A Visualization Framework For Earth Materials Studies Bijaya Bahadur Karki Graduate Students: Dipesh Bhattarai and Gaurav Khanduja Department of Computer.

A Visualization Framework For Earth Materials Studies

Bijaya Bahadur Karki

Graduate Students: Dipesh Bhattarai and Gaurav Khanduja

Department of Computer ScienceDepartment of Geology and Geophysics

Louisiana State University, Baton Rouge, LA 70803

2nd Workshop: Minneapolis, August 5-10, 2007

Studying Materials Problems

Simulation algorithms

Compute- and data-intensive

applications

Visualizationalgorithms

PWscf, VASPPCMD

Parallel and distributed computing

Mantle materials: Silicates and oxides

Rheology Liquids

Massive multivariate data:MDVSTMRReVis

Tezpur(15.3 TFlops, 360 nodes2 dual-core processor)

Queen Bee

(50.7 TFlops, 680 nodes2 quad-core processors)

Visualization: Definition

• Process of making a computer image for gaining insight onto data/information– Transform abstract, physical data/information to a form that can be seen (i.e., visual representation)

– Enhance cognitive process

Visualizing Materials Data

• Properties/processes of interest:– Microscopic

•Atomic structure, dynamics•Electronic structure

• Data characteristics:– Three-dimensional, time-dependent– Multivariate– Massiveness, multiple sets– Computational, experimental origin

– Macroscopic•EoS, elasticity, thermodynamics

Application-Based Approach

• Numerous visualization systems exist– None of them may be good enough– Lack of desired functionality and flexibility

• How to meet domain-specific needs– Presentation and interactivity– On-the-fly data processing– Multiple sets of data– Visualization with database– Remote and collaborative visualization– Visualization/computational steering

Current Visualization Activities

• Multiple datasets visualization (MDV)– Electron density distribution

• Space-time multi-resolution (STMR) visualization– Atomic structure and dynamics

• Remote visualization– Elastic moduli and wave propagation

Multiple Datasets Visualization

Simultaneous rendering of more than one set of data to examine cross-correlation among them

Isosurface extractionGPU-based visualizationAdaptive scalable approach

Example: Electronic Structure

Mg2- vacancy defect in MgSiO3 post-perovskite

Perfect Defect Defect - Perfect Difference in two images

Initial configurations

Final configurations (after relaxation)

Scalable Adaptive Isosurface Extraction

Octree data structure

High resolution Low resolutionDual resolution

Multiresolution approach

Originalcell

Octree nodes

Performance Analysis

Performance measurement on 64 sets of scalar volume data with size of 2563 and 5123

All-in-memory approach Scalable adaptive approach

GPU-Based Visualization

Graphics hardware assisted 3D textures

Interactive clippingIsosurface

Khanduja and Karki WSCG 2005GRAPP 2006WSCG 2007

Example: Electronic Structure

Mg2- vacancy defect in MgSiO3 post-perovskite

Perfect Defect Defect - Perfect Difference in images

Initial configurations

Final configurations (after relaxation)

MDV Example

25 sets of the scalar volume data of 2563 size in a planer clipped mode using 3D surface texture mapping

Electron density in liquid MgO as a function of time

Multi-scale color map:Blue: 0 to 0.05Blue and green: 0.05 to 0.5Red: above 0.5

Electron Density: Defects in MgSiO3 ppv

Mg Si O

Vacancies

Migrating ions

Electron Density: Defects in MgSiO3 ppv

Spheres and linesKarki and Khanduja, EPSL, 2007

Mg Si O

Vacancies

Migrating ions

Defects in MgSiO3 ppv: Atomic Structure

Mg: Green Si: Blue O: Red Vacancy site: Black

Mg Si O

Vacancies

Migratingions

Space-Time Multiresolution (STMR) Atomistic Visualization

Integration of visualization and complex analysis

On-the-fly extraction and rendering of a variety of data

Pair correlation, coordination and cluster structures

Dynamical behavior

Atomistic Visualization Modules

• Approach– Spatial proximity

– Temporal proximity

– Spatio-temporal analysis

• Model– Complete data rendering

– Local/extracted data rendering

Position-Time Series Data

O HMg SiAtomic Species:

Points: Complete data set Balls: Instantaneous configuration

Data: {P(jt) | 0 ≤ j ≤ N} where P(t) = {pi(t) | 1 ≤ i ≤ n}

Coordination EnvironmentGiven atomic system: Hydrous MgSiO3 liquid

Atomic species: spheres

16 different pair correlation structures

Cutoff distances from partial RDFs

Si-O

Coordination environment

Coordination stability

Coordination clusters

Radial distribution functions

Pair Correlation Matrix O HMg Si

O

H

Mg

Si

16 different types of nearest-neighbor pairs

Diagonal: like atoms Off-diagonal: unlike atoms

Radial Distribution Function

Spatial and temporal information on Si-O coordination

Coordination-Encoding

2 3 4 5 6Color map

Three-, four- and five-fold coordination

€

nniαβ = 1 ≤ j ≤ na : d(i, j) ≤ rmin

αβ ∧type( j) = β{ }

Coordination Stability

The lines (thickness encoding the bond stability) and center atoms (size encoding the coordination stability) are color-coded to represent, respectively, the length distribution and coordination states. The stability represents the fraction of the total simulation time over which a given bond or coordination state exists.

Bhattarai and Karki, ACMSE 2007

2 3 4 5 6Color map

Stability of Different Coordination

3 4 5 6

16 coordination states

Four types exist

0 1 2 3

4 5 6 7

8 9 10 11

12 13 14 15

Coordination Cluster

The lines (thickness encoding the bond stability) and center atoms (size encoding the coordination stability) are color-coded to represent, respectively, the length distribution and coordination states. The stability represents the fraction of the total simulation time over which a given bond or coordination state exists.

Spatial and temporal information on Si-O coordination

Bhattarai and Karki, ACMSE 2007

€

NN iαβ = nni

αβ ( jΔt)j =0

j =N

U

Coordination Cluster Per Atom

Spatial and temporal information on Si-O coordination

Coordination Visualization

16 different pair correlation structures

Cutoff distances from partial RDFs

Si-O

Coordination environment

Coordination stability

Coordination clusters

Radial distribution functions

Atomic species: spheres

Given atomic system: Hydrous MgSiO3 liquid

Visualizing Dynamics

Diffusion in 80-atoms liquid MgSiO3

Spheres for atomic displacements Ellipsoids for covariance matrices

Diffusion in 64-atoms liquid MgO

Bhattarai and Karki, ACMSE 2007

Elasticity visualization

Remote execution

Visualization and database server

Online data reposition

Elasticity Visualization - ElasViz

• Multivariate elastic moduli – Variation with pressure, temperature and composition

• Elastic wave propagation in an anisotropic medium– Velocity-direction surfaces– Anisotropic factors

Karki and Chennamsetty, Vis. Geosci., 2004

ReadData

CijPlot

Modules of ElasViz

GenerateDirection GenerateVelocity

AnPlotDrawVelocity

Other Modules Display

Global Visualization Mode

Selective Visualization Mode

Summary

• Visualization for gaining insight into a variety of datasets for important minerals properties and processes – Increasing amounts of data from simulations and other resources.

• Important visualization systems under development: – Elasticity, atomic and electronic data

• A lot needs to be done:– Adding more functionalities– Merging atomistic and electronic components– Extending for remote and distributed access– Adopting in virtual (immersive) environment.

Support from NSF (EAR 0347204, ATM 0426601 and EAR 0409074).