Chandra absolute astrometric accuracy

Summary

This page gives the latest information the absolute positional accuracy of source coordinates in Chandra observations. This is based on measuring the distances between the Chandra X-ray source positions and corresponding optical/radio counterpart positions from two accurate catalogs ( Tycho2 or ICRS).

Distribution of aspect offsets for each detector

The four plots below show the distribution of aspect offsets for observations from launch through 2003 to the present. The aspect offset is simply the distance from X-ray source centroid to the accurately known position of the optical or radio counterpart. This gives a good estimate of astrometric accuracy for observations without well-known counterpart positions.

Off-axis sources or observations with non-zero SIM offset

It is important to note that observations on ACIS or HRC-S at large off-nominal SIM-Z can suffer additional residual aspect offsets of up to 0.5 arcsec. This is due primarily to distortion of fid light images by the retroreflector collimator in the fiducial transfer system. HRC-I observations at large off-nominal SIM-Z can have residual aspect offsets of up to 3 arcsec due to the fid light distortion as well as a possible rotation of the detector relative to the geometry assumed in processing.

Sources which are at HRMA off-axis angles greater than 3 arcmin have additional offsets due to PSF blurring and asymmetry. A quick rule of thumb in this case is that the uncertainty in absolute astrometry is roughly 1/4 of the 50% encircled energy radius at the corresponding off-axis angle. This is based on the data tables available in the memo Plate Scale and Relative Chip Postions. A more accurate determination of uncertainties could be made using raytrace simulations with MARX or ChaRT

Time history of aspect offsets

The time-history of offsets for recent Chandra observations is shown below. There are no significant time-dependent residuals. For old archival data it should be noted that ACIS observations taken in 1999, and particularly before ~November 1999 could suffer larger offsets. Shortly after launch the alignment drifted rapidly due to outgassing of the ACIS mount tabs, and the calibration does not account for this effect.

For every "qualifying" observation, the positional accuracy monitor cross-correlates detected X-ray sources with several accurate optical/radio catalogs. For grating observations, the central 0.4 arcmin is used, while for imaging observations sources in the central 4 arcmin are considered. Currently a qualifying observation has:

• Integration time less than 100 ksec. This will probably be relaxed.
• Science category which is not "SN, SNR, and isolated NS", "Clusters of Galaxies", or "Solar System and Misc". Observations in these categories are likely to confound celldetect and produce too many spurious sources.

How is the offset defined?

# Define delta RA and dec in arcseconds
delta_RA  = (RA_catalog - RA_Xray) * cos(dec_catalog) * ARCSEC_PER_DEGREE
delta_dec = (dec_catalog - dec_Xray) * ARCSEC_PER_DEGREE

cr = cos(ROLL_NOM)   # Nominal ROLL taken from observation header
sr = sin(ROLL_NOM)

DZ = cr * delta_dec - sr * delta_RA;   
DY = sr * delta_dec + cr * delta_RA;

• ICRS - Sources from tables 3-5 in Ma et al. 1998, AJ 116, 516 ("The International Celestial Reference Frame as Realized by Very Long Baseline Interferometry"). Astrometric accuracy is ~1 mas.
• Tycho2 - Reference catalog of 2.5 million stars observed by the Tycho instrument abord the ESA Hipparcos satellite. Astrometric accuracy ~25 mas with stars down to ~11.5 mag. More information available in the Guide to the Tycho2 catalog (PDF).
• 2MASS - Two micron all-sky survey: 162,213,354 million point sources from 19,600 square degrees of sky. Astrometric accuracy is better than 0.2 arcsec.
• USNO-A2.0 - U.S. Naval Observatory Precision Measuring Machine project. 526,280,881 stars over the full sky. Astrometric precision is non-uniform, depending on position on Schmidt plate. Typically better than 1 arcsec.

Archival notes and announcements

Note about aspect offset in early 2007: (May-2007)