Chandra Source Catalog Release 2.1
Release 2.1 Overview
The Chandra Source Catalog (CSC) is the definitive catalog of X-ray sources detected by the Chandra X-ray Observatory. By combining Chandra's sub-arcsecond on-axis spatial resolution and low instrumental background with consistent data processing, the CSC delivers a wide variety of uniformly calibrated properties and science-ready data products for detected sources over four decades of flux.
CSC 2.1 includes measured properties for 407,806 unique compact and extended X-ray sources in the sky, allowing statistical analysis of large samples, as well as individual source studies. Extracted properties are provided for 1,304,376 individual observation detections identified in Chandra ACIS and HRC-I imaging observations released publicly through the end of 2021. Photometric properties for 1,717 highly extended (≳ 30″) sources are included as an "alpha" release, together with surface brightness polygons for several contour levels.
The sensitivity limit for compact sources in CSC 2.1 is ~5 net counts, achieved by using a two-stage approach that involves co-adding multiple observations of the same field prior to source detection, and then using an optimized source detection method.
For each X-ray detection and source, the catalog provides a detailed set of more than 100 tabulated positional, spatial, photometric, spectral, and temporal properties (each with associated lower and upper confidence intervals and measured in multiple energy bands). The catalog Bayesian aperture photometry code produces robust photometric probability density functions (PDFs), even in crowded fields and for low count detections. A Bayesian Blocks analysis identifies multiple observations of the same source that have similar photometric properties, and these are analyzed simultaneously to improve the signal to noise.
CSC 2.1 additionally provides an extensive selection of individual observation, stacked-observation, detection region, and master source FITS data products (e.g., responses, PSFs, spectra, light curves, aperture photometry PDFs) that are immediately usable for further detailed scientific analysis.
CSC 2.1 By the Numbers
| Master sources | Compact sources | 406,089 |
| Highly extended (“convex hull”) sources | 1,717 | |
| Total sources | 407,806 | |
| Stacked observations | Number of observation stacks | 10,034 |
| Detections | 493,236 | |
| Detections and photometric upper limits | 855,402 | |
| Individual observations | Number of individual observations | 15,533 |
| Detections | 1,304,376 | |
| Detections and photometric upper limits | 2,143,847 |
Release 2.1 Features
Observation Stacking
Exploiting the unique resolution and very low background of Chandra data, the limiting sensitivity of the catalog is enhanced significantly by stacking (co-adding) multiple observations of the same field prior to source detection. To minimize the impact of the variation of the Chandra point spread function (PSF) with off-axis angle, source detection is constrained to run on stacks of observations that have telescope pointings that are co-located within 60″, and that were obtained using the same instrument (ACIS or HRC-I).
Stacked-Observation Absolute Astrometry
CSC 2.1 observation stacks with sufficient X-ray source/Gaia optical source crossmatches (~98% of stacks) are tied to the Gaia-CRF3 astrometric reference frame, which provides an absolute astrometric reference for the entire catalog.
Voronoi Tessellation Background Maps
CSC 2.1 uses a Voronoi tessellation background tool (mkvtbkg) to create background maps prior to source detection. These background maps perform well in regions where the background intensity is changing rapidly (e.g., in areas of extended emission near the centers of galaxies) and at large off-axis angles.
Robust Two-step Compact Source Detection
The CSC 2.1 source detection method can reliably detect compact sources down to roughly 5 net counts over much of the field-of-view. The Voronoi tessellation background maps enable compact source detection in areas with diffuse emission and near the edge of fields. Source detection is performed first using the CIAO wavelet source detection tool (wavdetect) with tool parameters set to identify faint candidate detections, albeit with an unacceptably high false detection rate. Candidate compact detections superimposed on regions with bright extended emission are identified as a side-effect of running the mkvtbkg tool.
The combined set of candidate detections are further evaluated by a maximum likelihood estimator tool (mle) by fitting the local, energy band dependent PSF model to each candidate detection to the counts data to evaluate the likelihood that the detection is real. Candidate detections included in CSC 2.1 are classified as either TRUE or MARGINAL in the catalog, by comparing their likelihood estimates with a pair of thresholds that calibrate the permissible false detection rates.
The likelihoods of closely separated candidate detections are evaluated by fitting simultaneously with a model that accounts for the multiplicity of the detections. This makes the fits efficient and robust, even in cases where the nearby detections have similar numbers of counts.
Detailed Local PSF Modeling
Since the Chandra PSF is highly position dependent, the local PSF model is calculated separately for each energy band for each detection position prior to use by the mle tool. Fitting the local PSF to the photon splash of the detection improves source astrometry, particularly for larger off-axis angles where PSF asymmetries can bias centroid-based position determinations. The PSF models are available to for end users as FITS format data products, and each model includes ~50,000 counts, which is adequate for many types of studies.
Inclusion of Bright Extended Sources
In addition to background determination and identification of candidate compact detections, the mkvtbkg tool identifies regions of extended emission. Bright, extended sources are recorded using bounding convex hull polygons in CSC 2.1. However, sets of quasi-surface brightness polygons with multiple intensity thresholds are available as FITS format data products for end users who wish to perform more detailed analysis of detected extended sources. Extended sources are an "alpha" release and may not satisfy the same rigorous level of quality assurance that the compact sources meet.
Handling of Edge Effects
The false source rate is significantly higher for detections near the edges of the (stacked) field-of-view, in the gaps between ACIS back-illuminated and front-illuminated CCDs (on the ACIS-S array), and on readout streaks associated with saturated, bright sources, compared to the rest of the field. Detections in these regions are excluded from CSC 2.1 through the application of a pixel mask that removes those pixels from consideration.
Limiting Sensitivity Information
CSC 2.1 includes multi-band limiting sensitivity maps so that users can identify regions that are included in, or excluded from, the catalog. The limiting sensitivity for a point source to satisfy the catalog TRUE or MARGINAL detection thresholds in each energy band are recorded and computed on a HEALPIX grid of order 16 (with a pixel size of ~3.22″ × 3.22″).
Evaluation of Source and Detection Properties
CSC 2.1 includes numerous raw measurements for each detected source as well as scientifically useful properties derived from the observations in which the source is detected. All reported measurements and properties have associated two-sided confidence intervals.
Detection properties are provided at both the stacked-observation detection level and the per-observation detection level. The stacked-observation level allows composite properties to be reported from the co-added observations for detections that would otherwise not be visible or have poor S/N in individual observations, while for higher S/N detections the per-observation properties facilitate analysis of variable sources.
Reported positions include 95% error ellipses for brighter sources and circular errors for faint sources.
Multi-band aperture photometry is determined using a Bayesian approach to compute marginalized posterior probability density functions (MPDFs) for fluxes (and associated photometric quantities such as net count rates) that are subsequently used directly for computing such quantities as cross-band hardness ratios and temporal variability measures. Use of a robust No U-Turn MCMC Sampler improves the ability to provide fluxes in the low-count regime.
If a source is detected in one or more stacked-observations, photometric upper limits that are useful for temporal variability analyses are calculated for any overlapping stacked- and individual-observations in which the source is not detected.
CSC 2.1 uses a Bayesian Blocks analysis to identify multiple detections of the same source that have consistent multi-band aperture photometry. Such detections are analyzed simultaneously to improve S/N when determining detection properties. The properties for the longest exposure Bayesian Block are promoted to master source properties, but the properties for all blocks are recorded in an end-user FITS format data product.
Spectral fits computed using multiple models (absorbed power-law, absorbed black body, absorbed bremsstrahlung, and absorbed APEC) are provided for brighter sources and detections. The minimum counts required to compute spectral fits is 150 counts, but at the source level the analysis now includes simultaneous fits to multiple detections of the same source totaling at least 150 counts.
Intra- and inter-observation variability is evaluated from the individual observation MPDFs using several algorithms, including Gregory-Loredo for intra-observation variability, which also generates optimally binned light curves.
FITS Format Data Products
In addition to the tabulated properties, CSC 2.1 provides numerous FITS format data products that are immediately usable for further analysis. These products include full field event lists, multi-band images, exposure maps, limiting sensitivity maps, merged source lists, and extended source polygons. Source region data products include per-source-region event lists, multi-band images, exposure maps, pulse-invariant spectra, spectral response matrices, aperture photometry probability density functions, Bayesian block properties, position-error MCMC draws, and optimally binned light curves.