The Cosmological Evolution of AGN

X-ray Luminosity Function

Yoshihiro UedaYoshihiro Ueda (Kyoto University) (Kyoto University)

Guenther Hasinger (MPE)Guenther Hasinger (MPE)

Takamitsu Miyaji (Carnegie Mellon)Takamitsu Miyaji (Carnegie Mellon)

1 deg Log N log S relations (2-10 keV)

Kushino+ 02

The XRB is the integrated emission from all the AGNs in the universe, The XRB is the integrated emission from all the AGNs in the universe, telling us the formation history of supermassive black holes. telling us the formation history of supermassive black holes.

At least below 6-8 keV it is now almost completely resolved into At least below 6-8 keV it is now almost completely resolved into discrete sources, mainly AGNs. discrete sources, mainly AGNs.

Subaru-XMM Deep Survey fields

The X-ray Background (XRB)The X-ray Background (XRB)

The necessity of hard band The necessity of hard band surveyssurveys The energy density peaks at ~ 30 keVThe energy density peaks at ~ 30 keV

The shape of the XRB indicates that The shape of the XRB indicates that most of the AGNs are obscuredmost of the AGNs are obscured (such (such as Seyfert 2; Awaki et al. 1993). as Seyfert 2; Awaki et al. 1993).

Hard band surveys (above 2 keV) are indispensable to detect obscured Hard band surveys (above 2 keV) are indispensable to detect obscured AGNs. AGNs.

Sensitive surveys below 10 keV, currently available, can provide us with a Sensitive surveys below 10 keV, currently available, can provide us with a complete picture of “complete picture of “Compton thinCompton thin” AGN (log ” AGN (log NNHH<24) in the universe.<24) in the universe.

It is critical to establish the cosmological evolution of Compton thinIt is critical to establish the cosmological evolution of Compton thinAGNs, in order to evaluate the role of "AGNs, in order to evaluate the role of "Compton thickCompton thick" AGNs in the" AGNs in theaccretion history, a major theme in next generation X-ray astronomy.accretion history, a major theme in next generation X-ray astronomy.

Comastri+ 95

Solving the XRB originSolving the XRB origin

Two major elementsTwo major elements

1. X-ray Luminosity function (XLF)1. X-ray Luminosity function (XLF) The co-moving spatial number density of AGNs as a function of The co-moving spatial number density of AGNs as a function of

luminosity and redshiftluminosity and redshift Surveys in the 0.5-2 keV band: Surveys in the 0.5-2 keV band: mostly type1 AGNs (mostly type1 AGNs (log log NNH H <22)<22) Surveys in the 2-10 keV band: Surveys in the 2-10 keV band: type1+type2 Compton thin AGNs type1+type2 Compton thin AGNs (log (log NNH H <24)<24)

2. Absorption function (2. Absorption function (NNH H function) function)

Probability distribution function of absorption column density of Probability distribution function of absorption column density of an AGN at a given luminosity and a redshift, giving the fraction an AGN at a given luminosity and a redshift, giving the fraction of absorbed (log of absorbed (log NNH H >22) AGNs.>22) AGNs.

A fundmental quantity for understanding of the AGN A fundmental quantity for understanding of the AGN phenomenaphenomena

Is “unified scheme” correct? Is “unified scheme” correct? Evolution?Evolution?

Hard band sample is requiredHard band sample is required

NNHH distribution of AGNs distribution of AGNs in the local universein the local universe

The Swift/BAT and Integral hard X-ray surveysThe Swift/BAT and Integral hard X-ray surveys above 15 keV show above 15 keV show that absorbed AGNs are indeed a major population. The fraction of that absorbed AGNs are indeed a major population. The fraction of

absorbed sources with (log absorbed sources with (log NNH H > 22) is 0.5-0.6. > 22) is 0.5-0.6. The results of softer-band surveys are consistent with the Swift The results of softer-band surveys are consistent with the Swift

result after correcting for selection bias against absorbed sourcesresult after correcting for selection bias against absorbed sources

HEAO1 (2-10 keV)Shinozaki+ 2006

RXTE/ASM (3-20 keV)Szanov & Revnivtsev 2004

Swift/BAT (15-200 keV)Markwardt+ 2005

20 21 22 23 24

log NH

< 22 23 24

log NH

20 22 24 26

log NH

NNHH distribution of AGNs distribution of AGNsin the high redshift universein the high redshift universe

Due to the statistical fluctuation in photon counts, an Due to the statistical fluctuation in photon counts, an unabsorbed unabsorbed AGN AGN easily appear as an easily appear as an absorbed absorbed source if we rely on the best-fit source if we rely on the best-fit NNHH. More . More problematic at higher redshift. (see also Akylas+06)problematic at higher redshift. (see also Akylas+06)

In small photon statistics data (especially in Chandra surveys), evaluation In small photon statistics data (especially in Chandra surveys), evaluation of systematic errors in the observed of systematic errors in the observed NNHH distribution is indispensable distribution is indispensable

Observed NH distribution of the CDFSHard-band selected AGNs

Simulated histogram of NH distribution for an “unabsorbed” AGN at z=2

Tozzi+ 2005

Absorption Fraction Absorption Fraction The most straightforward way to avoid this is to use only

sources that have spectral information with sufficient photon statistics. XMM data are particularly useful for Chandra deep survey sources.

Our present analysis: 1. HEAO1 + ASCA/XMM follow-up (Piccinotti+82, Shinozaki+06)

2. XMM Hard Bright Sample (Caccianiga+ 04)3. XMM Lockman Hole 800 ks (Hasinger+02, Matteos+05)4. CDFS + XMM 400 ks (Giacconi+02, Streblyanska+06 in prep)

While the luminosity dependence of absorption fraction is significant, its redshift dependence is not (Hasinger’s talk this morning)

Construction of an ultimate Construction of an ultimate XLFXLF

combined analysis of the HXLF combined analysis of the HXLF and SXLFand SXLF There is still room for improvement of the population There is still room for improvement of the population

synthesis model after Ueda+ 03 (e.g., more precise synthesis model after Ueda+ 03 (e.g., more precise modeling of LF and absorption function, Hasinger’s talk)modeling of LF and absorption function, Hasinger’s talk)

Best constrain the rest-frame 2-10 keV LF of all Compton-Best constrain the rest-frame 2-10 keV LF of all Compton-thin AGNs using all the heritage of X-ray surveys with thin AGNs using all the heritage of X-ray surveys with various depth, width, and energy bands performed up to various depth, width, and energy bands performed up to date.date.

While hard band surveys above 2 keV are indispensable to While hard band surveys above 2 keV are indispensable to detect type-2 AGNs at low redshifts, soft band (0.5-2 keV) detect type-2 AGNs at low redshifts, soft band (0.5-2 keV) surveys are also effective to detect type-2 AGNs at high surveys are also effective to detect type-2 AGNs at high redshifts thanks to the “K-correction” effect. Sensitivities redshifts thanks to the “K-correction” effect. Sensitivities achieved by X-ray reflecting mirrors are generally better in achieved by X-ray reflecting mirrors are generally better in the soft band than in the hard band, leading to a deeper the soft band than in the hard band, leading to a deeper flux limit even for type 2 AGNs.flux limit even for type 2 AGNs.

Utilize only samples with high identification completeness Utilize only samples with high identification completeness (>90%)(>90%)

The analysis methodThe analysis method Maximum likelihood methodMaximum likelihood method applied to the most direct applied to the most direct

observationalobservationalquantities i.e.,list of quantities i.e.,list of count ratecount rate (flux)(flux) and and redshift redshift without any without any correction.correction.

For a given model of luminosity function plus absorption For a given model of luminosity function plus absorption function, we can calculate an expected count rate vs redshift function, we can calculate an expected count rate vs redshift distribution for each survey, by fully taking account of the distribution for each survey, by fully taking account of the energy response of instruments used for the survey.energy response of instruments used for the survey. Spectral shape of an AGN and its varianceSpectral shape of an AGN and its variance can be incorporated can be incorporated into the input model. into the input model.

Find a solution that maximize the total probability of finding Find a solution that maximize the total probability of finding thetheobserved quantities. Soft and hard bands can be regarded as observed quantities. Soft and hard bands can be regarded as statistically independent measurement even for common statistically independent measurement even for common sources.sources.

In principle, luminosity function and absorption function can beIn principle, luminosity function and absorption function can beconstrained simutaneously. constrained simutaneously.

Sample: 1514 detectionsSample: 1514 detections Survey N flux limitSurvey N flux limit HEAO-1 49 1.7x10HEAO-1 49 1.7x10-11-11

ASCA MSS/LSS 125 1x10ASCA MSS/LSS 125 1x10-13-13

XMM LH 84 5x10XMM LH 84 5x10-15 -15 CLASXS, CDFN/S 208 1.1x10CLASXS, CDFN/S 208 1.1x10-15-15

ROSAT/XMM/Chandra 1048 1.1x10ROSAT/XMM/Chandra 1048 1.1x10-16-16

Redshift distributionRedshift distribution

Hard band detected sample Soft band detected sample

-1.5 -1 -0.5 1 1.5 -1.5 -1 -0.5 1 1.5Log z Log z

Log N log S relationsLog N log S relations

Hard band detected sample Soft band detected sample

The latest rest-frame 2-10 keV XLF of all Compton-thin AGNs, showing the “LDDE” behavior.

HXLF HXLF

The AGN number density as a The AGN number density as a function of redshiftfunction of redshift

As previously found by Ueda+03 As previously found by Ueda+03 (type1+2) and Hasinger+05 (type1+2) and Hasinger+05 (type 1), the cosmological (type 1), the cosmological evolution of the whole AGN is evolution of the whole AGN is described with a luminosity described with a luminosity dependent density evolution dependent density evolution (LDDE) where the cut-off redshift (LDDE) where the cut-off redshift increases with the luminosity increases with the luminosity

Luminous AGNs have a density Luminous AGNs have a density peak earlier in the cosmic time peak earlier in the cosmic time than less luminous AGNs. than less luminous AGNs.

By assuming L~M, more By assuming L~M, more massive SMBHs formed earlier massive SMBHs formed earlier than less massive SMBHsthan less massive SMBHs (“anti-hierarchical evolution” or (“anti-hierarchical evolution” or

“down sizing”)“down sizing”)

Predicted XRB spectrumPredicted XRB spectrum The XRB intensity at 10 keV is ~10 % lower than The XRB intensity at 10 keV is ~10 % lower than

the previous model, which does not utilize the the previous model, which does not utilize the CDFS sampleCDFS sample

Ueda+ 03 (only Compton thin AGNs) This work (only Compton thin AGNs)

SummarySummary The fraction of X-ray absorbed-AGN decreses The fraction of X-ray absorbed-AGN decreses

with luminosity, while its redshift dependence with luminosity, while its redshift dependence is not significant. Fully consistent with the is not significant. Fully consistent with the result from classification utilizing optical result from classification utilizing optical spectra (Hasinger’s talk).spectra (Hasinger’s talk).

An ultimate 2-10 keV XLF of all Compton-thin An ultimate 2-10 keV XLF of all Compton-thin AGNs is being derived, confirming the “down-AGNs is being derived, confirming the “down-sizing” nature of BH growth.sizing” nature of BH growth.

The XRB synthesis model is close to its The XRB synthesis model is close to its finalization, although there still remain issues:finalization, although there still remain issues: The true number density of Compton thick AGNs, The true number density of Compton thick AGNs,

which is coupled with an assumption of broad band which is coupled with an assumption of broad band spectra of type1 and type2 AGNs (especially the spectra of type1 and type2 AGNs (especially the amount of reflection) →sensitive hard X-ray (>10 amount of reflection) →sensitive hard X-ray (>10 keV) surveys are imortant.keV) surveys are imortant.