Student Posters Repository

Mosquito-borne pathogens are responsible for a massive global burden of disease: they cause approximately 700,000 deaths annually. Each year more than half of this total mortality is attributed to the protozoan parasites that transmit malaria. Current control measures range from anti-parasitic drugs to bed nets, insecticides, and the destruction of vector habitats.[1]  But while such methods have met with some success to date, drug and insecticide resistance is rising. Further, the changing precipitation patterns and temperatures brought by global warming will affect the incidence of malaria. This makes eradication of the disease unlikely if existing approaches are not augmented.

Fortunately, the advent of functional genetic engineering technologies such as RNA-guided CRISPR-Cas9 presents novel possibilities for complementing traditional malaria control methods: the genes of mosquitoes can be altered to inhibit the spread of malaria-causing pathogens to humans. This work conducts a study of the optimal release schedule for modified Aedes Aegypti mosquitoes, building from a mathematical model created by the Marshall Lab at the University of California, Berkeley.[2] This approach to vector control employs approximate dynamic programming (DP) methods to achieve "fixation", here defined as successful propagation of the modified genotype such that it is inherited by at least 50% of the male population. The use of HPC would drastically expand the scope of this work and enable methodological improvements, including through more detailed modelling of the DP formulation and a better quantification of relevant uncertainties.

Type Ia supernovae (SNe Ia) play a critical role in astrophysics, yet their origin remains mysterious. A crucial physical mechanism in any SN Ia model is the initiation of the detonation front which ultimately disrupts the white dwarf progenitor and leads to the SN Ia. We demonstrate, for the first time, how a carbon detonation may arise in a realistic three-dimensional turbulent electron-degenerate flow in a new mechanism we refer to as turbulently-driven detonation. Using both analytic estimates and three-dimensional numerical simulations, we show that strong turbulence in the distributed burning regime gives rise to intermittent turbulent dissipation which locally enhances the nuclear burning rate by orders of magnitude above the mean, and may lead to supersonic burning and a detonation front. As a result, turbulence plays a key role in preconditioning the carbon-oxygen fuel for a detonation. The turbulently-driven detonation initiation mechanism leads to a wider range of conditions for the onset of carbon detonation than previously thought possible, with important ramifications for SNe Ia models.

Performance events or performance monitoring counters (PMCs) have been originally conceived, and widely used to aid low-level performance analysis and tuning. Nevertheless, they were opportunistically adopted for energy predictive modeling owing to lack of a precise energy measurement mechanism in processors, and to address the need of determining the energy consumption at a component-level granularity in a processor. Over the years, they have come to dominate research works in this area.

Modern hardware processors provide a large set of PMCs. Determination of the best subset of PMCs for energy predictive modeling is a non-trivial task given the fact that all the PMCs cannot be determined using a single application run. Several techniques have been devised to address this challenge. While some techniques are based on a statistical methodology, some use expert advice to pick a subset (that may not necessarily be obtained in one application run) that, in experts' opinion, are signicant contributors to energy consumption. However, the existing techniques have not considered a fundamental property of predictor variables that should have been applied in the the first place to remove PMCs unt for modeling energy. 

We propose to address this oversight in this talk. We present a novel selection criterion for PMCs called additivity, which can be used to determine the subset of PMCs that can potentially be used for reliable energy predictive modeling. We also show that the use of non-additive PMCs in a model renders it inconsistent.

The Navier-Stokes Equations, incompressible or compressible, are a set of non-linear partial differential equations used to model important phenomena, e.g. fluid flow around ship hulls, wind turbine simulations, blood flow inside the body. We employ a finite element method called Hybridizable Discontinuous Galerkin method to reduce the continuous problem into successive solutions of linear systems of equations. The size of this linear system depends on the required fidelity of the solution. The size can easily grow to millions and we need highly parallelizable and efficient methods to solve them. For those sizes, it is either impossible or incredibly infeasible to use direct methods, like Gaussian elimination, and we would like to use the so-called Krylov subspace methods. However, the performance of these methods highly depends on the size of the problem and the speed of the fluid. The idea is to improve the performance by using a preconditioner. We generalized the Pressure Convection-Diffusion preconditioner which was developed for another finite element method and we are going to show the results of our investigation for three test problems.

Protein loops, the "nonregular" structures connecting secondary structures, have been considered the source of heterogeneity responsible for protein function. Recent studies suggest that molecular flexibility and the dipeptide makeup of proteins are ultimately responsible for protein structure and the rise of genetics. Similarly, folding speed, which is correlated to flexibility, was found to increase during protein evolution. Since loop flexibility maybe a major link between structure, function and evolution, its study at genomic level could unlock prediction of evolutionary trajectories of macromolecules and facilitate advances in synthetic biology and translational medicine. My dissertation research aims to address the following: (1) determine history in protein structures using phylogenetic and graph-theoretical frameworks that link structure and flexible loop regions, (2) study patterns of flexibility and function in the three domains of life as well as viruses, and (3) determine loop motions that are associated with functional diversity using molecular dynamic (MD) simulation. In order to achieve these objectives, we make use of advanced methods of phylogenomic analysis to dissect the genomic impact of the combinatorial rearrangement of loops and domains and patterns of molecular accretion in the emergence of molecular functions. Additionally, we use NCSA Blue Waters, to integrate biophysics and genomic evolution by linking changes unfolding at nanosecond to microsecond levels with those spanning billions of years of protein history. In summary, our study weaves together, the paradigms of genomics, proteomics and biophysics.

One main goal of this Ph.D. project at the University of Valencia (supervised by Prof. M. A. Aloy, CAMAP – Computer Aided Modeling of Astrophysical Plasmas) is the study of conditions leading to an efficient jet launching around (rapidly) rotating Kerr black holes. An initial focus of the project was the search for suitable electromagnetic field configurations in the vicinity of black holes (also known as the magnetosphere) as solutions to the so-called relativistic Grad-Shafranov equation. For this purpose, a numerical solver for the relativistic Grad-Shafranov equation was implemented and tested - we were able to analyze and improve the numerical techniques used to solve the relativistic Grad-Shafranov equation across its singular surfaces and provide a detailed review of convergence properties. Recent developments in the numerical simulation of black holes and their relativistic outflows underline the need for both reliable initial data, and evolution procedures for highly scalable 3D simulations. In the framework of the Einstein Toolkit, an evolution thorn for force-free electrodynamics has been implemented and is currently undergoing first test stages.

The implementation of the numerical Grad-Shafranov solver has been done in Fortran and relies on OpenMP parallelization as well as the computational architecture of the CoCoNuT code. The EinsteinToolkit (C/C++/Fortran) is a highly scalable, open-source environment for numerical General Relativity and relies on both OpenMP, and MPI parallelization. Recent development efforts include, e.g., its extension to GPU resources as well as adaptive mesh refinement on MPI framed grids.

The most fascinating property of diblock copolymers (DBCP) is their ability to self-assemble into a wide range of ordered structures. In the simplest case, DBCPs can assemble into two distinct spherical phases, the familiar body-centered and face-centered cubic phases. The situation changes upon introducing conformational asymmetry, in which the statistical length of the two monomer types can differ. Experimental and theoretical studies have found the emergence of complex spherical structures, the Frank-Kasper (FK) phases, in conformationally asymmetric DBCPs. Recent experimental work confirms that conformational asymmetry is a key factor to realizing the FK phases. However, the observed differences between experimental and theoretical results suggest that alongside conformational asymmetry, other mechanisms may play a role in the formation of these intricate structures. One candidate is polydispersity, a measure of how uniform the polymer chain lengths are. Using the self-consistent field theory, we examine how polydispersity affects the relative stability of the FK phases in conformationally asymmetric DBCPs. From the study, we find that the formation of the FK phases depends on the functional form of the chain length distribution.

It is estimated that 50% of the power consumption of an aircraft is due to the skin-friction drag, i.e. the friction between the fluid and the surface. Therefore, reducing the skin-friction has significant economic and environmental ramifications. Injecting air, even at very low speed, from a porous surface (it is called wall blowing) can reduce the skin-friction by 80% for subsonic and supersonic flows over the surface. However, the energy expenditure by the blowing system can be high, leading to loss or small saving of the total power. In the current study, Bayesian optimisation in conjunction with high fidelity numerical simulations of flow in the wall vicinity, which are conducted by INCOMPACT3D, is used to considerably reduce the skin-friction drag and maximise the net energy saving. Primary results showed that Bayesian optimisation converged rapidly, after 13 evaluations, to the optimum parameters. The associated total energy saving of the obtained parameters is 5% with local skin-friction reduction of 40%.

Random variability of paramaters and imprecision are two significant sources of uncertainty in electromagnetic compatibility(EMC) models. While random variability can be represented by probability distribution functions, imprecision is better accounted for by possibility distribution. As practical situations in EMC models often involve both types of uncertainty, a hybrid approach is used to combine such possibilistic parameters with classical probabilistic parameters described by random variables, so to achieve a comprehensive characterization of uncertainty in model outputs. As an illustrative example, the proposed approach is applied to the estimation of differential mode currents induced in the terminations of a twisted wire pair with an unknown number of twists. The obtained results are compared versus conventional Monte Carlo simulations.