphilic fusion proteins with other hydrophobic
peptides to result in interesting morphology. We designed a family of fusion variants of oleosin
and resilin-like-peptide (RLP). We used cryo-TEM to confirm self-assembled structure from
fusion variants. We studied the morphology of different variants as a function of their
hydrophobicity and geometry design. Our results indicate current oleosin-RLP variants would
self-assemble into rod-like-micelles and vesicles in a physiological buffer. We are also
incorporating integrin-binding motif RGDS that is protected by a protease recognition site to
achieve protease-mediated cellular uptake.

Bicontinuous interfacially jammed emulsion gels (bijels) are soft matter consisting of two
immiscible liquid phases arranged in an interwoven configuration where liquid-liquid interfaces
are covered by a monolayer of colloidal particles. Bijels are formed by inducing spinodal
decomposition and kinetically arresting this process via interfacial attachment and jamming of
colloidal particles. The unique morphology of bijels makes them ideal for applications requiring
inter-phase mass transfer or particle-laden interfaces. While bijel formation has traditionally used
thermal quenching of homogeneous binary liquid mixtures to induce phase separation,
limitations of this method have led to the development of a solvent transfer-induced phase
separation (STRIPS) bijel fabrication process. In STRIPS, a cosolvent compatibilizes two
immiscible liquids to obtain a homogeneous ternary mixture and phase separation is triggered by
removing the cosolvent via liquid phase diffusion. One benefit of STRIPS is the addition of
cosolvent-related processing parameters that provide new ways to exercise control over bijel
structures. However, the STRIPS bijel processing parameter space is large and thus makes the
experimental methods used to understand processing-structure relationships tedious. Therefore,
the goal of our research is to use simulations of STRIPS bijels to efficiently study large
parameter spaces and provide an enhanced understanding of STRIPS bijel processing-structure
relationships. This research seeks to develop a computational model that simulates STRIPS bijel
formation, capture the processing-structure relationship using machine learning, and perform the
inverse design and synthesis of custom STRIPS bijels. Current research emphasis is on
understanding the effect that cosolvent partitioning between the immiscible phases has on
interfacial tension and final morphology.

Polymer electrolytes are promising replacements for hazardous liquid electrolytes used in commercial
lithium-ion batteries given their favorable mechanical properties and stability. However, the strong
coupling of ion transport to slow polymer dynamics limits the conductivity of solid polymer electrolytes
to values below those required for commercialization. This report details our approach to decouple ion
transport: assemble ionic aggregates into layers and selectively swell the ionic layers with solvent. This
preferentially solvates the conducting cation by the fast-moving solvent rather than the slow-moving
polymer. Both computational and experimental work demonstrate how the addition of ethylene
carbonate solvent into a single-ion-conducting lithium sulfonate based multiblock copolymer (PES12Li)
improves ion transport and seek to elucidate the mechanisms behind how solvent selection dictates
electrolyte performance. We anticipate that further investigations of solvent-polymer systems can
produce electrolytes with decoupled ion transport and vastly enhanced conductivity at modest
temperatures for use in commercialized electrochemical devices.

Leukocytes undergo “Leukocyte Adhesion Cascade” to reach sites of inflammation.
Amidst this cascade, a fascinating phenomenon occurs – upstream migration. As
leukocytes adhere to endothelial surfaces through LFA-1 and ICAM-1 interactions,
they resist the prevailing flow and migrate against it. This phenomenon, observed
both in vivo and in vitro, remains elusive apart from the integrin-ligand pair that is
responsible for the motion.
Traction Force Microscopy, a technique revealing cellular traction stresses, offers a
window into this enigma. It employs the use of an elastic substrate that contains
beads that can be imaged. As cells apply traction forces, the substrate deforms,
causing bead displacement. This displacement is captured via microscopy and
computational algorithms reconstruct it into traction maps. In our project, we prepare
a polyacrylamide gel embedded with fluorescent beads as a substrate, functionalize
it with ICAM-1 molecules, and assemble it within a microfluidic chamber. After
introducing the cells into the chamber, controlled media flow is initiated, and capture
30-minute timelapses observations are captured. Utilizing a computational algorithm
developed by Wang and Dembo, we reconstruct the displacement of the beads
within the substrate into traction force maps, we reconstruct bead displacement into
traction force maps, unveiling the temporal dynamics of the traction stresses and
spatial movement of the cells.
Traction maps generated shows how the lamellipodium and uropod projections
interact to generate traction stresses that allow the cells to go upstream. Through
this, we gain key understanding into how cells exert traction stresses to migrate
upstream. By comprehending leukocyte motion, we unlock diverse engineering
possibilities and therapeutic innovations, ushering in new horizons for biological
engineering of leukocytes.