software project brent fultz california institute of technology software functions full experiment...
Post on 15-Jan-2016
220 views
TRANSCRIPT
Software Project Brent Fultz
California Institute of Technology
• Software Functions
• Full Experiment Simulations
• Inversions of Dynamics Models
Discussions and Planning Meetings
• Sept. 24, 2001 meeting at Argonne with video to Oak Ridge, McMaster U., Caltech
• Nov. 29, 2001 small meeting at Caltech with postdoc interview.
• March 15, 2002 Software Workshop
Software Roadmap v. 1.0
Software Lifetime Exceeds Hardware Lifetime
main (){FILE *fp;/*MEM pair 7/19/89 B. Fultz
This program generates kinetic paths in a binaryalloy with B2 structure.The level of approximation in the kinetic equations isequivalent to the pair (Bethe) approximation in thermodynamics.
Note date above. Code runs just fine,migrated through workstations:austrl, ulysses, hyperfine1, hyperfine2, ARCS
Software Engineering
Reality: Cannot rewrite all codes in one language such as C++, Java or FORTRAN90
Need New Approach:• Modular software objects connected by object-oriented scripting language
• Different high-level scripts for different experimenters
• Span from detectors to dispersions on common software framework
• Python
Writing Python Bindings
For example, given a solver routine such as
and a wrapper
double adlib::StableTimeStep();
char pyadlib_stableTimestep__name__[] = “stableTimestep”;PyObject * pyaldib_stableTimestep(PyObject *, PyObject * args){ double dt = adlib::StableTimeStep(“deformation”);
return Py_BuildValue(“d”, dt);}
dt = pyadlib.stableTimestep()
one can place the result of the routine in a python variable
Software for Data Acquisition
Within 30 MinutesCalib. FilesOkay ?New RunQuickLookDataS(Q,E)
Within 10 MinutesOkay ?New Run(sample, instrument info.)
QuickLookData
Software for Data Acquisition
DataViewerDetector FilesMark Ω,tInstrumentSimulationSelect ,Q E
Within 2 HoursDiscovery! A feature in the data!Is it magnon, phonon, spurion?
Software for Data Analysis
Before Publication
Relate Data to Dynamics Model
DynamicsModel
Sim. S(Q,E)S(Q,E)DataSim. DataISAW++GraphicsCalc. 4Dfeedbackfeedback
Summary: ARCS Software
• Tools Are Required During Data Acquisition Users want software like a smart manservant (to take initiative, but with discretion, and no pay) ARCS needs an engineer to consult on the “user experience”
• Software for Data Analysis Enables New Science:
Full Experiment Simulations (connect to theory)
Analytical Inversions of Data
Single Crystal Data
S(Q,E) from TOF Data
• Bare minimum for users to take home
• General – model independent
• 3D single crystals, with real-time decisions on sample orientation
• Absolute units [barns, meV, sr, atom, and Å-1]
• Calibration
• Handle resolution and background
• Data visualization
Working with S(Q,E)
• Visualization
• Comparisons of data sets, arithmetic operations
• Analytical results from the theory of thermal neutron scattering by condensed matter
• Inversions of measured data to obtain force constants or exchange energies.
• Data mining — e.g., recognition of dispersions
Full Experiment Simulations
Primary Flight Path<kf|H|ki>Secondary Flight PathDetectorSourceSampleSimulationComparisonDataPhysicsModulesSample
ModulesConvergence
Modules
Space
Neutron Wave (or wavepacket)
TimeSpacer (t )1 0r (t )1 1r (t )1 2r (t )1 3r (t )1 4r (t )1 5r (t )2 0r (t )3 0r (t )4 0r (t )5 0
TimeSpacer (t )1 0r (t )1 1r (t )1 2r (t )1 3r (t )1 4r (t )1 5r (t )2 0r (t )3 0r (t )4 0r (t )5 0
EAM Simulations of Phonon DOS from Ni3Al
Large molecular dynamics model. Computationally challenging. Calculated:
PRB 57 (2): 862-869 JAN 1 1998
Scattering Simulations ofLocal Magnetic Dynamics in a Disordered System
Model Inversions — Incoherent Inelastic Scattering from 57Fe
400
300
200
100
0
Cou
nts
6040200-20-40-60
Energy (meV)
3 GPa 12 GPa 17 GPa
• Nuclear resonant scattering of x-rays by 57Fe at APS 3ID
• Pd357Fe in diamond-anvil
cell
• Same analysis as incoherent neutron scattering
• Invert data from 57Fe to obtain full lattice dynamics
Result of Inversion — Phonon Partial DOS Curves
Phonon Partial DOS Curves Under Pressure
0.4
0.3
0.2
0.1
0.0
Par
tial D
OS
403020100
Energy (meV)
0 GPa
3 GPa
17 GPa
Fe Pd
Result of Inversion — Radial 1nn Force Constants vs. Pressure
50
40
30
20
10
0
Rad
ial 1
nn F
orce
Con
stan
t (N
/m)
151050
Pressure (GPa)
Pd-Fe
Pd-Pd
L12 Pd3Fe
Note: Pd-Pd force constants from 57Fe scattering!
Fifty initial conditions for Pd3Fe
Try different numbers of nearest neighborsQuality of fit to experimental data
1nn Pd-Pd and Pd-Fe
Pd3Fe force constants from inversions
A case of bad 1nn Pd3Fe force constants
Future -- S(Q,E) inversions using coherent scattering
Inversions of incoherent inelastic scattering from 57Fe overcame the neutron-weighting problem in Fe alloys.
More information is available through the Q- or Q-dependence of coherent scattering.
S(Q,E) inversions using coherent scattering
Polycrystalline Ni3Al measured at HFIR with 4 values of Q(We were looking at order-disorder phenomena...)
Calculated incoherent scattering with Born-von Karman code.
Calculated coherent scattering from all orientations of crystallites w.r.t. Q
Coherent scattering was considered a nuisance to be overcome sowe could produce a phonon DOS.
S(Q,E) inversions using coherent scattering
Polycrystalline Ni3Al measured at HFIR with 4 values of Q
S(Q,E) inversions using coherent scattering
Isotropic average of S(Q,E) of polycrystalline Ce, a coherent scatterer.
1.5
1.0
0.5
0.03.02.52.01.51.00.50.0
Frequency (THz)
Q in A-1 3.121 3.335 3.613 3.777
Summary
• A roadmap exists for the ARCS software project, but it is still subject to change.
• Individual roads are modules connected by Python: S(Q,E) visualization fits to models
full experiment simulation
• Work is underway on phonon dynamics inversions incoherent scattering is useful coherent scattering should be even better
• The year 2002 will be an assessment of existing packages. Please let us know your views!