Community College Participants and Projects: 2020
Despite the ongoing COVID-19 pandemic and campus closure, the Center for Energy Efficient Electronics Science hosted eight students for an online summer research internship. The program was the only summer research internship at UC Berkeley that was not cancelled in 2020, and participants shared that they had a very positive experience. All eight participants have been invited to present their research at the 2020 SACNAS Diversity in STEM conference, and a subset have been invited to continue their research work into the 2020-2021 academic year.
The 2020 TTE Research Symposium can be viewed here.
Undergraduate Researcher: Austin Culp
Intended Major: Chemical Engineering
Home Institution: Los Angeles Valley College
Research Project: Sub-nanometer Control of Low-Dimensional Tantalum Tellurides via Encapsulation within Carbon Nanotubes
Faculty Advisor: Prof. Alex Zettl
Mentor: Scott Stonemeyer
UC Berkeley Department: Department of Physics
Project Abstract: The study of certain transition metal chalcogenides has been hindered by their instability in air or a difficulty in isolating a pure form of them. Synthesizing tantalum telluride crystals within carbon nanotubes would allow for both protection of the crystals from air and atomic control of the crystal growth, creating either 1-D nanoribbons of Tantalum Ditelluride, TaTe2, or 1-D trigonal prismatic chains of Tantalum Tritelluride, TaTe3. These two crystal forms can be preferentially selected by altering the temperature during the chemical vapor transport synthesis process. Scanning transmission electron microscopy, accompanied with a Java-based image processing program, reveals striking differences between the TaTe2 nanoribbons and the TaTe3 chains. Achieving such precise control over the form of crystal obtained, either nanoribbons or chains, at the atomic scale allows for full exploration of how the structural and electronic properties of these low-dimensional structures could be altered by encapsulation. Project Poster
Undergraduate Researcher: Cameron (Czara) Baker
Intended Major: Chemistry
Home Institution: Fullerton College
Research Project: Investigating Sequence Features of eIF3 and eIF4A Target mRNAs
Faculty Advisor: Prof. Jamie Cate
Mentor: Angelica Gonzalez-Sanchez
UC Berkeley Department: Department of Chemistry
Project Abstract: Translation regulation is critical for maintaining cell homeostasis and its misregulation leads to diseases such as cancer. Translation initiation is highly regulated and requires several eukaryotic initiation factors (eIFs). eIF3 has been found to serve as both a scaffold in preinitiation complexes and a regulator of translation for certain mRNAs. eIF4A unwinds mRNA in preparation for recruitment to initiation complexes. Both eIF3 and eIF4A have been found to bind to and regulate translation of the JUN mRNA. This suggests a new mechanism for translation regulation that can also act on other mRNAs in the cell. From this analysis, we were able to identify common sequence features of eIF3 and eIF4A target mRNAs. As a whole, this study broadens the understanding of the mechanisms of translation regulation mediated by eIF3 and eIF4A, which serves as a blueprint for targeting disease. Project Poster
Undergraduate Researcher: Jack Wyatt Jebef
Intended Major: Electrical Engineering and Computer Science (EECS)
Home Institution: Santa Barbara City College
Research Project: Understanding the Raman Spectra of Graphene Nanoribbons for Device Fabrication
Faculty Advisor: Prof. Jeffrey Bokor
Mentor: Dr. Zafer Mutlu
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: Graphene nanoribbons (GNRs) are quasi-one-dimensional carbon-based semiconductors that possess a tunable bandgap contingent on ribbon width and edge topology and are therefore a promising alternative for silicon channels in future transistors. The bottom-up synthesis of GNRs provides ultimate control over ribbon design and thus their electronic properties. Scanning-tunneling microscopy (STM) and Raman spectroscopy are the two standard techniques used for characterizing bottom-up synthesized GNRs. Although STM is an effective method to obtain atomic-scale information, its scope is localized and is limited to GNR samples grown on metallic substrates. Contrastingly, Raman spectroscopy is a fast and non-invasive analytical approach that provides complementary information about GNRs on both metallic and insulating substrates at a macroscopic scale. However, most contemporary studies only use Raman spectroscopy to identify GNR type. Herein, we investigate the effect of growth substrates, doping, and the transfer process on the Raman features of seven-atom wide armchair GNRs (7-AGNRs). The 7-AGNRs are grown on gold and copper in a UHV chamber and are moved to insulating substrates via a well-established wet-transfer process for the fabrication of field-effect transistors (FETs). The GNRs/gold samples are doped with chemicals used in the wet-transfer procedure to understand the effect of doping on the electrical and electronic properties of GNRs. This study represents significant progress towards our main goal: the large-scale integration of GNRs into high-performance transistors. Project Poster
Undergraduate Researcher: Menucha (Mimi) Winchell
Intended Major: Computer Engineering
Home Institution: Los Angeles Valley College
Research Project: Current-Induced Switching in Antiferromagnetic Multilayer System
Faculty Advisor: Prof. Jeffrey Bokor
Mentor: Dr. Sucheta Mondal
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: This research study utilizes ‘Object Oriented Micromagnetic Framework’ (OOMMF), a micromagnetic simulator, to layer nanodots consisting of antiferromagnetic, ferromagnetic, and nonmagnetic layers. Static magnetic properties are observed through varying the Zeeman field surrounding the multilayer system, thus numerically describing antiferromagnetic coupling within the system. The manipulation of magnetization within the system is studied by injecting an electrical current through the nonmagnetic layer. The flow of charge current results in the generation of pure spin currents, exerting spin torque on the interfaces. This torque influences the switching of the ferromagnet’s magnetization while resulting in ultrafast dynamics within the antiferromagnetic sublattices. This experiment aims to observe the conditions in which the Ne’el vector switches within the multilayer framework. Project Poster
Undergraduate Researcher: Musaiel (Moose) Gebremariam
Intended Major: Mechanical Engineering
Home Institution: Santa Rosa Junior College
Research Project: Design and Optimization of LiDAR Beam Scanners
Faculty Advisor: Prof. Ming C. Wu
Mentor: Xiaosheng Zhang
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: Light detection and ranging (LiDAR) technologies are being used in the industry to generate 3D images by emitting a laser to measure a scene of interest point-by-point. In commercial LiDAR products, mechanical beam scanners are currently used, whereas solid-state beam scanners that are constructed with silicon photonics and micro-electromechanical (MEMS) technologies are under intensive research and development. In this research, we determined how to improve the efficiency of mechanical and solid-state scanners and understood the correspondence and trade-off between each design. We analyzed the properties of the galvanometer mirror, such as the mirror size and scanning speed, and investigated the optomechanical package design of a solid-state focal plane switch array beam scanner. Project Poster
Undergraduate Researcher: Nguyen Thanh Vi (Vi) Tran
Intended Major: Computer Science
Home Institution: Orange Coast College
Research Project: Unsupervised Deep Learning on Lensless Imaging
Faculty Advisor: Prof. Laura Waller
Mentors: Grace Kuo and Kristina Monakhova
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: Lensless camera systems replace the lens with a light-weight diffuser to encode light information onto the camera sensor, which allows the imaging system to be compact and cheap. Traditionally, the scene is recovered from the multiplexed measurement by solving an inverse problem. However, the reconstructed image often suffers from model mismatch and image artifacts. In this research, we explore the unsupervised deep learning approach, which trains neural networks on the reconstruction task using single image measurement, without the need for ground truth labels or large dataset. Since ground truth data is hard to acquire for many mask-based imagers, this approach has the potential to produce high-quality reconstructions in the absence of ground truth training images. Project Poster
Undergraduate Researcher: Rosario Martinez
Intended Major: Computer Science
Home Institution: Long Beach City College
Research Project: An Automatic Data Processing Pipeline to Reconstruct Coherent 3D Image Volumes from Cardiac Cine MRI Data
Faculty Advisor: Prof. Shawn Shadden
Mentor: Fanwei Kong
UC Berkeley Department: Department of Mechanical Engineering
Abstract: Computational fluid dynamics (CFD) simulations of left ventricle (LV) flow combined with patient medical imaging data may facilitate a better understanding of cardiovascular diseases so patients can have improved diagnosis and treatments. These simulations usually require geometric models of the heart constructed from patient-specific cardiac image scans. The model reconstruction pipeline includes segmentation and surface reconstruction in 3D and is traditionally a time-consuming process when done manually. Deep learning-based methods can speed up this process. Cine magnetic resonance imaging (MRI) is a conventional image modality that is used for cardiac function evaluation which produces several time-series 2D image slices from different locations of the heart. These images must be reconstructed into a 3D image volume before a deep learning model can utilize it. Thus, we are creating a data processing pipeline that can automatically take in all 2D image slices for a patient and align them to produce a 3D image volume. We propose to apply inter-slice registration to reduce slice misalignments due to the 2D acquisition of cine MRI. The pipeline is validated on a large public dataset, Data Science Bowl Cardiac Challenge Data. By converting 2D image slices into coherent 3D volumes, the proposed pipeline facilitates the use of cardiac cine MRI scans by deep-learning-based methods for constructing CFD-ready LV models and may ultimately enhance the treatment of cardiovascular disease. Project Poster
Undergraduate Researcher: Samantha Childers
Intended Major: Computer Science
Home Institution: Citrus College
Research Project: The Computational Apprentice: Exploring a New Approach for AI Assistance
Faculty Advisor: Prof. Björn Hartmann
Mentor: J.D. Zamfirescu-Pereira
UC Berkeley Department: Department of Electrical Engineering and Computer Sciences
Project Abstract: Future intelligent systems often envision AI agents that aid users through mixed-initiative interaction. To make such agents effective, an understanding of the ways agents can learn from users’ actions and explanations is necessary. To help develop this understanding, this work describes and explores the interactions between an AI apprentice and a human mentor through a series of studies. Prior work in Human Computer Interaction (HCI) typically conceptualizes AI agents as “experts” in narrow domains that observe event streams and offer assistance when they have high confidence about the user’s goal and that the agent’s assistance will be welcome. This study explores how a computational apprentice can engage in apprentice-style questioning and elicit a deeper and better-connected model of the user’s tasks, actions, and goals. Beginning with insights collected from recorded interactions between two humans taking on the role of mentor and apprentice in a programming task, and culminating in the evaluation of an a human-in-the-loop simulation of an AI apprentice, these experiments provide insight into how apprentice agents should be designed, and how their design influences how human mentors use them and feel towards them. We found that Participants were willing to answer, via voice, even an Ai agent, and expected the agent to have a basic understanding of the environment and answer questions about it. Participants did not find the actual goal of the agent, to comment participants’ code, particularly helpful, and many participants wanted more control over when comments were placed where, and found the AI’s prompting neither particularly useful nor intrusive. Project Poster