X-UDS: The Chandra Legacy Survey of the UKIDSS Ultra Deep Survey Field

The first scientific goal of X-UDS is to identify a sizable number of heavily obscured AGNs and determine their obscuration and host galaxy properties in the intrinsic luminosity range L_X ∼ 10⁴⁴ to 10^45.5 erg s⁻¹ up to z ∼ 2, covering the peak activity of SMBH growth in the universe. Population synthesis analyses of the CXB require a substantial fraction of heavily obscured CTAGNs; however, their cosmic fraction and the demographics of their host galaxies are still uncertain (e.g., Comastri et al. 1995; Worsley et al. 2005; Gilli et al. 2007; Akylas et al. 2012; Ueda et al. 2014). These AGNs represent a key phase in the life cycle of galaxies as the majority of SMBH growth is expected to be enshrouded in obscuring gas and dust (Hopkins et al. 2006). CTAGNs are challenging to detect given the strong absorption of their X-ray signals (by factors of 10–100) at energies <10 keV. However, even the most obscured objects with N_H > 10²⁴ cm⁻² can be identified by sensitive X-ray spectroscopy due to nuclear emission that is Compton-scattered into our line of sight. This "reflected" emission has a characteristic spectral shape consisting of a flat continuum and a high-equivalent-width Fe Kα fluorescence line (Matt et al. 1996). Identifying obscured AGNs becomes easier at higher redshifts as key spectral features associated with CTAGNs, such as the Fe Kα line at 6.4 keV and the Compton reflection bump, which peaks at 30 keV, redshift into Chandra's 2–8 keV band at z ∼ 2. Recently, Brightman et al. (2014) identified 100 CTAGNs in the Chandra Deep Field South, EGS, and COSMOS fields using spectral models from Brightman & Nandra (2011), which correctly account for emission from Compton scattering and the geometry of the absorbing material and include a self-consistent treatment for Fe Kα emission.

To facilitate the identification of CTAGNs out to z ∼ 2, the exposure pattern of the X-UDS observations are designed in a way to achieve both a deep coverage (∼600 ks) over the central CANDELS region and wide coverage to facilitate the CXB/CIB fluctuation work. The UDS field has deep Hubble, Spitzer, and Herschel observations, allowing us to determine the morphologies, masses, and star formation rates (SFR) of AGN hosts as a function of obscuration to z ∼ 3. The X-UDS observations, in conjunction with the extensive multiwavelength data already available in the field, will allow us to address (1) the evolution of the CTAGN fraction with redshift and (2) the mechanisms that trigger obscured SMBH growth over cosmic time.

Regarding the CTAGN fraction, recent progress in determining the evolution of the AGN hard X-ray luminosity function (Vito et al. 2014; Aird et al. 2015) and population synthesis models of the X-ray background spectrum (Ueda et al. 2014), as well as new spectral analysis models (Brightman & Nandra 2011), has allowed us to start exploring the redshift and luminosity evolution of CTAGN fraction. However, the statistical and systematic uncertainties of the current samples are still substantial. Several studies have reported an increase in the fraction of heavily obscured AGNs with redshift (e.g., Hasinger 2008), but the extent of this increase is still debated (Treister et al. 2009; Brightman & Ueda 2012; Buchner et al. 2015). Populating the CTAGN demographics with more data and a better handle on systematic selection effects will be key to improving our understanding of how the CTAGN population evolves with redshift. We estimate that combining the intrinsically bright CTAGNs detected in EGS and UDS with fainter sources from the ultradeep CDFS/N data will allow for a robust determination of the CTAGN X-ray luminosity function (XLF) out to z = 3 in at least three redshift bins.

Regarding the triggering mechanisms of obscured SMBH growth, galaxy mergers have long been proposed as a means to fuel AGN activity (e.g., Hopkins & Hernquist 2006), yet studies of X-ray-selected AGNs out to z ∼ 2 have failed to find the predicted growth of dark matter host halos (Allevato et al. 2011) or the morphological signatures indicative of recent merger activity (Schawinski et al. 2011; Kocevski et al. 2012; Rosario et al. 2015). However, past studies have not been sensitive to CTAGNs due to their extremely attenuated X-ray emission. Hydrodynamical merger simulations predict that an obscured AGN phase should coincide with the most morphologically disturbed phase of a galaxy interaction (Cattaneo et al. 2005; Hopkins et al. 2008). It is therefore likely that many studies have systematically missed the AGN–merger connection by not sampling the obscured AGN population well. We aim to use the existing CANDELS imaging in the UDS to search for signs of disturbed host morphologies and recent merger activity among the CTAGN population. Based on existing samples of CTAGNs in CDFS, EGS, and COSMOS, Kocevski et al. (2015) recently reported a 3.8σ excess of disturbed morphologies among the heavily obscured AGNs at z ∼ 1 compared to unobscured AGNs with similar X-ray luminosities. This is part of an increasing body of work that now connects obscured AGNs with galaxy interactions (e.g., Koss et al. 2011; Glikman et al. 2015, Donley et al. 2018, Goulding et al. 2018, Ricci et al. 2017). The X-UDS observations will increase the sample of CTAGNs that have been imaged with Hubble/WFC3 and allow for this work to be extended to z ∼ 2 for the first time, where mergers are predicted to play an even greater role in triggering obscured AGNs.

The CIB is the collective radiation emitted throughout cosmic history as observed at infrared wavelengths. The intensity, spectrum, and spatial fluctuations of the CIB all provide valuable information about sources too faint to be directly detected (e.g., Kashlinsky et al. 2005, 2007, 2012; Cooray et al. 2012; Helgason et al. 2012). A salient feature of the CIB fluctuations is that their spatial power spectrum rises a factor of 10 above the expected contribution from all known sources at angular scales larger than 20'' (Kashlinsky et al. 2012). It has been proposed that these fluctuations in the CIB arise from objects within the first 0.5 Gyr of the universe (z > 6). This is evidenced by the fact that there are no large-scale correlations between the source-subtracted Spitzer/IRAC imaging and faint Hubble/ACS sources in the GOODS regions observed to m_AB ∼ 28 (Kashlinsky et al. 2007). This implies that unless the CIB anisotropies arise in more local but extremely faint and so far unobserved galaxies (i.e., m_AB > 28), the sources responsible for the near-infrared emission must be redshifted beyond z = 6, putting the Lyman break redward of the longest ACS wavelength (∼9000 Å). The CIB fluctuations are also isotropic, described by the ΛCDM model density field at high z; this clustering is significantly different from that of known galaxy populations at recent epochs. Similar results have been obtained with AKARI by Matsumoto et al. (2011), who showed that the spectral energy distribution (SED) of the fluctuations has a roughly λ⁻³ dependence out to 2.4 μm, which is consistent with the superposition of Rayleigh–Jeans spectra from high-z sources.

Recently, Cappelluti et al. (2013) measured a significant cross-correlation signal between the source-subtracted CIB and CXB fluctuations in the Extended Groth Strip. The Chandra 0.5–2 keV unresolved CXB fluctuations on scales 20''–800'' are highly coherent with the CIB, with an overall significance of ∼3.8σ and ∼5.6σ for the 3.6 and 4.5 μm IRAC bands, respectively, suggesting significant active BH populations among the CIB sources. The measured cross-power indicates that objects associated with powerful X-ray emitters produce 15%–25% of the CIB power. The coherence of the CXB–CIB cross-correlation signal is well above that from known sources (Helgason et al. 2012), and it is therefore likely that a substantial growing BH population contributes a large fraction of the CIB in the early universe (e.g., Mirabel et al. 2011).

If the sources responsible for the CXB and CIB fluctuations are at high redshift (i.e., z > 6) and distributed according to a ΛCDM density field, their angular CXB/CIB fluctuation spectrum should dominate in the region around 1000'' (4 Mpc), depending on the epoch of the sources. Thus far, the joint CXB and CIB fluctuations have been studied on scales <800'' in the EGS field, but, due to its elongated configuration (8' × 45'), scales exceeding ∼300'' are poorly sampled. The UDS field has been observed homogeneously with warm Spitzer IRAC observations as part of the SEDS survey that are of depth similar to those in EGS, but in a geometry (22' × 22') that is much better suited for the determination of the cosmologically interesting signal at large angular scales.

The X-UDS observations are designed to match the geometry of the existing Spitzer observations, allowing for the cross-correlation of the unresolved diffuse CIB and CXB signals. This will provide a unique insight into populations of the early universe unobtainable by other means, yielding information of fundamental importance to cosmology. A primary goal of X-UDS is, therefore, to measure the shape of the CXB/CIB fluctuation spectrum, determine whether the large-scale fluctuations are due to high-z sources, and extract information on the first luminous sources in the universe.

The X-UDS observations were carried out with Chandra's Advanced CCD Imaging Spectrometer (ACIS; Garmire et al. 2003) in the time period 2015 July 6 to 2015 October 4. The observations consist of 25 pointing positions that cover a total area of roughly 35' × 25' in size. The exposure times and roll angles of the individual observations range from 702 to 1062 and 44.69 to 51.19 ks, respectively. A summary of the observational parameters of the X-UDS exposures is listed in Table 1. Each pointing was imaged with the 16'9 × 16'9 ACIS-I array, with the aim point located on the ACIS-I3 chip. The ACIS-S2 chip was also active during the observations, but due to its large off-axis angle and reduced effective area, we do not make use of it in this analysis. All observations were carried out in the FAINT telemetry mode with the nominal 3.2 s CCD frame time. The individual pointings are mosaicked to provide an average exposure time of 200 ks over the outskirts of the field and 600 ks in the central region that overlaps the CANDELS Hubble/WFC3 imaging. An exposure map of the combined X-UDS observations is shown in Figure 1.

Zoom In Zoom Out Reset image size

The data reduction was performed using the CIAO data analysis software v4.7 (Fruscione et al. 2007), closely following the prescription described in Laird et al. (2009; L09) and Nandra et al. (2015). Briefly, for each individual observation (hereafter ObsID), we corrected the level 1 event files for aspect offsets, applied destreaking algorithms, and identified hot pixels and cosmic-ray afterglows for removal using the acis_find_afterglow task. New level 2 event files were then created after correcting for charge transfer inefficiency (CTI) and gain effects using the acis_process_events task. To identify periods of anomalously high background, we created light curves for each ObsID in the 0.5–7 keV band with a bin size of 50 s. This analysis was restricted to ACIS chips 0–3 and ASCA-style event grades 0, 2, 4, and 6. Periods of high background were rejected using the procedure of Nandra et al. (2005), adopting a threshold of 1.4 times the quiescent background level, determined as the count rate at which the background shows zero excess variance over that expected from statistical fluctuations alone.

Following this basic reduction, we corrected the astrometry of the individual image frames using a reference catalog. For this task, we made use of the UKIDSS (Lawrence et al. 2007) DR10 K-band catalog²³ to register the X-UDS images to the near-infrared reference frame. We first ran the Chandra wavelet source detection task wavdetect on the 0.5–7 keV image, using a detection threshold of 10⁻⁶, then used the CIAO task reproject_aspect to correct the astrometry compared to the reference image and create new aspect solution files. The new aspect solutions were then applied to the event files using the task reproject_events. The parameters used for reproject_aspect were a source match radius of 2'' and a residual limit of 050. Typically, ∼40 sources were matched in each ObsID and used in the reprojection. The absolute value of the applied offsets at this step was consistently small ($\lt 0\buildrel{\prime\prime}\over{.} 5$).

Zoom In Zoom Out Reset image size

After this astrometric calibration, we created event files, exposure maps, and point-spread function (PSF) maps for each ObsID in the full (FB; 0.5–7 keV), soft (SB; 0.5–2 keV), hard (HB; 2–7 keV), and ultrahard (UB; 4–7 keV) bands. For the purpose of producing the color mosaic shown in Figure 3, we also produced images in the 2–4 keV and 4–8 keV bands. The exposure maps were created using the task fluximage for each of our four primary bands using weights appropriate for a Γ = 1.4 power-law spectrum; these weights are listed in Table 2. The PSF maps were created using mkpsfmap using an enclosed counts fraction of 0.3. The individual event files were then combined using the reproject_obs task, while the exposure maps were stacked together using the dmregrid task. The individual PSF maps were combined using the task dmimgfilt so as to return the minimum PSF value at each location in the combined mosaic. The final data product of this reduction is a merged event file that covers the entire X-UDS region. A mosaic image of the merged event file in the full band is shown in Figure 2. An adaptively smoothed, "true-color" image of the center of the field, created using the csmooth algorithm (Ebeling et al. 2006), is shown in Figure 3. The effective exposure time as a function of survey area is shown in Figure 4.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Source detection proceeded in the same fashion as that described in Nandra et al. (2015). We start by creating a "seed" source catalog using wavdetect, which is run with a low false-positive probability threshold (sigthresh = 10⁻⁴) in order to capture all potential sources. Here wavdetect was run on the combined event files in our four energy bands with the standard $\sqrt{2}$ set of wavelet scales (i.e., 1, 1.41, 2, 2.82 ... 16). We also provided to wavdetect the minimum PSF maps created earlier for each energy band. The resulting seed catalogs contain 2850, 2021, 2306, and 1625 sources in the full, soft, hard, and ultrahard bands, respectively.

We then extract counts from the merged images at the positions of each of our candidate sources. For the purpose of source validation, counts were measured using a circular aperture with radius equal to the exposure-weighted 50% encircled energy fraction (EEF) of the Chandra PSF. Here, the PSF size at the location of each candidate source, in each separate ObsID, was taken from a lookup table calculated using the MARX simulation software as described in L09. An exposure-weighted average PSF size is then computed for each source in the merged image. The background near each source was determined using an annulus with an inner radius equal to 1.5 times the 95% EEF at the source position and an outer radius of 100 pixels larger than this, excluding detected sources. An average exposure value was also calculated for the source and background areas. The counts in the background area were then scaled to the source region by the ratio of the source and background areas and the ratio of the source and background average exposures, after masking out the 95% EEF region of other candidate sources.

Next the Poisson false probability of observing the total counts measured, given the expected background, was calculated for each source. A significance threshold of 1 × 10⁻⁴ was then applied,²⁴ and a further detection iteration was performed to mask out only sources more significant than this. This second iteration ensures that the background is not underestimated due to the masking of random positive variations identified as candidate sources by wavdetect. Any source detected at this 1 × 10⁻⁴ probability level in the second iteration in any individual band was included in the final catalog. As a one-sided p value, our threshold of 1 × 10⁻⁴ roughly corresponds to a 3.7σ detection above the expected background.

This process was carried out separately for each of our four energy bands, and sources considered significant were band-merged using a matching radius that depends on the exposure-weighted average off-axis angle of the source. Our adopted cross-band matching radii are $1\buildrel{\prime\prime}\over{.} 30$, $2\buildrel{\prime\prime}\over{.} 44$, $4\buildrel{\prime\prime}\over{.} 79$, and $7\buildrel{\prime\prime}\over{.} 08$ for sources with off-axis angles of 0–3, 3–6, 6–9, and >9 arcmin, respectively.²⁵ These matching radii are roughly 4 times the 1σ positional uncertainties at these off-axis angles, as determined from MARX simulations (see L09).

Photometry was then performed to estimate source fluxes in several energy ranges, even if the source was not considered a significant detection in that particular band. For these measurements, we used circular apertures to extract the counts from the combined images, using the exposure-weighted 90% EEF PSF appropriate for each source. Fluxes and 1σ confidence limits were estimated using the Bayesian methodology described in L09, using a spectral slope of Γ = 1.4 with Galactic N_H of 2.54 × 10²⁰ cm² (Dickey & Lockman 1990). Using this method, the best estimate of the source flux is obtained by finding the mode of the posterior distribution function, which is the product of the prior probability distribution and the Poisson likelihood of obtaining the observed total counts for a given source rate and background. We assumed a prior probability distribution for source fluxes that is based on the observed log N–log S relation reported in Georgakakis et al. (2008). This approach helps to correct for Eddington bias (Eddington 1940), which causes the flux of faint sources to be generally overestimated using classical approaches. The resulting fluxes are then extrapolated to the standard energy bands of 0.5–10 keV, 2–10 keV, and 5–10 keV. All fluxes are on-axis values and corrected for aperture size.

In addition, we also calculated fluxes using the classical method of converting count rates to fluxes. Effective on-axis source count rates were calculated by dividing the net counts with the average value of the exposure map (in units of count photon⁻¹ cm² s) and aperture correcting. These count rates were converted to fluxes using energy conversion factors of 4.25, 1.71, 8.83, and 1.26 × 10⁻⁹ energy photon⁻¹ in the full, soft, hard, and ultrahard bands, which are appropriate for a Γ = 1.4 spectrum. We then extrapolate the fluxes in the full, hard, and ultrahard bands out to 10 keV. Finally, hardness ratios (HRs) were calculated using the Bayesian methodology of Park et al. (2006). For this, we employed the BEHR package,²⁶ which models the detected counts as a Poisson distribution and gives error bars and reliable HRs for sources with both low and high counts. For comparison, we also calculated HRs using the classical method of HR = (H − S)/(H + S), where H and S are the hard and soft band net counts, respectively, corrected to on-axis values. Sources that only have upper limits on their soft or hard band counts have HR values set to +1 and −1, respectively.

To determine the limiting flux of our observations, we have computed sensitivity maps using the method described in Georgakakis et al. (2008) and implemented more recently in Nandra et al. (2015). In short, we begin by computing the average local background per pixel over the entire mosaic in each energy band, after masking out detected sources. With knowledge of the exposure-weighted PSF size at any location in the image, we then determine the minimum integer number of source counts needed to produce a false-positive Poisson probability that is less than our adopted threshold of 1 × 10⁻⁴ given the local background. This approach accounts for incompleteness and Eddington bias in the sensitivity calculation and is performed in a manner that is also fully consistent with our source detection procedure. After accounting for the exposure time in the detection cell and the fraction of the total source counts in the cell due to the PSF size, we can determine the minimum flux needed for detection in each band. Our sensitivity maps are 2D images of this minimum flux over the X-UDS footprint. The flux limits for the X-UDS survey as a function of area are shown for various energy bands in Figure 5. We define the limiting flux of our observations as the flux to which at least 1% of the survey area is sensitive. Using this definition, we find the limiting fluxes to be 4.4 × 10⁻¹⁶ (FB; 0.5–10 keV), 1.4 × 10⁻¹⁶ (SB; 0.5–2 keV), 6.5 × 10⁻¹⁶ (HB; 2–10 keV), and 9.2 ×10⁻¹⁵ (UB; 5–10 keV) erg cm⁻² s⁻¹.

Zoom In Zoom Out Reset image size

Following the approach of Nandra et al. (2015), we can constrain the number of spurious detections in our source catalog using our ultrahard band images. Due to Chandra's decreased sensitivity at ultrahard (4–7 keV) energies and the expected spectral shape of the source population, we do not expect sources to be detected exclusively in the ultrahard band without corresponding detections in the full, soft, or hard bands. Even the most obscured, Compton-thick sources are often detected at soft energies due to the presence of scattered X-ray light (e.g., Brightman & Nandra 2012). Therefore, we expect the number of ultrahard-only detections to provide insight on the number of spurious sources in each band. In our final catalog, nine sources are detected in the ultrahard band that pass our threshold cut and also lack detections in any other band. This implies that over our four detection bands, 36 sources may be spurious, or roughly 4.1% of our final catalog of 868 sources.

The final X-UDS source catalog of 868 sources is presented in Tables 4 and 5. Table 4 provides basic source properties such as source position, counts, and detection information. A more detailed description of each column in Table 4 is reported below.

Table 5.  Chandra X-UDS Source Catalog: Source Fluxes and HRs

ID	Bayesian Flux				Classical Flux				Bayes.	Class.	Phot.
	F0.5–10	F0.5–2	F2–10	F5–10	F0.5–10	F0.5–2	F2–10	F5–10	HR	HR	flag
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)
xuds_001	${109.70}_{-3.89}^{+4.03}$	${11.89}_{-0.95}^{+1.03}$	${126.67}_{-5.00}^{+5.19}$	${81.59}_{-5.10}^{+5.41}$	${110.04}_{-3.89}^{+4.03}$	${12.07}_{-0.95}^{+1.03}$	${127.16}_{-5.00}^{+5.19}$	${82.38}_{-5.10}^{+5.41}$	${0.51}_{-0.03}^{+0.03}$	0.51	0
xuds_002	${15.72}_{-2.80}^{+3.10}$	${4.86}_{-1.07}^{+1.24}$	${8.34}_{-2.52}^{+2.93}$	<2.92	${16.54}_{-2.98}^{+3.48}$	${5.28}_{-1.21}^{+1.50}$	${9.65}_{-2.83}^{+3.54}$	<5.46	$-{0.29}_{-0.16}^{+0.18}$	−0.30	0
xuds_003	${2.77}_{-1.54}^{+1.69}$	${1.03}_{-0.52}^{+0.66}$	<1.53	<2.26	${3.94}_{-1.60}^{+2.08}$	${1.47}_{-0.62}^{+0.89}$	$\lt 4.03$	<5.31	$-{0.42}_{-0.38}^{+0.27}$	−1.00	0
xuds_004	${95.53}_{-2.95}^{+3.04}$	${28.38}_{-1.15}^{+1.19}$	${59.61}_{-2.88}^{+3.02}$	${28.22}_{-2.58}^{+2.80}$	${95.75}_{-2.95}^{+3.04}$	${28.50}_{-1.15}^{+1.19}$	${59.96}_{-2.88}^{+3.02}$	${28.80}_{-2.58}^{+2.80}$	$-{0.24}_{-0.03}^{+0.03}$	−0.24	0
xuds_005	${4.15}_{-0.96}^{+1.05}$	${1.23}_{-0.33}^{+0.37}$	${2.50}_{-1.18}^{+1.24}$	<1.28	${4.49}_{-1.00}^{+1.14}$	${1.38}_{-0.35}^{+0.43}$	${3.21}_{-1.14}^{+1.36}$	<2.95	$-{0.19}_{-0.22}^{+0.21}$	−0.19	0
xuds_006	${12.85}_{-1.86}^{+2.02}$	${4.43}_{-0.72}^{+0.80}$	${5.36}_{-1.74}^{+1.99}$	<3.57	${13.28}_{-1.95}^{+2.21}$	${4.62}_{-0.76}^{+0.90}$	${6.26}_{-1.87}^{+2.30}$	<2.76	$-{0.42}_{-0.14}^{+0.13}$	−0.43	0
xuds_007	${43.22}_{-1.89}^{+1.97}$	${10.06}_{-0.64}^{+0.68}$	${35.32}_{-2.16}^{+2.29}$	${18.62}_{-2.02}^{+2.23}$	${43.43}_{-1.89}^{+1.97}$	${10.16}_{-0.64}^{+0.68}$	${35.65}_{-2.16}^{+2.29}$	${19.16}_{-2.02}^{+2.23}$	${0.01}_{-0.05}^{+0.04}$	0.01	0
xuds_008	${21.83}_{-1.39}^{+1.47}$	${5.24}_{-0.47}^{+0.51}$	${17.50}_{-1.59}^{+1.73}$	${6.79}_{-1.23}^{+1.36}$	${21.96}_{-1.39}^{+1.47}$	${5.30}_{-0.47}^{+0.51}$	${17.86}_{-1.59}^{+1.73}$	${7.18}_{-1.34}^{+1.56}$	$-{0.01}_{-0.06}^{+0.06}$	−0.01	0
xuds_009	${22.01}_{-1.38}^{+1.46}$	${6.40}_{-0.52}^{+0.56}$	${13.02}_{-1.37}^{+1.51}$	${6.05}_{-1.15}^{+1.28}$	${22.14}_{-1.38}^{+1.46}$	${6.50}_{-0.52}^{+0.56}$	${13.37}_{-1.37}^{+1.51}$	${6.42}_{-1.24}^{+1.45}$	$-{0.25}_{-0.06}^{+0.06}$	−0.25	0
xuds_010	${14.90}_{-2.08}^{+2.25}$	${2.78}_{-0.61}^{+0.70}$	${12.30}_{-2.23}^{+2.60}$	<3.09	${15.37}_{-2.18}^{+2.45}$	${3.00}_{-0.66}^{+0.81}$	${13.38}_{-2.57}^{+3.00}$	<4.28	${0.13}_{-0.14}^{+0.15}$	0.13	0

Note. (1) Source identification number, (2–5) source fluxes in units of 10⁻¹⁵ erg cm⁻² s⁻¹ in our four analysis bands calculated using the Bayesian methodology of L09, (6–9) source fluxes in units of 10⁻¹⁵ erg cm⁻² s⁻¹ in our four analysis bands calculated using the classical method of converting count rates to fluxes, (10–11) source hardness ratios calculated using the Bayesian and classical methods, (12) photometric quality flag indicating possible ("1") and likely ("2") contamination from a nearby source; all other sources have flag values of "0". See Section 4.3 for details.

Only a portion of this table is shown here to demonstrate its form and content. A machine-readable version of the full table is available.

Download table as:  DataTypeset image

• 1.  
  Column 1: Source identification number
• 2.  
  Column 2: Chandra source name, following the standard IAU convention with prefix "CXOUDS" for "Chandra X-ray Observatory UDS survey."
• 3.  
  Columns 3–4: R.A. and decl. in the J2000 coordinate system.
• 4.  
  Column 5: Positional error in arcseconds. Based on simulations carried out by L09, the positional error assigned to each source is dependent on its average off-axis angle (OAA) and net counts in the Full band. These errors range from 014 for bright sources detected at low OAAs (OAA < 4' and N > 100 counts) to 133 for faint sources at larger OAAs (OAA > 8' and N < 50 counts).
• 5.  
  Column 6: Exposure-weighted, average off-axis angle in arcminutes.
• 6.  
  Columns 7–14: Total source counts (N) and background counts (B) in the four analysis bands. Counts are given regardless of whether a source was detected in the band.
• 7.  
  Column 15: List of the bands in which a source is detected with Poisson false detection probability <1 × 10⁻⁴, where the bands are full (F), soft (S), hard (H), and ultrahard (U).
• 8.  
  Column 16: Log of the minimum Poisson probability among the four analysis bands. Probabilities lower than 10⁻⁸ are listed as −8.0.

Table 5 presents the flux and HR information for each source in the final catalog. A more detailed description of each column in Table 5 is reported below.

• 1.  
  Column 1: Source identification number
• 2.  
  Columns 2–5: Observed frame source fluxes in our four analysis bands calculated using the Bayesian method described in L09, which corrects for Eddington bias. Calculations were done using a spectral slope of Γ = 1.4 and Galactic N_H of 2.54 × 10²⁰ cm² (Dickey & Lockman 1990). The reported errors are 1σ values. Where sources are not detected in a particular band, we report the 68% upper limit. All fluxes have units of 10⁻¹⁵ erg cm⁻² s⁻¹ and have not been corrected for intrinsic source absorption.
• 3.  
  Columns 6–9: Observed frame source fluxes in our four analysis bands calculated using the classical method of converting count rates to fluxes. Calculations were done using a spectral slope of Γ = 1.4 and Galactic N_H of 2.54 × 10²⁰ cm² (Dickey & Lockman 1990). Reported errors are 1σ values. All fluxes have units of 10⁻¹⁵ erg cm⁻² s⁻¹ and have not been corrected for intrinsic source absorption.
• 4.  
  Columns 10 and 11: Source hardness ratios calculated using the Bayesian and classical methods. The reported errors on the Bayesian HRs are 1σ values. For sources only detected in the full band, and with upper limits in both hard and soft bands, the classical HR cannot be determined and is set to −99.
• 5.  
  Column 12: Photometry quality flag. A value of "1" indicates the presence of a nearby source that may be contaminating the photometry. A flag of "2" indicates that another source was detected within the 90% EEF and that the photometry is likely heavily contaminated and the source position uncertain. All other sources have a flag of "0".