Wilkinson Microwave Anisotropy Probe
This is the software used by the WMAP team to compute Fisher and Master matrices, and to compute the likelihoods of various
models.
This software is written in Fortran 90; it has been tested with the SGI MIPSpro,
NAG, and Intel Fortran 90 compilers. It requires the following libraries:
- The cfitsio library:
http://heasarc.nasa.gov/fitsio/fitsio.html
- A set of Cholesky factorization routines. On the SGI systems, the SCSL
Science Library contains the relevant routines; the LAPACK library can provide
the routines on other systems.
Two tarballs are supplied.
- The first contains all the supporting data the WMAP team used. It includes 'standard-cleaned' TB/EB data and
Gibbs polarization cleaned data. The compressed file is 988MB in size, and it decompresses to 1705MB.
- The third contains only the software; no supporting data is included, limiting its usefulness but resulting in
a MUCH smaller file. The compressed file is 35kB in size, and it decompresses to 180kB.
To build it, you should edit the Makefile to support your current environment;
the provided Makefile shows several possible configurations; these configurations
have been tested.
Then:
make # Builds the likelihood software library.
make check # Builds and the test program, and diff's the output
# to the correct results.
Output differences of ~0.001 between platforms are normal.
A patch was provided to LAMBDA by Ted Kisner of the Lawrence Berkeley National Laboratory
to allow the data directory to be identified through an environment variable. To install,
retrieve the patch tarball from the download page, unpack it, and copy the patch file to the
top level likelihood software directory. Execute the command:
patch -p1 patch_wmap_likelihood_p3v1
Then follow the build instructions above.
Modification History:
- 2006-Mar-16 - Version 2.0 - Initial release.
- 2006-Mar-17 - Version 2.1 - Additional error reporting routines have been added to
provide safe exit from the likelihood routine in the
event of errors.
- 2006-Oct-16 - Version 2.2 - The main new features are:
- The ability to evaluate the TT likelihood in pixel space up to l=30.
- The point source correction was reduced by 18% and its uncertainty doubled.
- The HEALPix library is no longer required to build this software.
- 2006-Oct-18 - Version 2.2.1 - Failures in the beam error module are now treated as a warning
instead of an error. Warnings are not fatal to a processing
run; instead the beam correction is set to zero, the problem
is reported, and processing continues.
- 2006-Nov-24 - Version 2.2.2 - This version includes a faster low-l TT likelihood module,
suggested by Jon Sievers of CITA. It used signal-to-noise eignemode compression to reduce the
dimensionality of the covariance matrices, while retaining numerical precision. It should produce
numerically equivalent results to previous releases of version 2.2. This version also fixes some
memory allocation and compiler-related problems. More details concerning the speed improvement
may be found in http://lambda.gsfc.nasa.gov/product/map/dr2/likelihood_faster_v2p2p2/
- 2008-Mar - Version 3.0 - The 5-year data release. Note that this version has
two options for computing the low-l TT likelihood (pixel and gibbs),
and an option to compute the likelihood for TB and EB spectra.
- 2008-Oct - Version 3.1 - A data directory identification patch was provided by T. Kisner of the
Lawrence Berkeley National Laboratory.
- 2009-Mar - Version 3.2 - Software optimization was improved through the reordering of many nested DO
loops throughout the software, as suggested by Raphael Flauger. The 2008-Oct
patch has not yet been integrated.
- 2010-Jan - Version 4.0 - The 7-year data initial release.
- 2010-May - Version 4.1 - Correction of two small bugs affecting high-l TT, TE, TB, and point
source power spectra.
See "CHANGES" for more details describing each release.
Acknowledgements:
We thank the following individuals for useful feedback on previous versions of this code:
- Rachel Bean
- Olivier Dore
- Hans Kristian Eriksen
- Raphael Flauger
- Kevin Huffenberger
- Antony Lewis
- Hiranya Peiris
- Jon Sievers
- Licia Verde
Tarball contents
The tarball contains a data subdirectory that contains various input files.
These input files are either ASCII text files containing columns of data,
or FITS files containing binary data.
The tarball also contains the following files:
- README.txt
A descriptive file.
- test.f90
A wrapper program shows you how to call the likelihood code and allow you to
run and test it. For the test data provided you should get the output stored in
output.std so you may compare your output to this with a simple diff command.
- WMAP_7yr_likelihood.F90
This is the central likelihood routine from which others are called.
- MASTER codes for TTTT (l=2-1000) and TETE (l=2-450) + a log determinant term for TETE
- The option of using a pixel based likelihood for TT 2≤l≤30 instead of MASTER
- The option of using a gibbs-sampling likelihood for TT 2≤l≤32 instead of MASTER
- The option of using a pixel based likelihood for TE, EE and BB 2≤l≤24 (this
substitutes for MASTER TETE 2<l<24)
- The option of using a pixel based likelihood for TE, TB, EE, BB and EB 2≤l≤24 (this
substitutes for MASTER TETE 2<l<24).
- The option of using a master likelihood for TB (l=2-450) + a log determinant term for TBTB
- wmap_7yr_options.F90
The options/parameters for use in the code are all contained the WMAP_options module; a sample
is shown below.
- WMAP_7yr_teeebb_pixlike.F90
The low l TE/EE/BB combined pixel based analysis for l=2-23
- WMAP_7yr_tetbeebbeb_pixlike.F90
The low l TE/TB/EE/BB/EB combined pixel based analysis for l=2-23
- WMAP_7yr_tt_beam_ptsrc_chisq.f90
Extra chisq corrections in TTTT from the beams and point source uncertainties.
- WMAP_7yr_gibbs.F90
The low l TT gibbs-sampling likelihood code for l≤32.
- WMAP_7yr_tt_pixlike.F90
The low l TT pixel based likelihood code for l≤30.
- wmap_7yr_util.f90
IO tools and error checker.
- read_archive_map.f90/read_archive_map.fh
A subroutine that read a skymap from a binary table in a FITS file.
- read_fits.f90/read_fits.fh
A collections of subroutines that read two and three dimension floating point and complex
arrays from the primary header/data unit of a FITS file.
- br_mod_dist.f90
An F90 module for computing the Blackwell-Rao estimator given signal samples
from the posterior [Written by Hans Kristian Eriksen]
Used by the low l TT gibbs-sampling likelihood option.
- Makefile
This is a standard makefile to compile the code.
- data/
A directory containing data necessary to the software.
- CHANGES
The list of changes made since the initial release.
Additional Information
- Seven-Year Wilkinson
Microwave Anisotropy Probe (WMAP) Observations: Sky Maps, Systematic Errors, and Basic Results,
N. Jarosik, et.al. (2010)
- Seven-Year Wilkinson
Microwave Anisotropy Probe (WMAP) Observations: Galactic Foreground Emission,
B. Gold, et.al (2010).
- Seven-Year Wilkinson
Microwave Anisotropy Probe (WMAP) Observations: Power Spectra and WMAP-Derived Parameters,
D. Larson, et.al. (2010)
- Seven-Year Wilkinson
Microwave Anisotropy Probe (WMAP) Observations: Assessment of Anomalous CMB Results from WMAP,
C. Bennett, et.al. (2010)
- Seven-Year Wilkinson
Microwave Anisotropy Probe (WMAP) Observations: Cosmological Interpretation,
E. Komatsu, et.al. (2010)
- Seven-Year Wilkinson
Microwave Anisotropy Probe (WMAP) Observations: Mars, Jupiter and Saturn,
J. Weiland, et.al (2010).
- WMAP Seven Year Explanatory Supplement
Back to Product Page