Past research has discovered a technique for photocatalyst synthesis utilizing never-dried cellulose fiber in order to produce non-agglomerating nanocrystals. This new technique allows the production of semiconductors in a flexible form. This form is useful in many areas, including in spacecraft technology, where solar panels must be kept in a compact form before deployment. One possible catalyst for this is gallium arsenide (GaAs), which holds many traits that are favorable to semiconductor devices in electronics. With a band gap that is receptive to visible light and many industry applications in radio frequency, light sensors and LEDs, gallium arsenide is an attractive candidate for storage in cellulose fiber. In particular, this storage method allows for the synthesis of long strands containing GaAs that are useful in solar cells. In order to form an electrostatic region necessary for electronic applications, a P-N junction must be formed within the compound. This is usually done by heating the compound and injecting ions into the structure so that P-N junctions are naturally formed. However, since the fibers are unable to be heated to the temperatures needed without compromising the storage, we instead propose that a magnetic field can be placed across the semiconductor during formation, with doping ions placed within the fiber so that the junctions naturally form within the GaAs structure. This new method for semiconductor production has the potential to offer a new strategy for electronic production that can exist in a flexible format.

Water can have many contaminants that are harmful to humans and animals alike. Since water is so essential to life on this planet, it is important to know whether or not a water source is contaminated. However, testing water samples can be expensive. In order to make it easier to test water for contaminants, we are working to create a device to sense this contamination within a matter of hours by using yeast that produce a fluorescent protein in the presence of various water contaminants. In order to make the yeast easier to transport, we first need to dry the yeast before packaging it, similar to dry yeast that can be bought at the grocery store. To test a sample, the device utilizes a light to frequency sensor to convert the intensity of light from the glowing yeast into a signal that can be processed by an arduino board contained within the device. Our device can be used as a reusable, low cost diagnostics machine for testing water for contamination. This could provide access to advanced technology in parts of the world that don’t have access to expensive testing equipment, as well as  be used by an ecologist in the field.

Our current project is one stage in a larger research goal: to build multiple yeast strains capable of detecting toxic water contaminants. We searched for the optimal contamination concentrations for yeast response to obtain robust, accurate RNA sequencing analysis results. This data will then be used to determine which contaminants elicit a strong enough genetic response for promoter attachment. The fluorescent yeast will emit a recognizable signal in this manner, and allow for reliable water assessment without the use of expensive equipment or testing services. Thus, we investigated the optimal times for RNA extraction, when the cells are expressing the most contaminant-induced genes but are still sustaining metabolic processes. To this end, we must consider factors such as contaminant concentration and time after exposure. At this stage, data collection involves the use of a plate reader, which periodically measures absorbance of the yeast solution. Absorbance, which is proportional to population density, is limited by the solution’s carrying capacity. We then generate population trend graphs and identify the greatest instantaneous growth time in each sample. Through repeated experimentation, we compile a spectrum of contaminant types, concentrations, and yeast carrying capacities. Analysis of these growth curves indicate the time of incubation at which RNA expression is ideal for sequencing. Future work will involve identifying additional contaminants of interest and additional modification of our algorithms to obtain more precise, accurate, and meaningful data.

Water pollution is a serious problem that not only threatens wildlife, but also the health and safety of human populations around the world. While the commonly held perception is that water pollution only affects vulnerable groups in developing countries, the 2014 water crisis in Flint, Michigan is a stark reminder that even developed nations are not impervious. Given the scope and weight of this problem, there has yet to be an accurate and affordable way to reliably test for heavy metals and other water-soluble contaminants. Current solutions such as handheld digital instruments can be expensive and require calibration and electricity, making them less practical in low-resource settings. Test strips, while deployable, are often inaccurate. Hence, our aim is to create an inexpensive, deployable yeast-based diagnostic tool that can detect various water pollutants with high sensitivity and report results with a fluorescent or color output. At the beginning stages of our project, we are using Next-Generation RNA Sequencing and experimenting with various analysis techniques to screen for genes in Saccharomyces cerevisiae that exhibit a unique and differential expression profile after exposure to a particular contaminant. For now, we are limiting the scope of our research to zinc, copper sulfate and caffeine because there are existing gene sets in literature for us to reference, but as more data are acquired, we hope to apply this pipeline to a wider range of chemical pollutants, including pesticides, toxins and hormones.

90% of drugs fail in clinical trials. As a result, a successful FDA-approved drug costs an average of $2.6 billion to develop and takes a decade to reach the market. Pharmaceutical companies are unable to comprehensively evaluate drug toxicity before clinical trials because no methods exist to screen the thousands of protein-protein interactions (PPI) that a drug could unintentionally disrupt in the human body. The ability to disqualify drug candidates with off-target effects before clinical trials would greatly increase the speed at which useful drugs are introduced to the market and reduce the expense of developing new therapeutics for those who need them. Here, we demonstrate a novel synthetic biology-based method, SynAg, that enables thorough drug characterization by linking PPI binding strength to the cellular agglutination of engineered S. cerevisiae cells. SynAg is a library-on-library approach that conserves the accuracy of pairwise PPI toxicity screening in a high-throughput, one-pot format. Specifically, we construct a SynAg library to characterize a 20-by-20 protein interaction network involved in cellular apoptosis to demonstrate the accurate profiling of a complex and highly interconnected family of proteins. Finally, we use the SynAg library to characterize the on- and off-target effects of small-molecule cancer therapeutics, which are shown to be consistent with previous studies. 

The Jasmonic Acid (JA) sensitive system is an important stress and pathogen resilience pathway in plants. Understanding the kinetics of the components of this system allows for a more rational approach to improving crop resilience than traditional breeding.  MYC transcription factors upregulate the expression of particular defense or stress response genes, and these are inhibited by JASMONATE-ZIM-DOMAIN (JAZ) proteins, which are in turn degraded in the presence of JA-Ile, the bioactive form of JA. This leads to a system whereby MYC transcription factors are only active when JA-Ile is present. The variable defenses plants are capable of mounting may be regulated by differential interactions between some of the 12 JAZ proteins and 3 relevant MYC proteins in Arabidopsis thaliana. We have used a fluorescent yeast model to characterize the differences in those interactions in order to predict changes in plant responses when particular JAZ proteins are up or down regulated. Although we have been unable to replicate the canonical direct repression of MYC transcription factors by JAZ proteins, repression was observed when the JAZ protein was fused with an Mxi1 repression domain. This allows us to determine the relative effect of each JAZ protein on MYC transcription factors. A rough model involving related Hill functions is built using this information which can be used to predict the effect on the downstream response based on the abundance of the different JAZ proteins. The model can then be verified in planta using a dCas9 transcription factor with an appropriate gRNA, and a MYC-responsive GUS gene in a transient transformation.

One issue in many medical applications is detecting the presence of forced exhalations or breaths through audio data. For example, analysing sleep disorder breathing or analysing a person's lung function from audio of their exhalation first requires localization of a breath event in an audio file. Past work has focused on extensive feature extraction from sound and machine learning methods to detect these events. However, these black box methods require significant feature engineering dependent on the type of exhalation and environment the audio data is recorded to find the start and end of an exhalation event. We present a novel method for detecting the start and end of forced exhalations for spirometry, a medical test of lung function, in audio files. The algorithm requires an initial set of audio files containing exhalations to be manually labelled for locations where exhalations can be heard. Then, we build an envelope which encapsulates the shape of our exhalation signal. We take the cross-correlation of this initial envelope with vast amounts of unlabelled audio data containing exhalations, and use the estimated locations of exhalation events to recalculate the envelope. We repeat this process with the new envelopes until convergence. The converged envelope is then used to autocorrelate an audio signal to detect the start and end of exhalation. While preliminary results show that this method only achieves comparable performance to ad-hoc approaches currently used compared to human labels, it has the benefit of being simpler and more interpretable, enabling easier improvements and runtime of the algorithm.

Seismic reflection surveys use acoustic energy to image the structure beneath the seafloor by broadcasting broadband impulsive sound signals (called airguns) that reflect off the sea floor and are recorded through long arrays of hydrophones (called seismic streamers). During these surveys it is important to have a model to predict how much noise will be added to the ocean environment. The complexity of local geology, seafloor topography, and uncertainty in water properties makes it difficult to mathematically model acoustic propagation in ocean. However, ocean data like many other branches of science is experiencing an explosion in the amount of data collected and available for analysis. The goal of this research is to use previously recorded ocean data to (1) calculate the sound power levels generated by airguns during seismic reflection surveys and (2) use machine learning to create a predictive model of airgun noise considering various variables such as ocean depth and distance from the airgun. Several regression methods in Python were used and the method that resulted in the most accurate regressor was chosen. The data used in this experiment spanned a wide range of water depths from the continental shelf (∼40 m) to deep water (∼2600 m). In this presentation, the performance of a machine learning algorithm trained by the sound power levels calculated through one track line of this cruise is investigated.

The Nanoscribe is a high-resolution (500nm) Nano 3D printer that is new to the Washington Nanofabrication Facility. It was acquired through a National Science Foundation (NSF) Major Research Instrumentation (MRI) grant. The tool uses a technology called two-photon polymerization. This involves the use of photoresist, which is a light sensitive material, and an infrared laser that directs brief pulses of light through a microscope objective in different patterns to create 3D objects. The highly focused laser beam allows for dimensions of less than 500 nanometers in X and Y directions to be achieved. One major limitation of the tool is that unlike most photolithography equipment, it is not programmed to do basic alignment. Most fabrication applications require at least two layers of materials to be aligned. For example, when making electronic circuits, there is a top and bottom layer at a minimum, and a vertical layer in between, thereby requiring at least two alignments to the first layer. Our research is focused on understanding and testing test code to add pattern recognition alignment capabilities to the Nanoscribe instrument. For our test application, we applied test aligning microneedles to pre-patterned electrodes fabricated using conventional photolithography. We then 3D printed microneedles that are subsequently used to release drug therapies in test wells via an applied electrical potential (voltage). The alignment process that we developed can be used as a basic construct and capability for other applications being developed on the tool.

Large data centers, such as those built by Google, Amazon, and other information technology leaders consume about 1.3% of the world’s energy, of which about 40% is used on electronics cooling [1, 2]. This amounts to 245 TWh per year, which, with the average US price of 12 cents per kWh, amounts to about $29.4 billion dollars spent per year on cooling high heat semiconductors [3, 4]. The work presented here proposes an innovative way to improve this cooling process. The proposed concept features a levitating inner rotor using fluid bearings that result in no physical contact between solid parts, eliminating friction. For the first time, precision-manufactured plastic parts are utilized to achieve both a low cost and a high reliability. The micropump is expected to last in operation for over one million hours Mean Time to Failure. This work emphasizes model-based design verification and optimization to ensure adequate performance for different form factors – so that a drop-in replacement of an air fan passive heat sink can be quickly developed for every microelectronics product. Twenty-four designs and prototypes were used in evaluation of two key criteria in order to optimize the pump’s design. Three separate herringbone geometries, square, beveled-step, and circular, of herringbone grooves were prototyped based on experimentation of optimum groove parameters. These findings helped determine the optimal layer height of 100 micron for use in the micropump design. Finally, the application of a sensorless, brushless DC motor reduces overall cost of the pump and increases efficiency due to the removal of friction.

Deep Reactive Ion Etching (DRIE) is used for high aspect ratio etches in silicon wafers. The process is effective at achieving deep etches with high anisotropy and selectivity, however more process development is required such that side wall profile angle is selectable based on etch parameters. The goal of this study is to manipulate process variables (including pressure, etch time, and temperature) to allow the production of tapered, reentrant, or vertical side walls as needed. The SPTS-DRIE will be used to etch silicon targets with a repeating array of test patterns using a standard etch and deposition recipe. The standard etch recipe used will cycle between polymer deposition steps and plasma etching steps at low pressure using C4F8 and SF6 gases. Etch depths will be verified using a Bruker DekTak profilometer and sidewall angles will be measured using a Jeol scanning electron microscope. The parameters of the etch recipe will then be adjusted depending on the previous etch profile. Depending on the profile and aspect ratio of the etch, different parameters will be studied, including the process pressure, temperature, gas flow rates, and etch/deposition times. This process will be repeated until results within a reasonable margin of the target are reached. The goal will be to obtain an etch profile with sidewall angles of 90° ± 0.1° for use in industry and to determine parameters for reentrant and tapered profiles.

Etching is one of the critical process steps in the fabrication of micro- and nano-scaled devices. Prior to etching, a pattern, usually defined in a temporary photo-sensitive polymer mask material called photoresist, is transferred onto a silicon wafer. One way to etch material, is through the use of plasma. Inductively Coupled Plasma (ICP), is distinct for its coil that shapes plasma and DC bias coupling for directionality. The machine has high etch rates and is used to etch oxide. Specifically for this project ICP-F (Fluorine) has been used. During etching, a photoresist reticulation problem has occurred. With this reticulation, features on the wafer are degraded, when reticulation is wide spread, the whole wafer must be stripped and restarted. This problem affects the speed and quality of production of wafers. In attempts to fix this problem, different variables such as the DC bias, chuck temperature, He backing, and RF power are monitored. Furthermore, the effects of varying pre-etch conditions, such as wafer bakes and different wait times, are examined to determine if there is a correlation between moisture in the photoresist film and resist reticulation or if the issues are predominately related to the etching system. Different inspection steps are made to ensure the quality of the etch, such as visual inspections and measurements of etch depths. These inspection steps concludes how much reticulation has occurred and what it has affected. ICP-F etching is just one step of the process in of fabrication of micrometer scale inductors and capacitors directly on the top of silicon wafers referred to as Integrated Passive Devices (IPD). This process decreases the amount of space used and increases the bandwidth of radio frequency (RF) analog filtering and the speed of digital information transfers. 

In the fabrication of integrated circuits and microscale devices a technique known as electroplating is often used to deposit thick layers of metals such as gold and copper. The electroplating process requires a conductive layer to facilitate electroplating, commonly referred to as a “seed” layer. In many cases, the seed material will not bond well with the base materials, necessitating the use of an intermediate “adhesion layer” to improve adhesion between the substrate material and the plated metal. In this project, Electron Beam Physical Vapor Deposition (EBPVD) was investigated as a means to deposit a 25nm titanium adhesion layer followed by a 300nm copper seed layer to promote the adhesion of electroplated copper to a silicon dioxide substrate. EBPVD is a process in which a collimated beam of high energy electrons is used to boil metal atoms off of a target and deposit them on a substrate. These systems operate in high vacuum to increase the mean free path of metal atoms ejected from the target surface, allowing for highly anisotropic deposition on the substrate. The stress states of EBPVD deposited seed layers and electroplated wafers were also observed as residual stress can adversely affect the performance of active circuitry that IPDs are bonded to. The deposition of this seed layer, followed by electroplating of copper, is but one step in the fabrication of prototype signal filtering devices for mobile applications. These micrometer scale inductors and capacitors, referred to as “Integrated Passive Devices,” are built directly atop silicon wafers, with the intent of eliminating the need for larger surface mounted signal filtering devices in applications where working volume is at a premium, such as in mobile phones and tablets. We hope our research will allow for production of higher quality prototypes and accelerate development of this emerging technology. 

An Inductively Coupled Plasma fluorine (ICP-F) reactor is used to anisotropically etch silicon dioxide with octafluorocyclobutane (C4F8) and argon (Ar) gas. Microelectromechanical system (MEMS) and microelectronics applications require several micron wide features with high-resolution and precision features made on silicon wafers. The ICP-F is a fundamental of the pattern transfer process called a damascene process where a precision pattern is formed and then filled with metal and planarized to remove excess metal and leave the intended pattern. This step of the damascene process consists of lithographic resist patterning and subsequent plasma etching of the underlying material. In this work, C4F8 plasma with Ar additives are used to etch silicon dioxide to create patterns in the oxide that will later be used molds for copper electroplating. In this design, the ICP-F must have a consistent etch rate with high selectivity to achieve constant depth for the copper layer with high resolution patterns. Variables such as DC bias, Ar flow, and O2 flow were optimized to achieve a repeatable etch rate. Furthermore, wafer stress was measured at every step in the process as the varying stress of the IPD could adversely affect the electronic circuits that the IPDs are connected to. Compared to Discrete Surface Mount Devices (SMD), IPDs allow for high-density trench capacitors, Metal-Insulator-Metal Capacitors, and high-Quality inductors which are enabled by ICP-F technology. In addition, new radio frequency (RF) circuits for 5G applications are employing the use of Integrated Passive Devices (IPDs) for making RF filters.

Computer science and neurobiology have an undeniable entanglement with one another in the 21st century with the advent of advanced artificial intelligence and attempts to understand the black-box model of a brain. However, most neurobiology research has revolved around understanding existing behaviors in large animals, with commonly researched organisms including zebrafish and mice. Introducing novel behaviors to animals in a controlled environment is undeniably attractive in modern neurobiology. With the use of mass data-collection in the online Aquarium lab environment, this study shows the potential to model and produce novel behaviors in Hydra vulgaris (freshwater anemones with simple nervous systems). We aim to create a high-throughput, data-oriented, model for gene-based behavior analysis and editing in neurobiology research. By creating a single controlled environment to perform all research jobs on mass-produced H. vulgaris strains with clockwork husbandry, extraneous factors affecting the organisms are significantly reduced. By taking data on the success of every modification of these Hydra, we can model the effects of specific factors on these animals and progress toward creating a model organism for neurobiology research akin to knockout mice in medicinal genetics. With these methods we have been able to produce a significantly more efficient process of gene editing in Hydra and with several different kinds of DNA integrants. Over the long term, the wide availability of the online Aquarium lab can yield the discovery of unknown trends in genetic knockouts and the discovery of novel behaviors through analysis of all data collected in this controlled environment. Here we highlight successes thus far and future developments necessary to make this system a viable model for gene-based neural engineering over this century.

X-ray diffraction gratings are an array of equally spaced structures with defined width designed to diffract optical waves. These waves create a spectrum that allows for the characterization of the wavelength or angles created by each diffraction. X-ray diffraction is commonly used in x-ray crystallography, where x-rays are diffracted off crystal lattices and give information about its three-dimensional structure. The goal of this project is to develop a microfabrication process for high aspect ratio x-ray diffraction gratings. These devices will expand the x-ray diffraction capabilities of the University of Washington and the National Nanotechnology Coordinated Infrastructure (NNCI) at large. In developing this process, our main objectives are to attain smooth vertical sidewalls for the grating features, minimize gold waste in the plating process and to be highly repeatable. For our research, each fabrication step will be tested, troubleshot, and documented for future use. The microfabrication steps required in our process include photolithography, etching, grinding and polishing, wafer bonding, and electrochemistry metal plating. The main challenges of this project are producing a photoresist pattern thick enough to withstand deep etching without affecting the sidewall profile, attaining precise grinding thicknesses and overcoming wafer bonding complications. The work towards maturing each process through inspection and troubleshooting for fabricating high aspect ratio x-ray gratings will be discussed.

This project addresses the problem of community membership detection on social media using only text features in a scenario where a small number of positive labeled examples defines the community. For example, if we are given the tweets of chess grandmasters, then the community is defined as chess grandmasters, and the model we propose should identify people likely to be chess grandmasters. We introduce user embeddings (dense vector representations of the users) trained on an unsupervised proxy task: user re-identification. Experiments with 16 different communities show that the resulting embeddings are more effective for community membership identification than common unsupervised representations, measured both by average AUC and 1/MRR measure.

UW's electricity bill is around $1MM per month, making it Seattle City Light's second-largest customer. Solar power represents one way that UW can reduce load on the city's power grid. In 2017, UW Housing and Food Services completed installation of four solar panel arrays to its buildings on campus. How have those panels performed, and has UW benefited from their presence? To gain insights into the effect of these installations, we developed the UW Solar Power Monitor to collect and present data about solar power usage within these buildings. The data was analyzed, interpreted, and integrated into the dashboard. Our hope is that the tools we have developed will enable UW administrators to make informed decisions about power infrastructure on campus, educate the public and promote awareness about solar projects on campus, facilitate the study and analysis of solar power, and encourage increased investment in solar infrastructure at the University of Washington.

Raven is a research prototype surgical robot that helps a surgeon perform surgery remotely. The purpose of my project is to develop a program to keep the surgical camera automatically focusing on the current surgical location. I wrote a GUI (graphical user interface) to perform object tracking and eye detection through the programing language Python with the OpenCV library. I created a ROS (robot operating system) node to program Raven. I expect the result will make Raven a more convenient tool for performing surgery remotely in the future.

This project aims to influence the next generation of flying robots by first studying how moths use multi-sensory information to increase their agility. These different types of multi-sensory information include visual- and touch-based feedback, which are influenced by forces such as those from flapping wings and changes in body posture (which affect a moth’s inertial distribution). To visualize and analyze moth flight in a controlled environment, the goal is to implement a virtual environment, in which projected video will simulate a changing environment and provide a haptic-type virtual reality interface for the moths. Varying the dynamics of the moth’s motion (for example, angle of rotation and speed of movement) in this manner will enable an analysis of how the moths respond to inertial and aerodynamic forces. The first engineering challenge is to design and build a system to electronically control and monitor the motion of the moth while having a minimal affect on a moth’s inertia and natural flight patterns. This requires the design to have low mass, inertia and friction. After this, cameras and sensors will be used to record data that will contribute to a more realistic understanding of how the principles of animal flight can be used in robotics.

Modeling how electromagnetic waves scatter from a distribution of rough plates poses many applications. Certain systems may be easy to approximate with planar geometry, but use of numerical field solvers to determine the radiated fields could take a long time for nontrivial structures. We propose an iteration on existing formulas based on the Kirchhoff approximation. This iteration will consider the case of multiple rough, finite-sized rectangular plates. We developed software to determine the scattering of waves off of a distribution of rough plates, of arbitrary position and orientation between a transmitter and receiver. The method considers each plate individually, calculating the coherent and incoherent scattered fields. Provided all plates and the transmitter and receiver are sufficiently spaced, we calculate the total fields by summing the result from each individual plate. We verified this method via comparison to numerical data. This model can then be used for determining the radiated fields from unwanted reflectors in urban environments such as buildings or car rooftops. The software could be additionally used to optimize a distribution of reflective plates to achieve a desired radiation pattern.