XSEDE Scholars Program

2016 XSP Scholars

10 scholars chosen to engage with XSEDE!

Ten STEM students have been named to the XSEDE Scholars program, an outreach and educational effort that reveals the full reach and potential of XSEDE's resources and services to undergraduates and graduates. This 2016 summer cohort is the sixth generation of this program, which aims to engage with underrepresented groups in computational science.

As part of the XSEDE Scholars program, the scholars attended the Blue Waters Petascale Institute, participated at the XSEDE16 conference with four poster presentations and  leading the XSEDE Scholar Lightning Talks  (15 talks) and one invited talk. All the XSP scholars will participate in at least six online technical training and mentoring webinars open to XSEDE Scholars and the general public, will network with leaders in the XSEDE research community and will learn about research, internships and other career opportunities via an online community throughout the year.

Top row: Quentarius Moore, Janette Garcia, Jason Regina, Michael Taft, Juan Castro-Garcia, Jorge Alarcon Ochoa, Wanda Moses. Bottom row: Olivia Irving, Sidafa Conde, Efrain Vargas Ramos.

Seven names to receive $5,000 stipend who will be conducting year-long internships under the guidance of HPC mentors. The students are expected to present their research at XSEDE16/HPC event/webinar, and write a paper/final report.

Christina Davis

 (Vanderbilt University, Ph.D Student, Astrophysics)

Project Title: Understanding Galaxy Transformation from Flyby Encounters

Research Advisor: Kelly Holley-Bockelmann, Department of Physics and Astronomy, Vanderbilt

Galaxy flybys are transient events where two halos interpenetrate and later detach forever. Although these encounters are surprisingly common, their dynamical effects have been largely ignored. By examin- ing flybys within a cosmological context, we get a better picture of how the encounters shape the nearby galaxies. The Illustris Simulation is a high-resolution hydrodynamical simulation of a (106.5Mpc)3 vol- ume. With 136 snapshots over 13.8 billion years of evolution, the Illustris Simulation provides a large number of halos with a wide variety of interaction histories. The simulation includes physics of star formation and stellar feedback, supermassive black hole growth, AGN and supernova feedback, and gas cooling. We will create a publicly available database of all halos and their interactions that includes flybys. In addition, we will quantify physical changes on both the host and intruder galaxies in a flyby event. Being a state of the art cosmological simulation, the amount of data demands high performance computing techniques to be implemented. Illustris 1 snapshot data is a total of 204 TB with over 4 million galaxies being identified in the final snapshot. In order to identify the flybys in the simulation, we must keep track of every halo and all of it's interactions throughout time. By using parallelization and vectorization, the data can be quickly sorted and optimized for scientific usage. 

Daniel Dummett Torres

(University of Illinois at Urbana Champaign, PhD Student, Chemistry)

Project Title: DFT Study of Doping and Cation Exchange in CdSe Nanocrystals

Research Advisor: Prashant K. Jain, Department of Chemistry, University of Illinois at Urbana Champaign

My project is aimed at achieving advances in the control of material properties at the nanoscale such as chemical reactivity. Defects in nanoscale solids play an important role in their electronic transport properties, their optical properties, and their chemical reactivity. However, an exact mechanistic/electronic explanation for how defects tune the properties of these materials is largely missing, limiting present understanding. This motivates my theoretical investigation with density functional theory (DFT) in which I sample the potential energy surface made up by the intermediate structures that arise throughout transformations of CdSe and Cu2Se. These electronic structure calculations will help elucidate the origins of cooperativity that has been experimentally observed by our group in the cation exchange reaction of Cd2+ for Cu+ in CdSe and will also describe the cation migration pathways that characterize superionicity in Cu 2Se among other insights that we will gain into the rich properties of these materials.

Mario Bencomo

(Rice University; PhD Student; Computational and Applied Mathematics)

Project Title: Multipole Point Representation and Full Waveform Inversion for (Anisotropic) Seismic Sources

Research Advisor: William Symes, Department of Computational and Applied Mathematics, Rice University

The seismic inversion problem consists of determining information about subsurface geological structures given seismic data. This inversion is posed as a PDE constrained optimization problem where one seeks to find optimal medium (and source) parameters that minimize misfit between observed and predicted data. Accurate mathematical representation and estimation of sources is essential for the recoverability of medium parameters and the main focus of this work. I propose a representation that takes into account anisotropy based on approximating sources via a truncated series of multipole point-sources (MPS). Moreover, my proposed work focuses on two main aspects related to the source estimation subproblem: preconditioning and regularization. Incorporating source estimation in the inversion formulation will undoubtedly come at higher computational cost. It is therefore vital to exploit the parallelizability of core computations for reasonably sized problems: mainly, parallelization over sources for multiple source simulations, and domain decomposition for core finite difference solves.

My proposed work consists of implementing MPS parameter estimation algorithms as part of the seismic inversion software package IWave, developed by the Rice Inversion Project (TRIP) group at Rice University. Incorporating source parameters estimation in the inversion formulation will undoubtedly come at higher computational cost. It is therefore vital for the sake of computational feasibility to exploit the parallelizability of core computations, primarily that of finite difference solves, for reasonably sized problems. Two main modes of parallelization are considered: parallelization over sources for multiple source simulations, and domain decomposition for the core finite difference solvers. 

Maurice Fabien

 (Rice University, PhD Student, Computational and Applied Mathematics)

Project Title: Multilevel methods for Discontinuous Galerkin Discretizations

Research Advisors: Matthew Knepley and Béatrice Rivière, Department of Computational and Applied Mathematics, Rice University

The focus of my research is the application, theory, and computer implementation of numerical methods for partial differential equations that model multiphase flows. More specifically I will investigate the effectiveness of discontinous Galerkin (DG) geometric multigrid (GMG) discretization and solver for PDEs arising from two phase flow and magma dynamics. In addition, the mathematical and numerical analysis of DG–GMG will be explored. Finally, my computer implementation will utilize the many–core architecture that accelerators offer.

The DG discretization is an ideal pairing for GMG, because of its ability to localize computations, handle complex domains, conservation properties, as well as the ease of adapting to high order. High order DG discretizations ensure that accurate simulations are generated, and, also will keep the multi-core and many- core architectures arithmetically saturated.

The DG discretization of PDEs gives rise to large sparse linear systems. Solving these large linear systems efficiently poses a significant challenge. For the past 40 years, there has been no method that is compet itive with geometric multrigrid for solving elliptic PDEs. The geometric multigrid method is well known to offer optimal complexity, that is, it only requires O(N) floating point operations to reduce the error to discretization level. Heterogeneous computing is required if larger simulations are to be generated, or if we want simulations to run faster. Exploring GMG in a heterogeneous computing needs careful consideration because it is inherently a multiplicative algorithm. 

Jovian Lazare

(Clark Atlanta University, Master Student, Chemistry)

Project Title: Computational Study of the Binding of Single and Double Methane with Polycyclic Aromatic Hydrocarbons

Research Advisor: Dinadayalane Tandabany, Department of Chemistry, Clark Atlanta University

Methane storage and sensors are of high importance to world-wide and in particular, Environmental Protection Agency (EPA) because methane is a green-house gas but it is also a source of energy. About 90% of methane gas is produced during formation of coal and capturing for storage after its release due to mining and erosion is therefore a priority. The Environmental Defense Fund called the leak at the Aliso Canyon Gas Storage Field is "one of the largest U.S. natural gas leaks ever recorded". Thousands of families have been forced to move from their homes, with some people complaining of headaches and nosebleeds. According to the EPA, methane traps heat so well therefore its impact on climate change is 25 times greater than carbon dioxide. Most U.S. methane emissions (29 percent) come from the production, processing, storage and distribution of natural gas, according to the EPA, followed by agriculture (26 percent) and landfills (18 percent).

It is important to design materials for methane storage and better sensors. Carbon based materials in particular graphene or graphane could be a potential medium for methane storage and/or sensor. In order to understand the viability of graphene based materials, we propose to explore the single and double methane binding with smaller polycyclic aromatic hydrocarbons (benzene, pyrene, and coronene) that are building blocks for graphene. In our computational study, a systematic investigation of binding of one and two methane with the above-mentioned systems will be carried out by considering several possibilities. All our computations will be performed using the density functional theory (DFT) calculations with double and triple-z basis sets. Since the M06-2X functional is reliable for interactions dominating dispersion forces, we will use that functional. The scholar (Mr. Lazare) has recently started the project with mentor Dr. Tandabany. He is currently using NWChem program package. He has access to the PI, Dr. Tandabany's XSEDE campus champion allocation. In the past three (3) months, he has gained experience in running calculations using Stampede and Gordon. Since several calculations should be performed with various possibilities of binding modes of methane with three systems, XSEDE resources will be very useful for the successful completion of this project. Some of the possibilities of different binding modes of methane with carbon systems are (1) three C-H…p, two C-H…p, one C-H…p, both methane at the same side for pyrene and coronene and both methane at opposite sides of these two systems. After completion of this project, we intend to explore the binding of single and double methane with graphene. The results and the knowledge gained from the smaller polycyclic aromatic systems will be a reference guide as we move to the large graphene system. Our computational study will assist experimentalists in developing new carbon-based materials for methane storage or sensor.  

Quentarius Moore

(Jackson State University, Master Student, Computational Quantum Chemistry)

Project Title: Computational Investigation of the Structural & Electronic Properties of FeS2 Nanoparticle and Consequence on the Formation of Non-Pyrite Sulfide Phases

Research Advisor: Jerzy Leszczynski, Director of the Interdisciplinary Center for Nanotoxicity, Department of Chemistry, Jackson State University

HPC Resources: Stampede

The team of Quentarius Moore and Dr. Jerzy Leszczynski seek to discover knowledge about the binding interactions of two systems, a single-walled carbon nanotube (SWCNT) and  5,10,15,20-tetrakis(pentafluorophenyl)porphyrin (TPPF 20), to exploit the electronic and mechanical properties of SWCNTs. TPPF 20 ,itself a very aromatic material, is a part of a class of large aromatic dyes called porphyrinoids, which have applications in fields such as electronic devices, photovoltaic cells, medicine, and optical sensors.. The team hypothesizes that the two systems of interest will undergo this same non-covalent interaction and has a goal of studying those interactions in order to assess the potential for this novel material to be implemented for photovoltaic, bio-sensing and pollutant absorbent applications. This study will be carried out by utilizing quantum chemical calculations to model and predict the electronic structure and chemical properties of these systems as well as correct for basis set superposition error. Density functional theory (DFT) will be applied to reveal the binding interactions at different sites with the tool, Gaussian09, being used for executing such calculations. The XSEDE Scholars Program facilitates in the access to resources such as Stampede, which is a great improvement over current resource, as the project is of very large systems. More importantly, the program adds training of HPC, exposure to more fields in computational science, and the social aspect needed to take the research project to advanced levels.

Ronald Gonzales

(Idaho State University, Master Student, Mathematics)

Project Title: Modeling Cloud Formation with the Discontinuous Petrov-Galerkin Method

Research Advisor: Yuriy Gryazin, Department of Mathematics, Idaho State University

The goal of this research project is to develop and assess the performance of a new numerical technique that will increase the order of approximation and improve the accuracy of modeling the cloud development process. This project will begin the second week of May 2016 and will conclude mid-August 2016. A high level of mathematics and computer science will be applied in order to achieve this goal.

The learning goals are to obtain the ability to develop advanced numerical processes and gain the proper programming skills to be able to apply these processes to real world problems. As a student, the applicable experience I have is a two semester course in introductory numerical analysis and two introductory courses in C++. My tasks will be to assist in the proof of convergence of the new numerical technique and developing the proper parallelized algorithm.

The advanced numerical methods that this research project aims to develop and assess the performance of are very computationally intensive. This will require the use of C++ and MPI to efficiently parallelize the numerical methods. This will make use of multiple nodes on the Blue Waters supercomputer.

Bernard Scott Jr.

(Georgia State University, Master Student, Biophysical Chemistry)

Project Title: Dynamical Studies on Liver Receptor homolog-1 bound to small molecule effectors

Research Advisor: Ivaylo Ivanov, Department of Microbiology and Immunology, Georgia State University

The nuclear receptor liver receptor homolog-1 (LRH-1), is unique in its use of various phospholipids (PLs) as signaling molecules. Through interaction with PLs and co-regulating peptides, LRH-1 regulates several processes including cell-cycle progression, steroid synthesis, as well as lipid and glucose homeostasis. Specific activators such as dilauroylphosphatidylcholine (DLPC) have been shown to lower serum lipid levels and improve glucose tolerance in diabetic rat models1, making LRH-1 a prime pharmaceutical target. Understanding more about the structural and dynamical details of its regulation at the molecular level can enhance the success of such efforts.

Through molecular dynamics studies (using NAMD) have recently reported a previously unidentified allosteric network running through the core of the LRH-1 ligand binding domain (LBD) The network forms a communication tether between the mouth of the ligand binding pocket and the AF-H/AF2 coregulator-binding region of the LRH-1 LBD2. This tether allows for LRH-1 to coordinate coregulator binding with ligand characteristic (i.e. agonist or antagonist). Furthermore, using principal component analysis (PCA), we have determined that LRH-1's displacement along two specific principal components is determined by the species of bound ligand and coregulator. Taken together, we view this allosteric tether and the accompanied conformational characteristics as dynamical fingerprints that can be used to identify whether a ligand will act in an agonistic or antagonistic role when bound to LRH-1.

We have recently been provided an x-ray crystal structure of LRH-1 bound to a synthetic agonist, Ent-2. I intend to perform analyses on this system along with several others (bound to different compounds) in accord to our earlier work with LRH-1/phospholipid regulation. If the mechanism of activation of LRH-1 by our small molecule mimics that of natural substrates, then strong communication between the ligand binding pocket and AF-H/AF2 is expected. Similarly, if the synthetic agonist impacts LRH-1 in a manner analogous to that of natural agonists, LRH-1 can be expected to reside in a conformational basin consistent with co-activator binding. The principle methods applied to this problem will include classical molecular dynamics with PMEMD followed by dynamic network analysis, and principle component analysis (PCA). The molecular dynamics simulations for these systems will likely be run on Comet. 

Luis Niebla Rios

 (Arizona State University, Undergraduate Student, Earth and Space Exploration)

Project Title: Chemodynamical Cosmological Simulations with RAMSES and KROME

Research Advisor: Evan Scannapieco, School of Earth & Space Exploration, Arizona State University,

This project will provide a description of how the abundances evolve with time on galaxy scales at high redshift. Tables of these abundances as function of basic parameters, such as surface density of gas or SFR, on different time scales can then be implemented in models of the spectra, making valuable predictions for future observations large radio telescopes such as ALMA, JVLA, GBT and LMT, thereby increasing our knowledge of the ISM, in particlar, at the early stages of the Universe. 

Natalie Zimmer

 (Louisiana State University, Undergraduate Student, Physics & Astronomy)

Project Title:  Using GEANT4 for Transition Radiation Simulations to Reduce the Proton-Kaon Contamination Level

Research Advisor: Michael Cherry, Department of Physics and Astronomy,  Louisiana State University.

An upcoming project at CERN will observe high energy (multi-TeV) inelastic proton-proton and proton-heavy ion collisions in the very-forward direction- the region where particles are emitted at very small angles with respect to the incoming particle. In order to identify the protons, pions, and kaons produced in the collisions, a transition radiation detector array must be used. Transition radiation occurs when a charged particle traverses the interface between two media, producing photons in the process. By looking at the energy spectra produced in the transiton radiaiton detectors, protons, pions, and kaons can be distinguished from one another. For this experiment, the misidentification of one species from another must be on the order of 10-4. Preliminary calculations have estimated that pions can be identified reasonably well, but protons and kaons can only be separated at a level of ~10-2. In order to improve this number, the number of x-ray photons produced in the transition radiation of the setup must be increased. The best way to do this is to use GEANT4, a C++ based toolkit designed for building particle simulations, to build simulations of this setup to find the optimal parameters that will maximize the number of x-ray photons produced in the space of 8m allotted for the transition radiation detectors. Through the use of simulations, we will find the ideal transition radiation detector setup to most effectively identify protons, pions, and kaons, which will eventually be built into an experimental setup.