Video Clip (34Mb)

Prof. Chris Johnson Discusses the Golden Age of Scientific Computing

Video Clip (4.8Mb)

This image shows a proton beam moving along the beam pipe (z-axis) in the presence of an electron cloud. The proton beam is shown in red in the center of the pipe. A. Adelmann, PSI.

Video Clip (5.5Mb)

Electron Cloud simulation with electrons and protons rendered as particles. A. Adelmann, PSI.

Video Clip (5.7Mb)

Electron Cloud simulation with electrons rendered as volume density and protons rendered as particles. A. Adelmann, PSI.

Electron Cloud simulation: trajectories of electrons as the simulation progresses rendered as splines colored by the magnitude of the velocity. A. Adelmann, PSI.

Electron Cloud simulation: trajectories of electrons as the simulation progresses rendered as splines colored by the magnitude of the velocity. A. Adelmann, PSI.

Tracking particles along a beam line. A. Adelmann, PSI.

Tracking particles along a beam line. A. Adelmann, PSI.

Tracking particles along a beam line. A. Adelmann, PSI.

Video Clip (32Mb)

The heightfield represents particle density from a VORPAL dataset of a laser wakefield acceleration. The spheres are individual particles whose velocity exceeds a certain threshold. Since we only had access to a single timestep of data, we randomly perturbed the heightfield and spheres to show we can handle time-varying datasets. T. Ize, SCI. Data data provided by Tech-X corporation and Peter Messmer.

Visualization of Magneto-rotational instability and turbulent angular momentum transport.

Visualization of Magneto-rotational instability and turbulent angular momentum transport.

Video Clip (13.6Mb)

Visualization of Magneto-rotational instability and turbulent angular momentum transport.

The featured plot is a Volume plot of the logarithm of gas/dust density in an Enzo star and galaxy simulation. Regions of high density are white while less dense regions are more blue and also more transparent.

The data used to make this image were provided by Tom Abel Ph.D. and Matthew Turk of the Kavli Institute for Particle Astrophysics and Cosmology.

Visualizations of magnetically unstable cylindrical Couette flow. This image shows the enstrophy and regions of high hydrodynamic dissipation, and was created by Cristina Siegerist, LBNL, in a collaborative effort between VACET, the NERSC Analytics program (www.nersc.gov)
Simulations by F. Cattaneo(1,2), P. Fischer(1) & A. Obabko(2)
(1) Argonne National Laboratory
(2) University of Chicago
as part of DOE's INCITE program.

Visualizations of magnetically unstable cylindrical Couette flow. This image shows the enstrophy and regions of high hydrodynamic dissipation, and was created by Cristina Siegerist, LBNL, in a collaborative effort between VACET, the NERSC Analytics program (www.nersc.gov)
Simulations by F. Cattaneo(1,2), P. Fischer(1) & A. Obabko(2)
(1) Argonne National Laboratory
(2) University of Chicago
as part of DOE's INCITE program.

(Above) Time evolution visualizations of a September 1970 hurricane data set. For more information, images and video clips please see Cristina Siegerist's Project website.

Volume visualization from a weather simulation. Images shows a condensed water field 1008x1008x300. UCRL-JC-145931 Effects of Domain Size and Numerical Resolution on the Simulation of Shallow Cumulus Convection. David E. Stevens (LLNL), Andrew S. Ackerman (NASA AMES), and Christopher S. Bretherton (UW, Seattle)

The image shows the component of the atmospheric CO² concentration that results from the net ecosystem exchange (NEE), which is shown on the land surface. This "green CO²" is the flux due to the respiration of vegetation, respiration of soil microbes, and fire minus that taken up by ecosystem production.
Credits: Data produced by the CCSM climate simulation code. Visualization produced by Jamison R. Daniel (ORNL) for Warren Washington (NCAR), John Drake (ORNL), and Forrest Hoffman (ORNL).

Our ability to generate ever-larger, increasingly-complex data, has established the need for scalable methods that identify, and provide insight into, important variable trends and interactions. Query-driven methods are among the small subset of techniques that are able to address both large and highly complex datasets. We present a method that increases the utility of query-driven techniques by visually conveying statistical information about the trends that exist between variables in a query. In this method, correlation fields, created between pairs of variables, are used with the cumulative distribution functions of variables expressed in a user's query. This integrated use of cumulative distribution functions and correlation fields visually reveals, with respect to the solution space of the query, statistically important interactions between any three variables, and allows for trends between these variables to be readily identified.

The image sets below show the results of a new technique for revealing the relationships between variables in complex, multivariate datasets. We demonstrate our method by analyzing interactions between variables in two combustion simulations: a methane V-flame (image 1), and an ultra-lean premixed hydrogen flame (images 2 and 3).

These images depict increasing ((a) through (f)) isosurface values of temperature (isotherms) colored by values of the correlation field derived from water (H₂O) and ethylene (C₂H₄). As temperature values increase, the predominant correlation between H₂O and C₂H₄ along the isotherms shifts from strongly positive (red) in (a), to strongly negative (blue) in (f). This shift suggests that temperature is itself negatively correlated with the H₂ O-C₂ H₄ correlation. Images generated by Luke Gosink using VisIt. Data provided by Marc Day and John Bell from - Center for Computational Sciences and Engineering at LBNL

These images depict increasing ((a) through (f)) iso-concentrations of water (H₂O) colored by values from the correlation field of oxygen (O₂) and perhydroxyl radical (HO₂). As water concentration increases, the predominant correlation along the isosurface shifts from strongly negative (blue) in (a), to strongly positive (red) in (f). This shift suggests that H₂O concentration is itself positively correlated with the O₂-HO₂ correlation. Local variations in this observed correlation (e.g., the bottom of the isosurfaces transitioning from negative correlation to positive correlation faster than the top of the isosurfaces) are due to the fact that burning occurs unevenly along the isotherms. Such variations in combustion influences both the rates of reactions and the locations of reaction fronts. As such, transitions in correlation are expected to occur at different concentrations in the isosurfaces of H₂O. Images generated by Luke Gosink using VisIt. Data provided by Marc Day and John Bell from - Center for Computational Sciences and Engineering at LBNL

These images depict increasing ((a) through (f)) iso-concentrations of hydrogen radicals (H) colored by values from the correlation field of oxygen (O₂) and perhydroxyl radical (HO₂). Each isosurface exhibits striations in the correlation field (i.e., bands of negative, zero, and positive correlation), and as H concentration increases, correlation increases within each striation (i.e., negative correlation tends to become positive). This behavior suggests that H concentration is itself positively correlated with the O₂ -HO₂ correlation. Image (f) indicates the simultaneous existence of near-zero correlation (i.e., high entropy) and high H concentrations. Images generated by Luke Gosink using VisIt. Data provided by Marc Day and John Bell from - Center for Computational Sciences and Engineering at LBNL

These images show a new technique for revealing the relationships between variables in a complex, multivariate dataset. The left and middle images show false-colored slices of Water (H₂O) and ethylene (C₂H₄) concentrations from a methane combustion dataset; red corresponds to high density, blue to low density. The right image shows a false-colored slice of the derived correlation field for these two compounds; red corresponds to strong positive correlation, blue to strong negative correlation, and green to little or no correlation. The switch from strong positive correlation to strong negative correlation in the reaction region corresponds to the area in which C₂H₄ is both produced and consumed, and H₂O is produced, in the process of combustion. The strong correlation (both positive and negative) in the center of the flame, as well as the atmospheric region, demonstrates the correlation field’s ability to show fine-scale interactions. Authors: Luke Gosink, John C. Anderson, Wes Bethel, Ken Joy. Data provided by Marc Day and John Bell from - Center for Computational Sciences and Engineering at LBNL

Rod-stabilized V-flame.

Video Clip (5Mb)

Rod-stabilized V-flame.

Sectional view of a rupturing steel container that is filled with a plastic bonded explosive and heated by a fire.

CH4 pool fire simulation.

Gyrokinetic Particle Simulation for Magnetic Fusion

Particle tracking visualization of a global, gyrokinetic 3D particle-in-cell simulation of plasma microtubulence in a tokamak toroidal fusion device. Only 22 of the 400 million particles in the simulation are being displayed based on the number of times that they get magnetically trapped (red line) and de-trapped (blue line) in relation to the externally imposed magnetic field. In the absence of turbulence, a trapped particle would remain trapped during the whole simulation and an untrapped particle would do likewise. In the video clip (fig 1) we show how the developing turbulence affects the trapping state and radial diffusion of the particles. The large circle following one of the particles displays the electrostatic potential around that particle. As the potential fluctuations increase due to the growing turbulence, the particle starts to interact with the wave through a different kind of trapping and detrapping process, resulting in a large radial diffusion, the amount of which is represented by the size of the particle. This simulation demonstrate the collision-less transport of particles, and thus energy, due to microturbulence. Although the particle trajectories are projected on a 2D cross-section for visualization purposes, the simulation itself is in 3D, (fig 2-4). It was carried out by Stephane Ethier of the Princeton Plasma Physics Laboratory using the Gyrokinetic Toroidal Code (GTC) (Z.Lin et al., Science 281, p.1835, 1998) on the Opteron-based system Jacquard at the National Energy Research Scientific Computing Center (NERSC). This work was performed under the DOE SciDAC Center for Gyrokinetic Particle Simulation of Turbulent Transport in Burning Plasmas. The visualization was produced by Allen Sanderson of the Scientific Computing and Imaging Institute under the DOE SciDAC Visualization and Analytics Center for Enabling Technologies (VACET).

Video Clip (43.5Mb)
Fig 1. The developing turbulence affects the trapping state and radial diffusion of the particles. The large circle following one of the particles displays the electrostatic potential around that particle.

Fig 2. A 3D view of the simulation at the final time step.

Fig 3. A combined view showing the relationship of the 3D and 2D views of the simulation at the final time step.

Fig 4. An intermediate time step from the simulation showing the particles as they move through the electric potential field.

Image of a particle as it moves through the electrostatic potential that surrounds it. The particle interacts with the potential resulting in a large radial diffusion. By using a transparent rendering it is possible to see through the electrostatic potential at the same time the potential values near zero are removed which allows the structure of the mode waves with in the field to be seen.

A visualization of a series of atomic particles that are part of a 3D simulation of magnetically confined fusion energy, visualized by Thiago Ize of the University of Utah's Scientific Computing and Imaging Institute. Such research is essential for the development of new energy sources, resulted from a collaboration between Department of Energy researchers and the University of Utah (data courtesy Stephane Ethier of the Princeton Plasma Physics Laboratory).

Topological Analysis of Magnetic Islands in DIID-D Tokamak

Video Clip (47Mb)

Video Clip (9Mb)

In many fusion simulation codes, the identification of instabilities in the plasma flow is critical to understanding the nature of tokamak experiment design. This visualization from the SCIRun visualization system shows a Poincare plot of magnetic field instabilities.

In many fusion simulation codes, the identification of instabilities in the plasma flow is critical to understanding the nature of tokamak experiment design. This visualization from the SCIRun visualization system shows the surfaces swept out by two "islands" of magnetic field lines.

Topological Landscapes: A Terrain Metaphor for Scientific Data

Topological landscapes map a contour tree that describes the topology of high-dimensional data sets to 2D landscapes. Peaks and valleys in the terrain represent minima and maxima of the original data set. Reparameterization of the landscape supports mapping measures, such as the volume of a topological feature, to the area of the "proxy" peak or valley, while persistence (i.e., the "value range" of a topological feature) is shown as height of representing peaks or valleys. By displaying this topologically equivalent landscape together with the original data we harness the natural human proficiency in understanding terrain topography and make complex topological information easily accessible.

The images show topological landscapes for various test data sets commonly used in the visualization community. Left panels show landscapes, while the right panels show corresponding volume rendered images. For the engine, hydrogen atom, methane and nucleon data set, features are shown in the same color in both panels.

Hydrogen: Notice how the landscape helps appreciate the fact that the two lobes in the hydrogen orbital (red, green) have larger function range than the toroidal ring in the middle (blue), although the ring takes a larger volume. One can also clearly see a large region at nearly zero persistence (yellow), which is probably an artifact from the construction process. In the past we completely missed this region due to its low function value even though it occupies a large portion of the volume. Authors: Gunther H. Weber, Peer-Timo Bremer, Valerio Pascucci.

Methane: In this case it is interesting to see how the landscape immediately clarifies, which feature is built around a minimum/ maximum, something not evident from the volume rendering. In particular, the main feature related to the carbon atom is associated with a large minimum, while each hydrogen atom is associated with two maxima. Clearly, the combination of these two images, even without interactive exploration, provides a better explanation than either of them independently. Authors: Gunther H. Weber, Peer-Timo Bremer, Valerio Pascucci.

Nucleon: For the nucleon and engine datasets, considerations similar to those for the hydrogen atom and methane data sets apply. The topological landscape immediately relates which regions are associated with maxima and minima and their persistence. Moreover, one understands better, which features take a large portion of the volume even though the 3D rendering does not contain that information explicitly due to occlusion. Authors: Gunther H. Weber, Peer-Timo Bremer, Valerio Pascucci.

Engine: For the nucleon and engine datasets, considerations similar to those for the hydrogen atom and methane data sets apply. The topological landscape immediately relates which regions are associated with maxima and minima and their persistence. Moreover, one understands better, which features take a large portion of the volume even though the 3D rendering does not contain that information explicitly due to occlusion. Authors: Gunther H. Weber, Peer-Timo Bremer, Valerio Pascucci.

Silicium: The silicium, neghip, and fuel dataset visualizations show topological landscapes together with a more traditional volume rendering using a global transfer function. This leads to a reduced correlation between the two visualizations and therefore hampers the benefit of the simultaneous presentation. Nevertheless, one can derive information that would not be obvious otherwise. In the case of the silicium data set, for example, one notices that the structures forming the lattice of the crystal are a set of maxima and minima all of similar persistence and all occupying similar volumes. This is easy to see in the topological landscape and its flipped counterpart. The volume rendering complements this information with a sense of the geometric shape of the actual crystal. For the neghip dataset the topological landscape reveals that many features that are small in terms of volume but span a large function range. Their geometric distribution is highlighted by the volume rendering. Authors: Gunther H. Weber, Peer-Timo Bremer, Valerio Pascucci.

Neghip: The silicium, neghip, and fuel dataset visualizations show topological landscapes together with a more traditional volume rendering using a global transfer function. This leads to a reduced correlation between the two visualizations and therefore hampers the benefit of the simultaneous presentation. Nevertheless, one can derive information that would not be obvious otherwise. In the case of the silicium data set, for example, one notices that the structures forming the lattice of the crystal are a set of maxima and minima all of similar persistence and all occupying similar volumes. This is easy to see in the topological landscape and its flipped counterpart. The volume rendering complements this information with a sense of the geometric shape of the actual crystal. For the neghip dataset the topological landscape reveals that many features that are small in terms of volume but span a large function range. Their geometric distribution is highlighted by the volume rendering. Authors: Gunther H. Weber, Peer-Timo Bremer, Valerio Pascucci.

Fuel: The silicium, neghip, and fuel dataset visualizations show topological landscapes together with a more traditional volume rendering using a global transfer function. This leads to a reduced correlation between the two visualizations and therefore hampers the benefit of the simultaneous presentation. Nevertheless, one can derive information that would not be obvious otherwise. In the case of the silicium data set, for example, one notices that the structures forming the lattice of the crystal are a set of maxima and minima all of similar persistence and all occupying similar volumes. This is easy to see in the topological landscape and its flipped counterpart. The volume rendering complements this information with a sense of the geometric shape of the actual crystal. For the neghip dataset the topological landscape reveals that many features that are small in terms of volume but span a large function range. Their geometric distribution is highlighted by the volume rendering. Authors: Gunther H. Weber, Peer-Timo Bremer, Valerio Pascucci.

A visualization of a vortex roll-up from the impulsive Rayleigh-Taylor instability at a density interface that shows baroclinic vorticity production (color) superimposed on vorticity magnitude (height representation). The adaptive mesh was generated by LLNL's multiresolution, view-dependent terrain rendering system, SOAR, from a 2D data set produced by LLNL's Miranda code.
Authors: Paul Miller and Andy Cook (simulation), Peter Lindstrom (visualization)

This image, created by Hank Childs of LLNL, is a visualization of a data set consisting of a twenty seven billion grid cell Rayleigh-Taylor simulation, which models the turbulent mixing of fluids. The simulation was done by MIRANDA, which also originated in ASC, but is now also being developed jointly with the SciDAC Physics Research Projects "Simulations of Turbulent Flows with Strong Shocks and Density Variations."

This image is a visualization of a Rayleigh-Taylor instability calculation, MIRANDA, containing 27 billion grid cells. It was created by Hank Childs (LLNL) using VisIt, one of VACET's production-quality scalable visualization platforms.

Visualization of coherent flow structures in a large scale delta wing dataset: Volume rendering of regions of high forward (red) and backward (blue) Finite Time Lyapunov Exponent. Coherent Structures appear as surfaces corresponding to the major vortices developing over the wing along the leading edge. Occlusion is a limitation that can be addressed with cropping or clipping. This image: Wing edge separation and the primary attachment layer. Inner structures are occluded.
Authors: Christoph Garth, Florian Gerhardt, Xavier Tricoche, Hans Hagen. Data courtesy of Markus Rütten, DLR Göttingen, Germany.

Visualization of coherent flow structures in a large scale delta wing dataset: Volume rendering of regions of high forward (red) and backward (blue) Finite Time Lyapunov Exponent. Coherent Structures appear as surfaces corresponding to the major vortices developing over the wing along the leading edge. Occlusion is a limitation that can be addressed with cropping or clipping. This image: Crop along the middle third of the left wing edge. The interplay of separation and attachment structures is visible on the front face. The grey box highlights the separation structure that characterizes a vortex breakdown bubble.
Authors: Christoph Garth, Florian Gerhardt, Xavier Tricoche, Hans Hagen. Data courtesy of Markus Rütten, DLR Göttingen, Germany.

First image: Generalized streak line in flow past a cuboid dataset. The streak line (blue) starts from the moving position of the singularity. The positions, i.e. the path (turquoise), is located in the lower part of the cuboid. After the singularity reaches the Hopf bifurcation at the right end of the path the computation stops releasing new particles for the streak line. Thus the streak line separates from the singularity path (right most image). Authors: Alexander Wiebel, Xavier Tricoche, Dominic Schneider, Heike Jänicke, Gerik Scheuermann.

Second image: Snapshots taken from an animation of generalized streak line and volume rendering of pressure in a flow past an ellipsoid dataset. Low pressure, which is correlated to vortical activity, is mapped to high opacity. The generalized streak line starts from the blue sink path in the lower left of the images (see e.g. third image). Authors: Alexander Wiebel, Xavier Tricoche, Dominic Schneider, Heike Jänicke, Gerik Scheuermann.

Visualization by analogy. The user chooses a pair of visualizations to serve as an analogy template. In this case, the pair represents a change where a file downloaded from the WWW is smoothed. Then, the user chooses a set of other visualizations that will be used to derive new visualizations, with the same change. These new visualizations are derived automatically. The pipeline on the left reflects the original changes, and the one on the right reflects the changes when translated to the last visualization on the right. The pipeline pieces to be removed are portrayed in orange, and the ones to be added, in blue. Note that the surrounding modules do not match exactly: the system figures out the most likely match.
Authors: Carlos E. Scheidegger, Huy T. Vo, David Koop, Juliana Freire, Claudio T. Silva.

Interactive Visualization An example of a VisTrails pipeline used to create an interactive visualization within the spreadsheet of the visible human. An interactive cutting plane separates volume rendering and isosurfacing algorithms and is used to update orthogonal slices in separate windows.
Authors: Carlos E. Scheidegger, Huy T. Vo, David Koop, Juliana Freire, Claudio T. Silva.

Parameter Exploration A snapshot of the parameter exploration interface for a simple one-dimensional exploration of isosurface values in the visible human.
Authors: Carlos E. Scheidegger, Huy T. Vo, David Koop, Juliana Freire, Claudio T. Silva.

Mesh Comparison An example of exploratory visualization for comparing isosurface extraction techniques using VisTrails. Complete provenance of the creation and exploration process is displayed in a history tree on the left. One pipeline that combines five different software libraries is shown in the middle. A parameter exploration of this pipeline for three different datasets and isosurface parameters is shown in the Visualization Spreadsheet on the right.
Authors: Carlos E. Scheidegger, Huy T. Vo, David Koop, Juliana Freire, Claudio T. Silva.