Reimaging Lens Polarization

1
The NCAR/HAO Community Spectro-Polarimetric
Analysis Center (CSAC)
Bruce W. Lites 303 497 1517 lites_at_ucar.edu
2
What is CSAC?

• Repository for analysis software for
  spectro-polarimetric data analysis
• Community involvement/community access
• Full range of analysis
• Calibration
• Inversions
• Ambiguity resolution
• Data visualization
• Goals
• Tested, transportable, documented code
• Conformation to modern software standards
• Computational efficiency

3
Why CSAC?

• Spectro-Polarimetric (SP) data has the highest
  information content
• Allows comprehensive, quantitative measures of B
• Permits extraction of line-of-sight gradients
• BUT
• SP Data is intrinsically more difficult to reduce
• Higher information content means more detailed
  analysis
• The problem of the past
• Analysis codes have been cumbersome, opaque, not
  easily ported to other systems, and not
  particularly well documented
• CSAC aims to remedy these problems

4
Why CSAC?
Widespread interest in spectro-polarimetry many
polarimeters are new and under-development.
5
Some CSAC Priorities

• Data Reduction Routines Develop software for
  flat-fielding, polarization calibration, merging,
  rectification, fringe removal, etc. to prepare
  data sets for subsequent inversion.
• Milne-Eddington Inversion This is the workhorse
  of analysis of SP data to extract the magnetic
  field vector (and other associated properties of
  the magnetized atmosphere).
• LILIA Inversion Develop standardized, portable
  software based upon the SIR (Stokes Inversion by
  Response functions) procedure. This technique
  allows for variation of parameters along the
  line-of-sight.
• Rapid Inversion Techniques New techniques such
  as principal components analysis, neural
  networks, support vector machines offer
  meaningful inversions at a very large increase in
  speed.
• Ambiguity Resolution CSAC will serve codes for
  resolution of the 180º azimuth ambiguity.
• Data Visualization The AZAM utility, as well as
  other methods for visualization, will be
  maintained by CSAC.

6
CSAC Accomplishments

• Hosted First Ambuiguity Resolution Workshop
  (September 2005)
• Co-Hosted and Co-Sponsored Second Ambiguity
  Workshop in October 2006
• Completed development of new GRID Milne-Eddington
  inversion code

7
DATA REDUCTION
8
CSAC Data Reduction Routines

• Data Reduction is
• Determination of Dark- and flat-field corrections
• Rectification of spectra (i.e. align dispersion
  along a pixel row)
• Removal of spectral curvature
• Polarimetric calibration over the spectral,
  spatial field-of-view
• Merging data from Dual-Beam (Beam Exchange)
  polarimeters

9
CSAC Data Reduction Routines

• CSAC has developed codes for data reduction for
  several instruments (DLSP, Swedish SP, and now
  HINODE)
• Procedures both in IDL and FORTRAN
• FORTRAN routines much faster, allow for real-time
  processing and calibration
• Data-specific parameters external to the code
• Simplification of the calling process relieves
  the user of complex data processing sequences
• Commonality among instruments of processed data
  structure

10
INVERSIONS
11
A More Comprehensive List of CSAC Inversion
Methods
12
MERLIN Milne-Eddington Inversion Code

• Designed for parallel processing via GRID
  architecture
• Unlike many problems, the inversion of Stokes
  data consists of many separate but identical
  computational tasks that require no interaction
  among them
• This is known as a scatter-gather computing
  problem, amenable to GRID computation
• The GRID consists of a heterogeneous array of
  loosely interconnected individual nodes sharing
  common resources
• In our case the interconnect is via a network
  (LAN, or WWW)
• Code is named Milne-Eddington gRid Linear
  Inversion Network (MERLIN)

13
Inversion Components

• MERLIN serves as a model for subsequent, more
  sophisticated CSAC inversion codes. It consists
  of 4 components
• The Inversion Kernel (IK) A set of libraries of
  codes (mainly C, C) to perform the actual
  inversion computations
• The Grid Inversion Server (GIS) An executable,
  run on the separate nodes of the Grid, that
  listens for incoming commands from the Client to
  run inversions, then run the IK libraries for
  such commands
• The Grid data server (GDS) An executable that
  accesses requests for slices of a data cube, then
  serve them to the GIS. If accessed locally, it
  is a UNIX library. If remote, uses an OPEnDAP
  server.
• The Grid client (GC) A web server application
  that allows the user to initiate inversions of
  either local or remote data sets.

14
Stokes Data Inversion

• Typical SP data is 4-dimensional
• Spatial slit scan direction (x)
• Spatial dimension along slit (y)
• Wavelength (?)
• Polarization (I,Q,U,V)

15
Data Passed by GDS to GIS

• The individual unit of inversion is the set of
  Stokes spectra (at right)

• One or more of these may be passed to the GIS

• The GDS accesses the entire data volume, and
  selects the requested slice, then passes it to
  the various GIS

• Typically, each GIS receives data from one slit
  position (one x-position, all y-positions)

16
Grid Topology

• Example
• 2 local GIS running under Unix sockets
• 2 remote GIS running under TCP/IP
• 1 local GDS running under NFS
• 1 remote GDS running under TCP/IP

• If data sits at llnl.gov all 4 GIS will be used
• If data sits at ucar.edu only 2 local GIS may be
  used to access these data

17
Example of Grid Operations
The Client is a web server. For highest
efficiency it resides in proximity to the Grid
Inversion Servers so that it may communicate
rapidly and receive results.
18
Standards for the CSAC Library of Analysis Tools

• Codes are highly transportable (written in C,
  C), callable from IDL
• Supported under Linux, Solaris
• Efficient coding, appropriate for parallel
  architecture
• Well documented, commented, and tested
• Flexibility to accommodate data from a wide
  variety of instruments
• Standardized input/output
• Standards for presentation in solar coordinates
• Filters provided to convert input data from major
  instruments (HINODE, DLSP, SOLIS, etc.)
• Codes maintained at a HAO/NCAR
• Examples of input/output data provided
• Open source for user modification,
  experimentation, and community input
• Online access to all analysis tools
• User forum for suggested modifications, additions

19
Comparison of Inversion Results, MERLIN and ASP
(Data from DLSP)
Continuum
B, ASP Code
B, MERLIN Code
0ltBlt3000 G
0ltBlt3000 G
20
MERLIN Improves Fit at Every Pixel
(BMERLIN - BASP)
?2MERLIN / ?2ASP lt 1
100 Gauss
21
Advanced Inversion Methods
Advanced inversion techniques allow extraction of
the field vector from the photosphere into the
chromosphere from simultaneous measurements of
photospheric lines and the Ca II IRT lines
Measured electric current density at heights 200,
650, and 1600 km Socas-Navarro 2005, ApJ 633,
L57
22
DATA VISUALIZATION
23
The AZAM Utility
24
The AZAM Utility

• Interactive, manual resolution of the 180º
  azimuth ambiguity
• Flexible display of inversion parameters
• Color images
• Arrows
• Blinking images against one another
• Contour plots
• Interactive display of data images, spectral
  profiles,
• And much more

25
CSAC Outlook for the Coming Year

• Finish MERLIN Milne-Eddington Inversion, prepare
  for analysis of HINODE data
• Implement Artificial Neural Network
  initialization for MERLIN
• Implement GUI
• Integrate into HINODE database
• Implement LILIA detailed inversion
• Generalize AZAM to accept data from MERLIN (i.e.,
  from any data source)
• Implement the simulated annealing azimuth
  ambiguity resolution for automatic processing
• Host a community workshop to address community
  needs
• Sponsor graduate student visits