1. Matthew C. Fleenor Roanoke College
Interactive Visualization of Intercluster Galaxy Structures in the Horologium-Reticulum Supercluster
Jameson Miller
UNC Chapel Hill
Cory W. Quammen UNC Chapel Hill
Matthew C. Fleenor Roanoke College

2. Data description Galaxy positions in RA-DEC-cz coordinate system
Right-ascension (RA) ~ longitude
Declination (DEC) ~ latitude
cz – radial dimension (recessional velocity)
~2500 galaxy locations
~30 clusters

3. Domain questions What is the distribution of intercluster galaxies?
Are there large void regions? How many?
Does the supercluster have filaments?
How do clusters fit into the structure defined by intercluster galaxies?

4. Standard 2D plots

5. Overvieew

6. Prior 3D experience Collaborators viewed data in immersive environment
Got lost – no context

7. Previous interactive tools
Cosmic Explorer [Song1993]
SGI Explorer [Christensen1995]
PartiView [Levy2001]
AstroMD [Gheller2002]

8. What’s missing Ability to group galaxies into structures they define
Voids
Filaments
Reference axes in all three RA-DEC-cz dimensions

9. Data types Sparse 3D position data Nominal (categorical) data
Intercluster galaxies vs. clusters
User-defined groups indicating structure (filament, void boundary)
Group 1
Group 2
Intercluster Galaxy
Cluster …

10. Nominal encoding 3D glyphs Nominal color encoding
Enough screen real estate
Distinct shapes encode object type
Depth queues from perspective and occlusion
Nominal color encoding
Group membership
Encoded by 10 of 12 colors recommended by Ware2004

11. RA-DEC-cz reference axes
Orientation to dataset
Colored with just-noticeably-different color than background
Can turn sides, top, and bottom on or off

12. Curved drop lines
Explicit connection between galaxies and reference axes
Curved to fit RA-DEC-cz coordinate system
Allows comparison to standard plots

13. Structure perception Structure-from-motion Torsional rocking Stereo
Strongest shape cue
Torsional rocking
Structure-from-motion without interaction
Stereo
Complements other techniques
User can control eye-separation parameter

14. Confirmation of analysis
Quantitative analysis shows two separate overdensities
Overdensities pop out in visualization
~95% correspondence between grouping by hand and quantitative grouping

15. Positive result Quick identification of void regions
Selection of galaxies along rim defines bounds of void
Offline sphere-fitting refines estimation
Six voids identified
Known clusters reside around voids

16. Negative result
Previous 2D plots identified potential filament structure
When rotated in 3D, filament is shown to be two separate structures

17. Computation vs. visualization
Voids – could compute
Filaments – maybe could compute
Visualization helps astronomers know where to focus quantitative analysis

18. Effective techniques User-controlled interaction Torsional rocking
Critical for maintaining viewer orientation
Interactive scaling allows snapping between overview of data and local features, giving context when zooming in
Home key moves back to a familiar orientation
Torsional rocking
Rocking aids void definition when far galaxies peek out from closer galaxy
View angle optimization
Stereo
Stereo is helpful for finding galaxies bordering voids
Picking with mouse was hampered by stereo
Collaborators came across campus to use it!

19. Software Called GyVe (Galaxy Viewer) Built on VTK, Python, and Tkinter
Available at
Runs on Windows and Linux

20. Future work Isosurfaces – initial attempt unused
Image processing for identifying voids from density projections
Add interactive statistical tools
Haptic probes to feel around for structures

21. Acknowledgements NSF Grant AST 04-06443 (Fleenor) James A. Rose
Russell M. Taylor II