Seeing is understanding

Visualizations help scientists make sense of complex data

The Berkeley Lab/NERSC Visualization Group maintains a suite of commercial visualization software for NERSC users, provides assistance in using it, and conducts research and development to expand the capabilities of visualization tools. This page provides a sampling of the results of their collaborations with NERSC researchers during the past year. Other examples of their work can be found in Figures 2–5, Figures 2–4, Figure 9, and Figure 7, and at http://www-vis.lbl.gov/Vignettes/, where there are also links to several animations.

Electron Cloud Visualization

The SciDAC Advanced Computing for 21st Century Accelerator Science and Technology project, led by Rob Ryne of Berkeley Lab, is studying electron cloud instabilities that can disrupt the main proton accelerator beam. Collaborators Andreas Adelmann of the Paul Scherrer Institut in Switzerland and Miguel Furman of Berkeley Lab have developed a simulation tool that calculates the positions and velocities of protons and electrons, but the large and complex datasets are difficult to manipulate and understand.

The Visualization Group is working with this team to develop and apply several different approaches for exploration and analysis of accelerator modeling results. Using AVS/Express software has allowed the researchers to move beyond simple scatter plots and actually track the movement of groups of particles in three dimensions (Figures 1 and 2). Two other tools, PartView and PPaint, were developed by the Visualization Group specifically for this type of research. PartView (Figure 3) provides a simple, interactive interface for routine data visualization and analysis. PPaint (Figure 4) enables researchers to “paint” selected particles in order to track their position, orientation, and motion through the beamline, both forwards and backwards in time.

Figure 1. Electron cloud simulation with electrons rendered as volume density and protons rendered as particles.

Figure 2. Trajectories of electrons are rendered here as splines colored by the magnitude of the velocity, while protons are rendered as particles.

Figure 3. PartView provides an easy-to-use interface for routine analysis of accelerator simulation data.

Figure 4. PPaint allows users to “paint” selected groups of particles in one of the 2D projected views, then view the selected particle groups in other projections. This approach facilitates understanding of time-varying particle behavior.

Getting AMR Data into Existing Visualization Tools

The Visualization Group is assisting participants in the SciDAC Applied Partial Differential Equations Center (APDEC) by providing a variety of data analysis options for understanding adaptive mesh refinement (AMR) datasets. These options include AMR Volume Renderer, ChomboVis, and file format converters that enable scientists to visualize their AMR data with their choice of off-the-shelf applications (Figure 5). The converters are necessary because most visualization algorithms assume a uniform grid, while AMR uses a hierarchy of grids of different cell sizes. The conversion software converts the grid hierarchy into a hexahedral unstructured mesh.

Seeing How a Protein Folds

Protein folding is the process by which a protein molecule — a simple unbranched chain of amino acids — assumes the complex shape, called its “native state,” that determines its biological function. The native state is stable because its shape minimizes the interaction energy of the atoms in the molecule. Computational modeling of energy minimization is one way of identifying possible folding pathways for a particular protein.

The Visualization Group collaborated with Ricardo Oliva, Juan Meza, and Silvia Crivelli of Berkeley Lab to visualize the energy minimization and folding of protein T209 (Figure 6), calculated with OPT++ from an initial configuration created with ProteinShop. In this visualization, the color of the atoms shows the relative speed of their motion at any given point in the process (Figure 7). Several movies of these simulations can be found at http://www-vis.lbl.gov/Vignettes/.

Figure 6. Calculating the minimum interaction energies for regions of a protein molecule and for the molecule as a whole results in a sequence of data that can be translated into images of protein folding.

Figure 7. The color scale shows the speed at which atoms are repositioning themselves into lower-energy configurations.

A Better View of Plasma Flow

The Quasi-Poloidal Stellarator (QPS), an experimental fusion reactor that is expected to be built at Oak Ridge National Laboratory by 2008, is being designed with the help of a computer optimization model that uses an extensive suite of physics design tools. With its compact, highly shaped configuration (sometimes described as a twisted donut), the QPS is expected to eliminate the violent plasma disruptions common in conventional research tokamaks at high plasma pressures because it will have only a fraction of the plasma current. If successful, the QPS may usher in a new class of smaller and more economical fusion reactor designs.

Cristina Siegerist of the Visualization Group worked with Oak Ridge physicist Don Spong to resolve some difficulties in visualizing vector fields with AVS/Express. Using an AVS network that evolved out of the one Siegerist created, Spong produced this visualization of simulated plasma flow streamlines in the QPS (Figure 8).

Figure 8. Simulated plasma flow streamlines in the QPS