PaperlayerView exact revision on arXiv

Source-linked AI summary

The Subaru/XMM-Newton Deep Survey (SXDS): III. X-Ray Data

Yoshihiro Ueda, Michael G. Watson, Ian M. Stewart, Masayuki Akiyama, Axel D. Schwope, Georg Lamer, Jacobo Ebrero, Francisco J. Carrera, Kazuhiro Sekiguchi, Tohru Yamada, Chris Simpson, Guenther Hasinger, Silvia Mateos

arXiv:0806.2846v1astro-ph

TL;DR

The SXDS X-ray survey addresses open questions about AGN populations and clustering by cataloging sources across a continuous 1.14 deg² field. Its catalog and source-count results bridge deep and brighter surveys, while clustering detections differ between soft and hard bands.

  • Problem

    Open questions include how obscured AGN fractions and clustering evolve, and the properties of rare populations require more cosmic volume than existing deep surveys provide.

    Source: §1 · p. 2

  • Method

    The paper presents an XMM-Newton source catalog and source statistics, using combined exposure maps and maximum-likelihood detection.

    Source: §1 · p. 2, §3.2.2 · p. 3, §3.2.3 · p. 4

  • Results

    1245 sources comprise the catalog; the 0.5–2 keV sample has θc = 5.9′′+1.0′′−0.9′′, while the 2–10 keV sample has a 90% upper limit θc < 1.5′′.

    Source: §5.2 · p. 13, §7 · p. 15

  • Takeaways & Limitations

    The log N - log S relations bridge the flux range between Chandra deep surveys and brighter surveys.

    Source: §7 · p. 15

    Exposure-map count-rate weights assume photon indices of 1.5 above 2 keV and 2.0 below 2 keV, though dependence on these assumptions is negligible.

    Source: §3.2.2 · p. 3

Abstract

from arXiv · show

We present the X-ray source catalog in the Subaru/XMM-Newton deep survey. A continuous area of 1.14 deg^2 centered at R.A. = 02h18m and Dec. = -05d is mapped by seven pointings with XMM-Newton covering the 0.2-10 keV band. From the combined images of the EPIC pn and MOS cameras, we detect 866, 1114, 645, and 136 sources with sensitivity limits of 6x10^{-16}, 8x10^{-16}, 3x10^{-15}, and 5x10^{-15} erg cm^{-2} s^{-1} in the 0.5-2, 0.5-4.5, 2-10, and 4.5-10 keV bands, respectively, with detection likelihood >= 7 (corresponding to a confidence level of 99.91%). The catalog consists of 1245 sources in total including 32 extended-source candidates. The averaged log N-log S relations are in good agreement with previous results, bridging the flux range between Chandra deep surveys and brighter surveys. The log N-log S relations show significant spatial variation among pointings on a scale of 0.2 deg^2. Analyzing the auto correlation function, we detect significant clustering signals from the 0.5-2 keV band sample, which can be fit with a power law form (\theta/\theta_c)^{-0.8} with a correlation length of \theta_c=5.9^{+1.0}_{-0.9} arcsec when the integral constraint term is included. In the 2-10 keV band, however, the clustering is not significant with a 90% upper limit of \theta_c < 1.5 arcsec.

1. INTRODUCTION

The SXDS combines deep multiwavelength observations over a contiguous area exceeding 1 deg² to study cosmic properties and large-scale structure. Its X-ray survey addresses open questions about AGN populations and evolution while presenting a catalog and basic source statistics.

Survey context

  • The SXDS spans a contiguous region exceeding 1 deg² with multiwavelength data, including X-ray imaging and spectroscopy from 0.2–10 keV.The survey aims to measure global properties without cosmic-variance effects and reveal large-scale-structure evolution.Source: §1 · p. 1

X-ray motivation

  • Hard X-ray surveys above 2 keV efficiently detect obscured AGNs with less bias from photo-electric absorption and less stellar contamination.The passage notes that X-ray surveys provide samples including heavily obscured, low-luminosity AGNs.Source: §1 · p. 1

Evolution context

  • The SXDS studies AGN evolution against a background of luminosity-dependent downsizing and possible galaxy–SMBH co-evolution.The cited passage describes higher-redshift number-density peaks for more luminous AGNs and notes feedback as a possible explanation.Source: §1 · p. 2

Open questions

  • AGN surveys still face questions about obscured fractions, Compton-thick populations, high-redshift luminosity functions, and clustering by luminosity and type.The authors identify wide, deep SXDS coverage as a step toward these questions and the statistical properties of AGNs.Source: §1 · p. 2

Paper scope

  • The paper presents the SXDS X-ray data and source catalog, with basic statistics of detected sources, based on XMM-Newton observations from 2000 and 2003.The paper also reports log N - log S relations, autocorrelation functions, and long-term variability analyses.Source: §1 · p. 2

2. OBSERVATIONS

Seven XMM-Newton pointings map a continuous 1.14 deg² region, with one deeper central observation and six shallower surrounding observations.

Pointing layout

  • Seven XMM-Newton pointings cover a continuous area of about 1.14 deg², with small overlaps and no gaps.The combined map covers essentially the Subaru optical-image region, but sensitivity varies because of exposure differences and vignetting.Source: §2 · p. 2

Exposure design

  • The central SDS-1 pointing had a nominal 100 ksec exposure, while the six surrounding pointings had 50 ksec each.Observations used thin filters in all three EPIC cameras.Source: §2 · p. 2

Observation epochs

  • Observations took place in three epochs between July 2000 and January 2003, with the final epoch supplementing SDS-4 exposure.SDS-1, 2, 3, and 4 were observed in 2000; SDS-5, 6, and 7 in 2002; SDS-4 was supplemented in 2003.Source: §2 · p. 2

3.1. Image Production

EPIC observations were processed into sky images after background-flare screening, and astrometric offsets were corrected against Subaru R-band positions.

Image processing

  • Image production used SAS 6.1.0 for image creation and subsequent analysis, starting from event files processed by the XMM-Newton SSC pipeline.The event files were produced with an earlier SAS version similar to 5.3.3.Source: §3.1 · p. 2

Image creation

  • The analysis created separate sky images for each observation ID and detector in four energy bands, using 4-arcsec bins.The 7.9–8.3 keV interval was excluded to avoid instrumental Cu K-line background.Source: §3.1 · pp. 2–3

Screening and astrometry

  • Screening removed background flares, leaving net exposure of about 60–80% of allocated observing time.Astrometry was corrected by cross-correlating detected sources with the Subaru R-band catalog; offsets were 0–1.3 arcsec.Source: §3.1 · p. 3

3.2. Source Detection

Source detection combines detector images and background models, then fits candidates across energy bands with maximum likelihood while accounting for crowded sources and assessing extent.

3.2. Source Detection

  • Source detection uses summed pn, MOS1, and MOS2 images, optimized background interpolation, and special handling of neighboring sources.Lists are produced by pointing and merged after excluding overlaps to form a statistically independent sample.Source: §3.2 · p. 3

3.2.1. Background Maps

  • Background maps are spline-interpolated after excluding initial eboxdetect candidates, then checked against observed background profiles across the field of view.The modeled background includes unresolved X-ray background and non-X-ray background.Source: §3.2.1 · p. 3

3.2.2. Summation of Images

  • Combining pn, MOS1, and MOS2 images and background maps maximizes available sensitivity and avoids detector chip-gap complications.Exposure maps average instrument maps weighted by count rate, normalized to pn; assumed photon indices are 1.5 above 2 keV and 2.0 below 2 keV, with negligible dependence on those assumptions.Source: §3.2.2 · p. 3
  • The combined images and exposure maps are treated as a single pn detector, an approximation whose fluxes agree with MOS-axis results at the 1.3% level.The image's stronger outer color fluctuation is attributed to shorter exposure.Source: §3.2.2 · pp. 3–4

3.2.3. Maximum Likelihood Fit

  • emldetect fits PSF models to detector co-added images and returns source position, vignetting-corrected count rate, and detection likelihood.Candidate detection is iterated with 20-arcsec and 12-arcsec cells before the merged list is refit; up to three confused sources are fit simultaneously.Source: §3.2.3 · p. 4
  • Cash's C statistic fits Poisson event counts, and the detection likelihood is −ln(p), with eboxdetect and emldetect cutoffs of 3.2 and 7 balancing missed sources against computational cost.The procedure assesses source extent with a free Gaussian-profile parameter, but the resulting extended-source candidate list is incomplete because earlier detection is optimized for point sources.Source: §3.2.3 · p. 4, §3.2.4 · p. 4

3.3. Source List

The final source list merges detections across seven pointings, retaining sources with detection likelihood ML≥7 in any of six bands and removing duplicates in overlaps. It contains 1245 sources overall.

3.3. Source List

  • The XID and hard-band likelihood fits use summed images and backgrounds, with source positions fixed and no new sources considered.Fixing positions reduces the degrees of freedom; the authors account for this in the log N - log S and autocorrelation analyses.Source: §3.3 · p. 5
  • ML≥7 in any one of six bands qualifies a source for the final list, with the highest-sensitivity pointing used to remove duplicates in overlaps.The six bands are 0.3–0.5, 0.5–2, 2–4.5, 4.5–10, 0.5–4.5, and 2–10 keV.Source: §3.3 · p. 5
  • Table 3 lists extended-source candidates, whose spatial extent is measured by fitting a Gaussian profile with source positions fixed.The listed extent is the Gaussian profile’s 1σ extent.Source: Table 3 · p. 6, §3.3 · p. 6
  • 1245 sources make up the final SXDS catalog.The source list combines detections from the seven pointings.Source: §3.3 · pp. 4, 6
  • About 50 false sources are estimated across the sample at ML≥7; their fraction ranges from 0.6% in the 0.5–4.5 keV band to 5% in the 4.5–10 keV band.The estimate is about 1.5 false sources per pointing in each energy band.Source: §3.3 · p. 6

3.4. Spectral Properties of the Sources

Hardness-ratio analysis indicates that the mean spectra of the sources become harder at fainter fluxes, especially for the 0.5–4.5 keV sample. Color-color and flux-hardness plots compare the observed properties with model spectra.

3.4. Spectral Properties of the Sources

  • At fainter fluxes, source spectra harden with 99.99% (HR2) and 96% (HR3) confidence in the 0.5–4.5 keV sample, and 84% (HR2) and 90% (HR3) in the 2–10 keV sample.At these flux levels, the mean slope is similar to or harder than the X-ray background slope of Γ ≃1.4 over 0.5–10 keV.Source: §3.4 · p. 6
  • The color-color plot uses sources with HR2 and HR3 uncertainties below 0.2 to compare their colors with power-law spectra varying in photon index or absorption at z = 1.The plot includes 208 sources detected in the 0.5–4.5 and/or 2–10 keV bands.Source: §3.4 · p. 6
  • The flux and hardness-ratio histograms show distributions for the 0.5–4.5 and 2–10 keV selected samples.The flux histograms distinguish the samples, while the hardness-ratio histograms show HR2 and HR3 for the respective selections.Source: Figure 3 · p. 8
  • The flux-hardness plots pair 0.5–4.5 keV flux with HR2 and 2–10 keV flux with HR3; their dashed reference line represents a power law with photon index 1.8.HR2 and HR3 are defined from the vignetting-corrected count rates in adjacent energy bands.Source: Figure 4 · p. 9, Figure 5 · p. 9

4.1. Sensitivity Map

The sensitivity-map procedure combines simulations of the detection likelihood with analytical maps to estimate flux limits across the field. The resulting maps have estimated systematic errors of at most 10%.

4.1. Sensitivity Map

  • The method combines simulation and analytical calculation because sensitivity depends on background, exposure, and the point-spread function, making analytic calculation alone nontrivial for the maximum-likelihood algorithm.The figure maps the ML≥7 flux threshold by position; darker colors indicate lower flux limits.Source: §4.1 · p. 9, Figure 7 · p. 10
  • At grid points spaced 40″ apart, simulated point sources on input background maps establish an empirical relation between input flux and emldetect likelihood, without Poisson counting noise.That relation gives the sensitivity for a chosen likelihood threshold at each position.Source: §4.1 · p. 9
  • Full-resolution 4″-pixel maps interpolate between grid points using analytical esensmap sensitivity maps to account for position-dependent changes, including CCD gaps and hot pixels.The method maps emldetect likelihood to the cell-detection likelihood and then converts that threshold into a flux limit.Source: §4.1 · p. 9
  • Shifting the simulation grid points changes flux limits by less than 10% at maximum, which the authors adopt as the maximum relative systematic error in the sensitivity maps.The difference is mostly within a few percent.Source: §4.1 · p. 10

4.2. Results and Comparison with Other Work

The SXDS survey derives source-count relations across four X-ray bands, with fitted results broadly consistent with earlier work. Its counts also show band-specific differences from COSMOS and quantify the survey’s flux reach and resolved X-ray background.

Results and Comparison with Other Work

  • The area curves combine the seven pointings by selecting the most sensitive pointing at each position in overlapping regions.Differential counts are obtained by dividing the observed flux distribution by Ω(S); sources below the local sensitivity limit are excluded.Source: §4.2 · p. 10
  • The differential counts are fit with a broken power law, using a fixed upper slope in three bands and a single power law in 4.5–10 keV, where no break is evident.The best-fit parameters are consistent within uncertainties with previous results; Table 7 summarizes the fits and Figure 9 shows the relations and fitted curves.Source: §4.2 · p. 10, Table 7 · p. 10, Figure 9 · p. 11
  • Survey sensitivities are 6 × 10^-16, 8 × 10^-16, 3 × 10^-15, and 5 × 10^-15 erg cm^-2 s^-1 across the 0.5–2, 0.5–4.5, 2–10, and 4.5–10 keV bands, respectively.Above those fluxes, the corresponding resolved X-ray-background fractions are 88 ± 5%, 75 ± 3%, 74 ± 4%, and 52 ± 3%.Source: §4.2 · p. 10, Figure 8 · p. 11, Figure 10 · p. 12
  • In the 2–10 keV band, COSMOS source counts are 20–30% below SXDS and Carrera et al.; the authors suggest the converted 2–4.5 keV survey may miss hard-spectrum sources or underestimate fluxes.At S < 3×10^-14 erg cm^-2 s^-1, SXDS 0.5–2 keV counts are larger than both comparisons but consistent with the XMM-Newton Lockman Hole results.Source: §4.2 · p. 10
  • Across the four bands, SXDS log N–log S results agree with previous work within ≈10%.Figure 11 compares the SXDS relations with results from Carrera et al. and COSMOS.Source: §4.2 · p. 10, Figure 11 · p. 12

5.1. Spatial Distribution

The seven SXDS pointings show significant variation in X-ray source counts, demonstrating cosmic variance on the scale of an XMM-Newton field of view. X-ray clustering also provides a way to study AGN environments and their underlying dark-matter halos.

Spatial Distribution

  • X-ray AGN clustering can trace large-scale structure and help relate AGN environments to the mass of their underlying dark-matter halos.Source: §5.1 · p. 13
  • SDS7 has 0.5–2 and 0.5–4.5 keV source counts more than 30% below other fields at medium to high fluxes.The counts vary among pointings depending on flux limit, and the analysis excludes shorter exposures in overlapping regions so the results are statistically independent.Source: §5.1 · p. 13
  • The field-to-field variation indicates cosmic variance over an area scale of about 0.2 deg^2, larger than the Chandra field of view.The comparison uses source counts derived separately for the seven pointings.Source: §5.1 · p. 13

5.2. Auto Correlation Function

The angular correlation analysis detects clustering in the 0.5–2 keV sample but not significantly in the 2–10 keV sample. It estimates the soft-band correlation length and derives spatial clustering properties under a comoving-clustering assumption.

5.2. Auto Correlation Function

  • The ACF is estimated with the Landy–Szalay estimator and a 10-times-larger random sample, using a conservative 1.1 sensitivity safety factor.The authors report essentially the same results with a safety factor of 1.0.Source: §5.2 · p. 13
  • The fit fixes γ = 1.8 and derives θc over 0.5–10′, accounting for bin covariance and an integral-constraint offset.The fit range reflects the XMM-Newton PSF size.Source: §5.2 · p. 13
  • The study detects an ACF signal from the 0.5–2 keV selected sample, reporting a characteristic angular scale of θc = 5.9″ + 1.0″.Source: §5.2 · p. 13
  • The soft-band AGN have a comoving correlation length of 14.9±1.1 h−1 Mpc and a bias parameter of about 5.8±0.5 at median redshift 1.3.The hard-band comoving correlation length is < 7.6 h−1 Mpc at median redshift 1.1.Source: §5.2 · pp. 13–14
  • The soft-band correlation length agrees with earlier soft X-ray surveys but exceeds optically selected QSO estimates; no significant 4.5–10 keV ACF signal is found.The 4.5–10 keV result is more susceptible to statistical fluctuation because few sources are detected.Source: §5.2 · p. 14

6. LONG TERM X-RAY VARIABILITY OF SOURCES

Two SDS-4 observations about 2.5 years apart were used to compare source fluxes between epochs. Many intermediate-flux sources vary significantly, illustrating the relevance of variability to AGN studies.

6. LONG TERM X-RAY VARIABILITY OF SOURCES

  • The two SDS-4 observations were separated by about 2.5 years, had similar exposures exceeding 20 ksec, and were analyzed separately at fixed catalog positions.The sample required summed ML > 15 in the 0.5–2 and 2–4.5 keV bands and over 15 ksec of corrected exposure in both epochs.Source: §6 · pp. 14–15
  • Many intermediate-flux sources with small error bars vary significantly between epochs, consistent with prior deep-field reports of variable-source fractions of 80–90% or higher.The flux comparison is for the 0.5–4.5 keV band.Source: §6 · p. 15, Figure 16 · p. 15
  • The observed variability demonstrates its impact on studies of AGN properties.Source: §6 · p. 15

7. SUMMARY

The paper presents the first SXDS X-ray source catalog and basic source properties from seven XMM-Newton pointings covering 1.14 deg2. The catalog bridges survey flux ranges, and the analysis finds soft-band clustering and long-term variability.

7. SUMMARY

  • The catalog contains 1245 sources from seven XMM-Newton pointings covering 1.14 deg2 in the 0.2–10 keV band, with detection likelihood greater than 7.Source: §7 · p. 15
  • The log N - log S relations bridge the flux range between Chandra deep surveys and brighter surveys, while the soft-band sample clusters and the hard-band sample has only an upper limit.The soft-band correlation length is 5.9″ +1.0″/−0.9″; the 2–10 keV upper limit is θc < 1.5″ at 90% confidence.Source: §7 · p. 15
  • Two pointings separated by 2.5 years indicate the importance of flux variability on year timescales.The data establish average X-ray-source properties at fluxes where a majority of the X-ray background is produced and are described as least affected by cosmic variance.Source: §7 · p. 15
Loading 0806.2846v1…