From application to framework driven computing solutions:

1
From application to framework driven computing
solutions The present and future of TOF single
crystal diffraction at ISIS (a super user
perspective)
A. Daoud-Aladine M.J. Gutmann, L.C Chapon
PSD4C workshop, Paris, November 2008
2
Outline
Present the SXD2001 package
TOF Laue diffraction specificities Presentation
of SXD2001 Two typical success stories
Illustrative examples of typical
difficulties/limitations
Viewing/understanding 3D data coverage
Optimizing the data collection strategy for the
integration of magnetic Bragg intensities
Future solutions The Mantid project
The framework concept Plans for deployment on
Single crystal software
3
TOF Laue diffraction specificity
SXD
w
2D visualization (summed in TOF)
90(2)
sample
kf
37-det
90-det
37(3)
ki
1D visualization of single spectra

• - 11x(64x64) ZnS pixelated AREA detectors with
  each pixel recording a 1D 2000 bins TOF spectrum
• 1 Data set 70 Mbytes (compressed)
• 3D data sets in (x,z,time)

(0 -12 0)
4
TOF Laue diffraction specificity
Reciprocal space (Q)
Instrument view (kf)
TOF scan ? l scan ? Q scan
1 Data set 70 Mbytes (compressed)
2D visualization (summed in TOF)
Qkf(l)-ki
1/lmin
90-det
1/lmax
1D visualization of single spectra
90-det tilted down by 45
Picture made with Ewald3D, a Matlab utility.
Non-trivial Q-coverage due to 1) The
non-equatorial geometry 2) The goniometry 3)
Detector gaps
5
Data analysis software SXD2001
Before post data processing
A data reduction software
Bragg Scattering (integration)
Produces HKL lists for supported
packages SHELX,GSAS,FULLPROF JANA2000 (New,
JANA2006)
37-det
90-det
80-90 common tasks already present in SXD2001!
(0 -12 0)
Reciprocal space mapping (binning in Q,
hkl-space)
Binary/ASCII files for internal/external
visualisation
6
Outline
Present the SXD2001 package
TOF Laue diffraction specificities Presentation
of SXD2001 Two typical success stories
Illustrative examples of typical
difficulties/limitations
Viewing/understanding 3D data coverage
Optimizing the data collection strategy for the
integration of magnetic Bragg intensities
Future solutions The Mantid project
The framework concept Plans for deployment on
Single crystal software
7
A brief history of SXD2001

• ISIS raw data access encapsulated in so-called
  open GENIE dlls provided by the computing
  group
• Calculations, GUI, plotting 100 written in the
  IDL application language in order to satisfy 2
  essential user wishes
• Need for GUI-based applications
• Beneficiate from high level built in plotting
  tools
• Derived from (not portable) and historical
  Fortran/Vax code SXD98
• Much enhanced capabilities
• Complete analysis route from raw data to
  structure factors and/or reciprocal space
  exploration
• Stand-alone package
• Dynamically extended according to users wishes

M.J. Gutmann
8
SXD2001Peak integration
M.J. Gutmann

• Choice of three algorithms
• Shoebox
• Dynamic box
• 3D Gauss ellipsoid
• Information about resolution used
• Gives directly F2
• Propagation vectors can be used
• Manual integration
• Graphical diagnostic

9
SXD2001Reciprocal space sections
M.J. Gutmann

• Calculation of volumes in reciprocal space
• Flexible slicing
• Markers for space-group allowed, magnetic,
  incommensurate reflections
• Laue symmetry averaging
• Export to ASCII files
• Movie through volume

10
Outline
Present the SXD2001 package
TOF Laue diffraction specificities Presentation
of SXD2001 Two typical success stories
Illustrative examples of typical
difficulties/limitations
Viewing/understanding 3D data coverage
Optimizing the data collection strategy for the
integration of magnetic Bragg intensities
Future solutions The Mantid project
The framework concept Plans for deployment on
Single crystal software
11
Success story-I Proton migration in urea
phosphoric acid
Protonated
Old data/Old software
Recent data/SXD2001
5000 reflections
8200 reflections
Same result on deuterated urea phosphoric acid
C C Wilson, K Shankland N Shankland (2001). Z
Krist, 216, 303-306
C. Spanswick, C. R. Pulham, University of
Edinburgh A. Parkin, C. C. Wilson, University of
Glasgow
No planning here required for getting large
numbers of unique anonymous Bragg reflections
some necessary redundancy experiment made using
56 STANDARD goniometry settings
12
Success story-II Quantitative diffuse scattering
modeling
Rebinned data
Correct magnitudes of thermal displacements
arise from the fitting of the diffuse scattering
T. R. Welberry, D. J. Goossens et al. J. Appl.
Cryst. 36, 1440 (2003)
Scattering resulting from a theoretical distorted
atom configuration, obtained from a spring
model, whose force constants are refined against
the data
13
Outline
Present the SXD2001 package
TOF Laue diffraction specificities Presentation
of SXD2001 Two typical success stories
Illustrative examples of typical
difficulties/limitations
Viewing/understanding 3D data coverage
Optimizing the data collection strategy for the
integration of magnetic Bragg intensities
Future solutions The Mantid project
The framework concept Plans for deployment on
Single crystal software
14
Viewing/understanding 3D data coverage
One single crystal orientation (say, w10)
2D-sliced representation of the coverage of the
(b,c) plane
3D schematic representation
Goniometry settings
Choice of plane
15
Viewing/understanding 3D data coverage
TWO single crystal orientation (say, w10, and
w-30)
Data merged indistinctly
Recall one orientation result
Det 4
Det 4
Det 3
Det 3
?
Det 5
Det 5
16
Viewing/understanding 3D data coverage
TWO single crystal orientation (say, w10, and
w-30)
Info recovered only in 3D view
Data merged indistinctly
Det 4
3
4
Det 3
6
Det 6
5
Det 5
17
Outline
Present the SXD2001 package
TOF Laue diffraction specificities Presentation
of SXD2001 Two typical success stories
Illustrative examples of typical
difficulties/limitations
Viewing/understanding 3D data coverage
Optimizing the data collection strategy for the
integration of magnetic Bragg intensities
Future solutions The Mantid project
The framework concept Plans for deployment on
Single crystal software
18
The reverse problem
How many and which GONIO setting(s)? Instrument
coords?
w4
w1
w2
w3
From one HKL
?
PEAKSEARCH
Find CELL
w1
Next Run/ Orientiation
Refine UB with more reflections
Goniometer angles setting ltgt 1 data set
w2
w3
Absorption/extinction corrections
Separate outputs
Improve ext/abs as model improves
Export to GSAS etc.
Export to ASCII files
Merged HKL Lists with redondancy ie. for hkls
measured more than once Iobs(w1,l1) ? Iobs(w2,l2)
Refinement in GSAS, SHELX, etc
Diffuse modelling
Data merged indistinctly
UB/goniometry Info lost!
Visualisation of structure
ORTEP etc
Each HKL, each Q-point, can be in several data
corresponding to different orientations
19
EX Magnetic diffraction
Lots of holes of coverage at low Qs
TbMn2O5 at T30K
To solve the (T30K) magnetic structure, a
simulated annealing algorithm was used with
single-crystal data to search for starting
configurations. The phases were then fixed while
the magnitude and directions of the moments were
refined by Rietveld analysis of the final model
on powder data (Fig. 4), which are more complete
at low momentum transfer
k(1/2 0 1/4)
k(1/2 0 1/4)

Standard settings w-150 w-130 w-110 w-90
w-70 w-50
L. Chapon et al. PRL(2004)
20
EX Magnetic diffraction
TbMn2O5 at T27K
HKL list with Propagation vectors can be
used (FullProf Convention)
k(1/2 0 1/4)
k(1/2 0 1/4)

YMn2O5 at T27K
Standard settings w-150 w-130 w-110 w-90
w-70 w-50
Rnuc 10 (not shown) Rmag 20
w-149 w -89 w -54 w -29 w 4 w 97
170 reflections Igt3s
L. Chapon et al. PRL(2004)
A. Daoud-Aladine, M. Gutmann, L. Chapon, P. G.
Radaelli
21
Outline
Present the SXD2001 package
TOF Laue diffraction specificities Presentation
of SXD2001 Two typical success stories
Illustrative examples of typical
difficulties/limitations
Viewing/understanding 3D data coverage
Optimizing the data collection strategy for the
integration of magnetic Bragg intensities
(reverse problem hkl gt angles)
Future solutions The Mantid project
The framework concept Plans for deployment on
Single crystal software
22
MantidManipulation and Analysis Toolkit for ISIS
data

• Nick Draper
• 05/11/2008

23
Project Aims

• Aims
• To provide a framework for Data Analysis that is
  not instrument or technique/dependent.
• Support multiple target platforms (Windows,
  Linux).
• Easily extensible by Instruments
  Scientists/Users.
• Freely redistributable to visiting scientists.
• Provide low-level functionalities for
• Scripting
• Visualization
• Data transformation
• Implementing Algorithms
• Virtual Instrument Geometry

24
Architectural Design - Overview
Instrument log files
MantidScript Command line Scripting interface
Mantid Framework
RAW data files
API
Workspaces
Algorithms
MantidPlot Graphing and analysis
NEXUS data files
Matlab Interface
DAE direct access
Future Instrument specific UI
Future Instrument specific UI
25
Geometry

• Problem How to maintain visualise an accurate
  and fast representation of complex objects

• Mesh Based
• Easy, fast visualization
• Poor computational accuracy performance
• Hard to define complex shapes

• Surface Based
• Hard, slow visualization
• Good computational accuracy performance
• Easy to define objects using CSG

26
Geometry

• Constructive Solid Geometry
• Building of complex shapes fromintersections,
  unions and differencesof common primitives
• Easily understood by users

• Hybrid geometric model

GNU Triangulated Surface Library

• Surface CSG
• Used for
• Definition
• Calculations

• Mesh
• Used for
• Visualization
• Approximation

27
Instrument Visualisation
Instrument.exe
28
User Extensible

• Doesnt have the algorithm you need?
• Add it yourself

Plugin.exe
29
Outline
Present the SXD2001 package
TOF Laue diffraction specificities Presentation
of SXD2001 Two typical success stories
Illustrative examples of typical
difficulties/limitations
Viewing/understanding 3D data coverage
Optimizing the data collection strategy for the
integration of magnetic Bragg intensities
(reverse problem hkl gt angles)
Future solutions The Mantid project
The framework concept Plans for deployment on
Single crystal software peak search done!
30
The data analysis flowchart
Indexation Semi interactive Single dataset
treatment gt Creates an indexation
file containing refined UBs for one ore more
than one crystals and the goniometer angles for
each raw file
PEAKSEARCH
Find CELL
Next Run/ Orientiation
Refine UB with more reflections
Integration/binning Data treatment
sequentially done using many raw files and
indexation files
Absorption/extinction corrections
Improve ext/abs as model improves
Export to GSAS etc.
Export to ASCII files
Refinement in GSAS, SHELX, etc
Diffuse modelling
Limit of the SXD black box
Visualisation of structure
ORTEP etc