Student Posters Repository

Many applications, ranging from radar technology to bioelectromagnetism, have utilised broadband simulations using the finite-difference time-domain (FDTD) method over recent years. However, the main drawback of the FDTD method is the difficulty in modelling electrically-fine geometrical features, including thin layers and narrow slots. When the FDTD method is employed to resolve such geometrical features in a problem space, it requires very fine spatial sampling of the problem space, and thus it demands excessive computational resources in terms of memory and CPU time, due to the increase in the FDTD grid points and small time-step usage required to satisfy the Courant-Friedrichs-Lewy (CFL) condition. To enhance computational efficiency of the FDTD method, we introduced a subcell technique to model frequency-dependent thin layers in the FDTD method, where the frequency dependency is represented by the one-pole Debye model, and incorporated into the FDTD method by means of Auxiliary Differential Equation (ADE) formulation. This poster presents a practical application of the subcell thin-layer model to the Digital Human Phantom (DHP).  

The Beam Longitudinal Dynamics code BLonD is a unique simulation code developed at CERN to model beam motion in synchrotrons. Beam parameters for important upgrades and future studies are based on the results of BLonD simulations. The code is written in a modular fashion, so that the physics content of each simulation can be adapted to special purposes. The code was originally developed in python and has been now translated to C++ to apply various optimization techniques. The present CERN grid infrastructure restricts the complexity of physics problems that can be modelled due to performance limitations. A dedicated computer architecture is necessary to meet the present and growing future computational needs in the field of longitudinal beam dynamics.

Given these needs, studying computer architectures and hardware accelerators suitable specifically for complex simulators is essential. The interaction between different modules has to be investigated in detail. Selecting a few representative, physics cases, the feasibility and performance of different combinations of architecture including Xeon Phi, GPU, and FPGA components are to be evaluated. In the first stage, testing on the existing CERN OpenLab and TechLab infrastructures is envisaged. Eventually, prototype hardware acquisition for testing purposes is likely to be necessary. Ideally, the outcome of the PhD thesis will be a proposal for a concrete, dedicated computer architecture that will be purchased and used for longitudinal beam dynamics studies.

Recent technology has achieved nano scale fabrication of the electronic devices. The electronic transport properties in such atomic scale materials, where quantum features are remarkable, are much different from that in macro scale materials. Therefore, electronic transport properties in nanoscale devices need to be investigated on the basis of the first-principles calculation, which is a method of numerical calculations of electronic structure based on quantum mechanics. However, since even small electronic devices consist of more than one million atoms, it goes without saying that HPC is necessary to write program codes to investigate their properties. Recently, we proposed a first-principles electronic transport calculation method whose computational cost is linear to the system size. I will explain the details of the electronic transportation calculation method and a program code of the hyper-parallel and high speed first-principles calculation for electronic transportation.

Instead of common batch processing, real-time interactive simulation and visualization endows an immediate and flexible way to steer and investigate the simulations, 

in which the in situ methodology can avoid expensive data transfer and storage. Around interactive simulations as the central topic, this poster will contain three subtopics:

1) a real-time interaction way of using sense of touching in a coupled fluid-structure simulation.

2) a framework for real-time interactive remote simulations which allows distributed collaborative simulation with interactions from men or machines.

3) a special designed interconnection system for interactive simulations at medium cluster scale. 

A detailed analysis of the reaction of CH3CCH2 and CH3CHCH with molecular oxygen is presented. The C3H5O2 potential energy surface (PES) has been characterized using a combination of electronic structure methods. The majority of the stationary points on the PES were determined at the CCSD(T)-F12a/cc-pVTZ-F12//B2PLYPD3/cc-pVTZ level of theory, with the remaining transition states computing using multi-reference methods. Microcanonical rate theory and the master equation are used to determine the temperature- and pressure- dependent rate coefficients for each reaction channel. The main product channels are CH2O + CH3CO for CH3CCH2 and CH3CHO + CHO for CH3CHCH. In contrast to vinyl (C2H3) + O2, the methylvinyl + O2 reactions remain chain propagating, even at high temperatures. The new rate coefficients were included in a detailed mechanism taken from the literature. These changes have a modest effect on the ignition delay time and laminar flame speeds for propene combustion.

Performance portability of a program is an important issue as a program may have to run on a set of diverse computing platforms, and significant amount of time and effort has to be spent to get the best performance on each platform. The lack of consensus on the definition and metrics for performance portability leads to further difficulties in the assessment and understanding of this concept. My research aims to explore the concept of performance portability, understand what makes a code performance portable, propose a comprehensive definition for performance portability and a metric to quantify it. I will further explore how optimizations and code transformations by the compiler and/or runtime tools can improve or affect the performance portability of a code. At this point, I present some preliminary evaluations of existing definitions and metrics for performance portability and assess their weaknesses and strengths.

My transportation problem focuses on Origin-Destination distribution. A destination choice is one of the most difficult problem in transportation research field. An origin-destination (OD) matrix is important for a road network planning because the OD matrix influences the network congestion strongly.  Previous studies just obtain one OD matrix to calculate the network congestion because of computational cost.  Recently, we can calculate the congestion on many OD scenarios using many CPUs.  This research tries to obtain a distribution of OD matrix using Monte Carlo sampling.  This sampling algorithm has three components and needs to obtain a convergence state between these components.  First component samples according to destination choice probability which is formulated by economic theory.  Second component solves non-linear simultaneous equations to obtain attractiveness of destination.  Third component obtains a traffic equilibrium and its travel costs using shortest path search and optimization algorithm.

For the optimization of flapping flight for micro air vehicles, a fluid-structure interaction (FSI) analysis system has been developed and used for analyzing various flapping motions. The system employed partitioned iterative coupling method, which a fluid domain and a structural one are calculated separately by each parallel three-dimensional finite element method solver, and physical quantities are interpolated by a coupling platform. The two solvers use MPI and the coupling platform utilizes Socket. In order to analyze the motions precisely in morphology and kinematics, the number of nodes and elements increase in fluid and structural meshes. Therefore, more efficient parallel computing is desirable for FSI analysis.

Plasticizers are an additive molecule used to make plastics more pliable. Polyvinyl chloride (PVC), a top commercial plastic with its usage continuously growing, needs plasticizer to ease its process-ability and make it suitable for applications from industrial cables to sensitive medical equipment. In recent years, plasticizer migration is a topic of concern and an important field of study because of its hazardous properties and its poor understanding of movement through plastics. However, migration of the plasticizer has a tremendous impact on plastic and on human health. After some time, the plasticizer ends up moving through the plastic due to various forms of contact environments like: exposure to air, liquid, or other plastics. The European Union (EU) has already banned the use of certain plasticizers in some applications due to negative effects on cardiac and reproductive development. 

The goal is to find the optimal plasticizer design to minimize migration by optimizing compatibility of the plastic-plasticizer system, and effectiveness of the plasticizer to alter the mechanical properties of the plastic. Traditionally migration data is collected experimentally, however it is a time consuming and resource heavy process. With molecular dynamic (MD) simulation, a coarse-grained (CG) model - a simple bead-spring model that generalizes particles as a bead and connects them with a finite-spring - will be used to explore the impact of plasticizer design on its movement through the plastic. Using MD, especially with CG modeling, the data collection time is faster and the cost of resources is minimal. The long-term goal is to design a computer aided screening process to easily determine which plasticizer is best suitable for the plastic.