•
•
•
•
•
•
Atmospheric Science
Bioinformatics
Computational Chemistry
Crash Analysis/Simulation
Generalized Finite Element Method
Physics
Mission
• Our mission is to support
(1) Large-scale computing for research and
(2) Computational science research and
instruction
…by providing
(1) High-performance hardware,
(2) Software, and
(3) Support Services
Our Facility Online
http://sc.tamu.edu
Email: help@sc.tamu.edu
Services
•
•
•
•
•
•
Hardware
Software
Help Desk
Short Courses
Scientific Visualization
Accounts
Hardware
• IBM Regatta p690 (agave)
•
•
•
•
32-CPUs, 64 GB Ram, 1.1 TB Disk
SGI Origin 2000 (titan)
32 CPUs, 8 GB RAM, 300 GB Disk
SGI Origin 3800 (k2)
48 CPUs, 48 GB RAM, 1.2 TB Disk
EMASS AML/J w/SGI Archive Host
Amass/DataMgr HSM, 7 TB, 4 DLT Drives
9 Linux Workstations
Professor Lee Panetta
Department of Atmospheric Sciences
Atmospheric Sciences, Panetta
Overview
The nature of cross-jet transport is of interest to the
chemistry of the stratosphere and the biology of the upper
extratropical oceans. Here I use a combination of pseudospectral and particle tracking methods to investigate the
nature of transport across self-organized jets in rapidly
rotating stratified flow. Simulations indicate the presence
of an anomalous, subdiffusive scaling regime for single
particle dispersion which is intermediate between the
short-time “ballistic” and long-time “diffusive” regimes.
The regime is seen over a range of forcing strengths, but a
physically based scaling can be chosen which collapses
results to a single dispersion curve.
Atmospheric Sciences, Panetta
Mathematical Model
The two-layer system consists of two horizontal fluid layers, bounded above and below by a rigid horizontal
surface, and separated by an immiscible interface. The layers have slightly different densities, with the
denser (cold) fluid beneath the lighter (warm) fluid.
Key variables in the theory are the layer "potential vorticities" variables Q i, defined in terms of the nondimensional streamfunction Yi in layer i, by Qi = by + 2 Yi + (-1)i (Y1 - Y2 / 2). The system we integrate
numerically governs the evolution of deviations Qi from a specified state with an interface having a spatially
uniform structure, a tilt upward to the north. This is effectively a spatially uniform thermal forcing. The
evolution equations are
Results shown here use a 512 x 512 grid in each layer.
Atmospheric Sciences, Panetta
Cross-Jet Transport in Geophysical Turbulence
After a spin-up period, the solution settles into a quasi-steady state with a sequence of
turbulent jets oriented in the east-west direction. The mean latitude of each jet is
surprisingly persistent, even though eddy-to-mean-flow kinetic energy ratios can be well in
excess of unity. The figure below shows the evolution in time of the x-averaged eastward
component of the wind in the upper layer for in one simulation. There are four steady jets
in the north, and alternately one and two in the south. Adjustment of domain width can
remove this transience, which is a "quantization" effect due to the presence of a
dynamically determined jet scale.
Atmospheric Sciences, Panetta
Cross-Jet Transport in Geophysical Turbulence
Instantaneous fields of potential vorticity show narrowly concentrated regions of tight gradient,
corresponding to eastward jets, and compact vortices which arise from waves which form on
the jets and break. Both spatial structures play important roles in transport.
Atmospheric Sciences, Panetta
Cross-Jet Transport in Geophysical Turbulence
The figure below shows an instantaneous potential vorticity field for the upper layer, and two
groups of tracers whose positions are indicated by asterisks. The tracers were released a short time
before on lines of constant latitude halfway between westerly (i.e. eastward) jets, in regions of
relatively weak potential vorticity gradient (compare previous figure). Also indicated by arrows
on the side are the average positions of the nearest westerly jets.
Atmospheric Sciences, Panetta
Cross-Jet Transport in Geophysical Turbulence
The figure below shows a selection of four
tracer trajectories, with initial release
points indicated by asterisks. The dotted
lines indicate the narrow regions of (time
averaged) high gradients of potential
vorticity, which are the regions of strong
eastward jets. Two of the tracers shown
started in these regions, and were cast out,
and two started outside these regions. One
of the latter (green trajectory) actually
crossed a nearby jet. The animation
provided below shows how this jet
crossing occurs, namely in a wavebreaking event that results in a vortex
being formed. The tracer is initially carried
along by the vortex.
The statistics of such "cross-jet transport"
are of our principal interest. To study this,
information from repeated releases of
groups of tracers of the sort shown in Fig 3
was analyzed.
Atmospheric Sciences, Panetta
Cross-Jet Transport in Geophysical Turbulence
Results on single particle
dispersion rates from five
different choices of B are
summarized in Fig 5 below.
Smaller values of B
correspond to flows more
strongly driven. The
sampling strategy was that
for each value of B, 10
groups of 1024 tracers were
released halfway between
mean westerly jet positions.
To account for the difference
in energy levels in the flows,
distances were rescaled by
interjet spacing distance.
Atmospheric Sciences, Panetta
Animation of Results
The animation shows a number of features (see the first frame below):
•
•
•
•
The labels on the side are nondimensional length units in
the x (eastward) and y (northward) directions;
Colors indicate potential vorticity, with high values
indicated by red and low values by blue. The sharp
gradation between yellow and red, and between yellow
and blue, correspond to cores of two eastward jets;
The dashed curves are lines of constant streamfunction
values. Each jet is seen to be part of a street-like array;
eddies of locally high streamfunction values are the south,
and locally low values are to the north (see the "H" and
"L");
There are different symbols indicating positions (see
arrow) of the three tracers released in the flow at
nondimensional time "t=0". As time evolves these tracers
quickly become widely separated, and the tracer marked
by the asterisk is seen in the animation to actually cross
the southern jet in a wave-breaking event.
Atmospheric Sciences, Panetta
QuickTime™ and a Cinepak decompressor are needed to see this picture.
Professor David L. Adelson
Clare A. Gill
Bioinformatics, Adelson and Gill
Current projects
• Our most pressing project at present is our
contribution of bovine BAC clones to the
Bovine Genome Project.
• We have sent 50,000 BACs to Marco
Marra’s group for fingerprinting.
• We need to end sequence these clones as
well (bidirectional).
Bioinformatics, Adelson and Gill
For this we require laboratory automation
Biorobot
for DNA
isolations
High
throughput
PCR machine
96 capillary
DNA
Sequencer
Bioinformatics, Adelson and Gill
Automation consequences
• Sample tracking:
– Labeling needs to be automated.
– Sample sheets need to be automated.
• Results need to be collected and entered into a database.
• With over 500 sequences collected every day, an analysis
pipeline needs to be in place and at least semi-automated.
• Database needs to be integrated to manage not only DNA
preps and sequences, but phenotypes, genotypes and
BLAST results.
Bioinformatics, Adelson and Gill
Work flow for automated sequencing
• Select 384 well plate for growth.
• Automatically generate MegaBACE sample sheets (4x96)
and transfer over network to MegaBACE.
• Inoculate 4x96 well grow boxes and bar code.
• Grow clones.
• Robotic DNA isolation (bar code DNA plates).
• Robotic DNA sequencing reaction set up (bar code).
• Cycle sequencing
• Robotic sequencing clean up (bar code).
• Load sequencer (bar code reader).
• Auto file transfer to sequence analysis server.
Bioinformatics, Adelson and Gill
Sequence data pipeline
Trace files
(sequencer)
Tab delimited
parsed output
(LINUX box)
phred
(LINUX box)
MySQL table
(LINUX box)
HT-BLAST
SGI Origin 3800
48 cpu supercomputer
Web server
(LINUX box)
Bioinformatics, Adelson and Gill
SGI Origion 3800
• High throughput BLAST1,2 requires a
multiprocessor machine with large, shared
memory for maximum speed of search.
• BLAST parallelizes and scales well, allowing the
data pipeline to keep up with the new data
generated and update search results for previously
generated data.
1Altschul,
Stephen F., Thomas L. Madden, Alejandro A. Schaffer, Jinghui Zhang, Zheng Zhang, Webb
Miller, and David J.Lipman (1997), "Gapped BLAST and PSI-BLAST: a new generation of protein
database search programs", Nucleic Acids Res. 25:3389-3402.
2Camp, N., Cofer, H., and Gomperts, R. 1998. High-Throughput BLAST.
http://www.sgi.com/industries/sciences/chembio/resources/papers/HTBlast/HT_Whitepaper.html
Ref Type: Electronic Citation.
Bioinformatics, Adelson and Gill
Database structure
• Multiple databases at present, not a single
integrated database.
• Angleton data in one database
• Sequence similarity in another db.
• Sensory panel data (meat science) in
another db.
Bioinformatics, Adelson and Gill
Family database
Bioinformatics, Adelson and Gill
Meat sensory data
Bioinformatics, Adelson and Gill
Sequence similarity db
Bioinformatics, Adelson and Gill
In silico mapping
• We get two kinds of information back from
sequence similarity searches
– Location of homolog in reference genome.
– Functional properties of homolog.
• We would like to be able to represent both types of
information simultaneously.
• For now we can only provide the physical location
of the homolog in the reference genome (usually
the human genome).
Bioinformatics, Adelson and Gill
Overall db diagram
Texas A&M University
Department of Chemistry
Director: Prof. Michael B. Hall
Manager: Lisa M. Thomson
Contributors:
M. B. Hall
L. M. Thomson
J. D. Hoefelmeyer
F. Gabbaï
C. E. Webster
D. H. Russell
H. A. Sawyer
G. F. Verbeck
http://www.chem.tamu.edu/LMS
Computational Chemistry, Hall et al.
LMS Resources
• HARDWARE - All computationally intensive calculations are carried out
on the Texas A&M Supercomputing Facility’s systems:
– 32-cpu IBM Regatta p690
– 32-cpu SGI Origin 2000
– 48-cpu SGI Origin 3800
• SOFTWARE - The LMS uses a variety of molecular modeling software.
This software includes the following:
– Materials Studio
– AMPAC
– Molden
– Cerius2
– MOLPRO
– CHARMm
– Q-Chem
– Dalton
– Quanta
– Gaussian 98 (G98)
– Spock
– GaussView
– TINKER
– Insight II
– MacroModel
Computational Chemistry, Hall et al.
A Theoretical Study of the Primary Oxo Transfer Reaction of a Dioxo
Molybdenum(VI) Compound with Imine Thiolate Chelating Ligands: A
Molybdenum Oxotransferase Analogue (Thomson and Hall)
Molybdenum containing enzymes are a broad class of enzymes that are essential for the metabolism of carbon,
nitrogen and sulfur in a wide variety of organisms. In humans, sulfite oxidase is the enzyme responsible for the
metabolism of the toxin sulfite to sulfate. Analogue reaction systems have been developed to mimic the activity of
the molybdoenzymes. These analogue systems can be used to verify experimental data on the structure and
reaction mechanism of the complex enzyme systems. This study focuses on the elucidation of the reaction
mechanism of an analogue system.
Density functional calculations on MoO2(NHCHCH2SH)2 + P(CH3)3  MoO(NHCHCH2SH)2 + OP(CH3)3
were performed at the B3P86 level of theory as implemented in Gaussian 94/98, using a double- quality basis set
for all atoms and the inclusion of a polarization function on the phosphorus. The DFT results indicate that this
reaction proceeds through a two step mechanism via an associative intermediate shown in the Figure A. The
substrate was found to attack one of the terminal oxo groups to form an unusual 3c-4e- O-P-C bond in the first
transition state, TSI. The OP(CH3)3 group then rotates to almost lie in the MoO2 plane to form the intermediate,
INT. The second transition state, TSII, involves the weakening of the Mo-OP(CH3)3 bond and the concomitant
rearrangement of the ligands. Figure B shows an important anti-bonding interaction the help to eliminate the
product, OP(CH3)3. The overall exothermicity of this reaction is 32.7 kcal/mol (-Ho) and Go = -27.1 kcal/mol, a
value consistent with the equilibrium lying far to the right. The H‡ for the first step (rate determining) was found
to be 9.4 kcal/mol, and the second step had a H‡ = 3.3 kcal/mol. These results are within the uncertainty of the
experimental system, for which the rate determining H‡ = 9.6(6).
Computational Chemistry, Hall et al.
Figure A. The B3P86 results for the reaction of MoO2(NHCHCH2SH)2 +
P(CH3)3  MoO(NHCHCH2SH)2 + OP(CH3)3. These results indicate that this
reaction proceeds through a two step mechanism via an associative intermediate.
Computational Chemistry, Hall et al.
Figure B. 0.05 isodensity surface of a) REAC, b) TS1 and c) INT. (b) shows the
important anti-bonding interaction the help to eliminate the product, OP(CH3)3
Computational Chemistry, Hall et al.
An Intramolecular Boron-Boron One-Electron s-Bond
(Hoefelmeyer and Gabbaï)
Owing to their isoelectronic relationship to neutral methyl radicals, the chemistry of stable boron-centered radical
anions R3B•- (R=aryl rings) has been investigated intensely. Although delocalization of the radical over the aryl rings
accounts for the stability of such systems, EPR studies show that, in some instances, the unpaired electron is mainly
localized at boron. In organodiboranes, one-electron reduction leads to the formation of a one-electron s-bond formed
by the overlap of the parallel pz boron orbitals. Interestingly, the isolation of boron radicals in which the unpaired
electron occupies a molecular orbital formed by the combination of overlapping colinear atomic orbitals is much more
elusive. Motivated by the importance of stable radicals to the field of material science, we have set out to prepare a
stable boron radical, of the general form (R3B)2•-, and report on the formation of a radical that features a boron-boron
one electron -bond.
A single-crystal X-ray analysis of 1,8-bis(diphenylboryl)-naphthalene revealed the existence of a sterically
congested structure with a boron-boron distance of 3.002(2) Å (Figure A). A one-electron reduction of 1 affords the
radical anion 2 (Figure A). While it has so far not been possible to obtain single crystals of 2, we have performed a
series of DFT calculations on 1 and 2 with the B3LYP functional as implemented in Gaussian 98 ( Basis set: 6-31G on
C, and H, and 6-31+G* basis set on B). Examination of the B3LYP orbitals (Figure B) reveals that, in 1, the pz orbitals
of the neighboring boron centers overlap substantially and contribute to the Lowest Unoccupied Molecular Orbital
(LUMO). The calculated structure for 2 differs from that of 1 in several aspects, but most noteworthy is that the boronboron distance decreases substantially (3.16 Å in 1 to 2.82 Å in 2) in agreement with the presence of a bonding
interaction. Both boron atoms are the dominant contributors to the singly occupied Highest Occupied Molecular
Orbital (HOMO), which has a strong boron-boron -bond character (Figure B). This one-electron -bond can be
viewed as the occupation of the formerly vacant boron p z-orbitals upon one-electron reduction of 1. The minor
contributions of the ring carbon atoms substantiate the importance of the stabilizing effect provided by aryl
substituents in stable radicals.
Computational Chemistry, Hall et al.
An Intramolecular Boron-Boron One-Electron -Bond
(Hoefelmeyer and Gabbaï)
Figure A. X-ray crystal structure of 1,8bis(diphenylboryl)-naphthalene, 1, and B3P86
optimized structure of the reduced species, 2,
1,8-bis(diphenylboryl)-naphthalene anion.
Figure B. 0.05 isodensity surface for the
Lowest Unoccupied Molecular Orbital
(LUMO) of 1, and the Highest Occupied
Molecular Orbital (HOMO) for 2, illustrating
the overlap of pz orbitals of the neighboring
boron centers forming the strong boron-boron
-bond character in 2.
Computational Chemistry, Hall et al.
Theoretical Studies of Carbon-Hydrogen and Carbon-Carbon Bond Activation of
Cyclopropane by Cationic Ir(III) (Webster and Hall)
The reactions of cyclopropane with the coordinately unsaturated species produced by mild thermal activation
of [Cp*Ir(P(CH3)3)CH3]+L (L = Cl2CH2, OSO2CF3-) (shown in the scheme below) have been investigated with
density functional calculations (B3LYP). The pathway for the production of endo or exo 3-allyl complexes from
the reaction of cyclopropane with the IrIII model complex [CpIr(PH3)CH3]+ proceeds through C-H bond activated
IrV intermediates and CH4 elimination, followed by ring opening of the iridium cyclopropyl complexes through
an iridium carbene vinyl intermediate to their respective 3-allyl products. This unexpected mechanism breaks
two C-C bonds simultaneously and then re-forms one en route from the iridium cyclopropane complex to the
iridium allyl products. The interconversion between endo and exo 3-allyl can be assisted by solvent through an
1-allyl intermediate. Thermal rearrangement of the cyclopropyl kinetic product proceeds back through the same
s-agostic complex, producing the thermodynamically more stable metallocyclobutane complex.
+
+
+
-CH4
Ir
Ir
(H3C)3P
CH3
Ir
H
(H3C)3P
(H3C)3P
+
CH3
+
Ir
(H3C)3P
Ir
(H3C)3P
CH2
Computational Chemistry, Hall et al.
Theoretical Studies of Carbon-Hydrogen and Carbon-Carbon Bond Activation of
Cyclopropane by Cationic Ir(III) (Webster and Hall)
Schematic representation of the potential energy surface for the reaction of C 3H6 with (CpIrPH3CH3)+. Relative energies
are in kcal mol-1 and for structures 8 and 8' through 14 and 14' include the energy of CH4. Optimized structures can be
found on the following slides.
Computational Chemistry, Hall et al.
Theoretical Studies of Carbon-Hydrogen and Carbon-Carbon Bond Activation of
Cyclopropane by Cationic Ir(III) (Webster and Hall)
Optimized structures of the reactant, transition states, intermediates, and products for the
reaction of C3H6 with (CpIrPH3CH3)+.
Computational Chemistry, Hall et al.
Theoretical Studies of Carbon-Hydrogen and Carbon-Carbon Bond Activation of
Cyclopropane by Cationic Ir(III) (Webster and Hall)
Optimized structures of the
reactant, transition states,
intermediates, and
products for the reaction of
C3H6 with (CpIrPH3CH3)+.
Computational Chemistry, Hall et al.
Theoretical Studies of Carbon-Hydrogen and Carbon-Carbon Bond Activation of
Cyclopropane
by Cationic
(Webster
and Hall)
Conversion between
the Ir(III)
endo and
the exo-allyl
"flipping", exchanging
the position of only the
central carbon
"rotating", exchanging
the position of the two
end carbons
H
Ir
H
H3P
H
H3P
H
H
+10.0
H
Ir
formation of an 1-allyl
and allyl rotation
H
H
H
Ir H H
H
H3P
H
-5.5
H
H
-9.2
+49.3
-9.4
+33.8
~+30.
-39.3 kcal mol-1
endo
-39.3
-45.5 kcal mol-1
exo
Ir
H3P
Ir
H3P
-45.5
Reaction paths for the conversion
of endo and exo-allyl
Computational Chemistry, Hall et al.
Theoretical Studies of Carbon-Hydrogen and Carbon-Carbon Bond Activation of
Cyclopropane by Cationic Ir(III) (Webster and Hall)
Conversion between the endo and the exo-allyl
"rotating", exchanging
the position of the two
end carbons
formation of an 1-allyl
with coordinated ClCH3
H
H
H
Ir
H3P
H
H3CCl
H
Ir H H
H
PH3
H
H
-5.5
+17.0
+33.8
Ir
H3CCl
-39.3 kcal mol-1
endo
H3P
-45.5 kcal mol-1
exo
Ir
H3P
Ir
H3P
0.04 isodensity surface of Highest Occupied Molecular Orbital (HOMO) for the interconversion
of endo and exo-allyl via “rotating” . The energetic diagram illustrates that this is a high barrier
mechanism due to an increase in the energy of the HOMO.
Computational Chemistry, Hall et al.
Molecular Modeling and Ion Mobility Time-of-Flight Mass
Spectroscopy (Russell, Sawyer, Thomson, and Verbeck)
Dr. Russell’s group of the Laboratory for Biological Mass Spectrometry uses the
Supercomputing Facility in conjunction with the Laboratory for Molecular Simulation
(LMS), for predictive modeling of molecular ions drifting through a bath gas, usually
He, Ar, and N2. Experimental analysis is carried out using ion mobility time-of-flight
mass spectrometry.
Our first focus is on the separation of proteins and peptides due to conformational
differences in the drift tube (labeled A in the following figure). In order to accurately
analyze the peak profiles of the mobility spectra we use molecular
mechanics/dynamics to sample the conformational space of the peptides and then
calculate the energies of the different conformations using MOPAC calculations at the
semi-empirical (AM1) level of theory.
Our second focus is on the separation of of small organic molecules with the same
mass, but differ by electronic structure (labeled B). High-level ab initio calculations
are used to analyze the potential energy surface of small radical cation organic
molecules such that we can predict which radical cation reacts longer with the bath
gas.
Computational Chemistry, Hall et al.
Molecular Modeling and Ion Mobility Time-of-Flight Mass
Spectroscopy (Russell, Sawyer, Thomson, and Verbeck)
A
B
Professor Robert R. Lucchese
Department of Chemistry
Texas A&M University
Collaborators:
Ping Lin
Eric Stratmann
Alexandra Natalense
Robert Zurales
Shaleen Botting
Funding:
Welch Foundation
National Science Foundation
Texas A&M Supercomputer Facility
Computational Chemistry, Lucchese et al.
Molecular Photoionization
• When light of sufficient energy interacts with a molecule, a
photon can be absorbed leading to the ionization of the
molecule:
M  h  M   e
• The probability for ionization is proportional to the square
of the dipole matrix element which is an integral over the
wave functions
that represent the initial state, the final ion

state, and the photoelectron:
Computational Chemistry, Lucchese et al.
Scattering Equations
• The initial state, Yi, and the final ion state are, Yf, are
described using standard quantum chemistry techniques.
• The wave function for the photoelectron,
, is the
solution of a one-electron integral scattering equation:
• This equation is not solved directly, we instead use a
variational method to compute the required dipole matrix
elements.
Computational Chemistry, Lucchese et al.
Schwinger Variational Equations
• The dipole matrix element can be reduced to an integral
over the coordinates of a single electron which in the
bracket notation is:
• By expanding the wave function in a basis set, the matrix
elements can be approximated by the following Schwinger
variational matrix expression:
Computational Chemistry, Lucchese et al.
Single-Center Expansions
• All integrals are evaluated using single-center expansions
where each function is expanded as
Typical values: lmax = 60 for N2 and lmax = 120 for CS2
• With this expansion all three-dimensional integrals become
a sum over a set of radial integrals which are computed on
a radial grid:
R. E. Stratmann et al., J. Chem. Phys. 104, 8989 (1996), and references therein.
Computational Chemistry, Lucchese et al.
Two-Electron Integrals
• The most expensive computational step is the evaluation of
two-electron exchange operators:
• Repeated transformations between the partial-wave and
grid representations are used:
Computational Chemistry, Lucchese et al.
Molecular Frame Photoelectron
Angular Distributions (MFPADs)
• The most detailed information that can be measured and
computed is the molecular frame photoelectron angular
distribution (MFPAD).
• The MFPAD depends on the relative orientation of the
molecule and the polarization of the light (e. g. parallel,
perpendicular, or magic angle 54º)
• The MFPADs can be either represented as 3D plots or in
terms of four functions F00(qk), F20(qk), F21(qk), and
F22(qk), where qk is the angle between the molecular axis
and the direction of the photoelectron.
• Good agreement with experiment is only obtained when a
convolution with the apparatus function is performed.
R. R. Lucchese et al., Phys. Rev. A 65, 020702 (2002).
Computational Chemistry, Lucchese et al.
3D MFPAD in NO Photoionization
Leading to the c 3 State of NO+
Computational Chemistry, Lucchese et al.
MFPAD in NO Photoionization
Leading to the c 3 State of NO+
Dr. Akram Abu-Odeh
Center for Transportation Computational Mechanics
Texas Transportation Institute
Roadside Safety Applications, Abu Odeh
Research Activities at TTI
• Side Impact with Rigid Pole
• Side Impact with Slip-Base Luminaire Pole
Roadside Safety Applications, Abu Odeh
QuickTi me™ a nd a Cinep ak decompre ssor are n eede d to see thi s pi ctu re.
Roadside Safety Applications, Abu Odeh
QuickTime™ and a Cinepak decompressor are needed to see this picture.
Roadside Safety Applications, Abu Odeh
QuickTime™ and a Cinepak decompressor are needed to see this picture.
Roadside Safety Applications, Abu Odeh
Side Impact
QuickTi me™ a nd a Cinep ak decompre ssor are n eede d to see thi s pi ctu re.
QuickTime™ and a Cinepak decompressor are needed to see this picture.
Roadside Safety Applications, Abu Odeh
Crushing
Roadside Safety Applications, Abu Odeh
Total Force on Rigid Pole (35 km/hr)
0
-100
-150
-200
DYNA3D
Test
Time (seconds)
0.14
0.12
0.1
0.08
0.06
0.04
0.02
-250
0
Force (kN)
-50
Roadside Safety Applications, Abu Odeh
Total Force on Rigid Pole (50 km/hr )
-110
-190
-270
DYNA3D
Test
Time (seconds)
0.12
0.10
0.08
0.06
0.04
0.02
-350
0.00
Force (kN)
-30
Roadside Safety Applications, Abu Odeh
Slip-Away Base Luminaire Pole
Roadside Safety Applications, Abu Odeh
Slip-Away Base
Roadside Safety Applications, Abu Odeh
Slip-Away Base Plate
• Bolts are torqued to sustain certain
clamping force.
• Upper base moves upon impact.
• Lower base fixed to the ground.
• Bolts and nuts move out of the base.
Roadside Safety Applications, Abu Odeh
Slip-Base Impact Direction
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
Clamping Load Modeling
• Springs between the two base.
– Bolts, washers and nuts are not modeled.
• Springs between nut and bolt head.
– Make nut sliding on bolt shaft.
• Thermal loads.
– Drop bolt shaft temperature.
• Stress initialization.
– Apply initial stress in the bolt shaft.
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
Initial Load Magnitude?
• Higher Than the Sustained Clamping Force
• Washers, Bases and Nuts Are Loaded As
Well
• Some Sort of Relaxation Is Needed
• Thermal Initial Load Leads to Undesired
Bolt Shaft Transverse Deformation
• Higher Initial Spring Loads Lead to
Undesired Nut/bolt Head Deformation
Roadside Safety Applications, Abu Odeh
Stress Initialization
• INITIAL_STRESS_SOLID Command
• Applied at Every Integration Point for All
the Solid Elements in the Bolt Shaft
• Usually Reaches Sustained Clamping Force
Level in 8 Milliseconds
Roadside Safety Applications, Abu Odeh
Load Jump on Release
• Very detailed meshing of the nuts and slots
is needed.
• Localized yielding might play a factor.
• The model will be computationally
expensive.
Roadside Safety Applications, Abu Odeh
Sensitivity to Nut Chamfering
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
QuickTi me™ a nd a Cinep ak decompre ssor are n eede d to see thi s pi ctu re.
Roadside Safety Applications, Abu Odeh
QuickTime™ and a Cinepak decompressor are needed to see this picture.
Roadside Safety Applications, Abu Odeh
Roadside Safety Applications, Abu Odeh
QuickTi me™ a nd a Cinep ak decompre ssor are n eede d to see thi s pi ctu re.
Professor Strouboulis, Zhang and Babuska
Department of Aerospace Engineering
Generalized Finite Element Method: Strouboulis et al.
Overview
•
•
•
•
Goal of the research
Description of PUM
Description of meshless idea
Features of GFEM
–
–
–
–
Linear dependence
Solver
Integration
Special functions
• Performance of GFEM
– Numerically generated handbook functions
• Handbook-based GFEM
– Hierarchical handbooks
• Homogenization
Generalized Finite Element Method: Strouboulis et al.
Goal
• Use knowledge about the boundary value problem
• Avoid meshing complicated geometries
• Less degrees of freedom, more accuracy
Generalized Finite Element Method: Strouboulis et al.
Partition of Unity Method
Generalized Finite Element Method: Strouboulis et al.
Partition of Unity Method
Generalized Finite Element Method: Strouboulis et al.
Partition of Unity Method
Generalized Finite Element Method: Strouboulis et al.
Meshless Technique
• Computational mesh is obtained without considering the
geometry of the domain
Examples of GFEM mesh
Generalized Finite Element Method: Strouboulis et al.
Features of GFEM – Linear Dependence
Generalized Finite Element Method: Strouboulis et al.
Features of GFEM – Solver
Generalized Finite Element Method: Strouboulis et al.
Features of GFEM – Solver Performance
Generalized Finite Element Method: Strouboulis et al.
Features of GFEM – Integration
Generalized Finite Element Method: Strouboulis et al.
Features of GFEM – Integration
Generalized Finite Element Method: Strouboulis et al.
Features of GFEM – Integration
Generalized Finite Element Method: Strouboulis et al.
GFEM with Numerically Generated
Handbook Functions
Generalized Finite Element Method: Strouboulis et al.
Solutions of Handbook Problems
Generalized Finite Element Method: Strouboulis et al.
Usage of the Handbook Solutions in the GFEM
Generalized Finite Element Method: Strouboulis et al.
GFEM Solution Using Numerical
Handbook Functions
Generalized Finite Element Method: Strouboulis et al.
Bifurcated Crack: Handbook problem mesh
Generalized Finite Element Method: Strouboulis et al.
Bifurcated Crack: Handbook meshes on actual domain
Generalized Finite Element Method: Strouboulis et al.
Bifurcated Crack: Handbook problem solution
Generalized Finite Element Method: Strouboulis et al.
GFEM using Handbook Solution of Bifurcated Crack
Generalized Finite Element Method: Strouboulis et al.
Handbook-based GFEM
Generalized Finite Element Method: Strouboulis et al.
Hierarchical handbooks for more
complex problems
Generalized Finite Element Method: Strouboulis et al.
Illustration of hierarchical handbooks
(a) Mesh for the original
problem
(b) First level handbook
(c) Mesh for the first level
handbook
(d) Second level handbook
Generalized Finite Element Method: Strouboulis et al.
Overkill solution for Problem
The overkill mesh is
obtained by uniformly
refining the domain six
times. The overkill
solution is obtained by
using p = 5 and one
level handbook functions
with phb = 1, and the
handbook functions are
obtained by using p = 5
and pvoids = 1.
The figure shows the
overkill solution for m =
1.125.
Generalized Finite Element Method: Strouboulis et al.
Results for Problem
Professor Roland Allen, Dou, Dumitrica, Graves, Torralva
Department of Physics
Physics, Allen et al.
Overview
• The response of matter to ultra-fast and ultra-intense laser
pulses is a current frontier of science.
• New discoveries often result from the ability to explore a
new regime. Here one is exploring both extremely short
time scales (below one hundred femtoseconds) and
extremely high intensities (above one terawatt per square
centimeter).
• The usual approximations of theoretical physics and
chemistry break down under these conditions, and both
electrons and atoms exhibit new kinds of behavior.
Physics, Allen et al.
Nonthermal phase transition
in Ge (experiment)
• The above figure provides experimental evidence for an ultrafast and
nonthermal solid-to-liquid phase transition in Ge, followed by
recrystallization. These x-ray diffraction probe measurements
employed 1.54 Angstrom photons, following 100 fs, 800 nm pump
pulses. [After C. W. Siders et al., Science 286, 1340 (1999).]
Physics, Allen et al.
Nonthermal phase transition
in GaAs (simulation)
•
•
In this figure, the timedependent dielectric function is
shown for GaAs (the most
important compound
semiconductor) after it has been
subjected to a very short and
intense laser pulse.
The behavior here signals a
nonthermal phase transition, due
to destabilization of the atomic
bonding on a femtosecond time
scale.
[The work represented by this and all of the
following figures was supported by the Texas
A&M Supercomputing Facility, and was
performed by the group R. E. Allen, B. Torralva,
T. Dumitrica, J. S. Graves, R. Hamilton, Q. Gao,
S. Khosravi, A. Burzo, and Y. Dou in the Physics
Department.]
Physics, Allen et al.
Dielectric function shows electronic
structure
• Note three aspects in the behavior of
the dielectric function:
– (1) There is a loss of the original
structural features, signaling a loss of the
original tetrahedral bonding in the
semiconductor.
– (2) The imaginary part of the dielectric
function, which measures absorption,
becomes nonzero for photon energies
below the original band gap energy of
about 1.4 eV, In fact, one observes
metallic behavior, which demonstrates a
complete collapse of the band gap,
beyond about 250 fs.
–
(3) There is a "hump" which persists even after the band gap has collapsed, and which
appears to indicate that there are still bonding-to-antibonding transitions, even after the
long-range crystalline order has been lost.
Physics, Allen et al.
Nonlinear susceptibility shows atomic
structure
•
•
•
Time-dependent nonlinear
susceptibility for GaAs just above
the threshold laser intensity for a
nonthermal phase transition.
The nonlinear susceptibility probes
the atomic structure of the material,
and is thus complementary to the
linear dielectric function, which
probes the electronic structure.
Notice that the nonlinear
susceptibility falls to zero over the
entire range of photon energies,
signaling a loss of the original
symmetry of the GaAs lattice.
All these results of the theoretical group at Texas A&M are in agreement with the measurements of the experimental
groups at Harvard University, the University of Essen, Bell Labs, M. I. T., and the University of California at Berkeley.
Physics, Allen et al.
Photoisomerization in chemistry and
biology
• The primary process in vision is photoisomerization of
retinal molecules in the eye.
• We have not yet performed a simulation for this molecule,
but we have observed similar photoisomerization of the
simpler butadiene molecule.
[In our butadiene simulation, the photon energy was 2.0 eV, the
fluence 0.36 kJ per square meter, and the pulse duration 15 fs FWHM.]
Physics, Allen et al.
Experimental results of the M.I.T./Florida
group
• Excitation of the breathing mode
in C60 buckyballs by an ultrafast
laser pulse.
• This figure demonstrates coherent
phonon oscillations in K3 C60 at
300 K. The pump-probe data were
taken in reflectivity with a single
wavelength pump-probe setup
having a time resolution of about
20 fs. The larger inset shows the
Fourier transform power spectrum
with a sharp peak at 492.5 inverse
centimeters.
After Fleischer, Pevzner, Dougherty, Zeiger, Dresselhaus, Dresselhaus, Ippen, and Hebard,
Appl. Phys. Lett. 71, 2734 (1997).
Physics, Allen et al.
Experimental results of the Lawrence
Berkeley group
• Excitation of the breathing mode
in C60, and the pentagonal pinch
mode at higher frequency,
following an ultrafast laser pulse.
• The figure shows the lightinduced negative differential
transmittance of a C60 thin film
detected at 580 nm as a function
of time delay between 12 fs pump
and probe pulses centered on 620
nm. The Fourier power spectrum
of the oscillatory part of the
response is shown in the inset.
After Dexheimer, Mittleman, Schoenlein, Vareka, Xiang, Zettl, and Shank, in "Ultrafast Phenomena VIII",
edited by J. L. Martin, A. Migus, G. A. Mourou, and A. H. Zewail (Springer-Verlag, Berlin, 1993).
Physics, Allen et al.
Simulation at the Texas A&M
Supercomputer Center
• Our calculations resolve what
might appear to be a
discrepancy between the
results of the two experimental
groups, by demonstrating that
only the breathing mode is
seen at high laser intensity,
whereas both the breathing
mode and the pentagonal-pinch
mode are seen at lower
intensity.
[After B. Torralva, T. A. Niehaus, M. Elstner, S. Suhai,
Th. Frauenheim, and R. E. Allen, Physical Review B 64,
153105 (2001).]
Physics, Allen et al.
Photofragmentation of a C60 molecule
•
•
Following a 35 fs, 2.0 eV laser pulse
with a fluence of 1.17 kJ per square
meter
There are two particularly striking
features:
– The first is the dramatic opening of
both ends of the fullerene,
accompanied by the breaking of
many bonds.
– The second is the release of a dimer
(at t = 952 fs) with a kinetic energy
of 0.25 eV, which is slightly more
than half the value reported for the
experiments (Hohmann et al., Z.
Phys. D. 33, 143 (1995)).
•
Following emission of the dimer, the remaining 58 atoms tend to move back toward
one another and reform bonds with both hexagons and pentagons. However,
complete reformation of a closed structure is not observed, since there is still a large
population of electrons in excited states at the end of the simulation.
Physics, Allen et al.
Photofragmentation of a C60 molecule (higher
intensity laser pulse)
•
•
•
For a still higher fluence of 1.24 kJ
per square meter, we observe that
the cage begins to break apart at t =
299 fs, with the emission of both a
dimer and a trimer.
Experimentally, it is well established
that C60 should normally fragment
with the emission of even numbered
clusters. In the present case,
however, the cage continues to open
up, and at t = 998 fs another dimer is
emitted.
It is apparent that the remaining cluster is structurally unstable, and the atoms
continue to move apart. At t=2000 fs, two corannulene-like structures of 29 and 24
atoms are connected by only one bond. One therefore expects that that they will
eventually separate rather than reform as a stable odd-numbered cluster with greater
than 32 atoms.
Physics, Allen et al.
Collision of atom with C60
•
•
•
Here the first panel shows a
representative trajectory in a DFTED
(density-functional-based, tightbinding, electron-ion dynamics)
simulation: A 155 eV projectile carbon
atom impinges on a buckyball.
At t = 147 fs, we see that the backside
of the cage has already reformed, and
that two dimers have been released.
(The release of dimers in collisional
experiments with C60 has been widely
observed in experiments.) The
projectile atom is out of the picture at
this point, since it requires only 23 fs
to pass through the buckyball.
At t = 1500 fs, the remaining 56 atoms have joined to form a new closed-cage
structure, whose prominent features include the formation of a seven-membered
ring accompanied by a extra pentagon.
Physics, Allen et al.
Computer simulations show processes
in microscopic detail
•
•
•
Average displacement of GaAs
atoms from their equilibrium
positions for various intensities of
the applied laser pulse. The
nonthermal transition clearly
occurs above A = 1.50 gauss cm.
Many quantities like this can be
monitored during a simulation.
This illustrates the power of
computational science: Quantities that
are experimentally inaccessible can be studied in the simulations and used to
achieve further understanding. On the other hand, it is important to also calculate
quantities that can be compared with experiment, like the dielectric function and secondorder nonlinear susceptibility. [The full-width-at-half-maximum pulse duration is 70 fs,
and the complete pulse extends from 0 to 140 fs. The field intensity is indicated on the
upper left, with the amplitude measured in gauss cm.]
Physics, Allen et al.
Biological molecules are complicated
but within reach
• Our calculated spectrum
for chlorophyll, which
we also intend to treat in
the future.
• Since chlorophyll
absorbs strongly in the
red, with a secondary
absorption peak in the
blue, it is green by
reflected light.
Physics, Allen et al.
Photodissociation of a molecule
• Photodissociation
can be studied, as
when cyclobutane
dissociates to form
two ethylene
molecules in this
simulation.
Physics, Allen et al.
Photoisomerization
of a molecule
• One can observe the detailed dynamics during photoisomerization, as in
this cis-to-tran conversion of butadiene.
Physics, Allen et al.
Simulation of
symmetry-forbidden reaction
• Snapshots of a simulation of
two ethylene molecules
approaching each other, with
a kinetic energy of 0.2 eV. At
t = 43 fs, the two ethylenes
reach their closest point, with
a separation of only 1.63
Angstroms. However, they
bounce off each other rather
than bonding, and are 7.46
Angstroms apart at t = 81 fs.
Physics, Allen et al.
Toward laser control of chemical reactions
•
•
•
•
Snapshots of a simulation of the
approach of two ethylene
molecules following the
irradiation of the ethylene on the
right by a 5 fs, 5.5 eV laser pulse
with a fluence of 1.4 kJ per
square meter.
The laser-pulse interaction with
the ethylene molecule was
sufficient to change the
symmetry-forbidden reaction to
one that is symmetry-allowed.
I.e., molecules in excited states react differently from molecules in the ground
state, and electrons can be promoted to excited states in a selective way with
the use of tailored laser pulses.
In the bottom right panel, the structure has been equilibrated; all remaining
electrons occupying excited states have been forced into the bonding states,
and the excess kinetic energy has been removed. [The apparently missing
bonds in earlier frames are artifacts of the graphics program.]
Physics, Allen et al.
Photoisomerization of butadiene
Evolution of dihedral angles and
bond length. This is another
example of how processes can be
studied in detail at the atomic
level in computer simulations.
Physics, Allen et al.
Depopulation of excited states stabilizes new
structure
Evolution of highest occupied
molecular orbital, lowest
unoccupied molecular orbital,
and other states, in
photoisomerization of
butadiene. At several avoided
crossings, the electrons in
excited states are observed to
automatically undergo
transitions to lower-energy
states, reducing the electronic
energy of the molecule and
stabilizing its new structure.
Physics, Allen et al.
How the simulations are performed: ion and
electron dynamics
Physics, Allen et al.
How the simulations are performed: coupling of
electrons to laser pulse
Physics, Allen et al.
Dr. Allen’s Group
CCW from top right:
John (Trey) Graves
Traian Dumitrica
Ben Torralva
Yusheng Dou
Roland Allen
For more information,
go to
http://lightandmatter.net