August 29, 2015 Atomistic Characterization of Stable and Metastable Alumina Surfaces ECSS Symposium...
-
Upload
marshall-harmon -
Category
Documents
-
view
215 -
download
1
Transcript of August 29, 2015 Atomistic Characterization of Stable and Metastable Alumina Surfaces ECSS Symposium...
April 19, 2023
Atomistic Characterization of Stable and Metastable Alumina Surfaces
ECSS Symposium Jun 18, 2013
Sudhakar Pamidighantam
Atomistic Characterization of Stable and Metastable Alumina Surfaces
• PI Doug Spearot and co-PI Shawn Coleman at the University of Arkansas.
• Grant #: DMR130007• ECSS Team - Sudhakar Pamidighantam at NCSA - Yang Wang at PSC, - Mark Vanmoer at NCSA
2
Metastable Alumina
Because of their fine particle size, high surface area, and catalytic activity of their surfaces, the transition aluminas (especially the form) find applications in industry as adsorbents, catalysts or catalyst carriers, coatings, and soft abrasives. The excellent stoichiometry and stability of Al2O3 help to make it an important constituent of many protective oxide scales formed on the surface of high-temperature metals and alloys.
J. Am. Ceram. Soc., 81 [8] 1995–2012 (1998)
3
Stable Form
4
X-Ray
Crystal Lattice
Molecular and Solid State Structure
Characterization and Crystal Engineering
Function
Materials for Diverse Applications
Solid State Structure Determination and Function
XRD Compute In Lammps• The algorithm is sufficiently generic to be applicable to both
electron and x-ray diffraction conditions and• is integrated into the LAMMPS molecular dynamics
simulator • the algorithm is capable of performing diffraction analyses
either statically (single snap shot after energy minimization) or dynamically during a molecular dynamics simulation to produce time averaged diffraction patterns at finite temperature.
• A visualization procedure is developed to create SAD patterns and 2θ x-ray diffraction line profiles from the intensities computed using the atomistic simulation data.
XRD Algorithm• generates a three-dimensional mesh of points filling a volume of reciprocal
space constructed from the entire domain of the atomistic simulation cell.• The mesh of reciprocal lattice points is built on a rectilinear grid with spacing
cn |An|−1 along each reciprocal lattice axis Bn. Each reciprocal lattice axis Bn is determined from the associated vector An corresponding to the n = 1, 2, or 3 edge of the simulation cell.
• each reciprocal lattice point is associated with a reciprocal lattice vector K describing the deviation between the diffracted and incident wave vectors kD and kI
---- K = kD − kI = ξB1 + ηB2 + ζB3, sin (θ) = |K|---------- ------ --------- Braggs’s Law 1/dhkl = |KB| λ 2
Structure Factors Atomic Scattering Factor
(x-ray)
Parallel Approach to the Simulation of Diffraction Pattern
• X-ray virtual diffraction code (xrd) has been implemented as a LAMMPS compute, a plugin capable of computing the diffraction pattern for the structure generated by LAMMPS
• The parallization of xrd is atom based, and is implemented within the framework of LAMMPS
• Further parallization of xrd has been explored over the k-space mesh.
• Built a stand-alone xrd code as a test bed K-mesh loop
• Parallization based on a MPI implementation has shown significant performance enhancement
• Another parallelization level (on top of the parallelization over the atoms) and employ extra MPI (or openMP) processes to parallelize the k-mesh is planned
Ongoing Efforts in the K-mesh Parallelization of the XRD Compute
• Implement MIC offload directives in xrd to allow for taking advantage of the many core co-processors on Stampede
– A project under the XSEDE Student Engagement Program. Summer intern student: Paula Romero Bermudez
• Implement MPI group in xrd to allow for running on a large number of cores on a distributed memory system
– A XSEDE Campus Champion project. Campus Champion fellow for this project: Luis A. Cueva-Parra
Simulation Conditions
11
Virtual selected area electron diffraction (SAED) patterns are created by examining the region in reciprocal space intersecting the Ewald sphere of radius λ−1.
Simulated 200 kV electron radiation (λ = 0.0251Å) and Cu Kα x-rays (λ = 1.54178Å) are used to create SAED patterns and 2θ x-ray line profiles.
Embedded-atom method (EAM) potential is used for modeling atoms and their interaction.
Table. Parameters used to compute analytical approximations of the Ni atomic scattering factors for electron and x-ray diffraction as calculated via equations (5) and (6) respectively with sin θ/λ (Å−1).Electrona a1 a2 a3 a4 a5 b1 b2 b3 b4 b50.3860 1.1765 1.5451 2.0730 1.3814 0.2478 1.7660 6.3107 25.2204 74.3146
X-rayb a1 a2 a3 a4 b1 b2 b3 b4 c12.8376 3.8785 7.2920 0.2565 4.4438 12.1763 2.3800 66.3421 1.0341-------------------------------------------------------------------------------------------------------------a Parameters fit to equation (5) by Peng et al Peng L-M, Ren G, Dudarev S L and WhelanMJ 1996 Robust parameterization of elastic and absorptive electronatomic scattering factors Acta Crystallogr. A 52 257–76.b Parameters fit to equation (6) by Fox et al Fox A G, O’KeefeMA and TabbernorMA 1989 Relativistic Hartree–Fock x-ray and electron atomic scatteringfactors at high angles Acta Crystallogr. A 45 786–93
Virtual diffraction analysis of Ni [0 1 0] symmetric tilt grain boundariesS P Coleman, D E Spearot and L Capolungo, Modelling Simul. Mater. Sci. Eng. 21 (2013) 055020
Virtual Electron Diffraction Pattern and Visualization
12
Diffraction Intensity Lorentz Polarization Factor
Virtual selected area electron diffraction (SAED) patterns are created byexamining the region in reciprocal space intersecting the Ewald sphere of radius λ−1.
A thin hemispherical slice of the reciprocal lattice mesh lying near the surface of the Ewald sphere is isolated and viewed parallel to the zone axis.
The thickness of this slice is dependent on the resolution of the reciprocal space mesh and is chosen such that between 1-5 reciprocal lattice points
Integration Into SEAGrid Science Gateway
• Deployed Lammps_DS under the community access on Gordon• Registered the application in SEAGrid Applications registry• Registered ECSS team as users in SEAGrid – Separate PI ships for Mark and
Doug created• Added Restrictions to access to Lammps_DS to ECSS project team pending
public release - Demonstrating the access restriction to the PI - Ye Fan (NCSA) implemented this and additional output/log access for
debugging.• Created internal single resource workflow consisting of Lammps_DS execution—XRD Compute- SAED Parsing - Visit Visualization on
Gordon• Tested multiple production inputs for PI and Co-PI, with varying virtual
shared memory upto 760 GB over 16 nodes.
13
Integrating visualization
• Application output .saed converted to .xyz with sed script.
• Using OpenMPI build of VisIt (thanks to Amit Chourasia for building)
• VisIt launched with –s for Python script driver.• Driver checks if user uploaded a sessionfile,
uses RestoreSessionWithDifferentSources()• Otherwise scene rendered with default
threshold, clips and camera.
16
Visualization
• Reciprocal space of diffractions yields a point cloud in a spherical volume. Scalar threshold, clipped and color mapped. – Discovered multi-window sessionfile rendering bug (#1472), testing tiled
images instead
18
Next Phase in The ECSS Project
• Deploy same Workflow under apache Airavata-Xbaya• Deploy modified workflow to use stampede for computation
(Lammps_DS and XRD) and pipe data for VisIT visualization on Gordon
• Extend the infrastructure for VisIT session less and sessioned visualizations
• If time permits provide interactive VisIT steering by launching local VisIT application and contacting VisIT server at Gordon
19
Status of the Software• The diffraction code is not yet available in the full release of
LAMMPS• Similarly this implementation of the code is also not accessible to
community from SEAGrid Gateway• Details of some of the methods implemented are in the recent
paper Virtual diffraction analysis of Ni [0 1 0] symmetric tilt grain boundariesS P Coleman , D E Spearot and L Capolungo, Modelling Simul. Mater. Sci. Eng. 21 (2013) 055020 (16pp)• Please contact Prof. Doug Spearot ([email protected]) for more
information
20
Future Directions in Visualization
• (Plugin to load saed directly into VisIt)• Other rendering techniques – point sprites,
volume rendering• Computation, visualization on different
resources
21