Detection and Visualization of Defects in

3D Unstructured Modelsof Nematic Liquid Crystals

Ketan Mehta* & T. J. Jankun-Kelly

Viz Lab,

Computer Science & Engineering Department,

Mississippi State University

* Now at Vital Images. Inc

Presentation Outline

Introduction and Motivation Defect Detection Background Algorithm Case Studies Conclusion Future Goals Q&A

Nematic Liquid Crystals

Unique state of matter Sensitivity

Liquid Crystals Physics

Orientation measured via: scalar order parameter (S),

director (n) and Q-tensor.

Uniaxial Biaxial

Biosensor Design

Design a nematic liquid crystal (NLC)-based bio-sensor.

10-30nm

Nanostructured substrate

Receptor LC

10-30nm

Nanostructured substrateReceptor pesticide

Pesticide

Facilitate Bio-sensor Design Process

Computational simulation

(physics)

Existing analysis methods• Partial views• Cutting-plane • Isosurface

Extend and provide both correct and deeper

How does Visualization beneift design process?

Challenges

Explore unstructured grid

Identify defect cores

Visualize and present relevant information

Introduction and Motivation Defect Detection Background Algorithm Case Studies Conclusion Future Goals Q&A

Structural Defects – Disclination

Very high director (n) gradient. Defects influence physical properties.

+1/2 defect

-1/2 defect

Axial +1 defect

+1

+2

Defect Detection Methods – Physical Sciences

Toyoki and Zapotocky’s

2D approach.

Hobdell and Windel

extended 2D to 3D.

Fukuda et al. demonstrated

benefit of adaptive grids• Unstructured

motivation

[M. Zapotocky et al. Feb’95]

[J. Hobdell et al. ’97]

Defect Detection Methods – Scientific Visualization

Sparavigna et al. • Oriented LIC and

streamlines

[A. Sparavigna et al. Oct’99 ]

Slavin et al. used • Streamlines• Streamtubes• Ellipsoids• Smooth field

[V. Slavin et al. Oct’04]

Existing Approaches: Contour map, Isosurface and Streamlines

Tools used:Ensight and TecPLot

Existing investigation methods

Existing Approach: S-based Analysis

Cut-planeS-mapped to color

IsosurfaceS-based

Surface boundary

Validation method for our algorithm

Tool used: FieldView

Validation Approach

Defect detection algorithm validation using

• Well understood models

• Comparative analysis

• Insights from experts

Presentation Outline

Introduction and Motivation Defect Detection Background Algorithm Case Studies Conclusion Future Goals Q&A

Defect Core Identification in Unstructured Grid!

Problems?

Approaches not extensible

Defect classification fails!!

NNP Sorted Spiral Winding

Existing approaches on regular structure

Regularity fails in unstructured space

Ordered Winding

Nearest-neighbor-path traversals:

random (left) and ordered (right)

Nearest Neighbor - preserve spatial orientation

Defect Detection Algorithm –Visual Flow

DATA

GRID (HDF5)

Nearest Neighbor Sorted list Visualization of Defect Nodes

Pre-process Detect Visualize

Visualization

Defect structures and orientation vectors

Colored nodes and arrow

Analyte (impurity) outline visualized

as a mesh

Presentation Outline

Introduction and Motivation Defect Detection Background Algorithm Case studies Conclusion Future Goals Q&A

Effectiveness and Validation

Verification using comparative analysis

Based on three Case studies• Regular structured

• backward compatibility• Unstructured temporal

• detecting changing features• Unstructured complex model

• real-life application

I: Defect Detection: Structured Model

II: Defect Detection: Temporal

Time Line10000 15000 20000 30000

II: Defect Detection: Temporal

Time Line10000 15000 20000 30000

II: Defect Detection: Temporal

Time Line10000 15000 20000 30000

II: Defect Detection: Temporal

Time Line10000 15000 20000 30000

III: Defect Detection: Immunoglobulin G Model (IgG)

Complex bio-molecular model

Existing method- isosurfacing

Our

method

Summary So Far

Semi-automatic defect detection feasible• On structured and un-structured grids

Easier to navigate – reduces analysis time

Provides correct and deeper insight

Surface mesh (red) and isosurface (blue)

Existing Techniques: S-, b-parameter, Isosurface

Biaxial plot

Scalar order plot

Cut-plane (mesh)

Comparison of Existing and Proposed Technique

Defect nodes identified with our algorithm (new)

Surface mesh (red) and isosurface (blue) (existing)

Future Work

Integration with Q-tensor based analysis and NLCGlyph [Jankun-Kelly 06 – very soon].

Further user study based validation.

Defect classification scheme for unstructured grids.

Acknowledgement

Collaborators• Dr. Rajendran Mohanraj, • Huang Li

Faculty and Staff of the

Computer Science & Engineering Department, MSU

National Science Foundation EPSCoR program via award #0132618.

Vital Images, Inc.

Thank You!!