OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
Nanosciences
Bill SheltonMichael O’KeefeDerrick ManciniBahram Parvin
Rick RiedelIan Anderson
March 16, 2004
Data Management workshop
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
• A highly collaborative and multidisciplinary research center
• Co-located with the Spallation Neutron Source (SNS) and the Joint Institute for Neutron Sciences (JINS) on ORNL’s “new campus”
• JINS: Housing and dining facilities, auditorium, classrooms, for research visitors and students
• SNS: Will provide access to unique neutron scattering capabilities for nanoscience
• CNMS: Provides urgently needed capabilities for materials synthesis, nanofabrication, and modeling
Scientific Scope and Vision for CNMSCenter for Nanophase Materials Sciences
Nanofabrication Research Lab
CNMS Offices and Labs
The CNMS Concept:Create scientific synergies
to accelerate discoveryin nanoscale science
CNMSORNL’s
SNSCampus
JINS
SNSCLO
UnderstandingThe data chain
Data curation Data (scientific)
Data (raw)
Data (publication)
Data (metadata)Software
Hardware
Data analysis
Data visualization
Data treatment
Data diagnostics
Data acquisition
Measurement
Data simulation
Understanding
Measurement
The data chain
Data curation Data (scientific)
Data (raw)
Data (publication)
Data (metadata)Software
Hardware
Facilit
y
User
Data analysis
Data visualisation
Data treatment
Data diagnostics
Data acquisition
Data simulation
Ownership?
• Data and Databases
• Metadata and Data Curation
• Data Visualization
• Remote Collaboration and Remote Access
• Automation and Intelligent Control
• Simulation (‘in silico’ experimentation)
• Distributed Computing (Grids)
• Synergy
March 2004
Motivation
• Management and computational requirements of nano-science data are complex— Three dimensional structures represented at
nano (shape level) and sub-nano (atomic level)— Flexible topologies as a function of external
stress and atomic interactions (temporal evolution)
— Presence of real data for validation and refinement of the model parameters
— Multi-resolution information from sub-nanometer to micro-meter, computed quantitative data, meta data
— Variable data formats
March 2004
Atomic image reconstruction from observational data
160 Mbytes of image data per 2D reconstruction to atomic resolution.24 Gbytes per (future) 3D reconstruction to atomic resolution.
Image of 7nm Au nanoparticle supported on carbon substrate. Reconstructed to sub-nanometer resolution from 20 electron microscope images.
Columns of atoms viewed end-on (white dots) reveal the internal structure. The particle exhibits 5-fold twinning, with one twin disordered to take up strain (right).
Mike O’Keefe, Bahram Parvin, Larry Allard, “Structural characterization of nanoparticles”
March 2004
Atomic image simulation and comparison with observational data
Bahram Parvin, Mike O’Keefe et al, “Convergence of simulation and observational data at atomic resolution”
160 Mbytes of image data per reconstruction to atomic (sub-nanometer) resolution
On-line image simulated from atomic model available to operator at the microscope
Drag and drop capability for validation of experimental image with simulated “Virtual Electron Microscope” image
Atomic-resolution image of carbon atoms (white) in diamond structure
March 2004
Shape evolution at nano-scale
• Macro-level shape representation as a function of stress (1.5 Gbytes/10-minute experiment)— Automated tracking of nano-particles— Managing images, quantified nano-particle shape
representation, and time-varying stress data— Kinetics of macro-level shape
• Comparison to simulated models
Bahram Parvin, Mike O’Keefe et al, “Automated in-situ electron microscopy”
Below melting point
Computer-controlled tracking and shape characterization of Pb nano-particle in aluminum
Above melting point
March 2004
Issues on shape reconstruction and comparison at nano-scale
• 3D Reconstruction from sparse views (1 - 2 Gbytes/reconstruction)• 3D Geometric representation and comparison • Tracking computed geometries from macro to sub-nano-scale
Ge Cong and Bahram Parvin, “Shape from Equal Thickness Contours”, 2001
March 2004
Challenges
• Tracking three dimensional shape evolution of the range from macro to nano-scale
• Developing object level multi-scale representation of shape features for querying and comparative analysis
• Migrating toward structure-function informatics instead of more low-level-representation data management...
• Rapid simulation tsimulation << tmeasurement
• Intelligent Control• Synergy
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
A distributed approach?A distributed approach? ?
Data Acquisition System
Remote storage
Local users
High Speed Network
Supercomputers
Metadata
rawdata
Remote userswith local computing and storage
Remote users
Super computers Nanofabrication Research Lab
CNMS Offices and Labs
~50 TBytes/year/facility
~10 GBits/s
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
Facility
Instruments
Sample environment
Data treatment
Scientific results
Impact?
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
FacilityInstruments
Sample environment
Data treatment
Scientific results
Funding!
OAK RIDGE NATIONAL LABORATORYU. S. DEPARTMENT OF ENERGY
Oak Ridge on the NSF Teragrid