190 likes | 203 Views
A Case Study in the Visualization of Supernova Simulation Data. Ed Bachta Visualization and Interactive Spaces Lab. Overview. Introduction Lagrangian-Eulerian Advection Software Design Results Future Work. A Core-collapse Supernova. Begins with a star of 8+ solar masses
E N D
A Case Study in the Visualization of Supernova Simulation Data Ed Bachta Visualization and Interactive Spaces Lab
Overview • Introduction • Lagrangian-Eulerian Advection • Software Design • Results • Future Work
A Core-collapse Supernova • Begins with a star of 8+ solar masses • Eventually, fusion produces Fe in the core • Pressure from fusion loses to gravitation • Material falls inward, increasing density • Neutrinos radiated at a rate of 1057 /s • Strong force halts collapse • Remaining material rebounds off the core • Shock wave carries material away from the core
Simulation • Doug Swesty & Eric Myra, SUNY Stony Brook • Exploring the role of convection • Radiation hydro code scales to 1000s of procs • 2 spatial dimensions (soon to be extended to 3) • 20 groups of neutrinos at different energies
Lagrangian-Eulerian Advection • A process for visualizing vector fields, valid for unsteady flows • Noise is advected along the flow, generating an image as output • Results • Single frames portray instantaneous flow • Animations simulate motion of material in flow Vector plot LEA
“Lagrangian-Eulerian Advection for Unsteady Flow Visualizaion” v Noise at t-1 Lagrangian step Eulerian step • Particles seeded randomly each iteration • Backward integration finds upstream cell • Color of upstream cell advected forward • Results blended temporally with bias toward most recent • Jobard, Erlebacher, Hussaini (IEEE Vis & CG 2002 [8:3])
LEA Animated • Applied to velocity • Propagation of light and dark areas indicates direction of flow • Areas where noise remains have near-zero velocities
Software • Vis modules provided by the Visualization Tool Kit (VTK) • LEA filter for VTK developed at the Swiss National Supercomputing Centre • Scripts programmed in Python
Results • Combination of LEA with: • Scalar data representations • Via colormaps • Via iso-contours • Vector data comparisons • Via visualization of dot products
Velocity & Entropy • Shows the development of regions of high entropy in upper convective zones
LEA & Optical Depth • The iso-contour where optical depth = 1 describes the surface of last scattering 1 8 2 3 • Generated for each energy group • Our results show that these contours vary with energy group and evolve along with the shock
Advective vs. Radiative NeutrinoFlux • Radiative neutrino flux • Tendency to propagate outward • Advective neutrino flux • Effect of convection • Dot product indicates: • “Constructive” flux • “Destructive” flux • Orthogonal flux
Comparison of Gradients • Lagrangian multipliers: • ∂rf(r) = λ ∂rg(r) • Describes a set of points where the iso-contours of f(r) and g(r) are tangential • A positive λ indicates parallel gradients • A negative λ indicates anti-parallel gradients • Very similar to our dot product analysis • The dot product reveals orthogonal conditions
Entropy & Temp. • Using our visualization scheme, we can see: • Where the gradients are || • Where they are anti-|| • Where they are orthogonal • How this relates to the flow of a vector field
Future Work • Extending the framework to support iteration • Developing new visualization techniques • Enabling remote visualization • Intended for batch processing • Investigating Web Services