OpenAtom is a highly scalable application for quantum molecular dynamics which implements the Car-Parrinello ab-initio molecular dynamics (CPAIMD) method. OpenAtom is developed using the Charm++ parallel framework, allowing it to scale to hundreds of thousands of processors and remain portable across multiple platforms. The project has collaborators at the IBM Watson Research Center (Dr. Glenn Martyna), the Dept. of Chemistry at NYU (Prof. Mark Tuckerman), and the Dept. of Computer Science at UIUC (Prof. Laxmikant Kale).

Parallel applications may be represented as directed acyclic graphs with tasks as vertices and communicating tasks connected via edges. If vertex and edge weights represent task execution time and communication time, respectively, there exists an optimal task schedule and mapping of tasks to processors that minimizes the total execution time of the application. Normal MPI applications are not easily divided into tasks, however Charm++ applications are inherently divided into discrete tasks and can take advantage of such techniques. The goal of the project is to implement a framework to automatically find a schedule and mapping of tasks that minimizes total execution time for a Charm++ parallel application.

Volumetric particle tracking and velocimetry algorithms are useful in a variety of research contexts, including contagion tracking, identifying insect swarming patterns, observing material deformation, and aircraft wing foil testing. The input data into these algorithms is usually provided from cameras or lasers, but with high fidelity sensors (i.e. high resolution, high fps cameras) the post-processing necessary to accurately track particles is extremely time-consuming. We have developed GPU-enabled parallel algorithms to improve performance many times over and enable real-time calculation of particle trajectories.

Dynamic voltage and frequency scaling (DVS) is a common technique used by processors to conserve energy, however most of these techniques only vary CPU parameters and ignore other adjustable parameters (i.e. memory and bus voltages and frequencies). Given a set of tasks and knowledge of their relative use of each system component, we proposed a realistic energy model and an optimal static frequency assignment for each component. The proposed energy model was verified on a real system and the frequency assignment scheme showed up to 20% less energy consumption than other popular DVS schemes.

Fuel cell research is motivated by rising oil prices, the heavy dependence on foreign oil, and growing environment concerns over fossil fuels. Current proton exchange membrane (PEM) fuel cells require a source of clean hydrogen and are extremely sensitive to COx poisoning. Storing and transporting hydrogen has become a major obstacle in the wide deployment of PEM fuel cells, and a number of options have been proposed. Using the hydrogen bound in ammonia is a promising solution because ammonia has a much higher hydrogen weight percent than metal hydrides, and its hydrogen can be recovered without COx byproducts and at lower temperatures than methane. Additionally, a storage and transport infrastructure for ammonia is already in place and can be expanded with relatively little effort and cost. This study examined the activity of a ruthenium catalyst on a gamma-alumina support to recover the hydrogen in ammonia and produce clean hydrogen for use in fuel cells.

Polyethylene has become ubiquitous with applications ranging from high quality medical equipment to simple grocery bags. Its growing demand has created the necessity for efficient production of its constituent monomer, ethylene. A common method of ethylene production is via steam cracking of ethane, which is energy intensive and polluting. Typical operating temperatures for steam cracking are over 700C, requiring multiple heat recovery steps to maintain process efficiency, and side products include COx, NOx and SOx. Although the side products are inevitable due to impurities in the ethane stream (often collected from natural gas wells or oil reservoirs), heating requirements, conversions, and selectivities may be improved and finely controlled by the catalytic oxidative dehydrogenation of ethane in micro-reactors.

The addition of a catalyst allows the process to run auto-thermally with only pre-heating requirements on the input stream, and varying inlet composition (ethane:oxygen ratio) and reactor channel geometry allows finer control of conversions and selectivities. This study investigated the role of reactor geometry in ethylene selectivity and the auto-thermal capability of catalytic oxidative dehydrogenation.