Abstract
TitleofDissertation: Extragalactic X-ray Surveys of ULXs and AGNs
Lisa M. Winter, Doctor of Philosophy, 2008
Dissertation directed by: Professor Christopher S. Reynolds, Advisor
Dr. Richard F. Mushotzky, Co-advisor
Department of Astronomy
Extragalactic X-ray studies provide unique opportunities for studying accret-
ing black holes. In particular, they are necessary for studying phenomena not
easily selected or observed in other wavelengths. Among these objects, ultra-
luminous X-ray sources (ULXs) emit the vast majority of their luminosity in the
X-ray band and are very faint or confused in other wavebands. Similarly, heavily
obscured active galactic nuclei (AGN) with absorbing columns > 1024 cm?2 are
rarely detected in optical surveys, due to the extreme reddening. In my thesis, I
study both phenomenon in the local universe.
AtULX luminosities [LX (0.3 ? 10 keV)> 3?1039 ergs?1], the EuropeanSpace
Agency?s XMM-Newton satellite provides the spectral resolution and sensitivity
necessary to study the population of local ULXs. Thus, we conducted an XMM-
Newton archival study of a complete sample of the ULXs located within 8Mpc.
Our study confirmed key predictions ofthe intermediate mass black hole (IMBH)
hypothesis for local ULXs. We then followed-up this study by investigating high
signal-to-noise XMM-Newton observations of 14 ULX sources ? studying their
spectral shape, testing the validity of different accretion disk and power law
models, and then using absorption of their spectra to measure the oxygen and
iron abundances of the interstellar medium of their host galaxies.
New breakthroughs are expected in the study of heavily obscured AGN from
SWIFT. The SWIFT satellite, launched in 2004, has detected a sample of 153 AGN
with the Burst Alert Telescope (BAT) in the first 9-months of data. The BAT is
sensitive in the 14 ? 195keV band and the selected sources have an average red-
shift of ? 0.03. Thus, it detects local AGN without bias towards all but the most
obscured sources (nH > 1024 atomscm?2). The BAT AGN sources are the result
of an all-sky survey with a flux limit of FBAT > 10?11 ergs?1 cm?2. I analyze and
present the results of X-ray data from XMM-Newton, ASCA, as well as SWIFT?s
XRT (0.3 ? 10keV) and BAT, in order to understand the properties of obscured
and unobscured AGN in the local universe. Among our results, we show that the
new class of ?hidden?/buried AGN are a significant population of local AGN
(? 20%). We also find that our data supports the need for a modified AGN uni-
fied model ? one which includes a luminosity dependence.
Extragalactic X-ray Surveys of ULXs and AGNs
by
Lisa M. Winter
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland at College Park in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
2008
Advisory Committee:
Professor Christopher S. Reynolds, Advisor
Dr. Richard F. Mushotzky, Co-advisor, chair
Professor M. Coleman Miller
Professor Sylvain Veilleux
Professor Edward Sion
Professor Gregory Sullivan, Dean?s Representative
c? Lisa M. Winter 2008
Preface
Much of the work presented in this thesis has been published or presented at
a professional conference. The work presented in Chapter2 of this thesis has
been published in the Astrophysical Journal as XMM-Newton Archival Study of the
ULX Population in Nearby Galaxies (Winter et al. 2006b). Preliminary results from
this study, as well as additional work not presented in the thesis, were presented
as an oral talk at the 2004 Winter American Astronomical Society Meeting in
San Diego, CA, as Investigating The Ultraviolet Environment of Ultraluminous X-
ray sources (Winter et al. 2004). Additionally, work from this Chapter was pre-
sented as an oral talk at the 2005 Winter American Astronomical Society Meeting
in Washington, D.C., as Bolstering the IMBH hypothesis: Evidence for ?Low? and
?High? State ULX Sources from XMM-Newton (Winter et al. 2005).
The work presented in Chapter3 has been published in the Astrophysical Jour-
nal as Elemental Abundances of Nearby Galaxies through High Signal-to-Noise Ratio
XMM-Newton Observations of Ultraluminous X-Ray Sources (Winter et al. 2007b).
The results were also presented as a poster presentation (as well as in the con-
ference proceedings) entitled Oxygen and Iron Abundances in nearby Galaxies Using
Ultraluminous X-ray Sources at the 2005 European Space Agency?s The X-ray Uni-
verse Conference in El Escorial, Spain (Winter et al. 2006a).
The work presented in Chapter4 has been published in the Astrophysical Jour-
nalasX-rayPropertiesofanUnbiasedHardX-rayDetectedSampleofAGN(Winter et al.
ii
2008). Portions of this work were presented in an oral presentation at the 2006
Chandra Extragalactic Surveys meeting in Boston, MA, as Early Results from the
SWIFT BAT survey: XMM follow-up Observations for 22 Sources. Additionally, por-
tions of this work were presented in a poster presentation at the 2007 Summer
meeting of the American Astronomical Society Meeting in Honolulu, HI, as Early
Results From Swifts BAT AGN Survey (Winter et al. 2007c).
The work presented in Chapter5 has been submitted for review to the As-
trophysical Journal on April 30, 2008. A portion of this work as well as parts of
Chapter4 were presented as an oral dissertation presentation at the 2008 Win-
ter American Astronomical Society Meeting in Austin, TX, as X-ray and Optical
Properties of an Unbiased Sample of Local AGN (Winter et al. 2007a).
iii
Acknowledgements
As I finish my over 20 years of formal education, there are many peo-
plethatIwouldliketothankforhelpingmealongtheway. Firstofall,
as the people who helped me through the largest part of these years,
I?d like to thank my parents. They have offered me support in many
ways and have made countless sacrifices to give me and my sisters
and brother the opportunities that we have. In a special way, I?d like
to thank my dad for being a shining example of hard work and com-
mitment and my mom for always believing in me and always being
incredibly optimistic.
I would also like to thank my husband, Eric Barron, for his help and
support. He introduced me to the Python programming language,
which greatly improved my sanity in allowing me to easily manipu-
late and plot the data shown. He also helped to keep me grounded
throughout the thesis writing process.
My undergraduate years are especially dear to me and for that I?d
like to thank the faculty and staff of the Villanova University Astron-
iv
omy/Astrophysics Department, as well as my fellow classmates. I re-
ceived a truly excellent education from an excellent group of people.
In a special way, I?d like to thank Edward Sion for being a wonderful
role model as well as advisor. I?d also like to thank Drs. Ambruster,
Fitzpatrick, Guinan, Maloney, and McCook, in addition to Larry De-
Warf, Dr. Elizabeth Jewell, and Dr. Morgan Besson (Physics).
At the University of Maryland, I would first like to thank all of my
teachers, especially Derek Richardson who always went above and
beyond his duties. I also thank my fellow classmates, particularly El-
yse Casper, who I struggled together with from Radiative Processes
through Math Methods, as well as Katie McGleam and Ashley Za-
uderer. Without your moral support, homework as well as studying
for the qualifiers, would have been even more exhausting. Addition-
ally, I would like to thank Cole Miller, for encouragement as well as
reading multiple second year project drafts, and both Sylvain Veilleux
and Karen Lewis, for help with optical proposals and observations.
Further, I would like to thank the staff at the Astronomy Department,
particularly Mary Ann Phillips and Adrienne Newman, for all of their
help in filling out the various required forms.
Finally, I am especially grateful to my advisors, Richard Mushotzky
and Chris Reynolds for their guidance. Richard?s enthusiasm, creativ-
ity, and extensive knowledge of 25+ years of X-ray astronomy (as well
as nearly any other astronomy question I ask him) is a constant inspi-
ration. I would like to thank Chris for his willingness to fit me into his
extremely busy schedule, his patience in answering and explaining
my questions, and his constant optimistic support.
v
Contents
List of Tables x
List of Figures xii
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 X-ray spectra of Accreting Black Holes . . . . . . . . . . . . . . . . . 4
1.3 The X-ray Observing Tools . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.1 XMM-Newton . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3.2 The SWIFT Gamma-ray Observatory . . . . . . . . . . . . . 13
1.3.3 Additional X-ray Observatories . . . . . . . . . . . . . . . . . 15
1.3.4 Analysis of the X-ray Spectra . . . . . . . . . . . . . . . . . . 16
1.4 The Multiple Types of AGN . . . . . . . . . . . . . . . . . . . . . . . 20
2 ULX Survey 24
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.2 Observations and Data Reduction . . . . . . . . . . . . . . . . . . . 28
2.2.1 Sample Selection . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.2.2 Data Reduction . . . . . . . . . . . . . . . . . . . . . . . . . . 35
2.3 Spectral Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.3.1 Single-Component Sources . . . . . . . . . . . . . . . . . . . 37
2.3.2 Two-Component Sources . . . . . . . . . . . . . . . . . . . . 44
2.3.3 Additional Sources . . . . . . . . . . . . . . . . . . . . . . . . 53
2.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.4.1 Classification Criteria . . . . . . . . . . . . . . . . . . . . . . 56
2.4.2 Low-State ULXs . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.4.3 High-State ULXs . . . . . . . . . . . . . . . . . . . . . . . . . 61
2.4.4 Temperature Gap . . . . . . . . . . . . . . . . . . . . . . . . . 65
2.4.5 Comparison with Galactic HMXBs . . . . . . . . . . . . . . . 66
2.4.6 Galaxy Sample . . . . . . . . . . . . . . . . . . . . . . . . . . 69
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
vi
3 Elemental Abundances of Nearby ULX Host Galaxies 77
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.2 Source Selection and Data Reduction . . . . . . . . . . . . . . . . . . 80
3.3 Spectral Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.4 Nature of the Soft Component . . . . . . . . . . . . . . . . . . . . . . 89
3.4.1 Comparison of Thermal Disk Models . . . . . . . . . . . . . 93
3.4.2 Power law/Thermal disk Degeneracies . . . . . . . . . . . . 95
3.4.3 Physical Plausibility of the Accretion Disk models . . . . . . 99
3.5 Properties of the ISM in ULX Host Galaxies . . . . . . . . . . . . . . 102
3.5.1 Column Densities . . . . . . . . . . . . . . . . . . . . . . . . . 103
3.5.2 Elemental Abundances . . . . . . . . . . . . . . . . . . . . . . 106
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
4 A Representative Sub-Sample of the SWIFT BAT AGN Survey 119
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
4.1.1 Selection Effects . . . . . . . . . . . . . . . . . . . . . . . . . . 120
4.1.2 Other Hard X-ray Surveys . . . . . . . . . . . . . . . . . . . . 123
4.1.3 This Chapter . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.2 Observations and Data Analysis . . . . . . . . . . . . . . . . . . . . 125
4.2.1 XMM-Newton and SWIFT XRT Spectra . . . . . . . . . . . . 125
4.2.2 SWIFT BAT Spectra . . . . . . . . . . . . . . . . . . . . . . . . 131
4.3 Spectral Fitting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
4.3.1 Detailed Spectral Properties . . . . . . . . . . . . . . . . . . . 135
4.3.2 Variability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
4.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
4.4.1 Spectral Classification . . . . . . . . . . . . . . . . . . . . . . 172
4.4.2 X-ray Colors . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
4.4.3 Properties of the Spectral Components . . . . . . . . . . . . 179
4.4.4 Variability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
5 The Complete X-ray Properties of the BAT AGN 190
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190
5.2 The Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191
5.2.1 Source Selection . . . . . . . . . . . . . . . . . . . . . . . . . . 191
5.2.2 X-ray Data and Analysis . . . . . . . . . . . . . . . . . . . . . 193
5.3 Properties of the Spectra . . . . . . . . . . . . . . . . . . . . . . . . . 196
5.4 The ?Uniform? Sample . . . . . . . . . . . . . . . . . . . . . . . . . . 202
5.4.1 Distribution of z, nH, and ? . . . . . . . . . . . . . . . . . . . 202
5.4.2 Distribution of Lcorr2?10keV and Lcorr2?10keV /LEdd . . . . . . . . . . 207
5.4.3 Correlations of ? with Lcorr2?10keV and Lcorr2?10keV /LEdd . . . . . . 214
vii
5.4.4 The Fe-K? Feature . . . . . . . . . . . . . . . . . . . . . . . . 216
5.4.5 Simple Power Law Model Sources . . . . . . . . . . . . . . . 221
5.4.6 Complex Sources . . . . . . . . . . . . . . . . . . . . . . . . . 225
5.5 The Cosmic X-ray Background . . . . . . . . . . . . . . . . . . . . . 232
5.5.1 The Average X-ray Spectrum . . . . . . . . . . . . . . . . . . 232
5.5.2 The 2-10keV Log N - Log S Relationship . . . . . . . . . . . 235
5.6 Host Galaxy Properties . . . . . . . . . . . . . . . . . . . . . . . . . . 238
5.6.1 Host Inclination . . . . . . . . . . . . . . . . . . . . . . . . . . 238
5.6.2 Host Morphology . . . . . . . . . . . . . . . . . . . . . . . . . 241
5.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 249
6 Conclusions 254
6.1 ULX Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
6.1.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 255
6.1.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 257
6.2 SWIFT BAT-detected AGN Survey . . . . . . . . . . . . . . . . . . . 258
6.2.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 258
6.2.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
A Glossary 264
A.1 Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
A.2 Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
A.3 Units . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
A.4 Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 266
B Appendix to Chapter 2 267
B.1 Spectral Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 267
B.1.1 Two-component Model . . . . . . . . . . . . . . . . . . . . . 267
B.1.2 Simple Power Law Model . . . . . . . . . . . . . . . . . . . . 272
B.2 Additional Spectral Fits . . . . . . . . . . . . . . . . . . . . . . . . . 274
B.2.1 NGC300 XMM4 . . . . . . . . . . . . . . . . . . . . . . . . . . 274
B.2.2 NGC4631 XMM4 . . . . . . . . . . . . . . . . . . . . . . . . . 274
B.2.3 NGC4631 XMM5 . . . . . . . . . . . . . . . . . . . . . . . . . 275
B.2.4 NGC4945 XMM5 . . . . . . . . . . . . . . . . . . . . . . . . . 275
B.2.5 M51 XMM5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
B.2.6 M83 XMM2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
B.2.7 Inverse Compton Scattering Sources . . . . . . . . . . . . . . 276
B.3 Additional Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
C Appendix to Chapter 3 292
C.1 Spectral Simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
viii
D Appendix to Chapter 4 297
D.1 Details on Variability of Individual Sources . . . . . . . . . . . . . . 297
D.1.1 Low Column Density Sources . . . . . . . . . . . . . . . . . . 297
D.1.2 High Column Density/Complex Sources . . . . . . . . . . . 303
E Appendix to Chapter 5 307
E.1 Additional Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
E.2 Notes on Individual Sources . . . . . . . . . . . . . . . . . . . . . . . 325
Bibliography 330
ix
List of Tables
2.1 XMM-Newton Galaxy Observations . . . . . . . . . . . . . . . . . . . 30
2.2 XMM-Newton best fit: single component spectral fits . . . . . . . . . 38
2.3 Bolometric Luminosities of ULX sources . . . . . . . . . . . . . . . . 41
2.4 XMM-Newton best fit: blackbody and power law fits . . . . . . . . . 45
2.5 XMM-Newton two-component fits for sources with large uncertainty 51
3.1 ULX Observation Details . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.2 Model Fits to EPIC spectra with the grad Model . . . . . . . . . . . 92
3.3 Model Fits to the RGS spectra with the grad Model . . . . . . . . . 92
3.4 Alternative Thermal Disk Model Fits . . . . . . . . . . . . . . . . . . 93
3.5 Spectral Fits for Sources with Degenerate Solutions . . . . . . . . . 98
3.6 Alternative Column Density Measurements . . . . . . . . . . . . . 104
3.7 Spectral Fits to the Oxygen Edge . . . . . . . . . . . . . . . . . . . . 108
4.1 Details on the Sub-sample of 22 Sources . . . . . . . . . . . . . . . . 127
4.2 XMM-Newton EPIC and SWIFT XRT Observation Information . . . 128
4.3 XMM-Newton Spectral Analysis: Absorbed Power Law Fits . . . . 137
4.4 SWIFT XRT Spectral Analysis: Absorbed Power Law Fits . . . . . . 138
4.5 XMM-Newton Detailed Fits For Power Law Sources . . . . . . . . . 141
4.6 XMM-Newton Detailed Fits for Power Law + Blackbody Sources . 141
4.7 XMM-Newton Warm Absorber Model . . . . . . . . . . . . . . . . . 143
4.8 XMM-Newton Detailed Fits: Partial Covering Model . . . . . . . . 147
4.9 XMM-Newton Detailed Fits: Double Power Law Model . . . . . . . 148
4.10 Compton-thick Reflection Model . . . . . . . . . . . . . . . . . . . . 154
4.11 Variability in individual XMM-Newton Observations . . . . . . . . . 159
4.12 Variability between XMM-Newton and XRT Observations . . . . . . 166
5.1 K-band Derived Mass and Lcorr2?10keV /LEdd . . . . . . . . . . . . . . . 211
B.1 Bright Point Sources Examined . . . . . . . . . . . . . . . . . . . . . 278
B.2 Bright, Identifiable Background and Foreground Sources . . . . . . 285
B.3 XMM-Newton Power Law Fit for Best Fit Two-component Spectra . 286
B.4 Best-Fit Absorbed Comptonization Model Parameters . . . . . . . . 290
x
B.5 XMM-Newton Galaxy Observations . . . . . . . . . . . . . . . . . . . 291
C.1 Spectral Fits to Simulated HolmIX XMM1 Spectrum . . . . . . . . . 295
E.1 X-ray Observation Details . . . . . . . . . . . . . . . . . . . . . . . . 308
E.2 X-ray Fits of SWIFT XRT spectra . . . . . . . . . . . . . . . . . . . . 316
E.3 X-ray Spectral Fits from the Literature/Suzaku XIS/TARTARUS Fits 320
xi
List of Figures
1.1 A Model X-ray Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.2 The XMM-Newton Payload . . . . . . . . . . . . . . . . . . . . . . . . 10
1.3 EPIC CCDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.4 The Burst Alert Telescope . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5 The AGN Unified Model . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.1 Host Galaxy Distribution . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.2 Simple Single Component Models . . . . . . . . . . . . . . . . . . . 39
2.3 Example Two-Component Spectrum . . . . . . . . . . . . . . . . . . 48
2.4 Comparison of Single and Two Component Models . . . . . . . . . 49
2.5 ULX ratio of X-ray to Optical Flux . . . . . . . . . . . . . . . . . . . 55
2.6 Distribution of ULX spectral indices . . . . . . . . . . . . . . . . . . 59
2.7 ULX X-ray Luminosity vs. Spectral Index . . . . . . . . . . . . . . . 60
2.8 Distribution of Blackbody Temperature for High State ULXs . . . . 62
2.9 ULX Color-Color Diagram . . . . . . . . . . . . . . . . . . . . . . . . 68
2.10 Host Galaxy FIR Luminosity . . . . . . . . . . . . . . . . . . . . . . . 70
2.11 ULX Distribution of Column Densities . . . . . . . . . . . . . . . . . 73
3.1 EPIC spectrum of Holm IX XMM1 . . . . . . . . . . . . . . . . . . . 91
3.2 High Mass Model for NGC 5408 XMM1 . . . . . . . . . . . . . . . . 96
3.3 Low Mass Model for NGC 5408 XMM1 . . . . . . . . . . . . . . . . 97
3.4 MBH vs. Accretion Rate . . . . . . . . . . . . . . . . . . . . . . . . . 101
3.5 Comparison of Hydrogen Columns from Different Methods . . . . 105
3.6 Comparison of Oxygen Abundances from Different Methods . . . . 110
3.7 Host Galaxy MB versus ULX Oxygen Abundance . . . . . . . . . . 112
3.8 Oxygen Abundance versus Host Galaxy LFIR . . . . . . . . . . . . . 113
3.9 Iron Abundance versus Host Galaxy LFIR . . . . . . . . . . . . . . . 114
3.10 Iron Abundance versus Oxygen Abundance . . . . . . . . . . . . . 116
3.11 Oxygen Abundance versus ULX Position in the Host Galaxy . . . . 117
4.1 Comparison of XMM-Newton and SWIFT XRT S-N for Fe-K? . . . . 134
4.2 XMM-Newton Spectrum of NGC 1142 . . . . . . . . . . . . . . . . . 149
4.3 XMM-Newton and SWIFT BAT spectra of NGC 6860 . . . . . . . . . 155
xii
4.4 Spectra of Compton-thick Candidates . . . . . . . . . . . . . . . . . 156
4.5 XMM-Newton Light Curves Showing Short-term Variability . . . . 158
4.6 Flux Variation Between X-ray Observations . . . . . . . . . . . . . . 161
4.7 Distribution of the Variability Measurements . . . . . . . . . . . . . 164
4.8 Flux Variability vs. Time-scale for Variability . . . . . . . . . . . . . 165
4.9 XMM-Newton and SWIFT XRT spectra of MRK 352 . . . . . . . . . . 168
4.10 XMM-Newton and SWIFT XRT Spectra of ESO 362-G018 . . . . . . 169
4.11 SWIFT BAT Flux vs. nH for the 9-month Sample . . . . . . . . . . . 171
4.12 Column Density vs. Soft and Hard Colors . . . . . . . . . . . . . . . 177
4.13 Color-color Diagram Incorporating XMM-Newton and BAT Fluxes . 178
4.14 Fe K-? EW vs. L2?10keV . . . . . . . . . . . . . . . . . . . . . . . . . . 181
4.15 Measured Warm Absorption Features vs. L2?10keV . . . . . . . . . . 183
4.16 Variability in Spectral Parameters . . . . . . . . . . . . . . . . . . . . 185
5.1 Simple and Complex Models . . . . . . . . . . . . . . . . . . . . . . 197
5.2 Example X-ray Spectra . . . . . . . . . . . . . . . . . . . . . . . . . . 198
5.3 The Color-Color Plot for the Entire Sample . . . . . . . . . . . . . . 199
5.4 Redshift Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . 204
5.5 Column Density Distributions . . . . . . . . . . . . . . . . . . . . . . 205
5.6 ? Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
5.7 Absorption Corrected Lcorr2?10keV versus L14?195keV . . . . . . . . . . . 209
5.8 Distribution of Lcorr2?10/LEdd and Lcorr2?10 . . . . . . . . . . . . . . . . . . 212
5.9 ? versus Lcorr2?10 and Lcorr2?10/LEdd . . . . . . . . . . . . . . . . . . . . . . 215
5.10 Fe K-? EW versus Lcorr2?10 and Lcorr2?10/LEdd . . . . . . . . . . . . . . . . 218
5.11 Lcorr2?10/LEdd and Fe K EW vs. nH . . . . . . . . . . . . . . . . . . . . . 220
5.12 Blackbody Components . . . . . . . . . . . . . . . . . . . . . . . . . 223
5.13 L0.5?2.0 for Complex Sources . . . . . . . . . . . . . . . . . . . . . . . 227
5.14 Fraction of Obscured Sources . . . . . . . . . . . . . . . . . . . . . . 228
5.15 Average 0.1?10keV Spectrum . . . . . . . . . . . . . . . . . . . . . . 234
5.16 Log N?Log S in the 2?10keV Band . . . . . . . . . . . . . . . . . . . 236
5.17 Host Inclination vs. X-ray Column Density . . . . . . . . . . . . . . 239
5.18 Host Morphologies of the BAT AGN . . . . . . . . . . . . . . . . . . 243
5.19 DSS Images of Peculiar Hosts . . . . . . . . . . . . . . . . . . . . . . 244
5.20 Distribution of Lcorr2?10 and Lcorr2?10/LEdd . . . . . . . . . . . . . . . . . . 246
B.1 Distribution of ??2 from Simulations . . . . . . . . . . . . . . . . . 271
C.1 X-ray Counts Required to Determine Abundances . . . . . . . . . . 296
xiii
Chapter 1
Introduction
1.1 Motivation
The X-ray band provides one of the clearest views of the most powerful phenom-
ena in astronomy; accreting black holes. As well as being produced in copious
quantities, X-rays can easily penetrate surrounding matter and arrive at our de-
tectors millions of trillions of kilometers away. Therefore, systematic studies of
extragalactic X-ray sources supply the framework necessary for studying objects
that are too faint or difficult to identify in other wavelength bands. Two such
classes of objects are the ultra-luminous X-ray sources (ULXs) and obscured ac-
tive galactic nuclei (AGN).
ULXs are bright, non-nuclear X-ray sources with X-ray luminosities (0.3?
10keV) > 3 ? 1039 ergs?1 (Miller et al. 2004a). The Eddington luminosity, the
luminosity at which radiation force and gravitational force are exactly matched
for an object, is approximately 1.3?1038 M/M? ergs?1. If the X-ray luminosities
observed for the ULXs are both the Eddington luminosity and isotropic, i.e. not
the product of a beaming effect, the ULXs have masses > 20M?. If the ULXs
are scaled-up versions of Galactic black hole X-ray binaries, which typically have
1
masses ranging from 3 ? 15M? (Remillard & McClintock 2006), we would expect
their typical luminosities to be much lower, 0.01 ? 0.1LEdd (Gierli?nski & Done
2004). Thus, ULXs could represent a class of intermediate mass black holes
(IMBHs). As such, their masses are too high to be the product of normal stel-
lar evolution but well below those of the supermassive black holes residing in
the centers of most galaxies.
Despite their high X-ray luminosities, little progress has been made in iden-
tifying optical counterparts of ULXs. Of the known ULXs (well over 50), op-
tical counterparts were discovered for only a handful of sources at the closest
distances (NGC 5204 X-1 by Liu et al. (2004), M81 X-6 by Liu et al. (2002), and
M101 X-1 by Kuntz et al. (2005)). Unlike AGN, which typically have optical/X-
ray flux ratios of 1 (Anderson et al. 2003), Fopt/FX < 0.1 for ULXs (Winter et al.
2006b). The optical faintness of a ULX counterpart coupled with the distance to
the source makes opticalstudies extremely difficult. Thus, X-ray observations are
the best tools for studying these interesting sources.
Similar challenges exist for heavily obscured AGNs. While optical studies
reveal the strong narrow and broad emission lines indicative of AGN activity, X-
ray and optical surveys fail to select the same AGN samples (see the discussion
in Mushotzky (2004)). X-ray surveys in the soft (0.5 ? 2keV) and hard (2 ? 10keV)
bands identify more sources that are obscured by high column density absorb-
ing material in the line of sight. However, even 2?10keV X-rays are affected
by this absorption, making it difficult to detect sources with column densities
> 1024 cm?2 (Compton thick sources). This provides a major question for AGN
surveys: how many Compton thick sources exist? Matt et al. (2000) estimated
thatthenumbercouldbeashighasanorderofmagnitudemorethanunobscured
sources which are easily detected in optical and softer X-ray surveys. In order to
2
account for these additionalheavily obscured sources and determine their contri-
bution to the cosmic X-ray background, very hard X-ray (> 10keV) surveys are
needed. At these wavelengths, the AGN emission is penetrating enough to pass
through much of the gas and dust surrounding the AGN.
With the launch of the SWIFT satellite, we expect to gain much insight on the
nature of AGN sources. The Burst Alert Telescope (BAT), sensitive in the 14 ?
195keV band, detects unobscured high energy emission from AGNs throughout
the entire sky. With a much larger sample than other very hard X-ray missions
(such as HEAO-1 and Integral), BAT is the first sensitive, all-sky, very hard X-ray
survey in 28 years. The 9 month BAT AGN catalog consists of 153 AGN with an
averageredshift of0.03(Tueller et al.2008). Thus, analysisoftheX-ray properties
of these sources will provide a new picture of the population of AGN in the local
universe, including sources that were not selected in previous surveys.
This dissertation consists of two parts: 1) an X-ray survey of ULXs and 2)
an X-ray study of the SWIFT BAT-detected AGNs. The structure of the thesis
is as follows: a summary of the observed X-ray spectra of black holes and the
instruments used to examine them, as well as a description of AGN and associ-
ated problems, is discussed in Chapter 1. The survey on ULXs and results are
discussed in Chapter 2. Chapter 3 presents an application of this survey: us-
ing ULXs to study the properties of the ISM in their host galaxy. The next two
chapters (4 and 5) examine the properties of the SWIFT BAT-detected AGN in the
9-month survey. Chapter 4 is specifically a study of XMM-Newton spectra of a
representative sample of 22 sources whose X-ray spectra had never before been
studied. Chapter 5 is a larger study of the X-ray properties of the entire 9-month
BAT AGN catalog. Conclusions are then presented in Chapter 6, discussing both
the ULX and AGN studies.
3
1.2 X-ray spectra of Accreting Black Holes
Observationally, the main feature of an accreting black hole?s X-ray spectrum is
a power law. This is represented as N(E) ? E??, where ? is the photon index
and N(E) has units of photons per unit area per unit time. Alternatively, this is
represented as energy flux per unit area per unit time:
F(E) ? E ?E?? = E??+1 = E?(??1) (1.1)
Here, the value ??1 is sometimes referred to as ?, the spectral index.
The power law could be the result of either synchrotron emission from a jet
or inverse Compton scattering of the blackbody photons from the accretion disk.
In the first scenario, accelerated electrons spiraling in a magnetic field create the
synchrotron emission. In the second scenario, the power law continuum is pro-
duced in the hot corona of the accretion disk. The photons from the accretion
disk are up-scattered to higher energies from interactions with the hot corona
electrons.
Typically, accreting black holes have photon indices of ? = 1.5 ? 2.5. For the
Galactic X-ray binaries (stellar mass black holes), these values vary from ? ? 1.7
inthe?low/hard?stateto? ? 2.5inthe?veryhigh?state(Remillard & McClintock
2006). The low/hard state is associated with a lower accretion rate (L ? 0.1LEdd)
(Done & Gierli?nski 2003). Meanwhile, the very high state, as its name suggests,
corresponds to higher luminosities (> 0.1LEdd). Thus, ? seems to be correlated
with L/LEdd (the Eddington ratio). Similar relations between ? and Edding-
ton ratio are claimed for AGN spectra (Piconcelli et al. 2005; Porquet et al. 2004b;
Shemmer et al. 2006).
In addition to the power law component, X-ray spectra may exhibit other
4
spectral features such as a soft excess, a reflection component, emission lines,
and absorption features. We discuss these spectral features below.
(1) Soft Excess ? A soft excess is an excess of emission above the extrapolated
hard X-ray power law. Since this excess occurs below 1keV, it is referred to as
?soft? emission. Some possible causes are outlined below:
Accretion Disk: Thermal emission (blackbody emission) from the inner regions
of an accretion disk is observed in the high/soft or thermal dominated state of
X-ray binaries (accreting stellar mass black holes). For a black hole system, the
accretion disk emission is more likely the combination of blackbody emission
from multiple temperature regions in the inner disk. The peak temperature of
this emission is related to the mass of the central black hole, since the mass of the
black hole determines the inner radius of the disk (R ? M).
Forablackhole, asimplecalculationcanbemadetoestimatethedisk temper-
ature expected for a given black hole mass. Equating the black body luminosity
(L = 4?R2?T4) with the Eddington luminosity (L = 1.3 ? 1038M/M? ergs?1)
shows that T ? M?1/4 for a black hole. In units appropriate for X-ray observa-
tions, the temperature for a stellar mass black hole accretion disk is kT ? 1keV,
while a 1000M? black hole has kT ? 0.1keV. Extending this to supermassive
black holes, the peak disk temperature is lower and thus in the ultraviolet.
Therefore, it is unlikely that the soft excess seen in AGN, thought to be accret-
ing supermassive (M > 105M?) black holes, is from the disk. Here, the measured
temperature of the soft excess (kT ? 0.1keV) is too hot, the shape is much more
gradual than a Wien rollover, and the same effective temperature component is
seen for many AGN (kT = 0.12 ? 0.02keV) (Gierli?nski & Done 2004). A few
alternate possibilities are discussed in the literature. One possibility is that the
thermal emission is from the disk, as a result of Compton up-scattering of ther-
5
mal emission. In other words, the disk blackbody emission is up-scattered from
the optical/UV in a hotter region of the disk?s corona. Other possibilities include
the soft excess being the result of partially ionized material close to the disk mov-
ing at relativistic speeds. This can be explained as blurred absorption from a
wind above the accretion disk (Gierli?nski & Done 2004) or blurred emission from
matter reflected off of the disk.
Ionizedgasfromcircumstellarmatter/starformation: ForAGN,asoftexcesscould
be the result of ionized gas from circumstellar matter or star formation. A soft ex-
cessfromthisphenomenonismorelikelyintheX-rayspectrumoflowluminosity
AGN (LLAGN). In this case, the emission lines from the hot, ionized gas around
young stars can compete with the LLAGN emission. If the AGN were more lu-
minous, the spectral contribution from the hot gas (with LX ? 1041 ergss?1 cm?2)
would be lost to the much more powerful AGN emission.
A soft excess could also be seen if the host galaxy is undergoing very active
star formation. Here, the lines produced would be strong enough to compete
with the AGN emission. With lower spectral resolution, the emission lines from
hot ionized gas are indistinguishable and blend to appear as a thermal spectrum.
In addition to spectral resolution, the angular resolution of the telescope also
plays a part in viewing a soft excess. If the angular resolution is insufficient to
separate the AGN point source from nearby galaxy/star burst emission, a soft
excess will be seen.
(2) Reflected Emission: In addition to the more or less direct emission, which
is well-modeled by a power law, some portion of the direct emission may be
blocked but reflected into our line of sight. Thus, the emission we see is from
photons which are reflected from matter surrounding the black hole. Some of the
main features of the resultant reflection spectrum include an iron line from flu-
6
orescence, the iron edge, and a Compton hump resulting from Compton down-
scattering of photons due to electron recoil. There is also a thermal cutoff which
is related to the temperature of the corona (thought to be at about 200keV). This
cutoff occurs due to the reduction in energy exchange as the photon energy ap-
proaches the average energy of the electrons in the corona. In Figure 1.1, an ex-
ample is given for a model including reflection. The most prominent features
include the Fe emission line at 6.4keV, the Compton hump, and the cutoff at
200keV. In the figure, a power law model is also shown to illustrate more clearly
the significant differences between the models at higher energies (above 10keV).
In the model, the Fe emission line is clearly a very prominent feature in the
spectrum below 10keV. The Fe lines typically observed are 6.4keV (from cold
Fe I reflection), 6.67keV (Fe XXV), and 6.97keV (Fe XXVI). The last two Fe lines
are both from reflection of thermally excited or photoionized material. The pro-
cess which produces the 6.4keV line is known as fluorescence. Fluorescence is
the phenomena in which the absorption of a higher-energy photon results in re-
emission of a lower energy photon. Here, an inner level electron (at X-ray ener-
gies this is often the K shell or innermost level) is given enough energy from a
photon to leave the atom. An electron from a higher energy shell cascades down
to take its place, losing energy in the form of a photon. Since iron has a high fluo-
rescentyieldandrelativelyhighcosmicabundance, thisistheelementcommonly
seen.
(3) Absorption
Cold/neutral Photoelectric absorption: Cold or neutral photoelectric absorption
is the result of absorbing material in the line of sight. This material can be from
the interstellar medium (ISM) in our own galaxy, the ISM of the host galaxy (if
the object is extragalactic), and also absorbing material local to the X-ray source.
7
0.1 1 10 10010
?4
10
?3
0.01
keV (keV/cm
2  s keV)
Energy
Reflection Model
Figure 1.1: Shown is a model fit to an X-ray spectrum. Included in this model
is a thermal component at soft energies (blackbody emission) along with direct
power law emission and a reflection spectrum, all modulated by photoelectric
absorption. The reflection spectrum has a characteristic hump, the Compton
hump, and a cutoff energy at 200keV.
The primary effect it has on the X-ray spectrum is to cut off the emission at low
energies. This cutoff is determined by the amount of material, corresponding to
higher energies for higher absorbing columns. In Figure 1.1, the effect of absorp-
tionisshown onthepower lawand reflected emission. Inthisexample, a column
density of nH = 1021 atoms per cm2 is shown.
Ionized/warm absorption: Warm ionized gas, from optically-thin photo-ionized
8
gas in the line of sight, is seen in the spectra of some type 1 AGN. From obser-
vations with the ASCA X-ray satellite, Reynolds (1997) and George et al. (1998)
found that half of their samples of unabsorbed AGN (type 1) spectra statistically
preferred a warm ionized absorber model. The main signatures of a warm ab-
sorber are the O VII and O VIII K edges at 0.74 keV and 0.87 keV. However,
with higher quality grating data, more features (in particular, resonance absorp-
tion lines of ionized oxygen and other elements) are distinguished and photo-
ionization models can be used to determine the temperature and ionization pa-
rameters.
The origin of the warm absorber is unclear, but may be matter blown off the
accretion disk by radiation pressure. High resolution data, including UV spec-
tra from the Hubble Space Telescope and X-ray grating data from Chandra and
XMM-Newton have led to great advances in understanding the warm aborber
properties. These observations have revealed that the gas is from an outflow-
ing wind with column densities from 1021?23 cm?2 and many different ionization
states (Crenshaw et al. 2003).
1.3 The X-ray Observing Tools
Having described the various features of black hole X-ray spectra, it is now nec-
essary to discuss the tools used to obtain the spectra. The bulk of this disserta-
tion relies on observations from XMM-Newton and SWIFT. Therefore, these ob-
servatories are discussed in the greatest detail. Additionally, ASCA, Chandra, and
Suzaku are discussed.
9
Figure 1.2: Shown here is the XMM-Newton payload. This image was obtained
from the XMM-Newton User?s Guide and is credited to Dornier Satellitensysteme
GmbH.
1.3.1 XMM-Newton
A majority of the observations analyzed in this thesis, particularly the data in
Chapters 2?4, come from the European Space Agency?s (ESA?s) XMM-Newton, X-
ray Multi-Mirror Mission. The observatory was launched in December of 1999 as
a part of ESA?s Horizon 2000 Program. XMM-Newton is truly a multi-wavelength
observatory, consisting of 4 co-aligned telescopes: three high-throughput 7.5m
focal length X-ray telescopes and a 30-cm optical/UV telescope. A sketch of the
payload, obtained from the XMM-Newton User?s Manual1, is shown in Figure 1.2.
Threetypesofscienceinstrumentsareincluded: theEuropeanPhoton-Imaging
Camera (EPIC), the Reflection Grating Spectrometer (RGS), and the Optical Mon-
itor (OM). EPIC includes 3 CCD cameras for moderate resolution X-ray spec-
1http://heasarc.nasa.gov/docs/xmm/uhb/XMM_UHB.html
10
troscopy and X-ray photometry. These CCDs include one pn camera (shown in
violet towards the upper right in Figure 1.2) and two MOS cameras (shown as
the green and black ?horns?). Two RGS are included (shown in pink) for high
resolution X-ray spectroscopy. Finally, the OM provides optical/UV photome-
try (with filters centered at 5430 ?A, 4500 ?A, 3440 ?A, 2910 ?A, 2310 ?A, and 2120 ?A) and
grism spectroscopy. Since the X-ray observations are the main focus of this work,
particularly using EPIC, the focus now turns to the X-ray capabilities.
To begin, since X-rays are absorbed and not reflected in normal incidence, X-
ray telescopes use a ?grazing? incidence design. The critical angle, ?c, for total
external reflection is proportional to both X-ray wavelength and the square root
ofthedensityofatomsinthematerialused. Typically, goldisusedforthemirrors
(?c corresponds to 3.72?at 1keV for both), but iridium is used for the Chandra X-
ray Observatory mirrors. For XMM-Newton, the reflective surface is gold and
the mirrors are most efficient between energies of 0.1 ? 10keV with a 30? field of
view. One of the telescopes is arranged such that all of the light reflected from the
mirrors is focused to the pn CCD. For each of the remaining two telescopes, 44%
of the light is focused on the primary focus, a MOS CCD. Further, 40% passes
through an array of gratings before being focused on a CCD strip (RGS). The
remaining light is absorbed by the support structures. The biggest advantage of
this set-up is that it allows for simultaneous observations with all 5 of the X-ray
instruments (as well as the OM).
For the EPIC detectors, the cameras perform imaging in energies from 0.15
? 15keV with a FWHM of 6?? and a 30? field of view. The spectral resolution is
E/?E ? 20 ? 50. In Figure 1.3, the configurationof the MOS and pn CCD chips is
shown. The pn consists of 12 CCDs that are 13.6??4.4?. Each of the MOS cameras
consist of 7 10.9? ? 10.9? chips. Although, due to a micrometeoroid event, CCD6
11
Figure 1.3: The configuration of the MOS and PN CCD arrays are shown. The
figure is taken from the XMM-Newton User?s Manual.
on MOS1 is no longer operational as of March 9, 2005.
The RGS, unlike the EPIC detectors, are geared towards high resolution spec-
troscopy (E/?E ? 100 to 500). However, it is sensitive in a far narrower energy
band (0.33 ? 2.5keV). The field of view is also smaller, 5?, as determined by the
width of the 9 MOS CCDs used for the readout of the dispersed spectra. In Chap-
ter 3, two RGS spectra are examined. However, the bulk of the analysis relied on
EPIC spectra (? 100 spectra described in Chapters 2 ? 4).
12
1.3.2 The SWIFT Gamma-ray Observatory
The SWIFT Gamma-ray Burst Observatory was launched into low-Earth orbit on
November 20, 2004. Like XMM-Newton, SWIFT is a multi-wavelength observa-
tory. However, since the main science mission is to detect and follow-up the in-
credibly powerful gamma-ray bursts, SWIFT includes a very hard X-ray detector
with a large field of view. This allows SWIFT to scan much of the sky in search of
bright triggers, which can then be rapidly followed up by additional instruments
onboard as well as on the ground.
The very hard X-ray detector, the Burst Alert Telescope (BAT), is sensitive
from 14 ? 195keV. The field of view is 1.4steradians with ? 4? accuracy in source
positions, depending on the source signal-to-noise. The aperture used is a coded
aperture mask, consisting of ? 54000 lead tiles (5 ? 5 ? 1mm). Below this, the
detector consists of 256 modules of 128 elements (CdZnTe). BAT runs in two
operational modes, burst mode and survey mode. In burst mode, BAT reads
data in a photon counting mode. When not triggered on a burst, the instrument
collects data by count rate in 64 energy bins. Every five minutes, a new detector
plane map is made.
The technique for producing spectra with BAT (high quality imaging is not
possible) is very different from the lower energy X-ray telescopes. The lead tiles
in the coded aperture mask block light from bright very hard X-ray sources,
casting a ?shadow?. The patterns formed by the shadow (really multiple shad-
ows since multiple sources are typically in the field of view) are then separated
to determine the position of the source. Depending on the brightness of the
source, it may take many observations to construct a spectrum. The spectra
constructed for the BAT AGN survey, for example, were made using data from
13
Figure 1.4: Image from the SWIFT BAT website on HEASARC showing a cut-
away view of the BAT instrument.
22months of observations (Tueller et al. 2008). These analyses are done by scien-
tists at NASA Goddard Space Flight Center, particularly Drs. Craig Markwardt
and Jack Tueller.
In addition to BAT, SWIFT also consists of an X-ray telescope (XRT) and an
optical/ultra-violet telescope (UVOT). UVOT is essentially a copy of the XMM?s
OM. It allows for rapid optical/UV follow-ups of bursts as well as redshift deter-
minations with the optical grism. The XRT is a grazing-incidence telescope with
a focal length of 3.5m and a 23? field of view. XRT is sensitive in the 0.2 ? 10keV
energy band, providing X-ray imaging, spectroscopy, and timing. The CCD is an
EEV CCD-22, with a 600?600pixel area. Since the XRT position accuracy of 3?? is
very good, this allows for accurate positions for BAT-detected sources. This is es-
sential for obtaining proper follow-ups with additional instruments (for instance
ground based optical spectroscopy).
14
1.3.3 Additional X-ray Observatories
In addition to XMM-Newton and SWIFT, in Chapter 5 we describe X-ray spec-
tra from additional observatories. The Chandra X-ray Observatory is perhaps
the most famous, as the X-ray observatory of NASA?s four Great Observatories.
Chandra was launched July 23, 1999, only several months before XMM-Newton.
The Chandra X-ray Observatory consists of an X-ray telescope with 4 differ-
ent X-ray instruments. Unlike XMM-Newton, all of the science instruments can
not be used simultaneously. These instruments include the Advanced CCD and
Imaging Spectrometer (ACIS) and the High Resolution Camera (HRC), which are
used in the primary focal-plane, and both the High Energy Transmission Grating
(HETG) and Low Energy Transmission Grating (LETG), which are used in com-
bination with either of the focal plane instruments.
In many ways, the Chandra and XMM-Newton observatories are complemen-
tary. The mirror effective area is much higher for XMM-Newton, 4650cm2 at
1keV to Chandra?s 555cm2. This allows XMM-Newton to collect more photons
in a given amount of time, obtaining a higher quality spectrum. Meanwhile, the
full-width half maximum point spread function for Chandra?s mirrors is much
smaller, 0.2?? compared to XMM-Newton?s 6??. This gives Chandra superior spatial
resolution.
The Suzaku X-ray Observatory is another complementary observatory, not
onlyforXMM-NewtonbutparticularlyforSWIFT.ThisJapanesemission,launched
July 10, 2005, is sensitive from 0.2 ? 600keV. The observatory consists of two
instruments, the X-ray Imaging Spectrometer (XIS) and the hard X-ray detector
(HXD). Unfortunately the third instrument, the X-ray Spectrometer (XRS), an X-
ray micro-calorimeter which would have provided superior spectral resolution,
15
lost its cryogen soon after launch. Four co-aligned foil X-ray telescopes focus X-
rays onto each of the four XIS CCDs, each with a 19? field of view. A fifth is used
for the non-functioning XRS, which was the primary instrument. A key advan-
tage of this satellite is that it can obtain simultaneous X-ray (0.2?12keV) and very
hard X-ray (10?600keV) spectra. However, the field of view for the HXD is rather
large (0.56?), so it is important to select sources where source confusion will not
be an issue.
The final X-ray observatory from which we obtained X-ray spectra is another
Japanesesatellite,theAdvancedSatelliteforCosmologyandAstrophysics(ASCA).
ASCA was launched on February 20, 1993, and was operational into 1999. ASCA
consisted of four 3.5m X-ray telescopes with effective area of 1300cm2 at 1keV, a
24? field of view, and 2.9? spatial resolution. The detectors include 2 Gas Imaging
Spectrometers (GIS) and 2 sets of Solid-state Imaging Spectrometers (SIS). The
SIS are CCDs while the GIS are proportional counters. We used ASCA data in
this study because there are many papers detailing the ASCA X-ray spectra of
AGN in a uniform way, using simple models that we employ in Chapters 4 and
5. Additionally,averynicedatabase(Tartarus2)existswithprocessed spectra and
images available for many of the observed AGNs.
1.3.4 Analysis of the X-ray Spectra
In modern day X-ray astronomy (? 0.1?12keV), the main detectors used in con-
temporarymissionsareCCDs. Sincethephotonenergiesaresignificantbutcount
rates are small for cosmic X-ray sources (a count rate of 1 X-ray photon per sec-
ond is considered a strong source), much information can be garnered with the
CCD observations. In particular, the energy, arrival time, and position are mea-
2http://tartarus.gsfc.nasa.gov/
16
sured for each incoming photon. This allows for the generation of time-averaged
spectra, light curves, and images. However, the analyses of X-ray spectra are not
altogether straightforward since the observed spectrum is really the source spec-
trum convolved withthe instrumental response. Further, the toolsused to extract
and analyze the data differ by the satellite.
The general X-ray data analysis procedure begins with an examination of the
image. The source is identified and both a source and background region are
specified. Next,theobservedsourceandbackgroundspectraareextracted. These
spectra consist of the raw number of counts in detector channels (the observed
spectrum, C(I)). In order to convert the counts per detector channels into flux
per discrete energy, a response file is necessary. Typically, the instrumental re-
sponse consists of two files: the auxilary response file (arf, or A(E)) and the re-
distribution matrix file (rmf, or R(E,I)). The arf file contains efficiency versus
energy, which is a combination of the telescope effective area, filter transmission
efficiency, and detector quantum efficiency as a function of energy. Meanwhile,
the rmf file contains the detector response.
In general, the observed spectrum can be described as:
C(I) =
integraldisplay
S(E)R(E,I)A(E)dE. (1.2)
Since the response matrices are really the probability that a photon of a given
energy is registered in a given channel, it makes more sense to fold a model spec-
trum through the telescope response than trying to deconvolve the response and
obtainthe?true spectrum? (S(E)). Therefore, this is themethod employed for fit-
ting X-ray spectra. The standard analysis software, which folds the model spec-
trum through the response and fits it to the data, is called XSPEC (Arnaud 1996).
In XSPEC, the observed spectrum is computed as the background subtracted
spectrum, corrected for differences in the area and exposure time between
17
the source and background spectra. The fitting algorithm used is a modified
Levenberg-Marquardt algorithm (Bevington 1969). Use of the algorithminvolves
aniterative procedure. The user provides a set of initialguesses for the modelpa-
rameters (M(E,p)), which are changed by an indicated step size. Additionally,
lower and upper limits are also supplied. At each iteration, the parameters are
varied within the allowed range and a statistic is calculated to indicate the good-
ness of fit. Typically, the ?2 statistic is used, however, other statistics can also be
chosen (e.g. C-statistic). The definition of this statistic is:
?2 = summationdisplay(C(I)?M(I))2/?(I)2, (1.3)
where M(I) is the model folded through the instrumental response and ? is the
error in the given channel, often
radicalBig
C(I). The process is continued until a mini-
mum is found in ?2-space. The best-fit parameters are then returned, as well as
the fit statistic.
In order to determine goodness of fit, the reduced ?2 value is computed as
?2/?. Here, ? is the degrees of freedom or the number of free model parameters
subtracted from the number of data points. A good fit is indicated by ?2/? close
to 1. If the value is much higher, the model is not a good fit to the data. Alter-
natively, very low values indicate poor sampling or over-estimated errors on the
data points.
Once the best fit model is obtained, it is important to also find the associated
errors on each free parameter. In XSPEC, this is calculated with the error com-
mand. Here, the fitting procedure is again repeated in order to find the range of
values corresponding to a user supplied confidence interval. For instance, a user
may calculate the range of values corresponding to the 0.68, 0.90, or 0.99 confi-
dence level for the change in one parameter. This involves calculating the values
corresponding to ??2 of 1.00, 2.71, or 6.63 (Avni 1976), respectively, depending
18
on the number of model parameters. So, if the 0.90 confidence level is chosen, the
model parameters will be changed, both above and below the best-fit value, and
the ?2 statistic computed until the ??2 value is 2.71.
Often, it is also important to compare the goodness of fit between multiple
models for the same observed spectrum. If both models have the same degrees
of freedom, the comparison is trivial ? the lowest ?2 corresponds to the best sta-
tistical fit. If one model is an extension of the other, for example a power law and
blackbody model as an extension of a power law model, the F-test may be used.
Here, the F statistic for the addition of a co-efficient to a model (Bevington 1969)
is computed as:
F? = ?
2(n?1)??2(n)
?2(n)/(N ?n?1) =
??2
?2/?. (1.4)
In this case, the extended model (i.e. blackbody + power law model) has n + 1
parameters (n coefficients and 1 constant term), whose ?2 value corresponds to
?2(n). The original model (i.e. power law model) has a ?2 value of ?2(n ? 1).
Finally, there are N data points and a Gaussian or normal distribution (G(x) =
1
??2pi exp?
1
2?2(x??)
2) is assumed for the ?2 distributions of each model. A prob-
ability is then computed for the given F-statistic, indicating the probability that
the two distributions come from the same distribution. Potential problems as-
sociated with using the F-test are identified by Protassov et al. (2002). However,
if the threshold of significance is required to be sufficiently high (i.e. 99%), the
effects are minimal.
19
Figure1.5: Aschematicmodelof theAGNunifiedmodelisshown. Imagecredit:
NASA.
1.4 The Multiple Types of AGN
Classifications of AGN can be confusing, with different terminologies used based
on the wavelength through which they are viewed. In the optical, AGN are clas-
sified asbroad-line Seyfert 1s, blazars, or narrowline Seyfert 2s, amongothers. In
the X-ray, there are type 1/unabsorbed and type 2/absorbed sources. Further, in
the radio bands some sources are discovered to output huge amounts of energy
in jets, others show evidence of giant radio lobes. Connecting the properties in
various bands can be a daunting task, which in some instances brings challenges
to the standard unified AGN model.
20
The AGN unified model predicts that changes in the properties of AGN are
the result of differences in our viewing angle to the central source. An illustra-
tion is shown in Figure 1.5. In the model, the central supermassive black hole
is surrounded by an accretion disk. As previously mentioned, the source of the
X-ray emission is believed to be either synchrotron emission from a jet or inverse
Compton scattering of the accretion disk emission in a hot corona above the disk.
Beyondthedisk,thereisabroadlineregion(widthsof? 10000kms?1)composed
of clouds of gas or a wind from the accretion disk. Since the broad line region is
relatively close to the central black hole, these clouds/matter in the wind are/is
affected by Doppler motions causing the resultant emission lines from gas in the
region (particularly hydrogen and helium) to be broadened.
At parsec scale distances from the black hole, the AGN is thought to be sur-
rounded by an optically thick obscuring torus (Krolik & Begelman 1988). In the
X-ray band, it is the torus which is believed to contribute much of the X-ray ob-
scuration. The term covering fraction is typically used to distinguish the amount
of toroidal material that blocks the central source. As such, the covering fraction
measures the probability that our line of sight to the source is obscured. Thus,
a heavily obscured source most likely has a large covering fraction while an un-
obscured source typically has a small covering fraction. However, since this is
a probability, there are some sources that are obscured along our line of sight
which have small covering fractions and unobscured sources with large cover-
ing fractions. In the optical band, the presence or absence of the broad emission
lines, from the broad line region, is associated with a viewing angle that is not/is
blocked by the torus. Finally, at distances far from the central black hole, clouds
of gas comprise the narrow line region, responsible for the narrow emission lines
(widths ? 1000kms?1) seen in the optical band.
21
When sources are ?radio quiet?, with a low ratio of radio to X-ray flux and ei-
ther weak or no sign of radio jets or lobes, the designation Seyfert 1 applies to op-
tical spectra exhibiting both broad and narrow emission lines. A Seyfert 2 source
is one where the narrow lines alone are seen. Radio quiet sources comprise the
majority of all AGN, such that only 10% are ?radio loud?. For this 10%, sources
are designated broad line or narrow line radio galaxies. However, sources with
a featureless optical spectrum are called blazars and are thought to be sources
viewed directly along the jet.
The first firm observational support for this unified theory came with
the detection of polarized broad lines for a Seyfert 2 source, NGC 1068
(Antonucci & Miller 1985). The widths of these broad lines were typical for a
Seyfert 1 source. Further, the polarized light comprised 16% of the AGN flux,
supporting the idea that a fraction of the direct emission from a hidden Seyfert
1 nucleus was being reflected or scattered into our line of sight. Observations
of this source, thus, showed that in this case, at least, a change in viewing angle
would change our classification from Seyfert 2 to Seyfert 1.
However, other AGN observations have provided challenging to this model.
One such challenge involves a mismatch between X-ray and optical classifica-
tions for some AGN. A discussion of this issue is found in Matt (2002). In brief, a
class of broad-line QSOs with strong X-ray absorption has been identified. Sim-
ilarly, there are X-ray absorbed sources with optical broad lines. This mismatch
does not immediately fit with the unified model. Some theories of how this dis-
parity is possible include the idea that the X-ray absorption originates close to
the nucleus while the optical absorber is at further distances. Other ideas include
a different dust/gas ratio in the host ISM from our Milky Way or differences in
dust grain size (Maiolino et al. 2001).
22
Another challenge is posed from observations of variability in the X-ray ab-
sorbers. Variability on the scale of several 1021 cm?2 was discovered to be com-
mon in a sample of 23 X-ray type 1 and type 2 sources by Malizia et al. (1997).
Since changes in the absorbers take place on time scales of a month or less, this
suggests the absorber is close to the black hole. Later, Risaliti et al. (2002) found
that this variability is ubiquitous in Seyfert 2s on time scales of months to a year.
From their study, they find that the variability time scale is too low to be ex-
plained by changes in a torus located 1?3pc away. Instead, they find that the
X-ray absorbers must be clumpy and much closer to the black hole.
Probably the most challenging observations, however, involve the relative
lack of obscured AGN at high X-ray luminosities. From a sample of 41 hard X-
ray sources (2?10keV) detected by XMM-Newton, Piconcelli et al. (2002) found a
lower fraction of absorbed sources than predicted from cosmic X-ray background
(CXB) synthesis models. Further, both Steffen et al. (2003) and Ueda et al. (2003)
foundthatthefractionofobscuredAGNislowerathigh2?10keVbandluminosi-
ties. These results in particular suggest that the unified model must be modified
to include a luminosity dependence. In particular, the covering fraction of the
torus seems to be dependent on luminosity. This could be the result of radiation
pressure changing the shape of the torus, i.e. when the source is luminous the
covering fraction is small and the source is unabsorbed.
The main point to keep in mind is that recent studies on AGN uncover many
remaining questions. While we know that accreting super massive black holes
lie at the heart of the phenomena, the standard unified model is too simplistic.
Thus, with an unbiased local AGN sample from SWIFT?s Burst Alert Telescope,
we hope to answer some questions and perhaps pose new ones. In Chapters 4-6,
we will touch on these issues again in the context of our X-ray analyses.
23
Chapter 2
ULX Survey
2.1 Introduction
In thischapter, we present theresults of anX-ray analysisof UltraluminousX-ray
sources (ULXs). As mentioned in ? 1.1, ULXs are pointlike, non-nuclear sources
with observed X-ray luminosities greater than 1039 ergs?1 (Miller et al. 2004a).
The most interesting of these sources have bolometric luminosities above the Ed-
dington limit for a 20 M? black hole, or Lbol > 2.8?1039 ergs?1, which is believed
to be the black hole mass limit from normal stellar evolution. While the true na-
ture of ULXs is still unclear, they are likely a composite class of several different
types of objects.
One likely explanation for extremely bright X-ray sources is that they are low
luminosity AGNs or bright, distant blazars. However, while this may be the case
for some individual sources, there is evidence that not all ULXs are AGNs. While
some ULXs are located within a few parsecs of their host galaxy?s dynamical cen-
ter, they do not exhibit many of the characteristics of AGN. The most important
of these is the ratio of X-ray to optical flux. ULXs are fairly easy to recognize
in X-ray imaging data, but finding their optical counterparts is very difficult ? a
24
property that is typically not true for AGN (except for heavily obscured sources).
In particular, the ratio of X-ray to optical flux is a factor of 10 greater than that
of AGN (Anderson et al. 2003; Stocke et al. 1983), illustrating a major difference
between these objects. However, there is evidence to suggest that at least some of
the cataloged ULXs really are background QSOs at high redshift (Burbidge et al.
2003).
Assuming that the Eddington limit is obeyed by black hole accretion, the ex-
istence of luminous non-AGN sources presents a puzzle. Several models have
been proposed to account for the high luminosities of the ULXs. Among these
are relativistic and non-relativistic beaming from stellar-mass black hole systems
(King et al.2001;K?ording et al.2002;Reynolds et al.1997)andaccretionofmatter
into intermediate mass black holes (IMBHs).
In several ULX systems (e.g., NGC 1313 X-2, M81 X-9), detection of emission
nebulae surrounding the ULX supports the claim that the emission is isotropic
(Pakull & Mirioni 2003) and can therefore not be described through beaming.
Further, a number of ULX (e.g., NGC1313 X-1) X-ray spectra are best fit with
combined multi-component blackbody (MCD) and power law fits, similar to
Galactic black holes in their high-state. Recently, Miller et al. (2004a) found that
many ULX spectral fits required cool accretion disk temperatures of approxi-
mately100eV.Sincethetheoreticalrelationshipbetweenblackholemassanddisk
temperature(T? M?1/4)hasbeenobservedtoholdtrueforstellarmass(typically
around1keV)andsupermassive(around10-100eV)blackholes(Gierli?nski & Done
2004; Makishima et al. 2000; Porquet et al. 2004b), application of this scaling law
to these cool accretion disk ULXs would correspond to a population of high-state
IMBHs with masses of ? 16?104 M?.
Still, it is likely that not all ULXs obey the Eddington limit. In this case, their
25
extreme X-ray luminosities could be the result of an outburst, similar to those
observed for low and high mass X-ray binaries within our own Galaxy. For
instance, Jonker & Nelemans (2004) find evidence for approximately 5 Galactic
black hole X-ray binaries which exhibit luminosities in the ULX range during
outbursts. Since the typical time scale for outburst of Galactic X-ray transients is
afew daystorise fromquiescent levelwithadecline frompeak brightnessto qui-
escentvalueof30-40days(Chen et al.1997), thesesourceswouldappearastran-
sient ULXs. Another possible explanation is super-Eddington emission from ac-
cretion disks surrounding stellar mass black holes (Begelman 2002; Ebisawa et al.
2003). Sources of this typewould be expected to havesoft X-raycomponents well
modeled by hot accretion disks (? 1.3 ? 2.0keV), similar to superluminal X-ray
sources in the Galaxy (e.g. Belloni et al. (1997)).
Likely,ULXsincludeavarietyofdifferentobjectswithbothisotropicandnon-
isotropic emitters. However, if some ULXs do indeed represent a class of high-
stateIMBHs, similartothehigh/soft(thermaldominated)statestellarmassblack
holes in our own Galaxy, we also expect to see their low-state counterparts. This
is the main thrust of this chapter, searching for high and low state ULXs.
In Galactic black hole systems, the low-state is generally characterized by
lower luminosity, with L< 0.1LEdd (Done & Gierli?nski 2003), and a power law
photon spectrum, typically with index ? ? 1.7 (Remillard & McClintock 2006).
Indeed, the existence of some ULX sources (IC 342 X-1, NGC 5204 X-1) as possi-
ble low-hard (pure power law) state IMBHs, well-fit by simple absorbed power
laws, have been noted from Chandra observations by Roberts et al. (2004). In this
studyweseektotestadirectpredictionoftheIMBHhypothesis; namely,whether
there is a class of sources with properties consistent with what we expect of low-
state IMBHs. This requires two major assumptions: (1) that the emission from
26
ULXs is isotropic and (2) that IMBHs exhibit the same states (whose classifica-
tion was based on luminosity and spectral form) as stellar mass black holes. Our
goal is to find these ?low-state? sources, if they exist, classify the properties of
both high-state and low-state ULXs, and test whether these data are consistent or
inconsistent with the predictions of the IMBH hypothesis.
Therefore, in this chapter we present the results of a detailed analysis of ULXs
innearbygalaxiesobservedwiththeEuropeanSpaceAgency?sXMM-Newtonob-
servatory. Only XMM-Newton provides the count rates and bandpass necessary
todistinguishdifferentspectralmodelsformostULXs,accuratelydetermineboth
the temperature of the thermal component expected for high-state objects, and
determine whether this component is required in the spectral modeling of these
objects. Since the XMM-Newton X-ray spectra of ULXs are similar in quality to
spectra for Galactic X-ray binaries obtained in the 1980s, our spectral classifica-
tion in this chapter will remain purely schematic. Thus, our classifications as
low and high state objects are a first approximation, based on the quality of the
spectra available.
In ? 2.2, we detail the observations examined from the XMM-Newton archives
and explain the data analysisfor the individualpointsources. In?2.3, we discuss
the spectral fitting technique as well as simulations we conducted to determine
their validity. We discuss the implications of our results in ? 2.4. A summary of
the results of this chapter are provided in ? 2.5.
27
2.2 Observations and Data Reduction
2.2.1 Sample Selection
The data used in this investigation were drawn from the XMM-Newton public
data archive. Assuming that low-state ULXs exist in the luminosity range of
1038?39 ergs?1, we conducted simulations to determine the optimum criteria for
observations capable of resolving point sources of this luminosity. These simula-
tions provided a guide for choosing which of the vast number of archival XMM-
Newton datasets we should examine. The base luminosity of 1038?39 ergs?1 was
chosen on the assumption that an approximately 100M? black hole would radi-
ate at ? 10% of the LEdd in the low-state (Done & Gierli?nski 2003).
Within the luminosity range of interest (LX > 1038 ergs?1), there are a num-
ber of known objects that could be confused with ULX sources. One class of
sources is supernova remnants (SNRs). These sources are often easy to distin-
guish based on their characteristic spectrum: with poor signal to noise we expect
a steep power law and as the signal to noise increases, emission lines become
clearly visible. Super-Eddington accreting neutron stars have been observed to
have luminosities within this range for a short period of time. Neutron star X-ray
binaries often have spectra well fit by a hot multi-color disk blackbody model,
or with low signal to noise, by a bremsstrahlung model. Both models have sim-
ilar curvature and a 0.7-2.0keV blackbody model is indistinguishable from the
bremsstrahlung model. We chose to use the bremsstrahlung model because it is
the simpler model and gives an adequate qualitative description to the data. We
expect that for low temperature bremsstrahlung sources, the spectrum should be
easily distinguishable from a power law with ? ? 1.7 (as is expected for a low-
28
Figure 2.1: Distribution of galaxies by Hubble type among our archival XMM-
Newton sample of nearby (< 8Mpc) galaxies. Our sample consists of spirals and
irregulars.
state object). If, however, the neutron star spectrum has kT > 5keV, as observed
for some NS X-ray binaries, our simulations show that we can not distinguish
between the power law and bremsstrahlung models.
29
Ta
ble
2.1
:X
M
M
-N
ew
ton
Ga
lax
yO
bs
erv
ati
on
s
Ga
lax
y
Ty
pe
a
nH
b
di
sta
nc
ec
ref
ob
si
dd
du
rat
ion
(s)
co
mm
en
ts
NG
C2
47
SA
B(
s)d
1.5
4
3.0
9
???
01
10
99
03
01
14
53
6
-
NG
C2
53
SA
B(
s)c
;H
II
1.4
0
3.7
3
???
01
10
90
01
01
,0
15
20
20
10
1
30
71
1,
11
05
91
Sta
rb
ur
st
NG
C3
00
SA
(s)
d
3.1
1
2.5
6
???
01
12
80
01
01
43
96
7
-
NG
C6
25
SB
(s)
m?
sp
;H
II
2.1
5
2.6
2
???
00
85
10
01
01
26
28
8
-
NG
C1
31
3
SB
(s)
d;
HI
I
4.0
4.1
7
???
01
06
86
01
01
41
31
0
-
IC
03
42
SA
B(
rs)
cd
;H
II
30
.3
3.9
1
00
93
64
09
01
11
21
7
-
NG
C1
56
9
IB
m
21
.7
1.6
1
01
12
29
08
01
15
58
2
Sta
rb
ur
st
NG
C1
70
5
SA
0-
pe
c;
HI
I
3.9
5.1
2
01
48
65
01
01
58
92
6
Sta
rb
ur
st
MR
K
71
Sm
;H
II
3.9
3.4
3
01
41
15
02
01
45
91
9
ga
lax
yp
air
NG
C2
40
3
SA
B(
s)c
d;
HI
I
4.1
5
3.5
6
???
01
50
65
12
01
11
41
5
-
Ho
lm
be
rg
II
Im
3.4
2
2.7
0
???
01
12
52
07
01
,0
11
25
20
90
1
13
52
8,
68
60
-
Ho
lm
be
rg
I
IA
B(
s)m
3.4
9
3.6
4
00
26
34
01
01
26
28
0
-
M8
1
SA
(s)
ab
;LI
NE
R
4.1
2
3.6
4
01
11
80
01
01
12
79
13
Ho
lIX
als
oi
nfie
ld
of
vie
w
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
30
Ta
ble
2.1
?C
on
tin
ue
d
Ga
lax
y
Ty
pe
a
nH
b
di
sta
nc
ec
ref
ob
si
dd
du
rat
ion
(s)
co
mm
en
ts
M8
2
I0;
HI
I
4.1
4
3.9
5
01
12
29
02
01
29
38
7
Sta
rb
ur
st
Ho
lm
be
rg
IX
Im
4.0
3.6
4
01
12
52
10
01
10
35
0
M8
1a
lso
in
fie
ld
of
vie
w
Se
xta
ns
A
IB
m
3.8
5
1.4
6
00
26
34
02
01
21
61
8
-
IC
25
74
SA
B(
s)m
2.2
9
3.6
7
00
26
34
03
01
24
26
3
bu
rst
ing
sta
r-f
or
ma
tio
n
NG
C
42
14
IA
B(
s)m
;H
II
1.4
9
2.7
???
00
35
94
02
01
14
74
4
-
NG
C
42
58
SA
B(
s)b
c;L
IN
ER
1.2
7.2
???
00
59
14
09
01
,0
11
09
20
10
1
16
14
6,
21
89
5
-
NG
C4
39
5
SA
(s)
m;
LI
NE
R
1.3
3
4
???
01
12
52
19
01
15
84
2
-
NG
C4
44
9
IB
m;
HI
I
1.3
9
3.0
8
???
01
12
52
17
01
15
52
2
-
NG
C4
49
0
SB
(s)
d
1.7
8
7.8
1
01
12
28
02
01
17
75
4
int
era
cti
ng
wi
th
NG
C4
48
5
NG
C4
63
1
SB
(s)
d
1.2
8
7.5
1
01
10
90
02
01
53
85
0
-
NG
C4
73
6
(R
)SA
(r)
ab
;LI
NE
R
1.4
3
4.3
1
00
94
36
06
01
23
46
1
-
NG
C4
94
5
SB
(s)
cd
;S
y2
15
.9
3.1
???
01
12
31
03
01
23
06
2
-
NG
C
52
04
SA
(s)
m;
HI
I
1.4
2
4.8
1
01
42
77
01
01
,0
14
27
70
30
1
19
20
5,
16
38
7
-
M5
1
Sc
;S
y2
1.5
5
7.2
1
01
12
84
02
01
20
92
4
Ga
lax
yp
air
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
31
Ta
ble
2.1
?C
on
tin
ue
d
Ga
lax
y
Ty
pe
a
nH
b
di
sta
nc
ec
ref
ob
si
dd
du
rat
ion
(s)
co
mm
en
ts
M8
3
SA
B(
s)c
;H
II
3.9
4
6.2
???
01
10
91
02
01
30
62
7
Sta
rb
ur
st
NG
C5
25
3
Im
pe
c;H
II
3.7
7
3.2
1
00
35
94
03
01
47
21
6
Sta
rb
ur
st
M1
01
SA
B(
rs)
cd
1.1
7
7.4
8
01
04
26
01
01
43
01
9
-
NG
C5
40
8
IB
(s)
m;
HI
I
5.7
3
4.8
9
01
12
29
06
01
77
57
-
Ci
rci
nu
s
SA
(s)
b;
Sy
2
57
.8
4
10
01
11
24
01
01
11
04
96
-
a f
ro
m
th
eN
AS
A/
IPA
C
Ex
tra
ga
lac
tic
Da
tab
as
e(
NE
D)
b c
olu
mn
de
ns
ity
in
un
its
of
10
20
cm
?
2 ,
ob
tai
ne
df
ro
m
th
ew
eb
ve
rsi
on
of
th
en
H
FT
OO
L
c d
ist
an
ce
in
M
pc
(if
no
ref
ere
nc
ei
sg
ive
n,
ob
tai
ne
df
ro
m
th
ed
ist
an
ce
mo
du
lus
giv
en
in
LE
DA
)
d X
M
M
-N
ew
ton
ob
se
rv
ati
on
id
sf
or
th
ed
ata
ex
am
ine
di
nt
his
su
rv
ey
Ta
bl
eR
efe
ren
ce
s:
(1)
Tu
lly
(19
88
);(
2)
To
si
et
al.
(20
01
);(
3)
To
lst
oy
et
al.
(19
95
);(
4)
Fr
ee
dm
an
et
al.
(19
94
);(
5)
Sa
ka
i&
M
ad
or
e(
19
99
);(
6)
Sa
ka
ie
ta
l.(
19
96
);(
7)
Sh
ap
ley
et
al.
(20
01
);(
8)
Ke
lso
ne
ta
l.(
19
96
);(
9)
Ka
rac
he
nt
se
ve
ta
l.(
20
02
);(
10
)F
ree
ma
ne
ta
l.(
19
77
).
32
The most common sources we expected to find in this luminosity range were
the analogs to Galactic black hole X-ray binaries in their high-soft (thermal dom-
inated) state. These sources typically have spectra well fit by a blackbody with
temperature of ? 1.0keV combined with a power law with index ? ? 2.5. Our
simulations sought to determine the number of photons required to distinguish
between spectral fits corresponding to a power law model with ? ? 1.7 and a
combined blackbody and power law model. These models qualitatively corre-
spond to those of a low-state (pure power law spectrum) and high-state (thermal
dominated spectrum) X-ray binary. Since we do not know the proper normal-
ization between the blackbody and power law components for high-state objects
(it varies from source to source), we tested whether each of the components sep-
arately, e.g. blackbody or a steep power law, could be distinguished from the
simulated ?low-state? spectrum. We chose to simulate spectra in XSPEC using
the command fakeit none with generic response and ancillary response ma-
trices. Simulating a power law model with a ? = 1.7, we found that for 200, 400,
and 1000 counts, these models were distinguishable at > 99% confidence from a
blackbody source (with kT constrained to the range of 0.6 to 1.3keV, similar to
that of Galactic black holes). We found that for a lower number of counts the dis-
tribution in ? values increases to include a larger range of ? values (i.e. ? = 1.3 ?
2.0 compared to ? = 1.5 ? 1.7). Simulating a power law with ? = 2.5, we find the
same trend. We determined that at roughly 400 counts the distributions of ? from
a ? = 1.7 and ? = 2.5 power law become entirely separable at > 99% confidence.
Inorder todistinguishbetween thedifferent spectralfitsforobjects withLX ?
2 ? 1038 ergs?1, we select all galaxies that were observed for at least 10 ks (with
the exception of the bright ULX in NGC 5408, which had enough photons for
analysis despite the low exposure time) with XMM-Newton and that are closer
33
than 8Mpc. We estimated that these criteria would give us a minimum of 400
counts for objects with LX > 2 ? 1038 ergs?1. We emphasize that the criteria
quoted, based on the simulations, were used as a guide in choosing the sample of
galaxies examined in this study. These simulations are not used as the statistical
basis for our object-by-object analysis (discussed in ? 2.3).
Oursample ofgalaxiesis selective in thatit represents objects of interest in the
X-ray band. We include details on these host galaxies in Table 2.1. XMM-Newton
spectral information of individual X-ray sources had previously been published
for approximately 60% of the host galaxies. While we include references in the
alternate ID column and footnotes of Table B.1, we do not compare our results
with these previous studies on a source by source basis.
For 13 of the 32 galaxies examined, abstracts describing the proposals for
XMM-Newton observations were publicly available. Of these 13, only one ob-
servation cited the motive as a study of ULXs (NGC 1313). However, 7 of the
remaining 19 galaxies contained sources classified as IXOs, intermediate X-ray
objects, by Colbert & Ptak (2002). If the remaining galaxies were not studied due
to their ULX population, the effects of bias are small with roughly 25% of the
sources studied explicitly due to their connection with ULX sources. In terms
of host type, the general makeup of our host galaxy sample includes only spi-
rals and irregulars. We show the distribution of galaxy type in Figure 2.1. Here,
we find that a majority of our galaxies are irregulars (31%). Since the 2MASS
galaxy survey shows a much smaller number of irregulars which are clearly not
the dominating type (4% of the 100 largest galaxies) (Jarrett et al. 2003), this may
introduce a bias. However, it is unclear what the nature of the bias may be.
34
2.2.2 Data Reduction
Data reduction of our observations proceeded using the XMM-Newton Science
Analysis System (SAS) version 6.0.0. Since the processed pipeline products (PPS)
were created with earlier versions of SAS, the observation data files (ODF) were
used to produce calibrated photon event files for the EPIC-MOS and pn cameras
using the commands emchain and epchain. Following this, the events tables
werefilteredusingthestandardcriteriaoutlinedintheXMMABCGuide1. Forthe
MOS data (both MOS1 and MOS2 cameras), good events constitute those with a
pulseheightintherangeof0.2to12keVandeventpatternsthatarecharacterized
as 0-12 (single, double, triple, and quadruple pixel events). For the pn camera,
only patterns of 0-4 (single and double pixel events) are kept, with the energy
range for the pulse height set between 0.2 and 15 keV. Bad pixels and events too
close to the edges of the CCD chips were rejected using the stringent selection
expression ?FLAG == 0?.
Time filtering was applied as needed by editing the light curve produced in
xmmselect for the entire observation. Flare events (distinguished by their high
count rate) detected in all three cameras, were cut using the tabgtigen task
as outlined in the ABC Guide. Typical count rate parameters for filtering were
?RATE < 5? for MOS detectors and ?RATE < 20? for the pn detector. Such filtering
was only done as needed. Pre-filtered exposure times are listed in Table 2.1. The
number of counts from the filtered net exposure times for the individual sources
are listed in Table B.1. We note that the filtered data are not always sufficiently
clean such that a high signal-to-noise is maintained up to 10keV. Sources with a
high background flux level, relative to the source spectrum, show poorer signal-
1http://heasarc.gsfc.nasa.gov/docs/xmm/abc/
35
to-noise in the spectrum above 1keV.
Before extracting spectra of the brightest sources, contour maps of the X-ray
observation were overlaid on Digital Sky Survey (DSS) images. This ensured that
bright foreground stars and background AGN were easily distinguished, and
thereby not included in our sample. Also, we checked the XMM-Newton source
positions with NED and SIMBAD to determine if they coincide with any known
background galaxies or QSOs. A list of these bright fore-ground or background
sources is included in Table B.2.
2.3 Spectral Fitting
Spectraforthebrightpointsourceswereextractedbasedontheirapparentbright-
ness in the CCD images. Therefore, no explicit source detection algorithm was
necessary. To extract the spectra, we used the SAS task especget. With this
task we created spectra (for both the source and background), response matri-
ces, and ancillary response files for all three EPIC cameras, when possible. The
typical extraction radius was 20 arcseconds, but depending on both the size and
proximity of a source to another source, the extraction radius ranged from 9 - 87
arcseconds. Background spectra were extracted either in an annulus centered on
the source, or in a circle of appropriate size away from the source, depending on
the proximity of the candidate source to other X-ray sources. Annuli were used
for sources that were not located within a few arcseconds of another source, thus
annular background extraction radii were not used for sources with small extrac-
tion radii. For sources in crowded regions, we used circular extraction radii close
to the source. We extracted backgrounds close to the source in order to correct for
emission local to the ULX. Once the spectra were obtained, they were rebinned to
36
requireatleast20countsperbinsuchthatthedataobeyGaussianstatistics, using
thecommandgrpphainLHEASOFT.Thelistofsources, withpositionandcount
information, is included in Table B.1. We only included sources for our spectral
studies that had 400 or greater pn counts (or MOS for sources in NGC 253 and
M81 and NGC 4945 XMM3 and NGC 2403 XMM1, for which pn spectra were not
available).
The extracted spectra were fit with standard models in XSPEC v11.3.1. For
each source, we fit the pn and MOS spectra simultaneously in the 0.3-10 keV
range. We allowed a free normalization constant to account for the differences in
flux calibration between the three cameras (similar to Jenkins et al. (2004)). Each
source was first fit with an absorbed single component model. In all cases we
used the standard absorption model wabs, leaving the column density as a free
parameter. For those sources where the hydrogen column density was uncon-
strained, we fixed the value to the Galactic foreground value listed in Table 2.1.
2.3.1 Single-Component Sources
Results of the single-component fits are seen in Table 2.2. We include in this
table only the best-fit parameters for those sources best described by a single-
component model. The addition of a blackbody component to these single-
component fits changes the ?2 value by a negligible amount and therefore is not
statistically significant. More specifically, the addition of a blackbody model to
the power law fit corresponds to a ??2 < 2.3, which is the ? 68% confidence
level using the F-test for two degrees of freedom.
The flux values quoted represent the unabsorbed flux in the pn spectra, in
the 0.3-10 keV band. All errors quoted, here and subsequently, correspond to the
90% confidence level for one degree of freedom (??2 = 2.71). The luminosities
37
Table 2.2. XMM-Newton best fit: single component spectral fits
Powerlaw Bremsstrahlung
Source nHa ? ?2/dof nHa kT (keV) ?2/dof FXb LXc
NGC247 XMM2 1.4+1.8?1.1 2.29+1.02?0.57 47.7/54 < 0.65 2.55+6.90?1.61 48.8/54 0.33 0.38
NGC 253 XMM2 (obs 1) 1.6+0.4?0.3 2.51+0.18?0.17 69.1/74 0.5+0.2?0.3 2.12+0.52?0.37 74.7/74 0.52 0.87
NGC300 XMM4d 2.5 9.07 90.6/45 0.27 0.14 117.6/45 - -
NGC1313 XMM4 1.86+0.5?0.4 1.8+0.07?0.12 141.7/149 1.2+0.3?0.3 6.62+2.3?1.48 140.1/149 0.33 0.69
IC0342 XMM1 5.8+0.6?0.3 1.68+0.08?0.08 159.5/185 4.9+0.5?0.4 10.5+3.3?1.9 160/185 3.5 6.37
IC0342 XMM2 23.9+4.0?3.6 1.85+0.22?0.20 77.5/85 21+3.0?2.8 8.5+5.0?2.4 74.9/85 4.64 8.44
IC0342 XMM4 5.3+1.4?1.2 2.02+0.20?0.19 64/58 4.2+0.99?0.85 4.44+1.68?0.74 56.9/58 0.69 1.26
MRK71 XMM1 0.47+0.30?0.32 1.69+0.11?0.13 55.3/54 0.04+0.26?0.04 7.98+4.90?2.92 59/56 0.19 0.27
NGC2403 XMM4 1.7+0.8?0.7 1.89+0.30?0.25 62.3/71 1.1+0.5?0.3 4.59+4.1?1.5 62.3/71 0.31 0.48
HolmII XMM1 (obs 2) 1.5+0.2?0.2 3.09+0.15?0.12 266.7/252 0.31+0.12?0.15 1.13+0.11?0.11 309.4/252 3.5 3.1
Holm I XMM2 0.35e 2.13+0.16?0.15 39.2/45 0.35e 2.0+0.57?0.48 68.8/45 0.10 0.16
Holm I XMM3 0.35e 2.05+0.19?0.18 34.4/32 0.35e 2.03+0.85?0.56 42.1/32 0.12 0.19
IC2574 XMM1 1.3+0.40?0.30 1.97+0.07?0.10 120.9/103 0.69+0.23?0.25 4.1+0.89?0.67 107.5/103 0.35 0.47
IC2574 XMM2 0.4+0.4?0.3 2.2+0.21?0.09 45.7/51 0.229e 1.71+0.33?0.27 65.4/52 0.22 0.34
IC2574 XMM3 0.15+0.35?0.14 2.43+0.27?0.18 40.3/49 0.229e 0.97+0.18?0.14 76.5/49 0.22 0.34
NGC4214 XMM1 1.1+0.52?0.47 1.87+0.26?0.21 41.9/38 0.54+0.41?0.35 4.86+4.52?1.66 44.5/38 0.25 0.22
NGC4258 XMM2 (obs 2) 6.7+2.6?1.5 2.49+0.44?0.33 83.6/57 4.8+0.9?1.3 2.61+1.22?0.72 85.5/57 0.30 1.9
NGC4258 XMM3 1.4+0.69?0.64 2.32+0.34?0.24 38.9/37 0.49+0.44?0.38 2.48+1.09?0.74 41.3/37 0.20 1.2
??? 3.8 1.82 4/11 2.7 7.14 5/11 0.077 0.48
NGC4258 XMM4 0.68+0.24?0.42 1.97+0.22?0.19 41.1/48 0.06+0.31?0.05 4.07+1.6?1.2 45.2/48 0.39 2.4
??? 1.9+0.78?0.60 2.24+0.29?0.24 77.03/77 0.9+0.6?0.4 2.82+1.2?0.8 77.8/77 0.33 2.0
NGC4395 XMM2 0.33+0.6?0.3 2.75+0.45?0.33 38.6/36 0.133e 0.79+0.13?0.13 52/37 0.15 0.28
NGC4395 XMM4 0.3+0.6?0.3 2.08+0.39?0.30 16/25 0.133e 1.97+0.99?0.60 21.6/26 0.15 0.28
NGC4449 XMM2 1.5+0.3?0.3 2.81+0.16?0.14 103.5/112 0.25+0.2?0.2 1.65+0.22?0.21 112.1/112 0.29 0.33
NGC4490 XMM4 10.2+2.3?1.8 2.09+0.23?0.19 51.6/50 8.3+1.3?1.5 4.75+1.82?0.90 50.3/50 0.84 6.1
NGC4490 XMM5 3.9+0.94?0.81 2.31+0.22?0.20 60.1/65 2.5+0.54?0.59 3.08+0.89?0.62 61.6/65 0.41 2.98
NGC4631 XMM4d 7.8 9.50 261.5/74 2.9 0.17 207.8/74 - -
NGC4631 XMM5f 1.3 1.03 641.8/153 1.3 199 659/153 - -
NGC4945 XMM3 3.3+1.3?0.9 1.82+0.12?0.20 30.1/30 2.5+0.83?0.90 6.07+4.50?1.71 30.3/30 0.38 0.43
NGC5204 XMM2 0.89+0.49?0.53 1.98+0.25?0.20 42.37/42 0.23+0.3?0.22 4.05+1.51?0.95 42.2/42 0.15 0.41
??? 0.75+0.45?0.45 1.63+0.20?0.17 41.4/47 0.42+0.46?0.38 7.82+5.03?2.70 39.4/47 0.25 0.69
M51 XMM3 0.6+0.30?0.40 1.86+0.09?0.15 63.2/72 0.05+0.3?0.02 5.22+2.26?1.41 69.2/72 0.18 1.1
M51 XMM4 0.4+0.20?0.30 1.55+0.08?0.13 34.8/37 0.01+0.17?0.13 11.1+0.32?0.25 34.8/37 0.16 0.99
atotal column density in units of 1021 cm?2
bunabsorbed flux in the 0.3-10 keV band in units of 10?12 ergcm?2 s?1
cunabsorbed luminosity in the 0.3-10 keV band, using the distances quoted in Table 2.1, in units of 1039 ergs?1
dsee appendix; super-soft X-ray source best fit by single-component blackbody
eabsorbtion column density fixed to the galactic column density found in Table 2.1
fsource is best fit by a combined power law and vapec model; see appendix
38
1 100.5 2 5
10
?5
10
?4
10
?3
Photons cm
?2
 s?1
 keV
?1
Energy (keV)
pow
1 100.5 2 5
10
?5
10
?4
10
?3
Photons cm
?2
 s?1
 keV
?1
Energy (keV)
bremss
Figure 2.2: Shown are the simple single component models used: an absorbed
power law and an absorbed bremsstrahlung model. Notice the difference in the
curvature and cut-off of the bremsstrahlung model. Here, ? = 2.7 for the power
law model and kT = 1.50keV for the bremsstrahlung model.
were calculated from the unabsorbed flux using the distances quoted in Table 2.1.
Both flux and luminosity correspond to those of the best fit model (power law or
bremsstrahlung). It should be noted that since our selection criteria was based on
a count rate cutoff, due to the variety of spectral forms, the inferred luminosity
cutoff will not be uniform.
In Table 2.2 we denote the single component model we choose as the better fit
inbold. Thisnotationalsoindicatesthemodel(powerlaworbremsstrahlung, see
Figure 2.2)used to compute thequoted flux. For? 46%ofthe power lawsources,
the?2 difference(< 2)betweenthepowerlawandbremsstrahlungmodelsisonly
marginally different. Of these sources, the average kT value for the bremsstrah-
lung fit is 5.54keV. From our simulations we find that at high temperatures the
39
bremsstrahlung fit becomes indistinguishable from a simple power law. Thus,
given the high temperatures of the bremsstrahlung fits for these sources, they
are equally well described by the power law model. Typical kT bremsstrahlung
values for accreting neutron stars are from 3.0 to 7.0keV (Jones 1977).
?Low State? ULXs
From these single-component sources, we classify 16 sources as low-state ULX
sources. This classification is based on (1) the shape of the spectrum, well-fit
by an absorbed power law, (2) the luminosity of the sources (they needed to be
luminous enough to be included in our sample, LX > 1038 ergs?1), and (3) the
X-ray location of the object within the optical galaxy (based on Digital Sky Sur-
vey (DSS) images). The third criterion was important in limiting the effects of
contamination from fore-ground and background sources within our sample of
ULX sources. Thus, we overlaid X-ray contours from the XMM image on the
DSS images, determining the location of the X-ray source as within or outside of
the optical extent of the galaxy. We note that for two of the sources classified as
low-state ULXs (Holmberg I XMM2 and NGC 2403 XMM4), there is uncertainty
of whether the X-ray source is in fact within the optical galaxy due to the qual-
ity of the optical DSS images. The images used to determine the third criterion
are available online (http://www.astro.umd.edu/?lwinter/second3.html). We
further discuss these criteria in ? 2.4.
We state that the first criterion for classification as a low/hard state ULX is
a spectrum well-fit by a power law. Of the 30 sources in Table 2.2, three sources
were clearly notwell-fit by either the power law or bremsstrahlung model. Seven
of the remaining 27 sources were clearly not within the optical extent of the host
galaxy. Of the remaining sources excluded from classification as a low-hard
40
state object IC 2574 XMM1 was better fit with a bremsstrahlung model (with
??2 = 13.4). In an additional observation of the source NGC 4258 XMM2, a
simple power law model is not an adequate fit to the data (while the X-ray lu-
minosity of the source is below the formal ULX range of L ? 3 ? 1039 ergs?1 in
both observations). For the remaining sources, NGC 247 XMM2 and NGC 253
XMM2, there was sufficient doubt on the spectral form where the bremsstrah-
lung and power law model as well as the addition of a thermal component all
yielded adequate fits to the spectra. Therefore, we excluded these sources from a
classification, noting the ambiguity of the model fits for these sources.
Table 2.3: Bolometric Luminosities of ULX sources
Source Lcutoffpla Lbol b MEddc classd
NGC247 XMM1 13.4258 7.07734 54 HS ULX
NGC253 XMM1 9.31469 2.44574 19 HS ULX
NGC253 XMM2 4.3701 2.15292 17 HS ULX
NGC253 XMM6 5.05828 3.92514 30 HS ULX
NGC1313 XMM3 37.0364 27.9692 215 HS ULX
NGC1313 XMM4 1.50345 ??? 116 LS IMBH cand.
IC0342 XMM1 14.1215 ??? 1086 LS ULX
IC0342 XMM2 19.8129 ??? 1524 LS ULX
IC0342 XMM3 114.015 95.4068 734 HS ULX
MRK71 XMM1 0.2993 ??? 23 LS IMBH cand.
NGC2403 XMM1 4.1497 2.14873 17 HS ULX
NGC2403 XMM4 0.57068 ??? 44 LS IMBH cand.
Holmberg II XMM1 0.88906 ??? 68 LS IMBH cand.
Continued on Next Page...
41
Table 2.3 ? Continued
Source Lcutoffpla Lbol b MEddc classd
??? 16.8335 11.4543 88 HS ULX
Holmberg I XMM2 10.5158 ??? 809 LS ULX
M81 XMM1 15.7004 3.17932 24 HS ULX
Holmberg IX XMM1 31.0582 28.1445 216 HS ULX
NGC4214 XMM1 0.26699 ??? 21 LS IMBH cand.
NGC4258 XMM3 0.46503 ??? 36 LS IMBH cand.
NGC4395 XMM1 9.04609 2.94683 23 HS ULX
NGC4449 XMM2 2.48586 ??? 191 LS IMBH cand.
NGC4490 XMM1 16.8513 3.21972 25 HS ULX
NGC4490 XMM2 7.36612 4.51554 35 HS ULX
NGC4490 XMM3 240.653 176.04 1354 HS ULX
NGC4490 XMM4 1.66829 ??? 128 LS IMBH cand.
NGC4490 XMM5 12.7136 ??? 978 LS ULX
NGC4631 XMM1 10.6527 8.59661 66 HS ULX
NGC4736 XMM1 31.6561 27.3664 211 HS ULX
NGC4945 XMM3 0.44985 ??? 35 LS IMBH cand.
NGC5204 XMM1 22.4756 2.20492 17 HS ULX
NGC5204 XMM2 5.57769 ??? 429 LS ULX
M51 XMM2 4.25898 3.57502 28 HS ULX
M51 XMM3 2.10133 ??? 162 LS IMBH cand.
M51 XMM4 2.56064 ??? 197 LS IMBH cand.
M51 XMM6 46.7642 39.5189 304 HS ULX
M101 XMM1 8.04916 7.68224 59 HS ULX
Continued on Next Page...
42
Table 2.3 ? Continued
Source Lcutoffpla Lbol b MEddc classd
M101 XMM2 7.54268 4.96709 38 HS ULX
M101 XMM3 10.9792 1.03659 8 HS ULX
NGC5408 XMM1 20.9211 11.5369 89 HS ULX
Circinus XMM1 70.3579 56.1033 432 HS ULX
Circinus XMM2 208.746 0.69929 5 HS ULX
Circinus XMM3 771.157 0.212957 2 HS ULX
aestimation of the bolometric luminosity, determined with an exponential cut-off in the power
law at high energy (see text)
bbolometricluminosity estimate for high-state ULXs wherethe power law is cut attwice kTin (see
text); units for both luminosity measurements in 1039 erg s?1
cmass computed for objects radiating at 0.1?LEdd (low-state objects) or LEdd (high-state objects;
using Lbol), in units of M?
dclassification based on the criteria set forward in the text: high-state ULX (HS ULX), low-state
ULX (LS ULX), and low-state IMBH candidate (low-state object with 1038 ergs?1 < Lbol < 3 ?
1039 ergs?1)
For those sources we classify as ULXs, we include computed bolomet-
ric luminosities in Table 2.3. To compute the bolometric luminosities for
these ULX sources, we used the exponentially cutoff power law spectrum of
Magdziarz & Zdziarski (1995), model cutoffpl in XSPEC, with a cutoff en-
ergy of 10keV. From observational studies of Galactic X-ray binaries, it has
been observed that low-state objects have spectra that cut-off at high energies
43
(? 10 ? 200keV) (Zdziarski & Gierli?nski 2004). Thus we chose the exponential
model cutoffpl over a simple power law. This also minimizes the total lumi-
nosity for flat power law sources. We computed an unabsorbed bolometric flux
in the 0.1 ? 100keV range through use of the dummyresp command (which ex-
tends the model beyond the observation?s energy range). The luminosity was
then computed using the distances listed in Table 2.1. We quote these values as
Lcutoffpl (theluminosity obtainedfromextrapolatingthepowerlawportionofthe
spectrum as an exponentially cut-off power law) in Table 2.3. We note that these
values represent an upper limit on the bolometric luminosity for steep power
law (? > 2) objects, since we would expect the power law component to cutoff
at some low energy. However, for shallow spectrum (? < 2) sources Lcutoffpl is
a lower limit. This is because, schematically, a steep power law diverges at low
energies while a shallow power law diverges at high energies.
2.3.2 Two-Component Sources
For a number of sources, we found that an improvement in reduced ?2 was
achieved through fitting their spectra with an absorbed two-component black-
body and power law model. We chose a simple blackbody model over the multi-
component disk model, diskbb, for purely schematic reasons. Namely, obser-
vations of galactic X-ray binary systems were fit with blackbody models in the
1980s, when the signal-to-noise of these objects was comparable to that for our
XMM-Newton data for ULX sources. We also note that the diskbb model does
not give an entirely accurate physical description of the data as it neglects the ef-
fects of general relativity. As a schematic model, the blackbody model is simpler
than diskbb, with the same number of degrees of freedom. In addition, for low
temperatures both models yield virtually identical temperatures. For this study,
44
we chose the simpler model. We defer to Chapter 3 a discussion of the different
models for the thermal component.
Table 2.4: XMM-Newton best fit: blackbody and power law fits
Source nHa kT (keV) ? ?2/dof ??2b FXc LXd
NGC247 XMM1 4.1+1.9?1.5 0.12+0.03?0.02 4.18+1.79?2.52 86.5/93 25.7 6.2 7.1
NGC253 XMM1 2.7+0.4?0.4 0.80+0.12?0.09 1.74+0.17?0.14 225.9/230 36.7 2.7 4.5
??? 7.3+1.1?0.9 1.14+0.07?0.10 2.54+0.27?0.22 567/580 44.6 3.4 5.7
NGC253 XMM2 (obs 2) 2.0+0.3?0.2 0.71+0.10?0.10 2.14+0.05?0.08 460.3/498 47.1 1.6 2.7
NGC253 XMM3 3.1+4.8?0.5 0.75+0.13?0.10 2.47+2.99?0.41 68.5/82 23.4 0.60 1.0
??? 3.2+0.7?0.5 0.67+0.13?0.09 2.07+0.14?0.20 347.4/407 34.4 0.80 1.3
NGC253 XMM4 20+10.8?7.6 0.11+0.03?0.03 2.51+0.49?0.30 66.7/57 6.9 15 25
??? 4.5+1.2?1.9 0.09+0.02?0.01 2.33+0.27?0.22 309.3/291 12.1 1.4 2.2
NGC253 XMM5 1.5+7.2?1.5 0.96+0.24?0.32 2.43+3.06?1.36 26.5/23 5.3 0.26 0.43
??? 4.6+1.1?0.7 0.16+0.02?0.03 1.95+0.14?0.11 223.7/296 60.1 1.4 2.2
NGC253 XMM6 6.3+2.1?1.1 0.12+0.02?0.02 2.26+0.18?0.12 417.9/407 17.1 1.9 3.1
NGC253 XMM7 6.3+0.9?1.1 0.69+0.11?0.12 2.40+0.41?0.17 335.8/339 21.2 1.1 1.8
NGC300 XMM1 1.7+0.20?0.30 0.98+0.14?0.10 3.41+0.06?0.26 443.7/420 26.1 1.3 1.0
NGC300 XMM2 3.8+1.7?1.4 0.09+0.01?0.01 2.87+0.34?0.38 102.6/97 31.34 1.1 0.86
NGC300 XMM3 4.4+1.0?0.8 0.04+0.25?0.01 1.98+0.1?0.1 87.7/79 14.2 1.2 0.93
NGC300 XMM6 2.3+2.6?1.3 0.84+0.25?0.19 4.9+1.97?0.7 34.6/35 13 0.27 0.20
NGC1313 XMM1 3.0+1.2?0.9 0.13+0.03?0.02 1.75+0.14?0.11 194.1/201 35.4 0.64 1.3
NGC1313 XMM2 3.1+0.4?0.3 0.16+0.04?0.02 2.27+0.10?0.14 425.2/419 38.9 2.0 4.2
NGC1313 XMM3 6.2+0.8?0.6 0.11+0.01?0.01 2.76+0.10?0.11 441.7/424 336.6 10 22
IC0342 XMM3 9.7+1.8?2.1 0.09+0.02?0.01 2.69+0.16?0.23 129.5/107 56.3 31 56.4
NGC1705 XMM1 0.29+0.39?0.27 1.01+0.41?0.29 2.31+0.89?0.48 53/85 8.9 0.10 0.44
NGC1705 XMM3 < 1.44 1.07+0.20?0.15 2.23+0.70?0.56 69.8/65 11.1 0.15 0.48
NGC2403 XMM1 2.3+1.2?1.1 0.66+0.16?0.18 2.18+0.41?0.59 81.4/79 10.8 1.99 3.1
NGC2403 XMM2 1.8+0.8?0.6 0.62+0.16?0.11 1.95+0.26?0.42 163.1/151 16.4 1.0 1.6
NGC2403 XMM3 1.7+1.1?0.8 0.74+0.23?0.21 2.15+0.66?0.40 84.2/105 8.4 0.64 1.1
Holm II XMM1(obs 1) 1.6+0.1?0.2 0.14+0.02?0.01 2.35+0.05?0.11 997.5/976 136.7 12 10
M81 XMM1 3.3+0.17?0.08 0.90+0.03?0.03 2.52+0.03?0.04 1273.7/1243 533.1 4.5 7.0
??? 3.5+0.4?0.6 1.13+0.13?0.14 2.34+0.29?0.36 203.5/204 21.4 4.8 7.4
M81 XMM2 7.4+0.5?0.7 0.1+0.004?0.004 2.87+0.16?0.17 833.9/616 524.3 13 22
M81 XMM4 1.1+1.6?1.0 2.51+1.11?0.73 2.31+1.22?1.05 48.9/50 28.2 0.43 0.70
M81 XMM5 0.15+0.69?0.13 0.62+0.19?0.11 1.26+0.22?0.20 89/80 8.5 0.38 0.59
Continued on Next Page...
45
Table 2.4 ? Continued
Source nHa kT (keV) ? ?2/dof ??2b FXc LXd
Holm IX XMM1 2.1+0.2?0.2 0.17+0.02?0.02 1.72+0.04?0.03 866.6/878 134.3 10 16
NGC4258 XMM1 0.38+0.96?0.3 0.54+0.17?0.08 1.51+0.4?0.4 91.1/76 10.3 0.34 2.1
NGC4258 XMM2 (obs 1) 1.9+2.4?0.4 0.78+0.12?0.13 2.02+0.65?1.8 73.4/61 24.1 0.31 1.9
NGC4395 XMM1 2.0+0.08?0.07 0.14+0.02?0.02 3.44+0.54?0.56 168.2/154 26.9 1.4 2.7
NGC4449 XMM1 8.7+4.8?2.1 0.19+0.13?0.07 2.21+0.33?0.29 111.2/116 4.3 1.56 1.8
NGC4449 XMM3 3.5+1.3?0.9 0.15+0.03?0.03 2.52+0.36?0.39 119.9/87 34.1 1.1 1.3
NGC4490 XMM1 5.8+2.96?2.96 0.77+0.08?0.095 2.89+1.77?0.61 66.5/63 35 0.88 6.4
NGC4631 XMM1 3+0.9?0.5 0.12+0.03?0.02 2.12+0.03?0.02 371.3/345 12.1 0.96 6.5
NGC4631 XMM2 2.3+1.4?0.3 0.18+0.05?0.06 1.80+0.12?0.09 107.4/97 12.1 0.25 1.7
NGC4631 XMM3 1.1+1.1?0.8 1.01+0.12?0.1 2.45+1?0.62 127.1/96 18.9 0.15 1.0
NGC4945 XMM1 3.5+2.1?1.1 0.77+0.27?0.10 1.60+0.40?0.31 96.1/120 20 0.59 0.68
NGC4945 XMM2 3.2+1.1?0.7 1.15+0.28?0.33 1.80+0.20?0.30 105.8/113 8.7 0.66 0.76
NGC4945 XMM4 4.0+2.0?1.1 0.61+0.10?0.10 2.82+1.06?0.58 58.4/60 17.1 0.38 0.44
NGC5204 XMM1 0.66+0.35?0.08 0.16+0.02?0.03 1.92+0.12?0.06 543.0/559 49.1 1.98 5.6
??? 1.1+0.08?0.14 0.16+0.02?0.02 2.03+0.12?0.12 461.4/496 71.6 2.92 8.0
M51 XMM1 0.95+1.10?0.18 0.16+0.03?0.05 2.15+0.42?0.17 97/80 13.4 0.31 1.9
M51 XMM5 10.4+1.7?3.7 0.078+0.01?0.01 2.26+0.26?0.25 59.8/70 196.2 220 1900
M83 XMM4 1.77+3.9?1.77 0.54+0.18?0.09 1.61+0.96?0.31 84.8/89 6.6 0.2 0.92
M101 XMM1 0.22+0.12?0.15 0.21+0.03?0.04 1.42+0.14?0.05 249.9/231 53.1 0.45 2.9
M101 XMM2 1.6+0.46?0.21 0.76+0.14?0.10 1.88+0.25?0.11 251.6/261 37.2 0.7 4.6
NGC5408 XMM1 0.9+0.21?0.16 0.14+0.01?0.01 2.71+0.16?0.20 316.4/337 80.4 3.97 10.9
CIRCINUS XMM1 10.1+1.2?1.2 0.10+0.01?0.01 2.30+0.08?0.08 749.4/861 13.5 12 23
CIRCINUS XMM2 11.2+2.4?1.7 0.53+0.03?0.03 4.71+0.94?0.49 438.5/430 79.4 5.6 10.7
CIRCINUS XMM3 13.5+5.5?5.6 0.67+0.10?0.08 5.77+2.24?2.3 269.3/260 15.9 7.6 14.5
atotal column density in units of 1021 cm?2
bimprovement in ?2 over the single-component power law model
cunabsorbed flux in the 0.3?10 keV band in units of 10?12 erg cm?2 s?1
dunabsorbed luminosity in the 0.3?10 keV band, using the distances quoted in Table 2.1, in units
of 1039 erg s?1
46
In Table 2.4 we present the results for the sources which are fit significantly
better by the two-component model, these are sources where the improvement
in ?2 is greater than 8 (determined from our simulations in Appendix B). We
include in Table 2.4 the improvement in ?2 of the two-component fit over the
simple power law. We include the power law best fits to these sources in the
appendix for comparison with other analyses. In Figure 2.3, we show the PN
spectrum of the two-component model fit to Holmberg II XMM1. For this source,
the two component fit was clearly a better fit with ??2 = 136.7. In Figure 2.4, we
also supply the ratio of data to model points for the two-component and single-
component fit to this source. Here, it is clear that the disk model is a much better
fit, particularly at energies below 2keV.
In order to determine whether the blackbody component is statistically signif-
icant for all of the sources fit with a two-component model, we simulated spectra
based on accurate modeling of some of the brightest sources: NGC 247 XMM1,
NGC 5408 XMM1, and Holmberg II XMM1. These sources span the observed
range of the ratios of the blackbody to power law component and thus represent
those from our sample with a weak blackbody relative to the power law compo-
nent, intermediate case, and a strong blackbody, respectively. Our simulations
are described in full in Appendix B. We found that, using a ??2 > 8 criterion,
which corresponds to the 99% significance level as according to the F-test for the
addition of two extra parameters, we can readily detect the strong and interme-
diate thermal components in all spectra with more than 400 counts. The weak
thermal emission cannot be detected in 400 count spectra, but is readily detected
in 2000 count spectra. This gives us confidence that our results are statistically
meaningful.
47
1 100.5 2 510
?5
10
?4
10
?3
0.01
Photons cm
?2
 s?1
 keV
?1
Energy (keV)
Figure 2.3: Shown is the unfolded pn X-ray spectrum of a two component model
source, Holmberg II XMM1. The two-component model is shown in red. Both
the absorbed power law and absorbed blackbody contributions are shown in
green. This represents a source with an intermediate strength blackbody compo-
nent.
48
1 100.5 2 5
0.6
0.8
1
1.2
1.4
ratio
Energy (keV)
data/model
1 100.5 2 5
0.6
0.8
1
1.2
1.4
ratio
Energy (keV)
data/model
Figure 2.4: Shown is the ratio of the data to the model for the absorbed power
law and blackbody model (left) and the absorbed power law model (right) for
Holmberg II XMM1. Clearly, the two-component model is the better fit.
?High-State ULXs?
Of the sources in Table 2.4, we classified high-state (or thermal dominated) ULX
sources based on three criteria: (1) spectra characterized by an absorbed power
law and blackbody model, (2) luminosity, and (3) X-ray source within the optical
extent of the host galaxy. The luminosity criteria required that these sources have
unabsorbed luminosities LX ? 3 ? 1039 ergs?1 (we used LX = 2.7 ? 1039 ergs?1
as our hard cutoff). If the sources are radiating at the Eddington luminosity, this
cutoff luminosity corresponds to objects with masses greater than 20M?.
From Table 2.4, 27 observations are recorded with LX > 2.7?1039 ergs?1. The
addition of a thermal component to these sources is statistically significant over a
49
purepower lawmodel. Ofthe27observations, 3correspond to multipleobserva-
tionsofasinglesource. FromananalysisoftheDSSimages,all24ofthesesources
are within the optical extent of their host galaxies. However, M51 XMM5 appears
tobecoincidentwiththecenterofitshost,adwarfcompaniongalaxytoM51. The
location, coupled with the high luminosity (LX = 1.9 ? 1042 ergs?1) leads us to
classify this source as an AGN. We also excluded two sources (NGC 1313 XMM2
and M81 XMM2) from our sample of high/soft state ULXs due to their previous
identification as supernovae. Of the remaining 21 sources, NGC 253 XMM4 had
a luminosity of 2.5 ? 1040 ergs?1 in one observation and 2.2 ? 1039 ergs?1 in a
second. This significant change in luminosity, with one observation below our
luminosity cutoff and another a factor of ? 10 higher than the other, led us to
exclude this source as a high/soft state ULX. It is likely that this source is a stellar
mass X-ray binary within its host galaxy, where one of the observations captured
the source in an outburst.
In Table 2.5, we list sources that have ??2 values less than 8 for a single ob-
servation. Most of these sources have weak blackbody normalizations compared
to the power law normalization. We classify these sources as being well-fit by a
two-component model while acknowledging the uncertainty in the fit as deter-
mined by the simulations. The addition of the thermal component is not signif-
icant enough for these sources to be classified with certainty in either Table 2.2
or 2.4. The simple power law fits for these sources are included with those for
sources in Table 2.4 in Table B.3. We note that due to their high luminosity we
included six of these sources (NGC 4490 XMM2, NGC 4490 XMM3, NGC 4736
XMM1, M51 XMM2, M51 XMM6, and M101 XMM3) with uncertain fit parame-
ters as ULX high-state sources. Two of these sources (NGC 4490 XMM3 and M51
XMM6) had unabsorbed luminosities > 10 times the 3 ? 1039 ergs?1 cutoff used
50
Table 2.5. XMM-Newton two-component fits for sources with large uncertainty
Source nHa kT (keV) ? ?2/dof ??2b FXc LXd
NGC300 XMM5 0.41+0.60?0.30 1.06+0.37?0.20 2.78+0.61?0.65 46.6/53 7.6 0.17 0.13
NGC1705 XMM2 0.96+0.97?0.32 0.23+0.10?0.11 1.60+1.97?0.27 85.5/74 6.5 0.09 0.27
Holm I XMM1 0.4+0.5?0.3 1.97+0.66?0.89 2.46+0.44?0.40 97.4/93 5.4 0.6 0.93
M81 XMM3 3.7+2.4?2.1 0.11+0.05?0.02 1.69+0.27?0.33 77.1/78 4.25 0.53 0.82
Sextans A XMM1 0.4+0.7?0.1 1.05+2.3?0.07 2.6+0.8?0.2 269.1/271 2.3 0.60 0.14
NGC4214 XMM2 1.8+1.9?0.6 0.81+0.56?0.21 3.95+1.81?1.05 46.4/44 4.5 0.4 0.35
NGC4395 XMM3 0.5+0.9?0.3 1.10+0.67?0.18 2.66+1.05?0.77 52/56 3.9 0.29 0.56
NGC4490 XMM2 4.4+1.9?1.9 0.60+0.20?0.12 2.13+0.50?0.70 42.4/54 7.1 0.65 4.7
NGC4490 XMM3 13+9.6?2.5 0.09+0.02?0.02 3.21+0.52?0.17 72.1/78 4.6 12 87.4
NGC4736 XMM1 6.3+3.0?3.7 0.08+0.03?0.02 2.41+0.34?0.27 54.9/51 7.9 8.1 17.9
M51 XMM2 1.3+0.6?0.5 0.26+0.07?0.08 1.80+0.61?0.92 70.7/68 4.5 0.36 3.0
M51 XMM6 8.2+3.5?5.6 0.08+0.05?0.02 3.0+0.37?0.43 36.9/41 4.07 5.6 35
M51 XMM7 2.8+3.4?2.1 0.10+0.03?0.03 1.97+0.43?0.30 31.7/29 6.1 0.26 1.6
M83 XMM1 1.6+0.48?0.45 0.74+0.23?0.26 2.58+0.60?0.24 177.7/209 4.7 0.63 2.5
M101 XMM3 1.98+1.0?0.61 0.63+0.22?0.20 2.93+0.15?0.26 145.5/131 3.4 0.56 3.7
M101 XMM4 1.8+0.17?0.15 0.54+0.11?0.07 2.22+0.12?0.08 158.2/138 7.5 0.34 2.2
M101 XMM5 1.3+1.2?0.2 0.18+0.05?0.06 1.95+0.3?0.22 45.1/44 2.8 0.13 0.85
atotal column density in units of 1021 cm?2
bimprovement in ?2 over the single-component power law model
cunabsorbed flux in the 0.3-10 keV band in units of 10?12 erg cm?2 s?1
dunabsorbed luminosity in the 0.3-10 keV band, using the distances quoted in Ta-
ble 2.1, in units of 1039 erg s?1
for high-state ULX classification. The other four sources had luminosities above
the threshold, as well as weak blackbody components compared to the power
law (see the Appendix B for simulations). We used these points to justify includ-
ing these sources with the Table 2.4 sources in the following discussions with the
proviso that their spectral fits do not indicate absolutely the necessity of the addi-
tionalthermalcomponent. Forthisreason, wedenotethesesources withaspecial
symbol (a circle) in subsequent figures while including them as ?high-state? ULX
objects.
51
For our ULX sources modeled by a combined blackbody and exponentially
cutoff power law, we computed bolometric luminosities using two methods. The
first method is recorded as Lcutoffpl in Table 2.3. We computed the flux from 0.1
- 100keV using an unabsorbed blackbody and exponentially cutoff power law
modelusingtheXSPECcommanddummyresp. Forthesecondmethod,recorded
as Lbol in Table 2.3, we estimate a more accurate bolometric luminosity calculated
from the flux in the range of 2?kT?100keV where kT is the blackbody temper-
ature obtained from the model. In Galactic X-ray binary systems, the power law
component of the X-ray spectrum is believed to be from Comptonization in a
corona. The photons supplying this energy originate from the blackbody con-
tinuum emanating from the accretion disk. Thus, a natural cutoff for this power
law component occurs at the peak emission of the blackbody (which is approxi-
mately 3?kT). The estimated values (obtained from cutting off the combined un-
absorbed blackbody and cutoff power law model at the value 2?kT) differ with
regard to the full estimate (flux from the fully integrated blackbody added to the
separate flux from the cutoff power law from 3?kT to 100keV) depending on the
normalization factors used (for both the blackbody temperature and the spectral
index ?). Choosing three sources displaying a range of blackbody to power law
strength (Holmberg II XMM1, NGC 253 XMM1, and IC 0342 XMM3) we found
that the estimated values were within 88.3, 95.1, and 96.8% of the more complete
estimation. Given their close proximity (within approximately 90%) we quote
these estimated values as a good approxmiation to the bolometric luminosity.
We note that our bolometric luminosities for all of the classified ULX sources,
on average, are a factor of 1.08 greater than the X-ray luminosities in the 0.3 ?
10keV band for the objects best fit by a combined blackbody and power law.
Thus, to good approximation, the X-ray luminosity is the bolometric luminosity.
52
However, for the objects best fit by a simple power law (low-state ULX sources),
the average bolometric luminosity is roughly a factor of 7 greater than the X-
ray luminosity in our band. This average is dominated by the steep power law
objects, in particular Holmberg II XMM1 (? = 3.09), since the total luminosity
depends sensitively on the lowest energy used. Excluding this object, we get an
averagebolometricluminositythatis2.8timestheX-rayfluxandmoreindicative
of the general properties of these power law-fit objects.
2.3.3 Additional Sources
In addition, in this large sample of point sources, we came across a number of ob-
jects whose spectra were not well fit by the models we employed. These sources
haveluminositiesexceeding LX ? 1038 ergs?1, iftheyareassociated withthehost
galaxy, and are placed in Tables 2.2 and 2.4 as well as Appendix B. These sources
include two supersoft sources, one possible AGN, and three sources well fit with
additional absorption models (including a partial covering model and a model of
hot gas). We briefly describe these sources in the appendix.
2.4 Discussion
We have determined best-fit spectral parameters of the bright X-ray sources in 32
nearby galaxies. In choosing three ?standard? models for our study, we hoped to
accurately separate high and low state ULXs from other types of luminous X-ray
sources. We specifically chose to fit the data with the bremsstrahlung model in
order to identify neutron star X-ray binaries within our sample. The models we
used are purely schematic, and they do not physically explain the phenomena
occurring, but are standard and qualitatively simple models often used to fit the
53
spectra of Galactic X-ray binaries.
We cross-referenced the X-ray positions of our sources with both NED and
SIMBAD in order to identify known supernovae, galaxies, and stars. In addition,
we examined the DSS optical images to place the position of our sources within
their respective galaxies. Such analysis aimed to minimize contamination of our
sample of ULXs with bright background and foreground sources.
Further,weexaminedXMM-Newton?sOpticalMonitordatainthevisualbands
(U, B, V). The XMM-Newton PPS contain point source detection files for the OM
data. We overlaid these point source detections with X-ray contour maps in or-
der to determine the brightest possible optical count rates for the X-ray sources,
which were then converted into fluxes using the OM calibration documentation.
In Figure 2.5, we plot the distribution of the logarithm of the X-ray to optical
flux for the brightest possible optical counterpart inside the XMM-Newton error
circle. Only 13 of the 32 host galaxies had visible band OM data during the ob-
servations. Of these 13 galaxies, 40 of the X-ray sources were in the range of the
OM data and only 14 were coincident with an optical point source. Therefore, the
majority of our sources have X-ray/optical flux ratios that are larger than those
displayed. Figure 2.5 illustrates the lowest possible X-ray/optical flux ratios and
also, by the sparsity of sources included in the diagram, it illustrates the fact that
a majority of the sources have no obvious optical counterpart in the OM data,
with a limiting magnitude of V ? 20 (Kuntz et al. 2008), and thus have very large
X-ray/optical flux ratios. We estimate the point source detection limit of the OM
U filter as approximately 1.24?10?14 ergcm?2 s?1. For an unabsorbed X-ray flux
of 1.0?10?12 ergcm?2 s?1, typical of objects with LX ? 2?1038 ergs?1 located at
a distance of 8Mpc, this corresponds to log(fx/fopt) = 1.9. Therefore, the average
value for our sources should fall around 2 or greater. The average distribution
54
Figure 2.5: The distribution of fx/fopt for the brightest possible optical point
source within the XMM-Newton error circle. We define fx as the unabsorbed X-
ray flux in the 0.3 ? 10keV range and fopt as the optical flux obtained from the
U filter of XMM?s OM (as described in text). These ratios do not represent the
actual fx/fopt of the sources but are an estimate of the minimum possible value.
A majority of the sources had no optical point source within the X-ray contour
and thus have ratios of fx/fopt far higher than those indicated in the plot.
for QSOs and AGN centers around 0 and 0.8 for BL Lacs (Anderson et al. 2003).
Our objects have ratios of Lx/Lopt at least 10 times higher than those of AGN and
100 times greater than stars.
55
Recently, Guti?errez & L?opez-Corredoira(2005) identify six ULXs from the cat-
alog of Colbert & Ptak (2002) as QSOs. They hypothesize that a large number of
ULXs may in fact be quasars at higher redshift than their supposed host galaxy.
However, unlike the objects studied in Guti?errez & L?opez-Corredoira (2005), our
ULX sources are all spatially coincident with the optical host galaxy. In addi-
tion, a majority of our ULXs are not in the proximity of a noticeable optical point
source. The X-ray/optical flux ratios of our sources are much larger, on aver-
age, than might be expected for a QSO. It is also worth noting that while some
cataloged ULXs may be QSOs, optical identifications have been made associat-
ing other ULXs with a type B supergiant companion (Kuntz et al. 2005; Liu et al.
2004).
2.4.1 Classification Criteria
The spectral fits indicate that to high statistical probability (see Appendix B) we
can distinguish a class of low-state ULXs from the high-state objects. This is as-
suming, as indicated in the introduction, that ULXs are isotropic emitters with
luminosity and spectral form similar to Galactic stellar-mass X-ray binaries. In
? 2.3, we stated that our ULX classification depends upon three criteria: (1) spec-
tral form, (2) luminosity, and (3) location of the X-ray source within the optical
host galaxy (as determined from the DSS images). We have chosen simple, para-
metric ?non-physical? models for the spectra because the signal to noise of most
of the observations does not allow anything else to be constrained.
Of the sources in Table 2.2, 16 are classified as ?low-state? objects or low/hard
state ULXs, having unabsorbed luminosities > 1038 ergs?1 and spectra that are
wellfitby powerlawmodels. Throughoutthischapter, weuse thetermlow-state
ULXs to include ?low-state IMBH candidates? (sources with LX ? 3?1039 ergs?1
56
and spectra well-fit by a simple absorbed power law) and low-state sources with
luminosities that clearly classify them as ULX sources (LX ? 3 ? 1039 ergs?1).
These low-state ULX sources are listed in Table 2.3.
In the Spectral Fitting section, we noted that a power law and high temper-
ature bremsstrahlung model are indistinguishable. Therefore, it is important to
consider the luminosity of these sources in the claim that they are not neutron
star X-ray binaries accreting at the Eddington luminosity. Of the low-state ULX
sources, only two of the 16 sources have bolometric luminosities below the Ed-
dington luminosity of a 3M? object (? 4 ? 1038 ergs?1), corresponding to the
maximum mass of a neutron star. All of the sources have values exceeding the
Eddington limit for a 2M? neutron star.
Further, 26 sources have unabsorbed LX ? 3 ? 1039 ergs?1, corresponding to
L ? LEdd at M > 20 M? as expected for ?high-state? IMBHs, and spectra that are
well fit by combined blackbody and power law models. These are ?high-state?
objects. The spectral fits for these sources are listed in Tables 2.4 and 2.5. In a sta-
tistical sense, we find that the greater the number of counts in the observation the
greater our confidence in the thermal component contributing to a better fitting
model. We explain our confidence levels obtained from spectral simulations in
the appendix.
In addition to these high and low state ULXs, we find a large number of
sources best fit by a combined blackbody and power law model but below our
threshold of LX ? 3?1039 ergs?1 for a high-state ULX (listed in Table 2.4). Many
of these sources may be accreting stellar mass black holes with M< 20M?. Some
of these non-?ULX? sources were found away from the optical extent of the tar-
geted galaxy (from our analysis of the DSS images), and therefore may be back-
ground AGN.
57
2.4.2 Low-State ULXs
For Galactic black hole X-ray binaries, spectral indices of low-state (or power
law dominated) objects are typically lower than those of high-state objects, with
< ? >low? 1.7 and < ? >high? 2.5 (Remillard & McClintock 2006). In Figure 2.6,
we plot the distribution of the spectral index for both high-state and low-state
objects. The spectral index for the high-state objects is the value of ? from the
two-component fit. As in the Galactic sources, it is clear that the spectral indices
of the high-state objects are indeed larger. Of further interest, the distribution of
spectral index for low-state objects looks remarkably similar to the distribution
of spectral index for moderate luminosity quasars, many of which are thought to
be the analogs of low-state black holes (Porquet et al. 2004b). This supports the
classification of these objects as accreting black holes.
For the high-state objects, we find mean values of ? = 2.46, with a root
mean square (rms) deviation of ? = 0.12 and mean LX = 1.4 ? 1040 ergs?1 with
log(?) = 1.6. This calculation excludes the 3 objects with spectral indices greater
than 3.5. For the low-state objects, we find mean values of ? = 2.09, with a
rms deviation of ? = 0.10, and LX = 2.2 ? 1039 ergs?1, with log(?) = 2.1. This
value of ? = 2.1 is softer than the typical hard-state value of ? 1.7, but within
the 1.5 < ? < 2.1 range used to classify this state for Galactic X-ray binaries
(Remillard & McClintock 2006). Computing a Kolmogorov-Smirnov two-sample
test, separating the sources into the category of low-state or high-state, we find
a likelihood of approximately 0.03 that the spectral indices belong to the same
distribution.
The low/hard X-ray state of X-ray binaries is associated with a low accretion
ratefromthecompanionobjectwithL? 0.1LEdd (Done & Gierli?nski2003). There-
58
Figure 2.6: Distribution of the spectral indices (?) for low-state (top) and
high-state (bottom) objects. For Galactic low-state objects, typically ? ?
2.0, similar to our sample, while the high-state objects have a steeper ?
(Remillard & McClintock 2006).
59
Figure 2.7: Relationship of luminosity vs. spectral index for low-state (rectan-
gle) and high-state (triangle) objects. Sources represented by a circle are those
where the ??2 value between the two-component and power law fits was very
small. As expected from observations of Galactic stellar-mass black hole sys-
tems (Remillard & McClintock 2006), the classified low-state ULXs in our sam-
plehave, onaverage, lowerX-rayluminositiesthanthecorrespondinghigh-state
ULXs. We plot the mean values for both high-state and low-state objects with er-
rorbars indicating the root mean square deviation. The outlying objects with
spectral indices greater than 3.5 were not included in the mean or deviation cal-
culations.
fore, on average, we expect the luminosities of the low-state objects to be lower
than the high-state objects. Figure 2.7 displays the luminosity of the objects as
a function of the spectral index. On average, the highest luminosity low-state
objects have luminosities lower than those of the high-state objects.
The lower LX values of the low-state objects imply that they may indeed be
accreting at a lower rate than the high-state objects. This can further be seen in
60
the bolometric luminosities listed in Table 2.3. If these objects are accreting at a
rate similar to Galactic low/hard state black holes (0.1?LEdd) (Done & Gierli?nski
2003), we can estimate their masses as
M
M? =
Lbol
0.1?LEdd (2.1)
with LEdd as the Eddington luminosity for a 1M? object (1.3 ? 1038 ergs?1). Our
mass estimations, based upon our limits to the bolometric luminosities, yield
masses of 20 ? 1500M? (see Table 2.3), precisely what we might expect for a
population of IMBHs.
2.4.3 High-State ULXs
Ifthehigh-state(thermaldominated)ULXsrepresentaclassofintermediatemass
black hole systems, their X-ray spectra should be well described by a combined
blackbody and power law model. Scaling for the mass of the black hole, we
would expect a relationship of T?M?1/4 (for sources radiating near the Edding-
ton limit) between black hole mass and blackbody temperature (Makishima et al.
2000). This would indicate a thermal component of ? 100eV for masses of ?
103 M?. A few objects have been reported to display this property (Miller et al.
2004a; Roberts et al. 2005). In Figure 2.8, we graph the distribution of the thermal
component for our classified high-state objects.
We find that there are two peaks in the distribution among the thermal com-
ponent, one at approximately 100eV and another centered close to 1keV. This
could indicate two different classes among the high-state objects. It is possible
that those objects with blackbody components near 100eV are indeed high-state
intermediate mass black holes. We note that the soft excess in PG quasars has
also been modeled as a blackbody with kTsoft ? 100eV, but it has been sug-
gested that this could be the result of a process not directly related to black hole
61
Figure2.8: (top)Distributionoftheblackbodytemperatureforhigh-stateobjects.
(bottom) Relationship of blackbody temperature vs. luminosity (in the 0.3-10
keV band) for high-state objects. We see two peaks arise in the distribution,
one centered around kT? 0.1 and another at kT? 1. The peak with a low disk
temperature also corresponds to the highest luminosities, suggesting that these
may be high-state IMBHs. The sources with higher disk temperature also have
lower luminosities. The spectra of these sources were also well fit by an inverse
Comptonizationmodel(amodelsuccesfullyusedtofitsomeoftheGalacticblack
hole X-ray binaries in the very high state).
62
accretion (such as the presence of a warm absorber: Gierli?nski & Done (2004)).
Another possible explanation is that the soft component is the result of ionized
reflection from the disk (Ross & Fabian 2005). While the possibility exists that the
?thermal? component of these 100eV sources is not directly related to black hole
accretion or is related in a ?non-thermal? (i.e. ionized reflection) sense, as may be
the case with the soft excess in PG quasars, we assume that the soft component
for the objects we classify as high-state ULXs originates from a thermal disk. We
use this assumption to test the IMBH hypothesis, thus speculation on the nature
of the soft component is beyond the scope of this study.
The second peak, centered around 1 keV, has a temperature reminiscent of the
Galacticblackholesystems inourowngalaxy. Thesesystems maythusbestellar-
mass black holes accreting matter near the Eddington limit. If this were the case,
we would expect the luminosities of the sources exhibiting a higher blackbody
temperature to be lower than those with cooler blackbody components. In the
second graph of Figure 2.8, we plot the relationship between blackbody temper-
ature and LX in the 0.3 - 10 keV band. Once again, two groups are seen in the
distribution of high-state ULXs. The most luminous objects are those with low
blackbody temperatures. On average, the less luminous sources exhibit higher
blackbody temperatures. For the sources with LX > 1040 ergs?1, the mean black-
body temperature is 0.31keV while the sources below this luminosity threshold
have a mean blackbody temperature of 0.61keV.
The second, low-luminosity, group in the distribution of high-state ULXs is
clearly distinguishable in both plots of Figure 2.8. We found that, with the excep-
tions of NGC 253 XMM1, M81 XMM1, and NGC 5204 XMM1, the spectra of these
objects could also be well-described by an absorbed Comptonization (compST)
model (Sunyaev & Titarchuk 1980) used to fit galactic black holes in the ?very
63
high? state when they are radiating at the Eddington limit. This model simulates
Compton scattering of cool photons on the hot electrons of a completely ionized
plasma. We present the best-fit parameters for the Comptonization model in Ta-
ble B.4.
This ?very high? state has been observed (Miyamoto et al. 1991) in a few
Galactic black holes. Yet another rubric for the very high state emerged in
Kubota et al. (2001) and Kubota & Makishima (2004), where they identify this as
the ?anomalous? state ? a state whose spectrum can be well fit by a Comptonized
scattering model. Regardless of the name, our best-fit Comptonization sources
likely fit into this category. The luminosities of these sources suggest that they
are stellar mass black hole systems in this anomalous/very high state.
As with the low-state, we include mass estimates for our high-state objects in
Table 2.3. We assume that the high-state objects are radiating at LEdd resulting in
a minimummassif there isno beaming. Wefind masses of 1.6?38M?, consistent
with ?normal? stellar mass BHs, for the sources well fit by the Comptonization
model. The other high-state ULXs masses range from 17 ? 1350M? based on
Eddington rates, analogous to the low-state ULX masses computed.
It is important to note that the initial simulations (appendix) and discussions
in the Spectral Simulations section need to be considered in relation to the impact
they pose to our classification scheme and the results presented in these sections.
While it is indeed possible that some of the objects with a weak blackbody com-
ponent and a relatively small number of counts would be mis-categorized as a
pure power law spectrum, one can ask what such a possible situation would
do to the correlations that we have seen. These putative objects, by assumption
would have lower luminosities, however their temperatures are unknown and
it is entirely unclear if they would destroy the kT?LX correlation. As we have
64
shown in our simulations it is unlikely that the fitted power law index would
change and thus the presence of a low state as indicated by the spectral index
would not change. This would create a new type of object, one with a flat power
law and a blackbody component, which is not seen in the Milky Way, nor among
the high signal to noise objects.
2.4.4 Temperature Gap
In addition to the existence of ULXs with low blackbody temperatures, the tem-
perature distribution of the high-state (thermal dominated) ULXs (Figure 2.8, left
panel) displays a ?gap? which is of particular interest ? there is a complete ab-
sence of objects with temperatures in the range 0.26keV to 0.50keV. It is tempt-
ing to take this as evidence for a gap in the mass distribution of these accreting
black holes. Since, for a given luminosity, we expect the temperature to vary as
T ? L1/4M?1/2, this factor of two gap in the temperature distribution translates
into a factor of four gap in the black hole mass distribution.
If this result is borne out by further study, it provides an important clue to
the origin and evolution of intermediate mass black holes. One popular idea is
that intermediate mass black holes formed from the collapse of massive Popula-
tion III stars (Madau & Rees 2001). Models suggest that Pop III stars with zero
age main sequence (ZAMS) masses in the range 25?140M? and above 260M?
collapse to produce black holes (Heger & Woosley 2002) whereas in the range of
ZAMS masses 140?260M?, pair-instability supernovae lead to the complete dis-
ruption of the stars (i.e., no remnant black hole remains). Hence, this model for
IMBH formation predicts a gap in the IMBH initial mass function in the range of
approximately 60?200M? (although this is uncertain on the low end due to the
effect of the pulsational pair-instability on the pre-collapse core). One possibility
65
is that the gap in our observed temperature distribution (and hence the inferred
gap in the mass function) is due to this effect of the pair instability supernovae in
Pop III stars. This would require that the current IMBH mass function is approxi-
matelythesameastheinitialIMBHmassfunction. Inotherwords,itrequiresthat
most IMBHs (especially those just below the gap) have not grown significantly
due to accretion since their formation and, hence, that the ULX phase represents
a short fraction of the life-time of an IMBH (f << tsal/tH, where tsal ? 45?0.1 Myr
is the e-folding timescale for Eddington limited black hole growth with radiative
efficiency ? = 0.1?0.1).
An alternative interpretation of the inferred mass gap is to suppose that two
fundamentally different modes of formation lead to a strong bi-modality in the
final black hole mass function. Black hole masses below the gap can be readily
understood through normal stellar processes. A separate and distinct population
of significantly more massive black holes may result from dynamicalprocesses in
the core of dense globular clusters (G?ultekin et al. 2004; Miller & Hamilton 2002).
2.4.5 Comparison with Galactic HMXBs
Supposing that the Galaxy?s bright X-ray population is representative of low-
redshift galaxies, we expected to find a number of sources similar to Galactic X-
ray binaries in our sample. In our sample, we find approximately 24 sources with
luminosities below our high-state ULX cutoff (?3?1039 ergs?1) , X-ray positions
within the optical extent of their host galaxy, and no obvious optical counterpart.
The unabsorbed luminosities for these sources range from 0.4?2.5?1039 ergs?1
(0.3 ? 10keV band). Two of these sources were transients in the XMM-Newton
data. Ofthe four hostgalaxieswithmultiple observationsexamined, two of these
galaxiescontainedsolelyULXsourcesinourluminosityregime(HolmbergIIand
66
NGC 5204). Each of the remaining two (NGC 253 and NGC 4258) had a transient
source best fit by a combined blackbody and a power law.
This suggests an interesting diagnostic in terms of distinguishing our ULX
sources from a normal HMXB population. In our own galaxy, most HMXBs
vary on timescales of days or less and most of the black holes in the Milky
Way are transients, though some HMXBs are indeed persistent. The figures in
Kalogera et al. (2004), determined through detailed mass-transfer calculations,
indicate that transient behavior should not be expected from a population of
IMBHs. Thus, on average, our ULX sources should remain X-ray bright in mul-
tiple observations. Through a literature search, we found that 37/42 of our ULX
sources are well detected in ROSAT observations and thus are luminous for
greater than 10 years and therefore are not transients. Examination of the long
term light curves show that most of these sources vary by less than a factor of 3
over the timescale from ROSAT to XMM. The sources that have been above the
Eddington limit in the Milky Way and the Magellenic clouds do so transiently,
for a small fraction of the time. As best as we can tell, from the light curves from
Einstein, ROSAT, ASCA, Chandra and XMM the ULXs are, rarely, transients, and
are almost always ?on?, unlike Galactic ?ultra-luminous? objects.
As a possible further diagnostic, we constructed a color-color diagram for our
ULX sources. We adopted the colors of Done & Gierli?nski (2003) in order to com-
pare our sample with their sample of Galactic X-ray sources. Thus, our colors
were constructed from unabsorbed model fluxes in four energy bands: 3-4, 4-
6.4, 6.4-9.7, and 9.7-16 keV. The XSPEC command dummyresp was used to cal-
culate a flux based on the model for the 10-16 keV range. We plot colors for
a pure unabsorbed power law (from ? = 1.5 ? 3.0) and an unabsorbed MCD
model (diskbb in XSPEC with kTin = 5.0 ? 0.2eV) for comparison. Comparing
67
Figure 2.9: Color-Color Diagram plotting soft vs. hard colors, as outlined
in Done & Gierli?nski (2003), for low-state (rectangle) and high-state (triangle)
ULXs. The filled circles represent the sources with low ??2 values between
the two-component and power law models. A large number of our sources
lie in the same range of this graph as the black hole sources examined by
Done & Gierli?nski (2003) (nearthe power lawdistribution, indicated bythe solid
line). The dashed line represents the color-color plot for a multi-colored disk
model with different disk temperatures. The sources approaching this line were
those well-fit by the Comptonization model. Done & Gierli?nski (2003) had no
black hole sources in this region, but atolls and Z-sources, which were also well-
fit by Comptonization models.
68
our Figure 2.9 with Figure 8 of Done & Gierli?nski (2003), we find that our ULX
sources largely lie along the same regions as their black hole sources. A few ULX
sources, however, lie in the region occupied by atoll and Z-sources in the plot of
Done & Gierli?nski (2003). These sources were those best fit by a Comptonization
model.
2.4.6 Galaxy Sample
In this section, we examine the environment in which ULX sources reside. We
investigate the claim that the ULX population is proportional to the host galaxy?s
star formation rate (SFR) (Grimm et al. 2003; Ranalli et al. 2003). Towards this
end, we use the far-infrared luminosity of the host galaxy as an indicator of the
SFR. In order to compare the ULX population of a galaxy with the SFR we fol-
lowed a similar approach to Swartz et al. (2004). We calculate the FIR flux from
observations taken by the Infrared Astronomical Satellite. As in Swartz et al. (2004),
the flux between 42.4 and 122.5 ?m is approximated as:
FFIR = 1.26?10?11(2.58S60 + S100), (2.2)
in units of ergcm?2 s?1. The values of the flux at 60?m (S60) and 100?m (S100)
were obtained from either Ho et al. (1997) or NED. Luminosities were calculated
using the distances quoted in Table 2.1. We list these values in addition to the
number of ULXs observed in individual galaxies in Table B.5. The number of
ULXs includes both the objects we classify as high and low state ULX as well as
those sources resolved by Chandra (see references in the Table B.1).
In Figure 2.10, we show two plots relating the number of ULXs to LFIR. It has
been suggested by Grimm et al. (2003) that the luminosity function in the X-ray
regime from HMXBs is related to SFR. In our first plot, we find that the galaxies
69
Figure 2.10: (top) Relationship of the far-infrared luminosity, as an indicator of
star formation rate, vs. the number of ULXs for each galaxy. If ULXs are asso-
ciated with star formation, we naively expect that the higher the FIR luminos-
ity the more ULXs the galaxy will host. (bottom) The distribution of the aver-
age number of ULXs / LFIR bin for spirals follows this expectation (marked by
solid lines). The distribution of irregulars (green histogram) is not so easily in-
terpreted. The numbers at the top indicate the number of spirals/irregulars in
each of the luminosity bins. More irregulars would need to be included in this
survey for meaningful statistics on this group.
70
with the highest LFIR seem to have fewer ULXs than may be expected from the
luminosity functions of Grimm et al. (2003), who present a relationship showing
a scaling of the number of HMXBs with luminosities over a set threshold with the
host galaxy?s SFR (see equation 7; Grimm et al. (2003)). Using this relationship,
we would expect that a galaxy with a SFR approximately equal to that of M51 (?
4M? yr?1 from their table 1) to have? 4.47 objects with luminosities greater than
1039 ergs?1. We find 5 objects with this luminosity in M51, consistent with their
result. However, for NGC 4945, a galaxy with approximately the same LFIR and
therefore SFR, we find only one source with a luminosity in this range. However,
we note that NGC 4945 is a Seyfert, implying that the LFIR may primarily be
caused by the AGN and not a direct indication of SFR. In addition to high LFIR
sources with few ULX we find a number of sources with very small SFR but
which contain a ULX. For sources with SFR < 0.2M? yr?1, which corresponds
roughly to sources with LFIR less than that of NGC 4736, we would expect < 0.22
sources with luminosities above 1039 ergs?1. However, there are a number of
bright ULXs in galaxies with very low SFRs (for instance Holmberg II, Holmberg
IX, NGC 5204, and NGC 5408). Thus, in a direct comparison, our results do not
agree with the predictions of Grimm et al. (2003).
Thesecond plotdisplaystheaveragenumber ofULXs/galaxy,binned accord-
ing to luminosity. This plot is extremely similar to Fig. 15 of Swartz et al. (2004)
for spiral galaxies. Thus, once again, it seems that the connection between SFR
and the ULX population in spirals is supported. For irregular galaxies, however,
there seems to be more of a spread in the distribution. This could be the result of
poor sampling ? most of the bins contain only one galaxy. Another possibility
is that there is no direct correlation in irregular galaxies or that the overall star
formation in these galaxies is less ordered or clumpier. If the latter is the case,
71
the overall SFR of the galaxy is only an average over a wide range of values.
We shall address this issue again in the next chapter, where we discuss the local
environments of the ULXs in our sample.
In Figure 2.11 we plot the distribution of column densities among the ULXs.
We subtracted the Galactic column density towards the galaxy (obtained from
the nH FTOOL and listed in Table 2.1) from the values obtained through spectral
fits. We note that, on average, the ULXs have large column densities. The typical
Galactic column density along a line of sight is ? 4?1020 cm?2. If the ULX is lo-
cated on the opposite side of its host galaxy, we might expect maximum column
densities of ? 1.2 ? 1021 cm?2. However, most of our sources have column den-
sities well above this value. This is in agreement with the analysis of 5 ULXs by
Roberts et al. (2004) and may imply, as they suggest, that the local environment
of the ULXs contains an extra source of absorption. We are investigating this fur-
ther, comparing the X-ray absorption column densities with HI data (see Chapter
3).
In order to better understand the relationship between SFR and the ULX pop-
ulation, it is necessary to extend ULX studies to other wavelengths. In particular,
it becomes important to analyze UV and IR images close to the ULX.
2.5 Summary
In this chapter, we presented the results of an XMM-Newton archival survey of
the ULX population in nearby galaxies. This study required an important as-
sumption ? that ULXs are isotropic emitters. For the sources we selected as ULX
sources (which excluded transientsources andsupernovae), this assumption was
supported by the finding that 37/42 of our ULXs were found to be ?on? in ROSAT
72
Figure 2.11: Distribution of the hydrogen column densities of ULX sources. The
nH values were obtained through spectral fits using the wabs model in XSPEC.
Galactic column densities towards the host galaxy were subtracted from the
spectral fit values. A majority of our ULX sources have high column densities
(> 1021 cm?2), suggesting that some of this absorption originates with the local
ULX environment. Bins to the left of the dashed line represent sources with col-
umn densities very close to the Galactic value and thus a simple subtraction is
not statistically representative of the true value.
73
observations. This implies that these sources exhibited high luminosities for time
scales of at least 10years, a property that is not seen in Galactic Eddington-limit
exceeding sources (such as black hole X-ray binaries undergoing an outburst).
We also assumed that if some ULX sources represent a class of IMBH X-ray bi-
naries, they would exhibit spectral states analogous to Galactic stellar mass black
hole X-ray binaries. This is the hypothesis we set out to test, classifying a source
as a ULX based on (1) spectral form, (2) luminosity, and (3) coincidence of the X-
ray source within the optical host galaxy. Due to the quality of spectra available
for these distant X-ray sources, our classification of spectral form is really a first
approximation describing the basic curvature of the spectrum.
Through this study, we have found that there exists a population of objects
whoseX-rayspectralpropertiescloselymatchthelow/hardstatespectraofGalac-
tic black holes, but whose luminosities lie in the range of Lbol ? 2 ? 1038 ? 1 ?
1040 ergs?1. In the Milky Way, black holes with these spectral properties radiate
at only ? 0.05 of the Eddington limit. If this is also true for this population, it
indirectly implies that these objects have a mass greater than ? 30M? ranging
up to 1500M? and thus should be IMBHs. The existence of such objects was
?predicted? on the basis that the ULXs previously studied shared the X-ray spec-
tral characteristics of high-state Galactic black holes; namely, an X-ray spectrum
best fit by a combined blackbody and a power law (Miller et al. 2004a), but with
much higher luminosities. If these objects are high-state IMBHs, the correspond-
ing low-state objects should also exist.
Our survey has also uncovered a large population of objects whose X-ray
spectra are well modeled by the canonical description of Galactic black holes in
the high-state (thermal dominated), a black hole with a steep power law, but
whose bolometric luminosities exceed 2?1039 ergs?1, ranging up to 1041.5 ergs?1
74
and whose blackbody temperatures are less than 0.3keV. If these objects are ra-
diating at ? 1/2 the Eddington limit like their Milky Way counterparts their im-
plied masses are from 30 ? 3000M?, a range very similar to that implied by the
low-state objects. Using the M?1/4 scaling of mass to temperature, the observed
spectral temperatures give masses of 500 ? 104 M? a considerably larger value.
In general agreement with the expectations of the IMBH hypothesis, the objects
with high-state spectra are more luminous than those with low-state spectra. We
note that these results have required the high signal to noise of XMM in order
to discern the spectrum of these objects. Many of these objects have also been
observed by Chandra and their spectra have been well-fitted by simple power
laws.
In addition to classification of the sources, we investigated some of the prop-
erties of the ULX sources. We found a gap in the temperature distribution of
high/soft state ULXs. This gap may indicate a gap in mass distribution, which
may provide clues to the nature of ULXs. We also found that our ULXs are per-
sistent sources (not transients) which occupy regions on the color-color diagram
of Done & Gierli?nski (2003) also occupied by Galactic black hole sources. Lastly,
the existence of a substantial population of ULXs in nearby dwarf and other low
star formation rate galaxies argues that (in agreement with Ptak & Colbert (2004);
Swartz et al. (2004)) there is more than one source term for the origin of ULXs,
with at least some of them not being associated with recent star formation, at
least statistically.
We conclude, from an X-ray spectral and luminosity point of view, that our
data are consistent with many of these objects having the properties expected of
an IMBH population. However, we also find two other populations of objects,
those whose blackbody temperature and luminosity correspond to that of stellar
75
mass black holes with kT? 1keV and logLX less than 2?1039 ergs?1 and a small
population of objects whose X-ray spectra and luminosities are consistent with
that of stellar mass black holes in the very high state, as expected from studies of
the Milky Way and M31. Thus, ULX selected purely on the basis of 0.3 ? 10keV
X-ray luminosities are a composite class with ? 1/4 being ?normal? stellar mass
black holes and the rest being consistent with a population of IMBHs.
76
Chapter 3
Elemental Abundances of Nearby ULX Host
Galaxies
3.1 Introduction
Long exposure XMM-Newton observations of nearby galaxies offer new opportu-
nities to study various properties of ULX spectra. In the previous chapter (Chap-
ter 2), we analyzed ULX spectra from sources found in 32 nearby galaxies. Based
on spectral form, luminosity, and location within the optical host galaxy, we clas-
sified a population of high/soft state and low/hard state ULXs. In our initial
study, we focused on spectral fits with simple schematic models. However, for
the spectra withthe greatest number of counts, more detailed modelscanbe used
to glean more information from their spectra. One important investigation that
such high signal-to-noise spectra can aid with is a study of absorption along the
line of sight to the ULX source. Thus, in this chapter we examine the highest
signal-to-noiseXMM-Newtonspectrainanattempttounderstandthelineofsight
absorption.
Many similar studies have been done within our own Milky Way, where
77
X-ray absorption models have been used to determine column densities and
abundances of the interstellar medium. The general procedure used in these
studies is to fit absorption features in the spectra of bright, background X-ray
sources. Using this method, successful determinations have been made using
both background galaxy clusters (Baumgartner & Mushotzky 2006) and X-ray bi-
naries (Juett et al. 2004). Specifically, these studies made use of a bright, X-ray
source as a background through which they observed the 542eV absorption edge
produced by photo-ionization of the inner K-shell electrons of oxygen. Analo-
gous studies have been used in the radio (see Dickey & Lockman (1990)) to op-
tical regime, using quasars, supernovae, or stars as a background for hydrogen
absorption and 21-cm emission, as a means to measure hydrogen column densi-
ties and metal abundances.
Inthisstudy,weextendtheX-rayabsorptionstudiestoexternalgalaxiesusing
ultraluminousX-raysources. DuetotheirextremebrightnessintheX-rayregime
and their non-nuclear location in external galaxies, these sources are ideal for
probing the ISM of their host galaxies. Typical ULXs, from our Chapter 2 study,
haveGalacticline-of-sightcolumndensitiesofafew1020 cm?2 (Dickey & Lockman
1990) and measured X-ray column densities greater than 1021 cm?2 (Figure 2.8)
for the combined ULX and host galaxy. Thus, if the local environment of the ULX
contributes little absorption, the X-ray column density is dominated by the host
galaxy. One goal of this study is to determine whether this absorption is that of
the hostgalaxyor local ULXenvironment. Therefore, we comparethe X-ray mea-
sured hydrogen column density with H I measurements from alternate methods.
In addition to the brightness of ULXs and the relatively small Milky Way con-
tributiontotheirX-rayhydrogencolumndensities, theirwellcharacterizedX-ray
spectra make ULXs ideal for measuring absorption features of the ISM. Bright
78
ULXs (e.g. NGC 1313 X-1) typically have spectra that are well-fit by an absorbed
multi-component blackbody and power law model. However, there is discussion
over whether this standard model is the most physical model for the ULXs (see,
for example, Stobbart et al. (2006) or Gonc?alves & Soria (2006)). Different models
applied to the base ULX spectra can affect the absorption measurements, par-
ticularly in the softer part of the spectrum. Thus, in this chapter we investigate
the effect different soft component models have on the X-ray measured hydrogen
column density and elemental abundances (through the oxygen K-shell edge at
542eV and the iron L-shell edge at 851eV).
We use high signal-to-noise XMM Newton observations of ULXs to measure
hydrogen column densities and elemental abundances of oxygen and iron. Lo-
catedinexternalgalaxies,theX-rayspectralresolutionofavailableULXspectrais
not as good as those of Galactic X-ray binaries, which often have grating spectra
available (e.g. Juett et al. (2004)). Therefore, in order to be able to distinguish the
oxygen K-shell edge as well as the iron L-shell edge located at 851eV, we needed
observations with a large number of counts (? 5000counts). The X-ray obser-
vatory XMM-Newton, having a larger collecting area than Chandra, provides the
countsnecessaryinordertoconductthisstudy. Further, withrecent100ksXMM-
Newton observations available for the host galaxies of two well-studied ULXs
(Holmberg II and Holmberg IX), these observations allow for the added analysis
of Reflection Grating Spectrometer (RGS) spectra in addition to spectra from the
European Photon Imaging Cameras (EPIC). The spectral resolution of the EPIC
and RGS allow us to test different soft component models for the ULX sources, to
determine the effect of the model on absorption and abundance measurements.
79
3.2 Source Selection and Data Reduction
InChapter2, we presented theresults ofanarchivalXMM-Newton study ofULXs
in 32 nearby (< 8 Mpc) galaxies. We had extracted spectra for the brightest
sources in the observations, corresponding to > 400counts. In this study, we
chose to further analyze the spectra of the objects with the highest number of
counts (> 5000counts1). In addition to the 11 sources from Chapter 2, we include
an analysis of 3 additional sources: the two ULXs in the spiral galaxy NGC 4559
(observation 0152170501)and the bright source in M33 (observation 0102640101).
A full list of the 14 ULX sources, with details of the observations (including ex-
posure times and count rates), is found in Table 3.1.
The two ULX sources in NGC 4559 were originally studied by Vogler et al.
(1997) using ROSAT and Cropper et al. (2004) with XMM-Newton. We follow the
naming convention established in these papers. Both of these sources (X7 and
X10) were not included in Chapter 2 because the host galaxy?s distance is greater
than the 8Mpc limit we initially required. However, we include these sources
now due to the high number of counts (> 5000counts) in their spectra. Initially
we did not include M33 X-8 in our ULX survey due to its location in the cen-
ter of its host galaxy. This source, however, shows no evidence of being a low-
luminosity AGN and is more likely a black hole X-ray binary (Takano et al. 1994).
Since the initial study of Chapter 2, longer exposure time XMM-Newton ob-
servations have become available for three of our sources from the original study.
1In Appendix C, we show results of simulations to determine the number of counts necessary
to detect the oxygen and iron absorption edges. We found that > 40000counts are needed to con-
strainironabundancewithouthavinglargeerrorsinthemeasurement. Foroxygen,> 5000counts
are needed to constrain the abundance without large errors. See the appendix for further details.
80
Ta
ble
3.1
.
UL
X
Ob
ser
va
tio
nD
eta
ils
So
ur
ce
RA
(h
m
s)a
De
c(
??
??)
a
nH
G
A
L
b
Ob
sI
D
Ex
po
su
re
Tim
e(
s)c
Co
un
tR
ate
(ct
ss
?
1 )
c
NG
C2
47
XM
M
1
00
47
03
.8
-20
47
46
.2
1.5
4
01
10
99
03
01
34
58
,1
38
9,
13
79
0.2
0,
0.0
6,
0.0
6
NG
C2
53
XM
M
2
00
47
22
.4
-25
20
55
.2
1.4
0
01
52
02
01
01
-,1
03
47
,1
03
04
-,0
.08
,0
.09
NG
C3
00
XM
M
1
00
55
09
.9
-37
42
13
.9
3.1
1
01
12
80
01
01
67
78
,2
24
8,
24
53
0.1
9,
0.0
5,
0.0
5
M
33
X-
8
01
33
50
.9
+3
03
93
6.1
5.5
8
01
02
64
01
01
11
91
9,
12
14
7,
12
14
2
0.0
6,
0.0
2,
0.0
2
NG
C1
31
3X
M
M
3
03
18
22
.5
-66
36
06
.2
4.0
01
06
86
01
01
69
60
,2
17
9,
17
93
0.2
33
,0
.08
,0
.07
Ho
lm
II
XM
M
1
08
19
28
.8
+7
04
22
0.3
3.4
2
01
12
52
07
01
31
05
2,
12
57
,1
08
07
2.7
2,
0.7
6,
0.7
3
??
?
??
?
??
?
??
?
02
00
47
01
01
56
98
7,6
57
58
,6
57
66
3.2
9,
0.9
5,
0.9
4
??
?
??
?
??
?
??
?
02
00
47
01
01
41
81
6,
41
80
2(
RG
S)
0.0
9,
0.1
1(
RG
S)
M
81
XM
M
1
09
55
32
.9
+6
90
03
4.8
4.1
2
01
11
80
01
01
50
78
8,
-,1
89
88
0.5
1,
-,0
.22
??
?
??
?
??
?
??
?
02
00
98
01
01
-,1
11
91
0,
11
19
30
-,0
.11
1,
0.1
04
Ho
lIX
XM
M
1
09
57
53
.3
+6
90
34
8.7
4.0
01
12
52
10
01
14
97
6,
65
46
,6
58
6
2.0
7,
0.6
4,
0.6
5
??
?
??
?
??
?
??
?
02
00
98
01
01
10
40
10
,1
11
76
0,
11
18
30
1.6
9,
0.5
1,
0.5
1
??
?
??
?
??
?
??
?
02
00
98
01
01
10
38
40
,1
03
83
0(
RG
S)
0.0
3,
0.0
4(
RG
S)
NG
C
45
59
X7
12
35
51
.8
+2
75
60
4
1.5
1
01
52
17
05
01
34
51
7,
-,-
0.3
18
,-,
-
NG
C
45
59
X1
0
12
35
58
.6
+2
75
74
0.8
1.5
1
01
52
17
05
01
34
50
9,
-,-
0.2
38
,-,
-
NG
C4
63
1X
M
M
1
12
41
55
.8
+3
23
21
4
1.2
8
01
10
90
02
01
50
93
,1
96
9,
17
62
0.1
3,
0.0
4,
0.0
4
NG
C5
20
4X
M
M
1
13
29
38
.5
+5
82
50
3.6
1.4
2
01
42
77
01
01
99
81
,3
35
2,
33
84
0.6
28
,0
.17
7,
0.1
79
??
?
??
?
??
?
??
?
01
42
77
03
01
92
31
,2
28
4,
23
49
0.8
55
,0
.24
7,
0.2
58
M
83
XM
M
1
13
37
19
.8
-29
53
49
.8
3.9
4
01
10
91
02
01
30
74
,9
27
,9
87
0.1
2,
0.0
33
,0
.02
5
NG
C5
40
8X
M
M
1d
14
03
19
.8
-41
22
59
.3
5.7
3
01
12
29
06
01
59
32
,2
03
6,
20
77
0.1
28
,0
.03
2,
0.0
33
a R
A
an
dD
ec
va
lue
sq
uo
ted
are
th
es
ou
rce
po
sit
ion
sf
ro
m
th
eE
PI
C-
pn
im
ag
es
.
b M
ilk
yW
ay
hy
dr
og
en
co
lum
nd
en
sit
ya
lon
gt
he
lin
eo
fs
igh
tf
ro
m
Di
ck
ey
&
Lo
ck
ma
n(
19
90
)in
un
its
of
10
20
cm
?
2
c E
xp
os
ur
et
im
es
an
dc
ou
nt
rat
es
are
lis
ted
for
th
eE
PI
C
pn
,M
OS
1,
M
OS
2,
an
dR
GS
(1
an
d2
),w
he
re
av
ail
ab
le.
No
te
th
at
for
so
ur
ce
sw
ith
mu
lti
ple
ob
se
rv
ati
on
s,
de
tai
ls
for
th
ea
dd
iti
on
al
ob
se
rv
ati
on
sa
re
lis
ted
be
low
th
efi
rst
ob
se
rv
ati
on
.D
eta
ils
of
th
e
sp
ec
ific
RG
S(
RG
S1
an
dR
GS
2)
ob
se
rv
ati
on
sa
re
ind
ica
ted
wi
th
(R
GS
).
d A
se
co
nd
ob
se
rv
ati
on
of
NG
C
54
08
XM
M
1i
sr
efe
rre
dt
oi
n?
3.4
.2.
Th
is
ob
se
rv
ati
on
(03
02
90
01
01
)w
ith
an
ex
po
su
re
tim
eo
f
13
03
35
sw
as
pr
op
rie
tar
ya
tt
he
tim
eo
fa
na
lys
is.
81
With the permission of Tod Strohmayer, we include data from the 100ks XMM-
Newton observation (0200980101) of Holmberg IX. These data include pn and
MOS spectra of Holmberg IX XMM1 and MOS spectra of M81 XMM1. We also
include a 100ks XMM-Newton observation (0200470101) of Holmberg II XMM1
that became public after the Chapter 2 study. The EPIC and RGS spectra from
this observation of Holmberg II XMM1 were first analyzed by Goad et al. (2006).
For Holmberg IX XMM1 and Holmberg II XMM1, the 100ks exposures provided
us with the opportunity to extract and analyze, in addition to the EPIC spectra,
spectra from the RGS detectors. Thus, we include an analysis of RGS spectra for
both Holmberg II XMM1 and Holmberg IX XMM1.
For the EPIC spectra we added to our original sample, we followed the same
reduction method as in Chapter 2. For observations that were processed with
an earlier version of the XMM-Newton Science Analysis System (SAS) (we used
SAS version 6.0), the observation data files (ODF) were used to produce cali-
brated photoneventfiles forthe EPIC-MOS andpncamerasusing thecommands
emchain and epchain. The events tables were filtered using the standard crite-
ria outlined in the XMM ABC Guide. For the MOS data (both MOS1 and MOS2
cameras), good events constitute those with a pulse height in the range of 0.2 to
12 keV and event patterns that are characterized as 0-12 (single, double, triple,
and quadruple pixel events). For the pn camera, only patterns of 0-4 (single and
double pixel events) are kept, with the energy range for the pulse height set be-
tween 0.2 and 15 keV. The selection expression ?FLAG == 0? was used to exclude
bad pixels and events too close to the edges of the CCD chips. Time filtering was
applied as needed by editing the light curve produced in xmmselect. For the
EPIC observations, time periods in the observation with high count rates (flares)
in the MOS and pn were cut using the command tabgtigen with the ?RATE<?
82
command set to 5ctss?1 for MOS detectors and 20ctss?1 for the pn detector.
We extracted source and background spectra along with response and ancil-
lary response matrices using the SAS task especget. The source spectra were
extracted from circular regions, typically with radii of 20??. This region was ad-
justed depending on the size of the source and the proximity of the source to
either another source or the edge of a CCD chip. We extracted background spec-
trafromannularregions, except whenthesource wasnearanother source ornear
the edge of a chip. In this case, in order to avoid source confusion, we extracted
background spectra from a circular region located near the source and on the
same CCD chip (as for Holmberg IX XMM1 and M81 XMM1).
FortheRGSspectraweextractedfirstandsecondorderspectraforthesources
Holmberg IX XMM1 (0200980101) and Holmberg II XMM1 (0200470101) using
the rgsproc command. The RA and Dec values used to extract the RGS spec-
tra were obtained from the EPIC pn data and are the values quoted in Chap-
ter 2. Time filtering was applied as for the EPIC data with tabgtigen, where
the ?RATE<? command was set to 0.5ctss?1 for Holmberg IX and 0.1ctss?1 for
Holmberg II. Once the spectra were obtained, for the RGS as well as EPIC data,
they were rebinned to require at least 20 counts per bin, using the command
grppha in LHEASOFT.
3.3 Spectral Fitting
Spectral fitting proceeded using XSPEC v11.3.1. For the RGS spectra, we simul-
taneously fit the first order spectra from both RGS1 and RGS2 in the RGS band
(0.33 - 2.5keV). For the EPIC spectra, we fit the pn and MOS spectra simultane-
ously in the 0.3-10keV energy range. We allowed a free normalization constant
83
to account for the differences in flux calibration between the three EPIC cameras.
Both the RGS and EPIC spectra were fit separately.
In the Chapter 2 study, we had fit all of the sources with three standard mod-
els: anabsorbedpowerlaw,anabsorbedbremsstrahlungmodel, andanabsorbed
combined blackbody and power law model. We used the XSPEC model wabs to
account for absorption from the Milky Way and the host galaxy/ULX contribu-
tion. This model is a photo-electric absorption model using the cross-sections of
Morrison & McCammon (1983) and the solar abundances of Anders & Ebihara
(1982). We found that the spectra of the brightest ULXs were typically best-
fit by an absorbed combined blackbody and power law model (fit in XSPEC
as wabs*wabs*(bbody + pow) where the first wabs model was fit to the
Dickey & Lockman (1990) Milky Way value and the second wabs model was fit
to the remaining host galaxy/ULX contribution). Likewise, we began our study
by fitting the additional sources (M33 X-8, NGC 4559 X-7, and NGC 4559 X-10)
with the same three models noted above. We found that these sources were well
fit by the absorbed blackbody and power law model with ?2/dof ? 1.0.
For M33 X-8, the source was well-fit with an absorbed blackbody and power
law with best-fit parameters: nH = 1.67+0.09?0.08 ? 1021 cm?2, kT= 0.74+0.02?0.02 keV, ? =
2.46+0.06?0.05, and ?2 = 1579.8/1533 dof, where nH represents the host galaxy/ULX
hydrogencolumn(theMilkyWaycontributionwasfixedtotheDickey & Lockman
(1990) value listed in Table 3.1). In Chapter 2 we noted that ULX sources well-fit
with the combined blackbody and power law model with a higher disk tem-
perature (? 1keV) and a lower flux were often well-fit by an absorbed Comp-
tonization model (XSPEC model compst). We note that for M33 X-8 the in-
verse Compton scattering model also fits the data well (wabs*compst) but with
a ?2 = 1614.9/1536 dof. Despite the larger ?2 value, the compst model bet-
84
ter fits the residuals from the spectra?s sloping high energy tail. The best-fit ab-
sorbed blackbody and power law parameters for NGC 4559 X7 (12h 35m 51.8s,
27? 56? 4??) were: nH = 1.51+0.04?0.01 ? 1021 cm?2, kT= 0.13+0.01?0.02 keV, ? = 2.16+0.10?0.04, and
?2 = 410.3/369 dof. For NGC 4559 X10 (12h 35m 58.6s, 27? 57? 40.8??), we found:
nH = 1.14+0.02?0.01 ? 1021 cm?2, kT= 0.96+0.26?0.20 keV, ? = 2.13+0.06?0.05, and ?2 = 292.3/330
dof. X10, as noted in Cropper et al. (2004), is also well fit by a Comptonization
model.
For the sources M33 X-8 and NGC 4559 X10, we noted that their spectra were
well-fit by either an absorbed blackbody and power law or an absorbed Comp-
tonization model. This brings up one issue surrounding ULX spectra. Namely,
ambiguity over their spectral form. In this chapter, we assume that there is an
analogy between ULXs and Galactic black hole binaries. Thus, we assume that
the accurate ULX spectrum for these bright sources is a hard component (well
modeled by a power law) and a soft component (which we assume as a ther-
mal contribution from an accretion disk). The form of the soft component, in
particular, will affect the measured absorption and abundance values. Thus, an
investigation of this component and its affect on the absorption model is impor-
tant.
The soft component of ULXs is most often modeled as a thermal component
originating from an accretion disk surrounding a central black hole. There are
numerous disk models applied to model this possibly thermal component. In
Chapter2, wemodeledthiscomponentasasimpleblackbody(inXSPECbbody).
While an accretion disk is expected to have a range of temperatures, empirically
a single blackbody is a good fit to low signal-to-noise spectra. A simple absorbed
blackbody and power law model was used for Galactic black hole X-ray binaries
in the 1980s when the quality of data for these sources was analogous to that for
85
ULX sources today. As the next step in accretion disk models, the soft component
of ULX spectra is often modeled as an optically thick, geometrically thin, multi-
component blackbody disk (a multi-component disk or MCD model, diskbb in
XSPEC) (Mitsuda et al. 1984).
Two disk models that are used to fit the soft component of ULX spectra with
more physical accuracy are the XSPEC models diskpn and grad. The diskpn
model is an extension of the MCD (diskbb) model which includes temperature
corrections for the disk close to the black hole. The grad model (Ebisawa et al.
1991; Hanawa 1989) is a multi-component disk model that unlike the diskbb
or diskpn model, incorporates the effects of general relativity. One of the ad-
vantages of the grad model is that it fits the spectra for mass (Mgrad) and mass
accretion rates ( ?M) given a few initial assumptions (distance to the source, disk
inclination angle, and the ratio of the color temperature to the effective tempera-
ture).
In addition to thermal models, other models have been suggested to explain
the soft component. Two of these models are ionized reflection and the warm ab-
sorber model (see ? 1.2). Both of these models have been applied to low-redshift
PG quasars, sources with blackbody temperatures of 150eV. This is relevant to
ULXs since many spectral fits of ULXs require cool accretion disk temperatures
ofapproximately100eV(Miller et al.2003,2004b;Roberts et al.2005;Winter et al.
2006b). The reflection model suggests that the soft component results from X-ray
ionizedreflection. Inthismodel, back-scattering andfluorescence ofX-raysinthe
disk, as well as radiative recombination, cause elements with smaller ionization
potentials (e.g. C, O, N) to become highly ionized. Ross & Fabian (2005) note that
a relativistically blurred X-ray ionization model folded through an XMM-Newton
pn response matrix, is well-fit by a blackbody with a temperature of 150eV, the
86
same value that is seen in PG quasars and many ULXs. In the warm absorber
model, absorption edges and lines from an absorbing material close to the X-ray
source appear as a thermal component in poorer quality spectra. With increased
spectral resolution, the numerous absorption edges and lines are distinguishable.
Thewarmabsorbermaybetheresultofastrong,mildlyrelativisticwindfromthe
disk as suggested by Gierli?nski & Done (2004). Indeed, the presence of a warm
absorber is well known and studied in many Seyfert galaxies (e.g. NGC 3783 and
MCG?6-30-15).
As noted above, for subsequent spectral fits we assume a thermal model for
the soft component of the ULX spectra. In the following section (? 3.4), we will
discuss the effect that different thermal models have on the measured absorp-
tion values. In order to measure the hydrogen column density, we fit the spec-
tra of sources listed in Table 3.1 with the more sophisticated Tuebingen-Boulder
ISM absorption model of Wilms et al. (2000) (tbabs, tbvarabs in XSPEC). This
model accounts for X-ray absorption resulting from contributions from X-ray ab-
sorption from the gas phase of the ISM, grains, and molecules. The model uses
updated solar abundances and photoionization cross-sections. For both the EPIC
and RGS spectra, we accounted for Galactic hydrogen absorption by setting the
column density of the tbabs model equal to the Milky Way hydrogen column
density along the line of sight to the host galaxy. These column densities are
quoted in Table 3.1 and are all less than 6?1020 cm?2.
Since the measured X-ray column densities for ULXs are typically an order of
magnitude higher than the Dickey & Lockman (1990) Milky Way values (Figure
2.8), we are confident that the additional absorption measured is not from the
Milky Way. To determine the host galaxy?s hydrogen column density and the
abundances of elements along the line of sight (oxygen and iron), we used the
87
tbvarabs model. The tbvarabs model accounts for X-ray absorption due to
photo-ionization. It includes the effects due to the H2 molecule and depletion
of metals in grains. The model allows for individual fits to abundant elements
(He through Ni), H2, and depletion of elements in grains. We initially allowed
the hydrogen column density to vary, fixing all other tbvarabs parameters to
their defaults. The Galactic column density remained fixed (using the tbabs
model),whiletheotherparameters(powerlawandblackbodycomponents)were
allowed to vary. This was modeled as tbabs*tbvarabs*(bbody + pow) in
XSPEC, wheretbabswas fixed to the Milky Way valueandtbvarabswas used
to fit the absorption from the ULX/host galaxy contribution. With the best-fit
hydrogen column density (from tbvarabs), we allowed the oxygen abundance
and then the iron abundance to vary from the solar abundances. Allowing these
parameters to float provides measurements of the depth of the oxygen K-shell
edge at 542eV and the iron L-shell edge at 851eV.
As in Baumgartner & Mushotzky (2006), we found that for our 14 sources, the
oxygen absorption values from the tbvarabsmodel yielded different values for
the EPIC pn and MOS spectra. They attribute this difference to an error in the re-
sponse matrices of the MOS detectors, possibly caused by outgassing of organic
material onto the surface of the MOS detectors2. Thus, we follow the procedure
of Baumgartner & Mushotzky (2006) in adding an edgemodel to account for the
differences. We add an extra edge component to the MOS1 and MOS2 detectors
at an energy of 0.53keV with optical depths of 0.22 and 0.20 respectively (mod-
eled astbabs*tbvarabs*edge*(bbody + pow)). This is the template model
used for all spectral fits mentioned throughout the rest of the chapter. However,
we also discuss models where we replace the simple blackbody model with the
2This effect is now calibrated as the MOS ?patch? effect as of SAS 7.0.
88
other thermal models discussed (diskpn, grad).
Anadditionalproblemwenoteinvolvesthehydrogencolumndensitiesalong
the line of sight. Baumgartner & Mushotzky (2006) note that a column density
(nH) greater than 5?1020 cm?2 is necessary to allow for good oxygen abundance
measurements. We also note that at very high nH values the oxygen edge is un-
detectable, where high absorption values cause the signal-to-noise to be too low
to detect the edge.
3.4 Nature of the Soft Component
In the previous section we outlined three possible origins for the soft component
in ULXs (X-ray reflection, thermal emission from an accretion disk, and a warm
absorber model). For our purposes, we assume a thermal component origin. We
chose the thermal model since, as described in Chapter 2, the spectra of ULXs
are well described with models used to fit Galactic black hole X-ray binaries. To
gauge the effect different disk models (i.e. bbody, diskbb, diskpn, grad) may
have on the measured abundances, we discuss different disk models in?3.4.1. In
? 3.4.3, we assess the usefulness of the model (absorbed disk and power law) for
measuring the galactic/ULX hydrogen column density and the oxygen and iron
abundances by discussing the physical plausibility of this model.
For sources with at least 5000 counts, the 14 sources listed in Table 3.1, we
fitted the grad model for the soft-component of the spectra (XSPEC model
tbabs*tbvarabs*edge*(grad + pow)). Two of the additional parameters
of the grad model include the disk inclination angle and the ratio of Tcolor/Teff.
We find that changing the inclination angle for Holmberg II XMM1 has a min-
imal change on ?2 (< 1), ?, and absorbing column density. Therefore, we set
89
the disk inclination angle to the default value of 60?for all of the sources. The
other parameter, Tcolor/Teff, was set to 1.7, the conventional value adopted for
stellar mass BH sources (Shimura & Takahara 1995; Sobczak et al. 1999). Here,
Teff is the effective temperature of the photosphere of the disk, while Tcolor is the
color temperature of the disk. The color temperature is the observed tempera-
ture, which is slightly modified from Teff as the photosphere spectrum is modi-
fied passing through the outer layers of the disk. We note however that there is
theoretical evidence supporting a non-constant ratio Tcolor/Teff, ranging from 1.7
to 3 and depending upon accretion rate (Merloni et al. 2000). Allowing the ratio
to vary for Holmberg II XMM1, we found the best-fit value corresponded to 1.7.
For the sake of uniformity of our models, we fixed the ratio to 1.7 for all of the
ULXs.
In Table 3.2, we list the best-fit parameters for EPIC spectra using this model.
Rows with no source name indicated represent an additional observation of the
previous ULX (as indicated in Table 3.1). Table 3.3 provides the best-fit param-
eters for the RGS spectra. The important measurements to note are the host
galaxy/ULX contribution to the hydrogen column density, oxygen abundance,
and iron abundance from the tbvarabs model. All errors quoted in this pa-
per are for the 90% confidence level for one degree of freedom (??2 = 2.76).
We provide a representative spectral fit in Figure 3.1 for the long observation of
Holmberg IX XMM1.
90
10
?4
10
?3
0.01
0.1
1
normalized counts
1 100.5 2 5
?0.2
?0.1
0
0.1
0.2
residuals
channel energy (keV)
Figure 3.1: EPIC spectrum of observation 0200980101 of Holmberg IX XMM1.
This plot shows the spectrum fit with the tbvarabs*tbabs*edge*(grad +
pow) model and the residuals from this fit. The best fit parameters are listed in
Table 3.2.
91
Table 3.2. Model Fits to EPIC spectra with the grad Model
Source nHa O abund.b Fe abund.b Mass (M?) ?M/ ?MEddc ? ?2/dof countsd
HolmII XMM1 0.12+0.01?0.01 0.90+0.11?0.11 0.0+0.15?0.0 141+21?32 0.15+0.02?0.09 2.36+0.05?0.05 1.11 342874
??? 0.17+0.02?0.04 1.27+0.23?0.42 3.15+1.29?1.74 791+209?363 0.07+0.03?0.03 2.29+0.07?0.10 0.98 43116
HolmIX XMM1 0.19+0.02?0.02 1.34+0.12?0.13 2.07+0.12?0.13 382+93?76 0.09+0.02?0.02 1.38+.025?.025 1.05 148061
??? 0.33+0.08?0.09 1.38+0.29?0.40 3.91+1.09?1.86 1181+1516?719 0.11+0.20?0.07 1.73+0.09?0.09 0.64 28108
M33 X-8 0.20+0.04?0.04 1.04+0.13?0.18 1.74+0.62?0.74 5.24+0.38?0.35 1.02+0.09?0.08 2.83+0.34?0.31 0.97 123903
M81 XMM1 0.66+0.11?0.13 1.20+0.08?0.10 2.04+0.44?0.57 8.61+33.58?5.03 2.32+44.8?1.31 4.84+0.58?0.66 1.02 69776
??? 0.27+0.21?0.09 1.42+0.58?0.50 1.67+1.61?1.67 4.2+7.4?1.8 8.31+4.36?5.98 2.08+1.58?0.75 0.91 31731
NGC253 XMM2 0.53+0.07?0.12 1.55+0.20?0.29 2.73+1.17?1.46 4140+860?2213 0.05+0.03?0.03 2.32+0.10?0.12 0.96 20651
NGC5204 XMM1 0.10+0.04?0.03 1.42+0.55?0.75 0.0+1.84?0.0 464+366?219 0.04+0.08?0.02 1.92+0.08?0.08 0.96 16717
??? 0.13+0.05?0.03 0.77+0.60?0.77 0.0+1.70?0.0 449+490?217 0.06+0.12?0.03 2.02+0.13?0.14 0.93 13864
NGC1313 XMM3 0.67+0.02?0.04 1.37+0.13?0.14 0.0+0.34?0.0 5000+0.03?1160 0.08+0?0.02 2.66+0.08?0.09 1.02 10932
NGC300 XMM1 0.15+0.05?0.04 2.46+0.45?0.43 0.0+1.77?0.0 417+312?175 0.02+0.03?0.01 2.46+0.10?0.11 1.01 11479
N4559 X-7 0.16+0.05?0.03 0.46+0.53?0.46 0.0+1.00?0.0 755+901?443 0.06+0.12?0.02 2.10+0.10?0.10 0.83 12506
NGC4631 XMM1 0.28+0.06?0.06 0.62+0.43?0.58 0.11+1.86?0.11 5.5+11.4?18.6 0.84+0.88?0.42 5.71+2.03?0.31 1.06 8824
NGC5408 X-1 0.09+0.03?0.03 1.99+0.26?0.52 5.0+0.0?1.68 1477+342?477 0.10+0.15?0.09 2.53+0.20?0.16 0.96 10045
NGC4559 X-10 0.12+0.05?0.04 1.28+0.61?0.81 0.0+2.30?0.0 5.34+7.88?2.75 2.74+2.52?1.86 2.09+0.52?0.28 0.77 8837
NGC247 XMM1 0.44+0.18?0.12 1.09+0.35?0.28 0.0+1.45?0.0 1717+3205?929 0.05+0.02?0.03 3.60+6.40?6.60 0.60 6226
M83 XMM1 0.13+0.09?0.10 1.70+1.12?1.65 0.0+3.92?0.0 6.34+13.4?4.3 0.92+0.70?0.70 2.64+0.88?0.64 0.83 4988
aHydrogen column density determined from tbvarabs in units of 1022 cm?2. The Galactic value of nH was fixed to the
Dickey & Lockman (1990) value with the tbabs model.
bElement abundance relative to the Wilms solar abundance from the tbvarabs model
cRatio of mass accretion rate from the grad model to Eddington accretion rate (see ? 3.4)
dTotal number of photon counts from pn and MOS detectors
Table 3.3. Model Fits to the RGS spectra with the grad Model
Source nHa O abund.b Fe abund.b Mass (M?) ?M/ ?MEddc ? ?2/dof countsd
HolmII XMM1 0.08+0.05?0.02 0.65+0.61?0.64 0.0+1.51 147+88?83 0.24+0.11?0.12 1.58+6.75?4.58 444.4/442 9521
HolmIX XMM1 0.29+0.12?0.08 0.68+0.30?0.34 0.36+1.36?0.36 774+4226?570 0.09+0.16?0.03 1.43+0.45?0.98 290.4/339 10807
aHydrogen column density determined from tbvarabs in units of 1022 cm?2. The Galactic value of nH was fixed to the
Dickey & Lockman (1990) value with the tbabs model.
bElement abundance relative to the Wilms solar abundance
cRatio of mass accretion rate from the grad model to Eddington accretion rate (see ? 3.4)
dTotal number of photon counts from RGS1 and RGS2 detectors
92
Table 3.4. Alternative Thermal Disk Model Fits
Source nHa Oxygen abundanceb Iron abundanceb kT ? ?2/dof countsc
tbabs*tbvarabs*edge*(bbody + pow)
HolmII XMM1 0.12+0.01?0.01 0.97+0.11?0.12 0.0+0.32 0.24+0.01?0.01 2.41+0.03?0.03 2216.5/2018 342874
??? 0.14+0.03?0.02 1.37+0.38?0.39 4.38+0.62?1.8 0.16+0.01?0.01 2.31+0.08?0.04 952.7/973 43116
HolmIX XMM1 0.14+0.02?0.01 1.51+0.19?0.15 3.37+0.99?0.80 0.19+0.01?0.01 1.40+0.02?0.01 2856.7/2752 148061
??? 0.23+0.10?0.04 1.37+0.43?0.44 4.75+0.25?1.94 0.17+0.02?0.03 1.71+0.14?0.06 559.5/880 28108
M33 X-8 0.22+0.01?0.14 1.24+0.10?0.10 2.38+0.31?0.60 0.76+0.02?0.02 2.51+0.07?0.06 1528.8/1533 123903
M81 XMM1 0.39+0.03?0.02 1.13+0.15?0.07 1.78+0.65?0.40 0.90+0.03?0.02 2.70+0.05?0.05 1240.2/1241 69776
??? 0.26+0.09?0.04 1.46+0.26?0.49 1.61+2.02?1.61 0.99+0.38?0.28 1.76+0.13?0.09 627.4/668 31731
NGC253 XMM2 0.22+0.05?0.05 1.21+0.40?0.37 1.53+1.67?1.53 0.73+0.14?0.11 2.16+0.08?0.23 460.8/496 20651
tbabs*tbvarabs*edge*(diskbb + pow)
HolmII XMM1 0.11+0.001?0.01 0.65+0.09?0.05 0.0+0.06 0.34+0.02?0.02 2.38+0.05?0.04 2241.8/2018 342874
??? 0.15+0.03?0.02 1.05+0.34?0.38 2.67+1.65?1.71 0.21+0.04?0.01 2.27+0.06?0.07 954.2/973 43116
HolmIX XMM1 0.19+0.02?0.02 1.38+0.11?0.16 2.39+0.58?0.84 0.24+0.02?0.01 1.39+0.02?0.03 2878.9/2752 148061
??? 0.30+0.11?0.07 1.36+0.32?0.38 4.03+0.97?1.65 0.20+0.04?0.03 1.72+0.10?0.07 559.8/880 28108
M33 X-8 0.19+0.04?0.03 1.04+0.15?0.16 1.75+0.68?0.71 1.17+0.05?0.05 2.47+0.21?0.15 1530.4/1533 123903
M81 XMM1 0.63+0.08?0.15 1.19+0.09?0.15 2.07+0.40?0.69 1.38+0.03?0.08 4.08+0.75?0.59 1247.7/1241 69776
??? 0.27+0.18?0.08 1.42+0.57?0.49 1.63+1.69?1.63 2.06+0.55?0.97 1.98+1.40?0.58 626.9/668 31731
NGC253 XMM2 0.21+0.10?0.07 1.16+0.53?0.37 1.64+1.64?1.64 1.27+0.22?0.27 2.27+0.83?0.62 461.3/496 20651
aHydrogen column density determined from tbvarabs in units of 1022 cm?2. The Galactic value of nH was fixed to the
Dickey & Lockman (1990) value with the tbabs model.
bElement abundance relative to the Wilms solar abundance from the tbvarabs model
cTotal number of photon counts from pn and MOS detectors
3.4.1 Comparison of Thermal Disk Models
We chose the general relativistic disk model, grad, because it is the most physi-
cally accurate of the various simple accretion disk models. Also, the grad model
requires few initial parameters while making a direct calculation of the mass and
mass accretion rate of the black hole. In order to compare the grad results with
an alternate model, we fit the highest signal-to-noise observation of Holmberg IX
XMM1 with thediskpnmodel (with the inner radius of the disk set to the radius
of marginal stability or 6 times the Schwarzschild radius) in place of the grad
model. We found that the values obtained agreed with those of the grad model.
The mass from the diskpnmodel was slightly lower than the grad model value
(382Msun for the grad compared to 358M?, corresponding to kT= 0.23keV for
93
diskpn) and the iron abundance was slightly higher at 2.33 compared to 2.07
with the grad model. Since the diskpn and grad models yield similar values,
either of these disk models would be sufficient for determining the hydrogen col-
umn density and oxygen and iron abundances.
In order to test the effect the thermal model has on the tbvarabs mea-
sured parameters (i.e. nH and abundances), we wanted to compare the re-
sults from the more physical disk model (grad) with less physical models
(bbody and diskbb). In order to compare the results from these three mod-
els, we chose to examine the best-fit parameters for spectra with the highest
signal-to-noise (using sources with at least 20000counts). Thus, in Table 3.4 we
list the best-fit parameters obtained using an absorbed blackbody and power
law (tbabs*tbvarabs*edge*(bbody + pow))and an absorbeddiskbband
power law model (tbabs*tbvarabs*edge*(diskbb + pow)) for sources
with at least 20000 counts. Subsequently in this chapter, when we refer to the
bbody, diskbb, or grad models, we are referring to the model fits used in Ta-
bles 3.2- 3.4.
Thecurvature, orshapeofthespectra atlowenergies, forthebbody,diskbb,
and grad models is different, presumably affecting the column density and ab-
sorption values. From a comparison of these models, comparing the mean values
from all of the observations with > 20000counts listed in Tables 3.2 and 3.4, we
find that the column densities obtained with the diskbb model are nearly iden-
tical to those with the grad model (the exception is the source NGC 253 XMM2).
The MCD (diskbb) model hydrogen column densities are approximately 6%
higher than those of the blackbody model (excluding the first M81 XMM1 ob-
servation). The average power law indices for the three models agree within a
factor of 5%, with the bbody model having the lowest and the grad having the
94
highest average power law index (excluding the first M81 XMM1 observation).
Thus, the difference in both power law index and hydrogen column density is
negligible on average, with the exceptions of NGC 253 XMM2 and M81 XMM1.
Foracomparisonofthederivedabundances,theoxygenabundanceisroughly
the same between the grad and bbody model. These values are approximately
10% higher than those derived from the MCD (diskbb) model. The iron abun-
dances, however, varymore from modelto model. The blackbody modelaverage
iron abundance is ? 13% higher than that of the grad model and ? 18% higher
than that of the diskbb model. Thus, there are no significant changes to the
model parameters in changing the base accretion disk model. The robustness of
our results are verified by simulations described in the appendix.
3.4.2 Power law/Thermal disk Degeneracies
Inmodelingoursourceswiththetbvarabs*tbabs*edge*(grad+pow)model,
wefound thatsome oftheULXsources were well-fitbyamodelwherethepower
law component dominates the low energy spectrum, also seen in Stobbart et al.
(2006). This type of spectral fit typically yields a steeper power law index (? >
3.0) and a low mass (M< 10M?). For two sources, M33 X-8 and M81 XMM1, this
model fit was a much better fit than a higher mass model with ??2 of 476 and
190 respectively. However, for some sources, there was a degeneracy between
the two models (high mass and low mass/steep power law). We illustrate this
degeneracy with observation 0112290601 of the source NGC 5408 XMM1, show-
ing the high mass model and spectral fit in Figure 3.2 and the low mass model
and spectral fit in Figure 3.3. For spectra exhibiting this degenerate solution, we
include the low-mass/steep power law fits in Table 3.5. These fits all exhibit, in
addition to low masses, solutions with ?M >> ?MEdd.
95
1 100.5 2 5
10
?6
10
?5
10
?4
10
?3
0.01
Photons/cm
2  s keV
channel energy (keV)
Figure 3.2: Plot of the EPIC unfolded spectrum for the NGC 5408 XMM1 ULX
with the high mass solution (see Table 3.2). In the plot, the absorbed grad and
absorbed power law model components are plotted with light gray lines while
the combined model is plotted in dark gray. The thermal component clearly
dominates the low energy spectrum with this model.
96
1 100.5 2 5
10
?6
10
?5
10
?4
10
?3
0.01
Photons/cm
2  s keV
channel energy (keV)
Figure 3.3: Plot of the EPIC unfolded spectrum for the NGC 5408 XMM1 ULX
with the low mass solution (see Table 3.5). In the plot, the absorbed grad and
absorbed power law model components are plotted with light gray lines while
the combinedmodel isplotted indark gray. Notice that thelow energyspectrum
is dominated by the power law component with a weak contribution from the
disk model (grad).
97
Table 3.5. Spectral Fits for Sources with Degenerate Solutions
Source nHa O abund.b Fe abund.b Mass (M?) ?M/ ?MEddc ? ?2/dof countsd
HolmII XMM1 0.20+0.04?0.04 0.83+0.17?0.27 1.11+0.83?1.11 2.24+1.52?1.24 7.79+1.52?1.95 3.10+0.31?0.30 0.99 43116
HolmIX XMM1 0.39+0.13?0.12 1.05+0.21?0.33 2.08+0.95?1.45 3.91+1.47?1.59 22.0+1.79?2.86 3.51+0.79?0.87 0.64 28108
NGC253 XMM2 0.18+0.17?0.08 1.10+0.42?0.55 1.32+1.90?1.32 6.82+4.19?2.22 1.27+0.71?0.40 2.15+0.84?0.58 0.93 20651
NGC5204 XMM1 0.18+0.05?0.04 1.00+0.30?0.40 0.0+0.59?0.0 3.21+0.98?0.95 6.86+0.67?0.09 3.28+0.39?0.38 0.95 16717
??? 0.20+0.06?0.05 0.63+0.35?0.48 0.0+0.62?0.0 2.95+2.13?1.84 7.37+1.35?2.08 3.16+0.50?0.44 0.94 13864
NGC300 XMM1 0.21+0.05?0.03 1.59+0.22?0.26 0.0+0.53?0.0 1.69+0.85?0.32 0.73+0.18?0.13 3.89+0.21?0.32 1.00 11479
N4559 X-7 0.20+0.06?0.05 0.34+0.38?0.34 0.0+0.48?0.0 5.79+2.82?2.75 6.63+1.44?2.42 3.13+0.55?0.50 0.83 12506
NGC5408 X-1 0.18+0.05?0.04 0.87+0.28?0.42 0.45+1.75?0.45 2.29+1.57?1.28 4.87+1.23?1.02 4.09+0.28?0.34 0.89 10045
aHydrogen column density determined from tbvarabs in units of 1022 cm?2. The Galactic value of nH was fixed to the
Dickey & Lockman (1990) value with the tbabs model.
bElement abundance relative to the Wilms solar abundance from the tbvarabs model
cRatio of mass accretion rate from the grad model to Eddington accretion rate (see ? 3.4)
dTotal number of photon counts from pn and MOS detectors
In Stobbart et al. (2006), the authors discussed the same issues in fitting the
XMM spectra of 13 ULXs. They found that two sources, M33 X-8 and NGC 2403
X-1, were best fitby the model with a power law fit to the low energy portion and
thermal model at higher energy. They also indicated six sources where an ambi-
guity existed between the two models. To understand the spectra of the sources
where both models provided good fits to the data, we further investigated the
spectra of sources with multiple observations.
For Holmberg II X-1 and Holmberg IX X-1 our spectral fits include multi-
ple observations (a shorter and a 100kilosecond observation). For each of these
sources, we found that the shorter observation could be fit with either a high
mass or low mass solution. When we fit the 100kilosecond observation, how-
ever, the high mass model was a much better fit. The ??2 values between the
high mass and low mass solutions for the 100kilosecond observations were 135
and 88, respectively. To test this further, we also fit a 100kilosecond observation
of NGC 5408 XMM1 with both of these models. This observation (0302900101)
98
was a proprietary observation at the time of publication for this chapter. The
spectral and temporal analysis are now available in Strohmayer et al. (2007). We
processed the pn data with SAS 6.5, following the same procedure as noted in the
Data Reduction section. Fitting the spectrum with both models (high mass and
low mass solution) we found a ??2 value of 170, favoring the high mass model.
Thus, we find that for sources that are well fit by either model (showing a
degenerate solution of either high mass or low mass/steep power law) the high
mass solution is the best fit when a higher count spectrum is obtained (as for
Holmberg IX XMM1, Holmberg II XMM1, and NGC 5408 X-1). Though we list
the alternate model column density and absorption values in Table 3.5, we use
the parameters listed in Table 3.2 throughout the paper (the high mass solutions).
As noted, M33 X-8 and M81 XMM1, sources with very highnumber ofcounts,
were not well-fit with a standard disk at low energy, power law at high energy
model. They were best fit with the steep low energy power law and hot disk
model shown in Figure 3.3. Along with these sources, NGC 4559 X-10 and M83
XMM1 were also well-fit by this model. We will further discuss these sources in
the following subsection.
3.4.3 Physical Plausibility of the Accretion Disk models
In Chapter 2, we had found that ULX spectra are consistent with the high/soft
and low/hard states of Galactic black holes. We had classified the sources stud-
ied in this chapter as consistent with the high/soft state. The high/soft state, as
stated earlier, is characterized by emission from an accretion disk and a Comp-
tonized power law tail. In order to investigate whether the mass and accretion
rate results from the grad model make physical sense in terms of the X-ray bi-
nary model, we present a comparison of the derived accretion rate in Eddington
99
accretion rate units versus the mass in Figure 3.4. The parameter ?MEdd was com-
puted as ?MEdd = 1.3?1038M?c2 gs?1 where ?, the efficiency factor, was set to 0.06 and
M is the value Mgrad. We find that for the sources with Mgrad > 100M?, the accre-
tion rate is computed to be below 40% of the Eddington rate. This is assuming,
as the grad model assumes, a Schwarzschild black hole. Noting that ?M/ ?MEdd is
equivalent to L/LEdd for most disk solutions, the L/LEdd values for the sources
with Mgrad > 100M? are consistent with those of Galactic black hole X-ray bi-
naries in the high state (L/LEdd ? 0.05 ? 1.0) (McClintock et al. 2004). Thus,
the sources with Mgrad > 100M? do have accretion rates that are predicted from
scaling up (in mass) observed high state Galactic black holes. In addition, the
spectral fit parameters for NGC 4631 XMM1, with an estimated mass of 5.5M?
and an accretion rate L/LEdd of 0.84, are also consistent with the standard high
state Galactic black hole model. This source is likely a normal stellar mass black
hole X-ray binary in an external galaxy.
The remaining sources with Mgrad << 100M? (M33 X-8, M81 XMM1, NGC
4559 X-10, and M83 XMM1) yielded L/LEdd ratios in the range of 1 ? 3. These
sources are also those described in the previous section where the power law
component fits the low energy spectrum. They are also well fit by a Comptoniza-
tion model and correspond to a sub-class of high luminosity ULXs described in
? 2.4.3. Due to the luminosity and modeled disk temperature (kT ? 1keV), we
suggested that these sources were very high state stellar mass black hole sys-
tems. In order to be consistent with the black hole accretion model assumed in
this study, the spectra of these sources should be the result of Comptonization
from a thermal disk spectrum.
To test this further, we fit the spectrum of the highest count source of this
type (M33 X-8) with an absorbed thermal disk and Comptonization model
100
Figure 3.4: Black hole mass versus mass accretion rate (per Eddington accretion
rate) obtained from the general relativistic disk model (grad). The sources with
model parameters indicating high masses correspond to accretion rates that are
below 40% of the Eddington accretion rate for the given mass. In Chapter 2, we
noted a temperature gap in the range of 0.26keV to 0.50keV from spectral fits
utilizinganabsorbedblackbodyandpowerlawmodel. Inthisfigure,weseethat
there is a gap in mass from the parameters of the absorbed general relativistic
disk and power law model. The values shown are results for the EPIC spectra
only, as recorded in Table 3.2 (for all sources in Table 3.2). Therefore, there are
multiple points for the sources with multiple observations. We plotted these
sources (Holmberg II XMM1, Holmberg IX XMM1, M81 XMM1, and NGC 5204
XMM1) twice to confirm that their masses, as determined by the grad model,
do not vary.
101
(tbabs*tbabs*(diskpn + comptt)). The comptt model has the parameters:
a seed temperature (keV), a plasma temperature (keV), and optical depth of the
medium. We used the diskpn model in place of the grad model since the for-
mer provides a disk temperature to which the comptt seed temperature can be
fixed. This provides a physical model, where the thermal disk supplies the en-
ergy for the Compton tail. M33 X-8 was well fit by this model with a ??2/dof =
1659.9/1534 (1.08). Thus, these sources are still consistent with the black hole ac-
cretion model. Replacing the tbabs model used for the galactic column density
with the tbvarabs, we measured the column density, oxygen abundance, and
iron abundance. With this thermal disk and Comptonization model we obtained
nH = 1.5+0.04?0.08?1021 cm?2, O/H = 1.26+0.08?0.15, and Fe/H = 2.83+0.71?0.76, with a ??2/dof
= 1577.3/1534(1.03). Within the error bars, these results are consistent with those
seen in Table 3.2.
3.5 Properties of the ISM in ULX Host Galaxies
ThemajorquestiontobeexaminedinusingULXsasprobesoftheISMiswhether
the hydrogen column density and element abundances are primarily from the
hostgalaxyorintrinsictothelocalenvironmentoftheULX.Beforewecananswer
thisquestion,itisimportanttounderstandtheintrinsicspectrumofthesource. In
the previous section we discussed the nature of the soft component in light of the
high signal-to-noise spectra of the 14 ULXsources we examined. We found that if
the spectrum is due to thermalemission froma disk, modeling thespectrum with
a variety of disk models (grad, diskbb, diskpn, bbody) does not significantly
change the measured oxygen abundance or hydrogen column density.
Assumingthereliabilityofthehydrogencolumndensityandabundancemea-
102
surements, based on their model independent values, we investigate the source
of the absorption in ULX spectra. In order to determine whether the model nH
values suggest the necessity of extra local absorption, we compare the model
values with column densities obtained from H I studies. We investigate this in
? 3.5.1. The oxygen abundances (as an indication of metallicity), which we exam-
ine in ? 3.5.2, can provide further clues of whether the absorption we see in the
X-ray spectrum is intrinsic to the source. We also examine possible connections
between the host galaxy?s star formation rate and elemental abundances.
3.5.1 Column Densities
TodeterminewhethertheULXX-rayhydrogencolumndensitiesrepresentlargely
galactic column densities or column densities local to the ULX, we compared
the X-ray values to those obtained from optical and radio studies. For a com-
parison to hydrogen column densities from optical studies, we used interstellar
reddening values. In a study of dust scattering X-ray halos surrounding point
sources and supernova remnants, Predehl & Schmitt (1995) derived a relation-
ship between hydrogen column density and interstellar reddening, using X-ray
data. TheyfoundthatnH = 5.3?1021cm?2EB?V . TheyalsofoundthattheseX-ray
derived column densities are not affected by the intrinsic absorption of the X-ray
source. Thus, the optical reddening, EB?V , becomes a useful tool in checking
our own X-ray derived column densities. Through a literature search, we found
EB?V values for the sources M33 X-8 (0.22; Long et al. (2002)) and M81 XMM1
(0.23; Kong et al. (2000)). The corresponding EB?V -derived nH values are plotted
in Figure 3.5 as triangles.
For a comparison of X-ray derived hydrogen column densities with radio val-
ues, we obtained H I column densities for four objects (Holmberg II XMM1,
103
Table 3.6. Alternative Column Density Measurements
Source nHa Methodb
NGC 247 XMM1 0.316 H I
M33 X-8 0.170 EB?V
HolmII XMM1 0.157 H I
M81 XMM1 0.122 EB?V
M81 XMM1 0.078 H I
HolmIX XMM1 0.148 H I
NGC 4559 X7 0.357 H I
NGC 4559 X10 0.300 H I
NGC 5204 XMM1 0.182 H I
aHydrogen column density (galac-
tic, not Milky Way) determined by the
methodindicatedinunitsof1022 cm?2.
bMethod used to compute the cor-
responding column densities, either
from reddening values (EB?V ) or from
radio H I measurements, see ? 3.5.1 for
details.
NGC 4559 X7, NGC 4559 X10, and NGC 5204 XMM1) from the WHISP cata-
log (Swaters et al. 2002). These are radio H I column densities within the host
galaxy (galactic), not the Galactic/Milky Way columns. Exact values of the H I
column densities were computed and given to us by Rob Swaters. Additionally,
we include H I column densities of NGC 247 XMM1, M81 XMM1, and Holmberg
IX XMM1 from Braun (1995). We obtained the FITS files of H I column density
maps from this paper (available on NED), where the pixel value corresponds to
the galactic column density in units of 1018 cm?2. The H I derived nH values from
both studies are plotted in Figure 3.5 as circles. The column densities, from the
radio and reddening studies, are listed in Table 3.6.
We find that the host galaxy column densities from alternate methods (optical
or radio studies) are not significantly different from the X-ray column densities.
104
Figure 3.5: Hydrogen column density from EB?V (red) or H I studies (blue) vs.
the hydrogen column density obtained from the tbvarabs model. H I column
densities were obtained for Holmberg II, NGC 5204, and NGC 4559 through the
WHISP survey (Swaters et al. 2002). Additional H I column densities were ob-
tained forNGC 247, HolmbergIX, andM81 throughVLA data (Braun1995). The
X-rayspectralfitcolumnsarenotbiasedtowardssignificantlyhighervaluesthan
the alternate method column densities. This implies that most of the matter in
the line of sight is from H I. This is not true, however, for the source M81 XMM1
(represented by the two outlying points in the lower right portion of the graph)
where the X-ray column density is much greater than those from the optical and
radio.
105
Particularly, the X-ray values are not skewed towards substantially higher values
than the optical/radio values. Thus, the X-ray columns are likely the galactic
values without any additional local absorption. The exception, however, is M81
XMM1 (represented by 2 points in Figure 3.5, one for each of the methods). The
EB?V value (1.22?1021 cm?2) and the H I value(0.78?1021 cm?2) are significantly
lower than the X-ray column density. This may indicate the presence of extra
absorption around this source.
The result that the X-ray hydrogen column densities are in good agreement
with those from H I studies and interstellar extinction values is interesting con-
sidering that the X-ray measured column densities are along a direct line of sight
to the ULXs while the H I measurements are an average over a larger beam area.
The agreement between the two measurements implies that the ULX sources,
with the exception of M81 XMM1, lie within roughly normal areas of their local
host galaxies (i.e. not in regions of higher column density such as a molecular
cloud). However, this only applies to the local galaxy region surrounding the
ULX and does not quantify whether the local galaxy region is different from the
galaxy at large (e.g. local star formation in a galaxy undergoing little star forma-
tion on average).
3.5.2 Elemental Abundances
Test for Oxygen Ionization Level
Before discussing implications of the determined oxygen and iron abun-
dances, we relate a test performed to determine whether we could dis-
tinguish between different ionization levels of oxygen. To do this, we
used the absorption edge model, edge, in XSPEC (using the full model:
106
tbabs*tbvarabs*edge*edge*(grad + pow)). We first checked to see that
the abundance values obtained with the edge model matched the values from
the tbvarabs model. We fixed the oxygen abundance in the tbvarabs model
at zero and added the edge model, allowing the threshold energy and absorption
depth (?) to float as free parameters (with an initial energy set to 0.543keV). We
fitthis modelto thelongest observationsofHolmberg IIXMM1 andHolmberg IX
XMM1 in addition to a source with a lower number of counts, NGC 5408 XMM1
(0112290601).
To find the oxygen column density (nO), we used the relationship that ? =
??nO, where? isthecrosssectionforphotoabsorption. Weused thecrosssection
values for neutral oxygen published in Reilman & Manson (1979) as an estimate.
This choice is supported by the results of Juett et al. (2004), who measured the
ratio of oxygen ionization states in the ISM as O II/O I? 0.1. In Table 3.7, the
hydrogen column density, threshold energy, and optical depth are listed for the
sources fit with this model. From Reilman & Manson (1979) we used the cross
section values of ? = 5.158 ? 10?19 cm2 (for E = 0.540keV) and ? = 4.789 ?
10?19 cm2 (forE = 0.570keV).AsseeninTable3.7,thehydrogencolumndensities
and the oxygen abundances obtained from this model are close to those from the
tbvarabs model. The [O/H] values, where [O/H] = 12 + log(O/H) and O and
H represent oxygen column density and hydrogen column density respectively,
between the two models vary by less than 1%.
To test whether the threshold energy from the edge model is affected by the
ionizationlevelofoxygen,wesimulatedspectraofanabsorbedpowerlawmodel
with an oxygen edge. Tim Kallmann (P.C.) provided us with an oxygen edge
model incorporating the cross sections of Garc??a et al. (2005). The model allows
for a variation of the ratio of O II/O I. Using the response and ancillary response
107
Table 3.7. Spectral Fits to the Oxygen Edge
Source nHa Eb ?c nOd [O/H]e ?2/dof countsf
HolmII XMM1 0.16+0.01?0.01 0.566+0.01?0.01 0.35+0.02?0.01 7.2 8.65/8.64 2543/2017 342874
HolmIX XMM1 0.19+0.02?0.02 0.543+0.01?0.01 0.62+0.11?0.12 12.0 8.81/8.82 2887/2751 148061
NGC 5408 XMM1 0.17+0.06?0.03 0.538+0.04?0.02 0.37+0.29?0.21 7.1 8.61/8.63 298/334 10045
aHydrogen column density determined from tbvarabs in units of 1022 cm?2. The Galactic value of
nH was fixed to the Dickey & Lockman (1990) value with the tbabs model.
bThreshold Energy obtained from the edge model in keV. Note that one edge model was used to correct
for the difference in the oxygenedge between the pn, MOS1, and MOS2 CCDs. The otheredgemodel was
used to measure the oxygen abundance from the 542eV K-shell edge.
cAbsorption depth obtained from the edge model
dColumn density of oxygen estimated from the edge model in units of 1017 cm?2
eAbundance of oxygen relative to hydrogen from the edge model versus the value quoted in Table 3.2,
[O/H] = 12 + log(O/H) where O is oxygen abundance and H is hydrogen abundance
fTotal number of photon counts from the pn and MOS detectors
matrices from the long Holmberg II ULX observation, we simulated spectra with
the XSPECfakeitcommand for a ? = 2.35power law. We simulated spectra for
O II/O I ratios of 0.0, 0.2, 0.4, 0.6, 0.8, and 1.0. The simulated spectra were binned
with a minimum of 20 counts/bin. The absorption edge component of the spec-
trum was then fit with the edge model. The fits to the threshold energies for the
simulated spectra yielded values ranging from 0.53-0.59keV, with no preference
of lower O II/O I ratios corresponding to lower threshold energies. Since the O I
absorption edge occurs at E= 0.543keV and the O II absorption edge occurs at
E= 0.57keV, we could not distinguish between these ionization states of oxygen.
Our simulations show that the oxygen edge measurements will be sensitive to
O I and O II but not to high ionization states (for instance O VIII which has an
edge energy of 0.87keV).
108
X-ray/optical [O/H] Comparison
As noted above, we tested the oxygen abundances obtained with the absorption
model tbvarabs against the abundances obtained from adding a photo-electric
absorption edge model, for three of the ULX sources. We found good agreement
(< 1% difference in [O/H] values) between both models for the X-ray spectra.
However, we found that it is not possible to distinguish between different low
ionization states of oxygen using the edge model.
We now discuss comparisons of our X-ray oxygen absorption values with
measurements in different wavelengths, based on a literature search for [O/H]
ratios. InFigure3.6wecompareour[O/H]ratioswiththoseofastudyconducted
by Pilyugin et al. (2004) (black). They provide a compilation of [O/H] ratios de-
terminedthroughspectrophotometricstudiesofH II regions. Their[O/H]values
are based on the radial distribution of oxygen abundance using the P-method.
This method uses an abundance indicator from easily measurable oxygen lines
(R23 =([O II]??3727, 3729 + [O III]??4959, 5007)/ H-?) along with an excitation
parameter. Usingthismethod, Pilyugin et al.(2004)determined[O/H]valuesfor
spiral galaxies where published spectra were available for at least 4 H II regions.
Inaddition,theyreference[O/H]valuesforirregularsobtainedthroughalternate
methods. Our [O/H] values are determined from the oxygen abundances listed
in Table 3.2. Thus, [O/H]= 12 + log(O ? 0.00049). O is the oxygen abundance
obtained from the model, which is multiplied by the Wilms relative abundance
of 4.9?10?4 of oxygen to hydrogen. We were able to compare [O/H] values for 8
sources with the P-method values (M33, NGC 253, NGC 300, M81, Holmberg II,
NGC 4559, and NGC 5408). We include the [O/H] value computed for the Holm-
berg IX ULX by Miller (1995) of 8.12. This value was computed from an optical
study of the surrounding H II region. Also, we add the [O/H] value of 8.4 for
109
Figure 3.6: Oxygen abundances from studies of H II regions vs. oxygen abun-
dances from the tbvarabs model. The [O/H] ratios are given for Holmberg II
XMM1, NGC 5408 XMM1, Holmberg IX XMM1, NGC 4559 XMM1, NGC 4559
XMM2, NGC1313 XMM3, M33 X-8, NGC 300XMM1, NGC253 XMM2, andM81
XMM1 using the P-calibration method (black). Also, [O/H] ratios are included
for M33 X-8, NGC 300 XMM1, M81 XMM1, NGC 253 XMM2, and M83 XMM1
using the R23 calibration method (green). Arrows indicate that the lower limit
for the [O/H] X-ray parameter is below the plotted graph region. Our values
are high compared to the P-method but in good agreement with the R23 method.
The sources with the largest difference between [O/H] values are those located
in irregular galaxies.
NGC 1313, determined separately by both Calzetti et al. (1994) and Walsh & Roy
(1997).
As shown in Figure 3.6, our [O/H] values are consistently high compared to
those obtained from the Pilyugin et al. (2004) H II study. Pilyugin et al. (2004) in-
clude a discussion of how their values, obtained by the P-calibration method, are
110
significantlylowerthanthoseobtainedbyGarnett(2002)usingtheR23-calibration
method. This method computes O/H at the B-band effective or half-light radius
of the disk. In Figure 3.6 we include [O/H] values for NGC 253, NGC 300, M33,
M81, and M83 from Garnett (2002) (green). Our oxygen abundances are in much
better agreement with the values from this R23-calibration method.
Our [O/H] values, which are consistent in the X-ray band between two
separate absorption models, (tbabs*tbvarabs*edge*(grad + pow) and an
tbabs*edge*edge*(grad + pow) model), are consistent with the values of
Garnett (2002). We further wish to compare them to the metallicity predicted
from Sloan Digital Sky Survey (SDSS) results. As a result of a SDSS study,
Tremonti et al. (2004) found a luminosity-metallicity relation for their sample of
star-forming galaxies of: 12+log(O/H) = ?0.185(?0.001)MB +5.238(?0.018). In
Figure 3.7 we compare our values with their results (represented by the line).
We obtained absolute magnitudes (MB) for the host galaxies using the total
apparent corrected B-magnitude recorded in the HyperLeda galaxy catalogue
(Paturel et al. 1989)(parameter btc) and the distances listed in Table 2.1. For M33
and NGC 4559, which were not included in the previous study, we used the dis-
tances of 0.7Mpc and 9.7Mpc from Ho et al. (1997). The majority of our sources
are consistent with the SDSS results. However, the sources in low luminosity
irregular galaxies have metallicities much higher than predicted.
Galaxy Properties
More luminous galaxies are sometimes expected to have higher star formation
rates, and thus higher metallicity. However, we found no evidence of this. Inves-
tigatingfurtherintotherelationshipbetweenstarformationrate(SFR)andmetal-
licity, we chose to look at a galactic luminosity diagnostic that is less dependent
111
Figure 3.7: Host galaxy MB vs. ULX X-ray [O/H] ratio. The solid line represents
the SDSS results from a study of star-forming galaxies (Tremonti et al. 2004). Ar-
rows indicate that the lower limit for the [O/H] X-ray parameter is below the
plotted graph region. Our values are largely consistent with the Sloan results,
with the exception of Holmberg IX and Holmberg II. Both of these galaxies are
irregulars. The X-ray [O/H] ratio derived from the ULX spectra is higher than
predicted for these objects.
on extinction from dust, the infrared galactic luminosity (LFIR). In Chapter 2,
we calculated LFIR using data from the Infrared Astronomical Satellite and the ap-
proach of Swartz et al. (2004). We quoted values of LFIR in Chapter 2. Using the
same method, with IRAS fluxes obtained from Ho et al. (1997), we find M33 to
have LFIR ? 1.7 ? 1042 ergs?1 and NGC 4559 to have LFIR ? 7.4 ? 1042 ergs?1.
We used these LFIR values to compare the SFR to both the oxygen and iron abun-
dances.
112
Figure 3.8: Oxygen abundance ([O/H]) as a function of the host galaxy?s FIR
luminosity, obtained from IRAS. The line represents the Wilms solar value of
[O/H] . Arrows indicate that the lower limit for the [O/H] X-ray parameter is
below the plotted graph region. There appears to be no correlation between
LFIR and oxygen abundance. In fact, the sources show roughly solar [O/H]
abundances regardless of host galaxy.
In Figure 3.8, we plot oxygen abundance relative to SFR. All sources with the
exception of Holmberg IX XMM1, which did not have available IRAS data, are
plotted. We see, as was also illustrated in Figure 3.7, that the luminosity of the
host galaxy does not determine the metallicity. The more luminous galaxies do
nothavemetallicitieshigherthanthelessluminousgalaxies. Infact,mostsources
have oxygen abundances thatare roughly the Wilms solar abundances (indicated
by the line).
In Figure 3.9 we see that there is no relationship between iron abundance (see
113
Figure 3.9: Iron abundance ([Fe/H]) as a function of the host galaxy?s FIR lumi-
nosity obtained from IRAS. Holm IX did not have corresponding IR data, how-
ever, being an irregular galaxy, its LFIR would probably be comparable to Holm
II. The ironabundancesof bothof theirregularsarewellabovethose of theULXs
in spiral galaxies. The sources in galaxies with larger LFIR have roughly equal
iron abundances, within the error bars. The iron abundances in all cases are well
above the Wilms solar abundance. Arrows indicate that the lower limit for the
[Fe/H] X-ray parameter is below the plotted graph region.
values in Tables 3.2 and 3.3 for values) and SFR. The Wilms solar abundance for
iron is only 2.69?10?5 relative the hydrogen abundance. The plots show that the
metallicity relationship is very flat, all of the sources have roughly solar abun-
dances. This is also seen in Figure 3.10. This plot shows the [Fe/H] ratios versus
the [O/H] values, both obtained through the tbvarabs model. The solar Wilms
values are represented on the plot by the open circle symbol. It appears that the
sources are slightly more abundant in iron than the solar value, however, the er-
114
ror bars are quite large. The oxygen abundances, as before stated, are roughly the
solar Wilms value. Roughly, the abundances appear solar.
With such a flat relationship between abundance and luminosity, we tested to
see if this result carried through in a comparison of abundance versus radial dis-
tance within the host galaxy. This follows upon an interesting property observed
in spiral galaxies. Namely, that abundances are typically higher in the center of
the galaxy and decrease with increasing radius (Searle 1971). We tested how our
results compare to this result in Figure 3.11. For the sources located within spiral
galaxies, we plotted the [O/H] ratio as a function of distance R from the dynam-
ical center (as reported by NED). We used the distances quoted in Table 2.1 to
translate angular distance on the sky into kpc from the galactic center. As evi-
denced in the plot, we do not see much variation in the [O/H] values. Clearly,
the expected scaling of higher oxygen abundance towards the center is not seen.
Since the sources lie in different host galaxies with different relative abundances,
it is possible that this trend may be detectable with a larger sample of galaxies.
However, the implication from Figures 3.7 through 3.11 is that the environment
of the ULX sources is relatively uniform in terms of metallicity. The ULX sources
appear to live in similar environments, with metallicities roughly solar.
3.6 Summary
Through our work, we conclude that X-ray spectral fits to ULX sources do pro-
vide a viable method of finding abundances in other galaxies. We have deter-
mined hydrogen column densities and oxygenabundances along the line of sight
to 14 ULX sources. To measure these values, we assumed a connection between
ULXs and Galactic black hole systems such that the ULX spectra used in this
115
Figure 3.10: [Fe/H] as a function of [O/H] from the tbvarabs model, using the
grad and power law fit. The Wilms solar abundance is indicated with an open
circle. The ratio of Fe/O abundances obtained through the X-ray spectral fits are
approximately the Wilms solar values. Arrows indicate that the lower limit for
the [Fe/H] X-ray parameter is below the plotted graph region.
study correspond to a high/soft state. Therefore, we modeled the sources with
an absorbed disk model and power law (in XSPEC tbabs *tbvarabs*(accretion
diskmodel+pow)). Wetestedtheeffectdifferentaccretiondiskmodelshaveonthe
measurement of the host galaxy?s hydrogen column density and elemental abun-
dance with the disk models: grad, diskpn, diskbb, and bbody. We found that
themeasured hydrogencolumndensity andabundancesaremodelindependent.
We also tested the physical plausibility of this model by comparing the mass
and mass accretion rates obtained from the grad model with expected results
based on Galactic black hole systems. We found that the ULX spectra were con-
116
Figure 3.11: Oxygen abundances from the tbvarabs model vs. distance of the
ULX from the host galaxy?s dynamicalcenter. The [O/H] ratios are given forspi-
ral galaxies only. There is no correlation between radius and oxygen abundance
in our sample. However, since the sources originate in different host galaxies
with different abundances, there is no conclusion that can be drawn from this
data. The host galaxy NGC 4559, containing two ULX sources studied, does ap-
pear to show a possible correlation (the source nearer the center has a higher
oxygen abundance), however, the error bars on the oxygen abundance are large
for both sources.
sistent with the high/soft state, with L/LEdd values < 1.0 for sources with a stan-
dard thermal model at low energies and power law dominated higher energy
spectrum. For four sources, the spectra were consistent with a heavily Comp-
tonized spectrum. These sources are more likely stellar mass black hole systems
in a very high state of accretion. We modeled their spectra with a power law
at low energy and disk model around 1keV. This model provided similar col-
117
umn density and abundance values to a more physical absorbed disk and Comp-
tonization model.
Comparing our X-ray measured column densities with those from optical and
H I studies for 8 of our sources, we find that 7 of the sources have X-ray column
densities approximately equal to those of the alternate methods. This implies
that the hydrogen columns towards most of our ULX sources represent that of
their host galaxy. Since the H I studies represent averages over a large beam area
where the X-ray column densities are directly along the line of sight to the ULX
source, this implies that the ULX sources lie within roughly normal areas of their
host galaxies, locally. However, without comparison with other lines of sight in
the galaxy, we can not determine whether the local galaxy regions are different
from the galaxy on average. The exception in this study was M81 XMM1, whose
X-ray hydrogen column density was large relative to the H I study. This suggests
that there is extra absorption intrinsic to this source.
The oxygen abundances appear to be roughly the Wilms solar values. For
five sources, the count rates were sufficient to determine iron abundances with-
out large error bars (see Figures 3.9 and 3.10). We found that iron abundances for
these sources were slightly overabundantrelative to the solar Wilms value. How-
ever, within error bars, the abundances appear solar. X-ray derived [O/H] values
are comparable to those from an optical study by (Garnett 2002), indicating that
the X-ray derived values are the same as the [O/H] values of H II regions within
the host galaxy. Luminosity-metallicity relationships for the ULX host galaxies
show a flat distribution, as does a radius-metallicity plot. Therefore, it appears
that the ULX sources exist in similar environments within their host galaxy, de-
spite the wide range of host galaxy properties.
118
Chapter 4
A Representative Sub-Sample of the SWIFT
BAT AGN Survey
4.1 Introduction
Breaking with the focus of the previous two chapters, ULXs, the following two
chapters discuss X-ray surveys of the SWIFT Burst Alert Telescope AGN survey.
AGN surveys are typically dominated by two selection effects: (1) dilution by
starlight from the host galaxy and (2) obscuration by dust and gas in the host
galaxy and/or the AGN itself (see Hewett & Foltz (1994); Mushotzky (2004)).
These factors previously kept an unbiased AGN sample from reach. However,
with the capabilities of SWIFT?s Burst Alert Telescope (BAT), this has changed.
The BAT surveys the sky in the hard X-ray range of 14-195keV. As of early 2006,
BAT detected 153 AGN with a median redshift of 0.03 (Markwardt et al. 2005;
Tueller et al. 2008). These AGN were selected purely by their hard X-ray flux,
and thus, all but the most heavily absorbed sources (nH > 3 ? 1024 cm?2) are
not affected by obscuration from gas and dust, which prevents them from being
easily detected in UV, optical, or soft band X-ray surveys. Thus, the BAT survey
119
will provide the information necessary to derive the true distribution of AGN
characteristics across the entire electromagnetic spectrum.
The 9-month BAT detection limit is ? 2 ? 10?11 ergs?1 cm?2. Thus, the BAT
AGNs are more powerful sources than the ULXs, with X-ray luminosities over
the full range of AGN luminosities. For the closest Seyfert galaxies (z ? 0.01), the
BATfluxlimitcorrespondstoa14?195keVluminosityofroughly3?1042 ergs?1.
Most of the sources (? 97%) have been optically detected as relatively bright
nearby objects with< z >? 0.03, while most of the sources without brightoptical
counterparts are blazars, of which there are 17 in the sample (Tueller et al. 2008).
Since the BAT sources are mostly bright and nearby, they are good sources with
which to study multi-wavelength properties.
Before delving into the survey and its results, it is important to understand
the importance of the BAT survey. Therefore, in ? 4.1.1, we elaborate on the se-
lection effects in various wavelength bands. We then discuss other very hard
X-ray surveys (HEAO-1 and Integral) and BAT?s advantages in ? 4.1.2. Finally, in
? 4.1.3, we outline the study presented in this chapter ? the X-ray properties of 22
previously unobserved (in X-rays) BAT AGN.
4.1.1 Selection Effects
As mentioned, selection effects are a common hazard in AGN surveys. The first
effect, dilution by starlight from the host galaxy, is most important in the opti-
cal/UV and infrared bands ? where stars output most of their energy. This di-
lution can even be present in the soft X-ray (< 2keV). Here, the contributions
of stars, particularly hot, young stars, supernova remnants, and ionized gas in
star forming regions, can dominate over comparatively weaker AGN emission.
In the optical/UV through IR, this becomes increasingly more important for a
120
galactic stellar luminosity of L ? 1044 ergs?1. In the soft X-rays, luminosities up
to L0.5?2 ? 1041 ergs?1 from stellar processes (Ranalli et al. 2003) can rival the
AGN emission in a low luminosity source. Observationally, this has the effect of
reducing the equivalent width of AGN emission lines and masking the contribu-
tion of the AGN to the measured flux/color (Mushotzky 2004).
The second important selection effect is the result of obscuration along our
line of sight to the AGN. Particularly for sources located in the Galactic plane,
dust and gas in our own Galaxy provides obscuration. Above the Galactic plane,
this obscuration largely comes from dust and gas in the AGN?s host and/or more
local to the AGN. From AGN models of the Cosmic X-ray background (CXB),
Fabian & Iwasawa (1999) found that the observed CXB can be replicated if we as-
sume that 85% of the direct X-ray emission from AGN is obscured. This suggests
that obscuration is an important effect, present in the majority of AGN.
In X-ray astronomy, a frequently used definition is that obscured sources are
those with nH ? 1022 atomscm?2. As the column density goes to higher values,
the photoelectric cut-off from absorption shifts to higher energies (see Figure 1.1).
Therefore, the direct emission from the AGN is cut-off and the corresponding
soft flux (< 2keV) is low. Thus, soft X-ray surveys are extremely sensitive to
the effects of obscuring material. Even above 2keV, the effect of obscuration can
make it difficult to impossible to select an unbiased sample of highly obscured
sources. In particular, AGN with columns nH ? 1.5 ? 1024 atomscm?2, become
extremely difficult to find. Such sources, termed Compton thick, have an optical
depth ? > 1 towards inverse Compton scattering. Therefore, the only emission
that reaches us is scattered or reflected emission, estimated to be ? 50?100 times
weaker than the direct emission (Comastri 2004).
Not only are Compton-thick sources faint, due to the lack of direct emission,
121
but a further complication arises in that obscured AGN are less luminous than
unobscured AGN. We already mentioned this in ? 1.4, as a challenge to the uni-
fied AGN model. In short, there are few luminous absorbed AGN in the 2?10keV
band (Steffen et al. 2003) and a larger fraction of absorbed AGN at lower lumi-
nosities (Ueda et al. 2003). Since all X-ray surveys are flux limited, this makes it
even more difficult to obtain an unbiased sample at a given redshift ? the true
ratio of absorbed to unabsorbed sources could be skewed by missing the lower
luminosity absorbed AGN.
Of course, the effect of obscuration is not limited to the X-ray bands. In fact, it
has a much stronger affect in the UV, optical, and near-IR. One evidence of this is
the fact that there are X-ray selected AGN showing none of the optical properties
ofAGN.Thiscanbebecauseofdilutionorobscuration. Inthecaseofobscuration,
the source is obscured to such a degree that none of the AGN emission lines can
be seen. As Mushotzky (2004) points out, columns above 1022 cm?2 atomscm?2
correspond to extinction of AV > 5, high enough that the UV/near-IR flux from
the AGN would not be seen.
Other effects that we havenotyet mentioned include both those from separat-
ing out star burst/AGN contributions and aperture size. In the optical/UV, line
diagnostics are effective at distinguishing between H II regions and AGN emis-
sion(seeVeilleux(2002)forareview). However, intheIR,itisdifficulttoseparate
out the AGN and stellar components without very high signal to noise spectra
(see Mel?endez et al. (2008)). As for aperture size, Veilleux et al. (1995) show the
significant effects aperture sizes have in the IR. Such effects are also important
in other bands (i.e. optical) and are somewhat related to the dilution effect. For
instance, a small aperture may include mostly light from the bulge region where
a large aperture may include the entire galaxy. This effect is also sensitive to the
122
redshift of the target, since redshift affects the angular size of the host galaxy.
In the radio band, a different kind of selection effect is at work. Namely, only
10%ofAGNareradio-loud,i.e. haveluminousradioemission(White et al.2000).
Thus, while finding AGN in the radio is easy, based on the presence of radio
jets or lobes, the majority of AGN are radio-quiet. Therefore, to examine the
radio properties of an unbiased sample, the sample must first be identified in an
alternate band and then followed up with radio observations.
Therefore, assuming that very hard X-rays (> 10keV) are present in all AGN,
they provide one of the best means to select an ?unbiased? sample. At these high
energies, there is no detectable emission from normal stars. Also, obscuration
plays a minimal effect for columns of nH ? 1024 atomscm?2. Further, AGN emis-
sion in the hard X-ray band is closely tied to the accretion processes, originating
from a compact region very close to the black hole. While, most other AGN sig-
natures detected in other bands are from indirect or reprocessed emission.
4.1.2 Other Hard X-ray Surveys
Beginning in 1977, NASA launched a series of High Energy Astronomy Ob-
servatories (HEAO). HEAO-1 was the first of these, surveying the sky ? 3
times in the 0.2keV?10MeV band over its 1.5 year lifetime. Among its achieve-
ments, HEAO-1 provided a measurement of the 3?50keV X-ray background
(Marshall et al. 1980) and a complete flux limited survey of high Galactic lati-
tude sources (Piccinotti et al. 1982). The HEAO-1 X-ray survey covered 8.2sr at
a sensitivity of ? 3.1? 10?11 ergs?1 cm?2 in the 2?10keV band. In the survey, 85
sources were detected, including ? 30 AGN (9 sources were unidentified).
More recently, Chandra and XMM-Newton have conducted deep X-ray sur-
veys in the 2?10keV band for relatively small regions of the sky. Unlike HEAO-1,
123
however, many of the detected sources are at much higher redshifts (z ? 1) than
theBATAGN.AgoodreviewofthesesurveyscanbefoundinBrandt & Hasinger
(2005). Based on redshift, the z ? 1 sources provide a good comparison with the
BAT AGN survey, particularly to search for evolutionary differences between the
two populations.
Yet another contemporary mission is Integral. The International Gamma-ray
Astrophysics Laboratory (INTEGRAL) is a European Space Agency led satel-
lite launched in October 2002. The main goals are spectroscopy, at a resolution
of E/?E = 500, and imaging, with angular resolution of 12?, in the 15keV to
10MeV range. The satellite consists of four science instruments: the spectrometer
(SPI), imager (IBIS), an X-ray detector sensitive in the 3?25keV band ? Joint Euro-
pean X-ray Monitor (JEM-X), and an optical monitoring camera (OMC) with a V
band filter (500?850nm). The Key Project of INTEGRAL is a study of gamma-ray
sources located in the Galactic plane. Therefore, the sky-coverage is particularly
good in this region and much poorer in other regions.
An INTEGRAL all-sky survey of AGN has now been completed in the
17?60keV band (Sazonov et al. 2007) to a limiting flux of F17?60 ? 1.4 ?
10?11 ergs?1 cm?2. There are a few clear advantages the BAT survey can claim.
For one, BAT is truly all-sky with a much higher sensitivity over a wider range
of the sky. Due to its larger field of view, BAT detects many more AGN than IN-
TEGRAL. After a 4 year survey, INTEGRAL detected 127 AGN. In the 9-month
survey, BAT has already detected more sources (153). Further, the BAT survey
continues to detect more sources with the 22-month survey adding an additional
? 100 AGN (Tueller et al., in prep). However, INTEGRAL clearly has the advan-
tage in the Galactic plane. A second advantage of BAT is that it covers a larger
and higher energy range (14?195keV) than INTEGRAL. Thus, it can potentially
124
detect more absorbed sources.
4.1.3 This Chapter
Despite the X-ray brightness of the BAT AGN, a number of these sources had yet
tobeobservedspectroscopicallyintheX-rays. Inthischapter, wepresentananal-
ysis of XMM-Newton EPIC spectra of 23 sources, 2 of which correspond to a pair
ofinteracting galaxieswhere theBATsource isconfused, i.e. likely acombination
of the AGN, for which spectra were obtained through discretionary project scien-
tist time. These sources had no previous X-ray spectrum, were clearly detected
by the BAT, and had clear optical counterparts from Digital Sky Survey (DSS) im-
ages. Many of these sources now also have SWIFT X-ray Telescope (XRT) spectra
available. Thus, we are able to compare the XRT spectra with the EPIC spec-
tra. In ? 4.2, we detail the observations and data analysis for the XMM-Newton
and SWIFT observations. In ? 4.3, we describe spectral fits to the X-ray spectra,
including a search for variability between the XRT and XMM-Newton observa-
tions. We discuss our results in ? 4.4. Finally, we summarize our results in ? 4.5.
Through this chapter, we make the claim that the properties of this sample are
representative ofthelarger9-monthAGN catalog. Thus, theresults become more
interesting as a first look at the BAT AGN properties, which is then expanded on
with the entire sample in Chapter 5.
4.2 Observations and Data Analysis
4.2.1 XMM-Newton and SWIFT XRT Spectra
We obtained proprietary XMM-Newton observations of 22 out of 26 proposed
BAT AGN sources through discretionary project scientist time. These particu-
125
lar BAT AGNs were selected based on: their detection in BAT with high signifi-
cance (? > 5), a clear optical counterpart to the BAT source in Digital Sky Survey
images, and no previous X-ray spectrum. A list of these sources, whose X-ray
spectra were observed for the first time, is in Table 4.1. Here we list the source
name, co-ordinates, redshift, Galactic column density towards the source, AGN
type, and host galaxy type. For one of the BAT sources, the ?BAT source? is an
interacting galaxy system with two AGNs, MCG +04-48-002 and NGC 6921. In
additiontotheXMM-Newtonobservations,wedownloadedarchivedSWIFTXRT
observations (48) for the 22 BAT AGNs from the NASA HEASARC archives. In
Table 4.2 we include details on the observations examined.
We reduced the XMM-Newton data using the Science Analysis System (SAS)
version 7.0. We created calibrated photon event files for the EPIC-MOS and pn
cameras using the observation data files (ODF) with the commands emchain
and epchain. Following this, the events tables were filtered using the standard
criteria outlined in the XMM-Newton ABC Guide1. For the MOS data (both MOS1
andMOS2cameras), goodeventsconstitutethosewithapulseheightintherange
of 0.2 to 12 keV and event patterns that are characterized as 0-12 (single, double,
triple,andquadruplepixelevents). Forthepncamera,onlypatternsof0-4(single
and double pixel events) are kept, with the energy range for the pulse height set
between 0.2 and 15 keV. Bad pixels and events too close to the edges of the CCD
chips were rejected using the stringent selection expression ?FLAG == 0?.
Light curves of the observations were produced with xmmselect and exam-
ined for flaring events (distinguished by high count rates). Time filtering was
required only for the sources SWIFT J0641.3+3257, SWIFT J0911.2+4533, MCG
+04-22-042, MRK 417, WKK 1263, and NGC 6921/MCG +04-48-002 (the interact-
1http://heasarc.gsfc.nasa.gov/docs/xmm/abc/
126
Table 4.1. Details on the Sub-sample of 22 Sources
Source RA (h m s) Dec (????) Redshift nH(GAL)1 Type2 Host Galaxy2
MRK 352 00 59 53.28 31 49 36.87 0.014864 5.59 Sy1 SA0
NGC 612 01 33 57.74 -36 29 35.70 0.029771 1.85 Gal SA0+ pec
SWIFT J0216.3+5128 02 16 29.84 51 26 24.70 ? 1.57 ? ?
NGC 1142 02 55 12.196 -00 11 00.81 0.028847 6.00 Sy2 S pec
SWIFT J0318.7+6828 03 18 18.98 68 29 31.42 0.090100 35.1 Sy2 S?
ESO 548-G081 03 42 03.72 -21 14 39.70 0.014480 3.04 Sy1 Sba
ESO 362-G018 05 19 35.82 -32 39 27.90 0.01264 1.78 Sy1.5 S0/a
ESO 490-G026 06 40 11.69 -25 53 43.30 0.024800 11.7 Sy1.2 Pec
SWIFT J0641.3+3257 06 41 23.04 32 55 38.60 0.017195 11.6 Sy2 E?
MRK 18 09 01 58.39 60 09 06.20 0.011088 4.49 Sy2 S
SWIFT J0904.3+5538 09 04 32.94 55 38 30.63 0.037142 2.78 Sy1 ?
SWIFT J0911.2+4533 09 11 29.97 45 28 05.00 0.026782 1.64 Sy2 S?
MCG +04-22-042 09 23 43.00 22 54 32.50 0.032349 3.37 Sy1.2 E
MRK 417 10 49 30.93 22 57 51.90 0.032756 2.06 Sy2 Sa
UGC 6728 11 45 16.02 79 40 53.42 0.015300 4.49 Sy1.2 SB0/a
SWIFT J1200.8+0650 12 00 57.92 06 48 23.11 0.036045 1.43 Sy2 S?
ESO 506-027 12 38 54.59 -27 18 28.20 0.025024 6.60 Sy2 S pec s
WKK 1263 12 41 25.74 -57 50 03.50 0.024430 35.5 Sy23 Sc
MCG +09-21-096 13 03 59.47 53 47 30.10 0.02988 1.53 Sy1 SABb
NGC 4992 13 09 12.95 11 38 45.32 0.025137 2.09 Gal Sa
NGC 6860 20 08 46.89 -61 06 00.70 0.014884 4.19 Sy1.54 SB ab
NGC 6921 20 28 28.86 25 43 24.30 0.014287 26.0 Sy2 SA 0/a
MCG+04-48-002 20 28 35.0 25 44 00.0 0.013900 26.0 Sy2 S
1Galactic column density towards the source, in units of 1020 cm?2, as obtained from the web version
of the nH FTOOL. These are the values from Dickey & Lockman (1990).
2AGN type and host galaxy type from Tueller et al. (2008). For AGN types, optical identifications are
listed, where available. Where ?Gal? is indicated, there are no optical emission lines indicative of the
presence of an AGN. The optical spectrum looks like a galaxy spectrum. Additional host galaxy classifi-
cationswereobtained fromthe LEDA database. Where??? isindicated, there isno available classification.
3While WKK 1263 is classified as a Sy2 in NED, we could find no optical spectrum to confirm this.
4NGC 6860 is classified as a Sy1.5 by Lipari et al. (1993) contrary to NED?s classification as a Sy1.
ingsystem). Thelightcurveswerefilteredusingthetasktabgtigen,asoutlined
in the SAS ABC guide, with the parameters ?RATE< 50? counts/s for the pn and
?RATE< 10? counts/s for MOS observations.
Spectra of the sources were extracted using the SAS task especget. This
routine extracts source and background spectra from a defined region as well as
response and ancillary response matrices. We extracted the source spectra from
circular regions with radii between ? 30?125??. The extraction radius depended
upon the location of the source with respect to other sources and the edge of a
127
CCDchip. Backgroundregionswereextractedfromannularregionssurrounding
the source, where possible. In cases were the source was near the edge of a chip
or another source, circular regions near the source were used, of the same size as
the source region and on the same chip.
Of the 22 target observations, there were a few cases where the position of the
source fell within a gap in the CCD chip (for either the pn or a MOS detector).
This occurred for the MOS1 observation of SWIFT J0641.3+3257 and the pn ob-
servation of SWIFT J0911.2+4533. Thus, there are no corresponding spectra from
these CCDs. Also, the position of source SWIFT J0904.3+5538 was located such
that a large fraction of the light (66%) was located in a gap in the pn CCD. Due
to a problem with the ODF files for NGC 6860, the pn observation of this source
could not be extracted.
Table 4.2: XMM-Newton EPIC and SWIFT XRT Observation Information
Source Telescope Observation ID Start Date Exposure Time (s)1 Total Counts1
MRK 352 XMM 0312190101 2006-01-24 9773, 12382, 12360 71296, 23572, 23175
MRK 352 XRT 00035243001 2006-05-29 11527 3456
MRK 352 XRT 00035243002 2006-05-30 19385 3983
NGC 612 XMM 0312190201 2006-06-26 9744, 12456, 12457 1164, 285, 352
NGC 612 XRT 00035627001 2006-06-02 6743 20
NGC 612 XRT 00035627002 2006-06-12 4875 20
SWIFT J0216.3+5128 XMM 0312190301 2006-01-24 8921, 11485 ,11495 14412, 6454, 6053
SWIFT J0216.3+5128 XRT 00035247001 2006-01-08 8291 1516
SWIFT J0216.3+5128 XRT 00035247002 2006-01-10 5559 982
SWIFT J0216.3+5128 XRT 00035247003 2006-02-05 2113 288
NGC 1142 XMM 0312190401 2006-01-28 8921, 11485, 11496 2481, 847, 907
NGC 1142 XRT 00035248001 2006-06-29 6808 245
NGC 1142 XRT 00035248002 2006-07-07 5434 201
NGC 1142 XRT 00035248003 2006-07-08 4670 145
NGC 1142 XRT 00035248004 2006-07-11 1888 40
SWIFT J0318.7+6828 XMM 0312190501 2006-01-29 6578, 11163, 11170 9654, 4512, 4491
Continued on next page ...
128
Table 4.2 ? Continued
Source Telescope Observation ID Start Date Exposure Time (s)1 Total Counts1
SWIFT J0318.7+6828 XRT 00035249001 2006-03-29 9247 413
SWIFT J0318.7+6828 XRT 00035249002 2006-04-05 8061 385
ESO 548-G081 XMM 0312190601 2006-01-28 8924,11485, 11499 106660, 32699, 32404
ESO 548-G081 XRT 00035250001 2006-01-28 3561 3632
ESO 548-G081 XRT 00035250002 2006-03-19 6605 7821
ESO 362-G018 XMM 0312190701 2006-01-28 8921, 11483, 11498 8497, 2905, 2884
ESO 362-G018 XRT 00035234001 2005-10-29 1379 772
ESO 362-G018 XRT 00035234002 2005-11-26 6979 2587
ESO 490-G026 XMM 0312190801 2006-03-07 9192, 11812, 11823 52246, 21057, 19713
ESO 490-G026 XRT 00035256001 2005-12-16 8246 4448
ESO 490-G026 XRT 00035256002 2006-03-23 2809 1434
ESO 490-G026 XRT 00035256003 2006-03-28 2747 954
SWIFT J0641.3+3257 XMM 0312190901 2006-03-11 10696, -, 13507 2686, -, 786
SWIFT J0641.3+3257 XRT 00035257002 2005-12-18 7784 20
SWIFT J0641.3+3257 XRT 00035257003 2005-12-26 14864 82
SWIFT J0641.3+3257 XRT 00035257006 2006-01-07 20008 60
MRK 18 XMM 0312191001 2006-03-23 9910, 13387, 13402 4990, 1089, 1029
MRK 18 XRT 00035259001 2005-12-18 5353 50
MRK 18 XRT 00035259002 2005-12-26 3088 20
SWIFT J0904.3+5538 XMM 0312191101 2006-03-31 7142, 12072 , 12089 10830, 10980, 11221
SWIFT J0904.3+5538 XRT 00035260001 2005-12-15 5706 916
SWIFT J0904.3+5538 XRT 00035260002 2006-01-06 6211 736
SWIFT J0911.2+4533 XMM 0312191201 2006-04-10 -, 11530, 11531 -, 615, 515
SWIFT J0911.2+4533 XRT 00035261001 2006-01-04 5487 21
SWIFT J0911.2+4533 XRT 00035261002 2006-01-22 8875 80
MCG +04-22-042 XMM 0312191401 2006-04-18 9012, 11809, 11824 126070, 39777 39220
MCG +04-22-042 XRT 00035263001 2005-12-10 8564 4986
MRK 417 XMM 0312191501 2006-06-15 7437, 348, 351 1075, 348, 351
MRK 417 XRT 00035264001 2005-12-12 6306 41
MRK 417 XRT 00035264002 2006-03-03 3534 20
MRK 417 XRT 00035264003 2006-06-24 16130 140
UGC 6728 XMM 0312191601 2006-02-23 7220, 11404, 11415 30705, 11420, 11109
UGC 6728 XRT 00035266001 2006-06-24 6331 3090
UGC 6728 XRT 00035266002 2006-06-29 3017 994
UGC 6728 XRT 00035266003 2006-07-07 1433 463
Continued on next page ...
129
Table 4.2 ? Continued
Source Telescope Observation ID Start Date Exposure Time (s)1 Total Counts1
SWIFT J1200.8+0650 XMM 0312191701 2006-06-26 9777, 12495, 12507 4543, 1590, 1751
SWIFT J1200.8+0650 XRT 00035267001 2005-12-11 14961 638
SWIFT J1200.8+0650 XRT 00035267002 2005-12-21 3156 120
ESO 506-G027 XMM 0312191801 2006-01-24 8162, 11139, 11150 2114, 630, 645
ESO 506-G027 XRT 00035273002 2005-06-15 3209 60
ESO 506-G027 XRT 00035273003 2005-08-15 1938 20
ESO 506-G027 XRT 00035273004 2005-08-28 10808 242
WKK 1263 XMM 0312191901 2006-02-01 8902, 11482, 11495 24195, 10198, 10120
WKK 1263 XRT 00035268001 2005-12-15 3253 564
WKK 1263 XRT 00035268002 2005-12-29 8690 1346
SWIFT J1303.8+5345 XMM 0312192001 2006-06-23 8408, 11470, 11490 86597, 27942, 27492
SWIFT J1303.8+5345 XRT 00035269001 2005-12-19 7468 7723
SWIFT J1303.8+5345 XRT 00035269004 2006-07-02 4752 3986
NGC 4992 XMM 0312192101 2006-06-27 12849, -, - 1756, -, -
NGC 6860 XMM 0312192201 2006-03-19 -, 11815, 11823 -, 5485, 4910
NGC 6860 XRT 00035275001 2005-12-12 3536 632
NGC 6921 XMM 0312192301 2006-04-23 8789, 11235, 11249 2243, 554, 513
NGC 6921 XRT 00035276001 2005-12-16 4071 41
NGC 6921 XRT 00035276002 2006-03-23 4600 40
NGC 6921 XRT 0003072201 2006-06-03 6885 65
MCG +04-48-002 XMM 0312192301 2006-04-23 8789, 11235, 11249 845, 180, 163
MCG +04-48-002 XRT 00035276001 2005-12-16 4071 41
MCG +04-48-002 XRT 00035276002 2006-03-23 4600 40
MCG +04-48-002 XRT 0003072201 2006-06-03 6885 65
1FortheXMMobservations,theexposuretimesandtotalcountsarelisted fortheEPICpn, MOS1,
and MOS2 respectively.
For the XRT data, we extracted spectra of target sources from the cleaned pho-
ton counting event files downloaded from the public High Energy Astrophysics
Science Archive Research Center (HEASARC) archive. We extracted spectra for
130
observationsthathadXRTexposuretimes ofatleast1000s, followingthe instruc-
tions outlined in The SWIFT XRT Data Reduction Guide2. Spectra were extracted
for the sources using the FTOOL XSELECT. With this tool, a source region was
created in DS9, with a radius of ? 50 ? 70??. A background region was created
in a source free region close to the source, with a radius of 95??. We used the
standard response and ancillary response files available for the photon counting
event files with grades 0 to 12. These are publicly available in the SWIFT XRT
calibration database (CALDB).
For all of the spectra, we binned the source and background spectrum and re-
sponse files for each observation with the Ftool grppha. All spectra were binned
with 20counts/bin. We then fit the spectra in the 0.3-10keV range with XSPEC
version 11. The details of these fits are described in the following section.
4.2.2 SWIFT BAT Spectra
BAT is a wide field (? 2 steradians) coded aperture hard X-ray instrument. Dur-
ing normal operations, it usually covers ? 60% of the sky each day at < 20 mil-
liCrab sensitivity. The BAT spectra were derived from an independent all sky
mosaic map in each energy bin, averaged over 22 months of data, beginning on
Dec 5 2004 (Tueller in preparation). The survey was processed using the BAT
Ftools and additional software to make mosaic maps. The energy bin edges are
14, 20, 24, 35, 50, 75, 100, 150, 195 keV. The energies are calibrated in-flight for
each detector using an on-board electronic pulser and the 59.5 keV gamma and
La L and M lines from a tagged 241Am source. The average count rate in the
map bin that corresponds to the known position of the counterpart was used.
Due to the strong correlation of the signal in adjacent map bins of the over-
2http://heasarc.nasa.gov/docs/swift/analysis/xrt_swguide_v1_2.pdf
131
sampled coded aperture image, it is not necessary to perform a fit to the PSF.
Each rate was normalized to the Crab nebula rate using an assumed spectrum
of 10.4 ? E?2.15 phcm?2 s?1 keV?1 for the BAT energy range. Due to the large
number of different pointings that contribute to any position in the map, this is
a good approximation of the response. This has been verified by fitting sources
with known spectra (Cas-A compared to BeppoSax and Integral, PSR1509-589
compared to Integral, NGC2110 and Cen-A compared to Suzaku) and generally
produces a good connection to X-ray spectra in sources that do not vary much
with time. Error estimates were derived directly from the mosaic images using
the RMS image noise in a region around the source of roughly 3 degrees in ra-
dius. This is the optimum procedure due to the residual systematic errors of 1.2
to 1.8 times statistical values in the current BAT mosaics. Analysis of the noise in
the images suggests that the variations in noise are small on this scale. Analysis
of negative fluctuations shows that the noise is very will fit by a Gaussian distri-
bution and that this normalization is very accurate on average. All fitting of the
BATdatawasperformedonthisnormalizeddatausing adiagonalinstrument re-
sponse matrix. This procedure correctly accounts for instrumental systematics in
sources with spectral indices similar to the Crab. While there may be significant
systematic errors for sources with spectra that are much flatter than the Crab, this
is not a significant problem for most of the sources presented in this chapter (the
exceptions being NGC 612 and MRK 417).
132
4.3 Spectral Fitting
In examining the X-ray spectra of these BAT AGNs, there are two main goals:
to determine how the sources vary between observations and to determine the
spectral properties of the source (the hydrogen column density, spectral shape,
and properties of the Fe K line and/or other lines if present). Since the extracted
XMM-Newton spectra have, on average, ten times the number of counts from
the XRT spectra, we will focus on the XMM-Newton spectra for a more detailed
analysis. The effective spectral resolution for the XRT is lower, since there are
fewer counts and at 20cts per bin there are few data points, such that Fe K lines
which are clearly visible in 10ks XMM-Newton observations are not resolved in
comparable XRT observations (see Figure 4.1). Thus, in order to examine the
variability between the XRT and XMM-Newton observations we need to rely on
simple models for the AGN spectra.
As a first fit to the spectra and as a means to compare variability between
XRT and XMM-Newton observations, we fit each observation separately with a
simple absorbed power law (absorption model * pow). In XSPEC, we used the
absorption model tbabs to account for cold absorption in the Milky Way. Thus,
we fixed the tbabsmodel hydrogen absorption to the Dickey & Lockman (1990)
value (see Table 4.1). We allowed the parameter of a second tbabsmodel to float
in order to account for the combination of absorption from the AGN host galaxy
and local environment.
To apply this model to the XMM-Newton EPIC observations, we simultane-
ously fit the pn and MOS observations, using a constant value to account for
differences in flux calibration. The best-fit spectral parameters for these obser-
vations are recorded in Table 4.3, including the observed flux values at soft (0.3-
133
0.01
0.1
normalized counts/sec/keV
50.5
1
1.5
2
2.5
ratio
channel energy (keV)
Figure 4.1: XMM-Newton pn and the highest quality XRT spectrum for NGC
1142 centered on the 6.4keV fluorescent Fe K line. The lines represent a simple
absorbed power law model. In the ? 10ks pn, the Fe K line is clearly distin-
guishable requiring the addition of a Gaussian component. However, the ? 7ks
XRT spectrum, binned as the pn spectrum with 20 photons/bin, does not have
the spectral resolution required to distinguish this feature. We found this to be
the case with all of the XRT spectra examined for these 22 BAT AGNs.
134
2keV) and hard (2-10keV) energies. We include the best-fit spectral parameters
for the XRT observations in Table 4.4. All quoted errors represent the 90% confi-
dence level. Weexcluded fromour spectralfits anyobservationwith? 50counts.
Also, we note that this simple power law model is a very poor fit to the XMM-
Newton spectra of NGC 612, NGC 1142, ESO 362-G018, MRK 417, ESO 506-G027,
and MCG +04-48-002. For each of these five sources the reduced ?2 value indi-
cates that the simple power law fit was not a good description of the data. Thus,
no errors were calculated for the spectral parameters. We will discuss the spectra
of these sources and more acceptable models below.
4.3.1 Detailed Spectral Properties
From our initial examination of the spectra, we found that simple power law
models (see Table 4.3) were not sufficient in describing the spectra of all of our
sources. In particular, there are three main features that were not accounted for
by a simple absorbed power law: a soft excess, line emission, and additional
complexity from emission with varying column densities. A ?soft excess? in
an AGN may be the result of thermal emission (for instance, from star forma-
tion), low temperature Comptonization, blurred reflection (Czerny et al. 2003;
Ross & Fabian 2005), or blurred absorption (Gierli?nski & Done 2004). Line emis-
sion can be produced by a number of mechanisms, in particular from fluorescent,
photo-ionized or collisionally ionized gas. Most importantly for AGN, is fluores-
cence from lowly ionized iron, with the strongest feature being the Fe K line at
6.4keV. The final type of ?feature?, a complex spectrum with different column
densities, is difficult to interpret. Such spectra, which appear to have a ?double
power law? shape, could be the result of contamination of the AGN light by a
less absorbed X-ray binary spectrum, a dusty environment where the AGN emis-
135
sion is partially covered by absorbing material, or scattering of some of the AGN
light.
In order to better characterize our sources, we first grouped our sources in
basic categories: (1) pure power law sources (with or without absorption beyond
the Galactic Milky Way values), (2) power law sources with a soft excess (with or
without absorption), and (3) more complex ?double power law? shaped spectra.
Based on the F-test, any additionalcomponents added to our spectral models im-
proved the fit by a ?2 of at least 8 (this is the 99% level for 2 additional degrees
of freedom). The only exception is the addition of a gaussian line (zgauss in
XSPEC) to characterize the Fe K-? line at 6.4keV. Where the addition of the line
was not significant, we derived upper limits on the strength and intensity (indi-
cated by the normalization which is the integrated photon flux of the line) of a
possible emission line. Therefore, for all of our sources, we fixed the energy of a
gaussian line to 6.4keV in the source?s rest frame with a fixed width (FWHM) of
0.01keV, corresponding to an unresolved line.
136
Table 4.3. XMM-Newton Spectral Analysis: Absorbed Power Law Fits
Source Observation ID nH1 ? ?2/dof Fsoft2 Fhard2
MRK 352 0312190101 ? 1.95+0.01?0.01 3104.4/1569 6.87 9.76
NGC 612 3 0312190201 79.90 0.31 274.2/83 0.00 1.66
SWIFT J0216.3+5128 0312190301 1.74+0.06?0.07 1.77+0.04?0.04 1069.7/985 0.90 8.44
NGC 1142 3 0312190401 65.68 1.63 942.8/197 0.00 3.00
SWIFT J0318.7+6828 0312190501 3.20+0.31?0.34 1.36+0.10?0.11 812.3/772 0.23 7.43
ESO 548-G081 0312190601 ? 2.03+0.01?0.01 2943.5/1642 11.34 13.48
ESO 362-G018 3 0312190701 ? 1.47 2534.6/549 0.69 1.83
ESO 490-G026 0312190801 0.24+0.01?0.01 1.59+0.02?0.02 1634.2/1757 5.30 19.79
SWIFT J0641.3+3257 0312190901 12.11+3.08?1.79 0.98+0.30?0.20 218.4/154 0.00 3.38
MRK 18 0312191001 13.06+7.20?3.52 1.26+0.72?0.38 413.4/327 0.00 1 .60
SWIFT J0904.3+5538 0312191101 ? 1.79+0.01?0.02 1126.3/865 3.19 5.41
SWIFT J0911.2+4533 0312191201 27.63+27.50?9.43 1.34+2.21?0.43 19.9/17 0.00 1.55
MCG +04-22-042 0312191401 ? 2.00+0.01?0.00 1835.8/1819 12.45 15.52
MRK 417 3 0312191501 41.76 0.56 565.6/82 0.00 1.52
UGC 6728 0312191601 0.01+0.00?0.01 1.78+0.02?0.02 1207.0/1160 3.59 6.49
SWIFT J1200.8+0650 0312191701 9.31+0.72?0.67 1.30+0.11?0.10 484.0/357 0.02 5.17
ESO 506-G027 3 0312191801 75.91 1.14 543.4/172 0.00 3.73
WKK 1263 0312191901 0.06+0.01?0.02 1.53+0.03?0.02 1257.3/1248 2.32 9.99
MCG +09-21-096 0312192001 ? 1.76+0.01?0.00 1291.2/1626 9.27 15.85
NGC 4992 0312192101 76.20+16.36?8.73 1.69+0.58?0.30 139.6/82 0.00 1.91
NGC 6860 0312192201 0.00+0.01?0.00 0.28+0.04?0.03 523.5/406 0.57 9.84
NGC 6921 0312192301 98.33+14.57?12.91 1.95+0.40?0.37 287.3/168 0.00 2.58
MCG +04-48-0023 0312192301 0.0 1.98 146.0/56 0.02 0.06
1Cold hydrogen column density in units of 1022 cm?2 from the tbabs model. This accounts for
absorption beyond the Galactic values which are listed in Table 4.1. A dash in this column indicates that no
extra absorption was necessary.
2The observed soft flux (0.3-2keV) and hard flux (2-10keV) are given in units of 10?12 ergcm?2 s?1.
3Errors at the 90% confidence range are not meaningful when ??2/dof > 2.0.
137
Table 4.4: SWIFT XRT Spectral Analysis: Absorbed Power Law Fits
Source Observation ID nH1 ? ?2/dof Fsoft2 Fhard2
MRK 352 00035243001 0.01+0.02?0.01 1.75+0.07?0.06 181.3/133 4.04 7.75
MRK 352 00035243002 0.00+0.01?0.00 1.65+0.04?0.04 152.0/152 2.69 5.95
SWIFT J0216.3+5128 00035247001 2.37+0.33?0.29 2.15+0.18?0.17 56.1/67 1.35 10.75
SWIFT J0216.3+5128 00035247002 1.15+0.25?0.23 1.77+0.21?0.20 46.0/44 1.30 11.10
SWIFT J0216.3+5128 00035247003 1.29+0.75?0.65 1.88+0.59?0.55 15.7/11 1.06 9.17
NGC 11423 00035248001 83.45 3.88 35.5/9 0.00 5.07
NGC 1142 00035248002 28.19+20.59?10.99 1.51+1.89?1.18 9.3/7 0.00 7.36
NGC 1142 00035248003 95.73+84.77?36.88 3.11+4.44?2.08 2.8/4 0.00 6.59
SWIFT J0318.7+6828 00035249001 3.70+1.74?1.83 1.73+0.48?0.55 17.4/17 0.15 4.11
SWIFT J0318.7+6828 00035249002 3.66+1.44?1.41 1.44+0.40?0.46 11.6/16 0.13 4.78
ESO 548-G081 00035250001 ? 1.92+0.04?0.05 131.7/136 13.91 19.62
ESO 548-G081 00035250002 ? 2.06+0.03?0.02 332.6/225 16.19 18.34
ESO 362-G018 00035234001 0.00+0.01?0.00 1.79+0.09?0.09 36.6/33 7.66 12.69
ESO 362-G018 00035234002 ? 1.45+0.05?0.05 159.7/108 4.38 12.16
ESO 490-G026 00035256001 0.27+0.04?0.03 1.88+0.07?0.07 143.8/177 7.07 18.91
ESO 490-G026 00035256002 0.27+0.09?0.09 1.70+0.14?0.13 59.2/62 6.32 21.07
ESO 490-G026 00035256003 0.49+0.15?0.14 1.88+0.19?0.18 27.2/42 4.28 14.49
Continued on Next Page...
138
Table 4.4 ? Continued
Source Observation ID nH1 ? ?2/dof Fsoft2 Fhard2
SWIFT J0904.3+5538 00035260001 0.00+0.01?0.00 1.54+0.08?0.08 39.9/40 2.04 5.02
SWIFT J0904.3+5538 00035260002 0.00+0.02?0.00 1.49+0.09?0.09 34.2/32 1.51 4.01
MCG +04-22-042 00035263001 0.01+0.00?0.01 1.90+0.05?0.05 173.9/185 7.99 11.69
MRK 4173 00035264003 ? -1.78 38.5/4 0.00 2.16
UGC 6728 00035266001 0.01+0.02?0.01 1.82+0.07?0.07 106.6/117 6.88 11.78
UGC 6728 00035266002 0.00+0.04?0.00 1.76+0.15?0.08 62.4/42 4.58 8.37
UGC 6728 00035266003 0.04+0.08?0.04 1.74+0.25?0.20 19.6/19 4.80 9.67
SWIFT J1200.8+0650 00035267001 11.23+3.50?2.62 1.60+0.48?0.41 39.5/28 0.01 4.99
SWIFT J1200.8+0650 00035267002 4.48+13.37?4.48 1.12+2.03?1.27 2.7/3 0.00 5.68
ESO 506-G0273 00035173004 36.38 -0.22 33.7/9 0.00 5.02
WKK 1263 00035268001 0.04+0.22?0.04 1.63+0.27?0.17 26.8/24 2.09 7.79
WKK 1263 00035268002 0.12+0.11?0.10 1.63+0.14?0.13 55.1/59 1.73 6.98
MCG +09-21-096 00035269001 0.00+0.01?0.00 1.80+0.04?0.03 250.2/240 13.67 22.13
MCG +09-21-096 00035269004 0.00+0.02?0.00 1.72+0.05?0.04 143.9/151 11.41 21.09
NGC 6860 00035275001 0.28+0.15?0.11 1.17+0.20?0.18 30.9/27 1.55 11.54
1Cold hydrogen column density in units of 1022 cm?2 from the tbabs model. This accounts for absorption beyond the
Galactic values which are listed in Table 4.1. A dash in this column indicates that no extra absorption was necessary.
2The observed soft flux (0.3-2keV) and hard flux (2-10keV) are given in units of 10?12 ergcm?2 s?1.
139
3Errors at the 90% confidence range can not be calculated for sources when ??2/dof > 2.0.
Simple Power Law Sources
A total of five of our sources (5/22) had XMM-Newton spectra best-fit by sim-
ple power law models. None of these sources showed evidence of a strong Fe K
line (EW < 150eV for all of these sources). We record the equivalent width and
normalization value for these lines in Table 4.5. Since the redshift for the source
SWIFTJ0216.3+5128is unknownand there is no evidence for a strong line, we ex-
clude this source from the analysis. We attempted to obtain an optical spectrum
for this source as part of our BAT AGN follow-up program with the 2.1-m tele-
scope at Kitt Peak National Observatory (Winter et al. in preparation). We found,
however, thattheprobableDigitalSkySurveyopticalcounterpartwithintheBAT
and XMM-Newton error circles was a star. Likely, the true optical counterpart of
thissourceisfaintandbelowtheDSSdetectionlimit. Basedonthelackofabright
optical counterpart and the featureless X-ray power law spectrum, this source is
probably a blazar, based on its similarity to other BAT blazars which have a high
X-ray/optical ratio.
Soft Excesses
Four sources (4/22) had XMM-Newton spectra best represented by a power law
with a soft excess. Since the spectra do not have the counts necessary to dis-
tinguish between a reflection or absorption model, we used a simple blackbody
(bbody in XSPEC) model to characterize this component. The best-fit spectral
parameters are shown in Table 4.6. The soft excess is significant for all of these
sources with ??2 values, between the simple power law and the power law with
a blackbody model, from ? 220?1400. As with the pure power law sources, we
140
Table 4.5. XMM-Newton Detailed Fits For Power Law Sources
Source nH1 ? Fe K EW2 Fe K norm.2 ?2/dof Fsoft3 Fhard3
MCG +04-22-042 ? 2.01+0.01?0.01 130.38+33.95?38.60 21.0+5.50?6.20 1801.9/1818 12.44 15.65
UGC 6728 0.01?0.01 1.78+0.03?0.02 83.33+56.58?45.66 5.73+3.89?3.14 1198.8/1159 3.59 6.51
WKK 1263 0.06+0.01?0.02 1.54+0.02?0.02 31.23+32.35?23.67 3.69+3.83?2.80 1253.0/1247 2.32 9.99
MCG +09-21-096 ? 1.77+0.01?0.01 55.94+31.09?29.09 9.83+5.47?5.11 1281.6/1625 9.26 15.89
1Cold hydrogen column density in units of 1022 cm?2 from the tbabs model. This accounts for absorption
beyond the Galactic values which are listed in Table 4.1. A dash in this column indicates that no extra absorption
was necessary.
2Equivalent width (in eV) and flux normalization for an inserted Gaussian line at 6.4keV (redshifted) with a set
FWHM of 0.01keV. The flux normalization is in units of 10?6? total photonscm?2 s?1 in the line.
3The observed soft flux (0.3-2keV) and hard flux (2-10keV) are given in units of 10?12 ergcm?2 s?1.
Table 4.6. XMM-Newton Detailed Fits for Power Law + Blackbody Sources
Source nH1 kT2 ? Fe K EW3 Fe K norm.3 ?2/dof Fsoft4 Fhard4
MRK 352 ? 0.096+0.003?0.003 1.70+0.02?0.01 60.29+28.09?27.89 7.78+3.62?3.60 1695.2/1566 6.50 11.89
ESO 548-G081 ? 0.087+0.003?0.003 1.85+0.01?0.01 114.35+38.41?22.56 18.7+4.60?4.70 1760.8/1639 10.84 15.83
ESO 490-G026 0.33+0.04?0.02 0.074+0.005?0.006 1.67+0.03?0.02 59.37+23.36?23.28 13.3+5.20?5.23 1542.9/1754 5.31 19.49
SWIFT J0904.3+5538 0.06+0.03?0.02 0.070+0.004?0.006 1.71+0.06?0.03 40.71+71.76?40.71 2.79+4.92?2.79 903.2/862 3.05 5.96
1Cold hydrogen column density in units of 1022 cm?2 from the tbabs model. This accounts for absorption beyond the Galactic
values which are listed in Table 4.1. A dash in this column indicates that no extra absorption was necessary.
2Temperature of the blackbody component (kT) in keV units.
3Equivalent width and flux normalization for an inserted Gaussian line at 6.4keV (redshifted) with a set FWHM (in eV) of 0.01keV.
The flux normalization is in units of 10?6? total photonscm?2 s?1 in the line.
4The observed soft flux (0.3-2keV) and hard flux (2-10keV) are given in units of 10?12 ergcm?2 s?1.
fit an Fe K line at 6.4keV with a Gaussian. The results are recorded in Table 4.6.
141
Warm Absorbers
In our models for AGN spectra, we assume that the absorption along the line of
sight is cold, neutral hydrogen (with the tbabs model). However, signatures of
warm, optically-thin absorption from photo-ionized gas, have been detected in
half of an ASCA Seyfert 1 AGN sample studied by Reynolds (1997) and 13/18
of the ASCA Seyfert 1 AGN sample of George et al. (1998). The main signatures
of a warm absorber are the O VII and O VIII K edges at 0.74keV and 0.87keV.
Since warm absorbers are seen in such a large fraction of ASCA observations
of Seyfert 1 sources and our data are of high enough quality to distinguish the
warm absorber signatures, or at least constrain upper limits, we looked for these
signatures among the sources with a high number of counts below 2keV (with
the exception of possible blazar SWIFT J0216.3+5128).
Following Reynolds (1997), we added two edge models (zedge in XSPEC)
to account for the O VII and O VIII K edges. We fixed the edge at the energies
of these warm absorber signatures (0.74keV and 0.87keV), allowing the optical
depth to vary. In Table 4.7, we record the errors on optical depth and the change
in ?2. In the Reynolds (1997) sample, optical depths for clearly detected edges
ranged from? 0.10?1.5. Fromour spectral fitting, halfof the sources haveupper
limits of ? < 0.10 for O VII and 7/8 have upper limits of ? < 0.10 for O VIII.
In fact, the only source with a clear detection of both edges is ESO 490-G026.
Clearly, much less than half of our low absorption/simple X-ray model sources
show evidence of warm absorbing material in their spectra. We will discuss our
results further in ? 4.4.
142
Table 4.7. XMM-Newton Warm Absorber Model
Source ? O VII a ? O VIIIa ??2b
MRK 352 0.017 0.005 -12.3
ESO 548-G081 0.032 0.005 -0.06
ESO 490-G026 0.233+0.048?0.083 0.095+0.047?0.043 25
SWIFT J0904.3+5538 0.186+0.098?0.091 0.039 9.9
MCG +04-22-042 0.063+0.022?0.029 0.035+0.026?0.025 41.5
UGC 6728 0.072 0.023 0.8
WKK 1263 0.046+0.087?0.046 0.020 1.0
MCG +09-21-096 0.036+0.023?0.025 0.008 5.7
aOptical depth with errors or upper limits for the ad-
ditions of edge models (zedge) at 0.74keV and 0.87keV.
bThe ?2 from the best fit model in Table 4.5 or 4.6 mi-
nus ?2 from the warm absorber model.
Complex X-ray spectra sources
The remaining 13 sources (? 1/2) had spectra we classified as ?double power
law? spectra, similar to Turner et al. (1997). Earlier, we mentioned three scenar-
ios that could create this type of a spectrum. Based on these three possibilities
(viewing anothersource(s) such asX-raybinaries/diffuse galacticemission along
with the absorbed AGN, cold gas partially covering the AGN light, and scatter-
ing of the AGN light from regions of different column densities) we decided to fit
the spectra of these sources with the partial covering model, pcfabs in XSPEC.
The partial covering model (see: Holt et al. (1980)) has two parameters: the hy-
143
drogen column density and the covering fraction. In addition to describing a
partial absorber spectrum, this model is also useful for describing a spectrum
where AGN light has been scattered, where the power law index of the direct,
heavily absorbed spectrum is the same as that of the scattered component (which
is not heavily absorbed). This model, to summarize, allows for flexibility in the
soft spectrum which can fit all of the three physical origins mentioned for a com-
plex spectrum (assuming the X-ray binary/diffuse emission has the same power
law slope as the AGN emission, which is not expected). Thus, we fit the spec-
tra of the remaining 13 sources with the model tbabs*pcfabs*(pow + zgauss)
(Table 4.8).
As we noted, the partial covering model gives an equivalent result to a model
with a power law modified by different amounts of absorption and two flux nor-
malizations. We also wanted to test whether the power law spectral indices for
these two components are the same or differ. To this end, we fit the sources with
the model tbabs*(tbabs*pow + tbabs*(pow + zgauss)). This model allows
the power law indices, normalizations, and column densities to vary for two sep-
arate power laws. The results do not differ significantly for half of the sources be-
tween the partial covering model (Table 4.8) and the separate power laws model
(Table 4.9), with ??2 < 8. Each of these seven sources show a ratio of the low ab-
sorption power law to the more highly absorbed power law flux (N?1/N?2) less
than 0.14 with an average value of 0.03. These results could be consistent with
any of the three possible physical models, where the portion of scattered light
or additional non-AGN emission or unabsorbed AGN light is very small. This
is true of all of the sources, with the exception of ESO 362-G018 and NGC 6860,
whose spectra are more complicated.
For the BAT source corresponding to the interacting system (NGC 6921 and
144
MCG +04-48-002), both sources are clearly absorbed. For MCG +04-48-002, the
absorbed power law component in the double power law model was not well
constrained due to the weak contribution from the low absorption power law
component. This source was weaker by an order of magnitude than NGC 6921 in
the XMM-Newton observation. Thus, throughout the remainder of the paper we
will distinguish NGC 6921 as ?the BAT source?. This will not change any results,
since both sources are absorbed sources with similar spectral results. We added
MCG+04-48-002tothespectralfitsbecause, thoughitisclearlytheweakersource
in the XMM observations, it is brighter than NGC 6921 in the XRT observations.
Further, in recently obtained Suzaku observations, which we are currently ana-
lyzing, we found that MCG +04-48-002 was the brighter source. These findings
and a detailed analysis will be discussed in a future paper.
Based on reduced ?2, three of the ?double power law? sources (NGC 1142,
ESO 362-G018, and ESO 506-G027) require additional/alternative models. For
NGC 1142, a soft excess is clearly present (see Figure 4.2). The addition of a
blackbody component with a temperature kT= 0.123keV improved the separate
power laws fit (Table 4.9), yielding an acceptable fit with ?2/dof of 217.13/191.
We note that this blackbody temperature, kT= 0.123keV, is similar to the values
seen in Table 4.6 for the sources fit with simple blackbody and power law mod-
els. The spectrum of ESO362-G018, however, is still even more complicated. This
spectrum appears to have well defined lines, particularly a strong line measured
with an energy of 0.56keV, which is consistent with O VII and improves the sep-
arate power laws fit by ??2 = 50. However, the power law spectral index for
this source is still extremely flat (? = 0.67) where the typical value for ? is ? 1.80
(Mushotzky 1982). This is also true of NGC 612, MRK 417, ESO 506-G027, and
NGC 6860.
145
Flat power law indices have been noted, in addition to high column densities
(nH ? 1024 cm?2) and strong Fe K lines (EW greater than a few hundred eV),
as indicators of Compton-thick AGN (Matt et al. 1996). For NGC 612, MRK 417,
and ESO 506-G027, all of these factors are met. However, even though ESO 362-
G018 has a strong Fe K line and a flat spectrum, the fitted column density is only
6.3 ? 1022 cm?2. The spectrum of NGC 6860 is even odder, with both power law
components (see Table 4.9) having a flat slope with very low hydrogen column
densities (nH < 1022 cm?2) and no strong Fe K line.
ThoughhighcolumndensitiesarepossibleindicatorsofCompton-thicksources,
it is also possible for the X-ray spectrum of a Compton-thick source to have a
lower measured column density. Such a model, as was proposed for the Seyfert
1 source MRK 231 by Maloney & Reynolds (2000), could be applicable for ESO
362-G018 and NGC 6860. In this model, the central power law source is blocked
by Compton-thick material. The resulting reflection component is then scattered
and absorbed elsewhere, outside the Compton-thick region. As a result, the mea-
sured X-ray column density is from this second absorbing region. Therefore, we
include ESO 362-G018 and NGC 6860 in our list of Compton-thick candidates,
despite their low column densities.
For our Compton-thick candidates, the sources with flat power law indices
(ESO 362-G018, NGC 6860, NGC 612, MRK 417, and ESO 506-G027), we simul-
taneously fit the 8-channel BAT spectrum along with the XMM-Newton spectra,
allowingaconstantfactortovaryfortheBATdata,aswedidfortheMOSspectra.
Increasing the energy range to 195keV allows for better constraints on the power
law component at high energies. Also, since a Compton-thick source spectrum
should be a heavily reflected spectrum, it is extremely important to have higher
energy data to determine the cutoff energy of the power law. This is evident con-
146
Ta
ble
4.8
.
XM
M
-N
ew
ton
De
tai
led
Fit
s:
Pa
rti
al
Co
ve
rin
gM
od
el
So
ur
ce
nH
1
fra
cti
on
1
?
Fe
K
eq
w2
Fe
K
no
rm
.2
?2
/d
of
Fs
of
t3
Fh
ar
d3
NG
C
61
2
12
9.7
0+
12
.9
0
?
8.
30
0.9
99
+
0.
00
1
?
0.
00
2
2.1
2+
0.
06
?
0.
33
10
8.1
0+
63
.7
4
?
62
.8
3
17
.9+
10
.6
?
10
.4
11
6.7
/8
1
0.0
2
1.5
0
NG
C
11
42
79
.75
+
5.
81
?
3.
05
0.9
96
+
0.
00
1
?
0.
00
1
2.2
7+
0.
09
?
0.
17
21
9.2
1+
57
.3
0
?
40
.8
7
39
.0+
10
.2
?
7.
27
26
0.5
/1
95
0.0
8
3.0
1
SW
IFT
J03
18
.7+
68
28
4.1
0+
0.
48
?
0.
41
0.9
67
+
0.
00
9
?
0.
00
9
1.5
2+
0.
12
?
0.
11
44
.15
+
41
.5
8
?
39
.4
3
5.3
5+
5.
04
?
4.
78
77
5.9
/7
70
0.2
4
7.3
1
ES
O
36
2-G
01
8
26
.64
+
2.
72
?
2.
49
0.9
13
+
0.
00
5
?
0.
01
1
2.1
3+
0.
04
?
0.
04
42
1.8
2+
35
.1
5
?
10
5.
96
28
.6+
2.
38
?
7.
18
82
4.1
/5
47
0.6
3
3.3
4
SW
IFT
J06
41
.3+
32
57
16
.01
+
2.
68
?
2.
28
0.9
82
+
0.
00
7
?
0.
00
9
1.2
4+
0.
26
?
0.
23
7.7
6+
46
.0
1
?
7.
76
0.5
1+
3.
00
?
0.
51
17
6.9
/1
52
0.0
2
3.3
1
M
RK
18
18
.25
+
3.
64
?
2.
71
0.9
7+
0.
02
?
0.
02
1.6
2+
0.
31
?
0.
22
17
8.0
9+
10
9.
77
?
10
7.
72
5.4
7+
3.
37
?
3.
31
32
2.4
/3
25
0.0
3
1.5
7
SW
IFT
J09
11
.2+
45
33
33
.02
+
11
.0
1
?
12
.7
6
0.9
94
+
0.
00
4
?
0.
00
9
2.4
7+
0.
98
?
1.
20
58
2.7
6+
42
5.
24
?
31
7.
37
23
.7+
17
.3
?
12
.9
9.3
/1
5
0.0
1
1.4
5
M
RK
41
7
85
.69
+
12
.7
3
?
6.
96
0.9
95
+
0.
00
2
?
0.
00
2
2.2
5+
0.
15
?
0.
17
11
4.7
7+
75
.7
8
?
80
.9
7
10
.8+
7.
11
?
7.
60
88
.1/
80
0.0
6
1.4
4
SW
IFT
J12
00
.8+
06
50
10
.80
+
0.
68
?
0.
82
0.9
91
+
0.
00
2
?
0.
00
3
1.4
7+
0.
10
?
0.
11
84
.18
+
47
.1
3
?
27
.1
5
7.6
9+
4.
31
?
2.
48
34
0.3
/3
55
0.0
4
5.0
8
ES
O
50
6-G
02
7
76
.82
+
7.
37
?
6.
79
0.9
86
+
0.
00
6
?
0.
00
3
0.9
1+
0.
23
?
0.
18
42
8.8
9+
50
.0
1
?
85
.0
1
72
.0+
8.
40
?
14
.3
22
8.9
/1
70
0.0
3
3.8
7
NG
C
49
92
69
.05
+
7.
39
?
2.
38
0.9
97
4+
0.
00
15
?
0.
00
16
1.6
1+
0.
13
?
0.
41
32
0.2
2+
80
.9
3
?
72
.5
6
29
.8+
5.
55
?
5.
39
12
0.8
/1
32
0.0
1
1.9
3
NG
C
68
60
4.5
3+
1.
33
?
1.
30
0.6
0+
0.
07
?
0.
10
0.7
9+
0.
11
?
0.
15
75
.64
+
62
.9
0
?
64
.4
8
9.4
5+
7.
86
?
8.
06
48
3.5
/4
04
0.5
4
8.9
9
NG
C
69
21
97
.42
+
19
.0
8
?
3.
78
0.9
98
4+
0.
00
07
?
0.
00
12
2.3
4+
0.
11
?
0.
21
54
.56
+
47
.6
8
?
53
.0
2
11
.9+
9.
8
?
10
.3
19
0.7
/1
54
0.0
3
2.5
3
M
CG
+0
4-4
8-0
02
96
.00
+
51
.9
7
?
27
.7
7
0.9
96
3+
0.
00
23
?
0.
01
33
2.4
7+
0.
32
?
0.
44
81
1+
12
19
?
32
9
42
.3+
63
.7
?
17
.2
62
.8/
52
0.0
25
0.3
1
1 C
old
hy
dr
og
en
co
lum
n
de
ns
ity
in
un
its
of
10
22
cm
?
2
fro
m
th
ep
cf
ab
s
mo
de
l.
Th
is
ac
co
un
ts
for
ab
so
rp
tio
n
be
yo
nd
th
eG
ala
cti
c
va
lue
sw
hic
ha
re
lis
ted
in
Ta
ble
4.1
.T
he
fra
cti
on
co
lum
ni
st
he
pa
rti
al
co
ve
rin
gf
rac
tio
n.
2 E
qu
iva
len
tw
id
th
(in
eV
)a
nd
flu
xn
or
ma
liz
ati
on
for
an
ins
ert
ed
Ga
us
sia
n
lin
ea
t6
.4
ke
V
(re
ds
hif
ted
)w
ith
as
et
FW
HM
of
0.0
1k
eV
.
Th
efl
ux
no
rm
ali
za
tio
ni
si
nu
nit
so
f1
0?
6 ?
tot
al
ph
oto
ns
cm
?
2
s?
1
in
th
el
ine
.
3 T
he
ob
se
rv
ed
so
ft
flu
x(
0.3
-2
ke
V)
an
dh
ard
flu
x(
2-1
0k
eV
)a
re
giv
en
in
un
its
of
10
?
12
erg
cm
?
2
s?
1 .
147
Ta
ble
4.9
.
XM
M
-N
ew
ton
De
tai
led
Fit
s:
Do
ub
le
Po
we
rL
aw
M
od
el
So
ur
ce
nH
1
?1
nH
?2
N?
1/
N?
22
Fe
K
eq
w3
Fe
K
no
rm
.3
?2
/d
of
Flu
x4
NG
C
61
2
0.3
2+
0.
09
?
0.
18
4.2
7+
1.
78
?
1.
15
79
.79
+
19
.4
7
?
12
.3
0
0.2
8+
0.
21
?
0.
19
0.2
15
17
7.0
0+
86
.9
6
?
74
.3
6
12
.8+
6.
25
?
5.
37
82
.0/
79
1.6
9
NG
C
11
42
0.1
5+
0.
07
?
0.
06
3.1
0+
0.
31
?
0.
71
62
.94
+
10
.8
7
?
5.
72
1.4
9+
0.
31
?
0.
17
0.0
28
26
8.1
0+
62
.4
8
?
52
.5
0
35
.4+
8.
31
?
6.
90
24
5.8
/1
93
3.1
9
SW
IFT
J03
18
.7+
68
28
1.1
1+
1.
42
?
1.
11
1.5
1+
2.
63
?
0.
72
6.1
0+
8.
15
?
6.
10
1.7
4+
1.
10
?
0.
24
0.1
39
45
.29
+
62
.8
9
?
37
.0
3
5.6
3+
5.
36
?
4.
79
77
2.8
/7
68
7.4
1
ES
O
36
2-G
01
8
0.0
0+
0.
01
?
0.
00
2.3
0+
0.
07
?
0.
06
6.3
0+
2.
32
?
2.
66
0.6
7+
0.
18
?
0.
23
1.7
53
43
2.2
3+
11
0.
50
?
32
.5
9
22
.1+
3.
31
?
3.
40
73
6.0
/5
45
4.2
6
SW
IFT
J06
41
.3+
32
57
0.0
0+
0.
04
?
0.
00
0.8
2+
1.
38
?
0.
76
16
.96
+
9.
75
?
3.
30
1.4
5+
0.
78
?
0.
42
0.0
12
1.5
8+
55
.0
2
?
1.
58
0.2
0+
3.
37
?
0.
20
17
6.5
/1
50
3.3
0
M
RK
18
0.3
6+
0.
57
?
0.
29
3.7
4+
2.
79
?
1.
67
18
.28
+
7.
22
?
5.
71
1.7
8+
0.
65
?
0.
56
0.0
48
18
8.5
7+
11
3.
34
?
10
9.
56
5.9
1+
3.
33
?
3.
49
31
5.4
/3
23
1.5
3
SW
IFT
J09
11
.2+
45
33
3.8
4+
9.
73
?
3.
84
2.5
6+
7.
44
?
5.
52
48
.03
+
28
.3
2
?
24
.6
1
3.3
5+
6.
59
?
2.
04
0.0
02
69
5.5
6+
12
13
.0
4
?
43
7.
52
32
.0+
55
.8
?
20
.1
8.2
/1
3
1.3
6
M
RK
41
7
0.0
0+
0.
04
?
0.
00
2.3
6+
0.
16
?
0.
16
54
.15
+
25
.0
4
?
11
.3
0
0.8
8+
0.
97
?
0.
42
0.1
01
17
9.2
0+
87
.7
1
?
86
.6
4
9.6
5+
4.
72
?
4.
67
78
.8/
78
1.6
2
SW
IFT
J12
00
.8+
06
50
0.0
0+
0.
06
?
0.
00
1.7
5+
0.
93
?
0.
43
10
.57
+
0.
79
?
0.
98
1.4
3+
0.
10
?
0.
08
0.0
09
94
.46
+
35
.7
0
?
35
.9
6
8.5
8+
3.
24
?
3.
27
33
8.9
/3
53
5.1
4
ES
O
50
6-G
02
7
0.0
0+
0.
08
?
0.
00
0.8
1+
0.
28
?
0.
31
80
.37
+
7.
77
?
8.
68
0.9
9+
0.
48
?
0.
24
0.0
11
41
8.2
2+
69
.3
9
?
59
.6
6
73
.6+
12
.2
?
10
.5
22
8.1
/1
68
3.8
9
NG
C
49
92
0.3
2+
0.
73
?
0.
31
2.6
8+
3.
73
?
1.
69
68
.47
+
14
.9
4
?
9.
62
1.4
1+
0.
53
?
0.
14
0.0
06
31
8.3
1+
90
.9
4
?
74
.9
1
27
.9+
7.
96
?
6.
56
70
.4/
78
1.9
7
NG
C
68
60
0.0
0+
0.
01
?
0.
00
0.4
7+
0.
10
?
0.
06
0.6
0+
9.
45
?
0.
13
0.4
8+
1.
56
?
0.
08
0.7
45
65
.36
+
60
.5
7
?
62
.1
5
8.1
7+
7.
57
?
7.
77
45
9.7
/4
02
9.9
7
NG
C
69
21
0.5
7+
0.
20
?
0.
21
5.0
8+
2.
53
?
1.
25
90
.25
+
10
.7
5
?
10
.6
3
1.7
1+
0.
53
?
0.
25
0.0
25
89
.01
+
54
.9
3
?
58
.3
7
14
.7+
9.
11
?
9.
67
14
9.3
/1
64
2.6
4
M
CG
+0
4-4
8-0
02
0.1
3+
0.
36
?
0.
13
3.0
6+
1.
07
?
0.
78
22
3.5
8+
48
.0
6
?
88
.1
6
10
.00
?
6.
63
1.6
e-9
49
6+
95
4
?
27
9
34
.2+
66
.6
?
19
.1
58
.4/
50
0.2
6
1 C
old
hy
dr
og
en
co
lum
nd
en
sit
yi
nu
nit
so
f1
02
2
cm
?
2
fro
m
th
et
ba
bs
mo
de
l.
Th
is
ac
co
un
ts
for
ab
so
rp
tio
nb
ey
on
d
th
eG
ala
cti
cv
alu
es
wh
ich
are
lis
ted
in
Ta
ble
4.1
.In
th
is
mo
de
l,t
he
re
is
as
ep
ara
te
co
lum
nd
en
sit
yc
om
po
ne
nt
for
ea
ch
of
th
et
wo
po
we
rl
aw
s.
2 R
ati
oo
ft
he
flu
xn
or
ma
liz
ati
on
va
lue
for
po
we
rl
aw
on
ev
ers
us
po
we
rl
aw
tw
o.
Fo
rn
ea
rly
all
of
th
es
es
ou
rce
s,
th
el
ow
ab
so
rp
tio
n
po
we
rl
aw
is
mu
ch
we
ak
er.
3 E
qu
iva
len
tw
id
th
(in
eV
)a
nd
flu
x
no
rm
ali
za
tio
n
for
an
ins
ert
ed
Ga
us
sia
n
lin
ea
t6
.4
ke
V
(re
ds
hif
ted
)w
ith
a
se
tF
W
HM
of
0.0
1k
eV
.T
he
flu
x
no
rm
ali
za
tio
ni
si
nu
nit
so
f1
0?
6 ?
tot
al
ph
oto
ns
cm
?
2
s?
1
in
th
el
ine
.
4 T
he
ob
se
rv
ed
tot
al
flu
x(
0.5
?1
0k
eV
)in
un
its
of
10
?
12
erg
cm
?
2
s?
1 .
148
10
?3
0.01
0.1
normalized counts/sec/keV
1 100.5 2 50
1
2
3
ratio
channel energy (keV)
Figure 4.2: XMM-Newton spectrum (pn, MOS1, and MOS2) of NGC 1142 fit with
the model tbabs*pcfabs*(pow + zgauss)*const. There is a clear soft excess
with possible unresolved lines (in pn). We find that a better fit to this source is
obtained with the addition of a blackbody model.
149
sidering that the reflection spectrum (pexrav in XSPEC (Magdziarz & Zdziarski
1995)) depends upon the cutoff energy of a power law, in addition to iron abun-
dance, reflection factor, and geometry of the system. When we use this model,
we fixed the iron abundance to the solar value and the inclination angle of the
system to the default (60?).
We note that in using the BAT spectra we are assuming that the individual 14
? 195keV spectra donotvaryover theperiod of22 monthsused to create the BAT
spectrum. Future AGN observations with Suzaku, which can obtain simultane-
ousspectrafrom0.3?200keV,willallowustotesttheaccuracyofthisassumption.
To this end, we have obtained and are processing the Suzaku spectrum for one of
our Compton-thick candidates, MRK 417. A paper is in preparation.
i. Low column density Compton-thick candidates
ESO 362-G018 The X-ray spectrum of ESO 362-G018 was not well-fit by either
the partial covering or double power law model. While the double power law
model provided the best fit, with a reduced ?2 value of 1.35, this is not satisfac-
tory. In the residuals from the fit, at least two emission line features were present.
Adding gaussians for these lines, the fit improved by ??2 = 72. The energies of
these lines (0.56keV and 0.90keV), which have fluxes on the order of the Fe K
line flux, correspond to helium-like oxygen and possibly helium-like neon lines.
The flat power law and strong Fe K line suggest a reflection spectrum. Simul-
taneously fitting the XMM-Newton spectra with the BAT spectrum, we replaced
the heavily absorbed power law model with a reflection model (pexrav). This
model is an acceptable fit to the data with a reduced ?2 of 1.14. However, though
the reflected power law component is more typical of AGN (? = 1.99+0.15?0.27), the
column density is extremely low for a Compton-thick source (nH = 6.2+9.0?2.6 ?
150
1021 cm?2). As mentioned earlier, the observed low column density could be the
result of scattering of the reflection spectrum through a second absorbing region
of lower column density. The details of this fit are listed in Table 4.10 along with
the other Compton-thick candidates.
From the HST observation of this source, Das et al. (2006) describe the im-
age as showing dusty lanes that are interspersed with star-forming regions. This
complex environment could partially cover some of the X-ray emission as well
as contribute line-emission from young stars. Similarly, ESO 362-G018 could be
a very Compton-thick source whose flat, reflection-dominated spectrum is scat-
tered and viewed through an absorber (Maloney & Reynolds 2000). From the
? 10ks XMM-Newton exposure, it is clear that the spectrum is complex. For
a more accurate description of the source spectrum, a longer observation with
higher signal-to-noise is necessary.
NGC 6860 From a literature search, we found that NGC 6860 is well studied
in the optical and IR. Bennert et al. (2006) find optical emission line diagnostics
of this source indicative of an intermediate state between an AGN and starburst
galaxy. Indeed, they state that the AGN dominates only in the inner 10?? in the
IR/optical. While it is unclear whether this is also true in the X-ray band, we
necessarily extracted a spectrum from a much larger, 85?? region. The Optical
Monitor pipeline processed images (U, UVW1, UVM2) also confirm the presence
of star formation, where the nucleus is seen surrounded by a ring of star forming
regions. From the evidence of the optical and IR observations, it is likely that the
spectral form of NGC 6860 is composite, with both star burst/star formation and
AGN contributions.
From the double power law model, we found that residuals to the fit indi-
151
cated a soft excess. Given the optical and IR evidence of star formation, we
added an apec model to fit this excess. The apec model in XSPEC is a model for
collisionally-ionized diffuse gas. Since the quality of the XMM-Newton spectra is
too low to distinguish between collisionally and photo-ionized gas, we used the
simpler, collisionally ionized model. As input parameters, the apec model re-
quires a plasma temperature, metal abundance, redshift, and normalization. We
fixed the redshift to the source?s value and set the abundance to the solar value.
Adding this model, with a best-fit plasma temperature of kT= 0.14keV and nor-
malization of 2?10?4, improved the double power law fit by ??2 = 6. Therefore,
this is not significant at the 90% level. One thing to note, however, is that the ad-
dition of theapecmodel causes ?1, the low absorption power law component, to
steepen from 0.47 to 3.60. The higher absorption power law model components,
column density and power law index, do not change.
Adding a reflection model to this fit (double power law with an apecmodel),
improves the fit by ??2 = 10. With this model, we could not constrain the reflec-
tion factor or folding energy. We then added the BAT data to the XMM-Newton
spectra. Still, the reflection factor and folding energy were not constrained (they
continued to increase to unphysical values). We fixed these values to a folding
energy of 100keV and complete reflection. Though this model does not make
physical sense, since a strong Fe K line is expected in a reflection dominated
spectrum, it was clear from the fitting that a high reflection factor is preferred
statistically. This reflection dominated model obtains a good statistical fit to the
data with ?1 = 1.14 and ?2 = 2.31 and ?2/dof = 444.5/408 (see Table 4.10). How-
ever, we stress that since there is no strong Fe K line and the column density is
low, we do not believe that this model is a good physical description of the data.
For this source, we also fit the data with a double partial covering model
152
(which could possibly be justified in a clumpy, dusty environment) and a model
where we replaced the neutral absorption model, tbabs, with an ionized ab-
sorber, absori. Both models fit the data with similar ?2 values as the reflection
model (reduced ?2 of 1.02 and 1.08, respectively). Both models also cause the fit-
ted power law indices to steepen to values typical of AGN sources. We conclude
that the spectrum of this source is too complicated (see Figure 4.3) to quantify
with the data available. A better signal-to-noise spectrum is required to under-
stand this source?s X-ray spectrum.
ii. High column density Compton-thick candidates
As mentioned, we simultaneously fit the BAT spectra with the XMM-Newton
spectra for sources with flat power law indices. Our three additional Compton-
thick candidate sources are NGC 612, MRK 417, and ESO 506-G027, all with
nH > 5?1023 cm?2. For each of these sources, we replaced the heavily absorbed
power law component in the double power law model (Table 4.9) with the re-
flection model (pexrav). We record the absorbed column density, power law
index for the reflection component, cutoff energy (which was not constrained for
ESO 506-G027), normalization factor for the BAT spectrum, and goodness of fit
in Table 4.10.
Allowing the BAT flux normalization to vary by a constant multiplicative fac-
tor, we found that the values of the factors for NGC 612 and MRK 417 were very
low (<< 0.50). Examining the BAT spectra, there is clear curvature in the BAT
energy spectrum of these two sources, which is not well fit by the pexravmodel
(see Figure 4.4). This curvature is not seen in the other three Compton-thick can-
didate spectra (see Figure 4.3 for the spectrum of NGC 6860). Of particular note,
the BAT spectrum of NGC 612 appears flat (well-modeled by a power law index
153
Table 4.10. Compton-thick Reflection Model
Source nH1 ? cutoff E2 BAT norm3 ?2/dof
NGC 6124 62.55+11.90?23.92 0.95+0.36?0.80 48.58+83.42?24.18 0.08+0.09?0.05 81.7/85
ESO 362-G018 0.62+0.90?0.26 1.99+0.15?0.27 100 0.782+0.27?0.22 622.8/547
MRK 4174 18.19+9.00?13.57 0.73+0.34?0.53 106.7+112.10?54.70 0.19+0.12?0.09 114.6/83
ESO 506-G027 65.49+11.93?10.40 1.88+0.22?0.26 100 0.94+0.24?0.10 226.7/174
NGC 6860 0.00+1.04 2.31+0.14?0.17 100 1.21+0.32?0.19 444.5/408
1Cold hydrogen column density in units of 1022 cm?2 from the tbabs model.
This accounts for absorption beyond the Galactic values which are listed in Ta-
ble 4.1. In this model, there is a separate column density component for each of
the two power laws. The column densities listed are for the more heavily ab-
sorbed source.
2Cutoff energy for the pexrav/reflection model. For this model, we assumed
that this component was a pure reflection component and allowed the cutoff en-
ergy to vary. Where the cutoff energy could not be constrained (the model pa-
rameter increases to very large, unphysical values), we fixed this parameter at
the default value of 100keV.
3Using the constant model, we allowed the BAT normalization to vary by a
multiplicative factor. The recorded value is the factor variation of the BAT flux
to the pn flux (normalized at 1), except for NGC 6860 which has no pn data (nor-
malized to MOS1).
4For NGC 612 and MRK 417, the best fit to the data, in terms of reduced ?2,
was a model with low BAT normalization. Both of these sources showed cur-
vature in the BAT spectrum, which was not well fit by the pexrav model. We
include, in the text, a discussion of this as well as a description of the fits with
BAT normalization set to 1 (similar to the best fits for the remaining 3 Compton-
thick candidates).
154
10
?5
10
?4
10
?3
0.01
0.1
normalized counts/sec/keV
1 10 1000
0.5
1
1.5
2
ratio
channel energy (keV)
Figure 4.3: XMM-Newton MOS1 (no pn data was available for this source) and
SWIFT BAT spectra of NGC 6860 fit with a reflection model. The spectrum of
this source is very complex and could be adequately fit by a few different mod-
els (such as a reflection model, ionized absorption in place of neutral absorption,
and a double partial covering model). However, residuals in the model point
to complexity that can not be explained without higher signal-to-noise observa-
tions.
155
1 10 100
10
?5
10
?4
10
?3
0.01
0.1
keV (keV/cm
2  s keV)
channel energy (keV)
1 10 100
10
?6
10
?5
10
?4
10
?3
0.01
keV (keV/cm
2  s keV)
channel energy (keV)
Figure 4.4: XMM-Newton and BAT spectra of the Compton-thick candi-
date sources NGC 612 (left) and MRK 417 (right). The model used is
tbabs*(tbabs*pow + tbabs*(pexrav + zgauss))*const. The unfolded spec-
trum is plotted (E2 f(E) vs. E, where f(E) is the model). The fits to the sources are
described in the text. These fits were obtained with the constant factor set to 1.0
(normalized to the pn spectrum). The BAT spectra show some curvature and are
not well fit by this model. This was not true of the remaining 3 Compton-thick
candidates.
<< 1.0). For MRK 417, fixing the BAT multiplicative factor to 1.0 (the same as
the pn spectrum) leads to a worse fit to the data with ?2/dof= 144.7/85. With
this fit, the cutoff energy for the pexrav model becomes unconstrained while
the column density and power law index increase (nH = 3.4+0.9?0.8 ? 1023 cm?2 and
? = 1.85+0.12?0.12). The same effect happens with the spectrum of NGC 612, where the
best fit gives ?2/dof= 126.4/86 with nH = 8.2+0.9?2.6 ? 1023 cm?2 and ? = 1.09+0.34?0.45.
It is possible that the curvature seen in the BAT spectrum is a real feature of the
spectrum above 10keV. Since the BAT diagonal response matrix is based on the
harder spectrum of the Crab (? ? 2), it is invalid in this case. Further, since the
156
BAT spectra are time averaged over months of observations, simultaneous ob-
servations for the 2 ? 10keV and 15 ? 200keV bands are needed to determine
whether the BAT spectra correctly represent the very hard X-ray spectrum. We
have already obtained Suzaku spectra for MRK 417 and are in the process of an-
alyzing the data, which will be presented in an upcoming paper.
4.3.2 Variability
The main focus of our variability study is determining how the sources vary
between the XMM-Newton and XRT observations, on a timeframe of hours to
months. However, with 10ks XMM-Newton observations, we also looked for
shorter variability by examining the light curves of our sources. To this end, we
extracted light curves from the filtered pn (or MOS1 where there was no pn data
available) event files with the FTOOL XSELECT. We extracted light curves from
the same regions used to extract spectra, binned by 100s. We also extracted a
background light curve from a region of the same size on the same chip as the
source. We excluded SWIFT J0911.2+4533 from our analysis due to the low aver-
age count rate in the MOS1 observation (the source is located in a gap in a chip
for the pn observation), < 0.1cts?1, which is on the order of the count rate in the
background spectrum.
For the remaining 21 sources, following the analysis of Nandra et al. (1997),
we computed the normalized excess variance and ?2 values, for the assumption
that the flux was constant, to quantify variability. We list these values as well as
the average count rate in Table 4.11. Within our sample, 8 sources were flagged
as possibly variable during the XMM-Newton observation, with reduced ?2 >
1.5, corresponding to a probability of < 1% of the count rates corresponding to
constant count rates. For each of these 8 sources, we examined both the source
157
Figure 4.5: Light curves binned by 100s for the XMM-Newton pn observations
of MRK 352, ESO 548-G081, and UGC 6728. These two were the only sources to
show significant variability above the background level(?2/dof > 1.5 compared
to a constant flux model). Each of these sources are among the brightest flux
sources in our sample (though not the three brightest), all are optical Seyfert 1
sources.
and background light curves. We found that for 5 of the sources the source and
background light curves showed identical variability. For each of these sources,
the ratio of average background count rates to average source count rates was
relatively large, between 0.2 and 0.8. Thus, the background rates were significant
compared to the source rates. The variability, also seen in the background light
curve, was not intrinsic to the source for these sources. This was not the case for
MRK 352, ESO 548-G081, and UGC 6728. The variability for these bright, Seyfert
1 sources is measured source variability.
In Figure 4.5, we include the light curves for each of the 3 sources with vari-
abilityduringtheXMM-Newtonobservation. Fromthelightcurves, weestimated
158
Table 4.11. Variability in individual XMM-Newton Observations
Source <Ct Rate>a ?2rmsb ?2/dofc
MRK 352 7.023 7.010?0.0014 576.71/99
NGC 612 0.119 8.583?5.640 123.93/107
SWIFT J0216.3+5128 1.470 1.043?0.0048 82.57/100
NGC 1142 0.283 6.755?0.8434 82.51/100
SWIFT J0318.7+6828 1.166 49.05?0.5085 727.06/71
ESO 548-G081 11.730 4.497?3?10?4 620.33/100
ESO 362-G018 0.949 1.067?0.01185 107.24/100
ESO 490-G026 5.128 0.00523?2?10?6 105.16/103
SWIFT J0641.3+3257 0.243 62.76?8.863 273.35/120
MRK 18 0.507 94.36?3.251 995.16/113
SWIFT J0904.3+5538 0.960 0.7704?7?10?4 133.70/121
MCG +04-22-042 13.151 0.08061?6?10?5 88.10/80
MRK 417 0.145 19.83?12.28 103.53/77
UGC 6728 3.982 29.45?0.02293 1069.19/81
SWIFT J1200.8+0650 0.430 1.860?0.09154 100.27/110
ESO 506-G027 0.268 4.805?0.6714 117.02/100
WKK 1263 2.446 0.1615?3?10?4 94.24/89
MCG +09-21-096 9.348 0.02412?3?10?6 96.63/99
NGC 4992 0.136 9.399?3.531 163.66/143
NGC 6860 0.478 49.60?1.823 377.40/119
NGC 6921 0.268 105.8?12.07 500.70/122
aAverage count rate for the XMM-Newton observation in the 0.3 ?
10keV band.
bCorresponding excess variability measurements, as defined in
Nandra et al. (1997) ?10?3.
c?2 value divided by the number of bins for variability.
159
an average change in count rate/time, or ? R/?T, where ?R=Rmax?Rmin and
?T is the corresponding change in time. These values are 2.5cts?1/2.4hr (MRK
352), 2.8cts?1/2.7hr (ESO 548-G081), and 2.4cts?1/0.8hr (UGC 6728). For both
MRK 352 and ESO 548-G081, these rates are minimums since the light curves
are decreasing/increasing monotonically. UGC 6728, however, shows a definite
maximum and is thus the most rapidly variable source, with count rate changing
appreciably over less than an hour.
Inordertocomparevariabilitybetweenobservations,ontimescalesofdaysto
months, we compared the XRT and XMM-Newton spectral fits listed in Tables 4.4
and 4.3. In Figure 4.6, we plotted the hard, 2 ? 10keV, (x) and soft, 0.5 ? 2keV,
(+) flux for multiple observations of our sources. We made the initial assumption
that any intrinsic differences in flux between the instruments is less than 10%.
From the figure, it is clear that variations greater than this level occurred for all of
the 16 sources with > 100counts in the XRT spectra. Of these, the most extreme
changes are seen for ESO 362-G018, where both the hard and soft flux drop by
an order of magnitude between the last two observations. However, without
error bars on the flux and with a simplified model that is not satisfactory for all
the sources, particularly for the high column density/complex spectra, a simple
comparison of the fluxes is only a starting point for our variability study.
In addition to the flux, both the power law index and hydrogen column den-
sityintroduceothersourcesofvariationmeasuredbythesimplepowerlawmodel.
Changes in these parameters also affect the measured flux. Thus, we computed
a statistic to quantify the flux variations between observations. To this end, we
determined the value(Fmax?Fmin)/Favg and the corresponding ?tmax (|tmax?tmin|
in days or the change in time for the greatest difference in observed flux between
two observations) for each source in both the hard and soft bands. These values
160
Figure4.6: Plotsofthefluxvariationinthesoft0.5?2keV(+)andhard2?10keV
(x) bands for sources with XMM-Newton and XRT observations (> 100counts).
These flux values were obtained from simple absorbed power law fits (see Ta-
bles 4.4 and 4.3). For the sources with few/no counts in the soft band, the soft
flux was unmeasurable.
161
arelisted inTable4.12and thedistributionsofthevaluesareplottedinFigure4.7.
From the histograms, there is no measured difference between the low column
density sources (simple sources) and the high column density (complex sources).
Thevaluesof(Fmax?Fmin)/Favg, however, aremuchsmallerinthehardbandthan
the soft. We note that the heavily absorbed sources, with much lower count rates
in the soft band, have much less accurate soft flux measurements as well as fewer
sources with observations > 100counts. For the low column density sources, we
find an average (Fmax?Fmin)/Favg value of 0.52 in the soft band and 0.37 in the
hard band.
Based onthe (Fmax?Fmin)/Favg values forour sources, our results indicatethat
the AGN spectra vary more in the soft band than the hard band. We tested this
by performing a Kolmogorov-Smirnov test on the soft and hard band values of
(Fmax?Fmin)/Favg, excluding the sources with values of 0 in the soft band. The
maximum difference between the cumulative distributions is 0.54 with a corre-
sponding probability of 0.028, suggesting that the distributions in the soft and
hard band differ. This claim was also made based on ASCA observations of
Seyfert 1 sources by Nandra et al. (1997). In addition to this result, we find that
the hard flux variability for the low absorption and more complex sources is sim-
ilar. Unfortunately, due to the lower number of counts in the XRT observations,
we can only compare half of the complex sources to the complete sample of low
column density sources.
In Figure 4.8, we plot the variability measure (Fmax?Fmin)/Favg versus the
change in time between the observations of Fmax and Fmin, ?tmax. From this plot,
it is clear that, as already stated, the variability measurement (Fmax?Fmin)/Favg
is smaller in the hard band than the soft band. However, there is no significant
difference in ?tmax for the total sample, with average values of 100.1 days for the
162
soft band and 81 days for the hard band.
As a next step in our analysis, we simultaneously fit the XMM-Newton pn
spectrum with all of the corresponding XRT observations for the sources listed in
Table 4.12. We began by fixing all XRT fit parameters to the best-fit XMM-Newton
pn power law (+ gaussian where there is a strong line and blackbody where it
was required) model. We accounted for absorption using the tbabs model for
low column sources and the pcfabs model for the heavily absorbed/complex
sources. Wethenallowedthefluxtovarybetweentheseobservationsbyaddinga
const model. Where the addition of this model significantly changed ?2 (??2 >
10) we flagged the source as having a varying flux. We then tested variability
in column density and power law photon index by allowing each of these pa-
rameters, along with their normalizations, to vary. Again, we noted significant
changes in ?2.
In order to measure the amount each model parameter varied between ob-
servations, we obtained error measurements for nH, ?, and the 0.3 ? 10keV
flux. We used the XSPEC model pegpwrlw in place of the pow model. The
pegpwrlw model is similar to the simple power law model, however, the pa-
rameters Emin and Emax are used to indicate the energy range for the power law
component. The normalization is then the flux from the pegged power law in
units of 10?12 ergs?1 cm?2. Since the normalization is a parameter in the model,
errors are easily computed for the flux with the XSPEC command err. For all
of the sources, we fixed Emin = 0.3keV and Emax = 10.0keV. We indicate vari-
ability based on our model fits in nH, ?, and flux (from the pegpwrlw model)
in Table 4.12. Details on the model fitting for the individual variable sources are
discussed in the appendix.
For the sources with low column densities, nH < 1023 cm?2, and simple spec-
163
Figure4.7: Distributionofvariabilitymeasurements,(Fmax?Fmin)/Favg,forallof
the sources listed in Table 4.12. The first histogram shows the values for the soft
(0.5 ? 2.0keV) band while the second set of histograms show the distribution in
the hard (2.0 ? 10.0keV) band. For the soft band, we show the distribution only
for the low columndensity/simple model sources. The more absorbed/complex
sources had fewer counts in the soft band, making the (Fmax?Fmin)/Favg unre-
liable for most of these sources. Further, only half of the absorbed sources had
XRTobservationswith> 100countsoverallbands,sothesampleisnotcomplete
even in the hard band. For the simple model sources, we find that the variability,
estimated from (Fmax?Fmin)/Favg, is higher for many of the sources in the soft
band than the hard band.
164
Figure 4.8: Plots of the flux variation in the soft 0.5 ? 2keV and hard 2 ? 10keV
bands, measured by (Fmax?Fmin)/Favg, versus number of days between obser-
vations for the maximum and minimum flux. In this figure, sources with X-ray
spectra best described by a simple model (triangles), complex model (circles),
and the two complex sources with nH < 1023 cm?2 (squares) are plotted. The
(Fmax?Fmin)/Favg values are lower in the hard band. In the soft band, three
complexsourcesarenotplottedduetouncertaintyinmeasuringtheirsoftfluxes.
The average value of ?tmax or |tmax?tmin| is about 100 days for both hard and
soft flux.
165
Table 4.12. Variability between XMM-Newton and XRT Observations
Source Soft var.1 ? tsoft2 Hard var.1 ? thard2 nH?3 ??3 Flux?3
MRK 352 0.92 125.7 0.49 125.7 yes no yes
SWIFT J0216.3+5128 0.39 16.5 0.27 14.7 no no yes
NGC 1142 0.00 151.9 0.79 159.4 no yes yes
SWIFT J0318.7+6828 0.59 65.5 0.61 59.4 no no yes
ESO 548-G081 0.35 49.7 0.36 49.7 no yes yes
ESO 362-G018 1.64 91.8 1.22 63.5 yes no yes
ESO 490-G026 0.49 101.7 0.35 4.8 yes yes yes
SWIFT J0904.3+5538 0.75 84.1 0.29 84.1 yes yes yes
MCG +04-22-042 0.44 129.1 0.28 129.1 yes no yes
MRK 417 0.00 8.3 0.35 8.3 no no no
UGC 6728 0.66 119.8 0.58 119.8 no no yes
SWIFT J1200.8+0650 2.00 188.4 0.13 9.7 no no no
ESO 506-G027 0.00 149.7 0.29 149.7 no no no
WKK 1263 0.29 35.0 0.36 35.0 no no yes
MCG +09-21-096 0.38 186.2 0.32 186.2 no yes yes
NGC 6860 0.92 97.7 0.16 97.7 yes yes yes
1Comparing soft and hard flux from the XMM-Newton and XRT observations listed
in Tables 4.4 and 4.3. As noted in the Variability section, we used the statistic (Fmax ?
Fmin)/Favg to compare the individual XMM-Newton and XRT fluxes in the soft (0.5 ?
2.0keV) and hard (2.0 ? 10.0keV) bands. For a few of the high column density sources
(nH > 1023 cm?2), thesoftbandfluxwasnotabletobemeasuredaccuratelyduetoalack
of counts. Therefore, the values in the Soft var. column are unreliable for these sources
(NGC 1142, MRK 417, SWIFT J1200.8+0650, and ESO 506-G027).
2The corresponding value tmax?tmin in days for the (Fmax?Fmin)/Favg values in each
band.
3Was there variability in each of the indicated parameters (nH, ?, Flux)? Details are
given in the appendix for each individual source.
166
tral shapes (Tables 4.5 and 4.6), all had at least one XRT observation to compare
with the XMM-Newton spectra. All of these sources showed some form of vari-
ability. As an example, Figure 4.9 shows the XRT and pn normalized observed
and unfolded spectra with the best-fit model for MRK 352. This source showed
more variability than any other low column density source. For MRK 352, the
XMM-Newton spectrum, taken five months prior to the XRT observations, shows
no absorption and is nearly double the flux level of the second XRT observation.
In the XRT observations, which are taken only a day apart, the flux changes by
40%. The column densities also change between these two observations, by ap-
proximately 30%, where the XRT columns are an order of magnitude higher than
the XMM-Newton observation?s measured column density. The observations for
the low column sources indicate variability in: flux for all of these sources (8),
column density for half, and power law index for 3 sources.
For the sources with hydrogen column densities higher than 1023 cm?2 (Ta-
ble 4.8), five of the AGN had XRT observations, all with less than 100counts,
while NGC 4992 had no XRT observations. Of the remaining six sources,
MRK 417, SWIFT J1200.8+0650, and ESO 506-G027 did not vary, in that, al-
lowing nH, the power law, and flux to vary yielded ??2 < 10. Using the
pegpwrlw model, the errors on the flux for each observation (XMM-Newton
and XRT) were within range of the other observations. For example, SWIFT
J1200.8+0650 showed an unabsorbed flux from the power law component rang-
ing from: 1.01 ? 1.25 ?10?11 ergs?1 cm?2 (XMM), 1.01 ? 1.29 ?10?11 ergs?1 cm?2
(XRT-1), and 1.04 ? 1.50 ?10?11 ergs?1 cm?2 (XRT-2) with ?2/dof= 235.7/241 for
the tbabs*pcfabs*(pegpwrlw)model.
The sources NGC 1142, SWIFT J0318.7+6828, ESO 362-G018, and NGC 6860
did show significant variability between the XMM-Newton and XRT observations
167
10?3
0.01
0.1
1
normalized counts s
?1
 keV
?1
10.5 2 50.5
1
1.5
2
2.5
ratio
Energy (keV)
10.5 2 5
10
?4
10
?3
0.01
keV
2  (Photons cm
?2
 s?1
 keV
?1
)
Energy (keV)
Figure 4.9: (left) XMM-Newton pn spectrum (black) with two XRT (red and
blue) observations for MRK 352 fit with the model tbabs*tbabs*(pegpwrlw
+ bbody). The best-fit model required both the hydrogen column density and
flux to vary between the three spectra. (right)The unfolded spectrum (E2 f(E) vs.
E, where f(E) is the model) is plotted for the same source, MRK 352.
(as detailed in the appendix). To summarize the variability, 3 showed no variabil-
ity, 4 had variable fluxes, 2 had varying column densities, and 3 showed varying
spectral indices. Given the complex shape of the spectra of ESO 362-G018 and
NGC 6860, we are uncertain of how to interpret the variability. We simply noted
the sources as varying under all of our criteria, but again note their complexity.
ESO 362-G018 showed the most variability of the objects in our study. We
conclude this section with a discussion of this source?s spectrum. In Figure 4.10a,
we plotted the observed spectra of this source. The shape of the spectra varied
considerably below 2keV for all observations. In the hard band, the XRT ob-
servations show no evidence of the Fe K line which is so prominent in the pn
spectrum. In Figure 4.10b, we plot the unfolded spectra. Here the y-axis cor-
168
0.01
0.1
1
normalized counts s
?1
 keV
?1
1 100.5 2 5
1
1.5
ratio
Energy (keV)
1 100.5 2 5
10
?4
10
?3
0.01
keV
2  (Photons cm
?2
 s?1
 keV
?1
)
Energy (keV)
Figure 4.10: (left) XMM-Newton pn spectrum (black) with two XRT
(red and blue) observations for ESO 362-G018 fit with the model
tbabs*pcfabs*(pegpwrlw + zgauss + zgauss). This source, with a
complex spectrum, varied considerably between the XRT observations, taken
approximately 2 months later, and XMM-Newton observation. The XMM-
Newton observation shows a strong Fe K line and a column density 10 times
that seen by the XRT observations. The flux is also lower by ? 50% in the
XMM-Newton observation (the spectra shown are normalized and exhibit the
observed spectrum). (right) The unfolded spectrum for ESO 362-G018 is plotted
(E2 f(E) vs. E). If the Fe K line remained at the same level as in the XMM-Newton
observation, the increased flux from the power law component would dominate
the line emission in the XRT observations. This is a possible explanation for the
appearance of the Fe K line. The observed spectrum is shown in Figure 4.9.
responds to E2 f(E). In this plot, we can see that the XRT spectra (taken about 2
months before the XMM-Newton observation) are much brighter. If the Fe K line
remained at the same flux level, it would be completely dominated by the power
law component. This is one possible explanation for the disappearance of the Fe
K line.
To summarize our findings, only a few sources (3/21) varied appreciably
169
on the ? 3hr time scale of the XMM-Newton observations. Those sources that
did vary were bright, X-ray sources with spectra well-fit by simple power law
models. From a comparison of the XRT and XMM-Newton spectra, taken a day
through months apart, it is clear that most of the sources vary on longer time
scales. In the extreme case of ESO 362-G018, the source varies drastically in flux,
column density, and overall shape in two months time. All of the low column
density sources varied in flux, while half showed evidence for varying column
densities. Unfortunately, given the lower count rates in comparable exposure
times, we have less data on the high column density sources. From a compar-
ison of the (Fmax?Fmin)/Favg distributions, they appear to vary similarly to low
column sources. However, higher quality data is necessary to draw firm conclu-
sions.
4.4 Discussion
In this study, we examined the X-ray properties of a sub-sample of BAT detected
AGN from the 9-month BAT catalog. These sources, selected based on their 14-
195keV flux, probable optical identifications with the Digital Sky Survey and
2MASS, and their lack of an archival X-ray spectrum, are probably representative
of the whole 9-month sample. Having a bright DSS or 2MASS counterpart does
not impart a significant selection effect, considering that all but 20 of the BAT
AGN 9-month sample fit this criteria. The 20 sources without a bright optical/IR
counterpart are mostly blazars, of which there are only 15 in the 9-month catalog.
Sinceblazarsareless than10%ofthesample, wedecidedtofocus onthemajority,
non-blazar sources. For the BAT AGN sample, Tueller et al. (2008) list 2MASS Ks
magnitudes for the sources with available values from NED. The average value
170
Figure 4.11: Hydrogen column density (cm?2) versus the 14-195keV flux
(ergs?1 cm?2) measured by SWIFT?s BAT instrument. These values are listed
in Tueller et al. (2008), with the circles representing 9-month catalog sources and
the squares sources from the 9-month catalog with XMM-Newton follow-ups de-
tailed in this paper. We note that for column densities higher than ? 1024 cm?2,
the spectrum is likely optically thick to Compton emission and thus there is a
greater uncertainty in the measured hydrogen column density.
for the BAT sample is 11.47 with a variance of 4.04. Our sub-sample is slightly
dimmer, with an average magnitude of 12.10 and a variance of 1.80, but well
within the distribution of BAT AGN magnitudes.
The column density distribution of our sub-sample, shown in Figure 4.11,
is representative of the larger sample. In the plot of hydrogen column density
versus 14-195keV flux (values from Tueller et al. (2008)), our 22 XMM-Newton
follow-up sources span the range of hydrogen column densities. Roughly half
171
of the sources have low column densities (nH < 1023 cm?2), while half are more
heavily absorbed. We find the same ratio of absorbed to non-absorbed sources
in our XMM-Newton follow-up sample. While the low absorption sources in our
sample span the lower range of hard X-ray fluxes (from BAT), this is expected,
since the sources were not previously studied in the X-ray regime. For the ab-
sorbed sources, however, we find that our sample spans the full range of BAT
X-ray fluxes.
Given that our sample of sources is a representative sample of the 9-month
BAT catalog, it is worthwhile to discuss the general properties of our sources. To
begin, the optical host galaxy classifications of our sources are listed in Table 4.1.
Examining the host galaxy classifications, 17/22 of the hosts are classified as spi-
rals or peculiar spirals. This result is interesting, considering that Grogin et al.
(2005)foundthehostsofX-rayselected, z? 0.4?1.3, ChandraDeep Fieldsources
to be dominated by ellipticals. If the BAT AGN hosts are predominately spi-
rals, as our sample suggests, this could imply an evolutionary effect in AGN host
galaxies between the z? 0.03 and z ? 0.4?1.3 universe.
4.4.1 Spectral Classification
From our detailed X-ray spectral fits, we found that 9 of the 22 sources had col-
umndensitiesbelow1023 cm?2 andspectrawell-fitbysimplepowerlaworpower
law with a soft excess models. Nearly half of these sources showed evidence of
having a soft excess. Optically, all of these sources except for SWIFT J0216.3+5128
and WKK 1263 are Seyfert 1 ? 1.2 sources. As discussed in the detailed spectral
fitting section, SWIFT J0216.3+5128 is most likely a blazar. Though, WKK 1263 is
identified as a Seyfert 2 in NED, no optical spectrum is available in the literature
to confirm this. It is possible that this source was simply misclassified.
172
The remaining sources in our study had more complex X-ray spectra. For
most of these sources, half of the sample, an absorbed power law component
model fit is unacceptable. This was particularly true for ESO 362-G018 and NGC
6860. The column densities and power law indices computed from the simple
model, for these sources, are drastically different than values from more complex
models. Both of these sources, optically Sy 1.5, had optical images indicating
dust clouds interspersed with star formation. Likely, the complex environment
contributed to the complexity seen in the X-ray observation. However, with-
out higher signal-to-noise observations we were unable to resolve the complex
spectral components. These results illustrate the danger of using low quality
data/simple models to determine the properties of complex sources.
The remaining 11 sources, half of the sample, had column densities clearly
above 1023 cm?2. We classified the observed spectra of these sources as having a
?doublepowerlaw?shape, similartoTurner et al.(1997). Optically,thesesources
are Seyfert 2s. The exceptions are NGC 612, a weak-lined, giant radio galaxy, and
NGC 4992, which have ?galaxy? spectra or optical spectra showing no emission
lines indicative of AGN emission. NGC 612 was specifically classified as a ?non-
LINER?, e.g. having non-AGN line ratios, by Lewis et al. (2003), while NGC 4992
is an INTEGRAL source whose optical spectrum led Masetti et al. (2006) to clas-
sify the source as an X-ray bright, optically normal galaxy. Considering the X-ray
column densities for these sources, the optical AGN emission could be hidden or
obscured by the high column of gas, nH > 5?1023 cm?2, in the line of sight.
ForallofthecomplexX-rayspectrasources, wefitthespectrawith(1)apartial
coveringabsorptionmodeland(2)adoublepowerlawmodel, where eachpower
law component had a separate absorption model. Most of the sources showed
no significant difference in ?2 between these two models. This makes it impos-
173
sible to determine whether the soft flux is the result of scattering of the AGN
light, partial covering of the AGN light, or other X-ray sources (such as X-ray
binaries or diffuse galaxy emission) contaminating the AGN spectrum. Particu-
larly for sources, such as NGC 612, with low 0.5 ? 2keV fluxes, e.g., F0.5?2keV =
2?10?14 ergs?1 cm?2 corresponding to a luminosity, L0.5?2keV ? 4?1036 ergs?1,
within the observed range of Galactic X-ray binaries, since the X-ray luminosity
function extends to ? 3?1038 ergs?1 (Grimm et al. 2002).
For sources with flat spectra (low photon index, ?), we fit the spectra with a
Compton thick model (a reflection dominated model, pexrav). We added the
BAT spectra to the XMM-Newton data in order to extend the energy range to
200keV. In addition to providing an adequate fit to the data, replacing the heav-
ily absorbed power law component with a reflection model resulted in steeper
spectral slopes more consistent with average AGN photon indices (where the
BAT spectra were not curved, see Table 4.10). Based on the detailed model fits,
NGC 612, ESO 362-G018, MRK 417, ESO 506-G027, and NGC 6860, are classi-
fied as Compton-thick candidates. Two of these sources, ESO 362-G018 and NGC
6860, havemeasured column densities nH << 1023 cm?2. While a column density
this low is not expected from reflection in a Compton-thick region, an alternate
model where the reflection component is scattered and then absorbed outside of
the Compton-thick region, such as employed for MRK 231 (Maloney & Reynolds
2000), could explain the spectra.
In addition to the Compton-thick candidates, four of the complex spectra
sources had high partial covering fractions (> 0.99) with the partial covering
model and very low ratios of the unabsorbed power law to the absorbed power
law component (N?1/N?2 < 0.02) with the double power law model. Thus,
emission from the sources SWIFT J0641.3+3257, SWIFT J0911.2+4533, SWIFT
174
J1200.8+0650, and NGC 4992, was extremely low in the soft band (0.5 ? 2keV)
compared to the hard band (2 ? 10keV). These sources are consistent with the
new class proposed by Ueda et al. (2007) of hidden or buried AGN. Ueda et al.
(2007) predict that these sources should have lower [O III] luminosities than typ-
ical Seyfert 2 sources. Archival optical spectra of NGC 4992 show very weak
[O III], in fact, so much so that, as discussed above, the spectrum of this source
appears as a typical galaxy. In our optical study (in preparation), we will ex-
plore this issue further. For now, it is important to note that more of these hidden
sources exist. If our sample, showing 4/22 hidden AGN, is representative of the
larger BAT sample, we expect that about 1/5th of local AGN have these same
properties, making them nearly undetectable in optical samples.
4.4.2 X-ray Colors
Having classified the sources into categories, we now describe the general prop-
erties of our sample as a whole. To begin, in Figure 4.12 we plotted the column
densities versus two different flux ratios (similar to a diagnostic in Malizia et al.
(2007)). The column densities we used are listed in Tables 4.5, 4.6, and 4.9,
where we used the column density of the more heavily absorbed power law
component for the complex spectra. The flux ratios plotted are the ratios of
F2?10keV /F14?195keV (medium/hard) and F0.5?2keV /F2?10keV (soft/medium). In
the plots, we represent the three classes of objects (simple power law or power
law and blackbody fit sources with low columns (triangles), complex heavily ab-
sorbed spectra (circles), andcomplex spectra thatrequire morecomplicated mod-
els (squares)). We find that the low absorption sources have average values of
medium/hard and soft/medium flux of 0.38 and 0.48, respectively. There is little
change between the medium/hard and soft/medium flux (? 20%) for the low
175
absorption sources. The heavily absorbed sources, however, have average values
of medium/hard and soft/medium flux of 0.08 and 0.02. This is a 75% change in
the values. Obviously, there is much less soft flux for the absorbed sources. The
complex sources with poorly defined spectral models, have intermediate values
of the medium/hard and soft/medium colors of 0.17 and 0.11.
In terms of use as a diagnostic, we find that the plot of column density ver-
sus ratio of F2?10keV /F14?195keV (medium/hard) is a good diagnostic of column
density, for nH < 1024 cm?2. Sources with similar column densities occupy areas
close to the regions of constant power law index plotted (for ? = 1.5 and 1.9).
This is not true for the plot of column density versus ratio of F0.5?2keV /F2?10keV
(soft/medium). This appears to be a poor diagnostic, despite its wide spread use
in deep X-ray surveys, with a large spread in the soft/medium color particularly
seen in the sources with nH > 1023 cm?2. These results are not surprising, since
the 2?10keV flux is also affected by absorption.
From these color diagrams, we decided to construct a color-color diagram of
the soft/medium flux ratio versus the hard/medium flux ratio in attempts to
construct a better diagnostic diagram for sources with too few counts to measure
columndensity. InFigure4.13, weplotthisdiagramusingthesamesymbolsasin
the previous diagram to indicate low absorption (triangle), complex (circle), and
more heavily absorbed (square) sources. In this figure, it is clear that the different
types of sources are clearly separated. The low absorption sources occupy the
left hand upper corner, where the soft/medium and hard/medium colors are
nearly equal. The heavily absorbed sources are closer to the right bottom corner,
where the hard/medium flux ratio is much higher than the soft/medium flux
ratio. Between these values, the complex sources as well as a source from each
of the other two categories, reside. All of these sources have measured column
176
Figure 4.12: Plots of nH vs. flux ratios for the XMM-Newton follow-up sources,
showing the ratio of the 2 ? 10keV to 14 ? 195kev (BAT) flux(top) and 0.5 ? 2keV
to 2 ? 10keV flux (bottom). The symbols represent the simple model sources
(triangles), complex spectra (circles), and complex spectra we could not interpret
(squares; ESO 362-G018 and NGC 6860). The unabsorbed sources are plotted as
having nH = 1020 cm?2. These sources had approximately the same ratio of hard
flux/BAT flux as soft flux/hard flux. For the more heavily absorbed sources, the
ratio of hard flux/BAT flux is clearly larger than the soft flux/hard flux ratio.
The lines represent column density vs. flux ratio for constant power law indices
(? = 1.9 and 1.5, as labeled).
177
Figure 4.13: Color-color diagram of soft/medium (0.5 ? 2keV / 2 ? 10keV flux)
andhard/medium(14?195kev(BAT)/2?10keVflux)colors. Theverticallines
represent values for constant power law indices (of ? = 1.9 and 1.5, as labeled)
with different absorbing columns. The other lines on the diagram represent con-
stant column densities for different power law indices (from top to bottom: 1020,
1021, 1022, and 1023 cm?2). The symbols represent the simple model sources (tri-
angles), complex spectra (circles), and complex spectra we could not interpret
(squares; ESO 362-G018 and NGC 6860). The unabsorbed sources clearly occupy
a region to the top left in the diagram while the more heavily absorbed sources
lie towards the bottom right. Sources with column densities in the middle (be-
tween 1022 and 1023 cm?2) lie between our unabsorbed and heavily absorbed
points. From this result, we present a new diagnostic to describe spectra with
low counts.
178
densities from 1022?1023 cm?2, intermediary between the two classes. This result
is very nice in that it provides a good diagnostic tool for observations with few
counts, but requires data above 15keV.
4.4.3 Properties of the Spectral Components
In addition to the flux and column density measurements, we have measured
power law indices, blackbody components (where present), and Fe K equivalent
widths (with the physical width of the line fixed to 0.01keV at 6.4keV). From our
sample, we found no correlation between the hard band (2 ? 10keV) luminosity
and power law indices. An important point to note is that the measured power
law index for the complex spectra depends very much on the model used. Com-
paringtheresultsofthepartialcoveringmodelwiththedoublepowerlawmodel
(Tables4.8and4.9), theaveragepowerlawindexforthepartialcoveringmodelis
significantlyhigher(< ? >= 1.74comparedto< ?2 >= 1.36)withsmallerassoci-
ated error bars. For the remaining 9 sources, the sources with spectra modeled by
absorbed simple power law or power law and blackbody models, < ? >= 1.75,
similar to the results from the partial covering model. The values for the sim-
ple model/low absorption sources and those from the partial covering model are
consistent with average photon indices for AGN (? 1.8 from Mushotzky (1982)).
Soft excesses, modeled with a blackbody component, were statistically signif-
icant in half of the spectra modeled by a simple power law model. We find a sig-
nificantly smaller fraction compared to the ROSAT sample of Gallo et al. (2007),
who find soft excesses in all of their sources, and a significantly larger fraction
compared to the Lockman Hole XMM-Newton survey (Mateos et al. 2005), where
only 11% of type 1 and 25% of type 2 sources show a soft excess. Due to the low
number of counts for our heavily absorbed sources, we can not quantify with cer-
179
tainty how many complex/heavily absorbed sources have this component, but at
least one source (NGC 1142) has a statistically significant soft excess. For the low
absorption sources, < kT >= 0.08keV, which is similar to but slightly lower than
that seen for PG selected QSOs (< kTBB >= 0.14 ? 0.02keV) (Piconcelli et al.
2005; Porquet et al. 2004b). If the soft excess is the result of a thermal process, the
lower kT values in our sample could be related to the lower luminosities of our
sample, compared to the PG QSOs. In fact, our average blackbody temperature is
directly in the range of those found for type 1 AGN in the Lockman Hole sample,
< kT >= 0.09?0.01keV (Mateos et al. 2005).
The final spectral component measured for our entire sample is the Fe K
equivalent width (EW) at 6.4keV. In Figure 4.14, we plot the Fe K EW versus the
hard band (2 ? 10keV) luminosity. We fit a line with the ordinary least squares
bisector method to the upper limits of the EW measurements (see plot), yielding
a fitof logEW = (?0.697?0.144)?logL2?10kev+(32.045?6.164). The significance
of this fit, indicated by R2 = 0.22, where R2 is the coefficient of determination, is
very low. Thus, our results show no indication of the X-ray Baldwin or IT ef-
fect (Iwasawa & Taniguchi 1993), an anti-correlation of Fe K EW and hard band
luminosity (see the following chapter).
For the low absorption/simple model sources, we measured the significance
of the O VII and O VIII K edges to search for evidence of a warm absorber (see
Table 4.7). All of the simple model sources, with the possible exception of WKK
1263, are classified optically as Seyfert 1s and thus can be directly compared with
theReynolds(1997)sample. InFigure4.15, weplotthevaluesofopticaldepthfor
each of the edges versus L2?10keV for our sources as well as the Reynolds (1997)
sources with luminosities in the same range. As the figure shows, the optical
depths we found for our sources are much lower than those from the Reynolds
180
Figure 4.14: Plot of the Fe K equivalent width measurements (eV) versus the
luminosity in the hard band (2 ? 10keV). The equivalent width measurements
are from the best-fit models for the sources shown in Tables 4.5, 4.6, and 4.9.
The simple power law/ power law and blackbody, unabsorbed/low absorption
sourcesare plotted astriangles. The squaresrepresentthe two sourceswith com-
plex spectra we could not interpret (ESO 362-G018 and NGC 6860). Finally, the
circles represent the absorbed sources with complex spectra. The source SWIFT
J0216.3+5128 is not plotted since there was no evidence of a line and the redshift
is unknown. The line is the ordinary least squares bisector fit to the data using
the upper limits on the Fe K equivalent widths.
181
sample. Only one source, ESO 490-G026, had a clear detection (??2 = 25) with
the optical depths of both edges having upper limits above 0.10. Thus, where
half of the Reynolds sample and 13/18 of the George et al. (1998) sample showed
evidence of a warm absorber in the line of sight, we find only 1/8 of our Seyfert 1
sources toshowsignificantevidence ofawarmabsorber. Thisresult couldbe due
to an incomplete or biased sample of Seyfert 1 sources, since the sources in this
study were among the low absorption sources with the lowest BAT flux in the
9-month sample. Alternatively, our result could be representative of the entire
BAT sample. In this case, it is possible that the previous AGN samples showed
more detections because they were from an optically selected/soft X-ray selected
sample. Thus, a possibility is that the emission that ionizes the gas, creates a
region of warm, ionized gas, that also destroys dust. An optical or softer X-ray
survey could preferentially select these sources, missing more obscured sources.
Analysis of the remaining BAT sources will allow us to verify whether our result
of few warm absorbers is consistent with the properties of the entire 9-month
catalog.
4.4.4 Variability
In addition to the spectral properties, we examined our sources for two types of
variability: (1) during the XMM-Newton observations and (2) variability in spec-
tral form and brightness between the XMM-Newton and XRT observations. For
the first type of variability, we created binned light curves for each object in our
sample. We found that 3/21 sources showed significant variability, with rates
varying by 2.4 ? 2.8cts?1 over time scales of 0.8 ? 2.7hr, in the pn observations.
The sources that varied the most were all Seyfert 1 sources with low absorption
and X-ray spectra well-fit by simple power law or power law and a blackbody
182
Figure 4.15: Plot of the optical depth of an added O VII, 0.74keV, and O VIII,
0.87keV, K edge vs. 2 ? 10keV luminosity. This model was added for the low
absorption/simple model sources(circles, solid lines forupper limits), whichare
optical Seyfert 1 sources. We compare our values to those from Reynolds (1997)
(squares, dashed lines for upper limits). We find much weaker optical depths
among our sample, most noticeably for O VIII where the upper limits are well
below ? = 0.1 for all but one source. Further, in Table 4.7, the addition of the two
edge models gives a statistically better fit (??2 > 10) for only two sources (ESO
490-G026 and MCG +04-22-042). Only ESO 490-G026, has upper limits for both
O VII and O VIII with ? > 0.1.
183
models. These sources were among the brightest in our sample. While only
three sources showed short term variability, during the XMM-Newton observa-
tion which lasted ? 10ks, nearly all of the sources (13/16) exhibited variability
on longer time scales, of hours to months, from comparisons of the XMM-Newton
and XRT observations. This result agrees with earlier studies which found AGN
more variable on long time scales than short time scales (Barr & Mushotzky 1986;
Nandra et al. 1997).
From our comparison of the XMM-Newton and XRT spectra, we found 13/16
sources had varyingfluxes, 6/16had varyingcolumn densities, and 6/16sources
had varying power law indices. Sources tended to vary more in the soft band
than the hard band (see the Variability section). Unfortunately, due to lower
count rates, measuring variability for the heavily absorbed sources was more un-
certain, particularly in the soft band (0.5 ? 2keV). In addition, we had fewer XRT
observations with > 100counts for these sources. Based on the result that our
sources varied more in the soft than hard band, it is likely that if we had more
counts in the soft band for the heavily absorbed sources, as well as more observa-
tions for comparison, our results would agree with other AGN variability studies
which found> 90%of their sources to varyover the timescale ofmonths to years
(such as theAGN fromtheLockmanHole(Mateos et al.2007)andChandraDeep
Fields (Bauer et al. 2003b; Paolillo et al. 2004) studies).
In Figure 4.16, we plot column density versus flux and photon index versus
flux forthesources thatvary. Theplotsshowtheobserved parameterfor eachob-
servation/average parameter for the source, where the observed values for each
source are plotted with a different symbol. In the column density figure, there is
clearly no correlation seen between the column density and flux. Similar results
were found by Risaliti et al. (2002) for a sample of Seyfert 2 galaxies, where they
184
Figure 4.16: Plots of varying column density (top) and photon index (bottom)
withflux. Forsourcesthatshowedvariationsincolumndensityorphotonindex,
we plotted the observation?s value divided by the average for the source (i.e.
Fi/ <Fsource), for each individual observation. The flux is the 0.3 ? 10keV flux
from the pegged power law component. The lines mark the area where each
parameter is 1.0 (where the observation value is the average value). A different
symbol is used for each source. No correlation is seen between column density
and flux, however, there is a strong correlation between photon index and flux.
185
conclude that the variations in column density are not caused by varying ioniza-
tion states but by a clumpy absorber. In the plot of spectral indices, however,
we do find a correlation between the spectral index and the flux. Therefore, we
find that higher fluxes correspond to higher spectral indices. This result has been
seen before for individual sources (Mushotzky et al. 1993). Based on a variability
study of the AGN sources in the Lockman Hole, Mateos et al. (2007) find no cor-
relation between spectral variability and flux variability. Further, they find flux
variability much more prevalent in their sample than spectral variability, find-
ing spectral variability in only 14 ? 8% of Seyfert 1s and 34 ? 14% of Seyfert 2s.
However, Mateos et al. (2007) note that the detection of spectral variability is re-
latedtothequalityofthespectrum. Whentheyconsider thisfactor, they predicta
higherfractionof? 40%toexhibitspectralvariability. Wedonothaveacomplete
sample for Seyfert 2 sources, due to the low number of counts in the XRT obser-
vations for heavily absorbed sources. However, we find 7/8 classified Seyfert 1
sources to exhibit a variation in either column density or power law index, much
higher than the Mateos et al. (2007) estimated value. Additionally, as we stated
earlier, there is a clear correlation between changing flux and power law index
for individual sources.
For the Compton-thick sources, variability or a lack thereof, gives clues to the
size and location of the Compton-thick gas. For our heavily obscured Compton-
thick candidates, only MRK 417 had enough counts in an XRT observation (>
100counts) to test for long term variability. We found no statistically significant
evidence of variability for this source between the XMM-Newton and XRT obser-
vations, taken 6 months apart. This lack of variability in two observations does
notgiveusmuchinformation. ForthelowabsorptionCompton-thickcandidates,
however, we find a great deal of variability between the XMM-Newton and XRT
186
observations. In particular, for NGC 6860, the flux and spectral index are higher
in the XRT observations, while the column density is lower. Since the XRT obser-
vation was taken 4 months earlier, this puts a limit on the suggested change from
Compton-thin to a reflection-dominated spectrum. Similarly, significant changes
are seen between the XRT and XMM-Newton observations of ESO 362-G018. In
this source, the most significant change is the disappearance of the strong Fe K
line seen in the XMM-Newton observation. A smaller time constraint is placed on
this source, 2 months between the last XRT and the XMM-Newton observation,
for a change from a Compton-thin to a reflection-dominated spectrum.
Changes from Compton-thin to Compton-thick spectra have been noted be-
fore, particularly by Matt et al. (2003). They discuss two possible scenarios to
explain the changes in spectra, a change in column density of the absorber
(Risaliti et al. 2002) and a ?switched-off? source, that is, a state where the emis-
sion from the central source drastically decreases below our detection threshold.
For both ESO 362-G018 and NGC 6860, Seyfert 1.5 sources embedded in dusty
host galaxies, a changing absorber is a more appealing explanation.
4.5 Summary
From our analysis of the XMM-Newton and XRT spectra of 22 BAT-selected AGN,
the complexity of the spectra of a large fraction of nearby (< z >? 0.03) AGN is
clear. Based on the range of X-ray column densities and BAT (14 ? 195keV) fluxes
(Tueller et al. 2008), our sources are a representative sample of the 9-month BAT
catalog. In analyzing their properties, we are presenting for the first time the
global X-ray properties of an unbiased, local AGN sample.
Within our sample, we find half of the sources to have low absorption (nH <
187
1023 cm?2) and spectra well-described by simple power law models. Half of these
sourcesstatisticallyshowevidenceofasoftexcess. Wetestedthesesourcesforthe
presence of a warm absorber, finding only one statistically significant detection
out of 8 low absorption sources. This is at odds with the studies of George et al.
(1998); Reynolds (1997) who found half or more of their samples consistent with
warmabsorbers. Ifour resultoffewwarmabsorbersisfound intheentire Seyfert
1 BAT AGN sample, the detection of a large number of warm absorbers is likely
a selection effect of optical/soft X-ray AGN samples.
The remaining 13 sources, which had too few soft counts to test for the pres-
ence of a warm absorber, have more complex spectra. Within the class of com-
plex sources, we find five Compton-thick candidates (based on a flat spectrum
above2keV),twoofthesesourceswithspectratoocomplextomodelsuccessfully
with the available signal-to-noise. Additionally, four other sources are consistent
with the hidden/buried AGN described in Ueda et al. (2007). Since ? 1/5 of our
sample fits in this category, we agree with Ueda et al. (2007) that these types of
sources are a significant fraction of local AGN. If these sources have weak [O III]
emission, as Ueda et al. (2007) predicts, they would be easily missed in optical
surveys and require very hard X-ray surveys, such as the BAT and Integral sur-
veys, for detection.
On short time scales, during the ? 3hr XMM-Newton observations, we found
that only 3/21 sources varied significantly, all of which were bright, low absorp-
tion X-ray sources. Comparing the XRT and XMM-Newton observations of 16/22
sources, which were separated by hours to months, we were able to compare
the spectra for longer time scale variability. Most of the sources varied in flux
(13/16), such that our results agree with previous studies which found AGN to
vary more on longer time scales than short time scales (Barr & Mushotzky 1986;
188
Nandra et al. 1997). In terms of spectral variability, nearly half of the sources var-
ied in both column density (6/16) and power law index (6/16). We found no
correlation between column density and flux between observations for the in-
dividual sources. However, there was a strong correlation between power law
index and 0.3 ? 10keV flux, where steeper slopes correspond to higher fluxes.
Contrary to the variability study by Mateos et al. (2007) who estimate ? 40% of
theirsampletovarywithrespecttospectralshapes, wefind7/8identifiedSeyfert
1s to vary in either column density or power law index. We note, as Mateos et al.
(2007) point out, that the detection of this variability depends on the quality of
the data. Thus, similar comparisons with Seyfert 2s were not plausible since the
data quality was much lower.
Optically, the Seyfert type of the sources match the X-ray column densities.
Thus, the Seyfert 1 sources have nH < 1022 cm?2 and the Seyfert 2 sources have
nH > 1022 cm?2. The two sources with no optical AGN signatures are heavily
absorbed sources with nH > 5 ? 1023 cm?2. The host galaxies of our sample
are mostly spirals, contrasting with the results of Grogin et al. (2005), who find
elliptical hosts dominating the z ? 0.4?1.3 universe.
189
Chapter 5
The Complete X-ray Properties of the BAT
AGN
5.1 Introduction
In Chapter 4, we analyzed the X-ray properties of a sub-sample of 22 AGN from
the BAT AGN catalog. Among our results, we found that 20% of our sample
had properties consistent with hidden/buried AGN, a class of object first identi-
fied through BAT and Suzakuobservations (Ueda et al. 2007). In this Chapter, we
present the results of an analysis of the X-ray properties of the entire 9-month cat-
alog. We provide the compiled properties from archival data, analyses presented
in the literature, and previously unanalyzed SWIFT XRT observations. Thus, for
the first time we show the collective properties of a 14?195keV band AGN sam-
ple ? probing energies high enough that there is no selection effect for absorbing
columns up to ? 1024 cm?2.
Our goals in this chapter are two-fold. First, we present the X-ray proper-
ties for the entire 9-month catalog (Tueller et al. 2008). Second, we examine in
more detail the collective properties of a uniform sub-sample. This sample con-
190
sists of non-beamed sources with Galactic latitudes ? 15?. In ? 5.2, we describe
the observations and the spectral fits, including data from ASCA, XMM-Newton,
Chandra, Suzaku, and SWIFT XRT. In ? 5.3, we describe the general properties
of the entire BAT 9-month AGN sample. In ? 5.4, we describe in more depth the
properties of our uniform sample. In ? 5.5, we present the average X-ray spec-
trum as well as the 2?10keV logN-logS relation. These X-ray properties can now
be used as input to X-ray background models for z ? 0. We then discuss the
properties of the host galaxies in ? 5.6. Finally, we summarize our results in ? 5.7.
5.2 The Data
5.2.1 Source Selection
The sources in the 9-month BAT AGN catalog were selected based on a detec-
tion at 4.8? or higher. This corresponds to fluxes in the 14?195keV band of
? 2 ? 10?11ergs?1 cm?2. The survey and the method of selection are described
in (Tueller et al. 2008). To summarize, the BAT positions for the detected sources
have an error of ? 6?at 4.8?. Therefore, the positions of identified sources were
compared with available optical, radio, and X-ray observations. Where the BAT
source could not be identified with a previously known AGN source, SWIFT
XRT observations were obtained. With an identified XRT counterpart, the po-
sition is narrowed to an error of ? 4??. The source identification was further
constrained by the requirement that each of the BAT AGN sources have a clear
optical/IR counterpart in the Digital Sky Survey/2MASS. Many of the sources
in this entire 9-month sample have identifications based on optical spectra ei-
ther from archived data, the literature, or our own follow-up data. The sample
consists of 17 blazars/BL Lacs, 49 Sy 1?1.2s, 34 Sy 1.5?1.9s, and 45 Sy 2s. Of
191
the remaining sources, 7 are unidentified, 2 have optical spectra showing a nor-
mal galaxy (i.e. no AGN emission lines; NGC 612 and NGC 4992), and 2MASX
J09180027+0425066 is identified as a type 2 quasar.
In order to study the properties of the local BAT-detected AGN, we identify a
uniform sample from the 9-month catalog. This requires the exclusion of beamed
sources aswellassources withintheGalacticplane(b < 15?). Weexcludebeamed
sources (the 17 blazars/BL Lacs), since they make up only ? 10% of AGN and
are at higher redshifts than our sample of local sources. Further, the physics
behind the spectra are different between beamed and unbeamed sources, since
these sources are jet dominated. Only high Galactic latitude sources are included
(b > 15?), since the identification of AGN in the Galactic plane by BAT are less
certain due to the higher background and large number of unidentified sources
(Galactic and extragalactic). This uniform sample consists of 102 sources. Among
these, 34areSy1?1.2s,28areSy1.5?1.9s,and36areSy2s,withnoneofthesources
without an identification. Since the normal galaxy spectra sources and the type
2 quasar show no broad lines, we include them as Sy2s in the following discus-
sions.
Since the 9-month survey sources are moderately bright, many of them were
well-known AGN sources with archival data/published papers detailing the X-
ray properties. Thus, in compiling the X-ray properties of this sample, we first
searched the literature for analyses of the X-ray spectra of our sources. In the
following section, we describe the X-ray data and analysis for the entire sample.
192
5.2.2 X-ray Data and Analysis
AsdiscussedinChapter1,theX-rayemissionfromAGNprimarilytakestheform
of an absorbed power law (Mushotzky et al. 1993). Additionalfeatures which are
present in the 0.3?10keV spectra of many AGN include a soft excess and an Fe
K-? line. The soft excess is often modeled as a blackbody component (kT ?
0.1keV) and its origin is believed to be either thermal emission from star for-
mation, photoionized gas, blurred reflection (Czerny et al. 2003; Ross & Fabian
2005), or blurred absorption (Gierli?nski & Done 2004). The Fe K-? line is a fluo-
rescentlinefromlowlyionizedironat6.41keV.Finally,manyAGNexhibitacom-
plex spectrum indicating emission components absorbed with different column
densities. This type of spectrum may be the result of scattering of direct AGN
emission, a dusty environment where the AGN emission is partially covered by
absorbing material, or contamination of the AGN spectrum by less-absorbed X-
ray binaries. In our effort to uniformly compare the properties of our sources,
we searched the literature for simple models fit to the X-ray data, based on these
components (power law, absorption (simple or complex), soft excess, and Fe K
line). We did not include models for reflection since higher signal-to-noise spec-
tra would be required for the entire sample.
In our study of 22 XMM-Newton follow-ups of BAT AGN (Chapter 4), we had
used the same components to classify the AGN X-ray spectra. We chose a sim-
ple and complex model that is well-used throughout the literature, allowing for
easy comparison with other AGN studies. In this study, we adopt these same
models. Our simple model is an absorbed power law model (Figure 5.1a), with
an additional soft blackbody and/or gaussian (for the Fe-K? line) model where
required. Using the standard X-ray software for fitting X-ray spectral models,
193
XSPEC (Arnaud 1996), our simple model is represented as tbabs*(pow + bbody
+ zgauss). Here, tbabs is a standard neutral absorption model (Wilms et al.
2000). We categorize a complex spectrum as one well-fitted with either a partial
covering model (Figure 5.1b) or a double power law model, both of which give
similar results (in ?2 and spectral parameters), along with an Fe-K? line. Our
complex/partialcoveringmodelisformallyimplementedinXSPECaspcfabs*(pow
+ zgauss) for partial covering or tbabs*(pow + tbabs*(pow + zgauss)) for a
double power law model. Further discussions of our choice of these models can
be found in Chapter 4.
For all of the 9-month BAT AGN, we list the source, position, optical type,
host galaxytype, and detailsof the observationin TableE.1. We include the X-ray
satellite used, references, exposure time, and count rate, where available. Where
count rates and exposure times are quoted, they correspond to the pn detector
for XMM-Newton and SIS0 for ASCA. For many sources, spectra were available
frommanydifferent satellites. Our preference wastochoose ASCAanalysesfirst,
followed by XMM-Newton and Chandra. Mostly, this is due to the uniform way
thespectralpropertiesofSeyfert1sandSeyfert2sarepresented inpapersanalyz-
ing ASCAdata, such as Nandra et al. (1997); Reynolds (1997); Turner et al. (1997),
and the availability of a uniform database (Tartarus1). Also, especially with the
increased resolution of the grating spectrometers and higher signal-to-noise CCD
data on XMM-Newton and Chandra, more accurate/complicated models are of-
ten used to analyze spectra from these satellites, particularly for observations
with a large number of counts. Where multiple observations were available, we
chose the longest observation below 50ks, to be comparable to the quality of
spectra collected for other sources.
1http://astro.ic.ac.uk/Research/Tartarus/
194
Wherereferenceswerenotavailable,orcomparablemodelswerenotused, we
analyzed either SWIFT XRT, ASCA (downloaded from the Tartarus database), or
Suzaku spectra (we obtained data for ? 10 sources through AO-1/AO-2 propos-
als). For the XRT spectra, we analyzed these data following the same procedure
as in Chapter 4. In XRT observations with few counts, if there were multiple ob-
servationsavailable, we combined thespectra fromthe longest observationswith
the FTOOL mathpha. The analysis of the Suzaku spectra is contained in Ueda
et al. (in prep), which includes detailed analyses of the spectra. The observation
and source properties for the SWIFT XRT, ASCA, and Suzaku observed sources
are also listed in Table E.1.
In Tables E.2 and E.3, we list the spectral properties from the observations.
Table E.2 includes our XSPEC fits to the XRT data using an absorbed power law
(tbabs*pow) or an absorbed partial covering model (pcfabs*pow). For each
of these fits, an additional neutral absorber (tbabs) was fixed to the Galactic
column density. For spectra with clear residuals from these models indicating
a soft component or an Fe K-? line, we added the necessary components and
included the details in the appendix. Table E.3 includes the compiled properties
from the literature as well as our own fits to ASCA spectra downloaded from
the Tartarus database. For these spectra, the model used is clearly specified in
the table. In both tables, we categorize the spectra as being simple (S: well-fit
by an absorbed power law or power law and blackbody) or complex (C), as in
Chapter 4. In Figure 5.2, we plot some examples of the XRT spectra from each
category. Notice that the spectrum of UGC 11871 has no data points below 1keV.
For this source, as well as 3 other sources in Table E.2, we classify the source as
S*. This indicates that it was best fit with a simple power law, however, with
longer exposure times, and thus more data points below 1keV, it would likely be
195
a complex source, showing the characteristic complex/double power law shape.
We have collected the X-ray properties for 145/154 AGN. Though there are
153 BAT sources, one BAT source (NGC 6921/MCG +04-48-002) is the combi-
nation of two interacting galaxies, each hosting AGN (Winter et al. 2008). In the
appendix, weinclude detailsoneach ofthe 9sources excluded from thisanalysis.
The sources are excluded due to a lack of data available (2) or complex spectra
(7). For instance, Cen A, an AGN hosted in a merger galaxy, is a complex source
that is excluded. In the following sections, we will describe the X-ray properties
for the remaining 145 sources.
5.3 Properties of the Spectra
In Tables E.1? E.3, we provide information on the 9-month BAT AGN sources,
including the X-ray spectral parameters and fluxes, all derived in a uniform way.
InordertostudythepropertiesofthelocalBAT-detectedAGN,weneedtolookat
auniformsample. Inthefollowingsections, wewillconcentrateontheproperties
of our uniform sample. However, it is first worth noting some of the general
properties of the sources detected in the 9 month catalog.
As a means to summarize these properties, we plot a color-color plot of
F0.5?2keV /F2?10keV vs. F14?195keV /F2?10keV (all observed fluxes), first used in Chap-
ter 4. This plot includes all sources excepting 3 AGN (the unabsorbed sources
ESO 416-G002, MCG -01-13-025, and Mrk 79) which had only broad band fluxes
available, the 4 sources which had no high-quality or publicly available X-ray
spectra, and the 4 very complex spectra sources (NGC 1275, Cen A, NGC 6240,
andGRS 1734-292). AsinChapter4, theunabsorbed/lowabsorptionsources (tri-
angles, nH ? 1022 cm?2)occupy theupperleftportion, withoneexception. CygA
196
1 100.5 2 510
?5
10
?4
10
?3
0.01
keV/cm
2  s keV
Energy
Simple Model
1 100.5 2 510
?5
10
?4
10
?3
0.01
keV/cm
2  s keV
Energy
Complex Model
Figure 5.1: The above figures are examples of the complex and simple models
used in this chapter. The simple model (left) is an absorbed power law model
(NH = 8 ? 1020 cm?2 and ? = 1.75). In addition to the power law, we include
a blackbody component (kT = 0.1keV, dashed line) and a 60EW Fe K-? line
at 6.41keV. The complex model (right) is an absorbed partial covering model
(or a double power law model, which gives similar results). Here, the same
power law and gaussian components are used as for the simple model, with
NH = 1?1023 cm?2 and a covering fraction of 0.95.
is a complex source with a complex spectrum, the product of an AGN in a recent
merger galaxy. A strong thermal component, mainly from the hot thermal emis-
sion of the rich galaxy cluster Cyg A inhabits, is seen at low energies, modeled by
a Raymond-Smith thermal model (Sambruna et al. 1999). This thermal emission
gives Cyg A its unique position on the diagram.
Forthe unabsorbed sources, the soft band mean (?) color is 0.55with standard
deviation (?) of 0.39. The hard color is much higher, with ? = 6.67 and ? =
17.50. Many of the heavily absorbed sources (circles, nH > 1023 cm?2) are located
197
10
?3
0.01
5?10
?4
2?10
?3
5?10
?3
normalized counts/sec/keV
NGC 235A
1 100.5 2 5
0.5
1
1.5
2
ratio
channel energy (keV)
10
?3
0.01
2?10
?3
5?10
?3
0.02
normalized counts/sec/keV
2MASX J04440903+2813003
1 100.5 2 5
0.8
1
1.2
ratio
channel energy (keV)
0.01
2?10
?3
5?10
?3
0.02
normalized counts/sec/keV
UGC 11871
1 100.5 2 50.5
1
1.5
ratio
channel energy (keV)
0.01
0.1
normalized counts/sec/keV
LEDA 138501
1 100.5 2 5
0.5
1
1.5
2
ratio
channel energy (keV)
10
?3
0.01
0.1
normalized counts/sec/keV
1RXS J112716.6+190914
1 100.5 2 50.5
1
1.5
2
ratio
channel energy (keV)
10
?3
0.01
0.1
normalized counts/sec/keV
2MASX J11454045?1827149
1 100.5 2 5
1
1.5
ratio
channel energy (keV)
Figure 5.2: Examples of XRT spectra for sources fit with simple and complex
models, along with residuals. The spectra of 2MASX J04440903+2813003 and
2MASX J11454045-1827149 were rebinned to better show the residuals from the
model. Details of each of the fits are recorded in Table E.2. The sources in the top
row are all well-fit with the complex model, while the lower row sources have
simple spectra (see Figure 5.1).
198
Figure 5.3: We plot all of the 9-month BAT AGN with soft and hard fluxes on the
color-color plot initially presented in Chapter 4. The triangles represent sources
with nH < 1022 cm?2, while the circles represent sources with nH > 1023 cm?2.
Squares indicate the intermediary hydrogen column sources. Finally, diamonds
are used to represent the 17 blazar/BLLac sources, whichall havelow measured
column densities. In the text, we describe Cyg A (the circle labeled in the plot)
and5othersources(NGC1365andNGC5728(circles),Mrk3(square),andNGC
4945 and NGC 6814 (triangles), within the box) to have unusual positions.
towards the lower right corner. These sources clearly have much less flux in
the soft band (with the exception of Cyg A). The mean and standard deviations
are: ? = 0.06, ? = 0.13 (soft) and ? = 20.43, ? = 16.16 (hard). Finally, the
squares represent the intermediary sources, with 1022 cm?2 < nH ? 1023 cm?2.
These sources have colors clearly in the middle, with soft colors (? = 0.07, ? =
0.07) similar to the more absorbed sources and hard colors (? = 6.68, ? = 8.48)
closer to the less absorbed sources.
199
In our color-color plot, there are some clear outliers towards the upper right.
The high column density sources include the complex ?changing-look? AGN,
NGC 1365, and NGC 5728, a strongly barred galaxy whose host is likely the main
contributer to the soft thermal emission. The remaining sources include Mrk 3
(square) and both NGC 4945 and NGC 6814 (triangles). Among these, Mrk 3 and
NGC 4945 have complex spectra. NGC 6814 is well fit by an unabsorbed power
law but has a strong Fe-K? line (EW= 545eV). There is very little flux in the
ASCA observation below 1keV. The strong emission in Fe K? and weak flux at
soft energies contribute to give it an unusual location in the plot. In short, all the
sources at this location have complex or atypical spectra. However, another fac-
tor which can lead to unusual positions in the color-color plot is time variability.
This variability could be over the softer X-ray band (< 10keV) or in the BAT ob-
servations. The complex source NGC 6814, for instance, is known to show X-ray
variability by at least 10 times over time scales of years (Mukai et al. 2003). In the
14?195keV band, the corresponding BAT fluxes are time averaged over months.
Therefore, sources that are variable in the X-ray or BAT bands may have unusual
hard colors. Analysis of the X-ray light curves, however, is beyond the scope of
this thesis.
We next summarize the X-ray spectral shape of the BAT AGNs. In Tables E.2
and E.3, we included a column with the type of model used (simple or complex).
Excluding the 17 blazars, 63 sources are best fit with a simple model while 65
require a more complex model. As stated earlier, our complex model referred
to a better fit (over the simple power law model) with either a partial covering
or double power law model. Among the complex sources, 5 corresponded to
low column density sources (nH ? 1022 cm?2), 20 to sources with intermediate
columns (1022 cm?2 < nH ? 1023 cm?2), and 40 to the heavily absorbed sources
200
(nH > 1023 cm?2). All of the heavily absorbed sources were better fit with a com-
plex model. This shows that our color-color plot is particularly good for selecting
complex sources, which lie to the right of the constant ? = 1.5 line (indicating
colors for a fixed spectral index with an absorbed power law model).
Amongthecomplexabsorbedsources, weclassify28AGNas?hidden?/buried
sources. ?Hidden? AGN were first distinguished as a new class discovered by
BATinUeda et al.(2007). Thesesources areextremely obscured, possiblybyage-
ometrically thick torus. Since the hidden sources were identified in the X-ray and
only two sources were identified in the discovery paper, the multi-wavelength
properties are as of yet not fully classified. The X-ray spectra are character-
ized by a very small scattering component (? 3%) in the soft band. In the par-
tial covering model, the model is a multiplicative model defined as M(E) =
f ? e?nH?(E) + (1 ? f). Here, f is the covering fraction, nH is the neutral hydro-
gen column density, and ?(E) is the photo-electric cross-section. The scattering
fraction is then the value (1 - f). For our spectral fits, the partial covering model
is applied to a simple power law spectrum (pcfabs*pow in XSPEC).
We adopted the criteria that a hidden source is one where the scattering com-
ponent is ? 0.03 and the ratio of soft (0.5 ? 2keV) to hard (2 ? 10keV) flux is
? 0.04. Of the 28 identified, 9 are from Chapter 4. The additional sources include
5 sources with ASCA spectra (NGC 788, NGC 3081, NGC 3281, NGC 4507, and
IC 5063) and 4 with Suzaku spectra (ESO 297-018, ESO 005-G004, and 2MASX
J09180027+0425066, and 3C 105). The remaining 10 sources have XRT spectra
(listed in Table E.2) which are identified as complex (C) or as well-fit with a sim-
ple model but with little/no soft flux (S*). The median column density of the
hidden sources is lognH = 23.57 and the median soft/hard flux is 0.013, consis-
tent with the properties of the previously identified hidden sources.
201
5.4 The ?Uniform? Sample
Having summarized the general properties of the entire 9-month BAT AGN sam-
ple, we now will focus on a uniform sample. Our uniform sample includes 102
Seyferts at |b| ? 15?, with 46 simple power law model sources and 56 complex
model (partial covering or double power law model) sources. In this section, we
present general properties of this uniform sample including the distribution of
nH, ?, Lcorr2?10keV , and Lcorr2?10keV /LEdd. We also discuss the existence of correlations
between these properties, particularly with the Fe-K? equivalent width. Finally,
we discuss the properties of our classified simple and complex sources in more
detail.
5.4.1 Distribution of z, nH, and ?
Inthissubsection, weshowthedistributionofredshift, columndensity, andspec-
tral index for the simple model and complex model sources. The average redshift
for the sample is 0.03 (Tueller et al. 2008). In Figure 5.4, we show the distribution
ofthelogarithmofz for thesimple(top)andcomplex (bottom)sources. Thecom-
plex sources are slightly closer than the simple model sources. The mean z and
standard deviations are: ? = 0.03, ? = 0.37 (simple) and ? = 0.02, ? = 0.40 (com-
plex). The redshift distributions indicate very local AGN and therefore this may
have an effect on differences in spectral properties/luminosities when we com-
pare our sources with the X-ray properties derived from higher redshift (z ? 1)
hard X-ray selected samples (i.e., the PG quasars), which also tend to be more
luminous than the BAT selected sources by 10?100 times.
In Figure 5.5, we plot the normalized column density distribution of the sim-
ple (top) and complex (bottom) sources. These column densities are the columns
202
in excess of the Galactic value (i.e. that from Dickey & Lockman (1990)), which
is not included in the plot. Where no additional absorption was necessary, we
set nH = 1020 cm?2. From the plots, it is clear that simple model sources have
much lower column densities than sources with spectra modeled with a com-
plex model. The mean and standard deviation for these lognH distributions are:
? = 20.58, ? = 0.74 (simple) and ? = 23.03, ? = 0.71 (complex). Notice that none
of the complex model sources have columns of lognH ? 21. As already noted
above, none of the simple power law model sources havecolumns of lognH ? 23.
With our simple model fits, we have determined the photon spectral index,
?, best fit to the continua of our sources. We excluded a treatment of reflection,
using a simpler model which is recorded more uniformly throughout the litera-
ture (absorbed power law/ partially covering absorbed power law models). This
allows us to use results from the literature for most ofour sources. If we had used
the reflection model (pexrav in XSPEC), many fewer literature results could be
included. In Figure 5.6, we plot the distributions for the simple power law model
(top) and the more complex model (bottom). For the complex sources, we ex-
clude 3C 452 whose flat spectrum is best characterized by reflection (Evans et al.
2006). The mean and standard deviation for the simple model is ? = 1.78 and
? = 0.24. This is consistent with the < ? >= 1.75 we obtained from our repre-
sentative sample (Chapter 4). For the complex sources, 10 had been best fit by a
double power law model while the rest were fit with a partial covering model.
Of these 10, all but 2 had details of the corresponding best-fit partial covering
model in Turner et al. (1997). We computed the average difference between the
two model fits (< ?pcfabs >=< ?dbl > /1.18) and applied this correction to es-
timate ?pcfabs for the 10 applicable sources. Thus, our distribution is computed
using the partial covering model spectral index. The mean and standard devia-
203
Figure 5.4: We plot the distribution of the log of redshift for our uniform sam-
ple of 102 sources. The top plot shows the simple model/power law sources,
where the mean redshift and standard deviation are ? = 0.03 and ? = 0.37. The
bottom plot shows the distribution of complex/partial covering/double power
law sources, where the mean redshift and standard deviation are ? = 0.02 and
? = 0.40. Ascanbeseen,thedistributionofcomplexmodelsourceshasaslightly
lower redshift than the simple model sources.
204
Figure 5.5: We plot the distribution of the log of column densities for our uni-
form sample of 102 sources. The top plot shows the simple model/power law
sources, wherethe meanand standarddeviation are? = 20.58 and? = 0.74. The
bottom plot shows the distribution of complex/partial covering/double power
law sources, where the mean and standard deviation are? = 23.03 and? = 0.71.
Clearly, the simple model sources are dominated by low column density sources
while the complex model sources mostly have higher column densities.
205
Figure 5.6: We plot the distribution of the spectral index, ?, for our uniform sam-
ple of 102 sources. The top plot shows the simple model/power law sources,
where the mean and standard deviation are ? = 1.78 and ? = 0.24. The bot-
tom plot shows the distribution of complex/partial covering/double power law
sources, where the mean and standard deviation are ? = 1.73 and ? = 0.45. The
mean of the simple and complex models are very close, with only a 0.05 differ-
ence. If the partial covering model is an accurate physical model, all of the AGN
have roughly the same power law description for their X-ray spectra.
206
tion for the complex partial covering model is ? = 1.73 and ? = 0.45. There is a
larger spread of values for the complex sources, compared to the simple power
law model sources. However, there is only a slight difference in the average spec-
tral index (0.05). This result is consistent with those of our representative sample
in Chapter 4, which agrees with the average AGN photon indices reported in
Mushotzky (1982) of ? 1.8 from HEAO-1 observations.
5.4.2 Distribution of Lcorr2?10keV and Lcorr2?10keV /LEdd
In this subsection, we test whether Sy 1 and Sy 2 sources have the same or dif-
ferent distributions of both hard band X-ray luminosity (2?10keV) and accretion
rate. Towards this end, we needed to compute an absorption independent mea-
sure of 2?10keV luminosity as well as the Eddington luminosity. In terms of 2?
10keV luminosity, absorption has little to no affect on the unabsorbed luminosity
for sources with nH < 1022 cm?2. Therefore, for these sources we used the quoted
2?10keV flux to compute Lcorr2?10keV (absorption corrected 2?10keV luminosity).
For all luminosity calculations, we used ? = 0.7 and H0 = 75kms?1 Mpc?1.
Above nH = 1022 cm?2, absorption has a significant affect on the 2?10keV flux.
For these sources, we used XSPEC to calculate the unabsorbed 2?10keV flux,
which we then used to compute Lcorr2?10keV .
In Figure 5.7a, we plot our absorption corrected 2?10keV luminosities versus
the 14?195keV luminosities. We find that the data is well fit (R2 = 0.85) by a line:
logLcorr2?10keV = (1.06? 0.05) ? logL14?195keV + (?3.08? 2.16). Therefore, the rela-
tionship is linear (Lcorr2?10keV ? L14?195keV ) showing the validity of our absorption
corrected2?10keVluminosities. Ontheplot,welabelthe5sourceswhichdeviate
the most from this relationship (sources with 1.7 ? L14?195keV/Lcorr2?10keV ? 1.2). A
likely explanation for these sources (NGC 931, 2MASX J09180027+0425066, NGC
207
2992, NGC 5728, and NGC 6814) deviating from the linear fit is variability in
the X-ray spectra, either in the BAT band or the 2?10keV band. Indeed, the 64
day 14?195keV light curves of each of these sources (Baumgartner et al. 2008)
show variability by at least a factor of 2. See ? 5.3, where we discuss the ef-
fects of time variability on hardness ratios. In Figure 5.7b, we plot the ratio of
Lcorr2?10keV/L14?195keV versus the total luminosity in both bands. As shown, there
is no relationship (R2 << 0.1). We also indicate on the plot values of spectral
index (?) for a constant ratio of Lcorr2?10keV/L14?195keV , assuming that ? is constant
between both bands (?corr2?10keV = ?14?195keV ).
To calculate the Eddington luminosity, we used the black hole mass com-
puted from stellar 2MASS K magnitudes. In Mushotzky et al. 2008 (submitted),
we calculated the stellar magnitude from the total 2MASS magnitude (obtained
from NED) and a nuclear magnitude, calculated by using the IRAF task qphot
to extract circular photometry from an aperture equal to the PSF FWHM of the
2MASS images. The stellar absolute magnitude was computed as: Mstellar =
2.5log(?/(?? 1)) + Mtot, where ? = Ftot/Fnuc = 10?0.4(Mtot?Mnuc). We then trans-
formed the measured stellar 2MASS K magnitudes to mass using the relation
set forward in Novak et al. (2006): logMBH = 8.19 + 0.524 ? (MK(stellar) ? 23).
The assumption made, in computing the mass, is that the K band stellar light is
dominated by the bulge. Further details can be found in Mushotzky et al. 2008
(submitted), where we show that the derived stellar luminosity is not correlated
with the X-ray luminosity (14?195keV) but that the nuclear luminosities are.
Comparing these black hole masses to the results from reverberation map-
ping (Peterson et al. 2004), we find that the values agree within the quoted mar-
gin of error for the reverberation study (a factor of 3). LEdd is then computed
as (MBH/M?) ? 1.3 ? 1038 ergss?1. As an estimate of accretion rate (L/LEdd),
208
Figure 5.7: (top) We plot the absorption corrected 2?10keV luminosity versus
the 14?195keV luminosity from BAT. The dashed line shows the best fit line to
the data. The fit shows a linear relationship (Lcorr2?10 ? L14?195) with high sig-
nificance (R2 = 0.85). The second line shows a slope of 2 in logarithmic space
(Lcorr2?10 ? 100 ? L14?195). The 5 sources which deviate the most from this fit
are labeled in the plot and discussed in the text. (bottom) We plot the ratio of
Lcorr2?10keV /L14?195keV versus the total luminosity in the 2?10keV and 14?195keV
bands. Thelinesshowvaluesofconstant?betweenbothbandsatconstantratios
of Lcorr2?10keV /L14?195keV . 209
we use the ratio of Lcorr2?10keV /LEdd. These quantities are related by the bolometric
correction, which could be as large as 100. Assuming a constant bolometric cor-
rection, however, Lcorr2?10keV is proportional to L. We list both black hole mass and
Lcorr2?10keV /LEdd in Table 5.1 for the ?uniform? sample.
InFigure5.8,weplotthedistributionofouraccretionrateproxy,Lcorr2?10keV /LEdd,
for the 34 Sy 1?1.2s (red) and 32 Sy 2s (blue). We find that there is a clear differ-
ence in the distributions, with the Sy 2s having lower ratios of Lcorr2?10keV /LEdd. To
quantify this, the mean and standard deviations for the logarithm of accretion
rate are ? = ?2.92 and ? = 0.57 for Sy 1?1.2s and ? = ?3.53 and ? = 0.75 for Sy
2s. Since we used unabsorbed luminosities to compute Lcorr2?10keV , this effect is not
simply due to the Sy 2s being absorbed and therefore less luminous. We find that
the difference in distributions also holds for luminosity alone, where ? = ?2.92
and ? = 0.57 for Sy 1?1.2s and ? = ?3.53 and ? = 0.75 for Sy 2s. Using the
Kolmogorov-Smirnov test, we find that the differences in distributions are sig-
nificant, both having a very small P value (< 0.001) or probability that the two
samples are drawn from the same parent population.
The result of Sy 2s having lower hard X-ray luminosities than Sy 1s is not
new, but the Eddington ratio result is. A lack of X-ray obscured sources at high
luminosities was noted by Reichert et al. (1986) and Piconcelli et al. (2002). The
results of Steffen et al. (2003) and Ueda et al. (2003) showed that the fraction of
obscured AGN is lower at high luminosities. Our luminosity distribution results
are new, however, in that the BAT AGN include the ?hidden? sources which had
not previously been identified as a class. Therefore, our study is much more
unbiased with respect to absorption than the previous studies.
Also, more importantly, we show that it is not simply the 2 ? 10keV luminosi-
ties which are different but also the accretion rates, estimated by Lcorr2?10keV /LEdd.
210
Table 5.1. K-band Derived Mass and Lcorr2?10keV /LEdd
No. Source M/M?1 LX/LEdd2 No. Source M/M?1 LX/LEdd 2
1 NGC 235A 8.76 -3.75 78 1RXS J112716.6+190914 9.00 -3.21
2 Mrk 348 7.97 -3.53 79 NGC 3783 8.21 -3.25
3 Mrk 352 7.26 -2.69 80 SBS 1136+594 7.53 -2.03
4 NGC 454 6.23 -2.08 81 UGC 06728 6.81 -3.22
5 Fairall 9 8.91 -3.13 82 2MASX J11454045-1827149 6.70 -1.36
6 NGC 526A 8.02 -2.75 83 CGCG 041-020 8.46 -3.32
7 NGC 612 8.47 -3.17 85 NGC 4051 7.27 -4.05
8 ESO 297-018 9.68 -4.66 86 Ark 347 8.12 -3.99
9 NGC 788 8.51 -3.71 88 NGC 4138 6.82 -3.79
10 Mrk 1018 8.94 -3.46 89 NGC 4151 7.69 -3.26
12 Mrk 590 8.87 -4.08 90 Mrk 766 7.85 -3.18
15 NGC 931 8.55 -4.25 91 NGC 4388 8.53 -4.24
16 NGC 985 8.94 -2.78 92 NGC 4395 5.30 -3.30
17 ESO 416-G002 9.02 -3.45 94 NGC 4507 8.39 -3.54
18 ESO 198-024 8.36 -2.85 95 ESO 506-G027 8.59 -3.51
20 NGC 1142 9.36 -4.01 96 XSS J12389-1614 8.88 -3.50
24 NGC 1365 8.88 -4.75 97 NGC 4593 8.61 -4.02
25 ESO 548-G081 8.94 -4.27 100 SBS 1301+540 7.54 -2.24
27 PGC 13946 8.75 -3.74 102 NGC 4992 8.56 -3.69
28 2MASX J03565655-4041453 8.64 -2.88 103 MCG -03-34-064 8.28 -3.56
29 3C 105 7.79 -2.00 105 MCG -06-30-015 7.36 -2.87
31 1H 0419-577 9.00 -2.83 106 NGC 5252 8.64 -3.91
32 3C 120 8.56 -2.71 108 IC 4329A 8.52 -3.05
34 MCG -01-13-025 8.06 -3.18 109 Mrk 279 8.62 -3.09
36 XSS J05054-2348 7.53 -2.04 110 NGC 5506 7.77 -3.12
38 Ark 120 8.74 -3.13 112 NGC 5548 8.42 -3.03
39 ESO 362-G018 9.00 -4.78 113 ESO 511-G030 8.66 -3.70
40 Pictor A 7.60 -2.30 115 NGC 5728 8.53 -4.90
45 NGC 2110 8.28 -3.94 116 Mrk 841 8.15 -2.88
47 EXO 055620-3820.2 8.44 -2.82 117 Mrk 290 7.68 -2.62
49 ESO 005-G004 7.89 -3.99 118 Mrk 1498 8.59 -3.10
50 Mrk 3 8.48 -3.66 124 1RXS J174538.1+290823 8.75 -2.70
51 ESO 121-G028 9.00 -3.81 125 3C 382 9.22 -3.05
53 2MASX J06403799-4321211 ? ? 126 ESO 103-035 7.73 -3.12
55 Mrk 6 8.24 -3.44 127 3C 390.3 8.52 -2.58
56 Mrk 79 8.42 -2.95 129 NGC 6814 8.15 -5.47
60 Mrk 18 7.45 -3.76 133 NGC 6860 8.24 -3.75
61 2MASX J0904699+5536025 7.70 -2.64 136 4C +74.26 9.00 -2.51
62 2MASX J09112999+4528060 7.53 -2.83 137 Mrk 509 8.59 -2.70
64 2MASX J09180027+0425066 8.57 -2.88 138 IC 5063 7.68 -3.08
65 MCG -01-24-012 7.16 -2.26 139 2MASX J21140128+8204483 8.81 -2.72
66 MCG +04-22-042 8.49 -3.13 144 UGC 11871 8.34 -3.54
67 Mrk 110 7.80 -2.11 145 NGC 7172 8.31 -3.58
68 NGC 2992 8.04 -4.52 146 NGC 7213 8.63 -4.46
69 MCG -05-23-016 7.66 -2.70 147 NGC 7314 7.84 -3.73
70 NGC 3081 7.96 -3.66 148 NGC 7319 8.54 -3.55
71 NGC 3227 7.83 -4.02 151 MR 2251-178 8.76 -2.56
72 NGC 3281 8.62 -4.32 152 NGC 7469 8.64 -3.55
75 Mrk 417 8.04 -2.87 153 Mrk 926 8.95 -2.93
77 NGC 3516 8.13 -3.64 154 NGC 7582 8.31 -4.51
1The logarithm of M/M? where mass is derived from the 2MASS K-band stellar magnitudes (see Mushotzky et al. 2008).
2The logarithm of Lcorr
2?10keV /LEdd. Here, L
corr
2?10keV is the value supplied in the X-ray spectral fitting tables for sources
with nH < 1022 cm?2 and the calculated unabsorbed flux for the higher column density sources. The Eddington luminosity is
calculated as 1.3 ? 1038?M/M?.
211
Figure 5.8: We plot the distribution of Lcorr2?10keV/LEdd (top) and unabsorbed
Lcorr2?10 (bottom) for our Seyfert 1 ? 1.2 (red) and Seyfert 2 (blue) sources. From
the plot, it is quite clear that the distribution of Seyfert 2s corresponds to lower
Lcorr2?10keV /LEdd. The X-ray luminosities are unabsorbed values and so the result
is not merely a result of absorption. For both sets of distributions, we used a
Kolmogorov-Smirnov test, finding a high probability of the distributions being
different (the P values were < 0.001).
212
These results provide a challenge to the unified AGN model. If all AGN were
essentially the same but viewed at different angles with respect to obscuring ma-
terial, there would be no difference in the distributions of accretion rate or intrin-
sic luminosity. The fact that we are seeing these differences suggests that there is
something fundamentally different besides the amountof obscuration. However,
if the bolometric correction for hard X-ray luminosities varies between absorbed
(correction of ? 85) and unabsorbed sources (correction of ? 35), as Barger et al.
(2005) suggest, then our distributions of accretion rates would be the same. How-
ever, the unabsorbed X-ray luminosities would still differ and thus there is still a
discrepancy with the unified model.
We are aware of at least three modifications to the unified model that can ex-
plain this difference in distributions between Lcorr2?10keV /LEdd in Sy 1s and Sy 2s.
The first possibility is a luminosity dependent opening angle for the molecular
torus, such as the receding torus model of Lawrence (1991). A second model is
presentedinNicastro(2000),wherethebroadlineregionisproducedfromawind
from the accretion disk. The wind is produced at a boundary between radiation
and gas dominated regions of the disk and its existence is dependent on reaching
(or exceeding) a critical accretion rate. Our observation of a lower distribution
of accretion rates in Sy 2s supports this claim. Their model also predicts that no
hidden broad line regions should exist for low accretion rate sources. Such a test
of the existence/non-existence of polarized broad line regions in the lowest ac-
cretion rate sources (Lbol/LEdd ? 10?3 (Nicastro et al. 2003)), is beyond the scope
of this thesis.
Another possibility is that, at lower luminosities, the host galaxy light is so
much brighter than the emission from the broad line region that it is completely
masked (Moran et al. 2002). Therefore, broad line galaxies should be brighter, as
213
we found. In Peterson et al. (2006), it is shown that the properties of high z ob-
jects may be modified by dilution of X-ray spectral features by star formation.
Given the arc min resolution of ASCA, used for most of our spectra, this allows
the blending of emission from star formation and other non-nuclear X-ray fea-
tures. We expect this effect to be most pronounced for low luminosity sources,
where the ratio of AGN emission to star formation/galaxy emission is lowest.
We quantify the possible effects of this dilution in ? 5.4.6.
5.4.3 Correlations of ? with Lcorr2?10keV and Lcorr2?10keV /LEdd
Basedonrecentstudies(Piconcelli et al.2005;Porquet et al.2004b;Shemmer et al.
2006), we wanted to test whether the X-ray power law index is correlated with
the 2?10keV luminosity or accretion rate. In Figure 5.9, we plot the photon index
versus Lcorr2?10keV (top left) and the ratio of Lcorr2?10keV /LEdd (bottom left). We see
no evidence of a correlation between ? and 2 ? 10keV luminosity or Eddington
ratio, even among the Seyfert 1 sources which have smaller associated error bars
on ?. Based on previous studies, correlations between ? and Lcorr2?10keV have been
seen in high-redshift samples (Dai et al. 2004; Saez et al. 2008), but not among the
low-redshift sample (z ? 0.1) of George et al. (2000). Thus, with our low redshift
sample (< z >= 0.03), we confirm earlier results showing no correlation.
While we did not find a correlation between ? and our Eddington ratio proxy
on average in Chapter 4, we had found a correlation between luminosity and
spectral index for individual sources. This correlation was found when we com-
pared multiple X-ray observations from XMM-Newton and XRT. Similar results
had been seen before for individual AGN (Mushotzky et al. 1993). Thus, these
results show that while on averagethere is no relationship between ? and a given
Eddington rate, on a source by source basis, the photon index becomes steeper
214
Figure 5.9: We plot the unabsorbed 2?10keV luminosity versus power law index
(top) and the ratio of unabsorbed 2?10keV luminosity to Eddington luminosity
versus power law index (bottom). Eddington luminosity is calculated from an
estimate of the black hole mass using 2MASS K band photometry. We see no
correlationbetween? andLcorr2?10keV orLcorr2?10keV /LEdd. AssumingthatLcorr2?10keV is
proportional to bolometric luminosity, our result shows no correlation between
? and accretion rate, contrary to those of Shemmer et al. (2006). The triangles
indicate optical Sy1?1.2 sources, the squares indicate Sy1.5?1.9 sources, and the
circles indicate Sy2 sources. 215
with higher accretion rate. We interpret the fact that we do not see a correlation
in the average plot as a result of each individual source having a broad range of
luminosities and accretion rates which they vary between. Since the individual
ranges overlap between the sources, the scatter is a natural result.
Comparing further with Shemmer et al. (2006), we see at least two possibili-
ties for our different result. First, the 30 objects selected by Shemmer et al. (2006)
are moderate to high luminosity RQQs. Our sample, however, includes lower
luminosity sources not present in their sample. It may be that the correlation is
only there for the most luminous AGNs. In this case, the more luminous AGN
may be more similar, perhaps having the same trigger (like a large-scale merger
event (Di Matteo et al. 2005)) which dictates a specific range of accretion rates.
Our sample is more heterogeneous with a larger range in properties and possi-
bly multiple triggers for AGN activity. Second, the Shemmer black hole masses
are estimated from the width of the optical H-? line (Porquet et al. (2004b) and
Piconcelli et al. (2005) use the luminosity at 5100 ?A) while we used the stellar K-
band flux. It is possible that there is a bias in the Shemmer black hole mass deter-
minations relative to ours. The H-? line width may have an explicit dependence
on ? (Laor et al. 1997), in addition to black hole mass and luminosity.
5.4.4 The Fe-K? Feature
Yet another important feature of AGN spectra is the Fe K? line. For 83/102
sources (81%), a measurement of the strength of this line, via the equivalent
width(EW),wasavailable. ThemissingobjectsincludemostoftheXRTobserved
sources as well as some from the literature. For these sources, the data were not
of sufficient quality to accurately measure the Fe K line strength. In Figure 5.10,
we plot the Fe-K? (6.4keV) EW versus absorption corrected 2 ? 10keV luminos-
216
ity and our proxy for the Eddington ratio. We note that for three bright Seyfert 1
sources (Fairall 9, 3C 120, and 3C 382), the ASCA data show a strong, broad, Fe
K-? line that is not seen in further observations. While we include the ASCA fit
parameters, the high EWs for these sources cause these three points to be clear
outliers.
The X-ray Baldwin/ ?Iwasawa-Taniguchi? effect (Iwasawa & Taniguchi 1993)
is the anti-correlation between Fe K-? and luminosity, quantified as EW ?
L?0.17?0.08 by Page et al. (2004). Recently, this effect has been reported in ra-
dio quiet samples of Jiang et al. (2006) and Bianchi et al. (2007). In the top left
panel of Figure 5.10, we see no evidence of the X-ray Baldwin effect. How-
ever, when we bin the sources by luminosity (excluding the 3 questionable
measurements) we do find a correlation (middle left panel). This correlation
is seen when we choose the average Fe K EW in each luminosity bin, with
logEW = (?0.23 ? 0.03) ? logLcorr2?10keV + (12.11 ? 1.17), and is similar to the
Page et al. (2004) measured slope. The significance of the anti-correlation, mea-
sured by a correlation co-efficient of R2 = 0.93 (99% significance), seems to con-
firm the X-ray Baldwin effect. However, the results are deceiving. We find that
when we alternatively use the median EW, the correlation becomes much weaker
(R2 = 0.63; lower left panel). This shows that the more absorbed, lower luminos-
ity sources ? among them, the hidden AGN clearly missed in other surveys ? are
skewing the results. Therefore, our data does not confirm the inverse Baldwin
effect.
Jiang et al. (2006) suggested that the X-ray Baldwin effect is driven by a cor-
relation of Fe K-? EW and Eddington ratio. Such an anti-correlation was found
by Bianchi et al. (2007), who found that EW ? Lbol/L?0.19?0.05Edd . In the top right
panel of Figure 5.10, we plot the EW versus our proxy for Eddington ratio. As
217
Figure 5.10: We plot the narrow Fe K? line equivalent width (EW) versus unab-
sorbed 2 ? 10keV luminosity (left) and our proxy for Eddington rate (right). In
the top panels, we plot the distribution for all of our sources. The triangles indi-
cate optical Sy1?1.2 sources, the squares indicate Sy1.5?1.9 sources, and the cir-
cles indicate Sy2 sources. The mean EW for the 76 sources plotted is ? = 200eV
with ? = 203eV. In the remaining panels, we plot the distribution after binning
the values by luminosity or Lcorr2?10keV /LEdd using the mean Fe K EW (middle) or
median EW (bottom). While anti-correlations are seen in both sets of plots using
the meanEW(middle), the relationshipweakenssubstantiallyfortheluminosity
plot when the median EW is used (bottom left). Thus, our data does not show
evidence of the X-ray Baldwin effect but does show an anti-correlation in the
binned EW vs. Eddington ratio relation.
218
withtheluminosity, wedo notimmediatelyfind anycorrelation. However, when
we again bin the values (excluding the 3 questionable measurements), we find an
anti-correlation. Using the mean values of Fe K-? EW, we find that logEW =
(?0.26?0.03)?logLcorr2?10keV/LEdd +(1.40?0.10), with R2 = 0.89. Further, unlike
in the luminosity plots, this relation also holds true when we use the median EW.
Thus, our data shows no correlation with unabsorbed 2?10keV luminosity but
does show an anti-correlation with Eddington rate.
This suggests that the primary relationship causing the X-ray Baldwin effect
is the relationship between Fe K EW and accretion rate. From Figure 5.8, we
found that the distribution of Lcorr2?10keV/LEdd is different between absorbed and
unabsorbed sources. The Sy 2s have lower accretion rates than Sy 1s. We also
find this in Figure 5.11, where we plot the Eddington rate proxy and the Fe K
EW versus hydrogen column density (nH). In Figure 5.11 (top), we see that while
there is a large range of accretion rates for a given column density, the higher
column sources tend to have lower accretion rates. In Figure 5.11 (bottom), we
findthatthehighernH sourcesalsohavehigherFeKEWs. Thisresultisexpected,
particularly if some of the sources are Compton thick. However, since there are
sources with higher EWs and lowernH, there is no direct correlation between EW
and column density.
Supporting the case that EW is correlated with accretion rate, Mattson et al.
(2007)reportedacorrelationbetween ?andFeKEW.Theirdatasampleconsisted
of 350 RXTE spectra of 12 Sy 1 ? 1.2 sources, thus, as shown in their Figure 3, it
appears that the anti-correlation (except for jet dominated 3C 273, where there is
a correlation) is seen mostly for multiple observations of the same source. Since
? steepens with increasing accretion rate, we claim that the primary relationship
is again accretion rate and EW.
219
Figure 5.11: We plot Lcorr2?10keV /LEdd (top) and the narrow Fe K? line equivalent
width (EW) (bottom) versus the X-ray measured hydrogen column density. We
find no correlation in either relation. Here, the triangles indicate optical Sy1?
1.2 sources, the squares indicate Sy1.5?1.9 sources, and the circles indicate Sy2
sources.
220
5.4.5 Simple Power Law Model Sources
Nearly half, 46/102, of the uniform sample were well-fit by a simple absorbed
power law model. As mentioned, none of these sources had X-ray column den-
sities ? 1023 cm?2. In this subsection, we detail additional properties of these
sources. In particular, from the data available we can compare optical Seyfert
type with the X-ray column density and examine the sources more closely for
soft excesses. While we hoped to examine sources with warm absorbers in detail,
given the non-uniform nature of analyses of warm absorbers in the literature, we
defer this topic to a later study.
First, we discuss the optical versus X-ray type. For each of the simple model
sources, an optical Seyfert type is listed in Table E.1. Most of the sources, 30/46
(65%), have optical classifications of Sy1 ? 1.2. The mean X-ray column density
for these sources corresponds to a low column density, lognH = 20.7. As ex-
pected, the Sy 1.5 ? 1.9 sources (13/46 or ? 28%) have a higher mean column
density of lognH = 21.8. The 3 Sy2 sources have a mean column density of
lognH = 22.8. Therefore, for the simple model sources, there is no large discrep-
ancy between the X-ray and optical classifications. In the unified model of AGN,
the presence/absence of optical broad emission lines is explained as an effect of
viewing angle. Therefore, sources whose optical spectra show broad emission
lines (Sy1s) would have little obscuring material blocking the central AGN emis-
sion while sources with no broad lines (Sy2s) would be more heavily obscured.
Thus, our results agree, in general, with the unified model.
Next, we will examine the spectral fits in more detail. As a first step, we
look at the goodness of fit for our sources through the ?2 parameter. While a
simple power law or power law and blackbody model was a better choice for
221
our sources than a more complex model (like the partial covering model), 4 of
the simple model sources have ?2/dof values of 1.3 or higher (in Tables E.2
and E.3). Of these, XSS J05054-2348 only had XRT data available. The number
of data points for this source is small, so the high ?2/dof value may be a product
of poor statistics. Of the remaining sources, the source with the highest ?2/dof
value (2.2) is Mrk 841. For this source, XMM-Newton observations analyzed by
Petrucci et al.(2007)showedanFelinecomplex. Inaddition,thissourceisknown
to have a strong soft excess. These features were not well fitted by the simple
models employed in this study. Additionally, Chandra observations of Mrk 279
reveal the presence of a weak absorbing outflow (Fields et al. 2007). Similarly,
a high-quality XMM-Newton spectrum of NGC 4593 reveals an ionized warm
absorber in addition to the soft excess, as well as cold and ionized Fe K lines
(Brenneman et al. 2007).
In the spectral fits for our simple model sources, we allowed for the addi-
tion of a blackbody component to model the presence of a soft excess. In our
uniform sample, 19/46 (? 41%) of the sources required a blackbody for a sta-
tistically improved fit. This is slightly less than the 50% found in Chapter 4,
but not significantly less. The mean blackbody temperature for our sources is
kT = 0.10keV with ? = 0.07keV. This value agrees with the blackbody tempera-
tures found for the PG quasars (Piconcelli et al. 2005; Porquet et al. 2004b) as well
as those for type 1 AGN in the Lockman Hole (Mateos et al. 2005). We note that
unliketheresults ofGierli?nski & Done(2004), who found< kT >= 0.12keV with
? = 0.02keV for the PG quasars, we do see scatter in the values of kT (which be-
comes apparent in Figure 5.12). However, since they used two Comptonization
models instead of a blackbody and power law model, it is hard to make a direct
comparison.
222
Figure 5.12: In the top row, we plot the blackbody temperature for the 19 sim-
ple model sources requiring a soft excess model versus Lcorr2?10keV /LEdd and black
hole mass, respectively. There is no correlation between these parameters. Sim-
ilarly, we find no correlation between the power law index and blackbody tem-
perature (bottom left). However, there is a correlation between the unabsorbed
luminosity in the power law component and that in the blackbody component
(bottom right). Fitting a line to the data, we find that the relationship is nearly
linear (Lpow ?LkT). Here, triangles represent Sy 1?1.2 sources and squares indi-
cate Sy 1.5?1.9 sources.
For the sources with measured soft excesses, we wanted to test whether there
was a relationship between the blackbody temperature (kT) and our proxy for
Eddington rate, black hole mass, and the photon index (?). In Figure 5.12, we
plot the results of these comparisons. As seen in the figure, there is no corre-
lation of any of these values. We tested this by calculating the coefficient of
correlation, R2, which was < 0.10 for each comparison. Based on current un-
derstanding of the soft excess, we did not expect to find correlations for these
223
parameters. If the soft excess were the result of a thermal process, blackbody
emission from a disk surrounding the black hole, we would see a correlation be-
tween black hole mass and the blackbody temperature (T ? M?1/4 L/LEdd1/4).
We do not find a correlation between kT and mass, kT and Lcorr2?10keV/LEdd, or kT
and Lcorr2?10keV/LEdd1/4 ?(M/M?)?1/4 (not shown in Figure 5.12).
In Figure 5.12, we also plot the luminosity in the power law component ver-
sus the luminosity in the blackbody component. Combining spectra from our
own analysis with downloaded spectra of sources with soft excesses previously
analyzed in the literature, we calculated unabsorbed fluxes for 17/19 sources in
the 0.3?10keV band for each component (power law and blackbody). As shown
in the plot, we did find a correlation between the two values. The correlation
is signficant, R2 = 0.48, with logLpow = (0.79 ? 0.14) ? logLkT + (9.34 ? 6.04).
Thus, we find that Lpow ? LkT. Comparing the flux values Fpow and FkT, we find
that the correlation is still present with R2 = 0.50 and logFpow = (0.43 ? 0.08) ?
logFkT + (?5.79?0.90).
Such a relationship between Lpow and LkT may provide a challenge to soft ex-
cess models where the excess is the result of an absorbing model, unless there is
an explicit luminosity dependence between the absorbing wind and source emis-
sion. In terms of a reflection origin, where the soft excess is the result of emission
from reprocessed disk emission, there is no implicit contradiction. However, in
this case, our results indicate that since the reprocessed emission (LkT) is on the
same order as the input spectrum (Lpow) the process would need to be highly
efficient, with the reprocessor occupying a large solid angle and a very high re-
processing efficiency.
224
5.4.6 Complex Sources
Slightly more than half of our sample (? 55%) consists of sources whose X-ray
spectra were not well-fit by a simple absorbed power law model. Here, we dis-
cuss the optical types of these sources. We also discuss the fraction of complex
sources and the nature of the complexity.
As expected, the optical types for the complex sources are opposite those of
the simple model sources. Here, only 4 sources are identified as Sy1 ? 1.2, where
the complexity in their spectra is a result of complex absorption (e.g. IC 4329A is
knowntohaveatleast7separateabsorbersintheX-rayspectrum(Steenbrugge et al.
2005)). The majority of sources are Sy2s (36/56 or 64%), with the remaining 16
sources Sy1.5 ? 1.9s (29%). As with the simple model sources, the optical types
roughly matched the X-ray column densities. The mean values of lognH corre-
sponded to 22.0 (Sy1 ? 1.2), 23.2 (Sy1.5 ? 1.9), and 23.6 (Sy2). Thus, we find no
substantial discrepancy between our data and the standard unified AGN model
for our entire uniform sample, with respect to X-ray/optical classifications.
One question remaining is the cause of the complexity. We fit the spectra in
this category with a partial covering model, but this model is flexible and can
be used to describe more than just a cloud or clouds of material blocking some
of the AGN light. As already mentioned, one cause of complexity is complex
absorption, e.g. warm absorption. The Seyfert 1?1.2 sources from our sample
with complex spectra are known to fit in this category. Another possible cause
of complexity is scattering of direct emission from an obscured region into our
line of sight, accounting for the soft emission (< 2keV). We can not easily test
either of these theories with our data. Yet another cause of complexity could be
that the soft emission is not from the AGN, but rather from X-ray binaries, star
225
formation, or hot ionized gas in the host galaxy. Since we do not expect this
emission to exceed a few?1041 ergss?1 (Ranalli et al. 2003), we can automatically
rule out this scenario for sources with higher luminosities in the soft emission.
InFigure5.13(top),weplotthedistributionofsoftX-rayluminosityinthe0.5?
2.0keV band for our complex sources. The mean luminosity is logL0.5?2.0keV = 41
with ? = 0.94. We find that only 13/54 sources have soft emission high enough
to exclude a simple explanation of galactic emission (logL0.5?2.0keV ? 41.5). For
these sources, which include all of the 4 Sy 1s as well as 3 ?hidden? AGN, it is
unlikely that the soft emission is from the host galaxy. The fact that a few of the
hidden sources can not be explained by this model suggests that an alternative
model, like scattering or partial covering, is more favorable. However, for 75% of
the complex sources, the luminosities are too low to exclude galactic emission. In
Figure 5.13 (bottom), we plot the observed soft luminosity versus the observed
hard luminosity. We would naively expect the two luminosities to be directly
correlated if they are related and not due to galactic emission. Of course, the
effects of obscuration in the 0.5?2keV band have not been considered to make
this plot, particularly since the nature of the soft emission is ill-determined. A
strong correlation is not seen, but this does not rule out any of the possibilities.
Unfortunately, the present data set has too low anangular resolution, on average,
to distinguish galactic sources of soft emission from the AGN. We know that in
some cases, e.g. Circinus, NGC 1365, and NGC 4151, the soft emission is due
to X-ray binaries, hot gas from star formation, and extended emission from AGN
cores, respectively. However, higherqualitydatawiththesuperiorspatialresolu-
tion of Chandra is needed to solve the problem of the origin of the soft emission.
Even with the low fluxes of many of our sources, not very long (? 10ks) Chandra
exposures would be required to obtain images of the soft emission.
226
Figure 5.13: Plotted in the top panel is the distribution of soft band X-ray lumi-
nosity for sources with spectra best fit by a complex model. The mean value is
logL0.5?2.0keV = 41 with ? = 0.94. This shows that for more than half of the
sources the luminosities are low enough that we can not rule out the idea that
the soft emission is from X-ray binaries/star formation/diffuse gas in the host
galaxy. Plotted in the bottom panel, we show the soft X-ray observed luminosity
versus the hard X-ray observed luminosity. A strong correlation is not seen, as
indicated by R2 = 0.51. This further suggests that we can not explain the soft
emission as a simple extension of the hard power law emission.
227
Figure 5.14: These plots show the fraction of absorbed sources in a given
Lcorr2?10keV bin (top) and accretion rate bin (bottom). N/Ntot is the number of ab-
sorbed sources in a given bin divided by the total number of sources in that bin.
The black bins show the fraction of sources with lognH ? 23. These are a subset
of thesourceswith lognH ? 22, picturedinblue. TherearenohighnH sourcesin
the highest Lcorr2?10keV (unabsorbed) bin. Further, the fraction of obscured sources
is lower at high luminosities. In the second plot, it is not clear that the fraction of
obscured sources increases with lower accretion rate (Lcorr2?10keV /LEdd). We find
that50%ofthesourcesinthehighestluminositybinareabsorbed, however,with
only 8 sources in this bin, this could be the result of poor sampling.
228
Earlier, we had shown that the distribution of 2?10keV luminosities and our
Eddington ratio proxy (Lcorr2?10keV/LEdd) was lower for Sy 2s than Sy 1s. Another
importantinvestigationthatwecannowmakeisthefractionofobscured AGNas
a function of luminosity and accretion rate. In Figure 5.14, we show the results of
the fraction of sources with column densities above lognH = 22 in each indicated
2?10keV unabsorbed luminosity bin and logLcorr2?10keV/LEdd bin. We also show the
subset fraction of sources with lognH ? 23 (in black). These plots show us that
there are clearly less obscured sources at high luminosity. The highest luminosity
bin is composed entirely of unabsorbed sources. One interesting thing to note,
however, is that the most absorbed sources (lognH ? 23) are not more numerous
at lower luminosity. Rather, they are merely a subset of the absorbed sources. In-
stead, the sources with 23 > lognH ? 22 dominate in the lowest luminosity bins.
It is unclear what this result implies. However, our results clearly support previ-
ousstudieswhichfoundthefractionofobscured sourceslowathighluminosities
and higher at lower luminosities (Barger et al. 2005; Steffen et al. 2003; Ueda et al.
2003). Our results argue even more strongly that there must be a modification to
the unified model which includes dependence on luminosity.
Our plot of the fraction of absorbed sources by binned accretion rate (Fig-
ure 5.14 (botom)), is more difficult to interpret. While absorbed sources do dom-
inate at the lowest values of logLcorr2?10keV/LEdd, the highest Eddington rate bin
also shows a large fraction of absorbed sources. However, we find that this bin
includes fewer sources (8), and the result could be a product of poor statistics.
We do note that, as with luminosity, the fraction of the most heavily absorbed
sources (lognH ? 23) does not appear to increase with decreasing Eddington
rate. Rather, it appears to remain nearly constant. Still, on the whole, the ab-
sorbed sources make up the largest fraction of sources at low Eddington rate and
229
a lower fraction at high Eddington rates.
Another important conclusion drawn from our analysis is that the fraction
of ?hidden? AGN, sources with low scattering fractions (? 0.03), is significant.
These sources comprise 45% of the complex sources and 24% of our uniform
sample. This highlights the importance of the BAT survey and its ability to find
obscured sources, since these objects have no indication of AGN activity in the
soft X-ray band. Now that these sources are firmly established as an important
subset of local AGN, it is important to understand their properties. In Chapter
4, we had noted that fitting the sources with a reflection model is problematic.
We were unable to constrain the reflection parameter or the cutoff energy, even
with the addition of the BAT spectrum (14?195keV). Further, while the partial
covering model provides a decent fit, this model is very flexible. One probable
explanation for the hidden/buried AGN, is that they are embedded in a very
geometrically thick torus (Ueda et al. 2007).
Fabian et al. (1998) proposed a model in which low-luminosity AGN were ob-
scured by nuclear starbursts. This is one possible origin for a geometrically thick
torus. Following Ueda et al. (2007), we used the 60?m and 100?m fluxes from
the Infrared Astronomical Satellite, obtained from NED, to estimate the far infrared
luminosity of the hosts of the hidden sources. We found these values for 16/24.
The mean value of logLFIR = 43.76 with ? = 0.42. We also computed the ratio
of Lcorr2?10keV/LFIR. Here, we find that the mean value is 0.26 with ? = 0.26. This
value is consistent with ratios for AGNs in the local universe, as pointed out by
Ueda et al. (2007). It is the sources with very small ratios, << 0.1, which indi-
cate the possibility of significant starburst activity. In our sample, only 6 sources
fall into this category, with the lowest ratio (0.007) corresponding to NGC 7582,
an object whose H and K band nuclear light is dominated by young supergiants
230
(Oliva et al. 1995). However, without higher quality IR observations of the nu-
clear region and an analysis which can separate out any AGN emission, little
conclusions can be drawn from this data.
Yetanother questionremaining ishowmanysources areCompton-thick. This
is not an easy question to answer, especially since different authors use different
definitions. In general, the term has been used to apply to: (1) heavily obscured
AGN (nH > 1.4 ? 1024 cm?2), (2) spectra with a high EW Fe K line, (3) spectra
with a flat power law continuum, and (4) spectra with a strong reflection hump.
The last three criteria are all indications of a reflection dominated spectrum. If we
taketheCompton-thickdefinitiontoapplytosourceswhosecolumndensitiesare
> 1.4 ? 1024 cm?2, none of the BAT-detected sources are Compton-thick. There
are, however, sources which come close (NGC 612, NGC 3281, NGC 1365, NGC
5728, NGC 6921, and NGC 7319), with nH ? 1024 cm?2. Without simultaneous
dataabove10keV, itis extremely difficultto definitivelydiscriminatebetween re-
flection models and partial covering models. Even with simultaneous data from
Suzaku, we are finding it impossible for some sources (Winter et al., in prep).
However, this is not the case for all sources, for instance, NGC 5728 strongly
prefers a reflection model and exhibits a strong iron K EW (? 800eV). Therefore,
it is clear that if other criteria are used we do find Compton-thick sources. For in-
stance, we find that6 sources exhibit a very flatspectrum (? ? 1.0). Alternatively,
6 sources have strong Fe K equivalent widths (EW ? 600eV). From our data, we
cannot test for the presence of a reflection hump, since the feature is predominant
above10keV.However,withtheBATspectrasuchstudiesareforthcoming. Com-
bining criteria, only 1 source shows both a flat continuum and a strong Fe K EW,
NGC 5728. However, the column density is still below the strict Compton-thick
limit. In this discussion, we have not considered the complicated spectra of Cen
231
A, NGC 1275, and the double nucleus NGC 6240, who may also have consistent
properties with the various Compton thick definitions.
5.5 The Cosmic X-ray Background
The most recent synthesis models for the cosmic X-ray background (CXB) find
that heavily obscured and Compton thick AGN are as important as unabsorbed
AGN (Gilli et al. 2007). With an unbiased sample of AGN from the BAT, we can
begintounderstandthecontributionsofbothtypesofAGNinthelocalUniverse.
In the previous section, we have presented the distribution of column densities
and power law indices from the 0.1 ? 12keV X-ray bands for our unique uniform
14 ? 195keV X-ray sample. These properties will provide a valuable input to the
CXB models for low redshift (z ? 0) AGN. In addition, in this section we provide
the 2 ? 10keV logN-logS relationship. With this relationship, we will comment
on the completeness of a very hard X-ray selected sample in the 2?10keV band
and its implications.
5.5.1 The Average X-ray Spectrum
Having compiled all of the properties of the 9-month BAT AGN sources, we
present the average 0.6?10keV X-ray spectrum for our uniform sample of 102
sources in Figure 5.15. Since the normalizationvalues are not uniformly recorded
in the literature, we could not simply add the spectra together. Instead, we con-
struct the average spectrum using the shape of the spectrum (from the absorbed
power law models) with 2?10keV flux used to weight the contribution to the
average spectrum.
To construct the average spectra, we have excluded three important AGN fea-
232
tures: blackbody components/soft excess, Fe K lines, and reflection. We do not
include blackbody components since the < kT >? 0.1keV components have
very little spectral effects at E > 0.6keV, the energy above which the CXB is well
measured (McCammon et al. 2002). It is almost impossible to measure the diffuse
background below this level because the Galaxy?s soft X-ray background is very
bright at E < 0.7keV. A significant fraction of our spectra are from XRT observa-
tions, where one cannot constrain Fe K line emission or reflection. Also many of
the ASCA, XMM, and other data sets were analyzed in many different ways with
respect to reflection and without re-analyzing the entire data set in a uniform
fashion one cannot model reflection correctly. In particular, we do not believe
that the exclusion of reflection has a significant effect on the average spectrum
< 10keV, since the effect of reflection is typically rather small at these energies
(for all but the largest reflection fractions). In fact, as Nandra & Pounds (1994)
show, a ? = 1.7 power law is almost a perfect match to a ? = 1.9 power law with
reflection in the 2?10keV band.
The average spectrum was constructed as: I(E) = summationtextAi ? Fi ? E1??i. For
each source, Fi is the observed 2?10keV flux. The measured spectral index was
used as the main component of the AGN emission. Finally, Ai represents the
absorbing column density. To apply the absorption term, we constructed a grid
of XSPEC absorption models using an input power law with a spectral index of
1.73 (the average of the complex model sources). The models included simple
absorption using tbabs with logNH = 20,20.5,21,21.5,22,22.5,23 and complex
absorption using pcfabs with a partial covering fraction of 0.95 and logNH =
21.5,22,22.5,23,23.5,24. The flux of the output XSPEC models corresponds to
normalized photonscm?2 s?1 keV?1. In order to compute A for each model, we
divided the model by the power law contribution (E1?1.73) and multiplied by
233
Figure 5.15: Shown here is the average spectrum constructed from the X-ray fits
toouruniformsampleof102AGN.Thesolidblackcurveistheaveragespectrum
of the AGN sources. The dashed lines show the contribution from the simple
absorption sources (magenta) and the complex absorption sources (green). We
also show a line fit to the average spectrum from 0.6 ? 10keV. The slope of the
line, ? = 1.369? 0.004, is consistent with the modeled CXB slope. The total flux
from our sources corresponds to only 0.29% of the entire 2?10keV CXB.
energy. We then applied the appropriate model for each of the sources based on
NH and the complexity flag.
In Figure 5.15, the contributions of the simple model and complex model
sources are shown. We note that we chose to neglect additional soft emission and
Fe K line signatures, as well as reflection. Fitting a line to our spectrum for ener-
gies above 0.6keV, we find logI(E) = (?0.369?0.004)logE + (?12.989?0.003),
with R2 = 0.97. Thus, we find ? ? 1.37. This result is quite remarkable. In 1980,
234
HEAO-1 found that the CXB could be modeled as a power law with ? = 1.4,
below 15keV (Marshall et al. 1980). However, the narrow distribution of power
law indices (? ? 1.7) for AGN led to a ?spectral paradox?, where it was unclear
how averaging over these dominant bright sources resulted in the flatter power
law index fit to the CXB (Boldt 1987). The paradox was resolved theoretically
by assuming that many AGN are absorbed, such that ? 85% of their light is ob-
scured (Fabian & Iwasawa 1999). With our simple estimate of the average X-ray
spectrum from the SWIFT BAT-detected AGN, we have now observationally re-
produced the measured CXB slope. The flux from these bright sources, with a
total flux of 1.79 ? 10?9ergs?1 cm?2over 74% of the sky, accounts for only 0.29%
of the CXB (2.0 ? 10?11ergs?1 cm?2 deg?2 Revnivtsev et al. (2008)). However, if
the distribution of source properties at z ? 1, where most of the CXB originates,
is similar to that of the BAT sources, the spectral paradox is resolved. A similar
conclusion was reached by Sazonov et al. (2008), who calculated the 3?300keV
SEDs of local AGN using Integral and RXTE data.
5.5.2 The 2-10keV Log N - Log S Relationship
Towards determining how complete a 14?195keV survey is in the 2?10keV band,
we plotted the logN-logS relationship for our uniform sample in the 2?10keV
band in Figure 5.16. Here, logN is the logarithm of the number of galaxies with
a 2?10keV flux above the associated logS value. We corrected this value for the
BAT sky coverage by using the BAT sky coverage map from Tueller et al. (2008)
and the 14?195keV flux for each source. We found that the sample is complete
in the 2?10keV band to logS ? ?11 (F2?10keV ? 1?10?11 ergcm?2 s?1). To show
this, we fit a line to the points above this threshold (shown in the plot). The
fit is very significant, with R2 = 0.97 and logN(S) = (?1.53 ? 0.12) ? logS ?
235
Figure 5.16: Here we plot the relation of logN-logS for our entire uniform sam-
ple (|b| ? 15? or 74% of the sky). The value logN corresponds to the number of
AGN with 2 ? 10keV fluxes above the indicated logS, corrected for sky coverage
using the BAT sky coverage maps and 14?195keV fluxes. The dashed line repre-
sents a fit to the points with logS ? ?11. The slope of this line (?1.53 ? 0.12) is
consistent with the cumulative distribution of a uniform density of objects (-1.5,
shown as the solid line). This plot further suggests that we are missing many
sources at fluxes below logS ? ?11.
(14.93?1.24).Thus, themeasuredslopeof?1.53?0.12isconsistentwiththevalue
?1.5 expected for a uniform density of objects. Our normalization, which corre-
sponds to 1.17?10?15 (ergscm?2 s?1)?1 sr?1, agrees very well with the normaliza-
tion from the 2?10keV logN-logS relation of HEAO-1 for AGN (2.2+0.3?0.2 ? 10?15
(Piccinotti et al. 1982) for AGN).
This shows that above ? 1.0?10?11 ergcm?2 s?1 (2 ? 10keV) we have a com-
236
plete sample, consisting of 51 sources. Only 9 of these AGN are Sy2s while 28
are Sy1 ? 1.2s, showing that the brightest sources correspond mostly to less ab-
sorbed sources even in a very hard X-ray selected sample. Re-examining the
average spectrum of this complete sample (with F2?10keV ? 10?11ergs?1 cm?2),
we find that the shape is the same as the larger sample (logI(E) = (?0.41 ?
0.005)logE + (?13.019 ? 0.003) with R2 = 0.973). Thus, the average spectrum
of these bright sources also replicate the measured CXB slope of ? ? 1.4. Of this
complete sample, all have measured NH < 5?1023 cm?2 and none are Compton-
thick. In fact, most of the sources have much lower column densities (only 6 have
NH ? 1023 cm?2). This suggests that there are no local Compton-thick objects
with 2?10keV fluxes above 1.0?10?11 ergcm?2 s?1.
However, we also find that below this flux threshold we are far from com-
plete. Particularly, towards logS = ?12 the plot suggests that we are missing
close to 3000 sources. Could some of these missing sources be Compton-thick?
From logN-logS in Tueller et al. (2008), we know that the 9-month BAT AGN
survey is complete in the 14 ? 195keV band above 2?10?11 ergcm?2 s?1. There-
fore, the missing sources in our sample must have 14 ? 195keV fluxes below
2 ? 10?11 ergcm?2 s?1 and 2 ? 10keV fluxes from ? 1 ? 10 ? 10?12 ergcm?2 s?1.
Since the ratio of F14?195keV/F2?10keV for the majority of missing sources (near
logS = ?12)is? 10, itislikely thatthese sources areheavily absorbed. They may
beanextensionofthe?hidden?AGN withhighercolumndensities (lognH ? 24).
We can not determine whether or not some of these sources are Compton-thick.
237
5.6 Host Galaxy Properties
In this section, we present a simple analysis of the host galaxy properties of the
uniform sample of BAT-detected AGN. Since the BAT AGN offer the first unbi-
ased AGN sample with respect to absorption (nH ? 1024 cm?2), studying their
host properties can provide major insight into the relationship between the AGN
and host. This is particularly important since the exact trigger of activity around
the black hole is not known.
Since the BAT AGN hosts are located nearby (< z >= 0.03), archived images
are easily available from the Digital Sky Survey (DSS), Sloan Digital Sky Survey
(SDSS), and2MASS.For ourstudy, weuse these dataaswellas publicly available
information from NED. First, we provide a simple comparison of the galaxy ma-
jor axis/minor axis to the X-ray column density. We then discuss the host galaxy
morphology types as well as the fraction in interacting galaxies.
5.6.1 Host Inclination
One of the questions left to be answered about AGNs is the nature of the obscur-
ingmaterial. Likely, therearemanydifferent answerstothisquestion, depending
on the host. In some cases, the obscuring material may largely be a product of
looking through much of the gas and dust in the host galaxy. In such case, we can
use the inclination of the host galaxy to determine how much of the obscuration
can be simply attributed to the host. To test this, we found measurements of the
major axis (a) and minor axis (b) in NED. These data were available for 85/102
sources in our uniform sample.
In Figure 5.17, we plot the ratio of minor to major axes (b/a) versus the X-ray
derived column density. From the plot, we find that sources of all types (Sy1 ?
238
Figure 5.17: We plot the measured host galaxy minor axis/major axis (b/a) from
NED versus the X-ray measured column density. As shown, only one Seyfert
1 (IC 4329A, a Sy 1.2) is in a highly inclined galaxy. However, this source has
a higher X-ray column than most other Seyfert 1s. The hosts of Seyfert 2s span
the entire range of b/a. Note that for sources requiring no additional absorption
over the Galactic, we used a standard value of nH = 1020 cm?2. Here, trian-
gles represent Sy 1?1.2 sources, squares indicate Sy 1.5?1.9 sources, and circles
indicate Sy2s.
2) have similar distributions of b/a. We find that a face-on galaxy (b/a = 1) could
host either a Sy1 or Sy2 source. However, the edge-on galaxies (b/a ? 0.4) only
host more absorbed AGNs. Among the 11 edge-on sources, only one is associ-
ated with a Sy1. However, while IC 4329A is an optical Sy1.2, the X-ray column
density is high for a Sy1 (6 ? 1021 cm?2) and the X-ray spectrum is complex. In
fact, the XMM-Newton spectrum of this source reveals signatures from 7 differ-
239
ent absorbing systems (Steenbrugge et al. 2005). We note that Kirhakos & Steiner
(1990)foundadeficiencyofedge-ongalaxiesintheiropticallyselectedAGNsam-
ple, which they explained as a lack of obscured sources in the optical survey. Our
results support this claim, since the unobscured sources are not associated with
edge-on galaxies.
For the edge-on sources, it is possible that much of the X-ray obscuration is
from the ISM in the host galaxy. In the Milky Way, we know that the column den-
sity looking towards the Galactic Center is a few ?1022 cm?2. Therefore, this can
be a plausible explanation for the 6 edge-on sources with X-ray columns below
1023 cm?2. For the other 5 sources, however, the columns are simply too large to
be attributed solely to the host galaxy.
Excluding the edge-on sources, we find the mean values for b/a as ? = 0.74
and ? = 0.13 for Sy 1 ? 1.2s, ? = 0.69 and ? = 0.14 for Sy 1.5 ? 1.9s, and ? = 0.70
and ? = 0.15 for Sy 2s. This shows that the distributions are effectively the same,
with a difference of only 0.04 in b/a between obscured and unobscured AGNs.
Thus, we can not explain differences in obscuration simply through the inclina-
tionof the host galaxyfor these sources. The fact that we see a rangeofb/avalues
causes us to disagree with Rigby et al. (2006), whose data on 9 optically active
AGN (AGN showing optical emission lines) at 0.5 ? z ? 0.8 led them to con-
clude that optically active AGN are only in the most face-on or spheroidal host
galaxies. However, they also conclude that optically dull AGN (from a sample
of 22 AGN) inhabit a range of b/a ratios, similar to our result for the entire sam-
ple. As for optically dull AGN in the 9-month BAT AGN sample, there are only
two sources (NGC 612 and NGC 4992), both of which are intermediate between
face-on and edge-on with b/a = 0.64 and 0.58, respectively.
240
5.6.2 Host Morphology
For our uniform sample, 74/102 sources had morphologies available from NED
or LEDA (listed in Table E.1). For all of these sources, we expressed the mor-
phology as a T-type, following the 2MASS large galaxy atlas (Jarrett et al. 2003)
and the de Vaucouleurs method. Here, each morphology is assigned a numerical
value (E = -1, S0 = 0, Sa = 1, Sab = 2, ..., Irr/Peculiars = 9.5). In Figure 5.18 (left),
we plotthe distribution of morphologies. For all of these galaxies, we also looked
at the accompanying DSS images as confirmation of the NED classification.
Additionally, these images are available for all of the 9-month surveys online
at http://swift.gsfc.nasa.gov/docs/swift/results/bs9mon/ (Baumgartner et al.
2008).
From the plot of the distribution of morphologies, we find that our sample
includes very few ellipticals (? 8%). This is in direct contrast to deeper AGN
surveys where the hosts are predominately in elliptical or red hosts. For instance,
the z ? 1 X-ray selected sample of Nandra et al. (2007) mostly have red colors
consistent with early type galaxies (ellipticals and lenticulars). Meanwhile, our
< z >= 0.03 sample includes a larger fraction of bluer hosts (spirals and pecu-
liars). This suggests that there may be an evolutionary change in the host proper-
ties between local AGN and those at higher redshift. However, the SDSS results
show a connection between Seyfert 2s and early type hosts but also young stellar
populations at 0.02 < z < 0.3 (Kauffmann et al. 2003). In order to compare more
directly with both the SDSS and X-ray selected samples, we need more robust
data including spectra and photometry. Thus, our results, based on NED clas-
sifications and DSS images, are preliminary and require higher quality images
and photometry (Koss et al. in prep) as well as spectra (Winter et al. in prep) to
241
substantiate these findings. Our main goal from our simple analysis is to identify
potentially interesting properties for further study.
In the morphology distribution plot, it is clear that a large fraction of the hosts
(33%) are irregular/peculiars, with an equal percentage of Sy 1-1.2s, Sy 1.5-1.9s,
and Sy2s with peculiar hosts. In some sense, the classification as peculiar in NED
can be somewhat subjective. However, we looked through all of the DSS images
to confirm the NED classifications. In a study of 256 nearby (z ? 0.035), opti-
cally selected galaxies with HST, Malkan et al. (1998) found only ? 20% of the
hosts ?normal? (i.e. axi-symmetric, a bulge with regular elliptical isophotes, a
thin, planar disk). This suggests that even more than the 33% identified would
fit within this category. In Figure 5.19, we show optical images of four represen-
tative cases. Some of these sources are quite famous and have high quality HST
images available showing large dust lanes and other disturbances (i.e. Cen A,
NGC 1275). In fact, many of these sources appear to be interacting with nearby
galaxies (i.e. ESO 490-G026, ESO 511-G030) or are identified as the result of a
merger (as for Cen A). More images can be seen at the indicated BAT 9-month
survey website.
This leads to the most important result we have found in our analysis of the
host galaxies. Namely, that a large fraction of the galaxies are interacting. In
Table E.1, we included a note of Int or Int? to indicate sources that we classify
as interacting or possibly interacting. Here, we use the term loosely to describe
sources which, from the available DSS images, have a nearby companion galaxy
or show a heavily distorted morphology, which has previously been identified
as the result of a merger in the literature. We supplement the list with sources
known to be the product of a merger (like Cen A and Cyg A). In addition to
ESO 490-G026, Cen A, and ESO 511-G030, shown in Figure 5.19, more images of
242
Figure 5.18: We plot the distribution of host galaxy type, from NED, for the uni-
form sample of BAT-detected AGNs (top). Also, we plot the distribution of host
galaxy type for ?interacting? sources in our uniform sample (bottom). Clearly,
there is no significant difference in the host type for these sources.
243
3:20:00.0 55.0 50.0 45.0 19:40.0 35.0
33:00.0
30.0
32:00.0
30.0
31:00.0
30.0
41:30:00.0
30.0
29:00.0
28:30.0
Right ascension
Declination
NGC 1275
22.0 18.0 14.0 6:40:10.0 06.0 02.0
51:30.0
52:00.0
30.0
53:00.0
30.0
-25:54:00.0
30.0
55:00.0
30.0
56:00.0
Right ascension
Declination
ESO 490-G026
40.0 35.0 13:25:30.0 25.0 20.0 15.0
-42:59:00.0
30.0
-43:00:00.0
30.0
01:00.0
30.0
02:00.0
30.0
03:00.0
30.0
Right ascension
Declination
Cen A
32.0 28.0 24.0 14:19:20.0 16.0 12.0
36:30.0
37:00.0
30.0
38:00.0
30.0
39:00.0
30.0
-26:40:00.0
30.0
41:00.0
Right ascension
Declination
ESO 511-G030
Figure 5.19: We show DSS images of 4 representative peculiar/irregular host
galaxies. For all images, we plot a 5??5? region. Both NGC 1275 and Cen A have
available HST images which reveal in detail large dust lanes. However, even in
the lower resolution DSS image, the dust lane in Cen A is clearly visible. Of the
other sources, ESO 490-G026 is really two colliding galaxies and ESO 511-G030
is a spiral with a small nearby companion galaxy (outside the image).
244
interacting hosts are found in Tueller et al. (2008). All six of the sources shown
in Figure 8 of that paper are classified as interacting. Additionally, NGC 454,
NGC 6240, and NGC 7469 are interacting sources with HST images released in
the recent ?Cosmic Collisions Galore!? news release. Just from these available
images, we account for 28% of the sources classified as interacting. Images of the
additional sources are available online at the BAT 9-month survey website.
Of the sources with a known morphology (including those listed as S or S?),
we classify 45 sources (54%) as interacting. The distinction between peculiars
and interacting sources is somewhat arbitrary, since a merger galaxy would be
expected to have a peculiar shape. However, not all of the classified peculiars
are also classified as interacting, though a large fraction are (66%). If interac-
tions cause disruptions to the galaxy, thereby making it peculiar, the percentage
of interacting galaxies is even higher ? including all of the peculiars. It is also
important to note that not all interacting galaxies are peculiar. The distinction
for these sources may be due to a companion that is much smaller than the AGN
host or at a greater distance.
The important thing to note is that a large fraction of local AGN have com-
panion galaxies or are the result of a merger (54%). This is significant considering
that 12-33% of optically selected Seyferts (Rafanelli et al. 1995) and 15% of more
distant X-ray selected AGN (0.2 ? z ? 1.2) (Pierce et al. 2007) are hosted in an
interacting galaxy. It is possible that our high percentage of interacting galaxies
indicates an evolutionary difference in AGN hosts. Another possibility is that
our value is higher simply because we can see smaller companions at low red-
shift. Thus, the sources at z ? 0.1 may have the same number of companions or
merger galaxies like Cen A and Cyg A but the available images may not be of
sufficient quality to distinguish this.
245
Figure 5.20: Here we plot the distributions of unabsorbed 2?10keV luminosity
andourEddingtonratioproxyforinteracting(black)andnon-interacting(green)
systems. There is little difference between the two distributions, showing that
AGN in galaxies with a close companion or which underwent a recent merger
have the same luminosities and accretion rates as AGN in non-interacting galax-
ies.
246
Sincesuchalargepercentageofhostsareinteracting,wewantedtotestwhether
the distribution of luminosities and Eddington ratios were the same or different
between AGN with interacting and non-interacting hosts. In Figure 5.20, we plot
the results. We find no difference in the AGN unabsorbed 2?10keV luminosity
or the Eddington rate proxy (Lcorr2?10keV/LEdd) between the two distributions. For
interacting systems, ? = 42.97 and ? = 0.80 for logLcorr2?10keV while ? = ?3.38 and
? = 0.76 for Lcorr2?10keV/LEdd. The non-interacting systems have a very similar dis-
tribution, with? = 43.12and? = 0.81forlogLcorr2?10keV and? = ?3.29and? = 0.70
for Lcorr2?10keV/LEdd. Using the Kolmogorov-Smirnov test, the P values (0.65 for lu-
minosity and 0.91 for Eddington ratio) are not small and therefore do not suggest
a difference between the distributions.
Further, when we examine the morphologies of the interacting host galaxies,
we find that they follow the same distribution as the larger sample (Figure 5.18
(bottom)). When we look at the distribution of optical Seyfert type within inter-
acting galaxies, we find no preference for absorbed (Sy 2) systems. Broken down
by optical type, 14 correspond to Sy 1 ? 1.2, 13 in Sy 1.5 ? 1.9, and 13 in Sy 2
sources. This is similar to the results of Rafanelli et al. (1995), who also found
no difference in the percentage of interacting sources between Sy 1s and Sy 2s.
However, this is somewhat confusing since we would expect optically-selected
samples to find less of the interacting Sy 2s due to heavy absorption. Since the
morphologies are a nearly even mix of Sy1 ? Sy2 sources, it is not surprising
that the mean 2?10keV luminosities and Eddington ratio proxies are intermedi-
ate between the Sy1 and Sy2 values quoted in ? 5.4.3 (for both interacting and
non-interacting hosts).
Based on simple analyses of publicly available data on the host galaxies, we
have found a few interesting results. We have found that the host inclination (ap-
247
proximated by b/a) does have an effect on the amount of obscuration we see in
the X-ray band, but that the effect is slight. In particular, there are no unabsorbed
sources hosted in an edge-on galaxy. However, there are heavily obscured AGN
in face-on galaxies. This shows that for the bulk of the obscuring medium, the
origin is most likely more local to the accreting black hole. This is further sup-
ported by HST observations of nearby AGN, which find Sy2s to be more likely
to have dust lanes or irregular or disturbed dust absorption through the galactic
nucleus (Malkan et al. 1998).
Also, we found that many of the hosts are in spirals and pecular/irregular
galaxies. We find very few (5/74) in ellipticals, contrasting with the results of
Nandra et al. (2007) who find red hosts for the majority of z ? 1, X-ray selected
AGN. This suggests an evolutionary change in AGN host properties, from red at
z ? 1 to blue at z ? 0.03. However, high quality photometry is necessary to make
a direct comparison with Nandra et al. (2007).
More than half of our sample is associated with a close companion or re-
cent/ongoing merger. We find that Sy 1 and Sy 2 sources are equally likely to
be hosted in interacting galaxies. Considering that optical surveys at similar red-
shift found between 12-33% (Rafanelli et al. 1995), it is tempting to attribute the
difference to selection effect, i.e. the heavily obscured sources missed in the opti-
cal. However, we find that the hosts of the interacting systems are not dominated
by absorbed sources.
We also find that the distribution of host morphologies are the same be-
tween interacting/non-interacting hosts. Further, the distributions of Lcorr2?10keV
and Lcorr2?10keV/LEdd are also the same. Though, the reason for the similar dis-
tributions is unclear. Likely, mergers are not the sole trigger for our sample.
The next step in understanding the differences between the interacting and non-
248
interacting systems is an in-depth source by source look at the local environ-
ments, with particular attention to star-formation and dust. This, however, is
beyond the scope of this thesis.
5.7 Summary
In this chapter, we present the X-ray properties of a uniform sample of very hard
X-ray (14 ? 195keV) selected AGN. We present a number of interesting results
that highlight the many uses of an unbiased very hard X-ray survey. This study
is complimentary to the 9-month AGN survey paper (Tueller et al. 2008), which
presents the 14?195keV properties of the sources. Additionally, this work con-
firms the results of our earlier study on XMM-Newton observations of a represen-
tative sample of the BAT AGN (Chapter 4). Among these, we show that: (1) the
X-ray and optical classifications agree, i.e. Sy 1s have low X-ray column densities
while Sy 2s are more obscured, (2) the average power law index, ? ? 1.8, agrees
with the results from HEAO-1 (Mushotzky 1982), (3) ?hidden? AGN are a sig-
nificant fraction of local AGN, where we can now quantify this value as ? 24%,
and (4) nearly half (45%) of local AGN are well-fit by a simple model (all with
lognH < 23) while the remaining sources (55%) require a more complex model.
In addition, this chapter presents a number of additional, important results.
Some of the most interesting results of the BAT AGN sample involve the na-
ture of the host galaxy. From examining the host galaxy properties, we found
that the majority of the X-ray obscuration is not simply from the host galaxy (by
comparing host inclination (b/a) to X-ray column density). The most surpris-
ing results, however, were that many of the host galaxies are peculiar/irregular
galaxies (33%). Further, an even larger fraction (54%) have either a close compan-
249
ion galaxy or are known mergers. This is observational proof that galaxy inter-
actions may be driving activity in local supermassive black holes. However, we
also find that the distribution of AGN 2?10keV luminosities and accretion rates,
as well as morphologies, are the same between interacting and non-interacting
hosts. While, it is unclear what these results mean, however, there appears to be
more than one trigger besides mergers for local AGN activity.
From our uniform sample (102 sources with |b| ? 15?), we found that the
distributions of both unabsorbed 2 ? 10keV luminosity and accretion rate are
significantly lower for Sy 2s than Sy 1s. While earlier studies found this con-
nection in 2 ? 10keV luminosity (Steffen et al. 2003; Ueda et al. 2003), this is the
first time it has been reported in accretion rate. We also showed that the frac-
tion of obscured AGN is indeed larger for lower luminosities (absorption cor-
rected Lcorr2?10keV ) and accretion rates. However, we note that the most heavily
obscured sources (lognH ? 23) do not dominate this relationship. Since the uni-
fied model predicts differences between absorbed and unabsorbed sources are a
product of viewing angle alone, our results provide a challenge, arguing in favor
of a luminosity-dependent AGN model.
Anotherresultinvolvesthecorrelationbetween accretionrateand?. InChap-
ter 4, we had found indication of a connection between Eddington ratio (or 2 ?
10keV luminosity) and ? using the spectral fits of multiple observations for indi-
vidualsources. Thefactthatwedidnotobserveacorrelationinourlargersample
seems to be a result of our sources having a larger range of Eddington ratios (or
2 ? 10keV luminosities). We suggest that previous studies, for instance by Shem-
mer et al. 2006, see this correlation because their samples have a narrower range
of properties (being mid- to high luminosity AGNs). The primary correlation ap-
pears to be with accretion rate and not hard band luminosity. Such a correlation
250
should appear when comparing ? to Lcorr2?10keV/LEdd for multiple observations of
individual sources or for a sample of sources with a narrow range of accretion
rates.
Inasimilarmanner,wefoundthatwhileoursampledidnotimmediatelycon-
firm the X-ray Baldwin effect, binning the sources by luminosity, we were able to
reproduce the anti-correlation between unabsorbed 2 ? 10keV luminosity and Fe
K EW. The primary anti-correlation, however, again appears to be with Edding-
tonrate. WhenwebinnedthevaluesbyourEddingtonratioproxy,wefoundthat
EW ? Lcorr2?10keV/LEdd?0.26?0.03 (agreeing with the results of Bianchi et al. (2007)).
Since both ? and Fe K EW are dependent on accretion rate, this suggests that the
?-EW correlation found by Mattson et al. (2007) is a result of the accretion rate
dependences.
Having classified the X-ray spectra of our sample into simple and complex
categories, we were able to examine the properties of the two sub-samples in
more detail. For the simple model sources, we found that 41% of the sources
exhibited a soft excess. Having modeled this parameter with a simple blackbody
model,wefoundtheaveragetemperaturetobekT = 0.10keV.Wealsofoundthat
there was a significant amount of scatter in this value (? = 0.07keV), contrasting
withtheGierli?nski & Done(2004)resultsforPGquasars. Wefoundnocorrelation
between the blackbody temperature and Eddington ratio, black hole mass, or
photon index. However, we did find a correlation between the luminosity of the
blackbody component and the luminosity in the power law. This relationship
is linear (Lpow ? LkT) and may provide a challenge to the current soft excess
models.
Examining the complex model sources, we found that the majority of these
sources included absorbed AGN. Of the 4 Sy 1s in this category, all have complex
251
absorption features in their X-ray spectra. For these sources, we showed that the
nature of the soft emission (L0.5?2keV ) for these sources is unclear. Over half have
soft band luminosities low enough to be the result of galactic emission from star
formation/X-ray binaries. However, of the sources with higher soft luminosities,
3 are ?hidden?/buried AGN. This argues that the soft emission may be scattered
AGN emission.
Animportantresult wefoundisthatthe?hidden?/buriedAGN,sources with
a high covering fraction, are a significant fraction of local AGN. Among the com-
plex sources, 45% are ?hidden?. For these sources, we found that the FIR lu-
minosity is not consistent with an increased star formation rate, as suggested
by Ueda et al. (2007). However, without higher quality X-ray spectra and multi-
wavelength observations, we are unable to further explore the nature of these
sources.
While BAT is quite good at finding heavily obscured sources, we found that
none of the 9month sources in our uniform sample have spectra consistent with
heavily obscured Compton-thick objects (nH > 1.4?1024 cm?2). However, we do
detect sources classified as Compton-thick in other studies based on a reflection
dominated spectrum or strong Fe K EW (for instance 3C 452 and NGC 4945).
Additionally, we did not include an analysis of the very complex source NGC
6240, which may also be classified as Compton thick. Since the Compton hump
lies above 10keV, spectral fits with and without reflection can be degenerate in
the 0.1?10keV band. Therefore, a full analysis of the Compton thick nature of the
BAT sources must be deferred to future studies.
One remarkable result we found came from the average spectrum we con-
structed in the 0.1?10keV band with the measured spectral properties of our uni-
form sample. Here, our data reproduce the measured slope of the CXB (? 1.4).
252
This highlights the importance of the BAT survey in selecting heavily absorbed
sources. More importantly, this is observational proof that the combination of
BAT-detected absorbed and unabsorbed local AGN replicate the shape of the
CXB. If the distribution of source properties at z ? 1, where much of the CXB
originates, is the same as that of the BAT-detected AGN, the spectral paradox is
resolved.
To test our completeness in the 2?10keV band, we plotted the distribution of
logN-logS for the entire uniform sample. This showed that while the sample is
complete in the 14?195keV band (Tueller et al. 2008), we are only complete above
logS = ?11 in the 2?10keV band. Further, this distribution suggests that we are
missing as many as 3000sources at logS = ?12, requiring thatthese sources have
14?195keV fluxes below the current flux limit of the BAT survey. Possibly these
sources are ?hidden? AGN with even higher X-ray columns (lognH ? 24). Such
sources must have a high ratio of F2?10keV/F14?195keV , like NGC 1068. Also, they
may or may not containCompton-thick sources, an answer to which our data can
not supply. These results, in addition to the X-ray properties (including column
densities and spectral indices) will provide important input for CXB models at
low redshift (z ? 0).
253
Chapter 6
Conclusions
In this dissertation, we have presented the results of X-ray surveys of accreting
black holes external to our Galaxy. For both the ULXs and AGN, possible inter-
mediate mass and super massive black holes, a great amount of energy is output
at these high energies. Further, the photons are energetic enough to penetrate
much of the gas and dust, which might otherwise hide the underlying black hole.
To understand the spectra of both ULXs and AGN, we employed a technique
of fitting the source spectra with simple schematic models (i.e., a black body
model, a power law model, absorption models, a gaussian line). It was impor-
tant to use such models for a few reasons. First, the true physical model for these
sources is not known. Therefore, any model that is used is an approximation on
some level. Further, in many cases it is not possible to tell which model is the
best fit to the data and the fits are often degenerate. For instance, we highlight
this degeneracy when discussing various disk models applied to ULXs in Chap-
ter 3 as well as in our discussion of the partial covering model versus double
power law and reflection models in Chapter 4. With so many uncertainties in the
?true models?, it becomes necessary to choose some base model which can be
applied to a large class of sources and be used as a comparison between the spec-
254
tra. Finally, since the data sample is not uniform (i.e. sources were observed with
different instruments and/or for different exposure times), use of a base simple
model allows for an easier comparison.
As presented in Chapters 2?5, a wealth of information can be obtained using
these simple modelanalyses. In this chapter, we willdetailthe highlightsof these
surveys including a discussion of our contributionand future directions. Inorder
to better facilitate the discussion, ULXs are discussed in ? 6.1 and the BAT AGN
Survey is discussed in ? 6.2.
6.1 ULX Studies
6.1.1 Conclusions
The study of ULXs is a relatively new endeavor. As such, the breadth of research
on these sources is not as extensive as the 40+ year history of AGN. When we
began our survey in the summer of 2004, a few ULX surveys had been com-
pleted with ROSAT (Colbert & Ptak 2002; Liu & Bregman 2005; Ptak & Colbert
2004) and Chandra (Swartz et al. 2004). With Chandra, there is less of an issue
of source confusion because of the superior sub-arcsecond angular resolution.
These studies were aimed primarily at finding ULXs in galaxies. However, in
order to conduct a comprehensive spectral survey, higher spectral resolution was
needed.
Following upon the work of Miller et al. (2003), who found the disk temper-
atures of two ULXs consistent with intermediate mass black holes (IMBHs), our
study sought to test another IMBH prediction. Namely, we used the superior
spectral sensitivity of XMM-Newton to look for source spectra consistent with
low/hard and high/soft state spectra; analogous to Galactic black hole systems.
255
As presented in Chapter 2, we did find sources consistent with low (16) and high
state (26) ULXs. We also found a population of sources who were consistent with
stellar mass black holes. Since a study of this kind had never before been con-
ducted, our results are significant and provide further support of at least some of
the ULXs as accreting IMBHs. In addition to providing a catalog of ULXs and the
X-raypropertiesof? 80sources, wealsoprovidedotherinformationusefultothe
community like the ratio of optical to X-ray flux for ULXs and the distribution of
column densities and host galaxy star formation rates.
In Chapter 3, we presented the result of a more in-depth look at the highest
quality XMM-Newton ULX spectra. We found, as in Chapter 2, that some ULXs
have spectra consistent with IMBHs and others with stellar mass black holes at
high accretion rates. We also found that we can use ULX spectra to probe the
ISM of their host galaxies. The column densities of the ULXs were largely in
agreement with H I studies, showing that only the M81 ULX required additional
absorption more local to the source. Further, we found that all of the ULXs ap-
peared to be in areas of roughly solar oxygen and iron abundances.
Recent work on ULXs appears to be centered along two tracks. The first in-
volves applying different theoretical models (particularly disk or Comptoniza-
tion models) to the ULX X-ray spectra, producing claims of sources having spec-
tra consistent with either stellar mass black holes or IMBHs. The progress along
this line appears to be hindered by the lack of higher quality spectra. Therefore,
since many models can be used to fit the spectra without a clear statistical pref-
erence, it appears that further progress requires higher quality data. The second
tract of ULX studies centers primarily on optical follow-ups of the sources. The
studies involve either looking for a clear optical counterpart (i.e. the companion
star) or the study of ionization nebulae. For a few ULXs, the discovery of a sur-
256
rounding ionization nebula has provided an exciting line of research and further
support of the IMBH hypothesis. Still, the holy grail of ULX studies ? observing
an eclipsing ULX system and measuring the stellar mass function ? continues to
elude researchers.
6.1.2 Future Work
There are many unanswered questions involving ULXs. While it is clear that
some sources are likely IMBHs, based on their X-ray spectra and the existence of
ionization nebulae, we still do not know how IMBHs form. Further, how many
ULXs are IMBHs versus stellar mass black holes and why/how long do they
accrete matter at this rate? If the ULXs are associated with high galactic star
formation rates, why are they not found in all star forming regions? Some future
projects that may help to answer some of these questions are:
? Studies of the local environments. Particularly, a study aimed at obtaining
local star formation rates for a large sample of ULXs, would provide valu-
able information on the local environments of ULXs. In Figure 2.10, we
showed that galaxies with the most ULXs also have the highest far infrared
luminosities (an indicator of star formation rate). However, we also found
galaxies in our sample with high LFIR and no ULXs. Further, we found
ULXs in irregular galaxies with low LFIR. Currently, we are involved in
a project classifying ULXs as near/outside/inside H II regions based on
XMM-Newton OM images (Berghea, Mushotzky, Weaver, Winter, in prep.).
However, ground-based studies measuring SFR indicators in the IR at the
ULXpositionandthroughoutthehostgalaxywouldgivedirectinformation
on whether or not ULXs are associated with heavy star formation. Further,
with this information we could separate the various types of ULXs (stel-
257
lar mass and probable IMBHs) looking for any differences between the two
sub-classes.
? Absorbed ULXs? From SWIFT and Integral, we know that there are a large
numberofobscuredAGN.IntegralisnowfindingaclassofobscuredGalac-
tic X-ray binaries (Chaty 2008), which SWIFT will likely also detect. From
our work and that of other authors, AGN studies are showing that there
is likely a physical mechanism related to luminosity which creates the dif-
ferences between obscured and unobscured sources. Assuming that there
is a luminosity difference between obscured and unobscured IMBHs, we
can look for obscured ULXs in really long XMM-Newton exposures. This
would be yet another piece of evidence in favor of IMBHs. However, it may
be difficult to tell whether the sources are IMBHs, stellar mass black holes,
or background AGN.
6.2 SWIFT BAT-detected AGN Survey
6.2.1 Conclusions
As described in ? 4.1, the BAT AGN survey is important because it is not affected
by the selection effects plaguing other bands. Also, it is the first sensitive all-sky
survey at 14?195keV. With a larger field of view, it detects many more sources
than Integral in a given amount of time. Therefore, with the continued lifetime
of the SWIFT mission the potential for more complete studies of local AGN in-
creases with ?t.
In Chapter 4, we presented XMM-Newton follow-ups of 22 BAT sources from
the 9-month catalog. As shown in Chapter 5, these sources are a representative
258
sampleof thecatalog. Thestatistical spectroscopic properties ofthe entire sample
are presented in Chapter 5. However, Chapter 4, in addition to providing a first-
look atthe BAT AGN, provides a timing analysis of the XMM-Newton and SWIFT
XRT spectra (on hour to month time scales).
From our AGN study, we have unveiled a number of interesting discover-
ies. Among these, we have quantified the fraction of ?hidden? or buried AGN.
Thisclasshadjustrecently beendiscovered throughSuzakufollow-upsofSWIFT
BAT-detected sources (Ueda et al. 2007). From our survey, we found that ? 20%
of our uniform sample had properties consistent with this class (i.e. a small frac-
tion of scattered emission/direct emission). This shows that a significant fraction
of local AGNs are only now being considered as a part of the general AGN zoo.
When we added the properties of these hidden sources, along with the unab-
sorbed and other absorbed sources, we reproduced the shape of the cosmic X-ray
background (? ? 1.4) (Marshall et al. 1980). This is quite remarkable since never
before were thereAGN X-ray observationswhichreproduced theCXB. However,
the bulk of the CXB is believed to be produced in the z ? 1 universe. Therefore,
provided the spectral properties of our low redshift sources match the higher
redshift sources, the CXB is resolved below ? 10keV.
At higher energies, reflection becomes more important making it necessary to
obtain high quality X-ray spectra above 10keV. We have found this to be chal-
lenging for the fainter sources. While we have BAT spectra for all of the sources,
the spectra currently available are 8 channel spectra and therefore difficult to
use to constrain reflection. Further, the response files were constructed using
the Crab spectrum as a basis (with a slope of ? 2), introducing error for spectra
which deviate from this shape. Suzaku allows for simultaneous observations in
the ? 0.2 ? 600keV band, however, we find that the spectra of our faint sources
259
are only ? 2?3% above the PIN background level, yielding poor quality spectra.
New background models have just been released, which will hopefully improve
the signal of the high energy spectra.
Another important result which we have presented is that the fraction of ob-
scured sources is larger at both low luminosity and low accretion rate. Other
X-ray surveys had noted this effect in luminosity (Barger et al. 2005; Steffen et al.
2003; Ueda et al. 2003), but this is the first time the effect was also seen in ac-
cretion rate. This result strongly argues for a modification of the unified AGN
model to include a dependence on luminosity/accretion rate. A physical expla-
nation may include a luminosity dependent opening angle to the torus due to
radiation pressure or a wind from the disk blowing matter away.
Since the BAT-detected AGN are mostly nearby sources (z ? 0.03), they are
close enough that we can study their host galaxy properties in detail. Using pub-
licly available data, we found that many of the sources were hosted in peculiar
galaxies (33%). Many, but not all, of these galaxies were classified as interacting
systems. In total, 54% of the hosts had nearby companions or were classified as
merger galaxies. This suggests that galaxy interactions are an important source
of fuel for local AGNs. Since the host galaxies appear to be bluer than higher
redshift AGN hosts, at z ? 1 hosts are primarily red or ellipticals (Nandra et al.
2007), this suggests an evolutionary change in the hosts of AGNs.
The most distant AGN (z >> 1) are intensely powerful X-ray sources. These
are believed to be fueled by large scale galaxy mergers ? mergers between galax-
ies of roughly equal size ? which not only fuel the central AGN, but also fuel a
phase of active star formation (Di Matteo et al. 2005). Following this stage, the
star formation is quenched by a wind from the AGN. With much of the AGN?s
fuel blown away, the accretion rate and luminosity lessens. Nandra et al. (2007)
260
find that their results of red/elliptical hosts at z ? 1 are consistent with this stage
of quenched star formation. Our results from the BAT-detected AGN suggest
another step in this evolution. In this phase, smaller scale interactions and per-
hapsother as yetunknowntriggers begina new stage ofstar formationandAGN
activity.
6.2.2 Future Work
TheresultsofourstudyoftheSWIFTBAT-detectedAGNshowthatthereismuch
to learn from an unbiased AGN survey. Further, they present direction towards
a number of areas the BAT survey can present new understanding. For instance:
? Optical Spectraof the BAT AGN. Nowthatwe havecompiled the complete X-
raypropertiesoftheBATAGNsample, animportantnextstep isto compile
the optical properties and make comparisons between the two bands. This
is particularly important for the sources which were not previously found
in the optical, the ?hidden?/buried AGN. Towards this end, I had obtained
(with Dr. Karen Lewis) optical spectra with the 2.1-m telescope at Kitt Peak
Observatory for ? 40 sources (AGN and template galaxies). Additionally,
? 30 more BAT AGN have publicly available spectra from the Sloan Digital
Sky Survey. The aim of this study will be subtracting the host galaxy from
the AGN light to measure the prominent emission features. With the AGN
lines, we will make diagnostic plots, estimate the amount of obscuration,
and compare optical and X-ray properties (i.e. compare the H-? line to the
X-ray slope).
? Optical Photometry of the BAT AGN. Based on our observations that the BAT
AGN hosts are comprised of many spirals, peculiars, and interacting sys-
261
tems (Chapter 5), we are obtaining optical images of the 9-month sources
(Koss et al., in prep). With this work, we will be able to quantify the claims
in Chapter 5 for a uniform sample. One major part will be obtaining col-
ors with which we can compare the BAT AGN hosts with those from SDSS
(Kauffmann et al. 2003) and the AEGIS X-ray sample at z ? 1 (Nandra et al.
2007; Pierce et al. 2007). This will tell us how the BAT hosts compare to
optically selected and hard X-ray selected (2?10keV) samples. We will be
able to determine if there are differences between the hosts of obscured and
unobscured sources and whether there is an evolutionary change between
the hosts at higher redshifts (the X-ray sample) and our own low redshift
sample.
Another interesting result will be quantifying the size and location of com-
panion sources in our effort to understand the interacting nature of the host
galaxies. From this, we can find if there is a difference in the type of interac-
tion between obscured and unobscured sources. We will also find whether
interactions are primarily between galaxies of similar size or a large host
and small companion galaxy.
? Multi-wavelength Observations. As mentioned in Chapter 5, many more in-
vestigations are currently underway to unveil the properties of the BAT
AGN.ASpitzerproposalhasbeenapprovedforIRspectroscopy ofmanyof
our sources (led by Dr. Weaver). An HST proposal has been submitted for
optical and IR imaging (PI, Dr. Weaver). Additionally, radio observations
are also underway and a VLA proposal is in preparation (PI, Dr. Ulvestad).
Further, optical polarimetry observations have been proposed for the hid-
den AGN sources (PI, Dr. Gandhi). Understanding these local AGN will
require data from all available sources. However, with these observations,
262
we can test whether the findings of other AGN surveys are consistent with
the properties of the BAT AGN. Since the BAT AGN are less biased towards
obscuration, we hope that their properties will reveal the true properties of
local AGN.
? Interesting Sources. In the BAT AGN survey, a number of new and interest-
ing sources have been discovered. In Chapter 4, we mentioned two sources
which had unusual spectra, NGC 6860 and ESO 362-G018. NGC 6860 had a
complex spectrum which was flat both above and below 2keV. We believe
thatthiscomplexshapeisduetocomplexabsorptionfeatures. ForESO362-
G018, the prominent Fe K line seen in the XMM-Newton spectrum appeared
to disappear in the much brighter XRT spectra. Such sources, and others
like them, are excellent targets for higher signal-to-noise observations. In
fact, for NGC 6860, we were approved for a 130ks XMM-Newton spectrum
to further investigate the spectrum with both RGS and EPIC.
263
Appendix A
Glossary
A.1 Abbreviations
2MASS: Two micron all sky survey
AV: total extinction in the optical V band
ASCA: the Advanced Satellite for Cosmology and Astrophysics, a Japanese
X ray satellite launched in 1993
AGN: active galactic nuclei
BAT: Burst Alert Telescope onboard SWIFT, sensitive in the 14?195 keV band
CCD: charge coupled device
CXB: cosmic X ray background
Dec: declination
dof: degrees of freedom
E: energy; measured in keV for the X ray band
EB?V: selective extinction between the optical B and V bands (AB ?AV )
EPIC: European photon imaging camera; instrument onboard XMM Newton
ESA: European Space Agency
F: flux; measured in ergs s?1 cm?2
FITS: Flexible Image Transport System; standard astronomical data format
endorsed by NASA and the IAU
FWHM: full width half maximum
H I: neutral hydrogen; observed in the radio through the 21 cm emission feature
HEASARC: NASA?s High Energy Astrophysics Science Archive Research
Center
HST: Hubble Space Telescope
IMBH: intermediate mass black hole
IR: infrared
ISM: interstellar medium; the gas and dust between stars
kT: measurement of the accretion disk temperature in units of energy; the
Boltzmann constant times the accretion disk temperature
L: luminosity; measured in ergs s?1
LEdd: Eddington luminosity
M: mass?
M: mass accretion rate
264
MCD: multi component disk; an accretion disk model consisting of black body
emission from different temperature regions in the disk
MOS: type of CCD used on XMM Newton
NASA: National Aeronautics and Space Administration
NED: NASA/IPAC Extragalactic Database
nH: column density of neutral hydrogen; measured in atoms cm?2
OM: Optical monitor; optical/UV telescope onboard XMM Newton
pn: type of CCD used on XMM Newton
PSF: point spread function
QSO: Quasi stellar object; designation for a type of AGN
RA: right ascension
RGS: Reflection grating spectrometer; detectors onboard XMM Newton
SAS: Science Analysis System; data analysis software for XMM Newton
SFR: star formation rate
Sy: Seyfert type; an optical classification scheme for AGN
Sy1: shows optical broad emission lines and narrow emission lines
Sy1.5: intermediary between a Sy1 and Sy2 source
Sy2: shows only narrow optical emission lines
ULX: ultra luminous X ray source
UV: ultra violet
UVOT: Ultraviolet/optical telescope; onboard SWIFT
VLA: Very Large Array; an array of radio telescopes located near Socorro, NM
WHISP: Westerbork observations of neutral Hydrogen in Irregular and SPiral
galaxies
XMM: X ray Multi mirror Mission
XRT: X ray telescope; onboard SWIFT
XSPEC: A data analysis and spectral fitting program used in X ray astronomy
z: redshift
A.2 Symbols
?2: a statistic used to compute goodness of fit for a model
?: efficiency factor for mass to energy conversion
?: photon spectral index measured from a power law model
?: mean/average value
?: degrees of freedom
?: signal to noise; alternatively, the measured error/standard deviation or a cross
section
?: optical depth
265
A.3 Units
?A: Angstrom; a unit of length
: 1?A= 10?10 m
Crab: a unit of intensity corresponding to the flux density of a bright X ray
source, the Crab nebula
: 1 Crab = 1060?J
: 1 Crab = 2.5652?10?9erg s?1 cm?2 keV?1
erg: a unit of energy
: 1 erg = 10?7 J
eV: electron volt, a unit of energy
: 1 eV = 1.60217653(14)?10?19 J
Jy: Jansky; a unit of flux density
: 1 Jy = 10?23erg s?1 cm?2 Hz?1
: 1 Jy = 2.42?10?18erg s?1 cm?2 keV?1
M?: solar mass
: 1M? = 1.9891?1030 kg
Mpc: Mega parsec; a unit of distance
: 1 Mpc = 3.08568025?1024 cm or 1.91735281?1019 miles
sr: steradian; SI unit of solid angle
A.4 Constants
c: speed of light in a vacuum; 2.99792458?108 m s?1
G: Gravitational constant; G = 6.67300?10?11 m3 kg?1 s?2
H0: Hubble?s constant; we adopt H0 = 75km s?1 Mpc?1
h: Planck?s constant; h = 6.626068?10?34 m2 kg s?1
k: Boltzmann?s constant; 8.617343(15)?10?5 eV K?1
?T: Thomson cross section; 6.652?1025 cm?2
266
Appendix B
Appendix to Chapter 2
B.1 Spectral Simulations
B.1.1 Two-component Model
In order to determine the number of counts required to distinguish
whether a blackbody component is statistically significant for the
sources fit with a two-component model, we simulated spectra based
on that of some of the brightest sources. We chose to simulate spectra
of bright two-component spectra exhibiting three different cases: (1) the
flux from the blackbody dominates over the power law component at
2kT, (2) an intermediary case, and (3) the flux from the power law dom-
inates over the blackbody component at 2kT . Such simulations would
allow us to determine the uncertainty in our claims of a combined fit
being a better descriptor of the data. This is necessary because there is
no a priori model which predicts the relative intensities of the two com-
ponents and, as we know from studies of Galactic black holes, these
components show a wide variety of relative intensities. To this end,
we simulated spectra using the best-fit absorbed blackbody and power
267
law model with the fakeit command in XSPEC. We chose (1) NGC
247 XMM1, (2) NGC 5408 XMM1, and (3) Holmberg II XMM1 as our
seed observations. These objects all have very high signal-to-noise and
thus the fits are robust. The respective ratios of powerlaw flux to black-
body flux contributions at 2kT are: (1) << 1.0, (2) 1.77, and (3) 3.52. All
of these sources have comparable blackbody temperatures indicative of
our high-state ULX candidates (roughly kT ? 0.15).
We simulated 500 spectra each, using the two-component model, for
each of 2000counts, 1000counts, 400counts, and 200counts for the PN.
Each simulated spectrum, based on the best-fit blackbody and power
law model, was fit with an absorbed blackbody and powerlaw model
as well as an absorbed pure-powerlaw model. We placed the constraint
thattheblackbodytemperaturemustliewithintherangeof0.07?4.0keV
(the range at which it would be detectable in the XMM-Newton band-
pass). We allowed the power law index to vary over the range 0? 4 for
the power law component of the combined blackbody and power law
model. However, we placed a constraint that the power law component
must lie within the range ? = 1.5?2.0 for the simple power law model
to be consistent with our fits to the sources we claim are best fit by sim-
ple power laws. This constraint ensures that the spectral index would
exhibit that of our classified ?low-state? objects.
When analysing and classifying our real spectra, we declared a detec-
tion of the thermal disk component if the addition of this component (to
a baseline powerlaw model) led to an improvement of the goodness of
fit parameter by at least ??2 = 8. We can use the above simulations
to address the detectability of a thermal disk component using this ??2
268
threshold as a function of the relative strength of the thermal compo-
nent and the number of counts in the spectrum. For each simulation,
we fit the spectrum with both a single absorbed power-law and a 2-
component powerlaw and thermal disk model and compute the quan-
tity ??2 = ?2pow ? ?2pow+disk. In Fig. A1, we plot the distribution of ??2
from our 500 simulations for the weak and strong blackbody component
for spectra with 400counts and 2000counts. It is clear that we cannot
detect a weak thermal component in a 400 count spectrum ? the vast
majority of the simulations (? 82%) result in ??2 < 8. However, even
a weak thermal component is easily detected in a 2000 count spectrum
(not a single simulation gave ??2 < 8). The strong blackbody case is
detectable with high significance even in a 400 count spectrum (< 1% of
the simulations resulted in ??2 > 8).
When we increase the upper limit of the range of the spectral index in
XSPEC for the power law model to ? = 3.0, our confidence levels de-
crease. For a weak thermal component with 400counts, all of the simu-
lations result in ??2 < 8. At 2000counts, only ? 30% of the simulations
for a weak thermal component have ??2 < 8. For a strong thermal
component, 20% of the simulations yield ??2 < 8 for 400counts while,
as was the case for ? = 2.0, none of the simulations gave ??2 < 8 for
2000counts.
Thusallowingthe?to?floatfreely?orremainunconstrainedfurtherde-
creases the confidence levels. When the ? parameter is allowed to float
the spectra are fit with higher ? values in order to compensate for the
missing blackbody component. When the upper limit was instituted at
2.0 or 3.0, we found that all 500 simulated spectra were fit with a ? = 2.0
269
or 3.0, respectively, for the simple power law model. This same affect is
not seen in the ? of the two-component model, where the value ranges
between 1.4 and 4.0 with a peak in the distribution near that of the orig-
inal model used to simulate the spectra. Thus, higher power law indices
(> 3.0) can indicate the necessity of an added blackbody component.
Since there are no Galactic black holes whose broad band spectra are
well fit by steep power laws it seems that restricting the allowed power
law indices is more consistent with the nature of Galactic black holes.
In fitting our sources with the three ?standard? models, we allowed the
? to float, thus the problem of a missing blackbody component being
compensated for by a steep power law should not have factored into
our classification criteria.
270
Figure B.1: Distribution of ??2 from simulations. The ??2 values represent the
difference between the unreduced ?2 of the absorbed power law model and the
combined blackbody and power law model. The left panel shows the results
for the ?weak? blackbody component relative to power law for 500 simulated
spectra at 400 counts (top) and 2000 counts (bottom). The right panel shows the
results for the ?strong? blackbody component. For a weak blackbody compo-
nent at 400 counts, the thermal component is undetectable. However, it is able
to be distinguished as the number of counts is increased. A strong blackbody
component is easily distinguishable at 400counts.
271
B.1.2 Simple Power Law Model
Our next set of simulations sought to determine our confidence in the
simple power law fit being an adequate descriptor of the spectra. Bin-
ning provides a problem in distinguishing between a powerlaw and a
curvature in the spectrum at the low energy range, since the binning
procedure can wash out a low kT blackbody from the spectrum. For this
reason, we chose to simulate unbinned PN spectra for a source we cate-
gorized as a low-state object, IC 0342 XMM1. IC 0342 XMM1 represents
characteristics typical of our low-state candidates, namely it is within
the proper luminosity range and it has a power law index of ? ? 1.7
(the median of the distribution for low-state objects is 2.03) and a hydro-
gen column density near the median of all the fitted values (where the
median value is ? 3 ? 1021 cm?2 and the value of IC 0342 X-1?s column
density is 5.8 ? 1021 cm?2). We chose this source for these reasons and
the high number of counts in its PN spectrum. Instead of using the ?2
statistic(usedforbinneddata),wechosetousethemaximum-likelihood
statistic, C-stat, in XSPEC (which uses unbinned data).
We simulated 500 spectra using the best-fit parameters for the simple
power law fit using the fakeit command. We fit the simulated spec-
tra with two models: the simple powerlaw and a combined blackbody
and power law model and computed the change in the goodness of fit,
??2. Forthetwo-componentmodel,onceagainweplacedtheconstraint
that the blackbody temperature remain in the range that it would be de-
tectable by XMM-Newton, 0.07 - 4keV. We followed this procedure for
1000, 2000, and 4000 counts. At both the 1000 and 2000 count level, the
272
addition of a second component has no effect on the C-statistic (the dis-
tributions in C-space for both the power law fit and the two-component
fit are indistinguishable).
At the 4000 count level, the C-statistic distributions for the two models
separate such that there is an 26% confidence that the two-component
model is a better fit to the data. We find, when we examine the model
parameters, that the power law index (?) for the two-component model
ranges between 1.19?2.12 with the mean value? 1.74. The mean value
for all three count levels used clustered around this value, though the
range in ? increased as the counts decreased. The mean blackbody
temperature for simulations with 4000 counts was 1.17keV (with the
range varying between the amount previously noted) with a median
of 1.06keV. For simulations with lower counts, the blackbody temper-
ature becomes higher (1.24keV for 1000 counts) with a higher median
(1.83keV for 1000 counts). This tells us that the fitting procedure tends
to approximate a pure power law spectrum as a two-component spec-
trum with ? equal to that of the true spectral index but with a black-
body temperature higher than those observed in our study (1.1keV or
higher) which in the XMM band can be approximated as a power law.
If we found spectra in our sample that were best fit with a low spec-
tral index and a high blackbody temperature, we might suspect that the
spectrum?s true nature is a power law. We also note that if the hydrogen
column density is large, much > 3 ? 1021 cm?2, a low temperature (kT)
blackbody component can be much more difficult to detect.
273
B.2 Additional Spectral Fits
The following sources were not best fit by the standard models em-
ployed in this study:
B.2.1 NGC300 XMM4
Thissourcewasclassifiedasasuper-softX-raysourcebyKong & Di Stefano
(2003). We find that the standard single-component absorbed blackbody
model is a much better model for this spectrum. In fact, the power law,
bremsstrahlung, and combined models do not fit the data within the
90% confidence range. Fitting an absorbed blackbody, we find the best
fit corresponds to the following parameters: nH = 1.38+0.27?0.55 ?1021 cm?2,
kT = 0.059+0.007?0.005 keV, and ?2/dof = 74.5/45. This fit yields an unab-
sorbed flux of 3.3?10?13 ergcm?2 s?1.
B.2.2 NGC4631 XMM4
The spectrum of this source clearly identifies it as a super-soft X-ray
source. As with NGC300 XMM4, the standard models employed in this
study did not adequately match the data. The best fitting model cor-
responds again to an absorbed blackbody. The corresponding parame-
ters are as follows: nH = 6.2+0.26?1.5 ? 1021 cm?2, kT = 0.07+0.01?0.01 keV, and
?2/dof = 142.3/74. This fit yields an unabsorbed flux of 9.5 ? 10?12
ergcm?2 s?1. The position of this source shows it to be coincident with
a globular cluster associated with that galaxy. This source was identi-
fied as a bulge X-ray source, possibly powered by accretion, in a ROSAT
study of NGC4631 (Vogler & Pietsch 1996).
274
B.2.3 NGC4631 XMM5
The spectrum of this source was best fit with an absorbed power law
+ an absorbed vapec model. This indicates the presence of hot gas,
indicating a possible thermal X-ray source.
B.2.4 NGC4945 XMM5
The spectrum of this source was not adequately fit with any of the stan-
dard models used in this investigation. The spectrum exhibits a promi-
nent Fe K line in the PN spectrum that is well fit by a gaussian (zgauss)
at6.4keV.Wefindthattheentirespectrumisbestfitwithapartialcover-
ing fractionabsorptionmodel(pcfabs)incombinationwiththenormal
absorption, a power law, and a gaussian. The best fit parameters yield:
absorption column density, nH = 1.79 ? 1021 cm?2, partial covering ab-
sorption, nH = 18.4?1021 cm?2, partialcovering fraction = 0.82, ? = 1.6,
and ?2/dof = 61.8/57. The source is clearly located within the optical
galaxy, and is thus unlikely to be a background AGN.
B.2.5 M51 XMM5
The spectrum and luminosity (LX ? 1.9?1042 ergs?1) ofthis source sug-
gests that it is an AGN. The location of the source, from the Digital Sky
Survey, places it within the dwarf companion of M51 making a value
of the optical flux hard to constrain. The best fit to this source was an
absorbed blackbody + power law and the spectral parameters are listed
in Table 2.4.
275
B.2.6 M83 XMM2
Like NGC4945 XMM5, this source was best fit by a partial absorption
model. However, this source showed no evidence of an Fe K line. We fit
this source?s spectra using a partial covering fraction absorption model
in combination with the normal absorption model and a power law.
The best fit parameters yield: absorption column density, nH = 2.1 ?
1021 cm?2, partial covering absorption, nH = 43.5 ? 1021 cm?2, partial
covering fraction = 0.86, ? = 2.95, and ?2/dof = 83.5/84. The unab-
sorbed flux in the range of 0.3-10 keV equals 1.37?10?12 ergcm?2 s?1.
B.2.7 Inverse Compton Scattering Sources
Table B.4 includes the parameters for the ?ULX? sources best fit by the
compST model. A discussion of these sources and interpretation of the
data is included in section 2.4.3.
B.3 Additional Tables
Inthisappendixweincludeadditionaltablescontainingsupportingma-
terial for the text of the paper. In Table B.1 we list details of the XMM-
Newton observations for all of the point sources examined in this study.
Table B.2 lists information for bright sources that we excluded from
our study, due to their classification as either foreground stars or back-
ground AGNs. Table B.3 includes single-component (absorbed power
law) model fits for the sources listed in Table 2.4, while Table B.4 in-
cludes Comptonization model fit parameters for the sources mentioned
in Section 2.4.3 (possible very high state stellar mass black holes with
276
luminosities in the ULX regime). Finally, Table B.5 includes the IR flux
information for ULX host galaxies used to construct Figure 2.10.
277
Ta
ble
B.1
:B
rig
ht
Po
int
So
ur
ce
sE
xa
mi
ne
di
nt
he
UL
X
Su
rv
ey
.
So
ur
ce
1
RA
(h
m
s)
De
c(
?
??
? )
To
tal
Co
un
ts
Co
un
tR
ate
2
ID
Lo
ca
tio
ni
ng
ala
xy
3
XM
M
ref
7
NG
C2
47
XM
M
1
00
47
03
.8
-20
47
46
.2
34
58
,1
38
9,
13
79
20
.33
,5
.8,
6.4
1R
XS
J00
47
04
.8-
20
47
43
sa
-
NG
C2
47
XM
M
2
00
47
03
.1
-20
37
02
.5
-,5
97
,6
00
-,1
.9,
1.4
-
sa
-
NG
C2
53
XM
M
1
00
47
32
.8
-25
17
52
.6
-,3
15
6,
29
85
-,1
1.3
8,
9.9
NG
C2
53
PS
X-
24
ne
ar
ce
nt
er
1
??
?
??
?
??
?
-,1
26
54
,1
28
12
-,8
.7,
9.1
??
?
-
NG
C2
53
XM
M
2
00
47
22
.4
-25
20
55
.2
-,8
25
,9
42
-,2
.8,
2.9
7
NG
C2
53
PS
X-
5
sa
1
??
?
??
?
??
?
-,1
03
47
,1
03
04
-,8
.2,
8.5
??
?
-
NG
C2
53
XM
M
3
00
47
35
.2
-25
15
13
.8
-,8
70
,1
06
5
-,3
,3
.41
NG
C2
53
PS
X-
7
sa
1
??
?
??
?
??
?
-,5
98
8,
61
31
-,4
.2,
4.6
??
?
-
NG
C2
53
XM
M
4
00
47
23
.3
-25
19
06
.5
-,6
49
,7
03
-,1
.4,
1.3
-
sa
-
??
?
??
?
??
?
-,3
82
3,3
73
8
-,1
.7,
1.9
5
-
-
NG
C2
53
XM
M
5
00
47
17
.6
-25
18
12
.1
-,2
95
,3
13
-,1
.08
,1
.04
NG
C2
53
PS
X-
4
sa
2
??
?
??
?
??
?
-,4
19
9,
43
03
-,3
.6,
3.8
??
?
-
NG
C2
53
XM
M
6
00
47
42
.8
-25
15
05
.5
-,6
08
1,
64
07
-,3
.8,
4.4
NG
C
02
53
[V
P9
9]
X4
0
sa
1
NG
C2
53
XM
M
75
00
47
09
.2
-25
21
21
.7
-,4
30
0,
44
54
-,2
.6,
3.0
-
sa
-
NG
C3
00
XM
M
1
00
55
09
.9
-37
42
13
.9
67
78
,2
24
8,
24
53
18
.6,
4.9
,5
.4
-
sa
-
NG
C3
00
XM
M
2
00
55
10
.6
-37
48
36
.7
13
64
,4
56
,4
63
3.1
,0
.9,
0.9
-
ed
ge
sa
?
-
NG
C3
00
XM
M
3
00
54
49
.7
-37
38
53
.8
91
5,
44
2,
43
5
2.2
,0
.9,
0.9
-
sa
6
-
NG
C3
00
XM
M
4
00
55
10
.9
-37
38
53
.8
74
5,
22
4,
23
3
1.9
,0
.3,
0.4
XM
M
U
J00
55
11
-37
49
;S
SS
sa
3
NG
C3
00
XM
M
5
00
55
21
.1
-37
29
19
.5
75
0,
24
7,
25
0
1.7
,0
.4,
0.5
-
ou
t
-
NG
C3
00
XM
M
6
00
54
44
.2
-37
51
04
.5
51
7,
18
7,
16
5
1.1
,0
.3,
0.3
-
ou
t
-
NG
C6
25
XM
M
18
01
35
06
.8
-41
26
17
.1
58
32
,5
77
,2
11
9
3,
0.6
,1
.4
-
sa
-
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
278
Ta
ble
B.1
?C
on
tin
ue
d
So
ur
ce
1
RA
(h
m
s)
De
c(
?
??
? )
To
tal
Co
un
ts
Co
un
tR
ate
2
ID
Lo
ca
tio
ni
ng
ala
xy
3
XM
M
ref
7
NG
C1
31
3X
M
M
1
03
18
19
.9
-66
29
10
.7
28
76
,9
00
,8
10
8.6
,3
.2,
2.8
NG
C1
31
3[
CP
S9
5]
X-
1;
IX
O
07
sa
4,5
NG
C1
31
3X
M
M
2
03
17
38
.8
-66
33
05
.3
75
68
,2
35
7,
21
08
25
.4,
8.8
,7
.8
NG
C1
31
3[
CP
S9
5]
X-
3;
SN
19
78
K
ed
ge
sa
6
NG
C1
31
3X
M
M
3
03
18
22
.5
-66
36
06
.2
69
60
,2
17
9,
17
93
23
.3,
8.1
,6
.6
NG
C1
31
3[
CP
S9
5]
X-
2;
IX
O
08
ed
ge
sa
?
4,5
NG
C1
31
3X
M
M
4
03
18
18
.5
-66
30
05
20
75
,6
59
,5
67
5.1
,1
.8,
1.5
NG
C1
31
3[
SP
C2
00
0]
X-
8
ne
ar
ce
nt
er
-
IC
03
42
XM
M
1
03
45
55
.8
+6
80
45
4.5
18
02
,1
21
6,
11
05
33
.6,
12
.4,
11
.1
IC
03
42
[R
W
20
00
]X
-1;
IX
O
22
sa
5,
7,
8
IC
03
42
XM
M
2
03
46
15
.0
+6
81
11
1.2
11
47
,5
41
,1
84
21
.1,
5.5
,1
.7
IC
03
42
[R
W
20
00
]X
-3
sa
7,
8
IC
03
42
XM
M
3
03
46
48
.6
+6
80
54
3.2
11
86
,6
70
,6
06
21
.7,
6.8
,6
.0
IC
03
42
[L
LJ
20
00
]X
-2
ne
ar
ce
nt
er
7,
8
IC
03
42
XM
M
4
03
46
57
.2
+6
80
62
0.2
55
1,
33
8,
37
7
9.5
,3
.9,
3.6
IC
03
42
[R
W
20
00
]X
-6
ne
ar
ce
nt
er
7,
8
NG
C1
56
99
??
?
??
?
??
?
??
?
??
?
??
?
??
?
NG
C1
70
5X
M
M
1
04
54
57
.6
-53
24
23
.5
11
74
,4
00
,3
71
2.4
,0
.6,
0.6
RX
J04
54
.9-
53
24
ou
t?
-
NG
C1
70
5X
M
M
2
04
54
19
.6
-53
20
41
.9
93
3,
37
5,
39
7
1.9
,0
.6,
0.6
??
?
ou
t?
-
NG
C1
70
5X
M
M
3
04
54
38
.1
-53
18
16
.2
69
8,
37
2,
41
8
1.4
,0
.6,
0.6
W
GA
J04
54
.7-
53
18
ou
t?
-
M
RK
71
XM
M
1
07
28
51
.8
+6
90
72
7
83
2,
22
5,
20
7
3.9
,0
.97
,0
.9
-
ou
t
-
NG
C2
40
3X
M
M
1
07
36
25
.6
+6
53
54
0
-,1
19
9,
67
2
-,1
.0,
0.6
0
NG
C2
40
3[
RW
20
00
]X
-1
ed
ge
sa
-
NG
C2
40
3X
M
M
2
07
36
50
.2
+6
53
60
2.1
19
64
,7
29
,6
72
1.9
9,
0.6
3,
0.6
0
-
ne
ar
ce
nt
er
-
NG
C2
40
3X
M
M
3
07
36
55
.4
+6
53
54
0.3
14
97
,4
89
,3
78
1.4
,0
.40
,0
.33
-
ne
ar
ce
nt
er
-
NG
C2
40
3X
M
M
4
07
37
02
.5
+6
53
93
5.2
10
04
,2
74
,2
88
0.5
2,
0.1
5,
0.2
1
NG
C2
40
3[
RW
20
00
]X
-4
ed
ge
sa
?
-
Ho
lII
XM
M
1
08
19
28
.8
+7
04
22
0.3
31
05
2,
12
57
,1
08
07
27
2.2
,7
5.5
,7
2.7
Ho
lm
II
X-
1;
IX
O
31
10
ne
ar
ce
nt
er?
9
??
?
??
?
??
?
38
53
,1
36
1,
14
52
78
.3,
18
.9,
20
.5
??
?
-
Ho
lI
XM
M
1
09
41
30
+7
11
23
4
68
7,
76
8,
75
4
2.9
,2
.7,
2.6
-
ou
t?
-
Ho
lI
XM
M
2
09
39
59
.7
+7
10
64
0.2
57
5,
20
3,
22
4
2.5
,0
.7,
0.7
1W
GA
J09
40
.0+
71
06
ou
t?
-
Ho
lI
XM
M
3
09
42
06
.7
+7
10
44
5.3
45
2,
14
1,
14
1
1.7
,0
.4,
0.4
-
ou
t?
-
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
279
Ta
ble
B.1
?C
on
tin
ue
d
So
ur
ce
1
RA
(h
m
s)
De
c(
?
??
? )
To
tal
Co
un
ts
Co
un
tR
ate
2
ID
Lo
ca
tio
ni
ng
ala
xy
3
XM
M
ref
7
M
81
XM
M
1
09
55
32
.9
+6
90
03
4.8
50
78
8,
-,1
89
88
51
.1,
-,2
1.7
NG
C3
03
1[
RW
20
00
]X
-11
sa
-
??
?
??
?
??
?
-,1
22
7,
-11
-,1
2,
13
.1
??
?
-
-
M
81
XM
M
2
09
55
24
.8
+6
90
11
1.7
17
87
1,
-,4
12
1
13
.2,
-,4
.4
SN
19
93
J
sa
10
M
81
XM
M
3
09
55
10
.6
+6
90
50
2.2
-,-
,1
97
0
-,-
,1
.5
NG
C3
03
1[
RW
20
00
]X
-05
sa
-
M
81
XM
M
4
09
55
24
.3
+6
91
00
0.2
-,-
,1
19
7
-,-
,1
.0
NG
C3
03
1[
RW
20
00
]X
-08
ed
ge
sa
-
M
81
XM
M
5
09
55
49
.2
+6
90
53
0.5
-,-
,2
07
7
-,-
,2
.3
-
sa
-
M
82
12
??
?
??
?
??
?
??
?
??
?
??
?
??
?
Ho
lIX
XM
M
1
09
57
53
.3
+6
90
34
8.7
14
97
6,
65
46
,6
58
6
20
7.3
,6
4.3
,6
5
Ho
lIX
X-
1;
IX
O
34
?
4,
5
Se
xta
ns
A
XM
M
1
10
11
24
.6
-04
42
17
.2
39
63
,1
32
3,
12
42
17
.2,
5,
4.5
-
ou
t
-
IC
25
74
XM
M
1
10
28
42
.4
+6
82
81
7.8
10
47
,6
73
,6
23
8.3
,2
.6,
2.5
-
sa
-
IC
25
74
XM
M
2
10
26
33
.5
+6
82
93
2.1
53
3,
33
5,
30
0
4.3
,1
.3,
1.2
-
ou
t
-
IC
25
74
XM
M
3
10
27
22
.2
+6
81
84
7.6
53
8,
29
3,
30
1
4.2
,1
.1,
1.2
-
ou
t
-
NG
C4
21
4X
M
M
1
12
15
37
.0
+3
61
92
9.4
43
4,
23
0,
22
5
2.9
,1
.6,
1.5
NG
C4
21
4[
HS
S2
00
4]
11
sa
-
NG
C4
21
4X
M
M
2
12
15
58
.2
+3
62
23
8.5
62
6,
16
0,
28
5
3.4
,0
.6,
1.5
-
ou
t
-
NG
C4
25
8X
M
M
15
12
18
47
.8
+4
72
05
1.7
82
8,
47
0,
44
4
7.0
,2
.8,
2.8
-
sa
-
??
?
??
?
??
?
??
?
??
?
??
?
-
-
NG
C4
25
8X
M
M
2
12
18
57
.8
+4
71
60
6.8
73
2,
33
7,
33
7
5.8
,2
,2
.1
NG
C4
25
8[
RW
20
00
]X
-7
sa
-
??
?
??
?
??
?
71
6,
26
8,
29
0
1.8
,1
,1
.0
??
?
-
NG
C4
25
8X
M
M
3
12
18
56
.5
+4
72
12
4.3
48
9,
18
2,
18
4
3.8
,0
.9,
1.1
NG
C4
25
8[
RW
20
00
]X
-5
sa
-
??
?
??
?
??
?
-,1
60
,1
42
-,0
.3,
0.3
??
?
-
NG
C4
25
8X
M
M
4
12
19
23
.2
+4
70
93
7.2
64
4,
27
7,
22
4
4.9
,1
.3,
1.3
HE
LL
AS
28
8
ed
ge
sa
?
-
??
?
??
?
??
?
96
4,
41
0,
38
2
0.3
,0
.2,
0.1
??
?
-
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
280
Ta
ble
B.1
?C
on
tin
ue
d
So
ur
ce
1
RA
(h
m
s)
De
c(
?
??
? )
To
tal
Co
un
ts
Co
un
tR
ate
2
ID
Lo
ca
tio
ni
ng
ala
xy
3
XM
M
ref
7
NG
C4
39
5X
M
M
1
12
26
01
.5
+3
33
12
9
21
62
,8
62
,9
21
15
.5,
5.1
,5
.6
NG
C4
39
5[
RW
20
00
]X
-1;
IX
O
53
sa
-
NG
C4
39
5X
M
M
2
12
25
25
.3
+3
33
64
6.4
39
2,
25
2,
20
8
2.6
,1
.5,
1.2
-
sa
?
-
NG
C4
39
5X
M
M
3
12
25
32
.6
+3
32
52
7.9
76
3,
27
7,
27
8
5.2
,1
.5,
1.6
-
ou
t?
-
NG
C4
39
5X
M
M
4
12
25
42
.7
+3
34
00
0.1
51
6,
60
,4
0
3.1
,0
.3,
0.2
-
ou
t
-
NG
C4
44
9X
M
M
1
12
28
18
+4
40
63
0.9
14
09
,6
08
,5
93
10
.6,
3.6
,3
.5
XR
B?
;N
GC
44
49
[R
W
20
00
]X
-7;
So
ur
ce
27
13
sa
-
NG
C4
44
9X
M
M
2
12
28
09
.3
+4
40
50
3.9
15
03
,5
27
,5
86
11
.3,
3.1
,3
.5
SN
R?
;N
GC
44
49
[R
W
20
00
]X
-1;
So
ur
ce
10
sa
-
NG
C4
44
9X
M
M
3
12
28
11
.1
+4
40
64
3.9
10
94
,5
49
,4
04
8.3
,3
.2,
2.3
SN
R;
So
ur
ce
15
sa
-
NG
C4
49
0X
M
M
1
12
30
32
.4
+4
13
91
4.6
74
6,
32
3,
39
3
6.0
,1
.9,
2.2
NG
C4
49
0[
RW
20
00
]X
-1
sa
-
NG
C4
49
0X
M
M
2
12
30
36
.5
+4
13
83
3.3
65
6,
29
9,
31
0
5.3
,1
.7,
1.7
NG
C4
49
0[
RW
20
00
]X
-2
ne
ar
ce
nt
er
-
NG
C4
49
0X
M
M
3
12
30
43
.3
+4
13
81
1.5
83
2,
50
1,
46
1
5.8
,2
.6,
2.2
NG
C4
49
0[
RW
20
00
]X
-4
sa
-
NG
C4
49
0X
M
M
4
12
30
31
.1
+4
13
90
8.1
54
6,
29
1,
28
6
4.4
,1
.7,
1.6
CX
OU
J12
30
30
.8
+4
13
91
1
sa
-
NG
C4
49
0X
M
M
5
12
30
30
.3
+4
14
14
0.3
41
3,
58
7,
48
2
2.7
,3
.1,
2.3
NG
C4
48
5[
RW
20
00
]X
-1
sa
-
NG
C4
63
1X
M
M
1
12
41
55
.8
+3
23
21
4
50
93
,1
96
9,
17
62
13
,4
.2,
3.8
NG
C4
63
1[
RW
20
00
]X
-1;
IX
O
68
sa
-
NG
C4
63
1X
M
M
2
12
41
57
.5
+3
23
20
1
12
73
,5
31
,4
00
3.3
,1
.2,
0.9
NG
C4
63
1[
RW
20
00
]X
-2
sa
-
NG
C4
63
1X
M
M
3
12
41
58
.2
+3
22
84
9.6
12
71
,4
43
,4
57
2.8
,0
.9,
0.9
CX
OU
SE
XS
IJ
12
41
58
.0+
32
28
51
ou
t
-
NG
C4
63
1X
M
M
4
12
42
16
.1
+3
23
24
8.8
95
7,
40
0,
42
9
2.3
,0
.8,
0.8
[V
P9
6]
H1
3
sa
-
NG
C4
63
1X
M
M
5
12
42
11
.2
+3
23
23
3.6
16
26
,8
94
,8
66
3.9
,1
.9,
1.8
[V
P9
6]
H1
2;
[H
FE
20
03
]P
SX
-01
sa
-
NG
C4
73
6X
M
M
1
12
50
50
.2
+4
10
71
2
71
3,
27
3,
22
7
6.7
,2
.1,
1.7
NG
C4
73
6X
-41
4
ne
ar
ce
nt
er
-
NG
C4
94
5X
M
M
1
13
05
33
.3
-49
27
36
.3
14
56
,6
46
,5
95
7.9
,2
.9,
2.6
NG
C4
94
5[
GM
B2
00
0]
X-
2
sa
-
NG
C4
94
5X
M
M
2
13
05
38
.4
-49
25
45
.3
13
93
,6
00
,5
23
7.6
,2
.7,
2.3
NG
C4
94
5[
R9
7]
X-
3
sa
-
NG
C4
94
5X
M
M
3
13
05
18
.8
-49
28
24
-,3
57
,3
62
-,1
.6,
1.6
??
?
sa
-
NG
C4
94
5X
M
M
4
13
05
22
.2
-49
28
26
.3
73
1,
33
1,
33
2
4.0
,1
.5,
1.5
NG
C4
94
5[
BI
R9
6]
X-
1?
sa
-
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
281
Ta
ble
B.1
?C
on
tin
ue
d
So
ur
ce
1
RA
(h
m
s)
De
c(
?
??
? )
To
tal
Co
un
ts
Co
un
tR
ate
2
ID
Lo
ca
tio
ni
ng
ala
xy
3
XM
M
ref
7
NG
C4
94
5X
M
M
5
13
05
25
.7
-49
28
30
.7
77
2,
26
7,
30
1
4.2
,1
.2,
1.3
2
??
?
sa
-
NG
C5
20
4X
M
M
1
13
29
38
.5
+5
82
50
3.6
99
81
,3
35
2,
33
84
62
.8,
17
.7,
17
.9
NG
C5
20
4[
RW
20
00
]X
-1;
IX
O
77
sa
11
??
?
??
?
??
?
92
31
,2
28
4,
23
49
85
.5,
24
.7,
25
.8
??
?
-
NG
C5
20
4X
M
M
2
13
29
27
.4
+5
82
53
1.8
57
3,
17
0,
23
1
3.4
,0
.8,
1.2
-
ed
ge
sa
-
??
?
??
?
??
?
77
2,
16
1,
12
1
5.0
,1
.7,
1.1
??
?
-
M
51
XM
M
1
13
29
40
+4
71
23
6.2
11
02
,3
67
,4
09
6.2
,1
.8,
2
NG
C5
19
4[
RW
20
00
]X
-1
sa
-
M
51
XM
M
2
13
30
07
.7
+4
71
10
4.8
51
4,
54
0,
54
9
2.6
,2
.2,
2.2
IX
O
81
sa
-
M
51
XM
M
3
13
30
01
.1
+4
71
34
1.4
10
04
,3
15
,3
11
4.1
,1
.1,
1.1
CX
OU
J13
30
01
.0
+4
71
34
4;
IX
O
80
sa
-
M
51
XM
M
4
13
30
06
+4
71
53
8.9
51
8,
18
3,
16
6
2.8
,0
.9,
0.8
NG
C
51
95
[R
W
20
00
]X
-1
sa
-
M
51
XM
M
5
13
29
59
.6
+4
71
55
4
10
79
,3
59
,2
57
5,
1.5
,1
CX
OU
J13
29
59
.5
+4
71
55
9
ne
ar
ce
nt
er
-
M
51
XM
M
6
13
29
57
.5
+4
71
04
5.3
53
6,
20
6,
24
7
1.9
,0
.6,
0.8
CX
OU
J13
29
57
.6
+4
71
04
8
sa
-
M
51
XM
M
7
13
29
53
.6
+4
71
43
1.5
45
2,
14
1,
14
3
2.4
,0
.6,
0.7
CX
OU
J13
29
53
.8
+4
71
43
2
ed
ge
sa
-
M
83
XM
M
1
13
37
19
.8
-29
53
49
.8
30
74
,9
27
,9
87
12
,3
.3,
2.5
RX
J13
37
19
-29
53
.6;
IX
O
82
sa
-
M
83
XM
M
2
13
36
59
.4
-29
49
57
.2
11
33
,3
71
,3
97
4,
1.3
,1
CX
OU
J13
36
59
.4
-29
49
59
sa
-
M
83
XM
M
31
5
13
37
04
.4
-29
51
24
17
24
,5
76
,4
59
7.2
,2
.3,
1.2
CX
OU
J13
37
04
.3
-29
51
21
sa
-
M
83
XM
M
4
13
37
01
.5
-29
53
26
12
89
,3
45
,4
01
4.9
,1
.3,
1.0
CX
OU
J13
37
01
.4
-29
53
26
sa
-
NG
C5
25
31
6
??
?
??
?
??
?
??
?
??
?
??
?
??
?
M
10
1X
M
M
1
14
03
14
.7
+5
41
80
5
26
90
,1
44
9,
14
17
10
.3,
3.3
,3
.3
CX
OU
J14
03
13
.9
+5
41
81
1;
XM
M
-21
7
sa
12
M
10
1X
M
M
2
14
03
03
.8
+5
42
73
7
28
25
,1
62
3,
15
51
10
.3,
3.6
,3
.3
XM
M
-1
ed
ge
sa
?
12
M
10
1X
M
M
3
14
04
14
.6
+5
42
60
4.4
14
60
,8
22
,7
17
5,
1.6
,1
.3
CX
OU
J14
04
14
.3
+5
42
60
4;
XM
M
-3
ed
ge
sa
?
12
M
10
1X
M
M
4
14
02
28
.5
+5
41
62
6.7
15
05
,8
77
,7
57
5.1
,1
.7,
1.5
CX
OU
J14
02
28
.3
+5
41
62
6
sa
12
M
10
1X
M
M
5
14
02
22
.5
+5
41
75
8
51
6,
24
5,
28
9
1.8
,0
.4,
0.6
CX
OU
J14
02
22
.2
+5
41
75
6;
XM
M
-6
sa
12
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
282
Ta
ble
B.1
?C
on
tin
ue
d
So
ur
ce
1
RA
(h
m
s)
De
c(
?
??
? )
To
tal
Co
un
ts
Co
un
tR
ate
2
ID
Lo
ca
tio
ni
ng
ala
xy
3
XM
M
ref
7
NG
C5
40
8X
M
M
1
14
03
19
.8
-41
22
59
.3
59
32
,2
03
6,
20
77
12
.8,
3.2
,3
.3
NG
C5
40
8[
KC
P2
00
3]
X-
1
sa
5
CI
RC
IN
US
XM
M
11
8
14
12
54
.2
-65
22
55
.3
16
22
0,
11
45
2,
-
14
.5,
11
.3,
-
-
ed
ge
sa
?
-
CI
RC
IN
US
XM
M
2
14
12
39
.2
-65
23
34
.3
87
41
,2
38
6,
-
5.7
,1
.9,
-
-
ed
ge
sa
?
-
CI
RC
IN
US
XM
M
3
14
13
28
.3
-65
18
08
.3
48
73
,1
03
1,
-
1.8
,0
.7,
-
-
ed
ge
sa
?
-
1 S
ou
rce
sl
ab
ele
dX
M
M
-n
in
or
de
ro
fa
pp
are
nt
br
igh
tn
es
sf
ro
m
th
efi
rst
ob
se
rv
ati
on
stu
di
ed
2 C
ou
nt
rat
eu
nit
so
f?
10
?
2
cts
s?
1
for
th
eP
N,
M
OS
1,
an
dM
OS
2s
pe
ctr
a
3 L
oc
ati
on
sp
ec
ifie
d
as
:i
ns
id
et
he
op
tic
al
ex
ten
to
ft
he
ga
lax
yo
ri
ns
pir
al
arm
s(
sa
),
ne
ar
th
ec
en
ter
of
th
eg
ala
xy
,a
tt
he
ed
ge
of
as
pir
al
arm
/
op
tic
al
ex
ten
to
fg
ala
xy
,o
ro
ut
sid
et
he
op
tic
al
ex
ten
to
ft
he
ga
lax
y.
Lo
ca
tio
nb
as
ed
on
DS
Si
ma
ge
s.
4 I
de
nt
ific
ati
on
for
NG
C0
25
3f
oll
ow
ing
Hu
mp
hr
ey
et
al.
(20
03
).
5 T
ran
sie
nt
.
6 A
pp
ea
rs
as
an
ex
ten
de
ds
ou
rce
in
HS
Ti
ma
ge
.
7 R
efe
ren
ce
st
os
tu
di
es
us
ing
XM
M
-N
ew
ton
sp
ec
tra
8 S
pe
ctr
at
oo
sca
tte
red
to
mo
de
l.
9 B
rig
ht
so
ur
ce
sc
oin
cid
ew
ith
nu
cle
us
(u
nr
es
olv
ab
le
sta
rc
lus
ter
an
dX
-ra
yb
ina
rie
s),
for
e-g
ro
un
ds
tar
,a
nd
ba
ck
gr
ou
nd
AG
N
(M
art
in
et
al.
20
02
).
10
IX
O
de
sig
na
tio
nf
ro
m
Co
lbe
rt
&
Pt
ak
(20
02
).
11
Sp
ec
tru
m
fro
m
Ho
lIX
ob
se
rv
ati
on
.
12
Br
igh
ts
ou
rce
is
too
clo
se
to
oth
er
so
ur
ce
s.
13
Ch
an
dr
ao
bs
erv
ati
on
so
fp
oin
ts
ou
rce
si
nN
GC
44
49
pu
bli
sh
ed
in
Su
mm
ers
et
al.
(20
03
);S
ou
rce
27
va
rie
df
ro
m
RO
SA
To
bs
erv
ati
on
s
14
Th
ree
oth
er
so
ur
ce
sn
ea
rt
he
nu
cle
us
are
un
res
olv
ea
ble
in
th
eX
M
M
ob
s.,
bu
ts
ee
nb
yC
ha
nd
ra
(E
rac
leo
us
et
al.
20
02
).
15
Un
ab
le
to
mo
de
ls
pr
ec
tru
m
du
et
oa
ne
rro
ri
n?
2 -
sp
ac
e
283
16
Br
igh
ts
ou
rce
sa
re
too
clo
se
,b
ut
res
olv
ab
le
by
Ch
an
dr
a(
Su
mm
ers
et
al.
20
04
).
17
Al
ter
na
te
ID
fro
m
Jen
kin
se
ta
l.(
20
04
).
18
Br
igh
ts
ou
rce
sn
ea
rt
he
nu
cle
us
are
un
res
olv
ab
le,
bu
ts
ee
nb
yC
ha
nd
ra
(Sm
ith
&
W
ils
on
20
01
).
Ta
bl
eR
efe
ren
ce
s:
(1)
Pie
tsc
he
ta
l.(
20
01
);(
2)
Pie
tsc
he
ta
l.(
20
03
);(
3)
Ko
ng
&
Di
Ste
fan
o(
20
03
);(
4)
M
ille
re
ta
l.(
20
04
b)
;(5
)W
an
ge
ta
l.(
20
04
);(
6)
Sc
hle
ge
le
ta
l.(
20
04
);(
7)
Ba
ue
re
ta
l.(
20
03
a);
(8)
Ko
ng
(20
03
);(
9)
Le
hm
an
ne
ta
l.(
20
05
);(
10
)Z
im
me
rm
an
n&
As
ch
en
ba
ch
(20
03
);(
11
)R
ob
ert
se
ta
l.(
20
05
);(
12
)J
en
kin
se
ta
l.(
20
04
).
284
Table B.2. Bright, Identifiable Background and Foreground Sources
Galaxy RA (h m s) Dec (????) Identification
NGC 247 0 46 51.7 -20 43 30 QSO B044-2059
NGC 300 0 55 26.7 -37 31 25.6 HD 5403 (Star)
NGC 625 01 34 42.4 -41 36 15.2 QSO B0132-4151
NGC 1569a 04 31 16.9 +64 49 50 CXOU J043116.8+644950 (Star)
NGC 1569 04 31 14.2 +64 51 07.9 CXOU 043114.0+645107 (Star)
NGC 1569 04 31 25.4 +64 51 53.8 CXOU 043125.1+645154 (AGN)
NGC 1705 04 54 01.2 -53 21 12.3 WGA J0454.0-5320 (M star or elliptical galaxy)
NGC 2403 07 35 09 +65 40 27.5 HD 59581 (Star)
NGC 4258 12 18 08.9 +47 16 08.3 QSO J1218+472
M83 13 36 45.6 -29 59 13.9 2MASX J13364579-2959122 (Galaxy)
M83 13 36 13.9 -29 56 13 RX J133615-2957.8 (Galaxy)
NGC 5253 13 39 50.6 -31 34 11.1 CD-30 10790 (Star)
M101 14 02 30 +54 21 18.2 [WIP99] H13 (Star)b
NGC 5408 14 03 27.5 -41 25 18.5 (Star)
aidentification for objects in NGC 1569 from Martin, Kobulnicky, & Heckman (2002)
bconfirmed by K. Kuntz using HST ACS
285
Table B.3: XMM-Newton Power Law Fit for Best Fit Two-component
Spectra (wabs*pow))
Source nHa ? ?2/dof FXc LXd
NGC247 XMM1 9.5+1.6?1.4 8.52+1.14?0.91 112.2/95 1900 2200
NGC253 XMM1 3.4+0.3?0.3 1.77+0.06?0.06 262.6/232 3.1 3.6
??? 6.9+0.4?0.4 1.98+0.05?0.05 611.6/582 2.9 3.3
NGC253 XMM2 (obs 2) 2.2+0.1?0.1 2.03+0.04?0.04 507.4/500 1.2 1.4
NGC253 XMM3 3.9+0.5?0.4 2.17+0.14?0.11 91.9/83 0.73 1.2
??? 4.0+0.4?0.3 2.06+0.09?0.07 381.8/409 0.98 1.6
NGC253 XMM4 8.5+3.0?2.3 2.09+0.33?0.28 73.6/59 0.52 0.85
??? 1.2+0.3?0.3 2.09+0.16?0.15 321.4/293 0.29 0.48
NGC253 XMM5 1.7+1.2?0.9 1.54+0.22?0.20 31.8/25 0.32 0.53
??? 3.4+0.2?0.2 2.17+0.7?0.7 283.8/298 1.1 1.3
NGC253 XMM6 3.9+0.3?0.3 2.21+0.84?0.80 435/409 0.93 1.5
NGC253 XMM7 7.1+0.7?0.7 2.15+0.11?0.11 357/342 1.3 2.2
NGC300 XMM1 0.97+0.11?0.11 2.67+0.06?0.06 469.8/422 0.81 0.6
NGC300 XMM2 1.7+0.40?0.40 3.20+0.32?0.24 133.98/99 0.27 0.21
NGC300 XMM3 3.4+0.8?0.6 1.86+0.15?0.13 101.9/81 0.19 0.15
NGC300 XMM5 0.311e 2.29+0.15?0.14 54.2/56 0.14 0.10
NGC300 XMM6 ?+??? 2.05+0.17?0.15 47.6/37 0.06 0.15
NGC1313 XMM1 1.5+0.2?0.2 1.81+0.08?0.09 219.8/203 0.42 0.88
NGC1313 XMM2 2.8+0.16?0.16 2.48+0.07?0.06 464.1/421 1.9 4.0
NGC1313 XMM3 3.6+0.2?0.2 3.2+0.09?0.09 778.3/426 3.3 6.9
IC0342 XMM3 3.8+0.4?0.4 2.58+0.15?0.14 185.8/109 1.7 3.1
NGC1705 XMM1 0.3e 1.93+0.11?0.10 61.9/88 0.12 0.37
Continued on Next Page...
286
Table B.3 ? Continued
Source nHa ? ?2/dof FXc LXd
NGC1705 XMM2 1.4+0.45?0.41 2.12+0.26?0.15 91/76 0.078 0.24
NGC1705 XMM3 0.6+0.36?0.40 1.36+0.12?0.13 80.9/67 0.17 0.53
NGC2403 XMM1 3.2+0.61?0.55 2.15+0.16?0.15 92.2/81 2.2 3.6
NGC2403 XMM2 2.7+0.37?0.34 2.07+0.11?0.11 179.5/151 1.3 2.0
NGC2403 XMM3 1.9+0.40?0.36 1.97+0.14?0.13 92.6/107 0.81 1.3
HolmII XMM1 (obs 1) 1.5+0.07?0.07 2.61+0.04?0.04 1134.2/976 12 10
Holm I XMM1 ?+??? 2.04+0.14?0.07 102.8/95 0.48 1.7
M81 XMM1 3.2+0.07?0.07 2.09+0.02?0.02 1849.9/1245 4.5 7.0
??? 3.0+0.3?0.3 1.79+0.07?0.07 224.9/208 4.3 6.7
M81 XMM2 7.3 6.13 1358.2/618 48.5 75.2
M81 XMM3 0.97+0.25?0.41 1.58+0.18?0.15 81.35/80 2.5 3.9
M81 XMM4 ?+??? 0.88+0.11?0.11 66.4/52 0.35 0.54
M81 XMM5 1.0+0.4?0.3 1.52+0.11?0.11 97.5/82 0.44 0.68
Holm IX XMM1 1.7+0.08?0.08 1.84+0.03?0.03 1000.9/882 9.4 15
Sextans A XMM1 0.18+0.23?0.16 2.25+0.12?0.07 271.4/275 0.56 0.13
NGC4214 XMM2 0.2+0.5?0.2 2.03+0.43?0.28 50.9/46 0.16 0.14
NGC4258 XMM1 1.6+0.4?0.4 1.9+0.14?0.13 101.4/78 0.06 0.04
NGC4258 XMM2 (obs 1) 3.5+0.9?0.7 1.88+0.16?0.15 97.5/63 0.43 2.6
NGC4395 XMM1 3.7+0.5?0.4 4.93+0.34?0.30 195.1/156 7.2 14
NGC4395 XMM3 ?+??? 1.86+0.14?0.09 55.9/58 0.25 0.48
NGC4449 XMM1 6.3+0.9?0.7 2.22+0.14?0.12 103/118 1.2 1.36
NGC4449 XMM3 3.3+0.5?0.4 3.36+0.29?0.23 154/89 1.3 1.5
NGC4490 XMM1 0.83+0.14?0.12 2.53+0.17?0.16 101.5/65 1.2 8.7
Continued on Next Page...
287
Table B.3 ? Continued
Source nHa ? ?2/dof FXc LXd
NGC4490 XMM2 6.3+1.3?1.0 2.36+0.19?0.17 49.5/56 0.92 6.7
NGC4490 XMM3 9.4+1.5?1.2 2.95+0.24?0.20 76.7/80 1.5 11
NGC4631 XMM1 2.3+0.16?0.15 2.13+0.06?0.05 383.4/347 0.76 5.1
NGC4631 XMM2 1.9+0.4?0.3 2.01+0.14?0.12 119.5/99 0.23 1.5
NGC4631 XMM3 0.63+0.3?0.2 1.53+0.1?0.08 146/98 0.15 1.0
NGC4736 XMM1 0.95+0.6?0.5 2.02+0.26?0.25 62.8/53 0.36 0.80
NGC4945 XMM1 5.8+0.8?0.7 1.88+0.08?0.10 116/122 0.9 1.0
NGC4945 XMM2 3.4+0.6?0.5 1.58+0.09?0.10 114.5/115 0.71 0.82
NGC4945 XMM4 5.2+0.8?0.7 2.59+0.19?0.17 75.5/62 0.49 0.56
NGC5204 XMM1 0.61+0.1?0.1 2.11+0.04?0.04 592.1/561 2.0 5.5
??? 1.1+0.1?0.1 2.41+0.07?0.07 533/498 3.0 8.3
M51 XMM1 1.1+0.30?0.27 2.67+0.20?0.16 110.5/82 0.34 2.8
M51 XMM2 2.3+0.50?0.30 2.50+0.22?0.20 75.2/70 0.52 3.3
M51 XMM5 2.7 3.08 256.0/72 0.43 2.7
M51 XMM6 2.0+0.83?0.72 2.50+0.33?0.25 40.97/43 0.13 0.83
M51 XMM7 0.5+0.39?0.46 1.95+0.23?0.18 37.8/31 0.11 0.66
M83 XMM1 1.9+0.34?0.31 2.32+0.13?0.12 210.9/211 0.64 2.8
M83 XMM4 6.0+1.8?1.3 2.54+0.26?0.23 91.4/91 0.4 1.8
M101 XMM1 0.56+0.15?0.14 1.98+0.08?0.08 303/233 0.45 2.9
M101 XMM2 2.2+0.25?0.23 1.85+0.07?0.07 288.8/263 0.81 5.3
M101 XMM3 1.5+0.4?0.3 2.70+0.21?0.17 148.9/133 0.51 3.4
M101 XMM4 2.2+0.45?0.42 2.25+0.20?0.17 165.7/140 0.38 2.5
M101 XMM5 1.3+0.2?0.3 2.28+0.12?0.11 47.9/46 0.12 0.8
Continued on Next Page...
288
Table B.3 ? Continued
Source nHa ? ?2/dof FXc LXd
NGC5408 XMM1 1.6+0.2?0.1 3.57+0.12?0.11 396.8/339 7.04 19.4
CIRCINUS XMM1 7.6+0.3?0.3 2.15+0.05?0.04 762.9/863 4.6 8.8
CIRCINUS XMM2 11.7+0.4?0.7 3.48+0.13?0.06 517.9/432 2.7 5.2
CIRCINUS XMM3 9.0+2.1?1.0 2.57+0.40?0.17 285.2/262 0.32 0.61
aTotal column density in units of 1021 cm?2
bImprovement in ?2 over the single-component power law model
cUnabsorbed flux in the 0.3-10 keV band in units of 10?12 erg cm?2 s?1
dUnabsorbed luminosity in the 0.3-10 keV band, using the distances
quoted in Table E.1, in units of 1039 erg s?1
eAbsorption level frozen at the Galactic level.
289
Table B.4. Best-Fit Absorbed Comptonization Model Parameters
ID nHa kTb tauc ?2 FXd
NGC 253 XMM2 1.8+0.08?0.16 1.28+0.13?0.12 19.59+2.0?1.6 464/498 1.47
NGC 2403 XMM1 1.95+1.2?0.6 0.98+0.16?0.15 25.4+7.2?8.3 82.8/85 1.4
NGC 4490 XMM1 4.7+1.1?2.1 0.96+0.13?0.16 27.0+18.3?4.9 66.5/64 0.66
NGC 4490 XMM2 5.0+1.6?1.4 1.21+0.21?0.31 18.8+8.6?9.7 45.7/55 0.67
M101 XMM2 1.6+0.26?0.24 1.24+0.09?0.15 23.3+3.4?2.8 256/262 0.65
M101 XMM3 1.1+0.43?0.40 1.13 15.2 128/132 0.41
Circinus XMM2 6.8+1.3?0.9 0.62+0.08?0.04 29.7+8.1?7.0 437.2/430 0.5
Circinus XMM3 6.7+1.6?2.4 0.93+0.28?0.24 23.1+17.5?5.2 273.1/261 0.17
atotal column density in units of 1021 cm?2
btemperature in keV
coptical depth
dunabsorbedfluxinthe0.3-10keV bandinunits of 10?12 erg cm?2 s?1
290
Table B.5. XMM-Newton Galaxy Observations
Galaxy S60 (Jy) S100 (Jy) FFIRa LFIRb No. of ULX
NGC247 7.93 27.32 0.602 0.687 1
NGC253 998.73 1861.67 55.92 93.10 3
NGC300 23.08 74.45 1.688 1.324 0
NGC625 5.09 9.08 0.280 0.230 0
NGC1313 35.97 92.00 2.329 4.845 2
IC0342 255.96 661.68 16.66 30.32 3
NGC1569 45.41 47.29 2.072 0.635 0
NGC1705 0.970 2.580 0.064 0.199 0
MRK 71 3.51 4.67 0.173 0.239 1
NGC2403 51.55 148.49 3.547 5.378 2
Holmberg II 1.15 2.62 0.070 0.061 1
Holmberg I ??? ??? ??? ??? 1
M81 44.73 174.02 3.647 5.655 1
M82 1271.32 1351.09 58.35 106.2 1
Holmberg IX ??? ??? ??? ??? 1
Sextans A 0.255 0.674 0.017 0.004 0
IC 2574 2.41 10.62 0.212 0.329 0
NGC 4214 17.87 29.04 0.947 0.826 1
NGC 4258 21.60 78.39 1.690 10.48 1
NGC4395 4.21 12.90 0.299 0.573 1
NGC4449 37.00 58.28 1.937 2.199 1
NGC4490 47.79 85.94 2.636 19.19 5
NGC4631 82.90 208.66 5.324 35.83 1
NGC4736 62.41 135.34 3.734 8.261 4
NGC4945 588.11 1415.5 36.95 42.49 0
NGC 5204 2.33 5.35 0.143 0.395 2
M51 108.68 292.08 7.213 44.74 5
M83 266.03 638.63 16.69 76.79 0
NGC5253 30.00 30.92 1.365 16.72 0
M101 88.04 252.84 6.048 39.63 4
NGC5408 2.825 2.958 0.129 0.356 1
Circinus 248.7 315.85 12.06 23.10 4
aflux in units of 10?9 ergcm?2 s?1
bfar-infrared luminosity in units of 1042 ergs?1
291
Appendix C
Appendix to Chapter 3
C.1 Spectral Simulations
We conducted spectral simulations in order to: (1) determine the num-
ber of counts needed to measure the oxygen and iron abundances and
(2) to verify the model independence of the galactic column density
and abundances with respect to the grad, diskbb, and bbody mod-
els (seen in a comparison of Tables 3.2 and 3.4). Towards this end, we
created simulated pn spectra based on the long Holmberg IX XMM1
observation?s (0200980101) unbinned, pn spectrum. We used the base
(grad + pow) model parameters as indicated in Table 3.2. We mod-
eled both the Galactic column density (Dickey & Lockman (1990) value:
nH = 4.0 ? 1020 cm?2) and host galaxy column density (nH = 1.9 ?
1021 cm?2) with individual tbabs models. Thus, all of the abundances
were set to the solar Wilms values. We used the XSPEC command
fakeit to create simulated spectra with 200000, 40000, 10000, 5000,
and 2000counts. The simulated spectra were binned with 20cts/bin
using grppha. We fit the binned simulated spectra with the mod-
292
els tbabs*tbvarabs*(grad + pow), tbabs*tbvarabs*(diskbb
+ pow), and tbabs*tbvarabs*(bbody + pow). This allowed us to
see the effects the different models have on the measured galactic hy-
drogen column density and abundances. The results for these fits are
seen in Table C.1. The range that the oxygen and iron abundances were
allowed to vary within was 0.0 (lower limit) to 5.0 (upper limit) with
respect to the solar values.
In Figure C.1, the number of simulated counts versus the er-
rors on the oxygen and iron abundances are plotted for the
tbabs*tbvarabs*(grad + pow)model. Here, [O/H]= 12+log(O?
0.00049) and [Fe/H]= 12 + log(Fe ? 0.0000269), using the solar Wilms
values for O/H and Fe/H. As seen in the plots, the errors in oxygen
abundance are much smaller for a given number of counts compared to
the errors in iron abundance. Further, the upper limits on the oxygen
abundance continue to be meaningful through 2000counts. This is not
true for the iron abundances, where the error bars extend through the
entire range of allowed values (from Fe/H = 0.0 - 5.0, or [Fe/H] up to
8.13). Thus, oursimulationsshowusthattheironabundance(frommea-
surements of the Fe L-shell edge at 851eV with the tbvarabs model)
requires at least 5000counts to be detected. At 5000counts the model
derived value (Fe/H = 1.36) is meaningful, however, the errors extend
throughout the entire allowable range (Fe/H = 0.0-5.0). The oxygen
abundance (from measurements of the O K-shell edge at 542eV with
the tbvarabs model) is detected down to 2000counts, but with large
errors below 10000counts.
From Table C.1, we find that the same trends described in Section 3.4.1
293
are present in our simulations. Namely, there is little variation between
the model derived abundances and column densities. A comparison of
the mean nH and oxygen abundance values shows ? 2% difference be-
tween the grad and diskbb model. The bbody nH values are roughly
26%higher while theoxygenabundances are? 11% higher. Comparing
the 40000 and 200000count spectra for the iron abundance, we find that
the grad and diskbb model values differ by ? 12% while the bbody
model results are larger by a factor of 50%. While the bbody model
yields lower column densities and higher abundances, the diskbb and
grad models are in agreement. The differences in the bbody results are
low for the column density and oxygen abundance, but appeared signif-
icant for the iron abundance.
294
Table C.1. Spectral Fits to Simulated HolmIX XMM1 Spectrum
Countsa nHb Oxygen abundancec Iron abundancec ?2/dof
tbabs*tbvarabs*(grad + pow)
200000 0.21+0.02?0.02 1.08+0.13?0.15 0.87+0.68?0.76 1687/1727
40000 0.19+0.04?0.03 1.15+0.26?0.33 0.97+1.41?0.97 933/914
10000 0.20+0.07?0.06 1.55+0.41?0.52 0.90+2.64?0.90 340.5/401
5000 0.16+0.13?0.07 1.37+0.75?1.34 1.36+3.64?1.36 187.2/201
2000 0.33+0.14?0.19 1.20+0.57?1.20 5.0+0.0?5.0 71.9/89
tbabs*tbvarabs*(diskbb + pow)
200000 0.21+0.02?0.02 1.09+0.15?0.19 0.99+0.94?0.76 1691/1727
40000 0.19+0.03?0.03 1.16+0.27?0.35 1.09+1.47?1.09 934/914
10000 0.19+0.03?0.05 1.58+0.43?0.50 1.01+2.80?1.01 341/401
5000 0.16+0.12?0.08 1.41+0.76?1.16 1.57+3.43?1.57 187.5/201
2000 0.32+0.18?0.19 1.23+0.59?1.23 5.0+0.0?5.0 71.7/89
tbabs*tbvarabs*(bbody + pow)
200000 0.17+0.02?0.02 1.24+0.16?0.20 1.65+0.85?1.01 1702/1727
40000 0.15+0.04?0.02 1.28+0.36?0.40 1.73+1.97?1.72 937/914
10000 0.15+0.07?0.04 1.77+0.61?0.72 1.22+3.78?1.22 343.7/401
5000 0.13+0.10?0.07 1.61+0.97?1.61 2.19+2.81?2.19 187.5/201
2000 0.21+0.19?0.12 1.16+0.92?1.16 5.0+0.0?5.0 71.8/89
aTotal number of photon counts for simulated pn spectrum. Simulated
spectra created with the XSPEC fakeitcommand, using the model param-
eters and response files from the long HolmIX XMM1 observation.
bHydrogen column density determined from tbvarabs in units of
1022 cm?2. The Galactic value of nH was fixed to the Dickey & Lockman
(1990) value with the tbabs model.
cElement abundance relative to the Wilms solar abundance from the
tbvarabs model
295
Figure C.1: The number of counts for simulated spectra versus the [O/H] value
(left) and [Fe/H] value (right) from the tbabs*tbvarabs*(grad + pow),
where errors represent the 90% confidence rate. The horizontal lines represent
the solar Wilms values ([O/H]? 8.69 and [Fe/H]? 7.43). Arrows are used to
represent errors that extend below the range of the plot. The tip of the arrow
point represents the [O/H] or [Fe/H] parameter from the XSPEC model (see Ta-
ble C.1).
296
Appendix D
Appendix to Chapter 4
D.1 DetailsonVariabilityofIndividualSources
Below we describe the variability between observations for the AGN
listed in Table 4.12. The low column density sources are those whose
best-fit spectra were fit by a simple absorbed power law or a power law
and blackbody model. The high column density/complex sources are
those that we fit with the partial covering/double power law models. In
the following discussion, XMM is used to denotethe PN spectrum while
XRT-1 denotes, for example, the first XRT observation for the source, as
listed in Table 4.2.
D.1.1 Low Column Density Sources
MRK 352 ? The PN and two XRT spectra for MRK 352 were not well
fit until the flux was allowed to vary. A varying flux improved the fit
by ??2 ? 6500. The fit was then greater improved by allowing the
column density to change (??2 = 320). Changing the power law in-
dex and normalizations improved the fit by ??2 = 30, however, the
297
power law photon indices were the same within the errorbars. The
best-fit tbabs*tbabs*(pegpwrlw + bbody) model is shown in Fig-
ure 4.9 where ?2/dof = 1289.01/1189. The black body and power law
components were those seen in Table 4.6. Hydrogen column density
changes between the observations as: 0.00 ? 0.02 ?1021 cm?2 (XMM),
0.83 ? 1.04 ?1021 cm?2 (XRT-1), 1.28 ? 1.51 ?1021 cm?2 (XRT-2). The
flux errors from the pegged power law component were: 1.96 ? 2.00
?10?11 ergs?1 cm?2 (XMM), 1.37 ? 1.47?10?11 ergs?1 cm?2 (XRT-1), and
0.98 ? 1.05 ?10?11 ergs?1 cm?2 (XRT-2). Considering that the XRT ob-
servations were taken only a day apart, it is clear that this source varies
considerably. Five months earlier, the XMM spectra show the source
nearly twice as bright with no absorption.
SWIFT J0216.3+5128 ? All of the observations for SWIFT J0216.3+5128
took place within the span of a month. While no variations were
seen between the column densities and power law indices (??2 <
3), the flux did vary. The flux errors from the pegged power law
component were: 1.63 ? 1.75 ?10?11 ergs?1 cm?2 (XMM), 2.25 ? 2.52
?10?11 ergs?1 cm?2 (XRT-1), 2.15?2.43?10?11 ergs?1 cm?2 (XRT-2), and
1.72 ? 2.12 ?10?11 ergs?1 cm?2 (XRT-3). From the first two observations,
the flux drops about 30% over two weeks and then remains at about the
same level through the last two observations.
298
ESO 548-G081 ? The spectra of ESO 548-G081 were found not to vary
in column density. However, they did vary in both flux (??2 = 625)
and ? (??2 = 60). We fit this source with a tbabs*tbabs*(pegpwrlw
+ bbody + zgauss) model with ?2/dof = 1428.1/1295. We note that the
blackbody parameters were not fixed for observation XRT-2. This spec-
trum showed a soft excess that was fit with a blackbody much lower
than the value for the XMM observation (see Table 4.6 with kT in the
range of 0.034 ? 0.049keV. There were remaining residuals in this fit for
XRT-2, leaving us unsure of the nature of this feature which could be
the result of a hot pixel or the instrument. The higher energy spectrum
(above 0.5keV) appeared well fit by the model used. We found the er-
ror on the photon index as: 1.851 ? 1.877 (XMM), 1.716 ? 1.802 (XRT-
1), and 1.930 ? 2.002 (XRT-2). The flux errors from the pegged power
law component were: 2.91 ? 2.96 ?10?11 ergs?1 cm?2 (XMM), 3.76 ? 4.04
?10?11 ergs?1 cm?2 (XRT-1), and 4.13 ? 4.33 ?10?11 ergs?1 cm?2 (XRT-
2). It is unclear whether the changes between the XMM observation and
XRT-1 reflect differences in the instruments or the source. Both observa-
tions were taken on the same day, nine hours apart. Two months later,
the XRT-2 observation shows the source to have a steeper power law
index and a higher flux.
ESO 490-G026 ? Allowing flux, column density, and power law in-
dices to vary between the four observations of ESO 490-G026 greatly
improved the fit with ??2 values of 171, 21, and 62. We used an
absorbed power law + blackbody model (kT set at the value in Ta-
ble 4.6) with an Fe K line. Using the pegpwrl for a power law com-
299
ponent, the best fit gave ?2/dof = 1281.8/1313. The errors on nH were:
2.96 ? 3.59 ?1021 cm?2 (XMM), 3.10 ? 3.81 ?1021 cm?2 (XRT-1), 2.54 ?
3.84 ?1021 cm?2 (XRT-2), and 4.12 ? 6.74 ?1021 cm?2 (XRT-3). The er-
rors on the photon index were: 1.66 ? 1.71 (XMM), 1.88 ? 2.00 (XRT-
1), 1.63 ? 1.86 (XRT-2), and 1.74 ? 2.09 (XRT-3). Finally, the errors
on the flux from the pegged power law component were: 3.23 ? 3.32
?10?11 ergs?1 cm?2 (XMM), 3.79 ? 4.09?10?11 ergs?1 cm?2 (XRT-1), 3.53
? 3.97 ?10?11 ergs?1 cm?2 (XRT-2), and 2.71 ? 3.37 ?10?11 ergs?1 cm?2
(XRT-3). Observation XRT-1 occurred about 3 months before the XMM
observation. In this time, the power law index flattened while the flux
decreased. The column density of the sources does not vary between
the first three observations. However, between XRT-2 and XRT-3, 5 days
apart, the column increased by nearly twice the previous amount with
the flux decreasing.
SWIFT J0904.3+5538 ? As with ESO 490-G026, all of the parameters
(column density, flux, and power law photon index) varied for SWIFT
J0904.3+5538. For this source, we fixed the parameters of the blackbody
component to the best-fit values of the XMM PN observation. No Fe K
line was required in this spectrum. The ??2 values allowing flux, nH,
and ? to vary were 523, 12, and 20. Errors for column density were:
0.61 ? 1.29 ?1021 cm?2 (XMM), 1.01 ? 1.93 ?1021 cm?2 (XRT-1), and 1.37
? 2.41 ?1021 cm?2 (XRT-2). The errors on the photon index were: 1.80
? 1.97 (XMM), 1.45 ? 1.69 (XRT-1), and 1.49 ? 1.75 (XRT-2). Finally, the
errors on the flux from the pegged power law component were: 0.94 ?
1.02?10?11 ergs?1 cm?2 (XMM), 0.76 ? 0.88?10?11 ergs?1 cm?2 (XRT-1),
300
and 0.60 ? 0.69 ?10?11 ergs?1 cm?2 (XRT-2). Between XRT-1 and XRT-2,
approximately a month apart, the source dimmed by ? 20% and then
brightened more than twice that amount 3 months later in the XMM
observation. With the higher flux, the XMM observation showed less
absorption and a steeper slope.
MCG +04-22-042 ? MCG +04-22-042 was well-fit with a simple ab-
sorbed power law (pegpwrlw) with ?2/dof = 1534.8/1190. This fit re-
quired flux and column density to vary between the XMM and XRT
observations with ??2 of 938 and 298, respectively. Errors for column
density were: 0.00 ? 0.02 ?1020 cm?2 (XMM) and 1.77 ? 3.36 ?1020 cm?2
(XRT-1). Though these values are small, there was clearly a change in
column density, evidenced by the very significant change in ?2. The er-
rors on the flux for the pegged power law component were: 3.38 ? 3.42
?10?11 ergs?1 cm?2 (XMM) and 2.29 ? 2.44?10?11 ergs?1 cm?2 (XRT-1).
The two observations were approximately 5 months apart, showing that
the flux and column changed while the photon index remained roughly
the same. The flux increased (in the XMM obs.) while the column den-
sity decreased.
UGC 6728 ? In the four UGC 6728 spectra, the only statistically signif-
icant variation is in flux (??2 = 680). The absorbed pegged power law
model yields a best-fit ?2/dof of 855.3/863. This source varies by a high
amount with errors in flux of: 1.15 ? 1.19 ?10?11 ergs?1 cm?2 (XMM),
2.12 ? 2.28 ?10?11 ergs?1 cm?2 (XRT-1), 1.40 ? 1.60 ?10?11 ergs?1 cm?2
(XRT-2), and 1.53 ? 1.89?10?11 ergs?1 cm?2 (XRT-3). Thus, the flux dou-
bled between the four months from the XMM observation and XRT-1. It
301
then decreased by about 25% over 5 days, remaining at about the same
level in the XRT-3 observation a week later.
WKK 1263 ? The only significant change in ?2 for the combined spec-
tralfits to WKK 1263 was in flux (??2 = 150). The best-fit pegged power
law fit had ?2/dof = 826.8/782. The errors on the flux from the pegged
powerlawcomponentwere: 1.52?1.57?10?11 ergs?1 cm?2 (XMM),1.18
? 1.39?10?11 ergs?1 cm?2 (XRT-1), and 1.07 ? 1.21 ?10?11 ergs?1 cm?2
(XRT-2). The source was brighter (by as much as 50%) in the XMM ob-
servation taken about a month after the XRT observations.
MCG+09-21-096? MCG+09-21-096showed significantvariationin?2
when allowing variations in flux and power law photon index (??2 =
797 and 12). The model (tbabs*pegpwrlw) was a good fit with ?2/dof
= 1289.2/1334. Errors on ? were: 1.78 ? 1.79 (XMM), 1.77 ? 1.82 (XRT-
1), and 1.67 ? 1.74 (XRT-2). The errors on the flux from the pegged
powerlawcomponentwere: 2.77?2.81?10?11 ergs?1 cm?2 (XMM),3.97
? 4.16?10?11 ergs?1 cm?2 (XRT-1), and 3.50 ? 3.74 ?10?11 ergs?1 cm?2
(XRT-2). Between theXRT-1 andXMM observations, 7months apart, the
photon index is the same while the flux decreased by about 40%. Then,
in the week between the XMM and XRT-2 observation, the source pho-
ton index flattened slightly while the flux increased to nearly the level
in the XRT-1 observation.
302
D.1.2 High Column Density/Complex Sources
NGC 1142 ? Since NGC 1142 has a strong Fe K line (see Table 4.8 for
EW and normalization), we fixed the Fe K parameters for the three XRT
observations to the best fit values for the PN spectrum. Allowing the
flux to vary between these observations improved the fit by ??2 = 130.
There is no evidence of variability in column density, however, vary-
ing power law components improved the fit by ??2 ? 30. Fitting the
spectra with a pegged power law model, the best-fit model has ?2/dof
= 132.6/105. Errors on the power law photon index, ?, were: 1.54 ? 2.29
(XMM), 2.34 ? 3.85 (XRT-1), 2.15 ? 3.98 (XRT-2), and 1.08 ? 2.94 (XRT-
3). Though the error bars for the photon index are large, due to the few
countsforthisheavilyabsorbedsource, itisclearthatobservationXRT-1
has a steeper power law component than the XMM observation. The er-
rors on thepegpwrlwflux were: 1.97 ? 6.01?10?11 ergs?1 cm?2 (XMM),
10.87?322.3?10?11 ergs?1 cm?2 (XRT-1),10.10?478?10?11 ergs?1 cm?2
(XRT-2), and 3.81 ? 43.13 ?10?11 ergs?1 cm?2 (XRT-3). Once again, the
XRT error bars are large due to the few counts. Noting that the XMM
observation occurred 6 months before XRT-1, clearly the flux is higher
in the XRT-1 observation while the photon index is steeper. No conclu-
sions can be drawn from the final two XRT observations.
303
SWIFT J0318.7+6828 ? SWIFT J0318.7+6828 showed no variability in
column density or power law component between the XMM and two
XRT observations (??2 < 3 allowing each to vary). The variability
was significant in flux with ??2 ? 40 when a constant model was
added. The errors on the pegpwrlw flux were computed as: 1.01 ? 1.25
?10?11 ergs?1 cm?2 (XMM), 0.76 ? 1.00?10?11 ergs?1 cm?2 (XRT-1), and
0.79 ? 1.04 ?10?11 ergs?1 cm?2 (XRT-2) with ?2/dof= 412.9/452. Thus,
the XRT observations, taken a week apart, did not vary. However, three
months earlier the XMM observations show the source to be brighter by
? 30%.
ESO 362-G018 ? As mentioned, the spectra of ESO 362-G018 showed
more variability than any other source in this sample. The value Fmax ?
Fmin/Favg for this source was 1.64 in the soft band and 1.22 in the hard
band. This source had a complex spectrum, described in the Detailed
Spectral Fitting section. We fit the XMM and XRT spectra of this source
withapartialcoveringandpowerlawmodel. However,weaddedgaus-
siancomponentstofitthestrongFeKline(intheXMMobservation)and
the helium-like oxygen edge. The best-fit model required flux, nH, and
? to vary (??2 = 1890, 254, and 40) with ?2/dof = 538.4/445. The errors
on column density were 24.54 ? 31.23 ?1022 cm?2 (XMM), 6.47 ? 269.3
?1022 cm?2 (XRT-1), and2.12?4.71?1022 cm?2 (XRT-2),withpartialcov-
ering fractions of 0.91 ? 0.93 (XMM), 0.10 ? 0.91 (XRT-1), and 0.45 ? 0.68
(XRT-2). Errors on the photon index were: 2.13 ? 2.23 (XMM), 1.76 ? 2.05
(XRT-1), and 1.69 ? 1.99 (XRT-2). Finally, errors on flux for the pegged
powerlawcomponentwere: 1.46?1.83?10?11 ergs?1 cm?2 (XMM),2.17
304
? 7.69 ?10?11 ergs?1 cm?2 (XRT-1), and 2.25 ? 2.95 ?10?11 ergs?1 cm?2
(XRT-2).
In Figure 4.10a, we show the normalized XMM and XRT spectra with
best-fit model. It is clear from this figure that this source varied a great
deal in these observations. Both XRT-1 and XRT-2, despite the large er-
ror bars on XRT-2, have similar spectra. These observations were taken
approximately a month apart. Taken two months later, the XMM obser-
vation looks like a different source altogether. In this time, the column
density increased by a factor of 10. Additionally, the photon index be-
came steeper and the flux dropped by about 50%. Along with these
changes, the Fe K line (not distinguishable in the XRT observations) be-
came extremely prominent. One likely explanation for the appearance
of the Fe K line is that it was simply too dim to be distinguishable at the
higher flux levels exhibited in the XRT observations. This is illustrated
in the unfolded spectrum shown in Figure 4.10b. In this plot, where the
y-axis shows E2 f(E), it is clear that if the Fe K line remained at the same
flux level as in the XMM observation it would be completely dominated
by the power law component.
NGC 6860 ? As mentioned in the Detailed Spectral Fitting section, the
spectrum of NGC 6860 is quite complex. Due to this complexity and a
lack of signal to noise (especially considering that the PN data was cor-
rupted), we areunsure of thetrue natureofthis spectrum. Therefore, we
decided to compare the XMM and XRT spectra with the pcfabs*pow
model. This may not be the most valid description of the data, but it
gives a basis to compare the spectra. Using this model, we fit both of the
305
MOS observations (with the parameters nH and ? tied together while
the flux was allowed to vary) simultaneously with the XRT observation.
Variations were statistically significant for flux, nH, and ? with ??2 =
83, 56, and 15, respectively. The best-fit for the partial covering, pegged
power law model yielded ?2/dof of 513.3/430. Errors on the column
density were 3.33 ? 6.09 ?1022 cm?2 (XMM) and 0.59 ? 1.73 ?1022 cm?2
(XRT) with covering fraction errors of 0.0 ? 0.66 (XMM) and 0.73 ? 0.92
(XRT). The photon index errors were: 0.64 ? 0.88 (XMM) and 1.16 ? 1.71
(XRT). Finally, errors on the flux from the pegged power law compo-
nent were: 1.06 ? 1.19 ?10?11 ergs?1 cm?2 (XMM MOS-1), 0.99 ? 1.12
?10?11 ergs?1 cm?2 (XMM MOS-2), and 1.44 ? 1.88 ?10?11 ergs?1 cm?2
(XRT). These observations were taken 4 months apart. Interestingly, the
XRT observation is well-fit by the partial covering model, giving pho-
ton index and covering fractions similar to the typical values seen in
Table 4.8. However, the column density is much lower (by a factor of 10
from the other sources). Between the XRT and XMM observations, the
column density seems to have doubled while the photonindex flattened
and the flux decreased by nearly half.
306
Appendix E
Appendix to Chapter 5
E.1 Additional Tables
In this appendix, we include several tables referred to in Chapter 5. Ta-
ble E.1 includes all of the SWIFT BAT AGN from the 9-month catalog
(Tueller et al. 2008). We include the optical type, host galaxy type, and
X-rayobservationdetailsforeachofthesesources. Thetableisdiscussed
in more detail throughout Chapter 5. Table E.2 includes the X-ray spec-
tral fits to SWIFT XRT observations of sources with no other available
data (or for which we could not easily obtain similar quality data to
the rest of the survey sources). Table E.3 includes the X-ray spectral fits
from the literature or our own analysis of ASCA data from the Tartarus
database. Explanations of the analysis are included in ? 5.2. Notes on
individual sources are found in the following section (? E.2).
307
Ta
ble
E.1
:X
-ra
yO
bs
erv
ati
on
De
tai
ls
No
.
So
ur
ce
RA
(h
m
s)
De
c(
?
??
? )
Ty
pe
1
Ho
st
Ga
lax
y1
Ob
s.2
Ob
sI
D/
ref
Ct
Ra
te
Ex
pT
im
e
1
NG
C
23
5A
00
42
52
.8
-23
32
27
.6
Sy
2
S0
pe
c,
In
t
XR
T
00
03
56
48
00
1,
02
0.0
21
96
21
40
2
2
M
rk
34
8
00
48
47
.1
31
57
25
.2
Sy
2
SA
(s)
0/
a,
In
t
AS
CA
Aw
ak
ie
ta
l.(
20
00
)
0.0
27
28
00
0
3
M
rk
35
2
00
59
53
.3
31
49
36
.8
Sy
1
SA
0
XM
M
W
int
er
et
al.
(20
08
)
7.3
97
73
4
NG
C
45
4
01
14
22
.7
-55
23
55
.0
Sy
2
Pe
c,
In
t
XR
T
00
03
52
44
00
1,
02
,0
3
0.0
17
20
29
0
5
Fa
ira
ll9
01
23
45
.8
-58
48
20
.5
Sy
1
S,
In
t?
AS
CA
Re
yn
old
s(
19
97
)
1.0
01
22
00
0
6
NG
C
52
6A
01
23
54
.4
-35
03
55
.4
Sy
1.5
S0
pe
c?,
In
t
AS
CA
Tu
rn
er
et
al.
(19
97
)
0.6
41
43
00
0
7
NG
C
61
2
01
33
57
.7
-36
29
35
.9
Ga
l
SA
0+
pe
c,
In
t
XM
M
W
int
er
et
al.
(20
08
)
0.1
2
97
44
8
ES
O
29
7-0
18
01
38
37
.2
-40
00
41
.0
Sy
2
Sa
sp
,In
t
Su
za
ku
9
NG
C
78
8
02
01
06
.5
-06
48
55
.8
Sy
2
SA
(s)
0/
a
AS
CA
77
04
00
00
0.0
3
19
21
10
M
rk
10
18
02
06
16
.0
00
17
29
.0
Sy
1.5
S0
,In
t
XR
T
00
03
51
66
00
1
0.6
10
9
42
56
11
LE
DA
13
85
01
02
09
34
.3
52
26
33
.0
Sy
1
?
XR
T
00
03
52
46
00
2
0.3
65
64
35
.5
12
M
rk
59
0
02
14
33
.6
00
46
00
.1
Sy
1.2
SA
(s)
a
XM
M
Lo
ng
ino
tti
et
al.
(20
07
)
71
00
0
13
2M
AS
X
J02
16
29
87
+5
12
62
46
02
16
29
.8
51
26
24
.7
?
?
XM
M
W
int
er
et
al.
(20
08
)
1.6
2
89
21
14
[H
B8
9]
02
12
+7
35
02
17
30
.8
73
49
32
.5
BL
La
c
?
XR
T
00
03
52
41
00
2
0.0
83
07
64
85
.4
15
NG
C
93
1
02
28
14
.5
31
18
42
.1
Sy
1.5
Sb
c,
In
t
AS
CA
Re
yn
old
se
ta
l.(
19
95
)
0.1
33
18
90
0
16
NG
C
98
5
02
34
37
.8
-08
47
15
.4
Sy
1
SB
bc
?p
(R
ing
)
Ch
an
dr
a
Kr
on
go
ld
et
al.
(20
05
)
0.1
8
77
00
0
17
ES
O
41
6-G
00
2
02
35
13
.4
-29
36
16
.9
Sy
1.9
Sa
XM
M
Ga
llo
et
al.
(20
06
)
42
69
18
ES
O
19
8-0
24
02
38
19
.7
-52
11
32
.3
Sy
1
N,
In
t
XM
M
Po
rq
ue
te
ta
l.(
20
04
a)
22
90
0
19
QS
O
B0
24
1+
62
2
02
44
57
.7
62
28
06
.6
Sy
1
?
XR
T
00
03
63
05
00
1
0.4
48
5
10
61
8
20
NG
C
11
42
02
55
12
.2
00
11
01
.0
Sy
2
Sp
ec
(R
ing
B)
,In
t
XM
M
W
int
er
et
al.
(20
08
)
0.2
8
89
21
21
2M
AS
X
J03
18
18
99
+6
82
93
22
03
18
19
.0
68
29
31
.6
Sy
1.9
?
XM
M
W
int
er
et
al.
(20
08
)
1.4
7
65
78
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
308
Ta
ble
E.1
?C
on
tin
ue
d
No
.
So
ur
ce
RA
(h
m
s)
De
c(
?
??
? )
Ty
pe
1
Ho
st
Ga
lax
y1
Ob
s.2
Ob
sI
D/
ref
Ct
Ra
te
Ex
pT
im
e
22
NG
C
12
75
03
19
48
.2
41
30
42
.1
Sy
2
cD
;pe
c
23
PK
S0
32
6-2
88
03
28
36
.5
-28
41
48
.5
Sy
1.9
?
no
ne
24
NG
C
13
65
03
33
36
.4
-36
08
25
.4
Sy
1.8
(R
?)S
Bb
(s)
b,
In
t
XR
T
00
03
54
58
00
1,
02
0.0
27
19
43
4
25
ES
O
54
8-G
08
1
03
42
03
.7
-21
14
39
.8
Sy
1
SB
(rs
)a
pe
c?,
In
t?
XM
M
W
int
er
et
al.
(20
08
)
11
.95
89
24
26
RB
S4
76
03
49
23
.2
-11
59
26
.9
BL
La
c
?
AS
CA
Re
ev
es
&
Tu
rn
er
(20
00
)
10
90
00
27
PG
C
13
94
6
03
50
23
.8
-50
18
35
.6
?
?,
In
t
XR
T
00
03
52
51
00
3
0.0
2
88
08
28
2M
AS
X
J03
56
56
55
-40
41
45
3
03
56
56
.5
-40
41
45
.6
Sy
1.9
?
XR
T
00
03
56
00
00
1,
02
0.0
62
8
12
57
9
29
3C
10
5
04
07
16
.5
03
42
25
.6
Sy
2
?
XR
T
00
03
56
25
00
1,
02
,0
3,
04
0.0
07
22
39
8
30
3C
11
1
04
18
21
.3
38
01
35
.8
Sy
1
N
AS
CA
Re
yn
old
se
ta
l.(
19
98
)
0.7
02
32
80
0
31
1H
04
19
-57
7
04
26
00
.8
-57
12
00
.4
Sy
1
?
AS
CA
Tu
rn
er
et
al.
(19
99
)
24
00
0
32
3C
12
0
04
33
11
.1
05
21
15
.5
Sy
1
S0
,In
t?
AS
CA
Gr
an
di
et
al.
(19
97
)
1.8
61
50
00
0
33
2M
AS
X
J04
44
09
03
+2
81
30
03
04
44
09
.0
28
13
00
.5
Sy
2
S
XR
T
00
03
51
75
00
5,
06
,0
7
0.1
14
76
40
4
34
M
CG
-01
-13
-02
5
04
51
41
.5
-03
48
33
.8
Sy
1.2
SA
B(
s)0
+p
ec
XM
M
Ga
llo
et
al.
(20
06
)
23
63
35
1R
XS
J04
52
05
.0+
49
32
48
04
52
05
.0
49
32
45
.2
Sy
1
?
XR
T
00
03
52
81
00
2
0.9
5
22
00
36
XS
SJ
05
05
4-2
34
8
05
05
45
.7
-23
51
14
.0
Sy
2
?
XR
T
00
03
52
06
00
4
0.1
2
32
44
37
4U
05
17
+1
7
05
10
45
.5
16
29
55
.7
Sy
1.5
?
XM
M
no
tp
ub
lic
38
Ar
k1
20
05
16
11
.4
00
08
59
.3
Sy
1
Sb
/p
ec
,In
t
AS
CA
72
00
00
00
1.2
48
47
85
0
39
ES
O
36
2-G
01
8
05
19
35
.8
-32
39
28
.1
Sy
1.5
S0
/a
,In
t
XM
M
W
int
er
et
al.
(20
08
)
0.9
5
89
21
40
Pic
tor
A
05
19
49
.7
-45
46
44
.4
Sy
1
(R
?)S
A0
:pe
c
AS
CA
Er
ac
leo
us
&
Ha
lpe
rn
(19
98
)
0.5
6
61
87
7
41
ES
O
36
2-G
02
1
05
22
58
.0
-36
27
31
.0
BL
La
c
N
AS
CA
73
05
50
10
0.4
0
51
33
42
PK
S0
53
7-4
41
05
38
50
.4
-44
05
08
.9
BL
La
c
?
XR
T
00
05
01
50
01
1
0.3
17
22
57
0
43
[H
B8
9]
05
37
-28
6
05
39
54
.3
-28
39
55
.8
Bl
az
ar
?
AS
CA
Sie
be
rt
et
al.
(19
96
)
0.0
4
29
00
0
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
309
Ta
ble
E.1
?C
on
tin
ue
d
No
.
So
ur
ce
RA
(h
m
s)
De
c(
?
??
? )
Ty
pe
1
Ho
st
Ga
lax
y1
Ob
s.2
Ob
sI
D/
ref
Ct
Ra
te
Ex
pT
im
e
44
PK
S0
54
8-3
22
05
50
40
.8
-32
16
17
.8
BL
La
c
?,
In
t
AS
CA
Sa
mb
ru
na
&
M
us
ho
tzk
y(
19
98
)
1.7
0
31
53
3
45
NG
C
21
10
05
52
11
.4
-07
27
22
.3
Sy
2
SA
B0
-
AS
CA
Tu
rn
er
et
al.
(19
97
)
0.3
19
36
30
0
46
M
CG
+0
8-1
1-0
11
05
54
53
.6
46
26
21
.5
Sy
1.5
SB
0
AS
CA
Gr
an
di
et
al.
(19
98
)
0.5
54
10
00
0
47
EX
O
05
56
20
-38
20
.2
05
58
02
.0
-38
20
04
.6
Sy
1
?,
In
t
AS
CA
Tu
rn
er
et
al.
(19
96
)
45
00
0
48
IR
AS
05
58
9+
28
28
06
02
10
.7
28
28
22
.1
Sy
1
?
XR
T
00
03
52
55
00
1
0.3
5
64
46
49
ES
O
00
5-G
00
4
06
05
41
.6
-86
37
54
.8
Sy
2
Sb
Su
za
ku
50
M
rk
3
06
15
36
.4
71
02
15
.0
Sy
2
S0
AS
CA
Tu
rn
er
et
al.
(19
97
)
0.0
44
27
30
0
51
ES
O
12
1-G
02
8
06
23
45
.6
-60
58
44
.4
Sy
2
SB
(l)
0/
a?
,In
t
XR
T
00
03
62
96
00
7
0.0
2
87
16
52
ES
O
49
0-G
02
6
06
40
11
.7
-25
53
43
.4
Sy
1.2
Pe
c,
In
t
XM
M
W
int
er
et
al.
(20
08
)
5.6
8
91
92
53
2M
AS
X
J06
40
37
99
-43
21
21
1
06
40
38
.0
-43
21
20
.9
?
?,
In
t
XR
T
00
03
56
01
00
1,
02
0.0
1
18
89
7
54
2M
AS
X
J06
41
18
06
+3
24
93
13
06
41
18
.0
32
49
31
.4
Sy
2
?,
In
t
XM
M
W
int
er
et
al.
(20
08
)
0.2
5
10
69
6
55
M
rk
6
06
52
12
.2
74
25
37
.6
Sy
1.5
SA
B0
+
XR
T
00
03
54
61
00
3
0.1
14
53
67
56
M
rk
79
07
42
32
.8
49
48
34
.9
Sy
1.2
SB
b
XM
M
Ga
llo
et
al.
(20
05
)
12
.57
16
80
57
SD
SS
J07
46
25
.87
+2
54
90
2.2
07
46
25
.9
25
49
02
.3
Bl
az
ar
?
XR
T
00
03
54
22
00
2
0.1
43
13
36
5
58
IG
R
J07
59
7-3
84
2
07
59
41
.0
-38
45
36
.0
Sy
1.2
?
XR
T
00
03
52
23
00
4
0.6
05
5
75
34
59
4C
+7
1.0
7
08
41
24
.4
70
53
42
.4
Bl
az
ar
?
AS
CA
Ca
pp
ie
ta
l.(
19
97
)
0.3
0
10
54
0
60
M
rk
18
09
01
58
.4
60
09
06
.1
Sy
2
S?
,In
t
XM
M
W
int
er
et
al.
(20
08
)
0.5
0
99
10
61
2M
AS
X
J09
04
69
9+
55
36
02
5
09
04
36
.9
55
36
02
.5
Sy
1
?,
In
t
XM
M
W
int
er
et
al.
(20
08
)
1.5
2
71
42
62
2M
AS
X
J09
11
29
99
+4
52
80
60
09
11
30
.0
45
28
05
.9
Sy
2
?
XM
M
W
int
er
et
al.
(20
08
)
0.0
5(
mo
s)
11
53
0
63
IR
AS
09
14
9-6
20
6
09
16
08
.9
-62
19
29
.6
Sy
1
?
XR
T
00
03
52
33
00
1
0.2
1
73
99
64
2M
AS
X
J09
18
00
27
+0
42
50
66
09
18
00
.3
04
25
06
.2
QS
O2
?,
In
t?
Su
za
ku
70
20
76
01
0
0.0
29
(X
IS0
+3
)
49
89
65
M
CG
-01
-24
-01
2
09
20
46
.2
-08
03
22
.0
Sy
2
SA
B(
rs)
c,
In
t
XR
T
00
03
52
62
00
1
0.0
8
85
22
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
310
Ta
ble
E.1
?C
on
tin
ue
d
No
.
So
ur
ce
RA
(h
m
s)
De
c(
?
??
? )
Ty
pe
1
Ho
st
Ga
lax
y1
Ob
s.2
Ob
sI
D/
ref
Ct
Ra
te
Ex
pT
im
e
66
M
CG
+0
4-2
2-0
42
09
23
43
.0
22
54
32
.4
Sy
1.2
E?
XM
M
W
int
er
et
al.
(20
08
)
13
.99
90
12
67
M
rk
11
0
09
25
12
.9
52
17
10
.7
Sy
1
S?
,In
t
AS
CA
73
09
10
00
1.1
5
20
74
7
68
NG
C
29
92
09
45
42
.0
-14
19
35
.0
Sy
2
Sa
pe
c,
In
t
AS
CA
Tu
rn
er
et
al.
(19
97
)
0.0
39
27
30
0
69
M
CG
-05
-23
-01
6
09
47
40
.2
-30
56
56
.0
Sy
2
(R
L)
SA
(l)
AS
CA
Tu
rn
er
et
al.
(19
97
)
1.5
83
34
00
0
70
NG
C
30
81
09
59
29
.5
-22
49
34
.7
Sy
2
SA
B(
r)0
/a
AS
CA
74
04
30
00
0.0
3
34
73
8
71
NG
C
32
27
10
23
30
.6
19
51
54
.0
Sy
1.5
SA
B(
s)
pe
c,
In
t
AS
CA
73
06
80
00
0.4
2
34
31
2
72
NG
C
32
81
10
31
52
.1
-34
51
13
.3
Sy
2
SA
B(
rs+
)a
AS
CA
74
05
80
00
0.0
15
16
95
9
73
2M
AS
X
J10
38
45
20
-49
46
53
1
10
38
44
.5
-49
46
57
.4
Sy
1
?
XR
T
00
03
52
25
00
3
0.0
87
16
90
7
74
LE
DA
09
39
74
10
40
22
.5
-46
25
25
.7
Sy
2
S0
XR
T
00
03
53
45
00
3
0.0
58
14
37
7
75
M
rk
41
7
10
49
30
.9
22
57
51
.8
Sy
2
Sa
XM
M
W
int
er
et
al.
(20
08
)
0.1
45
74
37
76
M
rk
42
1
11
04
27
.3
38
12
31
.7
BL
La
c
?
AS
CA
70
00
80
00
2.2
0
32
76
9
77
NG
C
35
16
11
06
47
.5
72
34
07
.0
Sy
1.5
(R
)SB
(s)
AS
CA
71
00
70
00
2.3
8
29
84
2
78
1R
XS
J11
27
16
.6+
19
09
14
11
27
16
.3
19
09
20
.2
Sy
1.8
?
XR
T
00
03
70
88
00
2
0.1
1
10
06
2
79
NG
C
37
83
11
39
01
.7
-37
44
19
.0
Sy
1
(R
?)S
B(
r)a
AS
CA
74
05
40
10
1.7
1
18
71
6
80
SB
S1
13
6+
59
4
11
39
09
.0
59
11
54
.6
Sy
1.5
?
XR
T
00
03
52
65
00
1
0.3
4
89
91
81
UG
C
06
72
8
11
45
16
.0
79
40
53
.4
Sy
1.2
SB
0/
a,
In
t?
XM
M
W
int
er
et
al.
(20
08
)
4.2
5
72
20
82
2M
AS
X
J11
45
40
45
-18
27
14
9
11
45
40
.5
-18
27
15
.5
Sy
1
Irr
,In
t
XR
T
00
03
56
45
00
1
0.7
2
94
02
83
CG
CG
04
1-0
20
12
00
57
.9
06
48
23
.0
Sy
2
S?
XM
M
W
int
er
et
al.
(20
08
)
0.4
65
97
77
84
IG
R
J12
02
6-5
34
9
12
02
47
.6
-53
50
07
.8
Sy
2
SB
0p
ec
,In
t
ch
an
dr
an
ot
pu
bli
c
85
NG
C
40
51
12
03
09
.6
44
31
52
.7
Sy
1.5
SA
B(
rs)
bc
AS
CA
70
00
10
00
0.9
17
25
97
0
86
Ar
k3
47
12
04
29
.7
20
18
58
.3
Sy
2
S0
:p
ec
XR
T
00
03
55
99
00
1,
02
0.0
13
21
70
6
87
NG
C
41
02
12
06
23
.1
52
42
39
.2
LI
NE
R
SA
B(
s)b
?
Ch
an
dr
a
too
sh
or
t?
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
311
Ta
ble
E.1
?C
on
tin
ue
d
No
.
So
ur
ce
RA
(h
m
s)
De
c(
?
??
? )
Ty
pe
1
Ho
st
Ga
lax
y1
Ob
s.2
Ob
sI
D/
ref
Ct
Ra
te
Ex
pT
im
e
88
NG
C
41
38
12
09
29
.8
43
41
07
.1
Sy
1.9
SA
(r)
0+
XM
M
Ca
pp
ie
ta
l.(
20
06
)
88
56
89
NG
C
41
51
12
10
32
.6
39
24
20
.5
Sy
1.5
(R
?)S
AB
(rs
)ab
AS
CA
78
00
10
03
1.2
55
46
00
6
90
M
rk
76
6
12
18
26
.5
29
48
46
.4
Sy
1.5
(R
?)S
B(
s)a
AS
CA
71
04
60
00
0.8
19
32
97
0
91
NG
C
43
88
12
25
46
.8
12
39
43
.6
Sy
2
SA
(s)
b
AS
CA
73
07
30
00
0.0
53
26
04
0
92
NG
C
43
95
12
25
48
.9
33
32
48
.5
Sy
1.9
SA
(s)
m,
In
t
AS
CA
78
00
90
00
0.0
79
36
08
4
93
3C
27
3
12
29
06
.7
02
03
08
.6
Bl
az
ar
?
AS
CA
70
02
30
00
3.8
49
31
96
4
94
NG
C
45
07
12
35
36
.6
-39
54
33
.5
Sy
2
SA
B(
s)a
b
AS
CA
Tu
rn
er
et
al.
(19
97
)
0.1
42
1
24
80
0
95
ES
O
50
6-G
02
7
12
38
54
.6
-27
18
28
.1
Sy
2
Sp
ec
sp
XM
M
W
int
er
et
al.
(20
08
)
0.2
7
89
19
96
XS
SJ
12
38
9-1
61
4
12
39
06
.3
-16
10
47
.6
Sy
2
?
XR
T
00
03
52
08
00
2
0.1
03
66
22
97
NG
C
45
93
12
39
39
.4
-05
20
39
.1
Sy
1
(R
)SB
(rs
)b
AS
CA
71
02
40
00
1.4
78
21
18
5
98
W
KK
12
63
12
41
25
.7
-57
50
03
.5
Sy
2?
?
XM
M
W
int
er
et
al.
(20
08
)
2.7
2
89
02
99
3C
27
9
12
56
11
.2
-05
47
21
.5
Bl
az
ar
?
AS
CA
70
02
60
00
0.2
99
37
19
5
10
0
SB
S1
30
1+
54
0
13
03
59
.5
53
47
30
.1
Sy
1
?
XM
M
W
int
er
et
al.
(20
08
)
10
.30
84
08
10
1
NG
C
49
45
13
05
27
.5
-49
28
05
.5
Sy
2
SB
(s)
cd
AS
CA
Tu
rn
er
et
al.
(19
97
)
0.0
45
61
00
10
2
NG
C
49
92
13
09
13
.0
11
38
45
.2
Ga
l
Sa
XM
M
W
int
er
et
al.
(20
08
)
0.1
37
12
84
9
10
3
M
CG
-03
-34
-06
4
13
22
24
.5
-16
43
43
.0
Sy
1.8
SB
?,
In
t
AS
CA
73
02
90
00
0.0
44
36
17
4
10
4
Ce
nA
13
25
27
.6
-43
01
09
.1
Sy
2
S0
pe
c,
In
t
10
5
M
CG
-06
-30
-01
5
13
35
53
.8
-34
17
44
.2
Sy
1.2
E-
S0
AS
CA
70
01
60
10
1.1
8
30
02
6
10
6
NG
C
52
52
13
38
16
.0
04
32
33
.4
Sy
1.9
S0
AS
CA
Tu
rn
er
et
al.
(19
97
)
0.0
63
27
30
0
10
7
4U
13
44
-60
13
47
24
.0
-60
38
24
.0
Sy
1.5
S
XR
T
00
03
52
84
00
2
0.3
88
14
06
5
10
8
IC
43
29
A
13
49
19
.3
-30
18
34
.6
Sy
1.2
SA
0+
,In
t
AS
CA
70
00
50
00
2.0
25
27
21
2
10
9
M
rk
27
9
13
53
03
.5
69
18
29
.5
Sy
1.5
S0
,In
t
AS
CA
72
02
80
00
1.3
42
21
35
7
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
312
Ta
ble
E.1
?C
on
tin
ue
d
No
.
So
ur
ce
RA
(h
m
s)
De
c(
?
??
? )
Ty
pe
1
Ho
st
Ga
lax
y1
Ob
s.2
Ob
sI
D/
ref
Ct
Ra
te
Ex
pT
im
e
11
0
NG
C
55
06
14
13
14
.9
-03
12
27
.0
Sy
1.9
Sa
pe
c,
In
t?
AS
CA
75
03
30
00
0.9
31
6
39
27
2
11
1
RB
S1
36
6
14
17
56
.7
25
43
26
.4
BL
La
c
?
XR
T
00
03
52
70
00
1
0.6
23
79
30
11
2
NG
C
55
48
14
17
59
.5
25
08
12
.5
Sy
1.5
(R
?)S
A(
s)0
/a
AS
CA
76
02
90
00
2.6
33
21
20
0
11
3
ES
O
51
1-G
03
0
14
19
22
.4
-26
38
40
.9
Sy
1
SA
(rs
)c
pe
c,
In
t
AS
CA
76
06
70
00
0.4
67
16
35
8
11
4
RB
S1
39
9
14
28
32
.7
42
40
20
.6
BL
La
c
?
AS
CA
76
05
30
00
1.4
4
36
00
0
11
5
NG
C
57
28
14
42
23
.9
-17
15
11
.5
Sy
2
(R
)SA
B(
r)a
Ch
an
dr
a
Zh
an
ge
ta
l.(
20
06
)
18
70
0
11
6
M
rk
84
1
15
04
01
.2
10
26
16
.1
Sy
1
E
XM
M
Pe
tru
cci
et
al.
(20
07
)
17
.9
58
61
11
7
M
rk
29
0
15
35
52
.4
57
54
09
.4
Sy
1
E1
?,
In
t?
AS
CA
72
02
70
00
0.3
08
41
01
8
11
8
M
rk
14
98
16
28
04
.1
51
46
31
.4
Sy
1.9
S?
XR
T
00
03
52
71
00
2
0.0
61
3
19
94
8
11
9
2M
AS
X
J16
48
15
23
-30
35
03
7
16
48
15
.2
-30
35
04
.2
Sy
1
?
XR
T
00
03
53
48
00
2
0.2
44
92
03
12
0
NG
C
62
40
16
52
58
.9
02
24
02
.9
Sy
I0:
pe
c,
In
t
12
1
M
rk
50
1
16
53
52
.2
39
45
36
.7
BL
La
c
E?
AS
CA
76
05
20
00
5.5
5
59
04
12
2
NG
C
63
00
17
16
59
.5
-62
49
14
.2
Sy
2
SB
(rs
)b
XM
M
M
ats
um
oto
et
al.
(20
04
)
12
3
GR
S1
73
4-2
92
17
37
24
.3
-29
10
48
.0
Sy
1
?
12
4
1R
XS
J17
45
38
.1+
29
08
23
17
45
38
.3
29
08
22
.2
Sy
1
?
XR
T
00
03
52
73
00
1
0.2
37
93
44
12
5
3C
38
2
18
35
02
.2
32
41
50
.3
Sy
1
?
AS
CA
Sa
mb
ru
na
et
al.
(19
99
)
1.7
93
40
50
0
12
6
ES
O
10
3-0
35
18
38
20
.3
-65
25
39
.4
Sy
2
SA
0,
In
t
AS
CA
Fo
rst
er
et
al.
(19
99
)
17
.31
15
70
0
12
7
3C
39
0.3
18
42
09
.0
79
46
17
.0
Sy
1
?
AS
CA
Sa
mb
ru
na
et
al.
(19
99
)
0.5
28
19
20
0
12
8
NV
SS
19
30
13
+3
41
04
7
19
30
13
.3
34
10
46
.9
Sy
1
?
XR
T
00
03
52
74
00
1,
02
0.0
13
10
24
7
12
9
NG
C
68
14
19
42
40
.7
-10
19
24
.6
Sy
1.5
SA
B(
rs)
bc
AS
CA
70
01
20
00
0.0
32
39
08
2
13
0
3C
40
3
19
52
15
.8
02
30
24
.5
Sy
2
S0
Ch
an
dr
a
Kr
aft
et
al.
(20
05
)
45
90
0
13
1
Cy
gA
19
59
28
.3
40
44
02
.0
Sy
2
Sc
,In
t
AS
CA
Sa
mb
ru
na
et
al.
(19
99
)
0.9
11
21
10
0
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
313
Ta
ble
E.1
?C
on
tin
ue
d
No
.
So
ur
ce
RA
(h
m
s)
De
c(
?
??
? )
Ty
pe
1
Ho
st
Ga
lax
y1
Ob
s.2
Ob
sI
D/
ref
Ct
Ra
te
Ex
pT
im
e
13
2
2M
AS
X
J19
59
59
75
+6
50
85
47
19
59
59
.9
65
08
54
.6
BL
La
c
E
XR
T
00
03
50
25
00
6
3.5
0
47
9.5
13
3
NG
C
68
60
20
08
46
.9
-61
06
00
.7
Sy
1.5
(R
?)S
B(
r)a
b
XM
M
W
int
er
et
al.
(20
08
)
0.4
6
11
81
5(
mo
s)
13
4
NG
C
69
21
20
28
28
.9
25
43
24
.3
Sy
2
SA
(r)
0/
a,
In
t
XM
M
W
int
er
et
al.
(20
08
)
0.2
55
87
89
13
5
M
CG
+0
4-4
8-0
02
20
28
35
.1
25
44
01
.0
Sy
2
Sc
d,
In
t
XM
M
W
int
er
et
al.
(20
08
)
0.0
96
87
89
13
6
4C
+7
4.2
6
20
42
37
.3
75
08
02
.4
Sy
1
?
AS
CA
74
09
70
00
0.6
15
21
21
0
13
7
M
rk
50
9
20
44
09
.7
-10
43
24
.6
Sy
1.2
S?
AS
CA
71
01
30
00
1.9
6
40
07
1
13
8
IC
50
63
20
52
02
.3
-57
04
07
.7
Sy
2
SA
(s)
0+
AS
CA
Tu
rn
er
et
al.
(19
97
)
0.0
96
23
90
0
13
9
2M
AS
X
J21
14
01
28
+8
20
44
83
21
14
01
.2
82
04
48
.4
Sy
1
?
XR
T
00
03
56
24
00
2
0.4
48
51
79
14
0
IG
R
J21
24
7+
50
58
21
24
38
.1
50
58
58
.1
Sy
1
?
XR
T
00
03
56
26
00
2
0.7
08
66
90
14
1
IG
R
21
27
7+
56
56
21
27
45
.9
56
56
34
.4
Sy
1
?
XR
T
00
03
52
15
00
3
0.4
26
83
72
14
2
RX
J21
35
.9+
47
28
21
35
55
.0
47
28
23
.2
Sy
1
?
XR
T
00
03
56
28
00
3
0.1
38
91
15
14
3
PK
S2
14
9-3
06
21
51
55
.5
-30
27
54
.0
Bl
az
ar
?
AS
CA
Sie
be
rt
et
al.
(19
96
)
0.2
4
19
00
0
14
4
UG
C
11
87
1
22
00
41
.4
10
33
08
.6
Sy
1.9
Sb
,In
t
XR
T
00
03
52
78
00
3
0.0
78
89
28
14
5
NG
C
71
72
22
02
01
.9
-31
52
11
.3
Sy
2
Sa
pe
cs
p
AS
CA
Tu
rn
er
et
al.
(19
97
)
0.0
06
6
14
90
0
14
6
NG
C
72
13
22
09
16
.2
-47
10
00
.1
Sy
1.5
Sa
AS
CA
71
02
60
00
1.2
26
27
09
1
14
7
NG
C
73
14
22
35
46
.2
-26
03
00
.7
Sy
1.9
SA
B(
rs)
bc
,In
t
AS
CA
Tu
rn
er
et
al.
(19
97
)
0.9
28
4
45
30
0
14
8
NG
C
73
19
22
36
03
.6
33
58
32
.5
Sy
2
SB
(s)
bc
pe
c,
In
t
XR
T
00
03
50
83
00
4
0.0
12
14
73
1
14
9
3C
45
2
22
45
48
.8
39
41
15
.7
Sy
2
S?
Ch
an
dr
a
15
0
3C
45
4.3
22
53
57
.7
16
08
53
.5
Bl
az
ar
?
XR
T
00
03
50
30
00
1
1.1
51
13
66
8
15
1
M
R
22
51
-17
8
22
54
05
.8
-17
34
54
.8
Sy
1
?
AS
CA
74
02
80
00
0.6
49
17
76
0
15
2
NG
C
74
69
23
03
15
.6
08
52
26
.4
Sy
1.2
(R
?)S
AB
(rs
)a,
In
t
AS
CA
71
02
80
10
1.5
93
18
70
7
15
3
M
rk
92
6
23
04
43
.5
-08
41
08
.5
Sy
1.5
Sb
c,
In
t
AS
CA
75
04
90
00
1.2
18
37
22
1
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
314
Ta
ble
E.1
?C
on
tin
ue
d
No
.
So
ur
ce
RA
(h
m
s)
De
c(
?
??
? )
Ty
pe
1
Ho
st
Ga
lax
y1
Ob
s.2
Ob
sI
D/
ref
Ct
Ra
te
Ex
pT
im
e
15
4
NG
C
75
82
23
18
23
.5
-42
22
14
.2
Sy
2
(R
?)S
B(
s)a
b,
In
t
AS
CA
Tu
rn
er
et
al.
(19
97
)
0.0
90
6
17
80
0
1 A
GN
typ
ei
sf
ro
m
Tu
ell
er
et
al.
(20
08
).
Fo
rA
GN
typ
es
,o
pt
ica
lid
en
tifi
ca
tio
ns
are
lis
ted
,w
he
re
av
ail
ab
le.
W
he
re
?G
al?
is
ind
ica
ted
,th
ere
are
no
op
tic
al
em
iss
ion
lin
es
ind
ica
tiv
eo
f
th
ep
res
en
ce
of
an
AG
N.
Th
eo
pt
ica
ls
pe
ctr
um
loo
ks
lik
ea
ga
lax
ys
pe
ctr
um
.A
dd
iti
on
al
ho
st
ga
lax
yc
las
sifi
ca
tio
ns
we
re
ob
tai
ne
d
fro
m
th
eN
ED
an
d
LE
DA
da
tab
as
es
.W
he
re
??
?
is
ind
ica
ted
,th
ere
is
no
av
ail
ab
le
cla
ssi
fic
ati
on
.
2 O
bs
erv
ati
on
sw
ere
co
lle
cte
df
ro
m
arc
hiv
al
or
pu
bli
sh
ed
wo
rk
fro
m
AS
CA
,C
ha
nd
ra,
XM
M
-N
ew
ton
,S
uz
ak
u,
an
dt
he
SW
IFT
XR
Ti
ns
tru
me
nt
.
315
Ta
ble
E.2
:X
-ra
yF
its
of
SW
IFT
XR
Ts
pe
ctr
a
So
ur
ce
n H
(G
al)
1
n H
(a
bs
)
Cv
r.F
rac
?
?2
/d
of
FX
2
typ
e
NG
C
23
5A
0.0
14
28
.25
+3
.79
?3
.37
0.9
97
5+
0.0
02
1
?0
.00
29
2.8
5+
0.4
0
?0
.66
20
.6/
21
0.0
6,
2.9
2
C
NG
C
45
4
0.0
26
15
.90
+4
.91
?3
.64
0.9
97
+0
.00
3
?0
.00
4
2.8
1+
1.3
3
?0
.88
19
.7/
15
0.0
23
,1
.96
4
C
Mr
k1
01
8
0.0
25
8
0.0
0+
0.0
1
?0
.00
?
1.9
3+
0.0
7
?0
.05
99
.5/
99
8.7
3,
11
.89
S
LE
DA
13
85
01
0.1
69
0.0
30
+0
.04
0
?0
.03
0
?
1.8
12
+0
.09
4
?0
.08
9
87
.6/
10
0
5.0
8,
11
.38
S
[H
B8
9]
02
12
+7
35
0.2
59
0.3
50
+0
.22
0
?0
.20
0
?
1.3
76
+0
.25
9
?0
.24
5
30
.6/
23
0.6
86
,5
.54
S
QS
O
B0
24
1+
62
2
0.7
42
0.1
00
+0
.07
0
?0
.06
0
?
1.7
33
+0
.08
3
?0
.08
0
19
7.8
/1
85
3.8
9,
25
.82
S
NG
C
13
65
0.0
14
2
10
4.8
0.7
4+
0.2
5
?0
.15
1.7
1+
0.2
9
?0
.23
55
.0/
43
0.4
0,
0.8
8
C
PG
C
13
94
6
0.0
15
6
14
.4+
7.4 ?5.9
0.9
4+
0.0
5
?0
.19
1.3
5+
1.1
4
?0
.70
9/
6
0.0
8,
2.9
6
C
2MA
SX
J03
56
56
55
-40
41
45
3
0.0
26
3.2
7+
0.9
2
?0
.61
?
1.9
4+
0.4
2
?0
.30
29
.4/
36
0.1
9,
5.5
8
S
2MA
SX
J04
44
09
03
+2
81
30
03
0.1
97
3.3
9+
0.3
1
?0
.25
0.9
89
96
+0
.00
3
?0
.00
3
1.3
7+
0.1
1
?0
.08
32
9.8
/3
31
0.0
2,
12
.34
C
1R
XS
J04
52
05
.0+
49
32
48
0.5
31
0.0
01
+0
.22
?0
.00
1
?
1.9
5+
0.1
6
?0
.11
11
0.1
/8
6
11
.4,
37
.8
S
XS
SJ
05
05
4-2
34
8
0.0
22
8
6.5
7+
3.0
3
?2
.39
?
1.5
1+
0.5
6
?0
.51
21
.9/
16
0.1
5,
13
.4
S
PKS
05
37
-44
1
0.0
39
4
?
?
1.7
8+
0.0
3
?0
.03
27
4.8
/2
37
4.4
1,
7.5
7
S
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
316
Ta
ble
E.2
?C
on
tin
ue
d
So
ur
ce
n H
(G
al)
1
n H
(a
bs
)
Cv
r.F
rac
?
?2
/d
of
FX
2
typ
e
IR
AS
05
58
9+
28
28
0.4
62
0.0
0+
0.0
4
?
1.6
0+
0.0
6
?0
.06
10
6.4
/9
8
3.9
5,
16
.42
S
ES
O
12
1-G
02
8
0.0
46
9
16
.19
+1
2.6
?9
.4
?
2.2
0+
1.5
5
?1
.32
4/
6
0,
2.9
7
S*
2MA
SX
J06
40
37
99
-43
21
21
1
0.0
62
4
16
.13
+1
0.0
?1
2.3
?
1.7
9+
1.4
1
?1
.51
6.6
/7
0.0
00
6,
1.5
8
S*
Mr
k6
0.0
62
3
3.2
6+
1.3
3
?1
.19
0.9
09
+0
.04
1
?0
.07
5
1.3
1+
0.4
0
?0
.37
31
.6/
26
0.4
4,
11
.36
C
SD
SS
J07
46
25
.87
+2
54
90
2.2
0.0
4
0.0
62
+0
.04
?0
.03
?
1.2
9+
0.0
9
?0
.08
95
.8/
84
1.6
3,
6.6
3
S
IG
R
J07
59
7-3
84
2
0.6
2
0.0
+0
.03
?
1.8
8+
0.0
6
?0
.05
18
5/
18
1
6.4
7,
27
.81
S
IR
AS
09
14
9-6
20
6
0.1
89
0.8
5+
0.2
6
?0
.17
?
1.7
4+
0.2
0
?0
.14
75
.7/
68
1.8
1,
11
.59
S
MC
G
-01
-24
-01
2
0.0
35
5
11
.44
+2
.82
?2
.27
?
2.1
7+
0.5
0
?0
.44
26
.8/
29
0.0
4,
8.2
8
S*
2MA
SX
J10
38
45
20
-49
46
53
1
0.2
62
1.5
5+
0.5
6
?0
.58
0.8
80
+0
.06
4
?0
.06
3
1.4
4+
0.1
8
?0
.20
77
.8/
66
0.6
0,
6.1
7
C
LE
DA
09
39
74
0.1
41
4.2
4+
1.0
6
?1
.00
0.9
86
+0
.00
7
?0
.01
4
2.1
2+
0.4
3
?0
.40
57
.7/
35
0.1
6,
4.9
5
C
1R
XS
J11
27
16
.6+
19
09
14
0.0
17
0.2
5+
0.1
0
?0
.08
?
1.7
9+
0.1
5
?0
.15
58
.4/
48
1.3
8,
3.7
7
S
SB
S1
13
6+
59
4
0.0
11
2
?
?
1.9
4+
0.0
5
?0
.04
13
2.6
/1
15
4.7
6,
6.1
5
S
2MA
SX
J11
45
40
45
-18
27
14
9
0.0
34
9
?
?
1.9
2+
0.0
3
?0
.03
21
2.7
/2
14
10
.37
,1
4.3
3
S
Ar
k3
47
0.0
23
6
30
.0+
8.7
6
?8
.88
0.9
95
+0
.00
5
?0
.01
7
2.1
7+
1.0
3
?0
.80
6.3
/1
3
0.0
3,
1.8
0
C
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
317
Ta
ble
E.2
?C
on
tin
ue
d
So
ur
ce
n H
(G
al)
1
n H
(a
bs
)
Cv
r.F
rac
?
?2
/d
of
FX
2
typ
e
XS
SJ
12
38
9-1
61
4
0.0
36
9
3.2
1+
0.8
3
?0
.71
?
1.8
4+
0.3
7
?0
.35
19
.2/
30
0.3
0,
9.5
7
S
4U
13
44
-60
1.0
5
1.4
5+
0.2
0
?0
.19
?
1.6
9+
0.0
9
?0
.09
23
1.7
/2
16
2.0
9,
29
.20
S
RB
S1
36
6
0.0
16
7
0.0
5+
0.0
1
?0
.01
?
2.0
8+
0.0
6
?0
.06
19
5.1
/1
76
9.2
6,
11
.03
S
Mr
k1
49
8
0.0
21
17
.84
+2
.37
?1
.82
0.9
90
+0
.00
5
?0
.00
6
1.9
2+
0.2
9
?0
.22
55
.8/
55
0.1
1,
7.2
0
C
2MA
SX
J16
48
15
23
-30
35
03
7
0.1
73
0.2
4+
0.0
6
?0
.06
?
1.7
2+
0.1
0
?0
.10
10
6.3
/9
6
2.9
3,
9.9
5
S
1R
XS
J17
45
38
.1+
29
08
23
0.1
37
0.0
2+
0.0
7
?0
.02
?
1.8
5+
0.0
7
?0
.07
84
.8/
89
3.2
5,
6.2
0
S
NV
SS
19
30
13
+3
41
04
7
0.1
73
27
.47
+2
0.3
7
?1
3.7
4
0.9
26
+0
.06
5
?0
.42
6
0.9
58
+1
.56
?1
.54
0.7
4/
3
0.0
3,
2.2
2
C
2MA
SX
J19
59
59
75
+6
50
85
47
0.1
03
0.0
7+
0.0
4
?0
.04
?
1.9
3+
0.1
1
?0
.10
77
.9/
71
50
.19
,9
1.7
5
S
2MA
SX
J21
14
01
28
+8
20
44
83
0.0
73
6
0.0
47
+0
.02
3
?0
.02
1
?
1.8
9+
0.0
9
?0
.09
99
.9/
96
6.2
7,
11
.92
S
IG
R
J21
24
7+
50
58
1.1
7
1.3
0+
0.2
3
?0
.21
?
1.3
3+
0.0
9
?0
.09
20
2.5
/1
94
3.0
1,
63
.18
S
IG
R
21
27
7+
56
56
0.8
06
0.4
24
+0
.11
?0
.11
?
2.0
2+
0.1
0
?0
.10
14
1.9
/1
44
4.2
9,
20
.25
S
RX
J21
35
.9+
47
28
0.3
85
0.4
4+
0.1
7
?0
.15
?
1.6
6+
0.1
6
?0
.15
52
.1/
56
1.3
7,
7.3
7
S
UG
C
11
87
1
0.0
54
4
2.3
9+
0.7
1
?0
.56
?
1.5
3+
0.2
7
?0
.25
39
.1/
31
0.3
9,
6.3
5
S*
NG
C
73
19
0.0
83
86
.48
+3
0.8
7
?2
0.5
5
0.9
96
5+
0.0
02
7
?0
.00
55
2.8
7+
0.5
3
?0
.28
1.6
/4
0.1
3,
1.4
9
C
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
318
Ta
ble
E.2
?C
on
tin
ue
d
So
ur
ce
n H
(G
al)
1
n H
(a
bs
)
Cv
r.F
rac
?
?2
/d
of
FX
2
typ
e
3C
45
4.3
0.0
64
5
0.0
8+
0.0
1
?0
.01
?
1.4
9+
0.0
3
?0
.03
44
6.7
/4
07
14
.31
,4
6.3
3
S
1 Mi
lky
W
ay
Ga
lac
tic
ab
so
rp
tio
no
bta
ine
df
ro
m
th
en
H
FT
OO
Lo
nH
EA
SA
RC
in
un
its
of
10
22
cm
?2
.
2 O
bs
erv
ed
flu
xi
nt
he
0.5
?2
ke
V
an
d2
?1
0k
eV
ba
nd
si
nu
nit
so
f1
0?
12
erg
s?
1 c
m?
2 .
Th
es
tat
ist
ica
le
rro
rs
on
flu
xe
sa
re
ve
ry
sm
all
an
d
th
es
ys
tem
ati
ce
rro
rs
are
do
mi
na
ted
by
mo
de
lu
nc
ert
ain
tie
s.
319
Ta
ble
E.3
:X
-ra
yS
pe
ctr
al
Fit
sf
ro
m
th
eL
ite
rat
ur
e/
Su
za
ku
XI
S/
TA
RT
AR
US
Fit
s
So
ur
ce
n H
(a
bs
)
?
pa
rm
ode
l3
Fe
K
E
Fe
K
EW
?2
/d
of
FX
2
mo
de
l
typ
e
M
rk
34
8
16
+
4
?
3
1.6
9+
0.
36
?
0.
39
?
6.2
7+
0.
08
?
0.
07
21
5+
75
?
75
11
7.4
/1
16
0.0
7,
4.8
4
do
ub
le
po
w
C
Fa
ira
ll9
?
1.9
1+
0.
01
?
0.
01
?
6.3
7+
0.
11
?
0.
13
35
0+
15
0
?
10
0
12
83
/1
14
9
18
.24 ,
19
.0
po
w
S
NG
C
52
6A
1.5
0+
0.
14
?
0.
14
1.8
2+
0.
16
?
0.
14
32
.5
6.4
11
1+
33
?
56
16
76
/1
58
6
3.5
4 ,3
5.4
do
ub
le
po
w
C
ES
O
29
7-0
18
41
.71
+
4.
70
?
2.
90
1.2
8+
0.
14
?
0.
16
0.9
93
+
0.
00
4
?
0.
00
6
6.3
8+
0.
02
?
0.
02
27
6+
57
?
51
18
0.7
/1
76
0.0
3,
3.2
6
pc
fab
s
C
NG
C
78
8
46
.89
+
4.
68
?
4.
47
1.8
9+
0.
24
?
0.
16
0.9
93
+
0.
00
3
?
0.
00
5
6.4
1+
0.
09
?
0.
12
14
8+
11
4
?
45
16
76
/1
58
6
0.1
1,
4.9
5
pc
fab
s
C
M
rk
59
0
?
1.6
2+
0.
02
?
0.
02
?
6.3
9+
0.
02
?
0.
02
12
1+
11
?
16
94
8/
96
4
4.4
,6
.4
po
w
S
NG
C
93
1
0.3
6+
0.
08
?
0.
08
1.7
5+
0.
11
?
0.
11
0.0
66
+
0.
01
1
?
0.
03
6
6.4
1+
0.
15
?
0.
15
55
0+
25
0
?
25
0
34
2/
36
5
1.1
44 ,
5.1
po
w
+b
b
S
NG
C
98
5
?
1.6
0+
0.
03
?
0.
03
0.0
10
+
0.
01
?
0.
01
?
?
17
9/
17
4
63
,5
4
po
w
+b
b
S
ES
O
41
6-G
00
2
<
0.0
27
1.6
3+
0.
06
?
0.
04
0.1
05
+
0.
00
6
?
0.
00
4
?
?
64
7/
63
4
7.5
3
po
w
+b
b
S
ES
O
19
8-0
24
?
1.8
4+
0.
02
?
0.
02
0.1
72
+
0.
00
7
?
0.
00
7
6.4
1
77
+
23
?
23
16
67
.9/
15
35
15
.8,
11
.0
po
w
+b
b
S
RB
S4
76
?
1.8
7+
0.
2
?
0.
2
?
?
?
47
.2/
50
10
.30
,1
8.0
2
po
w
S
3C
10
5
28
.16
+
1.
41
?
3.
23
1.3
6+
0.
18
?
0.
36
0.9
73
+
0.
00
5
?
0.
00
4
6.4
2+
0.
04
?
0.
06
21
6+
98
?
61
31
5.4
/3
17
0.0
4,
2.0
8
pc
fab
s
C
3C
11
1
0.6
2+
0.
03
?
0.
02
1.7
2+
0.
03
?
0.
02
?
?
?
16
53
/1
72
4
5.0
,3
5
po
w
S
1H
04
19
-57
7
?
1.4
8+
0.
07
?
0.
07
?
?
?
28
0/
31
9
4.8
,9
.4
po
w
S
3C
12
0
0.1
6+
0.
01
?
0.
01
2.0
0+
0.
02
?
0.
02
?
6.0
6+
0.
26
?
0.
21
>
40
0
13
9.9
/1
26
23
.54 ,
45
.9
po
w
S
M
CG
-01
-13
-02
5
<
0.0
03
9
1.7
0+
0.
03
?
0.
03
0.1
39
+
0.
02
7
?
0.
03
2
?
?
81
9.3
/8
74
21
.3
po
w
+b
b
S
Ar
k1
20
?
1.9
0+
0.
02
?
0.
03
0.2
0+
0.
01
?
0.
01
6.4
6+
0.
06
?
0.
08
12
3+
38
?
36
55
3/
44
9
17
.1,
27
.8
po
w
+b
b
S
Pic
tor
A
0.0
6+
0.
02
?
0.
02
1.7
7+
0.
04
?
0.
05
?
?
?
11
3/
10
6
6.9
7,1
1.3
3
po
w
S
ES
O
36
2-G
02
1
?
1.5
8+
0.
03
?
0.
03
?
?
?
36
5.6
/3
36
4.6
5,1
1.0
2
po
w
S
[H
B8
9]
05
37
-28
6
?
1.6
0+
0.
07
?
0.
07
?
?
?
14
2.8
/1
40
0.5
7,1
.75
po
w
S
PK
S0
54
8-3
22
0.1
03
+
0.
00
4
?
0.
00
4
2.0
5+
0.
02
?
0.
02
?
?
?
86
2.9
/8
46
25
,2
8
po
w
S
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
320
Ta
ble
E.3
?C
on
tin
ue
d
So
ur
ce
n H
(a
bs
)
?
pa
rm
ode
l3
Fe
K
E
Fe
K
EW
?2
/d
of
FX
2
mo
de
l
typ
e
NG
C
21
10
2.8
4+
0.
19
?
0.
16
1.5
4+
0.
08
?
0.
07
0.9
52
+
0.
00
6
?
0.
00
2
6.4
10
3+
29
?
29
62
8/
68
2
0.8
64 ,
25
.5
pc
fab
s
C
M
CG
+0
8-1
1-0
11
0.2
5+
0.
01
6
?
0.
01
5
1.6
4+
0.
03
?
0.
03
?
6.4
3+
0.
08
?
0.
08
38
0+
10
2
?
83
89
6/
86
0
5.5
,2
0.3
po
w
S
EX
O
05
56
20
-38
20
.2
2.5
7+
0.
14
?
0.
14
1.8
2+
0.
06
?
0.
07
0.9
66
+
0.
00
5
?
0.
00
5
6.2
6
20
0+
10
0
?
10
0
11
77
/1
21
5
1.4
4,
22
.47
pc
fab
s
C
ES
O
00
5-G
00
4
5.5
8+
0.
16
?
0.
16
1.4
8+
0.
04
?
0.
04
0.9
84
+
0.
00
2
?
0.
00
3
6.3
9+
0.
05
?
0.
05
79
+
9
?
15
13
16
.1/
11
61
0.1
5,
10
.41
pc
fab
s
C
M
rk
3
1.2
4+
5.
67
?
1.
24
2.9
4+
0.
55
?
0.
39
0.0
76
6.4
97
7+
19
3
?
13
7
13
4.1
/2
53
0.5
84 ,
2.1
5
do
ub
le
po
w
C
M
rk
79
<
0.0
06
3
1.8
5+
0.
04
?
0.
04
0.1
22
+
0.
00
7
?
0.
00
7
6.3
3+
0.
08
?
0.
09
11
7+
0.
15
?
0.
15
43
6.8
/4
80
42
.75
po
w
+b
b
S
4C
+7
1.0
7
0.1
1+
0.
03
?
0.
03
1.4
5+
0.
05
?
0.
05
?
6.4
<
11
0
43
7.6
/5
03
4.2
5,
14
.0
po
w
S
2M
AS
X
J09
18
00
27
+0
42
50
66
11
.13
+
1.
55
?
1.
18
1.7
7+
0.
22
?
0.
15
0.9
87
+
0.
00
6
?
0.
00
5
6.3
9+
0.
05
?
0.
05
65
.6+
39
.4
?
30
.2
31
0.6
/3
33
0.0
2,
1.3
8
pc
fab
s
C
M
rk
11
0
0.0
2+
0.
01
?
0.
01
1.7
8+
0.
02
?
0.
02
?
?
?
48
4.4
/4
51
16
.55
,2
7.9
3
po
w
S
NG
C
29
92
1.1
9+
2.
21
?
0.
96
1.5
4+
0.
41
?
0.
25
0.4
76
+
0.
20
9
?
0.
36
8
6.4
65
6+
20
4
?
19
1
25
2/
25
7
0.8
64 ,
3.5
1
pc
fab
s
C
M
CG
-05
-23
-01
6
1.6
0+
0.
00
5
?
0.
00
6
1.9
0+
0.
04
?
0.
04
13
9
6.4
79
+
16
?
14
19
95
/1
74
4
7.8
44 ,
85
.1
do
ub
le
po
w
C
NG
C
30
81
94
.2+
6.
2
?
7.
2
1.6
0+
0.
12
?
0.
17
0.9
90
+
0.
01
1
?
0.
00
2
6.3
8+
0.
03
?
0.
05
19
1+
66
?
58
30
3/
24
5
0.1
8,
5.5
3
pc
fab
s
C
NG
C
32
27
1.7
4+
0.
12
?
0.
09
1.3
5+
0.
04
?
0.
04
0.8
52
+
0.
00
6
?
0.
01
0
6.3
6+
0.
03
?
0.
06
14
2+
22
?
40
48
8.1
/4
45
2.4
4,
28
.29
pc
fab
s
C
NG
C
32
81
86
.3+
16
.3
2
?
16
.1
2
1.6
7+
0.
44
?
0.
50
0.9
81
+
0.
01
2
?
0.
02
5
6.3
6+
0.
06
?
0.
06
61
7+
22
5
?
18
1
15
3.6
/1
35
0.0
1,
2.5
8
pc
fab
s
C
M
rk
42
1
0.0
77
+
0.
00
7
?
0.
00
7
2.9
5+
0.
01
?
0.
01
?
?
?
62
1.1
/4
39
61
.22
,2
1.1
9
po
w
S
NG
C
35
16
0.3
53
+
0.
32
?
0.
12
1.7
3+
0.
01
?
0.
01
0.9
99
98
+
2e
?
5
?
0.
04
62
6.3
2+
0.
09
?
0.
02
10
4+
20
?
20
58
4.2
/4
13
2.4
4,
28
.29
pc
fab
s,
bb
C
NG
C
37
83
0.5
7+
0.
21
?
0.
14
1.6
0+
0.
04
?
0.
03
0.7
22
+
0.
10
7
?
0.
08
2
6.3
4+
0.
07
?
0.
04
80
+
26
?
29
55
9.9
/4
21
23
.96
,7
0.4
9
pc
fab
s,
bb
C
NG
C
40
51
?
1.9
5+
0.
02
?
0.
02
0.0
87
+
0.
00
5
?
0.
00
3
6.4
3+
0.
02
?
0.
02
14
7+
41
?
40
50
9.7
/4
34
24
.11
,2
1.8
6
po
w
+b
b
S
NG
C
41
38
8.0
0+
1.
00
?
1.
00
1.5
+
0.
1
?
0.
1
0.3
+
0.
1
?
0.
1
6.4
83
+
30
?
30
40
8/
40
8
0.0
6,
5.5
po
w
+b
b
S*
NG
C
41
51
5.3
2+
0.
07
?
0.
08
1.4
6+
0.
02
?
0.
02
0.9
59
+
0.
00
2
?
0.
00
1
6.3
6+
0.
01
?
0.
04
13
7+
10
?
14
69
2/
41
3
3.3
1,
12
9.3
6
pc
fab
s
C
M
rk
76
6
?
1.7
8+
0.
01
?
0.
01
0.0
43
+
0.
00
7
?
0.
00
5
6.3
8+
0.
07
?
0.
06
75
+
40
?
32
49
8.2
/4
41
15
.65
,2
0.0
9
po
w
+b
b
S
NG
C
43
88
36
.17
+
3.
81
?
3.
82
1.7
9+
0.
17
?
0.
17
0.9
54
+
0.
01
3
?
0.
01
6
6.4
4+
0.
04
?
0.
04
46
6+
88
?
92
39
6.5
/3
44
0.5
2,
5.9
0
pc
fab
s
C
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
321
Ta
ble
E.3
?C
on
tin
ue
d
So
ur
ce
n H
(a
bs
)
?
pa
rm
ode
l3
Fe
K
E
Fe
K
EW
?2
/d
of
FX
2
mo
de
l
typ
e
NG
C
43
95
3.3
0+
0.
68
?
0.
67
1.0
7+
0.
12
?
0.
06
0.6
78
+
0.
05
6
?
0.
06
2
6.2
2+
0.
11
?
0.
16
10
1+
69
?
64
31
8.6
/3
59
0.5
1,
5.9
5
pc
fab
s
C
3C
27
3
?
1.6
1+
0.
01
?
0.
01
0.1
54
+
0.
02
1
?
0.
01
3
?
?
57
4.0
6/
45
2
57
.44
,1
16
.13
po
w
+b
b
S
NG
C
45
07
34
.28
+
4.
50
?
4.
57
1.6
5+
0.
35
?
0.
38
66
6.4
16
7+
34
?
33
56
4/
49
0
0.5
14 ,
22
.8
do
ub
le
po
w
C
NG
C
45
93
0.0
31
+
0.
01
1
?
0.
01
2
1.8
6+
0.
01
?
0.
02
?
6.4
1+
0.
06
?
0.
08
10
5+
49
?
23
58
3.8
/4
44
21
.11
,3
4.1
9
po
w
S
3C
27
9
0.0
20
+
0.
01
7
?
0.
01
6
1.7
8+
0.
04
?
0.
03
?
?
?
44
9.2
/4
30
4.5
9,
8.0
6
po
w
S
NG
C
49
45
0.7
93
+
0.
95
7
?
0.
49
5
1.7
5+
0.
31
?
0.
14
0.6
91
+
0.
21
9
?
0.
59
0
6.4
10
20
+
21
0
?
29
8
25
5/
23
3
0.7
64 ,
2.9
1
pc
fab
s
C
M
CG
-03
-34
-06
4
40
.73
+
4.
79
?
4.
30
1.8
3+
0.
28
?
0.
23
0.9
61
+
0.
01
4
?
0.
00
8
6.4
4+
0.
09
?
0.
10
19
1+
12
3
?
74
30
6.9
/2
61
0.4
1,
4.7
8
pc
fab
s,
ap
ec
C
M
CG
-06
-30
-01
5
0.1
9+
0.
03
?
0.
01
1.7
8+
0.
03
?
0.
02
0.0
57
+
0.
00
3
?
0.
00
3
6.4
1
71
.4+
29
?
40
48
8.2
/4
50
20
.41
,3
7.6
5
po
w
+b
b
S
NG
C
52
52
4.3
4+
0.
52
?
0.
42
1.5
5+
0.
16
?
0.
13
0.9
62
+
0.
00
9
?
0.
00
9
6.4
76
+
54
?
53
48
3/
47
8
0.1
73
4 ,5
.81
pc
fab
s
C
IC
43
29
A
0.6
1+
0.
03
?
0.
03
1.8
0+
0.
01
?
0.
01
0.9
41
+
0.
01
6
?
0.
01
5
6.4
0
99
+
12
3
?
18
86
2.8
/4
52
21
.91
,8
1.4
8
pc
fab
s
C
M
rk
27
9
<
0.0
13
2.0
4+
0.
02
?
0.
02
?
6.4
22
3+
28
?
66
64
4.9
/4
47
22
.01
,2
5.9
8
po
w
S
NG
C
55
06
2.7
8+
0.
05
?
0.
05
1.7
6+
0.
03
?
0.
02
0.9
89
3+
0.
00
1
?
0.
00
1
6.4
0
13
7+
14
?
24
42
6.4
/4
13
2.8
1,
67
.59
pc
fab
s
C
NG
C
55
48
0.0
7+
0.
04
?
0.
05
1.8
8+
0.
01
?
0.
01
0.0
31
+
0.
01
6
?
0.
00
3
6.4
1
73
+
14
?
44
50
2.3
/4
23
40
.34
,5
9.6
0
po
w
+b
b
S
ES
O
51
1-G
03
0
0.0
98
+
0.
02
1
?
0.
02
1
1.8
2+
0.
04
?
0.
04
?
6.4
1
12
5+
29
?
88
38
4.5
/3
88
6.2
6,
13
.04
po
w
S
RB
S1
39
9
0.1
02
+
0.
00
7
?
0.
00
7
1.9
5+
0.
01
?
0.
01
?
?
?
49
9.0
/4
17
19
.49
,3
3.7
9
po
w
S
NG
C
57
28
82
.0+
5.
3
?
5.
0
0.9
4+
0.
08
?
0.
07
0.0
11
6.3
2+
0.
04
?
0.
04
89
0
16
/1
6
0.3
0,
1.4
4
pc
fab
s
C
M
rk
84
1
?
1.8
1+
0.
07
?
0.
09
0.2
0+
0.
01
?
0.
01
6.2
5+
15
?
14
90
+
21
0
?
35
54
8/
24
4
16
,1
0
po
w
+b
b
S
M
rk
29
0
0.1
5+
0.
03
?
0.
05
1.6
1+
0.
04
?
0.
04
0.0
5+
0.
01
?
0.
01
6.4
17
6+
79
?
76
52
3.3
/4
40
4.8
4,
9.3
1
po
w
+b
b
S
M
rk
50
1
0.1
8+
0.
01
?
0.
01
2.0
1+
0.
01
?
0.
01
?
?
?
52
3.8
/4
16
74
.96
,1
26
.03
po
w
S
NG
C
63
00
21
.5+
0.
8
?
0.
9
1.8
3+
0.
08
?
0.
08
0.0
3
6.4
3+
0.
01
?
0.
02
14
8+
18
?
18
42
3.5
/4
20
0.0
37
,8
.6
**
C
3C
38
2
?
1.1
2+
0.
10
?
0.
29
1.9
4+
0.
24
?
0.
06
6.3
0+
0.
21
?
0.
32
48
9+
20
9
?
21
0
10
11
/9
82
18
.00
,3
1.0
0
po
w
+b
rem
s
S
ES
O
10
3-0
35
21
.6+
2.
6
?
2.
5
2.0
8+
0.
29
?
0.
28
0.0
76
6.3
63
+
0.
27
0
?
0.
14
4
20
0+
23
0
?
11
0
53
4/
52
7
0.1
04 ,
24
.10
do
ub
le
po
w
C
Co
nt
inu
ed
on
Ne
xt
Pa
ge
...
322
Ta
ble
E.3
?C
on
tin
ue
d
So
ur
ce
n H
(a
bs
)
?
pa
rm
ode
l3
Fe
K
E
Fe
K
EW
?2
/d
of
FX
2
mo
de
l
typ
e
3C
39
0.3
0.1
2+
0.
03
?
0.
03
1.5
0+
0.
03
?
0.
03
?
6.5
2+
0.
10
?
0.
10
10
8+
60
?
57
80
7.8
/7
92
7.4
84 ,
19
.4
po
w
S
NG
C
68
14
?
1.7
4+
0.
07
?
0.
06
?
6.4
54
5+
21
7
?
21
6
22
1.8
/2
50
0.5
3,
1.2
5
po
w
S
3C
40
3
45
.0+
7.
0
?
6.
0
1.7
6+
0.
2
?
0.
2
0.0
04
4
6.3
2+
0.
2
?
0.
2
24
4+
20
?
20
31
.1/
50
0.0
5,
0.7
55
do
ub
le
po
w
C
Cy
gA
11
.0+
21
.0
?
6.
0
1.8
0+
0.
28
?
0.
43
?
?
?
96
9.3
/9
32
16
.0,
24
.0
po
w
+R
-S
C
4C
+7
4.2
6
0.1
77
+
0.
03
5
?
0.
01
6
1.8
6+
0.
03
?
0.
03
1.0
0?
0.
09
1
6.4
11
0+
42
?
41
41
3/
40
5
6.6
7,
19
.61
pc
fab
s
C
M
rk
50
9
0.0
15
+
0.
00
8
?
0.
00
8
1.8
7+
0.
01
?
0.
01
?
6.4
86
.1+
50
?
15
57
3/
44
5
28
.32
,4
5.8
3
po
w
S
IC
50
63
21
.78
+
2.
24
?
2.
06
1.6
3+
0.
22
?
0.
22
63
.1
6.4
10
8+
56
?
54
52
6/
46
1
0.2
64 ,
13
.1
do
ub
le
po
w
C
PK
S2
14
9-3
06
0.0
7+
0.
02
?
0.
02
1.5
7+
0.
05
?
0.
05
?
?
?
36
4.6
/3
44
3.3
94 ,
10
.6
po
w
S
NG
C
71
72
8.1
9+
3.
42
?
3.
30
1.6
9+
0.
08
?
0.
09
?
6.4
68
+
36
?
35
69
0/
72
9
0.1
28
4 ,3
7.0
po
w
S
NG
C
72
13
0.0
25
+
0.
01
1
?
0.
01
2
1.7
5+
0.
02
?
0.
02
?
6.4
13
3+
49
?
24
51
9.9
/4
45
16
.65
,3
1.1
8
po
w
S
NG
C
73
14
1.1
6+
0.
01
?
0.
14
2.4
1+
0.
05
?
0.
02
18
.5
6.4
18
2+
46
?
47
17
02
/1
56
1
7.3
54 ,
35
.6
do
ub
le
po
w
C
3C
45
2
22
.98
+
2.
78
?
2.
83
-0.
08
+
0.
08
?
0.
25
0.9
36
+
0.
01
2
?
0.
03
4
6.4
10
0
13
5.5
/1
10
0.0
1,
2.1
7
pc
fab
s
C
M
R
22
51
-17
8
0.2
8+
0.
11
?
0.
08
1.4
1+
0.
04
?
0.
05
0.0
9+
0.
03
?
0.
06
6.4
67
40
6.8
/4
37
6.3
9,
27
.24
po
w
+b
b
S
NG
C
74
69
?
1.9
8+
0.
01
?
0.
01
?
6.4
13
1+
38
?
39
56
2.9
/4
44
24
.10
,3
2.5
5
po
w
S
M
rk
92
6
0.0
35
+
0.
01
?
0.
01
1.7
9+
0.
02
?
0.
02
?
6.4
57
+
28
?
28
57
4.3
/4
16
15
.70
,3
3.1
3
po
w
S
NG
C
75
82
7.3
9+
1.
46
?
1.
00
1.3
8+
0.
23
?
0.
20
23
.5
6.4
10
6+
66
?
64
49
7/
39
2
0.4
54 ,
13
.2
do
ub
le
po
w
C
1 M
ilk
yW
ay
Ga
lac
tic
ab
so
rp
tio
no
bta
ine
df
ro
m
th
en
H
FT
OO
Lo
nH
EA
SA
RC
in
un
its
of
10
22
cm
?2
.
2 O
bs
erv
ed
flu
xi
nt
he
0.5
?2
ke
V
an
d2
?1
0k
eV
ba
nd
si
nu
nit
so
f1
0?
12
erg
s?
1 c
m?
2 .
W
he
re
on
ly
on
en
um
be
ri
ss
ho
wn
,th
ev
alu
ei
st
he
0.3
?1
0k
eV
flu
x.
Th
es
tat
ist
ica
le
rro
rs
on
flu
xe
sa
re
ve
ry
sm
all
an
dt
he
sy
ste
ma
tic
err
or
sa
re
do
mi
na
ted
by
mo
de
lu
nc
ert
ain
tie
s.
3 T
his
pa
ram
ete
ra
cco
un
ts
fro
m
ex
tra
co
mp
on
en
ts
ref
err
ed
to
in
th
em
od
el
co
lum
n.
If
th
em
od
el
inc
lud
es
:b
bo
dy
th
en
th
is
is
th
et
em
pe
rat
ur
ei
n
323
ke
V
(kT
),d
ou
bl
e
po
wt
he
nt
his
is
th
er
ati
oo
fn
or
ma
liz
ati
on
sb
etw
ee
nt
he
fir
st
an
ds
ec
on
dp
ow
er
law
co
mp
on
en
ts,
an
dp
cf
ab
st
he
nt
his
is
th
e
co
ve
rin
gf
rac
tio
n.
4 E
sti
ma
te
of
0.5
-2
ke
V
flu
xf
ro
m
Ta
rta
ru
s,
on
ly
2-1
0k
eV
flu
xw
as
giv
en
in
or
igi
na
lp
ap
er.
324
E.2 Notes on Individual Sources
Here, we present notes on individual sources. These notes include com-
plexities associated with the sources. Particularly, in notes to Table E.1,
we include details on why sources needed to be excluded or why XRT
observations were used instead of data available from other missions.
Also, we include details on extra components added in order to model
the X-rayspectra, corresponding to the XRTanalysis(Notes to TableE.2)
or the ASCA/XMM-Newton/Chandra/Suzaku analyses (Notes to Ta-
ble E.3). Throughout, the sources are ordered in RA.
Notes to Table E.1
NGC 1275: This source is located within the Perseus cluster. As such,
emission from the cluster is difficult to separatefrom the AGN emission.
Therefore, we do not include an analysis of this source.
PKS 0326-288: The XRT observation of this source is too short to extract
an X-ray spectrum.
NGC 1365: This source is a ?changing-look? AGN, observed in both
Compton-thin and Compton-thick states. The source has been observed
extensively by XMM-Newton and Chandra with relevant papers ? for
instance Risaliti (2007) and Risaliti et al. (2005). For simplicity, we aver-
aged the XRT observations, taken within a day of each other. However,
the spectral parameters of this source, particularly column density, are
quite variable.
4U 0517+17: The XMM-Newton observation of this source is not yet
publicly available.
325
Mrk 6: This source is known to have a complex absorber, with ASCA
and XMM-Newton spectra well modeled by a double partial covering
model (Feldmeier et al. 1999; Immler et al. 2003). Due to its complex na-
ture, we only include our XRT analysis with a simple model, for easy
comparison with the other sources.
IGR J12026-5349: The Chandra observation of this source is not yet
publicly available.
NGC 4102: The Chandra observation is too short to compare with the
spectra of other sources in this survey.
Cen A: Cen A recently underwent a galaxy merger. The X-ray spectrum
is extremely complicated, as presented in (Evans et al. 2004). Due to the
complex form, we do not include an analysis of this source.
NGC 6240: This peculiar source hosts two active nuclei, as found by
Chandra (Komossa et al. 2003). Therefore, we do notinclude an analysis
of this source.
GRS 1734-292: This source is a Sy 1 seen through the Galactic plane in
a heavily obscured region. Therefore, the exact hydrogen column is not
known. An analysis of the spectrum of this source is not included.
Notes to Table E.2
NGC 1365: The spectral fit to this source also includes the addition of
an apec model with kT of 0.74+0.07?0.10 keV and an Fe K line at 6.54+0.08?0.09 keV
with EW ? 780eV. The fitted column density has large errors.
2MASX J04440903+2813003: The residuals to this fit show clear evi-
dence of an Fe K-? line (??2 = 20.9). Adding a gaussian for this com-
ponent, with ? fixed at 0.01keV, we find E = 6.41+0.05?0.03 keV and EW=
326
140+70?54.6 eV.
1RXS J045205.0+493248: The model includes a significant (??2 = 23)
soft excess modeled with a blackbody (kT= 0.10+0.05?0.03). The residuals
indicate the presence of possible additional lines (including Fe K-?) in
the spectrum, but a higher quality spectrum is necessary to fully resolve
these features.
Mrk 6: The residuals to this fit show clear evidence of an Fe K-? line
(??2 = 10.7). Adding a gaussian for this component, we find E =
6.41+0.07?0.06 keV, ? = 0.09+0.10?0.05 keV, and EW= 498+264?250 eV.
SDSS J074625.87+254902.2: Sambruna et al. (2006) provide an analysis
of the BAT + XRT spectra, however, we re-analyze an observation for
comparison of the parameters from a fit without the BAT spectrum.
IRAS 09149-6206: An unabsorbed blackbody component was required
(??2 = 100)atsoftenergies, probablyaforegroundGalacticobjectgiven
the AGN?s position in the Galactic plane, with kT = 0.13?0.02keV.
LEDA 093974: The residuals to this fit show evidence of an Fe K-? line
(??2 = 5.4). Adding a gaussian for this component, with ? fixed at
0.01keV, we find E = 6.46+0.09?0.14 keV and EW= 297+326?187 eV.
1RXSJ112716.6+190914: Ablackbodycomponentwasalsoadded(??2 ?
20) with kT= 52+13?12 eV. There are some residuals in the fit but a higher
quality spectrum is necessary to investigate these features.
1RXS J174538.1+290823: We model a significant soft excess (??2 = 20)
is added with kT= 36.1+8.2?9.6 eV.
NGC 7319: We also include an Fe K line in this fit with E fixed at 6.4keV
with ? = 0.01keV. The resultant EW = 319+313?296 eV.
327
Notes to Table E.3
RBS 476: We found that the ASCA observation of RBS 476 had the
wrong position in the ASCA catalog and that a large part of the emis-
sion was outside the observation?s field of view. Using archived XMM-
Newton data, wefound thata brokenpowerlawisthebest fittothedata,
with the indicated fluxes recorded in the table.
EXO 055620-3820.2: We note that the authors (see Table 1 for the refer-
ence) suggest that an ionized partial covering absorber is a better physi-
cal explanation than neutral absorption.
ESO 005-G004: In addition to the partial covering model, a thermal
model is also required (apec) at soft energies with kT = 0.57+0.09?0.12 keV.
Mrk110: ThespectrumisnotedtobemorecomplexintheXMM-Newton
spectra analyzed by Boller et al. (2007). However, for our purposes a
simple power law is a good fit.
NGC 3516: NGC 3516 has a very complex spectrum. A very significant
(??2 ? 600) blackbody component was also added with kT= 52.2+1.4?2.0 eV
and a normalization of 0.086. This source is extensively studied, partic-
ularly because of the complex absorption present.
NGC 3728: NGC 3783 has a complex spectrum, which has been ob-
served with a 900ks Chandra observation. A very significant (??2 =
214) blackbody component was also added with kT= 63.4+0.6?0.7 eV and a
normalization of 0.008.
NGC 4138: Alternatively, this source can be fit with a partial covering
model with a high amount of scattering (> 99%). Further, the black-
body component can alternatively be fit with an apec model of similar
328
temperature with no difference in ?2.
MCG -03-34-064: The spectrum also has a significant (??2 = 34) soft
component, which we fit with an apec model with kT= 0.82+0.11?0.09 keV.
A more complete analysis of this source is presented in Miniutti et al.
(2007), where they model the spectra as reflection-dominated with ther-
mal plasmas, a broad Fe K line, and additional Fe emission lines.
IC 4329A: This source has complex absorption, studied most recently in
the XMM-Newton spectrum analyzed in Steenbrugge et al. (2005). They
find signatures from 7 different absorbing systems.
NGC5506: AfulldescriptionoftheX-rayhistoryandpropertiesofNGC
5506 is found in Bianchi et al. (2003) where BeppoSax observations are
also presented (a strong reflection component is evident).
NGC 5728: Also included is a thermal model (mekal) with kT = 0.15?
0.01keV. A reflection model is also used, which is a worse fit to the
data. We performed an independent analysis of the spectrum confirm-
ing these results.
Mrk 841: This source has complex iron features. The authors (see Ta-
ble 1 for reference) get an acceptable fit when a relativistic disk line is
included along with the narrow gaussian fit to Fe-K?. The blackbody
component is also an unacceptable approximation for the soft excess.
NGC 6300: In the analysis, the authors (see Table 1 for reference) model
the soft and hard spectrum separately. The additional parameter is the
ratio of the hard to soft power law flux.
Cyg A: For Cyg A, the authors include a Raymond-Smith model to de-
scribe the soft emission in the spectrum with kT = 0.88+0.07?0.11 keV, Y =
0.53+4.57?0.19.
329
3C 452: 3C 452 has an extremely flat spectrum which is much better fit
with the addition of a reflection model (Evans et al. 2006). The values
quoted are from our own analysis of the data with a simplistic model.
330
Bibliography
Anders, E., & Ebihara, M. 1982, Geochim. Cosmochim. Acta, 46, 2363
Anderson, S. F., Voges, W., Margon, B., Tr?umper, J., Ag?ueros, M. A., Boller, T.,
Collinge, M. J., Homer, L., Stinson, G., Strauss, M. A., Annis, J., G?omez, P.,
Hall, P. B., Nichol, R. C., Richards, G. T., Schneider, D. P., Vanden Berk, D. E.,
Fan, X., Ivezi?c, ?Z., Munn, J. A., Newberg, H. J., Richmond, M. W., Weinberg,
D. H., Yanny, B., Bahcall, N. A., Brinkmann, J., Fukugita, M., & York, D. G.
2003, AJ, 126, 2209
Antonucci, R. R. J., & Miller, J. S. 1985, ApJ, 297, 621
Arnaud, K. 1996, Astronomical Data Analysis Software and Systems V, 101, 5
Avni, Y. 1976, ApJ, 210, 642
Awaki, H., Ueno, S., Taniguchi, Y., & Weaver, K. A. 2000, ApJ, 542, 175
Barger, A. J., Cowie, L. L., Mushotzky, R. F., Yang, Y., Wang, W.-H., Steffen, A. T.,
& Capak, P. 2005, AJ, 129, 578
Barr, P., & Mushotzky, R. F. 1986, Nature, 320, 421
Bauer, F. E., Brandt, W. N., & Lehmer, B. 2003a, AJ, 126, 2797
Bauer, F. E., Vignali, C., Alexander, D. M., Brandt, W. N., Garmire, G. P., Horn-
schemeier, A. E., Broos, P., Townsley, L., & Schneider, D. P. 2003b, Astronomis-
che Nachrichten, 324, 175
Baumgartner, W., Tueller, J., Mushotzky, R., Barthelmy, S., Cannizzo, J., Gehrels,
N., Markwardt, C., Skinner, G., & Winter, L. 2008, The Astronomer?s Telegram,
331
1429, 1
Baumgartner, W. H., & Mushotzky, R. F. 2006, ApJ, 639, 929
Begelman, M. C. 2002, ApJ, 568, L97
Belloni, T., Mendez, M., King, A. R., van der Klis, M., & van Paradijs, J. 1997, ApJ,
488, L109+
Bennert, N., Jungwiert, B., Komossa, S., Haas, M., & Chini, R. 2006, A&A, 459, 55
Bevington, P. R. 1969, Data reduction and error analysis for the physical sciences
(New York: McGraw-Hill, 1969)
Bianchi, S., Balestra, I., Matt, G., Guainazzi, M., & Perola, G. C. 2003, A&A, 402,
141
Bianchi, S., Guainazzi, M., Matt, G., & Fonseca Bonilla, N. 2007, A&A, 467, L19
Boldt, E. 1987, Phys. Rep., 146, 215
Boller, T., Balestra, I., & Kollatschny, W. 2007, A&A, 465, 87
Brandt, W. N., & Hasinger, G. 2005, ARA&A, 43, 827
Braun, R. 1995, A&AS, 114, 409
Brenneman, L. W., Reynolds, C. S., Wilms, J., & Kaiser, M. E. 2007, ApJ, 666, 817
Burbidge, G., Burbidge, E. M., & Arp, H. 2003, A&A, 400, L17
Calzetti, D., Kinney, A. L., & Storchi-Bergmann, T. 1994, ApJ, 429, 582
Cappi, M., Matsuoka, M., Comastri, A., Brinkmann, W., Elvis, M., Palumbo,
G. G. C., & Vignali, C. 1997, ApJ, 478, 492
Cappi, M., Panessa, F., Bassani, L., Dadina, M., Dicocco, G., Comastri, A., della
Ceca, R., Filippenko, A. V., Gianotti, F., Ho, L. C., Malaguti, G., Mulchaey, J. S.,
Palumbo, G. G. C., Piconcelli, E., Sargent, W. L. W., Stephen, J., Trifoglio, M., &
Weaver, K. A. 2006, A&A, 446, 459
Chaty, S. 2008, ArXiv e-prints, 805
Chen, W., Shrader, C. R., & Livio, M. 1997, ApJ, 491, 312
332
Colbert, E. J. M., & Ptak, A. F. 2002, ApJS, 143, 25
Comastri, A. 2004, in Astrophysics and Space Science Library, Vol. 308, Super-
massive Black Holes in the Distant Universe, ed. A. J. Barger, 245?+
Crenshaw, D. M., Kraemer, S. B., & George, I. M. 2003, ARA&A, 41, 117
Cropper, M., Soria, R., Mushotzky, R. F., Wu, K., Markwardt, C. B., & Pakull, M.
2004, MNRAS, 349, 39
Czerny, B., Niko?ajuk, M., R?o?za?nska, A., Dumont, A.-M., Loska, Z., & Zycki, P. T.
2003, A&A, 412, 317
Dai, X., Chartas, G., Eracleous, M., & Garmire, G. P. 2004, ApJ, 605, 45
Das, V., Crenshaw, D. M., Kraemer, S. B., & Deo, R. P. 2006, AJ, 132, 620
Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604
Dickey, J. M., & Lockman, F. J. 1990, ARA&A, 28, 215
Done, C., & Gierli?nski, M. 2003, MNRAS, 342, 1041
Ebisawa, K., Mitsuda, K., & Hanawa, T. 1991, ApJ, 367, 213
Ebisawa, K., ?Zycki, P., Kubota, A., Mizuno, T., & Watarai, K.-y. 2003, ApJ, 597, 780
Eracleous, M., & Halpern, J. P. 1998, ApJ, 505, 577
Eracleous, M., Shields, J. C., Chartas, G., & Moran, E. C. 2002, ApJ, 565, 108
Evans, D. A., Kraft, R. P., Worrall, D. M., Hardcastle, M. J., Jones, C., Forman,
W. R., & Murray, S. S. 2004, ApJ, 612, 786
Evans, D. A., Worrall, D. M., Hardcastle, M. J., Kraft, R. P., &Birkinshaw, M. 2006,
ApJ, 642, 96
Fabian, A. C., Barcons, X., Almaini, O., & Iwasawa, K. 1998, MNRAS, 297, L11+
Fabian, A. C., & Iwasawa, K. 1999, MNRAS, 303, L34
Feldmeier, J. J., Brandt, W. N., Elvis, M., Fabian, A. C., Iwasawa, K., & Mathur, S.
1999, ApJ, 510, 167
Fields, D. L., Mathur, S., Krongold, Y., Williams, R., & Nicastro, F. 2007, ApJ, 666,
333
828
Forster, K., Leighly, K. M., & Kay, L. E. 1999, ApJ, 523, 521
Freedman, W. L., Hughes, S. M., Madore, B. F., Mould, J. R., Lee, M. G., Stetson,
P., Kennicutt, R. C., Turner, A., Ferrarese, L., Ford, H., Graham, J. A., Hill, R.,
Hoessel, J. G., Huchra, J., & Illingworth, G. D. 1994, ApJ, 427, 628
Freeman, K. C., Karlsson, B., Lynga, G., Burrell, J. F., van Woerden, H., Goss,
W. M., & Mebold, U. 1977, A&A, 55, 445
Gallo, L. C., Brandt, W. N., Costantini, E., & Fabian, A. C. 2007, MNRAS, 377,
1375
Gallo, L. C., Fabian, A. C., Boller, T., & Pietsch, W. 2005, MNRAS, 363, 64
Gallo, L. C., Lehmann, I., Pietsch, W., Boller, T., Brinkmann, W., Friedrich, P., &
Grupe, D. 2006, MNRAS, 365, 688
Garc??a, J., Mendoza, C., Bautista, M. A., Gorczyca, T. W., Kallman, T. R., &
Palmeri, P. 2005, ApJS, 158, 68
Garnett, D. R. 2002, ApJ, 581, 1019
George, I. M., Turner, T. J., Netzer, H., Nandra, K., Mushotzky, R. F., & Yaqoob,
T. 1998, ApJS, 114, 73
George, I. M., Turner, T. J., Yaqoob, T., Netzer, H., Laor, A., Mushotzky, R. F.,
Nandra, K., & Takahashi, T. 2000, ApJ, 531, 52
Gierli?nski, M., & Done, C. 2004, MNRAS, 349, L7
Gilli, R., Comastri, A., & Hasinger, G. 2007, A&A, 463, 79
Goad, M. R., Roberts, T. P., Reeves, J. N., & Uttley, P. 2006, MNRAS, 365, 191
Gonc?alves, A. C., & Soria, R. 2006, MNRAS, 371, 673
Grandi, P., Haardt, F., Ghisellini, G., Grove, E. J., Maraschi, L., & Urry, C. M. 1998,
ApJ, 498, 220
Grandi, P., Sambruna, R. M., Maraschi, L., Matt, G., Urry, C. M., & Mushotzky,
334
R. F. 1997, ApJ, 487, 636
Grimm, H.-J., Gilfanov, M., & Sunyaev, R. 2002, A&A, 391, 923
?. 2003, MNRAS, 339, 793
Grogin, N. A., Conselice, C. J., Chatzichristou, E., Alexander, D. M., Bauer, F. E.,
Hornschemeier, A. E., Jogee, S., Koekemoer, A. M., Laidler, V. G., Livio, M.,
Lucas, R. A., Paolillo, M., Ravindranath, S., Schreier, E. J., Simmons, B. D., &
Urry, C. M. 2005, ApJ, 627, L97
G?ultekin, K., Miller, M. C., & Hamilton, D. P. 2004, ApJ, 616, 221
Guti?errez, C. M., & L?opez-Corredoira, M. 2005, ApJ, 622, L89
Hanawa, T. 1989, ApJ, 341, 948
Heger, A., & Woosley, S. E. 2002, ApJ, 567, 532
Hewett, P. C., & Foltz, C. B. 1994, PASP, 106, 113
Ho, L. C., Filippenko, A. V., & Sargent, W. L. W. 1997, ApJS, 112, 315
Holt, S. S., Mushotzky, R. F., Boldt, E. A., Serlemitsos, P. J., Becker, R. H.,
Szymkowiak, A. E., & White, N. E. 1980, ApJ, 241, L13
Humphrey, P. J., Fabbiano, G., Elvis, M., Church, M. J., & Ba?uci?nska-Church, M.
2003, MNRAS, 344, 134
Immler, S., Brandt, W. N., Vignali, C., Bauer, F. E., Crenshaw, D. M., Feldmeier,
J. J., & Kraemer, S. B. 2003, AJ, 126, 153
Iwasawa, K., & Taniguchi, Y. 1993, ApJ, 413, L15
Jarrett, T. H., Chester, T., Cutri, R., Schneider, S. E., & Huchra, J. P. 2003, AJ, 125,
525
Jenkins, L. P., Roberts, T. P., Warwick, R. S., Kilgard, R. E., & Ward, M. J. 2004,
MNRAS, 349, 404
Jiang, P., Wang, J. X., & Wang, T. G. 2006, ApJ, 644, 725
Jones, C. 1977, ApJ, 214, 856
335
Jonker, P. G., & Nelemans, G. 2004, MNRAS, 354, 355
Juett, A. M., Schulz, N. S., & Chakrabarty, D. 2004, ApJ, 612, 308
Kalogera, V., Henninger, M., Ivanova, N., & King, A. R. 2004, ApJ, 603, L41
Karachentsev, I. D., Sharina, M. E., Dolphin, A. E., Grebel, E. K., Geisler, D.,
Guhathakurta, P., Hodge, P. W., Karachentseva, V. E., Sarajedini, A., & Seitzer,
P. 2002, A&A, 385, 21
Kauffmann, G., Heckman, T. M., Tremonti, C., Brinchmann, J., Charlot, S., White,
S. D. M., Ridgway, S. E., Brinkmann, J., Fukugita, M., Hall, P. B., Ivezi?c, ?Z.,
Richards, G. T., & Schneider, D. P. 2003, MNRAS, 346, 1055
Kelson, D.D., Illingworth, G.D., Freedman, W.F., Graham,J.A., Hill, R., Madore,
B. F., Saha, A., Stetson, P. B., Kennicutt, Jr., R. C., Mould, J. R., Hughes, S. M.,
Ferrarese, L., Phelps, R., Turner, A., Cook, K. H., Ford, H., Hoessel, J. G., &
Huchra, J. 1996, ApJ, 463, 26
King, A. R., Davies, M. B., Ward, M. J., Fabbiano, G., & Elvis, M. 2001, ApJ, 552,
L109
Kirhakos, S. D., & Steiner, J. E. 1990, AJ, 99, 1435
Komossa, S., Burwitz, V., Hasinger, G., Predehl, P., Kaastra, J. S., & Ikebe, Y. 2003,
ApJ, 582, L15
Kong, A. K. H. 2003, MNRAS, 346, 265
Kong, A. K. H., & Di Stefano, R. 2003, ApJ, 590, L13
Kong, X., Zhou, X., Chen, J., Cheng, F., Jiang, Z., Zhu, J., Zheng, Z., Mao, S.,
Shang, Z., Fan, X., Byun, Y.-I., Chen, R., Chen, W.-p., Deng, L., Hester, J. J., Li,
Y., Lin, W., Su, H., Sun, W.-h., Tsay, W.-s., Windhorst, R. A., Wu, H., Xia, X., Xu,
W., Xue, S., Yan, H., Zheng, Z., & Zou, Z. 2000, AJ, 119, 2745
K?ording, E., Falcke, H., & Markoff, S. 2002, A&A, 382, L13
Kraft, R. P., Hardcastle, M. J., Worrall, D. M., & Murray, S. S. 2005, ApJ, 622, 149
336
Krolik, J. H., & Begelman, M. C. 1988, ApJ, 329, 702
Krongold, Y., Nicastro, F., Elvis, M., Brickhouse, N. S., Mathur, S., & Zezas, A.
2005, ApJ, 620, 165
Kubota, A., & Makishima, K. 2004, ApJ, 601, 428
Kubota, A., Makishima, K., & Ebisawa, K. 2001, ApJ, 560, L147
Kuntz, K. D., Gruendl, R. A., Chu, Y.-H., Chen, C.-H. R., Still, M., Mukai, K., &
Mushotzky, R. F. 2005, ApJ, 620, L31
Kuntz, K. D., Harrus, I., McGlynn, T. A., Mushotzky, R. F., & Snowden, S. L. 2008,
ArXiv e-prints, 805
Laor, A., Fiore, F., Elvis, M., Wilkes, B. J., & McDowell, J. C. 1997, ApJ, 477, 93
Lawrence, A. 1991, MNRAS, 252, 586
Lehmann, I., Becker, T., Fabrika, S., Roth, M., Miyaji, T., Afanasiev, V.,
Sholukhova, O., S?anchez, S. F., Greiner, J., Hasinger, G., Costantini, E., Surkov,
A., & Burenkov, A. 2005, A&A, 431, 847
Lewis, K. T., Eracleous, M., & Sambruna, R. M. 2003, ApJ, 593, 115
Lipari, S., Tsvetanov, Z., & Macchetto, F. 1993, ApJ, 405, 186
Liu, J.-F., & Bregman, J. N. 2005, ApJS, 157, 59
Liu, J.-F., Bregman, J. N., & Seitzer, P. 2002, ApJ, 580, L31
?. 2004, ApJ, 602, 249
Long, K. S., Charles, P. A., & Dubus, G. 2002, ApJ, 569, 204
Longinotti, A. L., Bianchi, S., Santos-Lleo, M., Rodr??guez-Pascual, P., Guainazzi,
M., Cardaci, M., & Pollock, A. M. T. 2007, A&A, 470, 73
Madau, P., & Rees, M. J. 2001, ApJ, 551, L27
Magdziarz, P., & Zdziarski, A. A. 1995, MNRAS, 273, 837
Maiolino, R., Marconi, A., Salvati, M., Risaliti, G., Severgnini, P., Oliva, E., La
Franca, F., & Vanzi, L. 2001, A&A, 365, 28
337
Makishima, K., Kubota, A., Mizuno, T., Ohnishi, T., Tashiro, M., Aruga, Y.,
Asai, K., Dotani, T., Mitsuda, K., Ueda, Y., Uno, S., Yamaoka, K., Ebisawa, K.,
Kohmura, Y., & Okada, K. 2000, ApJ, 535, 632
Malizia, A., Bassani, L., Stephen, J. B., Malaguti, G., & Palumbo, G. G. C. 1997,
ApJS, 113, 311
Malizia, A., Landi, R., Bassani, L., Bird, A. J., Molina, M., De Rosa, A., Fiocchi,
M., Gehrels, N., Kennea, J., & Perri, M. 2007, ApJ, 668, 81
Malkan, M. A., Gorjian, V., & Tam, R. 1998, ApJS, 117, 25
Maloney, P. R., & Reynolds, C. S. 2000, ApJ, 545, L23
Markwardt, C. B., Tueller, J., Skinner, G. K., Gehrels, N., Barthelmy, S. D., &
Mushotzky, R. F. 2005, ApJ, 633, L77
Marshall, F. E., Boldt, E. A., Holt, S. S., Miller, R. B., Mushotzky, R. F., Rose, L. A.,
Rothschild, R. E., & Serlemitsos, P. J. 1980, ApJ, 235, 4
Martin, C. L., Kobulnicky, H. A., & Heckman, T. M. 2002, ApJ, 574, 663
Masetti, N., Morelli, L., Palazzi, E., Galaz, G., Bassani, L., Bazzano, A., Bird,
A. J., Dean, A. J., Israel, G. L., Landi, R., Malizia, A., Minniti, D., Schiavone,
F., Stephen, J. B., Ubertini, P., & Walter, R. 2006, A&A, 459, 21
Mateos, S., Barcons, X., Carrera, F. J., Ceballos, M. T., Hasinger, G., Lehmann, I.,
Fabian, A. C., & Streblyanska, A. 2005, A&A, 444, 79
Mateos, S., Barcons, X., Carrera, F. J., Page, M. J., Ceballos, M. T., Hasinger, G., &
Fabian, A. C. 2007, A&A, 473, 105
Matsumoto, C., Nava, A., Maddox, L. A., Leighly, K. M., Grupe, D., Awaki, H., &
Ueno, S. 2004, ApJ, 617, 930
Matt, G. 2002, Royal Society of London Philosophical Transactions Series A, 360,
2045
Matt, G., Brandt, W. N., & Fabian, A. C. 1996, MNRAS, 280, 823
338
Matt, G., Fabian, A. C., Guainazzi, M., Iwasawa, K., Bassani, L., & Malaguti, G.
2000, MNRAS, 318, 173
Matt, G., Guainazzi, M., & Maiolino, R. 2003, MNRAS, 342, 422
Mattson, B. J., Weaver, K. A., & Reynolds, C. S. 2007, ApJ, 664, 101
McCammon, D., Almy, R., Apodaca, E., Bergmann Tiest, W., Cui, W., Deiker, S.,
Galeazzi,M.,Juda,M.,Lesser,A.,Mihara,T.,Morgenthaler,J.P.,Sanders,W.T.,
Zhang, J., Figueroa-Feliciano, E., Kelley, R. L., Moseley, S. H., Mushotzky, R. F.,
Porter, F. S., Stahle, C. K., & Szymkowiak, A. E. 2002, ApJ, 576, 188
McClintock, J. E., Narayan, R., & Rybicki, G. B. 2004, ApJ, 615, 402
Mel?endez, M., Kraemer, S. B., Armentrout, B. K., Deo, R. P., Crenshaw, D. M.,
Schmitt, H. R., Mushotzky, R. F., Tueller, J., Markwardt, C. B., & Winter, L.
2008, ArXiv e-prints, 804
Merloni, A., Fabian, A. C., & Ross, R. R. 2000, MNRAS, 313, 193
Miller, B. W. 1995, ApJ, 446, L75+
Miller, J. M., Fabbiano, G., Miller, M. C., & Fabian, A. C. 2003, ApJ, 585, L37
Miller, J. M., Fabian, A. C., & Miller, M. C. 2004a, ApJ, 614, L117
?. 2004b, ApJ, 607, 931
Miller, M. C., & Hamilton, D. P. 2002, MNRAS, 330, 232
Miniutti, G., Ponti, G., Dadina, M., Cappi, M., & Malaguti, G. 2007, MNRAS, 375,
227
Mitsuda, K., Inoue, H., Koyama, K., Makishima, K., Matsuoka, M., Ogawara, Y.,
Suzuki, K., Tanaka, Y., Shibazaki, N., & Hirano, T. 1984, PASJ, 36, 741
Miyamoto, S., Kimura, K., Kitamoto, S., Dotani, T., & Ebisawa, K. 1991, ApJ, 383,
784
Moran, E. C., Filippenko, A. V., & Chornock, R. 2002, ApJ, 579, L71
Morrison, R., & McCammon, D. 1983, ApJ, 270, 119
339
Mukai, K., Hellier, C., Madejski, G., Patterson, J., & Skillman, D. R. 2003, ApJ,
597, 479
Mushotzky, R. 2004, in Astrophysics and Space Science Library, Vol. 308, Super-
massive Black Holes in the Distant Universe, ed. A. J. Barger, 53?+
Mushotzky, R. F. 1982, ApJ, 256, 92
Mushotzky, R. F., Done, C., & Pounds, K. A. 1993, ARA&A, 31, 717
Nandra, K., Georgakakis, A., Willmer, C. N. A., Cooper, M. C., Croton, D. J.,
Davis, M., Faber, S. M., Koo, D. C., Laird, E. S., & Newman, J. A. 2007, ApJ, 660,
L11
Nandra, K., George, I. M., Mushotzky, R. F., Turner, T. J., & Yaqoob, T. 1997, ApJ,
476, 70
Nandra, K., & Pounds, K. A. 1994, MNRAS, 268, 405
Nicastro, F. 2000, ApJ, 530, L65
Nicastro, F., Martocchia, A., & Matt, G. 2003, ApJ, 589, L13
Novak, G. S., Faber, S. M., & Dekel, A. 2006, ApJ, 637, 96
Oliva, E., Origlia, L., Kotilainen, J. K., & Moorwood, A. F. M. 1995, A&A, 301, 55
Page, K. L., O?Brien, P. T., Reeves, J. N., & Turner, M. J. L. 2004, MNRAS, 347, 316
Pakull, M. W., & Mirioni, L. 2003, in Revista Mexicana de Astronomia y As-
trofisica, vol. 27, Vol. 15, Revista Mexicana de Astronomia y Astrofisica Con-
ference Series, ed. J. Arthur & W. J. Henney, 197?199
Paolillo, M., Schreier, E. J., Giacconi, R., Koekemoer, A. M., & Grogin, N. A. 2004,
ApJ, 611, 93
Paturel, G., Fouque, P., Bottinelli, L., & Gouguenheim, L. 1989, A&AS, 80, 299
Peterson, B. M., Ferrarese, L., Gilbert, K. M., Kaspi, S., Malkan, M. A., Maoz, D.,
Merritt, D., Netzer, H., Onken, C. A., Pogge, R. W., Vestergaard, M., & Wandel,
A. 2004, ApJ, 613, 682
340
Peterson, K. C., Gallagher, S. C., Hornschemeier, A. E., Muno, M. P., & Bullard,
E. C. 2006, AJ, 131, 133
Petrucci, P. O., Ponti, G., Matt, G., Longinotti, A. L., Malzac, J., Mouchet, M.,
Boisson, C., Maraschi, L., Nandra, K., & Ferrando, P. 2007, A&A, 470, 889
Piccinotti, G., Mushotzky, R. F., Boldt, E. A., Holt, S. S., Marshall, F. E., Serlemit-
sos, P. J., & Shafer, R. A. 1982, ApJ, 253, 485
Piconcelli, E., Cappi, M., Bassani, L., Fiore, F., Di Cocco, G., & Stephen, J. B. 2002,
A&A, 394, 835
Piconcelli, E., Jimenez-Bail?on, E., Guainazzi, M., Schartel, N., Rodr??guez-Pascual,
P. M., & Santos-Lle?o, M. 2005, A&A, 432, 15
Pierce, C. M., Lotz, J. M., Laird, E. S., Lin, L., Nandra, K., Primack, J. R., Faber,
S. M., Barmby, P., Park, S. Q., Willner, S. P., Gwyn, S., Koo, D. C., Coil, A. L.,
Cooper, M. C., Georgakakis, A., Koekemoer, A. M., Noeske, K. G., Weiner, B. J.,
& Willmer, C. N. A. 2007, ApJ, 660, L19
Pietsch, W., Haberl, F., & Vogler, A. 2003, A&A, 402, 457
Pietsch, W., Roberts, T. P., Sako, M., Freyberg, M. J., Read, A. M., Borozdin, K. N.,
Branduardi-Raymont, G., Cappi, M., Ehle, M., Ferrando, P., Kahn, S. M., Pon-
man, T. J., Ptak, A., Shirey, R. E., & Ward, M. 2001, A&A, 365, L174
Pilyugin, L. S., V??lchez, J. M., & Contini, T. 2004, A&A, 425, 849
Porquet, D., Kaastra, J. S., Page, K. L., O?Brien, P. T., Ward, M. J., & Dubau, J.
2004a, A&A, 413, 913
Porquet, D., Reeves, J. N., O?Brien, P., & Brinkmann, W. 2004b, A&A, 422, 85
Predehl, P., & Schmitt, J. H. M. M. 1995, A&A, 293, 889
Protassov, R., van Dyk, D. A., Connors, A., Kashyap, V. L., & Siemiginowska, A.
2002, ApJ, 571, 545
Ptak, A., & Colbert, E. 2004, ApJ, 606, 291
341
Rafanelli, P., Violato, M., & Baruffolo, A. 1995, AJ, 109, 1546
Ranalli, P., Comastri, A., & Setti, G. 2003, A&A, 399, 39
Reeves, J. N., & Turner, M. J. L. 2000, MNRAS, 316, 234
Reichert, G. A., Mushotzky, R. F., & Holt, S. S. 1986, ApJ, 303, 87
Reilman, R. F., & Manson, S. T. 1979, ApJS, 40, 815
Remillard, R. A., & McClintock, J. E. 2006, ARA&A, 44, 49
Revnivtsev, M., Molkov, S., & Sazonov, S. 2008, A&A, 483, 425
Reynolds, C. S. 1997, MNRAS, 286, 513
Reynolds, C. S., Fabian, A. C., & Inoue, H. 1995, MNRAS, 276, 1311
Reynolds, C. S., Iwasawa, K., Crawford, C. S., & Fabian, A. C. 1998, MNRAS, 299,
410
Reynolds, C. S., Loan, A. J., Fabian, A. C., Makishima, K., Brandt, W. N., &
Mizuno, T. 1997, MNRAS, 286, 349
Rigby, J. R., Rieke, G. H., Donley, J. L., Alonso-Herrero, A., & P?erez-Gonz?alez,
P. G. 2006, ApJ, 645, 115
Risaliti, G. 2007, in Astronomical Society of the Pacific Conference Series, Vol.
373, Astronomical Society of the Pacific Conference Series, ed. L. C. Ho & J.-W.
Wang, 458?+
Risaliti, G., Elvis, M., Fabbiano, G., Baldi, A., & Zezas, A. 2005, ApJ, 623, L93
Risaliti, G., Elvis, M., & Nicastro, F. 2002, ApJ, 571, 234
Roberts, T. P., Warwick, R. S., Ward, M. J., & Goad, M. R. 2004, MNRAS, 349, 1193
Roberts, T. P., Warwick, R. S., Ward, M. J., Goad, M. R., & Jenkins, L. P. 2005,
MNRAS, 357, 1363
Ross, R. R., & Fabian, A. C. 2005, MNRAS, 358, 211
Saez, C., Chartas, G., Brandt, W. N., Lehmer, B. D., Bauer, F. E., Dai, X., &
Garmire, G. P. 2008, AJ, 135, 1505
342
Sakai, S., & Madore, B. F. 1999, ApJ, 526, 599
Sakai, S., Madore, B. F., & Freedman, W. L. 1996, ApJ, 461, 713
Sambruna, R. M., Eracleous, M., & Mushotzky, R. F. 1999, ApJ, 526, 60
Sambruna, R. M., Markwardt, C. B., Mushotzky, R. F., Tueller, J., Hartman, R.,
Brandt, W. N., Schneider, D. P., Falcone, A., Cucchiara, A., Aller, M. F., Aller,
H. D., Torniainen, I., Tavecchio, F., Maraschi, L., Gliozzi, M., & Takahashi, T.
2006, ApJ, 646, 23
Sambruna, R. M., & Mushotzky, R. F. 1998, ApJ, 502, 630
Sazonov, S., Krivonos, R., Revnivtsev, M., Churazov, E., & Sunyaev, R. 2008,
A&A, 482, 517
Sazonov, S., Revnivtsev, M., Krivonos, R., Churazov, E., & Sunyaev, R. 2007,
A&A, 462, 57
Schlegel, E. M., Kong, A., Kaaret, P., DiStefano, R., & Murray, S. 2004, ApJ, 603,
644
Searle, L. 1971, ApJ, 168, 327
Shapley, A., Fabbiano, G., & Eskridge, P. B. 2001, ApJS, 137, 139
Shemmer, O., Brandt, W. N., Netzer, H., Maiolino, R., & Kaspi, S. 2006, ApJ, 646,
L29
Shimura, T., & Takahara, F. 1995, ApJ, 445, 780
Siebert, J., Matsuoka, M., Brinkmann, W., Cappi, M., Mihara, T., & Takahashi, T.
1996, A&A, 307, 8
Smith, D. A., & Wilson, A. S. 2001, ApJ, 557, 180
Sobczak, G. J., McClintock, J. E., Remillard, R. A., Bailyn, C. D., & Orosz, J. A.
1999, ApJ, 520, 776
Steenbrugge, K. C., Kaastra, J. S., Sako, M., Branduardi-Raymont, G., Behar, E.,
Paerels, F. B. S., Blustin, A. J., & Kahn, S. M. 2005, A&A, 432, 453
343
Steffen, A. T., Barger, A. J., Cowie, L. L., Mushotzky, R. F., & Yang, Y. 2003, ApJ,
596, L23
Stobbart, A.-M., Roberts, T. P., & Wilms, J. 2006, MNRAS, 368, 397
Stocke, J. T., Liebert, J., Gioia, I. M., Maccacaro, T., Griffiths, R. E., Danziger, I. J.,
Kunth, D., & Lub, J. 1983, ApJ, 273, 458
Strohmayer, T. E., Mushotzky, R. F., Winter, L., Soria, R., Uttley, P., & Cropper,
M. 2007, ApJ, 660, 580
Summers, L. K., Stevens, I. R., Strickland, D. K., & Heckman, T. M. 2003, MNRAS,
342, 690
?. 2004, MNRAS, 351, 1
Sunyaev, R. A., & Titarchuk, L. G. 1980, A&A, 86, 121
Swartz, D. A., Ghosh, K. K., Tennant, A. F., & Wu, K. 2004, ApJS, 154, 519
Swaters, R. A., van Albada, T. S., van der Hulst, J. M., & Sancisi, R. 2002, A&A,
390, 829
Takano, M., Mitsuda, K., Fukazawa, Y., & Nagase, F. 1994, ApJ, 436, L47
Tolstoy, E., Saha, A., Hoessel, J. G., & McQuade, K. 1995, AJ, 110, 1640
Tosi, M., Sabbi, E., Bellazzini, M., Aloisi, A., Greggio, L., Leitherer, C., & Monte-
griffo, P. 2001, AJ, 122, 1271
Tremonti, C. A., Heckman, T. M., Kauffmann, G., Brinchmann, J., Charlot, S.,
White, S. D. M., Seibert, M., Peng, E. W., Schlegel, D. J., Uomoto, A., Fukugita,
M., & Brinkmann, J. 2004, ApJ, 613, 898
Tueller, J., Mushotzky, R. F., Barthelmy, S., Cannizzo, J. K., Gehrels, N., Mark-
wardt, C. B., Skinner, G. K., & Winter, L. M. 2008, ApJ, 681, 113
Tully, R. B. 1988, Nearby galaxies catalog (Cambridge and New York, Cambridge
University Press, 1988, 221 p.)
Turner,T.J.,George, I.M., Grupe, D.,Nandra,K.,Remillard,R.A.,Leighly,K.M.,
344
Marshall, H. L., Kraemer, S. B., & Crenshaw, D. M. 1999, ApJ, 510, 178
Turner, T. J., George, I. M., Nandra, K., & Mushotzky, R. F. 1997, ApJS, 113, 23
Turner, T. J., Netzer, H., & George, I. M. 1996, ApJ, 463, 134
Ueda, Y., Akiyama, M., Ohta, K., & Miyaji, T. 2003, ApJ, 598, 886
Ueda, Y., Eguchi, S., Terashima, Y., Mushotzky, R., Tueller, J., Markwardt, C.,
Gehrels, N., Hashimoto, Y., & Potter, S. 2007, ApJ, 664, L79
Veilleux, S. 2002, in Astronomical Society of the Pacific Conference Series, Vol.
284, IAU Colloq. 184: AGN Surveys, ed. R. F. Green, E. Y. Khachikian, & D. B.
Sanders, 111?+
Veilleux, S., Kim, D.-C., Sanders, D. B., Mazzarella, J. M., & Soifer, B. T. 1995,
ApJS, 98, 171
Vogler, A., & Pietsch, W. 1996, A&A, 311, 35
Vogler, A., Pietsch, W., & Bertoldi, F. 1997, A&A, 318, 768
Walsh, J. R., & Roy, J.-R. 1997, MNRAS, 288, 726
Wang, Q. D., Yao, Y., Fukui, W., Zhang, S. N., & Williams, R. 2004, ApJ, 609, 113
White, R. L., Becker, R. H., Gregg, M. D., Laurent-Muehleisen, S. A., Brotherton,
M. S., Impey, C. D., Petry, C. E., Foltz, C. B., Chaffee, F. H., Richards, G. T.,
Oegerle, W. R., Helfand, D. J., McMahon, R. G., & Cabanela, J. E. 2000, ApJS,
126, 133
Wilms, J., Allen, A., & McCray, R. 2000, ApJ, 542, 914
Winter, L. M., Mushotzky, R., & Reynolds, C. S. 2004, in Bulletin of the American
Astronomical Society, Vol. 36, Bulletin of the American Astronomical Society,
1472?+
Winter, L. M., Mushotzky, R., Tueller, J., Lewis, K., & Reynolds, C. 2007a, in
American Astronomical Society Meeting Abstracts, Vol. 211, American Astro-
nomical Society Meeting Abstracts, 64.02?+
345
Winter, L. M., Mushotzky, R. F., & Reynolds, C. S. 2006a, in ESA Special Publica-
tion, Vol. 604, The X-ray Universe 2005, ed. A. Wilson, 377?378
Winter, L. M., Mushotzky, R. F., & Reynolds, C. S. 2006b, ApJ, 649, 730
?. 2007b, ApJ, 655, 163
Winter, L. M., Mushotzky, R. F., Tueller, J., & Markwardt, C. 2008, ApJ, 674, 686
Winter, L. M., Mushotzky, R. F., Tueller, J., Reynolds, C. S., & Markwardt, C.
2007c, in American Astronomical Society Meeting Abstracts, Vol. 210, Ameri-
can Astronomical Society Meeting Abstracts, 02.25?+
Winter, L. M., R. F. Mushotzky Collaboration, & C. S. Reynolds Collaboration.
2005, in Bulletin of the American Astronomical Society, Vol. 37, Bulletin of the
American Astronomical Society, 1318?+
Zdziarski, A. A., & Gierli?nski, M. 2004, Progress of Theoretical Physics Supple-
ment, 155, 99
Zhang, J. S., Henkel, C., Kadler, M., Greenhill, L. J., Nagar, N., Wilson, A. S., &
Braatz, J. A. 2006, A&A, 450, 933
Zimmermann, H.-U., & Aschenbach, B. 2003, A&A, 406, 969
346