Community College Participants and Projects: 2021
Despite the ongoing COVID-19 pandemic and campus closure, the Center for Energy Efficient Electronics Science hosted nineteen community college students for an online summer research internship. This is the biggest cohort the TTE program has ever hosted, and participants shared that they had a very positive experience. At least thirteen interns have been invited to present their research at the 2021 SACNAS Diversity in STEM conference, two have been formally invited to join the E-TERN program, and an additional subset will continue their research in the 2021-2022 academic year.
The 2021 TTE Research Symposium can be viewed here
Undergraduate Researcher: Adam Abraham
Intended Major: Civil Engineering
Home Institution: Los Angeles Pierce College
Research Project: Computational Image Processing for a Point-of-Use Soil Health Assay
Faculty Advisor: Dr. Romy Chakraborty
Mentor: Dr. Sara Gushgari-Doyle
Lawrence Berkeley National Laboratory
Project Abstract: Microbial respiration is an important biological metric that can be used to measure significant carbon dioxide (CO2) levels and indirectly reveal the health of microbial populations within the soil. The goal of this research is to provide real-time monitoring of soil health through the improvement of in-field soil health assays. This is accomplished through analyzing colorimetric assays that use the pH indicator cresol red to estimate the CO2 produced by microbial respiration from prospective soil samples in PCR tubes. We are building an iOS application that processes assay images, analyzes the data, and gives informative feedback on soil health, making it accessible to all via smartphone. Project Poster.
—
Undergraduate Researcher: Jacob Ahmed
Intended Major: Civil Engineering
Home Institution: Alameda City College
Research Project: Liquid Metal Bubble Synthesis and Simulation of Atomically Thin 2D Nitride Materials
Faculty Advisor: Prof. Zakaria Y. Al Balushi
Mentor: Jiayun Liang
UC Berkeley Department: Department of Materials Science and Engineering
Project Abstract: In the family of two-dimensional (2D) materials, group-III nitrides are extremely versatile in their applications in electronics and can be modified physically, electrically, and/or chemically. However, current synthesis approaches make it difficult to extract 2D materials that are in chemical spaces under extreme conditions. In this project we approach the novel synthesis method of 2D GaN, where 2D GaN is formed via a three-phase bubbling process through liquid metal as the synthesis medium. To provide theoretical guidance for future experiments, we will utilize COMSOL Multiphysics, a finite element simulation program, to simulate and analyse bubble formation in liquid gallium for the synthesis of 2D GaN. The goal of the simulation is to develop and understand the physics behind three-phase flow models in a Finite Element Simulation, apply them to the synthesis of 2D GaN, and analyse how the radius, initial velocity, and gas composition affect the bubble’s shape and the time needed for it to burst. Our final goal is to realize the synthesis of 2D GaN at room temperature and further apply it in various energy harvesting applications. Project Poster.
–
Undergraduate Researcher: Iretunde Akinsola
Intended Major: Computer Science
Home Institution: Las Positas College
Research Project: Visualizing Engineering Design Decisions
Faculty Advisor: Prof. Kosa Goucher-Lambert
Mentor: Yakira Mirabito
UC Berkeley Department: Department of Mechanical Engineering
Project Abstract: Concept selection is a vital point in the product development process as it has direct impact on product success and efficient design behavior. The decisions engineers take during concept selection, leads to designs with optimal performance. Not only does efficient design behavior increase the likelihood of developing new innovative concepts. Efficient design behavior also decreases the likelihood of product postponement and saves funds for redesigning. This research utilizes an existing dataset from a human subject study conducted in 2020 by the Goucher-Lambert Lab. In the study, 57 participants were given 10 minutes to engage in the concept selection process of a gripping surface for a robotic arm. Participants explored different options but were permitted to test no more than 5 designs which resulted in a unique design journey. In this study, the existing dataset was extracted then cleaned to isolate 4 participants and two variables. These 4 files served as the dataset for this research.The software, R, was the foundation for the data analysis portion of this research. There was a strong interest in analyzing specific decision points and their level of influence in the design journey. Thus, the extracted dataset was imported into R and 4 scatter plots were built to visualize each engineer’s design journey. Results indicate that each engineer had an influential point in their design journey that either led them to the optimal design or put them in position to find it. Project Poster.
–
Undergraduate Researcher: Mariam Alsaid
Intended Major: Environmental Science
Home Institution: Sacramento City College
Research Project: Enrichment of Subsurface Microbes with Differing Carbon Substrate
Faculty Advisor: Dr. Romy Chakraborty
Mentor: Dr. Kristine Cabugao
Lawrence Berkeley National Laboratory
Project Abstract: Carbon sequestration is a necessary mechanism that mitigates the threats of climate change by preventing the accumulation of potentially harmful carbon-containing molecules (ex. CO2) in the atmosphere. Although the majority of research on the carbon cycle consists of above-ground pathways, there is evidence that suggests that soil microbiota greatly contribute to the efficiency of carbon sequestration; this is due to their significant role in the production of soil organic matter, the largest terrestrial carbon pool. To investigate this, we enriched and isolated microbial community members from various soil depths in different enrichment cultures. We found that microbial community compositions will differ depending on the carbon source of each enrichment and initial soil core depth. This will hopefully pave the way for more research focused on manipulating the soil environment to promote the growth of efficient carbon-sequestering microbes. Project Poster.
–
Undergraduate Researcher: David Belman
Intended Major: Electrical Engineering
Home Institution: Santa Monica College
Research Project: Memory Optimization for Delay-Based Optoelectronic Reservoir Computing
Faculty Advisor: Prof. Ming C. Wu
Mentor: Phillip Jacobson
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: The resurgence in popularity behind AI/machine learning revolutionized the rate at which we could analyze data, and sparked interest in AI/machine learning tailored for specific tasks. This gave rise to reservoir computing (RC) systems, which implement the use of randomly initialized virtual nodes for training to be reduced to linear regression. We have adapted the traditional RC system with an electro-optic modulator as the single working physical node, and an optical fiber cable to introduce delayed feedback back into the system. Our RC system defines its memory via the length of the optical delay fiber and low-pass filtering. These features determine the system’s maximum storage capacity and dictate how much memory is accessible to the system while running, respectively. The purpose of our research is to leverage simulations to determine a zone of predetermined values where both the optical fiber length and low-pass filter allow the RC system to perform optimally. We hope to gain further insight into this system, and potentially utilize it as a working prototype as an image or video classifier. Project Poster.
–
Undergraduate Researcher: Christina Chau
Intended Major: Physics
Home Institution: College of Alameda
Research Project: Low Resistivity in Silver Complexes of Ag2Te(MS2)3
Faculty Advisor: Prof. James Analytis
Mentor: Daria Balatsky
UC Berkeley Department: Department of Physics
Project Abstract: The chalcogenide compounds Ag2Te(MS2)3 (M=V,Nb), first synthesized by the Kanatzidis’ group, are of interest because of their unique properties. These two crystals exhibit low resistivity and may give access to other types of transport. We synthesized and characterized the two crystals to henceforth better understand transport at lower temperatures, where classical transport theories break down. A solid state method was used to synthesize both crystals. This work added to the characterization work of the Kanatzidis’ Group including calculating the RRR and the Hall Effect. The RRR values of the V and Nb sample are 19.73 and 44.22 respectively. Hall bar measurements of the Vanadium crystal reveal the predominant charge carriers are holes. Project Poster.
–
Undergraduate Researcher: Jesse Horowitz
Intended Major: Mechanical Engineering
Home Institution: Cuesta College
Research Project: Pump-Probe Measurement of Coherent Phonons in Ge, GaAs, and MoS2
Faculty Advisor: Prof. Juanqiao Wu
Mentor: Sarah Warkander
UC Berkeley Department: Department of Material Science and Engineering
Project Abstract: When an ultrafast laser pulse is incident on a material, a sudden rapid increase in temperature generates coherent phonons that can be seen as a clear sub-nanosecond oscillatory signal. The amplitude, frequency, and dampening rate can be related to material properties such as directional speed of sound and refractive index. The goal of this work is to use a pump-probe system to explore the thermal, electrical, and optical properties of semiconductors including germanium (Ge), gallium arsenide (GaAs), and molybdenum disulfide (MoS2). Phase and magnitude data will be analyzed by fitting to simple mathematical models and results will be compared to tabulated material properties. The phase fit parameters in GaAs were used to yield a theoretical longitudinal sound velocity of 4755 m/s. Our analysis found that oscillation amplitude of the phase signal increased linearly with pump power. Further analysis of these trends will lead to a better understanding of the phonon properties in our sample materials. Project Poster.
–
Undergraduate Researcher: Christopher Keokot
Intended Major: Computer Engineering
Home Institution: Cosumnes River College
Research Project: End-to-End Design of a Snapshot Hyperspectral Microscopy System
Faculty Advisor: Prof. Laura Waller
Mentor: Eric Markley
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: In microscopy, hyperspectral imaging has been used to identify fluorescent labels in cells and tissues. Snapshot methods for macroscale hyperspectral imaging traditionally requires bulky benchtop setups or have low spatio-spectral resolution. Recently, we have shown the feasibility of high spatio-temporal resolution for macroscale hyperspectral imaging while maintaining a small form factor and sufficient spatio-spectral resolution by placing a diffuser in front of a camera sensor with a tiled hyperspectral filter array (HFA) and using computational imaging approaches with compressed sensing to reconstruct the hyperspectral datacube. In previous work, we used a randomly tiled HFA and off-the-shelf diffuser as part of the computational imaging setup; here, we propose adapting recent approaches for end-to-end learning in order to jointly optimize the HFA and the reconstruction algorithm. The setup uses a differentiable forward model for the imaging system as well as a differentiable reconstruction algorithm, allowing for the use of gradient-based optimization approaches. We explore the use of a neural-network- based reconstruction algorithm for recovery of the hyperspectral datacube. Ultimately, we demonstrate the efficacy of our design by simulating reconstructions of a hyperspectral datacube across 64 spectral bands and recovering the full resolution of the underlying camera sensor. Our end-to-end design shows clear improvements in comparison to systems designed heuristically, and in future work we will manufacture and assemble the device to image microbeads that fluoresce across a wide range of wavelengths. Project Poster.
–
Undergraduate Researcher: Afnan Khawaja
Intended Major: Computer Science
Home Institution: Napa Valley College
Research Project: Investigating the Effectiveness of Robust Estimations in Multiple Dimension
Faculty Advisor: Prof. Jiantao Jao
Mentor: Banghua Zhu
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: Robust statistics focuses on learning in the presence of outliers in the dataset and producing statistical methods that are not unduly affected by outliers. There has been great progress in developing computationally efficient algorithms for robust estimation from high-dimensional corrupted data, including tasks on robust mean estimation, robust covariance estimation, robust linear regression etc. We are addressing the problem of natural outliers and data poisoning attacks in machine learning in which a small variation of data can contaminate the whole result. In this project, we will be investigating the effectiveness of different robust estimators by implementing and comparing the well-established high-dimensional robust estimators, including Tukey median and filtering. We will compare these estimators on both synthetic and real datasets, for example biological data and machine learning datasets, to provide a thorough analysis and empirical validation for the theory. These robust estimators will help us accurately calculate the mean, variance, and parameters of a set of data even if it is corrupted. Project Poster.
–
Undergraduate Researcher: Amy Matthews
Intended Major: Electrical Engineering and Computer Science
Home Institution: San Diego Miramar College
Research Project: Dark Patterns of Online Targeted Diet and Weight-loss Ads
Faculty Advisor: Prof. Niloufar Salehi
Mentor: Liza Gak
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: Online targeted advertising can potentially cause emotional harm due to ad delivery’s hyper-individualized, seemingly invasive nature. Particularly in the case of diet ads, the pressure for women to lose weight and strive for an unattainable body shape in an inadequate amount of time leads to body dysmorphia and various eating disorders. The goal of the project aims to uncover the predatory nature of online advertising utilizing data analysis. As it stands, the term diet ad is ambiguous and pushed on users who may not actively be trying to lose weight, in turn contributing to low-self esteem and acts as a sort of “slow-violence”. Employing Python and data-analysis tools, such as optical character recognition scanner, we will examine the trends and patterns in a dataset of commercials to create a proper definition of the otherwise ambiguous “fad diet-ads”. We can then further streamline and filter these deteriorating commercials from reaching the general population through technology in the form of ad blockers in the future. ProjectPoster
–
Undergraduate Researcher: Kavan Mehrizi
Intended Major: Electrical Engineering and Computer Science
Home Institution: Diablo Valley College
Research Project: Quadrupedal Robotic Guide Dog with Vocal Human-Robot Interaction
Faculty Advisor: Prof. Koushil Sreenath
Mentor: Zhongyu Li
UC Berkeley Department: Department of Mechanical Engineering
Project Abstract: Guide dogs play a critical role in the lives of many, however training them is a time- and labor-intensive process. We are developing a method to allow an autonomous robot to physically guide humans using direct human-robot communication. The proposed algorithm will be deployed on a Unitree A1 quadrupedal robot and will autonomously navigate the person to their destination while communicating with the person using a speech interface compatible with the robot. This speech interface utilizes cloud based services such as Amazon Polly and Google Cloud to serve as the text-to-speech and speech-to-text engines. Project Poster.
–
Undergraduate Researcher: Aneri Mody
Intended Major: Computer Science
Home Institution: Long Beach City College
Research Project: Engaging Families in Co-Designing the Algorithmic Student Assignment Policy in San Francisco Unified School District
Faculty Advisor: Prof. Niloufar Salehi
Mentor: Tonya Nguyen
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: San Francisco Unified School District’s (SFUSD) student assignment system is a socio-technical algorithm designed for fair, diverse, and equitable placements. However, this system faced real-world issues in its deployment, and the district decided to do a complete redesign of the student assignment system since it failed to uphold its intended values: the prior algorithm was working on a few best-case scenarios from modelling assumptions of the ideal world and overlooking the average and the worst-case scenarios like the practical challenges of life. It is essential to realign the goals of this algorithm with the interests of the community in mind. We are working in collaboration with SFUSD in a community engagement process, which will guide the redesign of SFUSD’s algorithmic student assignment system. We aim to engage families as the stakeholders in the participatory design process. For the direct inclusion of families, we conducted community engagement interviews to understand their goals and choices. We further performed qualitative analysis of these interviews using the grounded theory, which informed future designs of the algorithmic system. Our findings revealed the gaps between the algorithmic assignment system and the lived experiences of marginalized parents, which the algorithm could not account for. Project Poster.
–
Undergraduate Researcher: Darren Munoz
Intended Major: Mechanical Engineering
Home Institution: Allan Hancock College
Research Project: Towards High Performance Field-Effect Transistors with Bottom-Up Synthesized Graphene Nanoribbons
Faculty Advisor: Prof. Jeffrey Bokor
Mentor: Dr. Zafer Mutlu
UC Berkeley Department: Department of Mechanical Engineering
Project Abstract: Graphene nanoribbons (GNRs) can exhibit a uniform electronic band gap and emergent electronic properties that are promising for nano-electronic devices, such as field effect transistors (FETs), when synthesized with atomic precision. Bottom-up, on-surface synthesis approaches can provide the necessary precision to access these desirable properties, but the potential of these bottom-up synthesized GNRs for electronic devices has remained unexplored. Herein, we study the electrical properties of the FETs based on bottom-up synthesized nine-atom wide armchair GNRs (9-AGNRs) with varying channel lengths (30-65 nm), a local back gate geometry of 5.5 nm HfO2 gate dielectric, and Palladium contacts. The 9-AGNR FETs exhibit high ON-state current (up to 13 µA) and excellent switching performance (up to ION /IOF F = 105) with a high working device yield (> 80%). The current-voltage characteristics indicate a strong correlation between the channel lengths and key performance metrics, such as subthreshold swing (SS), ON-state current (ION ), and ON-OFF current ratio (ION /IOF F ). This research provides important insights into the design of future high-performance GNR-based digital logic devices. Project Poster.
Undergraduate Researcher: Emanuel Navarro-Ortiz
Intended Major: Electrical Engineering and Computer Science
Home Institution: Cabrillo College
Research Project: Jointly Inferring Human Irrationality and Intent
Faculty Advisor: Prof. Anca Dragan
Mentor: Dr. Daniel Brown
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: Abstract—Inverse reinforcement learning enables robots to learn new tasks from human demonstrations by learning a reward function that explains the human’s intent. Typically, inverse reinforcement learning algorithms assume that the human demonstrator is rational; however, this is rarely true in practice. Humans often make biased decisions or provide noisy irrational demonstrations that can lead to sub-optimal learning experiences for robots. Treating these demonstrations as near optimal may lead the robot to misinterpret the human’s intent, resulting in unintended behavior. This concern is paramount with the projected increase of artificial intelligence integration and human-robot interactions. To address this problem, we proposed joint Bayesian inference over human rationality models and human reward functions. We studied joint inference over a incoherence bias. Our experimental results demonstrate that our novel joint Bayesian inference approach produces a better model of human intent, as it is intended to detect if the human demonstrator is systematically biased or irrational and compensate for the human’s irrationality when inferring the human’s true reward function. Thus, our research takes steps towards safer training of robots and more reliable human-robot interactions. Project Poster.
Undergraduate Researcher: Edrees Saied
Intended Major: Electrical Engineering and Computer Science
Home Institution: College of Alameda
Research Project: News Coverage Diversity and Source Specialty
Faculty Advisor: Prof. John Canny
Mentor: Phillipe Laban
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: Different news channels have their styles and perspectives embedded in their coverage of local and global events. The distinctions in news coverage from different channels on a particular event can be analyzed through the unique themes and focus highlighted in their published articles. My project goal is to automate the process of tagging various themes to news articles published by different news sources to help news readers get informed on source specialization for specific topics. The method we propose uses a Natural Language Processing (NLP) pipeline to compare articles from different sources side to side and discover unique coverage in each source. In addition, the project aims to build a recommendation system that can direct users to a specific news article based on its general theme, the occurrence, and the user’s interests. Project Poster.
Undergraduate Researcher: Rocco Scinto
Intended Major: Computer Engineering
Home Institution: Santa Rosa Junior College
Research Project: Simulation and Optimization of Graphene Nanoribbon Transistors
Faculty Advisor: Prof. Jeffrey Bokor
Mentor: Dr. Yuxuan Cosmi Lin
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: Using foundational mathematical models created by Landauer-Buttiker and Wentzel–Kramers–Brillouin (WKB) we ¨ can measure the performance of graphene nanoribbon field effect transistors (GNRFETs) through simulation. Using these previous models with some modern adaptations to calculate device per formance given a set of user defined parameters, we analyzed the effects of different materials and fabrication dimensions. In the process of running these simulations GNRFETs can be optimized with the most beneficial attributes that lead to improvements in carrier mobility, on-current, etc. This process of using iterative simulations saves time and material which will impact construction of GNRFETs for future use in logic-circuit applications. Project Poster.
Undergraduate Researcher: Fayaz Shaik
Intended Major: Electrical Engineering and Computer Sciences
Home Institution: Ohlone College
Research Project: User-Centered Data Population of Knowledge Graphs
Faculty Advisor: Prof. Sarah Chasins
Mentor: Justin Lubin
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: A knowledge graph is a data structure that describes structured relationships between entities. Knowledge graphs are widely used in artificial intelligence systems, but sometimes require data that can only be found on the internet in a semi-structured format such as a bulleted list. Unfortunately, semi-structured data can be difficult to parse, requiring users to write custom web-scraping programs for each web page to extract the necessary data. After presenting several ideas to domain experts for their opinions in a mock formative study, we built a prototype that prompts the user to highlight relevant information in a webpage, then uses these highlights as input-output examples for a custom program synthesizer that automatically generates web scraping scripts. These scripts are customized to each website and parse the semi structured data into a tabular format easily transferable to a knowledge graph. Such a tool heightens the level of automation accessible to domain experts so that they may better leverage the vast amount of data available online for use in artificial intelligence systems. Project Poster.
Undergraduate Researcher: Cara Yee
Intended Major: Molecular and Microbiology
Home Institution: Cosumnes River College
Research Project: Monitoring Microbial Community Diversity with Cytometric Fingerprinting
Faculty Advisor: Dr. Romy Chakraborty
Mentor: Dr. Shwetha Acharya
Lawrence Berkeley National Laboratory
Project Abstract: Identifying and monitoring microbial diversity of microbial communities within bioreactors are critical for their performance efficiency. Most commonly, 16S rRNA amplicon sequencing is used to understand microbial community structure and composition, but this process is labor and time intensive. With improvements to flow cytometry (FCM) data preprocessing and analysis packages, FCM has become a more efficient method for providing insight into microbial diversity and monitoring microbial communities. Using FCM data of two synthetic bacterial consortia grown in three different media for ten generations, we intend to examine currently available R-based pre-processing packages and cytometric fingerprinting analysis packages Phenoflow to develop a pipeline best suited to monitor these microbial communities. Using a workflow with Phenoflow we were able to perform diversity analyses, using Hill’s diversity index and NMDS and gain insight into the evolution of communities over time. Project Poster.
Undergraduate Researcher: Isaac Villa-Loyer
Intended Major: Electrical Engineering and Computer Science
Home Institution: City College of San Francisco
Research Project: Optimization of Oscillatory Back-End-of-Line Nanoelectromechanical Switches
Faculty Advisor: Prof. Tsu-Jae King Liu
Mentor: Lars Tatum
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: The proliferation of Big Data and AI have greatly increased computing resource needs. Today’s Integrated Circuit (IC) technology is limited by both manufacturing and physical constraints, and has not kept up with the requirements for these compute-intensive applications. It has recently been shown that systems of coupled oscillators will naturally find an optimal solution of the Ising problem, which can be mapped to solve nondeterministic polynomial (NP)-hard problems significantly faster and more efficiently than conventional computing frame works. A propitious approach to implementing these oscillators is the use of monolithically integrated micro/nanoelectromechanical (M/NEM) switches. These switches allow a theoretical zero off state leakage current, and operate at far lower voltages than today’s state-of-the-art electrical transistor technology. Under the proper conditions, these switches will experience stable oscillation with very low power consumption per cycle. NEM switches implemented in IC Back-End-of-Line (BEOL) metallization can be configured to create dense oscillator arrays monolithically integrated with leading-edge IC technology, enabling a single chip solution. In this work a novel oscillating BEOL NEM switch is designed in Coventor MEMS+ for use in an Ising machine and simulated in Cadence Spectre. Simulation results show that BEOL NEM switches are a promising approach to enable NP hard hardware accelerator cores. Project Poster.